2047 files changed, 228359 insertions, 71564 deletions

diff --git a/drivers/Makefile b/drivers/Makefile
index f42a03029b7c..91874e048552 100644
--- a/drivers/Makefile
+++ b/drivers/Makefile
@@ -10,6 +10,7 @@ obj-$(CONFIG_PCI)		+= pci/
 obj-$(CONFIG_PARISC)		+= parisc/
 obj-$(CONFIG_RAPIDIO)		+= rapidio/
 obj-y				+= video/
+obj-y				+= idle/
 obj-$(CONFIG_ACPI)		+= acpi/
 obj-$(CONFIG_SFI)		+= sfi/
 # PnP must come after ACPI since it will eventually need to check if acpi
@@ -91,7 +92,6 @@ obj-$(CONFIG_EISA)		+= eisa/
 obj-y				+= lguest/
 obj-$(CONFIG_CPU_FREQ)		+= cpufreq/
 obj-$(CONFIG_CPU_IDLE)		+= cpuidle/
-obj-y				+= idle/
 obj-$(CONFIG_MMC)		+= mmc/
 obj-$(CONFIG_MEMSTICK)		+= memstick/
 obj-$(CONFIG_NEW_LEDS)		+= leds/
diff --git a/drivers/acpi/Kconfig b/drivers/acpi/Kconfig
index 93d2c7971df6..746411518802 100644
--- a/drivers/acpi/Kconfig
+++ b/drivers/acpi/Kconfig
@@ -360,4 +360,13 @@ config ACPI_SBS
 	  To compile this driver as a module, choose M here:
 	  the modules will be called sbs and sbshc.
 
+config ACPI_HED
+	tristate "Hardware Error Device"
+	help
+	  This driver supports the Hardware Error Device (PNP0C33),
+	  which is used to report some hardware errors notified via
+	  SCI, mainly the corrected errors.
+
+source "drivers/acpi/apei/Kconfig"
+
 endif	# ACPI
diff --git a/drivers/acpi/Makefile b/drivers/acpi/Makefile
index a8d8998dd5c5..6ee33169e1dc 100644
--- a/drivers/acpi/Makefile
+++ b/drivers/acpi/Makefile
@@ -19,7 +19,7 @@ obj-y				+= acpi.o \
 
 # All the builtin files are in the "acpi." module_param namespace.
 acpi-y				+= osl.o utils.o reboot.o
-acpi-y				+= hest.o
+acpi-y				+= atomicio.o
 
 # sleep related files
 acpi-y				+= wakeup.o
@@ -59,6 +59,7 @@ obj-$(CONFIG_ACPI_BATTERY)	+= battery.o
 obj-$(CONFIG_ACPI_SBS)		+= sbshc.o
 obj-$(CONFIG_ACPI_SBS)		+= sbs.o
 obj-$(CONFIG_ACPI_POWER_METER)	+= power_meter.o
+obj-$(CONFIG_ACPI_HED)		+= hed.o
 
 # processor has its own "processor." module_param namespace
 processor-y			:= processor_driver.o processor_throttling.o
@@ -66,3 +67,5 @@ processor-y			+= processor_idle.o processor_thermal.o
 processor-$(CONFIG_CPU_FREQ)	+= processor_perflib.o
 
 obj-$(CONFIG_ACPI_PROCESSOR_AGGREGATOR) += acpi_pad.o
+
+obj-$(CONFIG_ACPI_APEI)		+= apei/
diff --git a/drivers/acpi/acpi_pad.c b/drivers/acpi/acpi_pad.c
index 62122134693b..d269a8f3329c 100644
--- a/drivers/acpi/acpi_pad.c
+++ b/drivers/acpi/acpi_pad.c
@@ -43,6 +43,10 @@ static DEFINE_MUTEX(isolated_cpus_lock);
 #define CPUID5_ECX_EXTENSIONS_SUPPORTED (0x1)
 #define CPUID5_ECX_INTERRUPT_BREAK	(0x2)
 static unsigned long power_saving_mwait_eax;
+
+static unsigned char tsc_detected_unstable;
+static unsigned char tsc_marked_unstable;
+
 static void power_saving_mwait_init(void)
 {
 	unsigned int eax, ebx, ecx, edx;
@@ -87,8 +91,8 @@ static void power_saving_mwait_init(void)
 
 		/*FALL THROUGH*/
 	default:
-		/* TSC could halt in idle, so notify users */
-		mark_tsc_unstable("TSC halts in idle");
+		/* TSC could halt in idle */
+		tsc_detected_unstable = 1;
 	}
 #endif
 }
@@ -168,16 +172,14 @@ static int power_saving_thread(void *data)
 
 		do_sleep = 0;
 
-		current_thread_info()->status &= ~TS_POLLING;
-		/*
-		 * TS_POLLING-cleared state must be visible before we test
-		 * NEED_RESCHED:
-		 */
-		smp_mb();
-
 		expire_time = jiffies + HZ * (100 - idle_pct) / 100;
 
 		while (!need_resched()) {
+			if (tsc_detected_unstable && !tsc_marked_unstable) {
+				/* TSC could halt in idle, so notify users */
+				mark_tsc_unstable("TSC halts in idle");
+				tsc_marked_unstable = 1;
+			}
 			local_irq_disable();
 			cpu = smp_processor_id();
 			clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_ENTER,
@@ -200,8 +202,6 @@ static int power_saving_thread(void *data)
 			}
 		}
 
-		current_thread_info()->status |= TS_POLLING;
-
 		/*
 		 * current sched_rt has threshold for rt task running time.
 		 * When a rt task uses 95% CPU time, the rt thread will be
diff --git a/drivers/acpi/acpica/evxfevnt.c b/drivers/acpi/acpica/evxfevnt.c
index 7c7bbb4d402c..d5a5efc043bf 100644
--- a/drivers/acpi/acpica/evxfevnt.c
+++ b/drivers/acpi/acpica/evxfevnt.c
@@ -69,7 +69,7 @@ acpi_ev_get_gpe_device(struct acpi_gpe_xrupt_info *gpe_xrupt_info,
 
 acpi_status acpi_enable(void)
 {
-	acpi_status status = AE_OK;
+	acpi_status status;
 
 	ACPI_FUNCTION_TRACE(acpi_enable);
 
@@ -84,21 +84,30 @@ acpi_status acpi_enable(void)
 	if (acpi_hw_get_mode() == ACPI_SYS_MODE_ACPI) {
 		ACPI_DEBUG_PRINT((ACPI_DB_INIT,
 				  "System is already in ACPI mode\n"));
-	} else {
-		/* Transition to ACPI mode */
+		return_ACPI_STATUS(AE_OK);
+	}
 
-		status = acpi_hw_set_mode(ACPI_SYS_MODE_ACPI);
-		if (ACPI_FAILURE(status)) {
-			ACPI_ERROR((AE_INFO,
-				    "Could not transition to ACPI mode"));
-			return_ACPI_STATUS(status);
-		}
+	/* Transition to ACPI mode */
 
-		ACPI_DEBUG_PRINT((ACPI_DB_INIT,
-				  "Transition to ACPI mode successful\n"));
+	status = acpi_hw_set_mode(ACPI_SYS_MODE_ACPI);
+	if (ACPI_FAILURE(status)) {
+		ACPI_ERROR((AE_INFO,
+			    "Could not transition to ACPI mode"));
+		return_ACPI_STATUS(status);
 	}
 
-	return_ACPI_STATUS(status);
+	/* Sanity check that transition succeeded */
+
+	if (acpi_hw_get_mode() != ACPI_SYS_MODE_ACPI) {
+		ACPI_ERROR((AE_INFO,
+			    "Hardware did not enter ACPI mode"));
+		return_ACPI_STATUS(AE_NO_HARDWARE_RESPONSE);
+	}
+
+	ACPI_DEBUG_PRINT((ACPI_DB_INIT,
+			  "Transition to ACPI mode successful\n"));
+
+	return_ACPI_STATUS(AE_OK);
 }
 
 ACPI_EXPORT_SYMBOL(acpi_enable)
diff --git a/drivers/acpi/acpica/hwacpi.c b/drivers/acpi/acpica/hwacpi.c
index 679a112a7d26..b44274a0b62c 100644
--- a/drivers/acpi/acpica/hwacpi.c
+++ b/drivers/acpi/acpica/hwacpi.c
@@ -63,7 +63,6 @@ acpi_status acpi_hw_set_mode(u32 mode)
 {
 
 	acpi_status status;
-	u32 retry;
 
 	ACPI_FUNCTION_TRACE(hw_set_mode);
 
@@ -125,24 +124,7 @@ acpi_status acpi_hw_set_mode(u32 mode)
 		return_ACPI_STATUS(status);
 	}
 
-	/*
-	 * Some hardware takes a LONG time to switch modes. Give them 3 sec to
-	 * do so, but allow faster systems to proceed more quickly.
-	 */
-	retry = 3000;
-	while (retry) {
-		if (acpi_hw_get_mode() == mode) {
-			ACPI_DEBUG_PRINT((ACPI_DB_INFO,
-					  "Mode %X successfully enabled\n",
-					  mode));
-			return_ACPI_STATUS(AE_OK);
-		}
-		acpi_os_stall(1000);
-		retry--;
-	}
-
-	ACPI_ERROR((AE_INFO, "Hardware did not change modes"));
-	return_ACPI_STATUS(AE_NO_HARDWARE_RESPONSE);
+	return_ACPI_STATUS(AE_OK);
 }
 
 /*******************************************************************************
diff --git a/drivers/acpi/apei/Kconfig b/drivers/acpi/apei/Kconfig
new file mode 100644
index 000000000000..f8c668f27b5a
--- /dev/null
+++ b/drivers/acpi/apei/Kconfig
@@ -0,0 +1,30 @@
+config ACPI_APEI
+	bool "ACPI Platform Error Interface (APEI)"
+	depends on X86
+	help
+	  APEI allows to report errors (for example from the chipset)
+	  to the operating system. This improves NMI handling
+	  especially. In addition it supports error serialization and
+	  error injection.
+
+config ACPI_APEI_GHES
+	tristate "APEI Generic Hardware Error Source"
+	depends on ACPI_APEI && X86
+	select ACPI_HED
+	help
+	  Generic Hardware Error Source provides a way to report
+	  platform hardware errors (such as that from chipset). It
+	  works in so called "Firmware First" mode, that is, hardware
+	  errors are reported to firmware firstly, then reported to
+	  Linux by firmware. This way, some non-standard hardware
+	  error registers or non-standard hardware link can be checked
+	  by firmware to produce more valuable hardware error
+	  information for Linux.
+
+config ACPI_APEI_EINJ
+	tristate "APEI Error INJection (EINJ)"
+	depends on ACPI_APEI && DEBUG_FS
+	help
+	  EINJ provides a hardware error injection mechanism, it is
+	  mainly used for debugging and testing the other parts of
+	  APEI and some other RAS features.
diff --git a/drivers/acpi/apei/Makefile b/drivers/acpi/apei/Makefile
new file mode 100644
index 000000000000..b13b03a17789
--- /dev/null
+++ b/drivers/acpi/apei/Makefile
@@ -0,0 +1,5 @@
+obj-$(CONFIG_ACPI_APEI)		+= apei.o
+obj-$(CONFIG_ACPI_APEI_GHES)	+= ghes.o
+obj-$(CONFIG_ACPI_APEI_EINJ)	+= einj.o
+
+apei-y := apei-base.o hest.o cper.o erst.o
diff --git a/drivers/acpi/apei/apei-base.c b/drivers/acpi/apei/apei-base.c
new file mode 100644
index 000000000000..db3946e9c66b
--- /dev/null
+++ b/drivers/acpi/apei/apei-base.c
@@ -0,0 +1,593 @@
+/*
+ * apei-base.c - ACPI Platform Error Interface (APEI) supporting
+ * infrastructure
+ *
+ * APEI allows to report errors (for example from the chipset) to the
+ * the operating system. This improves NMI handling especially. In
+ * addition it supports error serialization and error injection.
+ *
+ * For more information about APEI, please refer to ACPI Specification
+ * version 4.0, chapter 17.
+ *
+ * This file has Common functions used by more than one APEI table,
+ * including framework of interpreter for ERST and EINJ; resource
+ * management for APEI registers.
+ *
+ * Copyright (C) 2009, Intel Corp.
+ *	Author: Huang Ying <ying.huang@intel.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License version
+ * 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/acpi.h>
+#include <linux/io.h>
+#include <linux/kref.h>
+#include <linux/rculist.h>
+#include <linux/interrupt.h>
+#include <linux/debugfs.h>
+#include <acpi/atomicio.h>
+
+#include "apei-internal.h"
+
+#define APEI_PFX "APEI: "
+
+/*
+ * APEI ERST (Error Record Serialization Table) and EINJ (Error
+ * INJection) interpreter framework.
+ */
+
+#define APEI_EXEC_PRESERVE_REGISTER	0x1
+
+void apei_exec_ctx_init(struct apei_exec_context *ctx,
+			struct apei_exec_ins_type *ins_table,
+			u32 instructions,
+			struct acpi_whea_header *action_table,
+			u32 entries)
+{
+	ctx->ins_table = ins_table;
+	ctx->instructions = instructions;
+	ctx->action_table = action_table;
+	ctx->entries = entries;
+}
+EXPORT_SYMBOL_GPL(apei_exec_ctx_init);
+
+int __apei_exec_read_register(struct acpi_whea_header *entry, u64 *val)
+{
+	int rc;
+
+	rc = acpi_atomic_read(val, &entry->register_region);
+	if (rc)
+		return rc;
+	*val >>= entry->register_region.bit_offset;
+	*val &= entry->mask;
+
+	return 0;
+}
+
+int apei_exec_read_register(struct apei_exec_context *ctx,
+			    struct acpi_whea_header *entry)
+{
+	int rc;
+	u64 val = 0;
+
+	rc = __apei_exec_read_register(entry, &val);
+	if (rc)
+		return rc;
+	ctx->value = val;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(apei_exec_read_register);
+
+int apei_exec_read_register_value(struct apei_exec_context *ctx,
+				  struct acpi_whea_header *entry)
+{
+	int rc;
+
+	rc = apei_exec_read_register(ctx, entry);
+	if (rc)
+		return rc;
+	ctx->value = (ctx->value == entry->value);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(apei_exec_read_register_value);
+
+int __apei_exec_write_register(struct acpi_whea_header *entry, u64 val)
+{
+	int rc;
+
+	val &= entry->mask;
+	val <<= entry->register_region.bit_offset;
+	if (entry->flags & APEI_EXEC_PRESERVE_REGISTER) {
+		u64 valr = 0;
+		rc = acpi_atomic_read(&valr, &entry->register_region);
+		if (rc)
+			return rc;
+		valr &= ~(entry->mask << entry->register_region.bit_offset);
+		val |= valr;
+	}
+	rc = acpi_atomic_write(val, &entry->register_region);
+
+	return rc;
+}
+
+int apei_exec_write_register(struct apei_exec_context *ctx,
+			     struct acpi_whea_header *entry)
+{
+	return __apei_exec_write_register(entry, ctx->value);
+}
+EXPORT_SYMBOL_GPL(apei_exec_write_register);
+
+int apei_exec_write_register_value(struct apei_exec_context *ctx,
+				   struct acpi_whea_header *entry)
+{
+	int rc;
+
+	ctx->value = entry->value;
+	rc = apei_exec_write_register(ctx, entry);
+
+	return rc;
+}
+EXPORT_SYMBOL_GPL(apei_exec_write_register_value);
+
+int apei_exec_noop(struct apei_exec_context *ctx,
+		   struct acpi_whea_header *entry)
+{
+	return 0;
+}
+EXPORT_SYMBOL_GPL(apei_exec_noop);
+
+/*
+ * Interpret the specified action. Go through whole action table,
+ * execute all instructions belong to the action.
+ */
+int apei_exec_run(struct apei_exec_context *ctx, u8 action)
+{
+	int rc;
+	u32 i, ip;
+	struct acpi_whea_header *entry;
+	apei_exec_ins_func_t run;
+
+	ctx->ip = 0;
+
+	/*
+	 * "ip" is the instruction pointer of current instruction,
+	 * "ctx->ip" specifies the next instruction to executed,
+	 * instruction "run" function may change the "ctx->ip" to
+	 * implement "goto" semantics.
+	 */
+rewind:
+	ip = 0;
+	for (i = 0; i < ctx->entries; i++) {
+		entry = &ctx->action_table[i];
+		if (entry->action != action)
+			continue;
+		if (ip == ctx->ip) {
+			if (entry->instruction >= ctx->instructions ||
+			    !ctx->ins_table[entry->instruction].run) {
+				pr_warning(FW_WARN APEI_PFX
+			"Invalid action table, unknown instruction type: %d\n",
+					   entry->instruction);
+				return -EINVAL;
+			}
+			run = ctx->ins_table[entry->instruction].run;
+			rc = run(ctx, entry);
+			if (rc < 0)
+				return rc;
+			else if (rc != APEI_EXEC_SET_IP)
+				ctx->ip++;
+		}
+		ip++;
+		if (ctx->ip < ip)
+			goto rewind;
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(apei_exec_run);
+
+typedef int (*apei_exec_entry_func_t)(struct apei_exec_context *ctx,
+				      struct acpi_whea_header *entry,
+				      void *data);
+
+static int apei_exec_for_each_entry(struct apei_exec_context *ctx,
+				    apei_exec_entry_func_t func,
+				    void *data,
+				    int *end)
+{
+	u8 ins;
+	int i, rc;
+	struct acpi_whea_header *entry;
+	struct apei_exec_ins_type *ins_table = ctx->ins_table;
+
+	for (i = 0; i < ctx->entries; i++) {
+		entry = ctx->action_table + i;
+		ins = entry->instruction;
+		if (end)
+			*end = i;
+		if (ins >= ctx->instructions || !ins_table[ins].run) {
+			pr_warning(FW_WARN APEI_PFX
+			"Invalid action table, unknown instruction type: %d\n",
+				   ins);
+			return -EINVAL;
+		}
+		rc = func(ctx, entry, data);
+		if (rc)
+			return rc;
+	}
+
+	return 0;
+}
+
+static int pre_map_gar_callback(struct apei_exec_context *ctx,
+				struct acpi_whea_header *entry,
+				void *data)
+{
+	u8 ins = entry->instruction;
+
+	if (ctx->ins_table[ins].flags & APEI_EXEC_INS_ACCESS_REGISTER)
+		return acpi_pre_map_gar(&entry->register_region);
+
+	return 0;
+}
+
+/*
+ * Pre-map all GARs in action table to make it possible to access them
+ * in NMI handler.
+ */
+int apei_exec_pre_map_gars(struct apei_exec_context *ctx)
+{
+	int rc, end;
+
+	rc = apei_exec_for_each_entry(ctx, pre_map_gar_callback,
+				      NULL, &end);
+	if (rc) {
+		struct apei_exec_context ctx_unmap;
+		memcpy(&ctx_unmap, ctx, sizeof(*ctx));
+		ctx_unmap.entries = end;
+		apei_exec_post_unmap_gars(&ctx_unmap);
+	}
+
+	return rc;
+}
+EXPORT_SYMBOL_GPL(apei_exec_pre_map_gars);
+
+static int post_unmap_gar_callback(struct apei_exec_context *ctx,
+				   struct acpi_whea_header *entry,
+				   void *data)
+{
+	u8 ins = entry->instruction;
+
+	if (ctx->ins_table[ins].flags & APEI_EXEC_INS_ACCESS_REGISTER)
+		acpi_post_unmap_gar(&entry->register_region);
+
+	return 0;
+}
+
+/* Post-unmap all GAR in action table. */
+int apei_exec_post_unmap_gars(struct apei_exec_context *ctx)
+{
+	return apei_exec_for_each_entry(ctx, post_unmap_gar_callback,
+					NULL, NULL);
+}
+EXPORT_SYMBOL_GPL(apei_exec_post_unmap_gars);
+
+/*
+ * Resource management for GARs in APEI
+ */
+struct apei_res {
+	struct list_head list;
+	unsigned long start;
+	unsigned long end;
+};
+
+/* Collect all resources requested, to avoid conflict */
+struct apei_resources apei_resources_all = {
+	.iomem = LIST_HEAD_INIT(apei_resources_all.iomem),
+	.ioport = LIST_HEAD_INIT(apei_resources_all.ioport),
+};
+
+static int apei_res_add(struct list_head *res_list,
+			unsigned long start, unsigned long size)
+{
+	struct apei_res *res, *resn, *res_ins = NULL;
+	unsigned long end = start + size;
+
+	if (end <= start)
+		return 0;
+repeat:
+	list_for_each_entry_safe(res, resn, res_list, list) {
+		if (res->start > end || res->end < start)
+			continue;
+		else if (end <= res->end && start >= res->start) {
+			kfree(res_ins);
+			return 0;
+		}
+		list_del(&res->list);
+		res->start = start = min(res->start, start);
+		res->end = end = max(res->end, end);
+		kfree(res_ins);
+		res_ins = res;
+		goto repeat;
+	}
+
+	if (res_ins)
+		list_add(&res_ins->list, res_list);
+	else {
+		res_ins = kmalloc(sizeof(*res), GFP_KERNEL);
+		if (!res_ins)
+			return -ENOMEM;
+		res_ins->start = start;
+		res_ins->end = end;
+		list_add(&res_ins->list, res_list);
+	}
+
+	return 0;
+}
+
+static int apei_res_sub(struct list_head *res_list1,
+			struct list_head *res_list2)
+{
+	struct apei_res *res1, *resn1, *res2, *res;
+	res1 = list_entry(res_list1->next, struct apei_res, list);
+	resn1 = list_entry(res1->list.next, struct apei_res, list);
+	while (&res1->list != res_list1) {
+		list_for_each_entry(res2, res_list2, list) {
+			if (res1->start >= res2->end ||
+			    res1->end <= res2->start)
+				continue;
+			else if (res1->end <= res2->end &&
+				 res1->start >= res2->start) {
+				list_del(&res1->list);
+				kfree(res1);
+				break;
+			} else if (res1->end > res2->end &&
+				   res1->start < res2->start) {
+				res = kmalloc(sizeof(*res), GFP_KERNEL);
+				if (!res)
+					return -ENOMEM;
+				res->start = res2->end;
+				res->end = res1->end;
+				res1->end = res2->start;
+				list_add(&res->list, &res1->list);
+				resn1 = res;
+			} else {
+				if (res1->start < res2->start)
+					res1->end = res2->start;
+				else
+					res1->start = res2->end;
+			}
+		}
+		res1 = resn1;
+		resn1 = list_entry(resn1->list.next, struct apei_res, list);
+	}
+
+	return 0;
+}
+
+static void apei_res_clean(struct list_head *res_list)
+{
+	struct apei_res *res, *resn;
+
+	list_for_each_entry_safe(res, resn, res_list, list) {
+		list_del(&res->list);
+		kfree(res);
+	}
+}
+
+void apei_resources_fini(struct apei_resources *resources)
+{
+	apei_res_clean(&resources->iomem);
+	apei_res_clean(&resources->ioport);
+}
+EXPORT_SYMBOL_GPL(apei_resources_fini);
+
+static int apei_resources_merge(struct apei_resources *resources1,
+				struct apei_resources *resources2)
+{
+	int rc;
+	struct apei_res *res;
+
+	list_for_each_entry(res, &resources2->iomem, list) {
+		rc = apei_res_add(&resources1->iomem, res->start,
+				  res->end - res->start);
+		if (rc)
+			return rc;
+	}
+	list_for_each_entry(res, &resources2->ioport, list) {
+		rc = apei_res_add(&resources1->ioport, res->start,
+				  res->end - res->start);
+		if (rc)
+			return rc;
+	}
+
+	return 0;
+}
+
+/*
+ * EINJ has two groups of GARs (EINJ table entry and trigger table
+ * entry), so common resources are subtracted from the trigger table
+ * resources before the second requesting.
+ */
+int apei_resources_sub(struct apei_resources *resources1,
+		       struct apei_resources *resources2)
+{
+	int rc;
+
+	rc = apei_res_sub(&resources1->iomem, &resources2->iomem);
+	if (rc)
+		return rc;
+	return apei_res_sub(&resources1->ioport, &resources2->ioport);
+}
+EXPORT_SYMBOL_GPL(apei_resources_sub);
+
+/*
+ * IO memory/port rersource management mechanism is used to check
+ * whether memory/port area used by GARs conflicts with normal memory
+ * or IO memory/port of devices.
+ */
+int apei_resources_request(struct apei_resources *resources,
+			   const char *desc)
+{
+	struct apei_res *res, *res_bak;
+	struct resource *r;
+
+	apei_resources_sub(resources, &apei_resources_all);
+
+	list_for_each_entry(res, &resources->iomem, list) {
+		r = request_mem_region(res->start, res->end - res->start,
+				       desc);
+		if (!r) {
+			pr_err(APEI_PFX
+		"Can not request iomem region <%016llx-%016llx> for GARs.\n",
+			       (unsigned long long)res->start,
+			       (unsigned long long)res->end);
+			res_bak = res;
+			goto err_unmap_iomem;
+		}
+	}
+
+	list_for_each_entry(res, &resources->ioport, list) {
+		r = request_region(res->start, res->end - res->start, desc);
+		if (!r) {
+			pr_err(APEI_PFX
+		"Can not request ioport region <%016llx-%016llx> for GARs.\n",
+			       (unsigned long long)res->start,
+			       (unsigned long long)res->end);
+			res_bak = res;
+			goto err_unmap_ioport;
+		}
+	}
+
+	apei_resources_merge(&apei_resources_all, resources);
+
+	return 0;
+err_unmap_ioport:
+	list_for_each_entry(res, &resources->ioport, list) {
+		if (res == res_bak)
+			break;
+		release_mem_region(res->start, res->end - res->start);
+	}
+	res_bak = NULL;
+err_unmap_iomem:
+	list_for_each_entry(res, &resources->iomem, list) {
+		if (res == res_bak)
+			break;
+		release_region(res->start, res->end - res->start);
+	}
+	return -EINVAL;
+}
+EXPORT_SYMBOL_GPL(apei_resources_request);
+
+void apei_resources_release(struct apei_resources *resources)
+{
+	struct apei_res *res;
+
+	list_for_each_entry(res, &resources->iomem, list)
+		release_mem_region(res->start, res->end - res->start);
+	list_for_each_entry(res, &resources->ioport, list)
+		release_region(res->start, res->end - res->start);
+
+	apei_resources_sub(&apei_resources_all, resources);
+}
+EXPORT_SYMBOL_GPL(apei_resources_release);
+
+static int apei_check_gar(struct acpi_generic_address *reg, u64 *paddr)
+{
+	u32 width, space_id;
+
+	width = reg->bit_width;
+	space_id = reg->space_id;
+	/* Handle possible alignment issues */
+	memcpy(paddr, &reg->address, sizeof(*paddr));
+	if (!*paddr) {
+		pr_warning(FW_BUG APEI_PFX
+			   "Invalid physical address in GAR [0x%llx/%u/%u]\n",
+			   *paddr, width, space_id);
+		return -EINVAL;
+	}
+
+	if ((width != 8) && (width != 16) && (width != 32) && (width != 64)) {
+		pr_warning(FW_BUG APEI_PFX
+			   "Invalid bit width in GAR [0x%llx/%u/%u]\n",
+			   *paddr, width, space_id);
+		return -EINVAL;
+	}
+
+	if (space_id != ACPI_ADR_SPACE_SYSTEM_MEMORY &&
+	    space_id != ACPI_ADR_SPACE_SYSTEM_IO) {
+		pr_warning(FW_BUG APEI_PFX
+			   "Invalid address space type in GAR [0x%llx/%u/%u]\n",
+			   *paddr, width, space_id);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int collect_res_callback(struct apei_exec_context *ctx,
+				struct acpi_whea_header *entry,
+				void *data)
+{
+	struct apei_resources *resources = data;
+	struct acpi_generic_address *reg = &entry->register_region;
+	u8 ins = entry->instruction;
+	u64 paddr;
+	int rc;
+
+	if (!(ctx->ins_table[ins].flags & APEI_EXEC_INS_ACCESS_REGISTER))
+		return 0;
+
+	rc = apei_check_gar(reg, &paddr);
+	if (rc)
+		return rc;
+
+	switch (reg->space_id) {
+	case ACPI_ADR_SPACE_SYSTEM_MEMORY:
+		return apei_res_add(&resources->iomem, paddr,
+				    reg->bit_width / 8);
+	case ACPI_ADR_SPACE_SYSTEM_IO:
+		return apei_res_add(&resources->ioport, paddr,
+				    reg->bit_width / 8);
+	default:
+		return -EINVAL;
+	}
+}
+
+/*
+ * Same register may be used by multiple instructions in GARs, so
+ * resources are collected before requesting.
+ */
+int apei_exec_collect_resources(struct apei_exec_context *ctx,
+				struct apei_resources *resources)
+{
+	return apei_exec_for_each_entry(ctx, collect_res_callback,
+					resources, NULL);
+}
+EXPORT_SYMBOL_GPL(apei_exec_collect_resources);
+
+struct dentry *apei_get_debugfs_dir(void)
+{
+	static struct dentry *dapei;
+
+	if (!dapei)
+		dapei = debugfs_create_dir("apei", NULL);
+
+	return dapei;
+}
+EXPORT_SYMBOL_GPL(apei_get_debugfs_dir);
diff --git a/drivers/acpi/apei/apei-internal.h b/drivers/acpi/apei/apei-internal.h
new file mode 100644
index 000000000000..18df1e940276
--- /dev/null
+++ b/drivers/acpi/apei/apei-internal.h
@@ -0,0 +1,114 @@
+/*
+ * apei-internal.h - ACPI Platform Error Interface internal
+ * definations.
+ */
+
+#ifndef APEI_INTERNAL_H
+#define APEI_INTERNAL_H
+
+#include <linux/cper.h>
+
+struct apei_exec_context;
+
+typedef int (*apei_exec_ins_func_t)(struct apei_exec_context *ctx,
+				    struct acpi_whea_header *entry);
+
+#define APEI_EXEC_INS_ACCESS_REGISTER	0x0001
+
+struct apei_exec_ins_type {
+	u32 flags;
+	apei_exec_ins_func_t run;
+};
+
+struct apei_exec_context {
+	u32 ip;
+	u64 value;
+	u64 var1;
+	u64 var2;
+	u64 src_base;
+	u64 dst_base;
+	struct apei_exec_ins_type *ins_table;
+	u32 instructions;
+	struct acpi_whea_header *action_table;
+	u32 entries;
+};
+
+void apei_exec_ctx_init(struct apei_exec_context *ctx,
+			struct apei_exec_ins_type *ins_table,
+			u32 instructions,
+			struct acpi_whea_header *action_table,
+			u32 entries);
+
+static inline void apei_exec_ctx_set_input(struct apei_exec_context *ctx,
+					   u64 input)
+{
+	ctx->value = input;
+}
+
+static inline u64 apei_exec_ctx_get_output(struct apei_exec_context *ctx)
+{
+	return ctx->value;
+}
+
+int apei_exec_run(struct apei_exec_context *ctx, u8 action);
+
+/* Common instruction implementation */
+
+/* IP has been set in instruction function */
+#define APEI_EXEC_SET_IP	1
+
+int __apei_exec_read_register(struct acpi_whea_header *entry, u64 *val);
+int __apei_exec_write_register(struct acpi_whea_header *entry, u64 val);
+int apei_exec_read_register(struct apei_exec_context *ctx,
+			    struct acpi_whea_header *entry);
+int apei_exec_read_register_value(struct apei_exec_context *ctx,
+				  struct acpi_whea_header *entry);
+int apei_exec_write_register(struct apei_exec_context *ctx,
+			     struct acpi_whea_header *entry);
+int apei_exec_write_register_value(struct apei_exec_context *ctx,
+				   struct acpi_whea_header *entry);
+int apei_exec_noop(struct apei_exec_context *ctx,
+		   struct acpi_whea_header *entry);
+int apei_exec_pre_map_gars(struct apei_exec_context *ctx);
+int apei_exec_post_unmap_gars(struct apei_exec_context *ctx);
+
+struct apei_resources {
+	struct list_head iomem;
+	struct list_head ioport;
+};
+
+static inline void apei_resources_init(struct apei_resources *resources)
+{
+	INIT_LIST_HEAD(&resources->iomem);
+	INIT_LIST_HEAD(&resources->ioport);
+}
+
+void apei_resources_fini(struct apei_resources *resources);
+int apei_resources_sub(struct apei_resources *resources1,
+		       struct apei_resources *resources2);
+int apei_resources_request(struct apei_resources *resources,
+			   const char *desc);
+void apei_resources_release(struct apei_resources *resources);
+int apei_exec_collect_resources(struct apei_exec_context *ctx,
+				struct apei_resources *resources);
+
+struct dentry;
+struct dentry *apei_get_debugfs_dir(void);
+
+#define apei_estatus_for_each_section(estatus, section)			\
+	for (section = (struct acpi_hest_generic_data *)(estatus + 1);	\
+	     (void *)section - (void *)estatus < estatus->data_length;	\
+	     section = (void *)(section+1) + section->error_data_length)
+
+static inline u32 apei_estatus_len(struct acpi_hest_generic_status *estatus)
+{
+	if (estatus->raw_data_length)
+		return estatus->raw_data_offset + \
+			estatus->raw_data_length;
+	else
+		return sizeof(*estatus) + estatus->data_length;
+}
+
+int apei_estatus_check_header(const struct acpi_hest_generic_status *estatus);
+int apei_estatus_check(const struct acpi_hest_generic_status *estatus);
+#endif
diff --git a/drivers/acpi/apei/cper.c b/drivers/acpi/apei/cper.c
new file mode 100644
index 000000000000..f4cf2fc4c8c1
--- /dev/null
+++ b/drivers/acpi/apei/cper.c
@@ -0,0 +1,84 @@
+/*
+ * UEFI Common Platform Error Record (CPER) support
+ *
+ * Copyright (C) 2010, Intel Corp.
+ *	Author: Huang Ying <ying.huang@intel.com>
+ *
+ * CPER is the format used to describe platform hardware error by
+ * various APEI tables, such as ERST, BERT and HEST etc.
+ *
+ * For more information about CPER, please refer to Appendix N of UEFI
+ * Specification version 2.3.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License version
+ * 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/time.h>
+#include <linux/cper.h>
+#include <linux/acpi.h>
+
+/*
+ * CPER record ID need to be unique even after reboot, because record
+ * ID is used as index for ERST storage, while CPER records from
+ * multiple boot may co-exist in ERST.
+ */
+u64 cper_next_record_id(void)
+{
+	static atomic64_t seq;
+
+	if (!atomic64_read(&seq))
+		atomic64_set(&seq, ((u64)get_seconds()) << 32);
+
+	return atomic64_inc_return(&seq);
+}
+EXPORT_SYMBOL_GPL(cper_next_record_id);
+
+int apei_estatus_check_header(const struct acpi_hest_generic_status *estatus)
+{
+	if (estatus->data_length &&
+	    estatus->data_length < sizeof(struct acpi_hest_generic_data))
+		return -EINVAL;
+	if (estatus->raw_data_length &&
+	    estatus->raw_data_offset < sizeof(*estatus) + estatus->data_length)
+		return -EINVAL;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(apei_estatus_check_header);
+
+int apei_estatus_check(const struct acpi_hest_generic_status *estatus)
+{
+	struct acpi_hest_generic_data *gdata;
+	unsigned int data_len, gedata_len;
+	int rc;
+
+	rc = apei_estatus_check_header(estatus);
+	if (rc)
+		return rc;
+	data_len = estatus->data_length;
+	gdata = (struct acpi_hest_generic_data *)(estatus + 1);
+	while (data_len > sizeof(*gdata)) {
+		gedata_len = gdata->error_data_length;
+		if (gedata_len > data_len - sizeof(*gdata))
+			return -EINVAL;
+		data_len -= gedata_len + sizeof(*gdata);
+	}
+	if (data_len)
+		return -EINVAL;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(apei_estatus_check);
diff --git a/drivers/acpi/apei/einj.c b/drivers/acpi/apei/einj.c
new file mode 100644
index 000000000000..465c885938ee
--- /dev/null
+++ b/drivers/acpi/apei/einj.c
@@ -0,0 +1,548 @@
+/*
+ * APEI Error INJection support
+ *
+ * EINJ provides a hardware error injection mechanism, this is useful
+ * for debugging and testing of other APEI and RAS features.
+ *
+ * For more information about EINJ, please refer to ACPI Specification
+ * version 4.0, section 17.5.
+ *
+ * Copyright 2009-2010 Intel Corp.
+ *   Author: Huang Ying <ying.huang@intel.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License version
+ * 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/io.h>
+#include <linux/debugfs.h>
+#include <linux/seq_file.h>
+#include <linux/nmi.h>
+#include <linux/delay.h>
+#include <acpi/acpi.h>
+
+#include "apei-internal.h"
+
+#define EINJ_PFX "EINJ: "
+
+#define SPIN_UNIT		100			/* 100ns */
+/* Firmware should respond within 1 miliseconds */
+#define FIRMWARE_TIMEOUT	(1 * NSEC_PER_MSEC)
+
+/*
+ * Some BIOSes allow parameters to the SET_ERROR_TYPE entries in the
+ * EINJ table through an unpublished extension. Use with caution as
+ * most will ignore the parameter and make their own choice of address
+ * for error injection.
+ */
+struct einj_parameter {
+	u64 type;
+	u64 reserved1;
+	u64 reserved2;
+	u64 param1;
+	u64 param2;
+};
+
+#define EINJ_OP_BUSY			0x1
+#define EINJ_STATUS_SUCCESS		0x0
+#define EINJ_STATUS_FAIL		0x1
+#define EINJ_STATUS_INVAL		0x2
+
+#define EINJ_TAB_ENTRY(tab)						\
+	((struct acpi_whea_header *)((char *)(tab) +			\
+				    sizeof(struct acpi_table_einj)))
+
+static struct acpi_table_einj *einj_tab;
+
+static struct apei_resources einj_resources;
+
+static struct apei_exec_ins_type einj_ins_type[] = {
+	[ACPI_EINJ_READ_REGISTER] = {
+		.flags = APEI_EXEC_INS_ACCESS_REGISTER,
+		.run   = apei_exec_read_register,
+	},
+	[ACPI_EINJ_READ_REGISTER_VALUE] = {
+		.flags = APEI_EXEC_INS_ACCESS_REGISTER,
+		.run   = apei_exec_read_register_value,
+	},
+	[ACPI_EINJ_WRITE_REGISTER] = {
+		.flags = APEI_EXEC_INS_ACCESS_REGISTER,
+		.run   = apei_exec_write_register,
+	},
+	[ACPI_EINJ_WRITE_REGISTER_VALUE] = {
+		.flags = APEI_EXEC_INS_ACCESS_REGISTER,
+		.run   = apei_exec_write_register_value,
+	},
+	[ACPI_EINJ_NOOP] = {
+		.flags = 0,
+		.run   = apei_exec_noop,
+	},
+};
+
+/*
+ * Prevent EINJ interpreter to run simultaneously, because the
+ * corresponding firmware implementation may not work properly when
+ * invoked simultaneously.
+ */
+static DEFINE_MUTEX(einj_mutex);
+
+static struct einj_parameter *einj_param;
+
+static void einj_exec_ctx_init(struct apei_exec_context *ctx)
+{
+	apei_exec_ctx_init(ctx, einj_ins_type, ARRAY_SIZE(einj_ins_type),
+			   EINJ_TAB_ENTRY(einj_tab), einj_tab->entries);
+}
+
+static int __einj_get_available_error_type(u32 *type)
+{
+	struct apei_exec_context ctx;
+	int rc;
+
+	einj_exec_ctx_init(&ctx);
+	rc = apei_exec_run(&ctx, ACPI_EINJ_GET_ERROR_TYPE);
+	if (rc)
+		return rc;
+	*type = apei_exec_ctx_get_output(&ctx);
+
+	return 0;
+}
+
+/* Get error injection capabilities of the platform */
+static int einj_get_available_error_type(u32 *type)
+{
+	int rc;
+
+	mutex_lock(&einj_mutex);
+	rc = __einj_get_available_error_type(type);
+	mutex_unlock(&einj_mutex);
+
+	return rc;
+}
+
+static int einj_timedout(u64 *t)
+{
+	if ((s64)*t < SPIN_UNIT) {
+		pr_warning(FW_WARN EINJ_PFX
+			   "Firmware does not respond in time\n");
+		return 1;
+	}
+	*t -= SPIN_UNIT;
+	ndelay(SPIN_UNIT);
+	touch_nmi_watchdog();
+	return 0;
+}
+
+static u64 einj_get_parameter_address(void)
+{
+	int i;
+	u64 paddr = 0;
+	struct acpi_whea_header *entry;
+
+	entry = EINJ_TAB_ENTRY(einj_tab);
+	for (i = 0; i < einj_tab->entries; i++) {
+		if (entry->action == ACPI_EINJ_SET_ERROR_TYPE &&
+		    entry->instruction == ACPI_EINJ_WRITE_REGISTER &&
+		    entry->register_region.space_id ==
+		    ACPI_ADR_SPACE_SYSTEM_MEMORY)
+			memcpy(&paddr, &entry->register_region.address,
+			       sizeof(paddr));
+		entry++;
+	}
+
+	return paddr;
+}
+
+/* do sanity check to trigger table */
+static int einj_check_trigger_header(struct acpi_einj_trigger *trigger_tab)
+{
+	if (trigger_tab->header_size != sizeof(struct acpi_einj_trigger))
+		return -EINVAL;
+	if (trigger_tab->table_size > PAGE_SIZE ||
+	    trigger_tab->table_size <= trigger_tab->header_size)
+		return -EINVAL;
+	if (trigger_tab->entry_count !=
+	    (trigger_tab->table_size - trigger_tab->header_size) /
+	    sizeof(struct acpi_einj_entry))
+		return -EINVAL;
+
+	return 0;
+}
+
+/* Execute instructions in trigger error action table */
+static int __einj_error_trigger(u64 trigger_paddr)
+{
+	struct acpi_einj_trigger *trigger_tab = NULL;
+	struct apei_exec_context trigger_ctx;
+	struct apei_resources trigger_resources;
+	struct acpi_whea_header *trigger_entry;
+	struct resource *r;
+	u32 table_size;
+	int rc = -EIO;
+
+	r = request_mem_region(trigger_paddr, sizeof(*trigger_tab),
+			       "APEI EINJ Trigger Table");
+	if (!r) {
+		pr_err(EINJ_PFX
+	"Can not request iomem region <%016llx-%016llx> for Trigger table.\n",
+		       (unsigned long long)trigger_paddr,
+		       (unsigned long long)trigger_paddr+sizeof(*trigger_tab));
+		goto out;
+	}
+	trigger_tab = ioremap_cache(trigger_paddr, sizeof(*trigger_tab));
+	if (!trigger_tab) {
+		pr_err(EINJ_PFX "Failed to map trigger table!\n");
+		goto out_rel_header;
+	}
+	rc = einj_check_trigger_header(trigger_tab);
+	if (rc) {
+		pr_warning(FW_BUG EINJ_PFX
+			   "The trigger error action table is invalid\n");
+		goto out_rel_header;
+	}
+	rc = -EIO;
+	table_size = trigger_tab->table_size;
+	r = request_mem_region(trigger_paddr + sizeof(*trigger_tab),
+			       table_size - sizeof(*trigger_tab),
+			       "APEI EINJ Trigger Table");
+	if (!r) {
+		pr_err(EINJ_PFX
+"Can not request iomem region <%016llx-%016llx> for Trigger Table Entry.\n",
+		       (unsigned long long)trigger_paddr+sizeof(*trigger_tab),
+		       (unsigned long long)trigger_paddr + table_size);
+		goto out_rel_header;
+	}
+	iounmap(trigger_tab);
+	trigger_tab = ioremap_cache(trigger_paddr, table_size);
+	if (!trigger_tab) {
+		pr_err(EINJ_PFX "Failed to map trigger table!\n");
+		goto out_rel_entry;
+	}
+	trigger_entry = (struct acpi_whea_header *)
+		((char *)trigger_tab + sizeof(struct acpi_einj_trigger));
+	apei_resources_init(&trigger_resources);
+	apei_exec_ctx_init(&trigger_ctx, einj_ins_type,
+			   ARRAY_SIZE(einj_ins_type),
+			   trigger_entry, trigger_tab->entry_count);
+	rc = apei_exec_collect_resources(&trigger_ctx, &trigger_resources);
+	if (rc)
+		goto out_fini;
+	rc = apei_resources_sub(&trigger_resources, &einj_resources);
+	if (rc)
+		goto out_fini;
+	rc = apei_resources_request(&trigger_resources, "APEI EINJ Trigger");
+	if (rc)
+		goto out_fini;
+	rc = apei_exec_pre_map_gars(&trigger_ctx);
+	if (rc)
+		goto out_release;
+
+	rc = apei_exec_run(&trigger_ctx, ACPI_EINJ_TRIGGER_ERROR);
+
+	apei_exec_post_unmap_gars(&trigger_ctx);
+out_release:
+	apei_resources_release(&trigger_resources);
+out_fini:
+	apei_resources_fini(&trigger_resources);
+out_rel_entry:
+	release_mem_region(trigger_paddr + sizeof(*trigger_tab),
+			   table_size - sizeof(*trigger_tab));
+out_rel_header:
+	release_mem_region(trigger_paddr, sizeof(*trigger_tab));
+out:
+	if (trigger_tab)
+		iounmap(trigger_tab);
+
+	return rc;
+}
+
+static int __einj_error_inject(u32 type, u64 param1, u64 param2)
+{
+	struct apei_exec_context ctx;
+	u64 val, trigger_paddr, timeout = FIRMWARE_TIMEOUT;
+	int rc;
+
+	einj_exec_ctx_init(&ctx);
+
+	rc = apei_exec_run(&ctx, ACPI_EINJ_BEGIN_OPERATION);
+	if (rc)
+		return rc;
+	apei_exec_ctx_set_input(&ctx, type);
+	rc = apei_exec_run(&ctx, ACPI_EINJ_SET_ERROR_TYPE);
+	if (rc)
+		return rc;
+	if (einj_param) {
+		writeq(param1, &einj_param->param1);
+		writeq(param2, &einj_param->param2);
+	}
+	rc = apei_exec_run(&ctx, ACPI_EINJ_EXECUTE_OPERATION);
+	if (rc)
+		return rc;
+	for (;;) {
+		rc = apei_exec_run(&ctx, ACPI_EINJ_CHECK_BUSY_STATUS);
+		if (rc)
+			return rc;
+		val = apei_exec_ctx_get_output(&ctx);
+		if (!(val & EINJ_OP_BUSY))
+			break;
+		if (einj_timedout(&timeout))
+			return -EIO;
+	}
+	rc = apei_exec_run(&ctx, ACPI_EINJ_GET_COMMAND_STATUS);
+	if (rc)
+		return rc;
+	val = apei_exec_ctx_get_output(&ctx);
+	if (val != EINJ_STATUS_SUCCESS)
+		return -EBUSY;
+
+	rc = apei_exec_run(&ctx, ACPI_EINJ_GET_TRIGGER_TABLE);
+	if (rc)
+		return rc;
+	trigger_paddr = apei_exec_ctx_get_output(&ctx);
+	rc = __einj_error_trigger(trigger_paddr);
+	if (rc)
+		return rc;
+	rc = apei_exec_run(&ctx, ACPI_EINJ_END_OPERATION);
+
+	return rc;
+}
+
+/* Inject the specified hardware error */
+static int einj_error_inject(u32 type, u64 param1, u64 param2)
+{
+	int rc;
+
+	mutex_lock(&einj_mutex);
+	rc = __einj_error_inject(type, param1, param2);
+	mutex_unlock(&einj_mutex);
+
+	return rc;
+}
+
+static u32 error_type;
+static u64 error_param1;
+static u64 error_param2;
+static struct dentry *einj_debug_dir;
+
+static int available_error_type_show(struct seq_file *m, void *v)
+{
+	int rc;
+	u32 available_error_type = 0;
+
+	rc = einj_get_available_error_type(&available_error_type);
+	if (rc)
+		return rc;
+	if (available_error_type & 0x0001)
+		seq_printf(m, "0x00000001\tProcessor Correctable\n");
+	if (available_error_type & 0x0002)
+		seq_printf(m, "0x00000002\tProcessor Uncorrectable non-fatal\n");
+	if (available_error_type & 0x0004)
+		seq_printf(m, "0x00000004\tProcessor Uncorrectable fatal\n");
+	if (available_error_type & 0x0008)
+		seq_printf(m, "0x00000008\tMemory Correctable\n");
+	if (available_error_type & 0x0010)
+		seq_printf(m, "0x00000010\tMemory Uncorrectable non-fatal\n");
+	if (available_error_type & 0x0020)
+		seq_printf(m, "0x00000020\tMemory Uncorrectable fatal\n");
+	if (available_error_type & 0x0040)
+		seq_printf(m, "0x00000040\tPCI Express Correctable\n");
+	if (available_error_type & 0x0080)
+		seq_printf(m, "0x00000080\tPCI Express Uncorrectable non-fatal\n");
+	if (available_error_type & 0x0100)
+		seq_printf(m, "0x00000100\tPCI Express Uncorrectable fatal\n");
+	if (available_error_type & 0x0200)
+		seq_printf(m, "0x00000200\tPlatform Correctable\n");
+	if (available_error_type & 0x0400)
+		seq_printf(m, "0x00000400\tPlatform Uncorrectable non-fatal\n");
+	if (available_error_type & 0x0800)
+		seq_printf(m, "0x00000800\tPlatform Uncorrectable fatal\n");
+
+	return 0;
+}
+
+static int available_error_type_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, available_error_type_show, NULL);
+}
+
+static const struct file_operations available_error_type_fops = {
+	.open		= available_error_type_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+static int error_type_get(void *data, u64 *val)
+{
+	*val = error_type;
+
+	return 0;
+}
+
+static int error_type_set(void *data, u64 val)
+{
+	int rc;
+	u32 available_error_type = 0;
+
+	/* Only one error type can be specified */
+	if (val & (val - 1))
+		return -EINVAL;
+	rc = einj_get_available_error_type(&available_error_type);
+	if (rc)
+		return rc;
+	if (!(val & available_error_type))
+		return -EINVAL;
+	error_type = val;
+
+	return 0;
+}
+
+DEFINE_SIMPLE_ATTRIBUTE(error_type_fops, error_type_get,
+			error_type_set, "0x%llx\n");
+
+static int error_inject_set(void *data, u64 val)
+{
+	if (!error_type)
+		return -EINVAL;
+
+	return einj_error_inject(error_type, error_param1, error_param2);
+}
+
+DEFINE_SIMPLE_ATTRIBUTE(error_inject_fops, NULL,
+			error_inject_set, "%llu\n");
+
+static int einj_check_table(struct acpi_table_einj *einj_tab)
+{
+	if (einj_tab->header_length != sizeof(struct acpi_table_einj))
+		return -EINVAL;
+	if (einj_tab->header.length < sizeof(struct acpi_table_einj))
+		return -EINVAL;
+	if (einj_tab->entries !=
+	    (einj_tab->header.length - sizeof(struct acpi_table_einj)) /
+	    sizeof(struct acpi_einj_entry))
+		return -EINVAL;
+
+	return 0;
+}
+
+static int __init einj_init(void)
+{
+	int rc;
+	u64 param_paddr;
+	acpi_status status;
+	struct dentry *fentry;
+	struct apei_exec_context ctx;
+
+	if (acpi_disabled)
+		return -ENODEV;
+
+	status = acpi_get_table(ACPI_SIG_EINJ, 0,
+				(struct acpi_table_header **)&einj_tab);
+	if (status == AE_NOT_FOUND) {
+		pr_info(EINJ_PFX "Table is not found!\n");
+		return -ENODEV;
+	} else if (ACPI_FAILURE(status)) {
+		const char *msg = acpi_format_exception(status);
+		pr_err(EINJ_PFX "Failed to get table, %s\n", msg);
+		return -EINVAL;
+	}
+
+	rc = einj_check_table(einj_tab);
+	if (rc) {
+		pr_warning(FW_BUG EINJ_PFX "EINJ table is invalid\n");
+		return -EINVAL;
+	}
+
+	rc = -ENOMEM;
+	einj_debug_dir = debugfs_create_dir("einj", apei_get_debugfs_dir());
+	if (!einj_debug_dir)
+		goto err_cleanup;
+	fentry = debugfs_create_file("available_error_type", S_IRUSR,
+				     einj_debug_dir, NULL,
+				     &available_error_type_fops);
+	if (!fentry)
+		goto err_cleanup;
+	fentry = debugfs_create_file("error_type", S_IRUSR | S_IWUSR,
+				     einj_debug_dir, NULL, &error_type_fops);
+	if (!fentry)
+		goto err_cleanup;
+	fentry = debugfs_create_x64("param1", S_IRUSR | S_IWUSR,
+				    einj_debug_dir, &error_param1);
+	if (!fentry)
+		goto err_cleanup;
+	fentry = debugfs_create_x64("param2", S_IRUSR | S_IWUSR,
+				    einj_debug_dir, &error_param2);
+	if (!fentry)
+		goto err_cleanup;
+	fentry = debugfs_create_file("error_inject", S_IWUSR,
+				     einj_debug_dir, NULL, &error_inject_fops);
+	if (!fentry)
+		goto err_cleanup;
+
+	apei_resources_init(&einj_resources);
+	einj_exec_ctx_init(&ctx);
+	rc = apei_exec_collect_resources(&ctx, &einj_resources);
+	if (rc)
+		goto err_fini;
+	rc = apei_resources_request(&einj_resources, "APEI EINJ");
+	if (rc)
+		goto err_fini;
+	rc = apei_exec_pre_map_gars(&ctx);
+	if (rc)
+		goto err_release;
+	param_paddr = einj_get_parameter_address();
+	if (param_paddr) {
+		einj_param = ioremap(param_paddr, sizeof(*einj_param));
+		rc = -ENOMEM;
+		if (!einj_param)
+			goto err_unmap;
+	}
+
+	pr_info(EINJ_PFX "Error INJection is initialized.\n");
+
+	return 0;
+
+err_unmap:
+	apei_exec_post_unmap_gars(&ctx);
+err_release:
+	apei_resources_release(&einj_resources);
+err_fini:
+	apei_resources_fini(&einj_resources);
+err_cleanup:
+	debugfs_remove_recursive(einj_debug_dir);
+
+	return rc;
+}
+
+static void __exit einj_exit(void)
+{
+	struct apei_exec_context ctx;
+
+	if (einj_param)
+		iounmap(einj_param);
+	einj_exec_ctx_init(&ctx);
+	apei_exec_post_unmap_gars(&ctx);
+	apei_resources_release(&einj_resources);
+	apei_resources_fini(&einj_resources);
+	debugfs_remove_recursive(einj_debug_dir);
+}
+
+module_init(einj_init);
+module_exit(einj_exit);
+
+MODULE_AUTHOR("Huang Ying");
+MODULE_DESCRIPTION("APEI Error INJection support");
+MODULE_LICENSE("GPL");
diff --git a/drivers/acpi/apei/erst.c b/drivers/acpi/apei/erst.c
new file mode 100644
index 000000000000..2ebc39115507
--- /dev/null
+++ b/drivers/acpi/apei/erst.c
@@ -0,0 +1,855 @@
+/*
+ * APEI Error Record Serialization Table support
+ *
+ * ERST is a way provided by APEI to save and retrieve hardware error
+ * infomation to and from a persistent store.
+ *
+ * For more information about ERST, please refer to ACPI Specification
+ * version 4.0, section 17.4.
+ *
+ * Copyright 2010 Intel Corp.
+ *   Author: Huang Ying <ying.huang@intel.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License version
+ * 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/delay.h>
+#include <linux/io.h>
+#include <linux/acpi.h>
+#include <linux/uaccess.h>
+#include <linux/cper.h>
+#include <linux/nmi.h>
+#include <acpi/apei.h>
+
+#include "apei-internal.h"
+
+#define ERST_PFX "ERST: "
+
+/* ERST command status */
+#define ERST_STATUS_SUCCESS			0x0
+#define ERST_STATUS_NOT_ENOUGH_SPACE		0x1
+#define ERST_STATUS_HARDWARE_NOT_AVAILABLE	0x2
+#define ERST_STATUS_FAILED			0x3
+#define ERST_STATUS_RECORD_STORE_EMPTY		0x4
+#define ERST_STATUS_RECORD_NOT_FOUND		0x5
+
+#define ERST_TAB_ENTRY(tab)						\
+	((struct acpi_whea_header *)((char *)(tab) +			\
+				     sizeof(struct acpi_table_erst)))
+
+#define SPIN_UNIT		100			/* 100ns */
+/* Firmware should respond within 1 miliseconds */
+#define FIRMWARE_TIMEOUT	(1 * NSEC_PER_MSEC)
+#define FIRMWARE_MAX_STALL	50			/* 50us */
+
+int erst_disable;
+EXPORT_SYMBOL_GPL(erst_disable);
+
+static struct acpi_table_erst *erst_tab;
+
+/* ERST Error Log Address Range atrributes */
+#define ERST_RANGE_RESERVED	0x0001
+#define ERST_RANGE_NVRAM	0x0002
+#define ERST_RANGE_SLOW		0x0004
+
+/*
+ * ERST Error Log Address Range, used as buffer for reading/writing
+ * error records.
+ */
+static struct erst_erange {
+	u64 base;
+	u64 size;
+	void __iomem *vaddr;
+	u32 attr;
+} erst_erange;
+
+/*
+ * Prevent ERST interpreter to run simultaneously, because the
+ * corresponding firmware implementation may not work properly when
+ * invoked simultaneously.
+ *
+ * It is used to provide exclusive accessing for ERST Error Log
+ * Address Range too.
+ */
+static DEFINE_SPINLOCK(erst_lock);
+
+static inline int erst_errno(int command_status)
+{
+	switch (command_status) {
+	case ERST_STATUS_SUCCESS:
+		return 0;
+	case ERST_STATUS_HARDWARE_NOT_AVAILABLE:
+		return -ENODEV;
+	case ERST_STATUS_NOT_ENOUGH_SPACE:
+		return -ENOSPC;
+	case ERST_STATUS_RECORD_STORE_EMPTY:
+	case ERST_STATUS_RECORD_NOT_FOUND:
+		return -ENOENT;
+	default:
+		return -EINVAL;
+	}
+}
+
+static int erst_timedout(u64 *t, u64 spin_unit)
+{
+	if ((s64)*t < spin_unit) {
+		pr_warning(FW_WARN ERST_PFX
+			   "Firmware does not respond in time\n");
+		return 1;
+	}
+	*t -= spin_unit;
+	ndelay(spin_unit);
+	touch_nmi_watchdog();
+	return 0;
+}
+
+static int erst_exec_load_var1(struct apei_exec_context *ctx,
+			       struct acpi_whea_header *entry)
+{
+	return __apei_exec_read_register(entry, &ctx->var1);
+}
+
+static int erst_exec_load_var2(struct apei_exec_context *ctx,
+			       struct acpi_whea_header *entry)
+{
+	return __apei_exec_read_register(entry, &ctx->var2);
+}
+
+static int erst_exec_store_var1(struct apei_exec_context *ctx,
+				struct acpi_whea_header *entry)
+{
+	return __apei_exec_write_register(entry, ctx->var1);
+}
+
+static int erst_exec_add(struct apei_exec_context *ctx,
+			 struct acpi_whea_header *entry)
+{
+	ctx->var1 += ctx->var2;
+	return 0;
+}
+
+static int erst_exec_subtract(struct apei_exec_context *ctx,
+			      struct acpi_whea_header *entry)
+{
+	ctx->var1 -= ctx->var2;
+	return 0;
+}
+
+static int erst_exec_add_value(struct apei_exec_context *ctx,
+			       struct acpi_whea_header *entry)
+{
+	int rc;
+	u64 val;
+
+	rc = __apei_exec_read_register(entry, &val);
+	if (rc)
+		return rc;
+	val += ctx->value;
+	rc = __apei_exec_write_register(entry, val);
+	return rc;
+}
+
+static int erst_exec_subtract_value(struct apei_exec_context *ctx,
+				    struct acpi_whea_header *entry)
+{
+	int rc;
+	u64 val;
+
+	rc = __apei_exec_read_register(entry, &val);
+	if (rc)
+		return rc;
+	val -= ctx->value;
+	rc = __apei_exec_write_register(entry, val);
+	return rc;
+}
+
+static int erst_exec_stall(struct apei_exec_context *ctx,
+			   struct acpi_whea_header *entry)
+{
+	u64 stall_time;
+
+	if (ctx->value > FIRMWARE_MAX_STALL) {
+		if (!in_nmi())
+			pr_warning(FW_WARN ERST_PFX
+			"Too long stall time for stall instruction: %llx.\n",
+				   ctx->value);
+		stall_time = FIRMWARE_MAX_STALL;
+	} else
+		stall_time = ctx->value;
+	udelay(stall_time);
+	return 0;
+}
+
+static int erst_exec_stall_while_true(struct apei_exec_context *ctx,
+				      struct acpi_whea_header *entry)
+{
+	int rc;
+	u64 val;
+	u64 timeout = FIRMWARE_TIMEOUT;
+	u64 stall_time;
+
+	if (ctx->var1 > FIRMWARE_MAX_STALL) {
+		if (!in_nmi())
+			pr_warning(FW_WARN ERST_PFX
+		"Too long stall time for stall while true instruction: %llx.\n",
+				   ctx->var1);
+		stall_time = FIRMWARE_MAX_STALL;
+	} else
+		stall_time = ctx->var1;
+
+	for (;;) {
+		rc = __apei_exec_read_register(entry, &val);
+		if (rc)
+			return rc;
+		if (val != ctx->value)
+			break;
+		if (erst_timedout(&timeout, stall_time * NSEC_PER_USEC))
+			return -EIO;
+	}
+	return 0;
+}
+
+static int erst_exec_skip_next_instruction_if_true(
+	struct apei_exec_context *ctx,
+	struct acpi_whea_header *entry)
+{
+	int rc;
+	u64 val;
+
+	rc = __apei_exec_read_register(entry, &val);
+	if (rc)
+		return rc;
+	if (val == ctx->value) {
+		ctx->ip += 2;
+		return APEI_EXEC_SET_IP;
+	}
+
+	return 0;
+}
+
+static int erst_exec_goto(struct apei_exec_context *ctx,
+			  struct acpi_whea_header *entry)
+{
+	ctx->ip = ctx->value;
+	return APEI_EXEC_SET_IP;
+}
+
+static int erst_exec_set_src_address_base(struct apei_exec_context *ctx,
+					  struct acpi_whea_header *entry)
+{
+	return __apei_exec_read_register(entry, &ctx->src_base);
+}
+
+static int erst_exec_set_dst_address_base(struct apei_exec_context *ctx,
+					  struct acpi_whea_header *entry)
+{
+	return __apei_exec_read_register(entry, &ctx->dst_base);
+}
+
+static int erst_exec_move_data(struct apei_exec_context *ctx,
+			       struct acpi_whea_header *entry)
+{
+	int rc;
+	u64 offset;
+
+	rc = __apei_exec_read_register(entry, &offset);
+	if (rc)
+		return rc;
+	memmove((void *)ctx->dst_base + offset,
+		(void *)ctx->src_base + offset,
+		ctx->var2);
+
+	return 0;
+}
+
+static struct apei_exec_ins_type erst_ins_type[] = {
+	[ACPI_ERST_READ_REGISTER] = {
+		.flags = APEI_EXEC_INS_ACCESS_REGISTER,
+		.run = apei_exec_read_register,
+	},
+	[ACPI_ERST_READ_REGISTER_VALUE] = {
+		.flags = APEI_EXEC_INS_ACCESS_REGISTER,
+		.run = apei_exec_read_register_value,
+	},
+	[ACPI_ERST_WRITE_REGISTER] = {
+		.flags = APEI_EXEC_INS_ACCESS_REGISTER,
+		.run = apei_exec_write_register,
+	},
+	[ACPI_ERST_WRITE_REGISTER_VALUE] = {
+		.flags = APEI_EXEC_INS_ACCESS_REGISTER,
+		.run = apei_exec_write_register_value,
+	},
+	[ACPI_ERST_NOOP] = {
+		.flags = 0,
+		.run = apei_exec_noop,
+	},
+	[ACPI_ERST_LOAD_VAR1] = {
+		.flags = APEI_EXEC_INS_ACCESS_REGISTER,
+		.run = erst_exec_load_var1,
+	},
+	[ACPI_ERST_LOAD_VAR2] = {
+		.flags = APEI_EXEC_INS_ACCESS_REGISTER,
+		.run = erst_exec_load_var2,
+	},
+	[ACPI_ERST_STORE_VAR1] = {
+		.flags = APEI_EXEC_INS_ACCESS_REGISTER,
+		.run = erst_exec_store_var1,
+	},
+	[ACPI_ERST_ADD] = {
+		.flags = 0,
+		.run = erst_exec_add,
+	},
+	[ACPI_ERST_SUBTRACT] = {
+		.flags = 0,
+		.run = erst_exec_subtract,
+	},
+	[ACPI_ERST_ADD_VALUE] = {
+		.flags = APEI_EXEC_INS_ACCESS_REGISTER,
+		.run = erst_exec_add_value,
+	},
+	[ACPI_ERST_SUBTRACT_VALUE] = {
+		.flags = APEI_EXEC_INS_ACCESS_REGISTER,
+		.run = erst_exec_subtract_value,
+	},
+	[ACPI_ERST_STALL] = {
+		.flags = 0,
+		.run = erst_exec_stall,
+	},
+	[ACPI_ERST_STALL_WHILE_TRUE] = {
+		.flags = APEI_EXEC_INS_ACCESS_REGISTER,
+		.run = erst_exec_stall_while_true,
+	},
+	[ACPI_ERST_SKIP_NEXT_IF_TRUE] = {
+		.flags = APEI_EXEC_INS_ACCESS_REGISTER,
+		.run = erst_exec_skip_next_instruction_if_true,
+	},
+	[ACPI_ERST_GOTO] = {
+		.flags = 0,
+		.run = erst_exec_goto,
+	},
+	[ACPI_ERST_SET_SRC_ADDRESS_BASE] = {
+		.flags = APEI_EXEC_INS_ACCESS_REGISTER,
+		.run = erst_exec_set_src_address_base,
+	},
+	[ACPI_ERST_SET_DST_ADDRESS_BASE] = {
+		.flags = APEI_EXEC_INS_ACCESS_REGISTER,
+		.run = erst_exec_set_dst_address_base,
+	},
+	[ACPI_ERST_MOVE_DATA] = {
+		.flags = APEI_EXEC_INS_ACCESS_REGISTER,
+		.run = erst_exec_move_data,
+	},
+};
+
+static inline void erst_exec_ctx_init(struct apei_exec_context *ctx)
+{
+	apei_exec_ctx_init(ctx, erst_ins_type, ARRAY_SIZE(erst_ins_type),
+			   ERST_TAB_ENTRY(erst_tab), erst_tab->entries);
+}
+
+static int erst_get_erange(struct erst_erange *range)
+{
+	struct apei_exec_context ctx;
+	int rc;
+
+	erst_exec_ctx_init(&ctx);
+	rc = apei_exec_run(&ctx, ACPI_ERST_GET_ERROR_RANGE);
+	if (rc)
+		return rc;
+	range->base = apei_exec_ctx_get_output(&ctx);
+	rc = apei_exec_run(&ctx, ACPI_ERST_GET_ERROR_LENGTH);
+	if (rc)
+		return rc;
+	range->size = apei_exec_ctx_get_output(&ctx);
+	rc = apei_exec_run(&ctx, ACPI_ERST_GET_ERROR_ATTRIBUTES);
+	if (rc)
+		return rc;
+	range->attr = apei_exec_ctx_get_output(&ctx);
+
+	return 0;
+}
+
+static ssize_t __erst_get_record_count(void)
+{
+	struct apei_exec_context ctx;
+	int rc;
+
+	erst_exec_ctx_init(&ctx);
+	rc = apei_exec_run(&ctx, ACPI_ERST_GET_RECORD_COUNT);
+	if (rc)
+		return rc;
+	return apei_exec_ctx_get_output(&ctx);
+}
+
+ssize_t erst_get_record_count(void)
+{
+	ssize_t count;
+	unsigned long flags;
+
+	if (erst_disable)
+		return -ENODEV;
+
+	spin_lock_irqsave(&erst_lock, flags);
+	count = __erst_get_record_count();
+	spin_unlock_irqrestore(&erst_lock, flags);
+
+	return count;
+}
+EXPORT_SYMBOL_GPL(erst_get_record_count);
+
+static int __erst_get_next_record_id(u64 *record_id)
+{
+	struct apei_exec_context ctx;
+	int rc;
+
+	erst_exec_ctx_init(&ctx);
+	rc = apei_exec_run(&ctx, ACPI_ERST_GET_RECORD_ID);
+	if (rc)
+		return rc;
+	*record_id = apei_exec_ctx_get_output(&ctx);
+
+	return 0;
+}
+
+/*
+ * Get the record ID of an existing error record on the persistent
+ * storage. If there is no error record on the persistent storage, the
+ * returned record_id is APEI_ERST_INVALID_RECORD_ID.
+ */
+int erst_get_next_record_id(u64 *record_id)
+{
+	int rc;
+	unsigned long flags;
+
+	if (erst_disable)
+		return -ENODEV;
+
+	spin_lock_irqsave(&erst_lock, flags);
+	rc = __erst_get_next_record_id(record_id);
+	spin_unlock_irqrestore(&erst_lock, flags);
+
+	return rc;
+}
+EXPORT_SYMBOL_GPL(erst_get_next_record_id);
+
+static int __erst_write_to_storage(u64 offset)
+{
+	struct apei_exec_context ctx;
+	u64 timeout = FIRMWARE_TIMEOUT;
+	u64 val;
+	int rc;
+
+	erst_exec_ctx_init(&ctx);
+	rc = apei_exec_run(&ctx, ACPI_ERST_BEGIN_WRITE);
+	if (rc)
+		return rc;
+	apei_exec_ctx_set_input(&ctx, offset);
+	rc = apei_exec_run(&ctx, ACPI_ERST_SET_RECORD_OFFSET);
+	if (rc)
+		return rc;
+	rc = apei_exec_run(&ctx, ACPI_ERST_EXECUTE_OPERATION);
+	if (rc)
+		return rc;
+	for (;;) {
+		rc = apei_exec_run(&ctx, ACPI_ERST_CHECK_BUSY_STATUS);
+		if (rc)
+			return rc;
+		val = apei_exec_ctx_get_output(&ctx);
+		if (!val)
+			break;
+		if (erst_timedout(&timeout, SPIN_UNIT))
+			return -EIO;
+	}
+	rc = apei_exec_run(&ctx, ACPI_ERST_GET_COMMAND_STATUS);
+	if (rc)
+		return rc;
+	val = apei_exec_ctx_get_output(&ctx);
+	rc = apei_exec_run(&ctx, ACPI_ERST_END);
+	if (rc)
+		return rc;
+
+	return erst_errno(val);
+}
+
+static int __erst_read_from_storage(u64 record_id, u64 offset)
+{
+	struct apei_exec_context ctx;
+	u64 timeout = FIRMWARE_TIMEOUT;
+	u64 val;
+	int rc;
+
+	erst_exec_ctx_init(&ctx);
+	rc = apei_exec_run(&ctx, ACPI_ERST_BEGIN_READ);
+	if (rc)
+		return rc;
+	apei_exec_ctx_set_input(&ctx, offset);
+	rc = apei_exec_run(&ctx, ACPI_ERST_SET_RECORD_OFFSET);
+	if (rc)
+		return rc;
+	apei_exec_ctx_set_input(&ctx, record_id);
+	rc = apei_exec_run(&ctx, ACPI_ERST_SET_RECORD_ID);
+	if (rc)
+		return rc;
+	rc = apei_exec_run(&ctx, ACPI_ERST_EXECUTE_OPERATION);
+	if (rc)
+		return rc;
+	for (;;) {
+		rc = apei_exec_run(&ctx, ACPI_ERST_CHECK_BUSY_STATUS);
+		if (rc)
+			return rc;
+		val = apei_exec_ctx_get_output(&ctx);
+		if (!val)
+			break;
+		if (erst_timedout(&timeout, SPIN_UNIT))
+			return -EIO;
+	};
+	rc = apei_exec_run(&ctx, ACPI_ERST_GET_COMMAND_STATUS);
+	if (rc)
+		return rc;
+	val = apei_exec_ctx_get_output(&ctx);
+	rc = apei_exec_run(&ctx, ACPI_ERST_END);
+	if (rc)
+		return rc;
+
+	return erst_errno(val);
+}
+
+static int __erst_clear_from_storage(u64 record_id)
+{
+	struct apei_exec_context ctx;
+	u64 timeout = FIRMWARE_TIMEOUT;
+	u64 val;
+	int rc;
+
+	erst_exec_ctx_init(&ctx);
+	rc = apei_exec_run(&ctx, ACPI_ERST_BEGIN_CLEAR);
+	if (rc)
+		return rc;
+	apei_exec_ctx_set_input(&ctx, record_id);
+	rc = apei_exec_run(&ctx, ACPI_ERST_SET_RECORD_ID);
+	if (rc)
+		return rc;
+	rc = apei_exec_run(&ctx, ACPI_ERST_EXECUTE_OPERATION);
+	if (rc)
+		return rc;
+	for (;;) {
+		rc = apei_exec_run(&ctx, ACPI_ERST_CHECK_BUSY_STATUS);
+		if (rc)
+			return rc;
+		val = apei_exec_ctx_get_output(&ctx);
+		if (!val)
+			break;
+		if (erst_timedout(&timeout, SPIN_UNIT))
+			return -EIO;
+	}
+	rc = apei_exec_run(&ctx, ACPI_ERST_GET_COMMAND_STATUS);
+	if (rc)
+		return rc;
+	val = apei_exec_ctx_get_output(&ctx);
+	rc = apei_exec_run(&ctx, ACPI_ERST_END);
+	if (rc)
+		return rc;
+
+	return erst_errno(val);
+}
+
+/* NVRAM ERST Error Log Address Range is not supported yet */
+static void pr_unimpl_nvram(void)
+{
+	if (printk_ratelimit())
+		pr_warning(ERST_PFX
+		"NVRAM ERST Log Address Range is not implemented yet\n");
+}
+
+static int __erst_write_to_nvram(const struct cper_record_header *record)
+{
+	/* do not print message, because printk is not safe for NMI */
+	return -ENOSYS;
+}
+
+static int __erst_read_to_erange_from_nvram(u64 record_id, u64 *offset)
+{
+	pr_unimpl_nvram();
+	return -ENOSYS;
+}
+
+static int __erst_clear_from_nvram(u64 record_id)
+{
+	pr_unimpl_nvram();
+	return -ENOSYS;
+}
+
+int erst_write(const struct cper_record_header *record)
+{
+	int rc;
+	unsigned long flags;
+	struct cper_record_header *rcd_erange;
+
+	if (erst_disable)
+		return -ENODEV;
+
+	if (memcmp(record->signature, CPER_SIG_RECORD, CPER_SIG_SIZE))
+		return -EINVAL;
+
+	if (erst_erange.attr & ERST_RANGE_NVRAM) {
+		if (!spin_trylock_irqsave(&erst_lock, flags))
+			return -EBUSY;
+		rc = __erst_write_to_nvram(record);
+		spin_unlock_irqrestore(&erst_lock, flags);
+		return rc;
+	}
+
+	if (record->record_length > erst_erange.size)
+		return -EINVAL;
+
+	if (!spin_trylock_irqsave(&erst_lock, flags))
+		return -EBUSY;
+	memcpy(erst_erange.vaddr, record, record->record_length);
+	rcd_erange = erst_erange.vaddr;
+	/* signature for serialization system */
+	memcpy(&rcd_erange->persistence_information, "ER", 2);
+
+	rc = __erst_write_to_storage(0);
+	spin_unlock_irqrestore(&erst_lock, flags);
+
+	return rc;
+}
+EXPORT_SYMBOL_GPL(erst_write);
+
+static int __erst_read_to_erange(u64 record_id, u64 *offset)
+{
+	int rc;
+
+	if (erst_erange.attr & ERST_RANGE_NVRAM)
+		return __erst_read_to_erange_from_nvram(
+			record_id, offset);
+
+	rc = __erst_read_from_storage(record_id, 0);
+	if (rc)
+		return rc;
+	*offset = 0;
+
+	return 0;
+}
+
+static ssize_t __erst_read(u64 record_id, struct cper_record_header *record,
+			   size_t buflen)
+{
+	int rc;
+	u64 offset, len = 0;
+	struct cper_record_header *rcd_tmp;
+
+	rc = __erst_read_to_erange(record_id, &offset);
+	if (rc)
+		return rc;
+	rcd_tmp = erst_erange.vaddr + offset;
+	len = rcd_tmp->record_length;
+	if (len <= buflen)
+		memcpy(record, rcd_tmp, len);
+
+	return len;
+}
+
+/*
+ * If return value > buflen, the buffer size is not big enough,
+ * else if return value < 0, something goes wrong,
+ * else everything is OK, and return value is record length
+ */
+ssize_t erst_read(u64 record_id, struct cper_record_header *record,
+		  size_t buflen)
+{
+	ssize_t len;
+	unsigned long flags;
+
+	if (erst_disable)
+		return -ENODEV;
+
+	spin_lock_irqsave(&erst_lock, flags);
+	len = __erst_read(record_id, record, buflen);
+	spin_unlock_irqrestore(&erst_lock, flags);
+	return len;
+}
+EXPORT_SYMBOL_GPL(erst_read);
+
+/*
+ * If return value > buflen, the buffer size is not big enough,
+ * else if return value = 0, there is no more record to read,
+ * else if return value < 0, something goes wrong,
+ * else everything is OK, and return value is record length
+ */
+ssize_t erst_read_next(struct cper_record_header *record, size_t buflen)
+{
+	int rc;
+	ssize_t len;
+	unsigned long flags;
+	u64 record_id;
+
+	if (erst_disable)
+		return -ENODEV;
+
+	spin_lock_irqsave(&erst_lock, flags);
+	rc = __erst_get_next_record_id(&record_id);
+	if (rc) {
+		spin_unlock_irqrestore(&erst_lock, flags);
+		return rc;
+	}
+	/* no more record */
+	if (record_id == APEI_ERST_INVALID_RECORD_ID) {
+		spin_unlock_irqrestore(&erst_lock, flags);
+		return 0;
+	}
+
+	len = __erst_read(record_id, record, buflen);
+	spin_unlock_irqrestore(&erst_lock, flags);
+
+	return len;
+}
+EXPORT_SYMBOL_GPL(erst_read_next);
+
+int erst_clear(u64 record_id)
+{
+	int rc;
+	unsigned long flags;
+
+	if (erst_disable)
+		return -ENODEV;
+
+	spin_lock_irqsave(&erst_lock, flags);
+	if (erst_erange.attr & ERST_RANGE_NVRAM)
+		rc = __erst_clear_from_nvram(record_id);
+	else
+		rc = __erst_clear_from_storage(record_id);
+	spin_unlock_irqrestore(&erst_lock, flags);
+
+	return rc;
+}
+EXPORT_SYMBOL_GPL(erst_clear);
+
+static int __init setup_erst_disable(char *str)
+{
+	erst_disable = 1;
+	return 0;
+}
+
+__setup("erst_disable", setup_erst_disable);
+
+static int erst_check_table(struct acpi_table_erst *erst_tab)
+{
+	if (erst_tab->header_length != sizeof(struct acpi_table_erst))
+		return -EINVAL;
+	if (erst_tab->header.length < sizeof(struct acpi_table_erst))
+		return -EINVAL;
+	if (erst_tab->entries !=
+	    (erst_tab->header.length - sizeof(struct acpi_table_erst)) /
+	    sizeof(struct acpi_erst_entry))
+		return -EINVAL;
+
+	return 0;
+}
+
+static int __init erst_init(void)
+{
+	int rc = 0;
+	acpi_status status;
+	struct apei_exec_context ctx;
+	struct apei_resources erst_resources;
+	struct resource *r;
+
+	if (acpi_disabled)
+		goto err;
+
+	if (erst_disable) {
+		pr_info(ERST_PFX
+	"Error Record Serialization Table (ERST) support is disabled.\n");
+		goto err;
+	}
+
+	status = acpi_get_table(ACPI_SIG_ERST, 0,
+				(struct acpi_table_header **)&erst_tab);
+	if (status == AE_NOT_FOUND) {
+		pr_err(ERST_PFX "Table is not found!\n");
+		goto err;
+	} else if (ACPI_FAILURE(status)) {
+		const char *msg = acpi_format_exception(status);
+		pr_err(ERST_PFX "Failed to get table, %s\n", msg);
+		rc = -EINVAL;
+		goto err;
+	}
+
+	rc = erst_check_table(erst_tab);
+	if (rc) {
+		pr_err(FW_BUG ERST_PFX "ERST table is invalid\n");
+		goto err;
+	}
+
+	apei_resources_init(&erst_resources);
+	erst_exec_ctx_init(&ctx);
+	rc = apei_exec_collect_resources(&ctx, &erst_resources);
+	if (rc)
+		goto err_fini;
+	rc = apei_resources_request(&erst_resources, "APEI ERST");
+	if (rc)
+		goto err_fini;
+	rc = apei_exec_pre_map_gars(&ctx);
+	if (rc)
+		goto err_release;
+	rc = erst_get_erange(&erst_erange);
+	if (rc) {
+		if (rc == -ENODEV)
+			pr_info(ERST_PFX
+	"The corresponding hardware device or firmware implementation "
+	"is not available.\n");
+		else
+			pr_err(ERST_PFX
+			       "Failed to get Error Log Address Range.\n");
+		goto err_unmap_reg;
+	}
+
+	r = request_mem_region(erst_erange.base, erst_erange.size, "APEI ERST");
+	if (!r) {
+		pr_err(ERST_PFX
+		"Can not request iomem region <0x%16llx-0x%16llx> for ERST.\n",
+		(unsigned long long)erst_erange.base,
+		(unsigned long long)erst_erange.base + erst_erange.size);
+		rc = -EIO;
+		goto err_unmap_reg;
+	}
+	rc = -ENOMEM;
+	erst_erange.vaddr = ioremap_cache(erst_erange.base,
+					  erst_erange.size);
+	if (!erst_erange.vaddr)
+		goto err_release_erange;
+
+	pr_info(ERST_PFX
+	"Error Record Serialization Table (ERST) support is initialized.\n");
+
+	return 0;
+
+err_release_erange:
+	release_mem_region(erst_erange.base, erst_erange.size);
+err_unmap_reg:
+	apei_exec_post_unmap_gars(&ctx);
+err_release:
+	apei_resources_release(&erst_resources);
+err_fini:
+	apei_resources_fini(&erst_resources);
+err:
+	erst_disable = 1;
+	return rc;
+}
+
+device_initcall(erst_init);
diff --git a/drivers/acpi/apei/ghes.c b/drivers/acpi/apei/ghes.c
new file mode 100644
index 000000000000..fd0cc016a099
--- /dev/null
+++ b/drivers/acpi/apei/ghes.c
@@ -0,0 +1,427 @@
+/*
+ * APEI Generic Hardware Error Source support
+ *
+ * Generic Hardware Error Source provides a way to report platform
+ * hardware errors (such as that from chipset). It works in so called
+ * "Firmware First" mode, that is, hardware errors are reported to
+ * firmware firstly, then reported to Linux by firmware. This way,
+ * some non-standard hardware error registers or non-standard hardware
+ * link can be checked by firmware to produce more hardware error
+ * information for Linux.
+ *
+ * For more information about Generic Hardware Error Source, please
+ * refer to ACPI Specification version 4.0, section 17.3.2.6
+ *
+ * Now, only SCI notification type and memory errors are
+ * supported. More notification type and hardware error type will be
+ * added later.
+ *
+ * Copyright 2010 Intel Corp.
+ *   Author: Huang Ying <ying.huang@intel.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License version
+ * 2 as published by the Free Software Foundation;
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/acpi.h>
+#include <linux/io.h>
+#include <linux/interrupt.h>
+#include <linux/cper.h>
+#include <linux/kdebug.h>
+#include <acpi/apei.h>
+#include <acpi/atomicio.h>
+#include <acpi/hed.h>
+#include <asm/mce.h>
+
+#include "apei-internal.h"
+
+#define GHES_PFX	"GHES: "
+
+#define GHES_ESTATUS_MAX_SIZE		65536
+
+/*
+ * One struct ghes is created for each generic hardware error
+ * source.
+ *
+ * It provides the context for APEI hardware error timer/IRQ/SCI/NMI
+ * handler. Handler for one generic hardware error source is only
+ * triggered after the previous one is done. So handler can uses
+ * struct ghes without locking.
+ *
+ * estatus: memory buffer for error status block, allocated during
+ * HEST parsing.
+ */
+#define GHES_TO_CLEAR		0x0001
+
+struct ghes {
+	struct acpi_hest_generic *generic;
+	struct acpi_hest_generic_status *estatus;
+	struct list_head list;
+	u64 buffer_paddr;
+	unsigned long flags;
+};
+
+/*
+ * Error source lists, one list for each notification method. The
+ * members in lists are struct ghes.
+ *
+ * The list members are only added in HEST parsing and deleted during
+ * module_exit, that is, single-threaded. So no lock is needed for
+ * that.
+ *
+ * But the mutual exclusion is needed between members adding/deleting
+ * and timer/IRQ/SCI/NMI handler, which may traverse the list. RCU is
+ * used for that.
+ */
+static LIST_HEAD(ghes_sci);
+
+static struct ghes *ghes_new(struct acpi_hest_generic *generic)
+{
+	struct ghes *ghes;
+	unsigned int error_block_length;
+	int rc;
+
+	ghes = kzalloc(sizeof(*ghes), GFP_KERNEL);
+	if (!ghes)
+		return ERR_PTR(-ENOMEM);
+	ghes->generic = generic;
+	INIT_LIST_HEAD(&ghes->list);
+	rc = acpi_pre_map_gar(&generic->error_status_address);
+	if (rc)
+		goto err_free;
+	error_block_length = generic->error_block_length;
+	if (error_block_length > GHES_ESTATUS_MAX_SIZE) {
+		pr_warning(FW_WARN GHES_PFX
+			   "Error status block length is too long: %u for "
+			   "generic hardware error source: %d.\n",
+			   error_block_length, generic->header.source_id);
+		error_block_length = GHES_ESTATUS_MAX_SIZE;
+	}
+	ghes->estatus = kmalloc(error_block_length, GFP_KERNEL);
+	if (!ghes->estatus) {
+		rc = -ENOMEM;
+		goto err_unmap;
+	}
+
+	return ghes;
+
+err_unmap:
+	acpi_post_unmap_gar(&generic->error_status_address);
+err_free:
+	kfree(ghes);
+	return ERR_PTR(rc);
+}
+
+static void ghes_fini(struct ghes *ghes)
+{
+	kfree(ghes->estatus);
+	acpi_post_unmap_gar(&ghes->generic->error_status_address);
+}
+
+enum {
+	GHES_SER_NO = 0x0,
+	GHES_SER_CORRECTED = 0x1,
+	GHES_SER_RECOVERABLE = 0x2,
+	GHES_SER_PANIC = 0x3,
+};
+
+static inline int ghes_severity(int severity)
+{
+	switch (severity) {
+	case CPER_SER_INFORMATIONAL:
+		return GHES_SER_NO;
+	case CPER_SER_CORRECTED:
+		return GHES_SER_CORRECTED;
+	case CPER_SER_RECOVERABLE:
+		return GHES_SER_RECOVERABLE;
+	case CPER_SER_FATAL:
+		return GHES_SER_PANIC;
+	default:
+		/* Unkown, go panic */
+		return GHES_SER_PANIC;
+	}
+}
+
+/* SCI handler run in work queue, so ioremap can be used here */
+static int ghes_copy_tofrom_phys(void *buffer, u64 paddr, u32 len,
+				 int from_phys)
+{
+	void *vaddr;
+
+	vaddr = ioremap_cache(paddr, len);
+	if (!vaddr)
+		return -ENOMEM;
+	if (from_phys)
+		memcpy(buffer, vaddr, len);
+	else
+		memcpy(vaddr, buffer, len);
+	iounmap(vaddr);
+
+	return 0;
+}
+
+static int ghes_read_estatus(struct ghes *ghes, int silent)
+{
+	struct acpi_hest_generic *g = ghes->generic;
+	u64 buf_paddr;
+	u32 len;
+	int rc;
+
+	rc = acpi_atomic_read(&buf_paddr, &g->error_status_address);
+	if (rc) {
+		if (!silent && printk_ratelimit())
+			pr_warning(FW_WARN GHES_PFX
+"Failed to read error status block address for hardware error source: %d.\n",
+				   g->header.source_id);
+		return -EIO;
+	}
+	if (!buf_paddr)
+		return -ENOENT;
+
+	rc = ghes_copy_tofrom_phys(ghes->estatus, buf_paddr,
+				   sizeof(*ghes->estatus), 1);
+	if (rc)
+		return rc;
+	if (!ghes->estatus->block_status)
+		return -ENOENT;
+
+	ghes->buffer_paddr = buf_paddr;
+	ghes->flags |= GHES_TO_CLEAR;
+
+	rc = -EIO;
+	len = apei_estatus_len(ghes->estatus);
+	if (len < sizeof(*ghes->estatus))
+		goto err_read_block;
+	if (len > ghes->generic->error_block_length)
+		goto err_read_block;
+	if (apei_estatus_check_header(ghes->estatus))
+		goto err_read_block;
+	rc = ghes_copy_tofrom_phys(ghes->estatus + 1,
+				   buf_paddr + sizeof(*ghes->estatus),
+				   len - sizeof(*ghes->estatus), 1);
+	if (rc)
+		return rc;
+	if (apei_estatus_check(ghes->estatus))
+		goto err_read_block;
+	rc = 0;
+
+err_read_block:
+	if (rc && !silent)
+		pr_warning(FW_WARN GHES_PFX
+			   "Failed to read error status block!\n");
+	return rc;
+}
+
+static void ghes_clear_estatus(struct ghes *ghes)
+{
+	ghes->estatus->block_status = 0;
+	if (!(ghes->flags & GHES_TO_CLEAR))
+		return;
+	ghes_copy_tofrom_phys(ghes->estatus, ghes->buffer_paddr,
+			      sizeof(ghes->estatus->block_status), 0);
+	ghes->flags &= ~GHES_TO_CLEAR;
+}
+
+static void ghes_do_proc(struct ghes *ghes)
+{
+	int ser, processed = 0;
+	struct acpi_hest_generic_data *gdata;
+
+	ser = ghes_severity(ghes->estatus->error_severity);
+	apei_estatus_for_each_section(ghes->estatus, gdata) {
+#ifdef CONFIG_X86_MCE
+		if (!uuid_le_cmp(*(uuid_le *)gdata->section_type,
+				 CPER_SEC_PLATFORM_MEM)) {
+			apei_mce_report_mem_error(
+				ser == GHES_SER_CORRECTED,
+				(struct cper_sec_mem_err *)(gdata+1));
+			processed = 1;
+		}
+#endif
+	}
+
+	if (!processed && printk_ratelimit())
+		pr_warning(GHES_PFX
+		"Unknown error record from generic hardware error source: %d\n",
+			   ghes->generic->header.source_id);
+}
+
+static int ghes_proc(struct ghes *ghes)
+{
+	int rc;
+
+	rc = ghes_read_estatus(ghes, 0);
+	if (rc)
+		goto out;
+	ghes_do_proc(ghes);
+
+out:
+	ghes_clear_estatus(ghes);
+	return 0;
+}
+
+static int ghes_notify_sci(struct notifier_block *this,
+				  unsigned long event, void *data)
+{
+	struct ghes *ghes;
+	int ret = NOTIFY_DONE;
+
+	rcu_read_lock();
+	list_for_each_entry_rcu(ghes, &ghes_sci, list) {
+		if (!ghes_proc(ghes))
+			ret = NOTIFY_OK;
+	}
+	rcu_read_unlock();
+
+	return ret;
+}
+
+static struct notifier_block ghes_notifier_sci = {
+	.notifier_call = ghes_notify_sci,
+};
+
+static int hest_ghes_parse(struct acpi_hest_header *hest_hdr, void *data)
+{
+	struct acpi_hest_generic *generic;
+	struct ghes *ghes = NULL;
+	int rc = 0;
+
+	if (hest_hdr->type != ACPI_HEST_TYPE_GENERIC_ERROR)
+		return 0;
+
+	generic = (struct acpi_hest_generic *)hest_hdr;
+	if (!generic->enabled)
+		return 0;
+
+	if (generic->error_block_length <
+	    sizeof(struct acpi_hest_generic_status)) {
+		pr_warning(FW_BUG GHES_PFX
+"Invalid error block length: %u for generic hardware error source: %d\n",
+			   generic->error_block_length,
+			   generic->header.source_id);
+		goto err;
+	}
+	if (generic->records_to_preallocate == 0) {
+		pr_warning(FW_BUG GHES_PFX
+"Invalid records to preallocate: %u for generic hardware error source: %d\n",
+			   generic->records_to_preallocate,
+			   generic->header.source_id);
+		goto err;
+	}
+	ghes = ghes_new(generic);
+	if (IS_ERR(ghes)) {
+		rc = PTR_ERR(ghes);
+		ghes = NULL;
+		goto err;
+	}
+	switch (generic->notify.type) {
+	case ACPI_HEST_NOTIFY_POLLED:
+		pr_warning(GHES_PFX
+"Generic hardware error source: %d notified via POLL is not supported!\n",
+			   generic->header.source_id);
+		break;
+	case ACPI_HEST_NOTIFY_EXTERNAL:
+	case ACPI_HEST_NOTIFY_LOCAL:
+		pr_warning(GHES_PFX
+"Generic hardware error source: %d notified via IRQ is not supported!\n",
+			   generic->header.source_id);
+		break;
+	case ACPI_HEST_NOTIFY_SCI:
+		if (list_empty(&ghes_sci))
+			register_acpi_hed_notifier(&ghes_notifier_sci);
+		list_add_rcu(&ghes->list, &ghes_sci);
+		break;
+	case ACPI_HEST_NOTIFY_NMI:
+		pr_warning(GHES_PFX
+"Generic hardware error source: %d notified via NMI is not supported!\n",
+			   generic->header.source_id);
+		break;
+	default:
+		pr_warning(FW_WARN GHES_PFX
+	"Unknown notification type: %u for generic hardware error source: %d\n",
+			   generic->notify.type, generic->header.source_id);
+		break;
+	}
+
+	return 0;
+err:
+	if (ghes)
+		ghes_fini(ghes);
+	return rc;
+}
+
+static void ghes_cleanup(void)
+{
+	struct ghes *ghes, *nghes;
+
+	if (!list_empty(&ghes_sci))
+		unregister_acpi_hed_notifier(&ghes_notifier_sci);
+
+	synchronize_rcu();
+
+	list_for_each_entry_safe(ghes, nghes, &ghes_sci, list) {
+		list_del(&ghes->list);
+		ghes_fini(ghes);
+		kfree(ghes);
+	}
+}
+
+static int __init ghes_init(void)
+{
+	int rc;
+
+	if (acpi_disabled)
+		return -ENODEV;
+
+	if (hest_disable) {
+		pr_info(GHES_PFX "HEST is not enabled!\n");
+		return -EINVAL;
+	}
+
+	rc = apei_hest_parse(hest_ghes_parse, NULL);
+	if (rc) {
+		pr_err(GHES_PFX
+		"Error during parsing HEST generic hardware error sources.\n");
+		goto err_cleanup;
+	}
+
+	if (list_empty(&ghes_sci)) {
+		pr_info(GHES_PFX
+			"No functional generic hardware error sources.\n");
+		rc = -ENODEV;
+		goto err_cleanup;
+	}
+
+	pr_info(GHES_PFX
+		"Generic Hardware Error Source support is initialized.\n");
+
+	return 0;
+err_cleanup:
+	ghes_cleanup();
+	return rc;
+}
+
+static void __exit ghes_exit(void)
+{
+	ghes_cleanup();
+}
+
+module_init(ghes_init);
+module_exit(ghes_exit);
+
+MODULE_AUTHOR("Huang Ying");
+MODULE_DESCRIPTION("APEI Generic Hardware Error Source support");
+MODULE_LICENSE("GPL");
diff --git a/drivers/acpi/apei/hest.c b/drivers/acpi/apei/hest.c
new file mode 100644
index 000000000000..e7f40d362cb3
--- /dev/null
+++ b/drivers/acpi/apei/hest.c
@@ -0,0 +1,173 @@
+/*
+ * APEI Hardware Error Souce Table support
+ *
+ * HEST describes error sources in detail; communicates operational
+ * parameters (i.e. severity levels, masking bits, and threshold
+ * values) to Linux as necessary. It also allows the BIOS to report
+ * non-standard error sources to Linux (for example, chipset-specific
+ * error registers).
+ *
+ * For more information about HEST, please refer to ACPI Specification
+ * version 4.0, section 17.3.2.
+ *
+ * Copyright 2009 Intel Corp.
+ *   Author: Huang Ying <ying.huang@intel.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License version
+ * 2 as published by the Free Software Foundation;
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/acpi.h>
+#include <linux/kdebug.h>
+#include <linux/highmem.h>
+#include <linux/io.h>
+#include <acpi/apei.h>
+
+#include "apei-internal.h"
+
+#define HEST_PFX "HEST: "
+
+int hest_disable;
+EXPORT_SYMBOL_GPL(hest_disable);
+
+/* HEST table parsing */
+
+static struct acpi_table_hest *hest_tab;
+
+static int hest_void_parse(struct acpi_hest_header *hest_hdr, void *data)
+{
+	return 0;
+}
+
+static int hest_esrc_len_tab[ACPI_HEST_TYPE_RESERVED] = {
+	[ACPI_HEST_TYPE_IA32_CHECK] = -1,	/* need further calculation */
+	[ACPI_HEST_TYPE_IA32_CORRECTED_CHECK] = -1,
+	[ACPI_HEST_TYPE_IA32_NMI] = sizeof(struct acpi_hest_ia_nmi),
+	[ACPI_HEST_TYPE_AER_ROOT_PORT] = sizeof(struct acpi_hest_aer_root),
+	[ACPI_HEST_TYPE_AER_ENDPOINT] = sizeof(struct acpi_hest_aer),
+	[ACPI_HEST_TYPE_AER_BRIDGE] = sizeof(struct acpi_hest_aer_bridge),
+	[ACPI_HEST_TYPE_GENERIC_ERROR] = sizeof(struct acpi_hest_generic),
+};
+
+static int hest_esrc_len(struct acpi_hest_header *hest_hdr)
+{
+	u16 hest_type = hest_hdr->type;
+	int len;
+
+	if (hest_type >= ACPI_HEST_TYPE_RESERVED)
+		return 0;
+
+	len = hest_esrc_len_tab[hest_type];
+
+	if (hest_type == ACPI_HEST_TYPE_IA32_CORRECTED_CHECK) {
+		struct acpi_hest_ia_corrected *cmc;
+		cmc = (struct acpi_hest_ia_corrected *)hest_hdr;
+		len = sizeof(*cmc) + cmc->num_hardware_banks *
+			sizeof(struct acpi_hest_ia_error_bank);
+	} else if (hest_type == ACPI_HEST_TYPE_IA32_CHECK) {
+		struct acpi_hest_ia_machine_check *mc;
+		mc = (struct acpi_hest_ia_machine_check *)hest_hdr;
+		len = sizeof(*mc) + mc->num_hardware_banks *
+			sizeof(struct acpi_hest_ia_error_bank);
+	}
+	BUG_ON(len == -1);
+
+	return len;
+};
+
+int apei_hest_parse(apei_hest_func_t func, void *data)
+{
+	struct acpi_hest_header *hest_hdr;
+	int i, rc, len;
+
+	if (hest_disable)
+		return -EINVAL;
+
+	hest_hdr = (struct acpi_hest_header *)(hest_tab + 1);
+	for (i = 0; i < hest_tab->error_source_count; i++) {
+		len = hest_esrc_len(hest_hdr);
+		if (!len) {
+			pr_warning(FW_WARN HEST_PFX
+				   "Unknown or unused hardware error source "
+				   "type: %d for hardware error source: %d.\n",
+				   hest_hdr->type, hest_hdr->source_id);
+			return -EINVAL;
+		}
+		if ((void *)hest_hdr + len >
+		    (void *)hest_tab + hest_tab->header.length) {
+			pr_warning(FW_BUG HEST_PFX
+		"Table contents overflow for hardware error source: %d.\n",
+				hest_hdr->source_id);
+			return -EINVAL;
+		}
+
+		rc = func(hest_hdr, data);
+		if (rc)
+			return rc;
+
+		hest_hdr = (void *)hest_hdr + len;
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(apei_hest_parse);
+
+static int __init setup_hest_disable(char *str)
+{
+	hest_disable = 1;
+	return 0;
+}
+
+__setup("hest_disable", setup_hest_disable);
+
+static int __init hest_init(void)
+{
+	acpi_status status;
+	int rc = -ENODEV;
+
+	if (acpi_disabled)
+		goto err;
+
+	if (hest_disable) {
+		pr_info(HEST_PFX "HEST tabling parsing is disabled.\n");
+		goto err;
+	}
+
+	status = acpi_get_table(ACPI_SIG_HEST, 0,
+				(struct acpi_table_header **)&hest_tab);
+	if (status == AE_NOT_FOUND) {
+		pr_info(HEST_PFX "Table is not found!\n");
+		goto err;
+	} else if (ACPI_FAILURE(status)) {
+		const char *msg = acpi_format_exception(status);
+		pr_err(HEST_PFX "Failed to get table, %s\n", msg);
+		rc = -EINVAL;
+		goto err;
+	}
+
+	rc = apei_hest_parse(hest_void_parse, NULL);
+	if (rc)
+		goto err;
+
+	pr_info(HEST_PFX "HEST table parsing is initialized.\n");
+
+	return 0;
+err:
+	hest_disable = 1;
+	return rc;
+}
+
+subsys_initcall(hest_init);
diff --git a/drivers/acpi/atomicio.c b/drivers/acpi/atomicio.c
new file mode 100644
index 000000000000..814b19249616
--- /dev/null
+++ b/drivers/acpi/atomicio.c
@@ -0,0 +1,360 @@
+/*
+ * atomicio.c - ACPI IO memory pre-mapping/post-unmapping, then
+ * accessing in atomic context.
+ *
+ * This is used for NMI handler to access IO memory area, because
+ * ioremap/iounmap can not be used in NMI handler. The IO memory area
+ * is pre-mapped in process context and accessed in NMI handler.
+ *
+ * Copyright (C) 2009-2010, Intel Corp.
+ *	Author: Huang Ying <ying.huang@intel.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License version
+ * 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/acpi.h>
+#include <linux/io.h>
+#include <linux/kref.h>
+#include <linux/rculist.h>
+#include <linux/interrupt.h>
+#include <acpi/atomicio.h>
+
+#define ACPI_PFX "ACPI: "
+
+static LIST_HEAD(acpi_iomaps);
+/*
+ * Used for mutual exclusion between writers of acpi_iomaps list, for
+ * synchronization between readers and writer, RCU is used.
+ */
+static DEFINE_SPINLOCK(acpi_iomaps_lock);
+
+struct acpi_iomap {
+	struct list_head list;
+	void __iomem *vaddr;
+	unsigned long size;
+	phys_addr_t paddr;
+	struct kref ref;
+};
+
+/* acpi_iomaps_lock or RCU read lock must be held before calling */
+static struct acpi_iomap *__acpi_find_iomap(phys_addr_t paddr,
+					    unsigned long size)
+{
+	struct acpi_iomap *map;
+
+	list_for_each_entry_rcu(map, &acpi_iomaps, list) {
+		if (map->paddr + map->size >= paddr + size &&
+		    map->paddr <= paddr)
+			return map;
+	}
+	return NULL;
+}
+
+/*
+ * Atomic "ioremap" used by NMI handler, if the specified IO memory
+ * area is not pre-mapped, NULL will be returned.
+ *
+ * acpi_iomaps_lock or RCU read lock must be held before calling
+ */
+static void __iomem *__acpi_ioremap_fast(phys_addr_t paddr,
+					 unsigned long size)
+{
+	struct acpi_iomap *map;
+
+	map = __acpi_find_iomap(paddr, size);
+	if (map)
+		return map->vaddr + (paddr - map->paddr);
+	else
+		return NULL;
+}
+
+/* acpi_iomaps_lock must be held before calling */
+static void __iomem *__acpi_try_ioremap(phys_addr_t paddr,
+					unsigned long size)
+{
+	struct acpi_iomap *map;
+
+	map = __acpi_find_iomap(paddr, size);
+	if (map) {
+		kref_get(&map->ref);
+		return map->vaddr + (paddr - map->paddr);
+	} else
+		return NULL;
+}
+
+/*
+ * Used to pre-map the specified IO memory area. First try to find
+ * whether the area is already pre-mapped, if it is, increase the
+ * reference count (in __acpi_try_ioremap) and return; otherwise, do
+ * the real ioremap, and add the mapping into acpi_iomaps list.
+ */
+static void __iomem *acpi_pre_map(phys_addr_t paddr,
+				  unsigned long size)
+{
+	void __iomem *vaddr;
+	struct acpi_iomap *map;
+	unsigned long pg_sz, flags;
+	phys_addr_t pg_off;
+
+	spin_lock_irqsave(&acpi_iomaps_lock, flags);
+	vaddr = __acpi_try_ioremap(paddr, size);
+	spin_unlock_irqrestore(&acpi_iomaps_lock, flags);
+	if (vaddr)
+		return vaddr;
+
+	pg_off = paddr & PAGE_MASK;
+	pg_sz = ((paddr + size + PAGE_SIZE - 1) & PAGE_MASK) - pg_off;
+	vaddr = ioremap(pg_off, pg_sz);
+	if (!vaddr)
+		return NULL;
+	map = kmalloc(sizeof(*map), GFP_KERNEL);
+	if (!map)
+		goto err_unmap;
+	INIT_LIST_HEAD(&map->list);
+	map->paddr = pg_off;
+	map->size = pg_sz;
+	map->vaddr = vaddr;
+	kref_init(&map->ref);
+
+	spin_lock_irqsave(&acpi_iomaps_lock, flags);
+	vaddr = __acpi_try_ioremap(paddr, size);
+	if (vaddr) {
+		spin_unlock_irqrestore(&acpi_iomaps_lock, flags);
+		iounmap(map->vaddr);
+		kfree(map);
+		return vaddr;
+	}
+	list_add_tail_rcu(&map->list, &acpi_iomaps);
+	spin_unlock_irqrestore(&acpi_iomaps_lock, flags);
+
+	return vaddr + (paddr - pg_off);
+err_unmap:
+	iounmap(vaddr);
+	return NULL;
+}
+
+/* acpi_iomaps_lock must be held before calling */
+static void __acpi_kref_del_iomap(struct kref *ref)
+{
+	struct acpi_iomap *map;
+
+	map = container_of(ref, struct acpi_iomap, ref);
+	list_del_rcu(&map->list);
+}
+
+/*
+ * Used to post-unmap the specified IO memory area. The iounmap is
+ * done only if the reference count goes zero.
+ */
+static void acpi_post_unmap(phys_addr_t paddr, unsigned long size)
+{
+	struct acpi_iomap *map;
+	unsigned long flags;
+	int del;
+
+	spin_lock_irqsave(&acpi_iomaps_lock, flags);
+	map = __acpi_find_iomap(paddr, size);
+	BUG_ON(!map);
+	del = kref_put(&map->ref, __acpi_kref_del_iomap);
+	spin_unlock_irqrestore(&acpi_iomaps_lock, flags);
+
+	if (!del)
+		return;
+
+	synchronize_rcu();
+	iounmap(map->vaddr);
+	kfree(map);
+}
+
+/* In NMI handler, should set silent = 1 */
+static int acpi_check_gar(struct acpi_generic_address *reg,
+			  u64 *paddr, int silent)
+{
+	u32 width, space_id;
+
+	width = reg->bit_width;
+	space_id = reg->space_id;
+	/* Handle possible alignment issues */
+	memcpy(paddr, &reg->address, sizeof(*paddr));
+	if (!*paddr) {
+		if (!silent)
+			pr_warning(FW_BUG ACPI_PFX
+			"Invalid physical address in GAR [0x%llx/%u/%u]\n",
+				   *paddr, width, space_id);
+		return -EINVAL;
+	}
+
+	if ((width != 8) && (width != 16) && (width != 32) && (width != 64)) {
+		if (!silent)
+			pr_warning(FW_BUG ACPI_PFX
+				   "Invalid bit width in GAR [0x%llx/%u/%u]\n",
+				   *paddr, width, space_id);
+		return -EINVAL;
+	}
+
+	if (space_id != ACPI_ADR_SPACE_SYSTEM_MEMORY &&
+	    space_id != ACPI_ADR_SPACE_SYSTEM_IO) {
+		if (!silent)
+			pr_warning(FW_BUG ACPI_PFX
+			"Invalid address space type in GAR [0x%llx/%u/%u]\n",
+				   *paddr, width, space_id);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+/* Pre-map, working on GAR */
+int acpi_pre_map_gar(struct acpi_generic_address *reg)
+{
+	u64 paddr;
+	void __iomem *vaddr;
+	int rc;
+
+	if (reg->space_id != ACPI_ADR_SPACE_SYSTEM_MEMORY)
+		return 0;
+
+	rc = acpi_check_gar(reg, &paddr, 0);
+	if (rc)
+		return rc;
+
+	vaddr = acpi_pre_map(paddr, reg->bit_width / 8);
+	if (!vaddr)
+		return -EIO;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(acpi_pre_map_gar);
+
+/* Post-unmap, working on GAR */
+int acpi_post_unmap_gar(struct acpi_generic_address *reg)
+{
+	u64 paddr;
+	int rc;
+
+	if (reg->space_id != ACPI_ADR_SPACE_SYSTEM_MEMORY)
+		return 0;
+
+	rc = acpi_check_gar(reg, &paddr, 0);
+	if (rc)
+		return rc;
+
+	acpi_post_unmap(paddr, reg->bit_width / 8);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(acpi_post_unmap_gar);
+
+/*
+ * Can be used in atomic (including NMI) or process context. RCU read
+ * lock can only be released after the IO memory area accessing.
+ */
+static int acpi_atomic_read_mem(u64 paddr, u64 *val, u32 width)
+{
+	void __iomem *addr;
+
+	rcu_read_lock();
+	addr = __acpi_ioremap_fast(paddr, width);
+	switch (width) {
+	case 8:
+		*val = readb(addr);
+		break;
+	case 16:
+		*val = readw(addr);
+		break;
+	case 32:
+		*val = readl(addr);
+		break;
+	case 64:
+		*val = readq(addr);
+		break;
+	default:
+		return -EINVAL;
+	}
+	rcu_read_unlock();
+
+	return 0;
+}
+
+static int acpi_atomic_write_mem(u64 paddr, u64 val, u32 width)
+{
+	void __iomem *addr;
+
+	rcu_read_lock();
+	addr = __acpi_ioremap_fast(paddr, width);
+	switch (width) {
+	case 8:
+		writeb(val, addr);
+		break;
+	case 16:
+		writew(val, addr);
+		break;
+	case 32:
+		writel(val, addr);
+		break;
+	case 64:
+		writeq(val, addr);
+		break;
+	default:
+		return -EINVAL;
+	}
+	rcu_read_unlock();
+
+	return 0;
+}
+
+/* GAR accessing in atomic (including NMI) or process context */
+int acpi_atomic_read(u64 *val, struct acpi_generic_address *reg)
+{
+	u64 paddr;
+	int rc;
+
+	rc = acpi_check_gar(reg, &paddr, 1);
+	if (rc)
+		return rc;
+
+	*val = 0;
+	switch (reg->space_id) {
+	case ACPI_ADR_SPACE_SYSTEM_MEMORY:
+		return acpi_atomic_read_mem(paddr, val, reg->bit_width);
+	case ACPI_ADR_SPACE_SYSTEM_IO:
+		return acpi_os_read_port(paddr, (u32 *)val, reg->bit_width);
+	default:
+		return -EINVAL;
+	}
+}
+EXPORT_SYMBOL_GPL(acpi_atomic_read);
+
+int acpi_atomic_write(u64 val, struct acpi_generic_address *reg)
+{
+	u64 paddr;
+	int rc;
+
+	rc = acpi_check_gar(reg, &paddr, 1);
+	if (rc)
+		return rc;
+
+	switch (reg->space_id) {
+	case ACPI_ADR_SPACE_SYSTEM_MEMORY:
+		return acpi_atomic_write_mem(paddr, val, reg->bit_width);
+	case ACPI_ADR_SPACE_SYSTEM_IO:
+		return acpi_os_write_port(paddr, val, reg->bit_width);
+	default:
+		return -EINVAL;
+	}
+}
+EXPORT_SYMBOL_GPL(acpi_atomic_write);
diff --git a/drivers/acpi/bus.c b/drivers/acpi/bus.c
index 9042a8579668..c1d23cd71652 100644
--- a/drivers/acpi/bus.c
+++ b/drivers/acpi/bus.c
@@ -401,11 +401,6 @@ static void acpi_print_osc_error(acpi_handle handle,
 	printk("\n");
 }
 
-static u8 hex_val(unsigned char c)
-{
-	return isdigit(c) ? c - '0' : toupper(c) - 'A' + 10;
-}
-
 static acpi_status acpi_str_to_uuid(char *str, u8 *uuid)
 {
 	int i;
@@ -422,8 +417,8 @@ static acpi_status acpi_str_to_uuid(char *str, u8 *uuid)
 			return AE_BAD_PARAMETER;
 	}
 	for (i = 0; i < 16; i++) {
-		uuid[i] = hex_val(str[opc_map_to_uuid[i]]) << 4;
-		uuid[i] |= hex_val(str[opc_map_to_uuid[i] + 1]);
+		uuid[i] = hex_to_bin(str[opc_map_to_uuid[i]]) << 4;
+		uuid[i] |= hex_to_bin(str[opc_map_to_uuid[i] + 1]);
 	}
 	return AE_OK;
 }
diff --git a/drivers/acpi/ec.c b/drivers/acpi/ec.c
index f2234db85da0..5f2027d782e8 100644
--- a/drivers/acpi/ec.c
+++ b/drivers/acpi/ec.c
@@ -79,7 +79,7 @@ enum {
 	EC_FLAGS_GPE_STORM,		/* GPE storm detected */
 	EC_FLAGS_HANDLERS_INSTALLED,	/* Handlers for GPE and
 					 * OpReg are installed */
-	EC_FLAGS_FROZEN,		/* Transactions are suspended */
+	EC_FLAGS_BLOCKED,		/* Transactions are blocked */
 };
 
 /* If we find an EC via the ECDT, we need to keep a ptr to its context */
@@ -293,7 +293,7 @@ static int acpi_ec_transaction(struct acpi_ec *ec, struct transaction *t)
 	if (t->rdata)
 		memset(t->rdata, 0, t->rlen);
 	mutex_lock(&ec->lock);
-	if (test_bit(EC_FLAGS_FROZEN, &ec->flags)) {
+	if (test_bit(EC_FLAGS_BLOCKED, &ec->flags)) {
 		status = -EINVAL;
 		goto unlock;
 	}
@@ -459,7 +459,7 @@ int ec_transaction(u8 command,
 
 EXPORT_SYMBOL(ec_transaction);
 
-void acpi_ec_suspend_transactions(void)
+void acpi_ec_block_transactions(void)
 {
 	struct acpi_ec *ec = first_ec;
 
@@ -468,11 +468,11 @@ void acpi_ec_suspend_transactions(void)
 
 	mutex_lock(&ec->lock);
 	/* Prevent transactions from being carried out */
-	set_bit(EC_FLAGS_FROZEN, &ec->flags);
+	set_bit(EC_FLAGS_BLOCKED, &ec->flags);
 	mutex_unlock(&ec->lock);
 }
 
-void acpi_ec_resume_transactions(void)
+void acpi_ec_unblock_transactions(void)
 {
 	struct acpi_ec *ec = first_ec;
 
@@ -481,10 +481,20 @@ void acpi_ec_resume_transactions(void)
 
 	mutex_lock(&ec->lock);
 	/* Allow transactions to be carried out again */
-	clear_bit(EC_FLAGS_FROZEN, &ec->flags);
+	clear_bit(EC_FLAGS_BLOCKED, &ec->flags);
 	mutex_unlock(&ec->lock);
 }
 
+void acpi_ec_unblock_transactions_early(void)
+{
+	/*
+	 * Allow transactions to happen again (this function is called from
+	 * atomic context during wakeup, so we don't need to acquire the mutex).
+	 */
+	if (first_ec)
+		clear_bit(EC_FLAGS_BLOCKED, &first_ec->flags);
+}
+
 static int acpi_ec_query_unlocked(struct acpi_ec *ec, u8 * data)
 {
 	int result;
@@ -1027,10 +1037,9 @@ int __init acpi_ec_ecdt_probe(void)
 		/* Don't trust ECDT, which comes from ASUSTek */
 		if (!EC_FLAGS_VALIDATE_ECDT)
 			goto install;
-		saved_ec = kmalloc(sizeof(struct acpi_ec), GFP_KERNEL);
+		saved_ec = kmemdup(boot_ec, sizeof(struct acpi_ec), GFP_KERNEL);
 		if (!saved_ec)
 			return -ENOMEM;
-		memcpy(saved_ec, boot_ec, sizeof(struct acpi_ec));
 	/* fall through */
 	}
 
diff --git a/drivers/acpi/hed.c b/drivers/acpi/hed.c
new file mode 100644
index 000000000000..d0c1967f7597
--- /dev/null
+++ b/drivers/acpi/hed.c
@@ -0,0 +1,112 @@
+/*
+ * ACPI Hardware Error Device (PNP0C33) Driver
+ *
+ * Copyright (C) 2010, Intel Corp.
+ *	Author: Huang Ying <ying.huang@intel.com>
+ *
+ * ACPI Hardware Error Device is used to report some hardware errors
+ * notified via SCI, mainly the corrected errors.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License version
+ * 2 as published by the Free Software Foundation;
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/acpi.h>
+#include <acpi/acpi_bus.h>
+#include <acpi/acpi_drivers.h>
+#include <acpi/hed.h>
+
+static struct acpi_device_id acpi_hed_ids[] = {
+	{"PNP0C33", 0},
+	{"", 0},
+};
+MODULE_DEVICE_TABLE(acpi, acpi_hed_ids);
+
+static acpi_handle hed_handle;
+
+static BLOCKING_NOTIFIER_HEAD(acpi_hed_notify_list);
+
+int register_acpi_hed_notifier(struct notifier_block *nb)
+{
+	return blocking_notifier_chain_register(&acpi_hed_notify_list, nb);
+}
+EXPORT_SYMBOL_GPL(register_acpi_hed_notifier);
+
+void unregister_acpi_hed_notifier(struct notifier_block *nb)
+{
+	blocking_notifier_chain_unregister(&acpi_hed_notify_list, nb);
+}
+EXPORT_SYMBOL_GPL(unregister_acpi_hed_notifier);
+
+/*
+ * SCI to report hardware error is forwarded to the listeners of HED,
+ * it is used by HEST Generic Hardware Error Source with notify type
+ * SCI.
+ */
+static void acpi_hed_notify(struct acpi_device *device, u32 event)
+{
+	blocking_notifier_call_chain(&acpi_hed_notify_list, 0, NULL);
+}
+
+static int __devinit acpi_hed_add(struct acpi_device *device)
+{
+	/* Only one hardware error device */
+	if (hed_handle)
+		return -EINVAL;
+	hed_handle = device->handle;
+	return 0;
+}
+
+static int __devexit acpi_hed_remove(struct acpi_device *device, int type)
+{
+	hed_handle = NULL;
+	return 0;
+}
+
+static struct acpi_driver acpi_hed_driver = {
+	.name = "hardware_error_device",
+	.class = "hardware_error",
+	.ids = acpi_hed_ids,
+	.ops = {
+		.add = acpi_hed_add,
+		.remove = acpi_hed_remove,
+		.notify = acpi_hed_notify,
+	},
+};
+
+static int __init acpi_hed_init(void)
+{
+	if (acpi_disabled)
+		return -ENODEV;
+
+	if (acpi_bus_register_driver(&acpi_hed_driver) < 0)
+		return -ENODEV;
+
+	return 0;
+}
+
+static void __exit acpi_hed_exit(void)
+{
+	acpi_bus_unregister_driver(&acpi_hed_driver);
+}
+
+module_init(acpi_hed_init);
+module_exit(acpi_hed_exit);
+
+ACPI_MODULE_NAME("hed");
+MODULE_AUTHOR("Huang Ying");
+MODULE_DESCRIPTION("ACPI Hardware Error Device Driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/acpi/hest.c b/drivers/acpi/hest.c
deleted file mode 100644
index 1c527a192872..000000000000
--- a/drivers/acpi/hest.c
+++ /dev/null
@@ -1,139 +0,0 @@
-#include <linux/acpi.h>
-#include <linux/pci.h>
-
-#define PREFIX "ACPI: "
-
-static inline unsigned long parse_acpi_hest_ia_machine_check(struct acpi_hest_ia_machine_check *p)
-{
-	return sizeof(*p) +
-		(sizeof(struct acpi_hest_ia_error_bank) * p->num_hardware_banks);
-}
-
-static inline unsigned long parse_acpi_hest_ia_corrected(struct acpi_hest_ia_corrected *p)
-{
-	return sizeof(*p) +
-		(sizeof(struct acpi_hest_ia_error_bank) * p->num_hardware_banks);
-}
-
-static inline unsigned long parse_acpi_hest_ia_nmi(struct acpi_hest_ia_nmi *p)
-{
-	return sizeof(*p);
-}
-
-static inline unsigned long parse_acpi_hest_generic(struct acpi_hest_generic *p)
-{
-	return sizeof(*p);
-}
-
-static inline unsigned int hest_match_pci(struct acpi_hest_aer_common *p, struct pci_dev *pci)
-{
-	return	(0           == pci_domain_nr(pci->bus) &&
-		 p->bus      == pci->bus->number &&
-		 p->device   == PCI_SLOT(pci->devfn) &&
-		 p->function == PCI_FUNC(pci->devfn));
-}
-
-static unsigned long parse_acpi_hest_aer(void *hdr, int type, struct pci_dev *pci, int *firmware_first)
-{
-	struct acpi_hest_aer_common *p = hdr + sizeof(struct acpi_hest_header);
-	unsigned long rc=0;
-	u8 pcie_type = 0;
-	u8 bridge = 0;
-	switch (type) {
-	case ACPI_HEST_TYPE_AER_ROOT_PORT:
-		rc = sizeof(struct acpi_hest_aer_root);
-		pcie_type = PCI_EXP_TYPE_ROOT_PORT;
-		break;
-	case ACPI_HEST_TYPE_AER_ENDPOINT:
-		rc = sizeof(struct acpi_hest_aer);
-		pcie_type = PCI_EXP_TYPE_ENDPOINT;
-		break;
-	case ACPI_HEST_TYPE_AER_BRIDGE:
-		rc = sizeof(struct acpi_hest_aer_bridge);
-		if ((pci->class >> 16) == PCI_BASE_CLASS_BRIDGE)
-			bridge = 1;
-		break;
-	}
-
-	if (p->flags & ACPI_HEST_GLOBAL) {
-		if ((pci->is_pcie && (pci->pcie_type == pcie_type)) || bridge)
-			*firmware_first = !!(p->flags & ACPI_HEST_FIRMWARE_FIRST);
-	}
-	else
-		if (hest_match_pci(p, pci))
-			*firmware_first = !!(p->flags & ACPI_HEST_FIRMWARE_FIRST);
-	return rc;
-}
-
-static int acpi_hest_firmware_first(struct acpi_table_header *stdheader, struct pci_dev *pci)
-{
-	struct acpi_table_hest *hest = (struct acpi_table_hest *)stdheader;
-	void *p = (void *)hest + sizeof(*hest); /* defined by the ACPI 4.0 spec */
-	struct acpi_hest_header *hdr = p;
-
-	int i;
-	int firmware_first = 0;
-	static unsigned char printed_unused = 0;
-	static unsigned char printed_reserved = 0;
-
-	for (i=0, hdr=p; p < (((void *)hest) + hest->header.length) && i < hest->error_source_count; i++) {
-		switch (hdr->type) {
-		case ACPI_HEST_TYPE_IA32_CHECK:
-			p += parse_acpi_hest_ia_machine_check(p);
-			break;
-		case ACPI_HEST_TYPE_IA32_CORRECTED_CHECK:
-			p += parse_acpi_hest_ia_corrected(p);
-			break;
-		case ACPI_HEST_TYPE_IA32_NMI:
-			p += parse_acpi_hest_ia_nmi(p);
-			break;
-		/* These three should never appear */
-		case ACPI_HEST_TYPE_NOT_USED3:
-		case ACPI_HEST_TYPE_NOT_USED4:
-		case ACPI_HEST_TYPE_NOT_USED5:
-			if (!printed_unused) {
-				printk(KERN_DEBUG PREFIX
-				       "HEST Error Source list contains an obsolete type (%d).\n", hdr->type);
-				printed_unused = 1;
-			}
-			break;
-		case ACPI_HEST_TYPE_AER_ROOT_PORT:
-		case ACPI_HEST_TYPE_AER_ENDPOINT:
-		case ACPI_HEST_TYPE_AER_BRIDGE:
-			p += parse_acpi_hest_aer(p, hdr->type, pci, &firmware_first);
-			break;
-		case ACPI_HEST_TYPE_GENERIC_ERROR:
-			p += parse_acpi_hest_generic(p);
-			break;
-		/* These should never appear either */
-		case ACPI_HEST_TYPE_RESERVED:
-		default:
-			if (!printed_reserved) {
-				printk(KERN_DEBUG PREFIX
-				       "HEST Error Source list contains a reserved type (%d).\n", hdr->type);
-				printed_reserved = 1;
-			}
-			break;
-		}
-	}
-	return firmware_first;
-}
-
-int acpi_hest_firmware_first_pci(struct pci_dev *pci)
-{
-	acpi_status status = AE_NOT_FOUND;
-	struct acpi_table_header *hest = NULL;
-
-	if (acpi_disabled)
-		return 0;
-
-	status = acpi_get_table(ACPI_SIG_HEST, 1, &hest);
-
-	if (ACPI_SUCCESS(status)) {
-		if (acpi_hest_firmware_first(hest, pci)) {
-			return 1;
-		}
-	}
-	return 0;
-}
-EXPORT_SYMBOL_GPL(acpi_hest_firmware_first_pci);
diff --git a/drivers/acpi/internal.h b/drivers/acpi/internal.h
index e28411367239..f8f190ec066e 100644
--- a/drivers/acpi/internal.h
+++ b/drivers/acpi/internal.h
@@ -49,8 +49,9 @@ void acpi_early_processor_set_pdc(void);
 int acpi_ec_init(void);
 int acpi_ec_ecdt_probe(void);
 int acpi_boot_ec_enable(void);
-void acpi_ec_suspend_transactions(void);
-void acpi_ec_resume_transactions(void);
+void acpi_ec_block_transactions(void);
+void acpi_ec_unblock_transactions(void);
+void acpi_ec_unblock_transactions_early(void);
 
 /*--------------------------------------------------------------------------
                                   Suspend/Resume
diff --git a/drivers/acpi/osl.c b/drivers/acpi/osl.c
index 4bc1c4178f50..78418ce4fc78 100644
--- a/drivers/acpi/osl.c
+++ b/drivers/acpi/osl.c
@@ -1207,6 +1207,15 @@ int acpi_check_mem_region(resource_size_t start, resource_size_t n,
 EXPORT_SYMBOL(acpi_check_mem_region);
 
 /*
+ * Let drivers know whether the resource checks are effective
+ */
+int acpi_resources_are_enforced(void)
+{
+	return acpi_enforce_resources == ENFORCE_RESOURCES_STRICT;
+}
+EXPORT_SYMBOL(acpi_resources_are_enforced);
+
+/*
  * Acquire a spinlock.
  *
  * handle is a pointer to the spinlock_t.
diff --git a/drivers/acpi/pci_root.c b/drivers/acpi/pci_root.c
index aefce33f2a09..4eac59393edc 100644
--- a/drivers/acpi/pci_root.c
+++ b/drivers/acpi/pci_root.c
@@ -120,7 +120,8 @@ acpi_handle acpi_get_pci_rootbridge_handle(unsigned int seg, unsigned int bus)
 	struct acpi_pci_root *root;
 	
 	list_for_each_entry(root, &acpi_pci_roots, node)
-		if ((root->segment == (u16) seg) && (root->bus_nr == (u16) bus))
+		if ((root->segment == (u16) seg) &&
+		    (root->secondary.start == (u16) bus))
 			return root->device->handle;
 	return NULL;		
 }
@@ -154,7 +155,7 @@ EXPORT_SYMBOL_GPL(acpi_is_root_bridge);
 static acpi_status
 get_root_bridge_busnr_callback(struct acpi_resource *resource, void *data)
 {
-	int *busnr = data;
+	struct resource *res = data;
 	struct acpi_resource_address64 address;
 
 	if (resource->type != ACPI_RESOURCE_TYPE_ADDRESS16 &&
@@ -164,28 +165,27 @@ get_root_bridge_busnr_callback(struct acpi_resource *resource, void *data)
 
 	acpi_resource_to_address64(resource, &address);
 	if ((address.address_length > 0) &&
-	    (address.resource_type == ACPI_BUS_NUMBER_RANGE))
-		*busnr = address.minimum;
+	    (address.resource_type == ACPI_BUS_NUMBER_RANGE)) {
+		res->start = address.minimum;
+		res->end = address.minimum + address.address_length - 1;
+	}
 
 	return AE_OK;
 }
 
 static acpi_status try_get_root_bridge_busnr(acpi_handle handle,
-					     unsigned long long *bus)
+					     struct resource *res)
 {
 	acpi_status status;
-	int busnum;
 
-	busnum = -1;
+	res->start = -1;
 	status =
 	    acpi_walk_resources(handle, METHOD_NAME__CRS,
-				get_root_bridge_busnr_callback, &busnum);
+				get_root_bridge_busnr_callback, res);
 	if (ACPI_FAILURE(status))
 		return status;
-	/* Check if we really get a bus number from _CRS */
-	if (busnum == -1)
+	if (res->start == -1)
 		return AE_ERROR;
-	*bus = busnum;
 	return AE_OK;
 }
 
@@ -429,34 +429,47 @@ static int __devinit acpi_pci_root_add(struct acpi_device *device)
 	struct acpi_device *child;
 	u32 flags, base_flags;
 
+	root = kzalloc(sizeof(struct acpi_pci_root), GFP_KERNEL);
+	if (!root)
+		return -ENOMEM;
+
 	segment = 0;
 	status = acpi_evaluate_integer(device->handle, METHOD_NAME__SEG, NULL,
 				       &segment);
 	if (ACPI_FAILURE(status) && status != AE_NOT_FOUND) {
 		printk(KERN_ERR PREFIX "can't evaluate _SEG\n");
-		return -ENODEV;
+		result = -ENODEV;
+		goto end;
 	}
 
 	/* Check _CRS first, then _BBN.  If no _BBN, default to zero. */
-	bus = 0;
-	status = try_get_root_bridge_busnr(device->handle, &bus);
+	root->secondary.flags = IORESOURCE_BUS;
+	status = try_get_root_bridge_busnr(device->handle, &root->secondary);
 	if (ACPI_FAILURE(status)) {
+		/*
+		 * We need both the start and end of the downstream bus range
+		 * to interpret _CBA (MMCONFIG base address), so it really is
+		 * supposed to be in _CRS.  If we don't find it there, all we
+		 * can do is assume [_BBN-0xFF] or [0-0xFF].
+		 */
+		root->secondary.end = 0xFF;
+		printk(KERN_WARNING FW_BUG PREFIX
+		       "no secondary bus range in _CRS\n");
 		status = acpi_evaluate_integer(device->handle, METHOD_NAME__BBN,					       NULL, &bus);
-		if (ACPI_FAILURE(status) && status != AE_NOT_FOUND) {
-			printk(KERN_ERR PREFIX
-			     "no bus number in _CRS and can't evaluate _BBN\n");
-			return -ENODEV;
+		if (ACPI_SUCCESS(status))
+			root->secondary.start = bus;
+		else if (status == AE_NOT_FOUND)
+			root->secondary.start = 0;
+		else {
+			printk(KERN_ERR PREFIX "can't evaluate _BBN\n");
+			result = -ENODEV;
+			goto end;
 		}
 	}
 
-	root = kzalloc(sizeof(struct acpi_pci_root), GFP_KERNEL);
-	if (!root)
-		return -ENOMEM;
-
 	INIT_LIST_HEAD(&root->node);
 	root->device = device;
 	root->segment = segment & 0xFFFF;
-	root->bus_nr = bus & 0xFF;
 	strcpy(acpi_device_name(device), ACPI_PCI_ROOT_DEVICE_NAME);
 	strcpy(acpi_device_class(device), ACPI_PCI_ROOT_CLASS);
 	device->driver_data = root;
@@ -475,9 +488,9 @@ static int __devinit acpi_pci_root_add(struct acpi_device *device)
 	/* TBD: Locking */
 	list_add_tail(&root->node, &acpi_pci_roots);
 
-	printk(KERN_INFO PREFIX "%s [%s] (%04x:%02x)\n",
+	printk(KERN_INFO PREFIX "%s [%s] (domain %04x %pR)\n",
 	       acpi_device_name(device), acpi_device_bid(device),
-	       root->segment, root->bus_nr);
+	       root->segment, &root->secondary);
 
 	/*
 	 * Scan the Root Bridge
@@ -486,11 +499,11 @@ static int __devinit acpi_pci_root_add(struct acpi_device *device)
 	 * PCI namespace does not get created until this call is made (and 
 	 * thus the root bridge's pci_dev does not exist).
 	 */
-	root->bus = pci_acpi_scan_root(device, segment, bus);
+	root->bus = pci_acpi_scan_root(root);
 	if (!root->bus) {
 		printk(KERN_ERR PREFIX
 			    "Bus %04x:%02x not present in PCI namespace\n",
-			    root->segment, root->bus_nr);
+			    root->segment, (unsigned int)root->secondary.start);
 		result = -ENODEV;
 		goto end;
 	}
diff --git a/drivers/acpi/processor_driver.c b/drivers/acpi/processor_driver.c
index 5675d9747e87..b1034a9ada4e 100644
--- a/drivers/acpi/processor_driver.c
+++ b/drivers/acpi/processor_driver.c
@@ -616,7 +616,8 @@ static int __cpuinit acpi_processor_add(struct acpi_device *device)
 	acpi_processor_get_limit_info(pr);
 
 
-	acpi_processor_power_init(pr, device);
+	if (cpuidle_get_driver() == &acpi_idle_driver)
+		acpi_processor_power_init(pr, device);
 
 	pr->cdev = thermal_cooling_device_register("Processor", device,
 						&processor_cooling_ops);
@@ -920,9 +921,14 @@ static int __init acpi_processor_init(void)
 	if (!acpi_processor_dir)
 		return -ENOMEM;
 #endif
-	result = cpuidle_register_driver(&acpi_idle_driver);
-	if (result < 0)
-		goto out_proc;
+
+	if (!cpuidle_register_driver(&acpi_idle_driver)) {
+		printk(KERN_DEBUG "ACPI: %s registered with cpuidle\n",
+			acpi_idle_driver.name);
+	} else {
+		printk(KERN_DEBUG "ACPI: acpi_idle yielding to %s",
+			cpuidle_get_driver()->name);
+	}
 
 	result = acpi_bus_register_driver(&acpi_processor_driver);
 	if (result < 0)
@@ -941,7 +947,6 @@ static int __init acpi_processor_init(void)
 out_cpuidle:
 	cpuidle_unregister_driver(&acpi_idle_driver);
 
-out_proc:
 #ifdef CONFIG_ACPI_PROCFS
 	remove_proc_entry(ACPI_PROCESSOR_CLASS, acpi_root_dir);
 #endif
diff --git a/drivers/acpi/processor_idle.c b/drivers/acpi/processor_idle.c
index c3817e1f32c7..b1b385692f46 100644
--- a/drivers/acpi/processor_idle.c
+++ b/drivers/acpi/processor_idle.c
@@ -80,7 +80,7 @@ module_param(nocst, uint, 0000);
 static unsigned int latency_factor __read_mostly = 2;
 module_param(latency_factor, uint, 0644);
 
-static s64 us_to_pm_timer_ticks(s64 t)
+static u64 us_to_pm_timer_ticks(s64 t)
 {
 	return div64_u64(t * PM_TIMER_FREQUENCY, 1000000);
 }
@@ -727,24 +727,14 @@ static int acpi_processor_power_seq_show(struct seq_file *seq, void *offset)
 			break;
 		}
 
-		if (pr->power.states[i].promotion.state)
-			seq_printf(seq, "promotion[C%zd] ",
-				   (pr->power.states[i].promotion.state -
-				    pr->power.states));
-		else
-			seq_puts(seq, "promotion[--] ");
-
-		if (pr->power.states[i].demotion.state)
-			seq_printf(seq, "demotion[C%zd] ",
-				   (pr->power.states[i].demotion.state -
-				    pr->power.states));
-		else
-			seq_puts(seq, "demotion[--] ");
-
-		seq_printf(seq, "latency[%03d] usage[%08d] duration[%020llu]\n",
+		seq_puts(seq, "promotion[--] ");
+
+		seq_puts(seq, "demotion[--] ");
+
+		seq_printf(seq, "latency[%03d] usage[%08d] duration[%020Lu]\n",
 			   pr->power.states[i].latency,
 			   pr->power.states[i].usage,
-			   (unsigned long long)pr->power.states[i].time);
+			   us_to_pm_timer_ticks(pr->power.states[i].time));
 	}
 
       end:
@@ -869,8 +859,8 @@ static int acpi_idle_enter_simple(struct cpuidle_device *dev,
 	struct acpi_processor *pr;
 	struct acpi_processor_cx *cx = cpuidle_get_statedata(state);
 	ktime_t  kt1, kt2;
+	s64 idle_time_ns;
 	s64 idle_time;
-	s64 sleep_ticks = 0;
 
 	pr = __get_cpu_var(processors);
 
@@ -881,6 +871,7 @@ static int acpi_idle_enter_simple(struct cpuidle_device *dev,
 		return(acpi_idle_enter_c1(dev, state));
 
 	local_irq_disable();
+
 	if (cx->entry_method != ACPI_CSTATE_FFH) {
 		current_thread_info()->status &= ~TS_POLLING;
 		/*
@@ -888,12 +879,12 @@ static int acpi_idle_enter_simple(struct cpuidle_device *dev,
 		 * NEED_RESCHED:
 		 */
 		smp_mb();
-	}
 
-	if (unlikely(need_resched())) {
-		current_thread_info()->status |= TS_POLLING;
-		local_irq_enable();
-		return 0;
+		if (unlikely(need_resched())) {
+			current_thread_info()->status |= TS_POLLING;
+			local_irq_enable();
+			return 0;
+		}
 	}
 
 	/*
@@ -910,20 +901,21 @@ static int acpi_idle_enter_simple(struct cpuidle_device *dev,
 	sched_clock_idle_sleep_event();
 	acpi_idle_do_entry(cx);
 	kt2 = ktime_get_real();
-	idle_time =  ktime_to_us(ktime_sub(kt2, kt1));
-
-	sleep_ticks = us_to_pm_timer_ticks(idle_time);
+	idle_time_ns = ktime_to_ns(ktime_sub(kt2, kt1));
+	idle_time = idle_time_ns;
+	do_div(idle_time, NSEC_PER_USEC);
 
 	/* Tell the scheduler how much we idled: */
-	sched_clock_idle_wakeup_event(sleep_ticks*PM_TIMER_TICK_NS);
+	sched_clock_idle_wakeup_event(idle_time_ns);
 
 	local_irq_enable();
-	current_thread_info()->status |= TS_POLLING;
+	if (cx->entry_method != ACPI_CSTATE_FFH)
+		current_thread_info()->status |= TS_POLLING;
 
 	cx->usage++;
 
 	lapic_timer_state_broadcast(pr, cx, 0);
-	cx->time += sleep_ticks;
+	cx->time += idle_time;
 	return idle_time;
 }
 
@@ -943,8 +935,8 @@ static int acpi_idle_enter_bm(struct cpuidle_device *dev,
 	struct acpi_processor *pr;
 	struct acpi_processor_cx *cx = cpuidle_get_statedata(state);
 	ktime_t  kt1, kt2;
+	s64 idle_time_ns;
 	s64 idle_time;
-	s64 sleep_ticks = 0;
 
 
 	pr = __get_cpu_var(processors);
@@ -968,6 +960,7 @@ static int acpi_idle_enter_bm(struct cpuidle_device *dev,
 	}
 
 	local_irq_disable();
+
 	if (cx->entry_method != ACPI_CSTATE_FFH) {
 		current_thread_info()->status &= ~TS_POLLING;
 		/*
@@ -975,12 +968,12 @@ static int acpi_idle_enter_bm(struct cpuidle_device *dev,
 		 * NEED_RESCHED:
 		 */
 		smp_mb();
-	}
 
-	if (unlikely(need_resched())) {
-		current_thread_info()->status |= TS_POLLING;
-		local_irq_enable();
-		return 0;
+		if (unlikely(need_resched())) {
+			current_thread_info()->status |= TS_POLLING;
+			local_irq_enable();
+			return 0;
+		}
 	}
 
 	acpi_unlazy_tlb(smp_processor_id());
@@ -1025,19 +1018,21 @@ static int acpi_idle_enter_bm(struct cpuidle_device *dev,
 		spin_unlock(&c3_lock);
 	}
 	kt2 = ktime_get_real();
-	idle_time =  ktime_to_us(ktime_sub(kt2, kt1));
+	idle_time_ns = ktime_to_ns(ktime_sub(kt2, kt1));
+	idle_time = idle_time_ns;
+	do_div(idle_time, NSEC_PER_USEC);
 
-	sleep_ticks = us_to_pm_timer_ticks(idle_time);
 	/* Tell the scheduler how much we idled: */
-	sched_clock_idle_wakeup_event(sleep_ticks*PM_TIMER_TICK_NS);
+	sched_clock_idle_wakeup_event(idle_time_ns);
 
 	local_irq_enable();
-	current_thread_info()->status |= TS_POLLING;
+	if (cx->entry_method != ACPI_CSTATE_FFH)
+		current_thread_info()->status |= TS_POLLING;
 
 	cx->usage++;
 
 	lapic_timer_state_broadcast(pr, cx, 0);
-	cx->time += sleep_ticks;
+	cx->time += idle_time;
 	return idle_time;
 }
 
diff --git a/drivers/acpi/sleep.c b/drivers/acpi/sleep.c
index baa76bbf244a..3fb4bdea7e06 100644
--- a/drivers/acpi/sleep.c
+++ b/drivers/acpi/sleep.c
@@ -80,22 +80,6 @@ static int acpi_sleep_prepare(u32 acpi_state)
 
 #ifdef CONFIG_ACPI_SLEEP
 static u32 acpi_target_sleep_state = ACPI_STATE_S0;
-/*
- * According to the ACPI specification the BIOS should make sure that ACPI is
- * enabled and SCI_EN bit is set on wake-up from S1 - S3 sleep states.  Still,
- * some BIOSes don't do that and therefore we use acpi_enable() to enable ACPI
- * on such systems during resume.  Unfortunately that doesn't help in
- * particularly pathological cases in which SCI_EN has to be set directly on
- * resume, although the specification states very clearly that this flag is
- * owned by the hardware.  The set_sci_en_on_resume variable will be set in such
- * cases.
- */
-static bool set_sci_en_on_resume;
-
-void __init acpi_set_sci_en_on_resume(void)
-{
-	set_sci_en_on_resume = true;
-}
 
 /*
  * ACPI 1.0 wants us to execute _PTS before suspending devices, so we allow the
@@ -110,11 +94,13 @@ void __init acpi_old_suspend_ordering(void)
 }
 
 /**
- *	acpi_pm_disable_gpes - Disable the GPEs.
+ * acpi_pm_freeze - Disable the GPEs and suspend EC transactions.
  */
-static int acpi_pm_disable_gpes(void)
+static int acpi_pm_freeze(void)
 {
 	acpi_disable_all_gpes();
+	acpi_os_wait_events_complete(NULL);
+	acpi_ec_block_transactions();
 	return 0;
 }
 
@@ -142,7 +128,8 @@ static int acpi_pm_prepare(void)
 	int error = __acpi_pm_prepare();
 
 	if (!error)
-		acpi_disable_all_gpes();
+		acpi_pm_freeze();
+
 	return error;
 }
 
@@ -253,11 +240,8 @@ static int acpi_suspend_enter(suspend_state_t pm_state)
 		break;
 	}
 
-	/* If ACPI is not enabled by the BIOS, we need to enable it here. */
-	if (set_sci_en_on_resume)
-		acpi_write_bit_register(ACPI_BITREG_SCI_ENABLE, 1);
-	else
-		acpi_enable();
+	/* This violates the spec but is required for bug compatibility. */
+	acpi_write_bit_register(ACPI_BITREG_SCI_ENABLE, 1);
 
 	/* Reprogram control registers and execute _BFS */
 	acpi_leave_sleep_state_prep(acpi_state);
@@ -275,6 +259,8 @@ static int acpi_suspend_enter(suspend_state_t pm_state)
 	 * acpi_leave_sleep_state will reenable specific GPEs later
 	 */
 	acpi_disable_all_gpes();
+	/* Allow EC transactions to happen. */
+	acpi_ec_unblock_transactions_early();
 
 	local_irq_restore(flags);
 	printk(KERN_DEBUG "Back to C!\n");
@@ -286,6 +272,12 @@ static int acpi_suspend_enter(suspend_state_t pm_state)
 	return ACPI_SUCCESS(status) ? 0 : -EFAULT;
 }
 
+static void acpi_suspend_finish(void)
+{
+	acpi_ec_unblock_transactions();
+	acpi_pm_finish();
+}
+
 static int acpi_suspend_state_valid(suspend_state_t pm_state)
 {
 	u32 acpi_state;
@@ -307,7 +299,7 @@ static struct platform_suspend_ops acpi_suspend_ops = {
 	.begin = acpi_suspend_begin,
 	.prepare_late = acpi_pm_prepare,
 	.enter = acpi_suspend_enter,
-	.wake = acpi_pm_finish,
+	.wake = acpi_suspend_finish,
 	.end = acpi_pm_end,
 };
 
@@ -333,9 +325,9 @@ static int acpi_suspend_begin_old(suspend_state_t pm_state)
 static struct platform_suspend_ops acpi_suspend_ops_old = {
 	.valid = acpi_suspend_state_valid,
 	.begin = acpi_suspend_begin_old,
-	.prepare_late = acpi_pm_disable_gpes,
+	.prepare_late = acpi_pm_freeze,
 	.enter = acpi_suspend_enter,
-	.wake = acpi_pm_finish,
+	.wake = acpi_suspend_finish,
 	.end = acpi_pm_end,
 	.recover = acpi_pm_finish,
 };
@@ -346,12 +338,6 @@ static int __init init_old_suspend_ordering(const struct dmi_system_id *d)
 	return 0;
 }
 
-static int __init init_set_sci_en_on_resume(const struct dmi_system_id *d)
-{
-	set_sci_en_on_resume = true;
-	return 0;
-}
-
 static struct dmi_system_id __initdata acpisleep_dmi_table[] = {
 	{
 	.callback = init_old_suspend_ordering,
@@ -370,22 +356,6 @@ static struct dmi_system_id __initdata acpisleep_dmi_table[] = {
 		},
 	},
 	{
-	.callback = init_set_sci_en_on_resume,
-	.ident = "Apple MacBook 1,1",
-	.matches = {
-		DMI_MATCH(DMI_SYS_VENDOR, "Apple Computer, Inc."),
-		DMI_MATCH(DMI_PRODUCT_NAME, "MacBook1,1"),
-		},
-	},
-	{
-	.callback = init_set_sci_en_on_resume,
-	.ident = "Apple MacMini 1,1",
-	.matches = {
-		DMI_MATCH(DMI_SYS_VENDOR, "Apple Computer, Inc."),
-		DMI_MATCH(DMI_PRODUCT_NAME, "Macmini1,1"),
-		},
-	},
-	{
 	.callback = init_old_suspend_ordering,
 	.ident = "Asus Pundit P1-AH2 (M2N8L motherboard)",
 	.matches = {
@@ -394,94 +364,6 @@ static struct dmi_system_id __initdata acpisleep_dmi_table[] = {
 		},
 	},
 	{
-	.callback = init_set_sci_en_on_resume,
-	.ident = "Toshiba Satellite L300",
-	.matches = {
-		DMI_MATCH(DMI_SYS_VENDOR, "TOSHIBA"),
-		DMI_MATCH(DMI_PRODUCT_NAME, "Satellite L300"),
-		},
-	},
-	{
-	.callback = init_set_sci_en_on_resume,
-	.ident = "Hewlett-Packard HP G7000 Notebook PC",
-	.matches = {
-		DMI_MATCH(DMI_SYS_VENDOR, "Hewlett-Packard"),
-		DMI_MATCH(DMI_PRODUCT_NAME, "HP G7000 Notebook PC"),
-		},
-	},
-	{
-	.callback = init_set_sci_en_on_resume,
-	.ident = "Hewlett-Packard HP Pavilion dv3 Notebook PC",
-	.matches = {
-		DMI_MATCH(DMI_SYS_VENDOR, "Hewlett-Packard"),
-		DMI_MATCH(DMI_PRODUCT_NAME, "HP Pavilion dv3 Notebook PC"),
-		},
-	},
-	{
-	.callback = init_set_sci_en_on_resume,
-	.ident = "Hewlett-Packard Pavilion dv4",
-	.matches = {
-		DMI_MATCH(DMI_SYS_VENDOR, "Hewlett-Packard"),
-		DMI_MATCH(DMI_PRODUCT_NAME, "HP Pavilion dv4"),
-		},
-	},
-	{
-	.callback = init_set_sci_en_on_resume,
-	.ident = "Hewlett-Packard Pavilion dv7",
-	.matches = {
-		DMI_MATCH(DMI_SYS_VENDOR, "Hewlett-Packard"),
-		DMI_MATCH(DMI_PRODUCT_NAME, "HP Pavilion dv7"),
-		},
-	},
-	{
-	.callback = init_set_sci_en_on_resume,
-	.ident = "Hewlett-Packard Compaq Presario C700 Notebook PC",
-	.matches = {
-		DMI_MATCH(DMI_SYS_VENDOR, "Hewlett-Packard"),
-		DMI_MATCH(DMI_PRODUCT_NAME, "Compaq Presario C700 Notebook PC"),
-		},
-	},
-	{
-	.callback = init_set_sci_en_on_resume,
-	.ident = "Hewlett-Packard Compaq Presario CQ40 Notebook PC",
-	.matches = {
-		DMI_MATCH(DMI_SYS_VENDOR, "Hewlett-Packard"),
-		DMI_MATCH(DMI_PRODUCT_NAME, "Compaq Presario CQ40 Notebook PC"),
-		},
-	},
-	{
-	.callback = init_set_sci_en_on_resume,
-	.ident = "Lenovo ThinkPad T410",
-	.matches = {
-		DMI_MATCH(DMI_SYS_VENDOR, "LENOVO"),
-		DMI_MATCH(DMI_PRODUCT_VERSION, "ThinkPad T410"),
-		},
-	},
-	{
-	.callback = init_set_sci_en_on_resume,
-	.ident = "Lenovo ThinkPad T510",
-	.matches = {
-		DMI_MATCH(DMI_SYS_VENDOR, "LENOVO"),
-		DMI_MATCH(DMI_PRODUCT_VERSION, "ThinkPad T510"),
-		},
-	},
-	{
-	.callback = init_set_sci_en_on_resume,
-	.ident = "Lenovo ThinkPad W510",
-	.matches = {
-		DMI_MATCH(DMI_SYS_VENDOR, "LENOVO"),
-		DMI_MATCH(DMI_PRODUCT_VERSION, "ThinkPad W510"),
-		},
-	},
-	{
-	.callback = init_set_sci_en_on_resume,
-	.ident = "Lenovo ThinkPad X201[s]",
-	.matches = {
-		DMI_MATCH(DMI_SYS_VENDOR, "LENOVO"),
-		DMI_MATCH(DMI_PRODUCT_VERSION, "ThinkPad X201"),
-		},
-	},
-	{
 	.callback = init_old_suspend_ordering,
 	.ident = "Panasonic CF51-2L",
 	.matches = {
@@ -490,30 +372,6 @@ static struct dmi_system_id __initdata acpisleep_dmi_table[] = {
 		DMI_MATCH(DMI_BOARD_NAME, "CF51-2L"),
 		},
 	},
-	{
-	.callback = init_set_sci_en_on_resume,
-	.ident = "Dell Studio 1558",
-	.matches = {
-		DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
-		DMI_MATCH(DMI_PRODUCT_NAME, "Studio 1558"),
-		},
-	},
-	{
-	.callback = init_set_sci_en_on_resume,
-	.ident = "Dell Studio 1557",
-	.matches = {
-		DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
-		DMI_MATCH(DMI_PRODUCT_NAME, "Studio 1557"),
-		},
-	},
-	{
-	.callback = init_set_sci_en_on_resume,
-	.ident = "Dell Studio 1555",
-	.matches = {
-		DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
-		DMI_MATCH(DMI_PRODUCT_NAME, "Studio 1555"),
-		},
-	},
 	{},
 };
 #endif /* CONFIG_SUSPEND */
@@ -586,6 +444,7 @@ static int acpi_hibernation_enter(void)
 static void acpi_hibernation_finish(void)
 {
 	hibernate_nvs_free();
+	acpi_ec_unblock_transactions();
 	acpi_pm_finish();
 }
 
@@ -606,19 +465,13 @@ static void acpi_hibernation_leave(void)
 	}
 	/* Restore the NVS memory area */
 	hibernate_nvs_restore();
+	/* Allow EC transactions to happen. */
+	acpi_ec_unblock_transactions_early();
 }
 
-static int acpi_pm_pre_restore(void)
+static void acpi_pm_thaw(void)
 {
-	acpi_disable_all_gpes();
-	acpi_os_wait_events_complete(NULL);
-	acpi_ec_suspend_transactions();
-	return 0;
-}
-
-static void acpi_pm_restore_cleanup(void)
-{
-	acpi_ec_resume_transactions();
+	acpi_ec_unblock_transactions();
 	acpi_enable_all_runtime_gpes();
 }
 
@@ -630,8 +483,8 @@ static struct platform_hibernation_ops acpi_hibernation_ops = {
 	.prepare = acpi_pm_prepare,
 	.enter = acpi_hibernation_enter,
 	.leave = acpi_hibernation_leave,
-	.pre_restore = acpi_pm_pre_restore,
-	.restore_cleanup = acpi_pm_restore_cleanup,
+	.pre_restore = acpi_pm_freeze,
+	.restore_cleanup = acpi_pm_thaw,
 };
 
 /**
@@ -663,12 +516,9 @@ static int acpi_hibernation_begin_old(void)
 
 static int acpi_hibernation_pre_snapshot_old(void)
 {
-	int error = acpi_pm_disable_gpes();
-
-	if (!error)
-		hibernate_nvs_save();
-
-	return error;
+	acpi_pm_freeze();
+	hibernate_nvs_save();
+	return 0;
 }
 
 /*
@@ -680,11 +530,11 @@ static struct platform_hibernation_ops acpi_hibernation_ops_old = {
 	.end = acpi_pm_end,
 	.pre_snapshot = acpi_hibernation_pre_snapshot_old,
 	.finish = acpi_hibernation_finish,
-	.prepare = acpi_pm_disable_gpes,
+	.prepare = acpi_pm_freeze,
 	.enter = acpi_hibernation_enter,
 	.leave = acpi_hibernation_leave,
-	.pre_restore = acpi_pm_pre_restore,
-	.restore_cleanup = acpi_pm_restore_cleanup,
+	.pre_restore = acpi_pm_freeze,
+	.restore_cleanup = acpi_pm_thaw,
 	.recover = acpi_pm_finish,
 };
 #endif /* CONFIG_HIBERNATION */
diff --git a/drivers/acpi/sleep.h b/drivers/acpi/sleep.h
index 8a8f3b3382a6..25b8bd149284 100644
--- a/drivers/acpi/sleep.h
+++ b/drivers/acpi/sleep.h
@@ -1,6 +1,6 @@
 
 extern u8 sleep_states[];
-extern int acpi_suspend (u32 state);
+extern int acpi_suspend(u32 state);
 
 extern void acpi_enable_wakeup_device_prep(u8 sleep_state);
 extern void acpi_enable_wakeup_device(u8 sleep_state);
diff --git a/drivers/acpi/tables.c b/drivers/acpi/tables.c
index 8a0ed2800e63..f336bca7c450 100644
--- a/drivers/acpi/tables.c
+++ b/drivers/acpi/tables.c
@@ -213,7 +213,7 @@ acpi_table_parse_entries(char *id,
 	unsigned long table_end;
 	acpi_size tbl_size;
 
-	if (acpi_disabled && !acpi_ht)
+	if (acpi_disabled)
 		return -ENODEV;
 
 	if (!handler)
@@ -280,7 +280,7 @@ int __init acpi_table_parse(char *id, acpi_table_handler handler)
 	struct acpi_table_header *table = NULL;
 	acpi_size tbl_size;
 
-	if (acpi_disabled && !acpi_ht)
+	if (acpi_disabled)
 		return -ENODEV;
 
 	if (!handler)
diff --git a/drivers/acpi/video.c b/drivers/acpi/video.c
index a0c93b321482..9865d46f49a8 100644
--- a/drivers/acpi/video.c
+++ b/drivers/acpi/video.c
@@ -45,6 +45,7 @@
 #include <acpi/acpi_bus.h>
 #include <acpi/acpi_drivers.h>
 #include <linux/suspend.h>
+#include <acpi/video.h>
 
 #define PREFIX "ACPI: "
 
@@ -65,11 +66,6 @@
 
 #define MAX_NAME_LEN	20
 
-#define ACPI_VIDEO_DISPLAY_CRT	1
-#define ACPI_VIDEO_DISPLAY_TV	2
-#define ACPI_VIDEO_DISPLAY_DVI	3
-#define ACPI_VIDEO_DISPLAY_LCD	4
-
 #define _COMPONENT		ACPI_VIDEO_COMPONENT
 ACPI_MODULE_NAME("video");
 
@@ -1007,11 +1003,11 @@ static void acpi_video_device_find_cap(struct acpi_video_device *device)
 		result = acpi_video_init_brightness(device);
 		if (result)
 			return;
-		name = kzalloc(MAX_NAME_LEN, GFP_KERNEL);
+		name = kasprintf(GFP_KERNEL, "acpi_video%d", count);
 		if (!name)
 			return;
+		count++;
 
-		sprintf(name, "acpi_video%d", count++);
 		memset(&props, 0, sizeof(struct backlight_properties));
 		props.max_brightness = device->brightness->count - 3;
 		device->backlight = backlight_device_register(name, NULL, device,
@@ -1067,10 +1063,10 @@ static void acpi_video_device_find_cap(struct acpi_video_device *device)
 		if (device->cap._DCS && device->cap._DSS) {
 			static int count;
 			char *name;
-			name = kzalloc(MAX_NAME_LEN, GFP_KERNEL);
+			name = kasprintf(GFP_KERNEL, "acpi_video%d", count);
 			if (!name)
 				return;
-			sprintf(name, "acpi_video%d", count++);
+			count++;
 			device->output_dev = video_output_register(name,
 					NULL, device, &acpi_output_properties);
 			kfree(name);
@@ -1748,11 +1744,27 @@ acpi_video_get_device_attr(struct acpi_video_bus *video, unsigned long device_id
 }
 
 static int
+acpi_video_get_device_type(struct acpi_video_bus *video,
+			   unsigned long device_id)
+{
+	struct acpi_video_enumerated_device *ids;
+	int i;
+
+	for (i = 0; i < video->attached_count; i++) {
+		ids = &video->attached_array[i];
+		if ((ids->value.int_val & 0xffff) == device_id)
+			return ids->value.int_val;
+	}
+
+	return 0;
+}
+
+static int
 acpi_video_bus_get_one_device(struct acpi_device *device,
 			      struct acpi_video_bus *video)
 {
 	unsigned long long device_id;
-	int status;
+	int status, device_type;
 	struct acpi_video_device *data;
 	struct acpi_video_device_attrib* attribute;
 
@@ -1797,8 +1809,25 @@ acpi_video_bus_get_one_device(struct acpi_device *device,
 			}
 			if(attribute->bios_can_detect)
 				data->flags.bios = 1;
-		} else
-			data->flags.unknown = 1;
+		} else {
+			/* Check for legacy IDs */
+			device_type = acpi_video_get_device_type(video,
+								 device_id);
+			/* Ignore bits 16 and 18-20 */
+			switch (device_type & 0xffe2ffff) {
+			case ACPI_VIDEO_DISPLAY_LEGACY_MONITOR:
+				data->flags.crt = 1;
+				break;
+			case ACPI_VIDEO_DISPLAY_LEGACY_PANEL:
+				data->flags.lcd = 1;
+				break;
+			case ACPI_VIDEO_DISPLAY_LEGACY_TV:
+				data->flags.tvout = 1;
+				break;
+			default:
+				data->flags.unknown = 1;
+			}
+		}
 
 		acpi_video_device_bind(video, data);
 		acpi_video_device_find_cap(data);
@@ -2032,6 +2061,71 @@ out:
 	return result;
 }
 
+int acpi_video_get_edid(struct acpi_device *device, int type, int device_id,
+			void **edid)
+{
+	struct acpi_video_bus *video;
+	struct acpi_video_device *video_device;
+	union acpi_object *buffer = NULL;
+	acpi_status status;
+	int i, length;
+
+	if (!device || !acpi_driver_data(device))
+		return -EINVAL;
+
+	video = acpi_driver_data(device);
+
+	for (i = 0; i < video->attached_count; i++) {
+		video_device = video->attached_array[i].bind_info;
+		length = 256;
+
+		if (!video_device)
+			continue;
+
+		if (type) {
+			switch (type) {
+			case ACPI_VIDEO_DISPLAY_CRT:
+				if (!video_device->flags.crt)
+					continue;
+				break;
+			case ACPI_VIDEO_DISPLAY_TV:
+				if (!video_device->flags.tvout)
+					continue;
+				break;
+			case ACPI_VIDEO_DISPLAY_DVI:
+				if (!video_device->flags.dvi)
+					continue;
+				break;
+			case ACPI_VIDEO_DISPLAY_LCD:
+				if (!video_device->flags.lcd)
+					continue;
+				break;
+			}
+		} else if (video_device->device_id != device_id) {
+			continue;
+		}
+
+		status = acpi_video_device_EDID(video_device, &buffer, length);
+
+		if (ACPI_FAILURE(status) || !buffer ||
+		    buffer->type != ACPI_TYPE_BUFFER) {
+			length = 128;
+			status = acpi_video_device_EDID(video_device, &buffer,
+							length);
+			if (ACPI_FAILURE(status) || !buffer ||
+			    buffer->type != ACPI_TYPE_BUFFER) {
+				continue;
+			}
+		}
+
+		*edid = buffer->buffer.pointer;
+		return length;
+	}
+
+	return -ENODEV;
+}
+EXPORT_SYMBOL(acpi_video_get_edid);
+
 static int
 acpi_video_bus_get_devices(struct acpi_video_bus *video,
 			   struct acpi_device *device)
diff --git a/drivers/acpi/video_detect.c b/drivers/acpi/video_detect.c
index fc2f26b9b407..c5fef01b3c95 100644
--- a/drivers/acpi/video_detect.c
+++ b/drivers/acpi/video_detect.c
@@ -250,7 +250,7 @@ static int __init acpi_backlight(char *str)
 				ACPI_VIDEO_BACKLIGHT_FORCE_VENDOR;
 		if (!strcmp("video", str))
 			acpi_video_support |=
-				ACPI_VIDEO_OUTPUT_SWITCHING_FORCE_VIDEO;
+				ACPI_VIDEO_BACKLIGHT_FORCE_VIDEO;
 	}
 	return 1;
 }
diff --git a/drivers/ata/Kconfig b/drivers/ata/Kconfig
index e68541f662b9..aa85a98d3a4f 100644
--- a/drivers/ata/Kconfig
+++ b/drivers/ata/Kconfig
@@ -57,6 +57,8 @@ config SATA_PMP
 	  This option adds support for SATA Port Multipliers
 	  (the SATA version of an ethernet hub, or SAS expander).
 
+comment "Controllers with non-SFF native interface"
+
 config SATA_AHCI
 	tristate "AHCI SATA support"
 	depends on PCI
@@ -73,11 +75,12 @@ config SATA_AHCI_PLATFORM
 
 	  If unsure, say N.
 
-config SATA_SIL24
-	tristate "Silicon Image 3124/3132 SATA support"
-	depends on PCI
+config SATA_FSL
+	tristate "Freescale 3.0Gbps SATA support"
+	depends on FSL_SOC
 	help
-	  This option enables support for Silicon Image 3124/3132 Serial ATA.
+	  This option enables support for Freescale 3.0Gbps SATA controller.
+	  It can be found on MPC837x and MPC8315.
 
 	  If unsure, say N.
 
@@ -87,12 +90,11 @@ config SATA_INIC162X
 	help
 	  This option enables support for Initio 162x Serial ATA.
 
-config SATA_FSL
-	tristate "Freescale 3.0Gbps SATA support"
-	depends on FSL_SOC
+config SATA_SIL24
+	tristate "Silicon Image 3124/3132 SATA support"
+	depends on PCI
 	help
-	  This option enables support for Freescale 3.0Gbps SATA controller.
-	  It can be found on MPC837x and MPC8315.
+	  This option enables support for Silicon Image 3124/3132 Serial ATA.
 
 	  If unsure, say N.
 
@@ -116,15 +118,65 @@ config ATA_SFF
 
 if ATA_SFF
 
-config SATA_SVW
-	tristate "ServerWorks Frodo / Apple K2 SATA support"
+comment "SFF controllers with custom DMA interface"
+
+config PDC_ADMA
+	tristate "Pacific Digital ADMA support"
 	depends on PCI
 	help
-	  This option enables support for Broadcom/Serverworks/Apple K2
-	  SATA support.
+	  This option enables support for Pacific Digital ADMA controllers
+
+	  If unsure, say N.
+
+config PATA_MPC52xx
+	tristate "Freescale MPC52xx SoC internal IDE"
+	depends on PPC_MPC52xx && PPC_BESTCOMM
+	select PPC_BESTCOMM_ATA
+	help
+	  This option enables support for integrated IDE controller
+	  of the Freescale MPC52xx SoC.
+
+	  If unsure, say N.
+
+config PATA_OCTEON_CF
+	tristate "OCTEON Boot Bus Compact Flash support"
+	depends on CPU_CAVIUM_OCTEON
+	help
+	  This option enables a polled compact flash driver for use with
+	  compact flash cards attached to the OCTEON boot bus.
+
+	  If unsure, say N.
+
+config SATA_QSTOR
+	tristate "Pacific Digital SATA QStor support"
+	depends on PCI
+	help
+	  This option enables support for Pacific Digital Serial ATA QStor.
+
+	  If unsure, say N.
+
+config SATA_SX4
+	tristate "Promise SATA SX4 support (Experimental)"
+	depends on PCI && EXPERIMENTAL
+	help
+	  This option enables support for Promise Serial ATA SX4.
 
 	  If unsure, say N.
 
+config ATA_BMDMA
+	bool "ATA BMDMA support"
+	default y
+	help
+	  This option adds support for SFF ATA controllers with BMDMA
+	  capability.  BMDMA stands for bus-master DMA and is the
+	  de facto DMA interface for SFF controllers.
+
+	  If unsure, say Y.
+
+if ATA_BMDMA
+
+comment "SATA SFF controllers with BMDMA"
+
 config ATA_PIIX
 	tristate "Intel ESB, ICH, PIIX3, PIIX4 PATA/SATA support"
 	depends on PCI
@@ -152,22 +204,6 @@ config SATA_NV
 
 	  If unsure, say N.
 
-config PDC_ADMA
-	tristate "Pacific Digital ADMA support"
-	depends on PCI
-	help
-	  This option enables support for Pacific Digital ADMA controllers
-
-	  If unsure, say N.
-
-config SATA_QSTOR
-	tristate "Pacific Digital SATA QStor support"
-	depends on PCI
-	help
-	  This option enables support for Pacific Digital Serial ATA QStor.
-
-	  If unsure, say N.
-
 config SATA_PROMISE
 	tristate "Promise SATA TX2/TX4 support"
 	depends on PCI
@@ -176,14 +212,6 @@ config SATA_PROMISE
 
 	  If unsure, say N.
 
-config SATA_SX4
-	tristate "Promise SATA SX4 support (Experimental)"
-	depends on PCI && EXPERIMENTAL
-	help
-	  This option enables support for Promise Serial ATA SX4.
-
-	  If unsure, say N.
-
 config SATA_SIL
 	tristate "Silicon Image SATA support"
 	depends on PCI
@@ -203,6 +231,15 @@ config SATA_SIS
 	  enable the PATA_SIS driver in the config.
 	  If unsure, say N.
 
+config SATA_SVW
+	tristate "ServerWorks Frodo / Apple K2 SATA support"
+	depends on PCI
+	help
+	  This option enables support for Broadcom/Serverworks/Apple K2
+	  SATA support.
+
+	  If unsure, say N.
+
 config SATA_ULI
 	tristate "ULi Electronics SATA support"
 	depends on PCI
@@ -227,14 +264,7 @@ config SATA_VITESSE
 
 	  If unsure, say N.
 
-config PATA_ACPI
-	tristate "ACPI firmware driver for PATA"
-	depends on ATA_ACPI
-	help
-	  This option enables an ACPI method driver which drives
-	  motherboard PATA controller interfaces through the ACPI
-	  firmware in the BIOS. This driver can sometimes handle
-	  otherwise unsupported hardware.
+comment "PATA SFF controllers with BMDMA"
 
 config PATA_ALI
 	tristate "ALi PATA support"
@@ -262,40 +292,30 @@ config PATA_ARTOP
 
 	  If unsure, say N.
 
-config PATA_ATP867X
-	tristate "ARTOP/Acard ATP867X PATA support"
+config PATA_ATIIXP
+	tristate "ATI PATA support"
 	depends on PCI
 	help
-	  This option enables support for ARTOP/Acard ATP867X PATA
-	  controllers.
-
-	  If unsure, say N.
-
-config PATA_AT32
-	tristate "Atmel AVR32 PATA support (Experimental)"
-	depends on AVR32 && PLATFORM_AT32AP && EXPERIMENTAL
-	help
-	  This option enables support for the IDE devices on the
-	  Atmel AT32AP platform.
+	  This option enables support for the ATI ATA interfaces
+	  found on the many ATI chipsets.
 
 	  If unsure, say N.
 
-config PATA_ATIIXP
-	tristate "ATI PATA support"
+config PATA_ATP867X
+	tristate "ARTOP/Acard ATP867X PATA support"
 	depends on PCI
 	help
-	  This option enables support for the ATI ATA interfaces
-	  found on the many ATI chipsets.
+	  This option enables support for ARTOP/Acard ATP867X PATA
+	  controllers.
 
 	  If unsure, say N.
 
-config PATA_CMD640_PCI
-	tristate "CMD640 PCI PATA support (Experimental)"
-	depends on PCI && EXPERIMENTAL
+config PATA_BF54X
+	tristate "Blackfin 54x ATAPI support"
+	depends on BF542 || BF548 || BF549
 	help
-	  This option enables support for the CMD640 PCI IDE
-	  interface chip. Only the primary channel is currently
-	  supported.
+	  This option enables support for the built-in ATAPI controller on
+	  Blackfin 54x family chips.
 
 	  If unsure, say N.
 
@@ -362,15 +382,6 @@ config PATA_EFAR
 
 	  If unsure, say N.
 
-config ATA_GENERIC
-	tristate "Generic ATA support"
-	depends on PCI
-	help
-	  This option enables support for generic BIOS configured
-	  ATA controllers via the new ATA layer
-
-	  If unsure, say N.
-
 config PATA_HPT366
 	tristate "HPT 366/368 PATA support"
 	depends on PCI
@@ -415,12 +426,20 @@ config PATA_HPT3X3_DMA
 	  controllers. Enable with care as there are still some
 	  problems with DMA on this chipset.
 
-config PATA_ISAPNP
-	tristate "ISA Plug and Play PATA support"
-	depends on ISAPNP
+config PATA_ICSIDE
+	tristate "Acorn ICS PATA support"
+	depends on ARM && ARCH_ACORN
 	help
-	  This option enables support for ISA plug & play ATA
-	  controllers such as those found on old soundcards.
+	  On Acorn systems, say Y here if you wish to use the ICS PATA
+	  interface card.  This is not required for ICS partition support.
+	  If you are unsure, say N to this.
+
+config PATA_IT8213
+	tristate "IT8213 PATA support (Experimental)"
+	depends on PCI && EXPERIMENTAL
+	help
+	  This option enables support for the ITE 821 PATA
+          controllers via the new ATA layer.
 
 	  If unsure, say N.
 
@@ -434,15 +453,6 @@ config PATA_IT821X
 
 	  If unsure, say N.
 
-config PATA_IT8213
-	tristate "IT8213 PATA support (Experimental)"
-	depends on PCI && EXPERIMENTAL
-	help
-	  This option enables support for the ITE 821 PATA
-          controllers via the new ATA layer.
-
-	  If unsure, say N.
-
 config PATA_JMICRON
 	tristate "JMicron PATA support"
 	depends on PCI
@@ -452,23 +462,14 @@ config PATA_JMICRON
 
 	  If unsure, say N.
 
-config PATA_LEGACY
-	tristate "Legacy ISA PATA support (Experimental)"
-	depends on (ISA || PCI)  && EXPERIMENTAL
-	help
-	  This option enables support for ISA/VLB/PCI bus legacy PATA
-	  ports and allows them to be accessed via the new ATA layer.
-
-	  If unsure, say N.
-
-config PATA_TRIFLEX
-	tristate "Compaq Triflex PATA support"
-	depends on PCI
+config PATA_MACIO
+	tristate "Apple PowerMac/PowerBook internal 'MacIO' IDE"
+	depends on PPC_PMAC
 	help
-	  Enable support for the Compaq 'Triflex' IDE controller as found
-	  on many Compaq Pentium-Pro systems, via the new ATA layer.
-
-	  If unsure, say N.
+	  Most IDE capable PowerMacs have IDE busses driven by a variant
+          of this controller which is part of the Apple chipset used on
+          most PowerMac models. Some models have multiple busses using
+          different chipsets, though generally, MacIO is one of them.
 
 config PATA_MARVELL
 	tristate "Marvell PATA support via legacy mode"
@@ -481,32 +482,6 @@ config PATA_MARVELL
 
 	  If unsure, say N.
 
-config PATA_MPC52xx
-	tristate "Freescale MPC52xx SoC internal IDE"
-	depends on PPC_MPC52xx && PPC_BESTCOMM
-	select PPC_BESTCOMM_ATA
-	help
-	  This option enables support for integrated IDE controller
-	  of the Freescale MPC52xx SoC.
-
-	  If unsure, say N.
-
-config PATA_MPIIX
-	tristate "Intel PATA MPIIX support"
-	depends on PCI
-	help
-	  This option enables support for MPIIX PATA support.
-
-	  If unsure, say N.
-
-config PATA_OLDPIIX
-	tristate "Intel PATA old PIIX support"
-	depends on PCI
-	help
-	  This option enables support for early PIIX PATA support.
-
-	  If unsure, say N.
-
 config PATA_NETCELL
 	tristate "NETCELL Revolution RAID support"
 	depends on PCI
@@ -525,15 +500,6 @@ config PATA_NINJA32
 
 	  If unsure, say N.
 
-config PATA_NS87410
-	tristate "Nat Semi NS87410 PATA support"
-	depends on PCI
-	help
-	  This option enables support for the National Semiconductor
-	  NS87410 PCI-IDE controller.
-
-	  If unsure, say N.
-
 config PATA_NS87415
 	tristate "Nat Semi NS87415 PATA support"
 	depends on PCI
@@ -543,12 +509,11 @@ config PATA_NS87415
 
 	  If unsure, say N.
 
-config PATA_OPTI
-	tristate "OPTI621/6215 PATA support (Very Experimental)"
-	depends on PCI && EXPERIMENTAL
+config PATA_OLDPIIX
+	tristate "Intel PATA old PIIX support"
+	depends on PCI
 	help
-	  This option enables full PIO support for the early Opti ATA
-	  controllers found on some old motherboards.
+	  This option enables support for early PIIX PATA support.
 
 	  If unsure, say N.
 
@@ -562,24 +527,6 @@ config PATA_OPTIDMA
 
 	  If unsure, say N.
 
-config PATA_PALMLD
-	tristate "Palm LifeDrive PATA support"
-	depends on MACH_PALMLD
-	help
-	  This option enables support for Palm LifeDrive's internal ATA
-	  port via the new ATA layer.
-
-	  If unsure, say N.
-
-config PATA_PCMCIA
-	tristate "PCMCIA PATA support"
-	depends on PCMCIA
-	help
-	  This option enables support for PCMCIA ATA interfaces, including
-	  compact flash card adapters via the new ATA layer.
-
-	  If unsure, say N.
-
 config PATA_PDC2027X
 	tristate "Promise PATA 2027x support"
 	depends on PCI
@@ -597,12 +544,6 @@ config PATA_PDC_OLD
 
 	  If unsure, say N.
 
-config PATA_QDI
-	tristate "QDI VLB PATA support"
-	depends on ISA
-	help
-	  Support for QDI 6500 and 6580 PATA controllers on VESA local bus.
-
 config PATA_RADISYS
 	tristate "RADISYS 82600 PATA support (Experimental)"
 	depends on PCI && EXPERIMENTAL
@@ -612,15 +553,6 @@ config PATA_RADISYS
 
 	  If unsure, say N.
 
-config PATA_RB532
-	tristate "RouterBoard 532 PATA CompactFlash support"
-	depends on MIKROTIK_RB532
-	help
-	  This option enables support for the RouterBoard 532
-	  PATA CompactFlash controller.
-
-	  If unsure, say N.
-
 config PATA_RDC
 	tristate "RDC PATA support"
 	depends on PCI
@@ -631,21 +563,30 @@ config PATA_RDC
 
 	  If unsure, say N.
 
-config PATA_RZ1000
-	tristate "PC Tech RZ1000 PATA support"
+config PATA_SC1200
+	tristate "SC1200 PATA support"
 	depends on PCI
 	help
-	  This option enables basic support for the PC Tech RZ1000/1
-	  PATA controllers via the new ATA layer
+	  This option enables support for the NatSemi/AMD SC1200 SoC
+	  companion chip used with the Geode processor family.
 
 	  If unsure, say N.
 
-config PATA_SC1200
-	tristate "SC1200 PATA support"
+config PATA_SCC
+	tristate "Toshiba's Cell Reference Set IDE support"
+	depends on PCI && PPC_CELLEB
+	help
+	  This option enables support for the built-in IDE controller on
+	  Toshiba Cell Reference Board.
+
+	  If unsure, say N.
+
+config PATA_SCH
+	tristate "Intel SCH PATA support"
 	depends on PCI
 	help
-	  This option enables support for the NatSemi/AMD SC1200 SoC
-	  companion chip used with the Geode processor family.
+	  This option enables support for Intel SCH PATA on the Intel
+	  SCH (US15W, US15L, UL11L) series host controllers.
 
 	  If unsure, say N.
 
@@ -683,6 +624,15 @@ config PATA_TOSHIBA
 
 	  If unsure, say N.
 
+config PATA_TRIFLEX
+	tristate "Compaq Triflex PATA support"
+	depends on PCI
+	help
+	  Enable support for the Compaq 'Triflex' IDE controller as found
+	  on many Compaq Pentium-Pro systems, via the new ATA layer.
+
+	  If unsure, say N.
+
 config PATA_VIA
 	tristate "VIA PATA support"
 	depends on PCI
@@ -701,12 +651,99 @@ config PATA_WINBOND
 
 	  If unsure, say N.
 
-config PATA_WINBOND_VLB
-	tristate "Winbond W83759A VLB PATA support (Experimental)"
-	depends on ISA && EXPERIMENTAL
+endif # ATA_BMDMA
+
+comment "PIO-only SFF controllers"
+
+config PATA_AT32
+	tristate "Atmel AVR32 PATA support (Experimental)"
+	depends on AVR32 && PLATFORM_AT32AP && EXPERIMENTAL
 	help
-	  Support for the Winbond W83759A controller on Vesa Local Bus
-	  systems.
+	  This option enables support for the IDE devices on the
+	  Atmel AT32AP platform.
+
+	  If unsure, say N.
+
+config PATA_AT91
+	tristate "PATA support for AT91SAM9260"
+	depends on ARM && ARCH_AT91
+	help
+	  This option enables support for IDE devices on the Atmel AT91SAM9260 SoC.
+
+	  If unsure, say N.
+
+config PATA_CMD640_PCI
+	tristate "CMD640 PCI PATA support (Experimental)"
+	depends on PCI && EXPERIMENTAL
+	help
+	  This option enables support for the CMD640 PCI IDE
+	  interface chip. Only the primary channel is currently
+	  supported.
+
+	  If unsure, say N.
+
+config PATA_ISAPNP
+	tristate "ISA Plug and Play PATA support"
+	depends on ISAPNP
+	help
+	  This option enables support for ISA plug & play ATA
+	  controllers such as those found on old soundcards.
+
+	  If unsure, say N.
+
+config PATA_IXP4XX_CF
+	tristate "IXP4XX Compact Flash support"
+	depends on ARCH_IXP4XX
+	help
+	  This option enables support for a Compact Flash connected on
+	  the ixp4xx expansion bus. This driver had been written for
+	  Loft/Avila boards in mind but can work with others.
+
+	  If unsure, say N.
+
+config PATA_MPIIX
+	tristate "Intel PATA MPIIX support"
+	depends on PCI
+	help
+	  This option enables support for MPIIX PATA support.
+
+	  If unsure, say N.
+
+config PATA_NS87410
+	tristate "Nat Semi NS87410 PATA support"
+	depends on PCI
+	help
+	  This option enables support for the National Semiconductor
+	  NS87410 PCI-IDE controller.
+
+	  If unsure, say N.
+
+config PATA_OPTI
+	tristate "OPTI621/6215 PATA support (Very Experimental)"
+	depends on PCI && EXPERIMENTAL
+	help
+	  This option enables full PIO support for the early Opti ATA
+	  controllers found on some old motherboards.
+
+	  If unsure, say N.
+
+config PATA_PALMLD
+	tristate "Palm LifeDrive PATA support"
+	depends on MACH_PALMLD
+	help
+	  This option enables support for Palm LifeDrive's internal ATA
+	  port via the new ATA layer.
+
+	  If unsure, say N.
+
+config PATA_PCMCIA
+	tristate "PCMCIA PATA support"
+	depends on PCMCIA
+	help
+	  This option enables support for PCMCIA ATA interfaces, including
+	  compact flash card adapters via the new ATA layer.
+
+	  If unsure, say N.
 
 config HAVE_PATA_PLATFORM
 	bool
@@ -725,14 +762,6 @@ config PATA_PLATFORM
 
 	  If unsure, say N.
 
-config PATA_AT91
-	tristate "PATA support for AT91SAM9260"
-	depends on ARM && ARCH_AT91
-	help
-	  This option enables support for IDE devices on the Atmel AT91SAM9260 SoC.
-
-	  If unsure, say N.
-
 config PATA_OF_PLATFORM
 	tristate "OpenFirmware platform device PATA support"
 	depends on PATA_PLATFORM && PPC_OF
@@ -743,69 +772,65 @@ config PATA_OF_PLATFORM
 
 	  If unsure, say N.
 
-config PATA_ICSIDE
-	tristate "Acorn ICS PATA support"
-	depends on ARM && ARCH_ACORN
+config PATA_QDI
+	tristate "QDI VLB PATA support"
+	depends on ISA
 	help
-	  On Acorn systems, say Y here if you wish to use the ICS PATA
-	  interface card.  This is not required for ICS partition support.
-	  If you are unsure, say N to this.
+	  Support for QDI 6500 and 6580 PATA controllers on VESA local bus.
 
-config PATA_IXP4XX_CF
-	tristate "IXP4XX Compact Flash support"
-	depends on ARCH_IXP4XX
+config PATA_RB532
+	tristate "RouterBoard 532 PATA CompactFlash support"
+	depends on MIKROTIK_RB532
 	help
-	  This option enables support for a Compact Flash connected on
-	  the ixp4xx expansion bus. This driver had been written for
-	  Loft/Avila boards in mind but can work with others.
+	  This option enables support for the RouterBoard 532
+	  PATA CompactFlash controller.
 
 	  If unsure, say N.
 
-config PATA_OCTEON_CF
-	tristate "OCTEON Boot Bus Compact Flash support"
-	depends on CPU_CAVIUM_OCTEON
+config PATA_RZ1000
+	tristate "PC Tech RZ1000 PATA support"
+	depends on PCI
 	help
-	  This option enables a polled compact flash driver for use with
-	  compact flash cards attached to the OCTEON boot bus.
+	  This option enables basic support for the PC Tech RZ1000/1
+	  PATA controllers via the new ATA layer
 
 	  If unsure, say N.
 
-config PATA_SCC
-	tristate "Toshiba's Cell Reference Set IDE support"
-	depends on PCI && PPC_CELLEB
+config PATA_WINBOND_VLB
+	tristate "Winbond W83759A VLB PATA support (Experimental)"
+	depends on ISA && EXPERIMENTAL
 	help
-	  This option enables support for the built-in IDE controller on
-	  Toshiba Cell Reference Board.
+	  Support for the Winbond W83759A controller on Vesa Local Bus
+	  systems.
 
-	  If unsure, say N.
+comment "Generic fallback / legacy drivers"
 
-config PATA_SCH
-	tristate "Intel SCH PATA support"
-	depends on PCI
+config PATA_ACPI
+	tristate "ACPI firmware driver for PATA"
+	depends on ATA_ACPI && ATA_BMDMA
 	help
-	  This option enables support for Intel SCH PATA on the Intel
-	  SCH (US15W, US15L, UL11L) series host controllers.
-
-	  If unsure, say N.
+	  This option enables an ACPI method driver which drives
+	  motherboard PATA controller interfaces through the ACPI
+	  firmware in the BIOS. This driver can sometimes handle
+	  otherwise unsupported hardware.
 
-config PATA_BF54X
-	tristate "Blackfin 54x ATAPI support"
-	depends on BF542 || BF548 || BF549
+config ATA_GENERIC
+	tristate "Generic ATA support"
+	depends on PCI && ATA_BMDMA
 	help
-	  This option enables support for the built-in ATAPI controller on
-	  Blackfin 54x family chips.
+	  This option enables support for generic BIOS configured
+	  ATA controllers via the new ATA layer
 
 	  If unsure, say N.
 
-config PATA_MACIO
-	tristate "Apple PowerMac/PowerBook internal 'MacIO' IDE"
-	depends on PPC_PMAC
+config PATA_LEGACY
+	tristate "Legacy ISA PATA support (Experimental)"
+	depends on (ISA || PCI) && EXPERIMENTAL
 	help
-	  Most IDE capable PowerMacs have IDE busses driven by a variant
-          of this controller which is part of the Apple chipset used on
-          most PowerMac models. Some models have multiple busses using
-          different chipsets, though generally, MacIO is one of them.
+	  This option enables support for ISA/VLB/PCI bus legacy PATA
+	  ports and allows them to be accessed via the new ATA layer.
 
+	  If unsure, say N.
 
 endif # ATA_SFF
 endif # ATA
diff --git a/drivers/ata/Makefile b/drivers/ata/Makefile
index d0a93c4ad3ec..7ef89d73df63 100644
--- a/drivers/ata/Makefile
+++ b/drivers/ata/Makefile
@@ -1,33 +1,39 @@
 
 obj-$(CONFIG_ATA)		+= libata.o
 
+# non-SFF interface
 obj-$(CONFIG_SATA_AHCI)		+= ahci.o libahci.o
 obj-$(CONFIG_SATA_AHCI_PLATFORM) += ahci_platform.o libahci.o
-obj-$(CONFIG_SATA_SVW)		+= sata_svw.o
+obj-$(CONFIG_SATA_FSL)		+= sata_fsl.o
+obj-$(CONFIG_SATA_INIC162X)	+= sata_inic162x.o
+obj-$(CONFIG_SATA_SIL24)	+= sata_sil24.o
+
+# SFF w/ custom DMA
+obj-$(CONFIG_PDC_ADMA)		+= pdc_adma.o
+obj-$(CONFIG_PATA_MPC52xx)	+= pata_mpc52xx.o
+obj-$(CONFIG_PATA_OCTEON_CF)	+= pata_octeon_cf.o
+obj-$(CONFIG_SATA_QSTOR)	+= sata_qstor.o
+obj-$(CONFIG_SATA_SX4)		+= sata_sx4.o
+
+# SFF SATA w/ BMDMA
 obj-$(CONFIG_ATA_PIIX)		+= ata_piix.o
+obj-$(CONFIG_SATA_MV)		+= sata_mv.o
+obj-$(CONFIG_SATA_NV)		+= sata_nv.o
 obj-$(CONFIG_SATA_PROMISE)	+= sata_promise.o
-obj-$(CONFIG_SATA_QSTOR)	+= sata_qstor.o
 obj-$(CONFIG_SATA_SIL)		+= sata_sil.o
-obj-$(CONFIG_SATA_SIL24)	+= sata_sil24.o
-obj-$(CONFIG_SATA_VIA)		+= sata_via.o
-obj-$(CONFIG_SATA_VITESSE)	+= sata_vsc.o
 obj-$(CONFIG_SATA_SIS)		+= sata_sis.o
-obj-$(CONFIG_SATA_SX4)		+= sata_sx4.o
-obj-$(CONFIG_SATA_NV)		+= sata_nv.o
+obj-$(CONFIG_SATA_SVW)		+= sata_svw.o
 obj-$(CONFIG_SATA_ULI)		+= sata_uli.o
-obj-$(CONFIG_SATA_MV)		+= sata_mv.o
-obj-$(CONFIG_SATA_INIC162X)	+= sata_inic162x.o
-obj-$(CONFIG_PDC_ADMA)		+= pdc_adma.o
-obj-$(CONFIG_SATA_FSL)		+= sata_fsl.o
-obj-$(CONFIG_PATA_MACIO)	+= pata_macio.o
+obj-$(CONFIG_SATA_VIA)		+= sata_via.o
+obj-$(CONFIG_SATA_VITESSE)	+= sata_vsc.o
 
+# SFF PATA w/ BMDMA
 obj-$(CONFIG_PATA_ALI)		+= pata_ali.o
 obj-$(CONFIG_PATA_AMD)		+= pata_amd.o
 obj-$(CONFIG_PATA_ARTOP)	+= pata_artop.o
-obj-$(CONFIG_PATA_ATP867X)	+= pata_atp867x.o
-obj-$(CONFIG_PATA_AT32)		+= pata_at32.o
 obj-$(CONFIG_PATA_ATIIXP)	+= pata_atiixp.o
-obj-$(CONFIG_PATA_CMD640_PCI)	+= pata_cmd640.o
+obj-$(CONFIG_PATA_ATP867X)	+= pata_atp867x.o
+obj-$(CONFIG_PATA_BF54X)	+= pata_bf54x.o
 obj-$(CONFIG_PATA_CMD64X)	+= pata_cmd64x.o
 obj-$(CONFIG_PATA_CS5520)	+= pata_cs5520.o
 obj-$(CONFIG_PATA_CS5530)	+= pata_cs5530.o
@@ -39,47 +45,50 @@ obj-$(CONFIG_PATA_HPT366)	+= pata_hpt366.o
 obj-$(CONFIG_PATA_HPT37X)	+= pata_hpt37x.o
 obj-$(CONFIG_PATA_HPT3X2N)	+= pata_hpt3x2n.o
 obj-$(CONFIG_PATA_HPT3X3)	+= pata_hpt3x3.o
-obj-$(CONFIG_PATA_ISAPNP)	+= pata_isapnp.o
-obj-$(CONFIG_PATA_IT821X)	+= pata_it821x.o
+obj-$(CONFIG_PATA_ICSIDE)	+= pata_icside.o
 obj-$(CONFIG_PATA_IT8213)	+= pata_it8213.o
+obj-$(CONFIG_PATA_IT821X)	+= pata_it821x.o
 obj-$(CONFIG_PATA_JMICRON)	+= pata_jmicron.o
+obj-$(CONFIG_PATA_MACIO)	+= pata_macio.o
+obj-$(CONFIG_PATA_MARVELL)	+= pata_marvell.o
 obj-$(CONFIG_PATA_NETCELL)	+= pata_netcell.o
 obj-$(CONFIG_PATA_NINJA32)	+= pata_ninja32.o
-obj-$(CONFIG_PATA_NS87410)	+= pata_ns87410.o
 obj-$(CONFIG_PATA_NS87415)	+= pata_ns87415.o
-obj-$(CONFIG_PATA_OPTI)		+= pata_opti.o
-obj-$(CONFIG_PATA_OPTIDMA)	+= pata_optidma.o
-obj-$(CONFIG_PATA_MPC52xx)	+= pata_mpc52xx.o
-obj-$(CONFIG_PATA_MARVELL)	+= pata_marvell.o
-obj-$(CONFIG_PATA_MPIIX)	+= pata_mpiix.o
 obj-$(CONFIG_PATA_OLDPIIX)	+= pata_oldpiix.o
-obj-$(CONFIG_PATA_PALMLD)	+= pata_palmld.o
-obj-$(CONFIG_PATA_PCMCIA)	+= pata_pcmcia.o
+obj-$(CONFIG_PATA_OPTIDMA)	+= pata_optidma.o
 obj-$(CONFIG_PATA_PDC2027X)	+= pata_pdc2027x.o
 obj-$(CONFIG_PATA_PDC_OLD)	+= pata_pdc202xx_old.o
-obj-$(CONFIG_PATA_QDI)		+= pata_qdi.o
 obj-$(CONFIG_PATA_RADISYS)	+= pata_radisys.o
-obj-$(CONFIG_PATA_RB532)	+= pata_rb532_cf.o
 obj-$(CONFIG_PATA_RDC)		+= pata_rdc.o
-obj-$(CONFIG_PATA_RZ1000)	+= pata_rz1000.o
 obj-$(CONFIG_PATA_SC1200)	+= pata_sc1200.o
+obj-$(CONFIG_PATA_SCC)		+= pata_scc.o
+obj-$(CONFIG_PATA_SCH)		+= pata_sch.o
 obj-$(CONFIG_PATA_SERVERWORKS)	+= pata_serverworks.o
 obj-$(CONFIG_PATA_SIL680)	+= pata_sil680.o
+obj-$(CONFIG_PATA_SIS)		+= pata_sis.o
 obj-$(CONFIG_PATA_TOSHIBA)	+= pata_piccolo.o
+obj-$(CONFIG_PATA_TRIFLEX)	+= pata_triflex.o
 obj-$(CONFIG_PATA_VIA)		+= pata_via.o
 obj-$(CONFIG_PATA_WINBOND)	+= pata_sl82c105.o
-obj-$(CONFIG_PATA_WINBOND_VLB)	+= pata_winbond.o
-obj-$(CONFIG_PATA_SIS)		+= pata_sis.o
-obj-$(CONFIG_PATA_TRIFLEX)	+= pata_triflex.o
+
+# SFF PIO only
+obj-$(CONFIG_PATA_AT32)		+= pata_at32.o
+obj-$(CONFIG_PATA_AT91)		+= pata_at91.o
+obj-$(CONFIG_PATA_CMD640_PCI)	+= pata_cmd640.o
+obj-$(CONFIG_PATA_ISAPNP)	+= pata_isapnp.o
 obj-$(CONFIG_PATA_IXP4XX_CF)	+= pata_ixp4xx_cf.o
-obj-$(CONFIG_PATA_SCC)		+= pata_scc.o
-obj-$(CONFIG_PATA_SCH)		+= pata_sch.o
-obj-$(CONFIG_PATA_BF54X)	+= pata_bf54x.o
-obj-$(CONFIG_PATA_OCTEON_CF)	+= pata_octeon_cf.o
+obj-$(CONFIG_PATA_MPIIX)	+= pata_mpiix.o
+obj-$(CONFIG_PATA_NS87410)	+= pata_ns87410.o
+obj-$(CONFIG_PATA_OPTI)		+= pata_opti.o
+obj-$(CONFIG_PATA_PCMCIA)	+= pata_pcmcia.o
+obj-$(CONFIG_PATA_PALMLD)	+= pata_palmld.o
 obj-$(CONFIG_PATA_PLATFORM)	+= pata_platform.o
-obj-$(CONFIG_PATA_AT91)	+= pata_at91.o
 obj-$(CONFIG_PATA_OF_PLATFORM)	+= pata_of_platform.o
-obj-$(CONFIG_PATA_ICSIDE)	+= pata_icside.o
+obj-$(CONFIG_PATA_QDI)		+= pata_qdi.o
+obj-$(CONFIG_PATA_RB532)	+= pata_rb532_cf.o
+obj-$(CONFIG_PATA_RZ1000)	+= pata_rz1000.o
+obj-$(CONFIG_PATA_WINBOND_VLB)	+= pata_winbond.o
+
 # Should be last but two libata driver
 obj-$(CONFIG_PATA_ACPI)		+= pata_acpi.o
 # Should be last but one libata driver
diff --git a/drivers/ata/ata_generic.c b/drivers/ata/ata_generic.c
index 33fb614f9784..573158a9668d 100644
--- a/drivers/ata/ata_generic.c
+++ b/drivers/ata/ata_generic.c
@@ -155,7 +155,7 @@ static int ata_generic_init_one(struct pci_dev *dev, const struct pci_device_id
 			return rc;
 		pcim_pin_device(dev);
 	}
-	return ata_pci_sff_init_one(dev, ppi, &generic_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(dev, ppi, &generic_sht, NULL, 0);
 }
 
 static struct pci_device_id ata_generic[] = {
diff --git a/drivers/ata/ata_piix.c b/drivers/ata/ata_piix.c
index ec52fc618763..7409f98d2ae6 100644
--- a/drivers/ata/ata_piix.c
+++ b/drivers/ata/ata_piix.c
@@ -1589,7 +1589,7 @@ static int __devinit piix_init_one(struct pci_dev *pdev,
 		hpriv->map = piix_init_sata_map(pdev, port_info,
 					piix_map_db_table[ent->driver_data]);
 
-	rc = ata_pci_sff_prepare_host(pdev, ppi, &host);
+	rc = ata_pci_bmdma_prepare_host(pdev, ppi, &host);
 	if (rc)
 		return rc;
 	host->private_data = hpriv;
@@ -1626,7 +1626,7 @@ static int __devinit piix_init_one(struct pci_dev *pdev,
 	host->flags |= ATA_HOST_PARALLEL_SCAN;
 
 	pci_set_master(pdev);
-	return ata_pci_sff_activate_host(host, ata_sff_interrupt, &piix_sht);
+	return ata_pci_sff_activate_host(host, ata_bmdma_interrupt, &piix_sht);
 }
 
 static void piix_remove_one(struct pci_dev *pdev)
diff --git a/drivers/ata/libahci.c b/drivers/ata/libahci.c
index 1984a6e89e84..261f86d102e8 100644
--- a/drivers/ata/libahci.c
+++ b/drivers/ata/libahci.c
@@ -541,29 +541,11 @@ static int ahci_scr_write(struct ata_link *link, unsigned int sc_reg, u32 val)
 	return -EINVAL;
 }
 
-static int ahci_is_device_present(void __iomem *port_mmio)
-{
-	u8 status = readl(port_mmio + PORT_TFDATA) & 0xff;
-
-	/* Make sure PxTFD.STS.BSY and PxTFD.STS.DRQ are 0 */
-	if (status & (ATA_BUSY | ATA_DRQ))
-		return 0;
-
-	/* Make sure PxSSTS.DET is 3h */
-	status = readl(port_mmio + PORT_SCR_STAT) & 0xf;
-	if (status != 3)
-		return 0;
-	return 1;
-}
-
 void ahci_start_engine(struct ata_port *ap)
 {
 	void __iomem *port_mmio = ahci_port_base(ap);
 	u32 tmp;
 
-	if (!ahci_is_device_present(port_mmio))
-		return;
-
 	/* start DMA */
 	tmp = readl(port_mmio + PORT_CMD);
 	tmp |= PORT_CMD_START;
@@ -1892,6 +1874,9 @@ static void ahci_error_handler(struct ata_port *ap)
 	}
 
 	sata_pmp_error_handler(ap);
+
+	if (!ata_dev_enabled(ap->link.device))
+		ahci_stop_engine(ap);
 }
 
 static void ahci_post_internal_cmd(struct ata_queued_cmd *qc)
diff --git a/drivers/ata/libata-core.c b/drivers/ata/libata-core.c
index c47373f01f89..ddf8e4862787 100644
--- a/drivers/ata/libata-core.c
+++ b/drivers/ata/libata-core.c
@@ -160,6 +160,10 @@ int libata_allow_tpm = 0;
 module_param_named(allow_tpm, libata_allow_tpm, int, 0444);
 MODULE_PARM_DESC(allow_tpm, "Permit the use of TPM commands (0=off [default], 1=on)");
 
+static int atapi_an;
+module_param(atapi_an, int, 0444);
+MODULE_PARM_DESC(atapi_an, "Enable ATAPI AN media presence notification (0=0ff [default], 1=on)");
+
 MODULE_AUTHOR("Jeff Garzik");
 MODULE_DESCRIPTION("Library module for ATA devices");
 MODULE_LICENSE("GPL");
@@ -2122,6 +2126,14 @@ retry:
 		goto err_out;
 	}
 
+	if (dev->horkage & ATA_HORKAGE_DUMP_ID) {
+		ata_dev_printk(dev, KERN_DEBUG, "dumping IDENTIFY data, "
+			       "class=%d may_fallback=%d tried_spinup=%d\n",
+			       class, may_fallback, tried_spinup);
+		print_hex_dump(KERN_DEBUG, "", DUMP_PREFIX_OFFSET,
+			       16, 2, id, ATA_ID_WORDS * sizeof(*id), true);
+	}
+
 	/* Falling back doesn't make sense if ID data was read
 	 * successfully at least once.
 	 */
@@ -2510,7 +2522,8 @@ int ata_dev_configure(struct ata_device *dev)
 		 * to enable ATAPI AN to discern between PHY status
 		 * changed notifications and ATAPI ANs.
 		 */
-		if ((ap->flags & ATA_FLAG_AN) && ata_id_has_atapi_AN(id) &&
+		if (atapi_an &&
+		    (ap->flags & ATA_FLAG_AN) && ata_id_has_atapi_AN(id) &&
 		    (!sata_pmp_attached(ap) ||
 		     sata_scr_read(&ap->link, SCR_NOTIFICATION, &sntf) == 0)) {
 			unsigned int err_mask;
@@ -4106,9 +4119,8 @@ int ata_dev_revalidate(struct ata_device *dev, unsigned int new_class,
 	    dev->n_sectors > n_sectors && dev->n_sectors == n_native_sectors) {
 		ata_dev_printk(dev, KERN_WARNING,
 			       "new n_sectors matches native, probably "
-			       "late HPA unlock, continuing\n");
-		/* keep using the old n_sectors */
-		dev->n_sectors = n_sectors;
+			       "late HPA unlock, n_sectors updated\n");
+		/* use the larger n_sectors */
 		return 0;
 	}
 
@@ -6372,6 +6384,7 @@ static int __init ata_parse_force_one(char **cur,
 		{ "3.0Gbps",	.spd_limit	= 2 },
 		{ "noncq",	.horkage_on	= ATA_HORKAGE_NONCQ },
 		{ "ncq",	.horkage_off	= ATA_HORKAGE_NONCQ },
+		{ "dump_id",	.horkage_on	= ATA_HORKAGE_DUMP_ID },
 		{ "pio0",	.xfer_mask	= 1 << (ATA_SHIFT_PIO + 0) },
 		{ "pio1",	.xfer_mask	= 1 << (ATA_SHIFT_PIO + 1) },
 		{ "pio2",	.xfer_mask	= 1 << (ATA_SHIFT_PIO + 2) },
@@ -6655,6 +6668,7 @@ EXPORT_SYMBOL_GPL(ata_dummy_port_info);
 EXPORT_SYMBOL_GPL(ata_link_next);
 EXPORT_SYMBOL_GPL(ata_dev_next);
 EXPORT_SYMBOL_GPL(ata_std_bios_param);
+EXPORT_SYMBOL_GPL(ata_scsi_unlock_native_capacity);
 EXPORT_SYMBOL_GPL(ata_host_init);
 EXPORT_SYMBOL_GPL(ata_host_alloc);
 EXPORT_SYMBOL_GPL(ata_host_alloc_pinfo);
diff --git a/drivers/ata/libata-scsi.c b/drivers/ata/libata-scsi.c
index cfa9dd3d7253..a54273d2c3c6 100644
--- a/drivers/ata/libata-scsi.c
+++ b/drivers/ata/libata-scsi.c
@@ -415,6 +415,35 @@ int ata_std_bios_param(struct scsi_device *sdev, struct block_device *bdev,
 }
 
 /**
+ *	ata_scsi_unlock_native_capacity - unlock native capacity
+ *	@sdev: SCSI device to adjust device capacity for
+ *
+ *	This function is called if a partition on @sdev extends beyond
+ *	the end of the device.  It requests EH to unlock HPA.
+ *
+ *	LOCKING:
+ *	Defined by the SCSI layer.  Might sleep.
+ */
+void ata_scsi_unlock_native_capacity(struct scsi_device *sdev)
+{
+	struct ata_port *ap = ata_shost_to_port(sdev->host);
+	struct ata_device *dev;
+	unsigned long flags;
+
+	spin_lock_irqsave(ap->lock, flags);
+
+	dev = ata_scsi_find_dev(ap, sdev);
+	if (dev && dev->n_sectors < dev->n_native_sectors) {
+		dev->flags |= ATA_DFLAG_UNLOCK_HPA;
+		dev->link->eh_info.action |= ATA_EH_RESET;
+		ata_port_schedule_eh(ap);
+	}
+
+	spin_unlock_irqrestore(ap->lock, flags);
+	ata_port_wait_eh(ap);
+}
+
+/**
  *	ata_get_identity - Handler for HDIO_GET_IDENTITY ioctl
  *	@ap: target port
  *	@sdev: SCSI device to get identify data for
diff --git a/drivers/ata/libata-sff.c b/drivers/ata/libata-sff.c
index 19ddf924944f..efa4a18cfb9d 100644
--- a/drivers/ata/libata-sff.c
+++ b/drivers/ata/libata-sff.c
@@ -63,7 +63,6 @@ const struct ata_port_operations ata_sff_port_ops = {
 	.sff_tf_read		= ata_sff_tf_read,
 	.sff_exec_command	= ata_sff_exec_command,
 	.sff_data_xfer		= ata_sff_data_xfer,
-	.sff_irq_clear		= ata_sff_irq_clear,
 	.sff_drain_fifo		= ata_sff_drain_fifo,
 
 	.lost_interrupt		= ata_sff_lost_interrupt,
@@ -395,33 +394,12 @@ void ata_sff_irq_on(struct ata_port *ap)
 		ata_sff_set_devctl(ap, ap->ctl);
 	ata_wait_idle(ap);
 
-	ap->ops->sff_irq_clear(ap);
+	if (ap->ops->sff_irq_clear)
+		ap->ops->sff_irq_clear(ap);
 }
 EXPORT_SYMBOL_GPL(ata_sff_irq_on);
 
 /**
- *	ata_sff_irq_clear - Clear PCI IDE BMDMA interrupt.
- *	@ap: Port associated with this ATA transaction.
- *
- *	Clear interrupt and error flags in DMA status register.
- *
- *	May be used as the irq_clear() entry in ata_port_operations.
- *
- *	LOCKING:
- *	spin_lock_irqsave(host lock)
- */
-void ata_sff_irq_clear(struct ata_port *ap)
-{
-	void __iomem *mmio = ap->ioaddr.bmdma_addr;
-
-	if (!mmio)
-		return;
-
-	iowrite8(ioread8(mmio + ATA_DMA_STATUS), mmio + ATA_DMA_STATUS);
-}
-EXPORT_SYMBOL_GPL(ata_sff_irq_clear);
-
-/**
  *	ata_sff_tf_load - send taskfile registers to host controller
  *	@ap: Port to which output is sent
  *	@tf: ATA taskfile register set
@@ -820,11 +798,15 @@ static void atapi_send_cdb(struct ata_port *ap, struct ata_queued_cmd *qc)
 	case ATAPI_PROT_NODATA:
 		ap->hsm_task_state = HSM_ST_LAST;
 		break;
+#ifdef CONFIG_ATA_BMDMA
 	case ATAPI_PROT_DMA:
 		ap->hsm_task_state = HSM_ST_LAST;
 		/* initiate bmdma */
 		ap->ops->bmdma_start(qc);
 		break;
+#endif /* CONFIG_ATA_BMDMA */
+	default:
+		BUG();
 	}
 }
 
@@ -1491,27 +1473,27 @@ bool ata_sff_qc_fill_rtf(struct ata_queued_cmd *qc)
 }
 EXPORT_SYMBOL_GPL(ata_sff_qc_fill_rtf);
 
-/**
- *	ata_sff_host_intr - Handle host interrupt for given (port, task)
- *	@ap: Port on which interrupt arrived (possibly...)
- *	@qc: Taskfile currently active in engine
- *
- *	Handle host interrupt for given queued command.  Currently,
- *	only DMA interrupts are handled.  All other commands are
- *	handled via polling with interrupts disabled (nIEN bit).
- *
- *	LOCKING:
- *	spin_lock_irqsave(host lock)
- *
- *	RETURNS:
- *	One if interrupt was handled, zero if not (shared irq).
- */
-unsigned int ata_sff_host_intr(struct ata_port *ap,
-				      struct ata_queued_cmd *qc)
+static unsigned int ata_sff_idle_irq(struct ata_port *ap)
 {
-	struct ata_eh_info *ehi = &ap->link.eh_info;
-	u8 status, host_stat = 0;
-	bool bmdma_stopped = false;
+	ap->stats.idle_irq++;
+
+#ifdef ATA_IRQ_TRAP
+	if ((ap->stats.idle_irq % 1000) == 0) {
+		ap->ops->sff_check_status(ap);
+		if (ap->ops->sff_irq_clear)
+			ap->ops->sff_irq_clear(ap);
+		ata_port_printk(ap, KERN_WARNING, "irq trap\n");
+		return 1;
+	}
+#endif
+	return 0;	/* irq not handled */
+}
+
+static unsigned int __ata_sff_port_intr(struct ata_port *ap,
+					struct ata_queued_cmd *qc,
+					bool hsmv_on_idle)
+{
+	u8 status;
 
 	VPRINTK("ata%u: protocol %d task_state %d\n",
 		ap->print_id, qc->tf.protocol, ap->hsm_task_state);
@@ -1528,90 +1510,56 @@ unsigned int ata_sff_host_intr(struct ata_port *ap,
 		 * need to check ata_is_atapi(qc->tf.protocol) again.
 		 */
 		if (!(qc->dev->flags & ATA_DFLAG_CDB_INTR))
-			goto idle_irq;
-		break;
-	case HSM_ST_LAST:
-		if (qc->tf.protocol == ATA_PROT_DMA ||
-		    qc->tf.protocol == ATAPI_PROT_DMA) {
-			/* check status of DMA engine */
-			host_stat = ap->ops->bmdma_status(ap);
-			VPRINTK("ata%u: host_stat 0x%X\n",
-				ap->print_id, host_stat);
-
-			/* if it's not our irq... */
-			if (!(host_stat & ATA_DMA_INTR))
-				goto idle_irq;
-
-			/* before we do anything else, clear DMA-Start bit */
-			ap->ops->bmdma_stop(qc);
-			bmdma_stopped = true;
-
-			if (unlikely(host_stat & ATA_DMA_ERR)) {
-				/* error when transfering data to/from memory */
-				qc->err_mask |= AC_ERR_HOST_BUS;
-				ap->hsm_task_state = HSM_ST_ERR;
-			}
-		}
+			return ata_sff_idle_irq(ap);
 		break;
 	case HSM_ST:
+	case HSM_ST_LAST:
 		break;
 	default:
-		goto idle_irq;
+		return ata_sff_idle_irq(ap);
 	}
 
-
 	/* check main status, clearing INTRQ if needed */
 	status = ata_sff_irq_status(ap);
 	if (status & ATA_BUSY) {
-		if (bmdma_stopped) {
+		if (hsmv_on_idle) {
 			/* BMDMA engine is already stopped, we're screwed */
 			qc->err_mask |= AC_ERR_HSM;
 			ap->hsm_task_state = HSM_ST_ERR;
 		} else
-			goto idle_irq;
+			return ata_sff_idle_irq(ap);
 	}
 
 	/* clear irq events */
-	ap->ops->sff_irq_clear(ap);
+	if (ap->ops->sff_irq_clear)
+		ap->ops->sff_irq_clear(ap);
 
 	ata_sff_hsm_move(ap, qc, status, 0);
 
-	if (unlikely(qc->err_mask) && (qc->tf.protocol == ATA_PROT_DMA ||
-				       qc->tf.protocol == ATAPI_PROT_DMA))
-		ata_ehi_push_desc(ehi, "BMDMA stat 0x%x", host_stat);
-
 	return 1;	/* irq handled */
-
-idle_irq:
-	ap->stats.idle_irq++;
-
-#ifdef ATA_IRQ_TRAP
-	if ((ap->stats.idle_irq % 1000) == 0) {
-		ap->ops->sff_check_status(ap);
-		ap->ops->sff_irq_clear(ap);
-		ata_port_printk(ap, KERN_WARNING, "irq trap\n");
-		return 1;
-	}
-#endif
-	return 0;	/* irq not handled */
 }
-EXPORT_SYMBOL_GPL(ata_sff_host_intr);
 
 /**
- *	ata_sff_interrupt - Default ATA host interrupt handler
- *	@irq: irq line (unused)
- *	@dev_instance: pointer to our ata_host information structure
+ *	ata_sff_port_intr - Handle SFF port interrupt
+ *	@ap: Port on which interrupt arrived (possibly...)
+ *	@qc: Taskfile currently active in engine
  *
- *	Default interrupt handler for PCI IDE devices.  Calls
- *	ata_sff_host_intr() for each port that is not disabled.
+ *	Handle port interrupt for given queued command.
  *
  *	LOCKING:
- *	Obtains host lock during operation.
+ *	spin_lock_irqsave(host lock)
  *
  *	RETURNS:
- *	IRQ_NONE or IRQ_HANDLED.
+ *	One if interrupt was handled, zero if not (shared irq).
  */
-irqreturn_t ata_sff_interrupt(int irq, void *dev_instance)
+unsigned int ata_sff_port_intr(struct ata_port *ap, struct ata_queued_cmd *qc)
+{
+	return __ata_sff_port_intr(ap, qc, false);
+}
+EXPORT_SYMBOL_GPL(ata_sff_port_intr);
+
+static inline irqreturn_t __ata_sff_interrupt(int irq, void *dev_instance,
+	unsigned int (*port_intr)(struct ata_port *, struct ata_queued_cmd *))
 {
 	struct ata_host *host = dev_instance;
 	bool retried = false;
@@ -1631,7 +1579,7 @@ retry:
 		qc = ata_qc_from_tag(ap, ap->link.active_tag);
 		if (qc) {
 			if (!(qc->tf.flags & ATA_TFLAG_POLLING))
-				handled |= ata_sff_host_intr(ap, qc);
+				handled |= port_intr(ap, qc);
 			else
 				polling |= 1 << i;
 		} else
@@ -1658,7 +1606,8 @@ retry:
 
 			if (idle & (1 << i)) {
 				ap->ops->sff_check_status(ap);
-				ap->ops->sff_irq_clear(ap);
+				if (ap->ops->sff_irq_clear)
+					ap->ops->sff_irq_clear(ap);
 			} else {
 				/* clear INTRQ and check if BUSY cleared */
 				if (!(ap->ops->sff_check_status(ap) & ATA_BUSY))
@@ -1680,6 +1629,25 @@ retry:
 
 	return IRQ_RETVAL(handled);
 }
+
+/**
+ *	ata_sff_interrupt - Default SFF ATA host interrupt handler
+ *	@irq: irq line (unused)
+ *	@dev_instance: pointer to our ata_host information structure
+ *
+ *	Default interrupt handler for PCI IDE devices.  Calls
+ *	ata_sff_port_intr() for each port that is not disabled.
+ *
+ *	LOCKING:
+ *	Obtains host lock during operation.
+ *
+ *	RETURNS:
+ *	IRQ_NONE or IRQ_HANDLED.
+ */
+irqreturn_t ata_sff_interrupt(int irq, void *dev_instance)
+{
+	return __ata_sff_interrupt(irq, dev_instance, ata_sff_port_intr);
+}
 EXPORT_SYMBOL_GPL(ata_sff_interrupt);
 
 /**
@@ -1717,7 +1685,7 @@ void ata_sff_lost_interrupt(struct ata_port *ap)
 								status);
 	/* Run the host interrupt logic as if the interrupt had not been
 	   lost */
-	ata_sff_host_intr(ap, qc);
+	ata_sff_port_intr(ap, qc);
 }
 EXPORT_SYMBOL_GPL(ata_sff_lost_interrupt);
 
@@ -1744,7 +1712,8 @@ void ata_sff_freeze(struct ata_port *ap)
 	 */
 	ap->ops->sff_check_status(ap);
 
-	ap->ops->sff_irq_clear(ap);
+	if (ap->ops->sff_irq_clear)
+		ap->ops->sff_irq_clear(ap);
 }
 EXPORT_SYMBOL_GPL(ata_sff_freeze);
 
@@ -1761,7 +1730,8 @@ void ata_sff_thaw(struct ata_port *ap)
 {
 	/* clear & re-enable interrupts */
 	ap->ops->sff_check_status(ap);
-	ap->ops->sff_irq_clear(ap);
+	if (ap->ops->sff_irq_clear)
+		ap->ops->sff_irq_clear(ap);
 	ata_sff_irq_on(ap);
 }
 EXPORT_SYMBOL_GPL(ata_sff_thaw);
@@ -2349,13 +2319,13 @@ int ata_pci_sff_init_host(struct ata_host *host)
 EXPORT_SYMBOL_GPL(ata_pci_sff_init_host);
 
 /**
- *	ata_pci_sff_prepare_host - helper to prepare native PCI ATA host
+ *	ata_pci_sff_prepare_host - helper to prepare PCI PIO-only SFF ATA host
  *	@pdev: target PCI device
  *	@ppi: array of port_info, must be enough for two ports
  *	@r_host: out argument for the initialized ATA host
  *
- *	Helper to allocate ATA host for @pdev, acquire all native PCI
- *	resources and initialize it accordingly in one go.
+ *	Helper to allocate PIO-only SFF ATA host for @pdev, acquire
+ *	all PCI resources and initialize it accordingly in one go.
  *
  *	LOCKING:
  *	Inherited from calling layer (may sleep).
@@ -2385,9 +2355,6 @@ int ata_pci_sff_prepare_host(struct pci_dev *pdev,
 	if (rc)
 		goto err_out;
 
-	/* init DMA related stuff */
-	ata_pci_bmdma_init(host);
-
 	devres_remove_group(&pdev->dev, NULL);
 	*r_host = host;
 	return 0;
@@ -2492,8 +2459,21 @@ out:
 }
 EXPORT_SYMBOL_GPL(ata_pci_sff_activate_host);
 
+static const struct ata_port_info *ata_sff_find_valid_pi(
+					const struct ata_port_info * const *ppi)
+{
+	int i;
+
+	/* look up the first valid port_info */
+	for (i = 0; i < 2 && ppi[i]; i++)
+		if (ppi[i]->port_ops != &ata_dummy_port_ops)
+			return ppi[i];
+
+	return NULL;
+}
+
 /**
- *	ata_pci_sff_init_one - Initialize/register PCI IDE host controller
+ *	ata_pci_sff_init_one - Initialize/register PIO-only PCI IDE controller
  *	@pdev: Controller to be initialized
  *	@ppi: array of port_info, must be enough for two ports
  *	@sht: scsi_host_template to use when registering the host
@@ -2502,11 +2482,7 @@ EXPORT_SYMBOL_GPL(ata_pci_sff_activate_host);
  *
  *	This is a helper function which can be called from a driver's
  *	xxx_init_one() probe function if the hardware uses traditional
- *	IDE taskfile registers.
- *
- *	This function calls pci_enable_device(), reserves its register
- *	regions, sets the dma mask, enables bus master mode, and calls
- *	ata_device_add()
+ *	IDE taskfile registers and is PIO only.
  *
  *	ASSUMPTION:
  *	Nobody makes a single channel controller that appears solely as
@@ -2523,20 +2499,13 @@ int ata_pci_sff_init_one(struct pci_dev *pdev,
 		 struct scsi_host_template *sht, void *host_priv, int hflag)
 {
 	struct device *dev = &pdev->dev;
-	const struct ata_port_info *pi = NULL;
+	const struct ata_port_info *pi;
 	struct ata_host *host = NULL;
-	int i, rc;
+	int rc;
 
 	DPRINTK("ENTER\n");
 
-	/* look up the first valid port_info */
-	for (i = 0; i < 2 && ppi[i]; i++) {
-		if (ppi[i]->port_ops != &ata_dummy_port_ops) {
-			pi = ppi[i];
-			break;
-		}
-	}
-
+	pi = ata_sff_find_valid_pi(ppi);
 	if (!pi) {
 		dev_printk(KERN_ERR, &pdev->dev,
 			   "no valid port_info specified\n");
@@ -2557,7 +2526,6 @@ int ata_pci_sff_init_one(struct pci_dev *pdev,
 	host->private_data = host_priv;
 	host->flags |= hflag;
 
-	pci_set_master(pdev);
 	rc = ata_pci_sff_activate_host(host, ata_sff_interrupt, sht);
 out:
 	if (rc == 0)
@@ -2571,6 +2539,12 @@ EXPORT_SYMBOL_GPL(ata_pci_sff_init_one);
 
 #endif /* CONFIG_PCI */
 
+/*
+ *	BMDMA support
+ */
+
+#ifdef CONFIG_ATA_BMDMA
+
 const struct ata_port_operations ata_bmdma_port_ops = {
 	.inherits		= &ata_sff_port_ops,
 
@@ -2580,6 +2554,7 @@ const struct ata_port_operations ata_bmdma_port_ops = {
 	.qc_prep		= ata_bmdma_qc_prep,
 	.qc_issue		= ata_bmdma_qc_issue,
 
+	.sff_irq_clear		= ata_bmdma_irq_clear,
 	.bmdma_setup		= ata_bmdma_setup,
 	.bmdma_start		= ata_bmdma_start,
 	.bmdma_stop		= ata_bmdma_stop,
@@ -2804,6 +2779,75 @@ unsigned int ata_bmdma_qc_issue(struct ata_queued_cmd *qc)
 EXPORT_SYMBOL_GPL(ata_bmdma_qc_issue);
 
 /**
+ *	ata_bmdma_port_intr - Handle BMDMA port interrupt
+ *	@ap: Port on which interrupt arrived (possibly...)
+ *	@qc: Taskfile currently active in engine
+ *
+ *	Handle port interrupt for given queued command.
+ *
+ *	LOCKING:
+ *	spin_lock_irqsave(host lock)
+ *
+ *	RETURNS:
+ *	One if interrupt was handled, zero if not (shared irq).
+ */
+unsigned int ata_bmdma_port_intr(struct ata_port *ap, struct ata_queued_cmd *qc)
+{
+	struct ata_eh_info *ehi = &ap->link.eh_info;
+	u8 host_stat = 0;
+	bool bmdma_stopped = false;
+	unsigned int handled;
+
+	if (ap->hsm_task_state == HSM_ST_LAST && ata_is_dma(qc->tf.protocol)) {
+		/* check status of DMA engine */
+		host_stat = ap->ops->bmdma_status(ap);
+		VPRINTK("ata%u: host_stat 0x%X\n", ap->print_id, host_stat);
+
+		/* if it's not our irq... */
+		if (!(host_stat & ATA_DMA_INTR))
+			return ata_sff_idle_irq(ap);
+
+		/* before we do anything else, clear DMA-Start bit */
+		ap->ops->bmdma_stop(qc);
+		bmdma_stopped = true;
+
+		if (unlikely(host_stat & ATA_DMA_ERR)) {
+			/* error when transfering data to/from memory */
+			qc->err_mask |= AC_ERR_HOST_BUS;
+			ap->hsm_task_state = HSM_ST_ERR;
+		}
+	}
+
+	handled = __ata_sff_port_intr(ap, qc, bmdma_stopped);
+
+	if (unlikely(qc->err_mask) && ata_is_dma(qc->tf.protocol))
+		ata_ehi_push_desc(ehi, "BMDMA stat 0x%x", host_stat);
+
+	return handled;
+}
+EXPORT_SYMBOL_GPL(ata_bmdma_port_intr);
+
+/**
+ *	ata_bmdma_interrupt - Default BMDMA ATA host interrupt handler
+ *	@irq: irq line (unused)
+ *	@dev_instance: pointer to our ata_host information structure
+ *
+ *	Default interrupt handler for PCI IDE devices.  Calls
+ *	ata_bmdma_port_intr() for each port that is not disabled.
+ *
+ *	LOCKING:
+ *	Obtains host lock during operation.
+ *
+ *	RETURNS:
+ *	IRQ_NONE or IRQ_HANDLED.
+ */
+irqreturn_t ata_bmdma_interrupt(int irq, void *dev_instance)
+{
+	return __ata_sff_interrupt(irq, dev_instance, ata_bmdma_port_intr);
+}
+EXPORT_SYMBOL_GPL(ata_bmdma_interrupt);
+
+/**
  *	ata_bmdma_error_handler - Stock error handler for BMDMA controller
  *	@ap: port to handle error for
  *
@@ -2848,7 +2892,8 @@ void ata_bmdma_error_handler(struct ata_port *ap)
 		/* if we're gonna thaw, make sure IRQ is clear */
 		if (thaw) {
 			ap->ops->sff_check_status(ap);
-			ap->ops->sff_irq_clear(ap);
+			if (ap->ops->sff_irq_clear)
+				ap->ops->sff_irq_clear(ap);
 		}
 	}
 
@@ -2882,6 +2927,28 @@ void ata_bmdma_post_internal_cmd(struct ata_queued_cmd *qc)
 EXPORT_SYMBOL_GPL(ata_bmdma_post_internal_cmd);
 
 /**
+ *	ata_bmdma_irq_clear - Clear PCI IDE BMDMA interrupt.
+ *	@ap: Port associated with this ATA transaction.
+ *
+ *	Clear interrupt and error flags in DMA status register.
+ *
+ *	May be used as the irq_clear() entry in ata_port_operations.
+ *
+ *	LOCKING:
+ *	spin_lock_irqsave(host lock)
+ */
+void ata_bmdma_irq_clear(struct ata_port *ap)
+{
+	void __iomem *mmio = ap->ioaddr.bmdma_addr;
+
+	if (!mmio)
+		return;
+
+	iowrite8(ioread8(mmio + ATA_DMA_STATUS), mmio + ATA_DMA_STATUS);
+}
+EXPORT_SYMBOL_GPL(ata_bmdma_irq_clear);
+
+/**
  *	ata_bmdma_setup - Set up PCI IDE BMDMA transaction
  *	@qc: Info associated with this ATA transaction.
  *
@@ -3137,7 +3204,100 @@ void ata_pci_bmdma_init(struct ata_host *host)
 }
 EXPORT_SYMBOL_GPL(ata_pci_bmdma_init);
 
+/**
+ *	ata_pci_bmdma_prepare_host - helper to prepare PCI BMDMA ATA host
+ *	@pdev: target PCI device
+ *	@ppi: array of port_info, must be enough for two ports
+ *	@r_host: out argument for the initialized ATA host
+ *
+ *	Helper to allocate BMDMA ATA host for @pdev, acquire all PCI
+ *	resources and initialize it accordingly in one go.
+ *
+ *	LOCKING:
+ *	Inherited from calling layer (may sleep).
+ *
+ *	RETURNS:
+ *	0 on success, -errno otherwise.
+ */
+int ata_pci_bmdma_prepare_host(struct pci_dev *pdev,
+			       const struct ata_port_info * const * ppi,
+			       struct ata_host **r_host)
+{
+	int rc;
+
+	rc = ata_pci_sff_prepare_host(pdev, ppi, r_host);
+	if (rc)
+		return rc;
+
+	ata_pci_bmdma_init(*r_host);
+	return 0;
+}
+EXPORT_SYMBOL_GPL(ata_pci_bmdma_prepare_host);
+
+/**
+ *	ata_pci_bmdma_init_one - Initialize/register BMDMA PCI IDE controller
+ *	@pdev: Controller to be initialized
+ *	@ppi: array of port_info, must be enough for two ports
+ *	@sht: scsi_host_template to use when registering the host
+ *	@host_priv: host private_data
+ *	@hflags: host flags
+ *
+ *	This function is similar to ata_pci_sff_init_one() but also
+ *	takes care of BMDMA initialization.
+ *
+ *	LOCKING:
+ *	Inherited from PCI layer (may sleep).
+ *
+ *	RETURNS:
+ *	Zero on success, negative on errno-based value on error.
+ */
+int ata_pci_bmdma_init_one(struct pci_dev *pdev,
+			   const struct ata_port_info * const * ppi,
+			   struct scsi_host_template *sht, void *host_priv,
+			   int hflags)
+{
+	struct device *dev = &pdev->dev;
+	const struct ata_port_info *pi;
+	struct ata_host *host = NULL;
+	int rc;
+
+	DPRINTK("ENTER\n");
+
+	pi = ata_sff_find_valid_pi(ppi);
+	if (!pi) {
+		dev_printk(KERN_ERR, &pdev->dev,
+			   "no valid port_info specified\n");
+		return -EINVAL;
+	}
+
+	if (!devres_open_group(dev, NULL, GFP_KERNEL))
+		return -ENOMEM;
+
+	rc = pcim_enable_device(pdev);
+	if (rc)
+		goto out;
+
+	/* prepare and activate BMDMA host */
+	rc = ata_pci_bmdma_prepare_host(pdev, ppi, &host);
+	if (rc)
+		goto out;
+	host->private_data = host_priv;
+	host->flags |= hflags;
+
+	pci_set_master(pdev);
+	rc = ata_pci_sff_activate_host(host, ata_bmdma_interrupt, sht);
+ out:
+	if (rc == 0)
+		devres_remove_group(&pdev->dev, NULL);
+	else
+		devres_release_group(&pdev->dev, NULL);
+
+	return rc;
+}
+EXPORT_SYMBOL_GPL(ata_pci_bmdma_init_one);
+
 #endif /* CONFIG_PCI */
+#endif /* CONFIG_ATA_BMDMA */
 
 /**
  *	ata_sff_port_init - Initialize SFF/BMDMA ATA port
diff --git a/drivers/ata/pata_acpi.c b/drivers/ata/pata_acpi.c
index 066b9f301ed5..c8d47034d5e9 100644
--- a/drivers/ata/pata_acpi.c
+++ b/drivers/ata/pata_acpi.c
@@ -260,7 +260,7 @@ static int pacpi_init_one (struct pci_dev *pdev, const struct pci_device_id *id)
 			return rc;
 		pcim_pin_device(pdev);
 	}
-	return ata_pci_sff_init_one(pdev, ppi, &pacpi_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(pdev, ppi, &pacpi_sht, NULL, 0);
 }
 
 static const struct pci_device_id pacpi_pci_tbl[] = {
diff --git a/drivers/ata/pata_ali.c b/drivers/ata/pata_ali.c
index f306e10c748d..794ec6e3275d 100644
--- a/drivers/ata/pata_ali.c
+++ b/drivers/ata/pata_ali.c
@@ -583,7 +583,10 @@ static int ali_init_one(struct pci_dev *pdev, const struct pci_device_id *id)
 	        	ppi[0] = &info_20_udma;
 	}
 
-	return ata_pci_sff_init_one(pdev, ppi, &ali_sht, NULL, 0);
+	if (!ppi[0]->mwdma_mask && !ppi[0]->udma_mask)
+		return ata_pci_sff_init_one(pdev, ppi, &ali_sht, NULL, 0);
+	else
+		return ata_pci_bmdma_init_one(pdev, ppi, &ali_sht, NULL, 0);
 }
 
 #ifdef CONFIG_PM
diff --git a/drivers/ata/pata_amd.c b/drivers/ata/pata_amd.c
index d95eca9c547e..620a07cabe31 100644
--- a/drivers/ata/pata_amd.c
+++ b/drivers/ata/pata_amd.c
@@ -574,7 +574,7 @@ static int amd_init_one(struct pci_dev *pdev, const struct pci_device_id *id)
 	}
 
 	/* And fire it up */
-	return ata_pci_sff_init_one(pdev, ppi, &amd_sht, hpriv, 0);
+	return ata_pci_bmdma_init_one(pdev, ppi, &amd_sht, hpriv, 0);
 }
 
 #ifdef CONFIG_PM
diff --git a/drivers/ata/pata_artop.c b/drivers/ata/pata_artop.c
index 4d066d6c30fa..ba43f0f8c880 100644
--- a/drivers/ata/pata_artop.c
+++ b/drivers/ata/pata_artop.c
@@ -421,7 +421,7 @@ static int artop_init_one (struct pci_dev *pdev, const struct pci_device_id *id)
 
 	BUG_ON(ppi[0] == NULL);
 
-	return ata_pci_sff_init_one(pdev, ppi, &artop_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(pdev, ppi, &artop_sht, NULL, 0);
 }
 
 static const struct pci_device_id artop_pci_tbl[] = {
diff --git a/drivers/ata/pata_atiixp.c b/drivers/ata/pata_atiixp.c
index 44d88b380ddd..43755616dc5a 100644
--- a/drivers/ata/pata_atiixp.c
+++ b/drivers/ata/pata_atiixp.c
@@ -246,8 +246,8 @@ static int atiixp_init_one(struct pci_dev *pdev, const struct pci_device_id *id)
 		if (!pci_test_config_bits(pdev, &atiixp_enable_bits[i]))
 			ppi[i] = &ata_dummy_port_info;
 
-	return ata_pci_sff_init_one(pdev, ppi, &atiixp_sht, NULL,
-						ATA_HOST_PARALLEL_SCAN);
+	return ata_pci_bmdma_init_one(pdev, ppi, &atiixp_sht, NULL,
+				      ATA_HOST_PARALLEL_SCAN);
 }
 
 static const struct pci_device_id atiixp[] = {
diff --git a/drivers/ata/pata_atp867x.c b/drivers/ata/pata_atp867x.c
index bb6e0746e07d..95295935dd95 100644
--- a/drivers/ata/pata_atp867x.c
+++ b/drivers/ata/pata_atp867x.c
@@ -525,7 +525,7 @@ static int atp867x_init_one(struct pci_dev *pdev,
 
 	pci_set_master(pdev);
 
-	rc = ata_host_activate(host, pdev->irq, ata_sff_interrupt,
+	rc = ata_host_activate(host, pdev->irq, ata_bmdma_interrupt,
 				IRQF_SHARED, &atp867x_sht);
 	if (rc)
 		dev_printk(KERN_ERR, &pdev->dev, "failed to activate host\n");
diff --git a/drivers/ata/pata_bf54x.c b/drivers/ata/pata_bf54x.c
index 6422cfd13d0d..9cae65de750e 100644
--- a/drivers/ata/pata_bf54x.c
+++ b/drivers/ata/pata_bf54x.c
@@ -1214,7 +1214,7 @@ static unsigned int bfin_data_xfer(struct ata_device *dev, unsigned char *buf,
  *	bfin_irq_clear - Clear ATAPI interrupt.
  *	@ap: Port associated with this ATA transaction.
  *
- *	Note: Original code is ata_sff_irq_clear().
+ *	Note: Original code is ata_bmdma_irq_clear().
  */
 
 static void bfin_irq_clear(struct ata_port *ap)
diff --git a/drivers/ata/pata_cmd64x.c b/drivers/ata/pata_cmd64x.c
index 4c81a71b8877..9f5da1c7454b 100644
--- a/drivers/ata/pata_cmd64x.c
+++ b/drivers/ata/pata_cmd64x.c
@@ -367,7 +367,7 @@ static int cmd64x_init_one(struct pci_dev *pdev, const struct pci_device_id *id)
 	pci_write_config_byte(pdev, UDIDETCR0, 0xF0);
 #endif
 
-	return ata_pci_sff_init_one(pdev, ppi, &cmd64x_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(pdev, ppi, &cmd64x_sht, NULL, 0);
 }
 
 #ifdef CONFIG_PM
diff --git a/drivers/ata/pata_cs5520.c b/drivers/ata/pata_cs5520.c
index 17c5f346ff01..030952f1f97c 100644
--- a/drivers/ata/pata_cs5520.c
+++ b/drivers/ata/pata_cs5520.c
@@ -221,7 +221,7 @@ static int __devinit cs5520_init_one(struct pci_dev *pdev, const struct pci_devi
 			continue;
 
 		rc = devm_request_irq(&pdev->dev, irq[ap->port_no],
-				      ata_sff_interrupt, 0, DRV_NAME, host);
+				      ata_bmdma_interrupt, 0, DRV_NAME, host);
 		if (rc)
 			return rc;
 
diff --git a/drivers/ata/pata_cs5530.c b/drivers/ata/pata_cs5530.c
index e809a4233a81..f792330f0d8e 100644
--- a/drivers/ata/pata_cs5530.c
+++ b/drivers/ata/pata_cs5530.c
@@ -324,7 +324,7 @@ static int cs5530_init_one(struct pci_dev *pdev, const struct pci_device_id *id)
 		ppi[1] = &info_palmax_secondary;
 
 	/* Now kick off ATA set up */
-	return ata_pci_sff_init_one(pdev, ppi, &cs5530_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(pdev, ppi, &cs5530_sht, NULL, 0);
 }
 
 #ifdef CONFIG_PM
diff --git a/drivers/ata/pata_cs5535.c b/drivers/ata/pata_cs5535.c
index a02e6459fdcc..03a93186aa19 100644
--- a/drivers/ata/pata_cs5535.c
+++ b/drivers/ata/pata_cs5535.c
@@ -198,7 +198,7 @@ static int cs5535_init_one(struct pci_dev *dev, const struct pci_device_id *id)
 	rdmsr(ATAC_CH0D1_PIO, timings, dummy);
 	if (CS5535_BAD_PIO(timings))
 		wrmsr(ATAC_CH0D1_PIO, 0xF7F4F7F4UL, 0);
-	return ata_pci_sff_init_one(dev, ppi, &cs5535_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(dev, ppi, &cs5535_sht, NULL, 0);
 }
 
 static const struct pci_device_id cs5535[] = {
diff --git a/drivers/ata/pata_cs5536.c b/drivers/ata/pata_cs5536.c
index 914ae3506ff5..21ee23f89e88 100644
--- a/drivers/ata/pata_cs5536.c
+++ b/drivers/ata/pata_cs5536.c
@@ -260,7 +260,7 @@ static int cs5536_init_one(struct pci_dev *dev, const struct pci_device_id *id)
 		return -ENODEV;
 	}
 
-	return ata_pci_sff_init_one(dev, ppi, &cs5536_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(dev, ppi, &cs5536_sht, NULL, 0);
 }
 
 static const struct pci_device_id cs5536[] = {
diff --git a/drivers/ata/pata_cypress.c b/drivers/ata/pata_cypress.c
index 0fcc096b8dac..6d915b063d93 100644
--- a/drivers/ata/pata_cypress.c
+++ b/drivers/ata/pata_cypress.c
@@ -138,7 +138,7 @@ static int cy82c693_init_one(struct pci_dev *pdev, const struct pci_device_id *i
 	if (PCI_FUNC(pdev->devfn) != 1)
 		return -ENODEV;
 
-	return ata_pci_sff_init_one(pdev, ppi, &cy82c693_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(pdev, ppi, &cy82c693_sht, NULL, 0);
 }
 
 static const struct pci_device_id cy82c693[] = {
diff --git a/drivers/ata/pata_efar.c b/drivers/ata/pata_efar.c
index 3bac0e079691..a08834758ea2 100644
--- a/drivers/ata/pata_efar.c
+++ b/drivers/ata/pata_efar.c
@@ -277,8 +277,8 @@ static int efar_init_one (struct pci_dev *pdev, const struct pci_device_id *ent)
 		dev_printk(KERN_DEBUG, &pdev->dev,
 			   "version " DRV_VERSION "\n");
 
-	return ata_pci_sff_init_one(pdev, ppi, &efar_sht, NULL,
-					ATA_HOST_PARALLEL_SCAN);
+	return ata_pci_bmdma_init_one(pdev, ppi, &efar_sht, NULL,
+				      ATA_HOST_PARALLEL_SCAN);
 }
 
 static const struct pci_device_id efar_pci_tbl[] = {
diff --git a/drivers/ata/pata_hpt366.c b/drivers/ata/pata_hpt366.c
index 8580eb3cd54d..7688868557b9 100644
--- a/drivers/ata/pata_hpt366.c
+++ b/drivers/ata/pata_hpt366.c
@@ -361,7 +361,7 @@ static int hpt36x_init_one(struct pci_dev *dev, const struct pci_device_id *id)
 			break;
 	}
 	/* Now kick off ATA set up */
-	return ata_pci_sff_init_one(dev, ppi, &hpt36x_sht, hpriv, 0);
+	return ata_pci_bmdma_init_one(dev, ppi, &hpt36x_sht, hpriv, 0);
 }
 
 #ifdef CONFIG_PM
diff --git a/drivers/ata/pata_hpt37x.c b/drivers/ata/pata_hpt37x.c
index 98b498b6907c..9ae4c0830577 100644
--- a/drivers/ata/pata_hpt37x.c
+++ b/drivers/ata/pata_hpt37x.c
@@ -987,7 +987,7 @@ static int hpt37x_init_one(struct pci_dev *dev, const struct pci_device_id *id)
 	}
 
 	/* Now kick off ATA set up */
-	return ata_pci_sff_init_one(dev, ppi, &hpt37x_sht, private_data, 0);
+	return ata_pci_bmdma_init_one(dev, ppi, &hpt37x_sht, private_data, 0);
 }
 
 static const struct pci_device_id hpt37x[] = {
diff --git a/drivers/ata/pata_hpt3x2n.c b/drivers/ata/pata_hpt3x2n.c
index 8b95aeba0e74..32f3463216b8 100644
--- a/drivers/ata/pata_hpt3x2n.c
+++ b/drivers/ata/pata_hpt3x2n.c
@@ -548,7 +548,7 @@ static int hpt3x2n_init_one(struct pci_dev *dev, const struct pci_device_id *id)
 		outb(inb(iobase + 0x9c) | 0x04, iobase + 0x9c);
 
 	/* Now kick off ATA set up */
-	return ata_pci_sff_init_one(dev, ppi, &hpt3x2n_sht, hpriv, 0);
+	return ata_pci_bmdma_init_one(dev, ppi, &hpt3x2n_sht, hpriv, 0);
 }
 
 static const struct pci_device_id hpt3x2n[] = {
diff --git a/drivers/ata/pata_hpt3x3.c b/drivers/ata/pata_hpt3x3.c
index 727a81ce4c9f..b63d5e2d4628 100644
--- a/drivers/ata/pata_hpt3x3.c
+++ b/drivers/ata/pata_hpt3x3.c
@@ -248,7 +248,7 @@ static int hpt3x3_init_one(struct pci_dev *pdev, const struct pci_device_id *id)
 		ata_port_pbar_desc(ap, 4, offset_cmd[i], "cmd");
 	}
 	pci_set_master(pdev);
-	return ata_host_activate(host, pdev->irq, ata_sff_interrupt,
+	return ata_host_activate(host, pdev->irq, ata_bmdma_interrupt,
 				 IRQF_SHARED, &hpt3x3_sht);
 }
 
diff --git a/drivers/ata/pata_icside.c b/drivers/ata/pata_icside.c
index b56e8f722d20..9f2889fe43b2 100644
--- a/drivers/ata/pata_icside.c
+++ b/drivers/ata/pata_icside.c
@@ -470,7 +470,7 @@ static int __devinit pata_icside_add_ports(struct pata_icside_info *info)
 		pata_icside_setup_ioaddr(ap, info->base, info, info->port[i]);
 	}
 
-	return ata_host_activate(host, ec->irq, ata_sff_interrupt, 0,
+	return ata_host_activate(host, ec->irq, ata_bmdma_interrupt, 0,
 				 &pata_icside_sht);
 }
 
diff --git a/drivers/ata/pata_it8213.c b/drivers/ata/pata_it8213.c
index f971f0de88e6..4d142a2ab8fd 100644
--- a/drivers/ata/pata_it8213.c
+++ b/drivers/ata/pata_it8213.c
@@ -273,7 +273,7 @@ static int it8213_init_one (struct pci_dev *pdev, const struct pci_device_id *en
 		dev_printk(KERN_DEBUG, &pdev->dev,
 			   "version " DRV_VERSION "\n");
 
-	return ata_pci_sff_init_one(pdev, ppi, &it8213_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(pdev, ppi, &it8213_sht, NULL, 0);
 }
 
 static const struct pci_device_id it8213_pci_tbl[] = {
diff --git a/drivers/ata/pata_it821x.c b/drivers/ata/pata_it821x.c
index 2bd2b002d14a..bf88f71a21f4 100644
--- a/drivers/ata/pata_it821x.c
+++ b/drivers/ata/pata_it821x.c
@@ -933,7 +933,7 @@ static int it821x_init_one(struct pci_dev *pdev, const struct pci_device_id *id)
 		else
 			ppi[0] = &info_smart;
 	}
-	return ata_pci_sff_init_one(pdev, ppi, &it821x_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(pdev, ppi, &it821x_sht, NULL, 0);
 }
 
 #ifdef CONFIG_PM
diff --git a/drivers/ata/pata_jmicron.c b/drivers/ata/pata_jmicron.c
index 565e01e6ac7c..cb3babbb7035 100644
--- a/drivers/ata/pata_jmicron.c
+++ b/drivers/ata/pata_jmicron.c
@@ -144,7 +144,7 @@ static int jmicron_init_one (struct pci_dev *pdev, const struct pci_device_id *i
 	};
 	const struct ata_port_info *ppi[] = { &info, NULL };
 
-	return ata_pci_sff_init_one(pdev, ppi, &jmicron_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(pdev, ppi, &jmicron_sht, NULL, 0);
 }
 
 static const struct pci_device_id jmicron_pci_tbl[] = {
diff --git a/drivers/ata/pata_macio.c b/drivers/ata/pata_macio.c
index 25df50f51c04..75b49d01780b 100644
--- a/drivers/ata/pata_macio.c
+++ b/drivers/ata/pata_macio.c
@@ -1110,7 +1110,7 @@ static int __devinit pata_macio_common_init(struct pata_macio_priv	*priv,
 
 	/* Start it up */
 	priv->irq = irq;
-	return ata_host_activate(priv->host, irq, ata_sff_interrupt, 0,
+	return ata_host_activate(priv->host, irq, ata_bmdma_interrupt, 0,
 				 &pata_macio_sht);
 }
 
@@ -1140,7 +1140,7 @@ static int __devinit pata_macio_attach(struct macio_dev *mdev,
 			"Failed to allocate private memory\n");
 		return -ENOMEM;
 	}
-	priv->node = of_node_get(mdev->ofdev.node);
+	priv->node = of_node_get(mdev->ofdev.dev.of_node);
 	priv->mdev = mdev;
 	priv->dev = &mdev->ofdev.dev;
 
@@ -1355,8 +1355,11 @@ static struct of_device_id pata_macio_match[] =
 
 static struct macio_driver pata_macio_driver =
 {
-	.name 		= "pata-macio",
-	.match_table	= pata_macio_match,
+	.driver = {
+		.name 		= "pata-macio",
+		.owner		= THIS_MODULE,
+		.of_match_table	= pata_macio_match,
+	},
 	.probe		= pata_macio_attach,
 	.remove		= pata_macio_detach,
 #ifdef CONFIG_PM
@@ -1366,9 +1369,6 @@ static struct macio_driver pata_macio_driver =
 #ifdef CONFIG_PMAC_MEDIABAY
 	.mediabay_event	= pata_macio_mb_event,
 #endif
-	.driver = {
-		.owner		= THIS_MODULE,
-	},
 };
 
 static const struct pci_device_id pata_macio_pci_match[] = {
diff --git a/drivers/ata/pata_marvell.c b/drivers/ata/pata_marvell.c
index e8ca02e5a71d..dd38083dcbeb 100644
--- a/drivers/ata/pata_marvell.c
+++ b/drivers/ata/pata_marvell.c
@@ -153,7 +153,7 @@ static int marvell_init_one (struct pci_dev *pdev, const struct pci_device_id *i
 		return -ENODEV;
 	}
 #endif
-	return ata_pci_sff_init_one(pdev, ppi, &marvell_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(pdev, ppi, &marvell_sht, NULL, 0);
 }
 
 static const struct pci_device_id marvell_pci_tbl[] = {
diff --git a/drivers/ata/pata_mpc52xx.c b/drivers/ata/pata_mpc52xx.c
index 96b11b604ae0..f087ab55b1df 100644
--- a/drivers/ata/pata_mpc52xx.c
+++ b/drivers/ata/pata_mpc52xx.c
@@ -659,7 +659,7 @@ mpc52xx_ata_init_one(struct device *dev, struct mpc52xx_ata_priv *priv,
 	ata_port_desc(ap, "ata_regs 0x%lx", raw_ata_regs);
 
 	/* activate host */
-	return ata_host_activate(host, priv->ata_irq, ata_sff_interrupt, 0,
+	return ata_host_activate(host, priv->ata_irq, ata_bmdma_interrupt, 0,
 				 &mpc52xx_ata_sht);
 }
 
@@ -694,7 +694,7 @@ mpc52xx_ata_probe(struct of_device *op, const struct of_device_id *match)
 	struct bcom_task *dmatsk = NULL;
 
 	/* Get ipb frequency */
-	ipb_freq = mpc5xxx_get_bus_frequency(op->node);
+	ipb_freq = mpc5xxx_get_bus_frequency(op->dev.of_node);
 	if (!ipb_freq) {
 		dev_err(&op->dev, "could not determine IPB bus frequency\n");
 		return -ENODEV;
@@ -702,7 +702,7 @@ mpc52xx_ata_probe(struct of_device *op, const struct of_device_id *match)
 
 	/* Get device base address from device tree, request the region
 	 * and ioremap it. */
-	rv = of_address_to_resource(op->node, 0, &res_mem);
+	rv = of_address_to_resource(op->dev.of_node, 0, &res_mem);
 	if (rv) {
 		dev_err(&op->dev, "could not determine device base address\n");
 		return rv;
@@ -735,14 +735,14 @@ mpc52xx_ata_probe(struct of_device *op, const struct of_device_id *match)
 	 * The MPC5200 ATA controller supports MWDMA modes 0, 1 and 2 and
 	 * UDMA modes 0, 1 and 2.
 	 */
-	prop = of_get_property(op->node, "mwdma-mode", &proplen);
+	prop = of_get_property(op->dev.of_node, "mwdma-mode", &proplen);
 	if ((prop) && (proplen >= 4))
 		mwdma_mask = ATA_MWDMA2 & ((1 << (*prop + 1)) - 1);
-	prop = of_get_property(op->node, "udma-mode", &proplen);
+	prop = of_get_property(op->dev.of_node, "udma-mode", &proplen);
 	if ((prop) && (proplen >= 4))
 		udma_mask = ATA_UDMA2 & ((1 << (*prop + 1)) - 1);
 
-	ata_irq = irq_of_parse_and_map(op->node, 0);
+	ata_irq = irq_of_parse_and_map(op->dev.of_node, 0);
 	if (ata_irq == NO_IRQ) {
 		dev_err(&op->dev, "error mapping irq\n");
 		return -EINVAL;
@@ -884,9 +884,6 @@ static struct of_device_id mpc52xx_ata_of_match[] = {
 
 
 static struct of_platform_driver mpc52xx_ata_of_platform_driver = {
-	.owner		= THIS_MODULE,
-	.name		= DRV_NAME,
-	.match_table	= mpc52xx_ata_of_match,
 	.probe		= mpc52xx_ata_probe,
 	.remove		= mpc52xx_ata_remove,
 #ifdef CONFIG_PM
@@ -896,6 +893,7 @@ static struct of_platform_driver mpc52xx_ata_of_platform_driver = {
 	.driver		= {
 		.name	= DRV_NAME,
 		.owner	= THIS_MODULE,
+		.of_match_table = mpc52xx_ata_of_match,
 	},
 };
 
diff --git a/drivers/ata/pata_netcell.c b/drivers/ata/pata_netcell.c
index 94f979a7f4f7..3eb921c746a1 100644
--- a/drivers/ata/pata_netcell.c
+++ b/drivers/ata/pata_netcell.c
@@ -82,7 +82,7 @@ static int netcell_init_one (struct pci_dev *pdev, const struct pci_device_id *e
 	ata_pci_bmdma_clear_simplex(pdev);
 
 	/* And let the library code do the work */
-	return ata_pci_sff_init_one(pdev, port_info, &netcell_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(pdev, port_info, &netcell_sht, NULL, 0);
 }
 
 static const struct pci_device_id netcell_pci_tbl[] = {
diff --git a/drivers/ata/pata_ninja32.c b/drivers/ata/pata_ninja32.c
index dd53a66b19e3..cc50bd09aa26 100644
--- a/drivers/ata/pata_ninja32.c
+++ b/drivers/ata/pata_ninja32.c
@@ -149,7 +149,7 @@ static int ninja32_init_one(struct pci_dev *dev, const struct pci_device_id *id)
 
 	ninja32_program(base);
 	/* FIXME: Should we disable them at remove ? */
-	return ata_host_activate(host, dev->irq, ata_sff_interrupt,
+	return ata_host_activate(host, dev->irq, ata_bmdma_interrupt,
 				 IRQF_SHARED, &ninja32_sht);
 }
 
diff --git a/drivers/ata/pata_ns87415.c b/drivers/ata/pata_ns87415.c
index fdbba2d76d3e..605f198f958c 100644
--- a/drivers/ata/pata_ns87415.c
+++ b/drivers/ata/pata_ns87415.c
@@ -380,7 +380,7 @@ static int ns87415_init_one (struct pci_dev *pdev, const struct pci_device_id *e
 
 	ns87415_fixup(pdev);
 
-	return ata_pci_sff_init_one(pdev, ppi, &ns87415_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(pdev, ppi, &ns87415_sht, NULL, 0);
 }
 
 static const struct pci_device_id ns87415_pci_tbl[] = {
diff --git a/drivers/ata/pata_octeon_cf.c b/drivers/ata/pata_octeon_cf.c
index 3001109352ea..06ddd91ffeda 100644
--- a/drivers/ata/pata_octeon_cf.c
+++ b/drivers/ata/pata_octeon_cf.c
@@ -750,20 +750,6 @@ static void octeon_cf_dev_config(struct ata_device *dev)
 }
 
 /*
- * Trap if driver tries to do standard bmdma commands.  They are not
- * supported.
- */
-static void unreachable_qc(struct ata_queued_cmd *qc)
-{
-	BUG();
-}
-
-static u8 unreachable_port(struct ata_port *ap)
-{
-	BUG();
-}
-
-/*
  * We don't do ATAPI DMA so return 0.
  */
 static int octeon_cf_check_atapi_dma(struct ata_queued_cmd *qc)
@@ -804,10 +790,6 @@ static struct ata_port_operations octeon_cf_ops = {
 	.sff_dev_select		= octeon_cf_dev_select,
 	.sff_irq_on		= octeon_cf_irq_on,
 	.sff_irq_clear		= octeon_cf_irq_clear,
-	.bmdma_setup		= unreachable_qc,
-	.bmdma_start		= unreachable_qc,
-	.bmdma_stop		= unreachable_qc,
-	.bmdma_status		= unreachable_port,
 	.cable_detect		= ata_cable_40wire,
 	.set_piomode		= octeon_cf_set_piomode,
 	.set_dmamode		= octeon_cf_set_dmamode,
diff --git a/drivers/ata/pata_of_platform.c b/drivers/ata/pata_of_platform.c
index 1f18ad9e4fe1..5a1b82c08be9 100644
--- a/drivers/ata/pata_of_platform.c
+++ b/drivers/ata/pata_of_platform.c
@@ -18,7 +18,7 @@ static int __devinit pata_of_platform_probe(struct of_device *ofdev,
 					    const struct of_device_id *match)
 {
 	int ret;
-	struct device_node *dn = ofdev->node;
+	struct device_node *dn = ofdev->dev.of_node;
 	struct resource io_res;
 	struct resource ctl_res;
 	struct resource irq_res;
@@ -91,8 +91,11 @@ static struct of_device_id pata_of_platform_match[] = {
 MODULE_DEVICE_TABLE(of, pata_of_platform_match);
 
 static struct of_platform_driver pata_of_platform_driver = {
-	.name		= "pata_of_platform",
-	.match_table	= pata_of_platform_match,
+	.driver = {
+		.name = "pata_of_platform",
+		.owner = THIS_MODULE,
+		.of_match_table = pata_of_platform_match,
+	},
 	.probe		= pata_of_platform_probe,
 	.remove		= __devexit_p(pata_of_platform_remove),
 };
diff --git a/drivers/ata/pata_oldpiix.c b/drivers/ata/pata_oldpiix.c
index 988ef2627be3..b811c1636204 100644
--- a/drivers/ata/pata_oldpiix.c
+++ b/drivers/ata/pata_oldpiix.c
@@ -248,7 +248,7 @@ static int oldpiix_init_one (struct pci_dev *pdev, const struct pci_device_id *e
 		dev_printk(KERN_DEBUG, &pdev->dev,
 			   "version " DRV_VERSION "\n");
 
-	return ata_pci_sff_init_one(pdev, ppi, &oldpiix_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(pdev, ppi, &oldpiix_sht, NULL, 0);
 }
 
 static const struct pci_device_id oldpiix_pci_tbl[] = {
diff --git a/drivers/ata/pata_optidma.c b/drivers/ata/pata_optidma.c
index 76b7d12b1e8d..0852cd07de08 100644
--- a/drivers/ata/pata_optidma.c
+++ b/drivers/ata/pata_optidma.c
@@ -429,7 +429,7 @@ static int optidma_init_one(struct pci_dev *dev, const struct pci_device_id *id)
 	if (optiplus_with_udma(dev))
 		ppi[0] = &info_82c700_udma;
 
-	return ata_pci_sff_init_one(dev, ppi, &optidma_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(dev, ppi, &optidma_sht, NULL, 0);
 }
 
 static const struct pci_device_id optidma[] = {
diff --git a/drivers/ata/pata_pdc2027x.c b/drivers/ata/pata_pdc2027x.c
index 09f1f22c0307..b18351122525 100644
--- a/drivers/ata/pata_pdc2027x.c
+++ b/drivers/ata/pata_pdc2027x.c
@@ -754,7 +754,7 @@ static int __devinit pdc2027x_init_one(struct pci_dev *pdev, const struct pci_de
 		return -EIO;
 
 	pci_set_master(pdev);
-	return ata_host_activate(host, pdev->irq, ata_sff_interrupt,
+	return ata_host_activate(host, pdev->irq, ata_bmdma_interrupt,
 				 IRQF_SHARED, &pdc2027x_sht);
 }
 
diff --git a/drivers/ata/pata_pdc202xx_old.c b/drivers/ata/pata_pdc202xx_old.c
index fa1e2f3bc0fd..c39f213e1bbc 100644
--- a/drivers/ata/pata_pdc202xx_old.c
+++ b/drivers/ata/pata_pdc202xx_old.c
@@ -337,7 +337,7 @@ static int pdc202xx_init_one(struct pci_dev *dev, const struct pci_device_id *id
 				return -ENODEV;
 		}
 	}
-	return ata_pci_sff_init_one(dev, ppi, &pdc202xx_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(dev, ppi, &pdc202xx_sht, NULL, 0);
 }
 
 static const struct pci_device_id pdc202xx[] = {
diff --git a/drivers/ata/pata_piccolo.c b/drivers/ata/pata_piccolo.c
index 981615414849..cb01bf9496fe 100644
--- a/drivers/ata/pata_piccolo.c
+++ b/drivers/ata/pata_piccolo.c
@@ -95,7 +95,7 @@ static int ata_tosh_init_one(struct pci_dev *dev, const struct pci_device_id *id
 	};
 	const struct ata_port_info *ppi[] = { &info, &ata_dummy_port_info };
 	/* Just one port for the moment */
-	return ata_pci_sff_init_one(dev, ppi, &tosh_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(dev, ppi, &tosh_sht, NULL, 0);
 }
 
 static struct pci_device_id ata_tosh[] = {
diff --git a/drivers/ata/pata_radisys.c b/drivers/ata/pata_radisys.c
index a5fa388e5398..8574b31f1773 100644
--- a/drivers/ata/pata_radisys.c
+++ b/drivers/ata/pata_radisys.c
@@ -227,7 +227,7 @@ static int radisys_init_one (struct pci_dev *pdev, const struct pci_device_id *e
 		dev_printk(KERN_DEBUG, &pdev->dev,
 			   "version " DRV_VERSION "\n");
 
-	return ata_pci_sff_init_one(pdev, ppi, &radisys_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(pdev, ppi, &radisys_sht, NULL, 0);
 }
 
 static const struct pci_device_id radisys_pci_tbl[] = {
diff --git a/drivers/ata/pata_rdc.c b/drivers/ata/pata_rdc.c
index 37092cfd7bc6..5fbe9b166c69 100644
--- a/drivers/ata/pata_rdc.c
+++ b/drivers/ata/pata_rdc.c
@@ -344,7 +344,7 @@ static int __devinit rdc_init_one(struct pci_dev *pdev,
 	 */
 	pci_read_config_dword(pdev, 0x54, &hpriv->saved_iocfg);
 
-	rc = ata_pci_sff_prepare_host(pdev, ppi, &host);
+	rc = ata_pci_bmdma_prepare_host(pdev, ppi, &host);
 	if (rc)
 		return rc;
 	host->private_data = hpriv;
@@ -354,7 +354,7 @@ static int __devinit rdc_init_one(struct pci_dev *pdev,
 	host->flags |= ATA_HOST_PARALLEL_SCAN;
 
 	pci_set_master(pdev);
-	return ata_pci_sff_activate_host(host, ata_sff_interrupt, &rdc_sht);
+	return ata_pci_sff_activate_host(host, ata_bmdma_interrupt, &rdc_sht);
 }
 
 static void rdc_remove_one(struct pci_dev *pdev)
diff --git a/drivers/ata/pata_sc1200.c b/drivers/ata/pata_sc1200.c
index 6b5b63a2fd8e..e2c18257adff 100644
--- a/drivers/ata/pata_sc1200.c
+++ b/drivers/ata/pata_sc1200.c
@@ -237,7 +237,7 @@ static int sc1200_init_one(struct pci_dev *dev, const struct pci_device_id *id)
 	};
 	const struct ata_port_info *ppi[] = { &info, NULL };
 
-	return ata_pci_sff_init_one(dev, ppi, &sc1200_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(dev, ppi, &sc1200_sht, NULL, 0);
 }
 
 static const struct pci_device_id sc1200[] = {
diff --git a/drivers/ata/pata_scc.c b/drivers/ata/pata_scc.c
index 6f6193b707cb..d9db3f8d60ef 100644
--- a/drivers/ata/pata_scc.c
+++ b/drivers/ata/pata_scc.c
@@ -875,7 +875,7 @@ static void scc_postreset(struct ata_link *link, unsigned int *classes)
  *	scc_irq_clear - Clear PCI IDE BMDMA interrupt.
  *	@ap: Port associated with this ATA transaction.
  *
- *	Note: Original code is ata_sff_irq_clear().
+ *	Note: Original code is ata_bmdma_irq_clear().
  */
 
 static void scc_irq_clear (struct ata_port *ap)
@@ -1105,7 +1105,7 @@ static int scc_init_one (struct pci_dev *pdev, const struct pci_device_id *ent)
 	if (rc)
 		return rc;
 
-	return ata_host_activate(host, pdev->irq, ata_sff_interrupt,
+	return ata_host_activate(host, pdev->irq, ata_bmdma_interrupt,
 				 IRQF_SHARED, &scc_sht);
 }
 
diff --git a/drivers/ata/pata_sch.c b/drivers/ata/pata_sch.c
index 86b3d0133c7c..e97b32f03a6e 100644
--- a/drivers/ata/pata_sch.c
+++ b/drivers/ata/pata_sch.c
@@ -179,7 +179,7 @@ static int __devinit sch_init_one(struct pci_dev *pdev,
 		dev_printk(KERN_DEBUG, &pdev->dev,
 			   "version " DRV_VERSION "\n");
 
-	return ata_pci_sff_init_one(pdev, ppi, &sch_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(pdev, ppi, &sch_sht, NULL, 0);
 }
 
 static int __init sch_init(void)
diff --git a/drivers/ata/pata_serverworks.c b/drivers/ata/pata_serverworks.c
index 43ea389df2b3..86dd714e3e1d 100644
--- a/drivers/ata/pata_serverworks.c
+++ b/drivers/ata/pata_serverworks.c
@@ -460,7 +460,7 @@ static int serverworks_init_one(struct pci_dev *pdev, const struct pci_device_id
 	if (pdev->device == PCI_DEVICE_ID_SERVERWORKS_CSB5IDE)
 		ata_pci_bmdma_clear_simplex(pdev);
 
-	return ata_pci_sff_init_one(pdev, ppi, &serverworks_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(pdev, ppi, &serverworks_sht, NULL, 0);
 }
 
 #ifdef CONFIG_PM
diff --git a/drivers/ata/pata_sil680.c b/drivers/ata/pata_sil680.c
index 43faf106f647..d3190d7ec304 100644
--- a/drivers/ata/pata_sil680.c
+++ b/drivers/ata/pata_sil680.c
@@ -374,11 +374,11 @@ static int __devinit sil680_init_one(struct pci_dev *pdev,
 	ata_sff_std_ports(&host->ports[1]->ioaddr);
 
 	/* Register & activate */
-	return ata_host_activate(host, pdev->irq, ata_sff_interrupt,
+	return ata_host_activate(host, pdev->irq, ata_bmdma_interrupt,
 				 IRQF_SHARED, &sil680_sht);
 
 use_ioports:
-	return ata_pci_sff_init_one(pdev, ppi, &sil680_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(pdev, ppi, &sil680_sht, NULL, 0);
 }
 
 #ifdef CONFIG_PM
diff --git a/drivers/ata/pata_sis.c b/drivers/ata/pata_sis.c
index b6708032f321..60cea13cccce 100644
--- a/drivers/ata/pata_sis.c
+++ b/drivers/ata/pata_sis.c
@@ -826,7 +826,7 @@ static int sis_init_one (struct pci_dev *pdev, const struct pci_device_id *ent)
 
 	sis_fixup(pdev, chipset);
 
-	return ata_pci_sff_init_one(pdev, ppi, &sis_sht, chipset, 0);
+	return ata_pci_bmdma_init_one(pdev, ppi, &sis_sht, chipset, 0);
 }
 
 #ifdef CONFIG_PM
diff --git a/drivers/ata/pata_sl82c105.c b/drivers/ata/pata_sl82c105.c
index 733b042a7469..98548f640c8e 100644
--- a/drivers/ata/pata_sl82c105.c
+++ b/drivers/ata/pata_sl82c105.c
@@ -316,7 +316,7 @@ static int sl82c105_init_one(struct pci_dev *dev, const struct pci_device_id *id
 	val |= CTRL_P0EN | CTRL_P0F16 | CTRL_P1F16;
 	pci_write_config_dword(dev, 0x40, val);
 
-	return ata_pci_sff_init_one(dev, ppi, &sl82c105_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(dev, ppi, &sl82c105_sht, NULL, 0);
 }
 
 static const struct pci_device_id sl82c105[] = {
diff --git a/drivers/ata/pata_triflex.c b/drivers/ata/pata_triflex.c
index 48f50600ed2a..0d1f89e571dd 100644
--- a/drivers/ata/pata_triflex.c
+++ b/drivers/ata/pata_triflex.c
@@ -201,7 +201,7 @@ static int triflex_init_one(struct pci_dev *dev, const struct pci_device_id *id)
 	if (!printed_version++)
 		dev_printk(KERN_DEBUG, &dev->dev, "version " DRV_VERSION "\n");
 
-	return ata_pci_sff_init_one(dev, ppi, &triflex_sht, NULL, 0);
+	return ata_pci_bmdma_init_one(dev, ppi, &triflex_sht, NULL, 0);
 }
 
 static const struct pci_device_id triflex[] = {
diff --git a/drivers/ata/pata_via.c b/drivers/ata/pata_via.c
index 7e3e0a5598b7..5e659885de16 100644
--- a/drivers/ata/pata_via.c
+++ b/drivers/ata/pata_via.c
@@ -627,7 +627,7 @@ static int via_init_one(struct pci_dev *pdev, const struct pci_device_id *id)
 	}
 
 	/* We have established the device type, now fire it up */
-	return ata_pci_sff_init_one(pdev, ppi, &via_sht, (void *)config, 0);
+	return ata_pci_bmdma_init_one(pdev, ppi, &via_sht, (void *)config, 0);
 }
 
 #ifdef CONFIG_PM
diff --git a/drivers/ata/sata_fsl.c b/drivers/ata/sata_fsl.c
index a69192b38b43..61c89b54ea23 100644
--- a/drivers/ata/sata_fsl.c
+++ b/drivers/ata/sata_fsl.c
@@ -1313,7 +1313,7 @@ static int sata_fsl_probe(struct of_device *ofdev,
 	dev_printk(KERN_INFO, &ofdev->dev,
 		   "Sata FSL Platform/CSB Driver init\n");
 
-	hcr_base = of_iomap(ofdev->node, 0);
+	hcr_base = of_iomap(ofdev->dev.of_node, 0);
 	if (!hcr_base)
 		goto error_exit_with_cleanup;
 
@@ -1332,7 +1332,7 @@ static int sata_fsl_probe(struct of_device *ofdev,
 	host_priv->ssr_base = ssr_base;
 	host_priv->csr_base = csr_base;
 
-	irq = irq_of_parse_and_map(ofdev->node, 0);
+	irq = irq_of_parse_and_map(ofdev->dev.of_node, 0);
 	if (irq < 0) {
 		dev_printk(KERN_ERR, &ofdev->dev, "invalid irq from platform\n");
 		goto error_exit_with_cleanup;
@@ -1427,8 +1427,11 @@ static struct of_device_id fsl_sata_match[] = {
 MODULE_DEVICE_TABLE(of, fsl_sata_match);
 
 static struct of_platform_driver fsl_sata_driver = {
-	.name		= "fsl-sata",
-	.match_table	= fsl_sata_match,
+	.driver = {
+		.name = "fsl-sata",
+		.owner = THIS_MODULE,
+		.of_match_table = fsl_sata_match,
+	},
 	.probe		= sata_fsl_probe,
 	.remove		= sata_fsl_remove,
 #ifdef CONFIG_PM
diff --git a/drivers/ata/sata_mv.c b/drivers/ata/sata_mv.c
index f3471bc949d3..a476cd99b95d 100644
--- a/drivers/ata/sata_mv.c
+++ b/drivers/ata/sata_mv.c
@@ -675,8 +675,6 @@ static struct ata_port_operations mv5_ops = {
 	.freeze			= mv_eh_freeze,
 	.thaw			= mv_eh_thaw,
 	.hardreset		= mv_hardreset,
-	.error_handler		= ata_std_error_handler, /* avoid SFF EH */
-	.post_internal_cmd	= ATA_OP_NULL,
 
 	.scr_read		= mv5_scr_read,
 	.scr_write		= mv5_scr_write,
@@ -2813,7 +2811,7 @@ static void mv_port_intr(struct ata_port *ap, u32 port_cause)
 	} else if (!edma_was_enabled) {
 		struct ata_queued_cmd *qc = mv_get_active_qc(ap);
 		if (qc)
-			ata_sff_host_intr(ap, qc);
+			ata_bmdma_port_intr(ap, qc);
 		else
 			mv_unexpected_intr(ap, edma_was_enabled);
 	}
diff --git a/drivers/ata/sata_nv.c b/drivers/ata/sata_nv.c
index baa8f0d2c86f..21161136cad0 100644
--- a/drivers/ata/sata_nv.c
+++ b/drivers/ata/sata_nv.c
@@ -920,7 +920,7 @@ static int nv_host_intr(struct ata_port *ap, u8 irq_stat)
 	}
 
 	/* handle interrupt */
-	return ata_sff_host_intr(ap, qc);
+	return ata_bmdma_port_intr(ap, qc);
 }
 
 static irqreturn_t nv_adma_interrupt(int irq, void *dev_instance)
@@ -1100,7 +1100,7 @@ static void nv_adma_irq_clear(struct ata_port *ap)
 	u32 notifier_clears[2];
 
 	if (pp->flags & NV_ADMA_ATAPI_SETUP_COMPLETE) {
-		ata_sff_irq_clear(ap);
+		ata_bmdma_irq_clear(ap);
 		return;
 	}
 
@@ -1505,7 +1505,7 @@ static irqreturn_t nv_generic_interrupt(int irq, void *dev_instance)
 
 		qc = ata_qc_from_tag(ap, ap->link.active_tag);
 		if (qc && (!(qc->tf.flags & ATA_TFLAG_POLLING))) {
-			handled += ata_sff_host_intr(ap, qc);
+			handled += ata_bmdma_port_intr(ap, qc);
 		} else {
 			/*
 			 * No request pending?  Clear interrupt status
@@ -1669,7 +1669,6 @@ static void nv_mcp55_freeze(struct ata_port *ap)
 	mask = readl(mmio_base + NV_INT_ENABLE_MCP55);
 	mask &= ~(NV_INT_ALL_MCP55 << shift);
 	writel(mask, mmio_base + NV_INT_ENABLE_MCP55);
-	ata_sff_freeze(ap);
 }
 
 static void nv_mcp55_thaw(struct ata_port *ap)
@@ -1683,7 +1682,6 @@ static void nv_mcp55_thaw(struct ata_port *ap)
 	mask = readl(mmio_base + NV_INT_ENABLE_MCP55);
 	mask |= (NV_INT_MASK_MCP55 << shift);
 	writel(mask, mmio_base + NV_INT_ENABLE_MCP55);
-	ata_sff_thaw(ap);
 }
 
 static void nv_adma_error_handler(struct ata_port *ap)
@@ -2430,7 +2428,7 @@ static int nv_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 
 	ppi[0] = &nv_port_info[type];
 	ipriv = ppi[0]->private_data;
-	rc = ata_pci_sff_prepare_host(pdev, ppi, &host);
+	rc = ata_pci_bmdma_prepare_host(pdev, ppi, &host);
 	if (rc)
 		return rc;
 
diff --git a/drivers/ata/sata_qstor.c b/drivers/ata/sata_qstor.c
index d533b3d20ca1..daeebf19a6a9 100644
--- a/drivers/ata/sata_qstor.c
+++ b/drivers/ata/sata_qstor.c
@@ -120,8 +120,6 @@ static void qs_host_stop(struct ata_host *host);
 static void qs_qc_prep(struct ata_queued_cmd *qc);
 static unsigned int qs_qc_issue(struct ata_queued_cmd *qc);
 static int qs_check_atapi_dma(struct ata_queued_cmd *qc);
-static void qs_bmdma_stop(struct ata_queued_cmd *qc);
-static u8 qs_bmdma_status(struct ata_port *ap);
 static void qs_freeze(struct ata_port *ap);
 static void qs_thaw(struct ata_port *ap);
 static int qs_prereset(struct ata_link *link, unsigned long deadline);
@@ -137,8 +135,6 @@ static struct ata_port_operations qs_ata_ops = {
 	.inherits		= &ata_sff_port_ops,
 
 	.check_atapi_dma	= qs_check_atapi_dma,
-	.bmdma_stop		= qs_bmdma_stop,
-	.bmdma_status		= qs_bmdma_status,
 	.qc_prep		= qs_qc_prep,
 	.qc_issue		= qs_qc_issue,
 
@@ -190,16 +186,6 @@ static int qs_check_atapi_dma(struct ata_queued_cmd *qc)
 	return 1;	/* ATAPI DMA not supported */
 }
 
-static void qs_bmdma_stop(struct ata_queued_cmd *qc)
-{
-	/* nothing */
-}
-
-static u8 qs_bmdma_status(struct ata_port *ap)
-{
-	return 0;
-}
-
 static inline void qs_enter_reg_mode(struct ata_port *ap)
 {
 	u8 __iomem *chan = qs_mmio_base(ap->host) + (ap->port_no * 0x4000);
@@ -454,7 +440,7 @@ static inline unsigned int qs_intr_mmio(struct ata_host *host)
 		if (!pp || pp->state != qs_state_mmio)
 			continue;
 		if (!(qc->tf.flags & ATA_TFLAG_POLLING))
-			handled |= ata_sff_host_intr(ap, qc);
+			handled |= ata_sff_port_intr(ap, qc);
 	}
 	return handled;
 }
diff --git a/drivers/ata/sata_sil.c b/drivers/ata/sata_sil.c
index 2dda312b6b9a..3a4f84219719 100644
--- a/drivers/ata/sata_sil.c
+++ b/drivers/ata/sata_sil.c
@@ -503,7 +503,7 @@ static void sil_host_intr(struct ata_port *ap, u32 bmdma2)
 		goto err_hsm;
 
 	/* ack bmdma irq events */
-	ata_sff_irq_clear(ap);
+	ata_bmdma_irq_clear(ap);
 
 	/* kick HSM in the ass */
 	ata_sff_hsm_move(ap, qc, status, 0);
@@ -584,7 +584,7 @@ static void sil_thaw(struct ata_port *ap)
 
 	/* clear IRQ */
 	ap->ops->sff_check_status(ap);
-	ata_sff_irq_clear(ap);
+	ata_bmdma_irq_clear(ap);
 
 	/* turn on SATA IRQ if supported */
 	if (!(ap->flags & SIL_FLAG_NO_SATA_IRQ))
diff --git a/drivers/ata/sata_sil24.c b/drivers/ata/sata_sil24.c
index e9250514734b..be7726d7686d 100644
--- a/drivers/ata/sata_sil24.c
+++ b/drivers/ata/sata_sil24.c
@@ -539,12 +539,12 @@ static void sil24_config_port(struct ata_port *ap)
 		writel(PORT_CS_IRQ_WOC, port + PORT_CTRL_CLR);
 
 	/* zero error counters. */
-	writel(0x8000, port + PORT_DECODE_ERR_THRESH);
-	writel(0x8000, port + PORT_CRC_ERR_THRESH);
-	writel(0x8000, port + PORT_HSHK_ERR_THRESH);
-	writel(0x0000, port + PORT_DECODE_ERR_CNT);
-	writel(0x0000, port + PORT_CRC_ERR_CNT);
-	writel(0x0000, port + PORT_HSHK_ERR_CNT);
+	writew(0x8000, port + PORT_DECODE_ERR_THRESH);
+	writew(0x8000, port + PORT_CRC_ERR_THRESH);
+	writew(0x8000, port + PORT_HSHK_ERR_THRESH);
+	writew(0x0000, port + PORT_DECODE_ERR_CNT);
+	writew(0x0000, port + PORT_CRC_ERR_CNT);
+	writew(0x0000, port + PORT_HSHK_ERR_CNT);
 
 	/* always use 64bit activation */
 	writel(PORT_CS_32BIT_ACTV, port + PORT_CTRL_CLR);
@@ -622,6 +622,11 @@ static int sil24_exec_polled_cmd(struct ata_port *ap, int pmp,
 	irq_enabled = readl(port + PORT_IRQ_ENABLE_SET);
 	writel(PORT_IRQ_COMPLETE | PORT_IRQ_ERROR, port + PORT_IRQ_ENABLE_CLR);
 
+	/*
+	 * The barrier is required to ensure that writes to cmd_block reach
+	 * the memory before the write to PORT_CMD_ACTIVATE.
+	 */
+	wmb();
 	writel((u32)paddr, port + PORT_CMD_ACTIVATE);
 	writel((u64)paddr >> 32, port + PORT_CMD_ACTIVATE + 4);
 
@@ -865,7 +870,7 @@ static void sil24_qc_prep(struct ata_queued_cmd *qc)
 	} else {
 		prb = &cb->atapi.prb;
 		sge = cb->atapi.sge;
-		memset(cb->atapi.cdb, 0, 32);
+		memset(cb->atapi.cdb, 0, sizeof(cb->atapi.cdb));
 		memcpy(cb->atapi.cdb, qc->cdb, qc->dev->cdb_len);
 
 		if (ata_is_data(qc->tf.protocol)) {
@@ -895,6 +900,11 @@ static unsigned int sil24_qc_issue(struct ata_queued_cmd *qc)
 	paddr = pp->cmd_block_dma + tag * sizeof(*pp->cmd_block);
 	activate = port + PORT_CMD_ACTIVATE + tag * 8;
 
+	/*
+	 * The barrier is required to ensure that writes to cmd_block reach
+	 * the memory before the write to PORT_CMD_ACTIVATE.
+	 */
+	wmb();
 	writel((u32)paddr, activate);
 	writel((u64)paddr >> 32, activate + 4);
 
diff --git a/drivers/ata/sata_sis.c b/drivers/ata/sata_sis.c
index f8a91bfd66a8..2bfe3ae03976 100644
--- a/drivers/ata/sata_sis.c
+++ b/drivers/ata/sata_sis.c
@@ -279,7 +279,7 @@ static int sis_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 		break;
 	}
 
-	rc = ata_pci_sff_prepare_host(pdev, ppi, &host);
+	rc = ata_pci_bmdma_prepare_host(pdev, ppi, &host);
 	if (rc)
 		return rc;
 
@@ -308,7 +308,7 @@ static int sis_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 
 	pci_set_master(pdev);
 	pci_intx(pdev, 1);
-	return ata_host_activate(host, pdev->irq, ata_sff_interrupt,
+	return ata_host_activate(host, pdev->irq, ata_bmdma_interrupt,
 				 IRQF_SHARED, &sis_sht);
 }
 
diff --git a/drivers/ata/sata_svw.c b/drivers/ata/sata_svw.c
index 101fd6a19829..7d9db4aaf07e 100644
--- a/drivers/ata/sata_svw.c
+++ b/drivers/ata/sata_svw.c
@@ -502,7 +502,7 @@ static int k2_sata_init_one(struct pci_dev *pdev, const struct pci_device_id *en
 	writel(0x0, mmio_base + K2_SATA_SIM_OFFSET);
 
 	pci_set_master(pdev);
-	return ata_host_activate(host, pdev->irq, ata_sff_interrupt,
+	return ata_host_activate(host, pdev->irq, ata_bmdma_interrupt,
 				 IRQF_SHARED, &k2_sata_sht);
 }
 
diff --git a/drivers/ata/sata_uli.c b/drivers/ata/sata_uli.c
index d8dac17dc2c8..b8578c32d344 100644
--- a/drivers/ata/sata_uli.c
+++ b/drivers/ata/sata_uli.c
@@ -242,7 +242,7 @@ static int uli_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 
 	pci_set_master(pdev);
 	pci_intx(pdev, 1);
-	return ata_host_activate(host, pdev->irq, ata_sff_interrupt,
+	return ata_host_activate(host, pdev->irq, ata_bmdma_interrupt,
 				 IRQF_SHARED, &uli_sht);
 }
 
diff --git a/drivers/ata/sata_via.c b/drivers/ata/sata_via.c
index 08f65492cc81..4730c42a5ee5 100644
--- a/drivers/ata/sata_via.c
+++ b/drivers/ata/sata_via.c
@@ -308,7 +308,7 @@ static void svia_noop_freeze(struct ata_port *ap)
 	 * certain way.  Leave it alone and just clear pending IRQ.
 	 */
 	ap->ops->sff_check_status(ap);
-	ata_sff_irq_clear(ap);
+	ata_bmdma_irq_clear(ap);
 }
 
 /**
@@ -463,7 +463,7 @@ static int vt6420_prepare_host(struct pci_dev *pdev, struct ata_host **r_host)
 	struct ata_host *host;
 	int rc;
 
-	rc = ata_pci_sff_prepare_host(pdev, ppi, &host);
+	rc = ata_pci_bmdma_prepare_host(pdev, ppi, &host);
 	if (rc)
 		return rc;
 	*r_host = host;
@@ -520,7 +520,7 @@ static int vt8251_prepare_host(struct pci_dev *pdev, struct ata_host **r_host)
 	struct ata_host *host;
 	int i, rc;
 
-	rc = ata_pci_sff_prepare_host(pdev, ppi, &host);
+	rc = ata_pci_bmdma_prepare_host(pdev, ppi, &host);
 	if (rc)
 		return rc;
 	*r_host = host;
@@ -575,6 +575,33 @@ static void svia_configure(struct pci_dev *pdev)
 		tmp8 |= NATIVE_MODE_ALL;
 		pci_write_config_byte(pdev, SATA_NATIVE_MODE, tmp8);
 	}
+
+	/*
+	 * vt6421 has problems talking to some drives.  The following
+	 * is the fix from Joseph Chan <JosephChan@via.com.tw>.
+	 *
+	 * When host issues HOLD, device may send up to 20DW of data
+	 * before acknowledging it with HOLDA and the host should be
+	 * able to buffer them in FIFO.  Unfortunately, some WD drives
+	 * send upto 40DW before acknowledging HOLD and, in the
+	 * default configuration, this ends up overflowing vt6421's
+	 * FIFO, making the controller abort the transaction with
+	 * R_ERR.
+	 *
+	 * Rx52[2] is the internal 128DW FIFO Flow control watermark
+	 * adjusting mechanism enable bit and the default value 0
+	 * means host will issue HOLD to device when the left FIFO
+	 * size goes below 32DW.  Setting it to 1 makes the watermark
+	 * 64DW.
+	 *
+	 * https://bugzilla.kernel.org/show_bug.cgi?id=15173
+	 * http://article.gmane.org/gmane.linux.ide/46352
+	 */
+	if (pdev->device == 0x3249) {
+		pci_read_config_byte(pdev, 0x52, &tmp8);
+		tmp8 |= 1 << 2;
+		pci_write_config_byte(pdev, 0x52, tmp8);
+	}
 }
 
 static int svia_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
@@ -628,7 +655,7 @@ static int svia_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	svia_configure(pdev);
 
 	pci_set_master(pdev);
-	return ata_host_activate(host, pdev->irq, ata_sff_interrupt,
+	return ata_host_activate(host, pdev->irq, ata_bmdma_interrupt,
 				 IRQF_SHARED, &svia_sht);
 }
 
diff --git a/drivers/ata/sata_vsc.c b/drivers/ata/sata_vsc.c
index 2107952ebff1..b777176ff494 100644
--- a/drivers/ata/sata_vsc.c
+++ b/drivers/ata/sata_vsc.c
@@ -245,7 +245,7 @@ static void vsc_port_intr(u8 port_status, struct ata_port *ap)
 
 	qc = ata_qc_from_tag(ap, ap->link.active_tag);
 	if (qc && likely(!(qc->tf.flags & ATA_TFLAG_POLLING)))
-		handled = ata_sff_host_intr(ap, qc);
+		handled = ata_bmdma_port_intr(ap, qc);
 
 	/* We received an interrupt during a polled command,
 	 * or some other spurious condition.  Interrupt reporting
diff --git a/drivers/atm/fore200e.c b/drivers/atm/fore200e.c
index f7d6ebaa0418..da8f176c051e 100644
--- a/drivers/atm/fore200e.c
+++ b/drivers/atm/fore200e.c
@@ -789,7 +789,7 @@ static int __init fore200e_sba_map(struct fore200e *fore200e)
 	fore200e->bus->write(0x02, fore200e->regs.sba.isr); /* XXX hardwired interrupt level */
 
 	/* get the supported DVMA burst sizes */
-	bursts = of_getintprop_default(op->node->parent, "burst-sizes", 0x00);
+	bursts = of_getintprop_default(op->dev.of_node->parent, "burst-sizes", 0x00);
 
 	if (sbus_can_dma_64bit())
 		sbus_set_sbus64(&op->dev, bursts);
@@ -820,18 +820,20 @@ static int __init fore200e_sba_prom_read(struct fore200e *fore200e, struct prom_
 	const u8 *prop;
 	int len;
 
-	prop = of_get_property(op->node, "madaddrlo2", &len);
+	prop = of_get_property(op->dev.of_node, "madaddrlo2", &len);
 	if (!prop)
 		return -ENODEV;
 	memcpy(&prom->mac_addr[4], prop, 4);
 
-	prop = of_get_property(op->node, "madaddrhi4", &len);
+	prop = of_get_property(op->dev.of_node, "madaddrhi4", &len);
 	if (!prop)
 		return -ENODEV;
 	memcpy(&prom->mac_addr[2], prop, 4);
 
-	prom->serial_number = of_getintprop_default(op->node, "serialnumber", 0);
-	prom->hw_revision = of_getintprop_default(op->node, "promversion", 0);
+	prom->serial_number = of_getintprop_default(op->dev.of_node,
+						    "serialnumber", 0);
+	prom->hw_revision = of_getintprop_default(op->dev.of_node,
+						  "promversion", 0);
     
 	return 0;
 }
@@ -841,10 +843,10 @@ static int fore200e_sba_proc_read(struct fore200e *fore200e, char *page)
 	struct of_device *op = fore200e->bus_dev;
 	const struct linux_prom_registers *regs;
 
-	regs = of_get_property(op->node, "reg", NULL);
+	regs = of_get_property(op->dev.of_node, "reg", NULL);
 
 	return sprintf(page, "   SBUS slot/device:\t\t%d/'%s'\n",
-		       (regs ? regs->which_io : 0), op->node->name);
+		       (regs ? regs->which_io : 0), op->dev.of_node->name);
 }
 #endif /* CONFIG_SBUS */
 
@@ -2693,8 +2695,11 @@ static const struct of_device_id fore200e_sba_match[] = {
 MODULE_DEVICE_TABLE(of, fore200e_sba_match);
 
 static struct of_platform_driver fore200e_sba_driver = {
-	.name		= "fore_200e",
-	.match_table	= fore200e_sba_match,
+	.driver = {
+		.name = "fore_200e",
+		.owner = THIS_MODULE,
+		.of_match_table = fore200e_sba_match,
+	},
 	.probe		= fore200e_sba_probe,
 	.remove		= __devexit_p(fore200e_sba_remove),
 };
diff --git a/drivers/auxdisplay/cfag12864bfb.c b/drivers/auxdisplay/cfag12864bfb.c
index 3fecfb446d90..5ad3bad2b0a5 100644
--- a/drivers/auxdisplay/cfag12864bfb.c
+++ b/drivers/auxdisplay/cfag12864bfb.c
@@ -37,7 +37,7 @@
 
 #define CFAG12864BFB_NAME "cfag12864bfb"
 
-static struct fb_fix_screeninfo cfag12864bfb_fix __initdata = {
+static struct fb_fix_screeninfo cfag12864bfb_fix __devinitdata = {
 	.id = "cfag12864b",
 	.type = FB_TYPE_PACKED_PIXELS,
 	.visual = FB_VISUAL_MONO10,
@@ -48,7 +48,7 @@ static struct fb_fix_screeninfo cfag12864bfb_fix __initdata = {
 	.accel = FB_ACCEL_NONE,
 };
 
-static struct fb_var_screeninfo cfag12864bfb_var __initdata = {
+static struct fb_var_screeninfo cfag12864bfb_var __devinitdata = {
 	.xres = CFAG12864B_WIDTH,
 	.yres = CFAG12864B_HEIGHT,
 	.xres_virtual = CFAG12864B_WIDTH,
@@ -114,7 +114,7 @@ none:
 	return ret;
 }
 
-static int cfag12864bfb_remove(struct platform_device *device)
+static int __devexit cfag12864bfb_remove(struct platform_device *device)
 {
 	struct fb_info *info = platform_get_drvdata(device);
 
@@ -128,7 +128,7 @@ static int cfag12864bfb_remove(struct platform_device *device)
 
 static struct platform_driver cfag12864bfb_driver = {
 	.probe	= cfag12864bfb_probe,
-	.remove = cfag12864bfb_remove,
+	.remove = __devexit_p(cfag12864bfb_remove),
 	.driver = {
 		.name	= CFAG12864BFB_NAME,
 	},
diff --git a/drivers/base/node.c b/drivers/base/node.c
index 057979a19eea..2bdd8a94ec94 100644
--- a/drivers/base/node.c
+++ b/drivers/base/node.c
@@ -9,6 +9,7 @@
 #include <linux/memory.h>
 #include <linux/node.h>
 #include <linux/hugetlb.h>
+#include <linux/compaction.h>
 #include <linux/cpumask.h>
 #include <linux/topology.h>
 #include <linux/nodemask.h>
@@ -246,6 +247,8 @@ int register_node(struct node *node, int num, struct node *parent)
 		scan_unevictable_register_node(node);
 
 		hugetlb_register_node(node);
+
+		compaction_register_node(node);
 	}
 	return error;
 }
diff --git a/drivers/base/topology.c b/drivers/base/topology.c
index bf6b13206d00..9fc630ce1ddb 100644
--- a/drivers/base/topology.c
+++ b/drivers/base/topology.c
@@ -162,7 +162,7 @@ static int __cpuinit topology_cpu_callback(struct notifier_block *nfb,
 		topology_remove_dev(cpu);
 		break;
 	}
-	return rc ? NOTIFY_BAD : NOTIFY_OK;
+	return notifier_from_errno(rc);
 }
 
 static int __cpuinit topology_sysfs_init(void)
diff --git a/drivers/block/Kconfig b/drivers/block/Kconfig
index 77bfce52e9ca..de277689da61 100644
--- a/drivers/block/Kconfig
+++ b/drivers/block/Kconfig
@@ -76,6 +76,17 @@ config BLK_DEV_XD
 
 	  It's pretty unlikely that you have one of these: say N.
 
+config GDROM
+	tristate "SEGA Dreamcast GD-ROM drive"
+	depends on SH_DREAMCAST
+	help
+	  A standard SEGA Dreamcast comes with a modified CD ROM drive called a
+	  "GD-ROM" by SEGA to signify it is capable of reading special disks
+	  with up to 1 GB of data. This drive will also read standard CD ROM
+	  disks. Select this option to access any disks in your GD ROM drive.
+	  Most users will want to say "Y" here.
+	  You can also build this as a module which will be called gdrom.
+
 config PARIDE
 	tristate "Parallel port IDE device support"
 	depends on PARPORT_PC
@@ -103,17 +114,6 @@ config PARIDE
 	  "MicroSolutions backpack protocol", "DataStor Commuter protocol"
 	  etc.).
 
-config GDROM
-	tristate "SEGA Dreamcast GD-ROM drive"
-	depends on SH_DREAMCAST
-	help
-	  A standard SEGA Dreamcast comes with a modified CD ROM drive called a
-	  "GD-ROM" by SEGA to signify it is capable of reading special disks
-	  with up to 1 GB of data. This drive will also read standard CD ROM
-	  disks. Select this option to access any disks in your GD ROM drive.
-	  Most users will want to say "Y" here.
-	  You can also build this as a module which will be called gdrom.
-
 source "drivers/block/paride/Kconfig"
 
 config BLK_CPQ_DA
diff --git a/drivers/block/brd.c b/drivers/block/brd.c
index 6081e81d5738..f1bf79d9bc0a 100644
--- a/drivers/block/brd.c
+++ b/drivers/block/brd.c
@@ -133,6 +133,28 @@ static struct page *brd_insert_page(struct brd_device *brd, sector_t sector)
 	return page;
 }
 
+static void brd_free_page(struct brd_device *brd, sector_t sector)
+{
+	struct page *page;
+	pgoff_t idx;
+
+	spin_lock(&brd->brd_lock);
+	idx = sector >> PAGE_SECTORS_SHIFT;
+	page = radix_tree_delete(&brd->brd_pages, idx);
+	spin_unlock(&brd->brd_lock);
+	if (page)
+		__free_page(page);
+}
+
+static void brd_zero_page(struct brd_device *brd, sector_t sector)
+{
+	struct page *page;
+
+	page = brd_lookup_page(brd, sector);
+	if (page)
+		clear_highpage(page);
+}
+
 /*
  * Free all backing store pages and radix tree. This must only be called when
  * there are no other users of the device.
@@ -189,6 +211,24 @@ static int copy_to_brd_setup(struct brd_device *brd, sector_t sector, size_t n)
 	return 0;
 }
 
+static void discard_from_brd(struct brd_device *brd,
+			sector_t sector, size_t n)
+{
+	while (n >= PAGE_SIZE) {
+		/*
+		 * Don't want to actually discard pages here because
+		 * re-allocating the pages can result in writeback
+		 * deadlocks under heavy load.
+		 */
+		if (0)
+			brd_free_page(brd, sector);
+		else
+			brd_zero_page(brd, sector);
+		sector += PAGE_SIZE >> SECTOR_SHIFT;
+		n -= PAGE_SIZE;
+	}
+}
+
 /*
  * Copy n bytes from src to the brd starting at sector. Does not sleep.
  */
@@ -300,6 +340,12 @@ static int brd_make_request(struct request_queue *q, struct bio *bio)
 						get_capacity(bdev->bd_disk))
 		goto out;
 
+	if (unlikely(bio_rw_flagged(bio, BIO_RW_DISCARD))) {
+		err = 0;
+		discard_from_brd(brd, sector, bio->bi_size);
+		goto out;
+	}
+
 	rw = bio_rw(bio);
 	if (rw == READA)
 		rw = READ;
@@ -320,7 +366,7 @@ out:
 }
 
 #ifdef CONFIG_BLK_DEV_XIP
-static int brd_direct_access (struct block_device *bdev, sector_t sector,
+static int brd_direct_access(struct block_device *bdev, sector_t sector,
 			void **kaddr, unsigned long *pfn)
 {
 	struct brd_device *brd = bdev->bd_disk->private_data;
@@ -437,6 +483,11 @@ static struct brd_device *brd_alloc(int i)
 	blk_queue_max_hw_sectors(brd->brd_queue, 1024);
 	blk_queue_bounce_limit(brd->brd_queue, BLK_BOUNCE_ANY);
 
+	brd->brd_queue->limits.discard_granularity = PAGE_SIZE;
+	brd->brd_queue->limits.max_discard_sectors = UINT_MAX;
+	brd->brd_queue->limits.discard_zeroes_data = 1;
+	queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, brd->brd_queue);
+
 	disk = brd->brd_disk = alloc_disk(1 << part_shift);
 	if (!disk)
 		goto out_free_queue;
diff --git a/drivers/block/cciss_scsi.c b/drivers/block/cciss_scsi.c
index e1d0e2cfec72..3381505c8a6c 100644
--- a/drivers/block/cciss_scsi.c
+++ b/drivers/block/cciss_scsi.c
@@ -188,11 +188,11 @@ scsi_cmd_free(ctlr_info_t *h, CommandList_struct *cmd)
 
 	sa = h->scsi_ctlr;
 	stk = &sa->cmd_stack; 
+	stk->top++;
 	if (stk->top >= CMD_STACK_SIZE) {
 		printk("cciss: scsi_cmd_free called too many times.\n");
 		BUG();
 	}
-	stk->top++;
 	stk->elem[stk->top] = (struct cciss_scsi_cmd_stack_elem_t *) cmd;
 }
 
diff --git a/drivers/block/drbd/drbd_bitmap.c b/drivers/block/drbd/drbd_bitmap.c
index 3390716898d5..e3f88d6e1412 100644
--- a/drivers/block/drbd/drbd_bitmap.c
+++ b/drivers/block/drbd/drbd_bitmap.c
@@ -84,6 +84,9 @@ struct drbd_bitmap {
 #define BM_MD_IO_ERROR  1
 #define BM_P_VMALLOCED  2
 
+static int __bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s,
+			       unsigned long e, int val, const enum km_type km);
+
 static int bm_is_locked(struct drbd_bitmap *b)
 {
 	return test_bit(BM_LOCKED, &b->bm_flags);
@@ -441,7 +444,7 @@ static void bm_memset(struct drbd_bitmap *b, size_t offset, int c, size_t len)
  * In case this is actually a resize, we copy the old bitmap into the new one.
  * Otherwise, the bitmap is initialized to all bits set.
  */
-int drbd_bm_resize(struct drbd_conf *mdev, sector_t capacity)
+int drbd_bm_resize(struct drbd_conf *mdev, sector_t capacity, int set_new_bits)
 {
 	struct drbd_bitmap *b = mdev->bitmap;
 	unsigned long bits, words, owords, obits, *p_addr, *bm;
@@ -516,7 +519,7 @@ int drbd_bm_resize(struct drbd_conf *mdev, sector_t capacity)
 	obits  = b->bm_bits;
 
 	growing = bits > obits;
-	if (opages)
+	if (opages && growing && set_new_bits)
 		bm_set_surplus(b);
 
 	b->bm_pages = npages;
@@ -526,8 +529,12 @@ int drbd_bm_resize(struct drbd_conf *mdev, sector_t capacity)
 	b->bm_dev_capacity = capacity;
 
 	if (growing) {
-		bm_memset(b, owords, 0xff, words-owords);
-		b->bm_set += bits - obits;
+		if (set_new_bits) {
+			bm_memset(b, owords, 0xff, words-owords);
+			b->bm_set += bits - obits;
+		} else
+			bm_memset(b, owords, 0x00, words-owords);
+
 	}
 
 	if (want < have) {
@@ -773,7 +780,7 @@ static void bm_page_io_async(struct drbd_conf *mdev, struct drbd_bitmap *b, int
 	/* nothing to do, on disk == in memory */
 # define bm_cpu_to_lel(x) ((void)0)
 # else
-void bm_cpu_to_lel(struct drbd_bitmap *b)
+static void bm_cpu_to_lel(struct drbd_bitmap *b)
 {
 	/* need to cpu_to_lel all the pages ...
 	 * this may be optimized by using
@@ -1015,7 +1022,7 @@ unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_f
  * wants bitnr, not sector.
  * expected to be called for only a few bits (e - s about BITS_PER_LONG).
  * Must hold bitmap lock already. */
-int __bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s,
+static int __bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s,
 	unsigned long e, int val, const enum km_type km)
 {
 	struct drbd_bitmap *b = mdev->bitmap;
@@ -1053,7 +1060,7 @@ int __bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s,
  * for val != 0, we change 0 -> 1, return code positive
  * for val == 0, we change 1 -> 0, return code negative
  * wants bitnr, not sector */
-int bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s,
+static int bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s,
 	const unsigned long e, int val)
 {
 	unsigned long flags;
diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index e5e86a781820..485ed8c7d623 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -132,6 +132,7 @@ enum {
 	DRBD_FAULT_DT_RA = 6,	/* data read ahead */
 	DRBD_FAULT_BM_ALLOC = 7,	/* bitmap allocation */
 	DRBD_FAULT_AL_EE = 8,	/* alloc ee */
+	DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */
 
 	DRBD_FAULT_MAX,
 };
@@ -208,8 +209,11 @@ enum drbd_packets {
 	P_RS_IS_IN_SYNC	      = 0x22, /* meta socket */
 	P_SYNC_PARAM89	      = 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */
 	P_COMPRESSED_BITMAP   = 0x24, /* compressed or otherwise encoded bitmap transfer */
+	/* P_CKPT_FENCE_REQ      = 0x25, * currently reserved for protocol D */
+	/* P_CKPT_DISABLE_REQ    = 0x26, * currently reserved for protocol D */
+	P_DELAY_PROBE         = 0x27, /* is used on BOTH sockets */
 
-	P_MAX_CMD	      = 0x25,
+	P_MAX_CMD	      = 0x28,
 	P_MAY_IGNORE	      = 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
 	P_MAX_OPT_CMD	      = 0x101,
 
@@ -264,6 +268,7 @@ static inline const char *cmdname(enum drbd_packets cmd)
 		[P_CSUM_RS_REQUEST]     = "CsumRSRequest",
 		[P_RS_IS_IN_SYNC]	= "CsumRSIsInSync",
 		[P_COMPRESSED_BITMAP]   = "CBitmap",
+		[P_DELAY_PROBE]         = "DelayProbe",
 		[P_MAX_CMD]	        = NULL,
 	};
 
@@ -481,7 +486,8 @@ struct p_sizes {
 	u64	    u_size;  /* user requested size */
 	u64	    c_size;  /* current exported size */
 	u32	    max_segment_size;  /* Maximal size of a BIO */
-	u32	    queue_order_type;
+	u16	    queue_order_type;  /* not yet implemented in DRBD*/
+	u16	    dds_flags; /* use enum dds_flags here. */
 } __packed;
 
 struct p_state {
@@ -538,6 +544,18 @@ struct p_compressed_bm {
 	u8 code[0];
 } __packed;
 
+struct p_delay_probe {
+	struct p_header head;
+	u32	seq_num; /* sequence number to match the two probe packets */
+	u32	offset;	 /* usecs the probe got sent after the reference time point */
+} __packed;
+
+struct delay_probe {
+	struct list_head list;
+	unsigned int seq_num;
+	struct timeval time;
+};
+
 /* DCBP: Drbd Compressed Bitmap Packet ... */
 static inline enum drbd_bitmap_code
 DCBP_get_code(struct p_compressed_bm *p)
@@ -722,22 +740,6 @@ enum epoch_event {
 	EV_CLEANUP = 32, /* used as flag */
 };
 
-struct drbd_epoch_entry {
-	struct drbd_work    w;
-	struct drbd_conf *mdev;
-	struct bio *private_bio;
-	struct hlist_node colision;
-	sector_t sector;
-	unsigned int size;
-	struct drbd_epoch *epoch;
-
-	/* up to here, the struct layout is identical to drbd_request;
-	 * we might be able to use that to our advantage...  */
-
-	unsigned int flags;
-	u64    block_id;
-};
-
 struct drbd_wq_barrier {
 	struct drbd_work w;
 	struct completion done;
@@ -748,17 +750,49 @@ struct digest_info {
 	void *digest;
 };
 
-/* ee flag bits */
+struct drbd_epoch_entry {
+	struct drbd_work w;
+	struct hlist_node colision;
+	struct drbd_epoch *epoch;
+	struct drbd_conf *mdev;
+	struct page *pages;
+	atomic_t pending_bios;
+	unsigned int size;
+	/* see comments on ee flag bits below */
+	unsigned long flags;
+	sector_t sector;
+	u64 block_id;
+};
+
+/* ee flag bits.
+ * While corresponding bios are in flight, the only modification will be
+ * set_bit WAS_ERROR, which has to be atomic.
+ * If no bios are in flight yet, or all have been completed,
+ * non-atomic modification to ee->flags is ok.
+ */
 enum {
 	__EE_CALL_AL_COMPLETE_IO,
-	__EE_CONFLICT_PENDING,
 	__EE_MAY_SET_IN_SYNC,
+
+	/* This epoch entry closes an epoch using a barrier.
+	 * On sucessful completion, the epoch is released,
+	 * and the P_BARRIER_ACK send. */
 	__EE_IS_BARRIER,
+
+	/* In case a barrier failed,
+	 * we need to resubmit without the barrier flag. */
+	__EE_RESUBMITTED,
+
+	/* we may have several bios per epoch entry.
+	 * if any of those fail, we set this flag atomically
+	 * from the endio callback */
+	__EE_WAS_ERROR,
 };
 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
-#define EE_CONFLICT_PENDING    (1<<__EE_CONFLICT_PENDING)
 #define EE_MAY_SET_IN_SYNC     (1<<__EE_MAY_SET_IN_SYNC)
 #define EE_IS_BARRIER          (1<<__EE_IS_BARRIER)
+#define	EE_RESUBMITTED         (1<<__EE_RESUBMITTED)
+#define EE_WAS_ERROR           (1<<__EE_WAS_ERROR)
 
 /* global flag bits */
 enum {
@@ -908,9 +942,11 @@ struct drbd_conf {
 	unsigned int ko_count;
 	struct drbd_work  resync_work,
 			  unplug_work,
-			  md_sync_work;
+			  md_sync_work,
+			  delay_probe_work;
 	struct timer_list resync_timer;
 	struct timer_list md_sync_timer;
+	struct timer_list delay_probe_timer;
 
 	/* Used after attach while negotiating new disk state. */
 	union drbd_state new_state_tmp;
@@ -1026,6 +1062,12 @@ struct drbd_conf {
 	u64 ed_uuid; /* UUID of the exposed data */
 	struct mutex state_mutex;
 	char congestion_reason;  /* Why we where congested... */
+	struct list_head delay_probes; /* protected by peer_seq_lock */
+	int data_delay;   /* Delay of packets on the data-sock behind meta-sock */
+	unsigned int delay_seq; /* To generate sequence numbers of delay probes */
+	struct timeval dps_time; /* delay-probes-start-time */
+	unsigned int dp_volume_last;  /* send_cnt of last delay probe */
+	int c_sync_rate; /* current resync rate after delay_probe magic */
 };
 
 static inline struct drbd_conf *minor_to_mdev(unsigned int minor)
@@ -1081,6 +1123,11 @@ enum chg_state_flags {
 	CS_ORDERED      = CS_WAIT_COMPLETE + CS_SERIALIZE,
 };
 
+enum dds_flags {
+	DDSF_FORCED    = 1,
+	DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
+};
+
 extern void drbd_init_set_defaults(struct drbd_conf *mdev);
 extern int drbd_change_state(struct drbd_conf *mdev, enum chg_state_flags f,
 			union drbd_state mask, union drbd_state val);
@@ -1113,7 +1160,7 @@ extern int drbd_send_protocol(struct drbd_conf *mdev);
 extern int drbd_send_uuids(struct drbd_conf *mdev);
 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev);
 extern int drbd_send_sync_uuid(struct drbd_conf *mdev, u64 val);
-extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply);
+extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags);
 extern int _drbd_send_state(struct drbd_conf *mdev);
 extern int drbd_send_state(struct drbd_conf *mdev);
 extern int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock,
@@ -1311,7 +1358,7 @@ struct bm_extent {
 #define APP_R_HSIZE 15
 
 extern int  drbd_bm_init(struct drbd_conf *mdev);
-extern int  drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors);
+extern int  drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors, int set_new_bits);
 extern void drbd_bm_cleanup(struct drbd_conf *mdev);
 extern void drbd_bm_set_all(struct drbd_conf *mdev);
 extern void drbd_bm_clear_all(struct drbd_conf *mdev);
@@ -1383,7 +1430,7 @@ extern void drbd_resume_io(struct drbd_conf *mdev);
 extern char *ppsize(char *buf, unsigned long long size);
 extern sector_t drbd_new_dev_size(struct drbd_conf *, struct drbd_backing_dev *, int);
 enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 };
-extern enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *, int force) __must_hold(local);
+extern enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *, enum dds_flags) __must_hold(local);
 extern void resync_after_online_grow(struct drbd_conf *);
 extern void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int) __must_hold(local);
 extern int drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role,
@@ -1414,7 +1461,8 @@ static inline void ov_oos_print(struct drbd_conf *mdev)
 }
 
 
-extern void drbd_csum(struct drbd_conf *, struct crypto_hash *, struct bio *, void *);
+extern void drbd_csum_bio(struct drbd_conf *, struct crypto_hash *, struct bio *, void *);
+extern void drbd_csum_ee(struct drbd_conf *, struct crypto_hash *, struct drbd_epoch_entry *, void *);
 /* worker callbacks */
 extern int w_req_cancel_conflict(struct drbd_conf *, struct drbd_work *, int);
 extern int w_read_retry_remote(struct drbd_conf *, struct drbd_work *, int);
@@ -1426,7 +1474,6 @@ extern int w_e_end_ov_req(struct drbd_conf *, struct drbd_work *, int);
 extern int w_ov_finished(struct drbd_conf *, struct drbd_work *, int);
 extern int w_resync_inactive(struct drbd_conf *, struct drbd_work *, int);
 extern int w_resume_next_sg(struct drbd_conf *, struct drbd_work *, int);
-extern int w_io_error(struct drbd_conf *, struct drbd_work *, int);
 extern int w_send_write_hint(struct drbd_conf *, struct drbd_work *, int);
 extern int w_make_resync_request(struct drbd_conf *, struct drbd_work *, int);
 extern int w_send_dblock(struct drbd_conf *, struct drbd_work *, int);
@@ -1438,6 +1485,8 @@ extern int w_e_reissue(struct drbd_conf *, struct drbd_work *, int);
 extern void resync_timer_fn(unsigned long data);
 
 /* drbd_receiver.c */
+extern int drbd_submit_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e,
+		const unsigned rw, const int fault_type);
 extern int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list);
 extern struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev,
 					    u64 id,
@@ -1490,7 +1539,7 @@ static inline void drbd_tcp_nodelay(struct socket *sock)
 
 static inline void drbd_tcp_quickack(struct socket *sock)
 {
-	int __user val = 1;
+	int __user val = 2;
 	(void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
 			(char __user *)&val, sizeof(val));
 }
@@ -1593,6 +1642,41 @@ void drbd_bcast_ee(struct drbd_conf *mdev,
  * inline helper functions
  *************************/
 
+/* see also page_chain_add and friends in drbd_receiver.c */
+static inline struct page *page_chain_next(struct page *page)
+{
+	return (struct page *)page_private(page);
+}
+#define page_chain_for_each(page) \
+	for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
+			page = page_chain_next(page))
+#define page_chain_for_each_safe(page, n) \
+	for (; page && ({ n = page_chain_next(page); 1; }); page = n)
+
+static inline int drbd_bio_has_active_page(struct bio *bio)
+{
+	struct bio_vec *bvec;
+	int i;
+
+	__bio_for_each_segment(bvec, bio, i, 0) {
+		if (page_count(bvec->bv_page) > 1)
+			return 1;
+	}
+
+	return 0;
+}
+
+static inline int drbd_ee_has_active_page(struct drbd_epoch_entry *e)
+{
+	struct page *page = e->pages;
+	page_chain_for_each(page) {
+		if (page_count(page) > 1)
+			return 1;
+	}
+	return 0;
+}
+
+
 static inline void drbd_state_lock(struct drbd_conf *mdev)
 {
 	wait_event(mdev->misc_wait,
@@ -1641,7 +1725,7 @@ static inline void __drbd_chk_io_error_(struct drbd_conf *mdev, int forcedetach,
 	switch (mdev->ldev->dc.on_io_error) {
 	case EP_PASS_ON:
 		if (!forcedetach) {
-			if (printk_ratelimit())
+			if (__ratelimit(&drbd_ratelimit_state))
 				dev_err(DEV, "Local IO failed in %s."
 					     "Passing error on...\n", where);
 			break;
@@ -2138,7 +2222,7 @@ static inline int __inc_ap_bio_cond(struct drbd_conf *mdev)
 /* I'd like to use wait_event_lock_irq,
  * but I'm not sure when it got introduced,
  * and not sure when it has 3 or 4 arguments */
-static inline void inc_ap_bio(struct drbd_conf *mdev, int one_or_two)
+static inline void inc_ap_bio(struct drbd_conf *mdev, int count)
 {
 	/* compare with after_state_ch,
 	 * os.conn != C_WF_BITMAP_S && ns.conn == C_WF_BITMAP_S */
@@ -2160,7 +2244,7 @@ static inline void inc_ap_bio(struct drbd_conf *mdev, int one_or_two)
 		finish_wait(&mdev->misc_wait, &wait);
 		spin_lock_irq(&mdev->req_lock);
 	}
-	atomic_add(one_or_two, &mdev->ap_bio_cnt);
+	atomic_add(count, &mdev->ap_bio_cnt);
 	spin_unlock_irq(&mdev->req_lock);
 }
 
@@ -2251,7 +2335,8 @@ static inline void drbd_md_flush(struct drbd_conf *mdev)
 	if (test_bit(MD_NO_BARRIER, &mdev->flags))
 		return;
 
-	r = blkdev_issue_flush(mdev->ldev->md_bdev, NULL);
+	r = blkdev_issue_flush(mdev->ldev->md_bdev, GFP_KERNEL, NULL,
+			BLKDEV_IFL_WAIT);
 	if (r) {
 		set_bit(MD_NO_BARRIER, &mdev->flags);
 		dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r);
diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c
index 93d1f9b469d4..6b077f93acc6 100644
--- a/drivers/block/drbd/drbd_main.c
+++ b/drivers/block/drbd/drbd_main.c
@@ -684,6 +684,9 @@ static int is_valid_state(struct drbd_conf *mdev, union drbd_state ns)
 	else if (ns.conn > C_CONNECTED && ns.pdsk < D_INCONSISTENT)
 		rv = SS_NO_REMOTE_DISK;
 
+	else if (ns.conn > C_CONNECTED && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE)
+		rv = SS_NO_UP_TO_DATE_DISK;
+
 	else if ((ns.conn == C_CONNECTED ||
 		  ns.conn == C_WF_BITMAP_S ||
 		  ns.conn == C_SYNC_SOURCE ||
@@ -840,7 +843,12 @@ static union drbd_state sanitize_state(struct drbd_conf *mdev, union drbd_state
 			break;
 		case C_WF_BITMAP_S:
 		case C_PAUSED_SYNC_S:
-			ns.pdsk = D_OUTDATED;
+			/* remap any consistent state to D_OUTDATED,
+			 * but disallow "upgrade" of not even consistent states.
+			 */
+			ns.pdsk =
+				(D_DISKLESS < os.pdsk && os.pdsk < D_OUTDATED)
+				? os.pdsk : D_OUTDATED;
 			break;
 		case C_SYNC_SOURCE:
 			ns.pdsk = D_INCONSISTENT;
@@ -1205,8 +1213,6 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os,
 	&&  (ns.pdsk < D_INCONSISTENT ||
 	     ns.pdsk == D_UNKNOWN ||
 	     ns.pdsk == D_OUTDATED)) {
-		kfree(mdev->p_uuid);
-		mdev->p_uuid = NULL;
 		if (get_ldev(mdev)) {
 			if ((ns.role == R_PRIMARY || ns.peer == R_PRIMARY) &&
 			    mdev->ldev->md.uuid[UI_BITMAP] == 0 && ns.disk >= D_UP_TO_DATE) {
@@ -1232,7 +1238,7 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os,
 	    os.disk == D_ATTACHING && ns.disk == D_NEGOTIATING) {
 		kfree(mdev->p_uuid); /* We expect to receive up-to-date UUIDs soon. */
 		mdev->p_uuid = NULL; /* ...to not use the old ones in the mean time */
-		drbd_send_sizes(mdev, 0);  /* to start sync... */
+		drbd_send_sizes(mdev, 0, 0);  /* to start sync... */
 		drbd_send_uuids(mdev);
 		drbd_send_state(mdev);
 	}
@@ -1755,7 +1761,7 @@ int drbd_send_sync_uuid(struct drbd_conf *mdev, u64 val)
 			     (struct p_header *)&p, sizeof(p));
 }
 
-int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply)
+int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags)
 {
 	struct p_sizes p;
 	sector_t d_size, u_size;
@@ -1767,7 +1773,6 @@ int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply)
 		d_size = drbd_get_max_capacity(mdev->ldev);
 		u_size = mdev->ldev->dc.disk_size;
 		q_order_type = drbd_queue_order_type(mdev);
-		p.queue_order_type = cpu_to_be32(drbd_queue_order_type(mdev));
 		put_ldev(mdev);
 	} else {
 		d_size = 0;
@@ -1779,7 +1784,8 @@ int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply)
 	p.u_size = cpu_to_be64(u_size);
 	p.c_size = cpu_to_be64(trigger_reply ? 0 : drbd_get_capacity(mdev->this_bdev));
 	p.max_segment_size = cpu_to_be32(queue_max_segment_size(mdev->rq_queue));
-	p.queue_order_type = cpu_to_be32(q_order_type);
+	p.queue_order_type = cpu_to_be16(q_order_type);
+	p.dds_flags = cpu_to_be16(flags);
 
 	ok = drbd_send_cmd(mdev, USE_DATA_SOCKET, P_SIZES,
 			   (struct p_header *)&p, sizeof(p));
@@ -2180,6 +2186,43 @@ int drbd_send_ov_request(struct drbd_conf *mdev, sector_t sector, int size)
 	return ok;
 }
 
+static int drbd_send_delay_probe(struct drbd_conf *mdev, struct drbd_socket *ds)
+{
+	struct p_delay_probe dp;
+	int offset, ok = 0;
+	struct timeval now;
+
+	mutex_lock(&ds->mutex);
+	if (likely(ds->socket)) {
+		do_gettimeofday(&now);
+		offset = now.tv_usec - mdev->dps_time.tv_usec +
+			 (now.tv_sec - mdev->dps_time.tv_sec) * 1000000;
+		dp.seq_num  = cpu_to_be32(mdev->delay_seq);
+		dp.offset   = cpu_to_be32(offset);
+
+		ok = _drbd_send_cmd(mdev, ds->socket, P_DELAY_PROBE,
+				    (struct p_header *)&dp, sizeof(dp), 0);
+	}
+	mutex_unlock(&ds->mutex);
+
+	return ok;
+}
+
+static int drbd_send_delay_probes(struct drbd_conf *mdev)
+{
+	int ok;
+
+	mdev->delay_seq++;
+	do_gettimeofday(&mdev->dps_time);
+	ok = drbd_send_delay_probe(mdev, &mdev->meta);
+	ok = ok && drbd_send_delay_probe(mdev, &mdev->data);
+
+	mdev->dp_volume_last = mdev->send_cnt;
+	mod_timer(&mdev->delay_probe_timer, jiffies + mdev->sync_conf.dp_interval * HZ / 10);
+
+	return ok;
+}
+
 /* called on sndtimeo
  * returns FALSE if we should retry,
  * TRUE if we think connection is dead
@@ -2229,9 +2272,9 @@ static int we_should_drop_the_connection(struct drbd_conf *mdev, struct socket *
  * with page_count == 0 or PageSlab.
  */
 static int _drbd_no_send_page(struct drbd_conf *mdev, struct page *page,
-		   int offset, size_t size)
+		   int offset, size_t size, unsigned msg_flags)
 {
-	int sent = drbd_send(mdev, mdev->data.socket, kmap(page) + offset, size, 0);
+	int sent = drbd_send(mdev, mdev->data.socket, kmap(page) + offset, size, msg_flags);
 	kunmap(page);
 	if (sent == size)
 		mdev->send_cnt += size>>9;
@@ -2239,7 +2282,7 @@ static int _drbd_no_send_page(struct drbd_conf *mdev, struct page *page,
 }
 
 static int _drbd_send_page(struct drbd_conf *mdev, struct page *page,
-		    int offset, size_t size)
+		    int offset, size_t size, unsigned msg_flags)
 {
 	mm_segment_t oldfs = get_fs();
 	int sent, ok;
@@ -2252,14 +2295,15 @@ static int _drbd_send_page(struct drbd_conf *mdev, struct page *page,
 	 * __page_cache_release a page that would actually still be referenced
 	 * by someone, leading to some obscure delayed Oops somewhere else. */
 	if (disable_sendpage || (page_count(page) < 1) || PageSlab(page))
-		return _drbd_no_send_page(mdev, page, offset, size);
+		return _drbd_no_send_page(mdev, page, offset, size, msg_flags);
 
+	msg_flags |= MSG_NOSIGNAL;
 	drbd_update_congested(mdev);
 	set_fs(KERNEL_DS);
 	do {
 		sent = mdev->data.socket->ops->sendpage(mdev->data.socket, page,
 							offset, len,
-							MSG_NOSIGNAL);
+							msg_flags);
 		if (sent == -EAGAIN) {
 			if (we_should_drop_the_connection(mdev,
 							  mdev->data.socket))
@@ -2288,9 +2332,11 @@ static int _drbd_send_bio(struct drbd_conf *mdev, struct bio *bio)
 {
 	struct bio_vec *bvec;
 	int i;
+	/* hint all but last page with MSG_MORE */
 	__bio_for_each_segment(bvec, bio, i, 0) {
 		if (!_drbd_no_send_page(mdev, bvec->bv_page,
-				     bvec->bv_offset, bvec->bv_len))
+				     bvec->bv_offset, bvec->bv_len,
+				     i == bio->bi_vcnt -1 ? 0 : MSG_MORE))
 			return 0;
 	}
 	return 1;
@@ -2300,15 +2346,56 @@ static int _drbd_send_zc_bio(struct drbd_conf *mdev, struct bio *bio)
 {
 	struct bio_vec *bvec;
 	int i;
+	/* hint all but last page with MSG_MORE */
 	__bio_for_each_segment(bvec, bio, i, 0) {
 		if (!_drbd_send_page(mdev, bvec->bv_page,
-				     bvec->bv_offset, bvec->bv_len))
+				     bvec->bv_offset, bvec->bv_len,
+				     i == bio->bi_vcnt -1 ? 0 : MSG_MORE))
+			return 0;
+	}
+	return 1;
+}
+
+static int _drbd_send_zc_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e)
+{
+	struct page *page = e->pages;
+	unsigned len = e->size;
+	/* hint all but last page with MSG_MORE */
+	page_chain_for_each(page) {
+		unsigned l = min_t(unsigned, len, PAGE_SIZE);
+		if (!_drbd_send_page(mdev, page, 0, l,
+				page_chain_next(page) ? MSG_MORE : 0))
 			return 0;
+		len -= l;
 	}
+	return 1;
+}
+
+static void consider_delay_probes(struct drbd_conf *mdev)
+{
+	if (mdev->state.conn != C_SYNC_SOURCE || mdev->agreed_pro_version < 93)
+		return;
+
+	if (mdev->dp_volume_last + mdev->sync_conf.dp_volume * 2 < mdev->send_cnt)
+		drbd_send_delay_probes(mdev);
+}
+
+static int w_delay_probes(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
+{
+	if (!cancel && mdev->state.conn == C_SYNC_SOURCE)
+		drbd_send_delay_probes(mdev);
 
 	return 1;
 }
 
+static void delay_probe_timer_fn(unsigned long data)
+{
+	struct drbd_conf *mdev = (struct drbd_conf *) data;
+
+	if (list_empty(&mdev->delay_probe_work.list))
+		drbd_queue_work(&mdev->data.work, &mdev->delay_probe_work);
+}
+
 /* Used to send write requests
  * R_PRIMARY -> Peer	(P_DATA)
  */
@@ -2357,11 +2444,11 @@ int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req)
 	p.dp_flags = cpu_to_be32(dp_flags);
 	set_bit(UNPLUG_REMOTE, &mdev->flags);
 	ok = (sizeof(p) ==
-		drbd_send(mdev, mdev->data.socket, &p, sizeof(p), MSG_MORE));
+		drbd_send(mdev, mdev->data.socket, &p, sizeof(p), dgs ? MSG_MORE : 0));
 	if (ok && dgs) {
 		dgb = mdev->int_dig_out;
-		drbd_csum(mdev, mdev->integrity_w_tfm, req->master_bio, dgb);
-		ok = drbd_send(mdev, mdev->data.socket, dgb, dgs, MSG_MORE);
+		drbd_csum_bio(mdev, mdev->integrity_w_tfm, req->master_bio, dgb);
+		ok = drbd_send(mdev, mdev->data.socket, dgb, dgs, 0);
 	}
 	if (ok) {
 		if (mdev->net_conf->wire_protocol == DRBD_PROT_A)
@@ -2371,6 +2458,10 @@ int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req)
 	}
 
 	drbd_put_data_sock(mdev);
+
+	if (ok)
+		consider_delay_probes(mdev);
+
 	return ok;
 }
 
@@ -2406,16 +2497,20 @@ int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd,
 		return 0;
 
 	ok = sizeof(p) == drbd_send(mdev, mdev->data.socket, &p,
-					sizeof(p), MSG_MORE);
+					sizeof(p), dgs ? MSG_MORE : 0);
 	if (ok && dgs) {
 		dgb = mdev->int_dig_out;
-		drbd_csum(mdev, mdev->integrity_w_tfm, e->private_bio, dgb);
-		ok = drbd_send(mdev, mdev->data.socket, dgb, dgs, MSG_MORE);
+		drbd_csum_ee(mdev, mdev->integrity_w_tfm, e, dgb);
+		ok = drbd_send(mdev, mdev->data.socket, dgb, dgs, 0);
 	}
 	if (ok)
-		ok = _drbd_send_zc_bio(mdev, e->private_bio);
+		ok = _drbd_send_zc_ee(mdev, e);
 
 	drbd_put_data_sock(mdev);
+
+	if (ok)
+		consider_delay_probes(mdev);
+
 	return ok;
 }
 
@@ -2628,16 +2723,24 @@ void drbd_init_set_defaults(struct drbd_conf *mdev)
 	INIT_LIST_HEAD(&mdev->unplug_work.list);
 	INIT_LIST_HEAD(&mdev->md_sync_work.list);
 	INIT_LIST_HEAD(&mdev->bm_io_work.w.list);
+	INIT_LIST_HEAD(&mdev->delay_probes);
+	INIT_LIST_HEAD(&mdev->delay_probe_work.list);
+
 	mdev->resync_work.cb  = w_resync_inactive;
 	mdev->unplug_work.cb  = w_send_write_hint;
 	mdev->md_sync_work.cb = w_md_sync;
 	mdev->bm_io_work.w.cb = w_bitmap_io;
+	mdev->delay_probe_work.cb = w_delay_probes;
 	init_timer(&mdev->resync_timer);
 	init_timer(&mdev->md_sync_timer);
+	init_timer(&mdev->delay_probe_timer);
 	mdev->resync_timer.function = resync_timer_fn;
 	mdev->resync_timer.data = (unsigned long) mdev;
 	mdev->md_sync_timer.function = md_sync_timer_fn;
 	mdev->md_sync_timer.data = (unsigned long) mdev;
+	mdev->delay_probe_timer.function = delay_probe_timer_fn;
+	mdev->delay_probe_timer.data = (unsigned long) mdev;
+
 
 	init_waitqueue_head(&mdev->misc_wait);
 	init_waitqueue_head(&mdev->state_wait);
@@ -2680,7 +2783,7 @@ void drbd_mdev_cleanup(struct drbd_conf *mdev)
 	drbd_set_my_capacity(mdev, 0);
 	if (mdev->bitmap) {
 		/* maybe never allocated. */
-		drbd_bm_resize(mdev, 0);
+		drbd_bm_resize(mdev, 0, 1);
 		drbd_bm_cleanup(mdev);
 	}
 
@@ -3129,7 +3232,7 @@ int __init drbd_init(void)
 	if (err)
 		goto Enomem;
 
-	drbd_proc = proc_create("drbd", S_IFREG | S_IRUGO , NULL, &drbd_proc_fops);
+	drbd_proc = proc_create_data("drbd", S_IFREG | S_IRUGO , NULL, &drbd_proc_fops, NULL);
 	if (!drbd_proc)	{
 		printk(KERN_ERR "drbd: unable to register proc file\n");
 		goto Enomem;
@@ -3660,7 +3763,8 @@ _drbd_fault_str(unsigned int type) {
 		[DRBD_FAULT_DT_RD] = "Data read",
 		[DRBD_FAULT_DT_RA] = "Data read ahead",
 		[DRBD_FAULT_BM_ALLOC] = "BM allocation",
-		[DRBD_FAULT_AL_EE] = "EE allocation"
+		[DRBD_FAULT_AL_EE] = "EE allocation",
+		[DRBD_FAULT_RECEIVE] = "receive data corruption",
 	};
 
 	return (type < DRBD_FAULT_MAX) ? _faults[type] : "**Unknown**";
@@ -3679,7 +3783,7 @@ _drbd_insert_fault(struct drbd_conf *mdev, unsigned int type)
 	if (ret) {
 		fault_count++;
 
-		if (printk_ratelimit())
+		if (__ratelimit(&drbd_ratelimit_state))
 			dev_warn(DEV, "***Simulating %s failure\n",
 				_drbd_fault_str(type));
 	}
diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c
index 6429d2b19e06..632e3245d1bb 100644
--- a/drivers/block/drbd/drbd_nl.c
+++ b/drivers/block/drbd/drbd_nl.c
@@ -510,7 +510,7 @@ void drbd_resume_io(struct drbd_conf *mdev)
  * Returns 0 on success, negative return values indicate errors.
  * You should call drbd_md_sync() after calling this function.
  */
-enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *mdev, int force) __must_hold(local)
+enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *mdev, enum dds_flags flags) __must_hold(local)
 {
 	sector_t prev_first_sect, prev_size; /* previous meta location */
 	sector_t la_size;
@@ -541,12 +541,12 @@ enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *mdev, int force
 	/* TODO: should only be some assert here, not (re)init... */
 	drbd_md_set_sector_offsets(mdev, mdev->ldev);
 
-	size = drbd_new_dev_size(mdev, mdev->ldev, force);
+	size = drbd_new_dev_size(mdev, mdev->ldev, flags & DDSF_FORCED);
 
 	if (drbd_get_capacity(mdev->this_bdev) != size ||
 	    drbd_bm_capacity(mdev) != size) {
 		int err;
-		err = drbd_bm_resize(mdev, size);
+		err = drbd_bm_resize(mdev, size, !(flags & DDSF_NO_RESYNC));
 		if (unlikely(err)) {
 			/* currently there is only one error: ENOMEM! */
 			size = drbd_bm_capacity(mdev)>>1;
@@ -704,9 +704,6 @@ void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int max_seg_s) __mu
 	struct request_queue * const b = mdev->ldev->backing_bdev->bd_disk->queue;
 	int max_segments = mdev->ldev->dc.max_bio_bvecs;
 
-	if (b->merge_bvec_fn && !mdev->ldev->dc.use_bmbv)
-		max_seg_s = PAGE_SIZE;
-
 	max_seg_s = min(queue_max_sectors(b) * queue_logical_block_size(b), max_seg_s);
 
 	blk_queue_max_hw_sectors(q, max_seg_s >> 9);
@@ -1199,13 +1196,12 @@ static int drbd_nl_net_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
 	}
 
 	/* allocation not in the IO path, cqueue thread context */
-	new_conf = kmalloc(sizeof(struct net_conf), GFP_KERNEL);
+	new_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL);
 	if (!new_conf) {
 		retcode = ERR_NOMEM;
 		goto fail;
 	}
 
-	memset(new_conf, 0, sizeof(struct net_conf));
 	new_conf->timeout	   = DRBD_TIMEOUT_DEF;
 	new_conf->try_connect_int  = DRBD_CONNECT_INT_DEF;
 	new_conf->ping_int	   = DRBD_PING_INT_DEF;
@@ -1477,8 +1473,8 @@ static int drbd_nl_resize(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
 {
 	struct resize rs;
 	int retcode = NO_ERROR;
-	int ldsc = 0; /* local disk size changed */
 	enum determine_dev_size dd;
+	enum dds_flags ddsf;
 
 	memset(&rs, 0, sizeof(struct resize));
 	if (!resize_from_tags(mdev, nlp->tag_list, &rs)) {
@@ -1502,13 +1498,17 @@ static int drbd_nl_resize(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
 		goto fail;
 	}
 
-	if (mdev->ldev->known_size != drbd_get_capacity(mdev->ldev->backing_bdev)) {
-		mdev->ldev->known_size = drbd_get_capacity(mdev->ldev->backing_bdev);
-		ldsc = 1;
+	if (rs.no_resync && mdev->agreed_pro_version < 93) {
+		retcode = ERR_NEED_APV_93;
+		goto fail;
 	}
 
+	if (mdev->ldev->known_size != drbd_get_capacity(mdev->ldev->backing_bdev))
+		mdev->ldev->known_size = drbd_get_capacity(mdev->ldev->backing_bdev);
+
 	mdev->ldev->dc.disk_size = (sector_t)rs.resize_size;
-	dd = drbd_determin_dev_size(mdev, rs.resize_force);
+	ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0);
+	dd = drbd_determin_dev_size(mdev, ddsf);
 	drbd_md_sync(mdev);
 	put_ldev(mdev);
 	if (dd == dev_size_error) {
@@ -1516,12 +1516,12 @@ static int drbd_nl_resize(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
 		goto fail;
 	}
 
-	if (mdev->state.conn == C_CONNECTED && (dd != unchanged || ldsc)) {
+	if (mdev->state.conn == C_CONNECTED) {
 		if (dd == grew)
 			set_bit(RESIZE_PENDING, &mdev->flags);
 
 		drbd_send_uuids(mdev);
-		drbd_send_sizes(mdev, 1);
+		drbd_send_sizes(mdev, 1, ddsf);
 	}
 
  fail:
@@ -1551,6 +1551,10 @@ static int drbd_nl_syncer_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *n
 		sc.rate       = DRBD_RATE_DEF;
 		sc.after      = DRBD_AFTER_DEF;
 		sc.al_extents = DRBD_AL_EXTENTS_DEF;
+		sc.dp_volume  = DRBD_DP_VOLUME_DEF;
+		sc.dp_interval = DRBD_DP_INTERVAL_DEF;
+		sc.throttle_th = DRBD_RS_THROTTLE_TH_DEF;
+		sc.hold_off_th = DRBD_RS_HOLD_OFF_TH_DEF;
 	} else
 		memcpy(&sc, &mdev->sync_conf, sizeof(struct syncer_conf));
 
@@ -2207,9 +2211,9 @@ void drbd_bcast_ee(struct drbd_conf *mdev,
 {
 	struct cn_msg *cn_reply;
 	struct drbd_nl_cfg_reply *reply;
-	struct bio_vec *bvec;
 	unsigned short *tl;
-	int i;
+	struct page *page;
+	unsigned len;
 
 	if (!e)
 		return;
@@ -2247,11 +2251,15 @@ void drbd_bcast_ee(struct drbd_conf *mdev,
 	put_unaligned(T_ee_data, tl++);
 	put_unaligned(e->size, tl++);
 
-	__bio_for_each_segment(bvec, e->private_bio, i, 0) {
-		void *d = kmap(bvec->bv_page);
-		memcpy(tl, d + bvec->bv_offset, bvec->bv_len);
-		kunmap(bvec->bv_page);
-		tl=(unsigned short*)((char*)tl + bvec->bv_len);
+	len = e->size;
+	page = e->pages;
+	page_chain_for_each(page) {
+		void *d = kmap_atomic(page, KM_USER0);
+		unsigned l = min_t(unsigned, len, PAGE_SIZE);
+		memcpy(tl, d, l);
+		kunmap_atomic(d, KM_USER0);
+		tl = (unsigned short*)((char*)tl + l);
+		len -= l;
 	}
 	put_unaligned(TT_END, tl++); /* Close the tag list */
 
diff --git a/drivers/block/drbd/drbd_proc.c b/drivers/block/drbd/drbd_proc.c
index be3374b68460..d0f1767ea4c3 100644
--- a/drivers/block/drbd/drbd_proc.c
+++ b/drivers/block/drbd/drbd_proc.c
@@ -73,14 +73,21 @@ static void drbd_syncer_progress(struct drbd_conf *mdev, struct seq_file *seq)
 	seq_printf(seq, "sync'ed:%3u.%u%% ", res / 10, res % 10);
 	/* if more than 1 GB display in MB */
 	if (mdev->rs_total > 0x100000L)
-		seq_printf(seq, "(%lu/%lu)M\n\t",
+		seq_printf(seq, "(%lu/%lu)M",
 			    (unsigned long) Bit2KB(rs_left >> 10),
 			    (unsigned long) Bit2KB(mdev->rs_total >> 10));
 	else
-		seq_printf(seq, "(%lu/%lu)K\n\t",
+		seq_printf(seq, "(%lu/%lu)K",
 			    (unsigned long) Bit2KB(rs_left),
 			    (unsigned long) Bit2KB(mdev->rs_total));
 
+	if (mdev->state.conn == C_SYNC_TARGET)
+		seq_printf(seq, " queue_delay: %d.%d ms\n\t",
+			   mdev->data_delay / 1000,
+			   (mdev->data_delay % 1000) / 100);
+	else if (mdev->state.conn == C_SYNC_SOURCE)
+		seq_printf(seq, " delay_probe: %u\n\t", mdev->delay_seq);
+
 	/* see drivers/md/md.c
 	 * We do not want to overflow, so the order of operands and
 	 * the * 100 / 100 trick are important. We do a +1 to be
@@ -128,6 +135,14 @@ static void drbd_syncer_progress(struct drbd_conf *mdev, struct seq_file *seq)
 	else
 		seq_printf(seq, " (%ld)", dbdt);
 
+	if (mdev->state.conn == C_SYNC_TARGET) {
+		if (mdev->c_sync_rate > 1000)
+			seq_printf(seq, " want: %d,%03d",
+				   mdev->c_sync_rate / 1000, mdev->c_sync_rate % 1000);
+		else
+			seq_printf(seq, " want: %d", mdev->c_sync_rate);
+	}
+
 	seq_printf(seq, " K/sec\n");
 }
 
diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index 3f096e7959b4..dff48701b84d 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -42,7 +42,6 @@
 #include <linux/unistd.h>
 #include <linux/vmalloc.h>
 #include <linux/random.h>
-#include <linux/mm.h>
 #include <linux/string.h>
 #include <linux/scatterlist.h>
 #include "drbd_int.h"
@@ -80,30 +79,128 @@ static struct drbd_epoch *previous_epoch(struct drbd_conf *mdev, struct drbd_epo
 
 #define GFP_TRY	(__GFP_HIGHMEM | __GFP_NOWARN)
 
-static struct page *drbd_pp_first_page_or_try_alloc(struct drbd_conf *mdev)
+/*
+ * some helper functions to deal with single linked page lists,
+ * page->private being our "next" pointer.
+ */
+
+/* If at least n pages are linked at head, get n pages off.
+ * Otherwise, don't modify head, and return NULL.
+ * Locking is the responsibility of the caller.
+ */
+static struct page *page_chain_del(struct page **head, int n)
+{
+	struct page *page;
+	struct page *tmp;
+
+	BUG_ON(!n);
+	BUG_ON(!head);
+
+	page = *head;
+
+	if (!page)
+		return NULL;
+
+	while (page) {
+		tmp = page_chain_next(page);
+		if (--n == 0)
+			break; /* found sufficient pages */
+		if (tmp == NULL)
+			/* insufficient pages, don't use any of them. */
+			return NULL;
+		page = tmp;
+	}
+
+	/* add end of list marker for the returned list */
+	set_page_private(page, 0);
+	/* actual return value, and adjustment of head */
+	page = *head;
+	*head = tmp;
+	return page;
+}
+
+/* may be used outside of locks to find the tail of a (usually short)
+ * "private" page chain, before adding it back to a global chain head
+ * with page_chain_add() under a spinlock. */
+static struct page *page_chain_tail(struct page *page, int *len)
+{
+	struct page *tmp;
+	int i = 1;
+	while ((tmp = page_chain_next(page)))
+		++i, page = tmp;
+	if (len)
+		*len = i;
+	return page;
+}
+
+static int page_chain_free(struct page *page)
+{
+	struct page *tmp;
+	int i = 0;
+	page_chain_for_each_safe(page, tmp) {
+		put_page(page);
+		++i;
+	}
+	return i;
+}
+
+static void page_chain_add(struct page **head,
+		struct page *chain_first, struct page *chain_last)
+{
+#if 1
+	struct page *tmp;
+	tmp = page_chain_tail(chain_first, NULL);
+	BUG_ON(tmp != chain_last);
+#endif
+
+	/* add chain to head */
+	set_page_private(chain_last, (unsigned long)*head);
+	*head = chain_first;
+}
+
+static struct page *drbd_pp_first_pages_or_try_alloc(struct drbd_conf *mdev, int number)
 {
 	struct page *page = NULL;
+	struct page *tmp = NULL;
+	int i = 0;
 
 	/* Yes, testing drbd_pp_vacant outside the lock is racy.
 	 * So what. It saves a spin_lock. */
-	if (drbd_pp_vacant > 0) {
+	if (drbd_pp_vacant >= number) {
 		spin_lock(&drbd_pp_lock);
-		page = drbd_pp_pool;
-		if (page) {
-			drbd_pp_pool = (struct page *)page_private(page);
-			set_page_private(page, 0); /* just to be polite */
-			drbd_pp_vacant--;
-		}
+		page = page_chain_del(&drbd_pp_pool, number);
+		if (page)
+			drbd_pp_vacant -= number;
 		spin_unlock(&drbd_pp_lock);
+		if (page)
+			return page;
 	}
+
 	/* GFP_TRY, because we must not cause arbitrary write-out: in a DRBD
 	 * "criss-cross" setup, that might cause write-out on some other DRBD,
 	 * which in turn might block on the other node at this very place.  */
-	if (!page)
-		page = alloc_page(GFP_TRY);
-	if (page)
-		atomic_inc(&mdev->pp_in_use);
-	return page;
+	for (i = 0; i < number; i++) {
+		tmp = alloc_page(GFP_TRY);
+		if (!tmp)
+			break;
+		set_page_private(tmp, (unsigned long)page);
+		page = tmp;
+	}
+
+	if (i == number)
+		return page;
+
+	/* Not enough pages immediately available this time.
+	 * No need to jump around here, drbd_pp_alloc will retry this
+	 * function "soon". */
+	if (page) {
+		tmp = page_chain_tail(page, NULL);
+		spin_lock(&drbd_pp_lock);
+		page_chain_add(&drbd_pp_pool, page, tmp);
+		drbd_pp_vacant += i;
+		spin_unlock(&drbd_pp_lock);
+	}
+	return NULL;
 }
 
 /* kick lower level device, if we have more than (arbitrary number)
@@ -127,7 +224,7 @@ static void reclaim_net_ee(struct drbd_conf *mdev, struct list_head *to_be_freed
 
 	list_for_each_safe(le, tle, &mdev->net_ee) {
 		e = list_entry(le, struct drbd_epoch_entry, w.list);
-		if (drbd_bio_has_active_page(e->private_bio))
+		if (drbd_ee_has_active_page(e))
 			break;
 		list_move(le, to_be_freed);
 	}
@@ -148,32 +245,34 @@ static void drbd_kick_lo_and_reclaim_net(struct drbd_conf *mdev)
 }
 
 /**
- * drbd_pp_alloc() - Returns a page, fails only if a signal comes in
+ * drbd_pp_alloc() - Returns @number pages, retries forever (or until signalled)
  * @mdev:	DRBD device.
- * @retry:	whether or not to retry allocation forever (or until signalled)
+ * @number:	number of pages requested
+ * @retry:	whether to retry, if not enough pages are available right now
+ *
+ * Tries to allocate number pages, first from our own page pool, then from
+ * the kernel, unless this allocation would exceed the max_buffers setting.
+ * Possibly retry until DRBD frees sufficient pages somewhere else.
  *
- * Tries to allocate a page, first from our own page pool, then from the
- * kernel, unless this allocation would exceed the max_buffers setting.
- * If @retry is non-zero, retry until DRBD frees a page somewhere else.
+ * Returns a page chain linked via page->private.
  */
-static struct page *drbd_pp_alloc(struct drbd_conf *mdev, int retry)
+static struct page *drbd_pp_alloc(struct drbd_conf *mdev, unsigned number, bool retry)
 {
 	struct page *page = NULL;
 	DEFINE_WAIT(wait);
 
-	if (atomic_read(&mdev->pp_in_use) < mdev->net_conf->max_buffers) {
-		page = drbd_pp_first_page_or_try_alloc(mdev);
-		if (page)
-			return page;
-	}
+	/* Yes, we may run up to @number over max_buffers. If we
+	 * follow it strictly, the admin will get it wrong anyways. */
+	if (atomic_read(&mdev->pp_in_use) < mdev->net_conf->max_buffers)
+		page = drbd_pp_first_pages_or_try_alloc(mdev, number);
 
-	for (;;) {
+	while (page == NULL) {
 		prepare_to_wait(&drbd_pp_wait, &wait, TASK_INTERRUPTIBLE);
 
 		drbd_kick_lo_and_reclaim_net(mdev);
 
 		if (atomic_read(&mdev->pp_in_use) < mdev->net_conf->max_buffers) {
-			page = drbd_pp_first_page_or_try_alloc(mdev);
+			page = drbd_pp_first_pages_or_try_alloc(mdev, number);
 			if (page)
 				break;
 		}
@@ -190,62 +289,32 @@ static struct page *drbd_pp_alloc(struct drbd_conf *mdev, int retry)
 	}
 	finish_wait(&drbd_pp_wait, &wait);
 
+	if (page)
+		atomic_add(number, &mdev->pp_in_use);
 	return page;
 }
 
 /* Must not be used from irq, as that may deadlock: see drbd_pp_alloc.
- * Is also used from inside an other spin_lock_irq(&mdev->req_lock) */
+ * Is also used from inside an other spin_lock_irq(&mdev->req_lock);
+ * Either links the page chain back to the global pool,
+ * or returns all pages to the system. */
 static void drbd_pp_free(struct drbd_conf *mdev, struct page *page)
 {
-	int free_it;
-
-	spin_lock(&drbd_pp_lock);
-	if (drbd_pp_vacant > (DRBD_MAX_SEGMENT_SIZE/PAGE_SIZE)*minor_count) {
-		free_it = 1;
-	} else {
-		set_page_private(page, (unsigned long)drbd_pp_pool);
-		drbd_pp_pool = page;
-		drbd_pp_vacant++;
-		free_it = 0;
-	}
-	spin_unlock(&drbd_pp_lock);
-
-	atomic_dec(&mdev->pp_in_use);
-
-	if (free_it)
-		__free_page(page);
-
-	wake_up(&drbd_pp_wait);
-}
-
-static void drbd_pp_free_bio_pages(struct drbd_conf *mdev, struct bio *bio)
-{
-	struct page *p_to_be_freed = NULL;
-	struct page *page;
-	struct bio_vec *bvec;
 	int i;
-
-	spin_lock(&drbd_pp_lock);
-	__bio_for_each_segment(bvec, bio, i, 0) {
-		if (drbd_pp_vacant > (DRBD_MAX_SEGMENT_SIZE/PAGE_SIZE)*minor_count) {
-			set_page_private(bvec->bv_page, (unsigned long)p_to_be_freed);
-			p_to_be_freed = bvec->bv_page;
-		} else {
-			set_page_private(bvec->bv_page, (unsigned long)drbd_pp_pool);
-			drbd_pp_pool = bvec->bv_page;
-			drbd_pp_vacant++;
-		}
-	}
-	spin_unlock(&drbd_pp_lock);
-	atomic_sub(bio->bi_vcnt, &mdev->pp_in_use);
-
-	while (p_to_be_freed) {
-		page = p_to_be_freed;
-		p_to_be_freed = (struct page *)page_private(page);
-		set_page_private(page, 0); /* just to be polite */
-		put_page(page);
+	if (drbd_pp_vacant > (DRBD_MAX_SEGMENT_SIZE/PAGE_SIZE)*minor_count)
+		i = page_chain_free(page);
+	else {
+		struct page *tmp;
+		tmp = page_chain_tail(page, &i);
+		spin_lock(&drbd_pp_lock);
+		page_chain_add(&drbd_pp_pool, page, tmp);
+		drbd_pp_vacant += i;
+		spin_unlock(&drbd_pp_lock);
 	}
-
+	atomic_sub(i, &mdev->pp_in_use);
+	i = atomic_read(&mdev->pp_in_use);
+	if (i < 0)
+		dev_warn(DEV, "ASSERTION FAILED: pp_in_use: %d < 0\n", i);
 	wake_up(&drbd_pp_wait);
 }
 
@@ -270,11 +339,9 @@ struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev,
 				     unsigned int data_size,
 				     gfp_t gfp_mask) __must_hold(local)
 {
-	struct request_queue *q;
 	struct drbd_epoch_entry *e;
 	struct page *page;
-	struct bio *bio;
-	unsigned int ds;
+	unsigned nr_pages = (data_size + PAGE_SIZE -1) >> PAGE_SHIFT;
 
 	if (FAULT_ACTIVE(mdev, DRBD_FAULT_AL_EE))
 		return NULL;
@@ -286,84 +353,32 @@ struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev,
 		return NULL;
 	}
 
-	bio = bio_alloc(gfp_mask & ~__GFP_HIGHMEM, div_ceil(data_size, PAGE_SIZE));
-	if (!bio) {
-		if (!(gfp_mask & __GFP_NOWARN))
-			dev_err(DEV, "alloc_ee: Allocation of a bio failed\n");
-		goto fail1;
-	}
-
-	bio->bi_bdev = mdev->ldev->backing_bdev;
-	bio->bi_sector = sector;
-
-	ds = data_size;
-	while (ds) {
-		page = drbd_pp_alloc(mdev, (gfp_mask & __GFP_WAIT));
-		if (!page) {
-			if (!(gfp_mask & __GFP_NOWARN))
-				dev_err(DEV, "alloc_ee: Allocation of a page failed\n");
-			goto fail2;
-		}
-		if (!bio_add_page(bio, page, min_t(int, ds, PAGE_SIZE), 0)) {
-			drbd_pp_free(mdev, page);
-			dev_err(DEV, "alloc_ee: bio_add_page(s=%llu,"
-			    "data_size=%u,ds=%u) failed\n",
-			    (unsigned long long)sector, data_size, ds);
-
-			q = bdev_get_queue(bio->bi_bdev);
-			if (q->merge_bvec_fn) {
-				struct bvec_merge_data bvm = {
-					.bi_bdev = bio->bi_bdev,
-					.bi_sector = bio->bi_sector,
-					.bi_size = bio->bi_size,
-					.bi_rw = bio->bi_rw,
-				};
-				int l = q->merge_bvec_fn(q, &bvm,
-						&bio->bi_io_vec[bio->bi_vcnt]);
-				dev_err(DEV, "merge_bvec_fn() = %d\n", l);
-			}
-
-			/* dump more of the bio. */
-			dev_err(DEV, "bio->bi_max_vecs = %d\n", bio->bi_max_vecs);
-			dev_err(DEV, "bio->bi_vcnt = %d\n", bio->bi_vcnt);
-			dev_err(DEV, "bio->bi_size = %d\n", bio->bi_size);
-			dev_err(DEV, "bio->bi_phys_segments = %d\n", bio->bi_phys_segments);
-
-			goto fail2;
-			break;
-		}
-		ds -= min_t(int, ds, PAGE_SIZE);
-	}
-
-	D_ASSERT(data_size == bio->bi_size);
-
-	bio->bi_private = e;
-	e->mdev = mdev;
-	e->sector = sector;
-	e->size = bio->bi_size;
+	page = drbd_pp_alloc(mdev, nr_pages, (gfp_mask & __GFP_WAIT));
+	if (!page)
+		goto fail;
 
-	e->private_bio = bio;
-	e->block_id = id;
 	INIT_HLIST_NODE(&e->colision);
 	e->epoch = NULL;
+	e->mdev = mdev;
+	e->pages = page;
+	atomic_set(&e->pending_bios, 0);
+	e->size = data_size;
 	e->flags = 0;
+	e->sector = sector;
+	e->sector = sector;
+	e->block_id = id;
 
 	return e;
 
- fail2:
-	drbd_pp_free_bio_pages(mdev, bio);
-	bio_put(bio);
- fail1:
+ fail:
 	mempool_free(e, drbd_ee_mempool);
-
 	return NULL;
 }
 
 void drbd_free_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e)
 {
-	struct bio *bio = e->private_bio;
-	drbd_pp_free_bio_pages(mdev, bio);
-	bio_put(bio);
+	drbd_pp_free(mdev, e->pages);
+	D_ASSERT(atomic_read(&e->pending_bios) == 0);
 	D_ASSERT(hlist_unhashed(&e->colision));
 	mempool_free(e, drbd_ee_mempool);
 }
@@ -555,6 +570,25 @@ static int drbd_recv(struct drbd_conf *mdev, void *buf, size_t size)
 	return rv;
 }
 
+/* quoting tcp(7):
+ *   On individual connections, the socket buffer size must be set prior to the
+ *   listen(2) or connect(2) calls in order to have it take effect.
+ * This is our wrapper to do so.
+ */
+static void drbd_setbufsize(struct socket *sock, unsigned int snd,
+		unsigned int rcv)
+{
+	/* open coded SO_SNDBUF, SO_RCVBUF */
+	if (snd) {
+		sock->sk->sk_sndbuf = snd;
+		sock->sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
+	}
+	if (rcv) {
+		sock->sk->sk_rcvbuf = rcv;
+		sock->sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
+	}
+}
+
 static struct socket *drbd_try_connect(struct drbd_conf *mdev)
 {
 	const char *what;
@@ -576,6 +610,8 @@ static struct socket *drbd_try_connect(struct drbd_conf *mdev)
 
 	sock->sk->sk_rcvtimeo =
 	sock->sk->sk_sndtimeo =  mdev->net_conf->try_connect_int*HZ;
+	drbd_setbufsize(sock, mdev->net_conf->sndbuf_size,
+			mdev->net_conf->rcvbuf_size);
 
        /* explicitly bind to the configured IP as source IP
 	*  for the outgoing connections.
@@ -654,6 +690,8 @@ static struct socket *drbd_wait_for_connect(struct drbd_conf *mdev)
 	s_listen->sk->sk_reuse    = 1; /* SO_REUSEADDR */
 	s_listen->sk->sk_rcvtimeo = timeo;
 	s_listen->sk->sk_sndtimeo = timeo;
+	drbd_setbufsize(s_listen, mdev->net_conf->sndbuf_size,
+			mdev->net_conf->rcvbuf_size);
 
 	what = "bind before listen";
 	err = s_listen->ops->bind(s_listen,
@@ -840,16 +878,6 @@ retry:
 	sock->sk->sk_priority = TC_PRIO_INTERACTIVE_BULK;
 	msock->sk->sk_priority = TC_PRIO_INTERACTIVE;
 
-	if (mdev->net_conf->sndbuf_size) {
-		sock->sk->sk_sndbuf = mdev->net_conf->sndbuf_size;
-		sock->sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
-	}
-
-	if (mdev->net_conf->rcvbuf_size) {
-		sock->sk->sk_rcvbuf = mdev->net_conf->rcvbuf_size;
-		sock->sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
-	}
-
 	/* NOT YET ...
 	 * sock->sk->sk_sndtimeo = mdev->net_conf->timeout*HZ/10;
 	 * sock->sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
@@ -902,7 +930,7 @@ retry:
 	if (!drbd_send_protocol(mdev))
 		return -1;
 	drbd_send_sync_param(mdev, &mdev->sync_conf);
-	drbd_send_sizes(mdev, 0);
+	drbd_send_sizes(mdev, 0, 0);
 	drbd_send_uuids(mdev);
 	drbd_send_state(mdev);
 	clear_bit(USE_DEGR_WFC_T, &mdev->flags);
@@ -946,7 +974,8 @@ static enum finish_epoch drbd_flush_after_epoch(struct drbd_conf *mdev, struct d
 	int rv;
 
 	if (mdev->write_ordering >= WO_bdev_flush && get_ldev(mdev)) {
-		rv = blkdev_issue_flush(mdev->ldev->backing_bdev, NULL);
+		rv = blkdev_issue_flush(mdev->ldev->backing_bdev, GFP_KERNEL,
+					NULL, BLKDEV_IFL_WAIT);
 		if (rv) {
 			dev_err(DEV, "local disk flush failed with status %d\n", rv);
 			/* would rather check on EOPNOTSUPP, but that is not reliable.
@@ -1120,6 +1149,90 @@ void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo)
 }
 
 /**
+ * drbd_submit_ee()
+ * @mdev:	DRBD device.
+ * @e:		epoch entry
+ * @rw:		flag field, see bio->bi_rw
+ */
+/* TODO allocate from our own bio_set. */
+int drbd_submit_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e,
+		const unsigned rw, const int fault_type)
+{
+	struct bio *bios = NULL;
+	struct bio *bio;
+	struct page *page = e->pages;
+	sector_t sector = e->sector;
+	unsigned ds = e->size;
+	unsigned n_bios = 0;
+	unsigned nr_pages = (ds + PAGE_SIZE -1) >> PAGE_SHIFT;
+
+	/* In most cases, we will only need one bio.  But in case the lower
+	 * level restrictions happen to be different at this offset on this
+	 * side than those of the sending peer, we may need to submit the
+	 * request in more than one bio. */
+next_bio:
+	bio = bio_alloc(GFP_NOIO, nr_pages);
+	if (!bio) {
+		dev_err(DEV, "submit_ee: Allocation of a bio failed\n");
+		goto fail;
+	}
+	/* > e->sector, unless this is the first bio */
+	bio->bi_sector = sector;
+	bio->bi_bdev = mdev->ldev->backing_bdev;
+	/* we special case some flags in the multi-bio case, see below
+	 * (BIO_RW_UNPLUG, BIO_RW_BARRIER) */
+	bio->bi_rw = rw;
+	bio->bi_private = e;
+	bio->bi_end_io = drbd_endio_sec;
+
+	bio->bi_next = bios;
+	bios = bio;
+	++n_bios;
+
+	page_chain_for_each(page) {
+		unsigned len = min_t(unsigned, ds, PAGE_SIZE);
+		if (!bio_add_page(bio, page, len, 0)) {
+			/* a single page must always be possible! */
+			BUG_ON(bio->bi_vcnt == 0);
+			goto next_bio;
+		}
+		ds -= len;
+		sector += len >> 9;
+		--nr_pages;
+	}
+	D_ASSERT(page == NULL);
+	D_ASSERT(ds == 0);
+
+	atomic_set(&e->pending_bios, n_bios);
+	do {
+		bio = bios;
+		bios = bios->bi_next;
+		bio->bi_next = NULL;
+
+		/* strip off BIO_RW_UNPLUG unless it is the last bio */
+		if (bios)
+			bio->bi_rw &= ~(1<<BIO_RW_UNPLUG);
+
+		drbd_generic_make_request(mdev, fault_type, bio);
+
+		/* strip off BIO_RW_BARRIER,
+		 * unless it is the first or last bio */
+		if (bios && bios->bi_next)
+			bios->bi_rw &= ~(1<<BIO_RW_BARRIER);
+	} while (bios);
+	maybe_kick_lo(mdev);
+	return 0;
+
+fail:
+	while (bios) {
+		bio = bios;
+		bios = bios->bi_next;
+		bio_put(bio);
+	}
+	return -ENOMEM;
+}
+
+/**
  * w_e_reissue() - Worker callback; Resubmit a bio, without BIO_RW_BARRIER set
  * @mdev:	DRBD device.
  * @w:		work object.
@@ -1128,8 +1241,6 @@ void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo)
 int w_e_reissue(struct drbd_conf *mdev, struct drbd_work *w, int cancel) __releases(local)
 {
 	struct drbd_epoch_entry *e = (struct drbd_epoch_entry *)w;
-	struct bio *bio = e->private_bio;
-
 	/* We leave DE_CONTAINS_A_BARRIER and EE_IS_BARRIER in place,
 	   (and DE_BARRIER_IN_NEXT_EPOCH_ISSUED in the previous Epoch)
 	   so that we can finish that epoch in drbd_may_finish_epoch().
@@ -1143,33 +1254,17 @@ int w_e_reissue(struct drbd_conf *mdev, struct drbd_work *w, int cancel) __relea
 	if (previous_epoch(mdev, e->epoch))
 		dev_warn(DEV, "Write ordering was not enforced (one time event)\n");
 
-	/* prepare bio for re-submit,
-	 * re-init volatile members */
 	/* we still have a local reference,
 	 * get_ldev was done in receive_Data. */
-	bio->bi_bdev = mdev->ldev->backing_bdev;
-	bio->bi_sector = e->sector;
-	bio->bi_size = e->size;
-	bio->bi_idx = 0;
-
-	bio->bi_flags &= ~(BIO_POOL_MASK - 1);
-	bio->bi_flags |= 1 << BIO_UPTODATE;
-
-	/* don't know whether this is necessary: */
-	bio->bi_phys_segments = 0;
-	bio->bi_next = NULL;
-
-	/* these should be unchanged: */
-	/* bio->bi_end_io = drbd_endio_write_sec; */
-	/* bio->bi_vcnt = whatever; */
 
 	e->w.cb = e_end_block;
-
-	/* This is no longer a barrier request. */
-	bio->bi_rw &= ~(1UL << BIO_RW_BARRIER);
-
-	drbd_generic_make_request(mdev, DRBD_FAULT_DT_WR, bio);
-
+	if (drbd_submit_ee(mdev, e, WRITE, DRBD_FAULT_DT_WR) != 0) {
+		/* drbd_submit_ee fails for one reason only:
+		 * if was not able to allocate sufficient bios.
+		 * requeue, try again later. */
+		e->w.cb = w_e_reissue;
+		drbd_queue_work(&mdev->data.work, &e->w);
+	}
 	return 1;
 }
 
@@ -1261,13 +1356,13 @@ static int receive_Barrier(struct drbd_conf *mdev, struct p_header *h)
 static struct drbd_epoch_entry *
 read_in_block(struct drbd_conf *mdev, u64 id, sector_t sector, int data_size) __must_hold(local)
 {
+	const sector_t capacity = drbd_get_capacity(mdev->this_bdev);
 	struct drbd_epoch_entry *e;
-	struct bio_vec *bvec;
 	struct page *page;
-	struct bio *bio;
-	int dgs, ds, i, rr;
+	int dgs, ds, rr;
 	void *dig_in = mdev->int_dig_in;
 	void *dig_vv = mdev->int_dig_vv;
+	unsigned long *data;
 
 	dgs = (mdev->agreed_pro_version >= 87 && mdev->integrity_r_tfm) ?
 		crypto_hash_digestsize(mdev->integrity_r_tfm) : 0;
@@ -1286,29 +1381,44 @@ read_in_block(struct drbd_conf *mdev, u64 id, sector_t sector, int data_size) __
 	ERR_IF(data_size &  0x1ff) return NULL;
 	ERR_IF(data_size >  DRBD_MAX_SEGMENT_SIZE) return NULL;
 
+	/* even though we trust out peer,
+	 * we sometimes have to double check. */
+	if (sector + (data_size>>9) > capacity) {
+		dev_err(DEV, "capacity: %llus < sector: %llus + size: %u\n",
+			(unsigned long long)capacity,
+			(unsigned long long)sector, data_size);
+		return NULL;
+	}
+
 	/* GFP_NOIO, because we must not cause arbitrary write-out: in a DRBD
 	 * "criss-cross" setup, that might cause write-out on some other DRBD,
 	 * which in turn might block on the other node at this very place.  */
 	e = drbd_alloc_ee(mdev, id, sector, data_size, GFP_NOIO);
 	if (!e)
 		return NULL;
-	bio = e->private_bio;
+
 	ds = data_size;
-	bio_for_each_segment(bvec, bio, i) {
-		page = bvec->bv_page;
-		rr = drbd_recv(mdev, kmap(page), min_t(int, ds, PAGE_SIZE));
+	page = e->pages;
+	page_chain_for_each(page) {
+		unsigned len = min_t(int, ds, PAGE_SIZE);
+		data = kmap(page);
+		rr = drbd_recv(mdev, data, len);
+		if (FAULT_ACTIVE(mdev, DRBD_FAULT_RECEIVE)) {
+			dev_err(DEV, "Fault injection: Corrupting data on receive\n");
+			data[0] = data[0] ^ (unsigned long)-1;
+		}
 		kunmap(page);
-		if (rr != min_t(int, ds, PAGE_SIZE)) {
+		if (rr != len) {
 			drbd_free_ee(mdev, e);
 			dev_warn(DEV, "short read receiving data: read %d expected %d\n",
-			     rr, min_t(int, ds, PAGE_SIZE));
+			     rr, len);
 			return NULL;
 		}
 		ds -= rr;
 	}
 
 	if (dgs) {
-		drbd_csum(mdev, mdev->integrity_r_tfm, bio, dig_vv);
+		drbd_csum_ee(mdev, mdev->integrity_r_tfm, e, dig_vv);
 		if (memcmp(dig_in, dig_vv, dgs)) {
 			dev_err(DEV, "Digest integrity check FAILED.\n");
 			drbd_bcast_ee(mdev, "digest failed",
@@ -1330,7 +1440,10 @@ static int drbd_drain_block(struct drbd_conf *mdev, int data_size)
 	int rr, rv = 1;
 	void *data;
 
-	page = drbd_pp_alloc(mdev, 1);
+	if (!data_size)
+		return TRUE;
+
+	page = drbd_pp_alloc(mdev, 1, 1);
 
 	data = kmap(page);
 	while (data_size) {
@@ -1394,7 +1507,7 @@ static int recv_dless_read(struct drbd_conf *mdev, struct drbd_request *req,
 	}
 
 	if (dgs) {
-		drbd_csum(mdev, mdev->integrity_r_tfm, bio, dig_vv);
+		drbd_csum_bio(mdev, mdev->integrity_r_tfm, bio, dig_vv);
 		if (memcmp(dig_in, dig_vv, dgs)) {
 			dev_err(DEV, "Digest integrity check FAILED. Broken NICs?\n");
 			return 0;
@@ -1415,7 +1528,7 @@ static int e_end_resync_block(struct drbd_conf *mdev, struct drbd_work *w, int u
 
 	D_ASSERT(hlist_unhashed(&e->colision));
 
-	if (likely(drbd_bio_uptodate(e->private_bio))) {
+	if (likely((e->flags & EE_WAS_ERROR) == 0)) {
 		drbd_set_in_sync(mdev, sector, e->size);
 		ok = drbd_send_ack(mdev, P_RS_WRITE_ACK, e);
 	} else {
@@ -1434,30 +1547,28 @@ static int recv_resync_read(struct drbd_conf *mdev, sector_t sector, int data_si
 	struct drbd_epoch_entry *e;
 
 	e = read_in_block(mdev, ID_SYNCER, sector, data_size);
-	if (!e) {
-		put_ldev(mdev);
-		return FALSE;
-	}
+	if (!e)
+		goto fail;
 
 	dec_rs_pending(mdev);
 
-	e->private_bio->bi_end_io = drbd_endio_write_sec;
-	e->private_bio->bi_rw = WRITE;
-	e->w.cb = e_end_resync_block;
-
 	inc_unacked(mdev);
 	/* corresponding dec_unacked() in e_end_resync_block()
 	 * respective _drbd_clear_done_ee */
 
+	e->w.cb = e_end_resync_block;
+
 	spin_lock_irq(&mdev->req_lock);
 	list_add(&e->w.list, &mdev->sync_ee);
 	spin_unlock_irq(&mdev->req_lock);
 
-	drbd_generic_make_request(mdev, DRBD_FAULT_RS_WR, e->private_bio);
-	/* accounting done in endio */
+	if (drbd_submit_ee(mdev, e, WRITE, DRBD_FAULT_RS_WR) == 0)
+		return TRUE;
 
-	maybe_kick_lo(mdev);
-	return TRUE;
+	drbd_free_ee(mdev, e);
+fail:
+	put_ldev(mdev);
+	return FALSE;
 }
 
 static int receive_DataReply(struct drbd_conf *mdev, struct p_header *h)
@@ -1552,7 +1663,7 @@ static int e_end_block(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 	}
 
 	if (mdev->net_conf->wire_protocol == DRBD_PROT_C) {
-		if (likely(drbd_bio_uptodate(e->private_bio))) {
+		if (likely((e->flags & EE_WAS_ERROR) == 0)) {
 			pcmd = (mdev->state.conn >= C_SYNC_SOURCE &&
 				mdev->state.conn <= C_PAUSED_SYNC_T &&
 				e->flags & EE_MAY_SET_IN_SYNC) ?
@@ -1698,7 +1809,6 @@ static int receive_Data(struct drbd_conf *mdev, struct p_header *h)
 		return FALSE;
 	}
 
-	e->private_bio->bi_end_io = drbd_endio_write_sec;
 	e->w.cb = e_end_block;
 
 	spin_lock(&mdev->epoch_lock);
@@ -1894,12 +2004,8 @@ static int receive_Data(struct drbd_conf *mdev, struct p_header *h)
 		drbd_al_begin_io(mdev, e->sector);
 	}
 
-	e->private_bio->bi_rw = rw;
-	drbd_generic_make_request(mdev, DRBD_FAULT_DT_WR, e->private_bio);
-	/* accounting done in endio */
-
-	maybe_kick_lo(mdev);
-	return TRUE;
+	if (drbd_submit_ee(mdev, e, rw, DRBD_FAULT_DT_WR) == 0)
+		return TRUE;
 
 out_interrupted:
 	/* yes, the epoch_size now is imbalanced.
@@ -1945,7 +2051,7 @@ static int receive_DataRequest(struct drbd_conf *mdev, struct p_header *h)
 			    "no local data.\n");
 		drbd_send_ack_rp(mdev, h->command == P_DATA_REQUEST ? P_NEG_DREPLY :
 				 P_NEG_RS_DREPLY , p);
-		return TRUE;
+		return drbd_drain_block(mdev, h->length - brps);
 	}
 
 	/* GFP_NOIO, because we must not cause arbitrary write-out: in a DRBD
@@ -1957,9 +2063,6 @@ static int receive_DataRequest(struct drbd_conf *mdev, struct p_header *h)
 		return FALSE;
 	}
 
-	e->private_bio->bi_rw = READ;
-	e->private_bio->bi_end_io = drbd_endio_read_sec;
-
 	switch (h->command) {
 	case P_DATA_REQUEST:
 		e->w.cb = w_e_end_data_req;
@@ -2053,10 +2156,8 @@ static int receive_DataRequest(struct drbd_conf *mdev, struct p_header *h)
 
 	inc_unacked(mdev);
 
-	drbd_generic_make_request(mdev, fault_type, e->private_bio);
-	maybe_kick_lo(mdev);
-
-	return TRUE;
+	if (drbd_submit_ee(mdev, e, READ, fault_type) == 0)
+		return TRUE;
 
 out_free_e:
 	kfree(di);
@@ -2473,6 +2574,9 @@ static enum drbd_conns drbd_sync_handshake(struct drbd_conf *mdev, enum drbd_rol
 		     hg > 0 ? "source" : "target");
 	}
 
+	if (abs(hg) == 100)
+		drbd_khelper(mdev, "initial-split-brain");
+
 	if (hg == 100 || (hg == -100 && mdev->net_conf->always_asbp)) {
 		int pcount = (mdev->state.role == R_PRIMARY)
 			   + (peer_role == R_PRIMARY);
@@ -2518,7 +2622,7 @@ static enum drbd_conns drbd_sync_handshake(struct drbd_conf *mdev, enum drbd_rol
 		 * after an attempted attach on a diskless node.
 		 * We just refuse to attach -- well, we drop the "connection"
 		 * to that disk, in a way... */
-		dev_alert(DEV, "Split-Brain detected, dropping connection!\n");
+		dev_alert(DEV, "Split-Brain detected but unresolved, dropping connection!\n");
 		drbd_khelper(mdev, "split-brain");
 		return C_MASK;
 	}
@@ -2849,7 +2953,7 @@ static int receive_sizes(struct drbd_conf *mdev, struct p_header *h)
 	unsigned int max_seg_s;
 	sector_t p_size, p_usize, my_usize;
 	int ldsc = 0; /* local disk size changed */
-	enum drbd_conns nconn;
+	enum dds_flags ddsf;
 
 	ERR_IF(h->length != (sizeof(*p)-sizeof(*h))) return FALSE;
 	if (drbd_recv(mdev, h->payload, h->length) != h->length)
@@ -2905,8 +3009,9 @@ static int receive_sizes(struct drbd_conf *mdev, struct p_header *h)
 	}
 #undef min_not_zero
 
+	ddsf = be16_to_cpu(p->dds_flags);
 	if (get_ldev(mdev)) {
-	  dd = drbd_determin_dev_size(mdev, 0);
+		dd = drbd_determin_dev_size(mdev, ddsf);
 		put_ldev(mdev);
 		if (dd == dev_size_error)
 			return FALSE;
@@ -2916,33 +3021,21 @@ static int receive_sizes(struct drbd_conf *mdev, struct p_header *h)
 		drbd_set_my_capacity(mdev, p_size);
 	}
 
-	if (mdev->p_uuid && mdev->state.conn <= C_CONNECTED && get_ldev(mdev)) {
-		nconn = drbd_sync_handshake(mdev,
-				mdev->state.peer, mdev->state.pdsk);
-		put_ldev(mdev);
-
-		if (nconn == C_MASK) {
-			drbd_force_state(mdev, NS(conn, C_DISCONNECTING));
-			return FALSE;
-		}
-
-		if (drbd_request_state(mdev, NS(conn, nconn)) < SS_SUCCESS) {
-			drbd_force_state(mdev, NS(conn, C_DISCONNECTING));
-			return FALSE;
-		}
-	}
-
 	if (get_ldev(mdev)) {
 		if (mdev->ldev->known_size != drbd_get_capacity(mdev->ldev->backing_bdev)) {
 			mdev->ldev->known_size = drbd_get_capacity(mdev->ldev->backing_bdev);
 			ldsc = 1;
 		}
 
-		max_seg_s = be32_to_cpu(p->max_segment_size);
+		if (mdev->agreed_pro_version < 94)
+			max_seg_s = be32_to_cpu(p->max_segment_size);
+		else /* drbd 8.3.8 onwards */
+			max_seg_s = DRBD_MAX_SEGMENT_SIZE;
+
 		if (max_seg_s != queue_max_segment_size(mdev->rq_queue))
 			drbd_setup_queue_param(mdev, max_seg_s);
 
-		drbd_setup_order_type(mdev, be32_to_cpu(p->queue_order_type));
+		drbd_setup_order_type(mdev, be16_to_cpu(p->queue_order_type));
 		put_ldev(mdev);
 	}
 
@@ -2951,14 +3044,17 @@ static int receive_sizes(struct drbd_conf *mdev, struct p_header *h)
 		    drbd_get_capacity(mdev->this_bdev) || ldsc) {
 			/* we have different sizes, probably peer
 			 * needs to know my new size... */
-			drbd_send_sizes(mdev, 0);
+			drbd_send_sizes(mdev, 0, ddsf);
 		}
 		if (test_and_clear_bit(RESIZE_PENDING, &mdev->flags) ||
 		    (dd == grew && mdev->state.conn == C_CONNECTED)) {
 			if (mdev->state.pdsk >= D_INCONSISTENT &&
-			    mdev->state.disk >= D_INCONSISTENT)
-				resync_after_online_grow(mdev);
-			else
+			    mdev->state.disk >= D_INCONSISTENT) {
+				if (ddsf & DDSF_NO_RESYNC)
+					dev_info(DEV, "Resync of new storage suppressed with --assume-clean\n");
+				else
+					resync_after_online_grow(mdev);
+			} else
 				set_bit(RESYNC_AFTER_NEG, &mdev->flags);
 		}
 	}
@@ -3490,6 +3586,92 @@ static int receive_UnplugRemote(struct drbd_conf *mdev, struct p_header *h)
 	return TRUE;
 }
 
+static void timeval_sub_us(struct timeval* tv, unsigned int us)
+{
+	tv->tv_sec -= us / 1000000;
+	us = us % 1000000;
+	if (tv->tv_usec > us) {
+		tv->tv_usec += 1000000;
+		tv->tv_sec--;
+	}
+	tv->tv_usec -= us;
+}
+
+static void got_delay_probe(struct drbd_conf *mdev, int from, struct p_delay_probe *p)
+{
+	struct delay_probe *dp;
+	struct list_head *le;
+	struct timeval now;
+	int seq_num;
+	int offset;
+	int data_delay;
+
+	seq_num = be32_to_cpu(p->seq_num);
+	offset  = be32_to_cpu(p->offset);
+
+	spin_lock(&mdev->peer_seq_lock);
+	if (!list_empty(&mdev->delay_probes)) {
+		if (from == USE_DATA_SOCKET)
+			le = mdev->delay_probes.next;
+		else
+			le = mdev->delay_probes.prev;
+
+		dp = list_entry(le, struct delay_probe, list);
+
+		if (dp->seq_num == seq_num) {
+			list_del(le);
+			spin_unlock(&mdev->peer_seq_lock);
+			do_gettimeofday(&now);
+			timeval_sub_us(&now, offset);
+			data_delay =
+				now.tv_usec - dp->time.tv_usec +
+				(now.tv_sec - dp->time.tv_sec) * 1000000;
+
+			if (data_delay > 0)
+				mdev->data_delay = data_delay;
+
+			kfree(dp);
+			return;
+		}
+
+		if (dp->seq_num > seq_num) {
+			spin_unlock(&mdev->peer_seq_lock);
+			dev_warn(DEV, "Previous allocation failure of struct delay_probe?\n");
+			return; /* Do not alloca a struct delay_probe.... */
+		}
+	}
+	spin_unlock(&mdev->peer_seq_lock);
+
+	dp = kmalloc(sizeof(struct delay_probe), GFP_NOIO);
+	if (!dp) {
+		dev_warn(DEV, "Failed to allocate a struct delay_probe, do not worry.\n");
+		return;
+	}
+
+	dp->seq_num = seq_num;
+	do_gettimeofday(&dp->time);
+	timeval_sub_us(&dp->time, offset);
+
+	spin_lock(&mdev->peer_seq_lock);
+	if (from == USE_DATA_SOCKET)
+		list_add(&dp->list, &mdev->delay_probes);
+	else
+		list_add_tail(&dp->list, &mdev->delay_probes);
+	spin_unlock(&mdev->peer_seq_lock);
+}
+
+static int receive_delay_probe(struct drbd_conf *mdev, struct p_header *h)
+{
+	struct p_delay_probe *p = (struct p_delay_probe *)h;
+
+	ERR_IF(h->length != (sizeof(*p)-sizeof(*h))) return FALSE;
+	if (drbd_recv(mdev, h->payload, h->length) != h->length)
+		return FALSE;
+
+	got_delay_probe(mdev, USE_DATA_SOCKET, p);
+	return TRUE;
+}
+
 typedef int (*drbd_cmd_handler_f)(struct drbd_conf *, struct p_header *);
 
 static drbd_cmd_handler_f drbd_default_handler[] = {
@@ -3513,6 +3695,7 @@ static drbd_cmd_handler_f drbd_default_handler[] = {
 	[P_OV_REQUEST]      = receive_DataRequest,
 	[P_OV_REPLY]        = receive_DataRequest,
 	[P_CSUM_RS_REQUEST]    = receive_DataRequest,
+	[P_DELAY_PROBE]     = receive_delay_probe,
 	/* anything missing from this table is in
 	 * the asender_tbl, see get_asender_cmd */
 	[P_MAX_CMD]	    = NULL,
@@ -3739,7 +3922,7 @@ static void drbd_disconnect(struct drbd_conf *mdev)
 		dev_info(DEV, "net_ee not empty, killed %u entries\n", i);
 	i = atomic_read(&mdev->pp_in_use);
 	if (i)
-		dev_info(DEV, "pp_in_use = %u, expected 0\n", i);
+		dev_info(DEV, "pp_in_use = %d, expected 0\n", i);
 
 	D_ASSERT(list_empty(&mdev->read_ee));
 	D_ASSERT(list_empty(&mdev->active_ee));
@@ -4232,7 +4415,6 @@ static int got_NegRSDReply(struct drbd_conf *mdev, struct p_header *h)
 
 	sector = be64_to_cpu(p->sector);
 	size = be32_to_cpu(p->blksize);
-	D_ASSERT(p->block_id == ID_SYNCER);
 
 	update_peer_seq(mdev, be32_to_cpu(p->seq_num));
 
@@ -4290,6 +4472,14 @@ static int got_OVResult(struct drbd_conf *mdev, struct p_header *h)
 	return TRUE;
 }
 
+static int got_delay_probe_m(struct drbd_conf *mdev, struct p_header *h)
+{
+	struct p_delay_probe *p = (struct p_delay_probe *)h;
+
+	got_delay_probe(mdev, USE_META_SOCKET, p);
+	return TRUE;
+}
+
 struct asender_cmd {
 	size_t pkt_size;
 	int (*process)(struct drbd_conf *mdev, struct p_header *h);
@@ -4314,6 +4504,7 @@ static struct asender_cmd *get_asender_cmd(int cmd)
 	[P_BARRIER_ACK]	    = { sizeof(struct p_barrier_ack), got_BarrierAck },
 	[P_STATE_CHG_REPLY] = { sizeof(struct p_req_state_reply), got_RqSReply },
 	[P_RS_IS_IN_SYNC]   = { sizeof(struct p_block_ack), got_IsInSync },
+	[P_DELAY_PROBE]     = { sizeof(struct p_delay_probe), got_delay_probe_m },
 	[P_MAX_CMD]	    = { 0, NULL },
 	};
 	if (cmd > P_MAX_CMD || asender_tbl[cmd].process == NULL)
diff --git a/drivers/block/drbd/drbd_req.c b/drivers/block/drbd/drbd_req.c
index de81ab7b4627..654f1ef5cbb0 100644
--- a/drivers/block/drbd/drbd_req.c
+++ b/drivers/block/drbd/drbd_req.c
@@ -102,32 +102,7 @@ static void _req_is_done(struct drbd_conf *mdev, struct drbd_request *req, const
 		}
 	}
 
-	/* if it was a local io error, we want to notify our
-	 * peer about that, and see if we need to
-	 * detach the disk and stuff.
-	 * to avoid allocating some special work
-	 * struct, reuse the request. */
-
-	/* THINK
-	 * why do we do this not when we detect the error,
-	 * but delay it until it is "done", i.e. possibly
-	 * until the next barrier ack? */
-
-	if (rw == WRITE &&
-	    ((s & RQ_LOCAL_MASK) && !(s & RQ_LOCAL_OK))) {
-		if (!(req->w.list.next == LIST_POISON1 ||
-		      list_empty(&req->w.list))) {
-			/* DEBUG ASSERT only; if this triggers, we
-			 * probably corrupt the worker list here */
-			dev_err(DEV, "req->w.list.next = %p\n", req->w.list.next);
-			dev_err(DEV, "req->w.list.prev = %p\n", req->w.list.prev);
-		}
-		req->w.cb = w_io_error;
-		drbd_queue_work(&mdev->data.work, &req->w);
-		/* drbd_req_free() is done in w_io_error */
-	} else {
-		drbd_req_free(req);
-	}
+	drbd_req_free(req);
 }
 
 static void queue_barrier(struct drbd_conf *mdev)
@@ -453,9 +428,6 @@ void __req_mod(struct drbd_request *req, enum drbd_req_event what,
 		req->rq_state |= RQ_LOCAL_COMPLETED;
 		req->rq_state &= ~RQ_LOCAL_PENDING;
 
-		dev_alert(DEV, "Local WRITE failed sec=%llus size=%u\n",
-		      (unsigned long long)req->sector, req->size);
-		/* and now: check how to handle local io error. */
 		__drbd_chk_io_error(mdev, FALSE);
 		_req_may_be_done(req, m);
 		put_ldev(mdev);
@@ -475,22 +447,21 @@ void __req_mod(struct drbd_request *req, enum drbd_req_event what,
 		req->rq_state |= RQ_LOCAL_COMPLETED;
 		req->rq_state &= ~RQ_LOCAL_PENDING;
 
-		dev_alert(DEV, "Local READ failed sec=%llus size=%u\n",
-		      (unsigned long long)req->sector, req->size);
-		/* _req_mod(req,to_be_send); oops, recursion... */
 		D_ASSERT(!(req->rq_state & RQ_NET_MASK));
-		req->rq_state |= RQ_NET_PENDING;
-		inc_ap_pending(mdev);
 
 		__drbd_chk_io_error(mdev, FALSE);
 		put_ldev(mdev);
-		/* NOTE: if we have no connection,
-		 * or know the peer has no good data either,
-		 * then we don't actually need to "queue_for_net_read",
-		 * but we do so anyways, since the drbd_io_error()
-		 * and the potential state change to "Diskless"
-		 * needs to be done from process context */
 
+		/* no point in retrying if there is no good remote data,
+		 * or we have no connection. */
+		if (mdev->state.pdsk != D_UP_TO_DATE) {
+			_req_may_be_done(req, m);
+			break;
+		}
+
+		/* _req_mod(req,to_be_send); oops, recursion... */
+		req->rq_state |= RQ_NET_PENDING;
+		inc_ap_pending(mdev);
 		/* fall through: _req_mod(req,queue_for_net_read); */
 
 	case queue_for_net_read:
@@ -600,6 +571,9 @@ void __req_mod(struct drbd_request *req, enum drbd_req_event what,
 		_req_may_be_done(req, m);
 		break;
 
+	case read_retry_remote_canceled:
+		req->rq_state &= ~RQ_NET_QUEUED;
+		/* fall through, in case we raced with drbd_disconnect */
 	case connection_lost_while_pending:
 		/* transfer log cleanup after connection loss */
 		/* assert something? */
@@ -722,6 +696,7 @@ static int drbd_make_request_common(struct drbd_conf *mdev, struct bio *bio)
 	struct drbd_request *req;
 	int local, remote;
 	int err = -EIO;
+	int ret = 0;
 
 	/* allocate outside of all locks; */
 	req = drbd_req_new(mdev, bio);
@@ -784,7 +759,7 @@ static int drbd_make_request_common(struct drbd_conf *mdev, struct bio *bio)
 			    (mdev->state.pdsk == D_INCONSISTENT &&
 			     mdev->state.conn >= C_CONNECTED));
 
-	if (!(local || remote)) {
+	if (!(local || remote) && !mdev->state.susp) {
 		dev_err(DEV, "IO ERROR: neither local nor remote disk\n");
 		goto fail_free_complete;
 	}
@@ -810,6 +785,16 @@ allocate_barrier:
 	/* GOOD, everything prepared, grab the spin_lock */
 	spin_lock_irq(&mdev->req_lock);
 
+	if (mdev->state.susp) {
+		/* If we got suspended, use the retry mechanism of
+		   generic_make_request() to restart processing of this
+		   bio. In the next call to drbd_make_request_26
+		   we sleep in inc_ap_bio() */
+		ret = 1;
+		spin_unlock_irq(&mdev->req_lock);
+		goto fail_free_complete;
+	}
+
 	if (remote) {
 		remote = (mdev->state.pdsk == D_UP_TO_DATE ||
 			    (mdev->state.pdsk == D_INCONSISTENT &&
@@ -947,12 +932,14 @@ fail_and_free_req:
 		req->private_bio = NULL;
 		put_ldev(mdev);
 	}
-	bio_endio(bio, err);
+	if (!ret)
+		bio_endio(bio, err);
+
 	drbd_req_free(req);
 	dec_ap_bio(mdev);
 	kfree(b);
 
-	return 0;
+	return ret;
 }
 
 /* helper function for drbd_make_request
@@ -962,11 +949,6 @@ fail_and_free_req:
  */
 static int drbd_fail_request_early(struct drbd_conf *mdev, int is_write)
 {
-	/* Unconfigured */
-	if (mdev->state.conn == C_DISCONNECTING &&
-	    mdev->state.disk == D_DISKLESS)
-		return 1;
-
 	if (mdev->state.role != R_PRIMARY &&
 		(!allow_oos || is_write)) {
 		if (__ratelimit(&drbd_ratelimit_state)) {
@@ -1070,15 +1052,21 @@ int drbd_make_request_26(struct request_queue *q, struct bio *bio)
 
 		/* we need to get a "reference count" (ap_bio_cnt)
 		 * to avoid races with the disconnect/reconnect/suspend code.
-		 * In case we need to split the bio here, we need to get two references
+		 * In case we need to split the bio here, we need to get three references
 		 * atomically, otherwise we might deadlock when trying to submit the
 		 * second one! */
-		inc_ap_bio(mdev, 2);
+		inc_ap_bio(mdev, 3);
 
 		D_ASSERT(e_enr == s_enr + 1);
 
-		drbd_make_request_common(mdev, &bp->bio1);
-		drbd_make_request_common(mdev, &bp->bio2);
+		while (drbd_make_request_common(mdev, &bp->bio1))
+			inc_ap_bio(mdev, 1);
+
+		while (drbd_make_request_common(mdev, &bp->bio2))
+			inc_ap_bio(mdev, 1);
+
+		dec_ap_bio(mdev);
+
 		bio_pair_release(bp);
 	}
 	return 0;
@@ -1115,7 +1103,7 @@ int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct
 	} else if (limit && get_ldev(mdev)) {
 		struct request_queue * const b =
 			mdev->ldev->backing_bdev->bd_disk->queue;
-		if (b->merge_bvec_fn && mdev->ldev->dc.use_bmbv) {
+		if (b->merge_bvec_fn) {
 			backing_limit = b->merge_bvec_fn(b, bvm, bvec);
 			limit = min(limit, backing_limit);
 		}
diff --git a/drivers/block/drbd/drbd_req.h b/drivers/block/drbd/drbd_req.h
index 16119d7056cc..02d575d24518 100644
--- a/drivers/block/drbd/drbd_req.h
+++ b/drivers/block/drbd/drbd_req.h
@@ -91,6 +91,7 @@ enum drbd_req_event {
 	send_failed,
 	handed_over_to_network,
 	connection_lost_while_pending,
+	read_retry_remote_canceled,
 	recv_acked_by_peer,
 	write_acked_by_peer,
 	write_acked_by_peer_and_sis, /* and set_in_sync */
diff --git a/drivers/block/drbd/drbd_strings.c b/drivers/block/drbd/drbd_strings.c
index 76863e3f05be..85179e1fb50a 100644
--- a/drivers/block/drbd/drbd_strings.c
+++ b/drivers/block/drbd/drbd_strings.c
@@ -70,7 +70,7 @@ static const char *drbd_disk_s_names[] = {
 
 static const char *drbd_state_sw_errors[] = {
 	[-SS_TWO_PRIMARIES] = "Multiple primaries not allowed by config",
-	[-SS_NO_UP_TO_DATE_DISK] = "Refusing to be Primary without at least one UpToDate disk",
+	[-SS_NO_UP_TO_DATE_DISK] = "Need access to UpToDate data",
 	[-SS_NO_LOCAL_DISK] = "Can not resync without local disk",
 	[-SS_NO_REMOTE_DISK] = "Can not resync without remote disk",
 	[-SS_CONNECTED_OUTDATES] = "Refusing to be Outdated while Connected",
diff --git a/drivers/block/drbd/drbd_worker.c b/drivers/block/drbd/drbd_worker.c
index d48a1dfd7b24..b623ceee2a4a 100644
--- a/drivers/block/drbd/drbd_worker.c
+++ b/drivers/block/drbd/drbd_worker.c
@@ -47,8 +47,7 @@ static int w_make_ov_request(struct drbd_conf *mdev, struct drbd_work *w, int ca
 
 /* defined here:
    drbd_md_io_complete
-   drbd_endio_write_sec
-   drbd_endio_read_sec
+   drbd_endio_sec
    drbd_endio_pri
 
  * more endio handlers:
@@ -85,27 +84,10 @@ void drbd_md_io_complete(struct bio *bio, int error)
 /* reads on behalf of the partner,
  * "submitted" by the receiver
  */
-void drbd_endio_read_sec(struct bio *bio, int error) __releases(local)
+void drbd_endio_read_sec_final(struct drbd_epoch_entry *e) __releases(local)
 {
 	unsigned long flags = 0;
-	struct drbd_epoch_entry *e = NULL;
-	struct drbd_conf *mdev;
-	int uptodate = bio_flagged(bio, BIO_UPTODATE);
-
-	e = bio->bi_private;
-	mdev = e->mdev;
-
-	if (error)
-		dev_warn(DEV, "read: error=%d s=%llus\n", error,
-				(unsigned long long)e->sector);
-	if (!error && !uptodate) {
-		dev_warn(DEV, "read: setting error to -EIO s=%llus\n",
-				(unsigned long long)e->sector);
-		/* strange behavior of some lower level drivers...
-		 * fail the request by clearing the uptodate flag,
-		 * but do not return any error?! */
-		error = -EIO;
-	}
+	struct drbd_conf *mdev = e->mdev;
 
 	D_ASSERT(e->block_id != ID_VACANT);
 
@@ -114,49 +96,38 @@ void drbd_endio_read_sec(struct bio *bio, int error) __releases(local)
 	list_del(&e->w.list);
 	if (list_empty(&mdev->read_ee))
 		wake_up(&mdev->ee_wait);
+	if (test_bit(__EE_WAS_ERROR, &e->flags))
+		__drbd_chk_io_error(mdev, FALSE);
 	spin_unlock_irqrestore(&mdev->req_lock, flags);
 
-	drbd_chk_io_error(mdev, error, FALSE);
 	drbd_queue_work(&mdev->data.work, &e->w);
 	put_ldev(mdev);
 }
 
+static int is_failed_barrier(int ee_flags)
+{
+	return (ee_flags & (EE_IS_BARRIER|EE_WAS_ERROR|EE_RESUBMITTED))
+			== (EE_IS_BARRIER|EE_WAS_ERROR);
+}
+
 /* writes on behalf of the partner, or resync writes,
- * "submitted" by the receiver.
- */
-void drbd_endio_write_sec(struct bio *bio, int error) __releases(local)
+ * "submitted" by the receiver, final stage.  */
+static void drbd_endio_write_sec_final(struct drbd_epoch_entry *e) __releases(local)
 {
 	unsigned long flags = 0;
-	struct drbd_epoch_entry *e = NULL;
-	struct drbd_conf *mdev;
+	struct drbd_conf *mdev = e->mdev;
 	sector_t e_sector;
 	int do_wake;
 	int is_syncer_req;
 	int do_al_complete_io;
-	int uptodate = bio_flagged(bio, BIO_UPTODATE);
-	int is_barrier = bio_rw_flagged(bio, BIO_RW_BARRIER);
-
-	e = bio->bi_private;
-	mdev = e->mdev;
 
-	if (error)
-		dev_warn(DEV, "write: error=%d s=%llus\n", error,
-				(unsigned long long)e->sector);
-	if (!error && !uptodate) {
-		dev_warn(DEV, "write: setting error to -EIO s=%llus\n",
-				(unsigned long long)e->sector);
-		/* strange behavior of some lower level drivers...
-		 * fail the request by clearing the uptodate flag,
-		 * but do not return any error?! */
-		error = -EIO;
-	}
-
-	/* error == -ENOTSUPP would be a better test,
-	 * alas it is not reliable */
-	if (error && is_barrier && e->flags & EE_IS_BARRIER) {
+	/* if this is a failed barrier request, disable use of barriers,
+	 * and schedule for resubmission */
+	if (is_failed_barrier(e->flags)) {
 		drbd_bump_write_ordering(mdev, WO_bdev_flush);
 		spin_lock_irqsave(&mdev->req_lock, flags);
 		list_del(&e->w.list);
+		e->flags = (e->flags & ~EE_WAS_ERROR) | EE_RESUBMITTED;
 		e->w.cb = w_e_reissue;
 		/* put_ldev actually happens below, once we come here again. */
 		__release(local);
@@ -167,17 +138,16 @@ void drbd_endio_write_sec(struct bio *bio, int error) __releases(local)
 
 	D_ASSERT(e->block_id != ID_VACANT);
 
-	spin_lock_irqsave(&mdev->req_lock, flags);
-	mdev->writ_cnt += e->size >> 9;
-	is_syncer_req = is_syncer_block_id(e->block_id);
-
 	/* after we moved e to done_ee,
 	 * we may no longer access it,
 	 * it may be freed/reused already!
 	 * (as soon as we release the req_lock) */
 	e_sector = e->sector;
 	do_al_complete_io = e->flags & EE_CALL_AL_COMPLETE_IO;
+	is_syncer_req = is_syncer_block_id(e->block_id);
 
+	spin_lock_irqsave(&mdev->req_lock, flags);
+	mdev->writ_cnt += e->size >> 9;
 	list_del(&e->w.list); /* has been on active_ee or sync_ee */
 	list_add_tail(&e->w.list, &mdev->done_ee);
 
@@ -190,7 +160,7 @@ void drbd_endio_write_sec(struct bio *bio, int error) __releases(local)
 		? list_empty(&mdev->sync_ee)
 		: list_empty(&mdev->active_ee);
 
-	if (error)
+	if (test_bit(__EE_WAS_ERROR, &e->flags))
 		__drbd_chk_io_error(mdev, FALSE);
 	spin_unlock_irqrestore(&mdev->req_lock, flags);
 
@@ -205,7 +175,42 @@ void drbd_endio_write_sec(struct bio *bio, int error) __releases(local)
 
 	wake_asender(mdev);
 	put_ldev(mdev);
+}
+
+/* writes on behalf of the partner, or resync writes,
+ * "submitted" by the receiver.
+ */
+void drbd_endio_sec(struct bio *bio, int error)
+{
+	struct drbd_epoch_entry *e = bio->bi_private;
+	struct drbd_conf *mdev = e->mdev;
+	int uptodate = bio_flagged(bio, BIO_UPTODATE);
+	int is_write = bio_data_dir(bio) == WRITE;
+
+	if (error)
+		dev_warn(DEV, "%s: error=%d s=%llus\n",
+				is_write ? "write" : "read", error,
+				(unsigned long long)e->sector);
+	if (!error && !uptodate) {
+		dev_warn(DEV, "%s: setting error to -EIO s=%llus\n",
+				is_write ? "write" : "read",
+				(unsigned long long)e->sector);
+		/* strange behavior of some lower level drivers...
+		 * fail the request by clearing the uptodate flag,
+		 * but do not return any error?! */
+		error = -EIO;
+	}
+
+	if (error)
+		set_bit(__EE_WAS_ERROR, &e->flags);
 
+	bio_put(bio); /* no need for the bio anymore */
+	if (atomic_dec_and_test(&e->pending_bios)) {
+		if (is_write)
+			drbd_endio_write_sec_final(e);
+		else
+			drbd_endio_read_sec_final(e);
+	}
 }
 
 /* read, readA or write requests on R_PRIMARY coming from drbd_make_request
@@ -219,9 +224,6 @@ void drbd_endio_pri(struct bio *bio, int error)
 	enum drbd_req_event what;
 	int uptodate = bio_flagged(bio, BIO_UPTODATE);
 
-	if (error)
-		dev_warn(DEV, "p %s: error=%d\n",
-			 bio_data_dir(bio) == WRITE ? "write" : "read", error);
 	if (!error && !uptodate) {
 		dev_warn(DEV, "p %s: setting error to -EIO\n",
 			 bio_data_dir(bio) == WRITE ? "write" : "read");
@@ -252,20 +254,6 @@ void drbd_endio_pri(struct bio *bio, int error)
 		complete_master_bio(mdev, &m);
 }
 
-int w_io_error(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
-{
-	struct drbd_request *req = container_of(w, struct drbd_request, w);
-
-	/* NOTE: mdev->ldev can be NULL by the time we get here! */
-	/* D_ASSERT(mdev->ldev->dc.on_io_error != EP_PASS_ON); */
-
-	/* the only way this callback is scheduled is from _req_may_be_done,
-	 * when it is done and had a local write error, see comments there */
-	drbd_req_free(req);
-
-	return TRUE;
-}
-
 int w_read_retry_remote(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 {
 	struct drbd_request *req = container_of(w, struct drbd_request, w);
@@ -275,12 +263,9 @@ int w_read_retry_remote(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 	 * to give the disk the chance to relocate that block */
 
 	spin_lock_irq(&mdev->req_lock);
-	if (cancel ||
-	    mdev->state.conn < C_CONNECTED ||
-	    mdev->state.pdsk <= D_INCONSISTENT) {
-		_req_mod(req, send_canceled);
+	if (cancel || mdev->state.pdsk != D_UP_TO_DATE) {
+		_req_mod(req, read_retry_remote_canceled);
 		spin_unlock_irq(&mdev->req_lock);
-		dev_alert(DEV, "WE ARE LOST. Local IO failure, no peer.\n");
 		return 1;
 	}
 	spin_unlock_irq(&mdev->req_lock);
@@ -295,7 +280,34 @@ int w_resync_inactive(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 	return 1; /* Simply ignore this! */
 }
 
-void drbd_csum(struct drbd_conf *mdev, struct crypto_hash *tfm, struct bio *bio, void *digest)
+void drbd_csum_ee(struct drbd_conf *mdev, struct crypto_hash *tfm, struct drbd_epoch_entry *e, void *digest)
+{
+	struct hash_desc desc;
+	struct scatterlist sg;
+	struct page *page = e->pages;
+	struct page *tmp;
+	unsigned len;
+
+	desc.tfm = tfm;
+	desc.flags = 0;
+
+	sg_init_table(&sg, 1);
+	crypto_hash_init(&desc);
+
+	while ((tmp = page_chain_next(page))) {
+		/* all but the last page will be fully used */
+		sg_set_page(&sg, page, PAGE_SIZE, 0);
+		crypto_hash_update(&desc, &sg, sg.length);
+		page = tmp;
+	}
+	/* and now the last, possibly only partially used page */
+	len = e->size & (PAGE_SIZE - 1);
+	sg_set_page(&sg, page, len ?: PAGE_SIZE, 0);
+	crypto_hash_update(&desc, &sg, sg.length);
+	crypto_hash_final(&desc, digest);
+}
+
+void drbd_csum_bio(struct drbd_conf *mdev, struct crypto_hash *tfm, struct bio *bio, void *digest)
 {
 	struct hash_desc desc;
 	struct scatterlist sg;
@@ -329,11 +341,11 @@ static int w_e_send_csum(struct drbd_conf *mdev, struct drbd_work *w, int cancel
 		return 1;
 	}
 
-	if (likely(drbd_bio_uptodate(e->private_bio))) {
+	if (likely((e->flags & EE_WAS_ERROR) == 0)) {
 		digest_size = crypto_hash_digestsize(mdev->csums_tfm);
 		digest = kmalloc(digest_size, GFP_NOIO);
 		if (digest) {
-			drbd_csum(mdev, mdev->csums_tfm, e->private_bio, digest);
+			drbd_csum_ee(mdev, mdev->csums_tfm, e, digest);
 
 			inc_rs_pending(mdev);
 			ok = drbd_send_drequest_csum(mdev,
@@ -369,23 +381,21 @@ static int read_for_csum(struct drbd_conf *mdev, sector_t sector, int size)
 	/* GFP_TRY, because if there is no memory available right now, this may
 	 * be rescheduled for later. It is "only" background resync, after all. */
 	e = drbd_alloc_ee(mdev, DRBD_MAGIC+0xbeef, sector, size, GFP_TRY);
-	if (!e) {
-		put_ldev(mdev);
-		return 2;
-	}
+	if (!e)
+		goto fail;
 
 	spin_lock_irq(&mdev->req_lock);
 	list_add(&e->w.list, &mdev->read_ee);
 	spin_unlock_irq(&mdev->req_lock);
 
-	e->private_bio->bi_end_io = drbd_endio_read_sec;
-	e->private_bio->bi_rw = READ;
 	e->w.cb = w_e_send_csum;
+	if (drbd_submit_ee(mdev, e, READ, DRBD_FAULT_RS_RD) == 0)
+		return 1;
 
-	mdev->read_cnt += size >> 9;
-	drbd_generic_make_request(mdev, DRBD_FAULT_RS_RD, e->private_bio);
-
-	return 1;
+	drbd_free_ee(mdev, e);
+fail:
+	put_ldev(mdev);
+	return 2;
 }
 
 void resync_timer_fn(unsigned long data)
@@ -414,13 +424,25 @@ void resync_timer_fn(unsigned long data)
 		drbd_queue_work(&mdev->data.work, &mdev->resync_work);
 }
 
+static int calc_resync_rate(struct drbd_conf *mdev)
+{
+	int d = mdev->data_delay / 1000; /* us -> ms */
+	int td = mdev->sync_conf.throttle_th * 100;  /* 0.1s -> ms */
+	int hd = mdev->sync_conf.hold_off_th * 100;  /* 0.1s -> ms */
+	int cr = mdev->sync_conf.rate;
+
+	return d <= td ? cr :
+		d >= hd ? 0 :
+		cr + (cr * (td - d) / (hd - td));
+}
+
 int w_make_resync_request(struct drbd_conf *mdev,
 		struct drbd_work *w, int cancel)
 {
 	unsigned long bit;
 	sector_t sector;
 	const sector_t capacity = drbd_get_capacity(mdev->this_bdev);
-	int max_segment_size = queue_max_segment_size(mdev->rq_queue);
+	int max_segment_size;
 	int number, i, size, pe, mx;
 	int align, queued, sndbuf;
 
@@ -446,7 +468,13 @@ int w_make_resync_request(struct drbd_conf *mdev,
 		return 1;
 	}
 
-	number = SLEEP_TIME * mdev->sync_conf.rate / ((BM_BLOCK_SIZE/1024)*HZ);
+	/* starting with drbd 8.3.8, we can handle multi-bio EEs,
+	 * if it should be necessary */
+	max_segment_size = mdev->agreed_pro_version < 94 ?
+		queue_max_segment_size(mdev->rq_queue) : DRBD_MAX_SEGMENT_SIZE;
+
+	mdev->c_sync_rate = calc_resync_rate(mdev);
+	number = SLEEP_TIME * mdev->c_sync_rate  / ((BM_BLOCK_SIZE / 1024) * HZ);
 	pe = atomic_read(&mdev->rs_pending_cnt);
 
 	mutex_lock(&mdev->data.mutex);
@@ -509,12 +537,6 @@ next_sector:
 		 *
 		 * Additionally always align bigger requests, in order to
 		 * be prepared for all stripe sizes of software RAIDs.
-		 *
-		 * we _do_ care about the agreed-upon q->max_segment_size
-		 * here, as splitting up the requests on the other side is more
-		 * difficult.  the consequence is, that on lvm and md and other
-		 * "indirect" devices, this is dead code, since
-		 * q->max_segment_size will be PAGE_SIZE.
 		 */
 		align = 1;
 		for (;;) {
@@ -806,7 +828,7 @@ out:
 /* helper */
 static void move_to_net_ee_or_free(struct drbd_conf *mdev, struct drbd_epoch_entry *e)
 {
-	if (drbd_bio_has_active_page(e->private_bio)) {
+	if (drbd_ee_has_active_page(e)) {
 		/* This might happen if sendpage() has not finished */
 		spin_lock_irq(&mdev->req_lock);
 		list_add_tail(&e->w.list, &mdev->net_ee);
@@ -832,7 +854,7 @@ int w_e_end_data_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 		return 1;
 	}
 
-	if (likely(drbd_bio_uptodate(e->private_bio))) {
+	if (likely((e->flags & EE_WAS_ERROR) == 0)) {
 		ok = drbd_send_block(mdev, P_DATA_REPLY, e);
 	} else {
 		if (__ratelimit(&drbd_ratelimit_state))
@@ -873,7 +895,7 @@ int w_e_end_rsdata_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 		put_ldev(mdev);
 	}
 
-	if (likely(drbd_bio_uptodate(e->private_bio))) {
+	if (likely((e->flags & EE_WAS_ERROR) == 0)) {
 		if (likely(mdev->state.pdsk >= D_INCONSISTENT)) {
 			inc_rs_pending(mdev);
 			ok = drbd_send_block(mdev, P_RS_DATA_REPLY, e);
@@ -921,7 +943,7 @@ int w_e_end_csum_rs_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 
 	di = (struct digest_info *)(unsigned long)e->block_id;
 
-	if (likely(drbd_bio_uptodate(e->private_bio))) {
+	if (likely((e->flags & EE_WAS_ERROR) == 0)) {
 		/* quick hack to try to avoid a race against reconfiguration.
 		 * a real fix would be much more involved,
 		 * introducing more locking mechanisms */
@@ -931,7 +953,7 @@ int w_e_end_csum_rs_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 			digest = kmalloc(digest_size, GFP_NOIO);
 		}
 		if (digest) {
-			drbd_csum(mdev, mdev->csums_tfm, e->private_bio, digest);
+			drbd_csum_ee(mdev, mdev->csums_tfm, e, digest);
 			eq = !memcmp(digest, di->digest, digest_size);
 			kfree(digest);
 		}
@@ -973,14 +995,14 @@ int w_e_end_ov_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 	if (unlikely(cancel))
 		goto out;
 
-	if (unlikely(!drbd_bio_uptodate(e->private_bio)))
+	if (unlikely((e->flags & EE_WAS_ERROR) != 0))
 		goto out;
 
 	digest_size = crypto_hash_digestsize(mdev->verify_tfm);
 	/* FIXME if this allocation fails, online verify will not terminate! */
 	digest = kmalloc(digest_size, GFP_NOIO);
 	if (digest) {
-		drbd_csum(mdev, mdev->verify_tfm, e->private_bio, digest);
+		drbd_csum_ee(mdev, mdev->verify_tfm, e, digest);
 		inc_rs_pending(mdev);
 		ok = drbd_send_drequest_csum(mdev, e->sector, e->size,
 					     digest, digest_size, P_OV_REPLY);
@@ -1029,11 +1051,11 @@ int w_e_end_ov_reply(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 
 	di = (struct digest_info *)(unsigned long)e->block_id;
 
-	if (likely(drbd_bio_uptodate(e->private_bio))) {
+	if (likely((e->flags & EE_WAS_ERROR) == 0)) {
 		digest_size = crypto_hash_digestsize(mdev->verify_tfm);
 		digest = kmalloc(digest_size, GFP_NOIO);
 		if (digest) {
-			drbd_csum(mdev, mdev->verify_tfm, e->private_bio, digest);
+			drbd_csum_ee(mdev, mdev->verify_tfm, e, digest);
 
 			D_ASSERT(digest_size == di->digest_size);
 			eq = !memcmp(digest, di->digest, digest_size);
diff --git a/drivers/block/drbd/drbd_wrappers.h b/drivers/block/drbd/drbd_wrappers.h
index f93fa111ce50..defdb5013ea3 100644
--- a/drivers/block/drbd/drbd_wrappers.h
+++ b/drivers/block/drbd/drbd_wrappers.h
@@ -18,23 +18,9 @@ static inline void drbd_set_my_capacity(struct drbd_conf *mdev,
 
 #define drbd_bio_uptodate(bio) bio_flagged(bio, BIO_UPTODATE)
 
-static inline int drbd_bio_has_active_page(struct bio *bio)
-{
-	struct bio_vec *bvec;
-	int i;
-
-	__bio_for_each_segment(bvec, bio, i, 0) {
-		if (page_count(bvec->bv_page) > 1)
-			return 1;
-	}
-
-	return 0;
-}
-
 /* bi_end_io handlers */
 extern void drbd_md_io_complete(struct bio *bio, int error);
-extern void drbd_endio_read_sec(struct bio *bio, int error);
-extern void drbd_endio_write_sec(struct bio *bio, int error);
+extern void drbd_endio_sec(struct bio *bio, int error);
 extern void drbd_endio_pri(struct bio *bio, int error);
 
 /*
diff --git a/drivers/block/loop.c b/drivers/block/loop.c
index a90e83c9be96..6120922f459f 100644
--- a/drivers/block/loop.c
+++ b/drivers/block/loop.c
@@ -485,7 +485,7 @@ static int do_bio_filebacked(struct loop_device *lo, struct bio *bio)
 				goto out;
 			}
 
-			ret = vfs_fsync(file, file->f_path.dentry, 0);
+			ret = vfs_fsync(file, 0);
 			if (unlikely(ret)) {
 				ret = -EIO;
 				goto out;
@@ -495,7 +495,7 @@ static int do_bio_filebacked(struct loop_device *lo, struct bio *bio)
 		ret = lo_send(lo, bio, pos);
 
 		if (barrier && !ret) {
-			ret = vfs_fsync(file, file->f_path.dentry, 0);
+			ret = vfs_fsync(file, 0);
 			if (unlikely(ret))
 				ret = -EIO;
 		}
diff --git a/drivers/block/swim3.c b/drivers/block/swim3.c
index 59ca2b77b574..ed6fb91123ab 100644
--- a/drivers/block/swim3.c
+++ b/drivers/block/swim3.c
@@ -1004,7 +1004,7 @@ static const struct block_device_operations floppy_fops = {
 
 static int swim3_add_device(struct macio_dev *mdev, int index)
 {
-	struct device_node *swim = mdev->ofdev.node;
+	struct device_node *swim = mdev->ofdev.dev.of_node;
 	struct floppy_state *fs = &floppy_states[index];
 	int rc = -EBUSY;
 
@@ -1159,8 +1159,10 @@ static struct of_device_id swim3_match[] =
 
 static struct macio_driver swim3_driver =
 {
-	.name 		= "swim3",
-	.match_table	= swim3_match,
+	.driver = {
+		.name 		= "swim3",
+		.of_match_table	= swim3_match,
+	},
 	.probe		= swim3_attach,
 #if 0
 	.suspend	= swim3_suspend,
diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
index 2138a7ae050c..258bc2ae2885 100644
--- a/drivers/block/virtio_blk.c
+++ b/drivers/block/virtio_blk.c
@@ -50,7 +50,7 @@ static void blk_done(struct virtqueue *vq)
 	unsigned long flags;
 
 	spin_lock_irqsave(&vblk->lock, flags);
-	while ((vbr = vblk->vq->vq_ops->get_buf(vblk->vq, &len)) != NULL) {
+	while ((vbr = virtqueue_get_buf(vblk->vq, &len)) != NULL) {
 		int error;
 
 		switch (vbr->status) {
@@ -70,6 +70,8 @@ static void blk_done(struct virtqueue *vq)
 			vbr->req->sense_len = vbr->in_hdr.sense_len;
 			vbr->req->errors = vbr->in_hdr.errors;
 		}
+		if (blk_special_request(vbr->req))
+			vbr->req->errors = (error != 0);
 
 		__blk_end_request_all(vbr->req, error);
 		list_del(&vbr->list);
@@ -103,6 +105,11 @@ static bool do_req(struct request_queue *q, struct virtio_blk *vblk,
 		vbr->out_hdr.sector = 0;
 		vbr->out_hdr.ioprio = req_get_ioprio(vbr->req);
 		break;
+	case REQ_TYPE_SPECIAL:
+		vbr->out_hdr.type = VIRTIO_BLK_T_GET_ID;
+		vbr->out_hdr.sector = 0;
+		vbr->out_hdr.ioprio = req_get_ioprio(vbr->req);
+		break;
 	case REQ_TYPE_LINUX_BLOCK:
 		if (req->cmd[0] == REQ_LB_OP_FLUSH) {
 			vbr->out_hdr.type = VIRTIO_BLK_T_FLUSH;
@@ -151,7 +158,7 @@ static bool do_req(struct request_queue *q, struct virtio_blk *vblk,
 		}
 	}
 
-	if (vblk->vq->vq_ops->add_buf(vblk->vq, vblk->sg, out, in, vbr) < 0) {
+	if (virtqueue_add_buf(vblk->vq, vblk->sg, out, in, vbr) < 0) {
 		mempool_free(vbr, vblk->pool);
 		return false;
 	}
@@ -180,7 +187,7 @@ static void do_virtblk_request(struct request_queue *q)
 	}
 
 	if (issued)
-		vblk->vq->vq_ops->kick(vblk->vq);
+		virtqueue_kick(vblk->vq);
 }
 
 static void virtblk_prepare_flush(struct request_queue *q, struct request *req)
@@ -189,12 +196,45 @@ static void virtblk_prepare_flush(struct request_queue *q, struct request *req)
 	req->cmd[0] = REQ_LB_OP_FLUSH;
 }
 
+/* return id (s/n) string for *disk to *id_str
+ */
+static int virtblk_get_id(struct gendisk *disk, char *id_str)
+{
+	struct virtio_blk *vblk = disk->private_data;
+	struct request *req;
+	struct bio *bio;
+
+	bio = bio_map_kern(vblk->disk->queue, id_str, VIRTIO_BLK_ID_BYTES,
+			   GFP_KERNEL);
+	if (IS_ERR(bio))
+		return PTR_ERR(bio);
+
+	req = blk_make_request(vblk->disk->queue, bio, GFP_KERNEL);
+	if (IS_ERR(req)) {
+		bio_put(bio);
+		return PTR_ERR(req);
+	}
+
+	req->cmd_type = REQ_TYPE_SPECIAL;
+	return blk_execute_rq(vblk->disk->queue, vblk->disk, req, false);
+}
+
 static int virtblk_ioctl(struct block_device *bdev, fmode_t mode,
 			 unsigned cmd, unsigned long data)
 {
 	struct gendisk *disk = bdev->bd_disk;
 	struct virtio_blk *vblk = disk->private_data;
 
+	if (cmd == 0x56424944) { /* 'VBID' */
+		void __user *usr_data = (void __user *)data;
+		char id_str[VIRTIO_BLK_ID_BYTES];
+		int err;
+
+		err = virtblk_get_id(disk, id_str);
+		if (!err && copy_to_user(usr_data, id_str, VIRTIO_BLK_ID_BYTES))
+			err = -EFAULT;
+		return err;
+	}
 	/*
 	 * Only allow the generic SCSI ioctls if the host can support it.
 	 */
@@ -258,7 +298,9 @@ static int __devinit virtblk_probe(struct virtio_device *vdev)
 	err = virtio_config_val(vdev, VIRTIO_BLK_F_SEG_MAX,
 				offsetof(struct virtio_blk_config, seg_max),
 				&sg_elems);
-	if (err)
+
+	/* We need at least one SG element, whatever they say. */
+	if (err || !sg_elems)
 		sg_elems = 1;
 
 	/* We need an extra sg elements at head and tail. */
diff --git a/drivers/block/xsysace.c b/drivers/block/xsysace.c
index e1c95e208a66..a7b83c0a7eb5 100644
--- a/drivers/block/xsysace.c
+++ b/drivers/block/xsysace.c
@@ -1198,10 +1198,10 @@ ace_of_probe(struct of_device *op, const struct of_device_id *match)
 	dev_dbg(&op->dev, "ace_of_probe(%p, %p)\n", op, match);
 
 	/* device id */
-	id = of_get_property(op->node, "port-number", NULL);
+	id = of_get_property(op->dev.of_node, "port-number", NULL);
 
 	/* physaddr */
-	rc = of_address_to_resource(op->node, 0, &res);
+	rc = of_address_to_resource(op->dev.of_node, 0, &res);
 	if (rc) {
 		dev_err(&op->dev, "invalid address\n");
 		return rc;
@@ -1209,11 +1209,11 @@ ace_of_probe(struct of_device *op, const struct of_device_id *match)
 	physaddr = res.start;
 
 	/* irq */
-	irq = irq_of_parse_and_map(op->node, 0);
+	irq = irq_of_parse_and_map(op->dev.of_node, 0);
 
 	/* bus width */
 	bus_width = ACE_BUS_WIDTH_16;
-	if (of_find_property(op->node, "8-bit", NULL))
+	if (of_find_property(op->dev.of_node, "8-bit", NULL))
 		bus_width = ACE_BUS_WIDTH_8;
 
 	/* Call the bus-independant setup code */
@@ -1237,13 +1237,12 @@ static const struct of_device_id ace_of_match[] __devinitconst = {
 MODULE_DEVICE_TABLE(of, ace_of_match);
 
 static struct of_platform_driver ace_of_driver = {
-	.owner = THIS_MODULE,
-	.name = "xsysace",
-	.match_table = ace_of_match,
 	.probe = ace_of_probe,
 	.remove = __devexit_p(ace_of_remove),
 	.driver = {
 		.name = "xsysace",
+		.owner = THIS_MODULE,
+		.of_match_table = ace_of_match,
 	},
 };
 
diff --git a/drivers/cdrom/viocd.c b/drivers/cdrom/viocd.c
index cc435be0bc13..451cd7071b1d 100644
--- a/drivers/cdrom/viocd.c
+++ b/drivers/cdrom/viocd.c
@@ -567,7 +567,7 @@ static int viocd_probe(struct vio_dev *vdev, const struct vio_device_id *id)
 	struct disk_info *d;
 	struct cdrom_device_info *c;
 	struct request_queue *q;
-	struct device_node *node = vdev->dev.archdata.of_node;
+	struct device_node *node = vdev->dev.of_node;
 
 	deviceno = vdev->unit_address;
 	if (deviceno >= VIOCD_MAX_CD)
diff --git a/drivers/char/Kconfig b/drivers/char/Kconfig
index e21175be25d0..7cfcc629a7fd 100644
--- a/drivers/char/Kconfig
+++ b/drivers/char/Kconfig
@@ -1121,5 +1121,13 @@ config DEVPORT
 
 source "drivers/s390/char/Kconfig"
 
+config RAMOOPS
+	tristate "Log panic/oops to a RAM buffer"
+	depends on HAS_IOMEM
+	default n
+	help
+	  This enables panic and oops messages to be logged to a circular
+	  buffer in RAM where it can be read back at some later point.
+
 endmenu
 
diff --git a/drivers/char/Makefile b/drivers/char/Makefile
index d39be4cf1f5d..88d6eac69754 100644
--- a/drivers/char/Makefile
+++ b/drivers/char/Makefile
@@ -108,6 +108,7 @@ obj-$(CONFIG_HANGCHECK_TIMER)	+= hangcheck-timer.o
 obj-$(CONFIG_TCG_TPM)		+= tpm/
 
 obj-$(CONFIG_PS3_FLASH)		+= ps3flash.o
+obj-$(CONFIG_RAMOOPS)		+= ramoops.o
 
 obj-$(CONFIG_JS_RTC)		+= js-rtc.o
 js-rtc-y = rtc.o
diff --git a/drivers/char/agp/amd64-agp.c b/drivers/char/agp/amd64-agp.c
index 67ea3a60de74..70312da4c968 100644
--- a/drivers/char/agp/amd64-agp.c
+++ b/drivers/char/agp/amd64-agp.c
@@ -384,7 +384,7 @@ static int __devinit uli_agp_init(struct pci_dev *pdev)
 {
 	u32 httfea,baseaddr,enuscr;
 	struct pci_dev *dev1;
-	int i;
+	int i, ret;
 	unsigned size = amd64_fetch_size();
 
 	dev_info(&pdev->dev, "setting up ULi AGP\n");
@@ -400,15 +400,18 @@ static int __devinit uli_agp_init(struct pci_dev *pdev)
 
 	if (i == ARRAY_SIZE(uli_sizes)) {
 		dev_info(&pdev->dev, "no ULi size found for %d\n", size);
-		return -ENODEV;
+		ret = -ENODEV;
+		goto put;
 	}
 
 	/* shadow x86-64 registers into ULi registers */
 	pci_read_config_dword (k8_northbridges[0], AMD64_GARTAPERTUREBASE, &httfea);
 
 	/* if x86-64 aperture base is beyond 4G, exit here */
-	if ((httfea & 0x7fff) >> (32 - 25))
-		return -ENODEV;
+	if ((httfea & 0x7fff) >> (32 - 25)) {
+		ret = -ENODEV;
+		goto put;
+	}
 
 	httfea = (httfea& 0x7fff) << 25;
 
@@ -420,9 +423,10 @@ static int __devinit uli_agp_init(struct pci_dev *pdev)
 	enuscr= httfea+ (size * 1024 * 1024) - 1;
 	pci_write_config_dword(dev1, ULI_X86_64_HTT_FEA_REG, httfea);
 	pci_write_config_dword(dev1, ULI_X86_64_ENU_SCR_REG, enuscr);
-
+	ret = 0;
+put:
 	pci_dev_put(dev1);
-	return 0;
+	return ret;
 }
 
 
@@ -441,7 +445,7 @@ static int nforce3_agp_init(struct pci_dev *pdev)
 {
 	u32 tmp, apbase, apbar, aplimit;
 	struct pci_dev *dev1;
-	int i;
+	int i, ret;
 	unsigned size = amd64_fetch_size();
 
 	dev_info(&pdev->dev, "setting up Nforce3 AGP\n");
@@ -458,7 +462,8 @@ static int nforce3_agp_init(struct pci_dev *pdev)
 
 	if (i == ARRAY_SIZE(nforce3_sizes)) {
 		dev_info(&pdev->dev, "no NForce3 size found for %d\n", size);
-		return -ENODEV;
+		ret = -ENODEV;
+		goto put;
 	}
 
 	pci_read_config_dword(dev1, NVIDIA_X86_64_1_APSIZE, &tmp);
@@ -472,7 +477,8 @@ static int nforce3_agp_init(struct pci_dev *pdev)
 	/* if x86-64 aperture base is beyond 4G, exit here */
 	if ( (apbase & 0x7fff) >> (32 - 25) ) {
 		dev_info(&pdev->dev, "aperture base > 4G\n");
-		return -ENODEV;
+		ret = -ENODEV;
+		goto put;
 	}
 
 	apbase = (apbase & 0x7fff) << 25;
@@ -488,9 +494,11 @@ static int nforce3_agp_init(struct pci_dev *pdev)
 	pci_write_config_dword(dev1, NVIDIA_X86_64_1_APBASE2, apbase);
 	pci_write_config_dword(dev1, NVIDIA_X86_64_1_APLIMIT2, aplimit);
 
+	ret = 0;
+put:
 	pci_dev_put(dev1);
 
-	return 0;
+	return ret;
 }
 
 static int __devinit agp_amd64_probe(struct pci_dev *pdev,
diff --git a/drivers/char/agp/intel-gtt.c b/drivers/char/agp/intel-gtt.c
index e8ea6825822c..9344216183a4 100644
--- a/drivers/char/agp/intel-gtt.c
+++ b/drivers/char/agp/intel-gtt.c
@@ -1059,7 +1059,7 @@ static void intel_i9xx_setup_flush(void)
 	}
 }
 
-static int intel_i915_configure(void)
+static int intel_i9xx_configure(void)
 {
 	struct aper_size_info_fixed *current_size;
 	u32 temp;
@@ -1207,6 +1207,38 @@ static int intel_i9xx_fetch_size(void)
 	return 0;
 }
 
+static int intel_i915_get_gtt_size(void)
+{
+	int size;
+
+	if (IS_G33) {
+		u16 gmch_ctrl;
+
+		/* G33's GTT size defined in gmch_ctrl */
+		pci_read_config_word(agp_bridge->dev, I830_GMCH_CTRL, &gmch_ctrl);
+		switch (gmch_ctrl & G33_PGETBL_SIZE_MASK) {
+		case G33_PGETBL_SIZE_1M:
+			size = 1024;
+			break;
+		case G33_PGETBL_SIZE_2M:
+			size = 2048;
+			break;
+		default:
+			dev_info(&agp_bridge->dev->dev,
+				 "unknown page table size 0x%x, assuming 512KB\n",
+				(gmch_ctrl & G33_PGETBL_SIZE_MASK));
+			size = 512;
+		}
+	} else {
+		/* On previous hardware, the GTT size was just what was
+		 * required to map the aperture.
+		 */
+		size = agp_bridge->driver->fetch_size();
+	}
+
+	return KB(size);
+}
+
 /* The intel i915 automatically initializes the agp aperture during POST.
  * Use the memory already set aside for in the GTT.
  */
@@ -1216,7 +1248,7 @@ static int intel_i915_create_gatt_table(struct agp_bridge_data *bridge)
 	struct aper_size_info_fixed *size;
 	int num_entries;
 	u32 temp, temp2;
-	int gtt_map_size = 256 * 1024;
+	int gtt_map_size;
 
 	size = agp_bridge->current_size;
 	page_order = size->page_order;
@@ -1226,8 +1258,8 @@ static int intel_i915_create_gatt_table(struct agp_bridge_data *bridge)
 	pci_read_config_dword(intel_private.pcidev, I915_MMADDR, &temp);
 	pci_read_config_dword(intel_private.pcidev, I915_PTEADDR, &temp2);
 
-	if (IS_G33)
-	    gtt_map_size = 1024 * 1024; /* 1M on G33 */
+	gtt_map_size = intel_i915_get_gtt_size();
+
 	intel_private.gtt = ioremap(temp2, gtt_map_size);
 	if (!intel_private.gtt)
 		return -ENOMEM;
@@ -1422,7 +1454,7 @@ static const struct agp_bridge_driver intel_915_driver = {
 	.size_type		= FIXED_APER_SIZE,
 	.num_aperture_sizes	= 4,
 	.needs_scratch_page	= true,
-	.configure		= intel_i915_configure,
+	.configure		= intel_i9xx_configure,
 	.fetch_size		= intel_i9xx_fetch_size,
 	.cleanup		= intel_i915_cleanup,
 	.mask_memory		= intel_i810_mask_memory,
@@ -1455,7 +1487,7 @@ static const struct agp_bridge_driver intel_i965_driver = {
 	.size_type		= FIXED_APER_SIZE,
 	.num_aperture_sizes	= 4,
 	.needs_scratch_page	= true,
-	.configure		= intel_i915_configure,
+	.configure		= intel_i9xx_configure,
 	.fetch_size		= intel_i9xx_fetch_size,
 	.cleanup		= intel_i915_cleanup,
 	.mask_memory		= intel_i965_mask_memory,
@@ -1488,7 +1520,7 @@ static const struct agp_bridge_driver intel_g33_driver = {
 	.size_type		= FIXED_APER_SIZE,
 	.num_aperture_sizes	= 4,
 	.needs_scratch_page	= true,
-	.configure		= intel_i915_configure,
+	.configure		= intel_i9xx_configure,
 	.fetch_size		= intel_i9xx_fetch_size,
 	.cleanup		= intel_i915_cleanup,
 	.mask_memory		= intel_i965_mask_memory,
diff --git a/drivers/char/agp/uninorth-agp.c b/drivers/char/agp/uninorth-agp.c
index 95db71360d24..f845a8f718b3 100644
--- a/drivers/char/agp/uninorth-agp.c
+++ b/drivers/char/agp/uninorth-agp.c
@@ -415,7 +415,7 @@ static int uninorth_create_gatt_table(struct agp_bridge_data *bridge)
 	bridge->gatt_table_real = (u32 *) table;
 	/* Need to clear out any dirty data still sitting in caches */
 	flush_dcache_range((unsigned long)table,
-			   (unsigned long)(table_end + PAGE_SIZE));
+			   (unsigned long)table_end + 1);
 	bridge->gatt_table = vmap(pages, (1 << page_order), 0, PAGE_KERNEL_NCG);
 
 	if (bridge->gatt_table == NULL)
diff --git a/drivers/char/amiserial.c b/drivers/char/amiserial.c
index 56b27671adc4..4f8d60c25a98 100644
--- a/drivers/char/amiserial.c
+++ b/drivers/char/amiserial.c
@@ -84,6 +84,7 @@ static char *serial_version = "4.30";
 #include <linux/smp_lock.h>
 #include <linux/init.h>
 #include <linux/bitops.h>
+#include <linux/platform_device.h>
 
 #include <asm/setup.h>
 
@@ -1954,29 +1955,16 @@ static const struct tty_operations serial_ops = {
 /*
  * The serial driver boot-time initialization code!
  */
-static int __init rs_init(void)
+static int __init amiga_serial_probe(struct platform_device *pdev)
 {
 	unsigned long flags;
 	struct serial_state * state;
 	int error;
 
-	if (!MACH_IS_AMIGA || !AMIGAHW_PRESENT(AMI_SERIAL))
-		return -ENODEV;
-
 	serial_driver = alloc_tty_driver(1);
 	if (!serial_driver)
 		return -ENOMEM;
 
-	/*
-	 *  We request SERDAT and SERPER only, because the serial registers are
-	 *  too spreaded over the custom register space
-	 */
-	if (!request_mem_region(CUSTOM_PHYSADDR+0x30, 4,
-				"amiserial [Paula]")) {
-		error = -EBUSY;
-		goto fail_put_tty_driver;
-	}
-
 	IRQ_ports = NULL;
 
 	show_serial_version();
@@ -1998,7 +1986,7 @@ static int __init rs_init(void)
 
 	error = tty_register_driver(serial_driver);
 	if (error)
-		goto fail_release_mem_region;
+		goto fail_put_tty_driver;
 
 	state = rs_table;
 	state->magic = SSTATE_MAGIC;
@@ -2050,23 +2038,24 @@ static int __init rs_init(void)
 	ciab.ddra |= (SER_DTR | SER_RTS);   /* outputs */
 	ciab.ddra &= ~(SER_DCD | SER_CTS | SER_DSR);  /* inputs */
 
+	platform_set_drvdata(pdev, state);
+
 	return 0;
 
 fail_free_irq:
 	free_irq(IRQ_AMIGA_TBE, state);
 fail_unregister:
 	tty_unregister_driver(serial_driver);
-fail_release_mem_region:
-	release_mem_region(CUSTOM_PHYSADDR+0x30, 4);
 fail_put_tty_driver:
 	put_tty_driver(serial_driver);
 	return error;
 }
 
-static __exit void rs_exit(void) 
+static int __exit amiga_serial_remove(struct platform_device *pdev)
 {
 	int error;
-	struct async_struct *info = rs_table[0].info;
+	struct serial_state *state = platform_get_drvdata(pdev);
+	struct async_struct *info = state->info;
 
 	/* printk("Unloading %s: version %s\n", serial_name, serial_version); */
 	tasklet_kill(&info->tlet);
@@ -2075,19 +2064,38 @@ static __exit void rs_exit(void)
 		       error);
 	put_tty_driver(serial_driver);
 
-	if (info) {
-	  rs_table[0].info = NULL;
-	  kfree(info);
-	}
+	rs_table[0].info = NULL;
+	kfree(info);
 
 	free_irq(IRQ_AMIGA_TBE, rs_table);
 	free_irq(IRQ_AMIGA_RBF, rs_table);
 
-	release_mem_region(CUSTOM_PHYSADDR+0x30, 4);
+	platform_set_drvdata(pdev, NULL);
+
+	return error;
+}
+
+static struct platform_driver amiga_serial_driver = {
+	.remove = __exit_p(amiga_serial_remove),
+	.driver   = {
+		.name	= "amiga-serial",
+		.owner	= THIS_MODULE,
+	},
+};
+
+static int __init amiga_serial_init(void)
+{
+	return platform_driver_probe(&amiga_serial_driver, amiga_serial_probe);
+}
+
+module_init(amiga_serial_init);
+
+static void __exit amiga_serial_exit(void)
+{
+	platform_driver_unregister(&amiga_serial_driver);
 }
 
-module_init(rs_init)
-module_exit(rs_exit)
+module_exit(amiga_serial_exit);
 
 
 #if defined(CONFIG_SERIAL_CONSOLE) && !defined(MODULE)
@@ -2154,3 +2162,4 @@ console_initcall(amiserial_console_init);
 #endif /* CONFIG_SERIAL_CONSOLE && !MODULE */
 
 MODULE_LICENSE("GPL");
+MODULE_ALIAS("platform:amiga-serial");
diff --git a/drivers/char/apm-emulation.c b/drivers/char/apm-emulation.c
index 4f568cb9af3f..033e1505fca9 100644
--- a/drivers/char/apm-emulation.c
+++ b/drivers/char/apm-emulation.c
@@ -265,8 +265,8 @@ static unsigned int apm_poll(struct file *fp, poll_table * wait)
  *   Only when everyone who has opened /dev/apm_bios with write permission
  *   has acknowledge does the actual suspend happen.
  */
-static int
-apm_ioctl(struct inode * inode, struct file *filp, u_int cmd, u_long arg)
+static long
+apm_ioctl(struct file *filp, u_int cmd, u_long arg)
 {
 	struct apm_user *as = filp->private_data;
 	int err = -EINVAL;
@@ -274,6 +274,7 @@ apm_ioctl(struct inode * inode, struct file *filp, u_int cmd, u_long arg)
 	if (!as->suser || !as->writer)
 		return -EPERM;
 
+	lock_kernel();
 	switch (cmd) {
 	case APM_IOC_SUSPEND:
 		mutex_lock(&state_lock);
@@ -334,6 +335,7 @@ apm_ioctl(struct inode * inode, struct file *filp, u_int cmd, u_long arg)
 		mutex_unlock(&state_lock);
 		break;
 	}
+	unlock_kernel();
 
 	return err;
 }
@@ -397,7 +399,7 @@ static const struct file_operations apm_bios_fops = {
 	.owner		= THIS_MODULE,
 	.read		= apm_read,
 	.poll		= apm_poll,
-	.ioctl		= apm_ioctl,
+	.unlocked_ioctl	= apm_ioctl,
 	.open		= apm_open,
 	.release	= apm_release,
 };
diff --git a/drivers/char/applicom.c b/drivers/char/applicom.c
index a7424bf7eacf..f4ae0e0fb631 100644
--- a/drivers/char/applicom.c
+++ b/drivers/char/applicom.c
@@ -26,6 +26,7 @@
 #include <linux/sched.h>
 #include <linux/slab.h>
 #include <linux/errno.h>
+#include <linux/smp_lock.h>
 #include <linux/miscdevice.h>
 #include <linux/pci.h>
 #include <linux/wait.h>
@@ -106,8 +107,7 @@ static unsigned int DeviceErrorCount;	/* number of device error     */
 
 static ssize_t ac_read (struct file *, char __user *, size_t, loff_t *);
 static ssize_t ac_write (struct file *, const char __user *, size_t, loff_t *);
-static int ac_ioctl(struct inode *, struct file *, unsigned int,
-		    unsigned long);
+static long ac_ioctl(struct file *, unsigned int, unsigned long);
 static irqreturn_t ac_interrupt(int, void *);
 
 static const struct file_operations ac_fops = {
@@ -115,7 +115,7 @@ static const struct file_operations ac_fops = {
 	.llseek = no_llseek,
 	.read = ac_read,
 	.write = ac_write,
-	.ioctl = ac_ioctl,
+	.unlocked_ioctl = ac_ioctl,
 };
 
 static struct miscdevice ac_miscdev = {
@@ -689,7 +689,7 @@ static irqreturn_t ac_interrupt(int vec, void *dev_instance)
 
 
 
-static int ac_ioctl(struct inode *inode, struct file *file, unsigned int cmd, unsigned long arg)
+static long ac_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
      
 {				/* @ ADG ou ATO selon le cas */
 	int i;
@@ -703,15 +703,11 @@ static int ac_ioctl(struct inode *inode, struct file *file, unsigned int cmd, un
 	/* In general, the device is only openable by root anyway, so we're not
 	   particularly concerned that bogus ioctls can flood the console. */
 
-	adgl = kmalloc(sizeof(struct st_ram_io), GFP_KERNEL);
-	if (!adgl)
-		return -ENOMEM;
+	adgl = memdup_user(argp, sizeof(struct st_ram_io));
+	if (IS_ERR(adgl))
+		return PTR_ERR(adgl);
 
-	if (copy_from_user(adgl, argp, sizeof(struct st_ram_io))) {
-		kfree(adgl);
-		return -EFAULT;
-	}
-	
+	lock_kernel();	
 	IndexCard = adgl->num_card-1;
 	 
 	if(cmd != 6 && ((IndexCard >= MAX_BOARD) || !apbs[IndexCard].RamIO)) {
@@ -721,6 +717,7 @@ static int ac_ioctl(struct inode *inode, struct file *file, unsigned int cmd, un
 			warncount--;
 		}
 		kfree(adgl);
+		unlock_kernel();
 		return -EINVAL;
 	}
 
@@ -838,6 +835,7 @@ static int ac_ioctl(struct inode *inode, struct file *file, unsigned int cmd, un
 	}
 	Dummy = readb(apbs[IndexCard].RamIO + VERS);
 	kfree(adgl);
+	unlock_kernel();
 	return 0;
 }
 
diff --git a/drivers/char/ds1620.c b/drivers/char/ds1620.c
index 61f0146e215d..dbee8688f75c 100644
--- a/drivers/char/ds1620.c
+++ b/drivers/char/ds1620.c
@@ -232,7 +232,7 @@ ds1620_read(struct file *file, char __user *buf, size_t count, loff_t *ptr)
 }
 
 static int
-ds1620_ioctl(struct inode *inode, struct file *file, unsigned int cmd, unsigned long arg)
+ds1620_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
 {
 	struct therm therm;
 	union {
@@ -316,6 +316,18 @@ ds1620_ioctl(struct inode *inode, struct file *file, unsigned int cmd, unsigned
 	return 0;
 }
 
+static long
+ds1620_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
+{
+	int ret;
+
+	lock_kernel();
+	ret = ds1620_ioctl(file, cmd, arg);
+	unlock_kernel();
+
+	return ret;
+}
+
 #ifdef THERM_USE_PROC
 static int
 proc_therm_ds1620_read(char *buf, char **start, off_t offset,
@@ -344,7 +356,7 @@ static const struct file_operations ds1620_fops = {
 	.owner		= THIS_MODULE,
 	.open		= ds1620_open,
 	.read		= ds1620_read,
-	.ioctl		= ds1620_ioctl,
+	.unlocked_ioctl	= ds1620_unlocked_ioctl,
 };
 
 static struct miscdevice ds1620_miscdev = {
diff --git a/drivers/char/dtlk.c b/drivers/char/dtlk.c
index 045c930e6320..e3859d4eaead 100644
--- a/drivers/char/dtlk.c
+++ b/drivers/char/dtlk.c
@@ -93,8 +93,8 @@ static ssize_t dtlk_write(struct file *, const char __user *,
 static unsigned int dtlk_poll(struct file *, poll_table *);
 static int dtlk_open(struct inode *, struct file *);
 static int dtlk_release(struct inode *, struct file *);
-static int dtlk_ioctl(struct inode *inode, struct file *file,
-		      unsigned int cmd, unsigned long arg);
+static long dtlk_ioctl(struct file *file,
+		       unsigned int cmd, unsigned long arg);
 
 static const struct file_operations dtlk_fops =
 {
@@ -102,7 +102,7 @@ static const struct file_operations dtlk_fops =
 	.read		= dtlk_read,
 	.write		= dtlk_write,
 	.poll		= dtlk_poll,
-	.ioctl		= dtlk_ioctl,
+	.unlocked_ioctl	= dtlk_ioctl,
 	.open		= dtlk_open,
 	.release	= dtlk_release,
 };
@@ -263,10 +263,9 @@ static void dtlk_timer_tick(unsigned long data)
 	wake_up_interruptible(&dtlk_process_list);
 }
 
-static int dtlk_ioctl(struct inode *inode,
-		      struct file *file,
-		      unsigned int cmd,
-		      unsigned long arg)
+static long dtlk_ioctl(struct file *file,
+		       unsigned int cmd,
+		       unsigned long arg)
 {
 	char __user *argp = (char __user *)arg;
 	struct dtlk_settings *sp;
@@ -276,7 +275,9 @@ static int dtlk_ioctl(struct inode *inode,
 	switch (cmd) {
 
 	case DTLK_INTERROGATE:
+		lock_kernel();
 		sp = dtlk_interrogate();
+		unlock_kernel();
 		if (copy_to_user(argp, sp, sizeof(struct dtlk_settings)))
 			return -EINVAL;
 		return 0;
diff --git a/drivers/char/generic_nvram.c b/drivers/char/generic_nvram.c
index fda4181b5e67..82b5a88a82d7 100644
--- a/drivers/char/generic_nvram.c
+++ b/drivers/char/generic_nvram.c
@@ -19,6 +19,7 @@
 #include <linux/miscdevice.h>
 #include <linux/fcntl.h>
 #include <linux/init.h>
+#include <linux/smp_lock.h>
 #include <asm/uaccess.h>
 #include <asm/nvram.h>
 #ifdef CONFIG_PPC_PMAC
@@ -84,8 +85,7 @@ static ssize_t write_nvram(struct file *file, const char __user *buf,
 	return p - buf;
 }
 
-static int nvram_ioctl(struct inode *inode, struct file *file,
-	unsigned int cmd, unsigned long arg)
+static int nvram_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
 {
 	switch(cmd) {
 #ifdef CONFIG_PPC_PMAC
@@ -116,12 +116,23 @@ static int nvram_ioctl(struct inode *inode, struct file *file,
 	return 0;
 }
 
+static long nvram_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
+{
+	int ret;
+
+	lock_kernel();
+	ret = nvram_ioctl(file, cmd, arg);
+	unlock_kernel();
+
+	return ret;
+}
+
 const struct file_operations nvram_fops = {
 	.owner		= THIS_MODULE,
 	.llseek		= nvram_llseek,
 	.read		= read_nvram,
 	.write		= write_nvram,
-	.ioctl		= nvram_ioctl,
+	.unlocked_ioctl	= nvram_unlocked_ioctl,
 };
 
 static struct miscdevice nvram_dev = {
diff --git a/drivers/char/genrtc.c b/drivers/char/genrtc.c
index 31e7c91c2d9d..b6c2cc167c11 100644
--- a/drivers/char/genrtc.c
+++ b/drivers/char/genrtc.c
@@ -262,7 +262,7 @@ static inline int gen_set_rtc_irq_bit(unsigned char bit)
 #endif
 }
 
-static int gen_rtc_ioctl(struct inode *inode, struct file *file,
+static int gen_rtc_ioctl(struct file *file,
 			 unsigned int cmd, unsigned long arg)
 {
 	struct rtc_time wtime;
@@ -332,6 +332,18 @@ static int gen_rtc_ioctl(struct inode *inode, struct file *file,
 	return -EINVAL;
 }
 
+static long gen_rtc_unlocked_ioctl(struct file *file, unsigned int cmd,
+				   unsigned long arg)
+{
+	int ret;
+
+	lock_kernel();
+	ret = gen_rtc_ioctl(file, cmd, arg);
+	unlock_kernel();
+
+	return ret;
+}
+
 /*
  *	We enforce only one user at a time here with the open/close.
  *	Also clear the previous interrupt data on an open, and clean
@@ -482,7 +494,7 @@ static const struct file_operations gen_rtc_fops = {
 	.read		= gen_rtc_read,
 	.poll		= gen_rtc_poll,
 #endif
-	.ioctl		= gen_rtc_ioctl,
+	.unlocked_ioctl	= gen_rtc_unlocked_ioctl,
 	.open		= gen_rtc_open,
 	.release	= gen_rtc_release,
 };
diff --git a/drivers/char/hangcheck-timer.c b/drivers/char/hangcheck-timer.c
index 712d9f271aa6..e0249722d25f 100644
--- a/drivers/char/hangcheck-timer.c
+++ b/drivers/char/hangcheck-timer.c
@@ -49,8 +49,9 @@
 #include <asm/uaccess.h>
 #include <linux/sysrq.h>
 #include <linux/timer.h>
+#include <linux/time.h>
 
-#define VERSION_STR "0.9.0"
+#define VERSION_STR "0.9.1"
 
 #define DEFAULT_IOFENCE_MARGIN 60	/* Default fudge factor, in seconds */
 #define DEFAULT_IOFENCE_TICK 180	/* Default timer timeout, in seconds */
@@ -119,10 +120,8 @@ __setup("hcheck_dump_tasks", hangcheck_parse_dump_tasks);
 #if defined(CONFIG_S390)
 # define HAVE_MONOTONIC
 # define TIMER_FREQ 1000000000ULL
-#elif defined(CONFIG_IA64)
-# define TIMER_FREQ ((unsigned long long)local_cpu_data->itc_freq)
 #else
-# define TIMER_FREQ (HZ*loops_per_jiffy)
+# define TIMER_FREQ 1000000000ULL
 #endif
 
 #ifdef HAVE_MONOTONIC
@@ -130,7 +129,9 @@ extern unsigned long long monotonic_clock(void);
 #else
 static inline unsigned long long monotonic_clock(void)
 {
-	return get_cycles();
+	struct timespec ts;
+	getrawmonotonic(&ts);
+	return timespec_to_ns(&ts);
 }
 #endif  /* HAVE_MONOTONIC */
 
@@ -168,6 +169,13 @@ static void hangcheck_fire(unsigned long data)
 			printk(KERN_CRIT "Hangcheck: hangcheck value past margin!\n");
 		}
 	}
+#if 0
+	/*
+	 * Enable to investigate delays in detail
+	 */
+	printk("Hangcheck: called %Ld ns since last time (%Ld ns overshoot)\n",
+			tsc_diff, tsc_diff - hangcheck_tick*TIMER_FREQ);
+#endif
 	mod_timer(&hangcheck_ticktock, jiffies + (hangcheck_tick*HZ));
 	hangcheck_tsc = monotonic_clock();
 }
@@ -180,7 +188,7 @@ static int __init hangcheck_init(void)
 #if defined (HAVE_MONOTONIC)
 	printk("Hangcheck: Using monotonic_clock().\n");
 #else
-	printk("Hangcheck: Using get_cycles().\n");
+	printk("Hangcheck: Using getrawmonotonic().\n");
 #endif  /* HAVE_MONOTONIC */
 	hangcheck_tsc_margin =
 		(unsigned long long)(hangcheck_margin + hangcheck_tick);
diff --git a/drivers/char/hpet.c b/drivers/char/hpet.c
index 9ded667625ac..a0a1829d3198 100644
--- a/drivers/char/hpet.c
+++ b/drivers/char/hpet.c
@@ -431,14 +431,18 @@ static int hpet_release(struct inode *inode, struct file *file)
 
 static int hpet_ioctl_common(struct hpet_dev *, int, unsigned long, int);
 
-static int
-hpet_ioctl(struct inode *inode, struct file *file, unsigned int cmd,
-	   unsigned long arg)
+static long hpet_ioctl(struct file *file, unsigned int cmd,
+			unsigned long arg)
 {
 	struct hpet_dev *devp;
+	int ret;
 
 	devp = file->private_data;
-	return hpet_ioctl_common(devp, cmd, arg, 0);
+	lock_kernel();
+	ret = hpet_ioctl_common(devp, cmd, arg, 0);
+	unlock_kernel();
+
+	return ret;
 }
 
 static int hpet_ioctl_ieon(struct hpet_dev *devp)
@@ -654,7 +658,7 @@ static const struct file_operations hpet_fops = {
 	.llseek = no_llseek,
 	.read = hpet_read,
 	.poll = hpet_poll,
-	.ioctl = hpet_ioctl,
+	.unlocked_ioctl = hpet_ioctl,
 	.open = hpet_open,
 	.release = hpet_release,
 	.fasync = hpet_fasync,
diff --git a/drivers/char/hvsi.c b/drivers/char/hvsi.c
index 793b236c9266..d4b14ff1c4c1 100644
--- a/drivers/char/hvsi.c
+++ b/drivers/char/hvsi.c
@@ -194,10 +194,8 @@ static inline void print_state(struct hvsi_struct *hp)
 		"HVSI_WAIT_FOR_MCTRL_RESPONSE",
 		"HVSI_FSP_DIED",
 	};
-	const char *name = state_names[hp->state];
-
-	if (hp->state > ARRAY_SIZE(state_names))
-		name = "UNKNOWN";
+	const char *name = (hp->state < ARRAY_SIZE(state_names))
+		? state_names[hp->state] : "UNKNOWN";
 
 	pr_debug("hvsi%i: state = %s\n", hp->index, name);
 #endif /* DEBUG */
diff --git a/drivers/char/hw_random/n2-drv.c b/drivers/char/hw_random/n2-drv.c
index 10f868eefaa6..0f9cbf1aaf15 100644
--- a/drivers/char/hw_random/n2-drv.c
+++ b/drivers/char/hw_random/n2-drv.c
@@ -660,7 +660,7 @@ static int __devinit n2rng_probe(struct of_device *op,
 				np->hvapi_major);
 			goto out_hvapi_unregister;
 		}
-		np->num_units = of_getintprop_default(op->node,
+		np->num_units = of_getintprop_default(op->dev.of_node,
 						      "rng-#units", 0);
 		if (!np->num_units) {
 			dev_err(&op->dev, "VF RNG lacks rng-#units property\n");
@@ -751,8 +751,11 @@ static const struct of_device_id n2rng_match[] = {
 MODULE_DEVICE_TABLE(of, n2rng_match);
 
 static struct of_platform_driver n2rng_driver = {
-	.name		= "n2rng",
-	.match_table	= n2rng_match,
+	.driver = {
+		.name = "n2rng",
+		.owner = THIS_MODULE,
+		.of_match_table = n2rng_match,
+	},
 	.probe		= n2rng_probe,
 	.remove		= __devexit_p(n2rng_remove),
 };
diff --git a/drivers/char/hw_random/nomadik-rng.c b/drivers/char/hw_random/nomadik-rng.c
index a8b4c4010144..a348c7e9aa0b 100644
--- a/drivers/char/hw_random/nomadik-rng.c
+++ b/drivers/char/hw_random/nomadik-rng.c
@@ -15,6 +15,10 @@
 #include <linux/amba/bus.h>
 #include <linux/hw_random.h>
 #include <linux/io.h>
+#include <linux/clk.h>
+#include <linux/err.h>
+
+static struct clk *rng_clk;
 
 static int nmk_rng_read(struct hwrng *rng, void *data, size_t max, bool wait)
 {
@@ -40,6 +44,15 @@ static int nmk_rng_probe(struct amba_device *dev, struct amba_id *id)
 	void __iomem *base;
 	int ret;
 
+	rng_clk = clk_get(&dev->dev, NULL);
+	if (IS_ERR(rng_clk)) {
+		dev_err(&dev->dev, "could not get rng clock\n");
+		ret = PTR_ERR(rng_clk);
+		return ret;
+	}
+
+	clk_enable(rng_clk);
+
 	ret = amba_request_regions(dev, dev->dev.init_name);
 	if (ret)
 		return ret;
@@ -57,6 +70,8 @@ out_unmap:
 	iounmap(base);
 out_release:
 	amba_release_regions(dev);
+	clk_disable(rng_clk);
+	clk_put(rng_clk);
 	return ret;
 }
 
@@ -66,6 +81,8 @@ static int nmk_rng_remove(struct amba_device *dev)
 	hwrng_unregister(&nmk_rng);
 	iounmap(base);
 	amba_release_regions(dev);
+	clk_disable(rng_clk);
+	clk_put(rng_clk);
 	return 0;
 }
 
diff --git a/drivers/char/hw_random/pasemi-rng.c b/drivers/char/hw_random/pasemi-rng.c
index 7fa61dd1d9d9..261ba8f22b8b 100644
--- a/drivers/char/hw_random/pasemi-rng.c
+++ b/drivers/char/hw_random/pasemi-rng.c
@@ -98,7 +98,7 @@ static int __devinit rng_probe(struct of_device *ofdev,
 			       const struct of_device_id *match)
 {
 	void __iomem *rng_regs;
-	struct device_node *rng_np = ofdev->node;
+	struct device_node *rng_np = ofdev->dev.of_node;
 	struct resource res;
 	int err = 0;
 
@@ -140,8 +140,11 @@ static struct of_device_id rng_match[] = {
 };
 
 static struct of_platform_driver rng_driver = {
-	.name		= "pasemi-rng",
-	.match_table	= rng_match,
+	.driver = {
+		.name = "pasemi-rng",
+		.owner = THIS_MODULE,
+		.of_match_table = rng_match,
+	},
 	.probe		= rng_probe,
 	.remove		= rng_remove,
 };
diff --git a/drivers/char/hw_random/virtio-rng.c b/drivers/char/hw_random/virtio-rng.c
index 64fe0a793efd..75f1cbd61c17 100644
--- a/drivers/char/hw_random/virtio-rng.c
+++ b/drivers/char/hw_random/virtio-rng.c
@@ -32,7 +32,7 @@ static bool busy;
 static void random_recv_done(struct virtqueue *vq)
 {
 	/* We can get spurious callbacks, e.g. shared IRQs + virtio_pci. */
-	if (!vq->vq_ops->get_buf(vq, &data_avail))
+	if (!virtqueue_get_buf(vq, &data_avail))
 		return;
 
 	complete(&have_data);
@@ -46,10 +46,10 @@ static void register_buffer(u8 *buf, size_t size)
 	sg_init_one(&sg, buf, size);
 
 	/* There should always be room for one buffer. */
-	if (vq->vq_ops->add_buf(vq, &sg, 0, 1, buf) < 0)
+	if (virtqueue_add_buf(vq, &sg, 0, 1, buf) < 0)
 		BUG();
 
-	vq->vq_ops->kick(vq);
+	virtqueue_kick(vq);
 }
 
 static int virtio_read(struct hwrng *rng, void *buf, size_t size, bool wait)
diff --git a/drivers/char/ipmi/ipmi_devintf.c b/drivers/char/ipmi/ipmi_devintf.c
index 65545de3dbf4..d8ec92a38980 100644
--- a/drivers/char/ipmi/ipmi_devintf.c
+++ b/drivers/char/ipmi/ipmi_devintf.c
@@ -228,8 +228,7 @@ static int handle_send_req(ipmi_user_t     user,
 	return rv;
 }
 
-static int ipmi_ioctl(struct inode  *inode,
-		      struct file   *file,
+static int ipmi_ioctl(struct file   *file,
 		      unsigned int  cmd,
 		      unsigned long data)
 {
@@ -630,6 +629,23 @@ static int ipmi_ioctl(struct inode  *inode,
 	return rv;
 }
 
+/*
+ * Note: it doesn't make sense to take the BKL here but
+ *       not in compat_ipmi_ioctl. -arnd
+ */
+static long ipmi_unlocked_ioctl(struct file   *file,
+			        unsigned int  cmd,
+			        unsigned long data)
+{
+	int ret;
+
+	lock_kernel();
+	ret = ipmi_ioctl(file, cmd, data);
+	unlock_kernel();
+
+	return ret;
+}
+
 #ifdef CONFIG_COMPAT
 
 /*
@@ -802,7 +818,7 @@ static long compat_ipmi_ioctl(struct file *filep, unsigned int cmd,
 		if (copy_to_user(precv64, &recv64, sizeof(recv64)))
 			return -EFAULT;
 
-		rc = ipmi_ioctl(filep->f_path.dentry->d_inode, filep,
+		rc = ipmi_ioctl(filep,
 				((cmd == COMPAT_IPMICTL_RECEIVE_MSG)
 				 ? IPMICTL_RECEIVE_MSG
 				 : IPMICTL_RECEIVE_MSG_TRUNC),
@@ -819,14 +835,14 @@ static long compat_ipmi_ioctl(struct file *filep, unsigned int cmd,
 		return rc;
 	}
 	default:
-		return ipmi_ioctl(filep->f_path.dentry->d_inode, filep, cmd, arg);
+		return ipmi_ioctl(filep, cmd, arg);
 	}
 }
 #endif
 
 static const struct file_operations ipmi_fops = {
 	.owner		= THIS_MODULE,
-	.ioctl		= ipmi_ioctl,
+	.unlocked_ioctl	= ipmi_unlocked_ioctl,
 #ifdef CONFIG_COMPAT
 	.compat_ioctl   = compat_ipmi_ioctl,
 #endif
diff --git a/drivers/char/ipmi/ipmi_msghandler.c b/drivers/char/ipmi/ipmi_msghandler.c
index c6ad4234378d..4f3f8c9ec262 100644
--- a/drivers/char/ipmi/ipmi_msghandler.c
+++ b/drivers/char/ipmi/ipmi_msghandler.c
@@ -2505,12 +2505,11 @@ static int ipmi_bmc_register(ipmi_smi_t intf, int ifnum,
 			return rv;
 		}
 
-		printk(KERN_INFO
-		       "ipmi: Found new BMC (man_id: 0x%6.6x, "
-		       " prod_id: 0x%4.4x, dev_id: 0x%2.2x)\n",
-		       bmc->id.manufacturer_id,
-		       bmc->id.product_id,
-		       bmc->id.device_id);
+		dev_info(intf->si_dev, "Found new BMC (man_id: 0x%6.6x, "
+			 "prod_id: 0x%4.4x, dev_id: 0x%2.2x)\n",
+			 bmc->id.manufacturer_id,
+			 bmc->id.product_id,
+			 bmc->id.device_id);
 	}
 
 	/*
@@ -4037,8 +4036,8 @@ static void ipmi_request_event(void)
 
 static struct timer_list ipmi_timer;
 
-/* Call every ~100 ms. */
-#define IPMI_TIMEOUT_TIME	100
+/* Call every ~1000 ms. */
+#define IPMI_TIMEOUT_TIME	1000
 
 /* How many jiffies does it take to get to the timeout time. */
 #define IPMI_TIMEOUT_JIFFIES	((IPMI_TIMEOUT_TIME * HZ) / 1000)
diff --git a/drivers/char/ipmi/ipmi_si_intf.c b/drivers/char/ipmi/ipmi_si_intf.c
index 4462b113ba3f..35603dd4e6c5 100644
--- a/drivers/char/ipmi/ipmi_si_intf.c
+++ b/drivers/char/ipmi/ipmi_si_intf.c
@@ -107,6 +107,14 @@ enum si_type {
 };
 static char *si_to_str[] = { "kcs", "smic", "bt" };
 
+enum ipmi_addr_src {
+	SI_INVALID = 0, SI_HOTMOD, SI_HARDCODED, SI_SPMI, SI_ACPI, SI_SMBIOS,
+	SI_PCI,	SI_DEVICETREE, SI_DEFAULT
+};
+static char *ipmi_addr_src_to_str[] = { NULL, "hotmod", "hardcoded", "SPMI",
+					"ACPI", "SMBIOS", "PCI",
+					"device-tree", "default" };
+
 #define DEVICE_NAME "ipmi_si"
 
 static struct platform_driver ipmi_driver = {
@@ -188,7 +196,7 @@ struct smi_info {
 	int (*irq_setup)(struct smi_info *info);
 	void (*irq_cleanup)(struct smi_info *info);
 	unsigned int io_size;
-	char *addr_source; /* ACPI, PCI, SMBIOS, hardcode, default. */
+	enum ipmi_addr_src addr_source; /* ACPI, PCI, SMBIOS, hardcode, etc. */
 	void (*addr_source_cleanup)(struct smi_info *info);
 	void *addr_source_data;
 
@@ -300,6 +308,7 @@ static int num_max_busy_us;
 
 static int unload_when_empty = 1;
 
+static int add_smi(struct smi_info *smi);
 static int try_smi_init(struct smi_info *smi);
 static void cleanup_one_si(struct smi_info *to_clean);
 
@@ -314,9 +323,14 @@ static void deliver_recv_msg(struct smi_info *smi_info,
 {
 	/* Deliver the message to the upper layer with the lock
 	   released. */
-	spin_unlock(&(smi_info->si_lock));
-	ipmi_smi_msg_received(smi_info->intf, msg);
-	spin_lock(&(smi_info->si_lock));
+
+	if (smi_info->run_to_completion) {
+		ipmi_smi_msg_received(smi_info->intf, msg);
+	} else {
+		spin_unlock(&(smi_info->si_lock));
+		ipmi_smi_msg_received(smi_info->intf, msg);
+		spin_lock(&(smi_info->si_lock));
+	}
 }
 
 static void return_hosed_msg(struct smi_info *smi_info, int cCode)
@@ -445,6 +459,9 @@ static inline void disable_si_irq(struct smi_info *smi_info)
 	if ((smi_info->irq) && (!smi_info->interrupt_disabled)) {
 		start_disable_irq(smi_info);
 		smi_info->interrupt_disabled = 1;
+		if (!atomic_read(&smi_info->stop_operation))
+			mod_timer(&smi_info->si_timer,
+				  jiffies + SI_TIMEOUT_JIFFIES);
 	}
 }
 
@@ -576,9 +593,8 @@ static void handle_transaction_done(struct smi_info *smi_info)
 		smi_info->handlers->get_result(smi_info->si_sm, msg, 3);
 		if (msg[2] != 0) {
 			/* Error clearing flags */
-			printk(KERN_WARNING
-			       "ipmi_si: Error clearing flags: %2.2x\n",
-			       msg[2]);
+			dev_warn(smi_info->dev,
+				 "Error clearing flags: %2.2x\n", msg[2]);
 		}
 		if (smi_info->si_state == SI_CLEARING_FLAGS_THEN_SET_IRQ)
 			start_enable_irq(smi_info);
@@ -670,9 +686,8 @@ static void handle_transaction_done(struct smi_info *smi_info)
 		/* We got the flags from the SMI, now handle them. */
 		smi_info->handlers->get_result(smi_info->si_sm, msg, 4);
 		if (msg[2] != 0) {
-			printk(KERN_WARNING
-			       "ipmi_si: Could not enable interrupts"
-			       ", failed get, using polled mode.\n");
+			dev_warn(smi_info->dev, "Could not enable interrupts"
+				 ", failed get, using polled mode.\n");
 			smi_info->si_state = SI_NORMAL;
 		} else {
 			msg[0] = (IPMI_NETFN_APP_REQUEST << 2);
@@ -693,11 +708,11 @@ static void handle_transaction_done(struct smi_info *smi_info)
 
 		/* We got the flags from the SMI, now handle them. */
 		smi_info->handlers->get_result(smi_info->si_sm, msg, 4);
-		if (msg[2] != 0) {
-			printk(KERN_WARNING
-			       "ipmi_si: Could not enable interrupts"
-			       ", failed set, using polled mode.\n");
-		}
+		if (msg[2] != 0)
+			dev_warn(smi_info->dev, "Could not enable interrupts"
+				 ", failed set, using polled mode.\n");
+		else
+			smi_info->interrupt_disabled = 0;
 		smi_info->si_state = SI_NORMAL;
 		break;
 	}
@@ -709,9 +724,8 @@ static void handle_transaction_done(struct smi_info *smi_info)
 		/* We got the flags from the SMI, now handle them. */
 		smi_info->handlers->get_result(smi_info->si_sm, msg, 4);
 		if (msg[2] != 0) {
-			printk(KERN_WARNING
-			       "ipmi_si: Could not disable interrupts"
-			       ", failed get.\n");
+			dev_warn(smi_info->dev, "Could not disable interrupts"
+				 ", failed get.\n");
 			smi_info->si_state = SI_NORMAL;
 		} else {
 			msg[0] = (IPMI_NETFN_APP_REQUEST << 2);
@@ -733,9 +747,8 @@ static void handle_transaction_done(struct smi_info *smi_info)
 		/* We got the flags from the SMI, now handle them. */
 		smi_info->handlers->get_result(smi_info->si_sm, msg, 4);
 		if (msg[2] != 0) {
-			printk(KERN_WARNING
-			       "ipmi_si: Could not disable interrupts"
-			       ", failed set.\n");
+			dev_warn(smi_info->dev, "Could not disable interrupts"
+				 ", failed set.\n");
 		}
 		smi_info->si_state = SI_NORMAL;
 		break;
@@ -877,6 +890,11 @@ static void sender(void                *send_info,
 	printk("**Enqueue: %d.%9.9d\n", t.tv_sec, t.tv_usec);
 #endif
 
+	mod_timer(&smi_info->si_timer, jiffies + SI_TIMEOUT_JIFFIES);
+
+	if (smi_info->thread)
+		wake_up_process(smi_info->thread);
+
 	if (smi_info->run_to_completion) {
 		/*
 		 * If we are running to completion, then throw it in
@@ -997,6 +1015,8 @@ static int ipmi_thread(void *data)
 			; /* do nothing */
 		else if (smi_result == SI_SM_CALL_WITH_DELAY && busy_wait)
 			schedule();
+		else if (smi_result == SI_SM_IDLE)
+			schedule_timeout_interruptible(100);
 		else
 			schedule_timeout_interruptible(0);
 	}
@@ -1039,6 +1059,7 @@ static void smi_timeout(unsigned long data)
 	unsigned long     flags;
 	unsigned long     jiffies_now;
 	long              time_diff;
+	long		  timeout;
 #ifdef DEBUG_TIMING
 	struct timeval    t;
 #endif
@@ -1059,9 +1080,9 @@ static void smi_timeout(unsigned long data)
 
 	if ((smi_info->irq) && (!smi_info->interrupt_disabled)) {
 		/* Running with interrupts, only do long timeouts. */
-		smi_info->si_timer.expires = jiffies + SI_TIMEOUT_JIFFIES;
+		timeout = jiffies + SI_TIMEOUT_JIFFIES;
 		smi_inc_stat(smi_info, long_timeouts);
-		goto do_add_timer;
+		goto do_mod_timer;
 	}
 
 	/*
@@ -1070,14 +1091,15 @@ static void smi_timeout(unsigned long data)
 	 */
 	if (smi_result == SI_SM_CALL_WITH_DELAY) {
 		smi_inc_stat(smi_info, short_timeouts);
-		smi_info->si_timer.expires = jiffies + 1;
+		timeout = jiffies + 1;
 	} else {
 		smi_inc_stat(smi_info, long_timeouts);
-		smi_info->si_timer.expires = jiffies + SI_TIMEOUT_JIFFIES;
+		timeout = jiffies + SI_TIMEOUT_JIFFIES;
 	}
 
- do_add_timer:
-	add_timer(&(smi_info->si_timer));
+ do_mod_timer:
+	if (smi_result != SI_SM_IDLE)
+		mod_timer(&(smi_info->si_timer), timeout);
 }
 
 static irqreturn_t si_irq_handler(int irq, void *data)
@@ -1144,10 +1166,10 @@ static int smi_start_processing(void       *send_info,
 		new_smi->thread = kthread_run(ipmi_thread, new_smi,
 					      "kipmi%d", new_smi->intf_num);
 		if (IS_ERR(new_smi->thread)) {
-			printk(KERN_NOTICE "ipmi_si_intf: Could not start"
-			       " kernel thread due to error %ld, only using"
-			       " timers to drive the interface\n",
-			       PTR_ERR(new_smi->thread));
+			dev_notice(new_smi->dev, "Could not start"
+				   " kernel thread due to error %ld, only using"
+				   " timers to drive the interface\n",
+				   PTR_ERR(new_smi->thread));
 			new_smi->thread = NULL;
 		}
 	}
@@ -1308,14 +1330,13 @@ static int std_irq_setup(struct smi_info *info)
 				 DEVICE_NAME,
 				 info);
 	if (rv) {
-		printk(KERN_WARNING
-		       "ipmi_si: %s unable to claim interrupt %d,"
-		       " running polled\n",
-		       DEVICE_NAME, info->irq);
+		dev_warn(info->dev, "%s unable to claim interrupt %d,"
+			 " running polled\n",
+			 DEVICE_NAME, info->irq);
 		info->irq = 0;
 	} else {
 		info->irq_cleanup = std_irq_cleanup;
-		printk("  Using irq %d\n", info->irq);
+		dev_info(info->dev, "Using irq %d\n", info->irq);
 	}
 
 	return rv;
@@ -1406,8 +1427,8 @@ static int port_setup(struct smi_info *info)
 		info->io.outputb = port_outl;
 		break;
 	default:
-		printk(KERN_WARNING "ipmi_si: Invalid register size: %d\n",
-		       info->io.regsize);
+		dev_warn(info->dev, "Invalid register size: %d\n",
+			 info->io.regsize);
 		return -EINVAL;
 	}
 
@@ -1529,8 +1550,8 @@ static int mem_setup(struct smi_info *info)
 		break;
 #endif
 	default:
-		printk(KERN_WARNING "ipmi_si: Invalid register size: %d\n",
-		       info->io.regsize);
+		dev_warn(info->dev, "Invalid register size: %d\n",
+			 info->io.regsize);
 		return -EINVAL;
 	}
 
@@ -1755,7 +1776,7 @@ static int hotmod_handler(const char *val, struct kernel_param *kp)
 				goto out;
 			}
 
-			info->addr_source = "hotmod";
+			info->addr_source = SI_HOTMOD;
 			info->si_type = si_type;
 			info->io.addr_data = addr;
 			info->io.addr_type = addr_space;
@@ -1777,7 +1798,9 @@ static int hotmod_handler(const char *val, struct kernel_param *kp)
 				info->irq_setup = std_irq_setup;
 			info->slave_addr = ipmb;
 
-			try_smi_init(info);
+			if (!add_smi(info))
+				if (try_smi_init(info))
+					cleanup_one_si(info);
 		} else {
 			/* remove */
 			struct smi_info *e, *tmp_e;
@@ -1813,7 +1836,8 @@ static __devinit void hardcode_find_bmc(void)
 		if (!info)
 			return;
 
-		info->addr_source = "hardcoded";
+		info->addr_source = SI_HARDCODED;
+		printk(KERN_INFO PFX "probing via hardcoded address\n");
 
 		if (!si_type[i] || strcmp(si_type[i], "kcs") == 0) {
 			info->si_type = SI_KCS;
@@ -1822,8 +1846,7 @@ static __devinit void hardcode_find_bmc(void)
 		} else if (strcmp(si_type[i], "bt") == 0) {
 			info->si_type = SI_BT;
 		} else {
-			printk(KERN_WARNING
-			       "ipmi_si: Interface type specified "
+			printk(KERN_WARNING PFX "Interface type specified "
 			       "for interface %d, was invalid: %s\n",
 			       i, si_type[i]);
 			kfree(info);
@@ -1841,11 +1864,9 @@ static __devinit void hardcode_find_bmc(void)
 			info->io.addr_data = addrs[i];
 			info->io.addr_type = IPMI_MEM_ADDR_SPACE;
 		} else {
-			printk(KERN_WARNING
-			       "ipmi_si: Interface type specified "
-			       "for interface %d, "
-			       "but port and address were not set or "
-			       "set to zero.\n", i);
+			printk(KERN_WARNING PFX "Interface type specified "
+			       "for interface %d, but port and address were "
+			       "not set or set to zero.\n", i);
 			kfree(info);
 			continue;
 		}
@@ -1863,7 +1884,9 @@ static __devinit void hardcode_find_bmc(void)
 			info->irq_setup = std_irq_setup;
 		info->slave_addr = slave_addrs[i];
 
-		try_smi_init(info);
+		if (!add_smi(info))
+			if (try_smi_init(info))
+				cleanup_one_si(info);
 	}
 }
 
@@ -1923,15 +1946,13 @@ static int acpi_gpe_irq_setup(struct smi_info *info)
 					  &ipmi_acpi_gpe,
 					  info);
 	if (status != AE_OK) {
-		printk(KERN_WARNING
-		       "ipmi_si: %s unable to claim ACPI GPE %d,"
-		       " running polled\n",
-		       DEVICE_NAME, info->irq);
+		dev_warn(info->dev, "%s unable to claim ACPI GPE %d,"
+			 " running polled\n", DEVICE_NAME, info->irq);
 		info->irq = 0;
 		return -EINVAL;
 	} else {
 		info->irq_cleanup = acpi_gpe_irq_cleanup;
-		printk("  Using ACPI GPE %d\n", info->irq);
+		dev_info(info->dev, "Using ACPI GPE %d\n", info->irq);
 		return 0;
 	}
 }
@@ -1989,8 +2010,8 @@ static __devinit int try_init_spmi(struct SPMITable *spmi)
 	u8 		 addr_space;
 
 	if (spmi->IPMIlegacy != 1) {
-	    printk(KERN_INFO "IPMI: Bad SPMI legacy %d\n", spmi->IPMIlegacy);
-	    return -ENODEV;
+		printk(KERN_INFO PFX "Bad SPMI legacy %d\n", spmi->IPMIlegacy);
+		return -ENODEV;
 	}
 
 	if (spmi->addr.space_id == ACPI_ADR_SPACE_SYSTEM_MEMORY)
@@ -2000,11 +2021,12 @@ static __devinit int try_init_spmi(struct SPMITable *spmi)
 
 	info = kzalloc(sizeof(*info), GFP_KERNEL);
 	if (!info) {
-		printk(KERN_ERR "ipmi_si: Could not allocate SI data (3)\n");
+		printk(KERN_ERR PFX "Could not allocate SI data (3)\n");
 		return -ENOMEM;
 	}
 
-	info->addr_source = "SPMI";
+	info->addr_source = SI_SPMI;
+	printk(KERN_INFO PFX "probing via SPMI\n");
 
 	/* Figure out the interface type. */
 	switch (spmi->InterfaceType) {
@@ -2018,8 +2040,8 @@ static __devinit int try_init_spmi(struct SPMITable *spmi)
 		info->si_type = SI_BT;
 		break;
 	default:
-		printk(KERN_INFO "ipmi_si: Unknown ACPI/SPMI SI type %d\n",
-			spmi->InterfaceType);
+		printk(KERN_INFO PFX "Unknown ACPI/SPMI SI type %d\n",
+		       spmi->InterfaceType);
 		kfree(info);
 		return -EIO;
 	}
@@ -2055,13 +2077,12 @@ static __devinit int try_init_spmi(struct SPMITable *spmi)
 		info->io.addr_type = IPMI_IO_ADDR_SPACE;
 	} else {
 		kfree(info);
-		printk(KERN_WARNING
-		       "ipmi_si: Unknown ACPI I/O Address type\n");
+		printk(KERN_WARNING PFX "Unknown ACPI I/O Address type\n");
 		return -EIO;
 	}
 	info->io.addr_data = spmi->addr.address;
 
-	try_smi_init(info);
+	add_smi(info);
 
 	return 0;
 }
@@ -2093,6 +2114,7 @@ static int __devinit ipmi_pnp_probe(struct pnp_dev *dev,
 {
 	struct acpi_device *acpi_dev;
 	struct smi_info *info;
+	struct resource *res;
 	acpi_handle handle;
 	acpi_status status;
 	unsigned long long tmp;
@@ -2105,7 +2127,8 @@ static int __devinit ipmi_pnp_probe(struct pnp_dev *dev,
 	if (!info)
 		return -ENOMEM;
 
-	info->addr_source = "ACPI";
+	info->addr_source = SI_ACPI;
+	printk(KERN_INFO PFX "probing via ACPI\n");
 
 	handle = acpi_dev->handle;
 
@@ -2125,22 +2148,26 @@ static int __devinit ipmi_pnp_probe(struct pnp_dev *dev,
 		info->si_type = SI_BT;
 		break;
 	default:
-		dev_info(&dev->dev, "unknown interface type %lld\n", tmp);
+		dev_info(&dev->dev, "unknown IPMI type %lld\n", tmp);
 		goto err_free;
 	}
 
-	if (pnp_port_valid(dev, 0)) {
+	res = pnp_get_resource(dev, IORESOURCE_IO, 0);
+	if (res) {
 		info->io_setup = port_setup;
 		info->io.addr_type = IPMI_IO_ADDR_SPACE;
-		info->io.addr_data = pnp_port_start(dev, 0);
-	} else if (pnp_mem_valid(dev, 0)) {
-		info->io_setup = mem_setup;
-		info->io.addr_type = IPMI_MEM_ADDR_SPACE;
-		info->io.addr_data = pnp_mem_start(dev, 0);
 	} else {
+		res = pnp_get_resource(dev, IORESOURCE_MEM, 0);
+		if (res) {
+			info->io_setup = mem_setup;
+			info->io.addr_type = IPMI_MEM_ADDR_SPACE;
+		}
+	}
+	if (!res) {
 		dev_err(&dev->dev, "no I/O or memory address\n");
 		goto err_free;
 	}
+	info->io.addr_data = res->start;
 
 	info->io.regspacing = DEFAULT_REGSPACING;
 	info->io.regsize = DEFAULT_REGSPACING;
@@ -2156,10 +2183,14 @@ static int __devinit ipmi_pnp_probe(struct pnp_dev *dev,
 		info->irq_setup = std_irq_setup;
 	}
 
-	info->dev = &acpi_dev->dev;
+	info->dev = &dev->dev;
 	pnp_set_drvdata(dev, info);
 
-	return try_smi_init(info);
+	dev_info(info->dev, "%pR regsize %d spacing %d irq %d\n",
+		 res, info->io.regsize, info->io.regspacing,
+		 info->irq);
+
+	return add_smi(info);
 
 err_free:
 	kfree(info);
@@ -2264,12 +2295,12 @@ static __devinit void try_init_dmi(struct dmi_ipmi_data *ipmi_data)
 
 	info = kzalloc(sizeof(*info), GFP_KERNEL);
 	if (!info) {
-		printk(KERN_ERR
-		       "ipmi_si: Could not allocate SI data\n");
+		printk(KERN_ERR PFX "Could not allocate SI data\n");
 		return;
 	}
 
-	info->addr_source = "SMBIOS";
+	info->addr_source = SI_SMBIOS;
+	printk(KERN_INFO PFX "probing via SMBIOS\n");
 
 	switch (ipmi_data->type) {
 	case 0x01: /* KCS */
@@ -2299,8 +2330,7 @@ static __devinit void try_init_dmi(struct dmi_ipmi_data *ipmi_data)
 
 	default:
 		kfree(info);
-		printk(KERN_WARNING
-		       "ipmi_si: Unknown SMBIOS I/O Address type: %d.\n",
+		printk(KERN_WARNING PFX "Unknown SMBIOS I/O Address type: %d\n",
 		       ipmi_data->addr_space);
 		return;
 	}
@@ -2318,7 +2348,7 @@ static __devinit void try_init_dmi(struct dmi_ipmi_data *ipmi_data)
 	if (info->irq)
 		info->irq_setup = std_irq_setup;
 
-	try_smi_init(info);
+	add_smi(info);
 }
 
 static void __devinit dmi_find_bmc(void)
@@ -2368,7 +2398,8 @@ static int __devinit ipmi_pci_probe(struct pci_dev *pdev,
 	if (!info)
 		return -ENOMEM;
 
-	info->addr_source = "PCI";
+	info->addr_source = SI_PCI;
+	dev_info(&pdev->dev, "probing via PCI");
 
 	switch (class_type) {
 	case PCI_ERMC_CLASSCODE_TYPE_SMIC:
@@ -2385,15 +2416,13 @@ static int __devinit ipmi_pci_probe(struct pci_dev *pdev,
 
 	default:
 		kfree(info);
-		printk(KERN_INFO "ipmi_si: %s: Unknown IPMI type: %d\n",
-		       pci_name(pdev), class_type);
+		dev_info(&pdev->dev, "Unknown IPMI type: %d\n", class_type);
 		return -ENOMEM;
 	}
 
 	rv = pci_enable_device(pdev);
 	if (rv) {
-		printk(KERN_ERR "ipmi_si: %s: couldn't enable PCI device\n",
-		       pci_name(pdev));
+		dev_err(&pdev->dev, "couldn't enable PCI device\n");
 		kfree(info);
 		return rv;
 	}
@@ -2421,7 +2450,11 @@ static int __devinit ipmi_pci_probe(struct pci_dev *pdev,
 	info->dev = &pdev->dev;
 	pci_set_drvdata(pdev, info);
 
-	return try_smi_init(info);
+	dev_info(&pdev->dev, "%pR regsize %d spacing %d irq %d\n",
+		&pdev->resource[0], info->io.regsize, info->io.regspacing,
+		info->irq);
+
+	return add_smi(info);
 }
 
 static void __devexit ipmi_pci_remove(struct pci_dev *pdev)
@@ -2469,11 +2502,11 @@ static int __devinit ipmi_of_probe(struct of_device *dev,
 	struct smi_info *info;
 	struct resource resource;
 	const int *regsize, *regspacing, *regshift;
-	struct device_node *np = dev->node;
+	struct device_node *np = dev->dev.of_node;
 	int ret;
 	int proplen;
 
-	dev_info(&dev->dev, PFX "probing via device tree\n");
+	dev_info(&dev->dev, "probing via device tree\n");
 
 	ret = of_address_to_resource(np, 0, &resource);
 	if (ret) {
@@ -2503,12 +2536,12 @@ static int __devinit ipmi_of_probe(struct of_device *dev,
 
 	if (!info) {
 		dev_err(&dev->dev,
-			PFX "could not allocate memory for OF probe\n");
+			"could not allocate memory for OF probe\n");
 		return -ENOMEM;
 	}
 
 	info->si_type		= (enum si_type) match->data;
-	info->addr_source	= "device-tree";
+	info->addr_source	= SI_DEVICETREE;
 	info->irq_setup		= std_irq_setup;
 
 	if (resource.flags & IORESOURCE_IO) {
@@ -2525,16 +2558,16 @@ static int __devinit ipmi_of_probe(struct of_device *dev,
 	info->io.regspacing	= regspacing ? *regspacing : DEFAULT_REGSPACING;
 	info->io.regshift	= regshift ? *regshift : 0;
 
-	info->irq		= irq_of_parse_and_map(dev->node, 0);
+	info->irq		= irq_of_parse_and_map(dev->dev.of_node, 0);
 	info->dev		= &dev->dev;
 
-	dev_dbg(&dev->dev, "addr 0x%lx regsize %d spacing %d irq %x\n",
+	dev_dbg(&dev->dev, "addr 0x%lx regsize %d spacing %d irq %d\n",
 		info->io.addr_data, info->io.regsize, info->io.regspacing,
 		info->irq);
 
 	dev_set_drvdata(&dev->dev, info);
 
-	return try_smi_init(info);
+	return add_smi(info);
 }
 
 static int __devexit ipmi_of_remove(struct of_device *dev)
@@ -2555,8 +2588,11 @@ static struct of_device_id ipmi_match[] =
 };
 
 static struct of_platform_driver ipmi_of_platform_driver = {
-	.name		= "ipmi",
-	.match_table	= ipmi_match,
+	.driver = {
+		.name = "ipmi",
+		.owner = THIS_MODULE,
+		.of_match_table = ipmi_match,
+	},
 	.probe		= ipmi_of_probe,
 	.remove		= __devexit_p(ipmi_of_remove),
 };
@@ -2640,9 +2676,8 @@ static int try_enable_event_buffer(struct smi_info *smi_info)
 
 	rv = wait_for_msg_done(smi_info);
 	if (rv) {
-		printk(KERN_WARNING
-		       "ipmi_si: Error getting response from get global,"
-		       " enables command, the event buffer is not"
+		printk(KERN_WARNING PFX "Error getting response from get"
+		       " global enables command, the event buffer is not"
 		       " enabled.\n");
 		goto out;
 	}
@@ -2654,10 +2689,8 @@ static int try_enable_event_buffer(struct smi_info *smi_info)
 			resp[0] != (IPMI_NETFN_APP_REQUEST | 1) << 2 ||
 			resp[1] != IPMI_GET_BMC_GLOBAL_ENABLES_CMD   ||
 			resp[2] != 0) {
-		printk(KERN_WARNING
-		       "ipmi_si: Invalid return from get global"
-		       " enables command, cannot enable the event"
-		       " buffer.\n");
+		printk(KERN_WARNING PFX "Invalid return from get global"
+		       " enables command, cannot enable the event buffer.\n");
 		rv = -EINVAL;
 		goto out;
 	}
@@ -2673,9 +2706,8 @@ static int try_enable_event_buffer(struct smi_info *smi_info)
 
 	rv = wait_for_msg_done(smi_info);
 	if (rv) {
-		printk(KERN_WARNING
-		       "ipmi_si: Error getting response from set global,"
-		       " enables command, the event buffer is not"
+		printk(KERN_WARNING PFX "Error getting response from set"
+		       " global, enables command, the event buffer is not"
 		       " enabled.\n");
 		goto out;
 	}
@@ -2686,10 +2718,8 @@ static int try_enable_event_buffer(struct smi_info *smi_info)
 	if (resp_len < 3 ||
 			resp[0] != (IPMI_NETFN_APP_REQUEST | 1) << 2 ||
 			resp[1] != IPMI_SET_BMC_GLOBAL_ENABLES_CMD) {
-		printk(KERN_WARNING
-		       "ipmi_si: Invalid return from get global,"
-		       "enables command, not enable the event"
-		       " buffer.\n");
+		printk(KERN_WARNING PFX "Invalid return from get global,"
+		       "enables command, not enable the event buffer.\n");
 		rv = -EINVAL;
 		goto out;
 	}
@@ -2948,7 +2978,7 @@ static __devinit void default_find_bmc(void)
 		if (!info)
 			return;
 
-		info->addr_source = NULL;
+		info->addr_source = SI_DEFAULT;
 
 		info->si_type = ipmi_defaults[i].type;
 		info->io_setup = port_setup;
@@ -2960,14 +2990,16 @@ static __devinit void default_find_bmc(void)
 		info->io.regsize = DEFAULT_REGSPACING;
 		info->io.regshift = 0;
 
-		if (try_smi_init(info) == 0) {
-			/* Found one... */
-			printk(KERN_INFO "ipmi_si: Found default %s state"
-			       " machine at %s address 0x%lx\n",
-			       si_to_str[info->si_type],
-			       addr_space_to_str[info->io.addr_type],
-			       info->io.addr_data);
-			return;
+		if (add_smi(info) == 0) {
+			if ((try_smi_init(info)) == 0) {
+				/* Found one... */
+				printk(KERN_INFO PFX "Found default %s"
+				" state machine at %s address 0x%lx\n",
+				si_to_str[info->si_type],
+				addr_space_to_str[info->io.addr_type],
+				info->io.addr_data);
+			} else
+				cleanup_one_si(info);
 		}
 	}
 }
@@ -2986,34 +3018,48 @@ static int is_new_interface(struct smi_info *info)
 	return 1;
 }
 
-static int try_smi_init(struct smi_info *new_smi)
+static int add_smi(struct smi_info *new_smi)
 {
-	int rv;
-	int i;
-
-	if (new_smi->addr_source) {
-		printk(KERN_INFO "ipmi_si: Trying %s-specified %s state"
-		       " machine at %s address 0x%lx, slave address 0x%x,"
-		       " irq %d\n",
-		       new_smi->addr_source,
-		       si_to_str[new_smi->si_type],
-		       addr_space_to_str[new_smi->io.addr_type],
-		       new_smi->io.addr_data,
-		       new_smi->slave_addr, new_smi->irq);
-	}
+	int rv = 0;
 
+	printk(KERN_INFO PFX "Adding %s-specified %s state machine",
+			ipmi_addr_src_to_str[new_smi->addr_source],
+			si_to_str[new_smi->si_type]);
 	mutex_lock(&smi_infos_lock);
 	if (!is_new_interface(new_smi)) {
-		printk(KERN_WARNING "ipmi_si: duplicate interface\n");
+		printk(KERN_CONT PFX "duplicate interface\n");
 		rv = -EBUSY;
 		goto out_err;
 	}
 
+	printk(KERN_CONT "\n");
+
 	/* So we know not to free it unless we have allocated one. */
 	new_smi->intf = NULL;
 	new_smi->si_sm = NULL;
 	new_smi->handlers = NULL;
 
+	list_add_tail(&new_smi->link, &smi_infos);
+
+out_err:
+	mutex_unlock(&smi_infos_lock);
+	return rv;
+}
+
+static int try_smi_init(struct smi_info *new_smi)
+{
+	int rv = 0;
+	int i;
+
+	printk(KERN_INFO PFX "Trying %s-specified %s state"
+	       " machine at %s address 0x%lx, slave address 0x%x,"
+	       " irq %d\n",
+	       ipmi_addr_src_to_str[new_smi->addr_source],
+	       si_to_str[new_smi->si_type],
+	       addr_space_to_str[new_smi->io.addr_type],
+	       new_smi->io.addr_data,
+	       new_smi->slave_addr, new_smi->irq);
+
 	switch (new_smi->si_type) {
 	case SI_KCS:
 		new_smi->handlers = &kcs_smi_handlers;
@@ -3036,7 +3082,8 @@ static int try_smi_init(struct smi_info *new_smi)
 	/* Allocate the state machine's data and initialize it. */
 	new_smi->si_sm = kmalloc(new_smi->handlers->size(), GFP_KERNEL);
 	if (!new_smi->si_sm) {
-		printk(KERN_ERR "Could not allocate state machine memory\n");
+		printk(KERN_ERR PFX
+		       "Could not allocate state machine memory\n");
 		rv = -ENOMEM;
 		goto out_err;
 	}
@@ -3046,7 +3093,7 @@ static int try_smi_init(struct smi_info *new_smi)
 	/* Now that we know the I/O size, we can set up the I/O. */
 	rv = new_smi->io_setup(new_smi);
 	if (rv) {
-		printk(KERN_ERR "Could not set up I/O space\n");
+		printk(KERN_ERR PFX "Could not set up I/O space\n");
 		goto out_err;
 	}
 
@@ -3056,8 +3103,7 @@ static int try_smi_init(struct smi_info *new_smi)
 	/* Do low-level detection first. */
 	if (new_smi->handlers->detect(new_smi->si_sm)) {
 		if (new_smi->addr_source)
-			printk(KERN_INFO "ipmi_si: Interface detection"
-			       " failed\n");
+			printk(KERN_INFO PFX "Interface detection failed\n");
 		rv = -ENODEV;
 		goto out_err;
 	}
@@ -3069,7 +3115,7 @@ static int try_smi_init(struct smi_info *new_smi)
 	rv = try_get_dev_id(new_smi);
 	if (rv) {
 		if (new_smi->addr_source)
-			printk(KERN_INFO "ipmi_si: There appears to be no BMC"
+			printk(KERN_INFO PFX "There appears to be no BMC"
 			       " at this location\n");
 		goto out_err;
 	}
@@ -3085,7 +3131,7 @@ static int try_smi_init(struct smi_info *new_smi)
 	for (i = 0; i < SI_NUM_STATS; i++)
 		atomic_set(&new_smi->stats[i], 0);
 
-	new_smi->interrupt_disabled = 0;
+	new_smi->interrupt_disabled = 1;
 	atomic_set(&new_smi->stop_operation, 0);
 	new_smi->intf_num = smi_num;
 	smi_num++;
@@ -3111,9 +3157,8 @@ static int try_smi_init(struct smi_info *new_smi)
 		new_smi->pdev = platform_device_alloc("ipmi_si",
 						      new_smi->intf_num);
 		if (!new_smi->pdev) {
-			printk(KERN_ERR
-			       "ipmi_si_intf:"
-			       " Unable to allocate platform device\n");
+			printk(KERN_ERR PFX
+			       "Unable to allocate platform device\n");
 			goto out_err;
 		}
 		new_smi->dev = &new_smi->pdev->dev;
@@ -3121,9 +3166,8 @@ static int try_smi_init(struct smi_info *new_smi)
 
 		rv = platform_device_add(new_smi->pdev);
 		if (rv) {
-			printk(KERN_ERR
-			       "ipmi_si_intf:"
-			       " Unable to register system interface device:"
+			printk(KERN_ERR PFX
+			       "Unable to register system interface device:"
 			       " %d\n",
 			       rv);
 			goto out_err;
@@ -3138,9 +3182,8 @@ static int try_smi_init(struct smi_info *new_smi)
 			       "bmc",
 			       new_smi->slave_addr);
 	if (rv) {
-		printk(KERN_ERR
-		       "ipmi_si: Unable to register device: error %d\n",
-		       rv);
+		dev_err(new_smi->dev, "Unable to register device: error %d\n",
+			rv);
 		goto out_err_stop_timer;
 	}
 
@@ -3148,9 +3191,7 @@ static int try_smi_init(struct smi_info *new_smi)
 				     type_file_read_proc,
 				     new_smi);
 	if (rv) {
-		printk(KERN_ERR
-		       "ipmi_si: Unable to create proc entry: %d\n",
-		       rv);
+		dev_err(new_smi->dev, "Unable to create proc entry: %d\n", rv);
 		goto out_err_stop_timer;
 	}
 
@@ -3158,9 +3199,7 @@ static int try_smi_init(struct smi_info *new_smi)
 				     stat_file_read_proc,
 				     new_smi);
 	if (rv) {
-		printk(KERN_ERR
-		       "ipmi_si: Unable to create proc entry: %d\n",
-		       rv);
+		dev_err(new_smi->dev, "Unable to create proc entry: %d\n", rv);
 		goto out_err_stop_timer;
 	}
 
@@ -3168,18 +3207,12 @@ static int try_smi_init(struct smi_info *new_smi)
 				     param_read_proc,
 				     new_smi);
 	if (rv) {
-		printk(KERN_ERR
-		       "ipmi_si: Unable to create proc entry: %d\n",
-		       rv);
+		dev_err(new_smi->dev, "Unable to create proc entry: %d\n", rv);
 		goto out_err_stop_timer;
 	}
 
-	list_add_tail(&new_smi->link, &smi_infos);
-
-	mutex_unlock(&smi_infos_lock);
-
-	printk(KERN_INFO "IPMI %s interface initialized\n",
-	       si_to_str[new_smi->si_type]);
+	dev_info(new_smi->dev, "IPMI %s interface initialized\n",
+		 si_to_str[new_smi->si_type]);
 
 	return 0;
 
@@ -3188,11 +3221,17 @@ static int try_smi_init(struct smi_info *new_smi)
 	wait_for_timer_and_thread(new_smi);
 
  out_err:
-	if (new_smi->intf)
+	new_smi->interrupt_disabled = 1;
+
+	if (new_smi->intf) {
 		ipmi_unregister_smi(new_smi->intf);
+		new_smi->intf = NULL;
+	}
 
-	if (new_smi->irq_cleanup)
+	if (new_smi->irq_cleanup) {
 		new_smi->irq_cleanup(new_smi);
+		new_smi->irq_cleanup = NULL;
+	}
 
 	/*
 	 * Wait until we know that we are out of any interrupt
@@ -3205,18 +3244,21 @@ static int try_smi_init(struct smi_info *new_smi)
 		if (new_smi->handlers)
 			new_smi->handlers->cleanup(new_smi->si_sm);
 		kfree(new_smi->si_sm);
+		new_smi->si_sm = NULL;
 	}
-	if (new_smi->addr_source_cleanup)
+	if (new_smi->addr_source_cleanup) {
 		new_smi->addr_source_cleanup(new_smi);
-	if (new_smi->io_cleanup)
+		new_smi->addr_source_cleanup = NULL;
+	}
+	if (new_smi->io_cleanup) {
 		new_smi->io_cleanup(new_smi);
+		new_smi->io_cleanup = NULL;
+	}
 
-	if (new_smi->dev_registered)
+	if (new_smi->dev_registered) {
 		platform_device_unregister(new_smi->pdev);
-
-	kfree(new_smi);
-
-	mutex_unlock(&smi_infos_lock);
+		new_smi->dev_registered = 0;
+	}
 
 	return rv;
 }
@@ -3226,6 +3268,8 @@ static __devinit int init_ipmi_si(void)
 	int  i;
 	char *str;
 	int  rv;
+	struct smi_info *e;
+	enum ipmi_addr_src type = SI_INVALID;
 
 	if (initialized)
 		return 0;
@@ -3234,9 +3278,7 @@ static __devinit int init_ipmi_si(void)
 	/* Register the device drivers. */
 	rv = driver_register(&ipmi_driver.driver);
 	if (rv) {
-		printk(KERN_ERR
-		       "init_ipmi_si: Unable to register driver: %d\n",
-		       rv);
+		printk(KERN_ERR PFX "Unable to register driver: %d\n", rv);
 		return rv;
 	}
 
@@ -3260,38 +3302,81 @@ static __devinit int init_ipmi_si(void)
 
 	hardcode_find_bmc();
 
-#ifdef CONFIG_DMI
-	dmi_find_bmc();
-#endif
+	/* If the user gave us a device, they presumably want us to use it */
+	mutex_lock(&smi_infos_lock);
+	if (!list_empty(&smi_infos)) {
+		mutex_unlock(&smi_infos_lock);
+		return 0;
+	}
+	mutex_unlock(&smi_infos_lock);
 
-#ifdef CONFIG_ACPI
-	spmi_find_bmc();
+#ifdef CONFIG_PCI
+	rv = pci_register_driver(&ipmi_pci_driver);
+	if (rv)
+		printk(KERN_ERR PFX "Unable to register PCI driver: %d\n", rv);
 #endif
+
 #ifdef CONFIG_ACPI
 	pnp_register_driver(&ipmi_pnp_driver);
 #endif
 
-#ifdef CONFIG_PCI
-	rv = pci_register_driver(&ipmi_pci_driver);
-	if (rv)
-		printk(KERN_ERR
-		       "init_ipmi_si: Unable to register PCI driver: %d\n",
-		       rv);
+#ifdef CONFIG_DMI
+	dmi_find_bmc();
+#endif
+
+#ifdef CONFIG_ACPI
+	spmi_find_bmc();
 #endif
 
 #ifdef CONFIG_PPC_OF
 	of_register_platform_driver(&ipmi_of_platform_driver);
 #endif
 
+	/* We prefer devices with interrupts, but in the case of a machine
+	   with multiple BMCs we assume that there will be several instances
+	   of a given type so if we succeed in registering a type then also
+	   try to register everything else of the same type */
+
+	mutex_lock(&smi_infos_lock);
+	list_for_each_entry(e, &smi_infos, link) {
+		/* Try to register a device if it has an IRQ and we either
+		   haven't successfully registered a device yet or this
+		   device has the same type as one we successfully registered */
+		if (e->irq && (!type || e->addr_source == type)) {
+			if (!try_smi_init(e)) {
+				type = e->addr_source;
+			}
+		}
+	}
+
+	/* type will only have been set if we successfully registered an si */
+	if (type) {
+		mutex_unlock(&smi_infos_lock);
+		return 0;
+	}
+
+	/* Fall back to the preferred device */
+
+	list_for_each_entry(e, &smi_infos, link) {
+		if (!e->irq && (!type || e->addr_source == type)) {
+			if (!try_smi_init(e)) {
+				type = e->addr_source;
+			}
+		}
+	}
+	mutex_unlock(&smi_infos_lock);
+
+	if (type)
+		return 0;
+
 	if (si_trydefaults) {
 		mutex_lock(&smi_infos_lock);
 		if (list_empty(&smi_infos)) {
 			/* No BMC was found, try defaults. */
 			mutex_unlock(&smi_infos_lock);
 			default_find_bmc();
-		} else {
+		} else
 			mutex_unlock(&smi_infos_lock);
-		}
 	}
 
 	mutex_lock(&smi_infos_lock);
@@ -3305,8 +3390,8 @@ static __devinit int init_ipmi_si(void)
 		of_unregister_platform_driver(&ipmi_of_platform_driver);
 #endif
 		driver_unregister(&ipmi_driver.driver);
-		printk(KERN_WARNING
-		       "ipmi_si: Unable to find any System Interface(s)\n");
+		printk(KERN_WARNING PFX
+		       "Unable to find any System Interface(s)\n");
 		return -ENODEV;
 	} else {
 		mutex_unlock(&smi_infos_lock);
@@ -3317,7 +3402,7 @@ module_init(init_ipmi_si);
 
 static void cleanup_one_si(struct smi_info *to_clean)
 {
-	int           rv;
+	int           rv = 0;
 	unsigned long flags;
 
 	if (!to_clean)
@@ -3361,14 +3446,16 @@ static void cleanup_one_si(struct smi_info *to_clean)
 		schedule_timeout_uninterruptible(1);
 	}
 
-	rv = ipmi_unregister_smi(to_clean->intf);
+	if (to_clean->intf)
+		rv = ipmi_unregister_smi(to_clean->intf);
+
 	if (rv) {
-		printk(KERN_ERR
-		       "ipmi_si: Unable to unregister device: errno=%d\n",
+		printk(KERN_ERR PFX "Unable to unregister device: errno=%d\n",
 		       rv);
 	}
 
-	to_clean->handlers->cleanup(to_clean->si_sm);
+	if (to_clean->handlers)
+		to_clean->handlers->cleanup(to_clean->si_sm);
 
 	kfree(to_clean->si_sm);
 
diff --git a/drivers/char/ipmi/ipmi_watchdog.c b/drivers/char/ipmi/ipmi_watchdog.c
index a4d57e31f713..82bcdb262a3a 100644
--- a/drivers/char/ipmi/ipmi_watchdog.c
+++ b/drivers/char/ipmi/ipmi_watchdog.c
@@ -659,7 +659,7 @@ static struct watchdog_info ident = {
 	.identity	= "IPMI"
 };
 
-static int ipmi_ioctl(struct inode *inode, struct file *file,
+static int ipmi_ioctl(struct file *file,
 		      unsigned int cmd, unsigned long arg)
 {
 	void __user *argp = (void __user *)arg;
@@ -730,6 +730,19 @@ static int ipmi_ioctl(struct inode *inode, struct file *file,
 	}
 }
 
+static long ipmi_unlocked_ioctl(struct file *file,
+				unsigned int cmd,
+				unsigned long arg)
+{
+	int ret;
+
+	lock_kernel();
+	ret = ipmi_ioctl(file, cmd, arg);
+	unlock_kernel();
+
+	return ret;
+}
+
 static ssize_t ipmi_write(struct file *file,
 			  const char  __user *buf,
 			  size_t      len,
@@ -880,7 +893,7 @@ static const struct file_operations ipmi_wdog_fops = {
 	.read    = ipmi_read,
 	.poll    = ipmi_poll,
 	.write   = ipmi_write,
-	.ioctl   = ipmi_ioctl,
+	.unlocked_ioctl = ipmi_unlocked_ioctl,
 	.open    = ipmi_open,
 	.release = ipmi_close,
 	.fasync  = ipmi_fasync,
diff --git a/drivers/char/misc.c b/drivers/char/misc.c
index 92ab03d28294..cd650ca8c679 100644
--- a/drivers/char/misc.c
+++ b/drivers/char/misc.c
@@ -144,6 +144,7 @@ static int misc_open(struct inode * inode, struct file * file)
 	old_fops = file->f_op;
 	file->f_op = new_fops;
 	if (file->f_op->open) {
+		file->private_data = c;
 		err=file->f_op->open(inode,file);
 		if (err) {
 			fops_put(file->f_op);
diff --git a/drivers/char/n_gsm.c b/drivers/char/n_gsm.c
index c4161d5e053d..e4089c432f15 100644
--- a/drivers/char/n_gsm.c
+++ b/drivers/char/n_gsm.c
@@ -904,9 +904,7 @@ static void gsm_dlci_data_sweep(struct gsm_mux *gsm)
 	int len;
 	/* Priority ordering: We should do priority with RR of the groups */
 	int i = 1;
-	unsigned long flags;
 
-	spin_lock_irqsave(&gsm->tx_lock, flags);
 	while (i < NUM_DLCI) {
 		struct gsm_dlci *dlci;
 
@@ -927,7 +925,6 @@ static void gsm_dlci_data_sweep(struct gsm_mux *gsm)
 		if (len == 0)
 			i++;
 	}
-	spin_unlock_irqrestore(&gsm->tx_lock, flags);
 }
 
 /**
@@ -2230,12 +2227,16 @@ static int gsmld_open(struct tty_struct *tty)
 static void gsmld_write_wakeup(struct tty_struct *tty)
 {
 	struct gsm_mux *gsm = tty->disc_data;
+	unsigned long flags;
 
 	/* Queue poll */
 	clear_bit(TTY_DO_WRITE_WAKEUP, &tty->flags);
 	gsm_data_kick(gsm);
-	if (gsm->tx_bytes < TX_THRESH_LO)
+	if (gsm->tx_bytes < TX_THRESH_LO) {
+		spin_lock_irqsave(&gsm->tx_lock, flags);
 		gsm_dlci_data_sweep(gsm);
+		spin_unlock_irqrestore(&gsm->tx_lock, flags);
+	}
 }
 
 /**
diff --git a/drivers/char/nvram.c b/drivers/char/nvram.c
index 47e8f7b0e4c1..66d2917b003f 100644
--- a/drivers/char/nvram.c
+++ b/drivers/char/nvram.c
@@ -296,8 +296,8 @@ checksum_err:
 	return -EIO;
 }
 
-static int nvram_ioctl(struct inode *inode, struct file *file,
-					unsigned int cmd, unsigned long arg)
+static long nvram_ioctl(struct file *file, unsigned int cmd,
+			unsigned long arg)
 {
 	int i;
 
@@ -308,6 +308,7 @@ static int nvram_ioctl(struct inode *inode, struct file *file,
 		if (!capable(CAP_SYS_ADMIN))
 			return -EACCES;
 
+		lock_kernel();
 		spin_lock_irq(&rtc_lock);
 
 		for (i = 0; i < NVRAM_BYTES; ++i)
@@ -315,6 +316,7 @@ static int nvram_ioctl(struct inode *inode, struct file *file,
 		__nvram_set_checksum();
 
 		spin_unlock_irq(&rtc_lock);
+		unlock_kernel();
 		return 0;
 
 	case NVRAM_SETCKS:
@@ -323,9 +325,11 @@ static int nvram_ioctl(struct inode *inode, struct file *file,
 		if (!capable(CAP_SYS_ADMIN))
 			return -EACCES;
 
+		lock_kernel();
 		spin_lock_irq(&rtc_lock);
 		__nvram_set_checksum();
 		spin_unlock_irq(&rtc_lock);
+		unlock_kernel();
 		return 0;
 
 	default:
@@ -422,7 +426,7 @@ static const struct file_operations nvram_fops = {
 	.llseek		= nvram_llseek,
 	.read		= nvram_read,
 	.write		= nvram_write,
-	.ioctl		= nvram_ioctl,
+	.unlocked_ioctl	= nvram_ioctl,
 	.open		= nvram_open,
 	.release	= nvram_release,
 };
diff --git a/drivers/char/nwflash.c b/drivers/char/nwflash.c
index f80810901db6..043a1c7b86be 100644
--- a/drivers/char/nwflash.c
+++ b/drivers/char/nwflash.c
@@ -94,8 +94,9 @@ static int get_flash_id(void)
 	return c2;
 }
 
-static int flash_ioctl(struct inode *inodep, struct file *filep, unsigned int cmd, unsigned long arg)
+static long flash_ioctl(struct file *filep, unsigned int cmd, unsigned long arg)
 {
+	lock_kernel();
 	switch (cmd) {
 	case CMD_WRITE_DISABLE:
 		gbWriteBase64Enable = 0;
@@ -113,8 +114,10 @@ static int flash_ioctl(struct inode *inodep, struct file *filep, unsigned int cm
 	default:
 		gbWriteBase64Enable = 0;
 		gbWriteEnable = 0;
+		unlock_kernel();
 		return -EINVAL;
 	}
+	unlock_kernel();
 	return 0;
 }
 
@@ -631,7 +634,7 @@ static const struct file_operations flash_fops =
 	.llseek		= flash_llseek,
 	.read		= flash_read,
 	.write		= flash_write,
-	.ioctl		= flash_ioctl,
+	.unlocked_ioctl	= flash_ioctl,
 };
 
 static struct miscdevice flash_miscdev =
diff --git a/drivers/char/ppdev.c b/drivers/char/ppdev.c
index fdd37543aa79..02abfddce45a 100644
--- a/drivers/char/ppdev.c
+++ b/drivers/char/ppdev.c
@@ -287,12 +287,10 @@ static int register_device (int minor, struct pp_struct *pp)
 	char *name;
 	int fl;
 
-	name = kmalloc (strlen (CHRDEV) + 3, GFP_KERNEL);
+	name = kasprintf(GFP_KERNEL, CHRDEV "%x", minor);
 	if (name == NULL)
 		return -ENOMEM;
 
-	sprintf (name, CHRDEV "%x", minor);
-
 	port = parport_find_number (minor);
 	if (!port) {
 		printk (KERN_WARNING "%s: no associated port!\n", name);
diff --git a/drivers/char/ps3flash.c b/drivers/char/ps3flash.c
index 606048b72bcf..85c004a518ee 100644
--- a/drivers/char/ps3flash.c
+++ b/drivers/char/ps3flash.c
@@ -305,8 +305,7 @@ static int ps3flash_flush(struct file *file, fl_owner_t id)
 	return ps3flash_writeback(ps3flash_dev);
 }
 
-static int ps3flash_fsync(struct file *file, struct dentry *dentry,
-			  int datasync)
+static int ps3flash_fsync(struct file *file, int datasync)
 {
 	return ps3flash_writeback(ps3flash_dev);
 }
diff --git a/drivers/char/ramoops.c b/drivers/char/ramoops.c
new file mode 100644
index 000000000000..74f00b5ffa36
--- /dev/null
+++ b/drivers/char/ramoops.c
@@ -0,0 +1,162 @@
+/*
+ * RAM Oops/Panic logger
+ *
+ * Copyright (C) 2010 Marco Stornelli <marco.stornelli@gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
+ * 02110-1301 USA
+ *
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/kmsg_dump.h>
+#include <linux/time.h>
+#include <linux/io.h>
+#include <linux/ioport.h>
+
+#define RAMOOPS_KERNMSG_HDR "===="
+#define RAMOOPS_HEADER_SIZE   (5 + sizeof(struct timeval))
+
+#define RECORD_SIZE 4096
+
+static ulong mem_address;
+module_param(mem_address, ulong, 0400);
+MODULE_PARM_DESC(mem_address,
+		"start of reserved RAM used to store oops/panic logs");
+
+static ulong mem_size;
+module_param(mem_size, ulong, 0400);
+MODULE_PARM_DESC(mem_size,
+		"size of reserved RAM used to store oops/panic logs");
+
+static int dump_oops = 1;
+module_param(dump_oops, int, 0600);
+MODULE_PARM_DESC(dump_oops,
+		"set to 1 to dump oopses, 0 to only dump panics (default 1)");
+
+static struct ramoops_context {
+	struct kmsg_dumper dump;
+	void *virt_addr;
+	phys_addr_t phys_addr;
+	unsigned long size;
+	int count;
+	int max_count;
+} oops_cxt;
+
+static void ramoops_do_dump(struct kmsg_dumper *dumper,
+		enum kmsg_dump_reason reason, const char *s1, unsigned long l1,
+		const char *s2, unsigned long l2)
+{
+	struct ramoops_context *cxt = container_of(dumper,
+			struct ramoops_context, dump);
+	unsigned long s1_start, s2_start;
+	unsigned long l1_cpy, l2_cpy;
+	int res;
+	char *buf;
+	struct timeval timestamp;
+
+	/* Only dump oopses if dump_oops is set */
+	if (reason == KMSG_DUMP_OOPS && !dump_oops)
+		return;
+
+	buf = (char *)(cxt->virt_addr + (cxt->count * RECORD_SIZE));
+	memset(buf, '\0', RECORD_SIZE);
+	res = sprintf(buf, "%s", RAMOOPS_KERNMSG_HDR);
+	buf += res;
+	do_gettimeofday(&timestamp);
+	res = sprintf(buf, "%lu.%lu\n", (long)timestamp.tv_sec, (long)timestamp.tv_usec);
+	buf += res;
+
+	l2_cpy = min(l2, (unsigned long)(RECORD_SIZE - RAMOOPS_HEADER_SIZE));
+	l1_cpy = min(l1, (unsigned long)(RECORD_SIZE - RAMOOPS_HEADER_SIZE) - l2_cpy);
+
+	s2_start = l2 - l2_cpy;
+	s1_start = l1 - l1_cpy;
+
+	memcpy(buf, s1 + s1_start, l1_cpy);
+	memcpy(buf + l1_cpy, s2 + s2_start, l2_cpy);
+
+	cxt->count = (cxt->count + 1) % cxt->max_count;
+}
+
+static int __init ramoops_init(void)
+{
+	struct ramoops_context *cxt = &oops_cxt;
+	int err = -EINVAL;
+
+	if (!mem_size) {
+		printk(KERN_ERR "ramoops: invalid size specification");
+		goto fail3;
+	}
+
+	rounddown_pow_of_two(mem_size);
+
+	if (mem_size < RECORD_SIZE) {
+		printk(KERN_ERR "ramoops: size too small");
+		goto fail3;
+	}
+
+	cxt->max_count = mem_size / RECORD_SIZE;
+	cxt->count = 0;
+	cxt->size = mem_size;
+	cxt->phys_addr = mem_address;
+
+	if (!request_mem_region(cxt->phys_addr, cxt->size, "ramoops")) {
+		printk(KERN_ERR "ramoops: request mem region failed");
+		err = -EINVAL;
+		goto fail3;
+	}
+
+	cxt->virt_addr = ioremap(cxt->phys_addr,  cxt->size);
+	if (!cxt->virt_addr) {
+		printk(KERN_ERR "ramoops: ioremap failed");
+		goto fail2;
+	}
+
+	cxt->dump.dump = ramoops_do_dump;
+	err = kmsg_dump_register(&cxt->dump);
+	if (err) {
+		printk(KERN_ERR "ramoops: registering kmsg dumper failed");
+		goto fail1;
+	}
+
+	return 0;
+
+fail1:
+	iounmap(cxt->virt_addr);
+fail2:
+	release_mem_region(cxt->phys_addr, cxt->size);
+fail3:
+	return err;
+}
+
+static void __exit ramoops_exit(void)
+{
+	struct ramoops_context *cxt = &oops_cxt;
+
+	if (kmsg_dump_unregister(&cxt->dump) < 0)
+		printk(KERN_WARNING "ramoops: could not unregister kmsg_dumper");
+
+	iounmap(cxt->virt_addr);
+	release_mem_region(cxt->phys_addr, cxt->size);
+}
+
+
+module_init(ramoops_init);
+module_exit(ramoops_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Marco Stornelli <marco.stornelli@gmail.com>");
+MODULE_DESCRIPTION("RAM Oops/Panic logger/driver");
diff --git a/drivers/char/random.c b/drivers/char/random.c
index 2fd3d39995d5..8d85587b6d4f 100644
--- a/drivers/char/random.c
+++ b/drivers/char/random.c
@@ -257,6 +257,7 @@
 #define INPUT_POOL_WORDS 128
 #define OUTPUT_POOL_WORDS 32
 #define SEC_XFER_SIZE 512
+#define EXTRACT_SIZE 10
 
 /*
  * The minimum number of bits of entropy before we wake up a read on
@@ -414,7 +415,7 @@ struct entropy_store {
 	unsigned add_ptr;
 	int entropy_count;
 	int input_rotate;
-	__u8 *last_data;
+	__u8 last_data[EXTRACT_SIZE];
 };
 
 static __u32 input_pool_data[INPUT_POOL_WORDS];
@@ -714,8 +715,6 @@ void add_disk_randomness(struct gendisk *disk)
 }
 #endif
 
-#define EXTRACT_SIZE 10
-
 /*********************************************************************
  *
  * Entropy extraction routines
@@ -862,7 +861,7 @@ static ssize_t extract_entropy(struct entropy_store *r, void *buf,
 	while (nbytes) {
 		extract_buf(r, tmp);
 
-		if (r->last_data) {
+		if (fips_enabled) {
 			spin_lock_irqsave(&r->lock, flags);
 			if (!memcmp(tmp, r->last_data, EXTRACT_SIZE))
 				panic("Hardware RNG duplicated output!\n");
@@ -951,9 +950,6 @@ static void init_std_data(struct entropy_store *r)
 	now = ktime_get_real();
 	mix_pool_bytes(r, &now, sizeof(now));
 	mix_pool_bytes(r, utsname(), sizeof(*(utsname())));
-	/* Enable continuous test in fips mode */
-	if (fips_enabled)
-		r->last_data = kmalloc(EXTRACT_SIZE, GFP_KERNEL);
 }
 
 static int rand_initialize(void)
diff --git a/drivers/char/raw.c b/drivers/char/raw.c
index 8756ab0daa8b..b38942f6bf31 100644
--- a/drivers/char/raw.c
+++ b/drivers/char/raw.c
@@ -121,13 +121,17 @@ static int raw_release(struct inode *inode, struct file *filp)
 /*
  * Forward ioctls to the underlying block device.
  */
-static int
-raw_ioctl(struct inode *inode, struct file *filp,
-		  unsigned int command, unsigned long arg)
+static long
+raw_ioctl(struct file *filp, unsigned int command, unsigned long arg)
 {
 	struct block_device *bdev = filp->private_data;
+	int ret;
+
+	lock_kernel();
+	ret = blkdev_ioctl(bdev, 0, command, arg);
+	unlock_kernel();
 
-	return blkdev_ioctl(bdev, 0, command, arg);
+	return ret;
 }
 
 static void bind_device(struct raw_config_request *rq)
@@ -141,13 +145,14 @@ static void bind_device(struct raw_config_request *rq)
  * Deal with ioctls against the raw-device control interface, to bind
  * and unbind other raw devices.
  */
-static int raw_ctl_ioctl(struct inode *inode, struct file *filp,
-			unsigned int command, unsigned long arg)
+static long raw_ctl_ioctl(struct file *filp, unsigned int command,
+			  unsigned long arg)
 {
 	struct raw_config_request rq;
 	struct raw_device_data *rawdev;
 	int err = 0;
 
+	lock_kernel();
 	switch (command) {
 	case RAW_SETBIND:
 	case RAW_GETBIND:
@@ -240,25 +245,26 @@ static int raw_ctl_ioctl(struct inode *inode, struct file *filp,
 		break;
 	}
 out:
+	unlock_kernel();
 	return err;
 }
 
 static const struct file_operations raw_fops = {
-	.read	=	do_sync_read,
-	.aio_read = 	generic_file_aio_read,
-	.write	=	do_sync_write,
-	.aio_write =	blkdev_aio_write,
-	.fsync	=	blkdev_fsync,
-	.open	=	raw_open,
-	.release=	raw_release,
-	.ioctl	=	raw_ioctl,
-	.owner	=	THIS_MODULE,
+	.read		= do_sync_read,
+	.aio_read	= generic_file_aio_read,
+	.write		= do_sync_write,
+	.aio_write	= blkdev_aio_write,
+	.fsync		= blkdev_fsync,
+	.open		= raw_open,
+	.release	= raw_release,
+	.unlocked_ioctl = raw_ioctl,
+	.owner		= THIS_MODULE,
 };
 
 static const struct file_operations raw_ctl_fops = {
-	.ioctl	=	raw_ctl_ioctl,
-	.open	=	raw_open,
-	.owner	=	THIS_MODULE,
+	.unlocked_ioctl = raw_ctl_ioctl,
+	.open		= raw_open,
+	.owner		= THIS_MODULE,
 };
 
 static struct cdev raw_cdev;
diff --git a/drivers/char/viotape.c b/drivers/char/viotape.c
index 1144a04cda6e..42f7fa442ff8 100644
--- a/drivers/char/viotape.c
+++ b/drivers/char/viotape.c
@@ -866,7 +866,7 @@ static int viotape_probe(struct vio_dev *vdev, const struct vio_device_id *id)
 {
 	int i = vdev->unit_address;
 	int j;
-	struct device_node *node = vdev->dev.archdata.of_node;
+	struct device_node *node = vdev->dev.of_node;
 
 	if (i >= VIOTAPE_MAX_TAPE)
 		return -ENODEV;
diff --git a/drivers/char/virtio_console.c b/drivers/char/virtio_console.c
index 196428c2287a..942a9826bd23 100644
--- a/drivers/char/virtio_console.c
+++ b/drivers/char/virtio_console.c
@@ -33,35 +33,6 @@
 #include <linux/workqueue.h>
 #include "hvc_console.h"
 
-/* Moved here from .h file in order to disable MULTIPORT. */
-#define VIRTIO_CONSOLE_F_MULTIPORT 1	/* Does host provide multiple ports? */
-
-struct virtio_console_multiport_conf {
-	struct virtio_console_config config;
-	/* max. number of ports this device can hold */
-	__u32 max_nr_ports;
-	/* number of ports added so far */
-	__u32 nr_ports;
-} __attribute__((packed));
-
-/*
- * A message that's passed between the Host and the Guest for a
- * particular port.
- */
-struct virtio_console_control {
-	__u32 id;		/* Port number */
-	__u16 event;		/* The kind of control event (see below) */
-	__u16 value;		/* Extra information for the key */
-};
-
-/* Some events for control messages */
-#define VIRTIO_CONSOLE_PORT_READY	0
-#define VIRTIO_CONSOLE_CONSOLE_PORT	1
-#define VIRTIO_CONSOLE_RESIZE		2
-#define VIRTIO_CONSOLE_PORT_OPEN	3
-#define VIRTIO_CONSOLE_PORT_NAME	4
-#define VIRTIO_CONSOLE_PORT_REMOVE	5
-
 /*
  * This is a global struct for storing common data for all the devices
  * this driver handles.
@@ -107,6 +78,9 @@ struct console {
 	/* The hvc device associated with this console port */
 	struct hvc_struct *hvc;
 
+	/* The size of the console */
+	struct winsize ws;
+
 	/*
 	 * This number identifies the number that we used to register
 	 * with hvc in hvc_instantiate() and hvc_alloc(); this is the
@@ -139,7 +113,6 @@ struct ports_device {
 	 * notification
 	 */
 	struct work_struct control_work;
-	struct work_struct config_work;
 
 	struct list_head ports;
 
@@ -150,7 +123,7 @@ struct ports_device {
 	spinlock_t cvq_lock;
 
 	/* The current config space is stored here */
-	struct virtio_console_multiport_conf config;
+	struct virtio_console_config config;
 
 	/* The virtio device we're associated with */
 	struct virtio_device *vdev;
@@ -189,6 +162,9 @@ struct port {
 	 */
 	spinlock_t inbuf_lock;
 
+	/* Protect the operations on the out_vq. */
+	spinlock_t outvq_lock;
+
 	/* The IO vqs for this port */
 	struct virtqueue *in_vq, *out_vq;
 
@@ -214,6 +190,8 @@ struct port {
 	/* The 'id' to identify the port with the Host */
 	u32 id;
 
+	bool outvq_full;
+
 	/* Is the host device open */
 	bool host_connected;
 
@@ -328,7 +306,7 @@ static void *get_inbuf(struct port *port)
 	unsigned int len;
 
 	vq = port->in_vq;
-	buf = vq->vq_ops->get_buf(vq, &len);
+	buf = virtqueue_get_buf(vq, &len);
 	if (buf) {
 		buf->len = len;
 		buf->offset = 0;
@@ -349,8 +327,8 @@ static int add_inbuf(struct virtqueue *vq, struct port_buffer *buf)
 
 	sg_init_one(sg, buf->buf, buf->size);
 
-	ret = vq->vq_ops->add_buf(vq, sg, 0, 1, buf);
-	vq->vq_ops->kick(vq);
+	ret = virtqueue_add_buf(vq, sg, 0, 1, buf);
+	virtqueue_kick(vq);
 	return ret;
 }
 
@@ -366,7 +344,7 @@ static void discard_port_data(struct port *port)
 	if (port->inbuf)
 		buf = port->inbuf;
 	else
-		buf = vq->vq_ops->get_buf(vq, &len);
+		buf = virtqueue_get_buf(vq, &len);
 
 	ret = 0;
 	while (buf) {
@@ -374,7 +352,7 @@ static void discard_port_data(struct port *port)
 			ret++;
 			free_buf(buf);
 		}
-		buf = vq->vq_ops->get_buf(vq, &len);
+		buf = virtqueue_get_buf(vq, &len);
 	}
 	port->inbuf = NULL;
 	if (ret)
@@ -403,57 +381,96 @@ out:
 	return ret;
 }
 
-static ssize_t send_control_msg(struct port *port, unsigned int event,
-				unsigned int value)
+static ssize_t __send_control_msg(struct ports_device *portdev, u32 port_id,
+				  unsigned int event, unsigned int value)
 {
 	struct scatterlist sg[1];
 	struct virtio_console_control cpkt;
 	struct virtqueue *vq;
 	unsigned int len;
 
-	if (!use_multiport(port->portdev))
+	if (!use_multiport(portdev))
 		return 0;
 
-	cpkt.id = port->id;
+	cpkt.id = port_id;
 	cpkt.event = event;
 	cpkt.value = value;
 
-	vq = port->portdev->c_ovq;
+	vq = portdev->c_ovq;
 
 	sg_init_one(sg, &cpkt, sizeof(cpkt));
-	if (vq->vq_ops->add_buf(vq, sg, 1, 0, &cpkt) >= 0) {
-		vq->vq_ops->kick(vq);
-		while (!vq->vq_ops->get_buf(vq, &len))
+	if (virtqueue_add_buf(vq, sg, 1, 0, &cpkt) >= 0) {
+		virtqueue_kick(vq);
+		while (!virtqueue_get_buf(vq, &len))
 			cpu_relax();
 	}
 	return 0;
 }
 
-static ssize_t send_buf(struct port *port, void *in_buf, size_t in_count)
+static ssize_t send_control_msg(struct port *port, unsigned int event,
+				unsigned int value)
+{
+	return __send_control_msg(port->portdev, port->id, event, value);
+}
+
+/* Callers must take the port->outvq_lock */
+static void reclaim_consumed_buffers(struct port *port)
+{
+	void *buf;
+	unsigned int len;
+
+	while ((buf = virtqueue_get_buf(port->out_vq, &len))) {
+		kfree(buf);
+		port->outvq_full = false;
+	}
+}
+
+static ssize_t send_buf(struct port *port, void *in_buf, size_t in_count,
+			bool nonblock)
 {
 	struct scatterlist sg[1];
 	struct virtqueue *out_vq;
 	ssize_t ret;
+	unsigned long flags;
 	unsigned int len;
 
 	out_vq = port->out_vq;
 
+	spin_lock_irqsave(&port->outvq_lock, flags);
+
+	reclaim_consumed_buffers(port);
+
 	sg_init_one(sg, in_buf, in_count);
-	ret = out_vq->vq_ops->add_buf(out_vq, sg, 1, 0, in_buf);
+	ret = virtqueue_add_buf(out_vq, sg, 1, 0, in_buf);
 
 	/* Tell Host to go! */
-	out_vq->vq_ops->kick(out_vq);
+	virtqueue_kick(out_vq);
 
 	if (ret < 0) {
 		in_count = 0;
-		goto fail;
+		goto done;
 	}
 
-	/* Wait till the host acknowledges it pushed out the data we sent. */
-	while (!out_vq->vq_ops->get_buf(out_vq, &len))
+	if (ret == 0)
+		port->outvq_full = true;
+
+	if (nonblock)
+		goto done;
+
+	/*
+	 * Wait till the host acknowledges it pushed out the data we
+	 * sent.  This is done for ports in blocking mode or for data
+	 * from the hvc_console; the tty operations are performed with
+	 * spinlocks held so we can't sleep here.
+	 */
+	while (!virtqueue_get_buf(out_vq, &len))
 		cpu_relax();
-fail:
-	/* We're expected to return the amount of data we wrote */
+done:
+	spin_unlock_irqrestore(&port->outvq_lock, flags);
+	/*
+	 * We're expected to return the amount of data we wrote -- all
+	 * of it
+	 */
 	return in_count;
 }
 
@@ -503,9 +520,32 @@ static ssize_t fill_readbuf(struct port *port, char *out_buf, size_t out_count,
 }
 
 /* The condition that must be true for polling to end */
-static bool wait_is_over(struct port *port)
+static bool will_read_block(struct port *port)
 {
-	return port_has_data(port) || !port->host_connected;
+	return !port_has_data(port) && port->host_connected;
+}
+
+static bool will_write_block(struct port *port)
+{
+	bool ret;
+
+	if (!port->guest_connected) {
+		/* Port got hot-unplugged. Let's exit. */
+		return false;
+	}
+	if (!port->host_connected)
+		return true;
+
+	spin_lock_irq(&port->outvq_lock);
+	/*
+	 * Check if the Host has consumed any buffers since we last
+	 * sent data (this is only applicable for nonblocking ports).
+	 */
+	reclaim_consumed_buffers(port);
+	ret = port->outvq_full;
+	spin_unlock_irq(&port->outvq_lock);
+
+	return ret;
 }
 
 static ssize_t port_fops_read(struct file *filp, char __user *ubuf,
@@ -528,7 +568,7 @@ static ssize_t port_fops_read(struct file *filp, char __user *ubuf,
 			return -EAGAIN;
 
 		ret = wait_event_interruptible(port->waitqueue,
-					       wait_is_over(port));
+					       !will_read_block(port));
 		if (ret < 0)
 			return ret;
 	}
@@ -554,9 +594,22 @@ static ssize_t port_fops_write(struct file *filp, const char __user *ubuf,
 	struct port *port;
 	char *buf;
 	ssize_t ret;
+	bool nonblock;
 
 	port = filp->private_data;
 
+	nonblock = filp->f_flags & O_NONBLOCK;
+
+	if (will_write_block(port)) {
+		if (nonblock)
+			return -EAGAIN;
+
+		ret = wait_event_interruptible(port->waitqueue,
+					       !will_write_block(port));
+		if (ret < 0)
+			return ret;
+	}
+
 	count = min((size_t)(32 * 1024), count);
 
 	buf = kmalloc(count, GFP_KERNEL);
@@ -569,9 +622,14 @@ static ssize_t port_fops_write(struct file *filp, const char __user *ubuf,
 		goto free_buf;
 	}
 
-	ret = send_buf(port, buf, count);
+	ret = send_buf(port, buf, count, nonblock);
+
+	if (nonblock && ret > 0)
+		goto out;
+
 free_buf:
 	kfree(buf);
+out:
 	return ret;
 }
 
@@ -586,7 +644,7 @@ static unsigned int port_fops_poll(struct file *filp, poll_table *wait)
 	ret = 0;
 	if (port->inbuf)
 		ret |= POLLIN | POLLRDNORM;
-	if (port->host_connected)
+	if (!will_write_block(port))
 		ret |= POLLOUT;
 	if (!port->host_connected)
 		ret |= POLLHUP;
@@ -610,6 +668,10 @@ static int port_fops_release(struct inode *inode, struct file *filp)
 
 	spin_unlock_irq(&port->inbuf_lock);
 
+	spin_lock_irq(&port->outvq_lock);
+	reclaim_consumed_buffers(port);
+	spin_unlock_irq(&port->outvq_lock);
+
 	return 0;
 }
 
@@ -638,6 +700,15 @@ static int port_fops_open(struct inode *inode, struct file *filp)
 	port->guest_connected = true;
 	spin_unlock_irq(&port->inbuf_lock);
 
+	spin_lock_irq(&port->outvq_lock);
+	/*
+	 * There might be a chance that we missed reclaiming a few
+	 * buffers in the window of the port getting previously closed
+	 * and opening now.
+	 */
+	reclaim_consumed_buffers(port);
+	spin_unlock_irq(&port->outvq_lock);
+
 	/* Notify host of port being opened */
 	send_control_msg(filp->private_data, VIRTIO_CONSOLE_PORT_OPEN, 1);
 
@@ -676,9 +747,9 @@ static int put_chars(u32 vtermno, const char *buf, int count)
 
 	port = find_port_by_vtermno(vtermno);
 	if (!port)
-		return 0;
+		return -EPIPE;
 
-	return send_buf(port, (void *)buf, count);
+	return send_buf(port, (void *)buf, count, false);
 }
 
 /*
@@ -692,9 +763,13 @@ static int get_chars(u32 vtermno, char *buf, int count)
 {
 	struct port *port;
 
+	/* If we've not set up the port yet, we have no input to give. */
+	if (unlikely(early_put_chars))
+		return 0;
+
 	port = find_port_by_vtermno(vtermno);
 	if (!port)
-		return 0;
+		return -EPIPE;
 
 	/* If we don't have an input queue yet, we can't get input. */
 	BUG_ON(!port->in_vq);
@@ -705,22 +780,14 @@ static int get_chars(u32 vtermno, char *buf, int count)
 static void resize_console(struct port *port)
 {
 	struct virtio_device *vdev;
-	struct winsize ws;
 
 	/* The port could have been hot-unplugged */
-	if (!port)
+	if (!port || !is_console_port(port))
 		return;
 
 	vdev = port->portdev->vdev;
-	if (virtio_has_feature(vdev, VIRTIO_CONSOLE_F_SIZE)) {
-		vdev->config->get(vdev,
-				  offsetof(struct virtio_console_config, cols),
-				  &ws.ws_col, sizeof(u16));
-		vdev->config->get(vdev,
-				  offsetof(struct virtio_console_config, rows),
-				  &ws.ws_row, sizeof(u16));
-		hvc_resize(port->cons.hvc, ws);
-	}
+	if (virtio_has_feature(vdev, VIRTIO_CONSOLE_F_SIZE))
+		hvc_resize(port->cons.hvc, port->cons.ws);
 }
 
 /* We set the configuration at this point, since we now have a tty */
@@ -804,6 +871,13 @@ int init_port_console(struct port *port)
 	spin_unlock_irq(&pdrvdata_lock);
 	port->guest_connected = true;
 
+	/*
+	 * Start using the new console output if this is the first
+	 * console to come up.
+	 */
+	if (early_put_chars)
+		early_put_chars = NULL;
+
 	/* Notify host of port being opened */
 	send_control_msg(port, VIRTIO_CONSOLE_PORT_OPEN, 1);
 
@@ -859,6 +933,8 @@ static ssize_t debugfs_read(struct file *filp, char __user *ubuf,
 	out_offset += snprintf(buf + out_offset, out_count - out_offset,
 			       "host_connected: %d\n", port->host_connected);
 	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+			       "outvq_full: %d\n", port->outvq_full);
+	out_offset += snprintf(buf + out_offset, out_count - out_offset,
 			       "is_console: %s\n",
 			       is_console_port(port) ? "yes" : "no");
 	out_offset += snprintf(buf + out_offset, out_count - out_offset,
@@ -875,11 +951,165 @@ static const struct file_operations port_debugfs_ops = {
 	.read  = debugfs_read,
 };
 
+static void set_console_size(struct port *port, u16 rows, u16 cols)
+{
+	if (!port || !is_console_port(port))
+		return;
+
+	port->cons.ws.ws_row = rows;
+	port->cons.ws.ws_col = cols;
+}
+
+static unsigned int fill_queue(struct virtqueue *vq, spinlock_t *lock)
+{
+	struct port_buffer *buf;
+	unsigned int nr_added_bufs;
+	int ret;
+
+	nr_added_bufs = 0;
+	do {
+		buf = alloc_buf(PAGE_SIZE);
+		if (!buf)
+			break;
+
+		spin_lock_irq(lock);
+		ret = add_inbuf(vq, buf);
+		if (ret < 0) {
+			spin_unlock_irq(lock);
+			free_buf(buf);
+			break;
+		}
+		nr_added_bufs++;
+		spin_unlock_irq(lock);
+	} while (ret > 0);
+
+	return nr_added_bufs;
+}
+
+static int add_port(struct ports_device *portdev, u32 id)
+{
+	char debugfs_name[16];
+	struct port *port;
+	struct port_buffer *buf;
+	dev_t devt;
+	unsigned int nr_added_bufs;
+	int err;
+
+	port = kmalloc(sizeof(*port), GFP_KERNEL);
+	if (!port) {
+		err = -ENOMEM;
+		goto fail;
+	}
+
+	port->portdev = portdev;
+	port->id = id;
+
+	port->name = NULL;
+	port->inbuf = NULL;
+	port->cons.hvc = NULL;
+
+	port->cons.ws.ws_row = port->cons.ws.ws_col = 0;
+
+	port->host_connected = port->guest_connected = false;
+
+	port->outvq_full = false;
+
+	port->in_vq = portdev->in_vqs[port->id];
+	port->out_vq = portdev->out_vqs[port->id];
+
+	cdev_init(&port->cdev, &port_fops);
+
+	devt = MKDEV(portdev->chr_major, id);
+	err = cdev_add(&port->cdev, devt, 1);
+	if (err < 0) {
+		dev_err(&port->portdev->vdev->dev,
+			"Error %d adding cdev for port %u\n", err, id);
+		goto free_port;
+	}
+	port->dev = device_create(pdrvdata.class, &port->portdev->vdev->dev,
+				  devt, port, "vport%up%u",
+				  port->portdev->drv_index, id);
+	if (IS_ERR(port->dev)) {
+		err = PTR_ERR(port->dev);
+		dev_err(&port->portdev->vdev->dev,
+			"Error %d creating device for port %u\n",
+			err, id);
+		goto free_cdev;
+	}
+
+	spin_lock_init(&port->inbuf_lock);
+	spin_lock_init(&port->outvq_lock);
+	init_waitqueue_head(&port->waitqueue);
+
+	/* Fill the in_vq with buffers so the host can send us data. */
+	nr_added_bufs = fill_queue(port->in_vq, &port->inbuf_lock);
+	if (!nr_added_bufs) {
+		dev_err(port->dev, "Error allocating inbufs\n");
+		err = -ENOMEM;
+		goto free_device;
+	}
+
+	/*
+	 * If we're not using multiport support, this has to be a console port
+	 */
+	if (!use_multiport(port->portdev)) {
+		err = init_port_console(port);
+		if (err)
+			goto free_inbufs;
+	}
+
+	spin_lock_irq(&portdev->ports_lock);
+	list_add_tail(&port->list, &port->portdev->ports);
+	spin_unlock_irq(&portdev->ports_lock);
+
+	/*
+	 * Tell the Host we're set so that it can send us various
+	 * configuration parameters for this port (eg, port name,
+	 * caching, whether this is a console port, etc.)
+	 */
+	send_control_msg(port, VIRTIO_CONSOLE_PORT_READY, 1);
+
+	if (pdrvdata.debugfs_dir) {
+		/*
+		 * Finally, create the debugfs file that we can use to
+		 * inspect a port's state at any time
+		 */
+		sprintf(debugfs_name, "vport%up%u",
+			port->portdev->drv_index, id);
+		port->debugfs_file = debugfs_create_file(debugfs_name, 0444,
+							 pdrvdata.debugfs_dir,
+							 port,
+							 &port_debugfs_ops);
+	}
+	return 0;
+
+free_inbufs:
+	while ((buf = virtqueue_detach_unused_buf(port->in_vq)))
+		free_buf(buf);
+free_device:
+	device_destroy(pdrvdata.class, port->dev->devt);
+free_cdev:
+	cdev_del(&port->cdev);
+free_port:
+	kfree(port);
+fail:
+	/* The host might want to notify management sw about port add failure */
+	__send_control_msg(portdev, id, VIRTIO_CONSOLE_PORT_READY, 0);
+	return err;
+}
+
 /* Remove all port-specific data. */
 static int remove_port(struct port *port)
 {
 	struct port_buffer *buf;
 
+	if (port->guest_connected) {
+		port->guest_connected = false;
+		port->host_connected = false;
+		wake_up_interruptible(&port->waitqueue);
+		send_control_msg(port, VIRTIO_CONSOLE_PORT_OPEN, 0);
+	}
+
 	spin_lock_irq(&port->portdev->ports_lock);
 	list_del(&port->list);
 	spin_unlock_irq(&port->portdev->ports_lock);
@@ -888,11 +1118,19 @@ static int remove_port(struct port *port)
 		spin_lock_irq(&pdrvdata_lock);
 		list_del(&port->cons.list);
 		spin_unlock_irq(&pdrvdata_lock);
+#if 0
+		/*
+		 * hvc_remove() not called as removing one hvc port
+		 * results in other hvc ports getting frozen.
+		 *
+		 * Once this is resolved in hvc, this functionality
+		 * will be enabled.  Till that is done, the -EPIPE
+		 * return from get_chars() above will help
+		 * hvc_console.c to clean up on ports we remove here.
+		 */
 		hvc_remove(port->cons.hvc);
+#endif
 	}
-	if (port->guest_connected)
-		send_control_msg(port, VIRTIO_CONSOLE_PORT_OPEN, 0);
-
 	sysfs_remove_group(&port->dev->kobj, &port_attribute_group);
 	device_destroy(pdrvdata.class, port->dev->devt);
 	cdev_del(&port->cdev);
@@ -900,8 +1138,10 @@ static int remove_port(struct port *port)
 	/* Remove unused data this port might have received. */
 	discard_port_data(port);
 
+	reclaim_consumed_buffers(port);
+
 	/* Remove buffers we queued up for the Host to send us data in. */
-	while ((buf = port->in_vq->vq_ops->detach_unused_buf(port->in_vq)))
+	while ((buf = virtqueue_detach_unused_buf(port->in_vq)))
 		free_buf(buf);
 
 	kfree(port->name);
@@ -924,7 +1164,7 @@ static void handle_control_message(struct ports_device *portdev,
 	cpkt = (struct virtio_console_control *)(buf->buf + buf->offset);
 
 	port = find_port_by_id(portdev, cpkt->id);
-	if (!port) {
+	if (!port && cpkt->event != VIRTIO_CONSOLE_PORT_ADD) {
 		/* No valid header at start of buffer.  Drop it. */
 		dev_dbg(&portdev->vdev->dev,
 			"Invalid index %u in control packet\n", cpkt->id);
@@ -932,6 +1172,24 @@ static void handle_control_message(struct ports_device *portdev,
 	}
 
 	switch (cpkt->event) {
+	case VIRTIO_CONSOLE_PORT_ADD:
+		if (port) {
+			dev_dbg(&portdev->vdev->dev,
+				"Port %u already added\n", port->id);
+			send_control_msg(port, VIRTIO_CONSOLE_PORT_READY, 1);
+			break;
+		}
+		if (cpkt->id >= portdev->config.max_nr_ports) {
+			dev_warn(&portdev->vdev->dev,
+				"Request for adding port with out-of-bound id %u, max. supported id: %u\n",
+				cpkt->id, portdev->config.max_nr_ports - 1);
+			break;
+		}
+		add_port(portdev, cpkt->id);
+		break;
+	case VIRTIO_CONSOLE_PORT_REMOVE:
+		remove_port(port);
+		break;
 	case VIRTIO_CONSOLE_CONSOLE_PORT:
 		if (!cpkt->value)
 			break;
@@ -944,15 +1202,34 @@ static void handle_control_message(struct ports_device *portdev,
 		 * have to notify the host first.
 		 */
 		break;
-	case VIRTIO_CONSOLE_RESIZE:
+	case VIRTIO_CONSOLE_RESIZE: {
+		struct {
+			__u16 rows;
+			__u16 cols;
+		} size;
+
 		if (!is_console_port(port))
 			break;
+
+		memcpy(&size, buf->buf + buf->offset + sizeof(*cpkt),
+		       sizeof(size));
+		set_console_size(port, size.rows, size.cols);
+
 		port->cons.hvc->irq_requested = 1;
 		resize_console(port);
 		break;
+	}
 	case VIRTIO_CONSOLE_PORT_OPEN:
 		port->host_connected = cpkt->value;
 		wake_up_interruptible(&port->waitqueue);
+		/*
+		 * If the host port got closed and the host had any
+		 * unconsumed buffers, we'll be able to reclaim them
+		 * now.
+		 */
+		spin_lock_irq(&port->outvq_lock);
+		reclaim_consumed_buffers(port);
+		spin_unlock_irq(&port->outvq_lock);
 		break;
 	case VIRTIO_CONSOLE_PORT_NAME:
 		/*
@@ -990,32 +1267,6 @@ static void handle_control_message(struct ports_device *portdev,
 			kobject_uevent(&port->dev->kobj, KOBJ_CHANGE);
 		}
 		break;
-	case VIRTIO_CONSOLE_PORT_REMOVE:
-		/*
-		 * Hot unplug the port.  We don't decrement nr_ports
-		 * since we don't want to deal with extra complexities
-		 * of using the lowest-available port id: We can just
-		 * pick up the nr_ports number as the id and not have
-		 * userspace send it to us.  This helps us in two
-		 * ways:
-		 *
-		 * - We don't need to have a 'port_id' field in the
-		 *   config space when a port is hot-added.  This is a
-		 *   good thing as we might queue up multiple hotplug
-		 *   requests issued in our workqueue.
-		 *
-		 * - Another way to deal with this would have been to
-		 *   use a bitmap of the active ports and select the
-		 *   lowest non-active port from that map.  That
-		 *   bloats the already tight config space and we
-		 *   would end up artificially limiting the
-		 *   max. number of ports to sizeof(bitmap).  Right
-		 *   now we can support 2^32 ports (as the port id is
-		 *   stored in a u32 type).
-		 *
-		 */
-		remove_port(port);
-		break;
 	}
 }
 
@@ -1030,7 +1281,7 @@ static void control_work_handler(struct work_struct *work)
 	vq = portdev->c_ivq;
 
 	spin_lock(&portdev->cvq_lock);
-	while ((buf = vq->vq_ops->get_buf(vq, &len))) {
+	while ((buf = virtqueue_get_buf(vq, &len))) {
 		spin_unlock(&portdev->cvq_lock);
 
 		buf->len = len;
@@ -1092,204 +1343,29 @@ static void config_intr(struct virtio_device *vdev)
 	struct ports_device *portdev;
 
 	portdev = vdev->priv;
-	if (use_multiport(portdev)) {
-		/* Handle port hot-add */
-		schedule_work(&portdev->config_work);
-	}
-	/*
-	 * We'll use this way of resizing only for legacy support.
-	 * For newer userspace (VIRTIO_CONSOLE_F_MULTPORT+), use
-	 * control messages to indicate console size changes so that
-	 * it can be done per-port
-	 */
-	resize_console(find_port_by_id(portdev, 0));
-}
-
-static unsigned int fill_queue(struct virtqueue *vq, spinlock_t *lock)
-{
-	struct port_buffer *buf;
-	unsigned int nr_added_bufs;
-	int ret;
-
-	nr_added_bufs = 0;
-	do {
-		buf = alloc_buf(PAGE_SIZE);
-		if (!buf)
-			break;
-
-		spin_lock_irq(lock);
-		ret = add_inbuf(vq, buf);
-		if (ret < 0) {
-			spin_unlock_irq(lock);
-			free_buf(buf);
-			break;
-		}
-		nr_added_bufs++;
-		spin_unlock_irq(lock);
-	} while (ret > 0);
-
-	return nr_added_bufs;
-}
-
-static int add_port(struct ports_device *portdev, u32 id)
-{
-	char debugfs_name[16];
-	struct port *port;
-	struct port_buffer *buf;
-	dev_t devt;
-	unsigned int nr_added_bufs;
-	int err;
-
-	port = kmalloc(sizeof(*port), GFP_KERNEL);
-	if (!port) {
-		err = -ENOMEM;
-		goto fail;
-	}
-
-	port->portdev = portdev;
-	port->id = id;
-
-	port->name = NULL;
-	port->inbuf = NULL;
-	port->cons.hvc = NULL;
-
-	port->host_connected = port->guest_connected = false;
-
-	port->in_vq = portdev->in_vqs[port->id];
-	port->out_vq = portdev->out_vqs[port->id];
-
-	cdev_init(&port->cdev, &port_fops);
-
-	devt = MKDEV(portdev->chr_major, id);
-	err = cdev_add(&port->cdev, devt, 1);
-	if (err < 0) {
-		dev_err(&port->portdev->vdev->dev,
-			"Error %d adding cdev for port %u\n", err, id);
-		goto free_port;
-	}
-	port->dev = device_create(pdrvdata.class, &port->portdev->vdev->dev,
-				  devt, port, "vport%up%u",
-				  port->portdev->drv_index, id);
-	if (IS_ERR(port->dev)) {
-		err = PTR_ERR(port->dev);
-		dev_err(&port->portdev->vdev->dev,
-			"Error %d creating device for port %u\n",
-			err, id);
-		goto free_cdev;
-	}
-
-	spin_lock_init(&port->inbuf_lock);
-	init_waitqueue_head(&port->waitqueue);
-
-	/* Fill the in_vq with buffers so the host can send us data. */
-	nr_added_bufs = fill_queue(port->in_vq, &port->inbuf_lock);
-	if (!nr_added_bufs) {
-		dev_err(port->dev, "Error allocating inbufs\n");
-		err = -ENOMEM;
-		goto free_device;
-	}
-
-	/*
-	 * If we're not using multiport support, this has to be a console port
-	 */
-	if (!use_multiport(port->portdev)) {
-		err = init_port_console(port);
-		if (err)
-			goto free_inbufs;
-	}
-
-	spin_lock_irq(&portdev->ports_lock);
-	list_add_tail(&port->list, &port->portdev->ports);
-	spin_unlock_irq(&portdev->ports_lock);
-
-	/*
-	 * Tell the Host we're set so that it can send us various
-	 * configuration parameters for this port (eg, port name,
-	 * caching, whether this is a console port, etc.)
-	 */
-	send_control_msg(port, VIRTIO_CONSOLE_PORT_READY, 1);
-
-	if (pdrvdata.debugfs_dir) {
-		/*
-		 * Finally, create the debugfs file that we can use to
-		 * inspect a port's state at any time
-		 */
-		sprintf(debugfs_name, "vport%up%u",
-			port->portdev->drv_index, id);
-		port->debugfs_file = debugfs_create_file(debugfs_name, 0444,
-							 pdrvdata.debugfs_dir,
-							 port,
-							 &port_debugfs_ops);
-	}
-	return 0;
 
-free_inbufs:
-	while ((buf = port->in_vq->vq_ops->detach_unused_buf(port->in_vq)))
-		free_buf(buf);
-free_device:
-	device_destroy(pdrvdata.class, port->dev->devt);
-free_cdev:
-	cdev_del(&port->cdev);
-free_port:
-	kfree(port);
-fail:
-	return err;
-}
-
-/*
- * The workhandler for config-space updates.
- *
- * This is called when ports are hot-added.
- */
-static void config_work_handler(struct work_struct *work)
-{
-	struct virtio_console_multiport_conf virtconconf;
-	struct ports_device *portdev;
-	struct virtio_device *vdev;
-	int err;
+	if (!use_multiport(portdev)) {
+		struct port *port;
+		u16 rows, cols;
 
-	portdev = container_of(work, struct ports_device, config_work);
+		vdev->config->get(vdev,
+				  offsetof(struct virtio_console_config, cols),
+				  &cols, sizeof(u16));
+		vdev->config->get(vdev,
+				  offsetof(struct virtio_console_config, rows),
+				  &rows, sizeof(u16));
 
-	vdev = portdev->vdev;
-	vdev->config->get(vdev,
-			  offsetof(struct virtio_console_multiport_conf,
-				   nr_ports),
-			  &virtconconf.nr_ports,
-			  sizeof(virtconconf.nr_ports));
+		port = find_port_by_id(portdev, 0);
+		set_console_size(port, rows, cols);
 
-	if (portdev->config.nr_ports == virtconconf.nr_ports) {
 		/*
-		 * Port 0 got hot-added.  Since we already did all the
-		 * other initialisation for it, just tell the Host
-		 * that the port is ready if we find the port.  In
-		 * case the port was hot-removed earlier, we call
-		 * add_port to add the port.
+		 * We'll use this way of resizing only for legacy
+		 * support.  For newer userspace
+		 * (VIRTIO_CONSOLE_F_MULTPORT+), use control messages
+		 * to indicate console size changes so that it can be
+		 * done per-port.
 		 */
-		struct port *port;
-
-		port = find_port_by_id(portdev, 0);
-		if (!port)
-			add_port(portdev, 0);
-		else
-			send_control_msg(port, VIRTIO_CONSOLE_PORT_READY, 1);
-		return;
-	}
-	if (virtconconf.nr_ports > portdev->config.max_nr_ports) {
-		dev_warn(&vdev->dev,
-			 "More ports specified (%u) than allowed (%u)",
-			 portdev->config.nr_ports + 1,
-			 portdev->config.max_nr_ports);
-		return;
-	}
-	if (virtconconf.nr_ports < portdev->config.nr_ports)
-		return;
-
-	/* Hot-add ports */
-	while (virtconconf.nr_ports - portdev->config.nr_ports) {
-		err = add_port(portdev, portdev->config.nr_ports);
-		if (err)
-			break;
-		portdev->config.nr_ports++;
+		resize_console(port);
 	}
 }
 
@@ -1414,7 +1490,6 @@ static const struct file_operations portdev_fops = {
 static int __devinit virtcons_probe(struct virtio_device *vdev)
 {
 	struct ports_device *portdev;
-	u32 i;
 	int err;
 	bool multiport;
 
@@ -1443,37 +1518,19 @@ static int __devinit virtcons_probe(struct virtio_device *vdev)
 	}
 
 	multiport = false;
-	portdev->config.nr_ports = 1;
 	portdev->config.max_nr_ports = 1;
-#if 0 /* Multiport is not quite ready yet --RR */
 	if (virtio_has_feature(vdev, VIRTIO_CONSOLE_F_MULTIPORT)) {
 		multiport = true;
 		vdev->features[0] |= 1 << VIRTIO_CONSOLE_F_MULTIPORT;
 
-		vdev->config->get(vdev,
-				  offsetof(struct virtio_console_multiport_conf,
-					   nr_ports),
-				  &portdev->config.nr_ports,
-				  sizeof(portdev->config.nr_ports));
-		vdev->config->get(vdev,
-				  offsetof(struct virtio_console_multiport_conf,
-					   max_nr_ports),
+		vdev->config->get(vdev, offsetof(struct virtio_console_config,
+						 max_nr_ports),
 				  &portdev->config.max_nr_ports,
 				  sizeof(portdev->config.max_nr_ports));
-		if (portdev->config.nr_ports > portdev->config.max_nr_ports) {
-			dev_warn(&vdev->dev,
-				 "More ports (%u) specified than allowed (%u). Will init %u ports.",
-				 portdev->config.nr_ports,
-				 portdev->config.max_nr_ports,
-				 portdev->config.max_nr_ports);
-
-			portdev->config.nr_ports = portdev->config.max_nr_ports;
-		}
 	}
 
 	/* Let the Host know we support multiple ports.*/
 	vdev->config->finalize_features(vdev);
-#endif
 
 	err = init_vqs(portdev);
 	if (err < 0) {
@@ -1489,7 +1546,6 @@ static int __devinit virtcons_probe(struct virtio_device *vdev)
 
 		spin_lock_init(&portdev->cvq_lock);
 		INIT_WORK(&portdev->control_work, &control_work_handler);
-		INIT_WORK(&portdev->config_work, &config_work_handler);
 
 		nr_added_bufs = fill_queue(portdev->c_ivq, &portdev->cvq_lock);
 		if (!nr_added_bufs) {
@@ -1498,16 +1554,22 @@ static int __devinit virtcons_probe(struct virtio_device *vdev)
 			err = -ENOMEM;
 			goto free_vqs;
 		}
+	} else {
+		/*
+		 * For backward compatibility: Create a console port
+		 * if we're running on older host.
+		 */
+		add_port(portdev, 0);
 	}
 
-	for (i = 0; i < portdev->config.nr_ports; i++)
-		add_port(portdev, i);
-
-	/* Start using the new console output. */
-	early_put_chars = NULL;
+	__send_control_msg(portdev, VIRTIO_CONSOLE_BAD_ID,
+			   VIRTIO_CONSOLE_DEVICE_READY, 1);
 	return 0;
 
 free_vqs:
+	/* The host might want to notify mgmt sw about device add failure */
+	__send_control_msg(portdev, VIRTIO_CONSOLE_BAD_ID,
+			   VIRTIO_CONSOLE_DEVICE_READY, 0);
 	vdev->config->del_vqs(vdev);
 	kfree(portdev->in_vqs);
 	kfree(portdev->out_vqs);
@@ -1529,17 +1591,16 @@ static void virtcons_remove(struct virtio_device *vdev)
 	portdev = vdev->priv;
 
 	cancel_work_sync(&portdev->control_work);
-	cancel_work_sync(&portdev->config_work);
 
 	list_for_each_entry_safe(port, port2, &portdev->ports, list)
 		remove_port(port);
 
 	unregister_chrdev(portdev->chr_major, "virtio-portsdev");
 
-	while ((buf = portdev->c_ivq->vq_ops->get_buf(portdev->c_ivq, &len)))
+	while ((buf = virtqueue_get_buf(portdev->c_ivq, &len)))
 		free_buf(buf);
 
-	while ((buf = portdev->c_ivq->vq_ops->detach_unused_buf(portdev->c_ivq)))
+	while ((buf = virtqueue_detach_unused_buf(portdev->c_ivq)))
 		free_buf(buf);
 
 	vdev->config->del_vqs(vdev);
@@ -1556,6 +1617,7 @@ static struct virtio_device_id id_table[] = {
 
 static unsigned int features[] = {
 	VIRTIO_CONSOLE_F_SIZE,
+	VIRTIO_CONSOLE_F_MULTIPORT,
 };
 
 static struct virtio_driver virtio_console = {
diff --git a/drivers/char/vt.c b/drivers/char/vt.c
index bd1d1164fec5..7cdb6ee569cd 100644
--- a/drivers/char/vt.c
+++ b/drivers/char/vt.c
@@ -3967,13 +3967,9 @@ static int con_font_set(struct vc_data *vc, struct console_font_op *op)
 	font.charcount = op->charcount;
 	font.height = op->height;
 	font.width = op->width;
-	font.data = kmalloc(size, GFP_KERNEL);
-	if (!font.data)
-		return -ENOMEM;
-	if (copy_from_user(font.data, op->data, size)) {
-		kfree(font.data);
-		return -EFAULT;
-	}
+	font.data = memdup_user(op->data, size);
+	if (IS_ERR(font.data))
+		return PTR_ERR(font.data);
 	acquire_console_sem();
 	if (vc->vc_sw->con_font_set)
 		rc = vc->vc_sw->con_font_set(vc, &font, op->flags);
diff --git a/drivers/char/vt_ioctl.c b/drivers/char/vt_ioctl.c
index 6aa10284104a..cb19dbc52136 100644
--- a/drivers/char/vt_ioctl.c
+++ b/drivers/char/vt_ioctl.c
@@ -1303,7 +1303,9 @@ int vt_ioctl(struct tty_struct *tty, struct file * file,
 		if (!perm)
 			goto eperm;
 		ret = copy_from_user(&ui, up, sizeof(struct unimapinit));
-		if (!ret)
+		if (ret)
+			ret = -EFAULT;
+		else
 			con_clear_unimap(vc, &ui);
 		break;
 	      }
diff --git a/drivers/char/xilinx_hwicap/xilinx_hwicap.c b/drivers/char/xilinx_hwicap/xilinx_hwicap.c
index 7261b8d9087c..ed8a9cec2a05 100644
--- a/drivers/char/xilinx_hwicap/xilinx_hwicap.c
+++ b/drivers/char/xilinx_hwicap/xilinx_hwicap.c
@@ -772,18 +772,18 @@ hwicap_of_probe(struct of_device *op, const struct of_device_id *match)
 
 	dev_dbg(&op->dev, "hwicap_of_probe(%p, %p)\n", op, match);
 
-	rc = of_address_to_resource(op->node, 0, &res);
+	rc = of_address_to_resource(op->dev.of_node, 0, &res);
 	if (rc) {
 		dev_err(&op->dev, "invalid address\n");
 		return rc;
 	}
 
-	id = of_get_property(op->node, "port-number", NULL);
+	id = of_get_property(op->dev.of_node, "port-number", NULL);
 
 	/* It's most likely that we're using V4, if the family is not
 	   specified */
 	regs = &v4_config_registers;
-	family = of_get_property(op->node, "xlnx,family", NULL);
+	family = of_get_property(op->dev.of_node, "xlnx,family", NULL);
 
 	if (family) {
 		if (!strcmp(family, "virtex2p")) {
@@ -812,13 +812,12 @@ static const struct of_device_id __devinitconst hwicap_of_match[] = {
 MODULE_DEVICE_TABLE(of, hwicap_of_match);
 
 static struct of_platform_driver hwicap_of_driver = {
-	.owner = THIS_MODULE,
-	.name = DRIVER_NAME,
-	.match_table = hwicap_of_match,
 	.probe = hwicap_of_probe,
 	.remove = __devexit_p(hwicap_of_remove),
 	.driver = {
 		.name = DRIVER_NAME,
+		.owner = THIS_MODULE,
+		.of_match_table = hwicap_of_match,
 	},
 };
 
diff --git a/drivers/clocksource/sh_cmt.c b/drivers/clocksource/sh_cmt.c
index f6677cb19789..f3d3898898ed 100644
--- a/drivers/clocksource/sh_cmt.c
+++ b/drivers/clocksource/sh_cmt.c
@@ -412,18 +412,10 @@ static cycle_t sh_cmt_clocksource_read(struct clocksource *cs)
 static int sh_cmt_clocksource_enable(struct clocksource *cs)
 {
 	struct sh_cmt_priv *p = cs_to_sh_cmt(cs);
-	int ret;
 
 	p->total_cycles = 0;
 
-	ret = sh_cmt_start(p, FLAG_CLOCKSOURCE);
-	if (ret)
-		return ret;
-
-	/* TODO: calculate good shift from rate and counter bit width */
-	cs->shift = 0;
-	cs->mult = clocksource_hz2mult(p->rate, cs->shift);
-	return 0;
+	return sh_cmt_start(p, FLAG_CLOCKSOURCE);
 }
 
 static void sh_cmt_clocksource_disable(struct clocksource *cs)
@@ -450,8 +442,20 @@ static int sh_cmt_register_clocksource(struct sh_cmt_priv *p,
 	cs->resume = sh_cmt_clocksource_resume;
 	cs->mask = CLOCKSOURCE_MASK(sizeof(unsigned long) * 8);
 	cs->flags = CLOCK_SOURCE_IS_CONTINUOUS;
+
+	/* clk_get_rate() needs an enabled clock */
+	clk_enable(p->clk);
+	p->rate = clk_get_rate(p->clk) / (p->width == 16) ? 512 : 8;
+	clk_disable(p->clk);
+
+	/* TODO: calculate good shift from rate and counter bit width */
+	cs->shift = 10;
+	cs->mult = clocksource_hz2mult(p->rate, cs->shift);
+
 	dev_info(&p->pdev->dev, "used as clock source\n");
+
 	clocksource_register(cs);
+
 	return 0;
 }
 
diff --git a/drivers/clocksource/sh_tmu.c b/drivers/clocksource/sh_tmu.c
index 8e44e14ec4c2..de715901b82a 100644
--- a/drivers/clocksource/sh_tmu.c
+++ b/drivers/clocksource/sh_tmu.c
@@ -199,16 +199,8 @@ static cycle_t sh_tmu_clocksource_read(struct clocksource *cs)
 static int sh_tmu_clocksource_enable(struct clocksource *cs)
 {
 	struct sh_tmu_priv *p = cs_to_sh_tmu(cs);
-	int ret;
-
-	ret = sh_tmu_enable(p);
-	if (ret)
-		return ret;
 
-	/* TODO: calculate good shift from rate and counter bit width */
-	cs->shift = 10;
-	cs->mult = clocksource_hz2mult(p->rate, cs->shift);
-	return 0;
+	return sh_tmu_enable(p);
 }
 
 static void sh_tmu_clocksource_disable(struct clocksource *cs)
@@ -228,6 +220,16 @@ static int sh_tmu_register_clocksource(struct sh_tmu_priv *p,
 	cs->disable = sh_tmu_clocksource_disable;
 	cs->mask = CLOCKSOURCE_MASK(32);
 	cs->flags = CLOCK_SOURCE_IS_CONTINUOUS;
+
+	/* clk_get_rate() needs an enabled clock */
+	clk_enable(p->clk);
+	/* channel will be configured at parent clock / 4 */
+	p->rate = clk_get_rate(p->clk) / 4;
+	clk_disable(p->clk);
+	/* TODO: calculate good shift from rate and counter bit width */
+	cs->shift = 10;
+	cs->mult = clocksource_hz2mult(p->rate, cs->shift);
+
 	dev_info(&p->pdev->dev, "used as clock source\n");
 	clocksource_register(cs);
 	return 0;
diff --git a/drivers/cpuidle/cpuidle.c b/drivers/cpuidle/cpuidle.c
index 12fdd3987a36..199488576a05 100644
--- a/drivers/cpuidle/cpuidle.c
+++ b/drivers/cpuidle/cpuidle.c
@@ -156,7 +156,7 @@ int cpuidle_enable_device(struct cpuidle_device *dev)
 
 	if (dev->enabled)
 		return 0;
-	if (!cpuidle_curr_driver || !cpuidle_curr_governor)
+	if (!cpuidle_get_driver() || !cpuidle_curr_governor)
 		return -EIO;
 	if (!dev->state_count)
 		return -EINVAL;
@@ -207,7 +207,7 @@ void cpuidle_disable_device(struct cpuidle_device *dev)
 {
 	if (!dev->enabled)
 		return;
-	if (!cpuidle_curr_driver || !cpuidle_curr_governor)
+	if (!cpuidle_get_driver() || !cpuidle_curr_governor)
 		return;
 
 	dev->enabled = 0;
@@ -271,10 +271,11 @@ static int __cpuidle_register_device(struct cpuidle_device *dev)
 {
 	int ret;
 	struct sys_device *sys_dev = get_cpu_sysdev((unsigned long)dev->cpu);
+	struct cpuidle_driver *cpuidle_driver = cpuidle_get_driver();
 
 	if (!sys_dev)
 		return -EINVAL;
-	if (!try_module_get(cpuidle_curr_driver->owner))
+	if (!try_module_get(cpuidle_driver->owner))
 		return -EINVAL;
 
 	init_completion(&dev->kobj_unregister);
@@ -284,7 +285,7 @@ static int __cpuidle_register_device(struct cpuidle_device *dev)
 	per_cpu(cpuidle_devices, dev->cpu) = dev;
 	list_add(&dev->device_list, &cpuidle_detected_devices);
 	if ((ret = cpuidle_add_sysfs(sys_dev))) {
-		module_put(cpuidle_curr_driver->owner);
+		module_put(cpuidle_driver->owner);
 		return ret;
 	}
 
@@ -325,6 +326,7 @@ EXPORT_SYMBOL_GPL(cpuidle_register_device);
 void cpuidle_unregister_device(struct cpuidle_device *dev)
 {
 	struct sys_device *sys_dev = get_cpu_sysdev((unsigned long)dev->cpu);
+	struct cpuidle_driver *cpuidle_driver = cpuidle_get_driver();
 
 	if (dev->registered == 0)
 		return;
@@ -340,7 +342,7 @@ void cpuidle_unregister_device(struct cpuidle_device *dev)
 
 	cpuidle_resume_and_unlock();
 
-	module_put(cpuidle_curr_driver->owner);
+	module_put(cpuidle_driver->owner);
 }
 
 EXPORT_SYMBOL_GPL(cpuidle_unregister_device);
diff --git a/drivers/cpuidle/cpuidle.h b/drivers/cpuidle/cpuidle.h
index 9476ba33ee2c..33e50d556f17 100644
--- a/drivers/cpuidle/cpuidle.h
+++ b/drivers/cpuidle/cpuidle.h
@@ -9,7 +9,6 @@
 
 /* For internal use only */
 extern struct cpuidle_governor *cpuidle_curr_governor;
-extern struct cpuidle_driver *cpuidle_curr_driver;
 extern struct list_head cpuidle_governors;
 extern struct list_head cpuidle_detected_devices;
 extern struct mutex cpuidle_lock;
diff --git a/drivers/cpuidle/driver.c b/drivers/cpuidle/driver.c
index 2257004fe33d..fd1601e3d125 100644
--- a/drivers/cpuidle/driver.c
+++ b/drivers/cpuidle/driver.c
@@ -14,7 +14,7 @@
 
 #include "cpuidle.h"
 
-struct cpuidle_driver *cpuidle_curr_driver;
+static struct cpuidle_driver *cpuidle_curr_driver;
 DEFINE_SPINLOCK(cpuidle_driver_lock);
 
 /**
@@ -40,13 +40,25 @@ int cpuidle_register_driver(struct cpuidle_driver *drv)
 EXPORT_SYMBOL_GPL(cpuidle_register_driver);
 
 /**
+ * cpuidle_get_driver - return the current driver
+ */
+struct cpuidle_driver *cpuidle_get_driver(void)
+{
+	return cpuidle_curr_driver;
+}
+EXPORT_SYMBOL_GPL(cpuidle_get_driver);
+
+/**
  * cpuidle_unregister_driver - unregisters a driver
  * @drv: the driver
  */
 void cpuidle_unregister_driver(struct cpuidle_driver *drv)
 {
-	if (!drv)
+	if (drv != cpuidle_curr_driver) {
+		WARN(1, "invalid cpuidle_unregister_driver(%s)\n",
+			drv->name);
 		return;
+	}
 
 	spin_lock(&cpuidle_driver_lock);
 	cpuidle_curr_driver = NULL;
diff --git a/drivers/cpuidle/governors/menu.c b/drivers/cpuidle/governors/menu.c
index b81ad9c731ae..52ff8aa63f84 100644
--- a/drivers/cpuidle/governors/menu.c
+++ b/drivers/cpuidle/governors/menu.c
@@ -21,9 +21,12 @@
 #include <linux/math64.h>
 
 #define BUCKETS 12
+#define INTERVALS 8
 #define RESOLUTION 1024
-#define DECAY 4
+#define DECAY 8
 #define MAX_INTERESTING 50000
+#define STDDEV_THRESH 400
+
 
 /*
  * Concepts and ideas behind the menu governor
@@ -64,6 +67,16 @@
  * indexed based on the magnitude of the expected duration as well as the
  * "is IO outstanding" property.
  *
+ * Repeatable-interval-detector
+ * ----------------------------
+ * There are some cases where "next timer" is a completely unusable predictor:
+ * Those cases where the interval is fixed, for example due to hardware
+ * interrupt mitigation, but also due to fixed transfer rate devices such as
+ * mice.
+ * For this, we use a different predictor: We track the duration of the last 8
+ * intervals and if the stand deviation of these 8 intervals is below a
+ * threshold value, we use the average of these intervals as prediction.
+ *
  * Limiting Performance Impact
  * ---------------------------
  * C states, especially those with large exit latencies, can have a real
@@ -104,6 +117,8 @@ struct menu_device {
 	unsigned int	exit_us;
 	unsigned int	bucket;
 	u64		correction_factor[BUCKETS];
+	u32		intervals[INTERVALS];
+	int		interval_ptr;
 };
 
 
@@ -175,6 +190,42 @@ static u64 div_round64(u64 dividend, u32 divisor)
 	return div_u64(dividend + (divisor / 2), divisor);
 }
 
+/*
+ * Try detecting repeating patterns by keeping track of the last 8
+ * intervals, and checking if the standard deviation of that set
+ * of points is below a threshold. If it is... then use the
+ * average of these 8 points as the estimated value.
+ */
+static void detect_repeating_patterns(struct menu_device *data)
+{
+	int i;
+	uint64_t avg = 0;
+	uint64_t stddev = 0; /* contains the square of the std deviation */
+
+	/* first calculate average and standard deviation of the past */
+	for (i = 0; i < INTERVALS; i++)
+		avg += data->intervals[i];
+	avg = avg / INTERVALS;
+
+	/* if the avg is beyond the known next tick, it's worthless */
+	if (avg > data->expected_us)
+		return;
+
+	for (i = 0; i < INTERVALS; i++)
+		stddev += (data->intervals[i] - avg) *
+			  (data->intervals[i] - avg);
+
+	stddev = stddev / INTERVALS;
+
+	/*
+	 * now.. if stddev is small.. then assume we have a
+	 * repeating pattern and predict we keep doing this.
+	 */
+
+	if (avg && stddev < STDDEV_THRESH)
+		data->predicted_us = avg;
+}
+
 /**
  * menu_select - selects the next idle state to enter
  * @dev: the CPU
@@ -218,6 +269,8 @@ static int menu_select(struct cpuidle_device *dev)
 	data->predicted_us = div_round64(data->expected_us * data->correction_factor[data->bucket],
 					 RESOLUTION * DECAY);
 
+	detect_repeating_patterns(data);
+
 	/*
 	 * We want to default to C1 (hlt), not to busy polling
 	 * unless the timer is happening really really soon.
@@ -310,6 +363,11 @@ static void menu_update(struct cpuidle_device *dev)
 		new_factor = 1;
 
 	data->correction_factor[data->bucket] = new_factor;
+
+	/* update the repeating-pattern data */
+	data->intervals[data->interval_ptr++] = last_idle_us;
+	if (data->interval_ptr >= INTERVALS)
+		data->interval_ptr = 0;
 }
 
 /**
diff --git a/drivers/cpuidle/sysfs.c b/drivers/cpuidle/sysfs.c
index 0ba9c8b8ee74..0310ffaec9df 100644
--- a/drivers/cpuidle/sysfs.c
+++ b/drivers/cpuidle/sysfs.c
@@ -47,10 +47,11 @@ static ssize_t show_current_driver(struct sysdev_class *class,
 				   char *buf)
 {
 	ssize_t ret;
+	struct cpuidle_driver *cpuidle_driver = cpuidle_get_driver();
 
 	spin_lock(&cpuidle_driver_lock);
-	if (cpuidle_curr_driver)
-		ret = sprintf(buf, "%s\n", cpuidle_curr_driver->name);
+	if (cpuidle_driver)
+		ret = sprintf(buf, "%s\n", cpuidle_driver->name);
 	else
 		ret = sprintf(buf, "none\n");
 	spin_unlock(&cpuidle_driver_lock);
diff --git a/drivers/crypto/Kconfig b/drivers/crypto/Kconfig
index b08403d7d1ca..fbf94cf496f0 100644
--- a/drivers/crypto/Kconfig
+++ b/drivers/crypto/Kconfig
@@ -170,6 +170,18 @@ config CRYPTO_DEV_MV_CESA
 
 	  Currently the driver supports AES in ECB and CBC mode without DMA.
 
+config CRYPTO_DEV_NIAGARA2
+       tristate "Niagara2 Stream Processing Unit driver"
+       select CRYPTO_ALGAPI
+       depends on SPARC64
+       help
+	  Each core of a Niagara2 processor contains a Stream
+	  Processing Unit, which itself contains several cryptographic
+	  sub-units.  One set provides the Modular Arithmetic Unit,
+	  used for SSL offload.  The other set provides the Cipher
+	  Group, which can perform encryption, decryption, hashing,
+	  checksumming, and raw copies.
+
 config CRYPTO_DEV_HIFN_795X
 	tristate "Driver HIFN 795x crypto accelerator chips"
 	select CRYPTO_DES
@@ -222,4 +234,13 @@ config CRYPTO_DEV_PPC4XX
 	help
 	  This option allows you to have support for AMCC crypto acceleration.
 
+config CRYPTO_DEV_OMAP_SHAM
+	tristate "Support for OMAP SHA1/MD5 hw accelerator"
+	depends on ARCH_OMAP2 || ARCH_OMAP3
+	select CRYPTO_SHA1
+	select CRYPTO_MD5
+	help
+	  OMAP processors have SHA1/MD5 hw accelerator. Select this if you
+	  want to use the OMAP module for SHA1/MD5 algorithms.
+
 endif # CRYPTO_HW
diff --git a/drivers/crypto/Makefile b/drivers/crypto/Makefile
index 6ffcb3f7f942..6dbbe00c4524 100644
--- a/drivers/crypto/Makefile
+++ b/drivers/crypto/Makefile
@@ -1,8 +1,12 @@
 obj-$(CONFIG_CRYPTO_DEV_PADLOCK_AES) += padlock-aes.o
 obj-$(CONFIG_CRYPTO_DEV_PADLOCK_SHA) += padlock-sha.o
 obj-$(CONFIG_CRYPTO_DEV_GEODE) += geode-aes.o
+obj-$(CONFIG_CRYPTO_DEV_NIAGARA2) += n2_crypto.o
+n2_crypto-objs := n2_core.o n2_asm.o
 obj-$(CONFIG_CRYPTO_DEV_HIFN_795X) += hifn_795x.o
 obj-$(CONFIG_CRYPTO_DEV_MV_CESA) += mv_cesa.o
 obj-$(CONFIG_CRYPTO_DEV_TALITOS) += talitos.o
 obj-$(CONFIG_CRYPTO_DEV_IXP4XX) += ixp4xx_crypto.o
 obj-$(CONFIG_CRYPTO_DEV_PPC4XX) += amcc/
+obj-$(CONFIG_CRYPTO_DEV_OMAP_SHAM) += omap-sham.o
+
diff --git a/drivers/crypto/amcc/crypto4xx_core.c b/drivers/crypto/amcc/crypto4xx_core.c
index 6c4c8b7ce3aa..983530ba04a7 100644
--- a/drivers/crypto/amcc/crypto4xx_core.c
+++ b/drivers/crypto/amcc/crypto4xx_core.c
@@ -1158,7 +1158,7 @@ static int __init crypto4xx_probe(struct of_device *ofdev,
 	struct device *dev = &ofdev->dev;
 	struct crypto4xx_core_device *core_dev;
 
-	rc = of_address_to_resource(ofdev->node, 0, &res);
+	rc = of_address_to_resource(ofdev->dev.of_node, 0, &res);
 	if (rc)
 		return -ENODEV;
 
@@ -1215,13 +1215,13 @@ static int __init crypto4xx_probe(struct of_device *ofdev,
 		     (unsigned long) dev);
 
 	/* Register for Crypto isr, Crypto Engine IRQ */
-	core_dev->irq = irq_of_parse_and_map(ofdev->node, 0);
+	core_dev->irq = irq_of_parse_and_map(ofdev->dev.of_node, 0);
 	rc = request_irq(core_dev->irq, crypto4xx_ce_interrupt_handler, 0,
 			 core_dev->dev->name, dev);
 	if (rc)
 		goto err_request_irq;
 
-	core_dev->dev->ce_base = of_iomap(ofdev->node, 0);
+	core_dev->dev->ce_base = of_iomap(ofdev->dev.of_node, 0);
 	if (!core_dev->dev->ce_base) {
 		dev_err(dev, "failed to of_iomap\n");
 		goto err_iomap;
@@ -1281,8 +1281,11 @@ static const struct of_device_id crypto4xx_match[] = {
 };
 
 static struct of_platform_driver crypto4xx_driver = {
-	.name		= "crypto4xx",
-	.match_table	= crypto4xx_match,
+	.driver = {
+		.name = "crypto4xx",
+		.owner = THIS_MODULE,
+		.of_match_table = crypto4xx_match,
+	},
 	.probe		= crypto4xx_probe,
 	.remove		= crypto4xx_remove,
 };
diff --git a/drivers/crypto/geode-aes.c b/drivers/crypto/geode-aes.c
index c7a5a43ba691..09389dd2f96b 100644
--- a/drivers/crypto/geode-aes.c
+++ b/drivers/crypto/geode-aes.c
@@ -15,14 +15,14 @@
 #include <crypto/algapi.h>
 #include <crypto/aes.h>
 
-#include <asm/io.h>
-#include <asm/delay.h>
+#include <linux/io.h>
+#include <linux/delay.h>
 
 #include "geode-aes.h"
 
 /* Static structures */
 
-static void __iomem * _iobase;
+static void __iomem *_iobase;
 static spinlock_t lock;
 
 /* Write a 128 bit field (either a writable key or IV) */
@@ -30,7 +30,7 @@ static inline void
 _writefield(u32 offset, void *value)
 {
 	int i;
-	for(i = 0; i < 4; i++)
+	for (i = 0; i < 4; i++)
 		iowrite32(((u32 *) value)[i], _iobase + offset + (i * 4));
 }
 
@@ -39,7 +39,7 @@ static inline void
 _readfield(u32 offset, void *value)
 {
 	int i;
-	for(i = 0; i < 4; i++)
+	for (i = 0; i < 4; i++)
 		((u32 *) value)[i] = ioread32(_iobase + offset + (i * 4));
 }
 
@@ -59,7 +59,7 @@ do_crypt(void *src, void *dst, int len, u32 flags)
 	do {
 		status = ioread32(_iobase + AES_INTR_REG);
 		cpu_relax();
-	} while(!(status & AES_INTRA_PENDING) && --counter);
+	} while (!(status & AES_INTRA_PENDING) && --counter);
 
 	/* Clear the event */
 	iowrite32((status & 0xFF) | AES_INTRA_PENDING, _iobase + AES_INTR_REG);
@@ -317,7 +317,7 @@ geode_cbc_decrypt(struct blkcipher_desc *desc,
 	err = blkcipher_walk_virt(desc, &walk);
 	op->iv = walk.iv;
 
-	while((nbytes = walk.nbytes)) {
+	while ((nbytes = walk.nbytes)) {
 		op->src = walk.src.virt.addr,
 		op->dst = walk.dst.virt.addr;
 		op->mode = AES_MODE_CBC;
@@ -349,7 +349,7 @@ geode_cbc_encrypt(struct blkcipher_desc *desc,
 	err = blkcipher_walk_virt(desc, &walk);
 	op->iv = walk.iv;
 
-	while((nbytes = walk.nbytes)) {
+	while ((nbytes = walk.nbytes)) {
 		op->src = walk.src.virt.addr,
 		op->dst = walk.dst.virt.addr;
 		op->mode = AES_MODE_CBC;
@@ -429,7 +429,7 @@ geode_ecb_decrypt(struct blkcipher_desc *desc,
 	blkcipher_walk_init(&walk, dst, src, nbytes);
 	err = blkcipher_walk_virt(desc, &walk);
 
-	while((nbytes = walk.nbytes)) {
+	while ((nbytes = walk.nbytes)) {
 		op->src = walk.src.virt.addr,
 		op->dst = walk.dst.virt.addr;
 		op->mode = AES_MODE_ECB;
@@ -459,7 +459,7 @@ geode_ecb_encrypt(struct blkcipher_desc *desc,
 	blkcipher_walk_init(&walk, dst, src, nbytes);
 	err = blkcipher_walk_virt(desc, &walk);
 
-	while((nbytes = walk.nbytes)) {
+	while ((nbytes = walk.nbytes)) {
 		op->src = walk.src.virt.addr,
 		op->dst = walk.dst.virt.addr;
 		op->mode = AES_MODE_ECB;
@@ -518,11 +518,12 @@ static int __devinit
 geode_aes_probe(struct pci_dev *dev, const struct pci_device_id *id)
 {
 	int ret;
-
-	if ((ret = pci_enable_device(dev)))
+	ret = pci_enable_device(dev);
+	if (ret)
 		return ret;
 
-	if ((ret = pci_request_regions(dev, "geode-aes")))
+	ret = pci_request_regions(dev, "geode-aes");
+	if (ret)
 		goto eenable;
 
 	_iobase = pci_iomap(dev, 0, 0);
@@ -537,13 +538,16 @@ geode_aes_probe(struct pci_dev *dev, const struct pci_device_id *id)
 	/* Clear any pending activity */
 	iowrite32(AES_INTR_PENDING | AES_INTR_MASK, _iobase + AES_INTR_REG);
 
-	if ((ret = crypto_register_alg(&geode_alg)))
+	ret = crypto_register_alg(&geode_alg);
+	if (ret)
 		goto eiomap;
 
-	if ((ret = crypto_register_alg(&geode_ecb_alg)))
+	ret = crypto_register_alg(&geode_ecb_alg);
+	if (ret)
 		goto ealg;
 
-	if ((ret = crypto_register_alg(&geode_cbc_alg)))
+	ret = crypto_register_alg(&geode_cbc_alg);
+	if (ret)
 		goto eecb;
 
 	printk(KERN_NOTICE "geode-aes: GEODE AES engine enabled.\n");
diff --git a/drivers/crypto/hifn_795x.c b/drivers/crypto/hifn_795x.c
index 73e8b1713b54..16fce3aadf4d 100644
--- a/drivers/crypto/hifn_795x.c
+++ b/drivers/crypto/hifn_795x.c
@@ -638,7 +638,7 @@ struct hifn_crypto_alg
 
 #define ASYNC_FLAGS_MISALIGNED	(1<<0)
 
-struct ablkcipher_walk
+struct hifn_cipher_walk
 {
 	struct scatterlist	cache[ASYNC_SCATTERLIST_CACHE];
 	u32			flags;
@@ -657,7 +657,7 @@ struct hifn_request_context
 	u8			*iv;
 	unsigned int		ivsize;
 	u8			op, type, mode, unused;
-	struct ablkcipher_walk	walk;
+	struct hifn_cipher_walk	walk;
 };
 
 #define crypto_alg_to_hifn(a)	container_of(a, struct hifn_crypto_alg, alg)
@@ -1417,7 +1417,7 @@ static int hifn_setup_dma(struct hifn_device *dev,
 	return 0;
 }
 
-static int ablkcipher_walk_init(struct ablkcipher_walk *w,
+static int hifn_cipher_walk_init(struct hifn_cipher_walk *w,
 		int num, gfp_t gfp_flags)
 {
 	int i;
@@ -1442,7 +1442,7 @@ static int ablkcipher_walk_init(struct ablkcipher_walk *w,
 	return i;
 }
 
-static void ablkcipher_walk_exit(struct ablkcipher_walk *w)
+static void hifn_cipher_walk_exit(struct hifn_cipher_walk *w)
 {
 	int i;
 
@@ -1486,8 +1486,8 @@ static int ablkcipher_add(unsigned int *drestp, struct scatterlist *dst,
 	return idx;
 }
 
-static int ablkcipher_walk(struct ablkcipher_request *req,
-		struct ablkcipher_walk *w)
+static int hifn_cipher_walk(struct ablkcipher_request *req,
+		struct hifn_cipher_walk *w)
 {
 	struct scatterlist *dst, *t;
 	unsigned int nbytes = req->nbytes, offset, copy, diff;
@@ -1600,12 +1600,12 @@ static int hifn_setup_session(struct ablkcipher_request *req)
 	}
 
 	if (rctx->walk.flags & ASYNC_FLAGS_MISALIGNED) {
-		err = ablkcipher_walk_init(&rctx->walk, idx, GFP_ATOMIC);
+		err = hifn_cipher_walk_init(&rctx->walk, idx, GFP_ATOMIC);
 		if (err < 0)
 			return err;
 	}
 
-	sg_num = ablkcipher_walk(req, &rctx->walk);
+	sg_num = hifn_cipher_walk(req, &rctx->walk);
 	if (sg_num < 0) {
 		err = sg_num;
 		goto err_out_exit;
@@ -1806,7 +1806,7 @@ static void hifn_process_ready(struct ablkcipher_request *req, int error)
 			kunmap_atomic(saddr, KM_SOFTIRQ0);
 		}
 
-		ablkcipher_walk_exit(&rctx->walk);
+		hifn_cipher_walk_exit(&rctx->walk);
 	}
 
 	req->base.complete(&req->base, error);
diff --git a/drivers/crypto/mv_cesa.c b/drivers/crypto/mv_cesa.c
index 6f29012bcc43..e095422b58dd 100644
--- a/drivers/crypto/mv_cesa.c
+++ b/drivers/crypto/mv_cesa.c
@@ -15,8 +15,14 @@
 #include <linux/platform_device.h>
 #include <linux/scatterlist.h>
 #include <linux/slab.h>
+#include <crypto/internal/hash.h>
+#include <crypto/sha.h>
 
 #include "mv_cesa.h"
+
+#define MV_CESA	"MV-CESA:"
+#define MAX_HW_HASH_SIZE	0xFFFF
+
 /*
  * STM:
  *   /---------------------------------------\
@@ -39,10 +45,12 @@ enum engine_status {
  * @dst_sg_it:		sg iterator for dst
  * @sg_src_left:	bytes left in src to process (scatter list)
  * @src_start:		offset to add to src start position (scatter list)
- * @crypt_len:		length of current crypt process
+ * @crypt_len:		length of current hw crypt/hash process
+ * @hw_nbytes:		total bytes to process in hw for this request
+ * @copy_back:		whether to copy data back (crypt) or not (hash)
  * @sg_dst_left:	bytes left dst to process in this scatter list
  * @dst_start:		offset to add to dst start position (scatter list)
- * @total_req_bytes:	total number of bytes processed (request).
+ * @hw_processed_bytes:	number of bytes processed by hw (request).
  *
  * sg helper are used to iterate over the scatterlist. Since the size of the
  * SRAM may be less than the scatter size, this struct struct is used to keep
@@ -51,15 +59,19 @@ enum engine_status {
 struct req_progress {
 	struct sg_mapping_iter src_sg_it;
 	struct sg_mapping_iter dst_sg_it;
+	void (*complete) (void);
+	void (*process) (int is_first);
 
 	/* src mostly */
 	int sg_src_left;
 	int src_start;
 	int crypt_len;
+	int hw_nbytes;
 	/* dst mostly */
+	int copy_back;
 	int sg_dst_left;
 	int dst_start;
-	int total_req_bytes;
+	int hw_processed_bytes;
 };
 
 struct crypto_priv {
@@ -72,10 +84,12 @@ struct crypto_priv {
 	spinlock_t lock;
 	struct crypto_queue queue;
 	enum engine_status eng_st;
-	struct ablkcipher_request *cur_req;
+	struct crypto_async_request *cur_req;
 	struct req_progress p;
 	int max_req_size;
 	int sram_size;
+	int has_sha1;
+	int has_hmac_sha1;
 };
 
 static struct crypto_priv *cpg;
@@ -97,6 +111,31 @@ struct mv_req_ctx {
 	int decrypt;
 };
 
+enum hash_op {
+	COP_SHA1,
+	COP_HMAC_SHA1
+};
+
+struct mv_tfm_hash_ctx {
+	struct crypto_shash *fallback;
+	struct crypto_shash *base_hash;
+	u32 ivs[2 * SHA1_DIGEST_SIZE / 4];
+	int count_add;
+	enum hash_op op;
+};
+
+struct mv_req_hash_ctx {
+	u64 count;
+	u32 state[SHA1_DIGEST_SIZE / 4];
+	u8 buffer[SHA1_BLOCK_SIZE];
+	int first_hash;		/* marks that we don't have previous state */
+	int last_chunk;		/* marks that this is the 'final' request */
+	int extra_bytes;	/* unprocessed bytes in buffer */
+	enum hash_op op;
+	int count_add;
+	struct scatterlist dummysg;
+};
+
 static void compute_aes_dec_key(struct mv_ctx *ctx)
 {
 	struct crypto_aes_ctx gen_aes_key;
@@ -144,32 +183,51 @@ static int mv_setkey_aes(struct crypto_ablkcipher *cipher, const u8 *key,
 	return 0;
 }
 
-static void setup_data_in(struct ablkcipher_request *req)
+static void copy_src_to_buf(struct req_progress *p, char *dbuf, int len)
 {
 	int ret;
-	void *buf;
+	void *sbuf;
+	int copied = 0;
 
-	if (!cpg->p.sg_src_left) {
-		ret = sg_miter_next(&cpg->p.src_sg_it);
-		BUG_ON(!ret);
-		cpg->p.sg_src_left = cpg->p.src_sg_it.length;
-		cpg->p.src_start = 0;
-	}
-
-	cpg->p.crypt_len = min(cpg->p.sg_src_left, cpg->max_req_size);
-
-	buf = cpg->p.src_sg_it.addr;
-	buf += cpg->p.src_start;
+	while (1) {
+		if (!p->sg_src_left) {
+			ret = sg_miter_next(&p->src_sg_it);
+			BUG_ON(!ret);
+			p->sg_src_left = p->src_sg_it.length;
+			p->src_start = 0;
+		}
 
-	memcpy(cpg->sram + SRAM_DATA_IN_START, buf, cpg->p.crypt_len);
+		sbuf = p->src_sg_it.addr + p->src_start;
+
+		if (p->sg_src_left <= len - copied) {
+			memcpy(dbuf + copied, sbuf, p->sg_src_left);
+			copied += p->sg_src_left;
+			p->sg_src_left = 0;
+			if (copied >= len)
+				break;
+		} else {
+			int copy_len = len - copied;
+			memcpy(dbuf + copied, sbuf, copy_len);
+			p->src_start += copy_len;
+			p->sg_src_left -= copy_len;
+			break;
+		}
+	}
+}
 
-	cpg->p.sg_src_left -= cpg->p.crypt_len;
-	cpg->p.src_start += cpg->p.crypt_len;
+static void setup_data_in(void)
+{
+	struct req_progress *p = &cpg->p;
+	int data_in_sram =
+	    min(p->hw_nbytes - p->hw_processed_bytes, cpg->max_req_size);
+	copy_src_to_buf(p, cpg->sram + SRAM_DATA_IN_START + p->crypt_len,
+			data_in_sram - p->crypt_len);
+	p->crypt_len = data_in_sram;
 }
 
 static void mv_process_current_q(int first_block)
 {
-	struct ablkcipher_request *req = cpg->cur_req;
+	struct ablkcipher_request *req = ablkcipher_request_cast(cpg->cur_req);
 	struct mv_ctx *ctx = crypto_tfm_ctx(req->base.tfm);
 	struct mv_req_ctx *req_ctx = ablkcipher_request_ctx(req);
 	struct sec_accel_config op;
@@ -179,6 +237,7 @@ static void mv_process_current_q(int first_block)
 		op.config = CFG_OP_CRYPT_ONLY | CFG_ENCM_AES | CFG_ENC_MODE_ECB;
 		break;
 	case COP_AES_CBC:
+	default:
 		op.config = CFG_OP_CRYPT_ONLY | CFG_ENCM_AES | CFG_ENC_MODE_CBC;
 		op.enc_iv = ENC_IV_POINT(SRAM_DATA_IV) |
 			ENC_IV_BUF_POINT(SRAM_DATA_IV_BUF);
@@ -211,7 +270,7 @@ static void mv_process_current_q(int first_block)
 		ENC_P_DST(SRAM_DATA_OUT_START);
 	op.enc_key_p = SRAM_DATA_KEY_P;
 
-	setup_data_in(req);
+	setup_data_in();
 	op.enc_len = cpg->p.crypt_len;
 	memcpy(cpg->sram + SRAM_CONFIG, &op,
 			sizeof(struct sec_accel_config));
@@ -228,91 +287,294 @@ static void mv_process_current_q(int first_block)
 
 static void mv_crypto_algo_completion(void)
 {
-	struct ablkcipher_request *req = cpg->cur_req;
+	struct ablkcipher_request *req = ablkcipher_request_cast(cpg->cur_req);
 	struct mv_req_ctx *req_ctx = ablkcipher_request_ctx(req);
 
+	sg_miter_stop(&cpg->p.src_sg_it);
+	sg_miter_stop(&cpg->p.dst_sg_it);
+
 	if (req_ctx->op != COP_AES_CBC)
 		return ;
 
 	memcpy(req->info, cpg->sram + SRAM_DATA_IV_BUF, 16);
 }
 
+static void mv_process_hash_current(int first_block)
+{
+	struct ahash_request *req = ahash_request_cast(cpg->cur_req);
+	struct mv_req_hash_ctx *req_ctx = ahash_request_ctx(req);
+	struct req_progress *p = &cpg->p;
+	struct sec_accel_config op = { 0 };
+	int is_last;
+
+	switch (req_ctx->op) {
+	case COP_SHA1:
+	default:
+		op.config = CFG_OP_MAC_ONLY | CFG_MACM_SHA1;
+		break;
+	case COP_HMAC_SHA1:
+		op.config = CFG_OP_MAC_ONLY | CFG_MACM_HMAC_SHA1;
+		break;
+	}
+
+	op.mac_src_p =
+		MAC_SRC_DATA_P(SRAM_DATA_IN_START) | MAC_SRC_TOTAL_LEN((u32)
+		req_ctx->
+		count);
+
+	setup_data_in();
+
+	op.mac_digest =
+		MAC_DIGEST_P(SRAM_DIGEST_BUF) | MAC_FRAG_LEN(p->crypt_len);
+	op.mac_iv =
+		MAC_INNER_IV_P(SRAM_HMAC_IV_IN) |
+		MAC_OUTER_IV_P(SRAM_HMAC_IV_OUT);
+
+	is_last = req_ctx->last_chunk
+		&& (p->hw_processed_bytes + p->crypt_len >= p->hw_nbytes)
+		&& (req_ctx->count <= MAX_HW_HASH_SIZE);
+	if (req_ctx->first_hash) {
+		if (is_last)
+			op.config |= CFG_NOT_FRAG;
+		else
+			op.config |= CFG_FIRST_FRAG;
+
+		req_ctx->first_hash = 0;
+	} else {
+		if (is_last)
+			op.config |= CFG_LAST_FRAG;
+		else
+			op.config |= CFG_MID_FRAG;
+	}
+
+	memcpy(cpg->sram + SRAM_CONFIG, &op, sizeof(struct sec_accel_config));
+
+	writel(SRAM_CONFIG, cpg->reg + SEC_ACCEL_DESC_P0);
+	/* GO */
+	writel(SEC_CMD_EN_SEC_ACCL0, cpg->reg + SEC_ACCEL_CMD);
+
+	/*
+	* XXX: add timer if the interrupt does not occur for some mystery
+	* reason
+	*/
+}
+
+static inline int mv_hash_import_sha1_ctx(const struct mv_req_hash_ctx *ctx,
+					  struct shash_desc *desc)
+{
+	int i;
+	struct sha1_state shash_state;
+
+	shash_state.count = ctx->count + ctx->count_add;
+	for (i = 0; i < 5; i++)
+		shash_state.state[i] = ctx->state[i];
+	memcpy(shash_state.buffer, ctx->buffer, sizeof(shash_state.buffer));
+	return crypto_shash_import(desc, &shash_state);
+}
+
+static int mv_hash_final_fallback(struct ahash_request *req)
+{
+	const struct mv_tfm_hash_ctx *tfm_ctx = crypto_tfm_ctx(req->base.tfm);
+	struct mv_req_hash_ctx *req_ctx = ahash_request_ctx(req);
+	struct {
+		struct shash_desc shash;
+		char ctx[crypto_shash_descsize(tfm_ctx->fallback)];
+	} desc;
+	int rc;
+
+	desc.shash.tfm = tfm_ctx->fallback;
+	desc.shash.flags = CRYPTO_TFM_REQ_MAY_SLEEP;
+	if (unlikely(req_ctx->first_hash)) {
+		crypto_shash_init(&desc.shash);
+		crypto_shash_update(&desc.shash, req_ctx->buffer,
+				    req_ctx->extra_bytes);
+	} else {
+		/* only SHA1 for now....
+		 */
+		rc = mv_hash_import_sha1_ctx(req_ctx, &desc.shash);
+		if (rc)
+			goto out;
+	}
+	rc = crypto_shash_final(&desc.shash, req->result);
+out:
+	return rc;
+}
+
+static void mv_hash_algo_completion(void)
+{
+	struct ahash_request *req = ahash_request_cast(cpg->cur_req);
+	struct mv_req_hash_ctx *ctx = ahash_request_ctx(req);
+
+	if (ctx->extra_bytes)
+		copy_src_to_buf(&cpg->p, ctx->buffer, ctx->extra_bytes);
+	sg_miter_stop(&cpg->p.src_sg_it);
+
+	ctx->state[0] = readl(cpg->reg + DIGEST_INITIAL_VAL_A);
+	ctx->state[1] = readl(cpg->reg + DIGEST_INITIAL_VAL_B);
+	ctx->state[2] = readl(cpg->reg + DIGEST_INITIAL_VAL_C);
+	ctx->state[3] = readl(cpg->reg + DIGEST_INITIAL_VAL_D);
+	ctx->state[4] = readl(cpg->reg + DIGEST_INITIAL_VAL_E);
+
+	if (likely(ctx->last_chunk)) {
+		if (likely(ctx->count <= MAX_HW_HASH_SIZE)) {
+			memcpy(req->result, cpg->sram + SRAM_DIGEST_BUF,
+			       crypto_ahash_digestsize(crypto_ahash_reqtfm
+						       (req)));
+		} else
+			mv_hash_final_fallback(req);
+	}
+}
+
 static void dequeue_complete_req(void)
 {
-	struct ablkcipher_request *req = cpg->cur_req;
+	struct crypto_async_request *req = cpg->cur_req;
 	void *buf;
 	int ret;
+	cpg->p.hw_processed_bytes += cpg->p.crypt_len;
+	if (cpg->p.copy_back) {
+		int need_copy_len = cpg->p.crypt_len;
+		int sram_offset = 0;
+		do {
+			int dst_copy;
+
+			if (!cpg->p.sg_dst_left) {
+				ret = sg_miter_next(&cpg->p.dst_sg_it);
+				BUG_ON(!ret);
+				cpg->p.sg_dst_left = cpg->p.dst_sg_it.length;
+				cpg->p.dst_start = 0;
+			}
 
-	cpg->p.total_req_bytes += cpg->p.crypt_len;
-	do {
-		int dst_copy;
-
-		if (!cpg->p.sg_dst_left) {
-			ret = sg_miter_next(&cpg->p.dst_sg_it);
-			BUG_ON(!ret);
-			cpg->p.sg_dst_left = cpg->p.dst_sg_it.length;
-			cpg->p.dst_start = 0;
-		}
-
-		buf = cpg->p.dst_sg_it.addr;
-		buf += cpg->p.dst_start;
+			buf = cpg->p.dst_sg_it.addr;
+			buf += cpg->p.dst_start;
 
-		dst_copy = min(cpg->p.crypt_len, cpg->p.sg_dst_left);
+			dst_copy = min(need_copy_len, cpg->p.sg_dst_left);
 
-		memcpy(buf, cpg->sram + SRAM_DATA_OUT_START, dst_copy);
+			memcpy(buf,
+			       cpg->sram + SRAM_DATA_OUT_START + sram_offset,
+			       dst_copy);
+			sram_offset += dst_copy;
+			cpg->p.sg_dst_left -= dst_copy;
+			need_copy_len -= dst_copy;
+			cpg->p.dst_start += dst_copy;
+		} while (need_copy_len > 0);
+	}
 
-		cpg->p.sg_dst_left -= dst_copy;
-		cpg->p.crypt_len -= dst_copy;
-		cpg->p.dst_start += dst_copy;
-	} while (cpg->p.crypt_len > 0);
+	cpg->p.crypt_len = 0;
 
 	BUG_ON(cpg->eng_st != ENGINE_W_DEQUEUE);
-	if (cpg->p.total_req_bytes < req->nbytes) {
+	if (cpg->p.hw_processed_bytes < cpg->p.hw_nbytes) {
 		/* process next scatter list entry */
 		cpg->eng_st = ENGINE_BUSY;
-		mv_process_current_q(0);
+		cpg->p.process(0);
 	} else {
-		sg_miter_stop(&cpg->p.src_sg_it);
-		sg_miter_stop(&cpg->p.dst_sg_it);
-		mv_crypto_algo_completion();
+		cpg->p.complete();
 		cpg->eng_st = ENGINE_IDLE;
-		req->base.complete(&req->base, 0);
+		local_bh_disable();
+		req->complete(req, 0);
+		local_bh_enable();
 	}
 }
 
 static int count_sgs(struct scatterlist *sl, unsigned int total_bytes)
 {
 	int i = 0;
-
-	do {
-		total_bytes -= sl[i].length;
-		i++;
-
-	} while (total_bytes > 0);
+	size_t cur_len;
+
+	while (1) {
+		cur_len = sl[i].length;
+		++i;
+		if (total_bytes > cur_len)
+			total_bytes -= cur_len;
+		else
+			break;
+	}
 
 	return i;
 }
 
-static void mv_enqueue_new_req(struct ablkcipher_request *req)
+static void mv_start_new_crypt_req(struct ablkcipher_request *req)
 {
+	struct req_progress *p = &cpg->p;
 	int num_sgs;
 
-	cpg->cur_req = req;
-	memset(&cpg->p, 0, sizeof(struct req_progress));
+	cpg->cur_req = &req->base;
+	memset(p, 0, sizeof(struct req_progress));
+	p->hw_nbytes = req->nbytes;
+	p->complete = mv_crypto_algo_completion;
+	p->process = mv_process_current_q;
+	p->copy_back = 1;
 
 	num_sgs = count_sgs(req->src, req->nbytes);
-	sg_miter_start(&cpg->p.src_sg_it, req->src, num_sgs, SG_MITER_FROM_SG);
+	sg_miter_start(&p->src_sg_it, req->src, num_sgs, SG_MITER_FROM_SG);
 
 	num_sgs = count_sgs(req->dst, req->nbytes);
-	sg_miter_start(&cpg->p.dst_sg_it, req->dst, num_sgs, SG_MITER_TO_SG);
+	sg_miter_start(&p->dst_sg_it, req->dst, num_sgs, SG_MITER_TO_SG);
+
 	mv_process_current_q(1);
 }
 
+static void mv_start_new_hash_req(struct ahash_request *req)
+{
+	struct req_progress *p = &cpg->p;
+	struct mv_req_hash_ctx *ctx = ahash_request_ctx(req);
+	const struct mv_tfm_hash_ctx *tfm_ctx = crypto_tfm_ctx(req->base.tfm);
+	int num_sgs, hw_bytes, old_extra_bytes, rc;
+	cpg->cur_req = &req->base;
+	memset(p, 0, sizeof(struct req_progress));
+	hw_bytes = req->nbytes + ctx->extra_bytes;
+	old_extra_bytes = ctx->extra_bytes;
+
+	if (unlikely(ctx->extra_bytes)) {
+		memcpy(cpg->sram + SRAM_DATA_IN_START, ctx->buffer,
+		       ctx->extra_bytes);
+		p->crypt_len = ctx->extra_bytes;
+	}
+
+	memcpy(cpg->sram + SRAM_HMAC_IV_IN, tfm_ctx->ivs, sizeof(tfm_ctx->ivs));
+
+	if (unlikely(!ctx->first_hash)) {
+		writel(ctx->state[0], cpg->reg + DIGEST_INITIAL_VAL_A);
+		writel(ctx->state[1], cpg->reg + DIGEST_INITIAL_VAL_B);
+		writel(ctx->state[2], cpg->reg + DIGEST_INITIAL_VAL_C);
+		writel(ctx->state[3], cpg->reg + DIGEST_INITIAL_VAL_D);
+		writel(ctx->state[4], cpg->reg + DIGEST_INITIAL_VAL_E);
+	}
+
+	ctx->extra_bytes = hw_bytes % SHA1_BLOCK_SIZE;
+	if (ctx->extra_bytes != 0
+	    && (!ctx->last_chunk || ctx->count > MAX_HW_HASH_SIZE))
+		hw_bytes -= ctx->extra_bytes;
+	else
+		ctx->extra_bytes = 0;
+
+	num_sgs = count_sgs(req->src, req->nbytes);
+	sg_miter_start(&p->src_sg_it, req->src, num_sgs, SG_MITER_FROM_SG);
+
+	if (hw_bytes) {
+		p->hw_nbytes = hw_bytes;
+		p->complete = mv_hash_algo_completion;
+		p->process = mv_process_hash_current;
+
+		mv_process_hash_current(1);
+	} else {
+		copy_src_to_buf(p, ctx->buffer + old_extra_bytes,
+				ctx->extra_bytes - old_extra_bytes);
+		sg_miter_stop(&p->src_sg_it);
+		if (ctx->last_chunk)
+			rc = mv_hash_final_fallback(req);
+		else
+			rc = 0;
+		cpg->eng_st = ENGINE_IDLE;
+		local_bh_disable();
+		req->base.complete(&req->base, rc);
+		local_bh_enable();
+	}
+}
+
 static int queue_manag(void *data)
 {
 	cpg->eng_st = ENGINE_IDLE;
 	do {
-		struct ablkcipher_request *req;
 		struct crypto_async_request *async_req = NULL;
 		struct crypto_async_request *backlog;
 
@@ -338,9 +600,18 @@ static int queue_manag(void *data)
 		}
 
 		if (async_req) {
-			req = container_of(async_req,
-					struct ablkcipher_request, base);
-			mv_enqueue_new_req(req);
+			if (async_req->tfm->__crt_alg->cra_type !=
+			    &crypto_ahash_type) {
+				struct ablkcipher_request *req =
+				    container_of(async_req,
+						 struct ablkcipher_request,
+						 base);
+				mv_start_new_crypt_req(req);
+			} else {
+				struct ahash_request *req =
+				    ahash_request_cast(async_req);
+				mv_start_new_hash_req(req);
+			}
 			async_req = NULL;
 		}
 
@@ -350,13 +621,13 @@ static int queue_manag(void *data)
 	return 0;
 }
 
-static int mv_handle_req(struct ablkcipher_request *req)
+static int mv_handle_req(struct crypto_async_request *req)
 {
 	unsigned long flags;
 	int ret;
 
 	spin_lock_irqsave(&cpg->lock, flags);
-	ret = ablkcipher_enqueue_request(&cpg->queue, req);
+	ret = crypto_enqueue_request(&cpg->queue, req);
 	spin_unlock_irqrestore(&cpg->lock, flags);
 	wake_up_process(cpg->queue_th);
 	return ret;
@@ -369,7 +640,7 @@ static int mv_enc_aes_ecb(struct ablkcipher_request *req)
 	req_ctx->op = COP_AES_ECB;
 	req_ctx->decrypt = 0;
 
-	return mv_handle_req(req);
+	return mv_handle_req(&req->base);
 }
 
 static int mv_dec_aes_ecb(struct ablkcipher_request *req)
@@ -381,7 +652,7 @@ static int mv_dec_aes_ecb(struct ablkcipher_request *req)
 	req_ctx->decrypt = 1;
 
 	compute_aes_dec_key(ctx);
-	return mv_handle_req(req);
+	return mv_handle_req(&req->base);
 }
 
 static int mv_enc_aes_cbc(struct ablkcipher_request *req)
@@ -391,7 +662,7 @@ static int mv_enc_aes_cbc(struct ablkcipher_request *req)
 	req_ctx->op = COP_AES_CBC;
 	req_ctx->decrypt = 0;
 
-	return mv_handle_req(req);
+	return mv_handle_req(&req->base);
 }
 
 static int mv_dec_aes_cbc(struct ablkcipher_request *req)
@@ -403,7 +674,7 @@ static int mv_dec_aes_cbc(struct ablkcipher_request *req)
 	req_ctx->decrypt = 1;
 
 	compute_aes_dec_key(ctx);
-	return mv_handle_req(req);
+	return mv_handle_req(&req->base);
 }
 
 static int mv_cra_init(struct crypto_tfm *tfm)
@@ -412,6 +683,215 @@ static int mv_cra_init(struct crypto_tfm *tfm)
 	return 0;
 }
 
+static void mv_init_hash_req_ctx(struct mv_req_hash_ctx *ctx, int op,
+				 int is_last, unsigned int req_len,
+				 int count_add)
+{
+	memset(ctx, 0, sizeof(*ctx));
+	ctx->op = op;
+	ctx->count = req_len;
+	ctx->first_hash = 1;
+	ctx->last_chunk = is_last;
+	ctx->count_add = count_add;
+}
+
+static void mv_update_hash_req_ctx(struct mv_req_hash_ctx *ctx, int is_last,
+				   unsigned req_len)
+{
+	ctx->last_chunk = is_last;
+	ctx->count += req_len;
+}
+
+static int mv_hash_init(struct ahash_request *req)
+{
+	const struct mv_tfm_hash_ctx *tfm_ctx = crypto_tfm_ctx(req->base.tfm);
+	mv_init_hash_req_ctx(ahash_request_ctx(req), tfm_ctx->op, 0, 0,
+			     tfm_ctx->count_add);
+	return 0;
+}
+
+static int mv_hash_update(struct ahash_request *req)
+{
+	if (!req->nbytes)
+		return 0;
+
+	mv_update_hash_req_ctx(ahash_request_ctx(req), 0, req->nbytes);
+	return mv_handle_req(&req->base);
+}
+
+static int mv_hash_final(struct ahash_request *req)
+{
+	struct mv_req_hash_ctx *ctx = ahash_request_ctx(req);
+	/* dummy buffer of 4 bytes */
+	sg_init_one(&ctx->dummysg, ctx->buffer, 4);
+	/* I think I'm allowed to do that... */
+	ahash_request_set_crypt(req, &ctx->dummysg, req->result, 0);
+	mv_update_hash_req_ctx(ctx, 1, 0);
+	return mv_handle_req(&req->base);
+}
+
+static int mv_hash_finup(struct ahash_request *req)
+{
+	if (!req->nbytes)
+		return mv_hash_final(req);
+
+	mv_update_hash_req_ctx(ahash_request_ctx(req), 1, req->nbytes);
+	return mv_handle_req(&req->base);
+}
+
+static int mv_hash_digest(struct ahash_request *req)
+{
+	const struct mv_tfm_hash_ctx *tfm_ctx = crypto_tfm_ctx(req->base.tfm);
+	mv_init_hash_req_ctx(ahash_request_ctx(req), tfm_ctx->op, 1,
+			     req->nbytes, tfm_ctx->count_add);
+	return mv_handle_req(&req->base);
+}
+
+static void mv_hash_init_ivs(struct mv_tfm_hash_ctx *ctx, const void *istate,
+			     const void *ostate)
+{
+	const struct sha1_state *isha1_state = istate, *osha1_state = ostate;
+	int i;
+	for (i = 0; i < 5; i++) {
+		ctx->ivs[i] = cpu_to_be32(isha1_state->state[i]);
+		ctx->ivs[i + 5] = cpu_to_be32(osha1_state->state[i]);
+	}
+}
+
+static int mv_hash_setkey(struct crypto_ahash *tfm, const u8 * key,
+			  unsigned int keylen)
+{
+	int rc;
+	struct mv_tfm_hash_ctx *ctx = crypto_tfm_ctx(&tfm->base);
+	int bs, ds, ss;
+
+	if (!ctx->base_hash)
+		return 0;
+
+	rc = crypto_shash_setkey(ctx->fallback, key, keylen);
+	if (rc)
+		return rc;
+
+	/* Can't see a way to extract the ipad/opad from the fallback tfm
+	   so I'm basically copying code from the hmac module */
+	bs = crypto_shash_blocksize(ctx->base_hash);
+	ds = crypto_shash_digestsize(ctx->base_hash);
+	ss = crypto_shash_statesize(ctx->base_hash);
+
+	{
+		struct {
+			struct shash_desc shash;
+			char ctx[crypto_shash_descsize(ctx->base_hash)];
+		} desc;
+		unsigned int i;
+		char ipad[ss];
+		char opad[ss];
+
+		desc.shash.tfm = ctx->base_hash;
+		desc.shash.flags = crypto_shash_get_flags(ctx->base_hash) &
+		    CRYPTO_TFM_REQ_MAY_SLEEP;
+
+		if (keylen > bs) {
+			int err;
+
+			err =
+			    crypto_shash_digest(&desc.shash, key, keylen, ipad);
+			if (err)
+				return err;
+
+			keylen = ds;
+		} else
+			memcpy(ipad, key, keylen);
+
+		memset(ipad + keylen, 0, bs - keylen);
+		memcpy(opad, ipad, bs);
+
+		for (i = 0; i < bs; i++) {
+			ipad[i] ^= 0x36;
+			opad[i] ^= 0x5c;
+		}
+
+		rc = crypto_shash_init(&desc.shash) ? :
+		    crypto_shash_update(&desc.shash, ipad, bs) ? :
+		    crypto_shash_export(&desc.shash, ipad) ? :
+		    crypto_shash_init(&desc.shash) ? :
+		    crypto_shash_update(&desc.shash, opad, bs) ? :
+		    crypto_shash_export(&desc.shash, opad);
+
+		if (rc == 0)
+			mv_hash_init_ivs(ctx, ipad, opad);
+
+		return rc;
+	}
+}
+
+static int mv_cra_hash_init(struct crypto_tfm *tfm, const char *base_hash_name,
+			    enum hash_op op, int count_add)
+{
+	const char *fallback_driver_name = tfm->__crt_alg->cra_name;
+	struct mv_tfm_hash_ctx *ctx = crypto_tfm_ctx(tfm);
+	struct crypto_shash *fallback_tfm = NULL;
+	struct crypto_shash *base_hash = NULL;
+	int err = -ENOMEM;
+
+	ctx->op = op;
+	ctx->count_add = count_add;
+
+	/* Allocate a fallback and abort if it failed. */
+	fallback_tfm = crypto_alloc_shash(fallback_driver_name, 0,
+					  CRYPTO_ALG_NEED_FALLBACK);
+	if (IS_ERR(fallback_tfm)) {
+		printk(KERN_WARNING MV_CESA
+		       "Fallback driver '%s' could not be loaded!\n",
+		       fallback_driver_name);
+		err = PTR_ERR(fallback_tfm);
+		goto out;
+	}
+	ctx->fallback = fallback_tfm;
+
+	if (base_hash_name) {
+		/* Allocate a hash to compute the ipad/opad of hmac. */
+		base_hash = crypto_alloc_shash(base_hash_name, 0,
+					       CRYPTO_ALG_NEED_FALLBACK);
+		if (IS_ERR(base_hash)) {
+			printk(KERN_WARNING MV_CESA
+			       "Base driver '%s' could not be loaded!\n",
+			       base_hash_name);
+			err = PTR_ERR(fallback_tfm);
+			goto err_bad_base;
+		}
+	}
+	ctx->base_hash = base_hash;
+
+	crypto_ahash_set_reqsize(__crypto_ahash_cast(tfm),
+				 sizeof(struct mv_req_hash_ctx) +
+				 crypto_shash_descsize(ctx->fallback));
+	return 0;
+err_bad_base:
+	crypto_free_shash(fallback_tfm);
+out:
+	return err;
+}
+
+static void mv_cra_hash_exit(struct crypto_tfm *tfm)
+{
+	struct mv_tfm_hash_ctx *ctx = crypto_tfm_ctx(tfm);
+
+	crypto_free_shash(ctx->fallback);
+	if (ctx->base_hash)
+		crypto_free_shash(ctx->base_hash);
+}
+
+static int mv_cra_hash_sha1_init(struct crypto_tfm *tfm)
+{
+	return mv_cra_hash_init(tfm, NULL, COP_SHA1, 0);
+}
+
+static int mv_cra_hash_hmac_sha1_init(struct crypto_tfm *tfm)
+{
+	return mv_cra_hash_init(tfm, "sha1", COP_HMAC_SHA1, SHA1_BLOCK_SIZE);
+}
+
 irqreturn_t crypto_int(int irq, void *priv)
 {
 	u32 val;
@@ -474,6 +954,53 @@ struct crypto_alg mv_aes_alg_cbc = {
 	},
 };
 
+struct ahash_alg mv_sha1_alg = {
+	.init = mv_hash_init,
+	.update = mv_hash_update,
+	.final = mv_hash_final,
+	.finup = mv_hash_finup,
+	.digest = mv_hash_digest,
+	.halg = {
+		 .digestsize = SHA1_DIGEST_SIZE,
+		 .base = {
+			  .cra_name = "sha1",
+			  .cra_driver_name = "mv-sha1",
+			  .cra_priority = 300,
+			  .cra_flags =
+			  CRYPTO_ALG_ASYNC | CRYPTO_ALG_NEED_FALLBACK,
+			  .cra_blocksize = SHA1_BLOCK_SIZE,
+			  .cra_ctxsize = sizeof(struct mv_tfm_hash_ctx),
+			  .cra_init = mv_cra_hash_sha1_init,
+			  .cra_exit = mv_cra_hash_exit,
+			  .cra_module = THIS_MODULE,
+			  }
+		 }
+};
+
+struct ahash_alg mv_hmac_sha1_alg = {
+	.init = mv_hash_init,
+	.update = mv_hash_update,
+	.final = mv_hash_final,
+	.finup = mv_hash_finup,
+	.digest = mv_hash_digest,
+	.setkey = mv_hash_setkey,
+	.halg = {
+		 .digestsize = SHA1_DIGEST_SIZE,
+		 .base = {
+			  .cra_name = "hmac(sha1)",
+			  .cra_driver_name = "mv-hmac-sha1",
+			  .cra_priority = 300,
+			  .cra_flags =
+			  CRYPTO_ALG_ASYNC | CRYPTO_ALG_NEED_FALLBACK,
+			  .cra_blocksize = SHA1_BLOCK_SIZE,
+			  .cra_ctxsize = sizeof(struct mv_tfm_hash_ctx),
+			  .cra_init = mv_cra_hash_hmac_sha1_init,
+			  .cra_exit = mv_cra_hash_exit,
+			  .cra_module = THIS_MODULE,
+			  }
+		 }
+};
+
 static int mv_probe(struct platform_device *pdev)
 {
 	struct crypto_priv *cp;
@@ -482,7 +1009,7 @@ static int mv_probe(struct platform_device *pdev)
 	int ret;
 
 	if (cpg) {
-		printk(KERN_ERR "Second crypto dev?\n");
+		printk(KERN_ERR MV_CESA "Second crypto dev?\n");
 		return -EEXIST;
 	}
 
@@ -496,7 +1023,7 @@ static int mv_probe(struct platform_device *pdev)
 
 	spin_lock_init(&cp->lock);
 	crypto_init_queue(&cp->queue, 50);
-	cp->reg = ioremap(res->start, res->end - res->start + 1);
+	cp->reg = ioremap(res->start, resource_size(res));
 	if (!cp->reg) {
 		ret = -ENOMEM;
 		goto err;
@@ -507,7 +1034,7 @@ static int mv_probe(struct platform_device *pdev)
 		ret = -ENXIO;
 		goto err_unmap_reg;
 	}
-	cp->sram_size = res->end - res->start + 1;
+	cp->sram_size = resource_size(res);
 	cp->max_req_size = cp->sram_size - SRAM_CFG_SPACE;
 	cp->sram = ioremap(res->start, cp->sram_size);
 	if (!cp->sram) {
@@ -546,6 +1073,21 @@ static int mv_probe(struct platform_device *pdev)
 	ret = crypto_register_alg(&mv_aes_alg_cbc);
 	if (ret)
 		goto err_unreg_ecb;
+
+	ret = crypto_register_ahash(&mv_sha1_alg);
+	if (ret == 0)
+		cpg->has_sha1 = 1;
+	else
+		printk(KERN_WARNING MV_CESA "Could not register sha1 driver\n");
+
+	ret = crypto_register_ahash(&mv_hmac_sha1_alg);
+	if (ret == 0) {
+		cpg->has_hmac_sha1 = 1;
+	} else {
+		printk(KERN_WARNING MV_CESA
+		       "Could not register hmac-sha1 driver\n");
+	}
+
 	return 0;
 err_unreg_ecb:
 	crypto_unregister_alg(&mv_aes_alg_ecb);
@@ -570,6 +1112,10 @@ static int mv_remove(struct platform_device *pdev)
 
 	crypto_unregister_alg(&mv_aes_alg_ecb);
 	crypto_unregister_alg(&mv_aes_alg_cbc);
+	if (cp->has_sha1)
+		crypto_unregister_ahash(&mv_sha1_alg);
+	if (cp->has_hmac_sha1)
+		crypto_unregister_ahash(&mv_hmac_sha1_alg);
 	kthread_stop(cp->queue_th);
 	free_irq(cp->irq, cp);
 	memset(cp->sram, 0, cp->sram_size);
diff --git a/drivers/crypto/mv_cesa.h b/drivers/crypto/mv_cesa.h
index c3e25d3bb171..08fcb1116d90 100644
--- a/drivers/crypto/mv_cesa.h
+++ b/drivers/crypto/mv_cesa.h
@@ -1,6 +1,10 @@
 #ifndef __MV_CRYPTO_H__
 
 #define DIGEST_INITIAL_VAL_A	0xdd00
+#define DIGEST_INITIAL_VAL_B	0xdd04
+#define DIGEST_INITIAL_VAL_C	0xdd08
+#define DIGEST_INITIAL_VAL_D	0xdd0c
+#define DIGEST_INITIAL_VAL_E	0xdd10
 #define DES_CMD_REG		0xdd58
 
 #define SEC_ACCEL_CMD		0xde00
@@ -70,6 +74,10 @@ struct sec_accel_config {
 #define CFG_AES_LEN_128		(0 << 24)
 #define CFG_AES_LEN_192		(1 << 24)
 #define CFG_AES_LEN_256		(2 << 24)
+#define CFG_NOT_FRAG		(0 << 30)
+#define CFG_FIRST_FRAG		(1 << 30)
+#define CFG_LAST_FRAG		(2 << 30)
+#define CFG_MID_FRAG		(3 << 30)
 
 	u32 enc_p;
 #define ENC_P_SRC(x)		(x)
@@ -90,7 +98,11 @@ struct sec_accel_config {
 #define MAC_SRC_TOTAL_LEN(x)	((x) << 16)
 
 	u32 mac_digest;
+#define MAC_DIGEST_P(x)	(x)
+#define MAC_FRAG_LEN(x)	((x) << 16)
 	u32 mac_iv;
+#define MAC_INNER_IV_P(x)	(x)
+#define MAC_OUTER_IV_P(x)	((x) << 16)
 }__attribute__ ((packed));
 	/*
 	 * /-----------\ 0
@@ -101,19 +113,37 @@ struct sec_accel_config {
 	 * |  IV   IN  |	4 * 4
 	 * |-----------| 0x40 (inplace)
 	 * |  IV BUF   |	4 * 4
-	 * |-----------| 0x50
+	 * |-----------| 0x80
 	 * |  DATA IN  |	16 * x (max ->max_req_size)
-	 * |-----------| 0x50 (inplace operation)
+	 * |-----------| 0x80 (inplace operation)
 	 * |  DATA OUT |	16 * x (max ->max_req_size)
 	 * \-----------/ SRAM size
 	 */
+
+	/* Hashing memory map:
+	 * /-----------\ 0
+	 * | ACCEL CFG |        4 * 8
+	 * |-----------| 0x20
+	 * | Inner IV  |        5 * 4
+	 * |-----------| 0x34
+	 * | Outer IV  |        5 * 4
+	 * |-----------| 0x48
+	 * | Output BUF|        5 * 4
+	 * |-----------| 0x80
+	 * |  DATA IN  |        64 * x (max ->max_req_size)
+	 * \-----------/ SRAM size
+	 */
 #define SRAM_CONFIG		0x00
 #define SRAM_DATA_KEY_P		0x20
 #define SRAM_DATA_IV		0x40
 #define SRAM_DATA_IV_BUF	0x40
-#define SRAM_DATA_IN_START	0x50
-#define SRAM_DATA_OUT_START	0x50
+#define SRAM_DATA_IN_START	0x80
+#define SRAM_DATA_OUT_START	0x80
+
+#define SRAM_HMAC_IV_IN		0x20
+#define SRAM_HMAC_IV_OUT	0x34
+#define SRAM_DIGEST_BUF		0x48
 
-#define SRAM_CFG_SPACE		0x50
+#define SRAM_CFG_SPACE		0x80
 
 #endif
diff --git a/drivers/crypto/n2_asm.S b/drivers/crypto/n2_asm.S
new file mode 100644
index 000000000000..f7c793745a1e
--- /dev/null
+++ b/drivers/crypto/n2_asm.S
@@ -0,0 +1,95 @@
+/* n2_asm.S: Hypervisor calls for NCS support.
+ *
+ * Copyright (C) 2009 David S. Miller <davem@davemloft.net>
+ */
+
+#include <linux/linkage.h>
+#include <asm/hypervisor.h>
+#include "n2_core.h"
+
+	/* o0: queue type
+	 * o1: RA of queue
+	 * o2: num entries in queue
+	 * o3: address of queue handle return
+	 */
+ENTRY(sun4v_ncs_qconf)
+	mov	HV_FAST_NCS_QCONF, %o5
+	ta	HV_FAST_TRAP
+	stx	%o1, [%o3]
+	retl
+	 nop
+ENDPROC(sun4v_ncs_qconf)
+
+	/* %o0: queue handle
+	 * %o1: address of queue type return
+	 * %o2: address of queue base address return
+	 * %o3: address of queue num entries return
+	 */
+ENTRY(sun4v_ncs_qinfo)
+	mov	%o1, %g1
+	mov	%o2, %g2
+	mov	%o3, %g3
+	mov	HV_FAST_NCS_QINFO, %o5
+	ta	HV_FAST_TRAP
+	stx	%o1, [%g1]
+	stx	%o2, [%g2]
+	stx	%o3, [%g3]
+	retl
+	 nop
+ENDPROC(sun4v_ncs_qinfo)
+
+	/* %o0: queue handle
+	 * %o1: address of head offset return
+	 */
+ENTRY(sun4v_ncs_gethead)
+	mov	%o1, %o2
+	mov	HV_FAST_NCS_GETHEAD, %o5
+	ta	HV_FAST_TRAP
+	stx	%o1, [%o2]
+	retl
+	 nop
+ENDPROC(sun4v_ncs_gethead)
+
+	/* %o0: queue handle
+	 * %o1: address of tail offset return
+	 */
+ENTRY(sun4v_ncs_gettail)
+	mov	%o1, %o2
+	mov	HV_FAST_NCS_GETTAIL, %o5
+	ta	HV_FAST_TRAP
+	stx	%o1, [%o2]
+	retl
+	 nop
+ENDPROC(sun4v_ncs_gettail)
+
+	/* %o0: queue handle
+	 * %o1: new tail offset
+	 */
+ENTRY(sun4v_ncs_settail)
+	mov	HV_FAST_NCS_SETTAIL, %o5
+	ta	HV_FAST_TRAP
+	retl
+	 nop
+ENDPROC(sun4v_ncs_settail)
+
+	/* %o0: queue handle
+	 * %o1: address of devino return
+	 */
+ENTRY(sun4v_ncs_qhandle_to_devino)
+	mov	%o1, %o2
+	mov	HV_FAST_NCS_QHANDLE_TO_DEVINO, %o5
+	ta	HV_FAST_TRAP
+	stx	%o1, [%o2]
+	retl
+	 nop
+ENDPROC(sun4v_ncs_qhandle_to_devino)
+
+	/* %o0: queue handle
+	 * %o1: new head offset
+	 */
+ENTRY(sun4v_ncs_sethead_marker)
+	mov	HV_FAST_NCS_SETHEAD_MARKER, %o5
+	ta	HV_FAST_TRAP
+	retl
+	 nop
+ENDPROC(sun4v_ncs_sethead_marker)
diff --git a/drivers/crypto/n2_core.c b/drivers/crypto/n2_core.c
new file mode 100644
index 000000000000..23163fda5035
--- /dev/null
+++ b/drivers/crypto/n2_core.c
@@ -0,0 +1,2090 @@
+/* n2_core.c: Niagara2 Stream Processing Unit (SPU) crypto support.
+ *
+ * Copyright (C) 2010 David S. Miller <davem@davemloft.net>
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/of_device.h>
+#include <linux/cpumask.h>
+#include <linux/slab.h>
+#include <linux/interrupt.h>
+#include <linux/crypto.h>
+#include <crypto/md5.h>
+#include <crypto/sha.h>
+#include <crypto/aes.h>
+#include <crypto/des.h>
+#include <linux/mutex.h>
+#include <linux/delay.h>
+#include <linux/sched.h>
+
+#include <crypto/internal/hash.h>
+#include <crypto/scatterwalk.h>
+#include <crypto/algapi.h>
+
+#include <asm/hypervisor.h>
+#include <asm/mdesc.h>
+
+#include "n2_core.h"
+
+#define DRV_MODULE_NAME		"n2_crypto"
+#define DRV_MODULE_VERSION	"0.1"
+#define DRV_MODULE_RELDATE	"April 29, 2010"
+
+static char version[] __devinitdata =
+	DRV_MODULE_NAME ".c:v" DRV_MODULE_VERSION " (" DRV_MODULE_RELDATE ")\n";
+
+MODULE_AUTHOR("David S. Miller (davem@davemloft.net)");
+MODULE_DESCRIPTION("Niagara2 Crypto driver");
+MODULE_LICENSE("GPL");
+MODULE_VERSION(DRV_MODULE_VERSION);
+
+#define N2_CRA_PRIORITY		300
+
+static DEFINE_MUTEX(spu_lock);
+
+struct spu_queue {
+	cpumask_t		sharing;
+	unsigned long		qhandle;
+
+	spinlock_t		lock;
+	u8			q_type;
+	void			*q;
+	unsigned long		head;
+	unsigned long		tail;
+	struct list_head	jobs;
+
+	unsigned long		devino;
+
+	char			irq_name[32];
+	unsigned int		irq;
+
+	struct list_head	list;
+};
+
+static struct spu_queue **cpu_to_cwq;
+static struct spu_queue **cpu_to_mau;
+
+static unsigned long spu_next_offset(struct spu_queue *q, unsigned long off)
+{
+	if (q->q_type == HV_NCS_QTYPE_MAU) {
+		off += MAU_ENTRY_SIZE;
+		if (off == (MAU_ENTRY_SIZE * MAU_NUM_ENTRIES))
+			off = 0;
+	} else {
+		off += CWQ_ENTRY_SIZE;
+		if (off == (CWQ_ENTRY_SIZE * CWQ_NUM_ENTRIES))
+			off = 0;
+	}
+	return off;
+}
+
+struct n2_request_common {
+	struct list_head	entry;
+	unsigned int		offset;
+};
+#define OFFSET_NOT_RUNNING	(~(unsigned int)0)
+
+/* An async job request records the final tail value it used in
+ * n2_request_common->offset, test to see if that offset is in
+ * the range old_head, new_head, inclusive.
+ */
+static inline bool job_finished(struct spu_queue *q, unsigned int offset,
+				unsigned long old_head, unsigned long new_head)
+{
+	if (old_head <= new_head) {
+		if (offset > old_head && offset <= new_head)
+			return true;
+	} else {
+		if (offset > old_head || offset <= new_head)
+			return true;
+	}
+	return false;
+}
+
+/* When the HEAD marker is unequal to the actual HEAD, we get
+ * a virtual device INO interrupt.  We should process the
+ * completed CWQ entries and adjust the HEAD marker to clear
+ * the IRQ.
+ */
+static irqreturn_t cwq_intr(int irq, void *dev_id)
+{
+	unsigned long off, new_head, hv_ret;
+	struct spu_queue *q = dev_id;
+
+	pr_err("CPU[%d]: Got CWQ interrupt for qhdl[%lx]\n",
+	       smp_processor_id(), q->qhandle);
+
+	spin_lock(&q->lock);
+
+	hv_ret = sun4v_ncs_gethead(q->qhandle, &new_head);
+
+	pr_err("CPU[%d]: CWQ gethead[%lx] hv_ret[%lu]\n",
+	       smp_processor_id(), new_head, hv_ret);
+
+	for (off = q->head; off != new_head; off = spu_next_offset(q, off)) {
+		/* XXX ... XXX */
+	}
+
+	hv_ret = sun4v_ncs_sethead_marker(q->qhandle, new_head);
+	if (hv_ret == HV_EOK)
+		q->head = new_head;
+
+	spin_unlock(&q->lock);
+
+	return IRQ_HANDLED;
+}
+
+static irqreturn_t mau_intr(int irq, void *dev_id)
+{
+	struct spu_queue *q = dev_id;
+	unsigned long head, hv_ret;
+
+	spin_lock(&q->lock);
+
+	pr_err("CPU[%d]: Got MAU interrupt for qhdl[%lx]\n",
+	       smp_processor_id(), q->qhandle);
+
+	hv_ret = sun4v_ncs_gethead(q->qhandle, &head);
+
+	pr_err("CPU[%d]: MAU gethead[%lx] hv_ret[%lu]\n",
+	       smp_processor_id(), head, hv_ret);
+
+	sun4v_ncs_sethead_marker(q->qhandle, head);
+
+	spin_unlock(&q->lock);
+
+	return IRQ_HANDLED;
+}
+
+static void *spu_queue_next(struct spu_queue *q, void *cur)
+{
+	return q->q + spu_next_offset(q, cur - q->q);
+}
+
+static int spu_queue_num_free(struct spu_queue *q)
+{
+	unsigned long head = q->head;
+	unsigned long tail = q->tail;
+	unsigned long end = (CWQ_ENTRY_SIZE * CWQ_NUM_ENTRIES);
+	unsigned long diff;
+
+	if (head > tail)
+		diff = head - tail;
+	else
+		diff = (end - tail) + head;
+
+	return (diff / CWQ_ENTRY_SIZE) - 1;
+}
+
+static void *spu_queue_alloc(struct spu_queue *q, int num_entries)
+{
+	int avail = spu_queue_num_free(q);
+
+	if (avail >= num_entries)
+		return q->q + q->tail;
+
+	return NULL;
+}
+
+static unsigned long spu_queue_submit(struct spu_queue *q, void *last)
+{
+	unsigned long hv_ret, new_tail;
+
+	new_tail = spu_next_offset(q, last - q->q);
+
+	hv_ret = sun4v_ncs_settail(q->qhandle, new_tail);
+	if (hv_ret == HV_EOK)
+		q->tail = new_tail;
+	return hv_ret;
+}
+
+static u64 control_word_base(unsigned int len, unsigned int hmac_key_len,
+			     int enc_type, int auth_type,
+			     unsigned int hash_len,
+			     bool sfas, bool sob, bool eob, bool encrypt,
+			     int opcode)
+{
+	u64 word = (len - 1) & CONTROL_LEN;
+
+	word |= ((u64) opcode << CONTROL_OPCODE_SHIFT);
+	word |= ((u64) enc_type << CONTROL_ENC_TYPE_SHIFT);
+	word |= ((u64) auth_type << CONTROL_AUTH_TYPE_SHIFT);
+	if (sfas)
+		word |= CONTROL_STORE_FINAL_AUTH_STATE;
+	if (sob)
+		word |= CONTROL_START_OF_BLOCK;
+	if (eob)
+		word |= CONTROL_END_OF_BLOCK;
+	if (encrypt)
+		word |= CONTROL_ENCRYPT;
+	if (hmac_key_len)
+		word |= ((u64) (hmac_key_len - 1)) << CONTROL_HMAC_KEY_LEN_SHIFT;
+	if (hash_len)
+		word |= ((u64) (hash_len - 1)) << CONTROL_HASH_LEN_SHIFT;
+
+	return word;
+}
+
+#if 0
+static inline bool n2_should_run_async(struct spu_queue *qp, int this_len)
+{
+	if (this_len >= 64 ||
+	    qp->head != qp->tail)
+		return true;
+	return false;
+}
+#endif
+
+struct n2_base_ctx {
+	struct list_head		list;
+};
+
+static void n2_base_ctx_init(struct n2_base_ctx *ctx)
+{
+	INIT_LIST_HEAD(&ctx->list);
+}
+
+struct n2_hash_ctx {
+	struct n2_base_ctx		base;
+
+	struct crypto_ahash		*fallback_tfm;
+};
+
+struct n2_hash_req_ctx {
+	union {
+		struct md5_state	md5;
+		struct sha1_state	sha1;
+		struct sha256_state	sha256;
+	} u;
+
+	unsigned char			hash_key[64];
+	unsigned char			keyed_zero_hash[32];
+
+	struct ahash_request		fallback_req;
+};
+
+static int n2_hash_async_init(struct ahash_request *req)
+{
+	struct n2_hash_req_ctx *rctx = ahash_request_ctx(req);
+	struct crypto_ahash *tfm = crypto_ahash_reqtfm(req);
+	struct n2_hash_ctx *ctx = crypto_ahash_ctx(tfm);
+
+	ahash_request_set_tfm(&rctx->fallback_req, ctx->fallback_tfm);
+	rctx->fallback_req.base.flags = req->base.flags & CRYPTO_TFM_REQ_MAY_SLEEP;
+
+	return crypto_ahash_init(&rctx->fallback_req);
+}
+
+static int n2_hash_async_update(struct ahash_request *req)
+{
+	struct n2_hash_req_ctx *rctx = ahash_request_ctx(req);
+	struct crypto_ahash *tfm = crypto_ahash_reqtfm(req);
+	struct n2_hash_ctx *ctx = crypto_ahash_ctx(tfm);
+
+	ahash_request_set_tfm(&rctx->fallback_req, ctx->fallback_tfm);
+	rctx->fallback_req.base.flags = req->base.flags & CRYPTO_TFM_REQ_MAY_SLEEP;
+	rctx->fallback_req.nbytes = req->nbytes;
+	rctx->fallback_req.src = req->src;
+
+	return crypto_ahash_update(&rctx->fallback_req);
+}
+
+static int n2_hash_async_final(struct ahash_request *req)
+{
+	struct n2_hash_req_ctx *rctx = ahash_request_ctx(req);
+	struct crypto_ahash *tfm = crypto_ahash_reqtfm(req);
+	struct n2_hash_ctx *ctx = crypto_ahash_ctx(tfm);
+
+	ahash_request_set_tfm(&rctx->fallback_req, ctx->fallback_tfm);
+	rctx->fallback_req.base.flags = req->base.flags & CRYPTO_TFM_REQ_MAY_SLEEP;
+	rctx->fallback_req.result = req->result;
+
+	return crypto_ahash_final(&rctx->fallback_req);
+}
+
+static int n2_hash_async_finup(struct ahash_request *req)
+{
+	struct n2_hash_req_ctx *rctx = ahash_request_ctx(req);
+	struct crypto_ahash *tfm = crypto_ahash_reqtfm(req);
+	struct n2_hash_ctx *ctx = crypto_ahash_ctx(tfm);
+
+	ahash_request_set_tfm(&rctx->fallback_req, ctx->fallback_tfm);
+	rctx->fallback_req.base.flags = req->base.flags & CRYPTO_TFM_REQ_MAY_SLEEP;
+	rctx->fallback_req.nbytes = req->nbytes;
+	rctx->fallback_req.src = req->src;
+	rctx->fallback_req.result = req->result;
+
+	return crypto_ahash_finup(&rctx->fallback_req);
+}
+
+static int n2_hash_cra_init(struct crypto_tfm *tfm)
+{
+	const char *fallback_driver_name = tfm->__crt_alg->cra_name;
+	struct crypto_ahash *ahash = __crypto_ahash_cast(tfm);
+	struct n2_hash_ctx *ctx = crypto_ahash_ctx(ahash);
+	struct crypto_ahash *fallback_tfm;
+	int err;
+
+	fallback_tfm = crypto_alloc_ahash(fallback_driver_name, 0,
+					  CRYPTO_ALG_NEED_FALLBACK);
+	if (IS_ERR(fallback_tfm)) {
+		pr_warning("Fallback driver '%s' could not be loaded!\n",
+			   fallback_driver_name);
+		err = PTR_ERR(fallback_tfm);
+		goto out;
+	}
+
+	crypto_ahash_set_reqsize(ahash, (sizeof(struct n2_hash_req_ctx) +
+					 crypto_ahash_reqsize(fallback_tfm)));
+
+	ctx->fallback_tfm = fallback_tfm;
+	return 0;
+
+out:
+	return err;
+}
+
+static void n2_hash_cra_exit(struct crypto_tfm *tfm)
+{
+	struct crypto_ahash *ahash = __crypto_ahash_cast(tfm);
+	struct n2_hash_ctx *ctx = crypto_ahash_ctx(ahash);
+
+	crypto_free_ahash(ctx->fallback_tfm);
+}
+
+static unsigned long wait_for_tail(struct spu_queue *qp)
+{
+	unsigned long head, hv_ret;
+
+	do {
+		hv_ret = sun4v_ncs_gethead(qp->qhandle, &head);
+		if (hv_ret != HV_EOK) {
+			pr_err("Hypervisor error on gethead\n");
+			break;
+		}
+		if (head == qp->tail) {
+			qp->head = head;
+			break;
+		}
+	} while (1);
+	return hv_ret;
+}
+
+static unsigned long submit_and_wait_for_tail(struct spu_queue *qp,
+					      struct cwq_initial_entry *ent)
+{
+	unsigned long hv_ret = spu_queue_submit(qp, ent);
+
+	if (hv_ret == HV_EOK)
+		hv_ret = wait_for_tail(qp);
+
+	return hv_ret;
+}
+
+static int n2_hash_async_digest(struct ahash_request *req,
+				unsigned int auth_type, unsigned int digest_size,
+				unsigned int result_size, void *hash_loc)
+{
+	struct crypto_ahash *tfm = crypto_ahash_reqtfm(req);
+	struct n2_hash_ctx *ctx = crypto_ahash_ctx(tfm);
+	struct cwq_initial_entry *ent;
+	struct crypto_hash_walk walk;
+	struct spu_queue *qp;
+	unsigned long flags;
+	int err = -ENODEV;
+	int nbytes, cpu;
+
+	/* The total effective length of the operation may not
+	 * exceed 2^16.
+	 */
+	if (unlikely(req->nbytes > (1 << 16))) {
+		struct n2_hash_req_ctx *rctx = ahash_request_ctx(req);
+
+		ahash_request_set_tfm(&rctx->fallback_req, ctx->fallback_tfm);
+		rctx->fallback_req.base.flags =
+			req->base.flags & CRYPTO_TFM_REQ_MAY_SLEEP;
+		rctx->fallback_req.nbytes = req->nbytes;
+		rctx->fallback_req.src = req->src;
+		rctx->fallback_req.result = req->result;
+
+		return crypto_ahash_digest(&rctx->fallback_req);
+	}
+
+	n2_base_ctx_init(&ctx->base);
+
+	nbytes = crypto_hash_walk_first(req, &walk);
+
+	cpu = get_cpu();
+	qp = cpu_to_cwq[cpu];
+	if (!qp)
+		goto out;
+
+	spin_lock_irqsave(&qp->lock, flags);
+
+	/* XXX can do better, improve this later by doing a by-hand scatterlist
+	 * XXX walk, etc.
+	 */
+	ent = qp->q + qp->tail;
+
+	ent->control = control_word_base(nbytes, 0, 0,
+					 auth_type, digest_size,
+					 false, true, false, false,
+					 OPCODE_INPLACE_BIT |
+					 OPCODE_AUTH_MAC);
+	ent->src_addr = __pa(walk.data);
+	ent->auth_key_addr = 0UL;
+	ent->auth_iv_addr = __pa(hash_loc);
+	ent->final_auth_state_addr = 0UL;
+	ent->enc_key_addr = 0UL;
+	ent->enc_iv_addr = 0UL;
+	ent->dest_addr = __pa(hash_loc);
+
+	nbytes = crypto_hash_walk_done(&walk, 0);
+	while (nbytes > 0) {
+		ent = spu_queue_next(qp, ent);
+
+		ent->control = (nbytes - 1);
+		ent->src_addr = __pa(walk.data);
+		ent->auth_key_addr = 0UL;
+		ent->auth_iv_addr = 0UL;
+		ent->final_auth_state_addr = 0UL;
+		ent->enc_key_addr = 0UL;
+		ent->enc_iv_addr = 0UL;
+		ent->dest_addr = 0UL;
+
+		nbytes = crypto_hash_walk_done(&walk, 0);
+	}
+	ent->control |= CONTROL_END_OF_BLOCK;
+
+	if (submit_and_wait_for_tail(qp, ent) != HV_EOK)
+		err = -EINVAL;
+	else
+		err = 0;
+
+	spin_unlock_irqrestore(&qp->lock, flags);
+
+	if (!err)
+		memcpy(req->result, hash_loc, result_size);
+out:
+	put_cpu();
+
+	return err;
+}
+
+static int n2_md5_async_digest(struct ahash_request *req)
+{
+	struct n2_hash_req_ctx *rctx = ahash_request_ctx(req);
+	struct md5_state *m = &rctx->u.md5;
+
+	if (unlikely(req->nbytes == 0)) {
+		static const char md5_zero[MD5_DIGEST_SIZE] = {
+			0xd4, 0x1d, 0x8c, 0xd9, 0x8f, 0x00, 0xb2, 0x04,
+			0xe9, 0x80, 0x09, 0x98, 0xec, 0xf8, 0x42, 0x7e,
+		};
+
+		memcpy(req->result, md5_zero, MD5_DIGEST_SIZE);
+		return 0;
+	}
+	m->hash[0] = cpu_to_le32(0x67452301);
+	m->hash[1] = cpu_to_le32(0xefcdab89);
+	m->hash[2] = cpu_to_le32(0x98badcfe);
+	m->hash[3] = cpu_to_le32(0x10325476);
+
+	return n2_hash_async_digest(req, AUTH_TYPE_MD5,
+				    MD5_DIGEST_SIZE, MD5_DIGEST_SIZE,
+				    m->hash);
+}
+
+static int n2_sha1_async_digest(struct ahash_request *req)
+{
+	struct n2_hash_req_ctx *rctx = ahash_request_ctx(req);
+	struct sha1_state *s = &rctx->u.sha1;
+
+	if (unlikely(req->nbytes == 0)) {
+		static const char sha1_zero[SHA1_DIGEST_SIZE] = {
+			0xda, 0x39, 0xa3, 0xee, 0x5e, 0x6b, 0x4b, 0x0d, 0x32,
+			0x55, 0xbf, 0xef, 0x95, 0x60, 0x18, 0x90, 0xaf, 0xd8,
+			0x07, 0x09
+		};
+
+		memcpy(req->result, sha1_zero, SHA1_DIGEST_SIZE);
+		return 0;
+	}
+	s->state[0] = SHA1_H0;
+	s->state[1] = SHA1_H1;
+	s->state[2] = SHA1_H2;
+	s->state[3] = SHA1_H3;
+	s->state[4] = SHA1_H4;
+
+	return n2_hash_async_digest(req, AUTH_TYPE_SHA1,
+				    SHA1_DIGEST_SIZE, SHA1_DIGEST_SIZE,
+				    s->state);
+}
+
+static int n2_sha256_async_digest(struct ahash_request *req)
+{
+	struct n2_hash_req_ctx *rctx = ahash_request_ctx(req);
+	struct sha256_state *s = &rctx->u.sha256;
+
+	if (req->nbytes == 0) {
+		static const char sha256_zero[SHA256_DIGEST_SIZE] = {
+			0xe3, 0xb0, 0xc4, 0x42, 0x98, 0xfc, 0x1c, 0x14, 0x9a,
+			0xfb, 0xf4, 0xc8, 0x99, 0x6f, 0xb9, 0x24, 0x27, 0xae,
+			0x41, 0xe4, 0x64, 0x9b, 0x93, 0x4c, 0xa4, 0x95, 0x99,
+			0x1b, 0x78, 0x52, 0xb8, 0x55
+		};
+
+		memcpy(req->result, sha256_zero, SHA256_DIGEST_SIZE);
+		return 0;
+	}
+	s->state[0] = SHA256_H0;
+	s->state[1] = SHA256_H1;
+	s->state[2] = SHA256_H2;
+	s->state[3] = SHA256_H3;
+	s->state[4] = SHA256_H4;
+	s->state[5] = SHA256_H5;
+	s->state[6] = SHA256_H6;
+	s->state[7] = SHA256_H7;
+
+	return n2_hash_async_digest(req, AUTH_TYPE_SHA256,
+				    SHA256_DIGEST_SIZE, SHA256_DIGEST_SIZE,
+				    s->state);
+}
+
+static int n2_sha224_async_digest(struct ahash_request *req)
+{
+	struct n2_hash_req_ctx *rctx = ahash_request_ctx(req);
+	struct sha256_state *s = &rctx->u.sha256;
+
+	if (req->nbytes == 0) {
+		static const char sha224_zero[SHA224_DIGEST_SIZE] = {
+			0xd1, 0x4a, 0x02, 0x8c, 0x2a, 0x3a, 0x2b, 0xc9, 0x47,
+			0x61, 0x02, 0xbb, 0x28, 0x82, 0x34, 0xc4, 0x15, 0xa2,
+			0xb0, 0x1f, 0x82, 0x8e, 0xa6, 0x2a, 0xc5, 0xb3, 0xe4,
+			0x2f
+		};
+
+		memcpy(req->result, sha224_zero, SHA224_DIGEST_SIZE);
+		return 0;
+	}
+	s->state[0] = SHA224_H0;
+	s->state[1] = SHA224_H1;
+	s->state[2] = SHA224_H2;
+	s->state[3] = SHA224_H3;
+	s->state[4] = SHA224_H4;
+	s->state[5] = SHA224_H5;
+	s->state[6] = SHA224_H6;
+	s->state[7] = SHA224_H7;
+
+	return n2_hash_async_digest(req, AUTH_TYPE_SHA256,
+				    SHA256_DIGEST_SIZE, SHA224_DIGEST_SIZE,
+				    s->state);
+}
+
+struct n2_cipher_context {
+	int			key_len;
+	int			enc_type;
+	union {
+		u8		aes[AES_MAX_KEY_SIZE];
+		u8		des[DES_KEY_SIZE];
+		u8		des3[3 * DES_KEY_SIZE];
+		u8		arc4[258]; /* S-box, X, Y */
+	} key;
+};
+
+#define N2_CHUNK_ARR_LEN	16
+
+struct n2_crypto_chunk {
+	struct list_head	entry;
+	unsigned long		iv_paddr : 44;
+	unsigned long		arr_len : 20;
+	unsigned long		dest_paddr;
+	unsigned long		dest_final;
+	struct {
+		unsigned long	src_paddr : 44;
+		unsigned long	src_len : 20;
+	} arr[N2_CHUNK_ARR_LEN];
+};
+
+struct n2_request_context {
+	struct ablkcipher_walk	walk;
+	struct list_head	chunk_list;
+	struct n2_crypto_chunk	chunk;
+	u8			temp_iv[16];
+};
+
+/* The SPU allows some level of flexibility for partial cipher blocks
+ * being specified in a descriptor.
+ *
+ * It merely requires that every descriptor's length field is at least
+ * as large as the cipher block size.  This means that a cipher block
+ * can span at most 2 descriptors.  However, this does not allow a
+ * partial block to span into the final descriptor as that would
+ * violate the rule (since every descriptor's length must be at lest
+ * the block size).  So, for example, assuming an 8 byte block size:
+ *
+ *	0xe --> 0xa --> 0x8
+ *
+ * is a valid length sequence, whereas:
+ *
+ *	0xe --> 0xb --> 0x7
+ *
+ * is not a valid sequence.
+ */
+
+struct n2_cipher_alg {
+	struct list_head	entry;
+	u8			enc_type;
+	struct crypto_alg	alg;
+};
+
+static inline struct n2_cipher_alg *n2_cipher_alg(struct crypto_tfm *tfm)
+{
+	struct crypto_alg *alg = tfm->__crt_alg;
+
+	return container_of(alg, struct n2_cipher_alg, alg);
+}
+
+struct n2_cipher_request_context {
+	struct ablkcipher_walk	walk;
+};
+
+static int n2_aes_setkey(struct crypto_ablkcipher *cipher, const u8 *key,
+			 unsigned int keylen)
+{
+	struct crypto_tfm *tfm = crypto_ablkcipher_tfm(cipher);
+	struct n2_cipher_context *ctx = crypto_tfm_ctx(tfm);
+	struct n2_cipher_alg *n2alg = n2_cipher_alg(tfm);
+
+	ctx->enc_type = (n2alg->enc_type & ENC_TYPE_CHAINING_MASK);
+
+	switch (keylen) {
+	case AES_KEYSIZE_128:
+		ctx->enc_type |= ENC_TYPE_ALG_AES128;
+		break;
+	case AES_KEYSIZE_192:
+		ctx->enc_type |= ENC_TYPE_ALG_AES192;
+		break;
+	case AES_KEYSIZE_256:
+		ctx->enc_type |= ENC_TYPE_ALG_AES256;
+		break;
+	default:
+		crypto_ablkcipher_set_flags(cipher, CRYPTO_TFM_RES_BAD_KEY_LEN);
+		return -EINVAL;
+	}
+
+	ctx->key_len = keylen;
+	memcpy(ctx->key.aes, key, keylen);
+	return 0;
+}
+
+static int n2_des_setkey(struct crypto_ablkcipher *cipher, const u8 *key,
+			 unsigned int keylen)
+{
+	struct crypto_tfm *tfm = crypto_ablkcipher_tfm(cipher);
+	struct n2_cipher_context *ctx = crypto_tfm_ctx(tfm);
+	struct n2_cipher_alg *n2alg = n2_cipher_alg(tfm);
+	u32 tmp[DES_EXPKEY_WORDS];
+	int err;
+
+	ctx->enc_type = n2alg->enc_type;
+
+	if (keylen != DES_KEY_SIZE) {
+		crypto_ablkcipher_set_flags(cipher, CRYPTO_TFM_RES_BAD_KEY_LEN);
+		return -EINVAL;
+	}
+
+	err = des_ekey(tmp, key);
+	if (err == 0 && (tfm->crt_flags & CRYPTO_TFM_REQ_WEAK_KEY)) {
+		tfm->crt_flags |= CRYPTO_TFM_RES_WEAK_KEY;
+		return -EINVAL;
+	}
+
+	ctx->key_len = keylen;
+	memcpy(ctx->key.des, key, keylen);
+	return 0;
+}
+
+static int n2_3des_setkey(struct crypto_ablkcipher *cipher, const u8 *key,
+			  unsigned int keylen)
+{
+	struct crypto_tfm *tfm = crypto_ablkcipher_tfm(cipher);
+	struct n2_cipher_context *ctx = crypto_tfm_ctx(tfm);
+	struct n2_cipher_alg *n2alg = n2_cipher_alg(tfm);
+
+	ctx->enc_type = n2alg->enc_type;
+
+	if (keylen != (3 * DES_KEY_SIZE)) {
+		crypto_ablkcipher_set_flags(cipher, CRYPTO_TFM_RES_BAD_KEY_LEN);
+		return -EINVAL;
+	}
+	ctx->key_len = keylen;
+	memcpy(ctx->key.des3, key, keylen);
+	return 0;
+}
+
+static int n2_arc4_setkey(struct crypto_ablkcipher *cipher, const u8 *key,
+			  unsigned int keylen)
+{
+	struct crypto_tfm *tfm = crypto_ablkcipher_tfm(cipher);
+	struct n2_cipher_context *ctx = crypto_tfm_ctx(tfm);
+	struct n2_cipher_alg *n2alg = n2_cipher_alg(tfm);
+	u8 *s = ctx->key.arc4;
+	u8 *x = s + 256;
+	u8 *y = x + 1;
+	int i, j, k;
+
+	ctx->enc_type = n2alg->enc_type;
+
+	j = k = 0;
+	*x = 0;
+	*y = 0;
+	for (i = 0; i < 256; i++)
+		s[i] = i;
+	for (i = 0; i < 256; i++) {
+		u8 a = s[i];
+		j = (j + key[k] + a) & 0xff;
+		s[i] = s[j];
+		s[j] = a;
+		if (++k >= keylen)
+			k = 0;
+	}
+
+	return 0;
+}
+
+static inline int cipher_descriptor_len(int nbytes, unsigned int block_size)
+{
+	int this_len = nbytes;
+
+	this_len -= (nbytes & (block_size - 1));
+	return this_len > (1 << 16) ? (1 << 16) : this_len;
+}
+
+static int __n2_crypt_chunk(struct crypto_tfm *tfm, struct n2_crypto_chunk *cp,
+			    struct spu_queue *qp, bool encrypt)
+{
+	struct n2_cipher_context *ctx = crypto_tfm_ctx(tfm);
+	struct cwq_initial_entry *ent;
+	bool in_place;
+	int i;
+
+	ent = spu_queue_alloc(qp, cp->arr_len);
+	if (!ent) {
+		pr_info("queue_alloc() of %d fails\n",
+			cp->arr_len);
+		return -EBUSY;
+	}
+
+	in_place = (cp->dest_paddr == cp->arr[0].src_paddr);
+
+	ent->control = control_word_base(cp->arr[0].src_len,
+					 0, ctx->enc_type, 0, 0,
+					 false, true, false, encrypt,
+					 OPCODE_ENCRYPT |
+					 (in_place ? OPCODE_INPLACE_BIT : 0));
+	ent->src_addr = cp->arr[0].src_paddr;
+	ent->auth_key_addr = 0UL;
+	ent->auth_iv_addr = 0UL;
+	ent->final_auth_state_addr = 0UL;
+	ent->enc_key_addr = __pa(&ctx->key);
+	ent->enc_iv_addr = cp->iv_paddr;
+	ent->dest_addr = (in_place ? 0UL : cp->dest_paddr);
+
+	for (i = 1; i < cp->arr_len; i++) {
+		ent = spu_queue_next(qp, ent);
+
+		ent->control = cp->arr[i].src_len - 1;
+		ent->src_addr = cp->arr[i].src_paddr;
+		ent->auth_key_addr = 0UL;
+		ent->auth_iv_addr = 0UL;
+		ent->final_auth_state_addr = 0UL;
+		ent->enc_key_addr = 0UL;
+		ent->enc_iv_addr = 0UL;
+		ent->dest_addr = 0UL;
+	}
+	ent->control |= CONTROL_END_OF_BLOCK;
+
+	return (spu_queue_submit(qp, ent) != HV_EOK) ? -EINVAL : 0;
+}
+
+static int n2_compute_chunks(struct ablkcipher_request *req)
+{
+	struct n2_request_context *rctx = ablkcipher_request_ctx(req);
+	struct ablkcipher_walk *walk = &rctx->walk;
+	struct n2_crypto_chunk *chunk;
+	unsigned long dest_prev;
+	unsigned int tot_len;
+	bool prev_in_place;
+	int err, nbytes;
+
+	ablkcipher_walk_init(walk, req->dst, req->src, req->nbytes);
+	err = ablkcipher_walk_phys(req, walk);
+	if (err)
+		return err;
+
+	INIT_LIST_HEAD(&rctx->chunk_list);
+
+	chunk = &rctx->chunk;
+	INIT_LIST_HEAD(&chunk->entry);
+
+	chunk->iv_paddr = 0UL;
+	chunk->arr_len = 0;
+	chunk->dest_paddr = 0UL;
+
+	prev_in_place = false;
+	dest_prev = ~0UL;
+	tot_len = 0;
+
+	while ((nbytes = walk->nbytes) != 0) {
+		unsigned long dest_paddr, src_paddr;
+		bool in_place;
+		int this_len;
+
+		src_paddr = (page_to_phys(walk->src.page) +
+			     walk->src.offset);
+		dest_paddr = (page_to_phys(walk->dst.page) +
+			      walk->dst.offset);
+		in_place = (src_paddr == dest_paddr);
+		this_len = cipher_descriptor_len(nbytes, walk->blocksize);
+
+		if (chunk->arr_len != 0) {
+			if (in_place != prev_in_place ||
+			    (!prev_in_place &&
+			     dest_paddr != dest_prev) ||
+			    chunk->arr_len == N2_CHUNK_ARR_LEN ||
+			    tot_len + this_len > (1 << 16)) {
+				chunk->dest_final = dest_prev;
+				list_add_tail(&chunk->entry,
+					      &rctx->chunk_list);
+				chunk = kzalloc(sizeof(*chunk), GFP_ATOMIC);
+				if (!chunk) {
+					err = -ENOMEM;
+					break;
+				}
+				INIT_LIST_HEAD(&chunk->entry);
+			}
+		}
+		if (chunk->arr_len == 0) {
+			chunk->dest_paddr = dest_paddr;
+			tot_len = 0;
+		}
+		chunk->arr[chunk->arr_len].src_paddr = src_paddr;
+		chunk->arr[chunk->arr_len].src_len = this_len;
+		chunk->arr_len++;
+
+		dest_prev = dest_paddr + this_len;
+		prev_in_place = in_place;
+		tot_len += this_len;
+
+		err = ablkcipher_walk_done(req, walk, nbytes - this_len);
+		if (err)
+			break;
+	}
+	if (!err && chunk->arr_len != 0) {
+		chunk->dest_final = dest_prev;
+		list_add_tail(&chunk->entry, &rctx->chunk_list);
+	}
+
+	return err;
+}
+
+static void n2_chunk_complete(struct ablkcipher_request *req, void *final_iv)
+{
+	struct n2_request_context *rctx = ablkcipher_request_ctx(req);
+	struct n2_crypto_chunk *c, *tmp;
+
+	if (final_iv)
+		memcpy(rctx->walk.iv, final_iv, rctx->walk.blocksize);
+
+	ablkcipher_walk_complete(&rctx->walk);
+	list_for_each_entry_safe(c, tmp, &rctx->chunk_list, entry) {
+		list_del(&c->entry);
+		if (unlikely(c != &rctx->chunk))
+			kfree(c);
+	}
+
+}
+
+static int n2_do_ecb(struct ablkcipher_request *req, bool encrypt)
+{
+	struct n2_request_context *rctx = ablkcipher_request_ctx(req);
+	struct crypto_tfm *tfm = req->base.tfm;
+	int err = n2_compute_chunks(req);
+	struct n2_crypto_chunk *c, *tmp;
+	unsigned long flags, hv_ret;
+	struct spu_queue *qp;
+
+	if (err)
+		return err;
+
+	qp = cpu_to_cwq[get_cpu()];
+	err = -ENODEV;
+	if (!qp)
+		goto out;
+
+	spin_lock_irqsave(&qp->lock, flags);
+
+	list_for_each_entry_safe(c, tmp, &rctx->chunk_list, entry) {
+		err = __n2_crypt_chunk(tfm, c, qp, encrypt);
+		if (err)
+			break;
+		list_del(&c->entry);
+		if (unlikely(c != &rctx->chunk))
+			kfree(c);
+	}
+	if (!err) {
+		hv_ret = wait_for_tail(qp);
+		if (hv_ret != HV_EOK)
+			err = -EINVAL;
+	}
+
+	spin_unlock_irqrestore(&qp->lock, flags);
+
+	put_cpu();
+
+out:
+	n2_chunk_complete(req, NULL);
+	return err;
+}
+
+static int n2_encrypt_ecb(struct ablkcipher_request *req)
+{
+	return n2_do_ecb(req, true);
+}
+
+static int n2_decrypt_ecb(struct ablkcipher_request *req)
+{
+	return n2_do_ecb(req, false);
+}
+
+static int n2_do_chaining(struct ablkcipher_request *req, bool encrypt)
+{
+	struct n2_request_context *rctx = ablkcipher_request_ctx(req);
+	struct crypto_tfm *tfm = req->base.tfm;
+	unsigned long flags, hv_ret, iv_paddr;
+	int err = n2_compute_chunks(req);
+	struct n2_crypto_chunk *c, *tmp;
+	struct spu_queue *qp;
+	void *final_iv_addr;
+
+	final_iv_addr = NULL;
+
+	if (err)
+		return err;
+
+	qp = cpu_to_cwq[get_cpu()];
+	err = -ENODEV;
+	if (!qp)
+		goto out;
+
+	spin_lock_irqsave(&qp->lock, flags);
+
+	if (encrypt) {
+		iv_paddr = __pa(rctx->walk.iv);
+		list_for_each_entry_safe(c, tmp, &rctx->chunk_list,
+					 entry) {
+			c->iv_paddr = iv_paddr;
+			err = __n2_crypt_chunk(tfm, c, qp, true);
+			if (err)
+				break;
+			iv_paddr = c->dest_final - rctx->walk.blocksize;
+			list_del(&c->entry);
+			if (unlikely(c != &rctx->chunk))
+				kfree(c);
+		}
+		final_iv_addr = __va(iv_paddr);
+	} else {
+		list_for_each_entry_safe_reverse(c, tmp, &rctx->chunk_list,
+						 entry) {
+			if (c == &rctx->chunk) {
+				iv_paddr = __pa(rctx->walk.iv);
+			} else {
+				iv_paddr = (tmp->arr[tmp->arr_len-1].src_paddr +
+					    tmp->arr[tmp->arr_len-1].src_len -
+					    rctx->walk.blocksize);
+			}
+			if (!final_iv_addr) {
+				unsigned long pa;
+
+				pa = (c->arr[c->arr_len-1].src_paddr +
+				      c->arr[c->arr_len-1].src_len -
+				      rctx->walk.blocksize);
+				final_iv_addr = rctx->temp_iv;
+				memcpy(rctx->temp_iv, __va(pa),
+				       rctx->walk.blocksize);
+			}
+			c->iv_paddr = iv_paddr;
+			err = __n2_crypt_chunk(tfm, c, qp, false);
+			if (err)
+				break;
+			list_del(&c->entry);
+			if (unlikely(c != &rctx->chunk))
+				kfree(c);
+		}
+	}
+	if (!err) {
+		hv_ret = wait_for_tail(qp);
+		if (hv_ret != HV_EOK)
+			err = -EINVAL;
+	}
+
+	spin_unlock_irqrestore(&qp->lock, flags);
+
+	put_cpu();
+
+out:
+	n2_chunk_complete(req, err ? NULL : final_iv_addr);
+	return err;
+}
+
+static int n2_encrypt_chaining(struct ablkcipher_request *req)
+{
+	return n2_do_chaining(req, true);
+}
+
+static int n2_decrypt_chaining(struct ablkcipher_request *req)
+{
+	return n2_do_chaining(req, false);
+}
+
+struct n2_cipher_tmpl {
+	const char		*name;
+	const char		*drv_name;
+	u8			block_size;
+	u8			enc_type;
+	struct ablkcipher_alg	ablkcipher;
+};
+
+static const struct n2_cipher_tmpl cipher_tmpls[] = {
+	/* ARC4: only ECB is supported (chaining bits ignored) */
+	{	.name		= "ecb(arc4)",
+		.drv_name	= "ecb-arc4",
+		.block_size	= 1,
+		.enc_type	= (ENC_TYPE_ALG_RC4_STREAM |
+				   ENC_TYPE_CHAINING_ECB),
+		.ablkcipher	= {
+			.min_keysize	= 1,
+			.max_keysize	= 256,
+			.setkey		= n2_arc4_setkey,
+			.encrypt	= n2_encrypt_ecb,
+			.decrypt	= n2_decrypt_ecb,
+		},
+	},
+
+	/* DES: ECB CBC and CFB are supported */
+	{	.name		= "ecb(des)",
+		.drv_name	= "ecb-des",
+		.block_size	= DES_BLOCK_SIZE,
+		.enc_type	= (ENC_TYPE_ALG_DES |
+				   ENC_TYPE_CHAINING_ECB),
+		.ablkcipher	= {
+			.min_keysize	= DES_KEY_SIZE,
+			.max_keysize	= DES_KEY_SIZE,
+			.setkey		= n2_des_setkey,
+			.encrypt	= n2_encrypt_ecb,
+			.decrypt	= n2_decrypt_ecb,
+		},
+	},
+	{	.name		= "cbc(des)",
+		.drv_name	= "cbc-des",
+		.block_size	= DES_BLOCK_SIZE,
+		.enc_type	= (ENC_TYPE_ALG_DES |
+				   ENC_TYPE_CHAINING_CBC),
+		.ablkcipher	= {
+			.ivsize		= DES_BLOCK_SIZE,
+			.min_keysize	= DES_KEY_SIZE,
+			.max_keysize	= DES_KEY_SIZE,
+			.setkey		= n2_des_setkey,
+			.encrypt	= n2_encrypt_chaining,
+			.decrypt	= n2_decrypt_chaining,
+		},
+	},
+	{	.name		= "cfb(des)",
+		.drv_name	= "cfb-des",
+		.block_size	= DES_BLOCK_SIZE,
+		.enc_type	= (ENC_TYPE_ALG_DES |
+				   ENC_TYPE_CHAINING_CFB),
+		.ablkcipher	= {
+			.min_keysize	= DES_KEY_SIZE,
+			.max_keysize	= DES_KEY_SIZE,
+			.setkey		= n2_des_setkey,
+			.encrypt	= n2_encrypt_chaining,
+			.decrypt	= n2_decrypt_chaining,
+		},
+	},
+
+	/* 3DES: ECB CBC and CFB are supported */
+	{	.name		= "ecb(des3_ede)",
+		.drv_name	= "ecb-3des",
+		.block_size	= DES_BLOCK_SIZE,
+		.enc_type	= (ENC_TYPE_ALG_3DES |
+				   ENC_TYPE_CHAINING_ECB),
+		.ablkcipher	= {
+			.min_keysize	= 3 * DES_KEY_SIZE,
+			.max_keysize	= 3 * DES_KEY_SIZE,
+			.setkey		= n2_3des_setkey,
+			.encrypt	= n2_encrypt_ecb,
+			.decrypt	= n2_decrypt_ecb,
+		},
+	},
+	{	.name		= "cbc(des3_ede)",
+		.drv_name	= "cbc-3des",
+		.block_size	= DES_BLOCK_SIZE,
+		.enc_type	= (ENC_TYPE_ALG_3DES |
+				   ENC_TYPE_CHAINING_CBC),
+		.ablkcipher	= {
+			.ivsize		= DES_BLOCK_SIZE,
+			.min_keysize	= 3 * DES_KEY_SIZE,
+			.max_keysize	= 3 * DES_KEY_SIZE,
+			.setkey		= n2_3des_setkey,
+			.encrypt	= n2_encrypt_chaining,
+			.decrypt	= n2_decrypt_chaining,
+		},
+	},
+	{	.name		= "cfb(des3_ede)",
+		.drv_name	= "cfb-3des",
+		.block_size	= DES_BLOCK_SIZE,
+		.enc_type	= (ENC_TYPE_ALG_3DES |
+				   ENC_TYPE_CHAINING_CFB),
+		.ablkcipher	= {
+			.min_keysize	= 3 * DES_KEY_SIZE,
+			.max_keysize	= 3 * DES_KEY_SIZE,
+			.setkey		= n2_3des_setkey,
+			.encrypt	= n2_encrypt_chaining,
+			.decrypt	= n2_decrypt_chaining,
+		},
+	},
+	/* AES: ECB CBC and CTR are supported */
+	{	.name		= "ecb(aes)",
+		.drv_name	= "ecb-aes",
+		.block_size	= AES_BLOCK_SIZE,
+		.enc_type	= (ENC_TYPE_ALG_AES128 |
+				   ENC_TYPE_CHAINING_ECB),
+		.ablkcipher	= {
+			.min_keysize	= AES_MIN_KEY_SIZE,
+			.max_keysize	= AES_MAX_KEY_SIZE,
+			.setkey		= n2_aes_setkey,
+			.encrypt	= n2_encrypt_ecb,
+			.decrypt	= n2_decrypt_ecb,
+		},
+	},
+	{	.name		= "cbc(aes)",
+		.drv_name	= "cbc-aes",
+		.block_size	= AES_BLOCK_SIZE,
+		.enc_type	= (ENC_TYPE_ALG_AES128 |
+				   ENC_TYPE_CHAINING_CBC),
+		.ablkcipher	= {
+			.ivsize		= AES_BLOCK_SIZE,
+			.min_keysize	= AES_MIN_KEY_SIZE,
+			.max_keysize	= AES_MAX_KEY_SIZE,
+			.setkey		= n2_aes_setkey,
+			.encrypt	= n2_encrypt_chaining,
+			.decrypt	= n2_decrypt_chaining,
+		},
+	},
+	{	.name		= "ctr(aes)",
+		.drv_name	= "ctr-aes",
+		.block_size	= AES_BLOCK_SIZE,
+		.enc_type	= (ENC_TYPE_ALG_AES128 |
+				   ENC_TYPE_CHAINING_COUNTER),
+		.ablkcipher	= {
+			.ivsize		= AES_BLOCK_SIZE,
+			.min_keysize	= AES_MIN_KEY_SIZE,
+			.max_keysize	= AES_MAX_KEY_SIZE,
+			.setkey		= n2_aes_setkey,
+			.encrypt	= n2_encrypt_chaining,
+			.decrypt	= n2_encrypt_chaining,
+		},
+	},
+
+};
+#define NUM_CIPHER_TMPLS ARRAY_SIZE(cipher_tmpls)
+
+static LIST_HEAD(cipher_algs);
+
+struct n2_hash_tmpl {
+	const char	*name;
+	int		(*digest)(struct ahash_request *req);
+	u8		digest_size;
+	u8		block_size;
+};
+static const struct n2_hash_tmpl hash_tmpls[] = {
+	{ .name		= "md5",
+	  .digest	= n2_md5_async_digest,
+	  .digest_size	= MD5_DIGEST_SIZE,
+	  .block_size	= MD5_HMAC_BLOCK_SIZE },
+	{ .name		= "sha1",
+	  .digest	= n2_sha1_async_digest,
+	  .digest_size	= SHA1_DIGEST_SIZE,
+	  .block_size	= SHA1_BLOCK_SIZE },
+	{ .name		= "sha256",
+	  .digest	= n2_sha256_async_digest,
+	  .digest_size	= SHA256_DIGEST_SIZE,
+	  .block_size	= SHA256_BLOCK_SIZE },
+	{ .name		= "sha224",
+	  .digest	= n2_sha224_async_digest,
+	  .digest_size	= SHA224_DIGEST_SIZE,
+	  .block_size	= SHA224_BLOCK_SIZE },
+};
+#define NUM_HASH_TMPLS ARRAY_SIZE(hash_tmpls)
+
+struct n2_ahash_alg {
+	struct list_head	entry;
+	struct ahash_alg	alg;
+};
+static LIST_HEAD(ahash_algs);
+
+static int algs_registered;
+
+static void __n2_unregister_algs(void)
+{
+	struct n2_cipher_alg *cipher, *cipher_tmp;
+	struct n2_ahash_alg *alg, *alg_tmp;
+
+	list_for_each_entry_safe(cipher, cipher_tmp, &cipher_algs, entry) {
+		crypto_unregister_alg(&cipher->alg);
+		list_del(&cipher->entry);
+		kfree(cipher);
+	}
+	list_for_each_entry_safe(alg, alg_tmp, &ahash_algs, entry) {
+		crypto_unregister_ahash(&alg->alg);
+		list_del(&alg->entry);
+		kfree(alg);
+	}
+}
+
+static int n2_cipher_cra_init(struct crypto_tfm *tfm)
+{
+	tfm->crt_ablkcipher.reqsize = sizeof(struct n2_request_context);
+	return 0;
+}
+
+static int __devinit __n2_register_one_cipher(const struct n2_cipher_tmpl *tmpl)
+{
+	struct n2_cipher_alg *p = kzalloc(sizeof(*p), GFP_KERNEL);
+	struct crypto_alg *alg;
+	int err;
+
+	if (!p)
+		return -ENOMEM;
+
+	alg = &p->alg;
+
+	snprintf(alg->cra_name, CRYPTO_MAX_ALG_NAME, "%s", tmpl->name);
+	snprintf(alg->cra_driver_name, CRYPTO_MAX_ALG_NAME, "%s-n2", tmpl->drv_name);
+	alg->cra_priority = N2_CRA_PRIORITY;
+	alg->cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER | CRYPTO_ALG_ASYNC;
+	alg->cra_blocksize = tmpl->block_size;
+	p->enc_type = tmpl->enc_type;
+	alg->cra_ctxsize = sizeof(struct n2_cipher_context);
+	alg->cra_type = &crypto_ablkcipher_type;
+	alg->cra_u.ablkcipher = tmpl->ablkcipher;
+	alg->cra_init = n2_cipher_cra_init;
+	alg->cra_module = THIS_MODULE;
+
+	list_add(&p->entry, &cipher_algs);
+	err = crypto_register_alg(alg);
+	if (err) {
+		list_del(&p->entry);
+		kfree(p);
+	}
+	return err;
+}
+
+static int __devinit __n2_register_one_ahash(const struct n2_hash_tmpl *tmpl)
+{
+	struct n2_ahash_alg *p = kzalloc(sizeof(*p), GFP_KERNEL);
+	struct hash_alg_common *halg;
+	struct crypto_alg *base;
+	struct ahash_alg *ahash;
+	int err;
+
+	if (!p)
+		return -ENOMEM;
+
+	ahash = &p->alg;
+	ahash->init = n2_hash_async_init;
+	ahash->update = n2_hash_async_update;
+	ahash->final = n2_hash_async_final;
+	ahash->finup = n2_hash_async_finup;
+	ahash->digest = tmpl->digest;
+
+	halg = &ahash->halg;
+	halg->digestsize = tmpl->digest_size;
+
+	base = &halg->base;
+	snprintf(base->cra_name, CRYPTO_MAX_ALG_NAME, "%s", tmpl->name);
+	snprintf(base->cra_driver_name, CRYPTO_MAX_ALG_NAME, "%s-n2", tmpl->name);
+	base->cra_priority = N2_CRA_PRIORITY;
+	base->cra_flags = CRYPTO_ALG_TYPE_AHASH | CRYPTO_ALG_NEED_FALLBACK;
+	base->cra_blocksize = tmpl->block_size;
+	base->cra_ctxsize = sizeof(struct n2_hash_ctx);
+	base->cra_module = THIS_MODULE;
+	base->cra_init = n2_hash_cra_init;
+	base->cra_exit = n2_hash_cra_exit;
+
+	list_add(&p->entry, &ahash_algs);
+	err = crypto_register_ahash(ahash);
+	if (err) {
+		list_del(&p->entry);
+		kfree(p);
+	}
+	return err;
+}
+
+static int __devinit n2_register_algs(void)
+{
+	int i, err = 0;
+
+	mutex_lock(&spu_lock);
+	if (algs_registered++)
+		goto out;
+
+	for (i = 0; i < NUM_HASH_TMPLS; i++) {
+		err = __n2_register_one_ahash(&hash_tmpls[i]);
+		if (err) {
+			__n2_unregister_algs();
+			goto out;
+		}
+	}
+	for (i = 0; i < NUM_CIPHER_TMPLS; i++) {
+		err = __n2_register_one_cipher(&cipher_tmpls[i]);
+		if (err) {
+			__n2_unregister_algs();
+			goto out;
+		}
+	}
+
+out:
+	mutex_unlock(&spu_lock);
+	return err;
+}
+
+static void __exit n2_unregister_algs(void)
+{
+	mutex_lock(&spu_lock);
+	if (!--algs_registered)
+		__n2_unregister_algs();
+	mutex_unlock(&spu_lock);
+}
+
+/* To map CWQ queues to interrupt sources, the hypervisor API provides
+ * a devino.  This isn't very useful to us because all of the
+ * interrupts listed in the of_device node have been translated to
+ * Linux virtual IRQ cookie numbers.
+ *
+ * So we have to back-translate, going through the 'intr' and 'ino'
+ * property tables of the n2cp MDESC node, matching it with the OF
+ * 'interrupts' property entries, in order to to figure out which
+ * devino goes to which already-translated IRQ.
+ */
+static int find_devino_index(struct of_device *dev, struct spu_mdesc_info *ip,
+			     unsigned long dev_ino)
+{
+	const unsigned int *dev_intrs;
+	unsigned int intr;
+	int i;
+
+	for (i = 0; i < ip->num_intrs; i++) {
+		if (ip->ino_table[i].ino == dev_ino)
+			break;
+	}
+	if (i == ip->num_intrs)
+		return -ENODEV;
+
+	intr = ip->ino_table[i].intr;
+
+	dev_intrs = of_get_property(dev->dev.of_node, "interrupts", NULL);
+	if (!dev_intrs)
+		return -ENODEV;
+
+	for (i = 0; i < dev->num_irqs; i++) {
+		if (dev_intrs[i] == intr)
+			return i;
+	}
+
+	return -ENODEV;
+}
+
+static int spu_map_ino(struct of_device *dev, struct spu_mdesc_info *ip,
+		       const char *irq_name, struct spu_queue *p,
+		       irq_handler_t handler)
+{
+	unsigned long herr;
+	int index;
+
+	herr = sun4v_ncs_qhandle_to_devino(p->qhandle, &p->devino);
+	if (herr)
+		return -EINVAL;
+
+	index = find_devino_index(dev, ip, p->devino);
+	if (index < 0)
+		return index;
+
+	p->irq = dev->irqs[index];
+
+	sprintf(p->irq_name, "%s-%d", irq_name, index);
+
+	return request_irq(p->irq, handler, IRQF_SAMPLE_RANDOM,
+			   p->irq_name, p);
+}
+
+static struct kmem_cache *queue_cache[2];
+
+static void *new_queue(unsigned long q_type)
+{
+	return kmem_cache_zalloc(queue_cache[q_type - 1], GFP_KERNEL);
+}
+
+static void free_queue(void *p, unsigned long q_type)
+{
+	return kmem_cache_free(queue_cache[q_type - 1], p);
+}
+
+static int queue_cache_init(void)
+{
+	if (!queue_cache[HV_NCS_QTYPE_MAU - 1])
+		queue_cache[HV_NCS_QTYPE_MAU - 1] =
+			kmem_cache_create("mau_queue",
+					  (MAU_NUM_ENTRIES *
+					   MAU_ENTRY_SIZE),
+					  MAU_ENTRY_SIZE, 0, NULL);
+	if (!queue_cache[HV_NCS_QTYPE_MAU - 1])
+		return -ENOMEM;
+
+	if (!queue_cache[HV_NCS_QTYPE_CWQ - 1])
+		queue_cache[HV_NCS_QTYPE_CWQ - 1] =
+			kmem_cache_create("cwq_queue",
+					  (CWQ_NUM_ENTRIES *
+					   CWQ_ENTRY_SIZE),
+					  CWQ_ENTRY_SIZE, 0, NULL);
+	if (!queue_cache[HV_NCS_QTYPE_CWQ - 1]) {
+		kmem_cache_destroy(queue_cache[HV_NCS_QTYPE_MAU - 1]);
+		return -ENOMEM;
+	}
+	return 0;
+}
+
+static void queue_cache_destroy(void)
+{
+	kmem_cache_destroy(queue_cache[HV_NCS_QTYPE_MAU - 1]);
+	kmem_cache_destroy(queue_cache[HV_NCS_QTYPE_CWQ - 1]);
+}
+
+static int spu_queue_register(struct spu_queue *p, unsigned long q_type)
+{
+	cpumask_var_t old_allowed;
+	unsigned long hv_ret;
+
+	if (cpumask_empty(&p->sharing))
+		return -EINVAL;
+
+	if (!alloc_cpumask_var(&old_allowed, GFP_KERNEL))
+		return -ENOMEM;
+
+	cpumask_copy(old_allowed, &current->cpus_allowed);
+
+	set_cpus_allowed_ptr(current, &p->sharing);
+
+	hv_ret = sun4v_ncs_qconf(q_type, __pa(p->q),
+				 CWQ_NUM_ENTRIES, &p->qhandle);
+	if (!hv_ret)
+		sun4v_ncs_sethead_marker(p->qhandle, 0);
+
+	set_cpus_allowed_ptr(current, old_allowed);
+
+	free_cpumask_var(old_allowed);
+
+	return (hv_ret ? -EINVAL : 0);
+}
+
+static int spu_queue_setup(struct spu_queue *p)
+{
+	int err;
+
+	p->q = new_queue(p->q_type);
+	if (!p->q)
+		return -ENOMEM;
+
+	err = spu_queue_register(p, p->q_type);
+	if (err) {
+		free_queue(p->q, p->q_type);
+		p->q = NULL;
+	}
+
+	return err;
+}
+
+static void spu_queue_destroy(struct spu_queue *p)
+{
+	unsigned long hv_ret;
+
+	if (!p->q)
+		return;
+
+	hv_ret = sun4v_ncs_qconf(p->q_type, p->qhandle, 0, &p->qhandle);
+
+	if (!hv_ret)
+		free_queue(p->q, p->q_type);
+}
+
+static void spu_list_destroy(struct list_head *list)
+{
+	struct spu_queue *p, *n;
+
+	list_for_each_entry_safe(p, n, list, list) {
+		int i;
+
+		for (i = 0; i < NR_CPUS; i++) {
+			if (cpu_to_cwq[i] == p)
+				cpu_to_cwq[i] = NULL;
+		}
+
+		if (p->irq) {
+			free_irq(p->irq, p);
+			p->irq = 0;
+		}
+		spu_queue_destroy(p);
+		list_del(&p->list);
+		kfree(p);
+	}
+}
+
+/* Walk the backward arcs of a CWQ 'exec-unit' node,
+ * gathering cpu membership information.
+ */
+static int spu_mdesc_walk_arcs(struct mdesc_handle *mdesc,
+			       struct of_device *dev,
+			       u64 node, struct spu_queue *p,
+			       struct spu_queue **table)
+{
+	u64 arc;
+
+	mdesc_for_each_arc(arc, mdesc, node, MDESC_ARC_TYPE_BACK) {
+		u64 tgt = mdesc_arc_target(mdesc, arc);
+		const char *name = mdesc_node_name(mdesc, tgt);
+		const u64 *id;
+
+		if (strcmp(name, "cpu"))
+			continue;
+		id = mdesc_get_property(mdesc, tgt, "id", NULL);
+		if (table[*id] != NULL) {
+			dev_err(&dev->dev, "%s: SPU cpu slot already set.\n",
+				dev->dev.of_node->full_name);
+			return -EINVAL;
+		}
+		cpu_set(*id, p->sharing);
+		table[*id] = p;
+	}
+	return 0;
+}
+
+/* Process an 'exec-unit' MDESC node of type 'cwq'.  */
+static int handle_exec_unit(struct spu_mdesc_info *ip, struct list_head *list,
+			    struct of_device *dev, struct mdesc_handle *mdesc,
+			    u64 node, const char *iname, unsigned long q_type,
+			    irq_handler_t handler, struct spu_queue **table)
+{
+	struct spu_queue *p;
+	int err;
+
+	p = kzalloc(sizeof(struct spu_queue), GFP_KERNEL);
+	if (!p) {
+		dev_err(&dev->dev, "%s: Could not allocate SPU queue.\n",
+			dev->dev.of_node->full_name);
+		return -ENOMEM;
+	}
+
+	cpus_clear(p->sharing);
+	spin_lock_init(&p->lock);
+	p->q_type = q_type;
+	INIT_LIST_HEAD(&p->jobs);
+	list_add(&p->list, list);
+
+	err = spu_mdesc_walk_arcs(mdesc, dev, node, p, table);
+	if (err)
+		return err;
+
+	err = spu_queue_setup(p);
+	if (err)
+		return err;
+
+	return spu_map_ino(dev, ip, iname, p, handler);
+}
+
+static int spu_mdesc_scan(struct mdesc_handle *mdesc, struct of_device *dev,
+			  struct spu_mdesc_info *ip, struct list_head *list,
+			  const char *exec_name, unsigned long q_type,
+			  irq_handler_t handler, struct spu_queue **table)
+{
+	int err = 0;
+	u64 node;
+
+	mdesc_for_each_node_by_name(mdesc, node, "exec-unit") {
+		const char *type;
+
+		type = mdesc_get_property(mdesc, node, "type", NULL);
+		if (!type || strcmp(type, exec_name))
+			continue;
+
+		err = handle_exec_unit(ip, list, dev, mdesc, node,
+				       exec_name, q_type, handler, table);
+		if (err) {
+			spu_list_destroy(list);
+			break;
+		}
+	}
+
+	return err;
+}
+
+static int __devinit get_irq_props(struct mdesc_handle *mdesc, u64 node,
+				   struct spu_mdesc_info *ip)
+{
+	const u64 *intr, *ino;
+	int intr_len, ino_len;
+	int i;
+
+	intr = mdesc_get_property(mdesc, node, "intr", &intr_len);
+	if (!intr)
+		return -ENODEV;
+
+	ino = mdesc_get_property(mdesc, node, "ino", &ino_len);
+	if (!intr)
+		return -ENODEV;
+
+	if (intr_len != ino_len)
+		return -EINVAL;
+
+	ip->num_intrs = intr_len / sizeof(u64);
+	ip->ino_table = kzalloc((sizeof(struct ino_blob) *
+				 ip->num_intrs),
+				GFP_KERNEL);
+	if (!ip->ino_table)
+		return -ENOMEM;
+
+	for (i = 0; i < ip->num_intrs; i++) {
+		struct ino_blob *b = &ip->ino_table[i];
+		b->intr = intr[i];
+		b->ino = ino[i];
+	}
+
+	return 0;
+}
+
+static int __devinit grab_mdesc_irq_props(struct mdesc_handle *mdesc,
+					  struct of_device *dev,
+					  struct spu_mdesc_info *ip,
+					  const char *node_name)
+{
+	const unsigned int *reg;
+	u64 node;
+
+	reg = of_get_property(dev->dev.of_node, "reg", NULL);
+	if (!reg)
+		return -ENODEV;
+
+	mdesc_for_each_node_by_name(mdesc, node, "virtual-device") {
+		const char *name;
+		const u64 *chdl;
+
+		name = mdesc_get_property(mdesc, node, "name", NULL);
+		if (!name || strcmp(name, node_name))
+			continue;
+		chdl = mdesc_get_property(mdesc, node, "cfg-handle", NULL);
+		if (!chdl || (*chdl != *reg))
+			continue;
+		ip->cfg_handle = *chdl;
+		return get_irq_props(mdesc, node, ip);
+	}
+
+	return -ENODEV;
+}
+
+static unsigned long n2_spu_hvapi_major;
+static unsigned long n2_spu_hvapi_minor;
+
+static int __devinit n2_spu_hvapi_register(void)
+{
+	int err;
+
+	n2_spu_hvapi_major = 2;
+	n2_spu_hvapi_minor = 0;
+
+	err = sun4v_hvapi_register(HV_GRP_NCS,
+				   n2_spu_hvapi_major,
+				   &n2_spu_hvapi_minor);
+
+	if (!err)
+		pr_info("Registered NCS HVAPI version %lu.%lu\n",
+			n2_spu_hvapi_major,
+			n2_spu_hvapi_minor);
+
+	return err;
+}
+
+static void n2_spu_hvapi_unregister(void)
+{
+	sun4v_hvapi_unregister(HV_GRP_NCS);
+}
+
+static int global_ref;
+
+static int __devinit grab_global_resources(void)
+{
+	int err = 0;
+
+	mutex_lock(&spu_lock);
+
+	if (global_ref++)
+		goto out;
+
+	err = n2_spu_hvapi_register();
+	if (err)
+		goto out;
+
+	err = queue_cache_init();
+	if (err)
+		goto out_hvapi_release;
+
+	err = -ENOMEM;
+	cpu_to_cwq = kzalloc(sizeof(struct spu_queue *) * NR_CPUS,
+			     GFP_KERNEL);
+	if (!cpu_to_cwq)
+		goto out_queue_cache_destroy;
+
+	cpu_to_mau = kzalloc(sizeof(struct spu_queue *) * NR_CPUS,
+			     GFP_KERNEL);
+	if (!cpu_to_mau)
+		goto out_free_cwq_table;
+
+	err = 0;
+
+out:
+	if (err)
+		global_ref--;
+	mutex_unlock(&spu_lock);
+	return err;
+
+out_free_cwq_table:
+	kfree(cpu_to_cwq);
+	cpu_to_cwq = NULL;
+
+out_queue_cache_destroy:
+	queue_cache_destroy();
+
+out_hvapi_release:
+	n2_spu_hvapi_unregister();
+	goto out;
+}
+
+static void release_global_resources(void)
+{
+	mutex_lock(&spu_lock);
+	if (!--global_ref) {
+		kfree(cpu_to_cwq);
+		cpu_to_cwq = NULL;
+
+		kfree(cpu_to_mau);
+		cpu_to_mau = NULL;
+
+		queue_cache_destroy();
+		n2_spu_hvapi_unregister();
+	}
+	mutex_unlock(&spu_lock);
+}
+
+static struct n2_crypto * __devinit alloc_n2cp(void)
+{
+	struct n2_crypto *np = kzalloc(sizeof(struct n2_crypto), GFP_KERNEL);
+
+	if (np)
+		INIT_LIST_HEAD(&np->cwq_list);
+
+	return np;
+}
+
+static void free_n2cp(struct n2_crypto *np)
+{
+	if (np->cwq_info.ino_table) {
+		kfree(np->cwq_info.ino_table);
+		np->cwq_info.ino_table = NULL;
+	}
+
+	kfree(np);
+}
+
+static void __devinit n2_spu_driver_version(void)
+{
+	static int n2_spu_version_printed;
+
+	if (n2_spu_version_printed++ == 0)
+		pr_info("%s", version);
+}
+
+static int __devinit n2_crypto_probe(struct of_device *dev,
+				     const struct of_device_id *match)
+{
+	struct mdesc_handle *mdesc;
+	const char *full_name;
+	struct n2_crypto *np;
+	int err;
+
+	n2_spu_driver_version();
+
+	full_name = dev->dev.of_node->full_name;
+	pr_info("Found N2CP at %s\n", full_name);
+
+	np = alloc_n2cp();
+	if (!np) {
+		dev_err(&dev->dev, "%s: Unable to allocate n2cp.\n",
+			full_name);
+		return -ENOMEM;
+	}
+
+	err = grab_global_resources();
+	if (err) {
+		dev_err(&dev->dev, "%s: Unable to grab "
+			"global resources.\n", full_name);
+		goto out_free_n2cp;
+	}
+
+	mdesc = mdesc_grab();
+
+	if (!mdesc) {
+		dev_err(&dev->dev, "%s: Unable to grab MDESC.\n",
+			full_name);
+		err = -ENODEV;
+		goto out_free_global;
+	}
+	err = grab_mdesc_irq_props(mdesc, dev, &np->cwq_info, "n2cp");
+	if (err) {
+		dev_err(&dev->dev, "%s: Unable to grab IRQ props.\n",
+			full_name);
+		mdesc_release(mdesc);
+		goto out_free_global;
+	}
+
+	err = spu_mdesc_scan(mdesc, dev, &np->cwq_info, &np->cwq_list,
+			     "cwq", HV_NCS_QTYPE_CWQ, cwq_intr,
+			     cpu_to_cwq);
+	mdesc_release(mdesc);
+
+	if (err) {
+		dev_err(&dev->dev, "%s: CWQ MDESC scan failed.\n",
+			full_name);
+		goto out_free_global;
+	}
+
+	err = n2_register_algs();
+	if (err) {
+		dev_err(&dev->dev, "%s: Unable to register algorithms.\n",
+			full_name);
+		goto out_free_spu_list;
+	}
+
+	dev_set_drvdata(&dev->dev, np);
+
+	return 0;
+
+out_free_spu_list:
+	spu_list_destroy(&np->cwq_list);
+
+out_free_global:
+	release_global_resources();
+
+out_free_n2cp:
+	free_n2cp(np);
+
+	return err;
+}
+
+static int __devexit n2_crypto_remove(struct of_device *dev)
+{
+	struct n2_crypto *np = dev_get_drvdata(&dev->dev);
+
+	n2_unregister_algs();
+
+	spu_list_destroy(&np->cwq_list);
+
+	release_global_resources();
+
+	free_n2cp(np);
+
+	return 0;
+}
+
+static struct n2_mau * __devinit alloc_ncp(void)
+{
+	struct n2_mau *mp = kzalloc(sizeof(struct n2_mau), GFP_KERNEL);
+
+	if (mp)
+		INIT_LIST_HEAD(&mp->mau_list);
+
+	return mp;
+}
+
+static void free_ncp(struct n2_mau *mp)
+{
+	if (mp->mau_info.ino_table) {
+		kfree(mp->mau_info.ino_table);
+		mp->mau_info.ino_table = NULL;
+	}
+
+	kfree(mp);
+}
+
+static int __devinit n2_mau_probe(struct of_device *dev,
+				     const struct of_device_id *match)
+{
+	struct mdesc_handle *mdesc;
+	const char *full_name;
+	struct n2_mau *mp;
+	int err;
+
+	n2_spu_driver_version();
+
+	full_name = dev->dev.of_node->full_name;
+	pr_info("Found NCP at %s\n", full_name);
+
+	mp = alloc_ncp();
+	if (!mp) {
+		dev_err(&dev->dev, "%s: Unable to allocate ncp.\n",
+			full_name);
+		return -ENOMEM;
+	}
+
+	err = grab_global_resources();
+	if (err) {
+		dev_err(&dev->dev, "%s: Unable to grab "
+			"global resources.\n", full_name);
+		goto out_free_ncp;
+	}
+
+	mdesc = mdesc_grab();
+
+	if (!mdesc) {
+		dev_err(&dev->dev, "%s: Unable to grab MDESC.\n",
+			full_name);
+		err = -ENODEV;
+		goto out_free_global;
+	}
+
+	err = grab_mdesc_irq_props(mdesc, dev, &mp->mau_info, "ncp");
+	if (err) {
+		dev_err(&dev->dev, "%s: Unable to grab IRQ props.\n",
+			full_name);
+		mdesc_release(mdesc);
+		goto out_free_global;
+	}
+
+	err = spu_mdesc_scan(mdesc, dev, &mp->mau_info, &mp->mau_list,
+			     "mau", HV_NCS_QTYPE_MAU, mau_intr,
+			     cpu_to_mau);
+	mdesc_release(mdesc);
+
+	if (err) {
+		dev_err(&dev->dev, "%s: MAU MDESC scan failed.\n",
+			full_name);
+		goto out_free_global;
+	}
+
+	dev_set_drvdata(&dev->dev, mp);
+
+	return 0;
+
+out_free_global:
+	release_global_resources();
+
+out_free_ncp:
+	free_ncp(mp);
+
+	return err;
+}
+
+static int __devexit n2_mau_remove(struct of_device *dev)
+{
+	struct n2_mau *mp = dev_get_drvdata(&dev->dev);
+
+	spu_list_destroy(&mp->mau_list);
+
+	release_global_resources();
+
+	free_ncp(mp);
+
+	return 0;
+}
+
+static struct of_device_id n2_crypto_match[] = {
+	{
+		.name = "n2cp",
+		.compatible = "SUNW,n2-cwq",
+	},
+	{
+		.name = "n2cp",
+		.compatible = "SUNW,vf-cwq",
+	},
+	{},
+};
+
+MODULE_DEVICE_TABLE(of, n2_crypto_match);
+
+static struct of_platform_driver n2_crypto_driver = {
+	.driver = {
+		.name		=	"n2cp",
+		.owner		=	THIS_MODULE,
+		.of_match_table	=	n2_crypto_match,
+	},
+	.probe		=	n2_crypto_probe,
+	.remove		=	__devexit_p(n2_crypto_remove),
+};
+
+static struct of_device_id n2_mau_match[] = {
+	{
+		.name = "ncp",
+		.compatible = "SUNW,n2-mau",
+	},
+	{
+		.name = "ncp",
+		.compatible = "SUNW,vf-mau",
+	},
+	{},
+};
+
+MODULE_DEVICE_TABLE(of, n2_mau_match);
+
+static struct of_platform_driver n2_mau_driver = {
+	.driver = {
+		.name		=	"ncp",
+		.owner		=	THIS_MODULE,
+		.of_match_table	=	n2_mau_match,
+	},
+	.probe		=	n2_mau_probe,
+	.remove		=	__devexit_p(n2_mau_remove),
+};
+
+static int __init n2_init(void)
+{
+	int err = of_register_driver(&n2_crypto_driver, &of_bus_type);
+
+	if (!err) {
+		err = of_register_driver(&n2_mau_driver, &of_bus_type);
+		if (err)
+			of_unregister_driver(&n2_crypto_driver);
+	}
+	return err;
+}
+
+static void __exit n2_exit(void)
+{
+	of_unregister_driver(&n2_mau_driver);
+	of_unregister_driver(&n2_crypto_driver);
+}
+
+module_init(n2_init);
+module_exit(n2_exit);
diff --git a/drivers/crypto/n2_core.h b/drivers/crypto/n2_core.h
new file mode 100644
index 000000000000..4bcbbeae98f5
--- /dev/null
+++ b/drivers/crypto/n2_core.h
@@ -0,0 +1,231 @@
+#ifndef _N2_CORE_H
+#define _N2_CORE_H
+
+#ifndef __ASSEMBLY__
+
+struct ino_blob {
+	u64			intr;
+	u64			ino;
+};
+
+struct spu_mdesc_info {
+	u64			cfg_handle;
+	struct ino_blob		*ino_table;
+	int			num_intrs;
+};
+
+struct n2_crypto {
+	struct spu_mdesc_info	cwq_info;
+	struct list_head	cwq_list;
+};
+
+struct n2_mau {
+	struct spu_mdesc_info	mau_info;
+	struct list_head	mau_list;
+};
+
+#define CWQ_ENTRY_SIZE		64
+#define CWQ_NUM_ENTRIES		64
+
+#define MAU_ENTRY_SIZE		64
+#define MAU_NUM_ENTRIES		64
+
+struct cwq_initial_entry {
+	u64			control;
+	u64			src_addr;
+	u64			auth_key_addr;
+	u64			auth_iv_addr;
+	u64			final_auth_state_addr;
+	u64			enc_key_addr;
+	u64			enc_iv_addr;
+	u64			dest_addr;
+};
+
+struct cwq_ext_entry {
+	u64			len;
+	u64			src_addr;
+	u64			resv1;
+	u64			resv2;
+	u64			resv3;
+	u64			resv4;
+	u64			resv5;
+	u64			resv6;
+};
+
+struct cwq_final_entry {
+	u64			control;
+	u64			src_addr;
+	u64			resv1;
+	u64			resv2;
+	u64			resv3;
+	u64			resv4;
+	u64			resv5;
+	u64			resv6;
+};
+
+#define CONTROL_LEN			0x000000000000ffffULL
+#define CONTROL_LEN_SHIFT		0
+#define CONTROL_HMAC_KEY_LEN		0x0000000000ff0000ULL
+#define CONTROL_HMAC_KEY_LEN_SHIFT	16
+#define CONTROL_ENC_TYPE		0x00000000ff000000ULL
+#define CONTROL_ENC_TYPE_SHIFT		24
+#define  ENC_TYPE_ALG_RC4_STREAM	0x00ULL
+#define  ENC_TYPE_ALG_RC4_NOSTREAM	0x04ULL
+#define  ENC_TYPE_ALG_DES		0x08ULL
+#define  ENC_TYPE_ALG_3DES		0x0cULL
+#define  ENC_TYPE_ALG_AES128		0x10ULL
+#define  ENC_TYPE_ALG_AES192		0x14ULL
+#define  ENC_TYPE_ALG_AES256		0x18ULL
+#define  ENC_TYPE_ALG_RESERVED		0x1cULL
+#define  ENC_TYPE_ALG_MASK		0x1cULL
+#define  ENC_TYPE_CHAINING_ECB		0x00ULL
+#define  ENC_TYPE_CHAINING_CBC		0x01ULL
+#define  ENC_TYPE_CHAINING_CFB		0x02ULL
+#define  ENC_TYPE_CHAINING_COUNTER	0x03ULL
+#define  ENC_TYPE_CHAINING_MASK		0x03ULL
+#define CONTROL_AUTH_TYPE		0x0000001f00000000ULL
+#define CONTROL_AUTH_TYPE_SHIFT		32
+#define  AUTH_TYPE_RESERVED		0x00ULL
+#define  AUTH_TYPE_MD5			0x01ULL
+#define  AUTH_TYPE_SHA1			0x02ULL
+#define  AUTH_TYPE_SHA256		0x03ULL
+#define  AUTH_TYPE_CRC32		0x04ULL
+#define  AUTH_TYPE_HMAC_MD5		0x05ULL
+#define  AUTH_TYPE_HMAC_SHA1		0x06ULL
+#define  AUTH_TYPE_HMAC_SHA256		0x07ULL
+#define  AUTH_TYPE_TCP_CHECKSUM		0x08ULL
+#define  AUTH_TYPE_SSL_HMAC_MD5		0x09ULL
+#define  AUTH_TYPE_SSL_HMAC_SHA1	0x0aULL
+#define  AUTH_TYPE_SSL_HMAC_SHA256	0x0bULL
+#define CONTROL_STRAND			0x000000e000000000ULL
+#define CONTROL_STRAND_SHIFT		37
+#define CONTROL_HASH_LEN		0x0000ff0000000000ULL
+#define CONTROL_HASH_LEN_SHIFT		40
+#define CONTROL_INTERRUPT		0x0001000000000000ULL
+#define CONTROL_STORE_FINAL_AUTH_STATE	0x0002000000000000ULL
+#define CONTROL_RESERVED		0x001c000000000000ULL
+#define CONTROL_HV_DONE			0x0004000000000000ULL
+#define CONTROL_HV_PROTOCOL_ERROR	0x0008000000000000ULL
+#define CONTROL_HV_HARDWARE_ERROR	0x0010000000000000ULL
+#define CONTROL_END_OF_BLOCK		0x0020000000000000ULL
+#define CONTROL_START_OF_BLOCK		0x0040000000000000ULL
+#define CONTROL_ENCRYPT			0x0080000000000000ULL
+#define CONTROL_OPCODE			0xff00000000000000ULL
+#define CONTROL_OPCODE_SHIFT		56
+#define  OPCODE_INPLACE_BIT		0x80ULL
+#define  OPCODE_SSL_KEYBLOCK		0x10ULL
+#define  OPCODE_COPY			0x20ULL
+#define  OPCODE_ENCRYPT			0x40ULL
+#define  OPCODE_AUTH_MAC		0x41ULL
+
+#endif /* !(__ASSEMBLY__) */
+
+/* NCS v2.0 hypervisor interfaces */
+#define HV_NCS_QTYPE_MAU		0x01
+#define HV_NCS_QTYPE_CWQ		0x02
+
+/* ncs_qconf()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_NCS_QCONF
+ * ARG0:	Queue type (HV_NCS_QTYPE_{MAU,CWQ})
+ * ARG1:	Real address of queue, or handle for unconfigure
+ * ARG2:	Number of entries in queue, zero for unconfigure
+ * RET0:	status
+ * RET1:	queue handle
+ *
+ * Configure a queue in the stream processing unit.
+ *
+ * The real address given as the base must be 64-byte
+ * aligned.
+ *
+ * The queue size can range from a minimum of 2 to a maximum
+ * of 64.  The queue size must be a power of two.
+ *
+ * To unconfigure a queue, specify a length of zero and place
+ * the queue handle into ARG1.
+ *
+ * On configure success the hypervisor will set the FIRST, HEAD,
+ * and TAIL registers to the address of the first entry in the
+ * queue.  The LAST register will be set to point to the last
+ * entry in the queue.
+ */
+#define HV_FAST_NCS_QCONF		0x111
+
+/* ncs_qinfo()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_NCS_QINFO
+ * ARG0:	Queue handle
+ * RET0:	status
+ * RET1:	Queue type (HV_NCS_QTYPE_{MAU,CWQ})
+ * RET2:	Queue base address
+ * RET3:	Number of entries
+ */
+#define HV_FAST_NCS_QINFO		0x112
+
+/* ncs_gethead()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_NCS_GETHEAD
+ * ARG0:	Queue handle
+ * RET0:	status
+ * RET1:	queue head offset
+ */
+#define HV_FAST_NCS_GETHEAD		0x113
+
+/* ncs_gettail()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_NCS_GETTAIL
+ * ARG0:	Queue handle
+ * RET0:	status
+ * RET1:	queue tail offset
+ */
+#define HV_FAST_NCS_GETTAIL		0x114
+
+/* ncs_settail()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_NCS_SETTAIL
+ * ARG0:	Queue handle
+ * ARG1:	New tail offset
+ * RET0:	status
+ */
+#define HV_FAST_NCS_SETTAIL		0x115
+
+/* ncs_qhandle_to_devino()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_NCS_QHANDLE_TO_DEVINO
+ * ARG0:	Queue handle
+ * RET0:	status
+ * RET1:	devino
+ */
+#define HV_FAST_NCS_QHANDLE_TO_DEVINO	0x116
+
+/* ncs_sethead_marker()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_NCS_SETHEAD_MARKER
+ * ARG0:	Queue handle
+ * ARG1:	New head offset
+ * RET0:	status
+ */
+#define HV_FAST_NCS_SETHEAD_MARKER	0x117
+
+#ifndef __ASSEMBLY__
+extern unsigned long sun4v_ncs_qconf(unsigned long queue_type,
+				     unsigned long queue_ra,
+				     unsigned long num_entries,
+				     unsigned long *qhandle);
+extern unsigned long sun4v_ncs_qinfo(unsigned long qhandle,
+				     unsigned long *queue_type,
+				     unsigned long *queue_ra,
+				     unsigned long *num_entries);
+extern unsigned long sun4v_ncs_gethead(unsigned long qhandle,
+				       unsigned long *head);
+extern unsigned long sun4v_ncs_gettail(unsigned long qhandle,
+				       unsigned long *tail);
+extern unsigned long sun4v_ncs_settail(unsigned long qhandle,
+				       unsigned long tail);
+extern unsigned long sun4v_ncs_qhandle_to_devino(unsigned long qhandle,
+						 unsigned long *devino);
+extern unsigned long sun4v_ncs_sethead_marker(unsigned long qhandle,
+					      unsigned long head);
+#endif /* !(__ASSEMBLY__) */
+
+#endif /* _N2_CORE_H */
diff --git a/drivers/crypto/omap-sham.c b/drivers/crypto/omap-sham.c
new file mode 100644
index 000000000000..8b034337793f
--- /dev/null
+++ b/drivers/crypto/omap-sham.c
@@ -0,0 +1,1259 @@
+/*
+ * Cryptographic API.
+ *
+ * Support for OMAP SHA1/MD5 HW acceleration.
+ *
+ * Copyright (c) 2010 Nokia Corporation
+ * Author: Dmitry Kasatkin <dmitry.kasatkin@nokia.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as published
+ * by the Free Software Foundation.
+ *
+ * Some ideas are from old omap-sha1-md5.c driver.
+ */
+
+#define pr_fmt(fmt) "%s: " fmt, __func__
+
+#include <linux/version.h>
+#include <linux/err.h>
+#include <linux/device.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/errno.h>
+#include <linux/interrupt.h>
+#include <linux/kernel.h>
+#include <linux/clk.h>
+#include <linux/irq.h>
+#include <linux/io.h>
+#include <linux/platform_device.h>
+#include <linux/scatterlist.h>
+#include <linux/dma-mapping.h>
+#include <linux/delay.h>
+#include <linux/crypto.h>
+#include <linux/cryptohash.h>
+#include <crypto/scatterwalk.h>
+#include <crypto/algapi.h>
+#include <crypto/sha.h>
+#include <crypto/hash.h>
+#include <crypto/internal/hash.h>
+
+#include <plat/cpu.h>
+#include <plat/dma.h>
+#include <mach/irqs.h>
+
+#define SHA_REG_DIGEST(x)		(0x00 + ((x) * 0x04))
+#define SHA_REG_DIN(x)			(0x1C + ((x) * 0x04))
+
+#define SHA1_MD5_BLOCK_SIZE		SHA1_BLOCK_SIZE
+#define MD5_DIGEST_SIZE			16
+
+#define SHA_REG_DIGCNT			0x14
+
+#define SHA_REG_CTRL			0x18
+#define SHA_REG_CTRL_LENGTH		(0xFFFFFFFF << 5)
+#define SHA_REG_CTRL_CLOSE_HASH		(1 << 4)
+#define SHA_REG_CTRL_ALGO_CONST		(1 << 3)
+#define SHA_REG_CTRL_ALGO		(1 << 2)
+#define SHA_REG_CTRL_INPUT_READY	(1 << 1)
+#define SHA_REG_CTRL_OUTPUT_READY	(1 << 0)
+
+#define SHA_REG_REV			0x5C
+#define SHA_REG_REV_MAJOR		0xF0
+#define SHA_REG_REV_MINOR		0x0F
+
+#define SHA_REG_MASK			0x60
+#define SHA_REG_MASK_DMA_EN		(1 << 3)
+#define SHA_REG_MASK_IT_EN		(1 << 2)
+#define SHA_REG_MASK_SOFTRESET		(1 << 1)
+#define SHA_REG_AUTOIDLE		(1 << 0)
+
+#define SHA_REG_SYSSTATUS		0x64
+#define SHA_REG_SYSSTATUS_RESETDONE	(1 << 0)
+
+#define DEFAULT_TIMEOUT_INTERVAL	HZ
+
+#define FLAGS_FIRST		0x0001
+#define FLAGS_FINUP		0x0002
+#define FLAGS_FINAL		0x0004
+#define FLAGS_FAST		0x0008
+#define FLAGS_SHA1		0x0010
+#define FLAGS_DMA_ACTIVE	0x0020
+#define FLAGS_OUTPUT_READY	0x0040
+#define FLAGS_CLEAN		0x0080
+#define FLAGS_INIT		0x0100
+#define FLAGS_CPU		0x0200
+#define FLAGS_HMAC		0x0400
+
+/* 3rd byte */
+#define FLAGS_BUSY		16
+
+#define OP_UPDATE	1
+#define OP_FINAL	2
+
+struct omap_sham_dev;
+
+struct omap_sham_reqctx {
+	struct omap_sham_dev	*dd;
+	unsigned long		flags;
+	unsigned long		op;
+
+	size_t			digcnt;
+	u8			*buffer;
+	size_t			bufcnt;
+	size_t			buflen;
+	dma_addr_t		dma_addr;
+
+	/* walk state */
+	struct scatterlist	*sg;
+	unsigned int		offset;	/* offset in current sg */
+	unsigned int		total;	/* total request */
+};
+
+struct omap_sham_hmac_ctx {
+	struct crypto_shash	*shash;
+	u8			ipad[SHA1_MD5_BLOCK_SIZE];
+	u8			opad[SHA1_MD5_BLOCK_SIZE];
+};
+
+struct omap_sham_ctx {
+	struct omap_sham_dev	*dd;
+
+	unsigned long		flags;
+
+	/* fallback stuff */
+	struct crypto_shash	*fallback;
+
+	struct omap_sham_hmac_ctx base[0];
+};
+
+#define OMAP_SHAM_QUEUE_LENGTH	1
+
+struct omap_sham_dev {
+	struct list_head	list;
+	unsigned long		phys_base;
+	struct device		*dev;
+	void __iomem		*io_base;
+	int			irq;
+	struct clk		*iclk;
+	spinlock_t		lock;
+	int			dma;
+	int			dma_lch;
+	struct tasklet_struct	done_task;
+	struct tasklet_struct	queue_task;
+
+	unsigned long		flags;
+	struct crypto_queue	queue;
+	struct ahash_request	*req;
+};
+
+struct omap_sham_drv {
+	struct list_head	dev_list;
+	spinlock_t		lock;
+	unsigned long		flags;
+};
+
+static struct omap_sham_drv sham = {
+	.dev_list = LIST_HEAD_INIT(sham.dev_list),
+	.lock = __SPIN_LOCK_UNLOCKED(sham.lock),
+};
+
+static inline u32 omap_sham_read(struct omap_sham_dev *dd, u32 offset)
+{
+	return __raw_readl(dd->io_base + offset);
+}
+
+static inline void omap_sham_write(struct omap_sham_dev *dd,
+					u32 offset, u32 value)
+{
+	__raw_writel(value, dd->io_base + offset);
+}
+
+static inline void omap_sham_write_mask(struct omap_sham_dev *dd, u32 address,
+					u32 value, u32 mask)
+{
+	u32 val;
+
+	val = omap_sham_read(dd, address);
+	val &= ~mask;
+	val |= value;
+	omap_sham_write(dd, address, val);
+}
+
+static inline int omap_sham_wait(struct omap_sham_dev *dd, u32 offset, u32 bit)
+{
+	unsigned long timeout = jiffies + DEFAULT_TIMEOUT_INTERVAL;
+
+	while (!(omap_sham_read(dd, offset) & bit)) {
+		if (time_is_before_jiffies(timeout))
+			return -ETIMEDOUT;
+	}
+
+	return 0;
+}
+
+static void omap_sham_copy_hash(struct ahash_request *req, int out)
+{
+	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
+	u32 *hash = (u32 *)req->result;
+	int i;
+
+	if (likely(ctx->flags & FLAGS_SHA1)) {
+		/* SHA1 results are in big endian */
+		for (i = 0; i < SHA1_DIGEST_SIZE / sizeof(u32); i++)
+			if (out)
+				hash[i] = be32_to_cpu(omap_sham_read(ctx->dd,
+							SHA_REG_DIGEST(i)));
+			else
+				omap_sham_write(ctx->dd, SHA_REG_DIGEST(i),
+							cpu_to_be32(hash[i]));
+	} else {
+		/* MD5 results are in little endian */
+		for (i = 0; i < MD5_DIGEST_SIZE / sizeof(u32); i++)
+			if (out)
+				hash[i] = le32_to_cpu(omap_sham_read(ctx->dd,
+							SHA_REG_DIGEST(i)));
+			else
+				omap_sham_write(ctx->dd, SHA_REG_DIGEST(i),
+							cpu_to_le32(hash[i]));
+	}
+}
+
+static int omap_sham_write_ctrl(struct omap_sham_dev *dd, size_t length,
+				 int final, int dma)
+{
+	struct omap_sham_reqctx *ctx = ahash_request_ctx(dd->req);
+	u32 val = length << 5, mask;
+
+	if (unlikely(!ctx->digcnt)) {
+
+		clk_enable(dd->iclk);
+
+		if (!(dd->flags & FLAGS_INIT)) {
+			omap_sham_write_mask(dd, SHA_REG_MASK,
+				SHA_REG_MASK_SOFTRESET, SHA_REG_MASK_SOFTRESET);
+
+			if (omap_sham_wait(dd, SHA_REG_SYSSTATUS,
+						SHA_REG_SYSSTATUS_RESETDONE))
+				return -ETIMEDOUT;
+
+			dd->flags |= FLAGS_INIT;
+		}
+	} else {
+		omap_sham_write(dd, SHA_REG_DIGCNT, ctx->digcnt);
+	}
+
+	omap_sham_write_mask(dd, SHA_REG_MASK,
+		SHA_REG_MASK_IT_EN | (dma ? SHA_REG_MASK_DMA_EN : 0),
+		SHA_REG_MASK_IT_EN | SHA_REG_MASK_DMA_EN);
+	/*
+	 * Setting ALGO_CONST only for the first iteration
+	 * and CLOSE_HASH only for the last one.
+	 */
+	if (ctx->flags & FLAGS_SHA1)
+		val |= SHA_REG_CTRL_ALGO;
+	if (!ctx->digcnt)
+		val |= SHA_REG_CTRL_ALGO_CONST;
+	if (final)
+		val |= SHA_REG_CTRL_CLOSE_HASH;
+
+	mask = SHA_REG_CTRL_ALGO_CONST | SHA_REG_CTRL_CLOSE_HASH |
+			SHA_REG_CTRL_ALGO | SHA_REG_CTRL_LENGTH;
+
+	omap_sham_write_mask(dd, SHA_REG_CTRL, val, mask);
+
+	return 0;
+}
+
+static int omap_sham_xmit_cpu(struct omap_sham_dev *dd, const u8 *buf,
+			      size_t length, int final)
+{
+	struct omap_sham_reqctx *ctx = ahash_request_ctx(dd->req);
+	int err, count, len32;
+	const u32 *buffer = (const u32 *)buf;
+
+	dev_dbg(dd->dev, "xmit_cpu: digcnt: %d, length: %d, final: %d\n",
+						ctx->digcnt, length, final);
+
+	err = omap_sham_write_ctrl(dd, length, final, 0);
+	if (err)
+		return err;
+
+	if (omap_sham_wait(dd, SHA_REG_CTRL, SHA_REG_CTRL_INPUT_READY))
+		return -ETIMEDOUT;
+
+	ctx->digcnt += length;
+
+	if (final)
+		ctx->flags |= FLAGS_FINAL; /* catch last interrupt */
+
+	len32 = DIV_ROUND_UP(length, sizeof(u32));
+
+	for (count = 0; count < len32; count++)
+		omap_sham_write(dd, SHA_REG_DIN(count), buffer[count]);
+
+	return -EINPROGRESS;
+}
+
+static int omap_sham_xmit_dma(struct omap_sham_dev *dd, dma_addr_t dma_addr,
+			      size_t length, int final)
+{
+	struct omap_sham_reqctx *ctx = ahash_request_ctx(dd->req);
+	int err, len32;
+
+	dev_dbg(dd->dev, "xmit_dma: digcnt: %d, length: %d, final: %d\n",
+						ctx->digcnt, length, final);
+
+	/* flush cache entries related to our page */
+	if (dma_addr == ctx->dma_addr)
+		dma_sync_single_for_device(dd->dev, dma_addr, length,
+					   DMA_TO_DEVICE);
+
+	len32 = DIV_ROUND_UP(length, sizeof(u32));
+
+	omap_set_dma_transfer_params(dd->dma_lch, OMAP_DMA_DATA_TYPE_S32, len32,
+			1, OMAP_DMA_SYNC_PACKET, dd->dma, OMAP_DMA_DST_SYNC);
+
+	omap_set_dma_src_params(dd->dma_lch, 0, OMAP_DMA_AMODE_POST_INC,
+				dma_addr, 0, 0);
+
+	err = omap_sham_write_ctrl(dd, length, final, 1);
+	if (err)
+		return err;
+
+	ctx->digcnt += length;
+
+	if (final)
+		ctx->flags |= FLAGS_FINAL; /* catch last interrupt */
+
+	dd->flags |= FLAGS_DMA_ACTIVE;
+
+	omap_start_dma(dd->dma_lch);
+
+	return -EINPROGRESS;
+}
+
+static size_t omap_sham_append_buffer(struct omap_sham_reqctx *ctx,
+				const u8 *data, size_t length)
+{
+	size_t count = min(length, ctx->buflen - ctx->bufcnt);
+
+	count = min(count, ctx->total);
+	if (count <= 0)
+		return 0;
+	memcpy(ctx->buffer + ctx->bufcnt, data, count);
+	ctx->bufcnt += count;
+
+	return count;
+}
+
+static size_t omap_sham_append_sg(struct omap_sham_reqctx *ctx)
+{
+	size_t count;
+
+	while (ctx->sg) {
+		count = omap_sham_append_buffer(ctx,
+				sg_virt(ctx->sg) + ctx->offset,
+				ctx->sg->length - ctx->offset);
+		if (!count)
+			break;
+		ctx->offset += count;
+		ctx->total -= count;
+		if (ctx->offset == ctx->sg->length) {
+			ctx->sg = sg_next(ctx->sg);
+			if (ctx->sg)
+				ctx->offset = 0;
+			else
+				ctx->total = 0;
+		}
+	}
+
+	return 0;
+}
+
+static int omap_sham_update_dma_slow(struct omap_sham_dev *dd)
+{
+	struct omap_sham_reqctx *ctx = ahash_request_ctx(dd->req);
+	unsigned int final;
+	size_t count;
+
+	if (!ctx->total)
+		return 0;
+
+	omap_sham_append_sg(ctx);
+
+	final = (ctx->flags & FLAGS_FINUP) && !ctx->total;
+
+	dev_dbg(dd->dev, "slow: bufcnt: %u, digcnt: %d, final: %d\n",
+					 ctx->bufcnt, ctx->digcnt, final);
+
+	if (final || (ctx->bufcnt == ctx->buflen && ctx->total)) {
+		count = ctx->bufcnt;
+		ctx->bufcnt = 0;
+		return omap_sham_xmit_dma(dd, ctx->dma_addr, count, final);
+	}
+
+	return 0;
+}
+
+static int omap_sham_update_dma_fast(struct omap_sham_dev *dd)
+{
+	struct omap_sham_reqctx *ctx = ahash_request_ctx(dd->req);
+	unsigned int length;
+
+	ctx->flags |= FLAGS_FAST;
+
+	length = min(ctx->total, sg_dma_len(ctx->sg));
+	ctx->total = length;
+
+	if (!dma_map_sg(dd->dev, ctx->sg, 1, DMA_TO_DEVICE)) {
+		dev_err(dd->dev, "dma_map_sg  error\n");
+		return -EINVAL;
+	}
+
+	ctx->total -= length;
+
+	return omap_sham_xmit_dma(dd, sg_dma_address(ctx->sg), length, 1);
+}
+
+static int omap_sham_update_cpu(struct omap_sham_dev *dd)
+{
+	struct omap_sham_reqctx *ctx = ahash_request_ctx(dd->req);
+	int bufcnt;
+
+	omap_sham_append_sg(ctx);
+	bufcnt = ctx->bufcnt;
+	ctx->bufcnt = 0;
+
+	return omap_sham_xmit_cpu(dd, ctx->buffer, bufcnt, 1);
+}
+
+static int omap_sham_update_dma_stop(struct omap_sham_dev *dd)
+{
+	struct omap_sham_reqctx *ctx = ahash_request_ctx(dd->req);
+
+	omap_stop_dma(dd->dma_lch);
+	if (ctx->flags & FLAGS_FAST)
+		dma_unmap_sg(dd->dev, ctx->sg, 1, DMA_TO_DEVICE);
+
+	return 0;
+}
+
+static void omap_sham_cleanup(struct ahash_request *req)
+{
+	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
+	struct omap_sham_dev *dd = ctx->dd;
+	unsigned long flags;
+
+	spin_lock_irqsave(&dd->lock, flags);
+	if (ctx->flags & FLAGS_CLEAN) {
+		spin_unlock_irqrestore(&dd->lock, flags);
+		return;
+	}
+	ctx->flags |= FLAGS_CLEAN;
+	spin_unlock_irqrestore(&dd->lock, flags);
+
+	if (ctx->digcnt)
+		clk_disable(dd->iclk);
+
+	if (ctx->dma_addr)
+		dma_unmap_single(dd->dev, ctx->dma_addr, ctx->buflen,
+				 DMA_TO_DEVICE);
+
+	if (ctx->buffer)
+		free_page((unsigned long)ctx->buffer);
+
+	dev_dbg(dd->dev, "digcnt: %d, bufcnt: %d\n", ctx->digcnt, ctx->bufcnt);
+}
+
+static int omap_sham_init(struct ahash_request *req)
+{
+	struct crypto_ahash *tfm = crypto_ahash_reqtfm(req);
+	struct omap_sham_ctx *tctx = crypto_ahash_ctx(tfm);
+	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
+	struct omap_sham_dev *dd = NULL, *tmp;
+
+	spin_lock_bh(&sham.lock);
+	if (!tctx->dd) {
+		list_for_each_entry(tmp, &sham.dev_list, list) {
+			dd = tmp;
+			break;
+		}
+		tctx->dd = dd;
+	} else {
+		dd = tctx->dd;
+	}
+	spin_unlock_bh(&sham.lock);
+
+	ctx->dd = dd;
+
+	ctx->flags = 0;
+
+	ctx->flags |= FLAGS_FIRST;
+
+	dev_dbg(dd->dev, "init: digest size: %d\n",
+		crypto_ahash_digestsize(tfm));
+
+	if (crypto_ahash_digestsize(tfm) == SHA1_DIGEST_SIZE)
+		ctx->flags |= FLAGS_SHA1;
+
+	ctx->bufcnt = 0;
+	ctx->digcnt = 0;
+
+	ctx->buflen = PAGE_SIZE;
+	ctx->buffer = (void *)__get_free_page(
+				(req->base.flags & CRYPTO_TFM_REQ_MAY_SLEEP) ?
+				GFP_KERNEL : GFP_ATOMIC);
+	if (!ctx->buffer)
+		return -ENOMEM;
+
+	ctx->dma_addr = dma_map_single(dd->dev, ctx->buffer, ctx->buflen,
+					DMA_TO_DEVICE);
+	if (dma_mapping_error(dd->dev, ctx->dma_addr)) {
+		dev_err(dd->dev, "dma %u bytes error\n", ctx->buflen);
+		free_page((unsigned long)ctx->buffer);
+		return -EINVAL;
+	}
+
+	if (tctx->flags & FLAGS_HMAC) {
+		struct omap_sham_hmac_ctx *bctx = tctx->base;
+
+		memcpy(ctx->buffer, bctx->ipad, SHA1_MD5_BLOCK_SIZE);
+		ctx->bufcnt = SHA1_MD5_BLOCK_SIZE;
+		ctx->flags |= FLAGS_HMAC;
+	}
+
+	return 0;
+
+}
+
+static int omap_sham_update_req(struct omap_sham_dev *dd)
+{
+	struct ahash_request *req = dd->req;
+	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
+	int err;
+
+	dev_dbg(dd->dev, "update_req: total: %u, digcnt: %d, finup: %d\n",
+		 ctx->total, ctx->digcnt, (ctx->flags & FLAGS_FINUP) != 0);
+
+	if (ctx->flags & FLAGS_CPU)
+		err = omap_sham_update_cpu(dd);
+	else if (ctx->flags & FLAGS_FAST)
+		err = omap_sham_update_dma_fast(dd);
+	else
+		err = omap_sham_update_dma_slow(dd);
+
+	/* wait for dma completion before can take more data */
+	dev_dbg(dd->dev, "update: err: %d, digcnt: %d\n", err, ctx->digcnt);
+
+	return err;
+}
+
+static int omap_sham_final_req(struct omap_sham_dev *dd)
+{
+	struct ahash_request *req = dd->req;
+	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
+	int err = 0, use_dma = 1;
+
+	if (ctx->bufcnt <= 64)
+		/* faster to handle last block with cpu */
+		use_dma = 0;
+
+	if (use_dma)
+		err = omap_sham_xmit_dma(dd, ctx->dma_addr, ctx->bufcnt, 1);
+	else
+		err = omap_sham_xmit_cpu(dd, ctx->buffer, ctx->bufcnt, 1);
+
+	ctx->bufcnt = 0;
+
+	if (err != -EINPROGRESS)
+		omap_sham_cleanup(req);
+
+	dev_dbg(dd->dev, "final_req: err: %d\n", err);
+
+	return err;
+}
+
+static int omap_sham_finish_req_hmac(struct ahash_request *req)
+{
+	struct omap_sham_ctx *tctx = crypto_tfm_ctx(req->base.tfm);
+	struct omap_sham_hmac_ctx *bctx = tctx->base;
+	int bs = crypto_shash_blocksize(bctx->shash);
+	int ds = crypto_shash_digestsize(bctx->shash);
+	struct {
+		struct shash_desc shash;
+		char ctx[crypto_shash_descsize(bctx->shash)];
+	} desc;
+
+	desc.shash.tfm = bctx->shash;
+	desc.shash.flags = 0; /* not CRYPTO_TFM_REQ_MAY_SLEEP */
+
+	return crypto_shash_init(&desc.shash) ?:
+	       crypto_shash_update(&desc.shash, bctx->opad, bs) ?:
+	       crypto_shash_finup(&desc.shash, req->result, ds, req->result);
+}
+
+static void omap_sham_finish_req(struct ahash_request *req, int err)
+{
+	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
+
+	if (!err) {
+		omap_sham_copy_hash(ctx->dd->req, 1);
+		if (ctx->flags & FLAGS_HMAC)
+			err = omap_sham_finish_req_hmac(req);
+	}
+
+	if (ctx->flags & FLAGS_FINAL)
+		omap_sham_cleanup(req);
+
+	clear_bit(FLAGS_BUSY, &ctx->dd->flags);
+
+	if (req->base.complete)
+		req->base.complete(&req->base, err);
+}
+
+static int omap_sham_handle_queue(struct omap_sham_dev *dd)
+{
+	struct crypto_async_request *async_req, *backlog;
+	struct omap_sham_reqctx *ctx;
+	struct ahash_request *req, *prev_req;
+	unsigned long flags;
+	int err = 0;
+
+	if (test_and_set_bit(FLAGS_BUSY, &dd->flags))
+		return 0;
+
+	spin_lock_irqsave(&dd->lock, flags);
+	backlog = crypto_get_backlog(&dd->queue);
+	async_req = crypto_dequeue_request(&dd->queue);
+	if (!async_req)
+		clear_bit(FLAGS_BUSY, &dd->flags);
+	spin_unlock_irqrestore(&dd->lock, flags);
+
+	if (!async_req)
+		return 0;
+
+	if (backlog)
+		backlog->complete(backlog, -EINPROGRESS);
+
+	req = ahash_request_cast(async_req);
+
+	prev_req = dd->req;
+	dd->req = req;
+
+	ctx = ahash_request_ctx(req);
+
+	dev_dbg(dd->dev, "handling new req, op: %lu, nbytes: %d\n",
+						ctx->op, req->nbytes);
+
+	if (req != prev_req && ctx->digcnt)
+		/* request has changed - restore hash */
+		omap_sham_copy_hash(req, 0);
+
+	if (ctx->op == OP_UPDATE) {
+		err = omap_sham_update_req(dd);
+		if (err != -EINPROGRESS && (ctx->flags & FLAGS_FINUP))
+			/* no final() after finup() */
+			err = omap_sham_final_req(dd);
+	} else if (ctx->op == OP_FINAL) {
+		err = omap_sham_final_req(dd);
+	}
+
+	if (err != -EINPROGRESS) {
+		/* done_task will not finish it, so do it here */
+		omap_sham_finish_req(req, err);
+		tasklet_schedule(&dd->queue_task);
+	}
+
+	dev_dbg(dd->dev, "exit, err: %d\n", err);
+
+	return err;
+}
+
+static int omap_sham_enqueue(struct ahash_request *req, unsigned int op)
+{
+	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
+	struct omap_sham_ctx *tctx = crypto_tfm_ctx(req->base.tfm);
+	struct omap_sham_dev *dd = tctx->dd;
+	unsigned long flags;
+	int err;
+
+	ctx->op = op;
+
+	spin_lock_irqsave(&dd->lock, flags);
+	err = ahash_enqueue_request(&dd->queue, req);
+	spin_unlock_irqrestore(&dd->lock, flags);
+
+	omap_sham_handle_queue(dd);
+
+	return err;
+}
+
+static int omap_sham_update(struct ahash_request *req)
+{
+	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
+
+	if (!req->nbytes)
+		return 0;
+
+	ctx->total = req->nbytes;
+	ctx->sg = req->src;
+	ctx->offset = 0;
+
+	if (ctx->flags & FLAGS_FINUP) {
+		if ((ctx->digcnt + ctx->bufcnt + ctx->total) < 9) {
+			/*
+			* OMAP HW accel works only with buffers >= 9
+			* will switch to bypass in final()
+			* final has the same request and data
+			*/
+			omap_sham_append_sg(ctx);
+			return 0;
+		} else if (ctx->bufcnt + ctx->total <= 64) {
+			ctx->flags |= FLAGS_CPU;
+		} else if (!ctx->bufcnt && sg_is_last(ctx->sg)) {
+			/* may be can use faster functions */
+			int aligned = IS_ALIGNED((u32)ctx->sg->offset,
+								sizeof(u32));
+
+			if (aligned && (ctx->flags & FLAGS_FIRST))
+				/* digest: first and final */
+				ctx->flags |= FLAGS_FAST;
+
+			ctx->flags &= ~FLAGS_FIRST;
+		}
+	} else if (ctx->bufcnt + ctx->total <= ctx->buflen) {
+		/* if not finaup -> not fast */
+		omap_sham_append_sg(ctx);
+		return 0;
+	}
+
+	return omap_sham_enqueue(req, OP_UPDATE);
+}
+
+static int omap_sham_shash_digest(struct crypto_shash *shash, u32 flags,
+				  const u8 *data, unsigned int len, u8 *out)
+{
+	struct {
+		struct shash_desc shash;
+		char ctx[crypto_shash_descsize(shash)];
+	} desc;
+
+	desc.shash.tfm = shash;
+	desc.shash.flags = flags & CRYPTO_TFM_REQ_MAY_SLEEP;
+
+	return crypto_shash_digest(&desc.shash, data, len, out);
+}
+
+static int omap_sham_final_shash(struct ahash_request *req)
+{
+	struct omap_sham_ctx *tctx = crypto_tfm_ctx(req->base.tfm);
+	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
+
+	return omap_sham_shash_digest(tctx->fallback, req->base.flags,
+				      ctx->buffer, ctx->bufcnt, req->result);
+}
+
+static int omap_sham_final(struct ahash_request *req)
+{
+	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
+	int err = 0;
+
+	ctx->flags |= FLAGS_FINUP;
+
+	/* OMAP HW accel works only with buffers >= 9 */
+	/* HMAC is always >= 9 because of ipad */
+	if ((ctx->digcnt + ctx->bufcnt) < 9)
+		err = omap_sham_final_shash(req);
+	else if (ctx->bufcnt)
+		return omap_sham_enqueue(req, OP_FINAL);
+
+	omap_sham_cleanup(req);
+
+	return err;
+}
+
+static int omap_sham_finup(struct ahash_request *req)
+{
+	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
+	int err1, err2;
+
+	ctx->flags |= FLAGS_FINUP;
+
+	err1 = omap_sham_update(req);
+	if (err1 == -EINPROGRESS)
+		return err1;
+	/*
+	 * final() has to be always called to cleanup resources
+	 * even if udpate() failed, except EINPROGRESS
+	 */
+	err2 = omap_sham_final(req);
+
+	return err1 ?: err2;
+}
+
+static int omap_sham_digest(struct ahash_request *req)
+{
+	return omap_sham_init(req) ?: omap_sham_finup(req);
+}
+
+static int omap_sham_setkey(struct crypto_ahash *tfm, const u8 *key,
+		      unsigned int keylen)
+{
+	struct omap_sham_ctx *tctx = crypto_ahash_ctx(tfm);
+	struct omap_sham_hmac_ctx *bctx = tctx->base;
+	int bs = crypto_shash_blocksize(bctx->shash);
+	int ds = crypto_shash_digestsize(bctx->shash);
+	int err, i;
+	err = crypto_shash_setkey(tctx->fallback, key, keylen);
+	if (err)
+		return err;
+
+	if (keylen > bs) {
+		err = omap_sham_shash_digest(bctx->shash,
+				crypto_shash_get_flags(bctx->shash),
+				key, keylen, bctx->ipad);
+		if (err)
+			return err;
+		keylen = ds;
+	} else {
+		memcpy(bctx->ipad, key, keylen);
+	}
+
+	memset(bctx->ipad + keylen, 0, bs - keylen);
+	memcpy(bctx->opad, bctx->ipad, bs);
+
+	for (i = 0; i < bs; i++) {
+		bctx->ipad[i] ^= 0x36;
+		bctx->opad[i] ^= 0x5c;
+	}
+
+	return err;
+}
+
+static int omap_sham_cra_init_alg(struct crypto_tfm *tfm, const char *alg_base)
+{
+	struct omap_sham_ctx *tctx = crypto_tfm_ctx(tfm);
+	const char *alg_name = crypto_tfm_alg_name(tfm);
+
+	/* Allocate a fallback and abort if it failed. */
+	tctx->fallback = crypto_alloc_shash(alg_name, 0,
+					    CRYPTO_ALG_NEED_FALLBACK);
+	if (IS_ERR(tctx->fallback)) {
+		pr_err("omap-sham: fallback driver '%s' "
+				"could not be loaded.\n", alg_name);
+		return PTR_ERR(tctx->fallback);
+	}
+
+	crypto_ahash_set_reqsize(__crypto_ahash_cast(tfm),
+				 sizeof(struct omap_sham_reqctx));
+
+	if (alg_base) {
+		struct omap_sham_hmac_ctx *bctx = tctx->base;
+		tctx->flags |= FLAGS_HMAC;
+		bctx->shash = crypto_alloc_shash(alg_base, 0,
+						CRYPTO_ALG_NEED_FALLBACK);
+		if (IS_ERR(bctx->shash)) {
+			pr_err("omap-sham: base driver '%s' "
+					"could not be loaded.\n", alg_base);
+			crypto_free_shash(tctx->fallback);
+			return PTR_ERR(bctx->shash);
+		}
+
+	}
+
+	return 0;
+}
+
+static int omap_sham_cra_init(struct crypto_tfm *tfm)
+{
+	return omap_sham_cra_init_alg(tfm, NULL);
+}
+
+static int omap_sham_cra_sha1_init(struct crypto_tfm *tfm)
+{
+	return omap_sham_cra_init_alg(tfm, "sha1");
+}
+
+static int omap_sham_cra_md5_init(struct crypto_tfm *tfm)
+{
+	return omap_sham_cra_init_alg(tfm, "md5");
+}
+
+static void omap_sham_cra_exit(struct crypto_tfm *tfm)
+{
+	struct omap_sham_ctx *tctx = crypto_tfm_ctx(tfm);
+
+	crypto_free_shash(tctx->fallback);
+	tctx->fallback = NULL;
+
+	if (tctx->flags & FLAGS_HMAC) {
+		struct omap_sham_hmac_ctx *bctx = tctx->base;
+		crypto_free_shash(bctx->shash);
+	}
+}
+
+static struct ahash_alg algs[] = {
+{
+	.init		= omap_sham_init,
+	.update		= omap_sham_update,
+	.final		= omap_sham_final,
+	.finup		= omap_sham_finup,
+	.digest		= omap_sham_digest,
+	.halg.digestsize	= SHA1_DIGEST_SIZE,
+	.halg.base	= {
+		.cra_name		= "sha1",
+		.cra_driver_name	= "omap-sha1",
+		.cra_priority		= 100,
+		.cra_flags		= CRYPTO_ALG_TYPE_AHASH |
+						CRYPTO_ALG_ASYNC |
+						CRYPTO_ALG_NEED_FALLBACK,
+		.cra_blocksize		= SHA1_BLOCK_SIZE,
+		.cra_ctxsize		= sizeof(struct omap_sham_ctx),
+		.cra_alignmask		= 0,
+		.cra_module		= THIS_MODULE,
+		.cra_init		= omap_sham_cra_init,
+		.cra_exit		= omap_sham_cra_exit,
+	}
+},
+{
+	.init		= omap_sham_init,
+	.update		= omap_sham_update,
+	.final		= omap_sham_final,
+	.finup		= omap_sham_finup,
+	.digest		= omap_sham_digest,
+	.halg.digestsize	= MD5_DIGEST_SIZE,
+	.halg.base	= {
+		.cra_name		= "md5",
+		.cra_driver_name	= "omap-md5",
+		.cra_priority		= 100,
+		.cra_flags		= CRYPTO_ALG_TYPE_AHASH |
+						CRYPTO_ALG_ASYNC |
+						CRYPTO_ALG_NEED_FALLBACK,
+		.cra_blocksize		= SHA1_BLOCK_SIZE,
+		.cra_ctxsize		= sizeof(struct omap_sham_ctx),
+		.cra_alignmask		= 0,
+		.cra_module		= THIS_MODULE,
+		.cra_init		= omap_sham_cra_init,
+		.cra_exit		= omap_sham_cra_exit,
+	}
+},
+{
+	.init		= omap_sham_init,
+	.update		= omap_sham_update,
+	.final		= omap_sham_final,
+	.finup		= omap_sham_finup,
+	.digest		= omap_sham_digest,
+	.setkey		= omap_sham_setkey,
+	.halg.digestsize	= SHA1_DIGEST_SIZE,
+	.halg.base	= {
+		.cra_name		= "hmac(sha1)",
+		.cra_driver_name	= "omap-hmac-sha1",
+		.cra_priority		= 100,
+		.cra_flags		= CRYPTO_ALG_TYPE_AHASH |
+						CRYPTO_ALG_ASYNC |
+						CRYPTO_ALG_NEED_FALLBACK,
+		.cra_blocksize		= SHA1_BLOCK_SIZE,
+		.cra_ctxsize		= sizeof(struct omap_sham_ctx) +
+					sizeof(struct omap_sham_hmac_ctx),
+		.cra_alignmask		= 0,
+		.cra_module		= THIS_MODULE,
+		.cra_init		= omap_sham_cra_sha1_init,
+		.cra_exit		= omap_sham_cra_exit,
+	}
+},
+{
+	.init		= omap_sham_init,
+	.update		= omap_sham_update,
+	.final		= omap_sham_final,
+	.finup		= omap_sham_finup,
+	.digest		= omap_sham_digest,
+	.setkey		= omap_sham_setkey,
+	.halg.digestsize	= MD5_DIGEST_SIZE,
+	.halg.base	= {
+		.cra_name		= "hmac(md5)",
+		.cra_driver_name	= "omap-hmac-md5",
+		.cra_priority		= 100,
+		.cra_flags		= CRYPTO_ALG_TYPE_AHASH |
+						CRYPTO_ALG_ASYNC |
+						CRYPTO_ALG_NEED_FALLBACK,
+		.cra_blocksize		= SHA1_BLOCK_SIZE,
+		.cra_ctxsize		= sizeof(struct omap_sham_ctx) +
+					sizeof(struct omap_sham_hmac_ctx),
+		.cra_alignmask		= 0,
+		.cra_module		= THIS_MODULE,
+		.cra_init		= omap_sham_cra_md5_init,
+		.cra_exit		= omap_sham_cra_exit,
+	}
+}
+};
+
+static void omap_sham_done_task(unsigned long data)
+{
+	struct omap_sham_dev *dd = (struct omap_sham_dev *)data;
+	struct ahash_request *req = dd->req;
+	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
+	int ready = 1;
+
+	if (ctx->flags & FLAGS_OUTPUT_READY) {
+		ctx->flags &= ~FLAGS_OUTPUT_READY;
+		ready = 1;
+	}
+
+	if (dd->flags & FLAGS_DMA_ACTIVE) {
+		dd->flags &= ~FLAGS_DMA_ACTIVE;
+		omap_sham_update_dma_stop(dd);
+		omap_sham_update_dma_slow(dd);
+	}
+
+	if (ready && !(dd->flags & FLAGS_DMA_ACTIVE)) {
+		dev_dbg(dd->dev, "update done\n");
+		/* finish curent request */
+		omap_sham_finish_req(req, 0);
+		/* start new request */
+		omap_sham_handle_queue(dd);
+	}
+}
+
+static void omap_sham_queue_task(unsigned long data)
+{
+	struct omap_sham_dev *dd = (struct omap_sham_dev *)data;
+
+	omap_sham_handle_queue(dd);
+}
+
+static irqreturn_t omap_sham_irq(int irq, void *dev_id)
+{
+	struct omap_sham_dev *dd = dev_id;
+	struct omap_sham_reqctx *ctx = ahash_request_ctx(dd->req);
+
+	if (!ctx) {
+		dev_err(dd->dev, "unknown interrupt.\n");
+		return IRQ_HANDLED;
+	}
+
+	if (unlikely(ctx->flags & FLAGS_FINAL))
+		/* final -> allow device to go to power-saving mode */
+		omap_sham_write_mask(dd, SHA_REG_CTRL, 0, SHA_REG_CTRL_LENGTH);
+
+	omap_sham_write_mask(dd, SHA_REG_CTRL, SHA_REG_CTRL_OUTPUT_READY,
+				 SHA_REG_CTRL_OUTPUT_READY);
+	omap_sham_read(dd, SHA_REG_CTRL);
+
+	ctx->flags |= FLAGS_OUTPUT_READY;
+	tasklet_schedule(&dd->done_task);
+
+	return IRQ_HANDLED;
+}
+
+static void omap_sham_dma_callback(int lch, u16 ch_status, void *data)
+{
+	struct omap_sham_dev *dd = data;
+
+	if (likely(lch == dd->dma_lch))
+		tasklet_schedule(&dd->done_task);
+}
+
+static int omap_sham_dma_init(struct omap_sham_dev *dd)
+{
+	int err;
+
+	dd->dma_lch = -1;
+
+	err = omap_request_dma(dd->dma, dev_name(dd->dev),
+			omap_sham_dma_callback, dd, &dd->dma_lch);
+	if (err) {
+		dev_err(dd->dev, "Unable to request DMA channel\n");
+		return err;
+	}
+	omap_set_dma_dest_params(dd->dma_lch, 0,
+			OMAP_DMA_AMODE_CONSTANT,
+			dd->phys_base + SHA_REG_DIN(0), 0, 16);
+
+	omap_set_dma_dest_burst_mode(dd->dma_lch,
+			OMAP_DMA_DATA_BURST_16);
+
+	return 0;
+}
+
+static void omap_sham_dma_cleanup(struct omap_sham_dev *dd)
+{
+	if (dd->dma_lch >= 0) {
+		omap_free_dma(dd->dma_lch);
+		dd->dma_lch = -1;
+	}
+}
+
+static int __devinit omap_sham_probe(struct platform_device *pdev)
+{
+	struct omap_sham_dev *dd;
+	struct device *dev = &pdev->dev;
+	struct resource *res;
+	int err, i, j;
+
+	dd = kzalloc(sizeof(struct omap_sham_dev), GFP_KERNEL);
+	if (dd == NULL) {
+		dev_err(dev, "unable to alloc data struct.\n");
+		err = -ENOMEM;
+		goto data_err;
+	}
+	dd->dev = dev;
+	platform_set_drvdata(pdev, dd);
+
+	INIT_LIST_HEAD(&dd->list);
+	spin_lock_init(&dd->lock);
+	tasklet_init(&dd->done_task, omap_sham_done_task, (unsigned long)dd);
+	tasklet_init(&dd->queue_task, omap_sham_queue_task, (unsigned long)dd);
+	crypto_init_queue(&dd->queue, OMAP_SHAM_QUEUE_LENGTH);
+
+	dd->irq = -1;
+
+	/* Get the base address */
+	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	if (!res) {
+		dev_err(dev, "no MEM resource info\n");
+		err = -ENODEV;
+		goto res_err;
+	}
+	dd->phys_base = res->start;
+
+	/* Get the DMA */
+	res = platform_get_resource(pdev, IORESOURCE_DMA, 0);
+	if (!res) {
+		dev_err(dev, "no DMA resource info\n");
+		err = -ENODEV;
+		goto res_err;
+	}
+	dd->dma = res->start;
+
+	/* Get the IRQ */
+	dd->irq = platform_get_irq(pdev,  0);
+	if (dd->irq < 0) {
+		dev_err(dev, "no IRQ resource info\n");
+		err = dd->irq;
+		goto res_err;
+	}
+
+	err = request_irq(dd->irq, omap_sham_irq,
+			IRQF_TRIGGER_LOW, dev_name(dev), dd);
+	if (err) {
+		dev_err(dev, "unable to request irq.\n");
+		goto res_err;
+	}
+
+	err = omap_sham_dma_init(dd);
+	if (err)
+		goto dma_err;
+
+	/* Initializing the clock */
+	dd->iclk = clk_get(dev, "ick");
+	if (!dd->iclk) {
+		dev_err(dev, "clock intialization failed.\n");
+		err = -ENODEV;
+		goto clk_err;
+	}
+
+	dd->io_base = ioremap(dd->phys_base, SZ_4K);
+	if (!dd->io_base) {
+		dev_err(dev, "can't ioremap\n");
+		err = -ENOMEM;
+		goto io_err;
+	}
+
+	clk_enable(dd->iclk);
+	dev_info(dev, "hw accel on OMAP rev %u.%u\n",
+		(omap_sham_read(dd, SHA_REG_REV) & SHA_REG_REV_MAJOR) >> 4,
+		omap_sham_read(dd, SHA_REG_REV) & SHA_REG_REV_MINOR);
+	clk_disable(dd->iclk);
+
+	spin_lock(&sham.lock);
+	list_add_tail(&dd->list, &sham.dev_list);
+	spin_unlock(&sham.lock);
+
+	for (i = 0; i < ARRAY_SIZE(algs); i++) {
+		err = crypto_register_ahash(&algs[i]);
+		if (err)
+			goto err_algs;
+	}
+
+	return 0;
+
+err_algs:
+	for (j = 0; j < i; j++)
+		crypto_unregister_ahash(&algs[j]);
+	iounmap(dd->io_base);
+io_err:
+	clk_put(dd->iclk);
+clk_err:
+	omap_sham_dma_cleanup(dd);
+dma_err:
+	if (dd->irq >= 0)
+		free_irq(dd->irq, dd);
+res_err:
+	kfree(dd);
+	dd = NULL;
+data_err:
+	dev_err(dev, "initialization failed.\n");
+
+	return err;
+}
+
+static int __devexit omap_sham_remove(struct platform_device *pdev)
+{
+	static struct omap_sham_dev *dd;
+	int i;
+
+	dd = platform_get_drvdata(pdev);
+	if (!dd)
+		return -ENODEV;
+	spin_lock(&sham.lock);
+	list_del(&dd->list);
+	spin_unlock(&sham.lock);
+	for (i = 0; i < ARRAY_SIZE(algs); i++)
+		crypto_unregister_ahash(&algs[i]);
+	tasklet_kill(&dd->done_task);
+	tasklet_kill(&dd->queue_task);
+	iounmap(dd->io_base);
+	clk_put(dd->iclk);
+	omap_sham_dma_cleanup(dd);
+	if (dd->irq >= 0)
+		free_irq(dd->irq, dd);
+	kfree(dd);
+	dd = NULL;
+
+	return 0;
+}
+
+static struct platform_driver omap_sham_driver = {
+	.probe	= omap_sham_probe,
+	.remove	= omap_sham_remove,
+	.driver	= {
+		.name	= "omap-sham",
+		.owner	= THIS_MODULE,
+	},
+};
+
+static int __init omap_sham_mod_init(void)
+{
+	pr_info("loading %s driver\n", "omap-sham");
+
+	if (!cpu_class_is_omap2() ||
+		omap_type() != OMAP2_DEVICE_TYPE_SEC) {
+		pr_err("Unsupported cpu\n");
+		return -ENODEV;
+	}
+
+	return platform_driver_register(&omap_sham_driver);
+}
+
+static void __exit omap_sham_mod_exit(void)
+{
+	platform_driver_unregister(&omap_sham_driver);
+}
+
+module_init(omap_sham_mod_init);
+module_exit(omap_sham_mod_exit);
+
+MODULE_DESCRIPTION("OMAP SHA1/MD5 hw acceleration support.");
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Dmitry Kasatkin");
diff --git a/drivers/crypto/talitos.c b/drivers/crypto/talitos.c
index dc558a097311..637c105f53d2 100644
--- a/drivers/crypto/talitos.c
+++ b/drivers/crypto/talitos.c
@@ -1,7 +1,7 @@
 /*
  * talitos - Freescale Integrated Security Engine (SEC) device driver
  *
- * Copyright (c) 2008 Freescale Semiconductor, Inc.
+ * Copyright (c) 2008-2010 Freescale Semiconductor, Inc.
  *
  * Scatterlist Crypto API glue code copied from files with the following:
  * Copyright (c) 2006-2007 Herbert Xu <herbert@gondor.apana.org.au>
@@ -43,9 +43,12 @@
 #include <crypto/aes.h>
 #include <crypto/des.h>
 #include <crypto/sha.h>
+#include <crypto/md5.h>
 #include <crypto/aead.h>
 #include <crypto/authenc.h>
 #include <crypto/skcipher.h>
+#include <crypto/hash.h>
+#include <crypto/internal/hash.h>
 #include <crypto/scatterwalk.h>
 
 #include "talitos.h"
@@ -65,6 +68,13 @@ struct talitos_ptr {
 	__be32 ptr;	/* address */
 };
 
+static const struct talitos_ptr zero_entry = {
+	.len = 0,
+	.j_extent = 0,
+	.eptr = 0,
+	.ptr = 0
+};
+
 /* descriptor */
 struct talitos_desc {
 	__be32 hdr;			/* header high bits */
@@ -146,6 +156,7 @@ struct talitos_private {
 /* .features flag */
 #define TALITOS_FTR_SRC_LINK_TBL_LEN_INCLUDES_EXTENT 0x00000001
 #define TALITOS_FTR_HW_AUTH_CHECK 0x00000002
+#define TALITOS_FTR_SHA224_HWINIT 0x00000004
 
 static void to_talitos_ptr(struct talitos_ptr *talitos_ptr, dma_addr_t dma_addr)
 {
@@ -692,7 +703,7 @@ static void talitos_unregister_rng(struct device *dev)
 #define TALITOS_MAX_KEY_SIZE		64
 #define TALITOS_MAX_IV_LENGTH		16 /* max of AES_BLOCK_SIZE, DES3_EDE_BLOCK_SIZE */
 
-#define MD5_DIGEST_SIZE   16
+#define MD5_BLOCK_SIZE    64
 
 struct talitos_ctx {
 	struct device *dev;
@@ -705,6 +716,23 @@ struct talitos_ctx {
 	unsigned int authsize;
 };
 
+#define HASH_MAX_BLOCK_SIZE		SHA512_BLOCK_SIZE
+#define TALITOS_MDEU_MAX_CONTEXT_SIZE	TALITOS_MDEU_CONTEXT_SIZE_SHA384_SHA512
+
+struct talitos_ahash_req_ctx {
+	u64 count;
+	u32 hw_context[TALITOS_MDEU_MAX_CONTEXT_SIZE / sizeof(u32)];
+	unsigned int hw_context_size;
+	u8 buf[HASH_MAX_BLOCK_SIZE];
+	u8 bufnext[HASH_MAX_BLOCK_SIZE];
+	unsigned int swinit;
+	unsigned int first;
+	unsigned int last;
+	unsigned int to_hash_later;
+	struct scatterlist bufsl[2];
+	struct scatterlist *psrc;
+};
+
 static int aead_setauthsize(struct crypto_aead *authenc,
 			    unsigned int authsize)
 {
@@ -821,10 +849,14 @@ static void talitos_sg_unmap(struct device *dev,
 		else
 			dma_unmap_sg(dev, src, src_nents, DMA_TO_DEVICE);
 
-		if (edesc->dst_is_chained)
-			talitos_unmap_sg_chain(dev, dst, DMA_FROM_DEVICE);
-		else
-			dma_unmap_sg(dev, dst, dst_nents, DMA_FROM_DEVICE);
+		if (dst) {
+			if (edesc->dst_is_chained)
+				talitos_unmap_sg_chain(dev, dst,
+						       DMA_FROM_DEVICE);
+			else
+				dma_unmap_sg(dev, dst, dst_nents,
+					     DMA_FROM_DEVICE);
+		}
 	} else
 		if (edesc->src_is_chained)
 			talitos_unmap_sg_chain(dev, src, DMA_BIDIRECTIONAL);
@@ -1114,12 +1146,67 @@ static int sg_count(struct scatterlist *sg_list, int nbytes, int *chained)
 	return sg_nents;
 }
 
+/**
+ * sg_copy_end_to_buffer - Copy end data from SG list to a linear buffer
+ * @sgl:		 The SG list
+ * @nents:		 Number of SG entries
+ * @buf:		 Where to copy to
+ * @buflen:		 The number of bytes to copy
+ * @skip:		 The number of bytes to skip before copying.
+ *                       Note: skip + buflen should equal SG total size.
+ *
+ * Returns the number of copied bytes.
+ *
+ **/
+static size_t sg_copy_end_to_buffer(struct scatterlist *sgl, unsigned int nents,
+				    void *buf, size_t buflen, unsigned int skip)
+{
+	unsigned int offset = 0;
+	unsigned int boffset = 0;
+	struct sg_mapping_iter miter;
+	unsigned long flags;
+	unsigned int sg_flags = SG_MITER_ATOMIC;
+	size_t total_buffer = buflen + skip;
+
+	sg_flags |= SG_MITER_FROM_SG;
+
+	sg_miter_start(&miter, sgl, nents, sg_flags);
+
+	local_irq_save(flags);
+
+	while (sg_miter_next(&miter) && offset < total_buffer) {
+		unsigned int len;
+		unsigned int ignore;
+
+		if ((offset + miter.length) > skip) {
+			if (offset < skip) {
+				/* Copy part of this segment */
+				ignore = skip - offset;
+				len = miter.length - ignore;
+				memcpy(buf + boffset, miter.addr + ignore, len);
+			} else {
+				/* Copy all of this segment */
+				len = miter.length;
+				memcpy(buf + boffset, miter.addr, len);
+			}
+			boffset += len;
+		}
+		offset += miter.length;
+	}
+
+	sg_miter_stop(&miter);
+
+	local_irq_restore(flags);
+	return boffset;
+}
+
 /*
  * allocate and map the extended descriptor
  */
 static struct talitos_edesc *talitos_edesc_alloc(struct device *dev,
 						 struct scatterlist *src,
 						 struct scatterlist *dst,
+						 int hash_result,
 						 unsigned int cryptlen,
 						 unsigned int authsize,
 						 int icv_stashing,
@@ -1139,11 +1226,16 @@ static struct talitos_edesc *talitos_edesc_alloc(struct device *dev,
 	src_nents = sg_count(src, cryptlen + authsize, &src_chained);
 	src_nents = (src_nents == 1) ? 0 : src_nents;
 
-	if (dst == src) {
-		dst_nents = src_nents;
+	if (hash_result) {
+		dst_nents = 0;
 	} else {
-		dst_nents = sg_count(dst, cryptlen + authsize, &dst_chained);
-		dst_nents = (dst_nents == 1) ? 0 : dst_nents;
+		if (dst == src) {
+			dst_nents = src_nents;
+		} else {
+			dst_nents = sg_count(dst, cryptlen + authsize,
+					     &dst_chained);
+			dst_nents = (dst_nents == 1) ? 0 : dst_nents;
+		}
 	}
 
 	/*
@@ -1172,8 +1264,10 @@ static struct talitos_edesc *talitos_edesc_alloc(struct device *dev,
 	edesc->src_is_chained = src_chained;
 	edesc->dst_is_chained = dst_chained;
 	edesc->dma_len = dma_len;
-	edesc->dma_link_tbl = dma_map_single(dev, &edesc->link_tbl[0],
-					     edesc->dma_len, DMA_BIDIRECTIONAL);
+	if (dma_len)
+		edesc->dma_link_tbl = dma_map_single(dev, &edesc->link_tbl[0],
+						     edesc->dma_len,
+						     DMA_BIDIRECTIONAL);
 
 	return edesc;
 }
@@ -1184,7 +1278,7 @@ static struct talitos_edesc *aead_edesc_alloc(struct aead_request *areq,
 	struct crypto_aead *authenc = crypto_aead_reqtfm(areq);
 	struct talitos_ctx *ctx = crypto_aead_ctx(authenc);
 
-	return talitos_edesc_alloc(ctx->dev, areq->src, areq->dst,
+	return talitos_edesc_alloc(ctx->dev, areq->src, areq->dst, 0,
 				   areq->cryptlen, ctx->authsize, icv_stashing,
 				   areq->base.flags);
 }
@@ -1441,8 +1535,8 @@ static struct talitos_edesc *ablkcipher_edesc_alloc(struct ablkcipher_request *
 	struct crypto_ablkcipher *cipher = crypto_ablkcipher_reqtfm(areq);
 	struct talitos_ctx *ctx = crypto_ablkcipher_ctx(cipher);
 
-	return talitos_edesc_alloc(ctx->dev, areq->src, areq->dst, areq->nbytes,
-				   0, 0, areq->base.flags);
+	return talitos_edesc_alloc(ctx->dev, areq->src, areq->dst, 0,
+				   areq->nbytes, 0, 0, areq->base.flags);
 }
 
 static int ablkcipher_encrypt(struct ablkcipher_request *areq)
@@ -1478,15 +1572,329 @@ static int ablkcipher_decrypt(struct ablkcipher_request *areq)
 	return common_nonsnoop(edesc, areq, NULL, ablkcipher_done);
 }
 
+static void common_nonsnoop_hash_unmap(struct device *dev,
+				       struct talitos_edesc *edesc,
+				       struct ahash_request *areq)
+{
+	struct talitos_ahash_req_ctx *req_ctx = ahash_request_ctx(areq);
+
+	unmap_single_talitos_ptr(dev, &edesc->desc.ptr[5], DMA_FROM_DEVICE);
+
+	/* When using hashctx-in, must unmap it. */
+	if (edesc->desc.ptr[1].len)
+		unmap_single_talitos_ptr(dev, &edesc->desc.ptr[1],
+					 DMA_TO_DEVICE);
+
+	if (edesc->desc.ptr[2].len)
+		unmap_single_talitos_ptr(dev, &edesc->desc.ptr[2],
+					 DMA_TO_DEVICE);
+
+	talitos_sg_unmap(dev, edesc, req_ctx->psrc, NULL);
+
+	if (edesc->dma_len)
+		dma_unmap_single(dev, edesc->dma_link_tbl, edesc->dma_len,
+				 DMA_BIDIRECTIONAL);
+
+}
+
+static void ahash_done(struct device *dev,
+		       struct talitos_desc *desc, void *context,
+		       int err)
+{
+	struct ahash_request *areq = context;
+	struct talitos_edesc *edesc =
+		 container_of(desc, struct talitos_edesc, desc);
+	struct talitos_ahash_req_ctx *req_ctx = ahash_request_ctx(areq);
+
+	if (!req_ctx->last && req_ctx->to_hash_later) {
+		/* Position any partial block for next update/final/finup */
+		memcpy(req_ctx->buf, req_ctx->bufnext, req_ctx->to_hash_later);
+	}
+	common_nonsnoop_hash_unmap(dev, edesc, areq);
+
+	kfree(edesc);
+
+	areq->base.complete(&areq->base, err);
+}
+
+static int common_nonsnoop_hash(struct talitos_edesc *edesc,
+				struct ahash_request *areq, unsigned int length,
+				void (*callback) (struct device *dev,
+						  struct talitos_desc *desc,
+						  void *context, int error))
+{
+	struct crypto_ahash *tfm = crypto_ahash_reqtfm(areq);
+	struct talitos_ctx *ctx = crypto_ahash_ctx(tfm);
+	struct talitos_ahash_req_ctx *req_ctx = ahash_request_ctx(areq);
+	struct device *dev = ctx->dev;
+	struct talitos_desc *desc = &edesc->desc;
+	int sg_count, ret;
+
+	/* first DWORD empty */
+	desc->ptr[0] = zero_entry;
+
+	/* hash context in */
+	if (!req_ctx->first || req_ctx->swinit) {
+		map_single_talitos_ptr(dev, &desc->ptr[1],
+				       req_ctx->hw_context_size,
+				       (char *)req_ctx->hw_context, 0,
+				       DMA_TO_DEVICE);
+		req_ctx->swinit = 0;
+	} else {
+		desc->ptr[1] = zero_entry;
+		/* Indicate next op is not the first. */
+		req_ctx->first = 0;
+	}
+
+	/* HMAC key */
+	if (ctx->keylen)
+		map_single_talitos_ptr(dev, &desc->ptr[2], ctx->keylen,
+				       (char *)&ctx->key, 0, DMA_TO_DEVICE);
+	else
+		desc->ptr[2] = zero_entry;
+
+	/*
+	 * data in
+	 */
+	desc->ptr[3].len = cpu_to_be16(length);
+	desc->ptr[3].j_extent = 0;
+
+	sg_count = talitos_map_sg(dev, req_ctx->psrc,
+				  edesc->src_nents ? : 1,
+				  DMA_TO_DEVICE,
+				  edesc->src_is_chained);
+
+	if (sg_count == 1) {
+		to_talitos_ptr(&desc->ptr[3], sg_dma_address(req_ctx->psrc));
+	} else {
+		sg_count = sg_to_link_tbl(req_ctx->psrc, sg_count, length,
+					  &edesc->link_tbl[0]);
+		if (sg_count > 1) {
+			desc->ptr[3].j_extent |= DESC_PTR_LNKTBL_JUMP;
+			to_talitos_ptr(&desc->ptr[3], edesc->dma_link_tbl);
+			dma_sync_single_for_device(ctx->dev,
+						   edesc->dma_link_tbl,
+						   edesc->dma_len,
+						   DMA_BIDIRECTIONAL);
+		} else {
+			/* Only one segment now, so no link tbl needed */
+			to_talitos_ptr(&desc->ptr[3],
+				       sg_dma_address(req_ctx->psrc));
+		}
+	}
+
+	/* fifth DWORD empty */
+	desc->ptr[4] = zero_entry;
+
+	/* hash/HMAC out -or- hash context out */
+	if (req_ctx->last)
+		map_single_talitos_ptr(dev, &desc->ptr[5],
+				       crypto_ahash_digestsize(tfm),
+				       areq->result, 0, DMA_FROM_DEVICE);
+	else
+		map_single_talitos_ptr(dev, &desc->ptr[5],
+				       req_ctx->hw_context_size,
+				       req_ctx->hw_context, 0, DMA_FROM_DEVICE);
+
+	/* last DWORD empty */
+	desc->ptr[6] = zero_entry;
+
+	ret = talitos_submit(dev, desc, callback, areq);
+	if (ret != -EINPROGRESS) {
+		common_nonsnoop_hash_unmap(dev, edesc, areq);
+		kfree(edesc);
+	}
+	return ret;
+}
+
+static struct talitos_edesc *ahash_edesc_alloc(struct ahash_request *areq,
+					       unsigned int nbytes)
+{
+	struct crypto_ahash *tfm = crypto_ahash_reqtfm(areq);
+	struct talitos_ctx *ctx = crypto_ahash_ctx(tfm);
+	struct talitos_ahash_req_ctx *req_ctx = ahash_request_ctx(areq);
+
+	return talitos_edesc_alloc(ctx->dev, req_ctx->psrc, NULL, 1,
+				   nbytes, 0, 0, areq->base.flags);
+}
+
+static int ahash_init(struct ahash_request *areq)
+{
+	struct crypto_ahash *tfm = crypto_ahash_reqtfm(areq);
+	struct talitos_ahash_req_ctx *req_ctx = ahash_request_ctx(areq);
+
+	/* Initialize the context */
+	req_ctx->count = 0;
+	req_ctx->first = 1; /* first indicates h/w must init its context */
+	req_ctx->swinit = 0; /* assume h/w init of context */
+	req_ctx->hw_context_size =
+		(crypto_ahash_digestsize(tfm) <= SHA256_DIGEST_SIZE)
+			? TALITOS_MDEU_CONTEXT_SIZE_MD5_SHA1_SHA256
+			: TALITOS_MDEU_CONTEXT_SIZE_SHA384_SHA512;
+
+	return 0;
+}
+
+/*
+ * on h/w without explicit sha224 support, we initialize h/w context
+ * manually with sha224 constants, and tell it to run sha256.
+ */
+static int ahash_init_sha224_swinit(struct ahash_request *areq)
+{
+	struct talitos_ahash_req_ctx *req_ctx = ahash_request_ctx(areq);
+
+	ahash_init(areq);
+	req_ctx->swinit = 1;/* prevent h/w initting context with sha256 values*/
+
+	req_ctx->hw_context[0] = cpu_to_be32(SHA224_H0);
+	req_ctx->hw_context[1] = cpu_to_be32(SHA224_H1);
+	req_ctx->hw_context[2] = cpu_to_be32(SHA224_H2);
+	req_ctx->hw_context[3] = cpu_to_be32(SHA224_H3);
+	req_ctx->hw_context[4] = cpu_to_be32(SHA224_H4);
+	req_ctx->hw_context[5] = cpu_to_be32(SHA224_H5);
+	req_ctx->hw_context[6] = cpu_to_be32(SHA224_H6);
+	req_ctx->hw_context[7] = cpu_to_be32(SHA224_H7);
+
+	/* init 64-bit count */
+	req_ctx->hw_context[8] = 0;
+	req_ctx->hw_context[9] = 0;
+
+	return 0;
+}
+
+static int ahash_process_req(struct ahash_request *areq, unsigned int nbytes)
+{
+	struct crypto_ahash *tfm = crypto_ahash_reqtfm(areq);
+	struct talitos_ctx *ctx = crypto_ahash_ctx(tfm);
+	struct talitos_ahash_req_ctx *req_ctx = ahash_request_ctx(areq);
+	struct talitos_edesc *edesc;
+	unsigned int blocksize =
+			crypto_tfm_alg_blocksize(crypto_ahash_tfm(tfm));
+	unsigned int nbytes_to_hash;
+	unsigned int to_hash_later;
+	unsigned int index;
+	int chained;
+
+	index = req_ctx->count & (blocksize - 1);
+	req_ctx->count += nbytes;
+
+	if (!req_ctx->last && (index + nbytes) < blocksize) {
+		/* Buffer the partial block */
+		sg_copy_to_buffer(areq->src,
+				  sg_count(areq->src, nbytes, &chained),
+				  req_ctx->buf + index, nbytes);
+		return 0;
+	}
+
+	if (index) {
+		/* partial block from previous update; chain it in. */
+		sg_init_table(req_ctx->bufsl, (nbytes) ? 2 : 1);
+		sg_set_buf(req_ctx->bufsl, req_ctx->buf, index);
+		if (nbytes)
+			scatterwalk_sg_chain(req_ctx->bufsl, 2,
+					     areq->src);
+		req_ctx->psrc = req_ctx->bufsl;
+	} else {
+		req_ctx->psrc = areq->src;
+	}
+	nbytes_to_hash =  index + nbytes;
+	if (!req_ctx->last) {
+		to_hash_later = (nbytes_to_hash & (blocksize - 1));
+		if (to_hash_later) {
+			int nents;
+			/* Must copy to_hash_later bytes from the end
+			 * to bufnext (a partial block) for later.
+			 */
+			nents = sg_count(areq->src, nbytes, &chained);
+			sg_copy_end_to_buffer(areq->src, nents,
+					      req_ctx->bufnext,
+					      to_hash_later,
+					      nbytes - to_hash_later);
+
+			/* Adjust count for what will be hashed now */
+			nbytes_to_hash -= to_hash_later;
+		}
+		req_ctx->to_hash_later = to_hash_later;
+	}
+
+	/* allocate extended descriptor */
+	edesc = ahash_edesc_alloc(areq, nbytes_to_hash);
+	if (IS_ERR(edesc))
+		return PTR_ERR(edesc);
+
+	edesc->desc.hdr = ctx->desc_hdr_template;
+
+	/* On last one, request SEC to pad; otherwise continue */
+	if (req_ctx->last)
+		edesc->desc.hdr |= DESC_HDR_MODE0_MDEU_PAD;
+	else
+		edesc->desc.hdr |= DESC_HDR_MODE0_MDEU_CONT;
+
+	/* request SEC to INIT hash. */
+	if (req_ctx->first && !req_ctx->swinit)
+		edesc->desc.hdr |= DESC_HDR_MODE0_MDEU_INIT;
+
+	/* When the tfm context has a keylen, it's an HMAC.
+	 * A first or last (ie. not middle) descriptor must request HMAC.
+	 */
+	if (ctx->keylen && (req_ctx->first || req_ctx->last))
+		edesc->desc.hdr |= DESC_HDR_MODE0_MDEU_HMAC;
+
+	return common_nonsnoop_hash(edesc, areq, nbytes_to_hash,
+				    ahash_done);
+}
+
+static int ahash_update(struct ahash_request *areq)
+{
+	struct talitos_ahash_req_ctx *req_ctx = ahash_request_ctx(areq);
+
+	req_ctx->last = 0;
+
+	return ahash_process_req(areq, areq->nbytes);
+}
+
+static int ahash_final(struct ahash_request *areq)
+{
+	struct talitos_ahash_req_ctx *req_ctx = ahash_request_ctx(areq);
+
+	req_ctx->last = 1;
+
+	return ahash_process_req(areq, 0);
+}
+
+static int ahash_finup(struct ahash_request *areq)
+{
+	struct talitos_ahash_req_ctx *req_ctx = ahash_request_ctx(areq);
+
+	req_ctx->last = 1;
+
+	return ahash_process_req(areq, areq->nbytes);
+}
+
+static int ahash_digest(struct ahash_request *areq)
+{
+	struct talitos_ahash_req_ctx *req_ctx = ahash_request_ctx(areq);
+	struct crypto_ahash *ahash = crypto_ahash_reqtfm(areq);
+
+	ahash->init(areq);
+	req_ctx->last = 1;
+
+	return ahash_process_req(areq, areq->nbytes);
+}
+
 struct talitos_alg_template {
-	struct crypto_alg alg;
+	u32 type;
+	union {
+		struct crypto_alg crypto;
+		struct ahash_alg hash;
+	} alg;
 	__be32 desc_hdr_template;
 };
 
 static struct talitos_alg_template driver_algs[] = {
 	/* AEAD algorithms.  These use a single-pass ipsec_esp descriptor */
-	{
-		.alg = {
+	{	.type = CRYPTO_ALG_TYPE_AEAD,
+		.alg.crypto = {
 			.cra_name = "authenc(hmac(sha1),cbc(aes))",
 			.cra_driver_name = "authenc-hmac-sha1-cbc-aes-talitos",
 			.cra_blocksize = AES_BLOCK_SIZE,
@@ -1511,8 +1919,8 @@ static struct talitos_alg_template driver_algs[] = {
 		                     DESC_HDR_MODE1_MDEU_PAD |
 		                     DESC_HDR_MODE1_MDEU_SHA1_HMAC,
 	},
-	{
-		.alg = {
+	{	.type = CRYPTO_ALG_TYPE_AEAD,
+		.alg.crypto = {
 			.cra_name = "authenc(hmac(sha1),cbc(des3_ede))",
 			.cra_driver_name = "authenc-hmac-sha1-cbc-3des-talitos",
 			.cra_blocksize = DES3_EDE_BLOCK_SIZE,
@@ -1538,8 +1946,8 @@ static struct talitos_alg_template driver_algs[] = {
 		                     DESC_HDR_MODE1_MDEU_PAD |
 		                     DESC_HDR_MODE1_MDEU_SHA1_HMAC,
 	},
-	{
-		.alg = {
+	{	.type = CRYPTO_ALG_TYPE_AEAD,
+		.alg.crypto = {
 			.cra_name = "authenc(hmac(sha256),cbc(aes))",
 			.cra_driver_name = "authenc-hmac-sha256-cbc-aes-talitos",
 			.cra_blocksize = AES_BLOCK_SIZE,
@@ -1564,8 +1972,8 @@ static struct talitos_alg_template driver_algs[] = {
 		                     DESC_HDR_MODE1_MDEU_PAD |
 		                     DESC_HDR_MODE1_MDEU_SHA256_HMAC,
 	},
-	{
-		.alg = {
+	{	.type = CRYPTO_ALG_TYPE_AEAD,
+		.alg.crypto = {
 			.cra_name = "authenc(hmac(sha256),cbc(des3_ede))",
 			.cra_driver_name = "authenc-hmac-sha256-cbc-3des-talitos",
 			.cra_blocksize = DES3_EDE_BLOCK_SIZE,
@@ -1591,8 +1999,8 @@ static struct talitos_alg_template driver_algs[] = {
 		                     DESC_HDR_MODE1_MDEU_PAD |
 		                     DESC_HDR_MODE1_MDEU_SHA256_HMAC,
 	},
-	{
-		.alg = {
+	{	.type = CRYPTO_ALG_TYPE_AEAD,
+		.alg.crypto = {
 			.cra_name = "authenc(hmac(md5),cbc(aes))",
 			.cra_driver_name = "authenc-hmac-md5-cbc-aes-talitos",
 			.cra_blocksize = AES_BLOCK_SIZE,
@@ -1617,8 +2025,8 @@ static struct talitos_alg_template driver_algs[] = {
 		                     DESC_HDR_MODE1_MDEU_PAD |
 		                     DESC_HDR_MODE1_MDEU_MD5_HMAC,
 	},
-	{
-		.alg = {
+	{	.type = CRYPTO_ALG_TYPE_AEAD,
+		.alg.crypto = {
 			.cra_name = "authenc(hmac(md5),cbc(des3_ede))",
 			.cra_driver_name = "authenc-hmac-md5-cbc-3des-talitos",
 			.cra_blocksize = DES3_EDE_BLOCK_SIZE,
@@ -1645,8 +2053,8 @@ static struct talitos_alg_template driver_algs[] = {
 		                     DESC_HDR_MODE1_MDEU_MD5_HMAC,
 	},
 	/* ABLKCIPHER algorithms. */
-	{
-		.alg = {
+	{	.type = CRYPTO_ALG_TYPE_ABLKCIPHER,
+		.alg.crypto = {
 			.cra_name = "cbc(aes)",
 			.cra_driver_name = "cbc-aes-talitos",
 			.cra_blocksize = AES_BLOCK_SIZE,
@@ -1667,8 +2075,8 @@ static struct talitos_alg_template driver_algs[] = {
 				     DESC_HDR_SEL0_AESU |
 				     DESC_HDR_MODE0_AESU_CBC,
 	},
-	{
-		.alg = {
+	{	.type = CRYPTO_ALG_TYPE_ABLKCIPHER,
+		.alg.crypto = {
 			.cra_name = "cbc(des3_ede)",
 			.cra_driver_name = "cbc-3des-talitos",
 			.cra_blocksize = DES3_EDE_BLOCK_SIZE,
@@ -1689,14 +2097,140 @@ static struct talitos_alg_template driver_algs[] = {
 			             DESC_HDR_SEL0_DEU |
 		                     DESC_HDR_MODE0_DEU_CBC |
 		                     DESC_HDR_MODE0_DEU_3DES,
-	}
+	},
+	/* AHASH algorithms. */
+	{	.type = CRYPTO_ALG_TYPE_AHASH,
+		.alg.hash = {
+			.init = ahash_init,
+			.update = ahash_update,
+			.final = ahash_final,
+			.finup = ahash_finup,
+			.digest = ahash_digest,
+			.halg.digestsize = MD5_DIGEST_SIZE,
+			.halg.base = {
+				.cra_name = "md5",
+				.cra_driver_name = "md5-talitos",
+				.cra_blocksize = MD5_BLOCK_SIZE,
+				.cra_flags = CRYPTO_ALG_TYPE_AHASH |
+					     CRYPTO_ALG_ASYNC,
+				.cra_type = &crypto_ahash_type
+			}
+		},
+		.desc_hdr_template = DESC_HDR_TYPE_COMMON_NONSNOOP_NO_AFEU |
+				     DESC_HDR_SEL0_MDEUA |
+				     DESC_HDR_MODE0_MDEU_MD5,
+	},
+	{	.type = CRYPTO_ALG_TYPE_AHASH,
+		.alg.hash = {
+			.init = ahash_init,
+			.update = ahash_update,
+			.final = ahash_final,
+			.finup = ahash_finup,
+			.digest = ahash_digest,
+			.halg.digestsize = SHA1_DIGEST_SIZE,
+			.halg.base = {
+				.cra_name = "sha1",
+				.cra_driver_name = "sha1-talitos",
+				.cra_blocksize = SHA1_BLOCK_SIZE,
+				.cra_flags = CRYPTO_ALG_TYPE_AHASH |
+					     CRYPTO_ALG_ASYNC,
+				.cra_type = &crypto_ahash_type
+			}
+		},
+		.desc_hdr_template = DESC_HDR_TYPE_COMMON_NONSNOOP_NO_AFEU |
+				     DESC_HDR_SEL0_MDEUA |
+				     DESC_HDR_MODE0_MDEU_SHA1,
+	},
+	{	.type = CRYPTO_ALG_TYPE_AHASH,
+		.alg.hash = {
+			.init = ahash_init,
+			.update = ahash_update,
+			.final = ahash_final,
+			.finup = ahash_finup,
+			.digest = ahash_digest,
+			.halg.digestsize = SHA224_DIGEST_SIZE,
+			.halg.base = {
+				.cra_name = "sha224",
+				.cra_driver_name = "sha224-talitos",
+				.cra_blocksize = SHA224_BLOCK_SIZE,
+				.cra_flags = CRYPTO_ALG_TYPE_AHASH |
+					     CRYPTO_ALG_ASYNC,
+				.cra_type = &crypto_ahash_type
+			}
+		},
+		.desc_hdr_template = DESC_HDR_TYPE_COMMON_NONSNOOP_NO_AFEU |
+				     DESC_HDR_SEL0_MDEUA |
+				     DESC_HDR_MODE0_MDEU_SHA224,
+	},
+	{	.type = CRYPTO_ALG_TYPE_AHASH,
+		.alg.hash = {
+			.init = ahash_init,
+			.update = ahash_update,
+			.final = ahash_final,
+			.finup = ahash_finup,
+			.digest = ahash_digest,
+			.halg.digestsize = SHA256_DIGEST_SIZE,
+			.halg.base = {
+				.cra_name = "sha256",
+				.cra_driver_name = "sha256-talitos",
+				.cra_blocksize = SHA256_BLOCK_SIZE,
+				.cra_flags = CRYPTO_ALG_TYPE_AHASH |
+					     CRYPTO_ALG_ASYNC,
+				.cra_type = &crypto_ahash_type
+			}
+		},
+		.desc_hdr_template = DESC_HDR_TYPE_COMMON_NONSNOOP_NO_AFEU |
+				     DESC_HDR_SEL0_MDEUA |
+				     DESC_HDR_MODE0_MDEU_SHA256,
+	},
+	{	.type = CRYPTO_ALG_TYPE_AHASH,
+		.alg.hash = {
+			.init = ahash_init,
+			.update = ahash_update,
+			.final = ahash_final,
+			.finup = ahash_finup,
+			.digest = ahash_digest,
+			.halg.digestsize = SHA384_DIGEST_SIZE,
+			.halg.base = {
+				.cra_name = "sha384",
+				.cra_driver_name = "sha384-talitos",
+				.cra_blocksize = SHA384_BLOCK_SIZE,
+				.cra_flags = CRYPTO_ALG_TYPE_AHASH |
+					     CRYPTO_ALG_ASYNC,
+				.cra_type = &crypto_ahash_type
+			}
+		},
+		.desc_hdr_template = DESC_HDR_TYPE_COMMON_NONSNOOP_NO_AFEU |
+				     DESC_HDR_SEL0_MDEUB |
+				     DESC_HDR_MODE0_MDEUB_SHA384,
+	},
+	{	.type = CRYPTO_ALG_TYPE_AHASH,
+		.alg.hash = {
+			.init = ahash_init,
+			.update = ahash_update,
+			.final = ahash_final,
+			.finup = ahash_finup,
+			.digest = ahash_digest,
+			.halg.digestsize = SHA512_DIGEST_SIZE,
+			.halg.base = {
+				.cra_name = "sha512",
+				.cra_driver_name = "sha512-talitos",
+				.cra_blocksize = SHA512_BLOCK_SIZE,
+				.cra_flags = CRYPTO_ALG_TYPE_AHASH |
+					     CRYPTO_ALG_ASYNC,
+				.cra_type = &crypto_ahash_type
+			}
+		},
+		.desc_hdr_template = DESC_HDR_TYPE_COMMON_NONSNOOP_NO_AFEU |
+				     DESC_HDR_SEL0_MDEUB |
+				     DESC_HDR_MODE0_MDEUB_SHA512,
+	},
 };
 
 struct talitos_crypto_alg {
 	struct list_head entry;
 	struct device *dev;
-	__be32 desc_hdr_template;
-	struct crypto_alg crypto_alg;
+	struct talitos_alg_template algt;
 };
 
 static int talitos_cra_init(struct crypto_tfm *tfm)
@@ -1705,13 +2239,28 @@ static int talitos_cra_init(struct crypto_tfm *tfm)
 	struct talitos_crypto_alg *talitos_alg;
 	struct talitos_ctx *ctx = crypto_tfm_ctx(tfm);
 
-	talitos_alg =  container_of(alg, struct talitos_crypto_alg, crypto_alg);
+	if ((alg->cra_flags & CRYPTO_ALG_TYPE_MASK) == CRYPTO_ALG_TYPE_AHASH)
+		talitos_alg = container_of(__crypto_ahash_alg(alg),
+					   struct talitos_crypto_alg,
+					   algt.alg.hash);
+	else
+		talitos_alg = container_of(alg, struct talitos_crypto_alg,
+					   algt.alg.crypto);
 
 	/* update context with ptr to dev */
 	ctx->dev = talitos_alg->dev;
 
 	/* copy descriptor header template value */
-	ctx->desc_hdr_template = talitos_alg->desc_hdr_template;
+	ctx->desc_hdr_template = talitos_alg->algt.desc_hdr_template;
+
+	return 0;
+}
+
+static int talitos_cra_init_aead(struct crypto_tfm *tfm)
+{
+	struct talitos_ctx *ctx = crypto_tfm_ctx(tfm);
+
+	talitos_cra_init(tfm);
 
 	/* random first IV */
 	get_random_bytes(ctx->iv, TALITOS_MAX_IV_LENGTH);
@@ -1719,6 +2268,19 @@ static int talitos_cra_init(struct crypto_tfm *tfm)
 	return 0;
 }
 
+static int talitos_cra_init_ahash(struct crypto_tfm *tfm)
+{
+	struct talitos_ctx *ctx = crypto_tfm_ctx(tfm);
+
+	talitos_cra_init(tfm);
+
+	ctx->keylen = 0;
+	crypto_ahash_set_reqsize(__crypto_ahash_cast(tfm),
+				 sizeof(struct talitos_ahash_req_ctx));
+
+	return 0;
+}
+
 /*
  * given the alg's descriptor header template, determine whether descriptor
  * type and primary/secondary execution units required match the hw
@@ -1747,7 +2309,15 @@ static int talitos_remove(struct of_device *ofdev)
 	int i;
 
 	list_for_each_entry_safe(t_alg, n, &priv->alg_list, entry) {
-		crypto_unregister_alg(&t_alg->crypto_alg);
+		switch (t_alg->algt.type) {
+		case CRYPTO_ALG_TYPE_ABLKCIPHER:
+		case CRYPTO_ALG_TYPE_AEAD:
+			crypto_unregister_alg(&t_alg->algt.alg.crypto);
+			break;
+		case CRYPTO_ALG_TYPE_AHASH:
+			crypto_unregister_ahash(&t_alg->algt.alg.hash);
+			break;
+		}
 		list_del(&t_alg->entry);
 		kfree(t_alg);
 	}
@@ -1781,6 +2351,7 @@ static struct talitos_crypto_alg *talitos_alg_alloc(struct device *dev,
 						    struct talitos_alg_template
 						           *template)
 {
+	struct talitos_private *priv = dev_get_drvdata(dev);
 	struct talitos_crypto_alg *t_alg;
 	struct crypto_alg *alg;
 
@@ -1788,16 +2359,36 @@ static struct talitos_crypto_alg *talitos_alg_alloc(struct device *dev,
 	if (!t_alg)
 		return ERR_PTR(-ENOMEM);
 
-	alg = &t_alg->crypto_alg;
-	*alg = template->alg;
+	t_alg->algt = *template;
+
+	switch (t_alg->algt.type) {
+	case CRYPTO_ALG_TYPE_ABLKCIPHER:
+		alg = &t_alg->algt.alg.crypto;
+		alg->cra_init = talitos_cra_init;
+		break;
+	case CRYPTO_ALG_TYPE_AEAD:
+		alg = &t_alg->algt.alg.crypto;
+		alg->cra_init = talitos_cra_init_aead;
+		break;
+	case CRYPTO_ALG_TYPE_AHASH:
+		alg = &t_alg->algt.alg.hash.halg.base;
+		alg->cra_init = talitos_cra_init_ahash;
+		if (!(priv->features & TALITOS_FTR_SHA224_HWINIT) &&
+		    !strcmp(alg->cra_name, "sha224")) {
+			t_alg->algt.alg.hash.init = ahash_init_sha224_swinit;
+			t_alg->algt.desc_hdr_template =
+					DESC_HDR_TYPE_COMMON_NONSNOOP_NO_AFEU |
+					DESC_HDR_SEL0_MDEUA |
+					DESC_HDR_MODE0_MDEU_SHA256;
+		}
+		break;
+	}
 
 	alg->cra_module = THIS_MODULE;
-	alg->cra_init = talitos_cra_init;
 	alg->cra_priority = TALITOS_CRA_PRIORITY;
 	alg->cra_alignmask = 0;
 	alg->cra_ctxsize = sizeof(struct talitos_ctx);
 
-	t_alg->desc_hdr_template = template->desc_hdr_template;
 	t_alg->dev = dev;
 
 	return t_alg;
@@ -1807,7 +2398,7 @@ static int talitos_probe(struct of_device *ofdev,
 			 const struct of_device_id *match)
 {
 	struct device *dev = &ofdev->dev;
-	struct device_node *np = ofdev->node;
+	struct device_node *np = ofdev->dev.of_node;
 	struct talitos_private *priv;
 	const unsigned int *prop;
 	int i, err;
@@ -1877,7 +2468,8 @@ static int talitos_probe(struct of_device *ofdev,
 		priv->features |= TALITOS_FTR_SRC_LINK_TBL_LEN_INCLUDES_EXTENT;
 
 	if (of_device_is_compatible(np, "fsl,sec2.1"))
-		priv->features |= TALITOS_FTR_HW_AUTH_CHECK;
+		priv->features |= TALITOS_FTR_HW_AUTH_CHECK |
+				  TALITOS_FTR_SHA224_HWINIT;
 
 	priv->chan = kzalloc(sizeof(struct talitos_channel) *
 			     priv->num_channels, GFP_KERNEL);
@@ -1931,6 +2523,7 @@ static int talitos_probe(struct of_device *ofdev,
 	for (i = 0; i < ARRAY_SIZE(driver_algs); i++) {
 		if (hw_supports(dev, driver_algs[i].desc_hdr_template)) {
 			struct talitos_crypto_alg *t_alg;
+			char *name = NULL;
 
 			t_alg = talitos_alg_alloc(dev, &driver_algs[i]);
 			if (IS_ERR(t_alg)) {
@@ -1938,15 +2531,27 @@ static int talitos_probe(struct of_device *ofdev,
 				goto err_out;
 			}
 
-			err = crypto_register_alg(&t_alg->crypto_alg);
+			switch (t_alg->algt.type) {
+			case CRYPTO_ALG_TYPE_ABLKCIPHER:
+			case CRYPTO_ALG_TYPE_AEAD:
+				err = crypto_register_alg(
+						&t_alg->algt.alg.crypto);
+				name = t_alg->algt.alg.crypto.cra_driver_name;
+				break;
+			case CRYPTO_ALG_TYPE_AHASH:
+				err = crypto_register_ahash(
+						&t_alg->algt.alg.hash);
+				name =
+				 t_alg->algt.alg.hash.halg.base.cra_driver_name;
+				break;
+			}
 			if (err) {
 				dev_err(dev, "%s alg registration failed\n",
-					t_alg->crypto_alg.cra_driver_name);
+					name);
 				kfree(t_alg);
 			} else {
 				list_add_tail(&t_alg->entry, &priv->alg_list);
-				dev_info(dev, "%s\n",
-					 t_alg->crypto_alg.cra_driver_name);
+				dev_info(dev, "%s\n", name);
 			}
 		}
 	}
@@ -1968,8 +2573,11 @@ static const struct of_device_id talitos_match[] = {
 MODULE_DEVICE_TABLE(of, talitos_match);
 
 static struct of_platform_driver talitos_driver = {
-	.name = "talitos",
-	.match_table = talitos_match,
+	.driver = {
+		.name = "talitos",
+		.owner = THIS_MODULE,
+		.of_match_table = talitos_match,
+	},
 	.probe = talitos_probe,
 	.remove = talitos_remove,
 };
diff --git a/drivers/crypto/talitos.h b/drivers/crypto/talitos.h
index ff5a1450e145..0b746aca4587 100644
--- a/drivers/crypto/talitos.h
+++ b/drivers/crypto/talitos.h
@@ -1,7 +1,7 @@
 /*
  * Freescale SEC (talitos) device register and descriptor header defines
  *
- * Copyright (c) 2006-2008 Freescale Semiconductor, Inc.
+ * Copyright (c) 2006-2010 Freescale Semiconductor, Inc.
  *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
@@ -130,6 +130,9 @@
 #define TALITOS_CRCUISR			0xf030 /* cyclic redundancy check unit*/
 #define TALITOS_CRCUISR_LO		0xf034
 
+#define TALITOS_MDEU_CONTEXT_SIZE_MD5_SHA1_SHA256	0x28
+#define TALITOS_MDEU_CONTEXT_SIZE_SHA384_SHA512		0x48
+
 /*
  * talitos descriptor header (hdr) bits
  */
@@ -157,12 +160,16 @@
 #define	DESC_HDR_MODE0_AESU_CBC		cpu_to_be32(0x00200000)
 #define	DESC_HDR_MODE0_DEU_CBC		cpu_to_be32(0x00400000)
 #define	DESC_HDR_MODE0_DEU_3DES		cpu_to_be32(0x00200000)
+#define	DESC_HDR_MODE0_MDEU_CONT	cpu_to_be32(0x08000000)
 #define	DESC_HDR_MODE0_MDEU_INIT	cpu_to_be32(0x01000000)
 #define	DESC_HDR_MODE0_MDEU_HMAC	cpu_to_be32(0x00800000)
 #define	DESC_HDR_MODE0_MDEU_PAD		cpu_to_be32(0x00400000)
+#define	DESC_HDR_MODE0_MDEU_SHA224	cpu_to_be32(0x00300000)
 #define	DESC_HDR_MODE0_MDEU_MD5		cpu_to_be32(0x00200000)
 #define	DESC_HDR_MODE0_MDEU_SHA256	cpu_to_be32(0x00100000)
 #define	DESC_HDR_MODE0_MDEU_SHA1	cpu_to_be32(0x00000000)
+#define	DESC_HDR_MODE0_MDEUB_SHA384	cpu_to_be32(0x00000000)
+#define	DESC_HDR_MODE0_MDEUB_SHA512	cpu_to_be32(0x00200000)
 #define	DESC_HDR_MODE0_MDEU_MD5_HMAC	(DESC_HDR_MODE0_MDEU_MD5 | \
 					 DESC_HDR_MODE0_MDEU_HMAC)
 #define	DESC_HDR_MODE0_MDEU_SHA256_HMAC	(DESC_HDR_MODE0_MDEU_SHA256 | \
@@ -181,9 +188,12 @@
 #define	DESC_HDR_MODE1_MDEU_INIT	cpu_to_be32(0x00001000)
 #define	DESC_HDR_MODE1_MDEU_HMAC	cpu_to_be32(0x00000800)
 #define	DESC_HDR_MODE1_MDEU_PAD		cpu_to_be32(0x00000400)
+#define	DESC_HDR_MODE1_MDEU_SHA224	cpu_to_be32(0x00000300)
 #define	DESC_HDR_MODE1_MDEU_MD5		cpu_to_be32(0x00000200)
 #define	DESC_HDR_MODE1_MDEU_SHA256	cpu_to_be32(0x00000100)
 #define	DESC_HDR_MODE1_MDEU_SHA1	cpu_to_be32(0x00000000)
+#define	DESC_HDR_MODE1_MDEUB_SHA384	cpu_to_be32(0x00000000)
+#define	DESC_HDR_MODE1_MDEUB_SHA512	cpu_to_be32(0x00000200)
 #define	DESC_HDR_MODE1_MDEU_MD5_HMAC	(DESC_HDR_MODE1_MDEU_MD5 | \
 					 DESC_HDR_MODE1_MDEU_HMAC)
 #define	DESC_HDR_MODE1_MDEU_SHA256_HMAC	(DESC_HDR_MODE1_MDEU_SHA256 | \
diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
index c27f80e5d531..9e01e96fee94 100644
--- a/drivers/dma/Kconfig
+++ b/drivers/dma/Kconfig
@@ -141,6 +141,13 @@ config COH901318
 	help
 	  Enable support for ST-Ericsson COH 901 318 DMA.
 
+config STE_DMA40
+	bool "ST-Ericsson DMA40 support"
+	depends on ARCH_U8500
+	select DMA_ENGINE
+	help
+	  Support for ST-Ericsson DMA40 controller
+
 config AMCC_PPC440SPE_ADMA
 	tristate "AMCC PPC440SPe ADMA support"
 	depends on 440SPe || 440SP
@@ -149,9 +156,25 @@ config AMCC_PPC440SPE_ADMA
 	help
 	  Enable support for the AMCC PPC440SPe RAID engines.
 
+config TIMB_DMA
+	tristate "Timberdale FPGA DMA support"
+	depends on MFD_TIMBERDALE || HAS_IOMEM
+	select DMA_ENGINE
+	help
+	  Enable support for the Timberdale FPGA DMA engine.
+
 config ARCH_HAS_ASYNC_TX_FIND_CHANNEL
 	bool
 
+config PL330_DMA
+	tristate "DMA API Driver for PL330"
+	select DMA_ENGINE
+	depends on PL330
+	help
+	  Select if your platform has one or more PL330 DMACs.
+	  You need to provide platform specific settings via
+	  platform_data for a dma-pl330 device.
+
 config DMA_ENGINE
 	bool
 
diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
index 22bba3d5e2b6..0fe5ebbfda5d 100644
--- a/drivers/dma/Makefile
+++ b/drivers/dma/Makefile
@@ -20,3 +20,6 @@ obj-$(CONFIG_TXX9_DMAC) += txx9dmac.o
 obj-$(CONFIG_SH_DMAE) += shdma.o
 obj-$(CONFIG_COH901318) += coh901318.o coh901318_lli.o
 obj-$(CONFIG_AMCC_PPC440SPE_ADMA) += ppc4xx/
+obj-$(CONFIG_TIMB_DMA) += timb_dma.o
+obj-$(CONFIG_STE_DMA40) += ste_dma40.o ste_dma40_ll.o
+obj-$(CONFIG_PL330_DMA) += pl330.o
diff --git a/drivers/dma/at_hdmac.c b/drivers/dma/at_hdmac.c
index 308ab320e20b..e88076022a7a 100644
--- a/drivers/dma/at_hdmac.c
+++ b/drivers/dma/at_hdmac.c
@@ -760,13 +760,18 @@ err_desc_get:
 	return NULL;
 }
 
-static void atc_terminate_all(struct dma_chan *chan)
+static int atc_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+		       unsigned long arg)
 {
 	struct at_dma_chan	*atchan = to_at_dma_chan(chan);
 	struct at_dma		*atdma = to_at_dma(chan->device);
 	struct at_desc		*desc, *_desc;
 	LIST_HEAD(list);
 
+	/* Only supports DMA_TERMINATE_ALL */
+	if (cmd != DMA_TERMINATE_ALL)
+		return -ENXIO;
+
 	/*
 	 * This is only called when something went wrong elsewhere, so
 	 * we don't really care about the data. Just disable the
@@ -790,32 +795,30 @@ static void atc_terminate_all(struct dma_chan *chan)
 	/* Flush all pending and queued descriptors */
 	list_for_each_entry_safe(desc, _desc, &list, desc_node)
 		atc_chain_complete(atchan, desc);
+
+	return 0;
 }
 
 /**
- * atc_is_tx_complete - poll for transaction completion
+ * atc_tx_status - poll for transaction completion
  * @chan: DMA channel
  * @cookie: transaction identifier to check status of
- * @done: if not %NULL, updated with last completed transaction
- * @used: if not %NULL, updated with last used transaction
+ * @txstate: if not %NULL updated with transaction state
  *
- * If @done and @used are passed in, upon return they reflect the driver
+ * If @txstate is passed in, upon return it reflect the driver
  * internal state and can be used with dma_async_is_complete() to check
  * the status of multiple cookies without re-checking hardware state.
  */
 static enum dma_status
-atc_is_tx_complete(struct dma_chan *chan,
+atc_tx_status(struct dma_chan *chan,
 		dma_cookie_t cookie,
-		dma_cookie_t *done, dma_cookie_t *used)
+		struct dma_tx_state *txstate)
 {
 	struct at_dma_chan	*atchan = to_at_dma_chan(chan);
 	dma_cookie_t		last_used;
 	dma_cookie_t		last_complete;
 	enum dma_status		ret;
 
-	dev_vdbg(chan2dev(chan), "is_tx_complete: %d (d%d, u%d)\n",
-			cookie, done ? *done : 0, used ? *used : 0);
-
 	spin_lock_bh(&atchan->lock);
 
 	last_complete = atchan->completed_cookie;
@@ -833,10 +836,10 @@ atc_is_tx_complete(struct dma_chan *chan,
 
 	spin_unlock_bh(&atchan->lock);
 
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
+	dev_vdbg(chan2dev(chan), "tx_status: %d (d%d, u%d)\n",
+		 cookie, last_complete ? last_complete : 0,
+		 last_used ? last_used : 0);
 
 	return ret;
 }
@@ -1082,7 +1085,7 @@ static int __init at_dma_probe(struct platform_device *pdev)
 	/* set base routines */
 	atdma->dma_common.device_alloc_chan_resources = atc_alloc_chan_resources;
 	atdma->dma_common.device_free_chan_resources = atc_free_chan_resources;
-	atdma->dma_common.device_is_tx_complete = atc_is_tx_complete;
+	atdma->dma_common.device_tx_status = atc_tx_status;
 	atdma->dma_common.device_issue_pending = atc_issue_pending;
 	atdma->dma_common.dev = &pdev->dev;
 
@@ -1092,7 +1095,7 @@ static int __init at_dma_probe(struct platform_device *pdev)
 
 	if (dma_has_cap(DMA_SLAVE, atdma->dma_common.cap_mask)) {
 		atdma->dma_common.device_prep_slave_sg = atc_prep_slave_sg;
-		atdma->dma_common.device_terminate_all = atc_terminate_all;
+		atdma->dma_common.device_control = atc_control;
 	}
 
 	dma_writel(atdma, EN, AT_DMA_ENABLE);
diff --git a/drivers/dma/coh901318.c b/drivers/dma/coh901318.c
index 1656fdcdb6c2..a724e6be1b4d 100644
--- a/drivers/dma/coh901318.c
+++ b/drivers/dma/coh901318.c
@@ -37,7 +37,7 @@ struct coh901318_desc {
 	struct list_head node;
 	struct scatterlist *sg;
 	unsigned int sg_len;
-	struct coh901318_lli *data;
+	struct coh901318_lli *lli;
 	enum dma_data_direction dir;
 	unsigned long flags;
 };
@@ -283,7 +283,7 @@ static int coh901318_start(struct coh901318_chan *cohc)
 }
 
 static int coh901318_prep_linked_list(struct coh901318_chan *cohc,
-				      struct coh901318_lli *data)
+				      struct coh901318_lli *lli)
 {
 	int channel = cohc->id;
 	void __iomem *virtbase = cohc->base->virtbase;
@@ -292,18 +292,18 @@ static int coh901318_prep_linked_list(struct coh901318_chan *cohc,
 		     COH901318_CX_STAT_SPACING*channel) &
 	       COH901318_CX_STAT_ACTIVE);
 
-	writel(data->src_addr,
+	writel(lli->src_addr,
 	       virtbase + COH901318_CX_SRC_ADDR +
 	       COH901318_CX_SRC_ADDR_SPACING * channel);
 
-	writel(data->dst_addr, virtbase +
+	writel(lli->dst_addr, virtbase +
 	       COH901318_CX_DST_ADDR +
 	       COH901318_CX_DST_ADDR_SPACING * channel);
 
-	writel(data->link_addr, virtbase + COH901318_CX_LNK_ADDR +
+	writel(lli->link_addr, virtbase + COH901318_CX_LNK_ADDR +
 	       COH901318_CX_LNK_ADDR_SPACING * channel);
 
-	writel(data->control, virtbase + COH901318_CX_CTRL +
+	writel(lli->control, virtbase + COH901318_CX_CTRL +
 	       COH901318_CX_CTRL_SPACING * channel);
 
 	return 0;
@@ -408,33 +408,107 @@ coh901318_first_queued(struct coh901318_chan *cohc)
 	return d;
 }
 
+static inline u32 coh901318_get_bytes_in_lli(struct coh901318_lli *in_lli)
+{
+	struct coh901318_lli *lli = in_lli;
+	u32 bytes = 0;
+
+	while (lli) {
+		bytes += lli->control & COH901318_CX_CTRL_TC_VALUE_MASK;
+		lli = lli->virt_link_addr;
+	}
+	return bytes;
+}
+
 /*
- * DMA start/stop controls
+ * Get the number of bytes left to transfer on this channel,
+ * it is unwise to call this before stopping the channel for
+ * absolute measures, but for a rough guess you can still call
+ * it.
  */
-u32 coh901318_get_bytes_left(struct dma_chan *chan)
+static u32 coh901318_get_bytes_left(struct dma_chan *chan)
 {
-	unsigned long flags;
-	u32 ret;
 	struct coh901318_chan *cohc = to_coh901318_chan(chan);
+	struct coh901318_desc *cohd;
+	struct list_head *pos;
+	unsigned long flags;
+	u32 left = 0;
+	int i = 0;
 
 	spin_lock_irqsave(&cohc->lock, flags);
 
-	/* Read transfer count value */
-	ret = readl(cohc->base->virtbase +
-		    COH901318_CX_CTRL+COH901318_CX_CTRL_SPACING *
-		    cohc->id) & COH901318_CX_CTRL_TC_VALUE_MASK;
+	/*
+	 * If there are many queued jobs, we iterate and add the
+	 * size of them all. We take a special look on the first
+	 * job though, since it is probably active.
+	 */
+	list_for_each(pos, &cohc->active) {
+		/*
+		 * The first job in the list will be working on the
+		 * hardware. The job can be stopped but still active,
+		 * so that the transfer counter is somewhere inside
+		 * the buffer.
+		 */
+		cohd = list_entry(pos, struct coh901318_desc, node);
+
+		if (i == 0) {
+			struct coh901318_lli *lli;
+			dma_addr_t ladd;
+
+			/* Read current transfer count value */
+			left = readl(cohc->base->virtbase +
+				     COH901318_CX_CTRL +
+				     COH901318_CX_CTRL_SPACING * cohc->id) &
+				COH901318_CX_CTRL_TC_VALUE_MASK;
+
+			/* See if the transfer is linked... */
+			ladd = readl(cohc->base->virtbase +
+				     COH901318_CX_LNK_ADDR +
+				     COH901318_CX_LNK_ADDR_SPACING *
+				     cohc->id) &
+				~COH901318_CX_LNK_LINK_IMMEDIATE;
+			/* Single transaction */
+			if (!ladd)
+				continue;
+
+			/*
+			 * Linked transaction, follow the lli, find the
+			 * currently processing lli, and proceed to the next
+			 */
+			lli = cohd->lli;
+			while (lli && lli->link_addr != ladd)
+				lli = lli->virt_link_addr;
+
+			if (lli)
+				lli = lli->virt_link_addr;
+
+			/*
+			 * Follow remaining lli links around to count the total
+			 * number of bytes left
+			 */
+			left += coh901318_get_bytes_in_lli(lli);
+		} else {
+			left += coh901318_get_bytes_in_lli(cohd->lli);
+		}
+		i++;
+	}
+
+	/* Also count bytes in the queued jobs */
+	list_for_each(pos, &cohc->queue) {
+		cohd = list_entry(pos, struct coh901318_desc, node);
+		left += coh901318_get_bytes_in_lli(cohd->lli);
+	}
 
 	spin_unlock_irqrestore(&cohc->lock, flags);
 
-	return ret;
+	return left;
 }
-EXPORT_SYMBOL(coh901318_get_bytes_left);
-
 
-/* Stops a transfer without losing data. Enables power save.
-   Use this function in conjunction with coh901318_continue(..)
-*/
-void coh901318_stop(struct dma_chan *chan)
+/*
+ * Pauses a transfer without losing data. Enables power save.
+ * Use this function in conjunction with coh901318_resume.
+ */
+static void coh901318_pause(struct dma_chan *chan)
 {
 	u32 val;
 	unsigned long flags;
@@ -475,12 +549,11 @@ void coh901318_stop(struct dma_chan *chan)
 
 	spin_unlock_irqrestore(&cohc->lock, flags);
 }
-EXPORT_SYMBOL(coh901318_stop);
 
-/* Continues a transfer that has been stopped via 300_dma_stop(..).
+/* Resumes a transfer that has been stopped via 300_dma_stop(..).
    Power save is handled.
 */
-void coh901318_continue(struct dma_chan *chan)
+static void coh901318_resume(struct dma_chan *chan)
 {
 	u32 val;
 	unsigned long flags;
@@ -506,7 +579,6 @@ void coh901318_continue(struct dma_chan *chan)
 
 	spin_unlock_irqrestore(&cohc->lock, flags);
 }
-EXPORT_SYMBOL(coh901318_continue);
 
 bool coh901318_filter_id(struct dma_chan *chan, void *chan_id)
 {
@@ -565,29 +637,30 @@ static int coh901318_config(struct coh901318_chan *cohc,
  */
 static struct coh901318_desc *coh901318_queue_start(struct coh901318_chan *cohc)
 {
-	struct coh901318_desc *cohd_que;
+	struct coh901318_desc *cohd;
 
-	/* start queued jobs, if any
+	/*
+	 * start queued jobs, if any
 	 * TODO: transmit all queued jobs in one go
 	 */
-	cohd_que = coh901318_first_queued(cohc);
+	cohd = coh901318_first_queued(cohc);
 
-	if (cohd_que != NULL) {
+	if (cohd != NULL) {
 		/* Remove from queue */
-		coh901318_desc_remove(cohd_que);
+		coh901318_desc_remove(cohd);
 		/* initiate DMA job */
 		cohc->busy = 1;
 
-		coh901318_desc_submit(cohc, cohd_que);
+		coh901318_desc_submit(cohc, cohd);
 
-		coh901318_prep_linked_list(cohc, cohd_que->data);
+		coh901318_prep_linked_list(cohc, cohd->lli);
 
-		/* start dma job */
+		/* start dma job on this channel */
 		coh901318_start(cohc);
 
 	}
 
-	return cohd_que;
+	return cohd;
 }
 
 /*
@@ -622,7 +695,7 @@ static void dma_tasklet(unsigned long data)
 	cohc->completed = cohd_fin->desc.cookie;
 
 	/* release the lli allocation and remove the descriptor */
-	coh901318_lli_free(&cohc->base->pool, &cohd_fin->data);
+	coh901318_lli_free(&cohc->base->pool, &cohd_fin->lli);
 
 	/* return desc to free-list */
 	coh901318_desc_remove(cohd_fin);
@@ -666,23 +739,44 @@ static void dma_tasklet(unsigned long data)
 /* called from interrupt context */
 static void dma_tc_handle(struct coh901318_chan *cohc)
 {
-	BUG_ON(!cohc->allocated && (list_empty(&cohc->active) ||
-				    list_empty(&cohc->queue)));
-
-	if (!cohc->allocated)
+	/*
+	 * If the channel is not allocated, then we shouldn't have
+	 * any TC interrupts on it.
+	 */
+	if (!cohc->allocated) {
+		dev_err(COHC_2_DEV(cohc), "spurious interrupt from "
+			"unallocated channel\n");
 		return;
+	}
 
 	spin_lock(&cohc->lock);
 
+	/*
+	 * When we reach this point, at least one queue item
+	 * should have been moved over from cohc->queue to
+	 * cohc->active and run to completion, that is why we're
+	 * getting a terminal count interrupt is it not?
+	 * If you get this BUG() the most probable cause is that
+	 * the individual nodes in the lli chain have IRQ enabled,
+	 * so check your platform config for lli chain ctrl.
+	 */
+	BUG_ON(list_empty(&cohc->active));
+
 	cohc->nbr_active_done++;
 
+	/*
+	 * This attempt to take a job from cohc->queue, put it
+	 * into cohc->active and start it.
+	 */
 	if (coh901318_queue_start(cohc) == NULL)
 		cohc->busy = 0;
 
-	BUG_ON(list_empty(&cohc->active));
-
 	spin_unlock(&cohc->lock);
 
+	/*
+	 * This tasklet will remove items from cohc->active
+	 * and thus terminates them.
+	 */
 	if (cohc_chan_conf(cohc)->priority_high)
 		tasklet_hi_schedule(&cohc->tasklet);
 	else
@@ -809,6 +903,7 @@ static irqreturn_t dma_irq_handler(int irq, void *dev_id)
 static int coh901318_alloc_chan_resources(struct dma_chan *chan)
 {
 	struct coh901318_chan	*cohc = to_coh901318_chan(chan);
+	unsigned long flags;
 
 	dev_vdbg(COHC_2_DEV(cohc), "[%s] DMA channel %d\n",
 		 __func__, cohc->id);
@@ -816,11 +911,15 @@ static int coh901318_alloc_chan_resources(struct dma_chan *chan)
 	if (chan->client_count > 1)
 		return -EBUSY;
 
+	spin_lock_irqsave(&cohc->lock, flags);
+
 	coh901318_config(cohc, NULL);
 
 	cohc->allocated = 1;
 	cohc->completed = chan->cookie = 1;
 
+	spin_unlock_irqrestore(&cohc->lock, flags);
+
 	return 1;
 }
 
@@ -843,7 +942,7 @@ coh901318_free_chan_resources(struct dma_chan *chan)
 
 	spin_unlock_irqrestore(&cohc->lock, flags);
 
-	chan->device->device_terminate_all(chan);
+	chan->device->device_control(chan, DMA_TERMINATE_ALL, 0);
 }
 
 
@@ -870,7 +969,7 @@ static struct dma_async_tx_descriptor *
 coh901318_prep_memcpy(struct dma_chan *chan, dma_addr_t dest, dma_addr_t src,
 		      size_t size, unsigned long flags)
 {
-	struct coh901318_lli *data;
+	struct coh901318_lli *lli;
 	struct coh901318_desc *cohd;
 	unsigned long flg;
 	struct coh901318_chan *cohc = to_coh901318_chan(chan);
@@ -892,23 +991,23 @@ coh901318_prep_memcpy(struct dma_chan *chan, dma_addr_t dest, dma_addr_t src,
 	if ((lli_len << MAX_DMA_PACKET_SIZE_SHIFT) < size)
 		lli_len++;
 
-	data = coh901318_lli_alloc(&cohc->base->pool, lli_len);
+	lli = coh901318_lli_alloc(&cohc->base->pool, lli_len);
 
-	if (data == NULL)
+	if (lli == NULL)
 		goto err;
 
 	ret = coh901318_lli_fill_memcpy(
-		&cohc->base->pool, data, src, size, dest,
+		&cohc->base->pool, lli, src, size, dest,
 		cohc_chan_param(cohc)->ctrl_lli_chained,
 		ctrl_last);
 	if (ret)
 		goto err;
 
-	COH_DBG(coh901318_list_print(cohc, data));
+	COH_DBG(coh901318_list_print(cohc, lli));
 
 	/* Pick a descriptor to handle this transfer */
 	cohd = coh901318_desc_get(cohc);
-	cohd->data = data;
+	cohd->lli = lli;
 	cohd->flags = flags;
 	cohd->desc.tx_submit = coh901318_tx_submit;
 
@@ -926,7 +1025,7 @@ coh901318_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
 			unsigned long flags)
 {
 	struct coh901318_chan *cohc = to_coh901318_chan(chan);
-	struct coh901318_lli *data;
+	struct coh901318_lli *lli;
 	struct coh901318_desc *cohd;
 	const struct coh901318_params *params;
 	struct scatterlist *sg;
@@ -999,13 +1098,13 @@ coh901318_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
 	}
 
 	pr_debug("Allocate %d lli:s for this transfer\n", len);
-	data = coh901318_lli_alloc(&cohc->base->pool, len);
+	lli = coh901318_lli_alloc(&cohc->base->pool, len);
 
-	if (data == NULL)
+	if (lli == NULL)
 		goto err_dma_alloc;
 
-	/* initiate allocated data list */
-	ret = coh901318_lli_fill_sg(&cohc->base->pool, data, sgl, sg_len,
+	/* initiate allocated lli list */
+	ret = coh901318_lli_fill_sg(&cohc->base->pool, lli, sgl, sg_len,
 				    cohc_dev_addr(cohc),
 				    ctrl_chained,
 				    ctrl,
@@ -1014,14 +1113,14 @@ coh901318_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
 	if (ret)
 		goto err_lli_fill;
 
-	COH_DBG(coh901318_list_print(cohc, data));
+	COH_DBG(coh901318_list_print(cohc, lli));
 
 	/* Pick a descriptor to handle this transfer */
 	cohd = coh901318_desc_get(cohc);
 	cohd->dir = direction;
 	cohd->flags = flags;
 	cohd->desc.tx_submit = coh901318_tx_submit;
-	cohd->data = data;
+	cohd->lli = lli;
 
 	spin_unlock_irqrestore(&cohc->lock, flg);
 
@@ -1035,9 +1134,8 @@ coh901318_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
 }
 
 static enum dma_status
-coh901318_is_tx_complete(struct dma_chan *chan,
-			 dma_cookie_t cookie, dma_cookie_t *done,
-			 dma_cookie_t *used)
+coh901318_tx_status(struct dma_chan *chan, dma_cookie_t cookie,
+		 struct dma_tx_state *txstate)
 {
 	struct coh901318_chan *cohc = to_coh901318_chan(chan);
 	dma_cookie_t last_used;
@@ -1049,10 +1147,10 @@ coh901318_is_tx_complete(struct dma_chan *chan,
 
 	ret = dma_async_is_complete(cookie, last_complete, last_used);
 
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used,
+			 coh901318_get_bytes_left(chan));
+	if (ret == DMA_IN_PROGRESS && cohc->stopped)
+		ret = DMA_PAUSED;
 
 	return ret;
 }
@@ -1065,23 +1163,42 @@ coh901318_issue_pending(struct dma_chan *chan)
 
 	spin_lock_irqsave(&cohc->lock, flags);
 
-	/* Busy means that pending jobs are already being processed */
+	/*
+	 * Busy means that pending jobs are already being processed,
+	 * and then there is no point in starting the queue: the
+	 * terminal count interrupt on the channel will take the next
+	 * job on the queue and execute it anyway.
+	 */
 	if (!cohc->busy)
 		coh901318_queue_start(cohc);
 
 	spin_unlock_irqrestore(&cohc->lock, flags);
 }
 
-static void
-coh901318_terminate_all(struct dma_chan *chan)
+static int
+coh901318_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+		  unsigned long arg)
 {
 	unsigned long flags;
 	struct coh901318_chan *cohc = to_coh901318_chan(chan);
 	struct coh901318_desc *cohd;
 	void __iomem *virtbase = cohc->base->virtbase;
 
-	coh901318_stop(chan);
+	if (cmd == DMA_PAUSE) {
+		coh901318_pause(chan);
+		return 0;
+	}
+
+	if (cmd == DMA_RESUME) {
+		coh901318_resume(chan);
+		return 0;
+	}
 
+	if (cmd != DMA_TERMINATE_ALL)
+		return -ENXIO;
+
+	/* The remainder of this function terminates the transfer */
+	coh901318_pause(chan);
 	spin_lock_irqsave(&cohc->lock, flags);
 
 	/* Clear any pending BE or TC interrupt */
@@ -1099,7 +1216,7 @@ coh901318_terminate_all(struct dma_chan *chan)
 
 	while ((cohd = coh901318_first_active_get(cohc))) {
 		/* release the lli allocation*/
-		coh901318_lli_free(&cohc->base->pool, &cohd->data);
+		coh901318_lli_free(&cohc->base->pool, &cohd->lli);
 
 		/* return desc to free-list */
 		coh901318_desc_remove(cohd);
@@ -1108,7 +1225,7 @@ coh901318_terminate_all(struct dma_chan *chan)
 
 	while ((cohd = coh901318_first_queued(cohc))) {
 		/* release the lli allocation*/
-		coh901318_lli_free(&cohc->base->pool, &cohd->data);
+		coh901318_lli_free(&cohc->base->pool, &cohd->lli);
 
 		/* return desc to free-list */
 		coh901318_desc_remove(cohd);
@@ -1120,6 +1237,8 @@ coh901318_terminate_all(struct dma_chan *chan)
 	cohc->busy = 0;
 
 	spin_unlock_irqrestore(&cohc->lock, flags);
+
+	return 0;
 }
 void coh901318_base_init(struct dma_device *dma, const int *pick_chans,
 			 struct coh901318_base *base)
@@ -1235,9 +1354,9 @@ static int __init coh901318_probe(struct platform_device *pdev)
 	base->dma_slave.device_alloc_chan_resources = coh901318_alloc_chan_resources;
 	base->dma_slave.device_free_chan_resources = coh901318_free_chan_resources;
 	base->dma_slave.device_prep_slave_sg = coh901318_prep_slave_sg;
-	base->dma_slave.device_is_tx_complete = coh901318_is_tx_complete;
+	base->dma_slave.device_tx_status = coh901318_tx_status;
 	base->dma_slave.device_issue_pending = coh901318_issue_pending;
-	base->dma_slave.device_terminate_all = coh901318_terminate_all;
+	base->dma_slave.device_control = coh901318_control;
 	base->dma_slave.dev = &pdev->dev;
 
 	err = dma_async_device_register(&base->dma_slave);
@@ -1255,9 +1374,9 @@ static int __init coh901318_probe(struct platform_device *pdev)
 	base->dma_memcpy.device_alloc_chan_resources = coh901318_alloc_chan_resources;
 	base->dma_memcpy.device_free_chan_resources = coh901318_free_chan_resources;
 	base->dma_memcpy.device_prep_dma_memcpy = coh901318_prep_memcpy;
-	base->dma_memcpy.device_is_tx_complete = coh901318_is_tx_complete;
+	base->dma_memcpy.device_tx_status = coh901318_tx_status;
 	base->dma_memcpy.device_issue_pending = coh901318_issue_pending;
-	base->dma_memcpy.device_terminate_all = coh901318_terminate_all;
+	base->dma_memcpy.device_control = coh901318_control;
 	base->dma_memcpy.dev = &pdev->dev;
 	/*
 	 * This controller can only access address at even 32bit boundaries,
diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
index d18b5d069d7e..9d31d5eb95c1 100644
--- a/drivers/dma/dmaengine.c
+++ b/drivers/dma/dmaengine.c
@@ -515,7 +515,6 @@ struct dma_chan *__dma_request_channel(dma_cap_mask_t *mask, dma_filter_fn fn, v
 				break;
 			if (--device->privatecnt == 0)
 				dma_cap_clear(DMA_PRIVATE, device->cap_mask);
-			chan->private = NULL;
 			chan = NULL;
 		}
 	}
@@ -537,7 +536,6 @@ void dma_release_channel(struct dma_chan *chan)
 	/* drop PRIVATE cap enabled by __dma_request_channel() */
 	if (--chan->device->privatecnt == 0)
 		dma_cap_clear(DMA_PRIVATE, chan->device->cap_mask);
-	chan->private = NULL;
 	mutex_unlock(&dma_list_mutex);
 }
 EXPORT_SYMBOL_GPL(dma_release_channel);
@@ -695,11 +693,11 @@ int dma_async_device_register(struct dma_device *device)
 	BUG_ON(dma_has_cap(DMA_SLAVE, device->cap_mask) &&
 		!device->device_prep_slave_sg);
 	BUG_ON(dma_has_cap(DMA_SLAVE, device->cap_mask) &&
-		!device->device_terminate_all);
+		!device->device_control);
 
 	BUG_ON(!device->device_alloc_chan_resources);
 	BUG_ON(!device->device_free_chan_resources);
-	BUG_ON(!device->device_is_tx_complete);
+	BUG_ON(!device->device_tx_status);
 	BUG_ON(!device->device_issue_pending);
 	BUG_ON(!device->dev);
 
@@ -978,7 +976,9 @@ void dma_async_tx_descriptor_init(struct dma_async_tx_descriptor *tx,
 	struct dma_chan *chan)
 {
 	tx->chan = chan;
+	#ifndef CONFIG_ASYNC_TX_DISABLE_CHANNEL_SWITCH
 	spin_lock_init(&tx->lock);
+	#endif
 }
 EXPORT_SYMBOL(dma_async_tx_descriptor_init);
 
@@ -1011,7 +1011,7 @@ EXPORT_SYMBOL_GPL(dma_wait_for_async_tx);
  */
 void dma_run_dependencies(struct dma_async_tx_descriptor *tx)
 {
-	struct dma_async_tx_descriptor *dep = tx->next;
+	struct dma_async_tx_descriptor *dep = txd_next(tx);
 	struct dma_async_tx_descriptor *dep_next;
 	struct dma_chan *chan;
 
@@ -1019,7 +1019,7 @@ void dma_run_dependencies(struct dma_async_tx_descriptor *tx)
 		return;
 
 	/* we'll submit tx->next now, so clear the link */
-	tx->next = NULL;
+	txd_clear_next(tx);
 	chan = dep->chan;
 
 	/* keep submitting up until a channel switch is detected
@@ -1027,14 +1027,14 @@ void dma_run_dependencies(struct dma_async_tx_descriptor *tx)
 	 * processing the interrupt from async_tx_channel_switch
 	 */
 	for (; dep; dep = dep_next) {
-		spin_lock_bh(&dep->lock);
-		dep->parent = NULL;
-		dep_next = dep->next;
+		txd_lock(dep);
+		txd_clear_parent(dep);
+		dep_next = txd_next(dep);
 		if (dep_next && dep_next->chan == chan)
-			dep->next = NULL; /* ->next will be submitted */
+			txd_clear_next(dep); /* ->next will be submitted */
 		else
 			dep_next = NULL; /* submit current dep and terminate */
-		spin_unlock_bh(&dep->lock);
+		txd_unlock(dep);
 
 		dep->tx_submit(dep);
 	}
diff --git a/drivers/dma/dw_dmac.c b/drivers/dma/dw_dmac.c
index d28369f7afd2..a3991ab0d67e 100644
--- a/drivers/dma/dw_dmac.c
+++ b/drivers/dma/dw_dmac.c
@@ -781,13 +781,18 @@ err_desc_get:
 	return NULL;
 }
 
-static void dwc_terminate_all(struct dma_chan *chan)
+static int dwc_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+		       unsigned long arg)
 {
 	struct dw_dma_chan	*dwc = to_dw_dma_chan(chan);
 	struct dw_dma		*dw = to_dw_dma(chan->device);
 	struct dw_desc		*desc, *_desc;
 	LIST_HEAD(list);
 
+	/* Only supports DMA_TERMINATE_ALL */
+	if (cmd != DMA_TERMINATE_ALL)
+		return -ENXIO;
+
 	/*
 	 * This is only called when something went wrong elsewhere, so
 	 * we don't really care about the data. Just disable the
@@ -810,12 +815,14 @@ static void dwc_terminate_all(struct dma_chan *chan)
 	/* Flush all pending and queued descriptors */
 	list_for_each_entry_safe(desc, _desc, &list, desc_node)
 		dwc_descriptor_complete(dwc, desc);
+
+	return 0;
 }
 
 static enum dma_status
-dwc_is_tx_complete(struct dma_chan *chan,
-		dma_cookie_t cookie,
-		dma_cookie_t *done, dma_cookie_t *used)
+dwc_tx_status(struct dma_chan *chan,
+	      dma_cookie_t cookie,
+	      struct dma_tx_state *txstate)
 {
 	struct dw_dma_chan	*dwc = to_dw_dma_chan(chan);
 	dma_cookie_t		last_used;
@@ -835,10 +842,7 @@ dwc_is_tx_complete(struct dma_chan *chan,
 		ret = dma_async_is_complete(cookie, last_complete, last_used);
 	}
 
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 
 	return ret;
 }
@@ -1338,9 +1342,9 @@ static int __init dw_probe(struct platform_device *pdev)
 	dw->dma.device_prep_dma_memcpy = dwc_prep_dma_memcpy;
 
 	dw->dma.device_prep_slave_sg = dwc_prep_slave_sg;
-	dw->dma.device_terminate_all = dwc_terminate_all;
+	dw->dma.device_control = dwc_control;
 
-	dw->dma.device_is_tx_complete = dwc_is_tx_complete;
+	dw->dma.device_tx_status = dwc_tx_status;
 	dw->dma.device_issue_pending = dwc_issue_pending;
 
 	dma_writel(dw, CFG, DW_CFG_DMA_EN);
diff --git a/drivers/dma/fsldma.c b/drivers/dma/fsldma.c
index 88f470f0d820..8088b14ba5f7 100644
--- a/drivers/dma/fsldma.c
+++ b/drivers/dma/fsldma.c
@@ -775,13 +775,18 @@ fail:
 	return NULL;
 }
 
-static void fsl_dma_device_terminate_all(struct dma_chan *dchan)
+static int fsl_dma_device_control(struct dma_chan *dchan,
+				  enum dma_ctrl_cmd cmd, unsigned long arg)
 {
 	struct fsldma_chan *chan;
 	unsigned long flags;
 
+	/* Only supports DMA_TERMINATE_ALL */
+	if (cmd != DMA_TERMINATE_ALL)
+		return -ENXIO;
+
 	if (!dchan)
-		return;
+		return -EINVAL;
 
 	chan = to_fsl_chan(dchan);
 
@@ -795,6 +800,8 @@ static void fsl_dma_device_terminate_all(struct dma_chan *dchan)
 	fsldma_free_desc_list(chan, &chan->ld_running);
 
 	spin_unlock_irqrestore(&chan->desc_lock, flags);
+
+	return 0;
 }
 
 /**
@@ -965,13 +972,12 @@ static void fsl_dma_memcpy_issue_pending(struct dma_chan *dchan)
 }
 
 /**
- * fsl_dma_is_complete - Determine the DMA status
+ * fsl_tx_status - Determine the DMA status
  * @chan : Freescale DMA channel
  */
-static enum dma_status fsl_dma_is_complete(struct dma_chan *dchan,
+static enum dma_status fsl_tx_status(struct dma_chan *dchan,
 					dma_cookie_t cookie,
-					dma_cookie_t *done,
-					dma_cookie_t *used)
+					struct dma_tx_state *txstate)
 {
 	struct fsldma_chan *chan = to_fsl_chan(dchan);
 	dma_cookie_t last_used;
@@ -982,11 +988,7 @@ static enum dma_status fsl_dma_is_complete(struct dma_chan *dchan,
 	last_used = dchan->cookie;
 	last_complete = chan->completed_cookie;
 
-	if (done)
-		*done = last_complete;
-
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 
 	return dma_async_is_complete(cookie, last_complete, last_used);
 }
@@ -1313,7 +1315,7 @@ static int __devinit fsldma_of_probe(struct of_device *op,
 	INIT_LIST_HEAD(&fdev->common.channels);
 
 	/* ioremap the registers for use */
-	fdev->regs = of_iomap(op->node, 0);
+	fdev->regs = of_iomap(op->dev.of_node, 0);
 	if (!fdev->regs) {
 		dev_err(&op->dev, "unable to ioremap registers\n");
 		err = -ENOMEM;
@@ -1321,7 +1323,7 @@ static int __devinit fsldma_of_probe(struct of_device *op,
 	}
 
 	/* map the channel IRQ if it exists, but don't hookup the handler yet */
-	fdev->irq = irq_of_parse_and_map(op->node, 0);
+	fdev->irq = irq_of_parse_and_map(op->dev.of_node, 0);
 
 	dma_cap_set(DMA_MEMCPY, fdev->common.cap_mask);
 	dma_cap_set(DMA_INTERRUPT, fdev->common.cap_mask);
@@ -1330,10 +1332,10 @@ static int __devinit fsldma_of_probe(struct of_device *op,
 	fdev->common.device_free_chan_resources = fsl_dma_free_chan_resources;
 	fdev->common.device_prep_dma_interrupt = fsl_dma_prep_interrupt;
 	fdev->common.device_prep_dma_memcpy = fsl_dma_prep_memcpy;
-	fdev->common.device_is_tx_complete = fsl_dma_is_complete;
+	fdev->common.device_tx_status = fsl_tx_status;
 	fdev->common.device_issue_pending = fsl_dma_memcpy_issue_pending;
 	fdev->common.device_prep_slave_sg = fsl_dma_prep_slave_sg;
-	fdev->common.device_terminate_all = fsl_dma_device_terminate_all;
+	fdev->common.device_control = fsl_dma_device_control;
 	fdev->common.dev = &op->dev;
 
 	dev_set_drvdata(&op->dev, fdev);
@@ -1343,7 +1345,7 @@ static int __devinit fsldma_of_probe(struct of_device *op,
 	 * of_platform_bus_remove(). Instead, we manually instantiate every DMA
 	 * channel object.
 	 */
-	for_each_child_of_node(op->node, child) {
+	for_each_child_of_node(op->dev.of_node, child) {
 		if (of_device_is_compatible(child, "fsl,eloplus-dma-channel")) {
 			fsl_dma_chan_probe(fdev, child,
 				FSL_DMA_IP_85XX | FSL_DMA_BIG_ENDIAN,
@@ -1409,10 +1411,13 @@ static const struct of_device_id fsldma_of_ids[] = {
 };
 
 static struct of_platform_driver fsldma_of_driver = {
-	.name		= "fsl-elo-dma",
-	.match_table	= fsldma_of_ids,
-	.probe		= fsldma_of_probe,
-	.remove		= fsldma_of_remove,
+	.driver = {
+		.name = "fsl-elo-dma",
+		.owner = THIS_MODULE,
+		.of_match_table = fsldma_of_ids,
+	},
+	.probe = fsldma_of_probe,
+	.remove = fsldma_of_remove,
 };
 
 /*----------------------------------------------------------------------------*/
diff --git a/drivers/dma/ioat/dma.c b/drivers/dma/ioat/dma.c
index 3e5a8005c62b..c9213ead4a26 100644
--- a/drivers/dma/ioat/dma.c
+++ b/drivers/dma/ioat/dma.c
@@ -727,18 +727,18 @@ static void ioat1_timer_event(unsigned long data)
 }
 
 enum dma_status
-ioat_is_dma_complete(struct dma_chan *c, dma_cookie_t cookie,
-		      dma_cookie_t *done, dma_cookie_t *used)
+ioat_dma_tx_status(struct dma_chan *c, dma_cookie_t cookie,
+		   struct dma_tx_state *txstate)
 {
 	struct ioat_chan_common *chan = to_chan_common(c);
 	struct ioatdma_device *device = chan->device;
 
-	if (ioat_is_complete(c, cookie, done, used) == DMA_SUCCESS)
+	if (ioat_tx_status(c, cookie, txstate) == DMA_SUCCESS)
 		return DMA_SUCCESS;
 
 	device->cleanup_fn((unsigned long) c);
 
-	return ioat_is_complete(c, cookie, done, used);
+	return ioat_tx_status(c, cookie, txstate);
 }
 
 static void ioat1_dma_start_null_desc(struct ioat_dma_chan *ioat)
@@ -858,7 +858,7 @@ int __devinit ioat_dma_self_test(struct ioatdma_device *device)
 	tmo = wait_for_completion_timeout(&cmp, msecs_to_jiffies(3000));
 
 	if (tmo == 0 ||
-	    dma->device_is_tx_complete(dma_chan, cookie, NULL, NULL)
+	    dma->device_tx_status(dma_chan, cookie, NULL)
 					!= DMA_SUCCESS) {
 		dev_err(dev, "Self-test copy timed out, disabling\n");
 		err = -ENODEV;
@@ -1199,7 +1199,7 @@ int __devinit ioat1_dma_probe(struct ioatdma_device *device, int dca)
 	dma->device_issue_pending = ioat1_dma_memcpy_issue_pending;
 	dma->device_alloc_chan_resources = ioat1_dma_alloc_chan_resources;
 	dma->device_free_chan_resources = ioat1_dma_free_chan_resources;
-	dma->device_is_tx_complete = ioat_is_dma_complete;
+	dma->device_tx_status = ioat_dma_tx_status;
 
 	err = ioat_probe(device);
 	if (err)
diff --git a/drivers/dma/ioat/dma.h b/drivers/dma/ioat/dma.h
index 86b97ac8774e..6d3a73b57e54 100644
--- a/drivers/dma/ioat/dma.h
+++ b/drivers/dma/ioat/dma.h
@@ -96,6 +96,7 @@ struct ioat_chan_common {
 	#define IOAT_COMPLETION_ACK 1
 	#define IOAT_RESET_PENDING 2
 	#define IOAT_KOBJ_INIT_FAIL 3
+	#define IOAT_RESHAPE_PENDING 4
 	struct timer_list timer;
 	#define COMPLETION_TIMEOUT msecs_to_jiffies(100)
 	#define IDLE_TIMEOUT msecs_to_jiffies(2000)
@@ -142,15 +143,14 @@ static inline struct ioat_dma_chan *to_ioat_chan(struct dma_chan *c)
 }
 
 /**
- * ioat_is_complete - poll the status of an ioat transaction
+ * ioat_tx_status - poll the status of an ioat transaction
  * @c: channel handle
  * @cookie: transaction identifier
- * @done: if set, updated with last completed transaction
- * @used: if set, updated with last used transaction
+ * @txstate: if set, updated with the transaction state
  */
 static inline enum dma_status
-ioat_is_complete(struct dma_chan *c, dma_cookie_t cookie,
-		 dma_cookie_t *done, dma_cookie_t *used)
+ioat_tx_status(struct dma_chan *c, dma_cookie_t cookie,
+		 struct dma_tx_state *txstate)
 {
 	struct ioat_chan_common *chan = to_chan_common(c);
 	dma_cookie_t last_used;
@@ -159,10 +159,7 @@ ioat_is_complete(struct dma_chan *c, dma_cookie_t cookie,
 	last_used = c->cookie;
 	last_complete = chan->completed_cookie;
 
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 
 	return dma_async_is_complete(cookie, last_complete, last_used);
 }
@@ -338,8 +335,8 @@ struct dca_provider * __devinit ioat_dca_init(struct pci_dev *pdev,
 unsigned long ioat_get_current_completion(struct ioat_chan_common *chan);
 void ioat_init_channel(struct ioatdma_device *device,
 		       struct ioat_chan_common *chan, int idx);
-enum dma_status ioat_is_dma_complete(struct dma_chan *c, dma_cookie_t cookie,
-				     dma_cookie_t *done, dma_cookie_t *used);
+enum dma_status ioat_dma_tx_status(struct dma_chan *c, dma_cookie_t cookie,
+				   struct dma_tx_state *txstate);
 void ioat_dma_unmap(struct ioat_chan_common *chan, enum dma_ctrl_flags flags,
 		    size_t len, struct ioat_dma_descriptor *hw);
 bool ioat_cleanup_preamble(struct ioat_chan_common *chan,
diff --git a/drivers/dma/ioat/dma_v2.c b/drivers/dma/ioat/dma_v2.c
index b5ae56c211e6..3c8b32a83794 100644
--- a/drivers/dma/ioat/dma_v2.c
+++ b/drivers/dma/ioat/dma_v2.c
@@ -56,8 +56,6 @@ void __ioat2_issue_pending(struct ioat2_dma_chan *ioat)
 
 	ioat->dmacount += ioat2_ring_pending(ioat);
 	ioat->issued = ioat->head;
-	/* make descriptor updates globally visible before notifying channel */
-	wmb();
 	writew(ioat->dmacount, chan->reg_base + IOAT_CHAN_DMACOUNT_OFFSET);
 	dev_dbg(to_dev(chan),
 		"%s: head: %#x tail: %#x issued: %#x count: %#x\n",
@@ -69,9 +67,9 @@ void ioat2_issue_pending(struct dma_chan *c)
 	struct ioat2_dma_chan *ioat = to_ioat2_chan(c);
 
 	if (ioat2_ring_pending(ioat)) {
-		spin_lock_bh(&ioat->ring_lock);
+		spin_lock_bh(&ioat->prep_lock);
 		__ioat2_issue_pending(ioat);
-		spin_unlock_bh(&ioat->ring_lock);
+		spin_unlock_bh(&ioat->prep_lock);
 	}
 }
 
@@ -80,7 +78,7 @@ void ioat2_issue_pending(struct dma_chan *c)
  * @ioat: ioat2+ channel
  *
  * Check if the number of unsubmitted descriptors has exceeded the
- * watermark.  Called with ring_lock held
+ * watermark.  Called with prep_lock held
  */
 static void ioat2_update_pending(struct ioat2_dma_chan *ioat)
 {
@@ -92,7 +90,6 @@ static void __ioat2_start_null_desc(struct ioat2_dma_chan *ioat)
 {
 	struct ioat_ring_ent *desc;
 	struct ioat_dma_descriptor *hw;
-	int idx;
 
 	if (ioat2_ring_space(ioat) < 1) {
 		dev_err(to_dev(&ioat->base),
@@ -102,8 +99,7 @@ static void __ioat2_start_null_desc(struct ioat2_dma_chan *ioat)
 
 	dev_dbg(to_dev(&ioat->base), "%s: head: %#x tail: %#x issued: %#x\n",
 		__func__, ioat->head, ioat->tail, ioat->issued);
-	idx = ioat2_desc_alloc(ioat, 1);
-	desc = ioat2_get_ring_ent(ioat, idx);
+	desc = ioat2_get_ring_ent(ioat, ioat->head);
 
 	hw = desc->hw;
 	hw->ctl = 0;
@@ -117,14 +113,16 @@ static void __ioat2_start_null_desc(struct ioat2_dma_chan *ioat)
 	async_tx_ack(&desc->txd);
 	ioat2_set_chainaddr(ioat, desc->txd.phys);
 	dump_desc_dbg(ioat, desc);
+	wmb();
+	ioat->head += 1;
 	__ioat2_issue_pending(ioat);
 }
 
 static void ioat2_start_null_desc(struct ioat2_dma_chan *ioat)
 {
-	spin_lock_bh(&ioat->ring_lock);
+	spin_lock_bh(&ioat->prep_lock);
 	__ioat2_start_null_desc(ioat);
-	spin_unlock_bh(&ioat->ring_lock);
+	spin_unlock_bh(&ioat->prep_lock);
 }
 
 static void __cleanup(struct ioat2_dma_chan *ioat, unsigned long phys_complete)
@@ -134,15 +132,16 @@ static void __cleanup(struct ioat2_dma_chan *ioat, unsigned long phys_complete)
 	struct ioat_ring_ent *desc;
 	bool seen_current = false;
 	u16 active;
-	int i;
+	int idx = ioat->tail, i;
 
 	dev_dbg(to_dev(chan), "%s: head: %#x tail: %#x issued: %#x\n",
 		__func__, ioat->head, ioat->tail, ioat->issued);
 
 	active = ioat2_ring_active(ioat);
 	for (i = 0; i < active && !seen_current; i++) {
-		prefetch(ioat2_get_ring_ent(ioat, ioat->tail + i + 1));
-		desc = ioat2_get_ring_ent(ioat, ioat->tail + i);
+		smp_read_barrier_depends();
+		prefetch(ioat2_get_ring_ent(ioat, idx + i + 1));
+		desc = ioat2_get_ring_ent(ioat, idx + i);
 		tx = &desc->txd;
 		dump_desc_dbg(ioat, desc);
 		if (tx->cookie) {
@@ -158,11 +157,12 @@ static void __cleanup(struct ioat2_dma_chan *ioat, unsigned long phys_complete)
 		if (tx->phys == phys_complete)
 			seen_current = true;
 	}
-	ioat->tail += i;
+	smp_mb(); /* finish all descriptor reads before incrementing tail */
+	ioat->tail = idx + i;
 	BUG_ON(active && !seen_current); /* no active descs have written a completion? */
 
 	chan->last_completion = phys_complete;
-	if (ioat->head == ioat->tail) {
+	if (active - i == 0) {
 		dev_dbg(to_dev(chan), "%s: cancel completion timeout\n",
 			__func__);
 		clear_bit(IOAT_COMPLETION_PENDING, &chan->state);
@@ -179,24 +179,9 @@ static void ioat2_cleanup(struct ioat2_dma_chan *ioat)
 	struct ioat_chan_common *chan = &ioat->base;
 	unsigned long phys_complete;
 
-	prefetch(chan->completion);
-
-	if (!spin_trylock_bh(&chan->cleanup_lock))
-		return;
-
-	if (!ioat_cleanup_preamble(chan, &phys_complete)) {
-		spin_unlock_bh(&chan->cleanup_lock);
-		return;
-	}
-
-	if (!spin_trylock_bh(&ioat->ring_lock)) {
-		spin_unlock_bh(&chan->cleanup_lock);
-		return;
-	}
-
-	__cleanup(ioat, phys_complete);
-
-	spin_unlock_bh(&ioat->ring_lock);
+	spin_lock_bh(&chan->cleanup_lock);
+	if (ioat_cleanup_preamble(chan, &phys_complete))
+		__cleanup(ioat, phys_complete);
 	spin_unlock_bh(&chan->cleanup_lock);
 }
 
@@ -287,12 +272,10 @@ void ioat2_timer_event(unsigned long data)
 	struct ioat2_dma_chan *ioat = to_ioat2_chan((void *) data);
 	struct ioat_chan_common *chan = &ioat->base;
 
-	spin_lock_bh(&chan->cleanup_lock);
 	if (test_bit(IOAT_COMPLETION_PENDING, &chan->state)) {
 		unsigned long phys_complete;
 		u64 status;
 
-		spin_lock_bh(&ioat->ring_lock);
 		status = ioat_chansts(chan);
 
 		/* when halted due to errors check for channel
@@ -311,26 +294,31 @@ void ioat2_timer_event(unsigned long data)
 		 * acknowledged a pending completion once, then be more
 		 * forceful with a restart
 		 */
-		if (ioat_cleanup_preamble(chan, &phys_complete))
+		spin_lock_bh(&chan->cleanup_lock);
+		if (ioat_cleanup_preamble(chan, &phys_complete)) {
 			__cleanup(ioat, phys_complete);
-		else if (test_bit(IOAT_COMPLETION_ACK, &chan->state))
+		} else if (test_bit(IOAT_COMPLETION_ACK, &chan->state)) {
+			spin_lock_bh(&ioat->prep_lock);
 			ioat2_restart_channel(ioat);
-		else {
+			spin_unlock_bh(&ioat->prep_lock);
+		} else {
 			set_bit(IOAT_COMPLETION_ACK, &chan->state);
 			mod_timer(&chan->timer, jiffies + COMPLETION_TIMEOUT);
 		}
-		spin_unlock_bh(&ioat->ring_lock);
+		spin_unlock_bh(&chan->cleanup_lock);
 	} else {
 		u16 active;
 
 		/* if the ring is idle, empty, and oversized try to step
 		 * down the size
 		 */
-		spin_lock_bh(&ioat->ring_lock);
+		spin_lock_bh(&chan->cleanup_lock);
+		spin_lock_bh(&ioat->prep_lock);
 		active = ioat2_ring_active(ioat);
 		if (active == 0 && ioat->alloc_order > ioat_get_alloc_order())
 			reshape_ring(ioat, ioat->alloc_order-1);
-		spin_unlock_bh(&ioat->ring_lock);
+		spin_unlock_bh(&ioat->prep_lock);
+		spin_unlock_bh(&chan->cleanup_lock);
 
 		/* keep shrinking until we get back to our minimum
 		 * default size
@@ -338,7 +326,6 @@ void ioat2_timer_event(unsigned long data)
 		if (ioat->alloc_order > ioat_get_alloc_order())
 			mod_timer(&chan->timer, jiffies + IDLE_TIMEOUT);
 	}
-	spin_unlock_bh(&chan->cleanup_lock);
 }
 
 static int ioat2_reset_hw(struct ioat_chan_common *chan)
@@ -392,7 +379,7 @@ int ioat2_enumerate_channels(struct ioatdma_device *device)
 
 		ioat_init_channel(device, &ioat->base, i);
 		ioat->xfercap_log = xfercap_log;
-		spin_lock_init(&ioat->ring_lock);
+		spin_lock_init(&ioat->prep_lock);
 		if (device->reset_hw(&ioat->base)) {
 			i = 0;
 			break;
@@ -418,8 +405,17 @@ static dma_cookie_t ioat2_tx_submit_unlock(struct dma_async_tx_descriptor *tx)
 
 	if (!test_and_set_bit(IOAT_COMPLETION_PENDING, &chan->state))
 		mod_timer(&chan->timer, jiffies + COMPLETION_TIMEOUT);
+
+	/* make descriptor updates visible before advancing ioat->head,
+	 * this is purposefully not smp_wmb() since we are also
+	 * publishing the descriptor updates to a dma device
+	 */
+	wmb();
+
+	ioat->head += ioat->produce;
+
 	ioat2_update_pending(ioat);
-	spin_unlock_bh(&ioat->ring_lock);
+	spin_unlock_bh(&ioat->prep_lock);
 
 	return cookie;
 }
@@ -531,13 +527,15 @@ int ioat2_alloc_chan_resources(struct dma_chan *c)
 	if (!ring)
 		return -ENOMEM;
 
-	spin_lock_bh(&ioat->ring_lock);
+	spin_lock_bh(&chan->cleanup_lock);
+	spin_lock_bh(&ioat->prep_lock);
 	ioat->ring = ring;
 	ioat->head = 0;
 	ioat->issued = 0;
 	ioat->tail = 0;
 	ioat->alloc_order = order;
-	spin_unlock_bh(&ioat->ring_lock);
+	spin_unlock_bh(&ioat->prep_lock);
+	spin_unlock_bh(&chan->cleanup_lock);
 
 	tasklet_enable(&chan->cleanup_task);
 	ioat2_start_null_desc(ioat);
@@ -553,7 +551,7 @@ bool reshape_ring(struct ioat2_dma_chan *ioat, int order)
 	 */
 	struct ioat_chan_common *chan = &ioat->base;
 	struct dma_chan *c = &chan->common;
-	const u16 curr_size = ioat2_ring_mask(ioat) + 1;
+	const u16 curr_size = ioat2_ring_size(ioat);
 	const u16 active = ioat2_ring_active(ioat);
 	const u16 new_size = 1 << order;
 	struct ioat_ring_ent **ring;
@@ -653,54 +651,61 @@ bool reshape_ring(struct ioat2_dma_chan *ioat, int order)
 }
 
 /**
- * ioat2_alloc_and_lock - common descriptor alloc boilerplate for ioat2,3 ops
- * @idx: gets starting descriptor index on successful allocation
+ * ioat2_check_space_lock - verify space and grab ring producer lock
  * @ioat: ioat2,3 channel (ring) to operate on
  * @num_descs: allocation length
  */
-int ioat2_alloc_and_lock(u16 *idx, struct ioat2_dma_chan *ioat, int num_descs)
+int ioat2_check_space_lock(struct ioat2_dma_chan *ioat, int num_descs)
 {
 	struct ioat_chan_common *chan = &ioat->base;
+	bool retry;
 
-	spin_lock_bh(&ioat->ring_lock);
+ retry:
+	spin_lock_bh(&ioat->prep_lock);
 	/* never allow the last descriptor to be consumed, we need at
 	 * least one free at all times to allow for on-the-fly ring
 	 * resizing.
 	 */
-	while (unlikely(ioat2_ring_space(ioat) <= num_descs)) {
-		if (reshape_ring(ioat, ioat->alloc_order + 1) &&
-		    ioat2_ring_space(ioat) > num_descs)
-				break;
-
-		if (printk_ratelimit())
-			dev_dbg(to_dev(chan),
-				"%s: ring full! num_descs: %d (%x:%x:%x)\n",
-				__func__, num_descs, ioat->head, ioat->tail,
-				ioat->issued);
-		spin_unlock_bh(&ioat->ring_lock);
-
-		/* progress reclaim in the allocation failure case we
-		 * may be called under bh_disabled so we need to trigger
-		 * the timer event directly
-		 */
-		spin_lock_bh(&chan->cleanup_lock);
-		if (jiffies > chan->timer.expires &&
-		    timer_pending(&chan->timer)) {
-			struct ioatdma_device *device = chan->device;
-
-			mod_timer(&chan->timer, jiffies + COMPLETION_TIMEOUT);
-			spin_unlock_bh(&chan->cleanup_lock);
-			device->timer_fn((unsigned long) &chan->common);
-		} else
-			spin_unlock_bh(&chan->cleanup_lock);
-		return -ENOMEM;
+	if (likely(ioat2_ring_space(ioat) > num_descs)) {
+		dev_dbg(to_dev(chan), "%s: num_descs: %d (%x:%x:%x)\n",
+			__func__, num_descs, ioat->head, ioat->tail, ioat->issued);
+		ioat->produce = num_descs;
+		return 0;  /* with ioat->prep_lock held */
 	}
+	retry = test_and_set_bit(IOAT_RESHAPE_PENDING, &chan->state);
+	spin_unlock_bh(&ioat->prep_lock);
 
-	dev_dbg(to_dev(chan), "%s: num_descs: %d (%x:%x:%x)\n",
-		__func__, num_descs, ioat->head, ioat->tail, ioat->issued);
+	/* is another cpu already trying to expand the ring? */
+	if (retry)
+		goto retry;
 
-	*idx = ioat2_desc_alloc(ioat, num_descs);
-	return 0;  /* with ioat->ring_lock held */
+	spin_lock_bh(&chan->cleanup_lock);
+	spin_lock_bh(&ioat->prep_lock);
+	retry = reshape_ring(ioat, ioat->alloc_order + 1);
+	clear_bit(IOAT_RESHAPE_PENDING, &chan->state);
+	spin_unlock_bh(&ioat->prep_lock);
+	spin_unlock_bh(&chan->cleanup_lock);
+
+	/* if we were able to expand the ring retry the allocation */
+	if (retry)
+		goto retry;
+
+	if (printk_ratelimit())
+		dev_dbg(to_dev(chan), "%s: ring full! num_descs: %d (%x:%x:%x)\n",
+			__func__, num_descs, ioat->head, ioat->tail, ioat->issued);
+
+	/* progress reclaim in the allocation failure case we may be
+	 * called under bh_disabled so we need to trigger the timer
+	 * event directly
+	 */
+	if (jiffies > chan->timer.expires && timer_pending(&chan->timer)) {
+		struct ioatdma_device *device = chan->device;
+
+		mod_timer(&chan->timer, jiffies + COMPLETION_TIMEOUT);
+		device->timer_fn((unsigned long) &chan->common);
+	}
+
+	return -ENOMEM;
 }
 
 struct dma_async_tx_descriptor *
@@ -713,14 +718,11 @@ ioat2_dma_prep_memcpy_lock(struct dma_chan *c, dma_addr_t dma_dest,
 	dma_addr_t dst = dma_dest;
 	dma_addr_t src = dma_src;
 	size_t total_len = len;
-	int num_descs;
-	u16 idx;
-	int i;
+	int num_descs, idx, i;
 
 	num_descs = ioat2_xferlen_to_descs(ioat, len);
-	if (likely(num_descs) &&
-	    ioat2_alloc_and_lock(&idx, ioat, num_descs) == 0)
-		/* pass */;
+	if (likely(num_descs) && ioat2_check_space_lock(ioat, num_descs) == 0)
+		idx = ioat->head;
 	else
 		return NULL;
 	i = 0;
@@ -777,7 +779,8 @@ void ioat2_free_chan_resources(struct dma_chan *c)
 	device->cleanup_fn((unsigned long) c);
 	device->reset_hw(chan);
 
-	spin_lock_bh(&ioat->ring_lock);
+	spin_lock_bh(&chan->cleanup_lock);
+	spin_lock_bh(&ioat->prep_lock);
 	descs = ioat2_ring_space(ioat);
 	dev_dbg(to_dev(chan), "freeing %d idle descriptors\n", descs);
 	for (i = 0; i < descs; i++) {
@@ -800,7 +803,8 @@ void ioat2_free_chan_resources(struct dma_chan *c)
 	ioat->alloc_order = 0;
 	pci_pool_free(device->completion_pool, chan->completion,
 		      chan->completion_dma);
-	spin_unlock_bh(&ioat->ring_lock);
+	spin_unlock_bh(&ioat->prep_lock);
+	spin_unlock_bh(&chan->cleanup_lock);
 
 	chan->last_completion = 0;
 	chan->completion_dma = 0;
@@ -855,7 +859,7 @@ int __devinit ioat2_dma_probe(struct ioatdma_device *device, int dca)
 	dma->device_issue_pending = ioat2_issue_pending;
 	dma->device_alloc_chan_resources = ioat2_alloc_chan_resources;
 	dma->device_free_chan_resources = ioat2_free_chan_resources;
-	dma->device_is_tx_complete = ioat_is_dma_complete;
+	dma->device_tx_status = ioat_tx_status;
 
 	err = ioat_probe(device);
 	if (err)
diff --git a/drivers/dma/ioat/dma_v2.h b/drivers/dma/ioat/dma_v2.h
index ef2871fd7868..a2c413b2b8d8 100644
--- a/drivers/dma/ioat/dma_v2.h
+++ b/drivers/dma/ioat/dma_v2.h
@@ -22,6 +22,7 @@
 #define IOATDMA_V2_H
 
 #include <linux/dmaengine.h>
+#include <linux/circ_buf.h>
 #include "dma.h"
 #include "hw.h"
 
@@ -49,8 +50,9 @@ extern int ioat_ring_alloc_order;
  * @tail: cleanup index
  * @dmacount: identical to 'head' except for occasionally resetting to zero
  * @alloc_order: log2 of the number of allocated descriptors
+ * @produce: number of descriptors to produce at submit time
  * @ring: software ring buffer implementation of hardware ring
- * @ring_lock: protects ring attributes
+ * @prep_lock: serializes descriptor preparation (producers)
  */
 struct ioat2_dma_chan {
 	struct ioat_chan_common base;
@@ -60,8 +62,9 @@ struct ioat2_dma_chan {
 	u16 tail;
 	u16 dmacount;
 	u16 alloc_order;
+	u16 produce;
 	struct ioat_ring_ent **ring;
-	spinlock_t ring_lock;
+	spinlock_t prep_lock;
 };
 
 static inline struct ioat2_dma_chan *to_ioat2_chan(struct dma_chan *c)
@@ -71,38 +74,26 @@ static inline struct ioat2_dma_chan *to_ioat2_chan(struct dma_chan *c)
 	return container_of(chan, struct ioat2_dma_chan, base);
 }
 
-static inline u16 ioat2_ring_mask(struct ioat2_dma_chan *ioat)
+static inline u16 ioat2_ring_size(struct ioat2_dma_chan *ioat)
 {
-	return (1 << ioat->alloc_order) - 1;
+	return 1 << ioat->alloc_order;
 }
 
 /* count of descriptors in flight with the engine */
 static inline u16 ioat2_ring_active(struct ioat2_dma_chan *ioat)
 {
-	return (ioat->head - ioat->tail) & ioat2_ring_mask(ioat);
+	return CIRC_CNT(ioat->head, ioat->tail, ioat2_ring_size(ioat));
 }
 
 /* count of descriptors pending submission to hardware */
 static inline u16 ioat2_ring_pending(struct ioat2_dma_chan *ioat)
 {
-	return (ioat->head - ioat->issued) & ioat2_ring_mask(ioat);
+	return CIRC_CNT(ioat->head, ioat->issued, ioat2_ring_size(ioat));
 }
 
 static inline u16 ioat2_ring_space(struct ioat2_dma_chan *ioat)
 {
-	u16 num_descs = ioat2_ring_mask(ioat) + 1;
-	u16 active = ioat2_ring_active(ioat);
-
-	BUG_ON(active > num_descs);
-
-	return num_descs - active;
-}
-
-/* assumes caller already checked space */
-static inline u16 ioat2_desc_alloc(struct ioat2_dma_chan *ioat, u16 len)
-{
-	ioat->head += len;
-	return ioat->head - len;
+	return ioat2_ring_size(ioat) - ioat2_ring_active(ioat);
 }
 
 static inline u16 ioat2_xferlen_to_descs(struct ioat2_dma_chan *ioat, size_t len)
@@ -151,7 +142,7 @@ struct ioat_ring_ent {
 static inline struct ioat_ring_ent *
 ioat2_get_ring_ent(struct ioat2_dma_chan *ioat, u16 idx)
 {
-	return ioat->ring[idx & ioat2_ring_mask(ioat)];
+	return ioat->ring[idx & (ioat2_ring_size(ioat) - 1)];
 }
 
 static inline void ioat2_set_chainaddr(struct ioat2_dma_chan *ioat, u64 addr)
@@ -168,7 +159,7 @@ int __devinit ioat2_dma_probe(struct ioatdma_device *dev, int dca);
 int __devinit ioat3_dma_probe(struct ioatdma_device *dev, int dca);
 struct dca_provider * __devinit ioat2_dca_init(struct pci_dev *pdev, void __iomem *iobase);
 struct dca_provider * __devinit ioat3_dca_init(struct pci_dev *pdev, void __iomem *iobase);
-int ioat2_alloc_and_lock(u16 *idx, struct ioat2_dma_chan *ioat, int num_descs);
+int ioat2_check_space_lock(struct ioat2_dma_chan *ioat, int num_descs);
 int ioat2_enumerate_channels(struct ioatdma_device *device);
 struct dma_async_tx_descriptor *
 ioat2_dma_prep_memcpy_lock(struct dma_chan *c, dma_addr_t dma_dest,
diff --git a/drivers/dma/ioat/dma_v3.c b/drivers/dma/ioat/dma_v3.c
index 6740e319c9cf..1cdd22e1051b 100644
--- a/drivers/dma/ioat/dma_v3.c
+++ b/drivers/dma/ioat/dma_v3.c
@@ -260,8 +260,8 @@ static void __cleanup(struct ioat2_dma_chan *ioat, unsigned long phys_complete)
 	struct ioat_chan_common *chan = &ioat->base;
 	struct ioat_ring_ent *desc;
 	bool seen_current = false;
+	int idx = ioat->tail, i;
 	u16 active;
-	int i;
 
 	dev_dbg(to_dev(chan), "%s: head: %#x tail: %#x issued: %#x\n",
 		__func__, ioat->head, ioat->tail, ioat->issued);
@@ -270,13 +270,14 @@ static void __cleanup(struct ioat2_dma_chan *ioat, unsigned long phys_complete)
 	for (i = 0; i < active && !seen_current; i++) {
 		struct dma_async_tx_descriptor *tx;
 
-		prefetch(ioat2_get_ring_ent(ioat, ioat->tail + i + 1));
-		desc = ioat2_get_ring_ent(ioat, ioat->tail + i);
+		smp_read_barrier_depends();
+		prefetch(ioat2_get_ring_ent(ioat, idx + i + 1));
+		desc = ioat2_get_ring_ent(ioat, idx + i);
 		dump_desc_dbg(ioat, desc);
 		tx = &desc->txd;
 		if (tx->cookie) {
 			chan->completed_cookie = tx->cookie;
-			ioat3_dma_unmap(ioat, desc, ioat->tail + i);
+			ioat3_dma_unmap(ioat, desc, idx + i);
 			tx->cookie = 0;
 			if (tx->callback) {
 				tx->callback(tx->callback_param);
@@ -293,69 +294,30 @@ static void __cleanup(struct ioat2_dma_chan *ioat, unsigned long phys_complete)
 			i++;
 		}
 	}
-	ioat->tail += i;
+	smp_mb(); /* finish all descriptor reads before incrementing tail */
+	ioat->tail = idx + i;
 	BUG_ON(active && !seen_current); /* no active descs have written a completion? */
 	chan->last_completion = phys_complete;
 
-	active = ioat2_ring_active(ioat);
-	if (active == 0) {
+	if (active - i == 0) {
 		dev_dbg(to_dev(chan), "%s: cancel completion timeout\n",
 			__func__);
 		clear_bit(IOAT_COMPLETION_PENDING, &chan->state);
 		mod_timer(&chan->timer, jiffies + IDLE_TIMEOUT);
 	}
 	/* 5 microsecond delay per pending descriptor */
-	writew(min((5 * active), IOAT_INTRDELAY_MASK),
+	writew(min((5 * (active - i)), IOAT_INTRDELAY_MASK),
 	       chan->device->reg_base + IOAT_INTRDELAY_OFFSET);
 }
 
-/* try to cleanup, but yield (via spin_trylock) to incoming submissions
- * with the expectation that we will immediately poll again shortly
- */
-static void ioat3_cleanup_poll(struct ioat2_dma_chan *ioat)
+static void ioat3_cleanup(struct ioat2_dma_chan *ioat)
 {
 	struct ioat_chan_common *chan = &ioat->base;
 	unsigned long phys_complete;
 
-	prefetch(chan->completion);
-
-	if (!spin_trylock_bh(&chan->cleanup_lock))
-		return;
-
-	if (!ioat_cleanup_preamble(chan, &phys_complete)) {
-		spin_unlock_bh(&chan->cleanup_lock);
-		return;
-	}
-
-	if (!spin_trylock_bh(&ioat->ring_lock)) {
-		spin_unlock_bh(&chan->cleanup_lock);
-		return;
-	}
-
-	__cleanup(ioat, phys_complete);
-
-	spin_unlock_bh(&ioat->ring_lock);
-	spin_unlock_bh(&chan->cleanup_lock);
-}
-
-/* run cleanup now because we already delayed the interrupt via INTRDELAY */
-static void ioat3_cleanup_sync(struct ioat2_dma_chan *ioat)
-{
-	struct ioat_chan_common *chan = &ioat->base;
-	unsigned long phys_complete;
-
-	prefetch(chan->completion);
-
 	spin_lock_bh(&chan->cleanup_lock);
-	if (!ioat_cleanup_preamble(chan, &phys_complete)) {
-		spin_unlock_bh(&chan->cleanup_lock);
-		return;
-	}
-	spin_lock_bh(&ioat->ring_lock);
-
-	__cleanup(ioat, phys_complete);
-
-	spin_unlock_bh(&ioat->ring_lock);
+	if (ioat_cleanup_preamble(chan, &phys_complete))
+		__cleanup(ioat, phys_complete);
 	spin_unlock_bh(&chan->cleanup_lock);
 }
 
@@ -363,7 +325,7 @@ static void ioat3_cleanup_event(unsigned long data)
 {
 	struct ioat2_dma_chan *ioat = to_ioat2_chan((void *) data);
 
-	ioat3_cleanup_sync(ioat);
+	ioat3_cleanup(ioat);
 	writew(IOAT_CHANCTRL_RUN, ioat->base.reg_base + IOAT_CHANCTRL_OFFSET);
 }
 
@@ -384,12 +346,10 @@ static void ioat3_timer_event(unsigned long data)
 	struct ioat2_dma_chan *ioat = to_ioat2_chan((void *) data);
 	struct ioat_chan_common *chan = &ioat->base;
 
-	spin_lock_bh(&chan->cleanup_lock);
 	if (test_bit(IOAT_COMPLETION_PENDING, &chan->state)) {
 		unsigned long phys_complete;
 		u64 status;
 
-		spin_lock_bh(&ioat->ring_lock);
 		status = ioat_chansts(chan);
 
 		/* when halted due to errors check for channel
@@ -408,26 +368,31 @@ static void ioat3_timer_event(unsigned long data)
 		 * acknowledged a pending completion once, then be more
 		 * forceful with a restart
 		 */
+		spin_lock_bh(&chan->cleanup_lock);
 		if (ioat_cleanup_preamble(chan, &phys_complete))
 			__cleanup(ioat, phys_complete);
-		else if (test_bit(IOAT_COMPLETION_ACK, &chan->state))
+		else if (test_bit(IOAT_COMPLETION_ACK, &chan->state)) {
+			spin_lock_bh(&ioat->prep_lock);
 			ioat3_restart_channel(ioat);
-		else {
+			spin_unlock_bh(&ioat->prep_lock);
+		} else {
 			set_bit(IOAT_COMPLETION_ACK, &chan->state);
 			mod_timer(&chan->timer, jiffies + COMPLETION_TIMEOUT);
 		}
-		spin_unlock_bh(&ioat->ring_lock);
+		spin_unlock_bh(&chan->cleanup_lock);
 	} else {
 		u16 active;
 
 		/* if the ring is idle, empty, and oversized try to step
 		 * down the size
 		 */
-		spin_lock_bh(&ioat->ring_lock);
+		spin_lock_bh(&chan->cleanup_lock);
+		spin_lock_bh(&ioat->prep_lock);
 		active = ioat2_ring_active(ioat);
 		if (active == 0 && ioat->alloc_order > ioat_get_alloc_order())
 			reshape_ring(ioat, ioat->alloc_order-1);
-		spin_unlock_bh(&ioat->ring_lock);
+		spin_unlock_bh(&ioat->prep_lock);
+		spin_unlock_bh(&chan->cleanup_lock);
 
 		/* keep shrinking until we get back to our minimum
 		 * default size
@@ -435,21 +400,20 @@ static void ioat3_timer_event(unsigned long data)
 		if (ioat->alloc_order > ioat_get_alloc_order())
 			mod_timer(&chan->timer, jiffies + IDLE_TIMEOUT);
 	}
-	spin_unlock_bh(&chan->cleanup_lock);
 }
 
 static enum dma_status
-ioat3_is_complete(struct dma_chan *c, dma_cookie_t cookie,
-		  dma_cookie_t *done, dma_cookie_t *used)
+ioat3_tx_status(struct dma_chan *c, dma_cookie_t cookie,
+		struct dma_tx_state *txstate)
 {
 	struct ioat2_dma_chan *ioat = to_ioat2_chan(c);
 
-	if (ioat_is_complete(c, cookie, done, used) == DMA_SUCCESS)
+	if (ioat_tx_status(c, cookie, txstate) == DMA_SUCCESS)
 		return DMA_SUCCESS;
 
-	ioat3_cleanup_poll(ioat);
+	ioat3_cleanup(ioat);
 
-	return ioat_is_complete(c, cookie, done, used);
+	return ioat_tx_status(c, cookie, txstate);
 }
 
 static struct dma_async_tx_descriptor *
@@ -460,15 +424,12 @@ ioat3_prep_memset_lock(struct dma_chan *c, dma_addr_t dest, int value,
 	struct ioat_ring_ent *desc;
 	size_t total_len = len;
 	struct ioat_fill_descriptor *fill;
-	int num_descs;
 	u64 src_data = (0x0101010101010101ULL) * (value & 0xff);
-	u16 idx;
-	int i;
+	int num_descs, idx, i;
 
 	num_descs = ioat2_xferlen_to_descs(ioat, len);
-	if (likely(num_descs) &&
-	    ioat2_alloc_and_lock(&idx, ioat, num_descs) == 0)
-		/* pass */;
+	if (likely(num_descs) && ioat2_check_space_lock(ioat, num_descs) == 0)
+		idx = ioat->head;
 	else
 		return NULL;
 	i = 0;
@@ -513,11 +474,8 @@ __ioat3_prep_xor_lock(struct dma_chan *c, enum sum_check_flags *result,
 	struct ioat_xor_descriptor *xor;
 	struct ioat_xor_ext_descriptor *xor_ex = NULL;
 	struct ioat_dma_descriptor *hw;
+	int num_descs, with_ext, idx, i;
 	u32 offset = 0;
-	int num_descs;
-	int with_ext;
-	int i;
-	u16 idx;
 	u8 op = result ? IOAT_OP_XOR_VAL : IOAT_OP_XOR;
 
 	BUG_ON(src_cnt < 2);
@@ -537,9 +495,8 @@ __ioat3_prep_xor_lock(struct dma_chan *c, enum sum_check_flags *result,
 	 * (legacy) descriptor to ensure all completion writes arrive in
 	 * order.
 	 */
-	if (likely(num_descs) &&
-	    ioat2_alloc_and_lock(&idx, ioat, num_descs+1) == 0)
-		/* pass */;
+	if (likely(num_descs) && ioat2_check_space_lock(ioat, num_descs+1) == 0)
+		idx = ioat->head;
 	else
 		return NULL;
 	i = 0;
@@ -657,11 +614,8 @@ __ioat3_prep_pq_lock(struct dma_chan *c, enum sum_check_flags *result,
 	struct ioat_pq_ext_descriptor *pq_ex = NULL;
 	struct ioat_dma_descriptor *hw;
 	u32 offset = 0;
-	int num_descs;
-	int with_ext;
-	int i, s;
-	u16 idx;
 	u8 op = result ? IOAT_OP_PQ_VAL : IOAT_OP_PQ;
+	int i, s, idx, with_ext, num_descs;
 
 	dev_dbg(to_dev(chan), "%s\n", __func__);
 	/* the engine requires at least two sources (we provide
@@ -687,8 +641,8 @@ __ioat3_prep_pq_lock(struct dma_chan *c, enum sum_check_flags *result,
 	 * order.
 	 */
 	if (likely(num_descs) &&
-	    ioat2_alloc_and_lock(&idx, ioat, num_descs+1) == 0)
-		/* pass */;
+	    ioat2_check_space_lock(ioat, num_descs+1) == 0)
+		idx = ioat->head;
 	else
 		return NULL;
 	i = 0;
@@ -851,10 +805,9 @@ ioat3_prep_interrupt_lock(struct dma_chan *c, unsigned long flags)
 	struct ioat2_dma_chan *ioat = to_ioat2_chan(c);
 	struct ioat_ring_ent *desc;
 	struct ioat_dma_descriptor *hw;
-	u16 idx;
 
-	if (ioat2_alloc_and_lock(&idx, ioat, 1) == 0)
-		desc = ioat2_get_ring_ent(ioat, idx);
+	if (ioat2_check_space_lock(ioat, 1) == 0)
+		desc = ioat2_get_ring_ent(ioat, ioat->head);
 	else
 		return NULL;
 
@@ -977,7 +930,7 @@ static int __devinit ioat_xor_val_self_test(struct ioatdma_device *device)
 
 	tmo = wait_for_completion_timeout(&cmp, msecs_to_jiffies(3000));
 
-	if (dma->device_is_tx_complete(dma_chan, cookie, NULL, NULL) != DMA_SUCCESS) {
+	if (dma->device_tx_status(dma_chan, cookie, NULL) != DMA_SUCCESS) {
 		dev_err(dev, "Self-test xor timed out\n");
 		err = -ENODEV;
 		goto free_resources;
@@ -1031,7 +984,7 @@ static int __devinit ioat_xor_val_self_test(struct ioatdma_device *device)
 
 	tmo = wait_for_completion_timeout(&cmp, msecs_to_jiffies(3000));
 
-	if (dma->device_is_tx_complete(dma_chan, cookie, NULL, NULL) != DMA_SUCCESS) {
+	if (dma->device_tx_status(dma_chan, cookie, NULL) != DMA_SUCCESS) {
 		dev_err(dev, "Self-test validate timed out\n");
 		err = -ENODEV;
 		goto free_resources;
@@ -1072,7 +1025,7 @@ static int __devinit ioat_xor_val_self_test(struct ioatdma_device *device)
 
 	tmo = wait_for_completion_timeout(&cmp, msecs_to_jiffies(3000));
 
-	if (dma->device_is_tx_complete(dma_chan, cookie, NULL, NULL) != DMA_SUCCESS) {
+	if (dma->device_tx_status(dma_chan, cookie, NULL) != DMA_SUCCESS) {
 		dev_err(dev, "Self-test memset timed out\n");
 		err = -ENODEV;
 		goto free_resources;
@@ -1115,7 +1068,7 @@ static int __devinit ioat_xor_val_self_test(struct ioatdma_device *device)
 
 	tmo = wait_for_completion_timeout(&cmp, msecs_to_jiffies(3000));
 
-	if (dma->device_is_tx_complete(dma_chan, cookie, NULL, NULL) != DMA_SUCCESS) {
+	if (dma->device_tx_status(dma_chan, cookie, NULL) != DMA_SUCCESS) {
 		dev_err(dev, "Self-test 2nd validate timed out\n");
 		err = -ENODEV;
 		goto free_resources;
@@ -1222,7 +1175,7 @@ int __devinit ioat3_dma_probe(struct ioatdma_device *device, int dca)
 	if (cap & IOAT_CAP_XOR) {
 		is_raid_device = true;
 		dma->max_xor = 8;
-		dma->xor_align = 2;
+		dma->xor_align = 6;
 
 		dma_cap_set(DMA_XOR, dma->cap_mask);
 		dma->device_prep_dma_xor = ioat3_prep_xor;
@@ -1233,7 +1186,7 @@ int __devinit ioat3_dma_probe(struct ioatdma_device *device, int dca)
 	if (cap & IOAT_CAP_PQ) {
 		is_raid_device = true;
 		dma_set_maxpq(dma, 8, 0);
-		dma->pq_align = 2;
+		dma->pq_align = 6;
 
 		dma_cap_set(DMA_PQ, dma->cap_mask);
 		dma->device_prep_dma_pq = ioat3_prep_pq;
@@ -1243,7 +1196,7 @@ int __devinit ioat3_dma_probe(struct ioatdma_device *device, int dca)
 
 		if (!(cap & IOAT_CAP_XOR)) {
 			dma->max_xor = 8;
-			dma->xor_align = 2;
+			dma->xor_align = 6;
 
 			dma_cap_set(DMA_XOR, dma->cap_mask);
 			dma->device_prep_dma_xor = ioat3_prep_pqxor;
@@ -1259,11 +1212,11 @@ int __devinit ioat3_dma_probe(struct ioatdma_device *device, int dca)
 
 
 	if (is_raid_device) {
-		dma->device_is_tx_complete = ioat3_is_complete;
+		dma->device_tx_status = ioat3_tx_status;
 		device->cleanup_fn = ioat3_cleanup_event;
 		device->timer_fn = ioat3_timer_event;
 	} else {
-		dma->device_is_tx_complete = ioat_is_dma_complete;
+		dma->device_tx_status = ioat_dma_tx_status;
 		device->cleanup_fn = ioat2_cleanup_event;
 		device->timer_fn = ioat2_timer_event;
 	}
diff --git a/drivers/dma/ioat/pci.c b/drivers/dma/ioat/pci.c
index 99ec26725bae..fab37d1cf48d 100644
--- a/drivers/dma/ioat/pci.c
+++ b/drivers/dma/ioat/pci.c
@@ -138,15 +138,10 @@ static int __devinit ioat_pci_probe(struct pci_dev *pdev, const struct pci_devic
 	if (err)
 		return err;
 
-	device = devm_kzalloc(dev, sizeof(*device), GFP_KERNEL);
-	if (!device)
-		return -ENOMEM;
-
-	pci_set_master(pdev);
-
 	device = alloc_ioatdma(pdev, iomap[IOAT_MMIO_BAR]);
 	if (!device)
 		return -ENOMEM;
+	pci_set_master(pdev);
 	pci_set_drvdata(pdev, device);
 
 	device->version = readb(device->reg_base + IOAT_VER_OFFSET);
diff --git a/drivers/dma/iop-adma.c b/drivers/dma/iop-adma.c
index 1ebc801678b0..161c452923b8 100644
--- a/drivers/dma/iop-adma.c
+++ b/drivers/dma/iop-adma.c
@@ -894,14 +894,14 @@ static void iop_adma_free_chan_resources(struct dma_chan *chan)
 }
 
 /**
- * iop_adma_is_complete - poll the status of an ADMA transaction
+ * iop_adma_status - poll the status of an ADMA transaction
  * @chan: ADMA channel handle
  * @cookie: ADMA transaction identifier
+ * @txstate: a holder for the current state of the channel or NULL
  */
-static enum dma_status iop_adma_is_complete(struct dma_chan *chan,
+static enum dma_status iop_adma_status(struct dma_chan *chan,
 					dma_cookie_t cookie,
-					dma_cookie_t *done,
-					dma_cookie_t *used)
+					struct dma_tx_state *txstate)
 {
 	struct iop_adma_chan *iop_chan = to_iop_adma_chan(chan);
 	dma_cookie_t last_used;
@@ -910,12 +910,7 @@ static enum dma_status iop_adma_is_complete(struct dma_chan *chan,
 
 	last_used = chan->cookie;
 	last_complete = iop_chan->completed_cookie;
-
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
-
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 	ret = dma_async_is_complete(cookie, last_complete, last_used);
 	if (ret == DMA_SUCCESS)
 		return ret;
@@ -924,11 +919,7 @@ static enum dma_status iop_adma_is_complete(struct dma_chan *chan,
 
 	last_used = chan->cookie;
 	last_complete = iop_chan->completed_cookie;
-
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 
 	return dma_async_is_complete(cookie, last_complete, last_used);
 }
@@ -1043,7 +1034,7 @@ static int __devinit iop_adma_memcpy_self_test(struct iop_adma_device *device)
 	iop_adma_issue_pending(dma_chan);
 	msleep(1);
 
-	if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) !=
+	if (iop_adma_status(dma_chan, cookie, NULL) !=
 			DMA_SUCCESS) {
 		dev_printk(KERN_ERR, dma_chan->device->dev,
 			"Self-test copy timed out, disabling\n");
@@ -1143,7 +1134,7 @@ iop_adma_xor_val_self_test(struct iop_adma_device *device)
 	iop_adma_issue_pending(dma_chan);
 	msleep(8);
 
-	if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) !=
+	if (iop_adma_status(dma_chan, cookie, NULL) !=
 		DMA_SUCCESS) {
 		dev_printk(KERN_ERR, dma_chan->device->dev,
 			"Self-test xor timed out, disabling\n");
@@ -1190,7 +1181,7 @@ iop_adma_xor_val_self_test(struct iop_adma_device *device)
 	iop_adma_issue_pending(dma_chan);
 	msleep(8);
 
-	if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) != DMA_SUCCESS) {
+	if (iop_adma_status(dma_chan, cookie, NULL) != DMA_SUCCESS) {
 		dev_printk(KERN_ERR, dma_chan->device->dev,
 			"Self-test zero sum timed out, disabling\n");
 		err = -ENODEV;
@@ -1214,7 +1205,7 @@ iop_adma_xor_val_self_test(struct iop_adma_device *device)
 	iop_adma_issue_pending(dma_chan);
 	msleep(8);
 
-	if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) != DMA_SUCCESS) {
+	if (iop_adma_status(dma_chan, cookie, NULL) != DMA_SUCCESS) {
 		dev_printk(KERN_ERR, dma_chan->device->dev,
 			"Self-test memset timed out, disabling\n");
 		err = -ENODEV;
@@ -1246,7 +1237,7 @@ iop_adma_xor_val_self_test(struct iop_adma_device *device)
 	iop_adma_issue_pending(dma_chan);
 	msleep(8);
 
-	if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) != DMA_SUCCESS) {
+	if (iop_adma_status(dma_chan, cookie, NULL) != DMA_SUCCESS) {
 		dev_printk(KERN_ERR, dma_chan->device->dev,
 			"Self-test non-zero sum timed out, disabling\n");
 		err = -ENODEV;
@@ -1341,7 +1332,7 @@ iop_adma_pq_zero_sum_self_test(struct iop_adma_device *device)
 	iop_adma_issue_pending(dma_chan);
 	msleep(8);
 
-	if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) !=
+	if (iop_adma_status(dma_chan, cookie, NULL) !=
 		DMA_SUCCESS) {
 		dev_err(dev, "Self-test pq timed out, disabling\n");
 		err = -ENODEV;
@@ -1378,7 +1369,7 @@ iop_adma_pq_zero_sum_self_test(struct iop_adma_device *device)
 	iop_adma_issue_pending(dma_chan);
 	msleep(8);
 
-	if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) !=
+	if (iop_adma_status(dma_chan, cookie, NULL) !=
 		DMA_SUCCESS) {
 		dev_err(dev, "Self-test pq-zero-sum timed out, disabling\n");
 		err = -ENODEV;
@@ -1410,7 +1401,7 @@ iop_adma_pq_zero_sum_self_test(struct iop_adma_device *device)
 	iop_adma_issue_pending(dma_chan);
 	msleep(8);
 
-	if (iop_adma_is_complete(dma_chan, cookie, NULL, NULL) !=
+	if (iop_adma_status(dma_chan, cookie, NULL) !=
 		DMA_SUCCESS) {
 		dev_err(dev, "Self-test !pq-zero-sum timed out, disabling\n");
 		err = -ENODEV;
@@ -1508,7 +1499,7 @@ static int __devinit iop_adma_probe(struct platform_device *pdev)
 	/* set base routines */
 	dma_dev->device_alloc_chan_resources = iop_adma_alloc_chan_resources;
 	dma_dev->device_free_chan_resources = iop_adma_free_chan_resources;
-	dma_dev->device_is_tx_complete = iop_adma_is_complete;
+	dma_dev->device_tx_status = iop_adma_status;
 	dma_dev->device_issue_pending = iop_adma_issue_pending;
 	dma_dev->dev = &pdev->dev;
 
diff --git a/drivers/dma/ipu/ipu_idmac.c b/drivers/dma/ipu/ipu_idmac.c
index 2a446397c884..cb26ee9773d6 100644
--- a/drivers/dma/ipu/ipu_idmac.c
+++ b/drivers/dma/ipu/ipu_idmac.c
@@ -1472,13 +1472,18 @@ static void idmac_issue_pending(struct dma_chan *chan)
 	 */
 }
 
-static void __idmac_terminate_all(struct dma_chan *chan)
+static int __idmac_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+			   unsigned long arg)
 {
 	struct idmac_channel *ichan = to_idmac_chan(chan);
 	struct idmac *idmac = to_idmac(chan->device);
 	unsigned long flags;
 	int i;
 
+	/* Only supports DMA_TERMINATE_ALL */
+	if (cmd != DMA_TERMINATE_ALL)
+		return -ENXIO;
+
 	ipu_disable_channel(idmac, ichan,
 			    ichan->status >= IPU_CHANNEL_ENABLED);
 
@@ -1505,17 +1510,23 @@ static void __idmac_terminate_all(struct dma_chan *chan)
 	tasklet_enable(&to_ipu(idmac)->tasklet);
 
 	ichan->status = IPU_CHANNEL_INITIALIZED;
+
+	return 0;
 }
 
-static void idmac_terminate_all(struct dma_chan *chan)
+static int idmac_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+			 unsigned long arg)
 {
 	struct idmac_channel *ichan = to_idmac_chan(chan);
+	int ret;
 
 	mutex_lock(&ichan->chan_mutex);
 
-	__idmac_terminate_all(chan);
+	ret = __idmac_control(chan, cmd, arg);
 
 	mutex_unlock(&ichan->chan_mutex);
+
+	return ret;
 }
 
 #ifdef DEBUG
@@ -1607,7 +1618,7 @@ static void idmac_free_chan_resources(struct dma_chan *chan)
 
 	mutex_lock(&ichan->chan_mutex);
 
-	__idmac_terminate_all(chan);
+	__idmac_control(chan, DMA_TERMINATE_ALL, 0);
 
 	if (ichan->status > IPU_CHANNEL_FREE) {
 #ifdef DEBUG
@@ -1637,15 +1648,12 @@ static void idmac_free_chan_resources(struct dma_chan *chan)
 	tasklet_schedule(&to_ipu(idmac)->tasklet);
 }
 
-static enum dma_status idmac_is_tx_complete(struct dma_chan *chan,
-		dma_cookie_t cookie, dma_cookie_t *done, dma_cookie_t *used)
+static enum dma_status idmac_tx_status(struct dma_chan *chan,
+		       dma_cookie_t cookie, struct dma_tx_state *txstate)
 {
 	struct idmac_channel *ichan = to_idmac_chan(chan);
 
-	if (done)
-		*done = ichan->completed;
-	if (used)
-		*used = chan->cookie;
+	dma_set_tx_state(txstate, ichan->completed, chan->cookie, 0);
 	if (cookie != chan->cookie)
 		return DMA_ERROR;
 	return DMA_SUCCESS;
@@ -1664,12 +1672,12 @@ static int __init ipu_idmac_init(struct ipu *ipu)
 	dma->dev				= ipu->dev;
 	dma->device_alloc_chan_resources	= idmac_alloc_chan_resources;
 	dma->device_free_chan_resources		= idmac_free_chan_resources;
-	dma->device_is_tx_complete		= idmac_is_tx_complete;
+	dma->device_tx_status			= idmac_tx_status;
 	dma->device_issue_pending		= idmac_issue_pending;
 
 	/* Compulsory for DMA_SLAVE fields */
 	dma->device_prep_slave_sg		= idmac_prep_slave_sg;
-	dma->device_terminate_all		= idmac_terminate_all;
+	dma->device_control			= idmac_control;
 
 	INIT_LIST_HEAD(&dma->channels);
 	for (i = 0; i < IPU_CHANNELS_NUM; i++) {
@@ -1703,7 +1711,7 @@ static void __exit ipu_idmac_exit(struct ipu *ipu)
 	for (i = 0; i < IPU_CHANNELS_NUM; i++) {
 		struct idmac_channel *ichan = ipu->channel + i;
 
-		idmac_terminate_all(&ichan->dma_chan);
+		idmac_control(&ichan->dma_chan, DMA_TERMINATE_ALL, 0);
 		idmac_prep_slave_sg(&ichan->dma_chan, NULL, 0, DMA_NONE, 0);
 	}
 
diff --git a/drivers/dma/mpc512x_dma.c b/drivers/dma/mpc512x_dma.c
index bbbd58566625..14a8c0f1698e 100644
--- a/drivers/dma/mpc512x_dma.c
+++ b/drivers/dma/mpc512x_dma.c
@@ -541,8 +541,8 @@ static void mpc_dma_issue_pending(struct dma_chan *chan)
 
 /* Check request completion status */
 static enum dma_status
-mpc_dma_is_tx_complete(struct dma_chan *chan, dma_cookie_t cookie,
-					dma_cookie_t *done, dma_cookie_t *used)
+mpc_dma_tx_status(struct dma_chan *chan, dma_cookie_t cookie,
+	       struct dma_tx_state *txstate)
 {
 	struct mpc_dma_chan *mchan = dma_chan_to_mpc_dma_chan(chan);
 	unsigned long flags;
@@ -554,12 +554,7 @@ mpc_dma_is_tx_complete(struct dma_chan *chan, dma_cookie_t cookie,
 	last_complete = mchan->completed_cookie;
 	spin_unlock_irqrestore(&mchan->lock, flags);
 
-	if (done)
-		*done = last_complete;
-
-	if (used)
-		*used = last_used;
-
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 	return dma_async_is_complete(cookie, last_complete, last_used);
 }
 
@@ -635,7 +630,7 @@ mpc_dma_prep_memcpy(struct dma_chan *chan, dma_addr_t dst, dma_addr_t src,
 static int __devinit mpc_dma_probe(struct of_device *op,
 					const struct of_device_id *match)
 {
-	struct device_node *dn = op->node;
+	struct device_node *dn = op->dev.of_node;
 	struct device *dev = &op->dev;
 	struct dma_device *dma;
 	struct mpc_dma *mdma;
@@ -663,7 +658,7 @@ static int __devinit mpc_dma_probe(struct of_device *op,
 	}
 
 	regs_start = res.start;
-	regs_size = res.end - res.start + 1;
+	regs_size = resource_size(&res);
 
 	if (!devm_request_mem_region(dev, regs_start, regs_size, DRV_NAME)) {
 		dev_err(dev, "Error requesting memory region!\n");
@@ -694,7 +689,7 @@ static int __devinit mpc_dma_probe(struct of_device *op,
 	dma->device_alloc_chan_resources = mpc_dma_alloc_chan_resources;
 	dma->device_free_chan_resources = mpc_dma_free_chan_resources;
 	dma->device_issue_pending = mpc_dma_issue_pending;
-	dma->device_is_tx_complete = mpc_dma_is_tx_complete;
+	dma->device_tx_status = mpc_dma_tx_status;
 	dma->device_prep_dma_memcpy = mpc_dma_prep_memcpy;
 
 	INIT_LIST_HEAD(&dma->channels);
@@ -776,12 +771,12 @@ static struct of_device_id mpc_dma_match[] = {
 };
 
 static struct of_platform_driver mpc_dma_driver = {
-	.match_table	= mpc_dma_match,
 	.probe		= mpc_dma_probe,
 	.remove		= __devexit_p(mpc_dma_remove),
-	.driver		= {
-		.name	= DRV_NAME,
-		.owner	= THIS_MODULE,
+	.driver = {
+		.name = DRV_NAME,
+		.owner = THIS_MODULE,
+		.of_match_table	= mpc_dma_match,
 	},
 };
 
diff --git a/drivers/dma/mv_xor.c b/drivers/dma/mv_xor.c
index e2fd34da64f2..86c5ae9fde34 100644
--- a/drivers/dma/mv_xor.c
+++ b/drivers/dma/mv_xor.c
@@ -810,14 +810,14 @@ static void mv_xor_free_chan_resources(struct dma_chan *chan)
 }
 
 /**
- * mv_xor_is_complete - poll the status of an XOR transaction
+ * mv_xor_status - poll the status of an XOR transaction
  * @chan: XOR channel handle
  * @cookie: XOR transaction identifier
+ * @txstate: XOR transactions state holder (or NULL)
  */
-static enum dma_status mv_xor_is_complete(struct dma_chan *chan,
+static enum dma_status mv_xor_status(struct dma_chan *chan,
 					  dma_cookie_t cookie,
-					  dma_cookie_t *done,
-					  dma_cookie_t *used)
+					  struct dma_tx_state *txstate)
 {
 	struct mv_xor_chan *mv_chan = to_mv_xor_chan(chan);
 	dma_cookie_t last_used;
@@ -827,10 +827,7 @@ static enum dma_status mv_xor_is_complete(struct dma_chan *chan,
 	last_used = chan->cookie;
 	last_complete = mv_chan->completed_cookie;
 	mv_chan->is_complete_cookie = cookie;
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 
 	ret = dma_async_is_complete(cookie, last_complete, last_used);
 	if (ret == DMA_SUCCESS) {
@@ -842,11 +839,7 @@ static enum dma_status mv_xor_is_complete(struct dma_chan *chan,
 	last_used = chan->cookie;
 	last_complete = mv_chan->completed_cookie;
 
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
-
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 	return dma_async_is_complete(cookie, last_complete, last_used);
 }
 
@@ -975,7 +968,7 @@ static int __devinit mv_xor_memcpy_self_test(struct mv_xor_device *device)
 	async_tx_ack(tx);
 	msleep(1);
 
-	if (mv_xor_is_complete(dma_chan, cookie, NULL, NULL) !=
+	if (mv_xor_status(dma_chan, cookie, NULL) !=
 	    DMA_SUCCESS) {
 		dev_printk(KERN_ERR, dma_chan->device->dev,
 			   "Self-test copy timed out, disabling\n");
@@ -1073,7 +1066,7 @@ mv_xor_xor_self_test(struct mv_xor_device *device)
 	async_tx_ack(tx);
 	msleep(8);
 
-	if (mv_xor_is_complete(dma_chan, cookie, NULL, NULL) !=
+	if (mv_xor_status(dma_chan, cookie, NULL) !=
 	    DMA_SUCCESS) {
 		dev_printk(KERN_ERR, dma_chan->device->dev,
 			   "Self-test xor timed out, disabling\n");
@@ -1168,7 +1161,7 @@ static int __devinit mv_xor_probe(struct platform_device *pdev)
 	/* set base routines */
 	dma_dev->device_alloc_chan_resources = mv_xor_alloc_chan_resources;
 	dma_dev->device_free_chan_resources = mv_xor_free_chan_resources;
-	dma_dev->device_is_tx_complete = mv_xor_is_complete;
+	dma_dev->device_tx_status = mv_xor_status;
 	dma_dev->device_issue_pending = mv_xor_issue_pending;
 	dma_dev->dev = &pdev->dev;
 
diff --git a/drivers/dma/pl330.c b/drivers/dma/pl330.c
new file mode 100644
index 000000000000..7c50f6dfd3f4
--- /dev/null
+++ b/drivers/dma/pl330.c
@@ -0,0 +1,866 @@
+/* linux/drivers/dma/pl330.c
+ *
+ * Copyright (C) 2010 Samsung Electronics Co. Ltd.
+ *	Jaswinder Singh <jassi.brar@samsung.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ */
+
+#include <linux/io.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/module.h>
+#include <linux/dmaengine.h>
+#include <linux/interrupt.h>
+#include <linux/amba/bus.h>
+#include <linux/amba/pl330.h>
+
+#define NR_DEFAULT_DESC	16
+
+enum desc_status {
+	/* In the DMAC pool */
+	FREE,
+	/*
+	 * Allocted to some channel during prep_xxx
+	 * Also may be sitting on the work_list.
+	 */
+	PREP,
+	/*
+	 * Sitting on the work_list and already submitted
+	 * to the PL330 core. Not more than two descriptors
+	 * of a channel can be BUSY at any time.
+	 */
+	BUSY,
+	/*
+	 * Sitting on the channel work_list but xfer done
+	 * by PL330 core
+	 */
+	DONE,
+};
+
+struct dma_pl330_chan {
+	/* Schedule desc completion */
+	struct tasklet_struct task;
+
+	/* DMA-Engine Channel */
+	struct dma_chan chan;
+
+	/* Last completed cookie */
+	dma_cookie_t completed;
+
+	/* List of to be xfered descriptors */
+	struct list_head work_list;
+
+	/* Pointer to the DMAC that manages this channel,
+	 * NULL if the channel is available to be acquired.
+	 * As the parent, this DMAC also provides descriptors
+	 * to the channel.
+	 */
+	struct dma_pl330_dmac *dmac;
+
+	/* To protect channel manipulation */
+	spinlock_t lock;
+
+	/* Token of a hardware channel thread of PL330 DMAC
+	 * NULL if the channel is available to be acquired.
+	 */
+	void *pl330_chid;
+};
+
+struct dma_pl330_dmac {
+	struct pl330_info pif;
+
+	/* DMA-Engine Device */
+	struct dma_device ddma;
+
+	/* Pool of descriptors available for the DMAC's channels */
+	struct list_head desc_pool;
+	/* To protect desc_pool manipulation */
+	spinlock_t pool_lock;
+
+	/* Peripheral channels connected to this DMAC */
+	struct dma_pl330_chan peripherals[0]; /* keep at end */
+};
+
+struct dma_pl330_desc {
+	/* To attach to a queue as child */
+	struct list_head node;
+
+	/* Descriptor for the DMA Engine API */
+	struct dma_async_tx_descriptor txd;
+
+	/* Xfer for PL330 core */
+	struct pl330_xfer px;
+
+	struct pl330_reqcfg rqcfg;
+	struct pl330_req req;
+
+	enum desc_status status;
+
+	/* The channel which currently holds this desc */
+	struct dma_pl330_chan *pchan;
+};
+
+static inline struct dma_pl330_chan *
+to_pchan(struct dma_chan *ch)
+{
+	if (!ch)
+		return NULL;
+
+	return container_of(ch, struct dma_pl330_chan, chan);
+}
+
+static inline struct dma_pl330_desc *
+to_desc(struct dma_async_tx_descriptor *tx)
+{
+	return container_of(tx, struct dma_pl330_desc, txd);
+}
+
+static inline void free_desc_list(struct list_head *list)
+{
+	struct dma_pl330_dmac *pdmac;
+	struct dma_pl330_desc *desc;
+	struct dma_pl330_chan *pch;
+	unsigned long flags;
+
+	if (list_empty(list))
+		return;
+
+	/* Finish off the work list */
+	list_for_each_entry(desc, list, node) {
+		dma_async_tx_callback callback;
+		void *param;
+
+		/* All desc in a list belong to same channel */
+		pch = desc->pchan;
+		callback = desc->txd.callback;
+		param = desc->txd.callback_param;
+
+		if (callback)
+			callback(param);
+
+		desc->pchan = NULL;
+	}
+
+	pdmac = pch->dmac;
+
+	spin_lock_irqsave(&pdmac->pool_lock, flags);
+	list_splice_tail_init(list, &pdmac->desc_pool);
+	spin_unlock_irqrestore(&pdmac->pool_lock, flags);
+}
+
+static inline void fill_queue(struct dma_pl330_chan *pch)
+{
+	struct dma_pl330_desc *desc;
+	int ret;
+
+	list_for_each_entry(desc, &pch->work_list, node) {
+
+		/* If already submitted */
+		if (desc->status == BUSY)
+			break;
+
+		ret = pl330_submit_req(pch->pl330_chid,
+						&desc->req);
+		if (!ret) {
+			desc->status = BUSY;
+			break;
+		} else if (ret == -EAGAIN) {
+			/* QFull or DMAC Dying */
+			break;
+		} else {
+			/* Unacceptable request */
+			desc->status = DONE;
+			dev_err(pch->dmac->pif.dev, "%s:%d Bad Desc(%d)\n",
+					__func__, __LINE__, desc->txd.cookie);
+			tasklet_schedule(&pch->task);
+		}
+	}
+}
+
+static void pl330_tasklet(unsigned long data)
+{
+	struct dma_pl330_chan *pch = (struct dma_pl330_chan *)data;
+	struct dma_pl330_desc *desc, *_dt;
+	unsigned long flags;
+	LIST_HEAD(list);
+
+	spin_lock_irqsave(&pch->lock, flags);
+
+	/* Pick up ripe tomatoes */
+	list_for_each_entry_safe(desc, _dt, &pch->work_list, node)
+		if (desc->status == DONE) {
+			pch->completed = desc->txd.cookie;
+			list_move_tail(&desc->node, &list);
+		}
+
+	/* Try to submit a req imm. next to the last completed cookie */
+	fill_queue(pch);
+
+	/* Make sure the PL330 Channel thread is active */
+	pl330_chan_ctrl(pch->pl330_chid, PL330_OP_START);
+
+	spin_unlock_irqrestore(&pch->lock, flags);
+
+	free_desc_list(&list);
+}
+
+static void dma_pl330_rqcb(void *token, enum pl330_op_err err)
+{
+	struct dma_pl330_desc *desc = token;
+	struct dma_pl330_chan *pch = desc->pchan;
+	unsigned long flags;
+
+	/* If desc aborted */
+	if (!pch)
+		return;
+
+	spin_lock_irqsave(&pch->lock, flags);
+
+	desc->status = DONE;
+
+	spin_unlock_irqrestore(&pch->lock, flags);
+
+	tasklet_schedule(&pch->task);
+}
+
+static int pl330_alloc_chan_resources(struct dma_chan *chan)
+{
+	struct dma_pl330_chan *pch = to_pchan(chan);
+	struct dma_pl330_dmac *pdmac = pch->dmac;
+	unsigned long flags;
+
+	spin_lock_irqsave(&pch->lock, flags);
+
+	pch->completed = chan->cookie = 1;
+
+	pch->pl330_chid = pl330_request_channel(&pdmac->pif);
+	if (!pch->pl330_chid) {
+		spin_unlock_irqrestore(&pch->lock, flags);
+		return 0;
+	}
+
+	tasklet_init(&pch->task, pl330_tasklet, (unsigned long) pch);
+
+	spin_unlock_irqrestore(&pch->lock, flags);
+
+	return 1;
+}
+
+static int pl330_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd, unsigned long arg)
+{
+	struct dma_pl330_chan *pch = to_pchan(chan);
+	struct dma_pl330_desc *desc;
+	unsigned long flags;
+
+	/* Only supports DMA_TERMINATE_ALL */
+	if (cmd != DMA_TERMINATE_ALL)
+		return -ENXIO;
+
+	spin_lock_irqsave(&pch->lock, flags);
+
+	/* FLUSH the PL330 Channel thread */
+	pl330_chan_ctrl(pch->pl330_chid, PL330_OP_FLUSH);
+
+	/* Mark all desc done */
+	list_for_each_entry(desc, &pch->work_list, node)
+		desc->status = DONE;
+
+	spin_unlock_irqrestore(&pch->lock, flags);
+
+	pl330_tasklet((unsigned long) pch);
+
+	return 0;
+}
+
+static void pl330_free_chan_resources(struct dma_chan *chan)
+{
+	struct dma_pl330_chan *pch = to_pchan(chan);
+	unsigned long flags;
+
+	spin_lock_irqsave(&pch->lock, flags);
+
+	tasklet_kill(&pch->task);
+
+	pl330_release_channel(pch->pl330_chid);
+	pch->pl330_chid = NULL;
+
+	spin_unlock_irqrestore(&pch->lock, flags);
+}
+
+static enum dma_status
+pl330_tx_status(struct dma_chan *chan, dma_cookie_t cookie,
+		 struct dma_tx_state *txstate)
+{
+	struct dma_pl330_chan *pch = to_pchan(chan);
+	dma_cookie_t last_done, last_used;
+	int ret;
+
+	last_done = pch->completed;
+	last_used = chan->cookie;
+
+	ret = dma_async_is_complete(cookie, last_done, last_used);
+
+	dma_set_tx_state(txstate, last_done, last_used, 0);
+
+	return ret;
+}
+
+static void pl330_issue_pending(struct dma_chan *chan)
+{
+	pl330_tasklet((unsigned long) to_pchan(chan));
+}
+
+/*
+ * We returned the last one of the circular list of descriptor(s)
+ * from prep_xxx, so the argument to submit corresponds to the last
+ * descriptor of the list.
+ */
+static dma_cookie_t pl330_tx_submit(struct dma_async_tx_descriptor *tx)
+{
+	struct dma_pl330_desc *desc, *last = to_desc(tx);
+	struct dma_pl330_chan *pch = to_pchan(tx->chan);
+	dma_cookie_t cookie;
+	unsigned long flags;
+
+	spin_lock_irqsave(&pch->lock, flags);
+
+	/* Assign cookies to all nodes */
+	cookie = tx->chan->cookie;
+
+	while (!list_empty(&last->node)) {
+		desc = list_entry(last->node.next, struct dma_pl330_desc, node);
+
+		if (++cookie < 0)
+			cookie = 1;
+		desc->txd.cookie = cookie;
+
+		list_move_tail(&desc->node, &pch->work_list);
+	}
+
+	if (++cookie < 0)
+		cookie = 1;
+	last->txd.cookie = cookie;
+
+	list_add_tail(&last->node, &pch->work_list);
+
+	tx->chan->cookie = cookie;
+
+	spin_unlock_irqrestore(&pch->lock, flags);
+
+	return cookie;
+}
+
+static inline void _init_desc(struct dma_pl330_desc *desc)
+{
+	desc->pchan = NULL;
+	desc->req.x = &desc->px;
+	desc->req.token = desc;
+	desc->rqcfg.swap = SWAP_NO;
+	desc->rqcfg.privileged = 0;
+	desc->rqcfg.insnaccess = 0;
+	desc->rqcfg.scctl = SCCTRL0;
+	desc->rqcfg.dcctl = DCCTRL0;
+	desc->req.cfg = &desc->rqcfg;
+	desc->req.xfer_cb = dma_pl330_rqcb;
+	desc->txd.tx_submit = pl330_tx_submit;
+
+	INIT_LIST_HEAD(&desc->node);
+}
+
+/* Returns the number of descriptors added to the DMAC pool */
+int add_desc(struct dma_pl330_dmac *pdmac, gfp_t flg, int count)
+{
+	struct dma_pl330_desc *desc;
+	unsigned long flags;
+	int i;
+
+	if (!pdmac)
+		return 0;
+
+	desc = kmalloc(count * sizeof(*desc), flg);
+	if (!desc)
+		return 0;
+
+	spin_lock_irqsave(&pdmac->pool_lock, flags);
+
+	for (i = 0; i < count; i++) {
+		_init_desc(&desc[i]);
+		list_add_tail(&desc[i].node, &pdmac->desc_pool);
+	}
+
+	spin_unlock_irqrestore(&pdmac->pool_lock, flags);
+
+	return count;
+}
+
+static struct dma_pl330_desc *
+pluck_desc(struct dma_pl330_dmac *pdmac)
+{
+	struct dma_pl330_desc *desc = NULL;
+	unsigned long flags;
+
+	if (!pdmac)
+		return NULL;
+
+	spin_lock_irqsave(&pdmac->pool_lock, flags);
+
+	if (!list_empty(&pdmac->desc_pool)) {
+		desc = list_entry(pdmac->desc_pool.next,
+				struct dma_pl330_desc, node);
+
+		list_del_init(&desc->node);
+
+		desc->status = PREP;
+		desc->txd.callback = NULL;
+	}
+
+	spin_unlock_irqrestore(&pdmac->pool_lock, flags);
+
+	return desc;
+}
+
+static struct dma_pl330_desc *pl330_get_desc(struct dma_pl330_chan *pch)
+{
+	struct dma_pl330_dmac *pdmac = pch->dmac;
+	struct dma_pl330_peri *peri = pch->chan.private;
+	struct dma_pl330_desc *desc;
+
+	/* Pluck one desc from the pool of DMAC */
+	desc = pluck_desc(pdmac);
+
+	/* If the DMAC pool is empty, alloc new */
+	if (!desc) {
+		if (!add_desc(pdmac, GFP_ATOMIC, 1))
+			return NULL;
+
+		/* Try again */
+		desc = pluck_desc(pdmac);
+		if (!desc) {
+			dev_err(pch->dmac->pif.dev,
+				"%s:%d ALERT!\n", __func__, __LINE__);
+			return NULL;
+		}
+	}
+
+	/* Initialize the descriptor */
+	desc->pchan = pch;
+	desc->txd.cookie = 0;
+	async_tx_ack(&desc->txd);
+
+	desc->req.rqtype = peri->rqtype;
+	desc->req.peri = peri->peri_id;
+
+	dma_async_tx_descriptor_init(&desc->txd, &pch->chan);
+
+	return desc;
+}
+
+static inline void fill_px(struct pl330_xfer *px,
+		dma_addr_t dst, dma_addr_t src, size_t len)
+{
+	px->next = NULL;
+	px->bytes = len;
+	px->dst_addr = dst;
+	px->src_addr = src;
+}
+
+static struct dma_pl330_desc *
+__pl330_prep_dma_memcpy(struct dma_pl330_chan *pch, dma_addr_t dst,
+		dma_addr_t src, size_t len)
+{
+	struct dma_pl330_desc *desc = pl330_get_desc(pch);
+
+	if (!desc) {
+		dev_err(pch->dmac->pif.dev, "%s:%d Unable to fetch desc\n",
+			__func__, __LINE__);
+		return NULL;
+	}
+
+	/*
+	 * Ideally we should lookout for reqs bigger than
+	 * those that can be programmed with 256 bytes of
+	 * MC buffer, but considering a req size is seldom
+	 * going to be word-unaligned and more than 200MB,
+	 * we take it easy.
+	 * Also, should the limit is reached we'd rather
+	 * have the platform increase MC buffer size than
+	 * complicating this API driver.
+	 */
+	fill_px(&desc->px, dst, src, len);
+
+	return desc;
+}
+
+/* Call after fixing burst size */
+static inline int get_burst_len(struct dma_pl330_desc *desc, size_t len)
+{
+	struct dma_pl330_chan *pch = desc->pchan;
+	struct pl330_info *pi = &pch->dmac->pif;
+	int burst_len;
+
+	burst_len = pi->pcfg.data_bus_width / 8;
+	burst_len *= pi->pcfg.data_buf_dep;
+	burst_len >>= desc->rqcfg.brst_size;
+
+	/* src/dst_burst_len can't be more than 16 */
+	if (burst_len > 16)
+		burst_len = 16;
+
+	while (burst_len > 1) {
+		if (!(len % (burst_len << desc->rqcfg.brst_size)))
+			break;
+		burst_len--;
+	}
+
+	return burst_len;
+}
+
+static struct dma_async_tx_descriptor *
+pl330_prep_dma_memcpy(struct dma_chan *chan, dma_addr_t dst,
+		dma_addr_t src, size_t len, unsigned long flags)
+{
+	struct dma_pl330_desc *desc;
+	struct dma_pl330_chan *pch = to_pchan(chan);
+	struct dma_pl330_peri *peri = chan->private;
+	struct pl330_info *pi;
+	int burst;
+
+	if (unlikely(!pch || !len || !peri))
+		return NULL;
+
+	if (peri->rqtype != MEMTOMEM)
+		return NULL;
+
+	pi = &pch->dmac->pif;
+
+	desc = __pl330_prep_dma_memcpy(pch, dst, src, len);
+	if (!desc)
+		return NULL;
+
+	desc->rqcfg.src_inc = 1;
+	desc->rqcfg.dst_inc = 1;
+
+	/* Select max possible burst size */
+	burst = pi->pcfg.data_bus_width / 8;
+
+	while (burst > 1) {
+		if (!(len % burst))
+			break;
+		burst /= 2;
+	}
+
+	desc->rqcfg.brst_size = 0;
+	while (burst != (1 << desc->rqcfg.brst_size))
+		desc->rqcfg.brst_size++;
+
+	desc->rqcfg.brst_len = get_burst_len(desc, len);
+
+	desc->txd.flags = flags;
+
+	return &desc->txd;
+}
+
+static struct dma_async_tx_descriptor *
+pl330_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
+		unsigned int sg_len, enum dma_data_direction direction,
+		unsigned long flg)
+{
+	struct dma_pl330_desc *first, *desc = NULL;
+	struct dma_pl330_chan *pch = to_pchan(chan);
+	struct dma_pl330_peri *peri = chan->private;
+	struct scatterlist *sg;
+	unsigned long flags;
+	int i, burst_size;
+	dma_addr_t addr;
+
+	if (unlikely(!pch || !sgl || !sg_len))
+		return NULL;
+
+	/* Make sure the direction is consistent */
+	if ((direction == DMA_TO_DEVICE &&
+				peri->rqtype != MEMTODEV) ||
+			(direction == DMA_FROM_DEVICE &&
+				peri->rqtype != DEVTOMEM)) {
+		dev_err(pch->dmac->pif.dev, "%s:%d Invalid Direction\n",
+				__func__, __LINE__);
+		return NULL;
+	}
+
+	addr = peri->fifo_addr;
+	burst_size = peri->burst_sz;
+
+	first = NULL;
+
+	for_each_sg(sgl, sg, sg_len, i) {
+
+		desc = pl330_get_desc(pch);
+		if (!desc) {
+			struct dma_pl330_dmac *pdmac = pch->dmac;
+
+			dev_err(pch->dmac->pif.dev,
+				"%s:%d Unable to fetch desc\n",
+				__func__, __LINE__);
+			if (!first)
+				return NULL;
+
+			spin_lock_irqsave(&pdmac->pool_lock, flags);
+
+			while (!list_empty(&first->node)) {
+				desc = list_entry(first->node.next,
+						struct dma_pl330_desc, node);
+				list_move_tail(&desc->node, &pdmac->desc_pool);
+			}
+
+			list_move_tail(&first->node, &pdmac->desc_pool);
+
+			spin_unlock_irqrestore(&pdmac->pool_lock, flags);
+
+			return NULL;
+		}
+
+		if (!first)
+			first = desc;
+		else
+			list_add_tail(&desc->node, &first->node);
+
+		if (direction == DMA_TO_DEVICE) {
+			desc->rqcfg.src_inc = 1;
+			desc->rqcfg.dst_inc = 0;
+			fill_px(&desc->px,
+				addr, sg_dma_address(sg), sg_dma_len(sg));
+		} else {
+			desc->rqcfg.src_inc = 0;
+			desc->rqcfg.dst_inc = 1;
+			fill_px(&desc->px,
+				sg_dma_address(sg), addr, sg_dma_len(sg));
+		}
+
+		desc->rqcfg.brst_size = burst_size;
+		desc->rqcfg.brst_len = 1;
+	}
+
+	/* Return the last desc in the chain */
+	desc->txd.flags = flg;
+	return &desc->txd;
+}
+
+static irqreturn_t pl330_irq_handler(int irq, void *data)
+{
+	if (pl330_update(data))
+		return IRQ_HANDLED;
+	else
+		return IRQ_NONE;
+}
+
+static int __devinit
+pl330_probe(struct amba_device *adev, struct amba_id *id)
+{
+	struct dma_pl330_platdata *pdat;
+	struct dma_pl330_dmac *pdmac;
+	struct dma_pl330_chan *pch;
+	struct pl330_info *pi;
+	struct dma_device *pd;
+	struct resource *res;
+	int i, ret, irq;
+
+	pdat = adev->dev.platform_data;
+
+	if (!pdat || !pdat->nr_valid_peri) {
+		dev_err(&adev->dev, "platform data missing\n");
+		return -ENODEV;
+	}
+
+	/* Allocate a new DMAC and its Channels */
+	pdmac = kzalloc(pdat->nr_valid_peri * sizeof(*pch)
+				+ sizeof(*pdmac), GFP_KERNEL);
+	if (!pdmac) {
+		dev_err(&adev->dev, "unable to allocate mem\n");
+		return -ENOMEM;
+	}
+
+	pi = &pdmac->pif;
+	pi->dev = &adev->dev;
+	pi->pl330_data = NULL;
+	pi->mcbufsz = pdat->mcbuf_sz;
+
+	res = &adev->res;
+	request_mem_region(res->start, resource_size(res), "dma-pl330");
+
+	pi->base = ioremap(res->start, resource_size(res));
+	if (!pi->base) {
+		ret = -ENXIO;
+		goto probe_err1;
+	}
+
+	irq = adev->irq[0];
+	ret = request_irq(irq, pl330_irq_handler, 0,
+			dev_name(&adev->dev), pi);
+	if (ret)
+		goto probe_err2;
+
+	ret = pl330_add(pi);
+	if (ret)
+		goto probe_err3;
+
+	INIT_LIST_HEAD(&pdmac->desc_pool);
+	spin_lock_init(&pdmac->pool_lock);
+
+	/* Create a descriptor pool of default size */
+	if (!add_desc(pdmac, GFP_KERNEL, NR_DEFAULT_DESC))
+		dev_warn(&adev->dev, "unable to allocate desc\n");
+
+	pd = &pdmac->ddma;
+	INIT_LIST_HEAD(&pd->channels);
+
+	/* Initialize channel parameters */
+	for (i = 0; i < pdat->nr_valid_peri; i++) {
+		struct dma_pl330_peri *peri = &pdat->peri[i];
+		pch = &pdmac->peripherals[i];
+
+		switch (peri->rqtype) {
+		case MEMTOMEM:
+			dma_cap_set(DMA_MEMCPY, pd->cap_mask);
+			break;
+		case MEMTODEV:
+		case DEVTOMEM:
+			dma_cap_set(DMA_SLAVE, pd->cap_mask);
+			break;
+		default:
+			dev_err(&adev->dev, "DEVTODEV Not Supported\n");
+			continue;
+		}
+
+		INIT_LIST_HEAD(&pch->work_list);
+		spin_lock_init(&pch->lock);
+		pch->pl330_chid = NULL;
+		pch->chan.private = peri;
+		pch->chan.device = pd;
+		pch->chan.chan_id = i;
+		pch->dmac = pdmac;
+
+		/* Add the channel to the DMAC list */
+		pd->chancnt++;
+		list_add_tail(&pch->chan.device_node, &pd->channels);
+	}
+
+	pd->dev = &adev->dev;
+
+	pd->device_alloc_chan_resources = pl330_alloc_chan_resources;
+	pd->device_free_chan_resources = pl330_free_chan_resources;
+	pd->device_prep_dma_memcpy = pl330_prep_dma_memcpy;
+	pd->device_tx_status = pl330_tx_status;
+	pd->device_prep_slave_sg = pl330_prep_slave_sg;
+	pd->device_control = pl330_control;
+	pd->device_issue_pending = pl330_issue_pending;
+
+	ret = dma_async_device_register(pd);
+	if (ret) {
+		dev_err(&adev->dev, "unable to register DMAC\n");
+		goto probe_err4;
+	}
+
+	amba_set_drvdata(adev, pdmac);
+
+	dev_info(&adev->dev,
+		"Loaded driver for PL330 DMAC-%d\n", adev->periphid);
+	dev_info(&adev->dev,
+		"\tDBUFF-%ux%ubytes Num_Chans-%u Num_Peri-%u Num_Events-%u\n",
+		pi->pcfg.data_buf_dep,
+		pi->pcfg.data_bus_width / 8, pi->pcfg.num_chan,
+		pi->pcfg.num_peri, pi->pcfg.num_events);
+
+	return 0;
+
+probe_err4:
+	pl330_del(pi);
+probe_err3:
+	free_irq(irq, pi);
+probe_err2:
+	iounmap(pi->base);
+probe_err1:
+	release_mem_region(res->start, resource_size(res));
+	kfree(pdmac);
+
+	return ret;
+}
+
+static int __devexit pl330_remove(struct amba_device *adev)
+{
+	struct dma_pl330_dmac *pdmac = amba_get_drvdata(adev);
+	struct dma_pl330_chan *pch, *_p;
+	struct pl330_info *pi;
+	struct resource *res;
+	int irq;
+
+	if (!pdmac)
+		return 0;
+
+	amba_set_drvdata(adev, NULL);
+
+	/* Idle the DMAC */
+	list_for_each_entry_safe(pch, _p, &pdmac->ddma.channels,
+			chan.device_node) {
+
+		/* Remove the channel */
+		list_del(&pch->chan.device_node);
+
+		/* Flush the channel */
+		pl330_control(&pch->chan, DMA_TERMINATE_ALL, 0);
+		pl330_free_chan_resources(&pch->chan);
+	}
+
+	pi = &pdmac->pif;
+
+	pl330_del(pi);
+
+	irq = adev->irq[0];
+	free_irq(irq, pi);
+
+	iounmap(pi->base);
+
+	res = &adev->res;
+	release_mem_region(res->start, resource_size(res));
+
+	kfree(pdmac);
+
+	return 0;
+}
+
+static struct amba_id pl330_ids[] = {
+	{
+		.id	= 0x00041330,
+		.mask	= 0x000fffff,
+	},
+	{ 0, 0 },
+};
+
+static struct amba_driver pl330_driver = {
+	.drv = {
+		.owner = THIS_MODULE,
+		.name = "dma-pl330",
+	},
+	.id_table = pl330_ids,
+	.probe = pl330_probe,
+	.remove = pl330_remove,
+};
+
+static int __init pl330_init(void)
+{
+	return amba_driver_register(&pl330_driver);
+}
+module_init(pl330_init);
+
+static void __exit pl330_exit(void)
+{
+	amba_driver_unregister(&pl330_driver);
+	return;
+}
+module_exit(pl330_exit);
+
+MODULE_AUTHOR("Jaswinder Singh <jassi.brar@samsung.com>");
+MODULE_DESCRIPTION("API Driver for PL330 DMAC");
+MODULE_LICENSE("GPL");
diff --git a/drivers/dma/ppc4xx/adma.c b/drivers/dma/ppc4xx/adma.c
index d44626fa35ad..5a22ca6927e5 100644
--- a/drivers/dma/ppc4xx/adma.c
+++ b/drivers/dma/ppc4xx/adma.c
@@ -3935,12 +3935,13 @@ static void ppc440spe_adma_free_chan_resources(struct dma_chan *chan)
 }
 
 /**
- * ppc440spe_adma_is_complete - poll the status of an ADMA transaction
+ * ppc440spe_adma_tx_status - poll the status of an ADMA transaction
  * @chan: ADMA channel handle
  * @cookie: ADMA transaction identifier
+ * @txstate: a holder for the current state of the channel
  */
-static enum dma_status ppc440spe_adma_is_complete(struct dma_chan *chan,
-	dma_cookie_t cookie, dma_cookie_t *done, dma_cookie_t *used)
+static enum dma_status ppc440spe_adma_tx_status(struct dma_chan *chan,
+			dma_cookie_t cookie, struct dma_tx_state *txstate)
 {
 	struct ppc440spe_adma_chan *ppc440spe_chan;
 	dma_cookie_t last_used;
@@ -3951,10 +3952,7 @@ static enum dma_status ppc440spe_adma_is_complete(struct dma_chan *chan,
 	last_used = chan->cookie;
 	last_complete = ppc440spe_chan->completed_cookie;
 
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 
 	ret = dma_async_is_complete(cookie, last_complete, last_used);
 	if (ret == DMA_SUCCESS)
@@ -3965,10 +3963,7 @@ static enum dma_status ppc440spe_adma_is_complete(struct dma_chan *chan,
 	last_used = chan->cookie;
 	last_complete = ppc440spe_chan->completed_cookie;
 
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 
 	return dma_async_is_complete(cookie, last_complete, last_used);
 }
@@ -4180,7 +4175,7 @@ static void ppc440spe_adma_init_capabilities(struct ppc440spe_adma_device *adev)
 				ppc440spe_adma_alloc_chan_resources;
 	adev->common.device_free_chan_resources =
 				ppc440spe_adma_free_chan_resources;
-	adev->common.device_is_tx_complete = ppc440spe_adma_is_complete;
+	adev->common.device_tx_status = ppc440spe_adma_tx_status;
 	adev->common.device_issue_pending = ppc440spe_adma_issue_pending;
 
 	/* Set prep routines based on capability */
@@ -4399,7 +4394,7 @@ static void ppc440spe_adma_release_irqs(struct ppc440spe_adma_device *adev,
 static int __devinit ppc440spe_adma_probe(struct of_device *ofdev,
 					  const struct of_device_id *match)
 {
-	struct device_node *np = ofdev->node;
+	struct device_node *np = ofdev->dev.of_node;
 	struct resource res;
 	struct ppc440spe_adma_device *adev;
 	struct ppc440spe_adma_chan *chan;
@@ -4631,7 +4626,7 @@ out:
 static int __devexit ppc440spe_adma_remove(struct of_device *ofdev)
 {
 	struct ppc440spe_adma_device *adev = dev_get_drvdata(&ofdev->dev);
-	struct device_node *np = ofdev->node;
+	struct device_node *np = ofdev->dev.of_node;
 	struct resource res;
 	struct dma_chan *chan, *_chan;
 	struct ppc_dma_chan_ref *ref, *_ref;
@@ -4949,12 +4944,12 @@ static const struct of_device_id ppc440spe_adma_of_match[] __devinitconst = {
 MODULE_DEVICE_TABLE(of, ppc440spe_adma_of_match);
 
 static struct of_platform_driver ppc440spe_adma_driver = {
-	.match_table = ppc440spe_adma_of_match,
 	.probe = ppc440spe_adma_probe,
 	.remove = __devexit_p(ppc440spe_adma_remove),
 	.driver = {
 		.name = "PPC440SP(E)-ADMA",
 		.owner = THIS_MODULE,
+		.of_match_table = ppc440spe_adma_of_match,
 	},
 };
 
diff --git a/drivers/dma/shdma.c b/drivers/dma/shdma.c
index 323afef77802..a2a519fd2a24 100644
--- a/drivers/dma/shdma.c
+++ b/drivers/dma/shdma.c
@@ -597,12 +597,17 @@ static struct dma_async_tx_descriptor *sh_dmae_prep_slave_sg(
 			       direction, flags);
 }
 
-static void sh_dmae_terminate_all(struct dma_chan *chan)
+static int sh_dmae_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+			   unsigned long arg)
 {
 	struct sh_dmae_chan *sh_chan = to_sh_chan(chan);
 
+	/* Only supports DMA_TERMINATE_ALL */
+	if (cmd != DMA_TERMINATE_ALL)
+		return -ENXIO;
+
 	if (!chan)
-		return;
+		return -EINVAL;
 
 	dmae_halt(sh_chan);
 
@@ -618,6 +623,8 @@ static void sh_dmae_terminate_all(struct dma_chan *chan)
 	spin_unlock_bh(&sh_chan->desc_lock);
 
 	sh_dmae_chan_ld_cleanup(sh_chan, true);
+
+	return 0;
 }
 
 static dma_async_tx_callback __ld_cleanup(struct sh_dmae_chan *sh_chan, bool all)
@@ -715,6 +722,10 @@ static void sh_dmae_chan_ld_cleanup(struct sh_dmae_chan *sh_chan, bool all)
 {
 	while (__ld_cleanup(sh_chan, all))
 		;
+
+	if (all)
+		/* Terminating - forgive uncompleted cookies */
+		sh_chan->completed_cookie = sh_chan->common.cookie;
 }
 
 static void sh_chan_xfer_ld_queue(struct sh_dmae_chan *sh_chan)
@@ -749,10 +760,9 @@ static void sh_dmae_memcpy_issue_pending(struct dma_chan *chan)
 	sh_chan_xfer_ld_queue(sh_chan);
 }
 
-static enum dma_status sh_dmae_is_complete(struct dma_chan *chan,
+static enum dma_status sh_dmae_tx_status(struct dma_chan *chan,
 					dma_cookie_t cookie,
-					dma_cookie_t *done,
-					dma_cookie_t *used)
+					struct dma_tx_state *txstate)
 {
 	struct sh_dmae_chan *sh_chan = to_sh_chan(chan);
 	dma_cookie_t last_used;
@@ -764,12 +774,7 @@ static enum dma_status sh_dmae_is_complete(struct dma_chan *chan,
 	last_used = chan->cookie;
 	last_complete = sh_chan->completed_cookie;
 	BUG_ON(last_complete < 0);
-
-	if (done)
-		*done = last_complete;
-
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 
 	spin_lock_bh(&sh_chan->desc_lock);
 
@@ -1041,12 +1046,12 @@ static int __init sh_dmae_probe(struct platform_device *pdev)
 		= sh_dmae_alloc_chan_resources;
 	shdev->common.device_free_chan_resources = sh_dmae_free_chan_resources;
 	shdev->common.device_prep_dma_memcpy = sh_dmae_prep_memcpy;
-	shdev->common.device_is_tx_complete = sh_dmae_is_complete;
+	shdev->common.device_tx_status = sh_dmae_tx_status;
 	shdev->common.device_issue_pending = sh_dmae_memcpy_issue_pending;
 
 	/* Compulsory for DMA_SLAVE fields */
 	shdev->common.device_prep_slave_sg = sh_dmae_prep_slave_sg;
-	shdev->common.device_terminate_all = sh_dmae_terminate_all;
+	shdev->common.device_control = sh_dmae_control;
 
 	shdev->common.dev = &pdev->dev;
 	/* Default transfer size of 32 bytes requires 32-byte alignment */
@@ -1187,6 +1192,7 @@ static struct platform_driver sh_dmae_driver = {
 	.remove		= __exit_p(sh_dmae_remove),
 	.shutdown	= sh_dmae_shutdown,
 	.driver = {
+		.owner	= THIS_MODULE,
 		.name	= "sh-dma-engine",
 	},
 };
diff --git a/drivers/dma/ste_dma40.c b/drivers/dma/ste_dma40.c
new file mode 100644
index 000000000000..c426829f6ab8
--- /dev/null
+++ b/drivers/dma/ste_dma40.c
@@ -0,0 +1,2657 @@
+/*
+ * driver/dma/ste_dma40.c
+ *
+ * Copyright (C) ST-Ericsson 2007-2010
+ * License terms: GNU General Public License (GPL) version 2
+ * Author: Per Friden <per.friden@stericsson.com>
+ * Author: Jonas Aaberg <jonas.aberg@stericsson.com>
+ *
+ */
+
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/dmaengine.h>
+#include <linux/platform_device.h>
+#include <linux/clk.h>
+#include <linux/delay.h>
+
+#include <plat/ste_dma40.h>
+
+#include "ste_dma40_ll.h"
+
+#define D40_NAME "dma40"
+
+#define D40_PHY_CHAN -1
+
+/* For masking out/in 2 bit channel positions */
+#define D40_CHAN_POS(chan)  (2 * (chan / 2))
+#define D40_CHAN_POS_MASK(chan) (0x3 << D40_CHAN_POS(chan))
+
+/* Maximum iterations taken before giving up suspending a channel */
+#define D40_SUSPEND_MAX_IT 500
+
+#define D40_ALLOC_FREE		(1 << 31)
+#define D40_ALLOC_PHY		(1 << 30)
+#define D40_ALLOC_LOG_FREE	0
+
+/* The number of free d40_desc to keep in memory before starting
+ * to kfree() them */
+#define D40_DESC_CACHE_SIZE 50
+
+/* Hardware designer of the block */
+#define D40_PERIPHID2_DESIGNER 0x8
+
+/**
+ * enum 40_command - The different commands and/or statuses.
+ *
+ * @D40_DMA_STOP: DMA channel command STOP or status STOPPED,
+ * @D40_DMA_RUN: The DMA channel is RUNNING of the command RUN.
+ * @D40_DMA_SUSPEND_REQ: Request the DMA to SUSPEND as soon as possible.
+ * @D40_DMA_SUSPENDED: The DMA channel is SUSPENDED.
+ */
+enum d40_command {
+	D40_DMA_STOP		= 0,
+	D40_DMA_RUN		= 1,
+	D40_DMA_SUSPEND_REQ	= 2,
+	D40_DMA_SUSPENDED	= 3
+};
+
+/**
+ * struct d40_lli_pool - Structure for keeping LLIs in memory
+ *
+ * @base: Pointer to memory area when the pre_alloc_lli's are not large
+ * enough, IE bigger than the most common case, 1 dst and 1 src. NULL if
+ * pre_alloc_lli is used.
+ * @size: The size in bytes of the memory at base or the size of pre_alloc_lli.
+ * @pre_alloc_lli: Pre allocated area for the most common case of transfers,
+ * one buffer to one buffer.
+ */
+struct d40_lli_pool {
+	void	*base;
+	int	size;
+	/* Space for dst and src, plus an extra for padding */
+	u8	pre_alloc_lli[3 * sizeof(struct d40_phy_lli)];
+};
+
+/**
+ * struct d40_desc - A descriptor is one DMA job.
+ *
+ * @lli_phy: LLI settings for physical channel. Both src and dst=
+ * points into the lli_pool, to base if lli_len > 1 or to pre_alloc_lli if
+ * lli_len equals one.
+ * @lli_log: Same as above but for logical channels.
+ * @lli_pool: The pool with two entries pre-allocated.
+ * @lli_len: Number of LLI's in lli_pool
+ * @lli_tcount: Number of LLIs processed in the transfer. When equals lli_len
+ * then this transfer job is done.
+ * @txd: DMA engine struct. Used for among other things for communication
+ * during a transfer.
+ * @node: List entry.
+ * @dir: The transfer direction of this job.
+ * @is_in_client_list: true if the client owns this descriptor.
+ *
+ * This descriptor is used for both logical and physical transfers.
+ */
+
+struct d40_desc {
+	/* LLI physical */
+	struct d40_phy_lli_bidir	 lli_phy;
+	/* LLI logical */
+	struct d40_log_lli_bidir	 lli_log;
+
+	struct d40_lli_pool		 lli_pool;
+	u32				 lli_len;
+	u32				 lli_tcount;
+
+	struct dma_async_tx_descriptor	 txd;
+	struct list_head		 node;
+
+	enum dma_data_direction		 dir;
+	bool				 is_in_client_list;
+};
+
+/**
+ * struct d40_lcla_pool - LCLA pool settings and data.
+ *
+ * @base: The virtual address of LCLA.
+ * @phy: Physical base address of LCLA.
+ * @base_size: size of lcla.
+ * @lock: Lock to protect the content in this struct.
+ * @alloc_map: Mapping between physical channel and LCLA entries.
+ * @num_blocks: The number of entries of alloc_map. Equals to the
+ * number of physical channels.
+ */
+struct d40_lcla_pool {
+	void		*base;
+	dma_addr_t	 phy;
+	resource_size_t  base_size;
+	spinlock_t	 lock;
+	u32		*alloc_map;
+	int		 num_blocks;
+};
+
+/**
+ * struct d40_phy_res - struct for handling eventlines mapped to physical
+ * channels.
+ *
+ * @lock: A lock protection this entity.
+ * @num: The physical channel number of this entity.
+ * @allocated_src: Bit mapped to show which src event line's are mapped to
+ * this physical channel. Can also be free or physically allocated.
+ * @allocated_dst: Same as for src but is dst.
+ * allocated_dst and allocated_src uses the D40_ALLOC* defines as well as
+ * event line number. Both allocated_src and allocated_dst can not be
+ * allocated to a physical channel, since the interrupt handler has then
+ * no way of figure out which one the interrupt belongs to.
+ */
+struct d40_phy_res {
+	spinlock_t lock;
+	int	   num;
+	u32	   allocated_src;
+	u32	   allocated_dst;
+};
+
+struct d40_base;
+
+/**
+ * struct d40_chan - Struct that describes a channel.
+ *
+ * @lock: A spinlock to protect this struct.
+ * @log_num: The logical number, if any of this channel.
+ * @completed: Starts with 1, after first interrupt it is set to dma engine's
+ * current cookie.
+ * @pending_tx: The number of pending transfers. Used between interrupt handler
+ * and tasklet.
+ * @busy: Set to true when transfer is ongoing on this channel.
+ * @phy_chan: Pointer to physical channel which this instance runs on.
+ * @chan: DMA engine handle.
+ * @tasklet: Tasklet that gets scheduled from interrupt context to complete a
+ * transfer and call client callback.
+ * @client: Cliented owned descriptor list.
+ * @active: Active descriptor.
+ * @queue: Queued jobs.
+ * @free: List of free descripts, ready to be reused.
+ * @free_len: Number of descriptors in the free list.
+ * @dma_cfg: The client configuration of this dma channel.
+ * @base: Pointer to the device instance struct.
+ * @src_def_cfg: Default cfg register setting for src.
+ * @dst_def_cfg: Default cfg register setting for dst.
+ * @log_def: Default logical channel settings.
+ * @lcla: Space for one dst src pair for logical channel transfers.
+ * @lcpa: Pointer to dst and src lcpa settings.
+ *
+ * This struct can either "be" a logical or a physical channel.
+ */
+struct d40_chan {
+	spinlock_t			 lock;
+	int				 log_num;
+	/* ID of the most recent completed transfer */
+	int				 completed;
+	int				 pending_tx;
+	bool				 busy;
+	struct d40_phy_res		*phy_chan;
+	struct dma_chan			 chan;
+	struct tasklet_struct		 tasklet;
+	struct list_head		 client;
+	struct list_head		 active;
+	struct list_head		 queue;
+	struct list_head		 free;
+	int				 free_len;
+	struct stedma40_chan_cfg	 dma_cfg;
+	struct d40_base			*base;
+	/* Default register configurations */
+	u32				 src_def_cfg;
+	u32				 dst_def_cfg;
+	struct d40_def_lcsp		 log_def;
+	struct d40_lcla_elem		 lcla;
+	struct d40_log_lli_full		*lcpa;
+};
+
+/**
+ * struct d40_base - The big global struct, one for each probe'd instance.
+ *
+ * @interrupt_lock: Lock used to make sure one interrupt is handle a time.
+ * @execmd_lock: Lock for execute command usage since several channels share
+ * the same physical register.
+ * @dev: The device structure.
+ * @virtbase: The virtual base address of the DMA's register.
+ * @clk: Pointer to the DMA clock structure.
+ * @phy_start: Physical memory start of the DMA registers.
+ * @phy_size: Size of the DMA register map.
+ * @irq: The IRQ number.
+ * @num_phy_chans: The number of physical channels. Read from HW. This
+ * is the number of available channels for this driver, not counting "Secure
+ * mode" allocated physical channels.
+ * @num_log_chans: The number of logical channels. Calculated from
+ * num_phy_chans.
+ * @dma_both: dma_device channels that can do both memcpy and slave transfers.
+ * @dma_slave: dma_device channels that can do only do slave transfers.
+ * @dma_memcpy: dma_device channels that can do only do memcpy transfers.
+ * @phy_chans: Room for all possible physical channels in system.
+ * @log_chans: Room for all possible logical channels in system.
+ * @lookup_log_chans: Used to map interrupt number to logical channel. Points
+ * to log_chans entries.
+ * @lookup_phy_chans: Used to map interrupt number to physical channel. Points
+ * to phy_chans entries.
+ * @plat_data: Pointer to provided platform_data which is the driver
+ * configuration.
+ * @phy_res: Vector containing all physical channels.
+ * @lcla_pool: lcla pool settings and data.
+ * @lcpa_base: The virtual mapped address of LCPA.
+ * @phy_lcpa: The physical address of the LCPA.
+ * @lcpa_size: The size of the LCPA area.
+ */
+struct d40_base {
+	spinlock_t			 interrupt_lock;
+	spinlock_t			 execmd_lock;
+	struct device			 *dev;
+	void __iomem			 *virtbase;
+	struct clk			 *clk;
+	phys_addr_t			  phy_start;
+	resource_size_t			  phy_size;
+	int				  irq;
+	int				  num_phy_chans;
+	int				  num_log_chans;
+	struct dma_device		  dma_both;
+	struct dma_device		  dma_slave;
+	struct dma_device		  dma_memcpy;
+	struct d40_chan			 *phy_chans;
+	struct d40_chan			 *log_chans;
+	struct d40_chan			**lookup_log_chans;
+	struct d40_chan			**lookup_phy_chans;
+	struct stedma40_platform_data	 *plat_data;
+	/* Physical half channels */
+	struct d40_phy_res		 *phy_res;
+	struct d40_lcla_pool		  lcla_pool;
+	void				 *lcpa_base;
+	dma_addr_t			  phy_lcpa;
+	resource_size_t			  lcpa_size;
+};
+
+/**
+ * struct d40_interrupt_lookup - lookup table for interrupt handler
+ *
+ * @src: Interrupt mask register.
+ * @clr: Interrupt clear register.
+ * @is_error: true if this is an error interrupt.
+ * @offset: start delta in the lookup_log_chans in d40_base. If equals to
+ * D40_PHY_CHAN, the lookup_phy_chans shall be used instead.
+ */
+struct d40_interrupt_lookup {
+	u32 src;
+	u32 clr;
+	bool is_error;
+	int offset;
+};
+
+/**
+ * struct d40_reg_val - simple lookup struct
+ *
+ * @reg: The register.
+ * @val: The value that belongs to the register in reg.
+ */
+struct d40_reg_val {
+	unsigned int reg;
+	unsigned int val;
+};
+
+static int d40_pool_lli_alloc(struct d40_desc *d40d,
+			      int lli_len, bool is_log)
+{
+	u32 align;
+	void *base;
+
+	if (is_log)
+		align = sizeof(struct d40_log_lli);
+	else
+		align = sizeof(struct d40_phy_lli);
+
+	if (lli_len == 1) {
+		base = d40d->lli_pool.pre_alloc_lli;
+		d40d->lli_pool.size = sizeof(d40d->lli_pool.pre_alloc_lli);
+		d40d->lli_pool.base = NULL;
+	} else {
+		d40d->lli_pool.size = ALIGN(lli_len * 2 * align, align);
+
+		base = kmalloc(d40d->lli_pool.size + align, GFP_NOWAIT);
+		d40d->lli_pool.base = base;
+
+		if (d40d->lli_pool.base == NULL)
+			return -ENOMEM;
+	}
+
+	if (is_log) {
+		d40d->lli_log.src = PTR_ALIGN((struct d40_log_lli *) base,
+					      align);
+		d40d->lli_log.dst = PTR_ALIGN(d40d->lli_log.src + lli_len,
+					      align);
+	} else {
+		d40d->lli_phy.src = PTR_ALIGN((struct d40_phy_lli *)base,
+					      align);
+		d40d->lli_phy.dst = PTR_ALIGN(d40d->lli_phy.src + lli_len,
+					      align);
+
+		d40d->lli_phy.src_addr = virt_to_phys(d40d->lli_phy.src);
+		d40d->lli_phy.dst_addr = virt_to_phys(d40d->lli_phy.dst);
+	}
+
+	return 0;
+}
+
+static void d40_pool_lli_free(struct d40_desc *d40d)
+{
+	kfree(d40d->lli_pool.base);
+	d40d->lli_pool.base = NULL;
+	d40d->lli_pool.size = 0;
+	d40d->lli_log.src = NULL;
+	d40d->lli_log.dst = NULL;
+	d40d->lli_phy.src = NULL;
+	d40d->lli_phy.dst = NULL;
+	d40d->lli_phy.src_addr = 0;
+	d40d->lli_phy.dst_addr = 0;
+}
+
+static dma_cookie_t d40_assign_cookie(struct d40_chan *d40c,
+				      struct d40_desc *desc)
+{
+	dma_cookie_t cookie = d40c->chan.cookie;
+
+	if (++cookie < 0)
+		cookie = 1;
+
+	d40c->chan.cookie = cookie;
+	desc->txd.cookie = cookie;
+
+	return cookie;
+}
+
+static void d40_desc_reset(struct d40_desc *d40d)
+{
+	d40d->lli_tcount = 0;
+}
+
+static void d40_desc_remove(struct d40_desc *d40d)
+{
+	list_del(&d40d->node);
+}
+
+static struct d40_desc *d40_desc_get(struct d40_chan *d40c)
+{
+	struct d40_desc *desc;
+	struct d40_desc *d;
+	struct d40_desc *_d;
+
+	if (!list_empty(&d40c->client)) {
+		list_for_each_entry_safe(d, _d, &d40c->client, node)
+			if (async_tx_test_ack(&d->txd)) {
+				d40_pool_lli_free(d);
+				d40_desc_remove(d);
+				desc = d;
+				goto out;
+			}
+	}
+
+	if (list_empty(&d40c->free)) {
+		/* Alloc new desc because we're out of used ones */
+		desc = kzalloc(sizeof(struct d40_desc), GFP_NOWAIT);
+		if (desc == NULL)
+			goto out;
+		INIT_LIST_HEAD(&desc->node);
+	} else {
+		/* Reuse an old desc. */
+		desc = list_first_entry(&d40c->free,
+					struct d40_desc,
+					node);
+		list_del(&desc->node);
+		d40c->free_len--;
+	}
+out:
+	return desc;
+}
+
+static void d40_desc_free(struct d40_chan *d40c, struct d40_desc *d40d)
+{
+	if (d40c->free_len < D40_DESC_CACHE_SIZE) {
+		list_add_tail(&d40d->node, &d40c->free);
+		d40c->free_len++;
+	} else
+		kfree(d40d);
+}
+
+static void d40_desc_submit(struct d40_chan *d40c, struct d40_desc *desc)
+{
+	list_add_tail(&desc->node, &d40c->active);
+}
+
+static struct d40_desc *d40_first_active_get(struct d40_chan *d40c)
+{
+	struct d40_desc *d;
+
+	if (list_empty(&d40c->active))
+		return NULL;
+
+	d = list_first_entry(&d40c->active,
+			     struct d40_desc,
+			     node);
+	return d;
+}
+
+static void d40_desc_queue(struct d40_chan *d40c, struct d40_desc *desc)
+{
+	list_add_tail(&desc->node, &d40c->queue);
+}
+
+static struct d40_desc *d40_first_queued(struct d40_chan *d40c)
+{
+	struct d40_desc *d;
+
+	if (list_empty(&d40c->queue))
+		return NULL;
+
+	d = list_first_entry(&d40c->queue,
+			     struct d40_desc,
+			     node);
+	return d;
+}
+
+/* Support functions for logical channels */
+
+static int d40_lcla_id_get(struct d40_chan *d40c,
+			   struct d40_lcla_pool *pool)
+{
+	int src_id = 0;
+	int dst_id = 0;
+	struct d40_log_lli *lcla_lidx_base =
+		pool->base + d40c->phy_chan->num * 1024;
+	int i;
+	int lli_per_log = d40c->base->plat_data->llis_per_log;
+
+	if (d40c->lcla.src_id >= 0 && d40c->lcla.dst_id >= 0)
+		return 0;
+
+	if (pool->num_blocks > 32)
+		return -EINVAL;
+
+	spin_lock(&pool->lock);
+
+	for (i = 0; i < pool->num_blocks; i++) {
+		if (!(pool->alloc_map[d40c->phy_chan->num] & (0x1 << i))) {
+			pool->alloc_map[d40c->phy_chan->num] |= (0x1 << i);
+			break;
+		}
+	}
+	src_id = i;
+	if (src_id >= pool->num_blocks)
+		goto err;
+
+	for (; i < pool->num_blocks; i++) {
+		if (!(pool->alloc_map[d40c->phy_chan->num] & (0x1 << i))) {
+			pool->alloc_map[d40c->phy_chan->num] |= (0x1 << i);
+			break;
+		}
+	}
+
+	dst_id = i;
+	if (dst_id == src_id)
+		goto err;
+
+	d40c->lcla.src_id = src_id;
+	d40c->lcla.dst_id = dst_id;
+	d40c->lcla.dst = lcla_lidx_base + dst_id * lli_per_log + 1;
+	d40c->lcla.src = lcla_lidx_base + src_id * lli_per_log + 1;
+
+
+	spin_unlock(&pool->lock);
+	return 0;
+err:
+	spin_unlock(&pool->lock);
+	return -EINVAL;
+}
+
+static void d40_lcla_id_put(struct d40_chan *d40c,
+			    struct d40_lcla_pool *pool,
+			    int id)
+{
+	if (id < 0)
+		return;
+
+	d40c->lcla.src_id = -1;
+	d40c->lcla.dst_id = -1;
+
+	spin_lock(&pool->lock);
+	pool->alloc_map[d40c->phy_chan->num] &= (~(0x1 << id));
+	spin_unlock(&pool->lock);
+}
+
+static int d40_channel_execute_command(struct d40_chan *d40c,
+				       enum d40_command command)
+{
+	int status, i;
+	void __iomem *active_reg;
+	int ret = 0;
+	unsigned long flags;
+
+	spin_lock_irqsave(&d40c->base->execmd_lock, flags);
+
+	if (d40c->phy_chan->num % 2 == 0)
+		active_reg = d40c->base->virtbase + D40_DREG_ACTIVE;
+	else
+		active_reg = d40c->base->virtbase + D40_DREG_ACTIVO;
+
+	if (command == D40_DMA_SUSPEND_REQ) {
+		status = (readl(active_reg) &
+			  D40_CHAN_POS_MASK(d40c->phy_chan->num)) >>
+			D40_CHAN_POS(d40c->phy_chan->num);
+
+		if (status == D40_DMA_SUSPENDED || status == D40_DMA_STOP)
+			goto done;
+	}
+
+	writel(command << D40_CHAN_POS(d40c->phy_chan->num), active_reg);
+
+	if (command == D40_DMA_SUSPEND_REQ) {
+
+		for (i = 0 ; i < D40_SUSPEND_MAX_IT; i++) {
+			status = (readl(active_reg) &
+				  D40_CHAN_POS_MASK(d40c->phy_chan->num)) >>
+				D40_CHAN_POS(d40c->phy_chan->num);
+
+			cpu_relax();
+			/*
+			 * Reduce the number of bus accesses while
+			 * waiting for the DMA to suspend.
+			 */
+			udelay(3);
+
+			if (status == D40_DMA_STOP ||
+			    status == D40_DMA_SUSPENDED)
+				break;
+		}
+
+		if (i == D40_SUSPEND_MAX_IT) {
+			dev_err(&d40c->chan.dev->device,
+				"[%s]: unable to suspend the chl %d (log: %d) status %x\n",
+				__func__, d40c->phy_chan->num, d40c->log_num,
+				status);
+			dump_stack();
+			ret = -EBUSY;
+		}
+
+	}
+done:
+	spin_unlock_irqrestore(&d40c->base->execmd_lock, flags);
+	return ret;
+}
+
+static void d40_term_all(struct d40_chan *d40c)
+{
+	struct d40_desc *d40d;
+	struct d40_desc *d;
+	struct d40_desc *_d;
+
+	/* Release active descriptors */
+	while ((d40d = d40_first_active_get(d40c))) {
+		d40_desc_remove(d40d);
+
+		/* Return desc to free-list */
+		d40_desc_free(d40c, d40d);
+	}
+
+	/* Release queued descriptors waiting for transfer */
+	while ((d40d = d40_first_queued(d40c))) {
+		d40_desc_remove(d40d);
+
+		/* Return desc to free-list */
+		d40_desc_free(d40c, d40d);
+	}
+
+	/* Release client owned descriptors */
+	if (!list_empty(&d40c->client))
+		list_for_each_entry_safe(d, _d, &d40c->client, node) {
+			d40_pool_lli_free(d);
+			d40_desc_remove(d);
+			/* Return desc to free-list */
+			d40_desc_free(d40c, d40d);
+		}
+
+	d40_lcla_id_put(d40c, &d40c->base->lcla_pool,
+			d40c->lcla.src_id);
+	d40_lcla_id_put(d40c, &d40c->base->lcla_pool,
+			d40c->lcla.dst_id);
+
+	d40c->pending_tx = 0;
+	d40c->busy = false;
+}
+
+static void d40_config_set_event(struct d40_chan *d40c, bool do_enable)
+{
+	u32 val;
+	unsigned long flags;
+
+	if (do_enable)
+		val = D40_ACTIVATE_EVENTLINE;
+	else
+		val = D40_DEACTIVATE_EVENTLINE;
+
+	spin_lock_irqsave(&d40c->phy_chan->lock, flags);
+
+	/* Enable event line connected to device (or memcpy) */
+	if ((d40c->dma_cfg.dir ==  STEDMA40_PERIPH_TO_MEM) ||
+	    (d40c->dma_cfg.dir == STEDMA40_PERIPH_TO_PERIPH)) {
+		u32 event = D40_TYPE_TO_EVENT(d40c->dma_cfg.src_dev_type);
+
+		writel((val << D40_EVENTLINE_POS(event)) |
+		       ~D40_EVENTLINE_MASK(event),
+		       d40c->base->virtbase + D40_DREG_PCBASE +
+		       d40c->phy_chan->num * D40_DREG_PCDELTA +
+		       D40_CHAN_REG_SSLNK);
+	}
+	if (d40c->dma_cfg.dir !=  STEDMA40_PERIPH_TO_MEM) {
+		u32 event = D40_TYPE_TO_EVENT(d40c->dma_cfg.dst_dev_type);
+
+		writel((val << D40_EVENTLINE_POS(event)) |
+		       ~D40_EVENTLINE_MASK(event),
+		       d40c->base->virtbase + D40_DREG_PCBASE +
+		       d40c->phy_chan->num * D40_DREG_PCDELTA +
+		       D40_CHAN_REG_SDLNK);
+	}
+
+	spin_unlock_irqrestore(&d40c->phy_chan->lock, flags);
+}
+
+static u32 d40_chan_has_events(struct d40_chan *d40c)
+{
+	u32 val = 0;
+
+	/* If SSLNK or SDLNK is zero all events are disabled */
+	if ((d40c->dma_cfg.dir ==  STEDMA40_PERIPH_TO_MEM) ||
+	    (d40c->dma_cfg.dir == STEDMA40_PERIPH_TO_PERIPH))
+		val = readl(d40c->base->virtbase + D40_DREG_PCBASE +
+			    d40c->phy_chan->num * D40_DREG_PCDELTA +
+			    D40_CHAN_REG_SSLNK);
+
+	if (d40c->dma_cfg.dir !=  STEDMA40_PERIPH_TO_MEM)
+		val = readl(d40c->base->virtbase + D40_DREG_PCBASE +
+			    d40c->phy_chan->num * D40_DREG_PCDELTA +
+			    D40_CHAN_REG_SDLNK);
+	return val;
+}
+
+static void d40_config_enable_lidx(struct d40_chan *d40c)
+{
+	/* Set LIDX for lcla */
+	writel((d40c->phy_chan->num << D40_SREG_ELEM_LOG_LIDX_POS) &
+	       D40_SREG_ELEM_LOG_LIDX_MASK,
+	       d40c->base->virtbase + D40_DREG_PCBASE +
+	       d40c->phy_chan->num * D40_DREG_PCDELTA + D40_CHAN_REG_SDELT);
+
+	writel((d40c->phy_chan->num << D40_SREG_ELEM_LOG_LIDX_POS) &
+	       D40_SREG_ELEM_LOG_LIDX_MASK,
+	       d40c->base->virtbase + D40_DREG_PCBASE +
+	       d40c->phy_chan->num * D40_DREG_PCDELTA + D40_CHAN_REG_SSELT);
+}
+
+static int d40_config_write(struct d40_chan *d40c)
+{
+	u32 addr_base;
+	u32 var;
+	int res;
+
+	res = d40_channel_execute_command(d40c, D40_DMA_SUSPEND_REQ);
+	if (res)
+		return res;
+
+	/* Odd addresses are even addresses + 4 */
+	addr_base = (d40c->phy_chan->num % 2) * 4;
+	/* Setup channel mode to logical or physical */
+	var = ((u32)(d40c->log_num != D40_PHY_CHAN) + 1) <<
+		D40_CHAN_POS(d40c->phy_chan->num);
+	writel(var, d40c->base->virtbase + D40_DREG_PRMSE + addr_base);
+
+	/* Setup operational mode option register */
+	var = ((d40c->dma_cfg.channel_type >> STEDMA40_INFO_CH_MODE_OPT_POS) &
+	       0x3) << D40_CHAN_POS(d40c->phy_chan->num);
+
+	writel(var, d40c->base->virtbase + D40_DREG_PRMOE + addr_base);
+
+	if (d40c->log_num != D40_PHY_CHAN) {
+		/* Set default config for CFG reg */
+		writel(d40c->src_def_cfg,
+		       d40c->base->virtbase + D40_DREG_PCBASE +
+		       d40c->phy_chan->num * D40_DREG_PCDELTA +
+		       D40_CHAN_REG_SSCFG);
+		writel(d40c->dst_def_cfg,
+		       d40c->base->virtbase + D40_DREG_PCBASE +
+		       d40c->phy_chan->num * D40_DREG_PCDELTA +
+		       D40_CHAN_REG_SDCFG);
+
+		d40_config_enable_lidx(d40c);
+	}
+	return res;
+}
+
+static void d40_desc_load(struct d40_chan *d40c, struct d40_desc *d40d)
+{
+
+	if (d40d->lli_phy.dst && d40d->lli_phy.src) {
+		d40_phy_lli_write(d40c->base->virtbase,
+				  d40c->phy_chan->num,
+				  d40d->lli_phy.dst,
+				  d40d->lli_phy.src);
+		d40d->lli_tcount = d40d->lli_len;
+	} else if (d40d->lli_log.dst && d40d->lli_log.src) {
+		u32 lli_len;
+		struct d40_log_lli *src = d40d->lli_log.src;
+		struct d40_log_lli *dst = d40d->lli_log.dst;
+
+		src += d40d->lli_tcount;
+		dst += d40d->lli_tcount;
+
+		if (d40d->lli_len <= d40c->base->plat_data->llis_per_log)
+			lli_len = d40d->lli_len;
+		else
+			lli_len = d40c->base->plat_data->llis_per_log;
+		d40d->lli_tcount += lli_len;
+		d40_log_lli_write(d40c->lcpa, d40c->lcla.src,
+				  d40c->lcla.dst,
+				  dst, src,
+				  d40c->base->plat_data->llis_per_log);
+	}
+}
+
+static dma_cookie_t d40_tx_submit(struct dma_async_tx_descriptor *tx)
+{
+	struct d40_chan *d40c = container_of(tx->chan,
+					     struct d40_chan,
+					     chan);
+	struct d40_desc *d40d = container_of(tx, struct d40_desc, txd);
+	unsigned long flags;
+
+	spin_lock_irqsave(&d40c->lock, flags);
+
+	tx->cookie = d40_assign_cookie(d40c, d40d);
+
+	d40_desc_queue(d40c, d40d);
+
+	spin_unlock_irqrestore(&d40c->lock, flags);
+
+	return tx->cookie;
+}
+
+static int d40_start(struct d40_chan *d40c)
+{
+	int err;
+
+	if (d40c->log_num != D40_PHY_CHAN) {
+		err = d40_channel_execute_command(d40c, D40_DMA_SUSPEND_REQ);
+		if (err)
+			return err;
+		d40_config_set_event(d40c, true);
+	}
+
+	err = d40_channel_execute_command(d40c, D40_DMA_RUN);
+
+	return err;
+}
+
+static struct d40_desc *d40_queue_start(struct d40_chan *d40c)
+{
+	struct d40_desc *d40d;
+	int err;
+
+	/* Start queued jobs, if any */
+	d40d = d40_first_queued(d40c);
+
+	if (d40d != NULL) {
+		d40c->busy = true;
+
+		/* Remove from queue */
+		d40_desc_remove(d40d);
+
+		/* Add to active queue */
+		d40_desc_submit(d40c, d40d);
+
+		/* Initiate DMA job */
+		d40_desc_load(d40c, d40d);
+
+		/* Start dma job */
+		err = d40_start(d40c);
+
+		if (err)
+			return NULL;
+	}
+
+	return d40d;
+}
+
+/* called from interrupt context */
+static void dma_tc_handle(struct d40_chan *d40c)
+{
+	struct d40_desc *d40d;
+
+	if (!d40c->phy_chan)
+		return;
+
+	/* Get first active entry from list */
+	d40d = d40_first_active_get(d40c);
+
+	if (d40d == NULL)
+		return;
+
+	if (d40d->lli_tcount < d40d->lli_len) {
+
+		d40_desc_load(d40c, d40d);
+		/* Start dma job */
+		(void) d40_start(d40c);
+		return;
+	}
+
+	if (d40_queue_start(d40c) == NULL)
+		d40c->busy = false;
+
+	d40c->pending_tx++;
+	tasklet_schedule(&d40c->tasklet);
+
+}
+
+static void dma_tasklet(unsigned long data)
+{
+	struct d40_chan *d40c = (struct d40_chan *) data;
+	struct d40_desc *d40d_fin;
+	unsigned long flags;
+	dma_async_tx_callback callback;
+	void *callback_param;
+
+	spin_lock_irqsave(&d40c->lock, flags);
+
+	/* Get first active entry from list */
+	d40d_fin = d40_first_active_get(d40c);
+
+	if (d40d_fin == NULL)
+		goto err;
+
+	d40c->completed = d40d_fin->txd.cookie;
+
+	/*
+	 * If terminating a channel pending_tx is set to zero.
+	 * This prevents any finished active jobs to return to the client.
+	 */
+	if (d40c->pending_tx == 0) {
+		spin_unlock_irqrestore(&d40c->lock, flags);
+		return;
+	}
+
+	/* Callback to client */
+	callback = d40d_fin->txd.callback;
+	callback_param = d40d_fin->txd.callback_param;
+
+	if (async_tx_test_ack(&d40d_fin->txd)) {
+		d40_pool_lli_free(d40d_fin);
+		d40_desc_remove(d40d_fin);
+		/* Return desc to free-list */
+		d40_desc_free(d40c, d40d_fin);
+	} else {
+		d40_desc_reset(d40d_fin);
+		if (!d40d_fin->is_in_client_list) {
+			d40_desc_remove(d40d_fin);
+			list_add_tail(&d40d_fin->node, &d40c->client);
+			d40d_fin->is_in_client_list = true;
+		}
+	}
+
+	d40c->pending_tx--;
+
+	if (d40c->pending_tx)
+		tasklet_schedule(&d40c->tasklet);
+
+	spin_unlock_irqrestore(&d40c->lock, flags);
+
+	if (callback)
+		callback(callback_param);
+
+	return;
+
+ err:
+	/* Rescue manouver if receiving double interrupts */
+	if (d40c->pending_tx > 0)
+		d40c->pending_tx--;
+	spin_unlock_irqrestore(&d40c->lock, flags);
+}
+
+static irqreturn_t d40_handle_interrupt(int irq, void *data)
+{
+	static const struct d40_interrupt_lookup il[] = {
+		{D40_DREG_LCTIS0, D40_DREG_LCICR0, false,  0},
+		{D40_DREG_LCTIS1, D40_DREG_LCICR1, false, 32},
+		{D40_DREG_LCTIS2, D40_DREG_LCICR2, false, 64},
+		{D40_DREG_LCTIS3, D40_DREG_LCICR3, false, 96},
+		{D40_DREG_LCEIS0, D40_DREG_LCICR0, true,   0},
+		{D40_DREG_LCEIS1, D40_DREG_LCICR1, true,  32},
+		{D40_DREG_LCEIS2, D40_DREG_LCICR2, true,  64},
+		{D40_DREG_LCEIS3, D40_DREG_LCICR3, true,  96},
+		{D40_DREG_PCTIS,  D40_DREG_PCICR,  false, D40_PHY_CHAN},
+		{D40_DREG_PCEIS,  D40_DREG_PCICR,  true,  D40_PHY_CHAN},
+	};
+
+	int i;
+	u32 regs[ARRAY_SIZE(il)];
+	u32 tmp;
+	u32 idx;
+	u32 row;
+	long chan = -1;
+	struct d40_chan *d40c;
+	unsigned long flags;
+	struct d40_base *base = data;
+
+	spin_lock_irqsave(&base->interrupt_lock, flags);
+
+	/* Read interrupt status of both logical and physical channels */
+	for (i = 0; i < ARRAY_SIZE(il); i++)
+		regs[i] = readl(base->virtbase + il[i].src);
+
+	for (;;) {
+
+		chan = find_next_bit((unsigned long *)regs,
+				     BITS_PER_LONG * ARRAY_SIZE(il), chan + 1);
+
+		/* No more set bits found? */
+		if (chan == BITS_PER_LONG * ARRAY_SIZE(il))
+			break;
+
+		row = chan / BITS_PER_LONG;
+		idx = chan & (BITS_PER_LONG - 1);
+
+		/* ACK interrupt */
+		tmp = readl(base->virtbase + il[row].clr);
+		tmp |= 1 << idx;
+		writel(tmp, base->virtbase + il[row].clr);
+
+		if (il[row].offset == D40_PHY_CHAN)
+			d40c = base->lookup_phy_chans[idx];
+		else
+			d40c = base->lookup_log_chans[il[row].offset + idx];
+		spin_lock(&d40c->lock);
+
+		if (!il[row].is_error)
+			dma_tc_handle(d40c);
+		else
+			dev_err(base->dev, "[%s] IRQ chan: %ld offset %d idx %d\n",
+				__func__, chan, il[row].offset, idx);
+
+		spin_unlock(&d40c->lock);
+	}
+
+	spin_unlock_irqrestore(&base->interrupt_lock, flags);
+
+	return IRQ_HANDLED;
+}
+
+
+static int d40_validate_conf(struct d40_chan *d40c,
+			     struct stedma40_chan_cfg *conf)
+{
+	int res = 0;
+	u32 dst_event_group = D40_TYPE_TO_GROUP(conf->dst_dev_type);
+	u32 src_event_group = D40_TYPE_TO_GROUP(conf->src_dev_type);
+	bool is_log = (conf->channel_type & STEDMA40_CHANNEL_IN_OPER_MODE)
+		== STEDMA40_CHANNEL_IN_LOG_MODE;
+
+	if (d40c->dma_cfg.dir == STEDMA40_MEM_TO_PERIPH &&
+	    dst_event_group == STEDMA40_DEV_DST_MEMORY) {
+		dev_err(&d40c->chan.dev->device, "[%s] Invalid dst\n",
+			__func__);
+		res = -EINVAL;
+	}
+
+	if (d40c->dma_cfg.dir == STEDMA40_PERIPH_TO_MEM &&
+	    src_event_group == STEDMA40_DEV_SRC_MEMORY) {
+		dev_err(&d40c->chan.dev->device, "[%s] Invalid src\n",
+			__func__);
+		res = -EINVAL;
+	}
+
+	if (src_event_group == STEDMA40_DEV_SRC_MEMORY &&
+	    dst_event_group == STEDMA40_DEV_DST_MEMORY && is_log) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] No event line\n", __func__);
+		res = -EINVAL;
+	}
+
+	if (conf->dir == STEDMA40_PERIPH_TO_PERIPH &&
+	    (src_event_group != dst_event_group)) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Invalid event group\n", __func__);
+		res = -EINVAL;
+	}
+
+	if (conf->dir == STEDMA40_PERIPH_TO_PERIPH) {
+		/*
+		 * DMAC HW supports it. Will be added to this driver,
+		 * in case any dma client requires it.
+		 */
+		dev_err(&d40c->chan.dev->device,
+			"[%s] periph to periph not supported\n",
+			__func__);
+		res = -EINVAL;
+	}
+
+	return res;
+}
+
+static bool d40_alloc_mask_set(struct d40_phy_res *phy, bool is_src,
+			       int log_event_line, bool is_log)
+{
+	unsigned long flags;
+	spin_lock_irqsave(&phy->lock, flags);
+	if (!is_log) {
+		/* Physical interrupts are masked per physical full channel */
+		if (phy->allocated_src == D40_ALLOC_FREE &&
+		    phy->allocated_dst == D40_ALLOC_FREE) {
+			phy->allocated_dst = D40_ALLOC_PHY;
+			phy->allocated_src = D40_ALLOC_PHY;
+			goto found;
+		} else
+			goto not_found;
+	}
+
+	/* Logical channel */
+	if (is_src) {
+		if (phy->allocated_src == D40_ALLOC_PHY)
+			goto not_found;
+
+		if (phy->allocated_src == D40_ALLOC_FREE)
+			phy->allocated_src = D40_ALLOC_LOG_FREE;
+
+		if (!(phy->allocated_src & (1 << log_event_line))) {
+			phy->allocated_src |= 1 << log_event_line;
+			goto found;
+		} else
+			goto not_found;
+	} else {
+		if (phy->allocated_dst == D40_ALLOC_PHY)
+			goto not_found;
+
+		if (phy->allocated_dst == D40_ALLOC_FREE)
+			phy->allocated_dst = D40_ALLOC_LOG_FREE;
+
+		if (!(phy->allocated_dst & (1 << log_event_line))) {
+			phy->allocated_dst |= 1 << log_event_line;
+			goto found;
+		} else
+			goto not_found;
+	}
+
+not_found:
+	spin_unlock_irqrestore(&phy->lock, flags);
+	return false;
+found:
+	spin_unlock_irqrestore(&phy->lock, flags);
+	return true;
+}
+
+static bool d40_alloc_mask_free(struct d40_phy_res *phy, bool is_src,
+			       int log_event_line)
+{
+	unsigned long flags;
+	bool is_free = false;
+
+	spin_lock_irqsave(&phy->lock, flags);
+	if (!log_event_line) {
+		/* Physical interrupts are masked per physical full channel */
+		phy->allocated_dst = D40_ALLOC_FREE;
+		phy->allocated_src = D40_ALLOC_FREE;
+		is_free = true;
+		goto out;
+	}
+
+	/* Logical channel */
+	if (is_src) {
+		phy->allocated_src &= ~(1 << log_event_line);
+		if (phy->allocated_src == D40_ALLOC_LOG_FREE)
+			phy->allocated_src = D40_ALLOC_FREE;
+	} else {
+		phy->allocated_dst &= ~(1 << log_event_line);
+		if (phy->allocated_dst == D40_ALLOC_LOG_FREE)
+			phy->allocated_dst = D40_ALLOC_FREE;
+	}
+
+	is_free = ((phy->allocated_src | phy->allocated_dst) ==
+		   D40_ALLOC_FREE);
+
+out:
+	spin_unlock_irqrestore(&phy->lock, flags);
+
+	return is_free;
+}
+
+static int d40_allocate_channel(struct d40_chan *d40c)
+{
+	int dev_type;
+	int event_group;
+	int event_line;
+	struct d40_phy_res *phys;
+	int i;
+	int j;
+	int log_num;
+	bool is_src;
+	bool is_log = (d40c->dma_cfg.channel_type & STEDMA40_CHANNEL_IN_OPER_MODE)
+		== STEDMA40_CHANNEL_IN_LOG_MODE;
+
+
+	phys = d40c->base->phy_res;
+
+	if (d40c->dma_cfg.dir == STEDMA40_PERIPH_TO_MEM) {
+		dev_type = d40c->dma_cfg.src_dev_type;
+		log_num = 2 * dev_type;
+		is_src = true;
+	} else if (d40c->dma_cfg.dir == STEDMA40_MEM_TO_PERIPH ||
+		   d40c->dma_cfg.dir == STEDMA40_MEM_TO_MEM) {
+		/* dst event lines are used for logical memcpy */
+		dev_type = d40c->dma_cfg.dst_dev_type;
+		log_num = 2 * dev_type + 1;
+		is_src = false;
+	} else
+		return -EINVAL;
+
+	event_group = D40_TYPE_TO_GROUP(dev_type);
+	event_line = D40_TYPE_TO_EVENT(dev_type);
+
+	if (!is_log) {
+		if (d40c->dma_cfg.dir == STEDMA40_MEM_TO_MEM) {
+			/* Find physical half channel */
+			for (i = 0; i < d40c->base->num_phy_chans; i++) {
+
+				if (d40_alloc_mask_set(&phys[i], is_src,
+						       0, is_log))
+					goto found_phy;
+			}
+		} else
+			for (j = 0; j < d40c->base->num_phy_chans; j += 8) {
+				int phy_num = j  + event_group * 2;
+				for (i = phy_num; i < phy_num + 2; i++) {
+					if (d40_alloc_mask_set(&phys[i], is_src,
+							       0, is_log))
+						goto found_phy;
+				}
+			}
+		return -EINVAL;
+found_phy:
+		d40c->phy_chan = &phys[i];
+		d40c->log_num = D40_PHY_CHAN;
+		goto out;
+	}
+	if (dev_type == -1)
+		return -EINVAL;
+
+	/* Find logical channel */
+	for (j = 0; j < d40c->base->num_phy_chans; j += 8) {
+		int phy_num = j + event_group * 2;
+		/*
+		 * Spread logical channels across all available physical rather
+		 * than pack every logical channel at the first available phy
+		 * channels.
+		 */
+		if (is_src) {
+			for (i = phy_num; i < phy_num + 2; i++) {
+				if (d40_alloc_mask_set(&phys[i], is_src,
+						       event_line, is_log))
+					goto found_log;
+			}
+		} else {
+			for (i = phy_num + 1; i >= phy_num; i--) {
+				if (d40_alloc_mask_set(&phys[i], is_src,
+						       event_line, is_log))
+					goto found_log;
+			}
+		}
+	}
+	return -EINVAL;
+
+found_log:
+	d40c->phy_chan = &phys[i];
+	d40c->log_num = log_num;
+out:
+
+	if (is_log)
+		d40c->base->lookup_log_chans[d40c->log_num] = d40c;
+	else
+		d40c->base->lookup_phy_chans[d40c->phy_chan->num] = d40c;
+
+	return 0;
+
+}
+
+static int d40_config_chan(struct d40_chan *d40c,
+			   struct stedma40_chan_cfg *info)
+{
+
+	/* Fill in basic CFG register values */
+	d40_phy_cfg(&d40c->dma_cfg, &d40c->src_def_cfg,
+		    &d40c->dst_def_cfg, d40c->log_num != D40_PHY_CHAN);
+
+	if (d40c->log_num != D40_PHY_CHAN) {
+		d40_log_cfg(&d40c->dma_cfg,
+			    &d40c->log_def.lcsp1, &d40c->log_def.lcsp3);
+
+		if (d40c->dma_cfg.dir == STEDMA40_PERIPH_TO_MEM)
+			d40c->lcpa = d40c->base->lcpa_base +
+				d40c->dma_cfg.src_dev_type * 32;
+		else
+			d40c->lcpa = d40c->base->lcpa_base +
+				d40c->dma_cfg.dst_dev_type * 32 + 16;
+	}
+
+	/* Write channel configuration to the DMA */
+	return d40_config_write(d40c);
+}
+
+static int d40_config_memcpy(struct d40_chan *d40c)
+{
+	dma_cap_mask_t cap = d40c->chan.device->cap_mask;
+
+	if (dma_has_cap(DMA_MEMCPY, cap) && !dma_has_cap(DMA_SLAVE, cap)) {
+		d40c->dma_cfg = *d40c->base->plat_data->memcpy_conf_log;
+		d40c->dma_cfg.src_dev_type = STEDMA40_DEV_SRC_MEMORY;
+		d40c->dma_cfg.dst_dev_type = d40c->base->plat_data->
+			memcpy[d40c->chan.chan_id];
+
+	} else if (dma_has_cap(DMA_MEMCPY, cap) &&
+		   dma_has_cap(DMA_SLAVE, cap)) {
+		d40c->dma_cfg = *d40c->base->plat_data->memcpy_conf_phy;
+	} else {
+		dev_err(&d40c->chan.dev->device, "[%s] No memcpy\n",
+			__func__);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+
+static int d40_free_dma(struct d40_chan *d40c)
+{
+
+	int res = 0;
+	u32 event, dir;
+	struct d40_phy_res *phy = d40c->phy_chan;
+	bool is_src;
+
+	/* Terminate all queued and active transfers */
+	d40_term_all(d40c);
+
+	if (phy == NULL) {
+		dev_err(&d40c->chan.dev->device, "[%s] phy == null\n",
+			__func__);
+		return -EINVAL;
+	}
+
+	if (phy->allocated_src == D40_ALLOC_FREE &&
+	    phy->allocated_dst == D40_ALLOC_FREE) {
+		dev_err(&d40c->chan.dev->device, "[%s] channel already free\n",
+			__func__);
+		return -EINVAL;
+	}
+
+
+	res = d40_channel_execute_command(d40c, D40_DMA_SUSPEND_REQ);
+	if (res) {
+		dev_err(&d40c->chan.dev->device, "[%s] suspend\n",
+			__func__);
+		return res;
+	}
+
+	if (d40c->dma_cfg.dir == STEDMA40_MEM_TO_PERIPH ||
+	    d40c->dma_cfg.dir == STEDMA40_MEM_TO_MEM) {
+		event = D40_TYPE_TO_EVENT(d40c->dma_cfg.dst_dev_type);
+		dir = D40_CHAN_REG_SDLNK;
+		is_src = false;
+	} else if (d40c->dma_cfg.dir == STEDMA40_PERIPH_TO_MEM) {
+		event = D40_TYPE_TO_EVENT(d40c->dma_cfg.src_dev_type);
+		dir = D40_CHAN_REG_SSLNK;
+		is_src = true;
+	} else {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Unknown direction\n", __func__);
+		return -EINVAL;
+	}
+
+	if (d40c->log_num != D40_PHY_CHAN) {
+		/*
+		 * Release logical channel, deactivate the event line during
+		 * the time physical res is suspended.
+		 */
+		writel((D40_DEACTIVATE_EVENTLINE << D40_EVENTLINE_POS(event)) &
+		       D40_EVENTLINE_MASK(event),
+		       d40c->base->virtbase + D40_DREG_PCBASE +
+		       phy->num * D40_DREG_PCDELTA + dir);
+
+		d40c->base->lookup_log_chans[d40c->log_num] = NULL;
+
+		/*
+		 * Check if there are more logical allocation
+		 * on this phy channel.
+		 */
+		if (!d40_alloc_mask_free(phy, is_src, event)) {
+			/* Resume the other logical channels if any */
+			if (d40_chan_has_events(d40c)) {
+				res = d40_channel_execute_command(d40c,
+								  D40_DMA_RUN);
+				if (res) {
+					dev_err(&d40c->chan.dev->device,
+						"[%s] Executing RUN command\n",
+						__func__);
+					return res;
+				}
+			}
+			return 0;
+		}
+	} else
+		d40_alloc_mask_free(phy, is_src, 0);
+
+	/* Release physical channel */
+	res = d40_channel_execute_command(d40c, D40_DMA_STOP);
+	if (res) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Failed to stop channel\n", __func__);
+		return res;
+	}
+	d40c->phy_chan = NULL;
+	/* Invalidate channel type */
+	d40c->dma_cfg.channel_type = 0;
+	d40c->base->lookup_phy_chans[phy->num] = NULL;
+
+	return 0;
+
+
+}
+
+static int d40_pause(struct dma_chan *chan)
+{
+	struct d40_chan *d40c =
+		container_of(chan, struct d40_chan, chan);
+	int res;
+
+	unsigned long flags;
+
+	spin_lock_irqsave(&d40c->lock, flags);
+
+	res = d40_channel_execute_command(d40c, D40_DMA_SUSPEND_REQ);
+	if (res == 0) {
+		if (d40c->log_num != D40_PHY_CHAN) {
+			d40_config_set_event(d40c, false);
+			/* Resume the other logical channels if any */
+			if (d40_chan_has_events(d40c))
+				res = d40_channel_execute_command(d40c,
+								  D40_DMA_RUN);
+		}
+	}
+
+	spin_unlock_irqrestore(&d40c->lock, flags);
+	return res;
+}
+
+static bool d40_is_paused(struct d40_chan *d40c)
+{
+	bool is_paused = false;
+	unsigned long flags;
+	void __iomem *active_reg;
+	u32 status;
+	u32 event;
+	int res;
+
+	spin_lock_irqsave(&d40c->lock, flags);
+
+	if (d40c->log_num == D40_PHY_CHAN) {
+		if (d40c->phy_chan->num % 2 == 0)
+			active_reg = d40c->base->virtbase + D40_DREG_ACTIVE;
+		else
+			active_reg = d40c->base->virtbase + D40_DREG_ACTIVO;
+
+		status = (readl(active_reg) &
+			  D40_CHAN_POS_MASK(d40c->phy_chan->num)) >>
+			D40_CHAN_POS(d40c->phy_chan->num);
+		if (status == D40_DMA_SUSPENDED || status == D40_DMA_STOP)
+			is_paused = true;
+
+		goto _exit;
+	}
+
+	res = d40_channel_execute_command(d40c, D40_DMA_SUSPEND_REQ);
+	if (res != 0)
+		goto _exit;
+
+	if (d40c->dma_cfg.dir == STEDMA40_MEM_TO_PERIPH ||
+	    d40c->dma_cfg.dir == STEDMA40_MEM_TO_MEM)
+		event = D40_TYPE_TO_EVENT(d40c->dma_cfg.dst_dev_type);
+	else if (d40c->dma_cfg.dir == STEDMA40_PERIPH_TO_MEM)
+		event = D40_TYPE_TO_EVENT(d40c->dma_cfg.src_dev_type);
+	else {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Unknown direction\n", __func__);
+		goto _exit;
+	}
+	status = d40_chan_has_events(d40c);
+	status = (status & D40_EVENTLINE_MASK(event)) >>
+		D40_EVENTLINE_POS(event);
+
+	if (status != D40_DMA_RUN)
+		is_paused = true;
+
+	/* Resume the other logical channels if any */
+	if (d40_chan_has_events(d40c))
+		res = d40_channel_execute_command(d40c,
+						  D40_DMA_RUN);
+
+_exit:
+	spin_unlock_irqrestore(&d40c->lock, flags);
+	return is_paused;
+
+}
+
+
+static bool d40_tx_is_linked(struct d40_chan *d40c)
+{
+	bool is_link;
+
+	if (d40c->log_num != D40_PHY_CHAN)
+		is_link = readl(&d40c->lcpa->lcsp3) &  D40_MEM_LCSP3_DLOS_MASK;
+	else
+		is_link = readl(d40c->base->virtbase + D40_DREG_PCBASE +
+				d40c->phy_chan->num * D40_DREG_PCDELTA +
+				D40_CHAN_REG_SDLNK) &
+			D40_SREG_LNK_PHYS_LNK_MASK;
+	return is_link;
+}
+
+static u32 d40_residue(struct d40_chan *d40c)
+{
+	u32 num_elt;
+
+	if (d40c->log_num != D40_PHY_CHAN)
+		num_elt = (readl(&d40c->lcpa->lcsp2) &  D40_MEM_LCSP2_ECNT_MASK)
+			>> D40_MEM_LCSP2_ECNT_POS;
+	else
+		num_elt = (readl(d40c->base->virtbase + D40_DREG_PCBASE +
+				 d40c->phy_chan->num * D40_DREG_PCDELTA +
+				 D40_CHAN_REG_SDELT) &
+			   D40_SREG_ELEM_PHY_ECNT_MASK) >> D40_SREG_ELEM_PHY_ECNT_POS;
+	return num_elt * (1 << d40c->dma_cfg.dst_info.data_width);
+}
+
+static int d40_resume(struct dma_chan *chan)
+{
+	struct d40_chan *d40c =
+		container_of(chan, struct d40_chan, chan);
+	int res = 0;
+	unsigned long flags;
+
+	spin_lock_irqsave(&d40c->lock, flags);
+
+	if (d40c->log_num != D40_PHY_CHAN) {
+		res = d40_channel_execute_command(d40c, D40_DMA_SUSPEND_REQ);
+		if (res)
+			goto out;
+
+		/* If bytes left to transfer or linked tx resume job */
+		if (d40_residue(d40c) || d40_tx_is_linked(d40c)) {
+			d40_config_set_event(d40c, true);
+			res = d40_channel_execute_command(d40c, D40_DMA_RUN);
+		}
+	} else if (d40_residue(d40c) || d40_tx_is_linked(d40c))
+		res = d40_channel_execute_command(d40c, D40_DMA_RUN);
+
+out:
+	spin_unlock_irqrestore(&d40c->lock, flags);
+	return res;
+}
+
+static u32 stedma40_residue(struct dma_chan *chan)
+{
+	struct d40_chan *d40c =
+		container_of(chan, struct d40_chan, chan);
+	u32 bytes_left;
+	unsigned long flags;
+
+	spin_lock_irqsave(&d40c->lock, flags);
+	bytes_left = d40_residue(d40c);
+	spin_unlock_irqrestore(&d40c->lock, flags);
+
+	return bytes_left;
+}
+
+/* Public DMA functions in addition to the DMA engine framework */
+
+int stedma40_set_psize(struct dma_chan *chan,
+		       int src_psize,
+		       int dst_psize)
+{
+	struct d40_chan *d40c =
+		container_of(chan, struct d40_chan, chan);
+	unsigned long flags;
+
+	spin_lock_irqsave(&d40c->lock, flags);
+
+	if (d40c->log_num != D40_PHY_CHAN) {
+		d40c->log_def.lcsp1 &= ~D40_MEM_LCSP1_SCFG_PSIZE_MASK;
+		d40c->log_def.lcsp3 &= ~D40_MEM_LCSP1_SCFG_PSIZE_MASK;
+		d40c->log_def.lcsp1 |= src_psize << D40_MEM_LCSP1_SCFG_PSIZE_POS;
+		d40c->log_def.lcsp3 |= dst_psize << D40_MEM_LCSP1_SCFG_PSIZE_POS;
+		goto out;
+	}
+
+	if (src_psize == STEDMA40_PSIZE_PHY_1)
+		d40c->src_def_cfg &= ~(1 << D40_SREG_CFG_PHY_PEN_POS);
+	else {
+		d40c->src_def_cfg |= 1 << D40_SREG_CFG_PHY_PEN_POS;
+		d40c->src_def_cfg &= ~(STEDMA40_PSIZE_PHY_16 <<
+				       D40_SREG_CFG_PSIZE_POS);
+		d40c->src_def_cfg |= src_psize << D40_SREG_CFG_PSIZE_POS;
+	}
+
+	if (dst_psize == STEDMA40_PSIZE_PHY_1)
+		d40c->dst_def_cfg &= ~(1 << D40_SREG_CFG_PHY_PEN_POS);
+	else {
+		d40c->dst_def_cfg |= 1 << D40_SREG_CFG_PHY_PEN_POS;
+		d40c->dst_def_cfg &= ~(STEDMA40_PSIZE_PHY_16 <<
+				       D40_SREG_CFG_PSIZE_POS);
+		d40c->dst_def_cfg |= dst_psize << D40_SREG_CFG_PSIZE_POS;
+	}
+out:
+	spin_unlock_irqrestore(&d40c->lock, flags);
+	return 0;
+}
+EXPORT_SYMBOL(stedma40_set_psize);
+
+struct dma_async_tx_descriptor *stedma40_memcpy_sg(struct dma_chan *chan,
+						   struct scatterlist *sgl_dst,
+						   struct scatterlist *sgl_src,
+						   unsigned int sgl_len,
+						   unsigned long flags)
+{
+	int res;
+	struct d40_desc *d40d;
+	struct d40_chan *d40c = container_of(chan, struct d40_chan,
+					     chan);
+	unsigned long flg;
+	int lli_max = d40c->base->plat_data->llis_per_log;
+
+
+	spin_lock_irqsave(&d40c->lock, flg);
+	d40d = d40_desc_get(d40c);
+
+	if (d40d == NULL)
+		goto err;
+
+	memset(d40d, 0, sizeof(struct d40_desc));
+	d40d->lli_len = sgl_len;
+
+	d40d->txd.flags = flags;
+
+	if (d40c->log_num != D40_PHY_CHAN) {
+		if (sgl_len > 1)
+			/*
+			 * Check if there is space available in lcla. If not,
+			 * split list into 1-length and run only in lcpa
+			 * space.
+			 */
+			if (d40_lcla_id_get(d40c,
+					    &d40c->base->lcla_pool) != 0)
+				lli_max = 1;
+
+		if (d40_pool_lli_alloc(d40d, sgl_len, true) < 0) {
+			dev_err(&d40c->chan.dev->device,
+				"[%s] Out of memory\n", __func__);
+			goto err;
+		}
+
+		(void) d40_log_sg_to_lli(d40c->lcla.src_id,
+					 sgl_src,
+					 sgl_len,
+					 d40d->lli_log.src,
+					 d40c->log_def.lcsp1,
+					 d40c->dma_cfg.src_info.data_width,
+					 flags & DMA_PREP_INTERRUPT, lli_max,
+					 d40c->base->plat_data->llis_per_log);
+
+		(void) d40_log_sg_to_lli(d40c->lcla.dst_id,
+					 sgl_dst,
+					 sgl_len,
+					 d40d->lli_log.dst,
+					 d40c->log_def.lcsp3,
+					 d40c->dma_cfg.dst_info.data_width,
+					 flags & DMA_PREP_INTERRUPT, lli_max,
+					 d40c->base->plat_data->llis_per_log);
+
+
+	} else {
+		if (d40_pool_lli_alloc(d40d, sgl_len, false) < 0) {
+			dev_err(&d40c->chan.dev->device,
+				"[%s] Out of memory\n", __func__);
+			goto err;
+		}
+
+		res = d40_phy_sg_to_lli(sgl_src,
+					sgl_len,
+					0,
+					d40d->lli_phy.src,
+					d40d->lli_phy.src_addr,
+					d40c->src_def_cfg,
+					d40c->dma_cfg.src_info.data_width,
+					d40c->dma_cfg.src_info.psize,
+					true);
+
+		if (res < 0)
+			goto err;
+
+		res = d40_phy_sg_to_lli(sgl_dst,
+					sgl_len,
+					0,
+					d40d->lli_phy.dst,
+					d40d->lli_phy.dst_addr,
+					d40c->dst_def_cfg,
+					d40c->dma_cfg.dst_info.data_width,
+					d40c->dma_cfg.dst_info.psize,
+					true);
+
+		if (res < 0)
+			goto err;
+
+		(void) dma_map_single(d40c->base->dev, d40d->lli_phy.src,
+				      d40d->lli_pool.size, DMA_TO_DEVICE);
+	}
+
+	dma_async_tx_descriptor_init(&d40d->txd, chan);
+
+	d40d->txd.tx_submit = d40_tx_submit;
+
+	spin_unlock_irqrestore(&d40c->lock, flg);
+
+	return &d40d->txd;
+err:
+	spin_unlock_irqrestore(&d40c->lock, flg);
+	return NULL;
+}
+EXPORT_SYMBOL(stedma40_memcpy_sg);
+
+bool stedma40_filter(struct dma_chan *chan, void *data)
+{
+	struct stedma40_chan_cfg *info = data;
+	struct d40_chan *d40c =
+		container_of(chan, struct d40_chan, chan);
+	int err;
+
+	if (data) {
+		err = d40_validate_conf(d40c, info);
+		if (!err)
+			d40c->dma_cfg = *info;
+	} else
+		err = d40_config_memcpy(d40c);
+
+	return err == 0;
+}
+EXPORT_SYMBOL(stedma40_filter);
+
+/* DMA ENGINE functions */
+static int d40_alloc_chan_resources(struct dma_chan *chan)
+{
+	int err;
+	unsigned long flags;
+	struct d40_chan *d40c =
+		container_of(chan, struct d40_chan, chan);
+
+	spin_lock_irqsave(&d40c->lock, flags);
+
+	d40c->completed = chan->cookie = 1;
+
+	/*
+	 * If no dma configuration is set (channel_type == 0)
+	 * use default configuration
+	 */
+	if (d40c->dma_cfg.channel_type == 0) {
+		err = d40_config_memcpy(d40c);
+		if (err)
+			goto err_alloc;
+	}
+
+	err = d40_allocate_channel(d40c);
+	if (err) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Failed to allocate channel\n", __func__);
+		goto err_alloc;
+	}
+
+	err = d40_config_chan(d40c, &d40c->dma_cfg);
+	if (err) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Failed to configure channel\n",
+			__func__);
+		goto err_config;
+	}
+
+	spin_unlock_irqrestore(&d40c->lock, flags);
+	return 0;
+
+ err_config:
+	(void) d40_free_dma(d40c);
+ err_alloc:
+	spin_unlock_irqrestore(&d40c->lock, flags);
+	dev_err(&d40c->chan.dev->device,
+		"[%s] Channel allocation failed\n", __func__);
+	return -EINVAL;
+}
+
+static void d40_free_chan_resources(struct dma_chan *chan)
+{
+	struct d40_chan *d40c =
+		container_of(chan, struct d40_chan, chan);
+	int err;
+	unsigned long flags;
+
+	spin_lock_irqsave(&d40c->lock, flags);
+
+	err = d40_free_dma(d40c);
+
+	if (err)
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Failed to free channel\n", __func__);
+	spin_unlock_irqrestore(&d40c->lock, flags);
+}
+
+static struct dma_async_tx_descriptor *d40_prep_memcpy(struct dma_chan *chan,
+						       dma_addr_t dst,
+						       dma_addr_t src,
+						       size_t size,
+						       unsigned long flags)
+{
+	struct d40_desc *d40d;
+	struct d40_chan *d40c = container_of(chan, struct d40_chan,
+					     chan);
+	unsigned long flg;
+	int err = 0;
+
+	spin_lock_irqsave(&d40c->lock, flg);
+	d40d = d40_desc_get(d40c);
+
+	if (d40d == NULL) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Descriptor is NULL\n", __func__);
+		goto err;
+	}
+
+	memset(d40d, 0, sizeof(struct d40_desc));
+
+	d40d->txd.flags = flags;
+
+	dma_async_tx_descriptor_init(&d40d->txd, chan);
+
+	d40d->txd.tx_submit = d40_tx_submit;
+
+	if (d40c->log_num != D40_PHY_CHAN) {
+
+		if (d40_pool_lli_alloc(d40d, 1, true) < 0) {
+			dev_err(&d40c->chan.dev->device,
+				"[%s] Out of memory\n", __func__);
+			goto err;
+		}
+		d40d->lli_len = 1;
+
+		d40_log_fill_lli(d40d->lli_log.src,
+				 src,
+				 size,
+				 0,
+				 d40c->log_def.lcsp1,
+				 d40c->dma_cfg.src_info.data_width,
+				 true, true);
+
+		d40_log_fill_lli(d40d->lli_log.dst,
+				 dst,
+				 size,
+				 0,
+				 d40c->log_def.lcsp3,
+				 d40c->dma_cfg.dst_info.data_width,
+				 true, true);
+
+	} else {
+
+		if (d40_pool_lli_alloc(d40d, 1, false) < 0) {
+			dev_err(&d40c->chan.dev->device,
+				"[%s] Out of memory\n", __func__);
+			goto err;
+		}
+
+		err = d40_phy_fill_lli(d40d->lli_phy.src,
+				       src,
+				       size,
+				       d40c->dma_cfg.src_info.psize,
+				       0,
+				       d40c->src_def_cfg,
+				       true,
+				       d40c->dma_cfg.src_info.data_width,
+				       false);
+		if (err)
+			goto err_fill_lli;
+
+		err = d40_phy_fill_lli(d40d->lli_phy.dst,
+				       dst,
+				       size,
+				       d40c->dma_cfg.dst_info.psize,
+				       0,
+				       d40c->dst_def_cfg,
+				       true,
+				       d40c->dma_cfg.dst_info.data_width,
+				       false);
+
+		if (err)
+			goto err_fill_lli;
+
+		(void) dma_map_single(d40c->base->dev, d40d->lli_phy.src,
+				      d40d->lli_pool.size, DMA_TO_DEVICE);
+	}
+
+	spin_unlock_irqrestore(&d40c->lock, flg);
+	return &d40d->txd;
+
+err_fill_lli:
+	dev_err(&d40c->chan.dev->device,
+		"[%s] Failed filling in PHY LLI\n", __func__);
+	d40_pool_lli_free(d40d);
+err:
+	spin_unlock_irqrestore(&d40c->lock, flg);
+	return NULL;
+}
+
+static int d40_prep_slave_sg_log(struct d40_desc *d40d,
+				 struct d40_chan *d40c,
+				 struct scatterlist *sgl,
+				 unsigned int sg_len,
+				 enum dma_data_direction direction,
+				 unsigned long flags)
+{
+	dma_addr_t dev_addr = 0;
+	int total_size;
+	int lli_max = d40c->base->plat_data->llis_per_log;
+
+	if (d40_pool_lli_alloc(d40d, sg_len, true) < 0) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Out of memory\n", __func__);
+		return -ENOMEM;
+	}
+
+	d40d->lli_len = sg_len;
+	d40d->lli_tcount = 0;
+
+	if (sg_len > 1)
+		/*
+		 * Check if there is space available in lcla.
+		 * If not, split list into 1-length and run only
+		 * in lcpa space.
+		 */
+		if (d40_lcla_id_get(d40c, &d40c->base->lcla_pool) != 0)
+			lli_max = 1;
+
+	if (direction == DMA_FROM_DEVICE) {
+		dev_addr = d40c->base->plat_data->dev_rx[d40c->dma_cfg.src_dev_type];
+		total_size = d40_log_sg_to_dev(&d40c->lcla,
+					       sgl, sg_len,
+					       &d40d->lli_log,
+					       &d40c->log_def,
+					       d40c->dma_cfg.src_info.data_width,
+					       d40c->dma_cfg.dst_info.data_width,
+					       direction,
+					       flags & DMA_PREP_INTERRUPT,
+					       dev_addr, lli_max,
+					       d40c->base->plat_data->llis_per_log);
+	} else if (direction == DMA_TO_DEVICE) {
+		dev_addr = d40c->base->plat_data->dev_tx[d40c->dma_cfg.dst_dev_type];
+		total_size = d40_log_sg_to_dev(&d40c->lcla,
+					       sgl, sg_len,
+					       &d40d->lli_log,
+					       &d40c->log_def,
+					       d40c->dma_cfg.src_info.data_width,
+					       d40c->dma_cfg.dst_info.data_width,
+					       direction,
+					       flags & DMA_PREP_INTERRUPT,
+					       dev_addr, lli_max,
+					       d40c->base->plat_data->llis_per_log);
+	} else
+		return -EINVAL;
+	if (total_size < 0)
+		return -EINVAL;
+
+	return 0;
+}
+
+static int d40_prep_slave_sg_phy(struct d40_desc *d40d,
+				 struct d40_chan *d40c,
+				 struct scatterlist *sgl,
+				 unsigned int sgl_len,
+				 enum dma_data_direction direction,
+				 unsigned long flags)
+{
+	dma_addr_t src_dev_addr;
+	dma_addr_t dst_dev_addr;
+	int res;
+
+	if (d40_pool_lli_alloc(d40d, sgl_len, false) < 0) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Out of memory\n", __func__);
+		return -ENOMEM;
+	}
+
+	d40d->lli_len = sgl_len;
+	d40d->lli_tcount = 0;
+
+	if (direction == DMA_FROM_DEVICE) {
+		dst_dev_addr = 0;
+		src_dev_addr = d40c->base->plat_data->dev_rx[d40c->dma_cfg.src_dev_type];
+	} else if (direction == DMA_TO_DEVICE) {
+		dst_dev_addr = d40c->base->plat_data->dev_tx[d40c->dma_cfg.dst_dev_type];
+		src_dev_addr = 0;
+	} else
+		return -EINVAL;
+
+	res = d40_phy_sg_to_lli(sgl,
+				sgl_len,
+				src_dev_addr,
+				d40d->lli_phy.src,
+				d40d->lli_phy.src_addr,
+				d40c->src_def_cfg,
+				d40c->dma_cfg.src_info.data_width,
+				d40c->dma_cfg.src_info.psize,
+				true);
+	if (res < 0)
+		return res;
+
+	res = d40_phy_sg_to_lli(sgl,
+				sgl_len,
+				dst_dev_addr,
+				d40d->lli_phy.dst,
+				d40d->lli_phy.dst_addr,
+				d40c->dst_def_cfg,
+				d40c->dma_cfg.dst_info.data_width,
+				d40c->dma_cfg.dst_info.psize,
+				 true);
+	if (res < 0)
+		return res;
+
+	(void) dma_map_single(d40c->base->dev, d40d->lli_phy.src,
+			      d40d->lli_pool.size, DMA_TO_DEVICE);
+	return 0;
+}
+
+static struct dma_async_tx_descriptor *d40_prep_slave_sg(struct dma_chan *chan,
+							 struct scatterlist *sgl,
+							 unsigned int sg_len,
+							 enum dma_data_direction direction,
+							 unsigned long flags)
+{
+	struct d40_desc *d40d;
+	struct d40_chan *d40c = container_of(chan, struct d40_chan,
+					     chan);
+	unsigned long flg;
+	int err;
+
+	if (d40c->dma_cfg.pre_transfer)
+		d40c->dma_cfg.pre_transfer(chan,
+					   d40c->dma_cfg.pre_transfer_data,
+					   sg_dma_len(sgl));
+
+	spin_lock_irqsave(&d40c->lock, flg);
+	d40d = d40_desc_get(d40c);
+	spin_unlock_irqrestore(&d40c->lock, flg);
+
+	if (d40d == NULL)
+		return NULL;
+
+	memset(d40d, 0, sizeof(struct d40_desc));
+
+	if (d40c->log_num != D40_PHY_CHAN)
+		err = d40_prep_slave_sg_log(d40d, d40c, sgl, sg_len,
+					    direction, flags);
+	else
+		err = d40_prep_slave_sg_phy(d40d, d40c, sgl, sg_len,
+					    direction, flags);
+	if (err) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Failed to prepare %s slave sg job: %d\n",
+			__func__,
+			d40c->log_num != D40_PHY_CHAN ? "log" : "phy", err);
+		return NULL;
+	}
+
+	d40d->txd.flags = flags;
+
+	dma_async_tx_descriptor_init(&d40d->txd, chan);
+
+	d40d->txd.tx_submit = d40_tx_submit;
+
+	return &d40d->txd;
+}
+
+static enum dma_status d40_tx_status(struct dma_chan *chan,
+				     dma_cookie_t cookie,
+				     struct dma_tx_state *txstate)
+{
+	struct d40_chan *d40c = container_of(chan, struct d40_chan, chan);
+	dma_cookie_t last_used;
+	dma_cookie_t last_complete;
+	int ret;
+
+	last_complete = d40c->completed;
+	last_used = chan->cookie;
+
+	if (d40_is_paused(d40c))
+		ret = DMA_PAUSED;
+	else
+		ret = dma_async_is_complete(cookie, last_complete, last_used);
+
+	dma_set_tx_state(txstate, last_complete, last_used,
+			 stedma40_residue(chan));
+
+	return ret;
+}
+
+static void d40_issue_pending(struct dma_chan *chan)
+{
+	struct d40_chan *d40c = container_of(chan, struct d40_chan, chan);
+	unsigned long flags;
+
+	spin_lock_irqsave(&d40c->lock, flags);
+
+	/* Busy means that pending jobs are already being processed */
+	if (!d40c->busy)
+		(void) d40_queue_start(d40c);
+
+	spin_unlock_irqrestore(&d40c->lock, flags);
+}
+
+static int d40_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+		       unsigned long arg)
+{
+	unsigned long flags;
+	struct d40_chan *d40c = container_of(chan, struct d40_chan, chan);
+
+	switch (cmd) {
+	case DMA_TERMINATE_ALL:
+		spin_lock_irqsave(&d40c->lock, flags);
+		d40_term_all(d40c);
+		spin_unlock_irqrestore(&d40c->lock, flags);
+		return 0;
+	case DMA_PAUSE:
+		return d40_pause(chan);
+	case DMA_RESUME:
+		return d40_resume(chan);
+	}
+
+	/* Other commands are unimplemented */
+	return -ENXIO;
+}
+
+/* Initialization functions */
+
+static void __init d40_chan_init(struct d40_base *base, struct dma_device *dma,
+				 struct d40_chan *chans, int offset,
+				 int num_chans)
+{
+	int i = 0;
+	struct d40_chan *d40c;
+
+	INIT_LIST_HEAD(&dma->channels);
+
+	for (i = offset; i < offset + num_chans; i++) {
+		d40c = &chans[i];
+		d40c->base = base;
+		d40c->chan.device = dma;
+
+		/* Invalidate lcla element */
+		d40c->lcla.src_id = -1;
+		d40c->lcla.dst_id = -1;
+
+		spin_lock_init(&d40c->lock);
+
+		d40c->log_num = D40_PHY_CHAN;
+
+		INIT_LIST_HEAD(&d40c->free);
+		INIT_LIST_HEAD(&d40c->active);
+		INIT_LIST_HEAD(&d40c->queue);
+		INIT_LIST_HEAD(&d40c->client);
+
+		d40c->free_len = 0;
+
+		tasklet_init(&d40c->tasklet, dma_tasklet,
+			     (unsigned long) d40c);
+
+		list_add_tail(&d40c->chan.device_node,
+			      &dma->channels);
+	}
+}
+
+static int __init d40_dmaengine_init(struct d40_base *base,
+				     int num_reserved_chans)
+{
+	int err ;
+
+	d40_chan_init(base, &base->dma_slave, base->log_chans,
+		      0, base->num_log_chans);
+
+	dma_cap_zero(base->dma_slave.cap_mask);
+	dma_cap_set(DMA_SLAVE, base->dma_slave.cap_mask);
+
+	base->dma_slave.device_alloc_chan_resources = d40_alloc_chan_resources;
+	base->dma_slave.device_free_chan_resources = d40_free_chan_resources;
+	base->dma_slave.device_prep_dma_memcpy = d40_prep_memcpy;
+	base->dma_slave.device_prep_slave_sg = d40_prep_slave_sg;
+	base->dma_slave.device_tx_status = d40_tx_status;
+	base->dma_slave.device_issue_pending = d40_issue_pending;
+	base->dma_slave.device_control = d40_control;
+	base->dma_slave.dev = base->dev;
+
+	err = dma_async_device_register(&base->dma_slave);
+
+	if (err) {
+		dev_err(base->dev,
+			"[%s] Failed to register slave channels\n",
+			__func__);
+		goto failure1;
+	}
+
+	d40_chan_init(base, &base->dma_memcpy, base->log_chans,
+		      base->num_log_chans, base->plat_data->memcpy_len);
+
+	dma_cap_zero(base->dma_memcpy.cap_mask);
+	dma_cap_set(DMA_MEMCPY, base->dma_memcpy.cap_mask);
+
+	base->dma_memcpy.device_alloc_chan_resources = d40_alloc_chan_resources;
+	base->dma_memcpy.device_free_chan_resources = d40_free_chan_resources;
+	base->dma_memcpy.device_prep_dma_memcpy = d40_prep_memcpy;
+	base->dma_memcpy.device_prep_slave_sg = d40_prep_slave_sg;
+	base->dma_memcpy.device_tx_status = d40_tx_status;
+	base->dma_memcpy.device_issue_pending = d40_issue_pending;
+	base->dma_memcpy.device_control = d40_control;
+	base->dma_memcpy.dev = base->dev;
+	/*
+	 * This controller can only access address at even
+	 * 32bit boundaries, i.e. 2^2
+	 */
+	base->dma_memcpy.copy_align = 2;
+
+	err = dma_async_device_register(&base->dma_memcpy);
+
+	if (err) {
+		dev_err(base->dev,
+			"[%s] Failed to regsiter memcpy only channels\n",
+			__func__);
+		goto failure2;
+	}
+
+	d40_chan_init(base, &base->dma_both, base->phy_chans,
+		      0, num_reserved_chans);
+
+	dma_cap_zero(base->dma_both.cap_mask);
+	dma_cap_set(DMA_SLAVE, base->dma_both.cap_mask);
+	dma_cap_set(DMA_MEMCPY, base->dma_both.cap_mask);
+
+	base->dma_both.device_alloc_chan_resources = d40_alloc_chan_resources;
+	base->dma_both.device_free_chan_resources = d40_free_chan_resources;
+	base->dma_both.device_prep_dma_memcpy = d40_prep_memcpy;
+	base->dma_both.device_prep_slave_sg = d40_prep_slave_sg;
+	base->dma_both.device_tx_status = d40_tx_status;
+	base->dma_both.device_issue_pending = d40_issue_pending;
+	base->dma_both.device_control = d40_control;
+	base->dma_both.dev = base->dev;
+	base->dma_both.copy_align = 2;
+	err = dma_async_device_register(&base->dma_both);
+
+	if (err) {
+		dev_err(base->dev,
+			"[%s] Failed to register logical and physical capable channels\n",
+			__func__);
+		goto failure3;
+	}
+	return 0;
+failure3:
+	dma_async_device_unregister(&base->dma_memcpy);
+failure2:
+	dma_async_device_unregister(&base->dma_slave);
+failure1:
+	return err;
+}
+
+/* Initialization functions. */
+
+static int __init d40_phy_res_init(struct d40_base *base)
+{
+	int i;
+	int num_phy_chans_avail = 0;
+	u32 val[2];
+	int odd_even_bit = -2;
+
+	val[0] = readl(base->virtbase + D40_DREG_PRSME);
+	val[1] = readl(base->virtbase + D40_DREG_PRSMO);
+
+	for (i = 0; i < base->num_phy_chans; i++) {
+		base->phy_res[i].num = i;
+		odd_even_bit += 2 * ((i % 2) == 0);
+		if (((val[i % 2] >> odd_even_bit) & 3) == 1) {
+			/* Mark security only channels as occupied */
+			base->phy_res[i].allocated_src = D40_ALLOC_PHY;
+			base->phy_res[i].allocated_dst = D40_ALLOC_PHY;
+		} else {
+			base->phy_res[i].allocated_src = D40_ALLOC_FREE;
+			base->phy_res[i].allocated_dst = D40_ALLOC_FREE;
+			num_phy_chans_avail++;
+		}
+		spin_lock_init(&base->phy_res[i].lock);
+	}
+	dev_info(base->dev, "%d of %d physical DMA channels available\n",
+		 num_phy_chans_avail, base->num_phy_chans);
+
+	/* Verify settings extended vs standard */
+	val[0] = readl(base->virtbase + D40_DREG_PRTYP);
+
+	for (i = 0; i < base->num_phy_chans; i++) {
+
+		if (base->phy_res[i].allocated_src == D40_ALLOC_FREE &&
+		    (val[0] & 0x3) != 1)
+			dev_info(base->dev,
+				 "[%s] INFO: channel %d is misconfigured (%d)\n",
+				 __func__, i, val[0] & 0x3);
+
+		val[0] = val[0] >> 2;
+	}
+
+	return num_phy_chans_avail;
+}
+
+static struct d40_base * __init d40_hw_detect_init(struct platform_device *pdev)
+{
+	static const struct d40_reg_val dma_id_regs[] = {
+		/* Peripheral Id */
+		{ .reg = D40_DREG_PERIPHID0, .val = 0x0040},
+		{ .reg = D40_DREG_PERIPHID1, .val = 0x0000},
+		/*
+		 * D40_DREG_PERIPHID2 Depends on HW revision:
+		 *  MOP500/HREF ED has 0x0008,
+		 *  ? has 0x0018,
+		 *  HREF V1 has 0x0028
+		 */
+		{ .reg = D40_DREG_PERIPHID3, .val = 0x0000},
+
+		/* PCell Id */
+		{ .reg = D40_DREG_CELLID0, .val = 0x000d},
+		{ .reg = D40_DREG_CELLID1, .val = 0x00f0},
+		{ .reg = D40_DREG_CELLID2, .val = 0x0005},
+		{ .reg = D40_DREG_CELLID3, .val = 0x00b1}
+	};
+	struct stedma40_platform_data *plat_data;
+	struct clk *clk = NULL;
+	void __iomem *virtbase = NULL;
+	struct resource *res = NULL;
+	struct d40_base *base = NULL;
+	int num_log_chans = 0;
+	int num_phy_chans;
+	int i;
+
+	clk = clk_get(&pdev->dev, NULL);
+
+	if (IS_ERR(clk)) {
+		dev_err(&pdev->dev, "[%s] No matching clock found\n",
+			__func__);
+		goto failure;
+	}
+
+	clk_enable(clk);
+
+	/* Get IO for DMAC base address */
+	res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "base");
+	if (!res)
+		goto failure;
+
+	if (request_mem_region(res->start, resource_size(res),
+			       D40_NAME " I/O base") == NULL)
+		goto failure;
+
+	virtbase = ioremap(res->start, resource_size(res));
+	if (!virtbase)
+		goto failure;
+
+	/* HW version check */
+	for (i = 0; i < ARRAY_SIZE(dma_id_regs); i++) {
+		if (dma_id_regs[i].val !=
+		    readl(virtbase + dma_id_regs[i].reg)) {
+			dev_err(&pdev->dev,
+				"[%s] Unknown hardware! Expected 0x%x at 0x%x but got 0x%x\n",
+				__func__,
+				dma_id_regs[i].val,
+				dma_id_regs[i].reg,
+				readl(virtbase + dma_id_regs[i].reg));
+			goto failure;
+		}
+	}
+
+	i = readl(virtbase + D40_DREG_PERIPHID2);
+
+	if ((i & 0xf) != D40_PERIPHID2_DESIGNER) {
+		dev_err(&pdev->dev,
+			"[%s] Unknown designer! Got %x wanted %x\n",
+			__func__, i & 0xf, D40_PERIPHID2_DESIGNER);
+		goto failure;
+	}
+
+	/* The number of physical channels on this HW */
+	num_phy_chans = 4 * (readl(virtbase + D40_DREG_ICFG) & 0x7) + 4;
+
+	dev_info(&pdev->dev, "hardware revision: %d @ 0x%x\n",
+		 (i >> 4) & 0xf, res->start);
+
+	plat_data = pdev->dev.platform_data;
+
+	/* Count the number of logical channels in use */
+	for (i = 0; i < plat_data->dev_len; i++)
+		if (plat_data->dev_rx[i] != 0)
+			num_log_chans++;
+
+	for (i = 0; i < plat_data->dev_len; i++)
+		if (plat_data->dev_tx[i] != 0)
+			num_log_chans++;
+
+	base = kzalloc(ALIGN(sizeof(struct d40_base), 4) +
+		       (num_phy_chans + num_log_chans + plat_data->memcpy_len) *
+		       sizeof(struct d40_chan), GFP_KERNEL);
+
+	if (base == NULL) {
+		dev_err(&pdev->dev, "[%s] Out of memory\n", __func__);
+		goto failure;
+	}
+
+	base->clk = clk;
+	base->num_phy_chans = num_phy_chans;
+	base->num_log_chans = num_log_chans;
+	base->phy_start = res->start;
+	base->phy_size = resource_size(res);
+	base->virtbase = virtbase;
+	base->plat_data = plat_data;
+	base->dev = &pdev->dev;
+	base->phy_chans = ((void *)base) + ALIGN(sizeof(struct d40_base), 4);
+	base->log_chans = &base->phy_chans[num_phy_chans];
+
+	base->phy_res = kzalloc(num_phy_chans * sizeof(struct d40_phy_res),
+				GFP_KERNEL);
+	if (!base->phy_res)
+		goto failure;
+
+	base->lookup_phy_chans = kzalloc(num_phy_chans *
+					 sizeof(struct d40_chan *),
+					 GFP_KERNEL);
+	if (!base->lookup_phy_chans)
+		goto failure;
+
+	if (num_log_chans + plat_data->memcpy_len) {
+		/*
+		 * The max number of logical channels are event lines for all
+		 * src devices and dst devices
+		 */
+		base->lookup_log_chans = kzalloc(plat_data->dev_len * 2 *
+						 sizeof(struct d40_chan *),
+						 GFP_KERNEL);
+		if (!base->lookup_log_chans)
+			goto failure;
+	}
+	base->lcla_pool.alloc_map = kzalloc(num_phy_chans * sizeof(u32),
+					    GFP_KERNEL);
+	if (!base->lcla_pool.alloc_map)
+		goto failure;
+
+	return base;
+
+failure:
+	if (clk) {
+		clk_disable(clk);
+		clk_put(clk);
+	}
+	if (virtbase)
+		iounmap(virtbase);
+	if (res)
+		release_mem_region(res->start,
+				   resource_size(res));
+	if (virtbase)
+		iounmap(virtbase);
+
+	if (base) {
+		kfree(base->lcla_pool.alloc_map);
+		kfree(base->lookup_log_chans);
+		kfree(base->lookup_phy_chans);
+		kfree(base->phy_res);
+		kfree(base);
+	}
+
+	return NULL;
+}
+
+static void __init d40_hw_init(struct d40_base *base)
+{
+
+	static const struct d40_reg_val dma_init_reg[] = {
+		/* Clock every part of the DMA block from start */
+		{ .reg = D40_DREG_GCC,    .val = 0x0000ff01},
+
+		/* Interrupts on all logical channels */
+		{ .reg = D40_DREG_LCMIS0, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCMIS1, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCMIS2, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCMIS3, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCICR0, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCICR1, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCICR2, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCICR3, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCTIS0, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCTIS1, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCTIS2, .val = 0xFFFFFFFF},
+		{ .reg = D40_DREG_LCTIS3, .val = 0xFFFFFFFF}
+	};
+	int i;
+	u32 prmseo[2] = {0, 0};
+	u32 activeo[2] = {0xFFFFFFFF, 0xFFFFFFFF};
+	u32 pcmis = 0;
+	u32 pcicr = 0;
+
+	for (i = 0; i < ARRAY_SIZE(dma_init_reg); i++)
+		writel(dma_init_reg[i].val,
+		       base->virtbase + dma_init_reg[i].reg);
+
+	/* Configure all our dma channels to default settings */
+	for (i = 0; i < base->num_phy_chans; i++) {
+
+		activeo[i % 2] = activeo[i % 2] << 2;
+
+		if (base->phy_res[base->num_phy_chans - i - 1].allocated_src
+		    == D40_ALLOC_PHY) {
+			activeo[i % 2] |= 3;
+			continue;
+		}
+
+		/* Enable interrupt # */
+		pcmis = (pcmis << 1) | 1;
+
+		/* Clear interrupt # */
+		pcicr = (pcicr << 1) | 1;
+
+		/* Set channel to physical mode */
+		prmseo[i % 2] = prmseo[i % 2] << 2;
+		prmseo[i % 2] |= 1;
+
+	}
+
+	writel(prmseo[1], base->virtbase + D40_DREG_PRMSE);
+	writel(prmseo[0], base->virtbase + D40_DREG_PRMSO);
+	writel(activeo[1], base->virtbase + D40_DREG_ACTIVE);
+	writel(activeo[0], base->virtbase + D40_DREG_ACTIVO);
+
+	/* Write which interrupt to enable */
+	writel(pcmis, base->virtbase + D40_DREG_PCMIS);
+
+	/* Write which interrupt to clear */
+	writel(pcicr, base->virtbase + D40_DREG_PCICR);
+
+}
+
+static int __init d40_probe(struct platform_device *pdev)
+{
+	int err;
+	int ret = -ENOENT;
+	struct d40_base *base;
+	struct resource *res = NULL;
+	int num_reserved_chans;
+	u32 val;
+
+	base = d40_hw_detect_init(pdev);
+
+	if (!base)
+		goto failure;
+
+	num_reserved_chans = d40_phy_res_init(base);
+
+	platform_set_drvdata(pdev, base);
+
+	spin_lock_init(&base->interrupt_lock);
+	spin_lock_init(&base->execmd_lock);
+
+	/* Get IO for logical channel parameter address */
+	res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "lcpa");
+	if (!res) {
+		ret = -ENOENT;
+		dev_err(&pdev->dev,
+			"[%s] No \"lcpa\" memory resource\n",
+			__func__);
+		goto failure;
+	}
+	base->lcpa_size = resource_size(res);
+	base->phy_lcpa = res->start;
+
+	if (request_mem_region(res->start, resource_size(res),
+			       D40_NAME " I/O lcpa") == NULL) {
+		ret = -EBUSY;
+		dev_err(&pdev->dev,
+			"[%s] Failed to request LCPA region 0x%x-0x%x\n",
+			__func__, res->start, res->end);
+		goto failure;
+	}
+
+	/* We make use of ESRAM memory for this. */
+	val = readl(base->virtbase + D40_DREG_LCPA);
+	if (res->start != val && val != 0) {
+		dev_warn(&pdev->dev,
+			 "[%s] Mismatch LCPA dma 0x%x, def 0x%x\n",
+			 __func__, val, res->start);
+	} else
+		writel(res->start, base->virtbase + D40_DREG_LCPA);
+
+	base->lcpa_base = ioremap(res->start, resource_size(res));
+	if (!base->lcpa_base) {
+		ret = -ENOMEM;
+		dev_err(&pdev->dev,
+			"[%s] Failed to ioremap LCPA region\n",
+			__func__);
+		goto failure;
+	}
+	/* Get IO for logical channel link address */
+	res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "lcla");
+	if (!res) {
+		ret = -ENOENT;
+		dev_err(&pdev->dev,
+			"[%s] No \"lcla\" resource defined\n",
+			__func__);
+		goto failure;
+	}
+
+	base->lcla_pool.base_size = resource_size(res);
+	base->lcla_pool.phy = res->start;
+
+	if (request_mem_region(res->start, resource_size(res),
+			       D40_NAME " I/O lcla") == NULL) {
+		ret = -EBUSY;
+		dev_err(&pdev->dev,
+			"[%s] Failed to request LCLA region 0x%x-0x%x\n",
+			__func__, res->start, res->end);
+		goto failure;
+	}
+	val = readl(base->virtbase + D40_DREG_LCLA);
+	if (res->start != val && val != 0) {
+		dev_warn(&pdev->dev,
+			 "[%s] Mismatch LCLA dma 0x%x, def 0x%x\n",
+			 __func__, val, res->start);
+	} else
+		writel(res->start, base->virtbase + D40_DREG_LCLA);
+
+	base->lcla_pool.base = ioremap(res->start, resource_size(res));
+	if (!base->lcla_pool.base) {
+		ret = -ENOMEM;
+		dev_err(&pdev->dev,
+			"[%s] Failed to ioremap LCLA 0x%x-0x%x\n",
+			__func__, res->start, res->end);
+		goto failure;
+	}
+
+	spin_lock_init(&base->lcla_pool.lock);
+
+	base->lcla_pool.num_blocks = base->num_phy_chans;
+
+	base->irq = platform_get_irq(pdev, 0);
+
+	ret = request_irq(base->irq, d40_handle_interrupt, 0, D40_NAME, base);
+
+	if (ret) {
+		dev_err(&pdev->dev, "[%s] No IRQ defined\n", __func__);
+		goto failure;
+	}
+
+	err = d40_dmaengine_init(base, num_reserved_chans);
+	if (err)
+		goto failure;
+
+	d40_hw_init(base);
+
+	dev_info(base->dev, "initialized\n");
+	return 0;
+
+failure:
+	if (base) {
+		if (base->virtbase)
+			iounmap(base->virtbase);
+		if (base->lcla_pool.phy)
+			release_mem_region(base->lcla_pool.phy,
+					   base->lcla_pool.base_size);
+		if (base->phy_lcpa)
+			release_mem_region(base->phy_lcpa,
+					   base->lcpa_size);
+		if (base->phy_start)
+			release_mem_region(base->phy_start,
+					   base->phy_size);
+		if (base->clk) {
+			clk_disable(base->clk);
+			clk_put(base->clk);
+		}
+
+		kfree(base->lcla_pool.alloc_map);
+		kfree(base->lookup_log_chans);
+		kfree(base->lookup_phy_chans);
+		kfree(base->phy_res);
+		kfree(base);
+	}
+
+	dev_err(&pdev->dev, "[%s] probe failed\n", __func__);
+	return ret;
+}
+
+static struct platform_driver d40_driver = {
+	.driver = {
+		.owner = THIS_MODULE,
+		.name  = D40_NAME,
+	},
+};
+
+int __init stedma40_init(void)
+{
+	return platform_driver_probe(&d40_driver, d40_probe);
+}
+arch_initcall(stedma40_init);
diff --git a/drivers/dma/ste_dma40_ll.c b/drivers/dma/ste_dma40_ll.c
new file mode 100644
index 000000000000..561fdd8a80c1
--- /dev/null
+++ b/drivers/dma/ste_dma40_ll.c
@@ -0,0 +1,454 @@
+/*
+ * driver/dma/ste_dma40_ll.c
+ *
+ * Copyright (C) ST-Ericsson 2007-2010
+ * License terms: GNU General Public License (GPL) version 2
+ * Author: Per Friden <per.friden@stericsson.com>
+ * Author: Jonas Aaberg <jonas.aberg@stericsson.com>
+ */
+
+#include <linux/kernel.h>
+#include <plat/ste_dma40.h>
+
+#include "ste_dma40_ll.h"
+
+/* Sets up proper LCSP1 and LCSP3 register for a logical channel */
+void d40_log_cfg(struct stedma40_chan_cfg *cfg,
+		 u32 *lcsp1, u32 *lcsp3)
+{
+	u32 l3 = 0; /* dst */
+	u32 l1 = 0; /* src */
+
+	/* src is mem? -> increase address pos */
+	if (cfg->dir ==  STEDMA40_MEM_TO_PERIPH ||
+	    cfg->dir ==  STEDMA40_MEM_TO_MEM)
+		l1 |= 1 << D40_MEM_LCSP1_SCFG_INCR_POS;
+
+	/* dst is mem? -> increase address pos */
+	if (cfg->dir ==  STEDMA40_PERIPH_TO_MEM ||
+	    cfg->dir ==  STEDMA40_MEM_TO_MEM)
+		l3 |= 1 << D40_MEM_LCSP3_DCFG_INCR_POS;
+
+	/* src is hw? -> master port 1 */
+	if (cfg->dir ==  STEDMA40_PERIPH_TO_MEM ||
+	    cfg->dir ==  STEDMA40_PERIPH_TO_PERIPH)
+		l1 |= 1 << D40_MEM_LCSP1_SCFG_MST_POS;
+
+	/* dst is hw? -> master port 1 */
+	if (cfg->dir ==  STEDMA40_MEM_TO_PERIPH ||
+	    cfg->dir ==  STEDMA40_PERIPH_TO_PERIPH)
+		l3 |= 1 << D40_MEM_LCSP3_DCFG_MST_POS;
+
+	l3 |= 1 << D40_MEM_LCSP3_DCFG_TIM_POS;
+	l3 |= 1 << D40_MEM_LCSP3_DCFG_EIM_POS;
+	l3 |= cfg->dst_info.psize << D40_MEM_LCSP3_DCFG_PSIZE_POS;
+	l3 |= cfg->dst_info.data_width << D40_MEM_LCSP3_DCFG_ESIZE_POS;
+	l3 |= 1 << D40_MEM_LCSP3_DTCP_POS;
+
+	l1 |= 1 << D40_MEM_LCSP1_SCFG_EIM_POS;
+	l1 |= cfg->src_info.psize << D40_MEM_LCSP1_SCFG_PSIZE_POS;
+	l1 |= cfg->src_info.data_width << D40_MEM_LCSP1_SCFG_ESIZE_POS;
+	l1 |= 1 << D40_MEM_LCSP1_STCP_POS;
+
+	*lcsp1 = l1;
+	*lcsp3 = l3;
+
+}
+
+/* Sets up SRC and DST CFG register for both logical and physical channels */
+void d40_phy_cfg(struct stedma40_chan_cfg *cfg,
+		 u32 *src_cfg, u32 *dst_cfg, bool is_log)
+{
+	u32 src = 0;
+	u32 dst = 0;
+
+	if (!is_log) {
+		/* Physical channel */
+		if ((cfg->dir ==  STEDMA40_PERIPH_TO_MEM) ||
+		    (cfg->dir == STEDMA40_PERIPH_TO_PERIPH)) {
+			/* Set master port to 1 */
+			src |= 1 << D40_SREG_CFG_MST_POS;
+			src |= D40_TYPE_TO_EVENT(cfg->src_dev_type);
+
+			if (cfg->src_info.flow_ctrl == STEDMA40_NO_FLOW_CTRL)
+				src |= 1 << D40_SREG_CFG_PHY_TM_POS;
+			else
+				src |= 3 << D40_SREG_CFG_PHY_TM_POS;
+		}
+		if ((cfg->dir ==  STEDMA40_MEM_TO_PERIPH) ||
+		    (cfg->dir == STEDMA40_PERIPH_TO_PERIPH)) {
+			/* Set master port to 1 */
+			dst |= 1 << D40_SREG_CFG_MST_POS;
+			dst |= D40_TYPE_TO_EVENT(cfg->dst_dev_type);
+
+			if (cfg->dst_info.flow_ctrl == STEDMA40_NO_FLOW_CTRL)
+				dst |= 1 << D40_SREG_CFG_PHY_TM_POS;
+			else
+				dst |= 3 << D40_SREG_CFG_PHY_TM_POS;
+		}
+		/* Interrupt on end of transfer for destination */
+		dst |= 1 << D40_SREG_CFG_TIM_POS;
+
+		/* Generate interrupt on error */
+		src |= 1 << D40_SREG_CFG_EIM_POS;
+		dst |= 1 << D40_SREG_CFG_EIM_POS;
+
+		/* PSIZE */
+		if (cfg->src_info.psize != STEDMA40_PSIZE_PHY_1) {
+			src |= 1 << D40_SREG_CFG_PHY_PEN_POS;
+			src |= cfg->src_info.psize << D40_SREG_CFG_PSIZE_POS;
+		}
+		if (cfg->dst_info.psize != STEDMA40_PSIZE_PHY_1) {
+			dst |= 1 << D40_SREG_CFG_PHY_PEN_POS;
+			dst |= cfg->dst_info.psize << D40_SREG_CFG_PSIZE_POS;
+		}
+
+		/* Element size */
+		src |= cfg->src_info.data_width << D40_SREG_CFG_ESIZE_POS;
+		dst |= cfg->dst_info.data_width << D40_SREG_CFG_ESIZE_POS;
+
+	} else {
+		/* Logical channel */
+		dst |= 1 << D40_SREG_CFG_LOG_GIM_POS;
+		src |= 1 << D40_SREG_CFG_LOG_GIM_POS;
+	}
+
+	if (cfg->channel_type & STEDMA40_HIGH_PRIORITY_CHANNEL) {
+		src |= 1 << D40_SREG_CFG_PRI_POS;
+		dst |= 1 << D40_SREG_CFG_PRI_POS;
+	}
+
+	src |= cfg->src_info.endianess << D40_SREG_CFG_LBE_POS;
+	dst |= cfg->dst_info.endianess << D40_SREG_CFG_LBE_POS;
+
+	*src_cfg = src;
+	*dst_cfg = dst;
+}
+
+int d40_phy_fill_lli(struct d40_phy_lli *lli,
+		     dma_addr_t data,
+		     u32 data_size,
+		     int psize,
+		     dma_addr_t next_lli,
+		     u32 reg_cfg,
+		     bool term_int,
+		     u32 data_width,
+		     bool is_device)
+{
+	int num_elems;
+
+	if (psize == STEDMA40_PSIZE_PHY_1)
+		num_elems = 1;
+	else
+		num_elems = 2 << psize;
+
+	/*
+	 * Size is 16bit. data_width is 8, 16, 32 or 64 bit
+	 * Block large than 64 KiB must be split.
+	 */
+	if (data_size > (0xffff << data_width))
+		return -EINVAL;
+
+	/* Must be aligned */
+	if (!IS_ALIGNED(data, 0x1 << data_width))
+		return -EINVAL;
+
+	/* Transfer size can't be smaller than (num_elms * elem_size) */
+	if (data_size < num_elems * (0x1 << data_width))
+		return -EINVAL;
+
+	/* The number of elements. IE now many chunks */
+	lli->reg_elt = (data_size >> data_width) << D40_SREG_ELEM_PHY_ECNT_POS;
+
+	/*
+	 * Distance to next element sized entry.
+	 * Usually the size of the element unless you want gaps.
+	 */
+	if (!is_device)
+		lli->reg_elt |= (0x1 << data_width) <<
+			D40_SREG_ELEM_PHY_EIDX_POS;
+
+	/* Where the data is */
+	lli->reg_ptr = data;
+	lli->reg_cfg = reg_cfg;
+
+	/* If this scatter list entry is the last one, no next link */
+	if (next_lli == 0)
+		lli->reg_lnk = 0x1 << D40_SREG_LNK_PHY_TCP_POS;
+	else
+		lli->reg_lnk = next_lli;
+
+	/* Set/clear interrupt generation on this link item.*/
+	if (term_int)
+		lli->reg_cfg |= 0x1 << D40_SREG_CFG_TIM_POS;
+	else
+		lli->reg_cfg &= ~(0x1 << D40_SREG_CFG_TIM_POS);
+
+	/* Post link */
+	lli->reg_lnk |= 0 << D40_SREG_LNK_PHY_PRE_POS;
+
+	return 0;
+}
+
+int d40_phy_sg_to_lli(struct scatterlist *sg,
+		      int sg_len,
+		      dma_addr_t target,
+		      struct d40_phy_lli *lli,
+		      dma_addr_t lli_phys,
+		      u32 reg_cfg,
+		      u32 data_width,
+		      int psize,
+		      bool term_int)
+{
+	int total_size = 0;
+	int i;
+	struct scatterlist *current_sg = sg;
+	dma_addr_t next_lli_phys;
+	dma_addr_t dst;
+	int err = 0;
+
+	for_each_sg(sg, current_sg, sg_len, i) {
+
+		total_size += sg_dma_len(current_sg);
+
+		/* If this scatter list entry is the last one, no next link */
+		if (sg_len - 1 == i)
+			next_lli_phys = 0;
+		else
+			next_lli_phys = ALIGN(lli_phys + (i + 1) *
+					      sizeof(struct d40_phy_lli),
+					      D40_LLI_ALIGN);
+
+		if (target)
+			dst = target;
+		else
+			dst = sg_phys(current_sg);
+
+		err = d40_phy_fill_lli(&lli[i],
+				       dst,
+				       sg_dma_len(current_sg),
+				       psize,
+				       next_lli_phys,
+				       reg_cfg,
+				       !next_lli_phys,
+				       data_width,
+				       target == dst);
+		if (err)
+			goto err;
+	}
+
+	return total_size;
+ err:
+	return err;
+}
+
+
+void d40_phy_lli_write(void __iomem *virtbase,
+		       u32 phy_chan_num,
+		       struct d40_phy_lli *lli_dst,
+		       struct d40_phy_lli *lli_src)
+{
+
+	writel(lli_src->reg_cfg, virtbase + D40_DREG_PCBASE +
+	       phy_chan_num * D40_DREG_PCDELTA + D40_CHAN_REG_SSCFG);
+	writel(lli_src->reg_elt, virtbase + D40_DREG_PCBASE +
+	       phy_chan_num * D40_DREG_PCDELTA + D40_CHAN_REG_SSELT);
+	writel(lli_src->reg_ptr, virtbase + D40_DREG_PCBASE +
+	       phy_chan_num * D40_DREG_PCDELTA + D40_CHAN_REG_SSPTR);
+	writel(lli_src->reg_lnk, virtbase + D40_DREG_PCBASE +
+	       phy_chan_num * D40_DREG_PCDELTA + D40_CHAN_REG_SSLNK);
+
+	writel(lli_dst->reg_cfg, virtbase + D40_DREG_PCBASE +
+	       phy_chan_num * D40_DREG_PCDELTA + D40_CHAN_REG_SDCFG);
+	writel(lli_dst->reg_elt, virtbase + D40_DREG_PCBASE +
+	       phy_chan_num * D40_DREG_PCDELTA + D40_CHAN_REG_SDELT);
+	writel(lli_dst->reg_ptr, virtbase + D40_DREG_PCBASE +
+	       phy_chan_num * D40_DREG_PCDELTA + D40_CHAN_REG_SDPTR);
+	writel(lli_dst->reg_lnk, virtbase + D40_DREG_PCBASE +
+	       phy_chan_num * D40_DREG_PCDELTA + D40_CHAN_REG_SDLNK);
+
+}
+
+/* DMA logical lli operations */
+
+void d40_log_fill_lli(struct d40_log_lli *lli,
+		      dma_addr_t data, u32 data_size,
+		      u32 lli_next_off, u32 reg_cfg,
+		      u32 data_width,
+		      bool term_int, bool addr_inc)
+{
+	lli->lcsp13 = reg_cfg;
+
+	/* The number of elements to transfer */
+	lli->lcsp02 = ((data_size >> data_width) <<
+		       D40_MEM_LCSP0_ECNT_POS) & D40_MEM_LCSP0_ECNT_MASK;
+	/* 16 LSBs address of the current element */
+	lli->lcsp02 |= data & D40_MEM_LCSP0_SPTR_MASK;
+	/* 16 MSBs address of the current element */
+	lli->lcsp13 |= data & D40_MEM_LCSP1_SPTR_MASK;
+
+	if (addr_inc)
+		lli->lcsp13 |= D40_MEM_LCSP1_SCFG_INCR_MASK;
+
+	lli->lcsp13 |= D40_MEM_LCSP3_DTCP_MASK;
+	/* If this scatter list entry is the last one, no next link */
+	lli->lcsp13 |= (lli_next_off << D40_MEM_LCSP1_SLOS_POS) &
+		D40_MEM_LCSP1_SLOS_MASK;
+
+	if (term_int)
+		lli->lcsp13 |= D40_MEM_LCSP1_SCFG_TIM_MASK;
+	else
+		lli->lcsp13 &= ~D40_MEM_LCSP1_SCFG_TIM_MASK;
+}
+
+int d40_log_sg_to_dev(struct d40_lcla_elem *lcla,
+		      struct scatterlist *sg,
+		      int sg_len,
+		      struct d40_log_lli_bidir *lli,
+		      struct d40_def_lcsp *lcsp,
+		      u32 src_data_width,
+		      u32 dst_data_width,
+		      enum dma_data_direction direction,
+		      bool term_int, dma_addr_t dev_addr, int max_len,
+		      int llis_per_log)
+{
+	int total_size = 0;
+	struct scatterlist *current_sg = sg;
+	int i;
+	u32 next_lli_off_dst;
+	u32 next_lli_off_src;
+
+	next_lli_off_src = 0;
+	next_lli_off_dst = 0;
+
+	for_each_sg(sg, current_sg, sg_len, i) {
+		total_size += sg_dma_len(current_sg);
+
+		/*
+		 * If this scatter list entry is the last one or
+		 * max length, terminate link.
+		 */
+		if (sg_len - 1 == i || ((i+1) % max_len == 0)) {
+			next_lli_off_src = 0;
+			next_lli_off_dst = 0;
+		} else {
+			if (next_lli_off_dst == 0 &&
+			    next_lli_off_src == 0) {
+				/* The first lli will be at next_lli_off */
+				next_lli_off_dst = (lcla->dst_id *
+						    llis_per_log + 1);
+				next_lli_off_src = (lcla->src_id *
+						    llis_per_log + 1);
+			} else {
+				next_lli_off_dst++;
+				next_lli_off_src++;
+			}
+		}
+
+		if (direction == DMA_TO_DEVICE) {
+			d40_log_fill_lli(&lli->src[i],
+					 sg_phys(current_sg),
+					 sg_dma_len(current_sg),
+					 next_lli_off_src,
+					 lcsp->lcsp1, src_data_width,
+					 term_int && !next_lli_off_src,
+					 true);
+			d40_log_fill_lli(&lli->dst[i],
+					 dev_addr,
+					 sg_dma_len(current_sg),
+					 next_lli_off_dst,
+					 lcsp->lcsp3, dst_data_width,
+					 /* No next == terminal interrupt */
+					 term_int && !next_lli_off_dst,
+					 false);
+		} else {
+			d40_log_fill_lli(&lli->dst[i],
+					 sg_phys(current_sg),
+					 sg_dma_len(current_sg),
+					 next_lli_off_dst,
+					 lcsp->lcsp3, dst_data_width,
+					 /* No next == terminal interrupt */
+					 term_int && !next_lli_off_dst,
+					 true);
+			d40_log_fill_lli(&lli->src[i],
+					 dev_addr,
+					 sg_dma_len(current_sg),
+					 next_lli_off_src,
+					 lcsp->lcsp1, src_data_width,
+					 term_int && !next_lli_off_src,
+					 false);
+		}
+	}
+	return total_size;
+}
+
+int d40_log_sg_to_lli(int lcla_id,
+		      struct scatterlist *sg,
+		      int sg_len,
+		      struct d40_log_lli *lli_sg,
+		      u32 lcsp13, /* src or dst*/
+		      u32 data_width,
+		      bool term_int, int max_len, int llis_per_log)
+{
+	int total_size = 0;
+	struct scatterlist *current_sg = sg;
+	int i;
+	u32 next_lli_off = 0;
+
+	for_each_sg(sg, current_sg, sg_len, i) {
+		total_size += sg_dma_len(current_sg);
+
+		/*
+		 * If this scatter list entry is the last one or
+		 * max length, terminate link.
+		 */
+		if (sg_len - 1 == i || ((i+1) % max_len == 0))
+			next_lli_off = 0;
+		else {
+			if (next_lli_off == 0)
+				/* The first lli will be at next_lli_off */
+				next_lli_off = lcla_id * llis_per_log + 1;
+			else
+				next_lli_off++;
+		}
+
+		d40_log_fill_lli(&lli_sg[i],
+				 sg_phys(current_sg),
+				 sg_dma_len(current_sg),
+				 next_lli_off,
+				 lcsp13, data_width,
+				 term_int && !next_lli_off,
+				 true);
+	}
+	return total_size;
+}
+
+void d40_log_lli_write(struct d40_log_lli_full *lcpa,
+		       struct d40_log_lli *lcla_src,
+		       struct d40_log_lli *lcla_dst,
+		       struct d40_log_lli *lli_dst,
+		       struct d40_log_lli *lli_src,
+		       int llis_per_log)
+{
+	u32 slos = 0;
+	u32 dlos = 0;
+	int i;
+
+	lcpa->lcsp0 = lli_src->lcsp02;
+	lcpa->lcsp1 = lli_src->lcsp13;
+	lcpa->lcsp2 = lli_dst->lcsp02;
+	lcpa->lcsp3 = lli_dst->lcsp13;
+
+	slos = lli_src->lcsp13 & D40_MEM_LCSP1_SLOS_MASK;
+	dlos = lli_dst->lcsp13 & D40_MEM_LCSP3_DLOS_MASK;
+
+	for (i = 0; (i < llis_per_log) && slos && dlos; i++) {
+		writel(lli_src[i+1].lcsp02, &lcla_src[i].lcsp02);
+		writel(lli_src[i+1].lcsp13, &lcla_src[i].lcsp13);
+		writel(lli_dst[i+1].lcsp02, &lcla_dst[i].lcsp02);
+		writel(lli_dst[i+1].lcsp13, &lcla_dst[i].lcsp13);
+
+		slos = lli_src[i+1].lcsp13 & D40_MEM_LCSP1_SLOS_MASK;
+		dlos = lli_dst[i+1].lcsp13 & D40_MEM_LCSP3_DLOS_MASK;
+	}
+}
diff --git a/drivers/dma/ste_dma40_ll.h b/drivers/dma/ste_dma40_ll.h
new file mode 100644
index 000000000000..2029280cb332
--- /dev/null
+++ b/drivers/dma/ste_dma40_ll.h
@@ -0,0 +1,354 @@
+/*
+ * driver/dma/ste_dma40_ll.h
+ *
+ * Copyright (C) ST-Ericsson 2007-2010
+ * License terms: GNU General Public License (GPL) version 2
+ * Author: Per Friden <per.friden@stericsson.com>
+ * Author: Jonas Aaberg <jonas.aberg@stericsson.com>
+ */
+#ifndef STE_DMA40_LL_H
+#define STE_DMA40_LL_H
+
+#define D40_DREG_PCBASE		0x400
+#define D40_DREG_PCDELTA	(8 * 4)
+#define D40_LLI_ALIGN		16 /* LLI alignment must be 16 bytes. */
+
+#define D40_TYPE_TO_GROUP(type) (type / 16)
+#define D40_TYPE_TO_EVENT(type) (type % 16)
+
+/* Most bits of the CFG register are the same in log as in phy mode */
+#define D40_SREG_CFG_MST_POS		15
+#define D40_SREG_CFG_TIM_POS		14
+#define D40_SREG_CFG_EIM_POS		13
+#define D40_SREG_CFG_LOG_INCR_POS	12
+#define D40_SREG_CFG_PHY_PEN_POS	12
+#define D40_SREG_CFG_PSIZE_POS		10
+#define D40_SREG_CFG_ESIZE_POS		 8
+#define D40_SREG_CFG_PRI_POS		 7
+#define D40_SREG_CFG_LBE_POS		 6
+#define D40_SREG_CFG_LOG_GIM_POS	 5
+#define D40_SREG_CFG_LOG_MFU_POS	 4
+#define D40_SREG_CFG_PHY_TM_POS		 4
+#define D40_SREG_CFG_PHY_EVTL_POS	 0
+
+
+/* Standard channel parameters - basic mode (element register) */
+#define D40_SREG_ELEM_PHY_ECNT_POS	16
+#define D40_SREG_ELEM_PHY_EIDX_POS	 0
+
+#define D40_SREG_ELEM_PHY_ECNT_MASK	(0xFFFF << D40_SREG_ELEM_PHY_ECNT_POS)
+
+/* Standard channel parameters - basic mode (Link register) */
+#define D40_SREG_LNK_PHY_TCP_POS	0
+#define D40_SREG_LNK_PHY_LMP_POS	1
+#define D40_SREG_LNK_PHY_PRE_POS	2
+/*
+ * Source  destination link address. Contains the
+ * 29-bit byte word aligned address of the reload area.
+ */
+#define D40_SREG_LNK_PHYS_LNK_MASK	0xFFFFFFF8UL
+
+/* Standard basic channel logical mode */
+
+/* Element register */
+#define D40_SREG_ELEM_LOG_ECNT_POS	16
+#define D40_SREG_ELEM_LOG_LIDX_POS	 8
+#define D40_SREG_ELEM_LOG_LOS_POS	 1
+#define D40_SREG_ELEM_LOG_TCP_POS	 0
+
+#define D40_SREG_ELEM_LOG_LIDX_MASK	(0xFF << D40_SREG_ELEM_LOG_LIDX_POS)
+
+/* Link register */
+#define D40_DEACTIVATE_EVENTLINE	0x0
+#define D40_ACTIVATE_EVENTLINE		0x1
+#define D40_EVENTLINE_POS(i)		(2 * i)
+#define D40_EVENTLINE_MASK(i)		(0x3 << D40_EVENTLINE_POS(i))
+
+/* Standard basic channel logical params in memory */
+
+/* LCSP0 */
+#define D40_MEM_LCSP0_ECNT_POS		16
+#define D40_MEM_LCSP0_SPTR_POS		 0
+
+#define D40_MEM_LCSP0_ECNT_MASK		(0xFFFF << D40_MEM_LCSP0_ECNT_POS)
+#define D40_MEM_LCSP0_SPTR_MASK		(0xFFFF << D40_MEM_LCSP0_SPTR_POS)
+
+/* LCSP1 */
+#define D40_MEM_LCSP1_SPTR_POS		16
+#define D40_MEM_LCSP1_SCFG_MST_POS	15
+#define D40_MEM_LCSP1_SCFG_TIM_POS	14
+#define D40_MEM_LCSP1_SCFG_EIM_POS	13
+#define D40_MEM_LCSP1_SCFG_INCR_POS	12
+#define D40_MEM_LCSP1_SCFG_PSIZE_POS	10
+#define D40_MEM_LCSP1_SCFG_ESIZE_POS	 8
+#define D40_MEM_LCSP1_SLOS_POS		 1
+#define D40_MEM_LCSP1_STCP_POS		 0
+
+#define D40_MEM_LCSP1_SPTR_MASK		(0xFFFF << D40_MEM_LCSP1_SPTR_POS)
+#define D40_MEM_LCSP1_SCFG_TIM_MASK	(0x1 << D40_MEM_LCSP1_SCFG_TIM_POS)
+#define D40_MEM_LCSP1_SCFG_INCR_MASK	(0x1 << D40_MEM_LCSP1_SCFG_INCR_POS)
+#define D40_MEM_LCSP1_SCFG_PSIZE_MASK	(0x3 << D40_MEM_LCSP1_SCFG_PSIZE_POS)
+#define D40_MEM_LCSP1_SLOS_MASK		(0x7F << D40_MEM_LCSP1_SLOS_POS)
+#define D40_MEM_LCSP1_STCP_MASK		(0x1 << D40_MEM_LCSP1_STCP_POS)
+
+/* LCSP2 */
+#define D40_MEM_LCSP2_ECNT_POS		16
+
+#define D40_MEM_LCSP2_ECNT_MASK		(0xFFFF << D40_MEM_LCSP2_ECNT_POS)
+
+/* LCSP3 */
+#define D40_MEM_LCSP3_DCFG_MST_POS	15
+#define D40_MEM_LCSP3_DCFG_TIM_POS	14
+#define D40_MEM_LCSP3_DCFG_EIM_POS	13
+#define D40_MEM_LCSP3_DCFG_INCR_POS	12
+#define D40_MEM_LCSP3_DCFG_PSIZE_POS	10
+#define D40_MEM_LCSP3_DCFG_ESIZE_POS	 8
+#define D40_MEM_LCSP3_DLOS_POS		 1
+#define D40_MEM_LCSP3_DTCP_POS		 0
+
+#define D40_MEM_LCSP3_DLOS_MASK		(0x7F << D40_MEM_LCSP3_DLOS_POS)
+#define D40_MEM_LCSP3_DTCP_MASK		(0x1 << D40_MEM_LCSP3_DTCP_POS)
+
+
+/* Standard channel parameter register offsets */
+#define D40_CHAN_REG_SSCFG	0x00
+#define D40_CHAN_REG_SSELT	0x04
+#define D40_CHAN_REG_SSPTR	0x08
+#define D40_CHAN_REG_SSLNK	0x0C
+#define D40_CHAN_REG_SDCFG	0x10
+#define D40_CHAN_REG_SDELT	0x14
+#define D40_CHAN_REG_SDPTR	0x18
+#define D40_CHAN_REG_SDLNK	0x1C
+
+/* DMA Register Offsets */
+#define D40_DREG_GCC		0x000
+#define D40_DREG_PRTYP		0x004
+#define D40_DREG_PRSME		0x008
+#define D40_DREG_PRSMO		0x00C
+#define D40_DREG_PRMSE		0x010
+#define D40_DREG_PRMSO		0x014
+#define D40_DREG_PRMOE		0x018
+#define D40_DREG_PRMOO		0x01C
+#define D40_DREG_LCPA		0x020
+#define D40_DREG_LCLA		0x024
+#define D40_DREG_ACTIVE		0x050
+#define D40_DREG_ACTIVO		0x054
+#define D40_DREG_FSEB1		0x058
+#define D40_DREG_FSEB2		0x05C
+#define D40_DREG_PCMIS		0x060
+#define D40_DREG_PCICR		0x064
+#define D40_DREG_PCTIS		0x068
+#define D40_DREG_PCEIS		0x06C
+#define D40_DREG_LCMIS0		0x080
+#define D40_DREG_LCMIS1		0x084
+#define D40_DREG_LCMIS2		0x088
+#define D40_DREG_LCMIS3		0x08C
+#define D40_DREG_LCICR0		0x090
+#define D40_DREG_LCICR1		0x094
+#define D40_DREG_LCICR2		0x098
+#define D40_DREG_LCICR3		0x09C
+#define D40_DREG_LCTIS0		0x0A0
+#define D40_DREG_LCTIS1		0x0A4
+#define D40_DREG_LCTIS2		0x0A8
+#define D40_DREG_LCTIS3		0x0AC
+#define D40_DREG_LCEIS0		0x0B0
+#define D40_DREG_LCEIS1		0x0B4
+#define D40_DREG_LCEIS2		0x0B8
+#define D40_DREG_LCEIS3		0x0BC
+#define D40_DREG_STFU		0xFC8
+#define D40_DREG_ICFG		0xFCC
+#define D40_DREG_PERIPHID0	0xFE0
+#define D40_DREG_PERIPHID1	0xFE4
+#define D40_DREG_PERIPHID2	0xFE8
+#define D40_DREG_PERIPHID3	0xFEC
+#define D40_DREG_CELLID0	0xFF0
+#define D40_DREG_CELLID1	0xFF4
+#define D40_DREG_CELLID2	0xFF8
+#define D40_DREG_CELLID3	0xFFC
+
+/* LLI related structures */
+
+/**
+ * struct d40_phy_lli - The basic configration register for each physical
+ * channel.
+ *
+ * @reg_cfg: The configuration register.
+ * @reg_elt: The element register.
+ * @reg_ptr: The pointer register.
+ * @reg_lnk: The link register.
+ *
+ * These registers are set up for both physical and logical transfers
+ * Note that the bit in each register means differently in logical and
+ * physical(standard) mode.
+ *
+ * This struct must be 16 bytes aligned, and only contain physical registers
+ * since it will be directly accessed by the DMA.
+ */
+struct d40_phy_lli {
+	u32 reg_cfg;
+	u32 reg_elt;
+	u32 reg_ptr;
+	u32 reg_lnk;
+};
+
+/**
+ * struct d40_phy_lli_bidir - struct for a transfer.
+ *
+ * @src: Register settings for src channel.
+ * @dst: Register settings for dst channel.
+ * @dst_addr: Physical destination address.
+ * @src_addr: Physical source address.
+ *
+ * All DMA transfers have a source and a destination.
+ */
+
+struct d40_phy_lli_bidir {
+	struct d40_phy_lli	*src;
+	struct d40_phy_lli	*dst;
+	dma_addr_t		 dst_addr;
+	dma_addr_t		 src_addr;
+};
+
+
+/**
+ * struct d40_log_lli - logical lli configuration
+ *
+ * @lcsp02: Either maps to register lcsp0 if src or lcsp2 if dst.
+ * @lcsp13: Either maps to register lcsp1 if src or lcsp3 if dst.
+ *
+ * This struct must be 8 bytes aligned since it will be accessed directy by
+ * the DMA. Never add any none hw mapped registers to this struct.
+ */
+
+struct d40_log_lli {
+	u32 lcsp02;
+	u32 lcsp13;
+};
+
+/**
+ * struct d40_log_lli_bidir - For both src and dst
+ *
+ * @src: pointer to src lli configuration.
+ * @dst: pointer to dst lli configuration.
+ *
+ * You always have a src and a dst when doing DMA transfers.
+ */
+
+struct d40_log_lli_bidir {
+	struct d40_log_lli *src;
+	struct d40_log_lli *dst;
+};
+
+/**
+ * struct d40_log_lli_full - LCPA layout
+ *
+ * @lcsp0: Logical Channel Standard Param 0 - Src.
+ * @lcsp1: Logical Channel Standard Param 1 - Src.
+ * @lcsp2: Logical Channel Standard Param 2 - Dst.
+ * @lcsp3: Logical Channel Standard Param 3 - Dst.
+ *
+ * This struct maps to LCPA physical memory layout. Must map to
+ * the hw.
+ */
+struct d40_log_lli_full {
+	u32 lcsp0;
+	u32 lcsp1;
+	u32 lcsp2;
+	u32 lcsp3;
+};
+
+/**
+ * struct d40_def_lcsp - Default LCSP1 and LCSP3 settings
+ *
+ * @lcsp3: The default configuration for dst.
+ * @lcsp1: The default configuration for src.
+ */
+struct d40_def_lcsp {
+	u32 lcsp3;
+	u32 lcsp1;
+};
+
+/**
+ * struct d40_lcla_elem - Info for one LCA element.
+ *
+ * @src_id: logical channel src id
+ * @dst_id: logical channel dst id
+ * @src: LCPA formated src parameters
+ * @dst: LCPA formated dst parameters
+ *
+ */
+struct d40_lcla_elem {
+	int			src_id;
+	int			dst_id;
+	struct d40_log_lli     *src;
+	struct d40_log_lli     *dst;
+};
+
+/* Physical channels */
+
+void d40_phy_cfg(struct stedma40_chan_cfg *cfg,
+		 u32 *src_cfg, u32 *dst_cfg, bool is_log);
+
+void d40_log_cfg(struct stedma40_chan_cfg *cfg,
+		 u32 *lcsp1, u32 *lcsp2);
+
+int d40_phy_sg_to_lli(struct scatterlist *sg,
+		      int sg_len,
+		      dma_addr_t target,
+		      struct d40_phy_lli *lli,
+		      dma_addr_t lli_phys,
+		      u32 reg_cfg,
+		      u32 data_width,
+		      int psize,
+		      bool term_int);
+
+int d40_phy_fill_lli(struct d40_phy_lli *lli,
+		     dma_addr_t data,
+		     u32 data_size,
+		     int psize,
+		     dma_addr_t next_lli,
+		     u32 reg_cfg,
+		     bool term_int,
+		     u32 data_width,
+		     bool is_device);
+
+void d40_phy_lli_write(void __iomem *virtbase,
+		       u32 phy_chan_num,
+		       struct d40_phy_lli *lli_dst,
+		       struct d40_phy_lli *lli_src);
+
+/* Logical channels */
+
+void d40_log_fill_lli(struct d40_log_lli *lli,
+		      dma_addr_t data, u32 data_size,
+		      u32 lli_next_off, u32 reg_cfg,
+		      u32 data_width,
+		      bool term_int, bool addr_inc);
+
+int d40_log_sg_to_dev(struct d40_lcla_elem *lcla,
+		      struct scatterlist *sg,
+		      int sg_len,
+		      struct d40_log_lli_bidir *lli,
+		      struct d40_def_lcsp *lcsp,
+		      u32 src_data_width,
+		      u32 dst_data_width,
+		      enum dma_data_direction direction,
+		      bool term_int, dma_addr_t dev_addr, int max_len,
+		      int llis_per_log);
+
+void d40_log_lli_write(struct d40_log_lli_full *lcpa,
+		       struct d40_log_lli *lcla_src,
+		       struct d40_log_lli *lcla_dst,
+		       struct d40_log_lli *lli_dst,
+		       struct d40_log_lli *lli_src,
+		       int llis_per_log);
+
+int d40_log_sg_to_lli(int lcla_id,
+		      struct scatterlist *sg,
+		      int sg_len,
+		      struct d40_log_lli *lli_sg,
+		      u32 lcsp13, /* src or dst*/
+		      u32 data_width,
+		      bool term_int, int max_len, int llis_per_log);
+
+#endif /* STE_DMA40_LLI_H */
diff --git a/drivers/dma/timb_dma.c b/drivers/dma/timb_dma.c
new file mode 100644
index 000000000000..a1bf77c1993f
--- /dev/null
+++ b/drivers/dma/timb_dma.c
@@ -0,0 +1,860 @@
+/*
+ * timb_dma.c timberdale FPGA DMA driver
+ * Copyright (c) 2010 Intel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+/* Supports:
+ * Timberdale FPGA DMA engine
+ */
+
+#include <linux/dmaengine.h>
+#include <linux/dma-mapping.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/slab.h>
+
+#include <linux/timb_dma.h>
+
+#define DRIVER_NAME "timb-dma"
+
+/* Global DMA registers */
+#define TIMBDMA_ACR		0x34
+#define TIMBDMA_32BIT_ADDR	0x01
+
+#define TIMBDMA_ISR		0x080000
+#define TIMBDMA_IPR		0x080004
+#define TIMBDMA_IER		0x080008
+
+/* Channel specific registers */
+/* RX instances base addresses are 0x00, 0x40, 0x80 ...
+ * TX instances base addresses are 0x18, 0x58, 0x98 ...
+ */
+#define TIMBDMA_INSTANCE_OFFSET		0x40
+#define TIMBDMA_INSTANCE_TX_OFFSET	0x18
+
+/* RX registers, relative the instance base */
+#define TIMBDMA_OFFS_RX_DHAR	0x00
+#define TIMBDMA_OFFS_RX_DLAR	0x04
+#define TIMBDMA_OFFS_RX_LR	0x0C
+#define TIMBDMA_OFFS_RX_BLR	0x10
+#define TIMBDMA_OFFS_RX_ER	0x14
+#define TIMBDMA_RX_EN		0x01
+/* bytes per Row, video specific register
+ * which is placed after the TX registers...
+ */
+#define TIMBDMA_OFFS_RX_BPRR	0x30
+
+/* TX registers, relative the instance base */
+#define TIMBDMA_OFFS_TX_DHAR	0x00
+#define TIMBDMA_OFFS_TX_DLAR	0x04
+#define TIMBDMA_OFFS_TX_BLR	0x0C
+#define TIMBDMA_OFFS_TX_LR	0x14
+
+
+#define TIMB_DMA_DESC_SIZE	8
+
+struct timb_dma_desc {
+	struct list_head		desc_node;
+	struct dma_async_tx_descriptor	txd;
+	u8				*desc_list;
+	unsigned int			desc_list_len;
+	bool				interrupt;
+};
+
+struct timb_dma_chan {
+	struct dma_chan		chan;
+	void __iomem		*membase;
+	spinlock_t		lock; /* Used to protect data structures,
+					especially the lists and descriptors,
+					from races between the tasklet and calls
+					from above */
+	dma_cookie_t		last_completed_cookie;
+	bool			ongoing;
+	struct list_head	active_list;
+	struct list_head	queue;
+	struct list_head	free_list;
+	unsigned int		bytes_per_line;
+	enum dma_data_direction	direction;
+	unsigned int		descs; /* Descriptors to allocate */
+	unsigned int		desc_elems; /* number of elems per descriptor */
+};
+
+struct timb_dma {
+	struct dma_device	dma;
+	void __iomem		*membase;
+	struct tasklet_struct	tasklet;
+	struct timb_dma_chan	channels[0];
+};
+
+static struct device *chan2dev(struct dma_chan *chan)
+{
+	return &chan->dev->device;
+}
+static struct device *chan2dmadev(struct dma_chan *chan)
+{
+	return chan2dev(chan)->parent->parent;
+}
+
+static struct timb_dma *tdchantotd(struct timb_dma_chan *td_chan)
+{
+	int id = td_chan->chan.chan_id;
+	return (struct timb_dma *)((u8 *)td_chan -
+		id * sizeof(struct timb_dma_chan) - sizeof(struct timb_dma));
+}
+
+/* Must be called with the spinlock held */
+static void __td_enable_chan_irq(struct timb_dma_chan *td_chan)
+{
+	int id = td_chan->chan.chan_id;
+	struct timb_dma *td = tdchantotd(td_chan);
+	u32 ier;
+
+	/* enable interrupt for this channel */
+	ier = ioread32(td->membase + TIMBDMA_IER);
+	ier |= 1 << id;
+	dev_dbg(chan2dev(&td_chan->chan), "Enabling irq: %d, IER: 0x%x\n", id,
+		ier);
+	iowrite32(ier, td->membase + TIMBDMA_IER);
+}
+
+/* Should be called with the spinlock held */
+static bool __td_dma_done_ack(struct timb_dma_chan *td_chan)
+{
+	int id = td_chan->chan.chan_id;
+	struct timb_dma *td = (struct timb_dma *)((u8 *)td_chan -
+		id * sizeof(struct timb_dma_chan) - sizeof(struct timb_dma));
+	u32 isr;
+	bool done = false;
+
+	dev_dbg(chan2dev(&td_chan->chan), "Checking irq: %d, td: %p\n", id, td);
+
+	isr = ioread32(td->membase + TIMBDMA_ISR) & (1 << id);
+	if (isr) {
+		iowrite32(isr, td->membase + TIMBDMA_ISR);
+		done = true;
+	}
+
+	return done;
+}
+
+static void __td_unmap_desc(struct timb_dma_chan *td_chan, const u8 *dma_desc,
+	bool single)
+{
+	dma_addr_t addr;
+	int len;
+
+	addr = (dma_desc[7] << 24) | (dma_desc[6] << 16) | (dma_desc[5] << 8) |
+		dma_desc[4];
+
+	len = (dma_desc[3] << 8) | dma_desc[2];
+
+	if (single)
+		dma_unmap_single(chan2dev(&td_chan->chan), addr, len,
+			td_chan->direction);
+	else
+		dma_unmap_page(chan2dev(&td_chan->chan), addr, len,
+			td_chan->direction);
+}
+
+static void __td_unmap_descs(struct timb_dma_desc *td_desc, bool single)
+{
+	struct timb_dma_chan *td_chan = container_of(td_desc->txd.chan,
+		struct timb_dma_chan, chan);
+	u8 *descs;
+
+	for (descs = td_desc->desc_list; ; descs += TIMB_DMA_DESC_SIZE) {
+		__td_unmap_desc(td_chan, descs, single);
+		if (descs[0] & 0x02)
+			break;
+	}
+}
+
+static int td_fill_desc(struct timb_dma_chan *td_chan, u8 *dma_desc,
+	struct scatterlist *sg, bool last)
+{
+	if (sg_dma_len(sg) > USHRT_MAX) {
+		dev_err(chan2dev(&td_chan->chan), "Too big sg element\n");
+		return -EINVAL;
+	}
+
+	/* length must be word aligned */
+	if (sg_dma_len(sg) % sizeof(u32)) {
+		dev_err(chan2dev(&td_chan->chan), "Incorrect length: %d\n",
+			sg_dma_len(sg));
+		return -EINVAL;
+	}
+
+	dev_dbg(chan2dev(&td_chan->chan), "desc: %p, addr: %p\n",
+		dma_desc, (void *)sg_dma_address(sg));
+
+	dma_desc[7] = (sg_dma_address(sg) >> 24) & 0xff;
+	dma_desc[6] = (sg_dma_address(sg) >> 16) & 0xff;
+	dma_desc[5] = (sg_dma_address(sg) >> 8) & 0xff;
+	dma_desc[4] = (sg_dma_address(sg) >> 0) & 0xff;
+
+	dma_desc[3] = (sg_dma_len(sg) >> 8) & 0xff;
+	dma_desc[2] = (sg_dma_len(sg) >> 0) & 0xff;
+
+	dma_desc[1] = 0x00;
+	dma_desc[0] = 0x21 | (last ? 0x02 : 0); /* tran, valid */
+
+	return 0;
+}
+
+/* Must be called with the spinlock held */
+static void __td_start_dma(struct timb_dma_chan *td_chan)
+{
+	struct timb_dma_desc *td_desc;
+
+	if (td_chan->ongoing) {
+		dev_err(chan2dev(&td_chan->chan),
+			"Transfer already ongoing\n");
+		return;
+	}
+
+	td_desc = list_entry(td_chan->active_list.next, struct timb_dma_desc,
+		desc_node);
+
+	dev_dbg(chan2dev(&td_chan->chan),
+		"td_chan: %p, chan: %d, membase: %p\n",
+		td_chan, td_chan->chan.chan_id, td_chan->membase);
+
+	if (td_chan->direction == DMA_FROM_DEVICE) {
+
+		/* descriptor address */
+		iowrite32(0, td_chan->membase + TIMBDMA_OFFS_RX_DHAR);
+		iowrite32(td_desc->txd.phys, td_chan->membase +
+			TIMBDMA_OFFS_RX_DLAR);
+		/* Bytes per line */
+		iowrite32(td_chan->bytes_per_line, td_chan->membase +
+			TIMBDMA_OFFS_RX_BPRR);
+		/* enable RX */
+		iowrite32(TIMBDMA_RX_EN, td_chan->membase + TIMBDMA_OFFS_RX_ER);
+	} else {
+		/* address high */
+		iowrite32(0, td_chan->membase + TIMBDMA_OFFS_TX_DHAR);
+		iowrite32(td_desc->txd.phys, td_chan->membase +
+			TIMBDMA_OFFS_TX_DLAR);
+	}
+
+	td_chan->ongoing = true;
+
+	if (td_desc->interrupt)
+		__td_enable_chan_irq(td_chan);
+}
+
+static void __td_finish(struct timb_dma_chan *td_chan)
+{
+	dma_async_tx_callback		callback;
+	void				*param;
+	struct dma_async_tx_descriptor	*txd;
+	struct timb_dma_desc		*td_desc;
+
+	/* can happen if the descriptor is canceled */
+	if (list_empty(&td_chan->active_list))
+		return;
+
+	td_desc = list_entry(td_chan->active_list.next, struct timb_dma_desc,
+		desc_node);
+	txd = &td_desc->txd;
+
+	dev_dbg(chan2dev(&td_chan->chan), "descriptor %u complete\n",
+		txd->cookie);
+
+	/* make sure to stop the transfer */
+	if (td_chan->direction == DMA_FROM_DEVICE)
+		iowrite32(0, td_chan->membase + TIMBDMA_OFFS_RX_ER);
+/* Currently no support for stopping DMA transfers
+	else
+		iowrite32(0, td_chan->membase + TIMBDMA_OFFS_TX_DLAR);
+*/
+	td_chan->last_completed_cookie = txd->cookie;
+	td_chan->ongoing = false;
+
+	callback = txd->callback;
+	param = txd->callback_param;
+
+	list_move(&td_desc->desc_node, &td_chan->free_list);
+
+	if (!(txd->flags & DMA_COMPL_SKIP_SRC_UNMAP))
+		__td_unmap_descs(td_desc,
+			txd->flags & DMA_COMPL_SRC_UNMAP_SINGLE);
+
+	/*
+	 * The API requires that no submissions are done from a
+	 * callback, so we don't need to drop the lock here
+	 */
+	if (callback)
+		callback(param);
+}
+
+static u32 __td_ier_mask(struct timb_dma *td)
+{
+	int i;
+	u32 ret = 0;
+
+	for (i = 0; i < td->dma.chancnt; i++) {
+		struct timb_dma_chan *td_chan = td->channels + i;
+		if (td_chan->ongoing) {
+			struct timb_dma_desc *td_desc =
+				list_entry(td_chan->active_list.next,
+				struct timb_dma_desc, desc_node);
+			if (td_desc->interrupt)
+				ret |= 1 << i;
+		}
+	}
+
+	return ret;
+}
+
+static void __td_start_next(struct timb_dma_chan *td_chan)
+{
+	struct timb_dma_desc *td_desc;
+
+	BUG_ON(list_empty(&td_chan->queue));
+	BUG_ON(td_chan->ongoing);
+
+	td_desc = list_entry(td_chan->queue.next, struct timb_dma_desc,
+		desc_node);
+
+	dev_dbg(chan2dev(&td_chan->chan), "%s: started %u\n",
+		__func__, td_desc->txd.cookie);
+
+	list_move(&td_desc->desc_node, &td_chan->active_list);
+	__td_start_dma(td_chan);
+}
+
+static dma_cookie_t td_tx_submit(struct dma_async_tx_descriptor *txd)
+{
+	struct timb_dma_desc *td_desc = container_of(txd, struct timb_dma_desc,
+		txd);
+	struct timb_dma_chan *td_chan = container_of(txd->chan,
+		struct timb_dma_chan, chan);
+	dma_cookie_t cookie;
+
+	spin_lock_bh(&td_chan->lock);
+
+	cookie = txd->chan->cookie;
+	if (++cookie < 0)
+		cookie = 1;
+	txd->chan->cookie = cookie;
+	txd->cookie = cookie;
+
+	if (list_empty(&td_chan->active_list)) {
+		dev_dbg(chan2dev(txd->chan), "%s: started %u\n", __func__,
+			txd->cookie);
+		list_add_tail(&td_desc->desc_node, &td_chan->active_list);
+		__td_start_dma(td_chan);
+	} else {
+		dev_dbg(chan2dev(txd->chan), "tx_submit: queued %u\n",
+			txd->cookie);
+
+		list_add_tail(&td_desc->desc_node, &td_chan->queue);
+	}
+
+	spin_unlock_bh(&td_chan->lock);
+
+	return cookie;
+}
+
+static struct timb_dma_desc *td_alloc_init_desc(struct timb_dma_chan *td_chan)
+{
+	struct dma_chan *chan = &td_chan->chan;
+	struct timb_dma_desc *td_desc;
+	int err;
+
+	td_desc = kzalloc(sizeof(struct timb_dma_desc), GFP_KERNEL);
+	if (!td_desc) {
+		dev_err(chan2dev(chan), "Failed to alloc descriptor\n");
+		goto err;
+	}
+
+	td_desc->desc_list_len = td_chan->desc_elems * TIMB_DMA_DESC_SIZE;
+
+	td_desc->desc_list = kzalloc(td_desc->desc_list_len, GFP_KERNEL);
+	if (!td_desc->desc_list) {
+		dev_err(chan2dev(chan), "Failed to alloc descriptor\n");
+		goto err;
+	}
+
+	dma_async_tx_descriptor_init(&td_desc->txd, chan);
+	td_desc->txd.tx_submit = td_tx_submit;
+	td_desc->txd.flags = DMA_CTRL_ACK;
+
+	td_desc->txd.phys = dma_map_single(chan2dmadev(chan),
+		td_desc->desc_list, td_desc->desc_list_len, DMA_TO_DEVICE);
+
+	err = dma_mapping_error(chan2dmadev(chan), td_desc->txd.phys);
+	if (err) {
+		dev_err(chan2dev(chan), "DMA mapping error: %d\n", err);
+		goto err;
+	}
+
+	return td_desc;
+err:
+	kfree(td_desc->desc_list);
+	kfree(td_desc);
+
+	return NULL;
+
+}
+
+static void td_free_desc(struct timb_dma_desc *td_desc)
+{
+	dev_dbg(chan2dev(td_desc->txd.chan), "Freeing desc: %p\n", td_desc);
+	dma_unmap_single(chan2dmadev(td_desc->txd.chan), td_desc->txd.phys,
+		td_desc->desc_list_len, DMA_TO_DEVICE);
+
+	kfree(td_desc->desc_list);
+	kfree(td_desc);
+}
+
+static void td_desc_put(struct timb_dma_chan *td_chan,
+	struct timb_dma_desc *td_desc)
+{
+	dev_dbg(chan2dev(&td_chan->chan), "Putting desc: %p\n", td_desc);
+
+	spin_lock_bh(&td_chan->lock);
+	list_add(&td_desc->desc_node, &td_chan->free_list);
+	spin_unlock_bh(&td_chan->lock);
+}
+
+static struct timb_dma_desc *td_desc_get(struct timb_dma_chan *td_chan)
+{
+	struct timb_dma_desc *td_desc, *_td_desc;
+	struct timb_dma_desc *ret = NULL;
+
+	spin_lock_bh(&td_chan->lock);
+	list_for_each_entry_safe(td_desc, _td_desc, &td_chan->free_list,
+		desc_node) {
+		if (async_tx_test_ack(&td_desc->txd)) {
+			list_del(&td_desc->desc_node);
+			ret = td_desc;
+			break;
+		}
+		dev_dbg(chan2dev(&td_chan->chan), "desc %p not ACKed\n",
+			td_desc);
+	}
+	spin_unlock_bh(&td_chan->lock);
+
+	return ret;
+}
+
+static int td_alloc_chan_resources(struct dma_chan *chan)
+{
+	struct timb_dma_chan *td_chan =
+		container_of(chan, struct timb_dma_chan, chan);
+	int i;
+
+	dev_dbg(chan2dev(chan), "%s: entry\n", __func__);
+
+	BUG_ON(!list_empty(&td_chan->free_list));
+	for (i = 0; i < td_chan->descs; i++) {
+		struct timb_dma_desc *td_desc = td_alloc_init_desc(td_chan);
+		if (!td_desc) {
+			if (i)
+				break;
+			else {
+				dev_err(chan2dev(chan),
+					"Couldnt allocate any descriptors\n");
+				return -ENOMEM;
+			}
+		}
+
+		td_desc_put(td_chan, td_desc);
+	}
+
+	spin_lock_bh(&td_chan->lock);
+	td_chan->last_completed_cookie = 1;
+	chan->cookie = 1;
+	spin_unlock_bh(&td_chan->lock);
+
+	return 0;
+}
+
+static void td_free_chan_resources(struct dma_chan *chan)
+{
+	struct timb_dma_chan *td_chan =
+		container_of(chan, struct timb_dma_chan, chan);
+	struct timb_dma_desc *td_desc, *_td_desc;
+	LIST_HEAD(list);
+
+	dev_dbg(chan2dev(chan), "%s: Entry\n", __func__);
+
+	/* check that all descriptors are free */
+	BUG_ON(!list_empty(&td_chan->active_list));
+	BUG_ON(!list_empty(&td_chan->queue));
+
+	spin_lock_bh(&td_chan->lock);
+	list_splice_init(&td_chan->free_list, &list);
+	spin_unlock_bh(&td_chan->lock);
+
+	list_for_each_entry_safe(td_desc, _td_desc, &list, desc_node) {
+		dev_dbg(chan2dev(chan), "%s: Freeing desc: %p\n", __func__,
+			td_desc);
+		td_free_desc(td_desc);
+	}
+}
+
+static enum dma_status td_tx_status(struct dma_chan *chan, dma_cookie_t cookie,
+				    struct dma_tx_state *txstate)
+{
+	struct timb_dma_chan *td_chan =
+		container_of(chan, struct timb_dma_chan, chan);
+	dma_cookie_t		last_used;
+	dma_cookie_t		last_complete;
+	int			ret;
+
+	dev_dbg(chan2dev(chan), "%s: Entry\n", __func__);
+
+	last_complete = td_chan->last_completed_cookie;
+	last_used = chan->cookie;
+
+	ret = dma_async_is_complete(cookie, last_complete, last_used);
+
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
+
+	dev_dbg(chan2dev(chan),
+		"%s: exit, ret: %d, last_complete: %d, last_used: %d\n",
+		__func__, ret, last_complete, last_used);
+
+	return ret;
+}
+
+static void td_issue_pending(struct dma_chan *chan)
+{
+	struct timb_dma_chan *td_chan =
+		container_of(chan, struct timb_dma_chan, chan);
+
+	dev_dbg(chan2dev(chan), "%s: Entry\n", __func__);
+	spin_lock_bh(&td_chan->lock);
+
+	if (!list_empty(&td_chan->active_list))
+		/* transfer ongoing */
+		if (__td_dma_done_ack(td_chan))
+			__td_finish(td_chan);
+
+	if (list_empty(&td_chan->active_list) && !list_empty(&td_chan->queue))
+		__td_start_next(td_chan);
+
+	spin_unlock_bh(&td_chan->lock);
+}
+
+static struct dma_async_tx_descriptor *td_prep_slave_sg(struct dma_chan *chan,
+	struct scatterlist *sgl, unsigned int sg_len,
+	enum dma_data_direction direction, unsigned long flags)
+{
+	struct timb_dma_chan *td_chan =
+		container_of(chan, struct timb_dma_chan, chan);
+	struct timb_dma_desc *td_desc;
+	struct scatterlist *sg;
+	unsigned int i;
+	unsigned int desc_usage = 0;
+
+	if (!sgl || !sg_len) {
+		dev_err(chan2dev(chan), "%s: No SG list\n", __func__);
+		return NULL;
+	}
+
+	/* even channels are for RX, odd for TX */
+	if (td_chan->direction != direction) {
+		dev_err(chan2dev(chan),
+			"Requesting channel in wrong direction\n");
+		return NULL;
+	}
+
+	td_desc = td_desc_get(td_chan);
+	if (!td_desc) {
+		dev_err(chan2dev(chan), "Not enough descriptors available\n");
+		return NULL;
+	}
+
+	td_desc->interrupt = (flags & DMA_PREP_INTERRUPT) != 0;
+
+	for_each_sg(sgl, sg, sg_len, i) {
+		int err;
+		if (desc_usage > td_desc->desc_list_len) {
+			dev_err(chan2dev(chan), "No descriptor space\n");
+			return NULL;
+		}
+
+		err = td_fill_desc(td_chan, td_desc->desc_list + desc_usage, sg,
+			i == (sg_len - 1));
+		if (err) {
+			dev_err(chan2dev(chan), "Failed to update desc: %d\n",
+				err);
+			td_desc_put(td_chan, td_desc);
+			return NULL;
+		}
+		desc_usage += TIMB_DMA_DESC_SIZE;
+	}
+
+	dma_sync_single_for_device(chan2dmadev(chan), td_desc->txd.phys,
+		td_desc->desc_list_len, DMA_TO_DEVICE);
+
+	return &td_desc->txd;
+}
+
+static int td_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+		      unsigned long arg)
+{
+	struct timb_dma_chan *td_chan =
+		container_of(chan, struct timb_dma_chan, chan);
+	struct timb_dma_desc *td_desc, *_td_desc;
+
+	dev_dbg(chan2dev(chan), "%s: Entry\n", __func__);
+
+	if (cmd != DMA_TERMINATE_ALL)
+		return -ENXIO;
+
+	/* first the easy part, put the queue into the free list */
+	spin_lock_bh(&td_chan->lock);
+	list_for_each_entry_safe(td_desc, _td_desc, &td_chan->queue,
+		desc_node)
+		list_move(&td_desc->desc_node, &td_chan->free_list);
+
+	/* now tear down the runnning */
+	__td_finish(td_chan);
+	spin_unlock_bh(&td_chan->lock);
+
+	return 0;
+}
+
+static void td_tasklet(unsigned long data)
+{
+	struct timb_dma *td = (struct timb_dma *)data;
+	u32 isr;
+	u32 ipr;
+	u32 ier;
+	int i;
+
+	isr = ioread32(td->membase + TIMBDMA_ISR);
+	ipr = isr & __td_ier_mask(td);
+
+	/* ack the interrupts */
+	iowrite32(ipr, td->membase + TIMBDMA_ISR);
+
+	for (i = 0; i < td->dma.chancnt; i++)
+		if (ipr & (1 << i)) {
+			struct timb_dma_chan *td_chan = td->channels + i;
+			spin_lock(&td_chan->lock);
+			__td_finish(td_chan);
+			if (!list_empty(&td_chan->queue))
+				__td_start_next(td_chan);
+			spin_unlock(&td_chan->lock);
+		}
+
+	ier = __td_ier_mask(td);
+	iowrite32(ier, td->membase + TIMBDMA_IER);
+}
+
+
+static irqreturn_t td_irq(int irq, void *devid)
+{
+	struct timb_dma *td = devid;
+	u32 ipr = ioread32(td->membase + TIMBDMA_IPR);
+
+	if (ipr) {
+		/* disable interrupts, will be re-enabled in tasklet */
+		iowrite32(0, td->membase + TIMBDMA_IER);
+
+		tasklet_schedule(&td->tasklet);
+
+		return IRQ_HANDLED;
+	} else
+		return IRQ_NONE;
+}
+
+
+static int __devinit td_probe(struct platform_device *pdev)
+{
+	struct timb_dma_platform_data *pdata = pdev->dev.platform_data;
+	struct timb_dma *td;
+	struct resource *iomem;
+	int irq;
+	int err;
+	int i;
+
+	if (!pdata) {
+		dev_err(&pdev->dev, "No platform data\n");
+		return -EINVAL;
+	}
+
+	iomem = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	if (!iomem)
+		return -EINVAL;
+
+	irq = platform_get_irq(pdev, 0);
+	if (irq < 0)
+		return irq;
+
+	if (!request_mem_region(iomem->start, resource_size(iomem),
+		DRIVER_NAME))
+		return -EBUSY;
+
+	td  = kzalloc(sizeof(struct timb_dma) +
+		sizeof(struct timb_dma_chan) * pdata->nr_channels, GFP_KERNEL);
+	if (!td) {
+		err = -ENOMEM;
+		goto err_release_region;
+	}
+
+	dev_dbg(&pdev->dev, "Allocated TD: %p\n", td);
+
+	td->membase = ioremap(iomem->start, resource_size(iomem));
+	if (!td->membase) {
+		dev_err(&pdev->dev, "Failed to remap I/O memory\n");
+		err = -ENOMEM;
+		goto err_free_mem;
+	}
+
+	/* 32bit addressing */
+	iowrite32(TIMBDMA_32BIT_ADDR, td->membase + TIMBDMA_ACR);
+
+	/* disable and clear any interrupts */
+	iowrite32(0x0, td->membase + TIMBDMA_IER);
+	iowrite32(0xFFFFFFFF, td->membase + TIMBDMA_ISR);
+
+	tasklet_init(&td->tasklet, td_tasklet, (unsigned long)td);
+
+	err = request_irq(irq, td_irq, IRQF_SHARED, DRIVER_NAME, td);
+	if (err) {
+		dev_err(&pdev->dev, "Failed to request IRQ\n");
+		goto err_tasklet_kill;
+	}
+
+	td->dma.device_alloc_chan_resources	= td_alloc_chan_resources;
+	td->dma.device_free_chan_resources	= td_free_chan_resources;
+	td->dma.device_tx_status		= td_tx_status;
+	td->dma.device_issue_pending		= td_issue_pending;
+
+	dma_cap_set(DMA_SLAVE, td->dma.cap_mask);
+	dma_cap_set(DMA_PRIVATE, td->dma.cap_mask);
+	td->dma.device_prep_slave_sg = td_prep_slave_sg;
+	td->dma.device_control = td_control;
+
+	td->dma.dev = &pdev->dev;
+
+	INIT_LIST_HEAD(&td->dma.channels);
+
+	for (i = 0; i < pdata->nr_channels; i++, td->dma.chancnt++) {
+		struct timb_dma_chan *td_chan = &td->channels[i];
+		struct timb_dma_platform_data_channel *pchan =
+			pdata->channels + i;
+
+		/* even channels are RX, odd are TX */
+		if (((i % 2) && pchan->rx) || (!(i % 2) && !pchan->rx)) {
+			dev_err(&pdev->dev, "Wrong channel configuration\n");
+			err = -EINVAL;
+			goto err_tasklet_kill;
+		}
+
+		td_chan->chan.device = &td->dma;
+		td_chan->chan.cookie = 1;
+		td_chan->chan.chan_id = i;
+		spin_lock_init(&td_chan->lock);
+		INIT_LIST_HEAD(&td_chan->active_list);
+		INIT_LIST_HEAD(&td_chan->queue);
+		INIT_LIST_HEAD(&td_chan->free_list);
+
+		td_chan->descs = pchan->descriptors;
+		td_chan->desc_elems = pchan->descriptor_elements;
+		td_chan->bytes_per_line = pchan->bytes_per_line;
+		td_chan->direction = pchan->rx ? DMA_FROM_DEVICE :
+			DMA_TO_DEVICE;
+
+		td_chan->membase = td->membase +
+			(i / 2) * TIMBDMA_INSTANCE_OFFSET +
+			(pchan->rx ? 0 : TIMBDMA_INSTANCE_TX_OFFSET);
+
+		dev_dbg(&pdev->dev, "Chan: %d, membase: %p\n",
+			i, td_chan->membase);
+
+		list_add_tail(&td_chan->chan.device_node, &td->dma.channels);
+	}
+
+	err = dma_async_device_register(&td->dma);
+	if (err) {
+		dev_err(&pdev->dev, "Failed to register async device\n");
+		goto err_free_irq;
+	}
+
+	platform_set_drvdata(pdev, td);
+
+	dev_dbg(&pdev->dev, "Probe result: %d\n", err);
+	return err;
+
+err_free_irq:
+	free_irq(irq, td);
+err_tasklet_kill:
+	tasklet_kill(&td->tasklet);
+	iounmap(td->membase);
+err_free_mem:
+	kfree(td);
+err_release_region:
+	release_mem_region(iomem->start, resource_size(iomem));
+
+	return err;
+
+}
+
+static int __devexit td_remove(struct platform_device *pdev)
+{
+	struct timb_dma *td = platform_get_drvdata(pdev);
+	struct resource *iomem = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	int irq = platform_get_irq(pdev, 0);
+
+	dma_async_device_unregister(&td->dma);
+	free_irq(irq, td);
+	tasklet_kill(&td->tasklet);
+	iounmap(td->membase);
+	kfree(td);
+	release_mem_region(iomem->start, resource_size(iomem));
+
+	platform_set_drvdata(pdev, NULL);
+
+	dev_dbg(&pdev->dev, "Removed...\n");
+	return 0;
+}
+
+static struct platform_driver td_driver = {
+	.driver = {
+		.name	= DRIVER_NAME,
+		.owner  = THIS_MODULE,
+	},
+	.probe	= td_probe,
+	.remove	= __exit_p(td_remove),
+};
+
+static int __init td_init(void)
+{
+	return platform_driver_register(&td_driver);
+}
+module_init(td_init);
+
+static void __exit td_exit(void)
+{
+	platform_driver_unregister(&td_driver);
+}
+module_exit(td_exit);
+
+MODULE_LICENSE("GPL v2");
+MODULE_DESCRIPTION("Timberdale DMA controller driver");
+MODULE_AUTHOR("Pelagicore AB <info@pelagicore.com>");
+MODULE_ALIAS("platform:"DRIVER_NAME);
diff --git a/drivers/dma/txx9dmac.c b/drivers/dma/txx9dmac.c
index 75fcf1ac8bb7..cbd83e362b5e 100644
--- a/drivers/dma/txx9dmac.c
+++ b/drivers/dma/txx9dmac.c
@@ -938,12 +938,17 @@ txx9dmac_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
 	return &first->txd;
 }
 
-static void txx9dmac_terminate_all(struct dma_chan *chan)
+static int txx9dmac_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+			    unsigned long arg)
 {
 	struct txx9dmac_chan *dc = to_txx9dmac_chan(chan);
 	struct txx9dmac_desc *desc, *_desc;
 	LIST_HEAD(list);
 
+	/* Only supports DMA_TERMINATE_ALL */
+	if (cmd != DMA_TERMINATE_ALL)
+		return -EINVAL;
+
 	dev_vdbg(chan2dev(chan), "terminate_all\n");
 	spin_lock_bh(&dc->lock);
 
@@ -958,12 +963,13 @@ static void txx9dmac_terminate_all(struct dma_chan *chan)
 	/* Flush all pending and queued descriptors */
 	list_for_each_entry_safe(desc, _desc, &list, desc_node)
 		txx9dmac_descriptor_complete(dc, desc);
+
+	return 0;
 }
 
 static enum dma_status
-txx9dmac_is_tx_complete(struct dma_chan *chan,
-			dma_cookie_t cookie,
-		dma_cookie_t *done, dma_cookie_t *used)
+txx9dmac_tx_status(struct dma_chan *chan, dma_cookie_t cookie,
+		   struct dma_tx_state *txstate)
 {
 	struct txx9dmac_chan *dc = to_txx9dmac_chan(chan);
 	dma_cookie_t last_used;
@@ -985,10 +991,7 @@ txx9dmac_is_tx_complete(struct dma_chan *chan,
 		ret = dma_async_is_complete(cookie, last_complete, last_used);
 	}
 
-	if (done)
-		*done = last_complete;
-	if (used)
-		*used = last_used;
+	dma_set_tx_state(txstate, last_complete, last_used, 0);
 
 	return ret;
 }
@@ -1153,8 +1156,8 @@ static int __init txx9dmac_chan_probe(struct platform_device *pdev)
 	dc->dma.dev = &pdev->dev;
 	dc->dma.device_alloc_chan_resources = txx9dmac_alloc_chan_resources;
 	dc->dma.device_free_chan_resources = txx9dmac_free_chan_resources;
-	dc->dma.device_terminate_all = txx9dmac_terminate_all;
-	dc->dma.device_is_tx_complete = txx9dmac_is_tx_complete;
+	dc->dma.device_control = txx9dmac_control;
+	dc->dma.device_tx_status = txx9dmac_tx_status;
 	dc->dma.device_issue_pending = txx9dmac_issue_pending;
 	if (pdata && pdata->memcpy_chan == ch) {
 		dc->dma.device_prep_dma_memcpy = txx9dmac_prep_dma_memcpy;
diff --git a/drivers/edac/Kconfig b/drivers/edac/Kconfig
index 55c9c59b3f71..aedef7941b22 100644
--- a/drivers/edac/Kconfig
+++ b/drivers/edac/Kconfig
@@ -69,6 +69,9 @@ config EDAC_MM_EDAC
 	  occurred so that a particular failing memory module can be
 	  replaced.  If unsure, select 'Y'.
 
+config EDAC_MCE
+	bool
+
 config EDAC_AMD64
 	tristate "AMD64 (Opteron, Athlon64) K8, F10h, F11h"
 	depends on EDAC_MM_EDAC && K8_NB && X86_64 && PCI && EDAC_DECODE_MCE
@@ -166,6 +169,16 @@ config EDAC_I5400
 	  Support for error detection and correction the Intel
 	  i5400 MCH chipset (Seaburg).
 
+config EDAC_I7CORE
+	tristate "Intel i7 Core (Nehalem) processors"
+	depends on EDAC_MM_EDAC && PCI && X86
+	select EDAC_MCE
+	help
+	  Support for error detection and correction the Intel
+	  i7 Core (Nehalem) Integrated Memory Controller that exists on
+	  newer processors like i7 Core, i7 Core Extreme, Xeon 35xx
+	  and Xeon 55xx processors.
+
 config EDAC_I82860
 	tristate "Intel 82860"
 	depends on EDAC_MM_EDAC && PCI && X86_32
diff --git a/drivers/edac/Makefile b/drivers/edac/Makefile
index bc5dc232a0fb..ca6b1bb24ccc 100644
--- a/drivers/edac/Makefile
+++ b/drivers/edac/Makefile
@@ -8,6 +8,7 @@
 
 obj-$(CONFIG_EDAC)			:= edac_stub.o
 obj-$(CONFIG_EDAC_MM_EDAC)		+= edac_core.o
+obj-$(CONFIG_EDAC_MCE)			+= edac_mce.o
 
 edac_core-objs	:= edac_mc.o edac_device.o edac_mc_sysfs.o edac_pci_sysfs.o
 edac_core-objs	+= edac_module.o edac_device_sysfs.o
@@ -23,6 +24,7 @@ obj-$(CONFIG_EDAC_CPC925)		+= cpc925_edac.o
 obj-$(CONFIG_EDAC_I5000)		+= i5000_edac.o
 obj-$(CONFIG_EDAC_I5100)		+= i5100_edac.o
 obj-$(CONFIG_EDAC_I5400)		+= i5400_edac.o
+obj-$(CONFIG_EDAC_I7CORE)		+= i7core_edac.o
 obj-$(CONFIG_EDAC_E7XXX)		+= e7xxx_edac.o
 obj-$(CONFIG_EDAC_E752X)		+= e752x_edac.o
 obj-$(CONFIG_EDAC_I82443BXGX)		+= i82443bxgx_edac.o
diff --git a/drivers/edac/amd76x_edac.c b/drivers/edac/amd76x_edac.c
index f2330f81cb5e..cace0a7b707a 100644
--- a/drivers/edac/amd76x_edac.c
+++ b/drivers/edac/amd76x_edac.c
@@ -294,7 +294,7 @@ static int __devinit amd76x_init_one(struct pci_dev *pdev,
 {
 	debugf0("%s()\n", __func__);
 
-	/* don't need to call pci_device_enable() */
+	/* don't need to call pci_enable_device() */
 	return amd76x_probe1(pdev, ent->driver_data);
 }
 
diff --git a/drivers/edac/edac_core.h b/drivers/edac/edac_core.h
index 001b2e797fb3..efca9343d26a 100644
--- a/drivers/edac/edac_core.h
+++ b/drivers/edac/edac_core.h
@@ -341,12 +341,30 @@ struct csrow_info {
 	struct channel_info *channels;
 };
 
+struct mcidev_sysfs_group {
+	const char *name;				/* group name */
+	struct mcidev_sysfs_attribute *mcidev_attr;	/* group attributes */
+};
+
+struct mcidev_sysfs_group_kobj {
+	struct list_head list;		/* list for all instances within a mc */
+
+	struct kobject kobj;		/* kobj for the group */
+
+	struct mcidev_sysfs_group *grp;	/* group description table */
+	struct mem_ctl_info *mci;	/* the parent */
+};
+
 /* mcidev_sysfs_attribute structure
  *	used for driver sysfs attributes and in mem_ctl_info
  * 	sysfs top level entries
  */
 struct mcidev_sysfs_attribute {
-        struct attribute attr;
+	/* It should use either attr or grp */
+	struct attribute attr;
+	struct mcidev_sysfs_group *grp;	/* Points to a group of attributes */
+
+	/* Ops for show/store values at the attribute - not used on group */
         ssize_t (*show)(struct mem_ctl_info *,char *);
         ssize_t (*store)(struct mem_ctl_info *, const char *,size_t);
 };
@@ -424,6 +442,9 @@ struct mem_ctl_info {
 	/* edac sysfs device control */
 	struct kobject edac_mci_kobj;
 
+	/* list for all grp instances within a mc */
+	struct list_head grp_kobj_list;
+
 	/* Additional top controller level attributes, but specified
 	 * by the low level driver.
 	 *
diff --git a/drivers/edac/edac_mc_sysfs.c b/drivers/edac/edac_mc_sysfs.c
index 418b65f1a1da..c200c2fd43ea 100644
--- a/drivers/edac/edac_mc_sysfs.c
+++ b/drivers/edac/edac_mc_sysfs.c
@@ -557,6 +557,8 @@ static ssize_t mcidev_show(struct kobject *kobj, struct attribute *attr,
 	struct mem_ctl_info *mem_ctl_info = to_mci(kobj);
 	struct mcidev_sysfs_attribute *mcidev_attr = to_mcidev_attr(attr);
 
+	debugf1("%s() mem_ctl_info %p\n", __func__, mem_ctl_info);
+
 	if (mcidev_attr->show)
 		return mcidev_attr->show(mem_ctl_info, buffer);
 
@@ -569,6 +571,8 @@ static ssize_t mcidev_store(struct kobject *kobj, struct attribute *attr,
 	struct mem_ctl_info *mem_ctl_info = to_mci(kobj);
 	struct mcidev_sysfs_attribute *mcidev_attr = to_mcidev_attr(attr);
 
+	debugf1("%s() mem_ctl_info %p\n", __func__, mem_ctl_info);
+
 	if (mcidev_attr->store)
 		return mcidev_attr->store(mem_ctl_info, buffer, count);
 
@@ -726,28 +730,118 @@ void edac_mc_unregister_sysfs_main_kobj(struct mem_ctl_info *mci)
 
 #define EDAC_DEVICE_SYMLINK	"device"
 
+#define grp_to_mci(k) (container_of(k, struct mcidev_sysfs_group_kobj, kobj)->mci)
+
+/* MCI show/store functions for top most object */
+static ssize_t inst_grp_show(struct kobject *kobj, struct attribute *attr,
+			char *buffer)
+{
+	struct mem_ctl_info *mem_ctl_info = grp_to_mci(kobj);
+	struct mcidev_sysfs_attribute *mcidev_attr = to_mcidev_attr(attr);
+
+	debugf1("%s() mem_ctl_info %p\n", __func__, mem_ctl_info);
+
+	if (mcidev_attr->show)
+		return mcidev_attr->show(mem_ctl_info, buffer);
+
+	return -EIO;
+}
+
+static ssize_t inst_grp_store(struct kobject *kobj, struct attribute *attr,
+			const char *buffer, size_t count)
+{
+	struct mem_ctl_info *mem_ctl_info = grp_to_mci(kobj);
+	struct mcidev_sysfs_attribute *mcidev_attr = to_mcidev_attr(attr);
+
+	debugf1("%s() mem_ctl_info %p\n", __func__, mem_ctl_info);
+
+	if (mcidev_attr->store)
+		return mcidev_attr->store(mem_ctl_info, buffer, count);
+
+	return -EIO;
+}
+
+/* No memory to release for this kobj */
+static void edac_inst_grp_release(struct kobject *kobj)
+{
+	struct mcidev_sysfs_group_kobj *grp;
+	struct mem_ctl_info *mci;
+
+	debugf1("%s()\n", __func__);
+
+	grp = container_of(kobj, struct mcidev_sysfs_group_kobj, kobj);
+	mci = grp->mci;
+
+	kobject_put(&mci->edac_mci_kobj);
+}
+
+/* Intermediate show/store table */
+static struct sysfs_ops inst_grp_ops = {
+	.show = inst_grp_show,
+	.store = inst_grp_store
+};
+
+/* the kobj_type instance for a instance group */
+static struct kobj_type ktype_inst_grp = {
+	.release = edac_inst_grp_release,
+	.sysfs_ops = &inst_grp_ops,
+};
+
+
 /*
  * edac_create_mci_instance_attributes
- *	create MC driver specific attributes at the topmost level
- *	directory of this mci instance.
+ *	create MC driver specific attributes bellow an specified kobj
+ * This routine calls itself recursively, in order to create an entire
+ * object tree.
  */
-static int edac_create_mci_instance_attributes(struct mem_ctl_info *mci)
+static int edac_create_mci_instance_attributes(struct mem_ctl_info *mci,
+				struct mcidev_sysfs_attribute *sysfs_attrib,
+				struct kobject *kobj)
 {
 	int err;
-	struct mcidev_sysfs_attribute *sysfs_attrib;
 
-	/* point to the start of the array and iterate over it
-	 * adding each attribute listed to this mci instance's kobject
-	 */
-	sysfs_attrib = mci->mc_driver_sysfs_attributes;
+	debugf1("%s()\n", __func__);
+
+	while (sysfs_attrib) {
+		if (sysfs_attrib->grp) {
+			struct mcidev_sysfs_group_kobj *grp_kobj;
+
+			grp_kobj = kzalloc(sizeof(*grp_kobj), GFP_KERNEL);
+			if (!grp_kobj)
+				return -ENOMEM;
+
+			list_add_tail(&grp_kobj->list, &mci->grp_kobj_list);
+
+			grp_kobj->grp = sysfs_attrib->grp;
+			grp_kobj->mci = mci;
+
+			debugf0("%s() grp %s, mci %p\n", __func__,
+				sysfs_attrib->grp->name, mci);
+
+			err = kobject_init_and_add(&grp_kobj->kobj,
+						&ktype_inst_grp,
+						&mci->edac_mci_kobj,
+						sysfs_attrib->grp->name);
+			if (err)
+				return err;
+
+			err = edac_create_mci_instance_attributes(mci,
+					grp_kobj->grp->mcidev_attr,
+					&grp_kobj->kobj);
+
+			if (err)
+				return err;
+		} else if (sysfs_attrib->attr.name) {
+			debugf0("%s() file %s\n", __func__,
+				sysfs_attrib->attr.name);
+
+			err = sysfs_create_file(kobj, &sysfs_attrib->attr);
+		} else
+			break;
 
-	while (sysfs_attrib && sysfs_attrib->attr.name) {
-		err = sysfs_create_file(&mci->edac_mci_kobj,
-					(struct attribute*) sysfs_attrib);
 		if (err) {
 			return err;
 		}
-
 		sysfs_attrib++;
 	}
 
@@ -759,21 +853,44 @@ static int edac_create_mci_instance_attributes(struct mem_ctl_info *mci)
  *	remove MC driver specific attributes at the topmost level
  *	directory of this mci instance.
  */
-static void edac_remove_mci_instance_attributes(struct mem_ctl_info *mci)
+static void edac_remove_mci_instance_attributes(struct mem_ctl_info *mci,
+				struct mcidev_sysfs_attribute *sysfs_attrib,
+				struct kobject *kobj, int count)
 {
-	struct mcidev_sysfs_attribute *sysfs_attrib;
+	struct mcidev_sysfs_group_kobj *grp_kobj, *tmp;
 
-	/* point to the start of the array and iterate over it
-	 * adding each attribute listed to this mci instance's kobject
-	 */
-	sysfs_attrib = mci->mc_driver_sysfs_attributes;
+	debugf1("%s()\n", __func__);
 
-	/* loop if there are attributes and until we hit a NULL entry */
-	while (sysfs_attrib && sysfs_attrib->attr.name) {
-		sysfs_remove_file(&mci->edac_mci_kobj,
-					(struct attribute *) sysfs_attrib);
+	/*
+	 * loop if there are attributes and until we hit a NULL entry
+	 * Remove first all the atributes
+	 */
+	while (sysfs_attrib) {
+		if (sysfs_attrib->grp) {
+			list_for_each_entry(grp_kobj, &mci->grp_kobj_list,
+					    list)
+				if (grp_kobj->grp == sysfs_attrib->grp)
+					edac_remove_mci_instance_attributes(mci,
+						    grp_kobj->grp->mcidev_attr,
+						    &grp_kobj->kobj, count + 1);
+		} else if (sysfs_attrib->attr.name) {
+			debugf0("%s() file %s\n", __func__,
+				sysfs_attrib->attr.name);
+			sysfs_remove_file(kobj, &sysfs_attrib->attr);
+		} else
+			break;
 		sysfs_attrib++;
 	}
+
+	/*
+	 * Now that all attributes got removed, it is save to remove all groups
+	 */
+	if (!count)
+		list_for_each_entry_safe(grp_kobj, tmp, &mci->grp_kobj_list,
+					 list) {
+			debugf0("%s() grp %s\n", __func__, grp_kobj->grp->name);
+			kobject_put(&grp_kobj->kobj);
+		}
 }
 
 
@@ -794,6 +911,8 @@ int edac_create_sysfs_mci_device(struct mem_ctl_info *mci)
 
 	debugf0("%s() idx=%d\n", __func__, mci->mc_idx);
 
+	INIT_LIST_HEAD(&mci->grp_kobj_list);
+
 	/* create a symlink for the device */
 	err = sysfs_create_link(kobj_mci, &mci->dev->kobj,
 				EDAC_DEVICE_SYMLINK);
@@ -806,7 +925,9 @@ int edac_create_sysfs_mci_device(struct mem_ctl_info *mci)
 	 * then create them now for the driver.
 	 */
 	if (mci->mc_driver_sysfs_attributes) {
-		err = edac_create_mci_instance_attributes(mci);
+		err = edac_create_mci_instance_attributes(mci,
+					mci->mc_driver_sysfs_attributes,
+					&mci->edac_mci_kobj);
 		if (err) {
 			debugf1("%s() failure to create mci attributes\n",
 				__func__);
@@ -841,7 +962,8 @@ fail1:
 	}
 
 	/* remove the mci instance's attributes, if any */
-	edac_remove_mci_instance_attributes(mci);
+	edac_remove_mci_instance_attributes(mci,
+		mci->mc_driver_sysfs_attributes, &mci->edac_mci_kobj, 0);
 
 	/* remove the symlink */
 	sysfs_remove_link(kobj_mci, EDAC_DEVICE_SYMLINK);
@@ -875,8 +997,9 @@ void edac_remove_sysfs_mci_device(struct mem_ctl_info *mci)
 	debugf0("%s()  remove_mci_instance\n", __func__);
 
 	/* remove this mci instance's attribtes */
-	edac_remove_mci_instance_attributes(mci);
-
+	edac_remove_mci_instance_attributes(mci,
+					    mci->mc_driver_sysfs_attributes,
+					    &mci->edac_mci_kobj, 0);
 	debugf0("%s()  unregister this mci kobj\n", __func__);
 
 	/* unregister this instance's kobject */
diff --git a/drivers/edac/edac_mce.c b/drivers/edac/edac_mce.c
new file mode 100644
index 000000000000..9ccdc5b140e7
--- /dev/null
+++ b/drivers/edac/edac_mce.c
@@ -0,0 +1,61 @@
+/* Provides edac interface to mcelog events
+ *
+ * This file may be distributed under the terms of the
+ * GNU General Public License version 2.
+ *
+ * Copyright (c) 2009 by:
+ *	 Mauro Carvalho Chehab <mchehab@redhat.com>
+ *
+ * Red Hat Inc. http://www.redhat.com
+ */
+
+#include <linux/module.h>
+#include <linux/edac_mce.h>
+#include <asm/mce.h>
+
+int edac_mce_enabled;
+EXPORT_SYMBOL_GPL(edac_mce_enabled);
+
+
+/*
+ * Extension interface
+ */
+
+static LIST_HEAD(edac_mce_list);
+static DEFINE_MUTEX(edac_mce_lock);
+
+int edac_mce_register(struct edac_mce *edac_mce)
+{
+	mutex_lock(&edac_mce_lock);
+	list_add_tail(&edac_mce->list, &edac_mce_list);
+	mutex_unlock(&edac_mce_lock);
+	return 0;
+}
+EXPORT_SYMBOL(edac_mce_register);
+
+void edac_mce_unregister(struct edac_mce *edac_mce)
+{
+	mutex_lock(&edac_mce_lock);
+	list_del(&edac_mce->list);
+	mutex_unlock(&edac_mce_lock);
+}
+EXPORT_SYMBOL(edac_mce_unregister);
+
+int edac_mce_parse(struct mce *mce)
+{
+	struct edac_mce *edac_mce;
+
+	list_for_each_entry(edac_mce, &edac_mce_list, list) {
+		if (edac_mce->check_error(edac_mce->priv, mce))
+			return 1;
+	}
+
+	/* Nobody queued the error */
+	return 0;
+}
+EXPORT_SYMBOL_GPL(edac_mce_parse);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Mauro Carvalho Chehab <mchehab@redhat.com>");
+MODULE_AUTHOR("Red Hat Inc. (http://www.redhat.com)");
+MODULE_DESCRIPTION("EDAC Driver for mcelog captured errors");
diff --git a/drivers/edac/i5000_edac.c b/drivers/edac/i5000_edac.c
index adc10a2ac5f6..996c1bdb5a34 100644
--- a/drivers/edac/i5000_edac.c
+++ b/drivers/edac/i5000_edac.c
@@ -774,7 +774,7 @@ static void i5000_clear_error(struct mem_ctl_info *mci)
 static void i5000_check_error(struct mem_ctl_info *mci)
 {
 	struct i5000_error_info info;
-	debugf4("MC%d: " __FILE__ ": %s()\n", mci->mc_idx, __func__);
+	debugf4("MC%d: %s: %s()\n", mci->mc_idx, __FILE__, __func__);
 	i5000_get_error_info(mci, &info);
 	i5000_process_error_info(mci, &info, 1);
 }
@@ -1353,8 +1353,8 @@ static int i5000_probe1(struct pci_dev *pdev, int dev_idx)
 	int num_dimms_per_channel;
 	int num_csrows;
 
-	debugf0("MC: " __FILE__ ": %s(), pdev bus %u dev=0x%x fn=0x%x\n",
-		__func__,
+	debugf0("MC: %s: %s(), pdev bus %u dev=0x%x fn=0x%x\n",
+		__FILE__, __func__,
 		pdev->bus->number,
 		PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
 
@@ -1389,7 +1389,7 @@ static int i5000_probe1(struct pci_dev *pdev, int dev_idx)
 		return -ENOMEM;
 
 	kobject_get(&mci->edac_mci_kobj);
-	debugf0("MC: " __FILE__ ": %s(): mci = %p\n", __func__, mci);
+	debugf0("MC: %s: %s(): mci = %p\n", __FILE__, __func__, mci);
 
 	mci->dev = &pdev->dev;	/* record ptr  to the generic device */
 
@@ -1432,8 +1432,8 @@ static int i5000_probe1(struct pci_dev *pdev, int dev_idx)
 
 	/* add this new MC control structure to EDAC's list of MCs */
 	if (edac_mc_add_mc(mci)) {
-		debugf0("MC: " __FILE__
-			": %s(): failed edac_mc_add_mc()\n", __func__);
+		debugf0("MC: %s: %s(): failed edac_mc_add_mc()\n",
+			__FILE__, __func__);
 		/* FIXME: perhaps some code should go here that disables error
 		 * reporting if we just enabled it
 		 */
@@ -1478,7 +1478,7 @@ static int __devinit i5000_init_one(struct pci_dev *pdev,
 {
 	int rc;
 
-	debugf0("MC: " __FILE__ ": %s()\n", __func__);
+	debugf0("MC: %s: %s()\n", __FILE__, __func__);
 
 	/* wake up device */
 	rc = pci_enable_device(pdev);
@@ -1497,7 +1497,7 @@ static void __devexit i5000_remove_one(struct pci_dev *pdev)
 {
 	struct mem_ctl_info *mci;
 
-	debugf0(__FILE__ ": %s()\n", __func__);
+	debugf0("%s: %s()\n", __FILE__, __func__);
 
 	if (i5000_pci)
 		edac_pci_release_generic_ctl(i5000_pci);
@@ -1544,7 +1544,7 @@ static int __init i5000_init(void)
 {
 	int pci_rc;
 
-	debugf2("MC: " __FILE__ ": %s()\n", __func__);
+	debugf2("MC: %s: %s()\n", __FILE__, __func__);
 
        /* Ensure that the OPSTATE is set correctly for POLL or NMI */
        opstate_init();
@@ -1560,7 +1560,7 @@ static int __init i5000_init(void)
  */
 static void __exit i5000_exit(void)
 {
-	debugf2("MC: " __FILE__ ": %s()\n", __func__);
+	debugf2("MC: %s: %s()\n", __FILE__, __func__);
 	pci_unregister_driver(&i5000_driver);
 }
 
diff --git a/drivers/edac/i5400_edac.c b/drivers/edac/i5400_edac.c
index f99d10655ed4..010c1d6526f5 100644
--- a/drivers/edac/i5400_edac.c
+++ b/drivers/edac/i5400_edac.c
@@ -694,7 +694,7 @@ static void i5400_clear_error(struct mem_ctl_info *mci)
 static void i5400_check_error(struct mem_ctl_info *mci)
 {
 	struct i5400_error_info info;
-	debugf4("MC%d: " __FILE__ ": %s()\n", mci->mc_idx, __func__);
+	debugf4("MC%d: %s: %s()\n", mci->mc_idx, __FILE__, __func__);
 	i5400_get_error_info(mci, &info);
 	i5400_process_error_info(mci, &info);
 }
@@ -1227,8 +1227,8 @@ static int i5400_probe1(struct pci_dev *pdev, int dev_idx)
 	if (dev_idx >= ARRAY_SIZE(i5400_devs))
 		return -EINVAL;
 
-	debugf0("MC: " __FILE__ ": %s(), pdev bus %u dev=0x%x fn=0x%x\n",
-		__func__,
+	debugf0("MC: %s: %s(), pdev bus %u dev=0x%x fn=0x%x\n",
+		__FILE__, __func__,
 		pdev->bus->number,
 		PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
 
@@ -1256,7 +1256,7 @@ static int i5400_probe1(struct pci_dev *pdev, int dev_idx)
 	if (mci == NULL)
 		return -ENOMEM;
 
-	debugf0("MC: " __FILE__ ": %s(): mci = %p\n", __func__, mci);
+	debugf0("MC: %s: %s(): mci = %p\n", __FILE__, __func__, mci);
 
 	mci->dev = &pdev->dev;	/* record ptr  to the generic device */
 
@@ -1299,8 +1299,8 @@ static int i5400_probe1(struct pci_dev *pdev, int dev_idx)
 
 	/* add this new MC control structure to EDAC's list of MCs */
 	if (edac_mc_add_mc(mci)) {
-		debugf0("MC: " __FILE__
-			": %s(): failed edac_mc_add_mc()\n", __func__);
+		debugf0("MC: %s: %s(): failed edac_mc_add_mc()\n",
+			__FILE__, __func__);
 		/* FIXME: perhaps some code should go here that disables error
 		 * reporting if we just enabled it
 		 */
@@ -1344,7 +1344,7 @@ static int __devinit i5400_init_one(struct pci_dev *pdev,
 {
 	int rc;
 
-	debugf0("MC: " __FILE__ ": %s()\n", __func__);
+	debugf0("MC: %s: %s()\n", __FILE__, __func__);
 
 	/* wake up device */
 	rc = pci_enable_device(pdev);
@@ -1363,7 +1363,7 @@ static void __devexit i5400_remove_one(struct pci_dev *pdev)
 {
 	struct mem_ctl_info *mci;
 
-	debugf0(__FILE__ ": %s()\n", __func__);
+	debugf0("%s: %s()\n", __FILE__, __func__);
 
 	if (i5400_pci)
 		edac_pci_release_generic_ctl(i5400_pci);
@@ -1409,7 +1409,7 @@ static int __init i5400_init(void)
 {
 	int pci_rc;
 
-	debugf2("MC: " __FILE__ ": %s()\n", __func__);
+	debugf2("MC: %s: %s()\n", __FILE__, __func__);
 
 	/* Ensure that the OPSTATE is set correctly for POLL or NMI */
 	opstate_init();
@@ -1425,7 +1425,7 @@ static int __init i5400_init(void)
  */
 static void __exit i5400_exit(void)
 {
-	debugf2("MC: " __FILE__ ": %s()\n", __func__);
+	debugf2("MC: %s: %s()\n", __FILE__, __func__);
 	pci_unregister_driver(&i5400_driver);
 }
 
diff --git a/drivers/edac/i7core_edac.c b/drivers/edac/i7core_edac.c
new file mode 100644
index 000000000000..6b8b7b41ec5f
--- /dev/null
+++ b/drivers/edac/i7core_edac.c
@@ -0,0 +1,2078 @@
+/* Intel i7 core/Nehalem Memory Controller kernel module
+ *
+ * This driver supports yhe memory controllers found on the Intel
+ * processor families i7core, i7core 7xx/8xx, i5core, Xeon 35xx,
+ * Xeon 55xx and Xeon 56xx also known as Nehalem, Nehalem-EP, Lynnfield
+ * and Westmere-EP.
+ *
+ * This file may be distributed under the terms of the
+ * GNU General Public License version 2 only.
+ *
+ * Copyright (c) 2009-2010 by:
+ *	 Mauro Carvalho Chehab <mchehab@redhat.com>
+ *
+ * Red Hat Inc. http://www.redhat.com
+ *
+ * Forked and adapted from the i5400_edac driver
+ *
+ * Based on the following public Intel datasheets:
+ * Intel Core i7 Processor Extreme Edition and Intel Core i7 Processor
+ * Datasheet, Volume 2:
+ *	http://download.intel.com/design/processor/datashts/320835.pdf
+ * Intel Xeon Processor 5500 Series Datasheet Volume 2
+ *	http://www.intel.com/Assets/PDF/datasheet/321322.pdf
+ * also available at:
+ * 	http://www.arrownac.com/manufacturers/intel/s/nehalem/5500-datasheet-v2.pdf
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/slab.h>
+#include <linux/delay.h>
+#include <linux/edac.h>
+#include <linux/mmzone.h>
+#include <linux/edac_mce.h>
+#include <linux/smp.h>
+#include <asm/processor.h>
+
+#include "edac_core.h"
+
+/*
+ * This is used for Nehalem-EP and Nehalem-EX devices, where the non-core
+ * registers start at bus 255, and are not reported by BIOS.
+ * We currently find devices with only 2 sockets. In order to support more QPI
+ * Quick Path Interconnect, just increment this number.
+ */
+#define MAX_SOCKET_BUSES	2
+
+
+/*
+ * Alter this version for the module when modifications are made
+ */
+#define I7CORE_REVISION    " Ver: 1.0.0 " __DATE__
+#define EDAC_MOD_STR      "i7core_edac"
+
+/*
+ * Debug macros
+ */
+#define i7core_printk(level, fmt, arg...)			\
+	edac_printk(level, "i7core", fmt, ##arg)
+
+#define i7core_mc_printk(mci, level, fmt, arg...)		\
+	edac_mc_chipset_printk(mci, level, "i7core", fmt, ##arg)
+
+/*
+ * i7core Memory Controller Registers
+ */
+
+	/* OFFSETS for Device 0 Function 0 */
+
+#define MC_CFG_CONTROL	0x90
+
+	/* OFFSETS for Device 3 Function 0 */
+
+#define MC_CONTROL	0x48
+#define MC_STATUS	0x4c
+#define MC_MAX_DOD	0x64
+
+/*
+ * OFFSETS for Device 3 Function 4, as inicated on Xeon 5500 datasheet:
+ * http://www.arrownac.com/manufacturers/intel/s/nehalem/5500-datasheet-v2.pdf
+ */
+
+#define MC_TEST_ERR_RCV1	0x60
+  #define DIMM2_COR_ERR(r)			((r) & 0x7fff)
+
+#define MC_TEST_ERR_RCV0	0x64
+  #define DIMM1_COR_ERR(r)			(((r) >> 16) & 0x7fff)
+  #define DIMM0_COR_ERR(r)			((r) & 0x7fff)
+
+/* OFFSETS for Device 3 Function 2, as inicated on Xeon 5500 datasheet */
+#define MC_COR_ECC_CNT_0	0x80
+#define MC_COR_ECC_CNT_1	0x84
+#define MC_COR_ECC_CNT_2	0x88
+#define MC_COR_ECC_CNT_3	0x8c
+#define MC_COR_ECC_CNT_4	0x90
+#define MC_COR_ECC_CNT_5	0x94
+
+#define DIMM_TOP_COR_ERR(r)			(((r) >> 16) & 0x7fff)
+#define DIMM_BOT_COR_ERR(r)			((r) & 0x7fff)
+
+
+	/* OFFSETS for Devices 4,5 and 6 Function 0 */
+
+#define MC_CHANNEL_DIMM_INIT_PARAMS 0x58
+  #define THREE_DIMMS_PRESENT		(1 << 24)
+  #define SINGLE_QUAD_RANK_PRESENT	(1 << 23)
+  #define QUAD_RANK_PRESENT		(1 << 22)
+  #define REGISTERED_DIMM		(1 << 15)
+
+#define MC_CHANNEL_MAPPER	0x60
+  #define RDLCH(r, ch)		((((r) >> (3 + (ch * 6))) & 0x07) - 1)
+  #define WRLCH(r, ch)		((((r) >> (ch * 6)) & 0x07) - 1)
+
+#define MC_CHANNEL_RANK_PRESENT 0x7c
+  #define RANK_PRESENT_MASK		0xffff
+
+#define MC_CHANNEL_ADDR_MATCH	0xf0
+#define MC_CHANNEL_ERROR_MASK	0xf8
+#define MC_CHANNEL_ERROR_INJECT	0xfc
+  #define INJECT_ADDR_PARITY	0x10
+  #define INJECT_ECC		0x08
+  #define MASK_CACHELINE	0x06
+  #define MASK_FULL_CACHELINE	0x06
+  #define MASK_MSB32_CACHELINE	0x04
+  #define MASK_LSB32_CACHELINE	0x02
+  #define NO_MASK_CACHELINE	0x00
+  #define REPEAT_EN		0x01
+
+	/* OFFSETS for Devices 4,5 and 6 Function 1 */
+
+#define MC_DOD_CH_DIMM0		0x48
+#define MC_DOD_CH_DIMM1		0x4c
+#define MC_DOD_CH_DIMM2		0x50
+  #define RANKOFFSET_MASK	((1 << 12) | (1 << 11) | (1 << 10))
+  #define RANKOFFSET(x)		((x & RANKOFFSET_MASK) >> 10)
+  #define DIMM_PRESENT_MASK	(1 << 9)
+  #define DIMM_PRESENT(x)	(((x) & DIMM_PRESENT_MASK) >> 9)
+  #define MC_DOD_NUMBANK_MASK		((1 << 8) | (1 << 7))
+  #define MC_DOD_NUMBANK(x)		(((x) & MC_DOD_NUMBANK_MASK) >> 7)
+  #define MC_DOD_NUMRANK_MASK		((1 << 6) | (1 << 5))
+  #define MC_DOD_NUMRANK(x)		(((x) & MC_DOD_NUMRANK_MASK) >> 5)
+  #define MC_DOD_NUMROW_MASK		((1 << 4) | (1 << 3) | (1 << 2))
+  #define MC_DOD_NUMROW(x)		(((x) & MC_DOD_NUMROW_MASK) >> 2)
+  #define MC_DOD_NUMCOL_MASK		3
+  #define MC_DOD_NUMCOL(x)		((x) & MC_DOD_NUMCOL_MASK)
+
+#define MC_RANK_PRESENT		0x7c
+
+#define MC_SAG_CH_0	0x80
+#define MC_SAG_CH_1	0x84
+#define MC_SAG_CH_2	0x88
+#define MC_SAG_CH_3	0x8c
+#define MC_SAG_CH_4	0x90
+#define MC_SAG_CH_5	0x94
+#define MC_SAG_CH_6	0x98
+#define MC_SAG_CH_7	0x9c
+
+#define MC_RIR_LIMIT_CH_0	0x40
+#define MC_RIR_LIMIT_CH_1	0x44
+#define MC_RIR_LIMIT_CH_2	0x48
+#define MC_RIR_LIMIT_CH_3	0x4C
+#define MC_RIR_LIMIT_CH_4	0x50
+#define MC_RIR_LIMIT_CH_5	0x54
+#define MC_RIR_LIMIT_CH_6	0x58
+#define MC_RIR_LIMIT_CH_7	0x5C
+#define MC_RIR_LIMIT_MASK	((1 << 10) - 1)
+
+#define MC_RIR_WAY_CH		0x80
+  #define MC_RIR_WAY_OFFSET_MASK	(((1 << 14) - 1) & ~0x7)
+  #define MC_RIR_WAY_RANK_MASK		0x7
+
+/*
+ * i7core structs
+ */
+
+#define NUM_CHANS 3
+#define MAX_DIMMS 3		/* Max DIMMS per channel */
+#define MAX_MCR_FUNC  4
+#define MAX_CHAN_FUNC 3
+
+struct i7core_info {
+	u32	mc_control;
+	u32	mc_status;
+	u32	max_dod;
+	u32	ch_map;
+};
+
+
+struct i7core_inject {
+	int	enable;
+
+	u32	section;
+	u32	type;
+	u32	eccmask;
+
+	/* Error address mask */
+	int channel, dimm, rank, bank, page, col;
+};
+
+struct i7core_channel {
+	u32		ranks;
+	u32		dimms;
+};
+
+struct pci_id_descr {
+	int			dev;
+	int			func;
+	int 			dev_id;
+	int			optional;
+};
+
+struct pci_id_table {
+	struct pci_id_descr	*descr;
+	int			n_devs;
+};
+
+struct i7core_dev {
+	struct list_head	list;
+	u8			socket;
+	struct pci_dev		**pdev;
+	int			n_devs;
+	struct mem_ctl_info	*mci;
+};
+
+struct i7core_pvt {
+	struct pci_dev	*pci_noncore;
+	struct pci_dev	*pci_mcr[MAX_MCR_FUNC + 1];
+	struct pci_dev	*pci_ch[NUM_CHANS][MAX_CHAN_FUNC + 1];
+
+	struct i7core_dev *i7core_dev;
+
+	struct i7core_info	info;
+	struct i7core_inject	inject;
+	struct i7core_channel	channel[NUM_CHANS];
+
+	int		channels; /* Number of active channels */
+
+	int		ce_count_available;
+	int 		csrow_map[NUM_CHANS][MAX_DIMMS];
+
+			/* ECC corrected errors counts per udimm */
+	unsigned long	udimm_ce_count[MAX_DIMMS];
+	int		udimm_last_ce_count[MAX_DIMMS];
+			/* ECC corrected errors counts per rdimm */
+	unsigned long	rdimm_ce_count[NUM_CHANS][MAX_DIMMS];
+	int		rdimm_last_ce_count[NUM_CHANS][MAX_DIMMS];
+
+	unsigned int	is_registered;
+
+	/* mcelog glue */
+	struct edac_mce		edac_mce;
+
+	/* Fifo double buffers */
+	struct mce		mce_entry[MCE_LOG_LEN];
+	struct mce		mce_outentry[MCE_LOG_LEN];
+
+	/* Fifo in/out counters */
+	unsigned		mce_in, mce_out;
+
+	/* Count indicator to show errors not got */
+	unsigned		mce_overrun;
+};
+
+/* Static vars */
+static LIST_HEAD(i7core_edac_list);
+static DEFINE_MUTEX(i7core_edac_lock);
+
+#define PCI_DESCR(device, function, device_id)	\
+	.dev = (device),			\
+	.func = (function),			\
+	.dev_id = (device_id)
+
+struct pci_id_descr pci_dev_descr_i7core_nehalem[] = {
+		/* Memory controller */
+	{ PCI_DESCR(3, 0, PCI_DEVICE_ID_INTEL_I7_MCR)     },
+	{ PCI_DESCR(3, 1, PCI_DEVICE_ID_INTEL_I7_MC_TAD)  },
+			/* Exists only for RDIMM */
+	{ PCI_DESCR(3, 2, PCI_DEVICE_ID_INTEL_I7_MC_RAS), .optional = 1  },
+	{ PCI_DESCR(3, 4, PCI_DEVICE_ID_INTEL_I7_MC_TEST) },
+
+		/* Channel 0 */
+	{ PCI_DESCR(4, 0, PCI_DEVICE_ID_INTEL_I7_MC_CH0_CTRL) },
+	{ PCI_DESCR(4, 1, PCI_DEVICE_ID_INTEL_I7_MC_CH0_ADDR) },
+	{ PCI_DESCR(4, 2, PCI_DEVICE_ID_INTEL_I7_MC_CH0_RANK) },
+	{ PCI_DESCR(4, 3, PCI_DEVICE_ID_INTEL_I7_MC_CH0_TC)   },
+
+		/* Channel 1 */
+	{ PCI_DESCR(5, 0, PCI_DEVICE_ID_INTEL_I7_MC_CH1_CTRL) },
+	{ PCI_DESCR(5, 1, PCI_DEVICE_ID_INTEL_I7_MC_CH1_ADDR) },
+	{ PCI_DESCR(5, 2, PCI_DEVICE_ID_INTEL_I7_MC_CH1_RANK) },
+	{ PCI_DESCR(5, 3, PCI_DEVICE_ID_INTEL_I7_MC_CH1_TC)   },
+
+		/* Channel 2 */
+	{ PCI_DESCR(6, 0, PCI_DEVICE_ID_INTEL_I7_MC_CH2_CTRL) },
+	{ PCI_DESCR(6, 1, PCI_DEVICE_ID_INTEL_I7_MC_CH2_ADDR) },
+	{ PCI_DESCR(6, 2, PCI_DEVICE_ID_INTEL_I7_MC_CH2_RANK) },
+	{ PCI_DESCR(6, 3, PCI_DEVICE_ID_INTEL_I7_MC_CH2_TC)   },
+
+		/* Generic Non-core registers */
+	/*
+	 * This is the PCI device on i7core and on Xeon 35xx (8086:2c41)
+	 * On Xeon 55xx, however, it has a different id (8086:2c40). So,
+	 * the probing code needs to test for the other address in case of
+	 * failure of this one
+	 */
+	{ PCI_DESCR(0, 0, PCI_DEVICE_ID_INTEL_I7_NONCORE)  },
+
+};
+
+struct pci_id_descr pci_dev_descr_lynnfield[] = {
+	{ PCI_DESCR( 3, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MCR)         },
+	{ PCI_DESCR( 3, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_TAD)      },
+	{ PCI_DESCR( 3, 4, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_TEST)     },
+
+	{ PCI_DESCR( 4, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_CTRL) },
+	{ PCI_DESCR( 4, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_ADDR) },
+	{ PCI_DESCR( 4, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_RANK) },
+	{ PCI_DESCR( 4, 3, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_TC)   },
+
+	{ PCI_DESCR( 5, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_CTRL) },
+	{ PCI_DESCR( 5, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_ADDR) },
+	{ PCI_DESCR( 5, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_RANK) },
+	{ PCI_DESCR( 5, 3, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_TC)   },
+
+	/*
+	 * This is the PCI device has an alternate address on some
+	 * processors like Core i7 860
+	 */
+	{ PCI_DESCR( 0, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE)     },
+};
+
+struct pci_id_descr pci_dev_descr_i7core_westmere[] = {
+		/* Memory controller */
+	{ PCI_DESCR(3, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MCR_REV2)     },
+	{ PCI_DESCR(3, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_TAD_REV2)  },
+			/* Exists only for RDIMM */
+	{ PCI_DESCR(3, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_RAS_REV2), .optional = 1  },
+	{ PCI_DESCR(3, 4, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_TEST_REV2) },
+
+		/* Channel 0 */
+	{ PCI_DESCR(4, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_CTRL_REV2) },
+	{ PCI_DESCR(4, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_ADDR_REV2) },
+	{ PCI_DESCR(4, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_RANK_REV2) },
+	{ PCI_DESCR(4, 3, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH0_TC_REV2)   },
+
+		/* Channel 1 */
+	{ PCI_DESCR(5, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_CTRL_REV2) },
+	{ PCI_DESCR(5, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_ADDR_REV2) },
+	{ PCI_DESCR(5, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_RANK_REV2) },
+	{ PCI_DESCR(5, 3, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH1_TC_REV2)   },
+
+		/* Channel 2 */
+	{ PCI_DESCR(6, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH2_CTRL_REV2) },
+	{ PCI_DESCR(6, 1, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH2_ADDR_REV2) },
+	{ PCI_DESCR(6, 2, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH2_RANK_REV2) },
+	{ PCI_DESCR(6, 3, PCI_DEVICE_ID_INTEL_LYNNFIELD_MC_CH2_TC_REV2)   },
+
+		/* Generic Non-core registers */
+	{ PCI_DESCR(0, 0, PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE_REV2)  },
+
+};
+
+#define PCI_ID_TABLE_ENTRY(A) { A, ARRAY_SIZE(A) }
+struct pci_id_table pci_dev_table[] = {
+	PCI_ID_TABLE_ENTRY(pci_dev_descr_i7core_nehalem),
+	PCI_ID_TABLE_ENTRY(pci_dev_descr_lynnfield),
+	PCI_ID_TABLE_ENTRY(pci_dev_descr_i7core_westmere),
+};
+
+/*
+ *	pci_device_id	table for which devices we are looking for
+ */
+static const struct pci_device_id i7core_pci_tbl[] __devinitdata = {
+	{PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_X58_HUB_MGMT)},
+	{PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_LYNNFIELD_QPI_LINK0)},
+	{0,}			/* 0 terminated list. */
+};
+
+static struct edac_pci_ctl_info *i7core_pci;
+
+/****************************************************************************
+			Anciliary status routines
+ ****************************************************************************/
+
+	/* MC_CONTROL bits */
+#define CH_ACTIVE(pvt, ch)	((pvt)->info.mc_control & (1 << (8 + ch)))
+#define ECCx8(pvt)		((pvt)->info.mc_control & (1 << 1))
+
+	/* MC_STATUS bits */
+#define ECC_ENABLED(pvt)	((pvt)->info.mc_status & (1 << 4))
+#define CH_DISABLED(pvt, ch)	((pvt)->info.mc_status & (1 << ch))
+
+	/* MC_MAX_DOD read functions */
+static inline int numdimms(u32 dimms)
+{
+	return (dimms & 0x3) + 1;
+}
+
+static inline int numrank(u32 rank)
+{
+	static int ranks[4] = { 1, 2, 4, -EINVAL };
+
+	return ranks[rank & 0x3];
+}
+
+static inline int numbank(u32 bank)
+{
+	static int banks[4] = { 4, 8, 16, -EINVAL };
+
+	return banks[bank & 0x3];
+}
+
+static inline int numrow(u32 row)
+{
+	static int rows[8] = {
+		1 << 12, 1 << 13, 1 << 14, 1 << 15,
+		1 << 16, -EINVAL, -EINVAL, -EINVAL,
+	};
+
+	return rows[row & 0x7];
+}
+
+static inline int numcol(u32 col)
+{
+	static int cols[8] = {
+		1 << 10, 1 << 11, 1 << 12, -EINVAL,
+	};
+	return cols[col & 0x3];
+}
+
+static struct i7core_dev *get_i7core_dev(u8 socket)
+{
+	struct i7core_dev *i7core_dev;
+
+	list_for_each_entry(i7core_dev, &i7core_edac_list, list) {
+		if (i7core_dev->socket == socket)
+			return i7core_dev;
+	}
+
+	return NULL;
+}
+
+/****************************************************************************
+			Memory check routines
+ ****************************************************************************/
+static struct pci_dev *get_pdev_slot_func(u8 socket, unsigned slot,
+					  unsigned func)
+{
+	struct i7core_dev *i7core_dev = get_i7core_dev(socket);
+	int i;
+
+	if (!i7core_dev)
+		return NULL;
+
+	for (i = 0; i < i7core_dev->n_devs; i++) {
+		if (!i7core_dev->pdev[i])
+			continue;
+
+		if (PCI_SLOT(i7core_dev->pdev[i]->devfn) == slot &&
+		    PCI_FUNC(i7core_dev->pdev[i]->devfn) == func) {
+			return i7core_dev->pdev[i];
+		}
+	}
+
+	return NULL;
+}
+
+/**
+ * i7core_get_active_channels() - gets the number of channels and csrows
+ * @socket:	Quick Path Interconnect socket
+ * @channels:	Number of channels that will be returned
+ * @csrows:	Number of csrows found
+ *
+ * Since EDAC core needs to know in advance the number of available channels
+ * and csrows, in order to allocate memory for csrows/channels, it is needed
+ * to run two similar steps. At the first step, implemented on this function,
+ * it checks the number of csrows/channels present at one socket.
+ * this is used in order to properly allocate the size of mci components.
+ *
+ * It should be noticed that none of the current available datasheets explain
+ * or even mention how csrows are seen by the memory controller. So, we need
+ * to add a fake description for csrows.
+ * So, this driver is attributing one DIMM memory for one csrow.
+ */
+static int i7core_get_active_channels(u8 socket, unsigned *channels,
+				      unsigned *csrows)
+{
+	struct pci_dev *pdev = NULL;
+	int i, j;
+	u32 status, control;
+
+	*channels = 0;
+	*csrows = 0;
+
+	pdev = get_pdev_slot_func(socket, 3, 0);
+	if (!pdev) {
+		i7core_printk(KERN_ERR, "Couldn't find socket %d fn 3.0!!!\n",
+			      socket);
+		return -ENODEV;
+	}
+
+	/* Device 3 function 0 reads */
+	pci_read_config_dword(pdev, MC_STATUS, &status);
+	pci_read_config_dword(pdev, MC_CONTROL, &control);
+
+	for (i = 0; i < NUM_CHANS; i++) {
+		u32 dimm_dod[3];
+		/* Check if the channel is active */
+		if (!(control & (1 << (8 + i))))
+			continue;
+
+		/* Check if the channel is disabled */
+		if (status & (1 << i))
+			continue;
+
+		pdev = get_pdev_slot_func(socket, i + 4, 1);
+		if (!pdev) {
+			i7core_printk(KERN_ERR, "Couldn't find socket %d "
+						"fn %d.%d!!!\n",
+						socket, i + 4, 1);
+			return -ENODEV;
+		}
+		/* Devices 4-6 function 1 */
+		pci_read_config_dword(pdev,
+				MC_DOD_CH_DIMM0, &dimm_dod[0]);
+		pci_read_config_dword(pdev,
+				MC_DOD_CH_DIMM1, &dimm_dod[1]);
+		pci_read_config_dword(pdev,
+				MC_DOD_CH_DIMM2, &dimm_dod[2]);
+
+		(*channels)++;
+
+		for (j = 0; j < 3; j++) {
+			if (!DIMM_PRESENT(dimm_dod[j]))
+				continue;
+			(*csrows)++;
+		}
+	}
+
+	debugf0("Number of active channels on socket %d: %d\n",
+		socket, *channels);
+
+	return 0;
+}
+
+static int get_dimm_config(struct mem_ctl_info *mci, int *csrow)
+{
+	struct i7core_pvt *pvt = mci->pvt_info;
+	struct csrow_info *csr;
+	struct pci_dev *pdev;
+	int i, j;
+	unsigned long last_page = 0;
+	enum edac_type mode;
+	enum mem_type mtype;
+
+	/* Get data from the MC register, function 0 */
+	pdev = pvt->pci_mcr[0];
+	if (!pdev)
+		return -ENODEV;
+
+	/* Device 3 function 0 reads */
+	pci_read_config_dword(pdev, MC_CONTROL, &pvt->info.mc_control);
+	pci_read_config_dword(pdev, MC_STATUS, &pvt->info.mc_status);
+	pci_read_config_dword(pdev, MC_MAX_DOD, &pvt->info.max_dod);
+	pci_read_config_dword(pdev, MC_CHANNEL_MAPPER, &pvt->info.ch_map);
+
+	debugf0("QPI %d control=0x%08x status=0x%08x dod=0x%08x map=0x%08x\n",
+		pvt->i7core_dev->socket, pvt->info.mc_control, pvt->info.mc_status,
+		pvt->info.max_dod, pvt->info.ch_map);
+
+	if (ECC_ENABLED(pvt)) {
+		debugf0("ECC enabled with x%d SDCC\n", ECCx8(pvt) ? 8 : 4);
+		if (ECCx8(pvt))
+			mode = EDAC_S8ECD8ED;
+		else
+			mode = EDAC_S4ECD4ED;
+	} else {
+		debugf0("ECC disabled\n");
+		mode = EDAC_NONE;
+	}
+
+	/* FIXME: need to handle the error codes */
+	debugf0("DOD Max limits: DIMMS: %d, %d-ranked, %d-banked "
+		"x%x x 0x%x\n",
+		numdimms(pvt->info.max_dod),
+		numrank(pvt->info.max_dod >> 2),
+		numbank(pvt->info.max_dod >> 4),
+		numrow(pvt->info.max_dod >> 6),
+		numcol(pvt->info.max_dod >> 9));
+
+	for (i = 0; i < NUM_CHANS; i++) {
+		u32 data, dimm_dod[3], value[8];
+
+		if (!pvt->pci_ch[i][0])
+			continue;
+
+		if (!CH_ACTIVE(pvt, i)) {
+			debugf0("Channel %i is not active\n", i);
+			continue;
+		}
+		if (CH_DISABLED(pvt, i)) {
+			debugf0("Channel %i is disabled\n", i);
+			continue;
+		}
+
+		/* Devices 4-6 function 0 */
+		pci_read_config_dword(pvt->pci_ch[i][0],
+				MC_CHANNEL_DIMM_INIT_PARAMS, &data);
+
+		pvt->channel[i].ranks = (data & QUAD_RANK_PRESENT) ?
+						4 : 2;
+
+		if (data & REGISTERED_DIMM)
+			mtype = MEM_RDDR3;
+		else
+			mtype = MEM_DDR3;
+#if 0
+		if (data & THREE_DIMMS_PRESENT)
+			pvt->channel[i].dimms = 3;
+		else if (data & SINGLE_QUAD_RANK_PRESENT)
+			pvt->channel[i].dimms = 1;
+		else
+			pvt->channel[i].dimms = 2;
+#endif
+
+		/* Devices 4-6 function 1 */
+		pci_read_config_dword(pvt->pci_ch[i][1],
+				MC_DOD_CH_DIMM0, &dimm_dod[0]);
+		pci_read_config_dword(pvt->pci_ch[i][1],
+				MC_DOD_CH_DIMM1, &dimm_dod[1]);
+		pci_read_config_dword(pvt->pci_ch[i][1],
+				MC_DOD_CH_DIMM2, &dimm_dod[2]);
+
+		debugf0("Ch%d phy rd%d, wr%d (0x%08x): "
+			"%d ranks, %cDIMMs\n",
+			i,
+			RDLCH(pvt->info.ch_map, i), WRLCH(pvt->info.ch_map, i),
+			data,
+			pvt->channel[i].ranks,
+			(data & REGISTERED_DIMM) ? 'R' : 'U');
+
+		for (j = 0; j < 3; j++) {
+			u32 banks, ranks, rows, cols;
+			u32 size, npages;
+
+			if (!DIMM_PRESENT(dimm_dod[j]))
+				continue;
+
+			banks = numbank(MC_DOD_NUMBANK(dimm_dod[j]));
+			ranks = numrank(MC_DOD_NUMRANK(dimm_dod[j]));
+			rows = numrow(MC_DOD_NUMROW(dimm_dod[j]));
+			cols = numcol(MC_DOD_NUMCOL(dimm_dod[j]));
+
+			/* DDR3 has 8 I/O banks */
+			size = (rows * cols * banks * ranks) >> (20 - 3);
+
+			pvt->channel[i].dimms++;
+
+			debugf0("\tdimm %d %d Mb offset: %x, "
+				"bank: %d, rank: %d, row: %#x, col: %#x\n",
+				j, size,
+				RANKOFFSET(dimm_dod[j]),
+				banks, ranks, rows, cols);
+
+#if PAGE_SHIFT > 20
+			npages = size >> (PAGE_SHIFT - 20);
+#else
+			npages = size << (20 - PAGE_SHIFT);
+#endif
+
+			csr = &mci->csrows[*csrow];
+			csr->first_page = last_page + 1;
+			last_page += npages;
+			csr->last_page = last_page;
+			csr->nr_pages = npages;
+
+			csr->page_mask = 0;
+			csr->grain = 8;
+			csr->csrow_idx = *csrow;
+			csr->nr_channels = 1;
+
+			csr->channels[0].chan_idx = i;
+			csr->channels[0].ce_count = 0;
+
+			pvt->csrow_map[i][j] = *csrow;
+
+			switch (banks) {
+			case 4:
+				csr->dtype = DEV_X4;
+				break;
+			case 8:
+				csr->dtype = DEV_X8;
+				break;
+			case 16:
+				csr->dtype = DEV_X16;
+				break;
+			default:
+				csr->dtype = DEV_UNKNOWN;
+			}
+
+			csr->edac_mode = mode;
+			csr->mtype = mtype;
+
+			(*csrow)++;
+		}
+
+		pci_read_config_dword(pdev, MC_SAG_CH_0, &value[0]);
+		pci_read_config_dword(pdev, MC_SAG_CH_1, &value[1]);
+		pci_read_config_dword(pdev, MC_SAG_CH_2, &value[2]);
+		pci_read_config_dword(pdev, MC_SAG_CH_3, &value[3]);
+		pci_read_config_dword(pdev, MC_SAG_CH_4, &value[4]);
+		pci_read_config_dword(pdev, MC_SAG_CH_5, &value[5]);
+		pci_read_config_dword(pdev, MC_SAG_CH_6, &value[6]);
+		pci_read_config_dword(pdev, MC_SAG_CH_7, &value[7]);
+		debugf1("\t[%i] DIVBY3\tREMOVED\tOFFSET\n", i);
+		for (j = 0; j < 8; j++)
+			debugf1("\t\t%#x\t%#x\t%#x\n",
+				(value[j] >> 27) & 0x1,
+				(value[j] >> 24) & 0x7,
+				(value[j] && ((1 << 24) - 1)));
+	}
+
+	return 0;
+}
+
+/****************************************************************************
+			Error insertion routines
+ ****************************************************************************/
+
+/* The i7core has independent error injection features per channel.
+   However, to have a simpler code, we don't allow enabling error injection
+   on more than one channel.
+   Also, since a change at an inject parameter will be applied only at enable,
+   we're disabling error injection on all write calls to the sysfs nodes that
+   controls the error code injection.
+ */
+static int disable_inject(struct mem_ctl_info *mci)
+{
+	struct i7core_pvt *pvt = mci->pvt_info;
+
+	pvt->inject.enable = 0;
+
+	if (!pvt->pci_ch[pvt->inject.channel][0])
+		return -ENODEV;
+
+	pci_write_config_dword(pvt->pci_ch[pvt->inject.channel][0],
+				MC_CHANNEL_ERROR_INJECT, 0);
+
+	return 0;
+}
+
+/*
+ * i7core inject inject.section
+ *
+ *	accept and store error injection inject.section value
+ *	bit 0 - refers to the lower 32-byte half cacheline
+ *	bit 1 - refers to the upper 32-byte half cacheline
+ */
+static ssize_t i7core_inject_section_store(struct mem_ctl_info *mci,
+					   const char *data, size_t count)
+{
+	struct i7core_pvt *pvt = mci->pvt_info;
+	unsigned long value;
+	int rc;
+
+	if (pvt->inject.enable)
+		disable_inject(mci);
+
+	rc = strict_strtoul(data, 10, &value);
+	if ((rc < 0) || (value > 3))
+		return -EIO;
+
+	pvt->inject.section = (u32) value;
+	return count;
+}
+
+static ssize_t i7core_inject_section_show(struct mem_ctl_info *mci,
+					      char *data)
+{
+	struct i7core_pvt *pvt = mci->pvt_info;
+	return sprintf(data, "0x%08x\n", pvt->inject.section);
+}
+
+/*
+ * i7core inject.type
+ *
+ *	accept and store error injection inject.section value
+ *	bit 0 - repeat enable - Enable error repetition
+ *	bit 1 - inject ECC error
+ *	bit 2 - inject parity error
+ */
+static ssize_t i7core_inject_type_store(struct mem_ctl_info *mci,
+					const char *data, size_t count)
+{
+	struct i7core_pvt *pvt = mci->pvt_info;
+	unsigned long value;
+	int rc;
+
+	if (pvt->inject.enable)
+		disable_inject(mci);
+
+	rc = strict_strtoul(data, 10, &value);
+	if ((rc < 0) || (value > 7))
+		return -EIO;
+
+	pvt->inject.type = (u32) value;
+	return count;
+}
+
+static ssize_t i7core_inject_type_show(struct mem_ctl_info *mci,
+					      char *data)
+{
+	struct i7core_pvt *pvt = mci->pvt_info;
+	return sprintf(data, "0x%08x\n", pvt->inject.type);
+}
+
+/*
+ * i7core_inject_inject.eccmask_store
+ *
+ * The type of error (UE/CE) will depend on the inject.eccmask value:
+ *   Any bits set to a 1 will flip the corresponding ECC bit
+ *   Correctable errors can be injected by flipping 1 bit or the bits within
+ *   a symbol pair (2 consecutive aligned 8-bit pairs - i.e. 7:0 and 15:8 or
+ *   23:16 and 31:24). Flipping bits in two symbol pairs will cause an
+ *   uncorrectable error to be injected.
+ */
+static ssize_t i7core_inject_eccmask_store(struct mem_ctl_info *mci,
+					const char *data, size_t count)
+{
+	struct i7core_pvt *pvt = mci->pvt_info;
+	unsigned long value;
+	int rc;
+
+	if (pvt->inject.enable)
+		disable_inject(mci);
+
+	rc = strict_strtoul(data, 10, &value);
+	if (rc < 0)
+		return -EIO;
+
+	pvt->inject.eccmask = (u32) value;
+	return count;
+}
+
+static ssize_t i7core_inject_eccmask_show(struct mem_ctl_info *mci,
+					      char *data)
+{
+	struct i7core_pvt *pvt = mci->pvt_info;
+	return sprintf(data, "0x%08x\n", pvt->inject.eccmask);
+}
+
+/*
+ * i7core_addrmatch
+ *
+ * The type of error (UE/CE) will depend on the inject.eccmask value:
+ *   Any bits set to a 1 will flip the corresponding ECC bit
+ *   Correctable errors can be injected by flipping 1 bit or the bits within
+ *   a symbol pair (2 consecutive aligned 8-bit pairs - i.e. 7:0 and 15:8 or
+ *   23:16 and 31:24). Flipping bits in two symbol pairs will cause an
+ *   uncorrectable error to be injected.
+ */
+
+#define DECLARE_ADDR_MATCH(param, limit)			\
+static ssize_t i7core_inject_store_##param(			\
+		struct mem_ctl_info *mci,			\
+		const char *data, size_t count)			\
+{								\
+	struct i7core_pvt *pvt;					\
+	long value;						\
+	int rc;							\
+								\
+	debugf1("%s()\n", __func__);				\
+	pvt = mci->pvt_info;					\
+								\
+	if (pvt->inject.enable)					\
+		disable_inject(mci);				\
+								\
+	if (!strcasecmp(data, "any") || !strcasecmp(data, "any\n"))\
+		value = -1;					\
+	else {							\
+		rc = strict_strtoul(data, 10, &value);		\
+		if ((rc < 0) || (value >= limit))		\
+			return -EIO;				\
+	}							\
+								\
+	pvt->inject.param = value;				\
+								\
+	return count;						\
+}								\
+								\
+static ssize_t i7core_inject_show_##param(			\
+		struct mem_ctl_info *mci,			\
+		char *data)					\
+{								\
+	struct i7core_pvt *pvt;					\
+								\
+	pvt = mci->pvt_info;					\
+	debugf1("%s() pvt=%p\n", __func__, pvt);		\
+	if (pvt->inject.param < 0)				\
+		return sprintf(data, "any\n");			\
+	else							\
+		return sprintf(data, "%d\n", pvt->inject.param);\
+}
+
+#define ATTR_ADDR_MATCH(param)					\
+	{							\
+		.attr = {					\
+			.name = #param,				\
+			.mode = (S_IRUGO | S_IWUSR)		\
+		},						\
+		.show  = i7core_inject_show_##param,		\
+		.store = i7core_inject_store_##param,		\
+	}
+
+DECLARE_ADDR_MATCH(channel, 3);
+DECLARE_ADDR_MATCH(dimm, 3);
+DECLARE_ADDR_MATCH(rank, 4);
+DECLARE_ADDR_MATCH(bank, 32);
+DECLARE_ADDR_MATCH(page, 0x10000);
+DECLARE_ADDR_MATCH(col, 0x4000);
+
+static int write_and_test(struct pci_dev *dev, int where, u32 val)
+{
+	u32 read;
+	int count;
+
+	debugf0("setting pci %02x:%02x.%x reg=%02x value=%08x\n",
+		dev->bus->number, PCI_SLOT(dev->devfn), PCI_FUNC(dev->devfn),
+		where, val);
+
+	for (count = 0; count < 10; count++) {
+		if (count)
+			msleep(100);
+		pci_write_config_dword(dev, where, val);
+		pci_read_config_dword(dev, where, &read);
+
+		if (read == val)
+			return 0;
+	}
+
+	i7core_printk(KERN_ERR, "Error during set pci %02x:%02x.%x reg=%02x "
+		"write=%08x. Read=%08x\n",
+		dev->bus->number, PCI_SLOT(dev->devfn), PCI_FUNC(dev->devfn),
+		where, val, read);
+
+	return -EINVAL;
+}
+
+/*
+ * This routine prepares the Memory Controller for error injection.
+ * The error will be injected when some process tries to write to the
+ * memory that matches the given criteria.
+ * The criteria can be set in terms of a mask where dimm, rank, bank, page
+ * and col can be specified.
+ * A -1 value for any of the mask items will make the MCU to ignore
+ * that matching criteria for error injection.
+ *
+ * It should be noticed that the error will only happen after a write operation
+ * on a memory that matches the condition. if REPEAT_EN is not enabled at
+ * inject mask, then it will produce just one error. Otherwise, it will repeat
+ * until the injectmask would be cleaned.
+ *
+ * FIXME: This routine assumes that MAXNUMDIMMS value of MC_MAX_DOD
+ *    is reliable enough to check if the MC is using the
+ *    three channels. However, this is not clear at the datasheet.
+ */
+static ssize_t i7core_inject_enable_store(struct mem_ctl_info *mci,
+				       const char *data, size_t count)
+{
+	struct i7core_pvt *pvt = mci->pvt_info;
+	u32 injectmask;
+	u64 mask = 0;
+	int  rc;
+	long enable;
+
+	if (!pvt->pci_ch[pvt->inject.channel][0])
+		return 0;
+
+	rc = strict_strtoul(data, 10, &enable);
+	if ((rc < 0))
+		return 0;
+
+	if (enable) {
+		pvt->inject.enable = 1;
+	} else {
+		disable_inject(mci);
+		return count;
+	}
+
+	/* Sets pvt->inject.dimm mask */
+	if (pvt->inject.dimm < 0)
+		mask |= 1LL << 41;
+	else {
+		if (pvt->channel[pvt->inject.channel].dimms > 2)
+			mask |= (pvt->inject.dimm & 0x3LL) << 35;
+		else
+			mask |= (pvt->inject.dimm & 0x1LL) << 36;
+	}
+
+	/* Sets pvt->inject.rank mask */
+	if (pvt->inject.rank < 0)
+		mask |= 1LL << 40;
+	else {
+		if (pvt->channel[pvt->inject.channel].dimms > 2)
+			mask |= (pvt->inject.rank & 0x1LL) << 34;
+		else
+			mask |= (pvt->inject.rank & 0x3LL) << 34;
+	}
+
+	/* Sets pvt->inject.bank mask */
+	if (pvt->inject.bank < 0)
+		mask |= 1LL << 39;
+	else
+		mask |= (pvt->inject.bank & 0x15LL) << 30;
+
+	/* Sets pvt->inject.page mask */
+	if (pvt->inject.page < 0)
+		mask |= 1LL << 38;
+	else
+		mask |= (pvt->inject.page & 0xffff) << 14;
+
+	/* Sets pvt->inject.column mask */
+	if (pvt->inject.col < 0)
+		mask |= 1LL << 37;
+	else
+		mask |= (pvt->inject.col & 0x3fff);
+
+	/*
+	 * bit    0: REPEAT_EN
+	 * bits 1-2: MASK_HALF_CACHELINE
+	 * bit    3: INJECT_ECC
+	 * bit    4: INJECT_ADDR_PARITY
+	 */
+
+	injectmask = (pvt->inject.type & 1) |
+		     (pvt->inject.section & 0x3) << 1 |
+		     (pvt->inject.type & 0x6) << (3 - 1);
+
+	/* Unlock writes to registers - this register is write only */
+	pci_write_config_dword(pvt->pci_noncore,
+			       MC_CFG_CONTROL, 0x2);
+
+	write_and_test(pvt->pci_ch[pvt->inject.channel][0],
+			       MC_CHANNEL_ADDR_MATCH, mask);
+	write_and_test(pvt->pci_ch[pvt->inject.channel][0],
+			       MC_CHANNEL_ADDR_MATCH + 4, mask >> 32L);
+
+	write_and_test(pvt->pci_ch[pvt->inject.channel][0],
+			       MC_CHANNEL_ERROR_MASK, pvt->inject.eccmask);
+
+	write_and_test(pvt->pci_ch[pvt->inject.channel][0],
+			       MC_CHANNEL_ERROR_INJECT, injectmask);
+
+	/*
+	 * This is something undocumented, based on my tests
+	 * Without writing 8 to this register, errors aren't injected. Not sure
+	 * why.
+	 */
+	pci_write_config_dword(pvt->pci_noncore,
+			       MC_CFG_CONTROL, 8);
+
+	debugf0("Error inject addr match 0x%016llx, ecc 0x%08x,"
+		" inject 0x%08x\n",
+		mask, pvt->inject.eccmask, injectmask);
+
+
+	return count;
+}
+
+static ssize_t i7core_inject_enable_show(struct mem_ctl_info *mci,
+					char *data)
+{
+	struct i7core_pvt *pvt = mci->pvt_info;
+	u32 injectmask;
+
+	if (!pvt->pci_ch[pvt->inject.channel][0])
+		return 0;
+
+	pci_read_config_dword(pvt->pci_ch[pvt->inject.channel][0],
+			       MC_CHANNEL_ERROR_INJECT, &injectmask);
+
+	debugf0("Inject error read: 0x%018x\n", injectmask);
+
+	if (injectmask & 0x0c)
+		pvt->inject.enable = 1;
+
+	return sprintf(data, "%d\n", pvt->inject.enable);
+}
+
+#define DECLARE_COUNTER(param)					\
+static ssize_t i7core_show_counter_##param(			\
+		struct mem_ctl_info *mci,			\
+		char *data)					\
+{								\
+	struct i7core_pvt *pvt = mci->pvt_info;			\
+								\
+	debugf1("%s() \n", __func__);				\
+	if (!pvt->ce_count_available || (pvt->is_registered))	\
+		return sprintf(data, "data unavailable\n");	\
+	return sprintf(data, "%lu\n",				\
+			pvt->udimm_ce_count[param]);		\
+}
+
+#define ATTR_COUNTER(param)					\
+	{							\
+		.attr = {					\
+			.name = __stringify(udimm##param),	\
+			.mode = (S_IRUGO | S_IWUSR)		\
+		},						\
+		.show  = i7core_show_counter_##param		\
+	}
+
+DECLARE_COUNTER(0);
+DECLARE_COUNTER(1);
+DECLARE_COUNTER(2);
+
+/*
+ * Sysfs struct
+ */
+
+
+static struct mcidev_sysfs_attribute i7core_addrmatch_attrs[] = {
+	ATTR_ADDR_MATCH(channel),
+	ATTR_ADDR_MATCH(dimm),
+	ATTR_ADDR_MATCH(rank),
+	ATTR_ADDR_MATCH(bank),
+	ATTR_ADDR_MATCH(page),
+	ATTR_ADDR_MATCH(col),
+	{ .attr = { .name = NULL } }
+};
+
+static struct mcidev_sysfs_group i7core_inject_addrmatch = {
+	.name  = "inject_addrmatch",
+	.mcidev_attr = i7core_addrmatch_attrs,
+};
+
+static struct mcidev_sysfs_attribute i7core_udimm_counters_attrs[] = {
+	ATTR_COUNTER(0),
+	ATTR_COUNTER(1),
+	ATTR_COUNTER(2),
+};
+
+static struct mcidev_sysfs_group i7core_udimm_counters = {
+	.name  = "all_channel_counts",
+	.mcidev_attr = i7core_udimm_counters_attrs,
+};
+
+static struct mcidev_sysfs_attribute i7core_sysfs_attrs[] = {
+	{
+		.attr = {
+			.name = "inject_section",
+			.mode = (S_IRUGO | S_IWUSR)
+		},
+		.show  = i7core_inject_section_show,
+		.store = i7core_inject_section_store,
+	}, {
+		.attr = {
+			.name = "inject_type",
+			.mode = (S_IRUGO | S_IWUSR)
+		},
+		.show  = i7core_inject_type_show,
+		.store = i7core_inject_type_store,
+	}, {
+		.attr = {
+			.name = "inject_eccmask",
+			.mode = (S_IRUGO | S_IWUSR)
+		},
+		.show  = i7core_inject_eccmask_show,
+		.store = i7core_inject_eccmask_store,
+	}, {
+		.grp = &i7core_inject_addrmatch,
+	}, {
+		.attr = {
+			.name = "inject_enable",
+			.mode = (S_IRUGO | S_IWUSR)
+		},
+		.show  = i7core_inject_enable_show,
+		.store = i7core_inject_enable_store,
+	},
+	{ .attr = { .name = NULL } },	/* Reserved for udimm counters */
+	{ .attr = { .name = NULL } }
+};
+
+/****************************************************************************
+	Device initialization routines: put/get, init/exit
+ ****************************************************************************/
+
+/*
+ *	i7core_put_devices	'put' all the devices that we have
+ *				reserved via 'get'
+ */
+static void i7core_put_devices(struct i7core_dev *i7core_dev)
+{
+	int i;
+
+	debugf0(__FILE__ ": %s()\n", __func__);
+	for (i = 0; i < i7core_dev->n_devs; i++) {
+		struct pci_dev *pdev = i7core_dev->pdev[i];
+		if (!pdev)
+			continue;
+		debugf0("Removing dev %02x:%02x.%d\n",
+			pdev->bus->number,
+			PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
+		pci_dev_put(pdev);
+	}
+	kfree(i7core_dev->pdev);
+	list_del(&i7core_dev->list);
+	kfree(i7core_dev);
+}
+
+static void i7core_put_all_devices(void)
+{
+	struct i7core_dev *i7core_dev, *tmp;
+
+	list_for_each_entry_safe(i7core_dev, tmp, &i7core_edac_list, list)
+		i7core_put_devices(i7core_dev);
+}
+
+static void __init i7core_xeon_pci_fixup(struct pci_id_table *table)
+{
+	struct pci_dev *pdev = NULL;
+	int i;
+	/*
+	 * On Xeon 55xx, the Intel Quckpath Arch Generic Non-core pci buses
+	 * aren't announced by acpi. So, we need to use a legacy scan probing
+	 * to detect them
+	 */
+	while (table && table->descr) {
+		pdev = pci_get_device(PCI_VENDOR_ID_INTEL, table->descr[0].dev_id, NULL);
+		if (unlikely(!pdev)) {
+			for (i = 0; i < MAX_SOCKET_BUSES; i++)
+				pcibios_scan_specific_bus(255-i);
+		}
+		table++;
+	}
+}
+
+/*
+ *	i7core_get_devices	Find and perform 'get' operation on the MCH's
+ *			device/functions we want to reference for this driver
+ *
+ *			Need to 'get' device 16 func 1 and func 2
+ */
+int i7core_get_onedevice(struct pci_dev **prev, int devno,
+			 struct pci_id_descr *dev_descr, unsigned n_devs)
+{
+	struct i7core_dev *i7core_dev;
+
+	struct pci_dev *pdev = NULL;
+	u8 bus = 0;
+	u8 socket = 0;
+
+	pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+			      dev_descr->dev_id, *prev);
+
+	/*
+	 * On Xeon 55xx, the Intel Quckpath Arch Generic Non-core regs
+	 * is at addr 8086:2c40, instead of 8086:2c41. So, we need
+	 * to probe for the alternate address in case of failure
+	 */
+	if (dev_descr->dev_id == PCI_DEVICE_ID_INTEL_I7_NONCORE && !pdev)
+		pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+				      PCI_DEVICE_ID_INTEL_I7_NONCORE_ALT, *prev);
+
+	if (dev_descr->dev_id == PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE && !pdev)
+		pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+				      PCI_DEVICE_ID_INTEL_LYNNFIELD_NONCORE_ALT,
+				      *prev);
+
+	if (!pdev) {
+		if (*prev) {
+			*prev = pdev;
+			return 0;
+		}
+
+		if (dev_descr->optional)
+			return 0;
+
+		if (devno == 0)
+			return -ENODEV;
+
+		i7core_printk(KERN_ERR,
+			"Device not found: dev %02x.%d PCI ID %04x:%04x\n",
+			dev_descr->dev, dev_descr->func,
+			PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
+
+		/* End of list, leave */
+		return -ENODEV;
+	}
+	bus = pdev->bus->number;
+
+	if (bus == 0x3f)
+		socket = 0;
+	else
+		socket = 255 - bus;
+
+	i7core_dev = get_i7core_dev(socket);
+	if (!i7core_dev) {
+		i7core_dev = kzalloc(sizeof(*i7core_dev), GFP_KERNEL);
+		if (!i7core_dev)
+			return -ENOMEM;
+		i7core_dev->pdev = kzalloc(sizeof(*i7core_dev->pdev) * n_devs,
+					   GFP_KERNEL);
+		if (!i7core_dev->pdev) {
+			kfree(i7core_dev);
+			return -ENOMEM;
+		}
+		i7core_dev->socket = socket;
+		i7core_dev->n_devs = n_devs;
+		list_add_tail(&i7core_dev->list, &i7core_edac_list);
+	}
+
+	if (i7core_dev->pdev[devno]) {
+		i7core_printk(KERN_ERR,
+			"Duplicated device for "
+			"dev %02x:%02x.%d PCI ID %04x:%04x\n",
+			bus, dev_descr->dev, dev_descr->func,
+			PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
+		pci_dev_put(pdev);
+		return -ENODEV;
+	}
+
+	i7core_dev->pdev[devno] = pdev;
+
+	/* Sanity check */
+	if (unlikely(PCI_SLOT(pdev->devfn) != dev_descr->dev ||
+			PCI_FUNC(pdev->devfn) != dev_descr->func)) {
+		i7core_printk(KERN_ERR,
+			"Device PCI ID %04x:%04x "
+			"has dev %02x:%02x.%d instead of dev %02x:%02x.%d\n",
+			PCI_VENDOR_ID_INTEL, dev_descr->dev_id,
+			bus, PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn),
+			bus, dev_descr->dev, dev_descr->func);
+		return -ENODEV;
+	}
+
+	/* Be sure that the device is enabled */
+	if (unlikely(pci_enable_device(pdev) < 0)) {
+		i7core_printk(KERN_ERR,
+			"Couldn't enable "
+			"dev %02x:%02x.%d PCI ID %04x:%04x\n",
+			bus, dev_descr->dev, dev_descr->func,
+			PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
+		return -ENODEV;
+	}
+
+	debugf0("Detected socket %d dev %02x:%02x.%d PCI ID %04x:%04x\n",
+		socket, bus, dev_descr->dev,
+		dev_descr->func,
+		PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
+
+	*prev = pdev;
+
+	return 0;
+}
+
+static int i7core_get_devices(struct pci_id_table *table)
+{
+	int i, rc;
+	struct pci_dev *pdev = NULL;
+	struct pci_id_descr *dev_descr;
+
+	while (table && table->descr) {
+		dev_descr = table->descr;
+		for (i = 0; i < table->n_devs; i++) {
+			pdev = NULL;
+			do {
+				rc = i7core_get_onedevice(&pdev, i, &dev_descr[i],
+							  table->n_devs);
+				if (rc < 0) {
+					if (i == 0) {
+						i = table->n_devs;
+						break;
+					}
+					i7core_put_all_devices();
+					return -ENODEV;
+				}
+			} while (pdev);
+		}
+		table++;
+	}
+
+	return 0;
+	return 0;
+}
+
+static int mci_bind_devs(struct mem_ctl_info *mci,
+			 struct i7core_dev *i7core_dev)
+{
+	struct i7core_pvt *pvt = mci->pvt_info;
+	struct pci_dev *pdev;
+	int i, func, slot;
+
+	/* Associates i7core_dev and mci for future usage */
+	pvt->i7core_dev = i7core_dev;
+	i7core_dev->mci = mci;
+
+	pvt->is_registered = 0;
+	for (i = 0; i < i7core_dev->n_devs; i++) {
+		pdev = i7core_dev->pdev[i];
+		if (!pdev)
+			continue;
+
+		func = PCI_FUNC(pdev->devfn);
+		slot = PCI_SLOT(pdev->devfn);
+		if (slot == 3) {
+			if (unlikely(func > MAX_MCR_FUNC))
+				goto error;
+			pvt->pci_mcr[func] = pdev;
+		} else if (likely(slot >= 4 && slot < 4 + NUM_CHANS)) {
+			if (unlikely(func > MAX_CHAN_FUNC))
+				goto error;
+			pvt->pci_ch[slot - 4][func] = pdev;
+		} else if (!slot && !func)
+			pvt->pci_noncore = pdev;
+		else
+			goto error;
+
+		debugf0("Associated fn %d.%d, dev = %p, socket %d\n",
+			PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn),
+			pdev, i7core_dev->socket);
+
+		if (PCI_SLOT(pdev->devfn) == 3 &&
+			PCI_FUNC(pdev->devfn) == 2)
+			pvt->is_registered = 1;
+	}
+
+	/*
+	 * Add extra nodes to count errors on udimm
+	 * For registered memory, this is not needed, since the counters
+	 * are already displayed at the standard locations
+	 */
+	if (!pvt->is_registered)
+		i7core_sysfs_attrs[ARRAY_SIZE(i7core_sysfs_attrs)-2].grp =
+			&i7core_udimm_counters;
+
+	return 0;
+
+error:
+	i7core_printk(KERN_ERR, "Device %d, function %d "
+		      "is out of the expected range\n",
+		      slot, func);
+	return -EINVAL;
+}
+
+/****************************************************************************
+			Error check routines
+ ****************************************************************************/
+static void i7core_rdimm_update_csrow(struct mem_ctl_info *mci,
+					 int chan, int dimm, int add)
+{
+	char *msg;
+	struct i7core_pvt *pvt = mci->pvt_info;
+	int row = pvt->csrow_map[chan][dimm], i;
+
+	for (i = 0; i < add; i++) {
+		msg = kasprintf(GFP_KERNEL, "Corrected error "
+				"(Socket=%d channel=%d dimm=%d)",
+				pvt->i7core_dev->socket, chan, dimm);
+
+		edac_mc_handle_fbd_ce(mci, row, 0, msg);
+		kfree (msg);
+	}
+}
+
+static void i7core_rdimm_update_ce_count(struct mem_ctl_info *mci,
+			int chan, int new0, int new1, int new2)
+{
+	struct i7core_pvt *pvt = mci->pvt_info;
+	int add0 = 0, add1 = 0, add2 = 0;
+	/* Updates CE counters if it is not the first time here */
+	if (pvt->ce_count_available) {
+		/* Updates CE counters */
+
+		add2 = new2 - pvt->rdimm_last_ce_count[chan][2];
+		add1 = new1 - pvt->rdimm_last_ce_count[chan][1];
+		add0 = new0 - pvt->rdimm_last_ce_count[chan][0];
+
+		if (add2 < 0)
+			add2 += 0x7fff;
+		pvt->rdimm_ce_count[chan][2] += add2;
+
+		if (add1 < 0)
+			add1 += 0x7fff;
+		pvt->rdimm_ce_count[chan][1] += add1;
+
+		if (add0 < 0)
+			add0 += 0x7fff;
+		pvt->rdimm_ce_count[chan][0] += add0;
+	} else
+		pvt->ce_count_available = 1;
+
+	/* Store the new values */
+	pvt->rdimm_last_ce_count[chan][2] = new2;
+	pvt->rdimm_last_ce_count[chan][1] = new1;
+	pvt->rdimm_last_ce_count[chan][0] = new0;
+
+	/*updated the edac core */
+	if (add0 != 0)
+		i7core_rdimm_update_csrow(mci, chan, 0, add0);
+	if (add1 != 0)
+		i7core_rdimm_update_csrow(mci, chan, 1, add1);
+	if (add2 != 0)
+		i7core_rdimm_update_csrow(mci, chan, 2, add2);
+
+}
+
+static void i7core_rdimm_check_mc_ecc_err(struct mem_ctl_info *mci)
+{
+	struct i7core_pvt *pvt = mci->pvt_info;
+	u32 rcv[3][2];
+	int i, new0, new1, new2;
+
+	/*Read DEV 3: FUN 2:  MC_COR_ECC_CNT regs directly*/
+	pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_0,
+								&rcv[0][0]);
+	pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_1,
+								&rcv[0][1]);
+	pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_2,
+								&rcv[1][0]);
+	pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_3,
+								&rcv[1][1]);
+	pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_4,
+								&rcv[2][0]);
+	pci_read_config_dword(pvt->pci_mcr[2], MC_COR_ECC_CNT_5,
+								&rcv[2][1]);
+	for (i = 0 ; i < 3; i++) {
+		debugf3("MC_COR_ECC_CNT%d = 0x%x; MC_COR_ECC_CNT%d = 0x%x\n",
+			(i * 2), rcv[i][0], (i * 2) + 1, rcv[i][1]);
+		/*if the channel has 3 dimms*/
+		if (pvt->channel[i].dimms > 2) {
+			new0 = DIMM_BOT_COR_ERR(rcv[i][0]);
+			new1 = DIMM_TOP_COR_ERR(rcv[i][0]);
+			new2 = DIMM_BOT_COR_ERR(rcv[i][1]);
+		} else {
+			new0 = DIMM_TOP_COR_ERR(rcv[i][0]) +
+					DIMM_BOT_COR_ERR(rcv[i][0]);
+			new1 = DIMM_TOP_COR_ERR(rcv[i][1]) +
+					DIMM_BOT_COR_ERR(rcv[i][1]);
+			new2 = 0;
+		}
+
+		i7core_rdimm_update_ce_count(mci, i, new0, new1, new2);
+	}
+}
+
+/* This function is based on the device 3 function 4 registers as described on:
+ * Intel Xeon Processor 5500 Series Datasheet Volume 2
+ *	http://www.intel.com/Assets/PDF/datasheet/321322.pdf
+ * also available at:
+ * 	http://www.arrownac.com/manufacturers/intel/s/nehalem/5500-datasheet-v2.pdf
+ */
+static void i7core_udimm_check_mc_ecc_err(struct mem_ctl_info *mci)
+{
+	struct i7core_pvt *pvt = mci->pvt_info;
+	u32 rcv1, rcv0;
+	int new0, new1, new2;
+
+	if (!pvt->pci_mcr[4]) {
+		debugf0("%s MCR registers not found\n", __func__);
+		return;
+	}
+
+	/* Corrected test errors */
+	pci_read_config_dword(pvt->pci_mcr[4], MC_TEST_ERR_RCV1, &rcv1);
+	pci_read_config_dword(pvt->pci_mcr[4], MC_TEST_ERR_RCV0, &rcv0);
+
+	/* Store the new values */
+	new2 = DIMM2_COR_ERR(rcv1);
+	new1 = DIMM1_COR_ERR(rcv0);
+	new0 = DIMM0_COR_ERR(rcv0);
+
+	/* Updates CE counters if it is not the first time here */
+	if (pvt->ce_count_available) {
+		/* Updates CE counters */
+		int add0, add1, add2;
+
+		add2 = new2 - pvt->udimm_last_ce_count[2];
+		add1 = new1 - pvt->udimm_last_ce_count[1];
+		add0 = new0 - pvt->udimm_last_ce_count[0];
+
+		if (add2 < 0)
+			add2 += 0x7fff;
+		pvt->udimm_ce_count[2] += add2;
+
+		if (add1 < 0)
+			add1 += 0x7fff;
+		pvt->udimm_ce_count[1] += add1;
+
+		if (add0 < 0)
+			add0 += 0x7fff;
+		pvt->udimm_ce_count[0] += add0;
+
+		if (add0 | add1 | add2)
+			i7core_printk(KERN_ERR, "New Corrected error(s): "
+				      "dimm0: +%d, dimm1: +%d, dimm2 +%d\n",
+				      add0, add1, add2);
+	} else
+		pvt->ce_count_available = 1;
+
+	/* Store the new values */
+	pvt->udimm_last_ce_count[2] = new2;
+	pvt->udimm_last_ce_count[1] = new1;
+	pvt->udimm_last_ce_count[0] = new0;
+}
+
+/*
+ * According with tables E-11 and E-12 of chapter E.3.3 of Intel 64 and IA-32
+ * Architectures Software Developer’s Manual Volume 3B.
+ * Nehalem are defined as family 0x06, model 0x1a
+ *
+ * The MCA registers used here are the following ones:
+ *     struct mce field	MCA Register
+ *     m->status	MSR_IA32_MC8_STATUS
+ *     m->addr		MSR_IA32_MC8_ADDR
+ *     m->misc		MSR_IA32_MC8_MISC
+ * In the case of Nehalem, the error information is masked at .status and .misc
+ * fields
+ */
+static void i7core_mce_output_error(struct mem_ctl_info *mci,
+				    struct mce *m)
+{
+	struct i7core_pvt *pvt = mci->pvt_info;
+	char *type, *optype, *err, *msg;
+	unsigned long error = m->status & 0x1ff0000l;
+	u32 optypenum = (m->status >> 4) & 0x07;
+	u32 core_err_cnt = (m->status >> 38) && 0x7fff;
+	u32 dimm = (m->misc >> 16) & 0x3;
+	u32 channel = (m->misc >> 18) & 0x3;
+	u32 syndrome = m->misc >> 32;
+	u32 errnum = find_first_bit(&error, 32);
+	int csrow;
+
+	if (m->mcgstatus & 1)
+		type = "FATAL";
+	else
+		type = "NON_FATAL";
+
+	switch (optypenum) {
+	case 0:
+		optype = "generic undef request";
+		break;
+	case 1:
+		optype = "read error";
+		break;
+	case 2:
+		optype = "write error";
+		break;
+	case 3:
+		optype = "addr/cmd error";
+		break;
+	case 4:
+		optype = "scrubbing error";
+		break;
+	default:
+		optype = "reserved";
+		break;
+	}
+
+	switch (errnum) {
+	case 16:
+		err = "read ECC error";
+		break;
+	case 17:
+		err = "RAS ECC error";
+		break;
+	case 18:
+		err = "write parity error";
+		break;
+	case 19:
+		err = "redundacy loss";
+		break;
+	case 20:
+		err = "reserved";
+		break;
+	case 21:
+		err = "memory range error";
+		break;
+	case 22:
+		err = "RTID out of range";
+		break;
+	case 23:
+		err = "address parity error";
+		break;
+	case 24:
+		err = "byte enable parity error";
+		break;
+	default:
+		err = "unknown";
+	}
+
+	/* FIXME: should convert addr into bank and rank information */
+	msg = kasprintf(GFP_ATOMIC,
+		"%s (addr = 0x%08llx, cpu=%d, Dimm=%d, Channel=%d, "
+		"syndrome=0x%08x, count=%d, Err=%08llx:%08llx (%s: %s))\n",
+		type, (long long) m->addr, m->cpu, dimm, channel,
+		syndrome, core_err_cnt, (long long)m->status,
+		(long long)m->misc, optype, err);
+
+	debugf0("%s", msg);
+
+	csrow = pvt->csrow_map[channel][dimm];
+
+	/* Call the helper to output message */
+	if (m->mcgstatus & 1)
+		edac_mc_handle_fbd_ue(mci, csrow, 0,
+				0 /* FIXME: should be channel here */, msg);
+	else if (!pvt->is_registered)
+		edac_mc_handle_fbd_ce(mci, csrow,
+				0 /* FIXME: should be channel here */, msg);
+
+	kfree(msg);
+}
+
+/*
+ *	i7core_check_error	Retrieve and process errors reported by the
+ *				hardware. Called by the Core module.
+ */
+static void i7core_check_error(struct mem_ctl_info *mci)
+{
+	struct i7core_pvt *pvt = mci->pvt_info;
+	int i;
+	unsigned count = 0;
+	struct mce *m;
+
+	/*
+	 * MCE first step: Copy all mce errors into a temporary buffer
+	 * We use a double buffering here, to reduce the risk of
+	 * loosing an error.
+	 */
+	smp_rmb();
+	count = (pvt->mce_out + MCE_LOG_LEN - pvt->mce_in)
+		% MCE_LOG_LEN;
+	if (!count)
+		goto check_ce_error;
+
+	m = pvt->mce_outentry;
+	if (pvt->mce_in + count > MCE_LOG_LEN) {
+		unsigned l = MCE_LOG_LEN - pvt->mce_in;
+
+		memcpy(m, &pvt->mce_entry[pvt->mce_in], sizeof(*m) * l);
+		smp_wmb();
+		pvt->mce_in = 0;
+		count -= l;
+		m += l;
+	}
+	memcpy(m, &pvt->mce_entry[pvt->mce_in], sizeof(*m) * count);
+	smp_wmb();
+	pvt->mce_in += count;
+
+	smp_rmb();
+	if (pvt->mce_overrun) {
+		i7core_printk(KERN_ERR, "Lost %d memory errors\n",
+			      pvt->mce_overrun);
+		smp_wmb();
+		pvt->mce_overrun = 0;
+	}
+
+	/*
+	 * MCE second step: parse errors and display
+	 */
+	for (i = 0; i < count; i++)
+		i7core_mce_output_error(mci, &pvt->mce_outentry[i]);
+
+	/*
+	 * Now, let's increment CE error counts
+	 */
+check_ce_error:
+	if (!pvt->is_registered)
+		i7core_udimm_check_mc_ecc_err(mci);
+	else
+		i7core_rdimm_check_mc_ecc_err(mci);
+}
+
+/*
+ * i7core_mce_check_error	Replicates mcelog routine to get errors
+ *				This routine simply queues mcelog errors, and
+ *				return. The error itself should be handled later
+ *				by i7core_check_error.
+ * WARNING: As this routine should be called at NMI time, extra care should
+ * be taken to avoid deadlocks, and to be as fast as possible.
+ */
+static int i7core_mce_check_error(void *priv, struct mce *mce)
+{
+	struct mem_ctl_info *mci = priv;
+	struct i7core_pvt *pvt = mci->pvt_info;
+
+	/*
+	 * Just let mcelog handle it if the error is
+	 * outside the memory controller
+	 */
+	if (((mce->status & 0xffff) >> 7) != 1)
+		return 0;
+
+	/* Bank 8 registers are the only ones that we know how to handle */
+	if (mce->bank != 8)
+		return 0;
+
+#ifdef CONFIG_SMP
+	/* Only handle if it is the right mc controller */
+	if (cpu_data(mce->cpu).phys_proc_id != pvt->i7core_dev->socket)
+		return 0;
+#endif
+
+	smp_rmb();
+	if ((pvt->mce_out + 1) % MCE_LOG_LEN == pvt->mce_in) {
+		smp_wmb();
+		pvt->mce_overrun++;
+		return 0;
+	}
+
+	/* Copy memory error at the ringbuffer */
+	memcpy(&pvt->mce_entry[pvt->mce_out], mce, sizeof(*mce));
+	smp_wmb();
+	pvt->mce_out = (pvt->mce_out + 1) % MCE_LOG_LEN;
+
+	/* Handle fatal errors immediately */
+	if (mce->mcgstatus & 1)
+		i7core_check_error(mci);
+
+	/* Advice mcelog that the error were handled */
+	return 1;
+}
+
+static int i7core_register_mci(struct i7core_dev *i7core_dev,
+			       int num_channels, int num_csrows)
+{
+	struct mem_ctl_info *mci;
+	struct i7core_pvt *pvt;
+	int csrow = 0;
+	int rc;
+
+	/* allocate a new MC control structure */
+	mci = edac_mc_alloc(sizeof(*pvt), num_csrows, num_channels,
+			    i7core_dev->socket);
+	if (unlikely(!mci))
+		return -ENOMEM;
+
+	debugf0("MC: " __FILE__ ": %s(): mci = %p\n", __func__, mci);
+
+	/* record ptr to the generic device */
+	mci->dev = &i7core_dev->pdev[0]->dev;
+
+	pvt = mci->pvt_info;
+	memset(pvt, 0, sizeof(*pvt));
+
+	/*
+	 * FIXME: how to handle RDDR3 at MCI level? It is possible to have
+	 * Mixed RDDR3/UDDR3 with Nehalem, provided that they are on different
+	 * memory channels
+	 */
+	mci->mtype_cap = MEM_FLAG_DDR3;
+	mci->edac_ctl_cap = EDAC_FLAG_NONE;
+	mci->edac_cap = EDAC_FLAG_NONE;
+	mci->mod_name = "i7core_edac.c";
+	mci->mod_ver = I7CORE_REVISION;
+	mci->ctl_name = kasprintf(GFP_KERNEL, "i7 core #%d",
+				  i7core_dev->socket);
+	mci->dev_name = pci_name(i7core_dev->pdev[0]);
+	mci->ctl_page_to_phys = NULL;
+	mci->mc_driver_sysfs_attributes = i7core_sysfs_attrs;
+	/* Set the function pointer to an actual operation function */
+	mci->edac_check = i7core_check_error;
+
+	/* Store pci devices at mci for faster access */
+	rc = mci_bind_devs(mci, i7core_dev);
+	if (unlikely(rc < 0))
+		goto fail;
+
+	/* Get dimm basic config */
+	get_dimm_config(mci, &csrow);
+
+	/* add this new MC control structure to EDAC's list of MCs */
+	if (unlikely(edac_mc_add_mc(mci))) {
+		debugf0("MC: " __FILE__
+			": %s(): failed edac_mc_add_mc()\n", __func__);
+		/* FIXME: perhaps some code should go here that disables error
+		 * reporting if we just enabled it
+		 */
+
+		rc = -EINVAL;
+		goto fail;
+	}
+
+	/* allocating generic PCI control info */
+	i7core_pci = edac_pci_create_generic_ctl(&i7core_dev->pdev[0]->dev,
+						 EDAC_MOD_STR);
+	if (unlikely(!i7core_pci)) {
+		printk(KERN_WARNING
+			"%s(): Unable to create PCI control\n",
+			__func__);
+		printk(KERN_WARNING
+			"%s(): PCI error report via EDAC not setup\n",
+			__func__);
+	}
+
+	/* Default error mask is any memory */
+	pvt->inject.channel = 0;
+	pvt->inject.dimm = -1;
+	pvt->inject.rank = -1;
+	pvt->inject.bank = -1;
+	pvt->inject.page = -1;
+	pvt->inject.col = -1;
+
+	/* Registers on edac_mce in order to receive memory errors */
+	pvt->edac_mce.priv = mci;
+	pvt->edac_mce.check_error = i7core_mce_check_error;
+
+	rc = edac_mce_register(&pvt->edac_mce);
+	if (unlikely(rc < 0)) {
+		debugf0("MC: " __FILE__
+			": %s(): failed edac_mce_register()\n", __func__);
+	}
+
+fail:
+	if (rc < 0)
+		edac_mc_free(mci);
+	return rc;
+}
+
+/*
+ *	i7core_probe	Probe for ONE instance of device to see if it is
+ *			present.
+ *	return:
+ *		0 for FOUND a device
+ *		< 0 for error code
+ */
+static int __devinit i7core_probe(struct pci_dev *pdev,
+				  const struct pci_device_id *id)
+{
+	int dev_idx = id->driver_data;
+	int rc;
+	struct i7core_dev *i7core_dev;
+
+	/*
+	 * All memory controllers are allocated at the first pass.
+	 */
+	if (unlikely(dev_idx >= 1))
+		return -EINVAL;
+
+	/* get the pci devices we want to reserve for our use */
+	mutex_lock(&i7core_edac_lock);
+
+	rc = i7core_get_devices(pci_dev_table);
+	if (unlikely(rc < 0))
+		goto fail0;
+
+	list_for_each_entry(i7core_dev, &i7core_edac_list, list) {
+		int channels;
+		int csrows;
+
+		/* Check the number of active and not disabled channels */
+		rc = i7core_get_active_channels(i7core_dev->socket,
+						&channels, &csrows);
+		if (unlikely(rc < 0))
+			goto fail1;
+
+		rc = i7core_register_mci(i7core_dev, channels, csrows);
+		if (unlikely(rc < 0))
+			goto fail1;
+	}
+
+	i7core_printk(KERN_INFO, "Driver loaded.\n");
+
+	mutex_unlock(&i7core_edac_lock);
+	return 0;
+
+fail1:
+	i7core_put_all_devices();
+fail0:
+	mutex_unlock(&i7core_edac_lock);
+	return rc;
+}
+
+/*
+ *	i7core_remove	destructor for one instance of device
+ *
+ */
+static void __devexit i7core_remove(struct pci_dev *pdev)
+{
+	struct mem_ctl_info *mci;
+	struct i7core_dev *i7core_dev, *tmp;
+
+	debugf0(__FILE__ ": %s()\n", __func__);
+
+	if (i7core_pci)
+		edac_pci_release_generic_ctl(i7core_pci);
+
+	/*
+	 * we have a trouble here: pdev value for removal will be wrong, since
+	 * it will point to the X58 register used to detect that the machine
+	 * is a Nehalem or upper design. However, due to the way several PCI
+	 * devices are grouped together to provide MC functionality, we need
+	 * to use a different method for releasing the devices
+	 */
+
+	mutex_lock(&i7core_edac_lock);
+	list_for_each_entry_safe(i7core_dev, tmp, &i7core_edac_list, list) {
+		mci = edac_mc_del_mc(&i7core_dev->pdev[0]->dev);
+		if (mci) {
+			struct i7core_pvt *pvt = mci->pvt_info;
+
+			i7core_dev = pvt->i7core_dev;
+			edac_mce_unregister(&pvt->edac_mce);
+			kfree(mci->ctl_name);
+			edac_mc_free(mci);
+			i7core_put_devices(i7core_dev);
+		} else {
+			i7core_printk(KERN_ERR,
+				      "Couldn't find mci for socket %d\n",
+				      i7core_dev->socket);
+		}
+	}
+	mutex_unlock(&i7core_edac_lock);
+}
+
+MODULE_DEVICE_TABLE(pci, i7core_pci_tbl);
+
+/*
+ *	i7core_driver	pci_driver structure for this module
+ *
+ */
+static struct pci_driver i7core_driver = {
+	.name     = "i7core_edac",
+	.probe    = i7core_probe,
+	.remove   = __devexit_p(i7core_remove),
+	.id_table = i7core_pci_tbl,
+};
+
+/*
+ *	i7core_init		Module entry function
+ *			Try to initialize this module for its devices
+ */
+static int __init i7core_init(void)
+{
+	int pci_rc;
+
+	debugf2("MC: " __FILE__ ": %s()\n", __func__);
+
+	/* Ensure that the OPSTATE is set correctly for POLL or NMI */
+	opstate_init();
+
+	i7core_xeon_pci_fixup(pci_dev_table);
+
+	pci_rc = pci_register_driver(&i7core_driver);
+
+	if (pci_rc >= 0)
+		return 0;
+
+	i7core_printk(KERN_ERR, "Failed to register device with error %d.\n",
+		      pci_rc);
+
+	return pci_rc;
+}
+
+/*
+ *	i7core_exit()	Module exit function
+ *			Unregister the driver
+ */
+static void __exit i7core_exit(void)
+{
+	debugf2("MC: " __FILE__ ": %s()\n", __func__);
+	pci_unregister_driver(&i7core_driver);
+}
+
+module_init(i7core_init);
+module_exit(i7core_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Mauro Carvalho Chehab <mchehab@redhat.com>");
+MODULE_AUTHOR("Red Hat Inc. (http://www.redhat.com)");
+MODULE_DESCRIPTION("MC Driver for Intel i7 Core memory controllers - "
+		   I7CORE_REVISION);
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
diff --git a/drivers/edac/i82443bxgx_edac.c b/drivers/edac/i82443bxgx_edac.c
index 7f3884fcbd46..a2fa1feed724 100644
--- a/drivers/edac/i82443bxgx_edac.c
+++ b/drivers/edac/i82443bxgx_edac.c
@@ -178,7 +178,7 @@ static void i82443bxgx_edacmc_check(struct mem_ctl_info *mci)
 {
 	struct i82443bxgx_edacmc_error_info info;
 
-	debugf1("MC%d: " __FILE__ ": %s()\n", mci->mc_idx, __func__);
+	debugf1("MC%d: %s: %s()\n", mci->mc_idx, __FILE__, __func__);
 	i82443bxgx_edacmc_get_error_info(mci, &info);
 	i82443bxgx_edacmc_process_error_info(mci, &info, 1);
 }
@@ -198,13 +198,13 @@ static void i82443bxgx_init_csrows(struct mem_ctl_info *mci,
 	for (index = 0; index < mci->nr_csrows; index++) {
 		csrow = &mci->csrows[index];
 		pci_read_config_byte(pdev, I82443BXGX_DRB + index, &drbar);
-		debugf1("MC%d: " __FILE__ ": %s() Row=%d DRB = %#0x\n",
-			mci->mc_idx, __func__, index, drbar);
+		debugf1("MC%d: %s: %s() Row=%d DRB = %#0x\n",
+			mci->mc_idx, __FILE__, __func__, index, drbar);
 		row_high_limit = ((u32) drbar << 23);
 		/* find the DRAM Chip Select Base address and mask */
-		debugf1("MC%d: " __FILE__ ": %s() Row=%d, "
-			"Boundry Address=%#0x, Last = %#0x \n",
-			mci->mc_idx, __func__, index, row_high_limit,
+		debugf1("MC%d: %s: %s() Row=%d, "
+			"Boundry Address=%#0x, Last = %#0x\n",
+			mci->mc_idx, __FILE__, __func__, index, row_high_limit,
 			row_high_limit_last);
 
 		/* 440GX goes to 2GB, represented with a DRB of 0. */
@@ -237,7 +237,7 @@ static int i82443bxgx_edacmc_probe1(struct pci_dev *pdev, int dev_idx)
 	enum mem_type mtype;
 	enum edac_type edac_mode;
 
-	debugf0("MC: " __FILE__ ": %s()\n", __func__);
+	debugf0("MC: %s: %s()\n", __FILE__, __func__);
 
 	/* Something is really hosed if PCI config space reads from
 	 * the MC aren't working.
@@ -250,7 +250,7 @@ static int i82443bxgx_edacmc_probe1(struct pci_dev *pdev, int dev_idx)
 	if (mci == NULL)
 		return -ENOMEM;
 
-	debugf0("MC: " __FILE__ ": %s(): mci = %p\n", __func__, mci);
+	debugf0("MC: %s: %s(): mci = %p\n", __FILE__, __func__, mci);
 	mci->dev = &pdev->dev;
 	mci->mtype_cap = MEM_FLAG_EDO | MEM_FLAG_SDR | MEM_FLAG_RDR;
 	mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_EC | EDAC_FLAG_SECDED;
@@ -336,7 +336,7 @@ static int i82443bxgx_edacmc_probe1(struct pci_dev *pdev, int dev_idx)
 			__func__);
 	}
 
-	debugf3("MC: " __FILE__ ": %s(): success\n", __func__);
+	debugf3("MC: %s: %s(): success\n", __FILE__, __func__);
 	return 0;
 
 fail:
@@ -352,9 +352,9 @@ static int __devinit i82443bxgx_edacmc_init_one(struct pci_dev *pdev,
 {
 	int rc;
 
-	debugf0("MC: " __FILE__ ": %s()\n", __func__);
+	debugf0("MC: %s: %s()\n", __FILE__, __func__);
 
-	/* don't need to call pci_device_enable() */
+	/* don't need to call pci_enable_device() */
 	rc = i82443bxgx_edacmc_probe1(pdev, ent->driver_data);
 
 	if (mci_pdev == NULL)
@@ -367,7 +367,7 @@ static void __devexit i82443bxgx_edacmc_remove_one(struct pci_dev *pdev)
 {
 	struct mem_ctl_info *mci;
 
-	debugf0(__FILE__ ": %s()\n", __func__);
+	debugf0("%s: %s()\n", __FILE__, __func__);
 
 	if (i82443bxgx_pci)
 		edac_pci_release_generic_ctl(i82443bxgx_pci);
diff --git a/drivers/edac/mpc85xx_edac.c b/drivers/edac/mpc85xx_edac.c
index 4471647b4807..52ca09bf4726 100644
--- a/drivers/edac/mpc85xx_edac.c
+++ b/drivers/edac/mpc85xx_edac.c
@@ -229,7 +229,7 @@ static int __devinit mpc85xx_pci_err_probe(struct of_device *op,
 
 	pdata->edac_idx = edac_pci_idx++;
 
-	res = of_address_to_resource(op->node, 0, &r);
+	res = of_address_to_resource(op->dev.of_node, 0, &r);
 	if (res) {
 		printk(KERN_ERR "%s: Unable to get resource for "
 		       "PCI err regs\n", __func__);
@@ -274,7 +274,7 @@ static int __devinit mpc85xx_pci_err_probe(struct of_device *op,
 	}
 
 	if (edac_op_state == EDAC_OPSTATE_INT) {
-		pdata->irq = irq_of_parse_and_map(op->node, 0);
+		pdata->irq = irq_of_parse_and_map(op->dev.of_node, 0);
 		res = devm_request_irq(&op->dev, pdata->irq,
 				       mpc85xx_pci_isr, IRQF_DISABLED,
 				       "[EDAC] PCI err", pci);
@@ -338,15 +338,13 @@ static struct of_device_id mpc85xx_pci_err_of_match[] = {
 };
 
 static struct of_platform_driver mpc85xx_pci_err_driver = {
-	.owner = THIS_MODULE,
-	.name = "mpc85xx_pci_err",
-	.match_table = mpc85xx_pci_err_of_match,
 	.probe = mpc85xx_pci_err_probe,
 	.remove = __devexit_p(mpc85xx_pci_err_remove),
 	.driver = {
-		   .name = "mpc85xx_pci_err",
-		   .owner = THIS_MODULE,
-		   },
+		.name = "mpc85xx_pci_err",
+		.owner = THIS_MODULE,
+		.of_match_table = mpc85xx_pci_err_of_match,
+	},
 };
 
 #endif				/* CONFIG_PCI */
@@ -531,7 +529,7 @@ static int __devinit mpc85xx_l2_err_probe(struct of_device *op,
 	edac_dev->ctl_name = pdata->name;
 	edac_dev->dev_name = pdata->name;
 
-	res = of_address_to_resource(op->node, 0, &r);
+	res = of_address_to_resource(op->dev.of_node, 0, &r);
 	if (res) {
 		printk(KERN_ERR "%s: Unable to get resource for "
 		       "L2 err regs\n", __func__);
@@ -578,7 +576,7 @@ static int __devinit mpc85xx_l2_err_probe(struct of_device *op,
 	}
 
 	if (edac_op_state == EDAC_OPSTATE_INT) {
-		pdata->irq = irq_of_parse_and_map(op->node, 0);
+		pdata->irq = irq_of_parse_and_map(op->dev.of_node, 0);
 		res = devm_request_irq(&op->dev, pdata->irq,
 				       mpc85xx_l2_isr, IRQF_DISABLED,
 				       "[EDAC] L2 err", edac_dev);
@@ -654,15 +652,13 @@ static struct of_device_id mpc85xx_l2_err_of_match[] = {
 };
 
 static struct of_platform_driver mpc85xx_l2_err_driver = {
-	.owner = THIS_MODULE,
-	.name = "mpc85xx_l2_err",
-	.match_table = mpc85xx_l2_err_of_match,
 	.probe = mpc85xx_l2_err_probe,
 	.remove = mpc85xx_l2_err_remove,
 	.driver = {
-		   .name = "mpc85xx_l2_err",
-		   .owner = THIS_MODULE,
-		   },
+		.name = "mpc85xx_l2_err",
+		.owner = THIS_MODULE,
+		.of_match_table = mpc85xx_l2_err_of_match,
+	},
 };
 
 /**************************** MC Err device ***************************/
@@ -982,7 +978,7 @@ static int __devinit mpc85xx_mc_err_probe(struct of_device *op,
 	mci->ctl_name = pdata->name;
 	mci->dev_name = pdata->name;
 
-	res = of_address_to_resource(op->node, 0, &r);
+	res = of_address_to_resource(op->dev.of_node, 0, &r);
 	if (res) {
 		printk(KERN_ERR "%s: Unable to get resource for MC err regs\n",
 		       __func__);
@@ -1056,7 +1052,7 @@ static int __devinit mpc85xx_mc_err_probe(struct of_device *op,
 		out_be32(pdata->mc_vbase + MPC85XX_MC_ERR_SBE, 0x10000);
 
 		/* register interrupts */
-		pdata->irq = irq_of_parse_and_map(op->node, 0);
+		pdata->irq = irq_of_parse_and_map(op->dev.of_node, 0);
 		res = devm_request_irq(&op->dev, pdata->irq,
 				       mpc85xx_mc_isr,
 					IRQF_DISABLED | IRQF_SHARED,
@@ -1131,15 +1127,13 @@ static struct of_device_id mpc85xx_mc_err_of_match[] = {
 };
 
 static struct of_platform_driver mpc85xx_mc_err_driver = {
-	.owner = THIS_MODULE,
-	.name = "mpc85xx_mc_err",
-	.match_table = mpc85xx_mc_err_of_match,
 	.probe = mpc85xx_mc_err_probe,
 	.remove = mpc85xx_mc_err_remove,
 	.driver = {
-		   .name = "mpc85xx_mc_err",
-		   .owner = THIS_MODULE,
-		   },
+		.name = "mpc85xx_mc_err",
+		.owner = THIS_MODULE,
+		.of_match_table = mpc85xx_mc_err_of_match,
+	},
 };
 
 #ifdef CONFIG_MPC85xx
diff --git a/drivers/edac/ppc4xx_edac.c b/drivers/edac/ppc4xx_edac.c
index 11f2172aa1e6..e78839e89a06 100644
--- a/drivers/edac/ppc4xx_edac.c
+++ b/drivers/edac/ppc4xx_edac.c
@@ -202,13 +202,13 @@ static struct of_device_id ppc4xx_edac_match[] = {
 };
 
 static struct of_platform_driver ppc4xx_edac_driver = {
-	.match_table		= ppc4xx_edac_match,
 	.probe			= ppc4xx_edac_probe,
 	.remove			= ppc4xx_edac_remove,
-	.driver			= {
-		.owner	= THIS_MODULE,
-		.name	= PPC4XX_EDAC_MODULE_NAME
-	}
+	.driver = {
+		.owner = THIS_MODULE,
+		.name = PPC4XX_EDAC_MODULE_NAME
+		.of_match_table = ppc4xx_edac_match,
+	},
 };
 
 /*
@@ -1022,7 +1022,7 @@ ppc4xx_edac_mc_init(struct mem_ctl_info *mci,
 	int status = 0;
 	const u32 memcheck = (mcopt1 & SDRAM_MCOPT1_MCHK_MASK);
 	struct ppc4xx_edac_pdata *pdata = NULL;
-	const struct device_node *np = op->node;
+	const struct device_node *np = op->dev.of_node;
 
 	if (match == NULL)
 		return -EINVAL;
@@ -1113,7 +1113,7 @@ ppc4xx_edac_register_irq(struct of_device *op, struct mem_ctl_info *mci)
 	int status = 0;
 	int ded_irq, sec_irq;
 	struct ppc4xx_edac_pdata *pdata = mci->pvt_info;
-	struct device_node *np = op->node;
+	struct device_node *np = op->dev.of_node;
 
 	ded_irq = irq_of_parse_and_map(np, INTMAP_ECCDED_INDEX);
 	sec_irq = irq_of_parse_and_map(np, INTMAP_ECCSEC_INDEX);
@@ -1243,7 +1243,7 @@ ppc4xx_edac_probe(struct of_device *op, const struct of_device_id *match)
 	int status = 0;
 	u32 mcopt1, memcheck;
 	dcr_host_t dcr_host;
-	const struct device_node *np = op->node;
+	const struct device_node *np = op->dev.of_node;
 	struct mem_ctl_info *mci = NULL;
 	static int ppc4xx_edac_instance;
 
diff --git a/drivers/edac/r82600_edac.c b/drivers/edac/r82600_edac.c
index d55f8e9de788..6a822c631ef5 100644
--- a/drivers/edac/r82600_edac.c
+++ b/drivers/edac/r82600_edac.c
@@ -354,7 +354,7 @@ static int __devinit r82600_init_one(struct pci_dev *pdev,
 {
 	debugf0("%s()\n", __func__);
 
-	/* don't need to call pci_device_enable() */
+	/* don't need to call pci_enable_device() */
 	return r82600_probe1(pdev, ent->driver_data);
 }
 
diff --git a/drivers/firewire/core-card.c b/drivers/firewire/core-card.c
index 5045156c5313..371713ff0266 100644
--- a/drivers/firewire/core-card.c
+++ b/drivers/firewire/core-card.c
@@ -30,7 +30,6 @@
 #include <linux/module.h>
 #include <linux/mutex.h>
 #include <linux/spinlock.h>
-#include <linux/timer.h>
 #include <linux/workqueue.h>
 
 #include <asm/atomic.h>
@@ -63,7 +62,7 @@ static size_t config_rom_length = 1 + 4 + 1 + 1;
 #define BIB_CRC(v)		((v) <<  0)
 #define BIB_CRC_LENGTH(v)	((v) << 16)
 #define BIB_INFO_LENGTH(v)	((v) << 24)
-
+#define BIB_BUS_NAME		0x31333934 /* "1394" */
 #define BIB_LINK_SPEED(v)	((v) <<  0)
 #define BIB_GENERATION(v)	((v) <<  4)
 #define BIB_MAX_ROM(v)		((v) <<  8)
@@ -73,7 +72,8 @@ static size_t config_rom_length = 1 + 4 + 1 + 1;
 #define BIB_BMC			((1) << 28)
 #define BIB_ISC			((1) << 29)
 #define BIB_CMC			((1) << 30)
-#define BIB_IMC			((1) << 31)
+#define BIB_IRMC		((1) << 31)
+#define NODE_CAPABILITIES	0x0c0083c0 /* per IEEE 1394 clause 8.3.2.6.5.2 */
 
 static void generate_config_rom(struct fw_card *card, __be32 *config_rom)
 {
@@ -91,18 +91,18 @@ static void generate_config_rom(struct fw_card *card, __be32 *config_rom)
 
 	config_rom[0] = cpu_to_be32(
 		BIB_CRC_LENGTH(4) | BIB_INFO_LENGTH(4) | BIB_CRC(0));
-	config_rom[1] = cpu_to_be32(0x31333934);
+	config_rom[1] = cpu_to_be32(BIB_BUS_NAME);
 	config_rom[2] = cpu_to_be32(
 		BIB_LINK_SPEED(card->link_speed) |
 		BIB_GENERATION(card->config_rom_generation++ % 14 + 2) |
 		BIB_MAX_ROM(2) |
 		BIB_MAX_RECEIVE(card->max_receive) |
-		BIB_BMC | BIB_ISC | BIB_CMC | BIB_IMC);
+		BIB_BMC | BIB_ISC | BIB_CMC | BIB_IRMC);
 	config_rom[3] = cpu_to_be32(card->guid >> 32);
 	config_rom[4] = cpu_to_be32(card->guid);
 
 	/* Generate root directory. */
-	config_rom[6] = cpu_to_be32(0x0c0083c0); /* node capabilities */
+	config_rom[6] = cpu_to_be32(NODE_CAPABILITIES);
 	i = 7;
 	j = 7 + descriptor_count;
 
@@ -231,7 +231,7 @@ void fw_schedule_bm_work(struct fw_card *card, unsigned long delay)
 static void fw_card_bm_work(struct work_struct *work)
 {
 	struct fw_card *card = container_of(work, struct fw_card, work.work);
-	struct fw_device *root_device;
+	struct fw_device *root_device, *irm_device;
 	struct fw_node *root_node;
 	unsigned long flags;
 	int root_id, new_root_id, irm_id, local_id;
@@ -239,6 +239,7 @@ static void fw_card_bm_work(struct work_struct *work)
 	bool do_reset = false;
 	bool root_device_is_running;
 	bool root_device_is_cmc;
+	bool irm_is_1394_1995_only;
 
 	spin_lock_irqsave(&card->lock, flags);
 
@@ -248,12 +249,18 @@ static void fw_card_bm_work(struct work_struct *work)
 	}
 
 	generation = card->generation;
+
 	root_node = card->root_node;
 	fw_node_get(root_node);
 	root_device = root_node->data;
 	root_device_is_running = root_device &&
 			atomic_read(&root_device->state) == FW_DEVICE_RUNNING;
 	root_device_is_cmc = root_device && root_device->cmc;
+
+	irm_device = card->irm_node->data;
+	irm_is_1394_1995_only = irm_device && irm_device->config_rom &&
+			(irm_device->config_rom[2] & 0x000000f0) == 0;
+
 	root_id  = root_node->node_id;
 	irm_id   = card->irm_node->node_id;
 	local_id = card->local_node->node_id;
@@ -276,8 +283,15 @@ static void fw_card_bm_work(struct work_struct *work)
 
 		if (!card->irm_node->link_on) {
 			new_root_id = local_id;
-			fw_notify("IRM has link off, making local node (%02x) root.\n",
-				  new_root_id);
+			fw_notify("%s, making local node (%02x) root.\n",
+				  "IRM has link off", new_root_id);
+			goto pick_me;
+		}
+
+		if (irm_is_1394_1995_only) {
+			new_root_id = local_id;
+			fw_notify("%s, making local node (%02x) root.\n",
+				  "IRM is not 1394a compliant", new_root_id);
 			goto pick_me;
 		}
 
@@ -316,8 +330,8 @@ static void fw_card_bm_work(struct work_struct *work)
 			 * root, and thus, IRM.
 			 */
 			new_root_id = local_id;
-			fw_notify("BM lock failed, making local node (%02x) root.\n",
-				  new_root_id);
+			fw_notify("%s, making local node (%02x) root.\n",
+				  "BM lock failed", new_root_id);
 			goto pick_me;
 		}
 	} else if (card->bm_generation != generation) {
@@ -407,13 +421,6 @@ static void fw_card_bm_work(struct work_struct *work)
 	fw_card_put(card);
 }
 
-static void flush_timer_callback(unsigned long data)
-{
-	struct fw_card *card = (struct fw_card *)data;
-
-	fw_flush_transactions(card);
-}
-
 void fw_card_initialize(struct fw_card *card,
 			const struct fw_card_driver *driver,
 			struct device *device)
@@ -432,8 +439,6 @@ void fw_card_initialize(struct fw_card *card,
 	init_completion(&card->done);
 	INIT_LIST_HEAD(&card->transaction_list);
 	spin_lock_init(&card->lock);
-	setup_timer(&card->flush_timer,
-		    flush_timer_callback, (unsigned long)card);
 
 	card->local_node = NULL;
 
@@ -558,7 +563,6 @@ void fw_core_remove_card(struct fw_card *card)
 	wait_for_completion(&card->done);
 
 	WARN_ON(!list_empty(&card->transaction_list));
-	del_timer_sync(&card->flush_timer);
 }
 EXPORT_SYMBOL(fw_core_remove_card);
 
diff --git a/drivers/firewire/core-cdev.c b/drivers/firewire/core-cdev.c
index 14a34d99eea2..5bf106b9d791 100644
--- a/drivers/firewire/core-cdev.c
+++ b/drivers/firewire/core-cdev.c
@@ -227,7 +227,7 @@ static int fw_device_op_open(struct inode *inode, struct file *file)
 	list_add_tail(&client->link, &device->client_list);
 	mutex_unlock(&device->client_list_mutex);
 
-	return 0;
+	return nonseekable_open(inode, file);
 }
 
 static void queue_event(struct client *client, struct event *event,
@@ -1496,13 +1496,13 @@ static unsigned int fw_device_op_poll(struct file *file, poll_table * pt)
 
 const struct file_operations fw_device_ops = {
 	.owner		= THIS_MODULE,
+	.llseek		= no_llseek,
 	.open		= fw_device_op_open,
 	.read		= fw_device_op_read,
 	.unlocked_ioctl	= fw_device_op_ioctl,
-	.poll		= fw_device_op_poll,
-	.release	= fw_device_op_release,
 	.mmap		= fw_device_op_mmap,
-
+	.release	= fw_device_op_release,
+	.poll		= fw_device_op_poll,
 #ifdef CONFIG_COMPAT
 	.compat_ioctl	= fw_device_op_compat_ioctl,
 #endif
diff --git a/drivers/firewire/core-transaction.c b/drivers/firewire/core-transaction.c
index 673b03f8b4ec..fdc33ff06dc1 100644
--- a/drivers/firewire/core-transaction.c
+++ b/drivers/firewire/core-transaction.c
@@ -81,7 +81,7 @@ static int close_transaction(struct fw_transaction *transaction,
 	spin_lock_irqsave(&card->lock, flags);
 	list_for_each_entry(t, &card->transaction_list, link) {
 		if (t == transaction) {
-			list_del(&t->link);
+			list_del_init(&t->link);
 			card->tlabel_mask &= ~(1ULL << t->tlabel);
 			break;
 		}
@@ -89,6 +89,7 @@ static int close_transaction(struct fw_transaction *transaction,
 	spin_unlock_irqrestore(&card->lock, flags);
 
 	if (&t->link != &card->transaction_list) {
+		del_timer_sync(&t->split_timeout_timer);
 		t->callback(card, rcode, NULL, 0, t->callback_data);
 		return 0;
 	}
@@ -121,6 +122,31 @@ int fw_cancel_transaction(struct fw_card *card,
 }
 EXPORT_SYMBOL(fw_cancel_transaction);
 
+static void split_transaction_timeout_callback(unsigned long data)
+{
+	struct fw_transaction *t = (struct fw_transaction *)data;
+	struct fw_card *card = t->card;
+	unsigned long flags;
+
+	spin_lock_irqsave(&card->lock, flags);
+	if (list_empty(&t->link)) {
+		spin_unlock_irqrestore(&card->lock, flags);
+		return;
+	}
+	list_del(&t->link);
+	card->tlabel_mask &= ~(1ULL << t->tlabel);
+	spin_unlock_irqrestore(&card->lock, flags);
+
+	card->driver->cancel_packet(card, &t->packet);
+
+	/*
+	 * At this point cancel_packet will never call the transaction
+	 * callback, since we just took the transaction out of the list.
+	 * So do it here.
+	 */
+	t->callback(card, RCODE_CANCELLED, NULL, 0, t->callback_data);
+}
+
 static void transmit_complete_callback(struct fw_packet *packet,
 				       struct fw_card *card, int status)
 {
@@ -229,6 +255,23 @@ static void fw_fill_request(struct fw_packet *packet, int tcode, int tlabel,
 	packet->payload_mapped = false;
 }
 
+static int allocate_tlabel(struct fw_card *card)
+{
+	int tlabel;
+
+	tlabel = card->current_tlabel;
+	while (card->tlabel_mask & (1ULL << tlabel)) {
+		tlabel = (tlabel + 1) & 0x3f;
+		if (tlabel == card->current_tlabel)
+			return -EBUSY;
+	}
+
+	card->current_tlabel = (tlabel + 1) & 0x3f;
+	card->tlabel_mask |= 1ULL << tlabel;
+
+	return tlabel;
+}
+
 /**
  * This function provides low-level access to the IEEE1394 transaction
  * logic.  Most C programs would use either fw_read(), fw_write() or
@@ -277,31 +320,26 @@ void fw_send_request(struct fw_card *card, struct fw_transaction *t, int tcode,
 	int tlabel;
 
 	/*
-	 * Bump the flush timer up 100ms first of all so we
-	 * don't race with a flush timer callback.
-	 */
-
-	mod_timer(&card->flush_timer, jiffies + DIV_ROUND_UP(HZ, 10));
-
-	/*
 	 * Allocate tlabel from the bitmap and put the transaction on
 	 * the list while holding the card spinlock.
 	 */
 
 	spin_lock_irqsave(&card->lock, flags);
 
-	tlabel = card->current_tlabel;
-	if (card->tlabel_mask & (1ULL << tlabel)) {
+	tlabel = allocate_tlabel(card);
+	if (tlabel < 0) {
 		spin_unlock_irqrestore(&card->lock, flags);
 		callback(card, RCODE_SEND_ERROR, NULL, 0, callback_data);
 		return;
 	}
 
-	card->current_tlabel = (card->current_tlabel + 1) & 0x3f;
-	card->tlabel_mask |= (1ULL << tlabel);
-
 	t->node_id = destination_id;
 	t->tlabel = tlabel;
+	t->card = card;
+	setup_timer(&t->split_timeout_timer,
+		    split_transaction_timeout_callback, (unsigned long)t);
+	/* FIXME: start this timer later, relative to t->timestamp */
+	mod_timer(&t->split_timeout_timer, jiffies + DIV_ROUND_UP(HZ, 10));
 	t->callback = callback;
 	t->callback_data = callback_data;
 
@@ -347,11 +385,13 @@ int fw_run_transaction(struct fw_card *card, int tcode, int destination_id,
 	struct transaction_callback_data d;
 	struct fw_transaction t;
 
+	init_timer_on_stack(&t.split_timeout_timer);
 	init_completion(&d.done);
 	d.payload = payload;
 	fw_send_request(card, &t, tcode, destination_id, generation, speed,
 			offset, payload, length, transaction_callback, &d);
 	wait_for_completion(&d.done);
+	destroy_timer_on_stack(&t.split_timeout_timer);
 
 	return d.rcode;
 }
@@ -394,30 +434,6 @@ void fw_send_phy_config(struct fw_card *card,
 	mutex_unlock(&phy_config_mutex);
 }
 
-void fw_flush_transactions(struct fw_card *card)
-{
-	struct fw_transaction *t, *next;
-	struct list_head list;
-	unsigned long flags;
-
-	INIT_LIST_HEAD(&list);
-	spin_lock_irqsave(&card->lock, flags);
-	list_splice_init(&card->transaction_list, &list);
-	card->tlabel_mask = 0;
-	spin_unlock_irqrestore(&card->lock, flags);
-
-	list_for_each_entry_safe(t, next, &list, link) {
-		card->driver->cancel_packet(card, &t->packet);
-
-		/*
-		 * At this point cancel_packet will never call the
-		 * transaction callback, since we just took all the
-		 * transactions out of the list.  So do it here.
-		 */
-		t->callback(card, RCODE_CANCELLED, NULL, 0, t->callback_data);
-	}
-}
-
 static struct fw_address_handler *lookup_overlapping_address_handler(
 	struct list_head *list, unsigned long long offset, size_t length)
 {
@@ -827,8 +843,8 @@ void fw_core_handle_response(struct fw_card *card, struct fw_packet *p)
 	spin_lock_irqsave(&card->lock, flags);
 	list_for_each_entry(t, &card->transaction_list, link) {
 		if (t->node_id == source && t->tlabel == tlabel) {
-			list_del(&t->link);
-			card->tlabel_mask &= ~(1 << t->tlabel);
+			list_del_init(&t->link);
+			card->tlabel_mask &= ~(1ULL << t->tlabel);
 			break;
 		}
 	}
@@ -869,6 +885,8 @@ void fw_core_handle_response(struct fw_card *card, struct fw_packet *p)
 		break;
 	}
 
+	del_timer_sync(&t->split_timeout_timer);
+
 	/*
 	 * The response handler may be executed while the request handler
 	 * is still pending.  Cancel the request handler.
diff --git a/drivers/firewire/core.h b/drivers/firewire/core.h
index fb0321300cce..0ecfcd95f4c5 100644
--- a/drivers/firewire/core.h
+++ b/drivers/firewire/core.h
@@ -27,7 +27,12 @@ struct fw_packet;
 #define PHY_LINK_ACTIVE		0x80
 #define PHY_CONTENDER		0x40
 #define PHY_BUS_RESET		0x40
+#define PHY_EXTENDED_REGISTERS	0xe0
 #define PHY_BUS_SHORT_RESET	0x40
+#define PHY_INT_STATUS_BITS	0x3c
+#define PHY_ENABLE_ACCEL	0x02
+#define PHY_ENABLE_MULTI	0x01
+#define PHY_PAGE_SELECT		0xe0
 
 #define BANDWIDTH_AVAILABLE_INITIAL	4915
 #define BROADCAST_CHANNEL_INITIAL	(1 << 31 | 31)
@@ -215,7 +220,6 @@ void fw_core_handle_request(struct fw_card *card, struct fw_packet *request);
 void fw_core_handle_response(struct fw_card *card, struct fw_packet *packet);
 void fw_fill_response(struct fw_packet *response, u32 *request_header,
 		      int rcode, void *payload, size_t length);
-void fw_flush_transactions(struct fw_card *card);
 void fw_send_phy_config(struct fw_card *card,
 			int node_id, int generation, int gap_count);
 
diff --git a/drivers/firewire/ohci.c b/drivers/firewire/ohci.c
index a3b083a7403a..9f627e758cfc 100644
--- a/drivers/firewire/ohci.c
+++ b/drivers/firewire/ohci.c
@@ -236,13 +236,15 @@ static char ohci_driver_name[] = KBUILD_MODNAME;
 #define QUIRK_CYCLE_TIMER		1
 #define QUIRK_RESET_PACKET		2
 #define QUIRK_BE_HEADERS		4
+#define QUIRK_NO_1394A			8
 
 /* In case of multiple matches in ohci_quirks[], only the first one is used. */
 static const struct {
 	unsigned short vendor, device, flags;
 } ohci_quirks[] = {
 	{PCI_VENDOR_ID_TI,	PCI_DEVICE_ID_TI_TSB12LV22, QUIRK_CYCLE_TIMER |
-							    QUIRK_RESET_PACKET},
+							    QUIRK_RESET_PACKET |
+							    QUIRK_NO_1394A},
 	{PCI_VENDOR_ID_TI,	PCI_ANY_ID,	QUIRK_RESET_PACKET},
 	{PCI_VENDOR_ID_AL,	PCI_ANY_ID,	QUIRK_CYCLE_TIMER},
 	{PCI_VENDOR_ID_NEC,	PCI_ANY_ID,	QUIRK_CYCLE_TIMER},
@@ -257,15 +259,16 @@ MODULE_PARM_DESC(quirks, "Chip quirks (default = 0"
 	", nonatomic cycle timer = "	__stringify(QUIRK_CYCLE_TIMER)
 	", reset packet generation = "	__stringify(QUIRK_RESET_PACKET)
 	", AR/selfID endianess = "	__stringify(QUIRK_BE_HEADERS)
+	", no 1394a enhancements = "	__stringify(QUIRK_NO_1394A)
 	")");
 
-#ifdef CONFIG_FIREWIRE_OHCI_DEBUG
-
 #define OHCI_PARAM_DEBUG_AT_AR		1
 #define OHCI_PARAM_DEBUG_SELFIDS	2
 #define OHCI_PARAM_DEBUG_IRQS		4
 #define OHCI_PARAM_DEBUG_BUSRESETS	8 /* only effective before chip init */
 
+#ifdef CONFIG_FIREWIRE_OHCI_DEBUG
+
 static int param_debug;
 module_param_named(debug, param_debug, int, 0644);
 MODULE_PARM_DESC(debug, "Verbose logging (default = 0"
@@ -438,9 +441,10 @@ static void log_ar_at_event(char dir, int speed, u32 *header, int evt)
 
 #else
 
-#define log_irqs(evt)
-#define log_selfids(node_id, generation, self_id_count, sid)
-#define log_ar_at_event(dir, speed, header, evt)
+#define param_debug 0
+static inline void log_irqs(u32 evt) {}
+static inline void log_selfids(int node_id, int generation, int self_id_count, u32 *s) {}
+static inline void log_ar_at_event(char dir, int speed, u32 *header, int evt) {}
 
 #endif /* CONFIG_FIREWIRE_OHCI_DEBUG */
 
@@ -460,27 +464,71 @@ static inline void flush_writes(const struct fw_ohci *ohci)
 	reg_read(ohci, OHCI1394_Version);
 }
 
-static int ohci_update_phy_reg(struct fw_card *card, int addr,
-			       int clear_bits, int set_bits)
+static int read_phy_reg(struct fw_ohci *ohci, int addr)
 {
-	struct fw_ohci *ohci = fw_ohci(card);
-	u32 val, old;
+	u32 val;
+	int i;
 
 	reg_write(ohci, OHCI1394_PhyControl, OHCI1394_PhyControl_Read(addr));
-	flush_writes(ohci);
-	msleep(2);
-	val = reg_read(ohci, OHCI1394_PhyControl);
-	if ((val & OHCI1394_PhyControl_ReadDone) == 0) {
-		fw_error("failed to set phy reg bits.\n");
-		return -EBUSY;
+	for (i = 0; i < 10; i++) {
+		val = reg_read(ohci, OHCI1394_PhyControl);
+		if (val & OHCI1394_PhyControl_ReadDone)
+			return OHCI1394_PhyControl_ReadData(val);
+
+		msleep(1);
 	}
+	fw_error("failed to read phy reg\n");
+
+	return -EBUSY;
+}
+
+static int write_phy_reg(const struct fw_ohci *ohci, int addr, u32 val)
+{
+	int i;
 
-	old = OHCI1394_PhyControl_ReadData(val);
-	old = (old & ~clear_bits) | set_bits;
 	reg_write(ohci, OHCI1394_PhyControl,
-		  OHCI1394_PhyControl_Write(addr, old));
+		  OHCI1394_PhyControl_Write(addr, val));
+	for (i = 0; i < 100; i++) {
+		val = reg_read(ohci, OHCI1394_PhyControl);
+		if (!(val & OHCI1394_PhyControl_WritePending))
+			return 0;
 
-	return 0;
+		msleep(1);
+	}
+	fw_error("failed to write phy reg\n");
+
+	return -EBUSY;
+}
+
+static int ohci_update_phy_reg(struct fw_card *card, int addr,
+			       int clear_bits, int set_bits)
+{
+	struct fw_ohci *ohci = fw_ohci(card);
+	int ret;
+
+	ret = read_phy_reg(ohci, addr);
+	if (ret < 0)
+		return ret;
+
+	/*
+	 * The interrupt status bits are cleared by writing a one bit.
+	 * Avoid clearing them unless explicitly requested in set_bits.
+	 */
+	if (addr == 5)
+		clear_bits |= PHY_INT_STATUS_BITS;
+
+	return write_phy_reg(ohci, addr, (ret & ~clear_bits) | set_bits);
+}
+
+static int read_paged_phy_reg(struct fw_ohci *ohci, int page, int addr)
+{
+	int ret;
+
+	ret = ohci_update_phy_reg(&ohci->card, 7, PHY_PAGE_SELECT, page << 5);
+	if (ret < 0)
+		return ret;
+
+	return read_phy_reg(ohci, addr);
 }
 
 static int ar_context_add_page(struct ar_context *ctx)
@@ -1495,13 +1543,64 @@ static void copy_config_rom(__be32 *dest, const __be32 *src, size_t length)
 		memset(&dest[length], 0, CONFIG_ROM_SIZE - size);
 }
 
+static int configure_1394a_enhancements(struct fw_ohci *ohci)
+{
+	bool enable_1394a;
+	int ret, clear, set, offset;
+
+	/* Check if the driver should configure link and PHY. */
+	if (!(reg_read(ohci, OHCI1394_HCControlSet) &
+	      OHCI1394_HCControl_programPhyEnable))
+		return 0;
+
+	/* Paranoia: check whether the PHY supports 1394a, too. */
+	enable_1394a = false;
+	ret = read_phy_reg(ohci, 2);
+	if (ret < 0)
+		return ret;
+	if ((ret & PHY_EXTENDED_REGISTERS) == PHY_EXTENDED_REGISTERS) {
+		ret = read_paged_phy_reg(ohci, 1, 8);
+		if (ret < 0)
+			return ret;
+		if (ret >= 1)
+			enable_1394a = true;
+	}
+
+	if (ohci->quirks & QUIRK_NO_1394A)
+		enable_1394a = false;
+
+	/* Configure PHY and link consistently. */
+	if (enable_1394a) {
+		clear = 0;
+		set = PHY_ENABLE_ACCEL | PHY_ENABLE_MULTI;
+	} else {
+		clear = PHY_ENABLE_ACCEL | PHY_ENABLE_MULTI;
+		set = 0;
+	}
+	ret = ohci_update_phy_reg(&ohci->card, 5, clear, set);
+	if (ret < 0)
+		return ret;
+
+	if (enable_1394a)
+		offset = OHCI1394_HCControlSet;
+	else
+		offset = OHCI1394_HCControlClear;
+	reg_write(ohci, offset, OHCI1394_HCControl_aPhyEnhanceEnable);
+
+	/* Clean up: configuration has been taken care of. */
+	reg_write(ohci, OHCI1394_HCControlClear,
+		  OHCI1394_HCControl_programPhyEnable);
+
+	return 0;
+}
+
 static int ohci_enable(struct fw_card *card,
 		       const __be32 *config_rom, size_t length)
 {
 	struct fw_ohci *ohci = fw_ohci(card);
 	struct pci_dev *dev = to_pci_dev(card->device);
 	u32 lps;
-	int i;
+	int i, ret;
 
 	if (software_reset(ohci)) {
 		fw_error("Failed to reset ohci card.\n");
@@ -1565,10 +1664,14 @@ static int ohci_enable(struct fw_card *card,
 	if (param_debug & OHCI_PARAM_DEBUG_BUSRESETS)
 		reg_write(ohci, OHCI1394_IntMaskSet, OHCI1394_busReset);
 
+	ret = configure_1394a_enhancements(ohci);
+	if (ret < 0)
+		return ret;
+
 	/* Activate link_on bit and contender bit in our self ID packets.*/
-	if (ohci_update_phy_reg(card, 4, 0,
-				PHY_LINK_ACTIVE | PHY_CONTENDER) < 0)
-		return -EIO;
+	ret = ohci_update_phy_reg(card, 4, 0, PHY_LINK_ACTIVE | PHY_CONTENDER);
+	if (ret < 0)
+		return ret;
 
 	/*
 	 * When the link is not yet enabled, the atomic config rom
@@ -2304,7 +2407,7 @@ static const struct fw_card_driver ohci_driver = {
 };
 
 #ifdef CONFIG_PPC_PMAC
-static void ohci_pmac_on(struct pci_dev *dev)
+static void pmac_ohci_on(struct pci_dev *dev)
 {
 	if (machine_is(powermac)) {
 		struct device_node *ofn = pci_device_to_OF_node(dev);
@@ -2316,7 +2419,7 @@ static void ohci_pmac_on(struct pci_dev *dev)
 	}
 }
 
-static void ohci_pmac_off(struct pci_dev *dev)
+static void pmac_ohci_off(struct pci_dev *dev)
 {
 	if (machine_is(powermac)) {
 		struct device_node *ofn = pci_device_to_OF_node(dev);
@@ -2328,15 +2431,15 @@ static void ohci_pmac_off(struct pci_dev *dev)
 	}
 }
 #else
-#define ohci_pmac_on(dev)
-#define ohci_pmac_off(dev)
+static inline void pmac_ohci_on(struct pci_dev *dev) {}
+static inline void pmac_ohci_off(struct pci_dev *dev) {}
 #endif /* CONFIG_PPC_PMAC */
 
 static int __devinit pci_probe(struct pci_dev *dev,
 			       const struct pci_device_id *ent)
 {
 	struct fw_ohci *ohci;
-	u32 bus_options, max_receive, link_speed, version;
+	u32 bus_options, max_receive, link_speed, version, link_enh;
 	u64 guid;
 	int i, err, n_ir, n_it;
 	size_t size;
@@ -2349,7 +2452,7 @@ static int __devinit pci_probe(struct pci_dev *dev,
 
 	fw_card_initialize(&ohci->card, &ohci_driver, &dev->dev);
 
-	ohci_pmac_on(dev);
+	pmac_ohci_on(dev);
 
 	err = pci_enable_device(dev);
 	if (err) {
@@ -2389,6 +2492,23 @@ static int __devinit pci_probe(struct pci_dev *dev,
 	if (param_quirks)
 		ohci->quirks = param_quirks;
 
+	/* TI OHCI-Lynx and compatible: set recommended configuration bits. */
+	if (dev->vendor == PCI_VENDOR_ID_TI) {
+		pci_read_config_dword(dev, PCI_CFG_TI_LinkEnh, &link_enh);
+
+		/* adjust latency of ATx FIFO: use 1.7 KB threshold */
+		link_enh &= ~TI_LinkEnh_atx_thresh_mask;
+		link_enh |= TI_LinkEnh_atx_thresh_1_7K;
+
+		/* use priority arbitration for asynchronous responses */
+		link_enh |= TI_LinkEnh_enab_unfair;
+
+		/* required for aPhyEnhanceEnable to work */
+		link_enh |= TI_LinkEnh_enab_accel;
+
+		pci_write_config_dword(dev, PCI_CFG_TI_LinkEnh, link_enh);
+	}
+
 	ar_context_init(&ohci->ar_request_ctx, ohci,
 			OHCI1394_AsReqRcvContextControlSet);
 
@@ -2466,7 +2586,7 @@ static int __devinit pci_probe(struct pci_dev *dev,
 	pci_disable_device(dev);
  fail_free:
 	kfree(&ohci->card);
-	ohci_pmac_off(dev);
+	pmac_ohci_off(dev);
  fail:
 	if (err == -ENOMEM)
 		fw_error("Out of memory\n");
@@ -2509,7 +2629,7 @@ static void pci_remove(struct pci_dev *dev)
 	pci_release_region(dev, 0);
 	pci_disable_device(dev);
 	kfree(&ohci->card);
-	ohci_pmac_off(dev);
+	pmac_ohci_off(dev);
 
 	fw_notify("Removed fw-ohci device.\n");
 }
@@ -2530,7 +2650,7 @@ static int pci_suspend(struct pci_dev *dev, pm_message_t state)
 	err = pci_set_power_state(dev, pci_choose_state(dev, state));
 	if (err)
 		fw_error("pci_set_power_state failed with %d\n", err);
-	ohci_pmac_off(dev);
+	pmac_ohci_off(dev);
 
 	return 0;
 }
@@ -2540,7 +2660,7 @@ static int pci_resume(struct pci_dev *dev)
 	struct fw_ohci *ohci = pci_get_drvdata(dev);
 	int err;
 
-	ohci_pmac_on(dev);
+	pmac_ohci_on(dev);
 	pci_set_power_state(dev, PCI_D0);
 	pci_restore_state(dev);
 	err = pci_enable_device(dev);
diff --git a/drivers/firewire/ohci.h b/drivers/firewire/ohci.h
index ba492d85c516..3bc9a5d744eb 100644
--- a/drivers/firewire/ohci.h
+++ b/drivers/firewire/ohci.h
@@ -67,7 +67,7 @@
 #define   OHCI1394_PhyControl_ReadDone		0x80000000
 #define   OHCI1394_PhyControl_ReadData(r)	(((r) & 0x00ff0000) >> 16)
 #define   OHCI1394_PhyControl_Write(addr, data)	(((addr) << 8) | (data) | 0x00004000)
-#define   OHCI1394_PhyControl_WriteDone		0x00004000
+#define   OHCI1394_PhyControl_WritePending	0x00004000
 #define OHCI1394_IsochronousCycleTimer        0x0F0
 #define OHCI1394_AsReqFilterHiSet             0x100
 #define OHCI1394_AsReqFilterHiClear           0x104
@@ -154,4 +154,12 @@
 
 #define OHCI1394_phy_tcode		0xe
 
+/* TI extensions */
+
+#define PCI_CFG_TI_LinkEnh		0xf4
+#define  TI_LinkEnh_enab_accel		0x00000002
+#define  TI_LinkEnh_enab_unfair		0x00000080
+#define  TI_LinkEnh_atx_thresh_mask	0x00003000
+#define  TI_LinkEnh_atx_thresh_1_7K	0x00001000
+
 #endif /* _FIREWIRE_OHCI_H */
diff --git a/drivers/gpio/Kconfig b/drivers/gpio/Kconfig
index fee678f74a19..724038dab4ca 100644
--- a/drivers/gpio/Kconfig
+++ b/drivers/gpio/Kconfig
@@ -139,6 +139,13 @@ config GPIO_MAX732X
 	  Board setup code must specify the model to use, and the start
 	  number for these GPIOs.
 
+config GPIO_MAX732X_IRQ
+	bool "Interrupt controller support for MAX732x"
+	depends on GPIO_MAX732X=y && GENERIC_HARDIRQS
+	help
+	  Say yes here to enable the max732x to be used as an interrupt
+	  controller. It requires the driver to be built in the kernel.
+
 config GPIO_PCA953X
 	tristate "PCA953x, PCA955x, TCA64xx, and MAX7310 I/O ports"
 	depends on I2C
@@ -188,6 +195,13 @@ config GPIO_PCF857X
 	  This driver provides an in-kernel interface to those GPIOs using
 	  platform-neutral GPIO calls.
 
+config GPIO_TC35892
+	bool "TC35892 GPIOs"
+	depends on MFD_TC35892
+	help
+	  This enables support for the GPIOs found on the TC35892
+	  I/O Expander.
+
 config GPIO_TWL4030
 	tristate "TWL4030, TWL5030, and TPS659x0 GPIOs"
 	depends on TWL4030_CORE
@@ -264,10 +278,10 @@ config GPIO_BT8XX
 	  If unsure, say N.
 
 config GPIO_LANGWELL
-	bool "Intel Moorestown Platform Langwell GPIO support"
+	bool "Intel Langwell/Penwell GPIO support"
 	depends on PCI
 	help
-	  Say Y here to support Intel Moorestown platform GPIO.
+	  Say Y here to support Intel Langwell/Penwell GPIO.
 
 config GPIO_TIMBERDALE
 	bool "Support for timberdale GPIO IP"
@@ -275,6 +289,15 @@ config GPIO_TIMBERDALE
 	---help---
 	Add support for the GPIO IP in the timberdale FPGA.
 
+config GPIO_RDC321X
+	tristate "RDC R-321x GPIO support"
+	depends on PCI && GPIOLIB
+	select MFD_CORE
+	select MFD_RDC321X
+	help
+	  Support for the RDC R321x SoC GPIOs over southbridge
+	  PCI configuration space.
+
 comment "SPI GPIO expanders:"
 
 config GPIO_MAX7301
@@ -310,4 +333,14 @@ config GPIO_UCB1400
 	  To compile this driver as a module, choose M here: the
 	  module will be called ucb1400_gpio.
 
+comment "MODULbus GPIO expanders:"
+
+config GPIO_JANZ_TTL
+	tristate "Janz VMOD-TTL Digital IO Module"
+	depends on MFD_JANZ_CMODIO
+	help
+	  This enables support for the Janz VMOD-TTL Digital IO module.
+	  This driver provides support for driving the pins in output
+	  mode only. Input mode is not supported.
+
 endif
diff --git a/drivers/gpio/Makefile b/drivers/gpio/Makefile
index 10f3f8d958b1..51c3cdd41b5a 100644
--- a/drivers/gpio/Makefile
+++ b/drivers/gpio/Makefile
@@ -16,6 +16,7 @@ obj-$(CONFIG_GPIO_MCP23S08)	+= mcp23s08.o
 obj-$(CONFIG_GPIO_PCA953X)	+= pca953x.o
 obj-$(CONFIG_GPIO_PCF857X)	+= pcf857x.o
 obj-$(CONFIG_GPIO_PL061)	+= pl061.o
+obj-$(CONFIG_GPIO_TC35892)	+= tc35892-gpio.o
 obj-$(CONFIG_GPIO_TIMBERDALE)	+= timbgpio.o
 obj-$(CONFIG_GPIO_TWL4030)	+= twl4030-gpio.o
 obj-$(CONFIG_GPIO_UCB1400)	+= ucb1400_gpio.o
@@ -27,4 +28,6 @@ obj-$(CONFIG_GPIO_VR41XX)	+= vr41xx_giu.o
 obj-$(CONFIG_GPIO_WM831X)	+= wm831x-gpio.o
 obj-$(CONFIG_GPIO_WM8350)	+= wm8350-gpiolib.o
 obj-$(CONFIG_GPIO_WM8994)	+= wm8994-gpio.o
-obj-$(CONFIG_GPIO_SCH)		+= sch_gpio.o
\ No newline at end of file
+obj-$(CONFIG_GPIO_SCH)		+= sch_gpio.o
+obj-$(CONFIG_GPIO_RDC321X)	+= rdc321x-gpio.o
+obj-$(CONFIG_GPIO_JANZ_TTL)	+= janz-ttl.o
diff --git a/drivers/gpio/cs5535-gpio.c b/drivers/gpio/cs5535-gpio.c
index 0c3c498f2260..f73a1555e49d 100644
--- a/drivers/gpio/cs5535-gpio.c
+++ b/drivers/gpio/cs5535-gpio.c
@@ -197,7 +197,7 @@ static int chip_direction_output(struct gpio_chip *c, unsigned offset, int val)
 	return 0;
 }
 
-static char *cs5535_gpio_names[] = {
+static const char * const cs5535_gpio_names[] = {
 	"GPIO0", "GPIO1", "GPIO2", "GPIO3",
 	"GPIO4", "GPIO5", "GPIO6", "GPIO7",
 	"GPIO8", "GPIO9", "GPIO10", "GPIO11",
diff --git a/drivers/gpio/gpiolib.c b/drivers/gpio/gpiolib.c
index cae1b8c5b08c..3ca36542e338 100644
--- a/drivers/gpio/gpiolib.c
+++ b/drivers/gpio/gpiolib.c
@@ -722,7 +722,7 @@ int gpio_export(unsigned gpio, bool direction_may_change)
 	unsigned long		flags;
 	struct gpio_desc	*desc;
 	int			status = -EINVAL;
-	char			*ioname = NULL;
+	const char		*ioname = NULL;
 
 	/* can't export until sysfs is available ... */
 	if (!gpio_class.p) {
@@ -753,7 +753,7 @@ int gpio_export(unsigned gpio, bool direction_may_change)
 		struct device	*dev;
 
 		dev = device_create(&gpio_class, desc->chip->dev, MKDEV(0, 0),
-				desc, ioname ? ioname : "gpio%d", gpio);
+				desc, ioname ? ioname : "gpio%u", gpio);
 		if (!IS_ERR(dev)) {
 			status = sysfs_create_group(&dev->kobj,
 						&gpio_attr_group);
@@ -1106,7 +1106,7 @@ unlock:
 fail:
 	/* failures here can mean systems won't boot... */
 	if (status)
-		pr_err("gpiochip_add: gpios %d..%d (%s) not registered\n",
+		pr_err("gpiochip_add: gpios %d..%d (%s) failed to register\n",
 			chip->base, chip->base + chip->ngpio - 1,
 			chip->label ? : "generic");
 	return status;
@@ -1447,6 +1447,49 @@ fail:
 }
 EXPORT_SYMBOL_GPL(gpio_direction_output);
 
+/**
+ * gpio_set_debounce - sets @debounce time for a @gpio
+ * @gpio: the gpio to set debounce time
+ * @debounce: debounce time is microseconds
+ */
+int gpio_set_debounce(unsigned gpio, unsigned debounce)
+{
+	unsigned long		flags;
+	struct gpio_chip	*chip;
+	struct gpio_desc	*desc = &gpio_desc[gpio];
+	int			status = -EINVAL;
+
+	spin_lock_irqsave(&gpio_lock, flags);
+
+	if (!gpio_is_valid(gpio))
+		goto fail;
+	chip = desc->chip;
+	if (!chip || !chip->set || !chip->set_debounce)
+		goto fail;
+	gpio -= chip->base;
+	if (gpio >= chip->ngpio)
+		goto fail;
+	status = gpio_ensure_requested(desc, gpio);
+	if (status < 0)
+		goto fail;
+
+	/* now we know the gpio is valid and chip won't vanish */
+
+	spin_unlock_irqrestore(&gpio_lock, flags);
+
+	might_sleep_if(extra_checks && chip->can_sleep);
+
+	return chip->set_debounce(chip, gpio, debounce);
+
+fail:
+	spin_unlock_irqrestore(&gpio_lock, flags);
+	if (status)
+		pr_debug("%s: gpio-%d status %d\n",
+			__func__, gpio, status);
+
+	return status;
+}
+EXPORT_SYMBOL_GPL(gpio_set_debounce);
 
 /* I/O calls are only valid after configuration completed; the relevant
  * "is this a valid GPIO" error checks should already have been done.
diff --git a/drivers/gpio/it8761e_gpio.c b/drivers/gpio/it8761e_gpio.c
index 41a9388f2fde..48fc43c4bdd1 100644
--- a/drivers/gpio/it8761e_gpio.c
+++ b/drivers/gpio/it8761e_gpio.c
@@ -217,7 +217,10 @@ gpiochip_add_err:
 static void __exit it8761e_gpio_exit(void)
 {
 	if (gpio_ba) {
-		gpiochip_remove(&it8761e_gpio_chip);
+		int ret = gpiochip_remove(&it8761e_gpio_chip);
+
+		WARN(ret, "%s(): gpiochip_remove() failed, ret=%d\n",
+				__func__, ret);
 
 		release_region(gpio_ba, GPIO_IOSIZE);
 		gpio_ba = 0;
diff --git a/drivers/gpio/janz-ttl.c b/drivers/gpio/janz-ttl.c
new file mode 100644
index 000000000000..813ac077e5d7
--- /dev/null
+++ b/drivers/gpio/janz-ttl.c
@@ -0,0 +1,258 @@
+/*
+ * Janz MODULbus VMOD-TTL GPIO Driver
+ *
+ * Copyright (c) 2010 Ira W. Snyder <iws@ovro.caltech.edu>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the
+ * Free Software Foundation; either version 2 of the License, or (at your
+ * option) any later version.
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/delay.h>
+#include <linux/platform_device.h>
+#include <linux/io.h>
+#include <linux/gpio.h>
+#include <linux/slab.h>
+
+#include <linux/mfd/janz.h>
+
+#define DRV_NAME "janz-ttl"
+
+#define PORTA_DIRECTION		0x23
+#define PORTB_DIRECTION		0x2B
+#define PORTC_DIRECTION		0x06
+#define PORTA_IOCTL		0x24
+#define PORTB_IOCTL		0x2C
+#define PORTC_IOCTL		0x07
+
+#define MASTER_INT_CTL		0x00
+#define MASTER_CONF_CTL		0x01
+
+#define CONF_PAE		(1 << 2)
+#define CONF_PBE		(1 << 7)
+#define CONF_PCE		(1 << 4)
+
+struct ttl_control_regs {
+	__be16 portc;
+	__be16 portb;
+	__be16 porta;
+	__be16 control;
+};
+
+struct ttl_module {
+	struct gpio_chip gpio;
+
+	/* base address of registers */
+	struct ttl_control_regs __iomem *regs;
+
+	u8 portc_shadow;
+	u8 portb_shadow;
+	u8 porta_shadow;
+
+	spinlock_t lock;
+};
+
+static int ttl_get_value(struct gpio_chip *gpio, unsigned offset)
+{
+	struct ttl_module *mod = dev_get_drvdata(gpio->dev);
+	u8 *shadow;
+	int ret;
+
+	if (offset < 8) {
+		shadow = &mod->porta_shadow;
+	} else if (offset < 16) {
+		shadow = &mod->portb_shadow;
+		offset -= 8;
+	} else {
+		shadow = &mod->portc_shadow;
+		offset -= 16;
+	}
+
+	spin_lock(&mod->lock);
+	ret = *shadow & (1 << offset);
+	spin_unlock(&mod->lock);
+	return ret;
+}
+
+static void ttl_set_value(struct gpio_chip *gpio, unsigned offset, int value)
+{
+	struct ttl_module *mod = dev_get_drvdata(gpio->dev);
+	void __iomem *port;
+	u8 *shadow;
+
+	if (offset < 8) {
+		port = &mod->regs->porta;
+		shadow = &mod->porta_shadow;
+	} else if (offset < 16) {
+		port = &mod->regs->portb;
+		shadow = &mod->portb_shadow;
+		offset -= 8;
+	} else {
+		port = &mod->regs->portc;
+		shadow = &mod->portc_shadow;
+		offset -= 16;
+	}
+
+	spin_lock(&mod->lock);
+	if (value)
+		*shadow |= (1 << offset);
+	else
+		*shadow &= ~(1 << offset);
+
+	iowrite16be(*shadow, port);
+	spin_unlock(&mod->lock);
+}
+
+static void __devinit ttl_write_reg(struct ttl_module *mod, u8 reg, u16 val)
+{
+	iowrite16be(reg, &mod->regs->control);
+	iowrite16be(val, &mod->regs->control);
+}
+
+static void __devinit ttl_setup_device(struct ttl_module *mod)
+{
+	/* reset the device to a known state */
+	iowrite16be(0x0000, &mod->regs->control);
+	iowrite16be(0x0001, &mod->regs->control);
+	iowrite16be(0x0000, &mod->regs->control);
+
+	/* put all ports in open-drain mode */
+	ttl_write_reg(mod, PORTA_IOCTL, 0x00ff);
+	ttl_write_reg(mod, PORTB_IOCTL, 0x00ff);
+	ttl_write_reg(mod, PORTC_IOCTL, 0x000f);
+
+	/* set all ports as outputs */
+	ttl_write_reg(mod, PORTA_DIRECTION, 0x0000);
+	ttl_write_reg(mod, PORTB_DIRECTION, 0x0000);
+	ttl_write_reg(mod, PORTC_DIRECTION, 0x0000);
+
+	/* set all ports to drive zeroes */
+	iowrite16be(0x0000, &mod->regs->porta);
+	iowrite16be(0x0000, &mod->regs->portb);
+	iowrite16be(0x0000, &mod->regs->portc);
+
+	/* enable all ports */
+	ttl_write_reg(mod, MASTER_CONF_CTL, CONF_PAE | CONF_PBE | CONF_PCE);
+}
+
+static int __devinit ttl_probe(struct platform_device *pdev)
+{
+	struct janz_platform_data *pdata;
+	struct device *dev = &pdev->dev;
+	struct ttl_module *mod;
+	struct gpio_chip *gpio;
+	struct resource *res;
+	int ret;
+
+	pdata = pdev->dev.platform_data;
+	if (!pdata) {
+		dev_err(dev, "no platform data\n");
+		ret = -ENXIO;
+		goto out_return;
+	}
+
+	mod = kzalloc(sizeof(*mod), GFP_KERNEL);
+	if (!mod) {
+		dev_err(dev, "unable to allocate private data\n");
+		ret = -ENOMEM;
+		goto out_return;
+	}
+
+	platform_set_drvdata(pdev, mod);
+	spin_lock_init(&mod->lock);
+
+	/* get access to the MODULbus registers for this module */
+	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	if (!res) {
+		dev_err(dev, "MODULbus registers not found\n");
+		ret = -ENODEV;
+		goto out_free_mod;
+	}
+
+	mod->regs = ioremap(res->start, resource_size(res));
+	if (!mod->regs) {
+		dev_err(dev, "MODULbus registers not ioremap\n");
+		ret = -ENOMEM;
+		goto out_free_mod;
+	}
+
+	ttl_setup_device(mod);
+
+	/* Initialize the GPIO data structures */
+	gpio = &mod->gpio;
+	gpio->dev = &pdev->dev;
+	gpio->label = pdev->name;
+	gpio->get = ttl_get_value;
+	gpio->set = ttl_set_value;
+	gpio->owner = THIS_MODULE;
+
+	/* request dynamic allocation */
+	gpio->base = -1;
+	gpio->ngpio = 20;
+
+	ret = gpiochip_add(gpio);
+	if (ret) {
+		dev_err(dev, "unable to add GPIO chip\n");
+		goto out_iounmap_regs;
+	}
+
+	dev_info(&pdev->dev, "module %d: registered GPIO device\n",
+			     pdata->modno);
+	return 0;
+
+out_iounmap_regs:
+	iounmap(mod->regs);
+out_free_mod:
+	kfree(mod);
+out_return:
+	return ret;
+}
+
+static int __devexit ttl_remove(struct platform_device *pdev)
+{
+	struct ttl_module *mod = platform_get_drvdata(pdev);
+	struct device *dev = &pdev->dev;
+	int ret;
+
+	ret = gpiochip_remove(&mod->gpio);
+	if (ret) {
+		dev_err(dev, "unable to remove GPIO chip\n");
+		return ret;
+	}
+
+	iounmap(mod->regs);
+	kfree(mod);
+	return 0;
+}
+
+static struct platform_driver ttl_driver = {
+	.driver		= {
+		.name	= DRV_NAME,
+		.owner	= THIS_MODULE,
+	},
+	.probe		= ttl_probe,
+	.remove		= __devexit_p(ttl_remove),
+};
+
+static int __init ttl_init(void)
+{
+	return platform_driver_register(&ttl_driver);
+}
+
+static void __exit ttl_exit(void)
+{
+	platform_driver_unregister(&ttl_driver);
+}
+
+MODULE_AUTHOR("Ira W. Snyder <iws@ovro.caltech.edu>");
+MODULE_DESCRIPTION("Janz MODULbus VMOD-TTL Driver");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS("platform:janz-ttl");
+
+module_init(ttl_init);
+module_exit(ttl_exit);
diff --git a/drivers/gpio/langwell_gpio.c b/drivers/gpio/langwell_gpio.c
index 00c3a14127af..8383a8d7f994 100644
--- a/drivers/gpio/langwell_gpio.c
+++ b/drivers/gpio/langwell_gpio.c
@@ -17,6 +17,7 @@
 
 /* Supports:
  * Moorestown platform Langwell chip.
+ * Medfield platform Penwell chip.
  */
 
 #include <linux/module.h>
@@ -31,44 +32,65 @@
 #include <linux/gpio.h>
 #include <linux/slab.h>
 
-struct lnw_gpio_register {
-	u32	GPLR[2];
-	u32	GPDR[2];
-	u32	GPSR[2];
-	u32	GPCR[2];
-	u32	GRER[2];
-	u32	GFER[2];
-	u32	GEDR[2];
+/*
+ * Langwell chip has 64 pins and thus there are 2 32bit registers to control
+ * each feature, while Penwell chip has 96 pins for each block, and need 3 32bit
+ * registers to control them, so we only define the order here instead of a
+ * structure, to get a bit offset for a pin (use GPDR as an example):
+ *
+ * nreg = ngpio / 32;
+ * reg = offset / 32;
+ * bit = offset % 32;
+ * reg_addr = reg_base + GPDR * nreg * 4 + reg * 4;
+ *
+ * so the bit of reg_addr is to control pin offset's GPDR feature
+*/
+
+enum GPIO_REG {
+	GPLR = 0,	/* pin level read-only */
+	GPDR,		/* pin direction */
+	GPSR,		/* pin set */
+	GPCR,		/* pin clear */
+	GRER,		/* rising edge detect */
+	GFER,		/* falling edge detect */
+	GEDR,		/* edge detect result */
 };
 
 struct lnw_gpio {
 	struct gpio_chip		chip;
-	struct lnw_gpio_register 	*reg_base;
+	void				*reg_base;
 	spinlock_t			lock;
 	unsigned			irq_base;
 };
 
-static int lnw_gpio_get(struct gpio_chip *chip, unsigned offset)
+static void __iomem *gpio_reg(struct gpio_chip *chip, unsigned offset,
+			enum GPIO_REG reg_type)
 {
 	struct lnw_gpio *lnw = container_of(chip, struct lnw_gpio, chip);
+	unsigned nreg = chip->ngpio / 32;
 	u8 reg = offset / 32;
-	void __iomem *gplr;
+	void __iomem *ptr;
+
+	ptr = (void __iomem *)(lnw->reg_base + reg_type * nreg * 4 + reg * 4);
+	return ptr;
+}
+
+static int lnw_gpio_get(struct gpio_chip *chip, unsigned offset)
+{
+	void __iomem *gplr = gpio_reg(chip, offset, GPLR);
 
-	gplr = (void __iomem *)(&lnw->reg_base->GPLR[reg]);
 	return readl(gplr) & BIT(offset % 32);
 }
 
 static void lnw_gpio_set(struct gpio_chip *chip, unsigned offset, int value)
 {
-	struct lnw_gpio *lnw = container_of(chip, struct lnw_gpio, chip);
-	u8 reg = offset / 32;
 	void __iomem *gpsr, *gpcr;
 
 	if (value) {
-		gpsr = (void __iomem *)(&lnw->reg_base->GPSR[reg]);
+		gpsr = gpio_reg(chip, offset, GPSR);
 		writel(BIT(offset % 32), gpsr);
 	} else {
-		gpcr = (void __iomem *)(&lnw->reg_base->GPCR[reg]);
+		gpcr = gpio_reg(chip, offset, GPCR);
 		writel(BIT(offset % 32), gpcr);
 	}
 }
@@ -76,12 +98,10 @@ static void lnw_gpio_set(struct gpio_chip *chip, unsigned offset, int value)
 static int lnw_gpio_direction_input(struct gpio_chip *chip, unsigned offset)
 {
 	struct lnw_gpio *lnw = container_of(chip, struct lnw_gpio, chip);
-	u8 reg = offset / 32;
+	void __iomem *gpdr = gpio_reg(chip, offset, GPDR);
 	u32 value;
 	unsigned long flags;
-	void __iomem *gpdr;
 
-	gpdr = (void __iomem *)(&lnw->reg_base->GPDR[reg]);
 	spin_lock_irqsave(&lnw->lock, flags);
 	value = readl(gpdr);
 	value &= ~BIT(offset % 32);
@@ -94,12 +114,10 @@ static int lnw_gpio_direction_output(struct gpio_chip *chip,
 			unsigned offset, int value)
 {
 	struct lnw_gpio *lnw = container_of(chip, struct lnw_gpio, chip);
-	u8 reg = offset / 32;
+	void __iomem *gpdr = gpio_reg(chip, offset, GPDR);
 	unsigned long flags;
-	void __iomem *gpdr;
 
 	lnw_gpio_set(chip, offset, value);
-	gpdr = (void __iomem *)(&lnw->reg_base->GPDR[reg]);
 	spin_lock_irqsave(&lnw->lock, flags);
 	value = readl(gpdr);
 	value |= BIT(offset % 32);;
@@ -118,11 +136,10 @@ static int lnw_irq_type(unsigned irq, unsigned type)
 {
 	struct lnw_gpio *lnw = get_irq_chip_data(irq);
 	u32 gpio = irq - lnw->irq_base;
-	u8 reg = gpio / 32;
 	unsigned long flags;
 	u32 value;
-	void __iomem *grer = (void __iomem *)(&lnw->reg_base->GRER[reg]);
-	void __iomem *gfer = (void __iomem *)(&lnw->reg_base->GFER[reg]);
+	void __iomem *grer = gpio_reg(&lnw->chip, gpio, GRER);
+	void __iomem *gfer = gpio_reg(&lnw->chip, gpio, GFER);
 
 	if (gpio >= lnw->chip.ngpio)
 		return -EINVAL;
@@ -158,8 +175,10 @@ static struct irq_chip lnw_irqchip = {
 	.set_type	= lnw_irq_type,
 };
 
-static struct pci_device_id lnw_gpio_ids[] = {
-	{ PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x080f) },
+static DEFINE_PCI_DEVICE_TABLE(lnw_gpio_ids) = {   /* pin number */
+	{ PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x080f), .driver_data = 64 },
+	{ PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x081f), .driver_data = 96 },
+	{ PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x081a), .driver_data = 96 },
 	{ 0, }
 };
 MODULE_DEVICE_TABLE(pci, lnw_gpio_ids);
@@ -167,17 +186,17 @@ MODULE_DEVICE_TABLE(pci, lnw_gpio_ids);
 static void lnw_irq_handler(unsigned irq, struct irq_desc *desc)
 {
 	struct lnw_gpio *lnw = (struct lnw_gpio *)get_irq_data(irq);
-	u32 reg, gpio;
+	u32 base, gpio;
 	void __iomem *gedr;
 	u32 gedr_v;
 
 	/* check GPIO controller to check which pin triggered the interrupt */
-	for (reg = 0; reg < lnw->chip.ngpio / 32; reg++) {
-		gedr = (void __iomem *)(&lnw->reg_base->GEDR[reg]);
+	for (base = 0; base < lnw->chip.ngpio; base += 32) {
+		gedr = gpio_reg(&lnw->chip, base, GEDR);
 		gedr_v = readl(gedr);
 		if (!gedr_v)
 			continue;
-		for (gpio = reg*32; gpio < reg*32+32; gpio++)
+		for (gpio = base; gpio < base + 32; gpio++)
 			if (gedr_v & BIT(gpio % 32)) {
 				pr_debug("pin %d triggered\n", gpio);
 				generic_handle_irq(lnw->irq_base + gpio);
@@ -245,7 +264,7 @@ static int __devinit lnw_gpio_probe(struct pci_dev *pdev,
 	lnw->chip.set = lnw_gpio_set;
 	lnw->chip.to_irq = lnw_gpio_to_irq;
 	lnw->chip.base = gpio_base;
-	lnw->chip.ngpio = 64;
+	lnw->chip.ngpio = id->driver_data;
 	lnw->chip.can_sleep = 0;
 	pci_set_drvdata(pdev, lnw);
 	retval = gpiochip_add(&lnw->chip);
diff --git a/drivers/gpio/max732x.c b/drivers/gpio/max732x.c
index f7868243af89..9cad60f9e962 100644
--- a/drivers/gpio/max732x.c
+++ b/drivers/gpio/max732x.c
@@ -17,7 +17,8 @@
 #include <linux/slab.h>
 #include <linux/string.h>
 #include <linux/gpio.h>
-
+#include <linux/interrupt.h>
+#include <linux/irq.h>
 #include <linux/i2c.h>
 #include <linux/i2c/max732x.h>
 
@@ -31,7 +32,8 @@
  *   - Open Drain I/O
  *
  * designated by 'O', 'I' and 'P' individually according to MAXIM's
- * datasheets.
+ * datasheets. 'I' and 'P' ports are interrupt capables, some with
+ * a dedicated interrupt mask.
  *
  * There are two groups of I/O ports, each group usually includes
  * up to 8 I/O ports, and is accessed by a specific I2C address:
@@ -44,7 +46,8 @@
  *
  * Within each group of ports, there are five known combinations of
  * I/O ports: 4I4O, 4P4O, 8I, 8P, 8O, see the definitions below for
- * the detailed organization of these ports.
+ * the detailed organization of these ports. Only Goup A is interrupt
+ * capable.
  *
  * GPIO numbers start from 'gpio_base + 0' to 'gpio_base + 8/16',
  * and GPIOs from GROUP_A are numbered before those from GROUP_B
@@ -68,16 +71,47 @@
 #define GROUP_A(x)	((x) & 0xffff)	/* I2C Addr: 0b'110xxxx */
 #define GROUP_B(x)	((x) << 16)	/* I2C Addr: 0b'101xxxx */
 
+#define INT_NONE	0x0	/* No interrupt capability */
+#define INT_NO_MASK	0x1	/* Has interrupts, no mask */
+#define INT_INDEP_MASK	0x2	/* Has interrupts, independent mask */
+#define INT_MERGED_MASK 0x3	/* Has interrupts, merged mask */
+
+#define INT_CAPS(x)	(((uint64_t)(x)) << 32)
+
+enum {
+	MAX7319,
+	MAX7320,
+	MAX7321,
+	MAX7322,
+	MAX7323,
+	MAX7324,
+	MAX7325,
+	MAX7326,
+	MAX7327,
+};
+
+static uint64_t max732x_features[] = {
+	[MAX7319] = GROUP_A(IO_8I) | INT_CAPS(INT_MERGED_MASK),
+	[MAX7320] = GROUP_B(IO_8O),
+	[MAX7321] = GROUP_A(IO_8P) | INT_CAPS(INT_NO_MASK),
+	[MAX7322] = GROUP_A(IO_4I4O) | INT_CAPS(INT_MERGED_MASK),
+	[MAX7323] = GROUP_A(IO_4P4O) | INT_CAPS(INT_INDEP_MASK),
+	[MAX7324] = GROUP_A(IO_8I) | GROUP_B(IO_8O) | INT_CAPS(INT_MERGED_MASK),
+	[MAX7325] = GROUP_A(IO_8P) | GROUP_B(IO_8O) | INT_CAPS(INT_NO_MASK),
+	[MAX7326] = GROUP_A(IO_4I4O) | GROUP_B(IO_8O) | INT_CAPS(INT_MERGED_MASK),
+	[MAX7327] = GROUP_A(IO_4P4O) | GROUP_B(IO_8O) | INT_CAPS(INT_NO_MASK),
+};
+
 static const struct i2c_device_id max732x_id[] = {
-	{ "max7319", GROUP_A(IO_8I) },
-	{ "max7320", GROUP_B(IO_8O) },
-	{ "max7321", GROUP_A(IO_8P) },
-	{ "max7322", GROUP_A(IO_4I4O) },
-	{ "max7323", GROUP_A(IO_4P4O) },
-	{ "max7324", GROUP_A(IO_8I) | GROUP_B(IO_8O) },
-	{ "max7325", GROUP_A(IO_8P) | GROUP_B(IO_8O) },
-	{ "max7326", GROUP_A(IO_4I4O) | GROUP_B(IO_8O) },
-	{ "max7327", GROUP_A(IO_4P4O) | GROUP_B(IO_8O) },
+	{ "max7319", MAX7319 },
+	{ "max7320", MAX7320 },
+	{ "max7321", MAX7321 },
+	{ "max7322", MAX7322 },
+	{ "max7323", MAX7323 },
+	{ "max7324", MAX7324 },
+	{ "max7325", MAX7325 },
+	{ "max7326", MAX7326 },
+	{ "max7327", MAX7327 },
 	{ },
 };
 MODULE_DEVICE_TABLE(i2c, max732x_id);
@@ -96,9 +130,19 @@ struct max732x_chip {
 
 	struct mutex	lock;
 	uint8_t		reg_out[2];
+
+#ifdef CONFIG_GPIO_MAX732X_IRQ
+	struct mutex	irq_lock;
+	int		irq_base;
+	uint8_t		irq_mask;
+	uint8_t		irq_mask_cur;
+	uint8_t		irq_trig_raise;
+	uint8_t		irq_trig_fall;
+	uint8_t		irq_features;
+#endif
 };
 
-static int max732x_write(struct max732x_chip *chip, int group_a, uint8_t val)
+static int max732x_writeb(struct max732x_chip *chip, int group_a, uint8_t val)
 {
 	struct i2c_client *client;
 	int ret;
@@ -113,7 +157,7 @@ static int max732x_write(struct max732x_chip *chip, int group_a, uint8_t val)
 	return 0;
 }
 
-static int max732x_read(struct max732x_chip *chip, int group_a, uint8_t *val)
+static int max732x_readb(struct max732x_chip *chip, int group_a, uint8_t *val)
 {
 	struct i2c_client *client;
 	int ret;
@@ -142,7 +186,7 @@ static int max732x_gpio_get_value(struct gpio_chip *gc, unsigned off)
 
 	chip = container_of(gc, struct max732x_chip, gpio_chip);
 
-	ret = max732x_read(chip, is_group_a(chip, off), &reg_val);
+	ret = max732x_readb(chip, is_group_a(chip, off), &reg_val);
 	if (ret < 0)
 		return 0;
 
@@ -162,7 +206,7 @@ static void max732x_gpio_set_value(struct gpio_chip *gc, unsigned off, int val)
 	reg_out = (off > 7) ? chip->reg_out[1] : chip->reg_out[0];
 	reg_out = (val) ? reg_out | mask : reg_out & ~mask;
 
-	ret = max732x_write(chip, is_group_a(chip, off), reg_out);
+	ret = max732x_writeb(chip, is_group_a(chip, off), reg_out);
 	if (ret < 0)
 		goto out;
 
@@ -188,6 +232,13 @@ static int max732x_gpio_direction_input(struct gpio_chip *gc, unsigned off)
 		return -EACCES;
 	}
 
+	/*
+	 * Open-drain pins must be set to high impedance (which is
+	 * equivalent to output-high) to be turned into an input.
+	 */
+	if ((mask & chip->dir_output))
+		max732x_gpio_set_value(gc, off, 1);
+
 	return 0;
 }
 
@@ -209,12 +260,278 @@ static int max732x_gpio_direction_output(struct gpio_chip *gc,
 	return 0;
 }
 
+#ifdef CONFIG_GPIO_MAX732X_IRQ
+static int max732x_writew(struct max732x_chip *chip, uint16_t val)
+{
+	int ret;
+
+	val = cpu_to_le16(val);
+
+	ret = i2c_master_send(chip->client_group_a, (char *)&val, 2);
+	if (ret < 0) {
+		dev_err(&chip->client_group_a->dev, "failed writing\n");
+		return ret;
+	}
+
+	return 0;
+}
+
+static int max732x_readw(struct max732x_chip *chip, uint16_t *val)
+{
+	int ret;
+
+	ret = i2c_master_recv(chip->client_group_a, (char *)val, 2);
+	if (ret < 0) {
+		dev_err(&chip->client_group_a->dev, "failed reading\n");
+		return ret;
+	}
+
+	*val = le16_to_cpu(*val);
+	return 0;
+}
+
+static void max732x_irq_update_mask(struct max732x_chip *chip)
+{
+	uint16_t msg;
+
+	if (chip->irq_mask == chip->irq_mask_cur)
+		return;
+
+	chip->irq_mask = chip->irq_mask_cur;
+
+	if (chip->irq_features == INT_NO_MASK)
+		return;
+
+	mutex_lock(&chip->lock);
+
+	switch (chip->irq_features) {
+	case INT_INDEP_MASK:
+		msg = (chip->irq_mask << 8) | chip->reg_out[0];
+		max732x_writew(chip, msg);
+		break;
+
+	case INT_MERGED_MASK:
+		msg = chip->irq_mask | chip->reg_out[0];
+		max732x_writeb(chip, 1, (uint8_t)msg);
+		break;
+	}
+
+	mutex_unlock(&chip->lock);
+}
+
+static int max732x_gpio_to_irq(struct gpio_chip *gc, unsigned off)
+{
+	struct max732x_chip *chip;
+
+	chip = container_of(gc, struct max732x_chip, gpio_chip);
+	return chip->irq_base + off;
+}
+
+static void max732x_irq_mask(unsigned int irq)
+{
+	struct max732x_chip *chip = get_irq_chip_data(irq);
+
+	chip->irq_mask_cur &= ~(1 << (irq - chip->irq_base));
+}
+
+static void max732x_irq_unmask(unsigned int irq)
+{
+	struct max732x_chip *chip = get_irq_chip_data(irq);
+
+	chip->irq_mask_cur |= 1 << (irq - chip->irq_base);
+}
+
+static void max732x_irq_bus_lock(unsigned int irq)
+{
+	struct max732x_chip *chip = get_irq_chip_data(irq);
+
+	mutex_lock(&chip->irq_lock);
+	chip->irq_mask_cur = chip->irq_mask;
+}
+
+static void max732x_irq_bus_sync_unlock(unsigned int irq)
+{
+	struct max732x_chip *chip = get_irq_chip_data(irq);
+
+	max732x_irq_update_mask(chip);
+	mutex_unlock(&chip->irq_lock);
+}
+
+static int max732x_irq_set_type(unsigned int irq, unsigned int type)
+{
+	struct max732x_chip *chip = get_irq_chip_data(irq);
+	uint16_t off = irq - chip->irq_base;
+	uint16_t mask = 1 << off;
+
+	if (!(mask & chip->dir_input)) {
+		dev_dbg(&chip->client->dev, "%s port %d is output only\n",
+			chip->client->name, off);
+		return -EACCES;
+	}
+
+	if (!(type & IRQ_TYPE_EDGE_BOTH)) {
+		dev_err(&chip->client->dev, "irq %d: unsupported type %d\n",
+			irq, type);
+		return -EINVAL;
+	}
+
+	if (type & IRQ_TYPE_EDGE_FALLING)
+		chip->irq_trig_fall |= mask;
+	else
+		chip->irq_trig_fall &= ~mask;
+
+	if (type & IRQ_TYPE_EDGE_RISING)
+		chip->irq_trig_raise |= mask;
+	else
+		chip->irq_trig_raise &= ~mask;
+
+	return max732x_gpio_direction_input(&chip->gpio_chip, off);
+}
+
+static struct irq_chip max732x_irq_chip = {
+	.name			= "max732x",
+	.mask			= max732x_irq_mask,
+	.unmask			= max732x_irq_unmask,
+	.bus_lock		= max732x_irq_bus_lock,
+	.bus_sync_unlock	= max732x_irq_bus_sync_unlock,
+	.set_type		= max732x_irq_set_type,
+};
+
+static uint8_t max732x_irq_pending(struct max732x_chip *chip)
+{
+	uint8_t cur_stat;
+	uint8_t old_stat;
+	uint8_t trigger;
+	uint8_t pending;
+	uint16_t status;
+	int ret;
+
+	ret = max732x_readw(chip, &status);
+	if (ret)
+		return 0;
+
+	trigger = status >> 8;
+	trigger &= chip->irq_mask;
+
+	if (!trigger)
+		return 0;
+
+	cur_stat = status & 0xFF;
+	cur_stat &= chip->irq_mask;
+
+	old_stat = cur_stat ^ trigger;
+
+	pending = (old_stat & chip->irq_trig_fall) |
+		  (cur_stat & chip->irq_trig_raise);
+	pending &= trigger;
+
+	return pending;
+}
+
+static irqreturn_t max732x_irq_handler(int irq, void *devid)
+{
+	struct max732x_chip *chip = devid;
+	uint8_t pending;
+	uint8_t level;
+
+	pending = max732x_irq_pending(chip);
+
+	if (!pending)
+		return IRQ_HANDLED;
+
+	do {
+		level = __ffs(pending);
+		handle_nested_irq(level + chip->irq_base);
+
+		pending &= ~(1 << level);
+	} while (pending);
+
+	return IRQ_HANDLED;
+}
+
+static int max732x_irq_setup(struct max732x_chip *chip,
+			     const struct i2c_device_id *id)
+{
+	struct i2c_client *client = chip->client;
+	struct max732x_platform_data *pdata = client->dev.platform_data;
+	int has_irq = max732x_features[id->driver_data] >> 32;
+	int ret;
+
+	if (pdata->irq_base && has_irq != INT_NONE) {
+		int lvl;
+
+		chip->irq_base = pdata->irq_base;
+		chip->irq_features = has_irq;
+		mutex_init(&chip->irq_lock);
+
+		for (lvl = 0; lvl < chip->gpio_chip.ngpio; lvl++) {
+			int irq = lvl + chip->irq_base;
+
+			if (!(chip->dir_input & (1 << lvl)))
+				continue;
+
+			set_irq_chip_data(irq, chip);
+			set_irq_chip_and_handler(irq, &max732x_irq_chip,
+						 handle_edge_irq);
+			set_irq_nested_thread(irq, 1);
+#ifdef CONFIG_ARM
+			set_irq_flags(irq, IRQF_VALID);
+#else
+			set_irq_noprobe(irq);
+#endif
+		}
+
+		ret = request_threaded_irq(client->irq,
+					   NULL,
+					   max732x_irq_handler,
+					   IRQF_TRIGGER_FALLING | IRQF_ONESHOT,
+					   dev_name(&client->dev), chip);
+		if (ret) {
+			dev_err(&client->dev, "failed to request irq %d\n",
+				client->irq);
+			goto out_failed;
+		}
+
+		chip->gpio_chip.to_irq = max732x_gpio_to_irq;
+	}
+
+	return 0;
+
+out_failed:
+	chip->irq_base = 0;
+	return ret;
+}
+
+static void max732x_irq_teardown(struct max732x_chip *chip)
+{
+	if (chip->irq_base)
+		free_irq(chip->client->irq, chip);
+}
+#else /* CONFIG_GPIO_MAX732X_IRQ */
+static int max732x_irq_setup(struct max732x_chip *chip,
+			     const struct i2c_device_id *id)
+{
+	struct i2c_client *client = chip->client;
+	struct max732x_platform_data *pdata = client->dev.platform_data;
+	int has_irq = max732x_features[id->driver_data] >> 32;
+
+	if (pdata->irq_base && has_irq != INT_NONE)
+		dev_warn(&client->dev, "interrupt support not compiled in\n");
+
+	return 0;
+}
+
+static void max732x_irq_teardown(struct max732x_chip *chip)
+{
+}
+#endif
+
 static int __devinit max732x_setup_gpio(struct max732x_chip *chip,
 					const struct i2c_device_id *id,
 					unsigned gpio_start)
 {
 	struct gpio_chip *gc = &chip->gpio_chip;
-	uint32_t id_data = id->driver_data;
+	uint32_t id_data = (uint32_t)max732x_features[id->driver_data];
 	int i, port = 0;
 
 	for (i = 0; i < 16; i++, id_data >>= 2) {
@@ -285,14 +602,14 @@ static int __devinit max732x_probe(struct i2c_client *client,
 	switch (client->addr & 0x70) {
 	case 0x60:
 		chip->client_group_a = client;
-		if (nr_port > 7) {
+		if (nr_port > 8) {
 			c = i2c_new_dummy(client->adapter, addr_b);
 			chip->client_group_b = chip->client_dummy = c;
 		}
 		break;
 	case 0x50:
 		chip->client_group_b = client;
-		if (nr_port > 7) {
+		if (nr_port > 8) {
 			c = i2c_new_dummy(client->adapter, addr_a);
 			chip->client_group_a = chip->client_dummy = c;
 		}
@@ -306,9 +623,13 @@ static int __devinit max732x_probe(struct i2c_client *client,
 
 	mutex_init(&chip->lock);
 
-	max732x_read(chip, is_group_a(chip, 0), &chip->reg_out[0]);
-	if (nr_port > 7)
-		max732x_read(chip, is_group_a(chip, 8), &chip->reg_out[1]);
+	max732x_readb(chip, is_group_a(chip, 0), &chip->reg_out[0]);
+	if (nr_port > 8)
+		max732x_readb(chip, is_group_a(chip, 8), &chip->reg_out[1]);
+
+	ret = max732x_irq_setup(chip, id);
+	if (ret)
+		goto out_failed;
 
 	ret = gpiochip_add(&chip->gpio_chip);
 	if (ret)
@@ -325,6 +646,7 @@ static int __devinit max732x_probe(struct i2c_client *client,
 	return 0;
 
 out_failed:
+	max732x_irq_teardown(chip);
 	kfree(chip);
 	return ret;
 }
@@ -352,6 +674,8 @@ static int __devexit max732x_remove(struct i2c_client *client)
 		return ret;
 	}
 
+	max732x_irq_teardown(chip);
+
 	/* unregister any dummy i2c_client */
 	if (chip->client_dummy)
 		i2c_unregister_device(chip->client_dummy);
diff --git a/drivers/gpio/pca953x.c b/drivers/gpio/pca953x.c
index b827c976dc62..a2b12aa1f2b9 100644
--- a/drivers/gpio/pca953x.c
+++ b/drivers/gpio/pca953x.c
@@ -73,7 +73,7 @@ struct pca953x_chip {
 	struct i2c_client *client;
 	struct pca953x_platform_data *dyn_pdata;
 	struct gpio_chip gpio_chip;
-	char **names;
+	const char *const *names;
 };
 
 static int pca953x_write_reg(struct pca953x_chip *chip, int reg, uint16_t val)
@@ -449,7 +449,7 @@ pca953x_get_alt_pdata(struct i2c_client *client)
 	struct device_node *node;
 	const uint16_t *val;
 
-	node = dev_archdata_get_node(&client->dev.archdata);
+	node = client->dev.of_node;
 	if (node == NULL)
 		return NULL;
 
diff --git a/drivers/gpio/pl061.c b/drivers/gpio/pl061.c
index 105701a1f05b..ee568c8fcbd0 100644
--- a/drivers/gpio/pl061.c
+++ b/drivers/gpio/pl061.c
@@ -164,7 +164,7 @@ static int pl061_irq_type(unsigned irq, unsigned trigger)
 	unsigned long flags;
 	u8 gpiois, gpioibe, gpioiev;
 
-	if (offset < 0 || offset > PL061_GPIO_NR)
+	if (offset < 0 || offset >= PL061_GPIO_NR)
 		return -EINVAL;
 
 	spin_lock_irqsave(&chip->irq_lock, flags);
diff --git a/drivers/gpio/rdc321x-gpio.c b/drivers/gpio/rdc321x-gpio.c
new file mode 100644
index 000000000000..2762698e0204
--- /dev/null
+++ b/drivers/gpio/rdc321x-gpio.c
@@ -0,0 +1,246 @@
+/*
+ * RDC321x GPIO driver
+ *
+ * Copyright (C) 2008, Volker Weiss <dev@tintuc.de>
+ * Copyright (C) 2007-2010 Florian Fainelli <florian@openwrt.org>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ *
+ */
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/spinlock.h>
+#include <linux/platform_device.h>
+#include <linux/pci.h>
+#include <linux/gpio.h>
+#include <linux/mfd/rdc321x.h>
+#include <linux/slab.h>
+
+struct rdc321x_gpio {
+	spinlock_t		lock;
+	struct pci_dev		*sb_pdev;
+	u32			data_reg[2];
+	int			reg1_ctrl_base;
+	int			reg1_data_base;
+	int			reg2_ctrl_base;
+	int			reg2_data_base;
+	struct gpio_chip	chip;
+};
+
+/* read GPIO pin */
+static int rdc_gpio_get_value(struct gpio_chip *chip, unsigned gpio)
+{
+	struct rdc321x_gpio *gpch;
+	u32 value = 0;
+	int reg;
+
+	gpch = container_of(chip, struct rdc321x_gpio, chip);
+	reg = gpio < 32 ? gpch->reg1_data_base : gpch->reg2_data_base;
+
+	spin_lock(&gpch->lock);
+	pci_write_config_dword(gpch->sb_pdev, reg,
+					gpch->data_reg[gpio < 32 ? 0 : 1]);
+	pci_read_config_dword(gpch->sb_pdev, reg, &value);
+	spin_unlock(&gpch->lock);
+
+	return (1 << (gpio & 0x1f)) & value ? 1 : 0;
+}
+
+static void rdc_gpio_set_value_impl(struct gpio_chip *chip,
+				unsigned gpio, int value)
+{
+	struct rdc321x_gpio *gpch;
+	int reg = (gpio < 32) ? 0 : 1;
+
+	gpch = container_of(chip, struct rdc321x_gpio, chip);
+
+	if (value)
+		gpch->data_reg[reg] |= 1 << (gpio & 0x1f);
+	else
+		gpch->data_reg[reg] &= ~(1 << (gpio & 0x1f));
+
+	pci_write_config_dword(gpch->sb_pdev,
+			reg ? gpch->reg2_data_base : gpch->reg1_data_base,
+			gpch->data_reg[reg]);
+}
+
+/* set GPIO pin to value */
+static void rdc_gpio_set_value(struct gpio_chip *chip,
+				unsigned gpio, int value)
+{
+	struct rdc321x_gpio *gpch;
+
+	gpch = container_of(chip, struct rdc321x_gpio, chip);
+	spin_lock(&gpch->lock);
+	rdc_gpio_set_value_impl(chip, gpio, value);
+	spin_unlock(&gpch->lock);
+}
+
+static int rdc_gpio_config(struct gpio_chip *chip,
+				unsigned gpio, int value)
+{
+	struct rdc321x_gpio *gpch;
+	int err;
+	u32 reg;
+
+	gpch = container_of(chip, struct rdc321x_gpio, chip);
+
+	spin_lock(&gpch->lock);
+	err = pci_read_config_dword(gpch->sb_pdev, gpio < 32 ?
+			gpch->reg1_ctrl_base : gpch->reg2_ctrl_base, &reg);
+	if (err)
+		goto unlock;
+
+	reg |= 1 << (gpio & 0x1f);
+
+	err = pci_write_config_dword(gpch->sb_pdev, gpio < 32 ?
+			gpch->reg1_ctrl_base : gpch->reg2_ctrl_base, reg);
+	if (err)
+		goto unlock;
+
+	rdc_gpio_set_value_impl(chip, gpio, value);
+
+unlock:
+	spin_unlock(&gpch->lock);
+
+	return err;
+}
+
+/* configure GPIO pin as input */
+static int rdc_gpio_direction_input(struct gpio_chip *chip, unsigned gpio)
+{
+	return rdc_gpio_config(chip, gpio, 1);
+}
+
+/*
+ * Cache the initial value of both GPIO data registers
+ */
+static int __devinit rdc321x_gpio_probe(struct platform_device *pdev)
+{
+	int err;
+	struct resource *r;
+	struct rdc321x_gpio *rdc321x_gpio_dev;
+	struct rdc321x_gpio_pdata *pdata;
+
+	pdata = pdev->dev.platform_data;
+	if (!pdata) {
+		dev_err(&pdev->dev, "no platform data supplied\n");
+		return -ENODEV;
+	}
+
+	rdc321x_gpio_dev = kzalloc(sizeof(struct rdc321x_gpio), GFP_KERNEL);
+	if (!rdc321x_gpio_dev) {
+		dev_err(&pdev->dev, "failed to allocate private data\n");
+		return -ENOMEM;
+	}
+
+	r = platform_get_resource_byname(pdev, IORESOURCE_IO, "gpio-reg1");
+	if (!r) {
+		dev_err(&pdev->dev, "failed to get gpio-reg1 resource\n");
+		err = -ENODEV;
+		goto out_free;
+	}
+
+	spin_lock_init(&rdc321x_gpio_dev->lock);
+	rdc321x_gpio_dev->sb_pdev = pdata->sb_pdev;
+	rdc321x_gpio_dev->reg1_ctrl_base = r->start;
+	rdc321x_gpio_dev->reg1_data_base = r->start + 0x4;
+
+	r = platform_get_resource_byname(pdev, IORESOURCE_IO, "gpio-reg2");
+	if (!r) {
+		dev_err(&pdev->dev, "failed to get gpio-reg2 resource\n");
+		err = -ENODEV;
+		goto out_free;
+	}
+
+	rdc321x_gpio_dev->reg2_ctrl_base = r->start;
+	rdc321x_gpio_dev->reg2_data_base = r->start + 0x4;
+
+	rdc321x_gpio_dev->chip.label = "rdc321x-gpio";
+	rdc321x_gpio_dev->chip.direction_input = rdc_gpio_direction_input;
+	rdc321x_gpio_dev->chip.direction_output = rdc_gpio_config;
+	rdc321x_gpio_dev->chip.get = rdc_gpio_get_value;
+	rdc321x_gpio_dev->chip.set = rdc_gpio_set_value;
+	rdc321x_gpio_dev->chip.base = 0;
+	rdc321x_gpio_dev->chip.ngpio = pdata->max_gpios;
+
+	platform_set_drvdata(pdev, rdc321x_gpio_dev);
+
+	/* This might not be, what others (BIOS, bootloader, etc.)
+	   wrote to these registers before, but it's a good guess. Still
+	   better than just using 0xffffffff. */
+	err = pci_read_config_dword(rdc321x_gpio_dev->sb_pdev,
+					rdc321x_gpio_dev->reg1_data_base,
+					&rdc321x_gpio_dev->data_reg[0]);
+	if (err)
+		goto out_drvdata;
+
+	err = pci_read_config_dword(rdc321x_gpio_dev->sb_pdev,
+					rdc321x_gpio_dev->reg2_data_base,
+					&rdc321x_gpio_dev->data_reg[1]);
+	if (err)
+		goto out_drvdata;
+
+	dev_info(&pdev->dev, "registering %d GPIOs\n",
+					rdc321x_gpio_dev->chip.ngpio);
+	return gpiochip_add(&rdc321x_gpio_dev->chip);
+
+out_drvdata:
+	platform_set_drvdata(pdev, NULL);
+out_free:
+	kfree(rdc321x_gpio_dev);
+	return err;
+}
+
+static int __devexit rdc321x_gpio_remove(struct platform_device *pdev)
+{
+	int ret;
+	struct rdc321x_gpio *rdc321x_gpio_dev = platform_get_drvdata(pdev);
+
+	ret = gpiochip_remove(&rdc321x_gpio_dev->chip);
+	if (ret)
+		dev_err(&pdev->dev, "failed to unregister chip\n");
+
+	kfree(rdc321x_gpio_dev);
+	platform_set_drvdata(pdev, NULL);
+
+	return ret;
+}
+
+static struct platform_driver rdc321x_gpio_driver = {
+	.driver.name	= "rdc321x-gpio",
+	.driver.owner	= THIS_MODULE,
+	.probe		= rdc321x_gpio_probe,
+	.remove		= __devexit_p(rdc321x_gpio_remove),
+};
+
+static int __init rdc321x_gpio_init(void)
+{
+	return platform_driver_register(&rdc321x_gpio_driver);
+}
+
+static void __exit rdc321x_gpio_exit(void)
+{
+	platform_driver_unregister(&rdc321x_gpio_driver);
+}
+
+module_init(rdc321x_gpio_init);
+module_exit(rdc321x_gpio_exit);
+
+MODULE_AUTHOR("Florian Fainelli <florian@openwrt.org>");
+MODULE_DESCRIPTION("RDC321x GPIO driver");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS("platform:rdc321x-gpio");
diff --git a/drivers/gpio/tc35892-gpio.c b/drivers/gpio/tc35892-gpio.c
new file mode 100644
index 000000000000..1be6288780de
--- /dev/null
+++ b/drivers/gpio/tc35892-gpio.c
@@ -0,0 +1,381 @@
+/*
+ * Copyright (C) ST-Ericsson SA 2010
+ *
+ * License Terms: GNU General Public License, version 2
+ * Author: Hanumath Prasad <hanumath.prasad@stericsson.com> for ST-Ericsson
+ * Author: Rabin Vincent <rabin.vincent@stericsson.com> for ST-Ericsson
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/platform_device.h>
+#include <linux/slab.h>
+#include <linux/gpio.h>
+#include <linux/irq.h>
+#include <linux/interrupt.h>
+#include <linux/mfd/tc35892.h>
+
+/*
+ * These registers are modified under the irq bus lock and cached to avoid
+ * unnecessary writes in bus_sync_unlock.
+ */
+enum { REG_IBE, REG_IEV, REG_IS, REG_IE };
+
+#define CACHE_NR_REGS	4
+#define CACHE_NR_BANKS	3
+
+struct tc35892_gpio {
+	struct gpio_chip chip;
+	struct tc35892 *tc35892;
+	struct device *dev;
+	struct mutex irq_lock;
+
+	int irq_base;
+
+	/* Caches of interrupt control registers for bus_lock */
+	u8 regs[CACHE_NR_REGS][CACHE_NR_BANKS];
+	u8 oldregs[CACHE_NR_REGS][CACHE_NR_BANKS];
+};
+
+static inline struct tc35892_gpio *to_tc35892_gpio(struct gpio_chip *chip)
+{
+	return container_of(chip, struct tc35892_gpio, chip);
+}
+
+static int tc35892_gpio_get(struct gpio_chip *chip, unsigned offset)
+{
+	struct tc35892_gpio *tc35892_gpio = to_tc35892_gpio(chip);
+	struct tc35892 *tc35892 = tc35892_gpio->tc35892;
+	u8 reg = TC35892_GPIODATA0 + (offset / 8) * 2;
+	u8 mask = 1 << (offset % 8);
+	int ret;
+
+	ret = tc35892_reg_read(tc35892, reg);
+	if (ret < 0)
+		return ret;
+
+	return ret & mask;
+}
+
+static void tc35892_gpio_set(struct gpio_chip *chip, unsigned offset, int val)
+{
+	struct tc35892_gpio *tc35892_gpio = to_tc35892_gpio(chip);
+	struct tc35892 *tc35892 = tc35892_gpio->tc35892;
+	u8 reg = TC35892_GPIODATA0 + (offset / 8) * 2;
+	unsigned pos = offset % 8;
+	u8 data[] = {!!val << pos, 1 << pos};
+
+	tc35892_block_write(tc35892, reg, ARRAY_SIZE(data), data);
+}
+
+static int tc35892_gpio_direction_output(struct gpio_chip *chip,
+					 unsigned offset, int val)
+{
+	struct tc35892_gpio *tc35892_gpio = to_tc35892_gpio(chip);
+	struct tc35892 *tc35892 = tc35892_gpio->tc35892;
+	u8 reg = TC35892_GPIODIR0 + offset / 8;
+	unsigned pos = offset % 8;
+
+	tc35892_gpio_set(chip, offset, val);
+
+	return tc35892_set_bits(tc35892, reg, 1 << pos, 1 << pos);
+}
+
+static int tc35892_gpio_direction_input(struct gpio_chip *chip,
+					unsigned offset)
+{
+	struct tc35892_gpio *tc35892_gpio = to_tc35892_gpio(chip);
+	struct tc35892 *tc35892 = tc35892_gpio->tc35892;
+	u8 reg = TC35892_GPIODIR0 + offset / 8;
+	unsigned pos = offset % 8;
+
+	return tc35892_set_bits(tc35892, reg, 1 << pos, 0);
+}
+
+static int tc35892_gpio_to_irq(struct gpio_chip *chip, unsigned offset)
+{
+	struct tc35892_gpio *tc35892_gpio = to_tc35892_gpio(chip);
+
+	return tc35892_gpio->irq_base + offset;
+}
+
+static struct gpio_chip template_chip = {
+	.label			= "tc35892",
+	.owner			= THIS_MODULE,
+	.direction_input	= tc35892_gpio_direction_input,
+	.get			= tc35892_gpio_get,
+	.direction_output	= tc35892_gpio_direction_output,
+	.set			= tc35892_gpio_set,
+	.to_irq			= tc35892_gpio_to_irq,
+	.can_sleep		= 1,
+};
+
+static int tc35892_gpio_irq_set_type(unsigned int irq, unsigned int type)
+{
+	struct tc35892_gpio *tc35892_gpio = get_irq_chip_data(irq);
+	int offset = irq - tc35892_gpio->irq_base;
+	int regoffset = offset / 8;
+	int mask = 1 << (offset % 8);
+
+	if (type == IRQ_TYPE_EDGE_BOTH) {
+		tc35892_gpio->regs[REG_IBE][regoffset] |= mask;
+		return 0;
+	}
+
+	tc35892_gpio->regs[REG_IBE][regoffset] &= ~mask;
+
+	if (type == IRQ_TYPE_LEVEL_LOW || type == IRQ_TYPE_LEVEL_HIGH)
+		tc35892_gpio->regs[REG_IS][regoffset] |= mask;
+	else
+		tc35892_gpio->regs[REG_IS][regoffset] &= ~mask;
+
+	if (type == IRQ_TYPE_EDGE_RISING || type == IRQ_TYPE_LEVEL_HIGH)
+		tc35892_gpio->regs[REG_IEV][regoffset] |= mask;
+	else
+		tc35892_gpio->regs[REG_IEV][regoffset] &= ~mask;
+
+	return 0;
+}
+
+static void tc35892_gpio_irq_lock(unsigned int irq)
+{
+	struct tc35892_gpio *tc35892_gpio = get_irq_chip_data(irq);
+
+	mutex_lock(&tc35892_gpio->irq_lock);
+}
+
+static void tc35892_gpio_irq_sync_unlock(unsigned int irq)
+{
+	struct tc35892_gpio *tc35892_gpio = get_irq_chip_data(irq);
+	struct tc35892 *tc35892 = tc35892_gpio->tc35892;
+	static const u8 regmap[] = {
+		[REG_IBE]	= TC35892_GPIOIBE0,
+		[REG_IEV]	= TC35892_GPIOIEV0,
+		[REG_IS]	= TC35892_GPIOIS0,
+		[REG_IE]	= TC35892_GPIOIE0,
+	};
+	int i, j;
+
+	for (i = 0; i < CACHE_NR_REGS; i++) {
+		for (j = 0; j < CACHE_NR_BANKS; j++) {
+			u8 old = tc35892_gpio->oldregs[i][j];
+			u8 new = tc35892_gpio->regs[i][j];
+
+			if (new == old)
+				continue;
+
+			tc35892_gpio->oldregs[i][j] = new;
+			tc35892_reg_write(tc35892, regmap[i] + j * 8, new);
+		}
+	}
+
+	mutex_unlock(&tc35892_gpio->irq_lock);
+}
+
+static void tc35892_gpio_irq_mask(unsigned int irq)
+{
+	struct tc35892_gpio *tc35892_gpio = get_irq_chip_data(irq);
+	int offset = irq - tc35892_gpio->irq_base;
+	int regoffset = offset / 8;
+	int mask = 1 << (offset % 8);
+
+	tc35892_gpio->regs[REG_IE][regoffset] &= ~mask;
+}
+
+static void tc35892_gpio_irq_unmask(unsigned int irq)
+{
+	struct tc35892_gpio *tc35892_gpio = get_irq_chip_data(irq);
+	int offset = irq - tc35892_gpio->irq_base;
+	int regoffset = offset / 8;
+	int mask = 1 << (offset % 8);
+
+	tc35892_gpio->regs[REG_IE][regoffset] |= mask;
+}
+
+static struct irq_chip tc35892_gpio_irq_chip = {
+	.name			= "tc35892-gpio",
+	.bus_lock		= tc35892_gpio_irq_lock,
+	.bus_sync_unlock	= tc35892_gpio_irq_sync_unlock,
+	.mask			= tc35892_gpio_irq_mask,
+	.unmask			= tc35892_gpio_irq_unmask,
+	.set_type		= tc35892_gpio_irq_set_type,
+};
+
+static irqreturn_t tc35892_gpio_irq(int irq, void *dev)
+{
+	struct tc35892_gpio *tc35892_gpio = dev;
+	struct tc35892 *tc35892 = tc35892_gpio->tc35892;
+	u8 status[CACHE_NR_BANKS];
+	int ret;
+	int i;
+
+	ret = tc35892_block_read(tc35892, TC35892_GPIOMIS0,
+				 ARRAY_SIZE(status), status);
+	if (ret < 0)
+		return IRQ_NONE;
+
+	for (i = 0; i < ARRAY_SIZE(status); i++) {
+		unsigned int stat = status[i];
+		if (!stat)
+			continue;
+
+		while (stat) {
+			int bit = __ffs(stat);
+			int line = i * 8 + bit;
+
+			handle_nested_irq(tc35892_gpio->irq_base + line);
+			stat &= ~(1 << bit);
+		}
+
+		tc35892_reg_write(tc35892, TC35892_GPIOIC0 + i, status[i]);
+	}
+
+	return IRQ_HANDLED;
+}
+
+static int tc35892_gpio_irq_init(struct tc35892_gpio *tc35892_gpio)
+{
+	int base = tc35892_gpio->irq_base;
+	int irq;
+
+	for (irq = base; irq < base + tc35892_gpio->chip.ngpio; irq++) {
+		set_irq_chip_data(irq, tc35892_gpio);
+		set_irq_chip_and_handler(irq, &tc35892_gpio_irq_chip,
+					 handle_simple_irq);
+		set_irq_nested_thread(irq, 1);
+#ifdef CONFIG_ARM
+		set_irq_flags(irq, IRQF_VALID);
+#else
+		set_irq_noprobe(irq);
+#endif
+	}
+
+	return 0;
+}
+
+static void tc35892_gpio_irq_remove(struct tc35892_gpio *tc35892_gpio)
+{
+	int base = tc35892_gpio->irq_base;
+	int irq;
+
+	for (irq = base; irq < base + tc35892_gpio->chip.ngpio; irq++) {
+#ifdef CONFIG_ARM
+		set_irq_flags(irq, 0);
+#endif
+		set_irq_chip_and_handler(irq, NULL, NULL);
+		set_irq_chip_data(irq, NULL);
+	}
+}
+
+static int __devinit tc35892_gpio_probe(struct platform_device *pdev)
+{
+	struct tc35892 *tc35892 = dev_get_drvdata(pdev->dev.parent);
+	struct tc35892_gpio_platform_data *pdata;
+	struct tc35892_gpio *tc35892_gpio;
+	int ret;
+	int irq;
+
+	pdata = tc35892->pdata->gpio;
+	if (!pdata)
+		return -ENODEV;
+
+	irq = platform_get_irq(pdev, 0);
+	if (irq < 0)
+		return irq;
+
+	tc35892_gpio = kzalloc(sizeof(struct tc35892_gpio), GFP_KERNEL);
+	if (!tc35892_gpio)
+		return -ENOMEM;
+
+	mutex_init(&tc35892_gpio->irq_lock);
+
+	tc35892_gpio->dev = &pdev->dev;
+	tc35892_gpio->tc35892 = tc35892;
+
+	tc35892_gpio->chip = template_chip;
+	tc35892_gpio->chip.ngpio = tc35892->num_gpio;
+	tc35892_gpio->chip.dev = &pdev->dev;
+	tc35892_gpio->chip.base = pdata->gpio_base;
+
+	tc35892_gpio->irq_base = tc35892->irq_base + TC35892_INT_GPIO(0);
+
+	/* Bring the GPIO module out of reset */
+	ret = tc35892_set_bits(tc35892, TC35892_RSTCTRL,
+			       TC35892_RSTCTRL_GPIRST, 0);
+	if (ret < 0)
+		goto out_free;
+
+	ret = tc35892_gpio_irq_init(tc35892_gpio);
+	if (ret)
+		goto out_free;
+
+	ret = request_threaded_irq(irq, NULL, tc35892_gpio_irq, IRQF_ONESHOT,
+				   "tc35892-gpio", tc35892_gpio);
+	if (ret) {
+		dev_err(&pdev->dev, "unable to get irq: %d\n", ret);
+		goto out_removeirq;
+	}
+
+	ret = gpiochip_add(&tc35892_gpio->chip);
+	if (ret) {
+		dev_err(&pdev->dev, "unable to add gpiochip: %d\n", ret);
+		goto out_freeirq;
+	}
+
+	platform_set_drvdata(pdev, tc35892_gpio);
+
+	return 0;
+
+out_freeirq:
+	free_irq(irq, tc35892_gpio);
+out_removeirq:
+	tc35892_gpio_irq_remove(tc35892_gpio);
+out_free:
+	kfree(tc35892_gpio);
+	return ret;
+}
+
+static int __devexit tc35892_gpio_remove(struct platform_device *pdev)
+{
+	struct tc35892_gpio *tc35892_gpio = platform_get_drvdata(pdev);
+	int irq = platform_get_irq(pdev, 0);
+	int ret;
+
+	ret = gpiochip_remove(&tc35892_gpio->chip);
+	if (ret < 0) {
+		dev_err(tc35892_gpio->dev,
+			"unable to remove gpiochip: %d\n", ret);
+		return ret;
+	}
+
+	free_irq(irq, tc35892_gpio);
+	tc35892_gpio_irq_remove(tc35892_gpio);
+
+	platform_set_drvdata(pdev, NULL);
+	kfree(tc35892_gpio);
+
+	return 0;
+}
+
+static struct platform_driver tc35892_gpio_driver = {
+	.driver.name	= "tc35892-gpio",
+	.driver.owner	= THIS_MODULE,
+	.probe		= tc35892_gpio_probe,
+	.remove		= __devexit_p(tc35892_gpio_remove),
+};
+
+static int __init tc35892_gpio_init(void)
+{
+	return platform_driver_register(&tc35892_gpio_driver);
+}
+subsys_initcall(tc35892_gpio_init);
+
+static void __exit tc35892_gpio_exit(void)
+{
+	platform_driver_unregister(&tc35892_gpio_driver);
+}
+module_exit(tc35892_gpio_exit);
+
+MODULE_LICENSE("GPL v2");
+MODULE_DESCRIPTION("TC35892 GPIO driver");
+MODULE_AUTHOR("Hanumath Prasad, Rabin Vincent");
diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c
index 994d23beeb1d..57cea01c4ffb 100644
--- a/drivers/gpu/drm/drm_crtc.c
+++ b/drivers/gpu/drm/drm_crtc.c
@@ -1840,8 +1840,10 @@ int drm_mode_dirtyfb_ioctl(struct drm_device *dev,
 
 		ret = copy_from_user(clips, clips_ptr,
 				     num_clips * sizeof(*clips));
-		if (ret)
+		if (ret) {
+			ret = -EFAULT;
 			goto out_err2;
+		}
 	}
 
 	if (fb->funcs->dirty) {
diff --git a/drivers/gpu/drm/drm_crtc_helper.c b/drivers/gpu/drm/drm_crtc_helper.c
index 764401951041..9b2a54117c91 100644
--- a/drivers/gpu/drm/drm_crtc_helper.c
+++ b/drivers/gpu/drm/drm_crtc_helper.c
@@ -860,19 +860,24 @@ static void output_poll_execute(struct slow_work *work)
 	}
 }
 
-void drm_kms_helper_poll_init(struct drm_device *dev)
+void drm_kms_helper_poll_disable(struct drm_device *dev)
+{
+	if (!dev->mode_config.poll_enabled)
+		return;
+	delayed_slow_work_cancel(&dev->mode_config.output_poll_slow_work);
+}
+EXPORT_SYMBOL(drm_kms_helper_poll_disable);
+
+void drm_kms_helper_poll_enable(struct drm_device *dev)
 {
-	struct drm_connector *connector;
 	bool poll = false;
+	struct drm_connector *connector;
 	int ret;
 
 	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
 		if (connector->polled)
 			poll = true;
 	}
-	slow_work_register_user(THIS_MODULE);
-	delayed_slow_work_init(&dev->mode_config.output_poll_slow_work,
-			       &output_poll_ops);
 
 	if (poll) {
 		ret = delayed_slow_work_enqueue(&dev->mode_config.output_poll_slow_work, DRM_OUTPUT_POLL_PERIOD);
@@ -880,11 +885,22 @@ void drm_kms_helper_poll_init(struct drm_device *dev)
 			DRM_ERROR("delayed enqueue failed %d\n", ret);
 	}
 }
+EXPORT_SYMBOL(drm_kms_helper_poll_enable);
+
+void drm_kms_helper_poll_init(struct drm_device *dev)
+{
+	slow_work_register_user(THIS_MODULE);
+	delayed_slow_work_init(&dev->mode_config.output_poll_slow_work,
+			       &output_poll_ops);
+	dev->mode_config.poll_enabled = true;
+
+	drm_kms_helper_poll_enable(dev);
+}
 EXPORT_SYMBOL(drm_kms_helper_poll_init);
 
 void drm_kms_helper_poll_fini(struct drm_device *dev)
 {
-	delayed_slow_work_cancel(&dev->mode_config.output_poll_slow_work);
+	drm_kms_helper_poll_disable(dev);
 	slow_work_unregister_user(THIS_MODULE);
 }
 EXPORT_SYMBOL(drm_kms_helper_poll_fini);
diff --git a/drivers/gpu/drm/drm_edid.c b/drivers/gpu/drm/drm_edid.c
index f569ae88ab38..c1981861bbbd 100644
--- a/drivers/gpu/drm/drm_edid.c
+++ b/drivers/gpu/drm/drm_edid.c
@@ -147,7 +147,10 @@ drm_edid_block_valid(u8 *raw_edid)
 		csum += raw_edid[i];
 	if (csum) {
 		DRM_ERROR("EDID checksum is invalid, remainder is %d\n", csum);
-		goto bad;
+
+		/* allow CEA to slide through, switches mangle this */
+		if (raw_edid[0] != 0x02)
+			goto bad;
 	}
 
 	/* per-block-type checks */
diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c
index b3779d243aef..08c4c926e65f 100644
--- a/drivers/gpu/drm/drm_fb_helper.c
+++ b/drivers/gpu/drm/drm_fb_helper.c
@@ -264,7 +264,7 @@ bool drm_fb_helper_force_kernel_mode(void)
 int drm_fb_helper_panic(struct notifier_block *n, unsigned long ununsed,
 			void *panic_str)
 {
-	DRM_ERROR("panic occurred, switching back to text console\n");
+	printk(KERN_ERR "panic occurred, switching back to text console\n");
 	return drm_fb_helper_force_kernel_mode();
 	return 0;
 }
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index 95639017bdbe..da78f2c0d909 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -22,6 +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_ringbuffer.o \
 	  intel_overlay.o \
 	  dvo_ch7xxx.o \
 	  dvo_ch7017.o \
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 322070c0c631..52510ad8b25d 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -77,7 +77,7 @@ static int i915_gem_object_list_info(struct seq_file *m, void *data)
 	case ACTIVE_LIST:
 		seq_printf(m, "Active:\n");
 		lock = &dev_priv->mm.active_list_lock;
-		head = &dev_priv->mm.active_list;
+		head = &dev_priv->render_ring.active_list;
 		break;
 	case INACTIVE_LIST:
 		seq_printf(m, "Inactive:\n");
@@ -129,7 +129,8 @@ static int i915_gem_request_info(struct seq_file *m, void *data)
 	struct drm_i915_gem_request *gem_request;
 
 	seq_printf(m, "Request:\n");
-	list_for_each_entry(gem_request, &dev_priv->mm.request_list, list) {
+	list_for_each_entry(gem_request, &dev_priv->render_ring.request_list,
+			list) {
 		seq_printf(m, "    %d @ %d\n",
 			   gem_request->seqno,
 			   (int) (jiffies - gem_request->emitted_jiffies));
@@ -143,9 +144,9 @@ static int i915_gem_seqno_info(struct seq_file *m, void *data)
 	struct drm_device *dev = node->minor->dev;
 	drm_i915_private_t *dev_priv = dev->dev_private;
 
-	if (dev_priv->hw_status_page != NULL) {
+	if (dev_priv->render_ring.status_page.page_addr != NULL) {
 		seq_printf(m, "Current sequence: %d\n",
-			   i915_get_gem_seqno(dev));
+			   i915_get_gem_seqno(dev,  &dev_priv->render_ring));
 	} else {
 		seq_printf(m, "Current sequence: hws uninitialized\n");
 	}
@@ -195,9 +196,9 @@ static int i915_interrupt_info(struct seq_file *m, void *data)
 	}
 	seq_printf(m, "Interrupts received: %d\n",
 		   atomic_read(&dev_priv->irq_received));
-	if (dev_priv->hw_status_page != NULL) {
+	if (dev_priv->render_ring.status_page.page_addr != NULL) {
 		seq_printf(m, "Current sequence:    %d\n",
-			   i915_get_gem_seqno(dev));
+			   i915_get_gem_seqno(dev,  &dev_priv->render_ring));
 	} else {
 		seq_printf(m, "Current sequence:    hws uninitialized\n");
 	}
@@ -251,7 +252,7 @@ static int i915_hws_info(struct seq_file *m, void *data)
 	int i;
 	volatile u32 *hws;
 
-	hws = (volatile u32 *)dev_priv->hw_status_page;
+	hws = (volatile u32 *)dev_priv->render_ring.status_page.page_addr;
 	if (hws == NULL)
 		return 0;
 
@@ -287,7 +288,8 @@ static int i915_batchbuffer_info(struct seq_file *m, void *data)
 
 	spin_lock(&dev_priv->mm.active_list_lock);
 
-	list_for_each_entry(obj_priv, &dev_priv->mm.active_list, list) {
+	list_for_each_entry(obj_priv, &dev_priv->render_ring.active_list,
+			list) {
 		obj = &obj_priv->base;
 		if (obj->read_domains & I915_GEM_DOMAIN_COMMAND) {
 		    ret = i915_gem_object_get_pages(obj, 0);
@@ -317,14 +319,14 @@ static int i915_ringbuffer_data(struct seq_file *m, void *data)
 	u8 *virt;
 	uint32_t *ptr, off;
 
-	if (!dev_priv->ring.ring_obj) {
+	if (!dev_priv->render_ring.gem_object) {
 		seq_printf(m, "No ringbuffer setup\n");
 		return 0;
 	}
 
-	virt = dev_priv->ring.virtual_start;
+	virt = dev_priv->render_ring.virtual_start;
 
-	for (off = 0; off < dev_priv->ring.Size; off += 4) {
+	for (off = 0; off < dev_priv->render_ring.size; off += 4) {
 		ptr = (uint32_t *)(virt + off);
 		seq_printf(m, "%08x :  %08x\n", off, *ptr);
 	}
@@ -344,7 +346,7 @@ static int i915_ringbuffer_info(struct seq_file *m, void *data)
 
 	seq_printf(m, "RingHead :  %08x\n", head);
 	seq_printf(m, "RingTail :  %08x\n", tail);
-	seq_printf(m, "RingSize :  %08lx\n", dev_priv->ring.Size);
+	seq_printf(m, "RingSize :  %08lx\n", dev_priv->render_ring.size);
 	seq_printf(m, "Acthd :     %08x\n", I915_READ(IS_I965G(dev) ? ACTHD_I965 : ACTHD));
 
 	return 0;
@@ -489,11 +491,14 @@ static int i915_cur_delayinfo(struct seq_file *m, void *unused)
 	struct drm_device *dev = node->minor->dev;
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	u16 rgvswctl = I915_READ16(MEMSWCTL);
+	u16 rgvstat = I915_READ16(MEMSTAT_ILK);
 
-	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);
+	seq_printf(m, "Requested P-state: %d\n", (rgvswctl >> 8) & 0xf);
+	seq_printf(m, "Requested VID: %d\n", rgvswctl & 0x3f);
+	seq_printf(m, "Current VID: %d\n", (rgvstat & MEMSTAT_VID_MASK) >>
+		   MEMSTAT_VID_SHIFT);
+	seq_printf(m, "Current P-state: %d\n",
+		   (rgvstat & MEMSTAT_PSTATE_MASK) >> MEMSTAT_PSTATE_SHIFT);
 
 	return 0;
 }
@@ -508,7 +513,8 @@ static int i915_delayfreq_table(struct seq_file *m, void *unused)
 
 	for (i = 0; i < 16; i++) {
 		delayfreq = I915_READ(PXVFREQ_BASE + i * 4);
-		seq_printf(m, "P%02dVIDFREQ: 0x%08x\n", i, delayfreq);
+		seq_printf(m, "P%02dVIDFREQ: 0x%08x (VID: %d)\n", i, delayfreq,
+			   (delayfreq & PXVFREQ_PX_MASK) >> PXVFREQ_PX_SHIFT);
 	}
 
 	return 0;
@@ -541,6 +547,8 @@ static int i915_drpc_info(struct seq_file *m, void *unused)
 	struct drm_device *dev = node->minor->dev;
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	u32 rgvmodectl = I915_READ(MEMMODECTL);
+	u32 rstdbyctl = I915_READ(MCHBAR_RENDER_STANDBY);
+	u16 crstandvid = I915_READ16(CRSTANDVID);
 
 	seq_printf(m, "HD boost: %s\n", (rgvmodectl & MEMMODE_BOOST_EN) ?
 		   "yes" : "no");
@@ -555,9 +563,13 @@ static int i915_drpc_info(struct seq_file *m, void *unused)
 		   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",
+	seq_printf(m, "Max P-state: P%d\n",
 		   (rgvmodectl & MEMMODE_FMAX_MASK) >> MEMMODE_FMAX_SHIFT);
-	seq_printf(m, "Min frequency: P%d\n", (rgvmodectl & MEMMODE_FMIN_MASK));
+	seq_printf(m, "Min P-state: P%d\n", (rgvmodectl & MEMMODE_FMIN_MASK));
+	seq_printf(m, "RS1 VID: %d\n", (crstandvid & 0x3f));
+	seq_printf(m, "RS2 VID: %d\n", ((crstandvid >> 8) & 0x3f));
+	seq_printf(m, "Render standby enabled: %s\n",
+		   (rstdbyctl & RCX_SW_EXIT) ? "no" : "yes");
 
 	return 0;
 }
@@ -621,6 +633,36 @@ static int i915_sr_status(struct seq_file *m, void *unused)
 	return 0;
 }
 
+static int i915_emon_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;
+	unsigned long temp, chipset, gfx;
+
+	temp = i915_mch_val(dev_priv);
+	chipset = i915_chipset_val(dev_priv);
+	gfx = i915_gfx_val(dev_priv);
+
+	seq_printf(m, "GMCH temp: %ld\n", temp);
+	seq_printf(m, "Chipset power: %ld\n", chipset);
+	seq_printf(m, "GFX power: %ld\n", gfx);
+	seq_printf(m, "Total power: %ld\n", chipset + gfx);
+
+	return 0;
+}
+
+static int i915_gfxec(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;
+
+	seq_printf(m, "GFXEC: %ld\n", (unsigned long)I915_READ(0x112f4));
+
+	return 0;
+}
+
 static int
 i915_wedged_open(struct inode *inode,
 		 struct file *filp)
@@ -743,6 +785,8 @@ static struct drm_info_list i915_debugfs_list[] = {
 	{"i915_delayfreq_table", i915_delayfreq_table, 0},
 	{"i915_inttoext_table", i915_inttoext_table, 0},
 	{"i915_drpc_info", i915_drpc_info, 0},
+	{"i915_emon_status", i915_emon_status, 0},
+	{"i915_gfxec", i915_gfxec, 0},
 	{"i915_fbc_status", i915_fbc_status, 0},
 	{"i915_sr_status", i915_sr_status, 0},
 };
diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c
index 2a6b5de5ae5d..59a2bf8592ec 100644
--- a/drivers/gpu/drm/i915/i915_dma.c
+++ b/drivers/gpu/drm/i915/i915_dma.c
@@ -40,84 +40,6 @@
 #include <linux/vga_switcheroo.h>
 #include <linux/slab.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
- * the head pointer changes, so that EBUSY only happens if the ring
- * actually stalls for (eg) 3 seconds.
- */
-int i915_wait_ring(struct drm_device * dev, int n, const char *caller)
-{
-	drm_i915_private_t *dev_priv = dev->dev_private;
-	drm_i915_ring_buffer_t *ring = &(dev_priv->ring);
-	u32 acthd_reg = IS_I965G(dev) ? ACTHD_I965 : ACTHD;
-	u32 last_acthd = I915_READ(acthd_reg);
-	u32 acthd;
-	u32 last_head = I915_READ(PRB0_HEAD) & HEAD_ADDR;
-	int i;
-
-	trace_i915_ring_wait_begin (dev);
-
-	for (i = 0; i < 100000; i++) {
-		ring->head = I915_READ(PRB0_HEAD) & HEAD_ADDR;
-		acthd = I915_READ(acthd_reg);
-		ring->space = ring->head - (ring->tail + 8);
-		if (ring->space < 0)
-			ring->space += ring->Size;
-		if (ring->space >= n) {
-			trace_i915_ring_wait_end (dev);
-			return 0;
-		}
-
-		if (dev->primary->master) {
-			struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv;
-			if (master_priv->sarea_priv)
-				master_priv->sarea_priv->perf_boxes |= I915_BOX_WAIT;
-		}
-
-
-		if (ring->head != last_head)
-			i = 0;
-		if (acthd != last_acthd)
-			i = 0;
-
-		last_head = ring->head;
-		last_acthd = acthd;
-		msleep_interruptible(10);
-
-	}
-
-	trace_i915_ring_wait_end (dev);
-	return -EBUSY;
-}
-
-/* As a ringbuffer is only allowed to wrap between instructions, fill
- * the tail with NOOPs.
- */
-int i915_wrap_ring(struct drm_device *dev)
-{
-	drm_i915_private_t *dev_priv = dev->dev_private;
-	volatile unsigned int *virt;
-	int rem;
-
-	rem = dev_priv->ring.Size - dev_priv->ring.tail;
-	if (dev_priv->ring.space < rem) {
-		int ret = i915_wait_ring(dev, rem, __func__);
-		if (ret)
-			return ret;
-	}
-	dev_priv->ring.space -= rem;
-
-	virt = (unsigned int *)
-		(dev_priv->ring.virtual_start + dev_priv->ring.tail);
-	rem /= 4;
-	while (rem--)
-		*virt++ = MI_NOOP;
-
-	dev_priv->ring.tail = 0;
-
-	return 0;
-}
-
 /**
  * Sets up the hardware status page for devices that need a physical address
  * in the register.
@@ -133,10 +55,11 @@ static int i915_init_phys_hws(struct drm_device *dev)
 		DRM_ERROR("Can not allocate hardware status page\n");
 		return -ENOMEM;
 	}
-	dev_priv->hw_status_page = dev_priv->status_page_dmah->vaddr;
+	dev_priv->render_ring.status_page.page_addr
+		= dev_priv->status_page_dmah->vaddr;
 	dev_priv->dma_status_page = dev_priv->status_page_dmah->busaddr;
 
-	memset(dev_priv->hw_status_page, 0, PAGE_SIZE);
+	memset(dev_priv->render_ring.status_page.page_addr, 0, PAGE_SIZE);
 
 	if (IS_I965G(dev))
 		dev_priv->dma_status_page |= (dev_priv->dma_status_page >> 28) &
@@ -159,8 +82,8 @@ static void i915_free_hws(struct drm_device *dev)
 		dev_priv->status_page_dmah = NULL;
 	}
 
-	if (dev_priv->status_gfx_addr) {
-		dev_priv->status_gfx_addr = 0;
+	if (dev_priv->render_ring.status_page.gfx_addr) {
+		dev_priv->render_ring.status_page.gfx_addr = 0;
 		drm_core_ioremapfree(&dev_priv->hws_map, dev);
 	}
 
@@ -172,7 +95,7 @@ void i915_kernel_lost_context(struct drm_device * dev)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	struct drm_i915_master_private *master_priv;
-	drm_i915_ring_buffer_t *ring = &(dev_priv->ring);
+	struct intel_ring_buffer *ring = &dev_priv->render_ring;
 
 	/*
 	 * We should never lose context on the ring with modesetting
@@ -185,7 +108,7 @@ void i915_kernel_lost_context(struct drm_device * dev)
 	ring->tail = I915_READ(PRB0_TAIL) & TAIL_ADDR;
 	ring->space = ring->head - (ring->tail + 8);
 	if (ring->space < 0)
-		ring->space += ring->Size;
+		ring->space += ring->size;
 
 	if (!dev->primary->master)
 		return;
@@ -205,12 +128,9 @@ static int i915_dma_cleanup(struct drm_device * dev)
 	if (dev->irq_enabled)
 		drm_irq_uninstall(dev);
 
-	if (dev_priv->ring.virtual_start) {
-		drm_core_ioremapfree(&dev_priv->ring.map, dev);
-		dev_priv->ring.virtual_start = NULL;
-		dev_priv->ring.map.handle = NULL;
-		dev_priv->ring.map.size = 0;
-	}
+	intel_cleanup_ring_buffer(dev, &dev_priv->render_ring);
+	if (HAS_BSD(dev))
+		intel_cleanup_ring_buffer(dev, &dev_priv->bsd_ring);
 
 	/* Clear the HWS virtual address at teardown */
 	if (I915_NEED_GFX_HWS(dev))
@@ -233,24 +153,24 @@ static int i915_initialize(struct drm_device * dev, drm_i915_init_t * init)
 	}
 
 	if (init->ring_size != 0) {
-		if (dev_priv->ring.ring_obj != NULL) {
+		if (dev_priv->render_ring.gem_object != NULL) {
 			i915_dma_cleanup(dev);
 			DRM_ERROR("Client tried to initialize ringbuffer in "
 				  "GEM mode\n");
 			return -EINVAL;
 		}
 
-		dev_priv->ring.Size = init->ring_size;
+		dev_priv->render_ring.size = init->ring_size;
 
-		dev_priv->ring.map.offset = init->ring_start;
-		dev_priv->ring.map.size = init->ring_size;
-		dev_priv->ring.map.type = 0;
-		dev_priv->ring.map.flags = 0;
-		dev_priv->ring.map.mtrr = 0;
+		dev_priv->render_ring.map.offset = init->ring_start;
+		dev_priv->render_ring.map.size = init->ring_size;
+		dev_priv->render_ring.map.type = 0;
+		dev_priv->render_ring.map.flags = 0;
+		dev_priv->render_ring.map.mtrr = 0;
 
-		drm_core_ioremap_wc(&dev_priv->ring.map, dev);
+		drm_core_ioremap_wc(&dev_priv->render_ring.map, dev);
 
-		if (dev_priv->ring.map.handle == NULL) {
+		if (dev_priv->render_ring.map.handle == NULL) {
 			i915_dma_cleanup(dev);
 			DRM_ERROR("can not ioremap virtual address for"
 				  " ring buffer\n");
@@ -258,7 +178,7 @@ static int i915_initialize(struct drm_device * dev, drm_i915_init_t * init)
 		}
 	}
 
-	dev_priv->ring.virtual_start = dev_priv->ring.map.handle;
+	dev_priv->render_ring.virtual_start = dev_priv->render_ring.map.handle;
 
 	dev_priv->cpp = init->cpp;
 	dev_priv->back_offset = init->back_offset;
@@ -278,26 +198,29 @@ static int i915_dma_resume(struct drm_device * dev)
 {
 	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
 
+	struct intel_ring_buffer *ring;
 	DRM_DEBUG_DRIVER("%s\n", __func__);
 
-	if (dev_priv->ring.map.handle == NULL) {
+	ring = &dev_priv->render_ring;
+
+	if (ring->map.handle == NULL) {
 		DRM_ERROR("can not ioremap virtual address for"
 			  " ring buffer\n");
 		return -ENOMEM;
 	}
 
 	/* Program Hardware Status Page */
-	if (!dev_priv->hw_status_page) {
+	if (!ring->status_page.page_addr) {
 		DRM_ERROR("Can not find hardware status page\n");
 		return -EINVAL;
 	}
 	DRM_DEBUG_DRIVER("hw status page @ %p\n",
-				dev_priv->hw_status_page);
-
-	if (dev_priv->status_gfx_addr != 0)
-		I915_WRITE(HWS_PGA, dev_priv->status_gfx_addr);
+				ring->status_page.page_addr);
+	if (ring->status_page.gfx_addr != 0)
+		ring->setup_status_page(dev, ring);
 	else
 		I915_WRITE(HWS_PGA, dev_priv->dma_status_page);
+
 	DRM_DEBUG_DRIVER("Enabled hardware status page\n");
 
 	return 0;
@@ -407,9 +330,8 @@ static int i915_emit_cmds(struct drm_device * dev, int *buffer, int dwords)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	int i;
-	RING_LOCALS;
 
-	if ((dwords+1) * sizeof(int) >= dev_priv->ring.Size - 8)
+	if ((dwords+1) * sizeof(int) >= dev_priv->render_ring.size - 8)
 		return -EINVAL;
 
 	BEGIN_LP_RING((dwords+1)&~1);
@@ -442,9 +364,7 @@ i915_emit_box(struct drm_device *dev,
 	      struct drm_clip_rect *boxes,
 	      int i, int DR1, int DR4)
 {
-	drm_i915_private_t *dev_priv = dev->dev_private;
 	struct drm_clip_rect box = boxes[i];
-	RING_LOCALS;
 
 	if (box.y2 <= box.y1 || box.x2 <= box.x1 || box.y2 <= 0 || box.x2 <= 0) {
 		DRM_ERROR("Bad box %d,%d..%d,%d\n",
@@ -481,7 +401,6 @@ static void i915_emit_breadcrumb(struct drm_device *dev)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv;
-	RING_LOCALS;
 
 	dev_priv->counter++;
 	if (dev_priv->counter > 0x7FFFFFFFUL)
@@ -535,10 +454,8 @@ static int i915_dispatch_batchbuffer(struct drm_device * dev,
 				     drm_i915_batchbuffer_t * batch,
 				     struct drm_clip_rect *cliprects)
 {
-	drm_i915_private_t *dev_priv = dev->dev_private;
 	int nbox = batch->num_cliprects;
 	int i = 0, count;
-	RING_LOCALS;
 
 	if ((batch->start | batch->used) & 0x7) {
 		DRM_ERROR("alignment");
@@ -587,7 +504,6 @@ static int i915_dispatch_flip(struct drm_device * dev)
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	struct drm_i915_master_private *master_priv =
 		dev->primary->master->driver_priv;
-	RING_LOCALS;
 
 	if (!master_priv->sarea_priv)
 		return -EINVAL;
@@ -640,7 +556,8 @@ static int i915_quiescent(struct drm_device * dev)
 	drm_i915_private_t *dev_priv = dev->dev_private;
 
 	i915_kernel_lost_context(dev);
-	return i915_wait_ring(dev, dev_priv->ring.Size - 8, __func__);
+	return intel_wait_ring_buffer(dev, &dev_priv->render_ring,
+				      dev_priv->render_ring.size - 8);
 }
 
 static int i915_flush_ioctl(struct drm_device *dev, void *data,
@@ -827,6 +744,9 @@ static int i915_getparam(struct drm_device *dev, void *data,
 		/* depends on GEM */
 		value = dev_priv->has_gem;
 		break;
+	case I915_PARAM_HAS_BSD:
+		value = HAS_BSD(dev);
+		break;
 	default:
 		DRM_DEBUG_DRIVER("Unknown parameter %d\n",
 				 param->param);
@@ -882,6 +802,7 @@ static int i915_set_status_page(struct drm_device *dev, void *data,
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	drm_i915_hws_addr_t *hws = data;
+	struct intel_ring_buffer *ring = &dev_priv->render_ring;
 
 	if (!I915_NEED_GFX_HWS(dev))
 		return -EINVAL;
@@ -898,7 +819,7 @@ static int i915_set_status_page(struct drm_device *dev, void *data,
 
 	DRM_DEBUG_DRIVER("set status page addr 0x%08x\n", (u32)hws->addr);
 
-	dev_priv->status_gfx_addr = hws->addr & (0x1ffff<<12);
+	ring->status_page.gfx_addr = hws->addr & (0x1ffff<<12);
 
 	dev_priv->hws_map.offset = dev->agp->base + hws->addr;
 	dev_priv->hws_map.size = 4*1024;
@@ -909,19 +830,19 @@ static int i915_set_status_page(struct drm_device *dev, void *data,
 	drm_core_ioremap_wc(&dev_priv->hws_map, dev);
 	if (dev_priv->hws_map.handle == NULL) {
 		i915_dma_cleanup(dev);
-		dev_priv->status_gfx_addr = 0;
+		ring->status_page.gfx_addr = 0;
 		DRM_ERROR("can not ioremap virtual address for"
 				" G33 hw status page\n");
 		return -ENOMEM;
 	}
-	dev_priv->hw_status_page = dev_priv->hws_map.handle;
+	ring->status_page.page_addr = dev_priv->hws_map.handle;
+	memset(ring->status_page.page_addr, 0, PAGE_SIZE);
+	I915_WRITE(HWS_PGA, ring->status_page.gfx_addr);
 
-	memset(dev_priv->hw_status_page, 0, PAGE_SIZE);
-	I915_WRITE(HWS_PGA, dev_priv->status_gfx_addr);
 	DRM_DEBUG_DRIVER("load hws HWS_PGA with gfx mem 0x%x\n",
-				dev_priv->status_gfx_addr);
+			 ring->status_page.gfx_addr);
 	DRM_DEBUG_DRIVER("load hws at %p\n",
-				dev_priv->hw_status_page);
+			 ring->status_page.page_addr);
 	return 0;
 }
 
@@ -1399,12 +1320,14 @@ static void i915_switcheroo_set_state(struct pci_dev *pdev, enum vga_switcheroo_
 	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");
+		printk(KERN_INFO "i915: switched on\n");
 		/* i915 resume handler doesn't set to D0 */
 		pci_set_power_state(dev->pdev, PCI_D0);
 		i915_resume(dev);
+		drm_kms_helper_poll_enable(dev);
 	} else {
 		printk(KERN_ERR "i915: switched off\n");
+		drm_kms_helper_poll_disable(dev);
 		i915_suspend(dev, pmm);
 	}
 }
@@ -1479,19 +1402,19 @@ static int i915_load_modeset_init(struct drm_device *dev,
 	/* if we have > 1 VGA cards, then disable the radeon VGA resources */
 	ret = vga_client_register(dev->pdev, dev, NULL, i915_vga_set_decode);
 	if (ret)
-		goto destroy_ringbuffer;
+		goto cleanup_ringbuffer;
 
 	ret = vga_switcheroo_register_client(dev->pdev,
 					     i915_switcheroo_set_state,
 					     i915_switcheroo_can_switch);
 	if (ret)
-		goto destroy_ringbuffer;
+		goto cleanup_vga_client;
 
 	intel_modeset_init(dev);
 
 	ret = drm_irq_install(dev);
 	if (ret)
-		goto destroy_ringbuffer;
+		goto cleanup_vga_switcheroo;
 
 	/* Always safe in the mode setting case. */
 	/* FIXME: do pre/post-mode set stuff in core KMS code */
@@ -1503,11 +1426,20 @@ static int i915_load_modeset_init(struct drm_device *dev,
 
 	I915_WRITE(INSTPM, (1 << 5) | (1 << 21));
 
-	intel_fbdev_init(dev);
+	ret = intel_fbdev_init(dev);
+	if (ret)
+		goto cleanup_irq;
+
 	drm_kms_helper_poll_init(dev);
 	return 0;
 
-destroy_ringbuffer:
+cleanup_irq:
+	drm_irq_uninstall(dev);
+cleanup_vga_switcheroo:
+	vga_switcheroo_unregister_client(dev->pdev);
+cleanup_vga_client:
+	vga_client_register(dev->pdev, NULL, NULL, NULL);
+cleanup_ringbuffer:
 	mutex_lock(&dev->struct_mutex);
 	i915_gem_cleanup_ringbuffer(dev);
 	mutex_unlock(&dev->struct_mutex);
@@ -1539,14 +1471,11 @@ void i915_master_destroy(struct drm_device *dev, struct drm_master *master)
 	master->driver_priv = NULL;
 }
 
-static void i915_get_mem_freq(struct drm_device *dev)
+static void i915_pineview_get_mem_freq(struct drm_device *dev)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	u32 tmp;
 
-	if (!IS_PINEVIEW(dev))
-		return;
-
 	tmp = I915_READ(CLKCFG);
 
 	switch (tmp & CLKCFG_FSB_MASK) {
@@ -1575,7 +1504,524 @@ static void i915_get_mem_freq(struct drm_device *dev)
 		dev_priv->mem_freq = 800;
 		break;
 	}
+
+	/* detect pineview DDR3 setting */
+	tmp = I915_READ(CSHRDDR3CTL);
+	dev_priv->is_ddr3 = (tmp & CSHRDDR3CTL_DDR3) ? 1 : 0;
+}
+
+static void i915_ironlake_get_mem_freq(struct drm_device *dev)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	u16 ddrpll, csipll;
+
+	ddrpll = I915_READ16(DDRMPLL1);
+	csipll = I915_READ16(CSIPLL0);
+
+	switch (ddrpll & 0xff) {
+	case 0xc:
+		dev_priv->mem_freq = 800;
+		break;
+	case 0x10:
+		dev_priv->mem_freq = 1066;
+		break;
+	case 0x14:
+		dev_priv->mem_freq = 1333;
+		break;
+	case 0x18:
+		dev_priv->mem_freq = 1600;
+		break;
+	default:
+		DRM_DEBUG_DRIVER("unknown memory frequency 0x%02x\n",
+				 ddrpll & 0xff);
+		dev_priv->mem_freq = 0;
+		break;
+	}
+
+	dev_priv->r_t = dev_priv->mem_freq;
+
+	switch (csipll & 0x3ff) {
+	case 0x00c:
+		dev_priv->fsb_freq = 3200;
+		break;
+	case 0x00e:
+		dev_priv->fsb_freq = 3733;
+		break;
+	case 0x010:
+		dev_priv->fsb_freq = 4266;
+		break;
+	case 0x012:
+		dev_priv->fsb_freq = 4800;
+		break;
+	case 0x014:
+		dev_priv->fsb_freq = 5333;
+		break;
+	case 0x016:
+		dev_priv->fsb_freq = 5866;
+		break;
+	case 0x018:
+		dev_priv->fsb_freq = 6400;
+		break;
+	default:
+		DRM_DEBUG_DRIVER("unknown fsb frequency 0x%04x\n",
+				 csipll & 0x3ff);
+		dev_priv->fsb_freq = 0;
+		break;
+	}
+
+	if (dev_priv->fsb_freq == 3200) {
+		dev_priv->c_m = 0;
+	} else if (dev_priv->fsb_freq > 3200 && dev_priv->fsb_freq <= 4800) {
+		dev_priv->c_m = 1;
+	} else {
+		dev_priv->c_m = 2;
+	}
+}
+
+struct v_table {
+	u8 vid;
+	unsigned long vd; /* in .1 mil */
+	unsigned long vm; /* in .1 mil */
+	u8 pvid;
+};
+
+static struct v_table v_table[] = {
+	{ 0, 16125, 15000, 0x7f, },
+	{ 1, 16000, 14875, 0x7e, },
+	{ 2, 15875, 14750, 0x7d, },
+	{ 3, 15750, 14625, 0x7c, },
+	{ 4, 15625, 14500, 0x7b, },
+	{ 5, 15500, 14375, 0x7a, },
+	{ 6, 15375, 14250, 0x79, },
+	{ 7, 15250, 14125, 0x78, },
+	{ 8, 15125, 14000, 0x77, },
+	{ 9, 15000, 13875, 0x76, },
+	{ 10, 14875, 13750, 0x75, },
+	{ 11, 14750, 13625, 0x74, },
+	{ 12, 14625, 13500, 0x73, },
+	{ 13, 14500, 13375, 0x72, },
+	{ 14, 14375, 13250, 0x71, },
+	{ 15, 14250, 13125, 0x70, },
+	{ 16, 14125, 13000, 0x6f, },
+	{ 17, 14000, 12875, 0x6e, },
+	{ 18, 13875, 12750, 0x6d, },
+	{ 19, 13750, 12625, 0x6c, },
+	{ 20, 13625, 12500, 0x6b, },
+	{ 21, 13500, 12375, 0x6a, },
+	{ 22, 13375, 12250, 0x69, },
+	{ 23, 13250, 12125, 0x68, },
+	{ 24, 13125, 12000, 0x67, },
+	{ 25, 13000, 11875, 0x66, },
+	{ 26, 12875, 11750, 0x65, },
+	{ 27, 12750, 11625, 0x64, },
+	{ 28, 12625, 11500, 0x63, },
+	{ 29, 12500, 11375, 0x62, },
+	{ 30, 12375, 11250, 0x61, },
+	{ 31, 12250, 11125, 0x60, },
+	{ 32, 12125, 11000, 0x5f, },
+	{ 33, 12000, 10875, 0x5e, },
+	{ 34, 11875, 10750, 0x5d, },
+	{ 35, 11750, 10625, 0x5c, },
+	{ 36, 11625, 10500, 0x5b, },
+	{ 37, 11500, 10375, 0x5a, },
+	{ 38, 11375, 10250, 0x59, },
+	{ 39, 11250, 10125, 0x58, },
+	{ 40, 11125, 10000, 0x57, },
+	{ 41, 11000, 9875, 0x56, },
+	{ 42, 10875, 9750, 0x55, },
+	{ 43, 10750, 9625, 0x54, },
+	{ 44, 10625, 9500, 0x53, },
+	{ 45, 10500, 9375, 0x52, },
+	{ 46, 10375, 9250, 0x51, },
+	{ 47, 10250, 9125, 0x50, },
+	{ 48, 10125, 9000, 0x4f, },
+	{ 49, 10000, 8875, 0x4e, },
+	{ 50, 9875, 8750, 0x4d, },
+	{ 51, 9750, 8625, 0x4c, },
+	{ 52, 9625, 8500, 0x4b, },
+	{ 53, 9500, 8375, 0x4a, },
+	{ 54, 9375, 8250, 0x49, },
+	{ 55, 9250, 8125, 0x48, },
+	{ 56, 9125, 8000, 0x47, },
+	{ 57, 9000, 7875, 0x46, },
+	{ 58, 8875, 7750, 0x45, },
+	{ 59, 8750, 7625, 0x44, },
+	{ 60, 8625, 7500, 0x43, },
+	{ 61, 8500, 7375, 0x42, },
+	{ 62, 8375, 7250, 0x41, },
+	{ 63, 8250, 7125, 0x40, },
+	{ 64, 8125, 7000, 0x3f, },
+	{ 65, 8000, 6875, 0x3e, },
+	{ 66, 7875, 6750, 0x3d, },
+	{ 67, 7750, 6625, 0x3c, },
+	{ 68, 7625, 6500, 0x3b, },
+	{ 69, 7500, 6375, 0x3a, },
+	{ 70, 7375, 6250, 0x39, },
+	{ 71, 7250, 6125, 0x38, },
+	{ 72, 7125, 6000, 0x37, },
+	{ 73, 7000, 5875, 0x36, },
+	{ 74, 6875, 5750, 0x35, },
+	{ 75, 6750, 5625, 0x34, },
+	{ 76, 6625, 5500, 0x33, },
+	{ 77, 6500, 5375, 0x32, },
+	{ 78, 6375, 5250, 0x31, },
+	{ 79, 6250, 5125, 0x30, },
+	{ 80, 6125, 5000, 0x2f, },
+	{ 81, 6000, 4875, 0x2e, },
+	{ 82, 5875, 4750, 0x2d, },
+	{ 83, 5750, 4625, 0x2c, },
+	{ 84, 5625, 4500, 0x2b, },
+	{ 85, 5500, 4375, 0x2a, },
+	{ 86, 5375, 4250, 0x29, },
+	{ 87, 5250, 4125, 0x28, },
+	{ 88, 5125, 4000, 0x27, },
+	{ 89, 5000, 3875, 0x26, },
+	{ 90, 4875, 3750, 0x25, },
+	{ 91, 4750, 3625, 0x24, },
+	{ 92, 4625, 3500, 0x23, },
+	{ 93, 4500, 3375, 0x22, },
+	{ 94, 4375, 3250, 0x21, },
+	{ 95, 4250, 3125, 0x20, },
+	{ 96, 4125, 3000, 0x1f, },
+	{ 97, 4125, 3000, 0x1e, },
+	{ 98, 4125, 3000, 0x1d, },
+	{ 99, 4125, 3000, 0x1c, },
+	{ 100, 4125, 3000, 0x1b, },
+	{ 101, 4125, 3000, 0x1a, },
+	{ 102, 4125, 3000, 0x19, },
+	{ 103, 4125, 3000, 0x18, },
+	{ 104, 4125, 3000, 0x17, },
+	{ 105, 4125, 3000, 0x16, },
+	{ 106, 4125, 3000, 0x15, },
+	{ 107, 4125, 3000, 0x14, },
+	{ 108, 4125, 3000, 0x13, },
+	{ 109, 4125, 3000, 0x12, },
+	{ 110, 4125, 3000, 0x11, },
+	{ 111, 4125, 3000, 0x10, },
+	{ 112, 4125, 3000, 0x0f, },
+	{ 113, 4125, 3000, 0x0e, },
+	{ 114, 4125, 3000, 0x0d, },
+	{ 115, 4125, 3000, 0x0c, },
+	{ 116, 4125, 3000, 0x0b, },
+	{ 117, 4125, 3000, 0x0a, },
+	{ 118, 4125, 3000, 0x09, },
+	{ 119, 4125, 3000, 0x08, },
+	{ 120, 1125, 0, 0x07, },
+	{ 121, 1000, 0, 0x06, },
+	{ 122, 875, 0, 0x05, },
+	{ 123, 750, 0, 0x04, },
+	{ 124, 625, 0, 0x03, },
+	{ 125, 500, 0, 0x02, },
+	{ 126, 375, 0, 0x01, },
+	{ 127, 0, 0, 0x00, },
+};
+
+struct cparams {
+	int i;
+	int t;
+	int m;
+	int c;
+};
+
+static struct cparams cparams[] = {
+	{ 1, 1333, 301, 28664 },
+	{ 1, 1066, 294, 24460 },
+	{ 1, 800, 294, 25192 },
+	{ 0, 1333, 276, 27605 },
+	{ 0, 1066, 276, 27605 },
+	{ 0, 800, 231, 23784 },
+};
+
+unsigned long i915_chipset_val(struct drm_i915_private *dev_priv)
+{
+	u64 total_count, diff, ret;
+	u32 count1, count2, count3, m = 0, c = 0;
+	unsigned long now = jiffies_to_msecs(jiffies), diff1;
+	int i;
+
+	diff1 = now - dev_priv->last_time1;
+
+	count1 = I915_READ(DMIEC);
+	count2 = I915_READ(DDREC);
+	count3 = I915_READ(CSIEC);
+
+	total_count = count1 + count2 + count3;
+
+	/* FIXME: handle per-counter overflow */
+	if (total_count < dev_priv->last_count1) {
+		diff = ~0UL - dev_priv->last_count1;
+		diff += total_count;
+	} else {
+		diff = total_count - dev_priv->last_count1;
+	}
+
+	for (i = 0; i < ARRAY_SIZE(cparams); i++) {
+		if (cparams[i].i == dev_priv->c_m &&
+		    cparams[i].t == dev_priv->r_t) {
+			m = cparams[i].m;
+			c = cparams[i].c;
+			break;
+		}
+	}
+
+	div_u64(diff, diff1);
+	ret = ((m * diff) + c);
+	div_u64(ret, 10);
+
+	dev_priv->last_count1 = total_count;
+	dev_priv->last_time1 = now;
+
+	return ret;
+}
+
+unsigned long i915_mch_val(struct drm_i915_private *dev_priv)
+{
+	unsigned long m, x, b;
+	u32 tsfs;
+
+	tsfs = I915_READ(TSFS);
+
+	m = ((tsfs & TSFS_SLOPE_MASK) >> TSFS_SLOPE_SHIFT);
+	x = I915_READ8(TR1);
+
+	b = tsfs & TSFS_INTR_MASK;
+
+	return ((m * x) / 127) - b;
+}
+
+static unsigned long pvid_to_extvid(struct drm_i915_private *dev_priv, u8 pxvid)
+{
+	unsigned long val = 0;
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(v_table); i++) {
+		if (v_table[i].pvid == pxvid) {
+			if (IS_MOBILE(dev_priv->dev))
+				val = v_table[i].vm;
+			else
+				val = v_table[i].vd;
+		}
+	}
+
+	return val;
+}
+
+void i915_update_gfx_val(struct drm_i915_private *dev_priv)
+{
+	struct timespec now, diff1;
+	u64 diff;
+	unsigned long diffms;
+	u32 count;
+
+	getrawmonotonic(&now);
+	diff1 = timespec_sub(now, dev_priv->last_time2);
+
+	/* Don't divide by 0 */
+	diffms = diff1.tv_sec * 1000 + diff1.tv_nsec / 1000000;
+	if (!diffms)
+		return;
+
+	count = I915_READ(GFXEC);
+
+	if (count < dev_priv->last_count2) {
+		diff = ~0UL - dev_priv->last_count2;
+		diff += count;
+	} else {
+		diff = count - dev_priv->last_count2;
+	}
+
+	dev_priv->last_count2 = count;
+	dev_priv->last_time2 = now;
+
+	/* More magic constants... */
+	diff = diff * 1181;
+	div_u64(diff, diffms * 10);
+	dev_priv->gfx_power = diff;
+}
+
+unsigned long i915_gfx_val(struct drm_i915_private *dev_priv)
+{
+	unsigned long t, corr, state1, corr2, state2;
+	u32 pxvid, ext_v;
+
+	pxvid = I915_READ(PXVFREQ_BASE + (dev_priv->cur_delay * 4));
+	pxvid = (pxvid >> 24) & 0x7f;
+	ext_v = pvid_to_extvid(dev_priv, pxvid);
+
+	state1 = ext_v;
+
+	t = i915_mch_val(dev_priv);
+
+	/* Revel in the empirically derived constants */
+
+	/* Correction factor in 1/100000 units */
+	if (t > 80)
+		corr = ((t * 2349) + 135940);
+	else if (t >= 50)
+		corr = ((t * 964) + 29317);
+	else /* < 50 */
+		corr = ((t * 301) + 1004);
+
+	corr = corr * ((150142 * state1) / 10000 - 78642);
+	corr /= 100000;
+	corr2 = (corr * dev_priv->corr);
+
+	state2 = (corr2 * state1) / 10000;
+	state2 /= 100; /* convert to mW */
+
+	i915_update_gfx_val(dev_priv);
+
+	return dev_priv->gfx_power + state2;
+}
+
+/* Global for IPS driver to get at the current i915 device */
+static struct drm_i915_private *i915_mch_dev;
+/*
+ * Lock protecting IPS related data structures
+ *   - i915_mch_dev
+ *   - dev_priv->max_delay
+ *   - dev_priv->min_delay
+ *   - dev_priv->fmax
+ *   - dev_priv->gpu_busy
+ */
+DEFINE_SPINLOCK(mchdev_lock);
+
+/**
+ * i915_read_mch_val - return value for IPS use
+ *
+ * Calculate and return a value for the IPS driver to use when deciding whether
+ * we have thermal and power headroom to increase CPU or GPU power budget.
+ */
+unsigned long i915_read_mch_val(void)
+{
+  	struct drm_i915_private *dev_priv;
+	unsigned long chipset_val, graphics_val, ret = 0;
+
+  	spin_lock(&mchdev_lock);
+	if (!i915_mch_dev)
+		goto out_unlock;
+	dev_priv = i915_mch_dev;
+
+	chipset_val = i915_chipset_val(dev_priv);
+	graphics_val = i915_gfx_val(dev_priv);
+
+	ret = chipset_val + graphics_val;
+
+out_unlock:
+  	spin_unlock(&mchdev_lock);
+
+  	return ret;
+}
+EXPORT_SYMBOL_GPL(i915_read_mch_val);
+
+/**
+ * i915_gpu_raise - raise GPU frequency limit
+ *
+ * Raise the limit; IPS indicates we have thermal headroom.
+ */
+bool i915_gpu_raise(void)
+{
+  	struct drm_i915_private *dev_priv;
+	bool ret = true;
+
+  	spin_lock(&mchdev_lock);
+	if (!i915_mch_dev) {
+		ret = false;
+		goto out_unlock;
+	}
+	dev_priv = i915_mch_dev;
+
+	if (dev_priv->max_delay > dev_priv->fmax)
+		dev_priv->max_delay--;
+
+out_unlock:
+  	spin_unlock(&mchdev_lock);
+
+  	return ret;
+}
+EXPORT_SYMBOL_GPL(i915_gpu_raise);
+
+/**
+ * i915_gpu_lower - lower GPU frequency limit
+ *
+ * IPS indicates we're close to a thermal limit, so throttle back the GPU
+ * frequency maximum.
+ */
+bool i915_gpu_lower(void)
+{
+  	struct drm_i915_private *dev_priv;
+	bool ret = true;
+
+  	spin_lock(&mchdev_lock);
+	if (!i915_mch_dev) {
+		ret = false;
+		goto out_unlock;
+	}
+	dev_priv = i915_mch_dev;
+
+	if (dev_priv->max_delay < dev_priv->min_delay)
+		dev_priv->max_delay++;
+
+out_unlock:
+  	spin_unlock(&mchdev_lock);
+
+  	return ret;
 }
+EXPORT_SYMBOL_GPL(i915_gpu_lower);
+
+/**
+ * i915_gpu_busy - indicate GPU business to IPS
+ *
+ * Tell the IPS driver whether or not the GPU is busy.
+ */
+bool i915_gpu_busy(void)
+{
+  	struct drm_i915_private *dev_priv;
+	bool ret = false;
+
+  	spin_lock(&mchdev_lock);
+	if (!i915_mch_dev)
+		goto out_unlock;
+	dev_priv = i915_mch_dev;
+
+	ret = dev_priv->busy;
+
+out_unlock:
+  	spin_unlock(&mchdev_lock);
+
+  	return ret;
+}
+EXPORT_SYMBOL_GPL(i915_gpu_busy);
+
+/**
+ * i915_gpu_turbo_disable - disable graphics turbo
+ *
+ * Disable graphics turbo by resetting the max frequency and setting the
+ * current frequency to the default.
+ */
+bool i915_gpu_turbo_disable(void)
+{
+  	struct drm_i915_private *dev_priv;
+	bool ret = true;
+
+  	spin_lock(&mchdev_lock);
+	if (!i915_mch_dev) {
+		ret = false;
+		goto out_unlock;
+	}
+	dev_priv = i915_mch_dev;
+
+	dev_priv->max_delay = dev_priv->fstart;
+
+	if (!ironlake_set_drps(dev_priv->dev, dev_priv->fstart))
+		ret = false;
+
+out_unlock:
+  	spin_unlock(&mchdev_lock);
+
+  	return ret;
+}
+EXPORT_SYMBOL_GPL(i915_gpu_turbo_disable);
 
 /**
  * i915_driver_load - setup chip and create an initial config
@@ -1594,7 +2040,6 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)
 	resource_size_t base, size;
 	int ret = 0, mmio_bar;
 	uint32_t agp_size, prealloc_size, prealloc_start;
-
 	/* i915 has 4 more counters */
 	dev->counters += 4;
 	dev->types[6] = _DRM_STAT_IRQ;
@@ -1672,6 +2117,13 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)
 		dev_priv->has_gem = 0;
 	}
 
+	if (dev_priv->has_gem == 0 &&
+	    drm_core_check_feature(dev, DRIVER_MODESET)) {
+		DRM_ERROR("kernel modesetting requires GEM, disabling driver.\n");
+		ret = -ENODEV;
+		goto out_iomapfree;
+	}
+
 	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_IRONLAKE(dev) || IS_GEN6(dev)) {
@@ -1691,7 +2143,10 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)
 			goto out_workqueue_free;
 	}
 
-	i915_get_mem_freq(dev);
+	if (IS_PINEVIEW(dev))
+		i915_pineview_get_mem_freq(dev);
+	else if (IS_IRONLAKE(dev))
+		i915_ironlake_get_mem_freq(dev);
 
 	/* On the 945G/GM, the chipset reports the MSI capability on the
 	 * integrated graphics even though the support isn't actually there
@@ -1709,7 +2164,6 @@ 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);
@@ -1738,6 +2192,12 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)
 
 	setup_timer(&dev_priv->hangcheck_timer, i915_hangcheck_elapsed,
 		    (unsigned long) dev);
+
+	spin_lock(&mchdev_lock);
+	i915_mch_dev = dev_priv;
+	dev_priv->mchdev_lock = &mchdev_lock;
+	spin_unlock(&mchdev_lock);
+
 	return 0;
 
 out_workqueue_free:
@@ -1759,6 +2219,10 @@ int i915_driver_unload(struct drm_device *dev)
 
 	i915_destroy_error_state(dev);
 
+	spin_lock(&mchdev_lock);
+	i915_mch_dev = NULL;
+	spin_unlock(&mchdev_lock);
+
 	destroy_workqueue(dev_priv->wq);
 	del_timer_sync(&dev_priv->hangcheck_timer);
 
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index 5c51e45ab68d..423dc90c1e20 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -60,95 +60,95 @@ extern int intel_agp_enabled;
 	.subdevice = PCI_ANY_ID,		\
 	.driver_data = (unsigned long) info }
 
-const static struct intel_device_info intel_i830_info = {
+static const 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 = {
+static const struct intel_device_info intel_845g_info = {
 	.is_i8xx = 1,
 };
 
-const static struct intel_device_info intel_i85x_info = {
+static const struct intel_device_info intel_i85x_info = {
 	.is_i8xx = 1, .is_i85x = 1, .is_mobile = 1,
 	.cursor_needs_physical = 1,
 };
 
-const static struct intel_device_info intel_i865g_info = {
+static const struct intel_device_info intel_i865g_info = {
 	.is_i8xx = 1,
 };
 
-const static struct intel_device_info intel_i915g_info = {
+static const 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 = {
+static const struct intel_device_info intel_i915gm_info = {
 	.is_i9xx = 1,  .is_mobile = 1,
 	.cursor_needs_physical = 1,
 };
-const static struct intel_device_info intel_i945g_info = {
+static const 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 = {
+static const struct intel_device_info intel_i945gm_info = {
 	.is_i945gm = 1, .is_i9xx = 1, .is_mobile = 1,
 	.has_hotplug = 1, .cursor_needs_physical = 1,
 };
 
-const static struct intel_device_info intel_i965g_info = {
+static const 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 = {
+static const 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 = {
+static const 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 = {
+static const 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 = {
+static const 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 = {
+static const 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 = {
+static const 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 = {
+static const 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 = {
+static const struct intel_device_info intel_sandybridge_d_info = {
 	.is_i965g = 1, .is_i9xx = 1, .need_gfx_hws = 1,
 	.has_hotplug = 1, .is_gen6 = 1,
 };
 
-const static struct intel_device_info intel_sandybridge_m_info = {
+static const struct intel_device_info intel_sandybridge_m_info = {
 	.is_i965g = 1, .is_mobile = 1, .is_i9xx = 1, .need_gfx_hws = 1,
 	.has_hotplug = 1, .is_gen6 = 1,
 };
 
-const static struct pci_device_id pciidlist[] = {
+static const 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),
@@ -340,7 +340,7 @@ int i965_reset(struct drm_device *dev, u8 flags)
 	/*
 	 * Clear request list
 	 */
-	i915_gem_retire_requests(dev);
+	i915_gem_retire_requests(dev, &dev_priv->render_ring);
 
 	if (need_display)
 		i915_save_display(dev);
@@ -370,6 +370,7 @@ int i965_reset(struct drm_device *dev, u8 flags)
 		}
 	} else {
 		DRM_ERROR("Error occurred. Don't know how to reset this chip.\n");
+		mutex_unlock(&dev->struct_mutex);
 		return -ENODEV;
 	}
 
@@ -388,33 +389,10 @@ int i965_reset(struct drm_device *dev, u8 flags)
 	 * switched away).
 	 */
 	if (drm_core_check_feature(dev, DRIVER_MODESET) ||
-	    !dev_priv->mm.suspended) {
-		drm_i915_ring_buffer_t *ring = &dev_priv->ring;
-		struct drm_gem_object *obj = ring->ring_obj;
-		struct drm_i915_gem_object *obj_priv = to_intel_bo(obj);
+			!dev_priv->mm.suspended) {
+		struct intel_ring_buffer *ring = &dev_priv->render_ring;
 		dev_priv->mm.suspended = 0;
-
-		/* Stop the ring if it's running. */
-		I915_WRITE(PRB0_CTL, 0);
-		I915_WRITE(PRB0_TAIL, 0);
-		I915_WRITE(PRB0_HEAD, 0);
-
-		/* Initialize the ring. */
-		I915_WRITE(PRB0_START, obj_priv->gtt_offset);
-		I915_WRITE(PRB0_CTL,
-			   ((obj->size - 4096) & RING_NR_PAGES) |
-			   RING_NO_REPORT |
-			   RING_VALID);
-		if (!drm_core_check_feature(dev, DRIVER_MODESET))
-			i915_kernel_lost_context(dev);
-		else {
-			ring->head = I915_READ(PRB0_HEAD) & HEAD_ADDR;
-			ring->tail = I915_READ(PRB0_TAIL) & TAIL_ADDR;
-			ring->space = ring->head - (ring->tail + 8);
-			if (ring->space < 0)
-				ring->space += ring->Size;
-		}
-
+		ring->init(dev, ring);
 		mutex_unlock(&dev->struct_mutex);
 		drm_irq_uninstall(dev);
 		drm_irq_install(dev);
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 7f797ef1ab39..276583159847 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -32,6 +32,7 @@
 
 #include "i915_reg.h"
 #include "intel_bios.h"
+#include "intel_ringbuffer.h"
 #include <linux/io-mapping.h>
 
 /* General customization:
@@ -55,6 +56,8 @@ enum plane {
 
 #define I915_NUM_PIPE	2
 
+#define I915_GEM_GPU_DOMAINS	(~(I915_GEM_DOMAIN_CPU | I915_GEM_DOMAIN_GTT))
+
 /* Interface history:
  *
  * 1.1: Original.
@@ -89,16 +92,6 @@ struct drm_i915_gem_phys_object {
 	struct drm_gem_object *cur_obj;
 };
 
-typedef struct _drm_i915_ring_buffer {
-	unsigned long Size;
-	u8 *virtual_start;
-	int head;
-	int tail;
-	int space;
-	drm_local_map_t map;
-	struct drm_gem_object *ring_obj;
-} drm_i915_ring_buffer_t;
-
 struct mem_block {
 	struct mem_block *next;
 	struct mem_block *prev;
@@ -241,17 +234,15 @@ typedef struct drm_i915_private {
 	void __iomem *regs;
 
 	struct pci_dev *bridge_dev;
-	drm_i915_ring_buffer_t ring;
+	struct intel_ring_buffer render_ring;
+	struct intel_ring_buffer bsd_ring;
 
 	drm_dma_handle_t *status_page_dmah;
-	void *hw_status_page;
 	void *seqno_page;
 	dma_addr_t dma_status_page;
 	uint32_t counter;
-	unsigned int status_gfx_addr;
 	unsigned int seqno_gfx_addr;
 	drm_local_map_t hws_map;
-	struct drm_gem_object *hws_obj;
 	struct drm_gem_object *seqno_obj;
 	struct drm_gem_object *pwrctx;
 
@@ -267,8 +258,6 @@ typedef struct drm_i915_private {
 	atomic_t irq_received;
 	/** Protects user_irq_refcount and irq_mask_reg */
 	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;
@@ -289,6 +278,7 @@ typedef struct drm_i915_private {
 	struct mem_block *agp_heap;
 	unsigned int sr01, adpa, ppcr, dvob, dvoc, lvds;
 	int vblank_pipe;
+	int num_pipe;
 
 	/* For hangcheck timer */
 #define DRM_I915_HANGCHECK_PERIOD 75 /* in jiffies */
@@ -334,7 +324,7 @@ typedef struct drm_i915_private {
 	int fence_reg_start; /* 4 if userland hasn't ioctl'd us yet */
 	int num_fence_regs; /* 8 on pre-965, 16 otherwise */
 
-	unsigned int fsb_freq, mem_freq;
+	unsigned int fsb_freq, mem_freq, is_ddr3;
 
 	spinlock_t error_lock;
 	struct drm_i915_error_state *first_error;
@@ -514,18 +504,7 @@ typedef struct drm_i915_private {
 		 */
 		struct list_head shrink_list;
 
-		/**
-		 * List of objects currently involved in rendering from the
-		 * ringbuffer.
-		 *
-		 * Includes buffers having the contents of their GPU caches
-		 * flushed, not necessarily primitives.  last_rendering_seqno
-		 * represents when the rendering involved will be completed.
-		 *
-		 * A reference is held on the buffer while on this list.
-		 */
 		spinlock_t active_list_lock;
-		struct list_head active_list;
 
 		/**
 		 * List of objects which are not in the ringbuffer but which
@@ -563,12 +542,6 @@ typedef struct drm_i915_private {
 		struct list_head fence_list;
 
 		/**
-		 * List of breadcrumbs associated with GPU requests currently
-		 * outstanding.
-		 */
-		struct list_head request_list;
-
-		/**
 		 * We leave the user IRQ off as much as possible,
 		 * but this means that requests will finish and never
 		 * be retired once the system goes idle. Set a timer to
@@ -644,6 +617,18 @@ typedef struct drm_i915_private {
 	u8 cur_delay;
 	u8 min_delay;
 	u8 max_delay;
+	u8 fmax;
+	u8 fstart;
+
+ 	u64 last_count1;
+ 	unsigned long last_time1;
+ 	u64 last_count2;
+ 	struct timespec last_time2;
+ 	unsigned long gfx_power;
+ 	int c_m;
+ 	int r_t;
+ 	u8 corr;
+	spinlock_t *mchdev_lock;
 
 	enum no_fbc_reason no_fbc_reason;
 
@@ -671,19 +656,64 @@ struct drm_i915_gem_object {
 	 * (has pending rendering), and is not set if it's on inactive (ready
 	 * to be unbound).
 	 */
-	int active;
+	unsigned int active : 1;
 
 	/**
 	 * This is set if the object has been written to since last bound
 	 * to the GTT
 	 */
-	int dirty;
+	unsigned int dirty : 1;
+
+	/**
+	 * Fence register bits (if any) for this object.  Will be set
+	 * as needed when mapped into the GTT.
+	 * Protected by dev->struct_mutex.
+	 *
+	 * Size: 4 bits for 16 fences + sign (for FENCE_REG_NONE)
+	 */
+	int fence_reg : 5;
+
+	/**
+	 * Used for checking the object doesn't appear more than once
+	 * in an execbuffer object list.
+	 */
+	unsigned int in_execbuffer : 1;
+
+	/**
+	 * Advice: are the backing pages purgeable?
+	 */
+	unsigned int madv : 2;
+
+	/**
+	 * Refcount for the pages array. With the current locking scheme, there
+	 * are at most two concurrent users: Binding a bo to the gtt and
+	 * pwrite/pread using physical addresses. So two bits for a maximum
+	 * of two users are enough.
+	 */
+	unsigned int pages_refcount : 2;
+#define DRM_I915_GEM_OBJECT_MAX_PAGES_REFCOUNT 0x3
+
+	/**
+	 * Current tiling mode for the object.
+	 */
+	unsigned int tiling_mode : 2;
+
+	/** How many users have pinned this object in GTT space. The following
+	 * users can each hold at most one reference: pwrite/pread, pin_ioctl
+	 * (via user_pin_count), execbuffer (objects are not allowed multiple
+	 * times for the same batchbuffer), and the framebuffer code. When
+	 * switching/pageflipping, the framebuffer code has at most two buffers
+	 * pinned per crtc.
+	 *
+	 * In the worst case this is 1 + 1 + 1 + 2*2 = 7. That would fit into 3
+	 * bits with absolutely no headroom. So use 4 bits. */
+	int pin_count : 4;
+#define DRM_I915_GEM_OBJECT_MAX_PIN_COUNT 0xf
 
 	/** AGP memory structure for our GTT binding. */
 	DRM_AGP_MEM *agp_mem;
 
 	struct page **pages;
-	int pages_refcount;
 
 	/**
 	 * Current offset of the object in GTT space.
@@ -692,26 +722,18 @@ struct drm_i915_gem_object {
 	 */
 	uint32_t gtt_offset;
 
+	/* Which ring is refering to is this object */
+	struct intel_ring_buffer *ring;
+
 	/**
 	 * Fake offset for use by mmap(2)
 	 */
 	uint64_t mmap_offset;
 
-	/**
-	 * Fence register bits (if any) for this object.  Will be set
-	 * as needed when mapped into the GTT.
-	 * Protected by dev->struct_mutex.
-	 */
-	int fence_reg;
-
-	/** How many users have pinned this object in GTT space */
-	int pin_count;
-
 	/** Breadcrumb of last rendering to the buffer. */
 	uint32_t last_rendering_seqno;
 
-	/** Current tiling mode for the object. */
-	uint32_t tiling_mode;
+	/** Current tiling stride for the object, if it's tiled. */
 	uint32_t stride;
 
 	/** Record of address bit 17 of each page at last unbind. */
@@ -734,17 +756,6 @@ struct drm_i915_gem_object {
 	struct drm_i915_gem_phys_object *phys_obj;
 
 	/**
-	 * Used for checking the object doesn't appear more than once
-	 * in an execbuffer object list.
-	 */
-	int in_execbuffer;
-
-	/**
-	 * 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.
@@ -765,6 +776,9 @@ struct drm_i915_gem_object {
  * an emission time with seqnos for tracking how far ahead of the GPU we are.
  */
 struct drm_i915_gem_request {
+	/** On Which ring this request was generated */
+	struct intel_ring_buffer *ring;
+
 	/** GEM sequence number associated with this request. */
 	uint32_t seqno;
 
@@ -821,6 +835,11 @@ extern int i915_emit_box(struct drm_device *dev,
 			 struct drm_clip_rect *boxes,
 			 int i, int DR1, int DR4);
 extern int i965_reset(struct drm_device *dev, u8 flags);
+extern unsigned long i915_chipset_val(struct drm_i915_private *dev_priv);
+extern unsigned long i915_mch_val(struct drm_i915_private *dev_priv);
+extern unsigned long i915_gfx_val(struct drm_i915_private *dev_priv);
+extern void i915_update_gfx_val(struct drm_i915_private *dev_priv);
+
 
 /* i915_irq.c */
 void i915_hangcheck_elapsed(unsigned long data);
@@ -829,9 +848,7 @@ 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);
 
 extern irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS);
@@ -849,6 +866,11 @@ extern u32 gm45_get_vblank_counter(struct drm_device *dev, int crtc);
 extern int i915_vblank_swap(struct drm_device *dev, void *data,
 			    struct drm_file *file_priv);
 extern void i915_enable_irq(drm_i915_private_t *dev_priv, u32 mask);
+extern void i915_disable_irq(drm_i915_private_t *dev_priv, u32 mask);
+extern void ironlake_enable_graphics_irq(drm_i915_private_t *dev_priv,
+		u32 mask);
+extern void ironlake_disable_graphics_irq(drm_i915_private_t *dev_priv,
+		u32 mask);
 
 void
 i915_enable_pipestat(drm_i915_private_t *dev_priv, int pipe, u32 mask);
@@ -922,11 +944,13 @@ void i915_gem_object_unpin(struct drm_gem_object *obj);
 int i915_gem_object_unbind(struct drm_gem_object *obj);
 void i915_gem_release_mmap(struct drm_gem_object *obj);
 void i915_gem_lastclose(struct drm_device *dev);
-uint32_t i915_get_gem_seqno(struct drm_device *dev);
+uint32_t i915_get_gem_seqno(struct drm_device *dev,
+		struct intel_ring_buffer *ring);
 bool i915_seqno_passed(uint32_t seq1, uint32_t seq2);
 int i915_gem_object_get_fence_reg(struct drm_gem_object *obj);
 int i915_gem_object_put_fence_reg(struct drm_gem_object *obj);
-void i915_gem_retire_requests(struct drm_device *dev);
+void i915_gem_retire_requests(struct drm_device *dev,
+		 struct intel_ring_buffer *ring);
 void i915_gem_retire_work_handler(struct work_struct *work);
 void i915_gem_clflush_object(struct drm_gem_object *obj);
 int i915_gem_object_set_domain(struct drm_gem_object *obj,
@@ -937,9 +961,13 @@ 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);
+uint32_t i915_add_request(struct drm_device *dev,
+		struct drm_file *file_priv,
+		uint32_t flush_domains,
+		struct intel_ring_buffer *ring);
+int i915_do_wait_request(struct drm_device *dev,
+		uint32_t seqno, int interruptible,
+		struct intel_ring_buffer *ring);
 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);
@@ -1015,7 +1043,7 @@ extern void g4x_disable_fbc(struct drm_device *dev);
 extern void intel_disable_fbc(struct drm_device *dev);
 extern void intel_enable_fbc(struct drm_crtc *crtc, unsigned long interval);
 extern bool intel_fbc_enabled(struct drm_device *dev);
-
+extern bool ironlake_set_drps(struct drm_device *dev, u8 val);
 extern void intel_detect_pch (struct drm_device *dev);
 extern int intel_trans_dp_port_sel (struct drm_crtc *crtc);
 
@@ -1026,7 +1054,8 @@ extern int intel_trans_dp_port_sel (struct drm_crtc *crtc);
  * has access to the ring.
  */
 #define RING_LOCK_TEST_WITH_RETURN(dev, file_priv) do {			\
-	if (((drm_i915_private_t *)dev->dev_private)->ring.ring_obj == NULL) \
+	if (((drm_i915_private_t *)dev->dev_private)->render_ring.gem_object \
+			== NULL)					\
 		LOCK_TEST_WITH_RETURN(dev, file_priv);			\
 } while (0)
 
@@ -1039,35 +1068,31 @@ extern int intel_trans_dp_port_sel (struct drm_crtc *crtc);
 #define I915_WRITE64(reg, val)	writeq(val, dev_priv->regs + (reg))
 #define I915_READ64(reg)	readq(dev_priv->regs + (reg))
 #define POSTING_READ(reg)	(void)I915_READ(reg)
+#define POSTING_READ16(reg)	(void)I915_READ16(reg)
 
 #define I915_VERBOSE 0
 
-#define RING_LOCALS	volatile unsigned int *ring_virt__;
-
-#define BEGIN_LP_RING(n) do {						\
-	int bytes__ = 4*(n);						\
-	if (I915_VERBOSE) DRM_DEBUG("BEGIN_LP_RING(%d)\n", (n));	\
-	/* a wrap must occur between instructions so pad beforehand */	\
-	if (unlikely (dev_priv->ring.tail + bytes__ > dev_priv->ring.Size)) \
-		i915_wrap_ring(dev);					\
-	if (unlikely (dev_priv->ring.space < bytes__))			\
-		i915_wait_ring(dev, bytes__, __func__);			\
-	ring_virt__ = (unsigned int *)					\
-	        (dev_priv->ring.virtual_start + dev_priv->ring.tail);	\
-	dev_priv->ring.tail += bytes__;					\
-	dev_priv->ring.tail &= dev_priv->ring.Size - 1;			\
-	dev_priv->ring.space -= bytes__;				\
+#define BEGIN_LP_RING(n)  do { \
+	drm_i915_private_t *dev_priv = dev->dev_private;                \
+	if (I915_VERBOSE)						\
+		DRM_DEBUG("   BEGIN_LP_RING %x\n", (int)(n));		\
+	intel_ring_begin(dev, &dev_priv->render_ring, 4*(n));		\
 } while (0)
 
-#define OUT_RING(n) do {						\
-	if (I915_VERBOSE) DRM_DEBUG("   OUT_RING %x\n", (int)(n));	\
-	*ring_virt__++ = (n);						\
+
+#define OUT_RING(x) do {						\
+	drm_i915_private_t *dev_priv = dev->dev_private;		\
+	if (I915_VERBOSE)						\
+		DRM_DEBUG("   OUT_RING %x\n", (int)(x));		\
+	intel_ring_emit(dev, &dev_priv->render_ring, x);		\
 } while (0)
 
 #define ADVANCE_LP_RING() do {						\
+	drm_i915_private_t *dev_priv = dev->dev_private;                \
 	if (I915_VERBOSE)						\
-		DRM_DEBUG("ADVANCE_LP_RING %x\n", dev_priv->ring.tail);	\
-	I915_WRITE(PRB0_TAIL, dev_priv->ring.tail);			\
+		DRM_DEBUG("ADVANCE_LP_RING %x\n",			\
+				dev_priv->render_ring.tail);		\
+	intel_ring_advance(dev, &dev_priv->render_ring);		\
 } while(0)
 
 /**
@@ -1085,14 +1110,12 @@ extern int intel_trans_dp_port_sel (struct drm_crtc *crtc);
  *
  * The area from dword 0x20 to 0x3ff is available for driver usage.
  */
-#define READ_HWSP(dev_priv, reg)  (((volatile u32*)(dev_priv->hw_status_page))[reg])
+#define READ_HWSP(dev_priv, reg)  (((volatile u32 *)\
+			(dev_priv->render_ring.status_page.page_addr))[reg])
 #define READ_BREADCRUMB(dev_priv) READ_HWSP(dev_priv, I915_BREADCRUMB_INDEX)
 #define I915_GEM_HWS_INDEX		0x20
 #define I915_BREADCRUMB_INDEX		0x21
 
-extern int i915_wrap_ring(struct drm_device * dev);
-extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller);
-
 #define INTEL_INFO(dev)	(((struct drm_i915_private *) (dev)->dev_private)->info)
 
 #define IS_I830(dev)		((dev)->pci_device == 0x3577)
@@ -1138,6 +1161,7 @@ extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller);
 			 (dev)->pci_device == 0x2A42 ||		\
 			 (dev)->pci_device == 0x2E42)
 
+#define HAS_BSD(dev)            (IS_IRONLAKE(dev) || IS_G4X(dev))
 #define I915_NEED_GFX_HWS(dev)	(INTEL_INFO(dev)->need_gfx_hws)
 
 /* With the 945 and later, Y tiling got adjusted so that it was 32 128-byte
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 112699f71fa4..9ded3dae6c87 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -35,8 +35,6 @@
 #include <linux/swap.h>
 #include <linux/pci.h>
 
-#define I915_GEM_GPU_DOMAINS	(~(I915_GEM_DOMAIN_CPU | I915_GEM_DOMAIN_GTT))
-
 static void i915_gem_object_flush_gpu_write_domain(struct drm_gem_object *obj);
 static void i915_gem_object_flush_gtt_write_domain(struct drm_gem_object *obj);
 static void i915_gem_object_flush_cpu_write_domain(struct drm_gem_object *obj);
@@ -169,7 +167,7 @@ static int i915_gem_object_needs_bit17_swizzle(struct drm_gem_object *obj)
 		obj_priv->tiling_mode != I915_TILING_NONE;
 }
 
-static inline int
+static inline void
 slow_shmem_copy(struct page *dst_page,
 		int dst_offset,
 		struct page *src_page,
@@ -178,25 +176,16 @@ slow_shmem_copy(struct page *dst_page,
 {
 	char *dst_vaddr, *src_vaddr;
 
-	dst_vaddr = kmap_atomic(dst_page, KM_USER0);
-	if (dst_vaddr == NULL)
-		return -ENOMEM;
-
-	src_vaddr = kmap_atomic(src_page, KM_USER1);
-	if (src_vaddr == NULL) {
-		kunmap_atomic(dst_vaddr, KM_USER0);
-		return -ENOMEM;
-	}
+	dst_vaddr = kmap(dst_page);
+	src_vaddr = kmap(src_page);
 
 	memcpy(dst_vaddr + dst_offset, src_vaddr + src_offset, length);
 
-	kunmap_atomic(src_vaddr, KM_USER1);
-	kunmap_atomic(dst_vaddr, KM_USER0);
-
-	return 0;
+	kunmap(src_page);
+	kunmap(dst_page);
 }
 
-static inline int
+static inline void
 slow_shmem_bit17_copy(struct page *gpu_page,
 		      int gpu_offset,
 		      struct page *cpu_page,
@@ -216,15 +205,8 @@ slow_shmem_bit17_copy(struct page *gpu_page,
 					       cpu_page, cpu_offset, length);
 	}
 
-	gpu_vaddr = kmap_atomic(gpu_page, KM_USER0);
-	if (gpu_vaddr == NULL)
-		return -ENOMEM;
-
-	cpu_vaddr = kmap_atomic(cpu_page, KM_USER1);
-	if (cpu_vaddr == NULL) {
-		kunmap_atomic(gpu_vaddr, KM_USER0);
-		return -ENOMEM;
-	}
+	gpu_vaddr = kmap(gpu_page);
+	cpu_vaddr = kmap(cpu_page);
 
 	/* Copy the data, XORing A6 with A17 (1). The user already knows he's
 	 * XORing with the other bits (A9 for Y, A9 and A10 for X)
@@ -248,10 +230,8 @@ slow_shmem_bit17_copy(struct page *gpu_page,
 		length -= this_length;
 	}
 
-	kunmap_atomic(cpu_vaddr, KM_USER1);
-	kunmap_atomic(gpu_vaddr, KM_USER0);
-
-	return 0;
+	kunmap(cpu_page);
+	kunmap(gpu_page);
 }
 
 /**
@@ -427,21 +407,19 @@ i915_gem_shmem_pread_slow(struct drm_device *dev, struct drm_gem_object *obj,
 			page_length = PAGE_SIZE - data_page_offset;
 
 		if (do_bit17_swizzling) {
-			ret = slow_shmem_bit17_copy(obj_priv->pages[shmem_page_index],
-						    shmem_page_offset,
-						    user_pages[data_page_index],
-						    data_page_offset,
-						    page_length,
-						    1);
-		} else {
-			ret = slow_shmem_copy(user_pages[data_page_index],
-					      data_page_offset,
-					      obj_priv->pages[shmem_page_index],
+			slow_shmem_bit17_copy(obj_priv->pages[shmem_page_index],
 					      shmem_page_offset,
-					      page_length);
+					      user_pages[data_page_index],
+					      data_page_offset,
+					      page_length,
+					      1);
+		} else {
+			slow_shmem_copy(user_pages[data_page_index],
+					data_page_offset,
+					obj_priv->pages[shmem_page_index],
+					shmem_page_offset,
+					page_length);
 		}
-		if (ret)
-			goto fail_put_pages;
 
 		remain -= page_length;
 		data_ptr += page_length;
@@ -531,25 +509,24 @@ fast_user_write(struct io_mapping *mapping,
  * page faults
  */
 
-static inline int
+static inline void
 slow_kernel_write(struct io_mapping *mapping,
 		  loff_t gtt_base, int gtt_offset,
 		  struct page *user_page, int user_offset,
 		  int length)
 {
-	char *src_vaddr, *dst_vaddr;
-	unsigned long unwritten;
+	char __iomem *dst_vaddr;
+	char *src_vaddr;
 
-	dst_vaddr = io_mapping_map_atomic_wc(mapping, gtt_base);
-	src_vaddr = kmap_atomic(user_page, KM_USER1);
-	unwritten = __copy_from_user_inatomic_nocache(dst_vaddr + gtt_offset,
-						      src_vaddr + user_offset,
-						      length);
-	kunmap_atomic(src_vaddr, KM_USER1);
-	io_mapping_unmap_atomic(dst_vaddr);
-	if (unwritten)
-		return -EFAULT;
-	return 0;
+	dst_vaddr = io_mapping_map_wc(mapping, gtt_base);
+	src_vaddr = kmap(user_page);
+
+	memcpy_toio(dst_vaddr + gtt_offset,
+		    src_vaddr + user_offset,
+		    length);
+
+	kunmap(user_page);
+	io_mapping_unmap(dst_vaddr);
 }
 
 static inline int
@@ -722,18 +699,11 @@ i915_gem_gtt_pwrite_slow(struct drm_device *dev, struct drm_gem_object *obj,
 		if ((data_page_offset + page_length) > PAGE_SIZE)
 			page_length = PAGE_SIZE - data_page_offset;
 
-		ret = slow_kernel_write(dev_priv->mm.gtt_mapping,
-					gtt_page_base, gtt_page_offset,
-					user_pages[data_page_index],
-					data_page_offset,
-					page_length);
-
-		/* If we get a fault while copying data, then (presumably) our
-		 * source page isn't available.  Return the error and we'll
-		 * retry in the slow path.
-		 */
-		if (ret)
-			goto out_unpin_object;
+		slow_kernel_write(dev_priv->mm.gtt_mapping,
+				  gtt_page_base, gtt_page_offset,
+				  user_pages[data_page_index],
+				  data_page_offset,
+				  page_length);
 
 		remain -= page_length;
 		offset += page_length;
@@ -902,21 +872,19 @@ i915_gem_shmem_pwrite_slow(struct drm_device *dev, struct drm_gem_object *obj,
 			page_length = PAGE_SIZE - data_page_offset;
 
 		if (do_bit17_swizzling) {
-			ret = slow_shmem_bit17_copy(obj_priv->pages[shmem_page_index],
-						    shmem_page_offset,
-						    user_pages[data_page_index],
-						    data_page_offset,
-						    page_length,
-						    0);
-		} else {
-			ret = slow_shmem_copy(obj_priv->pages[shmem_page_index],
+			slow_shmem_bit17_copy(obj_priv->pages[shmem_page_index],
 					      shmem_page_offset,
 					      user_pages[data_page_index],
 					      data_page_offset,
-					      page_length);
+					      page_length,
+					      0);
+		} else {
+			slow_shmem_copy(obj_priv->pages[shmem_page_index],
+					shmem_page_offset,
+					user_pages[data_page_index],
+					data_page_offset,
+					page_length);
 		}
-		if (ret)
-			goto fail_put_pages;
 
 		remain -= page_length;
 		data_ptr += page_length;
@@ -973,7 +941,8 @@ i915_gem_pwrite_ioctl(struct drm_device *dev, void *data,
 	if (obj_priv->phys_obj)
 		ret = i915_gem_phys_pwrite(dev, obj, args, file_priv);
 	else if (obj_priv->tiling_mode == I915_TILING_NONE &&
-		 dev->gtt_total != 0) {
+		 dev->gtt_total != 0 &&
+		 obj->write_domain != I915_GEM_DOMAIN_CPU) {
 		ret = i915_gem_gtt_pwrite_fast(dev, obj, args, file_priv);
 		if (ret == -EFAULT) {
 			ret = i915_gem_gtt_pwrite_slow(dev, obj, args,
@@ -1484,11 +1453,14 @@ i915_gem_object_put_pages(struct drm_gem_object *obj)
 }
 
 static void
-i915_gem_object_move_to_active(struct drm_gem_object *obj, uint32_t seqno)
+i915_gem_object_move_to_active(struct drm_gem_object *obj, uint32_t seqno,
+			       struct intel_ring_buffer *ring)
 {
 	struct drm_device *dev = obj->dev;
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	struct drm_i915_gem_object *obj_priv = to_intel_bo(obj);
+	BUG_ON(ring == NULL);
+	obj_priv->ring = ring;
 
 	/* Add a reference if we're newly entering the active list. */
 	if (!obj_priv->active) {
@@ -1497,8 +1469,7 @@ i915_gem_object_move_to_active(struct drm_gem_object *obj, uint32_t seqno)
 	}
 	/* Move from whatever list we were on to the tail of execution. */
 	spin_lock(&dev_priv->mm.active_list_lock);
-	list_move_tail(&obj_priv->list,
-		       &dev_priv->mm.active_list);
+	list_move_tail(&obj_priv->list, &ring->active_list);
 	spin_unlock(&dev_priv->mm.active_list_lock);
 	obj_priv->last_rendering_seqno = seqno;
 }
@@ -1551,6 +1522,7 @@ i915_gem_object_move_to_inactive(struct drm_gem_object *obj)
 	BUG_ON(!list_empty(&obj_priv->gpu_write_list));
 
 	obj_priv->last_rendering_seqno = 0;
+	obj_priv->ring = NULL;
 	if (obj_priv->active) {
 		obj_priv->active = 0;
 		drm_gem_object_unreference(obj);
@@ -1560,7 +1532,8 @@ i915_gem_object_move_to_inactive(struct drm_gem_object *obj)
 
 static void
 i915_gem_process_flushing_list(struct drm_device *dev,
-			       uint32_t flush_domains, uint32_t seqno)
+			       uint32_t flush_domains, uint32_t seqno,
+			       struct intel_ring_buffer *ring)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	struct drm_i915_gem_object *obj_priv, *next;
@@ -1571,12 +1544,13 @@ i915_gem_process_flushing_list(struct drm_device *dev,
 		struct drm_gem_object *obj = &obj_priv->base;
 
 		if ((obj->write_domain & flush_domains) ==
-		    obj->write_domain) {
+		    obj->write_domain &&
+		    obj_priv->ring->ring_flag == ring->ring_flag) {
 			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);
+			i915_gem_object_move_to_active(obj, seqno, ring);
 
 			/* update the fence lru list */
 			if (obj_priv->fence_reg != I915_FENCE_REG_NONE) {
@@ -1593,31 +1567,15 @@ i915_gem_process_flushing_list(struct drm_device *dev,
 	}
 }
 
-#define PIPE_CONTROL_FLUSH(addr)					\
-	OUT_RING(GFX_OP_PIPE_CONTROL | PIPE_CONTROL_QW_WRITE |		\
-		 PIPE_CONTROL_DEPTH_STALL);				\
-	OUT_RING(addr | PIPE_CONTROL_GLOBAL_GTT);			\
-	OUT_RING(0);							\
-	OUT_RING(0);							\
-
-/**
- * Creates a new sequence number, emitting a write of it to the status page
- * plus an interrupt, which will trigger i915_user_interrupt_handler.
- *
- * Must be called with struct_lock held.
- *
- * Returned sequence numbers are nonzero on success.
- */
 uint32_t
 i915_add_request(struct drm_device *dev, struct drm_file *file_priv,
-		 uint32_t flush_domains)
+		 uint32_t flush_domains, struct intel_ring_buffer *ring)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	struct drm_i915_file_private *i915_file_priv = NULL;
 	struct drm_i915_gem_request *request;
 	uint32_t seqno;
 	int was_empty;
-	RING_LOCALS;
 
 	if (file_priv != NULL)
 		i915_file_priv = file_priv->driver_priv;
@@ -1626,62 +1584,14 @@ i915_add_request(struct drm_device *dev, struct drm_file *file_priv,
 	if (request == NULL)
 		return 0;
 
-	/* Grab the seqno we're going to make this request be, and bump the
-	 * next (skipping 0 so it can be the reserved no-seqno value).
-	 */
-	seqno = dev_priv->mm.next_gem_seqno;
-	dev_priv->mm.next_gem_seqno++;
-	if (dev_priv->mm.next_gem_seqno == 0)
-		dev_priv->mm.next_gem_seqno++;
-
-	if (HAS_PIPE_CONTROL(dev)) {
-		u32 scratch_addr = dev_priv->seqno_gfx_addr + 128;
-
-		/*
-		 * Workaround qword write incoherence by flushing the
-		 * PIPE_NOTIFY buffers out to memory before requesting
-		 * an interrupt.
-		 */
-		BEGIN_LP_RING(32);
-		OUT_RING(GFX_OP_PIPE_CONTROL | PIPE_CONTROL_QW_WRITE |
-			 PIPE_CONTROL_WC_FLUSH | PIPE_CONTROL_TC_FLUSH);
-		OUT_RING(dev_priv->seqno_gfx_addr | PIPE_CONTROL_GLOBAL_GTT);
-		OUT_RING(seqno);
-		OUT_RING(0);
-		PIPE_CONTROL_FLUSH(scratch_addr);
-		scratch_addr += 128; /* write to separate cachelines */
-		PIPE_CONTROL_FLUSH(scratch_addr);
-		scratch_addr += 128;
-		PIPE_CONTROL_FLUSH(scratch_addr);
-		scratch_addr += 128;
-		PIPE_CONTROL_FLUSH(scratch_addr);
-		scratch_addr += 128;
-		PIPE_CONTROL_FLUSH(scratch_addr);
-		scratch_addr += 128;
-		PIPE_CONTROL_FLUSH(scratch_addr);
-		OUT_RING(GFX_OP_PIPE_CONTROL | PIPE_CONTROL_QW_WRITE |
-			 PIPE_CONTROL_WC_FLUSH | PIPE_CONTROL_TC_FLUSH |
-			 PIPE_CONTROL_NOTIFY);
-		OUT_RING(dev_priv->seqno_gfx_addr | PIPE_CONTROL_GLOBAL_GTT);
-		OUT_RING(seqno);
-		OUT_RING(0);
-		ADVANCE_LP_RING();
-	} else {
-		BEGIN_LP_RING(4);
-		OUT_RING(MI_STORE_DWORD_INDEX);
-		OUT_RING(I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT);
-		OUT_RING(seqno);
-
-		OUT_RING(MI_USER_INTERRUPT);
-		ADVANCE_LP_RING();
-	}
-
-	DRM_DEBUG_DRIVER("%d\n", seqno);
+	seqno = ring->add_request(dev, ring, file_priv, flush_domains);
 
 	request->seqno = seqno;
+	request->ring = ring;
 	request->emitted_jiffies = jiffies;
-	was_empty = list_empty(&dev_priv->mm.request_list);
-	list_add_tail(&request->list, &dev_priv->mm.request_list);
+	was_empty = list_empty(&ring->request_list);
+	list_add_tail(&request->list, &ring->request_list);
+
 	if (i915_file_priv) {
 		list_add_tail(&request->client_list,
 			      &i915_file_priv->mm.request_list);
@@ -1693,7 +1603,7 @@ i915_add_request(struct drm_device *dev, struct drm_file *file_priv,
 	 * domain we're flushing with our flush.
 	 */
 	if (flush_domains != 0) 
-		i915_gem_process_flushing_list(dev, flush_domains, seqno);
+		i915_gem_process_flushing_list(dev, flush_domains, seqno, ring);
 
 	if (!dev_priv->mm.suspended) {
 		mod_timer(&dev_priv->hangcheck_timer, jiffies + DRM_I915_HANGCHECK_PERIOD);
@@ -1710,20 +1620,16 @@ i915_add_request(struct drm_device *dev, struct drm_file *file_priv,
  * before signalling the CPU
  */
 static uint32_t
-i915_retire_commands(struct drm_device *dev)
+i915_retire_commands(struct drm_device *dev, struct intel_ring_buffer *ring)
 {
-	drm_i915_private_t *dev_priv = dev->dev_private;
-	uint32_t cmd = MI_FLUSH | MI_NO_WRITE_FLUSH;
 	uint32_t flush_domains = 0;
-	RING_LOCALS;
 
 	/* The sampler always gets flushed on i965 (sigh) */
 	if (IS_I965G(dev))
 		flush_domains |= I915_GEM_DOMAIN_SAMPLER;
-	BEGIN_LP_RING(2);
-	OUT_RING(cmd);
-	OUT_RING(0); /* noop */
-	ADVANCE_LP_RING();
+
+	ring->flush(dev, ring,
+			I915_GEM_DOMAIN_COMMAND, flush_domains);
 	return flush_domains;
 }
 
@@ -1743,11 +1649,11 @@ i915_gem_retire_request(struct drm_device *dev,
 	 * by the ringbuffer to the flushing/inactive lists as appropriate.
 	 */
 	spin_lock(&dev_priv->mm.active_list_lock);
-	while (!list_empty(&dev_priv->mm.active_list)) {
+	while (!list_empty(&request->ring->active_list)) {
 		struct drm_gem_object *obj;
 		struct drm_i915_gem_object *obj_priv;
 
-		obj_priv = list_first_entry(&dev_priv->mm.active_list,
+		obj_priv = list_first_entry(&request->ring->active_list,
 					    struct drm_i915_gem_object,
 					    list);
 		obj = &obj_priv->base;
@@ -1794,35 +1700,33 @@ i915_seqno_passed(uint32_t seq1, uint32_t seq2)
 }
 
 uint32_t
-i915_get_gem_seqno(struct drm_device *dev)
+i915_get_gem_seqno(struct drm_device *dev,
+		   struct intel_ring_buffer *ring)
 {
-	drm_i915_private_t *dev_priv = dev->dev_private;
-
-	if (HAS_PIPE_CONTROL(dev))
-		return ((volatile u32 *)(dev_priv->seqno_page))[0];
-	else
-		return READ_HWSP(dev_priv, I915_GEM_HWS_INDEX);
+	return ring->get_gem_seqno(dev, ring);
 }
 
 /**
  * This function clears the request list as sequence numbers are passed.
  */
 void
-i915_gem_retire_requests(struct drm_device *dev)
+i915_gem_retire_requests(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	uint32_t seqno;
 
-	if (!dev_priv->hw_status_page || list_empty(&dev_priv->mm.request_list))
+	if (!ring->status_page.page_addr
+			|| list_empty(&ring->request_list))
 		return;
 
-	seqno = i915_get_gem_seqno(dev);
+	seqno = i915_get_gem_seqno(dev, ring);
 
-	while (!list_empty(&dev_priv->mm.request_list)) {
+	while (!list_empty(&ring->request_list)) {
 		struct drm_i915_gem_request *request;
 		uint32_t retiring_seqno;
 
-		request = list_first_entry(&dev_priv->mm.request_list,
+		request = list_first_entry(&ring->request_list,
 					   struct drm_i915_gem_request,
 					   list);
 		retiring_seqno = request->seqno;
@@ -1840,7 +1744,8 @@ i915_gem_retire_requests(struct drm_device *dev)
 
 	if (unlikely (dev_priv->trace_irq_seqno &&
 		      i915_seqno_passed(dev_priv->trace_irq_seqno, seqno))) {
-		i915_user_irq_put(dev);
+
+		ring->user_irq_put(dev, ring);
 		dev_priv->trace_irq_seqno = 0;
 	}
 }
@@ -1856,15 +1761,22 @@ i915_gem_retire_work_handler(struct work_struct *work)
 	dev = dev_priv->dev;
 
 	mutex_lock(&dev->struct_mutex);
-	i915_gem_retire_requests(dev);
+	i915_gem_retire_requests(dev, &dev_priv->render_ring);
+
+	if (HAS_BSD(dev))
+		i915_gem_retire_requests(dev, &dev_priv->bsd_ring);
+
 	if (!dev_priv->mm.suspended &&
-	    !list_empty(&dev_priv->mm.request_list))
+		(!list_empty(&dev_priv->render_ring.request_list) ||
+			(HAS_BSD(dev) &&
+			 !list_empty(&dev_priv->bsd_ring.request_list))))
 		queue_delayed_work(dev_priv->wq, &dev_priv->mm.retire_work, HZ);
 	mutex_unlock(&dev->struct_mutex);
 }
 
 int
-i915_do_wait_request(struct drm_device *dev, uint32_t seqno, int interruptible)
+i915_do_wait_request(struct drm_device *dev, uint32_t seqno,
+		int interruptible, struct intel_ring_buffer *ring)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	u32 ier;
@@ -1875,7 +1787,7 @@ i915_do_wait_request(struct drm_device *dev, uint32_t seqno, int interruptible)
 	if (atomic_read(&dev_priv->mm.wedged))
 		return -EIO;
 
-	if (!i915_seqno_passed(i915_get_gem_seqno(dev), seqno)) {
+	if (!i915_seqno_passed(ring->get_gem_seqno(dev, ring), seqno)) {
 		if (HAS_PCH_SPLIT(dev))
 			ier = I915_READ(DEIER) | I915_READ(GTIER);
 		else
@@ -1889,19 +1801,21 @@ i915_do_wait_request(struct drm_device *dev, uint32_t seqno, int interruptible)
 
 		trace_i915_gem_request_wait_begin(dev, seqno);
 
-		dev_priv->mm.waiting_gem_seqno = seqno;
-		i915_user_irq_get(dev);
+		ring->waiting_gem_seqno = seqno;
+		ring->user_irq_get(dev, ring);
 		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));
+			ret = wait_event_interruptible(ring->irq_queue,
+				i915_seqno_passed(
+					ring->get_gem_seqno(dev, ring), 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));
+			wait_event(ring->irq_queue,
+				i915_seqno_passed(
+					ring->get_gem_seqno(dev, ring), seqno)
+				|| atomic_read(&dev_priv->mm.wedged));
 
-		i915_user_irq_put(dev);
-		dev_priv->mm.waiting_gem_seqno = 0;
+		ring->user_irq_put(dev, ring);
+		ring->waiting_gem_seqno = 0;
 
 		trace_i915_gem_request_wait_end(dev, seqno);
 	}
@@ -1910,7 +1824,7 @@ i915_do_wait_request(struct drm_device *dev, uint32_t seqno, int interruptible)
 
 	if (ret && ret != -ERESTARTSYS)
 		DRM_ERROR("%s returns %d (awaiting %d at %d)\n",
-			  __func__, ret, seqno, i915_get_gem_seqno(dev));
+			  __func__, ret, seqno, ring->get_gem_seqno(dev, ring));
 
 	/* Directly dispatch request retiring.  While we have the work queue
 	 * to handle this, the waiter on a request often wants an associated
@@ -1918,7 +1832,7 @@ i915_do_wait_request(struct drm_device *dev, uint32_t seqno, int interruptible)
 	 * a separate wait queue to handle that.
 	 */
 	if (ret == 0)
-		i915_gem_retire_requests(dev);
+		i915_gem_retire_requests(dev, ring);
 
 	return ret;
 }
@@ -1928,9 +1842,10 @@ i915_do_wait_request(struct drm_device *dev, uint32_t seqno, int interruptible)
  * request and object lists appropriately for that event.
  */
 static int
-i915_wait_request(struct drm_device *dev, uint32_t seqno)
+i915_wait_request(struct drm_device *dev, uint32_t seqno,
+		struct intel_ring_buffer *ring)
 {
-	return i915_do_wait_request(dev, seqno, 1);
+	return i915_do_wait_request(dev, seqno, 1, ring);
 }
 
 static void
@@ -1939,71 +1854,29 @@ i915_gem_flush(struct drm_device *dev,
 	       uint32_t flush_domains)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
-	uint32_t cmd;
-	RING_LOCALS;
-
-#if WATCH_EXEC
-	DRM_INFO("%s: invalidate %08x flush %08x\n", __func__,
-		  invalidate_domains, flush_domains);
-#endif
-	trace_i915_gem_request_flush(dev, dev_priv->mm.next_gem_seqno,
-				     invalidate_domains, flush_domains);
-
 	if (flush_domains & I915_GEM_DOMAIN_CPU)
 		drm_agp_chipset_flush(dev);
+	dev_priv->render_ring.flush(dev, &dev_priv->render_ring,
+			invalidate_domains,
+			flush_domains);
+
+	if (HAS_BSD(dev))
+		dev_priv->bsd_ring.flush(dev, &dev_priv->bsd_ring,
+				invalidate_domains,
+				flush_domains);
+}
 
-	if ((invalidate_domains | flush_domains) & I915_GEM_GPU_DOMAINS) {
-		/*
-		 * read/write caches:
-		 *
-		 * I915_GEM_DOMAIN_RENDER is always invalidated, but is
-		 * only flushed if MI_NO_WRITE_FLUSH is unset.  On 965, it is
-		 * also flushed at 2d versus 3d pipeline switches.
-		 *
-		 * read-only caches:
-		 *
-		 * I915_GEM_DOMAIN_SAMPLER is flushed on pre-965 if
-		 * MI_READ_FLUSH is set, and is always flushed on 965.
-		 *
-		 * I915_GEM_DOMAIN_COMMAND may not exist?
-		 *
-		 * I915_GEM_DOMAIN_INSTRUCTION, which exists on 965, is
-		 * invalidated when MI_EXE_FLUSH is set.
-		 *
-		 * I915_GEM_DOMAIN_VERTEX, which exists on 965, is
-		 * invalidated with every MI_FLUSH.
-		 *
-		 * TLBs:
-		 *
-		 * On 965, TLBs associated with I915_GEM_DOMAIN_COMMAND
-		 * and I915_GEM_DOMAIN_CPU in are invalidated at PTE write and
-		 * I915_GEM_DOMAIN_RENDER and I915_GEM_DOMAIN_SAMPLER
-		 * are flushed at any MI_FLUSH.
-		 */
-
-		cmd = MI_FLUSH | MI_NO_WRITE_FLUSH;
-		if ((invalidate_domains|flush_domains) &
-		    I915_GEM_DOMAIN_RENDER)
-			cmd &= ~MI_NO_WRITE_FLUSH;
-		if (!IS_I965G(dev)) {
-			/*
-			 * On the 965, the sampler cache always gets flushed
-			 * and this bit is reserved.
-			 */
-			if (invalidate_domains & I915_GEM_DOMAIN_SAMPLER)
-				cmd |= MI_READ_FLUSH;
-		}
-		if (invalidate_domains & I915_GEM_DOMAIN_INSTRUCTION)
-			cmd |= MI_EXE_FLUSH;
-
-#if WATCH_EXEC
-		DRM_INFO("%s: queue flush %08x to ring\n", __func__, cmd);
-#endif
-		BEGIN_LP_RING(2);
-		OUT_RING(cmd);
-		OUT_RING(MI_NOOP);
-		ADVANCE_LP_RING();
-	}
+static void
+i915_gem_flush_ring(struct drm_device *dev,
+	       uint32_t invalidate_domains,
+	       uint32_t flush_domains,
+	       struct intel_ring_buffer *ring)
+{
+	if (flush_domains & I915_GEM_DOMAIN_CPU)
+		drm_agp_chipset_flush(dev);
+	ring->flush(dev, ring,
+			invalidate_domains,
+			flush_domains);
 }
 
 /**
@@ -2030,7 +1903,8 @@ i915_gem_object_wait_rendering(struct drm_gem_object *obj)
 		DRM_INFO("%s: object %p wait for seqno %08x\n",
 			  __func__, obj, obj_priv->last_rendering_seqno);
 #endif
-		ret = i915_wait_request(dev, obj_priv->last_rendering_seqno);
+		ret = i915_wait_request(dev,
+				obj_priv->last_rendering_seqno, obj_priv->ring);
 		if (ret != 0)
 			return ret;
 	}
@@ -2146,11 +2020,14 @@ i915_gpu_idle(struct drm_device *dev)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	bool lists_empty;
-	uint32_t seqno;
+	uint32_t seqno1, seqno2;
+	int ret;
 
 	spin_lock(&dev_priv->mm.active_list_lock);
-	lists_empty = list_empty(&dev_priv->mm.flushing_list) &&
-		      list_empty(&dev_priv->mm.active_list);
+	lists_empty = (list_empty(&dev_priv->mm.flushing_list) &&
+		       list_empty(&dev_priv->render_ring.active_list) &&
+		       (!HAS_BSD(dev) ||
+			list_empty(&dev_priv->bsd_ring.active_list)));
 	spin_unlock(&dev_priv->mm.active_list_lock);
 
 	if (lists_empty)
@@ -2158,11 +2035,25 @@ i915_gpu_idle(struct drm_device *dev)
 
 	/* 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)
+	seqno1 = i915_add_request(dev, NULL, I915_GEM_GPU_DOMAINS,
+			&dev_priv->render_ring);
+	if (seqno1 == 0)
 		return -ENOMEM;
+	ret = i915_wait_request(dev, seqno1, &dev_priv->render_ring);
+
+	if (HAS_BSD(dev)) {
+		seqno2 = i915_add_request(dev, NULL, I915_GEM_GPU_DOMAINS,
+				&dev_priv->bsd_ring);
+		if (seqno2 == 0)
+			return -ENOMEM;
+
+		ret = i915_wait_request(dev, seqno2, &dev_priv->bsd_ring);
+		if (ret)
+			return ret;
+	}
+
 
-	return i915_wait_request(dev, seqno);
+	return ret;
 }
 
 static int
@@ -2175,7 +2066,9 @@ i915_gem_evict_everything(struct drm_device *dev)
 	spin_lock(&dev_priv->mm.active_list_lock);
 	lists_empty = (list_empty(&dev_priv->mm.inactive_list) &&
 		       list_empty(&dev_priv->mm.flushing_list) &&
-		       list_empty(&dev_priv->mm.active_list));
+		       list_empty(&dev_priv->render_ring.active_list) &&
+		       (!HAS_BSD(dev)
+			|| list_empty(&dev_priv->bsd_ring.active_list)));
 	spin_unlock(&dev_priv->mm.active_list_lock);
 
 	if (lists_empty)
@@ -2195,7 +2088,9 @@ i915_gem_evict_everything(struct drm_device *dev)
 	spin_lock(&dev_priv->mm.active_list_lock);
 	lists_empty = (list_empty(&dev_priv->mm.inactive_list) &&
 		       list_empty(&dev_priv->mm.flushing_list) &&
-		       list_empty(&dev_priv->mm.active_list));
+		       list_empty(&dev_priv->render_ring.active_list) &&
+		       (!HAS_BSD(dev)
+			|| list_empty(&dev_priv->bsd_ring.active_list)));
 	spin_unlock(&dev_priv->mm.active_list_lock);
 	BUG_ON(!lists_empty);
 
@@ -2209,8 +2104,13 @@ i915_gem_evict_something(struct drm_device *dev, int min_size)
 	struct drm_gem_object *obj;
 	int ret;
 
+	struct intel_ring_buffer *render_ring = &dev_priv->render_ring;
+	struct intel_ring_buffer *bsd_ring = &dev_priv->bsd_ring;
 	for (;;) {
-		i915_gem_retire_requests(dev);
+		i915_gem_retire_requests(dev, render_ring);
+
+		if (HAS_BSD(dev))
+			i915_gem_retire_requests(dev, bsd_ring);
 
 		/* If there's an inactive buffer available now, grab it
 		 * and be done.
@@ -2234,14 +2134,30 @@ i915_gem_evict_something(struct drm_device *dev, int min_size)
 		 * things, wait for the next to finish and hopefully leave us
 		 * a buffer to evict.
 		 */
-		if (!list_empty(&dev_priv->mm.request_list)) {
+		if (!list_empty(&render_ring->request_list)) {
+			struct drm_i915_gem_request *request;
+
+			request = list_first_entry(&render_ring->request_list,
+						   struct drm_i915_gem_request,
+						   list);
+
+			ret = i915_wait_request(dev,
+					request->seqno, request->ring);
+			if (ret)
+				return ret;
+
+			continue;
+		}
+
+		if (HAS_BSD(dev) && !list_empty(&bsd_ring->request_list)) {
 			struct drm_i915_gem_request *request;
 
-			request = list_first_entry(&dev_priv->mm.request_list,
+			request = list_first_entry(&bsd_ring->request_list,
 						   struct drm_i915_gem_request,
 						   list);
 
-			ret = i915_wait_request(dev, request->seqno);
+			ret = i915_wait_request(dev,
+					request->seqno, request->ring);
 			if (ret)
 				return ret;
 
@@ -2268,10 +2184,13 @@ i915_gem_evict_something(struct drm_device *dev, int min_size)
 			if (obj != NULL) {
 				uint32_t seqno;
 
-				i915_gem_flush(dev,
+				i915_gem_flush_ring(dev,
+					       obj->write_domain,
 					       obj->write_domain,
-					       obj->write_domain);
-				seqno = i915_add_request(dev, NULL, obj->write_domain);
+					       obj_priv->ring);
+				seqno = i915_add_request(dev, NULL,
+						obj->write_domain,
+						obj_priv->ring);
 				if (seqno == 0)
 					return -ENOMEM;
 				continue;
@@ -2299,6 +2218,9 @@ i915_gem_object_get_pages(struct drm_gem_object *obj,
 	struct inode *inode;
 	struct page *page;
 
+	BUG_ON(obj_priv->pages_refcount
+			== DRM_I915_GEM_OBJECT_MAX_PAGES_REFCOUNT);
+
 	if (obj_priv->pages_refcount++ != 0)
 		return 0;
 
@@ -2697,6 +2619,14 @@ i915_gem_object_bind_to_gtt(struct drm_gem_object *obj, unsigned alignment)
 		return -EINVAL;
 	}
 
+	/* If the object is bigger than the entire aperture, reject it early
+	 * before evicting everything in a vain attempt to find space.
+	 */
+	if (obj->size > dev->gtt_total) {
+		DRM_ERROR("Attempting to bind an object larger than the aperture\n");
+		return -E2BIG;
+	}
+
  search_free:
 	free_space = drm_mm_search_free(&dev_priv->mm.gtt_space,
 					obj->size, alignment, 0);
@@ -2807,6 +2737,7 @@ i915_gem_object_flush_gpu_write_domain(struct drm_gem_object *obj)
 {
 	struct drm_device *dev = obj->dev;
 	uint32_t old_write_domain;
+	struct drm_i915_gem_object *obj_priv = to_intel_bo(obj);
 
 	if ((obj->write_domain & I915_GEM_GPU_DOMAINS) == 0)
 		return;
@@ -2814,7 +2745,7 @@ 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);
-	(void) i915_add_request(dev, NULL, obj->write_domain);
+	(void) i915_add_request(dev, NULL, obj->write_domain, obj_priv->ring);
 	BUG_ON(obj->write_domain);
 
 	trace_i915_gem_object_change_domain(obj,
@@ -2954,23 +2885,24 @@ i915_gem_object_set_to_display_plane(struct drm_gem_object *obj)
 		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);
+		ret = i915_do_wait_request(dev,
+				obj_priv->last_rendering_seqno,
+				0,
+				obj_priv->ring);
 		if (ret != 0)
 			return ret;
 	}
 
+	i915_gem_object_flush_cpu_write_domain(obj);
+
 	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->read_domains = I915_GEM_DOMAIN_GTT;
 	obj->write_domain = I915_GEM_DOMAIN_GTT;
 	obj_priv->dirty = 1;
 
@@ -3354,9 +3286,13 @@ i915_gem_object_pin_and_relocate(struct drm_gem_object *obj,
 	             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);
+	if (need_fence &&
+	    !i915_gem_object_fence_offset_ok(obj,
+					     obj_priv->tiling_mode)) {
+		ret = i915_gem_object_unbind(obj);
+		if (ret)
+			return ret;
+	}
 
 	/* Choose the GTT offset for our buffer and put it there. */
 	ret = i915_gem_object_pin(obj, (uint32_t) entry->alignment);
@@ -3370,9 +3306,6 @@ i915_gem_object_pin_and_relocate(struct drm_gem_object *obj,
 	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;
 		}
@@ -3545,62 +3478,6 @@ i915_gem_object_pin_and_relocate(struct drm_gem_object *obj,
 	return 0;
 }
 
-/** Dispatch a batchbuffer to the ring
- */
-static int
-i915_dispatch_gem_execbuffer(struct drm_device *dev,
-			      struct drm_i915_gem_execbuffer2 *exec,
-			      struct drm_clip_rect *cliprects,
-			      uint64_t exec_offset)
-{
-	drm_i915_private_t *dev_priv = dev->dev_private;
-	int nbox = exec->num_cliprects;
-	int i = 0, count;
-	uint32_t exec_start, exec_len;
-	RING_LOCALS;
-
-	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 + 1);
-
-	count = nbox ? nbox : 1;
-
-	for (i = 0; i < count; i++) {
-		if (i < nbox) {
-			int ret = i915_emit_box(dev, cliprects, i,
-						exec->DR1, exec->DR4);
-			if (ret)
-				return ret;
-		}
-
-		if (IS_I830(dev) || IS_845G(dev)) {
-			BEGIN_LP_RING(4);
-			OUT_RING(MI_BATCH_BUFFER);
-			OUT_RING(exec_start | MI_BATCH_NON_SECURE);
-			OUT_RING(exec_start + exec_len - 4);
-			OUT_RING(0);
-			ADVANCE_LP_RING();
-		} else {
-			BEGIN_LP_RING(2);
-			if (IS_I965G(dev)) {
-				OUT_RING(MI_BATCH_BUFFER_START |
-					 (2 << 6) |
-					 MI_BATCH_NON_SECURE_I965);
-				OUT_RING(exec_start);
-			} else {
-				OUT_RING(MI_BATCH_BUFFER_START |
-					 (2 << 6));
-				OUT_RING(exec_start | MI_BATCH_NON_SECURE);
-			}
-			ADVANCE_LP_RING();
-		}
-	}
-
-	/* XXX breadcrumb */
-	return 0;
-}
-
 /* Throttle our rendering by waiting until the ring has completed our requests
  * emitted over 20 msec ago.
  *
@@ -3629,7 +3506,7 @@ i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file_priv)
 		if (time_after_eq(request->emitted_jiffies, recent_enough))
 			break;
 
-		ret = i915_wait_request(dev, request->seqno);
+		ret = i915_wait_request(dev, request->seqno, request->ring);
 		if (ret != 0)
 			break;
 	}
@@ -3786,10 +3663,22 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 	uint32_t seqno, flush_domains, reloc_index;
 	int pin_tries, flips;
 
+	struct intel_ring_buffer *ring = NULL;
+
 #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->flags & I915_EXEC_BSD) {
+		if (!HAS_BSD(dev)) {
+			DRM_ERROR("execbuf with wrong flag\n");
+			return -EINVAL;
+		}
+		ring = &dev_priv->bsd_ring;
+	} else {
+		ring = &dev_priv->render_ring;
+	}
+
 
 	if (args->buffer_count < 1) {
 		DRM_ERROR("execbuf with %d buffers\n", args->buffer_count);
@@ -3902,11 +3791,19 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 		if (ret != -ENOSPC || pin_tries >= 1) {
 			if (ret != -ERESTARTSYS) {
 				unsigned long long total_size = 0;
-				for (i = 0; i < args->buffer_count; i++)
+				int num_fences = 0;
+				for (i = 0; i < args->buffer_count; i++) {
+					obj_priv = object_list[i]->driver_private;
+
 					total_size += object_list[i]->size;
-				DRM_ERROR("Failed to pin buffer %d of %d, total %llu bytes: %d\n",
+					num_fences +=
+						exec_list[i].flags & EXEC_OBJECT_NEEDS_FENCE &&
+						obj_priv->tiling_mode != I915_TILING_NONE;
+				}
+				DRM_ERROR("Failed to pin buffer %d of %d, total %llu bytes, %d fences: %d\n",
 					  pinned+1, args->buffer_count,
-					  total_size, ret);
+					  total_size, num_fences,
+					  ret);
 				DRM_ERROR("%d objects [%d pinned], "
 					  "%d object bytes [%d pinned], "
 					  "%d/%d gtt bytes\n",
@@ -3976,9 +3873,16 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 		i915_gem_flush(dev,
 			       dev->invalidate_domains,
 			       dev->flush_domains);
-		if (dev->flush_domains & I915_GEM_GPU_DOMAINS)
+		if (dev->flush_domains & I915_GEM_GPU_DOMAINS) {
 			(void)i915_add_request(dev, file_priv,
-					       dev->flush_domains);
+					dev->flush_domains,
+					&dev_priv->render_ring);
+
+			if (HAS_BSD(dev))
+				(void)i915_add_request(dev, file_priv,
+						dev->flush_domains,
+						&dev_priv->bsd_ring);
+		}
 	}
 
 	for (i = 0; i < args->buffer_count; i++) {
@@ -4015,7 +3919,8 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 #endif
 
 	/* Exec the batchbuffer */
-	ret = i915_dispatch_gem_execbuffer(dev, args, cliprects, exec_offset);
+	ret = ring->dispatch_gem_execbuffer(dev, ring, args,
+			cliprects, exec_offset);
 	if (ret) {
 		DRM_ERROR("dispatch failed %d\n", ret);
 		goto err;
@@ -4025,7 +3930,7 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 	 * Ensure that the commands in the batch buffer are
 	 * finished before the interrupt fires
 	 */
-	flush_domains = i915_retire_commands(dev);
+	flush_domains = i915_retire_commands(dev, ring);
 
 	i915_verify_inactive(dev, __FILE__, __LINE__);
 
@@ -4036,12 +3941,13 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 	 * *some* interrupts representing completion of buffers that we can
 	 * wait on when trying to clear up gtt space).
 	 */
-	seqno = i915_add_request(dev, file_priv, flush_domains);
+	seqno = i915_add_request(dev, file_priv, flush_domains, ring);
 	BUG_ON(seqno == 0);
 	for (i = 0; i < args->buffer_count; i++) {
 		struct drm_gem_object *obj = object_list[i];
+		obj_priv = to_intel_bo(obj);
 
-		i915_gem_object_move_to_active(obj, seqno);
+		i915_gem_object_move_to_active(obj, seqno, ring);
 #if WATCH_LRU
 		DRM_INFO("%s: move to exec list %p\n", __func__, obj);
 #endif
@@ -4153,7 +4059,7 @@ i915_gem_execbuffer(struct drm_device *dev, void *data,
 	exec2.DR4 = args->DR4;
 	exec2.num_cliprects = args->num_cliprects;
 	exec2.cliprects_ptr = args->cliprects_ptr;
-	exec2.flags = 0;
+	exec2.flags = I915_EXEC_RENDER;
 
 	ret = i915_gem_do_execbuffer(dev, data, file_priv, &exec2, exec2_list);
 	if (!ret) {
@@ -4239,7 +4145,20 @@ i915_gem_object_pin(struct drm_gem_object *obj, uint32_t alignment)
 	struct drm_i915_gem_object *obj_priv = to_intel_bo(obj);
 	int ret;
 
+	BUG_ON(obj_priv->pin_count == DRM_I915_GEM_OBJECT_MAX_PIN_COUNT);
+
 	i915_verify_inactive(dev, __FILE__, __LINE__);
+
+	if (obj_priv->gtt_space != NULL) {
+		if (alignment == 0)
+			alignment = i915_gem_get_gtt_alignment(obj);
+		if (obj_priv->gtt_offset & (alignment - 1)) {
+			ret = i915_gem_object_unbind(obj);
+			if (ret)
+				return ret;
+		}
+	}
+
 	if (obj_priv->gtt_space == NULL) {
 		ret = i915_gem_object_bind_to_gtt(obj, alignment);
 		if (ret)
@@ -4392,6 +4311,7 @@ i915_gem_busy_ioctl(struct drm_device *dev, void *data,
 	struct drm_i915_gem_busy *args = data;
 	struct drm_gem_object *obj;
 	struct drm_i915_gem_object *obj_priv;
+	drm_i915_private_t *dev_priv = dev->dev_private;
 
 	obj = drm_gem_object_lookup(dev, file_priv, args->handle);
 	if (obj == NULL) {
@@ -4406,7 +4326,10 @@ i915_gem_busy_ioctl(struct drm_device *dev, void *data,
 	 * actually unmasked, and our working set ends up being larger than
 	 * required.
 	 */
-	i915_gem_retire_requests(dev);
+	i915_gem_retire_requests(dev, &dev_priv->render_ring);
+
+	if (HAS_BSD(dev))
+		i915_gem_retire_requests(dev, &dev_priv->bsd_ring);
 
 	obj_priv = to_intel_bo(obj);
 	/* Don't count being on the flushing list against the object being
@@ -4573,7 +4496,10 @@ i915_gem_idle(struct drm_device *dev)
 
 	mutex_lock(&dev->struct_mutex);
 
-	if (dev_priv->mm.suspended || dev_priv->ring.ring_obj == NULL) {
+	if (dev_priv->mm.suspended ||
+			(dev_priv->render_ring.gem_object == NULL) ||
+			(HAS_BSD(dev) &&
+			 dev_priv->bsd_ring.gem_object == NULL)) {
 		mutex_unlock(&dev->struct_mutex);
 		return 0;
 	}
@@ -4654,71 +4580,6 @@ err:
 	return ret;
 }
 
-static int
-i915_gem_init_hws(struct drm_device *dev)
-{
-	drm_i915_private_t *dev_priv = dev->dev_private;
-	struct drm_gem_object *obj;
-	struct drm_i915_gem_object *obj_priv;
-	int ret;
-
-	/* If we need a physical address for the status page, it's already
-	 * initialized at driver load time.
-	 */
-	if (!I915_NEED_GFX_HWS(dev))
-		return 0;
-
-	obj = i915_gem_alloc_object(dev, 4096);
-	if (obj == NULL) {
-		DRM_ERROR("Failed to allocate status page\n");
-		ret = -ENOMEM;
-		goto err;
-	}
-	obj_priv = to_intel_bo(obj);
-	obj_priv->agp_type = AGP_USER_CACHED_MEMORY;
-
-	ret = i915_gem_object_pin(obj, 4096);
-	if (ret != 0) {
-		drm_gem_object_unreference(obj);
-		goto err_unref;
-	}
-
-	dev_priv->status_gfx_addr = obj_priv->gtt_offset;
-
-	dev_priv->hw_status_page = kmap(obj_priv->pages[0]);
-	if (dev_priv->hw_status_page == NULL) {
-		DRM_ERROR("Failed to map status page.\n");
-		memset(&dev_priv->hws_map, 0, sizeof(dev_priv->hws_map));
-		ret = -EINVAL;
-		goto err_unpin;
-	}
-
-	if (HAS_PIPE_CONTROL(dev)) {
-		ret = i915_gem_init_pipe_control(dev);
-		if (ret)
-			goto err_unpin;
-	}
-
-	dev_priv->hws_obj = obj;
-	memset(dev_priv->hw_status_page, 0, PAGE_SIZE);
-	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;
-
-err_unpin:
-	i915_gem_object_unpin(obj);
-err_unref:
-	drm_gem_object_unreference(obj);
-err:
-	return 0;
-}
 
 static void
 i915_gem_cleanup_pipe_control(struct drm_device *dev)
@@ -4737,146 +4598,46 @@ i915_gem_cleanup_pipe_control(struct drm_device *dev)
 	dev_priv->seqno_page = NULL;
 }
 
-static void
-i915_gem_cleanup_hws(struct drm_device *dev)
-{
-	drm_i915_private_t *dev_priv = dev->dev_private;
-	struct drm_gem_object *obj;
-	struct drm_i915_gem_object *obj_priv;
-
-	if (dev_priv->hws_obj == NULL)
-		return;
-
-	obj = dev_priv->hws_obj;
-	obj_priv = to_intel_bo(obj);
-
-	kunmap(obj_priv->pages[0]);
-	i915_gem_object_unpin(obj);
-	drm_gem_object_unreference(obj);
-	dev_priv->hws_obj = NULL;
-
-	memset(&dev_priv->hws_map, 0, sizeof(dev_priv->hws_map));
-	dev_priv->hw_status_page = NULL;
-
-	if (HAS_PIPE_CONTROL(dev))
-		i915_gem_cleanup_pipe_control(dev);
-
-	/* Write high address into HWS_PGA when disabling. */
-	I915_WRITE(HWS_PGA, 0x1ffff000);
-}
-
 int
 i915_gem_init_ringbuffer(struct drm_device *dev)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
-	struct drm_gem_object *obj;
-	struct drm_i915_gem_object *obj_priv;
-	drm_i915_ring_buffer_t *ring = &dev_priv->ring;
 	int ret;
-	u32 head;
-
-	ret = i915_gem_init_hws(dev);
-	if (ret != 0)
-		return ret;
 
-	obj = i915_gem_alloc_object(dev, 128 * 1024);
-	if (obj == NULL) {
-		DRM_ERROR("Failed to allocate ringbuffer\n");
-		i915_gem_cleanup_hws(dev);
-		return -ENOMEM;
-	}
-	obj_priv = to_intel_bo(obj);
+	dev_priv->render_ring = render_ring;
 
-	ret = i915_gem_object_pin(obj, 4096);
-	if (ret != 0) {
-		drm_gem_object_unreference(obj);
-		i915_gem_cleanup_hws(dev);
-		return ret;
+	if (!I915_NEED_GFX_HWS(dev)) {
+		dev_priv->render_ring.status_page.page_addr
+			= dev_priv->status_page_dmah->vaddr;
+		memset(dev_priv->render_ring.status_page.page_addr,
+				0, PAGE_SIZE);
 	}
 
-	/* Set up the kernel mapping for the ring. */
-	ring->Size = obj->size;
-
-	ring->map.offset = dev->agp->base + obj_priv->gtt_offset;
-	ring->map.size = obj->size;
-	ring->map.type = 0;
-	ring->map.flags = 0;
-	ring->map.mtrr = 0;
-
-	drm_core_ioremap_wc(&ring->map, dev);
-	if (ring->map.handle == NULL) {
-		DRM_ERROR("Failed to map ringbuffer.\n");
-		memset(&dev_priv->ring, 0, sizeof(dev_priv->ring));
-		i915_gem_object_unpin(obj);
-		drm_gem_object_unreference(obj);
-		i915_gem_cleanup_hws(dev);
-		return -EINVAL;
-	}
-	ring->ring_obj = obj;
-	ring->virtual_start = ring->map.handle;
-
-	/* Stop the ring if it's running. */
-	I915_WRITE(PRB0_CTL, 0);
-	I915_WRITE(PRB0_TAIL, 0);
-	I915_WRITE(PRB0_HEAD, 0);
-
-	/* Initialize the ring. */
-	I915_WRITE(PRB0_START, obj_priv->gtt_offset);
-	head = I915_READ(PRB0_HEAD) & HEAD_ADDR;
-
-	/* G45 ring initialization fails to reset head to zero */
-	if (head != 0) {
-		DRM_ERROR("Ring head not reset to zero "
-			  "ctl %08x head %08x tail %08x start %08x\n",
-			  I915_READ(PRB0_CTL),
-			  I915_READ(PRB0_HEAD),
-			  I915_READ(PRB0_TAIL),
-			  I915_READ(PRB0_START));
-		I915_WRITE(PRB0_HEAD, 0);
-
-		DRM_ERROR("Ring head forced to zero "
-			  "ctl %08x head %08x tail %08x start %08x\n",
-			  I915_READ(PRB0_CTL),
-			  I915_READ(PRB0_HEAD),
-			  I915_READ(PRB0_TAIL),
-			  I915_READ(PRB0_START));
-	}
-
-	I915_WRITE(PRB0_CTL,
-		   ((obj->size - 4096) & RING_NR_PAGES) |
-		   RING_NO_REPORT |
-		   RING_VALID);
-
-	head = I915_READ(PRB0_HEAD) & HEAD_ADDR;
-
-	/* If the head is still not zero, the ring is dead */
-	if (head != 0) {
-		DRM_ERROR("Ring initialization failed "
-			  "ctl %08x head %08x tail %08x start %08x\n",
-			  I915_READ(PRB0_CTL),
-			  I915_READ(PRB0_HEAD),
-			  I915_READ(PRB0_TAIL),
-			  I915_READ(PRB0_START));
-		return -EIO;
+	if (HAS_PIPE_CONTROL(dev)) {
+		ret = i915_gem_init_pipe_control(dev);
+		if (ret)
+			return ret;
 	}
 
-	/* Update our cache of the ring state */
-	if (!drm_core_check_feature(dev, DRIVER_MODESET))
-		i915_kernel_lost_context(dev);
-	else {
-		ring->head = I915_READ(PRB0_HEAD) & HEAD_ADDR;
-		ring->tail = I915_READ(PRB0_TAIL) & TAIL_ADDR;
-		ring->space = ring->head - (ring->tail + 8);
-		if (ring->space < 0)
-			ring->space += ring->Size;
-	}
+	ret = intel_init_ring_buffer(dev, &dev_priv->render_ring);
+	if (ret)
+		goto cleanup_pipe_control;
 
-	if (IS_I9XX(dev) && !IS_GEN3(dev)) {
-		I915_WRITE(MI_MODE,
-			   (VS_TIMER_DISPATCH) << 16 | VS_TIMER_DISPATCH);
+	if (HAS_BSD(dev)) {
+		dev_priv->bsd_ring = bsd_ring;
+		ret = intel_init_ring_buffer(dev, &dev_priv->bsd_ring);
+		if (ret)
+			goto cleanup_render_ring;
 	}
 
 	return 0;
+
+cleanup_render_ring:
+	intel_cleanup_ring_buffer(dev, &dev_priv->render_ring);
+cleanup_pipe_control:
+	if (HAS_PIPE_CONTROL(dev))
+		i915_gem_cleanup_pipe_control(dev);
+	return ret;
 }
 
 void
@@ -4884,17 +4645,11 @@ i915_gem_cleanup_ringbuffer(struct drm_device *dev)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
 
-	if (dev_priv->ring.ring_obj == NULL)
-		return;
-
-	drm_core_ioremapfree(&dev_priv->ring.map, dev);
-
-	i915_gem_object_unpin(dev_priv->ring.ring_obj);
-	drm_gem_object_unreference(dev_priv->ring.ring_obj);
-	dev_priv->ring.ring_obj = NULL;
-	memset(&dev_priv->ring, 0, sizeof(dev_priv->ring));
-
-	i915_gem_cleanup_hws(dev);
+	intel_cleanup_ring_buffer(dev, &dev_priv->render_ring);
+	if (HAS_BSD(dev))
+		intel_cleanup_ring_buffer(dev, &dev_priv->bsd_ring);
+	if (HAS_PIPE_CONTROL(dev))
+		i915_gem_cleanup_pipe_control(dev);
 }
 
 int
@@ -4922,12 +4677,14 @@ i915_gem_entervt_ioctl(struct drm_device *dev, void *data,
 	}
 
 	spin_lock(&dev_priv->mm.active_list_lock);
-	BUG_ON(!list_empty(&dev_priv->mm.active_list));
+	BUG_ON(!list_empty(&dev_priv->render_ring.active_list));
+	BUG_ON(HAS_BSD(dev) && !list_empty(&dev_priv->bsd_ring.active_list));
 	spin_unlock(&dev_priv->mm.active_list_lock);
 
 	BUG_ON(!list_empty(&dev_priv->mm.flushing_list));
 	BUG_ON(!list_empty(&dev_priv->mm.inactive_list));
-	BUG_ON(!list_empty(&dev_priv->mm.request_list));
+	BUG_ON(!list_empty(&dev_priv->render_ring.request_list));
+	BUG_ON(HAS_BSD(dev) && !list_empty(&dev_priv->bsd_ring.request_list));
 	mutex_unlock(&dev->struct_mutex);
 
 	drm_irq_install(dev);
@@ -4966,18 +4723,20 @@ i915_gem_load(struct drm_device *dev)
 	drm_i915_private_t *dev_priv = dev->dev_private;
 
 	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);
+	INIT_LIST_HEAD(&dev_priv->render_ring.active_list);
+	INIT_LIST_HEAD(&dev_priv->render_ring.request_list);
+	if (HAS_BSD(dev)) {
+		INIT_LIST_HEAD(&dev_priv->bsd_ring.active_list);
+		INIT_LIST_HEAD(&dev_priv->bsd_ring.request_list);
+	}
 	for (i = 0; i < 16; i++)
 		INIT_LIST_HEAD(&dev_priv->fence_regs[i].lru_list);
 	INIT_DELAYED_WORK(&dev_priv->mm.retire_work,
 			  i915_gem_retire_work_handler);
-	dev_priv->mm.next_gem_seqno = 1;
-
 	spin_lock(&shrink_list_lock);
 	list_add(&dev_priv->mm.shrink_list, &shrink_list);
 	spin_unlock(&shrink_list_lock);
@@ -5209,7 +4968,9 @@ i915_gpu_is_active(struct drm_device *dev)
 
 	spin_lock(&dev_priv->mm.active_list_lock);
 	lists_empty = list_empty(&dev_priv->mm.flushing_list) &&
-		      list_empty(&dev_priv->mm.active_list);
+		      list_empty(&dev_priv->render_ring.active_list);
+	if (HAS_BSD(dev))
+		lists_empty &= list_empty(&dev_priv->bsd_ring.active_list);
 	spin_unlock(&dev_priv->mm.active_list_lock);
 
 	return !lists_empty;
@@ -5254,8 +5015,10 @@ rescan:
 			continue;
 
 		spin_unlock(&shrink_list_lock);
+		i915_gem_retire_requests(dev, &dev_priv->render_ring);
 
-		i915_gem_retire_requests(dev);
+		if (HAS_BSD(dev))
+			i915_gem_retire_requests(dev, &dev_priv->bsd_ring);
 
 		list_for_each_entry_safe(obj_priv, next_obj,
 					 &dev_priv->mm.inactive_list,
diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
index 8c3f0802686d..2479be001e40 100644
--- a/drivers/gpu/drm/i915/i915_irq.c
+++ b/drivers/gpu/drm/i915/i915_irq.c
@@ -53,7 +53,7 @@
 	 I915_RENDER_COMMAND_PARSER_ERROR_INTERRUPT)
 
 /** Interrupts that we mask and unmask at runtime. */
-#define I915_INTERRUPT_ENABLE_VAR (I915_USER_INTERRUPT)
+#define I915_INTERRUPT_ENABLE_VAR (I915_USER_INTERRUPT | I915_BSD_USER_INTERRUPT)
 
 #define I915_PIPE_VBLANK_STATUS	(PIPE_START_VBLANK_INTERRUPT_STATUS |\
 				 PIPE_VBLANK_INTERRUPT_STATUS)
@@ -74,7 +74,7 @@ ironlake_enable_graphics_irq(drm_i915_private_t *dev_priv, u32 mask)
 	}
 }
 
-static inline void
+void
 ironlake_disable_graphics_irq(drm_i915_private_t *dev_priv, u32 mask)
 {
 	if ((dev_priv->gt_irq_mask_reg & mask) != mask) {
@@ -115,7 +115,7 @@ i915_enable_irq(drm_i915_private_t *dev_priv, u32 mask)
 	}
 }
 
-static inline void
+void
 i915_disable_irq(drm_i915_private_t *dev_priv, u32 mask)
 {
 	if ((dev_priv->irq_mask_reg & mask) != mask) {
@@ -278,10 +278,9 @@ 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);
+	I915_WRITE16(MEMINTRSTS, MEMINT_EVAL_CHG);
 	busy_up = I915_READ(RCPREVBSYTUPAVG);
 	busy_down = I915_READ(RCPREVBSYTDNAVG);
 	max_avg = I915_READ(RCBMAXAVG);
@@ -300,27 +299,8 @@ static void i915_handle_rps_change(struct drm_device *dev)
 			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");
+	if (ironlake_set_drps(dev, new_delay))
+		dev_priv->cur_delay = new_delay;
 
 	return;
 }
@@ -331,6 +311,7 @@ irqreturn_t ironlake_irq_handler(struct drm_device *dev)
 	int ret = IRQ_NONE;
 	u32 de_iir, gt_iir, de_ier, pch_iir;
 	struct drm_i915_master_private *master_priv;
+	struct intel_ring_buffer *render_ring = &dev_priv->render_ring;
 
 	/* disable master interrupt before clearing iir  */
 	de_ier = I915_READ(DEIER);
@@ -354,13 +335,16 @@ irqreturn_t ironlake_irq_handler(struct drm_device *dev)
 	}
 
 	if (gt_iir & GT_PIPE_NOTIFY) {
-		u32 seqno = i915_get_gem_seqno(dev);
-		dev_priv->mm.irq_gem_seqno = seqno;
+		u32 seqno = render_ring->get_gem_seqno(dev, render_ring);
+		render_ring->irq_gem_seqno = seqno;
 		trace_i915_gem_request_complete(dev, seqno);
-		DRM_WAKEUP(&dev_priv->irq_queue);
+		DRM_WAKEUP(&dev_priv->render_ring.irq_queue);
 		dev_priv->hangcheck_count = 0;
 		mod_timer(&dev_priv->hangcheck_timer, jiffies + DRM_I915_HANGCHECK_PERIOD);
 	}
+	if (gt_iir & GT_BSD_USER_INTERRUPT)
+		DRM_WAKEUP(&dev_priv->bsd_ring.irq_queue);
+
 
 	if (de_iir & DE_GSE)
 		ironlake_opregion_gse_intr(dev);
@@ -388,7 +372,7 @@ irqreturn_t ironlake_irq_handler(struct drm_device *dev)
 	}
 
 	if (de_iir & DE_PCU_EVENT) {
-		I915_WRITE(MEMINTRSTS, I915_READ(MEMINTRSTS));
+		I915_WRITE16(MEMINTRSTS, I915_READ(MEMINTRSTS));
 		i915_handle_rps_change(dev);
 	}
 
@@ -536,17 +520,18 @@ i915_ringbuffer_last_batch(struct drm_device *dev)
 	 */
 	bbaddr = 0;
 	head = I915_READ(PRB0_HEAD) & HEAD_ADDR;
-	ring = (u32 *)(dev_priv->ring.virtual_start + head);
+	ring = (u32 *)(dev_priv->render_ring.virtual_start + head);
 
-	while (--ring >= (u32 *)dev_priv->ring.virtual_start) {
+	while (--ring >= (u32 *)dev_priv->render_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) {
+		ring = (u32 *)(dev_priv->render_ring.virtual_start
+				+ dev_priv->render_ring.size);
+		while (--ring >= (u32 *)dev_priv->render_ring.virtual_start) {
 			bbaddr = i915_get_bbaddr(dev, ring);
 			if (bbaddr)
 				break;
@@ -587,7 +572,7 @@ static void i915_capture_error_state(struct drm_device *dev)
 		return;
 	}
 
-	error->seqno = i915_get_gem_seqno(dev);
+	error->seqno = i915_get_gem_seqno(dev, &dev_priv->render_ring);
 	error->eir = I915_READ(EIR);
 	error->pgtbl_er = I915_READ(PGTBL_ER);
 	error->pipeastat = I915_READ(PIPEASTAT);
@@ -615,7 +600,9 @@ static void i915_capture_error_state(struct drm_device *dev)
 	batchbuffer[0] = NULL;
 	batchbuffer[1] = NULL;
 	count = 0;
-	list_for_each_entry(obj_priv, &dev_priv->mm.active_list, list) {
+	list_for_each_entry(obj_priv,
+			&dev_priv->render_ring.active_list, list) {
+
 		struct drm_gem_object *obj = &obj_priv->base;
 
 		if (batchbuffer[0] == NULL &&
@@ -639,7 +626,8 @@ static void i915_capture_error_state(struct drm_device *dev)
 	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);
+	error->ringbuffer = i915_error_object_create(dev,
+			dev_priv->render_ring.gem_object);
 
 	/* Record buffers on the active list. */
 	error->active_bo = NULL;
@@ -651,7 +639,8 @@ static void i915_capture_error_state(struct drm_device *dev)
 
 	if (error->active_bo) {
 		int i = 0;
-		list_for_each_entry(obj_priv, &dev_priv->mm.active_list, list) {
+		list_for_each_entry(obj_priv,
+				&dev_priv->render_ring.active_list, list) {
 			struct drm_gem_object *obj = &obj_priv->base;
 
 			error->active_bo[i].size = obj->size;
@@ -703,24 +692,13 @@ void i915_destroy_error_state(struct drm_device *dev)
 		i915_error_state_free(dev, error);
 }
 
-/**
- * i915_handle_error - handle an error interrupt
- * @dev: drm device
- *
- * Do some basic checking of regsiter state at error interrupt time and
- * dump it to the syslog.  Also call i915_capture_error_state() to make
- * sure we get a record and make it available in debugfs.  Fire a uevent
- * so userspace knows something bad happened (should trigger collection
- * of a ring dump etc.).
- */
-static void i915_handle_error(struct drm_device *dev, bool wedged)
+static void i915_report_and_clear_eir(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	u32 eir = I915_READ(EIR);
-	u32 pipea_stats = I915_READ(PIPEASTAT);
-	u32 pipeb_stats = I915_READ(PIPEBSTAT);
 
-	i915_capture_error_state(dev);
+	if (!eir)
+		return;
 
 	printk(KERN_ERR "render error detected, EIR: 0x%08x\n",
 	       eir);
@@ -766,6 +744,9 @@ static void i915_handle_error(struct drm_device *dev, bool wedged)
 	}
 
 	if (eir & I915_ERROR_MEMORY_REFRESH) {
+		u32 pipea_stats = I915_READ(PIPEASTAT);
+		u32 pipeb_stats = I915_READ(PIPEBSTAT);
+
 		printk(KERN_ERR "memory refresh error\n");
 		printk(KERN_ERR "PIPEASTAT: 0x%08x\n",
 		       pipea_stats);
@@ -822,6 +803,24 @@ static void i915_handle_error(struct drm_device *dev, bool wedged)
 		I915_WRITE(EMR, I915_READ(EMR) | eir);
 		I915_WRITE(IIR, I915_RENDER_COMMAND_PARSER_ERROR_INTERRUPT);
 	}
+}
+
+/**
+ * i915_handle_error - handle an error interrupt
+ * @dev: drm device
+ *
+ * Do some basic checking of regsiter state at error interrupt time and
+ * dump it to the syslog.  Also call i915_capture_error_state() to make
+ * sure we get a record and make it available in debugfs.  Fire a uevent
+ * so userspace knows something bad happened (should trigger collection
+ * of a ring dump etc.).
+ */
+static void i915_handle_error(struct drm_device *dev, bool wedged)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+
+	i915_capture_error_state(dev);
+	i915_report_and_clear_eir(dev);
 
 	if (wedged) {
 		atomic_set(&dev_priv->mm.wedged, 1);
@@ -829,7 +828,7 @@ static void i915_handle_error(struct drm_device *dev, bool wedged)
 		/*
 		 * Wakeup waiting processes so they don't hang
 		 */
-		DRM_WAKEUP(&dev_priv->irq_queue);
+		DRM_WAKEUP(&dev_priv->render_ring.irq_queue);
 	}
 
 	queue_work(dev_priv->wq, &dev_priv->error_work);
@@ -848,6 +847,7 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS)
 	unsigned long irqflags;
 	int irq_received;
 	int ret = IRQ_NONE;
+	struct intel_ring_buffer *render_ring = &dev_priv->render_ring;
 
 	atomic_inc(&dev_priv->irq_received);
 
@@ -928,14 +928,18 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS)
 		}
 
 		if (iir & I915_USER_INTERRUPT) {
-			u32 seqno = i915_get_gem_seqno(dev);
-			dev_priv->mm.irq_gem_seqno = seqno;
+			u32 seqno =
+				render_ring->get_gem_seqno(dev, render_ring);
+			render_ring->irq_gem_seqno = seqno;
 			trace_i915_gem_request_complete(dev, seqno);
-			DRM_WAKEUP(&dev_priv->irq_queue);
+			DRM_WAKEUP(&dev_priv->render_ring.irq_queue);
 			dev_priv->hangcheck_count = 0;
 			mod_timer(&dev_priv->hangcheck_timer, jiffies + DRM_I915_HANGCHECK_PERIOD);
 		}
 
+		if (HAS_BSD(dev) && (iir & I915_BSD_USER_INTERRUPT))
+			DRM_WAKEUP(&dev_priv->bsd_ring.irq_queue);
+
 		if (iir & I915_DISPLAY_PLANE_A_FLIP_PENDING_INTERRUPT)
 			intel_prepare_page_flip(dev, 0);
 
@@ -984,7 +988,6 @@ static int i915_emit_irq(struct drm_device * dev)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
 	struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv;
-	RING_LOCALS;
 
 	i915_kernel_lost_context(dev);
 
@@ -1006,43 +1009,13 @@ static int i915_emit_irq(struct drm_device * dev)
 	return dev_priv->counter;
 }
 
-void i915_user_irq_get(struct drm_device *dev)
-{
-	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
-	unsigned long irqflags;
-
-	spin_lock_irqsave(&dev_priv->user_irq_lock, irqflags);
-	if (dev->irq_enabled && (++dev_priv->user_irq_refcount == 1)) {
-		if (HAS_PCH_SPLIT(dev))
-			ironlake_enable_graphics_irq(dev_priv, GT_PIPE_NOTIFY);
-		else
-			i915_enable_irq(dev_priv, I915_USER_INTERRUPT);
-	}
-	spin_unlock_irqrestore(&dev_priv->user_irq_lock, irqflags);
-}
-
-void i915_user_irq_put(struct drm_device *dev)
-{
-	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
-	unsigned long irqflags;
-
-	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 (HAS_PCH_SPLIT(dev))
-			ironlake_disable_graphics_irq(dev_priv, GT_PIPE_NOTIFY);
-		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;
+	struct intel_ring_buffer *render_ring = &dev_priv->render_ring;
 
 	if (dev_priv->trace_irq_seqno == 0)
-		i915_user_irq_get(dev);
+		render_ring->user_irq_get(dev, render_ring);
 
 	dev_priv->trace_irq_seqno = seqno;
 }
@@ -1052,6 +1025,7 @@ 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;
+	struct intel_ring_buffer *render_ring = &dev_priv->render_ring;
 
 	DRM_DEBUG_DRIVER("irq_nr=%d breadcrumb=%d\n", irq_nr,
 		  READ_BREADCRUMB(dev_priv));
@@ -1065,10 +1039,10 @@ static int i915_wait_irq(struct drm_device * dev, int irq_nr)
 	if (master_priv->sarea_priv)
 		master_priv->sarea_priv->perf_boxes |= I915_BOX_WAIT;
 
-	i915_user_irq_get(dev);
-	DRM_WAIT_ON(ret, dev_priv->irq_queue, 3 * DRM_HZ,
+	render_ring->user_irq_get(dev, render_ring);
+	DRM_WAIT_ON(ret, dev_priv->render_ring.irq_queue, 3 * DRM_HZ,
 		    READ_BREADCRUMB(dev_priv) >= irq_nr);
-	i915_user_irq_put(dev);
+	render_ring->user_irq_put(dev, render_ring);
 
 	if (ret == -EBUSY) {
 		DRM_ERROR("EBUSY -- rec: %d emitted: %d\n",
@@ -1087,7 +1061,7 @@ int i915_irq_emit(struct drm_device *dev, void *data,
 	drm_i915_irq_emit_t *emit = data;
 	int result;
 
-	if (!dev_priv || !dev_priv->ring.virtual_start) {
+	if (!dev_priv || !dev_priv->render_ring.virtual_start) {
 		DRM_ERROR("called with no initialization\n");
 		return -EINVAL;
 	}
@@ -1233,9 +1207,12 @@ int i915_vblank_swap(struct drm_device *dev, void *data,
 	return -EINVAL;
 }
 
-struct drm_i915_gem_request *i915_get_tail_request(struct drm_device *dev) {
+struct drm_i915_gem_request *
+i915_get_tail_request(struct drm_device *dev)
+{
 	drm_i915_private_t *dev_priv = dev->dev_private;
-	return list_entry(dev_priv->mm.request_list.prev, struct drm_i915_gem_request, list);
+	return list_entry(dev_priv->render_ring.request_list.prev,
+			struct drm_i915_gem_request, list);
 }
 
 /**
@@ -1260,8 +1237,10 @@ void i915_hangcheck_elapsed(unsigned long data)
 		acthd = I915_READ(ACTHD_I965);
 
 	/* If all work is done then ACTHD clearly hasn't advanced. */
-	if (list_empty(&dev_priv->mm.request_list) ||
-		       i915_seqno_passed(i915_get_gem_seqno(dev), i915_get_tail_request(dev)->seqno)) {
+	if (list_empty(&dev_priv->render_ring.request_list) ||
+		i915_seqno_passed(i915_get_gem_seqno(dev,
+				&dev_priv->render_ring),
+			i915_get_tail_request(dev)->seqno)) {
 		dev_priv->hangcheck_count = 0;
 		return;
 	}
@@ -1314,7 +1293,7 @@ static int ironlake_irq_postinstall(struct drm_device *dev)
 	/* enable kind of interrupts always enabled */
 	u32 display_mask = DE_MASTER_IRQ_CONTROL | DE_GSE | DE_PCH_EVENT |
 			   DE_PLANEA_FLIP_DONE | DE_PLANEB_FLIP_DONE;
-	u32 render_mask = GT_PIPE_NOTIFY;
+	u32 render_mask = GT_PIPE_NOTIFY | GT_BSD_USER_INTERRUPT;
 	u32 hotplug_mask = SDE_CRT_HOTPLUG | SDE_PORTB_HOTPLUG |
 			   SDE_PORTC_HOTPLUG | SDE_PORTD_HOTPLUG;
 
@@ -1328,7 +1307,7 @@ static int ironlake_irq_postinstall(struct drm_device *dev)
 	(void) I915_READ(DEIER);
 
 	/* user interrupt should be enabled, but masked initial */
-	dev_priv->gt_irq_mask_reg = 0xffffffff;
+	dev_priv->gt_irq_mask_reg = ~render_mask;
 	dev_priv->gt_irq_enable_reg = render_mask;
 
 	I915_WRITE(GTIIR, I915_READ(GTIIR));
@@ -1391,7 +1370,10 @@ int i915_driver_irq_postinstall(struct drm_device *dev)
 	u32 enable_mask = I915_INTERRUPT_ENABLE_FIX | I915_INTERRUPT_ENABLE_VAR;
 	u32 error_mask;
 
-	DRM_INIT_WAITQUEUE(&dev_priv->irq_queue);
+	DRM_INIT_WAITQUEUE(&dev_priv->render_ring.irq_queue);
+
+	if (HAS_BSD(dev))
+		DRM_INIT_WAITQUEUE(&dev_priv->bsd_ring.irq_queue);
 
 	dev_priv->vblank_pipe = DRM_I915_VBLANK_PIPE_A | DRM_I915_VBLANK_PIPE_B;
 
diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index f3e39cc46f0d..64b0a3afd92b 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -334,6 +334,7 @@
 #define   I915_DEBUG_INTERRUPT				(1<<2)
 #define   I915_USER_INTERRUPT				(1<<1)
 #define   I915_ASLE_INTERRUPT				(1<<0)
+#define   I915_BSD_USER_INTERRUPT                      (1<<25)
 #define EIR		0x020b0
 #define EMR		0x020b4
 #define ESR		0x020b8
@@ -368,6 +369,36 @@
 #define BB_ADDR		0x02140 /* 8 bytes */
 #define GFX_FLSH_CNTL	0x02170 /* 915+ only */
 
+/* GEN6 interrupt control */
+#define GEN6_RENDER_HWSTAM	0x2098
+#define GEN6_RENDER_IMR		0x20a8
+#define   GEN6_RENDER_CONTEXT_SWITCH_INTERRUPT		(1 << 8)
+#define   GEN6_RENDER_PPGTT_PAGE_FAULT			(1 << 7)
+#define   GEN6_RENDER TIMEOUT_COUNTER_EXPIRED		(1 << 6)
+#define   GEN6_RENDER_L3_PARITY_ERROR			(1 << 5)
+#define   GEN6_RENDER_PIPE_CONTROL_NOTIFY_INTERRUPT	(1 << 4)
+#define   GEN6_RENDER_COMMAND_PARSER_MASTER_ERROR	(1 << 3)
+#define   GEN6_RENDER_SYNC_STATUS			(1 << 2)
+#define   GEN6_RENDER_DEBUG_INTERRUPT			(1 << 1)
+#define   GEN6_RENDER_USER_INTERRUPT			(1 << 0)
+
+#define GEN6_BLITTER_HWSTAM	0x22098
+#define GEN6_BLITTER_IMR	0x220a8
+#define   GEN6_BLITTER_MI_FLUSH_DW_NOTIFY_INTERRUPT	(1 << 26)
+#define   GEN6_BLITTER_COMMAND_PARSER_MASTER_ERROR	(1 << 25)
+#define   GEN6_BLITTER_SYNC_STATUS			(1 << 24)
+#define   GEN6_BLITTER_USER_INTERRUPT			(1 << 22)
+/*
+ * BSD (bit stream decoder instruction and interrupt control register defines
+ * (G4X and Ironlake only)
+ */
+
+#define BSD_RING_TAIL          0x04030
+#define BSD_RING_HEAD          0x04034
+#define BSD_RING_START         0x04038
+#define BSD_RING_CTL           0x0403c
+#define BSD_RING_ACTHD         0x04074
+#define BSD_HWS_PGA            0x04080
 
 /*
  * Framebuffer compression (915+ only)
@@ -805,6 +836,10 @@
 #define DCC_CHANNEL_XOR_DISABLE				(1 << 10)
 #define DCC_CHANNEL_XOR_BIT_17				(1 << 9)
 
+/** Pineview MCH register contains DDR3 setting */
+#define CSHRDDR3CTL            0x101a8
+#define CSHRDDR3CTL_DDR3       (1 << 2)
+
 /** 965 MCH register controlling DRAM channel configuration */
 #define C0DRB3			0x10206
 #define C1DRB3			0x10606
@@ -826,6 +861,12 @@
 #define CLKCFG_MEM_800					(3 << 4)
 #define CLKCFG_MEM_MASK					(7 << 4)
 
+#define TR1			0x11006
+#define TSFS			0x11020
+#define   TSFS_SLOPE_MASK	0x0000ff00
+#define   TSFS_SLOPE_SHIFT	8
+#define   TSFS_INTR_MASK	0x000000ff
+
 #define CRSTANDVID		0x11100
 #define PXVFREQ_BASE		0x11110 /* P[0-15]VIDFREQ (0x1114c) (Ironlake) */
 #define   PXVFREQ_PX_MASK	0x7f000000
@@ -964,6 +1005,41 @@
 #define   MEMSTAT_SRC_CTL_STDBY 3
 #define RCPREVBSYTUPAVG		0x113b8
 #define RCPREVBSYTDNAVG		0x113bc
+#define SDEW			0x1124c
+#define CSIEW0			0x11250
+#define CSIEW1			0x11254
+#define CSIEW2			0x11258
+#define PEW			0x1125c
+#define DEW			0x11270
+#define MCHAFE			0x112c0
+#define CSIEC			0x112e0
+#define DMIEC			0x112e4
+#define DDREC			0x112e8
+#define PEG0EC			0x112ec
+#define PEG1EC			0x112f0
+#define GFXEC			0x112f4
+#define RPPREVBSYTUPAVG		0x113b8
+#define RPPREVBSYTDNAVG		0x113bc
+#define ECR			0x11600
+#define   ECR_GPFE		(1<<31)
+#define   ECR_IMONE		(1<<30)
+#define   ECR_CAP_MASK		0x0000001f /* Event range, 0-31 */
+#define OGW0			0x11608
+#define OGW1			0x1160c
+#define EG0			0x11610
+#define EG1			0x11614
+#define EG2			0x11618
+#define EG3			0x1161c
+#define EG4			0x11620
+#define EG5			0x11624
+#define EG6			0x11628
+#define EG7			0x1162c
+#define PXW			0x11664
+#define PXWL			0x11680
+#define LCFUSE02		0x116c0
+#define   LCFUSE_HIV_MASK	0x000000ff
+#define CSIPLL0			0x12c10
+#define DDRMPLL1		0X12c20
 #define PEG_BAND_GAP_DATA	0x14d68
 
 /*
@@ -1055,7 +1131,6 @@
 #define CRT_HOTPLUG_DETECT_VOLTAGE_325MV	(0 << 2)
 #define CRT_HOTPLUG_DETECT_VOLTAGE_475MV	(1 << 2)
 #define CRT_HOTPLUG_MASK			(0x3fc) /* Bits 9-2 */
-#define CRT_FORCE_HOTPLUG_MASK			0xfffffe1f
 
 #define PORT_HOTPLUG_STAT	0x61114
 #define   HDMIB_HOTPLUG_INT_STATUS		(1 << 29)
@@ -2355,6 +2430,8 @@
 #define GT_PIPE_NOTIFY		(1 << 4)
 #define GT_SYNC_STATUS          (1 << 2)
 #define GT_USER_INTERRUPT       (1 << 0)
+#define GT_BSD_USER_INTERRUPT   (1 << 5)
+
 
 #define GTISR   0x44010
 #define GTIMR   0x44014
@@ -2690,6 +2767,9 @@
 #define  SDVO_ENCODING          (0)
 #define  TMDS_ENCODING          (2 << 10)
 #define  NULL_PACKET_VSYNC_ENABLE       (1 << 9)
+/* CPT */
+#define  HDMI_MODE_SELECT	(1 << 9)
+#define  DVI_MODE_SELECT	(0)
 #define  SDVOB_BORDER_ENABLE    (1 << 7)
 #define  AUDIO_ENABLE           (1 << 6)
 #define  VSYNC_ACTIVE_HIGH      (1 << 4)
diff --git a/drivers/gpu/drm/i915/i915_trace.h b/drivers/gpu/drm/i915/i915_trace.h
index 9e4c45f68d6e..fab21760dd57 100644
--- a/drivers/gpu/drm/i915/i915_trace.h
+++ b/drivers/gpu/drm/i915/i915_trace.h
@@ -53,23 +53,6 @@ TRACE_EVENT(i915_gem_object_bind,
 		      __entry->obj, __entry->gtt_offset)
 );
 
-TRACE_EVENT(i915_gem_object_clflush,
-
-	    TP_PROTO(struct drm_gem_object *obj),
-
-	    TP_ARGS(obj),
-
-	    TP_STRUCT__entry(
-			     __field(struct drm_gem_object *, obj)
-			     ),
-
-	    TP_fast_assign(
-			   __entry->obj = obj;
-			   ),
-
-	    TP_printk("obj=%p", __entry->obj)
-);
-
 TRACE_EVENT(i915_gem_object_change_domain,
 
 	    TP_PROTO(struct drm_gem_object *obj, uint32_t old_read_domains, uint32_t old_write_domain),
@@ -132,6 +115,13 @@ DECLARE_EVENT_CLASS(i915_gem_object,
 	    TP_printk("obj=%p", __entry->obj)
 );
 
+DEFINE_EVENT(i915_gem_object, i915_gem_object_clflush,
+
+	    TP_PROTO(struct drm_gem_object *obj),
+
+	    TP_ARGS(obj)
+);
+
 DEFINE_EVENT(i915_gem_object, i915_gem_object_unbind,
 
 	    TP_PROTO(struct drm_gem_object *obj),
diff --git a/drivers/gpu/drm/i915/intel_bios.c b/drivers/gpu/drm/i915/intel_bios.c
index 4c748d8f73d6..96f75d7f6633 100644
--- a/drivers/gpu/drm/i915/intel_bios.c
+++ b/drivers/gpu/drm/i915/intel_bios.c
@@ -95,6 +95,16 @@ fill_detail_timing_data(struct drm_display_mode *panel_fixed_mode,
 	panel_fixed_mode->clock = dvo_timing->clock * 10;
 	panel_fixed_mode->type = DRM_MODE_TYPE_PREFERRED;
 
+	if (dvo_timing->hsync_positive)
+		panel_fixed_mode->flags |= DRM_MODE_FLAG_PHSYNC;
+	else
+		panel_fixed_mode->flags |= DRM_MODE_FLAG_NHSYNC;
+
+	if (dvo_timing->vsync_positive)
+		panel_fixed_mode->flags |= DRM_MODE_FLAG_PVSYNC;
+	else
+		panel_fixed_mode->flags |= DRM_MODE_FLAG_NVSYNC;
+
 	/* Some VBTs have bogus h/vtotal values */
 	if (panel_fixed_mode->hsync_end > panel_fixed_mode->htotal)
 		panel_fixed_mode->htotal = panel_fixed_mode->hsync_end + 1;
diff --git a/drivers/gpu/drm/i915/intel_crt.c b/drivers/gpu/drm/i915/intel_crt.c
index e16ac5a28c3c..22ff38455731 100644
--- a/drivers/gpu/drm/i915/intel_crt.c
+++ b/drivers/gpu/drm/i915/intel_crt.c
@@ -217,7 +217,8 @@ static bool intel_crt_detect_hotplug(struct drm_connector *connector)
 {
 	struct drm_device *dev = connector->dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
-	u32 hotplug_en;
+	u32 hotplug_en, orig, stat;
+	bool ret = false;
 	int i, tries = 0;
 
 	if (HAS_PCH_SPLIT(dev))
@@ -232,8 +233,8 @@ static bool intel_crt_detect_hotplug(struct drm_connector *connector)
 		tries = 2;
 	else
 		tries = 1;
-	hotplug_en = I915_READ(PORT_HOTPLUG_EN);
-	hotplug_en &= CRT_FORCE_HOTPLUG_MASK;
+	hotplug_en = orig = I915_READ(PORT_HOTPLUG_EN);
+	hotplug_en &= CRT_HOTPLUG_MASK;
 	hotplug_en |= CRT_HOTPLUG_FORCE_DETECT;
 
 	if (IS_G4X(dev))
@@ -255,11 +256,17 @@ 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_NONE)
-		return true;
+	stat = I915_READ(PORT_HOTPLUG_STAT);
+	if ((stat & CRT_HOTPLUG_MONITOR_MASK) != CRT_HOTPLUG_MONITOR_NONE)
+		ret = true;
+
+	/* clear the interrupt we just generated, if any */
+	I915_WRITE(PORT_HOTPLUG_STAT, CRT_HOTPLUG_INT_STATUS);
 
-	return false;
+	/* and put the bits back */
+	I915_WRITE(PORT_HOTPLUG_EN, orig);
+
+	return ret;
 }
 
 static bool intel_crt_detect_ddc(struct drm_encoder *encoder)
@@ -569,7 +576,7 @@ void intel_crt_init(struct drm_device *dev)
 				   (1 << INTEL_ANALOG_CLONE_BIT) |
 				   (1 << INTEL_SDVO_LVDS_CLONE_BIT);
 	intel_encoder->crtc_mask = (1 << 0) | (1 << 1);
-	connector->interlace_allowed = 0;
+	connector->interlace_allowed = 1;
 	connector->doublescan_allowed = 0;
 
 	drm_encoder_helper_add(&intel_encoder->enc, &intel_crt_helper_funcs);
diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index f469a84cacfd..cc8131ff319f 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -1029,19 +1029,28 @@ static void i8xx_enable_fbc(struct drm_crtc *crtc, unsigned long interval)
 void i8xx_disable_fbc(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
+	unsigned long timeout = jiffies + msecs_to_jiffies(1);
 	u32 fbc_ctl;
 
 	if (!I915_HAS_FBC(dev))
 		return;
 
+	if (!(I915_READ(FBC_CONTROL) & FBC_CTL_EN))
+		return;	/* Already off, just return */
+
 	/* Disable compression */
 	fbc_ctl = I915_READ(FBC_CONTROL);
 	fbc_ctl &= ~FBC_CTL_EN;
 	I915_WRITE(FBC_CONTROL, fbc_ctl);
 
 	/* Wait for compressing bit to clear */
-	while (I915_READ(FBC_STATUS) & FBC_STAT_COMPRESSING)
-		; /* nothing */
+	while (I915_READ(FBC_STATUS) & FBC_STAT_COMPRESSING) {
+		if (time_after(jiffies, timeout)) {
+			DRM_DEBUG_DRIVER("FBC idle timed out\n");
+			break;
+		}
+		; /* do nothing */
+	}
 
 	intel_wait_for_vblank(dev);
 
@@ -1239,10 +1248,11 @@ static void intel_update_fbc(struct drm_crtc *crtc,
 	return;
 
 out_disable:
-	DRM_DEBUG_KMS("unsupported config, disabling FBC\n");
 	/* Multiple disables should be harmless */
-	if (intel_fbc_enabled(dev))
+	if (intel_fbc_enabled(dev)) {
+		DRM_DEBUG_KMS("unsupported config, disabling FBC\n");
 		intel_disable_fbc(dev);
+	}
 }
 
 static int
@@ -1386,7 +1396,8 @@ 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_KMS("Writing base %08lX %08lX %d %d\n", Start, Offset, x, y);
+	DRM_DEBUG_KMS("Writing base %08lX %08lX %d %d %d\n",
+		      Start, Offset, x, y, crtc->fb->pitch);
 	I915_WRITE(dspstride, crtc->fb->pitch);
 	if (IS_I965G(dev)) {
 		I915_WRITE(dspbase, Offset);
@@ -2345,6 +2356,8 @@ static bool intel_crtc_mode_fixup(struct drm_crtc *crtc,
 		if (mode->clock * 3 > 27000 * 4)
 			return MODE_CLOCK_HIGH;
 	}
+
+	drm_mode_set_crtcinfo(adjusted_mode, 0);
 	return true;
 }
 
@@ -2629,6 +2642,7 @@ static unsigned long intel_calculate_wm(unsigned long clock_in_khz,
 
 struct cxsr_latency {
 	int is_desktop;
+	int is_ddr3;
 	unsigned long fsb_freq;
 	unsigned long mem_freq;
 	unsigned long display_sr;
@@ -2638,33 +2652,45 @@ struct cxsr_latency {
 };
 
 static struct cxsr_latency cxsr_latency_table[] = {
-	{1, 800, 400, 3382, 33382, 3983, 33983},    /* DDR2-400 SC */
-	{1, 800, 667, 3354, 33354, 3807, 33807},    /* DDR2-667 SC */
-	{1, 800, 800, 3347, 33347, 3763, 33763},    /* DDR2-800 SC */
-
-	{1, 667, 400, 3400, 33400, 4021, 34021},    /* DDR2-400 SC */
-	{1, 667, 667, 3372, 33372, 3845, 33845},    /* DDR2-667 SC */
-	{1, 667, 800, 3386, 33386, 3822, 33822},    /* DDR2-800 SC */
-
-	{1, 400, 400, 3472, 33472, 4173, 34173},    /* DDR2-400 SC */
-	{1, 400, 667, 3443, 33443, 3996, 33996},    /* DDR2-667 SC */
-	{1, 400, 800, 3430, 33430, 3946, 33946},    /* DDR2-800 SC */
-
-	{0, 800, 400, 3438, 33438, 4065, 34065},    /* DDR2-400 SC */
-	{0, 800, 667, 3410, 33410, 3889, 33889},    /* DDR2-667 SC */
-	{0, 800, 800, 3403, 33403, 3845, 33845},    /* DDR2-800 SC */
-
-	{0, 667, 400, 3456, 33456, 4103, 34106},    /* DDR2-400 SC */
-	{0, 667, 667, 3428, 33428, 3927, 33927},    /* DDR2-667 SC */
-	{0, 667, 800, 3443, 33443, 3905, 33905},    /* DDR2-800 SC */
-
-	{0, 400, 400, 3528, 33528, 4255, 34255},    /* DDR2-400 SC */
-	{0, 400, 667, 3500, 33500, 4079, 34079},    /* DDR2-667 SC */
-	{0, 400, 800, 3487, 33487, 4029, 34029},    /* DDR2-800 SC */
+	{1, 0, 800, 400, 3382, 33382, 3983, 33983},    /* DDR2-400 SC */
+	{1, 0, 800, 667, 3354, 33354, 3807, 33807},    /* DDR2-667 SC */
+	{1, 0, 800, 800, 3347, 33347, 3763, 33763},    /* DDR2-800 SC */
+	{1, 1, 800, 667, 6420, 36420, 6873, 36873},    /* DDR3-667 SC */
+	{1, 1, 800, 800, 5902, 35902, 6318, 36318},    /* DDR3-800 SC */
+
+	{1, 0, 667, 400, 3400, 33400, 4021, 34021},    /* DDR2-400 SC */
+	{1, 0, 667, 667, 3372, 33372, 3845, 33845},    /* DDR2-667 SC */
+	{1, 0, 667, 800, 3386, 33386, 3822, 33822},    /* DDR2-800 SC */
+	{1, 1, 667, 667, 6438, 36438, 6911, 36911},    /* DDR3-667 SC */
+	{1, 1, 667, 800, 5941, 35941, 6377, 36377},    /* DDR3-800 SC */
+
+	{1, 0, 400, 400, 3472, 33472, 4173, 34173},    /* DDR2-400 SC */
+	{1, 0, 400, 667, 3443, 33443, 3996, 33996},    /* DDR2-667 SC */
+	{1, 0, 400, 800, 3430, 33430, 3946, 33946},    /* DDR2-800 SC */
+	{1, 1, 400, 667, 6509, 36509, 7062, 37062},    /* DDR3-667 SC */
+	{1, 1, 400, 800, 5985, 35985, 6501, 36501},    /* DDR3-800 SC */
+
+	{0, 0, 800, 400, 3438, 33438, 4065, 34065},    /* DDR2-400 SC */
+	{0, 0, 800, 667, 3410, 33410, 3889, 33889},    /* DDR2-667 SC */
+	{0, 0, 800, 800, 3403, 33403, 3845, 33845},    /* DDR2-800 SC */
+	{0, 1, 800, 667, 6476, 36476, 6955, 36955},    /* DDR3-667 SC */
+	{0, 1, 800, 800, 5958, 35958, 6400, 36400},    /* DDR3-800 SC */
+
+	{0, 0, 667, 400, 3456, 33456, 4103, 34106},    /* DDR2-400 SC */
+	{0, 0, 667, 667, 3428, 33428, 3927, 33927},    /* DDR2-667 SC */
+	{0, 0, 667, 800, 3443, 33443, 3905, 33905},    /* DDR2-800 SC */
+	{0, 1, 667, 667, 6494, 36494, 6993, 36993},    /* DDR3-667 SC */
+	{0, 1, 667, 800, 5998, 35998, 6460, 36460},    /* DDR3-800 SC */
+
+	{0, 0, 400, 400, 3528, 33528, 4255, 34255},    /* DDR2-400 SC */
+	{0, 0, 400, 667, 3500, 33500, 4079, 34079},    /* DDR2-667 SC */
+	{0, 0, 400, 800, 3487, 33487, 4029, 34029},    /* DDR2-800 SC */
+	{0, 1, 400, 667, 6566, 36566, 7145, 37145},    /* DDR3-667 SC */
+	{0, 1, 400, 800, 6042, 36042, 6584, 36584},    /* DDR3-800 SC */
 };
 
-static struct cxsr_latency *intel_get_cxsr_latency(int is_desktop, int fsb,
-						   int mem)
+static struct cxsr_latency *intel_get_cxsr_latency(int is_desktop, int is_ddr3, 
+						   int fsb, int mem)
 {
 	int i;
 	struct cxsr_latency *latency;
@@ -2675,6 +2701,7 @@ static struct cxsr_latency *intel_get_cxsr_latency(int is_desktop, int fsb,
 	for (i = 0; i < ARRAY_SIZE(cxsr_latency_table); i++) {
 		latency = &cxsr_latency_table[i];
 		if (is_desktop == latency->is_desktop &&
+		    is_ddr3 == latency->is_ddr3 &&
 		    fsb == latency->fsb_freq && mem == latency->mem_freq)
 			return latency;
 	}
@@ -2789,8 +2816,8 @@ static void pineview_update_wm(struct drm_device *dev,  int planea_clock,
 	struct cxsr_latency *latency;
 	int sr_clock;
 
-	latency = intel_get_cxsr_latency(IS_PINEVIEW_G(dev), dev_priv->fsb_freq,
-					 dev_priv->mem_freq);
+	latency = intel_get_cxsr_latency(IS_PINEVIEW_G(dev), dev_priv->is_ddr3, 
+					 dev_priv->fsb_freq, dev_priv->mem_freq);
 	if (!latency) {
 		DRM_DEBUG_KMS("Unknown FSB/MEM found, disable CxSR\n");
 		pineview_disable_cxsr(dev);
@@ -3626,6 +3653,11 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 			pipeconf &= ~PIPEACONF_DOUBLE_WIDE;
 	}
 
+	dspcntr |= DISPLAY_PLANE_ENABLE;
+	pipeconf |= PIPEACONF_ENABLE;
+	dpll |= DPLL_VCO_ENABLE;
+
+
 	/* Disable the panel fitter if it was on our pipe */
 	if (!HAS_PCH_SPLIT(dev) && intel_panel_fitter_pipe(dev) == pipe)
 		I915_WRITE(PFIT_CONTROL, 0);
@@ -3772,6 +3804,18 @@ static int intel_crtc_mode_set(struct drm_crtc *crtc,
 		}
 	}
 
+	if (adjusted_mode->flags & DRM_MODE_FLAG_INTERLACE) {
+		pipeconf |= PIPECONF_INTERLACE_W_FIELD_INDICATION;
+		/* the chip adds 2 halflines automatically */
+		adjusted_mode->crtc_vdisplay -= 1;
+		adjusted_mode->crtc_vtotal -= 1;
+		adjusted_mode->crtc_vblank_start -= 1;
+		adjusted_mode->crtc_vblank_end -= 1;
+		adjusted_mode->crtc_vsync_end -= 1;
+		adjusted_mode->crtc_vsync_start -= 1;
+	} else
+		pipeconf &= ~PIPECONF_INTERLACE_W_FIELD_INDICATION; /* progressive */
+
 	I915_WRITE(htot_reg, (adjusted_mode->crtc_hdisplay - 1) |
 		   ((adjusted_mode->crtc_htotal - 1) << 16));
 	I915_WRITE(hblank_reg, (adjusted_mode->crtc_hblank_start - 1) |
@@ -3934,6 +3978,13 @@ static int intel_crtc_cursor_set(struct drm_crtc *crtc,
 			DRM_ERROR("failed to pin cursor bo\n");
 			goto fail_locked;
 		}
+
+		ret = i915_gem_object_set_to_gtt_domain(bo, 0);
+		if (ret) {
+			DRM_ERROR("failed to move cursor bo into the GTT\n");
+			goto fail_unpin;
+		}
+
 		addr = obj_priv->gtt_offset;
 	} else {
 		ret = i915_gem_attach_phys_object(dev, bo, (pipe == 0) ? I915_GEM_PHYS_CURSOR_0 : I915_GEM_PHYS_CURSOR_1);
@@ -3977,6 +4028,8 @@ static int intel_crtc_cursor_set(struct drm_crtc *crtc,
 	intel_crtc->cursor_bo = bo;
 
 	return 0;
+fail_unpin:
+	i915_gem_object_unpin(bo);
 fail_locked:
 	mutex_unlock(&dev->struct_mutex);
 fail:
@@ -4436,6 +4489,8 @@ static void intel_idle_update(struct work_struct *work)
 
 	mutex_lock(&dev->struct_mutex);
 
+	i915_update_gfx_val(dev_priv);
+
 	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);
@@ -4564,12 +4619,6 @@ void intel_finish_page_flip(struct drm_device *dev, int pipe)
 	spin_lock_irqsave(&dev->event_lock, flags);
 	work = intel_crtc->unpin_work;
 	if (work == NULL || !work->pending) {
-		if (work && !work->pending) {
-			obj_priv = to_intel_bo(work->pending_flip_obj);
-			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;
 	}
@@ -4629,14 +4678,11 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,
 	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);
@@ -4646,10 +4692,10 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,
 	/* 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);
+
+		DRM_DEBUG_DRIVER("flip queue: crtc already busy\n");
 		return -EBUSY;
 	}
 	intel_crtc->unpin_work = work;
@@ -4658,13 +4704,19 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,
 	intel_fb = to_intel_framebuffer(fb);
 	obj = intel_fb->obj;
 
+	mutex_lock(&dev->struct_mutex);
 	ret = intel_pin_and_fence_fb_obj(dev, obj);
 	if (ret != 0) {
-		DRM_DEBUG_DRIVER("flip queue: %p pin & fence failed\n",
-			  to_intel_bo(obj));
-		kfree(work);
-		intel_crtc->unpin_work = NULL;
 		mutex_unlock(&dev->struct_mutex);
+
+		spin_lock_irqsave(&dev->event_lock, flags);
+		intel_crtc->unpin_work = NULL;
+		spin_unlock_irqrestore(&dev->event_lock, flags);
+
+		kfree(work);
+
+		DRM_DEBUG_DRIVER("flip queue: %p pin & fence failed\n",
+				 to_intel_bo(obj));
 		return ret;
 	}
 
@@ -5023,10 +5075,32 @@ err_unref:
 	return NULL;
 }
 
+bool ironlake_set_drps(struct drm_device *dev, u8 val)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	u16 rgvswctl;
+
+	rgvswctl = I915_READ16(MEMSWCTL);
+	if (rgvswctl & MEMCTL_CMD_STS) {
+		DRM_DEBUG("gpu busy, RCS change rejected\n");
+		return false; /* still busy with another command */
+	}
+
+	rgvswctl = (MEMCTL_CMD_CHFREQ << MEMCTL_CMD_SHIFT) |
+		(val << MEMCTL_FREQ_SHIFT) | MEMCTL_SFCAVM;
+	I915_WRITE16(MEMSWCTL, rgvswctl);
+	POSTING_READ16(MEMSWCTL);
+
+	rgvswctl |= MEMCTL_CMD_STS;
+	I915_WRITE16(MEMSWCTL, rgvswctl);
+
+	return true;
+}
+
 void ironlake_enable_drps(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
-	u32 rgvmodectl = I915_READ(MEMMODECTL), rgvswctl;
+	u32 rgvmodectl = I915_READ(MEMMODECTL);
 	u8 fmax, fmin, fstart, vstart;
 	int i = 0;
 
@@ -5045,13 +5119,21 @@ void ironlake_enable_drps(struct drm_device *dev)
 	fmin = (rgvmodectl & MEMMODE_FMIN_MASK);
 	fstart = (rgvmodectl & MEMMODE_FSTART_MASK) >>
 		MEMMODE_FSTART_SHIFT;
+	fstart = fmax;
+
 	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->fmax = fstart; /* IPS callback will increase this */
+	dev_priv->fstart = fstart;
+
+	dev_priv->max_delay = fmax;
 	dev_priv->min_delay = fmin;
 	dev_priv->cur_delay = fstart;
 
+	DRM_DEBUG_DRIVER("fmax: %d, fmin: %d, fstart: %d\n", fmax, fmin,
+			 fstart);
+
 	I915_WRITE(MEMINTREN, MEMINT_CX_SUPR_EN | MEMINT_EVAL_CHG_EN);
 
 	/*
@@ -5073,20 +5155,19 @@ void ironlake_enable_drps(struct drm_device *dev)
 	}
 	msleep(1);
 
-	rgvswctl = (MEMCTL_CMD_CHFREQ << MEMCTL_CMD_SHIFT) |
-		(fstart << MEMCTL_FREQ_SHIFT) | MEMCTL_SFCAVM;
-	I915_WRITE(MEMSWCTL, rgvswctl);
-	POSTING_READ(MEMSWCTL);
+	ironlake_set_drps(dev, fstart);
 
-	rgvswctl |= MEMCTL_CMD_STS;
-	I915_WRITE(MEMSWCTL, rgvswctl);
+	dev_priv->last_count1 = I915_READ(0x112e4) + I915_READ(0x112e8) +
+		I915_READ(0x112e0);
+	dev_priv->last_time1 = jiffies_to_msecs(jiffies);
+	dev_priv->last_count2 = I915_READ(0x112f4);
+	getrawmonotonic(&dev_priv->last_time2);
 }
 
 void ironlake_disable_drps(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
-	u32 rgvswctl;
-	u8 fstart;
+	u16 rgvswctl = I915_READ16(MEMSWCTL);
 
 	/* Ack interrupts, disable EFC interrupt */
 	I915_WRITE(MEMINTREN, I915_READ(MEMINTREN) & ~MEMINT_EVAL_CHG_EN);
@@ -5096,11 +5177,7 @@ void ironlake_disable_drps(struct drm_device *dev)
 	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);
+	ironlake_set_drps(dev, dev_priv->fstart);
 	msleep(1);
 	rgvswctl |= MEMCTL_CMD_STS;
 	I915_WRITE(MEMSWCTL, rgvswctl);
@@ -5108,6 +5185,92 @@ void ironlake_disable_drps(struct drm_device *dev)
 
 }
 
+static unsigned long intel_pxfreq(u32 vidfreq)
+{
+	unsigned long freq;
+	int div = (vidfreq & 0x3f0000) >> 16;
+	int post = (vidfreq & 0x3000) >> 12;
+	int pre = (vidfreq & 0x7);
+
+	if (!pre)
+		return 0;
+
+	freq = ((div * 133333) / ((1<<post) * pre));
+
+	return freq;
+}
+
+void intel_init_emon(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	u32 lcfuse;
+	u8 pxw[16];
+	int i;
+
+	/* Disable to program */
+	I915_WRITE(ECR, 0);
+	POSTING_READ(ECR);
+
+	/* Program energy weights for various events */
+	I915_WRITE(SDEW, 0x15040d00);
+	I915_WRITE(CSIEW0, 0x007f0000);
+	I915_WRITE(CSIEW1, 0x1e220004);
+	I915_WRITE(CSIEW2, 0x04000004);
+
+	for (i = 0; i < 5; i++)
+		I915_WRITE(PEW + (i * 4), 0);
+	for (i = 0; i < 3; i++)
+		I915_WRITE(DEW + (i * 4), 0);
+
+	/* Program P-state weights to account for frequency power adjustment */
+	for (i = 0; i < 16; i++) {
+		u32 pxvidfreq = I915_READ(PXVFREQ_BASE + (i * 4));
+		unsigned long freq = intel_pxfreq(pxvidfreq);
+		unsigned long vid = (pxvidfreq & PXVFREQ_PX_MASK) >>
+			PXVFREQ_PX_SHIFT;
+		unsigned long val;
+
+		val = vid * vid;
+		val *= (freq / 1000);
+		val *= 255;
+		val /= (127*127*900);
+		if (val > 0xff)
+			DRM_ERROR("bad pxval: %ld\n", val);
+		pxw[i] = val;
+	}
+	/* Render standby states get 0 weight */
+	pxw[14] = 0;
+	pxw[15] = 0;
+
+	for (i = 0; i < 4; i++) {
+		u32 val = (pxw[i*4] << 24) | (pxw[(i*4)+1] << 16) |
+			(pxw[(i*4)+2] << 8) | (pxw[(i*4)+3]);
+		I915_WRITE(PXW + (i * 4), val);
+	}
+
+	/* Adjust magic regs to magic values (more experimental results) */
+	I915_WRITE(OGW0, 0);
+	I915_WRITE(OGW1, 0);
+	I915_WRITE(EG0, 0x00007f00);
+	I915_WRITE(EG1, 0x0000000e);
+	I915_WRITE(EG2, 0x000e0000);
+	I915_WRITE(EG3, 0x68000300);
+	I915_WRITE(EG4, 0x42000000);
+	I915_WRITE(EG5, 0x00140031);
+	I915_WRITE(EG6, 0);
+	I915_WRITE(EG7, 0);
+
+	for (i = 0; i < 8; i++)
+		I915_WRITE(PXWL + (i * 4), 0);
+
+	/* Enable PMON + select events */
+	I915_WRITE(ECR, 0x80000019);
+
+	lcfuse = I915_READ(LCFUSE02);
+
+	dev_priv->corr = (lcfuse & LCFUSE_HIV_MASK);
+}
+
 void intel_init_clock_gating(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -5277,11 +5440,13 @@ static void intel_init_display(struct drm_device *dev)
 			dev_priv->display.update_wm = NULL;
 	} else if (IS_PINEVIEW(dev)) {
 		if (!intel_get_cxsr_latency(IS_PINEVIEW_G(dev),
+					    dev_priv->is_ddr3,
 					    dev_priv->fsb_freq,
 					    dev_priv->mem_freq)) {
 			DRM_INFO("failed to find known CxSR latency "
-				 "(found fsb freq %d, mem freq %d), "
+				 "(found ddr%s fsb freq %d, mem freq %d), "
 				 "disabling CxSR\n",
+				 (dev_priv->is_ddr3 == 1) ? "3": "2",
 				 dev_priv->fsb_freq, dev_priv->mem_freq);
 			/* Disable CxSR and never update its watermark again */
 			pineview_disable_cxsr(dev);
@@ -5310,7 +5475,6 @@ static void intel_init_display(struct drm_device *dev)
 void intel_modeset_init(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
-	int num_pipe;
 	int i;
 
 	drm_mode_config_init(dev);
@@ -5340,13 +5504,13 @@ void intel_modeset_init(struct drm_device *dev)
 		dev->mode_config.fb_base = pci_resource_start(dev->pdev, 0);
 
 	if (IS_MOBILE(dev) || IS_I9XX(dev))
-		num_pipe = 2;
+		dev_priv->num_pipe = 2;
 	else
-		num_pipe = 1;
+		dev_priv->num_pipe = 1;
 	DRM_DEBUG_KMS("%d display pipe%s available.\n",
-		  num_pipe, num_pipe > 1 ? "s" : "");
+		      dev_priv->num_pipe, dev_priv->num_pipe > 1 ? "s" : "");
 
-	for (i = 0; i < num_pipe; i++) {
+	for (i = 0; i < dev_priv->num_pipe; i++) {
 		intel_crtc_init(dev, i);
 	}
 
@@ -5354,8 +5518,10 @@ void intel_modeset_init(struct drm_device *dev)
 
 	intel_init_clock_gating(dev);
 
-	if (IS_IRONLAKE_M(dev))
+	if (IS_IRONLAKE_M(dev)) {
 		ironlake_enable_drps(dev);
+		intel_init_emon(dev);
+	}
 
 	INIT_WORK(&dev_priv->idle_work, intel_idle_update);
 	setup_timer(&dev_priv->idle_timer, intel_gpu_idle_timer,
diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c
index 6b1c9a27c27a..49b54f05d3cf 100644
--- a/drivers/gpu/drm/i915/intel_dp.c
+++ b/drivers/gpu/drm/i915/intel_dp.c
@@ -576,7 +576,7 @@ intel_dp_set_m_n(struct drm_crtc *crtc, struct drm_display_mode *mode,
 		struct intel_encoder *intel_encoder;
 		struct intel_dp_priv *dp_priv;
 
-		if (!encoder || encoder->crtc != crtc)
+		if (encoder->crtc != crtc)
 			continue;
 
 		intel_encoder = enc_to_intel_encoder(encoder);
@@ -675,10 +675,9 @@ intel_dp_mode_set(struct drm_encoder *encoder, struct drm_display_mode *mode,
 	dp_priv->link_configuration[1] = dp_priv->lane_count;
 
 	/*
-	 * Check for DPCD version > 1.1,
-	 * enable enahanced frame stuff in that case
+	 * Check for DPCD version > 1.1 and enhanced framing support
 	 */
-	if (dp_priv->dpcd[0] >= 0x11) {
+	if (dp_priv->dpcd[0] >= 0x11 && (dp_priv->dpcd[2] & DP_ENHANCED_FRAME_CAP)) {
 		dp_priv->link_configuration[1] |= DP_LANE_COUNT_ENHANCED_FRAME_EN;
 		dp_priv->DP |= DP_ENHANCED_FRAMING;
 	}
@@ -1208,6 +1207,8 @@ ironlake_dp_detect(struct drm_connector *connector)
 		if (dp_priv->dpcd[0] != 0)
 			status = connector_status_connected;
 	}
+	DRM_DEBUG_KMS("DPCD: %hx%hx%hx%hx\n", dp_priv->dpcd[0],
+		      dp_priv->dpcd[1], dp_priv->dpcd[2], dp_priv->dpcd[3]);
 	return status;
 }
 
@@ -1352,7 +1353,7 @@ intel_trans_dp_port_sel (struct drm_crtc *crtc)
 	struct intel_encoder *intel_encoder = NULL;
 
 	list_for_each_entry(encoder, &mode_config->encoder_list, head) {
-		if (!encoder || encoder->crtc != crtc)
+		if (encoder->crtc != crtc)
 			continue;
 
 		intel_encoder = enc_to_intel_encoder(encoder);
diff --git a/drivers/gpu/drm/i915/intel_fb.c b/drivers/gpu/drm/i915/intel_fb.c
index 6f53cf7fbc50..c3c505244e07 100644
--- a/drivers/gpu/drm/i915/intel_fb.c
+++ b/drivers/gpu/drm/i915/intel_fb.c
@@ -105,7 +105,11 @@ static int intelfb_create(struct intel_fbdev *ifbdev,
 	}
 
 	/* Flush everything out, we'll be doing GTT only from now on */
-	i915_gem_object_set_to_gtt_domain(fbo, 1);
+	ret = i915_gem_object_set_to_gtt_domain(fbo, 1);
+	if (ret) {
+		DRM_ERROR("failed to bind fb: %d.\n", ret);
+		goto out_unpin;
+	}
 
 	info = framebuffer_alloc(0, device);
 	if (!info) {
@@ -241,6 +245,7 @@ int intel_fbdev_init(struct drm_device *dev)
 {
 	struct intel_fbdev *ifbdev;
 	drm_i915_private_t *dev_priv = dev->dev_private;
+	int ret;
 
 	ifbdev = kzalloc(sizeof(struct intel_fbdev), GFP_KERNEL);
 	if (!ifbdev)
@@ -249,8 +254,13 @@ int intel_fbdev_init(struct drm_device *dev)
 	dev_priv->fbdev = ifbdev;
 	ifbdev->helper.funcs = &intel_fb_helper_funcs;
 
-	drm_fb_helper_init(dev, &ifbdev->helper, 2,
-			   INTELFB_CONN_LIMIT);
+	ret = drm_fb_helper_init(dev, &ifbdev->helper,
+				 dev_priv->num_pipe,
+				 INTELFB_CONN_LIMIT);
+	if (ret) {
+		kfree(ifbdev);
+		return ret;
+	}
 
 	drm_fb_helper_single_add_all_connectors(&ifbdev->helper);
 	drm_fb_helper_initial_config(&ifbdev->helper, 32);
diff --git a/drivers/gpu/drm/i915/intel_hdmi.c b/drivers/gpu/drm/i915/intel_hdmi.c
index 65727f0a79a3..83bd764b000e 100644
--- a/drivers/gpu/drm/i915/intel_hdmi.c
+++ b/drivers/gpu/drm/i915/intel_hdmi.c
@@ -59,8 +59,11 @@ static void intel_hdmi_mode_set(struct drm_encoder *encoder,
 		SDVO_VSYNC_ACTIVE_HIGH |
 		SDVO_HSYNC_ACTIVE_HIGH;
 
-	if (hdmi_priv->has_hdmi_sink)
+	if (hdmi_priv->has_hdmi_sink) {
 		sdvox |= SDVO_AUDIO_ENABLE;
+		if (HAS_PCH_CPT(dev))
+			sdvox |= HDMI_MODE_SELECT;
+	}
 
 	if (intel_crtc->pipe == 1) {
 		if (HAS_PCH_CPT(dev))
diff --git a/drivers/gpu/drm/i915/intel_overlay.c b/drivers/gpu/drm/i915/intel_overlay.c
index b0e17b06eb6e..d7ad5139d17c 100644
--- a/drivers/gpu/drm/i915/intel_overlay.c
+++ b/drivers/gpu/drm/i915/intel_overlay.c
@@ -211,9 +211,8 @@ static void intel_overlay_unmap_regs_atomic(struct intel_overlay *overlay)
 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;
+	drm_i915_private_t *dev_priv = dev->dev_private;
 
 	BUG_ON(overlay->active);
 
@@ -227,11 +226,13 @@ static int intel_overlay_on(struct intel_overlay *overlay)
 	OUT_RING(MI_NOOP);
 	ADVANCE_LP_RING();
 
-	overlay->last_flip_req = i915_add_request(dev, NULL, 0);
+	overlay->last_flip_req =
+		i915_add_request(dev, NULL, 0, &dev_priv->render_ring);
 	if (overlay->last_flip_req == 0)
 		return -ENOMEM;
 
-	ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
+	ret = i915_do_wait_request(dev,
+			overlay->last_flip_req, 1, &dev_priv->render_ring);
 	if (ret != 0)
 		return ret;
 
@@ -248,7 +249,6 @@ static void intel_overlay_continue(struct intel_overlay *overlay,
         drm_i915_private_t *dev_priv = dev->dev_private;
 	u32 flip_addr = overlay->flip_addr;
 	u32 tmp;
-	RING_LOCALS;
 
 	BUG_ON(!overlay->active);
 
@@ -265,7 +265,8 @@ static void intel_overlay_continue(struct intel_overlay *overlay,
 	OUT_RING(flip_addr);
         ADVANCE_LP_RING();
 
-	overlay->last_flip_req = i915_add_request(dev, NULL, 0);
+	overlay->last_flip_req =
+		i915_add_request(dev, NULL, 0, &dev_priv->render_ring);
 }
 
 static int intel_overlay_wait_flip(struct intel_overlay *overlay)
@@ -274,10 +275,10 @@ static int intel_overlay_wait_flip(struct intel_overlay *overlay)
         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);
+		ret = i915_do_wait_request(dev, overlay->last_flip_req,
+				1, &dev_priv->render_ring);
 		if (ret == 0) {
 			overlay->last_flip_req = 0;
 
@@ -296,11 +297,13 @@ static int intel_overlay_wait_flip(struct intel_overlay *overlay)
         OUT_RING(MI_NOOP);
         ADVANCE_LP_RING();
 
-	overlay->last_flip_req = i915_add_request(dev, NULL, 0);
+	overlay->last_flip_req =
+		i915_add_request(dev, NULL, 0, &dev_priv->render_ring);
 	if (overlay->last_flip_req == 0)
 		return -ENOMEM;
 
-	ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
+	ret = i915_do_wait_request(dev, overlay->last_flip_req,
+			1, &dev_priv->render_ring);
 	if (ret != 0)
 		return ret;
 
@@ -314,9 +317,8 @@ 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;
+	drm_i915_private_t *dev_priv = dev->dev_private;
 	int ret;
-	RING_LOCALS;
 
 	BUG_ON(!overlay->active);
 
@@ -336,11 +338,13 @@ static int intel_overlay_off(struct intel_overlay *overlay)
         OUT_RING(MI_NOOP);
         ADVANCE_LP_RING();
 
-	overlay->last_flip_req = i915_add_request(dev, NULL, 0);
+	overlay->last_flip_req =
+		i915_add_request(dev, NULL, 0, &dev_priv->render_ring);
 	if (overlay->last_flip_req == 0)
 		return -ENOMEM;
 
-	ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
+	ret = i915_do_wait_request(dev, overlay->last_flip_req,
+			1, &dev_priv->render_ring);
 	if (ret != 0)
 		return ret;
 
@@ -354,11 +358,13 @@ static int intel_overlay_off(struct intel_overlay *overlay)
         OUT_RING(MI_NOOP);
 	ADVANCE_LP_RING();
 
-	overlay->last_flip_req = i915_add_request(dev, NULL, 0);
+	overlay->last_flip_req =
+		i915_add_request(dev, NULL, 0, &dev_priv->render_ring);
 	if (overlay->last_flip_req == 0)
 		return -ENOMEM;
 
-	ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
+	ret = i915_do_wait_request(dev, overlay->last_flip_req,
+			1, &dev_priv->render_ring);
 	if (ret != 0)
 		return ret;
 
@@ -390,22 +396,23 @@ 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;
+	drm_i915_private_t *dev_priv = dev->dev_private;
 	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);
+		overlay->last_flip_req =
+			i915_add_request(dev, NULL, 0, &dev_priv->render_ring);
 		if (overlay->last_flip_req == 0)
 			return -ENOMEM;
 	}
 
-	ret = i915_do_wait_request(dev, overlay->last_flip_req, interruptible);
+	ret = i915_do_wait_request(dev, overlay->last_flip_req,
+			interruptible, &dev_priv->render_ring);
 	if (ret != 0)
 		return ret;
 
@@ -429,12 +436,13 @@ int intel_overlay_recover_from_interrupt(struct intel_overlay *overlay,
 			OUT_RING(MI_NOOP);
 			ADVANCE_LP_RING();
 
-			overlay->last_flip_req = i915_add_request(dev, NULL, 0);
+			overlay->last_flip_req = i915_add_request(dev, NULL,
+					0, &dev_priv->render_ring);
 			if (overlay->last_flip_req == 0)
 				return -ENOMEM;
 
 			ret = i915_do_wait_request(dev, overlay->last_flip_req,
-					interruptible);
+					interruptible, &dev_priv->render_ring);
 			if (ret != 0)
 				return ret;
 
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
new file mode 100644
index 000000000000..cea4f1a8709e
--- /dev/null
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -0,0 +1,849 @@
+/*
+ * Copyright © 2008-2010 Intel Corporation
+ *
+ * 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:
+ *    Eric Anholt <eric@anholt.net>
+ *    Zou Nan hai <nanhai.zou@intel.com>
+ *    Xiang Hai hao<haihao.xiang@intel.com>
+ *
+ */
+
+#include "drmP.h"
+#include "drm.h"
+#include "i915_drv.h"
+#include "i915_drm.h"
+#include "i915_trace.h"
+
+static void
+render_ring_flush(struct drm_device *dev,
+		struct intel_ring_buffer *ring,
+		u32	invalidate_domains,
+		u32	flush_domains)
+{
+#if WATCH_EXEC
+	DRM_INFO("%s: invalidate %08x flush %08x\n", __func__,
+		  invalidate_domains, flush_domains);
+#endif
+	u32 cmd;
+	trace_i915_gem_request_flush(dev, ring->next_seqno,
+				     invalidate_domains, flush_domains);
+
+	if ((invalidate_domains | flush_domains) & I915_GEM_GPU_DOMAINS) {
+		/*
+		 * read/write caches:
+		 *
+		 * I915_GEM_DOMAIN_RENDER is always invalidated, but is
+		 * only flushed if MI_NO_WRITE_FLUSH is unset.  On 965, it is
+		 * also flushed at 2d versus 3d pipeline switches.
+		 *
+		 * read-only caches:
+		 *
+		 * I915_GEM_DOMAIN_SAMPLER is flushed on pre-965 if
+		 * MI_READ_FLUSH is set, and is always flushed on 965.
+		 *
+		 * I915_GEM_DOMAIN_COMMAND may not exist?
+		 *
+		 * I915_GEM_DOMAIN_INSTRUCTION, which exists on 965, is
+		 * invalidated when MI_EXE_FLUSH is set.
+		 *
+		 * I915_GEM_DOMAIN_VERTEX, which exists on 965, is
+		 * invalidated with every MI_FLUSH.
+		 *
+		 * TLBs:
+		 *
+		 * On 965, TLBs associated with I915_GEM_DOMAIN_COMMAND
+		 * and I915_GEM_DOMAIN_CPU in are invalidated at PTE write and
+		 * I915_GEM_DOMAIN_RENDER and I915_GEM_DOMAIN_SAMPLER
+		 * are flushed at any MI_FLUSH.
+		 */
+
+		cmd = MI_FLUSH | MI_NO_WRITE_FLUSH;
+		if ((invalidate_domains|flush_domains) &
+		    I915_GEM_DOMAIN_RENDER)
+			cmd &= ~MI_NO_WRITE_FLUSH;
+		if (!IS_I965G(dev)) {
+			/*
+			 * On the 965, the sampler cache always gets flushed
+			 * and this bit is reserved.
+			 */
+			if (invalidate_domains & I915_GEM_DOMAIN_SAMPLER)
+				cmd |= MI_READ_FLUSH;
+		}
+		if (invalidate_domains & I915_GEM_DOMAIN_INSTRUCTION)
+			cmd |= MI_EXE_FLUSH;
+
+#if WATCH_EXEC
+		DRM_INFO("%s: queue flush %08x to ring\n", __func__, cmd);
+#endif
+		intel_ring_begin(dev, ring, 8);
+		intel_ring_emit(dev, ring, cmd);
+		intel_ring_emit(dev, ring, MI_NOOP);
+		intel_ring_advance(dev, ring);
+	}
+}
+
+static unsigned int render_ring_get_head(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	return I915_READ(PRB0_HEAD) & HEAD_ADDR;
+}
+
+static unsigned int render_ring_get_tail(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	return I915_READ(PRB0_TAIL) & TAIL_ADDR;
+}
+
+static unsigned int render_ring_get_active_head(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	u32 acthd_reg = IS_I965G(dev) ? ACTHD_I965 : ACTHD;
+
+	return I915_READ(acthd_reg);
+}
+
+static void render_ring_advance_ring(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	I915_WRITE(PRB0_TAIL, ring->tail);
+}
+
+static int init_ring_common(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	u32 head;
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	struct drm_i915_gem_object *obj_priv;
+	obj_priv = to_intel_bo(ring->gem_object);
+
+	/* Stop the ring if it's running. */
+	I915_WRITE(ring->regs.ctl, 0);
+	I915_WRITE(ring->regs.head, 0);
+	I915_WRITE(ring->regs.tail, 0);
+
+	/* Initialize the ring. */
+	I915_WRITE(ring->regs.start, obj_priv->gtt_offset);
+	head = ring->get_head(dev, ring);
+
+	/* G45 ring initialization fails to reset head to zero */
+	if (head != 0) {
+		DRM_ERROR("%s head not reset to zero "
+				"ctl %08x head %08x tail %08x start %08x\n",
+				ring->name,
+				I915_READ(ring->regs.ctl),
+				I915_READ(ring->regs.head),
+				I915_READ(ring->regs.tail),
+				I915_READ(ring->regs.start));
+
+		I915_WRITE(ring->regs.head, 0);
+
+		DRM_ERROR("%s head forced to zero "
+				"ctl %08x head %08x tail %08x start %08x\n",
+				ring->name,
+				I915_READ(ring->regs.ctl),
+				I915_READ(ring->regs.head),
+				I915_READ(ring->regs.tail),
+				I915_READ(ring->regs.start));
+	}
+
+	I915_WRITE(ring->regs.ctl,
+			((ring->gem_object->size - PAGE_SIZE) & RING_NR_PAGES)
+			| RING_NO_REPORT | RING_VALID);
+
+	head = I915_READ(ring->regs.head) & HEAD_ADDR;
+	/* If the head is still not zero, the ring is dead */
+	if (head != 0) {
+		DRM_ERROR("%s initialization failed "
+				"ctl %08x head %08x tail %08x start %08x\n",
+				ring->name,
+				I915_READ(ring->regs.ctl),
+				I915_READ(ring->regs.head),
+				I915_READ(ring->regs.tail),
+				I915_READ(ring->regs.start));
+		return -EIO;
+	}
+
+	if (!drm_core_check_feature(dev, DRIVER_MODESET))
+		i915_kernel_lost_context(dev);
+	else {
+		ring->head = ring->get_head(dev, ring);
+		ring->tail = ring->get_tail(dev, ring);
+		ring->space = ring->head - (ring->tail + 8);
+		if (ring->space < 0)
+			ring->space += ring->size;
+	}
+	return 0;
+}
+
+static int init_render_ring(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	int ret = init_ring_common(dev, ring);
+	if (IS_I9XX(dev) && !IS_GEN3(dev)) {
+		I915_WRITE(MI_MODE,
+				(VS_TIMER_DISPATCH) << 16 | VS_TIMER_DISPATCH);
+	}
+	return ret;
+}
+
+#define PIPE_CONTROL_FLUSH(addr)					\
+do {									\
+	OUT_RING(GFX_OP_PIPE_CONTROL | PIPE_CONTROL_QW_WRITE |		\
+		 PIPE_CONTROL_DEPTH_STALL | 2);				\
+	OUT_RING(addr | PIPE_CONTROL_GLOBAL_GTT);			\
+	OUT_RING(0);							\
+	OUT_RING(0);							\
+} while (0)
+
+/**
+ * Creates a new sequence number, emitting a write of it to the status page
+ * plus an interrupt, which will trigger i915_user_interrupt_handler.
+ *
+ * Must be called with struct_lock held.
+ *
+ * Returned sequence numbers are nonzero on success.
+ */
+static u32
+render_ring_add_request(struct drm_device *dev,
+		struct intel_ring_buffer *ring,
+		struct drm_file *file_priv,
+		u32 flush_domains)
+{
+	u32 seqno;
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	seqno = intel_ring_get_seqno(dev, ring);
+
+	if (IS_GEN6(dev)) {
+		BEGIN_LP_RING(6);
+		OUT_RING(GFX_OP_PIPE_CONTROL | 3);
+		OUT_RING(PIPE_CONTROL_QW_WRITE |
+			 PIPE_CONTROL_WC_FLUSH | PIPE_CONTROL_IS_FLUSH |
+			 PIPE_CONTROL_NOTIFY);
+		OUT_RING(dev_priv->seqno_gfx_addr | PIPE_CONTROL_GLOBAL_GTT);
+		OUT_RING(seqno);
+		OUT_RING(0);
+		OUT_RING(0);
+		ADVANCE_LP_RING();
+	} else if (HAS_PIPE_CONTROL(dev)) {
+		u32 scratch_addr = dev_priv->seqno_gfx_addr + 128;
+
+		/*
+		 * Workaround qword write incoherence by flushing the
+		 * PIPE_NOTIFY buffers out to memory before requesting
+		 * an interrupt.
+		 */
+		BEGIN_LP_RING(32);
+		OUT_RING(GFX_OP_PIPE_CONTROL | PIPE_CONTROL_QW_WRITE |
+			 PIPE_CONTROL_WC_FLUSH | PIPE_CONTROL_TC_FLUSH);
+		OUT_RING(dev_priv->seqno_gfx_addr | PIPE_CONTROL_GLOBAL_GTT);
+		OUT_RING(seqno);
+		OUT_RING(0);
+		PIPE_CONTROL_FLUSH(scratch_addr);
+		scratch_addr += 128; /* write to separate cachelines */
+		PIPE_CONTROL_FLUSH(scratch_addr);
+		scratch_addr += 128;
+		PIPE_CONTROL_FLUSH(scratch_addr);
+		scratch_addr += 128;
+		PIPE_CONTROL_FLUSH(scratch_addr);
+		scratch_addr += 128;
+		PIPE_CONTROL_FLUSH(scratch_addr);
+		scratch_addr += 128;
+		PIPE_CONTROL_FLUSH(scratch_addr);
+		OUT_RING(GFX_OP_PIPE_CONTROL | PIPE_CONTROL_QW_WRITE |
+			 PIPE_CONTROL_WC_FLUSH | PIPE_CONTROL_TC_FLUSH |
+			 PIPE_CONTROL_NOTIFY);
+		OUT_RING(dev_priv->seqno_gfx_addr | PIPE_CONTROL_GLOBAL_GTT);
+		OUT_RING(seqno);
+		OUT_RING(0);
+		ADVANCE_LP_RING();
+	} else {
+		BEGIN_LP_RING(4);
+		OUT_RING(MI_STORE_DWORD_INDEX);
+		OUT_RING(I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT);
+		OUT_RING(seqno);
+
+		OUT_RING(MI_USER_INTERRUPT);
+		ADVANCE_LP_RING();
+	}
+	return seqno;
+}
+
+static u32
+render_ring_get_gem_seqno(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
+	if (HAS_PIPE_CONTROL(dev))
+		return ((volatile u32 *)(dev_priv->seqno_page))[0];
+	else
+		return intel_read_status_page(ring, I915_GEM_HWS_INDEX);
+}
+
+static void
+render_ring_get_user_irq(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
+	unsigned long irqflags;
+
+	spin_lock_irqsave(&dev_priv->user_irq_lock, irqflags);
+	if (dev->irq_enabled && (++ring->user_irq_refcount == 1)) {
+		if (HAS_PCH_SPLIT(dev))
+			ironlake_enable_graphics_irq(dev_priv, GT_PIPE_NOTIFY);
+		else
+			i915_enable_irq(dev_priv, I915_USER_INTERRUPT);
+	}
+	spin_unlock_irqrestore(&dev_priv->user_irq_lock, irqflags);
+}
+
+static void
+render_ring_put_user_irq(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
+	unsigned long irqflags;
+
+	spin_lock_irqsave(&dev_priv->user_irq_lock, irqflags);
+	BUG_ON(dev->irq_enabled && ring->user_irq_refcount <= 0);
+	if (dev->irq_enabled && (--ring->user_irq_refcount == 0)) {
+		if (HAS_PCH_SPLIT(dev))
+			ironlake_disable_graphics_irq(dev_priv, GT_PIPE_NOTIFY);
+		else
+			i915_disable_irq(dev_priv, I915_USER_INTERRUPT);
+	}
+	spin_unlock_irqrestore(&dev_priv->user_irq_lock, irqflags);
+}
+
+static void render_setup_status_page(struct drm_device *dev,
+	struct	intel_ring_buffer *ring)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	if (IS_GEN6(dev)) {
+		I915_WRITE(HWS_PGA_GEN6, ring->status_page.gfx_addr);
+		I915_READ(HWS_PGA_GEN6); /* posting read */
+	} else {
+		I915_WRITE(HWS_PGA, ring->status_page.gfx_addr);
+		I915_READ(HWS_PGA); /* posting read */
+	}
+
+}
+
+void
+bsd_ring_flush(struct drm_device *dev,
+		struct intel_ring_buffer *ring,
+		u32     invalidate_domains,
+		u32     flush_domains)
+{
+	intel_ring_begin(dev, ring, 8);
+	intel_ring_emit(dev, ring, MI_FLUSH);
+	intel_ring_emit(dev, ring, MI_NOOP);
+	intel_ring_advance(dev, ring);
+}
+
+static inline unsigned int bsd_ring_get_head(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	return I915_READ(BSD_RING_HEAD) & HEAD_ADDR;
+}
+
+static inline unsigned int bsd_ring_get_tail(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	return I915_READ(BSD_RING_TAIL) & TAIL_ADDR;
+}
+
+static inline unsigned int bsd_ring_get_active_head(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	return I915_READ(BSD_RING_ACTHD);
+}
+
+static inline void bsd_ring_advance_ring(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	I915_WRITE(BSD_RING_TAIL, ring->tail);
+}
+
+static int init_bsd_ring(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	return init_ring_common(dev, ring);
+}
+
+static u32
+bsd_ring_add_request(struct drm_device *dev,
+		struct intel_ring_buffer *ring,
+		struct drm_file *file_priv,
+		u32 flush_domains)
+{
+	u32 seqno;
+	seqno = intel_ring_get_seqno(dev, ring);
+	intel_ring_begin(dev, ring, 4);
+	intel_ring_emit(dev, ring, MI_STORE_DWORD_INDEX);
+	intel_ring_emit(dev, ring,
+			I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT);
+	intel_ring_emit(dev, ring, seqno);
+	intel_ring_emit(dev, ring, MI_USER_INTERRUPT);
+	intel_ring_advance(dev, ring);
+
+	DRM_DEBUG_DRIVER("%s %d\n", ring->name, seqno);
+
+	return seqno;
+}
+
+static void bsd_setup_status_page(struct drm_device *dev,
+		struct  intel_ring_buffer *ring)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	I915_WRITE(BSD_HWS_PGA, ring->status_page.gfx_addr);
+	I915_READ(BSD_HWS_PGA);
+}
+
+static void
+bsd_ring_get_user_irq(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	/* do nothing */
+}
+static void
+bsd_ring_put_user_irq(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	/* do nothing */
+}
+
+static u32
+bsd_ring_get_gem_seqno(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	return intel_read_status_page(ring, I915_GEM_HWS_INDEX);
+}
+
+static int
+bsd_ring_dispatch_gem_execbuffer(struct drm_device *dev,
+		struct intel_ring_buffer *ring,
+		struct drm_i915_gem_execbuffer2 *exec,
+		struct drm_clip_rect *cliprects,
+		uint64_t exec_offset)
+{
+	uint32_t exec_start;
+	exec_start = (uint32_t) exec_offset + exec->batch_start_offset;
+	intel_ring_begin(dev, ring, 2);
+	intel_ring_emit(dev, ring, MI_BATCH_BUFFER_START |
+			(2 << 6) | MI_BATCH_NON_SECURE_I965);
+	intel_ring_emit(dev, ring, exec_start);
+	intel_ring_advance(dev, ring);
+	return 0;
+}
+
+
+static int
+render_ring_dispatch_gem_execbuffer(struct drm_device *dev,
+		struct intel_ring_buffer *ring,
+		struct drm_i915_gem_execbuffer2 *exec,
+		struct drm_clip_rect *cliprects,
+		uint64_t exec_offset)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	int nbox = exec->num_cliprects;
+	int i = 0, count;
+	uint32_t exec_start, exec_len;
+	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 + 1);
+
+	count = nbox ? nbox : 1;
+
+	for (i = 0; i < count; i++) {
+		if (i < nbox) {
+			int ret = i915_emit_box(dev, cliprects, i,
+						exec->DR1, exec->DR4);
+			if (ret)
+				return ret;
+		}
+
+		if (IS_I830(dev) || IS_845G(dev)) {
+			intel_ring_begin(dev, ring, 4);
+			intel_ring_emit(dev, ring, MI_BATCH_BUFFER);
+			intel_ring_emit(dev, ring,
+					exec_start | MI_BATCH_NON_SECURE);
+			intel_ring_emit(dev, ring, exec_start + exec_len - 4);
+			intel_ring_emit(dev, ring, 0);
+		} else {
+			intel_ring_begin(dev, ring, 4);
+			if (IS_I965G(dev)) {
+				intel_ring_emit(dev, ring,
+						MI_BATCH_BUFFER_START | (2 << 6)
+						| MI_BATCH_NON_SECURE_I965);
+				intel_ring_emit(dev, ring, exec_start);
+			} else {
+				intel_ring_emit(dev, ring, MI_BATCH_BUFFER_START
+						| (2 << 6));
+				intel_ring_emit(dev, ring, exec_start |
+						MI_BATCH_NON_SECURE);
+			}
+		}
+		intel_ring_advance(dev, ring);
+	}
+
+	/* XXX breadcrumb */
+	return 0;
+}
+
+static void cleanup_status_page(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	struct drm_gem_object *obj;
+	struct drm_i915_gem_object *obj_priv;
+
+	obj = ring->status_page.obj;
+	if (obj == NULL)
+		return;
+	obj_priv = to_intel_bo(obj);
+
+	kunmap(obj_priv->pages[0]);
+	i915_gem_object_unpin(obj);
+	drm_gem_object_unreference(obj);
+	ring->status_page.obj = NULL;
+
+	memset(&dev_priv->hws_map, 0, sizeof(dev_priv->hws_map));
+}
+
+static int init_status_page(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	drm_i915_private_t *dev_priv = dev->dev_private;
+	struct drm_gem_object *obj;
+	struct drm_i915_gem_object *obj_priv;
+	int ret;
+
+	obj = i915_gem_alloc_object(dev, 4096);
+	if (obj == NULL) {
+		DRM_ERROR("Failed to allocate status page\n");
+		ret = -ENOMEM;
+		goto err;
+	}
+	obj_priv = to_intel_bo(obj);
+	obj_priv->agp_type = AGP_USER_CACHED_MEMORY;
+
+	ret = i915_gem_object_pin(obj, 4096);
+	if (ret != 0) {
+		goto err_unref;
+	}
+
+	ring->status_page.gfx_addr = obj_priv->gtt_offset;
+	ring->status_page.page_addr = kmap(obj_priv->pages[0]);
+	if (ring->status_page.page_addr == NULL) {
+		memset(&dev_priv->hws_map, 0, sizeof(dev_priv->hws_map));
+		goto err_unpin;
+	}
+	ring->status_page.obj = obj;
+	memset(ring->status_page.page_addr, 0, PAGE_SIZE);
+
+	ring->setup_status_page(dev, ring);
+	DRM_DEBUG_DRIVER("%s hws offset: 0x%08x\n",
+			ring->name, ring->status_page.gfx_addr);
+
+	return 0;
+
+err_unpin:
+	i915_gem_object_unpin(obj);
+err_unref:
+	drm_gem_object_unreference(obj);
+err:
+	return ret;
+}
+
+
+int intel_init_ring_buffer(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	int ret;
+	struct drm_i915_gem_object *obj_priv;
+	struct drm_gem_object *obj;
+	ring->dev = dev;
+
+	if (I915_NEED_GFX_HWS(dev)) {
+		ret = init_status_page(dev, ring);
+		if (ret)
+			return ret;
+	}
+
+	obj = i915_gem_alloc_object(dev, ring->size);
+	if (obj == NULL) {
+		DRM_ERROR("Failed to allocate ringbuffer\n");
+		ret = -ENOMEM;
+		goto cleanup;
+	}
+
+	ring->gem_object = obj;
+
+	ret = i915_gem_object_pin(obj, ring->alignment);
+	if (ret != 0) {
+		drm_gem_object_unreference(obj);
+		goto cleanup;
+	}
+
+	obj_priv = to_intel_bo(obj);
+	ring->map.size = ring->size;
+	ring->map.offset = dev->agp->base + obj_priv->gtt_offset;
+	ring->map.type = 0;
+	ring->map.flags = 0;
+	ring->map.mtrr = 0;
+
+	drm_core_ioremap_wc(&ring->map, dev);
+	if (ring->map.handle == NULL) {
+		DRM_ERROR("Failed to map ringbuffer.\n");
+		i915_gem_object_unpin(obj);
+		drm_gem_object_unreference(obj);
+		ret = -EINVAL;
+		goto cleanup;
+	}
+
+	ring->virtual_start = ring->map.handle;
+	ret = ring->init(dev, ring);
+	if (ret != 0) {
+		intel_cleanup_ring_buffer(dev, ring);
+		return ret;
+	}
+
+	if (!drm_core_check_feature(dev, DRIVER_MODESET))
+		i915_kernel_lost_context(dev);
+	else {
+		ring->head = ring->get_head(dev, ring);
+		ring->tail = ring->get_tail(dev, ring);
+		ring->space = ring->head - (ring->tail + 8);
+		if (ring->space < 0)
+			ring->space += ring->size;
+	}
+	INIT_LIST_HEAD(&ring->active_list);
+	INIT_LIST_HEAD(&ring->request_list);
+	return ret;
+cleanup:
+	cleanup_status_page(dev, ring);
+	return ret;
+}
+
+void intel_cleanup_ring_buffer(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	if (ring->gem_object == NULL)
+		return;
+
+	drm_core_ioremapfree(&ring->map, dev);
+
+	i915_gem_object_unpin(ring->gem_object);
+	drm_gem_object_unreference(ring->gem_object);
+	ring->gem_object = NULL;
+	cleanup_status_page(dev, ring);
+}
+
+int intel_wrap_ring_buffer(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	unsigned int *virt;
+	int rem;
+	rem = ring->size - ring->tail;
+
+	if (ring->space < rem) {
+		int ret = intel_wait_ring_buffer(dev, ring, rem);
+		if (ret)
+			return ret;
+	}
+
+	virt = (unsigned int *)(ring->virtual_start + ring->tail);
+	rem /= 4;
+	while (rem--)
+		*virt++ = MI_NOOP;
+
+	ring->tail = 0;
+
+	return 0;
+}
+
+int intel_wait_ring_buffer(struct drm_device *dev,
+		struct intel_ring_buffer *ring, int n)
+{
+	unsigned long end;
+
+	trace_i915_ring_wait_begin (dev);
+	end = jiffies + 3 * HZ;
+	do {
+		ring->head = ring->get_head(dev, ring);
+		ring->space = ring->head - (ring->tail + 8);
+		if (ring->space < 0)
+			ring->space += ring->size;
+		if (ring->space >= n) {
+			trace_i915_ring_wait_end (dev);
+			return 0;
+		}
+
+		if (dev->primary->master) {
+			struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv;
+			if (master_priv->sarea_priv)
+				master_priv->sarea_priv->perf_boxes |= I915_BOX_WAIT;
+		}
+
+		yield();
+	} while (!time_after(jiffies, end));
+	trace_i915_ring_wait_end (dev);
+	return -EBUSY;
+}
+
+void intel_ring_begin(struct drm_device *dev,
+		struct intel_ring_buffer *ring, int n)
+{
+	if (unlikely(ring->tail + n > ring->size))
+		intel_wrap_ring_buffer(dev, ring);
+	if (unlikely(ring->space < n))
+		intel_wait_ring_buffer(dev, ring, n);
+}
+
+void intel_ring_emit(struct drm_device *dev,
+		struct intel_ring_buffer *ring, unsigned int data)
+{
+	unsigned int *virt = ring->virtual_start + ring->tail;
+	*virt = data;
+	ring->tail += 4;
+	ring->tail &= ring->size - 1;
+	ring->space -= 4;
+}
+
+void intel_ring_advance(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	ring->advance_ring(dev, ring);
+}
+
+void intel_fill_struct(struct drm_device *dev,
+		struct intel_ring_buffer *ring,
+		void *data,
+		unsigned int len)
+{
+	unsigned int *virt = ring->virtual_start + ring->tail;
+	BUG_ON((len&~(4-1)) != 0);
+	intel_ring_begin(dev, ring, len);
+	memcpy(virt, data, len);
+	ring->tail += len;
+	ring->tail &= ring->size - 1;
+	ring->space -= len;
+	intel_ring_advance(dev, ring);
+}
+
+u32 intel_ring_get_seqno(struct drm_device *dev,
+		struct intel_ring_buffer *ring)
+{
+	u32 seqno;
+	seqno = ring->next_seqno;
+
+	/* reserve 0 for non-seqno */
+	if (++ring->next_seqno == 0)
+		ring->next_seqno = 1;
+	return seqno;
+}
+
+struct intel_ring_buffer render_ring = {
+	.name			= "render ring",
+	.regs                   = {
+		.ctl = PRB0_CTL,
+		.head = PRB0_HEAD,
+		.tail = PRB0_TAIL,
+		.start = PRB0_START
+	},
+	.ring_flag		= I915_EXEC_RENDER,
+	.size			= 32 * PAGE_SIZE,
+	.alignment		= PAGE_SIZE,
+	.virtual_start		= NULL,
+	.dev			= NULL,
+	.gem_object		= NULL,
+	.head			= 0,
+	.tail			= 0,
+	.space			= 0,
+	.next_seqno		= 1,
+	.user_irq_refcount	= 0,
+	.irq_gem_seqno		= 0,
+	.waiting_gem_seqno	= 0,
+	.setup_status_page	= render_setup_status_page,
+	.init			= init_render_ring,
+	.get_head		= render_ring_get_head,
+	.get_tail		= render_ring_get_tail,
+	.get_active_head	= render_ring_get_active_head,
+	.advance_ring		= render_ring_advance_ring,
+	.flush			= render_ring_flush,
+	.add_request		= render_ring_add_request,
+	.get_gem_seqno		= render_ring_get_gem_seqno,
+	.user_irq_get		= render_ring_get_user_irq,
+	.user_irq_put		= render_ring_put_user_irq,
+	.dispatch_gem_execbuffer = render_ring_dispatch_gem_execbuffer,
+	.status_page		= {NULL, 0, NULL},
+	.map			= {0,}
+};
+
+/* ring buffer for bit-stream decoder */
+
+struct intel_ring_buffer bsd_ring = {
+	.name                   = "bsd ring",
+	.regs			= {
+		.ctl = BSD_RING_CTL,
+		.head = BSD_RING_HEAD,
+		.tail = BSD_RING_TAIL,
+		.start = BSD_RING_START
+	},
+	.ring_flag		= I915_EXEC_BSD,
+	.size			= 32 * PAGE_SIZE,
+	.alignment		= PAGE_SIZE,
+	.virtual_start		= NULL,
+	.dev			= NULL,
+	.gem_object		= NULL,
+	.head			= 0,
+	.tail			= 0,
+	.space			= 0,
+	.next_seqno		= 1,
+	.user_irq_refcount	= 0,
+	.irq_gem_seqno		= 0,
+	.waiting_gem_seqno	= 0,
+	.setup_status_page	= bsd_setup_status_page,
+	.init			= init_bsd_ring,
+	.get_head		= bsd_ring_get_head,
+	.get_tail		= bsd_ring_get_tail,
+	.get_active_head	= bsd_ring_get_active_head,
+	.advance_ring		= bsd_ring_advance_ring,
+	.flush			= bsd_ring_flush,
+	.add_request		= bsd_ring_add_request,
+	.get_gem_seqno		= bsd_ring_get_gem_seqno,
+	.user_irq_get		= bsd_ring_get_user_irq,
+	.user_irq_put		= bsd_ring_put_user_irq,
+	.dispatch_gem_execbuffer = bsd_ring_dispatch_gem_execbuffer,
+	.status_page		= {NULL, 0, NULL},
+	.map			= {0,}
+};
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
new file mode 100644
index 000000000000..d5568d3766de
--- /dev/null
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
@@ -0,0 +1,124 @@
+#ifndef _INTEL_RINGBUFFER_H_
+#define _INTEL_RINGBUFFER_H_
+
+struct  intel_hw_status_page {
+	void		*page_addr;
+	unsigned int	gfx_addr;
+	struct		drm_gem_object *obj;
+};
+
+struct drm_i915_gem_execbuffer2;
+struct  intel_ring_buffer {
+	const char	*name;
+	struct		ring_regs {
+			u32 ctl;
+			u32 head;
+			u32 tail;
+			u32 start;
+	} regs;
+	unsigned int	ring_flag;
+	unsigned long	size;
+	unsigned int	alignment;
+	void		*virtual_start;
+	struct		drm_device *dev;
+	struct		drm_gem_object *gem_object;
+
+	unsigned int	head;
+	unsigned int	tail;
+	unsigned int	space;
+	u32		next_seqno;
+	struct intel_hw_status_page status_page;
+
+	u32		irq_gem_seqno;		/* last seq seem at irq time */
+	u32		waiting_gem_seqno;
+	int		user_irq_refcount;
+	void		(*user_irq_get)(struct drm_device *dev,
+			struct intel_ring_buffer *ring);
+	void		(*user_irq_put)(struct drm_device *dev,
+			struct intel_ring_buffer *ring);
+	void		(*setup_status_page)(struct drm_device *dev,
+			struct	intel_ring_buffer *ring);
+
+	int		(*init)(struct drm_device *dev,
+			struct intel_ring_buffer *ring);
+
+	unsigned int	(*get_head)(struct drm_device *dev,
+			struct intel_ring_buffer *ring);
+	unsigned int	(*get_tail)(struct drm_device *dev,
+			struct intel_ring_buffer *ring);
+	unsigned int	(*get_active_head)(struct drm_device *dev,
+			struct intel_ring_buffer *ring);
+	void		(*advance_ring)(struct drm_device *dev,
+			struct intel_ring_buffer *ring);
+	void		(*flush)(struct drm_device *dev,
+			struct intel_ring_buffer *ring,
+			u32	invalidate_domains,
+			u32	flush_domains);
+	u32		(*add_request)(struct drm_device *dev,
+			struct intel_ring_buffer *ring,
+			struct drm_file *file_priv,
+			u32 flush_domains);
+	u32		(*get_gem_seqno)(struct drm_device *dev,
+			struct intel_ring_buffer *ring);
+	int		(*dispatch_gem_execbuffer)(struct drm_device *dev,
+			struct intel_ring_buffer *ring,
+			struct drm_i915_gem_execbuffer2 *exec,
+			struct drm_clip_rect *cliprects,
+			uint64_t exec_offset);
+
+	/**
+	 * List of objects currently involved in rendering from the
+	 * ringbuffer.
+	 *
+	 * Includes buffers having the contents of their GPU caches
+	 * flushed, not necessarily primitives.  last_rendering_seqno
+	 * represents when the rendering involved will be completed.
+	 *
+	 * A reference is held on the buffer while on this list.
+	 */
+	struct list_head active_list;
+
+	/**
+	 * List of breadcrumbs associated with GPU requests currently
+	 * outstanding.
+	 */
+	struct list_head request_list;
+
+	wait_queue_head_t irq_queue;
+	drm_local_map_t map;
+};
+
+static inline u32
+intel_read_status_page(struct intel_ring_buffer *ring,
+		int reg)
+{
+	u32 *regs = ring->status_page.page_addr;
+	return regs[reg];
+}
+
+int intel_init_ring_buffer(struct drm_device *dev,
+		struct intel_ring_buffer *ring);
+void intel_cleanup_ring_buffer(struct drm_device *dev,
+		struct intel_ring_buffer *ring);
+int intel_wait_ring_buffer(struct drm_device *dev,
+		struct intel_ring_buffer *ring, int n);
+int intel_wrap_ring_buffer(struct drm_device *dev,
+		struct intel_ring_buffer *ring);
+void intel_ring_begin(struct drm_device *dev,
+		struct intel_ring_buffer *ring, int n);
+void intel_ring_emit(struct drm_device *dev,
+		struct intel_ring_buffer *ring, u32 data);
+void intel_fill_struct(struct drm_device *dev,
+		struct intel_ring_buffer *ring,
+		void *data,
+		unsigned int len);
+void intel_ring_advance(struct drm_device *dev,
+		struct intel_ring_buffer *ring);
+
+u32 intel_ring_get_seqno(struct drm_device *dev,
+		struct intel_ring_buffer *ring);
+
+extern struct intel_ring_buffer render_ring;
+extern struct intel_ring_buffer bsd_ring;
+
+#endif /* _INTEL_RINGBUFFER_H_ */
diff --git a/drivers/gpu/drm/i915/intel_sdvo.c b/drivers/gpu/drm/i915/intel_sdvo.c
index aba72c489a2f..76993ac16cc1 100644
--- a/drivers/gpu/drm/i915/intel_sdvo.c
+++ b/drivers/gpu/drm/i915/intel_sdvo.c
@@ -1479,7 +1479,7 @@ intel_find_analog_connector(struct drm_device *dev)
 		intel_encoder = enc_to_intel_encoder(encoder);
 		if (intel_encoder->type == INTEL_OUTPUT_ANALOG) {
 			list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
-				if (connector && encoder == intel_attached_encoder(connector))
+				if (encoder == intel_attached_encoder(connector))
 					return connector;
 			}
 		}
diff --git a/drivers/gpu/drm/nouveau/nouveau_acpi.c b/drivers/gpu/drm/nouveau/nouveau_acpi.c
index e13f6af0037a..d4bcca8a5133 100644
--- a/drivers/gpu/drm/nouveau/nouveau_acpi.c
+++ b/drivers/gpu/drm/nouveau/nouveau_acpi.c
@@ -34,7 +34,7 @@
 static struct nouveau_dsm_priv {
 	bool dsm_detected;
 	acpi_handle dhandle;
-	acpi_handle dsm_handle;
+	acpi_handle rom_handle;
 } nouveau_dsm_priv;
 
 static const char nouveau_dsm_muid[] = {
@@ -107,9 +107,9 @@ static int nouveau_dsm_set_discrete_state(acpi_handle handle, enum vga_switchero
 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);
+		return nouveau_dsm_switch_mux(nouveau_dsm_priv.dhandle, NOUVEAU_DSM_LED_STAMINA);
 	else
-		return nouveau_dsm_switch_mux(nouveau_dsm_priv.dsm_handle, NOUVEAU_DSM_LED_SPEED);
+		return nouveau_dsm_switch_mux(nouveau_dsm_priv.dhandle, NOUVEAU_DSM_LED_SPEED);
 }
 
 static int nouveau_dsm_power_state(enum vga_switcheroo_client_id id,
@@ -118,7 +118,7 @@ static int nouveau_dsm_power_state(enum vga_switcheroo_client_id id,
 	if (id == VGA_SWITCHEROO_IGD)
 		return 0;
 
-	return nouveau_dsm_set_discrete_state(nouveau_dsm_priv.dsm_handle, state);
+	return nouveau_dsm_set_discrete_state(nouveau_dsm_priv.dhandle, state);
 }
 
 static int nouveau_dsm_init(void)
@@ -151,18 +151,18 @@ static bool nouveau_dsm_pci_probe(struct pci_dev *pdev)
 	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);
+	ret = nouveau_dsm(dhandle, 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;
 }
 
@@ -173,6 +173,7 @@ static bool nouveau_dsm_detect(void)
 	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++;
 
@@ -180,7 +181,7 @@ static bool nouveau_dsm_detect(void)
 	}
 
 	if (vga_count == 2 && has_dsm) {
-		acpi_get_name(nouveau_dsm_priv.dsm_handle, ACPI_FULL_PATHNAME, &buffer);
+		acpi_get_name(nouveau_dsm_priv.dhandle, 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;
@@ -204,3 +205,57 @@ void nouveau_unregister_dsm_handler(void)
 {
 	vga_switcheroo_unregister_handler();
 }
+
+/* retrieve the ROM in 4k blocks */
+static int nouveau_rom_call(acpi_handle rom_handle, uint8_t *bios,
+			    int offset, int len)
+{
+	acpi_status status;
+	union acpi_object rom_arg_elements[2], *obj;
+	struct acpi_object_list rom_arg;
+	struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL};
+
+	rom_arg.count = 2;
+	rom_arg.pointer = &rom_arg_elements[0];
+
+	rom_arg_elements[0].type = ACPI_TYPE_INTEGER;
+	rom_arg_elements[0].integer.value = offset;
+
+	rom_arg_elements[1].type = ACPI_TYPE_INTEGER;
+	rom_arg_elements[1].integer.value = len;
+
+	status = acpi_evaluate_object(rom_handle, NULL, &rom_arg, &buffer);
+	if (ACPI_FAILURE(status)) {
+		printk(KERN_INFO "failed to evaluate ROM 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 nouveau_acpi_rom_supported(struct pci_dev *pdev)
+{
+	acpi_status status;
+	acpi_handle dhandle, rom_handle;
+
+	if (!nouveau_dsm_priv.dsm_detected)
+		return false;
+
+	dhandle = DEVICE_ACPI_HANDLE(&pdev->dev);
+	if (!dhandle)
+		return false;
+
+	status = acpi_get_handle(dhandle, "_ROM", &rom_handle);
+	if (ACPI_FAILURE(status))
+		return false;
+
+	nouveau_dsm_priv.rom_handle = rom_handle;
+	return true;
+}
+
+int nouveau_acpi_get_bios_chunk(uint8_t *bios, int offset, int len)
+{
+	return nouveau_rom_call(nouveau_dsm_priv.rom_handle, bios, offset, len);
+}
diff --git a/drivers/gpu/drm/nouveau/nouveau_bios.c b/drivers/gpu/drm/nouveau/nouveau_bios.c
index e7e69ccce5c9..fc924b649195 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bios.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bios.c
@@ -178,6 +178,25 @@ out:
 	pci_disable_rom(dev->pdev);
 }
 
+static void load_vbios_acpi(struct drm_device *dev, uint8_t *data)
+{
+	int i;
+	int ret;
+	int size = 64 * 1024;
+
+	if (!nouveau_acpi_rom_supported(dev->pdev))
+		return;
+
+	for (i = 0; i < (size / ROM_BIOS_PAGE); i++) {
+		ret = nouveau_acpi_get_bios_chunk(data,
+						  (i * ROM_BIOS_PAGE),
+						  ROM_BIOS_PAGE);
+		if (ret <= 0)
+			break;
+	}
+	return;
+}
+
 struct methods {
 	const char desc[8];
 	void (*loadbios)(struct drm_device *, uint8_t *);
@@ -191,6 +210,7 @@ static struct methods nv04_methods[] = {
 };
 
 static struct methods nv50_methods[] = {
+	{ "ACPI", load_vbios_acpi, true },
 	{ "PRAMIN", load_vbios_pramin, true },
 	{ "PROM", load_vbios_prom, false },
 	{ "PCIROM", load_vbios_pci, true },
@@ -814,7 +834,7 @@ init_i2c_device_find(struct drm_device *dev, int i2c_index)
 	if (i2c_index == 0x81)
 		i2c_index = (dcb->i2c_default_indices & 0xf0) >> 4;
 
-	if (i2c_index > DCB_MAX_NUM_I2C_ENTRIES) {
+	if (i2c_index >= DCB_MAX_NUM_I2C_ENTRIES) {
 		NV_ERROR(dev, "invalid i2c_index 0x%x\n", i2c_index);
 		return NULL;
 	}
@@ -2807,7 +2827,10 @@ init_gpio(struct nvbios *bios, uint16_t offset, struct init_exec *iexec)
 
 		BIOSLOG(bios, "0x%04X: Entry: 0x%08X\n", offset, gpio->entry);
 
-		nv50_gpio_set(bios->dev, gpio->tag, gpio->state_default);
+		BIOSLOG(bios, "0x%04X: set gpio 0x%02x, state %d\n",
+			offset, gpio->tag, gpio->state_default);
+		if (bios->execute)
+			nv50_gpio_set(bios->dev, gpio->tag, gpio->state_default);
 
 		/* The NVIDIA binary driver doesn't appear to actually do
 		 * any of this, my VBIOS does however.
@@ -3897,7 +3920,8 @@ int nouveau_bios_parse_lvds_table(struct drm_device *dev, int pxclk, bool *dl, b
 
 static uint8_t *
 bios_output_config_match(struct drm_device *dev, struct dcb_entry *dcbent,
-			 uint16_t record, int record_len, int record_nr)
+			 uint16_t record, int record_len, int record_nr,
+			 bool match_link)
 {
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	struct nvbios *bios = &dev_priv->vbios;
@@ -3905,12 +3929,28 @@ bios_output_config_match(struct drm_device *dev, struct dcb_entry *dcbent,
 	uint16_t table;
 	int i, v;
 
+	switch (dcbent->type) {
+	case OUTPUT_TMDS:
+	case OUTPUT_LVDS:
+	case OUTPUT_DP:
+		break;
+	default:
+		match_link = false;
+		break;
+	}
+
 	for (i = 0; i < record_nr; i++, record += record_len) {
 		table = ROM16(bios->data[record]);
 		if (!table)
 			continue;
 		entry = ROM32(bios->data[table]);
 
+		if (match_link) {
+			v = (entry & 0x00c00000) >> 22;
+			if (!(v & dcbent->sorconf.link))
+				continue;
+		}
+
 		v = (entry & 0x000f0000) >> 16;
 		if (!(v & dcbent->or))
 			continue;
@@ -3952,7 +3992,7 @@ nouveau_bios_dp_table(struct drm_device *dev, struct dcb_entry *dcbent,
 	*length = table[4];
 	return bios_output_config_match(dev, dcbent,
 					bios->display.dp_table_ptr + table[1],
-					table[2], table[3]);
+					table[2], table[3], table[0] >= 0x21);
 }
 
 int
@@ -4041,7 +4081,7 @@ nouveau_bios_run_display_table(struct drm_device *dev, struct dcb_entry *dcbent,
 			dcbent->type, dcbent->location, dcbent->or);
 	otable = bios_output_config_match(dev, dcbent, table[1] +
 					  bios->display.script_table_ptr,
-					  table[2], table[3]);
+					  table[2], table[3], table[0] >= 0x21);
 	if (!otable) {
 		NV_ERROR(dev, "Couldn't find matching output script table\n");
 		return 1;
@@ -5533,12 +5573,6 @@ parse_dcb20_entry(struct drm_device *dev, struct dcb_table *dcb,
 	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:
@@ -5622,6 +5656,16 @@ parse_dcb20_entry(struct drm_device *dev, struct dcb_table *dcb,
 		break;
 	}
 
+	if (dcb->version < 0x40) {
+		/* 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);
+	} else {
+		entry->duallink_possible = (entry->sorconf.link == 3);
+	}
+
 	/* unsure what DCB version introduces this, 3.0? */
 	if (conf & 0x100000)
 		entry->i2c_upper_default = true;
@@ -6205,6 +6249,30 @@ nouveau_bios_i2c_devices_takedown(struct drm_device *dev)
 		nouveau_i2c_fini(dev, entry);
 }
 
+static bool
+nouveau_bios_posted(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	bool was_locked;
+	unsigned htotal;
+
+	if (dev_priv->chipset >= NV_50) {
+		if (NVReadVgaCrtc(dev, 0, 0x00) == 0 &&
+		    NVReadVgaCrtc(dev, 0, 0x1a) == 0)
+			return false;
+		return true;
+	}
+
+	was_locked = NVLockVgaCrtcs(dev, false);
+	htotal  = NVReadVgaCrtc(dev, 0, 0x06);
+	htotal |= (NVReadVgaCrtc(dev, 0, 0x07) & 0x01) << 8;
+	htotal |= (NVReadVgaCrtc(dev, 0, 0x07) & 0x20) << 4;
+	htotal |= (NVReadVgaCrtc(dev, 0, 0x25) & 0x01) << 10;
+	htotal |= (NVReadVgaCrtc(dev, 0, 0x41) & 0x01) << 11;
+	NVLockVgaCrtcs(dev, was_locked);
+	return (htotal != 0);
+}
+
 int
 nouveau_bios_init(struct drm_device *dev)
 {
@@ -6239,11 +6307,9 @@ nouveau_bios_init(struct drm_device *dev)
 	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) {
+	if (!nouveau_bios_posted(dev)) {
 		NV_INFO(dev, "Adaptor not initialised\n");
-		if (dev_priv->card_type < NV_50) {
+		if (dev_priv->card_type < NV_40) {
 			NV_ERROR(dev, "Unable to POST this chipset\n");
 			return -ENODEV;
 		}
diff --git a/drivers/gpu/drm/nouveau/nouveau_connector.c b/drivers/gpu/drm/nouveau/nouveau_connector.c
index 7e663a79829f..149ed224c3cb 100644
--- a/drivers/gpu/drm/nouveau/nouveau_connector.c
+++ b/drivers/gpu/drm/nouveau/nouveau_connector.c
@@ -241,7 +241,8 @@ nouveau_connector_detect(struct drm_connector *connector)
 	if (nv_encoder && nv_connector->native_mode) {
 		unsigned status = connector_status_connected;
 
-#ifdef CONFIG_ACPI
+#if defined(CONFIG_ACPI_BUTTON) || \
+	(defined(CONFIG_ACPI_BUTTON_MODULE) && defined(MODULE))
 		if (!nouveau_ignorelid && !acpi_lid_open())
 			status = connector_status_unknown;
 #endif
@@ -431,24 +432,27 @@ nouveau_connector_set_property(struct drm_connector *connector,
 }
 
 static struct drm_display_mode *
-nouveau_connector_native_mode(struct nouveau_connector *connector)
+nouveau_connector_native_mode(struct drm_connector *connector)
 {
-	struct drm_device *dev = connector->base.dev;
+	struct drm_connector_helper_funcs *helper = connector->helper_private;
+	struct nouveau_connector *nv_connector = nouveau_connector(connector);
+	struct drm_device *dev = connector->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) {
+	list_for_each_entry(mode, &nv_connector->base.probed_modes, head) {
+		if (helper->mode_valid(connector, mode) != MODE_OK)
+			continue;
+
+		/* Use preferred mode if there is one.. */
 		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) {
+		/* Otherwise, take the resolution with the largest width, then
+		 * height, then vertical refresh
+		 */
 		if (mode->hdisplay < high_w)
 			continue;
 
@@ -552,7 +556,7 @@ nouveau_connector_get_modes(struct drm_connector *connector)
 	 */
 	if (!nv_connector->native_mode)
 		nv_connector->native_mode =
-			nouveau_connector_native_mode(nv_connector);
+			nouveau_connector_native_mode(connector);
 	if (ret == 0 && nv_connector->native_mode) {
 		struct drm_display_mode *mode;
 
@@ -583,9 +587,9 @@ nouveau_connector_mode_valid(struct drm_connector *connector,
 
 	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)
+		if (nv_connector->native_mode &&
+		    (mode->hdisplay > nv_connector->native_mode->hdisplay ||
+		     mode->vdisplay > nv_connector->native_mode->vdisplay))
 			return MODE_PANEL;
 
 		min_clock = 0;
@@ -593,8 +597,7 @@ nouveau_connector_mode_valid(struct drm_connector *connector,
 		break;
 	case OUTPUT_TMDS:
 		if ((dev_priv->card_type >= NV_50 && !nouveau_duallink) ||
-		    (dev_priv->card_type < NV_50 &&
-		     !nv_encoder->dcb->duallink_possible))
+		    !nv_encoder->dcb->duallink_possible)
 			max_clock = 165000;
 		else
 			max_clock = 330000;
@@ -728,7 +731,7 @@ nouveau_connector_create_lvds(struct drm_device *dev,
 	if (ret == 0)
 		goto out;
 	nv_connector->detected_encoder = nv_encoder;
-	nv_connector->native_mode = nouveau_connector_native_mode(nv_connector);
+	nv_connector->native_mode = nouveau_connector_native_mode(connector);
 	list_for_each_entry_safe(mode, temp, &connector->probed_modes, head)
 		drm_mode_remove(connector, mode);
 
diff --git a/drivers/gpu/drm/nouveau/nouveau_crtc.h b/drivers/gpu/drm/nouveau/nouveau_crtc.h
index 49fa7b2d257e..cb1ce2a09162 100644
--- a/drivers/gpu/drm/nouveau/nouveau_crtc.h
+++ b/drivers/gpu/drm/nouveau/nouveau_crtc.h
@@ -40,6 +40,8 @@ struct nouveau_crtc {
 	int sharpness;
 	int last_dpms;
 
+	int cursor_saved_x, cursor_saved_y;
+
 	struct {
 		int cpp;
 		bool blanked;
diff --git a/drivers/gpu/drm/nouveau/nouveau_drv.c b/drivers/gpu/drm/nouveau/nouveau_drv.c
index c6079e36669d..273770432298 100644
--- a/drivers/gpu/drm/nouveau/nouveau_drv.c
+++ b/drivers/gpu/drm/nouveau/nouveau_drv.c
@@ -175,6 +175,13 @@ nouveau_pci_suspend(struct pci_dev *pdev, pm_message_t pm_state)
 		nouveau_bo_unpin(nouveau_fb->nvbo);
 	}
 
+	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
+		struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+
+		nouveau_bo_unmap(nv_crtc->cursor.nvbo);
+		nouveau_bo_unpin(nv_crtc->cursor.nvbo);
+	}
+
 	NV_INFO(dev, "Evicting buffers...\n");
 	ttm_bo_evict_mm(&dev_priv->ttm.bdev, TTM_PL_VRAM);
 
@@ -314,12 +321,34 @@ nouveau_pci_resume(struct pci_dev *pdev)
 		nouveau_bo_pin(nouveau_fb->nvbo, TTM_PL_FLAG_VRAM);
 	}
 
+	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
+		struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+		int 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)
+			NV_ERROR(dev, "Could not pin/map cursor.\n");
+	}
+
 	if (dev_priv->card_type < NV_50) {
 		nv04_display_restore(dev);
 		NVLockVgaCrtcs(dev, false);
 	} else
 		nv50_display_init(dev);
 
+	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
+		struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
+
+		nv_crtc->cursor.set_offset(nv_crtc,
+					nv_crtc->cursor.nvbo->bo.offset -
+					dev_priv->vm_vram_base);
+
+		nv_crtc->cursor.set_pos(nv_crtc, nv_crtc->cursor_saved_x,
+			nv_crtc->cursor_saved_y);
+	}
+
 	/* 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);
diff --git a/drivers/gpu/drm/nouveau/nouveau_drv.h b/drivers/gpu/drm/nouveau/nouveau_drv.h
index 5b134438effe..c69719106489 100644
--- a/drivers/gpu/drm/nouveau/nouveau_drv.h
+++ b/drivers/gpu/drm/nouveau/nouveau_drv.h
@@ -851,12 +851,17 @@ extern int  nouveau_dma_init(struct nouveau_channel *);
 extern int  nouveau_dma_wait(struct nouveau_channel *, int slots, int size);
 
 /* nouveau_acpi.c */
+#define ROM_BIOS_PAGE 4096
 #if defined(CONFIG_ACPI)
 void nouveau_register_dsm_handler(void);
 void nouveau_unregister_dsm_handler(void);
+int nouveau_acpi_get_bios_chunk(uint8_t *bios, int offset, int len);
+bool nouveau_acpi_rom_supported(struct pci_dev *pdev);
 #else
 static inline void nouveau_register_dsm_handler(void) {}
 static inline void nouveau_unregister_dsm_handler(void) {}
+static inline bool nouveau_acpi_rom_supported(struct pci_dev *pdev) { return false; }
+static inline int nouveau_acpi_get_bios_chunk(uint8_t *bios, int offset, int len) { return -EINVAL; }
 #endif
 
 /* nouveau_backlight.c */
diff --git a/drivers/gpu/drm/nouveau/nouveau_fbcon.c b/drivers/gpu/drm/nouveau/nouveau_fbcon.c
index fd4a2df715e9..c9a4a0d2a115 100644
--- a/drivers/gpu/drm/nouveau/nouveau_fbcon.c
+++ b/drivers/gpu/drm/nouveau/nouveau_fbcon.c
@@ -377,6 +377,7 @@ int nouveau_fbcon_init(struct drm_device *dev)
 {
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	struct nouveau_fbdev *nfbdev;
+	int ret;
 
 	nfbdev = kzalloc(sizeof(struct nouveau_fbdev), GFP_KERNEL);
 	if (!nfbdev)
@@ -386,7 +387,12 @@ int nouveau_fbcon_init(struct drm_device *dev)
 	dev_priv->nfbdev = nfbdev;
 	nfbdev->helper.funcs = &nouveau_fbcon_helper_funcs;
 
-	drm_fb_helper_init(dev, &nfbdev->helper, 2, 4);
+	ret = drm_fb_helper_init(dev, &nfbdev->helper, 2, 4);
+	if (ret) {
+		kfree(nfbdev);
+		return ret;
+	}
+
 	drm_fb_helper_single_add_all_connectors(&nfbdev->helper);
 	drm_fb_helper_initial_config(&nfbdev->helper, 32);
 	return 0;
diff --git a/drivers/gpu/drm/nouveau/nouveau_mem.c b/drivers/gpu/drm/nouveau/nouveau_mem.c
index 775a7017af64..c1fd42b0dad1 100644
--- a/drivers/gpu/drm/nouveau/nouveau_mem.c
+++ b/drivers/gpu/drm/nouveau/nouveau_mem.c
@@ -540,7 +540,8 @@ nouveau_mem_detect(struct drm_device *dev)
 		dev_priv->vram_size  = nv_rd32(dev, NV04_FIFO_DATA);
 		dev_priv->vram_size &= NV10_FIFO_DATA_RAM_AMOUNT_MB_MASK;
 		if (dev_priv->chipset == 0xaa || dev_priv->chipset == 0xac)
-			dev_priv->vram_sys_base = nv_rd32(dev, 0x100e10) << 12;
+			dev_priv->vram_sys_base = nv_rd32(dev, 0x100e10);
+			dev_priv->vram_sys_base <<= 12;
 	}
 
 	NV_INFO(dev, "Detected %dMiB VRAM\n", (int)(dev_priv->vram_size >> 20));
diff --git a/drivers/gpu/drm/nouveau/nouveau_state.c b/drivers/gpu/drm/nouveau/nouveau_state.c
index e632339c323e..b02a231d6937 100644
--- a/drivers/gpu/drm/nouveau/nouveau_state.c
+++ b/drivers/gpu/drm/nouveau/nouveau_state.c
@@ -376,12 +376,15 @@ out_err:
 static void nouveau_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_ERR "VGA switcheroo: switched nouveau on\n");
 		nouveau_pci_resume(pdev);
+		drm_kms_helper_poll_enable(dev);
 	} else {
 		printk(KERN_ERR "VGA switcheroo: switched nouveau off\n");
+		drm_kms_helper_poll_disable(dev);
 		nouveau_pci_suspend(pdev, pmm);
 	}
 }
@@ -776,29 +779,24 @@ int nouveau_load(struct drm_device *dev, unsigned long flags)
 			return ret;
 	}
 
-	/* map larger RAMIN aperture on NV40 cards */
-	dev_priv->ramin  = NULL;
+	/* Map PRAMIN BAR, or on older cards, the aperture withing BAR0 */
 	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 =
+			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");
+			NV_ERROR(dev, "Failed to PRAMIN BAR");
+			return -ENOMEM;
 		}
-	}
-
-	/* On older cards (or if the above failed), create a map covering
-	 * the BAR0 PRAMIN aperture */
-	if (!dev_priv->ramin) {
+	} else {
 		dev_priv->ramin_size = 1 * 1024 * 1024;
 		dev_priv->ramin = ioremap(mmio_start_offs + NV_RAMIN,
-							dev_priv->ramin_size);
+					  dev_priv->ramin_size);
 		if (!dev_priv->ramin) {
 			NV_ERROR(dev, "Failed to map BAR0 PRAMIN.\n");
 			return -ENOMEM;
@@ -913,6 +911,9 @@ int nouveau_ioctl_getparam(struct drm_device *dev, void *data,
 	case NOUVEAU_GETPARAM_VM_VRAM_BASE:
 		getparam->value = dev_priv->vm_vram_base;
 		break;
+	case NOUVEAU_GETPARAM_PTIMER_TIME:
+		getparam->value = dev_priv->engine.timer.read(dev);
+		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
diff --git a/drivers/gpu/drm/nouveau/nv04_cursor.c b/drivers/gpu/drm/nouveau/nv04_cursor.c
index 89a91b9d8b25..aaf3de3bc816 100644
--- a/drivers/gpu/drm/nouveau/nv04_cursor.c
+++ b/drivers/gpu/drm/nouveau/nv04_cursor.c
@@ -20,6 +20,7 @@ nv04_cursor_hide(struct nouveau_crtc *nv_crtc, bool update)
 static void
 nv04_cursor_set_pos(struct nouveau_crtc *nv_crtc, int x, int y)
 {
+	nv_crtc->cursor_saved_x = x; nv_crtc->cursor_saved_y = y;
 	NVWriteRAMDAC(nv_crtc->base.dev, nv_crtc->index,
 		      NV_PRAMDAC_CU_START_POS,
 		      XLATE(y, 0, NV_PRAMDAC_CU_START_POS_Y) |
diff --git a/drivers/gpu/drm/nouveau/nv40_graph.c b/drivers/gpu/drm/nouveau/nv40_graph.c
index 0616c96e4b67..704a25d04ac9 100644
--- a/drivers/gpu/drm/nouveau/nv40_graph.c
+++ b/drivers/gpu/drm/nouveau/nv40_graph.c
@@ -253,7 +253,11 @@ nv40_graph_init(struct drm_device *dev)
 
 	if (!dev_priv->engine.graph.ctxprog) {
 		struct nouveau_grctx ctx = {};
-		uint32_t cp[256];
+		uint32_t *cp;
+
+		cp = kmalloc(sizeof(*cp) * 256, GFP_KERNEL);
+		if (!cp)
+			return -ENOMEM;
 
 		ctx.dev = dev;
 		ctx.mode = NOUVEAU_GRCTX_PROG;
@@ -265,6 +269,8 @@ nv40_graph_init(struct drm_device *dev)
 		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]);
+
+		kfree(cp);
 	}
 
 	/* No context present currently */
diff --git a/drivers/gpu/drm/nouveau/nv50_cursor.c b/drivers/gpu/drm/nouveau/nv50_cursor.c
index 753e723adb3a..03ad7ab14f09 100644
--- a/drivers/gpu/drm/nouveau/nv50_cursor.c
+++ b/drivers/gpu/drm/nouveau/nv50_cursor.c
@@ -107,6 +107,7 @@ nv50_cursor_set_pos(struct nouveau_crtc *nv_crtc, int x, int y)
 {
 	struct drm_device *dev = nv_crtc->base.dev;
 
+	nv_crtc->cursor_saved_x = x; nv_crtc->cursor_saved_y = y;
 	nv_wr32(dev, NV50_PDISPLAY_CURSOR_USER_POS(nv_crtc->index),
 		((y & 0xFFFF) << 16) | (x & 0xFFFF));
 	/* Needed to make the cursor move. */
diff --git a/drivers/gpu/drm/nouveau/nv50_fb.c b/drivers/gpu/drm/nouveau/nv50_fb.c
index a95e6941ba88..32611bd30e6d 100644
--- a/drivers/gpu/drm/nouveau/nv50_fb.c
+++ b/drivers/gpu/drm/nouveau/nv50_fb.c
@@ -6,10 +6,16 @@
 int
 nv50_fb_init(struct drm_device *dev)
 {
-	/* This is needed to get meaningful information from 100c90
-	 * on traps. No idea what these values mean exactly. */
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
 
+	/* Not a clue what this is exactly.  Without pointing it at a
+	 * scratch page, VRAM->GART blits with M2MF (as in DDX DFS)
+	 * cause IOMMU "read from address 0" errors (rh#561267)
+	 */
+	nv_wr32(dev, 0x100c08, dev_priv->gart_info.sg_dummy_bus >> 8);
+
+	/* This is needed to get meaningful information from 100c90
+	 * on traps. No idea what these values mean exactly. */
 	switch (dev_priv->chipset) {
 	case 0x50:
 		nv_wr32(dev, 0x100c90, 0x0707ff);
diff --git a/drivers/gpu/drm/nouveau/nv50_gpio.c b/drivers/gpu/drm/nouveau/nv50_gpio.c
index c61782b314e7..bb47ad737267 100644
--- a/drivers/gpu/drm/nouveau/nv50_gpio.c
+++ b/drivers/gpu/drm/nouveau/nv50_gpio.c
@@ -31,7 +31,7 @@ nv50_gpio_location(struct dcb_gpio_entry *gpio, uint32_t *reg, uint32_t *shift)
 {
 	const uint32_t nv50_gpio_reg[4] = { 0xe104, 0xe108, 0xe280, 0xe284 };
 
-	if (gpio->line > 32)
+	if (gpio->line >= 32)
 		return -EINVAL;
 
 	*reg = nv50_gpio_reg[gpio->line >> 3];
diff --git a/drivers/gpu/drm/nouveau/nv50_sor.c b/drivers/gpu/drm/nouveau/nv50_sor.c
index b11eaf9c5c7c..812778db76ac 100644
--- a/drivers/gpu/drm/nouveau/nv50_sor.c
+++ b/drivers/gpu/drm/nouveau/nv50_sor.c
@@ -274,7 +274,6 @@ static const struct drm_encoder_funcs nv50_sor_encoder_funcs = {
 int
 nv50_sor_create(struct drm_device *dev, struct dcb_entry *entry)
 {
-	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	struct nouveau_encoder *nv_encoder = NULL;
 	struct drm_encoder *encoder;
 	bool dum;
@@ -324,11 +323,7 @@ nv50_sor_create(struct drm_device *dev, struct dcb_entry *entry)
 		int or = nv_encoder->or, link = !(entry->dpconf.sor.link & 1);
 		uint32_t tmp;
 
-		if (dev_priv->chipset < 0x90 ||
-		    dev_priv->chipset == 0x92 || dev_priv->chipset == 0xa0)
-			tmp = nv_rd32(dev, NV50_PDISPLAY_SOR_MODE_CTRL_C(or));
-		else
-			tmp = nv_rd32(dev, NV90_PDISPLAY_SOR_MODE_CTRL_C(or));
+		tmp = nv_rd32(dev, 0x61c700 + (or * 0x800));
 
 		switch ((tmp & 0x00000f00) >> 8) {
 		case 8:
diff --git a/drivers/gpu/drm/radeon/Makefile b/drivers/gpu/drm/radeon/Makefile
index 3c91312dea9a..84b1f2729d43 100644
--- a/drivers/gpu/drm/radeon/Makefile
+++ b/drivers/gpu/drm/radeon/Makefile
@@ -33,6 +33,9 @@ $(obj)/rs600_reg_safe.h: $(src)/reg_srcs/rs600 $(obj)/mkregtable
 $(obj)/r600_reg_safe.h: $(src)/reg_srcs/r600 $(obj)/mkregtable
 	$(call if_changed,mkregtable)
 
+$(obj)/evergreen_reg_safe.h: $(src)/reg_srcs/evergreen $(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
@@ -47,6 +50,8 @@ $(obj)/rs600.o: $(obj)/rs600_reg_safe.h
 
 $(obj)/r600_cs.o: $(obj)/r600_reg_safe.h
 
+$(obj)/evergreen_cs.o: $(obj)/evergreen_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
@@ -60,7 +65,7 @@ radeon-y += radeon_device.o radeon_asic.o radeon_kms.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 radeon_pm.o atombios_dp.o r600_audio.o r600_hdmi.o \
-	evergreen.o
+	evergreen.o evergreen_cs.o
 
 radeon-$(CONFIG_COMPAT) += radeon_ioc32.o
 radeon-$(CONFIG_VGA_SWITCHEROO) += radeon_atpx_handler.o
diff --git a/drivers/gpu/drm/radeon/atombios_crtc.c b/drivers/gpu/drm/radeon/atombios_crtc.c
index 03dd6c41dc19..f3f2827017ef 100644
--- a/drivers/gpu/drm/radeon/atombios_crtc.c
+++ b/drivers/gpu/drm/radeon/atombios_crtc.c
@@ -707,6 +707,7 @@ static void atombios_crtc_set_pll(struct drm_crtc *crtc, struct drm_display_mode
 		break;
 	case ATOM_DCPLL:
 	case ATOM_PPLL_INVALID:
+	default:
 		pll = &rdev->clock.dcpll;
 		break;
 	}
diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c
index 8c8e4d3cbaa3..4b6623df3b96 100644
--- a/drivers/gpu/drm/radeon/evergreen.c
+++ b/drivers/gpu/drm/radeon/evergreen.c
@@ -41,7 +41,18 @@ void evergreen_fini(struct radeon_device *rdev);
 
 void evergreen_pm_misc(struct radeon_device *rdev)
 {
-
+	int req_ps_idx = rdev->pm.requested_power_state_index;
+	int req_cm_idx = rdev->pm.requested_clock_mode_index;
+	struct radeon_power_state *ps = &rdev->pm.power_state[req_ps_idx];
+	struct radeon_voltage *voltage = &ps->clock_info[req_cm_idx].voltage;
+
+	if ((voltage->type == VOLTAGE_SW) && voltage->voltage) {
+		if (voltage->voltage != rdev->pm.current_vddc) {
+			radeon_atom_set_voltage(rdev, voltage->voltage);
+			rdev->pm.current_vddc = voltage->voltage;
+			DRM_DEBUG("Setting: v: %d\n", voltage->voltage);
+		}
+	}
 }
 
 void evergreen_pm_prepare(struct radeon_device *rdev)
@@ -2148,7 +2159,7 @@ int evergreen_init(struct radeon_device *rdev)
 	if (r)
 		return r;
 
-	rdev->accel_working = false;
+	rdev->accel_working = true;
 	r = evergreen_startup(rdev);
 	if (r) {
 		dev_err(rdev->dev, "disabling GPU acceleration\n");
diff --git a/drivers/gpu/drm/radeon/evergreen_cs.c b/drivers/gpu/drm/radeon/evergreen_cs.c
new file mode 100644
index 000000000000..64516b950891
--- /dev/null
+++ b/drivers/gpu/drm/radeon/evergreen_cs.c
@@ -0,0 +1,1356 @@
+/*
+ * Copyright 2010 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
+ */
+#include "drmP.h"
+#include "radeon.h"
+#include "evergreend.h"
+#include "evergreen_reg_safe.h"
+
+static int evergreen_cs_packet_next_reloc(struct radeon_cs_parser *p,
+					  struct radeon_cs_reloc **cs_reloc);
+
+struct evergreen_cs_track {
+	u32			group_size;
+	u32			nbanks;
+	u32			npipes;
+	/* value we track */
+	u32			nsamples;
+	u32			cb_color_base_last[12];
+	struct radeon_bo	*cb_color_bo[12];
+	u32			cb_color_bo_offset[12];
+	struct radeon_bo	*cb_color_fmask_bo[8];
+	struct radeon_bo	*cb_color_cmask_bo[8];
+	u32			cb_color_info[12];
+	u32			cb_color_view[12];
+	u32			cb_color_pitch_idx[12];
+	u32			cb_color_slice_idx[12];
+	u32			cb_color_dim_idx[12];
+	u32			cb_color_dim[12];
+	u32			cb_color_pitch[12];
+	u32			cb_color_slice[12];
+	u32			cb_color_cmask_slice[8];
+	u32			cb_color_fmask_slice[8];
+	u32			cb_target_mask;
+	u32			cb_shader_mask;
+	u32			vgt_strmout_config;
+	u32			vgt_strmout_buffer_config;
+	u32			db_depth_control;
+	u32			db_depth_view;
+	u32			db_depth_size;
+	u32			db_depth_size_idx;
+	u32			db_z_info;
+	u32			db_z_idx;
+	u32			db_z_read_offset;
+	u32			db_z_write_offset;
+	struct radeon_bo	*db_z_read_bo;
+	struct radeon_bo	*db_z_write_bo;
+	u32			db_s_info;
+	u32			db_s_idx;
+	u32			db_s_read_offset;
+	u32			db_s_write_offset;
+	struct radeon_bo	*db_s_read_bo;
+	struct radeon_bo	*db_s_write_bo;
+};
+
+static void evergreen_cs_track_init(struct evergreen_cs_track *track)
+{
+	int i;
+
+	for (i = 0; i < 8; i++) {
+		track->cb_color_fmask_bo[i] = NULL;
+		track->cb_color_cmask_bo[i] = NULL;
+		track->cb_color_cmask_slice[i] = 0;
+		track->cb_color_fmask_slice[i] = 0;
+	}
+
+	for (i = 0; i < 12; i++) {
+		track->cb_color_base_last[i] = 0;
+		track->cb_color_bo[i] = NULL;
+		track->cb_color_bo_offset[i] = 0xFFFFFFFF;
+		track->cb_color_info[i] = 0;
+		track->cb_color_view[i] = 0;
+		track->cb_color_pitch_idx[i] = 0;
+		track->cb_color_slice_idx[i] = 0;
+		track->cb_color_dim[i] = 0;
+		track->cb_color_pitch[i] = 0;
+		track->cb_color_slice[i] = 0;
+		track->cb_color_dim[i] = 0;
+	}
+	track->cb_target_mask = 0xFFFFFFFF;
+	track->cb_shader_mask = 0xFFFFFFFF;
+
+	track->db_depth_view = 0xFFFFC000;
+	track->db_depth_size = 0xFFFFFFFF;
+	track->db_depth_size_idx = 0;
+	track->db_depth_control = 0xFFFFFFFF;
+	track->db_z_info = 0xFFFFFFFF;
+	track->db_z_idx = 0xFFFFFFFF;
+	track->db_z_read_offset = 0xFFFFFFFF;
+	track->db_z_write_offset = 0xFFFFFFFF;
+	track->db_z_read_bo = NULL;
+	track->db_z_write_bo = NULL;
+	track->db_s_info = 0xFFFFFFFF;
+	track->db_s_idx = 0xFFFFFFFF;
+	track->db_s_read_offset = 0xFFFFFFFF;
+	track->db_s_write_offset = 0xFFFFFFFF;
+	track->db_s_read_bo = NULL;
+	track->db_s_write_bo = NULL;
+}
+
+static inline int evergreen_cs_track_validate_cb(struct radeon_cs_parser *p, int i)
+{
+	/* XXX fill in */
+	return 0;
+}
+
+static int evergreen_cs_track_check(struct radeon_cs_parser *p)
+{
+	struct evergreen_cs_track *track = p->track;
+
+	/* we don't support stream out buffer yet */
+	if (track->vgt_strmout_config || track->vgt_strmout_buffer_config) {
+		dev_warn(p->dev, "this kernel doesn't support SMX output buffer\n");
+		return -EINVAL;
+	}
+
+	/* XXX fill in */
+	return 0;
+}
+
+/**
+ * evergreen_cs_packet_parse() - parse cp packet and point ib index to next packet
+ * @parser:	parser structure holding parsing context.
+ * @pkt:	where to store packet informations
+ *
+ * Assume that chunk_ib_index is properly set. Will return -EINVAL
+ * if packet is bigger than remaining ib size. or if packets is unknown.
+ **/
+int evergreen_cs_packet_parse(struct radeon_cs_parser *p,
+			      struct radeon_cs_packet *pkt,
+			      unsigned idx)
+{
+	struct radeon_cs_chunk *ib_chunk = &p->chunks[p->chunk_ib_idx];
+	uint32_t header;
+
+	if (idx >= ib_chunk->length_dw) {
+		DRM_ERROR("Can not parse packet at %d after CS end %d !\n",
+			  idx, ib_chunk->length_dw);
+		return -EINVAL;
+	}
+	header = radeon_get_ib_value(p, idx);
+	pkt->idx = idx;
+	pkt->type = CP_PACKET_GET_TYPE(header);
+	pkt->count = CP_PACKET_GET_COUNT(header);
+	pkt->one_reg_wr = 0;
+	switch (pkt->type) {
+	case PACKET_TYPE0:
+		pkt->reg = CP_PACKET0_GET_REG(header);
+		break;
+	case PACKET_TYPE3:
+		pkt->opcode = CP_PACKET3_GET_OPCODE(header);
+		break;
+	case PACKET_TYPE2:
+		pkt->count = -1;
+		break;
+	default:
+		DRM_ERROR("Unknown packet type %d at %d !\n", pkt->type, idx);
+		return -EINVAL;
+	}
+	if ((pkt->count + 1 + pkt->idx) >= ib_chunk->length_dw) {
+		DRM_ERROR("Packet (%d:%d:%d) end after CS buffer (%d) !\n",
+			  pkt->idx, pkt->type, pkt->count, ib_chunk->length_dw);
+		return -EINVAL;
+	}
+	return 0;
+}
+
+/**
+ * evergreen_cs_packet_next_reloc() - parse next packet which should be reloc packet3
+ * @parser:		parser structure holding parsing context.
+ * @data:		pointer to relocation data
+ * @offset_start:	starting offset
+ * @offset_mask:	offset mask (to align start offset on)
+ * @reloc:		reloc informations
+ *
+ * Check next packet is relocation packet3, do bo validation and compute
+ * GPU offset using the provided start.
+ **/
+static int evergreen_cs_packet_next_reloc(struct radeon_cs_parser *p,
+					  struct radeon_cs_reloc **cs_reloc)
+{
+	struct radeon_cs_chunk *relocs_chunk;
+	struct radeon_cs_packet p3reloc;
+	unsigned idx;
+	int r;
+
+	if (p->chunk_relocs_idx == -1) {
+		DRM_ERROR("No relocation chunk !\n");
+		return -EINVAL;
+	}
+	*cs_reloc = NULL;
+	relocs_chunk = &p->chunks[p->chunk_relocs_idx];
+	r = evergreen_cs_packet_parse(p, &p3reloc, p->idx);
+	if (r) {
+		return r;
+	}
+	p->idx += p3reloc.count + 2;
+	if (p3reloc.type != PACKET_TYPE3 || p3reloc.opcode != PACKET3_NOP) {
+		DRM_ERROR("No packet3 for relocation for packet at %d.\n",
+			  p3reloc.idx);
+		return -EINVAL;
+	}
+	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;
+	}
+	/* FIXME: we assume reloc size is 4 dwords */
+	*cs_reloc = p->relocs_ptr[(idx / 4)];
+	return 0;
+}
+
+/**
+ * evergreen_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 evergreen_cs_packet_next_is_pkt3_nop(struct radeon_cs_parser *p)
+{
+	struct radeon_cs_packet p3reloc;
+	int r;
+
+	r = evergreen_cs_packet_parse(p, &p3reloc, p->idx);
+	if (r) {
+		return 0;
+	}
+	if (p3reloc.type != PACKET_TYPE3 || p3reloc.opcode != PACKET3_NOP) {
+		return 0;
+	}
+	return 1;
+}
+
+/**
+ * evergreen_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 evergreen_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 = evergreen_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) != EVERGREEN_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) != EVERGREEN_VLINE_STAT) {
+		DRM_ERROR("vline WAIT_REG_MEM bad bit mask\n");
+		r = -EINVAL;
+		return r;
+	}
+
+	/* jump over the NOP */
+	r = evergreen_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 {
+		switch (reg) {
+		case EVERGREEN_VLINE_START_END:
+			header &= ~R600_CP_PACKET0_REG_MASK;
+			header |= (EVERGREEN_VLINE_START_END + radeon_crtc->crtc_offset) >> 2;
+			ib[h_idx] = header;
+			ib[h_idx + 4] = (EVERGREEN_VLINE_STATUS + radeon_crtc->crtc_offset) >> 2;
+			break;
+		default:
+			DRM_ERROR("unknown crtc reloc\n");
+			r = -EINVAL;
+			goto out;
+		}
+	}
+out:
+	mutex_unlock(&p->rdev->ddev->mode_config.mutex);
+	return r;
+}
+
+static int evergreen_packet0_check(struct radeon_cs_parser *p,
+				   struct radeon_cs_packet *pkt,
+				   unsigned idx, unsigned reg)
+{
+	int r;
+
+	switch (reg) {
+	case EVERGREEN_VLINE_START_END:
+		r = evergreen_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",
+		       reg, idx);
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static int evergreen_cs_parse_packet0(struct radeon_cs_parser *p,
+				      struct radeon_cs_packet *pkt)
+{
+	unsigned reg, i;
+	unsigned idx;
+	int r;
+
+	idx = pkt->idx + 1;
+	reg = pkt->reg;
+	for (i = 0; i <= pkt->count; i++, idx++, reg += 4) {
+		r = evergreen_packet0_check(p, pkt, idx, reg);
+		if (r) {
+			return r;
+		}
+	}
+	return 0;
+}
+
+/**
+ * evergreen_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 evergreen_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 evergreen_cs_check_reg(struct radeon_cs_parser *p, u32 reg, u32 idx)
+{
+	struct evergreen_cs_track *track = (struct evergreen_cs_track *)p->track;
+	struct radeon_cs_reloc *reloc;
+	u32 last_reg = ARRAY_SIZE(evergreen_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 (!(evergreen_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 SQ_ESGS_RING_SIZE:
+	case SQ_GSVS_RING_SIZE:
+	case SQ_ESTMP_RING_SIZE:
+	case SQ_GSTMP_RING_SIZE:
+	case SQ_HSTMP_RING_SIZE:
+	case SQ_LSTMP_RING_SIZE:
+	case SQ_PSTMP_RING_SIZE:
+	case SQ_VSTMP_RING_SIZE:
+	case SQ_ESGS_RING_ITEMSIZE:
+	case SQ_ESTMP_RING_ITEMSIZE:
+	case SQ_GSTMP_RING_ITEMSIZE:
+	case SQ_GSVS_RING_ITEMSIZE:
+	case SQ_GS_VERT_ITEMSIZE:
+	case SQ_GS_VERT_ITEMSIZE_1:
+	case SQ_GS_VERT_ITEMSIZE_2:
+	case SQ_GS_VERT_ITEMSIZE_3:
+	case SQ_GSVS_RING_OFFSET_1:
+	case SQ_GSVS_RING_OFFSET_2:
+	case SQ_GSVS_RING_OFFSET_3:
+	case SQ_HSTMP_RING_ITEMSIZE:
+	case SQ_LSTMP_RING_ITEMSIZE:
+	case SQ_PSTMP_RING_ITEMSIZE:
+	case SQ_VSTMP_RING_ITEMSIZE:
+	case VGT_TF_RING_SIZE:
+		/* get value to populate the IB don't remove */
+		tmp =radeon_get_ib_value(p, idx);
+		ib[idx] = 0;
+		break;
+	case DB_DEPTH_CONTROL:
+		track->db_depth_control = radeon_get_ib_value(p, idx);
+		break;
+	case DB_Z_INFO:
+		r = evergreen_cs_packet_next_reloc(p, &reloc);
+		if (r) {
+			dev_warn(p->dev, "bad SET_CONTEXT_REG "
+					"0x%04X\n", reg);
+			return -EINVAL;
+		}
+		track->db_z_info = radeon_get_ib_value(p, idx);
+		ib[idx] &= ~Z_ARRAY_MODE(0xf);
+		track->db_z_info &= ~Z_ARRAY_MODE(0xf);
+		if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) {
+			ib[idx] |= Z_ARRAY_MODE(ARRAY_2D_TILED_THIN1);
+			track->db_z_info |= Z_ARRAY_MODE(ARRAY_2D_TILED_THIN1);
+		} else {
+			ib[idx] |= Z_ARRAY_MODE(ARRAY_1D_TILED_THIN1);
+			track->db_z_info |= Z_ARRAY_MODE(ARRAY_1D_TILED_THIN1);
+		}
+		break;
+	case DB_STENCIL_INFO:
+		track->db_s_info = radeon_get_ib_value(p, idx);
+		break;
+	case DB_DEPTH_VIEW:
+		track->db_depth_view = radeon_get_ib_value(p, idx);
+		break;
+	case DB_DEPTH_SIZE:
+		track->db_depth_size = radeon_get_ib_value(p, idx);
+		track->db_depth_size_idx = idx;
+		break;
+	case DB_Z_READ_BASE:
+		r = evergreen_cs_packet_next_reloc(p, &reloc);
+		if (r) {
+			dev_warn(p->dev, "bad SET_CONTEXT_REG "
+					"0x%04X\n", reg);
+			return -EINVAL;
+		}
+		track->db_z_read_offset = radeon_get_ib_value(p, idx);
+		ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
+		track->db_z_read_bo = reloc->robj;
+		break;
+	case DB_Z_WRITE_BASE:
+		r = evergreen_cs_packet_next_reloc(p, &reloc);
+		if (r) {
+			dev_warn(p->dev, "bad SET_CONTEXT_REG "
+					"0x%04X\n", reg);
+			return -EINVAL;
+		}
+		track->db_z_write_offset = radeon_get_ib_value(p, idx);
+		ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
+		track->db_z_write_bo = reloc->robj;
+		break;
+	case DB_STENCIL_READ_BASE:
+		r = evergreen_cs_packet_next_reloc(p, &reloc);
+		if (r) {
+			dev_warn(p->dev, "bad SET_CONTEXT_REG "
+					"0x%04X\n", reg);
+			return -EINVAL;
+		}
+		track->db_s_read_offset = radeon_get_ib_value(p, idx);
+		ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
+		track->db_s_read_bo = reloc->robj;
+		break;
+	case DB_STENCIL_WRITE_BASE:
+		r = evergreen_cs_packet_next_reloc(p, &reloc);
+		if (r) {
+			dev_warn(p->dev, "bad SET_CONTEXT_REG "
+					"0x%04X\n", reg);
+			return -EINVAL;
+		}
+		track->db_s_write_offset = radeon_get_ib_value(p, idx);
+		ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
+		track->db_s_write_bo = reloc->robj;
+		break;
+	case VGT_STRMOUT_CONFIG:
+		track->vgt_strmout_config = radeon_get_ib_value(p, idx);
+		break;
+	case VGT_STRMOUT_BUFFER_CONFIG:
+		track->vgt_strmout_buffer_config = radeon_get_ib_value(p, idx);
+		break;
+	case CB_TARGET_MASK:
+		track->cb_target_mask = radeon_get_ib_value(p, idx);
+		break;
+	case CB_SHADER_MASK:
+		track->cb_shader_mask = radeon_get_ib_value(p, idx);
+		break;
+	case PA_SC_AA_CONFIG:
+		tmp = radeon_get_ib_value(p, idx) & MSAA_NUM_SAMPLES_MASK;
+		track->nsamples = 1 << tmp;
+		break;
+	case CB_COLOR0_VIEW:
+	case CB_COLOR1_VIEW:
+	case CB_COLOR2_VIEW:
+	case CB_COLOR3_VIEW:
+	case CB_COLOR4_VIEW:
+	case CB_COLOR5_VIEW:
+	case CB_COLOR6_VIEW:
+	case CB_COLOR7_VIEW:
+		tmp = (reg - CB_COLOR0_VIEW) / 0x3c;
+		track->cb_color_view[tmp] = radeon_get_ib_value(p, idx);
+		break;
+	case CB_COLOR8_VIEW:
+	case CB_COLOR9_VIEW:
+	case CB_COLOR10_VIEW:
+	case CB_COLOR11_VIEW:
+		tmp = ((reg - CB_COLOR8_VIEW) / 0x1c) + 8;
+		track->cb_color_view[tmp] = radeon_get_ib_value(p, idx);
+		break;
+	case CB_COLOR0_INFO:
+	case CB_COLOR1_INFO:
+	case CB_COLOR2_INFO:
+	case CB_COLOR3_INFO:
+	case CB_COLOR4_INFO:
+	case CB_COLOR5_INFO:
+	case CB_COLOR6_INFO:
+	case CB_COLOR7_INFO:
+		r = evergreen_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_INFO) / 0x3c;
+		track->cb_color_info[tmp] = radeon_get_ib_value(p, idx);
+		if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) {
+			ib[idx] |= CB_ARRAY_MODE(ARRAY_2D_TILED_THIN1);
+			track->cb_color_info[tmp] |= CB_ARRAY_MODE(ARRAY_2D_TILED_THIN1);
+		} else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) {
+			ib[idx] |= CB_ARRAY_MODE(ARRAY_1D_TILED_THIN1);
+			track->cb_color_info[tmp] |= CB_ARRAY_MODE(ARRAY_1D_TILED_THIN1);
+		}
+		break;
+	case CB_COLOR8_INFO:
+	case CB_COLOR9_INFO:
+	case CB_COLOR10_INFO:
+	case CB_COLOR11_INFO:
+		r = evergreen_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_COLOR8_INFO) / 0x1c) + 8;
+		track->cb_color_info[tmp] = radeon_get_ib_value(p, idx);
+		if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) {
+			ib[idx] |= CB_ARRAY_MODE(ARRAY_2D_TILED_THIN1);
+			track->cb_color_info[tmp] |= CB_ARRAY_MODE(ARRAY_2D_TILED_THIN1);
+		} else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) {
+			ib[idx] |= CB_ARRAY_MODE(ARRAY_1D_TILED_THIN1);
+			track->cb_color_info[tmp] |= CB_ARRAY_MODE(ARRAY_1D_TILED_THIN1);
+		}
+		break;
+	case CB_COLOR0_PITCH:
+	case CB_COLOR1_PITCH:
+	case CB_COLOR2_PITCH:
+	case CB_COLOR3_PITCH:
+	case CB_COLOR4_PITCH:
+	case CB_COLOR5_PITCH:
+	case CB_COLOR6_PITCH:
+	case CB_COLOR7_PITCH:
+		tmp = (reg - CB_COLOR0_PITCH) / 0x3c;
+		track->cb_color_pitch[tmp] = radeon_get_ib_value(p, idx);
+		track->cb_color_pitch_idx[tmp] = idx;
+		break;
+	case CB_COLOR8_PITCH:
+	case CB_COLOR9_PITCH:
+	case CB_COLOR10_PITCH:
+	case CB_COLOR11_PITCH:
+		tmp = ((reg - CB_COLOR8_PITCH) / 0x1c) + 8;
+		track->cb_color_pitch[tmp] = radeon_get_ib_value(p, idx);
+		track->cb_color_pitch_idx[tmp] = idx;
+		break;
+	case CB_COLOR0_SLICE:
+	case CB_COLOR1_SLICE:
+	case CB_COLOR2_SLICE:
+	case CB_COLOR3_SLICE:
+	case CB_COLOR4_SLICE:
+	case CB_COLOR5_SLICE:
+	case CB_COLOR6_SLICE:
+	case CB_COLOR7_SLICE:
+		tmp = (reg - CB_COLOR0_SLICE) / 0x3c;
+		track->cb_color_slice[tmp] = radeon_get_ib_value(p, idx);
+		track->cb_color_slice_idx[tmp] = idx;
+		break;
+	case CB_COLOR8_SLICE:
+	case CB_COLOR9_SLICE:
+	case CB_COLOR10_SLICE:
+	case CB_COLOR11_SLICE:
+		tmp = ((reg - CB_COLOR8_SLICE) / 0x1c) + 8;
+		track->cb_color_slice[tmp] = radeon_get_ib_value(p, idx);
+		track->cb_color_slice_idx[tmp] = idx;
+		break;
+	case CB_COLOR0_ATTRIB:
+	case CB_COLOR1_ATTRIB:
+	case CB_COLOR2_ATTRIB:
+	case CB_COLOR3_ATTRIB:
+	case CB_COLOR4_ATTRIB:
+	case CB_COLOR5_ATTRIB:
+	case CB_COLOR6_ATTRIB:
+	case CB_COLOR7_ATTRIB:
+	case CB_COLOR8_ATTRIB:
+	case CB_COLOR9_ATTRIB:
+	case CB_COLOR10_ATTRIB:
+	case CB_COLOR11_ATTRIB:
+		break;
+	case CB_COLOR0_DIM:
+	case CB_COLOR1_DIM:
+	case CB_COLOR2_DIM:
+	case CB_COLOR3_DIM:
+	case CB_COLOR4_DIM:
+	case CB_COLOR5_DIM:
+	case CB_COLOR6_DIM:
+	case CB_COLOR7_DIM:
+		tmp = (reg - CB_COLOR0_DIM) / 0x3c;
+		track->cb_color_dim[tmp] = radeon_get_ib_value(p, idx);
+		track->cb_color_dim_idx[tmp] = idx;
+		break;
+	case CB_COLOR8_DIM:
+	case CB_COLOR9_DIM:
+	case CB_COLOR10_DIM:
+	case CB_COLOR11_DIM:
+		tmp = ((reg - CB_COLOR8_DIM) / 0x1c) + 8;
+		track->cb_color_dim[tmp] = radeon_get_ib_value(p, idx);
+		track->cb_color_dim_idx[tmp] = idx;
+		break;
+	case CB_COLOR0_FMASK:
+	case CB_COLOR1_FMASK:
+	case CB_COLOR2_FMASK:
+	case CB_COLOR3_FMASK:
+	case CB_COLOR4_FMASK:
+	case CB_COLOR5_FMASK:
+	case CB_COLOR6_FMASK:
+	case CB_COLOR7_FMASK:
+		tmp = (reg - CB_COLOR0_FMASK) / 0x3c;
+		r = evergreen_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_fmask_bo[tmp] = reloc->robj;
+		break;
+	case CB_COLOR0_CMASK:
+	case CB_COLOR1_CMASK:
+	case CB_COLOR2_CMASK:
+	case CB_COLOR3_CMASK:
+	case CB_COLOR4_CMASK:
+	case CB_COLOR5_CMASK:
+	case CB_COLOR6_CMASK:
+	case CB_COLOR7_CMASK:
+		tmp = (reg - CB_COLOR0_CMASK) / 0x3c;
+		r = evergreen_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_cmask_bo[tmp] = reloc->robj;
+		break;
+	case CB_COLOR0_FMASK_SLICE:
+	case CB_COLOR1_FMASK_SLICE:
+	case CB_COLOR2_FMASK_SLICE:
+	case CB_COLOR3_FMASK_SLICE:
+	case CB_COLOR4_FMASK_SLICE:
+	case CB_COLOR5_FMASK_SLICE:
+	case CB_COLOR6_FMASK_SLICE:
+	case CB_COLOR7_FMASK_SLICE:
+		tmp = (reg - CB_COLOR0_FMASK_SLICE) / 0x3c;
+		track->cb_color_fmask_slice[tmp] = radeon_get_ib_value(p, idx);
+		break;
+	case CB_COLOR0_CMASK_SLICE:
+	case CB_COLOR1_CMASK_SLICE:
+	case CB_COLOR2_CMASK_SLICE:
+	case CB_COLOR3_CMASK_SLICE:
+	case CB_COLOR4_CMASK_SLICE:
+	case CB_COLOR5_CMASK_SLICE:
+	case CB_COLOR6_CMASK_SLICE:
+	case CB_COLOR7_CMASK_SLICE:
+		tmp = (reg - CB_COLOR0_CMASK_SLICE) / 0x3c;
+		track->cb_color_cmask_slice[tmp] = radeon_get_ib_value(p, idx);
+		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 = evergreen_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) / 0x3c;
+		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 CB_COLOR8_BASE:
+	case CB_COLOR9_BASE:
+	case CB_COLOR10_BASE:
+	case CB_COLOR11_BASE:
+		r = evergreen_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_COLOR8_BASE) / 0x1c) + 8;
+		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 CB_IMMED0_BASE:
+	case CB_IMMED1_BASE:
+	case CB_IMMED2_BASE:
+	case CB_IMMED3_BASE:
+	case CB_IMMED4_BASE:
+	case CB_IMMED5_BASE:
+	case CB_IMMED6_BASE:
+	case CB_IMMED7_BASE:
+	case CB_IMMED8_BASE:
+	case CB_IMMED9_BASE:
+	case CB_IMMED10_BASE:
+	case CB_IMMED11_BASE:
+	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:
+	case SQ_PGM_START_HS:
+	case SQ_PGM_START_LS:
+	case GDS_ADDR_BASE:
+	case SQ_CONST_MEM_BASE:
+	case SQ_ALU_CONST_CACHE_GS_0:
+	case SQ_ALU_CONST_CACHE_GS_1:
+	case SQ_ALU_CONST_CACHE_GS_2:
+	case SQ_ALU_CONST_CACHE_GS_3:
+	case SQ_ALU_CONST_CACHE_GS_4:
+	case SQ_ALU_CONST_CACHE_GS_5:
+	case SQ_ALU_CONST_CACHE_GS_6:
+	case SQ_ALU_CONST_CACHE_GS_7:
+	case SQ_ALU_CONST_CACHE_GS_8:
+	case SQ_ALU_CONST_CACHE_GS_9:
+	case SQ_ALU_CONST_CACHE_GS_10:
+	case SQ_ALU_CONST_CACHE_GS_11:
+	case SQ_ALU_CONST_CACHE_GS_12:
+	case SQ_ALU_CONST_CACHE_GS_13:
+	case SQ_ALU_CONST_CACHE_GS_14:
+	case SQ_ALU_CONST_CACHE_GS_15:
+	case SQ_ALU_CONST_CACHE_PS_0:
+	case SQ_ALU_CONST_CACHE_PS_1:
+	case SQ_ALU_CONST_CACHE_PS_2:
+	case SQ_ALU_CONST_CACHE_PS_3:
+	case SQ_ALU_CONST_CACHE_PS_4:
+	case SQ_ALU_CONST_CACHE_PS_5:
+	case SQ_ALU_CONST_CACHE_PS_6:
+	case SQ_ALU_CONST_CACHE_PS_7:
+	case SQ_ALU_CONST_CACHE_PS_8:
+	case SQ_ALU_CONST_CACHE_PS_9:
+	case SQ_ALU_CONST_CACHE_PS_10:
+	case SQ_ALU_CONST_CACHE_PS_11:
+	case SQ_ALU_CONST_CACHE_PS_12:
+	case SQ_ALU_CONST_CACHE_PS_13:
+	case SQ_ALU_CONST_CACHE_PS_14:
+	case SQ_ALU_CONST_CACHE_PS_15:
+	case SQ_ALU_CONST_CACHE_VS_0:
+	case SQ_ALU_CONST_CACHE_VS_1:
+	case SQ_ALU_CONST_CACHE_VS_2:
+	case SQ_ALU_CONST_CACHE_VS_3:
+	case SQ_ALU_CONST_CACHE_VS_4:
+	case SQ_ALU_CONST_CACHE_VS_5:
+	case SQ_ALU_CONST_CACHE_VS_6:
+	case SQ_ALU_CONST_CACHE_VS_7:
+	case SQ_ALU_CONST_CACHE_VS_8:
+	case SQ_ALU_CONST_CACHE_VS_9:
+	case SQ_ALU_CONST_CACHE_VS_10:
+	case SQ_ALU_CONST_CACHE_VS_11:
+	case SQ_ALU_CONST_CACHE_VS_12:
+	case SQ_ALU_CONST_CACHE_VS_13:
+	case SQ_ALU_CONST_CACHE_VS_14:
+	case SQ_ALU_CONST_CACHE_VS_15:
+	case SQ_ALU_CONST_CACHE_HS_0:
+	case SQ_ALU_CONST_CACHE_HS_1:
+	case SQ_ALU_CONST_CACHE_HS_2:
+	case SQ_ALU_CONST_CACHE_HS_3:
+	case SQ_ALU_CONST_CACHE_HS_4:
+	case SQ_ALU_CONST_CACHE_HS_5:
+	case SQ_ALU_CONST_CACHE_HS_6:
+	case SQ_ALU_CONST_CACHE_HS_7:
+	case SQ_ALU_CONST_CACHE_HS_8:
+	case SQ_ALU_CONST_CACHE_HS_9:
+	case SQ_ALU_CONST_CACHE_HS_10:
+	case SQ_ALU_CONST_CACHE_HS_11:
+	case SQ_ALU_CONST_CACHE_HS_12:
+	case SQ_ALU_CONST_CACHE_HS_13:
+	case SQ_ALU_CONST_CACHE_HS_14:
+	case SQ_ALU_CONST_CACHE_HS_15:
+	case SQ_ALU_CONST_CACHE_LS_0:
+	case SQ_ALU_CONST_CACHE_LS_1:
+	case SQ_ALU_CONST_CACHE_LS_2:
+	case SQ_ALU_CONST_CACHE_LS_3:
+	case SQ_ALU_CONST_CACHE_LS_4:
+	case SQ_ALU_CONST_CACHE_LS_5:
+	case SQ_ALU_CONST_CACHE_LS_6:
+	case SQ_ALU_CONST_CACHE_LS_7:
+	case SQ_ALU_CONST_CACHE_LS_8:
+	case SQ_ALU_CONST_CACHE_LS_9:
+	case SQ_ALU_CONST_CACHE_LS_10:
+	case SQ_ALU_CONST_CACHE_LS_11:
+	case SQ_ALU_CONST_CACHE_LS_12:
+	case SQ_ALU_CONST_CACHE_LS_13:
+	case SQ_ALU_CONST_CACHE_LS_14:
+	case SQ_ALU_CONST_CACHE_LS_15:
+		r = evergreen_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;
+}
+
+/**
+ * evergreen_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 evergreen_check_texture_resource(struct radeon_cs_parser *p,  u32 idx,
+						   struct radeon_bo *texture,
+						   struct radeon_bo *mipmap)
+{
+	/* XXX fill in */
+	return 0;
+}
+
+static int evergreen_packet3_check(struct radeon_cs_parser *p,
+				   struct radeon_cs_packet *pkt)
+{
+	struct radeon_cs_reloc *reloc;
+	struct evergreen_cs_track *track;
+	volatile u32 *ib;
+	unsigned idx;
+	unsigned i;
+	unsigned start_reg, end_reg, reg;
+	int r;
+	u32 idx_value;
+
+	track = (struct evergreen_cs_track *)p->track;
+	ib = p->ib->ptr;
+	idx = pkt->idx + 1;
+	idx_value = radeon_get_ib_value(p, idx);
+
+	switch (pkt->opcode) {
+	case PACKET3_CONTEXT_CONTROL:
+		if (pkt->count != 1) {
+			DRM_ERROR("bad CONTEXT_CONTROL\n");
+			return -EINVAL;
+		}
+		break;
+	case PACKET3_INDEX_TYPE:
+	case PACKET3_NUM_INSTANCES:
+	case PACKET3_CLEAR_STATE:
+		if (pkt->count) {
+			DRM_ERROR("bad INDEX_TYPE/NUM_INSTANCES/CLEAR_STATE\n");
+			return -EINVAL;
+		}
+		break;
+	case PACKET3_INDEX_BASE:
+		if (pkt->count != 1) {
+			DRM_ERROR("bad INDEX_BASE\n");
+			return -EINVAL;
+		}
+		r = evergreen_cs_packet_next_reloc(p, &reloc);
+		if (r) {
+			DRM_ERROR("bad INDEX_BASE\n");
+			return -EINVAL;
+		}
+		ib[idx+0] = idx_value + (u32)(reloc->lobj.gpu_offset & 0xffffffff);
+		ib[idx+1] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
+		r = evergreen_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:
+		if (pkt->count != 3) {
+			DRM_ERROR("bad DRAW_INDEX\n");
+			return -EINVAL;
+		}
+		r = evergreen_cs_packet_next_reloc(p, &reloc);
+		if (r) {
+			DRM_ERROR("bad DRAW_INDEX\n");
+			return -EINVAL;
+		}
+		ib[idx+0] = idx_value + (u32)(reloc->lobj.gpu_offset & 0xffffffff);
+		ib[idx+1] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
+		r = evergreen_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_2:
+		if (pkt->count != 4) {
+			DRM_ERROR("bad DRAW_INDEX_2\n");
+			return -EINVAL;
+		}
+		r = evergreen_cs_packet_next_reloc(p, &reloc);
+		if (r) {
+			DRM_ERROR("bad DRAW_INDEX_2\n");
+			return -EINVAL;
+		}
+		ib[idx+1] = idx_value + (u32)(reloc->lobj.gpu_offset & 0xffffffff);
+		ib[idx+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
+		r = evergreen_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 = evergreen_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_MULTI_AUTO:
+		if (pkt->count != 2) {
+			DRM_ERROR("bad DRAW_INDEX_MULTI_AUTO\n");
+			return -EINVAL;
+		}
+		r = evergreen_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:
+		if (pkt->count < 2) {
+			DRM_ERROR("bad DRAW_INDEX_IMMD\n");
+			return -EINVAL;
+		}
+		r = evergreen_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_OFFSET:
+		if (pkt->count != 2) {
+			DRM_ERROR("bad DRAW_INDEX_OFFSET\n");
+			return -EINVAL;
+		}
+		r = evergreen_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_OFFSET_2:
+		if (pkt->count != 3) {
+			DRM_ERROR("bad DRAW_INDEX_OFFSET_2\n");
+			return -EINVAL;
+		}
+		r = evergreen_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) {
+			DRM_ERROR("bad WAIT_REG_MEM\n");
+			return -EINVAL;
+		}
+		/* bit 4 is reg (0) or mem (1) */
+		if (idx_value & 0x10) {
+			r = evergreen_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;
+		}
+		break;
+	case PACKET3_SURFACE_SYNC:
+		if (pkt->count != 3) {
+			DRM_ERROR("bad SURFACE_SYNC\n");
+			return -EINVAL;
+		}
+		/* 0xffffffff/0x0 is flush all cache flag */
+		if (radeon_get_ib_value(p, idx + 1) != 0xffffffff ||
+		    radeon_get_ib_value(p, idx + 2) != 0) {
+			r = evergreen_cs_packet_next_reloc(p, &reloc);
+			if (r) {
+				DRM_ERROR("bad SURFACE_SYNC\n");
+				return -EINVAL;
+			}
+			ib[idx+2] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
+		}
+		break;
+	case PACKET3_EVENT_WRITE:
+		if (pkt->count != 2 && pkt->count != 0) {
+			DRM_ERROR("bad EVENT_WRITE\n");
+			return -EINVAL;
+		}
+		if (pkt->count) {
+			r = evergreen_cs_packet_next_reloc(p, &reloc);
+			if (r) {
+				DRM_ERROR("bad EVENT_WRITE\n");
+				return -EINVAL;
+			}
+			ib[idx+1] += (u32)(reloc->lobj.gpu_offset & 0xffffffff);
+			ib[idx+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
+		}
+		break;
+	case PACKET3_EVENT_WRITE_EOP:
+		if (pkt->count != 4) {
+			DRM_ERROR("bad EVENT_WRITE_EOP\n");
+			return -EINVAL;
+		}
+		r = evergreen_cs_packet_next_reloc(p, &reloc);
+		if (r) {
+			DRM_ERROR("bad EVENT_WRITE_EOP\n");
+			return -EINVAL;
+		}
+		ib[idx+1] += (u32)(reloc->lobj.gpu_offset & 0xffffffff);
+		ib[idx+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
+		break;
+	case PACKET3_EVENT_WRITE_EOS:
+		if (pkt->count != 3) {
+			DRM_ERROR("bad EVENT_WRITE_EOS\n");
+			return -EINVAL;
+		}
+		r = evergreen_cs_packet_next_reloc(p, &reloc);
+		if (r) {
+			DRM_ERROR("bad EVENT_WRITE_EOS\n");
+			return -EINVAL;
+		}
+		ib[idx+1] += (u32)(reloc->lobj.gpu_offset & 0xffffffff);
+		ib[idx+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
+		break;
+	case PACKET3_SET_CONFIG_REG:
+		start_reg = (idx_value << 2) + PACKET3_SET_CONFIG_REG_START;
+		end_reg = 4 * pkt->count + start_reg - 4;
+		if ((start_reg < PACKET3_SET_CONFIG_REG_START) ||
+		    (start_reg >= PACKET3_SET_CONFIG_REG_END) ||
+		    (end_reg >= PACKET3_SET_CONFIG_REG_END)) {
+			DRM_ERROR("bad PACKET3_SET_CONFIG_REG\n");
+			return -EINVAL;
+		}
+		for (i = 0; i < pkt->count; i++) {
+			reg = start_reg + (4 * i);
+			r = evergreen_cs_check_reg(p, reg, idx+1+i);
+			if (r)
+				return r;
+		}
+		break;
+	case PACKET3_SET_CONTEXT_REG:
+		start_reg = (idx_value << 2) + PACKET3_SET_CONTEXT_REG_START;
+		end_reg = 4 * pkt->count + start_reg - 4;
+		if ((start_reg < PACKET3_SET_CONTEXT_REG_START) ||
+		    (start_reg >= PACKET3_SET_CONTEXT_REG_END) ||
+		    (end_reg >= PACKET3_SET_CONTEXT_REG_END)) {
+			DRM_ERROR("bad PACKET3_SET_CONTEXT_REG\n");
+			return -EINVAL;
+		}
+		for (i = 0; i < pkt->count; i++) {
+			reg = start_reg + (4 * i);
+			r = evergreen_cs_check_reg(p, reg, idx+1+i);
+			if (r)
+				return r;
+		}
+		break;
+	case PACKET3_SET_RESOURCE:
+		if (pkt->count % 8) {
+			DRM_ERROR("bad SET_RESOURCE\n");
+			return -EINVAL;
+		}
+		start_reg = (idx_value << 2) + PACKET3_SET_RESOURCE_START;
+		end_reg = 4 * pkt->count + start_reg - 4;
+		if ((start_reg < PACKET3_SET_RESOURCE_START) ||
+		    (start_reg >= PACKET3_SET_RESOURCE_END) ||
+		    (end_reg >= PACKET3_SET_RESOURCE_END)) {
+			DRM_ERROR("bad SET_RESOURCE\n");
+			return -EINVAL;
+		}
+		for (i = 0; i < (pkt->count / 8); i++) {
+			struct radeon_bo *texture, *mipmap;
+			u32 size, offset;
+
+			switch (G__SQ_CONSTANT_TYPE(radeon_get_ib_value(p, idx+1+(i*8)+7))) {
+			case SQ_TEX_VTX_VALID_TEXTURE:
+				/* tex base */
+				r = evergreen_cs_packet_next_reloc(p, &reloc);
+				if (r) {
+					DRM_ERROR("bad SET_RESOURCE (tex)\n");
+					return -EINVAL;
+				}
+				ib[idx+1+(i*8)+3] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
+				if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO)
+					ib[idx+1+(i*8)+1] |= TEX_ARRAY_MODE(ARRAY_2D_TILED_THIN1);
+				else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
+					ib[idx+1+(i*8)+1] |= TEX_ARRAY_MODE(ARRAY_1D_TILED_THIN1);
+				texture = reloc->robj;
+				/* tex mip base */
+				r = evergreen_cs_packet_next_reloc(p, &reloc);
+				if (r) {
+					DRM_ERROR("bad SET_RESOURCE (tex)\n");
+					return -EINVAL;
+				}
+				ib[idx+1+(i*8)+4] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
+				mipmap = reloc->robj;
+				r = evergreen_check_texture_resource(p,  idx+1+(i*8),
+						texture, mipmap);
+				if (r)
+					return r;
+				break;
+			case SQ_TEX_VTX_VALID_BUFFER:
+				/* vtx base */
+				r = evergreen_cs_packet_next_reloc(p, &reloc);
+				if (r) {
+					DRM_ERROR("bad SET_RESOURCE (vtx)\n");
+					return -EINVAL;
+				}
+				offset = radeon_get_ib_value(p, idx+1+(i*8)+0);
+				size = radeon_get_ib_value(p, idx+1+(i*8)+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*8)+1] = radeon_bo_size(reloc->robj);
+				}
+				ib[idx+1+(i*8)+0] += (u32)((reloc->lobj.gpu_offset) & 0xffffffff);
+				ib[idx+1+(i*8)+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
+				break;
+			case SQ_TEX_VTX_INVALID_TEXTURE:
+			case SQ_TEX_VTX_INVALID_BUFFER:
+			default:
+				DRM_ERROR("bad SET_RESOURCE\n");
+				return -EINVAL;
+			}
+		}
+		break;
+	case PACKET3_SET_ALU_CONST:
+		/* XXX fix me ALU const buffers only */
+		break;
+	case PACKET3_SET_BOOL_CONST:
+		start_reg = (idx_value << 2) + PACKET3_SET_BOOL_CONST_START;
+		end_reg = 4 * pkt->count + start_reg - 4;
+		if ((start_reg < PACKET3_SET_BOOL_CONST_START) ||
+		    (start_reg >= PACKET3_SET_BOOL_CONST_END) ||
+		    (end_reg >= PACKET3_SET_BOOL_CONST_END)) {
+			DRM_ERROR("bad SET_BOOL_CONST\n");
+			return -EINVAL;
+		}
+		break;
+	case PACKET3_SET_LOOP_CONST:
+		start_reg = (idx_value << 2) + PACKET3_SET_LOOP_CONST_START;
+		end_reg = 4 * pkt->count + start_reg - 4;
+		if ((start_reg < PACKET3_SET_LOOP_CONST_START) ||
+		    (start_reg >= PACKET3_SET_LOOP_CONST_END) ||
+		    (end_reg >= PACKET3_SET_LOOP_CONST_END)) {
+			DRM_ERROR("bad SET_LOOP_CONST\n");
+			return -EINVAL;
+		}
+		break;
+	case PACKET3_SET_CTL_CONST:
+		start_reg = (idx_value << 2) + PACKET3_SET_CTL_CONST_START;
+		end_reg = 4 * pkt->count + start_reg - 4;
+		if ((start_reg < PACKET3_SET_CTL_CONST_START) ||
+		    (start_reg >= PACKET3_SET_CTL_CONST_END) ||
+		    (end_reg >= PACKET3_SET_CTL_CONST_END)) {
+			DRM_ERROR("bad SET_CTL_CONST\n");
+			return -EINVAL;
+		}
+		break;
+	case PACKET3_SET_SAMPLER:
+		if (pkt->count % 3) {
+			DRM_ERROR("bad SET_SAMPLER\n");
+			return -EINVAL;
+		}
+		start_reg = (idx_value << 2) + PACKET3_SET_SAMPLER_START;
+		end_reg = 4 * pkt->count + start_reg - 4;
+		if ((start_reg < PACKET3_SET_SAMPLER_START) ||
+		    (start_reg >= PACKET3_SET_SAMPLER_END) ||
+		    (end_reg >= PACKET3_SET_SAMPLER_END)) {
+			DRM_ERROR("bad SET_SAMPLER\n");
+			return -EINVAL;
+		}
+		break;
+	case PACKET3_NOP:
+		break;
+	default:
+		DRM_ERROR("Packet3 opcode %x not supported\n", pkt->opcode);
+		return -EINVAL;
+	}
+	return 0;
+}
+
+int evergreen_cs_parse(struct radeon_cs_parser *p)
+{
+	struct radeon_cs_packet pkt;
+	struct evergreen_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;
+		evergreen_cs_track_init(track);
+		track->npipes = p->rdev->config.evergreen.tiling_npipes;
+		track->nbanks = p->rdev->config.evergreen.tiling_nbanks;
+		track->group_size = p->rdev->config.evergreen.tiling_group_size;
+		p->track = track;
+	}
+	do {
+		r = evergreen_cs_packet_parse(p, &pkt, p->idx);
+		if (r) {
+			kfree(p->track);
+			p->track = NULL;
+			return r;
+		}
+		p->idx += pkt.count + 2;
+		switch (pkt.type) {
+		case PACKET_TYPE0:
+			r = evergreen_cs_parse_packet0(p, &pkt);
+			break;
+		case PACKET_TYPE2:
+			break;
+		case PACKET_TYPE3:
+			r = evergreen_packet3_check(p, &pkt);
+			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);
+#if 0
+	for (r = 0; r < p->ib->length_dw; r++) {
+		printk(KERN_INFO "%05d  0x%08X\n", r, p->ib->ptr[r]);
+		mdelay(1);
+	}
+#endif
+	kfree(p->track);
+	p->track = NULL;
+	return 0;
+}
+
diff --git a/drivers/gpu/drm/radeon/evergreen_reg.h b/drivers/gpu/drm/radeon/evergreen_reg.h
index af86af836f13..e028c1cd9d9b 100644
--- a/drivers/gpu/drm/radeon/evergreen_reg.h
+++ b/drivers/gpu/drm/radeon/evergreen_reg.h
@@ -151,6 +151,9 @@
 #define EVERGREEN_DATA_FORMAT                           0x6b00
 #       define EVERGREEN_INTERLEAVE_EN                  (1 << 0)
 #define EVERGREEN_DESKTOP_HEIGHT                        0x6b04
+#define EVERGREEN_VLINE_START_END                       0x6b08
+#define EVERGREEN_VLINE_STATUS                          0x6bb8
+#       define EVERGREEN_VLINE_STAT                     (1 << 12)
 
 #define EVERGREEN_VIEWPORT_START                        0x6d70
 #define EVERGREEN_VIEWPORT_SIZE                         0x6d74
diff --git a/drivers/gpu/drm/radeon/evergreend.h b/drivers/gpu/drm/radeon/evergreend.h
index 93e9e17ad54a..79683f6b4452 100644
--- a/drivers/gpu/drm/radeon/evergreend.h
+++ b/drivers/gpu/drm/radeon/evergreend.h
@@ -218,6 +218,8 @@
 #define		CLIP_VTX_REORDER_ENA				(1 << 0)
 #define		NUM_CLIP_SEQ(x)					((x) << 1)
 #define PA_SC_AA_CONFIG					0x28C04
+#define         MSAA_NUM_SAMPLES_SHIFT                  0
+#define         MSAA_NUM_SAMPLES_MASK                   0x3
 #define PA_SC_CLIPRECT_RULE				0x2820C
 #define	PA_SC_EDGERULE					0x28230
 #define	PA_SC_FIFO_SIZE					0x8BCC
@@ -553,4 +555,466 @@
 #       define DC_HPDx_RX_INT_TIMER(x)                    ((x) << 16)
 #       define DC_HPDx_EN                                 (1 << 28)
 
+/*
+ * PM4
+ */
+#define	PACKET_TYPE0	0
+#define	PACKET_TYPE1	1
+#define	PACKET_TYPE2	2
+#define	PACKET_TYPE3	3
+
+#define CP_PACKET_GET_TYPE(h) (((h) >> 30) & 3)
+#define CP_PACKET_GET_COUNT(h) (((h) >> 16) & 0x3FFF)
+#define CP_PACKET0_GET_REG(h) (((h) & 0xFFFF) << 2)
+#define CP_PACKET3_GET_OPCODE(h) (((h) >> 8) & 0xFF)
+#define PACKET0(reg, n)	((PACKET_TYPE0 << 30) |				\
+			 (((reg) >> 2) & 0xFFFF) |			\
+			 ((n) & 0x3FFF) << 16)
+#define CP_PACKET2			0x80000000
+#define		PACKET2_PAD_SHIFT		0
+#define		PACKET2_PAD_MASK		(0x3fffffff << 0)
+
+#define PACKET2(v)	(CP_PACKET2 | REG_SET(PACKET2_PAD, (v)))
+
+#define PACKET3(op, n)	((PACKET_TYPE3 << 30) |				\
+			 (((op) & 0xFF) << 8) |				\
+			 ((n) & 0x3FFF) << 16)
+
+/* Packet 3 types */
+#define	PACKET3_NOP					0x10
+#define	PACKET3_SET_BASE				0x11
+#define	PACKET3_CLEAR_STATE				0x12
+#define	PACKET3_INDIRECT_BUFFER_SIZE			0x13
+#define	PACKET3_DISPATCH_DIRECT				0x15
+#define	PACKET3_DISPATCH_INDIRECT			0x16
+#define	PACKET3_INDIRECT_BUFFER_END			0x17
+#define	PACKET3_SET_PREDICATION				0x20
+#define	PACKET3_REG_RMW					0x21
+#define	PACKET3_COND_EXEC				0x22
+#define	PACKET3_PRED_EXEC				0x23
+#define	PACKET3_DRAW_INDIRECT				0x24
+#define	PACKET3_DRAW_INDEX_INDIRECT			0x25
+#define	PACKET3_INDEX_BASE				0x26
+#define	PACKET3_DRAW_INDEX_2				0x27
+#define	PACKET3_CONTEXT_CONTROL				0x28
+#define	PACKET3_DRAW_INDEX_OFFSET			0x29
+#define	PACKET3_INDEX_TYPE				0x2A
+#define	PACKET3_DRAW_INDEX				0x2B
+#define	PACKET3_DRAW_INDEX_AUTO				0x2D
+#define	PACKET3_DRAW_INDEX_IMMD				0x2E
+#define	PACKET3_NUM_INSTANCES				0x2F
+#define	PACKET3_DRAW_INDEX_MULTI_AUTO			0x30
+#define	PACKET3_STRMOUT_BUFFER_UPDATE			0x34
+#define	PACKET3_DRAW_INDEX_OFFSET_2			0x35
+#define	PACKET3_DRAW_INDEX_MULTI_ELEMENT		0x36
+#define	PACKET3_MEM_SEMAPHORE				0x39
+#define	PACKET3_MPEG_INDEX				0x3A
+#define	PACKET3_WAIT_REG_MEM				0x3C
+#define	PACKET3_MEM_WRITE				0x3D
+#define	PACKET3_INDIRECT_BUFFER				0x32
+#define	PACKET3_SURFACE_SYNC				0x43
+#              define PACKET3_CB0_DEST_BASE_ENA    (1 << 6)
+#              define PACKET3_CB1_DEST_BASE_ENA    (1 << 7)
+#              define PACKET3_CB2_DEST_BASE_ENA    (1 << 8)
+#              define PACKET3_CB3_DEST_BASE_ENA    (1 << 9)
+#              define PACKET3_CB4_DEST_BASE_ENA    (1 << 10)
+#              define PACKET3_CB5_DEST_BASE_ENA    (1 << 11)
+#              define PACKET3_CB6_DEST_BASE_ENA    (1 << 12)
+#              define PACKET3_CB7_DEST_BASE_ENA    (1 << 13)
+#              define PACKET3_DB_DEST_BASE_ENA     (1 << 14)
+#              define PACKET3_CB8_DEST_BASE_ENA    (1 << 15)
+#              define PACKET3_CB9_DEST_BASE_ENA    (1 << 16)
+#              define PACKET3_CB10_DEST_BASE_ENA   (1 << 17)
+#              define PACKET3_CB11_DEST_BASE_ENA   (1 << 17)
+#              define PACKET3_FULL_CACHE_ENA       (1 << 20)
+#              define PACKET3_TC_ACTION_ENA        (1 << 23)
+#              define PACKET3_VC_ACTION_ENA        (1 << 24)
+#              define PACKET3_CB_ACTION_ENA        (1 << 25)
+#              define PACKET3_DB_ACTION_ENA        (1 << 26)
+#              define PACKET3_SH_ACTION_ENA        (1 << 27)
+#              define PACKET3_SMX_ACTION_ENA       (1 << 28)
+#define	PACKET3_ME_INITIALIZE				0x44
+#define		PACKET3_ME_INITIALIZE_DEVICE_ID(x) ((x) << 16)
+#define	PACKET3_COND_WRITE				0x45
+#define	PACKET3_EVENT_WRITE				0x46
+#define	PACKET3_EVENT_WRITE_EOP				0x47
+#define	PACKET3_EVENT_WRITE_EOS				0x48
+#define	PACKET3_PREAMBLE_CNTL				0x4A
+#define	PACKET3_RB_OFFSET				0x4B
+#define	PACKET3_ALU_PS_CONST_BUFFER_COPY		0x4C
+#define	PACKET3_ALU_VS_CONST_BUFFER_COPY		0x4D
+#define	PACKET3_ALU_PS_CONST_UPDATE		        0x4E
+#define	PACKET3_ALU_VS_CONST_UPDATE		        0x4F
+#define	PACKET3_ONE_REG_WRITE				0x57
+#define	PACKET3_SET_CONFIG_REG				0x68
+#define		PACKET3_SET_CONFIG_REG_START			0x00008000
+#define		PACKET3_SET_CONFIG_REG_END			0x0000ac00
+#define	PACKET3_SET_CONTEXT_REG				0x69
+#define		PACKET3_SET_CONTEXT_REG_START			0x00028000
+#define		PACKET3_SET_CONTEXT_REG_END			0x00029000
+#define	PACKET3_SET_ALU_CONST				0x6A
+/* alu const buffers only; no reg file */
+#define	PACKET3_SET_BOOL_CONST				0x6B
+#define		PACKET3_SET_BOOL_CONST_START			0x0003a500
+#define		PACKET3_SET_BOOL_CONST_END			0x0003a518
+#define	PACKET3_SET_LOOP_CONST				0x6C
+#define		PACKET3_SET_LOOP_CONST_START			0x0003a200
+#define		PACKET3_SET_LOOP_CONST_END			0x0003a500
+#define	PACKET3_SET_RESOURCE				0x6D
+#define		PACKET3_SET_RESOURCE_START			0x00030000
+#define		PACKET3_SET_RESOURCE_END			0x00038000
+#define	PACKET3_SET_SAMPLER				0x6E
+#define		PACKET3_SET_SAMPLER_START			0x0003c000
+#define		PACKET3_SET_SAMPLER_END				0x0003c600
+#define	PACKET3_SET_CTL_CONST				0x6F
+#define		PACKET3_SET_CTL_CONST_START			0x0003cff0
+#define		PACKET3_SET_CTL_CONST_END			0x0003ff0c
+#define	PACKET3_SET_RESOURCE_OFFSET			0x70
+#define	PACKET3_SET_ALU_CONST_VS			0x71
+#define	PACKET3_SET_ALU_CONST_DI			0x72
+#define	PACKET3_SET_CONTEXT_REG_INDIRECT		0x73
+#define	PACKET3_SET_RESOURCE_INDIRECT			0x74
+#define	PACKET3_SET_APPEND_CNT			        0x75
+
+#define	SQ_RESOURCE_CONSTANT_WORD7_0				0x3001c
+#define		S__SQ_CONSTANT_TYPE(x)			(((x) & 3) << 30)
+#define		G__SQ_CONSTANT_TYPE(x)			(((x) >> 30) & 3)
+#define			SQ_TEX_VTX_INVALID_TEXTURE			0x0
+#define			SQ_TEX_VTX_INVALID_BUFFER			0x1
+#define			SQ_TEX_VTX_VALID_TEXTURE			0x2
+#define			SQ_TEX_VTX_VALID_BUFFER				0x3
+
+#define SQ_CONST_MEM_BASE				0x8df8
+
+#define SQ_ESGS_RING_SIZE				0x8c44
+#define SQ_GSVS_RING_SIZE				0x8c4c
+#define SQ_ESTMP_RING_SIZE				0x8c54
+#define SQ_GSTMP_RING_SIZE				0x8c5c
+#define SQ_VSTMP_RING_SIZE				0x8c64
+#define SQ_PSTMP_RING_SIZE				0x8c6c
+#define SQ_LSTMP_RING_SIZE				0x8e14
+#define SQ_HSTMP_RING_SIZE				0x8e1c
+#define VGT_TF_RING_SIZE				0x8988
+
+#define SQ_ESGS_RING_ITEMSIZE				0x28900
+#define SQ_GSVS_RING_ITEMSIZE				0x28904
+#define SQ_ESTMP_RING_ITEMSIZE				0x28908
+#define SQ_GSTMP_RING_ITEMSIZE				0x2890c
+#define SQ_VSTMP_RING_ITEMSIZE				0x28910
+#define SQ_PSTMP_RING_ITEMSIZE				0x28914
+#define SQ_LSTMP_RING_ITEMSIZE				0x28830
+#define SQ_HSTMP_RING_ITEMSIZE				0x28834
+
+#define SQ_GS_VERT_ITEMSIZE				0x2891c
+#define SQ_GS_VERT_ITEMSIZE_1				0x28920
+#define SQ_GS_VERT_ITEMSIZE_2				0x28924
+#define SQ_GS_VERT_ITEMSIZE_3				0x28928
+#define SQ_GSVS_RING_OFFSET_1				0x2892c
+#define SQ_GSVS_RING_OFFSET_2				0x28930
+#define SQ_GSVS_RING_OFFSET_3				0x28934
+
+#define SQ_ALU_CONST_CACHE_PS_0				0x28940
+#define SQ_ALU_CONST_CACHE_PS_1				0x28944
+#define SQ_ALU_CONST_CACHE_PS_2				0x28948
+#define SQ_ALU_CONST_CACHE_PS_3				0x2894c
+#define SQ_ALU_CONST_CACHE_PS_4				0x28950
+#define SQ_ALU_CONST_CACHE_PS_5				0x28954
+#define SQ_ALU_CONST_CACHE_PS_6				0x28958
+#define SQ_ALU_CONST_CACHE_PS_7				0x2895c
+#define SQ_ALU_CONST_CACHE_PS_8				0x28960
+#define SQ_ALU_CONST_CACHE_PS_9				0x28964
+#define SQ_ALU_CONST_CACHE_PS_10			0x28968
+#define SQ_ALU_CONST_CACHE_PS_11			0x2896c
+#define SQ_ALU_CONST_CACHE_PS_12			0x28970
+#define SQ_ALU_CONST_CACHE_PS_13			0x28974
+#define SQ_ALU_CONST_CACHE_PS_14			0x28978
+#define SQ_ALU_CONST_CACHE_PS_15			0x2897c
+#define SQ_ALU_CONST_CACHE_VS_0				0x28980
+#define SQ_ALU_CONST_CACHE_VS_1				0x28984
+#define SQ_ALU_CONST_CACHE_VS_2				0x28988
+#define SQ_ALU_CONST_CACHE_VS_3				0x2898c
+#define SQ_ALU_CONST_CACHE_VS_4				0x28990
+#define SQ_ALU_CONST_CACHE_VS_5				0x28994
+#define SQ_ALU_CONST_CACHE_VS_6				0x28998
+#define SQ_ALU_CONST_CACHE_VS_7				0x2899c
+#define SQ_ALU_CONST_CACHE_VS_8				0x289a0
+#define SQ_ALU_CONST_CACHE_VS_9				0x289a4
+#define SQ_ALU_CONST_CACHE_VS_10			0x289a8
+#define SQ_ALU_CONST_CACHE_VS_11			0x289ac
+#define SQ_ALU_CONST_CACHE_VS_12			0x289b0
+#define SQ_ALU_CONST_CACHE_VS_13			0x289b4
+#define SQ_ALU_CONST_CACHE_VS_14			0x289b8
+#define SQ_ALU_CONST_CACHE_VS_15			0x289bc
+#define SQ_ALU_CONST_CACHE_GS_0				0x289c0
+#define SQ_ALU_CONST_CACHE_GS_1				0x289c4
+#define SQ_ALU_CONST_CACHE_GS_2				0x289c8
+#define SQ_ALU_CONST_CACHE_GS_3				0x289cc
+#define SQ_ALU_CONST_CACHE_GS_4				0x289d0
+#define SQ_ALU_CONST_CACHE_GS_5				0x289d4
+#define SQ_ALU_CONST_CACHE_GS_6				0x289d8
+#define SQ_ALU_CONST_CACHE_GS_7				0x289dc
+#define SQ_ALU_CONST_CACHE_GS_8				0x289e0
+#define SQ_ALU_CONST_CACHE_GS_9				0x289e4
+#define SQ_ALU_CONST_CACHE_GS_10			0x289e8
+#define SQ_ALU_CONST_CACHE_GS_11			0x289ec
+#define SQ_ALU_CONST_CACHE_GS_12			0x289f0
+#define SQ_ALU_CONST_CACHE_GS_13			0x289f4
+#define SQ_ALU_CONST_CACHE_GS_14			0x289f8
+#define SQ_ALU_CONST_CACHE_GS_15			0x289fc
+#define SQ_ALU_CONST_CACHE_HS_0				0x28f00
+#define SQ_ALU_CONST_CACHE_HS_1				0x28f04
+#define SQ_ALU_CONST_CACHE_HS_2				0x28f08
+#define SQ_ALU_CONST_CACHE_HS_3				0x28f0c
+#define SQ_ALU_CONST_CACHE_HS_4				0x28f10
+#define SQ_ALU_CONST_CACHE_HS_5				0x28f14
+#define SQ_ALU_CONST_CACHE_HS_6				0x28f18
+#define SQ_ALU_CONST_CACHE_HS_7				0x28f1c
+#define SQ_ALU_CONST_CACHE_HS_8				0x28f20
+#define SQ_ALU_CONST_CACHE_HS_9				0x28f24
+#define SQ_ALU_CONST_CACHE_HS_10			0x28f28
+#define SQ_ALU_CONST_CACHE_HS_11			0x28f2c
+#define SQ_ALU_CONST_CACHE_HS_12			0x28f30
+#define SQ_ALU_CONST_CACHE_HS_13			0x28f34
+#define SQ_ALU_CONST_CACHE_HS_14			0x28f38
+#define SQ_ALU_CONST_CACHE_HS_15			0x28f3c
+#define SQ_ALU_CONST_CACHE_LS_0				0x28f40
+#define SQ_ALU_CONST_CACHE_LS_1				0x28f44
+#define SQ_ALU_CONST_CACHE_LS_2				0x28f48
+#define SQ_ALU_CONST_CACHE_LS_3				0x28f4c
+#define SQ_ALU_CONST_CACHE_LS_4				0x28f50
+#define SQ_ALU_CONST_CACHE_LS_5				0x28f54
+#define SQ_ALU_CONST_CACHE_LS_6				0x28f58
+#define SQ_ALU_CONST_CACHE_LS_7				0x28f5c
+#define SQ_ALU_CONST_CACHE_LS_8				0x28f60
+#define SQ_ALU_CONST_CACHE_LS_9				0x28f64
+#define SQ_ALU_CONST_CACHE_LS_10			0x28f68
+#define SQ_ALU_CONST_CACHE_LS_11			0x28f6c
+#define SQ_ALU_CONST_CACHE_LS_12			0x28f70
+#define SQ_ALU_CONST_CACHE_LS_13			0x28f74
+#define SQ_ALU_CONST_CACHE_LS_14			0x28f78
+#define SQ_ALU_CONST_CACHE_LS_15			0x28f7c
+
+#define DB_DEPTH_CONTROL				0x28800
+#define DB_DEPTH_VIEW					0x28008
+#define DB_HTILE_DATA_BASE				0x28014
+#define DB_Z_INFO					0x28040
+#       define Z_ARRAY_MODE(x)                          ((x) << 4)
+#define DB_STENCIL_INFO					0x28044
+#define DB_Z_READ_BASE					0x28048
+#define DB_STENCIL_READ_BASE				0x2804c
+#define DB_Z_WRITE_BASE					0x28050
+#define DB_STENCIL_WRITE_BASE				0x28054
+#define DB_DEPTH_SIZE					0x28058
+
+#define SQ_PGM_START_PS					0x28840
+#define SQ_PGM_START_VS					0x2885c
+#define SQ_PGM_START_GS					0x28874
+#define SQ_PGM_START_ES					0x2888c
+#define SQ_PGM_START_FS					0x288a4
+#define SQ_PGM_START_HS					0x288b8
+#define SQ_PGM_START_LS					0x288d0
+
+#define VGT_STRMOUT_CONFIG				0x28b94
+#define VGT_STRMOUT_BUFFER_CONFIG			0x28b98
+
+#define CB_TARGET_MASK					0x28238
+#define CB_SHADER_MASK					0x2823c
+
+#define GDS_ADDR_BASE					0x28720
+
+#define	CB_IMMED0_BASE					0x28b9c
+#define	CB_IMMED1_BASE					0x28ba0
+#define	CB_IMMED2_BASE					0x28ba4
+#define	CB_IMMED3_BASE					0x28ba8
+#define	CB_IMMED4_BASE					0x28bac
+#define	CB_IMMED5_BASE					0x28bb0
+#define	CB_IMMED6_BASE					0x28bb4
+#define	CB_IMMED7_BASE					0x28bb8
+#define	CB_IMMED8_BASE					0x28bbc
+#define	CB_IMMED9_BASE					0x28bc0
+#define	CB_IMMED10_BASE					0x28bc4
+#define	CB_IMMED11_BASE					0x28bc8
+
+/* all 12 CB blocks have these regs */
+#define	CB_COLOR0_BASE					0x28c60
+#define	CB_COLOR0_PITCH					0x28c64
+#define	CB_COLOR0_SLICE					0x28c68
+#define	CB_COLOR0_VIEW					0x28c6c
+#define	CB_COLOR0_INFO					0x28c70
+#       define CB_ARRAY_MODE(x)                         ((x) << 8)
+#       define ARRAY_LINEAR_GENERAL                     0
+#       define ARRAY_LINEAR_ALIGNED                     1
+#       define ARRAY_1D_TILED_THIN1                     2
+#       define ARRAY_2D_TILED_THIN1                     4
+#define	CB_COLOR0_ATTRIB				0x28c74
+#define	CB_COLOR0_DIM					0x28c78
+/* only CB0-7 blocks have these regs */
+#define	CB_COLOR0_CMASK					0x28c7c
+#define	CB_COLOR0_CMASK_SLICE				0x28c80
+#define	CB_COLOR0_FMASK					0x28c84
+#define	CB_COLOR0_FMASK_SLICE				0x28c88
+#define	CB_COLOR0_CLEAR_WORD0				0x28c8c
+#define	CB_COLOR0_CLEAR_WORD1				0x28c90
+#define	CB_COLOR0_CLEAR_WORD2				0x28c94
+#define	CB_COLOR0_CLEAR_WORD3				0x28c98
+
+#define	CB_COLOR1_BASE					0x28c9c
+#define	CB_COLOR2_BASE					0x28cd8
+#define	CB_COLOR3_BASE					0x28d14
+#define	CB_COLOR4_BASE					0x28d50
+#define	CB_COLOR5_BASE					0x28d8c
+#define	CB_COLOR6_BASE					0x28dc8
+#define	CB_COLOR7_BASE					0x28e04
+#define	CB_COLOR8_BASE					0x28e40
+#define	CB_COLOR9_BASE					0x28e5c
+#define	CB_COLOR10_BASE					0x28e78
+#define	CB_COLOR11_BASE					0x28e94
+
+#define	CB_COLOR1_PITCH					0x28ca0
+#define	CB_COLOR2_PITCH					0x28cdc
+#define	CB_COLOR3_PITCH					0x28d18
+#define	CB_COLOR4_PITCH					0x28d54
+#define	CB_COLOR5_PITCH					0x28d90
+#define	CB_COLOR6_PITCH					0x28dcc
+#define	CB_COLOR7_PITCH					0x28e08
+#define	CB_COLOR8_PITCH					0x28e44
+#define	CB_COLOR9_PITCH					0x28e60
+#define	CB_COLOR10_PITCH				0x28e7c
+#define	CB_COLOR11_PITCH				0x28e98
+
+#define	CB_COLOR1_SLICE					0x28ca4
+#define	CB_COLOR2_SLICE					0x28ce0
+#define	CB_COLOR3_SLICE					0x28d1c
+#define	CB_COLOR4_SLICE					0x28d58
+#define	CB_COLOR5_SLICE					0x28d94
+#define	CB_COLOR6_SLICE					0x28dd0
+#define	CB_COLOR7_SLICE					0x28e0c
+#define	CB_COLOR8_SLICE					0x28e48
+#define	CB_COLOR9_SLICE					0x28e64
+#define	CB_COLOR10_SLICE				0x28e80
+#define	CB_COLOR11_SLICE				0x28e9c
+
+#define	CB_COLOR1_VIEW					0x28ca8
+#define	CB_COLOR2_VIEW					0x28ce4
+#define	CB_COLOR3_VIEW					0x28d20
+#define	CB_COLOR4_VIEW					0x28d5c
+#define	CB_COLOR5_VIEW					0x28d98
+#define	CB_COLOR6_VIEW					0x28dd4
+#define	CB_COLOR7_VIEW					0x28e10
+#define	CB_COLOR8_VIEW					0x28e4c
+#define	CB_COLOR9_VIEW					0x28e68
+#define	CB_COLOR10_VIEW					0x28e84
+#define	CB_COLOR11_VIEW					0x28ea0
+
+#define	CB_COLOR1_INFO					0x28cac
+#define	CB_COLOR2_INFO					0x28ce8
+#define	CB_COLOR3_INFO					0x28d24
+#define	CB_COLOR4_INFO					0x28d60
+#define	CB_COLOR5_INFO					0x28d9c
+#define	CB_COLOR6_INFO					0x28dd8
+#define	CB_COLOR7_INFO					0x28e14
+#define	CB_COLOR8_INFO					0x28e50
+#define	CB_COLOR9_INFO					0x28e6c
+#define	CB_COLOR10_INFO					0x28e88
+#define	CB_COLOR11_INFO					0x28ea4
+
+#define	CB_COLOR1_ATTRIB				0x28cb0
+#define	CB_COLOR2_ATTRIB				0x28cec
+#define	CB_COLOR3_ATTRIB				0x28d28
+#define	CB_COLOR4_ATTRIB				0x28d64
+#define	CB_COLOR5_ATTRIB				0x28da0
+#define	CB_COLOR6_ATTRIB				0x28ddc
+#define	CB_COLOR7_ATTRIB				0x28e18
+#define	CB_COLOR8_ATTRIB				0x28e54
+#define	CB_COLOR9_ATTRIB				0x28e70
+#define	CB_COLOR10_ATTRIB				0x28e8c
+#define	CB_COLOR11_ATTRIB				0x28ea8
+
+#define	CB_COLOR1_DIM					0x28cb4
+#define	CB_COLOR2_DIM					0x28cf0
+#define	CB_COLOR3_DIM					0x28d2c
+#define	CB_COLOR4_DIM					0x28d68
+#define	CB_COLOR5_DIM					0x28da4
+#define	CB_COLOR6_DIM					0x28de0
+#define	CB_COLOR7_DIM					0x28e1c
+#define	CB_COLOR8_DIM					0x28e58
+#define	CB_COLOR9_DIM					0x28e74
+#define	CB_COLOR10_DIM					0x28e90
+#define	CB_COLOR11_DIM					0x28eac
+
+#define	CB_COLOR1_CMASK					0x28cb8
+#define	CB_COLOR2_CMASK					0x28cf4
+#define	CB_COLOR3_CMASK					0x28d30
+#define	CB_COLOR4_CMASK					0x28d6c
+#define	CB_COLOR5_CMASK					0x28da8
+#define	CB_COLOR6_CMASK					0x28de4
+#define	CB_COLOR7_CMASK					0x28e20
+
+#define	CB_COLOR1_CMASK_SLICE				0x28cbc
+#define	CB_COLOR2_CMASK_SLICE				0x28cf8
+#define	CB_COLOR3_CMASK_SLICE				0x28d34
+#define	CB_COLOR4_CMASK_SLICE				0x28d70
+#define	CB_COLOR5_CMASK_SLICE				0x28dac
+#define	CB_COLOR6_CMASK_SLICE				0x28de8
+#define	CB_COLOR7_CMASK_SLICE				0x28e24
+
+#define	CB_COLOR1_FMASK					0x28cc0
+#define	CB_COLOR2_FMASK					0x28cfc
+#define	CB_COLOR3_FMASK					0x28d38
+#define	CB_COLOR4_FMASK					0x28d74
+#define	CB_COLOR5_FMASK					0x28db0
+#define	CB_COLOR6_FMASK					0x28dec
+#define	CB_COLOR7_FMASK					0x28e28
+
+#define	CB_COLOR1_FMASK_SLICE				0x28cc4
+#define	CB_COLOR2_FMASK_SLICE				0x28d00
+#define	CB_COLOR3_FMASK_SLICE				0x28d3c
+#define	CB_COLOR4_FMASK_SLICE				0x28d78
+#define	CB_COLOR5_FMASK_SLICE				0x28db4
+#define	CB_COLOR6_FMASK_SLICE				0x28df0
+#define	CB_COLOR7_FMASK_SLICE				0x28e2c
+
+#define	CB_COLOR1_CLEAR_WORD0				0x28cc8
+#define	CB_COLOR2_CLEAR_WORD0				0x28d04
+#define	CB_COLOR3_CLEAR_WORD0				0x28d40
+#define	CB_COLOR4_CLEAR_WORD0				0x28d7c
+#define	CB_COLOR5_CLEAR_WORD0				0x28db8
+#define	CB_COLOR6_CLEAR_WORD0				0x28df4
+#define	CB_COLOR7_CLEAR_WORD0				0x28e30
+
+#define	CB_COLOR1_CLEAR_WORD1				0x28ccc
+#define	CB_COLOR2_CLEAR_WORD1				0x28d08
+#define	CB_COLOR3_CLEAR_WORD1				0x28d44
+#define	CB_COLOR4_CLEAR_WORD1				0x28d80
+#define	CB_COLOR5_CLEAR_WORD1				0x28dbc
+#define	CB_COLOR6_CLEAR_WORD1				0x28df8
+#define	CB_COLOR7_CLEAR_WORD1				0x28e34
+
+#define	CB_COLOR1_CLEAR_WORD2				0x28cd0
+#define	CB_COLOR2_CLEAR_WORD2				0x28d0c
+#define	CB_COLOR3_CLEAR_WORD2				0x28d48
+#define	CB_COLOR4_CLEAR_WORD2				0x28d84
+#define	CB_COLOR5_CLEAR_WORD2				0x28dc0
+#define	CB_COLOR6_CLEAR_WORD2				0x28dfc
+#define	CB_COLOR7_CLEAR_WORD2				0x28e38
+
+#define	CB_COLOR1_CLEAR_WORD3				0x28cd4
+#define	CB_COLOR2_CLEAR_WORD3				0x28d10
+#define	CB_COLOR3_CLEAR_WORD3				0x28d4c
+#define	CB_COLOR4_CLEAR_WORD3				0x28d88
+#define	CB_COLOR5_CLEAR_WORD3				0x28dc4
+#define	CB_COLOR6_CLEAR_WORD3				0x28e00
+#define	CB_COLOR7_CLEAR_WORD3				0x28e3c
+
+#define SQ_TEX_RESOURCE_WORD0_0                         0x30000
+#define SQ_TEX_RESOURCE_WORD1_0                         0x30004
+#       define TEX_ARRAY_MODE(x)                        ((x) << 28)
+#define SQ_TEX_RESOURCE_WORD2_0                         0x30008
+#define SQ_TEX_RESOURCE_WORD3_0                         0x3000C
+#define SQ_TEX_RESOURCE_WORD4_0                         0x30010
+#define SQ_TEX_RESOURCE_WORD5_0                         0x30014
+#define SQ_TEX_RESOURCE_WORD6_0                         0x30018
+#define SQ_TEX_RESOURCE_WORD7_0                         0x3001c
+
+
 #endif
diff --git a/drivers/gpu/drm/radeon/r100.c b/drivers/gpu/drm/radeon/r100.c
index cc004b05d63e..cf89aa2eb28c 100644
--- a/drivers/gpu/drm/radeon/r100.c
+++ b/drivers/gpu/drm/radeon/r100.c
@@ -162,6 +162,11 @@ void r100_pm_init_profile(struct radeon_device *rdev)
 	rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = 0;
 	rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0;
 	rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0;
+	/* mid sh */
+	rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx = 0;
+	rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx = 0;
+	rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0;
+	rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 0;
 	/* high sh */
 	rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_ps_idx = 0;
 	rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
@@ -172,6 +177,11 @@ void r100_pm_init_profile(struct radeon_device *rdev)
 	rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
 	rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0;
 	rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0;
+	/* mid mh */
+	rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx = 0;
+	rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
+	rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0;
+	rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 0;
 	/* high mh */
 	rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_ps_idx = 0;
 	rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
diff --git a/drivers/gpu/drm/radeon/r420.c b/drivers/gpu/drm/radeon/r420.c
index 4415a5ee5871..e6c89142bb4d 100644
--- a/drivers/gpu/drm/radeon/r420.c
+++ b/drivers/gpu/drm/radeon/r420.c
@@ -45,9 +45,14 @@ void r420_pm_init_profile(struct radeon_device *rdev)
 	rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_cm_idx = 0;
 	/* low sh */
 	rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_ps_idx = 0;
-	rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = 1;
+	rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = 0;
 	rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0;
 	rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0;
+	/* mid sh */
+	rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx = 0;
+	rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx = 1;
+	rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0;
+	rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 0;
 	/* high sh */
 	rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_ps_idx = 0;
 	rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
@@ -58,6 +63,11 @@ void r420_pm_init_profile(struct radeon_device *rdev)
 	rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
 	rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0;
 	rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0;
+	/* mid mh */
+	rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx = 0;
+	rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
+	rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0;
+	rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 0;
 	/* high mh */
 	rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_ps_idx = 0;
 	rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c
index 44e96a2ae25a..0e91871f45be 100644
--- a/drivers/gpu/drm/radeon/r600.c
+++ b/drivers/gpu/drm/radeon/r600.c
@@ -291,6 +291,11 @@ void rs780_pm_init_profile(struct radeon_device *rdev)
 		rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = 0;
 		rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0;
 		rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0;
+		/* mid sh */
+		rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx = 0;
+		rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx = 0;
+		rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0;
+		rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 0;
 		/* high sh */
 		rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_ps_idx = 0;
 		rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_ps_idx = 1;
@@ -301,6 +306,11 @@ void rs780_pm_init_profile(struct radeon_device *rdev)
 		rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx = 0;
 		rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0;
 		rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0;
+		/* mid mh */
+		rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx = 0;
+		rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx = 0;
+		rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0;
+		rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 0;
 		/* high mh */
 		rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_ps_idx = 0;
 		rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_ps_idx = 1;
@@ -317,6 +327,11 @@ void rs780_pm_init_profile(struct radeon_device *rdev)
 		rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = 1;
 		rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0;
 		rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0;
+		/* mid sh */
+		rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx = 1;
+		rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx = 1;
+		rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0;
+		rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 0;
 		/* high sh */
 		rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_ps_idx = 1;
 		rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_ps_idx = 2;
@@ -327,6 +342,11 @@ void rs780_pm_init_profile(struct radeon_device *rdev)
 		rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx = 1;
 		rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0;
 		rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0;
+		/* mid mh */
+		rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx = 1;
+		rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx = 1;
+		rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0;
+		rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 0;
 		/* high mh */
 		rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_ps_idx = 1;
 		rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_ps_idx = 2;
@@ -343,6 +363,11 @@ void rs780_pm_init_profile(struct radeon_device *rdev)
 		rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = 2;
 		rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0;
 		rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0;
+		/* mid sh */
+		rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx = 2;
+		rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx = 2;
+		rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0;
+		rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 0;
 		/* high sh */
 		rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_ps_idx = 2;
 		rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_ps_idx = 3;
@@ -353,6 +378,11 @@ void rs780_pm_init_profile(struct radeon_device *rdev)
 		rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx = 0;
 		rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0;
 		rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0;
+		/* mid mh */
+		rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx = 2;
+		rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx = 0;
+		rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0;
+		rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 0;
 		/* high mh */
 		rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_ps_idx = 2;
 		rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_ps_idx = 3;
@@ -375,6 +405,11 @@ void r600_pm_init_profile(struct radeon_device *rdev)
 		rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
 		rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0;
 		rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0;
+		/* mid sh */
+		rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index;
+		rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
+		rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0;
+		rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 0;
 		/* high sh */
 		rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index;
 		rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
@@ -385,6 +420,11 @@ void r600_pm_init_profile(struct radeon_device *rdev)
 		rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
 		rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0;
 		rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0;
+		/* mid mh */
+		rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index;
+		rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
+		rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0;
+		rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 0;
 		/* high mh */
 		rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index;
 		rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
@@ -401,7 +441,12 @@ void r600_pm_init_profile(struct radeon_device *rdev)
 			rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_ps_idx = 1;
 			rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = 1;
 			rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0;
-			rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 1;
+			rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0;
+			/* mid sh */
+			rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx = 1;
+			rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx = 1;
+			rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0;
+			rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 1;
 			/* high sh */
 			rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_ps_idx = 1;
 			rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_ps_idx = 1;
@@ -411,7 +456,12 @@ void r600_pm_init_profile(struct radeon_device *rdev)
 			rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_ps_idx = 2;
 			rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx = 2;
 			rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0;
-			rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 1;
+			rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0;
+			/* low mh */
+			rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx = 2;
+			rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx = 2;
+			rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0;
+			rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 1;
 			/* high mh */
 			rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_ps_idx = 2;
 			rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_ps_idx = 2;
@@ -430,14 +480,30 @@ void r600_pm_init_profile(struct radeon_device *rdev)
 				rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx =
 					r600_pm_get_type_index(rdev, POWER_STATE_TYPE_BATTERY, 0);
 				rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0;
-				rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 1;
+				rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0;
 			} else {
 				rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_ps_idx =
 					r600_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 0);
 				rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx =
 					r600_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 0);
 				rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0;
-				rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 1;
+				rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0;
+			}
+			/* mid sh */
+			if (rdev->flags & RADEON_IS_MOBILITY) {
+				rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx =
+					r600_pm_get_type_index(rdev, POWER_STATE_TYPE_BATTERY, 0);
+				rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx =
+					r600_pm_get_type_index(rdev, POWER_STATE_TYPE_BATTERY, 0);
+				rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0;
+				rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 1;
+			} else {
+				rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx =
+					r600_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 0);
+				rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx =
+					r600_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 0);
+				rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0;
+				rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 1;
 			}
 			/* high sh */
 			rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_ps_idx =
@@ -453,14 +519,30 @@ void r600_pm_init_profile(struct radeon_device *rdev)
 				rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx =
 					r600_pm_get_type_index(rdev, POWER_STATE_TYPE_BATTERY, 1);
 				rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0;
-				rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 2;
+				rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0;
 			} else {
 				rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_ps_idx =
 					r600_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 1);
 				rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx =
 					r600_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 1);
 				rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0;
-				rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 1;
+				rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0;
+			}
+			/* mid mh */
+			if (rdev->flags & RADEON_IS_MOBILITY) {
+				rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx =
+					r600_pm_get_type_index(rdev, POWER_STATE_TYPE_BATTERY, 1);
+				rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx =
+					r600_pm_get_type_index(rdev, POWER_STATE_TYPE_BATTERY, 1);
+				rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0;
+				rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 1;
+			} else {
+				rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx =
+					r600_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 1);
+				rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx =
+					r600_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 1);
+				rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0;
+				rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 1;
 			}
 			/* high mh */
 			rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_ps_idx =
@@ -475,7 +557,18 @@ void r600_pm_init_profile(struct radeon_device *rdev)
 
 void r600_pm_misc(struct radeon_device *rdev)
 {
+	int req_ps_idx = rdev->pm.requested_power_state_index;
+	int req_cm_idx = rdev->pm.requested_clock_mode_index;
+	struct radeon_power_state *ps = &rdev->pm.power_state[req_ps_idx];
+	struct radeon_voltage *voltage = &ps->clock_info[req_cm_idx].voltage;
 
+	if ((voltage->type == VOLTAGE_SW) && voltage->voltage) {
+		if (voltage->voltage != rdev->pm.current_vddc) {
+			radeon_atom_set_voltage(rdev, voltage->voltage);
+			rdev->pm.current_vddc = voltage->voltage;
+			DRM_DEBUG("Setting: v: %d\n", voltage->voltage);
+		}
+	}
 }
 
 bool r600_gui_idle(struct radeon_device *rdev)
diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h
index 66a37fb75839..8e1d44ca26ec 100644
--- a/drivers/gpu/drm/radeon/radeon.h
+++ b/drivers/gpu/drm/radeon/radeon.h
@@ -176,6 +176,7 @@ void radeon_pm_suspend(struct radeon_device *rdev);
 void radeon_pm_resume(struct radeon_device *rdev);
 void radeon_combios_get_power_modes(struct radeon_device *rdev);
 void radeon_atombios_get_power_modes(struct radeon_device *rdev);
+void radeon_atom_set_voltage(struct radeon_device *rdev, u16 level);
 
 /*
  * Fences.
@@ -576,6 +577,7 @@ typedef int (*radeon_packet3_check_t)(struct radeon_cs_parser *p,
  */
 int radeon_agp_init(struct radeon_device *rdev);
 void radeon_agp_resume(struct radeon_device *rdev);
+void radeon_agp_suspend(struct radeon_device *rdev);
 void radeon_agp_fini(struct radeon_device *rdev);
 
 
@@ -646,15 +648,18 @@ enum radeon_pm_profile_type {
 	PM_PROFILE_DEFAULT,
 	PM_PROFILE_AUTO,
 	PM_PROFILE_LOW,
+	PM_PROFILE_MID,
 	PM_PROFILE_HIGH,
 };
 
 #define PM_PROFILE_DEFAULT_IDX 0
 #define PM_PROFILE_LOW_SH_IDX  1
-#define PM_PROFILE_HIGH_SH_IDX 2
-#define PM_PROFILE_LOW_MH_IDX  3
-#define PM_PROFILE_HIGH_MH_IDX 4
-#define PM_PROFILE_MAX         5
+#define PM_PROFILE_MID_SH_IDX  2
+#define PM_PROFILE_HIGH_SH_IDX 3
+#define PM_PROFILE_LOW_MH_IDX  4
+#define PM_PROFILE_MID_MH_IDX  5
+#define PM_PROFILE_HIGH_MH_IDX 6
+#define PM_PROFILE_MAX         7
 
 struct radeon_pm_profile {
 	int dpms_off_ps_idx;
@@ -743,6 +748,7 @@ struct radeon_pm {
 	int                     default_power_state_index;
 	u32                     current_sclk;
 	u32                     current_mclk;
+	u32                     current_vddc;
 	struct radeon_i2c_chan *i2c_bus;
 	/* selected pm method */
 	enum radeon_pm_method     pm_method;
diff --git a/drivers/gpu/drm/radeon/radeon_agp.c b/drivers/gpu/drm/radeon/radeon_agp.c
index 28e473f1f56f..f40dfb77f9b1 100644
--- a/drivers/gpu/drm/radeon/radeon_agp.c
+++ b/drivers/gpu/drm/radeon/radeon_agp.c
@@ -270,3 +270,8 @@ void radeon_agp_fini(struct radeon_device *rdev)
 	}
 #endif
 }
+
+void radeon_agp_suspend(struct radeon_device *rdev)
+{
+	radeon_agp_fini(rdev);
+}
diff --git a/drivers/gpu/drm/radeon/radeon_asic.c b/drivers/gpu/drm/radeon/radeon_asic.c
index e57df08d4aeb..87f7e2cc52d4 100644
--- a/drivers/gpu/drm/radeon/radeon_asic.c
+++ b/drivers/gpu/drm/radeon/radeon_asic.c
@@ -724,8 +724,8 @@ static struct radeon_asic evergreen_asic = {
 	.irq_set = &evergreen_irq_set,
 	.irq_process = &evergreen_irq_process,
 	.get_vblank_counter = &evergreen_get_vblank_counter,
-	.fence_ring_emit = NULL,
-	.cs_parse = NULL,
+	.fence_ring_emit = &r600_fence_ring_emit,
+	.cs_parse = &evergreen_cs_parse,
 	.copy_blit = NULL,
 	.copy_dma = NULL,
 	.copy = NULL,
diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeon/radeon_asic.h
index 5c40a3dfaca2..c0bbaa64157a 100644
--- a/drivers/gpu/drm/radeon/radeon_asic.h
+++ b/drivers/gpu/drm/radeon/radeon_asic.h
@@ -314,6 +314,7 @@ void evergreen_hpd_set_polarity(struct radeon_device *rdev,
 u32 evergreen_get_vblank_counter(struct radeon_device *rdev, int crtc);
 int evergreen_irq_set(struct radeon_device *rdev);
 int evergreen_irq_process(struct radeon_device *rdev);
+extern int evergreen_cs_parse(struct radeon_cs_parser *p);
 extern void evergreen_pm_misc(struct radeon_device *rdev);
 extern void evergreen_pm_prepare(struct radeon_device *rdev);
 extern void evergreen_pm_finish(struct radeon_device *rdev);
diff --git a/drivers/gpu/drm/radeon/radeon_atombios.c b/drivers/gpu/drm/radeon/radeon_atombios.c
index 6e733fdc3349..99bd8a9c56b3 100644
--- a/drivers/gpu/drm/radeon/radeon_atombios.c
+++ b/drivers/gpu/drm/radeon/radeon_atombios.c
@@ -680,11 +680,19 @@ bool radeon_get_atom_connector_info_from_supported_devices_table(struct
 	uint8_t dac;
 	union atom_supported_devices *supported_devices;
 	int i, j, max_device;
-	struct bios_connector bios_connectors[ATOM_MAX_SUPPORTED_DEVICE];
+	struct bios_connector *bios_connectors;
+	size_t bc_size = sizeof(*bios_connectors) * ATOM_MAX_SUPPORTED_DEVICE;
 
-	if (!atom_parse_data_header(ctx, index, &size, &frev, &crev, &data_offset))
+	bios_connectors = kzalloc(bc_size, GFP_KERNEL);
+	if (!bios_connectors)
 		return false;
 
+	if (!atom_parse_data_header(ctx, index, &size, &frev, &crev,
+				    &data_offset)) {
+		kfree(bios_connectors);
+		return false;
+	}
+
 	supported_devices =
 	    (union atom_supported_devices *)(ctx->bios + data_offset);
 
@@ -851,6 +859,7 @@ bool radeon_get_atom_connector_info_from_supported_devices_table(struct
 
 	radeon_link_encoder_connector(dev);
 
+	kfree(bios_connectors);
 	return true;
 }
 
@@ -1529,7 +1538,8 @@ void radeon_atombios_get_power_modes(struct radeon_device *rdev)
 					rdev->pm.power_state[state_index].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) {
+					if ((misc & ATOM_PM_MISCINFO_VOLTAGE_DROP_SUPPORT) ||
+					    (misc & ATOM_PM_MISCINFO_VOLTAGE_DROP_ACTIVE_HIGH)) {
 						rdev->pm.power_state[state_index].clock_info[0].voltage.type =
 							VOLTAGE_GPIO;
 						rdev->pm.power_state[state_index].clock_info[0].voltage.gpio =
@@ -1596,7 +1606,8 @@ void radeon_atombios_get_power_modes(struct radeon_device *rdev)
 						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) {
+					if ((misc & ATOM_PM_MISCINFO_VOLTAGE_DROP_SUPPORT) ||
+					    (misc & ATOM_PM_MISCINFO_VOLTAGE_DROP_ACTIVE_HIGH)) {
 						rdev->pm.power_state[state_index].clock_info[0].voltage.type =
 							VOLTAGE_GPIO;
 						rdev->pm.power_state[state_index].clock_info[0].voltage.gpio =
@@ -1670,7 +1681,8 @@ void radeon_atombios_get_power_modes(struct radeon_device *rdev)
 						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) {
+					if ((misc & ATOM_PM_MISCINFO_VOLTAGE_DROP_SUPPORT) ||
+					    (misc & ATOM_PM_MISCINFO_VOLTAGE_DROP_ACTIVE_HIGH)) {
 						rdev->pm.power_state[state_index].clock_info[0].voltage.type =
 							VOLTAGE_GPIO;
 						rdev->pm.power_state[state_index].clock_info[0].voltage.gpio =
@@ -1746,9 +1758,22 @@ void radeon_atombios_get_power_modes(struct radeon_device *rdev)
 				rdev->pm.power_state[state_index].misc2 = 0;
 			}
 		} else {
+			int fw_index = GetIndexIntoMasterTable(DATA, FirmwareInfo);
+			uint8_t fw_frev, fw_crev;
+			uint16_t fw_data_offset, vddc = 0;
+			union firmware_info *firmware_info;
+			ATOM_PPLIB_THERMALCONTROLLER *controller = &power_info->info_4.sThermalController;
+
+			if (atom_parse_data_header(mode_info->atom_context, fw_index, NULL,
+						   &fw_frev, &fw_crev, &fw_data_offset)) {
+				firmware_info =
+					(union firmware_info *)(mode_info->atom_context->bios +
+								fw_data_offset);
+				vddc = firmware_info->info_14.usBootUpVDDCVoltage;
+			}
+
 			/* add the i2c bus for thermal/fan chip */
 			/* no support for internal controller yet */
-			ATOM_PPLIB_THERMALCONTROLLER *controller = &power_info->info_4.sThermalController;
 			if (controller->ucType > 0) {
 				if ((controller->ucType == ATOM_PP_THERMALCONTROLLER_RV6xx) ||
 				    (controller->ucType == ATOM_PP_THERMALCONTROLLER_RV770) ||
@@ -1808,10 +1833,7 @@ void radeon_atombios_get_power_modes(struct radeon_device *rdev)
 						/* skip invalid modes */
 						if (rdev->pm.power_state[state_index].clock_info[mode_index].sclk == 0)
 							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;
+						/* voltage works differently on IGPs */
 						mode_index++;
 					} else if (ASIC_IS_DCE4(rdev)) {
 						struct _ATOM_PPLIB_EVERGREEN_CLOCK_INFO *clock_info =
@@ -1895,6 +1917,16 @@ void radeon_atombios_get_power_modes(struct radeon_device *rdev)
 						rdev->pm.default_power_state_index = state_index;
 						rdev->pm.power_state[state_index].default_clock_mode =
 							&rdev->pm.power_state[state_index].clock_info[mode_index - 1];
+						/* patch the table values with the default slck/mclk from firmware info */
+						for (j = 0; j < mode_index; j++) {
+							rdev->pm.power_state[state_index].clock_info[j].mclk =
+								rdev->clock.default_mclk;
+							rdev->pm.power_state[state_index].clock_info[j].sclk =
+								rdev->clock.default_sclk;
+							if (vddc)
+								rdev->pm.power_state[state_index].clock_info[j].voltage.voltage =
+									vddc;
+						}
 					}
 					state_index++;
 				}
@@ -1934,6 +1966,7 @@ void radeon_atombios_get_power_modes(struct radeon_device *rdev)
 
 	rdev->pm.current_power_state_index = rdev->pm.default_power_state_index;
 	rdev->pm.current_clock_mode_index = 0;
+	rdev->pm.current_vddc = rdev->pm.power_state[rdev->pm.default_power_state_index].clock_info[0].voltage.voltage;
 }
 
 void radeon_atom_set_clock_gating(struct radeon_device *rdev, int enable)
@@ -1989,6 +2022,42 @@ void radeon_atom_set_memory_clock(struct radeon_device *rdev,
 	atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args);
 }
 
+union set_voltage {
+	struct _SET_VOLTAGE_PS_ALLOCATION alloc;
+	struct _SET_VOLTAGE_PARAMETERS v1;
+	struct _SET_VOLTAGE_PARAMETERS_V2 v2;
+};
+
+void radeon_atom_set_voltage(struct radeon_device *rdev, u16 level)
+{
+	union set_voltage args;
+	int index = GetIndexIntoMasterTable(COMMAND, SetVoltage);
+	u8 frev, crev, volt_index = level;
+
+	if (!atom_parse_cmd_header(rdev->mode_info.atom_context, index, &frev, &crev))
+		return;
+
+	switch (crev) {
+	case 1:
+		args.v1.ucVoltageType = SET_VOLTAGE_TYPE_ASIC_VDDC;
+		args.v1.ucVoltageMode = SET_ASIC_VOLTAGE_MODE_ALL_SOURCE;
+		args.v1.ucVoltageIndex = volt_index;
+		break;
+	case 2:
+		args.v2.ucVoltageType = SET_VOLTAGE_TYPE_ASIC_VDDC;
+		args.v2.ucVoltageMode = SET_ASIC_VOLTAGE_MODE_SET_VOLTAGE;
+		args.v2.usVoltageLevel = cpu_to_le16(level);
+		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);
+}
+
+
+
 void radeon_atom_initialize_bios_scratch_regs(struct drm_device *dev)
 {
 	struct radeon_device *rdev = dev->dev_private;
diff --git a/drivers/gpu/drm/radeon/radeon_combios.c b/drivers/gpu/drm/radeon/radeon_combios.c
index 7b5e10d3e9c9..1bee2f9e24a5 100644
--- a/drivers/gpu/drm/radeon/radeon_combios.c
+++ b/drivers/gpu/drm/radeon/radeon_combios.c
@@ -2026,6 +2026,7 @@ bool radeon_get_legacy_connector_info_from_bios(struct drm_device *dev)
 					combios_setup_i2c_bus(rdev, RADEON_GPIO_CRT2_DDC);
 				break;
 			default:
+				ddc_i2c.valid = false;
 				break;
 			}
 
@@ -2339,6 +2340,7 @@ bool radeon_get_legacy_connector_info_from_bios(struct drm_device *dev)
 			if (RBIOS8(tv_info + 6) == 'T') {
 				if (radeon_apply_legacy_tv_quirks(dev)) {
 					hpd.hpd = RADEON_HPD_NONE;
+					ddc_i2c.valid = false;
 					radeon_add_legacy_encoder(dev,
 								  radeon_get_encoder_id
 								  (dev,
@@ -2454,7 +2456,12 @@ default_mode:
 	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 ((state_index > 0) &&
+	    (rdev->pm.power_state[0].clock_info[0].voltage.type == VOLTAGE_GPIO))
+		rdev->pm.power_state[state_index].clock_info[0].voltage =
+			rdev->pm.power_state[0].clock_info[0].voltage;
+	else
+		rdev->pm.power_state[state_index].clock_info[0].voltage.type = VOLTAGE_NONE;
 	rdev->pm.power_state[state_index].pcie_lanes = 16;
 	rdev->pm.power_state[state_index].flags = 0;
 	rdev->pm.default_power_state_index = state_index;
diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c
index a20b612ffe75..f10faed21567 100644
--- a/drivers/gpu/drm/radeon/radeon_device.c
+++ b/drivers/gpu/drm/radeon/radeon_device.c
@@ -546,8 +546,10 @@ static void radeon_switcheroo_set_state(struct pci_dev *pdev, enum vga_switchero
 		/* don't suspend or resume card normally */
 		rdev->powered_down = false;
 		radeon_resume_kms(dev);
+		drm_kms_helper_poll_enable(dev);
 	} else {
 		printk(KERN_INFO "radeon: switched off\n");
+		drm_kms_helper_poll_disable(dev);
 		radeon_suspend_kms(dev, pmm);
 		/* don't suspend or resume card normally */
 		rdev->powered_down = true;
@@ -711,6 +713,7 @@ int radeon_suspend_kms(struct drm_device *dev, pm_message_t state)
 {
 	struct radeon_device *rdev;
 	struct drm_crtc *crtc;
+	struct drm_connector *connector;
 	int r;
 
 	if (dev == NULL || dev->dev_private == NULL) {
@@ -723,6 +726,12 @@ int radeon_suspend_kms(struct drm_device *dev, pm_message_t state)
 
 	if (rdev->powered_down)
 		return 0;
+
+	/* turn off display hw */
+	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
+		drm_helper_connector_dpms(connector, DRM_MODE_DPMS_OFF);
+	}
+
 	/* unpin the front buffers */
 	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
 		struct radeon_framebuffer *rfb = to_radeon_framebuffer(crtc->fb);
@@ -754,6 +763,8 @@ int radeon_suspend_kms(struct drm_device *dev, pm_message_t state)
 	/* evict remaining vram memory */
 	radeon_bo_evict_vram(rdev);
 
+	radeon_agp_suspend(rdev);
+
 	pci_save_state(dev->pdev);
 	if (state.event == PM_EVENT_SUSPEND) {
 		/* Shut down the device */
diff --git a/drivers/gpu/drm/radeon/radeon_display.c b/drivers/gpu/drm/radeon/radeon_display.c
index 1006549d1570..8154cdf796e4 100644
--- a/drivers/gpu/drm/radeon/radeon_display.c
+++ b/drivers/gpu/drm/radeon/radeon_display.c
@@ -284,8 +284,7 @@ static const char *connector_names[15] = {
 	"eDP",
 };
 
-static const char *hpd_names[7] = {
-	"NONE",
+static const char *hpd_names[6] = {
 	"HPD1",
 	"HPD2",
 	"HPD3",
diff --git a/drivers/gpu/drm/radeon/radeon_drv.c b/drivers/gpu/drm/radeon/radeon_drv.c
index 902d1731a652..e166fe4d7c30 100644
--- a/drivers/gpu/drm/radeon/radeon_drv.c
+++ b/drivers/gpu/drm/radeon/radeon_drv.c
@@ -45,9 +45,10 @@
  * - 2.2.0 - add r6xx/r7xx const buffer support
  * - 2.3.0 - add MSPOS + 3D texture + r500 VAP regs
  * - 2.4.0 - add crtc id query
+ * - 2.5.0 - add get accel 2 to work around ddx breakage for evergreen
  */
 #define KMS_DRIVER_MAJOR	2
-#define KMS_DRIVER_MINOR	4
+#define KMS_DRIVER_MINOR	5
 #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);
diff --git a/drivers/gpu/drm/radeon/radeon_fb.c b/drivers/gpu/drm/radeon/radeon_fb.c
index e192acfbf0cd..dc1634bb0c11 100644
--- a/drivers/gpu/drm/radeon/radeon_fb.c
+++ b/drivers/gpu/drm/radeon/radeon_fb.c
@@ -363,6 +363,7 @@ int radeon_fbdev_init(struct radeon_device *rdev)
 {
 	struct radeon_fbdev *rfbdev;
 	int bpp_sel = 32;
+	int ret;
 
 	/* select 8 bpp console on RN50 or 16MB cards */
 	if (ASIC_IS_RN50(rdev) || rdev->mc.real_vram_size <= (32*1024*1024))
@@ -376,9 +377,14 @@ int radeon_fbdev_init(struct radeon_device *rdev)
 	rdev->mode_info.rfbdev = rfbdev;
 	rfbdev->helper.funcs = &radeon_fb_helper_funcs;
 
-	drm_fb_helper_init(rdev->ddev, &rfbdev->helper,
-			   rdev->num_crtc,
-			   RADEONFB_CONN_LIMIT);
+	ret = drm_fb_helper_init(rdev->ddev, &rfbdev->helper,
+				 rdev->num_crtc,
+				 RADEONFB_CONN_LIMIT);
+	if (ret) {
+		kfree(rfbdev);
+		return ret;
+	}
+
 	drm_fb_helper_single_add_all_connectors(&rfbdev->helper);
 	drm_fb_helper_initial_config(&rfbdev->helper, bpp_sel);
 	return 0;
diff --git a/drivers/gpu/drm/radeon/radeon_kms.c b/drivers/gpu/drm/radeon/radeon_kms.c
index 04068352ccd2..6a70c0dc7f92 100644
--- a/drivers/gpu/drm/radeon/radeon_kms.c
+++ b/drivers/gpu/drm/radeon/radeon_kms.c
@@ -118,7 +118,11 @@ int radeon_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		value = rdev->num_z_pipes;
 		break;
 	case RADEON_INFO_ACCEL_WORKING:
-		value = rdev->accel_working;
+		/* xf86-video-ati 6.13.0 relies on this being false for evergreen */
+		if ((rdev->family >= CHIP_CEDAR) && (rdev->family <= CHIP_HEMLOCK))
+			value = false;
+		else
+			value = rdev->accel_working;
 		break;
 	case RADEON_INFO_CRTC_FROM_ID:
 		for (i = 0, found = 0; i < rdev->num_crtc; i++) {
@@ -134,6 +138,9 @@ int radeon_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 			return -EINVAL;
 		}
 		break;
+	case RADEON_INFO_ACCEL_WORKING2:
+		value = rdev->accel_working;
+		break;
 	default:
 		DRM_DEBUG("Invalid request %d\n", info->request);
 		return -EINVAL;
diff --git a/drivers/gpu/drm/radeon/radeon_legacy_encoders.c b/drivers/gpu/drm/radeon/radeon_legacy_encoders.c
index 5a13b3eeef19..5b07b8848e09 100644
--- a/drivers/gpu/drm/radeon/radeon_legacy_encoders.c
+++ b/drivers/gpu/drm/radeon/radeon_legacy_encoders.c
@@ -1168,6 +1168,17 @@ static enum drm_connector_status radeon_legacy_tv_dac_detect(struct drm_encoder
 	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
 	struct radeon_encoder_tv_dac *tv_dac = radeon_encoder->enc_priv;
 	bool color = true;
+	struct drm_crtc *crtc;
+
+	/* find out if crtc2 is in use or if this encoder is using it */
+	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
+		struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
+		if ((radeon_crtc->crtc_id == 1) && crtc->enabled) {
+			if (encoder->crtc != crtc) {
+				return connector_status_disconnected;
+			}
+		}
+	}
 
 	if (connector->connector_type == DRM_MODE_CONNECTOR_SVIDEO ||
 	    connector->connector_type == DRM_MODE_CONNECTOR_Composite ||
diff --git a/drivers/gpu/drm/radeon/radeon_pm.c b/drivers/gpu/drm/radeon/radeon_pm.c
index a8d162c6f829..63f679a04b25 100644
--- a/drivers/gpu/drm/radeon/radeon_pm.c
+++ b/drivers/gpu/drm/radeon/radeon_pm.c
@@ -33,6 +33,14 @@
 #define RADEON_WAIT_VBLANK_TIMEOUT 200
 #define RADEON_WAIT_IDLE_TIMEOUT 200
 
+static const char *radeon_pm_state_type_name[5] = {
+	"Default",
+	"Powersave",
+	"Battery",
+	"Balanced",
+	"Performance",
+};
+
 static void radeon_dynpm_idle_work_handler(struct work_struct *work);
 static int radeon_debugfs_pm_init(struct radeon_device *rdev);
 static bool radeon_pm_in_vbl(struct radeon_device *rdev);
@@ -84,9 +92,9 @@ static void radeon_pm_update_profile(struct radeon_device *rdev)
 				rdev->pm.profile_index = PM_PROFILE_HIGH_SH_IDX;
 		} else {
 			if (rdev->pm.active_crtc_count > 1)
-				rdev->pm.profile_index = PM_PROFILE_LOW_MH_IDX;
+				rdev->pm.profile_index = PM_PROFILE_MID_MH_IDX;
 			else
-				rdev->pm.profile_index = PM_PROFILE_LOW_SH_IDX;
+				rdev->pm.profile_index = PM_PROFILE_MID_SH_IDX;
 		}
 		break;
 	case PM_PROFILE_LOW:
@@ -95,6 +103,12 @@ static void radeon_pm_update_profile(struct radeon_device *rdev)
 		else
 			rdev->pm.profile_index = PM_PROFILE_LOW_SH_IDX;
 		break;
+	case PM_PROFILE_MID:
+		if (rdev->pm.active_crtc_count > 1)
+			rdev->pm.profile_index = PM_PROFILE_MID_MH_IDX;
+		else
+			rdev->pm.profile_index = PM_PROFILE_MID_SH_IDX;
+		break;
 	case PM_PROFILE_HIGH:
 		if (rdev->pm.active_crtc_count > 1)
 			rdev->pm.profile_index = PM_PROFILE_HIGH_MH_IDX;
@@ -127,15 +141,6 @@ static void radeon_unmap_vram_bos(struct radeon_device *rdev)
 		if (bo->tbo.mem.mem_type == TTM_PL_VRAM)
 			ttm_bo_unmap_virtual(&bo->tbo);
 	}
-
-	if (rdev->gart.table.vram.robj)
-		ttm_bo_unmap_virtual(&rdev->gart.table.vram.robj->tbo);
-
-	if (rdev->stollen_vga_memory)
-		ttm_bo_unmap_virtual(&rdev->stollen_vga_memory->tbo);
-
-	if (rdev->r600_blit.shader_obj)
-		ttm_bo_unmap_virtual(&rdev->r600_blit.shader_obj->tbo);
 }
 
 static void radeon_sync_with_vblank(struct radeon_device *rdev)
@@ -151,6 +156,7 @@ static void radeon_sync_with_vblank(struct radeon_device *rdev)
 static void radeon_set_power_state(struct radeon_device *rdev)
 {
 	u32 sclk, mclk;
+	bool misc_after = false;
 
 	if ((rdev->pm.requested_clock_mode_index == rdev->pm.current_clock_mode_index) &&
 	    (rdev->pm.requested_power_state_index == rdev->pm.current_power_state_index))
@@ -167,55 +173,47 @@ static void radeon_set_power_state(struct radeon_device *rdev)
 		if (mclk > rdev->clock.default_mclk)
 			mclk = rdev->clock.default_mclk;
 
-		/* voltage, pcie lanes, etc.*/
-		radeon_pm_misc(rdev);
+		/* upvolt before raising clocks, downvolt after lowering clocks */
+		if (sclk < rdev->pm.current_sclk)
+			misc_after = true;
 
-		if (rdev->pm.pm_method == PM_METHOD_DYNPM) {
-			radeon_sync_with_vblank(rdev);
+		radeon_sync_with_vblank(rdev);
 
+		if (rdev->pm.pm_method == PM_METHOD_DYNPM) {
 			if (!radeon_pm_in_vbl(rdev))
 				return;
+		}
 
-			radeon_pm_prepare(rdev);
-			/* set engine clock */
-			if (sclk != rdev->pm.current_sclk) {
-				radeon_pm_debug_check_in_vbl(rdev, false);
-				radeon_set_engine_clock(rdev, sclk);
-				radeon_pm_debug_check_in_vbl(rdev, true);
-				rdev->pm.current_sclk = sclk;
-				DRM_DEBUG("Setting: e: %d\n", sclk);
-			}
+		radeon_pm_prepare(rdev);
 
-			/* set memory clock */
-			if (rdev->asic->set_memory_clock && (mclk != rdev->pm.current_mclk)) {
-				radeon_pm_debug_check_in_vbl(rdev, false);
-				radeon_set_memory_clock(rdev, mclk);
-				radeon_pm_debug_check_in_vbl(rdev, true);
-				rdev->pm.current_mclk = mclk;
-				DRM_DEBUG("Setting: m: %d\n", mclk);
-			}
-			radeon_pm_finish(rdev);
-		} else {
-			/* set engine clock */
-			if (sclk != rdev->pm.current_sclk) {
-				radeon_sync_with_vblank(rdev);
-				radeon_pm_prepare(rdev);
-				radeon_set_engine_clock(rdev, sclk);
-				radeon_pm_finish(rdev);
-				rdev->pm.current_sclk = sclk;
-				DRM_DEBUG("Setting: e: %d\n", sclk);
-			}
-			/* set memory clock */
-			if (rdev->asic->set_memory_clock && (mclk != rdev->pm.current_mclk)) {
-				radeon_sync_with_vblank(rdev);
-				radeon_pm_prepare(rdev);
-				radeon_set_memory_clock(rdev, mclk);
-				radeon_pm_finish(rdev);
-				rdev->pm.current_mclk = mclk;
-				DRM_DEBUG("Setting: m: %d\n", mclk);
-			}
+		if (!misc_after)
+			/* voltage, pcie lanes, etc.*/
+			radeon_pm_misc(rdev);
+
+		/* set engine clock */
+		if (sclk != rdev->pm.current_sclk) {
+			radeon_pm_debug_check_in_vbl(rdev, false);
+			radeon_set_engine_clock(rdev, sclk);
+			radeon_pm_debug_check_in_vbl(rdev, true);
+			rdev->pm.current_sclk = sclk;
+			DRM_DEBUG("Setting: e: %d\n", sclk);
 		}
 
+		/* set memory clock */
+		if (rdev->asic->set_memory_clock && (mclk != rdev->pm.current_mclk)) {
+			radeon_pm_debug_check_in_vbl(rdev, false);
+			radeon_set_memory_clock(rdev, mclk);
+			radeon_pm_debug_check_in_vbl(rdev, true);
+			rdev->pm.current_mclk = mclk;
+			DRM_DEBUG("Setting: m: %d\n", mclk);
+		}
+
+		if (misc_after)
+			/* voltage, pcie lanes, etc.*/
+			radeon_pm_misc(rdev);
+
+		radeon_pm_finish(rdev);
+
 		rdev->pm.current_power_state_index = rdev->pm.requested_power_state_index;
 		rdev->pm.current_clock_mode_index = rdev->pm.requested_clock_mode_index;
 	} else
@@ -288,6 +286,42 @@ static void radeon_pm_set_clocks(struct radeon_device *rdev)
 	mutex_unlock(&rdev->ddev->struct_mutex);
 }
 
+static void radeon_pm_print_states(struct radeon_device *rdev)
+{
+	int i, j;
+	struct radeon_power_state *power_state;
+	struct radeon_pm_clock_info *clock_info;
+
+	DRM_DEBUG("%d Power State(s)\n", rdev->pm.num_power_states);
+	for (i = 0; i < rdev->pm.num_power_states; i++) {
+		power_state = &rdev->pm.power_state[i];
+		DRM_DEBUG("State %d: %s\n", i,
+			radeon_pm_state_type_name[power_state->type]);
+		if (i == rdev->pm.default_power_state_index)
+			DRM_DEBUG("\tDefault");
+		if ((rdev->flags & RADEON_IS_PCIE) && !(rdev->flags & RADEON_IS_IGP))
+			DRM_DEBUG("\t%d PCIE Lanes\n", power_state->pcie_lanes);
+		if (power_state->flags & RADEON_PM_STATE_SINGLE_DISPLAY_ONLY)
+			DRM_DEBUG("\tSingle display only\n");
+		DRM_DEBUG("\t%d Clock Mode(s)\n", power_state->num_clock_modes);
+		for (j = 0; j < power_state->num_clock_modes; j++) {
+			clock_info = &(power_state->clock_info[j]);
+			if (rdev->flags & RADEON_IS_IGP)
+				DRM_DEBUG("\t\t%d e: %d%s\n",
+					j,
+					clock_info->sclk * 10,
+					clock_info->flags & RADEON_PM_MODE_NO_DISPLAY ? "\tNo display only" : "");
+			else
+				DRM_DEBUG("\t\t%d e: %d\tm: %d\tv: %d%s\n",
+					j,
+					clock_info->sclk * 10,
+					clock_info->mclk * 10,
+					clock_info->voltage.voltage,
+					clock_info->flags & RADEON_PM_MODE_NO_DISPLAY ? "\tNo display only" : "");
+		}
+	}
+}
+
 static ssize_t radeon_get_pm_profile(struct device *dev,
 				     struct device_attribute *attr,
 				     char *buf)
@@ -318,6 +352,8 @@ static ssize_t radeon_set_pm_profile(struct device *dev,
 			rdev->pm.profile = PM_PROFILE_AUTO;
 		else if (strncmp("low", buf, strlen("low")) == 0)
 			rdev->pm.profile = PM_PROFILE_LOW;
+		else if (strncmp("mid", buf, strlen("mid")) == 0)
+			rdev->pm.profile = PM_PROFILE_MID;
 		else if (strncmp("high", buf, strlen("high")) == 0)
 			rdev->pm.profile = PM_PROFILE_HIGH;
 		else {
@@ -384,15 +420,19 @@ void radeon_pm_suspend(struct radeon_device *rdev)
 {
 	mutex_lock(&rdev->pm.mutex);
 	cancel_delayed_work(&rdev->pm.dynpm_idle_work);
-	rdev->pm.current_power_state_index = -1;
-	rdev->pm.current_clock_mode_index = -1;
-	rdev->pm.current_sclk = 0;
-	rdev->pm.current_mclk = 0;
 	mutex_unlock(&rdev->pm.mutex);
 }
 
 void radeon_pm_resume(struct radeon_device *rdev)
 {
+	/* asic init will reset the default power state */
+	mutex_lock(&rdev->pm.mutex);
+	rdev->pm.current_power_state_index = rdev->pm.default_power_state_index;
+	rdev->pm.current_clock_mode_index = 0;
+	rdev->pm.current_sclk = rdev->clock.default_sclk;
+	rdev->pm.current_mclk = rdev->clock.default_mclk;
+	rdev->pm.current_vddc = rdev->pm.power_state[rdev->pm.default_power_state_index].clock_info[0].voltage.voltage;
+	mutex_unlock(&rdev->pm.mutex);
 	radeon_pm_compute_clocks(rdev);
 }
 
@@ -401,32 +441,24 @@ int radeon_pm_init(struct radeon_device *rdev)
 	int ret;
 	/* default to profile method */
 	rdev->pm.pm_method = PM_METHOD_PROFILE;
+	rdev->pm.profile = PM_PROFILE_DEFAULT;
 	rdev->pm.dynpm_state = DYNPM_STATE_DISABLED;
 	rdev->pm.dynpm_planned_action = DYNPM_ACTION_NONE;
 	rdev->pm.dynpm_can_upclock = true;
 	rdev->pm.dynpm_can_downclock = true;
-	rdev->pm.current_sclk = 0;
-	rdev->pm.current_mclk = 0;
+	rdev->pm.current_sclk = rdev->clock.default_sclk;
+	rdev->pm.current_mclk = rdev->clock.default_mclk;
 
 	if (rdev->bios) {
 		if (rdev->is_atom_bios)
 			radeon_atombios_get_power_modes(rdev);
 		else
 			radeon_combios_get_power_modes(rdev);
+		radeon_pm_print_states(rdev);
 		radeon_pm_init_profile(rdev);
-		rdev->pm.current_power_state_index = -1;
-		rdev->pm.current_clock_mode_index = -1;
 	}
 
 	if (rdev->pm.num_power_states > 1) {
-		if (rdev->pm.pm_method == PM_METHOD_PROFILE) {
-			mutex_lock(&rdev->pm.mutex);
-			rdev->pm.profile = PM_PROFILE_DEFAULT;
-			radeon_pm_update_profile(rdev);
-			radeon_pm_set_clocks(rdev);
-			mutex_unlock(&rdev->pm.mutex);
-		}
-
 		/* where's the best place to put these? */
 		ret = device_create_file(rdev->dev, &dev_attr_power_profile);
 		if (ret)
@@ -712,6 +744,8 @@ static int radeon_debugfs_pm_info(struct seq_file *m, void *data)
 	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->pm.current_vddc)
+		seq_printf(m, "voltage: %u mV\n", rdev->pm.current_vddc);
 	if (rdev->asic->get_pcie_lanes)
 		seq_printf(m, "PCIE lanes: %d\n", radeon_get_pcie_lanes(rdev));
 
diff --git a/drivers/gpu/drm/radeon/radeon_state.c b/drivers/gpu/drm/radeon/radeon_state.c
index cc5316dcf580..b3ba44c0a818 100644
--- a/drivers/gpu/drm/radeon/radeon_state.c
+++ b/drivers/gpu/drm/radeon/radeon_state.c
@@ -900,9 +900,10 @@ static void radeon_cp_dispatch_clear(struct drm_device * dev,
 			flags |= RADEON_FRONT;
 	}
 	if (flags & (RADEON_DEPTH|RADEON_STENCIL)) {
-		if (!dev_priv->have_z_offset)
+		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);
+			flags &= ~(RADEON_DEPTH | RADEON_STENCIL);
+		}
 	}
 
 	if (flags & (RADEON_FRONT | RADEON_BACK)) {
diff --git a/drivers/gpu/drm/radeon/reg_srcs/evergreen b/drivers/gpu/drm/radeon/reg_srcs/evergreen
new file mode 100644
index 000000000000..b5c757f68d3c
--- /dev/null
+++ b/drivers/gpu/drm/radeon/reg_srcs/evergreen
@@ -0,0 +1,611 @@
+evergreen 0x9400
+0x00008040 WAIT_UNTIL
+0x00008044 WAIT_UNTIL_POLL_CNTL
+0x00008048 WAIT_UNTIL_POLL_MASK
+0x0000804c WAIT_UNTIL_POLL_REFDATA
+0x000088B0 VGT_VTX_VECT_EJECT_REG
+0x000088C4 VGT_CACHE_INVALIDATION
+0x000088D4 VGT_GS_VERTEX_REUSE
+0x00008958 VGT_PRIMITIVE_TYPE
+0x0000895C VGT_INDEX_TYPE
+0x00008970 VGT_NUM_INDICES
+0x00008974 VGT_NUM_INSTANCES
+0x00008990 VGT_COMPUTE_DIM_X
+0x00008994 VGT_COMPUTE_DIM_Y
+0x00008998 VGT_COMPUTE_DIM_Z
+0x0000899C VGT_COMPUTE_START_X
+0x000089A0 VGT_COMPUTE_START_Y
+0x000089A4 VGT_COMPUTE_START_Z
+0x000089AC VGT_COMPUTE_THREAD_GOURP_SIZE
+0x00008A14 PA_CL_ENHANCE
+0x00008A60 PA_SC_LINE_STIPPLE_VALUE
+0x00008B10 PA_SC_LINE_STIPPLE_STATE
+0x00008BF0 PA_SC_ENHANCE
+0x00008D8C SQ_DYN_GPR_CNTL_PS_FLUSH_REQ
+0x00008C00 SQ_CONFIG
+0x00008C04 SQ_GPR_RESOURCE_MGMT_1
+0x00008C08 SQ_GPR_RESOURCE_MGMT_2
+0x00008C0C SQ_GPR_RESOURCE_MGMT_3
+0x00008C10 SQ_GLOBAL_GPR_RESOURCE_MGMT_1
+0x00008C14 SQ_GLOBAL_GPR_RESOURCE_MGMT_2
+0x00008C18 SQ_THREAD_RESOURCE_MGMT
+0x00008C1C SQ_THREAD_RESOURCE_MGMT_2
+0x00008C20 SQ_STACK_RESOURCE_MGMT_1
+0x00008C24 SQ_STACK_RESOURCE_MGMT_2
+0x00008C28 SQ_STACK_RESOURCE_MGMT_3
+0x00008DF8 SQ_CONST_MEM_BASE
+0x00008E48 SQ_EX_ALLOC_TABLE_SLOTS
+0x00009100 SPI_CONFIG_CNTL
+0x0000913C SPI_CONFIG_CNTL_1
+0x00009700 VC_CNTL
+0x00009714 VC_ENHANCE
+0x00009830 DB_DEBUG
+0x00009834 DB_DEBUG2
+0x00009838 DB_DEBUG3
+0x0000983C DB_DEBUG4
+0x00009854 DB_WATERMARKS
+0x0000A400 TD_PS_BORDER_COLOR_INDEX
+0x0000A404 TD_PS_BORDER_COLOR_RED
+0x0000A408 TD_PS_BORDER_COLOR_GREEN
+0x0000A40C TD_PS_BORDER_COLOR_BLUE
+0x0000A410 TD_PS_BORDER_COLOR_ALPHA
+0x0000A414 TD_VS_BORDER_COLOR_INDEX
+0x0000A418 TD_VS_BORDER_COLOR_RED
+0x0000A41C TD_VS_BORDER_COLOR_GREEN
+0x0000A420 TD_VS_BORDER_COLOR_BLUE
+0x0000A424 TD_VS_BORDER_COLOR_ALPHA
+0x0000A428 TD_GS_BORDER_COLOR_INDEX
+0x0000A42C TD_GS_BORDER_COLOR_RED
+0x0000A430 TD_GS_BORDER_COLOR_GREEN
+0x0000A434 TD_GS_BORDER_COLOR_BLUE
+0x0000A438 TD_GS_BORDER_COLOR_ALPHA
+0x0000A43C TD_HS_BORDER_COLOR_INDEX
+0x0000A440 TD_HS_BORDER_COLOR_RED
+0x0000A444 TD_HS_BORDER_COLOR_GREEN
+0x0000A448 TD_HS_BORDER_COLOR_BLUE
+0x0000A44C TD_HS_BORDER_COLOR_ALPHA
+0x0000A450 TD_LS_BORDER_COLOR_INDEX
+0x0000A454 TD_LS_BORDER_COLOR_RED
+0x0000A458 TD_LS_BORDER_COLOR_GREEN
+0x0000A45C TD_LS_BORDER_COLOR_BLUE
+0x0000A460 TD_LS_BORDER_COLOR_ALPHA
+0x0000A464 TD_CS_BORDER_COLOR_INDEX
+0x0000A468 TD_CS_BORDER_COLOR_RED
+0x0000A46C TD_CS_BORDER_COLOR_GREEN
+0x0000A470 TD_CS_BORDER_COLOR_BLUE
+0x0000A474 TD_CS_BORDER_COLOR_ALPHA
+0x00028000 DB_RENDER_CONTROL
+0x00028004 DB_COUNT_CONTROL
+0x0002800C DB_RENDER_OVERRIDE
+0x00028010 DB_RENDER_OVERRIDE2
+0x00028028 DB_STENCIL_CLEAR
+0x0002802C DB_DEPTH_CLEAR
+0x00028034 PA_SC_SCREEN_SCISSOR_BR
+0x00028030 PA_SC_SCREEN_SCISSOR_TL
+0x0002805C DB_DEPTH_SLICE
+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
+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
+0x00028200 PA_SC_WINDOW_OFFSET
+0x00028204 PA_SC_WINDOW_SCISSOR_TL
+0x00028208 PA_SC_WINDOW_SCISSOR_BR
+0x0002820C PA_SC_CLIPRECT_RULE
+0x00028210 PA_SC_CLIPRECT_0_TL
+0x00028214 PA_SC_CLIPRECT_0_BR
+0x00028218 PA_SC_CLIPRECT_1_TL
+0x0002821C PA_SC_CLIPRECT_1_BR
+0x00028220 PA_SC_CLIPRECT_2_TL
+0x00028224 PA_SC_CLIPRECT_2_BR
+0x00028228 PA_SC_CLIPRECT_3_TL
+0x0002822C PA_SC_CLIPRECT_3_BR
+0x00028230 PA_SC_EDGERULE
+0x00028234 PA_SU_HARDWARE_SCREEN_OFFSET
+0x00028240 PA_SC_GENERIC_SCISSOR_TL
+0x00028244 PA_SC_GENERIC_SCISSOR_BR
+0x00028250 PA_SC_VPORT_SCISSOR_0_TL
+0x00028254 PA_SC_VPORT_SCISSOR_0_BR
+0x00028258 PA_SC_VPORT_SCISSOR_1_TL
+0x0002825C PA_SC_VPORT_SCISSOR_1_BR
+0x00028260 PA_SC_VPORT_SCISSOR_2_TL
+0x00028264 PA_SC_VPORT_SCISSOR_2_BR
+0x00028268 PA_SC_VPORT_SCISSOR_3_TL
+0x0002826C PA_SC_VPORT_SCISSOR_3_BR
+0x00028270 PA_SC_VPORT_SCISSOR_4_TL
+0x00028274 PA_SC_VPORT_SCISSOR_4_BR
+0x00028278 PA_SC_VPORT_SCISSOR_5_TL
+0x0002827C PA_SC_VPORT_SCISSOR_5_BR
+0x00028280 PA_SC_VPORT_SCISSOR_6_TL
+0x00028284 PA_SC_VPORT_SCISSOR_6_BR
+0x00028288 PA_SC_VPORT_SCISSOR_7_TL
+0x0002828C PA_SC_VPORT_SCISSOR_7_BR
+0x00028290 PA_SC_VPORT_SCISSOR_8_TL
+0x00028294 PA_SC_VPORT_SCISSOR_8_BR
+0x00028298 PA_SC_VPORT_SCISSOR_9_TL
+0x0002829C PA_SC_VPORT_SCISSOR_9_BR
+0x000282A0 PA_SC_VPORT_SCISSOR_10_TL
+0x000282A4 PA_SC_VPORT_SCISSOR_10_BR
+0x000282A8 PA_SC_VPORT_SCISSOR_11_TL
+0x000282AC PA_SC_VPORT_SCISSOR_11_BR
+0x000282B0 PA_SC_VPORT_SCISSOR_12_TL
+0x000282B4 PA_SC_VPORT_SCISSOR_12_BR
+0x000282B8 PA_SC_VPORT_SCISSOR_13_TL
+0x000282BC PA_SC_VPORT_SCISSOR_13_BR
+0x000282C0 PA_SC_VPORT_SCISSOR_14_TL
+0x000282C4 PA_SC_VPORT_SCISSOR_14_BR
+0x000282C8 PA_SC_VPORT_SCISSOR_15_TL
+0x000282CC PA_SC_VPORT_SCISSOR_15_BR
+0x000282D0 PA_SC_VPORT_ZMIN_0
+0x000282D4 PA_SC_VPORT_ZMAX_0
+0x000282D8 PA_SC_VPORT_ZMIN_1
+0x000282DC PA_SC_VPORT_ZMAX_1
+0x000282E0 PA_SC_VPORT_ZMIN_2
+0x000282E4 PA_SC_VPORT_ZMAX_2
+0x000282E8 PA_SC_VPORT_ZMIN_3
+0x000282EC PA_SC_VPORT_ZMAX_3
+0x000282F0 PA_SC_VPORT_ZMIN_4
+0x000282F4 PA_SC_VPORT_ZMAX_4
+0x000282F8 PA_SC_VPORT_ZMIN_5
+0x000282FC PA_SC_VPORT_ZMAX_5
+0x00028300 PA_SC_VPORT_ZMIN_6
+0x00028304 PA_SC_VPORT_ZMAX_6
+0x00028308 PA_SC_VPORT_ZMIN_7
+0x0002830C PA_SC_VPORT_ZMAX_7
+0x00028310 PA_SC_VPORT_ZMIN_8
+0x00028314 PA_SC_VPORT_ZMAX_8
+0x00028318 PA_SC_VPORT_ZMIN_9
+0x0002831C PA_SC_VPORT_ZMAX_9
+0x00028320 PA_SC_VPORT_ZMIN_10
+0x00028324 PA_SC_VPORT_ZMAX_10
+0x00028328 PA_SC_VPORT_ZMIN_11
+0x0002832C PA_SC_VPORT_ZMAX_11
+0x00028330 PA_SC_VPORT_ZMIN_12
+0x00028334 PA_SC_VPORT_ZMAX_12
+0x00028338 PA_SC_VPORT_ZMIN_13
+0x0002833C PA_SC_VPORT_ZMAX_13
+0x00028340 PA_SC_VPORT_ZMIN_14
+0x00028344 PA_SC_VPORT_ZMAX_14
+0x00028348 PA_SC_VPORT_ZMIN_15
+0x0002834C PA_SC_VPORT_ZMAX_15
+0x00028350 SX_MISC
+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
+0x00028400 VGT_MAX_VTX_INDX
+0x00028404 VGT_MIN_VTX_INDX
+0x00028408 VGT_INDX_OFFSET
+0x0002840C VGT_MULTI_PRIM_IB_RESET_INDX
+0x00028410 SX_ALPHA_TEST_CONTROL
+0x00028414 CB_BLEND_RED
+0x00028418 CB_BLEND_GREEN
+0x0002841C CB_BLEND_BLUE
+0x00028420 CB_BLEND_ALPHA
+0x00028430 DB_STENCILREFMASK
+0x00028434 DB_STENCILREFMASK_BF
+0x00028438 SX_ALPHA_REF
+0x0002843C PA_CL_VPORT_XSCALE_0
+0x00028440 PA_CL_VPORT_XOFFSET_0
+0x00028444 PA_CL_VPORT_YSCALE_0
+0x00028448 PA_CL_VPORT_YOFFSET_0
+0x0002844C PA_CL_VPORT_ZSCALE_0
+0x00028450 PA_CL_VPORT_ZOFFSET_0
+0x00028454 PA_CL_VPORT_XSCALE_1
+0x00028458 PA_CL_VPORT_XOFFSET_1
+0x0002845C PA_CL_VPORT_YSCALE_1
+0x00028460 PA_CL_VPORT_YOFFSET_1
+0x00028464 PA_CL_VPORT_ZSCALE_1
+0x00028468 PA_CL_VPORT_ZOFFSET_1
+0x0002846C PA_CL_VPORT_XSCALE_2
+0x00028470 PA_CL_VPORT_XOFFSET_2
+0x00028474 PA_CL_VPORT_YSCALE_2
+0x00028478 PA_CL_VPORT_YOFFSET_2
+0x0002847C PA_CL_VPORT_ZSCALE_2
+0x00028480 PA_CL_VPORT_ZOFFSET_2
+0x00028484 PA_CL_VPORT_XSCALE_3
+0x00028488 PA_CL_VPORT_XOFFSET_3
+0x0002848C PA_CL_VPORT_YSCALE_3
+0x00028490 PA_CL_VPORT_YOFFSET_3
+0x00028494 PA_CL_VPORT_ZSCALE_3
+0x00028498 PA_CL_VPORT_ZOFFSET_3
+0x0002849C PA_CL_VPORT_XSCALE_4
+0x000284A0 PA_CL_VPORT_XOFFSET_4
+0x000284A4 PA_CL_VPORT_YSCALE_4
+0x000284A8 PA_CL_VPORT_YOFFSET_4
+0x000284AC PA_CL_VPORT_ZSCALE_4
+0x000284B0 PA_CL_VPORT_ZOFFSET_4
+0x000284B4 PA_CL_VPORT_XSCALE_5
+0x000284B8 PA_CL_VPORT_XOFFSET_5
+0x000284BC PA_CL_VPORT_YSCALE_5
+0x000284C0 PA_CL_VPORT_YOFFSET_5
+0x000284C4 PA_CL_VPORT_ZSCALE_5
+0x000284C8 PA_CL_VPORT_ZOFFSET_5
+0x000284CC PA_CL_VPORT_XSCALE_6
+0x000284D0 PA_CL_VPORT_XOFFSET_6
+0x000284D4 PA_CL_VPORT_YSCALE_6
+0x000284D8 PA_CL_VPORT_YOFFSET_6
+0x000284DC PA_CL_VPORT_ZSCALE_6
+0x000284E0 PA_CL_VPORT_ZOFFSET_6
+0x000284E4 PA_CL_VPORT_XSCALE_7
+0x000284E8 PA_CL_VPORT_XOFFSET_7
+0x000284EC PA_CL_VPORT_YSCALE_7
+0x000284F0 PA_CL_VPORT_YOFFSET_7
+0x000284F4 PA_CL_VPORT_ZSCALE_7
+0x000284F8 PA_CL_VPORT_ZOFFSET_7
+0x000284FC PA_CL_VPORT_XSCALE_8
+0x00028500 PA_CL_VPORT_XOFFSET_8
+0x00028504 PA_CL_VPORT_YSCALE_8
+0x00028508 PA_CL_VPORT_YOFFSET_8
+0x0002850C PA_CL_VPORT_ZSCALE_8
+0x00028510 PA_CL_VPORT_ZOFFSET_8
+0x00028514 PA_CL_VPORT_XSCALE_9
+0x00028518 PA_CL_VPORT_XOFFSET_9
+0x0002851C PA_CL_VPORT_YSCALE_9
+0x00028520 PA_CL_VPORT_YOFFSET_9
+0x00028524 PA_CL_VPORT_ZSCALE_9
+0x00028528 PA_CL_VPORT_ZOFFSET_9
+0x0002852C PA_CL_VPORT_XSCALE_10
+0x00028530 PA_CL_VPORT_XOFFSET_10
+0x00028534 PA_CL_VPORT_YSCALE_10
+0x00028538 PA_CL_VPORT_YOFFSET_10
+0x0002853C PA_CL_VPORT_ZSCALE_10
+0x00028540 PA_CL_VPORT_ZOFFSET_10
+0x00028544 PA_CL_VPORT_XSCALE_11
+0x00028548 PA_CL_VPORT_XOFFSET_11
+0x0002854C PA_CL_VPORT_YSCALE_11
+0x00028550 PA_CL_VPORT_YOFFSET_11
+0x00028554 PA_CL_VPORT_ZSCALE_11
+0x00028558 PA_CL_VPORT_ZOFFSET_11
+0x0002855C PA_CL_VPORT_XSCALE_12
+0x00028560 PA_CL_VPORT_XOFFSET_12
+0x00028564 PA_CL_VPORT_YSCALE_12
+0x00028568 PA_CL_VPORT_YOFFSET_12
+0x0002856C PA_CL_VPORT_ZSCALE_12
+0x00028570 PA_CL_VPORT_ZOFFSET_12
+0x00028574 PA_CL_VPORT_XSCALE_13
+0x00028578 PA_CL_VPORT_XOFFSET_13
+0x0002857C PA_CL_VPORT_YSCALE_13
+0x00028580 PA_CL_VPORT_YOFFSET_13
+0x00028584 PA_CL_VPORT_ZSCALE_13
+0x00028588 PA_CL_VPORT_ZOFFSET_13
+0x0002858C PA_CL_VPORT_XSCALE_14
+0x00028590 PA_CL_VPORT_XOFFSET_14
+0x00028594 PA_CL_VPORT_YSCALE_14
+0x00028598 PA_CL_VPORT_YOFFSET_14
+0x0002859C PA_CL_VPORT_ZSCALE_14
+0x000285A0 PA_CL_VPORT_ZOFFSET_14
+0x000285A4 PA_CL_VPORT_XSCALE_15
+0x000285A8 PA_CL_VPORT_XOFFSET_15
+0x000285AC PA_CL_VPORT_YSCALE_15
+0x000285B0 PA_CL_VPORT_YOFFSET_15
+0x000285B4 PA_CL_VPORT_ZSCALE_15
+0x000285B8 PA_CL_VPORT_ZOFFSET_15
+0x000285BC PA_CL_UCP_0_X
+0x000285C0 PA_CL_UCP_0_Y
+0x000285C4 PA_CL_UCP_0_Z
+0x000285C8 PA_CL_UCP_0_W
+0x000285CC PA_CL_UCP_1_X
+0x000285D0 PA_CL_UCP_1_Y
+0x000285D4 PA_CL_UCP_1_Z
+0x000285D8 PA_CL_UCP_1_W
+0x000285DC PA_CL_UCP_2_X
+0x000285E0 PA_CL_UCP_2_Y
+0x000285E4 PA_CL_UCP_2_Z
+0x000285E8 PA_CL_UCP_2_W
+0x000285EC PA_CL_UCP_3_X
+0x000285F0 PA_CL_UCP_3_Y
+0x000285F4 PA_CL_UCP_3_Z
+0x000285F8 PA_CL_UCP_3_W
+0x000285FC PA_CL_UCP_4_X
+0x00028600 PA_CL_UCP_4_Y
+0x00028604 PA_CL_UCP_4_Z
+0x00028608 PA_CL_UCP_4_W
+0x0002860C PA_CL_UCP_5_X
+0x00028610 PA_CL_UCP_5_Y
+0x00028614 PA_CL_UCP_5_Z
+0x00028618 PA_CL_UCP_5_W
+0x0002861C SPI_VS_OUT_ID_0
+0x00028620 SPI_VS_OUT_ID_1
+0x00028624 SPI_VS_OUT_ID_2
+0x00028628 SPI_VS_OUT_ID_3
+0x0002862C SPI_VS_OUT_ID_4
+0x00028630 SPI_VS_OUT_ID_5
+0x00028634 SPI_VS_OUT_ID_6
+0x00028638 SPI_VS_OUT_ID_7
+0x0002863C SPI_VS_OUT_ID_8
+0x00028640 SPI_VS_OUT_ID_9
+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
+0x000286C4 SPI_VS_OUT_CONFIG
+0x000286C8 SPI_THREAD_GROUPING
+0x000286CC SPI_PS_IN_CONTROL_0
+0x000286D0 SPI_PS_IN_CONTROL_1
+0x000286D4 SPI_INTERP_CONTROL_0
+0x000286D8 SPI_INPUT_Z
+0x000286DC SPI_FOG_CNTL
+0x000286E0 SPI_BARYC_CNTL
+0x000286E4 SPI_PS_IN_CONTROL_2
+0x000286E8 SPI_COMPUTE_INPUT_CNTL
+0x000286EC SPI_COMPUTE_NUM_THREAD_X
+0x000286F0 SPI_COMPUTE_NUM_THREAD_Y
+0x000286F4 SPI_COMPUTE_NUM_THREAD_Z
+0x000286F8 GDS_ADDR_SIZE
+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
+0x000287CC CS_COPY_STATE
+0x000287D0 GFX_COPY_STATE
+0x000287D4 PA_CL_POINT_X_RAD
+0x000287D8 PA_CL_POINT_Y_RAD
+0x000287DC PA_CL_POINT_SIZE
+0x000287E0 PA_CL_POINT_CULL_RAD
+0x00028808 CB_COLOR_CONTROL
+0x0002880C DB_SHADER_CONTROL
+0x00028810 PA_CL_CLIP_CNTL
+0x00028814 PA_SU_SC_MODE_CNTL
+0x00028818 PA_CL_VTE_CNTL
+0x0002881C PA_CL_VS_OUT_CNTL
+0x00028820 PA_CL_NANINF_CNTL
+0x00028824 PA_SU_LINE_STIPPLE_CNTL
+0x00028828 PA_SU_LINE_STIPPLE_SCALE
+0x0002882C PA_SU_PRIM_FILTER_CNTL
+0x00028838 SQ_DYN_GPR_RESOURCE_LIMIT_1
+0x00028844 SQ_PGM_RESOURCES_PS
+0x00028848 SQ_PGM_RESOURCES_2_PS
+0x0002884C SQ_PGM_EXPORTS_PS
+0x0002885C SQ_PGM_RESOURCES_VS
+0x00028860 SQ_PGM_RESOURCES_2_VS
+0x00028878 SQ_PGM_RESOURCES_GS
+0x0002887C SQ_PGM_RESOURCES_2_GS
+0x00028890 SQ_PGM_RESOURCES_ES
+0x00028894 SQ_PGM_RESOURCES_2_ES
+0x000288A8 SQ_PGM_RESOURCES_FS
+0x000288BC SQ_PGM_RESOURCES_HS
+0x000288C0 SQ_PGM_RESOURCES_2_HS
+0x000288D0 SQ_PGM_RESOURCES_LS
+0x000288D4 SQ_PGM_RESOURCES_2_LS
+0x000288E8 SQ_LDS_ALLOC
+0x000288EC SQ_LDS_ALLOC_PS
+0x000288F0 SQ_VTX_SEMANTIC_CLEAR
+0x00028A00 PA_SU_POINT_SIZE
+0x00028A04 PA_SU_POINT_MINMAX
+0x00028A08 PA_SU_LINE_CNTL
+0x00028A0C PA_SC_LINE_STIPPLE
+0x00028A10 VGT_OUTPUT_PATH_CNTL
+0x00028A14 VGT_HOS_CNTL
+0x00028A18 VGT_HOS_MAX_TESS_LEVEL
+0x00028A1C VGT_HOS_MIN_TESS_LEVEL
+0x00028A20 VGT_HOS_REUSE_DEPTH
+0x00028A24 VGT_GROUP_PRIM_TYPE
+0x00028A28 VGT_GROUP_FIRST_DECR
+0x00028A2C VGT_GROUP_DECR
+0x00028A30 VGT_GROUP_VECT_0_CNTL
+0x00028A34 VGT_GROUP_VECT_1_CNTL
+0x00028A38 VGT_GROUP_VECT_0_FMT_CNTL
+0x00028A3C VGT_GROUP_VECT_1_FMT_CNTL
+0x00028A40 VGT_GS_MODE
+0x00028A48 PA_SC_MODE_CNTL_0
+0x00028A4C PA_SC_MODE_CNTL_1
+0x00028A50 VGT_ENHANCE
+0x00028A54 VGT_GS_PER_ES
+0x00028A58 VGT_ES_PER_GS
+0x00028A5C VGT_GS_PER_VS
+0x00028A6C VGT_GS_OUT_PRIM_TYPE
+0x00028A84 VGT_PRIMITIVEID_EN
+0x00028A94 VGT_MULTI_PRIM_IB_RESET_EN
+0x00028AA0 VGT_INSTANCE_STEP_RATE_0
+0x00028AA4 VGT_INSTANCE_STEP_RATE_1
+0x00028AB4 VGT_REUSE_OFF
+0x00028AB8 VGT_VTX_CNT_EN
+0x00028ABC DB_HTILE_SURFACE
+0x00028AC0 DB_SRESULTS_COMPARE_STATE0
+0x00028AC4 DB_SRESULTS_COMPARE_STATE1
+0x00028AC8 DB_PRELOAD_CONTROL
+0x00028B38 VGT_GS_MAX_VERT_OUT
+0x00028B54 VGT_SHADER_STAGES_EN
+0x00028B58 VGT_LS_HS_CONFIG
+0x00028B5C VGT_LS_SIZE
+0x00028B60 VGT_HS_SIZE
+0x00028B64 VGT_LS_HS_ALLOC
+0x00028B68 VGT_HS_PATCH_CONST
+0x00028B6C VGT_TF_PARAM
+0x00028B70 DB_ALPHA_TO_MASK
+0x00028B74 VGT_DISPATCH_INITIATOR
+0x00028B78 PA_SU_POLY_OFFSET_DB_FMT_CNTL
+0x00028B7C PA_SU_POLY_OFFSET_CLAMP
+0x00028B80 PA_SU_POLY_OFFSET_FRONT_SCALE
+0x00028B84 PA_SU_POLY_OFFSET_FRONT_OFFSET
+0x00028B88 PA_SU_POLY_OFFSET_BACK_SCALE
+0x00028B8C PA_SU_POLY_OFFSET_BACK_OFFSET
+0x00028B74 VGT_GS_INSTANCE_CNT
+0x00028C00 PA_SC_LINE_CNTL
+0x00028C08 PA_SU_VTX_CNTL
+0x00028C0C PA_CL_GB_VERT_CLIP_ADJ
+0x00028C10 PA_CL_GB_VERT_DISC_ADJ
+0x00028C14 PA_CL_GB_HORZ_CLIP_ADJ
+0x00028C18 PA_CL_GB_HORZ_DISC_ADJ
+0x00028C1C PA_SC_AA_SAMPLE_LOCS_0
+0x00028C20 PA_SC_AA_SAMPLE_LOCS_1
+0x00028C24 PA_SC_AA_SAMPLE_LOCS_2
+0x00028C28 PA_SC_AA_SAMPLE_LOCS_3
+0x00028C2C PA_SC_AA_SAMPLE_LOCS_4
+0x00028C30 PA_SC_AA_SAMPLE_LOCS_5
+0x00028C34 PA_SC_AA_SAMPLE_LOCS_6
+0x00028C38 PA_SC_AA_SAMPLE_LOCS_7
+0x00028C3C PA_SC_AA_MASK
+0x00028C8C CB_COLOR0_CLEAR_WORD0
+0x00028C90 CB_COLOR0_CLEAR_WORD1
+0x00028C94 CB_COLOR0_CLEAR_WORD2
+0x00028C98 CB_COLOR0_CLEAR_WORD3
+0x00028CC8 CB_COLOR1_CLEAR_WORD0
+0x00028CCC CB_COLOR1_CLEAR_WORD1
+0x00028CD0 CB_COLOR1_CLEAR_WORD2
+0x00028CD4 CB_COLOR1_CLEAR_WORD3
+0x00028D04 CB_COLOR2_CLEAR_WORD0
+0x00028D08 CB_COLOR2_CLEAR_WORD1
+0x00028D0C CB_COLOR2_CLEAR_WORD2
+0x00028D10 CB_COLOR2_CLEAR_WORD3
+0x00028D40 CB_COLOR3_CLEAR_WORD0
+0x00028D44 CB_COLOR3_CLEAR_WORD1
+0x00028D48 CB_COLOR3_CLEAR_WORD2
+0x00028D4C CB_COLOR3_CLEAR_WORD3
+0x00028D7C CB_COLOR4_CLEAR_WORD0
+0x00028D80 CB_COLOR4_CLEAR_WORD1
+0x00028D84 CB_COLOR4_CLEAR_WORD2
+0x00028D88 CB_COLOR4_CLEAR_WORD3
+0x00028DB8 CB_COLOR5_CLEAR_WORD0
+0x00028DBC CB_COLOR5_CLEAR_WORD1
+0x00028DC0 CB_COLOR5_CLEAR_WORD2
+0x00028DC4 CB_COLOR5_CLEAR_WORD3
+0x00028DF4 CB_COLOR6_CLEAR_WORD0
+0x00028DF8 CB_COLOR6_CLEAR_WORD1
+0x00028DFC CB_COLOR6_CLEAR_WORD2
+0x00028E00 CB_COLOR6_CLEAR_WORD3
+0x00028E30 CB_COLOR7_CLEAR_WORD0
+0x00028E34 CB_COLOR7_CLEAR_WORD1
+0x00028E38 CB_COLOR7_CLEAR_WORD2
+0x00028E3C CB_COLOR7_CLEAR_WORD3
+0x00028F80 SQ_ALU_CONST_BUFFER_SIZE_HS_0
+0x00028F84 SQ_ALU_CONST_BUFFER_SIZE_HS_1
+0x00028F88 SQ_ALU_CONST_BUFFER_SIZE_HS_2
+0x00028F8C SQ_ALU_CONST_BUFFER_SIZE_HS_3
+0x00028F90 SQ_ALU_CONST_BUFFER_SIZE_HS_4
+0x00028F94 SQ_ALU_CONST_BUFFER_SIZE_HS_5
+0x00028F98 SQ_ALU_CONST_BUFFER_SIZE_HS_6
+0x00028F9C SQ_ALU_CONST_BUFFER_SIZE_HS_7
+0x00028FA0 SQ_ALU_CONST_BUFFER_SIZE_HS_8
+0x00028FA4 SQ_ALU_CONST_BUFFER_SIZE_HS_9
+0x00028FA8 SQ_ALU_CONST_BUFFER_SIZE_HS_10
+0x00028FAC SQ_ALU_CONST_BUFFER_SIZE_HS_11
+0x00028FB0 SQ_ALU_CONST_BUFFER_SIZE_HS_12
+0x00028FB4 SQ_ALU_CONST_BUFFER_SIZE_HS_13
+0x00028FB8 SQ_ALU_CONST_BUFFER_SIZE_HS_14
+0x00028FBC SQ_ALU_CONST_BUFFER_SIZE_HS_15
+0x00028FC0 SQ_ALU_CONST_BUFFER_SIZE_LS_0
+0x00028FC4 SQ_ALU_CONST_BUFFER_SIZE_LS_1
+0x00028FC8 SQ_ALU_CONST_BUFFER_SIZE_LS_2
+0x00028FCC SQ_ALU_CONST_BUFFER_SIZE_LS_3
+0x00028FD0 SQ_ALU_CONST_BUFFER_SIZE_LS_4
+0x00028FD4 SQ_ALU_CONST_BUFFER_SIZE_LS_5
+0x00028FD8 SQ_ALU_CONST_BUFFER_SIZE_LS_6
+0x00028FDC SQ_ALU_CONST_BUFFER_SIZE_LS_7
+0x00028FE0 SQ_ALU_CONST_BUFFER_SIZE_LS_8
+0x00028FE4 SQ_ALU_CONST_BUFFER_SIZE_LS_9
+0x00028FE8 SQ_ALU_CONST_BUFFER_SIZE_LS_10
+0x00028FEC SQ_ALU_CONST_BUFFER_SIZE_LS_11
+0x00028FF0 SQ_ALU_CONST_BUFFER_SIZE_LS_12
+0x00028FF4 SQ_ALU_CONST_BUFFER_SIZE_LS_13
+0x00028FF8 SQ_ALU_CONST_BUFFER_SIZE_LS_14
+0x00028FFC SQ_ALU_CONST_BUFFER_SIZE_LS_15
+0x0003CFF0 SQ_VTX_BASE_VTX_LOC
+0x0003CFF4 SQ_VTX_START_INST_LOC
+0x0003FF00 SQ_TEX_SAMPLER_CLEAR
+0x0003FF04 SQ_TEX_RESOURCE_CLEAR
+0x0003FF08 SQ_LOOP_BOOL_CLEAR
diff --git a/drivers/gpu/drm/radeon/rs600.c b/drivers/gpu/drm/radeon/rs600.c
index 79887cac5b54..7bb4c3e52f3b 100644
--- a/drivers/gpu/drm/radeon/rs600.c
+++ b/drivers/gpu/drm/radeon/rs600.c
@@ -74,7 +74,8 @@ void rs600_pm_misc(struct radeon_device *rdev)
 			if (voltage->delay)
 				udelay(voltage->delay);
 		}
-	}
+	} else if (voltage->type == VOLTAGE_VDDC)
+		radeon_atom_set_voltage(rdev, voltage->vddc_id);
 
 	dyn_pwrmgt_sclk_length = RREG32_PLL(DYN_PWRMGT_SCLK_LENGTH);
 	dyn_pwrmgt_sclk_length &= ~REDUCED_POWER_SCLK_HILEN(0xf);
diff --git a/drivers/gpu/drm/radeon/rv770.c b/drivers/gpu/drm/radeon/rv770.c
index 253f24aec031..cec536c222c5 100644
--- a/drivers/gpu/drm/radeon/rv770.c
+++ b/drivers/gpu/drm/radeon/rv770.c
@@ -44,7 +44,18 @@ void rv770_fini(struct radeon_device *rdev);
 
 void rv770_pm_misc(struct radeon_device *rdev)
 {
-
+	int req_ps_idx = rdev->pm.requested_power_state_index;
+	int req_cm_idx = rdev->pm.requested_clock_mode_index;
+	struct radeon_power_state *ps = &rdev->pm.power_state[req_ps_idx];
+	struct radeon_voltage *voltage = &ps->clock_info[req_cm_idx].voltage;
+
+	if ((voltage->type == VOLTAGE_SW) && voltage->voltage) {
+		if (voltage->voltage != rdev->pm.current_vddc) {
+			radeon_atom_set_voltage(rdev, voltage->voltage);
+			rdev->pm.current_vddc = voltage->voltage;
+			DRM_DEBUG("Setting: v: %d\n", voltage->voltage);
+		}
+	}
 }
 
 /*
diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc.c b/drivers/gpu/drm/ttm/ttm_page_alloc.c
index 0d9a42c2394f..ef910694bd63 100644
--- a/drivers/gpu/drm/ttm/ttm_page_alloc.c
+++ b/drivers/gpu/drm/ttm/ttm_page_alloc.c
@@ -77,7 +77,7 @@ struct ttm_page_pool {
 /**
  * Limits for the pool. They are handled without locks because only place where
  * they may change is in sysfs store. They won't have immediate effect anyway
- * so forcing serialiazation to access them is pointless.
+ * so forcing serialization to access them is pointless.
  */
 
 struct ttm_pool_opts {
@@ -165,16 +165,18 @@ static ssize_t ttm_pool_store(struct kobject *kobj,
 		m->options.small = val;
 	else if (attr == &ttm_page_pool_alloc_size) {
 		if (val > NUM_PAGES_TO_ALLOC*8) {
-			printk(KERN_ERR "[ttm] Setting allocation size to %lu "
-					"is not allowed. Recomended size is "
-					"%lu\n",
-					NUM_PAGES_TO_ALLOC*(PAGE_SIZE >> 7),
-					NUM_PAGES_TO_ALLOC*(PAGE_SIZE >> 10));
+			printk(KERN_ERR TTM_PFX
+			       "Setting allocation size to %lu "
+			       "is not allowed. Recommended size is "
+			       "%lu\n",
+			       NUM_PAGES_TO_ALLOC*(PAGE_SIZE >> 7),
+			       NUM_PAGES_TO_ALLOC*(PAGE_SIZE >> 10));
 			return size;
 		} else if (val > NUM_PAGES_TO_ALLOC) {
-			printk(KERN_WARNING "[ttm] Setting allocation size to "
-					"larger than %lu is not recomended.\n",
-					NUM_PAGES_TO_ALLOC*(PAGE_SIZE >> 10));
+			printk(KERN_WARNING TTM_PFX
+			       "Setting allocation size to "
+			       "larger than %lu is not recommended.\n",
+			       NUM_PAGES_TO_ALLOC*(PAGE_SIZE >> 10));
 		}
 		m->options.alloc_size = val;
 	}
@@ -277,7 +279,7 @@ static void ttm_pages_put(struct page *pages[], unsigned npages)
 {
 	unsigned i;
 	if (set_pages_array_wb(pages, npages))
-		printk(KERN_ERR "[ttm] Failed to set %d pages to wb!\n",
+		printk(KERN_ERR TTM_PFX "Failed to set %d pages to wb!\n",
 				npages);
 	for (i = 0; i < npages; ++i)
 		__free_page(pages[i]);
@@ -313,7 +315,8 @@ static int ttm_page_pool_free(struct ttm_page_pool *pool, unsigned nr_free)
 	pages_to_free = kmalloc(npages_to_free * sizeof(struct page *),
 			GFP_KERNEL);
 	if (!pages_to_free) {
-		printk(KERN_ERR "Failed to allocate memory for pool free operation.\n");
+		printk(KERN_ERR TTM_PFX
+		       "Failed to allocate memory for pool free operation.\n");
 		return 0;
 	}
 
@@ -390,7 +393,7 @@ static int ttm_pool_get_num_unused_pages(void)
 }
 
 /**
- * Calback for mm to request pool to reduce number of page held.
+ * Callback for mm to request pool to reduce number of page held.
  */
 static int ttm_pool_mm_shrink(int shrink_pages, gfp_t gfp_mask)
 {
@@ -433,14 +436,16 @@ static int ttm_set_pages_caching(struct page **pages,
 	case tt_uncached:
 		r = set_pages_array_uc(pages, cpages);
 		if (r)
-			printk(KERN_ERR "[ttm] Failed to set %d pages to uc!\n",
-					cpages);
+			printk(KERN_ERR TTM_PFX
+			       "Failed to set %d pages to uc!\n",
+			       cpages);
 		break;
 	case tt_wc:
 		r = set_pages_array_wc(pages, cpages);
 		if (r)
-			printk(KERN_ERR "[ttm] Failed to set %d pages to wc!\n",
-					cpages);
+			printk(KERN_ERR TTM_PFX
+			       "Failed to set %d pages to wc!\n",
+			       cpages);
 		break;
 	default:
 		break;
@@ -458,7 +463,7 @@ static void ttm_handle_caching_state_failure(struct list_head *pages,
 		struct page **failed_pages, unsigned cpages)
 {
 	unsigned i;
-	/* Failed pages has to be reed */
+	/* Failed pages have to be freed */
 	for (i = 0; i < cpages; ++i) {
 		list_del(&failed_pages[i]->lru);
 		__free_page(failed_pages[i]);
@@ -485,7 +490,8 @@ static int ttm_alloc_new_pages(struct list_head *pages, int gfp_flags,
 	caching_array = kmalloc(max_cpages*sizeof(struct page *), GFP_KERNEL);
 
 	if (!caching_array) {
-		printk(KERN_ERR "[ttm] unable to allocate table for new pages.");
+		printk(KERN_ERR TTM_PFX
+		       "Unable to allocate table for new pages.");
 		return -ENOMEM;
 	}
 
@@ -493,12 +499,13 @@ static int ttm_alloc_new_pages(struct list_head *pages, int gfp_flags,
 		p = alloc_page(gfp_flags);
 
 		if (!p) {
-			printk(KERN_ERR "[ttm] unable to get page %u\n", i);
+			printk(KERN_ERR TTM_PFX "Unable to get page %u.\n", i);
 
 			/* store already allocated pages in the pool after
 			 * setting the caching state */
 			if (cpages) {
-				r = ttm_set_pages_caching(caching_array, cstate, cpages);
+				r = ttm_set_pages_caching(caching_array,
+							  cstate, cpages);
 				if (r)
 					ttm_handle_caching_state_failure(pages,
 						ttm_flags, cstate,
@@ -590,7 +597,8 @@ static void ttm_page_pool_fill_locked(struct ttm_page_pool *pool,
 			++pool->nrefills;
 			pool->npages += alloc_size;
 		} else {
-			printk(KERN_ERR "[ttm] Failed to fill pool (%p).", pool);
+			printk(KERN_ERR TTM_PFX
+			       "Failed to fill pool (%p).", pool);
 			/* If we have any pages left put them to the pool. */
 			list_for_each_entry(p, &pool->list, lru) {
 				++cpages;
@@ -671,13 +679,14 @@ int ttm_get_pages(struct list_head *pages, int flags,
 		if (flags & TTM_PAGE_FLAG_DMA32)
 			gfp_flags |= GFP_DMA32;
 		else
-			gfp_flags |= __GFP_HIGHMEM;
+			gfp_flags |= GFP_HIGHUSER;
 
 		for (r = 0; r < count; ++r) {
 			p = alloc_page(gfp_flags);
 			if (!p) {
 
-				printk(KERN_ERR "[ttm] unable to allocate page.");
+				printk(KERN_ERR TTM_PFX
+				       "Unable to allocate page.");
 				return -ENOMEM;
 			}
 
@@ -709,8 +718,9 @@ int ttm_get_pages(struct list_head *pages, int flags,
 		if (r) {
 			/* If there is any pages in the list put them back to
 			 * the pool. */
-			printk(KERN_ERR "[ttm] Failed to allocate extra pages "
-					"for large request.");
+			printk(KERN_ERR TTM_PFX
+			       "Failed to allocate extra pages "
+			       "for large request.");
 			ttm_put_pages(pages, 0, flags, cstate);
 			return r;
 		}
@@ -778,7 +788,7 @@ int ttm_page_alloc_init(struct ttm_mem_global *glob, unsigned max_pages)
 	if (atomic_add_return(1, &_manager.page_alloc_inited) > 1)
 		return 0;
 
-	printk(KERN_INFO "[ttm] Initializing pool allocator.\n");
+	printk(KERN_INFO TTM_PFX "Initializing pool allocator.\n");
 
 	ttm_page_pool_init_locked(&_manager.wc_pool, GFP_HIGHUSER, "wc");
 
@@ -813,7 +823,7 @@ void ttm_page_alloc_fini()
 	if (atomic_sub_return(1, &_manager.page_alloc_inited) > 0)
 		return;
 
-	printk(KERN_INFO "[ttm] Finilizing pool allocator.\n");
+	printk(KERN_INFO TTM_PFX "Finalizing pool allocator.\n");
 	ttm_pool_mm_shrink_fini(&_manager);
 
 	for (i = 0; i < NUM_POOLS; ++i)
diff --git a/drivers/gpu/drm/vmwgfx/Makefile b/drivers/gpu/drm/vmwgfx/Makefile
index 1a3cb6816d1c..4505e17df3f5 100644
--- a/drivers/gpu/drm/vmwgfx/Makefile
+++ b/drivers/gpu/drm/vmwgfx/Makefile
@@ -4,6 +4,6 @@ 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
+	    vmwgfx_overlay.o vmwgfx_fence.o
 
 obj-$(CONFIG_DRM_VMWGFX) := vmwgfx.o
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
index 0c9c0811f42d..b793c8c9acb3 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
@@ -88,6 +88,9 @@
 #define DRM_IOCTL_VMW_FENCE_WAIT				\
 	DRM_IOWR(DRM_COMMAND_BASE + DRM_VMW_FENCE_WAIT,		\
 		 struct drm_vmw_fence_wait_arg)
+#define DRM_IOCTL_VMW_UPDATE_LAYOUT				\
+	DRM_IOWR(DRM_COMMAND_BASE + DRM_VMW_UPDATE_LAYOUT,	\
+		 struct drm_vmw_update_layout_arg)
 
 
 /**
@@ -135,7 +138,9 @@ static struct drm_ioctl_desc vmw_ioctls[] = {
 	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)
+		      DRM_AUTH | DRM_UNLOCKED),
+	VMW_IOCTL_DEF(DRM_IOCTL_VMW_UPDATE_LAYOUT, vmw_kms_update_layout_ioctl,
+		      DRM_MASTER | DRM_CONTROL_ALLOW | DRM_UNLOCKED)
 };
 
 static struct pci_device_id vmw_pci_id_list[] = {
@@ -318,6 +323,15 @@ static int vmw_driver_load(struct drm_device *dev, unsigned long chipset)
 		goto out_err3;
 	}
 
+	/* Need mmio memory to check for fifo pitchlock cap. */
+	if (!(dev_priv->capabilities & SVGA_CAP_DISPLAY_TOPOLOGY) &&
+	    !(dev_priv->capabilities & SVGA_CAP_PITCHLOCK) &&
+	    !vmw_fifo_have_pitchlock(dev_priv)) {
+		ret = -ENOSYS;
+		DRM_ERROR("Hardware has no pitchlock\n");
+		goto out_err4;
+	}
+
 	dev_priv->tdev = ttm_object_device_init
 	    (dev_priv->mem_global_ref.object, 12);
 
@@ -399,8 +413,6 @@ 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);
@@ -546,7 +558,6 @@ static int vmw_master_create(struct drm_device *dev,
 {
 	struct vmw_master *vmaster;
 
-	DRM_INFO("Master create.\n");
 	vmaster = kzalloc(sizeof(*vmaster), GFP_KERNEL);
 	if (unlikely(vmaster == NULL))
 		return -ENOMEM;
@@ -563,7 +574,6 @@ static void vmw_master_destroy(struct drm_device *dev,
 {
 	struct vmw_master *vmaster = vmw_master(master);
 
-	DRM_INFO("Master destroy.\n");
 	master->driver_priv = NULL;
 	kfree(vmaster);
 }
@@ -579,8 +589,6 @@ static int vmw_master_set(struct drm_device *dev,
 	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);
@@ -622,8 +630,6 @@ static void vmw_master_drop(struct drm_device *dev,
 	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.
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
index 356dc935ec13..eaad52095339 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
@@ -41,12 +41,13 @@
 
 #define VMWGFX_DRIVER_DATE "20100209"
 #define VMWGFX_DRIVER_MAJOR 1
-#define VMWGFX_DRIVER_MINOR 0
+#define VMWGFX_DRIVER_MINOR 2
 #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
+#define VMWGFX_MAX_DISPLAYS 16
 
 struct vmw_fpriv {
 	struct drm_master *locked_master;
@@ -102,6 +103,13 @@ struct vmw_surface {
 	struct vmw_cursor_snooper snooper;
 };
 
+struct vmw_fence_queue {
+	struct list_head head;
+	struct timespec lag;
+	struct timespec lag_time;
+	spinlock_t lock;
+};
+
 struct vmw_fifo_state {
 	unsigned long reserved_size;
 	__le32 *dynamic_buffer;
@@ -115,6 +123,7 @@ struct vmw_fifo_state {
 	uint32_t capabilities;
 	struct mutex fifo_mutex;
 	struct rw_semaphore rwsem;
+	struct vmw_fence_queue fence_queue;
 };
 
 struct vmw_relocation {
@@ -144,6 +153,14 @@ struct vmw_master {
 	struct ttm_lock lock;
 };
 
+struct vmw_vga_topology_state {
+	uint32_t width;
+	uint32_t height;
+	uint32_t primary;
+	uint32_t pos_x;
+	uint32_t pos_y;
+};
+
 struct vmw_private {
 	struct ttm_bo_device bdev;
 	struct ttm_bo_global_ref bo_global_ref;
@@ -171,14 +188,19 @@ struct vmw_private {
 	 * VGA registers.
 	 */
 
+	struct vmw_vga_topology_state vga_save[VMWGFX_MAX_DISPLAYS];
 	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;
+	uint32_t vga_blue_mask;
+	uint32_t vga_bpl;
+	uint32_t vga_pitchlock;
+
+	uint32_t num_displays;
 
 	/*
 	 * Framebuffer info.
@@ -393,6 +415,7 @@ extern int vmw_fifo_send_fence(struct vmw_private *dev_priv,
 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);
+extern bool vmw_fifo_have_pitchlock(struct vmw_private *dev_priv);
 
 /**
  * TTM glue - vmwgfx_ttm_glue.c
@@ -441,6 +464,23 @@ extern int vmw_fallback_wait(struct vmw_private *dev_priv,
 			     uint32_t sequence,
 			     bool interruptible,
 			     unsigned long timeout);
+extern void vmw_update_sequence(struct vmw_private *dev_priv,
+				struct vmw_fifo_state *fifo_state);
+
+
+/**
+ * Rudimentary fence objects currently used only for throttling -
+ * vmwgfx_fence.c
+ */
+
+extern void vmw_fence_queue_init(struct vmw_fence_queue *queue);
+extern void vmw_fence_queue_takedown(struct vmw_fence_queue *queue);
+extern int vmw_fence_push(struct vmw_fence_queue *queue,
+			  uint32_t sequence);
+extern int vmw_fence_pull(struct vmw_fence_queue *queue,
+			  uint32_t signaled_sequence);
+extern int vmw_wait_lag(struct vmw_private *dev_priv,
+			struct vmw_fence_queue *queue, uint32_t us);
 
 /**
  * Kernel framebuffer - vmwgfx_fb.c
@@ -466,6 +506,11 @@ void vmw_kms_cursor_snoop(struct vmw_surface *srf,
 			  struct ttm_object_file *tfile,
 			  struct ttm_buffer_object *bo,
 			  SVGA3dCmdHeader *header);
+void vmw_kms_write_svga(struct vmw_private *vmw_priv,
+			unsigned width, unsigned height, unsigned pitch,
+			unsigned bbp, unsigned depth);
+int vmw_kms_update_layout_ioctl(struct drm_device *dev, void *data,
+				struct drm_file *file_priv);
 
 /**
  * Overlay control - vmwgfx_overlay.c
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c
index dbd36b8910cf..8e396850513c 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c
@@ -644,6 +644,7 @@ int vmw_execbuf_ioctl(struct drm_device *dev, void *data,
 	ret = copy_from_user(cmd, user_cmd, arg->command_size);
 
 	if (unlikely(ret != 0)) {
+		ret = -EFAULT;
 		DRM_ERROR("Failed copying commands.\n");
 		goto out_commit;
 	}
@@ -669,6 +670,15 @@ int vmw_execbuf_ioctl(struct drm_device *dev, void *data,
 		goto out_err;
 
 	vmw_apply_relocations(sw_context);
+
+	if (arg->throttle_us) {
+		ret = vmw_wait_lag(dev_priv, &dev_priv->fifo.fence_queue,
+				   arg->throttle_us);
+
+		if (unlikely(ret != 0))
+			goto out_err;
+	}
+
 	vmw_fifo_commit(dev_priv, arg->command_size);
 
 	ret = vmw_fifo_send_fence(dev_priv, &sequence);
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_fb.c b/drivers/gpu/drm/vmwgfx/vmwgfx_fb.c
index 7421aaad8d09..b0866f04ec76 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_fb.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_fb.c
@@ -132,16 +132,14 @@ static int vmw_fb_check_var(struct fb_var_screeninfo *var,
 		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");
+	if (!(vmw_priv->capabilities & SVGA_CAP_DISPLAY_TOPOLOGY) &&
+	    (var->xoffset != 0 || var->yoffset != 0)) {
+		DRM_ERROR("Can not handle panning without display topology\n");
 		return -EINVAL;
 	}
 
-	if (var->xres > par->max_width ||
-	    var->yres > par->max_height) {
+	if ((var->xoffset + var->xres) > par->max_width ||
+	    (var->yoffset + var->yres) > par->max_height) {
 		DRM_ERROR("Requested geom can not fit in framebuffer\n");
 		return -EINVAL;
 	}
@@ -154,27 +152,11 @@ 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);
-
+	vmw_kms_write_svga(vmw_priv, info->var.xres, info->var.yres,
+			   info->fix.line_length,
+			   par->bpp, par->depth);
+	if (vmw_priv->capabilities & SVGA_CAP_DISPLAY_TOPOLOGY) {
 		/* 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);
@@ -183,13 +165,13 @@ static int vmw_fb_set_par(struct fb_info *info)
 		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 */
 	}
 
+	/* This is really helpful since if this fails the user
+	 * can probably not see anything on the screen.
+	 */
+	WARN_ON(vmw_read(vmw_priv, SVGA_REG_FB_OFFSET) != 0);
+
 	return 0;
 }
 
@@ -416,48 +398,23 @@ int vmw_fb_init(struct vmw_private *vmw_priv)
 	unsigned fb_bbp, fb_depth, fb_offset, fb_pitch, fb_size;
 	int ret;
 
+	/* XXX These shouldn't be hardcoded. */
 	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);
-	}
+	/* XXX As shouldn't these be as well. */
+	fb_width = min(vmw_priv->fb_max_width, (unsigned)2048);
+	fb_height = min(vmw_priv->fb_max_height, (unsigned)2048);
 
 	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_pitch = fb_width * fb_bbp / 8;
+	fb_size = fb_pitch * fb_height;
 	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)
@@ -659,6 +616,10 @@ int vmw_dmabuf_to_start_of_vram(struct vmw_private *vmw_priv,
 		goto err_unlock;
 
 	ret = ttm_bo_validate(bo, &ne_placement, false, false, false);
+
+	/* Could probably bug on */
+	WARN_ON(bo->offset != 0);
+
 	ttm_bo_unreserve(bo);
 err_unlock:
 	ttm_write_unlock(&vmw_priv->active_master->lock);
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c b/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c
new file mode 100644
index 000000000000..61eacc1b5ca3
--- /dev/null
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c
@@ -0,0 +1,173 @@
+/**************************************************************************
+ *
+ * Copyright (C) 2010 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"
+
+struct vmw_fence {
+	struct list_head head;
+	uint32_t sequence;
+	struct timespec submitted;
+};
+
+void vmw_fence_queue_init(struct vmw_fence_queue *queue)
+{
+	INIT_LIST_HEAD(&queue->head);
+	queue->lag = ns_to_timespec(0);
+	getrawmonotonic(&queue->lag_time);
+	spin_lock_init(&queue->lock);
+}
+
+void vmw_fence_queue_takedown(struct vmw_fence_queue *queue)
+{
+	struct vmw_fence *fence, *next;
+
+	spin_lock(&queue->lock);
+	list_for_each_entry_safe(fence, next, &queue->head, head) {
+		kfree(fence);
+	}
+	spin_unlock(&queue->lock);
+}
+
+int vmw_fence_push(struct vmw_fence_queue *queue,
+		   uint32_t sequence)
+{
+	struct vmw_fence *fence = kmalloc(sizeof(*fence), GFP_KERNEL);
+
+	if (unlikely(!fence))
+		return -ENOMEM;
+
+	fence->sequence = sequence;
+	getrawmonotonic(&fence->submitted);
+	spin_lock(&queue->lock);
+	list_add_tail(&fence->head, &queue->head);
+	spin_unlock(&queue->lock);
+
+	return 0;
+}
+
+int vmw_fence_pull(struct vmw_fence_queue *queue,
+		   uint32_t signaled_sequence)
+{
+	struct vmw_fence *fence, *next;
+	struct timespec now;
+	bool updated = false;
+
+	spin_lock(&queue->lock);
+	getrawmonotonic(&now);
+
+	if (list_empty(&queue->head)) {
+		queue->lag = ns_to_timespec(0);
+		queue->lag_time = now;
+		updated = true;
+		goto out_unlock;
+	}
+
+	list_for_each_entry_safe(fence, next, &queue->head, head) {
+		if (signaled_sequence - fence->sequence > (1 << 30))
+			continue;
+
+		queue->lag = timespec_sub(now, fence->submitted);
+		queue->lag_time = now;
+		updated = true;
+		list_del(&fence->head);
+		kfree(fence);
+	}
+
+out_unlock:
+	spin_unlock(&queue->lock);
+
+	return (updated) ? 0 : -EBUSY;
+}
+
+static struct timespec vmw_timespec_add(struct timespec t1,
+					struct timespec t2)
+{
+	t1.tv_sec += t2.tv_sec;
+	t1.tv_nsec += t2.tv_nsec;
+	if (t1.tv_nsec >= 1000000000L) {
+		t1.tv_sec += 1;
+		t1.tv_nsec -= 1000000000L;
+	}
+
+	return t1;
+}
+
+static struct timespec vmw_fifo_lag(struct vmw_fence_queue *queue)
+{
+	struct timespec now;
+
+	spin_lock(&queue->lock);
+	getrawmonotonic(&now);
+	queue->lag = vmw_timespec_add(queue->lag,
+				      timespec_sub(now, queue->lag_time));
+	queue->lag_time = now;
+	spin_unlock(&queue->lock);
+	return queue->lag;
+}
+
+
+static bool vmw_lag_lt(struct vmw_fence_queue *queue,
+		       uint32_t us)
+{
+	struct timespec lag, cond;
+
+	cond = ns_to_timespec((s64) us * 1000);
+	lag = vmw_fifo_lag(queue);
+	return (timespec_compare(&lag, &cond) < 1);
+}
+
+int vmw_wait_lag(struct vmw_private *dev_priv,
+		 struct vmw_fence_queue *queue, uint32_t us)
+{
+	struct vmw_fence *fence;
+	uint32_t sequence;
+	int ret;
+
+	while (!vmw_lag_lt(queue, us)) {
+		spin_lock(&queue->lock);
+		if (list_empty(&queue->head))
+			sequence = atomic_read(&dev_priv->fence_seq);
+		else {
+			fence = list_first_entry(&queue->head,
+						 struct vmw_fence, head);
+			sequence = fence->sequence;
+		}
+		spin_unlock(&queue->lock);
+
+		ret = vmw_wait_fence(dev_priv, false, sequence, true,
+				     3*HZ);
+
+		if (unlikely(ret != 0))
+			return ret;
+
+		(void) vmw_fence_pull(queue, sequence);
+	}
+	return 0;
+}
+
+
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_fifo.c b/drivers/gpu/drm/vmwgfx/vmwgfx_fifo.c
index 39d43a01d846..e6a1eb7ea954 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_fifo.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_fifo.c
@@ -34,6 +34,9 @@ bool vmw_fifo_have_3d(struct vmw_private *dev_priv)
 	__le32 __iomem *fifo_mem = dev_priv->mmio_virt;
 	uint32_t fifo_min, hwversion;
 
+	if (!(dev_priv->capabilities & SVGA_CAP_EXTENDED_FIFO))
+		return false;
+
 	fifo_min = ioread32(fifo_mem  + SVGA_FIFO_MIN);
 	if (fifo_min <= SVGA_FIFO_3D_HWVERSION * sizeof(unsigned int))
 		return false;
@@ -48,6 +51,21 @@ bool vmw_fifo_have_3d(struct vmw_private *dev_priv)
 	return true;
 }
 
+bool vmw_fifo_have_pitchlock(struct vmw_private *dev_priv)
+{
+	__le32 __iomem *fifo_mem = dev_priv->mmio_virt;
+	uint32_t caps;
+
+	if (!(dev_priv->capabilities & SVGA_CAP_EXTENDED_FIFO))
+		return false;
+
+	caps = ioread32(fifo_mem + SVGA_FIFO_CAPABILITIES);
+	if (caps & SVGA_FIFO_CAP_PITCHLOCK)
+		return true;
+
+	return false;
+}
+
 int vmw_fifo_init(struct vmw_private *dev_priv, struct vmw_fifo_state *fifo)
 {
 	__le32 __iomem *fifo_mem = dev_priv->mmio_virt;
@@ -120,7 +138,7 @@ int vmw_fifo_init(struct vmw_private *dev_priv, struct vmw_fifo_state *fifo)
 
 	atomic_set(&dev_priv->fence_seq, dev_priv->last_read_sequence);
 	iowrite32(dev_priv->last_read_sequence, fifo_mem + SVGA_FIFO_FENCE);
-
+	vmw_fence_queue_init(&fifo->fence_queue);
 	return vmw_fifo_send_fence(dev_priv, &dummy);
 out_err:
 	vfree(fifo->static_buffer);
@@ -159,6 +177,7 @@ void vmw_fifo_release(struct vmw_private *dev_priv, struct vmw_fifo_state *fifo)
 		  dev_priv->enable_state);
 
 	mutex_unlock(&dev_priv->hw_mutex);
+	vmw_fence_queue_takedown(&fifo->fence_queue);
 
 	if (likely(fifo->last_buffer != NULL)) {
 		vfree(fifo->last_buffer);
@@ -484,6 +503,8 @@ int vmw_fifo_send_fence(struct vmw_private *dev_priv, uint32_t *sequence)
 	fifo_state->last_buffer_add = true;
 	vmw_fifo_commit(dev_priv, bytes);
 	fifo_state->last_buffer_add = false;
+	(void) vmw_fence_push(&fifo_state->fence_queue, *sequence);
+	vmw_update_sequence(dev_priv, fifo_state);
 
 out_err:
 	return ret;
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_irq.c b/drivers/gpu/drm/vmwgfx/vmwgfx_irq.c
index 4d7cb5393860..e92298a6a383 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_irq.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_irq.c
@@ -64,22 +64,33 @@ static bool vmw_fifo_idle(struct vmw_private *dev_priv, uint32_t sequence)
 	return (busy == 0);
 }
 
+void vmw_update_sequence(struct vmw_private *dev_priv,
+			 struct vmw_fifo_state *fifo_state)
+{
+	__le32 __iomem *fifo_mem = dev_priv->mmio_virt;
+
+	uint32_t sequence = ioread32(fifo_mem + SVGA_FIFO_FENCE);
+
+	if (dev_priv->last_read_sequence != sequence) {
+		dev_priv->last_read_sequence = sequence;
+		vmw_fence_pull(&fifo_state->fence_queue, sequence);
+	}
+}
 
 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);
+	fifo_state = &dev_priv->fifo;
+	vmw_update_sequence(dev_priv, fifo_state);
 	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;
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c
index bbc7c4c30bc7..f1d626112415 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c
@@ -30,6 +30,8 @@
 /* Might need a hrtimer here? */
 #define VMWGFX_PRESENT_RATE ((HZ / 60 > 0) ? HZ / 60 : 1)
 
+static int vmw_surface_dmabuf_pin(struct vmw_framebuffer *vfb);
+static int vmw_surface_dmabuf_unpin(struct vmw_framebuffer *vfb);
 
 void vmw_display_unit_cleanup(struct vmw_display_unit *du)
 {
@@ -326,6 +328,7 @@ int vmw_framebuffer_create_handle(struct drm_framebuffer *fb,
 struct vmw_framebuffer_surface {
 	struct vmw_framebuffer base;
 	struct vmw_surface *surface;
+	struct vmw_dma_buffer *buffer;
 	struct delayed_work d_work;
 	struct mutex work_lock;
 	bool present_fs;
@@ -500,8 +503,8 @@ int vmw_kms_new_framebuffer_surface(struct vmw_private *dev_priv,
 	vfbs->base.base.depth = 24;
 	vfbs->base.base.width = width;
 	vfbs->base.base.height = height;
-	vfbs->base.pin = NULL;
-	vfbs->base.unpin = NULL;
+	vfbs->base.pin = &vmw_surface_dmabuf_pin;
+	vfbs->base.unpin = &vmw_surface_dmabuf_unpin;
 	vfbs->surface = surface;
 	mutex_init(&vfbs->work_lock);
 	INIT_DELAYED_WORK(&vfbs->d_work, &vmw_framebuffer_present_fs_callback);
@@ -589,6 +592,40 @@ static struct drm_framebuffer_funcs vmw_framebuffer_dmabuf_funcs = {
 	.create_handle = vmw_framebuffer_create_handle,
 };
 
+static int vmw_surface_dmabuf_pin(struct vmw_framebuffer *vfb)
+{
+	struct vmw_private *dev_priv = vmw_priv(vfb->base.dev);
+	struct vmw_framebuffer_surface *vfbs =
+		vmw_framebuffer_to_vfbs(&vfb->base);
+	unsigned long size = vfbs->base.base.pitch * vfbs->base.base.height;
+	int ret;
+
+	vfbs->buffer = kzalloc(sizeof(*vfbs->buffer), GFP_KERNEL);
+	if (unlikely(vfbs->buffer == NULL))
+		return -ENOMEM;
+
+	vmw_overlay_pause_all(dev_priv);
+	ret = vmw_dmabuf_init(dev_priv, vfbs->buffer, size,
+			       &vmw_vram_ne_placement,
+			       false, &vmw_dmabuf_bo_free);
+	vmw_overlay_resume_all(dev_priv);
+
+	return ret;
+}
+
+static int vmw_surface_dmabuf_unpin(struct vmw_framebuffer *vfb)
+{
+	struct ttm_buffer_object *bo;
+	struct vmw_framebuffer_surface *vfbs =
+		vmw_framebuffer_to_vfbs(&vfb->base);
+
+	bo = &vfbs->buffer->base;
+	ttm_bo_unref(&bo);
+	vfbs->buffer = NULL;
+
+	return 0;
+}
+
 static int vmw_framebuffer_dmabuf_pin(struct vmw_framebuffer *vfb)
 {
 	struct vmw_private *dev_priv = vmw_priv(vfb->base.dev);
@@ -596,33 +633,15 @@ static int vmw_framebuffer_dmabuf_pin(struct vmw_framebuffer *vfb)
 		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);
 
+	WARN_ON(ret != 0);
+
 	return 0;
 }
 
@@ -668,7 +687,7 @@ int vmw_kms_new_framebuffer_dmabuf(struct vmw_private *dev_priv,
 
 	/* 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.pitch = width * vfbd->base.base.bits_per_pixel / 8;
 	vfbd->base.base.depth = 24;
 	vfbd->base.base.width = width;
 	vfbd->base.base.height = height;
@@ -765,8 +784,9 @@ int vmw_kms_init(struct vmw_private *dev_priv)
 	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;
+	/* assumed largest fb size */
+	dev->mode_config.max_width = 8192;
+	dev->mode_config.max_height = 8192;
 
 	ret = vmw_kms_init_legacy_display_system(dev_priv);
 
@@ -826,49 +846,140 @@ out:
 	return ret;
 }
 
+void vmw_kms_write_svga(struct vmw_private *vmw_priv,
+			unsigned width, unsigned height, unsigned pitch,
+			unsigned bbp, unsigned depth)
+{
+	if (vmw_priv->capabilities & SVGA_CAP_PITCHLOCK)
+		vmw_write(vmw_priv, SVGA_REG_PITCHLOCK, pitch);
+	else if (vmw_fifo_have_pitchlock(vmw_priv))
+		iowrite32(pitch, vmw_priv->mmio_virt + SVGA_FIFO_PITCHLOCK);
+	vmw_write(vmw_priv, SVGA_REG_WIDTH, width);
+	vmw_write(vmw_priv, SVGA_REG_HEIGHT, height);
+	vmw_write(vmw_priv, SVGA_REG_BITS_PER_PIXEL, bbp);
+	vmw_write(vmw_priv, SVGA_REG_DEPTH, 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);
+}
+
 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);
-	}
+	struct vmw_vga_topology_state *save;
+	uint32_t i;
 
 	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_bpp = vmw_read(vmw_priv, SVGA_REG_BITS_PER_PIXEL);
 	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);
+	vmw_priv->vga_green_mask = vmw_read(vmw_priv, SVGA_REG_GREEN_MASK);
+	if (vmw_priv->capabilities & SVGA_CAP_PITCHLOCK)
+		vmw_priv->vga_pitchlock =
+		  vmw_read(vmw_priv, SVGA_REG_PITCHLOCK);
+	else if (vmw_fifo_have_pitchlock(vmw_priv))
+		vmw_priv->vga_pitchlock = ioread32(vmw_priv->mmio_virt +
+						       SVGA_FIFO_PITCHLOCK);
+
+	if (!(vmw_priv->capabilities & SVGA_CAP_DISPLAY_TOPOLOGY))
+		return 0;
 
+	vmw_priv->num_displays = vmw_read(vmw_priv,
+					  SVGA_REG_NUM_GUEST_DISPLAYS);
+
+	for (i = 0; i < vmw_priv->num_displays; ++i) {
+		save = &vmw_priv->vga_save[i];
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_ID, i);
+		save->primary = vmw_read(vmw_priv, SVGA_REG_DISPLAY_IS_PRIMARY);
+		save->pos_x = vmw_read(vmw_priv, SVGA_REG_DISPLAY_POSITION_X);
+		save->pos_y = vmw_read(vmw_priv, SVGA_REG_DISPLAY_POSITION_Y);
+		save->width = vmw_read(vmw_priv, SVGA_REG_DISPLAY_WIDTH);
+		save->height = vmw_read(vmw_priv, SVGA_REG_DISPLAY_HEIGHT);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_ID, SVGA_ID_INVALID);
+	}
 	return 0;
 }
 
 int vmw_kms_restore_vga(struct vmw_private *vmw_priv)
 {
+	struct vmw_vga_topology_state *save;
+	uint32_t i;
+
 	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_BITS_PER_PIXEL, vmw_priv->vga_bpp);
 	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);
+	if (vmw_priv->capabilities & SVGA_CAP_PITCHLOCK)
+		vmw_write(vmw_priv, SVGA_REG_PITCHLOCK,
+			  vmw_priv->vga_pitchlock);
+	else if (vmw_fifo_have_pitchlock(vmw_priv))
+		iowrite32(vmw_priv->vga_pitchlock,
+			  vmw_priv->mmio_virt + SVGA_FIFO_PITCHLOCK);
+
+	if (!(vmw_priv->capabilities & SVGA_CAP_DISPLAY_TOPOLOGY))
+		return 0;
 
-	/* TODO check for multimon */
-	vmw_write(vmw_priv, SVGA_REG_NUM_GUEST_DISPLAYS, 0);
+	for (i = 0; i < vmw_priv->num_displays; ++i) {
+		save = &vmw_priv->vga_save[i];
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_ID, i);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_IS_PRIMARY, save->primary);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_POSITION_X, save->pos_x);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_POSITION_Y, save->pos_y);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_WIDTH, save->width);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_HEIGHT, save->height);
+		vmw_write(vmw_priv, SVGA_REG_DISPLAY_ID, SVGA_ID_INVALID);
+	}
 
 	return 0;
 }
+
+int vmw_kms_update_layout_ioctl(struct drm_device *dev, void *data,
+				struct drm_file *file_priv)
+{
+	struct vmw_private *dev_priv = vmw_priv(dev);
+	struct drm_vmw_update_layout_arg *arg =
+		(struct drm_vmw_update_layout_arg *)data;
+	struct vmw_master *vmaster = vmw_master(file_priv->master);
+	void __user *user_rects;
+	struct drm_vmw_rect *rects;
+	unsigned rects_size;
+	int ret;
+
+	ret = ttm_read_lock(&vmaster->lock, true);
+	if (unlikely(ret != 0))
+		return ret;
+
+	if (!arg->num_outputs) {
+		struct drm_vmw_rect def_rect = {0, 0, 800, 600};
+		vmw_kms_ldu_update_layout(dev_priv, 1, &def_rect);
+		goto out_unlock;
+	}
+
+	rects_size = arg->num_outputs * sizeof(struct drm_vmw_rect);
+	rects = kzalloc(rects_size, GFP_KERNEL);
+	if (unlikely(!rects)) {
+		ret = -ENOMEM;
+		goto out_unlock;
+	}
+
+	user_rects = (void __user *)(unsigned long)arg->rects;
+	ret = copy_from_user(rects, user_rects, rects_size);
+	if (unlikely(ret != 0)) {
+		DRM_ERROR("Failed to get rects.\n");
+		goto out_free;
+	}
+
+	vmw_kms_ldu_update_layout(dev_priv, arg->num_outputs, rects);
+
+out_free:
+	kfree(rects);
+out_unlock:
+	ttm_read_unlock(&vmaster->lock);
+	return ret;
+}
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.h b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.h
index 8b95249f0531..8a398a0339b6 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.h
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.h
@@ -94,9 +94,11 @@ int vmw_du_crtc_cursor_set(struct drm_crtc *crtc, struct drm_file *file_priv,
 int vmw_du_crtc_cursor_move(struct drm_crtc *crtc, int x, int y);
 
 /*
- * Legacy display unit functions - vmwgfx_ldu.h
+ * Legacy display unit functions - vmwgfx_ldu.c
  */
 int vmw_kms_init_legacy_display_system(struct vmw_private *dev_priv);
 int vmw_kms_close_legacy_display_system(struct vmw_private *dev_priv);
+int vmw_kms_ldu_update_layout(struct vmw_private *dev_priv, unsigned num,
+			      struct drm_vmw_rect *rects);
 
 #endif
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ldu.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ldu.c
index 90891593bf6c..cfaf690a5b2f 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ldu.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ldu.c
@@ -38,6 +38,7 @@ struct vmw_legacy_display {
 	struct list_head active;
 
 	unsigned num_active;
+	unsigned last_num_active;
 
 	struct vmw_framebuffer *fb;
 };
@@ -48,9 +49,12 @@ struct vmw_legacy_display {
 struct vmw_legacy_display_unit {
 	struct vmw_display_unit base;
 
-	struct list_head active;
+	unsigned pref_width;
+	unsigned pref_height;
+	bool pref_active;
+	struct drm_display_mode *pref_mode;
 
-	unsigned unit;
+	struct list_head active;
 };
 
 static void vmw_ldu_destroy(struct vmw_legacy_display_unit *ldu)
@@ -88,23 +92,44 @@ 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;
+	struct drm_framebuffer *fb = NULL;
+	struct drm_crtc *crtc = NULL;
 	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);
+	/* If there is no display topology the host just assumes
+	 * that the guest will set the same layout as the host.
+	 */
+	if (!(dev_priv->capabilities & SVGA_CAP_DISPLAY_TOPOLOGY)) {
+		int w = 0, h = 0;
+		list_for_each_entry(entry, &lds->active, active) {
+			crtc = &entry->base.crtc;
+			w = max(w, crtc->x + crtc->mode.hdisplay);
+			h = max(h, crtc->y + crtc->mode.vdisplay);
+			i++;
+		}
+
+		if (crtc == NULL)
+			return 0;
+		fb = entry->base.crtc.fb;
+
+		vmw_kms_write_svga(dev_priv, w, h, fb->pitch,
+				   fb->bits_per_pixel, fb->depth);
+
+		return 0;
 	}
 
-	/* Now set the mode */
-	vmw_write(dev_priv, SVGA_REG_NUM_GUEST_DISPLAYS, lds->num_active);
+	if (!list_empty(&lds->active)) {
+		entry = list_entry(lds->active.next, typeof(*entry), active);
+		fb = entry->base.crtc.fb;
+
+		vmw_kms_write_svga(dev_priv, fb->width, fb->height, fb->pitch,
+				   fb->bits_per_pixel, fb->depth);
+	}
+
+	/* Make sure we always show something. */
+	vmw_write(dev_priv, SVGA_REG_NUM_GUEST_DISPLAYS,
+		  lds->num_active ? lds->num_active : 1);
+
 	i = 0;
 	list_for_each_entry(entry, &lds->active, active) {
 		crtc = &entry->base.crtc;
@@ -120,6 +145,10 @@ static int vmw_ldu_commit_list(struct vmw_private *dev_priv)
 		i++;
 	}
 
+	BUG_ON(i != lds->num_active);
+
+	lds->last_num_active = lds->num_active;
+
 	return 0;
 }
 
@@ -130,6 +159,7 @@ static int vmw_ldu_del_active(struct vmw_private *vmw_priv,
 	if (list_empty(&ldu->active))
 		return 0;
 
+	/* Must init otherwise list_empty(&ldu->active) will not work. */
 	list_del_init(&ldu->active);
 	if (--(ld->num_active) == 0) {
 		BUG_ON(!ld->fb);
@@ -149,24 +179,29 @@ static int vmw_ldu_add_active(struct vmw_private *vmw_priv,
 	struct vmw_legacy_display_unit *entry;
 	struct list_head *at;
 
+	BUG_ON(!ld->num_active && ld->fb);
+	if (vfb != ld->fb) {
+		if (ld->fb && ld->fb->unpin)
+			ld->fb->unpin(ld->fb);
+		if (vfb->pin)
+			vfb->pin(vfb);
+		ld->fb = vfb;
+	}
+
 	if (!list_empty(&ldu->active))
 		return 0;
 
 	at = &ld->active;
 	list_for_each_entry(entry, &ld->active, active) {
-		if (entry->unit > ldu->unit)
+		if (entry->base.unit > ldu->base.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;
-	}
+
+	ld->num_active++;
 
 	return 0;
 }
@@ -208,6 +243,8 @@ static int vmw_ldu_crtc_set_config(struct drm_mode_set *set)
 
 	/* ldu only supports one fb active at the time */
 	if (dev_priv->ldu_priv->fb && vfb &&
+	    !(dev_priv->ldu_priv->num_active == 1 &&
+	      !list_empty(&ldu->active)) &&
 	    dev_priv->ldu_priv->fb != vfb) {
 		DRM_ERROR("Multiple framebuffers not supported\n");
 		return -EINVAL;
@@ -300,8 +337,7 @@ 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)
+	if (vmw_connector_to_ldu(connector)->pref_active)
 		return connector_status_connected;
 	return connector_status_disconnected;
 }
@@ -312,10 +348,9 @@ static struct drm_display_mode vmw_ldu_connector_builtin[] = {
 		   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) },
+	{ DRM_MODE("800x600", DRM_MODE_TYPE_DRIVER, 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,
@@ -387,10 +422,34 @@ static struct drm_display_mode vmw_ldu_connector_builtin[] = {
 static int vmw_ldu_connector_fill_modes(struct drm_connector *connector,
 					uint32_t max_width, uint32_t max_height)
 {
+	struct vmw_legacy_display_unit *ldu = vmw_connector_to_ldu(connector);
 	struct drm_device *dev = connector->dev;
 	struct drm_display_mode *mode = NULL;
+	struct drm_display_mode prefmode = { DRM_MODE("preferred",
+		DRM_MODE_TYPE_DRIVER | DRM_MODE_TYPE_PREFERRED,
+		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+		DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_PVSYNC)
+	};
 	int i;
 
+	/* Add preferred mode */
+	{
+		mode = drm_mode_duplicate(dev, &prefmode);
+		if (!mode)
+			return 0;
+		mode->hdisplay = ldu->pref_width;
+		mode->vdisplay = ldu->pref_height;
+		mode->vrefresh = drm_mode_vrefresh(mode);
+		drm_mode_probed_add(connector, mode);
+
+		if (ldu->pref_mode) {
+			list_del_init(&ldu->pref_mode->head);
+			drm_mode_destroy(dev, ldu->pref_mode);
+		}
+
+		ldu->pref_mode = mode;
+	}
+
 	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)
@@ -443,18 +502,21 @@ static int vmw_ldu_init(struct vmw_private *dev_priv, unsigned unit)
 	if (!ldu)
 		return -ENOMEM;
 
-	ldu->unit = unit;
+	ldu->base.unit = unit;
 	crtc = &ldu->base.crtc;
 	encoder = &ldu->base.encoder;
 	connector = &ldu->base.connector;
 
+	INIT_LIST_HEAD(&ldu->active);
+
+	ldu->pref_active = (unit == 0);
+	ldu->pref_width = 800;
+	ldu->pref_height = 600;
+	ldu->pref_mode = NULL;
+
 	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;
+	connector->status = vmw_ldu_connector_detect(connector);
 
 	drm_encoder_init(dev, encoder, &vmw_legacy_encoder_funcs,
 			 DRM_MODE_ENCODER_LVDS);
@@ -462,8 +524,6 @@ static int vmw_ldu_init(struct vmw_private *dev_priv, unsigned unit)
 	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,
@@ -487,18 +547,22 @@ int vmw_kms_init_legacy_display_system(struct vmw_private *dev_priv)
 
 	INIT_LIST_HEAD(&dev_priv->ldu_priv->active);
 	dev_priv->ldu_priv->num_active = 0;
+	dev_priv->ldu_priv->last_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);
+	/* for old hardware without multimon only enable one display */
+	if (dev_priv->capabilities & SVGA_CAP_MULTIMON) {
+		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;
 }
@@ -514,3 +578,42 @@ int vmw_kms_close_legacy_display_system(struct vmw_private *dev_priv)
 
 	return 0;
 }
+
+int vmw_kms_ldu_update_layout(struct vmw_private *dev_priv, unsigned num,
+			      struct drm_vmw_rect *rects)
+{
+	struct drm_device *dev = dev_priv->dev;
+	struct vmw_legacy_display_unit *ldu;
+	struct drm_connector *con;
+	int i;
+
+	mutex_lock(&dev->mode_config.mutex);
+
+#if 0
+	DRM_INFO("%s: new layout ", __func__);
+	for (i = 0; i < (int)num; i++)
+		DRM_INFO("(%i, %i %ux%u) ", rects[i].x, rects[i].y,
+			 rects[i].w, rects[i].h);
+	DRM_INFO("\n");
+#else
+	(void)i;
+#endif
+
+	list_for_each_entry(con, &dev->mode_config.connector_list, head) {
+		ldu = vmw_connector_to_ldu(con);
+		if (num > ldu->base.unit) {
+			ldu->pref_width = rects[ldu->base.unit].w;
+			ldu->pref_height = rects[ldu->base.unit].h;
+			ldu->pref_active = true;
+		} else {
+			ldu->pref_width = 800;
+			ldu->pref_height = 600;
+			ldu->pref_active = false;
+		}
+		con->status = vmw_ldu_connector_detect(con);
+	}
+
+	mutex_unlock(&dev->mode_config.mutex);
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_overlay.c b/drivers/gpu/drm/vmwgfx/vmwgfx_overlay.c
index ad566c85b075..df2036ed18d5 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_overlay.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_overlay.c
@@ -358,6 +358,8 @@ static int vmw_overlay_update_stream(struct vmw_private *dev_priv,
 	if (stream->buf != buf)
 		stream->buf = vmw_dmabuf_reference(buf);
 	stream->saved = *arg;
+	/* stream is no longer stopped/paused */
+	stream->paused = false;
 
 	return 0;
 }
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
index 7745394c3e63..5f2d5df01e5c 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
@@ -597,8 +597,10 @@ int vmw_surface_define_ioctl(struct drm_device *dev, void *data,
 
 	ret = copy_from_user(srf->sizes, user_sizes,
 			     srf->num_sizes * sizeof(*srf->sizes));
-	if (unlikely(ret != 0))
+	if (unlikely(ret != 0)) {
+		ret = -EFAULT;
 		goto out_err1;
+	}
 
 	if (srf->scanout &&
 	    srf->num_sizes == 1 &&
@@ -697,9 +699,11 @@ int vmw_surface_reference_ioctl(struct drm_device *dev, void *data,
 	if (user_sizes)
 		ret = copy_to_user(user_sizes, srf->sizes,
 				   srf->num_sizes * sizeof(*srf->sizes));
-	if (unlikely(ret != 0))
+	if (unlikely(ret != 0)) {
 		DRM_ERROR("copy_to_user failed %p %u\n",
 			  user_sizes, srf->num_sizes);
+		ret = -EFAULT;
+	}
 out_bad_resource:
 out_no_reference:
 	ttm_base_object_unref(&base);
diff --git a/drivers/gpu/vga/vgaarb.c b/drivers/gpu/vga/vgaarb.c
index 441e38c95a85..b87569e96b16 100644
--- a/drivers/gpu/vga/vgaarb.c
+++ b/drivers/gpu/vga/vgaarb.c
@@ -1,12 +1,32 @@
 /*
- * vgaarb.c
+ * vgaarb.c: Implements the VGA arbitration. For details refer to
+ * Documentation/vgaarbiter.txt
+ *
  *
  * (C) Copyright 2005 Benjamin Herrenschmidt <benh@kernel.crashing.org>
  * (C) Copyright 2007 Paulo R. Zanoni <przanoni@gmail.com>
  * (C) Copyright 2007, 2009 Tiago Vignatti <vignatti@freedesktop.org>
  *
- * Implements the VGA arbitration. For details refer to
- * Documentation/vgaarbiter.txt
+ * 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 <linux/module.h>
@@ -155,8 +175,8 @@ static struct vga_device *__vga_tryget(struct vga_device *vgadev,
 	    (vgadev->decodes & VGA_RSRC_LEGACY_MEM))
 		rsrc |= VGA_RSRC_LEGACY_MEM;
 
-	pr_devel("%s: %d\n", __func__, rsrc);
-	pr_devel("%s: owns: %d\n", __func__, vgadev->owns);
+	pr_debug("%s: %d\n", __func__, rsrc);
+	pr_debug("%s: owns: %d\n", __func__, vgadev->owns);
 
 	/* Check what resources we need to acquire */
 	wants = rsrc & ~vgadev->owns;
@@ -268,7 +288,7 @@ static void __vga_put(struct vga_device *vgadev, unsigned int rsrc)
 {
 	unsigned int old_locks = vgadev->locks;
 
-	pr_devel("%s\n", __func__);
+	pr_debug("%s\n", __func__);
 
 	/* Update our counters, and account for equivalent legacy resources
 	 * if we decode them
@@ -575,6 +595,7 @@ static inline void vga_update_device_decodes(struct vga_device *vgadev,
 		else
 			vga_decode_count--;
 	}
+	pr_debug("vgaarb: decoding count now is: %d\n", vga_decode_count);
 }
 
 void __vga_set_legacy_decoding(struct pci_dev *pdev, unsigned int decodes, bool userspace)
@@ -831,7 +852,7 @@ static ssize_t vga_arb_write(struct file *file, const char __user * buf,
 		curr_pos += 5;
 		remaining -= 5;
 
-		pr_devel("client 0x%p called 'lock'\n", priv);
+		pr_debug("client 0x%p called 'lock'\n", priv);
 
 		if (!vga_str_to_iostate(curr_pos, remaining, &io_state)) {
 			ret_val = -EPROTO;
@@ -867,7 +888,7 @@ static ssize_t vga_arb_write(struct file *file, const char __user * buf,
 		curr_pos += 7;
 		remaining -= 7;
 
-		pr_devel("client 0x%p called 'unlock'\n", priv);
+		pr_debug("client 0x%p called 'unlock'\n", priv);
 
 		if (strncmp(curr_pos, "all", 3) == 0)
 			io_state = VGA_RSRC_LEGACY_IO | VGA_RSRC_LEGACY_MEM;
@@ -917,7 +938,7 @@ static ssize_t vga_arb_write(struct file *file, const char __user * buf,
 		curr_pos += 8;
 		remaining -= 8;
 
-		pr_devel("client 0x%p called 'trylock'\n", priv);
+		pr_debug("client 0x%p called 'trylock'\n", priv);
 
 		if (!vga_str_to_iostate(curr_pos, remaining, &io_state)) {
 			ret_val = -EPROTO;
@@ -961,7 +982,7 @@ static ssize_t vga_arb_write(struct file *file, const char __user * buf,
 
 		curr_pos += 7;
 		remaining -= 7;
-		pr_devel("client 0x%p called 'target'\n", priv);
+		pr_debug("client 0x%p called 'target'\n", priv);
 		/* if target is default */
 		if (!strncmp(curr_pos, "default", 7))
 			pdev = pci_dev_get(vga_default_device());
@@ -971,11 +992,11 @@ static ssize_t vga_arb_write(struct file *file, const char __user * buf,
 				ret_val = -EPROTO;
 				goto done;
 			}
-			pr_devel("vgaarb: %s ==> %x:%x:%x.%x\n", curr_pos,
+			pr_debug("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);
+			pr_debug("vgaarb: pbus %p\n", pbus);
 			if (pbus == NULL) {
 				pr_err("vgaarb: invalid PCI domain and/or bus address %x:%x\n",
 					domain, bus);
@@ -983,7 +1004,7 @@ static ssize_t vga_arb_write(struct file *file, const char __user * buf,
 				goto done;
 			}
 			pdev = pci_get_slot(pbus, devfn);
-			pr_devel("vgaarb: pdev %p\n", pdev);
+			pr_debug("vgaarb: pdev %p\n", pdev);
 			if (!pdev) {
 				pr_err("vgaarb: invalid PCI address %x:%x\n",
 					bus, devfn);
@@ -993,7 +1014,7 @@ static ssize_t vga_arb_write(struct file *file, const char __user * buf,
 		}
 
 		vgadev = vgadev_find(pdev);
-		pr_devel("vgaarb: vgadev %p\n", vgadev);
+		pr_debug("vgaarb: vgadev %p\n", vgadev);
 		if (vgadev == NULL) {
 			pr_err("vgaarb: this pci device is not a vga device\n");
 			pci_dev_put(pdev);
@@ -1029,7 +1050,7 @@ static ssize_t vga_arb_write(struct file *file, const char __user * buf,
 	} else if (strncmp(curr_pos, "decodes ", 8) == 0) {
 		curr_pos += 8;
 		remaining -= 8;
-		pr_devel("vgaarb: client 0x%p called 'decodes'\n", priv);
+		pr_debug("vgaarb: client 0x%p called 'decodes'\n", priv);
 
 		if (!vga_str_to_iostate(curr_pos, remaining, &io_state)) {
 			ret_val = -EPROTO;
@@ -1058,7 +1079,7 @@ static unsigned int vga_arb_fpoll(struct file *file, poll_table * wait)
 {
 	struct vga_arb_private *priv = file->private_data;
 
-	pr_devel("%s\n", __func__);
+	pr_debug("%s\n", __func__);
 
 	if (priv == NULL)
 		return -ENODEV;
@@ -1071,7 +1092,7 @@ static int vga_arb_open(struct inode *inode, struct file *file)
 	struct vga_arb_private *priv;
 	unsigned long flags;
 
-	pr_devel("%s\n", __func__);
+	pr_debug("%s\n", __func__);
 
 	priv = kmalloc(sizeof(struct vga_arb_private), GFP_KERNEL);
 	if (priv == NULL)
@@ -1101,7 +1122,7 @@ static int vga_arb_release(struct inode *inode, struct file *file)
 	unsigned long flags;
 	int i;
 
-	pr_devel("%s\n", __func__);
+	pr_debug("%s\n", __func__);
 
 	if (priv == NULL)
 		return -ENODEV;
@@ -1112,7 +1133,7 @@ static int vga_arb_release(struct inode *inode, struct file *file)
 		uc = &priv->cards[i];
 		if (uc->pdev == NULL)
 			continue;
-		pr_devel("uc->io_cnt == %d, uc->mem_cnt == %d\n",
+		pr_debug("uc->io_cnt == %d, uc->mem_cnt == %d\n",
 			 uc->io_cnt, uc->mem_cnt);
 		while (uc->io_cnt--)
 			vga_put(uc->pdev, VGA_RSRC_LEGACY_IO);
@@ -1165,7 +1186,7 @@ static int pci_notify(struct notifier_block *nb, unsigned long action,
 	struct pci_dev *pdev = to_pci_dev(dev);
 	bool notify = false;
 
-	pr_devel("%s\n", __func__);
+	pr_debug("%s\n", __func__);
 
 	/* For now we're only intereted in devices added and removed. I didn't
 	 * test this thing here, so someone needs to double check for the
diff --git a/drivers/hid/Kconfig b/drivers/hid/Kconfig
index 76ba59b9fea1..132278fa6240 100644
--- a/drivers/hid/Kconfig
+++ b/drivers/hid/Kconfig
@@ -347,6 +347,14 @@ config HID_QUANTA
 	---help---
 	Support for Quanta Optical Touch dual-touch panels.
 
+config HID_ROCCAT
+	tristate "Roccat special event support"
+	depends on USB_HID
+	---help---
+	Support for Roccat special events.
+	Say Y here if you have a Roccat mouse or keyboard and want OSD or
+	macro execution support.
+
 config HID_ROCCAT_KONE
 	tristate "Roccat Kone Mouse support"
 	depends on USB_HID
diff --git a/drivers/hid/Makefile b/drivers/hid/Makefile
index 22e47eaeea32..987fa0627367 100644
--- a/drivers/hid/Makefile
+++ b/drivers/hid/Makefile
@@ -48,6 +48,7 @@ obj-$(CONFIG_HID_QUANTA)	+= hid-quanta.o
 obj-$(CONFIG_HID_PANTHERLORD)	+= hid-pl.o
 obj-$(CONFIG_HID_PETALYNX)	+= hid-petalynx.o
 obj-$(CONFIG_HID_PICOLCD)	+= hid-picolcd.o
+obj-$(CONFIG_HID_ROCCAT)	+= hid-roccat.o
 obj-$(CONFIG_HID_ROCCAT_KONE)	+= hid-roccat-kone.o
 obj-$(CONFIG_HID_SAMSUNG)	+= hid-samsung.o
 obj-$(CONFIG_HID_SMARTJOYPLUS)	+= hid-sjoy.o
diff --git a/drivers/hid/hid-core.c b/drivers/hid/hid-core.c
index e10e314d38cc..aa0f7dcabcd7 100644
--- a/drivers/hid/hid-core.c
+++ b/drivers/hid/hid-core.c
@@ -1301,6 +1301,7 @@ static const struct hid_device_id hid_blacklist[] = {
 	{ HID_USB_DEVICE(USB_VENDOR_ID_GREENASIA, 0x0012) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_GYRATION, USB_DEVICE_ID_GYRATION_REMOTE) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_GYRATION, USB_DEVICE_ID_GYRATION_REMOTE_2) },
+	{ HID_USB_DEVICE(USB_VENDOR_ID_GYRATION, USB_DEVICE_ID_GYRATION_REMOTE_3) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_KENSINGTON, USB_DEVICE_ID_KS_SLIMBLADE) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_KYE, USB_DEVICE_ID_KYE_ERGO_525V) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_LABTEC, USB_DEVICE_ID_LABTEC_WIRELESS_KEYBOARD) },
diff --git a/drivers/hid/hid-debug.c b/drivers/hid/hid-debug.c
index 56f314fbd4f9..c94026768570 100644
--- a/drivers/hid/hid-debug.c
+++ b/drivers/hid/hid-debug.c
@@ -811,7 +811,7 @@ static const char *relatives[REL_MAX + 1] = {
 	[REL_WHEEL] = "Wheel",		[REL_MISC] = "Misc",
 };
 
-static const char *absolutes[ABS_MAX + 1] = {
+static const char *absolutes[ABS_CNT] = {
 	[ABS_X] = "X",			[ABS_Y] = "Y",
 	[ABS_Z] = "Z",			[ABS_RX] = "Rx",
 	[ABS_RY] = "Ry",		[ABS_RZ] = "Rz",
diff --git a/drivers/hid/hid-gyration.c b/drivers/hid/hid-gyration.c
index 62416e6baeca..3975e039c3dd 100644
--- a/drivers/hid/hid-gyration.c
+++ b/drivers/hid/hid-gyration.c
@@ -73,6 +73,7 @@ static int gyration_event(struct hid_device *hdev, struct hid_field *field,
 static const struct hid_device_id gyration_devices[] = {
 	{ HID_USB_DEVICE(USB_VENDOR_ID_GYRATION, USB_DEVICE_ID_GYRATION_REMOTE) },
 	{ HID_USB_DEVICE(USB_VENDOR_ID_GYRATION, USB_DEVICE_ID_GYRATION_REMOTE_2) },
+	{ HID_USB_DEVICE(USB_VENDOR_ID_GYRATION, USB_DEVICE_ID_GYRATION_REMOTE_3) },
 	{ }
 };
 MODULE_DEVICE_TABLE(hid, gyration_devices);
diff --git a/drivers/hid/hid-ids.h b/drivers/hid/hid-ids.h
index 9776896cc4fc..6af77ed0b555 100644
--- a/drivers/hid/hid-ids.h
+++ b/drivers/hid/hid-ids.h
@@ -282,6 +282,7 @@
 #define USB_VENDOR_ID_GYRATION		0x0c16
 #define USB_DEVICE_ID_GYRATION_REMOTE	0x0002
 #define USB_DEVICE_ID_GYRATION_REMOTE_2 0x0003
+#define USB_DEVICE_ID_GYRATION_REMOTE_3 0x0008
 
 #define USB_VENDOR_ID_HAPP		0x078b
 #define USB_DEVICE_ID_UGCI_DRIVING	0x0010
diff --git a/drivers/hid/hid-roccat-kone.c b/drivers/hid/hid-roccat-kone.c
index 66e694054ba2..17f2dc04f883 100644
--- a/drivers/hid/hid-roccat-kone.c
+++ b/drivers/hid/hid-roccat-kone.c
@@ -37,6 +37,7 @@
 #include <linux/module.h>
 #include <linux/slab.h>
 #include "hid-ids.h"
+#include "hid-roccat.h"
 #include "hid-roccat-kone.h"
 
 static void kone_set_settings_checksum(struct kone_settings *settings)
@@ -263,7 +264,7 @@ static int kone_get_firmware_version(struct usb_device *usb_dev, int *result)
 	return 0;
 }
 
-static ssize_t kone_sysfs_read_settings(struct kobject *kobj,
+static ssize_t kone_sysfs_read_settings(struct file *fp, struct kobject *kobj,
 		struct bin_attribute *attr, char *buf,
 		loff_t off, size_t count) {
 	struct device *dev = container_of(kobj, struct device, kobj);
@@ -287,7 +288,7 @@ static ssize_t kone_sysfs_read_settings(struct kobject *kobj,
  * This function keeps values in kone_device up to date and assumes that in
  * case of error the old data is still valid
  */
-static ssize_t kone_sysfs_write_settings(struct kobject *kobj,
+static ssize_t kone_sysfs_write_settings(struct file *fp, struct kobject *kobj,
 		struct bin_attribute *attr, char *buf,
 		loff_t off, size_t count) {
 	struct device *dev = container_of(kobj, struct device, kobj);
@@ -342,31 +343,31 @@ static ssize_t kone_sysfs_read_profilex(struct kobject *kobj,
 	return count;
 }
 
-static ssize_t kone_sysfs_read_profile1(struct kobject *kobj,
+static ssize_t kone_sysfs_read_profile1(struct file *fp, struct kobject *kobj,
 		struct bin_attribute *attr, char *buf,
 		loff_t off, size_t count) {
 	return kone_sysfs_read_profilex(kobj, attr, buf, off, count, 1);
 }
 
-static ssize_t kone_sysfs_read_profile2(struct kobject *kobj,
+static ssize_t kone_sysfs_read_profile2(struct file *fp, struct kobject *kobj,
 		struct bin_attribute *attr, char *buf,
 		loff_t off, size_t count) {
 	return kone_sysfs_read_profilex(kobj, attr, buf, off, count, 2);
 }
 
-static ssize_t kone_sysfs_read_profile3(struct kobject *kobj,
+static ssize_t kone_sysfs_read_profile3(struct file *fp, struct kobject *kobj,
 		struct bin_attribute *attr, char *buf,
 		loff_t off, size_t count) {
 	return kone_sysfs_read_profilex(kobj, attr, buf, off, count, 3);
 }
 
-static ssize_t kone_sysfs_read_profile4(struct kobject *kobj,
+static ssize_t kone_sysfs_read_profile4(struct file *fp, struct kobject *kobj,
 		struct bin_attribute *attr, char *buf,
 		loff_t off, size_t count) {
 	return kone_sysfs_read_profilex(kobj, attr, buf, off, count, 4);
 }
 
-static ssize_t kone_sysfs_read_profile5(struct kobject *kobj,
+static ssize_t kone_sysfs_read_profile5(struct file *fp, struct kobject *kobj,
 		struct bin_attribute *attr, char *buf,
 		loff_t off, size_t count) {
 	return kone_sysfs_read_profilex(kobj, attr, buf, off, count, 5);
@@ -404,31 +405,31 @@ static ssize_t kone_sysfs_write_profilex(struct kobject *kobj,
 	return sizeof(struct kone_profile);
 }
 
-static ssize_t kone_sysfs_write_profile1(struct kobject *kobj,
+static ssize_t kone_sysfs_write_profile1(struct file *fp, struct kobject *kobj,
 		struct bin_attribute *attr, char *buf,
 		loff_t off, size_t count) {
 	return kone_sysfs_write_profilex(kobj, attr, buf, off, count, 1);
 }
 
-static ssize_t kone_sysfs_write_profile2(struct kobject *kobj,
+static ssize_t kone_sysfs_write_profile2(struct file *fp, struct kobject *kobj,
 		struct bin_attribute *attr, char *buf,
 		loff_t off, size_t count) {
 	return kone_sysfs_write_profilex(kobj, attr, buf, off, count, 2);
 }
 
-static ssize_t kone_sysfs_write_profile3(struct kobject *kobj,
+static ssize_t kone_sysfs_write_profile3(struct file *fp, struct kobject *kobj,
 		struct bin_attribute *attr, char *buf,
 		loff_t off, size_t count) {
 	return kone_sysfs_write_profilex(kobj, attr, buf, off, count, 3);
 }
 
-static ssize_t kone_sysfs_write_profile4(struct kobject *kobj,
+static ssize_t kone_sysfs_write_profile4(struct file *fp, struct kobject *kobj,
 		struct bin_attribute *attr, char *buf,
 		loff_t off, size_t count) {
 	return kone_sysfs_write_profilex(kobj, attr, buf, off, count, 4);
 }
 
-static ssize_t kone_sysfs_write_profile5(struct kobject *kobj,
+static ssize_t kone_sysfs_write_profile5(struct file *fp, struct kobject *kobj,
 		struct bin_attribute *attr, char *buf,
 		loff_t off, size_t count) {
 	return kone_sysfs_write_profilex(kobj, attr, buf, off, count, 5);
@@ -849,6 +850,16 @@ static int kone_init_specials(struct hid_device *hdev)
 					"couldn't init struct kone_device\n");
 			goto exit_free;
 		}
+
+		retval = roccat_connect(hdev);
+		if (retval < 0) {
+			dev_err(&hdev->dev, "couldn't init char dev\n");
+			/* be tolerant about not getting chrdev */
+		} else {
+			kone->roccat_claimed = 1;
+			kone->chrdev_minor = retval;
+		}
+
 		retval = kone_create_sysfs_attributes(intf);
 		if (retval) {
 			dev_err(&hdev->dev, "cannot create sysfs files\n");
@@ -868,10 +879,14 @@ exit_free:
 static void kone_remove_specials(struct hid_device *hdev)
 {
 	struct usb_interface *intf = to_usb_interface(hdev->dev.parent);
+	struct kone_device *kone;
 
 	if (intf->cur_altsetting->desc.bInterfaceProtocol
 			== USB_INTERFACE_PROTOCOL_MOUSE) {
 		kone_remove_sysfs_attributes(intf);
+		kone = hid_get_drvdata(hdev);
+		if (kone->roccat_claimed)
+			roccat_disconnect(kone->chrdev_minor);
 		kfree(hid_get_drvdata(hdev));
 	}
 }
@@ -930,6 +945,37 @@ static void kone_keep_values_up_to_date(struct kone_device *kone,
 	}
 }
 
+static void kone_report_to_chrdev(struct kone_device const *kone,
+		struct kone_mouse_event const *event)
+{
+	struct kone_roccat_report roccat_report;
+
+	switch (event->event) {
+	case kone_mouse_event_switch_profile:
+	case kone_mouse_event_switch_dpi:
+	case kone_mouse_event_osd_profile:
+	case kone_mouse_event_osd_dpi:
+		roccat_report.event = event->event;
+		roccat_report.value = event->value;
+		roccat_report.key = 0;
+		roccat_report_event(kone->chrdev_minor,
+				(uint8_t *)&roccat_report,
+				sizeof(struct kone_roccat_report));
+		break;
+	case kone_mouse_event_call_overlong_macro:
+		if (event->value == kone_keystroke_action_press) {
+			roccat_report.event = kone_mouse_event_call_overlong_macro;
+			roccat_report.value = kone->actual_profile;
+			roccat_report.key = event->macro_key;
+			roccat_report_event(kone->chrdev_minor,
+					(uint8_t *)&roccat_report,
+					sizeof(struct kone_roccat_report));
+		}
+		break;
+	}
+
+}
+
 /*
  * Is called for keyboard- and mousepart.
  * Only mousepart gets informations about special events in its extended event
@@ -958,6 +1004,9 @@ static int kone_raw_event(struct hid_device *hdev, struct hid_report *report,
 
 	kone_keep_values_up_to_date(kone, event);
 
+	if (kone->roccat_claimed)
+		kone_report_to_chrdev(kone, event);
+
 	return 0; /* always do further processing */
 }
 
diff --git a/drivers/hid/hid-roccat-kone.h b/drivers/hid/hid-roccat-kone.h
index b413b10a7f8a..003e6f81c195 100644
--- a/drivers/hid/hid-roccat-kone.h
+++ b/drivers/hid/hid-roccat-kone.h
@@ -189,6 +189,12 @@ enum kone_commands {
 	kone_command_firmware = 0xe5a
 };
 
+struct kone_roccat_report {
+	uint8_t event;
+	uint8_t value; /* holds dpi or profile value */
+	uint8_t key; /* macro key on overlong macro execution */
+};
+
 #pragma pack(pop)
 
 struct kone_device {
@@ -219,6 +225,9 @@ struct kone_device {
 	 * so it's read only once
 	 */
 	int firmware_version;
+
+	int roccat_claimed;
+	int chrdev_minor;
 };
 
 #endif
diff --git a/drivers/hid/hid-roccat.c b/drivers/hid/hid-roccat.c
new file mode 100644
index 000000000000..e05d48edb66f
--- /dev/null
+++ b/drivers/hid/hid-roccat.c
@@ -0,0 +1,428 @@
+/*
+ * Roccat driver for Linux
+ *
+ * Copyright (c) 2010 Stefan Achatz <erazor_de@users.sourceforge.net>
+ */
+
+/*
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the Free
+ * Software Foundation; either version 2 of the License, or (at your option)
+ * any later version.
+ */
+
+/*
+ * Module roccat is a char device used to report special events of roccat
+ * hardware to userland. These events include requests for on-screen-display of
+ * profile or dpi settings or requests for execution of macro sequences that are
+ * not stored in device. The information in these events depends on hid device
+ * implementation and contains data that is not available in a single hid event
+ * or else hidraw could have been used.
+ * It is inspired by hidraw, but uses only one circular buffer for all readers.
+ */
+
+#include <linux/cdev.h>
+#include <linux/poll.h>
+#include <linux/sched.h>
+
+#include "hid-roccat.h"
+
+#define ROCCAT_FIRST_MINOR 0
+#define ROCCAT_MAX_DEVICES 8
+
+/* should be a power of 2 for performance reason */
+#define ROCCAT_CBUF_SIZE 16
+
+struct roccat_report {
+	uint8_t *value;
+	int len;
+};
+
+struct roccat_device {
+	unsigned int minor;
+	int open;
+	int exist;
+	wait_queue_head_t wait;
+	struct device *dev;
+	struct hid_device *hid;
+	struct list_head readers;
+	/* protects modifications of readers list */
+	struct mutex readers_lock;
+
+	/*
+	 * circular_buffer has one writer and multiple readers with their own
+	 * read pointers
+	 */
+	struct roccat_report cbuf[ROCCAT_CBUF_SIZE];
+	int cbuf_end;
+	struct mutex cbuf_lock;
+};
+
+struct roccat_reader {
+	struct list_head node;
+	struct roccat_device *device;
+	int cbuf_start;
+};
+
+static int roccat_major;
+static struct class *roccat_class;
+static struct cdev roccat_cdev;
+
+static struct roccat_device *devices[ROCCAT_MAX_DEVICES];
+/* protects modifications of devices array */
+static DEFINE_MUTEX(devices_lock);
+
+static ssize_t roccat_read(struct file *file, char __user *buffer,
+		size_t count, loff_t *ppos)
+{
+	struct roccat_reader *reader = file->private_data;
+	struct roccat_device *device = reader->device;
+	struct roccat_report *report;
+	ssize_t retval = 0, len;
+	DECLARE_WAITQUEUE(wait, current);
+
+	mutex_lock(&device->cbuf_lock);
+
+	/* no data? */
+	if (reader->cbuf_start == device->cbuf_end) {
+		add_wait_queue(&device->wait, &wait);
+		set_current_state(TASK_INTERRUPTIBLE);
+
+		/* wait for data */
+		while (reader->cbuf_start == device->cbuf_end) {
+			if (file->f_flags & O_NONBLOCK) {
+				retval = -EAGAIN;
+				break;
+			}
+			if (signal_pending(current)) {
+				retval = -ERESTARTSYS;
+				break;
+			}
+			if (!device->exist) {
+				retval = -EIO;
+				break;
+			}
+
+			mutex_unlock(&device->cbuf_lock);
+			schedule();
+			mutex_lock(&device->cbuf_lock);
+			set_current_state(TASK_INTERRUPTIBLE);
+		}
+
+		set_current_state(TASK_RUNNING);
+		remove_wait_queue(&device->wait, &wait);
+	}
+
+	/* here we either have data or a reason to return if retval is set */
+	if (retval)
+		goto exit_unlock;
+
+	report = &device->cbuf[reader->cbuf_start];
+	/*
+	 * If report is larger than requested amount of data, rest of report
+	 * is lost!
+	 */
+	len = report->len > count ? count : report->len;
+
+	if (copy_to_user(buffer, report->value, len)) {
+		retval = -EFAULT;
+		goto exit_unlock;
+	}
+	retval += len;
+	reader->cbuf_start = (reader->cbuf_start + 1) % ROCCAT_CBUF_SIZE;
+
+exit_unlock:
+	mutex_unlock(&device->cbuf_lock);
+	return retval;
+}
+
+static unsigned int roccat_poll(struct file *file, poll_table *wait)
+{
+	struct roccat_reader *reader = file->private_data;
+	poll_wait(file, &reader->device->wait, wait);
+	if (reader->cbuf_start != reader->device->cbuf_end)
+		return POLLIN | POLLRDNORM;
+	if (!reader->device->exist)
+		return POLLERR | POLLHUP;
+	return 0;
+}
+
+static int roccat_open(struct inode *inode, struct file *file)
+{
+	unsigned int minor = iminor(inode);
+	struct roccat_reader *reader;
+	struct roccat_device *device;
+	int error = 0;
+
+	reader = kzalloc(sizeof(struct roccat_reader), GFP_KERNEL);
+	if (!reader)
+		return -ENOMEM;
+
+	mutex_lock(&devices_lock);
+
+	device = devices[minor];
+
+	mutex_lock(&device->readers_lock);
+
+	if (!device) {
+		printk(KERN_EMERG "roccat device with minor %d doesn't exist\n",
+				minor);
+		error = -ENODEV;
+		goto exit_unlock;
+	}
+
+	if (!device->open++) {
+		/* power on device on adding first reader */
+		if (device->hid->ll_driver->power) {
+			error = device->hid->ll_driver->power(device->hid,
+					PM_HINT_FULLON);
+			if (error < 0) {
+				--device->open;
+				goto exit_unlock;
+			}
+		}
+		error = device->hid->ll_driver->open(device->hid);
+		if (error < 0) {
+			if (device->hid->ll_driver->power)
+				device->hid->ll_driver->power(device->hid,
+						PM_HINT_NORMAL);
+			--device->open;
+			goto exit_unlock;
+		}
+	}
+
+	reader->device = device;
+	/* new reader doesn't get old events */
+	reader->cbuf_start = device->cbuf_end;
+
+	list_add_tail(&reader->node, &device->readers);
+	file->private_data = reader;
+
+exit_unlock:
+	mutex_unlock(&device->readers_lock);
+	mutex_unlock(&devices_lock);
+	return error;
+}
+
+static int roccat_release(struct inode *inode, struct file *file)
+{
+	unsigned int minor = iminor(inode);
+	struct roccat_reader *reader = file->private_data;
+	struct roccat_device *device;
+
+	mutex_lock(&devices_lock);
+
+	device = devices[minor];
+	if (!device) {
+		mutex_unlock(&devices_lock);
+		printk(KERN_EMERG "roccat device with minor %d doesn't exist\n",
+				minor);
+		return -ENODEV;
+	}
+
+	mutex_lock(&device->readers_lock);
+	list_del(&reader->node);
+	mutex_unlock(&device->readers_lock);
+	kfree(reader);
+
+	if (!--device->open) {
+		/* removing last reader */
+		if (device->exist) {
+			if (device->hid->ll_driver->power)
+				device->hid->ll_driver->power(device->hid,
+						PM_HINT_NORMAL);
+			device->hid->ll_driver->close(device->hid);
+		} else {
+			kfree(device);
+		}
+	}
+
+	mutex_unlock(&devices_lock);
+
+	return 0;
+}
+
+/*
+ * roccat_report_event() - output data to readers
+ * @minor: minor device number returned by roccat_connect()
+ * @data: pointer to data
+ * @len: size of data
+ *
+ * Return value is zero on success, a negative error code on failure.
+ *
+ * This is called from interrupt handler.
+ */
+int roccat_report_event(int minor, u8 const *data, int len)
+{
+	struct roccat_device *device;
+	struct roccat_reader *reader;
+	struct roccat_report *report;
+	uint8_t *new_value;
+
+	new_value = kmemdup(data, len, GFP_ATOMIC);
+	if (!new_value)
+		return -ENOMEM;
+
+	device = devices[minor];
+
+	report = &device->cbuf[device->cbuf_end];
+
+	/* passing NULL is safe */
+	kfree(report->value);
+
+	report->value = new_value;
+	report->len = len;
+	device->cbuf_end = (device->cbuf_end + 1) % ROCCAT_CBUF_SIZE;
+
+	list_for_each_entry(reader, &device->readers, node) {
+		/*
+		 * As we already inserted one element, the buffer can't be
+		 * empty. If start and end are equal, buffer is full and we
+		 * increase start, so that slow reader misses one event, but
+		 * gets the newer ones in the right order.
+		 */
+		if (reader->cbuf_start == device->cbuf_end)
+			reader->cbuf_start = (reader->cbuf_start + 1) % ROCCAT_CBUF_SIZE;
+	}
+
+	wake_up_interruptible(&device->wait);
+	return 0;
+}
+EXPORT_SYMBOL_GPL(roccat_report_event);
+
+/*
+ * roccat_connect() - create a char device for special event output
+ * @hid: the hid device the char device should be connected to.
+ *
+ * Return value is minor device number in Range [0, ROCCAT_MAX_DEVICES] on
+ * success, a negative error code on failure.
+ */
+int roccat_connect(struct hid_device *hid)
+{
+	unsigned int minor;
+	struct roccat_device *device;
+	int temp;
+
+	device = kzalloc(sizeof(struct roccat_device), GFP_KERNEL);
+	if (!device)
+		return -ENOMEM;
+
+	mutex_lock(&devices_lock);
+
+	for (minor = 0; minor < ROCCAT_MAX_DEVICES; ++minor) {
+		if (devices[minor])
+			continue;
+		break;
+	}
+
+	if (minor < ROCCAT_MAX_DEVICES) {
+		devices[minor] = device;
+	} else {
+		mutex_unlock(&devices_lock);
+		kfree(device);
+		return -EINVAL;
+	}
+
+	device->dev = device_create(roccat_class, &hid->dev,
+			MKDEV(roccat_major, minor), NULL,
+			"%s%s%d", "roccat", hid->driver->name, minor);
+
+	if (IS_ERR(device->dev)) {
+		devices[minor] = NULL;
+		mutex_unlock(&devices_lock);
+		temp = PTR_ERR(device->dev);
+		kfree(device);
+		return temp;
+	}
+
+	mutex_unlock(&devices_lock);
+
+	init_waitqueue_head(&device->wait);
+	INIT_LIST_HEAD(&device->readers);
+	mutex_init(&device->readers_lock);
+	mutex_init(&device->cbuf_lock);
+	device->minor = minor;
+	device->hid = hid;
+	device->exist = 1;
+	device->cbuf_end = 0;
+
+	return minor;
+}
+EXPORT_SYMBOL_GPL(roccat_connect);
+
+/* roccat_disconnect() - remove char device from hid device
+ * @minor: the minor device number returned by roccat_connect()
+ */
+void roccat_disconnect(int minor)
+{
+	struct roccat_device *device;
+
+	mutex_lock(&devices_lock);
+	device = devices[minor];
+	devices[minor] = NULL;
+	mutex_unlock(&devices_lock);
+
+	device->exist = 0; /* TODO exist maybe not needed */
+
+	device_destroy(roccat_class, MKDEV(roccat_major, minor));
+
+	if (device->open) {
+		device->hid->ll_driver->close(device->hid);
+		wake_up_interruptible(&device->wait);
+	} else {
+		kfree(device);
+	}
+}
+EXPORT_SYMBOL_GPL(roccat_disconnect);
+
+static const struct file_operations roccat_ops = {
+	.owner = THIS_MODULE,
+	.read = roccat_read,
+	.poll = roccat_poll,
+	.open = roccat_open,
+	.release = roccat_release,
+};
+
+static int __init roccat_init(void)
+{
+	int retval;
+	dev_t dev_id;
+
+	retval = alloc_chrdev_region(&dev_id, ROCCAT_FIRST_MINOR,
+			ROCCAT_MAX_DEVICES, "roccat");
+
+	roccat_major = MAJOR(dev_id);
+
+	if (retval < 0) {
+		printk(KERN_WARNING "roccat: can't get major number\n");
+		return retval;
+	}
+
+	roccat_class = class_create(THIS_MODULE, "roccat");
+	if (IS_ERR(roccat_class)) {
+		retval = PTR_ERR(roccat_class);
+		unregister_chrdev_region(dev_id, ROCCAT_MAX_DEVICES);
+		return retval;
+	}
+
+	cdev_init(&roccat_cdev, &roccat_ops);
+	cdev_add(&roccat_cdev, dev_id, ROCCAT_MAX_DEVICES);
+
+	return 0;
+}
+
+static void __exit roccat_exit(void)
+{
+	dev_t dev_id = MKDEV(roccat_major, 0);
+
+	cdev_del(&roccat_cdev);
+	class_destroy(roccat_class);
+	unregister_chrdev_region(dev_id, ROCCAT_MAX_DEVICES);
+}
+
+module_init(roccat_init);
+module_exit(roccat_exit);
+
+MODULE_AUTHOR("Stefan Achatz");
+MODULE_DESCRIPTION("USB Roccat char device");
+MODULE_LICENSE("GPL v2");
diff --git a/drivers/hid/hid-roccat.h b/drivers/hid/hid-roccat.h
new file mode 100644
index 000000000000..d8aae0c1fa7e
--- /dev/null
+++ b/drivers/hid/hid-roccat.h
@@ -0,0 +1,31 @@
+#ifndef __HID_ROCCAT_H
+#define __HID_ROCCAT_H
+
+/*
+ * Copyright (c) 2010 Stefan Achatz <erazor_de@users.sourceforge.net>
+ */
+
+/*
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the Free
+ * Software Foundation; either version 2 of the License, or (at your option)
+ * any later version.
+ */
+
+#include <linux/hid.h>
+#include <linux/types.h>
+
+#if defined(CONFIG_HID_ROCCAT) || defined (CONFIG_HID_ROCCAT_MODULE)
+int roccat_connect(struct hid_device *hid);
+void roccat_disconnect(int minor);
+int roccat_report_event(int minor, u8 const *data, int len);
+#else
+static inline int roccat_connect(struct hid_device *hid) { return -1; }
+static inline void roccat_disconnect(int minor) {}
+static inline int roccat_report_event(int minor, u8 const *data, int len)
+{
+	return 0;
+}
+#endif
+
+#endif
diff --git a/drivers/hwmon/Kconfig b/drivers/hwmon/Kconfig
index 9be8e1754a0b..e19cf8eb6ccf 100644
--- a/drivers/hwmon/Kconfig
+++ b/drivers/hwmon/Kconfig
@@ -447,13 +447,14 @@ config SENSORS_IT87
 	  will be called it87.
 
 config SENSORS_LM63
-	tristate "National Semiconductor LM63"
+	tristate "National Semiconductor LM63 and LM64"
 	depends on I2C
 	help
-	  If you say yes here you get support for the National Semiconductor
-	  LM63 remote diode digital temperature sensor with integrated fan
-	  control.  Such chips are found on the Tyan S4882 (Thunder K8QS Pro)
-	  motherboard, among others.
+	  If you say yes here you get support for the National
+	  Semiconductor LM63 and LM64 remote diode digital temperature
+	  sensors with integrated fan control.  Such chips are found
+	  on the Tyan S4882 (Thunder K8QS Pro) motherboard, among
+	  others.
 
 	  This driver can also be built as a module.  If so, the module
 	  will be called lm63.
@@ -492,7 +493,8 @@ config SENSORS_LM75
 		- NXP's LM75A
 		- ST Microelectronics STDS75
 		- TelCom (now Microchip) TCN75
-		- Texas Instruments TMP100, TMP101, TMP75, TMP175, TMP275
+		- Texas Instruments TMP100, TMP101, TMP105, TMP75, TMP175,
+		  TMP275
 
 	  This driver supports driver model based binding through board
 	  specific I2C device tables.
@@ -749,6 +751,16 @@ config SENSORS_DME1737
 	  This driver can also be built as a module.  If so, the module
 	  will be called dme1737.
 
+config SENSORS_EMC1403
+	tristate "SMSC EMC1403 thermal sensor"
+	depends on I2C
+	help
+	  If you say yes here you get support for the SMSC EMC1403
+	  temperature monitoring chip.
+
+	  Threshold values can be configured using sysfs.
+	  Data from the different diodes are accessible via sysfs.
+
 config SENSORS_SMSC47M1
 	tristate "SMSC LPC47M10x and compatibles"
 	help
@@ -802,6 +814,15 @@ config SENSORS_ADS7828
 	  This driver can also be built as a module.  If so, the module
 	  will be called ads7828.
 
+config SENSORS_ADS7871
+	tristate "Texas Instruments ADS7871 A/D converter"
+	depends on SPI
+	help
+	  If you say yes here you get support for TI ADS7871 & ADS7870
+
+	  This driver can also be built as a module.  If so, the module
+	  will be called ads7871.
+
 config SENSORS_AMC6821
 	tristate "Texas Instruments AMC6821"
 	depends on I2C  && EXPERIMENTAL
@@ -822,6 +843,16 @@ config SENSORS_THMC50
 	  This driver can also be built as a module.  If so, the module
 	  will be called thmc50.
 
+config SENSORS_TMP102
+	tristate "Texas Instruments TMP102"
+	depends on I2C && EXPERIMENTAL
+	help
+	  If you say yes here you get support for Texas Instruments TMP102
+	  sensor chips.
+
+	  This driver can also be built as a module.  If so, the module
+	  will be called tmp102.
+
 config SENSORS_TMP401
 	tristate "Texas Instruments TMP401 and compatibles"
 	depends on I2C && EXPERIMENTAL
diff --git a/drivers/hwmon/Makefile b/drivers/hwmon/Makefile
index 4aa1a3d112ad..2138ceb1a713 100644
--- a/drivers/hwmon/Makefile
+++ b/drivers/hwmon/Makefile
@@ -29,6 +29,7 @@ obj-$(CONFIG_SENSORS_ADM1029)	+= adm1029.o
 obj-$(CONFIG_SENSORS_ADM1031)	+= adm1031.o
 obj-$(CONFIG_SENSORS_ADM9240)	+= adm9240.o
 obj-$(CONFIG_SENSORS_ADS7828)	+= ads7828.o
+obj-$(CONFIG_SENSORS_ADS7871)	+= ads7871.o
 obj-$(CONFIG_SENSORS_ADT7411)	+= adt7411.o
 obj-$(CONFIG_SENSORS_ADT7462)	+= adt7462.o
 obj-$(CONFIG_SENSORS_ADT7470)	+= adt7470.o
@@ -40,6 +41,7 @@ obj-$(CONFIG_SENSORS_ATXP1)	+= atxp1.o
 obj-$(CONFIG_SENSORS_CORETEMP)	+= coretemp.o
 obj-$(CONFIG_SENSORS_DME1737)	+= dme1737.o
 obj-$(CONFIG_SENSORS_DS1621)	+= ds1621.o
+obj-$(CONFIG_SENSORS_EMC1403)	+= emc1403.o
 obj-$(CONFIG_SENSORS_F71805F)	+= f71805f.o
 obj-$(CONFIG_SENSORS_F71882FG)	+= f71882fg.o
 obj-$(CONFIG_SENSORS_F75375S)	+= f75375s.o
@@ -89,6 +91,7 @@ obj-$(CONFIG_SENSORS_SMSC47M1)	+= smsc47m1.o
 obj-$(CONFIG_SENSORS_SMSC47M192)+= smsc47m192.o
 obj-$(CONFIG_SENSORS_AMC6821)	+= amc6821.o
 obj-$(CONFIG_SENSORS_THMC50)	+= thmc50.o
+obj-$(CONFIG_SENSORS_TMP102)	+= tmp102.o
 obj-$(CONFIG_SENSORS_TMP401)	+= tmp401.o
 obj-$(CONFIG_SENSORS_TMP421)	+= tmp421.o
 obj-$(CONFIG_SENSORS_VIA_CPUTEMP)+= via-cputemp.o
diff --git a/drivers/hwmon/adm1031.c b/drivers/hwmon/adm1031.c
index 1644b92e7cc4..15c1a9616af3 100644
--- a/drivers/hwmon/adm1031.c
+++ b/drivers/hwmon/adm1031.c
@@ -36,6 +36,7 @@
 #define ADM1031_REG_FAN_DIV(nr)		(0x20 + (nr))
 #define ADM1031_REG_PWM			(0x22)
 #define ADM1031_REG_FAN_MIN(nr)		(0x10 + (nr))
+#define ADM1031_REG_FAN_FILTER		(0x23)
 
 #define ADM1031_REG_TEMP_OFFSET(nr)	(0x0d + (nr))
 #define ADM1031_REG_TEMP_MAX(nr)	(0x14 + 4 * (nr))
@@ -61,6 +62,9 @@
 #define ADM1031_CONF2_TACH2_ENABLE	0x08
 #define ADM1031_CONF2_TEMP_ENABLE(chan)	(0x10 << (chan))
 
+#define ADM1031_UPDATE_RATE_MASK	0x1c
+#define ADM1031_UPDATE_RATE_SHIFT	2
+
 /* Addresses to scan */
 static const unsigned short normal_i2c[] = { 0x2c, 0x2d, 0x2e, I2C_CLIENT_END };
 
@@ -75,6 +79,7 @@ struct adm1031_data {
 	int chip_type;
 	char valid;		/* !=0 if following fields are valid */
 	unsigned long last_updated;	/* In jiffies */
+	unsigned int update_rate;	/* In milliseconds */
 	/* The chan_select_table contains the possible configurations for
 	 * auto fan control.
 	 */
@@ -738,6 +743,57 @@ static SENSOR_DEVICE_ATTR(temp3_crit_alarm, S_IRUGO, show_alarm, NULL, 12);
 static SENSOR_DEVICE_ATTR(temp3_fault, S_IRUGO, show_alarm, NULL, 13);
 static SENSOR_DEVICE_ATTR(temp1_crit_alarm, S_IRUGO, show_alarm, NULL, 14);
 
+/* Update Rate */
+static const unsigned int update_rates[] = {
+	16000, 8000, 4000, 2000, 1000, 500, 250, 125,
+};
+
+static ssize_t show_update_rate(struct device *dev,
+				struct device_attribute *attr, char *buf)
+{
+	struct i2c_client *client = to_i2c_client(dev);
+	struct adm1031_data *data = i2c_get_clientdata(client);
+
+	return sprintf(buf, "%u\n", data->update_rate);
+}
+
+static ssize_t set_update_rate(struct device *dev,
+			       struct device_attribute *attr,
+			       const char *buf, size_t count)
+{
+	struct i2c_client *client = to_i2c_client(dev);
+	struct adm1031_data *data = i2c_get_clientdata(client);
+	unsigned long val;
+	int i, err;
+	u8 reg;
+
+	err = strict_strtoul(buf, 10, &val);
+	if (err)
+		return err;
+
+	/* find the nearest update rate from the table */
+	for (i = 0; i < ARRAY_SIZE(update_rates) - 1; i++) {
+		if (val >= update_rates[i])
+			break;
+	}
+	/* if not found, we point to the last entry (lowest update rate) */
+
+	/* set the new update rate while preserving other settings */
+	reg = adm1031_read_value(client, ADM1031_REG_FAN_FILTER);
+	reg &= ~ADM1031_UPDATE_RATE_MASK;
+	reg |= i << ADM1031_UPDATE_RATE_SHIFT;
+	adm1031_write_value(client, ADM1031_REG_FAN_FILTER, reg);
+
+	mutex_lock(&data->update_lock);
+	data->update_rate = update_rates[i];
+	mutex_unlock(&data->update_lock);
+
+	return count;
+}
+
+static DEVICE_ATTR(update_rate, S_IRUGO | S_IWUSR, show_update_rate,
+		   set_update_rate);
+
 static struct attribute *adm1031_attributes[] = {
 	&sensor_dev_attr_fan1_input.dev_attr.attr,
 	&sensor_dev_attr_fan1_div.dev_attr.attr,
@@ -774,6 +830,7 @@ static struct attribute *adm1031_attributes[] = {
 
 	&sensor_dev_attr_auto_fan1_min_pwm.dev_attr.attr,
 
+	&dev_attr_update_rate.attr,
 	&dev_attr_alarms.attr,
 
 	NULL
@@ -900,6 +957,7 @@ static void adm1031_init_client(struct i2c_client *client)
 {
 	unsigned int read_val;
 	unsigned int mask;
+	int i;
 	struct adm1031_data *data = i2c_get_clientdata(client);
 
 	mask = (ADM1031_CONF2_PWM1_ENABLE | ADM1031_CONF2_TACH1_ENABLE);
@@ -919,18 +977,24 @@ static void adm1031_init_client(struct i2c_client *client)
 				ADM1031_CONF1_MONITOR_ENABLE);
 	}
 
+	/* Read the chip's update rate */
+	mask = ADM1031_UPDATE_RATE_MASK;
+	read_val = adm1031_read_value(client, ADM1031_REG_FAN_FILTER);
+	i = (read_val & mask) >> ADM1031_UPDATE_RATE_SHIFT;
+	data->update_rate = update_rates[i];
 }
 
 static struct adm1031_data *adm1031_update_device(struct device *dev)
 {
 	struct i2c_client *client = to_i2c_client(dev);
 	struct adm1031_data *data = i2c_get_clientdata(client);
+	unsigned long next_update;
 	int chan;
 
 	mutex_lock(&data->update_lock);
 
-	if (time_after(jiffies, data->last_updated + HZ + HZ / 2)
-	    || !data->valid) {
+	next_update = data->last_updated + msecs_to_jiffies(data->update_rate);
+	if (time_after(jiffies, next_update) || !data->valid) {
 
 		dev_dbg(&client->dev, "Starting adm1031 update\n");
 		for (chan = 0;
diff --git a/drivers/hwmon/ads7871.c b/drivers/hwmon/ads7871.c
new file mode 100644
index 000000000000..b300a2048af1
--- /dev/null
+++ b/drivers/hwmon/ads7871.c
@@ -0,0 +1,253 @@
+/*
+ *  ads7871 - driver for TI ADS7871 A/D converter
+ *
+ *  Copyright (c) 2010 Paul Thomas <pthomas8589@gmail.com>
+ *
+ *  This program is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ *  GNU General Public License for more details.
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License version 2 or
+ *  later as publishhed by the Free Software Foundation.
+ *
+ *	You need to have something like this in struct spi_board_info
+ *	{
+ *		.modalias	= "ads7871",
+ *		.max_speed_hz	= 2*1000*1000,
+ *		.chip_select	= 0,
+ *		.bus_num	= 1,
+ *	},
+ */
+
+/*From figure 18 in the datasheet*/
+/*Register addresses*/
+#define REG_LS_BYTE	0 /*A/D Output Data, LS Byte*/
+#define REG_MS_BYTE	1 /*A/D Output Data, MS Byte*/
+#define REG_PGA_VALID	2 /*PGA Valid Register*/
+#define REG_AD_CONTROL	3 /*A/D Control Register*/
+#define REG_GAIN_MUX	4 /*Gain/Mux Register*/
+#define REG_IO_STATE	5 /*Digital I/O State Register*/
+#define REG_IO_CONTROL	6 /*Digital I/O Control Register*/
+#define REG_OSC_CONTROL	7 /*Rev/Oscillator Control Register*/
+#define REG_SER_CONTROL 24 /*Serial Interface Control Register*/
+#define REG_ID		31 /*ID Register*/
+
+/*From figure 17 in the datasheet
+* These bits get ORed with the address to form
+* the instruction byte */
+/*Instruction Bit masks*/
+#define INST_MODE_bm	(1<<7)
+#define INST_READ_bm	(1<<6)
+#define INST_16BIT_bm	(1<<5)
+
+/*From figure 18 in the datasheet*/
+/*bit masks for Rev/Oscillator Control Register*/
+#define MUX_CNV_bv	7
+#define MUX_CNV_bm	(1<<MUX_CNV_bv)
+#define MUX_M3_bm	(1<<3) /*M3 selects single ended*/
+#define MUX_G_bv	4 /*allows for reg = (gain << MUX_G_bv) | ...*/
+
+/*From figure 18 in the datasheet*/
+/*bit masks for Rev/Oscillator Control Register*/
+#define OSC_OSCR_bm	(1<<5)
+#define OSC_OSCE_bm	(1<<4)
+#define OSC_REFE_bm	(1<<3)
+#define OSC_BUFE_bm	(1<<2)
+#define OSC_R2V_bm	(1<<1)
+#define OSC_RBG_bm	(1<<0)
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/spi/spi.h>
+#include <linux/hwmon.h>
+#include <linux/hwmon-sysfs.h>
+#include <linux/err.h>
+#include <linux/mutex.h>
+#include <linux/delay.h>
+
+#define DEVICE_NAME	"ads7871"
+
+struct ads7871_data {
+	struct device	*hwmon_dev;
+	struct mutex	update_lock;
+};
+
+static int ads7871_read_reg8(struct spi_device *spi, int reg)
+{
+	int ret;
+	reg = reg | INST_READ_bm;
+	ret = spi_w8r8(spi, reg);
+	return ret;
+}
+
+static int ads7871_read_reg16(struct spi_device *spi, int reg)
+{
+	int ret;
+	reg = reg | INST_READ_bm | INST_16BIT_bm;
+	ret = spi_w8r16(spi, reg);
+	return ret;
+}
+
+static int ads7871_write_reg8(struct spi_device *spi, int reg, u8 val)
+{
+	u8 tmp[2] = {reg, val};
+	return spi_write(spi, tmp, sizeof(tmp));
+}
+
+static ssize_t show_voltage(struct device *dev,
+		struct device_attribute *da, char *buf)
+{
+	struct spi_device *spi = to_spi_device(dev);
+	struct sensor_device_attribute *attr = to_sensor_dev_attr(da);
+	int ret, val, i = 0;
+	uint8_t channel, mux_cnv;
+
+	channel = attr->index;
+	/*TODO: add support for conversions
+	 *other than single ended with a gain of 1*/
+	/*MUX_M3_bm forces single ended*/
+	/*This is also where the gain of the PGA would be set*/
+	ads7871_write_reg8(spi, REG_GAIN_MUX,
+		(MUX_CNV_bm | MUX_M3_bm | channel));
+
+	ret = ads7871_read_reg8(spi, REG_GAIN_MUX);
+	mux_cnv = ((ret & MUX_CNV_bm)>>MUX_CNV_bv);
+	/*on 400MHz arm9 platform the conversion
+	 *is already done when we do this test*/
+	while ((i < 2) && mux_cnv) {
+		i++;
+		ret = ads7871_read_reg8(spi, REG_GAIN_MUX);
+		mux_cnv = ((ret & MUX_CNV_bm)>>MUX_CNV_bv);
+		msleep_interruptible(1);
+	}
+
+	if (mux_cnv == 0) {
+		val = ads7871_read_reg16(spi, REG_LS_BYTE);
+		/*result in volts*10000 = (val/8192)*2.5*10000*/
+		val = ((val>>2) * 25000) / 8192;
+		return sprintf(buf, "%d\n", val);
+	} else {
+		return -1;
+	}
+}
+
+static SENSOR_DEVICE_ATTR(in0_input, S_IRUGO, show_voltage, NULL, 0);
+static SENSOR_DEVICE_ATTR(in1_input, S_IRUGO, show_voltage, NULL, 1);
+static SENSOR_DEVICE_ATTR(in2_input, S_IRUGO, show_voltage, NULL, 2);
+static SENSOR_DEVICE_ATTR(in3_input, S_IRUGO, show_voltage, NULL, 3);
+static SENSOR_DEVICE_ATTR(in4_input, S_IRUGO, show_voltage, NULL, 4);
+static SENSOR_DEVICE_ATTR(in5_input, S_IRUGO, show_voltage, NULL, 5);
+static SENSOR_DEVICE_ATTR(in6_input, S_IRUGO, show_voltage, NULL, 6);
+static SENSOR_DEVICE_ATTR(in7_input, S_IRUGO, show_voltage, NULL, 7);
+
+static struct attribute *ads7871_attributes[] = {
+	&sensor_dev_attr_in0_input.dev_attr.attr,
+	&sensor_dev_attr_in1_input.dev_attr.attr,
+	&sensor_dev_attr_in2_input.dev_attr.attr,
+	&sensor_dev_attr_in3_input.dev_attr.attr,
+	&sensor_dev_attr_in4_input.dev_attr.attr,
+	&sensor_dev_attr_in5_input.dev_attr.attr,
+	&sensor_dev_attr_in6_input.dev_attr.attr,
+	&sensor_dev_attr_in7_input.dev_attr.attr,
+	NULL
+};
+
+static const struct attribute_group ads7871_group = {
+	.attrs = ads7871_attributes,
+};
+
+static int __devinit ads7871_probe(struct spi_device *spi)
+{
+	int status, ret, err = 0;
+	uint8_t val;
+	struct ads7871_data *pdata;
+
+	dev_dbg(&spi->dev, "probe\n");
+
+	pdata = kzalloc(sizeof(struct ads7871_data), GFP_KERNEL);
+	if (!pdata) {
+		err = -ENOMEM;
+		goto exit;
+	}
+
+	status = sysfs_create_group(&spi->dev.kobj, &ads7871_group);
+	if (status < 0)
+		goto error_free;
+
+	pdata->hwmon_dev = hwmon_device_register(&spi->dev);
+	if (IS_ERR(pdata->hwmon_dev)) {
+		err = PTR_ERR(pdata->hwmon_dev);
+		goto error_remove;
+	}
+
+	spi_set_drvdata(spi, pdata);
+
+	/* Configure the SPI bus */
+	spi->mode = (SPI_MODE_0);
+	spi->bits_per_word = 8;
+	spi_setup(spi);
+
+	ads7871_write_reg8(spi, REG_SER_CONTROL, 0);
+	ads7871_write_reg8(spi, REG_AD_CONTROL, 0);
+
+	val = (OSC_OSCR_bm | OSC_OSCE_bm | OSC_REFE_bm | OSC_BUFE_bm);
+	ads7871_write_reg8(spi, REG_OSC_CONTROL, val);
+	ret = ads7871_read_reg8(spi, REG_OSC_CONTROL);
+
+	dev_dbg(&spi->dev, "REG_OSC_CONTROL write:%x, read:%x\n", val, ret);
+	/*because there is no other error checking on an SPI bus
+	we need to make sure we really have a chip*/
+	if (val != ret) {
+		err = -ENODEV;
+		goto error_remove;
+	}
+
+	return 0;
+
+error_remove:
+	sysfs_remove_group(&spi->dev.kobj, &ads7871_group);
+error_free:
+	kfree(pdata);
+exit:
+	return err;
+}
+
+static int __devexit ads7871_remove(struct spi_device *spi)
+{
+	struct ads7871_data *pdata = spi_get_drvdata(spi);
+
+	hwmon_device_unregister(pdata->hwmon_dev);
+	sysfs_remove_group(&spi->dev.kobj, &ads7871_group);
+	kfree(pdata);
+	return 0;
+}
+
+static struct spi_driver ads7871_driver = {
+	.driver = {
+		.name = DEVICE_NAME,
+		.bus = &spi_bus_type,
+		.owner = THIS_MODULE,
+	},
+
+	.probe = ads7871_probe,
+	.remove = __devexit_p(ads7871_remove),
+};
+
+static int __init ads7871_init(void)
+{
+	return spi_register_driver(&ads7871_driver);
+}
+
+static void __exit ads7871_exit(void)
+{
+	spi_unregister_driver(&ads7871_driver);
+}
+
+module_init(ads7871_init);
+module_exit(ads7871_exit);
+
+MODULE_AUTHOR("Paul Thomas <pthomas8589@gmail.com>");
+MODULE_DESCRIPTION("TI ADS7871 A/D driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/hwmon/adt7411.c b/drivers/hwmon/adt7411.c
index 4086c7257f91..f13c843a2964 100644
--- a/drivers/hwmon/adt7411.c
+++ b/drivers/hwmon/adt7411.c
@@ -316,7 +316,6 @@ static int __devinit adt7411_probe(struct i2c_client *client,
  exit_remove:
 	sysfs_remove_group(&client->dev.kobj, &adt7411_attr_grp);
  exit_free:
-	i2c_set_clientdata(client, NULL);
 	kfree(data);
 	return ret;
 }
@@ -327,7 +326,6 @@ static int __devexit adt7411_remove(struct i2c_client *client)
 
 	hwmon_device_unregister(data->hwmon_dev);
 	sysfs_remove_group(&client->dev.kobj, &adt7411_attr_grp);
-	i2c_set_clientdata(client, NULL);
 	kfree(data);
 	return 0;
 }
diff --git a/drivers/hwmon/applesmc.c b/drivers/hwmon/applesmc.c
index f085c18d2905..b6598aa557a0 100644
--- a/drivers/hwmon/applesmc.c
+++ b/drivers/hwmon/applesmc.c
@@ -148,6 +148,20 @@ static const char *temperature_sensors_sets[][41] = {
 /* Set 18: MacBook Pro 2,2 */
 	{ "TB0T", "TC0D", "TC0P", "TG0H", "TG0P", "TG0T", "TM0P", "TTF0",
 	  "Th0H", "Th1H", "Tm0P", "Ts0P", NULL },
+/* Set 19: Macbook Pro 5,3 */
+	{ "TB0T", "TB1T", "TB2T", "TB3T", "TC0D", "TC0F", "TC0P", "TG0D",
+	  "TG0F", "TG0H", "TG0P", "TG0T", "TN0D", "TN0P", "TTF0", "Th2H",
+	  "Tm0P", "Ts0P", "Ts0S", NULL },
+/* Set 20: MacBook Pro 5,4 */
+	{ "TB0T", "TB1T", "TB2T", "TB3T", "TC0D", "TC0F", "TC0P", "TN0D",
+	  "TN0P", "TTF0", "Th2H", "Ts0P", "Ts0S", NULL },
+/* Set 21: MacBook Pro 6,2 */
+	{ "TB0T", "TB1T", "TB2T", "TC0C", "TC0D", "TC0P", "TC1C", "TG0D",
+	  "TG0P", "TG0T", "TMCD", "TP0P", "TPCD", "Th1H", "Th2H", "Tm0P",
+	  "Ts0P", "Ts0S", NULL },
+/* Set 22: MacBook Pro 7,1 */
+	{ "TB0T", "TB1T", "TB2T", "TC0D", "TC0P", "TN0D", "TN0P", "TN0S",
+	  "TN1D", "TN1F", "TN1G", "TN1S", "Th1H", "Ts0P", "Ts0S", NULL },
 };
 
 /* List of keys used to read/write fan speeds */
@@ -646,6 +660,17 @@ out:
 		return snprintf(sysfsbuf, PAGE_SIZE, "(%d,%d)\n", left, right);
 }
 
+/* Displays sensor key as label */
+static ssize_t applesmc_show_sensor_label(struct device *dev,
+			struct device_attribute *devattr, char *sysfsbuf)
+{
+	struct sensor_device_attribute *attr = to_sensor_dev_attr(devattr);
+	const char *key =
+		temperature_sensors_sets[applesmc_temperature_set][attr->index];
+
+	return snprintf(sysfsbuf, PAGE_SIZE, "%s\n", key);
+}
+
 /* Displays degree Celsius * 1000 */
 static ssize_t applesmc_show_temperature(struct device *dev,
 			struct device_attribute *devattr, char *sysfsbuf)
@@ -1113,6 +1138,86 @@ static const struct attribute_group fan_attribute_groups[] = {
 /*
  * Temperature sensors sysfs entries.
  */
+static SENSOR_DEVICE_ATTR(temp1_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 0);
+static SENSOR_DEVICE_ATTR(temp2_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 1);
+static SENSOR_DEVICE_ATTR(temp3_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 2);
+static SENSOR_DEVICE_ATTR(temp4_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 3);
+static SENSOR_DEVICE_ATTR(temp5_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 4);
+static SENSOR_DEVICE_ATTR(temp6_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 5);
+static SENSOR_DEVICE_ATTR(temp7_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 6);
+static SENSOR_DEVICE_ATTR(temp8_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 7);
+static SENSOR_DEVICE_ATTR(temp9_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 8);
+static SENSOR_DEVICE_ATTR(temp10_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 9);
+static SENSOR_DEVICE_ATTR(temp11_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 10);
+static SENSOR_DEVICE_ATTR(temp12_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 11);
+static SENSOR_DEVICE_ATTR(temp13_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 12);
+static SENSOR_DEVICE_ATTR(temp14_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 13);
+static SENSOR_DEVICE_ATTR(temp15_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 14);
+static SENSOR_DEVICE_ATTR(temp16_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 15);
+static SENSOR_DEVICE_ATTR(temp17_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 16);
+static SENSOR_DEVICE_ATTR(temp18_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 17);
+static SENSOR_DEVICE_ATTR(temp19_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 18);
+static SENSOR_DEVICE_ATTR(temp20_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 19);
+static SENSOR_DEVICE_ATTR(temp21_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 20);
+static SENSOR_DEVICE_ATTR(temp22_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 21);
+static SENSOR_DEVICE_ATTR(temp23_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 22);
+static SENSOR_DEVICE_ATTR(temp24_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 23);
+static SENSOR_DEVICE_ATTR(temp25_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 24);
+static SENSOR_DEVICE_ATTR(temp26_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 25);
+static SENSOR_DEVICE_ATTR(temp27_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 26);
+static SENSOR_DEVICE_ATTR(temp28_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 27);
+static SENSOR_DEVICE_ATTR(temp29_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 28);
+static SENSOR_DEVICE_ATTR(temp30_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 29);
+static SENSOR_DEVICE_ATTR(temp31_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 30);
+static SENSOR_DEVICE_ATTR(temp32_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 31);
+static SENSOR_DEVICE_ATTR(temp33_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 32);
+static SENSOR_DEVICE_ATTR(temp34_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 33);
+static SENSOR_DEVICE_ATTR(temp35_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 34);
+static SENSOR_DEVICE_ATTR(temp36_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 35);
+static SENSOR_DEVICE_ATTR(temp37_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 36);
+static SENSOR_DEVICE_ATTR(temp38_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 37);
+static SENSOR_DEVICE_ATTR(temp39_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 38);
+static SENSOR_DEVICE_ATTR(temp40_label, S_IRUGO,
+					applesmc_show_sensor_label, NULL, 39);
 static SENSOR_DEVICE_ATTR(temp1_input, S_IRUGO,
 					applesmc_show_temperature, NULL, 0);
 static SENSOR_DEVICE_ATTR(temp2_input, S_IRUGO,
@@ -1194,6 +1299,50 @@ static SENSOR_DEVICE_ATTR(temp39_input, S_IRUGO,
 static SENSOR_DEVICE_ATTR(temp40_input, S_IRUGO,
 					applesmc_show_temperature, NULL, 39);
 
+static struct attribute *label_attributes[] = {
+	&sensor_dev_attr_temp1_label.dev_attr.attr,
+	&sensor_dev_attr_temp2_label.dev_attr.attr,
+	&sensor_dev_attr_temp3_label.dev_attr.attr,
+	&sensor_dev_attr_temp4_label.dev_attr.attr,
+	&sensor_dev_attr_temp5_label.dev_attr.attr,
+	&sensor_dev_attr_temp6_label.dev_attr.attr,
+	&sensor_dev_attr_temp7_label.dev_attr.attr,
+	&sensor_dev_attr_temp8_label.dev_attr.attr,
+	&sensor_dev_attr_temp9_label.dev_attr.attr,
+	&sensor_dev_attr_temp10_label.dev_attr.attr,
+	&sensor_dev_attr_temp11_label.dev_attr.attr,
+	&sensor_dev_attr_temp12_label.dev_attr.attr,
+	&sensor_dev_attr_temp13_label.dev_attr.attr,
+	&sensor_dev_attr_temp14_label.dev_attr.attr,
+	&sensor_dev_attr_temp15_label.dev_attr.attr,
+	&sensor_dev_attr_temp16_label.dev_attr.attr,
+	&sensor_dev_attr_temp17_label.dev_attr.attr,
+	&sensor_dev_attr_temp18_label.dev_attr.attr,
+	&sensor_dev_attr_temp19_label.dev_attr.attr,
+	&sensor_dev_attr_temp20_label.dev_attr.attr,
+	&sensor_dev_attr_temp21_label.dev_attr.attr,
+	&sensor_dev_attr_temp22_label.dev_attr.attr,
+	&sensor_dev_attr_temp23_label.dev_attr.attr,
+	&sensor_dev_attr_temp24_label.dev_attr.attr,
+	&sensor_dev_attr_temp25_label.dev_attr.attr,
+	&sensor_dev_attr_temp26_label.dev_attr.attr,
+	&sensor_dev_attr_temp27_label.dev_attr.attr,
+	&sensor_dev_attr_temp28_label.dev_attr.attr,
+	&sensor_dev_attr_temp29_label.dev_attr.attr,
+	&sensor_dev_attr_temp30_label.dev_attr.attr,
+	&sensor_dev_attr_temp31_label.dev_attr.attr,
+	&sensor_dev_attr_temp32_label.dev_attr.attr,
+	&sensor_dev_attr_temp33_label.dev_attr.attr,
+	&sensor_dev_attr_temp34_label.dev_attr.attr,
+	&sensor_dev_attr_temp35_label.dev_attr.attr,
+	&sensor_dev_attr_temp36_label.dev_attr.attr,
+	&sensor_dev_attr_temp37_label.dev_attr.attr,
+	&sensor_dev_attr_temp38_label.dev_attr.attr,
+	&sensor_dev_attr_temp39_label.dev_attr.attr,
+	&sensor_dev_attr_temp40_label.dev_attr.attr,
+	NULL
+};
+
 static struct attribute *temperature_attributes[] = {
 	&sensor_dev_attr_temp1_input.dev_attr.attr,
 	&sensor_dev_attr_temp2_input.dev_attr.attr,
@@ -1241,6 +1390,10 @@ static struct attribute *temperature_attributes[] = {
 static const struct attribute_group temperature_attributes_group =
 	{ .attrs = temperature_attributes };
 
+static const struct attribute_group label_attributes_group = {
+	.attrs = label_attributes
+};
+
 /* Module stuff */
 
 /*
@@ -1363,6 +1516,14 @@ static __initdata struct dmi_match_data applesmc_dmi_data[] = {
 	{ .accelerometer = 0, .light = 0, .temperature_set = 17 },
 /* MacBook Pro 2,2: accelerometer, backlight and temperature set 18 */
 	{ .accelerometer = 1, .light = 1, .temperature_set = 18 },
+/* MacBook Pro 5,3: accelerometer, backlight and temperature set 19 */
+	{ .accelerometer = 1, .light = 1, .temperature_set = 19 },
+/* MacBook Pro 5,4: accelerometer, backlight and temperature set 20 */
+	{ .accelerometer = 1, .light = 1, .temperature_set = 20 },
+/* MacBook Pro 6,2: accelerometer, backlight and temperature set 21 */
+	{ .accelerometer = 1, .light = 1, .temperature_set = 21 },
+/* MacBook Pro 7,1: accelerometer, backlight and temperature set 22 */
+	{ .accelerometer = 1, .light = 1, .temperature_set = 22 },
 };
 
 /* Note that DMI_MATCH(...,"MacBook") will match "MacBookPro1,1".
@@ -1376,6 +1537,22 @@ static __initdata struct dmi_system_id applesmc_whitelist[] = {
 	  DMI_MATCH(DMI_BOARD_VENDOR, "Apple"),
 	  DMI_MATCH(DMI_PRODUCT_NAME, "MacBookAir") },
 		&applesmc_dmi_data[7]},
+	{ applesmc_dmi_match, "Apple MacBook Pro 7", {
+	  DMI_MATCH(DMI_BOARD_VENDOR, "Apple"),
+	  DMI_MATCH(DMI_PRODUCT_NAME, "MacBookPro7") },
+		&applesmc_dmi_data[22]},
+	{ applesmc_dmi_match, "Apple MacBook Pro 5,4", {
+	  DMI_MATCH(DMI_BOARD_VENDOR, "Apple"),
+	  DMI_MATCH(DMI_PRODUCT_NAME, "MacBookPro5,4") },
+		&applesmc_dmi_data[20]},
+	{ applesmc_dmi_match, "Apple MacBook Pro 5,3", {
+	  DMI_MATCH(DMI_BOARD_VENDOR, "Apple"),
+	  DMI_MATCH(DMI_PRODUCT_NAME, "MacBookPro5,3") },
+		&applesmc_dmi_data[19]},
+	{ applesmc_dmi_match, "Apple MacBook Pro 6", {
+	  DMI_MATCH(DMI_BOARD_VENDOR, "Apple"),
+	  DMI_MATCH(DMI_PRODUCT_NAME, "MacBookPro6") },
+		&applesmc_dmi_data[21]},
 	{ applesmc_dmi_match, "Apple MacBook Pro 5", {
 	  DMI_MATCH(DMI_BOARD_VENDOR, "Apple"),
 	  DMI_MATCH(DMI_PRODUCT_NAME, "MacBookPro5") },
@@ -1518,7 +1695,8 @@ static int __init applesmc_init(void)
 	for (i = 0;
 	     temperature_sensors_sets[applesmc_temperature_set][i] != NULL;
 	     i++) {
-		if (temperature_attributes[i] == NULL) {
+		if (temperature_attributes[i] == NULL ||
+		    label_attributes[i] == NULL) {
 			printk(KERN_ERR "applesmc: More temperature sensors "
 				"in temperature_sensors_sets (at least %i)"
 				"than available sysfs files in "
@@ -1530,6 +1708,10 @@ static int __init applesmc_init(void)
 						temperature_attributes[i]);
 		if (ret)
 			goto out_temperature;
+		ret = sysfs_create_file(&pdev->dev.kobj,
+						label_attributes[i]);
+		if (ret)
+			goto out_temperature;
 	}
 
 	if (applesmc_accelerometer) {
@@ -1580,6 +1762,7 @@ out_accelerometer:
 	if (applesmc_accelerometer)
 		applesmc_release_accelerometer();
 out_temperature:
+	sysfs_remove_group(&pdev->dev.kobj, &label_attributes_group);
 	sysfs_remove_group(&pdev->dev.kobj, &temperature_attributes_group);
 out_fans:
 	while (fans_handled)
@@ -1609,6 +1792,7 @@ static void __exit applesmc_exit(void)
 	}
 	if (applesmc_accelerometer)
 		applesmc_release_accelerometer();
+	sysfs_remove_group(&pdev->dev.kobj, &label_attributes_group);
 	sysfs_remove_group(&pdev->dev.kobj, &temperature_attributes_group);
 	while (fans_handled)
 		sysfs_remove_group(&pdev->dev.kobj,
diff --git a/drivers/hwmon/asc7621.c b/drivers/hwmon/asc7621.c
index 0f388adc6187..3b973f30b1f6 100644
--- a/drivers/hwmon/asc7621.c
+++ b/drivers/hwmon/asc7621.c
@@ -1141,7 +1141,6 @@ exit_remove:
 				   &(asc7621_params[i].sda.dev_attr));
 	}
 
-	i2c_set_clientdata(client, NULL);
 	kfree(data);
 	return err;
 }
@@ -1196,7 +1195,6 @@ static int asc7621_remove(struct i2c_client *client)
 				   &(asc7621_params[i].sda.dev_attr));
 	}
 
-	i2c_set_clientdata(client, NULL);
 	kfree(data);
 	return 0;
 }
diff --git a/drivers/hwmon/asus_atk0110.c b/drivers/hwmon/asus_atk0110.c
index 16c420240724..653db1bda934 100644
--- a/drivers/hwmon/asus_atk0110.c
+++ b/drivers/hwmon/asus_atk0110.c
@@ -1411,6 +1411,13 @@ static int __init atk0110_init(void)
 {
 	int ret;
 
+	/* Make sure it's safe to access the device through ACPI */
+	if (!acpi_resources_are_enforced()) {
+		pr_err("atk: Resources not safely usable due to "
+		       "acpi_enforce_resources kernel parameter\n");
+		return -EBUSY;
+	}
+
 	ret = acpi_bus_register_driver(&atk_driver);
 	if (ret)
 		pr_info("atk: acpi_bus_register_driver failed: %d\n", ret);
diff --git a/drivers/hwmon/coretemp.c b/drivers/hwmon/coretemp.c
index e9b7fbc5a447..2988da150ed6 100644
--- a/drivers/hwmon/coretemp.c
+++ b/drivers/hwmon/coretemp.c
@@ -241,6 +241,55 @@ static int __devinit adjust_tjmax(struct cpuinfo_x86 *c, u32 id, struct device *
 	return tjmax;
 }
 
+static int __devinit get_tjmax(struct cpuinfo_x86 *c, u32 id,
+			       struct device *dev)
+{
+	/* The 100C is default for both mobile and non mobile CPUs */
+	int err;
+	u32 eax, edx;
+	u32 val;
+
+	/* A new feature of current Intel(R) processors, the
+	   IA32_TEMPERATURE_TARGET contains the TjMax value */
+	err = rdmsr_safe_on_cpu(id, MSR_IA32_TEMPERATURE_TARGET, &eax, &edx);
+	if (err) {
+		dev_warn(dev, "Unable to read TjMax from CPU.\n");
+	} else {
+		val = (eax >> 16) & 0xff;
+		/*
+		 * If the TjMax is not plausible, an assumption
+		 * will be used
+		 */
+		if ((val > 80) && (val < 120)) {
+			dev_info(dev, "TjMax is %d C.\n", val);
+			return val * 1000;
+		}
+	}
+
+	/*
+	 * An assumption is made for early CPUs and unreadable MSR.
+	 * NOTE: the given value may not be correct.
+	 */
+
+	switch (c->x86_model) {
+	case 0xe:
+	case 0xf:
+	case 0x16:
+	case 0x1a:
+		dev_warn(dev, "TjMax is assumed as 100 C!\n");
+		return 100000;
+		break;
+	case 0x17:
+	case 0x1c:		/* Atom CPUs */
+		return adjust_tjmax(c, id, dev);
+		break;
+	default:
+		dev_warn(dev, "CPU (model=0x%x) is not supported yet,"
+			" using default TjMax of 100C.\n", c->x86_model);
+		return 100000;
+	}
+}
+
 static int __devinit coretemp_probe(struct platform_device *pdev)
 {
 	struct coretemp_data *data;
@@ -283,14 +332,18 @@ static int __devinit coretemp_probe(struct platform_device *pdev)
 		}
 	}
 
-	data->tjmax = adjust_tjmax(c, data->id, &pdev->dev);
+	data->tjmax = get_tjmax(c, data->id, &pdev->dev);
 	platform_set_drvdata(pdev, data);
 
-	/* read the still undocumented IA32_TEMPERATURE_TARGET it exists
-	   on older CPUs but not in this register, Atoms don't have it either */
+	/*
+	 * read the still undocumented IA32_TEMPERATURE_TARGET. It exists
+	 * on older CPUs but not in this register,
+	 * Atoms don't have it either.
+	 */
 
 	if ((c->x86_model > 0xe) && (c->x86_model != 0x1c)) {
-		err = rdmsr_safe_on_cpu(data->id, 0x1a2, &eax, &edx);
+		err = rdmsr_safe_on_cpu(data->id, MSR_IA32_TEMPERATURE_TARGET,
+		    &eax, &edx);
 		if (err) {
 			dev_warn(&pdev->dev, "Unable to read"
 					" IA32_TEMPERATURE_TARGET MSR\n");
@@ -451,28 +504,20 @@ static int __init coretemp_init(void)
 
 	for_each_online_cpu(i) {
 		struct cpuinfo_x86 *c = &cpu_data(i);
+		/*
+		 * CPUID.06H.EAX[0] indicates whether the CPU has thermal
+		 * sensors. We check this bit only, all the early CPUs
+		 * without thermal sensors will be filtered out.
+		 */
+		if (c->cpuid_level >= 6 && (cpuid_eax(0x06) & 0x01)) {
+			err = coretemp_device_add(i);
+			if (err)
+				goto exit_devices_unreg;
 
-		/* check if family 6, models 0xe (Pentium M DC),
-		  0xf (Core 2 DC 65nm), 0x16 (Core 2 SC 65nm),
-		  0x17 (Penryn 45nm), 0x1a (Nehalem), 0x1c (Atom),
-		  0x1e (Lynnfield) */
-		if ((c->cpuid_level < 0) || (c->x86 != 0x6) ||
-		    !((c->x86_model == 0xe) || (c->x86_model == 0xf) ||
-			(c->x86_model == 0x16) || (c->x86_model == 0x17) ||
-			(c->x86_model == 0x1a) || (c->x86_model == 0x1c) ||
-			(c->x86_model == 0x1e))) {
-
-			/* supported CPU not found, but report the unknown
-			   family 6 CPU */
-			if ((c->x86 == 0x6) && (c->x86_model > 0xf))
-				printk(KERN_WARNING DRVNAME ": Unknown CPU "
-					"model 0x%x\n", c->x86_model);
-			continue;
+		} else {
+			printk(KERN_INFO DRVNAME ": CPU (model=0x%x)"
+				" has no thermal sensor.\n", c->x86_model);
 		}
-
-		err = coretemp_device_add(i);
-		if (err)
-			goto exit_devices_unreg;
 	}
 	if (list_empty(&pdev_list)) {
 		err = -ENODEV;
diff --git a/drivers/hwmon/dme1737.c b/drivers/hwmon/dme1737.c
index 823dd28a902c..980c17d5eeae 100644
--- a/drivers/hwmon/dme1737.c
+++ b/drivers/hwmon/dme1737.c
@@ -1,12 +1,14 @@
 /*
- * dme1737.c - Driver for the SMSC DME1737, Asus A8000, SMSC SCH311x and
- *             SCH5027 Super-I/O chips integrated hardware monitoring features.
- * Copyright (c) 2007, 2008 Juerg Haefliger <juergh@gmail.com>
+ * dme1737.c - Driver for the SMSC DME1737, Asus A8000, SMSC SCH311x, SCH5027,
+ *             and SCH5127 Super-I/O chips integrated hardware monitoring
+ *             features.
+ * Copyright (c) 2007, 2008, 2009, 2010 Juerg Haefliger <juergh@gmail.com>
  *
  * This driver is an I2C/ISA hybrid, meaning that it uses the I2C bus to access
  * the chip registers if a DME1737, A8000, or SCH5027 is found and the ISA bus
- * if a SCH311x chip is found. Both types of chips have very similar hardware
- * monitoring capabilities but differ in the way they can be accessed.
+ * if a SCH311x or SCH5127 chip is found. Both types of chips have very
+ * similar hardware monitoring capabilities but differ in the way they can be
+ * accessed.
  *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
@@ -57,7 +59,7 @@ MODULE_PARM_DESC(probe_all_addr, "Include probing of non-standard LPC "
 /* Addresses to scan */
 static const unsigned short normal_i2c[] = {0x2c, 0x2d, 0x2e, I2C_CLIENT_END};
 
-enum chips { dme1737, sch5027, sch311x };
+enum chips { dme1737, sch5027, sch311x, sch5127 };
 
 /* ---------------------------------------------------------------------
  * Registers
@@ -164,10 +166,29 @@ static const u8 DME1737_BIT_ALARM_FAN[] = {10, 11, 12, 13, 22, 23};
 #define DME1737_VERSTEP_MASK	0xf8
 #define SCH311X_DEVICE		0x8c
 #define SCH5027_VERSTEP		0x69
+#define SCH5127_DEVICE		0x8e
+
+/* Device ID values (global configuration register index 0x20) */
+#define DME1737_ID_1	0x77
+#define DME1737_ID_2	0x78
+#define SCH3112_ID	0x7c
+#define SCH3114_ID	0x7d
+#define SCH3116_ID	0x7f
+#define SCH5027_ID	0x89
+#define SCH5127_ID	0x86
 
 /* Length of ISA address segment */
 #define DME1737_EXTENT	2
 
+/* chip-dependent features */
+#define HAS_TEMP_OFFSET		(1 << 0)		/* bit 0 */
+#define HAS_VID			(1 << 1)		/* bit 1 */
+#define HAS_ZONE3		(1 << 2)		/* bit 2 */
+#define HAS_ZONE_HYST		(1 << 3)		/* bit 3 */
+#define HAS_PWM_MIN		(1 << 4)		/* bit 4 */
+#define HAS_FAN(ix)		(1 << ((ix) + 5))	/* bits 5-10 */
+#define HAS_PWM(ix)		(1 << ((ix) + 11))	/* bits 11-16 */
+
 /* ---------------------------------------------------------------------
  * Data structures and manipulation thereof
  * --------------------------------------------------------------------- */
@@ -187,8 +208,7 @@ struct dme1737_data {
 
 	u8 vid;
 	u8 pwm_rr_en;
-	u8 has_pwm;
-	u8 has_fan;
+	u32 has_features;
 
 	/* Register values */
 	u16 in[7];
@@ -224,8 +244,11 @@ static const int IN_NOMINAL_SCH311x[] = {2500, 1500, 3300, 5000, 12000, 3300,
 					 3300};
 static const int IN_NOMINAL_SCH5027[] = {5000, 2250, 3300, 1125, 1125, 3300,
 					 3300};
+static const int IN_NOMINAL_SCH5127[] = {2500, 2250, 3300, 1125, 1125, 3300,
+					 3300};
 #define IN_NOMINAL(type)	((type) == sch311x ? IN_NOMINAL_SCH311x : \
 				 (type) == sch5027 ? IN_NOMINAL_SCH5027 : \
+				 (type) == sch5127 ? IN_NOMINAL_SCH5127 : \
 				 IN_NOMINAL_DME1737)
 
 /* Voltage input
@@ -568,7 +591,7 @@ static struct dme1737_data *dme1737_update_device(struct device *dev)
 
 	/* Sample register contents every 1 sec */
 	if (time_after(jiffies, data->last_update + HZ) || !data->valid) {
-		if (data->type == dme1737) {
+		if (data->has_features & HAS_VID) {
 			data->vid = dme1737_read(data, DME1737_REG_VID) &
 				0x3f;
 		}
@@ -599,7 +622,7 @@ static struct dme1737_data *dme1737_update_device(struct device *dev)
 					DME1737_REG_TEMP_MIN(ix));
 			data->temp_max[ix] = dme1737_read(data,
 					DME1737_REG_TEMP_MAX(ix));
-			if (data->type != sch5027) {
+			if (data->has_features & HAS_TEMP_OFFSET) {
 				data->temp_offset[ix] = dme1737_read(data,
 						DME1737_REG_TEMP_OFFSET(ix));
 			}
@@ -626,7 +649,7 @@ static struct dme1737_data *dme1737_update_device(struct device *dev)
 		for (ix = 0; ix < ARRAY_SIZE(data->fan); ix++) {
 			/* Skip reading registers if optional fans are not
 			 * present */
-			if (!(data->has_fan & (1 << ix))) {
+			if (!(data->has_features & HAS_FAN(ix))) {
 				continue;
 			}
 			data->fan[ix] = dme1737_read(data,
@@ -650,7 +673,7 @@ static struct dme1737_data *dme1737_update_device(struct device *dev)
 		for (ix = 0; ix < ARRAY_SIZE(data->pwm); ix++) {
 			/* Skip reading registers if optional PWMs are not
 			 * present */
-			if (!(data->has_pwm & (1 << ix))) {
+			if (!(data->has_features & HAS_PWM(ix))) {
 				continue;
 			}
 			data->pwm[ix] = dme1737_read(data,
@@ -672,12 +695,24 @@ static struct dme1737_data *dme1737_update_device(struct device *dev)
 
 		/* Thermal zone registers */
 		for (ix = 0; ix < ARRAY_SIZE(data->zone_low); ix++) {
-			data->zone_low[ix] = dme1737_read(data,
-					DME1737_REG_ZONE_LOW(ix));
-			data->zone_abs[ix] = dme1737_read(data,
-					DME1737_REG_ZONE_ABS(ix));
+			/* Skip reading registers if zone3 is not present */
+			if ((ix == 2) && !(data->has_features & HAS_ZONE3)) {
+				continue;
+			}
+			/* sch5127 zone2 registers are special */
+			if ((ix == 1) && (data->type == sch5127)) {
+				data->zone_low[1] = dme1737_read(data,
+						DME1737_REG_ZONE_LOW(2));
+				data->zone_abs[1] = dme1737_read(data,
+						DME1737_REG_ZONE_ABS(2));
+			} else {
+				data->zone_low[ix] = dme1737_read(data,
+						DME1737_REG_ZONE_LOW(ix));
+				data->zone_abs[ix] = dme1737_read(data,
+						DME1737_REG_ZONE_ABS(ix));
+			}
 		}
-		if (data->type != sch5027) {
+		if (data->has_features & HAS_ZONE_HYST) {
 			for (ix = 0; ix < ARRAY_SIZE(data->zone_hyst); ix++) {
 				data->zone_hyst[ix] = dme1737_read(data,
 						DME1737_REG_ZONE_HYST(ix));
@@ -1594,10 +1629,6 @@ static struct attribute *dme1737_attr[] ={
 	&sensor_dev_attr_zone2_auto_point2_temp.dev_attr.attr,
 	&sensor_dev_attr_zone2_auto_point3_temp.dev_attr.attr,
 	&sensor_dev_attr_zone2_auto_channels_temp.dev_attr.attr,
-	&sensor_dev_attr_zone3_auto_point1_temp.dev_attr.attr,
-	&sensor_dev_attr_zone3_auto_point2_temp.dev_attr.attr,
-	&sensor_dev_attr_zone3_auto_point3_temp.dev_attr.attr,
-	&sensor_dev_attr_zone3_auto_channels_temp.dev_attr.attr,
 	NULL
 };
 
@@ -1605,27 +1636,23 @@ static const struct attribute_group dme1737_group = {
 	.attrs = dme1737_attr,
 };
 
-/* The following struct holds misc attributes, which are not available in all
- * chips. Their creation depends on the chip type which is determined during
- * module load. */
-static struct attribute *dme1737_misc_attr[] = {
-	/* Temperatures */
+/* The following struct holds temp offset attributes, which are not available
+ * in all chips. The following chips support them:
+ * DME1737, SCH311x */
+static struct attribute *dme1737_temp_offset_attr[] = {
 	&sensor_dev_attr_temp1_offset.dev_attr.attr,
 	&sensor_dev_attr_temp2_offset.dev_attr.attr,
 	&sensor_dev_attr_temp3_offset.dev_attr.attr,
-	/* Zones */
-	&sensor_dev_attr_zone1_auto_point1_temp_hyst.dev_attr.attr,
-	&sensor_dev_attr_zone2_auto_point1_temp_hyst.dev_attr.attr,
-	&sensor_dev_attr_zone3_auto_point1_temp_hyst.dev_attr.attr,
 	NULL
 };
 
-static const struct attribute_group dme1737_misc_group = {
-	.attrs = dme1737_misc_attr,
+static const struct attribute_group dme1737_temp_offset_group = {
+	.attrs = dme1737_temp_offset_attr,
 };
 
-/* The following struct holds VID-related attributes. Their creation
-   depends on the chip type which is determined during module load. */
+/* The following struct holds VID related attributes, which are not available
+ * in all chips. The following chips support them:
+ * DME1737 */
 static struct attribute *dme1737_vid_attr[] = {
 	&dev_attr_vrm.attr,
 	&dev_attr_cpu0_vid.attr,
@@ -1636,6 +1663,36 @@ static const struct attribute_group dme1737_vid_group = {
 	.attrs = dme1737_vid_attr,
 };
 
+/* The following struct holds temp zone 3 related attributes, which are not
+ * available in all chips. The following chips support them:
+ * DME1737, SCH311x, SCH5027 */
+static struct attribute *dme1737_zone3_attr[] = {
+	&sensor_dev_attr_zone3_auto_point1_temp.dev_attr.attr,
+	&sensor_dev_attr_zone3_auto_point2_temp.dev_attr.attr,
+	&sensor_dev_attr_zone3_auto_point3_temp.dev_attr.attr,
+	&sensor_dev_attr_zone3_auto_channels_temp.dev_attr.attr,
+	NULL
+};
+
+static const struct attribute_group dme1737_zone3_group = {
+	.attrs = dme1737_zone3_attr,
+};
+
+
+/* The following struct holds temp zone hysteresis  related attributes, which
+ * are not available in all chips. The following chips support them:
+ * DME1737, SCH311x */
+static struct attribute *dme1737_zone_hyst_attr[] = {
+	&sensor_dev_attr_zone1_auto_point1_temp_hyst.dev_attr.attr,
+	&sensor_dev_attr_zone2_auto_point1_temp_hyst.dev_attr.attr,
+	&sensor_dev_attr_zone3_auto_point1_temp_hyst.dev_attr.attr,
+	NULL
+};
+
+static const struct attribute_group dme1737_zone_hyst_group = {
+	.attrs = dme1737_zone_hyst_attr,
+};
+
 /* The following structs hold the PWM attributes, some of which are optional.
  * Their creation depends on the chip configuration which is determined during
  * module load. */
@@ -1691,10 +1748,10 @@ static const struct attribute_group dme1737_pwm_group[] = {
 	{ .attrs = dme1737_pwm6_attr },
 };
 
-/* The following struct holds misc PWM attributes, which are not available in
- * all chips. Their creation depends on the chip type which is determined
+/* The following struct holds auto PWM min attributes, which are not available
+ * in all chips. Their creation depends on the chip type which is determined
  * during module load. */
-static struct attribute *dme1737_pwm_misc_attr[] = {
+static struct attribute *dme1737_auto_pwm_min_attr[] = {
 	&sensor_dev_attr_pwm1_auto_pwm_min.dev_attr.attr,
 	&sensor_dev_attr_pwm2_auto_pwm_min.dev_attr.attr,
 	&sensor_dev_attr_pwm3_auto_pwm_min.dev_attr.attr,
@@ -1764,14 +1821,25 @@ static struct attribute *dme1737_zone_chmod_attr[] = {
 	&sensor_dev_attr_zone2_auto_point1_temp.dev_attr.attr,
 	&sensor_dev_attr_zone2_auto_point2_temp.dev_attr.attr,
 	&sensor_dev_attr_zone2_auto_point3_temp.dev_attr.attr,
+	NULL
+};
+
+static const struct attribute_group dme1737_zone_chmod_group = {
+	.attrs = dme1737_zone_chmod_attr,
+};
+
+
+/* The permissions of the following zone 3 attributes are changed to read-
+ * writeable if the chip is *not* locked. Otherwise they stay read-only. */
+static struct attribute *dme1737_zone3_chmod_attr[] = {
 	&sensor_dev_attr_zone3_auto_point1_temp.dev_attr.attr,
 	&sensor_dev_attr_zone3_auto_point2_temp.dev_attr.attr,
 	&sensor_dev_attr_zone3_auto_point3_temp.dev_attr.attr,
 	NULL
 };
 
-static const struct attribute_group dme1737_zone_chmod_group = {
-	.attrs = dme1737_zone_chmod_attr,
+static const struct attribute_group dme1737_zone3_chmod_group = {
+	.attrs = dme1737_zone3_chmod_attr,
 };
 
 /* The permissions of the following PWM attributes are changed to read-
@@ -1887,30 +1955,35 @@ static void dme1737_remove_files(struct device *dev)
 	int ix;
 
 	for (ix = 0; ix < ARRAY_SIZE(dme1737_fan_group); ix++) {
-		if (data->has_fan & (1 << ix)) {
+		if (data->has_features & HAS_FAN(ix)) {
 			sysfs_remove_group(&dev->kobj,
 					   &dme1737_fan_group[ix]);
 		}
 	}
 
 	for (ix = 0; ix < ARRAY_SIZE(dme1737_pwm_group); ix++) {
-		if (data->has_pwm & (1 << ix)) {
+		if (data->has_features & HAS_PWM(ix)) {
 			sysfs_remove_group(&dev->kobj,
 					   &dme1737_pwm_group[ix]);
-			if (data->type != sch5027 && ix < 3) {
+			if ((data->has_features & HAS_PWM_MIN) && ix < 3) {
 				sysfs_remove_file(&dev->kobj,
-						  dme1737_pwm_misc_attr[ix]);
+						dme1737_auto_pwm_min_attr[ix]);
 			}
 		}
 	}
 
-	if (data->type != sch5027) {
-		sysfs_remove_group(&dev->kobj, &dme1737_misc_group);
+	if (data->has_features & HAS_TEMP_OFFSET) {
+		sysfs_remove_group(&dev->kobj, &dme1737_temp_offset_group);
 	}
-	if (data->type == dme1737) {
+	if (data->has_features & HAS_VID) {
 		sysfs_remove_group(&dev->kobj, &dme1737_vid_group);
 	}
-
+	if (data->has_features & HAS_ZONE3) {
+		sysfs_remove_group(&dev->kobj, &dme1737_zone3_group);
+	}
+	if (data->has_features & HAS_ZONE_HYST) {
+		sysfs_remove_group(&dev->kobj, &dme1737_zone_hyst_group);
+	}
 	sysfs_remove_group(&dev->kobj, &dme1737_group);
 
 	if (!data->client) {
@@ -1934,23 +2007,31 @@ static int dme1737_create_files(struct device *dev)
 		goto exit_remove;
 	}
 
-	/* Create misc sysfs attributes */
-	if ((data->type != sch5027) &&
+	/* Create chip-dependent sysfs attributes */
+	if ((data->has_features & HAS_TEMP_OFFSET) &&
 	    (err = sysfs_create_group(&dev->kobj,
-				      &dme1737_misc_group))) {
+				      &dme1737_temp_offset_group))) {
 		goto exit_remove;
 	}
-
-	/* Create VID-related sysfs attributes */
-	if ((data->type == dme1737) &&
+	if ((data->has_features & HAS_VID) &&
 	    (err = sysfs_create_group(&dev->kobj,
 				      &dme1737_vid_group))) {
 		goto exit_remove;
 	}
+	if ((data->has_features & HAS_ZONE3) &&
+	    (err = sysfs_create_group(&dev->kobj,
+				      &dme1737_zone3_group))) {
+		goto exit_remove;
+	}
+	if ((data->has_features & HAS_ZONE_HYST) &&
+	    (err = sysfs_create_group(&dev->kobj,
+				      &dme1737_zone_hyst_group))) {
+		goto exit_remove;
+	}
 
 	/* Create fan sysfs attributes */
 	for (ix = 0; ix < ARRAY_SIZE(dme1737_fan_group); ix++) {
-		if (data->has_fan & (1 << ix)) {
+		if (data->has_features & HAS_FAN(ix)) {
 			if ((err = sysfs_create_group(&dev->kobj,
 						&dme1737_fan_group[ix]))) {
 				goto exit_remove;
@@ -1960,14 +2041,14 @@ static int dme1737_create_files(struct device *dev)
 
 	/* Create PWM sysfs attributes */
 	for (ix = 0; ix < ARRAY_SIZE(dme1737_pwm_group); ix++) {
-		if (data->has_pwm & (1 << ix)) {
+		if (data->has_features & HAS_PWM(ix)) {
 			if ((err = sysfs_create_group(&dev->kobj,
 						&dme1737_pwm_group[ix]))) {
 				goto exit_remove;
 			}
-			if (data->type != sch5027 && ix < 3 &&
+			if ((data->has_features & HAS_PWM_MIN) && ix < 3 &&
 			    (err = sysfs_create_file(&dev->kobj,
-						dme1737_pwm_misc_attr[ix]))) {
+					dme1737_auto_pwm_min_attr[ix]))) {
 				goto exit_remove;
 			}
 		}
@@ -1983,21 +2064,30 @@ static int dme1737_create_files(struct device *dev)
 		dme1737_chmod_group(dev, &dme1737_zone_chmod_group,
 				    S_IRUGO | S_IWUSR);
 
-		/* Change permissions of misc sysfs attributes */
-		if (data->type != sch5027) {
-			dme1737_chmod_group(dev, &dme1737_misc_group,
+		/* Change permissions of chip-dependent sysfs attributes */
+		if (data->has_features & HAS_TEMP_OFFSET) {
+			dme1737_chmod_group(dev, &dme1737_temp_offset_group,
+					    S_IRUGO | S_IWUSR);
+		}
+		if (data->has_features & HAS_ZONE3) {
+			dme1737_chmod_group(dev, &dme1737_zone3_chmod_group,
+					    S_IRUGO | S_IWUSR);
+		}
+		if (data->has_features & HAS_ZONE_HYST) {
+			dme1737_chmod_group(dev, &dme1737_zone_hyst_group,
 					    S_IRUGO | S_IWUSR);
 		}
 
 		/* Change permissions of PWM sysfs attributes */
 		for (ix = 0; ix < ARRAY_SIZE(dme1737_pwm_chmod_group); ix++) {
-			if (data->has_pwm & (1 << ix)) {
+			if (data->has_features & HAS_PWM(ix)) {
 				dme1737_chmod_group(dev,
 						&dme1737_pwm_chmod_group[ix],
 						S_IRUGO | S_IWUSR);
-				if (data->type != sch5027 && ix < 3) {
+				if ((data->has_features & HAS_PWM_MIN) &&
+				    ix < 3) {
 					dme1737_chmod_file(dev,
-						dme1737_pwm_misc_attr[ix],
+						dme1737_auto_pwm_min_attr[ix],
 						S_IRUGO | S_IWUSR);
 				}
 			}
@@ -2005,7 +2095,7 @@ static int dme1737_create_files(struct device *dev)
 
 		/* Change permissions of pwm[1-3] if in manual mode */
 		for (ix = 0; ix < 3; ix++) {
-			if ((data->has_pwm & (1 << ix)) &&
+			if ((data->has_features & HAS_PWM(ix)) &&
 			    (PWM_EN_FROM_REG(data->pwm_config[ix]) == 1)) {
 				dme1737_chmod_file(dev,
 						dme1737_pwm_chmod_attr[ix],
@@ -2052,20 +2142,20 @@ static int dme1737_init_device(struct device *dev)
 		return -EFAULT;
 	}
 
-	/* Determine which optional fan and pwm features are enabled/present */
+	/* Determine which optional fan and pwm features are enabled (only
+	 * valid for I2C devices) */
 	if (client) {   /* I2C chip */
 		data->config2 = dme1737_read(data, DME1737_REG_CONFIG2);
 		/* Check if optional fan3 input is enabled */
 		if (data->config2 & 0x04) {
-			data->has_fan |= (1 << 2);
+			data->has_features |= HAS_FAN(2);
 		}
 
 		/* Fan4 and pwm3 are only available if the client's I2C address
 		 * is the default 0x2e. Otherwise the I/Os associated with
 		 * these functions are used for addr enable/select. */
 		if (client->addr == 0x2e) {
-			data->has_fan |= (1 << 3);
-			data->has_pwm |= (1 << 2);
+			data->has_features |= HAS_FAN(3) | HAS_PWM(2);
 		}
 
 		/* Determine which of the optional fan[5-6] and pwm[5-6]
@@ -2077,26 +2167,40 @@ static int dme1737_init_device(struct device *dev)
 			dev_warn(dev, "Failed to query Super-IO for optional "
 				 "features.\n");
 		}
-	} else {   /* ISA chip */
-		/* Fan3 and pwm3 are always available. Fan[4-5] and pwm[5-6]
-		 * don't exist in the ISA chip. */
-		data->has_fan |= (1 << 2);
-		data->has_pwm |= (1 << 2);
 	}
 
-	/* Fan1, fan2, pwm1, and pwm2 are always present */
-	data->has_fan |= 0x03;
-	data->has_pwm |= 0x03;
+	/* Fan[1-2] and pwm[1-2] are present in all chips */
+	data->has_features |= HAS_FAN(0) | HAS_FAN(1) | HAS_PWM(0) | HAS_PWM(1);
+
+	/* Chip-dependent features */
+	switch (data->type) {
+	case dme1737:
+		data->has_features |= HAS_TEMP_OFFSET | HAS_VID | HAS_ZONE3 |
+			HAS_ZONE_HYST | HAS_PWM_MIN;
+		break;
+	case sch311x:
+		data->has_features |= HAS_TEMP_OFFSET | HAS_ZONE3 |
+			HAS_ZONE_HYST | HAS_PWM_MIN | HAS_FAN(2) | HAS_PWM(2);
+		break;
+	case sch5027:
+		data->has_features |= HAS_ZONE3;
+		break;
+	case sch5127:
+		data->has_features |= HAS_FAN(2) | HAS_PWM(2);
+		break;
+	default:
+		break;
+	}
 
 	dev_info(dev, "Optional features: pwm3=%s, pwm5=%s, pwm6=%s, "
 		 "fan3=%s, fan4=%s, fan5=%s, fan6=%s.\n",
-		 (data->has_pwm & (1 << 2)) ? "yes" : "no",
-		 (data->has_pwm & (1 << 4)) ? "yes" : "no",
-		 (data->has_pwm & (1 << 5)) ? "yes" : "no",
-		 (data->has_fan & (1 << 2)) ? "yes" : "no",
-		 (data->has_fan & (1 << 3)) ? "yes" : "no",
-		 (data->has_fan & (1 << 4)) ? "yes" : "no",
-		 (data->has_fan & (1 << 5)) ? "yes" : "no");
+		 (data->has_features & HAS_PWM(2)) ? "yes" : "no",
+		 (data->has_features & HAS_PWM(4)) ? "yes" : "no",
+		 (data->has_features & HAS_PWM(5)) ? "yes" : "no",
+		 (data->has_features & HAS_FAN(2)) ? "yes" : "no",
+		 (data->has_features & HAS_FAN(3)) ? "yes" : "no",
+		 (data->has_features & HAS_FAN(4)) ? "yes" : "no",
+		 (data->has_features & HAS_FAN(5)) ? "yes" : "no");
 
 	reg = dme1737_read(data, DME1737_REG_TACH_PWM);
 	/* Inform if fan-to-pwm mapping differs from the default */
@@ -2122,7 +2226,7 @@ static int dme1737_init_device(struct device *dev)
 		for (ix = 0; ix < 3; ix++) {
 			data->pwm_config[ix] = dme1737_read(data,
 						DME1737_REG_PWM_CONFIG(ix));
-			if ((data->has_pwm & (1 << ix)) &&
+			if ((data->has_features & HAS_PWM(ix)) &&
 			    (PWM_EN_FROM_REG(data->pwm_config[ix]) == -1)) {
 				dev_info(dev, "Switching pwm%d to "
 					 "manual mode.\n", ix + 1);
@@ -2142,7 +2246,7 @@ static int dme1737_init_device(struct device *dev)
 	data->pwm_acz[2] = 4;	/* pwm3 -> zone3 */
 
 	/* Set VRM */
-	if (data->type == dme1737) {
+	if (data->has_features & HAS_VID) {
 		data->vrm = vid_which_vrm();
 	}
 
@@ -2163,10 +2267,10 @@ static int dme1737_i2c_get_features(int sio_cip, struct dme1737_data *data)
 	dme1737_sio_enter(sio_cip);
 
 	/* Check device ID
-	 * The DME1737 can return either 0x78 or 0x77 as its device ID.
-	 * The SCH5027 returns 0x89 as its device ID. */
+	 * We currently know about two kinds of DME1737 and SCH5027. */
 	reg = force_id ? force_id : dme1737_sio_inb(sio_cip, 0x20);
-	if (!(reg == 0x77 || reg == 0x78 || reg == 0x89)) {
+	if (!(reg == DME1737_ID_1 || reg == DME1737_ID_2 ||
+	      reg == SCH5027_ID)) {
 		err = -ENODEV;
 		goto exit;
 	}
@@ -2185,16 +2289,16 @@ static int dme1737_i2c_get_features(int sio_cip, struct dme1737_data *data)
 	 * are enabled and available. Bits [3:2] of registers 0x43-0x46 are set
 	 * to '10' if the respective feature is enabled. */
 	if ((inb(addr + 0x43) & 0x0c) == 0x08) { /* fan6 */
-		data->has_fan |= (1 << 5);
+		data->has_features |= HAS_FAN(5);
 	}
 	if ((inb(addr + 0x44) & 0x0c) == 0x08) { /* pwm6 */
-		data->has_pwm |= (1 << 5);
+		data->has_features |= HAS_PWM(5);
 	}
 	if ((inb(addr + 0x45) & 0x0c) == 0x08) { /* fan5 */
-		data->has_fan |= (1 << 4);
+		data->has_features |= HAS_FAN(4);
 	}
 	if ((inb(addr + 0x46) & 0x0c) == 0x08) { /* pwm5 */
-		data->has_pwm |= (1 << 4);
+		data->has_features |= HAS_PWM(4);
 	}
 
 exit:
@@ -2222,7 +2326,6 @@ static int dme1737_i2c_detect(struct i2c_client *client,
 	if (company == DME1737_COMPANY_SMSC &&
 	    verstep == SCH5027_VERSTEP) {
 		name = "sch5027";
-
 	} else if (company == DME1737_COMPANY_SMSC &&
 		   (verstep & DME1737_VERSTEP_MASK) == DME1737_VERSTEP) {
 		name = "dme1737";
@@ -2329,10 +2432,10 @@ static int __init dme1737_isa_detect(int sio_cip, unsigned short *addr)
 	dme1737_sio_enter(sio_cip);
 
 	/* Check device ID
-	 * We currently know about SCH3112 (0x7c), SCH3114 (0x7d), and
-	 * SCH3116 (0x7f). */
+	 * We currently know about SCH3112, SCH3114, SCH3116, and SCH5127 */
 	reg = force_id ? force_id : dme1737_sio_inb(sio_cip, 0x20);
-	if (!(reg == 0x7c || reg == 0x7d || reg == 0x7f)) {
+	if (!(reg == SCH3112_ID || reg == SCH3114_ID || reg == SCH3116_ID ||
+	      reg == SCH5127_ID)) {
 		err = -ENODEV;
 		goto exit;
 	}
@@ -2424,23 +2527,42 @@ static int __devinit dme1737_isa_probe(struct platform_device *pdev)
 	platform_set_drvdata(pdev, data);
 
 	/* Skip chip detection if module is loaded with force_id parameter */
-	if (!force_id) {
+	switch (force_id) {
+	case SCH3112_ID:
+	case SCH3114_ID:
+	case SCH3116_ID:
+		data->type = sch311x;
+		break;
+	case SCH5127_ID:
+		data->type = sch5127;
+		break;
+	default:
 		company = dme1737_read(data, DME1737_REG_COMPANY);
 		device = dme1737_read(data, DME1737_REG_DEVICE);
 
-		if (!((company == DME1737_COMPANY_SMSC) &&
-		      (device == SCH311X_DEVICE))) {
+		if ((company == DME1737_COMPANY_SMSC) &&
+		    (device == SCH311X_DEVICE)) {
+			data->type = sch311x;
+		} else if ((company == DME1737_COMPANY_SMSC) &&
+			   (device == SCH5127_DEVICE)) {
+			data->type = sch5127;
+		} else {
 			err = -ENODEV;
 			goto exit_kfree;
 		}
 	}
-	data->type = sch311x;
 
-	/* Fill in the remaining client fields and initialize the mutex */
-	data->name = "sch311x";
+	if (data->type == sch5127) {
+		data->name = "sch5127";
+	} else {
+		data->name = "sch311x";
+	}
+
+	/* Initialize the mutex */
 	mutex_init(&data->update_lock);
 
-	dev_info(dev, "Found a SCH311x chip at 0x%04x\n", data->addr);
+	dev_info(dev, "Found a %s chip at 0x%04x\n",
+		 data->type == sch5127 ? "SCH5127" : "SCH311x", data->addr);
 
 	/* Initialize the chip */
 	if ((err = dme1737_init_device(dev))) {
diff --git a/drivers/hwmon/emc1403.c b/drivers/hwmon/emc1403.c
new file mode 100644
index 000000000000..0e4b5642638d
--- /dev/null
+++ b/drivers/hwmon/emc1403.c
@@ -0,0 +1,344 @@
+/*
+ * emc1403.c - SMSC Thermal Driver
+ *
+ * Copyright (C) 2008 Intel Corp
+ *
+ *  ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write to the Free Software Foundation, Inc.,
+ * 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA.
+ * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+ *
+ * TODO
+ *	-	cache alarm and critical limit registers
+ *	-	add emc1404 support
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/i2c.h>
+#include <linux/hwmon.h>
+#include <linux/hwmon-sysfs.h>
+#include <linux/err.h>
+#include <linux/sysfs.h>
+#include <linux/mutex.h>
+
+#define THERMAL_PID_REG		0xfd
+#define THERMAL_SMSC_ID_REG	0xfe
+#define THERMAL_REVISION_REG	0xff
+
+struct thermal_data {
+	struct device *hwmon_dev;
+	struct mutex mutex;
+	/* Cache the hyst value so we don't keep re-reading it. In theory
+	   we could cache it forever as nobody else should be writing it. */
+	u8 cached_hyst;
+	unsigned long hyst_valid;
+};
+
+static ssize_t show_temp(struct device *dev,
+			struct device_attribute *attr, char *buf)
+{
+	struct i2c_client *client = to_i2c_client(dev);
+	struct sensor_device_attribute *sda = to_sensor_dev_attr(attr);
+	int retval = i2c_smbus_read_byte_data(client, sda->index);
+
+	if (retval < 0)
+		return retval;
+	return sprintf(buf, "%d000\n", retval);
+}
+
+static ssize_t show_bit(struct device *dev,
+			struct device_attribute *attr, char *buf)
+{
+	struct i2c_client *client = to_i2c_client(dev);
+	struct sensor_device_attribute_2 *sda = to_sensor_dev_attr_2(attr);
+	int retval = i2c_smbus_read_byte_data(client, sda->nr);
+
+	if (retval < 0)
+		return retval;
+	retval &= sda->index;
+	return sprintf(buf, "%d\n", retval ? 1 : 0);
+}
+
+static ssize_t store_temp(struct device *dev,
+		struct device_attribute *attr, const char *buf, size_t count)
+{
+	struct sensor_device_attribute *sda = to_sensor_dev_attr(attr);
+	struct i2c_client *client = to_i2c_client(dev);
+	unsigned long val;
+	int retval;
+
+	if (strict_strtoul(buf, 10, &val))
+		return -EINVAL;
+	retval = i2c_smbus_write_byte_data(client, sda->index,
+					DIV_ROUND_CLOSEST(val, 1000));
+	if (retval < 0)
+		return retval;
+	return count;
+}
+
+static ssize_t show_hyst(struct device *dev,
+			struct device_attribute *attr, char *buf)
+{
+	struct i2c_client *client = to_i2c_client(dev);
+	struct thermal_data *data = i2c_get_clientdata(client);
+	struct sensor_device_attribute *sda = to_sensor_dev_attr(attr);
+	int retval;
+	int hyst;
+
+	retval = i2c_smbus_read_byte_data(client, sda->index);
+	if (retval < 0)
+		return retval;
+
+	if (time_after(jiffies, data->hyst_valid)) {
+		hyst = i2c_smbus_read_byte_data(client, 0x21);
+		if (hyst < 0)
+			return retval;
+		data->cached_hyst = hyst;
+		data->hyst_valid = jiffies + HZ;
+	}
+	return sprintf(buf, "%d000\n", retval - data->cached_hyst);
+}
+
+static ssize_t store_hyst(struct device *dev,
+		struct device_attribute *attr, const char *buf, size_t count)
+{
+	struct i2c_client *client = to_i2c_client(dev);
+	struct thermal_data *data = i2c_get_clientdata(client);
+	struct sensor_device_attribute *sda = to_sensor_dev_attr(attr);
+	int retval;
+	int hyst;
+	unsigned long val;
+
+	if (strict_strtoul(buf, 10, &val))
+		return -EINVAL;
+
+	mutex_lock(&data->mutex);
+	retval = i2c_smbus_read_byte_data(client, sda->index);
+	if (retval < 0)
+		goto fail;
+
+	hyst = val - retval * 1000;
+	hyst = DIV_ROUND_CLOSEST(hyst, 1000);
+	if (hyst < 0 || hyst > 255) {
+		retval = -ERANGE;
+		goto fail;
+	}
+
+	retval = i2c_smbus_write_byte_data(client, 0x21, hyst);
+	if (retval == 0) {
+		retval = count;
+		data->cached_hyst = hyst;
+		data->hyst_valid = jiffies + HZ;
+	}
+fail:
+	mutex_unlock(&data->mutex);
+	return retval;
+}
+
+/*
+ *	Sensors. We pass the actual i2c register to the methods.
+ */
+
+static SENSOR_DEVICE_ATTR(temp1_min, S_IRUGO | S_IWUSR,
+	show_temp, store_temp, 0x06);
+static SENSOR_DEVICE_ATTR(temp1_max, S_IRUGO | S_IWUSR,
+	show_temp, store_temp, 0x05);
+static SENSOR_DEVICE_ATTR(temp1_crit, S_IRUGO | S_IWUSR,
+	show_temp, store_temp, 0x20);
+static SENSOR_DEVICE_ATTR(temp1_input, S_IRUGO, show_temp, NULL, 0x00);
+static SENSOR_DEVICE_ATTR_2(temp1_min_alarm, S_IRUGO,
+	show_bit, NULL, 0x36, 0x01);
+static SENSOR_DEVICE_ATTR_2(temp1_max_alarm, S_IRUGO,
+	show_bit, NULL, 0x35, 0x01);
+static SENSOR_DEVICE_ATTR_2(temp1_crit_alarm, S_IRUGO,
+	show_bit, NULL, 0x37, 0x01);
+static SENSOR_DEVICE_ATTR(temp1_crit_hyst, S_IRUGO | S_IWUSR,
+	show_hyst, store_hyst, 0x20);
+
+static SENSOR_DEVICE_ATTR(temp2_min, S_IRUGO | S_IWUSR,
+	show_temp, store_temp, 0x08);
+static SENSOR_DEVICE_ATTR(temp2_max, S_IRUGO | S_IWUSR,
+	show_temp, store_temp, 0x07);
+static SENSOR_DEVICE_ATTR(temp2_crit, S_IRUGO | S_IWUSR,
+	show_temp, store_temp, 0x19);
+static SENSOR_DEVICE_ATTR(temp2_input, S_IRUGO, show_temp, NULL, 0x01);
+static SENSOR_DEVICE_ATTR_2(temp2_min_alarm, S_IRUGO,
+	show_bit, NULL, 0x36, 0x02);
+static SENSOR_DEVICE_ATTR_2(temp2_max_alarm, S_IRUGO,
+	show_bit, NULL, 0x35, 0x02);
+static SENSOR_DEVICE_ATTR_2(temp2_crit_alarm, S_IRUGO,
+	show_bit, NULL, 0x37, 0x02);
+static SENSOR_DEVICE_ATTR(temp2_crit_hyst, S_IRUGO | S_IWUSR,
+	show_hyst, store_hyst, 0x19);
+
+static SENSOR_DEVICE_ATTR(temp3_min, S_IRUGO | S_IWUSR,
+	show_temp, store_temp, 0x16);
+static SENSOR_DEVICE_ATTR(temp3_max, S_IRUGO | S_IWUSR,
+	show_temp, store_temp, 0x15);
+static SENSOR_DEVICE_ATTR(temp3_crit, S_IRUGO | S_IWUSR,
+	show_temp, store_temp, 0x1A);
+static SENSOR_DEVICE_ATTR(temp3_input, S_IRUGO, show_temp, NULL, 0x23);
+static SENSOR_DEVICE_ATTR_2(temp3_min_alarm, S_IRUGO,
+	show_bit, NULL, 0x36, 0x04);
+static SENSOR_DEVICE_ATTR_2(temp3_max_alarm, S_IRUGO,
+	show_bit, NULL, 0x35, 0x04);
+static SENSOR_DEVICE_ATTR_2(temp3_crit_alarm, S_IRUGO,
+	show_bit, NULL, 0x37, 0x04);
+static SENSOR_DEVICE_ATTR(temp3_crit_hyst, S_IRUGO | S_IWUSR,
+	show_hyst, store_hyst, 0x1A);
+
+static struct attribute *mid_att_thermal[] = {
+	&sensor_dev_attr_temp1_min.dev_attr.attr,
+	&sensor_dev_attr_temp1_max.dev_attr.attr,
+	&sensor_dev_attr_temp1_crit.dev_attr.attr,
+	&sensor_dev_attr_temp1_input.dev_attr.attr,
+	&sensor_dev_attr_temp1_min_alarm.dev_attr.attr,
+	&sensor_dev_attr_temp1_max_alarm.dev_attr.attr,
+	&sensor_dev_attr_temp1_crit_alarm.dev_attr.attr,
+	&sensor_dev_attr_temp1_crit_hyst.dev_attr.attr,
+	&sensor_dev_attr_temp2_min.dev_attr.attr,
+	&sensor_dev_attr_temp2_max.dev_attr.attr,
+	&sensor_dev_attr_temp2_crit.dev_attr.attr,
+	&sensor_dev_attr_temp2_input.dev_attr.attr,
+	&sensor_dev_attr_temp2_min_alarm.dev_attr.attr,
+	&sensor_dev_attr_temp2_max_alarm.dev_attr.attr,
+	&sensor_dev_attr_temp2_crit_alarm.dev_attr.attr,
+	&sensor_dev_attr_temp2_crit_hyst.dev_attr.attr,
+	&sensor_dev_attr_temp3_min.dev_attr.attr,
+	&sensor_dev_attr_temp3_max.dev_attr.attr,
+	&sensor_dev_attr_temp3_crit.dev_attr.attr,
+	&sensor_dev_attr_temp3_input.dev_attr.attr,
+	&sensor_dev_attr_temp3_min_alarm.dev_attr.attr,
+	&sensor_dev_attr_temp3_max_alarm.dev_attr.attr,
+	&sensor_dev_attr_temp3_crit_alarm.dev_attr.attr,
+	&sensor_dev_attr_temp3_crit_hyst.dev_attr.attr,
+	NULL
+};
+
+static const struct attribute_group m_thermal_gr = {
+	.attrs = mid_att_thermal
+};
+
+static int emc1403_detect(struct i2c_client *client,
+			struct i2c_board_info *info)
+{
+	int id;
+	/* Check if thermal chip is SMSC and EMC1403 */
+
+	id = i2c_smbus_read_byte_data(client, THERMAL_SMSC_ID_REG);
+	if (id != 0x5d)
+		return -ENODEV;
+
+	/* Note: 0x25 is the 1404 which is very similar and this
+	   driver could be extended */
+	id = i2c_smbus_read_byte_data(client, THERMAL_PID_REG);
+	if (id != 0x21)
+		return -ENODEV;
+
+	id = i2c_smbus_read_byte_data(client, THERMAL_REVISION_REG);
+	if (id != 0x01)
+		return -ENODEV;
+
+	strlcpy(info->type, "emc1403", I2C_NAME_SIZE);
+	return 0;
+}
+
+static int emc1403_probe(struct i2c_client *client,
+			const struct i2c_device_id *id)
+{
+	int res;
+	struct thermal_data *data;
+
+	data = kzalloc(sizeof(struct thermal_data), GFP_KERNEL);
+	if (data == NULL) {
+		dev_warn(&client->dev, "out of memory");
+		return -ENOMEM;
+	}
+
+	i2c_set_clientdata(client, data);
+	mutex_init(&data->mutex);
+	data->hyst_valid = jiffies - 1;		/* Expired */
+
+	res = sysfs_create_group(&client->dev.kobj, &m_thermal_gr);
+	if (res) {
+		dev_warn(&client->dev, "create group failed\n");
+		hwmon_device_unregister(data->hwmon_dev);
+		goto thermal_error1;
+	}
+	data->hwmon_dev = hwmon_device_register(&client->dev);
+	if (IS_ERR(data->hwmon_dev)) {
+		res = PTR_ERR(data->hwmon_dev);
+		dev_warn(&client->dev, "register hwmon dev failed\n");
+		goto thermal_error2;
+	}
+	dev_info(&client->dev, "EMC1403 Thermal chip found\n");
+	return res;
+
+thermal_error2:
+	sysfs_remove_group(&client->dev.kobj, &m_thermal_gr);
+thermal_error1:
+	kfree(data);
+	return res;
+}
+
+static int emc1403_remove(struct i2c_client *client)
+{
+	struct thermal_data *data = i2c_get_clientdata(client);
+
+	hwmon_device_unregister(data->hwmon_dev);
+	sysfs_remove_group(&client->dev.kobj, &m_thermal_gr);
+	kfree(data);
+	return 0;
+}
+
+static const unsigned short emc1403_address_list[] = {
+	0x18, 0x2a, 0x4c, 0x4d, I2C_CLIENT_END
+};
+
+static const struct i2c_device_id emc1403_idtable[] = {
+	{ "emc1403", 0 },
+	{ }
+};
+MODULE_DEVICE_TABLE(i2c, emc1403_idtable);
+
+static struct i2c_driver sensor_emc1403 = {
+	.class = I2C_CLASS_HWMON,
+	.driver = {
+		.name = "emc1403",
+	},
+	.detect = emc1403_detect,
+	.probe = emc1403_probe,
+	.remove = emc1403_remove,
+	.id_table = emc1403_idtable,
+	.address_list = emc1403_address_list,
+};
+
+static int __init sensor_emc1403_init(void)
+{
+	return i2c_add_driver(&sensor_emc1403);
+}
+
+static void  __exit sensor_emc1403_exit(void)
+{
+	i2c_del_driver(&sensor_emc1403);
+}
+
+module_init(sensor_emc1403_init);
+module_exit(sensor_emc1403_exit);
+
+MODULE_AUTHOR("Kalhan Trisal <kalhan.trisal@intel.com");
+MODULE_DESCRIPTION("emc1403 Thermal Driver");
+MODULE_LICENSE("GPL v2");
diff --git a/drivers/hwmon/f71882fg.c b/drivers/hwmon/f71882fg.c
index a95fa4256caa..537841ef44b9 100644
--- a/drivers/hwmon/f71882fg.c
+++ b/drivers/hwmon/f71882fg.c
@@ -856,21 +856,19 @@ static inline int superio_inb(int base, int reg)
 static int superio_inw(int base, int reg)
 {
 	int val;
-	outb(reg++, base);
-	val = inb(base + 1) << 8;
-	outb(reg, base);
-	val |= inb(base + 1);
+	val  = superio_inb(base, reg) << 8;
+	val |= superio_inb(base, reg + 1);
 	return val;
 }
 
 static inline void superio_enter(int base)
 {
 	/* according to the datasheet the key must be send twice! */
-	outb( SIO_UNLOCK_KEY, base);
-	outb( SIO_UNLOCK_KEY, base);
+	outb(SIO_UNLOCK_KEY, base);
+	outb(SIO_UNLOCK_KEY, base);
 }
 
-static inline void superio_select( int base, int ld)
+static inline void superio_select(int base, int ld)
 {
 	outb(SIO_REG_LDSEL, base);
 	outb(ld, base + 1);
@@ -905,10 +903,8 @@ static u16 f71882fg_read16(struct f71882fg_data *data, u8 reg)
 {
 	u16 val;
 
-	outb(reg++, data->addr + ADDR_REG_OFFSET);
-	val = inb(data->addr + DATA_REG_OFFSET) << 8;
-	outb(reg, data->addr + ADDR_REG_OFFSET);
-	val |= inb(data->addr + DATA_REG_OFFSET);
+	val  = f71882fg_read8(data, reg) << 8;
+	val |= f71882fg_read8(data, reg + 1);
 
 	return val;
 }
@@ -921,10 +917,8 @@ static void f71882fg_write8(struct f71882fg_data *data, u8 reg, u8 val)
 
 static void f71882fg_write16(struct f71882fg_data *data, u8 reg, u16 val)
 {
-	outb(reg++, data->addr + ADDR_REG_OFFSET);
-	outb(val >> 8, data->addr + DATA_REG_OFFSET);
-	outb(reg, data->addr + ADDR_REG_OFFSET);
-	outb(val & 255, data->addr + DATA_REG_OFFSET);
+	f71882fg_write8(data, reg,     val >> 8);
+	f71882fg_write8(data, reg + 1, val & 0xff);
 }
 
 static u16 f71882fg_read_temp(struct f71882fg_data *data, int nr)
@@ -945,7 +939,7 @@ static struct f71882fg_data *f71882fg_update_device(struct device *dev)
 	mutex_lock(&data->update_lock);
 
 	/* Update once every 60 seconds */
-	if ( time_after(jiffies, data->last_limits + 60 * HZ ) ||
+	if (time_after(jiffies, data->last_limits + 60 * HZ) ||
 			!data->valid) {
 		if (data->type == f71882fg || data->type == f71889fg) {
 			data->in1_max =
@@ -1127,8 +1121,12 @@ static ssize_t store_fan_full_speed(struct device *dev,
 				    const char *buf, size_t count)
 {
 	struct f71882fg_data *data = dev_get_drvdata(dev);
-	int nr = to_sensor_dev_attr_2(devattr)->index;
-	long val = simple_strtol(buf, NULL, 10);
+	int err, nr = to_sensor_dev_attr_2(devattr)->index;
+	long val;
+
+	err = strict_strtol(buf, 10, &val);
+	if (err)
+		return err;
 
 	val = SENSORS_LIMIT(val, 23, 1500000);
 	val = fan_to_reg(val);
@@ -1157,8 +1155,12 @@ static ssize_t store_fan_beep(struct device *dev, struct device_attribute
 	*devattr, const char *buf, size_t count)
 {
 	struct f71882fg_data *data = dev_get_drvdata(dev);
-	int nr = to_sensor_dev_attr_2(devattr)->index;
-	unsigned long val = simple_strtoul(buf, NULL, 10);
+	int err, nr = to_sensor_dev_attr_2(devattr)->index;
+	unsigned long val;
+
+	err = strict_strtoul(buf, 10, &val);
+	if (err)
+		return err;
 
 	mutex_lock(&data->update_lock);
 	data->fan_beep = f71882fg_read8(data, F71882FG_REG_FAN_BEEP);
@@ -1206,7 +1208,14 @@ static ssize_t store_in_max(struct device *dev, struct device_attribute
 	*devattr, const char *buf, size_t count)
 {
 	struct f71882fg_data *data = dev_get_drvdata(dev);
-	long val = simple_strtol(buf, NULL, 10) / 8;
+	int err;
+	long val;
+
+	err = strict_strtol(buf, 10, &val);
+	if (err)
+		return err;
+
+	val /= 8;
 	val = SENSORS_LIMIT(val, 0, 255);
 
 	mutex_lock(&data->update_lock);
@@ -1233,8 +1242,12 @@ static ssize_t store_in_beep(struct device *dev, struct device_attribute
 	*devattr, const char *buf, size_t count)
 {
 	struct f71882fg_data *data = dev_get_drvdata(dev);
-	int nr = to_sensor_dev_attr_2(devattr)->index;
-	unsigned long val = simple_strtoul(buf, NULL, 10);
+	int err, nr = to_sensor_dev_attr_2(devattr)->index;
+	unsigned long val;
+
+	err = strict_strtoul(buf, 10, &val);
+	if (err)
+		return err;
 
 	mutex_lock(&data->update_lock);
 	data->in_beep = f71882fg_read8(data, F71882FG_REG_IN_BEEP);
@@ -1299,8 +1312,14 @@ static ssize_t store_temp_max(struct device *dev, struct device_attribute
 	*devattr, const char *buf, size_t count)
 {
 	struct f71882fg_data *data = dev_get_drvdata(dev);
-	int nr = to_sensor_dev_attr_2(devattr)->index;
-	long val = simple_strtol(buf, NULL, 10) / 1000;
+	int err, nr = to_sensor_dev_attr_2(devattr)->index;
+	long val;
+
+	err = strict_strtol(buf, 10, &val);
+	if (err)
+		return err;
+
+	val /= 1000;
 	val = SENSORS_LIMIT(val, 0, 255);
 
 	mutex_lock(&data->update_lock);
@@ -1333,10 +1352,16 @@ static ssize_t store_temp_max_hyst(struct device *dev, struct device_attribute
 	*devattr, const char *buf, size_t count)
 {
 	struct f71882fg_data *data = dev_get_drvdata(dev);
-	int nr = to_sensor_dev_attr_2(devattr)->index;
-	long val = simple_strtol(buf, NULL, 10) / 1000;
+	int err, nr = to_sensor_dev_attr_2(devattr)->index;
 	ssize_t ret = count;
 	u8 reg;
+	long val;
+
+	err = strict_strtol(buf, 10, &val);
+	if (err)
+		return err;
+
+	val /= 1000;
 
 	mutex_lock(&data->update_lock);
 
@@ -1372,8 +1397,14 @@ static ssize_t store_temp_crit(struct device *dev, struct device_attribute
 	*devattr, const char *buf, size_t count)
 {
 	struct f71882fg_data *data = dev_get_drvdata(dev);
-	int nr = to_sensor_dev_attr_2(devattr)->index;
-	long val = simple_strtol(buf, NULL, 10) / 1000;
+	int err, nr = to_sensor_dev_attr_2(devattr)->index;
+	long val;
+
+	err = strict_strtol(buf, 10, &val);
+	if (err)
+		return err;
+
+	val /= 1000;
 	val = SENSORS_LIMIT(val, 0, 255);
 
 	mutex_lock(&data->update_lock);
@@ -1427,8 +1458,12 @@ static ssize_t store_temp_beep(struct device *dev, struct device_attribute
 	*devattr, const char *buf, size_t count)
 {
 	struct f71882fg_data *data = dev_get_drvdata(dev);
-	int nr = to_sensor_dev_attr_2(devattr)->index;
-	unsigned long val = simple_strtoul(buf, NULL, 10);
+	int err, nr = to_sensor_dev_attr_2(devattr)->index;
+	unsigned long val;
+
+	err = strict_strtoul(buf, 10, &val);
+	if (err)
+		return err;
 
 	mutex_lock(&data->update_lock);
 	data->temp_beep = f71882fg_read8(data, F71882FG_REG_TEMP_BEEP);
@@ -1490,8 +1525,13 @@ static ssize_t store_pwm(struct device *dev,
 			 size_t count)
 {
 	struct f71882fg_data *data = dev_get_drvdata(dev);
-	int nr = to_sensor_dev_attr_2(devattr)->index;
-	long val = simple_strtol(buf, NULL, 10);
+	int err, nr = to_sensor_dev_attr_2(devattr)->index;
+	long val;
+
+	err = strict_strtol(buf, 10, &val);
+	if (err)
+		return err;
+
 	val = SENSORS_LIMIT(val, 0, 255);
 
 	mutex_lock(&data->update_lock);
@@ -1551,8 +1591,12 @@ static ssize_t store_pwm_enable(struct device *dev, struct device_attribute
 				*devattr, const char *buf, size_t count)
 {
 	struct f71882fg_data *data = dev_get_drvdata(dev);
-	int nr = to_sensor_dev_attr_2(devattr)->index;
-	long val = simple_strtol(buf, NULL, 10);
+	int err, nr = to_sensor_dev_attr_2(devattr)->index;
+	long val;
+
+	err = strict_strtol(buf, 10, &val);
+	if (err)
+		return err;
 
 	/* Special case for F8000 pwm channel 3 which only does auto mode */
 	if (data->type == f8000 && nr == 2 && val != 2)
@@ -1626,9 +1670,14 @@ static ssize_t store_pwm_auto_point_pwm(struct device *dev,
 					const char *buf, size_t count)
 {
 	struct f71882fg_data *data = dev_get_drvdata(dev);
-	int pwm = to_sensor_dev_attr_2(devattr)->index;
+	int err, pwm = to_sensor_dev_attr_2(devattr)->index;
 	int point = to_sensor_dev_attr_2(devattr)->nr;
-	long val = simple_strtol(buf, NULL, 10);
+	long val;
+
+	err = strict_strtol(buf, 10, &val);
+	if (err)
+		return err;
+
 	val = SENSORS_LIMIT(val, 0, 255);
 
 	mutex_lock(&data->update_lock);
@@ -1674,10 +1723,16 @@ static ssize_t store_pwm_auto_point_temp_hyst(struct device *dev,
 					      const char *buf, size_t count)
 {
 	struct f71882fg_data *data = dev_get_drvdata(dev);
-	int nr = to_sensor_dev_attr_2(devattr)->index;
+	int err, nr = to_sensor_dev_attr_2(devattr)->index;
 	int point = to_sensor_dev_attr_2(devattr)->nr;
-	long val = simple_strtol(buf, NULL, 10) / 1000;
 	u8 reg;
+	long val;
+
+	err = strict_strtol(buf, 10, &val);
+	if (err)
+		return err;
+
+	val /= 1000;
 
 	mutex_lock(&data->update_lock);
 	data->pwm_auto_point_temp[nr][point] =
@@ -1716,8 +1771,12 @@ static ssize_t store_pwm_interpolate(struct device *dev,
 				     const char *buf, size_t count)
 {
 	struct f71882fg_data *data = dev_get_drvdata(dev);
-	int nr = to_sensor_dev_attr_2(devattr)->index;
-	unsigned long val = simple_strtoul(buf, NULL, 10);
+	int err, nr = to_sensor_dev_attr_2(devattr)->index;
+	unsigned long val;
+
+	err = strict_strtoul(buf, 10, &val);
+	if (err)
+		return err;
 
 	mutex_lock(&data->update_lock);
 	data->pwm_auto_point_mapping[nr] =
@@ -1752,8 +1811,12 @@ static ssize_t store_pwm_auto_point_channel(struct device *dev,
 					    const char *buf, size_t count)
 {
 	struct f71882fg_data *data = dev_get_drvdata(dev);
-	int nr = to_sensor_dev_attr_2(devattr)->index;
-	long val = simple_strtol(buf, NULL, 10);
+	int err, nr = to_sensor_dev_attr_2(devattr)->index;
+	long val;
+
+	err = strict_strtol(buf, 10, &val);
+	if (err)
+		return err;
 
 	switch (val) {
 	case 1:
@@ -1798,9 +1861,15 @@ static ssize_t store_pwm_auto_point_temp(struct device *dev,
 					 const char *buf, size_t count)
 {
 	struct f71882fg_data *data = dev_get_drvdata(dev);
-	int pwm = to_sensor_dev_attr_2(devattr)->index;
+	int err, pwm = to_sensor_dev_attr_2(devattr)->index;
 	int point = to_sensor_dev_attr_2(devattr)->nr;
-	long val = simple_strtol(buf, NULL, 10) / 1000;
+	long val;
+
+	err = strict_strtol(buf, 10, &val);
+	if (err)
+		return err;
+
+	val /= 1000;
 
 	if (data->type == f71889fg)
 		val = SENSORS_LIMIT(val, -128, 127);
@@ -2109,6 +2178,13 @@ static int __init f71882fg_find(int sioaddr, unsigned short *address,
 	int err = -ENODEV;
 	u16 devid;
 
+	/* Don't step on other drivers' I/O space by accident */
+	if (!request_region(sioaddr, 2, DRVNAME)) {
+		printk(KERN_ERR DRVNAME ": I/O address 0x%04x already in use\n",
+				(int)sioaddr);
+		return -EBUSY;
+	}
+
 	superio_enter(sioaddr);
 
 	devid = superio_inw(sioaddr, SIO_REG_MANID);
@@ -2151,8 +2227,7 @@ static int __init f71882fg_find(int sioaddr, unsigned short *address,
 	}
 
 	*address = superio_inw(sioaddr, SIO_REG_ADDR);
-	if (*address == 0)
-	{
+	if (*address == 0) {
 		printk(KERN_WARNING DRVNAME ": Base address not set\n");
 		goto exit;
 	}
@@ -2164,6 +2239,7 @@ static int __init f71882fg_find(int sioaddr, unsigned short *address,
 		(int)superio_inb(sioaddr, SIO_REG_DEVREV));
 exit:
 	superio_exit(sioaddr);
+	release_region(sioaddr, 2);
 	return err;
 }
 
diff --git a/drivers/hwmon/f75375s.c b/drivers/hwmon/f75375s.c
index bad2cf3ef4a4..0f58ecc5334d 100644
--- a/drivers/hwmon/f75375s.c
+++ b/drivers/hwmon/f75375s.c
@@ -662,7 +662,6 @@ exit_remove:
 	sysfs_remove_group(&client->dev.kobj, &f75375_group);
 exit_free:
 	kfree(data);
-	i2c_set_clientdata(client, NULL);
 	return err;
 }
 
@@ -672,7 +671,6 @@ static int f75375_remove(struct i2c_client *client)
 	hwmon_device_unregister(data->hwmon_dev);
 	sysfs_remove_group(&client->dev.kobj, &f75375_group);
 	kfree(data);
-	i2c_set_clientdata(client, NULL);
 	return 0;
 }
 
diff --git a/drivers/hwmon/fschmd.c b/drivers/hwmon/fschmd.c
index 0627f7a5b9b8..b7ca2a9676cf 100644
--- a/drivers/hwmon/fschmd.c
+++ b/drivers/hwmon/fschmd.c
@@ -38,6 +38,7 @@
 #include <linux/i2c.h>
 #include <linux/hwmon.h>
 #include <linux/hwmon-sysfs.h>
+#include <linux/smp_lock.h>
 #include <linux/err.h>
 #include <linux/mutex.h>
 #include <linux/sysfs.h>
@@ -847,8 +848,7 @@ static ssize_t watchdog_write(struct file *filp, const char __user *buf,
 	return count;
 }
 
-static int watchdog_ioctl(struct inode *inode, struct file *filp,
-	unsigned int cmd, unsigned long arg)
+static long watchdog_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 {
 	static struct watchdog_info ident = {
 		.options = WDIOF_KEEPALIVEPING | WDIOF_SETTIMEOUT |
@@ -858,6 +858,7 @@ static int watchdog_ioctl(struct inode *inode, struct file *filp,
 	int i, ret = 0;
 	struct fschmd_data *data = filp->private_data;
 
+	lock_kernel();
 	switch (cmd) {
 	case WDIOC_GETSUPPORT:
 		ident.firmware_version = data->revision;
@@ -914,7 +915,7 @@ static int watchdog_ioctl(struct inode *inode, struct file *filp,
 	default:
 		ret = -ENOTTY;
 	}
-
+	unlock_kernel();
 	return ret;
 }
 
@@ -924,7 +925,7 @@ static const struct file_operations watchdog_fops = {
 	.open = watchdog_open,
 	.release = watchdog_release,
 	.write = watchdog_write,
-	.ioctl = watchdog_ioctl,
+	.unlocked_ioctl = watchdog_ioctl,
 };
 
 
diff --git a/drivers/hwmon/g760a.c b/drivers/hwmon/g760a.c
index 09ea12e0a551..1f63d1a3af5e 100644
--- a/drivers/hwmon/g760a.c
+++ b/drivers/hwmon/g760a.c
@@ -236,7 +236,6 @@ error_hwmon_device_register:
 	sysfs_remove_group(&client->dev.kobj, &g760a_group);
 error_sysfs_create_group:
 	kfree(data);
-	i2c_set_clientdata(client, NULL);
 
 	return err;
 }
@@ -247,7 +246,6 @@ static int g760a_remove(struct i2c_client *client)
 	hwmon_device_unregister(data->hwmon_dev);
 	sysfs_remove_group(&client->dev.kobj, &g760a_group);
 	kfree(data);
-	i2c_set_clientdata(client, NULL);
 
 	return 0;
 }
diff --git a/drivers/hwmon/lis3lv02d.c b/drivers/hwmon/lis3lv02d.c
index b2f2277cad3c..6138f036b159 100644
--- a/drivers/hwmon/lis3lv02d.c
+++ b/drivers/hwmon/lis3lv02d.c
@@ -41,6 +41,8 @@
 
 /* joystick device poll interval in milliseconds */
 #define MDPS_POLL_INTERVAL 50
+#define MDPS_POLL_MIN	   0
+#define MDPS_POLL_MAX	   2000
 /*
  * The sensor can also generate interrupts (DRDY) but it's pretty pointless
  * because they are generated even if the data do not change. So it's better
@@ -121,11 +123,9 @@ static void lis3lv02d_get_xyz(struct lis3lv02d *lis3, int *x, int *y, int *z)
 	int position[3];
 	int i;
 
-	mutex_lock(&lis3->mutex);
 	position[0] = lis3->read_data(lis3, OUTX);
 	position[1] = lis3->read_data(lis3, OUTY);
 	position[2] = lis3->read_data(lis3, OUTZ);
-	mutex_unlock(&lis3->mutex);
 
 	for (i = 0; i < 3; i++)
 		position[i] = (position[i] * lis3->scale) / LIS3_ACCURACY;
@@ -249,8 +249,24 @@ void lis3lv02d_poweron(struct lis3lv02d *lis3)
 EXPORT_SYMBOL_GPL(lis3lv02d_poweron);
 
 
+static void lis3lv02d_joystick_poll(struct input_polled_dev *pidev)
+{
+	int x, y, z;
+
+	mutex_lock(&lis3_dev.mutex);
+	lis3lv02d_get_xyz(&lis3_dev, &x, &y, &z);
+	input_report_abs(pidev->input, ABS_X, x);
+	input_report_abs(pidev->input, ABS_Y, y);
+	input_report_abs(pidev->input, ABS_Z, z);
+	input_sync(pidev->input);
+	mutex_unlock(&lis3_dev.mutex);
+}
+
 static irqreturn_t lis302dl_interrupt(int irq, void *dummy)
 {
+	if (!test_bit(0, &lis3_dev.misc_opened))
+		goto out;
+
 	/*
 	 * Be careful: on some HP laptops the bios force DD when on battery and
 	 * the lid is closed. This leads to interrupts as soon as a little move
@@ -260,44 +276,93 @@ static irqreturn_t lis302dl_interrupt(int irq, void *dummy)
 
 	wake_up_interruptible(&lis3_dev.misc_wait);
 	kill_fasync(&lis3_dev.async_queue, SIGIO, POLL_IN);
+out:
+	if (lis3_dev.whoami == WAI_8B && lis3_dev.idev &&
+	    lis3_dev.idev->input->users)
+		return IRQ_WAKE_THREAD;
 	return IRQ_HANDLED;
 }
 
-static int lis3lv02d_misc_open(struct inode *inode, struct file *file)
+static void lis302dl_interrupt_handle_click(struct lis3lv02d *lis3)
 {
-	int ret;
+	struct input_dev *dev = lis3->idev->input;
+	u8 click_src;
 
-	if (test_and_set_bit(0, &lis3_dev.misc_opened))
-		return -EBUSY; /* already open */
+	mutex_lock(&lis3->mutex);
+	lis3->read(lis3, CLICK_SRC, &click_src);
 
-	atomic_set(&lis3_dev.count, 0);
+	if (click_src & CLICK_SINGLE_X) {
+		input_report_key(dev, lis3->mapped_btns[0], 1);
+		input_report_key(dev, lis3->mapped_btns[0], 0);
+	}
 
-	/*
-	 * The sensor can generate interrupts for free-fall and direction
-	 * detection (distinguishable with FF_WU_SRC and DD_SRC) but to keep
-	 * the things simple and _fast_ we activate it only for free-fall, so
-	 * no need to read register (very slow with ACPI). For the same reason,
-	 * we forbid shared interrupts.
-	 *
-	 * IRQF_TRIGGER_RISING seems pointless on HP laptops because the
-	 * io-apic is not configurable (and generates a warning) but I keep it
-	 * in case of support for other hardware.
-	 */
-	ret = request_irq(lis3_dev.irq, lis302dl_interrupt, IRQF_TRIGGER_RISING,
-			  DRIVER_NAME, &lis3_dev);
+	if (click_src & CLICK_SINGLE_Y) {
+		input_report_key(dev, lis3->mapped_btns[1], 1);
+		input_report_key(dev, lis3->mapped_btns[1], 0);
+	}
 
-	if (ret) {
-		clear_bit(0, &lis3_dev.misc_opened);
-		printk(KERN_ERR DRIVER_NAME ": IRQ%d allocation failed\n", lis3_dev.irq);
-		return -EBUSY;
+	if (click_src & CLICK_SINGLE_Z) {
+		input_report_key(dev, lis3->mapped_btns[2], 1);
+		input_report_key(dev, lis3->mapped_btns[2], 0);
 	}
+	input_sync(dev);
+	mutex_unlock(&lis3->mutex);
+}
+
+static void lis302dl_interrupt_handle_ff_wu(struct lis3lv02d *lis3)
+{
+	u8 wu1_src;
+	u8 wu2_src;
+
+	lis3->read(lis3, FF_WU_SRC_1, &wu1_src);
+	lis3->read(lis3, FF_WU_SRC_2, &wu2_src);
+
+	wu1_src = wu1_src & FF_WU_SRC_IA ? wu1_src : 0;
+	wu2_src = wu2_src & FF_WU_SRC_IA ? wu2_src : 0;
+
+	/* joystick poll is internally protected by the lis3->mutex. */
+	if (wu1_src || wu2_src)
+		lis3lv02d_joystick_poll(lis3_dev.idev);
+}
+
+static irqreturn_t lis302dl_interrupt_thread1_8b(int irq, void *data)
+{
+
+	struct lis3lv02d *lis3 = data;
+
+	if ((lis3->pdata->irq_cfg & LIS3_IRQ1_MASK) == LIS3_IRQ1_CLICK)
+		lis302dl_interrupt_handle_click(lis3);
+	else
+		lis302dl_interrupt_handle_ff_wu(lis3);
+
+	return IRQ_HANDLED;
+}
+
+static irqreturn_t lis302dl_interrupt_thread2_8b(int irq, void *data)
+{
+
+	struct lis3lv02d *lis3 = data;
+
+	if ((lis3->pdata->irq_cfg & LIS3_IRQ2_MASK) == LIS3_IRQ2_CLICK)
+		lis302dl_interrupt_handle_click(lis3);
+	else
+		lis302dl_interrupt_handle_ff_wu(lis3);
+
+	return IRQ_HANDLED;
+}
+
+static int lis3lv02d_misc_open(struct inode *inode, struct file *file)
+{
+	if (test_and_set_bit(0, &lis3_dev.misc_opened))
+		return -EBUSY; /* already open */
+
+	atomic_set(&lis3_dev.count, 0);
 	return 0;
 }
 
 static int lis3lv02d_misc_release(struct inode *inode, struct file *file)
 {
 	fasync_helper(-1, file, 0, &lis3_dev.async_queue);
-	free_irq(lis3_dev.irq, &lis3_dev);
 	clear_bit(0, &lis3_dev.misc_opened); /* release the device */
 	return 0;
 }
@@ -380,22 +445,12 @@ static struct miscdevice lis3lv02d_misc_device = {
 	.fops    = &lis3lv02d_misc_fops,
 };
 
-static void lis3lv02d_joystick_poll(struct input_polled_dev *pidev)
-{
-	int x, y, z;
-
-	lis3lv02d_get_xyz(&lis3_dev, &x, &y, &z);
-	input_report_abs(pidev->input, ABS_X, x);
-	input_report_abs(pidev->input, ABS_Y, y);
-	input_report_abs(pidev->input, ABS_Z, z);
-	input_sync(pidev->input);
-}
-
 int lis3lv02d_joystick_enable(void)
 {
 	struct input_dev *input_dev;
 	int err;
 	int max_val, fuzz, flat;
+	int btns[] = {BTN_X, BTN_Y, BTN_Z};
 
 	if (lis3_dev.idev)
 		return -EINVAL;
@@ -406,6 +461,8 @@ int lis3lv02d_joystick_enable(void)
 
 	lis3_dev.idev->poll = lis3lv02d_joystick_poll;
 	lis3_dev.idev->poll_interval = MDPS_POLL_INTERVAL;
+	lis3_dev.idev->poll_interval_min = MDPS_POLL_MIN;
+	lis3_dev.idev->poll_interval_max = MDPS_POLL_MAX;
 	input_dev = lis3_dev.idev->input;
 
 	input_dev->name       = "ST LIS3LV02DL Accelerometer";
@@ -422,6 +479,10 @@ int lis3lv02d_joystick_enable(void)
 	input_set_abs_params(input_dev, ABS_Y, -max_val, max_val, fuzz, flat);
 	input_set_abs_params(input_dev, ABS_Z, -max_val, max_val, fuzz, flat);
 
+	lis3_dev.mapped_btns[0] = lis3lv02d_get_axis(abs(lis3_dev.ac.x), btns);
+	lis3_dev.mapped_btns[1] = lis3lv02d_get_axis(abs(lis3_dev.ac.y), btns);
+	lis3_dev.mapped_btns[2] = lis3lv02d_get_axis(abs(lis3_dev.ac.z), btns);
+
 	err = input_register_polled_device(lis3_dev.idev);
 	if (err) {
 		input_free_polled_device(lis3_dev.idev);
@@ -434,6 +495,11 @@ EXPORT_SYMBOL_GPL(lis3lv02d_joystick_enable);
 
 void lis3lv02d_joystick_disable(void)
 {
+	if (lis3_dev.irq)
+		free_irq(lis3_dev.irq, &lis3_dev);
+	if (lis3_dev.pdata && lis3_dev.pdata->irq2)
+		free_irq(lis3_dev.pdata->irq2, &lis3_dev);
+
 	if (!lis3_dev.idev)
 		return;
 
@@ -462,7 +528,9 @@ static ssize_t lis3lv02d_position_show(struct device *dev,
 {
 	int x, y, z;
 
+	mutex_lock(&lis3_dev.mutex);
 	lis3lv02d_get_xyz(&lis3_dev, &x, &y, &z);
+	mutex_unlock(&lis3_dev.mutex);
 	return sprintf(buf, "(%d,%d,%d)\n", x, y, z);
 }
 
@@ -521,12 +589,70 @@ int lis3lv02d_remove_fs(struct lis3lv02d *lis3)
 }
 EXPORT_SYMBOL_GPL(lis3lv02d_remove_fs);
 
+static void lis3lv02d_8b_configure(struct lis3lv02d *dev,
+				struct lis3lv02d_platform_data *p)
+{
+	int err;
+	int ctrl2 = p->hipass_ctrl;
+
+	if (p->click_flags) {
+		dev->write(dev, CLICK_CFG, p->click_flags);
+		dev->write(dev, CLICK_TIMELIMIT, p->click_time_limit);
+		dev->write(dev, CLICK_LATENCY, p->click_latency);
+		dev->write(dev, CLICK_WINDOW, p->click_window);
+		dev->write(dev, CLICK_THSZ, p->click_thresh_z & 0xf);
+		dev->write(dev, CLICK_THSY_X,
+			(p->click_thresh_x & 0xf) |
+			(p->click_thresh_y << 4));
+
+		if (dev->idev) {
+			struct input_dev *input_dev = lis3_dev.idev->input;
+			input_set_capability(input_dev, EV_KEY, BTN_X);
+			input_set_capability(input_dev, EV_KEY, BTN_Y);
+			input_set_capability(input_dev, EV_KEY, BTN_Z);
+		}
+	}
+
+	if (p->wakeup_flags) {
+		dev->write(dev, FF_WU_CFG_1, p->wakeup_flags);
+		dev->write(dev, FF_WU_THS_1, p->wakeup_thresh & 0x7f);
+		/* default to 2.5ms for now */
+		dev->write(dev, FF_WU_DURATION_1, 1);
+		ctrl2 ^= HP_FF_WU1; /* Xor to keep compatible with old pdata*/
+	}
+
+	if (p->wakeup_flags2) {
+		dev->write(dev, FF_WU_CFG_2, p->wakeup_flags2);
+		dev->write(dev, FF_WU_THS_2, p->wakeup_thresh2 & 0x7f);
+		/* default to 2.5ms for now */
+		dev->write(dev, FF_WU_DURATION_2, 1);
+		ctrl2 ^= HP_FF_WU2; /* Xor to keep compatible with old pdata*/
+	}
+	/* Configure hipass filters */
+	dev->write(dev, CTRL_REG2, ctrl2);
+
+	if (p->irq2) {
+		err = request_threaded_irq(p->irq2,
+					NULL,
+					lis302dl_interrupt_thread2_8b,
+					IRQF_TRIGGER_RISING |
+					IRQF_ONESHOT,
+					DRIVER_NAME, &lis3_dev);
+		if (err < 0)
+			printk(KERN_ERR DRIVER_NAME
+				"No second IRQ. Limited functionality\n");
+	}
+}
+
 /*
  * Initialise the accelerometer and the various subsystems.
  * Should be rather independent of the bus system.
  */
 int lis3lv02d_init_device(struct lis3lv02d *dev)
 {
+	int err;
+	irq_handler_t thread_fn;
+
 	dev->whoami = lis3lv02d_read_8(dev, WHO_AM_I);
 
 	switch (dev->whoami) {
@@ -567,25 +693,8 @@ int lis3lv02d_init_device(struct lis3lv02d *dev)
 	if (dev->pdata) {
 		struct lis3lv02d_platform_data *p = dev->pdata;
 
-		if (p->click_flags && (dev->whoami == WAI_8B)) {
-			dev->write(dev, CLICK_CFG, p->click_flags);
-			dev->write(dev, CLICK_TIMELIMIT, p->click_time_limit);
-			dev->write(dev, CLICK_LATENCY, p->click_latency);
-			dev->write(dev, CLICK_WINDOW, p->click_window);
-			dev->write(dev, CLICK_THSZ, p->click_thresh_z & 0xf);
-			dev->write(dev, CLICK_THSY_X,
-					(p->click_thresh_x & 0xf) |
-					(p->click_thresh_y << 4));
-		}
-
-		if (p->wakeup_flags && (dev->whoami == WAI_8B)) {
-			dev->write(dev, FF_WU_CFG_1, p->wakeup_flags);
-			dev->write(dev, FF_WU_THS_1, p->wakeup_thresh & 0x7f);
-			/* default to 2.5ms for now */
-			dev->write(dev, FF_WU_DURATION_1, 1);
-			/* enable high pass filter for both free-fall units */
-			dev->write(dev, CTRL_REG2, HP_FF_WU1 | HP_FF_WU2);
-		}
+		if (dev->whoami == WAI_8B)
+			lis3lv02d_8b_configure(dev, p);
 
 		if (p->irq_cfg)
 			dev->write(dev, CTRL_REG3, p->irq_cfg);
@@ -598,6 +707,32 @@ int lis3lv02d_init_device(struct lis3lv02d *dev)
 		goto out;
 	}
 
+	/*
+	 * The sensor can generate interrupts for free-fall and direction
+	 * detection (distinguishable with FF_WU_SRC and DD_SRC) but to keep
+	 * the things simple and _fast_ we activate it only for free-fall, so
+	 * no need to read register (very slow with ACPI). For the same reason,
+	 * we forbid shared interrupts.
+	 *
+	 * IRQF_TRIGGER_RISING seems pointless on HP laptops because the
+	 * io-apic is not configurable (and generates a warning) but I keep it
+	 * in case of support for other hardware.
+	 */
+	if (dev->whoami == WAI_8B)
+		thread_fn = lis302dl_interrupt_thread1_8b;
+	else
+		thread_fn = NULL;
+
+	err = request_threaded_irq(dev->irq, lis302dl_interrupt,
+				thread_fn,
+				IRQF_TRIGGER_RISING | IRQF_ONESHOT,
+				DRIVER_NAME, &lis3_dev);
+
+	if (err < 0) {
+		printk(KERN_ERR DRIVER_NAME "Cannot get IRQ\n");
+		goto out;
+	}
+
 	if (misc_register(&lis3lv02d_misc_device))
 		printk(KERN_ERR DRIVER_NAME ": misc_register failed\n");
 out:
diff --git a/drivers/hwmon/lis3lv02d.h b/drivers/hwmon/lis3lv02d.h
index e6a01f44709b..854091380e33 100644
--- a/drivers/hwmon/lis3lv02d.h
+++ b/drivers/hwmon/lis3lv02d.h
@@ -196,6 +196,16 @@ enum lis3lv02d_dd_src {
 	DD_SRC_IA	= 0x40,
 };
 
+enum lis3lv02d_click_src_8b {
+	CLICK_SINGLE_X	= 0x01,
+	CLICK_DOUBLE_X	= 0x02,
+	CLICK_SINGLE_Y	= 0x04,
+	CLICK_DOUBLE_Y	= 0x08,
+	CLICK_SINGLE_Z	= 0x10,
+	CLICK_DOUBLE_Z	= 0x20,
+	CLICK_IA	= 0x40,
+};
+
 struct axis_conversion {
 	s8	x;
 	s8	y;
@@ -223,6 +233,7 @@ struct lis3lv02d {
 	struct platform_device	*pdev;     /* platform device */
 	atomic_t		count;     /* interrupt count after last read */
 	struct axis_conversion	ac;        /* hw -> logical axis */
+	int			mapped_btns[3];
 
 	u32			irq;       /* IRQ number */
 	struct fasync_struct	*async_queue; /* queue for the misc device */
diff --git a/drivers/hwmon/lm63.c b/drivers/hwmon/lm63.c
index bf81aff7051d..776aeb3019d2 100644
--- a/drivers/hwmon/lm63.c
+++ b/drivers/hwmon/lm63.c
@@ -53,7 +53,7 @@
  * Address is fully defined internally and cannot be changed.
  */
 
-static const unsigned short normal_i2c[] = { 0x4c, I2C_CLIENT_END };
+static const unsigned short normal_i2c[] = { 0x18, 0x4c, 0x4e, I2C_CLIENT_END };
 
 /*
  * The LM63 registers
@@ -131,12 +131,15 @@ static struct lm63_data *lm63_update_device(struct device *dev);
 static int lm63_detect(struct i2c_client *client, struct i2c_board_info *info);
 static void lm63_init_client(struct i2c_client *client);
 
+enum chips { lm63, lm64 };
+
 /*
  * Driver data (common to all clients)
  */
 
 static const struct i2c_device_id lm63_id[] = {
-	{ "lm63", 0 },
+	{ "lm63", lm63 },
+	{ "lm64", lm64 },
 	{ }
 };
 MODULE_DEVICE_TABLE(i2c, lm63_id);
@@ -422,6 +425,7 @@ static int lm63_detect(struct i2c_client *new_client,
 	struct i2c_adapter *adapter = new_client->adapter;
 	u8 man_id, chip_id, reg_config1, reg_config2;
 	u8 reg_alert_status, reg_alert_mask;
+	int address = new_client->addr;
 
 	if (!i2c_check_functionality(adapter, I2C_FUNC_SMBUS_BYTE_DATA))
 		return -ENODEV;
@@ -439,7 +443,6 @@ static int lm63_detect(struct i2c_client *new_client,
 			 LM63_REG_ALERT_MASK);
 
 	if (man_id != 0x01 /* National Semiconductor */
-	 || chip_id != 0x41 /* LM63 */
 	 || (reg_config1 & 0x18) != 0x00
 	 || (reg_config2 & 0xF8) != 0x00
 	 || (reg_alert_status & 0x20) != 0x00
@@ -450,7 +453,12 @@ static int lm63_detect(struct i2c_client *new_client,
 		return -ENODEV;
 	}
 
-	strlcpy(info->type, "lm63", I2C_NAME_SIZE);
+	if (chip_id == 0x41 && address == 0x4c)
+		strlcpy(info->type, "lm63", I2C_NAME_SIZE);
+	else if (chip_id == 0x51 && (address == 0x18 || address == 0x4e))
+		strlcpy(info->type, "lm64", I2C_NAME_SIZE);
+	else
+		return -ENODEV;
 
 	return 0;
 }
diff --git a/drivers/hwmon/lm73.c b/drivers/hwmon/lm73.c
index 4d1b76bc8148..29b9030d42c3 100644
--- a/drivers/hwmon/lm73.c
+++ b/drivers/hwmon/lm73.c
@@ -136,7 +136,6 @@ static int lm73_remove(struct i2c_client *client)
 
 	hwmon_device_unregister(hwmon_dev);
 	sysfs_remove_group(&client->dev.kobj, &lm73_group);
-	i2c_set_clientdata(client, NULL);
 	return 0;
 }
 
diff --git a/drivers/hwmon/lm75.c b/drivers/hwmon/lm75.c
index 8ae2cfe2d827..393f354f92a4 100644
--- a/drivers/hwmon/lm75.c
+++ b/drivers/hwmon/lm75.c
@@ -46,6 +46,7 @@ enum lm75_type {		/* keep sorted in alphabetical order */
 	tcn75,
 	tmp100,
 	tmp101,
+	tmp105,
 	tmp175,
 	tmp275,
 	tmp75,
@@ -191,7 +192,6 @@ lm75_probe(struct i2c_client *client, const struct i2c_device_id *id)
 exit_remove:
 	sysfs_remove_group(&client->dev.kobj, &lm75_group);
 exit_free:
-	i2c_set_clientdata(client, NULL);
 	kfree(data);
 	return status;
 }
@@ -203,7 +203,6 @@ static int lm75_remove(struct i2c_client *client)
 	hwmon_device_unregister(data->hwmon_dev);
 	sysfs_remove_group(&client->dev.kobj, &lm75_group);
 	lm75_write_value(client, LM75_REG_CONF, data->orig_conf);
-	i2c_set_clientdata(client, NULL);
 	kfree(data);
 	return 0;
 }
@@ -220,6 +219,7 @@ static const struct i2c_device_id lm75_ids[] = {
 	{ "tcn75", tcn75, },
 	{ "tmp100", tmp100, },
 	{ "tmp101", tmp101, },
+	{ "tmp105", tmp105, },
 	{ "tmp175", tmp175, },
 	{ "tmp275", tmp275, },
 	{ "tmp75", tmp75, },
diff --git a/drivers/hwmon/lm90.c b/drivers/hwmon/lm90.c
index 7cc2708871ab..760ef72eea56 100644
--- a/drivers/hwmon/lm90.c
+++ b/drivers/hwmon/lm90.c
@@ -982,7 +982,8 @@ static struct lm90_data *lm90_update_device(struct device *dev)
 
 	mutex_lock(&data->update_lock);
 
-	if (time_after(jiffies, data->last_updated + HZ * 2) || !data->valid) {
+	if (time_after(jiffies, data->last_updated + HZ / 2 + HZ / 10)
+	 || !data->valid) {
 		u8 h, l;
 
 		dev_dbg(&client->dev, "Updating lm90 data.\n");
diff --git a/drivers/hwmon/lm95241.c b/drivers/hwmon/lm95241.c
index 8fc8eb8cba47..94741d42112d 100644
--- a/drivers/hwmon/lm95241.c
+++ b/drivers/hwmon/lm95241.c
@@ -399,7 +399,6 @@ static int lm95241_remove(struct i2c_client *client)
 	hwmon_device_unregister(data->hwmon_dev);
 	sysfs_remove_group(&client->dev.kobj, &lm95241_group);
 
-	i2c_set_clientdata(client, NULL);
 	kfree(data);
 	return 0;
 }
diff --git a/drivers/hwmon/ltc4245.c b/drivers/hwmon/ltc4245.c
index 65c232a9d0c5..21d201befc2c 100644
--- a/drivers/hwmon/ltc4245.c
+++ b/drivers/hwmon/ltc4245.c
@@ -45,9 +45,7 @@ enum ltc4245_cmd {
 	LTC4245_VEEIN			= 0x19,
 	LTC4245_VEESENSE		= 0x1a,
 	LTC4245_VEEOUT			= 0x1b,
-	LTC4245_GPIOADC1		= 0x1c,
-	LTC4245_GPIOADC2		= 0x1d,
-	LTC4245_GPIOADC3		= 0x1e,
+	LTC4245_GPIOADC			= 0x1c,
 };
 
 struct ltc4245_data {
@@ -61,7 +59,7 @@ struct ltc4245_data {
 	u8 cregs[0x08];
 
 	/* Voltage registers */
-	u8 vregs[0x0f];
+	u8 vregs[0x0d];
 };
 
 static struct ltc4245_data *ltc4245_update_device(struct device *dev)
@@ -86,7 +84,7 @@ static struct ltc4245_data *ltc4245_update_device(struct device *dev)
 				data->cregs[i] = val;
 		}
 
-		/* Read voltage registers -- 0x10 to 0x1f */
+		/* Read voltage registers -- 0x10 to 0x1c */
 		for (i = 0; i < ARRAY_SIZE(data->vregs); i++) {
 			val = i2c_smbus_read_byte_data(client, i+0x10);
 			if (unlikely(val < 0))
@@ -128,9 +126,7 @@ static int ltc4245_get_voltage(struct device *dev, u8 reg)
 	case LTC4245_VEEOUT:
 		voltage = regval * -55;
 		break;
-	case LTC4245_GPIOADC1:
-	case LTC4245_GPIOADC2:
-	case LTC4245_GPIOADC3:
+	case LTC4245_GPIOADC:
 		voltage = regval * 10;
 		break;
 	default:
@@ -297,9 +293,7 @@ LTC4245_ALARM(in7_min_alarm,	(1 << 2),	LTC4245_FAULT2);
 LTC4245_ALARM(in8_min_alarm,	(1 << 3),	LTC4245_FAULT2);
 
 /* GPIO voltages */
-LTC4245_VOLTAGE(in9_input,			LTC4245_GPIOADC1);
-LTC4245_VOLTAGE(in10_input,			LTC4245_GPIOADC2);
-LTC4245_VOLTAGE(in11_input,			LTC4245_GPIOADC3);
+LTC4245_VOLTAGE(in9_input,			LTC4245_GPIOADC);
 
 /* Power Consumption (virtual) */
 LTC4245_POWER(power1_input,			LTC4245_12VSENSE);
@@ -342,8 +336,6 @@ static struct attribute *ltc4245_attributes[] = {
 	&sensor_dev_attr_in8_min_alarm.dev_attr.attr,
 
 	&sensor_dev_attr_in9_input.dev_attr.attr,
-	&sensor_dev_attr_in10_input.dev_attr.attr,
-	&sensor_dev_attr_in11_input.dev_attr.attr,
 
 	&sensor_dev_attr_power1_input.dev_attr.attr,
 	&sensor_dev_attr_power2_input.dev_attr.attr,
diff --git a/drivers/hwmon/tmp102.c b/drivers/hwmon/tmp102.c
new file mode 100644
index 000000000000..93187c3cb5e7
--- /dev/null
+++ b/drivers/hwmon/tmp102.c
@@ -0,0 +1,319 @@
+/* Texas Instruments TMP102 SMBus temperature sensor driver
+ *
+ * Copyright (C) 2010 Steven King <sfking@fdwdc.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/i2c.h>
+#include <linux/hwmon.h>
+#include <linux/hwmon-sysfs.h>
+#include <linux/err.h>
+#include <linux/mutex.h>
+#include <linux/device.h>
+
+#define	DRIVER_NAME "tmp102"
+
+#define	TMP102_TEMP_REG			0x00
+#define	TMP102_CONF_REG			0x01
+/* note: these bit definitions are byte swapped */
+#define		TMP102_CONF_SD		0x0100
+#define		TMP102_CONF_TM		0x0200
+#define		TMP102_CONF_POL		0x0400
+#define		TMP102_CONF_F0		0x0800
+#define		TMP102_CONF_F1		0x1000
+#define		TMP102_CONF_R0		0x2000
+#define		TMP102_CONF_R1		0x4000
+#define		TMP102_CONF_OS		0x8000
+#define		TMP102_CONF_EM		0x0010
+#define		TMP102_CONF_AL		0x0020
+#define		TMP102_CONF_CR0		0x0040
+#define		TMP102_CONF_CR1		0x0080
+#define	TMP102_TLOW_REG			0x02
+#define	TMP102_THIGH_REG		0x03
+
+struct tmp102 {
+	struct device *hwmon_dev;
+	struct mutex lock;
+	u16 config_orig;
+	unsigned long last_update;
+	int temp[3];
+};
+
+/* SMBus specifies low byte first, but the TMP102 returns high byte first,
+ * so we have to swab16 the values */
+static inline int tmp102_read_reg(struct i2c_client *client, u8 reg)
+{
+	int result = i2c_smbus_read_word_data(client, reg);
+	return result < 0 ? result : swab16(result);
+}
+
+static inline int tmp102_write_reg(struct i2c_client *client, u8 reg, u16 val)
+{
+	return i2c_smbus_write_word_data(client, reg, swab16(val));
+}
+
+/* convert left adjusted 13-bit TMP102 register value to milliCelsius */
+static inline int tmp102_reg_to_mC(s16 val)
+{
+	return ((val & ~0x01) * 1000) / 128;
+}
+
+/* convert milliCelsius to left adjusted 13-bit TMP102 register value */
+static inline u16 tmp102_mC_to_reg(int val)
+{
+	return (val * 128) / 1000;
+}
+
+static const u8 tmp102_reg[] = {
+	TMP102_TEMP_REG,
+	TMP102_TLOW_REG,
+	TMP102_THIGH_REG,
+};
+
+static struct tmp102 *tmp102_update_device(struct i2c_client *client)
+{
+	struct tmp102 *tmp102 = i2c_get_clientdata(client);
+
+	mutex_lock(&tmp102->lock);
+	if (time_after(jiffies, tmp102->last_update + HZ / 3)) {
+		int i;
+		for (i = 0; i < ARRAY_SIZE(tmp102->temp); ++i) {
+			int status = tmp102_read_reg(client, tmp102_reg[i]);
+			if (status > -1)
+				tmp102->temp[i] = tmp102_reg_to_mC(status);
+		}
+		tmp102->last_update = jiffies;
+	}
+	mutex_unlock(&tmp102->lock);
+	return tmp102;
+}
+
+static ssize_t tmp102_show_temp(struct device *dev,
+				struct device_attribute *attr,
+				char *buf)
+{
+	struct sensor_device_attribute *sda = to_sensor_dev_attr(attr);
+	struct tmp102 *tmp102 = tmp102_update_device(to_i2c_client(dev));
+
+	return sprintf(buf, "%d\n", tmp102->temp[sda->index]);
+}
+
+static ssize_t tmp102_set_temp(struct device *dev,
+			       struct device_attribute *attr,
+			       const char *buf, size_t count)
+{
+	struct sensor_device_attribute *sda = to_sensor_dev_attr(attr);
+	struct i2c_client *client = to_i2c_client(dev);
+	struct tmp102 *tmp102 = i2c_get_clientdata(client);
+	long val;
+	int status;
+
+	if (strict_strtol(buf, 10, &val) < 0)
+		return -EINVAL;
+	val = SENSORS_LIMIT(val, -256000, 255000);
+
+	mutex_lock(&tmp102->lock);
+	tmp102->temp[sda->index] = val;
+	status = tmp102_write_reg(client, tmp102_reg[sda->index],
+				  tmp102_mC_to_reg(val));
+	mutex_unlock(&tmp102->lock);
+	return status ? : count;
+}
+
+static SENSOR_DEVICE_ATTR(temp1_input, S_IRUGO, tmp102_show_temp, NULL , 0);
+
+static SENSOR_DEVICE_ATTR(temp1_max_hyst, S_IWUSR | S_IRUGO, tmp102_show_temp,
+			  tmp102_set_temp, 1);
+
+static SENSOR_DEVICE_ATTR(temp1_max, S_IWUSR | S_IRUGO, tmp102_show_temp,
+			  tmp102_set_temp, 2);
+
+static struct attribute *tmp102_attributes[] = {
+	&sensor_dev_attr_temp1_input.dev_attr.attr,
+	&sensor_dev_attr_temp1_max_hyst.dev_attr.attr,
+	&sensor_dev_attr_temp1_max.dev_attr.attr,
+	NULL
+};
+
+static const struct attribute_group tmp102_attr_group = {
+	.attrs = tmp102_attributes,
+};
+
+#define TMP102_CONFIG  (TMP102_CONF_TM | TMP102_CONF_EM | TMP102_CONF_CR1)
+#define TMP102_CONFIG_RD_ONLY (TMP102_CONF_R0 | TMP102_CONF_R1 | TMP102_CONF_AL)
+
+static int __devinit tmp102_probe(struct i2c_client *client,
+				  const struct i2c_device_id *id)
+{
+	struct tmp102 *tmp102;
+	int status;
+
+	if (!i2c_check_functionality(client->adapter,
+				     I2C_FUNC_SMBUS_WORD_DATA)) {
+		dev_err(&client->dev, "adapter doesnt support SMBus word "
+			"transactions\n");
+		return -ENODEV;
+	}
+
+	tmp102 = kzalloc(sizeof(*tmp102), GFP_KERNEL);
+	if (!tmp102) {
+		dev_dbg(&client->dev, "kzalloc failed\n");
+		return -ENOMEM;
+	}
+	i2c_set_clientdata(client, tmp102);
+
+	status = tmp102_read_reg(client, TMP102_CONF_REG);
+	if (status < 0) {
+		dev_err(&client->dev, "error reading config register\n");
+		goto fail_free;
+	}
+	tmp102->config_orig = status;
+	status = tmp102_write_reg(client, TMP102_CONF_REG, TMP102_CONFIG);
+	if (status < 0) {
+		dev_err(&client->dev, "error writing config register\n");
+		goto fail_restore_config;
+	}
+	status = tmp102_read_reg(client, TMP102_CONF_REG);
+	if (status < 0) {
+		dev_err(&client->dev, "error reading config register\n");
+		goto fail_restore_config;
+	}
+	status &= ~TMP102_CONFIG_RD_ONLY;
+	if (status != TMP102_CONFIG) {
+		dev_err(&client->dev, "config settings did not stick\n");
+		status = -ENODEV;
+		goto fail_restore_config;
+	}
+	tmp102->last_update = jiffies - HZ;
+	mutex_init(&tmp102->lock);
+
+	status = sysfs_create_group(&client->dev.kobj, &tmp102_attr_group);
+	if (status) {
+		dev_dbg(&client->dev, "could not create sysfs files\n");
+		goto fail_restore_config;
+	}
+	tmp102->hwmon_dev = hwmon_device_register(&client->dev);
+	if (IS_ERR(tmp102->hwmon_dev)) {
+		dev_dbg(&client->dev, "unable to register hwmon device\n");
+		status = PTR_ERR(tmp102->hwmon_dev);
+		goto fail_remove_sysfs;
+	}
+
+	dev_info(&client->dev, "initialized\n");
+
+	return 0;
+
+fail_remove_sysfs:
+	sysfs_remove_group(&client->dev.kobj, &tmp102_attr_group);
+fail_restore_config:
+	tmp102_write_reg(client, TMP102_CONF_REG, tmp102->config_orig);
+fail_free:
+	kfree(tmp102);
+
+	return status;
+}
+
+static int __devexit tmp102_remove(struct i2c_client *client)
+{
+	struct tmp102 *tmp102 = i2c_get_clientdata(client);
+
+	hwmon_device_unregister(tmp102->hwmon_dev);
+	sysfs_remove_group(&client->dev.kobj, &tmp102_attr_group);
+
+	/* Stop monitoring if device was stopped originally */
+	if (tmp102->config_orig & TMP102_CONF_SD) {
+		int config;
+
+		config = tmp102_read_reg(client, TMP102_CONF_REG);
+		if (config >= 0)
+			tmp102_write_reg(client, TMP102_CONF_REG,
+					 config | TMP102_CONF_SD);
+	}
+
+	kfree(tmp102);
+
+	return 0;
+}
+
+#ifdef CONFIG_PM
+static int tmp102_suspend(struct device *dev)
+{
+	struct i2c_client *client = to_i2c_client(dev);
+	int config;
+
+	config = tmp102_read_reg(client, TMP102_CONF_REG);
+	if (config < 0)
+		return config;
+
+	config |= TMP102_CONF_SD;
+	return tmp102_write_reg(client, TMP102_CONF_REG, config);
+}
+
+static int tmp102_resume(struct device *dev)
+{
+	struct i2c_client *client = to_i2c_client(dev);
+	int config;
+
+	config = tmp102_read_reg(client, TMP102_CONF_REG);
+	if (config < 0)
+		return config;
+
+	config &= ~TMP102_CONF_SD;
+	return tmp102_write_reg(client, TMP102_CONF_REG, config);
+}
+
+static const struct dev_pm_ops tmp102_dev_pm_ops = {
+	.suspend	= tmp102_suspend,
+	.resume		= tmp102_resume,
+};
+
+#define TMP102_DEV_PM_OPS (&tmp102_dev_pm_ops)
+#else
+#define	TMP102_DEV_PM_OPS NULL
+#endif /* CONFIG_PM */
+
+static const struct i2c_device_id tmp102_id[] = {
+	{ "tmp102", 0 },
+	{ }
+};
+MODULE_DEVICE_TABLE(i2c, tmp102_id);
+
+static struct i2c_driver tmp102_driver = {
+	.driver.name	= DRIVER_NAME,
+	.driver.pm	= TMP102_DEV_PM_OPS,
+	.probe		= tmp102_probe,
+	.remove		= __devexit_p(tmp102_remove),
+	.id_table	= tmp102_id,
+};
+
+static int __init tmp102_init(void)
+{
+	return i2c_add_driver(&tmp102_driver);
+}
+module_init(tmp102_init);
+
+static void __exit tmp102_exit(void)
+{
+	i2c_del_driver(&tmp102_driver);
+}
+module_exit(tmp102_exit);
+
+MODULE_AUTHOR("Steven King <sfking@fdwdc.com>");
+MODULE_DESCRIPTION("Texas Instruments TMP102 temperature sensor driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/hwmon/tmp401.c b/drivers/hwmon/tmp401.c
index d14a1af9f550..ad8d535235c5 100644
--- a/drivers/hwmon/tmp401.c
+++ b/drivers/hwmon/tmp401.c
@@ -92,17 +92,6 @@ static const u8 TMP411_TEMP_HIGHEST_LSB[2]		= { 0x33, 0x37 };
 #define TMP411_DEVICE_ID			0x12
 
 /*
- * Functions declarations
- */
-
-static int tmp401_probe(struct i2c_client *client,
-			const struct i2c_device_id *id);
-static int tmp401_detect(struct i2c_client *client,
-			 struct i2c_board_info *info);
-static int tmp401_remove(struct i2c_client *client);
-static struct tmp401_data *tmp401_update_device(struct device *dev);
-
-/*
  * Driver data (common to all clients)
  */
 
@@ -113,18 +102,6 @@ static const struct i2c_device_id tmp401_id[] = {
 };
 MODULE_DEVICE_TABLE(i2c, tmp401_id);
 
-static struct i2c_driver tmp401_driver = {
-	.class		= I2C_CLASS_HWMON,
-	.driver = {
-		.name	= "tmp401",
-	},
-	.probe		= tmp401_probe,
-	.remove		= tmp401_remove,
-	.id_table	= tmp401_id,
-	.detect		= tmp401_detect,
-	.address_list	= normal_i2c,
-};
-
 /*
  * Client data (each client gets its own)
  */
@@ -194,6 +171,71 @@ static u8 tmp401_crit_temp_to_register(long temp, u8 config)
 	return (temp + 500) / 1000;
 }
 
+static struct tmp401_data *tmp401_update_device_reg16(
+	struct i2c_client *client, struct tmp401_data *data)
+{
+	int i;
+
+	for (i = 0; i < 2; i++) {
+		/*
+		 * High byte must be read first immediately followed
+		 * by the low byte
+		 */
+		data->temp[i] = i2c_smbus_read_byte_data(client,
+			TMP401_TEMP_MSB[i]) << 8;
+		data->temp[i] |= i2c_smbus_read_byte_data(client,
+			TMP401_TEMP_LSB[i]);
+		data->temp_low[i] = i2c_smbus_read_byte_data(client,
+			TMP401_TEMP_LOW_LIMIT_MSB_READ[i]) << 8;
+		data->temp_low[i] |= i2c_smbus_read_byte_data(client,
+			TMP401_TEMP_LOW_LIMIT_LSB[i]);
+		data->temp_high[i] = i2c_smbus_read_byte_data(client,
+			TMP401_TEMP_HIGH_LIMIT_MSB_READ[i]) << 8;
+		data->temp_high[i] |= i2c_smbus_read_byte_data(client,
+			TMP401_TEMP_HIGH_LIMIT_LSB[i]);
+		data->temp_crit[i] = i2c_smbus_read_byte_data(client,
+			TMP401_TEMP_CRIT_LIMIT[i]);
+
+		if (data->kind == tmp411) {
+			data->temp_lowest[i] = i2c_smbus_read_byte_data(client,
+				TMP411_TEMP_LOWEST_MSB[i]) << 8;
+			data->temp_lowest[i] |= i2c_smbus_read_byte_data(
+				client, TMP411_TEMP_LOWEST_LSB[i]);
+
+			data->temp_highest[i] = i2c_smbus_read_byte_data(
+				client, TMP411_TEMP_HIGHEST_MSB[i]) << 8;
+			data->temp_highest[i] |= i2c_smbus_read_byte_data(
+				client, TMP411_TEMP_HIGHEST_LSB[i]);
+		}
+	}
+	return data;
+}
+
+static struct tmp401_data *tmp401_update_device(struct device *dev)
+{
+	struct i2c_client *client = to_i2c_client(dev);
+	struct tmp401_data *data = i2c_get_clientdata(client);
+
+	mutex_lock(&data->update_lock);
+
+	if (time_after(jiffies, data->last_updated + HZ) || !data->valid) {
+		data->status = i2c_smbus_read_byte_data(client, TMP401_STATUS);
+		data->config = i2c_smbus_read_byte_data(client,
+						TMP401_CONFIG_READ);
+		tmp401_update_device_reg16(client, data);
+
+		data->temp_crit_hyst = i2c_smbus_read_byte_data(client,
+						TMP401_TEMP_CRIT_HYST);
+
+		data->last_updated = jiffies;
+		data->valid = 1;
+	}
+
+	mutex_unlock(&data->update_lock);
+
+	return data;
+}
+
 static ssize_t show_temp_value(struct device *dev,
 	struct device_attribute *devattr, char *buf)
 {
@@ -420,30 +462,36 @@ static ssize_t reset_temp_history(struct device *dev,
 }
 
 static struct sensor_device_attribute tmp401_attr[] = {
-	SENSOR_ATTR(temp1_input, 0444, show_temp_value, NULL, 0),
-	SENSOR_ATTR(temp1_min, 0644, show_temp_min, store_temp_min, 0),
-	SENSOR_ATTR(temp1_max, 0644, show_temp_max, store_temp_max, 0),
-	SENSOR_ATTR(temp1_crit, 0644, show_temp_crit, store_temp_crit, 0),
-	SENSOR_ATTR(temp1_crit_hyst, 0644, show_temp_crit_hyst,
+	SENSOR_ATTR(temp1_input, S_IRUGO, show_temp_value, NULL, 0),
+	SENSOR_ATTR(temp1_min, S_IWUSR | S_IRUGO, show_temp_min,
+		    store_temp_min, 0),
+	SENSOR_ATTR(temp1_max, S_IWUSR | S_IRUGO, show_temp_max,
+		    store_temp_max, 0),
+	SENSOR_ATTR(temp1_crit, S_IWUSR | S_IRUGO, show_temp_crit,
+		    store_temp_crit, 0),
+	SENSOR_ATTR(temp1_crit_hyst, S_IWUSR | S_IRUGO, show_temp_crit_hyst,
 		    store_temp_crit_hyst, 0),
-	SENSOR_ATTR(temp1_min_alarm, 0444, show_status, NULL,
+	SENSOR_ATTR(temp1_min_alarm, S_IRUGO, show_status, NULL,
 		    TMP401_STATUS_LOCAL_LOW),
-	SENSOR_ATTR(temp1_max_alarm, 0444, show_status, NULL,
+	SENSOR_ATTR(temp1_max_alarm, S_IRUGO, show_status, NULL,
 		    TMP401_STATUS_LOCAL_HIGH),
-	SENSOR_ATTR(temp1_crit_alarm, 0444, show_status, NULL,
+	SENSOR_ATTR(temp1_crit_alarm, S_IRUGO, show_status, NULL,
 		    TMP401_STATUS_LOCAL_CRIT),
-	SENSOR_ATTR(temp2_input, 0444, show_temp_value, NULL, 1),
-	SENSOR_ATTR(temp2_min, 0644, show_temp_min, store_temp_min, 1),
-	SENSOR_ATTR(temp2_max, 0644, show_temp_max, store_temp_max, 1),
-	SENSOR_ATTR(temp2_crit, 0644, show_temp_crit, store_temp_crit, 1),
-	SENSOR_ATTR(temp2_crit_hyst, 0444, show_temp_crit_hyst, NULL, 1),
-	SENSOR_ATTR(temp2_fault, 0444, show_status, NULL,
+	SENSOR_ATTR(temp2_input, S_IRUGO, show_temp_value, NULL, 1),
+	SENSOR_ATTR(temp2_min, S_IWUSR | S_IRUGO, show_temp_min,
+		    store_temp_min, 1),
+	SENSOR_ATTR(temp2_max, S_IWUSR | S_IRUGO, show_temp_max,
+		    store_temp_max, 1),
+	SENSOR_ATTR(temp2_crit, S_IWUSR | S_IRUGO, show_temp_crit,
+		    store_temp_crit, 1),
+	SENSOR_ATTR(temp2_crit_hyst, S_IRUGO, show_temp_crit_hyst, NULL, 1),
+	SENSOR_ATTR(temp2_fault, S_IRUGO, show_status, NULL,
 		    TMP401_STATUS_REMOTE_OPEN),
-	SENSOR_ATTR(temp2_min_alarm, 0444, show_status, NULL,
+	SENSOR_ATTR(temp2_min_alarm, S_IRUGO, show_status, NULL,
 		    TMP401_STATUS_REMOTE_LOW),
-	SENSOR_ATTR(temp2_max_alarm, 0444, show_status, NULL,
+	SENSOR_ATTR(temp2_max_alarm, S_IRUGO, show_status, NULL,
 		    TMP401_STATUS_REMOTE_HIGH),
-	SENSOR_ATTR(temp2_crit_alarm, 0444, show_status, NULL,
+	SENSOR_ATTR(temp2_crit_alarm, S_IRUGO, show_status, NULL,
 		    TMP401_STATUS_REMOTE_CRIT),
 };
 
@@ -455,11 +503,11 @@ static struct sensor_device_attribute tmp401_attr[] = {
  * and remote channels.
  */
 static struct sensor_device_attribute tmp411_attr[] = {
-	SENSOR_ATTR(temp1_highest, 0444, show_temp_highest, NULL, 0),
-	SENSOR_ATTR(temp1_lowest, 0444, show_temp_lowest, NULL, 0),
-	SENSOR_ATTR(temp2_highest, 0444, show_temp_highest, NULL, 1),
-	SENSOR_ATTR(temp2_lowest, 0444, show_temp_lowest, NULL, 1),
-	SENSOR_ATTR(temp_reset_history, 0200, NULL, reset_temp_history, 0),
+	SENSOR_ATTR(temp1_highest, S_IRUGO, show_temp_highest, NULL, 0),
+	SENSOR_ATTR(temp1_lowest, S_IRUGO, show_temp_lowest, NULL, 0),
+	SENSOR_ATTR(temp2_highest, S_IRUGO, show_temp_highest, NULL, 1),
+	SENSOR_ATTR(temp2_lowest, S_IRUGO, show_temp_lowest, NULL, 1),
+	SENSOR_ATTR(temp_reset_history, S_IWUSR, NULL, reset_temp_history, 0),
 };
 
 /*
@@ -529,6 +577,27 @@ static int tmp401_detect(struct i2c_client *client,
 	return 0;
 }
 
+static int tmp401_remove(struct i2c_client *client)
+{
+	struct tmp401_data *data = i2c_get_clientdata(client);
+	int i;
+
+	if (data->hwmon_dev)
+		hwmon_device_unregister(data->hwmon_dev);
+
+	for (i = 0; i < ARRAY_SIZE(tmp401_attr); i++)
+		device_remove_file(&client->dev, &tmp401_attr[i].dev_attr);
+
+	if (data->kind == tmp411) {
+		for (i = 0; i < ARRAY_SIZE(tmp411_attr); i++)
+			device_remove_file(&client->dev,
+					   &tmp411_attr[i].dev_attr);
+	}
+
+	kfree(data);
+	return 0;
+}
+
 static int tmp401_probe(struct i2c_client *client,
 			const struct i2c_device_id *id)
 {
@@ -581,91 +650,17 @@ exit_remove:
 	return err;
 }
 
-static int tmp401_remove(struct i2c_client *client)
-{
-	struct tmp401_data *data = i2c_get_clientdata(client);
-	int i;
-
-	if (data->hwmon_dev)
-		hwmon_device_unregister(data->hwmon_dev);
-
-	for (i = 0; i < ARRAY_SIZE(tmp401_attr); i++)
-		device_remove_file(&client->dev, &tmp401_attr[i].dev_attr);
-
-	if (data->kind == tmp411) {
-		for (i = 0; i < ARRAY_SIZE(tmp411_attr); i++)
-			device_remove_file(&client->dev,
-					   &tmp411_attr[i].dev_attr);
-	}
-
-	kfree(data);
-	return 0;
-}
-
-static struct tmp401_data *tmp401_update_device_reg16(
-	struct i2c_client *client, struct tmp401_data *data)
-{
-	int i;
-
-	for (i = 0; i < 2; i++) {
-		/*
-		 * High byte must be read first immediately followed
-		 * by the low byte
-		 */
-		data->temp[i] = i2c_smbus_read_byte_data(client,
-			TMP401_TEMP_MSB[i]) << 8;
-		data->temp[i] |= i2c_smbus_read_byte_data(client,
-			TMP401_TEMP_LSB[i]);
-		data->temp_low[i] = i2c_smbus_read_byte_data(client,
-			TMP401_TEMP_LOW_LIMIT_MSB_READ[i]) << 8;
-		data->temp_low[i] |= i2c_smbus_read_byte_data(client,
-			TMP401_TEMP_LOW_LIMIT_LSB[i]);
-		data->temp_high[i] = i2c_smbus_read_byte_data(client,
-			TMP401_TEMP_HIGH_LIMIT_MSB_READ[i]) << 8;
-		data->temp_high[i] |= i2c_smbus_read_byte_data(client,
-			TMP401_TEMP_HIGH_LIMIT_LSB[i]);
-		data->temp_crit[i] = i2c_smbus_read_byte_data(client,
-			TMP401_TEMP_CRIT_LIMIT[i]);
-
-		if (data->kind == tmp411) {
-			data->temp_lowest[i] = i2c_smbus_read_byte_data(client,
-				TMP411_TEMP_LOWEST_MSB[i]) << 8;
-			data->temp_lowest[i] |= i2c_smbus_read_byte_data(
-				client, TMP411_TEMP_LOWEST_LSB[i]);
-
-			data->temp_highest[i] = i2c_smbus_read_byte_data(
-				client, TMP411_TEMP_HIGHEST_MSB[i]) << 8;
-			data->temp_highest[i] |= i2c_smbus_read_byte_data(
-				client, TMP411_TEMP_HIGHEST_LSB[i]);
-		}
-	}
-	return data;
-}
-
-static struct tmp401_data *tmp401_update_device(struct device *dev)
-{
-	struct i2c_client *client = to_i2c_client(dev);
-	struct tmp401_data *data = i2c_get_clientdata(client);
-
-	mutex_lock(&data->update_lock);
-
-	if (time_after(jiffies, data->last_updated + HZ) || !data->valid) {
-		data->status = i2c_smbus_read_byte_data(client, TMP401_STATUS);
-		data->config = i2c_smbus_read_byte_data(client,
-						TMP401_CONFIG_READ);
-		tmp401_update_device_reg16(client, data);
-
-		data->temp_crit_hyst = i2c_smbus_read_byte_data(client,
-						TMP401_TEMP_CRIT_HYST);
-
-		data->last_updated = jiffies;
-		data->valid = 1;
-	}
-
-	mutex_unlock(&data->update_lock);
-
-	return data;
-}
+static struct i2c_driver tmp401_driver = {
+	.class		= I2C_CLASS_HWMON,
+	.driver = {
+		.name	= "tmp401",
+	},
+	.probe		= tmp401_probe,
+	.remove		= tmp401_remove,
+	.id_table	= tmp401_id,
+	.detect		= tmp401_detect,
+	.address_list	= normal_i2c,
+};
 
 static int __init tmp401_init(void)
 {
diff --git a/drivers/hwmon/tmp421.c b/drivers/hwmon/tmp421.c
index 738c472ece27..6b4165c12092 100644
--- a/drivers/hwmon/tmp421.c
+++ b/drivers/hwmon/tmp421.c
@@ -295,7 +295,6 @@ exit_remove:
 	sysfs_remove_group(&client->dev.kobj, &tmp421_group);
 
 exit_free:
-	i2c_set_clientdata(client, NULL);
 	kfree(data);
 
 	return err;
@@ -308,7 +307,6 @@ static int tmp421_remove(struct i2c_client *client)
 	hwmon_device_unregister(data->hwmon_dev);
 	sysfs_remove_group(&client->dev.kobj, &tmp421_group);
 
-	i2c_set_clientdata(client, NULL);
 	kfree(data);
 
 	return 0;
diff --git a/drivers/hwmon/ultra45_env.c b/drivers/hwmon/ultra45_env.c
index 68e90abeba96..5da5942cf970 100644
--- a/drivers/hwmon/ultra45_env.c
+++ b/drivers/hwmon/ultra45_env.c
@@ -300,8 +300,11 @@ static const struct of_device_id env_match[] = {
 MODULE_DEVICE_TABLE(of, env_match);
 
 static struct of_platform_driver env_driver = {
-	.name		= "ultra45_env",
-	.match_table	= env_match,
+	.driver = {
+		.name = "ultra45_env",
+		.owner = THIS_MODULE,
+		.of_match_table = env_match,
+	},
 	.probe		= env_probe,
 	.remove		= __devexit_p(env_remove),
 };
diff --git a/drivers/hwmon/w83781d.c b/drivers/hwmon/w83781d.c
index 32d4adee73db..c84b9b4e6960 100644
--- a/drivers/hwmon/w83781d.c
+++ b/drivers/hwmon/w83781d.c
@@ -1197,7 +1197,6 @@ ERROR4:
 	if (data->lm75[1])
 		i2c_unregister_device(data->lm75[1]);
 ERROR3:
-	i2c_set_clientdata(client, NULL);
 	kfree(data);
 ERROR1:
 	return err;
@@ -1219,7 +1218,6 @@ w83781d_remove(struct i2c_client *client)
 	if (data->lm75[1])
 		i2c_unregister_device(data->lm75[1]);
 
-	i2c_set_clientdata(client, NULL);
 	kfree(data);
 
 	return 0;
diff --git a/drivers/hwmon/w83793.c b/drivers/hwmon/w83793.c
index 612807d97155..697202e27891 100644
--- a/drivers/hwmon/w83793.c
+++ b/drivers/hwmon/w83793.c
@@ -35,6 +35,7 @@
 #include <linux/slab.h>
 #include <linux/i2c.h>
 #include <linux/hwmon.h>
+#include <linux/smp_lock.h>
 #include <linux/hwmon-vid.h>
 #include <linux/hwmon-sysfs.h>
 #include <linux/err.h>
@@ -1319,8 +1320,8 @@ static ssize_t watchdog_write(struct file *filp, const char __user *buf,
 	return count;
 }
 
-static int watchdog_ioctl(struct inode *inode, struct file *filp,
-			  unsigned int cmd, unsigned long arg)
+static long watchdog_ioctl(struct file *filp, unsigned int cmd,
+			   unsigned long arg)
 {
 	static struct watchdog_info ident = {
 		.options = WDIOF_KEEPALIVEPING |
@@ -1332,6 +1333,7 @@ static int watchdog_ioctl(struct inode *inode, struct file *filp,
 	int val, ret = 0;
 	struct w83793_data *data = filp->private_data;
 
+	lock_kernel();
 	switch (cmd) {
 	case WDIOC_GETSUPPORT:
 		if (!nowayout)
@@ -1385,7 +1387,7 @@ static int watchdog_ioctl(struct inode *inode, struct file *filp,
 	default:
 		ret = -ENOTTY;
 	}
-
+	unlock_kernel();
 	return ret;
 }
 
@@ -1395,7 +1397,7 @@ static const struct file_operations watchdog_fops = {
 	.open = watchdog_open,
 	.release = watchdog_close,
 	.write = watchdog_write,
-	.ioctl = watchdog_ioctl,
+	.unlocked_ioctl = watchdog_ioctl,
 };
 
 /*
diff --git a/drivers/i2c/busses/Kconfig b/drivers/i2c/busses/Kconfig
index 87ab0568bb0e..bceafbfa7268 100644
--- a/drivers/i2c/busses/Kconfig
+++ b/drivers/i2c/busses/Kconfig
@@ -475,6 +475,26 @@ config I2C_PASEMI
 	help
 	  Supports the PA Semi PWRficient on-chip SMBus interfaces.
 
+config I2C_PCA_PLATFORM
+	tristate "PCA9564/PCA9665 as platform device"
+	select I2C_ALGOPCA
+	default n
+	help
+	  This driver supports a memory mapped Philips PCA9564/PCA9665
+	  parallel bus to I2C bus controller.
+
+	  This driver can also be built as a module.  If so, the module
+	  will be called i2c-pca-platform.
+
+config I2C_PMCMSP
+	tristate "PMC MSP I2C TWI Controller"
+	depends on PMC_MSP
+	help
+	  This driver supports the PMC TWI controller on MSP devices.
+
+	  This driver can also be built as module. If so, the module
+	  will be called i2c-pmcmsp.
+
 config I2C_PNX
 	tristate "I2C bus support for Philips PNX targets"
 	depends on ARCH_PNX4008
@@ -711,26 +731,6 @@ config I2C_PCA_ISA
 	  delays when I2C/SMBus chip drivers are loaded (e.g. at boot
 	  time).  If unsure, say N.
 
-config I2C_PCA_PLATFORM
-	tristate "PCA9564/PCA9665 as platform device"
-	select I2C_ALGOPCA
-	default n
-	help
-	  This driver supports a memory mapped Philips PCA9564/PCA9665
-	  parallel bus to I2C bus controller.
-
-	  This driver can also be built as a module.  If so, the module
-	  will be called i2c-pca-platform.
-
-config I2C_PMCMSP
-	tristate "PMC MSP I2C TWI Controller"
-	depends on PMC_MSP
-	help
-	  This driver supports the PMC TWI controller on MSP devices.
-
-	  This driver can also be built as module. If so, the module
-	  will be called i2c-pmcmsp.
-
 config I2C_SIBYTE
 	tristate "SiByte SMBus interface"
 	depends on SIBYTE_SB1xxx_SOC
diff --git a/drivers/i2c/busses/Makefile b/drivers/i2c/busses/Makefile
index 097236f631e8..936880bd1dc5 100644
--- a/drivers/i2c/busses/Makefile
+++ b/drivers/i2c/busses/Makefile
@@ -27,7 +27,7 @@ obj-$(CONFIG_I2C_VIAPRO)	+= i2c-viapro.o
 obj-$(CONFIG_I2C_HYDRA)		+= i2c-hydra.o
 obj-$(CONFIG_I2C_POWERMAC)	+= i2c-powermac.o
 
-# Embebbed system I2C/SMBus host controller drivers
+# Embedded system I2C/SMBus host controller drivers
 obj-$(CONFIG_I2C_AT91)		+= i2c-at91.o
 obj-$(CONFIG_I2C_AU1550)	+= i2c-au1550.o
 obj-$(CONFIG_I2C_BLACKFIN_TWI)	+= i2c-bfin-twi.o
@@ -46,6 +46,8 @@ obj-$(CONFIG_I2C_NOMADIK)	+= i2c-nomadik.o
 obj-$(CONFIG_I2C_OCORES)	+= i2c-ocores.o
 obj-$(CONFIG_I2C_OMAP)		+= i2c-omap.o
 obj-$(CONFIG_I2C_PASEMI)	+= i2c-pasemi.o
+obj-$(CONFIG_I2C_PCA_PLATFORM)	+= i2c-pca-platform.o
+obj-$(CONFIG_I2C_PMCMSP)	+= i2c-pmcmsp.o
 obj-$(CONFIG_I2C_PNX)		+= i2c-pnx.o
 obj-$(CONFIG_I2C_PXA)		+= i2c-pxa.o
 obj-$(CONFIG_I2C_S3C2410)	+= i2c-s3c2410.o
@@ -68,8 +70,6 @@ obj-$(CONFIG_I2C_TINY_USB)	+= i2c-tiny-usb.o
 obj-$(CONFIG_I2C_ACORN)		+= i2c-acorn.o
 obj-$(CONFIG_I2C_ELEKTOR)	+= i2c-elektor.o
 obj-$(CONFIG_I2C_PCA_ISA)	+= i2c-pca-isa.o
-obj-$(CONFIG_I2C_PCA_PLATFORM)	+= i2c-pca-platform.o
-obj-$(CONFIG_I2C_PMCMSP)	+= i2c-pmcmsp.o
 obj-$(CONFIG_I2C_SIBYTE)	+= i2c-sibyte.o
 obj-$(CONFIG_I2C_STUB)		+= i2c-stub.o
 obj-$(CONFIG_SCx200_ACB)	+= scx200_acb.o
diff --git a/drivers/i2c/busses/i2c-cpm.c b/drivers/i2c/busses/i2c-cpm.c
index 16948db38973..b02b4533651d 100644
--- a/drivers/i2c/busses/i2c-cpm.c
+++ b/drivers/i2c/busses/i2c-cpm.c
@@ -440,7 +440,7 @@ static int __devinit cpm_i2c_setup(struct cpm_i2c *cpm)
 
 	init_waitqueue_head(&cpm->i2c_wait);
 
-	cpm->irq = of_irq_to_resource(ofdev->node, 0, NULL);
+	cpm->irq = of_irq_to_resource(ofdev->dev.of_node, 0, NULL);
 	if (!cpm->irq)
 		return -EINVAL;
 
@@ -451,13 +451,13 @@ static int __devinit cpm_i2c_setup(struct cpm_i2c *cpm)
 		return ret;
 
 	/* I2C parameter RAM */
-	i2c_base = of_iomap(ofdev->node, 1);
+	i2c_base = of_iomap(ofdev->dev.of_node, 1);
 	if (i2c_base == NULL) {
 		ret = -EINVAL;
 		goto out_irq;
 	}
 
-	if (of_device_is_compatible(ofdev->node, "fsl,cpm1-i2c")) {
+	if (of_device_is_compatible(ofdev->dev.of_node, "fsl,cpm1-i2c")) {
 
 		/* Check for and use a microcode relocation patch. */
 		cpm->i2c_ram = i2c_base;
@@ -474,7 +474,7 @@ static int __devinit cpm_i2c_setup(struct cpm_i2c *cpm)
 
 		cpm->version = 1;
 
-	} else if (of_device_is_compatible(ofdev->node, "fsl,cpm2-i2c")) {
+	} else if (of_device_is_compatible(ofdev->dev.of_node, "fsl,cpm2-i2c")) {
 		cpm->i2c_addr = cpm_muram_alloc(sizeof(struct i2c_ram), 64);
 		cpm->i2c_ram = cpm_muram_addr(cpm->i2c_addr);
 		out_be16(i2c_base, cpm->i2c_addr);
@@ -489,24 +489,24 @@ static int __devinit cpm_i2c_setup(struct cpm_i2c *cpm)
 	}
 
 	/* I2C control/status registers */
-	cpm->i2c_reg = of_iomap(ofdev->node, 0);
+	cpm->i2c_reg = of_iomap(ofdev->dev.of_node, 0);
 	if (cpm->i2c_reg == NULL) {
 		ret = -EINVAL;
 		goto out_ram;
 	}
 
-	data = of_get_property(ofdev->node, "fsl,cpm-command", &len);
+	data = of_get_property(ofdev->dev.of_node, "fsl,cpm-command", &len);
 	if (!data || len != 4) {
 		ret = -EINVAL;
 		goto out_reg;
 	}
 	cpm->cp_command = *data;
 
-	data = of_get_property(ofdev->node, "linux,i2c-class", &len);
+	data = of_get_property(ofdev->dev.of_node, "linux,i2c-class", &len);
 	if (data && len == 4)
 		cpm->adap.class = *data;
 
-	data = of_get_property(ofdev->node, "clock-frequency", &len);
+	data = of_get_property(ofdev->dev.of_node, "clock-frequency", &len);
 	if (data && len == 4)
 		cpm->freq = *data;
 	else
@@ -661,7 +661,7 @@ static int __devinit cpm_i2c_probe(struct of_device *ofdev,
 
 	/* register new adapter to i2c module... */
 
-	data = of_get_property(ofdev->node, "linux,i2c-index", &len);
+	data = of_get_property(ofdev->dev.of_node, "linux,i2c-index", &len);
 	if (data && len == 4) {
 		cpm->adap.nr = *data;
 		result = i2c_add_numbered_adapter(&cpm->adap);
@@ -679,7 +679,7 @@ static int __devinit cpm_i2c_probe(struct of_device *ofdev,
 	/*
 	 * register OF I2C devices
 	 */
-	of_register_i2c_devices(&cpm->adap, ofdev->node);
+	of_register_i2c_devices(&cpm->adap, ofdev->dev.of_node);
 
 	return 0;
 out_shut:
@@ -718,13 +718,13 @@ static const struct of_device_id cpm_i2c_match[] = {
 MODULE_DEVICE_TABLE(of, cpm_i2c_match);
 
 static struct of_platform_driver cpm_i2c_driver = {
-	.match_table	= cpm_i2c_match,
 	.probe		= cpm_i2c_probe,
 	.remove		= __devexit_p(cpm_i2c_remove),
-	.driver		= {
-		.name	= "fsl-i2c-cpm",
-		.owner	= THIS_MODULE,
-	}
+	.driver = {
+		.name = "fsl-i2c-cpm",
+		.owner = THIS_MODULE,
+		.of_match_table = cpm_i2c_match,
+	},
 };
 
 static int __init cpm_i2c_init(void)
diff --git a/drivers/i2c/busses/i2c-ibm_iic.c b/drivers/i2c/busses/i2c-ibm_iic.c
index f8ccc0fe95a8..bf344135647a 100644
--- a/drivers/i2c/busses/i2c-ibm_iic.c
+++ b/drivers/i2c/busses/i2c-ibm_iic.c
@@ -664,7 +664,7 @@ static inline u8 iic_clckdiv(unsigned int opb)
 static int __devinit iic_request_irq(struct of_device *ofdev,
 				     struct ibm_iic_private *dev)
 {
-	struct device_node *np = ofdev->node;
+	struct device_node *np = ofdev->dev.of_node;
 	int irq;
 
 	if (iic_force_poll)
@@ -695,7 +695,7 @@ static int __devinit iic_request_irq(struct of_device *ofdev,
 static int __devinit iic_probe(struct of_device *ofdev,
 			       const struct of_device_id *match)
 {
-	struct device_node *np = ofdev->node;
+	struct device_node *np = ofdev->dev.of_node;
 	struct ibm_iic_private *dev;
 	struct i2c_adapter *adap;
 	const u32 *freq;
@@ -807,8 +807,11 @@ static const struct of_device_id ibm_iic_match[] = {
 };
 
 static struct of_platform_driver ibm_iic_driver = {
-	.name	= "ibm-iic",
-	.match_table = ibm_iic_match,
+	.driver = {
+		.name = "ibm-iic",
+		.owner = THIS_MODULE,
+		.of_match_table = ibm_iic_match,
+	},
 	.probe	= iic_probe,
 	.remove	= __devexit_p(iic_remove),
 };
diff --git a/drivers/i2c/busses/i2c-mpc.c b/drivers/i2c/busses/i2c-mpc.c
index e86cef300c7d..df00eb1f11f9 100644
--- a/drivers/i2c/busses/i2c-mpc.c
+++ b/drivers/i2c/busses/i2c-mpc.c
@@ -560,14 +560,14 @@ static int __devinit fsl_i2c_probe(struct of_device *op,
 
 	init_waitqueue_head(&i2c->queue);
 
-	i2c->base = of_iomap(op->node, 0);
+	i2c->base = of_iomap(op->dev.of_node, 0);
 	if (!i2c->base) {
 		dev_err(i2c->dev, "failed to map controller\n");
 		result = -ENOMEM;
 		goto fail_map;
 	}
 
-	i2c->irq = irq_of_parse_and_map(op->node, 0);
+	i2c->irq = irq_of_parse_and_map(op->dev.of_node, 0);
 	if (i2c->irq) { /* no i2c->irq implies polling */
 		result = request_irq(i2c->irq, mpc_i2c_isr,
 				     IRQF_SHARED, "i2c-mpc", i2c);
@@ -577,21 +577,22 @@ static int __devinit fsl_i2c_probe(struct of_device *op,
 		}
 	}
 
-	if (of_get_property(op->node, "fsl,preserve-clocking", NULL)) {
+	if (of_get_property(op->dev.of_node, "fsl,preserve-clocking", NULL)) {
 		clock = MPC_I2C_CLOCK_PRESERVE;
 	} else {
-		prop = of_get_property(op->node, "clock-frequency", &plen);
+		prop = of_get_property(op->dev.of_node, "clock-frequency",
+					&plen);
 		if (prop && plen == sizeof(u32))
 			clock = *prop;
 	}
 
 	if (match->data) {
 		struct mpc_i2c_data *data = match->data;
-		data->setup(op->node, i2c, clock, data->prescaler);
+		data->setup(op->dev.of_node, i2c, clock, data->prescaler);
 	} else {
 		/* Backwards compatibility */
-		if (of_get_property(op->node, "dfsrr", NULL))
-			mpc_i2c_setup_8xxx(op->node, i2c, clock, 0);
+		if (of_get_property(op->dev.of_node, "dfsrr", NULL))
+			mpc_i2c_setup_8xxx(op->dev.of_node, i2c, clock, 0);
 	}
 
 	dev_set_drvdata(&op->dev, i2c);
@@ -605,7 +606,7 @@ static int __devinit fsl_i2c_probe(struct of_device *op,
 		dev_err(i2c->dev, "failed to add adapter\n");
 		goto fail_add;
 	}
-	of_register_i2c_devices(&i2c->adap, op->node);
+	of_register_i2c_devices(&i2c->adap, op->dev.of_node);
 
 	return result;
 
@@ -674,12 +675,12 @@ MODULE_DEVICE_TABLE(of, mpc_i2c_of_match);
 
 /* Structure for a device driver */
 static struct of_platform_driver mpc_i2c_driver = {
-	.match_table	= mpc_i2c_of_match,
 	.probe		= fsl_i2c_probe,
 	.remove		= __devexit_p(fsl_i2c_remove),
-	.driver		= {
-		.owner	= THIS_MODULE,
-		.name	= DRV_NAME,
+	.driver = {
+		.owner = THIS_MODULE,
+		.name = DRV_NAME,
+		.of_match_table = mpc_i2c_of_match,
 	},
 };
 
diff --git a/drivers/i2c/i2c-core.c b/drivers/i2c/i2c-core.c
index db3c9f3a7647..1cca2631e5b3 100644
--- a/drivers/i2c/i2c-core.c
+++ b/drivers/i2c/i2c-core.c
@@ -47,7 +47,6 @@ static DEFINE_MUTEX(core_lock);
 static DEFINE_IDR(i2c_adapter_idr);
 
 static struct device_type i2c_client_type;
-static int i2c_check_addr(struct i2c_adapter *adapter, int addr);
 static int i2c_detect(struct i2c_adapter *adapter, struct i2c_driver *driver);
 
 /* ------------------------------------------------------------------------- */
@@ -371,6 +370,59 @@ struct i2c_client *i2c_verify_client(struct device *dev)
 EXPORT_SYMBOL(i2c_verify_client);
 
 
+/* This is a permissive address validity check, I2C address map constraints
+ * are purposedly not enforced, except for the general call address. */
+static int i2c_check_client_addr_validity(const struct i2c_client *client)
+{
+	if (client->flags & I2C_CLIENT_TEN) {
+		/* 10-bit address, all values are valid */
+		if (client->addr > 0x3ff)
+			return -EINVAL;
+	} else {
+		/* 7-bit address, reject the general call address */
+		if (client->addr == 0x00 || client->addr > 0x7f)
+			return -EINVAL;
+	}
+	return 0;
+}
+
+/* And this is a strict address validity check, used when probing. If a
+ * device uses a reserved address, then it shouldn't be probed. 7-bit
+ * addressing is assumed, 10-bit address devices are rare and should be
+ * explicitly enumerated. */
+static int i2c_check_addr_validity(unsigned short addr)
+{
+	/*
+	 * Reserved addresses per I2C specification:
+	 *  0x00       General call address / START byte
+	 *  0x01       CBUS address
+	 *  0x02       Reserved for different bus format
+	 *  0x03       Reserved for future purposes
+	 *  0x04-0x07  Hs-mode master code
+	 *  0x78-0x7b  10-bit slave addressing
+	 *  0x7c-0x7f  Reserved for future purposes
+	 */
+	if (addr < 0x08 || addr > 0x77)
+		return -EINVAL;
+	return 0;
+}
+
+static int __i2c_check_addr_busy(struct device *dev, void *addrp)
+{
+	struct i2c_client	*client = i2c_verify_client(dev);
+	int			addr = *(int *)addrp;
+
+	if (client && client->addr == addr)
+		return -EBUSY;
+	return 0;
+}
+
+static int i2c_check_addr_busy(struct i2c_adapter *adapter, int addr)
+{
+	return device_for_each_child(&adapter->dev, &addr,
+				     __i2c_check_addr_busy);
+}
+
 /**
  * i2c_new_device - instantiate an i2c device
  * @adap: the adapter managing the device
@@ -410,14 +462,25 @@ i2c_new_device(struct i2c_adapter *adap, struct i2c_board_info const *info)
 
 	strlcpy(client->name, info->type, sizeof(client->name));
 
+	/* Check for address validity */
+	status = i2c_check_client_addr_validity(client);
+	if (status) {
+		dev_err(&adap->dev, "Invalid %d-bit I2C address 0x%02hx\n",
+			client->flags & I2C_CLIENT_TEN ? 10 : 7, client->addr);
+		goto out_err_silent;
+	}
+
 	/* Check for address business */
-	status = i2c_check_addr(adap, client->addr);
+	status = i2c_check_addr_busy(adap, client->addr);
 	if (status)
 		goto out_err;
 
 	client->dev.parent = &client->adapter->dev;
 	client->dev.bus = &i2c_bus_type;
 	client->dev.type = &i2c_client_type;
+#ifdef CONFIG_OF
+	client->dev.of_node = info->of_node;
+#endif
 
 	dev_set_name(&client->dev, "%d-%04x", i2c_adapter_id(adap),
 		     client->addr);
@@ -433,6 +496,7 @@ i2c_new_device(struct i2c_adapter *adap, struct i2c_board_info const *info)
 out_err:
 	dev_err(&adap->dev, "Failed to register i2c client %s at 0x%02x "
 		"(%d)\n", client->name, client->addr, status);
+out_err_silent:
 	kfree(client);
 	return NULL;
 }
@@ -558,15 +622,9 @@ i2c_sysfs_new_device(struct device *dev, struct device_attribute *attr,
 		return -EINVAL;
 	}
 
-	if (info.addr < 0x03 || info.addr > 0x77) {
-		dev_err(dev, "%s: Invalid I2C address 0x%hx\n", "new_device",
-			info.addr);
-		return -EINVAL;
-	}
-
 	client = i2c_new_device(adap, &info);
 	if (!client)
-		return -EEXIST;
+		return -EINVAL;
 
 	/* Keep track of the added device */
 	i2c_lock_adapter(adap);
@@ -1021,21 +1079,6 @@ EXPORT_SYMBOL(i2c_del_driver);
 
 /* ------------------------------------------------------------------------- */
 
-static int __i2c_check_addr(struct device *dev, void *addrp)
-{
-	struct i2c_client	*client = i2c_verify_client(dev);
-	int			addr = *(int *)addrp;
-
-	if (client && client->addr == addr)
-		return -EBUSY;
-	return 0;
-}
-
-static int i2c_check_addr(struct i2c_adapter *adapter, int addr)
-{
-	return device_for_each_child(&adapter->dev, &addr, __i2c_check_addr);
-}
-
 /**
  * i2c_use_client - increments the reference count of the i2c client structure
  * @client: the client being referenced
@@ -1274,6 +1317,41 @@ EXPORT_SYMBOL(i2c_master_recv);
  * ----------------------------------------------------
  */
 
+/*
+ * Legacy default probe function, mostly relevant for SMBus. The default
+ * probe method is a quick write, but it is known to corrupt the 24RF08
+ * EEPROMs due to a state machine bug, and could also irreversibly
+ * write-protect some EEPROMs, so for address ranges 0x30-0x37 and 0x50-0x5f,
+ * we use a short byte read instead. Also, some bus drivers don't implement
+ * quick write, so we fallback to a byte read in that case too.
+ * On x86, there is another special case for FSC hardware monitoring chips,
+ * which want regular byte reads (address 0x73.) Fortunately, these are the
+ * only known chips using this I2C address on PC hardware.
+ * Returns 1 if probe succeeded, 0 if not.
+ */
+static int i2c_default_probe(struct i2c_adapter *adap, unsigned short addr)
+{
+	int err;
+	union i2c_smbus_data dummy;
+
+#ifdef CONFIG_X86
+	if (addr == 0x73 && (adap->class & I2C_CLASS_HWMON)
+	 && i2c_check_functionality(adap, I2C_FUNC_SMBUS_READ_BYTE_DATA))
+		err = i2c_smbus_xfer(adap, addr, 0, I2C_SMBUS_READ, 0,
+				     I2C_SMBUS_BYTE_DATA, &dummy);
+	else
+#endif
+	if ((addr & ~0x07) == 0x30 || (addr & ~0x0f) == 0x50
+	 || !i2c_check_functionality(adap, I2C_FUNC_SMBUS_QUICK))
+		err = i2c_smbus_xfer(adap, addr, 0, I2C_SMBUS_READ, 0,
+				     I2C_SMBUS_BYTE, &dummy);
+	else
+		err = i2c_smbus_xfer(adap, addr, 0, I2C_SMBUS_WRITE, 0,
+				     I2C_SMBUS_QUICK, NULL);
+
+	return err >= 0;
+}
+
 static int i2c_detect_address(struct i2c_client *temp_client,
 			      struct i2c_driver *driver)
 {
@@ -1283,34 +1361,20 @@ static int i2c_detect_address(struct i2c_client *temp_client,
 	int err;
 
 	/* Make sure the address is valid */
-	if (addr < 0x03 || addr > 0x77) {
+	err = i2c_check_addr_validity(addr);
+	if (err) {
 		dev_warn(&adapter->dev, "Invalid probe address 0x%02x\n",
 			 addr);
-		return -EINVAL;
+		return err;
 	}
 
 	/* Skip if already in use */
-	if (i2c_check_addr(adapter, addr))
+	if (i2c_check_addr_busy(adapter, addr))
 		return 0;
 
 	/* Make sure there is something at this address */
-	if (addr == 0x73 && (adapter->class & I2C_CLASS_HWMON)) {
-		/* Special probe for FSC hwmon chips */
-		union i2c_smbus_data dummy;
-
-		if (i2c_smbus_xfer(adapter, addr, 0, I2C_SMBUS_READ, 0,
-				   I2C_SMBUS_BYTE_DATA, &dummy) < 0)
-			return 0;
-	} else {
-		if (i2c_smbus_xfer(adapter, addr, 0, I2C_SMBUS_WRITE, 0,
-				   I2C_SMBUS_QUICK, NULL) < 0)
-			return 0;
-
-		/* Prevent 24RF08 corruption */
-		if ((addr & ~0x0f) == 0x50)
-			i2c_smbus_xfer(adapter, addr, 0, I2C_SMBUS_WRITE, 0,
-				       I2C_SMBUS_QUICK, NULL);
-	}
+	if (!i2c_default_probe(adapter, addr))
+		return 0;
 
 	/* Finally call the custom detection function */
 	memset(&info, 0, sizeof(struct i2c_board_info));
@@ -1404,42 +1468,22 @@ i2c_new_probed_device(struct i2c_adapter *adap,
 
 	for (i = 0; addr_list[i] != I2C_CLIENT_END; i++) {
 		/* Check address validity */
-		if (addr_list[i] < 0x03 || addr_list[i] > 0x77) {
+		if (i2c_check_addr_validity(addr_list[i]) < 0) {
 			dev_warn(&adap->dev, "Invalid 7-bit address "
 				 "0x%02x\n", addr_list[i]);
 			continue;
 		}
 
 		/* Check address availability */
-		if (i2c_check_addr(adap, addr_list[i])) {
+		if (i2c_check_addr_busy(adap, addr_list[i])) {
 			dev_dbg(&adap->dev, "Address 0x%02x already in "
 				"use, not probing\n", addr_list[i]);
 			continue;
 		}
 
-		/* Test address responsiveness
-		   The default probe method is a quick write, but it is known
-		   to corrupt the 24RF08 EEPROMs due to a state machine bug,
-		   and could also irreversibly write-protect some EEPROMs, so
-		   for address ranges 0x30-0x37 and 0x50-0x5f, we use a byte
-		   read instead. Also, some bus drivers don't implement
-		   quick write, so we fallback to a byte read it that case
-		   too. */
-		if ((addr_list[i] & ~0x07) == 0x30
-		 || (addr_list[i] & ~0x0f) == 0x50
-		 || !i2c_check_functionality(adap, I2C_FUNC_SMBUS_QUICK)) {
-			union i2c_smbus_data data;
-
-			if (i2c_smbus_xfer(adap, addr_list[i], 0,
-					   I2C_SMBUS_READ, 0,
-					   I2C_SMBUS_BYTE, &data) >= 0)
-				break;
-		} else {
-			if (i2c_smbus_xfer(adap, addr_list[i], 0,
-					   I2C_SMBUS_WRITE, 0,
-					   I2C_SMBUS_QUICK, NULL) >= 0)
-				break;
-		}
+		/* Test address responsiveness */
+		if (i2c_default_probe(adap, addr_list[i]))
+			break;
 	}
 
 	if (addr_list[i] == I2C_CLIENT_END) {
diff --git a/drivers/i2c/i2c-smbus.c b/drivers/i2c/i2c-smbus.c
index a24e0bfe9201..f61ccc1e5ea3 100644
--- a/drivers/i2c/i2c-smbus.c
+++ b/drivers/i2c/i2c-smbus.c
@@ -173,7 +173,6 @@ static int smbalert_remove(struct i2c_client *ara)
 
 	cancel_work_sync(&alert->alert);
 
-	i2c_set_clientdata(ara, NULL);
 	kfree(alert);
 	return 0;
 }
diff --git a/drivers/ide/cmd640.c b/drivers/ide/cmd640.c
index d2b8b272bc27..cb10201a15ed 100644
--- a/drivers/ide/cmd640.c
+++ b/drivers/ide/cmd640.c
@@ -633,12 +633,10 @@ static void __init cmd640_init_dev(ide_drive_t *drive)
 
 static int cmd640_test_irq(ide_hwif_t *hwif)
 {
-	struct pci_dev *dev	= to_pci_dev(hwif->dev);
 	int irq_reg		= hwif->channel ? ARTTIM23 : CFR;
-	u8  irq_stat, irq_mask	= hwif->channel ? ARTTIM23_IDE23INTR :
+	u8  irq_mask		= hwif->channel ? ARTTIM23_IDE23INTR :
 						  CFR_IDE01INTR;
-
-	pci_read_config_byte(dev, irq_reg, &irq_stat);
+	u8  irq_stat		= get_cmd640_reg(irq_reg);
 
 	return (irq_stat & irq_mask) ? 1 : 0;
 }
diff --git a/drivers/ide/gayle.c b/drivers/ide/gayle.c
index b9e517de6a82..3feaa26410be 100644
--- a/drivers/ide/gayle.c
+++ b/drivers/ide/gayle.c
@@ -16,6 +16,7 @@
 #include <linux/init.h>
 #include <linux/zorro.h>
 #include <linux/module.h>
+#include <linux/platform_device.h>
 
 #include <asm/setup.h>
 #include <asm/amigahw.h>
@@ -24,15 +25,6 @@
 
 
     /*
-     *  Bases of the IDE interfaces
-     */
-
-#define GAYLE_BASE_4000	0xdd2020	/* A4000/A4000T */
-#define GAYLE_BASE_1200	0xda0000	/* A1200/A600 and E-Matrix 530 */
-
-#define GAYLE_IDEREG_SIZE	0x2000
-
-    /*
      *  Offsets from one of the above bases
      */
 
@@ -68,20 +60,20 @@ MODULE_PARM_DESC(doubler, "enable support for IDE doublers");
 
 static int gayle_test_irq(ide_hwif_t *hwif)
 {
-    unsigned char ch;
+	unsigned char ch;
 
-    ch = z_readb(hwif->io_ports.irq_addr);
-    if (!(ch & GAYLE_IRQ_IDE))
-	return 0;
-    return 1;
+	ch = z_readb(hwif->io_ports.irq_addr);
+	if (!(ch & GAYLE_IRQ_IDE))
+		return 0;
+	return 1;
 }
 
 static void gayle_a1200_clear_irq(ide_drive_t *drive)
 {
-    ide_hwif_t *hwif = drive->hwif;
+	ide_hwif_t *hwif = drive->hwif;
 
-    (void)z_readb(hwif->io_ports.status_addr);
-    z_writeb(0x7c, hwif->io_ports.irq_addr);
+	(void)z_readb(hwif->io_ports.status_addr);
+	z_writeb(0x7c, hwif->io_ports.irq_addr);
 }
 
 static void __init gayle_setup_ports(struct ide_hw *hw, unsigned long base,
@@ -122,64 +114,89 @@ static const struct ide_port_info gayle_port_info = {
      *  Probe for a Gayle IDE interface (and optionally for an IDE doubler)
      */
 
-static int __init gayle_init(void)
+static int __init amiga_gayle_ide_probe(struct platform_device *pdev)
 {
-    unsigned long phys_base, res_start, res_n;
-    unsigned long base, ctrlport, irqport;
-    int a4000, i, rc;
-    struct ide_hw hw[GAYLE_NUM_HWIFS], *hws[GAYLE_NUM_HWIFS];
-    struct ide_port_info d = gayle_port_info;
-
-    if (!MACH_IS_AMIGA)
-	return -ENODEV;
-
-    if ((a4000 = AMIGAHW_PRESENT(A4000_IDE)) || AMIGAHW_PRESENT(A1200_IDE))
-	goto found;
-
-#ifdef CONFIG_ZORRO
-    if (zorro_find_device(ZORRO_PROD_MTEC_VIPER_MK_V_E_MATRIX_530_SCSI_IDE,
-			  NULL))
-	goto found;
-#endif
-    return -ENODEV;
-
-found:
-	printk(KERN_INFO "ide: Gayle IDE controller (A%d style%s)\n",
-			 a4000 ? 4000 : 1200,
-			 ide_doubler ? ", IDE doubler" : "");
-
-	if (a4000) {
-	    phys_base = GAYLE_BASE_4000;
-	    irqport = (unsigned long)ZTWO_VADDR(GAYLE_IRQ_4000);
-	    d.port_ops = &gayle_a4000_port_ops;
-	} else {
-	    phys_base = GAYLE_BASE_1200;
-	    irqport = (unsigned long)ZTWO_VADDR(GAYLE_IRQ_1200);
-	    d.port_ops = &gayle_a1200_port_ops;
+	struct resource *res;
+	struct gayle_ide_platform_data *pdata;
+	unsigned long base, ctrlport, irqport;
+	unsigned int i;
+	int error;
+	struct ide_hw hw[GAYLE_NUM_HWIFS], *hws[GAYLE_NUM_HWIFS];
+	struct ide_port_info d = gayle_port_info;
+	struct ide_host *host;
+
+	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	if (!res)
+		return -ENODEV;
+
+	if (!request_mem_region(res->start, resource_size(res), "IDE"))
+		return -EBUSY;
+
+	pdata = pdev->dev.platform_data;
+	pr_info("ide: Gayle IDE controller (A%u style%s)\n",
+		pdata->explicit_ack ? 1200 : 4000,
+		ide_doubler ? ", IDE doubler" : "");
+
+	base = (unsigned long)ZTWO_VADDR(pdata->base);
+	ctrlport = 0;
+	irqport = (unsigned long)ZTWO_VADDR(pdata->irqport);
+	if (pdata->explicit_ack)
+		d.port_ops = &gayle_a1200_port_ops;
+	else
+		d.port_ops = &gayle_a4000_port_ops;
+
+	for (i = 0; i < GAYLE_NUM_PROBE_HWIFS; i++, base += GAYLE_NEXT_PORT) {
+		if (GAYLE_HAS_CONTROL_REG)
+			ctrlport = base + GAYLE_CONTROL;
+
+		gayle_setup_ports(&hw[i], base, ctrlport, irqport);
+		hws[i] = &hw[i];
 	}
 
-	res_start = ((unsigned long)phys_base) & ~(GAYLE_NEXT_PORT-1);
-	res_n = GAYLE_IDEREG_SIZE;
+	error = ide_host_add(&d, hws, i, &host);
+	if (error)
+		goto out;
 
-	if (!request_mem_region(res_start, res_n, "IDE"))
-		return -EBUSY;
+	platform_set_drvdata(pdev, host);
+	return 0;
 
-    for (i = 0; i < GAYLE_NUM_PROBE_HWIFS; i++) {
-	base = (unsigned long)ZTWO_VADDR(phys_base + i * GAYLE_NEXT_PORT);
-	ctrlport = GAYLE_HAS_CONTROL_REG ? (base + GAYLE_CONTROL) : 0;
+out:
+	release_mem_region(res->start, resource_size(res));
+	return error;
+}
+
+static int __exit amiga_gayle_ide_remove(struct platform_device *pdev)
+{
+	struct ide_host *host = platform_get_drvdata(pdev);
+	struct resource *res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+
+	ide_host_remove(host);
+	release_mem_region(res->start, resource_size(res));
+	return 0;
+}
 
-	gayle_setup_ports(&hw[i], base, ctrlport, irqport);
+static struct platform_driver amiga_gayle_ide_driver = {
+	.remove = __exit_p(amiga_gayle_ide_remove),
+	.driver   = {
+		.name	= "amiga-gayle-ide",
+		.owner	= THIS_MODULE,
+	},
+};
 
-	hws[i] = &hw[i];
-    }
+static int __init amiga_gayle_ide_init(void)
+{
+	return platform_driver_probe(&amiga_gayle_ide_driver,
+				     amiga_gayle_ide_probe);
+}
 
-    rc = ide_host_add(&d, hws, i, NULL);
-    if (rc)
-	release_mem_region(res_start, res_n);
+module_init(amiga_gayle_ide_init);
 
-    return rc;
+static void __exit amiga_gayle_ide_exit(void)
+{
+	platform_driver_unregister(&amiga_gayle_ide_driver);
 }
 
-module_init(gayle_init);
+module_exit(amiga_gayle_ide_exit);
 
 MODULE_LICENSE("GPL");
+MODULE_ALIAS("platform:amiga-gayle-ide");
diff --git a/drivers/ide/ide-disk.c b/drivers/ide/ide-disk.c
index 3b128dce9c3a..33d65039cce9 100644
--- a/drivers/ide/ide-disk.c
+++ b/drivers/ide/ide-disk.c
@@ -407,32 +407,24 @@ static int ide_disk_get_capacity(ide_drive_t *drive)
 	return 0;
 }
 
-static u64 ide_disk_set_capacity(ide_drive_t *drive, u64 capacity)
+static void ide_disk_unlock_native_capacity(ide_drive_t *drive)
 {
-	u64 set = min(capacity, drive->probed_capacity);
 	u16 *id = drive->id;
 	int lba48 = ata_id_lba48_enabled(id);
 
 	if ((drive->dev_flags & IDE_DFLAG_LBA) == 0 ||
 	    ata_id_hpa_enabled(id) == 0)
-		goto out;
+		return;
 
 	/*
 	 * according to the spec the SET MAX ADDRESS command shall be
 	 * immediately preceded by a READ NATIVE MAX ADDRESS command
 	 */
-	capacity = ide_disk_hpa_get_native_capacity(drive, lba48);
-	if (capacity == 0)
-		goto out;
-
-	set = ide_disk_hpa_set_capacity(drive, set, lba48);
-	if (set) {
-		/* needed for ->resume to disable HPA */
-		drive->dev_flags |= IDE_DFLAG_NOHPA;
-		return set;
-	}
-out:
-	return drive->capacity64;
+	if (!ide_disk_hpa_get_native_capacity(drive, lba48))
+		return;
+
+	if (ide_disk_hpa_set_capacity(drive, drive->probed_capacity, lba48))
+		drive->dev_flags |= IDE_DFLAG_NOHPA; /* disable HPA on resume */
 }
 
 static void idedisk_prepare_flush(struct request_queue *q, struct request *rq)
@@ -783,13 +775,13 @@ static int ide_disk_set_doorlock(ide_drive_t *drive, struct gendisk *disk,
 }
 
 const struct ide_disk_ops ide_ata_disk_ops = {
-	.check		= ide_disk_check,
-	.set_capacity	= ide_disk_set_capacity,
-	.get_capacity	= ide_disk_get_capacity,
-	.setup		= ide_disk_setup,
-	.flush		= ide_disk_flush,
-	.init_media	= ide_disk_init_media,
-	.set_doorlock	= ide_disk_set_doorlock,
-	.do_request	= ide_do_rw_disk,
-	.ioctl		= ide_disk_ioctl,
+	.check			= ide_disk_check,
+	.unlock_native_capacity	= ide_disk_unlock_native_capacity,
+	.get_capacity		= ide_disk_get_capacity,
+	.setup			= ide_disk_setup,
+	.flush			= ide_disk_flush,
+	.init_media		= ide_disk_init_media,
+	.set_doorlock		= ide_disk_set_doorlock,
+	.do_request		= ide_do_rw_disk,
+	.ioctl			= ide_disk_ioctl,
 };
diff --git a/drivers/ide/ide-gd.c b/drivers/ide/ide-gd.c
index 27d9fe08d80b..79399534782c 100644
--- a/drivers/ide/ide-gd.c
+++ b/drivers/ide/ide-gd.c
@@ -288,17 +288,14 @@ static int ide_gd_media_changed(struct gendisk *disk)
 	return ret;
 }
 
-static unsigned long long ide_gd_set_capacity(struct gendisk *disk,
-					      unsigned long long capacity)
+static void ide_gd_unlock_native_capacity(struct gendisk *disk)
 {
 	struct ide_disk_obj *idkp = ide_drv_g(disk, ide_disk_obj);
 	ide_drive_t *drive = idkp->drive;
 	const struct ide_disk_ops *disk_ops = drive->disk_ops;
 
-	if (disk_ops->set_capacity)
-		return disk_ops->set_capacity(drive, capacity);
-
-	return drive->capacity64;
+	if (disk_ops->unlock_native_capacity)
+		disk_ops->unlock_native_capacity(drive);
 }
 
 static int ide_gd_revalidate_disk(struct gendisk *disk)
@@ -329,7 +326,7 @@ static const struct block_device_operations ide_gd_ops = {
 	.locked_ioctl		= ide_gd_ioctl,
 	.getgeo			= ide_gd_getgeo,
 	.media_changed		= ide_gd_media_changed,
-	.set_capacity		= ide_gd_set_capacity,
+	.unlock_native_capacity	= ide_gd_unlock_native_capacity,
 	.revalidate_disk	= ide_gd_revalidate_disk
 };
 
diff --git a/drivers/ide/ide_platform.c b/drivers/ide/ide_platform.c
index 42965b3e30b9..542603b394e4 100644
--- a/drivers/ide/ide_platform.c
+++ b/drivers/ide/ide_platform.c
@@ -95,6 +95,7 @@ static int __devinit plat_ide_probe(struct platform_device *pdev)
 	plat_ide_setup_ports(&hw, base, alt_base, pdata, res_irq->start);
 	hw.dev = &pdev->dev;
 
+	d.irq_flags = res_irq->flags;
 	if (mmio)
 		d.host_flags |= IDE_HFLAG_MMIO;
 
diff --git a/drivers/ide/pdc202xx_old.c b/drivers/ide/pdc202xx_old.c
index c5f3841af360..3a35ec6193d2 100644
--- a/drivers/ide/pdc202xx_old.c
+++ b/drivers/ide/pdc202xx_old.c
@@ -93,13 +93,13 @@ static int pdc202xx_test_irq(ide_hwif_t *hwif)
 		 * bit 7: error, bit 6: interrupting,
 		 * bit 5: FIFO full, bit 4: FIFO empty
 		 */
-		return ((sc1d & 0x50) == 0x50) ? 1 : 0;
+		return (sc1d & 0x40) ? 1 : 0;
 	} else	{
 		/*
 		 * bit 3: error, bit 2: interrupting,
 		 * bit 1: FIFO full, bit 0: FIFO empty
 		 */
-		return ((sc1d & 0x05) == 0x05) ? 1 : 0;
+		return (sc1d & 0x04) ? 1 : 0;
 	}
 }
 
@@ -241,6 +241,7 @@ static const struct ide_port_ops pdc20246_port_ops = {
 static const struct ide_port_ops pdc2026x_port_ops = {
 	.set_pio_mode		= pdc202xx_set_pio_mode,
 	.set_dma_mode		= pdc202xx_set_mode,
+	.test_irq		= pdc202xx_test_irq,
 	.cable_detect		= pdc2026x_cable_detect,
 };
 
diff --git a/drivers/ide/pmac.c b/drivers/ide/pmac.c
index 159955d16c47..ebcf8e470a97 100644
--- a/drivers/ide/pmac.c
+++ b/drivers/ide/pmac.c
@@ -1153,7 +1153,7 @@ pmac_ide_macio_attach(struct macio_dev *mdev, const struct of_device_id *match)
 
 	if (macio_resource_count(mdev) == 0) {
 		printk(KERN_WARNING "ide-pmac: no address for %s\n",
-				    mdev->ofdev.node->full_name);
+				    mdev->ofdev.dev.of_node->full_name);
 		rc = -ENXIO;
 		goto out_free_pmif;
 	}
@@ -1161,7 +1161,7 @@ pmac_ide_macio_attach(struct macio_dev *mdev, const struct of_device_id *match)
 	/* Request memory resource for IO ports */
 	if (macio_request_resource(mdev, 0, "ide-pmac (ports)")) {
 		printk(KERN_ERR "ide-pmac: can't request MMIO resource for "
-				"%s!\n", mdev->ofdev.node->full_name);
+				"%s!\n", mdev->ofdev.dev.of_node->full_name);
 		rc = -EBUSY;
 		goto out_free_pmif;
 	}
@@ -1173,7 +1173,7 @@ pmac_ide_macio_attach(struct macio_dev *mdev, const struct of_device_id *match)
 	 */
 	if (macio_irq_count(mdev) == 0) {
 		printk(KERN_WARNING "ide-pmac: no intrs for device %s, using "
-				    "13\n", mdev->ofdev.node->full_name);
+				    "13\n", mdev->ofdev.dev.of_node->full_name);
 		irq = irq_create_mapping(NULL, 13);
 	} else
 		irq = macio_irq(mdev, 0);
@@ -1182,7 +1182,7 @@ pmac_ide_macio_attach(struct macio_dev *mdev, const struct of_device_id *match)
 	regbase = (unsigned long) base;
 
 	pmif->mdev = mdev;
-	pmif->node = mdev->ofdev.node;
+	pmif->node = mdev->ofdev.dev.of_node;
 	pmif->regbase = regbase;
 	pmif->irq = irq;
 	pmif->kauai_fcr = NULL;
@@ -1191,7 +1191,7 @@ pmac_ide_macio_attach(struct macio_dev *mdev, const struct of_device_id *match)
 		if (macio_request_resource(mdev, 1, "ide-pmac (dma)"))
 			printk(KERN_WARNING "ide-pmac: can't request DMA "
 					    "resource for %s!\n",
-					    mdev->ofdev.node->full_name);
+					    mdev->ofdev.dev.of_node->full_name);
 		else
 			pmif->dma_regs = ioremap(macio_resource_start(mdev, 1), 0x1000);
 	} else
@@ -1400,8 +1400,11 @@ static struct of_device_id pmac_ide_macio_match[] =
 
 static struct macio_driver pmac_ide_macio_driver = 
 {
-	.name 		= "ide-pmac",
-	.match_table	= pmac_ide_macio_match,
+	.driver = {
+		.name 		= "ide-pmac",
+		.owner		= THIS_MODULE,
+		.of_match_table	= pmac_ide_macio_match,
+	},
 	.probe		= pmac_ide_macio_attach,
 	.suspend	= pmac_ide_macio_suspend,
 	.resume		= pmac_ide_macio_resume,
diff --git a/drivers/idle/Kconfig b/drivers/idle/Kconfig
index f15e90a453d1..fb5c5186d4aa 100644
--- a/drivers/idle/Kconfig
+++ b/drivers/idle/Kconfig
@@ -1,3 +1,14 @@
+config INTEL_IDLE
+	tristate "Cpuidle Driver for Intel Processors"
+	depends on CPU_IDLE
+	depends on X86
+	depends on CPU_SUP_INTEL
+	depends on EXPERIMENTAL
+	help
+	  Enable intel_idle, a cpuidle driver that includes knowledge of
+	  native Intel hardware idle features.  The acpi_idle driver
+	  can be configured at the same time, in order to handle
+	  processors intel_idle does not support.
 
 menu "Memory power savings"
 depends on X86_64
diff --git a/drivers/idle/Makefile b/drivers/idle/Makefile
index 5f68fc377e21..23d295cf10f2 100644
--- a/drivers/idle/Makefile
+++ b/drivers/idle/Makefile
@@ -1,2 +1,3 @@
 obj-$(CONFIG_I7300_IDLE)			+= i7300_idle.o
+obj-$(CONFIG_INTEL_IDLE)			+= intel_idle.o
 
diff --git a/drivers/idle/intel_idle.c b/drivers/idle/intel_idle.c
new file mode 100755
index 000000000000..54f0fb4cd5d2
--- /dev/null
+++ b/drivers/idle/intel_idle.c
@@ -0,0 +1,461 @@
+/*
+ * intel_idle.c - native hardware idle loop for modern Intel processors
+ *
+ * Copyright (c) 2010, Intel Corporation.
+ * Len Brown <len.brown@intel.com>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc.,
+ * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
+ */
+
+/*
+ * intel_idle is a cpuidle driver that loads on specific Intel processors
+ * in lieu of the legacy ACPI processor_idle driver.  The intent is to
+ * make Linux more efficient on these processors, as intel_idle knows
+ * more than ACPI, as well as make Linux more immune to ACPI BIOS bugs.
+ */
+
+/*
+ * Design Assumptions
+ *
+ * All CPUs have same idle states as boot CPU
+ *
+ * Chipset BM_STS (bus master status) bit is a NOP
+ *	for preventing entry into deep C-stats
+ */
+
+/*
+ * Known limitations
+ *
+ * The driver currently initializes for_each_online_cpu() upon modprobe.
+ * It it unaware of subsequent processors hot-added to the system.
+ * This means that if you boot with maxcpus=n and later online
+ * processors above n, those processors will use C1 only.
+ *
+ * ACPI has a .suspend hack to turn off deep c-statees during suspend
+ * to avoid complications with the lapic timer workaround.
+ * Have not seen issues with suspend, but may need same workaround here.
+ *
+ * There is currently no kernel-based automatic probing/loading mechanism
+ * if the driver is built as a module.
+ */
+
+/* un-comment DEBUG to enable pr_debug() statements */
+#define DEBUG
+
+#include <linux/kernel.h>
+#include <linux/cpuidle.h>
+#include <linux/clockchips.h>
+#include <linux/hrtimer.h>	/* ktime_get_real() */
+#include <trace/events/power.h>
+#include <linux/sched.h>
+
+#define INTEL_IDLE_VERSION "0.4"
+#define PREFIX "intel_idle: "
+
+#define MWAIT_SUBSTATE_MASK	(0xf)
+#define MWAIT_CSTATE_MASK	(0xf)
+#define MWAIT_SUBSTATE_SIZE	(4)
+#define MWAIT_MAX_NUM_CSTATES	8
+#define CPUID_MWAIT_LEAF (5)
+#define CPUID5_ECX_EXTENSIONS_SUPPORTED (0x1)
+#define CPUID5_ECX_INTERRUPT_BREAK	(0x2)
+
+static struct cpuidle_driver intel_idle_driver = {
+	.name = "intel_idle",
+	.owner = THIS_MODULE,
+};
+/* intel_idle.max_cstate=0 disables driver */
+static int max_cstate = MWAIT_MAX_NUM_CSTATES - 1;
+static int power_policy = 7; /* 0 = max perf; 15 = max powersave */
+
+static unsigned int substates;
+static int (*choose_substate)(int);
+
+/* Reliable LAPIC Timer States, bit 1 for C1 etc.  */
+static unsigned int lapic_timer_reliable_states;
+
+static struct cpuidle_device *intel_idle_cpuidle_devices;
+static int intel_idle(struct cpuidle_device *dev, struct cpuidle_state *state);
+
+static struct cpuidle_state *cpuidle_state_table;
+
+/*
+ * States are indexed by the cstate number,
+ * which is also the index into the MWAIT hint array.
+ * Thus C0 is a dummy.
+ */
+static struct cpuidle_state nehalem_cstates[MWAIT_MAX_NUM_CSTATES] = {
+	{ /* MWAIT C0 */ },
+	{ /* MWAIT C1 */
+		.name = "NHM-C1",
+		.desc = "MWAIT 0x00",
+		.driver_data = (void *) 0x00,
+		.flags = CPUIDLE_FLAG_TIME_VALID,
+		.exit_latency = 3,
+		.power_usage = 1000,
+		.target_residency = 6,
+		.enter = &intel_idle },
+	{ /* MWAIT C2 */
+		.name = "NHM-C3",
+		.desc = "MWAIT 0x10",
+		.driver_data = (void *) 0x10,
+		.flags = CPUIDLE_FLAG_TIME_VALID,
+		.exit_latency = 20,
+		.power_usage = 500,
+		.target_residency = 80,
+		.enter = &intel_idle },
+	{ /* MWAIT C3 */
+		.name = "NHM-C6",
+		.desc = "MWAIT 0x20",
+		.driver_data = (void *) 0x20,
+		.flags = CPUIDLE_FLAG_TIME_VALID,
+		.exit_latency = 200,
+		.power_usage = 350,
+		.target_residency = 800,
+		.enter = &intel_idle },
+};
+
+static struct cpuidle_state atom_cstates[MWAIT_MAX_NUM_CSTATES] = {
+	{ /* MWAIT C0 */ },
+	{ /* MWAIT C1 */
+		.name = "ATM-C1",
+		.desc = "MWAIT 0x00",
+		.driver_data = (void *) 0x00,
+		.flags = CPUIDLE_FLAG_TIME_VALID,
+		.exit_latency = 1,
+		.power_usage = 1000,
+		.target_residency = 4,
+		.enter = &intel_idle },
+	{ /* MWAIT C2 */
+		.name = "ATM-C2",
+		.desc = "MWAIT 0x10",
+		.driver_data = (void *) 0x10,
+		.flags = CPUIDLE_FLAG_TIME_VALID,
+		.exit_latency = 20,
+		.power_usage = 500,
+		.target_residency = 80,
+		.enter = &intel_idle },
+	{ /* MWAIT C3 */ },
+	{ /* MWAIT C4 */
+		.name = "ATM-C4",
+		.desc = "MWAIT 0x30",
+		.driver_data = (void *) 0x30,
+		.flags = CPUIDLE_FLAG_TIME_VALID,
+		.exit_latency = 100,
+		.power_usage = 250,
+		.target_residency = 400,
+		.enter = &intel_idle },
+	{ /* MWAIT C5 */ },
+	{ /* MWAIT C6 */
+		.name = "ATM-C6",
+		.desc = "MWAIT 0x40",
+		.driver_data = (void *) 0x40,
+		.flags = CPUIDLE_FLAG_TIME_VALID,
+		.exit_latency = 200,
+		.power_usage = 150,
+		.target_residency = 800,
+		.enter = NULL },	/* disabled */
+};
+
+/*
+ * choose_tunable_substate()
+ *
+ * Run-time decision on which C-state substate to invoke
+ * If power_policy = 0, choose shallowest substate (0)
+ * If power_policy = 15, choose deepest substate
+ * If power_policy = middle, choose middle substate etc.
+ */
+static int choose_tunable_substate(int cstate)
+{
+	unsigned int num_substates;
+	unsigned int substate_choice;
+
+	power_policy &= 0xF;	/* valid range: 0-15 */
+	cstate &= 7;	/* valid range: 0-7 */
+
+	num_substates = (substates >> ((cstate) * 4)) & MWAIT_SUBSTATE_MASK;
+
+	if (num_substates <= 1)
+		return 0;
+
+	substate_choice = ((power_policy + (power_policy + 1) *
+				(num_substates - 1)) / 16);
+
+	return substate_choice;
+}
+
+/*
+ * choose_zero_substate()
+ */
+static int choose_zero_substate(int cstate)
+{
+	return 0;
+}
+
+/**
+ * intel_idle
+ * @dev: cpuidle_device
+ * @state: cpuidle state
+ *
+ */
+static int intel_idle(struct cpuidle_device *dev, struct cpuidle_state *state)
+{
+	unsigned long ecx = 1; /* break on interrupt flag */
+	unsigned long eax = (unsigned long)cpuidle_get_statedata(state);
+	unsigned int cstate;
+	ktime_t kt_before, kt_after;
+	s64 usec_delta;
+	int cpu = smp_processor_id();
+
+	cstate = (((eax) >> MWAIT_SUBSTATE_SIZE) & MWAIT_CSTATE_MASK) + 1;
+
+	eax = eax + (choose_substate)(cstate);
+
+	local_irq_disable();
+
+	if (!(lapic_timer_reliable_states & (1 << (cstate))))
+		clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_ENTER, &cpu);
+
+	kt_before = ktime_get_real();
+
+	stop_critical_timings();
+#ifndef MODULE
+	trace_power_start(POWER_CSTATE, (eax >> 4) + 1);
+#endif
+	if (!need_resched()) {
+
+		__monitor((void *)&current_thread_info()->flags, 0, 0);
+		smp_mb();
+		if (!need_resched())
+			__mwait(eax, ecx);
+	}
+
+	start_critical_timings();
+
+	kt_after = ktime_get_real();
+	usec_delta = ktime_to_us(ktime_sub(kt_after, kt_before));
+
+	local_irq_enable();
+
+	if (!(lapic_timer_reliable_states & (1 << (cstate))))
+		clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_EXIT, &cpu);
+
+	return usec_delta;
+}
+
+/*
+ * intel_idle_probe()
+ */
+static int intel_idle_probe(void)
+{
+	unsigned int eax, ebx, ecx, edx;
+
+	if (max_cstate == 0) {
+		pr_debug(PREFIX "disabled\n");
+		return -EPERM;
+	}
+
+	if (boot_cpu_data.x86_vendor != X86_VENDOR_INTEL)
+		return -ENODEV;
+
+	if (!boot_cpu_has(X86_FEATURE_MWAIT))
+		return -ENODEV;
+
+	if (boot_cpu_data.cpuid_level < CPUID_MWAIT_LEAF)
+		return -ENODEV;
+
+	cpuid(CPUID_MWAIT_LEAF, &eax, &ebx, &ecx, &edx);
+
+	if (!(ecx & CPUID5_ECX_EXTENSIONS_SUPPORTED) ||
+		!(ecx & CPUID5_ECX_INTERRUPT_BREAK))
+			return -ENODEV;
+#ifdef DEBUG
+	if (substates == 0)	/* can over-ride via modparam */
+#endif
+		substates = edx;
+
+	pr_debug(PREFIX "MWAIT substates: 0x%x\n", substates);
+
+	if (boot_cpu_has(X86_FEATURE_ARAT))	/* Always Reliable APIC Timer */
+		lapic_timer_reliable_states = 0xFFFFFFFF;
+
+	if (boot_cpu_data.x86 != 6)	/* family 6 */
+		return -ENODEV;
+
+	switch (boot_cpu_data.x86_model) {
+
+	case 0x1A:	/* Core i7, Xeon 5500 series */
+	case 0x1E:	/* Core i7 and i5 Processor - Lynnfield Jasper Forest */
+	case 0x1F:	/* Core i7 and i5 Processor - Nehalem */
+	case 0x2E:	/* Nehalem-EX Xeon */
+		lapic_timer_reliable_states = (1 << 1);	 /* C1 */
+
+	case 0x25:	/* Westmere */
+	case 0x2C:	/* Westmere */
+		cpuidle_state_table = nehalem_cstates;
+		choose_substate = choose_tunable_substate;
+		break;
+
+	case 0x1C:	/* 28 - Atom Processor */
+		lapic_timer_reliable_states = (1 << 2) | (1 << 1); /* C2, C1 */
+		cpuidle_state_table = atom_cstates;
+		choose_substate = choose_zero_substate;
+		break;
+#ifdef FUTURE_USE
+	case 0x17:	/* 23 - Core 2 Duo */
+		lapic_timer_reliable_states = (1 << 2) | (1 << 1); /* C2, C1 */
+#endif
+
+	default:
+		pr_debug(PREFIX "does not run on family %d model %d\n",
+			boot_cpu_data.x86, boot_cpu_data.x86_model);
+		return -ENODEV;
+	}
+
+	pr_debug(PREFIX "v" INTEL_IDLE_VERSION
+		" model 0x%X\n", boot_cpu_data.x86_model);
+
+	pr_debug(PREFIX "lapic_timer_reliable_states 0x%x\n",
+		lapic_timer_reliable_states);
+	return 0;
+}
+
+/*
+ * intel_idle_cpuidle_devices_uninit()
+ * unregister, free cpuidle_devices
+ */
+static void intel_idle_cpuidle_devices_uninit(void)
+{
+	int i;
+	struct cpuidle_device *dev;
+
+	for_each_online_cpu(i) {
+		dev = per_cpu_ptr(intel_idle_cpuidle_devices, i);
+		cpuidle_unregister_device(dev);
+	}
+
+	free_percpu(intel_idle_cpuidle_devices);
+	return;
+}
+/*
+ * intel_idle_cpuidle_devices_init()
+ * allocate, initialize, register cpuidle_devices
+ */
+static int intel_idle_cpuidle_devices_init(void)
+{
+	int i, cstate;
+	struct cpuidle_device *dev;
+
+	intel_idle_cpuidle_devices = alloc_percpu(struct cpuidle_device);
+	if (intel_idle_cpuidle_devices == NULL)
+		return -ENOMEM;
+
+	for_each_online_cpu(i) {
+		dev = per_cpu_ptr(intel_idle_cpuidle_devices, i);
+
+		dev->state_count = 1;
+
+		for (cstate = 1; cstate < MWAIT_MAX_NUM_CSTATES; ++cstate) {
+			int num_substates;
+
+			if (cstate > max_cstate) {
+				printk(PREFIX "max_cstate %d reached\n",
+					max_cstate);
+				break;
+			}
+
+			/* does the state exist in CPUID.MWAIT? */
+			num_substates = (substates >> ((cstate) * 4))
+						& MWAIT_SUBSTATE_MASK;
+			if (num_substates == 0)
+				continue;
+			/* is the state not enabled? */
+			if (cpuidle_state_table[cstate].enter == NULL) {
+				/* does the driver not know about the state? */
+				if (*cpuidle_state_table[cstate].name == '\0')
+					pr_debug(PREFIX "unaware of model 0x%x"
+						" MWAIT %d please"
+						" contact lenb@kernel.org",
+					boot_cpu_data.x86_model, cstate);
+				continue;
+			}
+
+			if ((cstate > 2) &&
+				!boot_cpu_has(X86_FEATURE_NONSTOP_TSC))
+				mark_tsc_unstable("TSC halts in idle"
+					" states deeper than C2");
+
+			dev->states[dev->state_count] =	/* structure copy */
+				cpuidle_state_table[cstate];
+
+			dev->state_count += 1;
+		}
+
+		dev->cpu = i;
+		if (cpuidle_register_device(dev)) {
+			pr_debug(PREFIX "cpuidle_register_device %d failed!\n",
+				 i);
+			intel_idle_cpuidle_devices_uninit();
+			return -EIO;
+		}
+	}
+
+	return 0;
+}
+
+
+static int __init intel_idle_init(void)
+{
+	int retval;
+
+	retval = intel_idle_probe();
+	if (retval)
+		return retval;
+
+	retval = cpuidle_register_driver(&intel_idle_driver);
+	if (retval) {
+		printk(KERN_DEBUG PREFIX "intel_idle yielding to %s",
+			cpuidle_get_driver()->name);
+		return retval;
+	}
+
+	retval = intel_idle_cpuidle_devices_init();
+	if (retval) {
+		cpuidle_unregister_driver(&intel_idle_driver);
+		return retval;
+	}
+
+	return 0;
+}
+
+static void __exit intel_idle_exit(void)
+{
+	intel_idle_cpuidle_devices_uninit();
+	cpuidle_unregister_driver(&intel_idle_driver);
+
+	return;
+}
+
+module_init(intel_idle_init);
+module_exit(intel_idle_exit);
+
+module_param(power_policy, int, 0644);
+module_param(max_cstate, int, 0444);
+#ifdef DEBUG
+module_param(substates, int, 0444);
+#endif
+
+MODULE_AUTHOR("Len Brown <len.brown@intel.com>");
+MODULE_DESCRIPTION("Cpuidle driver for Intel Hardware v" INTEL_IDLE_VERSION);
+MODULE_LICENSE("GPL");
diff --git a/drivers/ieee1394/dv1394.c b/drivers/ieee1394/dv1394.c
index 9fd4a0d3206e..adaefabc40e9 100644
--- a/drivers/ieee1394/dv1394.c
+++ b/drivers/ieee1394/dv1394.c
@@ -1824,7 +1824,7 @@ static int dv1394_open(struct inode *inode, struct file *file)
 	       "and will not be available in the new firewire driver stack. "
 	       "Try libraw1394 based programs instead.\n", current->comm);
 
-	return 0;
+	return nonseekable_open(inode, file);
 }
 
 
@@ -2153,17 +2153,18 @@ static struct cdev dv1394_cdev;
 static const struct file_operations dv1394_fops=
 {
 	.owner =	THIS_MODULE,
-	.poll =         dv1394_poll,
+	.poll =		dv1394_poll,
 	.unlocked_ioctl = dv1394_ioctl,
 #ifdef CONFIG_COMPAT
 	.compat_ioctl = dv1394_compat_ioctl,
 #endif
 	.mmap =		dv1394_mmap,
 	.open =		dv1394_open,
-	.write =        dv1394_write,
-	.read =         dv1394_read,
+	.write =	dv1394_write,
+	.read =		dv1394_read,
 	.release =	dv1394_release,
-	.fasync =       dv1394_fasync,
+	.fasync =	dv1394_fasync,
+	.llseek =	no_llseek,
 };
 
 
diff --git a/drivers/ieee1394/raw1394.c b/drivers/ieee1394/raw1394.c
index 8aa56ac07e29..b563d5e9fa2e 100644
--- a/drivers/ieee1394/raw1394.c
+++ b/drivers/ieee1394/raw1394.c
@@ -2834,7 +2834,7 @@ static int raw1394_open(struct inode *inode, struct file *file)
 
 	file->private_data = fi;
 
-	return 0;
+	return nonseekable_open(inode, file);
 }
 
 static int raw1394_release(struct inode *inode, struct file *file)
@@ -3035,6 +3035,7 @@ static const struct file_operations raw1394_fops = {
 	.poll = raw1394_poll,
 	.open = raw1394_open,
 	.release = raw1394_release,
+	.llseek = no_llseek,
 };
 
 static int __init init_raw1394(void)
diff --git a/drivers/ieee1394/video1394.c b/drivers/ieee1394/video1394.c
index 949064a05675..a42bd6893bcf 100644
--- a/drivers/ieee1394/video1394.c
+++ b/drivers/ieee1394/video1394.c
@@ -1239,7 +1239,7 @@ static int video1394_open(struct inode *inode, struct file *file)
 	ctx->current_ctx = NULL;
 	file->private_data = ctx;
 
-	return 0;
+	return nonseekable_open(inode, file);
 }
 
 static int video1394_release(struct inode *inode, struct file *file)
@@ -1287,7 +1287,8 @@ static const struct file_operations video1394_fops=
 	.poll =		video1394_poll,
 	.mmap =		video1394_mmap,
 	.open =		video1394_open,
-	.release =	video1394_release
+	.release =	video1394_release,
+	.llseek =	no_llseek,
 };
 
 /*** HOTPLUG STUFF **********************************************************/
diff --git a/drivers/infiniband/Kconfig b/drivers/infiniband/Kconfig
index 330d2a423362..89d70de5e235 100644
--- a/drivers/infiniband/Kconfig
+++ b/drivers/infiniband/Kconfig
@@ -43,6 +43,7 @@ config INFINIBAND_ADDR_TRANS
 
 source "drivers/infiniband/hw/mthca/Kconfig"
 source "drivers/infiniband/hw/ipath/Kconfig"
+source "drivers/infiniband/hw/qib/Kconfig"
 source "drivers/infiniband/hw/ehca/Kconfig"
 source "drivers/infiniband/hw/amso1100/Kconfig"
 source "drivers/infiniband/hw/cxgb3/Kconfig"
diff --git a/drivers/infiniband/Makefile b/drivers/infiniband/Makefile
index 0c4e589d746e..9cc7a47d3e67 100644
--- a/drivers/infiniband/Makefile
+++ b/drivers/infiniband/Makefile
@@ -1,6 +1,7 @@
 obj-$(CONFIG_INFINIBAND)		+= core/
 obj-$(CONFIG_INFINIBAND_MTHCA)		+= hw/mthca/
 obj-$(CONFIG_INFINIBAND_IPATH)		+= hw/ipath/
+obj-$(CONFIG_INFINIBAND_QIB)		+= hw/qib/
 obj-$(CONFIG_INFINIBAND_EHCA)		+= hw/ehca/
 obj-$(CONFIG_INFINIBAND_AMSO1100)	+= hw/amso1100/
 obj-$(CONFIG_INFINIBAND_CXGB3)		+= hw/cxgb3/
diff --git a/drivers/infiniband/core/core_priv.h b/drivers/infiniband/core/core_priv.h
index 05ac36e6acdb..a565af5c2d2e 100644
--- a/drivers/infiniband/core/core_priv.h
+++ b/drivers/infiniband/core/core_priv.h
@@ -38,7 +38,9 @@
 
 #include <rdma/ib_verbs.h>
 
-int  ib_device_register_sysfs(struct ib_device *device);
+int  ib_device_register_sysfs(struct ib_device *device,
+			      int (*port_callback)(struct ib_device *,
+						   u8, struct kobject *));
 void ib_device_unregister_sysfs(struct ib_device *device);
 
 int  ib_sysfs_setup(void);
diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c
index d1fba4153332..a19effad0811 100644
--- a/drivers/infiniband/core/device.c
+++ b/drivers/infiniband/core/device.c
@@ -267,7 +267,9 @@ out:
  * callback for each device that is added. @device must be allocated
  * with ib_alloc_device().
  */
-int ib_register_device(struct ib_device *device)
+int ib_register_device(struct ib_device *device,
+		       int (*port_callback)(struct ib_device *,
+					    u8, struct kobject *))
 {
 	int ret;
 
@@ -296,7 +298,7 @@ int ib_register_device(struct ib_device *device)
 		goto out;
 	}
 
-	ret = ib_device_register_sysfs(device);
+	ret = ib_device_register_sysfs(device, port_callback);
 	if (ret) {
 		printk(KERN_WARNING "Couldn't register device %s with driver model\n",
 		       device->name);
diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c
index 6dc7b77d5d29..ef1304f151dc 100644
--- a/drivers/infiniband/core/mad.c
+++ b/drivers/infiniband/core/mad.c
@@ -47,8 +47,8 @@ MODULE_DESCRIPTION("kernel IB MAD API");
 MODULE_AUTHOR("Hal Rosenstock");
 MODULE_AUTHOR("Sean Hefty");
 
-int mad_sendq_size = IB_MAD_QP_SEND_SIZE;
-int mad_recvq_size = IB_MAD_QP_RECV_SIZE;
+static int mad_sendq_size = IB_MAD_QP_SEND_SIZE;
+static int mad_recvq_size = IB_MAD_QP_RECV_SIZE;
 
 module_param_named(send_queue_size, mad_sendq_size, int, 0444);
 MODULE_PARM_DESC(send_queue_size, "Size of send queue in number of work requests");
diff --git a/drivers/infiniband/core/sysfs.c b/drivers/infiniband/core/sysfs.c
index f901957abc8b..3627300e2a10 100644
--- a/drivers/infiniband/core/sysfs.c
+++ b/drivers/infiniband/core/sysfs.c
@@ -475,7 +475,9 @@ err:
 	return NULL;
 }
 
-static int add_port(struct ib_device *device, int port_num)
+static int add_port(struct ib_device *device, int port_num,
+		    int (*port_callback)(struct ib_device *,
+					 u8, struct kobject *))
 {
 	struct ib_port *p;
 	struct ib_port_attr attr;
@@ -522,11 +524,20 @@ static int add_port(struct ib_device *device, int port_num)
 	if (ret)
 		goto err_free_pkey;
 
+	if (port_callback) {
+		ret = port_callback(device, port_num, &p->kobj);
+		if (ret)
+			goto err_remove_pkey;
+	}
+
 	list_add_tail(&p->kobj.entry, &device->port_list);
 
 	kobject_uevent(&p->kobj, KOBJ_ADD);
 	return 0;
 
+err_remove_pkey:
+	sysfs_remove_group(&p->kobj, &p->pkey_group);
+
 err_free_pkey:
 	for (i = 0; i < attr.pkey_tbl_len; ++i)
 		kfree(p->pkey_group.attrs[i]);
@@ -754,7 +765,9 @@ static struct attribute_group iw_stats_group = {
 	.attrs	= iw_proto_stats_attrs,
 };
 
-int ib_device_register_sysfs(struct ib_device *device)
+int ib_device_register_sysfs(struct ib_device *device,
+			     int (*port_callback)(struct ib_device *,
+						  u8, struct kobject *))
 {
 	struct device *class_dev = &device->dev;
 	int ret;
@@ -785,12 +798,12 @@ int ib_device_register_sysfs(struct ib_device *device)
 	}
 
 	if (device->node_type == RDMA_NODE_IB_SWITCH) {
-		ret = add_port(device, 0);
+		ret = add_port(device, 0, port_callback);
 		if (ret)
 			goto err_put;
 	} else {
 		for (i = 1; i <= device->phys_port_cnt; ++i) {
-			ret = add_port(device, i);
+			ret = add_port(device, i, port_callback);
 			if (ret)
 				goto err_put;
 		}
diff --git a/drivers/infiniband/core/ucm.c b/drivers/infiniband/core/ucm.c
index 46474842cfe9..08f948df8fa9 100644
--- a/drivers/infiniband/core/ucm.c
+++ b/drivers/infiniband/core/ucm.c
@@ -706,14 +706,9 @@ static int ib_ucm_alloc_data(const void **dest, u64 src, u32 len)
 	if (!len)
 		return 0;
 
-	data = kmalloc(len, GFP_KERNEL);
-	if (!data)
-		return -ENOMEM;
-
-	if (copy_from_user(data, (void __user *)(unsigned long)src, len)) {
-		kfree(data);
-		return -EFAULT;
-	}
+	data = memdup_user((void __user *)(unsigned long)src, len);
+	if (IS_ERR(data))
+		return PTR_ERR(data);
 
 	*dest = data;
 	return 0;
diff --git a/drivers/infiniband/hw/amso1100/c2_provider.c b/drivers/infiniband/hw/amso1100/c2_provider.c
index c47f618d12e8..aeebc4d37e33 100644
--- a/drivers/infiniband/hw/amso1100/c2_provider.c
+++ b/drivers/infiniband/hw/amso1100/c2_provider.c
@@ -865,7 +865,7 @@ int c2_register_device(struct c2_dev *dev)
 	dev->ibdev.iwcm->create_listen = c2_service_create;
 	dev->ibdev.iwcm->destroy_listen = c2_service_destroy;
 
-	ret = ib_register_device(&dev->ibdev);
+	ret = ib_register_device(&dev->ibdev, NULL);
 	if (ret)
 		goto out_free_iwcm;
 
diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c
index 19b1c4a62a23..fca0b4b747e4 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c
@@ -1428,7 +1428,7 @@ int iwch_register_device(struct iwch_dev *dev)
 	dev->ibdev.iwcm->rem_ref = iwch_qp_rem_ref;
 	dev->ibdev.iwcm->get_qp = iwch_get_qp;
 
-	ret = ib_register_device(&dev->ibdev);
+	ret = ib_register_device(&dev->ibdev, NULL);
 	if (ret)
 		goto bail1;
 
diff --git a/drivers/infiniband/hw/cxgb4/cq.c b/drivers/infiniband/hw/cxgb4/cq.c
index fb1aafcc294f..2447f5295482 100644
--- a/drivers/infiniband/hw/cxgb4/cq.c
+++ b/drivers/infiniband/hw/cxgb4/cq.c
@@ -373,6 +373,7 @@ static void create_read_req_cqe(struct t4_wq *wq, struct t4_cqe *hw_cqe,
 				 V_CQE_SWCQE(SW_CQE(hw_cqe)) |
 				 V_CQE_OPCODE(FW_RI_READ_REQ) |
 				 V_CQE_TYPE(1));
+	read_cqe->bits_type_ts = hw_cqe->bits_type_ts;
 }
 
 /*
@@ -780,6 +781,9 @@ struct ib_cq *c4iw_create_cq(struct ib_device *ibdev, int entries,
 	/* account for the status page. */
 	entries++;
 
+	/* IQ needs one extra entry to differentiate full vs empty. */
+	entries++;
+
 	/*
 	 * entries must be multiple of 16 for HW.
 	 */
@@ -801,7 +805,7 @@ struct ib_cq *c4iw_create_cq(struct ib_device *ibdev, int entries,
 
 	chp->rhp = rhp;
 	chp->cq.size--;				/* status page */
-	chp->ibcq.cqe = chp->cq.size;
+	chp->ibcq.cqe = chp->cq.size - 1;
 	spin_lock_init(&chp->lock);
 	atomic_set(&chp->refcnt, 1);
 	init_waitqueue_head(&chp->wait);
diff --git a/drivers/infiniband/hw/cxgb4/device.c b/drivers/infiniband/hw/cxgb4/device.c
index be23b5eab13b..d870f9c17c1e 100644
--- a/drivers/infiniband/hw/cxgb4/device.c
+++ b/drivers/infiniband/hw/cxgb4/device.c
@@ -306,7 +306,8 @@ static void c4iw_remove(struct c4iw_dev *dev)
 	PDBG("%s c4iw_dev %p\n", __func__,  dev);
 	cancel_delayed_work_sync(&dev->db_drop_task);
 	list_del(&dev->entry);
-	c4iw_unregister_device(dev);
+	if (dev->registered)
+		c4iw_unregister_device(dev);
 	c4iw_rdev_close(&dev->rdev);
 	idr_destroy(&dev->cqidr);
 	idr_destroy(&dev->qpidr);
@@ -343,12 +344,6 @@ static struct c4iw_dev *c4iw_alloc(const struct cxgb4_lld_info *infop)
 	list_add_tail(&devp->entry, &dev_list);
 	mutex_unlock(&dev_mutex);
 
-	if (c4iw_register_device(devp)) {
-		printk(KERN_ERR MOD "Unable to register device\n");
-		mutex_lock(&dev_mutex);
-		c4iw_remove(devp);
-		mutex_unlock(&dev_mutex);
-	}
 	if (c4iw_debugfs_root) {
 		devp->debugfs_root = debugfs_create_dir(
 					pci_name(devp->rdev.lldi.pdev),
@@ -379,9 +374,6 @@ static void *c4iw_uld_add(const struct cxgb4_lld_info *infop)
 
 	for (i = 0; i < dev->rdev.lldi.nrxq; i++)
 		PDBG("rxqid[%u] %u\n", i, dev->rdev.lldi.rxq_ids[i]);
-
-	printk(KERN_INFO MOD "Initialized device %s\n",
-	       pci_name(dev->rdev.lldi.pdev));
 out:
 	return dev;
 }
@@ -471,7 +463,41 @@ nomem:
 
 static int c4iw_uld_state_change(void *handle, enum cxgb4_state new_state)
 {
+	struct c4iw_dev *dev = handle;
+
 	PDBG("%s new_state %u\n", __func__, new_state);
+	switch (new_state) {
+	case CXGB4_STATE_UP:
+		printk(KERN_INFO MOD "%s: Up\n", pci_name(dev->rdev.lldi.pdev));
+		if (!dev->registered) {
+			int ret;
+			ret = c4iw_register_device(dev);
+			if (ret)
+				printk(KERN_ERR MOD
+				       "%s: RDMA registration failed: %d\n",
+				       pci_name(dev->rdev.lldi.pdev), ret);
+		}
+		break;
+	case CXGB4_STATE_DOWN:
+		printk(KERN_INFO MOD "%s: Down\n",
+		       pci_name(dev->rdev.lldi.pdev));
+		if (dev->registered)
+			c4iw_unregister_device(dev);
+		break;
+	case CXGB4_STATE_START_RECOVERY:
+		printk(KERN_INFO MOD "%s: Fatal Error\n",
+		       pci_name(dev->rdev.lldi.pdev));
+		if (dev->registered)
+			c4iw_unregister_device(dev);
+		break;
+	case CXGB4_STATE_DETACH:
+		printk(KERN_INFO MOD "%s: Detach\n",
+		       pci_name(dev->rdev.lldi.pdev));
+		mutex_lock(&dev_mutex);
+		c4iw_remove(dev);
+		mutex_unlock(&dev_mutex);
+		break;
+	}
 	return 0;
 }
 
@@ -504,14 +530,12 @@ static void __exit c4iw_exit_module(void)
 {
 	struct c4iw_dev *dev, *tmp;
 
-	cxgb4_unregister_uld(CXGB4_ULD_RDMA);
-
 	mutex_lock(&dev_mutex);
 	list_for_each_entry_safe(dev, tmp, &dev_list, entry) {
 		c4iw_remove(dev);
 	}
 	mutex_unlock(&dev_mutex);
-
+	cxgb4_unregister_uld(CXGB4_ULD_RDMA);
 	c4iw_cm_term();
 	debugfs_remove_recursive(c4iw_debugfs_root);
 }
diff --git a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
index a6269981e815..277ab589b44d 100644
--- a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
+++ b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
@@ -152,6 +152,7 @@ struct c4iw_dev {
 	struct list_head entry;
 	struct delayed_work db_drop_task;
 	struct dentry *debugfs_root;
+	u8 registered;
 };
 
 static inline struct c4iw_dev *to_c4iw_dev(struct ib_device *ibdev)
diff --git a/drivers/infiniband/hw/cxgb4/mem.c b/drivers/infiniband/hw/cxgb4/mem.c
index e54ff6d25691..7f94da1a2437 100644
--- a/drivers/infiniband/hw/cxgb4/mem.c
+++ b/drivers/infiniband/hw/cxgb4/mem.c
@@ -712,8 +712,10 @@ struct ib_mr *c4iw_alloc_fast_reg_mr(struct ib_pd *pd, int pbl_depth)
 	php = to_c4iw_pd(pd);
 	rhp = php->rhp;
 	mhp = kzalloc(sizeof(*mhp), GFP_KERNEL);
-	if (!mhp)
+	if (!mhp) {
+		ret = -ENOMEM;
 		goto err;
+	}
 
 	mhp->rhp = rhp;
 	ret = alloc_pbl(mhp, pbl_depth);
@@ -730,8 +732,10 @@ struct ib_mr *c4iw_alloc_fast_reg_mr(struct ib_pd *pd, int pbl_depth)
 	mhp->attr.state = 1;
 	mmid = (stag) >> 8;
 	mhp->ibmr.rkey = mhp->ibmr.lkey = stag;
-	if (insert_handle(rhp, &rhp->mmidr, mhp, mmid))
+	if (insert_handle(rhp, &rhp->mmidr, mhp, mmid)) {
+		ret = -ENOMEM;
 		goto err3;
+	}
 
 	PDBG("%s mmid 0x%x mhp %p stag 0x%x\n", __func__, mmid, mhp, stag);
 	return &(mhp->ibmr);
@@ -755,9 +759,6 @@ struct ib_fast_reg_page_list *c4iw_alloc_fastreg_pbl(struct ib_device *device,
 	dma_addr_t dma_addr;
 	int size = sizeof *c4pl + page_list_len * sizeof(u64);
 
-	if (page_list_len > T4_MAX_FR_DEPTH)
-		return ERR_PTR(-EINVAL);
-
 	c4pl = dma_alloc_coherent(&dev->rdev.lldi.pdev->dev, size,
 				  &dma_addr, GFP_KERNEL);
 	if (!c4pl)
diff --git a/drivers/infiniband/hw/cxgb4/provider.c b/drivers/infiniband/hw/cxgb4/provider.c
index dfc49020bb9c..8f645c83a125 100644
--- a/drivers/infiniband/hw/cxgb4/provider.c
+++ b/drivers/infiniband/hw/cxgb4/provider.c
@@ -486,7 +486,7 @@ int c4iw_register_device(struct c4iw_dev *dev)
 	dev->ibdev.iwcm->rem_ref = c4iw_qp_rem_ref;
 	dev->ibdev.iwcm->get_qp = c4iw_get_qp;
 
-	ret = ib_register_device(&dev->ibdev);
+	ret = ib_register_device(&dev->ibdev, NULL);
 	if (ret)
 		goto bail1;
 
@@ -496,6 +496,7 @@ int c4iw_register_device(struct c4iw_dev *dev)
 		if (ret)
 			goto bail2;
 	}
+	dev->registered = 1;
 	return 0;
 bail2:
 	ib_unregister_device(&dev->ibdev);
@@ -514,5 +515,6 @@ void c4iw_unregister_device(struct c4iw_dev *dev)
 				   c4iw_class_attributes[i]);
 	ib_unregister_device(&dev->ibdev);
 	kfree(dev->ibdev.iwcm);
+	dev->registered = 0;
 	return;
 }
diff --git a/drivers/infiniband/hw/cxgb4/qp.c b/drivers/infiniband/hw/cxgb4/qp.c
index b321835dcf6e..646a2a5711f2 100644
--- a/drivers/infiniband/hw/cxgb4/qp.c
+++ b/drivers/infiniband/hw/cxgb4/qp.c
@@ -572,9 +572,13 @@ int c4iw_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 			err = build_rdma_write(wqe, wr, &len16);
 			break;
 		case IB_WR_RDMA_READ:
+		case IB_WR_RDMA_READ_WITH_INV:
 			fw_opcode = FW_RI_RDMA_READ_WR;
 			swsqe->opcode = FW_RI_READ_REQ;
-			fw_flags = 0;
+			if (wr->opcode == IB_WR_RDMA_READ_WITH_INV)
+				fw_flags |= FW_RI_RDMA_READ_INVALIDATE;
+			else
+				fw_flags = 0;
 			err = build_rdma_read(wqe, wr, &len16);
 			if (err)
 				break;
@@ -588,6 +592,8 @@ int c4iw_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 			err = build_fastreg(wqe, wr, &len16);
 			break;
 		case IB_WR_LOCAL_INV:
+			if (wr->send_flags & IB_SEND_FENCE)
+				fw_flags |= FW_RI_LOCAL_FENCE_FLAG;
 			fw_opcode = FW_RI_INV_LSTAG_WR;
 			swsqe->opcode = FW_RI_LOCAL_INV;
 			err = build_inv_stag(wqe, wr, &len16);
@@ -1339,7 +1345,6 @@ int c4iw_destroy_qp(struct ib_qp *ib_qp)
 	wait_event(qhp->wait, !qhp->ep);
 
 	remove_handle(rhp, &rhp->qpidr, qhp->wq.sq.qid);
-	remove_handle(rhp, &rhp->qpidr, qhp->wq.rq.qid);
 	atomic_dec(&qhp->refcnt);
 	wait_event(qhp->wait, !atomic_read(&qhp->refcnt));
 
@@ -1442,30 +1447,26 @@ struct ib_qp *c4iw_create_qp(struct ib_pd *pd, struct ib_qp_init_attr *attrs,
 	if (ret)
 		goto err2;
 
-	ret = insert_handle(rhp, &rhp->qpidr, qhp, qhp->wq.rq.qid);
-	if (ret)
-		goto err3;
-
 	if (udata) {
 		mm1 = kmalloc(sizeof *mm1, GFP_KERNEL);
 		if (!mm1) {
 			ret = -ENOMEM;
-			goto err4;
+			goto err3;
 		}
 		mm2 = kmalloc(sizeof *mm2, GFP_KERNEL);
 		if (!mm2) {
 			ret = -ENOMEM;
-			goto err5;
+			goto err4;
 		}
 		mm3 = kmalloc(sizeof *mm3, GFP_KERNEL);
 		if (!mm3) {
 			ret = -ENOMEM;
-			goto err6;
+			goto err5;
 		}
 		mm4 = kmalloc(sizeof *mm4, GFP_KERNEL);
 		if (!mm4) {
 			ret = -ENOMEM;
-			goto err7;
+			goto err6;
 		}
 
 		uresp.qid_mask = rhp->rdev.qpmask;
@@ -1487,7 +1488,7 @@ struct ib_qp *c4iw_create_qp(struct ib_pd *pd, struct ib_qp_init_attr *attrs,
 		spin_unlock(&ucontext->mmap_lock);
 		ret = ib_copy_to_udata(udata, &uresp, sizeof uresp);
 		if (ret)
-			goto err8;
+			goto err7;
 		mm1->key = uresp.sq_key;
 		mm1->addr = virt_to_phys(qhp->wq.sq.queue);
 		mm1->len = PAGE_ALIGN(qhp->wq.sq.memsize);
@@ -1511,16 +1512,14 @@ struct ib_qp *c4iw_create_qp(struct ib_pd *pd, struct ib_qp_init_attr *attrs,
 	     __func__, qhp, qhp->attr.sq_num_entries, qhp->attr.rq_num_entries,
 	     qhp->wq.sq.qid);
 	return &qhp->ibqp;
-err8:
-	kfree(mm4);
 err7:
-	kfree(mm3);
+	kfree(mm4);
 err6:
-	kfree(mm2);
+	kfree(mm3);
 err5:
-	kfree(mm1);
+	kfree(mm2);
 err4:
-	remove_handle(rhp, &rhp->qpidr, qhp->wq.rq.qid);
+	kfree(mm1);
 err3:
 	remove_handle(rhp, &rhp->qpidr, qhp->wq.sq.qid);
 err2:
diff --git a/drivers/infiniband/hw/cxgb4/t4.h b/drivers/infiniband/hw/cxgb4/t4.h
index d0e8af352408..1057cb96302e 100644
--- a/drivers/infiniband/hw/cxgb4/t4.h
+++ b/drivers/infiniband/hw/cxgb4/t4.h
@@ -41,11 +41,13 @@
 #define T4_MAX_NUM_QP (1<<16)
 #define T4_MAX_NUM_CQ (1<<15)
 #define T4_MAX_NUM_PD (1<<15)
-#define T4_MAX_PBL_SIZE 256
-#define T4_MAX_RQ_SIZE 1024
-#define T4_MAX_SQ_SIZE 1024
-#define T4_MAX_QP_DEPTH (T4_MAX_RQ_SIZE-1)
-#define T4_MAX_CQ_DEPTH 8192
+#define T4_EQ_STATUS_ENTRIES (L1_CACHE_BYTES > 64 ? 2 : 1)
+#define T4_MAX_EQ_SIZE (65520 - T4_EQ_STATUS_ENTRIES)
+#define T4_MAX_IQ_SIZE (65520 - 1)
+#define T4_MAX_RQ_SIZE (8192 - T4_EQ_STATUS_ENTRIES)
+#define T4_MAX_SQ_SIZE (T4_MAX_EQ_SIZE - 1)
+#define T4_MAX_QP_DEPTH (T4_MAX_RQ_SIZE - 1)
+#define T4_MAX_CQ_DEPTH (T4_MAX_IQ_SIZE - 1)
 #define T4_MAX_NUM_STAG (1<<15)
 #define T4_MAX_MR_SIZE (~0ULL - 1)
 #define T4_PAGESIZE_MASK 0xffff000  /* 4KB-128MB */
@@ -79,12 +81,11 @@ struct t4_status_page {
 			sizeof(struct fw_ri_isgl)) / sizeof(struct fw_ri_sge))
 #define T4_MAX_FR_IMMD ((T4_SQ_NUM_BYTES - sizeof(struct fw_ri_fr_nsmr_wr) - \
 			sizeof(struct fw_ri_immd)))
-#define T4_MAX_FR_DEPTH 255
+#define T4_MAX_FR_DEPTH (T4_MAX_FR_IMMD / sizeof(u64))
 
 #define T4_RQ_NUM_SLOTS 2
 #define T4_RQ_NUM_BYTES (T4_EQ_SIZE * T4_RQ_NUM_SLOTS)
-#define T4_MAX_RECV_SGE ((T4_RQ_NUM_BYTES - sizeof(struct fw_ri_recv_wr) - \
-			sizeof(struct fw_ri_isgl)) / sizeof(struct fw_ri_sge))
+#define T4_MAX_RECV_SGE 4
 
 union t4_wr {
 	struct fw_ri_res_wr res;
@@ -434,7 +435,7 @@ struct t4_cq {
 	struct c4iw_rdev *rdev;
 	u64 ugts;
 	size_t memsize;
-	u64 timestamp;
+	__be64 bits_type_ts;
 	u32 cqid;
 	u16 size; /* including status page */
 	u16 cidx;
@@ -449,25 +450,17 @@ struct t4_cq {
 static inline int t4_arm_cq(struct t4_cq *cq, int se)
 {
 	u32 val;
-	u16 inc;
-
-	do {
-		/*
-		 * inc must be less the both the max update value -and-
-		 * the size of the CQ.
-		 */
-		inc = cq->cidx_inc <= CIDXINC_MASK ? cq->cidx_inc :
-						     CIDXINC_MASK;
-		inc = inc <= (cq->size - 1) ? inc : (cq->size - 1);
-		if (inc == cq->cidx_inc)
-			val = SEINTARM(se) | CIDXINC(inc) | TIMERREG(6) |
-			      INGRESSQID(cq->cqid);
-		else
-			val = SEINTARM(0) | CIDXINC(inc) | TIMERREG(7) |
-			      INGRESSQID(cq->cqid);
-		cq->cidx_inc -= inc;
+
+	while (cq->cidx_inc > CIDXINC_MASK) {
+		val = SEINTARM(0) | CIDXINC(CIDXINC_MASK) | TIMERREG(7) |
+		      INGRESSQID(cq->cqid);
 		writel(val, cq->gts);
-	} while (cq->cidx_inc);
+		cq->cidx_inc -= CIDXINC_MASK;
+	}
+	val = SEINTARM(se) | CIDXINC(cq->cidx_inc) | TIMERREG(6) |
+	      INGRESSQID(cq->cqid);
+	writel(val, cq->gts);
+	cq->cidx_inc = 0;
 	return 0;
 }
 
@@ -487,7 +480,9 @@ static inline void t4_swcq_consume(struct t4_cq *cq)
 
 static inline void t4_hwcq_consume(struct t4_cq *cq)
 {
-	cq->cidx_inc++;
+	cq->bits_type_ts = cq->queue[cq->cidx].bits_type_ts;
+	if (++cq->cidx_inc == cq->size)
+		cq->cidx_inc = 0;
 	if (++cq->cidx == cq->size) {
 		cq->cidx = 0;
 		cq->gen ^= 1;
@@ -501,20 +496,23 @@ static inline int t4_valid_cqe(struct t4_cq *cq, struct t4_cqe *cqe)
 
 static inline int t4_next_hw_cqe(struct t4_cq *cq, struct t4_cqe **cqe)
 {
-	int ret = 0;
-	u64 bits_type_ts = be64_to_cpu(cq->queue[cq->cidx].bits_type_ts);
+	int ret;
+	u16 prev_cidx;
 
-	if (G_CQE_GENBIT(bits_type_ts) == cq->gen) {
-		*cqe = &cq->queue[cq->cidx];
-		cq->timestamp = G_CQE_TS(bits_type_ts);
-	} else if (G_CQE_TS(bits_type_ts) > cq->timestamp)
-		ret = -EOVERFLOW;
+	if (cq->cidx == 0)
+		prev_cidx = cq->size - 1;
 	else
-		ret = -ENODATA;
-	if (ret == -EOVERFLOW) {
-		printk(KERN_ERR MOD "cq overflow cqid %u\n", cq->cqid);
+		prev_cidx = cq->cidx - 1;
+
+	if (cq->queue[prev_cidx].bits_type_ts != cq->bits_type_ts) {
+		ret = -EOVERFLOW;
 		cq->error = 1;
-	}
+		printk(KERN_ERR MOD "cq overflow cqid %u\n", cq->cqid);
+	} else if (t4_valid_cqe(cq, &cq->queue[cq->cidx])) {
+		*cqe = &cq->queue[cq->cidx];
+		ret = 0;
+	} else
+		ret = -ENODATA;
 	return ret;
 }
 
diff --git a/drivers/infiniband/hw/ehca/ehca_irq.c b/drivers/infiniband/hw/ehca/ehca_irq.c
index 07cae552cafb..e571e60ecb88 100644
--- a/drivers/infiniband/hw/ehca/ehca_irq.c
+++ b/drivers/infiniband/hw/ehca/ehca_irq.c
@@ -847,7 +847,7 @@ static int __cpuinit comp_pool_callback(struct notifier_block *nfb,
 		ehca_gen_dbg("CPU: %x (CPU_PREPARE)", cpu);
 		if (!create_comp_task(pool, cpu)) {
 			ehca_gen_err("Can't create comp_task for cpu: %x", cpu);
-			return NOTIFY_BAD;
+			return notifier_from_errno(-ENOMEM);
 		}
 		break;
 	case CPU_UP_CANCELED:
diff --git a/drivers/infiniband/hw/ehca/ehca_main.c b/drivers/infiniband/hw/ehca/ehca_main.c
index 129a6bebd6e3..ecb51b396c42 100644
--- a/drivers/infiniband/hw/ehca/ehca_main.c
+++ b/drivers/infiniband/hw/ehca/ehca_main.c
@@ -291,8 +291,9 @@ static int ehca_sense_attributes(struct ehca_shca *shca)
 	};
 
 	ehca_gen_dbg("Probing adapter %s...",
-		     shca->ofdev->node->full_name);
-	loc_code = of_get_property(shca->ofdev->node, "ibm,loc-code", NULL);
+		     shca->ofdev->dev.of_node->full_name);
+	loc_code = of_get_property(shca->ofdev->dev.of_node, "ibm,loc-code",
+				   NULL);
 	if (loc_code)
 		ehca_gen_dbg(" ... location lode=%s", loc_code);
 
@@ -720,16 +721,16 @@ static int __devinit ehca_probe(struct of_device *dev,
 	int ret, i, eq_size;
 	unsigned long flags;
 
-	handle = of_get_property(dev->node, "ibm,hca-handle", NULL);
+	handle = of_get_property(dev->dev.of_node, "ibm,hca-handle", NULL);
 	if (!handle) {
 		ehca_gen_err("Cannot get eHCA handle for adapter: %s.",
-			     dev->node->full_name);
+			     dev->dev.of_node->full_name);
 		return -ENODEV;
 	}
 
 	if (!(*handle)) {
 		ehca_gen_err("Wrong eHCA handle for adapter: %s.",
-			     dev->node->full_name);
+			     dev->dev.of_node->full_name);
 		return -ENODEV;
 	}
 
@@ -798,7 +799,7 @@ static int __devinit ehca_probe(struct of_device *dev,
 		goto probe5;
 	}
 
-	ret = ib_register_device(&shca->ib_device);
+	ret = ib_register_device(&shca->ib_device, NULL);
 	if (ret) {
 		ehca_err(&shca->ib_device,
 			 "ib_register_device() failed ret=%i", ret);
@@ -936,12 +937,13 @@ static struct of_device_id ehca_device_table[] =
 MODULE_DEVICE_TABLE(of, ehca_device_table);
 
 static struct of_platform_driver ehca_driver = {
-	.name        = "ehca",
-	.match_table = ehca_device_table,
 	.probe       = ehca_probe,
 	.remove      = ehca_remove,
-	.driver	     = {
+	.driver = {
+		.name = "ehca",
+		.owner = THIS_MODULE,
 		.groups = ehca_drv_attr_groups,
+		.of_match_table = ehca_device_table,
 	},
 };
 
diff --git a/drivers/infiniband/hw/ipath/Kconfig b/drivers/infiniband/hw/ipath/Kconfig
index 3c7968f25ec2..1d9bb115cbf6 100644
--- a/drivers/infiniband/hw/ipath/Kconfig
+++ b/drivers/infiniband/hw/ipath/Kconfig
@@ -1,9 +1,11 @@
 config INFINIBAND_IPATH
-	tristate "QLogic InfiniPath Driver"
-	depends on 64BIT && NET
+	tristate "QLogic HTX HCA support"
+	depends on 64BIT && NET && HT_IRQ
 	---help---
-	This is a driver for QLogic InfiniPath host channel adapters,
+	This is a driver for the obsolete QLogic Hyper-Transport
+	IB host channel adapter (model QHT7140),
 	including InfiniBand verbs support.  This driver allows these
 	devices to be used with both kernel upper level protocols such
 	as IP-over-InfiniBand as well as with userspace applications
 	(in conjunction with InfiniBand userspace access).
+	For QLogic PCIe QLE based cards, use the QIB driver instead.
diff --git a/drivers/infiniband/hw/ipath/Makefile b/drivers/infiniband/hw/ipath/Makefile
index bf9450061986..fa3df82681df 100644
--- a/drivers/infiniband/hw/ipath/Makefile
+++ b/drivers/infiniband/hw/ipath/Makefile
@@ -29,13 +29,9 @@ ib_ipath-y := \
 	ipath_user_pages.o \
 	ipath_user_sdma.o \
 	ipath_verbs_mcast.o \
-	ipath_verbs.o \
-	ipath_iba7220.o \
-	ipath_sd7220.o \
-	ipath_sd7220_img.o
+	ipath_verbs.o
 
 ib_ipath-$(CONFIG_HT_IRQ) += ipath_iba6110.o
-ib_ipath-$(CONFIG_PCI_MSI) += ipath_iba6120.o
 
 ib_ipath-$(CONFIG_X86_64) += ipath_wc_x86_64.o
 ib_ipath-$(CONFIG_PPC64) += ipath_wc_ppc64.o
diff --git a/drivers/infiniband/hw/ipath/ipath_driver.c b/drivers/infiniband/hw/ipath/ipath_driver.c
index 6302626d17f0..21337468c652 100644
--- a/drivers/infiniband/hw/ipath/ipath_driver.c
+++ b/drivers/infiniband/hw/ipath/ipath_driver.c
@@ -132,18 +132,13 @@ static int __devinit ipath_init_one(struct pci_dev *,
 
 /* Only needed for registration, nothing else needs this info */
 #define PCI_VENDOR_ID_PATHSCALE 0x1fc1
-#define PCI_VENDOR_ID_QLOGIC 0x1077
 #define PCI_DEVICE_ID_INFINIPATH_HT 0xd
-#define PCI_DEVICE_ID_INFINIPATH_PE800 0x10
-#define PCI_DEVICE_ID_INFINIPATH_7220 0x7220
 
 /* Number of seconds before our card status check...  */
 #define STATUS_TIMEOUT 60
 
 static const struct pci_device_id ipath_pci_tbl[] = {
 	{ PCI_DEVICE(PCI_VENDOR_ID_PATHSCALE, PCI_DEVICE_ID_INFINIPATH_HT) },
-	{ PCI_DEVICE(PCI_VENDOR_ID_PATHSCALE, PCI_DEVICE_ID_INFINIPATH_PE800) },
-	{ PCI_DEVICE(PCI_VENDOR_ID_QLOGIC, PCI_DEVICE_ID_INFINIPATH_7220) },
 	{ 0, }
 };
 
@@ -521,30 +516,9 @@ static int __devinit ipath_init_one(struct pci_dev *pdev,
 	/* setup the chip-specific functions, as early as possible. */
 	switch (ent->device) {
 	case PCI_DEVICE_ID_INFINIPATH_HT:
-#ifdef CONFIG_HT_IRQ
 		ipath_init_iba6110_funcs(dd);
 		break;
-#else
-		ipath_dev_err(dd, "QLogic HT device 0x%x cannot work if "
-			      "CONFIG_HT_IRQ is not enabled\n", ent->device);
-		return -ENODEV;
-#endif
-	case PCI_DEVICE_ID_INFINIPATH_PE800:
-#ifdef CONFIG_PCI_MSI
-		ipath_init_iba6120_funcs(dd);
-		break;
-#else
-		ipath_dev_err(dd, "QLogic PCIE device 0x%x cannot work if "
-			      "CONFIG_PCI_MSI is not enabled\n", ent->device);
-		return -ENODEV;
-#endif
-	case PCI_DEVICE_ID_INFINIPATH_7220:
-#ifndef CONFIG_PCI_MSI
-		ipath_dbg("CONFIG_PCI_MSI is not enabled, "
-			  "using INTx for unit %u\n", dd->ipath_unit);
-#endif
-		ipath_init_iba7220_funcs(dd);
-		break;
+
 	default:
 		ipath_dev_err(dd, "Found unknown QLogic deviceid 0x%x, "
 			      "failing\n", ent->device);
diff --git a/drivers/infiniband/hw/ipath/ipath_iba6120.c b/drivers/infiniband/hw/ipath/ipath_iba6120.c
deleted file mode 100644
index 4b4a30b0dabd..000000000000
--- a/drivers/infiniband/hw/ipath/ipath_iba6120.c
+++ /dev/null
@@ -1,1862 +0,0 @@
-/*
- * Copyright (c) 2006, 2007, 2008 QLogic Corporation. All rights reserved.
- * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
- *
- * This software is available to you under a choice of one of two
- * licenses.  You may choose to be licensed under the terms of the GNU
- * General Public License (GPL) Version 2, available from the file
- * COPYING in the main directory of this source tree, or the
- * OpenIB.org BSD license below:
- *
- *     Redistribution and use in source and binary forms, with or
- *     without modification, are permitted provided that the following
- *     conditions are met:
- *
- *      - Redistributions of source code must retain the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer.
- *
- *      - Redistributions in binary form must reproduce the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer in the documentation and/or other materials
- *        provided with the distribution.
- *
- * 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.
- */
-/*
- * This file contains all of the code that is specific to the
- * InfiniPath PCIe chip.
- */
-
-#include <linux/interrupt.h>
-#include <linux/pci.h>
-#include <linux/delay.h>
-#include <rdma/ib_verbs.h>
-
-#include "ipath_kernel.h"
-#include "ipath_registers.h"
-
-static void ipath_setup_pe_setextled(struct ipath_devdata *, u64, u64);
-
-/*
- * This file contains all the chip-specific register information and
- * access functions for the QLogic InfiniPath PCI-Express chip.
- *
- * This lists the InfiniPath registers, in the actual chip layout.
- * This structure should never be directly accessed.
- */
-struct _infinipath_do_not_use_kernel_regs {
-	unsigned long long Revision;
-	unsigned long long Control;
-	unsigned long long PageAlign;
-	unsigned long long PortCnt;
-	unsigned long long DebugPortSelect;
-	unsigned long long Reserved0;
-	unsigned long long SendRegBase;
-	unsigned long long UserRegBase;
-	unsigned long long CounterRegBase;
-	unsigned long long Scratch;
-	unsigned long long Reserved1;
-	unsigned long long Reserved2;
-	unsigned long long IntBlocked;
-	unsigned long long IntMask;
-	unsigned long long IntStatus;
-	unsigned long long IntClear;
-	unsigned long long ErrorMask;
-	unsigned long long ErrorStatus;
-	unsigned long long ErrorClear;
-	unsigned long long HwErrMask;
-	unsigned long long HwErrStatus;
-	unsigned long long HwErrClear;
-	unsigned long long HwDiagCtrl;
-	unsigned long long MDIO;
-	unsigned long long IBCStatus;
-	unsigned long long IBCCtrl;
-	unsigned long long ExtStatus;
-	unsigned long long ExtCtrl;
-	unsigned long long GPIOOut;
-	unsigned long long GPIOMask;
-	unsigned long long GPIOStatus;
-	unsigned long long GPIOClear;
-	unsigned long long RcvCtrl;
-	unsigned long long RcvBTHQP;
-	unsigned long long RcvHdrSize;
-	unsigned long long RcvHdrCnt;
-	unsigned long long RcvHdrEntSize;
-	unsigned long long RcvTIDBase;
-	unsigned long long RcvTIDCnt;
-	unsigned long long RcvEgrBase;
-	unsigned long long RcvEgrCnt;
-	unsigned long long RcvBufBase;
-	unsigned long long RcvBufSize;
-	unsigned long long RxIntMemBase;
-	unsigned long long RxIntMemSize;
-	unsigned long long RcvPartitionKey;
-	unsigned long long Reserved3;
-	unsigned long long RcvPktLEDCnt;
-	unsigned long long Reserved4[8];
-	unsigned long long SendCtrl;
-	unsigned long long SendPIOBufBase;
-	unsigned long long SendPIOSize;
-	unsigned long long SendPIOBufCnt;
-	unsigned long long SendPIOAvailAddr;
-	unsigned long long TxIntMemBase;
-	unsigned long long TxIntMemSize;
-	unsigned long long Reserved5;
-	unsigned long long PCIeRBufTestReg0;
-	unsigned long long PCIeRBufTestReg1;
-	unsigned long long Reserved51[6];
-	unsigned long long SendBufferError;
-	unsigned long long SendBufferErrorCONT1;
-	unsigned long long Reserved6SBE[6];
-	unsigned long long RcvHdrAddr0;
-	unsigned long long RcvHdrAddr1;
-	unsigned long long RcvHdrAddr2;
-	unsigned long long RcvHdrAddr3;
-	unsigned long long RcvHdrAddr4;
-	unsigned long long Reserved7RHA[11];
-	unsigned long long RcvHdrTailAddr0;
-	unsigned long long RcvHdrTailAddr1;
-	unsigned long long RcvHdrTailAddr2;
-	unsigned long long RcvHdrTailAddr3;
-	unsigned long long RcvHdrTailAddr4;
-	unsigned long long Reserved8RHTA[11];
-	unsigned long long Reserved9SW[8];
-	unsigned long long SerdesConfig0;
-	unsigned long long SerdesConfig1;
-	unsigned long long SerdesStatus;
-	unsigned long long XGXSConfig;
-	unsigned long long IBPLLCfg;
-	unsigned long long Reserved10SW2[3];
-	unsigned long long PCIEQ0SerdesConfig0;
-	unsigned long long PCIEQ0SerdesConfig1;
-	unsigned long long PCIEQ0SerdesStatus;
-	unsigned long long Reserved11;
-	unsigned long long PCIEQ1SerdesConfig0;
-	unsigned long long PCIEQ1SerdesConfig1;
-	unsigned long long PCIEQ1SerdesStatus;
-	unsigned long long Reserved12;
-};
-
-struct _infinipath_do_not_use_counters {
-	__u64 LBIntCnt;
-	__u64 LBFlowStallCnt;
-	__u64 Reserved1;
-	__u64 TxUnsupVLErrCnt;
-	__u64 TxDataPktCnt;
-	__u64 TxFlowPktCnt;
-	__u64 TxDwordCnt;
-	__u64 TxLenErrCnt;
-	__u64 TxMaxMinLenErrCnt;
-	__u64 TxUnderrunCnt;
-	__u64 TxFlowStallCnt;
-	__u64 TxDroppedPktCnt;
-	__u64 RxDroppedPktCnt;
-	__u64 RxDataPktCnt;
-	__u64 RxFlowPktCnt;
-	__u64 RxDwordCnt;
-	__u64 RxLenErrCnt;
-	__u64 RxMaxMinLenErrCnt;
-	__u64 RxICRCErrCnt;
-	__u64 RxVCRCErrCnt;
-	__u64 RxFlowCtrlErrCnt;
-	__u64 RxBadFormatCnt;
-	__u64 RxLinkProblemCnt;
-	__u64 RxEBPCnt;
-	__u64 RxLPCRCErrCnt;
-	__u64 RxBufOvflCnt;
-	__u64 RxTIDFullErrCnt;
-	__u64 RxTIDValidErrCnt;
-	__u64 RxPKeyMismatchCnt;
-	__u64 RxP0HdrEgrOvflCnt;
-	__u64 RxP1HdrEgrOvflCnt;
-	__u64 RxP2HdrEgrOvflCnt;
-	__u64 RxP3HdrEgrOvflCnt;
-	__u64 RxP4HdrEgrOvflCnt;
-	__u64 RxP5HdrEgrOvflCnt;
-	__u64 RxP6HdrEgrOvflCnt;
-	__u64 RxP7HdrEgrOvflCnt;
-	__u64 RxP8HdrEgrOvflCnt;
-	__u64 Reserved6;
-	__u64 Reserved7;
-	__u64 IBStatusChangeCnt;
-	__u64 IBLinkErrRecoveryCnt;
-	__u64 IBLinkDownedCnt;
-	__u64 IBSymbolErrCnt;
-};
-
-#define IPATH_KREG_OFFSET(field) (offsetof( \
-	struct _infinipath_do_not_use_kernel_regs, field) / sizeof(u64))
-#define IPATH_CREG_OFFSET(field) (offsetof( \
-	struct _infinipath_do_not_use_counters, field) / sizeof(u64))
-
-static const struct ipath_kregs ipath_pe_kregs = {
-	.kr_control = IPATH_KREG_OFFSET(Control),
-	.kr_counterregbase = IPATH_KREG_OFFSET(CounterRegBase),
-	.kr_debugportselect = IPATH_KREG_OFFSET(DebugPortSelect),
-	.kr_errorclear = IPATH_KREG_OFFSET(ErrorClear),
-	.kr_errormask = IPATH_KREG_OFFSET(ErrorMask),
-	.kr_errorstatus = IPATH_KREG_OFFSET(ErrorStatus),
-	.kr_extctrl = IPATH_KREG_OFFSET(ExtCtrl),
-	.kr_extstatus = IPATH_KREG_OFFSET(ExtStatus),
-	.kr_gpio_clear = IPATH_KREG_OFFSET(GPIOClear),
-	.kr_gpio_mask = IPATH_KREG_OFFSET(GPIOMask),
-	.kr_gpio_out = IPATH_KREG_OFFSET(GPIOOut),
-	.kr_gpio_status = IPATH_KREG_OFFSET(GPIOStatus),
-	.kr_hwdiagctrl = IPATH_KREG_OFFSET(HwDiagCtrl),
-	.kr_hwerrclear = IPATH_KREG_OFFSET(HwErrClear),
-	.kr_hwerrmask = IPATH_KREG_OFFSET(HwErrMask),
-	.kr_hwerrstatus = IPATH_KREG_OFFSET(HwErrStatus),
-	.kr_ibcctrl = IPATH_KREG_OFFSET(IBCCtrl),
-	.kr_ibcstatus = IPATH_KREG_OFFSET(IBCStatus),
-	.kr_intblocked = IPATH_KREG_OFFSET(IntBlocked),
-	.kr_intclear = IPATH_KREG_OFFSET(IntClear),
-	.kr_intmask = IPATH_KREG_OFFSET(IntMask),
-	.kr_intstatus = IPATH_KREG_OFFSET(IntStatus),
-	.kr_mdio = IPATH_KREG_OFFSET(MDIO),
-	.kr_pagealign = IPATH_KREG_OFFSET(PageAlign),
-	.kr_partitionkey = IPATH_KREG_OFFSET(RcvPartitionKey),
-	.kr_portcnt = IPATH_KREG_OFFSET(PortCnt),
-	.kr_rcvbthqp = IPATH_KREG_OFFSET(RcvBTHQP),
-	.kr_rcvbufbase = IPATH_KREG_OFFSET(RcvBufBase),
-	.kr_rcvbufsize = IPATH_KREG_OFFSET(RcvBufSize),
-	.kr_rcvctrl = IPATH_KREG_OFFSET(RcvCtrl),
-	.kr_rcvegrbase = IPATH_KREG_OFFSET(RcvEgrBase),
-	.kr_rcvegrcnt = IPATH_KREG_OFFSET(RcvEgrCnt),
-	.kr_rcvhdrcnt = IPATH_KREG_OFFSET(RcvHdrCnt),
-	.kr_rcvhdrentsize = IPATH_KREG_OFFSET(RcvHdrEntSize),
-	.kr_rcvhdrsize = IPATH_KREG_OFFSET(RcvHdrSize),
-	.kr_rcvintmembase = IPATH_KREG_OFFSET(RxIntMemBase),
-	.kr_rcvintmemsize = IPATH_KREG_OFFSET(RxIntMemSize),
-	.kr_rcvtidbase = IPATH_KREG_OFFSET(RcvTIDBase),
-	.kr_rcvtidcnt = IPATH_KREG_OFFSET(RcvTIDCnt),
-	.kr_revision = IPATH_KREG_OFFSET(Revision),
-	.kr_scratch = IPATH_KREG_OFFSET(Scratch),
-	.kr_sendbuffererror = IPATH_KREG_OFFSET(SendBufferError),
-	.kr_sendctrl = IPATH_KREG_OFFSET(SendCtrl),
-	.kr_sendpioavailaddr = IPATH_KREG_OFFSET(SendPIOAvailAddr),
-	.kr_sendpiobufbase = IPATH_KREG_OFFSET(SendPIOBufBase),
-	.kr_sendpiobufcnt = IPATH_KREG_OFFSET(SendPIOBufCnt),
-	.kr_sendpiosize = IPATH_KREG_OFFSET(SendPIOSize),
-	.kr_sendregbase = IPATH_KREG_OFFSET(SendRegBase),
-	.kr_txintmembase = IPATH_KREG_OFFSET(TxIntMemBase),
-	.kr_txintmemsize = IPATH_KREG_OFFSET(TxIntMemSize),
-	.kr_userregbase = IPATH_KREG_OFFSET(UserRegBase),
-	.kr_serdesconfig0 = IPATH_KREG_OFFSET(SerdesConfig0),
-	.kr_serdesconfig1 = IPATH_KREG_OFFSET(SerdesConfig1),
-	.kr_serdesstatus = IPATH_KREG_OFFSET(SerdesStatus),
-	.kr_xgxsconfig = IPATH_KREG_OFFSET(XGXSConfig),
-	.kr_ibpllcfg = IPATH_KREG_OFFSET(IBPLLCfg),
-
-	/*
-	 * These should not be used directly via ipath_write_kreg64(),
-	 * use them with ipath_write_kreg64_port(),
-	 */
-	.kr_rcvhdraddr = IPATH_KREG_OFFSET(RcvHdrAddr0),
-	.kr_rcvhdrtailaddr = IPATH_KREG_OFFSET(RcvHdrTailAddr0),
-
-	/* The rcvpktled register controls one of the debug port signals, so
-	 * a packet activity LED can be connected to it. */
-	.kr_rcvpktledcnt = IPATH_KREG_OFFSET(RcvPktLEDCnt),
-	.kr_pcierbuftestreg0 = IPATH_KREG_OFFSET(PCIeRBufTestReg0),
-	.kr_pcierbuftestreg1 = IPATH_KREG_OFFSET(PCIeRBufTestReg1),
-	.kr_pcieq0serdesconfig0 = IPATH_KREG_OFFSET(PCIEQ0SerdesConfig0),
-	.kr_pcieq0serdesconfig1 = IPATH_KREG_OFFSET(PCIEQ0SerdesConfig1),
-	.kr_pcieq0serdesstatus = IPATH_KREG_OFFSET(PCIEQ0SerdesStatus),
-	.kr_pcieq1serdesconfig0 = IPATH_KREG_OFFSET(PCIEQ1SerdesConfig0),
-	.kr_pcieq1serdesconfig1 = IPATH_KREG_OFFSET(PCIEQ1SerdesConfig1),
-	.kr_pcieq1serdesstatus = IPATH_KREG_OFFSET(PCIEQ1SerdesStatus)
-};
-
-static const struct ipath_cregs ipath_pe_cregs = {
-	.cr_badformatcnt = IPATH_CREG_OFFSET(RxBadFormatCnt),
-	.cr_erricrccnt = IPATH_CREG_OFFSET(RxICRCErrCnt),
-	.cr_errlinkcnt = IPATH_CREG_OFFSET(RxLinkProblemCnt),
-	.cr_errlpcrccnt = IPATH_CREG_OFFSET(RxLPCRCErrCnt),
-	.cr_errpkey = IPATH_CREG_OFFSET(RxPKeyMismatchCnt),
-	.cr_errrcvflowctrlcnt = IPATH_CREG_OFFSET(RxFlowCtrlErrCnt),
-	.cr_err_rlencnt = IPATH_CREG_OFFSET(RxLenErrCnt),
-	.cr_errslencnt = IPATH_CREG_OFFSET(TxLenErrCnt),
-	.cr_errtidfull = IPATH_CREG_OFFSET(RxTIDFullErrCnt),
-	.cr_errtidvalid = IPATH_CREG_OFFSET(RxTIDValidErrCnt),
-	.cr_errvcrccnt = IPATH_CREG_OFFSET(RxVCRCErrCnt),
-	.cr_ibstatuschange = IPATH_CREG_OFFSET(IBStatusChangeCnt),
-	.cr_intcnt = IPATH_CREG_OFFSET(LBIntCnt),
-	.cr_invalidrlencnt = IPATH_CREG_OFFSET(RxMaxMinLenErrCnt),
-	.cr_invalidslencnt = IPATH_CREG_OFFSET(TxMaxMinLenErrCnt),
-	.cr_lbflowstallcnt = IPATH_CREG_OFFSET(LBFlowStallCnt),
-	.cr_pktrcvcnt = IPATH_CREG_OFFSET(RxDataPktCnt),
-	.cr_pktrcvflowctrlcnt = IPATH_CREG_OFFSET(RxFlowPktCnt),
-	.cr_pktsendcnt = IPATH_CREG_OFFSET(TxDataPktCnt),
-	.cr_pktsendflowcnt = IPATH_CREG_OFFSET(TxFlowPktCnt),
-	.cr_portovflcnt = IPATH_CREG_OFFSET(RxP0HdrEgrOvflCnt),
-	.cr_rcvebpcnt = IPATH_CREG_OFFSET(RxEBPCnt),
-	.cr_rcvovflcnt = IPATH_CREG_OFFSET(RxBufOvflCnt),
-	.cr_senddropped = IPATH_CREG_OFFSET(TxDroppedPktCnt),
-	.cr_sendstallcnt = IPATH_CREG_OFFSET(TxFlowStallCnt),
-	.cr_sendunderruncnt = IPATH_CREG_OFFSET(TxUnderrunCnt),
-	.cr_wordrcvcnt = IPATH_CREG_OFFSET(RxDwordCnt),
-	.cr_wordsendcnt = IPATH_CREG_OFFSET(TxDwordCnt),
-	.cr_unsupvlcnt = IPATH_CREG_OFFSET(TxUnsupVLErrCnt),
-	.cr_rxdroppktcnt = IPATH_CREG_OFFSET(RxDroppedPktCnt),
-	.cr_iblinkerrrecovcnt = IPATH_CREG_OFFSET(IBLinkErrRecoveryCnt),
-	.cr_iblinkdowncnt = IPATH_CREG_OFFSET(IBLinkDownedCnt),
-	.cr_ibsymbolerrcnt = IPATH_CREG_OFFSET(IBSymbolErrCnt)
-};
-
-/* kr_control bits */
-#define INFINIPATH_C_RESET 1U
-
-/* kr_intstatus, kr_intclear, kr_intmask bits */
-#define INFINIPATH_I_RCVURG_MASK ((1U<<5)-1)
-#define INFINIPATH_I_RCVURG_SHIFT 0
-#define INFINIPATH_I_RCVAVAIL_MASK ((1U<<5)-1)
-#define INFINIPATH_I_RCVAVAIL_SHIFT 12
-
-/* kr_hwerrclear, kr_hwerrmask, kr_hwerrstatus, bits */
-#define INFINIPATH_HWE_PCIEMEMPARITYERR_MASK  0x000000000000003fULL
-#define INFINIPATH_HWE_PCIEMEMPARITYERR_SHIFT 0
-#define INFINIPATH_HWE_PCIEPOISONEDTLP      0x0000000010000000ULL
-#define INFINIPATH_HWE_PCIECPLTIMEOUT       0x0000000020000000ULL
-#define INFINIPATH_HWE_PCIEBUSPARITYXTLH    0x0000000040000000ULL
-#define INFINIPATH_HWE_PCIEBUSPARITYXADM    0x0000000080000000ULL
-#define INFINIPATH_HWE_PCIEBUSPARITYRADM    0x0000000100000000ULL
-#define INFINIPATH_HWE_COREPLL_FBSLIP       0x0080000000000000ULL
-#define INFINIPATH_HWE_COREPLL_RFSLIP       0x0100000000000000ULL
-#define INFINIPATH_HWE_PCIE1PLLFAILED       0x0400000000000000ULL
-#define INFINIPATH_HWE_PCIE0PLLFAILED       0x0800000000000000ULL
-#define INFINIPATH_HWE_SERDESPLLFAILED      0x1000000000000000ULL
-
-#define IBA6120_IBCS_LINKTRAININGSTATE_MASK 0xf
-#define IBA6120_IBCS_LINKSTATE_SHIFT 4
-
-/* kr_extstatus bits */
-#define INFINIPATH_EXTS_FREQSEL 0x2
-#define INFINIPATH_EXTS_SERDESSEL 0x4
-#define INFINIPATH_EXTS_MEMBIST_ENDTEST     0x0000000000004000
-#define INFINIPATH_EXTS_MEMBIST_FOUND       0x0000000000008000
-
-/* kr_xgxsconfig bits */
-#define INFINIPATH_XGXS_RESET          0x5ULL
-
-#define _IPATH_GPIO_SDA_NUM 1
-#define _IPATH_GPIO_SCL_NUM 0
-
-#define IPATH_GPIO_SDA (1ULL << \
-	(_IPATH_GPIO_SDA_NUM+INFINIPATH_EXTC_GPIOOE_SHIFT))
-#define IPATH_GPIO_SCL (1ULL << \
-	(_IPATH_GPIO_SCL_NUM+INFINIPATH_EXTC_GPIOOE_SHIFT))
-
-#define INFINIPATH_RT_BUFSIZE_MASK 0xe0000000ULL
-#define INFINIPATH_RT_BUFSIZE_SHIFTVAL(tid) \
-	((((tid) & INFINIPATH_RT_BUFSIZE_MASK) >> 29) + 11 - 1)
-#define INFINIPATH_RT_BUFSIZE(tid) (1 << INFINIPATH_RT_BUFSIZE_SHIFTVAL(tid))
-#define INFINIPATH_RT_IS_VALID(tid) \
-	(((tid) & INFINIPATH_RT_BUFSIZE_MASK) && \
-	 ((((tid) & INFINIPATH_RT_BUFSIZE_MASK) != INFINIPATH_RT_BUFSIZE_MASK)))
-#define INFINIPATH_RT_ADDR_MASK 0x1FFFFFFFULL /* 29 bits valid */
-#define INFINIPATH_RT_ADDR_SHIFT 10
-
-#define INFINIPATH_R_INTRAVAIL_SHIFT 16
-#define INFINIPATH_R_TAILUPD_SHIFT 31
-
-/* 6120 specific hardware errors... */
-static const struct ipath_hwerror_msgs ipath_6120_hwerror_msgs[] = {
-	INFINIPATH_HWE_MSG(PCIEPOISONEDTLP, "PCIe Poisoned TLP"),
-	INFINIPATH_HWE_MSG(PCIECPLTIMEOUT, "PCIe completion timeout"),
-	/*
-	 * In practice, it's unlikely wthat we'll see PCIe PLL, or bus
-	 * parity or memory parity error failures, because most likely we
-	 * won't be able to talk to the core of the chip.  Nonetheless, we
-	 * might see them, if they are in parts of the PCIe core that aren't
-	 * essential.
-	 */
-	INFINIPATH_HWE_MSG(PCIE1PLLFAILED, "PCIePLL1"),
-	INFINIPATH_HWE_MSG(PCIE0PLLFAILED, "PCIePLL0"),
-	INFINIPATH_HWE_MSG(PCIEBUSPARITYXTLH, "PCIe XTLH core parity"),
-	INFINIPATH_HWE_MSG(PCIEBUSPARITYXADM, "PCIe ADM TX core parity"),
-	INFINIPATH_HWE_MSG(PCIEBUSPARITYRADM, "PCIe ADM RX core parity"),
-	INFINIPATH_HWE_MSG(RXDSYNCMEMPARITYERR, "Rx Dsync"),
-	INFINIPATH_HWE_MSG(SERDESPLLFAILED, "SerDes PLL"),
-};
-
-#define TXE_PIO_PARITY ((INFINIPATH_HWE_TXEMEMPARITYERR_PIOBUF | \
-		        INFINIPATH_HWE_TXEMEMPARITYERR_PIOPBC) \
-		        << INFINIPATH_HWE_TXEMEMPARITYERR_SHIFT)
-#define RXE_EAGER_PARITY (INFINIPATH_HWE_RXEMEMPARITYERR_EAGERTID \
-			  << INFINIPATH_HWE_RXEMEMPARITYERR_SHIFT)
-
-static void ipath_pe_put_tid_2(struct ipath_devdata *, u64 __iomem *,
-			       u32, unsigned long);
-
-/*
- * On platforms using this chip, and not having ordered WC stores, we
- * can get TXE parity errors due to speculative reads to the PIO buffers,
- * and this, due to a chip bug can result in (many) false parity error
- * reports.  So it's a debug print on those, and an info print on systems
- * where the speculative reads don't occur.
- */
-static void ipath_pe_txe_recover(struct ipath_devdata *dd)
-{
-	if (ipath_unordered_wc())
-		ipath_dbg("Recovering from TXE PIO parity error\n");
-	else {
-		++ipath_stats.sps_txeparity;
-		dev_info(&dd->pcidev->dev,
-			"Recovering from TXE PIO parity error\n");
-	}
-}
-
-
-/**
- * ipath_pe_handle_hwerrors - display hardware errors.
- * @dd: the infinipath device
- * @msg: the output buffer
- * @msgl: the size of the output buffer
- *
- * Use same msg buffer as regular errors to avoid excessive stack
- * use.  Most hardware errors are catastrophic, but for right now,
- * we'll print them and continue.  We reuse the same message buffer as
- * ipath_handle_errors() to avoid excessive stack usage.
- */
-static void ipath_pe_handle_hwerrors(struct ipath_devdata *dd, char *msg,
-				     size_t msgl)
-{
-	ipath_err_t hwerrs;
-	u32 bits, ctrl;
-	int isfatal = 0;
-	char bitsmsg[64];
-	int log_idx;
-
-	hwerrs = ipath_read_kreg64(dd, dd->ipath_kregs->kr_hwerrstatus);
-	if (!hwerrs) {
-		/*
-		 * better than printing cofusing messages
-		 * This seems to be related to clearing the crc error, or
-		 * the pll error during init.
-		 */
-		ipath_cdbg(VERBOSE, "Called but no hardware errors set\n");
-		return;
-	} else if (hwerrs == ~0ULL) {
-		ipath_dev_err(dd, "Read of hardware error status failed "
-			      "(all bits set); ignoring\n");
-		return;
-	}
-	ipath_stats.sps_hwerrs++;
-
-	/* Always clear the error status register, except MEMBISTFAIL,
-	 * regardless of whether we continue or stop using the chip.
-	 * We want that set so we know it failed, even across driver reload.
-	 * We'll still ignore it in the hwerrmask.  We do this partly for
-	 * diagnostics, but also for support */
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_hwerrclear,
-			 hwerrs&~INFINIPATH_HWE_MEMBISTFAILED);
-
-	hwerrs &= dd->ipath_hwerrmask;
-
-	/* We log some errors to EEPROM, check if we have any of those. */
-	for (log_idx = 0; log_idx < IPATH_EEP_LOG_CNT; ++log_idx)
-		if (hwerrs & dd->ipath_eep_st_masks[log_idx].hwerrs_to_log)
-			ipath_inc_eeprom_err(dd, log_idx, 1);
-
-	/*
-	 * make sure we get this much out, unless told to be quiet,
-	 * or it's occurred within the last 5 seconds
-	 */
-	if ((hwerrs & ~(dd->ipath_lasthwerror | TXE_PIO_PARITY |
-			RXE_EAGER_PARITY)) ||
-	    (ipath_debug & __IPATH_VERBDBG))
-		dev_info(&dd->pcidev->dev, "Hardware error: hwerr=0x%llx "
-			 "(cleared)\n", (unsigned long long) hwerrs);
-	dd->ipath_lasthwerror |= hwerrs;
-
-	if (hwerrs & ~dd->ipath_hwe_bitsextant)
-		ipath_dev_err(dd, "hwerror interrupt with unknown errors "
-			      "%llx set\n", (unsigned long long)
-			      (hwerrs & ~dd->ipath_hwe_bitsextant));
-
-	ctrl = ipath_read_kreg32(dd, dd->ipath_kregs->kr_control);
-	if ((ctrl & INFINIPATH_C_FREEZEMODE) && !ipath_diag_inuse) {
-		/*
-		 * parity errors in send memory are recoverable,
-		 * just cancel the send (if indicated in * sendbuffererror),
-		 * count the occurrence, unfreeze (if no other handled
-		 * hardware error bits are set), and continue. They can
-		 * occur if a processor speculative read is done to the PIO
-		 * buffer while we are sending a packet, for example.
-		 */
-		if (hwerrs & TXE_PIO_PARITY) {
-			ipath_pe_txe_recover(dd);
-			hwerrs &= ~TXE_PIO_PARITY;
-		}
-		if (!hwerrs) {
-			static u32 freeze_cnt;
-
-			freeze_cnt++;
-			ipath_dbg("Clearing freezemode on ignored or recovered "
-				  "hardware error (%u)\n", freeze_cnt);
-			ipath_clear_freeze(dd);
-		}
-	}
-
-	*msg = '\0';
-
-	if (hwerrs & INFINIPATH_HWE_MEMBISTFAILED) {
-		strlcat(msg, "[Memory BIST test failed, InfiniPath hardware unusable]",
-			msgl);
-		/* ignore from now on, so disable until driver reloaded */
-		*dd->ipath_statusp |= IPATH_STATUS_HWERROR;
-		dd->ipath_hwerrmask &= ~INFINIPATH_HWE_MEMBISTFAILED;
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_hwerrmask,
-				 dd->ipath_hwerrmask);
-	}
-
-	ipath_format_hwerrors(hwerrs,
-			      ipath_6120_hwerror_msgs,
-			      sizeof(ipath_6120_hwerror_msgs)/
-			      sizeof(ipath_6120_hwerror_msgs[0]),
-			      msg, msgl);
-
-	if (hwerrs & (INFINIPATH_HWE_PCIEMEMPARITYERR_MASK
-		      << INFINIPATH_HWE_PCIEMEMPARITYERR_SHIFT)) {
-		bits = (u32) ((hwerrs >>
-			       INFINIPATH_HWE_PCIEMEMPARITYERR_SHIFT) &
-			      INFINIPATH_HWE_PCIEMEMPARITYERR_MASK);
-		snprintf(bitsmsg, sizeof bitsmsg,
-			 "[PCIe Mem Parity Errs %x] ", bits);
-		strlcat(msg, bitsmsg, msgl);
-	}
-
-#define _IPATH_PLL_FAIL (INFINIPATH_HWE_COREPLL_FBSLIP |	\
-			 INFINIPATH_HWE_COREPLL_RFSLIP )
-
-	if (hwerrs & _IPATH_PLL_FAIL) {
-		snprintf(bitsmsg, sizeof bitsmsg,
-			 "[PLL failed (%llx), InfiniPath hardware unusable]",
-			 (unsigned long long) hwerrs & _IPATH_PLL_FAIL);
-		strlcat(msg, bitsmsg, msgl);
-		/* ignore from now on, so disable until driver reloaded */
-		dd->ipath_hwerrmask &= ~(hwerrs & _IPATH_PLL_FAIL);
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_hwerrmask,
-				 dd->ipath_hwerrmask);
-	}
-
-	if (hwerrs & INFINIPATH_HWE_SERDESPLLFAILED) {
-		/*
-		 * If it occurs, it is left masked since the external
-		 * interface is unused
-		 */
-		dd->ipath_hwerrmask &= ~INFINIPATH_HWE_SERDESPLLFAILED;
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_hwerrmask,
-				 dd->ipath_hwerrmask);
-	}
-
-	if (hwerrs) {
-		/*
-		 * if any set that we aren't ignoring; only
-		 * make the complaint once, in case it's stuck
-		 * or recurring, and we get here multiple
-		 * times.
-		 */
-		ipath_dev_err(dd, "%s hardware error\n", msg);
-		if (dd->ipath_flags & IPATH_INITTED) {
-			ipath_set_linkstate(dd, IPATH_IB_LINKDOWN);
-			ipath_setup_pe_setextled(dd,
-				INFINIPATH_IBCS_L_STATE_DOWN,
-				INFINIPATH_IBCS_LT_STATE_DISABLED);
-			ipath_dev_err(dd, "Fatal Hardware Error (freeze "
-					  "mode), no longer usable, SN %.16s\n",
-					  dd->ipath_serial);
-			isfatal = 1;
-		}
-		*dd->ipath_statusp &= ~IPATH_STATUS_IB_READY;
-		/* mark as having had error */
-		*dd->ipath_statusp |= IPATH_STATUS_HWERROR;
-		/*
-		 * mark as not usable, at a minimum until driver
-		 * is reloaded, probably until reboot, since no
-		 * other reset is possible.
-		 */
-		dd->ipath_flags &= ~IPATH_INITTED;
-	} else
-		*msg = 0; /* recovered from all of them */
-
-	if (isfatal && !ipath_diag_inuse && dd->ipath_freezemsg && msg) {
-		/*
-		 * for /sys status file ; if no trailing brace is copied,
-		 * we'll know it was truncated.
-		 */
-		snprintf(dd->ipath_freezemsg, dd->ipath_freezelen,
-			 "{%s}", msg);
-	}
-}
-
-/**
- * ipath_pe_boardname - fill in the board name
- * @dd: the infinipath device
- * @name: the output buffer
- * @namelen: the size of the output buffer
- *
- * info is based on the board revision register
- */
-static int ipath_pe_boardname(struct ipath_devdata *dd, char *name,
-			      size_t namelen)
-{
-	char *n = NULL;
-	u8 boardrev = dd->ipath_boardrev;
-	int ret;
-
-	switch (boardrev) {
-	case 0:
-		n = "InfiniPath_Emulation";
-		break;
-	case 1:
-		n = "InfiniPath_QLE7140-Bringup";
-		break;
-	case 2:
-		n = "InfiniPath_QLE7140";
-		break;
-	case 3:
-		n = "InfiniPath_QMI7140";
-		break;
-	case 4:
-		n = "InfiniPath_QEM7140";
-		break;
-	case 5:
-		n = "InfiniPath_QMH7140";
-		break;
-	case 6:
-		n = "InfiniPath_QLE7142";
-		break;
-	default:
-		ipath_dev_err(dd,
-			      "Don't yet know about board with ID %u\n",
-			      boardrev);
-		snprintf(name, namelen, "Unknown_InfiniPath_PCIe_%u",
-			 boardrev);
-		break;
-	}
-	if (n)
-		snprintf(name, namelen, "%s", n);
-
-	if (dd->ipath_majrev != 4 || !dd->ipath_minrev || dd->ipath_minrev>2) {
-		ipath_dev_err(dd, "Unsupported InfiniPath hardware revision %u.%u!\n",
-			      dd->ipath_majrev, dd->ipath_minrev);
-		ret = 1;
-	} else {
-		ret = 0;
-		if (dd->ipath_minrev >= 2)
-			dd->ipath_f_put_tid = ipath_pe_put_tid_2;
-	}
-
-	/*
-	 * set here, not in ipath_init_*_funcs because we have to do
-	 * it after we can read chip registers.
-	 */
-	dd->ipath_ureg_align =
-		ipath_read_kreg32(dd, dd->ipath_kregs->kr_pagealign);
-
-	return ret;
-}
-
-/**
- * ipath_pe_init_hwerrors - enable hardware errors
- * @dd: the infinipath device
- *
- * now that we have finished initializing everything that might reasonably
- * cause a hardware error, and cleared those errors bits as they occur,
- * we can enable hardware errors in the mask (potentially enabling
- * freeze mode), and enable hardware errors as errors (along with
- * everything else) in errormask
- */
-static void ipath_pe_init_hwerrors(struct ipath_devdata *dd)
-{
-	ipath_err_t val;
-	u64 extsval;
-
-	extsval = ipath_read_kreg64(dd, dd->ipath_kregs->kr_extstatus);
-
-	if (!(extsval & INFINIPATH_EXTS_MEMBIST_ENDTEST))
-		ipath_dev_err(dd, "MemBIST did not complete!\n");
-	if (extsval & INFINIPATH_EXTS_MEMBIST_FOUND)
-		ipath_dbg("MemBIST corrected\n");
-
-	val = ~0ULL;	/* barring bugs, all hwerrors become interrupts, */
-
-	if (!dd->ipath_boardrev)	// no PLL for Emulator
-		val &= ~INFINIPATH_HWE_SERDESPLLFAILED;
-
-	if (dd->ipath_minrev < 2) {
-		/* workaround bug 9460 in internal interface bus parity
-		 * checking. Fixed (HW bug 9490) in Rev2.
-		 */
-		val &= ~INFINIPATH_HWE_PCIEBUSPARITYRADM;
-	}
-	dd->ipath_hwerrmask = val;
-}
-
-/**
- * ipath_pe_bringup_serdes - bring up the serdes
- * @dd: the infinipath device
- */
-static int ipath_pe_bringup_serdes(struct ipath_devdata *dd)
-{
-	u64 val, config1, prev_val;
-	int ret = 0;
-
-	ipath_dbg("Trying to bringup serdes\n");
-
-	if (ipath_read_kreg64(dd, dd->ipath_kregs->kr_hwerrstatus) &
-	    INFINIPATH_HWE_SERDESPLLFAILED) {
-		ipath_dbg("At start, serdes PLL failed bit set "
-			  "in hwerrstatus, clearing and continuing\n");
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_hwerrclear,
-				 INFINIPATH_HWE_SERDESPLLFAILED);
-	}
-
-	dd->ibdeltainprog = 1;
-	dd->ibsymsnap =
-	     ipath_read_creg32(dd, dd->ipath_cregs->cr_ibsymbolerrcnt);
-	dd->iblnkerrsnap =
-	     ipath_read_creg32(dd, dd->ipath_cregs->cr_iblinkerrrecovcnt);
-
-	val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_serdesconfig0);
-	config1 = ipath_read_kreg64(dd, dd->ipath_kregs->kr_serdesconfig1);
-
-	ipath_cdbg(VERBOSE, "SerDes status config0=%llx config1=%llx, "
-		   "xgxsconfig %llx\n", (unsigned long long) val,
-		   (unsigned long long) config1, (unsigned long long)
-		   ipath_read_kreg64(dd, dd->ipath_kregs->kr_xgxsconfig));
-
-	/*
-	 * Force reset on, also set rxdetect enable.  Must do before reading
-	 * serdesstatus at least for simulation, or some of the bits in
-	 * serdes status will come back as undefined and cause simulation
-	 * failures
-	 */
-	val |= INFINIPATH_SERDC0_RESET_PLL | INFINIPATH_SERDC0_RXDETECT_EN
-		| INFINIPATH_SERDC0_L1PWR_DN;
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_serdesconfig0, val);
-	/* be sure chip saw it */
-	ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
-	udelay(5);		/* need pll reset set at least for a bit */
-	/*
-	 * after PLL is reset, set the per-lane Resets and TxIdle and
-	 * clear the PLL reset and rxdetect (to get falling edge).
-	 * Leave L1PWR bits set (permanently)
-	 */
-	val &= ~(INFINIPATH_SERDC0_RXDETECT_EN | INFINIPATH_SERDC0_RESET_PLL
-		 | INFINIPATH_SERDC0_L1PWR_DN);
-	val |= INFINIPATH_SERDC0_RESET_MASK | INFINIPATH_SERDC0_TXIDLE;
-	ipath_cdbg(VERBOSE, "Clearing pll reset and setting lane resets "
-		   "and txidle (%llx)\n", (unsigned long long) val);
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_serdesconfig0, val);
-	/* be sure chip saw it */
-	ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
-	/* need PLL reset clear for at least 11 usec before lane
-	 * resets cleared; give it a few more to be sure */
-	udelay(15);
-	val &= ~(INFINIPATH_SERDC0_RESET_MASK | INFINIPATH_SERDC0_TXIDLE);
-
-	ipath_cdbg(VERBOSE, "Clearing lane resets and txidle "
-		   "(writing %llx)\n", (unsigned long long) val);
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_serdesconfig0, val);
-	/* be sure chip saw it */
-	val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
-
-	val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_xgxsconfig);
-	prev_val = val;
-	if (val & INFINIPATH_XGXS_RESET)
-		val &= ~INFINIPATH_XGXS_RESET;
-	if (((val >> INFINIPATH_XGXS_RX_POL_SHIFT) &
-	     INFINIPATH_XGXS_RX_POL_MASK) != dd->ipath_rx_pol_inv ) {
-		/* need to compensate for Tx inversion in partner */
-		val &= ~(INFINIPATH_XGXS_RX_POL_MASK <<
-		         INFINIPATH_XGXS_RX_POL_SHIFT);
-		val |= dd->ipath_rx_pol_inv <<
-			INFINIPATH_XGXS_RX_POL_SHIFT;
-	}
-	if (val != prev_val)
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_xgxsconfig, val);
-
-	val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_serdesconfig0);
-
-	/* clear current and de-emphasis bits */
-	config1 &= ~0x0ffffffff00ULL;
-	/* set current to 20ma */
-	config1 |= 0x00000000000ULL;
-	/* set de-emphasis to -5.68dB */
-	config1 |= 0x0cccc000000ULL;
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_serdesconfig1, config1);
-
-	ipath_cdbg(VERBOSE, "done: SerDes status config0=%llx "
-		   "config1=%llx, sstatus=%llx xgxs=%llx\n",
-		   (unsigned long long) val, (unsigned long long) config1,
-		   (unsigned long long)
-		   ipath_read_kreg64(dd, dd->ipath_kregs->kr_serdesstatus),
-		   (unsigned long long)
-		   ipath_read_kreg64(dd, dd->ipath_kregs->kr_xgxsconfig));
-
-	return ret;
-}
-
-/**
- * ipath_pe_quiet_serdes - set serdes to txidle
- * @dd: the infinipath device
- * Called when driver is being unloaded
- */
-static void ipath_pe_quiet_serdes(struct ipath_devdata *dd)
-{
-	u64 val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_serdesconfig0);
-
-	if (dd->ibsymdelta || dd->iblnkerrdelta ||
-	    dd->ibdeltainprog) {
-		u64 diagc;
-		/* enable counter writes */
-		diagc = ipath_read_kreg64(dd, dd->ipath_kregs->kr_hwdiagctrl);
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_hwdiagctrl,
-				 diagc | INFINIPATH_DC_COUNTERWREN);
-
-		if (dd->ibsymdelta || dd->ibdeltainprog) {
-			val = ipath_read_creg32(dd,
-					dd->ipath_cregs->cr_ibsymbolerrcnt);
-			if (dd->ibdeltainprog)
-				val -= val - dd->ibsymsnap;
-			val -= dd->ibsymdelta;
-			ipath_write_creg(dd,
-				  dd->ipath_cregs->cr_ibsymbolerrcnt, val);
-		}
-		if (dd->iblnkerrdelta || dd->ibdeltainprog) {
-			val = ipath_read_creg32(dd,
-					dd->ipath_cregs->cr_iblinkerrrecovcnt);
-			if (dd->ibdeltainprog)
-				val -= val - dd->iblnkerrsnap;
-			val -= dd->iblnkerrdelta;
-			ipath_write_creg(dd,
-				   dd->ipath_cregs->cr_iblinkerrrecovcnt, val);
-	     }
-
-	     /* and disable counter writes */
-	     ipath_write_kreg(dd, dd->ipath_kregs->kr_hwdiagctrl, diagc);
-	}
-	val |= INFINIPATH_SERDC0_TXIDLE;
-	ipath_dbg("Setting TxIdleEn on serdes (config0 = %llx)\n",
-		  (unsigned long long) val);
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_serdesconfig0, val);
-}
-
-static int ipath_pe_intconfig(struct ipath_devdata *dd)
-{
-	u32 chiprev;
-
-	/*
-	 * If the chip supports added error indication via GPIO pins,
-	 * enable interrupts on those bits so the interrupt routine
-	 * can count the events. Also set flag so interrupt routine
-	 * can know they are expected.
-	 */
-	chiprev = dd->ipath_revision >> INFINIPATH_R_CHIPREVMINOR_SHIFT;
-	if ((chiprev & INFINIPATH_R_CHIPREVMINOR_MASK) > 1) {
-		/* Rev2+ reports extra errors via internal GPIO pins */
-		dd->ipath_flags |= IPATH_GPIO_ERRINTRS;
-		dd->ipath_gpio_mask |= IPATH_GPIO_ERRINTR_MASK;
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_gpio_mask,
-				 dd->ipath_gpio_mask);
-	}
-	return 0;
-}
-
-/**
- * ipath_setup_pe_setextled - set the state of the two external LEDs
- * @dd: the infinipath device
- * @lst: the L state
- * @ltst: the LT state
-
- * These LEDs indicate the physical and logical state of IB link.
- * For this chip (at least with recommended board pinouts), LED1
- * is Yellow (logical state) and LED2 is Green (physical state),
- *
- * Note:  We try to match the Mellanox HCA LED behavior as best
- * we can.  Green indicates physical link state is OK (something is
- * plugged in, and we can train).
- * Amber indicates the link is logically up (ACTIVE).
- * Mellanox further blinks the amber LED to indicate data packet
- * activity, but we have no hardware support for that, so it would
- * require waking up every 10-20 msecs and checking the counters
- * on the chip, and then turning the LED off if appropriate.  That's
- * visible overhead, so not something we will do.
- *
- */
-static void ipath_setup_pe_setextled(struct ipath_devdata *dd, u64 lst,
-				     u64 ltst)
-{
-	u64 extctl;
-	unsigned long flags = 0;
-
-	/* the diags use the LED to indicate diag info, so we leave
-	 * the external LED alone when the diags are running */
-	if (ipath_diag_inuse)
-		return;
-
-	/* Allow override of LED display for, e.g. Locating system in rack */
-	if (dd->ipath_led_override) {
-		ltst = (dd->ipath_led_override & IPATH_LED_PHYS)
-			? INFINIPATH_IBCS_LT_STATE_LINKUP
-			: INFINIPATH_IBCS_LT_STATE_DISABLED;
-		lst = (dd->ipath_led_override & IPATH_LED_LOG)
-			? INFINIPATH_IBCS_L_STATE_ACTIVE
-			: INFINIPATH_IBCS_L_STATE_DOWN;
-	}
-
-	spin_lock_irqsave(&dd->ipath_gpio_lock, flags);
-	extctl = dd->ipath_extctrl & ~(INFINIPATH_EXTC_LED1PRIPORT_ON |
-				       INFINIPATH_EXTC_LED2PRIPORT_ON);
-
-	if (ltst == INFINIPATH_IBCS_LT_STATE_LINKUP)
-		extctl |= INFINIPATH_EXTC_LED2PRIPORT_ON;
-	if (lst == INFINIPATH_IBCS_L_STATE_ACTIVE)
-		extctl |= INFINIPATH_EXTC_LED1PRIPORT_ON;
-	dd->ipath_extctrl = extctl;
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_extctrl, extctl);
-	spin_unlock_irqrestore(&dd->ipath_gpio_lock, flags);
-}
-
-/**
- * ipath_setup_pe_cleanup - clean up any per-chip chip-specific stuff
- * @dd: the infinipath device
- *
- * This is called during driver unload.
- * We do the pci_disable_msi here, not in generic code, because it
- * isn't used for the HT chips. If we do end up needing pci_enable_msi
- * at some point in the future for HT, we'll move the call back
- * into the main init_one code.
- */
-static void ipath_setup_pe_cleanup(struct ipath_devdata *dd)
-{
-	dd->ipath_msi_lo = 0;	/* just in case unload fails */
-	pci_disable_msi(dd->pcidev);
-}
-
-static void ipath_6120_pcie_params(struct ipath_devdata *dd)
-{
-	u16 linkstat, speed;
-	int pos;
-
-	pos = pci_find_capability(dd->pcidev, PCI_CAP_ID_EXP);
-	if (!pos) {
-		ipath_dev_err(dd, "Can't find PCI Express capability!\n");
-		goto bail;
-	}
-
-	pci_read_config_word(dd->pcidev, pos + PCI_EXP_LNKSTA,
-			     &linkstat);
-	/*
-	 * speed is bits 0-4, linkwidth is bits 4-8
-	 * no defines for them in headers
-	 */
-	speed = linkstat & 0xf;
-	linkstat >>= 4;
-	linkstat &= 0x1f;
-	dd->ipath_lbus_width = linkstat;
-
-	switch (speed) {
-	case 1:
-		dd->ipath_lbus_speed = 2500; /* Gen1, 2.5GHz */
-		break;
-	case 2:
-		dd->ipath_lbus_speed = 5000; /* Gen1, 5GHz */
-		break;
-	default: /* not defined, assume gen1 */
-		dd->ipath_lbus_speed = 2500;
-		break;
-	}
-
-	if (linkstat < 8)
-		ipath_dev_err(dd,
-			"PCIe width %u (x8 HCA), performance reduced\n",
-			linkstat);
-	else
-		ipath_cdbg(VERBOSE, "PCIe speed %u width %u (x8 HCA)\n",
-			dd->ipath_lbus_speed, linkstat);
-
-	if (speed != 1)
-		ipath_dev_err(dd,
-			"PCIe linkspeed %u is incorrect; "
-			"should be 1 (2500)!\n", speed);
-bail:
-	/* fill in string, even on errors */
-	snprintf(dd->ipath_lbus_info, sizeof(dd->ipath_lbus_info),
-		"PCIe,%uMHz,x%u\n",
-		dd->ipath_lbus_speed,
-		dd->ipath_lbus_width);
-
-	return;
-}
-
-/**
- * ipath_setup_pe_config - setup PCIe config related stuff
- * @dd: the infinipath device
- * @pdev: the PCI device
- *
- * The pci_enable_msi() call will fail on systems with MSI quirks
- * such as those with AMD8131, even if the device of interest is not
- * attached to that device, (in the 2.6.13 - 2.6.15 kernels, at least, fixed
- * late in 2.6.16).
- * All that can be done is to edit the kernel source to remove the quirk
- * check until that is fixed.
- * We do not need to call enable_msi() for our HyperTransport chip,
- * even though it uses MSI, and we want to avoid the quirk warning, so
- * So we call enable_msi only for PCIe.  If we do end up needing
- * pci_enable_msi at some point in the future for HT, we'll move the
- * call back into the main init_one code.
- * We save the msi lo and hi values, so we can restore them after
- * chip reset (the kernel PCI infrastructure doesn't yet handle that
- * correctly).
- */
-static int ipath_setup_pe_config(struct ipath_devdata *dd,
-				 struct pci_dev *pdev)
-{
-	int pos, ret;
-
-	dd->ipath_msi_lo = 0;	/* used as a flag during reset processing */
-	ret = pci_enable_msi(dd->pcidev);
-	if (ret)
-		ipath_dev_err(dd, "pci_enable_msi failed: %d, "
-			      "interrupts may not work\n", ret);
-	/* continue even if it fails, we may still be OK... */
-	dd->ipath_irq = pdev->irq;
-
-	if ((pos = pci_find_capability(dd->pcidev, PCI_CAP_ID_MSI))) {
-		u16 control;
-		pci_read_config_dword(dd->pcidev, pos + PCI_MSI_ADDRESS_LO,
-				      &dd->ipath_msi_lo);
-		pci_read_config_dword(dd->pcidev, pos + PCI_MSI_ADDRESS_HI,
-				      &dd->ipath_msi_hi);
-		pci_read_config_word(dd->pcidev, pos + PCI_MSI_FLAGS,
-				     &control);
-		/* now save the data (vector) info */
-		pci_read_config_word(dd->pcidev,
-				     pos + ((control & PCI_MSI_FLAGS_64BIT)
-					    ? 12 : 8),
-				     &dd->ipath_msi_data);
-		ipath_cdbg(VERBOSE, "Read msi data 0x%x from config offset "
-			   "0x%x, control=0x%x\n", dd->ipath_msi_data,
-			   pos + ((control & PCI_MSI_FLAGS_64BIT) ? 12 : 8),
-			   control);
-		/* we save the cachelinesize also, although it doesn't
-		 * really matter */
-		pci_read_config_byte(dd->pcidev, PCI_CACHE_LINE_SIZE,
-				     &dd->ipath_pci_cacheline);
-	} else
-		ipath_dev_err(dd, "Can't find MSI capability, "
-			      "can't save MSI settings for reset\n");
-
-	ipath_6120_pcie_params(dd);
-
-	dd->ipath_link_width_supported = IB_WIDTH_1X | IB_WIDTH_4X;
-	dd->ipath_link_speed_supported = IPATH_IB_SDR;
-	dd->ipath_link_width_enabled = IB_WIDTH_4X;
-	dd->ipath_link_speed_enabled = dd->ipath_link_speed_supported;
-	/* these can't change for this chip, so set once */
-	dd->ipath_link_width_active = dd->ipath_link_width_enabled;
-	dd->ipath_link_speed_active = dd->ipath_link_speed_enabled;
-	return 0;
-}
-
-static void ipath_init_pe_variables(struct ipath_devdata *dd)
-{
-	/*
-	 * setup the register offsets, since they are different for each
-	 * chip
-	 */
-	dd->ipath_kregs = &ipath_pe_kregs;
-	dd->ipath_cregs = &ipath_pe_cregs;
-
-	/*
-	 * bits for selecting i2c direction and values,
-	 * used for I2C serial flash
-	 */
-	dd->ipath_gpio_sda_num = _IPATH_GPIO_SDA_NUM;
-	dd->ipath_gpio_scl_num = _IPATH_GPIO_SCL_NUM;
-	dd->ipath_gpio_sda = IPATH_GPIO_SDA;
-	dd->ipath_gpio_scl = IPATH_GPIO_SCL;
-
-	/*
-	 * Fill in data for field-values that change in newer chips.
-	 * We dynamically specify only the mask for LINKTRAININGSTATE
-	 * and only the shift for LINKSTATE, as they are the only ones
-	 * that change.  Also precalculate the 3 link states of interest
-	 * and the combined mask.
-	 */
-	dd->ibcs_ls_shift = IBA6120_IBCS_LINKSTATE_SHIFT;
-	dd->ibcs_lts_mask = IBA6120_IBCS_LINKTRAININGSTATE_MASK;
-	dd->ibcs_mask = (INFINIPATH_IBCS_LINKSTATE_MASK <<
-		dd->ibcs_ls_shift) | dd->ibcs_lts_mask;
-	dd->ib_init = (INFINIPATH_IBCS_LT_STATE_LINKUP <<
-		INFINIPATH_IBCS_LINKTRAININGSTATE_SHIFT) |
-		(INFINIPATH_IBCS_L_STATE_INIT << dd->ibcs_ls_shift);
-	dd->ib_arm = (INFINIPATH_IBCS_LT_STATE_LINKUP <<
-		INFINIPATH_IBCS_LINKTRAININGSTATE_SHIFT) |
-		(INFINIPATH_IBCS_L_STATE_ARM << dd->ibcs_ls_shift);
-	dd->ib_active = (INFINIPATH_IBCS_LT_STATE_LINKUP <<
-		INFINIPATH_IBCS_LINKTRAININGSTATE_SHIFT) |
-		(INFINIPATH_IBCS_L_STATE_ACTIVE << dd->ibcs_ls_shift);
-
-	/*
-	 * Fill in data for ibcc field-values that change in newer chips.
-	 * We dynamically specify only the mask for LINKINITCMD
-	 * and only the shift for LINKCMD and MAXPKTLEN, as they are
-	 * the only ones that change.
-	 */
-	dd->ibcc_lic_mask = INFINIPATH_IBCC_LINKINITCMD_MASK;
-	dd->ibcc_lc_shift = INFINIPATH_IBCC_LINKCMD_SHIFT;
-	dd->ibcc_mpl_shift = INFINIPATH_IBCC_MAXPKTLEN_SHIFT;
-
-	/* Fill in shifts for RcvCtrl. */
-	dd->ipath_r_portenable_shift = INFINIPATH_R_PORTENABLE_SHIFT;
-	dd->ipath_r_intravail_shift = INFINIPATH_R_INTRAVAIL_SHIFT;
-	dd->ipath_r_tailupd_shift = INFINIPATH_R_TAILUPD_SHIFT;
-	dd->ipath_r_portcfg_shift = 0; /* Not on IBA6120 */
-
-	/* variables for sanity checking interrupt and errors */
-	dd->ipath_hwe_bitsextant =
-		(INFINIPATH_HWE_RXEMEMPARITYERR_MASK <<
-		 INFINIPATH_HWE_RXEMEMPARITYERR_SHIFT) |
-		(INFINIPATH_HWE_TXEMEMPARITYERR_MASK <<
-		 INFINIPATH_HWE_TXEMEMPARITYERR_SHIFT) |
-		(INFINIPATH_HWE_PCIEMEMPARITYERR_MASK <<
-		 INFINIPATH_HWE_PCIEMEMPARITYERR_SHIFT) |
-		INFINIPATH_HWE_PCIE1PLLFAILED |
-		INFINIPATH_HWE_PCIE0PLLFAILED |
-		INFINIPATH_HWE_PCIEPOISONEDTLP |
-		INFINIPATH_HWE_PCIECPLTIMEOUT |
-		INFINIPATH_HWE_PCIEBUSPARITYXTLH |
-		INFINIPATH_HWE_PCIEBUSPARITYXADM |
-		INFINIPATH_HWE_PCIEBUSPARITYRADM |
-		INFINIPATH_HWE_MEMBISTFAILED |
-		INFINIPATH_HWE_COREPLL_FBSLIP |
-		INFINIPATH_HWE_COREPLL_RFSLIP |
-		INFINIPATH_HWE_SERDESPLLFAILED |
-		INFINIPATH_HWE_IBCBUSTOSPCPARITYERR |
-		INFINIPATH_HWE_IBCBUSFRSPCPARITYERR;
-	dd->ipath_i_bitsextant =
-		(INFINIPATH_I_RCVURG_MASK << INFINIPATH_I_RCVURG_SHIFT) |
-		(INFINIPATH_I_RCVAVAIL_MASK <<
-		 INFINIPATH_I_RCVAVAIL_SHIFT) |
-		INFINIPATH_I_ERROR | INFINIPATH_I_SPIOSENT |
-		INFINIPATH_I_SPIOBUFAVAIL | INFINIPATH_I_GPIO;
-	dd->ipath_e_bitsextant =
-		INFINIPATH_E_RFORMATERR | INFINIPATH_E_RVCRC |
-		INFINIPATH_E_RICRC | INFINIPATH_E_RMINPKTLEN |
-		INFINIPATH_E_RMAXPKTLEN | INFINIPATH_E_RLONGPKTLEN |
-		INFINIPATH_E_RSHORTPKTLEN | INFINIPATH_E_RUNEXPCHAR |
-		INFINIPATH_E_RUNSUPVL | INFINIPATH_E_REBP |
-		INFINIPATH_E_RIBFLOW | INFINIPATH_E_RBADVERSION |
-		INFINIPATH_E_RRCVEGRFULL | INFINIPATH_E_RRCVHDRFULL |
-		INFINIPATH_E_RBADTID | INFINIPATH_E_RHDRLEN |
-		INFINIPATH_E_RHDR | INFINIPATH_E_RIBLOSTLINK |
-		INFINIPATH_E_SMINPKTLEN | INFINIPATH_E_SMAXPKTLEN |
-		INFINIPATH_E_SUNDERRUN | INFINIPATH_E_SPKTLEN |
-		INFINIPATH_E_SDROPPEDSMPPKT | INFINIPATH_E_SDROPPEDDATAPKT |
-		INFINIPATH_E_SPIOARMLAUNCH | INFINIPATH_E_SUNEXPERRPKTNUM |
-		INFINIPATH_E_SUNSUPVL | INFINIPATH_E_IBSTATUSCHANGED |
-		INFINIPATH_E_INVALIDADDR | INFINIPATH_E_RESET |
-		INFINIPATH_E_HARDWARE;
-
-	dd->ipath_i_rcvavail_mask = INFINIPATH_I_RCVAVAIL_MASK;
-	dd->ipath_i_rcvurg_mask = INFINIPATH_I_RCVURG_MASK;
-	dd->ipath_i_rcvavail_shift = INFINIPATH_I_RCVAVAIL_SHIFT;
-	dd->ipath_i_rcvurg_shift = INFINIPATH_I_RCVURG_SHIFT;
-
-	/*
-	 * EEPROM error log 0 is TXE Parity errors. 1 is RXE Parity.
-	 * 2 is Some Misc, 3 is reserved for future.
-	 */
-	dd->ipath_eep_st_masks[0].hwerrs_to_log =
-		INFINIPATH_HWE_TXEMEMPARITYERR_MASK <<
-		INFINIPATH_HWE_TXEMEMPARITYERR_SHIFT;
-
-	/* Ignore errors in PIO/PBC on systems with unordered write-combining */
-	if (ipath_unordered_wc())
-		dd->ipath_eep_st_masks[0].hwerrs_to_log &= ~TXE_PIO_PARITY;
-
-	dd->ipath_eep_st_masks[1].hwerrs_to_log =
-		INFINIPATH_HWE_RXEMEMPARITYERR_MASK <<
-		INFINIPATH_HWE_RXEMEMPARITYERR_SHIFT;
-
-	dd->ipath_eep_st_masks[2].errs_to_log = INFINIPATH_E_RESET;
-	dd->delay_mult = 2; /* SDR, 4X, can't change */
-}
-
-/* setup the MSI stuff again after a reset.  I'd like to just call
- * pci_enable_msi() and request_irq() again, but when I do that,
- * the MSI enable bit doesn't get set in the command word, and
- * we switch to to a different interrupt vector, which is confusing,
- * so I instead just do it all inline.  Perhaps somehow can tie this
- * into the PCIe hotplug support at some point
- * Note, because I'm doing it all here, I don't call pci_disable_msi()
- * or free_irq() at the start of ipath_setup_pe_reset().
- */
-static int ipath_reinit_msi(struct ipath_devdata *dd)
-{
-	int pos;
-	u16 control;
-	int ret;
-
-	if (!dd->ipath_msi_lo) {
-		dev_info(&dd->pcidev->dev, "Can't restore MSI config, "
-			 "initial setup failed?\n");
-		ret = 0;
-		goto bail;
-	}
-
-	if (!(pos = pci_find_capability(dd->pcidev, PCI_CAP_ID_MSI))) {
-		ipath_dev_err(dd, "Can't find MSI capability, "
-			      "can't restore MSI settings\n");
-		ret = 0;
-		goto bail;
-	}
-	ipath_cdbg(VERBOSE, "Writing msi_lo 0x%x to config offset 0x%x\n",
-		   dd->ipath_msi_lo, pos + PCI_MSI_ADDRESS_LO);
-	pci_write_config_dword(dd->pcidev, pos + PCI_MSI_ADDRESS_LO,
-			       dd->ipath_msi_lo);
-	ipath_cdbg(VERBOSE, "Writing msi_lo 0x%x to config offset 0x%x\n",
-		   dd->ipath_msi_hi, pos + PCI_MSI_ADDRESS_HI);
-	pci_write_config_dword(dd->pcidev, pos + PCI_MSI_ADDRESS_HI,
-			       dd->ipath_msi_hi);
-	pci_read_config_word(dd->pcidev, pos + PCI_MSI_FLAGS, &control);
-	if (!(control & PCI_MSI_FLAGS_ENABLE)) {
-		ipath_cdbg(VERBOSE, "MSI control at off %x was %x, "
-			   "setting MSI enable (%x)\n", pos + PCI_MSI_FLAGS,
-			   control, control | PCI_MSI_FLAGS_ENABLE);
-		control |= PCI_MSI_FLAGS_ENABLE;
-		pci_write_config_word(dd->pcidev, pos + PCI_MSI_FLAGS,
-				      control);
-	}
-	/* now rewrite the data (vector) info */
-	pci_write_config_word(dd->pcidev, pos +
-			      ((control & PCI_MSI_FLAGS_64BIT) ? 12 : 8),
-			      dd->ipath_msi_data);
-	/* we restore the cachelinesize also, although it doesn't really
-	 * matter */
-	pci_write_config_byte(dd->pcidev, PCI_CACHE_LINE_SIZE,
-			      dd->ipath_pci_cacheline);
-	/* and now set the pci master bit again */
-	pci_set_master(dd->pcidev);
-	ret = 1;
-
-bail:
-	return ret;
-}
-
-/* This routine sleeps, so it can only be called from user context, not
- * from interrupt context.  If we need interrupt context, we can split
- * it into two routines.
-*/
-static int ipath_setup_pe_reset(struct ipath_devdata *dd)
-{
-	u64 val;
-	int i;
-	int ret;
-	u16 cmdval;
-
-	pci_read_config_word(dd->pcidev, PCI_COMMAND, &cmdval);
-
-	/* Use ERROR so it shows up in logs, etc. */
-	ipath_dev_err(dd, "Resetting InfiniPath unit %u\n", dd->ipath_unit);
-	/* keep chip from being accessed in a few places */
-	dd->ipath_flags &= ~(IPATH_INITTED|IPATH_PRESENT);
-	val = dd->ipath_control | INFINIPATH_C_RESET;
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_control, val);
-	mb();
-
-	for (i = 1; i <= 5; i++) {
-		int r;
-		/* allow MBIST, etc. to complete; longer on each retry.
-		 * We sometimes get machine checks from bus timeout if no
-		 * response, so for now, make it *really* long.
-		 */
-		msleep(1000 + (1 + i) * 2000);
-		if ((r =
-		     pci_write_config_dword(dd->pcidev, PCI_BASE_ADDRESS_0,
-					    dd->ipath_pcibar0)))
-			ipath_dev_err(dd, "rewrite of BAR0 failed: %d\n",
-				      r);
-		if ((r =
-		     pci_write_config_dword(dd->pcidev, PCI_BASE_ADDRESS_1,
-					    dd->ipath_pcibar1)))
-			ipath_dev_err(dd, "rewrite of BAR1 failed: %d\n",
-				      r);
-		/* now re-enable memory access */
-		pci_write_config_word(dd->pcidev, PCI_COMMAND, cmdval);
-		if ((r = pci_enable_device(dd->pcidev)))
-			ipath_dev_err(dd, "pci_enable_device failed after "
-				      "reset: %d\n", r);
-		/*
-		 * whether it fully enabled or not, mark as present,
-		 * again (but not INITTED)
-		 */
-		dd->ipath_flags |= IPATH_PRESENT;
-		val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_revision);
-		if (val == dd->ipath_revision) {
-			ipath_cdbg(VERBOSE, "Got matching revision "
-				   "register %llx on try %d\n",
-				   (unsigned long long) val, i);
-			ret = ipath_reinit_msi(dd);
-			goto bail;
-		}
-		/* Probably getting -1 back */
-		ipath_dbg("Didn't get expected revision register, "
-			  "got %llx, try %d\n", (unsigned long long) val,
-			  i + 1);
-	}
-	ret = 0; /* failed */
-
-bail:
-	if (ret)
-		ipath_6120_pcie_params(dd);
-	return ret;
-}
-
-/**
- * ipath_pe_put_tid - write a TID in chip
- * @dd: the infinipath device
- * @tidptr: pointer to the expected TID (in chip) to update
- * @tidtype: RCVHQ_RCV_TYPE_EAGER (1) for eager, RCVHQ_RCV_TYPE_EXPECTED (0) for expected
- * @pa: physical address of in memory buffer; ipath_tidinvalid if freeing
- *
- * This exists as a separate routine to allow for special locking etc.
- * It's used for both the full cleanup on exit, as well as the normal
- * setup and teardown.
- */
-static void ipath_pe_put_tid(struct ipath_devdata *dd, u64 __iomem *tidptr,
-			     u32 type, unsigned long pa)
-{
-	u32 __iomem *tidp32 = (u32 __iomem *)tidptr;
-	unsigned long flags = 0; /* keep gcc quiet */
-	int tidx;
-	spinlock_t *tidlockp;
-
-	if (!dd->ipath_kregbase)
-		return;
-
-	if (pa != dd->ipath_tidinvalid) {
-		if (pa & ((1U << 11) - 1)) {
-			dev_info(&dd->pcidev->dev, "BUG: physaddr %lx "
-				 "not 2KB aligned!\n", pa);
-			return;
-		}
-		pa >>= 11;
-		/* paranoia check */
-		if (pa & ~INFINIPATH_RT_ADDR_MASK)
-			ipath_dev_err(dd,
-				      "BUG: Physical page address 0x%lx "
-				      "has bits set in 31-29\n", pa);
-
-		if (type == RCVHQ_RCV_TYPE_EAGER)
-			pa |= dd->ipath_tidtemplate;
-		else /* for now, always full 4KB page */
-			pa |= 2 << 29;
-	}
-
-	/*
-	 * Workaround chip bug 9437 by writing the scratch register
-	 * before and after the TID, and with an io write barrier.
-	 * We use a spinlock around the writes, so they can't intermix
-	 * with other TID (eager or expected) writes (the chip bug
-	 * is triggered by back to back TID writes). Unfortunately, this
-	 * call can be done from interrupt level for the port 0 eager TIDs,
-	 * so we have to use irqsave locks.
-	 */
-	/*
-	 * Assumes tidptr always > ipath_egrtidbase
-	 * if type == RCVHQ_RCV_TYPE_EAGER.
-	 */
-	tidx = tidptr - dd->ipath_egrtidbase;
-
-	tidlockp = (type == RCVHQ_RCV_TYPE_EAGER && tidx < dd->ipath_rcvegrcnt)
-		? &dd->ipath_kernel_tid_lock : &dd->ipath_user_tid_lock;
-	spin_lock_irqsave(tidlockp, flags);
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_scratch, 0xfeeddeaf);
-	writel(pa, tidp32);
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_scratch, 0xdeadbeef);
-	mmiowb();
-	spin_unlock_irqrestore(tidlockp, flags);
-}
-
-/**
- * ipath_pe_put_tid_2 - write a TID in chip, Revision 2 or higher
- * @dd: the infinipath device
- * @tidptr: pointer to the expected TID (in chip) to update
- * @tidtype: RCVHQ_RCV_TYPE_EAGER (1) for eager, RCVHQ_RCV_TYPE_EXPECTED (0) for expected
- * @pa: physical address of in memory buffer; ipath_tidinvalid if freeing
- *
- * This exists as a separate routine to allow for selection of the
- * appropriate "flavor". The static calls in cleanup just use the
- * revision-agnostic form, as they are not performance critical.
- */
-static void ipath_pe_put_tid_2(struct ipath_devdata *dd, u64 __iomem *tidptr,
-			     u32 type, unsigned long pa)
-{
-	u32 __iomem *tidp32 = (u32 __iomem *)tidptr;
-	u32 tidx;
-
-	if (!dd->ipath_kregbase)
-		return;
-
-	if (pa != dd->ipath_tidinvalid) {
-		if (pa & ((1U << 11) - 1)) {
-			dev_info(&dd->pcidev->dev, "BUG: physaddr %lx "
-				 "not 2KB aligned!\n", pa);
-			return;
-		}
-		pa >>= 11;
-		/* paranoia check */
-		if (pa & ~INFINIPATH_RT_ADDR_MASK)
-			ipath_dev_err(dd,
-				      "BUG: Physical page address 0x%lx "
-				      "has bits set in 31-29\n", pa);
-
-		if (type == RCVHQ_RCV_TYPE_EAGER)
-			pa |= dd->ipath_tidtemplate;
-		else /* for now, always full 4KB page */
-			pa |= 2 << 29;
-	}
-	tidx = tidptr - dd->ipath_egrtidbase;
-	writel(pa, tidp32);
-	mmiowb();
-}
-
-
-/**
- * ipath_pe_clear_tid - clear all TID entries for a port, expected and eager
- * @dd: the infinipath device
- * @port: the port
- *
- * clear all TID entries for a port, expected and eager.
- * Used from ipath_close().  On this chip, TIDs are only 32 bits,
- * not 64, but they are still on 64 bit boundaries, so tidbase
- * is declared as u64 * for the pointer math, even though we write 32 bits
- */
-static void ipath_pe_clear_tids(struct ipath_devdata *dd, unsigned port)
-{
-	u64 __iomem *tidbase;
-	unsigned long tidinv;
-	int i;
-
-	if (!dd->ipath_kregbase)
-		return;
-
-	ipath_cdbg(VERBOSE, "Invalidate TIDs for port %u\n", port);
-
-	tidinv = dd->ipath_tidinvalid;
-	tidbase = (u64 __iomem *)
-		((char __iomem *)(dd->ipath_kregbase) +
-		 dd->ipath_rcvtidbase +
-		 port * dd->ipath_rcvtidcnt * sizeof(*tidbase));
-
-	for (i = 0; i < dd->ipath_rcvtidcnt; i++)
-		dd->ipath_f_put_tid(dd, &tidbase[i], RCVHQ_RCV_TYPE_EXPECTED,
-				 tidinv);
-
-	tidbase = (u64 __iomem *)
-		((char __iomem *)(dd->ipath_kregbase) +
-		 dd->ipath_rcvegrbase +
-		 port * dd->ipath_rcvegrcnt * sizeof(*tidbase));
-
-	for (i = 0; i < dd->ipath_rcvegrcnt; i++)
-		dd->ipath_f_put_tid(dd, &tidbase[i], RCVHQ_RCV_TYPE_EAGER,
-				 tidinv);
-}
-
-/**
- * ipath_pe_tidtemplate - setup constants for TID updates
- * @dd: the infinipath device
- *
- * We setup stuff that we use a lot, to avoid calculating each time
- */
-static void ipath_pe_tidtemplate(struct ipath_devdata *dd)
-{
-	u32 egrsize = dd->ipath_rcvegrbufsize;
-
-	/* For now, we always allocate 4KB buffers (at init) so we can
-	 * receive max size packets.  We may want a module parameter to
-	 * specify 2KB or 4KB and/or make be per port instead of per device
-	 * for those who want to reduce memory footprint.  Note that the
-	 * ipath_rcvhdrentsize size must be large enough to hold the largest
-	 * IB header (currently 96 bytes) that we expect to handle (plus of
-	 * course the 2 dwords of RHF).
-	 */
-	if (egrsize == 2048)
-		dd->ipath_tidtemplate = 1U << 29;
-	else if (egrsize == 4096)
-		dd->ipath_tidtemplate = 2U << 29;
-	else {
-		egrsize = 4096;
-		dev_info(&dd->pcidev->dev, "BUG: unsupported egrbufsize "
-			 "%u, using %u\n", dd->ipath_rcvegrbufsize,
-			 egrsize);
-		dd->ipath_tidtemplate = 2U << 29;
-	}
-	dd->ipath_tidinvalid = 0;
-}
-
-static int ipath_pe_early_init(struct ipath_devdata *dd)
-{
-	dd->ipath_flags |= IPATH_4BYTE_TID;
-	if (ipath_unordered_wc())
-		dd->ipath_flags |= IPATH_PIO_FLUSH_WC;
-
-	/*
-	 * For openfabrics, we need to be able to handle an IB header of
-	 * 24 dwords.  HT chip has arbitrary sized receive buffers, so we
-	 * made them the same size as the PIO buffers.  This chip does not
-	 * handle arbitrary size buffers, so we need the header large enough
-	 * to handle largest IB header, but still have room for a 2KB MTU
-	 * standard IB packet.
-	 */
-	dd->ipath_rcvhdrentsize = 24;
-	dd->ipath_rcvhdrsize = IPATH_DFLT_RCVHDRSIZE;
-	dd->ipath_rhf_offset = 0;
-	dd->ipath_egrtidbase = (u64 __iomem *)
-		((char __iomem *) dd->ipath_kregbase + dd->ipath_rcvegrbase);
-
-	dd->ipath_rcvegrbufsize = ipath_mtu4096 ? 4096 : 2048;
-	/*
-	 * the min() check here is currently a nop, but it may not always
-	 * be, depending on just how we do ipath_rcvegrbufsize
-	 */
-	dd->ipath_ibmaxlen = min(ipath_mtu4096 ? dd->ipath_piosize4k :
-				 dd->ipath_piosize2k,
-				 dd->ipath_rcvegrbufsize +
-				 (dd->ipath_rcvhdrentsize << 2));
-	dd->ipath_init_ibmaxlen = dd->ipath_ibmaxlen;
-
-	/*
-	 * We can request a receive interrupt for 1 or
-	 * more packets from current offset.  For now, we set this
-	 * up for a single packet.
-	 */
-	dd->ipath_rhdrhead_intr_off = 1ULL<<32;
-
-	ipath_get_eeprom_info(dd);
-
-	return 0;
-}
-
-int __attribute__((weak)) ipath_unordered_wc(void)
-{
-	return 0;
-}
-
-/**
- * ipath_init_pe_get_base_info - set chip-specific flags for user code
- * @pd: the infinipath port
- * @kbase: ipath_base_info pointer
- *
- * We set the PCIE flag because the lower bandwidth on PCIe vs
- * HyperTransport can affect some user packet algorithms.
- */
-static int ipath_pe_get_base_info(struct ipath_portdata *pd, void *kbase)
-{
-	struct ipath_base_info *kinfo = kbase;
-	struct ipath_devdata *dd;
-
-	if (ipath_unordered_wc()) {
-		kinfo->spi_runtime_flags |= IPATH_RUNTIME_FORCE_WC_ORDER;
-		ipath_cdbg(PROC, "Intel processor, forcing WC order\n");
-	}
-	else
-		ipath_cdbg(PROC, "Not Intel processor, WC ordered\n");
-
-	if (pd == NULL)
-		goto done;
-
-	dd = pd->port_dd;
-
-done:
-	kinfo->spi_runtime_flags |= IPATH_RUNTIME_PCIE |
-		IPATH_RUNTIME_FORCE_PIOAVAIL | IPATH_RUNTIME_PIO_REGSWAPPED;
-	return 0;
-}
-
-static void ipath_pe_free_irq(struct ipath_devdata *dd)
-{
-	free_irq(dd->ipath_irq, dd);
-	dd->ipath_irq = 0;
-}
-
-
-static struct ipath_message_header *
-ipath_pe_get_msgheader(struct ipath_devdata *dd, __le32 *rhf_addr)
-{
-	return (struct ipath_message_header *)
-		&rhf_addr[sizeof(u64) / sizeof(u32)];
-}
-
-static void ipath_pe_config_ports(struct ipath_devdata *dd, ushort cfgports)
-{
-	dd->ipath_portcnt =
-		ipath_read_kreg32(dd, dd->ipath_kregs->kr_portcnt);
-	dd->ipath_p0_rcvegrcnt =
-		ipath_read_kreg32(dd, dd->ipath_kregs->kr_rcvegrcnt);
-}
-
-static void ipath_pe_read_counters(struct ipath_devdata *dd,
-				   struct infinipath_counters *cntrs)
-{
-	cntrs->LBIntCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(LBIntCnt));
-	cntrs->LBFlowStallCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(LBFlowStallCnt));
-	cntrs->TxSDmaDescCnt = 0;
-	cntrs->TxUnsupVLErrCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxUnsupVLErrCnt));
-	cntrs->TxDataPktCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxDataPktCnt));
-	cntrs->TxFlowPktCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxFlowPktCnt));
-	cntrs->TxDwordCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxDwordCnt));
-	cntrs->TxLenErrCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxLenErrCnt));
-	cntrs->TxMaxMinLenErrCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxMaxMinLenErrCnt));
-	cntrs->TxUnderrunCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxUnderrunCnt));
-	cntrs->TxFlowStallCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxFlowStallCnt));
-	cntrs->TxDroppedPktCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(TxDroppedPktCnt));
-	cntrs->RxDroppedPktCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxDroppedPktCnt));
-	cntrs->RxDataPktCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxDataPktCnt));
-	cntrs->RxFlowPktCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxFlowPktCnt));
-	cntrs->RxDwordCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxDwordCnt));
-	cntrs->RxLenErrCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxLenErrCnt));
-	cntrs->RxMaxMinLenErrCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxMaxMinLenErrCnt));
-	cntrs->RxICRCErrCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxICRCErrCnt));
-	cntrs->RxVCRCErrCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxVCRCErrCnt));
-	cntrs->RxFlowCtrlErrCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxFlowCtrlErrCnt));
-	cntrs->RxBadFormatCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxBadFormatCnt));
-	cntrs->RxLinkProblemCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxLinkProblemCnt));
-	cntrs->RxEBPCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxEBPCnt));
-	cntrs->RxLPCRCErrCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxLPCRCErrCnt));
-	cntrs->RxBufOvflCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxBufOvflCnt));
-	cntrs->RxTIDFullErrCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxTIDFullErrCnt));
-	cntrs->RxTIDValidErrCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxTIDValidErrCnt));
-	cntrs->RxPKeyMismatchCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxPKeyMismatchCnt));
-	cntrs->RxP0HdrEgrOvflCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxP0HdrEgrOvflCnt));
-	cntrs->RxP1HdrEgrOvflCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxP1HdrEgrOvflCnt));
-	cntrs->RxP2HdrEgrOvflCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxP2HdrEgrOvflCnt));
-	cntrs->RxP3HdrEgrOvflCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxP3HdrEgrOvflCnt));
-	cntrs->RxP4HdrEgrOvflCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(RxP4HdrEgrOvflCnt));
-	cntrs->RxP5HdrEgrOvflCnt = 0;
-	cntrs->RxP6HdrEgrOvflCnt = 0;
-	cntrs->RxP7HdrEgrOvflCnt = 0;
-	cntrs->RxP8HdrEgrOvflCnt = 0;
-	cntrs->RxP9HdrEgrOvflCnt = 0;
-	cntrs->RxP10HdrEgrOvflCnt = 0;
-	cntrs->RxP11HdrEgrOvflCnt = 0;
-	cntrs->RxP12HdrEgrOvflCnt = 0;
-	cntrs->RxP13HdrEgrOvflCnt = 0;
-	cntrs->RxP14HdrEgrOvflCnt = 0;
-	cntrs->RxP15HdrEgrOvflCnt = 0;
-	cntrs->RxP16HdrEgrOvflCnt = 0;
-	cntrs->IBStatusChangeCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(IBStatusChangeCnt));
-	cntrs->IBLinkErrRecoveryCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(IBLinkErrRecoveryCnt));
-	cntrs->IBLinkDownedCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(IBLinkDownedCnt));
-	cntrs->IBSymbolErrCnt =
-		ipath_snap_cntr(dd, IPATH_CREG_OFFSET(IBSymbolErrCnt));
-	cntrs->RxVL15DroppedPktCnt = 0;
-	cntrs->RxOtherLocalPhyErrCnt = 0;
-	cntrs->PcieRetryBufDiagQwordCnt = 0;
-	cntrs->ExcessBufferOvflCnt = dd->ipath_overrun_thresh_errs;
-	cntrs->LocalLinkIntegrityErrCnt = dd->ipath_lli_errs;
-	cntrs->RxVlErrCnt = 0;
-	cntrs->RxDlidFltrCnt = 0;
-}
-
-
-/* no interrupt fallback for these chips */
-static int ipath_pe_nointr_fallback(struct ipath_devdata *dd)
-{
-	return 0;
-}
-
-
-/*
- * reset the XGXS (between serdes and IBC).  Slightly less intrusive
- * than resetting the IBC or external link state, and useful in some
- * cases to cause some retraining.  To do this right, we reset IBC
- * as well.
- */
-static void ipath_pe_xgxs_reset(struct ipath_devdata *dd)
-{
-	u64 val, prev_val;
-
-	prev_val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_xgxsconfig);
-	val = prev_val | INFINIPATH_XGXS_RESET;
-	prev_val &= ~INFINIPATH_XGXS_RESET; /* be sure */
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_control,
-			 dd->ipath_control & ~INFINIPATH_C_LINKENABLE);
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_xgxsconfig, val);
-	ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_xgxsconfig, prev_val);
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_control,
-			 dd->ipath_control);
-}
-
-
-static int ipath_pe_get_ib_cfg(struct ipath_devdata *dd, int which)
-{
-	int ret;
-
-	switch (which) {
-	case IPATH_IB_CFG_LWID:
-		ret = dd->ipath_link_width_active;
-		break;
-	case IPATH_IB_CFG_SPD:
-		ret = dd->ipath_link_speed_active;
-		break;
-	case IPATH_IB_CFG_LWID_ENB:
-		ret = dd->ipath_link_width_enabled;
-		break;
-	case IPATH_IB_CFG_SPD_ENB:
-		ret = dd->ipath_link_speed_enabled;
-		break;
-	default:
-		ret =  -ENOTSUPP;
-		break;
-	}
-	return ret;
-}
-
-
-/* we assume range checking is already done, if needed */
-static int ipath_pe_set_ib_cfg(struct ipath_devdata *dd, int which, u32 val)
-{
-	int ret = 0;
-
-	if (which == IPATH_IB_CFG_LWID_ENB)
-		dd->ipath_link_width_enabled = val;
-	else if (which == IPATH_IB_CFG_SPD_ENB)
-		dd->ipath_link_speed_enabled = val;
-	else
-		ret = -ENOTSUPP;
-	return ret;
-}
-
-static void ipath_pe_config_jint(struct ipath_devdata *dd, u16 a, u16 b)
-{
-}
-
-
-static int ipath_pe_ib_updown(struct ipath_devdata *dd, int ibup, u64 ibcs)
-{
-	if (ibup) {
-		if (dd->ibdeltainprog) {
-			dd->ibdeltainprog = 0;
-			dd->ibsymdelta +=
-				ipath_read_creg32(dd,
-				  dd->ipath_cregs->cr_ibsymbolerrcnt) -
-				dd->ibsymsnap;
-			dd->iblnkerrdelta +=
-				ipath_read_creg32(dd,
-				  dd->ipath_cregs->cr_iblinkerrrecovcnt) -
-				dd->iblnkerrsnap;
-		}
-	} else {
-		dd->ipath_lli_counter = 0;
-		if (!dd->ibdeltainprog) {
-			dd->ibdeltainprog = 1;
-			dd->ibsymsnap =
-				ipath_read_creg32(dd,
-				  dd->ipath_cregs->cr_ibsymbolerrcnt);
-			dd->iblnkerrsnap =
-				ipath_read_creg32(dd,
-				  dd->ipath_cregs->cr_iblinkerrrecovcnt);
-		}
-	}
-
-	ipath_setup_pe_setextled(dd, ipath_ib_linkstate(dd, ibcs),
-		ipath_ib_linktrstate(dd, ibcs));
-	return 0;
-}
-
-
-/**
- * ipath_init_iba6120_funcs - set up the chip-specific function pointers
- * @dd: the infinipath device
- *
- * This is global, and is called directly at init to set up the
- * chip-specific function pointers for later use.
- */
-void ipath_init_iba6120_funcs(struct ipath_devdata *dd)
-{
-	dd->ipath_f_intrsetup = ipath_pe_intconfig;
-	dd->ipath_f_bus = ipath_setup_pe_config;
-	dd->ipath_f_reset = ipath_setup_pe_reset;
-	dd->ipath_f_get_boardname = ipath_pe_boardname;
-	dd->ipath_f_init_hwerrors = ipath_pe_init_hwerrors;
-	dd->ipath_f_early_init = ipath_pe_early_init;
-	dd->ipath_f_handle_hwerrors = ipath_pe_handle_hwerrors;
-	dd->ipath_f_quiet_serdes = ipath_pe_quiet_serdes;
-	dd->ipath_f_bringup_serdes = ipath_pe_bringup_serdes;
-	dd->ipath_f_clear_tids = ipath_pe_clear_tids;
-	/*
-	 * _f_put_tid may get changed after we read the chip revision,
-	 * but we start with the safe version for all revs
-	 */
-	dd->ipath_f_put_tid = ipath_pe_put_tid;
-	dd->ipath_f_cleanup = ipath_setup_pe_cleanup;
-	dd->ipath_f_setextled = ipath_setup_pe_setextled;
-	dd->ipath_f_get_base_info = ipath_pe_get_base_info;
-	dd->ipath_f_free_irq = ipath_pe_free_irq;
-	dd->ipath_f_tidtemplate = ipath_pe_tidtemplate;
-	dd->ipath_f_intr_fallback = ipath_pe_nointr_fallback;
-	dd->ipath_f_xgxs_reset = ipath_pe_xgxs_reset;
-	dd->ipath_f_get_msgheader = ipath_pe_get_msgheader;
-	dd->ipath_f_config_ports = ipath_pe_config_ports;
-	dd->ipath_f_read_counters = ipath_pe_read_counters;
-	dd->ipath_f_get_ib_cfg = ipath_pe_get_ib_cfg;
-	dd->ipath_f_set_ib_cfg = ipath_pe_set_ib_cfg;
-	dd->ipath_f_config_jint = ipath_pe_config_jint;
-	dd->ipath_f_ib_updown = ipath_pe_ib_updown;
-
-
-	/* initialize chip-specific variables */
-	ipath_init_pe_variables(dd);
-}
-
diff --git a/drivers/infiniband/hw/ipath/ipath_iba7220.c b/drivers/infiniband/hw/ipath/ipath_iba7220.c
deleted file mode 100644
index 34b778ed97fc..000000000000
--- a/drivers/infiniband/hw/ipath/ipath_iba7220.c
+++ /dev/null
@@ -1,2631 +0,0 @@
-/*
- * Copyright (c) 2006, 2007, 2008 QLogic Corporation. All rights reserved.
- * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
- *
- * This software is available to you under a choice of one of two
- * licenses.  You may choose to be licensed under the terms of the GNU
- * General Public License (GPL) Version 2, available from the file
- * COPYING in the main directory of this source tree, or the
- * OpenIB.org BSD license below:
- *
- *     Redistribution and use in source and binary forms, with or
- *     without modification, are permitted provided that the following
- *     conditions are met:
- *
- *      - Redistributions of source code must retain the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer.
- *
- *      - Redistributions in binary form must reproduce the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer in the documentation and/or other materials
- *        provided with the distribution.
- *
- * 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.
- */
-/*
- * This file contains all of the code that is specific to the
- * InfiniPath 7220 chip (except that specific to the SerDes)
- */
-
-#include <linux/interrupt.h>
-#include <linux/pci.h>
-#include <linux/sched.h>
-#include <linux/delay.h>
-#include <linux/io.h>
-#include <rdma/ib_verbs.h>
-
-#include "ipath_kernel.h"
-#include "ipath_registers.h"
-#include "ipath_7220.h"
-
-static void ipath_setup_7220_setextled(struct ipath_devdata *, u64, u64);
-
-static unsigned ipath_compat_ddr_negotiate = 1;
-
-module_param_named(compat_ddr_negotiate, ipath_compat_ddr_negotiate, uint,
-			S_IWUSR | S_IRUGO);
-MODULE_PARM_DESC(compat_ddr_negotiate,
-		"Attempt pre-IBTA 1.2 DDR speed negotiation");
-
-static unsigned ipath_sdma_fetch_arb = 1;
-module_param_named(fetch_arb, ipath_sdma_fetch_arb, uint, S_IRUGO);
-MODULE_PARM_DESC(fetch_arb, "IBA7220: change SDMA descriptor arbitration");
-
-/*
- * This file contains almost all the chip-specific register information and
- * access functions for the QLogic InfiniPath 7220 PCI-Express chip, with the
- * exception of SerDes support, which in in ipath_sd7220.c.
- *
- * This lists the InfiniPath registers, in the actual chip layout.
- * This structure should never be directly accessed.
- */
-struct _infinipath_do_not_use_kernel_regs {
-	unsigned long long Revision;
-	unsigned long long Control;
-	unsigned long long PageAlign;
-	unsigned long long PortCnt;
-	unsigned long long DebugPortSelect;
-	unsigned long long DebugSigsIntSel; /* was Reserved0;*/
-	unsigned long long SendRegBase;
-	unsigned long long UserRegBase;
-	unsigned long long CounterRegBase;
-	unsigned long long Scratch;
-	unsigned long long EEPROMAddrCmd; /* was Reserved1; */
-	unsigned long long EEPROMData; /* was Reserved2; */
-	unsigned long long IntBlocked;
-	unsigned long long IntMask;
-	unsigned long long IntStatus;
-	unsigned long long IntClear;
-	unsigned long long ErrorMask;
-	unsigned long long ErrorStatus;
-	unsigned long long ErrorClear;
-	unsigned long long HwErrMask;
-	unsigned long long HwErrStatus;
-	unsigned long long HwErrClear;
-	unsigned long long HwDiagCtrl;
-	unsigned long long MDIO;
-	unsigned long long IBCStatus;
-	unsigned long long IBCCtrl;
-	unsigned long long ExtStatus;
-	unsigned long long ExtCtrl;
-	unsigned long long GPIOOut;
-	unsigned long long GPIOMask;
-	unsigned long long GPIOStatus;
-	unsigned long long GPIOClear;
-	unsigned long long RcvCtrl;
-	unsigned long long RcvBTHQP;
-	unsigned long long RcvHdrSize;
-	unsigned long long RcvHdrCnt;
-	unsigned long long RcvHdrEntSize;
-	unsigned long long RcvTIDBase;
-	unsigned long long RcvTIDCnt;
-	unsigned long long RcvEgrBase;
-	unsigned long long RcvEgrCnt;
-	unsigned long long RcvBufBase;
-	unsigned long long RcvBufSize;
-	unsigned long long RxIntMemBase;
-	unsigned long long RxIntMemSize;
-	unsigned long long RcvPartitionKey;
-	unsigned long long RcvQPMulticastPort;
-	unsigned long long RcvPktLEDCnt;
-	unsigned long long IBCDDRCtrl;
-	unsigned long long HRTBT_GUID;
-	unsigned long long IB_SDTEST_IF_TX;
-	unsigned long long IB_SDTEST_IF_RX;
-	unsigned long long IBCDDRCtrl2;
-	unsigned long long IBCDDRStatus;
-	unsigned long long JIntReload;
-	unsigned long long IBNCModeCtrl;
-	unsigned long long SendCtrl;
-	unsigned long long SendBufBase;
-	unsigned long long SendBufSize;
-	unsigned long long SendBufCnt;
-	unsigned long long SendAvailAddr;
-	unsigned long long TxIntMemBase;
-	unsigned long long TxIntMemSize;
-	unsigned long long SendDmaBase;
-	unsigned long long SendDmaLenGen;
-	unsigned long long SendDmaTail;
-	unsigned long long SendDmaHead;
-	unsigned long long SendDmaHeadAddr;
-	unsigned long long SendDmaBufMask0;
-	unsigned long long SendDmaBufMask1;
-	unsigned long long SendDmaBufMask2;
-	unsigned long long SendDmaStatus;
-	unsigned long long SendBufferError;
-	unsigned long long SendBufferErrorCONT1;
-	unsigned long long SendBufErr2; /* was Reserved6SBE[0/6] */
-	unsigned long long Reserved6L[2];
-	unsigned long long AvailUpdCount;
-	unsigned long long RcvHdrAddr0;
-	unsigned long long RcvHdrAddrs[16]; /* Why enumerate? */
-	unsigned long long Reserved7hdtl; /* Align next to 300 */
-	unsigned long long RcvHdrTailAddr0; /* 300, like others */
-	unsigned long long RcvHdrTailAddrs[16];
-	unsigned long long Reserved9SW[7]; /* was [8]; we have 17 ports */
-	unsigned long long IbsdEpbAccCtl; /* IB Serdes EPB access control */
-	unsigned long long IbsdEpbTransReg; /* IB Serdes EPB Transaction */
-	unsigned long long Reserved10sds; /* was SerdesStatus on */
-	unsigned long long XGXSConfig;
-	unsigned long long IBSerDesCtrl; /* Was IBPLLCfg on Monty */
-	unsigned long long EEPCtlStat; /* for "boot" EEPROM/FLASH */
-	unsigned long long EEPAddrCmd;
-	unsigned long long EEPData;
-	unsigned long long PcieEpbAccCtl;
-	unsigned long long PcieEpbTransCtl;
-	unsigned long long EfuseCtl; /* E-Fuse control */
-	unsigned long long EfuseData[4];
-	unsigned long long ProcMon;
-	/* this chip moves following two from previous 200, 208 */
-	unsigned long long PCIeRBufTestReg0;
-	unsigned long long PCIeRBufTestReg1;
-	/* added for this chip */
-	unsigned long long PCIeRBufTestReg2;
-	unsigned long long PCIeRBufTestReg3;
-	/* added for this chip, debug only */
-	unsigned long long SPC_JTAG_ACCESS_REG;
-	unsigned long long LAControlReg;
-	unsigned long long GPIODebugSelReg;
-	unsigned long long DebugPortValueReg;
-	/* added for this chip, DMA */
-	unsigned long long SendDmaBufUsed[3];
-	unsigned long long SendDmaReqTagUsed;
-	/*
-	 * added for this chip, EFUSE: note that these program 64-bit
-	 * words 2 and 3 */
-	unsigned long long efuse_pgm_data[2];
-	unsigned long long Reserved11LAalign[10]; /* Skip 4B0..4F8 */
-	/* we have 30 regs for DDS and RXEQ in IB SERDES */
-	unsigned long long SerDesDDSRXEQ[30];
-	unsigned long long Reserved12LAalign[2]; /* Skip 5F0, 5F8 */
-	/* added for LA debug support */
-	unsigned long long LAMemory[32];
-};
-
-struct _infinipath_do_not_use_counters {
-	__u64 LBIntCnt;
-	__u64 LBFlowStallCnt;
-	__u64 TxSDmaDescCnt;	/* was Reserved1 */
-	__u64 TxUnsupVLErrCnt;
-	__u64 TxDataPktCnt;
-	__u64 TxFlowPktCnt;
-	__u64 TxDwordCnt;
-	__u64 TxLenErrCnt;
-	__u64 TxMaxMinLenErrCnt;
-	__u64 TxUnderrunCnt;
-	__u64 TxFlowStallCnt;
-	__u64 TxDroppedPktCnt;
-	__u64 RxDroppedPktCnt;
-	__u64 RxDataPktCnt;
-	__u64 RxFlowPktCnt;
-	__u64 RxDwordCnt;
-	__u64 RxLenErrCnt;
-	__u64 RxMaxMinLenErrCnt;
-	__u64 RxICRCErrCnt;
-	__u64 RxVCRCErrCnt;
-	__u64 RxFlowCtrlErrCnt;
-	__u64 RxBadFormatCnt;
-	__u64 RxLinkProblemCnt;
-	__u64 RxEBPCnt;
-	__u64 RxLPCRCErrCnt;
-	__u64 RxBufOvflCnt;
-	__u64 RxTIDFullErrCnt;
-	__u64 RxTIDValidErrCnt;
-	__u64 RxPKeyMismatchCnt;
-	__u64 RxP0HdrEgrOvflCnt;
-	__u64 RxP1HdrEgrOvflCnt;
-	__u64 RxP2HdrEgrOvflCnt;
-	__u64 RxP3HdrEgrOvflCnt;
-	__u64 RxP4HdrEgrOvflCnt;
-	__u64 RxP5HdrEgrOvflCnt;
-	__u64 RxP6HdrEgrOvflCnt;
-	__u64 RxP7HdrEgrOvflCnt;
-	__u64 RxP8HdrEgrOvflCnt;
-	__u64 RxP9HdrEgrOvflCnt;	/* was Reserved6 */
-	__u64 RxP10HdrEgrOvflCnt;	/* was Reserved7 */
-	__u64 RxP11HdrEgrOvflCnt;	/* new for IBA7220 */
-	__u64 RxP12HdrEgrOvflCnt;	/* new for IBA7220 */
-	__u64 RxP13HdrEgrOvflCnt;	/* new for IBA7220 */
-	__u64 RxP14HdrEgrOvflCnt;	/* new for IBA7220 */
-	__u64 RxP15HdrEgrOvflCnt;	/* new for IBA7220 */
-	__u64 RxP16HdrEgrOvflCnt;	/* new for IBA7220 */
-	__u64 IBStatusChangeCnt;
-	__u64 IBLinkErrRecoveryCnt;
-	__u64 IBLinkDownedCnt;
-	__u64 IBSymbolErrCnt;
-	/* The following are new for IBA7220 */
-	__u64 RxVL15DroppedPktCnt;
-	__u64 RxOtherLocalPhyErrCnt;
-	__u64 PcieRetryBufDiagQwordCnt;
-	__u64 ExcessBufferOvflCnt;
-	__u64 LocalLinkIntegrityErrCnt;
-	__u64 RxVlErrCnt;
-	__u64 RxDlidFltrCnt;
-	__u64 Reserved8[7];
-	__u64 PSStat;
-	__u64 PSStart;
-	__u64 PSInterval;
-	__u64 PSRcvDataCount;
-	__u64 PSRcvPktsCount;
-	__u64 PSXmitDataCount;
-	__u64 PSXmitPktsCount;
-	__u64 PSXmitWaitCount;
-};
-
-#define IPATH_KREG_OFFSET(field) (offsetof( \
-	struct _infinipath_do_not_use_kernel_regs, field) / sizeof(u64))
-#define IPATH_CREG_OFFSET(field) (offsetof( \
-	struct _infinipath_do_not_use_counters, field) / sizeof(u64))
-
-static const struct ipath_kregs ipath_7220_kregs = {
-	.kr_control = IPATH_KREG_OFFSET(Control),
-	.kr_counterregbase = IPATH_KREG_OFFSET(CounterRegBase),
-	.kr_debugportselect = IPATH_KREG_OFFSET(DebugPortSelect),
-	.kr_errorclear = IPATH_KREG_OFFSET(ErrorClear),
-	.kr_errormask = IPATH_KREG_OFFSET(ErrorMask),
-	.kr_errorstatus = IPATH_KREG_OFFSET(ErrorStatus),
-	.kr_extctrl = IPATH_KREG_OFFSET(ExtCtrl),
-	.kr_extstatus = IPATH_KREG_OFFSET(ExtStatus),
-	.kr_gpio_clear = IPATH_KREG_OFFSET(GPIOClear),
-	.kr_gpio_mask = IPATH_KREG_OFFSET(GPIOMask),
-	.kr_gpio_out = IPATH_KREG_OFFSET(GPIOOut),
-	.kr_gpio_status = IPATH_KREG_OFFSET(GPIOStatus),
-	.kr_hwdiagctrl = IPATH_KREG_OFFSET(HwDiagCtrl),
-	.kr_hwerrclear = IPATH_KREG_OFFSET(HwErrClear),
-	.kr_hwerrmask = IPATH_KREG_OFFSET(HwErrMask),
-	.kr_hwerrstatus = IPATH_KREG_OFFSET(HwErrStatus),
-	.kr_ibcctrl = IPATH_KREG_OFFSET(IBCCtrl),
-	.kr_ibcstatus = IPATH_KREG_OFFSET(IBCStatus),
-	.kr_intblocked = IPATH_KREG_OFFSET(IntBlocked),
-	.kr_intclear = IPATH_KREG_OFFSET(IntClear),
-	.kr_intmask = IPATH_KREG_OFFSET(IntMask),
-	.kr_intstatus = IPATH_KREG_OFFSET(IntStatus),
-	.kr_mdio = IPATH_KREG_OFFSET(MDIO),
-	.kr_pagealign = IPATH_KREG_OFFSET(PageAlign),
-	.kr_partitionkey = IPATH_KREG_OFFSET(RcvPartitionKey),
-	.kr_portcnt = IPATH_KREG_OFFSET(PortCnt),
-	.kr_rcvbthqp = IPATH_KREG_OFFSET(RcvBTHQP),
-	.kr_rcvbufbase = IPATH_KREG_OFFSET(RcvBufBase),
-	.kr_rcvbufsize = IPATH_KREG_OFFSET(RcvBufSize),
-	.kr_rcvctrl = IPATH_KREG_OFFSET(RcvCtrl),
-	.kr_rcvegrbase = IPATH_KREG_OFFSET(RcvEgrBase),
-	.kr_rcvegrcnt = IPATH_KREG_OFFSET(RcvEgrCnt),
-	.kr_rcvhdrcnt = IPATH_KREG_OFFSET(RcvHdrCnt),
-	.kr_rcvhdrentsize = IPATH_KREG_OFFSET(RcvHdrEntSize),
-	.kr_rcvhdrsize = IPATH_KREG_OFFSET(RcvHdrSize),
-	.kr_rcvintmembase = IPATH_KREG_OFFSET(RxIntMemBase),
-	.kr_rcvintmemsize = IPATH_KREG_OFFSET(RxIntMemSize),
-	.kr_rcvtidbase = IPATH_KREG_OFFSET(RcvTIDBase),
-	.kr_rcvtidcnt = IPATH_KREG_OFFSET(RcvTIDCnt),
-	.kr_revision = IPATH_KREG_OFFSET(Revision),
-	.kr_scratch = IPATH_KREG_OFFSET(Scratch),
-	.kr_sendbuffererror = IPATH_KREG_OFFSET(SendBufferError),
-	.kr_sendctrl = IPATH_KREG_OFFSET(SendCtrl),
-	.kr_sendpioavailaddr = IPATH_KREG_OFFSET(SendAvailAddr),
-	.kr_sendpiobufbase = IPATH_KREG_OFFSET(SendBufBase),
-	.kr_sendpiobufcnt = IPATH_KREG_OFFSET(SendBufCnt),
-	.kr_sendpiosize = IPATH_KREG_OFFSET(SendBufSize),
-	.kr_sendregbase = IPATH_KREG_OFFSET(SendRegBase),
-	.kr_txintmembase = IPATH_KREG_OFFSET(TxIntMemBase),
-	.kr_txintmemsize = IPATH_KREG_OFFSET(TxIntMemSize),
-	.kr_userregbase = IPATH_KREG_OFFSET(UserRegBase),
-
-	.kr_xgxsconfig = IPATH_KREG_OFFSET(XGXSConfig),
-
-	/* send dma related regs */
-	.kr_senddmabase = IPATH_KREG_OFFSET(SendDmaBase),
-	.kr_senddmalengen = IPATH_KREG_OFFSET(SendDmaLenGen),
-	.kr_senddmatail = IPATH_KREG_OFFSET(SendDmaTail),
-	.kr_senddmahead = IPATH_KREG_OFFSET(SendDmaHead),
-	.kr_senddmaheadaddr = IPATH_KREG_OFFSET(SendDmaHeadAddr),
-	.kr_senddmabufmask0 = IPATH_KREG_OFFSET(SendDmaBufMask0),
-	.kr_senddmabufmask1 = IPATH_KREG_OFFSET(SendDmaBufMask1),
-	.kr_senddmabufmask2 = IPATH_KREG_OFFSET(SendDmaBufMask2),
-	.kr_senddmastatus = IPATH_KREG_OFFSET(SendDmaStatus),
-
-	/* SerDes related regs */
-	.kr_ibserdesctrl = IPATH_KREG_OFFSET(IBSerDesCtrl),
-	.kr_ib_epbacc = IPATH_KREG_OFFSET(IbsdEpbAccCtl),
-	.kr_ib_epbtrans = IPATH_KREG_OFFSET(IbsdEpbTransReg),
-	.kr_pcie_epbacc = IPATH_KREG_OFFSET(PcieEpbAccCtl),
-	.kr_pcie_epbtrans = IPATH_KREG_OFFSET(PcieEpbTransCtl),
-	.kr_ib_ddsrxeq = IPATH_KREG_OFFSET(SerDesDDSRXEQ),
-
-	/*
-	 * These should not be used directly via ipath_read_kreg64(),
-	 * use them with ipath_read_kreg64_port()
-	 */
-	.kr_rcvhdraddr = IPATH_KREG_OFFSET(RcvHdrAddr0),
-	.kr_rcvhdrtailaddr = IPATH_KREG_OFFSET(RcvHdrTailAddr0),
-
-	/*
-	 * The rcvpktled register controls one of the debug port signals, so
-	 * a packet activity LED can be connected to it.
-	 */
-	.kr_rcvpktledcnt = IPATH_KREG_OFFSET(RcvPktLEDCnt),
-	.kr_pcierbuftestreg0 = IPATH_KREG_OFFSET(PCIeRBufTestReg0),
-	.kr_pcierbuftestreg1 = IPATH_KREG_OFFSET(PCIeRBufTestReg1),
-
-	.kr_hrtbt_guid = IPATH_KREG_OFFSET(HRTBT_GUID),
-	.kr_ibcddrctrl = IPATH_KREG_OFFSET(IBCDDRCtrl),
-	.kr_ibcddrstatus = IPATH_KREG_OFFSET(IBCDDRStatus),
-	.kr_jintreload = IPATH_KREG_OFFSET(JIntReload)
-};
-
-static const struct ipath_cregs ipath_7220_cregs = {
-	.cr_badformatcnt = IPATH_CREG_OFFSET(RxBadFormatCnt),
-	.cr_erricrccnt = IPATH_CREG_OFFSET(RxICRCErrCnt),
-	.cr_errlinkcnt = IPATH_CREG_OFFSET(RxLinkProblemCnt),
-	.cr_errlpcrccnt = IPATH_CREG_OFFSET(RxLPCRCErrCnt),
-	.cr_errpkey = IPATH_CREG_OFFSET(RxPKeyMismatchCnt),
-	.cr_errrcvflowctrlcnt = IPATH_CREG_OFFSET(RxFlowCtrlErrCnt),
-	.cr_err_rlencnt = IPATH_CREG_OFFSET(RxLenErrCnt),
-	.cr_errslencnt = IPATH_CREG_OFFSET(TxLenErrCnt),
-	.cr_errtidfull = IPATH_CREG_OFFSET(RxTIDFullErrCnt),
-	.cr_errtidvalid = IPATH_CREG_OFFSET(RxTIDValidErrCnt),
-	.cr_errvcrccnt = IPATH_CREG_OFFSET(RxVCRCErrCnt),
-	.cr_ibstatuschange = IPATH_CREG_OFFSET(IBStatusChangeCnt),
-	.cr_intcnt = IPATH_CREG_OFFSET(LBIntCnt),
-	.cr_invalidrlencnt = IPATH_CREG_OFFSET(RxMaxMinLenErrCnt),
-	.cr_invalidslencnt = IPATH_CREG_OFFSET(TxMaxMinLenErrCnt),
-	.cr_lbflowstallcnt = IPATH_CREG_OFFSET(LBFlowStallCnt),
-	.cr_pktrcvcnt = IPATH_CREG_OFFSET(RxDataPktCnt),
-	.cr_pktrcvflowctrlcnt = IPATH_CREG_OFFSET(RxFlowPktCnt),
-	.cr_pktsendcnt = IPATH_CREG_OFFSET(TxDataPktCnt),
-	.cr_pktsendflowcnt = IPATH_CREG_OFFSET(TxFlowPktCnt),
-	.cr_portovflcnt = IPATH_CREG_OFFSET(RxP0HdrEgrOvflCnt),
-	.cr_rcvebpcnt = IPATH_CREG_OFFSET(RxEBPCnt),
-	.cr_rcvovflcnt = IPATH_CREG_OFFSET(RxBufOvflCnt),
-	.cr_senddropped = IPATH_CREG_OFFSET(TxDroppedPktCnt),
-	.cr_sendstallcnt = IPATH_CREG_OFFSET(TxFlowStallCnt),
-	.cr_sendunderruncnt = IPATH_CREG_OFFSET(TxUnderrunCnt),
-	.cr_wordrcvcnt = IPATH_CREG_OFFSET(RxDwordCnt),
-	.cr_wordsendcnt = IPATH_CREG_OFFSET(TxDwordCnt),
-	.cr_unsupvlcnt = IPATH_CREG_OFFSET(TxUnsupVLErrCnt),
-	.cr_rxdroppktcnt = IPATH_CREG_OFFSET(RxDroppedPktCnt),
-	.cr_iblinkerrrecovcnt = IPATH_CREG_OFFSET(IBLinkErrRecoveryCnt),
-	.cr_iblinkdowncnt = IPATH_CREG_OFFSET(IBLinkDownedCnt),
-	.cr_ibsymbolerrcnt = IPATH_CREG_OFFSET(IBSymbolErrCnt),
-	.cr_vl15droppedpktcnt = IPATH_CREG_OFFSET(RxVL15DroppedPktCnt),
-	.cr_rxotherlocalphyerrcnt =
-		IPATH_CREG_OFFSET(RxOtherLocalPhyErrCnt),
-	.cr_excessbufferovflcnt = IPATH_CREG_OFFSET(ExcessBufferOvflCnt),
-	.cr_locallinkintegrityerrcnt =
-		IPATH_CREG_OFFSET(LocalLinkIntegrityErrCnt),
-	.cr_rxvlerrcnt = IPATH_CREG_OFFSET(RxVlErrCnt),
-	.cr_rxdlidfltrcnt = IPATH_CREG_OFFSET(RxDlidFltrCnt),
-	.cr_psstat = IPATH_CREG_OFFSET(PSStat),
-	.cr_psstart = IPATH_CREG_OFFSET(PSStart),
-	.cr_psinterval = IPATH_CREG_OFFSET(PSInterval),
-	.cr_psrcvdatacount = IPATH_CREG_OFFSET(PSRcvDataCount),
-	.cr_psrcvpktscount = IPATH_CREG_OFFSET(PSRcvPktsCount),
-	.cr_psxmitdatacount = IPATH_CREG_OFFSET(PSXmitDataCount),
-	.cr_psxmitpktscount = IPATH_CREG_OFFSET(PSXmitPktsCount),
-	.cr_psxmitwaitcount = IPATH_CREG_OFFSET(PSXmitWaitCount),
-};
-
-/* kr_control bits */
-#define INFINIPATH_C_RESET (1U<<7)
-
-/* kr_intstatus, kr_intclear, kr_intmask bits */
-#define INFINIPATH_I_RCVURG_MASK ((1ULL<<17)-1)
-#define INFINIPATH_I_RCVURG_SHIFT 32
-#define INFINIPATH_I_RCVAVAIL_MASK ((1ULL<<17)-1)
-#define INFINIPATH_I_RCVAVAIL_SHIFT 0
-#define INFINIPATH_I_SERDESTRIMDONE (1ULL<<27)
-
-/* kr_hwerrclear, kr_hwerrmask, kr_hwerrstatus, bits */
-#define INFINIPATH_HWE_PCIEMEMPARITYERR_MASK  0x00000000000000ffULL
-#define INFINIPATH_HWE_PCIEMEMPARITYERR_SHIFT 0
-#define INFINIPATH_HWE_PCIEPOISONEDTLP      0x0000000010000000ULL
-#define INFINIPATH_HWE_PCIECPLTIMEOUT       0x0000000020000000ULL
-#define INFINIPATH_HWE_PCIEBUSPARITYXTLH    0x0000000040000000ULL
-#define INFINIPATH_HWE_PCIEBUSPARITYXADM    0x0000000080000000ULL
-#define INFINIPATH_HWE_PCIEBUSPARITYRADM    0x0000000100000000ULL
-#define INFINIPATH_HWE_COREPLL_FBSLIP       0x0080000000000000ULL
-#define INFINIPATH_HWE_COREPLL_RFSLIP       0x0100000000000000ULL
-#define INFINIPATH_HWE_PCIE1PLLFAILED       0x0400000000000000ULL
-#define INFINIPATH_HWE_PCIE0PLLFAILED       0x0800000000000000ULL
-#define INFINIPATH_HWE_SERDESPLLFAILED      0x1000000000000000ULL
-/* specific to this chip */
-#define INFINIPATH_HWE_PCIECPLDATAQUEUEERR         0x0000000000000040ULL
-#define INFINIPATH_HWE_PCIECPLHDRQUEUEERR          0x0000000000000080ULL
-#define INFINIPATH_HWE_SDMAMEMREADERR              0x0000000010000000ULL
-#define INFINIPATH_HWE_CLK_UC_PLLNOTLOCKED	   0x2000000000000000ULL
-#define INFINIPATH_HWE_PCIESERDESQ0PCLKNOTDETECT   0x0100000000000000ULL
-#define INFINIPATH_HWE_PCIESERDESQ1PCLKNOTDETECT   0x0200000000000000ULL
-#define INFINIPATH_HWE_PCIESERDESQ2PCLKNOTDETECT   0x0400000000000000ULL
-#define INFINIPATH_HWE_PCIESERDESQ3PCLKNOTDETECT   0x0800000000000000ULL
-#define INFINIPATH_HWE_DDSRXEQMEMORYPARITYERR	   0x0000008000000000ULL
-#define INFINIPATH_HWE_IB_UC_MEMORYPARITYERR	   0x0000004000000000ULL
-#define INFINIPATH_HWE_PCIE_UC_OCT0MEMORYPARITYERR 0x0000001000000000ULL
-#define INFINIPATH_HWE_PCIE_UC_OCT1MEMORYPARITYERR 0x0000002000000000ULL
-
-#define IBA7220_IBCS_LINKTRAININGSTATE_MASK 0x1F
-#define IBA7220_IBCS_LINKSTATE_SHIFT 5
-#define IBA7220_IBCS_LINKSPEED_SHIFT 8
-#define IBA7220_IBCS_LINKWIDTH_SHIFT 9
-
-#define IBA7220_IBCC_LINKINITCMD_MASK 0x7ULL
-#define IBA7220_IBCC_LINKCMD_SHIFT 19
-#define IBA7220_IBCC_MAXPKTLEN_SHIFT 21
-
-/* kr_ibcddrctrl bits */
-#define IBA7220_IBC_DLIDLMC_MASK	0xFFFFFFFFUL
-#define IBA7220_IBC_DLIDLMC_SHIFT	32
-#define IBA7220_IBC_HRTBT_MASK	3
-#define IBA7220_IBC_HRTBT_SHIFT	16
-#define IBA7220_IBC_HRTBT_ENB	0x10000UL
-#define IBA7220_IBC_LANE_REV_SUPPORTED (1<<8)
-#define IBA7220_IBC_LREV_MASK	1
-#define IBA7220_IBC_LREV_SHIFT	8
-#define IBA7220_IBC_RXPOL_MASK	1
-#define IBA7220_IBC_RXPOL_SHIFT	7
-#define IBA7220_IBC_WIDTH_SHIFT	5
-#define IBA7220_IBC_WIDTH_MASK	0x3
-#define IBA7220_IBC_WIDTH_1X_ONLY	(0<<IBA7220_IBC_WIDTH_SHIFT)
-#define IBA7220_IBC_WIDTH_4X_ONLY	(1<<IBA7220_IBC_WIDTH_SHIFT)
-#define IBA7220_IBC_WIDTH_AUTONEG	(2<<IBA7220_IBC_WIDTH_SHIFT)
-#define IBA7220_IBC_SPEED_AUTONEG	(1<<1)
-#define IBA7220_IBC_SPEED_SDR		(1<<2)
-#define IBA7220_IBC_SPEED_DDR		(1<<3)
-#define IBA7220_IBC_SPEED_AUTONEG_MASK  (0x7<<1)
-#define IBA7220_IBC_IBTA_1_2_MASK	(1)
-
-/* kr_ibcddrstatus */
-/* link latency shift is 0, don't bother defining */
-#define IBA7220_DDRSTAT_LINKLAT_MASK    0x3ffffff
-
-/* kr_extstatus bits */
-#define INFINIPATH_EXTS_FREQSEL 0x2
-#define INFINIPATH_EXTS_SERDESSEL 0x4
-#define INFINIPATH_EXTS_MEMBIST_ENDTEST     0x0000000000004000
-#define INFINIPATH_EXTS_MEMBIST_DISABLED    0x0000000000008000
-
-/* kr_xgxsconfig bits */
-#define INFINIPATH_XGXS_RESET          0x5ULL
-#define INFINIPATH_XGXS_FC_SAFE        (1ULL<<63)
-
-/* kr_rcvpktledcnt */
-#define IBA7220_LEDBLINK_ON_SHIFT 32 /* 4ns period on after packet */
-#define IBA7220_LEDBLINK_OFF_SHIFT 0 /* 4ns period off before next on */
-
-#define _IPATH_GPIO_SDA_NUM 1
-#define _IPATH_GPIO_SCL_NUM 0
-
-#define IPATH_GPIO_SDA (1ULL << \
-	(_IPATH_GPIO_SDA_NUM+INFINIPATH_EXTC_GPIOOE_SHIFT))
-#define IPATH_GPIO_SCL (1ULL << \
-	(_IPATH_GPIO_SCL_NUM+INFINIPATH_EXTC_GPIOOE_SHIFT))
-
-#define IBA7220_R_INTRAVAIL_SHIFT 17
-#define IBA7220_R_TAILUPD_SHIFT 35
-#define IBA7220_R_PORTCFG_SHIFT 36
-
-#define INFINIPATH_JINT_PACKETSHIFT 16
-#define INFINIPATH_JINT_DEFAULT_IDLE_TICKS  0
-#define INFINIPATH_JINT_DEFAULT_MAX_PACKETS 0
-
-#define IBA7220_HDRHEAD_PKTINT_SHIFT 32 /* interrupt cnt in upper 32 bits */
-
-/*
- * the size bits give us 2^N, in KB units.  0 marks as invalid,
- * and 7 is reserved.  We currently use only 2KB and 4KB
- */
-#define IBA7220_TID_SZ_SHIFT 37 /* shift to 3bit size selector */
-#define IBA7220_TID_SZ_2K (1UL<<IBA7220_TID_SZ_SHIFT) /* 2KB */
-#define IBA7220_TID_SZ_4K (2UL<<IBA7220_TID_SZ_SHIFT) /* 4KB */
-#define IBA7220_TID_PA_SHIFT 11U /* TID addr in chip stored w/o low bits */
-
-#define IPATH_AUTONEG_TRIES 5 /* sequential retries to negotiate DDR */
-
-static char int_type[16] = "auto";
-module_param_string(interrupt_type, int_type, sizeof(int_type), 0444);
-MODULE_PARM_DESC(int_type, " interrupt_type=auto|force_msi|force_intx");
-
-/* packet rate matching delay; chip has support */
-static u8 rate_to_delay[2][2] = {
-	/* 1x, 4x */
-	{   8, 2 }, /* SDR */
-	{   4, 1 }  /* DDR */
-};
-
-/* 7220 specific hardware errors... */
-static const struct ipath_hwerror_msgs ipath_7220_hwerror_msgs[] = {
-	INFINIPATH_HWE_MSG(PCIEPOISONEDTLP, "PCIe Poisoned TLP"),
-	INFINIPATH_HWE_MSG(PCIECPLTIMEOUT, "PCIe completion timeout"),
-	/*
-	 * In practice, it's unlikely wthat we'll see PCIe PLL, or bus
-	 * parity or memory parity error failures, because most likely we
-	 * won't be able to talk to the core of the chip.  Nonetheless, we
-	 * might see them, if they are in parts of the PCIe core that aren't
-	 * essential.
-	 */
-	INFINIPATH_HWE_MSG(PCIE1PLLFAILED, "PCIePLL1"),
-	INFINIPATH_HWE_MSG(PCIE0PLLFAILED, "PCIePLL0"),
-	INFINIPATH_HWE_MSG(PCIEBUSPARITYXTLH, "PCIe XTLH core parity"),
-	INFINIPATH_HWE_MSG(PCIEBUSPARITYXADM, "PCIe ADM TX core parity"),
-	INFINIPATH_HWE_MSG(PCIEBUSPARITYRADM, "PCIe ADM RX core parity"),
-	INFINIPATH_HWE_MSG(RXDSYNCMEMPARITYERR, "Rx Dsync"),
-	INFINIPATH_HWE_MSG(SERDESPLLFAILED, "SerDes PLL"),
-	INFINIPATH_HWE_MSG(PCIECPLDATAQUEUEERR, "PCIe cpl header queue"),
-	INFINIPATH_HWE_MSG(PCIECPLHDRQUEUEERR, "PCIe cpl data queue"),
-	INFINIPATH_HWE_MSG(SDMAMEMREADERR, "Send DMA memory read"),
-	INFINIPATH_HWE_MSG(CLK_UC_PLLNOTLOCKED, "uC PLL clock not locked"),
-	INFINIPATH_HWE_MSG(PCIESERDESQ0PCLKNOTDETECT,
-		"PCIe serdes Q0 no clock"),
-	INFINIPATH_HWE_MSG(PCIESERDESQ1PCLKNOTDETECT,
-		"PCIe serdes Q1 no clock"),
-	INFINIPATH_HWE_MSG(PCIESERDESQ2PCLKNOTDETECT,
-		"PCIe serdes Q2 no clock"),
-	INFINIPATH_HWE_MSG(PCIESERDESQ3PCLKNOTDETECT,
-		"PCIe serdes Q3 no clock"),
-	INFINIPATH_HWE_MSG(DDSRXEQMEMORYPARITYERR,
-		"DDS RXEQ memory parity"),
-	INFINIPATH_HWE_MSG(IB_UC_MEMORYPARITYERR, "IB uC memory parity"),
-	INFINIPATH_HWE_MSG(PCIE_UC_OCT0MEMORYPARITYERR,
-		"PCIe uC oct0 memory parity"),
-	INFINIPATH_HWE_MSG(PCIE_UC_OCT1MEMORYPARITYERR,
-		"PCIe uC oct1 memory parity"),
-};
-
-static void autoneg_work(struct work_struct *);
-
-/*
- * the offset is different for different configured port numbers, since
- * port0 is fixed in size, but others can vary.   Make it a function to
- * make the issue more obvious.
-*/
-static inline u32 port_egrtid_idx(struct ipath_devdata *dd, unsigned port)
-{
-	 return port ? dd->ipath_p0_rcvegrcnt +
-		 (port-1) * dd->ipath_rcvegrcnt : 0;
-}
-
-static void ipath_7220_txe_recover(struct ipath_devdata *dd)
-{
-	++ipath_stats.sps_txeparity;
-
-	dev_info(&dd->pcidev->dev,
-		"Recovering from TXE PIO parity error\n");
-	ipath_disarm_senderrbufs(dd);
-}
-
-
-/**
- * ipath_7220_handle_hwerrors - display hardware errors.
- * @dd: the infinipath device
- * @msg: the output buffer
- * @msgl: the size of the output buffer
- *
- * Use same msg buffer as regular errors to avoid excessive stack
- * use.  Most hardware errors are catastrophic, but for right now,
- * we'll print them and continue.  We reuse the same message buffer as
- * ipath_handle_errors() to avoid excessive stack usage.
- */
-static void ipath_7220_handle_hwerrors(struct ipath_devdata *dd, char *msg,
-				       size_t msgl)
-{
-	ipath_err_t hwerrs;
-	u32 bits, ctrl;
-	int isfatal = 0;
-	char bitsmsg[64];
-	int log_idx;
-
-	hwerrs = ipath_read_kreg64(dd, dd->ipath_kregs->kr_hwerrstatus);
-	if (!hwerrs) {
-		/*
-		 * better than printing cofusing messages
-		 * This seems to be related to clearing the crc error, or
-		 * the pll error during init.
-		 */
-		ipath_cdbg(VERBOSE, "Called but no hardware errors set\n");
-		goto bail;
-	} else if (hwerrs == ~0ULL) {
-		ipath_dev_err(dd, "Read of hardware error status failed "
-			      "(all bits set); ignoring\n");
-		goto bail;
-	}
-	ipath_stats.sps_hwerrs++;
-
-	/*
-	 * Always clear the error status register, except MEMBISTFAIL,
-	 * regardless of whether we continue or stop using the chip.
-	 * We want that set so we know it failed, even across driver reload.
-	 * We'll still ignore it in the hwerrmask.  We do this partly for
-	 * diagnostics, but also for support.
-	 */
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_hwerrclear,
-			 hwerrs&~INFINIPATH_HWE_MEMBISTFAILED);
-
-	hwerrs &= dd->ipath_hwerrmask;
-
-	/* We log some errors to EEPROM, check if we have any of those. */
-	for (log_idx = 0; log_idx < IPATH_EEP_LOG_CNT; ++log_idx)
-		if (hwerrs & dd->ipath_eep_st_masks[log_idx].hwerrs_to_log)
-			ipath_inc_eeprom_err(dd, log_idx, 1);
-	/*
-	 * Make sure we get this much out, unless told to be quiet,
-	 * or it's occurred within the last 5 seconds.
-	 */
-	if ((hwerrs & ~(dd->ipath_lasthwerror |
-			((INFINIPATH_HWE_TXEMEMPARITYERR_PIOBUF |
-			  INFINIPATH_HWE_TXEMEMPARITYERR_PIOPBC)
-			 << INFINIPATH_HWE_TXEMEMPARITYERR_SHIFT))) ||
-	    (ipath_debug & __IPATH_VERBDBG))
-		dev_info(&dd->pcidev->dev, "Hardware error: hwerr=0x%llx "
-			 "(cleared)\n", (unsigned long long) hwerrs);
-	dd->ipath_lasthwerror |= hwerrs;
-
-	if (hwerrs & ~dd->ipath_hwe_bitsextant)
-		ipath_dev_err(dd, "hwerror interrupt with unknown errors "
-			      "%llx set\n", (unsigned long long)
-			      (hwerrs & ~dd->ipath_hwe_bitsextant));
-
-	if (hwerrs & INFINIPATH_HWE_IB_UC_MEMORYPARITYERR)
-		ipath_sd7220_clr_ibpar(dd);
-
-	ctrl = ipath_read_kreg32(dd, dd->ipath_kregs->kr_control);
-	if ((ctrl & INFINIPATH_C_FREEZEMODE) && !ipath_diag_inuse) {
-		/*
-		 * Parity errors in send memory are recoverable by h/w
-		 * just do housekeeping, exit freeze mode and continue.
-		 */
-		if (hwerrs & ((INFINIPATH_HWE_TXEMEMPARITYERR_PIOBUF |
-			       INFINIPATH_HWE_TXEMEMPARITYERR_PIOPBC)
-			      << INFINIPATH_HWE_TXEMEMPARITYERR_SHIFT)) {
-			ipath_7220_txe_recover(dd);
-			hwerrs &= ~((INFINIPATH_HWE_TXEMEMPARITYERR_PIOBUF |
-				     INFINIPATH_HWE_TXEMEMPARITYERR_PIOPBC)
-				    << INFINIPATH_HWE_TXEMEMPARITYERR_SHIFT);
-		}
-		if (hwerrs) {
-			/*
-			 * If any set that we aren't ignoring only make the
-			 * complaint once, in case it's stuck or recurring,
-			 * and we get here multiple times
-			 * Force link down, so switch knows, and
-			 * LEDs are turned off.
-			 */
-			if (dd->ipath_flags & IPATH_INITTED) {
-				ipath_set_linkstate(dd, IPATH_IB_LINKDOWN);
-				ipath_setup_7220_setextled(dd,
-					INFINIPATH_IBCS_L_STATE_DOWN,
-					INFINIPATH_IBCS_LT_STATE_DISABLED);
-				ipath_dev_err(dd, "Fatal Hardware Error "
-					      "(freeze mode), no longer"
-					      " usable, SN %.16s\n",
-						  dd->ipath_serial);
-				isfatal = 1;
-			}
-			/*
-			 * Mark as having had an error for driver, and also
-			 * for /sys and status word mapped to user programs.
-			 * This marks unit as not usable, until reset.
-			 */
-			*dd->ipath_statusp &= ~IPATH_STATUS_IB_READY;
-			*dd->ipath_statusp |= IPATH_STATUS_HWERROR;
-			dd->ipath_flags &= ~IPATH_INITTED;
-		} else {
-			ipath_dbg("Clearing freezemode on ignored or "
-				"recovered hardware error\n");
-			ipath_clear_freeze(dd);
-		}
-	}
-
-	*msg = '\0';
-
-	if (hwerrs & INFINIPATH_HWE_MEMBISTFAILED) {
-		strlcat(msg, "[Memory BIST test failed, "
-			"InfiniPath hardware unusable]", msgl);
-		/* ignore from now on, so disable until driver reloaded */
-		*dd->ipath_statusp |= IPATH_STATUS_HWERROR;
-		dd->ipath_hwerrmask &= ~INFINIPATH_HWE_MEMBISTFAILED;
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_hwerrmask,
-				 dd->ipath_hwerrmask);
-	}
-
-	ipath_format_hwerrors(hwerrs,
-			      ipath_7220_hwerror_msgs,
-			      ARRAY_SIZE(ipath_7220_hwerror_msgs),
-			      msg, msgl);
-
-	if (hwerrs & (INFINIPATH_HWE_PCIEMEMPARITYERR_MASK
-		      << INFINIPATH_HWE_PCIEMEMPARITYERR_SHIFT)) {
-		bits = (u32) ((hwerrs >>
-			       INFINIPATH_HWE_PCIEMEMPARITYERR_SHIFT) &
-			      INFINIPATH_HWE_PCIEMEMPARITYERR_MASK);
-		snprintf(bitsmsg, sizeof bitsmsg,
-			 "[PCIe Mem Parity Errs %x] ", bits);
-		strlcat(msg, bitsmsg, msgl);
-	}
-
-#define _IPATH_PLL_FAIL (INFINIPATH_HWE_COREPLL_FBSLIP |	\
-			 INFINIPATH_HWE_COREPLL_RFSLIP)
-
-	if (hwerrs & _IPATH_PLL_FAIL) {
-		snprintf(bitsmsg, sizeof bitsmsg,
-			 "[PLL failed (%llx), InfiniPath hardware unusable]",
-			 (unsigned long long) hwerrs & _IPATH_PLL_FAIL);
-		strlcat(msg, bitsmsg, msgl);
-		/* ignore from now on, so disable until driver reloaded */
-		dd->ipath_hwerrmask &= ~(hwerrs & _IPATH_PLL_FAIL);
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_hwerrmask,
-				 dd->ipath_hwerrmask);
-	}
-
-	if (hwerrs & INFINIPATH_HWE_SERDESPLLFAILED) {
-		/*
-		 * If it occurs, it is left masked since the eternal
-		 * interface is unused.
-		 */
-		dd->ipath_hwerrmask &= ~INFINIPATH_HWE_SERDESPLLFAILED;
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_hwerrmask,
-				 dd->ipath_hwerrmask);
-	}
-
-	ipath_dev_err(dd, "%s hardware error\n", msg);
-	/*
-	 * For /sys status file. if no trailing } is copied, we'll
-	 * know it was truncated.
-	 */
-	if (isfatal && !ipath_diag_inuse && dd->ipath_freezemsg)
-		snprintf(dd->ipath_freezemsg, dd->ipath_freezelen,
-			 "{%s}", msg);
-bail:;
-}
-
-/**
- * ipath_7220_boardname - fill in the board name
- * @dd: the infinipath device
- * @name: the output buffer
- * @namelen: the size of the output buffer
- *
- * info is based on the board revision register
- */
-static int ipath_7220_boardname(struct ipath_devdata *dd, char *name,
-	size_t namelen)
-{
-	char *n = NULL;
-	u8 boardrev = dd->ipath_boardrev;
-	int ret;
-
-	if (boardrev == 15) {
-		/*
-		 * Emulator sometimes comes up all-ones, rather than zero.
-		 */
-		boardrev = 0;
-		dd->ipath_boardrev = boardrev;
-	}
-	switch (boardrev) {
-	case 0:
-		n = "InfiniPath_7220_Emulation";
-		break;
-	case 1:
-		n = "InfiniPath_QLE7240";
-		break;
-	case 2:
-		n = "InfiniPath_QLE7280";
-		break;
-	case 3:
-		n = "InfiniPath_QLE7242";
-		break;
-	case 4:
-		n = "InfiniPath_QEM7240";
-		break;
-	case 5:
-		n = "InfiniPath_QMI7240";
-		break;
-	case 6:
-		n = "InfiniPath_QMI7264";
-		break;
-	case 7:
-		n = "InfiniPath_QMH7240";
-		break;
-	case 8:
-		n = "InfiniPath_QME7240";
-		break;
-	case 9:
-		n = "InfiniPath_QLE7250";
-		break;
-	case 10:
-		n = "InfiniPath_QLE7290";
-		break;
-	case 11:
-		n = "InfiniPath_QEM7250";
-		break;
-	case 12:
-		n = "InfiniPath_QLE-Bringup";
-		break;
-	default:
-		ipath_dev_err(dd,
-			      "Don't yet know about board with ID %u\n",
-			      boardrev);
-		snprintf(name, namelen, "Unknown_InfiniPath_PCIe_%u",
-			 boardrev);
-		break;
-	}
-	if (n)
-		snprintf(name, namelen, "%s", n);
-
-	if (dd->ipath_majrev != 5 || !dd->ipath_minrev ||
-		dd->ipath_minrev > 2) {
-		ipath_dev_err(dd, "Unsupported InfiniPath hardware "
-			      "revision %u.%u!\n",
-			      dd->ipath_majrev, dd->ipath_minrev);
-		ret = 1;
-	} else if (dd->ipath_minrev == 1 &&
-		!(dd->ipath_flags & IPATH_INITTED)) {
-		/* Rev1 chips are prototype. Complain at init, but allow use */
-		ipath_dev_err(dd, "Unsupported hardware "
-			      "revision %u.%u, Contact support@qlogic.com\n",
-			      dd->ipath_majrev, dd->ipath_minrev);
-		ret = 0;
-	} else
-		ret = 0;
-
-	/*
-	 * Set here not in ipath_init_*_funcs because we have to do
-	 * it after we can read chip registers.
-	 */
-	dd->ipath_ureg_align = 0x10000;  /* 64KB alignment */
-
-	return ret;
-}
-
-/**
- * ipath_7220_init_hwerrors - enable hardware errors
- * @dd: the infinipath device
- *
- * now that we have finished initializing everything that might reasonably
- * cause a hardware error, and cleared those errors bits as they occur,
- * we can enable hardware errors in the mask (potentially enabling
- * freeze mode), and enable hardware errors as errors (along with
- * everything else) in errormask
- */
-static void ipath_7220_init_hwerrors(struct ipath_devdata *dd)
-{
-	ipath_err_t val;
-	u64 extsval;
-
-	extsval = ipath_read_kreg64(dd, dd->ipath_kregs->kr_extstatus);
-
-	if (!(extsval & (INFINIPATH_EXTS_MEMBIST_ENDTEST |
-			INFINIPATH_EXTS_MEMBIST_DISABLED)))
-		ipath_dev_err(dd, "MemBIST did not complete!\n");
-	if (extsval & INFINIPATH_EXTS_MEMBIST_DISABLED)
-		dev_info(&dd->pcidev->dev, "MemBIST is disabled.\n");
-
-	val = ~0ULL;	/* barring bugs, all hwerrors become interrupts, */
-
-	if (!dd->ipath_boardrev)	/* no PLL for Emulator */
-		val &= ~INFINIPATH_HWE_SERDESPLLFAILED;
-
-	if (dd->ipath_minrev == 1)
-		val &= ~(1ULL << 42); /* TXE LaunchFIFO Parity rev1 issue */
-
-	val &= ~INFINIPATH_HWE_IB_UC_MEMORYPARITYERR;
-	dd->ipath_hwerrmask = val;
-
-	/*
-	 * special trigger "error" is for debugging purposes. It
-	 * works around a processor/chipset problem.  The error
-	 * interrupt allows us to count occurrences, but we don't
-	 * want to pay the overhead for normal use.  Emulation only
-	 */
-	if (!dd->ipath_boardrev)
-		dd->ipath_maskederrs = INFINIPATH_E_SENDSPECIALTRIGGER;
-}
-
-/*
- * All detailed interaction with the SerDes has been moved to ipath_sd7220.c
- *
- * The portion of IBA7220-specific bringup_serdes() that actually deals with
- * registers and memory within the SerDes itself is ipath_sd7220_init().
- */
-
-/**
- * ipath_7220_bringup_serdes - bring up the serdes
- * @dd: the infinipath device
- */
-static int ipath_7220_bringup_serdes(struct ipath_devdata *dd)
-{
-	int ret = 0;
-	u64 val, prev_val, guid;
-	int was_reset;		/* Note whether uC was reset */
-
-	ipath_dbg("Trying to bringup serdes\n");
-
-	if (ipath_read_kreg64(dd, dd->ipath_kregs->kr_hwerrstatus) &
-	    INFINIPATH_HWE_SERDESPLLFAILED) {
-		ipath_dbg("At start, serdes PLL failed bit set "
-			  "in hwerrstatus, clearing and continuing\n");
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_hwerrclear,
-				 INFINIPATH_HWE_SERDESPLLFAILED);
-	}
-
-	dd->ibdeltainprog = 1;
-	dd->ibsymsnap =
-	     ipath_read_creg32(dd, dd->ipath_cregs->cr_ibsymbolerrcnt);
-	dd->iblnkerrsnap =
-	     ipath_read_creg32(dd, dd->ipath_cregs->cr_iblinkerrrecovcnt);
-
-	if (!dd->ipath_ibcddrctrl) {
-		/* not on re-init after reset */
-		dd->ipath_ibcddrctrl =
-			ipath_read_kreg64(dd, dd->ipath_kregs->kr_ibcddrctrl);
-
-		if (dd->ipath_link_speed_enabled ==
-			(IPATH_IB_SDR | IPATH_IB_DDR))
-			dd->ipath_ibcddrctrl |=
-				IBA7220_IBC_SPEED_AUTONEG_MASK |
-				IBA7220_IBC_IBTA_1_2_MASK;
-		else
-			dd->ipath_ibcddrctrl |=
-				dd->ipath_link_speed_enabled == IPATH_IB_DDR
-				?  IBA7220_IBC_SPEED_DDR :
-				IBA7220_IBC_SPEED_SDR;
-		if ((dd->ipath_link_width_enabled & (IB_WIDTH_1X |
-			IB_WIDTH_4X)) == (IB_WIDTH_1X | IB_WIDTH_4X))
-			dd->ipath_ibcddrctrl |= IBA7220_IBC_WIDTH_AUTONEG;
-		else
-			dd->ipath_ibcddrctrl |=
-				dd->ipath_link_width_enabled == IB_WIDTH_4X
-				? IBA7220_IBC_WIDTH_4X_ONLY :
-				IBA7220_IBC_WIDTH_1X_ONLY;
-
-		/* always enable these on driver reload, not sticky */
-		dd->ipath_ibcddrctrl |=
-			IBA7220_IBC_RXPOL_MASK << IBA7220_IBC_RXPOL_SHIFT;
-		dd->ipath_ibcddrctrl |=
-			IBA7220_IBC_HRTBT_MASK << IBA7220_IBC_HRTBT_SHIFT;
-		/*
-		 * automatic lane reversal detection for receive
-		 * doesn't work correctly in rev 1, so disable it
-		 * on that rev, otherwise enable (disabling not
-		 * sticky across reload for >rev1)
-		 */
-		if (dd->ipath_minrev == 1)
-			dd->ipath_ibcddrctrl &=
-			~IBA7220_IBC_LANE_REV_SUPPORTED;
-		else
-			dd->ipath_ibcddrctrl |=
-				IBA7220_IBC_LANE_REV_SUPPORTED;
-	}
-
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_ibcddrctrl,
-			dd->ipath_ibcddrctrl);
-
-	ipath_write_kreg(dd, IPATH_KREG_OFFSET(IBNCModeCtrl), 0Ull);
-
-	/* IBA7220 has SERDES MPU reset in D0 of what _was_ IBPLLCfg */
-	val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_ibserdesctrl);
-	/* remember if uC was in Reset or not, for dactrim */
-	was_reset = (val & 1);
-	ipath_cdbg(VERBOSE, "IBReset %s xgxsconfig %llx\n",
-		   was_reset ? "Asserted" : "Negated", (unsigned long long)
-		   ipath_read_kreg64(dd, dd->ipath_kregs->kr_xgxsconfig));
-
-	if (dd->ipath_boardrev) {
-		/*
-		 * Hardware is not emulator, and may have been reset. Init it.
-		 * Below will release reset, but needs to know if chip was
-		 * originally in reset, to only trim DACs on first time
-		 * after chip reset or powercycle (not driver reload)
-		 */
-		ret = ipath_sd7220_init(dd, was_reset);
-	}
-
-	val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_xgxsconfig);
-	prev_val = val;
-	val |= INFINIPATH_XGXS_FC_SAFE;
-	if (val != prev_val) {
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_xgxsconfig, val);
-		ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
-	}
-	if (val & INFINIPATH_XGXS_RESET)
-		val &= ~INFINIPATH_XGXS_RESET;
-	if (val != prev_val)
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_xgxsconfig, val);
-
-	ipath_cdbg(VERBOSE, "done: xgxs=%llx from %llx\n",
-		   (unsigned long long)
-		   ipath_read_kreg64(dd, dd->ipath_kregs->kr_xgxsconfig),
-		   (unsigned long long) prev_val);
-
-	guid = be64_to_cpu(dd->ipath_guid);
-
-	if (!guid) {
-		/* have to have something, so use likely unique tsc */
-		guid = get_cycles();
-		ipath_dbg("No GUID for heartbeat, faking %llx\n",
-			(unsigned long long)guid);
-	} else
-		ipath_cdbg(VERBOSE, "Wrote %llX to HRTBT_GUID\n",
-			(unsigned long long) guid);
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_hrtbt_guid, guid);
-	return ret;
-}
-
-static void ipath_7220_config_jint(struct ipath_devdata *dd,
-				   u16 idle_ticks, u16 max_packets)
-{
-
-	/*
-	 * We can request a receive interrupt for 1 or more packets
-	 * from current offset.
-	 */
-	if (idle_ticks == 0 || max_packets == 0)
-		/* interrupt after one packet if no mitigation */
-		dd->ipath_rhdrhead_intr_off =
-			1ULL << IBA7220_HDRHEAD_PKTINT_SHIFT;
-	else
-		/* Turn off RcvHdrHead interrupts if using mitigation */
-		dd->ipath_rhdrhead_intr_off = 0ULL;
-
-	/* refresh kernel RcvHdrHead registers... */
-	ipath_write_ureg(dd, ur_rcvhdrhead,
-			 dd->ipath_rhdrhead_intr_off |
-			 dd->ipath_pd[0]->port_head, 0);
-
-	dd->ipath_jint_max_packets = max_packets;
-	dd->ipath_jint_idle_ticks = idle_ticks;
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_jintreload,
-			 ((u64) max_packets << INFINIPATH_JINT_PACKETSHIFT) |
-			 idle_ticks);
-}
-
-/**
- * ipath_7220_quiet_serdes - set serdes to txidle
- * @dd: the infinipath device
- * Called when driver is being unloaded
- */
-static void ipath_7220_quiet_serdes(struct ipath_devdata *dd)
-{
-	u64 val;
-	if (dd->ibsymdelta || dd->iblnkerrdelta ||
-	    dd->ibdeltainprog) {
-		u64 diagc;
-		/* enable counter writes */
-		diagc = ipath_read_kreg64(dd, dd->ipath_kregs->kr_hwdiagctrl);
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_hwdiagctrl,
-				 diagc | INFINIPATH_DC_COUNTERWREN);
-
-		if (dd->ibsymdelta || dd->ibdeltainprog) {
-			val = ipath_read_creg32(dd,
-					dd->ipath_cregs->cr_ibsymbolerrcnt);
-			if (dd->ibdeltainprog)
-				val -= val - dd->ibsymsnap;
-			val -= dd->ibsymdelta;
-			ipath_write_creg(dd,
-				  dd->ipath_cregs->cr_ibsymbolerrcnt, val);
-		}
-		if (dd->iblnkerrdelta || dd->ibdeltainprog) {
-			val = ipath_read_creg32(dd,
-					dd->ipath_cregs->cr_iblinkerrrecovcnt);
-			if (dd->ibdeltainprog)
-				val -= val - dd->iblnkerrsnap;
-			val -= dd->iblnkerrdelta;
-			ipath_write_creg(dd,
-				   dd->ipath_cregs->cr_iblinkerrrecovcnt, val);
-	     }
-
-	     /* and disable counter writes */
-	     ipath_write_kreg(dd, dd->ipath_kregs->kr_hwdiagctrl, diagc);
-	}
-
-	dd->ipath_flags &= ~IPATH_IB_AUTONEG_INPROG;
-	wake_up(&dd->ipath_autoneg_wait);
-	cancel_delayed_work(&dd->ipath_autoneg_work);
-	flush_scheduled_work();
-	ipath_shutdown_relock_poll(dd);
-	val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_xgxsconfig);
-	val |= INFINIPATH_XGXS_RESET;
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_xgxsconfig, val);
-}
-
-static int ipath_7220_intconfig(struct ipath_devdata *dd)
-{
-	ipath_7220_config_jint(dd, dd->ipath_jint_idle_ticks,
-			       dd->ipath_jint_max_packets);
-	return 0;
-}
-
-/**
- * ipath_setup_7220_setextled - set the state of the two external LEDs
- * @dd: the infinipath device
- * @lst: the L state
- * @ltst: the LT state
- *
- * These LEDs indicate the physical and logical state of IB link.
- * For this chip (at least with recommended board pinouts), LED1
- * is Yellow (logical state) and LED2 is Green (physical state),
- *
- * Note:  We try to match the Mellanox HCA LED behavior as best
- * we can.  Green indicates physical link state is OK (something is
- * plugged in, and we can train).
- * Amber indicates the link is logically up (ACTIVE).
- * Mellanox further blinks the amber LED to indicate data packet
- * activity, but we have no hardware support for that, so it would
- * require waking up every 10-20 msecs and checking the counters
- * on the chip, and then turning the LED off if appropriate.  That's
- * visible overhead, so not something we will do.
- *
- */
-static void ipath_setup_7220_setextled(struct ipath_devdata *dd, u64 lst,
-				       u64 ltst)
-{
-	u64 extctl, ledblink = 0;
-	unsigned long flags = 0;
-
-	/* the diags use the LED to indicate diag info, so we leave
-	 * the external LED alone when the diags are running */
-	if (ipath_diag_inuse)
-		return;
-
-	/* Allow override of LED display for, e.g. Locating system in rack */
-	if (dd->ipath_led_override) {
-		ltst = (dd->ipath_led_override & IPATH_LED_PHYS)
-			? INFINIPATH_IBCS_LT_STATE_LINKUP
-			: INFINIPATH_IBCS_LT_STATE_DISABLED;
-		lst = (dd->ipath_led_override & IPATH_LED_LOG)
-			? INFINIPATH_IBCS_L_STATE_ACTIVE
-			: INFINIPATH_IBCS_L_STATE_DOWN;
-	}
-
-	spin_lock_irqsave(&dd->ipath_gpio_lock, flags);
-	extctl = dd->ipath_extctrl & ~(INFINIPATH_EXTC_LED1PRIPORT_ON |
-				       INFINIPATH_EXTC_LED2PRIPORT_ON);
-	if (ltst == INFINIPATH_IBCS_LT_STATE_LINKUP) {
-		extctl |= INFINIPATH_EXTC_LED1PRIPORT_ON;
-		/*
-		 * counts are in chip clock (4ns) periods.
-		 * This is 1/16 sec (66.6ms) on,
-		 * 3/16 sec (187.5 ms) off, with packets rcvd
-		 */
-		ledblink = ((66600*1000UL/4) << IBA7220_LEDBLINK_ON_SHIFT)
-			| ((187500*1000UL/4) << IBA7220_LEDBLINK_OFF_SHIFT);
-	}
-	if (lst == INFINIPATH_IBCS_L_STATE_ACTIVE)
-		extctl |= INFINIPATH_EXTC_LED2PRIPORT_ON;
-	dd->ipath_extctrl = extctl;
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_extctrl, extctl);
-	spin_unlock_irqrestore(&dd->ipath_gpio_lock, flags);
-
-	if (ledblink) /* blink the LED on packet receive */
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_rcvpktledcnt,
-			ledblink);
-}
-
-/*
- * Similar to pci_intx(pdev, 1), except that we make sure
- * msi is off...
- */
-static void ipath_enable_intx(struct pci_dev *pdev)
-{
-	u16 cw, new;
-	int pos;
-
-	/* first, turn on INTx */
-	pci_read_config_word(pdev, PCI_COMMAND, &cw);
-	new = cw & ~PCI_COMMAND_INTX_DISABLE;
-	if (new != cw)
-		pci_write_config_word(pdev, PCI_COMMAND, new);
-
-	/* then turn off MSI */
-	pos = pci_find_capability(pdev, PCI_CAP_ID_MSI);
-	if (pos) {
-		pci_read_config_word(pdev, pos + PCI_MSI_FLAGS, &cw);
-		new = cw & ~PCI_MSI_FLAGS_ENABLE;
-		if (new != cw)
-			pci_write_config_word(pdev, pos + PCI_MSI_FLAGS, new);
-	}
-}
-
-static int ipath_msi_enabled(struct pci_dev *pdev)
-{
-	int pos, ret = 0;
-
-	pos = pci_find_capability(pdev, PCI_CAP_ID_MSI);
-	if (pos) {
-		u16 cw;
-
-		pci_read_config_word(pdev, pos + PCI_MSI_FLAGS, &cw);
-		ret = !!(cw & PCI_MSI_FLAGS_ENABLE);
-	}
-	return ret;
-}
-
-/*
- * disable msi interrupt if enabled, and clear the flag.
- * flag is used primarily for the fallback to INTx, but
- * is also used in reinit after reset as a flag.
- */
-static void ipath_7220_nomsi(struct ipath_devdata *dd)
-{
-	dd->ipath_msi_lo = 0;
-
-	if (ipath_msi_enabled(dd->pcidev)) {
-		/*
-		 * free, but don't zero; later kernels require
-		 * it be freed before disable_msi, so the intx
-		 * setup has to request it again.
-		 */
-		 if (dd->ipath_irq)
-			free_irq(dd->ipath_irq, dd);
-		pci_disable_msi(dd->pcidev);
-	}
-}
-
-/*
- * ipath_setup_7220_cleanup - clean up any per-chip chip-specific stuff
- * @dd: the infinipath device
- *
- * Nothing but msi interrupt cleanup for now.
- *
- * This is called during driver unload.
- */
-static void ipath_setup_7220_cleanup(struct ipath_devdata *dd)
-{
-	ipath_7220_nomsi(dd);
-}
-
-
-static void ipath_7220_pcie_params(struct ipath_devdata *dd, u32 boardrev)
-{
-	u16 linkstat, minwidth, speed;
-	int pos;
-
-	pos = pci_find_capability(dd->pcidev, PCI_CAP_ID_EXP);
-	if (!pos) {
-		ipath_dev_err(dd, "Can't find PCI Express capability!\n");
-		goto bail;
-	}
-
-	pci_read_config_word(dd->pcidev, pos + PCI_EXP_LNKSTA,
-			     &linkstat);
-	/*
-	 * speed is bits 0-4, linkwidth is bits 4-8
-	 * no defines for them in headers
-	 */
-	speed = linkstat & 0xf;
-	linkstat >>= 4;
-	linkstat &= 0x1f;
-	dd->ipath_lbus_width = linkstat;
-	switch (boardrev) {
-	case 0:
-	case 2:
-	case 10:
-	case 12:
-		minwidth = 16; /* x16 capable boards */
-		break;
-	default:
-		minwidth = 8; /* x8 capable boards */
-		break;
-	}
-
-	switch (speed) {
-	case 1:
-		dd->ipath_lbus_speed = 2500; /* Gen1, 2.5GHz */
-		break;
-	case 2:
-		dd->ipath_lbus_speed = 5000; /* Gen1, 5GHz */
-		break;
-	default: /* not defined, assume gen1 */
-		dd->ipath_lbus_speed = 2500;
-		break;
-	}
-
-	if (linkstat < minwidth)
-		ipath_dev_err(dd,
-			"PCIe width %u (x%u HCA), performance "
-			"reduced\n", linkstat, minwidth);
-	else
-		ipath_cdbg(VERBOSE, "PCIe speed %u width %u (x%u HCA)\n",
-			dd->ipath_lbus_speed, linkstat, minwidth);
-
-	if (speed != 1)
-		ipath_dev_err(dd,
-			"PCIe linkspeed %u is incorrect; "
-			"should be 1 (2500)!\n", speed);
-
-bail:
-	/* fill in string, even on errors */
-	snprintf(dd->ipath_lbus_info, sizeof(dd->ipath_lbus_info),
-		"PCIe,%uMHz,x%u\n",
-		dd->ipath_lbus_speed,
-		dd->ipath_lbus_width);
-	return;
-}
-
-
-/**
- * ipath_setup_7220_config - setup PCIe config related stuff
- * @dd: the infinipath device
- * @pdev: the PCI device
- *
- * The pci_enable_msi() call will fail on systems with MSI quirks
- * such as those with AMD8131, even if the device of interest is not
- * attached to that device, (in the 2.6.13 - 2.6.15 kernels, at least, fixed
- * late in 2.6.16).
- * All that can be done is to edit the kernel source to remove the quirk
- * check until that is fixed.
- * We do not need to call enable_msi() for our HyperTransport chip,
- * even though it uses MSI, and we want to avoid the quirk warning, so
- * So we call enable_msi only for PCIe.  If we do end up needing
- * pci_enable_msi at some point in the future for HT, we'll move the
- * call back into the main init_one code.
- * We save the msi lo and hi values, so we can restore them after
- * chip reset (the kernel PCI infrastructure doesn't yet handle that
- * correctly).
- */
-static int ipath_setup_7220_config(struct ipath_devdata *dd,
-				   struct pci_dev *pdev)
-{
-	int pos, ret = -1;
-	u32 boardrev;
-
-	dd->ipath_msi_lo = 0;	/* used as a flag during reset processing */
-
-	pos = pci_find_capability(pdev, PCI_CAP_ID_MSI);
-	if (!strcmp(int_type, "force_msi") || !strcmp(int_type, "auto"))
-		ret = pci_enable_msi(pdev);
-	if (ret) {
-		if (!strcmp(int_type, "force_msi")) {
-			ipath_dev_err(dd, "pci_enable_msi failed: %d, "
-				      "force_msi is on, so not continuing.\n",
-				      ret);
-			return ret;
-		}
-
-		ipath_enable_intx(pdev);
-		if (!strcmp(int_type, "auto"))
-			ipath_dev_err(dd, "pci_enable_msi failed: %d, "
-				      "falling back to INTx\n", ret);
-	} else if (pos) {
-		u16 control;
-		pci_read_config_dword(pdev, pos + PCI_MSI_ADDRESS_LO,
-				      &dd->ipath_msi_lo);
-		pci_read_config_dword(pdev, pos + PCI_MSI_ADDRESS_HI,
-				      &dd->ipath_msi_hi);
-		pci_read_config_word(pdev, pos + PCI_MSI_FLAGS,
-				     &control);
-		/* now save the data (vector) info */
-		pci_read_config_word(pdev,
-				     pos + ((control & PCI_MSI_FLAGS_64BIT)
-					    ? PCI_MSI_DATA_64 :
-					    PCI_MSI_DATA_32),
-				     &dd->ipath_msi_data);
-	} else
-		ipath_dev_err(dd, "Can't find MSI capability, "
-			      "can't save MSI settings for reset\n");
-
-	dd->ipath_irq = pdev->irq;
-
-	/*
-	 * We save the cachelinesize also, although it doesn't
-	 * really matter.
-	 */
-	pci_read_config_byte(pdev, PCI_CACHE_LINE_SIZE,
-			     &dd->ipath_pci_cacheline);
-
-	/*
-	 * this function called early, ipath_boardrev not set yet.  Can't
-	 * use ipath_read_kreg64() yet, too early in init, so use readq()
-	 */
-	boardrev = (readq(&dd->ipath_kregbase[dd->ipath_kregs->kr_revision])
-		 >> INFINIPATH_R_BOARDID_SHIFT) & INFINIPATH_R_BOARDID_MASK;
-
-	ipath_7220_pcie_params(dd, boardrev);
-
-	dd->ipath_flags |= IPATH_NODMA_RTAIL | IPATH_HAS_SEND_DMA |
-		IPATH_HAS_PBC_CNT | IPATH_HAS_THRESH_UPDATE;
-	dd->ipath_pioupd_thresh = 4U; /* set default update threshold */
-	return 0;
-}
-
-static void ipath_init_7220_variables(struct ipath_devdata *dd)
-{
-	/*
-	 * setup the register offsets, since they are different for each
-	 * chip
-	 */
-	dd->ipath_kregs = &ipath_7220_kregs;
-	dd->ipath_cregs = &ipath_7220_cregs;
-
-	/*
-	 * bits for selecting i2c direction and values,
-	 * used for I2C serial flash
-	 */
-	dd->ipath_gpio_sda_num = _IPATH_GPIO_SDA_NUM;
-	dd->ipath_gpio_scl_num = _IPATH_GPIO_SCL_NUM;
-	dd->ipath_gpio_sda = IPATH_GPIO_SDA;
-	dd->ipath_gpio_scl = IPATH_GPIO_SCL;
-
-	/*
-	 * Fill in data for field-values that change in IBA7220.
-	 * We dynamically specify only the mask for LINKTRAININGSTATE
-	 * and only the shift for LINKSTATE, as they are the only ones
-	 * that change.  Also precalculate the 3 link states of interest
-	 * and the combined mask.
-	 */
-	dd->ibcs_ls_shift = IBA7220_IBCS_LINKSTATE_SHIFT;
-	dd->ibcs_lts_mask = IBA7220_IBCS_LINKTRAININGSTATE_MASK;
-	dd->ibcs_mask = (INFINIPATH_IBCS_LINKSTATE_MASK <<
-		dd->ibcs_ls_shift) | dd->ibcs_lts_mask;
-	dd->ib_init = (INFINIPATH_IBCS_LT_STATE_LINKUP <<
-		INFINIPATH_IBCS_LINKTRAININGSTATE_SHIFT) |
-		(INFINIPATH_IBCS_L_STATE_INIT << dd->ibcs_ls_shift);
-	dd->ib_arm = (INFINIPATH_IBCS_LT_STATE_LINKUP <<
-		INFINIPATH_IBCS_LINKTRAININGSTATE_SHIFT) |
-		(INFINIPATH_IBCS_L_STATE_ARM << dd->ibcs_ls_shift);
-	dd->ib_active = (INFINIPATH_IBCS_LT_STATE_LINKUP <<
-		INFINIPATH_IBCS_LINKTRAININGSTATE_SHIFT) |
-		(INFINIPATH_IBCS_L_STATE_ACTIVE << dd->ibcs_ls_shift);
-
-	/*
-	 * Fill in data for ibcc field-values that change in IBA7220.
-	 * We dynamically specify only the mask for LINKINITCMD
-	 * and only the shift for LINKCMD and MAXPKTLEN, as they are
-	 * the only ones that change.
-	 */
-	dd->ibcc_lic_mask = IBA7220_IBCC_LINKINITCMD_MASK;
-	dd->ibcc_lc_shift = IBA7220_IBCC_LINKCMD_SHIFT;
-	dd->ibcc_mpl_shift = IBA7220_IBCC_MAXPKTLEN_SHIFT;
-
-	/* Fill in shifts for RcvCtrl. */
-	dd->ipath_r_portenable_shift = INFINIPATH_R_PORTENABLE_SHIFT;
-	dd->ipath_r_intravail_shift = IBA7220_R_INTRAVAIL_SHIFT;
-	dd->ipath_r_tailupd_shift = IBA7220_R_TAILUPD_SHIFT;
-	dd->ipath_r_portcfg_shift = IBA7220_R_PORTCFG_SHIFT;
-
-	/* variables for sanity checking interrupt and errors */
-	dd->ipath_hwe_bitsextant =
-		(INFINIPATH_HWE_RXEMEMPARITYERR_MASK <<
-		 INFINIPATH_HWE_RXEMEMPARITYERR_SHIFT) |
-		(INFINIPATH_HWE_TXEMEMPARITYERR_MASK <<
-		 INFINIPATH_HWE_TXEMEMPARITYERR_SHIFT) |
-		(INFINIPATH_HWE_PCIEMEMPARITYERR_MASK <<
-		 INFINIPATH_HWE_PCIEMEMPARITYERR_SHIFT) |
-		INFINIPATH_HWE_PCIE1PLLFAILED |
-		INFINIPATH_HWE_PCIE0PLLFAILED |
-		INFINIPATH_HWE_PCIEPOISONEDTLP |
-		INFINIPATH_HWE_PCIECPLTIMEOUT |
-		INFINIPATH_HWE_PCIEBUSPARITYXTLH |
-		INFINIPATH_HWE_PCIEBUSPARITYXADM |
-		INFINIPATH_HWE_PCIEBUSPARITYRADM |
-		INFINIPATH_HWE_MEMBISTFAILED |
-		INFINIPATH_HWE_COREPLL_FBSLIP |
-		INFINIPATH_HWE_COREPLL_RFSLIP |
-		INFINIPATH_HWE_SERDESPLLFAILED |
-		INFINIPATH_HWE_IBCBUSTOSPCPARITYERR |
-		INFINIPATH_HWE_IBCBUSFRSPCPARITYERR |
-		INFINIPATH_HWE_PCIECPLDATAQUEUEERR |
-		INFINIPATH_HWE_PCIECPLHDRQUEUEERR |
-		INFINIPATH_HWE_SDMAMEMREADERR |
-		INFINIPATH_HWE_CLK_UC_PLLNOTLOCKED |
-		INFINIPATH_HWE_PCIESERDESQ0PCLKNOTDETECT |
-		INFINIPATH_HWE_PCIESERDESQ1PCLKNOTDETECT |
-		INFINIPATH_HWE_PCIESERDESQ2PCLKNOTDETECT |
-		INFINIPATH_HWE_PCIESERDESQ3PCLKNOTDETECT |
-		INFINIPATH_HWE_DDSRXEQMEMORYPARITYERR |
-		INFINIPATH_HWE_IB_UC_MEMORYPARITYERR |
-		INFINIPATH_HWE_PCIE_UC_OCT0MEMORYPARITYERR |
-		INFINIPATH_HWE_PCIE_UC_OCT1MEMORYPARITYERR;
-	dd->ipath_i_bitsextant =
-		INFINIPATH_I_SDMAINT | INFINIPATH_I_SDMADISABLED |
-		(INFINIPATH_I_RCVURG_MASK << INFINIPATH_I_RCVURG_SHIFT) |
-		(INFINIPATH_I_RCVAVAIL_MASK <<
-		 INFINIPATH_I_RCVAVAIL_SHIFT) |
-		INFINIPATH_I_ERROR | INFINIPATH_I_SPIOSENT |
-		INFINIPATH_I_SPIOBUFAVAIL | INFINIPATH_I_GPIO |
-		INFINIPATH_I_JINT | INFINIPATH_I_SERDESTRIMDONE;
-	dd->ipath_e_bitsextant =
-		INFINIPATH_E_RFORMATERR | INFINIPATH_E_RVCRC |
-		INFINIPATH_E_RICRC | INFINIPATH_E_RMINPKTLEN |
-		INFINIPATH_E_RMAXPKTLEN | INFINIPATH_E_RLONGPKTLEN |
-		INFINIPATH_E_RSHORTPKTLEN | INFINIPATH_E_RUNEXPCHAR |
-		INFINIPATH_E_RUNSUPVL | INFINIPATH_E_REBP |
-		INFINIPATH_E_RIBFLOW | INFINIPATH_E_RBADVERSION |
-		INFINIPATH_E_RRCVEGRFULL | INFINIPATH_E_RRCVHDRFULL |
-		INFINIPATH_E_RBADTID | INFINIPATH_E_RHDRLEN |
-		INFINIPATH_E_RHDR | INFINIPATH_E_RIBLOSTLINK |
-		INFINIPATH_E_SENDSPECIALTRIGGER |
-		INFINIPATH_E_SDMADISABLED | INFINIPATH_E_SMINPKTLEN |
-		INFINIPATH_E_SMAXPKTLEN | INFINIPATH_E_SUNDERRUN |
-		INFINIPATH_E_SPKTLEN | INFINIPATH_E_SDROPPEDSMPPKT |
-		INFINIPATH_E_SDROPPEDDATAPKT |
-		INFINIPATH_E_SPIOARMLAUNCH | INFINIPATH_E_SUNEXPERRPKTNUM |
-		INFINIPATH_E_SUNSUPVL | INFINIPATH_E_SENDBUFMISUSE |
-		INFINIPATH_E_SDMAGENMISMATCH | INFINIPATH_E_SDMAOUTOFBOUND |
-		INFINIPATH_E_SDMATAILOUTOFBOUND | INFINIPATH_E_SDMABASE |
-		INFINIPATH_E_SDMA1STDESC | INFINIPATH_E_SDMARPYTAG |
-		INFINIPATH_E_SDMADWEN | INFINIPATH_E_SDMAMISSINGDW |
-		INFINIPATH_E_SDMAUNEXPDATA |
-		INFINIPATH_E_IBSTATUSCHANGED | INFINIPATH_E_INVALIDADDR |
-		INFINIPATH_E_RESET | INFINIPATH_E_HARDWARE |
-		INFINIPATH_E_SDMADESCADDRMISALIGN |
-		INFINIPATH_E_INVALIDEEPCMD;
-
-	dd->ipath_i_rcvavail_mask = INFINIPATH_I_RCVAVAIL_MASK;
-	dd->ipath_i_rcvurg_mask = INFINIPATH_I_RCVURG_MASK;
-	dd->ipath_i_rcvavail_shift = INFINIPATH_I_RCVAVAIL_SHIFT;
-	dd->ipath_i_rcvurg_shift = INFINIPATH_I_RCVURG_SHIFT;
-	dd->ipath_flags |= IPATH_INTREG_64 | IPATH_HAS_MULT_IB_SPEED
-		| IPATH_HAS_LINK_LATENCY;
-
-	/*
-	 * EEPROM error log 0 is TXE Parity errors. 1 is RXE Parity.
-	 * 2 is Some Misc, 3 is reserved for future.
-	 */
-	dd->ipath_eep_st_masks[0].hwerrs_to_log =
-		INFINIPATH_HWE_TXEMEMPARITYERR_MASK <<
-		INFINIPATH_HWE_TXEMEMPARITYERR_SHIFT;
-
-	dd->ipath_eep_st_masks[1].hwerrs_to_log =
-		INFINIPATH_HWE_RXEMEMPARITYERR_MASK <<
-		INFINIPATH_HWE_RXEMEMPARITYERR_SHIFT;
-
-	dd->ipath_eep_st_masks[2].errs_to_log = INFINIPATH_E_RESET;
-
-	ipath_linkrecovery = 0;
-
-	init_waitqueue_head(&dd->ipath_autoneg_wait);
-	INIT_DELAYED_WORK(&dd->ipath_autoneg_work,  autoneg_work);
-
-	dd->ipath_link_width_supported = IB_WIDTH_1X | IB_WIDTH_4X;
-	dd->ipath_link_speed_supported = IPATH_IB_SDR | IPATH_IB_DDR;
-
-	dd->ipath_link_width_enabled = dd->ipath_link_width_supported;
-	dd->ipath_link_speed_enabled = dd->ipath_link_speed_supported;
-	/*
-	 * set the initial values to reasonable default, will be set
-	 * for real when link is up.
-	 */
-	dd->ipath_link_width_active = IB_WIDTH_4X;
-	dd->ipath_link_speed_active = IPATH_IB_SDR;
-	dd->delay_mult = rate_to_delay[0][1];
-}
-
-
-/*
- * Setup the MSI stuff again after a reset.  I'd like to just call
- * pci_enable_msi() and request_irq() again, but when I do that,
- * the MSI enable bit doesn't get set in the command word, and
- * we switch to to a different interrupt vector, which is confusing,
- * so I instead just do it all inline.  Perhaps somehow can tie this
- * into the PCIe hotplug support at some point
- * Note, because I'm doing it all here, I don't call pci_disable_msi()
- * or free_irq() at the start of ipath_setup_7220_reset().
- */
-static int ipath_reinit_msi(struct ipath_devdata *dd)
-{
-	int ret = 0;
-
-	int pos;
-	u16 control;
-	if (!dd->ipath_msi_lo) /* Using intX, or init problem */
-		goto bail;
-
-	pos = pci_find_capability(dd->pcidev, PCI_CAP_ID_MSI);
-	if (!pos) {
-		ipath_dev_err(dd, "Can't find MSI capability, "
-			      "can't restore MSI settings\n");
-		goto bail;
-	}
-	ipath_cdbg(VERBOSE, "Writing msi_lo 0x%x to config offset 0x%x\n",
-		   dd->ipath_msi_lo, pos + PCI_MSI_ADDRESS_LO);
-	pci_write_config_dword(dd->pcidev, pos + PCI_MSI_ADDRESS_LO,
-			       dd->ipath_msi_lo);
-	ipath_cdbg(VERBOSE, "Writing msi_lo 0x%x to config offset 0x%x\n",
-		   dd->ipath_msi_hi, pos + PCI_MSI_ADDRESS_HI);
-	pci_write_config_dword(dd->pcidev, pos + PCI_MSI_ADDRESS_HI,
-			       dd->ipath_msi_hi);
-	pci_read_config_word(dd->pcidev, pos + PCI_MSI_FLAGS, &control);
-	if (!(control & PCI_MSI_FLAGS_ENABLE)) {
-		ipath_cdbg(VERBOSE, "MSI control at off %x was %x, "
-			   "setting MSI enable (%x)\n", pos + PCI_MSI_FLAGS,
-			   control, control | PCI_MSI_FLAGS_ENABLE);
-		control |= PCI_MSI_FLAGS_ENABLE;
-		pci_write_config_word(dd->pcidev, pos + PCI_MSI_FLAGS,
-				      control);
-	}
-	/* now rewrite the data (vector) info */
-	pci_write_config_word(dd->pcidev, pos +
-			      ((control & PCI_MSI_FLAGS_64BIT) ? 12 : 8),
-			      dd->ipath_msi_data);
-	ret = 1;
-
-bail:
-	if (!ret) {
-		ipath_dbg("Using INTx, MSI disabled or not configured\n");
-		ipath_enable_intx(dd->pcidev);
-		ret = 1;
-	}
-	/*
-	 * We restore the cachelinesize also, although it doesn't really
-	 * matter.
-	 */
-	pci_write_config_byte(dd->pcidev, PCI_CACHE_LINE_SIZE,
-			      dd->ipath_pci_cacheline);
-	/* and now set the pci master bit again */
-	pci_set_master(dd->pcidev);
-
-	return ret;
-}
-
-/*
- * This routine sleeps, so it can only be called from user context, not
- * from interrupt context.  If we need interrupt context, we can split
- * it into two routines.
- */
-static int ipath_setup_7220_reset(struct ipath_devdata *dd)
-{
-	u64 val;
-	int i;
-	int ret;
-	u16 cmdval;
-
-	pci_read_config_word(dd->pcidev, PCI_COMMAND, &cmdval);
-
-	/* Use dev_err so it shows up in logs, etc. */
-	ipath_dev_err(dd, "Resetting InfiniPath unit %u\n", dd->ipath_unit);
-
-	/* keep chip from being accessed in a few places */
-	dd->ipath_flags &= ~(IPATH_INITTED | IPATH_PRESENT);
-	val = dd->ipath_control | INFINIPATH_C_RESET;
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_control, val);
-	mb();
-
-	for (i = 1; i <= 5; i++) {
-		int r;
-
-		/*
-		 * Allow MBIST, etc. to complete; longer on each retry.
-		 * We sometimes get machine checks from bus timeout if no
-		 * response, so for now, make it *really* long.
-		 */
-		msleep(1000 + (1 + i) * 2000);
-		r = pci_write_config_dword(dd->pcidev, PCI_BASE_ADDRESS_0,
-					   dd->ipath_pcibar0);
-		if (r)
-			ipath_dev_err(dd, "rewrite of BAR0 failed: %d\n", r);
-		r = pci_write_config_dword(dd->pcidev, PCI_BASE_ADDRESS_1,
-					   dd->ipath_pcibar1);
-		if (r)
-			ipath_dev_err(dd, "rewrite of BAR1 failed: %d\n", r);
-		/* now re-enable memory access */
-		pci_write_config_word(dd->pcidev, PCI_COMMAND, cmdval);
-		r = pci_enable_device(dd->pcidev);
-		if (r)
-			ipath_dev_err(dd, "pci_enable_device failed after "
-				      "reset: %d\n", r);
-		/*
-		 * whether it fully enabled or not, mark as present,
-		 * again (but not INITTED)
-		 */
-		dd->ipath_flags |= IPATH_PRESENT;
-		val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_revision);
-		if (val == dd->ipath_revision) {
-			ipath_cdbg(VERBOSE, "Got matching revision "
-				   "register %llx on try %d\n",
-				   (unsigned long long) val, i);
-			ret = ipath_reinit_msi(dd);
-			goto bail;
-		}
-		/* Probably getting -1 back */
-		ipath_dbg("Didn't get expected revision register, "
-			  "got %llx, try %d\n", (unsigned long long) val,
-			  i + 1);
-	}
-	ret = 0; /* failed */
-
-bail:
-	if (ret)
-		ipath_7220_pcie_params(dd, dd->ipath_boardrev);
-
-	return ret;
-}
-
-/**
- * ipath_7220_put_tid - write a TID to the chip
- * @dd: the infinipath device
- * @tidptr: pointer to the expected TID (in chip) to update
- * @tidtype: 0 for eager, 1 for expected
- * @pa: physical address of in memory buffer; ipath_tidinvalid if freeing
- *
- * This exists as a separate routine to allow for selection of the
- * appropriate "flavor". The static calls in cleanup just use the
- * revision-agnostic form, as they are not performance critical.
- */
-static void ipath_7220_put_tid(struct ipath_devdata *dd, u64 __iomem *tidptr,
-			     u32 type, unsigned long pa)
-{
-	if (pa != dd->ipath_tidinvalid) {
-		u64 chippa = pa >> IBA7220_TID_PA_SHIFT;
-
-		/* paranoia checks */
-		if (pa != (chippa << IBA7220_TID_PA_SHIFT)) {
-			dev_info(&dd->pcidev->dev, "BUG: physaddr %lx "
-				 "not 2KB aligned!\n", pa);
-			return;
-		}
-		if (chippa >= (1UL << IBA7220_TID_SZ_SHIFT)) {
-			ipath_dev_err(dd,
-				      "BUG: Physical page address 0x%lx "
-				      "larger than supported\n", pa);
-			return;
-		}
-
-		if (type == RCVHQ_RCV_TYPE_EAGER)
-			chippa |= dd->ipath_tidtemplate;
-		else /* for now, always full 4KB page */
-			chippa |= IBA7220_TID_SZ_4K;
-		writeq(chippa, tidptr);
-	} else
-		writeq(pa, tidptr);
-	mmiowb();
-}
-
-/**
- * ipath_7220_clear_tid - clear all TID entries for a port, expected and eager
- * @dd: the infinipath device
- * @port: the port
- *
- * clear all TID entries for a port, expected and eager.
- * Used from ipath_close().  On this chip, TIDs are only 32 bits,
- * not 64, but they are still on 64 bit boundaries, so tidbase
- * is declared as u64 * for the pointer math, even though we write 32 bits
- */
-static void ipath_7220_clear_tids(struct ipath_devdata *dd, unsigned port)
-{
-	u64 __iomem *tidbase;
-	unsigned long tidinv;
-	int i;
-
-	if (!dd->ipath_kregbase)
-		return;
-
-	ipath_cdbg(VERBOSE, "Invalidate TIDs for port %u\n", port);
-
-	tidinv = dd->ipath_tidinvalid;
-	tidbase = (u64 __iomem *)
-		((char __iomem *)(dd->ipath_kregbase) +
-		 dd->ipath_rcvtidbase +
-		 port * dd->ipath_rcvtidcnt * sizeof(*tidbase));
-
-	for (i = 0; i < dd->ipath_rcvtidcnt; i++)
-		ipath_7220_put_tid(dd, &tidbase[i], RCVHQ_RCV_TYPE_EXPECTED,
-				   tidinv);
-
-	tidbase = (u64 __iomem *)
-		((char __iomem *)(dd->ipath_kregbase) +
-		 dd->ipath_rcvegrbase + port_egrtid_idx(dd, port)
-		 * sizeof(*tidbase));
-
-	for (i = port ? dd->ipath_rcvegrcnt : dd->ipath_p0_rcvegrcnt; i; i--)
-		ipath_7220_put_tid(dd, &tidbase[i-1], RCVHQ_RCV_TYPE_EAGER,
-			tidinv);
-}
-
-/**
- * ipath_7220_tidtemplate - setup constants for TID updates
- * @dd: the infinipath device
- *
- * We setup stuff that we use a lot, to avoid calculating each time
- */
-static void ipath_7220_tidtemplate(struct ipath_devdata *dd)
-{
-	/* For now, we always allocate 4KB buffers (at init) so we can
-	 * receive max size packets.  We may want a module parameter to
-	 * specify 2KB or 4KB and/or make be per port instead of per device
-	 * for those who want to reduce memory footprint.  Note that the
-	 * ipath_rcvhdrentsize size must be large enough to hold the largest
-	 * IB header (currently 96 bytes) that we expect to handle (plus of
-	 * course the 2 dwords of RHF).
-	 */
-	if (dd->ipath_rcvegrbufsize == 2048)
-		dd->ipath_tidtemplate = IBA7220_TID_SZ_2K;
-	else if (dd->ipath_rcvegrbufsize == 4096)
-		dd->ipath_tidtemplate = IBA7220_TID_SZ_4K;
-	else {
-		dev_info(&dd->pcidev->dev, "BUG: unsupported egrbufsize "
-			 "%u, using %u\n", dd->ipath_rcvegrbufsize,
-			 4096);
-		dd->ipath_tidtemplate = IBA7220_TID_SZ_4K;
-	}
-	dd->ipath_tidinvalid = 0;
-}
-
-static int ipath_7220_early_init(struct ipath_devdata *dd)
-{
-	u32 i, s;
-
-	if (strcmp(int_type, "auto") &&
-	    strcmp(int_type, "force_msi") &&
-	    strcmp(int_type, "force_intx")) {
-		ipath_dev_err(dd, "Invalid interrupt_type: '%s', expecting "
-			      "auto, force_msi or force_intx\n", int_type);
-		return -EINVAL;
-	}
-
-	/*
-	 * Control[4] has been added to change the arbitration within
-	 * the SDMA engine between favoring data fetches over descriptor
-	 * fetches.  ipath_sdma_fetch_arb==0 gives data fetches priority.
-	 */
-	if (ipath_sdma_fetch_arb && (dd->ipath_minrev > 1))
-		dd->ipath_control |= 1<<4;
-
-	dd->ipath_flags |= IPATH_4BYTE_TID;
-
-	/*
-	 * For openfabrics, we need to be able to handle an IB header of
-	 * 24 dwords.  HT chip has arbitrary sized receive buffers, so we
-	 * made them the same size as the PIO buffers.  This chip does not
-	 * handle arbitrary size buffers, so we need the header large enough
-	 * to handle largest IB header, but still have room for a 2KB MTU
-	 * standard IB packet.
-	 */
-	dd->ipath_rcvhdrentsize = 24;
-	dd->ipath_rcvhdrsize = IPATH_DFLT_RCVHDRSIZE;
-	dd->ipath_rhf_offset =
-		dd->ipath_rcvhdrentsize - sizeof(u64) / sizeof(u32);
-
-	dd->ipath_rcvegrbufsize = ipath_mtu4096 ? 4096 : 2048;
-	/*
-	 * the min() check here is currently a nop, but it may not always
-	 * be, depending on just how we do ipath_rcvegrbufsize
-	 */
-	dd->ipath_ibmaxlen = min(ipath_mtu4096 ? dd->ipath_piosize4k :
-				 dd->ipath_piosize2k,
-				 dd->ipath_rcvegrbufsize +
-				 (dd->ipath_rcvhdrentsize << 2));
-	dd->ipath_init_ibmaxlen = dd->ipath_ibmaxlen;
-
-	ipath_7220_config_jint(dd, INFINIPATH_JINT_DEFAULT_IDLE_TICKS,
-			       INFINIPATH_JINT_DEFAULT_MAX_PACKETS);
-
-	if (dd->ipath_boardrev) /* no eeprom on emulator */
-		ipath_get_eeprom_info(dd);
-
-	/* start of code to check and print procmon */
-	s = ipath_read_kreg32(dd, IPATH_KREG_OFFSET(ProcMon));
-	s &= ~(1U<<31); /* clear done bit */
-	s |= 1U<<14; /* clear counter (write 1 to clear) */
-	ipath_write_kreg(dd, IPATH_KREG_OFFSET(ProcMon), s);
-	/* make sure clear_counter low long enough before start */
-	ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
-	ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
-
-	s &= ~(1U<<14); /* allow counter to count (before starting) */
-	ipath_write_kreg(dd, IPATH_KREG_OFFSET(ProcMon), s);
-	ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
-	ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
-	s = ipath_read_kreg32(dd, IPATH_KREG_OFFSET(ProcMon));
-
-	s |= 1U<<15; /* start the counter */
-	s &= ~(1U<<31); /* clear done bit */
-	s &= ~0x7ffU; /* clear frequency bits */
-	s |= 0xe29; /* set frequency bits, in case cleared */
-	ipath_write_kreg(dd, IPATH_KREG_OFFSET(ProcMon), s);
-
-	s = 0;
-	for (i = 500; i > 0 && !(s&(1ULL<<31)); i--) {
-		ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
-		s = ipath_read_kreg32(dd, IPATH_KREG_OFFSET(ProcMon));
-	}
-	if (!(s&(1U<<31)))
-		ipath_dev_err(dd, "ProcMon register not valid: 0x%x\n", s);
-	else
-		ipath_dbg("ProcMon=0x%x, count=0x%x\n", s, (s>>16)&0x1ff);
-
-	return 0;
-}
-
-/**
- * ipath_init_7220_get_base_info - set chip-specific flags for user code
- * @pd: the infinipath port
- * @kbase: ipath_base_info pointer
- *
- * We set the PCIE flag because the lower bandwidth on PCIe vs
- * HyperTransport can affect some user packet algorithims.
- */
-static int ipath_7220_get_base_info(struct ipath_portdata *pd, void *kbase)
-{
-	struct ipath_base_info *kinfo = kbase;
-
-	kinfo->spi_runtime_flags |=
-		IPATH_RUNTIME_PCIE | IPATH_RUNTIME_NODMA_RTAIL |
-		IPATH_RUNTIME_SDMA;
-
-	return 0;
-}
-
-static void ipath_7220_free_irq(struct ipath_devdata *dd)
-{
-	free_irq(dd->ipath_irq, dd);
-	dd->ipath_irq = 0;
-}
-
-static struct ipath_message_header *
-ipath_7220_get_msgheader(struct ipath_devdata *dd, __le32 *rhf_addr)
-{
-	u32 offset = ipath_hdrget_offset(rhf_addr);
-
-	return (struct ipath_message_header *)
-		(rhf_addr - dd->ipath_rhf_offset + offset);
-}
-
-static void ipath_7220_config_ports(struct ipath_devdata *dd, ushort cfgports)
-{
-	u32 nchipports;
-
-	nchipports = ipath_read_kreg32(dd, dd->ipath_kregs->kr_portcnt);
-	if (!cfgports) {
-		int ncpus = num_online_cpus();
-
-		if (ncpus <= 4)
-			dd->ipath_portcnt = 5;
-		else if (ncpus <= 8)
-			dd->ipath_portcnt = 9;
-		if (dd->ipath_portcnt)
-			ipath_dbg("Auto-configured for %u ports, %d cpus "
-				"online\n", dd->ipath_portcnt, ncpus);
-	} else if (cfgports <= nchipports)
-		dd->ipath_portcnt = cfgports;
-	if (!dd->ipath_portcnt) /* none of the above, set to max */
-		dd->ipath_portcnt = nchipports;
-	/*
-	 * chip can be configured for 5, 9, or 17 ports, and choice
-	 * affects number of eager TIDs per port (1K, 2K, 4K).
-	 */
-	if (dd->ipath_portcnt > 9)
-		dd->ipath_rcvctrl |= 2ULL << IBA7220_R_PORTCFG_SHIFT;
-	else if (dd->ipath_portcnt > 5)
-		dd->ipath_rcvctrl |= 1ULL << IBA7220_R_PORTCFG_SHIFT;
-	/* else configure for default 5 receive ports */
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_rcvctrl,
-			 dd->ipath_rcvctrl);
-	dd->ipath_p0_rcvegrcnt = 2048; /* always */
-	if (dd->ipath_flags & IPATH_HAS_SEND_DMA)
-		dd->ipath_pioreserved = 3; /* kpiobufs used for PIO */
-}
-
-
-static int ipath_7220_get_ib_cfg(struct ipath_devdata *dd, int which)
-{
-	int lsb, ret = 0;
-	u64 maskr; /* right-justified mask */
-
-	switch (which) {
-	case IPATH_IB_CFG_HRTBT: /* Get Heartbeat off/enable/auto */
-		lsb = IBA7220_IBC_HRTBT_SHIFT;
-		maskr = IBA7220_IBC_HRTBT_MASK;
-		break;
-
-	case IPATH_IB_CFG_LWID_ENB: /* Get allowed Link-width */
-		ret = dd->ipath_link_width_enabled;
-		goto done;
-
-	case IPATH_IB_CFG_LWID: /* Get currently active Link-width */
-		ret = dd->ipath_link_width_active;
-		goto done;
-
-	case IPATH_IB_CFG_SPD_ENB: /* Get allowed Link speeds */
-		ret = dd->ipath_link_speed_enabled;
-		goto done;
-
-	case IPATH_IB_CFG_SPD: /* Get current Link spd */
-		ret = dd->ipath_link_speed_active;
-		goto done;
-
-	case IPATH_IB_CFG_RXPOL_ENB: /* Get Auto-RX-polarity enable */
-		lsb = IBA7220_IBC_RXPOL_SHIFT;
-		maskr = IBA7220_IBC_RXPOL_MASK;
-		break;
-
-	case IPATH_IB_CFG_LREV_ENB: /* Get Auto-Lane-reversal enable */
-		lsb = IBA7220_IBC_LREV_SHIFT;
-		maskr = IBA7220_IBC_LREV_MASK;
-		break;
-
-	case IPATH_IB_CFG_LINKLATENCY:
-		ret = ipath_read_kreg64(dd, dd->ipath_kregs->kr_ibcddrstatus)
-			& IBA7220_DDRSTAT_LINKLAT_MASK;
-		goto done;
-
-	default:
-		ret = -ENOTSUPP;
-		goto done;
-	}
-	ret = (int)((dd->ipath_ibcddrctrl >> lsb) & maskr);
-done:
-	return ret;
-}
-
-static int ipath_7220_set_ib_cfg(struct ipath_devdata *dd, int which, u32 val)
-{
-	int lsb, ret = 0, setforce = 0;
-	u64 maskr; /* right-justified mask */
-
-	switch (which) {
-	case IPATH_IB_CFG_LIDLMC:
-		/*
-		 * Set LID and LMC. Combined to avoid possible hazard
-		 * caller puts LMC in 16MSbits, DLID in 16LSbits of val
-		 */
-		lsb = IBA7220_IBC_DLIDLMC_SHIFT;
-		maskr = IBA7220_IBC_DLIDLMC_MASK;
-		break;
-
-	case IPATH_IB_CFG_HRTBT: /* set Heartbeat off/enable/auto */
-		if (val & IPATH_IB_HRTBT_ON &&
-			(dd->ipath_flags & IPATH_NO_HRTBT))
-			goto bail;
-		lsb = IBA7220_IBC_HRTBT_SHIFT;
-		maskr = IBA7220_IBC_HRTBT_MASK;
-		break;
-
-	case IPATH_IB_CFG_LWID_ENB: /* set allowed Link-width */
-		/*
-		 * As with speed, only write the actual register if
-		 * the link is currently down, otherwise takes effect
-		 * on next link change.
-		 */
-		dd->ipath_link_width_enabled = val;
-		if ((dd->ipath_flags & (IPATH_LINKDOWN|IPATH_LINKINIT)) !=
-			IPATH_LINKDOWN)
-			goto bail;
-		/*
-		 * We set the IPATH_IB_FORCE_NOTIFY bit so updown
-		 * will get called because we want update
-		 * link_width_active, and the change may not take
-		 * effect for some time (if we are in POLL), so this
-		 * flag will force the updown routine to be called
-		 * on the next ibstatuschange down interrupt, even
-		 * if it's not an down->up transition.
-		 */
-		val--; /* convert from IB to chip */
-		maskr = IBA7220_IBC_WIDTH_MASK;
-		lsb = IBA7220_IBC_WIDTH_SHIFT;
-		setforce = 1;
-		dd->ipath_flags |= IPATH_IB_FORCE_NOTIFY;
-		break;
-
-	case IPATH_IB_CFG_SPD_ENB: /* set allowed Link speeds */
-		/*
-		 * If we turn off IB1.2, need to preset SerDes defaults,
-		 * but not right now. Set a flag for the next time
-		 * we command the link down.  As with width, only write the
-		 * actual register if the link is currently down, otherwise
-		 * takes effect on next link change.  Since setting is being
-		 * explictly requested (via MAD or sysfs), clear autoneg
-		 * failure status if speed autoneg is enabled.
-		 */
-		dd->ipath_link_speed_enabled = val;
-		if (dd->ipath_ibcddrctrl & IBA7220_IBC_IBTA_1_2_MASK &&
-		    !(val & (val - 1)))
-			dd->ipath_presets_needed = 1;
-		if ((dd->ipath_flags & (IPATH_LINKDOWN|IPATH_LINKINIT)) !=
-			IPATH_LINKDOWN)
-			goto bail;
-		/*
-		 * We set the IPATH_IB_FORCE_NOTIFY bit so updown
-		 * will get called because we want update
-		 * link_speed_active, and the change may not take
-		 * effect for some time (if we are in POLL), so this
-		 * flag will force the updown routine to be called
-		 * on the next ibstatuschange down interrupt, even
-		 * if it's not an down->up transition.  When setting
-		 * speed autoneg, clear AUTONEG_FAILED.
-		 */
-		if (val == (IPATH_IB_SDR | IPATH_IB_DDR)) {
-			val = IBA7220_IBC_SPEED_AUTONEG_MASK |
-				IBA7220_IBC_IBTA_1_2_MASK;
-			dd->ipath_flags &= ~IPATH_IB_AUTONEG_FAILED;
-		} else
-			val = val == IPATH_IB_DDR ?  IBA7220_IBC_SPEED_DDR
-				: IBA7220_IBC_SPEED_SDR;
-		maskr = IBA7220_IBC_SPEED_AUTONEG_MASK |
-			IBA7220_IBC_IBTA_1_2_MASK;
-		lsb = 0; /* speed bits are low bits */
-		setforce = 1;
-		break;
-
-	case IPATH_IB_CFG_RXPOL_ENB: /* set Auto-RX-polarity enable */
-		lsb = IBA7220_IBC_RXPOL_SHIFT;
-		maskr = IBA7220_IBC_RXPOL_MASK;
-		break;
-
-	case IPATH_IB_CFG_LREV_ENB: /* set Auto-Lane-reversal enable */
-		lsb = IBA7220_IBC_LREV_SHIFT;
-		maskr = IBA7220_IBC_LREV_MASK;
-		break;
-
-	default:
-		ret = -ENOTSUPP;
-		goto bail;
-	}
-	dd->ipath_ibcddrctrl &= ~(maskr << lsb);
-	dd->ipath_ibcddrctrl |= (((u64) val & maskr) << lsb);
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_ibcddrctrl,
-			 dd->ipath_ibcddrctrl);
-	if (setforce)
-		dd->ipath_flags |= IPATH_IB_FORCE_NOTIFY;
-bail:
-	return ret;
-}
-
-static void ipath_7220_read_counters(struct ipath_devdata *dd,
-				     struct infinipath_counters *cntrs)
-{
-	u64 *counters = (u64 *) cntrs;
-	int i;
-
-	for (i = 0; i < sizeof(*cntrs) / sizeof(u64); i++)
-		counters[i] = ipath_snap_cntr(dd, i);
-}
-
-/* if we are using MSI, try to fallback to INTx */
-static int ipath_7220_intr_fallback(struct ipath_devdata *dd)
-{
-	if (dd->ipath_msi_lo) {
-		dev_info(&dd->pcidev->dev, "MSI interrupt not detected,"
-			" trying INTx interrupts\n");
-		ipath_7220_nomsi(dd);
-		ipath_enable_intx(dd->pcidev);
-		/*
-		 * some newer kernels require free_irq before disable_msi,
-		 * and irq can be changed during disable and intx enable
-		 * and we need to therefore use the pcidev->irq value,
-		 * not our saved MSI value.
-		 */
-		dd->ipath_irq = dd->pcidev->irq;
-		if (request_irq(dd->ipath_irq, ipath_intr, IRQF_SHARED,
-			IPATH_DRV_NAME, dd))
-			ipath_dev_err(dd,
-				"Could not re-request_irq for INTx\n");
-		return 1;
-	}
-	return 0;
-}
-
-/*
- * reset the XGXS (between serdes and IBC).  Slightly less intrusive
- * than resetting the IBC or external link state, and useful in some
- * cases to cause some retraining.  To do this right, we reset IBC
- * as well.
- */
-static void ipath_7220_xgxs_reset(struct ipath_devdata *dd)
-{
-	u64 val, prev_val;
-
-	prev_val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_xgxsconfig);
-	val = prev_val | INFINIPATH_XGXS_RESET;
-	prev_val &= ~INFINIPATH_XGXS_RESET; /* be sure */
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_control,
-			 dd->ipath_control & ~INFINIPATH_C_LINKENABLE);
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_xgxsconfig, val);
-	ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_xgxsconfig, prev_val);
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_control,
-			 dd->ipath_control);
-}
-
-
-/* Still needs cleanup, too much hardwired stuff */
-static void autoneg_send(struct ipath_devdata *dd,
-	u32 *hdr, u32 dcnt, u32 *data)
-{
-	int i;
-	u64 cnt;
-	u32 __iomem *piobuf;
-	u32 pnum;
-
-	i = 0;
-	cnt = 7 + dcnt + 1; /* 7 dword header, dword data, icrc */
-	while (!(piobuf = ipath_getpiobuf(dd, cnt, &pnum))) {
-		if (i++ > 15) {
-			ipath_dbg("Couldn't get pio buffer for send\n");
-			return;
-		}
-		udelay(2);
-	}
-	if (dd->ipath_flags&IPATH_HAS_PBC_CNT)
-		cnt |= 0x80000000UL<<32; /* mark as VL15 */
-	writeq(cnt, piobuf);
-	ipath_flush_wc();
-	__iowrite32_copy(piobuf + 2, hdr, 7);
-	__iowrite32_copy(piobuf + 9, data, dcnt);
-	ipath_flush_wc();
-}
-
-/*
- * _start packet gets sent twice at start, _done gets sent twice at end
- */
-static void ipath_autoneg_send(struct ipath_devdata *dd, int which)
-{
-	static u32 swapped;
-	u32 dw, i, hcnt, dcnt, *data;
-	static u32 hdr[7] = { 0xf002ffff, 0x48ffff, 0x6400abba };
-	static u32 madpayload_start[0x40] = {
-		0x1810103, 0x1, 0x0, 0x0, 0x2c90000, 0x2c9, 0x0, 0x0,
-		0xffffffff, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0,
-		0x1, 0x1388, 0x15e, 0x1, /* rest 0's */
-		};
-	static u32 madpayload_done[0x40] = {
-		0x1810103, 0x1, 0x0, 0x0, 0x2c90000, 0x2c9, 0x0, 0x0,
-		0xffffffff, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0,
-		0x40000001, 0x1388, 0x15e, /* rest 0's */
-		};
-	dcnt = ARRAY_SIZE(madpayload_start);
-	hcnt = ARRAY_SIZE(hdr);
-	if (!swapped) {
-		/* for maintainability, do it at runtime */
-		for (i = 0; i < hcnt; i++) {
-			dw = (__force u32) cpu_to_be32(hdr[i]);
-			hdr[i] = dw;
-		}
-		for (i = 0; i < dcnt; i++) {
-			dw = (__force u32) cpu_to_be32(madpayload_start[i]);
-			madpayload_start[i] = dw;
-			dw = (__force u32) cpu_to_be32(madpayload_done[i]);
-			madpayload_done[i] = dw;
-		}
-		swapped = 1;
-	}
-
-	data = which ? madpayload_done : madpayload_start;
-	ipath_cdbg(PKT, "Sending %s special MADs\n", which?"done":"start");
-
-	autoneg_send(dd, hdr, dcnt, data);
-	ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
-	udelay(2);
-	autoneg_send(dd, hdr, dcnt, data);
-	ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
-	udelay(2);
-}
-
-
-
-/*
- * Do the absolute minimum to cause an IB speed change, and make it
- * ready, but don't actually trigger the change.   The caller will
- * do that when ready (if link is in Polling training state, it will
- * happen immediately, otherwise when link next goes down)
- *
- * This routine should only be used as part of the DDR autonegotation
- * code for devices that are not compliant with IB 1.2 (or code that
- * fixes things up for same).
- *
- * When link has gone down, and autoneg enabled, or autoneg has
- * failed and we give up until next time we set both speeds, and
- * then we want IBTA enabled as well as "use max enabled speed.
- */
-static void set_speed_fast(struct ipath_devdata *dd, u32 speed)
-{
-	dd->ipath_ibcddrctrl &= ~(IBA7220_IBC_SPEED_AUTONEG_MASK |
-		IBA7220_IBC_IBTA_1_2_MASK |
-		(IBA7220_IBC_WIDTH_MASK << IBA7220_IBC_WIDTH_SHIFT));
-
-	if (speed == (IPATH_IB_SDR | IPATH_IB_DDR))
-		dd->ipath_ibcddrctrl |= IBA7220_IBC_SPEED_AUTONEG_MASK |
-			IBA7220_IBC_IBTA_1_2_MASK;
-	else
-		dd->ipath_ibcddrctrl |= speed == IPATH_IB_DDR ?
-			IBA7220_IBC_SPEED_DDR : IBA7220_IBC_SPEED_SDR;
-
-	/*
-	 * Convert from IB-style 1 = 1x, 2 = 4x, 3 = auto
-	 * to chip-centric       0 = 1x, 1 = 4x, 2 = auto
-	 */
-	dd->ipath_ibcddrctrl |= (u64)(dd->ipath_link_width_enabled - 1) <<
-		IBA7220_IBC_WIDTH_SHIFT;
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_ibcddrctrl,
-			dd->ipath_ibcddrctrl);
-	ipath_cdbg(VERBOSE, "setup for IB speed (%x) done\n", speed);
-}
-
-
-/*
- * this routine is only used when we are not talking to another
- * IB 1.2-compliant device that we think can do DDR.
- * (This includes all existing switch chips as of Oct 2007.)
- * 1.2-compliant devices go directly to DDR prior to reaching INIT
- */
-static void try_auto_neg(struct ipath_devdata *dd)
-{
-	/*
-	 * required for older non-IB1.2 DDR switches.  Newer
-	 * non-IB-compliant switches don't need it, but so far,
-	 * aren't bothered by it either.  "Magic constant"
-	 */
-	ipath_write_kreg(dd, IPATH_KREG_OFFSET(IBNCModeCtrl),
-		0x3b9dc07);
-	dd->ipath_flags |= IPATH_IB_AUTONEG_INPROG;
-	ipath_autoneg_send(dd, 0);
-	set_speed_fast(dd, IPATH_IB_DDR);
-	ipath_toggle_rclkrls(dd);
-	/* 2 msec is minimum length of a poll cycle */
-	schedule_delayed_work(&dd->ipath_autoneg_work,
-		msecs_to_jiffies(2));
-}
-
-
-static int ipath_7220_ib_updown(struct ipath_devdata *dd, int ibup, u64 ibcs)
-{
-	int ret = 0, symadj = 0;
-	u32 ltstate = ipath_ib_linkstate(dd, ibcs);
-
-	dd->ipath_link_width_active =
-		((ibcs >> IBA7220_IBCS_LINKWIDTH_SHIFT) & 1) ?
-		    IB_WIDTH_4X : IB_WIDTH_1X;
-	dd->ipath_link_speed_active =
-		((ibcs >> IBA7220_IBCS_LINKSPEED_SHIFT) & 1) ?
-		    IPATH_IB_DDR : IPATH_IB_SDR;
-
-	if (!ibup) {
-		/*
-		 * when link goes down we don't want aeq running, so it
-		 * won't't interfere with IBC training, etc., and we need
-		 * to go back to the static SerDes preset values
-		 */
-		if (dd->ipath_x1_fix_tries &&
-			 ltstate <= INFINIPATH_IBCS_LT_STATE_SLEEPQUIET &&
-			ltstate != INFINIPATH_IBCS_LT_STATE_LINKUP)
-			dd->ipath_x1_fix_tries = 0;
-		if (!(dd->ipath_flags & (IPATH_IB_AUTONEG_FAILED |
-			IPATH_IB_AUTONEG_INPROG)))
-			set_speed_fast(dd, dd->ipath_link_speed_enabled);
-		if (!(dd->ipath_flags & IPATH_IB_AUTONEG_INPROG)) {
-			ipath_cdbg(VERBOSE, "Setting RXEQ defaults\n");
-			ipath_sd7220_presets(dd);
-		}
-		/* this might better in ipath_sd7220_presets() */
-		ipath_set_relock_poll(dd, ibup);
-	} else {
-		if (ipath_compat_ddr_negotiate &&
-		    !(dd->ipath_flags & (IPATH_IB_AUTONEG_FAILED |
-			IPATH_IB_AUTONEG_INPROG)) &&
-			dd->ipath_link_speed_active == IPATH_IB_SDR &&
-			(dd->ipath_link_speed_enabled &
-			    (IPATH_IB_DDR | IPATH_IB_SDR)) ==
-			    (IPATH_IB_DDR | IPATH_IB_SDR) &&
-			dd->ipath_autoneg_tries < IPATH_AUTONEG_TRIES) {
-			/* we are SDR, and DDR auto-negotiation enabled */
-			++dd->ipath_autoneg_tries;
-			ipath_dbg("DDR negotiation try, %u/%u\n",
-				dd->ipath_autoneg_tries,
-				IPATH_AUTONEG_TRIES);
-			if (!dd->ibdeltainprog) {
-				dd->ibdeltainprog = 1;
-				dd->ibsymsnap = ipath_read_creg32(dd,
-					dd->ipath_cregs->cr_ibsymbolerrcnt);
-				dd->iblnkerrsnap = ipath_read_creg32(dd,
-					dd->ipath_cregs->cr_iblinkerrrecovcnt);
-			}
-			try_auto_neg(dd);
-			ret = 1; /* no other IB status change processing */
-		} else if ((dd->ipath_flags & IPATH_IB_AUTONEG_INPROG)
-			&& dd->ipath_link_speed_active == IPATH_IB_SDR) {
-			ipath_autoneg_send(dd, 1);
-			set_speed_fast(dd, IPATH_IB_DDR);
-			udelay(2);
-			ipath_toggle_rclkrls(dd);
-			ret = 1; /* no other IB status change processing */
-		} else {
-			if ((dd->ipath_flags & IPATH_IB_AUTONEG_INPROG) &&
-				(dd->ipath_link_speed_active & IPATH_IB_DDR)) {
-				ipath_dbg("Got to INIT with DDR autoneg\n");
-				dd->ipath_flags &= ~(IPATH_IB_AUTONEG_INPROG
-					| IPATH_IB_AUTONEG_FAILED);
-				dd->ipath_autoneg_tries = 0;
-				/* re-enable SDR, for next link down */
-				set_speed_fast(dd,
-					dd->ipath_link_speed_enabled);
-				wake_up(&dd->ipath_autoneg_wait);
-				symadj = 1;
-			} else if (dd->ipath_flags & IPATH_IB_AUTONEG_FAILED) {
-				/*
-				 * clear autoneg failure flag, and do setup
-				 * so we'll try next time link goes down and
-				 * back to INIT (possibly connected to different
-				 * device).
-				 */
-				ipath_dbg("INIT %sDR after autoneg failure\n",
-					(dd->ipath_link_speed_active &
-					  IPATH_IB_DDR) ? "D" : "S");
-				dd->ipath_flags &= ~IPATH_IB_AUTONEG_FAILED;
-				dd->ipath_ibcddrctrl |=
-					IBA7220_IBC_IBTA_1_2_MASK;
-				ipath_write_kreg(dd,
-					IPATH_KREG_OFFSET(IBNCModeCtrl), 0);
-				symadj = 1;
-			}
-		}
-		/*
-		 * if we are in 1X on rev1 only, and are in autoneg width,
-		 * it could be due to an xgxs problem, so if we haven't
-		 * already tried, try twice to get to 4X; if we
-		 * tried, and couldn't, report it, since it will
-		 * probably not be what is desired.
-		 */
-		if (dd->ipath_minrev == 1 &&
-		    (dd->ipath_link_width_enabled & (IB_WIDTH_1X |
-			IB_WIDTH_4X)) == (IB_WIDTH_1X | IB_WIDTH_4X)
-			&& dd->ipath_link_width_active == IB_WIDTH_1X
-			&& dd->ipath_x1_fix_tries < 3) {
-		     if (++dd->ipath_x1_fix_tries == 3) {
-				dev_info(&dd->pcidev->dev,
-					"IB link is in 1X mode\n");
-				if (!(dd->ipath_flags &
-				      IPATH_IB_AUTONEG_INPROG))
-					symadj = 1;
-		     }
-			else {
-				ipath_cdbg(VERBOSE, "IB 1X in "
-					"auto-width, try %u to be "
-					"sure it's really 1X; "
-					"ltstate %u\n",
-					 dd->ipath_x1_fix_tries,
-					 ltstate);
-				dd->ipath_f_xgxs_reset(dd);
-				ret = 1; /* skip other processing */
-			}
-		} else if (!(dd->ipath_flags & IPATH_IB_AUTONEG_INPROG))
-			symadj = 1;
-
-		if (!ret) {
-			dd->delay_mult = rate_to_delay
-			    [(ibcs >> IBA7220_IBCS_LINKSPEED_SHIFT) & 1]
-			    [(ibcs >> IBA7220_IBCS_LINKWIDTH_SHIFT) & 1];
-
-			ipath_set_relock_poll(dd, ibup);
-		}
-	}
-
-	if (symadj) {
-		if (dd->ibdeltainprog) {
-			dd->ibdeltainprog = 0;
-			dd->ibsymdelta += ipath_read_creg32(dd,
-				dd->ipath_cregs->cr_ibsymbolerrcnt) -
-				dd->ibsymsnap;
-			dd->iblnkerrdelta += ipath_read_creg32(dd,
-				dd->ipath_cregs->cr_iblinkerrrecovcnt) -
-				dd->iblnkerrsnap;
-		}
-	} else if (!ibup && !dd->ibdeltainprog
-		   && !(dd->ipath_flags & IPATH_IB_AUTONEG_INPROG)) {
-		dd->ibdeltainprog = 1;
-		dd->ibsymsnap =	ipath_read_creg32(dd,
-				     dd->ipath_cregs->cr_ibsymbolerrcnt);
-		dd->iblnkerrsnap = ipath_read_creg32(dd,
-				     dd->ipath_cregs->cr_iblinkerrrecovcnt);
-	}
-
-	if (!ret)
-		ipath_setup_7220_setextled(dd, ipath_ib_linkstate(dd, ibcs),
-			ltstate);
-	return ret;
-}
-
-
-/*
- * Handle the empirically determined mechanism for auto-negotiation
- * of DDR speed with switches.
- */
-static void autoneg_work(struct work_struct *work)
-{
-	struct ipath_devdata *dd;
-	u64 startms;
-	u32 lastlts, i;
-
-	dd = container_of(work, struct ipath_devdata,
-		ipath_autoneg_work.work);
-
-	startms = jiffies_to_msecs(jiffies);
-
-	/*
-	 * busy wait for this first part, it should be at most a
-	 * few hundred usec, since we scheduled ourselves for 2msec.
-	 */
-	for (i = 0; i < 25; i++) {
-		lastlts = ipath_ib_linktrstate(dd, dd->ipath_lastibcstat);
-		if (lastlts == INFINIPATH_IBCS_LT_STATE_POLLQUIET) {
-			ipath_set_linkstate(dd, IPATH_IB_LINKDOWN_DISABLE);
-			break;
-		}
-		udelay(100);
-	}
-
-	if (!(dd->ipath_flags & IPATH_IB_AUTONEG_INPROG))
-		goto done; /* we got there early or told to stop */
-
-	/* we expect this to timeout */
-	if (wait_event_timeout(dd->ipath_autoneg_wait,
-		!(dd->ipath_flags & IPATH_IB_AUTONEG_INPROG),
-		msecs_to_jiffies(90)))
-		goto done;
-
-	ipath_toggle_rclkrls(dd);
-
-	/* we expect this to timeout */
-	if (wait_event_timeout(dd->ipath_autoneg_wait,
-		!(dd->ipath_flags & IPATH_IB_AUTONEG_INPROG),
-		msecs_to_jiffies(1700)))
-		goto done;
-
-	set_speed_fast(dd, IPATH_IB_SDR);
-	ipath_toggle_rclkrls(dd);
-
-	/*
-	 * wait up to 250 msec for link to train and get to INIT at DDR;
-	 * this should terminate early
-	 */
-	wait_event_timeout(dd->ipath_autoneg_wait,
-		!(dd->ipath_flags & IPATH_IB_AUTONEG_INPROG),
-		msecs_to_jiffies(250));
-done:
-	if (dd->ipath_flags & IPATH_IB_AUTONEG_INPROG) {
-		ipath_dbg("Did not get to DDR INIT (%x) after %Lu msecs\n",
-			ipath_ib_state(dd, dd->ipath_lastibcstat),
-			(unsigned long long) jiffies_to_msecs(jiffies)-startms);
-		dd->ipath_flags &= ~IPATH_IB_AUTONEG_INPROG;
-		if (dd->ipath_autoneg_tries == IPATH_AUTONEG_TRIES) {
-			dd->ipath_flags |= IPATH_IB_AUTONEG_FAILED;
-			ipath_dbg("Giving up on DDR until next IB "
-				"link Down\n");
-			dd->ipath_autoneg_tries = 0;
-		}
-		set_speed_fast(dd, dd->ipath_link_speed_enabled);
-	}
-}
-
-
-/**
- * ipath_init_iba7220_funcs - set up the chip-specific function pointers
- * @dd: the infinipath device
- *
- * This is global, and is called directly at init to set up the
- * chip-specific function pointers for later use.
- */
-void ipath_init_iba7220_funcs(struct ipath_devdata *dd)
-{
-	dd->ipath_f_intrsetup = ipath_7220_intconfig;
-	dd->ipath_f_bus = ipath_setup_7220_config;
-	dd->ipath_f_reset = ipath_setup_7220_reset;
-	dd->ipath_f_get_boardname = ipath_7220_boardname;
-	dd->ipath_f_init_hwerrors = ipath_7220_init_hwerrors;
-	dd->ipath_f_early_init = ipath_7220_early_init;
-	dd->ipath_f_handle_hwerrors = ipath_7220_handle_hwerrors;
-	dd->ipath_f_quiet_serdes = ipath_7220_quiet_serdes;
-	dd->ipath_f_bringup_serdes = ipath_7220_bringup_serdes;
-	dd->ipath_f_clear_tids = ipath_7220_clear_tids;
-	dd->ipath_f_put_tid = ipath_7220_put_tid;
-	dd->ipath_f_cleanup = ipath_setup_7220_cleanup;
-	dd->ipath_f_setextled = ipath_setup_7220_setextled;
-	dd->ipath_f_get_base_info = ipath_7220_get_base_info;
-	dd->ipath_f_free_irq = ipath_7220_free_irq;
-	dd->ipath_f_tidtemplate = ipath_7220_tidtemplate;
-	dd->ipath_f_intr_fallback = ipath_7220_intr_fallback;
-	dd->ipath_f_xgxs_reset = ipath_7220_xgxs_reset;
-	dd->ipath_f_get_ib_cfg = ipath_7220_get_ib_cfg;
-	dd->ipath_f_set_ib_cfg = ipath_7220_set_ib_cfg;
-	dd->ipath_f_config_jint = ipath_7220_config_jint;
-	dd->ipath_f_config_ports = ipath_7220_config_ports;
-	dd->ipath_f_read_counters = ipath_7220_read_counters;
-	dd->ipath_f_get_msgheader = ipath_7220_get_msgheader;
-	dd->ipath_f_ib_updown = ipath_7220_ib_updown;
-
-	/* initialize chip-specific variables */
-	ipath_init_7220_variables(dd);
-}
diff --git a/drivers/infiniband/hw/ipath/ipath_kernel.h b/drivers/infiniband/hw/ipath/ipath_kernel.h
index b3d7efcdf021..6559af60bffd 100644
--- a/drivers/infiniband/hw/ipath/ipath_kernel.h
+++ b/drivers/infiniband/hw/ipath/ipath_kernel.h
@@ -1030,8 +1030,6 @@ void ipath_free_data(struct ipath_portdata *dd);
 u32 __iomem *ipath_getpiobuf(struct ipath_devdata *, u32, u32 *);
 void ipath_chg_pioavailkernel(struct ipath_devdata *dd, unsigned start,
 				unsigned len, int avail);
-void ipath_init_iba7220_funcs(struct ipath_devdata *);
-void ipath_init_iba6120_funcs(struct ipath_devdata *);
 void ipath_init_iba6110_funcs(struct ipath_devdata *);
 void ipath_get_eeprom_info(struct ipath_devdata *);
 int ipath_update_eeprom_log(struct ipath_devdata *dd);
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.c b/drivers/infiniband/hw/ipath/ipath_verbs.c
index 559f39be0dcc..dd7f26d04d46 100644
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c
@@ -2182,7 +2182,7 @@ int ipath_register_ib_device(struct ipath_devdata *dd)
 	snprintf(dev->node_desc, sizeof(dev->node_desc),
 		 IPATH_IDSTR " %s", init_utsname()->nodename);
 
-	ret = ib_register_device(dev);
+	ret = ib_register_device(dev, NULL);
 	if (ret)
 		goto err_reg;
 
diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c
index 39051417054c..4e94e360e43b 100644
--- a/drivers/infiniband/hw/mlx4/main.c
+++ b/drivers/infiniband/hw/mlx4/main.c
@@ -662,7 +662,7 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
 	spin_lock_init(&ibdev->sm_lock);
 	mutex_init(&ibdev->cap_mask_mutex);
 
-	if (ib_register_device(&ibdev->ib_dev))
+	if (ib_register_device(&ibdev->ib_dev, NULL))
 		goto err_map;
 
 	if (mlx4_ib_mad_init(ibdev))
diff --git a/drivers/infiniband/hw/mthca/mthca_provider.c b/drivers/infiniband/hw/mthca/mthca_provider.c
index f080a784bc79..1e0b4b6074ad 100644
--- a/drivers/infiniband/hw/mthca/mthca_provider.c
+++ b/drivers/infiniband/hw/mthca/mthca_provider.c
@@ -1403,7 +1403,7 @@ int mthca_register_device(struct mthca_dev *dev)
 
 	mutex_init(&dev->cap_mask_mutex);
 
-	ret = ib_register_device(&dev->ib_dev);
+	ret = ib_register_device(&dev->ib_dev, NULL);
 	if (ret)
 		return ret;
 
diff --git a/drivers/infiniband/hw/nes/nes_hw.c b/drivers/infiniband/hw/nes/nes_hw.c
index 86acb7d57064..57874a165083 100644
--- a/drivers/infiniband/hw/nes/nes_hw.c
+++ b/drivers/infiniband/hw/nes/nes_hw.c
@@ -2584,7 +2584,6 @@ static void nes_process_mac_intr(struct nes_device *nesdev, u32 mac_number)
 				break;
 			}
 		}
-		spin_unlock_irqrestore(&nesadapter->phy_lock, flags);
 
 		if (phy_data & 0x0004) {
 			if (wide_ppm_offset &&
@@ -2639,6 +2638,8 @@ static void nes_process_mac_intr(struct nes_device *nesdev, u32 mac_number)
 		}
 	}
 
+	spin_unlock_irqrestore(&nesadapter->phy_lock, flags);
+
 	nesadapter->mac_sw_state[mac_number] = NES_MAC_SW_IDLE;
 }
 
@@ -3422,6 +3423,7 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev,
 	struct nes_adapter *nesadapter = nesdev->nesadapter;
 	u32 aeq_info;
 	u32 next_iwarp_state = 0;
+	u32 aeqe_cq_id;
 	u16 async_event_id;
 	u8 tcp_state;
 	u8 iwarp_state;
@@ -3449,6 +3451,14 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev,
 			le32_to_cpu(aeqe->aeqe_words[NES_AEQE_COMP_QP_CQ_ID_IDX]), aeqe,
 			nes_tcp_state_str[tcp_state], nes_iwarp_state_str[iwarp_state]);
 
+	aeqe_cq_id = le32_to_cpu(aeqe->aeqe_words[NES_AEQE_COMP_QP_CQ_ID_IDX]);
+	if (aeq_info & NES_AEQE_QP) {
+		if ((!nes_is_resource_allocated(nesadapter, nesadapter->allocated_qps,
+				aeqe_cq_id)) ||
+				(atomic_read(&nesqp->close_timer_started)))
+			return;
+	}
+
 	switch (async_event_id) {
 		case NES_AEQE_AEID_LLP_FIN_RECEIVED:
 			if (nesqp->term_flags)
diff --git a/drivers/infiniband/hw/nes/nes_nic.c b/drivers/infiniband/hw/nes/nes_nic.c
index e95e8d09ff38..5cc0a9ae5bb1 100644
--- a/drivers/infiniband/hw/nes/nes_nic.c
+++ b/drivers/infiniband/hw/nes/nes_nic.c
@@ -1001,6 +1001,7 @@ static int nes_netdev_change_mtu(struct net_device *netdev, int new_mtu)
 	return ret;
 }
 
+
 static const char nes_ethtool_stringset[][ETH_GSTRING_LEN] = {
 	"Link Change Interrupts",
 	"Linearized SKBs",
@@ -1015,11 +1016,15 @@ static const char nes_ethtool_stringset[][ETH_GSTRING_LEN] = {
 	"Rx Jabber Errors",
 	"Rx Oversized Frames",
 	"Rx Short Frames",
+	"Rx Length Errors",
+	"Rx CRC Errors",
+	"Rx Port Discard",
 	"Endnode Rx Discards",
 	"Endnode Rx Octets",
 	"Endnode Rx Frames",
 	"Endnode Tx Octets",
 	"Endnode Tx Frames",
+	"Tx Errors",
 	"mh detected",
 	"mh pauses",
 	"Retransmission Count",
@@ -1048,19 +1053,13 @@ static const char nes_ethtool_stringset[][ETH_GSTRING_LEN] = {
 	"CM Nodes Destroyed",
 	"CM Accel Drops",
 	"CM Resets Received",
+	"Free 4Kpbls",
+	"Free 256pbls",
 	"Timer Inits",
-	"CQ Depth 1",
-	"CQ Depth 4",
-	"CQ Depth 16",
-	"CQ Depth 24",
-	"CQ Depth 32",
-	"CQ Depth 128",
-	"CQ Depth 256",
 	"LRO aggregated",
 	"LRO flushed",
 	"LRO no_desc",
 };
-
 #define NES_ETHTOOL_STAT_COUNT  ARRAY_SIZE(nes_ethtool_stringset)
 
 /**
@@ -1120,12 +1119,14 @@ static void nes_netdev_get_strings(struct net_device *netdev, u32 stringset,
 /**
  * nes_netdev_get_ethtool_stats
  */
+
 static void nes_netdev_get_ethtool_stats(struct net_device *netdev,
 		struct ethtool_stats *target_ethtool_stats, u64 *target_stat_values)
 {
 	u64 u64temp;
 	struct nes_vnic *nesvnic = netdev_priv(netdev);
 	struct nes_device *nesdev = nesvnic->nesdev;
+	struct nes_adapter *nesadapter = nesdev->nesadapter;
 	u32 nic_count;
 	u32 u32temp;
 	u32 index = 0;
@@ -1154,6 +1155,46 @@ static void nes_netdev_get_ethtool_stats(struct net_device *netdev,
 	nesvnic->nesdev->port_tx_discards += u32temp;
 	nesvnic->netstats.tx_dropped += u32temp;
 
+	u32temp = nes_read_indexed(nesdev,
+			NES_IDX_MAC_RX_SHORT_FRAMES + (nesvnic->nesdev->mac_index*0x200));
+	nesvnic->netstats.rx_dropped += u32temp;
+	nesvnic->nesdev->mac_rx_errors += u32temp;
+	nesvnic->nesdev->mac_rx_short_frames += u32temp;
+
+	u32temp = nes_read_indexed(nesdev,
+			NES_IDX_MAC_RX_OVERSIZED_FRAMES + (nesvnic->nesdev->mac_index*0x200));
+	nesvnic->netstats.rx_dropped += u32temp;
+	nesvnic->nesdev->mac_rx_errors += u32temp;
+	nesvnic->nesdev->mac_rx_oversized_frames += u32temp;
+
+	u32temp = nes_read_indexed(nesdev,
+			NES_IDX_MAC_RX_JABBER_FRAMES + (nesvnic->nesdev->mac_index*0x200));
+	nesvnic->netstats.rx_dropped += u32temp;
+	nesvnic->nesdev->mac_rx_errors += u32temp;
+	nesvnic->nesdev->mac_rx_jabber_frames += u32temp;
+
+	u32temp = nes_read_indexed(nesdev,
+			NES_IDX_MAC_RX_SYMBOL_ERR_FRAMES + (nesvnic->nesdev->mac_index*0x200));
+	nesvnic->netstats.rx_dropped += u32temp;
+	nesvnic->nesdev->mac_rx_errors += u32temp;
+	nesvnic->nesdev->mac_rx_symbol_err_frames += u32temp;
+
+	u32temp = nes_read_indexed(nesdev,
+			NES_IDX_MAC_RX_LENGTH_ERR_FRAMES + (nesvnic->nesdev->mac_index*0x200));
+	nesvnic->netstats.rx_length_errors += u32temp;
+	nesvnic->nesdev->mac_rx_errors += u32temp;
+
+	u32temp = nes_read_indexed(nesdev,
+			NES_IDX_MAC_RX_CRC_ERR_FRAMES + (nesvnic->nesdev->mac_index*0x200));
+	nesvnic->nesdev->mac_rx_errors += u32temp;
+	nesvnic->nesdev->mac_rx_crc_errors += u32temp;
+	nesvnic->netstats.rx_crc_errors += u32temp;
+
+	u32temp = nes_read_indexed(nesdev,
+			NES_IDX_MAC_TX_ERRORS + (nesvnic->nesdev->mac_index*0x200));
+	nesvnic->nesdev->mac_tx_errors += u32temp;
+	nesvnic->netstats.tx_errors += u32temp;
+
 	for (nic_count = 0; nic_count < NES_MAX_PORT_COUNT; nic_count++) {
 		if (nesvnic->qp_nic_index[nic_count] == 0xf)
 			break;
@@ -1218,11 +1259,15 @@ static void nes_netdev_get_ethtool_stats(struct net_device *netdev,
 	target_stat_values[++index] = nesvnic->nesdev->mac_rx_jabber_frames;
 	target_stat_values[++index] = nesvnic->nesdev->mac_rx_oversized_frames;
 	target_stat_values[++index] = nesvnic->nesdev->mac_rx_short_frames;
+	target_stat_values[++index] = nesvnic->netstats.rx_length_errors;
+	target_stat_values[++index] = nesvnic->nesdev->mac_rx_crc_errors;
+	target_stat_values[++index] = nesvnic->nesdev->port_rx_discards;
 	target_stat_values[++index] = nesvnic->endnode_nstat_rx_discard;
 	target_stat_values[++index] = nesvnic->endnode_nstat_rx_octets;
 	target_stat_values[++index] = nesvnic->endnode_nstat_rx_frames;
 	target_stat_values[++index] = nesvnic->endnode_nstat_tx_octets;
 	target_stat_values[++index] = nesvnic->endnode_nstat_tx_frames;
+	target_stat_values[++index] = nesvnic->nesdev->mac_tx_errors;
 	target_stat_values[++index] = mh_detected;
 	target_stat_values[++index] = mh_pauses_sent;
 	target_stat_values[++index] = nesvnic->endnode_ipv4_tcp_retransmits;
@@ -1251,21 +1296,14 @@ static void nes_netdev_get_ethtool_stats(struct net_device *netdev,
 	target_stat_values[++index] = atomic_read(&cm_nodes_destroyed);
 	target_stat_values[++index] = atomic_read(&cm_accel_dropped_pkts);
 	target_stat_values[++index] = atomic_read(&cm_resets_recvd);
+	target_stat_values[++index] = nesadapter->free_4kpbl;
+	target_stat_values[++index] = nesadapter->free_256pbl;
 	target_stat_values[++index] = int_mod_timer_init;
-	target_stat_values[++index] = int_mod_cq_depth_1;
-	target_stat_values[++index] = int_mod_cq_depth_4;
-	target_stat_values[++index] = int_mod_cq_depth_16;
-	target_stat_values[++index] = int_mod_cq_depth_24;
-	target_stat_values[++index] = int_mod_cq_depth_32;
-	target_stat_values[++index] = int_mod_cq_depth_128;
-	target_stat_values[++index] = int_mod_cq_depth_256;
 	target_stat_values[++index] = nesvnic->lro_mgr.stats.aggregated;
 	target_stat_values[++index] = nesvnic->lro_mgr.stats.flushed;
 	target_stat_values[++index] = nesvnic->lro_mgr.stats.no_desc;
-
 }
 
-
 /**
  * nes_netdev_get_drvinfo
  */
diff --git a/drivers/infiniband/hw/nes/nes_verbs.c b/drivers/infiniband/hw/nes/nes_verbs.c
index 925e1f2d1d55..9bc2d744b2ea 100644
--- a/drivers/infiniband/hw/nes/nes_verbs.c
+++ b/drivers/infiniband/hw/nes/nes_verbs.c
@@ -3962,7 +3962,7 @@ int nes_register_ofa_device(struct nes_ib_device *nesibdev)
 	struct nes_adapter *nesadapter = nesdev->nesadapter;
 	int i, ret;
 
-	ret = ib_register_device(&nesvnic->nesibdev->ibdev);
+	ret = ib_register_device(&nesvnic->nesibdev->ibdev, NULL);
 	if (ret) {
 		return ret;
 	}
diff --git a/drivers/infiniband/hw/qib/Kconfig b/drivers/infiniband/hw/qib/Kconfig
new file mode 100644
index 000000000000..7c03a70c55a2
--- /dev/null
+++ b/drivers/infiniband/hw/qib/Kconfig
@@ -0,0 +1,7 @@
+config INFINIBAND_QIB
+	tristate "QLogic PCIe HCA support"
+	depends on 64BIT && NET
+	---help---
+	This is a low-level driver for QLogic PCIe QLE InfiniBand host
+	channel adapters.  This driver does not support the QLogic
+	HyperTransport card (model QHT7140).
diff --git a/drivers/infiniband/hw/qib/Makefile b/drivers/infiniband/hw/qib/Makefile
new file mode 100644
index 000000000000..c6515a1b9a6a
--- /dev/null
+++ b/drivers/infiniband/hw/qib/Makefile
@@ -0,0 +1,15 @@
+obj-$(CONFIG_INFINIBAND_QIB) += ib_qib.o
+
+ib_qib-y := qib_cq.o qib_diag.o qib_dma.o qib_driver.o qib_eeprom.o \
+	qib_file_ops.o qib_fs.o qib_init.o qib_intr.o qib_keys.o \
+	qib_mad.o qib_mmap.o qib_mr.o qib_pcie.o qib_pio_copy.o \
+	qib_qp.o qib_qsfp.o qib_rc.o qib_ruc.o qib_sdma.o qib_srq.o \
+	qib_sysfs.o qib_twsi.o qib_tx.o qib_uc.o qib_ud.o \
+	qib_user_pages.o qib_user_sdma.o qib_verbs_mcast.o qib_iba7220.o \
+	qib_sd7220.o qib_sd7220_img.o qib_iba7322.o qib_verbs.o
+
+# 6120 has no fallback if no MSI interrupts, others can do INTx
+ib_qib-$(CONFIG_PCI_MSI) += qib_iba6120.o
+
+ib_qib-$(CONFIG_X86_64) += qib_wc_x86_64.o
+ib_qib-$(CONFIG_PPC64) += qib_wc_ppc64.o
diff --git a/drivers/infiniband/hw/qib/qib.h b/drivers/infiniband/hw/qib/qib.h
new file mode 100644
index 000000000000..32d9208efcff
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib.h
@@ -0,0 +1,1439 @@
+#ifndef _QIB_KERNEL_H
+#define _QIB_KERNEL_H
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009, 2010 QLogic Corporation.
+ * All rights reserved.
+ * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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.
+ */
+
+/*
+ * This header file is the base header file for qlogic_ib kernel code
+ * qib_user.h serves a similar purpose for user code.
+ */
+
+#include <linux/interrupt.h>
+#include <linux/pci.h>
+#include <linux/dma-mapping.h>
+#include <linux/mutex.h>
+#include <linux/list.h>
+#include <linux/scatterlist.h>
+#include <linux/io.h>
+#include <linux/fs.h>
+#include <linux/completion.h>
+#include <linux/kref.h>
+#include <linux/sched.h>
+
+#include "qib_common.h"
+#include "qib_verbs.h"
+
+/* only s/w major version of QLogic_IB we can handle */
+#define QIB_CHIP_VERS_MAJ 2U
+
+/* don't care about this except printing */
+#define QIB_CHIP_VERS_MIN 0U
+
+/* The Organization Unique Identifier (Mfg code), and its position in GUID */
+#define QIB_OUI 0x001175
+#define QIB_OUI_LSB 40
+
+/*
+ * per driver stats, either not device nor port-specific, or
+ * summed over all of the devices and ports.
+ * They are described by name via ipathfs filesystem, so layout
+ * and number of elements can change without breaking compatibility.
+ * If members are added or deleted qib_statnames[] in qib_fs.c must
+ * change to match.
+ */
+struct qlogic_ib_stats {
+	__u64 sps_ints; /* number of interrupts handled */
+	__u64 sps_errints; /* number of error interrupts */
+	__u64 sps_txerrs; /* tx-related packet errors */
+	__u64 sps_rcverrs; /* non-crc rcv packet errors */
+	__u64 sps_hwerrs; /* hardware errors reported (parity, etc.) */
+	__u64 sps_nopiobufs; /* no pio bufs avail from kernel */
+	__u64 sps_ctxts; /* number of contexts currently open */
+	__u64 sps_lenerrs; /* number of kernel packets where RHF != LRH len */
+	__u64 sps_buffull;
+	__u64 sps_hdrfull;
+};
+
+extern struct qlogic_ib_stats qib_stats;
+extern struct pci_error_handlers qib_pci_err_handler;
+extern struct pci_driver qib_driver;
+
+#define QIB_CHIP_SWVERSION QIB_CHIP_VERS_MAJ
+/*
+ * First-cut critierion for "device is active" is
+ * two thousand dwords combined Tx, Rx traffic per
+ * 5-second interval. SMA packets are 64 dwords,
+ * and occur "a few per second", presumably each way.
+ */
+#define QIB_TRAFFIC_ACTIVE_THRESHOLD (2000)
+
+/*
+ * Struct used to indicate which errors are logged in each of the
+ * error-counters that are logged to EEPROM. A counter is incremented
+ * _once_ (saturating at 255) for each event with any bits set in
+ * the error or hwerror register masks below.
+ */
+#define QIB_EEP_LOG_CNT (4)
+struct qib_eep_log_mask {
+	u64 errs_to_log;
+	u64 hwerrs_to_log;
+};
+
+/*
+ * Below contains all data related to a single context (formerly called port).
+ */
+struct qib_ctxtdata {
+	void **rcvegrbuf;
+	dma_addr_t *rcvegrbuf_phys;
+	/* rcvhdrq base, needs mmap before useful */
+	void *rcvhdrq;
+	/* kernel virtual address where hdrqtail is updated */
+	void *rcvhdrtail_kvaddr;
+	/*
+	 * temp buffer for expected send setup, allocated at open, instead
+	 * of each setup call
+	 */
+	void *tid_pg_list;
+	/*
+	 * Shared page for kernel to signal user processes that send buffers
+	 * need disarming.  The process should call QIB_CMD_DISARM_BUFS
+	 * or QIB_CMD_ACK_EVENT with IPATH_EVENT_DISARM_BUFS set.
+	 */
+	unsigned long *user_event_mask;
+	/* when waiting for rcv or pioavail */
+	wait_queue_head_t wait;
+	/*
+	 * rcvegr bufs base, physical, must fit
+	 * in 44 bits so 32 bit programs mmap64 44 bit works)
+	 */
+	dma_addr_t rcvegr_phys;
+	/* mmap of hdrq, must fit in 44 bits */
+	dma_addr_t rcvhdrq_phys;
+	dma_addr_t rcvhdrqtailaddr_phys;
+
+	/*
+	 * number of opens (including slave sub-contexts) on this instance
+	 * (ignoring forks, dup, etc. for now)
+	 */
+	int cnt;
+	/*
+	 * how much space to leave at start of eager TID entries for
+	 * protocol use, on each TID
+	 */
+	/* instead of calculating it */
+	unsigned ctxt;
+	/* non-zero if ctxt is being shared. */
+	u16 subctxt_cnt;
+	/* non-zero if ctxt is being shared. */
+	u16 subctxt_id;
+	/* number of eager TID entries. */
+	u16 rcvegrcnt;
+	/* index of first eager TID entry. */
+	u16 rcvegr_tid_base;
+	/* number of pio bufs for this ctxt (all procs, if shared) */
+	u32 piocnt;
+	/* first pio buffer for this ctxt */
+	u32 pio_base;
+	/* chip offset of PIO buffers for this ctxt */
+	u32 piobufs;
+	/* how many alloc_pages() chunks in rcvegrbuf_pages */
+	u32 rcvegrbuf_chunks;
+	/* how many egrbufs per chunk */
+	u32 rcvegrbufs_perchunk;
+	/* order for rcvegrbuf_pages */
+	size_t rcvegrbuf_size;
+	/* rcvhdrq size (for freeing) */
+	size_t rcvhdrq_size;
+	/* per-context flags for fileops/intr communication */
+	unsigned long flag;
+	/* next expected TID to check when looking for free */
+	u32 tidcursor;
+	/* WAIT_RCV that timed out, no interrupt */
+	u32 rcvwait_to;
+	/* WAIT_PIO that timed out, no interrupt */
+	u32 piowait_to;
+	/* WAIT_RCV already happened, no wait */
+	u32 rcvnowait;
+	/* WAIT_PIO already happened, no wait */
+	u32 pionowait;
+	/* total number of polled urgent packets */
+	u32 urgent;
+	/* saved total number of polled urgent packets for poll edge trigger */
+	u32 urgent_poll;
+	/* pid of process using this ctxt */
+	pid_t pid;
+	pid_t subpid[QLOGIC_IB_MAX_SUBCTXT];
+	/* same size as task_struct .comm[], command that opened context */
+	char comm[16];
+	/* pkeys set by this use of this ctxt */
+	u16 pkeys[4];
+	/* so file ops can get at unit */
+	struct qib_devdata *dd;
+	/* so funcs that need physical port can get it easily */
+	struct qib_pportdata *ppd;
+	/* A page of memory for rcvhdrhead, rcvegrhead, rcvegrtail * N */
+	void *subctxt_uregbase;
+	/* An array of pages for the eager receive buffers * N */
+	void *subctxt_rcvegrbuf;
+	/* An array of pages for the eager header queue entries * N */
+	void *subctxt_rcvhdr_base;
+	/* The version of the library which opened this ctxt */
+	u32 userversion;
+	/* Bitmask of active slaves */
+	u32 active_slaves;
+	/* Type of packets or conditions we want to poll for */
+	u16 poll_type;
+	/* receive packet sequence counter */
+	u8 seq_cnt;
+	u8 redirect_seq_cnt;
+	/* ctxt rcvhdrq head offset */
+	u32 head;
+	u32 pkt_count;
+	/* QPs waiting for context processing */
+	struct list_head qp_wait_list;
+};
+
+struct qib_sge_state;
+
+struct qib_sdma_txreq {
+	int                 flags;
+	int                 sg_count;
+	dma_addr_t          addr;
+	void              (*callback)(struct qib_sdma_txreq *, int);
+	u16                 start_idx;  /* sdma private */
+	u16                 next_descq_idx;  /* sdma private */
+	struct list_head    list;       /* sdma private */
+};
+
+struct qib_sdma_desc {
+	__le64 qw[2];
+};
+
+struct qib_verbs_txreq {
+	struct qib_sdma_txreq   txreq;
+	struct qib_qp           *qp;
+	struct qib_swqe         *wqe;
+	u32                     dwords;
+	u16                     hdr_dwords;
+	u16                     hdr_inx;
+	struct qib_pio_header	*align_buf;
+	struct qib_mregion	*mr;
+	struct qib_sge_state    *ss;
+};
+
+#define QIB_SDMA_TXREQ_F_USELARGEBUF  0x1
+#define QIB_SDMA_TXREQ_F_HEADTOHOST   0x2
+#define QIB_SDMA_TXREQ_F_INTREQ       0x4
+#define QIB_SDMA_TXREQ_F_FREEBUF      0x8
+#define QIB_SDMA_TXREQ_F_FREEDESC     0x10
+
+#define QIB_SDMA_TXREQ_S_OK        0
+#define QIB_SDMA_TXREQ_S_SENDERROR 1
+#define QIB_SDMA_TXREQ_S_ABORTED   2
+#define QIB_SDMA_TXREQ_S_SHUTDOWN  3
+
+/*
+ * Get/Set IB link-level config parameters for f_get/set_ib_cfg()
+ * Mostly for MADs that set or query link parameters, also ipath
+ * config interfaces
+ */
+#define QIB_IB_CFG_LIDLMC 0 /* LID (LS16b) and Mask (MS16b) */
+#define QIB_IB_CFG_LWID_ENB 2 /* allowed Link-width */
+#define QIB_IB_CFG_LWID 3 /* currently active Link-width */
+#define QIB_IB_CFG_SPD_ENB 4 /* allowed Link speeds */
+#define QIB_IB_CFG_SPD 5 /* current Link spd */
+#define QIB_IB_CFG_RXPOL_ENB 6 /* Auto-RX-polarity enable */
+#define QIB_IB_CFG_LREV_ENB 7 /* Auto-Lane-reversal enable */
+#define QIB_IB_CFG_LINKLATENCY 8 /* Link Latency (IB1.2 only) */
+#define QIB_IB_CFG_HRTBT 9 /* IB heartbeat off/enable/auto; DDR/QDR only */
+#define QIB_IB_CFG_OP_VLS 10 /* operational VLs */
+#define QIB_IB_CFG_VL_HIGH_CAP 11 /* num of VL high priority weights */
+#define QIB_IB_CFG_VL_LOW_CAP 12 /* num of VL low priority weights */
+#define QIB_IB_CFG_OVERRUN_THRESH 13 /* IB overrun threshold */
+#define QIB_IB_CFG_PHYERR_THRESH 14 /* IB PHY error threshold */
+#define QIB_IB_CFG_LINKDEFAULT 15 /* IB link default (sleep/poll) */
+#define QIB_IB_CFG_PKEYS 16 /* update partition keys */
+#define QIB_IB_CFG_MTU 17 /* update MTU in IBC */
+#define QIB_IB_CFG_LSTATE 18 /* update linkcmd and linkinitcmd in IBC */
+#define QIB_IB_CFG_VL_HIGH_LIMIT 19
+#define QIB_IB_CFG_PMA_TICKS 20 /* PMA sample tick resolution */
+#define QIB_IB_CFG_PORT 21 /* switch port we are connected to */
+
+/*
+ * for CFG_LSTATE: LINKCMD in upper 16 bits, LINKINITCMD in lower 16
+ * IB_LINKINITCMD_POLL and SLEEP are also used as set/get values for
+ * QIB_IB_CFG_LINKDEFAULT cmd
+ */
+#define   IB_LINKCMD_DOWN   (0 << 16)
+#define   IB_LINKCMD_ARMED  (1 << 16)
+#define   IB_LINKCMD_ACTIVE (2 << 16)
+#define   IB_LINKINITCMD_NOP     0
+#define   IB_LINKINITCMD_POLL    1
+#define   IB_LINKINITCMD_SLEEP   2
+#define   IB_LINKINITCMD_DISABLE 3
+
+/*
+ * valid states passed to qib_set_linkstate() user call
+ */
+#define QIB_IB_LINKDOWN         0
+#define QIB_IB_LINKARM          1
+#define QIB_IB_LINKACTIVE       2
+#define QIB_IB_LINKDOWN_ONLY    3
+#define QIB_IB_LINKDOWN_SLEEP   4
+#define QIB_IB_LINKDOWN_DISABLE 5
+
+/*
+ * These 7 values (SDR, DDR, and QDR may be ORed for auto-speed
+ * negotiation) are used for the 3rd argument to path_f_set_ib_cfg
+ * with cmd QIB_IB_CFG_SPD_ENB, by direct calls or via sysfs.  They
+ * are also the the possible values for qib_link_speed_enabled and active
+ * The values were chosen to match values used within the IB spec.
+ */
+#define QIB_IB_SDR 1
+#define QIB_IB_DDR 2
+#define QIB_IB_QDR 4
+
+#define QIB_DEFAULT_MTU 4096
+
+/*
+ * Possible IB config parameters for f_get/set_ib_table()
+ */
+#define QIB_IB_TBL_VL_HIGH_ARB 1 /* Get/set VL high priority weights */
+#define QIB_IB_TBL_VL_LOW_ARB 2 /* Get/set VL low priority weights */
+
+/*
+ * Possible "operations" for f_rcvctrl(ppd, op, ctxt)
+ * these are bits so they can be combined, e.g.
+ * QIB_RCVCTRL_INTRAVAIL_ENB | QIB_RCVCTRL_CTXT_ENB
+ */
+#define QIB_RCVCTRL_TAILUPD_ENB 0x01
+#define QIB_RCVCTRL_TAILUPD_DIS 0x02
+#define QIB_RCVCTRL_CTXT_ENB 0x04
+#define QIB_RCVCTRL_CTXT_DIS 0x08
+#define QIB_RCVCTRL_INTRAVAIL_ENB 0x10
+#define QIB_RCVCTRL_INTRAVAIL_DIS 0x20
+#define QIB_RCVCTRL_PKEY_ENB 0x40  /* Note, default is enabled */
+#define QIB_RCVCTRL_PKEY_DIS 0x80
+#define QIB_RCVCTRL_BP_ENB 0x0100
+#define QIB_RCVCTRL_BP_DIS 0x0200
+#define QIB_RCVCTRL_TIDFLOW_ENB 0x0400
+#define QIB_RCVCTRL_TIDFLOW_DIS 0x0800
+
+/*
+ * Possible "operations" for f_sendctrl(ppd, op, var)
+ * these are bits so they can be combined, e.g.
+ * QIB_SENDCTRL_BUFAVAIL_ENB | QIB_SENDCTRL_ENB
+ * Some operations (e.g. DISARM, ABORT) are known to
+ * be "one-shot", so do not modify shadow.
+ */
+#define QIB_SENDCTRL_DISARM       (0x1000)
+#define QIB_SENDCTRL_DISARM_BUF(bufn) ((bufn) | QIB_SENDCTRL_DISARM)
+	/* available (0x2000) */
+#define QIB_SENDCTRL_AVAIL_DIS    (0x4000)
+#define QIB_SENDCTRL_AVAIL_ENB    (0x8000)
+#define QIB_SENDCTRL_AVAIL_BLIP  (0x10000)
+#define QIB_SENDCTRL_SEND_DIS    (0x20000)
+#define QIB_SENDCTRL_SEND_ENB    (0x40000)
+#define QIB_SENDCTRL_FLUSH       (0x80000)
+#define QIB_SENDCTRL_CLEAR      (0x100000)
+#define QIB_SENDCTRL_DISARM_ALL (0x200000)
+
+/*
+ * These are the generic indices for requesting per-port
+ * counter values via the f_portcntr function.  They
+ * are always returned as 64 bit values, although most
+ * are 32 bit counters.
+ */
+/* send-related counters */
+#define QIBPORTCNTR_PKTSEND         0U
+#define QIBPORTCNTR_WORDSEND        1U
+#define QIBPORTCNTR_PSXMITDATA      2U
+#define QIBPORTCNTR_PSXMITPKTS      3U
+#define QIBPORTCNTR_PSXMITWAIT      4U
+#define QIBPORTCNTR_SENDSTALL       5U
+/* receive-related counters */
+#define QIBPORTCNTR_PKTRCV          6U
+#define QIBPORTCNTR_PSRCVDATA       7U
+#define QIBPORTCNTR_PSRCVPKTS       8U
+#define QIBPORTCNTR_RCVEBP          9U
+#define QIBPORTCNTR_RCVOVFL         10U
+#define QIBPORTCNTR_WORDRCV         11U
+/* IB link related error counters */
+#define QIBPORTCNTR_RXLOCALPHYERR   12U
+#define QIBPORTCNTR_RXVLERR         13U
+#define QIBPORTCNTR_ERRICRC         14U
+#define QIBPORTCNTR_ERRVCRC         15U
+#define QIBPORTCNTR_ERRLPCRC        16U
+#define QIBPORTCNTR_BADFORMAT       17U
+#define QIBPORTCNTR_ERR_RLEN        18U
+#define QIBPORTCNTR_IBSYMBOLERR     19U
+#define QIBPORTCNTR_INVALIDRLEN     20U
+#define QIBPORTCNTR_UNSUPVL         21U
+#define QIBPORTCNTR_EXCESSBUFOVFL   22U
+#define QIBPORTCNTR_ERRLINK         23U
+#define QIBPORTCNTR_IBLINKDOWN      24U
+#define QIBPORTCNTR_IBLINKERRRECOV  25U
+#define QIBPORTCNTR_LLI             26U
+/* other error counters */
+#define QIBPORTCNTR_RXDROPPKT       27U
+#define QIBPORTCNTR_VL15PKTDROP     28U
+#define QIBPORTCNTR_ERRPKEY         29U
+#define QIBPORTCNTR_KHDROVFL        30U
+/* sampling counters (these are actually control registers) */
+#define QIBPORTCNTR_PSINTERVAL      31U
+#define QIBPORTCNTR_PSSTART         32U
+#define QIBPORTCNTR_PSSTAT          33U
+
+/* how often we check for packet activity for "power on hours (in seconds) */
+#define ACTIVITY_TIMER 5
+
+/* Below is an opaque struct. Each chip (device) can maintain
+ * private data needed for its operation, but not germane to the
+ * rest of the driver.  For convenience, we define another that
+ * is chip-specific, per-port
+ */
+struct qib_chip_specific;
+struct qib_chipport_specific;
+
+enum qib_sdma_states {
+	qib_sdma_state_s00_hw_down,
+	qib_sdma_state_s10_hw_start_up_wait,
+	qib_sdma_state_s20_idle,
+	qib_sdma_state_s30_sw_clean_up_wait,
+	qib_sdma_state_s40_hw_clean_up_wait,
+	qib_sdma_state_s50_hw_halt_wait,
+	qib_sdma_state_s99_running,
+};
+
+enum qib_sdma_events {
+	qib_sdma_event_e00_go_hw_down,
+	qib_sdma_event_e10_go_hw_start,
+	qib_sdma_event_e20_hw_started,
+	qib_sdma_event_e30_go_running,
+	qib_sdma_event_e40_sw_cleaned,
+	qib_sdma_event_e50_hw_cleaned,
+	qib_sdma_event_e60_hw_halted,
+	qib_sdma_event_e70_go_idle,
+	qib_sdma_event_e7220_err_halted,
+	qib_sdma_event_e7322_err_halted,
+	qib_sdma_event_e90_timer_tick,
+};
+
+extern char *qib_sdma_state_names[];
+extern char *qib_sdma_event_names[];
+
+struct sdma_set_state_action {
+	unsigned op_enable:1;
+	unsigned op_intenable:1;
+	unsigned op_halt:1;
+	unsigned op_drain:1;
+	unsigned go_s99_running_tofalse:1;
+	unsigned go_s99_running_totrue:1;
+};
+
+struct qib_sdma_state {
+	struct kref          kref;
+	struct completion    comp;
+	enum qib_sdma_states current_state;
+	struct sdma_set_state_action *set_state_action;
+	unsigned             current_op;
+	unsigned             go_s99_running;
+	unsigned             first_sendbuf;
+	unsigned             last_sendbuf; /* really last +1 */
+	/* debugging/devel */
+	enum qib_sdma_states previous_state;
+	unsigned             previous_op;
+	enum qib_sdma_events last_event;
+};
+
+struct xmit_wait {
+	struct timer_list timer;
+	u64 counter;
+	u8 flags;
+	struct cache {
+		u64 psxmitdata;
+		u64 psrcvdata;
+		u64 psxmitpkts;
+		u64 psrcvpkts;
+		u64 psxmitwait;
+	} counter_cache;
+};
+
+/*
+ * The structure below encapsulates data relevant to a physical IB Port.
+ * Current chips support only one such port, but the separation
+ * clarifies things a bit. Note that to conform to IB conventions,
+ * port-numbers are one-based. The first or only port is port1.
+ */
+struct qib_pportdata {
+	struct qib_ibport ibport_data;
+
+	struct qib_devdata *dd;
+	struct qib_chippport_specific *cpspec; /* chip-specific per-port */
+	struct kobject pport_kobj;
+	struct kobject sl2vl_kobj;
+	struct kobject diagc_kobj;
+
+	/* GUID for this interface, in network order */
+	__be64 guid;
+
+	/* QIB_POLL, etc. link-state specific flags, per port */
+	u32 lflags;
+	/* qib_lflags driver is waiting for */
+	u32 state_wanted;
+	spinlock_t lflags_lock;
+	/* number of (port-specific) interrupts for this port -- saturates... */
+	u32 int_counter;
+
+	/* ref count for each pkey */
+	atomic_t pkeyrefs[4];
+
+	/*
+	 * this address is mapped readonly into user processes so they can
+	 * get status cheaply, whenever they want.  One qword of status per port
+	 */
+	u64 *statusp;
+
+	/* SendDMA related entries */
+	spinlock_t            sdma_lock;
+	struct qib_sdma_state sdma_state;
+	unsigned long         sdma_buf_jiffies;
+	struct qib_sdma_desc *sdma_descq;
+	u64                   sdma_descq_added;
+	u64                   sdma_descq_removed;
+	u16                   sdma_descq_cnt;
+	u16                   sdma_descq_tail;
+	u16                   sdma_descq_head;
+	u16                   sdma_next_intr;
+	u16                   sdma_reset_wait;
+	u8                    sdma_generation;
+	struct tasklet_struct sdma_sw_clean_up_task;
+	struct list_head      sdma_activelist;
+
+	dma_addr_t       sdma_descq_phys;
+	volatile __le64 *sdma_head_dma; /* DMA'ed by chip */
+	dma_addr_t       sdma_head_phys;
+
+	wait_queue_head_t state_wait; /* for state_wanted */
+
+	/* HoL blocking for SMP replies */
+	unsigned          hol_state;
+	struct timer_list hol_timer;
+
+	/*
+	 * Shadow copies of registers; size indicates read access size.
+	 * Most of them are readonly, but some are write-only register,
+	 * where we manipulate the bits in the shadow copy, and then write
+	 * the shadow copy to qlogic_ib.
+	 *
+	 * We deliberately make most of these 32 bits, since they have
+	 * restricted range.  For any that we read, we won't to generate 32
+	 * bit accesses, since Opteron will generate 2 separate 32 bit HT
+	 * transactions for a 64 bit read, and we want to avoid unnecessary
+	 * bus transactions.
+	 */
+
+	/* This is the 64 bit group */
+	/* last ibcstatus.  opaque outside chip-specific code */
+	u64 lastibcstat;
+
+	/* these are the "32 bit" regs */
+
+	/*
+	 * the following two are 32-bit bitmasks, but {test,clear,set}_bit
+	 * all expect bit fields to be "unsigned long"
+	 */
+	unsigned long p_rcvctrl; /* shadow per-port rcvctrl */
+	unsigned long p_sendctrl; /* shadow per-port sendctrl */
+
+	u32 ibmtu; /* The MTU programmed for this unit */
+	/*
+	 * Current max size IB packet (in bytes) including IB headers, that
+	 * we can send. Changes when ibmtu changes.
+	 */
+	u32 ibmaxlen;
+	/*
+	 * ibmaxlen at init time, limited by chip and by receive buffer
+	 * size.  Not changed after init.
+	 */
+	u32 init_ibmaxlen;
+	/* LID programmed for this instance */
+	u16 lid;
+	/* list of pkeys programmed; 0 if not set */
+	u16 pkeys[4];
+	/* LID mask control */
+	u8 lmc;
+	u8 link_width_supported;
+	u8 link_speed_supported;
+	u8 link_width_enabled;
+	u8 link_speed_enabled;
+	u8 link_width_active;
+	u8 link_speed_active;
+	u8 vls_supported;
+	u8 vls_operational;
+	/* Rx Polarity inversion (compensate for ~tx on partner) */
+	u8 rx_pol_inv;
+
+	u8 hw_pidx;     /* physical port index */
+	u8 port;        /* IB port number and index into dd->pports - 1 */
+
+	u8 delay_mult;
+
+	/* used to override LED behavior */
+	u8 led_override;  /* Substituted for normal value, if non-zero */
+	u16 led_override_timeoff; /* delta to next timer event */
+	u8 led_override_vals[2]; /* Alternates per blink-frame */
+	u8 led_override_phase; /* Just counts, LSB picks from vals[] */
+	atomic_t led_override_timer_active;
+	/* Used to flash LEDs in override mode */
+	struct timer_list led_override_timer;
+	struct xmit_wait cong_stats;
+	struct timer_list symerr_clear_timer;
+};
+
+/* Observers. Not to be taken lightly, possibly not to ship. */
+/*
+ * If a diag read or write is to (bottom <= offset <= top),
+ * the "hoook" is called, allowing, e.g. shadows to be
+ * updated in sync with the driver. struct diag_observer
+ * is the "visible" part.
+ */
+struct diag_observer;
+
+typedef int (*diag_hook) (struct qib_devdata *dd,
+	const struct diag_observer *op,
+	u32 offs, u64 *data, u64 mask, int only_32);
+
+struct diag_observer {
+	diag_hook hook;
+	u32 bottom;
+	u32 top;
+};
+
+extern int qib_register_observer(struct qib_devdata *dd,
+	const struct diag_observer *op);
+
+/* Only declared here, not defined. Private to diags */
+struct diag_observer_list_elt;
+
+/* device data struct now contains only "general per-device" info.
+ * fields related to a physical IB port are in a qib_pportdata struct,
+ * described above) while fields only used by a particualr chip-type are in
+ * a qib_chipdata struct, whose contents are opaque to this file.
+ */
+struct qib_devdata {
+	struct qib_ibdev verbs_dev;     /* must be first */
+	struct list_head list;
+	/* pointers to related structs for this device */
+	/* pci access data structure */
+	struct pci_dev *pcidev;
+	struct cdev *user_cdev;
+	struct cdev *diag_cdev;
+	struct device *user_device;
+	struct device *diag_device;
+
+	/* mem-mapped pointer to base of chip regs */
+	u64 __iomem *kregbase;
+	/* end of mem-mapped chip space excluding sendbuf and user regs */
+	u64 __iomem *kregend;
+	/* physical address of chip for io_remap, etc. */
+	resource_size_t physaddr;
+	/* qib_cfgctxts pointers */
+	struct qib_ctxtdata **rcd; /* Receive Context Data */
+
+	/* qib_pportdata, points to array of (physical) port-specific
+	 * data structs, indexed by pidx (0..n-1)
+	 */
+	struct qib_pportdata *pport;
+	struct qib_chip_specific *cspec; /* chip-specific */
+
+	/* kvirt address of 1st 2k pio buffer */
+	void __iomem *pio2kbase;
+	/* kvirt address of 1st 4k pio buffer */
+	void __iomem *pio4kbase;
+	/* mem-mapped pointer to base of PIO buffers (if using WC PAT) */
+	void __iomem *piobase;
+	/* mem-mapped pointer to base of user chip regs (if using WC PAT) */
+	u64 __iomem *userbase;
+	/*
+	 * points to area where PIOavail registers will be DMA'ed.
+	 * Has to be on a page of it's own, because the page will be
+	 * mapped into user program space.  This copy is *ONLY* ever
+	 * written by DMA, not by the driver!  Need a copy per device
+	 * when we get to multiple devices
+	 */
+	volatile __le64 *pioavailregs_dma; /* DMA'ed by chip */
+	/* physical address where updates occur */
+	dma_addr_t pioavailregs_phys;
+
+	/* device-specific implementations of functions needed by
+	 * common code. Contrary to previous consensus, we can't
+	 * really just point to a device-specific table, because we
+	 * may need to "bend", e.g. *_f_put_tid
+	 */
+	/* fallback to alternate interrupt type if possible */
+	int (*f_intr_fallback)(struct qib_devdata *);
+	/* hard reset chip */
+	int (*f_reset)(struct qib_devdata *);
+	void (*f_quiet_serdes)(struct qib_pportdata *);
+	int (*f_bringup_serdes)(struct qib_pportdata *);
+	int (*f_early_init)(struct qib_devdata *);
+	void (*f_clear_tids)(struct qib_devdata *, struct qib_ctxtdata *);
+	void (*f_put_tid)(struct qib_devdata *, u64 __iomem*,
+				u32, unsigned long);
+	void (*f_cleanup)(struct qib_devdata *);
+	void (*f_setextled)(struct qib_pportdata *, u32);
+	/* fill out chip-specific fields */
+	int (*f_get_base_info)(struct qib_ctxtdata *, struct qib_base_info *);
+	/* free irq */
+	void (*f_free_irq)(struct qib_devdata *);
+	struct qib_message_header *(*f_get_msgheader)
+					(struct qib_devdata *, __le32 *);
+	void (*f_config_ctxts)(struct qib_devdata *);
+	int (*f_get_ib_cfg)(struct qib_pportdata *, int);
+	int (*f_set_ib_cfg)(struct qib_pportdata *, int, u32);
+	int (*f_set_ib_loopback)(struct qib_pportdata *, const char *);
+	int (*f_get_ib_table)(struct qib_pportdata *, int, void *);
+	int (*f_set_ib_table)(struct qib_pportdata *, int, void *);
+	u32 (*f_iblink_state)(u64);
+	u8 (*f_ibphys_portstate)(u64);
+	void (*f_xgxs_reset)(struct qib_pportdata *);
+	/* per chip actions needed for IB Link up/down changes */
+	int (*f_ib_updown)(struct qib_pportdata *, int, u64);
+	u32 __iomem *(*f_getsendbuf)(struct qib_pportdata *, u64, u32 *);
+	/* Read/modify/write of GPIO pins (potentially chip-specific */
+	int (*f_gpio_mod)(struct qib_devdata *dd, u32 out, u32 dir,
+		u32 mask);
+	/* Enable writes to config EEPROM (if supported) */
+	int (*f_eeprom_wen)(struct qib_devdata *dd, int wen);
+	/*
+	 * modify rcvctrl shadow[s] and write to appropriate chip-regs.
+	 * see above QIB_RCVCTRL_xxx_ENB/DIS for operations.
+	 * (ctxt == -1) means "all contexts", only meaningful for
+	 * clearing. Could remove if chip_spec shutdown properly done.
+	 */
+	void (*f_rcvctrl)(struct qib_pportdata *, unsigned int op,
+		int ctxt);
+	/* Read/modify/write sendctrl appropriately for op and port. */
+	void (*f_sendctrl)(struct qib_pportdata *, u32 op);
+	void (*f_set_intr_state)(struct qib_devdata *, u32);
+	void (*f_set_armlaunch)(struct qib_devdata *, u32);
+	void (*f_wantpiobuf_intr)(struct qib_devdata *, u32);
+	int (*f_late_initreg)(struct qib_devdata *);
+	int (*f_init_sdma_regs)(struct qib_pportdata *);
+	u16 (*f_sdma_gethead)(struct qib_pportdata *);
+	int (*f_sdma_busy)(struct qib_pportdata *);
+	void (*f_sdma_update_tail)(struct qib_pportdata *, u16);
+	void (*f_sdma_set_desc_cnt)(struct qib_pportdata *, unsigned);
+	void (*f_sdma_sendctrl)(struct qib_pportdata *, unsigned);
+	void (*f_sdma_hw_clean_up)(struct qib_pportdata *);
+	void (*f_sdma_hw_start_up)(struct qib_pportdata *);
+	void (*f_sdma_init_early)(struct qib_pportdata *);
+	void (*f_set_cntr_sample)(struct qib_pportdata *, u32, u32);
+	void (*f_update_usrhead)(struct qib_ctxtdata *, u64, u32, u32);
+	u32 (*f_hdrqempty)(struct qib_ctxtdata *);
+	u64 (*f_portcntr)(struct qib_pportdata *, u32);
+	u32 (*f_read_cntrs)(struct qib_devdata *, loff_t, char **,
+		u64 **);
+	u32 (*f_read_portcntrs)(struct qib_devdata *, loff_t, u32,
+		char **, u64 **);
+	u32 (*f_setpbc_control)(struct qib_pportdata *, u32, u8, u8);
+	void (*f_initvl15_bufs)(struct qib_devdata *);
+	void (*f_init_ctxt)(struct qib_ctxtdata *);
+	void (*f_txchk_change)(struct qib_devdata *, u32, u32, u32,
+		struct qib_ctxtdata *);
+	void (*f_writescratch)(struct qib_devdata *, u32);
+	int (*f_tempsense_rd)(struct qib_devdata *, int regnum);
+
+	char *boardname; /* human readable board info */
+
+	/* template for writing TIDs  */
+	u64 tidtemplate;
+	/* value to write to free TIDs */
+	u64 tidinvalid;
+
+	/* number of registers used for pioavail */
+	u32 pioavregs;
+	/* device (not port) flags, basically device capabilities */
+	u32 flags;
+	/* last buffer for user use */
+	u32 lastctxt_piobuf;
+
+	/* saturating counter of (non-port-specific) device interrupts */
+	u32 int_counter;
+
+	/* pio bufs allocated per ctxt */
+	u32 pbufsctxt;
+	/* if remainder on bufs/ctxt, ctxts < extrabuf get 1 extra */
+	u32 ctxts_extrabuf;
+	/*
+	 * number of ctxts configured as max; zero is set to number chip
+	 * supports, less gives more pio bufs/ctxt, etc.
+	 */
+	u32 cfgctxts;
+
+	/*
+	 * hint that we should update pioavailshadow before
+	 * looking for a PIO buffer
+	 */
+	u32 upd_pio_shadow;
+
+	/* internal debugging stats */
+	u32 maxpkts_call;
+	u32 avgpkts_call;
+	u64 nopiobufs;
+
+	/* PCI Vendor ID (here for NodeInfo) */
+	u16 vendorid;
+	/* PCI Device ID (here for NodeInfo) */
+	u16 deviceid;
+	/* for write combining settings */
+	unsigned long wc_cookie;
+	unsigned long wc_base;
+	unsigned long wc_len;
+
+	/* shadow copy of struct page *'s for exp tid pages */
+	struct page **pageshadow;
+	/* shadow copy of dma handles for exp tid pages */
+	dma_addr_t *physshadow;
+	u64 __iomem *egrtidbase;
+	spinlock_t sendctrl_lock; /* protect changes to sendctrl shadow */
+	/* around rcd and (user ctxts) ctxt_cnt use (intr vs free) */
+	spinlock_t uctxt_lock; /* rcd and user context changes */
+	/*
+	 * per unit status, see also portdata statusp
+	 * mapped readonly into user processes so they can get unit and
+	 * IB link status cheaply
+	 */
+	u64 *devstatusp;
+	char *freezemsg; /* freeze msg if hw error put chip in freeze */
+	u32 freezelen; /* max length of freezemsg */
+	/* timer used to prevent stats overflow, error throttling, etc. */
+	struct timer_list stats_timer;
+
+	/* timer to verify interrupts work, and fallback if possible */
+	struct timer_list intrchk_timer;
+	unsigned long ureg_align; /* user register alignment */
+
+	/*
+	 * Protects pioavailshadow, pioavailkernel, pio_need_disarm, and
+	 * pio_writing.
+	 */
+	spinlock_t pioavail_lock;
+
+	/*
+	 * Shadow copies of registers; size indicates read access size.
+	 * Most of them are readonly, but some are write-only register,
+	 * where we manipulate the bits in the shadow copy, and then write
+	 * the shadow copy to qlogic_ib.
+	 *
+	 * We deliberately make most of these 32 bits, since they have
+	 * restricted range.  For any that we read, we won't to generate 32
+	 * bit accesses, since Opteron will generate 2 separate 32 bit HT
+	 * transactions for a 64 bit read, and we want to avoid unnecessary
+	 * bus transactions.
+	 */
+
+	/* This is the 64 bit group */
+
+	unsigned long pioavailshadow[6];
+	/* bitmap of send buffers available for the kernel to use with PIO. */
+	unsigned long pioavailkernel[6];
+	/* bitmap of send buffers which need to be disarmed. */
+	unsigned long pio_need_disarm[3];
+	/* bitmap of send buffers which are being written to. */
+	unsigned long pio_writing[3];
+	/* kr_revision shadow */
+	u64 revision;
+	/* Base GUID for device (from eeprom, network order) */
+	__be64 base_guid;
+
+	/*
+	 * kr_sendpiobufbase value (chip offset of pio buffers), and the
+	 * base of the 2KB buffer s(user processes only use 2K)
+	 */
+	u64 piobufbase;
+	u32 pio2k_bufbase;
+
+	/* these are the "32 bit" regs */
+
+	/* number of GUIDs in the flash for this interface */
+	u32 nguid;
+	/*
+	 * the following two are 32-bit bitmasks, but {test,clear,set}_bit
+	 * all expect bit fields to be "unsigned long"
+	 */
+	unsigned long rcvctrl; /* shadow per device rcvctrl */
+	unsigned long sendctrl; /* shadow per device sendctrl */
+
+	/* value we put in kr_rcvhdrcnt */
+	u32 rcvhdrcnt;
+	/* value we put in kr_rcvhdrsize */
+	u32 rcvhdrsize;
+	/* value we put in kr_rcvhdrentsize */
+	u32 rcvhdrentsize;
+	/* kr_ctxtcnt value */
+	u32 ctxtcnt;
+	/* kr_pagealign value */
+	u32 palign;
+	/* number of "2KB" PIO buffers */
+	u32 piobcnt2k;
+	/* size in bytes of "2KB" PIO buffers */
+	u32 piosize2k;
+	/* max usable size in dwords of a "2KB" PIO buffer before going "4KB" */
+	u32 piosize2kmax_dwords;
+	/* number of "4KB" PIO buffers */
+	u32 piobcnt4k;
+	/* size in bytes of "4KB" PIO buffers */
+	u32 piosize4k;
+	/* kr_rcvegrbase value */
+	u32 rcvegrbase;
+	/* kr_rcvtidbase value */
+	u32 rcvtidbase;
+	/* kr_rcvtidcnt value */
+	u32 rcvtidcnt;
+	/* kr_userregbase */
+	u32 uregbase;
+	/* shadow the control register contents */
+	u32 control;
+
+	/* chip address space used by 4k pio buffers */
+	u32 align4k;
+	/* size of each rcvegrbuffer */
+	u32 rcvegrbufsize;
+	/* localbus width (1, 2,4,8,16,32) from config space  */
+	u32 lbus_width;
+	/* localbus speed in MHz */
+	u32 lbus_speed;
+	int unit; /* unit # of this chip */
+
+	/* start of CHIP_SPEC move to chipspec, but need code changes */
+	/* low and high portions of MSI capability/vector */
+	u32 msi_lo;
+	/* saved after PCIe init for restore after reset */
+	u32 msi_hi;
+	/* MSI data (vector) saved for restore */
+	u16 msi_data;
+	/* so we can rewrite it after a chip reset */
+	u32 pcibar0;
+	/* so we can rewrite it after a chip reset */
+	u32 pcibar1;
+	u64 rhdrhead_intr_off;
+
+	/*
+	 * ASCII serial number, from flash, large enough for original
+	 * all digit strings, and longer QLogic serial number format
+	 */
+	u8 serial[16];
+	/* human readable board version */
+	u8 boardversion[96];
+	u8 lbus_info[32]; /* human readable localbus info */
+	/* chip major rev, from qib_revision */
+	u8 majrev;
+	/* chip minor rev, from qib_revision */
+	u8 minrev;
+
+	/* Misc small ints */
+	/* Number of physical ports available */
+	u8 num_pports;
+	/* Lowest context number which can be used by user processes */
+	u8 first_user_ctxt;
+	u8 n_krcv_queues;
+	u8 qpn_mask;
+	u8 skip_kctxt_mask;
+
+	u16 rhf_offset; /* offset of RHF within receive header entry */
+
+	/*
+	 * GPIO pins for twsi-connected devices, and device code for eeprom
+	 */
+	u8 gpio_sda_num;
+	u8 gpio_scl_num;
+	u8 twsi_eeprom_dev;
+	u8 board_atten;
+
+	/* Support (including locks) for EEPROM logging of errors and time */
+	/* control access to actual counters, timer */
+	spinlock_t eep_st_lock;
+	/* control high-level access to EEPROM */
+	struct mutex eep_lock;
+	uint64_t traffic_wds;
+	/* active time is kept in seconds, but logged in hours */
+	atomic_t active_time;
+	/* Below are nominal shadow of EEPROM, new since last EEPROM update */
+	uint8_t eep_st_errs[QIB_EEP_LOG_CNT];
+	uint8_t eep_st_new_errs[QIB_EEP_LOG_CNT];
+	uint16_t eep_hrs;
+	/*
+	 * masks for which bits of errs, hwerrs that cause
+	 * each of the counters to increment.
+	 */
+	struct qib_eep_log_mask eep_st_masks[QIB_EEP_LOG_CNT];
+	struct qib_diag_client *diag_client;
+	spinlock_t qib_diag_trans_lock; /* protect diag observer ops */
+	struct diag_observer_list_elt *diag_observer_list;
+
+	u8 psxmitwait_supported;
+	/* cycle length of PS* counters in HW (in picoseconds) */
+	u16 psxmitwait_check_rate;
+};
+
+/* hol_state values */
+#define QIB_HOL_UP       0
+#define QIB_HOL_INIT     1
+
+#define QIB_SDMA_SENDCTRL_OP_ENABLE    (1U << 0)
+#define QIB_SDMA_SENDCTRL_OP_INTENABLE (1U << 1)
+#define QIB_SDMA_SENDCTRL_OP_HALT      (1U << 2)
+#define QIB_SDMA_SENDCTRL_OP_CLEANUP   (1U << 3)
+#define QIB_SDMA_SENDCTRL_OP_DRAIN     (1U << 4)
+
+/* operation types for f_txchk_change() */
+#define TXCHK_CHG_TYPE_DIS1  3
+#define TXCHK_CHG_TYPE_ENAB1 2
+#define TXCHK_CHG_TYPE_KERN  1
+#define TXCHK_CHG_TYPE_USER  0
+
+#define QIB_CHASE_TIME msecs_to_jiffies(145)
+#define QIB_CHASE_DIS_TIME msecs_to_jiffies(160)
+
+/* Private data for file operations */
+struct qib_filedata {
+	struct qib_ctxtdata *rcd;
+	unsigned subctxt;
+	unsigned tidcursor;
+	struct qib_user_sdma_queue *pq;
+	int rec_cpu_num; /* for cpu affinity; -1 if none */
+};
+
+extern struct list_head qib_dev_list;
+extern spinlock_t qib_devs_lock;
+extern struct qib_devdata *qib_lookup(int unit);
+extern u32 qib_cpulist_count;
+extern unsigned long *qib_cpulist;
+
+extern unsigned qib_wc_pat;
+int qib_init(struct qib_devdata *, int);
+int init_chip_wc_pat(struct qib_devdata *dd, u32);
+int qib_enable_wc(struct qib_devdata *dd);
+void qib_disable_wc(struct qib_devdata *dd);
+int qib_count_units(int *npresentp, int *nupp);
+int qib_count_active_units(void);
+
+int qib_cdev_init(int minor, const char *name,
+		  const struct file_operations *fops,
+		  struct cdev **cdevp, struct device **devp);
+void qib_cdev_cleanup(struct cdev **cdevp, struct device **devp);
+int qib_dev_init(void);
+void qib_dev_cleanup(void);
+
+int qib_diag_add(struct qib_devdata *);
+void qib_diag_remove(struct qib_devdata *);
+void qib_handle_e_ibstatuschanged(struct qib_pportdata *, u64);
+void qib_sdma_update_tail(struct qib_pportdata *, u16); /* hold sdma_lock */
+
+int qib_decode_err(struct qib_devdata *dd, char *buf, size_t blen, u64 err);
+void qib_bad_intrstatus(struct qib_devdata *);
+void qib_handle_urcv(struct qib_devdata *, u64);
+
+/* clean up any per-chip chip-specific stuff */
+void qib_chip_cleanup(struct qib_devdata *);
+/* clean up any chip type-specific stuff */
+void qib_chip_done(void);
+
+/* check to see if we have to force ordering for write combining */
+int qib_unordered_wc(void);
+void qib_pio_copy(void __iomem *to, const void *from, size_t count);
+
+void qib_disarm_piobufs(struct qib_devdata *, unsigned, unsigned);
+int qib_disarm_piobufs_ifneeded(struct qib_ctxtdata *);
+void qib_disarm_piobufs_set(struct qib_devdata *, unsigned long *, unsigned);
+void qib_cancel_sends(struct qib_pportdata *);
+
+int qib_create_rcvhdrq(struct qib_devdata *, struct qib_ctxtdata *);
+int qib_setup_eagerbufs(struct qib_ctxtdata *);
+void qib_set_ctxtcnt(struct qib_devdata *);
+int qib_create_ctxts(struct qib_devdata *dd);
+struct qib_ctxtdata *qib_create_ctxtdata(struct qib_pportdata *, u32);
+void qib_init_pportdata(struct qib_pportdata *, struct qib_devdata *, u8, u8);
+void qib_free_ctxtdata(struct qib_devdata *, struct qib_ctxtdata *);
+
+u32 qib_kreceive(struct qib_ctxtdata *, u32 *, u32 *);
+int qib_reset_device(int);
+int qib_wait_linkstate(struct qib_pportdata *, u32, int);
+int qib_set_linkstate(struct qib_pportdata *, u8);
+int qib_set_mtu(struct qib_pportdata *, u16);
+int qib_set_lid(struct qib_pportdata *, u32, u8);
+void qib_hol_down(struct qib_pportdata *);
+void qib_hol_init(struct qib_pportdata *);
+void qib_hol_up(struct qib_pportdata *);
+void qib_hol_event(unsigned long);
+void qib_disable_after_error(struct qib_devdata *);
+int qib_set_uevent_bits(struct qib_pportdata *, const int);
+
+/* for use in system calls, where we want to know device type, etc. */
+#define ctxt_fp(fp) \
+	(((struct qib_filedata *)(fp)->private_data)->rcd)
+#define subctxt_fp(fp) \
+	(((struct qib_filedata *)(fp)->private_data)->subctxt)
+#define tidcursor_fp(fp) \
+	(((struct qib_filedata *)(fp)->private_data)->tidcursor)
+#define user_sdma_queue_fp(fp) \
+	(((struct qib_filedata *)(fp)->private_data)->pq)
+
+static inline struct qib_devdata *dd_from_ppd(struct qib_pportdata *ppd)
+{
+	return ppd->dd;
+}
+
+static inline struct qib_devdata *dd_from_dev(struct qib_ibdev *dev)
+{
+	return container_of(dev, struct qib_devdata, verbs_dev);
+}
+
+static inline struct qib_devdata *dd_from_ibdev(struct ib_device *ibdev)
+{
+	return dd_from_dev(to_idev(ibdev));
+}
+
+static inline struct qib_pportdata *ppd_from_ibp(struct qib_ibport *ibp)
+{
+	return container_of(ibp, struct qib_pportdata, ibport_data);
+}
+
+static inline struct qib_ibport *to_iport(struct ib_device *ibdev, u8 port)
+{
+	struct qib_devdata *dd = dd_from_ibdev(ibdev);
+	unsigned pidx = port - 1; /* IB number port from 1, hdw from 0 */
+
+	WARN_ON(pidx >= dd->num_pports);
+	return &dd->pport[pidx].ibport_data;
+}
+
+/*
+ * values for dd->flags (_device_ related flags) and
+ */
+#define QIB_HAS_LINK_LATENCY  0x1 /* supports link latency (IB 1.2) */
+#define QIB_INITTED           0x2 /* chip and driver up and initted */
+#define QIB_DOING_RESET       0x4  /* in the middle of doing chip reset */
+#define QIB_PRESENT           0x8  /* chip accesses can be done */
+#define QIB_PIO_FLUSH_WC      0x10 /* Needs Write combining flush for PIO */
+#define QIB_HAS_THRESH_UPDATE 0x40
+#define QIB_HAS_SDMA_TIMEOUT  0x80
+#define QIB_USE_SPCL_TRIG     0x100 /* SpecialTrigger launch enabled */
+#define QIB_NODMA_RTAIL       0x200 /* rcvhdrtail register DMA enabled */
+#define QIB_HAS_INTX          0x800 /* Supports INTx interrupts */
+#define QIB_HAS_SEND_DMA      0x1000 /* Supports Send DMA */
+#define QIB_HAS_VLSUPP        0x2000 /* Supports multiple VLs; PBC different */
+#define QIB_HAS_HDRSUPP       0x4000 /* Supports header suppression */
+#define QIB_BADINTR           0x8000 /* severe interrupt problems */
+#define QIB_DCA_ENABLED       0x10000 /* Direct Cache Access enabled */
+#define QIB_HAS_QSFP          0x20000 /* device (card instance) has QSFP */
+
+/*
+ * values for ppd->lflags (_ib_port_ related flags)
+ */
+#define QIBL_LINKV             0x1 /* IB link state valid */
+#define QIBL_LINKDOWN          0x8 /* IB link is down */
+#define QIBL_LINKINIT          0x10 /* IB link level is up */
+#define QIBL_LINKARMED         0x20 /* IB link is ARMED */
+#define QIBL_LINKACTIVE        0x40 /* IB link is ACTIVE */
+/* leave a gap for more IB-link state */
+#define QIBL_IB_AUTONEG_INPROG 0x1000 /* non-IBTA DDR/QDR neg active */
+#define QIBL_IB_AUTONEG_FAILED 0x2000 /* non-IBTA DDR/QDR neg failed */
+#define QIBL_IB_LINK_DISABLED  0x4000 /* Linkdown-disable forced,
+				       * Do not try to bring up */
+#define QIBL_IB_FORCE_NOTIFY   0x8000 /* force notify on next ib change */
+
+/* IB dword length mask in PBC (lower 11 bits); same for all chips */
+#define QIB_PBC_LENGTH_MASK                     ((1 << 11) - 1)
+
+
+/* ctxt_flag bit offsets */
+		/* waiting for a packet to arrive */
+#define QIB_CTXT_WAITING_RCV   2
+		/* master has not finished initializing */
+#define QIB_CTXT_MASTER_UNINIT 4
+		/* waiting for an urgent packet to arrive */
+#define QIB_CTXT_WAITING_URG 5
+
+/* free up any allocated data at closes */
+void qib_free_data(struct qib_ctxtdata *dd);
+void qib_chg_pioavailkernel(struct qib_devdata *, unsigned, unsigned,
+			    u32, struct qib_ctxtdata *);
+struct qib_devdata *qib_init_iba7322_funcs(struct pci_dev *,
+					   const struct pci_device_id *);
+struct qib_devdata *qib_init_iba7220_funcs(struct pci_dev *,
+					   const struct pci_device_id *);
+struct qib_devdata *qib_init_iba6120_funcs(struct pci_dev *,
+					   const struct pci_device_id *);
+void qib_free_devdata(struct qib_devdata *);
+struct qib_devdata *qib_alloc_devdata(struct pci_dev *pdev, size_t extra);
+
+#define QIB_TWSI_NO_DEV 0xFF
+/* Below qib_twsi_ functions must be called with eep_lock held */
+int qib_twsi_reset(struct qib_devdata *dd);
+int qib_twsi_blk_rd(struct qib_devdata *dd, int dev, int addr, void *buffer,
+		    int len);
+int qib_twsi_blk_wr(struct qib_devdata *dd, int dev, int addr,
+		    const void *buffer, int len);
+void qib_get_eeprom_info(struct qib_devdata *);
+int qib_update_eeprom_log(struct qib_devdata *dd);
+void qib_inc_eeprom_err(struct qib_devdata *dd, u32 eidx, u32 incr);
+void qib_dump_lookup_output_queue(struct qib_devdata *);
+void qib_force_pio_avail_update(struct qib_devdata *);
+void qib_clear_symerror_on_linkup(unsigned long opaque);
+
+/*
+ * Set LED override, only the two LSBs have "public" meaning, but
+ * any non-zero value substitutes them for the Link and LinkTrain
+ * LED states.
+ */
+#define QIB_LED_PHYS 1 /* Physical (linktraining) GREEN LED */
+#define QIB_LED_LOG 2  /* Logical (link) YELLOW LED */
+void qib_set_led_override(struct qib_pportdata *ppd, unsigned int val);
+
+/* send dma routines */
+int qib_setup_sdma(struct qib_pportdata *);
+void qib_teardown_sdma(struct qib_pportdata *);
+void __qib_sdma_intr(struct qib_pportdata *);
+void qib_sdma_intr(struct qib_pportdata *);
+int qib_sdma_verbs_send(struct qib_pportdata *, struct qib_sge_state *,
+			u32, struct qib_verbs_txreq *);
+/* ppd->sdma_lock should be locked before calling this. */
+int qib_sdma_make_progress(struct qib_pportdata *dd);
+
+/* must be called under qib_sdma_lock */
+static inline u16 qib_sdma_descq_freecnt(const struct qib_pportdata *ppd)
+{
+	return ppd->sdma_descq_cnt -
+		(ppd->sdma_descq_added - ppd->sdma_descq_removed) - 1;
+}
+
+static inline int __qib_sdma_running(struct qib_pportdata *ppd)
+{
+	return ppd->sdma_state.current_state == qib_sdma_state_s99_running;
+}
+int qib_sdma_running(struct qib_pportdata *);
+
+void __qib_sdma_process_event(struct qib_pportdata *, enum qib_sdma_events);
+void qib_sdma_process_event(struct qib_pportdata *, enum qib_sdma_events);
+
+/*
+ * number of words used for protocol header if not set by qib_userinit();
+ */
+#define QIB_DFLT_RCVHDRSIZE 9
+
+/*
+ * We need to be able to handle an IB header of at least 24 dwords.
+ * We need the rcvhdrq large enough to handle largest IB header, but
+ * still have room for a 2KB MTU standard IB packet.
+ * Additionally, some processor/memory controller combinations
+ * benefit quite strongly from having the DMA'ed data be cacheline
+ * aligned and a cacheline multiple, so we set the size to 32 dwords
+ * (2 64-byte primary cachelines for pretty much all processors of
+ * interest).  The alignment hurts nothing, other than using somewhat
+ * more memory.
+ */
+#define QIB_RCVHDR_ENTSIZE 32
+
+int qib_get_user_pages(unsigned long, size_t, struct page **);
+void qib_release_user_pages(struct page **, size_t);
+int qib_eeprom_read(struct qib_devdata *, u8, void *, int);
+int qib_eeprom_write(struct qib_devdata *, u8, const void *, int);
+u32 __iomem *qib_getsendbuf_range(struct qib_devdata *, u32 *, u32, u32);
+void qib_sendbuf_done(struct qib_devdata *, unsigned);
+
+static inline void qib_clear_rcvhdrtail(const struct qib_ctxtdata *rcd)
+{
+	*((u64 *) rcd->rcvhdrtail_kvaddr) = 0ULL;
+}
+
+static inline u32 qib_get_rcvhdrtail(const struct qib_ctxtdata *rcd)
+{
+	/*
+	 * volatile because it's a DMA target from the chip, routine is
+	 * inlined, and don't want register caching or reordering.
+	 */
+	return (u32) le64_to_cpu(
+		*((volatile __le64 *)rcd->rcvhdrtail_kvaddr)); /* DMA'ed */
+}
+
+static inline u32 qib_get_hdrqtail(const struct qib_ctxtdata *rcd)
+{
+	const struct qib_devdata *dd = rcd->dd;
+	u32 hdrqtail;
+
+	if (dd->flags & QIB_NODMA_RTAIL) {
+		__le32 *rhf_addr;
+		u32 seq;
+
+		rhf_addr = (__le32 *) rcd->rcvhdrq +
+			rcd->head + dd->rhf_offset;
+		seq = qib_hdrget_seq(rhf_addr);
+		hdrqtail = rcd->head;
+		if (seq == rcd->seq_cnt)
+			hdrqtail++;
+	} else
+		hdrqtail = qib_get_rcvhdrtail(rcd);
+
+	return hdrqtail;
+}
+
+/*
+ * sysfs interface.
+ */
+
+extern const char ib_qib_version[];
+
+int qib_device_create(struct qib_devdata *);
+void qib_device_remove(struct qib_devdata *);
+
+int qib_create_port_files(struct ib_device *ibdev, u8 port_num,
+			  struct kobject *kobj);
+int qib_verbs_register_sysfs(struct qib_devdata *);
+void qib_verbs_unregister_sysfs(struct qib_devdata *);
+/* Hook for sysfs read of QSFP */
+extern int qib_qsfp_dump(struct qib_pportdata *ppd, char *buf, int len);
+
+int __init qib_init_qibfs(void);
+int __exit qib_exit_qibfs(void);
+
+int qibfs_add(struct qib_devdata *);
+int qibfs_remove(struct qib_devdata *);
+
+int qib_pcie_init(struct pci_dev *, const struct pci_device_id *);
+int qib_pcie_ddinit(struct qib_devdata *, struct pci_dev *,
+		    const struct pci_device_id *);
+void qib_pcie_ddcleanup(struct qib_devdata *);
+int qib_pcie_params(struct qib_devdata *, u32, u32 *, struct msix_entry *);
+int qib_reinit_intr(struct qib_devdata *);
+void qib_enable_intx(struct pci_dev *);
+void qib_nomsi(struct qib_devdata *);
+void qib_nomsix(struct qib_devdata *);
+void qib_pcie_getcmd(struct qib_devdata *, u16 *, u8 *, u8 *);
+void qib_pcie_reenable(struct qib_devdata *, u16, u8, u8);
+
+/*
+ * dma_addr wrappers - all 0's invalid for hw
+ */
+dma_addr_t qib_map_page(struct pci_dev *, struct page *, unsigned long,
+			  size_t, int);
+const char *qib_get_unit_name(int unit);
+
+/*
+ * Flush write combining store buffers (if present) and perform a write
+ * barrier.
+ */
+#if defined(CONFIG_X86_64)
+#define qib_flush_wc() asm volatile("sfence" : : : "memory")
+#else
+#define qib_flush_wc() wmb() /* no reorder around wc flush */
+#endif
+
+/* global module parameter variables */
+extern unsigned qib_ibmtu;
+extern ushort qib_cfgctxts;
+extern ushort qib_num_cfg_vls;
+extern ushort qib_mini_init; /* If set, do few (ideally 0) writes to chip */
+extern unsigned qib_n_krcv_queues;
+extern unsigned qib_sdma_fetch_arb;
+extern unsigned qib_compat_ddr_negotiate;
+extern int qib_special_trigger;
+
+extern struct mutex qib_mutex;
+
+/* Number of seconds before our card status check...  */
+#define STATUS_TIMEOUT 60
+
+#define QIB_DRV_NAME            "ib_qib"
+#define QIB_USER_MINOR_BASE     0
+#define QIB_TRACE_MINOR         127
+#define QIB_DIAGPKT_MINOR       128
+#define QIB_DIAG_MINOR_BASE     129
+#define QIB_NMINORS             255
+
+#define PCI_VENDOR_ID_PATHSCALE 0x1fc1
+#define PCI_VENDOR_ID_QLOGIC 0x1077
+#define PCI_DEVICE_ID_QLOGIC_IB_6120 0x10
+#define PCI_DEVICE_ID_QLOGIC_IB_7220 0x7220
+#define PCI_DEVICE_ID_QLOGIC_IB_7322 0x7322
+
+/*
+ * qib_early_err is used (only!) to print early errors before devdata is
+ * allocated, or when dd->pcidev may not be valid, and at the tail end of
+ * cleanup when devdata may have been freed, etc.  qib_dev_porterr is
+ * the same as qib_dev_err, but is used when the message really needs
+ * the IB port# to be definitive as to what's happening..
+ * All of these go to the trace log, and the trace log entry is done
+ * first to avoid possible serial port delays from printk.
+ */
+#define qib_early_err(dev, fmt, ...) \
+	do { \
+		dev_info(dev, KERN_ERR QIB_DRV_NAME ": " fmt, ##__VA_ARGS__); \
+	} while (0)
+
+#define qib_dev_err(dd, fmt, ...) \
+	do { \
+		dev_err(&(dd)->pcidev->dev, "%s: " fmt, \
+			qib_get_unit_name((dd)->unit), ##__VA_ARGS__); \
+	} while (0)
+
+#define qib_dev_porterr(dd, port, fmt, ...) \
+	do { \
+		dev_err(&(dd)->pcidev->dev, "%s: IB%u:%u " fmt, \
+			qib_get_unit_name((dd)->unit), (dd)->unit, (port), \
+			##__VA_ARGS__); \
+	} while (0)
+
+#define qib_devinfo(pcidev, fmt, ...) \
+	do { \
+		dev_info(&(pcidev)->dev, fmt, ##__VA_ARGS__); \
+	} while (0)
+
+/*
+ * this is used for formatting hw error messages...
+ */
+struct qib_hwerror_msgs {
+	u64 mask;
+	const char *msg;
+};
+
+#define QLOGIC_IB_HWE_MSG(a, b) { .mask = a, .msg = b }
+
+/* in qib_intr.c... */
+void qib_format_hwerrors(u64 hwerrs,
+			 const struct qib_hwerror_msgs *hwerrmsgs,
+			 size_t nhwerrmsgs, char *msg, size_t lmsg);
+#endif                          /* _QIB_KERNEL_H */
diff --git a/drivers/infiniband/hw/qib/qib_6120_regs.h b/drivers/infiniband/hw/qib/qib_6120_regs.h
new file mode 100644
index 000000000000..e16cb6f7de2c
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_6120_regs.h
@@ -0,0 +1,977 @@
+/*
+ * Copyright (c) 2008, 2009, 2010 QLogic Corporation. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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.
+ */
+
+/* This file is mechanically generated from RTL. Any hand-edits will be lost! */
+
+#define QIB_6120_Revision_OFFS 0x0
+#define QIB_6120_Revision_R_Simulator_LSB 0x3F
+#define QIB_6120_Revision_R_Simulator_RMASK 0x1
+#define QIB_6120_Revision_Reserved_LSB 0x28
+#define QIB_6120_Revision_Reserved_RMASK 0x7FFFFF
+#define QIB_6120_Revision_BoardID_LSB 0x20
+#define QIB_6120_Revision_BoardID_RMASK 0xFF
+#define QIB_6120_Revision_R_SW_LSB 0x18
+#define QIB_6120_Revision_R_SW_RMASK 0xFF
+#define QIB_6120_Revision_R_Arch_LSB 0x10
+#define QIB_6120_Revision_R_Arch_RMASK 0xFF
+#define QIB_6120_Revision_R_ChipRevMajor_LSB 0x8
+#define QIB_6120_Revision_R_ChipRevMajor_RMASK 0xFF
+#define QIB_6120_Revision_R_ChipRevMinor_LSB 0x0
+#define QIB_6120_Revision_R_ChipRevMinor_RMASK 0xFF
+
+#define QIB_6120_Control_OFFS 0x8
+#define QIB_6120_Control_TxLatency_LSB 0x4
+#define QIB_6120_Control_TxLatency_RMASK 0x1
+#define QIB_6120_Control_PCIERetryBufDiagEn_LSB 0x3
+#define QIB_6120_Control_PCIERetryBufDiagEn_RMASK 0x1
+#define QIB_6120_Control_LinkEn_LSB 0x2
+#define QIB_6120_Control_LinkEn_RMASK 0x1
+#define QIB_6120_Control_FreezeMode_LSB 0x1
+#define QIB_6120_Control_FreezeMode_RMASK 0x1
+#define QIB_6120_Control_SyncReset_LSB 0x0
+#define QIB_6120_Control_SyncReset_RMASK 0x1
+
+#define QIB_6120_PageAlign_OFFS 0x10
+
+#define QIB_6120_PortCnt_OFFS 0x18
+
+#define QIB_6120_SendRegBase_OFFS 0x30
+
+#define QIB_6120_UserRegBase_OFFS 0x38
+
+#define QIB_6120_CntrRegBase_OFFS 0x40
+
+#define QIB_6120_Scratch_OFFS 0x48
+#define QIB_6120_Scratch_TopHalf_LSB 0x20
+#define QIB_6120_Scratch_TopHalf_RMASK 0xFFFFFFFF
+#define QIB_6120_Scratch_BottomHalf_LSB 0x0
+#define QIB_6120_Scratch_BottomHalf_RMASK 0xFFFFFFFF
+
+#define QIB_6120_IntBlocked_OFFS 0x60
+#define QIB_6120_IntBlocked_ErrorIntBlocked_LSB 0x1F
+#define QIB_6120_IntBlocked_ErrorIntBlocked_RMASK 0x1
+#define QIB_6120_IntBlocked_PioSetIntBlocked_LSB 0x1E
+#define QIB_6120_IntBlocked_PioSetIntBlocked_RMASK 0x1
+#define QIB_6120_IntBlocked_PioBufAvailIntBlocked_LSB 0x1D
+#define QIB_6120_IntBlocked_PioBufAvailIntBlocked_RMASK 0x1
+#define QIB_6120_IntBlocked_assertGPIOIntBlocked_LSB 0x1C
+#define QIB_6120_IntBlocked_assertGPIOIntBlocked_RMASK 0x1
+#define QIB_6120_IntBlocked_Reserved_LSB 0xF
+#define QIB_6120_IntBlocked_Reserved_RMASK 0x1FFF
+#define QIB_6120_IntBlocked_RcvAvail4IntBlocked_LSB 0x10
+#define QIB_6120_IntBlocked_RcvAvail4IntBlocked_RMASK 0x1
+#define QIB_6120_IntBlocked_RcvAvail3IntBlocked_LSB 0xF
+#define QIB_6120_IntBlocked_RcvAvail3IntBlocked_RMASK 0x1
+#define QIB_6120_IntBlocked_RcvAvail2IntBlocked_LSB 0xE
+#define QIB_6120_IntBlocked_RcvAvail2IntBlocked_RMASK 0x1
+#define QIB_6120_IntBlocked_RcvAvail1IntBlocked_LSB 0xD
+#define QIB_6120_IntBlocked_RcvAvail1IntBlocked_RMASK 0x1
+#define QIB_6120_IntBlocked_RcvAvail0IntBlocked_LSB 0xC
+#define QIB_6120_IntBlocked_RcvAvail0IntBlocked_RMASK 0x1
+#define QIB_6120_IntBlocked_Reserved1_LSB 0x5
+#define QIB_6120_IntBlocked_Reserved1_RMASK 0x7F
+#define QIB_6120_IntBlocked_RcvUrg4IntBlocked_LSB 0x4
+#define QIB_6120_IntBlocked_RcvUrg4IntBlocked_RMASK 0x1
+#define QIB_6120_IntBlocked_RcvUrg3IntBlocked_LSB 0x3
+#define QIB_6120_IntBlocked_RcvUrg3IntBlocked_RMASK 0x1
+#define QIB_6120_IntBlocked_RcvUrg2IntBlocked_LSB 0x2
+#define QIB_6120_IntBlocked_RcvUrg2IntBlocked_RMASK 0x1
+#define QIB_6120_IntBlocked_RcvUrg1IntBlocked_LSB 0x1
+#define QIB_6120_IntBlocked_RcvUrg1IntBlocked_RMASK 0x1
+#define QIB_6120_IntBlocked_RcvUrg0IntBlocked_LSB 0x0
+#define QIB_6120_IntBlocked_RcvUrg0IntBlocked_RMASK 0x1
+
+#define QIB_6120_IntMask_OFFS 0x68
+#define QIB_6120_IntMask_ErrorIntMask_LSB 0x1F
+#define QIB_6120_IntMask_ErrorIntMask_RMASK 0x1
+#define QIB_6120_IntMask_PioSetIntMask_LSB 0x1E
+#define QIB_6120_IntMask_PioSetIntMask_RMASK 0x1
+#define QIB_6120_IntMask_PioBufAvailIntMask_LSB 0x1D
+#define QIB_6120_IntMask_PioBufAvailIntMask_RMASK 0x1
+#define QIB_6120_IntMask_assertGPIOIntMask_LSB 0x1C
+#define QIB_6120_IntMask_assertGPIOIntMask_RMASK 0x1
+#define QIB_6120_IntMask_Reserved_LSB 0x11
+#define QIB_6120_IntMask_Reserved_RMASK 0x7FF
+#define QIB_6120_IntMask_RcvAvail4IntMask_LSB 0x10
+#define QIB_6120_IntMask_RcvAvail4IntMask_RMASK 0x1
+#define QIB_6120_IntMask_RcvAvail3IntMask_LSB 0xF
+#define QIB_6120_IntMask_RcvAvail3IntMask_RMASK 0x1
+#define QIB_6120_IntMask_RcvAvail2IntMask_LSB 0xE
+#define QIB_6120_IntMask_RcvAvail2IntMask_RMASK 0x1
+#define QIB_6120_IntMask_RcvAvail1IntMask_LSB 0xD
+#define QIB_6120_IntMask_RcvAvail1IntMask_RMASK 0x1
+#define QIB_6120_IntMask_RcvAvail0IntMask_LSB 0xC
+#define QIB_6120_IntMask_RcvAvail0IntMask_RMASK 0x1
+#define QIB_6120_IntMask_Reserved1_LSB 0x5
+#define QIB_6120_IntMask_Reserved1_RMASK 0x7F
+#define QIB_6120_IntMask_RcvUrg4IntMask_LSB 0x4
+#define QIB_6120_IntMask_RcvUrg4IntMask_RMASK 0x1
+#define QIB_6120_IntMask_RcvUrg3IntMask_LSB 0x3
+#define QIB_6120_IntMask_RcvUrg3IntMask_RMASK 0x1
+#define QIB_6120_IntMask_RcvUrg2IntMask_LSB 0x2
+#define QIB_6120_IntMask_RcvUrg2IntMask_RMASK 0x1
+#define QIB_6120_IntMask_RcvUrg1IntMask_LSB 0x1
+#define QIB_6120_IntMask_RcvUrg1IntMask_RMASK 0x1
+#define QIB_6120_IntMask_RcvUrg0IntMask_LSB 0x0
+#define QIB_6120_IntMask_RcvUrg0IntMask_RMASK 0x1
+
+#define QIB_6120_IntStatus_OFFS 0x70
+#define QIB_6120_IntStatus_Error_LSB 0x1F
+#define QIB_6120_IntStatus_Error_RMASK 0x1
+#define QIB_6120_IntStatus_PioSent_LSB 0x1E
+#define QIB_6120_IntStatus_PioSent_RMASK 0x1
+#define QIB_6120_IntStatus_PioBufAvail_LSB 0x1D
+#define QIB_6120_IntStatus_PioBufAvail_RMASK 0x1
+#define QIB_6120_IntStatus_assertGPIO_LSB 0x1C
+#define QIB_6120_IntStatus_assertGPIO_RMASK 0x1
+#define QIB_6120_IntStatus_Reserved_LSB 0xF
+#define QIB_6120_IntStatus_Reserved_RMASK 0x1FFF
+#define QIB_6120_IntStatus_RcvAvail4_LSB 0x10
+#define QIB_6120_IntStatus_RcvAvail4_RMASK 0x1
+#define QIB_6120_IntStatus_RcvAvail3_LSB 0xF
+#define QIB_6120_IntStatus_RcvAvail3_RMASK 0x1
+#define QIB_6120_IntStatus_RcvAvail2_LSB 0xE
+#define QIB_6120_IntStatus_RcvAvail2_RMASK 0x1
+#define QIB_6120_IntStatus_RcvAvail1_LSB 0xD
+#define QIB_6120_IntStatus_RcvAvail1_RMASK 0x1
+#define QIB_6120_IntStatus_RcvAvail0_LSB 0xC
+#define QIB_6120_IntStatus_RcvAvail0_RMASK 0x1
+#define QIB_6120_IntStatus_Reserved1_LSB 0x5
+#define QIB_6120_IntStatus_Reserved1_RMASK 0x7F
+#define QIB_6120_IntStatus_RcvUrg4_LSB 0x4
+#define QIB_6120_IntStatus_RcvUrg4_RMASK 0x1
+#define QIB_6120_IntStatus_RcvUrg3_LSB 0x3
+#define QIB_6120_IntStatus_RcvUrg3_RMASK 0x1
+#define QIB_6120_IntStatus_RcvUrg2_LSB 0x2
+#define QIB_6120_IntStatus_RcvUrg2_RMASK 0x1
+#define QIB_6120_IntStatus_RcvUrg1_LSB 0x1
+#define QIB_6120_IntStatus_RcvUrg1_RMASK 0x1
+#define QIB_6120_IntStatus_RcvUrg0_LSB 0x0
+#define QIB_6120_IntStatus_RcvUrg0_RMASK 0x1
+
+#define QIB_6120_IntClear_OFFS 0x78
+#define QIB_6120_IntClear_ErrorIntClear_LSB 0x1F
+#define QIB_6120_IntClear_ErrorIntClear_RMASK 0x1
+#define QIB_6120_IntClear_PioSetIntClear_LSB 0x1E
+#define QIB_6120_IntClear_PioSetIntClear_RMASK 0x1
+#define QIB_6120_IntClear_PioBufAvailIntClear_LSB 0x1D
+#define QIB_6120_IntClear_PioBufAvailIntClear_RMASK 0x1
+#define QIB_6120_IntClear_assertGPIOIntClear_LSB 0x1C
+#define QIB_6120_IntClear_assertGPIOIntClear_RMASK 0x1
+#define QIB_6120_IntClear_Reserved_LSB 0xF
+#define QIB_6120_IntClear_Reserved_RMASK 0x1FFF
+#define QIB_6120_IntClear_RcvAvail4IntClear_LSB 0x10
+#define QIB_6120_IntClear_RcvAvail4IntClear_RMASK 0x1
+#define QIB_6120_IntClear_RcvAvail3IntClear_LSB 0xF
+#define QIB_6120_IntClear_RcvAvail3IntClear_RMASK 0x1
+#define QIB_6120_IntClear_RcvAvail2IntClear_LSB 0xE
+#define QIB_6120_IntClear_RcvAvail2IntClear_RMASK 0x1
+#define QIB_6120_IntClear_RcvAvail1IntClear_LSB 0xD
+#define QIB_6120_IntClear_RcvAvail1IntClear_RMASK 0x1
+#define QIB_6120_IntClear_RcvAvail0IntClear_LSB 0xC
+#define QIB_6120_IntClear_RcvAvail0IntClear_RMASK 0x1
+#define QIB_6120_IntClear_Reserved1_LSB 0x5
+#define QIB_6120_IntClear_Reserved1_RMASK 0x7F
+#define QIB_6120_IntClear_RcvUrg4IntClear_LSB 0x4
+#define QIB_6120_IntClear_RcvUrg4IntClear_RMASK 0x1
+#define QIB_6120_IntClear_RcvUrg3IntClear_LSB 0x3
+#define QIB_6120_IntClear_RcvUrg3IntClear_RMASK 0x1
+#define QIB_6120_IntClear_RcvUrg2IntClear_LSB 0x2
+#define QIB_6120_IntClear_RcvUrg2IntClear_RMASK 0x1
+#define QIB_6120_IntClear_RcvUrg1IntClear_LSB 0x1
+#define QIB_6120_IntClear_RcvUrg1IntClear_RMASK 0x1
+#define QIB_6120_IntClear_RcvUrg0IntClear_LSB 0x0
+#define QIB_6120_IntClear_RcvUrg0IntClear_RMASK 0x1
+
+#define QIB_6120_ErrMask_OFFS 0x80
+#define QIB_6120_ErrMask_Reserved_LSB 0x34
+#define QIB_6120_ErrMask_Reserved_RMASK 0xFFF
+#define QIB_6120_ErrMask_HardwareErrMask_LSB 0x33
+#define QIB_6120_ErrMask_HardwareErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_ResetNegatedMask_LSB 0x32
+#define QIB_6120_ErrMask_ResetNegatedMask_RMASK 0x1
+#define QIB_6120_ErrMask_InvalidAddrErrMask_LSB 0x31
+#define QIB_6120_ErrMask_InvalidAddrErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_IBStatusChangedMask_LSB 0x30
+#define QIB_6120_ErrMask_IBStatusChangedMask_RMASK 0x1
+#define QIB_6120_ErrMask_Reserved1_LSB 0x26
+#define QIB_6120_ErrMask_Reserved1_RMASK 0x3FF
+#define QIB_6120_ErrMask_SendUnsupportedVLErrMask_LSB 0x25
+#define QIB_6120_ErrMask_SendUnsupportedVLErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_SendUnexpectedPktNumErrMask_LSB 0x24
+#define QIB_6120_ErrMask_SendUnexpectedPktNumErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_SendPioArmLaunchErrMask_LSB 0x23
+#define QIB_6120_ErrMask_SendPioArmLaunchErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_SendDroppedDataPktErrMask_LSB 0x22
+#define QIB_6120_ErrMask_SendDroppedDataPktErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_SendDroppedSmpPktErrMask_LSB 0x21
+#define QIB_6120_ErrMask_SendDroppedSmpPktErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_SendPktLenErrMask_LSB 0x20
+#define QIB_6120_ErrMask_SendPktLenErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_SendUnderRunErrMask_LSB 0x1F
+#define QIB_6120_ErrMask_SendUnderRunErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_SendMaxPktLenErrMask_LSB 0x1E
+#define QIB_6120_ErrMask_SendMaxPktLenErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_SendMinPktLenErrMask_LSB 0x1D
+#define QIB_6120_ErrMask_SendMinPktLenErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_Reserved2_LSB 0x12
+#define QIB_6120_ErrMask_Reserved2_RMASK 0x7FF
+#define QIB_6120_ErrMask_RcvIBLostLinkErrMask_LSB 0x11
+#define QIB_6120_ErrMask_RcvIBLostLinkErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_RcvHdrErrMask_LSB 0x10
+#define QIB_6120_ErrMask_RcvHdrErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_RcvHdrLenErrMask_LSB 0xF
+#define QIB_6120_ErrMask_RcvHdrLenErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_RcvBadTidErrMask_LSB 0xE
+#define QIB_6120_ErrMask_RcvBadTidErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_RcvHdrFullErrMask_LSB 0xD
+#define QIB_6120_ErrMask_RcvHdrFullErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_RcvEgrFullErrMask_LSB 0xC
+#define QIB_6120_ErrMask_RcvEgrFullErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_RcvBadVersionErrMask_LSB 0xB
+#define QIB_6120_ErrMask_RcvBadVersionErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_RcvIBFlowErrMask_LSB 0xA
+#define QIB_6120_ErrMask_RcvIBFlowErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_RcvEBPErrMask_LSB 0x9
+#define QIB_6120_ErrMask_RcvEBPErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_RcvUnsupportedVLErrMask_LSB 0x8
+#define QIB_6120_ErrMask_RcvUnsupportedVLErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_RcvUnexpectedCharErrMask_LSB 0x7
+#define QIB_6120_ErrMask_RcvUnexpectedCharErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_RcvShortPktLenErrMask_LSB 0x6
+#define QIB_6120_ErrMask_RcvShortPktLenErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_RcvLongPktLenErrMask_LSB 0x5
+#define QIB_6120_ErrMask_RcvLongPktLenErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_RcvMaxPktLenErrMask_LSB 0x4
+#define QIB_6120_ErrMask_RcvMaxPktLenErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_RcvMinPktLenErrMask_LSB 0x3
+#define QIB_6120_ErrMask_RcvMinPktLenErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_RcvICRCErrMask_LSB 0x2
+#define QIB_6120_ErrMask_RcvICRCErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_RcvVCRCErrMask_LSB 0x1
+#define QIB_6120_ErrMask_RcvVCRCErrMask_RMASK 0x1
+#define QIB_6120_ErrMask_RcvFormatErrMask_LSB 0x0
+#define QIB_6120_ErrMask_RcvFormatErrMask_RMASK 0x1
+
+#define QIB_6120_ErrStatus_OFFS 0x88
+#define QIB_6120_ErrStatus_Reserved_LSB 0x34
+#define QIB_6120_ErrStatus_Reserved_RMASK 0xFFF
+#define QIB_6120_ErrStatus_HardwareErr_LSB 0x33
+#define QIB_6120_ErrStatus_HardwareErr_RMASK 0x1
+#define QIB_6120_ErrStatus_ResetNegated_LSB 0x32
+#define QIB_6120_ErrStatus_ResetNegated_RMASK 0x1
+#define QIB_6120_ErrStatus_InvalidAddrErr_LSB 0x31
+#define QIB_6120_ErrStatus_InvalidAddrErr_RMASK 0x1
+#define QIB_6120_ErrStatus_IBStatusChanged_LSB 0x30
+#define QIB_6120_ErrStatus_IBStatusChanged_RMASK 0x1
+#define QIB_6120_ErrStatus_Reserved1_LSB 0x26
+#define QIB_6120_ErrStatus_Reserved1_RMASK 0x3FF
+#define QIB_6120_ErrStatus_SendUnsupportedVLErr_LSB 0x25
+#define QIB_6120_ErrStatus_SendUnsupportedVLErr_RMASK 0x1
+#define QIB_6120_ErrStatus_SendUnexpectedPktNumErr_LSB 0x24
+#define QIB_6120_ErrStatus_SendUnexpectedPktNumErr_RMASK 0x1
+#define QIB_6120_ErrStatus_SendPioArmLaunchErr_LSB 0x23
+#define QIB_6120_ErrStatus_SendPioArmLaunchErr_RMASK 0x1
+#define QIB_6120_ErrStatus_SendDroppedDataPktErr_LSB 0x22
+#define QIB_6120_ErrStatus_SendDroppedDataPktErr_RMASK 0x1
+#define QIB_6120_ErrStatus_SendDroppedSmpPktErr_LSB 0x21
+#define QIB_6120_ErrStatus_SendDroppedSmpPktErr_RMASK 0x1
+#define QIB_6120_ErrStatus_SendPktLenErr_LSB 0x20
+#define QIB_6120_ErrStatus_SendPktLenErr_RMASK 0x1
+#define QIB_6120_ErrStatus_SendUnderRunErr_LSB 0x1F
+#define QIB_6120_ErrStatus_SendUnderRunErr_RMASK 0x1
+#define QIB_6120_ErrStatus_SendMaxPktLenErr_LSB 0x1E
+#define QIB_6120_ErrStatus_SendMaxPktLenErr_RMASK 0x1
+#define QIB_6120_ErrStatus_SendMinPktLenErr_LSB 0x1D
+#define QIB_6120_ErrStatus_SendMinPktLenErr_RMASK 0x1
+#define QIB_6120_ErrStatus_Reserved2_LSB 0x12
+#define QIB_6120_ErrStatus_Reserved2_RMASK 0x7FF
+#define QIB_6120_ErrStatus_RcvIBLostLinkErr_LSB 0x11
+#define QIB_6120_ErrStatus_RcvIBLostLinkErr_RMASK 0x1
+#define QIB_6120_ErrStatus_RcvHdrErr_LSB 0x10
+#define QIB_6120_ErrStatus_RcvHdrErr_RMASK 0x1
+#define QIB_6120_ErrStatus_RcvHdrLenErr_LSB 0xF
+#define QIB_6120_ErrStatus_RcvHdrLenErr_RMASK 0x1
+#define QIB_6120_ErrStatus_RcvBadTidErr_LSB 0xE
+#define QIB_6120_ErrStatus_RcvBadTidErr_RMASK 0x1
+#define QIB_6120_ErrStatus_RcvHdrFullErr_LSB 0xD
+#define QIB_6120_ErrStatus_RcvHdrFullErr_RMASK 0x1
+#define QIB_6120_ErrStatus_RcvEgrFullErr_LSB 0xC
+#define QIB_6120_ErrStatus_RcvEgrFullErr_RMASK 0x1
+#define QIB_6120_ErrStatus_RcvBadVersionErr_LSB 0xB
+#define QIB_6120_ErrStatus_RcvBadVersionErr_RMASK 0x1
+#define QIB_6120_ErrStatus_RcvIBFlowErr_LSB 0xA
+#define QIB_6120_ErrStatus_RcvIBFlowErr_RMASK 0x1
+#define QIB_6120_ErrStatus_RcvEBPErr_LSB 0x9
+#define QIB_6120_ErrStatus_RcvEBPErr_RMASK 0x1
+#define QIB_6120_ErrStatus_RcvUnsupportedVLErr_LSB 0x8
+#define QIB_6120_ErrStatus_RcvUnsupportedVLErr_RMASK 0x1
+#define QIB_6120_ErrStatus_RcvUnexpectedCharErr_LSB 0x7
+#define QIB_6120_ErrStatus_RcvUnexpectedCharErr_RMASK 0x1
+#define QIB_6120_ErrStatus_RcvShortPktLenErr_LSB 0x6
+#define QIB_6120_ErrStatus_RcvShortPktLenErr_RMASK 0x1
+#define QIB_6120_ErrStatus_RcvLongPktLenErr_LSB 0x5
+#define QIB_6120_ErrStatus_RcvLongPktLenErr_RMASK 0x1
+#define QIB_6120_ErrStatus_RcvMaxPktLenErr_LSB 0x4
+#define QIB_6120_ErrStatus_RcvMaxPktLenErr_RMASK 0x1
+#define QIB_6120_ErrStatus_RcvMinPktLenErr_LSB 0x3
+#define QIB_6120_ErrStatus_RcvMinPktLenErr_RMASK 0x1
+#define QIB_6120_ErrStatus_RcvICRCErr_LSB 0x2
+#define QIB_6120_ErrStatus_RcvICRCErr_RMASK 0x1
+#define QIB_6120_ErrStatus_RcvVCRCErr_LSB 0x1
+#define QIB_6120_ErrStatus_RcvVCRCErr_RMASK 0x1
+#define QIB_6120_ErrStatus_RcvFormatErr_LSB 0x0
+#define QIB_6120_ErrStatus_RcvFormatErr_RMASK 0x1
+
+#define QIB_6120_ErrClear_OFFS 0x90
+#define QIB_6120_ErrClear_Reserved_LSB 0x34
+#define QIB_6120_ErrClear_Reserved_RMASK 0xFFF
+#define QIB_6120_ErrClear_HardwareErrClear_LSB 0x33
+#define QIB_6120_ErrClear_HardwareErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_ResetNegatedClear_LSB 0x32
+#define QIB_6120_ErrClear_ResetNegatedClear_RMASK 0x1
+#define QIB_6120_ErrClear_InvalidAddrErrClear_LSB 0x31
+#define QIB_6120_ErrClear_InvalidAddrErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_IBStatusChangedClear_LSB 0x30
+#define QIB_6120_ErrClear_IBStatusChangedClear_RMASK 0x1
+#define QIB_6120_ErrClear_Reserved1_LSB 0x26
+#define QIB_6120_ErrClear_Reserved1_RMASK 0x3FF
+#define QIB_6120_ErrClear_SendUnsupportedVLErrClear_LSB 0x25
+#define QIB_6120_ErrClear_SendUnsupportedVLErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_SendUnexpectedPktNumErrClear_LSB 0x24
+#define QIB_6120_ErrClear_SendUnexpectedPktNumErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_SendPioArmLaunchErrClear_LSB 0x23
+#define QIB_6120_ErrClear_SendPioArmLaunchErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_SendDroppedDataPktErrClear_LSB 0x22
+#define QIB_6120_ErrClear_SendDroppedDataPktErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_SendDroppedSmpPktErrClear_LSB 0x21
+#define QIB_6120_ErrClear_SendDroppedSmpPktErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_SendPktLenErrClear_LSB 0x20
+#define QIB_6120_ErrClear_SendPktLenErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_SendUnderRunErrClear_LSB 0x1F
+#define QIB_6120_ErrClear_SendUnderRunErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_SendMaxPktLenErrClear_LSB 0x1E
+#define QIB_6120_ErrClear_SendMaxPktLenErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_SendMinPktLenErrClear_LSB 0x1D
+#define QIB_6120_ErrClear_SendMinPktLenErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_Reserved2_LSB 0x12
+#define QIB_6120_ErrClear_Reserved2_RMASK 0x7FF
+#define QIB_6120_ErrClear_RcvIBLostLinkErrClear_LSB 0x11
+#define QIB_6120_ErrClear_RcvIBLostLinkErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_RcvHdrErrClear_LSB 0x10
+#define QIB_6120_ErrClear_RcvHdrErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_RcvHdrLenErrClear_LSB 0xF
+#define QIB_6120_ErrClear_RcvHdrLenErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_RcvBadTidErrClear_LSB 0xE
+#define QIB_6120_ErrClear_RcvBadTidErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_RcvHdrFullErrClear_LSB 0xD
+#define QIB_6120_ErrClear_RcvHdrFullErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_RcvEgrFullErrClear_LSB 0xC
+#define QIB_6120_ErrClear_RcvEgrFullErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_RcvBadVersionErrClear_LSB 0xB
+#define QIB_6120_ErrClear_RcvBadVersionErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_RcvIBFlowErrClear_LSB 0xA
+#define QIB_6120_ErrClear_RcvIBFlowErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_RcvEBPErrClear_LSB 0x9
+#define QIB_6120_ErrClear_RcvEBPErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_RcvUnsupportedVLErrClear_LSB 0x8
+#define QIB_6120_ErrClear_RcvUnsupportedVLErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_RcvUnexpectedCharErrClear_LSB 0x7
+#define QIB_6120_ErrClear_RcvUnexpectedCharErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_RcvShortPktLenErrClear_LSB 0x6
+#define QIB_6120_ErrClear_RcvShortPktLenErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_RcvLongPktLenErrClear_LSB 0x5
+#define QIB_6120_ErrClear_RcvLongPktLenErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_RcvMaxPktLenErrClear_LSB 0x4
+#define QIB_6120_ErrClear_RcvMaxPktLenErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_RcvMinPktLenErrClear_LSB 0x3
+#define QIB_6120_ErrClear_RcvMinPktLenErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_RcvICRCErrClear_LSB 0x2
+#define QIB_6120_ErrClear_RcvICRCErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_RcvVCRCErrClear_LSB 0x1
+#define QIB_6120_ErrClear_RcvVCRCErrClear_RMASK 0x1
+#define QIB_6120_ErrClear_RcvFormatErrClear_LSB 0x0
+#define QIB_6120_ErrClear_RcvFormatErrClear_RMASK 0x1
+
+#define QIB_6120_HwErrMask_OFFS 0x98
+#define QIB_6120_HwErrMask_IBCBusFromSPCParityErrMask_LSB 0x3F
+#define QIB_6120_HwErrMask_IBCBusFromSPCParityErrMask_RMASK 0x1
+#define QIB_6120_HwErrMask_IBCBusToSPCParityErrMask_LSB 0x3E
+#define QIB_6120_HwErrMask_IBCBusToSPCParityErrMask_RMASK 0x1
+#define QIB_6120_HwErrMask_Reserved_LSB 0x3D
+#define QIB_6120_HwErrMask_Reserved_RMASK 0x1
+#define QIB_6120_HwErrMask_IBSerdesPClkNotDetectMask_LSB 0x3C
+#define QIB_6120_HwErrMask_IBSerdesPClkNotDetectMask_RMASK 0x1
+#define QIB_6120_HwErrMask_PCIESerdesQ0PClkNotDetectMask_LSB 0x3B
+#define QIB_6120_HwErrMask_PCIESerdesQ0PClkNotDetectMask_RMASK 0x1
+#define QIB_6120_HwErrMask_PCIESerdesQ1PClkNotDetectMask_LSB 0x3A
+#define QIB_6120_HwErrMask_PCIESerdesQ1PClkNotDetectMask_RMASK 0x1
+#define QIB_6120_HwErrMask_Reserved1_LSB 0x39
+#define QIB_6120_HwErrMask_Reserved1_RMASK 0x1
+#define QIB_6120_HwErrMask_IBPLLrfSlipMask_LSB 0x38
+#define QIB_6120_HwErrMask_IBPLLrfSlipMask_RMASK 0x1
+#define QIB_6120_HwErrMask_IBPLLfbSlipMask_LSB 0x37
+#define QIB_6120_HwErrMask_IBPLLfbSlipMask_RMASK 0x1
+#define QIB_6120_HwErrMask_PowerOnBISTFailedMask_LSB 0x36
+#define QIB_6120_HwErrMask_PowerOnBISTFailedMask_RMASK 0x1
+#define QIB_6120_HwErrMask_Reserved2_LSB 0x33
+#define QIB_6120_HwErrMask_Reserved2_RMASK 0x7
+#define QIB_6120_HwErrMask_RXEMemParityErrMask_LSB 0x2C
+#define QIB_6120_HwErrMask_RXEMemParityErrMask_RMASK 0x7F
+#define QIB_6120_HwErrMask_TXEMemParityErrMask_LSB 0x28
+#define QIB_6120_HwErrMask_TXEMemParityErrMask_RMASK 0xF
+#define QIB_6120_HwErrMask_Reserved3_LSB 0x22
+#define QIB_6120_HwErrMask_Reserved3_RMASK 0x3F
+#define QIB_6120_HwErrMask_PCIeBusParityErrMask_LSB 0x1F
+#define QIB_6120_HwErrMask_PCIeBusParityErrMask_RMASK 0x7
+#define QIB_6120_HwErrMask_PcieCplTimeoutMask_LSB 0x1E
+#define QIB_6120_HwErrMask_PcieCplTimeoutMask_RMASK 0x1
+#define QIB_6120_HwErrMask_PoisonedTLPMask_LSB 0x1D
+#define QIB_6120_HwErrMask_PoisonedTLPMask_RMASK 0x1
+#define QIB_6120_HwErrMask_Reserved4_LSB 0x6
+#define QIB_6120_HwErrMask_Reserved4_RMASK 0x7FFFFF
+#define QIB_6120_HwErrMask_PCIeMemParityErrMask_LSB 0x0
+#define QIB_6120_HwErrMask_PCIeMemParityErrMask_RMASK 0x3F
+
+#define QIB_6120_HwErrStatus_OFFS 0xA0
+#define QIB_6120_HwErrStatus_IBCBusFromSPCParityErr_LSB 0x3F
+#define QIB_6120_HwErrStatus_IBCBusFromSPCParityErr_RMASK 0x1
+#define QIB_6120_HwErrStatus_IBCBusToSPCParityErr_LSB 0x3E
+#define QIB_6120_HwErrStatus_IBCBusToSPCParityErr_RMASK 0x1
+#define QIB_6120_HwErrStatus_Reserved_LSB 0x3D
+#define QIB_6120_HwErrStatus_Reserved_RMASK 0x1
+#define QIB_6120_HwErrStatus_IBSerdesPClkNotDetect_LSB 0x3C
+#define QIB_6120_HwErrStatus_IBSerdesPClkNotDetect_RMASK 0x1
+#define QIB_6120_HwErrStatus_PCIESerdesQ0PClkNotDetect_LSB 0x3B
+#define QIB_6120_HwErrStatus_PCIESerdesQ0PClkNotDetect_RMASK 0x1
+#define QIB_6120_HwErrStatus_PCIESerdesQ1PClkNotDetect_LSB 0x3A
+#define QIB_6120_HwErrStatus_PCIESerdesQ1PClkNotDetect_RMASK 0x1
+#define QIB_6120_HwErrStatus_Reserved1_LSB 0x39
+#define QIB_6120_HwErrStatus_Reserved1_RMASK 0x1
+#define QIB_6120_HwErrStatus_IBPLLrfSlip_LSB 0x38
+#define QIB_6120_HwErrStatus_IBPLLrfSlip_RMASK 0x1
+#define QIB_6120_HwErrStatus_IBPLLfbSlip_LSB 0x37
+#define QIB_6120_HwErrStatus_IBPLLfbSlip_RMASK 0x1
+#define QIB_6120_HwErrStatus_PowerOnBISTFailed_LSB 0x36
+#define QIB_6120_HwErrStatus_PowerOnBISTFailed_RMASK 0x1
+#define QIB_6120_HwErrStatus_Reserved2_LSB 0x33
+#define QIB_6120_HwErrStatus_Reserved2_RMASK 0x7
+#define QIB_6120_HwErrStatus_RXEMemParity_LSB 0x2C
+#define QIB_6120_HwErrStatus_RXEMemParity_RMASK 0x7F
+#define QIB_6120_HwErrStatus_TXEMemParity_LSB 0x28
+#define QIB_6120_HwErrStatus_TXEMemParity_RMASK 0xF
+#define QIB_6120_HwErrStatus_Reserved3_LSB 0x22
+#define QIB_6120_HwErrStatus_Reserved3_RMASK 0x3F
+#define QIB_6120_HwErrStatus_PCIeBusParity_LSB 0x1F
+#define QIB_6120_HwErrStatus_PCIeBusParity_RMASK 0x7
+#define QIB_6120_HwErrStatus_PcieCplTimeout_LSB 0x1E
+#define QIB_6120_HwErrStatus_PcieCplTimeout_RMASK 0x1
+#define QIB_6120_HwErrStatus_PoisenedTLP_LSB 0x1D
+#define QIB_6120_HwErrStatus_PoisenedTLP_RMASK 0x1
+#define QIB_6120_HwErrStatus_Reserved4_LSB 0x6
+#define QIB_6120_HwErrStatus_Reserved4_RMASK 0x7FFFFF
+#define QIB_6120_HwErrStatus_PCIeMemParity_LSB 0x0
+#define QIB_6120_HwErrStatus_PCIeMemParity_RMASK 0x3F
+
+#define QIB_6120_HwErrClear_OFFS 0xA8
+#define QIB_6120_HwErrClear_IBCBusFromSPCParityErrClear_LSB 0x3F
+#define QIB_6120_HwErrClear_IBCBusFromSPCParityErrClear_RMASK 0x1
+#define QIB_6120_HwErrClear_IBCBusToSPCparityErrClear_LSB 0x3E
+#define QIB_6120_HwErrClear_IBCBusToSPCparityErrClear_RMASK 0x1
+#define QIB_6120_HwErrClear_Reserved_LSB 0x3D
+#define QIB_6120_HwErrClear_Reserved_RMASK 0x1
+#define QIB_6120_HwErrClear_IBSerdesPClkNotDetectClear_LSB 0x3C
+#define QIB_6120_HwErrClear_IBSerdesPClkNotDetectClear_RMASK 0x1
+#define QIB_6120_HwErrClear_PCIESerdesQ0PClkNotDetectClear_LSB 0x3B
+#define QIB_6120_HwErrClear_PCIESerdesQ0PClkNotDetectClear_RMASK 0x1
+#define QIB_6120_HwErrClear_PCIESerdesQ1PClkNotDetectClear_LSB 0x3A
+#define QIB_6120_HwErrClear_PCIESerdesQ1PClkNotDetectClear_RMASK 0x1
+#define QIB_6120_HwErrClear_Reserved1_LSB 0x39
+#define QIB_6120_HwErrClear_Reserved1_RMASK 0x1
+#define QIB_6120_HwErrClear_IBPLLrfSlipClear_LSB 0x38
+#define QIB_6120_HwErrClear_IBPLLrfSlipClear_RMASK 0x1
+#define QIB_6120_HwErrClear_IBPLLfbSlipClear_LSB 0x37
+#define QIB_6120_HwErrClear_IBPLLfbSlipClear_RMASK 0x1
+#define QIB_6120_HwErrClear_PowerOnBISTFailedClear_LSB 0x36
+#define QIB_6120_HwErrClear_PowerOnBISTFailedClear_RMASK 0x1
+#define QIB_6120_HwErrClear_Reserved2_LSB 0x33
+#define QIB_6120_HwErrClear_Reserved2_RMASK 0x7
+#define QIB_6120_HwErrClear_RXEMemParityClear_LSB 0x2C
+#define QIB_6120_HwErrClear_RXEMemParityClear_RMASK 0x7F
+#define QIB_6120_HwErrClear_TXEMemParityClear_LSB 0x28
+#define QIB_6120_HwErrClear_TXEMemParityClear_RMASK 0xF
+#define QIB_6120_HwErrClear_Reserved3_LSB 0x22
+#define QIB_6120_HwErrClear_Reserved3_RMASK 0x3F
+#define QIB_6120_HwErrClear_PCIeBusParityClr_LSB 0x1F
+#define QIB_6120_HwErrClear_PCIeBusParityClr_RMASK 0x7
+#define QIB_6120_HwErrClear_PcieCplTimeoutClear_LSB 0x1E
+#define QIB_6120_HwErrClear_PcieCplTimeoutClear_RMASK 0x1
+#define QIB_6120_HwErrClear_PoisonedTLPClear_LSB 0x1D
+#define QIB_6120_HwErrClear_PoisonedTLPClear_RMASK 0x1
+#define QIB_6120_HwErrClear_Reserved4_LSB 0x6
+#define QIB_6120_HwErrClear_Reserved4_RMASK 0x7FFFFF
+#define QIB_6120_HwErrClear_PCIeMemParityClr_LSB 0x0
+#define QIB_6120_HwErrClear_PCIeMemParityClr_RMASK 0x3F
+
+#define QIB_6120_HwDiagCtrl_OFFS 0xB0
+#define QIB_6120_HwDiagCtrl_ForceIBCBusFromSPCParityErr_LSB 0x3F
+#define QIB_6120_HwDiagCtrl_ForceIBCBusFromSPCParityErr_RMASK 0x1
+#define QIB_6120_HwDiagCtrl_ForceIBCBusToSPCParityErr_LSB 0x3E
+#define QIB_6120_HwDiagCtrl_ForceIBCBusToSPCParityErr_RMASK 0x1
+#define QIB_6120_HwDiagCtrl_CounterWrEnable_LSB 0x3D
+#define QIB_6120_HwDiagCtrl_CounterWrEnable_RMASK 0x1
+#define QIB_6120_HwDiagCtrl_CounterDisable_LSB 0x3C
+#define QIB_6120_HwDiagCtrl_CounterDisable_RMASK 0x1
+#define QIB_6120_HwDiagCtrl_Reserved_LSB 0x33
+#define QIB_6120_HwDiagCtrl_Reserved_RMASK 0x1FF
+#define QIB_6120_HwDiagCtrl_ForceRxMemParityErr_LSB 0x2C
+#define QIB_6120_HwDiagCtrl_ForceRxMemParityErr_RMASK 0x7F
+#define QIB_6120_HwDiagCtrl_ForceTxMemparityErr_LSB 0x28
+#define QIB_6120_HwDiagCtrl_ForceTxMemparityErr_RMASK 0xF
+#define QIB_6120_HwDiagCtrl_Reserved1_LSB 0x23
+#define QIB_6120_HwDiagCtrl_Reserved1_RMASK 0x1F
+#define QIB_6120_HwDiagCtrl_forcePCIeBusParity_LSB 0x1F
+#define QIB_6120_HwDiagCtrl_forcePCIeBusParity_RMASK 0xF
+#define QIB_6120_HwDiagCtrl_Reserved2_LSB 0x6
+#define QIB_6120_HwDiagCtrl_Reserved2_RMASK 0x1FFFFFF
+#define QIB_6120_HwDiagCtrl_forcePCIeMemParity_LSB 0x0
+#define QIB_6120_HwDiagCtrl_forcePCIeMemParity_RMASK 0x3F
+
+#define QIB_6120_IBCStatus_OFFS 0xC0
+#define QIB_6120_IBCStatus_TxCreditOk_LSB 0x1F
+#define QIB_6120_IBCStatus_TxCreditOk_RMASK 0x1
+#define QIB_6120_IBCStatus_TxReady_LSB 0x1E
+#define QIB_6120_IBCStatus_TxReady_RMASK 0x1
+#define QIB_6120_IBCStatus_Reserved_LSB 0x7
+#define QIB_6120_IBCStatus_Reserved_RMASK 0x7FFFFF
+#define QIB_6120_IBCStatus_LinkState_LSB 0x4
+#define QIB_6120_IBCStatus_LinkState_RMASK 0x7
+#define QIB_6120_IBCStatus_LinkTrainingState_LSB 0x0
+#define QIB_6120_IBCStatus_LinkTrainingState_RMASK 0xF
+
+#define QIB_6120_IBCCtrl_OFFS 0xC8
+#define QIB_6120_IBCCtrl_Loopback_LSB 0x3F
+#define QIB_6120_IBCCtrl_Loopback_RMASK 0x1
+#define QIB_6120_IBCCtrl_LinkDownDefaultState_LSB 0x3E
+#define QIB_6120_IBCCtrl_LinkDownDefaultState_RMASK 0x1
+#define QIB_6120_IBCCtrl_Reserved_LSB 0x2B
+#define QIB_6120_IBCCtrl_Reserved_RMASK 0x7FFFF
+#define QIB_6120_IBCCtrl_CreditScale_LSB 0x28
+#define QIB_6120_IBCCtrl_CreditScale_RMASK 0x7
+#define QIB_6120_IBCCtrl_OverrunThreshold_LSB 0x24
+#define QIB_6120_IBCCtrl_OverrunThreshold_RMASK 0xF
+#define QIB_6120_IBCCtrl_PhyerrThreshold_LSB 0x20
+#define QIB_6120_IBCCtrl_PhyerrThreshold_RMASK 0xF
+#define QIB_6120_IBCCtrl_Reserved1_LSB 0x1F
+#define QIB_6120_IBCCtrl_Reserved1_RMASK 0x1
+#define QIB_6120_IBCCtrl_MaxPktLen_LSB 0x14
+#define QIB_6120_IBCCtrl_MaxPktLen_RMASK 0x7FF
+#define QIB_6120_IBCCtrl_LinkCmd_LSB 0x12
+#define QIB_6120_IBCCtrl_LinkCmd_RMASK 0x3
+#define QIB_6120_IBCCtrl_LinkInitCmd_LSB 0x10
+#define QIB_6120_IBCCtrl_LinkInitCmd_RMASK 0x3
+#define QIB_6120_IBCCtrl_FlowCtrlWaterMark_LSB 0x8
+#define QIB_6120_IBCCtrl_FlowCtrlWaterMark_RMASK 0xFF
+#define QIB_6120_IBCCtrl_FlowCtrlPeriod_LSB 0x0
+#define QIB_6120_IBCCtrl_FlowCtrlPeriod_RMASK 0xFF
+
+#define QIB_6120_EXTStatus_OFFS 0xD0
+#define QIB_6120_EXTStatus_GPIOIn_LSB 0x30
+#define QIB_6120_EXTStatus_GPIOIn_RMASK 0xFFFF
+#define QIB_6120_EXTStatus_Reserved_LSB 0x20
+#define QIB_6120_EXTStatus_Reserved_RMASK 0xFFFF
+#define QIB_6120_EXTStatus_Reserved1_LSB 0x10
+#define QIB_6120_EXTStatus_Reserved1_RMASK 0xFFFF
+#define QIB_6120_EXTStatus_MemBISTFoundErr_LSB 0xF
+#define QIB_6120_EXTStatus_MemBISTFoundErr_RMASK 0x1
+#define QIB_6120_EXTStatus_MemBISTEndTest_LSB 0xE
+#define QIB_6120_EXTStatus_MemBISTEndTest_RMASK 0x1
+#define QIB_6120_EXTStatus_Reserved2_LSB 0x0
+#define QIB_6120_EXTStatus_Reserved2_RMASK 0x3FFF
+
+#define QIB_6120_EXTCtrl_OFFS 0xD8
+#define QIB_6120_EXTCtrl_GPIOOe_LSB 0x30
+#define QIB_6120_EXTCtrl_GPIOOe_RMASK 0xFFFF
+#define QIB_6120_EXTCtrl_GPIOInvert_LSB 0x20
+#define QIB_6120_EXTCtrl_GPIOInvert_RMASK 0xFFFF
+#define QIB_6120_EXTCtrl_Reserved_LSB 0x4
+#define QIB_6120_EXTCtrl_Reserved_RMASK 0xFFFFFFF
+#define QIB_6120_EXTCtrl_LEDPriPortGreenOn_LSB 0x3
+#define QIB_6120_EXTCtrl_LEDPriPortGreenOn_RMASK 0x1
+#define QIB_6120_EXTCtrl_LEDPriPortYellowOn_LSB 0x2
+#define QIB_6120_EXTCtrl_LEDPriPortYellowOn_RMASK 0x1
+#define QIB_6120_EXTCtrl_LEDGblOkGreenOn_LSB 0x1
+#define QIB_6120_EXTCtrl_LEDGblOkGreenOn_RMASK 0x1
+#define QIB_6120_EXTCtrl_LEDGblErrRedOff_LSB 0x0
+#define QIB_6120_EXTCtrl_LEDGblErrRedOff_RMASK 0x1
+
+#define QIB_6120_GPIOOut_OFFS 0xE0
+
+#define QIB_6120_GPIOMask_OFFS 0xE8
+
+#define QIB_6120_GPIOStatus_OFFS 0xF0
+
+#define QIB_6120_GPIOClear_OFFS 0xF8
+
+#define QIB_6120_RcvCtrl_OFFS 0x100
+#define QIB_6120_RcvCtrl_TailUpd_LSB 0x1F
+#define QIB_6120_RcvCtrl_TailUpd_RMASK 0x1
+#define QIB_6120_RcvCtrl_RcvPartitionKeyDisable_LSB 0x1E
+#define QIB_6120_RcvCtrl_RcvPartitionKeyDisable_RMASK 0x1
+#define QIB_6120_RcvCtrl_Reserved_LSB 0x15
+#define QIB_6120_RcvCtrl_Reserved_RMASK 0x1FF
+#define QIB_6120_RcvCtrl_IntrAvail_LSB 0x10
+#define QIB_6120_RcvCtrl_IntrAvail_RMASK 0x1F
+#define QIB_6120_RcvCtrl_Reserved1_LSB 0x9
+#define QIB_6120_RcvCtrl_Reserved1_RMASK 0x7F
+#define QIB_6120_RcvCtrl_Reserved2_LSB 0x5
+#define QIB_6120_RcvCtrl_Reserved2_RMASK 0xF
+#define QIB_6120_RcvCtrl_PortEnable_LSB 0x0
+#define QIB_6120_RcvCtrl_PortEnable_RMASK 0x1F
+
+#define QIB_6120_RcvBTHQP_OFFS 0x108
+#define QIB_6120_RcvBTHQP_BTHQP_Mask_LSB 0x1E
+#define QIB_6120_RcvBTHQP_BTHQP_Mask_RMASK 0x3
+#define QIB_6120_RcvBTHQP_Reserved_LSB 0x18
+#define QIB_6120_RcvBTHQP_Reserved_RMASK 0x3F
+#define QIB_6120_RcvBTHQP_RcvBTHQP_LSB 0x0
+#define QIB_6120_RcvBTHQP_RcvBTHQP_RMASK 0xFFFFFF
+
+#define QIB_6120_RcvHdrSize_OFFS 0x110
+
+#define QIB_6120_RcvHdrCnt_OFFS 0x118
+
+#define QIB_6120_RcvHdrEntSize_OFFS 0x120
+
+#define QIB_6120_RcvTIDBase_OFFS 0x128
+
+#define QIB_6120_RcvTIDCnt_OFFS 0x130
+
+#define QIB_6120_RcvEgrBase_OFFS 0x138
+
+#define QIB_6120_RcvEgrCnt_OFFS 0x140
+
+#define QIB_6120_RcvBufBase_OFFS 0x148
+
+#define QIB_6120_RcvBufSize_OFFS 0x150
+
+#define QIB_6120_RxIntMemBase_OFFS 0x158
+
+#define QIB_6120_RxIntMemSize_OFFS 0x160
+
+#define QIB_6120_RcvPartitionKey_OFFS 0x168
+
+#define QIB_6120_RcvPktLEDCnt_OFFS 0x178
+#define QIB_6120_RcvPktLEDCnt_ONperiod_LSB 0x20
+#define QIB_6120_RcvPktLEDCnt_ONperiod_RMASK 0xFFFFFFFF
+#define QIB_6120_RcvPktLEDCnt_OFFperiod_LSB 0x0
+#define QIB_6120_RcvPktLEDCnt_OFFperiod_RMASK 0xFFFFFFFF
+
+#define QIB_6120_SendCtrl_OFFS 0x1C0
+#define QIB_6120_SendCtrl_Disarm_LSB 0x1F
+#define QIB_6120_SendCtrl_Disarm_RMASK 0x1
+#define QIB_6120_SendCtrl_Reserved_LSB 0x17
+#define QIB_6120_SendCtrl_Reserved_RMASK 0xFF
+#define QIB_6120_SendCtrl_DisarmPIOBuf_LSB 0x10
+#define QIB_6120_SendCtrl_DisarmPIOBuf_RMASK 0x7F
+#define QIB_6120_SendCtrl_Reserved1_LSB 0x4
+#define QIB_6120_SendCtrl_Reserved1_RMASK 0xFFF
+#define QIB_6120_SendCtrl_PIOEnable_LSB 0x3
+#define QIB_6120_SendCtrl_PIOEnable_RMASK 0x1
+#define QIB_6120_SendCtrl_PIOBufAvailUpd_LSB 0x2
+#define QIB_6120_SendCtrl_PIOBufAvailUpd_RMASK 0x1
+#define QIB_6120_SendCtrl_PIOIntBufAvail_LSB 0x1
+#define QIB_6120_SendCtrl_PIOIntBufAvail_RMASK 0x1
+#define QIB_6120_SendCtrl_Abort_LSB 0x0
+#define QIB_6120_SendCtrl_Abort_RMASK 0x1
+
+#define QIB_6120_SendPIOBufBase_OFFS 0x1C8
+#define QIB_6120_SendPIOBufBase_Reserved_LSB 0x35
+#define QIB_6120_SendPIOBufBase_Reserved_RMASK 0x7FF
+#define QIB_6120_SendPIOBufBase_BaseAddr_LargePIO_LSB 0x20
+#define QIB_6120_SendPIOBufBase_BaseAddr_LargePIO_RMASK 0x1FFFFF
+#define QIB_6120_SendPIOBufBase_Reserved1_LSB 0x15
+#define QIB_6120_SendPIOBufBase_Reserved1_RMASK 0x7FF
+#define QIB_6120_SendPIOBufBase_BaseAddr_SmallPIO_LSB 0x0
+#define QIB_6120_SendPIOBufBase_BaseAddr_SmallPIO_RMASK 0x1FFFFF
+
+#define QIB_6120_SendPIOSize_OFFS 0x1D0
+#define QIB_6120_SendPIOSize_Reserved_LSB 0x2D
+#define QIB_6120_SendPIOSize_Reserved_RMASK 0xFFFFF
+#define QIB_6120_SendPIOSize_Size_LargePIO_LSB 0x20
+#define QIB_6120_SendPIOSize_Size_LargePIO_RMASK 0x1FFF
+#define QIB_6120_SendPIOSize_Reserved1_LSB 0xC
+#define QIB_6120_SendPIOSize_Reserved1_RMASK 0xFFFFF
+#define QIB_6120_SendPIOSize_Size_SmallPIO_LSB 0x0
+#define QIB_6120_SendPIOSize_Size_SmallPIO_RMASK 0xFFF
+
+#define QIB_6120_SendPIOBufCnt_OFFS 0x1D8
+#define QIB_6120_SendPIOBufCnt_Reserved_LSB 0x24
+#define QIB_6120_SendPIOBufCnt_Reserved_RMASK 0xFFFFFFF
+#define QIB_6120_SendPIOBufCnt_Num_LargePIO_LSB 0x20
+#define QIB_6120_SendPIOBufCnt_Num_LargePIO_RMASK 0xF
+#define QIB_6120_SendPIOBufCnt_Reserved1_LSB 0x9
+#define QIB_6120_SendPIOBufCnt_Reserved1_RMASK 0x7FFFFF
+#define QIB_6120_SendPIOBufCnt_Num_SmallPIO_LSB 0x0
+#define QIB_6120_SendPIOBufCnt_Num_SmallPIO_RMASK 0x1FF
+
+#define QIB_6120_SendPIOAvailAddr_OFFS 0x1E0
+#define QIB_6120_SendPIOAvailAddr_SendPIOAvailAddr_LSB 0x6
+#define QIB_6120_SendPIOAvailAddr_SendPIOAvailAddr_RMASK 0x3FFFFFFFF
+#define QIB_6120_SendPIOAvailAddr_Reserved_LSB 0x0
+#define QIB_6120_SendPIOAvailAddr_Reserved_RMASK 0x3F
+
+#define QIB_6120_SendBufErr0_OFFS 0x240
+#define QIB_6120_SendBufErr0_SendBufErrPIO_63_0_LSB 0x0
+#define QIB_6120_SendBufErr0_SendBufErrPIO_63_0_RMASK 0x0
+
+#define QIB_6120_RcvHdrAddr0_OFFS 0x280
+#define QIB_6120_RcvHdrAddr0_RcvHdrAddr0_LSB 0x2
+#define QIB_6120_RcvHdrAddr0_RcvHdrAddr0_RMASK 0x3FFFFFFFFF
+#define QIB_6120_RcvHdrAddr0_Reserved_LSB 0x0
+#define QIB_6120_RcvHdrAddr0_Reserved_RMASK 0x3
+
+#define QIB_6120_RcvHdrTailAddr0_OFFS 0x300
+#define QIB_6120_RcvHdrTailAddr0_RcvHdrTailAddr0_LSB 0x2
+#define QIB_6120_RcvHdrTailAddr0_RcvHdrTailAddr0_RMASK 0x3FFFFFFFFF
+#define QIB_6120_RcvHdrTailAddr0_Reserved_LSB 0x0
+#define QIB_6120_RcvHdrTailAddr0_Reserved_RMASK 0x3
+
+#define QIB_6120_SerdesCfg0_OFFS 0x3C0
+#define QIB_6120_SerdesCfg0_DisableIBTxIdleDetect_LSB 0x3F
+#define QIB_6120_SerdesCfg0_DisableIBTxIdleDetect_RMASK 0x1
+#define QIB_6120_SerdesCfg0_Reserved_LSB 0x38
+#define QIB_6120_SerdesCfg0_Reserved_RMASK 0x7F
+#define QIB_6120_SerdesCfg0_RxEqCtl_LSB 0x36
+#define QIB_6120_SerdesCfg0_RxEqCtl_RMASK 0x3
+#define QIB_6120_SerdesCfg0_TxTermAdj_LSB 0x34
+#define QIB_6120_SerdesCfg0_TxTermAdj_RMASK 0x3
+#define QIB_6120_SerdesCfg0_RxTermAdj_LSB 0x32
+#define QIB_6120_SerdesCfg0_RxTermAdj_RMASK 0x3
+#define QIB_6120_SerdesCfg0_TermAdj1_LSB 0x31
+#define QIB_6120_SerdesCfg0_TermAdj1_RMASK 0x1
+#define QIB_6120_SerdesCfg0_TermAdj0_LSB 0x30
+#define QIB_6120_SerdesCfg0_TermAdj0_RMASK 0x1
+#define QIB_6120_SerdesCfg0_LPBKA_LSB 0x2F
+#define QIB_6120_SerdesCfg0_LPBKA_RMASK 0x1
+#define QIB_6120_SerdesCfg0_LPBKB_LSB 0x2E
+#define QIB_6120_SerdesCfg0_LPBKB_RMASK 0x1
+#define QIB_6120_SerdesCfg0_LPBKC_LSB 0x2D
+#define QIB_6120_SerdesCfg0_LPBKC_RMASK 0x1
+#define QIB_6120_SerdesCfg0_LPBKD_LSB 0x2C
+#define QIB_6120_SerdesCfg0_LPBKD_RMASK 0x1
+#define QIB_6120_SerdesCfg0_PW_LSB 0x2B
+#define QIB_6120_SerdesCfg0_PW_RMASK 0x1
+#define QIB_6120_SerdesCfg0_RefSel_LSB 0x29
+#define QIB_6120_SerdesCfg0_RefSel_RMASK 0x3
+#define QIB_6120_SerdesCfg0_ParReset_LSB 0x28
+#define QIB_6120_SerdesCfg0_ParReset_RMASK 0x1
+#define QIB_6120_SerdesCfg0_ParLPBK_LSB 0x27
+#define QIB_6120_SerdesCfg0_ParLPBK_RMASK 0x1
+#define QIB_6120_SerdesCfg0_OffsetEn_LSB 0x26
+#define QIB_6120_SerdesCfg0_OffsetEn_RMASK 0x1
+#define QIB_6120_SerdesCfg0_Offset_LSB 0x1E
+#define QIB_6120_SerdesCfg0_Offset_RMASK 0xFF
+#define QIB_6120_SerdesCfg0_L2PwrDn_LSB 0x1D
+#define QIB_6120_SerdesCfg0_L2PwrDn_RMASK 0x1
+#define QIB_6120_SerdesCfg0_ResetPLL_LSB 0x1C
+#define QIB_6120_SerdesCfg0_ResetPLL_RMASK 0x1
+#define QIB_6120_SerdesCfg0_RxTermEnX_LSB 0x18
+#define QIB_6120_SerdesCfg0_RxTermEnX_RMASK 0xF
+#define QIB_6120_SerdesCfg0_BeaconTxEnX_LSB 0x14
+#define QIB_6120_SerdesCfg0_BeaconTxEnX_RMASK 0xF
+#define QIB_6120_SerdesCfg0_RxDetEnX_LSB 0x10
+#define QIB_6120_SerdesCfg0_RxDetEnX_RMASK 0xF
+#define QIB_6120_SerdesCfg0_TxIdeEnX_LSB 0xC
+#define QIB_6120_SerdesCfg0_TxIdeEnX_RMASK 0xF
+#define QIB_6120_SerdesCfg0_RxIdleEnX_LSB 0x8
+#define QIB_6120_SerdesCfg0_RxIdleEnX_RMASK 0xF
+#define QIB_6120_SerdesCfg0_L1PwrDnA_LSB 0x7
+#define QIB_6120_SerdesCfg0_L1PwrDnA_RMASK 0x1
+#define QIB_6120_SerdesCfg0_L1PwrDnB_LSB 0x6
+#define QIB_6120_SerdesCfg0_L1PwrDnB_RMASK 0x1
+#define QIB_6120_SerdesCfg0_L1PwrDnC_LSB 0x5
+#define QIB_6120_SerdesCfg0_L1PwrDnC_RMASK 0x1
+#define QIB_6120_SerdesCfg0_L1PwrDnD_LSB 0x4
+#define QIB_6120_SerdesCfg0_L1PwrDnD_RMASK 0x1
+#define QIB_6120_SerdesCfg0_ResetA_LSB 0x3
+#define QIB_6120_SerdesCfg0_ResetA_RMASK 0x1
+#define QIB_6120_SerdesCfg0_ResetB_LSB 0x2
+#define QIB_6120_SerdesCfg0_ResetB_RMASK 0x1
+#define QIB_6120_SerdesCfg0_ResetC_LSB 0x1
+#define QIB_6120_SerdesCfg0_ResetC_RMASK 0x1
+#define QIB_6120_SerdesCfg0_ResetD_LSB 0x0
+#define QIB_6120_SerdesCfg0_ResetD_RMASK 0x1
+
+#define QIB_6120_SerdesStat_OFFS 0x3D0
+#define QIB_6120_SerdesStat_Reserved_LSB 0xC
+#define QIB_6120_SerdesStat_Reserved_RMASK 0xFFFFFFFFFFFFF
+#define QIB_6120_SerdesStat_BeaconDetA_LSB 0xB
+#define QIB_6120_SerdesStat_BeaconDetA_RMASK 0x1
+#define QIB_6120_SerdesStat_BeaconDetB_LSB 0xA
+#define QIB_6120_SerdesStat_BeaconDetB_RMASK 0x1
+#define QIB_6120_SerdesStat_BeaconDetC_LSB 0x9
+#define QIB_6120_SerdesStat_BeaconDetC_RMASK 0x1
+#define QIB_6120_SerdesStat_BeaconDetD_LSB 0x8
+#define QIB_6120_SerdesStat_BeaconDetD_RMASK 0x1
+#define QIB_6120_SerdesStat_RxDetA_LSB 0x7
+#define QIB_6120_SerdesStat_RxDetA_RMASK 0x1
+#define QIB_6120_SerdesStat_RxDetB_LSB 0x6
+#define QIB_6120_SerdesStat_RxDetB_RMASK 0x1
+#define QIB_6120_SerdesStat_RxDetC_LSB 0x5
+#define QIB_6120_SerdesStat_RxDetC_RMASK 0x1
+#define QIB_6120_SerdesStat_RxDetD_LSB 0x4
+#define QIB_6120_SerdesStat_RxDetD_RMASK 0x1
+#define QIB_6120_SerdesStat_TxIdleDetA_LSB 0x3
+#define QIB_6120_SerdesStat_TxIdleDetA_RMASK 0x1
+#define QIB_6120_SerdesStat_TxIdleDetB_LSB 0x2
+#define QIB_6120_SerdesStat_TxIdleDetB_RMASK 0x1
+#define QIB_6120_SerdesStat_TxIdleDetC_LSB 0x1
+#define QIB_6120_SerdesStat_TxIdleDetC_RMASK 0x1
+#define QIB_6120_SerdesStat_TxIdleDetD_LSB 0x0
+#define QIB_6120_SerdesStat_TxIdleDetD_RMASK 0x1
+
+#define QIB_6120_XGXSCfg_OFFS 0x3D8
+#define QIB_6120_XGXSCfg_ArmLaunchErrorDisable_LSB 0x3F
+#define QIB_6120_XGXSCfg_ArmLaunchErrorDisable_RMASK 0x1
+#define QIB_6120_XGXSCfg_Reserved_LSB 0x17
+#define QIB_6120_XGXSCfg_Reserved_RMASK 0xFFFFFFFFFF
+#define QIB_6120_XGXSCfg_polarity_inv_LSB 0x13
+#define QIB_6120_XGXSCfg_polarity_inv_RMASK 0xF
+#define QIB_6120_XGXSCfg_link_sync_mask_LSB 0x9
+#define QIB_6120_XGXSCfg_link_sync_mask_RMASK 0x3FF
+#define QIB_6120_XGXSCfg_port_addr_LSB 0x4
+#define QIB_6120_XGXSCfg_port_addr_RMASK 0x1F
+#define QIB_6120_XGXSCfg_mdd_30_LSB 0x3
+#define QIB_6120_XGXSCfg_mdd_30_RMASK 0x1
+#define QIB_6120_XGXSCfg_xcv_resetn_LSB 0x2
+#define QIB_6120_XGXSCfg_xcv_resetn_RMASK 0x1
+#define QIB_6120_XGXSCfg_Reserved1_LSB 0x1
+#define QIB_6120_XGXSCfg_Reserved1_RMASK 0x1
+#define QIB_6120_XGXSCfg_tx_rx_resetn_LSB 0x0
+#define QIB_6120_XGXSCfg_tx_rx_resetn_RMASK 0x1
+
+#define QIB_6120_LBIntCnt_OFFS 0x12000
+
+#define QIB_6120_LBFlowStallCnt_OFFS 0x12008
+
+#define QIB_6120_TxUnsupVLErrCnt_OFFS 0x12018
+
+#define QIB_6120_TxDataPktCnt_OFFS 0x12020
+
+#define QIB_6120_TxFlowPktCnt_OFFS 0x12028
+
+#define QIB_6120_TxDwordCnt_OFFS 0x12030
+
+#define QIB_6120_TxLenErrCnt_OFFS 0x12038
+
+#define QIB_6120_TxMaxMinLenErrCnt_OFFS 0x12040
+
+#define QIB_6120_TxUnderrunCnt_OFFS 0x12048
+
+#define QIB_6120_TxFlowStallCnt_OFFS 0x12050
+
+#define QIB_6120_TxDroppedPktCnt_OFFS 0x12058
+
+#define QIB_6120_RxDroppedPktCnt_OFFS 0x12060
+
+#define QIB_6120_RxDataPktCnt_OFFS 0x12068
+
+#define QIB_6120_RxFlowPktCnt_OFFS 0x12070
+
+#define QIB_6120_RxDwordCnt_OFFS 0x12078
+
+#define QIB_6120_RxLenErrCnt_OFFS 0x12080
+
+#define QIB_6120_RxMaxMinLenErrCnt_OFFS 0x12088
+
+#define QIB_6120_RxICRCErrCnt_OFFS 0x12090
+
+#define QIB_6120_RxVCRCErrCnt_OFFS 0x12098
+
+#define QIB_6120_RxFlowCtrlErrCnt_OFFS 0x120A0
+
+#define QIB_6120_RxBadFormatCnt_OFFS 0x120A8
+
+#define QIB_6120_RxLinkProblemCnt_OFFS 0x120B0
+
+#define QIB_6120_RxEBPCnt_OFFS 0x120B8
+
+#define QIB_6120_RxLPCRCErrCnt_OFFS 0x120C0
+
+#define QIB_6120_RxBufOvflCnt_OFFS 0x120C8
+
+#define QIB_6120_RxTIDFullErrCnt_OFFS 0x120D0
+
+#define QIB_6120_RxTIDValidErrCnt_OFFS 0x120D8
+
+#define QIB_6120_RxPKeyMismatchCnt_OFFS 0x120E0
+
+#define QIB_6120_RxP0HdrEgrOvflCnt_OFFS 0x120E8
+
+#define QIB_6120_IBStatusChangeCnt_OFFS 0x12140
+
+#define QIB_6120_IBLinkErrRecoveryCnt_OFFS 0x12148
+
+#define QIB_6120_IBLinkDownedCnt_OFFS 0x12150
+
+#define QIB_6120_IBSymbolErrCnt_OFFS 0x12158
+
+#define QIB_6120_PcieRetryBufDiagQwordCnt_OFFS 0x12170
+
+#define QIB_6120_RcvEgrArray0_OFFS 0x14000
+
+#define QIB_6120_RcvTIDArray0_OFFS 0x54000
+
+#define QIB_6120_PIOLaunchFIFO_OFFS 0x64000
+
+#define QIB_6120_SendPIOpbcCache_OFFS 0x64800
+
+#define QIB_6120_RcvBuf1_OFFS 0x72000
+
+#define QIB_6120_RcvBuf2_OFFS 0x75000
+
+#define QIB_6120_RcvFlags_OFFS 0x77000
+
+#define QIB_6120_RcvLookupBuf1_OFFS 0x79000
+
+#define QIB_6120_RcvDMABuf_OFFS 0x7B000
+
+#define QIB_6120_MiscRXEIntMem_OFFS 0x7C000
+
+#define QIB_6120_PCIERcvBuf_OFFS 0x80000
+
+#define QIB_6120_PCIERetryBuf_OFFS 0x82000
+
+#define QIB_6120_PCIERcvBufRdToWrAddr_OFFS 0x84000
+
+#define QIB_6120_PIOBuf0_MA_OFFS 0x100000
diff --git a/drivers/infiniband/hw/qib/qib_7220.h b/drivers/infiniband/hw/qib/qib_7220.h
new file mode 100644
index 000000000000..ea0bfd896f92
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_7220.h
@@ -0,0 +1,156 @@
+#ifndef _QIB_7220_H
+#define _QIB_7220_H
+/*
+ * Copyright (c) 2007, 2009, 2010 QLogic Corporation. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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.
+ */
+
+/* grab register-defs auto-generated by HW */
+#include "qib_7220_regs.h"
+
+/* The number of eager receive TIDs for context zero. */
+#define IBA7220_KRCVEGRCNT      2048U
+
+#define IB_7220_LT_STATE_CFGRCVFCFG      0x09
+#define IB_7220_LT_STATE_CFGWAITRMT      0x0a
+#define IB_7220_LT_STATE_TXREVLANES      0x0d
+#define IB_7220_LT_STATE_CFGENH          0x10
+
+struct qib_chip_specific {
+	u64 __iomem *cregbase;
+	u64 *cntrs;
+	u64 *portcntrs;
+	spinlock_t sdepb_lock; /* serdes EPB bus */
+	spinlock_t rcvmod_lock; /* protect rcvctrl shadow changes */
+	spinlock_t gpio_lock; /* RMW of shadows/regs for ExtCtrl and GPIO */
+	u64 hwerrmask;
+	u64 errormask;
+	u64 gpio_out; /* shadow of kr_gpio_out, for rmw ops */
+	u64 gpio_mask; /* shadow the gpio mask register */
+	u64 extctrl; /* shadow the gpio output enable, etc... */
+	u32 ncntrs;
+	u32 nportcntrs;
+	u32 cntrnamelen;
+	u32 portcntrnamelen;
+	u32 numctxts;
+	u32 rcvegrcnt;
+	u32 autoneg_tries;
+	u32 serdes_first_init_done;
+	u32 sdmabufcnt;
+	u32 lastbuf_for_pio;
+	u32 updthresh; /* current AvailUpdThld */
+	u32 updthresh_dflt; /* default AvailUpdThld */
+	int irq;
+	u8 presets_needed;
+	u8 relock_timer_active;
+	char emsgbuf[128];
+	char sdmamsgbuf[192];
+	char bitsmsgbuf[64];
+	struct timer_list relock_timer;
+	unsigned int relock_interval; /* in jiffies */
+};
+
+struct qib_chippport_specific {
+	struct qib_pportdata pportdata;
+	wait_queue_head_t autoneg_wait;
+	struct delayed_work autoneg_work;
+	struct timer_list chase_timer;
+	/*
+	 * these 5 fields are used to establish deltas for IB symbol
+	 * errors and linkrecovery errors.  They can be reported on
+	 * some chips during link negotiation prior to INIT, and with
+	 * DDR when faking DDR negotiations with non-IBTA switches.
+	 * The chip counters are adjusted at driver unload if there is
+	 * a non-zero delta.
+	 */
+	u64 ibdeltainprog;
+	u64 ibsymdelta;
+	u64 ibsymsnap;
+	u64 iblnkerrdelta;
+	u64 iblnkerrsnap;
+	u64 ibcctrl; /* kr_ibcctrl shadow */
+	u64 ibcddrctrl; /* kr_ibcddrctrl shadow */
+	u64 chase_end;
+	u32 last_delay_mult;
+};
+
+/*
+ * This header file provides the declarations and common definitions
+ * for (mostly) manipulation of the SerDes blocks within the IBA7220.
+ * the functions declared should only be called from within other
+ * 7220-related files such as qib_iba7220.c or qib_sd7220.c.
+ */
+int qib_sd7220_presets(struct qib_devdata *dd);
+int qib_sd7220_init(struct qib_devdata *dd);
+int qib_sd7220_prog_ld(struct qib_devdata *dd, int sdnum, u8 *img,
+		       int len, int offset);
+int qib_sd7220_prog_vfy(struct qib_devdata *dd, int sdnum, const u8 *img,
+			int len, int offset);
+void qib_sd7220_clr_ibpar(struct qib_devdata *);
+/*
+ * Below used for sdnum parameter, selecting one of the two sections
+ * used for PCIe, or the single SerDes used for IB, which is the
+ * only one currently used
+ */
+#define IB_7220_SERDES 2
+
+int qib_sd7220_ib_load(struct qib_devdata *dd);
+int qib_sd7220_ib_vfy(struct qib_devdata *dd);
+
+static inline u32 qib_read_kreg32(const struct qib_devdata *dd,
+				  const u16 regno)
+{
+	if (!dd->kregbase || !(dd->flags & QIB_PRESENT))
+		return -1;
+	return readl((u32 __iomem *)&dd->kregbase[regno]);
+}
+
+static inline u64 qib_read_kreg64(const struct qib_devdata *dd,
+				  const u16 regno)
+{
+	if (!dd->kregbase || !(dd->flags & QIB_PRESENT))
+		return -1;
+
+	return readq(&dd->kregbase[regno]);
+}
+
+static inline void qib_write_kreg(const struct qib_devdata *dd,
+				  const u16 regno, u64 value)
+{
+	if (dd->kregbase)
+		writeq(value, &dd->kregbase[regno]);
+}
+
+void set_7220_relock_poll(struct qib_devdata *, int);
+void shutdown_7220_relock_poll(struct qib_devdata *);
+void toggle_7220_rclkrls(struct qib_devdata *);
+
+
+#endif /* _QIB_7220_H */
diff --git a/drivers/infiniband/hw/qib/qib_7220_regs.h b/drivers/infiniband/hw/qib/qib_7220_regs.h
new file mode 100644
index 000000000000..0da5bb750e52
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_7220_regs.h
@@ -0,0 +1,1496 @@
+/*
+ * Copyright (c) 2008, 2009, 2010 QLogic Corporation. All rights reserved.
+ *
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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.
+ *
+ */
+
+/* This file is mechanically generated from RTL. Any hand-edits will be lost! */
+
+#define QIB_7220_Revision_OFFS 0x0
+#define QIB_7220_Revision_R_Simulator_LSB 0x3F
+#define QIB_7220_Revision_R_Simulator_RMASK 0x1
+#define QIB_7220_Revision_R_Emulation_LSB 0x3E
+#define QIB_7220_Revision_R_Emulation_RMASK 0x1
+#define QIB_7220_Revision_R_Emulation_Revcode_LSB 0x28
+#define QIB_7220_Revision_R_Emulation_Revcode_RMASK 0x3FFFFF
+#define QIB_7220_Revision_BoardID_LSB 0x20
+#define QIB_7220_Revision_BoardID_RMASK 0xFF
+#define QIB_7220_Revision_R_SW_LSB 0x18
+#define QIB_7220_Revision_R_SW_RMASK 0xFF
+#define QIB_7220_Revision_R_Arch_LSB 0x10
+#define QIB_7220_Revision_R_Arch_RMASK 0xFF
+#define QIB_7220_Revision_R_ChipRevMajor_LSB 0x8
+#define QIB_7220_Revision_R_ChipRevMajor_RMASK 0xFF
+#define QIB_7220_Revision_R_ChipRevMinor_LSB 0x0
+#define QIB_7220_Revision_R_ChipRevMinor_RMASK 0xFF
+
+#define QIB_7220_Control_OFFS 0x8
+#define QIB_7220_Control_SyncResetExceptPcieIRAMRST_LSB 0x7
+#define QIB_7220_Control_SyncResetExceptPcieIRAMRST_RMASK 0x1
+#define QIB_7220_Control_PCIECplQDiagEn_LSB 0x6
+#define QIB_7220_Control_PCIECplQDiagEn_RMASK 0x1
+#define QIB_7220_Control_Reserved_LSB 0x5
+#define QIB_7220_Control_Reserved_RMASK 0x1
+#define QIB_7220_Control_TxLatency_LSB 0x4
+#define QIB_7220_Control_TxLatency_RMASK 0x1
+#define QIB_7220_Control_PCIERetryBufDiagEn_LSB 0x3
+#define QIB_7220_Control_PCIERetryBufDiagEn_RMASK 0x1
+#define QIB_7220_Control_LinkEn_LSB 0x2
+#define QIB_7220_Control_LinkEn_RMASK 0x1
+#define QIB_7220_Control_FreezeMode_LSB 0x1
+#define QIB_7220_Control_FreezeMode_RMASK 0x1
+#define QIB_7220_Control_SyncReset_LSB 0x0
+#define QIB_7220_Control_SyncReset_RMASK 0x1
+
+#define QIB_7220_PageAlign_OFFS 0x10
+
+#define QIB_7220_PortCnt_OFFS 0x18
+
+#define QIB_7220_SendRegBase_OFFS 0x30
+
+#define QIB_7220_UserRegBase_OFFS 0x38
+
+#define QIB_7220_CntrRegBase_OFFS 0x40
+
+#define QIB_7220_Scratch_OFFS 0x48
+
+#define QIB_7220_IntMask_OFFS 0x68
+#define QIB_7220_IntMask_SDmaIntMask_LSB 0x3F
+#define QIB_7220_IntMask_SDmaIntMask_RMASK 0x1
+#define QIB_7220_IntMask_SDmaDisabledMasked_LSB 0x3E
+#define QIB_7220_IntMask_SDmaDisabledMasked_RMASK 0x1
+#define QIB_7220_IntMask_Reserved_LSB 0x31
+#define QIB_7220_IntMask_Reserved_RMASK 0x1FFF
+#define QIB_7220_IntMask_RcvUrg16IntMask_LSB 0x30
+#define QIB_7220_IntMask_RcvUrg16IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvUrg15IntMask_LSB 0x2F
+#define QIB_7220_IntMask_RcvUrg15IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvUrg14IntMask_LSB 0x2E
+#define QIB_7220_IntMask_RcvUrg14IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvUrg13IntMask_LSB 0x2D
+#define QIB_7220_IntMask_RcvUrg13IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvUrg12IntMask_LSB 0x2C
+#define QIB_7220_IntMask_RcvUrg12IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvUrg11IntMask_LSB 0x2B
+#define QIB_7220_IntMask_RcvUrg11IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvUrg10IntMask_LSB 0x2A
+#define QIB_7220_IntMask_RcvUrg10IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvUrg9IntMask_LSB 0x29
+#define QIB_7220_IntMask_RcvUrg9IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvUrg8IntMask_LSB 0x28
+#define QIB_7220_IntMask_RcvUrg8IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvUrg7IntMask_LSB 0x27
+#define QIB_7220_IntMask_RcvUrg7IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvUrg6IntMask_LSB 0x26
+#define QIB_7220_IntMask_RcvUrg6IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvUrg5IntMask_LSB 0x25
+#define QIB_7220_IntMask_RcvUrg5IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvUrg4IntMask_LSB 0x24
+#define QIB_7220_IntMask_RcvUrg4IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvUrg3IntMask_LSB 0x23
+#define QIB_7220_IntMask_RcvUrg3IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvUrg2IntMask_LSB 0x22
+#define QIB_7220_IntMask_RcvUrg2IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvUrg1IntMask_LSB 0x21
+#define QIB_7220_IntMask_RcvUrg1IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvUrg0IntMask_LSB 0x20
+#define QIB_7220_IntMask_RcvUrg0IntMask_RMASK 0x1
+#define QIB_7220_IntMask_ErrorIntMask_LSB 0x1F
+#define QIB_7220_IntMask_ErrorIntMask_RMASK 0x1
+#define QIB_7220_IntMask_PioSetIntMask_LSB 0x1E
+#define QIB_7220_IntMask_PioSetIntMask_RMASK 0x1
+#define QIB_7220_IntMask_PioBufAvailIntMask_LSB 0x1D
+#define QIB_7220_IntMask_PioBufAvailIntMask_RMASK 0x1
+#define QIB_7220_IntMask_assertGPIOIntMask_LSB 0x1C
+#define QIB_7220_IntMask_assertGPIOIntMask_RMASK 0x1
+#define QIB_7220_IntMask_IBSerdesTrimDoneIntMask_LSB 0x1B
+#define QIB_7220_IntMask_IBSerdesTrimDoneIntMask_RMASK 0x1
+#define QIB_7220_IntMask_JIntMask_LSB 0x1A
+#define QIB_7220_IntMask_JIntMask_RMASK 0x1
+#define QIB_7220_IntMask_Reserved1_LSB 0x11
+#define QIB_7220_IntMask_Reserved1_RMASK 0x1FF
+#define QIB_7220_IntMask_RcvAvail16IntMask_LSB 0x10
+#define QIB_7220_IntMask_RcvAvail16IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvAvail15IntMask_LSB 0xF
+#define QIB_7220_IntMask_RcvAvail15IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvAvail14IntMask_LSB 0xE
+#define QIB_7220_IntMask_RcvAvail14IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvAvail13IntMask_LSB 0xD
+#define QIB_7220_IntMask_RcvAvail13IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvAvail12IntMask_LSB 0xC
+#define QIB_7220_IntMask_RcvAvail12IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvAvail11IntMask_LSB 0xB
+#define QIB_7220_IntMask_RcvAvail11IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvAvail10IntMask_LSB 0xA
+#define QIB_7220_IntMask_RcvAvail10IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvAvail9IntMask_LSB 0x9
+#define QIB_7220_IntMask_RcvAvail9IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvAvail8IntMask_LSB 0x8
+#define QIB_7220_IntMask_RcvAvail8IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvAvail7IntMask_LSB 0x7
+#define QIB_7220_IntMask_RcvAvail7IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvAvail6IntMask_LSB 0x6
+#define QIB_7220_IntMask_RcvAvail6IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvAvail5IntMask_LSB 0x5
+#define QIB_7220_IntMask_RcvAvail5IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvAvail4IntMask_LSB 0x4
+#define QIB_7220_IntMask_RcvAvail4IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvAvail3IntMask_LSB 0x3
+#define QIB_7220_IntMask_RcvAvail3IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvAvail2IntMask_LSB 0x2
+#define QIB_7220_IntMask_RcvAvail2IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvAvail1IntMask_LSB 0x1
+#define QIB_7220_IntMask_RcvAvail1IntMask_RMASK 0x1
+#define QIB_7220_IntMask_RcvAvail0IntMask_LSB 0x0
+#define QIB_7220_IntMask_RcvAvail0IntMask_RMASK 0x1
+
+#define QIB_7220_IntStatus_OFFS 0x70
+#define QIB_7220_IntStatus_SDmaInt_LSB 0x3F
+#define QIB_7220_IntStatus_SDmaInt_RMASK 0x1
+#define QIB_7220_IntStatus_SDmaDisabled_LSB 0x3E
+#define QIB_7220_IntStatus_SDmaDisabled_RMASK 0x1
+#define QIB_7220_IntStatus_Reserved_LSB 0x31
+#define QIB_7220_IntStatus_Reserved_RMASK 0x1FFF
+#define QIB_7220_IntStatus_RcvUrg16_LSB 0x30
+#define QIB_7220_IntStatus_RcvUrg16_RMASK 0x1
+#define QIB_7220_IntStatus_RcvUrg15_LSB 0x2F
+#define QIB_7220_IntStatus_RcvUrg15_RMASK 0x1
+#define QIB_7220_IntStatus_RcvUrg14_LSB 0x2E
+#define QIB_7220_IntStatus_RcvUrg14_RMASK 0x1
+#define QIB_7220_IntStatus_RcvUrg13_LSB 0x2D
+#define QIB_7220_IntStatus_RcvUrg13_RMASK 0x1
+#define QIB_7220_IntStatus_RcvUrg12_LSB 0x2C
+#define QIB_7220_IntStatus_RcvUrg12_RMASK 0x1
+#define QIB_7220_IntStatus_RcvUrg11_LSB 0x2B
+#define QIB_7220_IntStatus_RcvUrg11_RMASK 0x1
+#define QIB_7220_IntStatus_RcvUrg10_LSB 0x2A
+#define QIB_7220_IntStatus_RcvUrg10_RMASK 0x1
+#define QIB_7220_IntStatus_RcvUrg9_LSB 0x29
+#define QIB_7220_IntStatus_RcvUrg9_RMASK 0x1
+#define QIB_7220_IntStatus_RcvUrg8_LSB 0x28
+#define QIB_7220_IntStatus_RcvUrg8_RMASK 0x1
+#define QIB_7220_IntStatus_RcvUrg7_LSB 0x27
+#define QIB_7220_IntStatus_RcvUrg7_RMASK 0x1
+#define QIB_7220_IntStatus_RcvUrg6_LSB 0x26
+#define QIB_7220_IntStatus_RcvUrg6_RMASK 0x1
+#define QIB_7220_IntStatus_RcvUrg5_LSB 0x25
+#define QIB_7220_IntStatus_RcvUrg5_RMASK 0x1
+#define QIB_7220_IntStatus_RcvUrg4_LSB 0x24
+#define QIB_7220_IntStatus_RcvUrg4_RMASK 0x1
+#define QIB_7220_IntStatus_RcvUrg3_LSB 0x23
+#define QIB_7220_IntStatus_RcvUrg3_RMASK 0x1
+#define QIB_7220_IntStatus_RcvUrg2_LSB 0x22
+#define QIB_7220_IntStatus_RcvUrg2_RMASK 0x1
+#define QIB_7220_IntStatus_RcvUrg1_LSB 0x21
+#define QIB_7220_IntStatus_RcvUrg1_RMASK 0x1
+#define QIB_7220_IntStatus_RcvUrg0_LSB 0x20
+#define QIB_7220_IntStatus_RcvUrg0_RMASK 0x1
+#define QIB_7220_IntStatus_Error_LSB 0x1F
+#define QIB_7220_IntStatus_Error_RMASK 0x1
+#define QIB_7220_IntStatus_PioSent_LSB 0x1E
+#define QIB_7220_IntStatus_PioSent_RMASK 0x1
+#define QIB_7220_IntStatus_PioBufAvail_LSB 0x1D
+#define QIB_7220_IntStatus_PioBufAvail_RMASK 0x1
+#define QIB_7220_IntStatus_assertGPIO_LSB 0x1C
+#define QIB_7220_IntStatus_assertGPIO_RMASK 0x1
+#define QIB_7220_IntStatus_IBSerdesTrimDone_LSB 0x1B
+#define QIB_7220_IntStatus_IBSerdesTrimDone_RMASK 0x1
+#define QIB_7220_IntStatus_JInt_LSB 0x1A
+#define QIB_7220_IntStatus_JInt_RMASK 0x1
+#define QIB_7220_IntStatus_Reserved1_LSB 0x11
+#define QIB_7220_IntStatus_Reserved1_RMASK 0x1FF
+#define QIB_7220_IntStatus_RcvAvail16_LSB 0x10
+#define QIB_7220_IntStatus_RcvAvail16_RMASK 0x1
+#define QIB_7220_IntStatus_RcvAvail15_LSB 0xF
+#define QIB_7220_IntStatus_RcvAvail15_RMASK 0x1
+#define QIB_7220_IntStatus_RcvAvail14_LSB 0xE
+#define QIB_7220_IntStatus_RcvAvail14_RMASK 0x1
+#define QIB_7220_IntStatus_RcvAvail13_LSB 0xD
+#define QIB_7220_IntStatus_RcvAvail13_RMASK 0x1
+#define QIB_7220_IntStatus_RcvAvail12_LSB 0xC
+#define QIB_7220_IntStatus_RcvAvail12_RMASK 0x1
+#define QIB_7220_IntStatus_RcvAvail11_LSB 0xB
+#define QIB_7220_IntStatus_RcvAvail11_RMASK 0x1
+#define QIB_7220_IntStatus_RcvAvail10_LSB 0xA
+#define QIB_7220_IntStatus_RcvAvail10_RMASK 0x1
+#define QIB_7220_IntStatus_RcvAvail9_LSB 0x9
+#define QIB_7220_IntStatus_RcvAvail9_RMASK 0x1
+#define QIB_7220_IntStatus_RcvAvail8_LSB 0x8
+#define QIB_7220_IntStatus_RcvAvail8_RMASK 0x1
+#define QIB_7220_IntStatus_RcvAvail7_LSB 0x7
+#define QIB_7220_IntStatus_RcvAvail7_RMASK 0x1
+#define QIB_7220_IntStatus_RcvAvail6_LSB 0x6
+#define QIB_7220_IntStatus_RcvAvail6_RMASK 0x1
+#define QIB_7220_IntStatus_RcvAvail5_LSB 0x5
+#define QIB_7220_IntStatus_RcvAvail5_RMASK 0x1
+#define QIB_7220_IntStatus_RcvAvail4_LSB 0x4
+#define QIB_7220_IntStatus_RcvAvail4_RMASK 0x1
+#define QIB_7220_IntStatus_RcvAvail3_LSB 0x3
+#define QIB_7220_IntStatus_RcvAvail3_RMASK 0x1
+#define QIB_7220_IntStatus_RcvAvail2_LSB 0x2
+#define QIB_7220_IntStatus_RcvAvail2_RMASK 0x1
+#define QIB_7220_IntStatus_RcvAvail1_LSB 0x1
+#define QIB_7220_IntStatus_RcvAvail1_RMASK 0x1
+#define QIB_7220_IntStatus_RcvAvail0_LSB 0x0
+#define QIB_7220_IntStatus_RcvAvail0_RMASK 0x1
+
+#define QIB_7220_IntClear_OFFS 0x78
+#define QIB_7220_IntClear_SDmaIntClear_LSB 0x3F
+#define QIB_7220_IntClear_SDmaIntClear_RMASK 0x1
+#define QIB_7220_IntClear_SDmaDisabledClear_LSB 0x3E
+#define QIB_7220_IntClear_SDmaDisabledClear_RMASK 0x1
+#define QIB_7220_IntClear_Reserved_LSB 0x31
+#define QIB_7220_IntClear_Reserved_RMASK 0x1FFF
+#define QIB_7220_IntClear_RcvUrg16IntClear_LSB 0x30
+#define QIB_7220_IntClear_RcvUrg16IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvUrg15IntClear_LSB 0x2F
+#define QIB_7220_IntClear_RcvUrg15IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvUrg14IntClear_LSB 0x2E
+#define QIB_7220_IntClear_RcvUrg14IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvUrg13IntClear_LSB 0x2D
+#define QIB_7220_IntClear_RcvUrg13IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvUrg12IntClear_LSB 0x2C
+#define QIB_7220_IntClear_RcvUrg12IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvUrg11IntClear_LSB 0x2B
+#define QIB_7220_IntClear_RcvUrg11IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvUrg10IntClear_LSB 0x2A
+#define QIB_7220_IntClear_RcvUrg10IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvUrg9IntClear_LSB 0x29
+#define QIB_7220_IntClear_RcvUrg9IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvUrg8IntClear_LSB 0x28
+#define QIB_7220_IntClear_RcvUrg8IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvUrg7IntClear_LSB 0x27
+#define QIB_7220_IntClear_RcvUrg7IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvUrg6IntClear_LSB 0x26
+#define QIB_7220_IntClear_RcvUrg6IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvUrg5IntClear_LSB 0x25
+#define QIB_7220_IntClear_RcvUrg5IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvUrg4IntClear_LSB 0x24
+#define QIB_7220_IntClear_RcvUrg4IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvUrg3IntClear_LSB 0x23
+#define QIB_7220_IntClear_RcvUrg3IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvUrg2IntClear_LSB 0x22
+#define QIB_7220_IntClear_RcvUrg2IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvUrg1IntClear_LSB 0x21
+#define QIB_7220_IntClear_RcvUrg1IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvUrg0IntClear_LSB 0x20
+#define QIB_7220_IntClear_RcvUrg0IntClear_RMASK 0x1
+#define QIB_7220_IntClear_ErrorIntClear_LSB 0x1F
+#define QIB_7220_IntClear_ErrorIntClear_RMASK 0x1
+#define QIB_7220_IntClear_PioSetIntClear_LSB 0x1E
+#define QIB_7220_IntClear_PioSetIntClear_RMASK 0x1
+#define QIB_7220_IntClear_PioBufAvailIntClear_LSB 0x1D
+#define QIB_7220_IntClear_PioBufAvailIntClear_RMASK 0x1
+#define QIB_7220_IntClear_assertGPIOIntClear_LSB 0x1C
+#define QIB_7220_IntClear_assertGPIOIntClear_RMASK 0x1
+#define QIB_7220_IntClear_IBSerdesTrimDoneClear_LSB 0x1B
+#define QIB_7220_IntClear_IBSerdesTrimDoneClear_RMASK 0x1
+#define QIB_7220_IntClear_JIntClear_LSB 0x1A
+#define QIB_7220_IntClear_JIntClear_RMASK 0x1
+#define QIB_7220_IntClear_Reserved1_LSB 0x11
+#define QIB_7220_IntClear_Reserved1_RMASK 0x1FF
+#define QIB_7220_IntClear_RcvAvail16IntClear_LSB 0x10
+#define QIB_7220_IntClear_RcvAvail16IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvAvail15IntClear_LSB 0xF
+#define QIB_7220_IntClear_RcvAvail15IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvAvail14IntClear_LSB 0xE
+#define QIB_7220_IntClear_RcvAvail14IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvAvail13IntClear_LSB 0xD
+#define QIB_7220_IntClear_RcvAvail13IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvAvail12IntClear_LSB 0xC
+#define QIB_7220_IntClear_RcvAvail12IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvAvail11IntClear_LSB 0xB
+#define QIB_7220_IntClear_RcvAvail11IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvAvail10IntClear_LSB 0xA
+#define QIB_7220_IntClear_RcvAvail10IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvAvail9IntClear_LSB 0x9
+#define QIB_7220_IntClear_RcvAvail9IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvAvail8IntClear_LSB 0x8
+#define QIB_7220_IntClear_RcvAvail8IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvAvail7IntClear_LSB 0x7
+#define QIB_7220_IntClear_RcvAvail7IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvAvail6IntClear_LSB 0x6
+#define QIB_7220_IntClear_RcvAvail6IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvAvail5IntClear_LSB 0x5
+#define QIB_7220_IntClear_RcvAvail5IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvAvail4IntClear_LSB 0x4
+#define QIB_7220_IntClear_RcvAvail4IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvAvail3IntClear_LSB 0x3
+#define QIB_7220_IntClear_RcvAvail3IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvAvail2IntClear_LSB 0x2
+#define QIB_7220_IntClear_RcvAvail2IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvAvail1IntClear_LSB 0x1
+#define QIB_7220_IntClear_RcvAvail1IntClear_RMASK 0x1
+#define QIB_7220_IntClear_RcvAvail0IntClear_LSB 0x0
+#define QIB_7220_IntClear_RcvAvail0IntClear_RMASK 0x1
+
+#define QIB_7220_ErrMask_OFFS 0x80
+#define QIB_7220_ErrMask_Reserved_LSB 0x36
+#define QIB_7220_ErrMask_Reserved_RMASK 0x3FF
+#define QIB_7220_ErrMask_InvalidEEPCmdMask_LSB 0x35
+#define QIB_7220_ErrMask_InvalidEEPCmdMask_RMASK 0x1
+#define QIB_7220_ErrMask_SDmaDescAddrMisalignErrMask_LSB 0x34
+#define QIB_7220_ErrMask_SDmaDescAddrMisalignErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_HardwareErrMask_LSB 0x33
+#define QIB_7220_ErrMask_HardwareErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_ResetNegatedMask_LSB 0x32
+#define QIB_7220_ErrMask_ResetNegatedMask_RMASK 0x1
+#define QIB_7220_ErrMask_InvalidAddrErrMask_LSB 0x31
+#define QIB_7220_ErrMask_InvalidAddrErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_IBStatusChangedMask_LSB 0x30
+#define QIB_7220_ErrMask_IBStatusChangedMask_RMASK 0x1
+#define QIB_7220_ErrMask_SDmaUnexpDataErrMask_LSB 0x2F
+#define QIB_7220_ErrMask_SDmaUnexpDataErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_SDmaMissingDwErrMask_LSB 0x2E
+#define QIB_7220_ErrMask_SDmaMissingDwErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_SDmaDwEnErrMask_LSB 0x2D
+#define QIB_7220_ErrMask_SDmaDwEnErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_SDmaRpyTagErrMask_LSB 0x2C
+#define QIB_7220_ErrMask_SDmaRpyTagErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_SDma1stDescErrMask_LSB 0x2B
+#define QIB_7220_ErrMask_SDma1stDescErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_SDmaBaseErrMask_LSB 0x2A
+#define QIB_7220_ErrMask_SDmaBaseErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_SDmaTailOutOfBoundErrMask_LSB 0x29
+#define QIB_7220_ErrMask_SDmaTailOutOfBoundErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_SDmaOutOfBoundErrMask_LSB 0x28
+#define QIB_7220_ErrMask_SDmaOutOfBoundErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_SDmaGenMismatchErrMask_LSB 0x27
+#define QIB_7220_ErrMask_SDmaGenMismatchErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_SendBufMisuseErrMask_LSB 0x26
+#define QIB_7220_ErrMask_SendBufMisuseErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_SendUnsupportedVLErrMask_LSB 0x25
+#define QIB_7220_ErrMask_SendUnsupportedVLErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_SendUnexpectedPktNumErrMask_LSB 0x24
+#define QIB_7220_ErrMask_SendUnexpectedPktNumErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_SendPioArmLaunchErrMask_LSB 0x23
+#define QIB_7220_ErrMask_SendPioArmLaunchErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_SendDroppedDataPktErrMask_LSB 0x22
+#define QIB_7220_ErrMask_SendDroppedDataPktErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_SendDroppedSmpPktErrMask_LSB 0x21
+#define QIB_7220_ErrMask_SendDroppedSmpPktErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_SendPktLenErrMask_LSB 0x20
+#define QIB_7220_ErrMask_SendPktLenErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_SendUnderRunErrMask_LSB 0x1F
+#define QIB_7220_ErrMask_SendUnderRunErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_SendMaxPktLenErrMask_LSB 0x1E
+#define QIB_7220_ErrMask_SendMaxPktLenErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_SendMinPktLenErrMask_LSB 0x1D
+#define QIB_7220_ErrMask_SendMinPktLenErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_SDmaDisabledErrMask_LSB 0x1C
+#define QIB_7220_ErrMask_SDmaDisabledErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_SendSpecialTriggerErrMask_LSB 0x1B
+#define QIB_7220_ErrMask_SendSpecialTriggerErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_Reserved1_LSB 0x12
+#define QIB_7220_ErrMask_Reserved1_RMASK 0x1FF
+#define QIB_7220_ErrMask_RcvIBLostLinkErrMask_LSB 0x11
+#define QIB_7220_ErrMask_RcvIBLostLinkErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_RcvHdrErrMask_LSB 0x10
+#define QIB_7220_ErrMask_RcvHdrErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_RcvHdrLenErrMask_LSB 0xF
+#define QIB_7220_ErrMask_RcvHdrLenErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_RcvBadTidErrMask_LSB 0xE
+#define QIB_7220_ErrMask_RcvBadTidErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_RcvHdrFullErrMask_LSB 0xD
+#define QIB_7220_ErrMask_RcvHdrFullErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_RcvEgrFullErrMask_LSB 0xC
+#define QIB_7220_ErrMask_RcvEgrFullErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_RcvBadVersionErrMask_LSB 0xB
+#define QIB_7220_ErrMask_RcvBadVersionErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_RcvIBFlowErrMask_LSB 0xA
+#define QIB_7220_ErrMask_RcvIBFlowErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_RcvEBPErrMask_LSB 0x9
+#define QIB_7220_ErrMask_RcvEBPErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_RcvUnsupportedVLErrMask_LSB 0x8
+#define QIB_7220_ErrMask_RcvUnsupportedVLErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_RcvUnexpectedCharErrMask_LSB 0x7
+#define QIB_7220_ErrMask_RcvUnexpectedCharErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_RcvShortPktLenErrMask_LSB 0x6
+#define QIB_7220_ErrMask_RcvShortPktLenErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_RcvLongPktLenErrMask_LSB 0x5
+#define QIB_7220_ErrMask_RcvLongPktLenErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_RcvMaxPktLenErrMask_LSB 0x4
+#define QIB_7220_ErrMask_RcvMaxPktLenErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_RcvMinPktLenErrMask_LSB 0x3
+#define QIB_7220_ErrMask_RcvMinPktLenErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_RcvICRCErrMask_LSB 0x2
+#define QIB_7220_ErrMask_RcvICRCErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_RcvVCRCErrMask_LSB 0x1
+#define QIB_7220_ErrMask_RcvVCRCErrMask_RMASK 0x1
+#define QIB_7220_ErrMask_RcvFormatErrMask_LSB 0x0
+#define QIB_7220_ErrMask_RcvFormatErrMask_RMASK 0x1
+
+#define QIB_7220_ErrStatus_OFFS 0x88
+#define QIB_7220_ErrStatus_Reserved_LSB 0x36
+#define QIB_7220_ErrStatus_Reserved_RMASK 0x3FF
+#define QIB_7220_ErrStatus_InvalidEEPCmdErr_LSB 0x35
+#define QIB_7220_ErrStatus_InvalidEEPCmdErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SDmaDescAddrMisalignErr_LSB 0x34
+#define QIB_7220_ErrStatus_SDmaDescAddrMisalignErr_RMASK 0x1
+#define QIB_7220_ErrStatus_HardwareErr_LSB 0x33
+#define QIB_7220_ErrStatus_HardwareErr_RMASK 0x1
+#define QIB_7220_ErrStatus_ResetNegated_LSB 0x32
+#define QIB_7220_ErrStatus_ResetNegated_RMASK 0x1
+#define QIB_7220_ErrStatus_InvalidAddrErr_LSB 0x31
+#define QIB_7220_ErrStatus_InvalidAddrErr_RMASK 0x1
+#define QIB_7220_ErrStatus_IBStatusChanged_LSB 0x30
+#define QIB_7220_ErrStatus_IBStatusChanged_RMASK 0x1
+#define QIB_7220_ErrStatus_SDmaUnexpDataErr_LSB 0x2F
+#define QIB_7220_ErrStatus_SDmaUnexpDataErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SDmaMissingDwErr_LSB 0x2E
+#define QIB_7220_ErrStatus_SDmaMissingDwErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SDmaDwEnErr_LSB 0x2D
+#define QIB_7220_ErrStatus_SDmaDwEnErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SDmaRpyTagErr_LSB 0x2C
+#define QIB_7220_ErrStatus_SDmaRpyTagErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SDma1stDescErr_LSB 0x2B
+#define QIB_7220_ErrStatus_SDma1stDescErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SDmaBaseErr_LSB 0x2A
+#define QIB_7220_ErrStatus_SDmaBaseErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SDmaTailOutOfBoundErr_LSB 0x29
+#define QIB_7220_ErrStatus_SDmaTailOutOfBoundErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SDmaOutOfBoundErr_LSB 0x28
+#define QIB_7220_ErrStatus_SDmaOutOfBoundErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SDmaGenMismatchErr_LSB 0x27
+#define QIB_7220_ErrStatus_SDmaGenMismatchErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SendBufMisuseErr_LSB 0x26
+#define QIB_7220_ErrStatus_SendBufMisuseErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SendUnsupportedVLErr_LSB 0x25
+#define QIB_7220_ErrStatus_SendUnsupportedVLErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SendUnexpectedPktNumErr_LSB 0x24
+#define QIB_7220_ErrStatus_SendUnexpectedPktNumErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SendPioArmLaunchErr_LSB 0x23
+#define QIB_7220_ErrStatus_SendPioArmLaunchErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SendDroppedDataPktErr_LSB 0x22
+#define QIB_7220_ErrStatus_SendDroppedDataPktErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SendDroppedSmpPktErr_LSB 0x21
+#define QIB_7220_ErrStatus_SendDroppedSmpPktErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SendPktLenErr_LSB 0x20
+#define QIB_7220_ErrStatus_SendPktLenErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SendUnderRunErr_LSB 0x1F
+#define QIB_7220_ErrStatus_SendUnderRunErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SendMaxPktLenErr_LSB 0x1E
+#define QIB_7220_ErrStatus_SendMaxPktLenErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SendMinPktLenErr_LSB 0x1D
+#define QIB_7220_ErrStatus_SendMinPktLenErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SDmaDisabledErr_LSB 0x1C
+#define QIB_7220_ErrStatus_SDmaDisabledErr_RMASK 0x1
+#define QIB_7220_ErrStatus_SendSpecialTriggerErr_LSB 0x1B
+#define QIB_7220_ErrStatus_SendSpecialTriggerErr_RMASK 0x1
+#define QIB_7220_ErrStatus_Reserved1_LSB 0x12
+#define QIB_7220_ErrStatus_Reserved1_RMASK 0x1FF
+#define QIB_7220_ErrStatus_RcvIBLostLinkErr_LSB 0x11
+#define QIB_7220_ErrStatus_RcvIBLostLinkErr_RMASK 0x1
+#define QIB_7220_ErrStatus_RcvHdrErr_LSB 0x10
+#define QIB_7220_ErrStatus_RcvHdrErr_RMASK 0x1
+#define QIB_7220_ErrStatus_RcvHdrLenErr_LSB 0xF
+#define QIB_7220_ErrStatus_RcvHdrLenErr_RMASK 0x1
+#define QIB_7220_ErrStatus_RcvBadTidErr_LSB 0xE
+#define QIB_7220_ErrStatus_RcvBadTidErr_RMASK 0x1
+#define QIB_7220_ErrStatus_RcvHdrFullErr_LSB 0xD
+#define QIB_7220_ErrStatus_RcvHdrFullErr_RMASK 0x1
+#define QIB_7220_ErrStatus_RcvEgrFullErr_LSB 0xC
+#define QIB_7220_ErrStatus_RcvEgrFullErr_RMASK 0x1
+#define QIB_7220_ErrStatus_RcvBadVersionErr_LSB 0xB
+#define QIB_7220_ErrStatus_RcvBadVersionErr_RMASK 0x1
+#define QIB_7220_ErrStatus_RcvIBFlowErr_LSB 0xA
+#define QIB_7220_ErrStatus_RcvIBFlowErr_RMASK 0x1
+#define QIB_7220_ErrStatus_RcvEBPErr_LSB 0x9
+#define QIB_7220_ErrStatus_RcvEBPErr_RMASK 0x1
+#define QIB_7220_ErrStatus_RcvUnsupportedVLErr_LSB 0x8
+#define QIB_7220_ErrStatus_RcvUnsupportedVLErr_RMASK 0x1
+#define QIB_7220_ErrStatus_RcvUnexpectedCharErr_LSB 0x7
+#define QIB_7220_ErrStatus_RcvUnexpectedCharErr_RMASK 0x1
+#define QIB_7220_ErrStatus_RcvShortPktLenErr_LSB 0x6
+#define QIB_7220_ErrStatus_RcvShortPktLenErr_RMASK 0x1
+#define QIB_7220_ErrStatus_RcvLongPktLenErr_LSB 0x5
+#define QIB_7220_ErrStatus_RcvLongPktLenErr_RMASK 0x1
+#define QIB_7220_ErrStatus_RcvMaxPktLenErr_LSB 0x4
+#define QIB_7220_ErrStatus_RcvMaxPktLenErr_RMASK 0x1
+#define QIB_7220_ErrStatus_RcvMinPktLenErr_LSB 0x3
+#define QIB_7220_ErrStatus_RcvMinPktLenErr_RMASK 0x1
+#define QIB_7220_ErrStatus_RcvICRCErr_LSB 0x2
+#define QIB_7220_ErrStatus_RcvICRCErr_RMASK 0x1
+#define QIB_7220_ErrStatus_RcvVCRCErr_LSB 0x1
+#define QIB_7220_ErrStatus_RcvVCRCErr_RMASK 0x1
+#define QIB_7220_ErrStatus_RcvFormatErr_LSB 0x0
+#define QIB_7220_ErrStatus_RcvFormatErr_RMASK 0x1
+
+#define QIB_7220_ErrClear_OFFS 0x90
+#define QIB_7220_ErrClear_Reserved_LSB 0x36
+#define QIB_7220_ErrClear_Reserved_RMASK 0x3FF
+#define QIB_7220_ErrClear_InvalidEEPCmdErrClear_LSB 0x35
+#define QIB_7220_ErrClear_InvalidEEPCmdErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SDmaDescAddrMisalignErrClear_LSB 0x34
+#define QIB_7220_ErrClear_SDmaDescAddrMisalignErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_HardwareErrClear_LSB 0x33
+#define QIB_7220_ErrClear_HardwareErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_ResetNegatedClear_LSB 0x32
+#define QIB_7220_ErrClear_ResetNegatedClear_RMASK 0x1
+#define QIB_7220_ErrClear_InvalidAddrErrClear_LSB 0x31
+#define QIB_7220_ErrClear_InvalidAddrErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_IBStatusChangedClear_LSB 0x30
+#define QIB_7220_ErrClear_IBStatusChangedClear_RMASK 0x1
+#define QIB_7220_ErrClear_SDmaUnexpDataErrClear_LSB 0x2F
+#define QIB_7220_ErrClear_SDmaUnexpDataErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SDmaMissingDwErrClear_LSB 0x2E
+#define QIB_7220_ErrClear_SDmaMissingDwErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SDmaDwEnErrClear_LSB 0x2D
+#define QIB_7220_ErrClear_SDmaDwEnErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SDmaRpyTagErrClear_LSB 0x2C
+#define QIB_7220_ErrClear_SDmaRpyTagErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SDma1stDescErrClear_LSB 0x2B
+#define QIB_7220_ErrClear_SDma1stDescErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SDmaBaseErrClear_LSB 0x2A
+#define QIB_7220_ErrClear_SDmaBaseErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SDmaTailOutOfBoundErrClear_LSB 0x29
+#define QIB_7220_ErrClear_SDmaTailOutOfBoundErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SDmaOutOfBoundErrClear_LSB 0x28
+#define QIB_7220_ErrClear_SDmaOutOfBoundErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SDmaGenMismatchErrClear_LSB 0x27
+#define QIB_7220_ErrClear_SDmaGenMismatchErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SendBufMisuseErrClear_LSB 0x26
+#define QIB_7220_ErrClear_SendBufMisuseErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SendUnsupportedVLErrClear_LSB 0x25
+#define QIB_7220_ErrClear_SendUnsupportedVLErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SendUnexpectedPktNumErrClear_LSB 0x24
+#define QIB_7220_ErrClear_SendUnexpectedPktNumErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SendPioArmLaunchErrClear_LSB 0x23
+#define QIB_7220_ErrClear_SendPioArmLaunchErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SendDroppedDataPktErrClear_LSB 0x22
+#define QIB_7220_ErrClear_SendDroppedDataPktErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SendDroppedSmpPktErrClear_LSB 0x21
+#define QIB_7220_ErrClear_SendDroppedSmpPktErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SendPktLenErrClear_LSB 0x20
+#define QIB_7220_ErrClear_SendPktLenErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SendUnderRunErrClear_LSB 0x1F
+#define QIB_7220_ErrClear_SendUnderRunErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SendMaxPktLenErrClear_LSB 0x1E
+#define QIB_7220_ErrClear_SendMaxPktLenErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SendMinPktLenErrClear_LSB 0x1D
+#define QIB_7220_ErrClear_SendMinPktLenErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SDmaDisabledErrClear_LSB 0x1C
+#define QIB_7220_ErrClear_SDmaDisabledErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_SendSpecialTriggerErrClear_LSB 0x1B
+#define QIB_7220_ErrClear_SendSpecialTriggerErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_Reserved1_LSB 0x12
+#define QIB_7220_ErrClear_Reserved1_RMASK 0x1FF
+#define QIB_7220_ErrClear_RcvIBLostLinkErrClear_LSB 0x11
+#define QIB_7220_ErrClear_RcvIBLostLinkErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_RcvHdrErrClear_LSB 0x10
+#define QIB_7220_ErrClear_RcvHdrErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_RcvHdrLenErrClear_LSB 0xF
+#define QIB_7220_ErrClear_RcvHdrLenErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_RcvBadTidErrClear_LSB 0xE
+#define QIB_7220_ErrClear_RcvBadTidErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_RcvHdrFullErrClear_LSB 0xD
+#define QIB_7220_ErrClear_RcvHdrFullErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_RcvEgrFullErrClear_LSB 0xC
+#define QIB_7220_ErrClear_RcvEgrFullErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_RcvBadVersionErrClear_LSB 0xB
+#define QIB_7220_ErrClear_RcvBadVersionErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_RcvIBFlowErrClear_LSB 0xA
+#define QIB_7220_ErrClear_RcvIBFlowErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_RcvEBPErrClear_LSB 0x9
+#define QIB_7220_ErrClear_RcvEBPErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_RcvUnsupportedVLErrClear_LSB 0x8
+#define QIB_7220_ErrClear_RcvUnsupportedVLErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_RcvUnexpectedCharErrClear_LSB 0x7
+#define QIB_7220_ErrClear_RcvUnexpectedCharErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_RcvShortPktLenErrClear_LSB 0x6
+#define QIB_7220_ErrClear_RcvShortPktLenErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_RcvLongPktLenErrClear_LSB 0x5
+#define QIB_7220_ErrClear_RcvLongPktLenErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_RcvMaxPktLenErrClear_LSB 0x4
+#define QIB_7220_ErrClear_RcvMaxPktLenErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_RcvMinPktLenErrClear_LSB 0x3
+#define QIB_7220_ErrClear_RcvMinPktLenErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_RcvICRCErrClear_LSB 0x2
+#define QIB_7220_ErrClear_RcvICRCErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_RcvVCRCErrClear_LSB 0x1
+#define QIB_7220_ErrClear_RcvVCRCErrClear_RMASK 0x1
+#define QIB_7220_ErrClear_RcvFormatErrClear_LSB 0x0
+#define QIB_7220_ErrClear_RcvFormatErrClear_RMASK 0x1
+
+#define QIB_7220_HwErrMask_OFFS 0x98
+#define QIB_7220_HwErrMask_IBCBusFromSPCParityErrMask_LSB 0x3F
+#define QIB_7220_HwErrMask_IBCBusFromSPCParityErrMask_RMASK 0x1
+#define QIB_7220_HwErrMask_IBCBusToSPCParityErrMask_LSB 0x3E
+#define QIB_7220_HwErrMask_IBCBusToSPCParityErrMask_RMASK 0x1
+#define QIB_7220_HwErrMask_Clk_uC_PLLNotLockedMask_LSB 0x3D
+#define QIB_7220_HwErrMask_Clk_uC_PLLNotLockedMask_RMASK 0x1
+#define QIB_7220_HwErrMask_IBSerdesPClkNotDetectMask_LSB 0x3C
+#define QIB_7220_HwErrMask_IBSerdesPClkNotDetectMask_RMASK 0x1
+#define QIB_7220_HwErrMask_PCIESerdesQ3PClkNotDetectMask_LSB 0x3B
+#define QIB_7220_HwErrMask_PCIESerdesQ3PClkNotDetectMask_RMASK 0x1
+#define QIB_7220_HwErrMask_PCIESerdesQ2PClkNotDetectMask_LSB 0x3A
+#define QIB_7220_HwErrMask_PCIESerdesQ2PClkNotDetectMask_RMASK 0x1
+#define QIB_7220_HwErrMask_PCIESerdesQ1PClkNotDetectMask_LSB 0x39
+#define QIB_7220_HwErrMask_PCIESerdesQ1PClkNotDetectMask_RMASK 0x1
+#define QIB_7220_HwErrMask_PCIESerdesQ0PClkNotDetectMask_LSB 0x38
+#define QIB_7220_HwErrMask_PCIESerdesQ0PClkNotDetectMask_RMASK 0x1
+#define QIB_7220_HwErrMask_Reserved_LSB 0x37
+#define QIB_7220_HwErrMask_Reserved_RMASK 0x1
+#define QIB_7220_HwErrMask_PowerOnBISTFailedMask_LSB 0x36
+#define QIB_7220_HwErrMask_PowerOnBISTFailedMask_RMASK 0x1
+#define QIB_7220_HwErrMask_Reserved1_LSB 0x33
+#define QIB_7220_HwErrMask_Reserved1_RMASK 0x7
+#define QIB_7220_HwErrMask_RXEMemParityErrMask_LSB 0x2C
+#define QIB_7220_HwErrMask_RXEMemParityErrMask_RMASK 0x7F
+#define QIB_7220_HwErrMask_TXEMemParityErrMask_LSB 0x28
+#define QIB_7220_HwErrMask_TXEMemParityErrMask_RMASK 0xF
+#define QIB_7220_HwErrMask_DDSRXEQMemoryParityErrMask_LSB 0x27
+#define QIB_7220_HwErrMask_DDSRXEQMemoryParityErrMask_RMASK 0x1
+#define QIB_7220_HwErrMask_IB_uC_MemoryParityErrMask_LSB 0x26
+#define QIB_7220_HwErrMask_IB_uC_MemoryParityErrMask_RMASK 0x1
+#define QIB_7220_HwErrMask_PCIEOct1_uC_MemoryParityErrMask_LSB 0x25
+#define QIB_7220_HwErrMask_PCIEOct1_uC_MemoryParityErrMask_RMASK 0x1
+#define QIB_7220_HwErrMask_PCIEOct0_uC_MemoryParityErrMask_LSB 0x24
+#define QIB_7220_HwErrMask_PCIEOct0_uC_MemoryParityErrMask_RMASK 0x1
+#define QIB_7220_HwErrMask_Reserved2_LSB 0x22
+#define QIB_7220_HwErrMask_Reserved2_RMASK 0x3
+#define QIB_7220_HwErrMask_PCIeBusParityErrMask_LSB 0x1F
+#define QIB_7220_HwErrMask_PCIeBusParityErrMask_RMASK 0x7
+#define QIB_7220_HwErrMask_PcieCplTimeoutMask_LSB 0x1E
+#define QIB_7220_HwErrMask_PcieCplTimeoutMask_RMASK 0x1
+#define QIB_7220_HwErrMask_PoisonedTLPMask_LSB 0x1D
+#define QIB_7220_HwErrMask_PoisonedTLPMask_RMASK 0x1
+#define QIB_7220_HwErrMask_SDmaMemReadErrMask_LSB 0x1C
+#define QIB_7220_HwErrMask_SDmaMemReadErrMask_RMASK 0x1
+#define QIB_7220_HwErrMask_Reserved3_LSB 0x8
+#define QIB_7220_HwErrMask_Reserved3_RMASK 0xFFFFF
+#define QIB_7220_HwErrMask_PCIeMemParityErrMask_LSB 0x0
+#define QIB_7220_HwErrMask_PCIeMemParityErrMask_RMASK 0xFF
+
+#define QIB_7220_HwErrStatus_OFFS 0xA0
+#define QIB_7220_HwErrStatus_IBCBusFromSPCParityErr_LSB 0x3F
+#define QIB_7220_HwErrStatus_IBCBusFromSPCParityErr_RMASK 0x1
+#define QIB_7220_HwErrStatus_IBCBusToSPCParityErr_LSB 0x3E
+#define QIB_7220_HwErrStatus_IBCBusToSPCParityErr_RMASK 0x1
+#define QIB_7220_HwErrStatus_Clk_uC_PLLNotLocked_LSB 0x3D
+#define QIB_7220_HwErrStatus_Clk_uC_PLLNotLocked_RMASK 0x1
+#define QIB_7220_HwErrStatus_IBSerdesPClkNotDetect_LSB 0x3C
+#define QIB_7220_HwErrStatus_IBSerdesPClkNotDetect_RMASK 0x1
+#define QIB_7220_HwErrStatus_PCIESerdesQ3PClkNotDetect_LSB 0x3B
+#define QIB_7220_HwErrStatus_PCIESerdesQ3PClkNotDetect_RMASK 0x1
+#define QIB_7220_HwErrStatus_PCIESerdesQ2PClkNotDetect_LSB 0x3A
+#define QIB_7220_HwErrStatus_PCIESerdesQ2PClkNotDetect_RMASK 0x1
+#define QIB_7220_HwErrStatus_PCIESerdesQ1PClkNotDetect_LSB 0x39
+#define QIB_7220_HwErrStatus_PCIESerdesQ1PClkNotDetect_RMASK 0x1
+#define QIB_7220_HwErrStatus_PCIESerdesQ0PClkNotDetect_LSB 0x38
+#define QIB_7220_HwErrStatus_PCIESerdesQ0PClkNotDetect_RMASK 0x1
+#define QIB_7220_HwErrStatus_Reserved_LSB 0x37
+#define QIB_7220_HwErrStatus_Reserved_RMASK 0x1
+#define QIB_7220_HwErrStatus_PowerOnBISTFailed_LSB 0x36
+#define QIB_7220_HwErrStatus_PowerOnBISTFailed_RMASK 0x1
+#define QIB_7220_HwErrStatus_Reserved1_LSB 0x33
+#define QIB_7220_HwErrStatus_Reserved1_RMASK 0x7
+#define QIB_7220_HwErrStatus_RXEMemParity_LSB 0x2C
+#define QIB_7220_HwErrStatus_RXEMemParity_RMASK 0x7F
+#define QIB_7220_HwErrStatus_TXEMemParity_LSB 0x28
+#define QIB_7220_HwErrStatus_TXEMemParity_RMASK 0xF
+#define QIB_7220_HwErrStatus_DDSRXEQMemoryParityErr_LSB 0x27
+#define QIB_7220_HwErrStatus_DDSRXEQMemoryParityErr_RMASK 0x1
+#define QIB_7220_HwErrStatus_IB_uC_MemoryParityErr_LSB 0x26
+#define QIB_7220_HwErrStatus_IB_uC_MemoryParityErr_RMASK 0x1
+#define QIB_7220_HwErrStatus_PCIE_uC_Oct1MemoryParityErr_LSB 0x25
+#define QIB_7220_HwErrStatus_PCIE_uC_Oct1MemoryParityErr_RMASK 0x1
+#define QIB_7220_HwErrStatus_PCIE_uC_Oct0MemoryParityErr_LSB 0x24
+#define QIB_7220_HwErrStatus_PCIE_uC_Oct0MemoryParityErr_RMASK 0x1
+#define QIB_7220_HwErrStatus_Reserved2_LSB 0x22
+#define QIB_7220_HwErrStatus_Reserved2_RMASK 0x3
+#define QIB_7220_HwErrStatus_PCIeBusParity_LSB 0x1F
+#define QIB_7220_HwErrStatus_PCIeBusParity_RMASK 0x7
+#define QIB_7220_HwErrStatus_PcieCplTimeout_LSB 0x1E
+#define QIB_7220_HwErrStatus_PcieCplTimeout_RMASK 0x1
+#define QIB_7220_HwErrStatus_PoisenedTLP_LSB 0x1D
+#define QIB_7220_HwErrStatus_PoisenedTLP_RMASK 0x1
+#define QIB_7220_HwErrStatus_SDmaMemReadErr_LSB 0x1C
+#define QIB_7220_HwErrStatus_SDmaMemReadErr_RMASK 0x1
+#define QIB_7220_HwErrStatus_Reserved3_LSB 0x8
+#define QIB_7220_HwErrStatus_Reserved3_RMASK 0xFFFFF
+#define QIB_7220_HwErrStatus_PCIeMemParity_LSB 0x0
+#define QIB_7220_HwErrStatus_PCIeMemParity_RMASK 0xFF
+
+#define QIB_7220_HwErrClear_OFFS 0xA8
+#define QIB_7220_HwErrClear_IBCBusFromSPCParityErrClear_LSB 0x3F
+#define QIB_7220_HwErrClear_IBCBusFromSPCParityErrClear_RMASK 0x1
+#define QIB_7220_HwErrClear_IBCBusToSPCparityErrClear_LSB 0x3E
+#define QIB_7220_HwErrClear_IBCBusToSPCparityErrClear_RMASK 0x1
+#define QIB_7220_HwErrClear_Clk_uC_PLLNotLockedClear_LSB 0x3D
+#define QIB_7220_HwErrClear_Clk_uC_PLLNotLockedClear_RMASK 0x1
+#define QIB_7220_HwErrClear_IBSerdesPClkNotDetectClear_LSB 0x3C
+#define QIB_7220_HwErrClear_IBSerdesPClkNotDetectClear_RMASK 0x1
+#define QIB_7220_HwErrClear_PCIESerdesQ3PClkNotDetectClear_LSB 0x3B
+#define QIB_7220_HwErrClear_PCIESerdesQ3PClkNotDetectClear_RMASK 0x1
+#define QIB_7220_HwErrClear_PCIESerdesQ2PClkNotDetectClear_LSB 0x3A
+#define QIB_7220_HwErrClear_PCIESerdesQ2PClkNotDetectClear_RMASK 0x1
+#define QIB_7220_HwErrClear_PCIESerdesQ1PClkNotDetectClear_LSB 0x39
+#define QIB_7220_HwErrClear_PCIESerdesQ1PClkNotDetectClear_RMASK 0x1
+#define QIB_7220_HwErrClear_PCIESerdesQ0PClkNotDetectClear_LSB 0x38
+#define QIB_7220_HwErrClear_PCIESerdesQ0PClkNotDetectClear_RMASK 0x1
+#define QIB_7220_HwErrClear_Reserved_LSB 0x37
+#define QIB_7220_HwErrClear_Reserved_RMASK 0x1
+#define QIB_7220_HwErrClear_PowerOnBISTFailedClear_LSB 0x36
+#define QIB_7220_HwErrClear_PowerOnBISTFailedClear_RMASK 0x1
+#define QIB_7220_HwErrClear_Reserved1_LSB 0x33
+#define QIB_7220_HwErrClear_Reserved1_RMASK 0x7
+#define QIB_7220_HwErrClear_RXEMemParityClear_LSB 0x2C
+#define QIB_7220_HwErrClear_RXEMemParityClear_RMASK 0x7F
+#define QIB_7220_HwErrClear_TXEMemParityClear_LSB 0x28
+#define QIB_7220_HwErrClear_TXEMemParityClear_RMASK 0xF
+#define QIB_7220_HwErrClear_DDSRXEQMemoryParityErrClear_LSB 0x27
+#define QIB_7220_HwErrClear_DDSRXEQMemoryParityErrClear_RMASK 0x1
+#define QIB_7220_HwErrClear_IB_uC_MemoryParityErrClear_LSB 0x26
+#define QIB_7220_HwErrClear_IB_uC_MemoryParityErrClear_RMASK 0x1
+#define QIB_7220_HwErrClear_PCIE_uC_Oct1MemoryParityErrClear_LSB 0x25
+#define QIB_7220_HwErrClear_PCIE_uC_Oct1MemoryParityErrClear_RMASK 0x1
+#define QIB_7220_HwErrClear_PCIE_uC_Oct0MemoryParityErrClear_LSB 0x24
+#define QIB_7220_HwErrClear_PCIE_uC_Oct0MemoryParityErrClear_RMASK 0x1
+#define QIB_7220_HwErrClear_Reserved2_LSB 0x22
+#define QIB_7220_HwErrClear_Reserved2_RMASK 0x3
+#define QIB_7220_HwErrClear_PCIeBusParityClr_LSB 0x1F
+#define QIB_7220_HwErrClear_PCIeBusParityClr_RMASK 0x7
+#define QIB_7220_HwErrClear_PcieCplTimeoutClear_LSB 0x1E
+#define QIB_7220_HwErrClear_PcieCplTimeoutClear_RMASK 0x1
+#define QIB_7220_HwErrClear_PoisonedTLPClear_LSB 0x1D
+#define QIB_7220_HwErrClear_PoisonedTLPClear_RMASK 0x1
+#define QIB_7220_HwErrClear_SDmaMemReadErrClear_LSB 0x1C
+#define QIB_7220_HwErrClear_SDmaMemReadErrClear_RMASK 0x1
+#define QIB_7220_HwErrClear_Reserved3_LSB 0x8
+#define QIB_7220_HwErrClear_Reserved3_RMASK 0xFFFFF
+#define QIB_7220_HwErrClear_PCIeMemParityClr_LSB 0x0
+#define QIB_7220_HwErrClear_PCIeMemParityClr_RMASK 0xFF
+
+#define QIB_7220_HwDiagCtrl_OFFS 0xB0
+#define QIB_7220_HwDiagCtrl_ForceIBCBusFromSPCParityErr_LSB 0x3F
+#define QIB_7220_HwDiagCtrl_ForceIBCBusFromSPCParityErr_RMASK 0x1
+#define QIB_7220_HwDiagCtrl_ForceIBCBusToSPCParityErr_LSB 0x3E
+#define QIB_7220_HwDiagCtrl_ForceIBCBusToSPCParityErr_RMASK 0x1
+#define QIB_7220_HwDiagCtrl_CounterWrEnable_LSB 0x3D
+#define QIB_7220_HwDiagCtrl_CounterWrEnable_RMASK 0x1
+#define QIB_7220_HwDiagCtrl_CounterDisable_LSB 0x3C
+#define QIB_7220_HwDiagCtrl_CounterDisable_RMASK 0x1
+#define QIB_7220_HwDiagCtrl_Reserved_LSB 0x33
+#define QIB_7220_HwDiagCtrl_Reserved_RMASK 0x1FF
+#define QIB_7220_HwDiagCtrl_ForceRxMemParityErr_LSB 0x2C
+#define QIB_7220_HwDiagCtrl_ForceRxMemParityErr_RMASK 0x7F
+#define QIB_7220_HwDiagCtrl_ForceTxMemparityErr_LSB 0x28
+#define QIB_7220_HwDiagCtrl_ForceTxMemparityErr_RMASK 0xF
+#define QIB_7220_HwDiagCtrl_ForceDDSRXEQMemoryParityErr_LSB 0x27
+#define QIB_7220_HwDiagCtrl_ForceDDSRXEQMemoryParityErr_RMASK 0x1
+#define QIB_7220_HwDiagCtrl_ForceIB_uC_MemoryParityErr_LSB 0x26
+#define QIB_7220_HwDiagCtrl_ForceIB_uC_MemoryParityErr_RMASK 0x1
+#define QIB_7220_HwDiagCtrl_ForcePCIE_uC_Oct1MemoryParityErr_LSB 0x25
+#define QIB_7220_HwDiagCtrl_ForcePCIE_uC_Oct1MemoryParityErr_RMASK 0x1
+#define QIB_7220_HwDiagCtrl_ForcePCIE_uC_Oct0MemoryParityErr_LSB 0x24
+#define QIB_7220_HwDiagCtrl_ForcePCIE_uC_Oct0MemoryParityErr_RMASK 0x1
+#define QIB_7220_HwDiagCtrl_Reserved1_LSB 0x23
+#define QIB_7220_HwDiagCtrl_Reserved1_RMASK 0x1
+#define QIB_7220_HwDiagCtrl_forcePCIeBusParity_LSB 0x1F
+#define QIB_7220_HwDiagCtrl_forcePCIeBusParity_RMASK 0xF
+#define QIB_7220_HwDiagCtrl_Reserved2_LSB 0x8
+#define QIB_7220_HwDiagCtrl_Reserved2_RMASK 0x7FFFFF
+#define QIB_7220_HwDiagCtrl_forcePCIeMemParity_LSB 0x0
+#define QIB_7220_HwDiagCtrl_forcePCIeMemParity_RMASK 0xFF
+
+#define QIB_7220_REG_0000B8_OFFS 0xB8
+
+#define QIB_7220_IBCStatus_OFFS 0xC0
+#define QIB_7220_IBCStatus_TxCreditOk_LSB 0x1F
+#define QIB_7220_IBCStatus_TxCreditOk_RMASK 0x1
+#define QIB_7220_IBCStatus_TxReady_LSB 0x1E
+#define QIB_7220_IBCStatus_TxReady_RMASK 0x1
+#define QIB_7220_IBCStatus_Reserved_LSB 0xE
+#define QIB_7220_IBCStatus_Reserved_RMASK 0xFFFF
+#define QIB_7220_IBCStatus_IBTxLaneReversed_LSB 0xD
+#define QIB_7220_IBCStatus_IBTxLaneReversed_RMASK 0x1
+#define QIB_7220_IBCStatus_IBRxLaneReversed_LSB 0xC
+#define QIB_7220_IBCStatus_IBRxLaneReversed_RMASK 0x1
+#define QIB_7220_IBCStatus_IB_SERDES_TRIM_DONE_LSB 0xB
+#define QIB_7220_IBCStatus_IB_SERDES_TRIM_DONE_RMASK 0x1
+#define QIB_7220_IBCStatus_DDS_RXEQ_FAIL_LSB 0xA
+#define QIB_7220_IBCStatus_DDS_RXEQ_FAIL_RMASK 0x1
+#define QIB_7220_IBCStatus_LinkWidthActive_LSB 0x9
+#define QIB_7220_IBCStatus_LinkWidthActive_RMASK 0x1
+#define QIB_7220_IBCStatus_LinkSpeedActive_LSB 0x8
+#define QIB_7220_IBCStatus_LinkSpeedActive_RMASK 0x1
+#define QIB_7220_IBCStatus_LinkState_LSB 0x5
+#define QIB_7220_IBCStatus_LinkState_RMASK 0x7
+#define QIB_7220_IBCStatus_LinkTrainingState_LSB 0x0
+#define QIB_7220_IBCStatus_LinkTrainingState_RMASK 0x1F
+
+#define QIB_7220_IBCCtrl_OFFS 0xC8
+#define QIB_7220_IBCCtrl_Loopback_LSB 0x3F
+#define QIB_7220_IBCCtrl_Loopback_RMASK 0x1
+#define QIB_7220_IBCCtrl_LinkDownDefaultState_LSB 0x3E
+#define QIB_7220_IBCCtrl_LinkDownDefaultState_RMASK 0x1
+#define QIB_7220_IBCCtrl_Reserved_LSB 0x2B
+#define QIB_7220_IBCCtrl_Reserved_RMASK 0x7FFFF
+#define QIB_7220_IBCCtrl_CreditScale_LSB 0x28
+#define QIB_7220_IBCCtrl_CreditScale_RMASK 0x7
+#define QIB_7220_IBCCtrl_OverrunThreshold_LSB 0x24
+#define QIB_7220_IBCCtrl_OverrunThreshold_RMASK 0xF
+#define QIB_7220_IBCCtrl_PhyerrThreshold_LSB 0x20
+#define QIB_7220_IBCCtrl_PhyerrThreshold_RMASK 0xF
+#define QIB_7220_IBCCtrl_MaxPktLen_LSB 0x15
+#define QIB_7220_IBCCtrl_MaxPktLen_RMASK 0x7FF
+#define QIB_7220_IBCCtrl_LinkCmd_LSB 0x13
+#define QIB_7220_IBCCtrl_LinkCmd_RMASK 0x3
+#define QIB_7220_IBCCtrl_LinkInitCmd_LSB 0x10
+#define QIB_7220_IBCCtrl_LinkInitCmd_RMASK 0x7
+#define QIB_7220_IBCCtrl_FlowCtrlWaterMark_LSB 0x8
+#define QIB_7220_IBCCtrl_FlowCtrlWaterMark_RMASK 0xFF
+#define QIB_7220_IBCCtrl_FlowCtrlPeriod_LSB 0x0
+#define QIB_7220_IBCCtrl_FlowCtrlPeriod_RMASK 0xFF
+
+#define QIB_7220_EXTStatus_OFFS 0xD0
+#define QIB_7220_EXTStatus_GPIOIn_LSB 0x30
+#define QIB_7220_EXTStatus_GPIOIn_RMASK 0xFFFF
+#define QIB_7220_EXTStatus_Reserved_LSB 0x20
+#define QIB_7220_EXTStatus_Reserved_RMASK 0xFFFF
+#define QIB_7220_EXTStatus_Reserved1_LSB 0x10
+#define QIB_7220_EXTStatus_Reserved1_RMASK 0xFFFF
+#define QIB_7220_EXTStatus_MemBISTDisabled_LSB 0xF
+#define QIB_7220_EXTStatus_MemBISTDisabled_RMASK 0x1
+#define QIB_7220_EXTStatus_MemBISTEndTest_LSB 0xE
+#define QIB_7220_EXTStatus_MemBISTEndTest_RMASK 0x1
+#define QIB_7220_EXTStatus_Reserved2_LSB 0x0
+#define QIB_7220_EXTStatus_Reserved2_RMASK 0x3FFF
+
+#define QIB_7220_EXTCtrl_OFFS 0xD8
+#define QIB_7220_EXTCtrl_GPIOOe_LSB 0x30
+#define QIB_7220_EXTCtrl_GPIOOe_RMASK 0xFFFF
+#define QIB_7220_EXTCtrl_GPIOInvert_LSB 0x20
+#define QIB_7220_EXTCtrl_GPIOInvert_RMASK 0xFFFF
+#define QIB_7220_EXTCtrl_Reserved_LSB 0x4
+#define QIB_7220_EXTCtrl_Reserved_RMASK 0xFFFFFFF
+#define QIB_7220_EXTCtrl_LEDPriPortGreenOn_LSB 0x3
+#define QIB_7220_EXTCtrl_LEDPriPortGreenOn_RMASK 0x1
+#define QIB_7220_EXTCtrl_LEDPriPortYellowOn_LSB 0x2
+#define QIB_7220_EXTCtrl_LEDPriPortYellowOn_RMASK 0x1
+#define QIB_7220_EXTCtrl_LEDGblOkGreenOn_LSB 0x1
+#define QIB_7220_EXTCtrl_LEDGblOkGreenOn_RMASK 0x1
+#define QIB_7220_EXTCtrl_LEDGblErrRedOff_LSB 0x0
+#define QIB_7220_EXTCtrl_LEDGblErrRedOff_RMASK 0x1
+
+#define QIB_7220_GPIOOut_OFFS 0xE0
+
+#define QIB_7220_GPIOMask_OFFS 0xE8
+
+#define QIB_7220_GPIOStatus_OFFS 0xF0
+
+#define QIB_7220_GPIOClear_OFFS 0xF8
+
+#define QIB_7220_RcvCtrl_OFFS 0x100
+#define QIB_7220_RcvCtrl_Reserved_LSB 0x27
+#define QIB_7220_RcvCtrl_Reserved_RMASK 0x1FFFFFF
+#define QIB_7220_RcvCtrl_RcvQPMapEnable_LSB 0x26
+#define QIB_7220_RcvCtrl_RcvQPMapEnable_RMASK 0x1
+#define QIB_7220_RcvCtrl_PortCfg_LSB 0x24
+#define QIB_7220_RcvCtrl_PortCfg_RMASK 0x3
+#define QIB_7220_RcvCtrl_TailUpd_LSB 0x23
+#define QIB_7220_RcvCtrl_TailUpd_RMASK 0x1
+#define QIB_7220_RcvCtrl_RcvPartitionKeyDisable_LSB 0x22
+#define QIB_7220_RcvCtrl_RcvPartitionKeyDisable_RMASK 0x1
+#define QIB_7220_RcvCtrl_IntrAvail_LSB 0x11
+#define QIB_7220_RcvCtrl_IntrAvail_RMASK 0x1FFFF
+#define QIB_7220_RcvCtrl_PortEnable_LSB 0x0
+#define QIB_7220_RcvCtrl_PortEnable_RMASK 0x1FFFF
+
+#define QIB_7220_RcvBTHQP_OFFS 0x108
+#define QIB_7220_RcvBTHQP_Reserved_LSB 0x18
+#define QIB_7220_RcvBTHQP_Reserved_RMASK 0xFF
+#define QIB_7220_RcvBTHQP_RcvBTHQP_LSB 0x0
+#define QIB_7220_RcvBTHQP_RcvBTHQP_RMASK 0xFFFFFF
+
+#define QIB_7220_RcvHdrSize_OFFS 0x110
+
+#define QIB_7220_RcvHdrCnt_OFFS 0x118
+
+#define QIB_7220_RcvHdrEntSize_OFFS 0x120
+
+#define QIB_7220_RcvTIDBase_OFFS 0x128
+
+#define QIB_7220_RcvTIDCnt_OFFS 0x130
+
+#define QIB_7220_RcvEgrBase_OFFS 0x138
+
+#define QIB_7220_RcvEgrCnt_OFFS 0x140
+
+#define QIB_7220_RcvBufBase_OFFS 0x148
+
+#define QIB_7220_RcvBufSize_OFFS 0x150
+
+#define QIB_7220_RxIntMemBase_OFFS 0x158
+
+#define QIB_7220_RxIntMemSize_OFFS 0x160
+
+#define QIB_7220_RcvPartitionKey_OFFS 0x168
+
+#define QIB_7220_RcvQPMulticastPort_OFFS 0x170
+#define QIB_7220_RcvQPMulticastPort_Reserved_LSB 0x5
+#define QIB_7220_RcvQPMulticastPort_Reserved_RMASK 0x7FFFFFFFFFFFFFF
+#define QIB_7220_RcvQPMulticastPort_RcvQpMcPort_LSB 0x0
+#define QIB_7220_RcvQPMulticastPort_RcvQpMcPort_RMASK 0x1F
+
+#define QIB_7220_RcvPktLEDCnt_OFFS 0x178
+#define QIB_7220_RcvPktLEDCnt_ONperiod_LSB 0x20
+#define QIB_7220_RcvPktLEDCnt_ONperiod_RMASK 0xFFFFFFFF
+#define QIB_7220_RcvPktLEDCnt_OFFperiod_LSB 0x0
+#define QIB_7220_RcvPktLEDCnt_OFFperiod_RMASK 0xFFFFFFFF
+
+#define QIB_7220_IBCDDRCtrl_OFFS 0x180
+#define QIB_7220_IBCDDRCtrl_IB_DLID_MASK_LSB 0x30
+#define QIB_7220_IBCDDRCtrl_IB_DLID_MASK_RMASK 0xFFFF
+#define QIB_7220_IBCDDRCtrl_IB_DLID_LSB 0x20
+#define QIB_7220_IBCDDRCtrl_IB_DLID_RMASK 0xFFFF
+#define QIB_7220_IBCDDRCtrl_Reserved_LSB 0x1B
+#define QIB_7220_IBCDDRCtrl_Reserved_RMASK 0x1F
+#define QIB_7220_IBCDDRCtrl_HRTBT_REQ_LSB 0x1A
+#define QIB_7220_IBCDDRCtrl_HRTBT_REQ_RMASK 0x1
+#define QIB_7220_IBCDDRCtrl_HRTBT_PORT_LSB 0x12
+#define QIB_7220_IBCDDRCtrl_HRTBT_PORT_RMASK 0xFF
+#define QIB_7220_IBCDDRCtrl_HRTBT_AUTO_LSB 0x11
+#define QIB_7220_IBCDDRCtrl_HRTBT_AUTO_RMASK 0x1
+#define QIB_7220_IBCDDRCtrl_HRTBT_ENB_LSB 0x10
+#define QIB_7220_IBCDDRCtrl_HRTBT_ENB_RMASK 0x1
+#define QIB_7220_IBCDDRCtrl_SD_DDS_LSB 0xC
+#define QIB_7220_IBCDDRCtrl_SD_DDS_RMASK 0xF
+#define QIB_7220_IBCDDRCtrl_SD_DDSV_LSB 0xB
+#define QIB_7220_IBCDDRCtrl_SD_DDSV_RMASK 0x1
+#define QIB_7220_IBCDDRCtrl_SD_ADD_ENB_LSB 0xA
+#define QIB_7220_IBCDDRCtrl_SD_ADD_ENB_RMASK 0x1
+#define QIB_7220_IBCDDRCtrl_SD_RX_EQUAL_ENABLE_LSB 0x9
+#define QIB_7220_IBCDDRCtrl_SD_RX_EQUAL_ENABLE_RMASK 0x1
+#define QIB_7220_IBCDDRCtrl_IB_LANE_REV_SUPPORTED_LSB 0x8
+#define QIB_7220_IBCDDRCtrl_IB_LANE_REV_SUPPORTED_RMASK 0x1
+#define QIB_7220_IBCDDRCtrl_IB_POLARITY_REV_SUPP_LSB 0x7
+#define QIB_7220_IBCDDRCtrl_IB_POLARITY_REV_SUPP_RMASK 0x1
+#define QIB_7220_IBCDDRCtrl_IB_NUM_CHANNELS_LSB 0x5
+#define QIB_7220_IBCDDRCtrl_IB_NUM_CHANNELS_RMASK 0x3
+#define QIB_7220_IBCDDRCtrl_SD_SPEED_QDR_LSB 0x4
+#define QIB_7220_IBCDDRCtrl_SD_SPEED_QDR_RMASK 0x1
+#define QIB_7220_IBCDDRCtrl_SD_SPEED_DDR_LSB 0x3
+#define QIB_7220_IBCDDRCtrl_SD_SPEED_DDR_RMASK 0x1
+#define QIB_7220_IBCDDRCtrl_SD_SPEED_SDR_LSB 0x2
+#define QIB_7220_IBCDDRCtrl_SD_SPEED_SDR_RMASK 0x1
+#define QIB_7220_IBCDDRCtrl_SD_SPEED_LSB 0x1
+#define QIB_7220_IBCDDRCtrl_SD_SPEED_RMASK 0x1
+#define QIB_7220_IBCDDRCtrl_IB_ENHANCED_MODE_LSB 0x0
+#define QIB_7220_IBCDDRCtrl_IB_ENHANCED_MODE_RMASK 0x1
+
+#define QIB_7220_HRTBT_GUID_OFFS 0x188
+
+#define QIB_7220_IBCDDRCtrl2_OFFS 0x1A0
+#define QIB_7220_IBCDDRCtrl2_IB_BACK_PORCH_LSB 0x5
+#define QIB_7220_IBCDDRCtrl2_IB_BACK_PORCH_RMASK 0x1F
+#define QIB_7220_IBCDDRCtrl2_IB_FRONT_PORCH_LSB 0x0
+#define QIB_7220_IBCDDRCtrl2_IB_FRONT_PORCH_RMASK 0x1F
+
+#define QIB_7220_IBCDDRStatus_OFFS 0x1A8
+#define QIB_7220_IBCDDRStatus_heartbeat_timed_out_LSB 0x24
+#define QIB_7220_IBCDDRStatus_heartbeat_timed_out_RMASK 0x1
+#define QIB_7220_IBCDDRStatus_heartbeat_crosstalk_LSB 0x20
+#define QIB_7220_IBCDDRStatus_heartbeat_crosstalk_RMASK 0xF
+#define QIB_7220_IBCDDRStatus_RxEqLocalDevice_LSB 0x1E
+#define QIB_7220_IBCDDRStatus_RxEqLocalDevice_RMASK 0x3
+#define QIB_7220_IBCDDRStatus_ReqDDSLocalFromRmt_LSB 0x1A
+#define QIB_7220_IBCDDRStatus_ReqDDSLocalFromRmt_RMASK 0xF
+#define QIB_7220_IBCDDRStatus_LinkRoundTripLatency_LSB 0x0
+#define QIB_7220_IBCDDRStatus_LinkRoundTripLatency_RMASK 0x3FFFFFF
+
+#define QIB_7220_JIntReload_OFFS 0x1B0
+#define QIB_7220_JIntReload_J_limit_reload_LSB 0x10
+#define QIB_7220_JIntReload_J_limit_reload_RMASK 0xFFFF
+#define QIB_7220_JIntReload_J_reload_LSB 0x0
+#define QIB_7220_JIntReload_J_reload_RMASK 0xFFFF
+
+#define QIB_7220_IBNCModeCtrl_OFFS 0x1B8
+#define QIB_7220_IBNCModeCtrl_Reserved_LSB 0x1A
+#define QIB_7220_IBNCModeCtrl_Reserved_RMASK 0x3FFFFFFFFF
+#define QIB_7220_IBNCModeCtrl_TSMCode_TS2_LSB 0x11
+#define QIB_7220_IBNCModeCtrl_TSMCode_TS2_RMASK 0x1FF
+#define QIB_7220_IBNCModeCtrl_TSMCode_TS1_LSB 0x8
+#define QIB_7220_IBNCModeCtrl_TSMCode_TS1_RMASK 0x1FF
+#define QIB_7220_IBNCModeCtrl_Reserved1_LSB 0x3
+#define QIB_7220_IBNCModeCtrl_Reserved1_RMASK 0x1F
+#define QIB_7220_IBNCModeCtrl_TSMEnable_ignore_TSM_on_rx_LSB 0x2
+#define QIB_7220_IBNCModeCtrl_TSMEnable_ignore_TSM_on_rx_RMASK 0x1
+#define QIB_7220_IBNCModeCtrl_TSMEnable_send_TS2_LSB 0x1
+#define QIB_7220_IBNCModeCtrl_TSMEnable_send_TS2_RMASK 0x1
+#define QIB_7220_IBNCModeCtrl_TSMEnable_send_TS1_LSB 0x0
+#define QIB_7220_IBNCModeCtrl_TSMEnable_send_TS1_RMASK 0x1
+
+#define QIB_7220_SendCtrl_OFFS 0x1C0
+#define QIB_7220_SendCtrl_Disarm_LSB 0x1F
+#define QIB_7220_SendCtrl_Disarm_RMASK 0x1
+#define QIB_7220_SendCtrl_Reserved_LSB 0x1D
+#define QIB_7220_SendCtrl_Reserved_RMASK 0x3
+#define QIB_7220_SendCtrl_AvailUpdThld_LSB 0x18
+#define QIB_7220_SendCtrl_AvailUpdThld_RMASK 0x1F
+#define QIB_7220_SendCtrl_DisarmPIOBuf_LSB 0x10
+#define QIB_7220_SendCtrl_DisarmPIOBuf_RMASK 0xFF
+#define QIB_7220_SendCtrl_Reserved1_LSB 0xD
+#define QIB_7220_SendCtrl_Reserved1_RMASK 0x7
+#define QIB_7220_SendCtrl_SDmaHalt_LSB 0xC
+#define QIB_7220_SendCtrl_SDmaHalt_RMASK 0x1
+#define QIB_7220_SendCtrl_SDmaEnable_LSB 0xB
+#define QIB_7220_SendCtrl_SDmaEnable_RMASK 0x1
+#define QIB_7220_SendCtrl_SDmaSingleDescriptor_LSB 0xA
+#define QIB_7220_SendCtrl_SDmaSingleDescriptor_RMASK 0x1
+#define QIB_7220_SendCtrl_SDmaIntEnable_LSB 0x9
+#define QIB_7220_SendCtrl_SDmaIntEnable_RMASK 0x1
+#define QIB_7220_SendCtrl_Reserved2_LSB 0x5
+#define QIB_7220_SendCtrl_Reserved2_RMASK 0xF
+#define QIB_7220_SendCtrl_SSpecialTriggerEn_LSB 0x4
+#define QIB_7220_SendCtrl_SSpecialTriggerEn_RMASK 0x1
+#define QIB_7220_SendCtrl_SPioEnable_LSB 0x3
+#define QIB_7220_SendCtrl_SPioEnable_RMASK 0x1
+#define QIB_7220_SendCtrl_SendBufAvailUpd_LSB 0x2
+#define QIB_7220_SendCtrl_SendBufAvailUpd_RMASK 0x1
+#define QIB_7220_SendCtrl_SendIntBufAvail_LSB 0x1
+#define QIB_7220_SendCtrl_SendIntBufAvail_RMASK 0x1
+#define QIB_7220_SendCtrl_Abort_LSB 0x0
+#define QIB_7220_SendCtrl_Abort_RMASK 0x1
+
+#define QIB_7220_SendBufBase_OFFS 0x1C8
+#define QIB_7220_SendBufBase_Reserved_LSB 0x35
+#define QIB_7220_SendBufBase_Reserved_RMASK 0x7FF
+#define QIB_7220_SendBufBase_BaseAddr_LargePIO_LSB 0x20
+#define QIB_7220_SendBufBase_BaseAddr_LargePIO_RMASK 0x1FFFFF
+#define QIB_7220_SendBufBase_Reserved1_LSB 0x15
+#define QIB_7220_SendBufBase_Reserved1_RMASK 0x7FF
+#define QIB_7220_SendBufBase_BaseAddr_SmallPIO_LSB 0x0
+#define QIB_7220_SendBufBase_BaseAddr_SmallPIO_RMASK 0x1FFFFF
+
+#define QIB_7220_SendBufSize_OFFS 0x1D0
+#define QIB_7220_SendBufSize_Reserved_LSB 0x2D
+#define QIB_7220_SendBufSize_Reserved_RMASK 0xFFFFF
+#define QIB_7220_SendBufSize_Size_LargePIO_LSB 0x20
+#define QIB_7220_SendBufSize_Size_LargePIO_RMASK 0x1FFF
+#define QIB_7220_SendBufSize_Reserved1_LSB 0xC
+#define QIB_7220_SendBufSize_Reserved1_RMASK 0xFFFFF
+#define QIB_7220_SendBufSize_Size_SmallPIO_LSB 0x0
+#define QIB_7220_SendBufSize_Size_SmallPIO_RMASK 0xFFF
+
+#define QIB_7220_SendBufCnt_OFFS 0x1D8
+#define QIB_7220_SendBufCnt_Reserved_LSB 0x24
+#define QIB_7220_SendBufCnt_Reserved_RMASK 0xFFFFFFF
+#define QIB_7220_SendBufCnt_Num_LargeBuffers_LSB 0x20
+#define QIB_7220_SendBufCnt_Num_LargeBuffers_RMASK 0xF
+#define QIB_7220_SendBufCnt_Reserved1_LSB 0x9
+#define QIB_7220_SendBufCnt_Reserved1_RMASK 0x7FFFFF
+#define QIB_7220_SendBufCnt_Num_SmallBuffers_LSB 0x0
+#define QIB_7220_SendBufCnt_Num_SmallBuffers_RMASK 0x1FF
+
+#define QIB_7220_SendBufAvailAddr_OFFS 0x1E0
+#define QIB_7220_SendBufAvailAddr_SendBufAvailAddr_LSB 0x6
+#define QIB_7220_SendBufAvailAddr_SendBufAvailAddr_RMASK 0x3FFFFFFFF
+#define QIB_7220_SendBufAvailAddr_Reserved_LSB 0x0
+#define QIB_7220_SendBufAvailAddr_Reserved_RMASK 0x3F
+
+#define QIB_7220_TxIntMemBase_OFFS 0x1E8
+
+#define QIB_7220_TxIntMemSize_OFFS 0x1F0
+
+#define QIB_7220_SendDmaBase_OFFS 0x1F8
+#define QIB_7220_SendDmaBase_Reserved_LSB 0x30
+#define QIB_7220_SendDmaBase_Reserved_RMASK 0xFFFF
+#define QIB_7220_SendDmaBase_SendDmaBase_LSB 0x0
+#define QIB_7220_SendDmaBase_SendDmaBase_RMASK 0xFFFFFFFFFFFF
+
+#define QIB_7220_SendDmaLenGen_OFFS 0x200
+#define QIB_7220_SendDmaLenGen_Reserved_LSB 0x13
+#define QIB_7220_SendDmaLenGen_Reserved_RMASK 0x1FFFFFFFFFFF
+#define QIB_7220_SendDmaLenGen_Generation_LSB 0x10
+#define QIB_7220_SendDmaLenGen_Generation_MSB 0x12
+#define QIB_7220_SendDmaLenGen_Generation_RMASK 0x7
+#define QIB_7220_SendDmaLenGen_Length_LSB 0x0
+#define QIB_7220_SendDmaLenGen_Length_RMASK 0xFFFF
+
+#define QIB_7220_SendDmaTail_OFFS 0x208
+#define QIB_7220_SendDmaTail_Reserved_LSB 0x10
+#define QIB_7220_SendDmaTail_Reserved_RMASK 0xFFFFFFFFFFFF
+#define QIB_7220_SendDmaTail_SendDmaTail_LSB 0x0
+#define QIB_7220_SendDmaTail_SendDmaTail_RMASK 0xFFFF
+
+#define QIB_7220_SendDmaHead_OFFS 0x210
+#define QIB_7220_SendDmaHead_Reserved_LSB 0x30
+#define QIB_7220_SendDmaHead_Reserved_RMASK 0xFFFF
+#define QIB_7220_SendDmaHead_InternalSendDmaHead_LSB 0x20
+#define QIB_7220_SendDmaHead_InternalSendDmaHead_RMASK 0xFFFF
+#define QIB_7220_SendDmaHead_Reserved1_LSB 0x10
+#define QIB_7220_SendDmaHead_Reserved1_RMASK 0xFFFF
+#define QIB_7220_SendDmaHead_SendDmaHead_LSB 0x0
+#define QIB_7220_SendDmaHead_SendDmaHead_RMASK 0xFFFF
+
+#define QIB_7220_SendDmaHeadAddr_OFFS 0x218
+#define QIB_7220_SendDmaHeadAddr_Reserved_LSB 0x30
+#define QIB_7220_SendDmaHeadAddr_Reserved_RMASK 0xFFFF
+#define QIB_7220_SendDmaHeadAddr_SendDmaHeadAddr_LSB 0x0
+#define QIB_7220_SendDmaHeadAddr_SendDmaHeadAddr_RMASK 0xFFFFFFFFFFFF
+
+#define QIB_7220_SendDmaBufMask0_OFFS 0x220
+#define QIB_7220_SendDmaBufMask0_BufMask_63_0_LSB 0x0
+#define QIB_7220_SendDmaBufMask0_BufMask_63_0_RMASK 0x0
+
+#define QIB_7220_SendDmaStatus_OFFS 0x238
+#define QIB_7220_SendDmaStatus_ScoreBoardDrainInProg_LSB 0x3F
+#define QIB_7220_SendDmaStatus_ScoreBoardDrainInProg_RMASK 0x1
+#define QIB_7220_SendDmaStatus_AbortInProg_LSB 0x3E
+#define QIB_7220_SendDmaStatus_AbortInProg_RMASK 0x1
+#define QIB_7220_SendDmaStatus_InternalSDmaEnable_LSB 0x3D
+#define QIB_7220_SendDmaStatus_InternalSDmaEnable_RMASK 0x1
+#define QIB_7220_SendDmaStatus_ScbDescIndex_13_0_LSB 0x2F
+#define QIB_7220_SendDmaStatus_ScbDescIndex_13_0_RMASK 0x3FFF
+#define QIB_7220_SendDmaStatus_RpyLowAddr_6_0_LSB 0x28
+#define QIB_7220_SendDmaStatus_RpyLowAddr_6_0_RMASK 0x7F
+#define QIB_7220_SendDmaStatus_RpyTag_7_0_LSB 0x20
+#define QIB_7220_SendDmaStatus_RpyTag_7_0_RMASK 0xFF
+#define QIB_7220_SendDmaStatus_ScbFull_LSB 0x1F
+#define QIB_7220_SendDmaStatus_ScbFull_RMASK 0x1
+#define QIB_7220_SendDmaStatus_ScbEmpty_LSB 0x1E
+#define QIB_7220_SendDmaStatus_ScbEmpty_RMASK 0x1
+#define QIB_7220_SendDmaStatus_ScbEntryValid_LSB 0x1D
+#define QIB_7220_SendDmaStatus_ScbEntryValid_RMASK 0x1
+#define QIB_7220_SendDmaStatus_ScbFetchDescFlag_LSB 0x1C
+#define QIB_7220_SendDmaStatus_ScbFetchDescFlag_RMASK 0x1
+#define QIB_7220_SendDmaStatus_SplFifoReadyToGo_LSB 0x1B
+#define QIB_7220_SendDmaStatus_SplFifoReadyToGo_RMASK 0x1
+#define QIB_7220_SendDmaStatus_SplFifoDisarmed_LSB 0x1A
+#define QIB_7220_SendDmaStatus_SplFifoDisarmed_RMASK 0x1
+#define QIB_7220_SendDmaStatus_SplFifoEmpty_LSB 0x19
+#define QIB_7220_SendDmaStatus_SplFifoEmpty_RMASK 0x1
+#define QIB_7220_SendDmaStatus_SplFifoFull_LSB 0x18
+#define QIB_7220_SendDmaStatus_SplFifoFull_RMASK 0x1
+#define QIB_7220_SendDmaStatus_SplFifoBufNum_LSB 0x10
+#define QIB_7220_SendDmaStatus_SplFifoBufNum_RMASK 0xFF
+#define QIB_7220_SendDmaStatus_SplFifoDescIndex_LSB 0x0
+#define QIB_7220_SendDmaStatus_SplFifoDescIndex_RMASK 0xFFFF
+
+#define QIB_7220_SendBufErr0_OFFS 0x240
+#define QIB_7220_SendBufErr0_SendBufErr_63_0_LSB 0x0
+#define QIB_7220_SendBufErr0_SendBufErr_63_0_RMASK 0x0
+
+#define QIB_7220_RcvHdrAddr0_OFFS 0x270
+#define QIB_7220_RcvHdrAddr0_RcvHdrAddr0_LSB 0x2
+#define QIB_7220_RcvHdrAddr0_RcvHdrAddr0_RMASK 0x3FFFFFFFFF
+#define QIB_7220_RcvHdrAddr0_Reserved_LSB 0x0
+#define QIB_7220_RcvHdrAddr0_Reserved_RMASK 0x3
+
+#define QIB_7220_RcvHdrTailAddr0_OFFS 0x300
+#define QIB_7220_RcvHdrTailAddr0_RcvHdrTailAddr0_LSB 0x2
+#define QIB_7220_RcvHdrTailAddr0_RcvHdrTailAddr0_RMASK 0x3FFFFFFFFF
+#define QIB_7220_RcvHdrTailAddr0_Reserved_LSB 0x0
+#define QIB_7220_RcvHdrTailAddr0_Reserved_RMASK 0x3
+
+#define QIB_7220_ibsd_epb_access_ctrl_OFFS 0x3C0
+#define QIB_7220_ibsd_epb_access_ctrl_sw_ib_epb_req_granted_LSB 0x8
+#define QIB_7220_ibsd_epb_access_ctrl_sw_ib_epb_req_granted_RMASK 0x1
+#define QIB_7220_ibsd_epb_access_ctrl_Reserved_LSB 0x1
+#define QIB_7220_ibsd_epb_access_ctrl_Reserved_RMASK 0x7F
+#define QIB_7220_ibsd_epb_access_ctrl_sw_ib_epb_req_LSB 0x0
+#define QIB_7220_ibsd_epb_access_ctrl_sw_ib_epb_req_RMASK 0x1
+
+#define QIB_7220_ibsd_epb_transaction_reg_OFFS 0x3C8
+#define QIB_7220_ibsd_epb_transaction_reg_ib_epb_rdy_LSB 0x1F
+#define QIB_7220_ibsd_epb_transaction_reg_ib_epb_rdy_RMASK 0x1
+#define QIB_7220_ibsd_epb_transaction_reg_ib_epb_req_error_LSB 0x1E
+#define QIB_7220_ibsd_epb_transaction_reg_ib_epb_req_error_RMASK 0x1
+#define QIB_7220_ibsd_epb_transaction_reg_Reserved_LSB 0x1D
+#define QIB_7220_ibsd_epb_transaction_reg_Reserved_RMASK 0x1
+#define QIB_7220_ibsd_epb_transaction_reg_mem_data_parity_LSB 0x1C
+#define QIB_7220_ibsd_epb_transaction_reg_mem_data_parity_RMASK 0x1
+#define QIB_7220_ibsd_epb_transaction_reg_Reserved1_LSB 0x1B
+#define QIB_7220_ibsd_epb_transaction_reg_Reserved1_RMASK 0x1
+#define QIB_7220_ibsd_epb_transaction_reg_ib_epb_cs_LSB 0x19
+#define QIB_7220_ibsd_epb_transaction_reg_ib_epb_cs_RMASK 0x3
+#define QIB_7220_ibsd_epb_transaction_reg_ib_epb_read_write_LSB 0x18
+#define QIB_7220_ibsd_epb_transaction_reg_ib_epb_read_write_RMASK 0x1
+#define QIB_7220_ibsd_epb_transaction_reg_Reserved2_LSB 0x17
+#define QIB_7220_ibsd_epb_transaction_reg_Reserved2_RMASK 0x1
+#define QIB_7220_ibsd_epb_transaction_reg_ib_epb_address_LSB 0x8
+#define QIB_7220_ibsd_epb_transaction_reg_ib_epb_address_RMASK 0x7FFF
+#define QIB_7220_ibsd_epb_transaction_reg_ib_epb_data_LSB 0x0
+#define QIB_7220_ibsd_epb_transaction_reg_ib_epb_data_RMASK 0xFF
+
+#define QIB_7220_XGXSCfg_OFFS 0x3D8
+#define QIB_7220_XGXSCfg_sel_link_down_for_fctrl_lane_sync_reset_LSB 0x3F
+#define QIB_7220_XGXSCfg_sel_link_down_for_fctrl_lane_sync_reset_RMASK 0x1
+#define QIB_7220_XGXSCfg_Reserved_LSB 0x13
+#define QIB_7220_XGXSCfg_Reserved_RMASK 0xFFFFFFFFFFF
+#define QIB_7220_XGXSCfg_link_sync_mask_LSB 0x9
+#define QIB_7220_XGXSCfg_link_sync_mask_RMASK 0x3FF
+#define QIB_7220_XGXSCfg_Reserved1_LSB 0x3
+#define QIB_7220_XGXSCfg_Reserved1_RMASK 0x3F
+#define QIB_7220_XGXSCfg_xcv_reset_LSB 0x2
+#define QIB_7220_XGXSCfg_xcv_reset_RMASK 0x1
+#define QIB_7220_XGXSCfg_Reserved2_LSB 0x1
+#define QIB_7220_XGXSCfg_Reserved2_RMASK 0x1
+#define QIB_7220_XGXSCfg_tx_rx_reset_LSB 0x0
+#define QIB_7220_XGXSCfg_tx_rx_reset_RMASK 0x1
+
+#define QIB_7220_IBSerDesCtrl_OFFS 0x3E0
+#define QIB_7220_IBSerDesCtrl_Reserved_LSB 0x2D
+#define QIB_7220_IBSerDesCtrl_Reserved_RMASK 0x7FFFF
+#define QIB_7220_IBSerDesCtrl_INT_uC_LSB 0x2C
+#define QIB_7220_IBSerDesCtrl_INT_uC_RMASK 0x1
+#define QIB_7220_IBSerDesCtrl_CKSEL_uC_LSB 0x2A
+#define QIB_7220_IBSerDesCtrl_CKSEL_uC_RMASK 0x3
+#define QIB_7220_IBSerDesCtrl_PLLN_LSB 0x28
+#define QIB_7220_IBSerDesCtrl_PLLN_RMASK 0x3
+#define QIB_7220_IBSerDesCtrl_PLLM_LSB 0x25
+#define QIB_7220_IBSerDesCtrl_PLLM_RMASK 0x7
+#define QIB_7220_IBSerDesCtrl_TXOBPD_LSB 0x24
+#define QIB_7220_IBSerDesCtrl_TXOBPD_RMASK 0x1
+#define QIB_7220_IBSerDesCtrl_TWC_LSB 0x23
+#define QIB_7220_IBSerDesCtrl_TWC_RMASK 0x1
+#define QIB_7220_IBSerDesCtrl_RXIDLE_LSB 0x22
+#define QIB_7220_IBSerDesCtrl_RXIDLE_RMASK 0x1
+#define QIB_7220_IBSerDesCtrl_RXINV_LSB 0x21
+#define QIB_7220_IBSerDesCtrl_RXINV_RMASK 0x1
+#define QIB_7220_IBSerDesCtrl_TXINV_LSB 0x20
+#define QIB_7220_IBSerDesCtrl_TXINV_RMASK 0x1
+#define QIB_7220_IBSerDesCtrl_Reserved1_LSB 0x12
+#define QIB_7220_IBSerDesCtrl_Reserved1_RMASK 0x3FFF
+#define QIB_7220_IBSerDesCtrl_NumSerDesRegsToWrForRXEQ_LSB 0xD
+#define QIB_7220_IBSerDesCtrl_NumSerDesRegsToWrForRXEQ_RMASK 0x1F
+#define QIB_7220_IBSerDesCtrl_NumSerDesRegsToWrForDDS_LSB 0x8
+#define QIB_7220_IBSerDesCtrl_NumSerDesRegsToWrForDDS_RMASK 0x1F
+#define QIB_7220_IBSerDesCtrl_Reserved2_LSB 0x1
+#define QIB_7220_IBSerDesCtrl_Reserved2_RMASK 0x7F
+#define QIB_7220_IBSerDesCtrl_ResetIB_uC_Core_LSB 0x0
+#define QIB_7220_IBSerDesCtrl_ResetIB_uC_Core_RMASK 0x1
+
+#define QIB_7220_pciesd_epb_access_ctrl_OFFS 0x400
+#define QIB_7220_pciesd_epb_access_ctrl_sw_pcie_epb_req_granted_LSB 0x8
+#define QIB_7220_pciesd_epb_access_ctrl_sw_pcie_epb_req_granted_RMASK 0x1
+#define QIB_7220_pciesd_epb_access_ctrl_Reserved_LSB 0x3
+#define QIB_7220_pciesd_epb_access_ctrl_Reserved_RMASK 0x1F
+#define QIB_7220_pciesd_epb_access_ctrl_sw_pcieepb_star_en_LSB 0x1
+#define QIB_7220_pciesd_epb_access_ctrl_sw_pcieepb_star_en_RMASK 0x3
+#define QIB_7220_pciesd_epb_access_ctrl_sw_pcie_epb_req_LSB 0x0
+#define QIB_7220_pciesd_epb_access_ctrl_sw_pcie_epb_req_RMASK 0x1
+
+#define QIB_7220_pciesd_epb_transaction_reg_OFFS 0x408
+#define QIB_7220_pciesd_epb_transaction_reg_pcie_epb_rdy_LSB 0x1F
+#define QIB_7220_pciesd_epb_transaction_reg_pcie_epb_rdy_RMASK 0x1
+#define QIB_7220_pciesd_epb_transaction_reg_pcie_epb_req_error_LSB 0x1E
+#define QIB_7220_pciesd_epb_transaction_reg_pcie_epb_req_error_RMASK 0x1
+#define QIB_7220_pciesd_epb_transaction_reg_Reserved_LSB 0x1D
+#define QIB_7220_pciesd_epb_transaction_reg_Reserved_RMASK 0x1
+#define QIB_7220_pciesd_epb_transaction_reg_mem_data_parity_LSB 0x1C
+#define QIB_7220_pciesd_epb_transaction_reg_mem_data_parity_RMASK 0x1
+#define QIB_7220_pciesd_epb_transaction_reg_pcie_epb_cs_LSB 0x19
+#define QIB_7220_pciesd_epb_transaction_reg_pcie_epb_cs_RMASK 0x7
+#define QIB_7220_pciesd_epb_transaction_reg_pcie_epb_read_write_LSB 0x18
+#define QIB_7220_pciesd_epb_transaction_reg_pcie_epb_read_write_RMASK 0x1
+#define QIB_7220_pciesd_epb_transaction_reg_Reserved1_LSB 0x17
+#define QIB_7220_pciesd_epb_transaction_reg_Reserved1_RMASK 0x1
+#define QIB_7220_pciesd_epb_transaction_reg_pcie_epb_address_LSB 0x8
+#define QIB_7220_pciesd_epb_transaction_reg_pcie_epb_address_RMASK 0x7FFF
+#define QIB_7220_pciesd_epb_transaction_reg_pcie_epb_data_LSB 0x0
+#define QIB_7220_pciesd_epb_transaction_reg_pcie_epb_data_RMASK 0xFF
+
+#define QIB_7220_SerDes_DDSRXEQ0_OFFS 0x500
+#define QIB_7220_SerDes_DDSRXEQ0_reg_addr_LSB 0x4
+#define QIB_7220_SerDes_DDSRXEQ0_reg_addr_RMASK 0x3F
+#define QIB_7220_SerDes_DDSRXEQ0_element_num_LSB 0x0
+#define QIB_7220_SerDes_DDSRXEQ0_element_num_RMASK 0xF
+
+#define QIB_7220_LBIntCnt_OFFS 0x13000
+
+#define QIB_7220_LBFlowStallCnt_OFFS 0x13008
+
+#define QIB_7220_TxSDmaDescCnt_OFFS 0x13010
+
+#define QIB_7220_TxUnsupVLErrCnt_OFFS 0x13018
+
+#define QIB_7220_TxDataPktCnt_OFFS 0x13020
+
+#define QIB_7220_TxFlowPktCnt_OFFS 0x13028
+
+#define QIB_7220_TxDwordCnt_OFFS 0x13030
+
+#define QIB_7220_TxLenErrCnt_OFFS 0x13038
+
+#define QIB_7220_TxMaxMinLenErrCnt_OFFS 0x13040
+
+#define QIB_7220_TxUnderrunCnt_OFFS 0x13048
+
+#define QIB_7220_TxFlowStallCnt_OFFS 0x13050
+
+#define QIB_7220_TxDroppedPktCnt_OFFS 0x13058
+
+#define QIB_7220_RxDroppedPktCnt_OFFS 0x13060
+
+#define QIB_7220_RxDataPktCnt_OFFS 0x13068
+
+#define QIB_7220_RxFlowPktCnt_OFFS 0x13070
+
+#define QIB_7220_RxDwordCnt_OFFS 0x13078
+
+#define QIB_7220_RxLenErrCnt_OFFS 0x13080
+
+#define QIB_7220_RxMaxMinLenErrCnt_OFFS 0x13088
+
+#define QIB_7220_RxICRCErrCnt_OFFS 0x13090
+
+#define QIB_7220_RxVCRCErrCnt_OFFS 0x13098
+
+#define QIB_7220_RxFlowCtrlViolCnt_OFFS 0x130A0
+
+#define QIB_7220_RxVersionErrCnt_OFFS 0x130A8
+
+#define QIB_7220_RxLinkMalformCnt_OFFS 0x130B0
+
+#define QIB_7220_RxEBPCnt_OFFS 0x130B8
+
+#define QIB_7220_RxLPCRCErrCnt_OFFS 0x130C0
+
+#define QIB_7220_RxBufOvflCnt_OFFS 0x130C8
+
+#define QIB_7220_RxTIDFullErrCnt_OFFS 0x130D0
+
+#define QIB_7220_RxTIDValidErrCnt_OFFS 0x130D8
+
+#define QIB_7220_RxPKeyMismatchCnt_OFFS 0x130E0
+
+#define QIB_7220_RxP0HdrEgrOvflCnt_OFFS 0x130E8
+
+#define QIB_7220_IBStatusChangeCnt_OFFS 0x13170
+
+#define QIB_7220_IBLinkErrRecoveryCnt_OFFS 0x13178
+
+#define QIB_7220_IBLinkDownedCnt_OFFS 0x13180
+
+#define QIB_7220_IBSymbolErrCnt_OFFS 0x13188
+
+#define QIB_7220_RxVL15DroppedPktCnt_OFFS 0x13190
+
+#define QIB_7220_RxOtherLocalPhyErrCnt_OFFS 0x13198
+
+#define QIB_7220_PcieRetryBufDiagQwordCnt_OFFS 0x131A0
+
+#define QIB_7220_ExcessBufferOvflCnt_OFFS 0x131A8
+
+#define QIB_7220_LocalLinkIntegrityErrCnt_OFFS 0x131B0
+
+#define QIB_7220_RxVlErrCnt_OFFS 0x131B8
+
+#define QIB_7220_RxDlidFltrCnt_OFFS 0x131C0
+
+#define QIB_7220_CNT_0131C8_OFFS 0x131C8
+
+#define QIB_7220_PSStat_OFFS 0x13200
+
+#define QIB_7220_PSStart_OFFS 0x13208
+
+#define QIB_7220_PSInterval_OFFS 0x13210
+
+#define QIB_7220_PSRcvDataCount_OFFS 0x13218
+
+#define QIB_7220_PSRcvPktsCount_OFFS 0x13220
+
+#define QIB_7220_PSXmitDataCount_OFFS 0x13228
+
+#define QIB_7220_PSXmitPktsCount_OFFS 0x13230
+
+#define QIB_7220_PSXmitWaitCount_OFFS 0x13238
+
+#define QIB_7220_CNT_013240_OFFS 0x13240
+
+#define QIB_7220_RcvEgrArray_OFFS 0x14000
+
+#define QIB_7220_MEM_038000_OFFS 0x38000
+
+#define QIB_7220_RcvTIDArray0_OFFS 0x53000
+
+#define QIB_7220_PIOLaunchFIFO_OFFS 0x64000
+
+#define QIB_7220_MEM_064480_OFFS 0x64480
+
+#define QIB_7220_SendPIOpbcCache_OFFS 0x64800
+
+#define QIB_7220_MEM_064C80_OFFS 0x64C80
+
+#define QIB_7220_PreLaunchFIFO_OFFS 0x65000
+
+#define QIB_7220_MEM_065080_OFFS 0x65080
+
+#define QIB_7220_ScoreBoard_OFFS 0x65400
+
+#define QIB_7220_MEM_065440_OFFS 0x65440
+
+#define QIB_7220_DescriptorFIFO_OFFS 0x65800
+
+#define QIB_7220_MEM_065880_OFFS 0x65880
+
+#define QIB_7220_RcvBuf1_OFFS 0x72000
+
+#define QIB_7220_MEM_074800_OFFS 0x74800
+
+#define QIB_7220_RcvBuf2_OFFS 0x75000
+
+#define QIB_7220_MEM_076400_OFFS 0x76400
+
+#define QIB_7220_RcvFlags_OFFS 0x77000
+
+#define QIB_7220_MEM_078400_OFFS 0x78400
+
+#define QIB_7220_RcvLookupBuf1_OFFS 0x79000
+
+#define QIB_7220_MEM_07A400_OFFS 0x7A400
+
+#define QIB_7220_RcvDMADatBuf_OFFS 0x7B000
+
+#define QIB_7220_RcvDMAHdrBuf_OFFS 0x7B800
+
+#define QIB_7220_MiscRXEIntMem_OFFS 0x7C000
+
+#define QIB_7220_MEM_07D400_OFFS 0x7D400
+
+#define QIB_7220_PCIERcvBuf_OFFS 0x80000
+
+#define QIB_7220_PCIERetryBuf_OFFS 0x84000
+
+#define QIB_7220_PCIERcvBufRdToWrAddr_OFFS 0x88000
+
+#define QIB_7220_PCIECplBuf_OFFS 0x90000
+
+#define QIB_7220_IBSerDesMappTable_OFFS 0x94000
+
+#define QIB_7220_MEM_095000_OFFS 0x95000
+
+#define QIB_7220_SendBuf0_MA_OFFS 0x100000
+
+#define QIB_7220_MEM_1A0000_OFFS 0x1A0000
diff --git a/drivers/infiniband/hw/qib/qib_7322_regs.h b/drivers/infiniband/hw/qib/qib_7322_regs.h
new file mode 100644
index 000000000000..a97440ba924c
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_7322_regs.h
@@ -0,0 +1,3163 @@
+/*
+ * Copyright (c) 2008, 2009, 2010 QLogic Corporation. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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.
+ */
+
+/* This file is mechanically generated from RTL. Any hand-edits will be lost! */
+
+#define QIB_7322_Revision_OFFS 0x0
+#define QIB_7322_Revision_DEF 0x0000000002010601
+#define QIB_7322_Revision_R_Simulator_LSB 0x3F
+#define QIB_7322_Revision_R_Simulator_MSB 0x3F
+#define QIB_7322_Revision_R_Simulator_RMASK 0x1
+#define QIB_7322_Revision_R_Emulation_LSB 0x3E
+#define QIB_7322_Revision_R_Emulation_MSB 0x3E
+#define QIB_7322_Revision_R_Emulation_RMASK 0x1
+#define QIB_7322_Revision_R_Emulation_Revcode_LSB 0x28
+#define QIB_7322_Revision_R_Emulation_Revcode_MSB 0x3D
+#define QIB_7322_Revision_R_Emulation_Revcode_RMASK 0x3FFFFF
+#define QIB_7322_Revision_BoardID_LSB 0x20
+#define QIB_7322_Revision_BoardID_MSB 0x27
+#define QIB_7322_Revision_BoardID_RMASK 0xFF
+#define QIB_7322_Revision_R_SW_LSB 0x18
+#define QIB_7322_Revision_R_SW_MSB 0x1F
+#define QIB_7322_Revision_R_SW_RMASK 0xFF
+#define QIB_7322_Revision_R_Arch_LSB 0x10
+#define QIB_7322_Revision_R_Arch_MSB 0x17
+#define QIB_7322_Revision_R_Arch_RMASK 0xFF
+#define QIB_7322_Revision_R_ChipRevMajor_LSB 0x8
+#define QIB_7322_Revision_R_ChipRevMajor_MSB 0xF
+#define QIB_7322_Revision_R_ChipRevMajor_RMASK 0xFF
+#define QIB_7322_Revision_R_ChipRevMinor_LSB 0x0
+#define QIB_7322_Revision_R_ChipRevMinor_MSB 0x7
+#define QIB_7322_Revision_R_ChipRevMinor_RMASK 0xFF
+
+#define QIB_7322_Control_OFFS 0x8
+#define QIB_7322_Control_DEF 0x0000000000000000
+#define QIB_7322_Control_PCIECplQDiagEn_LSB 0x6
+#define QIB_7322_Control_PCIECplQDiagEn_MSB 0x6
+#define QIB_7322_Control_PCIECplQDiagEn_RMASK 0x1
+#define QIB_7322_Control_PCIEPostQDiagEn_LSB 0x5
+#define QIB_7322_Control_PCIEPostQDiagEn_MSB 0x5
+#define QIB_7322_Control_PCIEPostQDiagEn_RMASK 0x1
+#define QIB_7322_Control_SDmaDescFetchPriorityEn_LSB 0x4
+#define QIB_7322_Control_SDmaDescFetchPriorityEn_MSB 0x4
+#define QIB_7322_Control_SDmaDescFetchPriorityEn_RMASK 0x1
+#define QIB_7322_Control_PCIERetryBufDiagEn_LSB 0x3
+#define QIB_7322_Control_PCIERetryBufDiagEn_MSB 0x3
+#define QIB_7322_Control_PCIERetryBufDiagEn_RMASK 0x1
+#define QIB_7322_Control_FreezeMode_LSB 0x1
+#define QIB_7322_Control_FreezeMode_MSB 0x1
+#define QIB_7322_Control_FreezeMode_RMASK 0x1
+#define QIB_7322_Control_SyncReset_LSB 0x0
+#define QIB_7322_Control_SyncReset_MSB 0x0
+#define QIB_7322_Control_SyncReset_RMASK 0x1
+
+#define QIB_7322_PageAlign_OFFS 0x10
+#define QIB_7322_PageAlign_DEF 0x0000000000001000
+
+#define QIB_7322_ContextCnt_OFFS 0x18
+#define QIB_7322_ContextCnt_DEF 0x0000000000000012
+
+#define QIB_7322_Scratch_OFFS 0x20
+#define QIB_7322_Scratch_DEF 0x0000000000000000
+
+#define QIB_7322_CntrRegBase_OFFS 0x28
+#define QIB_7322_CntrRegBase_DEF 0x0000000000011000
+
+#define QIB_7322_SendRegBase_OFFS 0x30
+#define QIB_7322_SendRegBase_DEF 0x0000000000003000
+
+#define QIB_7322_UserRegBase_OFFS 0x38
+#define QIB_7322_UserRegBase_DEF 0x0000000000200000
+
+#define QIB_7322_IntMask_OFFS 0x68
+#define QIB_7322_IntMask_DEF 0x0000000000000000
+#define QIB_7322_IntMask_SDmaIntMask_1_LSB 0x3F
+#define QIB_7322_IntMask_SDmaIntMask_1_MSB 0x3F
+#define QIB_7322_IntMask_SDmaIntMask_1_RMASK 0x1
+#define QIB_7322_IntMask_SDmaIntMask_0_LSB 0x3E
+#define QIB_7322_IntMask_SDmaIntMask_0_MSB 0x3E
+#define QIB_7322_IntMask_SDmaIntMask_0_RMASK 0x1
+#define QIB_7322_IntMask_SDmaProgressIntMask_1_LSB 0x3D
+#define QIB_7322_IntMask_SDmaProgressIntMask_1_MSB 0x3D
+#define QIB_7322_IntMask_SDmaProgressIntMask_1_RMASK 0x1
+#define QIB_7322_IntMask_SDmaProgressIntMask_0_LSB 0x3C
+#define QIB_7322_IntMask_SDmaProgressIntMask_0_MSB 0x3C
+#define QIB_7322_IntMask_SDmaProgressIntMask_0_RMASK 0x1
+#define QIB_7322_IntMask_SDmaIdleIntMask_1_LSB 0x3B
+#define QIB_7322_IntMask_SDmaIdleIntMask_1_MSB 0x3B
+#define QIB_7322_IntMask_SDmaIdleIntMask_1_RMASK 0x1
+#define QIB_7322_IntMask_SDmaIdleIntMask_0_LSB 0x3A
+#define QIB_7322_IntMask_SDmaIdleIntMask_0_MSB 0x3A
+#define QIB_7322_IntMask_SDmaIdleIntMask_0_RMASK 0x1
+#define QIB_7322_IntMask_SDmaCleanupDoneMask_1_LSB 0x39
+#define QIB_7322_IntMask_SDmaCleanupDoneMask_1_MSB 0x39
+#define QIB_7322_IntMask_SDmaCleanupDoneMask_1_RMASK 0x1
+#define QIB_7322_IntMask_SDmaCleanupDoneMask_0_LSB 0x38
+#define QIB_7322_IntMask_SDmaCleanupDoneMask_0_MSB 0x38
+#define QIB_7322_IntMask_SDmaCleanupDoneMask_0_RMASK 0x1
+#define QIB_7322_IntMask_RcvUrg17IntMask_LSB 0x31
+#define QIB_7322_IntMask_RcvUrg17IntMask_MSB 0x31
+#define QIB_7322_IntMask_RcvUrg17IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvUrg16IntMask_LSB 0x30
+#define QIB_7322_IntMask_RcvUrg16IntMask_MSB 0x30
+#define QIB_7322_IntMask_RcvUrg16IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvUrg15IntMask_LSB 0x2F
+#define QIB_7322_IntMask_RcvUrg15IntMask_MSB 0x2F
+#define QIB_7322_IntMask_RcvUrg15IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvUrg14IntMask_LSB 0x2E
+#define QIB_7322_IntMask_RcvUrg14IntMask_MSB 0x2E
+#define QIB_7322_IntMask_RcvUrg14IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvUrg13IntMask_LSB 0x2D
+#define QIB_7322_IntMask_RcvUrg13IntMask_MSB 0x2D
+#define QIB_7322_IntMask_RcvUrg13IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvUrg12IntMask_LSB 0x2C
+#define QIB_7322_IntMask_RcvUrg12IntMask_MSB 0x2C
+#define QIB_7322_IntMask_RcvUrg12IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvUrg11IntMask_LSB 0x2B
+#define QIB_7322_IntMask_RcvUrg11IntMask_MSB 0x2B
+#define QIB_7322_IntMask_RcvUrg11IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvUrg10IntMask_LSB 0x2A
+#define QIB_7322_IntMask_RcvUrg10IntMask_MSB 0x2A
+#define QIB_7322_IntMask_RcvUrg10IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvUrg9IntMask_LSB 0x29
+#define QIB_7322_IntMask_RcvUrg9IntMask_MSB 0x29
+#define QIB_7322_IntMask_RcvUrg9IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvUrg8IntMask_LSB 0x28
+#define QIB_7322_IntMask_RcvUrg8IntMask_MSB 0x28
+#define QIB_7322_IntMask_RcvUrg8IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvUrg7IntMask_LSB 0x27
+#define QIB_7322_IntMask_RcvUrg7IntMask_MSB 0x27
+#define QIB_7322_IntMask_RcvUrg7IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvUrg6IntMask_LSB 0x26
+#define QIB_7322_IntMask_RcvUrg6IntMask_MSB 0x26
+#define QIB_7322_IntMask_RcvUrg6IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvUrg5IntMask_LSB 0x25
+#define QIB_7322_IntMask_RcvUrg5IntMask_MSB 0x25
+#define QIB_7322_IntMask_RcvUrg5IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvUrg4IntMask_LSB 0x24
+#define QIB_7322_IntMask_RcvUrg4IntMask_MSB 0x24
+#define QIB_7322_IntMask_RcvUrg4IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvUrg3IntMask_LSB 0x23
+#define QIB_7322_IntMask_RcvUrg3IntMask_MSB 0x23
+#define QIB_7322_IntMask_RcvUrg3IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvUrg2IntMask_LSB 0x22
+#define QIB_7322_IntMask_RcvUrg2IntMask_MSB 0x22
+#define QIB_7322_IntMask_RcvUrg2IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvUrg1IntMask_LSB 0x21
+#define QIB_7322_IntMask_RcvUrg1IntMask_MSB 0x21
+#define QIB_7322_IntMask_RcvUrg1IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvUrg0IntMask_LSB 0x20
+#define QIB_7322_IntMask_RcvUrg0IntMask_MSB 0x20
+#define QIB_7322_IntMask_RcvUrg0IntMask_RMASK 0x1
+#define QIB_7322_IntMask_ErrIntMask_1_LSB 0x1F
+#define QIB_7322_IntMask_ErrIntMask_1_MSB 0x1F
+#define QIB_7322_IntMask_ErrIntMask_1_RMASK 0x1
+#define QIB_7322_IntMask_ErrIntMask_0_LSB 0x1E
+#define QIB_7322_IntMask_ErrIntMask_0_MSB 0x1E
+#define QIB_7322_IntMask_ErrIntMask_0_RMASK 0x1
+#define QIB_7322_IntMask_ErrIntMask_LSB 0x1D
+#define QIB_7322_IntMask_ErrIntMask_MSB 0x1D
+#define QIB_7322_IntMask_ErrIntMask_RMASK 0x1
+#define QIB_7322_IntMask_AssertGPIOIntMask_LSB 0x1C
+#define QIB_7322_IntMask_AssertGPIOIntMask_MSB 0x1C
+#define QIB_7322_IntMask_AssertGPIOIntMask_RMASK 0x1
+#define QIB_7322_IntMask_SendDoneIntMask_1_LSB 0x19
+#define QIB_7322_IntMask_SendDoneIntMask_1_MSB 0x19
+#define QIB_7322_IntMask_SendDoneIntMask_1_RMASK 0x1
+#define QIB_7322_IntMask_SendDoneIntMask_0_LSB 0x18
+#define QIB_7322_IntMask_SendDoneIntMask_0_MSB 0x18
+#define QIB_7322_IntMask_SendDoneIntMask_0_RMASK 0x1
+#define QIB_7322_IntMask_SendBufAvailIntMask_LSB 0x17
+#define QIB_7322_IntMask_SendBufAvailIntMask_MSB 0x17
+#define QIB_7322_IntMask_SendBufAvailIntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvAvail17IntMask_LSB 0x11
+#define QIB_7322_IntMask_RcvAvail17IntMask_MSB 0x11
+#define QIB_7322_IntMask_RcvAvail17IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvAvail16IntMask_LSB 0x10
+#define QIB_7322_IntMask_RcvAvail16IntMask_MSB 0x10
+#define QIB_7322_IntMask_RcvAvail16IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvAvail15IntMask_LSB 0xF
+#define QIB_7322_IntMask_RcvAvail15IntMask_MSB 0xF
+#define QIB_7322_IntMask_RcvAvail15IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvAvail14IntMask_LSB 0xE
+#define QIB_7322_IntMask_RcvAvail14IntMask_MSB 0xE
+#define QIB_7322_IntMask_RcvAvail14IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvAvail13IntMask_LSB 0xD
+#define QIB_7322_IntMask_RcvAvail13IntMask_MSB 0xD
+#define QIB_7322_IntMask_RcvAvail13IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvAvail12IntMask_LSB 0xC
+#define QIB_7322_IntMask_RcvAvail12IntMask_MSB 0xC
+#define QIB_7322_IntMask_RcvAvail12IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvAvail11IntMask_LSB 0xB
+#define QIB_7322_IntMask_RcvAvail11IntMask_MSB 0xB
+#define QIB_7322_IntMask_RcvAvail11IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvAvail10IntMask_LSB 0xA
+#define QIB_7322_IntMask_RcvAvail10IntMask_MSB 0xA
+#define QIB_7322_IntMask_RcvAvail10IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvAvail9IntMask_LSB 0x9
+#define QIB_7322_IntMask_RcvAvail9IntMask_MSB 0x9
+#define QIB_7322_IntMask_RcvAvail9IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvAvail8IntMask_LSB 0x8
+#define QIB_7322_IntMask_RcvAvail8IntMask_MSB 0x8
+#define QIB_7322_IntMask_RcvAvail8IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvAvail7IntMask_LSB 0x7
+#define QIB_7322_IntMask_RcvAvail7IntMask_MSB 0x7
+#define QIB_7322_IntMask_RcvAvail7IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvAvail6IntMask_LSB 0x6
+#define QIB_7322_IntMask_RcvAvail6IntMask_MSB 0x6
+#define QIB_7322_IntMask_RcvAvail6IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvAvail5IntMask_LSB 0x5
+#define QIB_7322_IntMask_RcvAvail5IntMask_MSB 0x5
+#define QIB_7322_IntMask_RcvAvail5IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvAvail4IntMask_LSB 0x4
+#define QIB_7322_IntMask_RcvAvail4IntMask_MSB 0x4
+#define QIB_7322_IntMask_RcvAvail4IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvAvail3IntMask_LSB 0x3
+#define QIB_7322_IntMask_RcvAvail3IntMask_MSB 0x3
+#define QIB_7322_IntMask_RcvAvail3IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvAvail2IntMask_LSB 0x2
+#define QIB_7322_IntMask_RcvAvail2IntMask_MSB 0x2
+#define QIB_7322_IntMask_RcvAvail2IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvAvail1IntMask_LSB 0x1
+#define QIB_7322_IntMask_RcvAvail1IntMask_MSB 0x1
+#define QIB_7322_IntMask_RcvAvail1IntMask_RMASK 0x1
+#define QIB_7322_IntMask_RcvAvail0IntMask_LSB 0x0
+#define QIB_7322_IntMask_RcvAvail0IntMask_MSB 0x0
+#define QIB_7322_IntMask_RcvAvail0IntMask_RMASK 0x1
+
+#define QIB_7322_IntStatus_OFFS 0x70
+#define QIB_7322_IntStatus_DEF 0x0000000000000000
+#define QIB_7322_IntStatus_SDmaInt_1_LSB 0x3F
+#define QIB_7322_IntStatus_SDmaInt_1_MSB 0x3F
+#define QIB_7322_IntStatus_SDmaInt_1_RMASK 0x1
+#define QIB_7322_IntStatus_SDmaInt_0_LSB 0x3E
+#define QIB_7322_IntStatus_SDmaInt_0_MSB 0x3E
+#define QIB_7322_IntStatus_SDmaInt_0_RMASK 0x1
+#define QIB_7322_IntStatus_SDmaProgressInt_1_LSB 0x3D
+#define QIB_7322_IntStatus_SDmaProgressInt_1_MSB 0x3D
+#define QIB_7322_IntStatus_SDmaProgressInt_1_RMASK 0x1
+#define QIB_7322_IntStatus_SDmaProgressInt_0_LSB 0x3C
+#define QIB_7322_IntStatus_SDmaProgressInt_0_MSB 0x3C
+#define QIB_7322_IntStatus_SDmaProgressInt_0_RMASK 0x1
+#define QIB_7322_IntStatus_SDmaIdleInt_1_LSB 0x3B
+#define QIB_7322_IntStatus_SDmaIdleInt_1_MSB 0x3B
+#define QIB_7322_IntStatus_SDmaIdleInt_1_RMASK 0x1
+#define QIB_7322_IntStatus_SDmaIdleInt_0_LSB 0x3A
+#define QIB_7322_IntStatus_SDmaIdleInt_0_MSB 0x3A
+#define QIB_7322_IntStatus_SDmaIdleInt_0_RMASK 0x1
+#define QIB_7322_IntStatus_SDmaCleanupDone_1_LSB 0x39
+#define QIB_7322_IntStatus_SDmaCleanupDone_1_MSB 0x39
+#define QIB_7322_IntStatus_SDmaCleanupDone_1_RMASK 0x1
+#define QIB_7322_IntStatus_SDmaCleanupDone_0_LSB 0x38
+#define QIB_7322_IntStatus_SDmaCleanupDone_0_MSB 0x38
+#define QIB_7322_IntStatus_SDmaCleanupDone_0_RMASK 0x1
+#define QIB_7322_IntStatus_RcvUrg17_LSB 0x31
+#define QIB_7322_IntStatus_RcvUrg17_MSB 0x31
+#define QIB_7322_IntStatus_RcvUrg17_RMASK 0x1
+#define QIB_7322_IntStatus_RcvUrg16_LSB 0x30
+#define QIB_7322_IntStatus_RcvUrg16_MSB 0x30
+#define QIB_7322_IntStatus_RcvUrg16_RMASK 0x1
+#define QIB_7322_IntStatus_RcvUrg15_LSB 0x2F
+#define QIB_7322_IntStatus_RcvUrg15_MSB 0x2F
+#define QIB_7322_IntStatus_RcvUrg15_RMASK 0x1
+#define QIB_7322_IntStatus_RcvUrg14_LSB 0x2E
+#define QIB_7322_IntStatus_RcvUrg14_MSB 0x2E
+#define QIB_7322_IntStatus_RcvUrg14_RMASK 0x1
+#define QIB_7322_IntStatus_RcvUrg13_LSB 0x2D
+#define QIB_7322_IntStatus_RcvUrg13_MSB 0x2D
+#define QIB_7322_IntStatus_RcvUrg13_RMASK 0x1
+#define QIB_7322_IntStatus_RcvUrg12_LSB 0x2C
+#define QIB_7322_IntStatus_RcvUrg12_MSB 0x2C
+#define QIB_7322_IntStatus_RcvUrg12_RMASK 0x1
+#define QIB_7322_IntStatus_RcvUrg11_LSB 0x2B
+#define QIB_7322_IntStatus_RcvUrg11_MSB 0x2B
+#define QIB_7322_IntStatus_RcvUrg11_RMASK 0x1
+#define QIB_7322_IntStatus_RcvUrg10_LSB 0x2A
+#define QIB_7322_IntStatus_RcvUrg10_MSB 0x2A
+#define QIB_7322_IntStatus_RcvUrg10_RMASK 0x1
+#define QIB_7322_IntStatus_RcvUrg9_LSB 0x29
+#define QIB_7322_IntStatus_RcvUrg9_MSB 0x29
+#define QIB_7322_IntStatus_RcvUrg9_RMASK 0x1
+#define QIB_7322_IntStatus_RcvUrg8_LSB 0x28
+#define QIB_7322_IntStatus_RcvUrg8_MSB 0x28
+#define QIB_7322_IntStatus_RcvUrg8_RMASK 0x1
+#define QIB_7322_IntStatus_RcvUrg7_LSB 0x27
+#define QIB_7322_IntStatus_RcvUrg7_MSB 0x27
+#define QIB_7322_IntStatus_RcvUrg7_RMASK 0x1
+#define QIB_7322_IntStatus_RcvUrg6_LSB 0x26
+#define QIB_7322_IntStatus_RcvUrg6_MSB 0x26
+#define QIB_7322_IntStatus_RcvUrg6_RMASK 0x1
+#define QIB_7322_IntStatus_RcvUrg5_LSB 0x25
+#define QIB_7322_IntStatus_RcvUrg5_MSB 0x25
+#define QIB_7322_IntStatus_RcvUrg5_RMASK 0x1
+#define QIB_7322_IntStatus_RcvUrg4_LSB 0x24
+#define QIB_7322_IntStatus_RcvUrg4_MSB 0x24
+#define QIB_7322_IntStatus_RcvUrg4_RMASK 0x1
+#define QIB_7322_IntStatus_RcvUrg3_LSB 0x23
+#define QIB_7322_IntStatus_RcvUrg3_MSB 0x23
+#define QIB_7322_IntStatus_RcvUrg3_RMASK 0x1
+#define QIB_7322_IntStatus_RcvUrg2_LSB 0x22
+#define QIB_7322_IntStatus_RcvUrg2_MSB 0x22
+#define QIB_7322_IntStatus_RcvUrg2_RMASK 0x1
+#define QIB_7322_IntStatus_RcvUrg1_LSB 0x21
+#define QIB_7322_IntStatus_RcvUrg1_MSB 0x21
+#define QIB_7322_IntStatus_RcvUrg1_RMASK 0x1
+#define QIB_7322_IntStatus_RcvUrg0_LSB 0x20
+#define QIB_7322_IntStatus_RcvUrg0_MSB 0x20
+#define QIB_7322_IntStatus_RcvUrg0_RMASK 0x1
+#define QIB_7322_IntStatus_Err_1_LSB 0x1F
+#define QIB_7322_IntStatus_Err_1_MSB 0x1F
+#define QIB_7322_IntStatus_Err_1_RMASK 0x1
+#define QIB_7322_IntStatus_Err_0_LSB 0x1E
+#define QIB_7322_IntStatus_Err_0_MSB 0x1E
+#define QIB_7322_IntStatus_Err_0_RMASK 0x1
+#define QIB_7322_IntStatus_Err_LSB 0x1D
+#define QIB_7322_IntStatus_Err_MSB 0x1D
+#define QIB_7322_IntStatus_Err_RMASK 0x1
+#define QIB_7322_IntStatus_AssertGPIO_LSB 0x1C
+#define QIB_7322_IntStatus_AssertGPIO_MSB 0x1C
+#define QIB_7322_IntStatus_AssertGPIO_RMASK 0x1
+#define QIB_7322_IntStatus_SendDone_1_LSB 0x19
+#define QIB_7322_IntStatus_SendDone_1_MSB 0x19
+#define QIB_7322_IntStatus_SendDone_1_RMASK 0x1
+#define QIB_7322_IntStatus_SendDone_0_LSB 0x18
+#define QIB_7322_IntStatus_SendDone_0_MSB 0x18
+#define QIB_7322_IntStatus_SendDone_0_RMASK 0x1
+#define QIB_7322_IntStatus_SendBufAvail_LSB 0x17
+#define QIB_7322_IntStatus_SendBufAvail_MSB 0x17
+#define QIB_7322_IntStatus_SendBufAvail_RMASK 0x1
+#define QIB_7322_IntStatus_RcvAvail17_LSB 0x11
+#define QIB_7322_IntStatus_RcvAvail17_MSB 0x11
+#define QIB_7322_IntStatus_RcvAvail17_RMASK 0x1
+#define QIB_7322_IntStatus_RcvAvail16_LSB 0x10
+#define QIB_7322_IntStatus_RcvAvail16_MSB 0x10
+#define QIB_7322_IntStatus_RcvAvail16_RMASK 0x1
+#define QIB_7322_IntStatus_RcvAvail15_LSB 0xF
+#define QIB_7322_IntStatus_RcvAvail15_MSB 0xF
+#define QIB_7322_IntStatus_RcvAvail15_RMASK 0x1
+#define QIB_7322_IntStatus_RcvAvail14_LSB 0xE
+#define QIB_7322_IntStatus_RcvAvail14_MSB 0xE
+#define QIB_7322_IntStatus_RcvAvail14_RMASK 0x1
+#define QIB_7322_IntStatus_RcvAvail13_LSB 0xD
+#define QIB_7322_IntStatus_RcvAvail13_MSB 0xD
+#define QIB_7322_IntStatus_RcvAvail13_RMASK 0x1
+#define QIB_7322_IntStatus_RcvAvail12_LSB 0xC
+#define QIB_7322_IntStatus_RcvAvail12_MSB 0xC
+#define QIB_7322_IntStatus_RcvAvail12_RMASK 0x1
+#define QIB_7322_IntStatus_RcvAvail11_LSB 0xB
+#define QIB_7322_IntStatus_RcvAvail11_MSB 0xB
+#define QIB_7322_IntStatus_RcvAvail11_RMASK 0x1
+#define QIB_7322_IntStatus_RcvAvail10_LSB 0xA
+#define QIB_7322_IntStatus_RcvAvail10_MSB 0xA
+#define QIB_7322_IntStatus_RcvAvail10_RMASK 0x1
+#define QIB_7322_IntStatus_RcvAvail9_LSB 0x9
+#define QIB_7322_IntStatus_RcvAvail9_MSB 0x9
+#define QIB_7322_IntStatus_RcvAvail9_RMASK 0x1
+#define QIB_7322_IntStatus_RcvAvail8_LSB 0x8
+#define QIB_7322_IntStatus_RcvAvail8_MSB 0x8
+#define QIB_7322_IntStatus_RcvAvail8_RMASK 0x1
+#define QIB_7322_IntStatus_RcvAvail7_LSB 0x7
+#define QIB_7322_IntStatus_RcvAvail7_MSB 0x7
+#define QIB_7322_IntStatus_RcvAvail7_RMASK 0x1
+#define QIB_7322_IntStatus_RcvAvail6_LSB 0x6
+#define QIB_7322_IntStatus_RcvAvail6_MSB 0x6
+#define QIB_7322_IntStatus_RcvAvail6_RMASK 0x1
+#define QIB_7322_IntStatus_RcvAvail5_LSB 0x5
+#define QIB_7322_IntStatus_RcvAvail5_MSB 0x5
+#define QIB_7322_IntStatus_RcvAvail5_RMASK 0x1
+#define QIB_7322_IntStatus_RcvAvail4_LSB 0x4
+#define QIB_7322_IntStatus_RcvAvail4_MSB 0x4
+#define QIB_7322_IntStatus_RcvAvail4_RMASK 0x1
+#define QIB_7322_IntStatus_RcvAvail3_LSB 0x3
+#define QIB_7322_IntStatus_RcvAvail3_MSB 0x3
+#define QIB_7322_IntStatus_RcvAvail3_RMASK 0x1
+#define QIB_7322_IntStatus_RcvAvail2_LSB 0x2
+#define QIB_7322_IntStatus_RcvAvail2_MSB 0x2
+#define QIB_7322_IntStatus_RcvAvail2_RMASK 0x1
+#define QIB_7322_IntStatus_RcvAvail1_LSB 0x1
+#define QIB_7322_IntStatus_RcvAvail1_MSB 0x1
+#define QIB_7322_IntStatus_RcvAvail1_RMASK 0x1
+#define QIB_7322_IntStatus_RcvAvail0_LSB 0x0
+#define QIB_7322_IntStatus_RcvAvail0_MSB 0x0
+#define QIB_7322_IntStatus_RcvAvail0_RMASK 0x1
+
+#define QIB_7322_IntClear_OFFS 0x78
+#define QIB_7322_IntClear_DEF 0x0000000000000000
+#define QIB_7322_IntClear_SDmaIntClear_1_LSB 0x3F
+#define QIB_7322_IntClear_SDmaIntClear_1_MSB 0x3F
+#define QIB_7322_IntClear_SDmaIntClear_1_RMASK 0x1
+#define QIB_7322_IntClear_SDmaIntClear_0_LSB 0x3E
+#define QIB_7322_IntClear_SDmaIntClear_0_MSB 0x3E
+#define QIB_7322_IntClear_SDmaIntClear_0_RMASK 0x1
+#define QIB_7322_IntClear_SDmaProgressIntClear_1_LSB 0x3D
+#define QIB_7322_IntClear_SDmaProgressIntClear_1_MSB 0x3D
+#define QIB_7322_IntClear_SDmaProgressIntClear_1_RMASK 0x1
+#define QIB_7322_IntClear_SDmaProgressIntClear_0_LSB 0x3C
+#define QIB_7322_IntClear_SDmaProgressIntClear_0_MSB 0x3C
+#define QIB_7322_IntClear_SDmaProgressIntClear_0_RMASK 0x1
+#define QIB_7322_IntClear_SDmaIdleIntClear_1_LSB 0x3B
+#define QIB_7322_IntClear_SDmaIdleIntClear_1_MSB 0x3B
+#define QIB_7322_IntClear_SDmaIdleIntClear_1_RMASK 0x1
+#define QIB_7322_IntClear_SDmaIdleIntClear_0_LSB 0x3A
+#define QIB_7322_IntClear_SDmaIdleIntClear_0_MSB 0x3A
+#define QIB_7322_IntClear_SDmaIdleIntClear_0_RMASK 0x1
+#define QIB_7322_IntClear_SDmaCleanupDoneClear_1_LSB 0x39
+#define QIB_7322_IntClear_SDmaCleanupDoneClear_1_MSB 0x39
+#define QIB_7322_IntClear_SDmaCleanupDoneClear_1_RMASK 0x1
+#define QIB_7322_IntClear_SDmaCleanupDoneClear_0_LSB 0x38
+#define QIB_7322_IntClear_SDmaCleanupDoneClear_0_MSB 0x38
+#define QIB_7322_IntClear_SDmaCleanupDoneClear_0_RMASK 0x1
+#define QIB_7322_IntClear_RcvUrg17IntClear_LSB 0x31
+#define QIB_7322_IntClear_RcvUrg17IntClear_MSB 0x31
+#define QIB_7322_IntClear_RcvUrg17IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvUrg16IntClear_LSB 0x30
+#define QIB_7322_IntClear_RcvUrg16IntClear_MSB 0x30
+#define QIB_7322_IntClear_RcvUrg16IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvUrg15IntClear_LSB 0x2F
+#define QIB_7322_IntClear_RcvUrg15IntClear_MSB 0x2F
+#define QIB_7322_IntClear_RcvUrg15IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvUrg14IntClear_LSB 0x2E
+#define QIB_7322_IntClear_RcvUrg14IntClear_MSB 0x2E
+#define QIB_7322_IntClear_RcvUrg14IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvUrg13IntClear_LSB 0x2D
+#define QIB_7322_IntClear_RcvUrg13IntClear_MSB 0x2D
+#define QIB_7322_IntClear_RcvUrg13IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvUrg12IntClear_LSB 0x2C
+#define QIB_7322_IntClear_RcvUrg12IntClear_MSB 0x2C
+#define QIB_7322_IntClear_RcvUrg12IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvUrg11IntClear_LSB 0x2B
+#define QIB_7322_IntClear_RcvUrg11IntClear_MSB 0x2B
+#define QIB_7322_IntClear_RcvUrg11IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvUrg10IntClear_LSB 0x2A
+#define QIB_7322_IntClear_RcvUrg10IntClear_MSB 0x2A
+#define QIB_7322_IntClear_RcvUrg10IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvUrg9IntClear_LSB 0x29
+#define QIB_7322_IntClear_RcvUrg9IntClear_MSB 0x29
+#define QIB_7322_IntClear_RcvUrg9IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvUrg8IntClear_LSB 0x28
+#define QIB_7322_IntClear_RcvUrg8IntClear_MSB 0x28
+#define QIB_7322_IntClear_RcvUrg8IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvUrg7IntClear_LSB 0x27
+#define QIB_7322_IntClear_RcvUrg7IntClear_MSB 0x27
+#define QIB_7322_IntClear_RcvUrg7IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvUrg6IntClear_LSB 0x26
+#define QIB_7322_IntClear_RcvUrg6IntClear_MSB 0x26
+#define QIB_7322_IntClear_RcvUrg6IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvUrg5IntClear_LSB 0x25
+#define QIB_7322_IntClear_RcvUrg5IntClear_MSB 0x25
+#define QIB_7322_IntClear_RcvUrg5IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvUrg4IntClear_LSB 0x24
+#define QIB_7322_IntClear_RcvUrg4IntClear_MSB 0x24
+#define QIB_7322_IntClear_RcvUrg4IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvUrg3IntClear_LSB 0x23
+#define QIB_7322_IntClear_RcvUrg3IntClear_MSB 0x23
+#define QIB_7322_IntClear_RcvUrg3IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvUrg2IntClear_LSB 0x22
+#define QIB_7322_IntClear_RcvUrg2IntClear_MSB 0x22
+#define QIB_7322_IntClear_RcvUrg2IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvUrg1IntClear_LSB 0x21
+#define QIB_7322_IntClear_RcvUrg1IntClear_MSB 0x21
+#define QIB_7322_IntClear_RcvUrg1IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvUrg0IntClear_LSB 0x20
+#define QIB_7322_IntClear_RcvUrg0IntClear_MSB 0x20
+#define QIB_7322_IntClear_RcvUrg0IntClear_RMASK 0x1
+#define QIB_7322_IntClear_ErrIntClear_1_LSB 0x1F
+#define QIB_7322_IntClear_ErrIntClear_1_MSB 0x1F
+#define QIB_7322_IntClear_ErrIntClear_1_RMASK 0x1
+#define QIB_7322_IntClear_ErrIntClear_0_LSB 0x1E
+#define QIB_7322_IntClear_ErrIntClear_0_MSB 0x1E
+#define QIB_7322_IntClear_ErrIntClear_0_RMASK 0x1
+#define QIB_7322_IntClear_ErrIntClear_LSB 0x1D
+#define QIB_7322_IntClear_ErrIntClear_MSB 0x1D
+#define QIB_7322_IntClear_ErrIntClear_RMASK 0x1
+#define QIB_7322_IntClear_AssertGPIOIntClear_LSB 0x1C
+#define QIB_7322_IntClear_AssertGPIOIntClear_MSB 0x1C
+#define QIB_7322_IntClear_AssertGPIOIntClear_RMASK 0x1
+#define QIB_7322_IntClear_SendDoneIntClear_1_LSB 0x19
+#define QIB_7322_IntClear_SendDoneIntClear_1_MSB 0x19
+#define QIB_7322_IntClear_SendDoneIntClear_1_RMASK 0x1
+#define QIB_7322_IntClear_SendDoneIntClear_0_LSB 0x18
+#define QIB_7322_IntClear_SendDoneIntClear_0_MSB 0x18
+#define QIB_7322_IntClear_SendDoneIntClear_0_RMASK 0x1
+#define QIB_7322_IntClear_SendBufAvailIntClear_LSB 0x17
+#define QIB_7322_IntClear_SendBufAvailIntClear_MSB 0x17
+#define QIB_7322_IntClear_SendBufAvailIntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvAvail17IntClear_LSB 0x11
+#define QIB_7322_IntClear_RcvAvail17IntClear_MSB 0x11
+#define QIB_7322_IntClear_RcvAvail17IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvAvail16IntClear_LSB 0x10
+#define QIB_7322_IntClear_RcvAvail16IntClear_MSB 0x10
+#define QIB_7322_IntClear_RcvAvail16IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvAvail15IntClear_LSB 0xF
+#define QIB_7322_IntClear_RcvAvail15IntClear_MSB 0xF
+#define QIB_7322_IntClear_RcvAvail15IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvAvail14IntClear_LSB 0xE
+#define QIB_7322_IntClear_RcvAvail14IntClear_MSB 0xE
+#define QIB_7322_IntClear_RcvAvail14IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvAvail13IntClear_LSB 0xD
+#define QIB_7322_IntClear_RcvAvail13IntClear_MSB 0xD
+#define QIB_7322_IntClear_RcvAvail13IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvAvail12IntClear_LSB 0xC
+#define QIB_7322_IntClear_RcvAvail12IntClear_MSB 0xC
+#define QIB_7322_IntClear_RcvAvail12IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvAvail11IntClear_LSB 0xB
+#define QIB_7322_IntClear_RcvAvail11IntClear_MSB 0xB
+#define QIB_7322_IntClear_RcvAvail11IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvAvail10IntClear_LSB 0xA
+#define QIB_7322_IntClear_RcvAvail10IntClear_MSB 0xA
+#define QIB_7322_IntClear_RcvAvail10IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvAvail9IntClear_LSB 0x9
+#define QIB_7322_IntClear_RcvAvail9IntClear_MSB 0x9
+#define QIB_7322_IntClear_RcvAvail9IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvAvail8IntClear_LSB 0x8
+#define QIB_7322_IntClear_RcvAvail8IntClear_MSB 0x8
+#define QIB_7322_IntClear_RcvAvail8IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvAvail7IntClear_LSB 0x7
+#define QIB_7322_IntClear_RcvAvail7IntClear_MSB 0x7
+#define QIB_7322_IntClear_RcvAvail7IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvAvail6IntClear_LSB 0x6
+#define QIB_7322_IntClear_RcvAvail6IntClear_MSB 0x6
+#define QIB_7322_IntClear_RcvAvail6IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvAvail5IntClear_LSB 0x5
+#define QIB_7322_IntClear_RcvAvail5IntClear_MSB 0x5
+#define QIB_7322_IntClear_RcvAvail5IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvAvail4IntClear_LSB 0x4
+#define QIB_7322_IntClear_RcvAvail4IntClear_MSB 0x4
+#define QIB_7322_IntClear_RcvAvail4IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvAvail3IntClear_LSB 0x3
+#define QIB_7322_IntClear_RcvAvail3IntClear_MSB 0x3
+#define QIB_7322_IntClear_RcvAvail3IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvAvail2IntClear_LSB 0x2
+#define QIB_7322_IntClear_RcvAvail2IntClear_MSB 0x2
+#define QIB_7322_IntClear_RcvAvail2IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvAvail1IntClear_LSB 0x1
+#define QIB_7322_IntClear_RcvAvail1IntClear_MSB 0x1
+#define QIB_7322_IntClear_RcvAvail1IntClear_RMASK 0x1
+#define QIB_7322_IntClear_RcvAvail0IntClear_LSB 0x0
+#define QIB_7322_IntClear_RcvAvail0IntClear_MSB 0x0
+#define QIB_7322_IntClear_RcvAvail0IntClear_RMASK 0x1
+
+#define QIB_7322_ErrMask_OFFS 0x80
+#define QIB_7322_ErrMask_DEF 0x0000000000000000
+#define QIB_7322_ErrMask_ResetNegatedMask_LSB 0x3F
+#define QIB_7322_ErrMask_ResetNegatedMask_MSB 0x3F
+#define QIB_7322_ErrMask_ResetNegatedMask_RMASK 0x1
+#define QIB_7322_ErrMask_HardwareErrMask_LSB 0x3E
+#define QIB_7322_ErrMask_HardwareErrMask_MSB 0x3E
+#define QIB_7322_ErrMask_HardwareErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_InvalidAddrErrMask_LSB 0x3D
+#define QIB_7322_ErrMask_InvalidAddrErrMask_MSB 0x3D
+#define QIB_7322_ErrMask_InvalidAddrErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_SDmaVL15ErrMask_LSB 0x38
+#define QIB_7322_ErrMask_SDmaVL15ErrMask_MSB 0x38
+#define QIB_7322_ErrMask_SDmaVL15ErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_SBufVL15MisUseErrMask_LSB 0x37
+#define QIB_7322_ErrMask_SBufVL15MisUseErrMask_MSB 0x37
+#define QIB_7322_ErrMask_SBufVL15MisUseErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_InvalidEEPCmdMask_LSB 0x35
+#define QIB_7322_ErrMask_InvalidEEPCmdMask_MSB 0x35
+#define QIB_7322_ErrMask_InvalidEEPCmdMask_RMASK 0x1
+#define QIB_7322_ErrMask_RcvContextShareErrMask_LSB 0x34
+#define QIB_7322_ErrMask_RcvContextShareErrMask_MSB 0x34
+#define QIB_7322_ErrMask_RcvContextShareErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_SendVLMismatchErrMask_LSB 0x24
+#define QIB_7322_ErrMask_SendVLMismatchErrMask_MSB 0x24
+#define QIB_7322_ErrMask_SendVLMismatchErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_SendArmLaunchErrMask_LSB 0x23
+#define QIB_7322_ErrMask_SendArmLaunchErrMask_MSB 0x23
+#define QIB_7322_ErrMask_SendArmLaunchErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_SendSpecialTriggerErrMask_LSB 0x1B
+#define QIB_7322_ErrMask_SendSpecialTriggerErrMask_MSB 0x1B
+#define QIB_7322_ErrMask_SendSpecialTriggerErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_SDmaWrongPortErrMask_LSB 0x1A
+#define QIB_7322_ErrMask_SDmaWrongPortErrMask_MSB 0x1A
+#define QIB_7322_ErrMask_SDmaWrongPortErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_SDmaBufMaskDuplicateErrMask_LSB 0x19
+#define QIB_7322_ErrMask_SDmaBufMaskDuplicateErrMask_MSB 0x19
+#define QIB_7322_ErrMask_SDmaBufMaskDuplicateErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_RcvHdrFullErrMask_LSB 0xD
+#define QIB_7322_ErrMask_RcvHdrFullErrMask_MSB 0xD
+#define QIB_7322_ErrMask_RcvHdrFullErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_RcvEgrFullErrMask_LSB 0xC
+#define QIB_7322_ErrMask_RcvEgrFullErrMask_MSB 0xC
+#define QIB_7322_ErrMask_RcvEgrFullErrMask_RMASK 0x1
+
+#define QIB_7322_ErrStatus_OFFS 0x88
+#define QIB_7322_ErrStatus_DEF 0x0000000000000000
+#define QIB_7322_ErrStatus_ResetNegated_LSB 0x3F
+#define QIB_7322_ErrStatus_ResetNegated_MSB 0x3F
+#define QIB_7322_ErrStatus_ResetNegated_RMASK 0x1
+#define QIB_7322_ErrStatus_HardwareErr_LSB 0x3E
+#define QIB_7322_ErrStatus_HardwareErr_MSB 0x3E
+#define QIB_7322_ErrStatus_HardwareErr_RMASK 0x1
+#define QIB_7322_ErrStatus_InvalidAddrErr_LSB 0x3D
+#define QIB_7322_ErrStatus_InvalidAddrErr_MSB 0x3D
+#define QIB_7322_ErrStatus_InvalidAddrErr_RMASK 0x1
+#define QIB_7322_ErrStatus_SDmaVL15Err_LSB 0x38
+#define QIB_7322_ErrStatus_SDmaVL15Err_MSB 0x38
+#define QIB_7322_ErrStatus_SDmaVL15Err_RMASK 0x1
+#define QIB_7322_ErrStatus_SBufVL15MisUseErr_LSB 0x37
+#define QIB_7322_ErrStatus_SBufVL15MisUseErr_MSB 0x37
+#define QIB_7322_ErrStatus_SBufVL15MisUseErr_RMASK 0x1
+#define QIB_7322_ErrStatus_InvalidEEPCmdErr_LSB 0x35
+#define QIB_7322_ErrStatus_InvalidEEPCmdErr_MSB 0x35
+#define QIB_7322_ErrStatus_InvalidEEPCmdErr_RMASK 0x1
+#define QIB_7322_ErrStatus_RcvContextShareErr_LSB 0x34
+#define QIB_7322_ErrStatus_RcvContextShareErr_MSB 0x34
+#define QIB_7322_ErrStatus_RcvContextShareErr_RMASK 0x1
+#define QIB_7322_ErrStatus_SendVLMismatchErr_LSB 0x24
+#define QIB_7322_ErrStatus_SendVLMismatchErr_MSB 0x24
+#define QIB_7322_ErrStatus_SendVLMismatchErr_RMASK 0x1
+#define QIB_7322_ErrStatus_SendArmLaunchErr_LSB 0x23
+#define QIB_7322_ErrStatus_SendArmLaunchErr_MSB 0x23
+#define QIB_7322_ErrStatus_SendArmLaunchErr_RMASK 0x1
+#define QIB_7322_ErrStatus_SendSpecialTriggerErr_LSB 0x1B
+#define QIB_7322_ErrStatus_SendSpecialTriggerErr_MSB 0x1B
+#define QIB_7322_ErrStatus_SendSpecialTriggerErr_RMASK 0x1
+#define QIB_7322_ErrStatus_SDmaWrongPortErr_LSB 0x1A
+#define QIB_7322_ErrStatus_SDmaWrongPortErr_MSB 0x1A
+#define QIB_7322_ErrStatus_SDmaWrongPortErr_RMASK 0x1
+#define QIB_7322_ErrStatus_SDmaBufMaskDuplicateErr_LSB 0x19
+#define QIB_7322_ErrStatus_SDmaBufMaskDuplicateErr_MSB 0x19
+#define QIB_7322_ErrStatus_SDmaBufMaskDuplicateErr_RMASK 0x1
+#define QIB_7322_ErrStatus_RcvHdrFullErr_LSB 0xD
+#define QIB_7322_ErrStatus_RcvHdrFullErr_MSB 0xD
+#define QIB_7322_ErrStatus_RcvHdrFullErr_RMASK 0x1
+#define QIB_7322_ErrStatus_RcvEgrFullErr_LSB 0xC
+#define QIB_7322_ErrStatus_RcvEgrFullErr_MSB 0xC
+#define QIB_7322_ErrStatus_RcvEgrFullErr_RMASK 0x1
+
+#define QIB_7322_ErrClear_OFFS 0x90
+#define QIB_7322_ErrClear_DEF 0x0000000000000000
+#define QIB_7322_ErrClear_ResetNegatedClear_LSB 0x3F
+#define QIB_7322_ErrClear_ResetNegatedClear_MSB 0x3F
+#define QIB_7322_ErrClear_ResetNegatedClear_RMASK 0x1
+#define QIB_7322_ErrClear_HardwareErrClear_LSB 0x3E
+#define QIB_7322_ErrClear_HardwareErrClear_MSB 0x3E
+#define QIB_7322_ErrClear_HardwareErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_InvalidAddrErrClear_LSB 0x3D
+#define QIB_7322_ErrClear_InvalidAddrErrClear_MSB 0x3D
+#define QIB_7322_ErrClear_InvalidAddrErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_SDmaVL15ErrClear_LSB 0x38
+#define QIB_7322_ErrClear_SDmaVL15ErrClear_MSB 0x38
+#define QIB_7322_ErrClear_SDmaVL15ErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_SBufVL15MisUseErrClear_LSB 0x37
+#define QIB_7322_ErrClear_SBufVL15MisUseErrClear_MSB 0x37
+#define QIB_7322_ErrClear_SBufVL15MisUseErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_InvalidEEPCmdErrClear_LSB 0x35
+#define QIB_7322_ErrClear_InvalidEEPCmdErrClear_MSB 0x35
+#define QIB_7322_ErrClear_InvalidEEPCmdErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_RcvContextShareErrClear_LSB 0x34
+#define QIB_7322_ErrClear_RcvContextShareErrClear_MSB 0x34
+#define QIB_7322_ErrClear_RcvContextShareErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_SendVLMismatchErrMask_LSB 0x24
+#define QIB_7322_ErrClear_SendVLMismatchErrMask_MSB 0x24
+#define QIB_7322_ErrClear_SendVLMismatchErrMask_RMASK 0x1
+#define QIB_7322_ErrClear_SendArmLaunchErrClear_LSB 0x23
+#define QIB_7322_ErrClear_SendArmLaunchErrClear_MSB 0x23
+#define QIB_7322_ErrClear_SendArmLaunchErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_SendSpecialTriggerErrClear_LSB 0x1B
+#define QIB_7322_ErrClear_SendSpecialTriggerErrClear_MSB 0x1B
+#define QIB_7322_ErrClear_SendSpecialTriggerErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_SDmaWrongPortErrClear_LSB 0x1A
+#define QIB_7322_ErrClear_SDmaWrongPortErrClear_MSB 0x1A
+#define QIB_7322_ErrClear_SDmaWrongPortErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_SDmaBufMaskDuplicateErrClear_LSB 0x19
+#define QIB_7322_ErrClear_SDmaBufMaskDuplicateErrClear_MSB 0x19
+#define QIB_7322_ErrClear_SDmaBufMaskDuplicateErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_RcvHdrFullErrClear_LSB 0xD
+#define QIB_7322_ErrClear_RcvHdrFullErrClear_MSB 0xD
+#define QIB_7322_ErrClear_RcvHdrFullErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_RcvEgrFullErrClear_LSB 0xC
+#define QIB_7322_ErrClear_RcvEgrFullErrClear_MSB 0xC
+#define QIB_7322_ErrClear_RcvEgrFullErrClear_RMASK 0x1
+
+#define QIB_7322_HwErrMask_OFFS 0x98
+#define QIB_7322_HwErrMask_DEF 0x0000000000000000
+#define QIB_7322_HwErrMask_IBSerdesPClkNotDetectMask_1_LSB 0x3F
+#define QIB_7322_HwErrMask_IBSerdesPClkNotDetectMask_1_MSB 0x3F
+#define QIB_7322_HwErrMask_IBSerdesPClkNotDetectMask_1_RMASK 0x1
+#define QIB_7322_HwErrMask_IBSerdesPClkNotDetectMask_0_LSB 0x3E
+#define QIB_7322_HwErrMask_IBSerdesPClkNotDetectMask_0_MSB 0x3E
+#define QIB_7322_HwErrMask_IBSerdesPClkNotDetectMask_0_RMASK 0x1
+#define QIB_7322_HwErrMask_PCIESerdesPClkNotDetectMask_LSB 0x37
+#define QIB_7322_HwErrMask_PCIESerdesPClkNotDetectMask_MSB 0x37
+#define QIB_7322_HwErrMask_PCIESerdesPClkNotDetectMask_RMASK 0x1
+#define QIB_7322_HwErrMask_PowerOnBISTFailedMask_LSB 0x36
+#define QIB_7322_HwErrMask_PowerOnBISTFailedMask_MSB 0x36
+#define QIB_7322_HwErrMask_PowerOnBISTFailedMask_RMASK 0x1
+#define QIB_7322_HwErrMask_TempsenseTholdReachedMask_LSB 0x35
+#define QIB_7322_HwErrMask_TempsenseTholdReachedMask_MSB 0x35
+#define QIB_7322_HwErrMask_TempsenseTholdReachedMask_RMASK 0x1
+#define QIB_7322_HwErrMask_MemoryErrMask_LSB 0x30
+#define QIB_7322_HwErrMask_MemoryErrMask_MSB 0x30
+#define QIB_7322_HwErrMask_MemoryErrMask_RMASK 0x1
+#define QIB_7322_HwErrMask_pcie_phy_txParityErr_LSB 0x22
+#define QIB_7322_HwErrMask_pcie_phy_txParityErr_MSB 0x22
+#define QIB_7322_HwErrMask_pcie_phy_txParityErr_RMASK 0x1
+#define QIB_7322_HwErrMask_PCIeBusParityErrMask_LSB 0x1F
+#define QIB_7322_HwErrMask_PCIeBusParityErrMask_MSB 0x21
+#define QIB_7322_HwErrMask_PCIeBusParityErrMask_RMASK 0x7
+#define QIB_7322_HwErrMask_PcieCplTimeoutMask_LSB 0x1E
+#define QIB_7322_HwErrMask_PcieCplTimeoutMask_MSB 0x1E
+#define QIB_7322_HwErrMask_PcieCplTimeoutMask_RMASK 0x1
+#define QIB_7322_HwErrMask_PciePoisonedTLPMask_LSB 0x1D
+#define QIB_7322_HwErrMask_PciePoisonedTLPMask_MSB 0x1D
+#define QIB_7322_HwErrMask_PciePoisonedTLPMask_RMASK 0x1
+#define QIB_7322_HwErrMask_SDmaMemReadErrMask_1_LSB 0x1C
+#define QIB_7322_HwErrMask_SDmaMemReadErrMask_1_MSB 0x1C
+#define QIB_7322_HwErrMask_SDmaMemReadErrMask_1_RMASK 0x1
+#define QIB_7322_HwErrMask_SDmaMemReadErrMask_0_LSB 0x1B
+#define QIB_7322_HwErrMask_SDmaMemReadErrMask_0_MSB 0x1B
+#define QIB_7322_HwErrMask_SDmaMemReadErrMask_0_RMASK 0x1
+#define QIB_7322_HwErrMask_IBCBusFromSPCParityErrMask_1_LSB 0xF
+#define QIB_7322_HwErrMask_IBCBusFromSPCParityErrMask_1_MSB 0xF
+#define QIB_7322_HwErrMask_IBCBusFromSPCParityErrMask_1_RMASK 0x1
+#define QIB_7322_HwErrMask_statusValidNoEopMask_1_LSB 0xE
+#define QIB_7322_HwErrMask_statusValidNoEopMask_1_MSB 0xE
+#define QIB_7322_HwErrMask_statusValidNoEopMask_1_RMASK 0x1
+#define QIB_7322_HwErrMask_IBCBusFromSPCParityErrMask_0_LSB 0xD
+#define QIB_7322_HwErrMask_IBCBusFromSPCParityErrMask_0_MSB 0xD
+#define QIB_7322_HwErrMask_IBCBusFromSPCParityErrMask_0_RMASK 0x1
+#define QIB_7322_HwErrMask_statusValidNoEopMask_0_LSB 0xC
+#define QIB_7322_HwErrMask_statusValidNoEopMask_0_MSB 0xC
+#define QIB_7322_HwErrMask_statusValidNoEopMask_0_RMASK 0x1
+#define QIB_7322_HwErrMask_LATriggeredMask_LSB 0xB
+#define QIB_7322_HwErrMask_LATriggeredMask_MSB 0xB
+#define QIB_7322_HwErrMask_LATriggeredMask_RMASK 0x1
+
+#define QIB_7322_HwErrStatus_OFFS 0xA0
+#define QIB_7322_HwErrStatus_DEF 0x0000000000000000
+#define QIB_7322_HwErrStatus_IBSerdesPClkNotDetect_1_LSB 0x3F
+#define QIB_7322_HwErrStatus_IBSerdesPClkNotDetect_1_MSB 0x3F
+#define QIB_7322_HwErrStatus_IBSerdesPClkNotDetect_1_RMASK 0x1
+#define QIB_7322_HwErrStatus_IBSerdesPClkNotDetect_0_LSB 0x3E
+#define QIB_7322_HwErrStatus_IBSerdesPClkNotDetect_0_MSB 0x3E
+#define QIB_7322_HwErrStatus_IBSerdesPClkNotDetect_0_RMASK 0x1
+#define QIB_7322_HwErrStatus_PCIESerdesPClkNotDetect_LSB 0x37
+#define QIB_7322_HwErrStatus_PCIESerdesPClkNotDetect_MSB 0x37
+#define QIB_7322_HwErrStatus_PCIESerdesPClkNotDetect_RMASK 0x1
+#define QIB_7322_HwErrStatus_PowerOnBISTFailed_LSB 0x36
+#define QIB_7322_HwErrStatus_PowerOnBISTFailed_MSB 0x36
+#define QIB_7322_HwErrStatus_PowerOnBISTFailed_RMASK 0x1
+#define QIB_7322_HwErrStatus_TempsenseTholdReached_LSB 0x35
+#define QIB_7322_HwErrStatus_TempsenseTholdReached_MSB 0x35
+#define QIB_7322_HwErrStatus_TempsenseTholdReached_RMASK 0x1
+#define QIB_7322_HwErrStatus_MemoryErr_LSB 0x30
+#define QIB_7322_HwErrStatus_MemoryErr_MSB 0x30
+#define QIB_7322_HwErrStatus_MemoryErr_RMASK 0x1
+#define QIB_7322_HwErrStatus_pcie_phy_txParityErr_LSB 0x22
+#define QIB_7322_HwErrStatus_pcie_phy_txParityErr_MSB 0x22
+#define QIB_7322_HwErrStatus_pcie_phy_txParityErr_RMASK 0x1
+#define QIB_7322_HwErrStatus_PCIeBusParity_LSB 0x1F
+#define QIB_7322_HwErrStatus_PCIeBusParity_MSB 0x21
+#define QIB_7322_HwErrStatus_PCIeBusParity_RMASK 0x7
+#define QIB_7322_HwErrStatus_PcieCplTimeout_LSB 0x1E
+#define QIB_7322_HwErrStatus_PcieCplTimeout_MSB 0x1E
+#define QIB_7322_HwErrStatus_PcieCplTimeout_RMASK 0x1
+#define QIB_7322_HwErrStatus_PciePoisonedTLP_LSB 0x1D
+#define QIB_7322_HwErrStatus_PciePoisonedTLP_MSB 0x1D
+#define QIB_7322_HwErrStatus_PciePoisonedTLP_RMASK 0x1
+#define QIB_7322_HwErrStatus_SDmaMemReadErr_1_LSB 0x1C
+#define QIB_7322_HwErrStatus_SDmaMemReadErr_1_MSB 0x1C
+#define QIB_7322_HwErrStatus_SDmaMemReadErr_1_RMASK 0x1
+#define QIB_7322_HwErrStatus_SDmaMemReadErr_0_LSB 0x1B
+#define QIB_7322_HwErrStatus_SDmaMemReadErr_0_MSB 0x1B
+#define QIB_7322_HwErrStatus_SDmaMemReadErr_0_RMASK 0x1
+#define QIB_7322_HwErrStatus_IBCBusFromSPCParityErr_1_LSB 0xF
+#define QIB_7322_HwErrStatus_IBCBusFromSPCParityErr_1_MSB 0xF
+#define QIB_7322_HwErrStatus_IBCBusFromSPCParityErr_1_RMASK 0x1
+#define QIB_7322_HwErrStatus_statusValidNoEop_1_LSB 0xE
+#define QIB_7322_HwErrStatus_statusValidNoEop_1_MSB 0xE
+#define QIB_7322_HwErrStatus_statusValidNoEop_1_RMASK 0x1
+#define QIB_7322_HwErrStatus_IBCBusFromSPCParityErr_0_LSB 0xD
+#define QIB_7322_HwErrStatus_IBCBusFromSPCParityErr_0_MSB 0xD
+#define QIB_7322_HwErrStatus_IBCBusFromSPCParityErr_0_RMASK 0x1
+#define QIB_7322_HwErrStatus_statusValidNoEop_0_LSB 0xC
+#define QIB_7322_HwErrStatus_statusValidNoEop_0_MSB 0xC
+#define QIB_7322_HwErrStatus_statusValidNoEop_0_RMASK 0x1
+#define QIB_7322_HwErrStatus_LATriggered_LSB 0xB
+#define QIB_7322_HwErrStatus_LATriggered_MSB 0xB
+#define QIB_7322_HwErrStatus_LATriggered_RMASK 0x1
+
+#define QIB_7322_HwErrClear_OFFS 0xA8
+#define QIB_7322_HwErrClear_DEF 0x0000000000000000
+#define QIB_7322_HwErrClear_IBSerdesPClkNotDetectClear_1_LSB 0x3F
+#define QIB_7322_HwErrClear_IBSerdesPClkNotDetectClear_1_MSB 0x3F
+#define QIB_7322_HwErrClear_IBSerdesPClkNotDetectClear_1_RMASK 0x1
+#define QIB_7322_HwErrClear_IBSerdesPClkNotDetectClear_0_LSB 0x3E
+#define QIB_7322_HwErrClear_IBSerdesPClkNotDetectClear_0_MSB 0x3E
+#define QIB_7322_HwErrClear_IBSerdesPClkNotDetectClear_0_RMASK 0x1
+#define QIB_7322_HwErrClear_PCIESerdesPClkNotDetectClear_LSB 0x37
+#define QIB_7322_HwErrClear_PCIESerdesPClkNotDetectClear_MSB 0x37
+#define QIB_7322_HwErrClear_PCIESerdesPClkNotDetectClear_RMASK 0x1
+#define QIB_7322_HwErrClear_PowerOnBISTFailedClear_LSB 0x36
+#define QIB_7322_HwErrClear_PowerOnBISTFailedClear_MSB 0x36
+#define QIB_7322_HwErrClear_PowerOnBISTFailedClear_RMASK 0x1
+#define QIB_7322_HwErrClear_TempsenseTholdReachedClear_LSB 0x35
+#define QIB_7322_HwErrClear_TempsenseTholdReachedClear_MSB 0x35
+#define QIB_7322_HwErrClear_TempsenseTholdReachedClear_RMASK 0x1
+#define QIB_7322_HwErrClear_MemoryErrClear_LSB 0x30
+#define QIB_7322_HwErrClear_MemoryErrClear_MSB 0x30
+#define QIB_7322_HwErrClear_MemoryErrClear_RMASK 0x1
+#define QIB_7322_HwErrClear_pcie_phy_txParityErr_LSB 0x22
+#define QIB_7322_HwErrClear_pcie_phy_txParityErr_MSB 0x22
+#define QIB_7322_HwErrClear_pcie_phy_txParityErr_RMASK 0x1
+#define QIB_7322_HwErrClear_PCIeBusParityClear_LSB 0x1F
+#define QIB_7322_HwErrClear_PCIeBusParityClear_MSB 0x21
+#define QIB_7322_HwErrClear_PCIeBusParityClear_RMASK 0x7
+#define QIB_7322_HwErrClear_PcieCplTimeoutClear_LSB 0x1E
+#define QIB_7322_HwErrClear_PcieCplTimeoutClear_MSB 0x1E
+#define QIB_7322_HwErrClear_PcieCplTimeoutClear_RMASK 0x1
+#define QIB_7322_HwErrClear_PciePoisonedTLPClear_LSB 0x1D
+#define QIB_7322_HwErrClear_PciePoisonedTLPClear_MSB 0x1D
+#define QIB_7322_HwErrClear_PciePoisonedTLPClear_RMASK 0x1
+#define QIB_7322_HwErrClear_SDmaMemReadErrClear_1_LSB 0x1C
+#define QIB_7322_HwErrClear_SDmaMemReadErrClear_1_MSB 0x1C
+#define QIB_7322_HwErrClear_SDmaMemReadErrClear_1_RMASK 0x1
+#define QIB_7322_HwErrClear_SDmaMemReadErrClear_0_LSB 0x1B
+#define QIB_7322_HwErrClear_SDmaMemReadErrClear_0_MSB 0x1B
+#define QIB_7322_HwErrClear_SDmaMemReadErrClear_0_RMASK 0x1
+#define QIB_7322_HwErrClear_IBCBusFromSPCParityErrClear_1_LSB 0xF
+#define QIB_7322_HwErrClear_IBCBusFromSPCParityErrClear_1_MSB 0xF
+#define QIB_7322_HwErrClear_IBCBusFromSPCParityErrClear_1_RMASK 0x1
+#define QIB_7322_HwErrClear_IBCBusToSPCparityErrClear_1_LSB 0xE
+#define QIB_7322_HwErrClear_IBCBusToSPCparityErrClear_1_MSB 0xE
+#define QIB_7322_HwErrClear_IBCBusToSPCparityErrClear_1_RMASK 0x1
+#define QIB_7322_HwErrClear_IBCBusFromSPCParityErrClear_0_LSB 0xD
+#define QIB_7322_HwErrClear_IBCBusFromSPCParityErrClear_0_MSB 0xD
+#define QIB_7322_HwErrClear_IBCBusFromSPCParityErrClear_0_RMASK 0x1
+#define QIB_7322_HwErrClear_IBCBusToSPCparityErrClear_0_LSB 0xC
+#define QIB_7322_HwErrClear_IBCBusToSPCparityErrClear_0_MSB 0xC
+#define QIB_7322_HwErrClear_IBCBusToSPCparityErrClear_0_RMASK 0x1
+#define QIB_7322_HwErrClear_LATriggeredClear_LSB 0xB
+#define QIB_7322_HwErrClear_LATriggeredClear_MSB 0xB
+#define QIB_7322_HwErrClear_LATriggeredClear_RMASK 0x1
+
+#define QIB_7322_HwDiagCtrl_OFFS 0xB0
+#define QIB_7322_HwDiagCtrl_DEF 0x0000000000000000
+#define QIB_7322_HwDiagCtrl_Diagnostic_LSB 0x3F
+#define QIB_7322_HwDiagCtrl_Diagnostic_MSB 0x3F
+#define QIB_7322_HwDiagCtrl_Diagnostic_RMASK 0x1
+#define QIB_7322_HwDiagCtrl_CounterWrEnable_LSB 0x3D
+#define QIB_7322_HwDiagCtrl_CounterWrEnable_MSB 0x3D
+#define QIB_7322_HwDiagCtrl_CounterWrEnable_RMASK 0x1
+#define QIB_7322_HwDiagCtrl_CounterDisable_LSB 0x3C
+#define QIB_7322_HwDiagCtrl_CounterDisable_MSB 0x3C
+#define QIB_7322_HwDiagCtrl_CounterDisable_RMASK 0x1
+#define QIB_7322_HwDiagCtrl_forcePCIeBusParity_LSB 0x1F
+#define QIB_7322_HwDiagCtrl_forcePCIeBusParity_MSB 0x22
+#define QIB_7322_HwDiagCtrl_forcePCIeBusParity_RMASK 0xF
+#define QIB_7322_HwDiagCtrl_ForceIBCBusFromSPCParityErr_1_LSB 0xF
+#define QIB_7322_HwDiagCtrl_ForceIBCBusFromSPCParityErr_1_MSB 0xF
+#define QIB_7322_HwDiagCtrl_ForceIBCBusFromSPCParityErr_1_RMASK 0x1
+#define QIB_7322_HwDiagCtrl_ForcestatusValidNoEop_1_LSB 0xE
+#define QIB_7322_HwDiagCtrl_ForcestatusValidNoEop_1_MSB 0xE
+#define QIB_7322_HwDiagCtrl_ForcestatusValidNoEop_1_RMASK 0x1
+#define QIB_7322_HwDiagCtrl_ForceIBCBusFromSPCParityErr_0_LSB 0xD
+#define QIB_7322_HwDiagCtrl_ForceIBCBusFromSPCParityErr_0_MSB 0xD
+#define QIB_7322_HwDiagCtrl_ForceIBCBusFromSPCParityErr_0_RMASK 0x1
+#define QIB_7322_HwDiagCtrl_ForcestatusValidNoEop_0_LSB 0xC
+#define QIB_7322_HwDiagCtrl_ForcestatusValidNoEop_0_MSB 0xC
+#define QIB_7322_HwDiagCtrl_ForcestatusValidNoEop_0_RMASK 0x1
+
+#define QIB_7322_EXTStatus_OFFS 0xC0
+#define QIB_7322_EXTStatus_DEF 0x000000000000X000
+#define QIB_7322_EXTStatus_GPIOIn_LSB 0x30
+#define QIB_7322_EXTStatus_GPIOIn_MSB 0x3F
+#define QIB_7322_EXTStatus_GPIOIn_RMASK 0xFFFF
+#define QIB_7322_EXTStatus_MemBISTDisabled_LSB 0xF
+#define QIB_7322_EXTStatus_MemBISTDisabled_MSB 0xF
+#define QIB_7322_EXTStatus_MemBISTDisabled_RMASK 0x1
+#define QIB_7322_EXTStatus_MemBISTEndTest_LSB 0xE
+#define QIB_7322_EXTStatus_MemBISTEndTest_MSB 0xE
+#define QIB_7322_EXTStatus_MemBISTEndTest_RMASK 0x1
+
+#define QIB_7322_EXTCtrl_OFFS 0xC8
+#define QIB_7322_EXTCtrl_DEF 0x0000000000000000
+#define QIB_7322_EXTCtrl_GPIOOe_LSB 0x30
+#define QIB_7322_EXTCtrl_GPIOOe_MSB 0x3F
+#define QIB_7322_EXTCtrl_GPIOOe_RMASK 0xFFFF
+#define QIB_7322_EXTCtrl_GPIOInvert_LSB 0x20
+#define QIB_7322_EXTCtrl_GPIOInvert_MSB 0x2F
+#define QIB_7322_EXTCtrl_GPIOInvert_RMASK 0xFFFF
+#define QIB_7322_EXTCtrl_LEDPort1GreenOn_LSB 0x3
+#define QIB_7322_EXTCtrl_LEDPort1GreenOn_MSB 0x3
+#define QIB_7322_EXTCtrl_LEDPort1GreenOn_RMASK 0x1
+#define QIB_7322_EXTCtrl_LEDPort1YellowOn_LSB 0x2
+#define QIB_7322_EXTCtrl_LEDPort1YellowOn_MSB 0x2
+#define QIB_7322_EXTCtrl_LEDPort1YellowOn_RMASK 0x1
+#define QIB_7322_EXTCtrl_LEDPort0GreenOn_LSB 0x1
+#define QIB_7322_EXTCtrl_LEDPort0GreenOn_MSB 0x1
+#define QIB_7322_EXTCtrl_LEDPort0GreenOn_RMASK 0x1
+#define QIB_7322_EXTCtrl_LEDPort0YellowOn_LSB 0x0
+#define QIB_7322_EXTCtrl_LEDPort0YellowOn_MSB 0x0
+#define QIB_7322_EXTCtrl_LEDPort0YellowOn_RMASK 0x1
+
+#define QIB_7322_GPIOOut_OFFS 0xE0
+#define QIB_7322_GPIOOut_DEF 0x0000000000000000
+
+#define QIB_7322_GPIOMask_OFFS 0xE8
+#define QIB_7322_GPIOMask_DEF 0x0000000000000000
+
+#define QIB_7322_GPIOStatus_OFFS 0xF0
+#define QIB_7322_GPIOStatus_DEF 0x0000000000000000
+
+#define QIB_7322_GPIOClear_OFFS 0xF8
+#define QIB_7322_GPIOClear_DEF 0x0000000000000000
+
+#define QIB_7322_RcvCtrl_OFFS 0x100
+#define QIB_7322_RcvCtrl_DEF 0x0000000000000000
+#define QIB_7322_RcvCtrl_TidReDirect_LSB 0x30
+#define QIB_7322_RcvCtrl_TidReDirect_MSB 0x3F
+#define QIB_7322_RcvCtrl_TidReDirect_RMASK 0xFFFF
+#define QIB_7322_RcvCtrl_TailUpd_LSB 0x2F
+#define QIB_7322_RcvCtrl_TailUpd_MSB 0x2F
+#define QIB_7322_RcvCtrl_TailUpd_RMASK 0x1
+#define QIB_7322_RcvCtrl_XrcTypeCode_LSB 0x2C
+#define QIB_7322_RcvCtrl_XrcTypeCode_MSB 0x2E
+#define QIB_7322_RcvCtrl_XrcTypeCode_RMASK 0x7
+#define QIB_7322_RcvCtrl_TidFlowEnable_LSB 0x2B
+#define QIB_7322_RcvCtrl_TidFlowEnable_MSB 0x2B
+#define QIB_7322_RcvCtrl_TidFlowEnable_RMASK 0x1
+#define QIB_7322_RcvCtrl_ContextCfg_LSB 0x29
+#define QIB_7322_RcvCtrl_ContextCfg_MSB 0x2A
+#define QIB_7322_RcvCtrl_ContextCfg_RMASK 0x3
+#define QIB_7322_RcvCtrl_IntrAvail_LSB 0x14
+#define QIB_7322_RcvCtrl_IntrAvail_MSB 0x25
+#define QIB_7322_RcvCtrl_IntrAvail_RMASK 0x3FFFF
+#define QIB_7322_RcvCtrl_dontDropRHQFull_LSB 0x0
+#define QIB_7322_RcvCtrl_dontDropRHQFull_MSB 0x11
+#define QIB_7322_RcvCtrl_dontDropRHQFull_RMASK 0x3FFFF
+
+#define QIB_7322_RcvHdrSize_OFFS 0x110
+#define QIB_7322_RcvHdrSize_DEF 0x0000000000000000
+
+#define QIB_7322_RcvHdrCnt_OFFS 0x118
+#define QIB_7322_RcvHdrCnt_DEF 0x0000000000000000
+
+#define QIB_7322_RcvHdrEntSize_OFFS 0x120
+#define QIB_7322_RcvHdrEntSize_DEF 0x0000000000000000
+
+#define QIB_7322_RcvTIDBase_OFFS 0x128
+#define QIB_7322_RcvTIDBase_DEF 0x0000000000050000
+
+#define QIB_7322_RcvTIDCnt_OFFS 0x130
+#define QIB_7322_RcvTIDCnt_DEF 0x0000000000000200
+
+#define QIB_7322_RcvEgrBase_OFFS 0x138
+#define QIB_7322_RcvEgrBase_DEF 0x0000000000014000
+
+#define QIB_7322_RcvEgrCnt_OFFS 0x140
+#define QIB_7322_RcvEgrCnt_DEF 0x0000000000001000
+
+#define QIB_7322_RcvBufBase_OFFS 0x148
+#define QIB_7322_RcvBufBase_DEF 0x0000000000080000
+
+#define QIB_7322_RcvBufSize_OFFS 0x150
+#define QIB_7322_RcvBufSize_DEF 0x0000000000005000
+
+#define QIB_7322_RxIntMemBase_OFFS 0x158
+#define QIB_7322_RxIntMemBase_DEF 0x0000000000077000
+
+#define QIB_7322_RxIntMemSize_OFFS 0x160
+#define QIB_7322_RxIntMemSize_DEF 0x0000000000007000
+
+#define QIB_7322_feature_mask_OFFS 0x190
+#define QIB_7322_feature_mask_DEF 0x00000000000000XX
+
+#define QIB_7322_active_feature_mask_OFFS 0x198
+#define QIB_7322_active_feature_mask_DEF 0x00000000000000XX
+#define QIB_7322_active_feature_mask_Port1_QDR_Enabled_LSB 0x5
+#define QIB_7322_active_feature_mask_Port1_QDR_Enabled_MSB 0x5
+#define QIB_7322_active_feature_mask_Port1_QDR_Enabled_RMASK 0x1
+#define QIB_7322_active_feature_mask_Port1_DDR_Enabled_LSB 0x4
+#define QIB_7322_active_feature_mask_Port1_DDR_Enabled_MSB 0x4
+#define QIB_7322_active_feature_mask_Port1_DDR_Enabled_RMASK 0x1
+#define QIB_7322_active_feature_mask_Port1_SDR_Enabled_LSB 0x3
+#define QIB_7322_active_feature_mask_Port1_SDR_Enabled_MSB 0x3
+#define QIB_7322_active_feature_mask_Port1_SDR_Enabled_RMASK 0x1
+#define QIB_7322_active_feature_mask_Port0_QDR_Enabled_LSB 0x2
+#define QIB_7322_active_feature_mask_Port0_QDR_Enabled_MSB 0x2
+#define QIB_7322_active_feature_mask_Port0_QDR_Enabled_RMASK 0x1
+#define QIB_7322_active_feature_mask_Port0_DDR_Enabled_LSB 0x1
+#define QIB_7322_active_feature_mask_Port0_DDR_Enabled_MSB 0x1
+#define QIB_7322_active_feature_mask_Port0_DDR_Enabled_RMASK 0x1
+#define QIB_7322_active_feature_mask_Port0_SDR_Enabled_LSB 0x0
+#define QIB_7322_active_feature_mask_Port0_SDR_Enabled_MSB 0x0
+#define QIB_7322_active_feature_mask_Port0_SDR_Enabled_RMASK 0x1
+
+#define QIB_7322_SendCtrl_OFFS 0x1C0
+#define QIB_7322_SendCtrl_DEF 0x0000000000000000
+#define QIB_7322_SendCtrl_Disarm_LSB 0x1F
+#define QIB_7322_SendCtrl_Disarm_MSB 0x1F
+#define QIB_7322_SendCtrl_Disarm_RMASK 0x1
+#define QIB_7322_SendCtrl_SendBufAvailPad64Byte_LSB 0x1D
+#define QIB_7322_SendCtrl_SendBufAvailPad64Byte_MSB 0x1D
+#define QIB_7322_SendCtrl_SendBufAvailPad64Byte_RMASK 0x1
+#define QIB_7322_SendCtrl_AvailUpdThld_LSB 0x18
+#define QIB_7322_SendCtrl_AvailUpdThld_MSB 0x1C
+#define QIB_7322_SendCtrl_AvailUpdThld_RMASK 0x1F
+#define QIB_7322_SendCtrl_DisarmSendBuf_LSB 0x10
+#define QIB_7322_SendCtrl_DisarmSendBuf_MSB 0x17
+#define QIB_7322_SendCtrl_DisarmSendBuf_RMASK 0xFF
+#define QIB_7322_SendCtrl_SpecialTriggerEn_LSB 0x4
+#define QIB_7322_SendCtrl_SpecialTriggerEn_MSB 0x4
+#define QIB_7322_SendCtrl_SpecialTriggerEn_RMASK 0x1
+#define QIB_7322_SendCtrl_SendBufAvailUpd_LSB 0x2
+#define QIB_7322_SendCtrl_SendBufAvailUpd_MSB 0x2
+#define QIB_7322_SendCtrl_SendBufAvailUpd_RMASK 0x1
+#define QIB_7322_SendCtrl_SendIntBufAvail_LSB 0x1
+#define QIB_7322_SendCtrl_SendIntBufAvail_MSB 0x1
+#define QIB_7322_SendCtrl_SendIntBufAvail_RMASK 0x1
+
+#define QIB_7322_SendBufBase_OFFS 0x1C8
+#define QIB_7322_SendBufBase_DEF 0x0018000000100000
+#define QIB_7322_SendBufBase_BaseAddr_LargePIO_LSB 0x20
+#define QIB_7322_SendBufBase_BaseAddr_LargePIO_MSB 0x34
+#define QIB_7322_SendBufBase_BaseAddr_LargePIO_RMASK 0x1FFFFF
+#define QIB_7322_SendBufBase_BaseAddr_SmallPIO_LSB 0x0
+#define QIB_7322_SendBufBase_BaseAddr_SmallPIO_MSB 0x14
+#define QIB_7322_SendBufBase_BaseAddr_SmallPIO_RMASK 0x1FFFFF
+
+#define QIB_7322_SendBufSize_OFFS 0x1D0
+#define QIB_7322_SendBufSize_DEF 0x0000108000000880
+#define QIB_7322_SendBufSize_Size_LargePIO_LSB 0x20
+#define QIB_7322_SendBufSize_Size_LargePIO_MSB 0x2C
+#define QIB_7322_SendBufSize_Size_LargePIO_RMASK 0x1FFF
+#define QIB_7322_SendBufSize_Size_SmallPIO_LSB 0x0
+#define QIB_7322_SendBufSize_Size_SmallPIO_MSB 0xB
+#define QIB_7322_SendBufSize_Size_SmallPIO_RMASK 0xFFF
+
+#define QIB_7322_SendBufCnt_OFFS 0x1D8
+#define QIB_7322_SendBufCnt_DEF 0x0000002000000080
+#define QIB_7322_SendBufCnt_Num_LargeBuffers_LSB 0x20
+#define QIB_7322_SendBufCnt_Num_LargeBuffers_MSB 0x25
+#define QIB_7322_SendBufCnt_Num_LargeBuffers_RMASK 0x3F
+#define QIB_7322_SendBufCnt_Num_SmallBuffers_LSB 0x0
+#define QIB_7322_SendBufCnt_Num_SmallBuffers_MSB 0x8
+#define QIB_7322_SendBufCnt_Num_SmallBuffers_RMASK 0x1FF
+
+#define QIB_7322_SendBufAvailAddr_OFFS 0x1E0
+#define QIB_7322_SendBufAvailAddr_DEF 0x0000000000000000
+#define QIB_7322_SendBufAvailAddr_SendBufAvailAddr_LSB 0x6
+#define QIB_7322_SendBufAvailAddr_SendBufAvailAddr_MSB 0x27
+#define QIB_7322_SendBufAvailAddr_SendBufAvailAddr_RMASK 0x3FFFFFFFF
+
+#define QIB_7322_SendBufErr0_OFFS 0x240
+#define QIB_7322_SendBufErr0_DEF 0x0000000000000000
+#define QIB_7322_SendBufErr0_SendBufErr_63_0_LSB 0x0
+#define QIB_7322_SendBufErr0_SendBufErr_63_0_MSB 0x3F
+#define QIB_7322_SendBufErr0_SendBufErr_63_0_RMASK 0x0
+
+#define QIB_7322_AvailUpdCount_OFFS 0x268
+#define QIB_7322_AvailUpdCount_DEF 0x0000000000000000
+#define QIB_7322_AvailUpdCount_AvailUpdCount_LSB 0x0
+#define QIB_7322_AvailUpdCount_AvailUpdCount_MSB 0x4
+#define QIB_7322_AvailUpdCount_AvailUpdCount_RMASK 0x1F
+
+#define QIB_7322_RcvHdrAddr0_OFFS 0x280
+#define QIB_7322_RcvHdrAddr0_DEF 0x0000000000000000
+#define QIB_7322_RcvHdrAddr0_RcvHdrAddr_LSB 0x2
+#define QIB_7322_RcvHdrAddr0_RcvHdrAddr_MSB 0x27
+#define QIB_7322_RcvHdrAddr0_RcvHdrAddr_RMASK 0x3FFFFFFFFF
+
+#define QIB_7322_RcvHdrTailAddr0_OFFS 0x340
+#define QIB_7322_RcvHdrTailAddr0_DEF 0x0000000000000000
+#define QIB_7322_RcvHdrTailAddr0_RcvHdrTailAddr_LSB 0x2
+#define QIB_7322_RcvHdrTailAddr0_RcvHdrTailAddr_MSB 0x27
+#define QIB_7322_RcvHdrTailAddr0_RcvHdrTailAddr_RMASK 0x3FFFFFFFFF
+
+#define QIB_7322_ahb_access_ctrl_OFFS 0x460
+#define QIB_7322_ahb_access_ctrl_DEF 0x0000000000000000
+#define QIB_7322_ahb_access_ctrl_sw_sel_ahb_trgt_LSB 0x1
+#define QIB_7322_ahb_access_ctrl_sw_sel_ahb_trgt_MSB 0x2
+#define QIB_7322_ahb_access_ctrl_sw_sel_ahb_trgt_RMASK 0x3
+#define QIB_7322_ahb_access_ctrl_sw_ahb_sel_LSB 0x0
+#define QIB_7322_ahb_access_ctrl_sw_ahb_sel_MSB 0x0
+#define QIB_7322_ahb_access_ctrl_sw_ahb_sel_RMASK 0x1
+
+#define QIB_7322_ahb_transaction_reg_OFFS 0x468
+#define QIB_7322_ahb_transaction_reg_DEF 0x0000000080000000
+#define QIB_7322_ahb_transaction_reg_ahb_data_LSB 0x20
+#define QIB_7322_ahb_transaction_reg_ahb_data_MSB 0x3F
+#define QIB_7322_ahb_transaction_reg_ahb_data_RMASK 0xFFFFFFFF
+#define QIB_7322_ahb_transaction_reg_ahb_rdy_LSB 0x1F
+#define QIB_7322_ahb_transaction_reg_ahb_rdy_MSB 0x1F
+#define QIB_7322_ahb_transaction_reg_ahb_rdy_RMASK 0x1
+#define QIB_7322_ahb_transaction_reg_ahb_req_err_LSB 0x1E
+#define QIB_7322_ahb_transaction_reg_ahb_req_err_MSB 0x1E
+#define QIB_7322_ahb_transaction_reg_ahb_req_err_RMASK 0x1
+#define QIB_7322_ahb_transaction_reg_write_not_read_LSB 0x1B
+#define QIB_7322_ahb_transaction_reg_write_not_read_MSB 0x1B
+#define QIB_7322_ahb_transaction_reg_write_not_read_RMASK 0x1
+#define QIB_7322_ahb_transaction_reg_ahb_address_LSB 0x10
+#define QIB_7322_ahb_transaction_reg_ahb_address_MSB 0x1A
+#define QIB_7322_ahb_transaction_reg_ahb_address_RMASK 0x7FF
+
+#define QIB_7322_SPC_JTAG_ACCESS_REG_OFFS 0x470
+#define QIB_7322_SPC_JTAG_ACCESS_REG_DEF 0x0000000000000001
+#define QIB_7322_SPC_JTAG_ACCESS_REG_SPC_JTAG_ACCESS_EN_LSB 0xA
+#define QIB_7322_SPC_JTAG_ACCESS_REG_SPC_JTAG_ACCESS_EN_MSB 0xA
+#define QIB_7322_SPC_JTAG_ACCESS_REG_SPC_JTAG_ACCESS_EN_RMASK 0x1
+#define QIB_7322_SPC_JTAG_ACCESS_REG_bist_en_LSB 0x5
+#define QIB_7322_SPC_JTAG_ACCESS_REG_bist_en_MSB 0x9
+#define QIB_7322_SPC_JTAG_ACCESS_REG_bist_en_RMASK 0x1F
+#define QIB_7322_SPC_JTAG_ACCESS_REG_opcode_LSB 0x3
+#define QIB_7322_SPC_JTAG_ACCESS_REG_opcode_MSB 0x4
+#define QIB_7322_SPC_JTAG_ACCESS_REG_opcode_RMASK 0x3
+#define QIB_7322_SPC_JTAG_ACCESS_REG_tdi_LSB 0x2
+#define QIB_7322_SPC_JTAG_ACCESS_REG_tdi_MSB 0x2
+#define QIB_7322_SPC_JTAG_ACCESS_REG_tdi_RMASK 0x1
+#define QIB_7322_SPC_JTAG_ACCESS_REG_tdo_LSB 0x1
+#define QIB_7322_SPC_JTAG_ACCESS_REG_tdo_MSB 0x1
+#define QIB_7322_SPC_JTAG_ACCESS_REG_tdo_RMASK 0x1
+#define QIB_7322_SPC_JTAG_ACCESS_REG_rdy_LSB 0x0
+#define QIB_7322_SPC_JTAG_ACCESS_REG_rdy_MSB 0x0
+#define QIB_7322_SPC_JTAG_ACCESS_REG_rdy_RMASK 0x1
+
+#define QIB_7322_SendCheckMask0_OFFS 0x4C0
+#define QIB_7322_SendCheckMask0_DEF 0x0000000000000000
+#define QIB_7322_SendCheckMask0_SendCheckMask_63_32_LSB 0x0
+#define QIB_7322_SendCheckMask0_SendCheckMask_63_32_MSB 0x3F
+#define QIB_7322_SendCheckMask0_SendCheckMask_63_32_RMASK 0x0
+
+#define QIB_7322_SendGRHCheckMask0_OFFS 0x4E0
+#define QIB_7322_SendGRHCheckMask0_DEF 0x0000000000000000
+#define QIB_7322_SendGRHCheckMask0_SendGRHCheckMask_63_32_LSB 0x0
+#define QIB_7322_SendGRHCheckMask0_SendGRHCheckMask_63_32_MSB 0x3F
+#define QIB_7322_SendGRHCheckMask0_SendGRHCheckMask_63_32_RMASK 0x0
+
+#define QIB_7322_SendIBPacketMask0_OFFS 0x500
+#define QIB_7322_SendIBPacketMask0_DEF 0x0000000000000000
+#define QIB_7322_SendIBPacketMask0_SendIBPacketMask_63_32_LSB 0x0
+#define QIB_7322_SendIBPacketMask0_SendIBPacketMask_63_32_MSB 0x3F
+#define QIB_7322_SendIBPacketMask0_SendIBPacketMask_63_32_RMASK 0x0
+
+#define QIB_7322_IntRedirect0_OFFS 0x540
+#define QIB_7322_IntRedirect0_DEF 0x0000000000000000
+#define QIB_7322_IntRedirect0_vec11_LSB 0x37
+#define QIB_7322_IntRedirect0_vec11_MSB 0x3B
+#define QIB_7322_IntRedirect0_vec11_RMASK 0x1F
+#define QIB_7322_IntRedirect0_vec10_LSB 0x32
+#define QIB_7322_IntRedirect0_vec10_MSB 0x36
+#define QIB_7322_IntRedirect0_vec10_RMASK 0x1F
+#define QIB_7322_IntRedirect0_vec9_LSB 0x2D
+#define QIB_7322_IntRedirect0_vec9_MSB 0x31
+#define QIB_7322_IntRedirect0_vec9_RMASK 0x1F
+#define QIB_7322_IntRedirect0_vec8_LSB 0x28
+#define QIB_7322_IntRedirect0_vec8_MSB 0x2C
+#define QIB_7322_IntRedirect0_vec8_RMASK 0x1F
+#define QIB_7322_IntRedirect0_vec7_LSB 0x23
+#define QIB_7322_IntRedirect0_vec7_MSB 0x27
+#define QIB_7322_IntRedirect0_vec7_RMASK 0x1F
+#define QIB_7322_IntRedirect0_vec6_LSB 0x1E
+#define QIB_7322_IntRedirect0_vec6_MSB 0x22
+#define QIB_7322_IntRedirect0_vec6_RMASK 0x1F
+#define QIB_7322_IntRedirect0_vec5_LSB 0x19
+#define QIB_7322_IntRedirect0_vec5_MSB 0x1D
+#define QIB_7322_IntRedirect0_vec5_RMASK 0x1F
+#define QIB_7322_IntRedirect0_vec4_LSB 0x14
+#define QIB_7322_IntRedirect0_vec4_MSB 0x18
+#define QIB_7322_IntRedirect0_vec4_RMASK 0x1F
+#define QIB_7322_IntRedirect0_vec3_LSB 0xF
+#define QIB_7322_IntRedirect0_vec3_MSB 0x13
+#define QIB_7322_IntRedirect0_vec3_RMASK 0x1F
+#define QIB_7322_IntRedirect0_vec2_LSB 0xA
+#define QIB_7322_IntRedirect0_vec2_MSB 0xE
+#define QIB_7322_IntRedirect0_vec2_RMASK 0x1F
+#define QIB_7322_IntRedirect0_vec1_LSB 0x5
+#define QIB_7322_IntRedirect0_vec1_MSB 0x9
+#define QIB_7322_IntRedirect0_vec1_RMASK 0x1F
+#define QIB_7322_IntRedirect0_vec0_LSB 0x0
+#define QIB_7322_IntRedirect0_vec0_MSB 0x4
+#define QIB_7322_IntRedirect0_vec0_RMASK 0x1F
+
+#define QIB_7322_Int_Granted_OFFS 0x570
+#define QIB_7322_Int_Granted_DEF 0x0000000000000000
+
+#define QIB_7322_vec_clr_without_int_OFFS 0x578
+#define QIB_7322_vec_clr_without_int_DEF 0x0000000000000000
+
+#define QIB_7322_DCACtrlA_OFFS 0x580
+#define QIB_7322_DCACtrlA_DEF 0x0000000000000000
+#define QIB_7322_DCACtrlA_SendDMAHead1DCAEnable_LSB 0x4
+#define QIB_7322_DCACtrlA_SendDMAHead1DCAEnable_MSB 0x4
+#define QIB_7322_DCACtrlA_SendDMAHead1DCAEnable_RMASK 0x1
+#define QIB_7322_DCACtrlA_SendDMAHead0DCAEnable_LSB 0x3
+#define QIB_7322_DCACtrlA_SendDMAHead0DCAEnable_MSB 0x3
+#define QIB_7322_DCACtrlA_SendDMAHead0DCAEnable_RMASK 0x1
+#define QIB_7322_DCACtrlA_RcvTailUpdDCAEnable_LSB 0x2
+#define QIB_7322_DCACtrlA_RcvTailUpdDCAEnable_MSB 0x2
+#define QIB_7322_DCACtrlA_RcvTailUpdDCAEnable_RMASK 0x1
+#define QIB_7322_DCACtrlA_EagerDCAEnable_LSB 0x1
+#define QIB_7322_DCACtrlA_EagerDCAEnable_MSB 0x1
+#define QIB_7322_DCACtrlA_EagerDCAEnable_RMASK 0x1
+#define QIB_7322_DCACtrlA_RcvHdrqDCAEnable_LSB 0x0
+#define QIB_7322_DCACtrlA_RcvHdrqDCAEnable_MSB 0x0
+#define QIB_7322_DCACtrlA_RcvHdrqDCAEnable_RMASK 0x1
+
+#define QIB_7322_DCACtrlB_OFFS 0x588
+#define QIB_7322_DCACtrlB_DEF 0x0000000000000000
+#define QIB_7322_DCACtrlB_RcvHdrq3DCAXfrCnt_LSB 0x36
+#define QIB_7322_DCACtrlB_RcvHdrq3DCAXfrCnt_MSB 0x3B
+#define QIB_7322_DCACtrlB_RcvHdrq3DCAXfrCnt_RMASK 0x3F
+#define QIB_7322_DCACtrlB_RcvHdrq3DCAOPH_LSB 0x2E
+#define QIB_7322_DCACtrlB_RcvHdrq3DCAOPH_MSB 0x35
+#define QIB_7322_DCACtrlB_RcvHdrq3DCAOPH_RMASK 0xFF
+#define QIB_7322_DCACtrlB_RcvHdrq2DCAXfrCnt_LSB 0x28
+#define QIB_7322_DCACtrlB_RcvHdrq2DCAXfrCnt_MSB 0x2D
+#define QIB_7322_DCACtrlB_RcvHdrq2DCAXfrCnt_RMASK 0x3F
+#define QIB_7322_DCACtrlB_RcvHdrq2DCAOPH_LSB 0x20
+#define QIB_7322_DCACtrlB_RcvHdrq2DCAOPH_MSB 0x27
+#define QIB_7322_DCACtrlB_RcvHdrq2DCAOPH_RMASK 0xFF
+#define QIB_7322_DCACtrlB_RcvHdrq1DCAXfrCnt_LSB 0x16
+#define QIB_7322_DCACtrlB_RcvHdrq1DCAXfrCnt_MSB 0x1B
+#define QIB_7322_DCACtrlB_RcvHdrq1DCAXfrCnt_RMASK 0x3F
+#define QIB_7322_DCACtrlB_RcvHdrq1DCAOPH_LSB 0xE
+#define QIB_7322_DCACtrlB_RcvHdrq1DCAOPH_MSB 0x15
+#define QIB_7322_DCACtrlB_RcvHdrq1DCAOPH_RMASK 0xFF
+#define QIB_7322_DCACtrlB_RcvHdrq0DCAXfrCnt_LSB 0x8
+#define QIB_7322_DCACtrlB_RcvHdrq0DCAXfrCnt_MSB 0xD
+#define QIB_7322_DCACtrlB_RcvHdrq0DCAXfrCnt_RMASK 0x3F
+#define QIB_7322_DCACtrlB_RcvHdrq0DCAOPH_LSB 0x0
+#define QIB_7322_DCACtrlB_RcvHdrq0DCAOPH_MSB 0x7
+#define QIB_7322_DCACtrlB_RcvHdrq0DCAOPH_RMASK 0xFF
+
+#define QIB_7322_DCACtrlC_OFFS 0x590
+#define QIB_7322_DCACtrlC_DEF 0x0000000000000000
+#define QIB_7322_DCACtrlC_RcvHdrq7DCAXfrCnt_LSB 0x36
+#define QIB_7322_DCACtrlC_RcvHdrq7DCAXfrCnt_MSB 0x3B
+#define QIB_7322_DCACtrlC_RcvHdrq7DCAXfrCnt_RMASK 0x3F
+#define QIB_7322_DCACtrlC_RcvHdrq7DCAOPH_LSB 0x2E
+#define QIB_7322_DCACtrlC_RcvHdrq7DCAOPH_MSB 0x35
+#define QIB_7322_DCACtrlC_RcvHdrq7DCAOPH_RMASK 0xFF
+#define QIB_7322_DCACtrlC_RcvHdrq6DCAXfrCnt_LSB 0x28
+#define QIB_7322_DCACtrlC_RcvHdrq6DCAXfrCnt_MSB 0x2D
+#define QIB_7322_DCACtrlC_RcvHdrq6DCAXfrCnt_RMASK 0x3F
+#define QIB_7322_DCACtrlC_RcvHdrq6DCAOPH_LSB 0x20
+#define QIB_7322_DCACtrlC_RcvHdrq6DCAOPH_MSB 0x27
+#define QIB_7322_DCACtrlC_RcvHdrq6DCAOPH_RMASK 0xFF
+#define QIB_7322_DCACtrlC_RcvHdrq5DCAXfrCnt_LSB 0x16
+#define QIB_7322_DCACtrlC_RcvHdrq5DCAXfrCnt_MSB 0x1B
+#define QIB_7322_DCACtrlC_RcvHdrq5DCAXfrCnt_RMASK 0x3F
+#define QIB_7322_DCACtrlC_RcvHdrq5DCAOPH_LSB 0xE
+#define QIB_7322_DCACtrlC_RcvHdrq5DCAOPH_MSB 0x15
+#define QIB_7322_DCACtrlC_RcvHdrq5DCAOPH_RMASK 0xFF
+#define QIB_7322_DCACtrlC_RcvHdrq4DCAXfrCnt_LSB 0x8
+#define QIB_7322_DCACtrlC_RcvHdrq4DCAXfrCnt_MSB 0xD
+#define QIB_7322_DCACtrlC_RcvHdrq4DCAXfrCnt_RMASK 0x3F
+#define QIB_7322_DCACtrlC_RcvHdrq4DCAOPH_LSB 0x0
+#define QIB_7322_DCACtrlC_RcvHdrq4DCAOPH_MSB 0x7
+#define QIB_7322_DCACtrlC_RcvHdrq4DCAOPH_RMASK 0xFF
+
+#define QIB_7322_DCACtrlD_OFFS 0x598
+#define QIB_7322_DCACtrlD_DEF 0x0000000000000000
+#define QIB_7322_DCACtrlD_RcvHdrq11DCAXfrCnt_LSB 0x36
+#define QIB_7322_DCACtrlD_RcvHdrq11DCAXfrCnt_MSB 0x3B
+#define QIB_7322_DCACtrlD_RcvHdrq11DCAXfrCnt_RMASK 0x3F
+#define QIB_7322_DCACtrlD_RcvHdrq11DCAOPH_LSB 0x2E
+#define QIB_7322_DCACtrlD_RcvHdrq11DCAOPH_MSB 0x35
+#define QIB_7322_DCACtrlD_RcvHdrq11DCAOPH_RMASK 0xFF
+#define QIB_7322_DCACtrlD_RcvHdrq10DCAXfrCnt_LSB 0x28
+#define QIB_7322_DCACtrlD_RcvHdrq10DCAXfrCnt_MSB 0x2D
+#define QIB_7322_DCACtrlD_RcvHdrq10DCAXfrCnt_RMASK 0x3F
+#define QIB_7322_DCACtrlD_RcvHdrq10DCAOPH_LSB 0x20
+#define QIB_7322_DCACtrlD_RcvHdrq10DCAOPH_MSB 0x27
+#define QIB_7322_DCACtrlD_RcvHdrq10DCAOPH_RMASK 0xFF
+#define QIB_7322_DCACtrlD_RcvHdrq9DCAXfrCnt_LSB 0x16
+#define QIB_7322_DCACtrlD_RcvHdrq9DCAXfrCnt_MSB 0x1B
+#define QIB_7322_DCACtrlD_RcvHdrq9DCAXfrCnt_RMASK 0x3F
+#define QIB_7322_DCACtrlD_RcvHdrq9DCAOPH_LSB 0xE
+#define QIB_7322_DCACtrlD_RcvHdrq9DCAOPH_MSB 0x15
+#define QIB_7322_DCACtrlD_RcvHdrq9DCAOPH_RMASK 0xFF
+#define QIB_7322_DCACtrlD_RcvHdrq8DCAXfrCnt_LSB 0x8
+#define QIB_7322_DCACtrlD_RcvHdrq8DCAXfrCnt_MSB 0xD
+#define QIB_7322_DCACtrlD_RcvHdrq8DCAXfrCnt_RMASK 0x3F
+#define QIB_7322_DCACtrlD_RcvHdrq8DCAOPH_LSB 0x0
+#define QIB_7322_DCACtrlD_RcvHdrq8DCAOPH_MSB 0x7
+#define QIB_7322_DCACtrlD_RcvHdrq8DCAOPH_RMASK 0xFF
+
+#define QIB_7322_DCACtrlE_OFFS 0x5A0
+#define QIB_7322_DCACtrlE_DEF 0x0000000000000000
+#define QIB_7322_DCACtrlE_RcvHdrq15DCAXfrCnt_LSB 0x36
+#define QIB_7322_DCACtrlE_RcvHdrq15DCAXfrCnt_MSB 0x3B
+#define QIB_7322_DCACtrlE_RcvHdrq15DCAXfrCnt_RMASK 0x3F
+#define QIB_7322_DCACtrlE_RcvHdrq15DCAOPH_LSB 0x2E
+#define QIB_7322_DCACtrlE_RcvHdrq15DCAOPH_MSB 0x35
+#define QIB_7322_DCACtrlE_RcvHdrq15DCAOPH_RMASK 0xFF
+#define QIB_7322_DCACtrlE_RcvHdrq14DCAXfrCnt_LSB 0x28
+#define QIB_7322_DCACtrlE_RcvHdrq14DCAXfrCnt_MSB 0x2D
+#define QIB_7322_DCACtrlE_RcvHdrq14DCAXfrCnt_RMASK 0x3F
+#define QIB_7322_DCACtrlE_RcvHdrq14DCAOPH_LSB 0x20
+#define QIB_7322_DCACtrlE_RcvHdrq14DCAOPH_MSB 0x27
+#define QIB_7322_DCACtrlE_RcvHdrq14DCAOPH_RMASK 0xFF
+#define QIB_7322_DCACtrlE_RcvHdrq13DCAXfrCnt_LSB 0x16
+#define QIB_7322_DCACtrlE_RcvHdrq13DCAXfrCnt_MSB 0x1B
+#define QIB_7322_DCACtrlE_RcvHdrq13DCAXfrCnt_RMASK 0x3F
+#define QIB_7322_DCACtrlE_RcvHdrq13DCAOPH_LSB 0xE
+#define QIB_7322_DCACtrlE_RcvHdrq13DCAOPH_MSB 0x15
+#define QIB_7322_DCACtrlE_RcvHdrq13DCAOPH_RMASK 0xFF
+#define QIB_7322_DCACtrlE_RcvHdrq12DCAXfrCnt_LSB 0x8
+#define QIB_7322_DCACtrlE_RcvHdrq12DCAXfrCnt_MSB 0xD
+#define QIB_7322_DCACtrlE_RcvHdrq12DCAXfrCnt_RMASK 0x3F
+#define QIB_7322_DCACtrlE_RcvHdrq12DCAOPH_LSB 0x0
+#define QIB_7322_DCACtrlE_RcvHdrq12DCAOPH_MSB 0x7
+#define QIB_7322_DCACtrlE_RcvHdrq12DCAOPH_RMASK 0xFF
+
+#define QIB_7322_DCACtrlF_OFFS 0x5A8
+#define QIB_7322_DCACtrlF_DEF 0x0000000000000000
+#define QIB_7322_DCACtrlF_SendDma1DCAOPH_LSB 0x28
+#define QIB_7322_DCACtrlF_SendDma1DCAOPH_MSB 0x2F
+#define QIB_7322_DCACtrlF_SendDma1DCAOPH_RMASK 0xFF
+#define QIB_7322_DCACtrlF_SendDma0DCAOPH_LSB 0x20
+#define QIB_7322_DCACtrlF_SendDma0DCAOPH_MSB 0x27
+#define QIB_7322_DCACtrlF_SendDma0DCAOPH_RMASK 0xFF
+#define QIB_7322_DCACtrlF_RcvHdrq17DCAXfrCnt_LSB 0x16
+#define QIB_7322_DCACtrlF_RcvHdrq17DCAXfrCnt_MSB 0x1B
+#define QIB_7322_DCACtrlF_RcvHdrq17DCAXfrCnt_RMASK 0x3F
+#define QIB_7322_DCACtrlF_RcvHdrq17DCAOPH_LSB 0xE
+#define QIB_7322_DCACtrlF_RcvHdrq17DCAOPH_MSB 0x15
+#define QIB_7322_DCACtrlF_RcvHdrq17DCAOPH_RMASK 0xFF
+#define QIB_7322_DCACtrlF_RcvHdrq16DCAXfrCnt_LSB 0x8
+#define QIB_7322_DCACtrlF_RcvHdrq16DCAXfrCnt_MSB 0xD
+#define QIB_7322_DCACtrlF_RcvHdrq16DCAXfrCnt_RMASK 0x3F
+#define QIB_7322_DCACtrlF_RcvHdrq16DCAOPH_LSB 0x0
+#define QIB_7322_DCACtrlF_RcvHdrq16DCAOPH_MSB 0x7
+#define QIB_7322_DCACtrlF_RcvHdrq16DCAOPH_RMASK 0xFF
+
+#define QIB_7322_RcvAvailTimeOut0_OFFS 0xC00
+#define QIB_7322_RcvAvailTimeOut0_DEF 0x0000000000000000
+#define QIB_7322_RcvAvailTimeOut0_RcvAvailTOCount_LSB 0x10
+#define QIB_7322_RcvAvailTimeOut0_RcvAvailTOCount_MSB 0x1F
+#define QIB_7322_RcvAvailTimeOut0_RcvAvailTOCount_RMASK 0xFFFF
+#define QIB_7322_RcvAvailTimeOut0_RcvAvailTOReload_LSB 0x0
+#define QIB_7322_RcvAvailTimeOut0_RcvAvailTOReload_MSB 0xF
+#define QIB_7322_RcvAvailTimeOut0_RcvAvailTOReload_RMASK 0xFFFF
+
+#define QIB_7322_CntrRegBase_0_OFFS 0x1028
+#define QIB_7322_CntrRegBase_0_DEF 0x0000000000012000
+
+#define QIB_7322_ErrMask_0_OFFS 0x1080
+#define QIB_7322_ErrMask_0_DEF 0x0000000000000000
+#define QIB_7322_ErrMask_0_IBStatusChangedMask_LSB 0x3A
+#define QIB_7322_ErrMask_0_IBStatusChangedMask_MSB 0x3A
+#define QIB_7322_ErrMask_0_IBStatusChangedMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SHeadersErrMask_LSB 0x39
+#define QIB_7322_ErrMask_0_SHeadersErrMask_MSB 0x39
+#define QIB_7322_ErrMask_0_SHeadersErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_VL15BufMisuseErrMask_LSB 0x36
+#define QIB_7322_ErrMask_0_VL15BufMisuseErrMask_MSB 0x36
+#define QIB_7322_ErrMask_0_VL15BufMisuseErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SDmaHaltErrMask_LSB 0x31
+#define QIB_7322_ErrMask_0_SDmaHaltErrMask_MSB 0x31
+#define QIB_7322_ErrMask_0_SDmaHaltErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SDmaDescAddrMisalignErrMask_LSB 0x30
+#define QIB_7322_ErrMask_0_SDmaDescAddrMisalignErrMask_MSB 0x30
+#define QIB_7322_ErrMask_0_SDmaDescAddrMisalignErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SDmaUnexpDataErrMask_LSB 0x2F
+#define QIB_7322_ErrMask_0_SDmaUnexpDataErrMask_MSB 0x2F
+#define QIB_7322_ErrMask_0_SDmaUnexpDataErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SDmaMissingDwErrMask_LSB 0x2E
+#define QIB_7322_ErrMask_0_SDmaMissingDwErrMask_MSB 0x2E
+#define QIB_7322_ErrMask_0_SDmaMissingDwErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SDmaDwEnErrMask_LSB 0x2D
+#define QIB_7322_ErrMask_0_SDmaDwEnErrMask_MSB 0x2D
+#define QIB_7322_ErrMask_0_SDmaDwEnErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SDmaRpyTagErrMask_LSB 0x2C
+#define QIB_7322_ErrMask_0_SDmaRpyTagErrMask_MSB 0x2C
+#define QIB_7322_ErrMask_0_SDmaRpyTagErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SDma1stDescErrMask_LSB 0x2B
+#define QIB_7322_ErrMask_0_SDma1stDescErrMask_MSB 0x2B
+#define QIB_7322_ErrMask_0_SDma1stDescErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SDmaBaseErrMask_LSB 0x2A
+#define QIB_7322_ErrMask_0_SDmaBaseErrMask_MSB 0x2A
+#define QIB_7322_ErrMask_0_SDmaBaseErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SDmaTailOutOfBoundErrMask_LSB 0x29
+#define QIB_7322_ErrMask_0_SDmaTailOutOfBoundErrMask_MSB 0x29
+#define QIB_7322_ErrMask_0_SDmaTailOutOfBoundErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SDmaOutOfBoundErrMask_LSB 0x28
+#define QIB_7322_ErrMask_0_SDmaOutOfBoundErrMask_MSB 0x28
+#define QIB_7322_ErrMask_0_SDmaOutOfBoundErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SDmaGenMismatchErrMask_LSB 0x27
+#define QIB_7322_ErrMask_0_SDmaGenMismatchErrMask_MSB 0x27
+#define QIB_7322_ErrMask_0_SDmaGenMismatchErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SendBufMisuseErrMask_LSB 0x26
+#define QIB_7322_ErrMask_0_SendBufMisuseErrMask_MSB 0x26
+#define QIB_7322_ErrMask_0_SendBufMisuseErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SendUnsupportedVLErrMask_LSB 0x25
+#define QIB_7322_ErrMask_0_SendUnsupportedVLErrMask_MSB 0x25
+#define QIB_7322_ErrMask_0_SendUnsupportedVLErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SendUnexpectedPktNumErrMask_LSB 0x24
+#define QIB_7322_ErrMask_0_SendUnexpectedPktNumErrMask_MSB 0x24
+#define QIB_7322_ErrMask_0_SendUnexpectedPktNumErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SendDroppedDataPktErrMask_LSB 0x22
+#define QIB_7322_ErrMask_0_SendDroppedDataPktErrMask_MSB 0x22
+#define QIB_7322_ErrMask_0_SendDroppedDataPktErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SendDroppedSmpPktErrMask_LSB 0x21
+#define QIB_7322_ErrMask_0_SendDroppedSmpPktErrMask_MSB 0x21
+#define QIB_7322_ErrMask_0_SendDroppedSmpPktErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SendPktLenErrMask_LSB 0x20
+#define QIB_7322_ErrMask_0_SendPktLenErrMask_MSB 0x20
+#define QIB_7322_ErrMask_0_SendPktLenErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SendUnderRunErrMask_LSB 0x1F
+#define QIB_7322_ErrMask_0_SendUnderRunErrMask_MSB 0x1F
+#define QIB_7322_ErrMask_0_SendUnderRunErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SendMaxPktLenErrMask_LSB 0x1E
+#define QIB_7322_ErrMask_0_SendMaxPktLenErrMask_MSB 0x1E
+#define QIB_7322_ErrMask_0_SendMaxPktLenErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_SendMinPktLenErrMask_LSB 0x1D
+#define QIB_7322_ErrMask_0_SendMinPktLenErrMask_MSB 0x1D
+#define QIB_7322_ErrMask_0_SendMinPktLenErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_RcvIBLostLinkErrMask_LSB 0x11
+#define QIB_7322_ErrMask_0_RcvIBLostLinkErrMask_MSB 0x11
+#define QIB_7322_ErrMask_0_RcvIBLostLinkErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_RcvHdrErrMask_LSB 0x10
+#define QIB_7322_ErrMask_0_RcvHdrErrMask_MSB 0x10
+#define QIB_7322_ErrMask_0_RcvHdrErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_RcvHdrLenErrMask_LSB 0xF
+#define QIB_7322_ErrMask_0_RcvHdrLenErrMask_MSB 0xF
+#define QIB_7322_ErrMask_0_RcvHdrLenErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_RcvBadTidErrMask_LSB 0xE
+#define QIB_7322_ErrMask_0_RcvBadTidErrMask_MSB 0xE
+#define QIB_7322_ErrMask_0_RcvBadTidErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_RcvBadVersionErrMask_LSB 0xB
+#define QIB_7322_ErrMask_0_RcvBadVersionErrMask_MSB 0xB
+#define QIB_7322_ErrMask_0_RcvBadVersionErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_RcvIBFlowErrMask_LSB 0xA
+#define QIB_7322_ErrMask_0_RcvIBFlowErrMask_MSB 0xA
+#define QIB_7322_ErrMask_0_RcvIBFlowErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_RcvEBPErrMask_LSB 0x9
+#define QIB_7322_ErrMask_0_RcvEBPErrMask_MSB 0x9
+#define QIB_7322_ErrMask_0_RcvEBPErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_RcvUnsupportedVLErrMask_LSB 0x8
+#define QIB_7322_ErrMask_0_RcvUnsupportedVLErrMask_MSB 0x8
+#define QIB_7322_ErrMask_0_RcvUnsupportedVLErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_RcvUnexpectedCharErrMask_LSB 0x7
+#define QIB_7322_ErrMask_0_RcvUnexpectedCharErrMask_MSB 0x7
+#define QIB_7322_ErrMask_0_RcvUnexpectedCharErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_RcvShortPktLenErrMask_LSB 0x6
+#define QIB_7322_ErrMask_0_RcvShortPktLenErrMask_MSB 0x6
+#define QIB_7322_ErrMask_0_RcvShortPktLenErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_RcvLongPktLenErrMask_LSB 0x5
+#define QIB_7322_ErrMask_0_RcvLongPktLenErrMask_MSB 0x5
+#define QIB_7322_ErrMask_0_RcvLongPktLenErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_RcvMaxPktLenErrMask_LSB 0x4
+#define QIB_7322_ErrMask_0_RcvMaxPktLenErrMask_MSB 0x4
+#define QIB_7322_ErrMask_0_RcvMaxPktLenErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_RcvMinPktLenErrMask_LSB 0x3
+#define QIB_7322_ErrMask_0_RcvMinPktLenErrMask_MSB 0x3
+#define QIB_7322_ErrMask_0_RcvMinPktLenErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_RcvICRCErrMask_LSB 0x2
+#define QIB_7322_ErrMask_0_RcvICRCErrMask_MSB 0x2
+#define QIB_7322_ErrMask_0_RcvICRCErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_RcvVCRCErrMask_LSB 0x1
+#define QIB_7322_ErrMask_0_RcvVCRCErrMask_MSB 0x1
+#define QIB_7322_ErrMask_0_RcvVCRCErrMask_RMASK 0x1
+#define QIB_7322_ErrMask_0_RcvFormatErrMask_LSB 0x0
+#define QIB_7322_ErrMask_0_RcvFormatErrMask_MSB 0x0
+#define QIB_7322_ErrMask_0_RcvFormatErrMask_RMASK 0x1
+
+#define QIB_7322_ErrStatus_0_OFFS 0x1088
+#define QIB_7322_ErrStatus_0_DEF 0x0000000000000000
+#define QIB_7322_ErrStatus_0_IBStatusChanged_LSB 0x3A
+#define QIB_7322_ErrStatus_0_IBStatusChanged_MSB 0x3A
+#define QIB_7322_ErrStatus_0_IBStatusChanged_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SHeadersErr_LSB 0x39
+#define QIB_7322_ErrStatus_0_SHeadersErr_MSB 0x39
+#define QIB_7322_ErrStatus_0_SHeadersErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_VL15BufMisuseErr_LSB 0x36
+#define QIB_7322_ErrStatus_0_VL15BufMisuseErr_MSB 0x36
+#define QIB_7322_ErrStatus_0_VL15BufMisuseErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SDmaHaltErr_LSB 0x31
+#define QIB_7322_ErrStatus_0_SDmaHaltErr_MSB 0x31
+#define QIB_7322_ErrStatus_0_SDmaHaltErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SDmaDescAddrMisalignErr_LSB 0x30
+#define QIB_7322_ErrStatus_0_SDmaDescAddrMisalignErr_MSB 0x30
+#define QIB_7322_ErrStatus_0_SDmaDescAddrMisalignErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SDmaUnexpDataErr_LSB 0x2F
+#define QIB_7322_ErrStatus_0_SDmaUnexpDataErr_MSB 0x2F
+#define QIB_7322_ErrStatus_0_SDmaUnexpDataErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SDmaMissingDwErr_LSB 0x2E
+#define QIB_7322_ErrStatus_0_SDmaMissingDwErr_MSB 0x2E
+#define QIB_7322_ErrStatus_0_SDmaMissingDwErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SDmaDwEnErr_LSB 0x2D
+#define QIB_7322_ErrStatus_0_SDmaDwEnErr_MSB 0x2D
+#define QIB_7322_ErrStatus_0_SDmaDwEnErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SDmaRpyTagErr_LSB 0x2C
+#define QIB_7322_ErrStatus_0_SDmaRpyTagErr_MSB 0x2C
+#define QIB_7322_ErrStatus_0_SDmaRpyTagErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SDma1stDescErr_LSB 0x2B
+#define QIB_7322_ErrStatus_0_SDma1stDescErr_MSB 0x2B
+#define QIB_7322_ErrStatus_0_SDma1stDescErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SDmaBaseErr_LSB 0x2A
+#define QIB_7322_ErrStatus_0_SDmaBaseErr_MSB 0x2A
+#define QIB_7322_ErrStatus_0_SDmaBaseErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SDmaTailOutOfBoundErr_LSB 0x29
+#define QIB_7322_ErrStatus_0_SDmaTailOutOfBoundErr_MSB 0x29
+#define QIB_7322_ErrStatus_0_SDmaTailOutOfBoundErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SDmaOutOfBoundErr_LSB 0x28
+#define QIB_7322_ErrStatus_0_SDmaOutOfBoundErr_MSB 0x28
+#define QIB_7322_ErrStatus_0_SDmaOutOfBoundErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SDmaGenMismatchErr_LSB 0x27
+#define QIB_7322_ErrStatus_0_SDmaGenMismatchErr_MSB 0x27
+#define QIB_7322_ErrStatus_0_SDmaGenMismatchErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SendBufMisuseErr_LSB 0x26
+#define QIB_7322_ErrStatus_0_SendBufMisuseErr_MSB 0x26
+#define QIB_7322_ErrStatus_0_SendBufMisuseErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SendUnsupportedVLErr_LSB 0x25
+#define QIB_7322_ErrStatus_0_SendUnsupportedVLErr_MSB 0x25
+#define QIB_7322_ErrStatus_0_SendUnsupportedVLErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SendUnexpectedPktNumErr_LSB 0x24
+#define QIB_7322_ErrStatus_0_SendUnexpectedPktNumErr_MSB 0x24
+#define QIB_7322_ErrStatus_0_SendUnexpectedPktNumErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SendDroppedDataPktErr_LSB 0x22
+#define QIB_7322_ErrStatus_0_SendDroppedDataPktErr_MSB 0x22
+#define QIB_7322_ErrStatus_0_SendDroppedDataPktErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SendDroppedSmpPktErr_LSB 0x21
+#define QIB_7322_ErrStatus_0_SendDroppedSmpPktErr_MSB 0x21
+#define QIB_7322_ErrStatus_0_SendDroppedSmpPktErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SendPktLenErr_LSB 0x20
+#define QIB_7322_ErrStatus_0_SendPktLenErr_MSB 0x20
+#define QIB_7322_ErrStatus_0_SendPktLenErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SendUnderRunErr_LSB 0x1F
+#define QIB_7322_ErrStatus_0_SendUnderRunErr_MSB 0x1F
+#define QIB_7322_ErrStatus_0_SendUnderRunErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SendMaxPktLenErr_LSB 0x1E
+#define QIB_7322_ErrStatus_0_SendMaxPktLenErr_MSB 0x1E
+#define QIB_7322_ErrStatus_0_SendMaxPktLenErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_SendMinPktLenErr_LSB 0x1D
+#define QIB_7322_ErrStatus_0_SendMinPktLenErr_MSB 0x1D
+#define QIB_7322_ErrStatus_0_SendMinPktLenErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_RcvIBLostLinkErr_LSB 0x11
+#define QIB_7322_ErrStatus_0_RcvIBLostLinkErr_MSB 0x11
+#define QIB_7322_ErrStatus_0_RcvIBLostLinkErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_RcvHdrErr_LSB 0x10
+#define QIB_7322_ErrStatus_0_RcvHdrErr_MSB 0x10
+#define QIB_7322_ErrStatus_0_RcvHdrErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_RcvHdrLenErr_LSB 0xF
+#define QIB_7322_ErrStatus_0_RcvHdrLenErr_MSB 0xF
+#define QIB_7322_ErrStatus_0_RcvHdrLenErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_RcvBadTidErr_LSB 0xE
+#define QIB_7322_ErrStatus_0_RcvBadTidErr_MSB 0xE
+#define QIB_7322_ErrStatus_0_RcvBadTidErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_RcvBadVersionErr_LSB 0xB
+#define QIB_7322_ErrStatus_0_RcvBadVersionErr_MSB 0xB
+#define QIB_7322_ErrStatus_0_RcvBadVersionErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_RcvIBFlowErr_LSB 0xA
+#define QIB_7322_ErrStatus_0_RcvIBFlowErr_MSB 0xA
+#define QIB_7322_ErrStatus_0_RcvIBFlowErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_RcvEBPErr_LSB 0x9
+#define QIB_7322_ErrStatus_0_RcvEBPErr_MSB 0x9
+#define QIB_7322_ErrStatus_0_RcvEBPErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_RcvUnsupportedVLErr_LSB 0x8
+#define QIB_7322_ErrStatus_0_RcvUnsupportedVLErr_MSB 0x8
+#define QIB_7322_ErrStatus_0_RcvUnsupportedVLErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_RcvUnexpectedCharErr_LSB 0x7
+#define QIB_7322_ErrStatus_0_RcvUnexpectedCharErr_MSB 0x7
+#define QIB_7322_ErrStatus_0_RcvUnexpectedCharErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_RcvShortPktLenErr_LSB 0x6
+#define QIB_7322_ErrStatus_0_RcvShortPktLenErr_MSB 0x6
+#define QIB_7322_ErrStatus_0_RcvShortPktLenErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_RcvLongPktLenErr_LSB 0x5
+#define QIB_7322_ErrStatus_0_RcvLongPktLenErr_MSB 0x5
+#define QIB_7322_ErrStatus_0_RcvLongPktLenErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_RcvMaxPktLenErr_LSB 0x4
+#define QIB_7322_ErrStatus_0_RcvMaxPktLenErr_MSB 0x4
+#define QIB_7322_ErrStatus_0_RcvMaxPktLenErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_RcvMinPktLenErr_LSB 0x3
+#define QIB_7322_ErrStatus_0_RcvMinPktLenErr_MSB 0x3
+#define QIB_7322_ErrStatus_0_RcvMinPktLenErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_RcvICRCErr_LSB 0x2
+#define QIB_7322_ErrStatus_0_RcvICRCErr_MSB 0x2
+#define QIB_7322_ErrStatus_0_RcvICRCErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_RcvVCRCErr_LSB 0x1
+#define QIB_7322_ErrStatus_0_RcvVCRCErr_MSB 0x1
+#define QIB_7322_ErrStatus_0_RcvVCRCErr_RMASK 0x1
+#define QIB_7322_ErrStatus_0_RcvFormatErr_LSB 0x0
+#define QIB_7322_ErrStatus_0_RcvFormatErr_MSB 0x0
+#define QIB_7322_ErrStatus_0_RcvFormatErr_RMASK 0x1
+
+#define QIB_7322_ErrClear_0_OFFS 0x1090
+#define QIB_7322_ErrClear_0_DEF 0x0000000000000000
+#define QIB_7322_ErrClear_0_IBStatusChangedClear_LSB 0x3A
+#define QIB_7322_ErrClear_0_IBStatusChangedClear_MSB 0x3A
+#define QIB_7322_ErrClear_0_IBStatusChangedClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SHeadersErrClear_LSB 0x39
+#define QIB_7322_ErrClear_0_SHeadersErrClear_MSB 0x39
+#define QIB_7322_ErrClear_0_SHeadersErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_VL15BufMisuseErrClear_LSB 0x36
+#define QIB_7322_ErrClear_0_VL15BufMisuseErrClear_MSB 0x36
+#define QIB_7322_ErrClear_0_VL15BufMisuseErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SDmaHaltErrClear_LSB 0x31
+#define QIB_7322_ErrClear_0_SDmaHaltErrClear_MSB 0x31
+#define QIB_7322_ErrClear_0_SDmaHaltErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SDmaDescAddrMisalignErrClear_LSB 0x30
+#define QIB_7322_ErrClear_0_SDmaDescAddrMisalignErrClear_MSB 0x30
+#define QIB_7322_ErrClear_0_SDmaDescAddrMisalignErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SDmaUnexpDataErrClear_LSB 0x2F
+#define QIB_7322_ErrClear_0_SDmaUnexpDataErrClear_MSB 0x2F
+#define QIB_7322_ErrClear_0_SDmaUnexpDataErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SDmaMissingDwErrClear_LSB 0x2E
+#define QIB_7322_ErrClear_0_SDmaMissingDwErrClear_MSB 0x2E
+#define QIB_7322_ErrClear_0_SDmaMissingDwErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SDmaDwEnErrClear_LSB 0x2D
+#define QIB_7322_ErrClear_0_SDmaDwEnErrClear_MSB 0x2D
+#define QIB_7322_ErrClear_0_SDmaDwEnErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SDmaRpyTagErrClear_LSB 0x2C
+#define QIB_7322_ErrClear_0_SDmaRpyTagErrClear_MSB 0x2C
+#define QIB_7322_ErrClear_0_SDmaRpyTagErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SDma1stDescErrClear_LSB 0x2B
+#define QIB_7322_ErrClear_0_SDma1stDescErrClear_MSB 0x2B
+#define QIB_7322_ErrClear_0_SDma1stDescErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SDmaBaseErrClear_LSB 0x2A
+#define QIB_7322_ErrClear_0_SDmaBaseErrClear_MSB 0x2A
+#define QIB_7322_ErrClear_0_SDmaBaseErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SDmaTailOutOfBoundErrClear_LSB 0x29
+#define QIB_7322_ErrClear_0_SDmaTailOutOfBoundErrClear_MSB 0x29
+#define QIB_7322_ErrClear_0_SDmaTailOutOfBoundErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SDmaOutOfBoundErrClear_LSB 0x28
+#define QIB_7322_ErrClear_0_SDmaOutOfBoundErrClear_MSB 0x28
+#define QIB_7322_ErrClear_0_SDmaOutOfBoundErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SDmaGenMismatchErrClear_LSB 0x27
+#define QIB_7322_ErrClear_0_SDmaGenMismatchErrClear_MSB 0x27
+#define QIB_7322_ErrClear_0_SDmaGenMismatchErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SendBufMisuseErrClear_LSB 0x26
+#define QIB_7322_ErrClear_0_SendBufMisuseErrClear_MSB 0x26
+#define QIB_7322_ErrClear_0_SendBufMisuseErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SendUnsupportedVLErrClear_LSB 0x25
+#define QIB_7322_ErrClear_0_SendUnsupportedVLErrClear_MSB 0x25
+#define QIB_7322_ErrClear_0_SendUnsupportedVLErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SendUnexpectedPktNumErrClear_LSB 0x24
+#define QIB_7322_ErrClear_0_SendUnexpectedPktNumErrClear_MSB 0x24
+#define QIB_7322_ErrClear_0_SendUnexpectedPktNumErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SendDroppedDataPktErrClear_LSB 0x22
+#define QIB_7322_ErrClear_0_SendDroppedDataPktErrClear_MSB 0x22
+#define QIB_7322_ErrClear_0_SendDroppedDataPktErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SendDroppedSmpPktErrClear_LSB 0x21
+#define QIB_7322_ErrClear_0_SendDroppedSmpPktErrClear_MSB 0x21
+#define QIB_7322_ErrClear_0_SendDroppedSmpPktErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SendPktLenErrClear_LSB 0x20
+#define QIB_7322_ErrClear_0_SendPktLenErrClear_MSB 0x20
+#define QIB_7322_ErrClear_0_SendPktLenErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SendUnderRunErrClear_LSB 0x1F
+#define QIB_7322_ErrClear_0_SendUnderRunErrClear_MSB 0x1F
+#define QIB_7322_ErrClear_0_SendUnderRunErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SendMaxPktLenErrClear_LSB 0x1E
+#define QIB_7322_ErrClear_0_SendMaxPktLenErrClear_MSB 0x1E
+#define QIB_7322_ErrClear_0_SendMaxPktLenErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_SendMinPktLenErrClear_LSB 0x1D
+#define QIB_7322_ErrClear_0_SendMinPktLenErrClear_MSB 0x1D
+#define QIB_7322_ErrClear_0_SendMinPktLenErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_RcvIBLostLinkErrClear_LSB 0x11
+#define QIB_7322_ErrClear_0_RcvIBLostLinkErrClear_MSB 0x11
+#define QIB_7322_ErrClear_0_RcvIBLostLinkErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_RcvHdrErrClear_LSB 0x10
+#define QIB_7322_ErrClear_0_RcvHdrErrClear_MSB 0x10
+#define QIB_7322_ErrClear_0_RcvHdrErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_RcvHdrLenErrClear_LSB 0xF
+#define QIB_7322_ErrClear_0_RcvHdrLenErrClear_MSB 0xF
+#define QIB_7322_ErrClear_0_RcvHdrLenErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_RcvBadTidErrClear_LSB 0xE
+#define QIB_7322_ErrClear_0_RcvBadTidErrClear_MSB 0xE
+#define QIB_7322_ErrClear_0_RcvBadTidErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_RcvBadVersionErrClear_LSB 0xB
+#define QIB_7322_ErrClear_0_RcvBadVersionErrClear_MSB 0xB
+#define QIB_7322_ErrClear_0_RcvBadVersionErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_RcvIBFlowErrClear_LSB 0xA
+#define QIB_7322_ErrClear_0_RcvIBFlowErrClear_MSB 0xA
+#define QIB_7322_ErrClear_0_RcvIBFlowErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_RcvEBPErrClear_LSB 0x9
+#define QIB_7322_ErrClear_0_RcvEBPErrClear_MSB 0x9
+#define QIB_7322_ErrClear_0_RcvEBPErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_RcvUnsupportedVLErrClear_LSB 0x8
+#define QIB_7322_ErrClear_0_RcvUnsupportedVLErrClear_MSB 0x8
+#define QIB_7322_ErrClear_0_RcvUnsupportedVLErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_RcvUnexpectedCharErrClear_LSB 0x7
+#define QIB_7322_ErrClear_0_RcvUnexpectedCharErrClear_MSB 0x7
+#define QIB_7322_ErrClear_0_RcvUnexpectedCharErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_RcvShortPktLenErrClear_LSB 0x6
+#define QIB_7322_ErrClear_0_RcvShortPktLenErrClear_MSB 0x6
+#define QIB_7322_ErrClear_0_RcvShortPktLenErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_RcvLongPktLenErrClear_LSB 0x5
+#define QIB_7322_ErrClear_0_RcvLongPktLenErrClear_MSB 0x5
+#define QIB_7322_ErrClear_0_RcvLongPktLenErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_RcvMaxPktLenErrClear_LSB 0x4
+#define QIB_7322_ErrClear_0_RcvMaxPktLenErrClear_MSB 0x4
+#define QIB_7322_ErrClear_0_RcvMaxPktLenErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_RcvMinPktLenErrClear_LSB 0x3
+#define QIB_7322_ErrClear_0_RcvMinPktLenErrClear_MSB 0x3
+#define QIB_7322_ErrClear_0_RcvMinPktLenErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_RcvICRCErrClear_LSB 0x2
+#define QIB_7322_ErrClear_0_RcvICRCErrClear_MSB 0x2
+#define QIB_7322_ErrClear_0_RcvICRCErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_RcvVCRCErrClear_LSB 0x1
+#define QIB_7322_ErrClear_0_RcvVCRCErrClear_MSB 0x1
+#define QIB_7322_ErrClear_0_RcvVCRCErrClear_RMASK 0x1
+#define QIB_7322_ErrClear_0_RcvFormatErrClear_LSB 0x0
+#define QIB_7322_ErrClear_0_RcvFormatErrClear_MSB 0x0
+#define QIB_7322_ErrClear_0_RcvFormatErrClear_RMASK 0x1
+
+#define QIB_7322_TXEStatus_0_OFFS 0x10B8
+#define QIB_7322_TXEStatus_0_DEF 0x0000000XC00080FF
+#define QIB_7322_TXEStatus_0_TXE_IBC_Idle_LSB 0x1F
+#define QIB_7322_TXEStatus_0_TXE_IBC_Idle_MSB 0x1F
+#define QIB_7322_TXEStatus_0_TXE_IBC_Idle_RMASK 0x1
+#define QIB_7322_TXEStatus_0_RmFifoEmpty_LSB 0x1E
+#define QIB_7322_TXEStatus_0_RmFifoEmpty_MSB 0x1E
+#define QIB_7322_TXEStatus_0_RmFifoEmpty_RMASK 0x1
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL15_LSB 0xF
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL15_MSB 0xF
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL15_RMASK 0x1
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL7_LSB 0x7
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL7_MSB 0x7
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL7_RMASK 0x1
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL6_LSB 0x6
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL6_MSB 0x6
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL6_RMASK 0x1
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL5_LSB 0x5
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL5_MSB 0x5
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL5_RMASK 0x1
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL4_LSB 0x4
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL4_MSB 0x4
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL4_RMASK 0x1
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL3_LSB 0x3
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL3_MSB 0x3
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL3_RMASK 0x1
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL2_LSB 0x2
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL2_MSB 0x2
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL2_RMASK 0x1
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL1_LSB 0x1
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL1_MSB 0x1
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL1_RMASK 0x1
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL0_LSB 0x0
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL0_MSB 0x0
+#define QIB_7322_TXEStatus_0_LaFifoEmpty_VL0_RMASK 0x1
+
+#define QIB_7322_RcvCtrl_0_OFFS 0x1100
+#define QIB_7322_RcvCtrl_0_DEF 0x0000000000000000
+#define QIB_7322_RcvCtrl_0_RcvResetCredit_LSB 0x2A
+#define QIB_7322_RcvCtrl_0_RcvResetCredit_MSB 0x2A
+#define QIB_7322_RcvCtrl_0_RcvResetCredit_RMASK 0x1
+#define QIB_7322_RcvCtrl_0_RcvPartitionKeyDisable_LSB 0x29
+#define QIB_7322_RcvCtrl_0_RcvPartitionKeyDisable_MSB 0x29
+#define QIB_7322_RcvCtrl_0_RcvPartitionKeyDisable_RMASK 0x1
+#define QIB_7322_RcvCtrl_0_RcvQPMapEnable_LSB 0x28
+#define QIB_7322_RcvCtrl_0_RcvQPMapEnable_MSB 0x28
+#define QIB_7322_RcvCtrl_0_RcvQPMapEnable_RMASK 0x1
+#define QIB_7322_RcvCtrl_0_RcvIBPortEnable_LSB 0x27
+#define QIB_7322_RcvCtrl_0_RcvIBPortEnable_MSB 0x27
+#define QIB_7322_RcvCtrl_0_RcvIBPortEnable_RMASK 0x1
+#define QIB_7322_RcvCtrl_0_ContextEnableUser_LSB 0x2
+#define QIB_7322_RcvCtrl_0_ContextEnableUser_MSB 0x11
+#define QIB_7322_RcvCtrl_0_ContextEnableUser_RMASK 0xFFFF
+#define QIB_7322_RcvCtrl_0_ContextEnableKernel_LSB 0x0
+#define QIB_7322_RcvCtrl_0_ContextEnableKernel_MSB 0x0
+#define QIB_7322_RcvCtrl_0_ContextEnableKernel_RMASK 0x1
+
+#define QIB_7322_RcvBTHQP_0_OFFS 0x1108
+#define QIB_7322_RcvBTHQP_0_DEF 0x0000000000000000
+#define QIB_7322_RcvBTHQP_0_RcvBTHQP_LSB 0x0
+#define QIB_7322_RcvBTHQP_0_RcvBTHQP_MSB 0x17
+#define QIB_7322_RcvBTHQP_0_RcvBTHQP_RMASK 0xFFFFFF
+
+#define QIB_7322_RcvQPMapTableA_0_OFFS 0x1110
+#define QIB_7322_RcvQPMapTableA_0_DEF 0x0000000000000000
+#define QIB_7322_RcvQPMapTableA_0_RcvQPMapContext5_LSB 0x19
+#define QIB_7322_RcvQPMapTableA_0_RcvQPMapContext5_MSB 0x1D
+#define QIB_7322_RcvQPMapTableA_0_RcvQPMapContext5_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableA_0_RcvQPMapContext4_LSB 0x14
+#define QIB_7322_RcvQPMapTableA_0_RcvQPMapContext4_MSB 0x18
+#define QIB_7322_RcvQPMapTableA_0_RcvQPMapContext4_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableA_0_RcvQPMapContext3_LSB 0xF
+#define QIB_7322_RcvQPMapTableA_0_RcvQPMapContext3_MSB 0x13
+#define QIB_7322_RcvQPMapTableA_0_RcvQPMapContext3_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableA_0_RcvQPMapContext2_LSB 0xA
+#define QIB_7322_RcvQPMapTableA_0_RcvQPMapContext2_MSB 0xE
+#define QIB_7322_RcvQPMapTableA_0_RcvQPMapContext2_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableA_0_RcvQPMapContext1_LSB 0x5
+#define QIB_7322_RcvQPMapTableA_0_RcvQPMapContext1_MSB 0x9
+#define QIB_7322_RcvQPMapTableA_0_RcvQPMapContext1_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableA_0_RcvQPMapContext0_LSB 0x0
+#define QIB_7322_RcvQPMapTableA_0_RcvQPMapContext0_MSB 0x4
+#define QIB_7322_RcvQPMapTableA_0_RcvQPMapContext0_RMASK 0x1F
+
+#define QIB_7322_RcvQPMapTableB_0_OFFS 0x1118
+#define QIB_7322_RcvQPMapTableB_0_DEF 0x0000000000000000
+#define QIB_7322_RcvQPMapTableB_0_RcvQPMapContext11_LSB 0x19
+#define QIB_7322_RcvQPMapTableB_0_RcvQPMapContext11_MSB 0x1D
+#define QIB_7322_RcvQPMapTableB_0_RcvQPMapContext11_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableB_0_RcvQPMapContext10_LSB 0x14
+#define QIB_7322_RcvQPMapTableB_0_RcvQPMapContext10_MSB 0x18
+#define QIB_7322_RcvQPMapTableB_0_RcvQPMapContext10_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableB_0_RcvQPMapContext9_LSB 0xF
+#define QIB_7322_RcvQPMapTableB_0_RcvQPMapContext9_MSB 0x13
+#define QIB_7322_RcvQPMapTableB_0_RcvQPMapContext9_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableB_0_RcvQPMapContext8_LSB 0xA
+#define QIB_7322_RcvQPMapTableB_0_RcvQPMapContext8_MSB 0xE
+#define QIB_7322_RcvQPMapTableB_0_RcvQPMapContext8_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableB_0_RcvQPMapContext7_LSB 0x5
+#define QIB_7322_RcvQPMapTableB_0_RcvQPMapContext7_MSB 0x9
+#define QIB_7322_RcvQPMapTableB_0_RcvQPMapContext7_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableB_0_RcvQPMapContext6_LSB 0x0
+#define QIB_7322_RcvQPMapTableB_0_RcvQPMapContext6_MSB 0x4
+#define QIB_7322_RcvQPMapTableB_0_RcvQPMapContext6_RMASK 0x1F
+
+#define QIB_7322_RcvQPMapTableC_0_OFFS 0x1120
+#define QIB_7322_RcvQPMapTableC_0_DEF 0x0000000000000000
+#define QIB_7322_RcvQPMapTableC_0_RcvQPMapContext17_LSB 0x19
+#define QIB_7322_RcvQPMapTableC_0_RcvQPMapContext17_MSB 0x1D
+#define QIB_7322_RcvQPMapTableC_0_RcvQPMapContext17_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableC_0_RcvQPMapContext16_LSB 0x14
+#define QIB_7322_RcvQPMapTableC_0_RcvQPMapContext16_MSB 0x18
+#define QIB_7322_RcvQPMapTableC_0_RcvQPMapContext16_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableC_0_RcvQPMapContext15_LSB 0xF
+#define QIB_7322_RcvQPMapTableC_0_RcvQPMapContext15_MSB 0x13
+#define QIB_7322_RcvQPMapTableC_0_RcvQPMapContext15_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableC_0_RcvQPMapContext14_LSB 0xA
+#define QIB_7322_RcvQPMapTableC_0_RcvQPMapContext14_MSB 0xE
+#define QIB_7322_RcvQPMapTableC_0_RcvQPMapContext14_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableC_0_RcvQPMapContext13_LSB 0x5
+#define QIB_7322_RcvQPMapTableC_0_RcvQPMapContext13_MSB 0x9
+#define QIB_7322_RcvQPMapTableC_0_RcvQPMapContext13_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableC_0_RcvQPMapContext12_LSB 0x0
+#define QIB_7322_RcvQPMapTableC_0_RcvQPMapContext12_MSB 0x4
+#define QIB_7322_RcvQPMapTableC_0_RcvQPMapContext12_RMASK 0x1F
+
+#define QIB_7322_RcvQPMapTableD_0_OFFS 0x1128
+#define QIB_7322_RcvQPMapTableD_0_DEF 0x0000000000000000
+#define QIB_7322_RcvQPMapTableD_0_RcvQPMapContext23_LSB 0x19
+#define QIB_7322_RcvQPMapTableD_0_RcvQPMapContext23_MSB 0x1D
+#define QIB_7322_RcvQPMapTableD_0_RcvQPMapContext23_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableD_0_RcvQPMapContext22_LSB 0x14
+#define QIB_7322_RcvQPMapTableD_0_RcvQPMapContext22_MSB 0x18
+#define QIB_7322_RcvQPMapTableD_0_RcvQPMapContext22_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableD_0_RcvQPMapContext21_LSB 0xF
+#define QIB_7322_RcvQPMapTableD_0_RcvQPMapContext21_MSB 0x13
+#define QIB_7322_RcvQPMapTableD_0_RcvQPMapContext21_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableD_0_RcvQPMapContext20_LSB 0xA
+#define QIB_7322_RcvQPMapTableD_0_RcvQPMapContext20_MSB 0xE
+#define QIB_7322_RcvQPMapTableD_0_RcvQPMapContext20_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableD_0_RcvQPMapContext19_LSB 0x5
+#define QIB_7322_RcvQPMapTableD_0_RcvQPMapContext19_MSB 0x9
+#define QIB_7322_RcvQPMapTableD_0_RcvQPMapContext19_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableD_0_RcvQPMapContext18_LSB 0x0
+#define QIB_7322_RcvQPMapTableD_0_RcvQPMapContext18_MSB 0x4
+#define QIB_7322_RcvQPMapTableD_0_RcvQPMapContext18_RMASK 0x1F
+
+#define QIB_7322_RcvQPMapTableE_0_OFFS 0x1130
+#define QIB_7322_RcvQPMapTableE_0_DEF 0x0000000000000000
+#define QIB_7322_RcvQPMapTableE_0_RcvQPMapContext29_LSB 0x19
+#define QIB_7322_RcvQPMapTableE_0_RcvQPMapContext29_MSB 0x1D
+#define QIB_7322_RcvQPMapTableE_0_RcvQPMapContext29_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableE_0_RcvQPMapContext28_LSB 0x14
+#define QIB_7322_RcvQPMapTableE_0_RcvQPMapContext28_MSB 0x18
+#define QIB_7322_RcvQPMapTableE_0_RcvQPMapContext28_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableE_0_RcvQPMapContext27_LSB 0xF
+#define QIB_7322_RcvQPMapTableE_0_RcvQPMapContext27_MSB 0x13
+#define QIB_7322_RcvQPMapTableE_0_RcvQPMapContext27_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableE_0_RcvQPMapContext26_LSB 0xA
+#define QIB_7322_RcvQPMapTableE_0_RcvQPMapContext26_MSB 0xE
+#define QIB_7322_RcvQPMapTableE_0_RcvQPMapContext26_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableE_0_RcvQPMapContext25_LSB 0x5
+#define QIB_7322_RcvQPMapTableE_0_RcvQPMapContext25_MSB 0x9
+#define QIB_7322_RcvQPMapTableE_0_RcvQPMapContext25_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableE_0_RcvQPMapContext24_LSB 0x0
+#define QIB_7322_RcvQPMapTableE_0_RcvQPMapContext24_MSB 0x4
+#define QIB_7322_RcvQPMapTableE_0_RcvQPMapContext24_RMASK 0x1F
+
+#define QIB_7322_RcvQPMapTableF_0_OFFS 0x1138
+#define QIB_7322_RcvQPMapTableF_0_DEF 0x0000000000000000
+#define QIB_7322_RcvQPMapTableF_0_RcvQPMapContext31_LSB 0x5
+#define QIB_7322_RcvQPMapTableF_0_RcvQPMapContext31_MSB 0x9
+#define QIB_7322_RcvQPMapTableF_0_RcvQPMapContext31_RMASK 0x1F
+#define QIB_7322_RcvQPMapTableF_0_RcvQPMapContext30_LSB 0x0
+#define QIB_7322_RcvQPMapTableF_0_RcvQPMapContext30_MSB 0x4
+#define QIB_7322_RcvQPMapTableF_0_RcvQPMapContext30_RMASK 0x1F
+
+#define QIB_7322_PSStat_0_OFFS 0x1140
+#define QIB_7322_PSStat_0_DEF 0x0000000000000000
+
+#define QIB_7322_PSStart_0_OFFS 0x1148
+#define QIB_7322_PSStart_0_DEF 0x0000000000000000
+
+#define QIB_7322_PSInterval_0_OFFS 0x1150
+#define QIB_7322_PSInterval_0_DEF 0x0000000000000000
+
+#define QIB_7322_RcvStatus_0_OFFS 0x1160
+#define QIB_7322_RcvStatus_0_DEF 0x0000000000000000
+#define QIB_7322_RcvStatus_0_DmaeqBlockingContext_LSB 0x1
+#define QIB_7322_RcvStatus_0_DmaeqBlockingContext_MSB 0x5
+#define QIB_7322_RcvStatus_0_DmaeqBlockingContext_RMASK 0x1F
+#define QIB_7322_RcvStatus_0_RxPktInProgress_LSB 0x0
+#define QIB_7322_RcvStatus_0_RxPktInProgress_MSB 0x0
+#define QIB_7322_RcvStatus_0_RxPktInProgress_RMASK 0x1
+
+#define QIB_7322_RcvPartitionKey_0_OFFS 0x1168
+#define QIB_7322_RcvPartitionKey_0_DEF 0x0000000000000000
+
+#define QIB_7322_RcvQPMulticastContext_0_OFFS 0x1170
+#define QIB_7322_RcvQPMulticastContext_0_DEF 0x0000000000000000
+#define QIB_7322_RcvQPMulticastContext_0_RcvQpMcContext_LSB 0x0
+#define QIB_7322_RcvQPMulticastContext_0_RcvQpMcContext_MSB 0x4
+#define QIB_7322_RcvQPMulticastContext_0_RcvQpMcContext_RMASK 0x1F
+
+#define QIB_7322_RcvPktLEDCnt_0_OFFS 0x1178
+#define QIB_7322_RcvPktLEDCnt_0_DEF 0x0000000000000000
+#define QIB_7322_RcvPktLEDCnt_0_ONperiod_LSB 0x20
+#define QIB_7322_RcvPktLEDCnt_0_ONperiod_MSB 0x3F
+#define QIB_7322_RcvPktLEDCnt_0_ONperiod_RMASK 0xFFFFFFFF
+#define QIB_7322_RcvPktLEDCnt_0_OFFperiod_LSB 0x0
+#define QIB_7322_RcvPktLEDCnt_0_OFFperiod_MSB 0x1F
+#define QIB_7322_RcvPktLEDCnt_0_OFFperiod_RMASK 0xFFFFFFFF
+
+#define QIB_7322_SendDmaIdleCnt_0_OFFS 0x1180
+#define QIB_7322_SendDmaIdleCnt_0_DEF 0x0000000000000000
+#define QIB_7322_SendDmaIdleCnt_0_SendDmaIdleCnt_LSB 0x0
+#define QIB_7322_SendDmaIdleCnt_0_SendDmaIdleCnt_MSB 0xF
+#define QIB_7322_SendDmaIdleCnt_0_SendDmaIdleCnt_RMASK 0xFFFF
+
+#define QIB_7322_SendDmaReloadCnt_0_OFFS 0x1188
+#define QIB_7322_SendDmaReloadCnt_0_DEF 0x0000000000000000
+#define QIB_7322_SendDmaReloadCnt_0_SendDmaReloadCnt_LSB 0x0
+#define QIB_7322_SendDmaReloadCnt_0_SendDmaReloadCnt_MSB 0xF
+#define QIB_7322_SendDmaReloadCnt_0_SendDmaReloadCnt_RMASK 0xFFFF
+
+#define QIB_7322_SendDmaDescCnt_0_OFFS 0x1190
+#define QIB_7322_SendDmaDescCnt_0_DEF 0x0000000000000000
+#define QIB_7322_SendDmaDescCnt_0_SendDmaDescCnt_LSB 0x0
+#define QIB_7322_SendDmaDescCnt_0_SendDmaDescCnt_MSB 0xF
+#define QIB_7322_SendDmaDescCnt_0_SendDmaDescCnt_RMASK 0xFFFF
+
+#define QIB_7322_SendCtrl_0_OFFS 0x11C0
+#define QIB_7322_SendCtrl_0_DEF 0x0000000000000000
+#define QIB_7322_SendCtrl_0_IBVLArbiterEn_LSB 0xF
+#define QIB_7322_SendCtrl_0_IBVLArbiterEn_MSB 0xF
+#define QIB_7322_SendCtrl_0_IBVLArbiterEn_RMASK 0x1
+#define QIB_7322_SendCtrl_0_TxeDrainRmFifo_LSB 0xE
+#define QIB_7322_SendCtrl_0_TxeDrainRmFifo_MSB 0xE
+#define QIB_7322_SendCtrl_0_TxeDrainRmFifo_RMASK 0x1
+#define QIB_7322_SendCtrl_0_TxeDrainLaFifo_LSB 0xD
+#define QIB_7322_SendCtrl_0_TxeDrainLaFifo_MSB 0xD
+#define QIB_7322_SendCtrl_0_TxeDrainLaFifo_RMASK 0x1
+#define QIB_7322_SendCtrl_0_SDmaHalt_LSB 0xC
+#define QIB_7322_SendCtrl_0_SDmaHalt_MSB 0xC
+#define QIB_7322_SendCtrl_0_SDmaHalt_RMASK 0x1
+#define QIB_7322_SendCtrl_0_SDmaEnable_LSB 0xB
+#define QIB_7322_SendCtrl_0_SDmaEnable_MSB 0xB
+#define QIB_7322_SendCtrl_0_SDmaEnable_RMASK 0x1
+#define QIB_7322_SendCtrl_0_SDmaSingleDescriptor_LSB 0xA
+#define QIB_7322_SendCtrl_0_SDmaSingleDescriptor_MSB 0xA
+#define QIB_7322_SendCtrl_0_SDmaSingleDescriptor_RMASK 0x1
+#define QIB_7322_SendCtrl_0_SDmaIntEnable_LSB 0x9
+#define QIB_7322_SendCtrl_0_SDmaIntEnable_MSB 0x9
+#define QIB_7322_SendCtrl_0_SDmaIntEnable_RMASK 0x1
+#define QIB_7322_SendCtrl_0_SDmaCleanup_LSB 0x8
+#define QIB_7322_SendCtrl_0_SDmaCleanup_MSB 0x8
+#define QIB_7322_SendCtrl_0_SDmaCleanup_RMASK 0x1
+#define QIB_7322_SendCtrl_0_ForceCreditUpToDate_LSB 0x7
+#define QIB_7322_SendCtrl_0_ForceCreditUpToDate_MSB 0x7
+#define QIB_7322_SendCtrl_0_ForceCreditUpToDate_RMASK 0x1
+#define QIB_7322_SendCtrl_0_SendEnable_LSB 0x3
+#define QIB_7322_SendCtrl_0_SendEnable_MSB 0x3
+#define QIB_7322_SendCtrl_0_SendEnable_RMASK 0x1
+#define QIB_7322_SendCtrl_0_TxeBypassIbc_LSB 0x1
+#define QIB_7322_SendCtrl_0_TxeBypassIbc_MSB 0x1
+#define QIB_7322_SendCtrl_0_TxeBypassIbc_RMASK 0x1
+#define QIB_7322_SendCtrl_0_TxeAbortIbc_LSB 0x0
+#define QIB_7322_SendCtrl_0_TxeAbortIbc_MSB 0x0
+#define QIB_7322_SendCtrl_0_TxeAbortIbc_RMASK 0x1
+
+#define QIB_7322_SendDmaBase_0_OFFS 0x11F8
+#define QIB_7322_SendDmaBase_0_DEF 0x0000000000000000
+#define QIB_7322_SendDmaBase_0_SendDmaBase_LSB 0x0
+#define QIB_7322_SendDmaBase_0_SendDmaBase_MSB 0x2F
+#define QIB_7322_SendDmaBase_0_SendDmaBase_RMASK 0xFFFFFFFFFFFF
+
+#define QIB_7322_SendDmaLenGen_0_OFFS 0x1200
+#define QIB_7322_SendDmaLenGen_0_DEF 0x0000000000000000
+#define QIB_7322_SendDmaLenGen_0_Generation_LSB 0x10
+#define QIB_7322_SendDmaLenGen_0_Generation_MSB 0x12
+#define QIB_7322_SendDmaLenGen_0_Generation_RMASK 0x7
+#define QIB_7322_SendDmaLenGen_0_Length_LSB 0x0
+#define QIB_7322_SendDmaLenGen_0_Length_MSB 0xF
+#define QIB_7322_SendDmaLenGen_0_Length_RMASK 0xFFFF
+
+#define QIB_7322_SendDmaTail_0_OFFS 0x1208
+#define QIB_7322_SendDmaTail_0_DEF 0x0000000000000000
+#define QIB_7322_SendDmaTail_0_SendDmaTail_LSB 0x0
+#define QIB_7322_SendDmaTail_0_SendDmaTail_MSB 0xF
+#define QIB_7322_SendDmaTail_0_SendDmaTail_RMASK 0xFFFF
+
+#define QIB_7322_SendDmaHead_0_OFFS 0x1210
+#define QIB_7322_SendDmaHead_0_DEF 0x0000000000000000
+#define QIB_7322_SendDmaHead_0_InternalSendDmaHead_LSB 0x20
+#define QIB_7322_SendDmaHead_0_InternalSendDmaHead_MSB 0x2F
+#define QIB_7322_SendDmaHead_0_InternalSendDmaHead_RMASK 0xFFFF
+#define QIB_7322_SendDmaHead_0_SendDmaHead_LSB 0x0
+#define QIB_7322_SendDmaHead_0_SendDmaHead_MSB 0xF
+#define QIB_7322_SendDmaHead_0_SendDmaHead_RMASK 0xFFFF
+
+#define QIB_7322_SendDmaHeadAddr_0_OFFS 0x1218
+#define QIB_7322_SendDmaHeadAddr_0_DEF 0x0000000000000000
+#define QIB_7322_SendDmaHeadAddr_0_SendDmaHeadAddr_LSB 0x0
+#define QIB_7322_SendDmaHeadAddr_0_SendDmaHeadAddr_MSB 0x2F
+#define QIB_7322_SendDmaHeadAddr_0_SendDmaHeadAddr_RMASK 0xFFFFFFFFFFFF
+
+#define QIB_7322_SendDmaBufMask0_0_OFFS 0x1220
+#define QIB_7322_SendDmaBufMask0_0_DEF 0x0000000000000000
+#define QIB_7322_SendDmaBufMask0_0_BufMask_63_0_LSB 0x0
+#define QIB_7322_SendDmaBufMask0_0_BufMask_63_0_MSB 0x3F
+#define QIB_7322_SendDmaBufMask0_0_BufMask_63_0_RMASK 0x0
+
+#define QIB_7322_SendDmaStatus_0_OFFS 0x1238
+#define QIB_7322_SendDmaStatus_0_DEF 0x0000000042000000
+#define QIB_7322_SendDmaStatus_0_ScoreBoardDrainInProg_LSB 0x3F
+#define QIB_7322_SendDmaStatus_0_ScoreBoardDrainInProg_MSB 0x3F
+#define QIB_7322_SendDmaStatus_0_ScoreBoardDrainInProg_RMASK 0x1
+#define QIB_7322_SendDmaStatus_0_HaltInProg_LSB 0x3E
+#define QIB_7322_SendDmaStatus_0_HaltInProg_MSB 0x3E
+#define QIB_7322_SendDmaStatus_0_HaltInProg_RMASK 0x1
+#define QIB_7322_SendDmaStatus_0_InternalSDmaHalt_LSB 0x3D
+#define QIB_7322_SendDmaStatus_0_InternalSDmaHalt_MSB 0x3D
+#define QIB_7322_SendDmaStatus_0_InternalSDmaHalt_RMASK 0x1
+#define QIB_7322_SendDmaStatus_0_ScbDescIndex_13_0_LSB 0x2F
+#define QIB_7322_SendDmaStatus_0_ScbDescIndex_13_0_MSB 0x3C
+#define QIB_7322_SendDmaStatus_0_ScbDescIndex_13_0_RMASK 0x3FFF
+#define QIB_7322_SendDmaStatus_0_RpyLowAddr_6_0_LSB 0x28
+#define QIB_7322_SendDmaStatus_0_RpyLowAddr_6_0_MSB 0x2E
+#define QIB_7322_SendDmaStatus_0_RpyLowAddr_6_0_RMASK 0x7F
+#define QIB_7322_SendDmaStatus_0_RpyTag_7_0_LSB 0x20
+#define QIB_7322_SendDmaStatus_0_RpyTag_7_0_MSB 0x27
+#define QIB_7322_SendDmaStatus_0_RpyTag_7_0_RMASK 0xFF
+#define QIB_7322_SendDmaStatus_0_ScbFull_LSB 0x1F
+#define QIB_7322_SendDmaStatus_0_ScbFull_MSB 0x1F
+#define QIB_7322_SendDmaStatus_0_ScbFull_RMASK 0x1
+#define QIB_7322_SendDmaStatus_0_ScbEmpty_LSB 0x1E
+#define QIB_7322_SendDmaStatus_0_ScbEmpty_MSB 0x1E
+#define QIB_7322_SendDmaStatus_0_ScbEmpty_RMASK 0x1
+#define QIB_7322_SendDmaStatus_0_ScbEntryValid_LSB 0x1D
+#define QIB_7322_SendDmaStatus_0_ScbEntryValid_MSB 0x1D
+#define QIB_7322_SendDmaStatus_0_ScbEntryValid_RMASK 0x1
+#define QIB_7322_SendDmaStatus_0_ScbFetchDescFlag_LSB 0x1C
+#define QIB_7322_SendDmaStatus_0_ScbFetchDescFlag_MSB 0x1C
+#define QIB_7322_SendDmaStatus_0_ScbFetchDescFlag_RMASK 0x1
+#define QIB_7322_SendDmaStatus_0_SplFifoReadyToGo_LSB 0x1B
+#define QIB_7322_SendDmaStatus_0_SplFifoReadyToGo_MSB 0x1B
+#define QIB_7322_SendDmaStatus_0_SplFifoReadyToGo_RMASK 0x1
+#define QIB_7322_SendDmaStatus_0_SplFifoDisarmed_LSB 0x1A
+#define QIB_7322_SendDmaStatus_0_SplFifoDisarmed_MSB 0x1A
+#define QIB_7322_SendDmaStatus_0_SplFifoDisarmed_RMASK 0x1
+#define QIB_7322_SendDmaStatus_0_SplFifoEmpty_LSB 0x19
+#define QIB_7322_SendDmaStatus_0_SplFifoEmpty_MSB 0x19
+#define QIB_7322_SendDmaStatus_0_SplFifoEmpty_RMASK 0x1
+#define QIB_7322_SendDmaStatus_0_SplFifoFull_LSB 0x18
+#define QIB_7322_SendDmaStatus_0_SplFifoFull_MSB 0x18
+#define QIB_7322_SendDmaStatus_0_SplFifoFull_RMASK 0x1
+#define QIB_7322_SendDmaStatus_0_SplFifoBufNum_LSB 0x10
+#define QIB_7322_SendDmaStatus_0_SplFifoBufNum_MSB 0x17
+#define QIB_7322_SendDmaStatus_0_SplFifoBufNum_RMASK 0xFF
+#define QIB_7322_SendDmaStatus_0_SplFifoDescIndex_LSB 0x0
+#define QIB_7322_SendDmaStatus_0_SplFifoDescIndex_MSB 0xF
+#define QIB_7322_SendDmaStatus_0_SplFifoDescIndex_RMASK 0xFFFF
+
+#define QIB_7322_SendDmaPriorityThld_0_OFFS 0x1258
+#define QIB_7322_SendDmaPriorityThld_0_DEF 0x0000000000000000
+#define QIB_7322_SendDmaPriorityThld_0_PriorityThreshold_LSB 0x0
+#define QIB_7322_SendDmaPriorityThld_0_PriorityThreshold_MSB 0x3
+#define QIB_7322_SendDmaPriorityThld_0_PriorityThreshold_RMASK 0xF
+
+#define QIB_7322_SendHdrErrSymptom_0_OFFS 0x1260
+#define QIB_7322_SendHdrErrSymptom_0_DEF 0x0000000000000000
+#define QIB_7322_SendHdrErrSymptom_0_NonKeyPacket_LSB 0x6
+#define QIB_7322_SendHdrErrSymptom_0_NonKeyPacket_MSB 0x6
+#define QIB_7322_SendHdrErrSymptom_0_NonKeyPacket_RMASK 0x1
+#define QIB_7322_SendHdrErrSymptom_0_GRHFail_LSB 0x5
+#define QIB_7322_SendHdrErrSymptom_0_GRHFail_MSB 0x5
+#define QIB_7322_SendHdrErrSymptom_0_GRHFail_RMASK 0x1
+#define QIB_7322_SendHdrErrSymptom_0_PkeyFail_LSB 0x4
+#define QIB_7322_SendHdrErrSymptom_0_PkeyFail_MSB 0x4
+#define QIB_7322_SendHdrErrSymptom_0_PkeyFail_RMASK 0x1
+#define QIB_7322_SendHdrErrSymptom_0_QPFail_LSB 0x3
+#define QIB_7322_SendHdrErrSymptom_0_QPFail_MSB 0x3
+#define QIB_7322_SendHdrErrSymptom_0_QPFail_RMASK 0x1
+#define QIB_7322_SendHdrErrSymptom_0_SLIDFail_LSB 0x2
+#define QIB_7322_SendHdrErrSymptom_0_SLIDFail_MSB 0x2
+#define QIB_7322_SendHdrErrSymptom_0_SLIDFail_RMASK 0x1
+#define QIB_7322_SendHdrErrSymptom_0_RawIPV6_LSB 0x1
+#define QIB_7322_SendHdrErrSymptom_0_RawIPV6_MSB 0x1
+#define QIB_7322_SendHdrErrSymptom_0_RawIPV6_RMASK 0x1
+#define QIB_7322_SendHdrErrSymptom_0_PacketTooSmall_LSB 0x0
+#define QIB_7322_SendHdrErrSymptom_0_PacketTooSmall_MSB 0x0
+#define QIB_7322_SendHdrErrSymptom_0_PacketTooSmall_RMASK 0x1
+
+#define QIB_7322_RxCreditVL0_0_OFFS 0x1280
+#define QIB_7322_RxCreditVL0_0_DEF 0x0000000000000000
+#define QIB_7322_RxCreditVL0_0_RxBufrConsumedVL_LSB 0x10
+#define QIB_7322_RxCreditVL0_0_RxBufrConsumedVL_MSB 0x1B
+#define QIB_7322_RxCreditVL0_0_RxBufrConsumedVL_RMASK 0xFFF
+#define QIB_7322_RxCreditVL0_0_RxMaxCreditVL_LSB 0x0
+#define QIB_7322_RxCreditVL0_0_RxMaxCreditVL_MSB 0xB
+#define QIB_7322_RxCreditVL0_0_RxMaxCreditVL_RMASK 0xFFF
+
+#define QIB_7322_SendDmaBufUsed0_0_OFFS 0x1480
+#define QIB_7322_SendDmaBufUsed0_0_DEF 0x0000000000000000
+#define QIB_7322_SendDmaBufUsed0_0_BufUsed_63_0_LSB 0x0
+#define QIB_7322_SendDmaBufUsed0_0_BufUsed_63_0_MSB 0x3F
+#define QIB_7322_SendDmaBufUsed0_0_BufUsed_63_0_RMASK 0x0
+
+#define QIB_7322_SendCheckControl_0_OFFS 0x14A8
+#define QIB_7322_SendCheckControl_0_DEF 0x0000000000000000
+#define QIB_7322_SendCheckControl_0_PKey_En_LSB 0x4
+#define QIB_7322_SendCheckControl_0_PKey_En_MSB 0x4
+#define QIB_7322_SendCheckControl_0_PKey_En_RMASK 0x1
+#define QIB_7322_SendCheckControl_0_BTHQP_En_LSB 0x3
+#define QIB_7322_SendCheckControl_0_BTHQP_En_MSB 0x3
+#define QIB_7322_SendCheckControl_0_BTHQP_En_RMASK 0x1
+#define QIB_7322_SendCheckControl_0_SLID_En_LSB 0x2
+#define QIB_7322_SendCheckControl_0_SLID_En_MSB 0x2
+#define QIB_7322_SendCheckControl_0_SLID_En_RMASK 0x1
+#define QIB_7322_SendCheckControl_0_RawIPV6_En_LSB 0x1
+#define QIB_7322_SendCheckControl_0_RawIPV6_En_MSB 0x1
+#define QIB_7322_SendCheckControl_0_RawIPV6_En_RMASK 0x1
+#define QIB_7322_SendCheckControl_0_PacketTooSmall_En_LSB 0x0
+#define QIB_7322_SendCheckControl_0_PacketTooSmall_En_MSB 0x0
+#define QIB_7322_SendCheckControl_0_PacketTooSmall_En_RMASK 0x1
+
+#define QIB_7322_SendIBSLIDMask_0_OFFS 0x14B0
+#define QIB_7322_SendIBSLIDMask_0_DEF 0x0000000000000000
+#define QIB_7322_SendIBSLIDMask_0_SendIBSLIDMask_15_0_LSB 0x0
+#define QIB_7322_SendIBSLIDMask_0_SendIBSLIDMask_15_0_MSB 0xF
+#define QIB_7322_SendIBSLIDMask_0_SendIBSLIDMask_15_0_RMASK 0xFFFF
+
+#define QIB_7322_SendIBSLIDAssign_0_OFFS 0x14B8
+#define QIB_7322_SendIBSLIDAssign_0_DEF 0x0000000000000000
+#define QIB_7322_SendIBSLIDAssign_0_SendIBSLIDAssign_15_0_LSB 0x0
+#define QIB_7322_SendIBSLIDAssign_0_SendIBSLIDAssign_15_0_MSB 0xF
+#define QIB_7322_SendIBSLIDAssign_0_SendIBSLIDAssign_15_0_RMASK 0xFFFF
+
+#define QIB_7322_IBCStatusA_0_OFFS 0x1540
+#define QIB_7322_IBCStatusA_0_DEF 0x0000000000000X02
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL7_LSB 0x27
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL7_MSB 0x27
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL7_RMASK 0x1
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL6_LSB 0x26
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL6_MSB 0x26
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL6_RMASK 0x1
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL5_LSB 0x25
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL5_MSB 0x25
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL5_RMASK 0x1
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL4_LSB 0x24
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL4_MSB 0x24
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL4_RMASK 0x1
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL3_LSB 0x23
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL3_MSB 0x23
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL3_RMASK 0x1
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL2_LSB 0x22
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL2_MSB 0x22
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL2_RMASK 0x1
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL1_LSB 0x21
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL1_MSB 0x21
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL1_RMASK 0x1
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL0_LSB 0x20
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL0_MSB 0x20
+#define QIB_7322_IBCStatusA_0_TxCreditOk_VL0_RMASK 0x1
+#define QIB_7322_IBCStatusA_0_TxReady_LSB 0x1E
+#define QIB_7322_IBCStatusA_0_TxReady_MSB 0x1E
+#define QIB_7322_IBCStatusA_0_TxReady_RMASK 0x1
+#define QIB_7322_IBCStatusA_0_LinkSpeedQDR_LSB 0x1D
+#define QIB_7322_IBCStatusA_0_LinkSpeedQDR_MSB 0x1D
+#define QIB_7322_IBCStatusA_0_LinkSpeedQDR_RMASK 0x1
+#define QIB_7322_IBCStatusA_0_ScrambleCapRemote_LSB 0xF
+#define QIB_7322_IBCStatusA_0_ScrambleCapRemote_MSB 0xF
+#define QIB_7322_IBCStatusA_0_ScrambleCapRemote_RMASK 0x1
+#define QIB_7322_IBCStatusA_0_ScrambleEn_LSB 0xE
+#define QIB_7322_IBCStatusA_0_ScrambleEn_MSB 0xE
+#define QIB_7322_IBCStatusA_0_ScrambleEn_RMASK 0x1
+#define QIB_7322_IBCStatusA_0_IBTxLaneReversed_LSB 0xD
+#define QIB_7322_IBCStatusA_0_IBTxLaneReversed_MSB 0xD
+#define QIB_7322_IBCStatusA_0_IBTxLaneReversed_RMASK 0x1
+#define QIB_7322_IBCStatusA_0_IBRxLaneReversed_LSB 0xC
+#define QIB_7322_IBCStatusA_0_IBRxLaneReversed_MSB 0xC
+#define QIB_7322_IBCStatusA_0_IBRxLaneReversed_RMASK 0x1
+#define QIB_7322_IBCStatusA_0_DDS_RXEQ_FAIL_LSB 0xA
+#define QIB_7322_IBCStatusA_0_DDS_RXEQ_FAIL_MSB 0xA
+#define QIB_7322_IBCStatusA_0_DDS_RXEQ_FAIL_RMASK 0x1
+#define QIB_7322_IBCStatusA_0_LinkWidthActive_LSB 0x9
+#define QIB_7322_IBCStatusA_0_LinkWidthActive_MSB 0x9
+#define QIB_7322_IBCStatusA_0_LinkWidthActive_RMASK 0x1
+#define QIB_7322_IBCStatusA_0_LinkSpeedActive_LSB 0x8
+#define QIB_7322_IBCStatusA_0_LinkSpeedActive_MSB 0x8
+#define QIB_7322_IBCStatusA_0_LinkSpeedActive_RMASK 0x1
+#define QIB_7322_IBCStatusA_0_LinkState_LSB 0x5
+#define QIB_7322_IBCStatusA_0_LinkState_MSB 0x7
+#define QIB_7322_IBCStatusA_0_LinkState_RMASK 0x7
+#define QIB_7322_IBCStatusA_0_LinkTrainingState_LSB 0x0
+#define QIB_7322_IBCStatusA_0_LinkTrainingState_MSB 0x4
+#define QIB_7322_IBCStatusA_0_LinkTrainingState_RMASK 0x1F
+
+#define QIB_7322_IBCStatusB_0_OFFS 0x1548
+#define QIB_7322_IBCStatusB_0_DEF 0x00000000XXXXXXXX
+#define QIB_7322_IBCStatusB_0_ibsd_adaptation_timer_debug_LSB 0x27
+#define QIB_7322_IBCStatusB_0_ibsd_adaptation_timer_debug_MSB 0x27
+#define QIB_7322_IBCStatusB_0_ibsd_adaptation_timer_debug_RMASK 0x1
+#define QIB_7322_IBCStatusB_0_ibsd_adaptation_timer_reached_threshold_LSB 0x26
+#define QIB_7322_IBCStatusB_0_ibsd_adaptation_timer_reached_threshold_MSB 0x26
+#define QIB_7322_IBCStatusB_0_ibsd_adaptation_timer_reached_threshold_RMASK 0x1
+#define QIB_7322_IBCStatusB_0_ibsd_adaptation_timer_started_LSB 0x25
+#define QIB_7322_IBCStatusB_0_ibsd_adaptation_timer_started_MSB 0x25
+#define QIB_7322_IBCStatusB_0_ibsd_adaptation_timer_started_RMASK 0x1
+#define QIB_7322_IBCStatusB_0_heartbeat_timed_out_LSB 0x24
+#define QIB_7322_IBCStatusB_0_heartbeat_timed_out_MSB 0x24
+#define QIB_7322_IBCStatusB_0_heartbeat_timed_out_RMASK 0x1
+#define QIB_7322_IBCStatusB_0_heartbeat_crosstalk_LSB 0x20
+#define QIB_7322_IBCStatusB_0_heartbeat_crosstalk_MSB 0x23
+#define QIB_7322_IBCStatusB_0_heartbeat_crosstalk_RMASK 0xF
+#define QIB_7322_IBCStatusB_0_RxEqLocalDevice_LSB 0x1E
+#define QIB_7322_IBCStatusB_0_RxEqLocalDevice_MSB 0x1F
+#define QIB_7322_IBCStatusB_0_RxEqLocalDevice_RMASK 0x3
+#define QIB_7322_IBCStatusB_0_ReqDDSLocalFromRmt_LSB 0x1A
+#define QIB_7322_IBCStatusB_0_ReqDDSLocalFromRmt_MSB 0x1D
+#define QIB_7322_IBCStatusB_0_ReqDDSLocalFromRmt_RMASK 0xF
+#define QIB_7322_IBCStatusB_0_LinkRoundTripLatency_LSB 0x0
+#define QIB_7322_IBCStatusB_0_LinkRoundTripLatency_MSB 0x19
+#define QIB_7322_IBCStatusB_0_LinkRoundTripLatency_RMASK 0x3FFFFFF
+
+#define QIB_7322_IBCCtrlA_0_OFFS 0x1560
+#define QIB_7322_IBCCtrlA_0_DEF 0x0000000000000000
+#define QIB_7322_IBCCtrlA_0_Loopback_LSB 0x3F
+#define QIB_7322_IBCCtrlA_0_Loopback_MSB 0x3F
+#define QIB_7322_IBCCtrlA_0_Loopback_RMASK 0x1
+#define QIB_7322_IBCCtrlA_0_LinkDownDefaultState_LSB 0x3E
+#define QIB_7322_IBCCtrlA_0_LinkDownDefaultState_MSB 0x3E
+#define QIB_7322_IBCCtrlA_0_LinkDownDefaultState_RMASK 0x1
+#define QIB_7322_IBCCtrlA_0_IBLinkEn_LSB 0x3D
+#define QIB_7322_IBCCtrlA_0_IBLinkEn_MSB 0x3D
+#define QIB_7322_IBCCtrlA_0_IBLinkEn_RMASK 0x1
+#define QIB_7322_IBCCtrlA_0_IBStatIntReductionEn_LSB 0x3C
+#define QIB_7322_IBCCtrlA_0_IBStatIntReductionEn_MSB 0x3C
+#define QIB_7322_IBCCtrlA_0_IBStatIntReductionEn_RMASK 0x1
+#define QIB_7322_IBCCtrlA_0_NumVLane_LSB 0x30
+#define QIB_7322_IBCCtrlA_0_NumVLane_MSB 0x32
+#define QIB_7322_IBCCtrlA_0_NumVLane_RMASK 0x7
+#define QIB_7322_IBCCtrlA_0_OverrunThreshold_LSB 0x24
+#define QIB_7322_IBCCtrlA_0_OverrunThreshold_MSB 0x27
+#define QIB_7322_IBCCtrlA_0_OverrunThreshold_RMASK 0xF
+#define QIB_7322_IBCCtrlA_0_PhyerrThreshold_LSB 0x20
+#define QIB_7322_IBCCtrlA_0_PhyerrThreshold_MSB 0x23
+#define QIB_7322_IBCCtrlA_0_PhyerrThreshold_RMASK 0xF
+#define QIB_7322_IBCCtrlA_0_MaxPktLen_LSB 0x15
+#define QIB_7322_IBCCtrlA_0_MaxPktLen_MSB 0x1F
+#define QIB_7322_IBCCtrlA_0_MaxPktLen_RMASK 0x7FF
+#define QIB_7322_IBCCtrlA_0_LinkCmd_LSB 0x13
+#define QIB_7322_IBCCtrlA_0_LinkCmd_MSB 0x14
+#define QIB_7322_IBCCtrlA_0_LinkCmd_RMASK 0x3
+#define QIB_7322_IBCCtrlA_0_LinkInitCmd_LSB 0x10
+#define QIB_7322_IBCCtrlA_0_LinkInitCmd_MSB 0x12
+#define QIB_7322_IBCCtrlA_0_LinkInitCmd_RMASK 0x7
+#define QIB_7322_IBCCtrlA_0_FlowCtrlWaterMark_LSB 0x8
+#define QIB_7322_IBCCtrlA_0_FlowCtrlWaterMark_MSB 0xF
+#define QIB_7322_IBCCtrlA_0_FlowCtrlWaterMark_RMASK 0xFF
+#define QIB_7322_IBCCtrlA_0_FlowCtrlPeriod_LSB 0x0
+#define QIB_7322_IBCCtrlA_0_FlowCtrlPeriod_MSB 0x7
+#define QIB_7322_IBCCtrlA_0_FlowCtrlPeriod_RMASK 0xFF
+
+#define QIB_7322_IBCCtrlB_0_OFFS 0x1568
+#define QIB_7322_IBCCtrlB_0_DEF 0x00000000000305FF
+#define QIB_7322_IBCCtrlB_0_IB_DLID_MASK_LSB 0x30
+#define QIB_7322_IBCCtrlB_0_IB_DLID_MASK_MSB 0x3F
+#define QIB_7322_IBCCtrlB_0_IB_DLID_MASK_RMASK 0xFFFF
+#define QIB_7322_IBCCtrlB_0_IB_DLID_LSB 0x20
+#define QIB_7322_IBCCtrlB_0_IB_DLID_MSB 0x2F
+#define QIB_7322_IBCCtrlB_0_IB_DLID_RMASK 0xFFFF
+#define QIB_7322_IBCCtrlB_0_IB_ENABLE_FILT_DPKT_LSB 0x1B
+#define QIB_7322_IBCCtrlB_0_IB_ENABLE_FILT_DPKT_MSB 0x1B
+#define QIB_7322_IBCCtrlB_0_IB_ENABLE_FILT_DPKT_RMASK 0x1
+#define QIB_7322_IBCCtrlB_0_HRTBT_REQ_LSB 0x1A
+#define QIB_7322_IBCCtrlB_0_HRTBT_REQ_MSB 0x1A
+#define QIB_7322_IBCCtrlB_0_HRTBT_REQ_RMASK 0x1
+#define QIB_7322_IBCCtrlB_0_HRTBT_PORT_LSB 0x12
+#define QIB_7322_IBCCtrlB_0_HRTBT_PORT_MSB 0x19
+#define QIB_7322_IBCCtrlB_0_HRTBT_PORT_RMASK 0xFF
+#define QIB_7322_IBCCtrlB_0_HRTBT_AUTO_LSB 0x11
+#define QIB_7322_IBCCtrlB_0_HRTBT_AUTO_MSB 0x11
+#define QIB_7322_IBCCtrlB_0_HRTBT_AUTO_RMASK 0x1
+#define QIB_7322_IBCCtrlB_0_HRTBT_ENB_LSB 0x10
+#define QIB_7322_IBCCtrlB_0_HRTBT_ENB_MSB 0x10
+#define QIB_7322_IBCCtrlB_0_HRTBT_ENB_RMASK 0x1
+#define QIB_7322_IBCCtrlB_0_SD_DDS_LSB 0xC
+#define QIB_7322_IBCCtrlB_0_SD_DDS_MSB 0xF
+#define QIB_7322_IBCCtrlB_0_SD_DDS_RMASK 0xF
+#define QIB_7322_IBCCtrlB_0_SD_DDSV_LSB 0xB
+#define QIB_7322_IBCCtrlB_0_SD_DDSV_MSB 0xB
+#define QIB_7322_IBCCtrlB_0_SD_DDSV_RMASK 0x1
+#define QIB_7322_IBCCtrlB_0_SD_ADD_ENB_LSB 0xA
+#define QIB_7322_IBCCtrlB_0_SD_ADD_ENB_MSB 0xA
+#define QIB_7322_IBCCtrlB_0_SD_ADD_ENB_RMASK 0x1
+#define QIB_7322_IBCCtrlB_0_SD_RX_EQUAL_ENABLE_LSB 0x9
+#define QIB_7322_IBCCtrlB_0_SD_RX_EQUAL_ENABLE_MSB 0x9
+#define QIB_7322_IBCCtrlB_0_SD_RX_EQUAL_ENABLE_RMASK 0x1
+#define QIB_7322_IBCCtrlB_0_IB_LANE_REV_SUPPORTED_LSB 0x8
+#define QIB_7322_IBCCtrlB_0_IB_LANE_REV_SUPPORTED_MSB 0x8
+#define QIB_7322_IBCCtrlB_0_IB_LANE_REV_SUPPORTED_RMASK 0x1
+#define QIB_7322_IBCCtrlB_0_IB_POLARITY_REV_SUPP_LSB 0x7
+#define QIB_7322_IBCCtrlB_0_IB_POLARITY_REV_SUPP_MSB 0x7
+#define QIB_7322_IBCCtrlB_0_IB_POLARITY_REV_SUPP_RMASK 0x1
+#define QIB_7322_IBCCtrlB_0_IB_NUM_CHANNELS_LSB 0x5
+#define QIB_7322_IBCCtrlB_0_IB_NUM_CHANNELS_MSB 0x6
+#define QIB_7322_IBCCtrlB_0_IB_NUM_CHANNELS_RMASK 0x3
+#define QIB_7322_IBCCtrlB_0_SD_SPEED_QDR_LSB 0x4
+#define QIB_7322_IBCCtrlB_0_SD_SPEED_QDR_MSB 0x4
+#define QIB_7322_IBCCtrlB_0_SD_SPEED_QDR_RMASK 0x1
+#define QIB_7322_IBCCtrlB_0_SD_SPEED_DDR_LSB 0x3
+#define QIB_7322_IBCCtrlB_0_SD_SPEED_DDR_MSB 0x3
+#define QIB_7322_IBCCtrlB_0_SD_SPEED_DDR_RMASK 0x1
+#define QIB_7322_IBCCtrlB_0_SD_SPEED_SDR_LSB 0x2
+#define QIB_7322_IBCCtrlB_0_SD_SPEED_SDR_MSB 0x2
+#define QIB_7322_IBCCtrlB_0_SD_SPEED_SDR_RMASK 0x1
+#define QIB_7322_IBCCtrlB_0_SD_SPEED_LSB 0x1
+#define QIB_7322_IBCCtrlB_0_SD_SPEED_MSB 0x1
+#define QIB_7322_IBCCtrlB_0_SD_SPEED_RMASK 0x1
+#define QIB_7322_IBCCtrlB_0_IB_ENHANCED_MODE_LSB 0x0
+#define QIB_7322_IBCCtrlB_0_IB_ENHANCED_MODE_MSB 0x0
+#define QIB_7322_IBCCtrlB_0_IB_ENHANCED_MODE_RMASK 0x1
+
+#define QIB_7322_IBCCtrlC_0_OFFS 0x1570
+#define QIB_7322_IBCCtrlC_0_DEF 0x0000000000000301
+#define QIB_7322_IBCCtrlC_0_IB_BACK_PORCH_LSB 0x5
+#define QIB_7322_IBCCtrlC_0_IB_BACK_PORCH_MSB 0x9
+#define QIB_7322_IBCCtrlC_0_IB_BACK_PORCH_RMASK 0x1F
+#define QIB_7322_IBCCtrlC_0_IB_FRONT_PORCH_LSB 0x0
+#define QIB_7322_IBCCtrlC_0_IB_FRONT_PORCH_MSB 0x4
+#define QIB_7322_IBCCtrlC_0_IB_FRONT_PORCH_RMASK 0x1F
+
+#define QIB_7322_HRTBT_GUID_0_OFFS 0x1588
+#define QIB_7322_HRTBT_GUID_0_DEF 0x0000000000000000
+
+#define QIB_7322_IB_SDTEST_IF_TX_0_OFFS 0x1590
+#define QIB_7322_IB_SDTEST_IF_TX_0_DEF 0x0000000000000000
+#define QIB_7322_IB_SDTEST_IF_TX_0_TS_TX_RX_CFG_LSB 0x30
+#define QIB_7322_IB_SDTEST_IF_TX_0_TS_TX_RX_CFG_MSB 0x3F
+#define QIB_7322_IB_SDTEST_IF_TX_0_TS_TX_RX_CFG_RMASK 0xFFFF
+#define QIB_7322_IB_SDTEST_IF_TX_0_TS_TX_TX_CFG_LSB 0x20
+#define QIB_7322_IB_SDTEST_IF_TX_0_TS_TX_TX_CFG_MSB 0x2F
+#define QIB_7322_IB_SDTEST_IF_TX_0_TS_TX_TX_CFG_RMASK 0xFFFF
+#define QIB_7322_IB_SDTEST_IF_TX_0_TS_TX_SPEED_LSB 0xD
+#define QIB_7322_IB_SDTEST_IF_TX_0_TS_TX_SPEED_MSB 0xF
+#define QIB_7322_IB_SDTEST_IF_TX_0_TS_TX_SPEED_RMASK 0x7
+#define QIB_7322_IB_SDTEST_IF_TX_0_TS_TX_OPCODE_LSB 0xB
+#define QIB_7322_IB_SDTEST_IF_TX_0_TS_TX_OPCODE_MSB 0xC
+#define QIB_7322_IB_SDTEST_IF_TX_0_TS_TX_OPCODE_RMASK 0x3
+#define QIB_7322_IB_SDTEST_IF_TX_0_CREDIT_CHANGE_LSB 0x4
+#define QIB_7322_IB_SDTEST_IF_TX_0_CREDIT_CHANGE_MSB 0x4
+#define QIB_7322_IB_SDTEST_IF_TX_0_CREDIT_CHANGE_RMASK 0x1
+#define QIB_7322_IB_SDTEST_IF_TX_0_VL_CAP_LSB 0x2
+#define QIB_7322_IB_SDTEST_IF_TX_0_VL_CAP_MSB 0x3
+#define QIB_7322_IB_SDTEST_IF_TX_0_VL_CAP_RMASK 0x3
+#define QIB_7322_IB_SDTEST_IF_TX_0_TS_3_TX_VALID_LSB 0x1
+#define QIB_7322_IB_SDTEST_IF_TX_0_TS_3_TX_VALID_MSB 0x1
+#define QIB_7322_IB_SDTEST_IF_TX_0_TS_3_TX_VALID_RMASK 0x1
+#define QIB_7322_IB_SDTEST_IF_TX_0_TS_T_TX_VALID_LSB 0x0
+#define QIB_7322_IB_SDTEST_IF_TX_0_TS_T_TX_VALID_MSB 0x0
+#define QIB_7322_IB_SDTEST_IF_TX_0_TS_T_TX_VALID_RMASK 0x1
+
+#define QIB_7322_IB_SDTEST_IF_RX_0_OFFS 0x1598
+#define QIB_7322_IB_SDTEST_IF_RX_0_DEF 0x0000000000000000
+#define QIB_7322_IB_SDTEST_IF_RX_0_TS_RX_RX_CFG_LSB 0x30
+#define QIB_7322_IB_SDTEST_IF_RX_0_TS_RX_RX_CFG_MSB 0x3F
+#define QIB_7322_IB_SDTEST_IF_RX_0_TS_RX_RX_CFG_RMASK 0xFFFF
+#define QIB_7322_IB_SDTEST_IF_RX_0_TS_RX_TX_CFG_LSB 0x20
+#define QIB_7322_IB_SDTEST_IF_RX_0_TS_RX_TX_CFG_MSB 0x2F
+#define QIB_7322_IB_SDTEST_IF_RX_0_TS_RX_TX_CFG_RMASK 0xFFFF
+#define QIB_7322_IB_SDTEST_IF_RX_0_TS_RX_B_LSB 0x18
+#define QIB_7322_IB_SDTEST_IF_RX_0_TS_RX_B_MSB 0x1F
+#define QIB_7322_IB_SDTEST_IF_RX_0_TS_RX_B_RMASK 0xFF
+#define QIB_7322_IB_SDTEST_IF_RX_0_TS_RX_A_LSB 0x10
+#define QIB_7322_IB_SDTEST_IF_RX_0_TS_RX_A_MSB 0x17
+#define QIB_7322_IB_SDTEST_IF_RX_0_TS_RX_A_RMASK 0xFF
+#define QIB_7322_IB_SDTEST_IF_RX_0_TS_3_RX_VALID_LSB 0x1
+#define QIB_7322_IB_SDTEST_IF_RX_0_TS_3_RX_VALID_MSB 0x1
+#define QIB_7322_IB_SDTEST_IF_RX_0_TS_3_RX_VALID_RMASK 0x1
+#define QIB_7322_IB_SDTEST_IF_RX_0_TS_T_RX_VALID_LSB 0x0
+#define QIB_7322_IB_SDTEST_IF_RX_0_TS_T_RX_VALID_MSB 0x0
+#define QIB_7322_IB_SDTEST_IF_RX_0_TS_T_RX_VALID_RMASK 0x1
+
+#define QIB_7322_IBNCModeCtrl_0_OFFS 0x15B8
+#define QIB_7322_IBNCModeCtrl_0_DEF 0x0000000000000000
+#define QIB_7322_IBNCModeCtrl_0_ScrambleCapRemoteForce_LSB 0x22
+#define QIB_7322_IBNCModeCtrl_0_ScrambleCapRemoteForce_MSB 0x22
+#define QIB_7322_IBNCModeCtrl_0_ScrambleCapRemoteForce_RMASK 0x1
+#define QIB_7322_IBNCModeCtrl_0_ScrambleCapRemoteMask_LSB 0x21
+#define QIB_7322_IBNCModeCtrl_0_ScrambleCapRemoteMask_MSB 0x21
+#define QIB_7322_IBNCModeCtrl_0_ScrambleCapRemoteMask_RMASK 0x1
+#define QIB_7322_IBNCModeCtrl_0_ScrambleCapLocal_LSB 0x20
+#define QIB_7322_IBNCModeCtrl_0_ScrambleCapLocal_MSB 0x20
+#define QIB_7322_IBNCModeCtrl_0_ScrambleCapLocal_RMASK 0x1
+#define QIB_7322_IBNCModeCtrl_0_TSMCode_TS2_LSB 0x11
+#define QIB_7322_IBNCModeCtrl_0_TSMCode_TS2_MSB 0x19
+#define QIB_7322_IBNCModeCtrl_0_TSMCode_TS2_RMASK 0x1FF
+#define QIB_7322_IBNCModeCtrl_0_TSMCode_TS1_LSB 0x8
+#define QIB_7322_IBNCModeCtrl_0_TSMCode_TS1_MSB 0x10
+#define QIB_7322_IBNCModeCtrl_0_TSMCode_TS1_RMASK 0x1FF
+#define QIB_7322_IBNCModeCtrl_0_TSMEnable_ignore_TSM_on_rx_LSB 0x2
+#define QIB_7322_IBNCModeCtrl_0_TSMEnable_ignore_TSM_on_rx_MSB 0x2
+#define QIB_7322_IBNCModeCtrl_0_TSMEnable_ignore_TSM_on_rx_RMASK 0x1
+#define QIB_7322_IBNCModeCtrl_0_TSMEnable_send_TS2_LSB 0x1
+#define QIB_7322_IBNCModeCtrl_0_TSMEnable_send_TS2_MSB 0x1
+#define QIB_7322_IBNCModeCtrl_0_TSMEnable_send_TS2_RMASK 0x1
+#define QIB_7322_IBNCModeCtrl_0_TSMEnable_send_TS1_LSB 0x0
+#define QIB_7322_IBNCModeCtrl_0_TSMEnable_send_TS1_MSB 0x0
+#define QIB_7322_IBNCModeCtrl_0_TSMEnable_send_TS1_RMASK 0x1
+
+#define QIB_7322_IBSerdesStatus_0_OFFS 0x15D0
+#define QIB_7322_IBSerdesStatus_0_DEF 0x0000000000000000
+
+#define QIB_7322_IBPCSConfig_0_OFFS 0x15D8
+#define QIB_7322_IBPCSConfig_0_DEF 0x0000000000000007
+#define QIB_7322_IBPCSConfig_0_link_sync_mask_LSB 0x9
+#define QIB_7322_IBPCSConfig_0_link_sync_mask_MSB 0x12
+#define QIB_7322_IBPCSConfig_0_link_sync_mask_RMASK 0x3FF
+#define QIB_7322_IBPCSConfig_0_xcv_rreset_LSB 0x2
+#define QIB_7322_IBPCSConfig_0_xcv_rreset_MSB 0x2
+#define QIB_7322_IBPCSConfig_0_xcv_rreset_RMASK 0x1
+#define QIB_7322_IBPCSConfig_0_xcv_treset_LSB 0x1
+#define QIB_7322_IBPCSConfig_0_xcv_treset_MSB 0x1
+#define QIB_7322_IBPCSConfig_0_xcv_treset_RMASK 0x1
+#define QIB_7322_IBPCSConfig_0_tx_rx_reset_LSB 0x0
+#define QIB_7322_IBPCSConfig_0_tx_rx_reset_MSB 0x0
+#define QIB_7322_IBPCSConfig_0_tx_rx_reset_RMASK 0x1
+
+#define QIB_7322_IBSerdesCtrl_0_OFFS 0x15E0
+#define QIB_7322_IBSerdesCtrl_0_DEF 0x0000000000FFA00F
+#define QIB_7322_IBSerdesCtrl_0_DISABLE_RXLATOFF_QDR_LSB 0x1A
+#define QIB_7322_IBSerdesCtrl_0_DISABLE_RXLATOFF_QDR_MSB 0x1A
+#define QIB_7322_IBSerdesCtrl_0_DISABLE_RXLATOFF_QDR_RMASK 0x1
+#define QIB_7322_IBSerdesCtrl_0_DISABLE_RXLATOFF_DDR_LSB 0x19
+#define QIB_7322_IBSerdesCtrl_0_DISABLE_RXLATOFF_DDR_MSB 0x19
+#define QIB_7322_IBSerdesCtrl_0_DISABLE_RXLATOFF_DDR_RMASK 0x1
+#define QIB_7322_IBSerdesCtrl_0_DISABLE_RXLATOFF_SDR_LSB 0x18
+#define QIB_7322_IBSerdesCtrl_0_DISABLE_RXLATOFF_SDR_MSB 0x18
+#define QIB_7322_IBSerdesCtrl_0_DISABLE_RXLATOFF_SDR_RMASK 0x1
+#define QIB_7322_IBSerdesCtrl_0_CHANNEL_RESET_N_LSB 0x14
+#define QIB_7322_IBSerdesCtrl_0_CHANNEL_RESET_N_MSB 0x17
+#define QIB_7322_IBSerdesCtrl_0_CHANNEL_RESET_N_RMASK 0xF
+#define QIB_7322_IBSerdesCtrl_0_CGMODE_LSB 0x10
+#define QIB_7322_IBSerdesCtrl_0_CGMODE_MSB 0x13
+#define QIB_7322_IBSerdesCtrl_0_CGMODE_RMASK 0xF
+#define QIB_7322_IBSerdesCtrl_0_IB_LAT_MODE_LSB 0xF
+#define QIB_7322_IBSerdesCtrl_0_IB_LAT_MODE_MSB 0xF
+#define QIB_7322_IBSerdesCtrl_0_IB_LAT_MODE_RMASK 0x1
+#define QIB_7322_IBSerdesCtrl_0_RXLOSEN_LSB 0xD
+#define QIB_7322_IBSerdesCtrl_0_RXLOSEN_MSB 0xD
+#define QIB_7322_IBSerdesCtrl_0_RXLOSEN_RMASK 0x1
+#define QIB_7322_IBSerdesCtrl_0_LPEN_LSB 0xC
+#define QIB_7322_IBSerdesCtrl_0_LPEN_MSB 0xC
+#define QIB_7322_IBSerdesCtrl_0_LPEN_RMASK 0x1
+#define QIB_7322_IBSerdesCtrl_0_PLLPD_LSB 0xB
+#define QIB_7322_IBSerdesCtrl_0_PLLPD_MSB 0xB
+#define QIB_7322_IBSerdesCtrl_0_PLLPD_RMASK 0x1
+#define QIB_7322_IBSerdesCtrl_0_TXPD_LSB 0xA
+#define QIB_7322_IBSerdesCtrl_0_TXPD_MSB 0xA
+#define QIB_7322_IBSerdesCtrl_0_TXPD_RMASK 0x1
+#define QIB_7322_IBSerdesCtrl_0_RXPD_LSB 0x9
+#define QIB_7322_IBSerdesCtrl_0_RXPD_MSB 0x9
+#define QIB_7322_IBSerdesCtrl_0_RXPD_RMASK 0x1
+#define QIB_7322_IBSerdesCtrl_0_TXIDLE_LSB 0x8
+#define QIB_7322_IBSerdesCtrl_0_TXIDLE_MSB 0x8
+#define QIB_7322_IBSerdesCtrl_0_TXIDLE_RMASK 0x1
+#define QIB_7322_IBSerdesCtrl_0_CMODE_LSB 0x0
+#define QIB_7322_IBSerdesCtrl_0_CMODE_MSB 0x6
+#define QIB_7322_IBSerdesCtrl_0_CMODE_RMASK 0x7F
+
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_OFFS 0x1600
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_DEF 0x0000000000000000
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_tx_override_deemphasis_select_LSB 0x1F
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_tx_override_deemphasis_select_MSB 0x1F
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_tx_override_deemphasis_select_RMASK 0x1
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_reset_tx_deemphasis_override_LSB 0x1E
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_reset_tx_deemphasis_override_MSB 0x1E
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_reset_tx_deemphasis_override_RMASK 0x1
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_txampcntl_d2a_LSB 0xE
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_txampcntl_d2a_MSB 0x11
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_txampcntl_d2a_RMASK 0xF
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_txc0_ena_LSB 0x9
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_txc0_ena_MSB 0xD
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_txc0_ena_RMASK 0x1F
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_txcp1_ena_LSB 0x5
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_txcp1_ena_MSB 0x8
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_txcp1_ena_RMASK 0xF
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_txcn1_xtra_emph0_LSB 0x3
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_txcn1_xtra_emph0_MSB 0x4
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_txcn1_xtra_emph0_RMASK 0x3
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_txcn1_ena_LSB 0x0
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_txcn1_ena_MSB 0x2
+#define QIB_7322_IBSD_TX_DEEMPHASIS_OVERRIDE_0_txcn1_ena_RMASK 0x7
+
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_OFFS 0x1640
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_DEF 0x0000000000000000
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenagain_sdr_ch3_LSB 0x27
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenagain_sdr_ch3_MSB 0x27
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenagain_sdr_ch3_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenagain_sdr_ch2_LSB 0x26
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenagain_sdr_ch2_MSB 0x26
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenagain_sdr_ch2_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenagain_sdr_ch1_LSB 0x25
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenagain_sdr_ch1_MSB 0x25
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenagain_sdr_ch1_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenagain_sdr_ch0_LSB 0x24
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenagain_sdr_ch0_MSB 0x24
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenagain_sdr_ch0_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenale_sdr_ch3_LSB 0x23
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenale_sdr_ch3_MSB 0x23
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenale_sdr_ch3_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenale_sdr_ch2_LSB 0x22
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenale_sdr_ch2_MSB 0x22
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenale_sdr_ch2_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenale_sdr_ch1_LSB 0x21
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenale_sdr_ch1_MSB 0x21
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenale_sdr_ch1_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenale_sdr_ch0_LSB 0x20
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenale_sdr_ch0_MSB 0x20
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenale_sdr_ch0_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenadfe_sdr_ch3_LSB 0x18
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenadfe_sdr_ch3_MSB 0x1F
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenadfe_sdr_ch3_RMASK 0xFF
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenadfe_sdr_ch2_LSB 0x10
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenadfe_sdr_ch2_MSB 0x17
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenadfe_sdr_ch2_RMASK 0xFF
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenadfe_sdr_ch1_LSB 0x8
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenadfe_sdr_ch1_MSB 0xF
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenadfe_sdr_ch1_RMASK 0xFF
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenadfe_sdr_ch0_LSB 0x0
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenadfe_sdr_ch0_MSB 0x7
+#define QIB_7322_ADAPT_DISABLE_STATIC_SDR_0_static_disable_rxenadfe_sdr_ch0_RMASK 0xFF
+
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_OFFS 0x1648
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_DEF 0x0000000000000000
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenagain_sdr_ch3_LSB 0x27
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenagain_sdr_ch3_MSB 0x27
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenagain_sdr_ch3_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenagain_sdr_ch2_LSB 0x26
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenagain_sdr_ch2_MSB 0x26
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenagain_sdr_ch2_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenagain_sdr_ch1_LSB 0x25
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenagain_sdr_ch1_MSB 0x25
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenagain_sdr_ch1_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenagain_sdr_ch0_LSB 0x24
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenagain_sdr_ch0_MSB 0x24
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenagain_sdr_ch0_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenale_sdr_ch3_LSB 0x23
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenale_sdr_ch3_MSB 0x23
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenale_sdr_ch3_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenale_sdr_ch2_LSB 0x22
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenale_sdr_ch2_MSB 0x22
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenale_sdr_ch2_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenale_sdr_ch1_LSB 0x21
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenale_sdr_ch1_MSB 0x21
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenale_sdr_ch1_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenale_sdr_ch0_LSB 0x20
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenale_sdr_ch0_MSB 0x20
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenale_sdr_ch0_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenadfe_sdr_ch3_LSB 0x18
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenadfe_sdr_ch3_MSB 0x1F
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenadfe_sdr_ch3_RMASK 0xFF
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenadfe_sdr_ch2_LSB 0x10
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenadfe_sdr_ch2_MSB 0x17
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenadfe_sdr_ch2_RMASK 0xFF
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenadfe_sdr_ch1_LSB 0x8
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenadfe_sdr_ch1_MSB 0xF
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenadfe_sdr_ch1_RMASK 0xFF
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenadfe_sdr_ch0_LSB 0x0
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenadfe_sdr_ch0_MSB 0x7
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_SDR_0_dyn_disable_rxenadfe_sdr_ch0_RMASK 0xFF
+
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_OFFS 0x1650
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_DEF 0x0000000000000000
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenagain_ddr_ch3_LSB 0x27
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenagain_ddr_ch3_MSB 0x27
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenagain_ddr_ch3_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenagain_ddr_ch2_LSB 0x26
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenagain_ddr_ch2_MSB 0x26
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenagain_ddr_ch2_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenagain_ddr_ch1_LSB 0x25
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenagain_ddr_ch1_MSB 0x25
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenagain_ddr_ch1_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenagain_ddr_ch0_LSB 0x24
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenagain_ddr_ch0_MSB 0x24
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenagain_ddr_ch0_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenale_ddr_ch3_LSB 0x23
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenale_ddr_ch3_MSB 0x23
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenale_ddr_ch3_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenale_ddr_ch2_LSB 0x22
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenale_ddr_ch2_MSB 0x22
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenale_ddr_ch2_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenale_ddr_ch1_LSB 0x21
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenale_ddr_ch1_MSB 0x21
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenale_ddr_ch1_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenale_ddr_ch0_LSB 0x20
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenale_ddr_ch0_MSB 0x20
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenale_ddr_ch0_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenadfe_ddr_ch3_LSB 0x18
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenadfe_ddr_ch3_MSB 0x1F
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenadfe_ddr_ch3_RMASK 0xFF
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenadfe_ddr_ch2_LSB 0x10
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenadfe_ddr_ch2_MSB 0x17
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenadfe_ddr_ch2_RMASK 0xFF
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenadfe_ddr_ch1_LSB 0x8
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenadfe_ddr_ch1_MSB 0xF
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenadfe_ddr_ch1_RMASK 0xFF
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenadfe_ddr_ch0_LSB 0x0
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenadfe_ddr_ch0_MSB 0x7
+#define QIB_7322_ADAPT_DISABLE_STATIC_DDR_0_static_disable_rxenadfe_ddr_ch0_RMASK 0xFF
+
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_OFFS 0x1658
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_DEF 0x0000000000000000
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenagain_ddr_ch3_LSB 0x27
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenagain_ddr_ch3_MSB 0x27
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenagain_ddr_ch3_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenagain_ddr_ch2_LSB 0x26
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenagain_ddr_ch2_MSB 0x26
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenagain_ddr_ch2_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenagain_ddr_ch1_LSB 0x25
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenagain_ddr_ch1_MSB 0x25
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenagain_ddr_ch1_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenagain_ddr_ch0_LSB 0x24
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenagain_ddr_ch0_MSB 0x24
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenagain_ddr_ch0_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenale_ddr_ch3_LSB 0x23
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenale_ddr_ch3_MSB 0x23
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenale_ddr_ch3_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenale_ddr_ch2_LSB 0x22
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenale_ddr_ch2_MSB 0x22
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenale_ddr_ch2_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenale_ddr_ch1_LSB 0x21
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenale_ddr_ch1_MSB 0x21
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenale_ddr_ch1_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenale_ddr_ch0_LSB 0x20
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenale_ddr_ch0_MSB 0x20
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenale_ddr_ch0_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenadfe_ddr_ch3_LSB 0x18
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenadfe_ddr_ch3_MSB 0x1F
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenadfe_ddr_ch3_RMASK 0xFF
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenadfe_ddr_ch2_LSB 0x10
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenadfe_ddr_ch2_MSB 0x17
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenadfe_ddr_ch2_RMASK 0xFF
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenadfe_ddr_ch1_LSB 0x8
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenadfe_ddr_ch1_MSB 0xF
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenadfe_ddr_ch1_RMASK 0xFF
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenadfe_ddr_ch0_LSB 0x0
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenadfe_ddr_ch0_MSB 0x7
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_DDR_0_dyn_disable_rxenadfe_ddr_ch0_RMASK 0xFF
+
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_OFFS 0x1660
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_DEF 0x0000000000000000
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenagain_qdr_ch3_LSB 0x27
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenagain_qdr_ch3_MSB 0x27
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenagain_qdr_ch3_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenagain_qdr_ch2_LSB 0x26
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenagain_qdr_ch2_MSB 0x26
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenagain_qdr_ch2_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenagain_qdr_ch1_LSB 0x25
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenagain_qdr_ch1_MSB 0x25
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenagain_qdr_ch1_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenagain_qdr_ch0_LSB 0x24
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenagain_qdr_ch0_MSB 0x24
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenagain_qdr_ch0_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenale_qdr_ch3_LSB 0x23
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenale_qdr_ch3_MSB 0x23
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenale_qdr_ch3_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenale_qdr_ch2_LSB 0x22
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenale_qdr_ch2_MSB 0x22
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenale_qdr_ch2_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenale_qdr_ch1_LSB 0x21
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenale_qdr_ch1_MSB 0x21
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenale_qdr_ch1_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenale_qdr_ch0_LSB 0x20
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenale_qdr_ch0_MSB 0x20
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenale_qdr_ch0_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenadfe_qdr_ch3_LSB 0x18
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenadfe_qdr_ch3_MSB 0x1F
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenadfe_qdr_ch3_RMASK 0xFF
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenadfe_qdr_ch2_LSB 0x10
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenadfe_qdr_ch2_MSB 0x17
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenadfe_qdr_ch2_RMASK 0xFF
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenadfe_qdr_ch1_LSB 0x8
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenadfe_qdr_ch1_MSB 0xF
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenadfe_qdr_ch1_RMASK 0xFF
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenadfe_qdr_ch0_LSB 0x0
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenadfe_qdr_ch0_MSB 0x7
+#define QIB_7322_ADAPT_DISABLE_STATIC_QDR_0_static_disable_rxenadfe_qdr_ch0_RMASK 0xFF
+
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_OFFS 0x1668
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_DEF 0x0000000000000000
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenagain_qdr_ch3_LSB 0x27
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenagain_qdr_ch3_MSB 0x27
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenagain_qdr_ch3_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenagain_qdr_ch2_LSB 0x26
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenagain_qdr_ch2_MSB 0x26
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenagain_qdr_ch2_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenagain_qdr_ch1_LSB 0x25
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenagain_qdr_ch1_MSB 0x25
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenagain_qdr_ch1_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenagain_qdr_ch0_LSB 0x24
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenagain_qdr_ch0_MSB 0x24
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenagain_qdr_ch0_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenale_qdr_ch3_LSB 0x23
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenale_qdr_ch3_MSB 0x23
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenale_qdr_ch3_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenale_qdr_ch2_LSB 0x22
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenale_qdr_ch2_MSB 0x22
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenale_qdr_ch2_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenale_qdr_ch1_LSB 0x21
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenale_qdr_ch1_MSB 0x21
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenale_qdr_ch1_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenale_qdr_ch0_LSB 0x20
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenale_qdr_ch0_MSB 0x20
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenale_qdr_ch0_RMASK 0x1
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenadfe_qdr_ch3_LSB 0x18
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenadfe_qdr_ch3_MSB 0x1F
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenadfe_qdr_ch3_RMASK 0xFF
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenadfe_qdr_ch2_LSB 0x10
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenadfe_qdr_ch2_MSB 0x17
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenadfe_qdr_ch2_RMASK 0xFF
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenadfe_qdr_ch1_LSB 0x8
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenadfe_qdr_ch1_MSB 0xF
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenadfe_qdr_ch1_RMASK 0xFF
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenadfe_qdr_ch0_LSB 0x0
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenadfe_qdr_ch0_MSB 0x7
+#define QIB_7322_ADAPT_DISABLE_DYNAMIC_QDR_0_dyn_disable_rxenadfe_qdr_ch0_RMASK 0xFF
+
+#define QIB_7322_ADAPT_DISABLE_TIMER_THRESHOLD_0_OFFS 0x1670
+#define QIB_7322_ADAPT_DISABLE_TIMER_THRESHOLD_0_DEF 0x0000000000000000
+
+#define QIB_7322_HighPriorityLimit_0_OFFS 0x1BC0
+#define QIB_7322_HighPriorityLimit_0_DEF 0x0000000000000000
+#define QIB_7322_HighPriorityLimit_0_Limit_LSB 0x0
+#define QIB_7322_HighPriorityLimit_0_Limit_MSB 0x7
+#define QIB_7322_HighPriorityLimit_0_Limit_RMASK 0xFF
+
+#define QIB_7322_LowPriority0_0_OFFS 0x1C00
+#define QIB_7322_LowPriority0_0_DEF 0x0000000000000000
+#define QIB_7322_LowPriority0_0_VirtualLane_LSB 0x10
+#define QIB_7322_LowPriority0_0_VirtualLane_MSB 0x12
+#define QIB_7322_LowPriority0_0_VirtualLane_RMASK 0x7
+#define QIB_7322_LowPriority0_0_Weight_LSB 0x0
+#define QIB_7322_LowPriority0_0_Weight_MSB 0x7
+#define QIB_7322_LowPriority0_0_Weight_RMASK 0xFF
+
+#define QIB_7322_HighPriority0_0_OFFS 0x1E00
+#define QIB_7322_HighPriority0_0_DEF 0x0000000000000000
+#define QIB_7322_HighPriority0_0_VirtualLane_LSB 0x10
+#define QIB_7322_HighPriority0_0_VirtualLane_MSB 0x12
+#define QIB_7322_HighPriority0_0_VirtualLane_RMASK 0x7
+#define QIB_7322_HighPriority0_0_Weight_LSB 0x0
+#define QIB_7322_HighPriority0_0_Weight_MSB 0x7
+#define QIB_7322_HighPriority0_0_Weight_RMASK 0xFF
+
+#define QIB_7322_CntrRegBase_1_OFFS 0x2028
+#define QIB_7322_CntrRegBase_1_DEF 0x0000000000013000
+
+#define QIB_7322_RcvQPMulticastContext_1_OFFS 0x2170
+
+#define QIB_7322_SendCtrl_1_OFFS 0x21C0
+
+#define QIB_7322_SendBufAvail0_OFFS 0x3000
+#define QIB_7322_SendBufAvail0_DEF 0x0000000000000000
+#define QIB_7322_SendBufAvail0_SendBuf_31_0_LSB 0x0
+#define QIB_7322_SendBufAvail0_SendBuf_31_0_MSB 0x3F
+#define QIB_7322_SendBufAvail0_SendBuf_31_0_RMASK 0x0
+
+#define QIB_7322_MsixTable_OFFS 0x8000
+#define QIB_7322_MsixTable_DEF 0x0000000000000000
+
+#define QIB_7322_MsixPba_OFFS 0x9000
+#define QIB_7322_MsixPba_DEF 0x0000000000000000
+
+#define QIB_7322_LAMemory_OFFS 0xA000
+#define QIB_7322_LAMemory_DEF 0x0000000000000000
+
+#define QIB_7322_LBIntCnt_OFFS 0x11000
+#define QIB_7322_LBIntCnt_DEF 0x0000000000000000
+
+#define QIB_7322_LBFlowStallCnt_OFFS 0x11008
+#define QIB_7322_LBFlowStallCnt_DEF 0x0000000000000000
+
+#define QIB_7322_RxTIDFullErrCnt_OFFS 0x110D0
+#define QIB_7322_RxTIDFullErrCnt_DEF 0x0000000000000000
+
+#define QIB_7322_RxTIDValidErrCnt_OFFS 0x110D8
+#define QIB_7322_RxTIDValidErrCnt_DEF 0x0000000000000000
+
+#define QIB_7322_RxP0HdrEgrOvflCnt_OFFS 0x110E8
+#define QIB_7322_RxP0HdrEgrOvflCnt_DEF 0x0000000000000000
+
+#define QIB_7322_PcieRetryBufDiagQwordCnt_OFFS 0x111A0
+#define QIB_7322_PcieRetryBufDiagQwordCnt_DEF 0x0000000000000000
+
+#define QIB_7322_RxTidFlowDropCnt_OFFS 0x111E0
+#define QIB_7322_RxTidFlowDropCnt_DEF 0x0000000000000000
+
+#define QIB_7322_LBIntCnt_0_OFFS 0x12000
+#define QIB_7322_LBIntCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_TxCreditUpToDateTimeOut_0_OFFS 0x12008
+#define QIB_7322_TxCreditUpToDateTimeOut_0_DEF 0x0000000000000000
+
+#define QIB_7322_TxSDmaDescCnt_0_OFFS 0x12010
+#define QIB_7322_TxSDmaDescCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_TxUnsupVLErrCnt_0_OFFS 0x12018
+#define QIB_7322_TxUnsupVLErrCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_TxDataPktCnt_0_OFFS 0x12020
+#define QIB_7322_TxDataPktCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_TxFlowPktCnt_0_OFFS 0x12028
+#define QIB_7322_TxFlowPktCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_TxDwordCnt_0_OFFS 0x12030
+#define QIB_7322_TxDwordCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_TxLenErrCnt_0_OFFS 0x12038
+#define QIB_7322_TxLenErrCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_TxMaxMinLenErrCnt_0_OFFS 0x12040
+#define QIB_7322_TxMaxMinLenErrCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_TxUnderrunCnt_0_OFFS 0x12048
+#define QIB_7322_TxUnderrunCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_TxFlowStallCnt_0_OFFS 0x12050
+#define QIB_7322_TxFlowStallCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_TxDroppedPktCnt_0_OFFS 0x12058
+#define QIB_7322_TxDroppedPktCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxDroppedPktCnt_0_OFFS 0x12060
+#define QIB_7322_RxDroppedPktCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxDataPktCnt_0_OFFS 0x12068
+#define QIB_7322_RxDataPktCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxFlowPktCnt_0_OFFS 0x12070
+#define QIB_7322_RxFlowPktCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxDwordCnt_0_OFFS 0x12078
+#define QIB_7322_RxDwordCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxLenErrCnt_0_OFFS 0x12080
+#define QIB_7322_RxLenErrCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxMaxMinLenErrCnt_0_OFFS 0x12088
+#define QIB_7322_RxMaxMinLenErrCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxICRCErrCnt_0_OFFS 0x12090
+#define QIB_7322_RxICRCErrCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxVCRCErrCnt_0_OFFS 0x12098
+#define QIB_7322_RxVCRCErrCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxFlowCtrlViolCnt_0_OFFS 0x120A0
+#define QIB_7322_RxFlowCtrlViolCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxVersionErrCnt_0_OFFS 0x120A8
+#define QIB_7322_RxVersionErrCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxLinkMalformCnt_0_OFFS 0x120B0
+#define QIB_7322_RxLinkMalformCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxEBPCnt_0_OFFS 0x120B8
+#define QIB_7322_RxEBPCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxLPCRCErrCnt_0_OFFS 0x120C0
+#define QIB_7322_RxLPCRCErrCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxBufOvflCnt_0_OFFS 0x120C8
+#define QIB_7322_RxBufOvflCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxLenTruncateCnt_0_OFFS 0x120D0
+#define QIB_7322_RxLenTruncateCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxPKeyMismatchCnt_0_OFFS 0x120E0
+#define QIB_7322_RxPKeyMismatchCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_IBLinkDownedCnt_0_OFFS 0x12180
+#define QIB_7322_IBLinkDownedCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_IBSymbolErrCnt_0_OFFS 0x12188
+#define QIB_7322_IBSymbolErrCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_IBStatusChangeCnt_0_OFFS 0x12190
+#define QIB_7322_IBStatusChangeCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_IBLinkErrRecoveryCnt_0_OFFS 0x12198
+#define QIB_7322_IBLinkErrRecoveryCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_ExcessBufferOvflCnt_0_OFFS 0x121A8
+#define QIB_7322_ExcessBufferOvflCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_LocalLinkIntegrityErrCnt_0_OFFS 0x121B0
+#define QIB_7322_LocalLinkIntegrityErrCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxVlErrCnt_0_OFFS 0x121B8
+#define QIB_7322_RxVlErrCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxDlidFltrCnt_0_OFFS 0x121C0
+#define QIB_7322_RxDlidFltrCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxVL15DroppedPktCnt_0_OFFS 0x121C8
+#define QIB_7322_RxVL15DroppedPktCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxOtherLocalPhyErrCnt_0_OFFS 0x121D0
+#define QIB_7322_RxOtherLocalPhyErrCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_RxQPInvalidContextCnt_0_OFFS 0x121D8
+#define QIB_7322_RxQPInvalidContextCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_TxHeadersErrCnt_0_OFFS 0x121F8
+#define QIB_7322_TxHeadersErrCnt_0_DEF 0x0000000000000000
+
+#define QIB_7322_PSRcvDataCount_0_OFFS 0x12218
+#define QIB_7322_PSRcvDataCount_0_DEF 0x0000000000000000
+
+#define QIB_7322_PSRcvPktsCount_0_OFFS 0x12220
+#define QIB_7322_PSRcvPktsCount_0_DEF 0x0000000000000000
+
+#define QIB_7322_PSXmitDataCount_0_OFFS 0x12228
+#define QIB_7322_PSXmitDataCount_0_DEF 0x0000000000000000
+
+#define QIB_7322_PSXmitPktsCount_0_OFFS 0x12230
+#define QIB_7322_PSXmitPktsCount_0_DEF 0x0000000000000000
+
+#define QIB_7322_PSXmitWaitCount_0_OFFS 0x12238
+#define QIB_7322_PSXmitWaitCount_0_DEF 0x0000000000000000
+
+#define QIB_7322_LBIntCnt_1_OFFS 0x13000
+#define QIB_7322_LBIntCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_TxCreditUpToDateTimeOut_1_OFFS 0x13008
+#define QIB_7322_TxCreditUpToDateTimeOut_1_DEF 0x0000000000000000
+
+#define QIB_7322_TxSDmaDescCnt_1_OFFS 0x13010
+#define QIB_7322_TxSDmaDescCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_TxUnsupVLErrCnt_1_OFFS 0x13018
+#define QIB_7322_TxUnsupVLErrCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_TxDataPktCnt_1_OFFS 0x13020
+#define QIB_7322_TxDataPktCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_TxFlowPktCnt_1_OFFS 0x13028
+#define QIB_7322_TxFlowPktCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_TxDwordCnt_1_OFFS 0x13030
+#define QIB_7322_TxDwordCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_TxLenErrCnt_1_OFFS 0x13038
+#define QIB_7322_TxLenErrCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_TxMaxMinLenErrCnt_1_OFFS 0x13040
+#define QIB_7322_TxMaxMinLenErrCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_TxUnderrunCnt_1_OFFS 0x13048
+#define QIB_7322_TxUnderrunCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_TxFlowStallCnt_1_OFFS 0x13050
+#define QIB_7322_TxFlowStallCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_TxDroppedPktCnt_1_OFFS 0x13058
+#define QIB_7322_TxDroppedPktCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxDroppedPktCnt_1_OFFS 0x13060
+#define QIB_7322_RxDroppedPktCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxDataPktCnt_1_OFFS 0x13068
+#define QIB_7322_RxDataPktCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxFlowPktCnt_1_OFFS 0x13070
+#define QIB_7322_RxFlowPktCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxDwordCnt_1_OFFS 0x13078
+#define QIB_7322_RxDwordCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxLenErrCnt_1_OFFS 0x13080
+#define QIB_7322_RxLenErrCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxMaxMinLenErrCnt_1_OFFS 0x13088
+#define QIB_7322_RxMaxMinLenErrCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxICRCErrCnt_1_OFFS 0x13090
+#define QIB_7322_RxICRCErrCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxVCRCErrCnt_1_OFFS 0x13098
+#define QIB_7322_RxVCRCErrCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxFlowCtrlViolCnt_1_OFFS 0x130A0
+#define QIB_7322_RxFlowCtrlViolCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxVersionErrCnt_1_OFFS 0x130A8
+#define QIB_7322_RxVersionErrCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxLinkMalformCnt_1_OFFS 0x130B0
+#define QIB_7322_RxLinkMalformCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxEBPCnt_1_OFFS 0x130B8
+#define QIB_7322_RxEBPCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxLPCRCErrCnt_1_OFFS 0x130C0
+#define QIB_7322_RxLPCRCErrCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxBufOvflCnt_1_OFFS 0x130C8
+#define QIB_7322_RxBufOvflCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxLenTruncateCnt_1_OFFS 0x130D0
+#define QIB_7322_RxLenTruncateCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxPKeyMismatchCnt_1_OFFS 0x130E0
+#define QIB_7322_RxPKeyMismatchCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_IBLinkDownedCnt_1_OFFS 0x13180
+#define QIB_7322_IBLinkDownedCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_IBSymbolErrCnt_1_OFFS 0x13188
+#define QIB_7322_IBSymbolErrCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_IBStatusChangeCnt_1_OFFS 0x13190
+#define QIB_7322_IBStatusChangeCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_IBLinkErrRecoveryCnt_1_OFFS 0x13198
+#define QIB_7322_IBLinkErrRecoveryCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_ExcessBufferOvflCnt_1_OFFS 0x131A8
+#define QIB_7322_ExcessBufferOvflCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_LocalLinkIntegrityErrCnt_1_OFFS 0x131B0
+#define QIB_7322_LocalLinkIntegrityErrCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxVlErrCnt_1_OFFS 0x131B8
+#define QIB_7322_RxVlErrCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxDlidFltrCnt_1_OFFS 0x131C0
+#define QIB_7322_RxDlidFltrCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxVL15DroppedPktCnt_1_OFFS 0x131C8
+#define QIB_7322_RxVL15DroppedPktCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxOtherLocalPhyErrCnt_1_OFFS 0x131D0
+#define QIB_7322_RxOtherLocalPhyErrCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_RxQPInvalidContextCnt_1_OFFS 0x131D8
+#define QIB_7322_RxQPInvalidContextCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_TxHeadersErrCnt_1_OFFS 0x131F8
+#define QIB_7322_TxHeadersErrCnt_1_DEF 0x0000000000000000
+
+#define QIB_7322_PSRcvDataCount_1_OFFS 0x13218
+#define QIB_7322_PSRcvDataCount_1_DEF 0x0000000000000000
+
+#define QIB_7322_PSRcvPktsCount_1_OFFS 0x13220
+#define QIB_7322_PSRcvPktsCount_1_DEF 0x0000000000000000
+
+#define QIB_7322_PSXmitDataCount_1_OFFS 0x13228
+#define QIB_7322_PSXmitDataCount_1_DEF 0x0000000000000000
+
+#define QIB_7322_PSXmitPktsCount_1_OFFS 0x13230
+#define QIB_7322_PSXmitPktsCount_1_DEF 0x0000000000000000
+
+#define QIB_7322_PSXmitWaitCount_1_OFFS 0x13238
+#define QIB_7322_PSXmitWaitCount_1_DEF 0x0000000000000000
+
+#define QIB_7322_RcvEgrArray_OFFS 0x14000
+#define QIB_7322_RcvEgrArray_DEF 0x0000000000000000
+#define QIB_7322_RcvEgrArray_RT_BufSize_LSB 0x25
+#define QIB_7322_RcvEgrArray_RT_BufSize_MSB 0x27
+#define QIB_7322_RcvEgrArray_RT_BufSize_RMASK 0x7
+#define QIB_7322_RcvEgrArray_RT_Addr_LSB 0x0
+#define QIB_7322_RcvEgrArray_RT_Addr_MSB 0x24
+#define QIB_7322_RcvEgrArray_RT_Addr_RMASK 0x1FFFFFFFFF
+
+#define QIB_7322_RcvTIDArray0_OFFS 0x50000
+#define QIB_7322_RcvTIDArray0_DEF 0x0000000000000000
+#define QIB_7322_RcvTIDArray0_RT_BufSize_LSB 0x25
+#define QIB_7322_RcvTIDArray0_RT_BufSize_MSB 0x27
+#define QIB_7322_RcvTIDArray0_RT_BufSize_RMASK 0x7
+#define QIB_7322_RcvTIDArray0_RT_Addr_LSB 0x0
+#define QIB_7322_RcvTIDArray0_RT_Addr_MSB 0x24
+#define QIB_7322_RcvTIDArray0_RT_Addr_RMASK 0x1FFFFFFFFF
+
+#define QIB_7322_IBSD_DDS_MAP_TABLE_0_OFFS 0xD0000
+#define QIB_7322_IBSD_DDS_MAP_TABLE_0_DEF 0x0000000000000000
+
+#define QIB_7322_RcvHdrTail0_OFFS 0x200000
+#define QIB_7322_RcvHdrTail0_DEF 0x0000000000000000
+
+#define QIB_7322_RcvHdrHead0_OFFS 0x200008
+#define QIB_7322_RcvHdrHead0_DEF 0x0000000000000000
+#define QIB_7322_RcvHdrHead0_counter_LSB 0x20
+#define QIB_7322_RcvHdrHead0_counter_MSB 0x2F
+#define QIB_7322_RcvHdrHead0_counter_RMASK 0xFFFF
+#define QIB_7322_RcvHdrHead0_RcvHeadPointer_LSB 0x0
+#define QIB_7322_RcvHdrHead0_RcvHeadPointer_MSB 0x1F
+#define QIB_7322_RcvHdrHead0_RcvHeadPointer_RMASK 0xFFFFFFFF
+
+#define QIB_7322_RcvEgrIndexTail0_OFFS 0x200010
+#define QIB_7322_RcvEgrIndexTail0_DEF 0x0000000000000000
+
+#define QIB_7322_RcvEgrIndexHead0_OFFS 0x200018
+#define QIB_7322_RcvEgrIndexHead0_DEF 0x0000000000000000
+
+#define QIB_7322_RcvTIDFlowTable0_OFFS 0x201000
+#define QIB_7322_RcvTIDFlowTable0_DEF 0x0000000000000000
+#define QIB_7322_RcvTIDFlowTable0_GenMismatch_LSB 0x1C
+#define QIB_7322_RcvTIDFlowTable0_GenMismatch_MSB 0x1C
+#define QIB_7322_RcvTIDFlowTable0_GenMismatch_RMASK 0x1
+#define QIB_7322_RcvTIDFlowTable0_SeqMismatch_LSB 0x1B
+#define QIB_7322_RcvTIDFlowTable0_SeqMismatch_MSB 0x1B
+#define QIB_7322_RcvTIDFlowTable0_SeqMismatch_RMASK 0x1
+#define QIB_7322_RcvTIDFlowTable0_KeepOnGenErr_LSB 0x16
+#define QIB_7322_RcvTIDFlowTable0_KeepOnGenErr_MSB 0x16
+#define QIB_7322_RcvTIDFlowTable0_KeepOnGenErr_RMASK 0x1
+#define QIB_7322_RcvTIDFlowTable0_KeepAfterSeqErr_LSB 0x15
+#define QIB_7322_RcvTIDFlowTable0_KeepAfterSeqErr_MSB 0x15
+#define QIB_7322_RcvTIDFlowTable0_KeepAfterSeqErr_RMASK 0x1
+#define QIB_7322_RcvTIDFlowTable0_HdrSuppEnabled_LSB 0x14
+#define QIB_7322_RcvTIDFlowTable0_HdrSuppEnabled_MSB 0x14
+#define QIB_7322_RcvTIDFlowTable0_HdrSuppEnabled_RMASK 0x1
+#define QIB_7322_RcvTIDFlowTable0_FlowValid_LSB 0x13
+#define QIB_7322_RcvTIDFlowTable0_FlowValid_MSB 0x13
+#define QIB_7322_RcvTIDFlowTable0_FlowValid_RMASK 0x1
+#define QIB_7322_RcvTIDFlowTable0_GenVal_LSB 0xB
+#define QIB_7322_RcvTIDFlowTable0_GenVal_MSB 0x12
+#define QIB_7322_RcvTIDFlowTable0_GenVal_RMASK 0xFF
+#define QIB_7322_RcvTIDFlowTable0_SeqNum_LSB 0x0
+#define QIB_7322_RcvTIDFlowTable0_SeqNum_MSB 0xA
+#define QIB_7322_RcvTIDFlowTable0_SeqNum_RMASK 0x7FF
diff --git a/drivers/infiniband/hw/qib/qib_common.h b/drivers/infiniband/hw/qib/qib_common.h
new file mode 100644
index 000000000000..b3955ed8f794
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_common.h
@@ -0,0 +1,758 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009, 2010 QLogic Corporation.
+ * All rights reserved.
+ * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 _QIB_COMMON_H
+#define _QIB_COMMON_H
+
+/*
+ * This file contains defines, structures, etc. that are used
+ * to communicate between kernel and user code.
+ */
+
+/* This is the IEEE-assigned OUI for QLogic Inc. QLogic_IB */
+#define QIB_SRC_OUI_1 0x00
+#define QIB_SRC_OUI_2 0x11
+#define QIB_SRC_OUI_3 0x75
+
+/* version of protocol header (known to chip also). In the long run,
+ * we should be able to generate and accept a range of version numbers;
+ * for now we only accept one, and it's compiled in.
+ */
+#define IPS_PROTO_VERSION 2
+
+/*
+ * These are compile time constants that you may want to enable or disable
+ * if you are trying to debug problems with code or performance.
+ * QIB_VERBOSE_TRACING define as 1 if you want additional tracing in
+ * fastpath code
+ * QIB_TRACE_REGWRITES define as 1 if you want register writes to be
+ * traced in faspath code
+ * _QIB_TRACING define as 0 if you want to remove all tracing in a
+ * compilation unit
+ */
+
+/*
+ * The value in the BTH QP field that QLogic_IB uses to differentiate
+ * an qlogic_ib protocol IB packet vs standard IB transport
+ * This it needs to be even (0x656b78), because the LSB is sometimes
+ * used for the MSB of context. The change may cause a problem
+ * interoperating with older software.
+ */
+#define QIB_KD_QP 0x656b78
+
+/*
+ * These are the status bits readable (in ascii form, 64bit value)
+ * from the "status" sysfs file.  For binary compatibility, values
+ * must remain as is; removed states can be reused for different
+ * purposes.
+ */
+#define QIB_STATUS_INITTED       0x1    /* basic initialization done */
+/* Chip has been found and initted */
+#define QIB_STATUS_CHIP_PRESENT 0x20
+/* IB link is at ACTIVE, usable for data traffic */
+#define QIB_STATUS_IB_READY     0x40
+/* link is configured, LID, MTU, etc. have been set */
+#define QIB_STATUS_IB_CONF      0x80
+/* A Fatal hardware error has occurred. */
+#define QIB_STATUS_HWERROR     0x200
+
+/*
+ * The list of usermode accessible registers.  Also see Reg_* later in file.
+ */
+enum qib_ureg {
+	/* (RO)  DMA RcvHdr to be used next. */
+	ur_rcvhdrtail = 0,
+	/* (RW)  RcvHdr entry to be processed next by host. */
+	ur_rcvhdrhead = 1,
+	/* (RO)  Index of next Eager index to use. */
+	ur_rcvegrindextail = 2,
+	/* (RW)  Eager TID to be processed next */
+	ur_rcvegrindexhead = 3,
+	/* For internal use only; max register number. */
+	_QIB_UregMax
+};
+
+/* bit values for spi_runtime_flags */
+#define QIB_RUNTIME_PCIE                0x0002
+#define QIB_RUNTIME_FORCE_WC_ORDER      0x0004
+#define QIB_RUNTIME_RCVHDR_COPY         0x0008
+#define QIB_RUNTIME_MASTER              0x0010
+#define QIB_RUNTIME_RCHK                0x0020
+#define QIB_RUNTIME_NODMA_RTAIL         0x0080
+#define QIB_RUNTIME_SPECIAL_TRIGGER     0x0100
+#define QIB_RUNTIME_SDMA                0x0200
+#define QIB_RUNTIME_FORCE_PIOAVAIL      0x0400
+#define QIB_RUNTIME_PIO_REGSWAPPED      0x0800
+#define QIB_RUNTIME_CTXT_MSB_IN_QP      0x1000
+#define QIB_RUNTIME_CTXT_REDIRECT       0x2000
+#define QIB_RUNTIME_HDRSUPP             0x4000
+
+/*
+ * This structure is returned by qib_userinit() immediately after
+ * open to get implementation-specific info, and info specific to this
+ * instance.
+ *
+ * This struct must have explict pad fields where type sizes
+ * may result in different alignments between 32 and 64 bit
+ * programs, since the 64 bit * bit kernel requires the user code
+ * to have matching offsets
+ */
+struct qib_base_info {
+	/* version of hardware, for feature checking. */
+	__u32 spi_hw_version;
+	/* version of software, for feature checking. */
+	__u32 spi_sw_version;
+	/* QLogic_IB context assigned, goes into sent packets */
+	__u16 spi_ctxt;
+	__u16 spi_subctxt;
+	/*
+	 * IB MTU, packets IB data must be less than this.
+	 * The MTU is in bytes, and will be a multiple of 4 bytes.
+	 */
+	__u32 spi_mtu;
+	/*
+	 * Size of a PIO buffer.  Any given packet's total size must be less
+	 * than this (in words).  Included is the starting control word, so
+	 * if 513 is returned, then total pkt size is 512 words or less.
+	 */
+	__u32 spi_piosize;
+	/* size of the TID cache in qlogic_ib, in entries */
+	__u32 spi_tidcnt;
+	/* size of the TID Eager list in qlogic_ib, in entries */
+	__u32 spi_tidegrcnt;
+	/* size of a single receive header queue entry in words. */
+	__u32 spi_rcvhdrent_size;
+	/*
+	 * Count of receive header queue entries allocated.
+	 * This may be less than the spu_rcvhdrcnt passed in!.
+	 */
+	__u32 spi_rcvhdr_cnt;
+
+	/* per-chip and other runtime features bitmap (QIB_RUNTIME_*) */
+	__u32 spi_runtime_flags;
+
+	/* address where hardware receive header queue is mapped */
+	__u64 spi_rcvhdr_base;
+
+	/* user program. */
+
+	/* base address of eager TID receive buffers used by hardware. */
+	__u64 spi_rcv_egrbufs;
+
+	/* Allocated by initialization code, not by protocol. */
+
+	/*
+	 * Size of each TID buffer in host memory, starting at
+	 * spi_rcv_egrbufs.  The buffers are virtually contiguous.
+	 */
+	__u32 spi_rcv_egrbufsize;
+	/*
+	 * The special QP (queue pair) value that identifies an qlogic_ib
+	 * protocol packet from standard IB packets.  More, probably much
+	 * more, to be added.
+	 */
+	__u32 spi_qpair;
+
+	/*
+	 * User register base for init code, not to be used directly by
+	 * protocol or applications.  Always points to chip registers,
+	 * for normal or shared context.
+	 */
+	__u64 spi_uregbase;
+	/*
+	 * Maximum buffer size in bytes that can be used in a single TID
+	 * entry (assuming the buffer is aligned to this boundary).  This is
+	 * the minimum of what the hardware and software support Guaranteed
+	 * to be a power of 2.
+	 */
+	__u32 spi_tid_maxsize;
+	/*
+	 * alignment of each pio send buffer (byte count
+	 * to add to spi_piobufbase to get to second buffer)
+	 */
+	__u32 spi_pioalign;
+	/*
+	 * The index of the first pio buffer available to this process;
+	 * needed to do lookup in spi_pioavailaddr; not added to
+	 * spi_piobufbase.
+	 */
+	__u32 spi_pioindex;
+	 /* number of buffers mapped for this process */
+	__u32 spi_piocnt;
+
+	/*
+	 * Base address of writeonly pio buffers for this process.
+	 * Each buffer has spi_piosize words, and is aligned on spi_pioalign
+	 * boundaries.  spi_piocnt buffers are mapped from this address
+	 */
+	__u64 spi_piobufbase;
+
+	/*
+	 * Base address of readonly memory copy of the pioavail registers.
+	 * There are 2 bits for each buffer.
+	 */
+	__u64 spi_pioavailaddr;
+
+	/*
+	 * Address where driver updates a copy of the interface and driver
+	 * status (QIB_STATUS_*) as a 64 bit value.  It's followed by a
+	 * link status qword (formerly combined with driver status), then a
+	 * string indicating hardware error, if there was one.
+	 */
+	__u64 spi_status;
+
+	/* number of chip ctxts available to user processes */
+	__u32 spi_nctxts;
+	__u16 spi_unit; /* unit number of chip we are using */
+	__u16 spi_port; /* IB port number we are using */
+	/* num bufs in each contiguous set */
+	__u32 spi_rcv_egrperchunk;
+	/* size in bytes of each contiguous set */
+	__u32 spi_rcv_egrchunksize;
+	/* total size of mmap to cover full rcvegrbuffers */
+	__u32 spi_rcv_egrbuftotlen;
+	__u32 spi_rhf_offset; /* dword offset in hdrqent for rcvhdr flags */
+	/* address of readonly memory copy of the rcvhdrq tail register. */
+	__u64 spi_rcvhdr_tailaddr;
+
+	/*
+	 * shared memory pages for subctxts if ctxt is shared; these cover
+	 * all the processes in the group sharing a single context.
+	 * all have enough space for the num_subcontexts value on this job.
+	 */
+	__u64 spi_subctxt_uregbase;
+	__u64 spi_subctxt_rcvegrbuf;
+	__u64 spi_subctxt_rcvhdr_base;
+
+	/* shared memory page for send buffer disarm status */
+	__u64 spi_sendbuf_status;
+} __attribute__ ((aligned(8)));
+
+/*
+ * This version number is given to the driver by the user code during
+ * initialization in the spu_userversion field of qib_user_info, so
+ * the driver can check for compatibility with user code.
+ *
+ * The major version changes when data structures
+ * change in an incompatible way.  The driver must be the same or higher
+ * for initialization to succeed.  In some cases, a higher version
+ * driver will not interoperate with older software, and initialization
+ * will return an error.
+ */
+#define QIB_USER_SWMAJOR 1
+
+/*
+ * Minor version differences are always compatible
+ * a within a major version, however if user software is larger
+ * than driver software, some new features and/or structure fields
+ * may not be implemented; the user code must deal with this if it
+ * cares, or it must abort after initialization reports the difference.
+ */
+#define QIB_USER_SWMINOR 10
+
+#define QIB_USER_SWVERSION ((QIB_USER_SWMAJOR << 16) | QIB_USER_SWMINOR)
+
+#ifndef QIB_KERN_TYPE
+#define QIB_KERN_TYPE 0
+#define QIB_IDSTR "QLogic kernel.org driver"
+#endif
+
+/*
+ * Similarly, this is the kernel version going back to the user.  It's
+ * slightly different, in that we want to tell if the driver was built as
+ * part of a QLogic release, or from the driver from openfabrics.org,
+ * kernel.org, or a standard distribution, for support reasons.
+ * The high bit is 0 for non-QLogic and 1 for QLogic-built/supplied.
+ *
+ * It's returned by the driver to the user code during initialization in the
+ * spi_sw_version field of qib_base_info, so the user code can in turn
+ * check for compatibility with the kernel.
+*/
+#define QIB_KERN_SWVERSION ((QIB_KERN_TYPE << 31) | QIB_USER_SWVERSION)
+
+/*
+ * This structure is passed to qib_userinit() to tell the driver where
+ * user code buffers are, sizes, etc.   The offsets and sizes of the
+ * fields must remain unchanged, for binary compatibility.  It can
+ * be extended, if userversion is changed so user code can tell, if needed
+ */
+struct qib_user_info {
+	/*
+	 * version of user software, to detect compatibility issues.
+	 * Should be set to QIB_USER_SWVERSION.
+	 */
+	__u32 spu_userversion;
+
+	__u32 _spu_unused2;
+
+	/* size of struct base_info to write to */
+	__u32 spu_base_info_size;
+
+	__u32 _spu_unused3;
+
+	/*
+	 * If two or more processes wish to share a context, each process
+	 * must set the spu_subctxt_cnt and spu_subctxt_id to the same
+	 * values.  The only restriction on the spu_subctxt_id is that
+	 * it be unique for a given node.
+	 */
+	__u16 spu_subctxt_cnt;
+	__u16 spu_subctxt_id;
+
+	__u32 spu_port; /* IB port requested by user if > 0 */
+
+	/*
+	 * address of struct base_info to write to
+	 */
+	__u64 spu_base_info;
+
+} __attribute__ ((aligned(8)));
+
+/* User commands. */
+
+/* 16 available, was: old set up userspace (for old user code) */
+#define QIB_CMD_CTXT_INFO       17      /* find out what resources we got */
+#define QIB_CMD_RECV_CTRL       18      /* control receipt of packets */
+#define QIB_CMD_TID_UPDATE      19      /* update expected TID entries */
+#define QIB_CMD_TID_FREE        20      /* free expected TID entries */
+#define QIB_CMD_SET_PART_KEY    21      /* add partition key */
+/* 22 available, was: return info on slave processes (for old user code) */
+#define QIB_CMD_ASSIGN_CTXT     23      /* allocate HCA and ctxt */
+#define QIB_CMD_USER_INIT       24      /* set up userspace */
+#define QIB_CMD_UNUSED_1        25
+#define QIB_CMD_UNUSED_2        26
+#define QIB_CMD_PIOAVAILUPD     27      /* force an update of PIOAvail reg */
+#define QIB_CMD_POLL_TYPE       28      /* set the kind of polling we want */
+#define QIB_CMD_ARMLAUNCH_CTRL  29      /* armlaunch detection control */
+/* 30 is unused */
+#define QIB_CMD_SDMA_INFLIGHT   31      /* sdma inflight counter request */
+#define QIB_CMD_SDMA_COMPLETE   32      /* sdma completion counter request */
+/* 33 available, was a testing feature  */
+#define QIB_CMD_DISARM_BUFS     34      /* disarm send buffers w/ errors */
+#define QIB_CMD_ACK_EVENT       35      /* ack & clear bits */
+#define QIB_CMD_CPUS_LIST       36      /* list of cpus allocated, for pinned
+					 * processes: qib_cpus_list */
+
+/*
+ * QIB_CMD_ACK_EVENT obsoletes QIB_CMD_DISARM_BUFS, but we keep it for
+ * compatibility with libraries from previous release.   The ACK_EVENT
+ * will take appropriate driver action (if any, just DISARM for now),
+ * then clear the bits passed in as part of the mask.  These bits are
+ * in the first 64bit word at spi_sendbuf_status, and are passed to
+ * the driver in the event_mask union as well.
+ */
+#define _QIB_EVENT_DISARM_BUFS_BIT	0
+#define _QIB_EVENT_LINKDOWN_BIT		1
+#define _QIB_EVENT_LID_CHANGE_BIT	2
+#define _QIB_EVENT_LMC_CHANGE_BIT	3
+#define _QIB_EVENT_SL2VL_CHANGE_BIT	4
+#define _QIB_MAX_EVENT_BIT _QIB_EVENT_SL2VL_CHANGE_BIT
+
+#define QIB_EVENT_DISARM_BUFS_BIT	(1UL << _QIB_EVENT_DISARM_BUFS_BIT)
+#define QIB_EVENT_LINKDOWN_BIT		(1UL << _QIB_EVENT_LINKDOWN_BIT)
+#define QIB_EVENT_LID_CHANGE_BIT	(1UL << _QIB_EVENT_LID_CHANGE_BIT)
+#define QIB_EVENT_LMC_CHANGE_BIT	(1UL << _QIB_EVENT_LMC_CHANGE_BIT)
+#define QIB_EVENT_SL2VL_CHANGE_BIT	(1UL << _QIB_EVENT_SL2VL_CHANGE_BIT)
+
+
+/*
+ * Poll types
+ */
+#define QIB_POLL_TYPE_ANYRCV     0x0
+#define QIB_POLL_TYPE_URGENT     0x1
+
+struct qib_ctxt_info {
+	__u16 num_active;       /* number of active units */
+	__u16 unit;             /* unit (chip) assigned to caller */
+	__u16 port;             /* IB port assigned to caller (1-based) */
+	__u16 ctxt;             /* ctxt on unit assigned to caller */
+	__u16 subctxt;          /* subctxt on unit assigned to caller */
+	__u16 num_ctxts;        /* number of ctxts available on unit */
+	__u16 num_subctxts;     /* number of subctxts opened on ctxt */
+	__u16 rec_cpu;          /* cpu # for affinity (ffff if none) */
+};
+
+struct qib_tid_info {
+	__u32 tidcnt;
+	/* make structure same size in 32 and 64 bit */
+	__u32 tid__unused;
+	/* virtual address of first page in transfer */
+	__u64 tidvaddr;
+	/* pointer (same size 32/64 bit) to __u16 tid array */
+	__u64 tidlist;
+
+	/*
+	 * pointer (same size 32/64 bit) to bitmap of TIDs used
+	 * for this call; checked for being large enough at open
+	 */
+	__u64 tidmap;
+};
+
+struct qib_cmd {
+	__u32 type;                     /* command type */
+	union {
+		struct qib_tid_info tid_info;
+		struct qib_user_info user_info;
+
+		/*
+		 * address in userspace where we should put the sdma
+		 * inflight counter
+		 */
+		__u64 sdma_inflight;
+		/*
+		 * address in userspace where we should put the sdma
+		 * completion counter
+		 */
+		__u64 sdma_complete;
+		/* address in userspace of struct qib_ctxt_info to
+		   write result to */
+		__u64 ctxt_info;
+		/* enable/disable receipt of packets */
+		__u32 recv_ctrl;
+		/* enable/disable armlaunch errors (non-zero to enable) */
+		__u32 armlaunch_ctrl;
+		/* partition key to set */
+		__u16 part_key;
+		/* user address of __u32 bitmask of active slaves */
+		__u64 slave_mask_addr;
+		/* type of polling we want */
+		__u16 poll_type;
+		/* back pressure enable bit for one particular context */
+		__u8 ctxt_bp;
+		/* qib_user_event_ack(), IPATH_EVENT_* bits */
+		__u64 event_mask;
+	} cmd;
+};
+
+struct qib_iovec {
+	/* Pointer to data, but same size 32 and 64 bit */
+	__u64 iov_base;
+
+	/*
+	 * Length of data; don't need 64 bits, but want
+	 * qib_sendpkt to remain same size as before 32 bit changes, so...
+	 */
+	__u64 iov_len;
+};
+
+/*
+ * Describes a single packet for send.  Each packet can have one or more
+ * buffers, but the total length (exclusive of IB headers) must be less
+ * than the MTU, and if using the PIO method, entire packet length,
+ * including IB headers, must be less than the qib_piosize value (words).
+ * Use of this necessitates including sys/uio.h
+ */
+struct __qib_sendpkt {
+	__u32 sps_flags;        /* flags for packet (TBD) */
+	__u32 sps_cnt;          /* number of entries to use in sps_iov */
+	/* array of iov's describing packet. TEMPORARY */
+	struct qib_iovec sps_iov[4];
+};
+
+/*
+ * Diagnostics can send a packet by "writing" the following
+ * structs to the diag data special file.
+ * This allows a custom
+ * pbc (+ static rate) qword, so that special modes and deliberate
+ * changes to CRCs can be used. The elements were also re-ordered
+ * for better alignment and to avoid padding issues.
+ */
+#define _DIAG_XPKT_VERS 3
+struct qib_diag_xpkt {
+	__u16 version;
+	__u16 unit;
+	__u16 port;
+	__u16 len;
+	__u64 data;
+	__u64 pbc_wd;
+};
+
+/*
+ * Data layout in I2C flash (for GUID, etc.)
+ * All fields are little-endian binary unless otherwise stated
+ */
+#define QIB_FLASH_VERSION 2
+struct qib_flash {
+	/* flash layout version (QIB_FLASH_VERSION) */
+	__u8 if_fversion;
+	/* checksum protecting if_length bytes */
+	__u8 if_csum;
+	/*
+	 * valid length (in use, protected by if_csum), including
+	 * if_fversion and if_csum themselves)
+	 */
+	__u8 if_length;
+	/* the GUID, in network order */
+	__u8 if_guid[8];
+	/* number of GUIDs to use, starting from if_guid */
+	__u8 if_numguid;
+	/* the (last 10 characters of) board serial number, in ASCII */
+	char if_serial[12];
+	/* board mfg date (YYYYMMDD ASCII) */
+	char if_mfgdate[8];
+	/* last board rework/test date (YYYYMMDD ASCII) */
+	char if_testdate[8];
+	/* logging of error counts, TBD */
+	__u8 if_errcntp[4];
+	/* powered on hours, updated at driver unload */
+	__u8 if_powerhour[2];
+	/* ASCII free-form comment field */
+	char if_comment[32];
+	/* Backwards compatible prefix for longer QLogic Serial Numbers */
+	char if_sprefix[4];
+	/* 82 bytes used, min flash size is 128 bytes */
+	__u8 if_future[46];
+};
+
+/*
+ * These are the counters implemented in the chip, and are listed in order.
+ * The InterCaps naming is taken straight from the chip spec.
+ */
+struct qlogic_ib_counters {
+	__u64 LBIntCnt;
+	__u64 LBFlowStallCnt;
+	__u64 TxSDmaDescCnt;    /* was Reserved1 */
+	__u64 TxUnsupVLErrCnt;
+	__u64 TxDataPktCnt;
+	__u64 TxFlowPktCnt;
+	__u64 TxDwordCnt;
+	__u64 TxLenErrCnt;
+	__u64 TxMaxMinLenErrCnt;
+	__u64 TxUnderrunCnt;
+	__u64 TxFlowStallCnt;
+	__u64 TxDroppedPktCnt;
+	__u64 RxDroppedPktCnt;
+	__u64 RxDataPktCnt;
+	__u64 RxFlowPktCnt;
+	__u64 RxDwordCnt;
+	__u64 RxLenErrCnt;
+	__u64 RxMaxMinLenErrCnt;
+	__u64 RxICRCErrCnt;
+	__u64 RxVCRCErrCnt;
+	__u64 RxFlowCtrlErrCnt;
+	__u64 RxBadFormatCnt;
+	__u64 RxLinkProblemCnt;
+	__u64 RxEBPCnt;
+	__u64 RxLPCRCErrCnt;
+	__u64 RxBufOvflCnt;
+	__u64 RxTIDFullErrCnt;
+	__u64 RxTIDValidErrCnt;
+	__u64 RxPKeyMismatchCnt;
+	__u64 RxP0HdrEgrOvflCnt;
+	__u64 RxP1HdrEgrOvflCnt;
+	__u64 RxP2HdrEgrOvflCnt;
+	__u64 RxP3HdrEgrOvflCnt;
+	__u64 RxP4HdrEgrOvflCnt;
+	__u64 RxP5HdrEgrOvflCnt;
+	__u64 RxP6HdrEgrOvflCnt;
+	__u64 RxP7HdrEgrOvflCnt;
+	__u64 RxP8HdrEgrOvflCnt;
+	__u64 RxP9HdrEgrOvflCnt;
+	__u64 RxP10HdrEgrOvflCnt;
+	__u64 RxP11HdrEgrOvflCnt;
+	__u64 RxP12HdrEgrOvflCnt;
+	__u64 RxP13HdrEgrOvflCnt;
+	__u64 RxP14HdrEgrOvflCnt;
+	__u64 RxP15HdrEgrOvflCnt;
+	__u64 RxP16HdrEgrOvflCnt;
+	__u64 IBStatusChangeCnt;
+	__u64 IBLinkErrRecoveryCnt;
+	__u64 IBLinkDownedCnt;
+	__u64 IBSymbolErrCnt;
+	__u64 RxVL15DroppedPktCnt;
+	__u64 RxOtherLocalPhyErrCnt;
+	__u64 PcieRetryBufDiagQwordCnt;
+	__u64 ExcessBufferOvflCnt;
+	__u64 LocalLinkIntegrityErrCnt;
+	__u64 RxVlErrCnt;
+	__u64 RxDlidFltrCnt;
+};
+
+/*
+ * The next set of defines are for packet headers, and chip register
+ * and memory bits that are visible to and/or used by user-mode software.
+ */
+
+/* RcvHdrFlags bits */
+#define QLOGIC_IB_RHF_LENGTH_MASK 0x7FF
+#define QLOGIC_IB_RHF_LENGTH_SHIFT 0
+#define QLOGIC_IB_RHF_RCVTYPE_MASK 0x7
+#define QLOGIC_IB_RHF_RCVTYPE_SHIFT 11
+#define QLOGIC_IB_RHF_EGRINDEX_MASK 0xFFF
+#define QLOGIC_IB_RHF_EGRINDEX_SHIFT 16
+#define QLOGIC_IB_RHF_SEQ_MASK 0xF
+#define QLOGIC_IB_RHF_SEQ_SHIFT 0
+#define QLOGIC_IB_RHF_HDRQ_OFFSET_MASK 0x7FF
+#define QLOGIC_IB_RHF_HDRQ_OFFSET_SHIFT 4
+#define QLOGIC_IB_RHF_H_ICRCERR   0x80000000
+#define QLOGIC_IB_RHF_H_VCRCERR   0x40000000
+#define QLOGIC_IB_RHF_H_PARITYERR 0x20000000
+#define QLOGIC_IB_RHF_H_LENERR    0x10000000
+#define QLOGIC_IB_RHF_H_MTUERR    0x08000000
+#define QLOGIC_IB_RHF_H_IHDRERR   0x04000000
+#define QLOGIC_IB_RHF_H_TIDERR    0x02000000
+#define QLOGIC_IB_RHF_H_MKERR     0x01000000
+#define QLOGIC_IB_RHF_H_IBERR     0x00800000
+#define QLOGIC_IB_RHF_H_ERR_MASK  0xFF800000
+#define QLOGIC_IB_RHF_L_USE_EGR   0x80000000
+#define QLOGIC_IB_RHF_L_SWA       0x00008000
+#define QLOGIC_IB_RHF_L_SWB       0x00004000
+
+/* qlogic_ib header fields */
+#define QLOGIC_IB_I_VERS_MASK 0xF
+#define QLOGIC_IB_I_VERS_SHIFT 28
+#define QLOGIC_IB_I_CTXT_MASK 0xF
+#define QLOGIC_IB_I_CTXT_SHIFT 24
+#define QLOGIC_IB_I_TID_MASK 0x7FF
+#define QLOGIC_IB_I_TID_SHIFT 13
+#define QLOGIC_IB_I_OFFSET_MASK 0x1FFF
+#define QLOGIC_IB_I_OFFSET_SHIFT 0
+
+/* K_PktFlags bits */
+#define QLOGIC_IB_KPF_INTR 0x1
+#define QLOGIC_IB_KPF_SUBCTXT_MASK 0x3
+#define QLOGIC_IB_KPF_SUBCTXT_SHIFT 1
+
+#define QLOGIC_IB_MAX_SUBCTXT   4
+
+/* SendPIO per-buffer control */
+#define QLOGIC_IB_SP_TEST    0x40
+#define QLOGIC_IB_SP_TESTEBP 0x20
+#define QLOGIC_IB_SP_TRIGGER_SHIFT  15
+
+/* SendPIOAvail bits */
+#define QLOGIC_IB_SENDPIOAVAIL_BUSY_SHIFT 1
+#define QLOGIC_IB_SENDPIOAVAIL_CHECK_SHIFT 0
+
+/* qlogic_ib header format */
+struct qib_header {
+	/*
+	 * Version - 4 bits, Context - 4 bits, TID - 10 bits and Offset -
+	 * 14 bits before ECO change ~28 Dec 03.  After that, Vers 4,
+	 * Context 4, TID 11, offset 13.
+	 */
+	__le32 ver_ctxt_tid_offset;
+	__le16 chksum;
+	__le16 pkt_flags;
+};
+
+/*
+ * qlogic_ib user message header format.
+ * This structure contains the first 4 fields common to all protocols
+ * that employ qlogic_ib.
+ */
+struct qib_message_header {
+	__be16 lrh[4];
+	__be32 bth[3];
+	/* fields below this point are in host byte order */
+	struct qib_header iph;
+	__u8 sub_opcode;
+};
+
+/* IB - LRH header consts */
+#define QIB_LRH_GRH 0x0003      /* 1. word of IB LRH - next header: GRH */
+#define QIB_LRH_BTH 0x0002      /* 1. word of IB LRH - next header: BTH */
+
+/* misc. */
+#define SIZE_OF_CRC 1
+
+#define QIB_DEFAULT_P_KEY 0xFFFF
+#define QIB_PERMISSIVE_LID 0xFFFF
+#define QIB_AETH_CREDIT_SHIFT 24
+#define QIB_AETH_CREDIT_MASK 0x1F
+#define QIB_AETH_CREDIT_INVAL 0x1F
+#define QIB_PSN_MASK 0xFFFFFF
+#define QIB_MSN_MASK 0xFFFFFF
+#define QIB_QPN_MASK 0xFFFFFF
+#define QIB_MULTICAST_LID_BASE 0xC000
+#define QIB_EAGER_TID_ID QLOGIC_IB_I_TID_MASK
+#define QIB_MULTICAST_QPN 0xFFFFFF
+
+/* Receive Header Queue: receive type (from qlogic_ib) */
+#define RCVHQ_RCV_TYPE_EXPECTED  0
+#define RCVHQ_RCV_TYPE_EAGER     1
+#define RCVHQ_RCV_TYPE_NON_KD    2
+#define RCVHQ_RCV_TYPE_ERROR     3
+
+#define QIB_HEADER_QUEUE_WORDS 9
+
+/* functions for extracting fields from rcvhdrq entries for the driver.
+ */
+static inline __u32 qib_hdrget_err_flags(const __le32 *rbuf)
+{
+	return __le32_to_cpu(rbuf[1]) & QLOGIC_IB_RHF_H_ERR_MASK;
+}
+
+static inline __u32 qib_hdrget_rcv_type(const __le32 *rbuf)
+{
+	return (__le32_to_cpu(rbuf[0]) >> QLOGIC_IB_RHF_RCVTYPE_SHIFT) &
+		QLOGIC_IB_RHF_RCVTYPE_MASK;
+}
+
+static inline __u32 qib_hdrget_length_in_bytes(const __le32 *rbuf)
+{
+	return ((__le32_to_cpu(rbuf[0]) >> QLOGIC_IB_RHF_LENGTH_SHIFT) &
+		QLOGIC_IB_RHF_LENGTH_MASK) << 2;
+}
+
+static inline __u32 qib_hdrget_index(const __le32 *rbuf)
+{
+	return (__le32_to_cpu(rbuf[0]) >> QLOGIC_IB_RHF_EGRINDEX_SHIFT) &
+		QLOGIC_IB_RHF_EGRINDEX_MASK;
+}
+
+static inline __u32 qib_hdrget_seq(const __le32 *rbuf)
+{
+	return (__le32_to_cpu(rbuf[1]) >> QLOGIC_IB_RHF_SEQ_SHIFT) &
+		QLOGIC_IB_RHF_SEQ_MASK;
+}
+
+static inline __u32 qib_hdrget_offset(const __le32 *rbuf)
+{
+	return (__le32_to_cpu(rbuf[1]) >> QLOGIC_IB_RHF_HDRQ_OFFSET_SHIFT) &
+		QLOGIC_IB_RHF_HDRQ_OFFSET_MASK;
+}
+
+static inline __u32 qib_hdrget_use_egr_buf(const __le32 *rbuf)
+{
+	return __le32_to_cpu(rbuf[0]) & QLOGIC_IB_RHF_L_USE_EGR;
+}
+
+static inline __u32 qib_hdrget_qib_ver(__le32 hdrword)
+{
+	return (__le32_to_cpu(hdrword) >> QLOGIC_IB_I_VERS_SHIFT) &
+		QLOGIC_IB_I_VERS_MASK;
+}
+
+#endif                          /* _QIB_COMMON_H */
diff --git a/drivers/infiniband/hw/qib/qib_cq.c b/drivers/infiniband/hw/qib/qib_cq.c
new file mode 100644
index 000000000000..a86cbf880f98
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_cq.c
@@ -0,0 +1,484 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2010 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/err.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+
+#include "qib_verbs.h"
+
+/**
+ * qib_cq_enter - add a new entry to the completion queue
+ * @cq: completion queue
+ * @entry: work completion entry to add
+ * @sig: true if @entry is a solicitated entry
+ *
+ * This may be called with qp->s_lock held.
+ */
+void qib_cq_enter(struct qib_cq *cq, struct ib_wc *entry, int solicited)
+{
+	struct qib_cq_wc *wc;
+	unsigned long flags;
+	u32 head;
+	u32 next;
+
+	spin_lock_irqsave(&cq->lock, flags);
+
+	/*
+	 * Note that the head pointer might be writable by user processes.
+	 * Take care to verify it is a sane value.
+	 */
+	wc = cq->queue;
+	head = wc->head;
+	if (head >= (unsigned) cq->ibcq.cqe) {
+		head = cq->ibcq.cqe;
+		next = 0;
+	} else
+		next = head + 1;
+	if (unlikely(next == wc->tail)) {
+		spin_unlock_irqrestore(&cq->lock, flags);
+		if (cq->ibcq.event_handler) {
+			struct ib_event ev;
+
+			ev.device = cq->ibcq.device;
+			ev.element.cq = &cq->ibcq;
+			ev.event = IB_EVENT_CQ_ERR;
+			cq->ibcq.event_handler(&ev, cq->ibcq.cq_context);
+		}
+		return;
+	}
+	if (cq->ip) {
+		wc->uqueue[head].wr_id = entry->wr_id;
+		wc->uqueue[head].status = entry->status;
+		wc->uqueue[head].opcode = entry->opcode;
+		wc->uqueue[head].vendor_err = entry->vendor_err;
+		wc->uqueue[head].byte_len = entry->byte_len;
+		wc->uqueue[head].ex.imm_data =
+			(__u32 __force)entry->ex.imm_data;
+		wc->uqueue[head].qp_num = entry->qp->qp_num;
+		wc->uqueue[head].src_qp = entry->src_qp;
+		wc->uqueue[head].wc_flags = entry->wc_flags;
+		wc->uqueue[head].pkey_index = entry->pkey_index;
+		wc->uqueue[head].slid = entry->slid;
+		wc->uqueue[head].sl = entry->sl;
+		wc->uqueue[head].dlid_path_bits = entry->dlid_path_bits;
+		wc->uqueue[head].port_num = entry->port_num;
+		/* Make sure entry is written before the head index. */
+		smp_wmb();
+	} else
+		wc->kqueue[head] = *entry;
+	wc->head = next;
+
+	if (cq->notify == IB_CQ_NEXT_COMP ||
+	    (cq->notify == IB_CQ_SOLICITED && solicited)) {
+		cq->notify = IB_CQ_NONE;
+		cq->triggered++;
+		/*
+		 * This will cause send_complete() to be called in
+		 * another thread.
+		 */
+		queue_work(qib_cq_wq, &cq->comptask);
+	}
+
+	spin_unlock_irqrestore(&cq->lock, flags);
+}
+
+/**
+ * qib_poll_cq - poll for work completion entries
+ * @ibcq: the completion queue to poll
+ * @num_entries: the maximum number of entries to return
+ * @entry: pointer to array where work completions are placed
+ *
+ * Returns the number of completion entries polled.
+ *
+ * This may be called from interrupt context.  Also called by ib_poll_cq()
+ * in the generic verbs code.
+ */
+int qib_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry)
+{
+	struct qib_cq *cq = to_icq(ibcq);
+	struct qib_cq_wc *wc;
+	unsigned long flags;
+	int npolled;
+	u32 tail;
+
+	/* The kernel can only poll a kernel completion queue */
+	if (cq->ip) {
+		npolled = -EINVAL;
+		goto bail;
+	}
+
+	spin_lock_irqsave(&cq->lock, flags);
+
+	wc = cq->queue;
+	tail = wc->tail;
+	if (tail > (u32) cq->ibcq.cqe)
+		tail = (u32) cq->ibcq.cqe;
+	for (npolled = 0; npolled < num_entries; ++npolled, ++entry) {
+		if (tail == wc->head)
+			break;
+		/* The kernel doesn't need a RMB since it has the lock. */
+		*entry = wc->kqueue[tail];
+		if (tail >= cq->ibcq.cqe)
+			tail = 0;
+		else
+			tail++;
+	}
+	wc->tail = tail;
+
+	spin_unlock_irqrestore(&cq->lock, flags);
+
+bail:
+	return npolled;
+}
+
+static void send_complete(struct work_struct *work)
+{
+	struct qib_cq *cq = container_of(work, struct qib_cq, comptask);
+
+	/*
+	 * The completion handler will most likely rearm the notification
+	 * and poll for all pending entries.  If a new completion entry
+	 * is added while we are in this routine, queue_work()
+	 * won't call us again until we return so we check triggered to
+	 * see if we need to call the handler again.
+	 */
+	for (;;) {
+		u8 triggered = cq->triggered;
+
+		/*
+		 * IPoIB connected mode assumes the callback is from a
+		 * soft IRQ. We simulate this by blocking "bottom halves".
+		 * See the implementation for ipoib_cm_handle_tx_wc(),
+		 * netif_tx_lock_bh() and netif_tx_lock().
+		 */
+		local_bh_disable();
+		cq->ibcq.comp_handler(&cq->ibcq, cq->ibcq.cq_context);
+		local_bh_enable();
+
+		if (cq->triggered == triggered)
+			return;
+	}
+}
+
+/**
+ * qib_create_cq - create a completion queue
+ * @ibdev: the device this completion queue is attached to
+ * @entries: the minimum size of the completion queue
+ * @context: unused by the QLogic_IB driver
+ * @udata: user data for libibverbs.so
+ *
+ * Returns a pointer to the completion queue or negative errno values
+ * for failure.
+ *
+ * Called by ib_create_cq() in the generic verbs code.
+ */
+struct ib_cq *qib_create_cq(struct ib_device *ibdev, int entries,
+			    int comp_vector, struct ib_ucontext *context,
+			    struct ib_udata *udata)
+{
+	struct qib_ibdev *dev = to_idev(ibdev);
+	struct qib_cq *cq;
+	struct qib_cq_wc *wc;
+	struct ib_cq *ret;
+	u32 sz;
+
+	if (entries < 1 || entries > ib_qib_max_cqes) {
+		ret = ERR_PTR(-EINVAL);
+		goto done;
+	}
+
+	/* Allocate the completion queue structure. */
+	cq = kmalloc(sizeof(*cq), GFP_KERNEL);
+	if (!cq) {
+		ret = ERR_PTR(-ENOMEM);
+		goto done;
+	}
+
+	/*
+	 * Allocate the completion queue entries and head/tail pointers.
+	 * This is allocated separately so that it can be resized and
+	 * also mapped into user space.
+	 * We need to use vmalloc() in order to support mmap and large
+	 * numbers of entries.
+	 */
+	sz = sizeof(*wc);
+	if (udata && udata->outlen >= sizeof(__u64))
+		sz += sizeof(struct ib_uverbs_wc) * (entries + 1);
+	else
+		sz += sizeof(struct ib_wc) * (entries + 1);
+	wc = vmalloc_user(sz);
+	if (!wc) {
+		ret = ERR_PTR(-ENOMEM);
+		goto bail_cq;
+	}
+
+	/*
+	 * Return the address of the WC as the offset to mmap.
+	 * See qib_mmap() for details.
+	 */
+	if (udata && udata->outlen >= sizeof(__u64)) {
+		int err;
+
+		cq->ip = qib_create_mmap_info(dev, sz, context, wc);
+		if (!cq->ip) {
+			ret = ERR_PTR(-ENOMEM);
+			goto bail_wc;
+		}
+
+		err = ib_copy_to_udata(udata, &cq->ip->offset,
+				       sizeof(cq->ip->offset));
+		if (err) {
+			ret = ERR_PTR(err);
+			goto bail_ip;
+		}
+	} else
+		cq->ip = NULL;
+
+	spin_lock(&dev->n_cqs_lock);
+	if (dev->n_cqs_allocated == ib_qib_max_cqs) {
+		spin_unlock(&dev->n_cqs_lock);
+		ret = ERR_PTR(-ENOMEM);
+		goto bail_ip;
+	}
+
+	dev->n_cqs_allocated++;
+	spin_unlock(&dev->n_cqs_lock);
+
+	if (cq->ip) {
+		spin_lock_irq(&dev->pending_lock);
+		list_add(&cq->ip->pending_mmaps, &dev->pending_mmaps);
+		spin_unlock_irq(&dev->pending_lock);
+	}
+
+	/*
+	 * ib_create_cq() will initialize cq->ibcq except for cq->ibcq.cqe.
+	 * The number of entries should be >= the number requested or return
+	 * an error.
+	 */
+	cq->ibcq.cqe = entries;
+	cq->notify = IB_CQ_NONE;
+	cq->triggered = 0;
+	spin_lock_init(&cq->lock);
+	INIT_WORK(&cq->comptask, send_complete);
+	wc->head = 0;
+	wc->tail = 0;
+	cq->queue = wc;
+
+	ret = &cq->ibcq;
+
+	goto done;
+
+bail_ip:
+	kfree(cq->ip);
+bail_wc:
+	vfree(wc);
+bail_cq:
+	kfree(cq);
+done:
+	return ret;
+}
+
+/**
+ * qib_destroy_cq - destroy a completion queue
+ * @ibcq: the completion queue to destroy.
+ *
+ * Returns 0 for success.
+ *
+ * Called by ib_destroy_cq() in the generic verbs code.
+ */
+int qib_destroy_cq(struct ib_cq *ibcq)
+{
+	struct qib_ibdev *dev = to_idev(ibcq->device);
+	struct qib_cq *cq = to_icq(ibcq);
+
+	flush_work(&cq->comptask);
+	spin_lock(&dev->n_cqs_lock);
+	dev->n_cqs_allocated--;
+	spin_unlock(&dev->n_cqs_lock);
+	if (cq->ip)
+		kref_put(&cq->ip->ref, qib_release_mmap_info);
+	else
+		vfree(cq->queue);
+	kfree(cq);
+
+	return 0;
+}
+
+/**
+ * qib_req_notify_cq - change the notification type for a completion queue
+ * @ibcq: the completion queue
+ * @notify_flags: the type of notification to request
+ *
+ * Returns 0 for success.
+ *
+ * This may be called from interrupt context.  Also called by
+ * ib_req_notify_cq() in the generic verbs code.
+ */
+int qib_req_notify_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags notify_flags)
+{
+	struct qib_cq *cq = to_icq(ibcq);
+	unsigned long flags;
+	int ret = 0;
+
+	spin_lock_irqsave(&cq->lock, flags);
+	/*
+	 * Don't change IB_CQ_NEXT_COMP to IB_CQ_SOLICITED but allow
+	 * any other transitions (see C11-31 and C11-32 in ch. 11.4.2.2).
+	 */
+	if (cq->notify != IB_CQ_NEXT_COMP)
+		cq->notify = notify_flags & IB_CQ_SOLICITED_MASK;
+
+	if ((notify_flags & IB_CQ_REPORT_MISSED_EVENTS) &&
+	    cq->queue->head != cq->queue->tail)
+		ret = 1;
+
+	spin_unlock_irqrestore(&cq->lock, flags);
+
+	return ret;
+}
+
+/**
+ * qib_resize_cq - change the size of the CQ
+ * @ibcq: the completion queue
+ *
+ * Returns 0 for success.
+ */
+int qib_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
+{
+	struct qib_cq *cq = to_icq(ibcq);
+	struct qib_cq_wc *old_wc;
+	struct qib_cq_wc *wc;
+	u32 head, tail, n;
+	int ret;
+	u32 sz;
+
+	if (cqe < 1 || cqe > ib_qib_max_cqes) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	/*
+	 * Need to use vmalloc() if we want to support large #s of entries.
+	 */
+	sz = sizeof(*wc);
+	if (udata && udata->outlen >= sizeof(__u64))
+		sz += sizeof(struct ib_uverbs_wc) * (cqe + 1);
+	else
+		sz += sizeof(struct ib_wc) * (cqe + 1);
+	wc = vmalloc_user(sz);
+	if (!wc) {
+		ret = -ENOMEM;
+		goto bail;
+	}
+
+	/* Check that we can write the offset to mmap. */
+	if (udata && udata->outlen >= sizeof(__u64)) {
+		__u64 offset = 0;
+
+		ret = ib_copy_to_udata(udata, &offset, sizeof(offset));
+		if (ret)
+			goto bail_free;
+	}
+
+	spin_lock_irq(&cq->lock);
+	/*
+	 * Make sure head and tail are sane since they
+	 * might be user writable.
+	 */
+	old_wc = cq->queue;
+	head = old_wc->head;
+	if (head > (u32) cq->ibcq.cqe)
+		head = (u32) cq->ibcq.cqe;
+	tail = old_wc->tail;
+	if (tail > (u32) cq->ibcq.cqe)
+		tail = (u32) cq->ibcq.cqe;
+	if (head < tail)
+		n = cq->ibcq.cqe + 1 + head - tail;
+	else
+		n = head - tail;
+	if (unlikely((u32)cqe < n)) {
+		ret = -EINVAL;
+		goto bail_unlock;
+	}
+	for (n = 0; tail != head; n++) {
+		if (cq->ip)
+			wc->uqueue[n] = old_wc->uqueue[tail];
+		else
+			wc->kqueue[n] = old_wc->kqueue[tail];
+		if (tail == (u32) cq->ibcq.cqe)
+			tail = 0;
+		else
+			tail++;
+	}
+	cq->ibcq.cqe = cqe;
+	wc->head = n;
+	wc->tail = 0;
+	cq->queue = wc;
+	spin_unlock_irq(&cq->lock);
+
+	vfree(old_wc);
+
+	if (cq->ip) {
+		struct qib_ibdev *dev = to_idev(ibcq->device);
+		struct qib_mmap_info *ip = cq->ip;
+
+		qib_update_mmap_info(dev, ip, sz, wc);
+
+		/*
+		 * Return the offset to mmap.
+		 * See qib_mmap() for details.
+		 */
+		if (udata && udata->outlen >= sizeof(__u64)) {
+			ret = ib_copy_to_udata(udata, &ip->offset,
+					       sizeof(ip->offset));
+			if (ret)
+				goto bail;
+		}
+
+		spin_lock_irq(&dev->pending_lock);
+		if (list_empty(&ip->pending_mmaps))
+			list_add(&ip->pending_mmaps, &dev->pending_mmaps);
+		spin_unlock_irq(&dev->pending_lock);
+	}
+
+	ret = 0;
+	goto bail;
+
+bail_unlock:
+	spin_unlock_irq(&cq->lock);
+bail_free:
+	vfree(wc);
+bail:
+	return ret;
+}
diff --git a/drivers/infiniband/hw/qib/qib_diag.c b/drivers/infiniband/hw/qib/qib_diag.c
new file mode 100644
index 000000000000..ca98dd523752
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_diag.c
@@ -0,0 +1,894 @@
+/*
+ * Copyright (c) 2010 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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.
+ */
+
+/*
+ * This file contains support for diagnostic functions.  It is accessed by
+ * opening the qib_diag device, normally minor number 129.  Diagnostic use
+ * of the QLogic_IB chip may render the chip or board unusable until the
+ * driver is unloaded, or in some cases, until the system is rebooted.
+ *
+ * Accesses to the chip through this interface are not similar to going
+ * through the /sys/bus/pci resource mmap interface.
+ */
+
+#include <linux/io.h>
+#include <linux/pci.h>
+#include <linux/poll.h>
+#include <linux/vmalloc.h>
+#include <linux/fs.h>
+#include <linux/uaccess.h>
+
+#include "qib.h"
+#include "qib_common.h"
+
+/*
+ * Each client that opens the diag device must read then write
+ * offset 0, to prevent lossage from random cat or od. diag_state
+ * sequences this "handshake".
+ */
+enum diag_state { UNUSED = 0, OPENED, INIT, READY };
+
+/* State for an individual client. PID so children cannot abuse handshake */
+static struct qib_diag_client {
+	struct qib_diag_client *next;
+	struct qib_devdata *dd;
+	pid_t pid;
+	enum diag_state state;
+} *client_pool;
+
+/*
+ * Get a client struct. Recycled if possible, else kmalloc.
+ * Must be called with qib_mutex held
+ */
+static struct qib_diag_client *get_client(struct qib_devdata *dd)
+{
+	struct qib_diag_client *dc;
+
+	dc = client_pool;
+	if (dc)
+		/* got from pool remove it and use */
+		client_pool = dc->next;
+	else
+		/* None in pool, alloc and init */
+		dc = kmalloc(sizeof *dc, GFP_KERNEL);
+
+	if (dc) {
+		dc->next = NULL;
+		dc->dd = dd;
+		dc->pid = current->pid;
+		dc->state = OPENED;
+	}
+	return dc;
+}
+
+/*
+ * Return to pool. Must be called with qib_mutex held
+ */
+static void return_client(struct qib_diag_client *dc)
+{
+	struct qib_devdata *dd = dc->dd;
+	struct qib_diag_client *tdc, *rdc;
+
+	rdc = NULL;
+	if (dc == dd->diag_client) {
+		dd->diag_client = dc->next;
+		rdc = dc;
+	} else {
+		tdc = dc->dd->diag_client;
+		while (tdc) {
+			if (dc == tdc->next) {
+				tdc->next = dc->next;
+				rdc = dc;
+				break;
+			}
+			tdc = tdc->next;
+		}
+	}
+	if (rdc) {
+		rdc->state = UNUSED;
+		rdc->dd = NULL;
+		rdc->pid = 0;
+		rdc->next = client_pool;
+		client_pool = rdc;
+	}
+}
+
+static int qib_diag_open(struct inode *in, struct file *fp);
+static int qib_diag_release(struct inode *in, struct file *fp);
+static ssize_t qib_diag_read(struct file *fp, char __user *data,
+			     size_t count, loff_t *off);
+static ssize_t qib_diag_write(struct file *fp, const char __user *data,
+			      size_t count, loff_t *off);
+
+static const struct file_operations diag_file_ops = {
+	.owner = THIS_MODULE,
+	.write = qib_diag_write,
+	.read = qib_diag_read,
+	.open = qib_diag_open,
+	.release = qib_diag_release
+};
+
+static atomic_t diagpkt_count = ATOMIC_INIT(0);
+static struct cdev *diagpkt_cdev;
+static struct device *diagpkt_device;
+
+static ssize_t qib_diagpkt_write(struct file *fp, const char __user *data,
+				 size_t count, loff_t *off);
+
+static const struct file_operations diagpkt_file_ops = {
+	.owner = THIS_MODULE,
+	.write = qib_diagpkt_write,
+};
+
+int qib_diag_add(struct qib_devdata *dd)
+{
+	char name[16];
+	int ret = 0;
+
+	if (atomic_inc_return(&diagpkt_count) == 1) {
+		ret = qib_cdev_init(QIB_DIAGPKT_MINOR, "ipath_diagpkt",
+				    &diagpkt_file_ops, &diagpkt_cdev,
+				    &diagpkt_device);
+		if (ret)
+			goto done;
+	}
+
+	snprintf(name, sizeof(name), "ipath_diag%d", dd->unit);
+	ret = qib_cdev_init(QIB_DIAG_MINOR_BASE + dd->unit, name,
+			    &diag_file_ops, &dd->diag_cdev,
+			    &dd->diag_device);
+done:
+	return ret;
+}
+
+static void qib_unregister_observers(struct qib_devdata *dd);
+
+void qib_diag_remove(struct qib_devdata *dd)
+{
+	struct qib_diag_client *dc;
+
+	if (atomic_dec_and_test(&diagpkt_count))
+		qib_cdev_cleanup(&diagpkt_cdev, &diagpkt_device);
+
+	qib_cdev_cleanup(&dd->diag_cdev, &dd->diag_device);
+
+	/*
+	 * Return all diag_clients of this device. There should be none,
+	 * as we are "guaranteed" that no clients are still open
+	 */
+	while (dd->diag_client)
+		return_client(dd->diag_client);
+
+	/* Now clean up all unused client structs */
+	while (client_pool) {
+		dc = client_pool;
+		client_pool = dc->next;
+		kfree(dc);
+	}
+	/* Clean up observer list */
+	qib_unregister_observers(dd);
+}
+
+/* qib_remap_ioaddr32 - remap an offset into chip address space to __iomem *
+ *
+ * @dd: the qlogic_ib device
+ * @offs: the offset in chip-space
+ * @cntp: Pointer to max (byte) count for transfer starting at offset
+ * This returns a u32 __iomem * so it can be used for both 64 and 32-bit
+ * mapping. It is needed because with the use of PAT for control of
+ * write-combining, the logically contiguous address-space of the chip
+ * may be split into virtually non-contiguous spaces, with different
+ * attributes, which are them mapped to contiguous physical space
+ * based from the first BAR.
+ *
+ * The code below makes the same assumptions as were made in
+ * init_chip_wc_pat() (qib_init.c), copied here:
+ * Assumes chip address space looks like:
+ *		- kregs + sregs + cregs + uregs (in any order)
+ *		- piobufs (2K and 4K bufs in either order)
+ *	or:
+ *		- kregs + sregs + cregs (in any order)
+ *		- piobufs (2K and 4K bufs in either order)
+ *		- uregs
+ *
+ * If cntp is non-NULL, returns how many bytes from offset can be accessed
+ * Returns 0 if the offset is not mapped.
+ */
+static u32 __iomem *qib_remap_ioaddr32(struct qib_devdata *dd, u32 offset,
+				       u32 *cntp)
+{
+	u32 kreglen;
+	u32 snd_bottom, snd_lim = 0;
+	u32 __iomem *krb32 = (u32 __iomem *)dd->kregbase;
+	u32 __iomem *map = NULL;
+	u32 cnt = 0;
+
+	/* First, simplest case, offset is within the first map. */
+	kreglen = (dd->kregend - dd->kregbase) * sizeof(u64);
+	if (offset < kreglen) {
+		map = krb32 + (offset / sizeof(u32));
+		cnt = kreglen - offset;
+		goto mapped;
+	}
+
+	/*
+	 * Next check for user regs, the next most common case,
+	 * and a cheap check because if they are not in the first map
+	 * they are last in chip.
+	 */
+	if (dd->userbase) {
+		/* If user regs mapped, they are after send, so set limit. */
+		u32 ulim = (dd->cfgctxts * dd->ureg_align) + dd->uregbase;
+		snd_lim = dd->uregbase;
+		krb32 = (u32 __iomem *)dd->userbase;
+		if (offset >= dd->uregbase && offset < ulim) {
+			map = krb32 + (offset - dd->uregbase) / sizeof(u32);
+			cnt = ulim - offset;
+			goto mapped;
+		}
+	}
+
+	/*
+	 * Lastly, check for offset within Send Buffers.
+	 * This is gnarly because struct devdata is deliberately vague
+	 * about things like 7322 VL15 buffers, and we are not in
+	 * chip-specific code here, so should not make many assumptions.
+	 * The one we _do_ make is that the only chip that has more sndbufs
+	 * than we admit is the 7322, and it has userregs above that, so
+	 * we know the snd_lim.
+	 */
+	/* Assume 2K buffers are first. */
+	snd_bottom = dd->pio2k_bufbase;
+	if (snd_lim == 0) {
+		u32 tot2k = dd->piobcnt2k * ALIGN(dd->piosize2k, dd->palign);
+		snd_lim = snd_bottom + tot2k;
+	}
+	/* If 4k buffers exist, account for them by bumping
+	 * appropriate limit.
+	 */
+	if (dd->piobcnt4k) {
+		u32 tot4k = dd->piobcnt4k * dd->align4k;
+		u32 offs4k = dd->piobufbase >> 32;
+		if (snd_bottom > offs4k)
+			snd_bottom = offs4k;
+		else {
+			/* 4k above 2k. Bump snd_lim, if needed*/
+			if (!dd->userbase)
+				snd_lim = offs4k + tot4k;
+		}
+	}
+	/*
+	 * Judgement call: can we ignore the space between SendBuffs and
+	 * UserRegs, where we would like to see vl15 buffs, but not more?
+	 */
+	if (offset >= snd_bottom && offset < snd_lim) {
+		offset -= snd_bottom;
+		map = (u32 __iomem *)dd->piobase + (offset / sizeof(u32));
+		cnt = snd_lim - offset;
+	}
+
+mapped:
+	if (cntp)
+		*cntp = cnt;
+	return map;
+}
+
+/*
+ * qib_read_umem64 - read a 64-bit quantity from the chip into user space
+ * @dd: the qlogic_ib device
+ * @uaddr: the location to store the data in user memory
+ * @regoffs: the offset from BAR0 (_NOT_ full pointer, anymore)
+ * @count: number of bytes to copy (multiple of 32 bits)
+ *
+ * This function also localizes all chip memory accesses.
+ * The copy should be written such that we read full cacheline packets
+ * from the chip.  This is usually used for a single qword
+ *
+ * NOTE:  This assumes the chip address is 64-bit aligned.
+ */
+static int qib_read_umem64(struct qib_devdata *dd, void __user *uaddr,
+			   u32 regoffs, size_t count)
+{
+	const u64 __iomem *reg_addr;
+	const u64 __iomem *reg_end;
+	u32 limit;
+	int ret;
+
+	reg_addr = (const u64 __iomem *)qib_remap_ioaddr32(dd, regoffs, &limit);
+	if (reg_addr == NULL || limit == 0 || !(dd->flags & QIB_PRESENT)) {
+		ret = -EINVAL;
+		goto bail;
+	}
+	if (count >= limit)
+		count = limit;
+	reg_end = reg_addr + (count / sizeof(u64));
+
+	/* not very efficient, but it works for now */
+	while (reg_addr < reg_end) {
+		u64 data = readq(reg_addr);
+
+		if (copy_to_user(uaddr, &data, sizeof(u64))) {
+			ret = -EFAULT;
+			goto bail;
+		}
+		reg_addr++;
+		uaddr += sizeof(u64);
+	}
+	ret = 0;
+bail:
+	return ret;
+}
+
+/*
+ * qib_write_umem64 - write a 64-bit quantity to the chip from user space
+ * @dd: the qlogic_ib device
+ * @regoffs: the offset from BAR0 (_NOT_ full pointer, anymore)
+ * @uaddr: the source of the data in user memory
+ * @count: the number of bytes to copy (multiple of 32 bits)
+ *
+ * This is usually used for a single qword
+ * NOTE:  This assumes the chip address is 64-bit aligned.
+ */
+
+static int qib_write_umem64(struct qib_devdata *dd, u32 regoffs,
+			    const void __user *uaddr, size_t count)
+{
+	u64 __iomem *reg_addr;
+	const u64 __iomem *reg_end;
+	u32 limit;
+	int ret;
+
+	reg_addr = (u64 __iomem *)qib_remap_ioaddr32(dd, regoffs, &limit);
+	if (reg_addr == NULL || limit == 0 || !(dd->flags & QIB_PRESENT)) {
+		ret = -EINVAL;
+		goto bail;
+	}
+	if (count >= limit)
+		count = limit;
+	reg_end = reg_addr + (count / sizeof(u64));
+
+	/* not very efficient, but it works for now */
+	while (reg_addr < reg_end) {
+		u64 data;
+		if (copy_from_user(&data, uaddr, sizeof(data))) {
+			ret = -EFAULT;
+			goto bail;
+		}
+		writeq(data, reg_addr);
+
+		reg_addr++;
+		uaddr += sizeof(u64);
+	}
+	ret = 0;
+bail:
+	return ret;
+}
+
+/*
+ * qib_read_umem32 - read a 32-bit quantity from the chip into user space
+ * @dd: the qlogic_ib device
+ * @uaddr: the location to store the data in user memory
+ * @regoffs: the offset from BAR0 (_NOT_ full pointer, anymore)
+ * @count: number of bytes to copy
+ *
+ * read 32 bit values, not 64 bit; for memories that only
+ * support 32 bit reads; usually a single dword.
+ */
+static int qib_read_umem32(struct qib_devdata *dd, void __user *uaddr,
+			   u32 regoffs, size_t count)
+{
+	const u32 __iomem *reg_addr;
+	const u32 __iomem *reg_end;
+	u32 limit;
+	int ret;
+
+	reg_addr = qib_remap_ioaddr32(dd, regoffs, &limit);
+	if (reg_addr == NULL || limit == 0 || !(dd->flags & QIB_PRESENT)) {
+		ret = -EINVAL;
+		goto bail;
+	}
+	if (count >= limit)
+		count = limit;
+	reg_end = reg_addr + (count / sizeof(u32));
+
+	/* not very efficient, but it works for now */
+	while (reg_addr < reg_end) {
+		u32 data = readl(reg_addr);
+
+		if (copy_to_user(uaddr, &data, sizeof(data))) {
+			ret = -EFAULT;
+			goto bail;
+		}
+
+		reg_addr++;
+		uaddr += sizeof(u32);
+
+	}
+	ret = 0;
+bail:
+	return ret;
+}
+
+/*
+ * qib_write_umem32 - write a 32-bit quantity to the chip from user space
+ * @dd: the qlogic_ib device
+ * @regoffs: the offset from BAR0 (_NOT_ full pointer, anymore)
+ * @uaddr: the source of the data in user memory
+ * @count: number of bytes to copy
+ *
+ * write 32 bit values, not 64 bit; for memories that only
+ * support 32 bit write; usually a single dword.
+ */
+
+static int qib_write_umem32(struct qib_devdata *dd, u32 regoffs,
+			    const void __user *uaddr, size_t count)
+{
+	u32 __iomem *reg_addr;
+	const u32 __iomem *reg_end;
+	u32 limit;
+	int ret;
+
+	reg_addr = qib_remap_ioaddr32(dd, regoffs, &limit);
+	if (reg_addr == NULL || limit == 0 || !(dd->flags & QIB_PRESENT)) {
+		ret = -EINVAL;
+		goto bail;
+	}
+	if (count >= limit)
+		count = limit;
+	reg_end = reg_addr + (count / sizeof(u32));
+
+	while (reg_addr < reg_end) {
+		u32 data;
+
+		if (copy_from_user(&data, uaddr, sizeof(data))) {
+			ret = -EFAULT;
+			goto bail;
+		}
+		writel(data, reg_addr);
+
+		reg_addr++;
+		uaddr += sizeof(u32);
+	}
+	ret = 0;
+bail:
+	return ret;
+}
+
+static int qib_diag_open(struct inode *in, struct file *fp)
+{
+	int unit = iminor(in) - QIB_DIAG_MINOR_BASE;
+	struct qib_devdata *dd;
+	struct qib_diag_client *dc;
+	int ret;
+
+	mutex_lock(&qib_mutex);
+
+	dd = qib_lookup(unit);
+
+	if (dd == NULL || !(dd->flags & QIB_PRESENT) ||
+	    !dd->kregbase) {
+		ret = -ENODEV;
+		goto bail;
+	}
+
+	dc = get_client(dd);
+	if (!dc) {
+		ret = -ENOMEM;
+		goto bail;
+	}
+	dc->next = dd->diag_client;
+	dd->diag_client = dc;
+	fp->private_data = dc;
+	ret = 0;
+bail:
+	mutex_unlock(&qib_mutex);
+
+	return ret;
+}
+
+/**
+ * qib_diagpkt_write - write an IB packet
+ * @fp: the diag data device file pointer
+ * @data: qib_diag_pkt structure saying where to get the packet
+ * @count: size of data to write
+ * @off: unused by this code
+ */
+static ssize_t qib_diagpkt_write(struct file *fp,
+				 const char __user *data,
+				 size_t count, loff_t *off)
+{
+	u32 __iomem *piobuf;
+	u32 plen, clen, pbufn;
+	struct qib_diag_xpkt dp;
+	u32 *tmpbuf = NULL;
+	struct qib_devdata *dd;
+	struct qib_pportdata *ppd;
+	ssize_t ret = 0;
+
+	if (count != sizeof(dp)) {
+		ret = -EINVAL;
+		goto bail;
+	}
+	if (copy_from_user(&dp, data, sizeof(dp))) {
+		ret = -EFAULT;
+		goto bail;
+	}
+
+	dd = qib_lookup(dp.unit);
+	if (!dd || !(dd->flags & QIB_PRESENT) || !dd->kregbase) {
+		ret = -ENODEV;
+		goto bail;
+	}
+	if (!(dd->flags & QIB_INITTED)) {
+		/* no hardware, freeze, etc. */
+		ret = -ENODEV;
+		goto bail;
+	}
+
+	if (dp.version != _DIAG_XPKT_VERS) {
+		qib_dev_err(dd, "Invalid version %u for diagpkt_write\n",
+			    dp.version);
+		ret = -EINVAL;
+		goto bail;
+	}
+	/* send count must be an exact number of dwords */
+	if (dp.len & 3) {
+		ret = -EINVAL;
+		goto bail;
+	}
+	if (!dp.port || dp.port > dd->num_pports) {
+		ret = -EINVAL;
+		goto bail;
+	}
+	ppd = &dd->pport[dp.port - 1];
+
+	/* need total length before first word written */
+	/* +1 word is for the qword padding */
+	plen = sizeof(u32) + dp.len;
+	clen = dp.len >> 2;
+
+	if ((plen + 4) > ppd->ibmaxlen) {
+		ret = -EINVAL;
+		goto bail;      /* before writing pbc */
+	}
+	tmpbuf = vmalloc(plen);
+	if (!tmpbuf) {
+		qib_devinfo(dd->pcidev, "Unable to allocate tmp buffer, "
+			 "failing\n");
+		ret = -ENOMEM;
+		goto bail;
+	}
+
+	if (copy_from_user(tmpbuf,
+			   (const void __user *) (unsigned long) dp.data,
+			   dp.len)) {
+		ret = -EFAULT;
+		goto bail;
+	}
+
+	plen >>= 2;             /* in dwords */
+
+	if (dp.pbc_wd == 0)
+		dp.pbc_wd = plen;
+
+	piobuf = dd->f_getsendbuf(ppd, dp.pbc_wd, &pbufn);
+	if (!piobuf) {
+		ret = -EBUSY;
+		goto bail;
+	}
+	/* disarm it just to be extra sure */
+	dd->f_sendctrl(dd->pport, QIB_SENDCTRL_DISARM_BUF(pbufn));
+
+	/* disable header check on pbufn for this packet */
+	dd->f_txchk_change(dd, pbufn, 1, TXCHK_CHG_TYPE_DIS1, NULL);
+
+	writeq(dp.pbc_wd, piobuf);
+	/*
+	 * Copy all but the trigger word, then flush, so it's written
+	 * to chip before trigger word, then write trigger word, then
+	 * flush again, so packet is sent.
+	 */
+	if (dd->flags & QIB_PIO_FLUSH_WC) {
+		qib_flush_wc();
+		qib_pio_copy(piobuf + 2, tmpbuf, clen - 1);
+		qib_flush_wc();
+		__raw_writel(tmpbuf[clen - 1], piobuf + clen + 1);
+	} else
+		qib_pio_copy(piobuf + 2, tmpbuf, clen);
+
+	if (dd->flags & QIB_USE_SPCL_TRIG) {
+		u32 spcl_off = (pbufn >= dd->piobcnt2k) ? 2047 : 1023;
+
+		qib_flush_wc();
+		__raw_writel(0xaebecede, piobuf + spcl_off);
+	}
+
+	/*
+	 * Ensure buffer is written to the chip, then re-enable
+	 * header checks (if supported by chip).  The txchk
+	 * code will ensure seen by chip before returning.
+	 */
+	qib_flush_wc();
+	qib_sendbuf_done(dd, pbufn);
+	dd->f_txchk_change(dd, pbufn, 1, TXCHK_CHG_TYPE_ENAB1, NULL);
+
+	ret = sizeof(dp);
+
+bail:
+	vfree(tmpbuf);
+	return ret;
+}
+
+static int qib_diag_release(struct inode *in, struct file *fp)
+{
+	mutex_lock(&qib_mutex);
+	return_client(fp->private_data);
+	fp->private_data = NULL;
+	mutex_unlock(&qib_mutex);
+	return 0;
+}
+
+/*
+ * Chip-specific code calls to register its interest in
+ * a specific range.
+ */
+struct diag_observer_list_elt {
+	struct diag_observer_list_elt *next;
+	const struct diag_observer *op;
+};
+
+int qib_register_observer(struct qib_devdata *dd,
+			  const struct diag_observer *op)
+{
+	struct diag_observer_list_elt *olp;
+	int ret = -EINVAL;
+
+	if (!dd || !op)
+		goto bail;
+	ret = -ENOMEM;
+	olp = vmalloc(sizeof *olp);
+	if (!olp) {
+		printk(KERN_ERR QIB_DRV_NAME ": vmalloc for observer failed\n");
+		goto bail;
+	}
+	if (olp) {
+		unsigned long flags;
+
+		spin_lock_irqsave(&dd->qib_diag_trans_lock, flags);
+		olp->op = op;
+		olp->next = dd->diag_observer_list;
+		dd->diag_observer_list = olp;
+		spin_unlock_irqrestore(&dd->qib_diag_trans_lock, flags);
+		ret = 0;
+	}
+bail:
+	return ret;
+}
+
+/* Remove all registered observers when device is closed */
+static void qib_unregister_observers(struct qib_devdata *dd)
+{
+	struct diag_observer_list_elt *olp;
+	unsigned long flags;
+
+	spin_lock_irqsave(&dd->qib_diag_trans_lock, flags);
+	olp = dd->diag_observer_list;
+	while (olp) {
+		/* Pop one observer, let go of lock */
+		dd->diag_observer_list = olp->next;
+		spin_unlock_irqrestore(&dd->qib_diag_trans_lock, flags);
+		vfree(olp);
+		/* try again. */
+		spin_lock_irqsave(&dd->qib_diag_trans_lock, flags);
+		olp = dd->diag_observer_list;
+	}
+	spin_unlock_irqrestore(&dd->qib_diag_trans_lock, flags);
+}
+
+/*
+ * Find the observer, if any, for the specified address. Initial implementation
+ * is simple stack of observers. This must be called with diag transaction
+ * lock held.
+ */
+static const struct diag_observer *diag_get_observer(struct qib_devdata *dd,
+						     u32 addr)
+{
+	struct diag_observer_list_elt *olp;
+	const struct diag_observer *op = NULL;
+
+	olp = dd->diag_observer_list;
+	while (olp) {
+		op = olp->op;
+		if (addr >= op->bottom && addr <= op->top)
+			break;
+		olp = olp->next;
+	}
+	if (!olp)
+		op = NULL;
+
+	return op;
+}
+
+static ssize_t qib_diag_read(struct file *fp, char __user *data,
+			     size_t count, loff_t *off)
+{
+	struct qib_diag_client *dc = fp->private_data;
+	struct qib_devdata *dd = dc->dd;
+	void __iomem *kreg_base;
+	ssize_t ret;
+
+	if (dc->pid != current->pid) {
+		ret = -EPERM;
+		goto bail;
+	}
+
+	kreg_base = dd->kregbase;
+
+	if (count == 0)
+		ret = 0;
+	else if ((count % 4) || (*off % 4))
+		/* address or length is not 32-bit aligned, hence invalid */
+		ret = -EINVAL;
+	else if (dc->state < READY && (*off || count != 8))
+		ret = -EINVAL;  /* prevent cat /dev/qib_diag* */
+	else {
+		unsigned long flags;
+		u64 data64 = 0;
+		int use_32;
+		const struct diag_observer *op;
+
+		use_32 = (count % 8) || (*off % 8);
+		ret = -1;
+		spin_lock_irqsave(&dd->qib_diag_trans_lock, flags);
+		/*
+		 * Check for observer on this address range.
+		 * we only support a single 32 or 64-bit read
+		 * via observer, currently.
+		 */
+		op = diag_get_observer(dd, *off);
+		if (op) {
+			u32 offset = *off;
+			ret = op->hook(dd, op, offset, &data64, 0, use_32);
+		}
+		/*
+		 * We need to release lock before any copy_to_user(),
+		 * whether implicit in qib_read_umem* or explicit below.
+		 */
+		spin_unlock_irqrestore(&dd->qib_diag_trans_lock, flags);
+		if (!op) {
+			if (use_32)
+				/*
+				 * Address or length is not 64-bit aligned;
+				 * do 32-bit rd
+				 */
+				ret = qib_read_umem32(dd, data, (u32) *off,
+						      count);
+			else
+				ret = qib_read_umem64(dd, data, (u32) *off,
+						      count);
+		} else if (ret == count) {
+			/* Below finishes case where observer existed */
+			ret = copy_to_user(data, &data64, use_32 ?
+					   sizeof(u32) : sizeof(u64));
+			if (ret)
+				ret = -EFAULT;
+		}
+	}
+
+	if (ret >= 0) {
+		*off += count;
+		ret = count;
+		if (dc->state == OPENED)
+			dc->state = INIT;
+	}
+bail:
+	return ret;
+}
+
+static ssize_t qib_diag_write(struct file *fp, const char __user *data,
+			      size_t count, loff_t *off)
+{
+	struct qib_diag_client *dc = fp->private_data;
+	struct qib_devdata *dd = dc->dd;
+	void __iomem *kreg_base;
+	ssize_t ret;
+
+	if (dc->pid != current->pid) {
+		ret = -EPERM;
+		goto bail;
+	}
+
+	kreg_base = dd->kregbase;
+
+	if (count == 0)
+		ret = 0;
+	else if ((count % 4) || (*off % 4))
+		/* address or length is not 32-bit aligned, hence invalid */
+		ret = -EINVAL;
+	else if (dc->state < READY &&
+		((*off || count != 8) || dc->state != INIT))
+		/* No writes except second-step of init seq */
+		ret = -EINVAL;  /* before any other write allowed */
+	else {
+		unsigned long flags;
+		const struct diag_observer *op = NULL;
+		int use_32 =  (count % 8) || (*off % 8);
+
+		/*
+		 * Check for observer on this address range.
+		 * We only support a single 32 or 64-bit write
+		 * via observer, currently. This helps, because
+		 * we would otherwise have to jump through hoops
+		 * to make "diag transaction" meaningful when we
+		 * cannot do a copy_from_user while holding the lock.
+		 */
+		if (count == 4 || count == 8) {
+			u64 data64;
+			u32 offset = *off;
+			ret = copy_from_user(&data64, data, count);
+			if (ret) {
+				ret = -EFAULT;
+				goto bail;
+			}
+			spin_lock_irqsave(&dd->qib_diag_trans_lock, flags);
+			op = diag_get_observer(dd, *off);
+			if (op)
+				ret = op->hook(dd, op, offset, &data64, ~0Ull,
+					       use_32);
+			spin_unlock_irqrestore(&dd->qib_diag_trans_lock, flags);
+		}
+
+		if (!op) {
+			if (use_32)
+				/*
+				 * Address or length is not 64-bit aligned;
+				 * do 32-bit write
+				 */
+				ret = qib_write_umem32(dd, (u32) *off, data,
+						       count);
+			else
+				ret = qib_write_umem64(dd, (u32) *off, data,
+						       count);
+		}
+	}
+
+	if (ret >= 0) {
+		*off += count;
+		ret = count;
+		if (dc->state == INIT)
+			dc->state = READY; /* all read/write OK now */
+	}
+bail:
+	return ret;
+}
diff --git a/drivers/infiniband/hw/qib/qib_dma.c b/drivers/infiniband/hw/qib/qib_dma.c
new file mode 100644
index 000000000000..2920bb39a65b
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_dma.c
@@ -0,0 +1,182 @@
+/*
+ * Copyright (c) 2006, 2009, 2010 QLogic, Corporation. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/types.h>
+#include <linux/scatterlist.h>
+
+#include "qib_verbs.h"
+
+#define BAD_DMA_ADDRESS ((u64) 0)
+
+/*
+ * The following functions implement driver specific replacements
+ * for the ib_dma_*() functions.
+ *
+ * These functions return kernel virtual addresses instead of
+ * device bus addresses since the driver uses the CPU to copy
+ * data instead of using hardware DMA.
+ */
+
+static int qib_mapping_error(struct ib_device *dev, u64 dma_addr)
+{
+	return dma_addr == BAD_DMA_ADDRESS;
+}
+
+static u64 qib_dma_map_single(struct ib_device *dev, void *cpu_addr,
+			      size_t size, enum dma_data_direction direction)
+{
+	BUG_ON(!valid_dma_direction(direction));
+	return (u64) cpu_addr;
+}
+
+static void qib_dma_unmap_single(struct ib_device *dev, u64 addr, size_t size,
+				 enum dma_data_direction direction)
+{
+	BUG_ON(!valid_dma_direction(direction));
+}
+
+static u64 qib_dma_map_page(struct ib_device *dev, struct page *page,
+			    unsigned long offset, size_t size,
+			    enum dma_data_direction direction)
+{
+	u64 addr;
+
+	BUG_ON(!valid_dma_direction(direction));
+
+	if (offset + size > PAGE_SIZE) {
+		addr = BAD_DMA_ADDRESS;
+		goto done;
+	}
+
+	addr = (u64) page_address(page);
+	if (addr)
+		addr += offset;
+	/* TODO: handle highmem pages */
+
+done:
+	return addr;
+}
+
+static void qib_dma_unmap_page(struct ib_device *dev, u64 addr, size_t size,
+			       enum dma_data_direction direction)
+{
+	BUG_ON(!valid_dma_direction(direction));
+}
+
+static int qib_map_sg(struct ib_device *dev, struct scatterlist *sgl,
+		      int nents, enum dma_data_direction direction)
+{
+	struct scatterlist *sg;
+	u64 addr;
+	int i;
+	int ret = nents;
+
+	BUG_ON(!valid_dma_direction(direction));
+
+	for_each_sg(sgl, sg, nents, i) {
+		addr = (u64) page_address(sg_page(sg));
+		/* TODO: handle highmem pages */
+		if (!addr) {
+			ret = 0;
+			break;
+		}
+	}
+	return ret;
+}
+
+static void qib_unmap_sg(struct ib_device *dev,
+			 struct scatterlist *sg, int nents,
+			 enum dma_data_direction direction)
+{
+	BUG_ON(!valid_dma_direction(direction));
+}
+
+static u64 qib_sg_dma_address(struct ib_device *dev, struct scatterlist *sg)
+{
+	u64 addr = (u64) page_address(sg_page(sg));
+
+	if (addr)
+		addr += sg->offset;
+	return addr;
+}
+
+static unsigned int qib_sg_dma_len(struct ib_device *dev,
+				   struct scatterlist *sg)
+{
+	return sg->length;
+}
+
+static void qib_sync_single_for_cpu(struct ib_device *dev, u64 addr,
+				    size_t size, enum dma_data_direction dir)
+{
+}
+
+static void qib_sync_single_for_device(struct ib_device *dev, u64 addr,
+				       size_t size,
+				       enum dma_data_direction dir)
+{
+}
+
+static void *qib_dma_alloc_coherent(struct ib_device *dev, size_t size,
+				    u64 *dma_handle, gfp_t flag)
+{
+	struct page *p;
+	void *addr = NULL;
+
+	p = alloc_pages(flag, get_order(size));
+	if (p)
+		addr = page_address(p);
+	if (dma_handle)
+		*dma_handle = (u64) addr;
+	return addr;
+}
+
+static void qib_dma_free_coherent(struct ib_device *dev, size_t size,
+				  void *cpu_addr, u64 dma_handle)
+{
+	free_pages((unsigned long) cpu_addr, get_order(size));
+}
+
+struct ib_dma_mapping_ops qib_dma_mapping_ops = {
+	.mapping_error = qib_mapping_error,
+	.map_single = qib_dma_map_single,
+	.unmap_single = qib_dma_unmap_single,
+	.map_page = qib_dma_map_page,
+	.unmap_page = qib_dma_unmap_page,
+	.map_sg = qib_map_sg,
+	.unmap_sg = qib_unmap_sg,
+	.dma_address = qib_sg_dma_address,
+	.dma_len = qib_sg_dma_len,
+	.sync_single_for_cpu = qib_sync_single_for_cpu,
+	.sync_single_for_device = qib_sync_single_for_device,
+	.alloc_coherent = qib_dma_alloc_coherent,
+	.free_coherent = qib_dma_free_coherent
+};
diff --git a/drivers/infiniband/hw/qib/qib_driver.c b/drivers/infiniband/hw/qib/qib_driver.c
new file mode 100644
index 000000000000..f15ce076ac49
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_driver.c
@@ -0,0 +1,665 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/spinlock.h>
+#include <linux/pci.h>
+#include <linux/io.h>
+#include <linux/delay.h>
+#include <linux/netdevice.h>
+#include <linux/vmalloc.h>
+
+#include "qib.h"
+
+/*
+ * The size has to be longer than this string, so we can append
+ * board/chip information to it in the init code.
+ */
+const char ib_qib_version[] = QIB_IDSTR "\n";
+
+DEFINE_SPINLOCK(qib_devs_lock);
+LIST_HEAD(qib_dev_list);
+DEFINE_MUTEX(qib_mutex);	/* general driver use */
+
+unsigned qib_ibmtu;
+module_param_named(ibmtu, qib_ibmtu, uint, S_IRUGO);
+MODULE_PARM_DESC(ibmtu, "Set max IB MTU (0=2KB, 1=256, 2=512, ... 5=4096");
+
+unsigned qib_compat_ddr_negotiate = 1;
+module_param_named(compat_ddr_negotiate, qib_compat_ddr_negotiate, uint,
+		   S_IWUSR | S_IRUGO);
+MODULE_PARM_DESC(compat_ddr_negotiate,
+		 "Attempt pre-IBTA 1.2 DDR speed negotiation");
+
+MODULE_LICENSE("Dual BSD/GPL");
+MODULE_AUTHOR("QLogic <support@qlogic.com>");
+MODULE_DESCRIPTION("QLogic IB driver");
+
+/*
+ * QIB_PIO_MAXIBHDR is the max IB header size allowed for in our
+ * PIO send buffers.  This is well beyond anything currently
+ * defined in the InfiniBand spec.
+ */
+#define QIB_PIO_MAXIBHDR 128
+
+struct qlogic_ib_stats qib_stats;
+
+const char *qib_get_unit_name(int unit)
+{
+	static char iname[16];
+
+	snprintf(iname, sizeof iname, "infinipath%u", unit);
+	return iname;
+}
+
+/*
+ * Return count of units with at least one port ACTIVE.
+ */
+int qib_count_active_units(void)
+{
+	struct qib_devdata *dd;
+	struct qib_pportdata *ppd;
+	unsigned long flags;
+	int pidx, nunits_active = 0;
+
+	spin_lock_irqsave(&qib_devs_lock, flags);
+	list_for_each_entry(dd, &qib_dev_list, list) {
+		if (!(dd->flags & QIB_PRESENT) || !dd->kregbase)
+			continue;
+		for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+			ppd = dd->pport + pidx;
+			if (ppd->lid && (ppd->lflags & (QIBL_LINKINIT |
+					 QIBL_LINKARMED | QIBL_LINKACTIVE))) {
+				nunits_active++;
+				break;
+			}
+		}
+	}
+	spin_unlock_irqrestore(&qib_devs_lock, flags);
+	return nunits_active;
+}
+
+/*
+ * Return count of all units, optionally return in arguments
+ * the number of usable (present) units, and the number of
+ * ports that are up.
+ */
+int qib_count_units(int *npresentp, int *nupp)
+{
+	int nunits = 0, npresent = 0, nup = 0;
+	struct qib_devdata *dd;
+	unsigned long flags;
+	int pidx;
+	struct qib_pportdata *ppd;
+
+	spin_lock_irqsave(&qib_devs_lock, flags);
+
+	list_for_each_entry(dd, &qib_dev_list, list) {
+		nunits++;
+		if ((dd->flags & QIB_PRESENT) && dd->kregbase)
+			npresent++;
+		for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+			ppd = dd->pport + pidx;
+			if (ppd->lid && (ppd->lflags & (QIBL_LINKINIT |
+					 QIBL_LINKARMED | QIBL_LINKACTIVE)))
+				nup++;
+		}
+	}
+
+	spin_unlock_irqrestore(&qib_devs_lock, flags);
+
+	if (npresentp)
+		*npresentp = npresent;
+	if (nupp)
+		*nupp = nup;
+
+	return nunits;
+}
+
+/**
+ * qib_wait_linkstate - wait for an IB link state change to occur
+ * @dd: the qlogic_ib device
+ * @state: the state to wait for
+ * @msecs: the number of milliseconds to wait
+ *
+ * wait up to msecs milliseconds for IB link state change to occur for
+ * now, take the easy polling route.  Currently used only by
+ * qib_set_linkstate.  Returns 0 if state reached, otherwise
+ * -ETIMEDOUT state can have multiple states set, for any of several
+ * transitions.
+ */
+int qib_wait_linkstate(struct qib_pportdata *ppd, u32 state, int msecs)
+{
+	int ret;
+	unsigned long flags;
+
+	spin_lock_irqsave(&ppd->lflags_lock, flags);
+	if (ppd->state_wanted) {
+		spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+		ret = -EBUSY;
+		goto bail;
+	}
+	ppd->state_wanted = state;
+	spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+	wait_event_interruptible_timeout(ppd->state_wait,
+					 (ppd->lflags & state),
+					 msecs_to_jiffies(msecs));
+	spin_lock_irqsave(&ppd->lflags_lock, flags);
+	ppd->state_wanted = 0;
+	spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+
+	if (!(ppd->lflags & state))
+		ret = -ETIMEDOUT;
+	else
+		ret = 0;
+bail:
+	return ret;
+}
+
+int qib_set_linkstate(struct qib_pportdata *ppd, u8 newstate)
+{
+	u32 lstate;
+	int ret;
+	struct qib_devdata *dd = ppd->dd;
+	unsigned long flags;
+
+	switch (newstate) {
+	case QIB_IB_LINKDOWN_ONLY:
+		dd->f_set_ib_cfg(ppd, QIB_IB_CFG_LSTATE,
+				 IB_LINKCMD_DOWN | IB_LINKINITCMD_NOP);
+		/* don't wait */
+		ret = 0;
+		goto bail;
+
+	case QIB_IB_LINKDOWN:
+		dd->f_set_ib_cfg(ppd, QIB_IB_CFG_LSTATE,
+				 IB_LINKCMD_DOWN | IB_LINKINITCMD_POLL);
+		/* don't wait */
+		ret = 0;
+		goto bail;
+
+	case QIB_IB_LINKDOWN_SLEEP:
+		dd->f_set_ib_cfg(ppd, QIB_IB_CFG_LSTATE,
+				 IB_LINKCMD_DOWN | IB_LINKINITCMD_SLEEP);
+		/* don't wait */
+		ret = 0;
+		goto bail;
+
+	case QIB_IB_LINKDOWN_DISABLE:
+		dd->f_set_ib_cfg(ppd, QIB_IB_CFG_LSTATE,
+				 IB_LINKCMD_DOWN | IB_LINKINITCMD_DISABLE);
+		/* don't wait */
+		ret = 0;
+		goto bail;
+
+	case QIB_IB_LINKARM:
+		if (ppd->lflags & QIBL_LINKARMED) {
+			ret = 0;
+			goto bail;
+		}
+		if (!(ppd->lflags & (QIBL_LINKINIT | QIBL_LINKACTIVE))) {
+			ret = -EINVAL;
+			goto bail;
+		}
+		/*
+		 * Since the port can be ACTIVE when we ask for ARMED,
+		 * clear QIBL_LINKV so we can wait for a transition.
+		 * If the link isn't ARMED, then something else happened
+		 * and there is no point waiting for ARMED.
+		 */
+		spin_lock_irqsave(&ppd->lflags_lock, flags);
+		ppd->lflags &= ~QIBL_LINKV;
+		spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+		dd->f_set_ib_cfg(ppd, QIB_IB_CFG_LSTATE,
+				 IB_LINKCMD_ARMED | IB_LINKINITCMD_NOP);
+		lstate = QIBL_LINKV;
+		break;
+
+	case QIB_IB_LINKACTIVE:
+		if (ppd->lflags & QIBL_LINKACTIVE) {
+			ret = 0;
+			goto bail;
+		}
+		if (!(ppd->lflags & QIBL_LINKARMED)) {
+			ret = -EINVAL;
+			goto bail;
+		}
+		dd->f_set_ib_cfg(ppd, QIB_IB_CFG_LSTATE,
+				 IB_LINKCMD_ACTIVE | IB_LINKINITCMD_NOP);
+		lstate = QIBL_LINKACTIVE;
+		break;
+
+	default:
+		ret = -EINVAL;
+		goto bail;
+	}
+	ret = qib_wait_linkstate(ppd, lstate, 10);
+
+bail:
+	return ret;
+}
+
+/*
+ * Get address of eager buffer from it's index (allocated in chunks, not
+ * contiguous).
+ */
+static inline void *qib_get_egrbuf(const struct qib_ctxtdata *rcd, u32 etail)
+{
+	const u32 chunk = etail / rcd->rcvegrbufs_perchunk;
+	const u32 idx =  etail % rcd->rcvegrbufs_perchunk;
+
+	return rcd->rcvegrbuf[chunk] + idx * rcd->dd->rcvegrbufsize;
+}
+
+/*
+ * Returns 1 if error was a CRC, else 0.
+ * Needed for some chip's synthesized error counters.
+ */
+static u32 qib_rcv_hdrerr(struct qib_pportdata *ppd, u32 ctxt,
+			  u32 eflags, u32 l, u32 etail, __le32 *rhf_addr,
+			  struct qib_message_header *hdr)
+{
+	u32 ret = 0;
+
+	if (eflags & (QLOGIC_IB_RHF_H_ICRCERR | QLOGIC_IB_RHF_H_VCRCERR))
+		ret = 1;
+	return ret;
+}
+
+/*
+ * qib_kreceive - receive a packet
+ * @rcd: the qlogic_ib context
+ * @llic: gets count of good packets needed to clear lli,
+ *          (used with chips that need need to track crcs for lli)
+ *
+ * called from interrupt handler for errors or receive interrupt
+ * Returns number of CRC error packets, needed by some chips for
+ * local link integrity tracking.   crcs are adjusted down by following
+ * good packets, if any, and count of good packets is also tracked.
+ */
+u32 qib_kreceive(struct qib_ctxtdata *rcd, u32 *llic, u32 *npkts)
+{
+	struct qib_devdata *dd = rcd->dd;
+	struct qib_pportdata *ppd = rcd->ppd;
+	__le32 *rhf_addr;
+	void *ebuf;
+	const u32 rsize = dd->rcvhdrentsize;        /* words */
+	const u32 maxcnt = dd->rcvhdrcnt * rsize;   /* words */
+	u32 etail = -1, l, hdrqtail;
+	struct qib_message_header *hdr;
+	u32 eflags, etype, tlen, i = 0, updegr = 0, crcs = 0;
+	int last;
+	u64 lval;
+	struct qib_qp *qp, *nqp;
+
+	l = rcd->head;
+	rhf_addr = (__le32 *) rcd->rcvhdrq + l + dd->rhf_offset;
+	if (dd->flags & QIB_NODMA_RTAIL) {
+		u32 seq = qib_hdrget_seq(rhf_addr);
+		if (seq != rcd->seq_cnt)
+			goto bail;
+		hdrqtail = 0;
+	} else {
+		hdrqtail = qib_get_rcvhdrtail(rcd);
+		if (l == hdrqtail)
+			goto bail;
+		smp_rmb();  /* prevent speculative reads of dma'ed hdrq */
+	}
+
+	for (last = 0, i = 1; !last; i += !last) {
+		hdr = dd->f_get_msgheader(dd, rhf_addr);
+		eflags = qib_hdrget_err_flags(rhf_addr);
+		etype = qib_hdrget_rcv_type(rhf_addr);
+		/* total length */
+		tlen = qib_hdrget_length_in_bytes(rhf_addr);
+		ebuf = NULL;
+		if ((dd->flags & QIB_NODMA_RTAIL) ?
+		    qib_hdrget_use_egr_buf(rhf_addr) :
+		    (etype != RCVHQ_RCV_TYPE_EXPECTED)) {
+			etail = qib_hdrget_index(rhf_addr);
+			updegr = 1;
+			if (tlen > sizeof(*hdr) ||
+			    etype >= RCVHQ_RCV_TYPE_NON_KD)
+				ebuf = qib_get_egrbuf(rcd, etail);
+		}
+		if (!eflags) {
+			u16 lrh_len = be16_to_cpu(hdr->lrh[2]) << 2;
+
+			if (lrh_len != tlen) {
+				qib_stats.sps_lenerrs++;
+				goto move_along;
+			}
+		}
+		if (etype == RCVHQ_RCV_TYPE_NON_KD && !eflags &&
+		    ebuf == NULL &&
+		    tlen > (dd->rcvhdrentsize - 2 + 1 -
+				qib_hdrget_offset(rhf_addr)) << 2) {
+			goto move_along;
+		}
+
+		/*
+		 * Both tiderr and qibhdrerr are set for all plain IB
+		 * packets; only qibhdrerr should be set.
+		 */
+		if (unlikely(eflags))
+			crcs += qib_rcv_hdrerr(ppd, rcd->ctxt, eflags, l,
+					       etail, rhf_addr, hdr);
+		else if (etype == RCVHQ_RCV_TYPE_NON_KD) {
+			qib_ib_rcv(rcd, hdr, ebuf, tlen);
+			if (crcs)
+				crcs--;
+			else if (llic && *llic)
+				--*llic;
+		}
+move_along:
+		l += rsize;
+		if (l >= maxcnt)
+			l = 0;
+		rhf_addr = (__le32 *) rcd->rcvhdrq + l + dd->rhf_offset;
+		if (dd->flags & QIB_NODMA_RTAIL) {
+			u32 seq = qib_hdrget_seq(rhf_addr);
+
+			if (++rcd->seq_cnt > 13)
+				rcd->seq_cnt = 1;
+			if (seq != rcd->seq_cnt)
+				last = 1;
+		} else if (l == hdrqtail)
+			last = 1;
+		/*
+		 * Update head regs etc., every 16 packets, if not last pkt,
+		 * to help prevent rcvhdrq overflows, when many packets
+		 * are processed and queue is nearly full.
+		 * Don't request an interrupt for intermediate updates.
+		 */
+		lval = l;
+		if (!last && !(i & 0xf)) {
+			dd->f_update_usrhead(rcd, lval, updegr, etail);
+			updegr = 0;
+		}
+	}
+
+	rcd->head = l;
+	rcd->pkt_count += i;
+
+	/*
+	 * Iterate over all QPs waiting to respond.
+	 * The list won't change since the IRQ is only run on one CPU.
+	 */
+	list_for_each_entry_safe(qp, nqp, &rcd->qp_wait_list, rspwait) {
+		list_del_init(&qp->rspwait);
+		if (qp->r_flags & QIB_R_RSP_NAK) {
+			qp->r_flags &= ~QIB_R_RSP_NAK;
+			qib_send_rc_ack(qp);
+		}
+		if (qp->r_flags & QIB_R_RSP_SEND) {
+			unsigned long flags;
+
+			qp->r_flags &= ~QIB_R_RSP_SEND;
+			spin_lock_irqsave(&qp->s_lock, flags);
+			if (ib_qib_state_ops[qp->state] &
+					QIB_PROCESS_OR_FLUSH_SEND)
+				qib_schedule_send(qp);
+			spin_unlock_irqrestore(&qp->s_lock, flags);
+		}
+		if (atomic_dec_and_test(&qp->refcount))
+			wake_up(&qp->wait);
+	}
+
+bail:
+	/* Report number of packets consumed */
+	if (npkts)
+		*npkts = i;
+
+	/*
+	 * Always write head at end, and setup rcv interrupt, even
+	 * if no packets were processed.
+	 */
+	lval = (u64)rcd->head | dd->rhdrhead_intr_off;
+	dd->f_update_usrhead(rcd, lval, updegr, etail);
+	return crcs;
+}
+
+/**
+ * qib_set_mtu - set the MTU
+ * @ppd: the perport data
+ * @arg: the new MTU
+ *
+ * We can handle "any" incoming size, the issue here is whether we
+ * need to restrict our outgoing size.   For now, we don't do any
+ * sanity checking on this, and we don't deal with what happens to
+ * programs that are already running when the size changes.
+ * NOTE: changing the MTU will usually cause the IBC to go back to
+ * link INIT state...
+ */
+int qib_set_mtu(struct qib_pportdata *ppd, u16 arg)
+{
+	u32 piosize;
+	int ret, chk;
+
+	if (arg != 256 && arg != 512 && arg != 1024 && arg != 2048 &&
+	    arg != 4096) {
+		ret = -EINVAL;
+		goto bail;
+	}
+	chk = ib_mtu_enum_to_int(qib_ibmtu);
+	if (chk > 0 && arg > chk) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	piosize = ppd->ibmaxlen;
+	ppd->ibmtu = arg;
+
+	if (arg >= (piosize - QIB_PIO_MAXIBHDR)) {
+		/* Only if it's not the initial value (or reset to it) */
+		if (piosize != ppd->init_ibmaxlen) {
+			if (arg > piosize && arg <= ppd->init_ibmaxlen)
+				piosize = ppd->init_ibmaxlen - 2 * sizeof(u32);
+			ppd->ibmaxlen = piosize;
+		}
+	} else if ((arg + QIB_PIO_MAXIBHDR) != ppd->ibmaxlen) {
+		piosize = arg + QIB_PIO_MAXIBHDR - 2 * sizeof(u32);
+		ppd->ibmaxlen = piosize;
+	}
+
+	ppd->dd->f_set_ib_cfg(ppd, QIB_IB_CFG_MTU, 0);
+
+	ret = 0;
+
+bail:
+	return ret;
+}
+
+int qib_set_lid(struct qib_pportdata *ppd, u32 lid, u8 lmc)
+{
+	struct qib_devdata *dd = ppd->dd;
+	ppd->lid = lid;
+	ppd->lmc = lmc;
+
+	dd->f_set_ib_cfg(ppd, QIB_IB_CFG_LIDLMC,
+			 lid | (~((1U << lmc) - 1)) << 16);
+
+	qib_devinfo(dd->pcidev, "IB%u:%u got a lid: 0x%x\n",
+		    dd->unit, ppd->port, lid);
+
+	return 0;
+}
+
+/*
+ * Following deal with the "obviously simple" task of overriding the state
+ * of the LEDS, which normally indicate link physical and logical status.
+ * The complications arise in dealing with different hardware mappings
+ * and the board-dependent routine being called from interrupts.
+ * and then there's the requirement to _flash_ them.
+ */
+#define LED_OVER_FREQ_SHIFT 8
+#define LED_OVER_FREQ_MASK (0xFF<<LED_OVER_FREQ_SHIFT)
+/* Below is "non-zero" to force override, but both actual LEDs are off */
+#define LED_OVER_BOTH_OFF (8)
+
+static void qib_run_led_override(unsigned long opaque)
+{
+	struct qib_pportdata *ppd = (struct qib_pportdata *)opaque;
+	struct qib_devdata *dd = ppd->dd;
+	int timeoff;
+	int ph_idx;
+
+	if (!(dd->flags & QIB_INITTED))
+		return;
+
+	ph_idx = ppd->led_override_phase++ & 1;
+	ppd->led_override = ppd->led_override_vals[ph_idx];
+	timeoff = ppd->led_override_timeoff;
+
+	dd->f_setextled(ppd, 1);
+	/*
+	 * don't re-fire the timer if user asked for it to be off; we let
+	 * it fire one more time after they turn it off to simplify
+	 */
+	if (ppd->led_override_vals[0] || ppd->led_override_vals[1])
+		mod_timer(&ppd->led_override_timer, jiffies + timeoff);
+}
+
+void qib_set_led_override(struct qib_pportdata *ppd, unsigned int val)
+{
+	struct qib_devdata *dd = ppd->dd;
+	int timeoff, freq;
+
+	if (!(dd->flags & QIB_INITTED))
+		return;
+
+	/* First check if we are blinking. If not, use 1HZ polling */
+	timeoff = HZ;
+	freq = (val & LED_OVER_FREQ_MASK) >> LED_OVER_FREQ_SHIFT;
+
+	if (freq) {
+		/* For blink, set each phase from one nybble of val */
+		ppd->led_override_vals[0] = val & 0xF;
+		ppd->led_override_vals[1] = (val >> 4) & 0xF;
+		timeoff = (HZ << 4)/freq;
+	} else {
+		/* Non-blink set both phases the same. */
+		ppd->led_override_vals[0] = val & 0xF;
+		ppd->led_override_vals[1] = val & 0xF;
+	}
+	ppd->led_override_timeoff = timeoff;
+
+	/*
+	 * If the timer has not already been started, do so. Use a "quick"
+	 * timeout so the function will be called soon, to look at our request.
+	 */
+	if (atomic_inc_return(&ppd->led_override_timer_active) == 1) {
+		/* Need to start timer */
+		init_timer(&ppd->led_override_timer);
+		ppd->led_override_timer.function = qib_run_led_override;
+		ppd->led_override_timer.data = (unsigned long) ppd;
+		ppd->led_override_timer.expires = jiffies + 1;
+		add_timer(&ppd->led_override_timer);
+	} else {
+		if (ppd->led_override_vals[0] || ppd->led_override_vals[1])
+			mod_timer(&ppd->led_override_timer, jiffies + 1);
+		atomic_dec(&ppd->led_override_timer_active);
+	}
+}
+
+/**
+ * qib_reset_device - reset the chip if possible
+ * @unit: the device to reset
+ *
+ * Whether or not reset is successful, we attempt to re-initialize the chip
+ * (that is, much like a driver unload/reload).  We clear the INITTED flag
+ * so that the various entry points will fail until we reinitialize.  For
+ * now, we only allow this if no user contexts are open that use chip resources
+ */
+int qib_reset_device(int unit)
+{
+	int ret, i;
+	struct qib_devdata *dd = qib_lookup(unit);
+	struct qib_pportdata *ppd;
+	unsigned long flags;
+	int pidx;
+
+	if (!dd) {
+		ret = -ENODEV;
+		goto bail;
+	}
+
+	qib_devinfo(dd->pcidev, "Reset on unit %u requested\n", unit);
+
+	if (!dd->kregbase || !(dd->flags & QIB_PRESENT)) {
+		qib_devinfo(dd->pcidev, "Invalid unit number %u or "
+			    "not initialized or not present\n", unit);
+		ret = -ENXIO;
+		goto bail;
+	}
+
+	spin_lock_irqsave(&dd->uctxt_lock, flags);
+	if (dd->rcd)
+		for (i = dd->first_user_ctxt; i < dd->cfgctxts; i++) {
+			if (!dd->rcd[i] || !dd->rcd[i]->cnt)
+				continue;
+			spin_unlock_irqrestore(&dd->uctxt_lock, flags);
+			ret = -EBUSY;
+			goto bail;
+		}
+	spin_unlock_irqrestore(&dd->uctxt_lock, flags);
+
+	for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+		ppd = dd->pport + pidx;
+		if (atomic_read(&ppd->led_override_timer_active)) {
+			/* Need to stop LED timer, _then_ shut off LEDs */
+			del_timer_sync(&ppd->led_override_timer);
+			atomic_set(&ppd->led_override_timer_active, 0);
+		}
+
+		/* Shut off LEDs after we are sure timer is not running */
+		ppd->led_override = LED_OVER_BOTH_OFF;
+		dd->f_setextled(ppd, 0);
+		if (dd->flags & QIB_HAS_SEND_DMA)
+			qib_teardown_sdma(ppd);
+	}
+
+	ret = dd->f_reset(dd);
+	if (ret == 1)
+		ret = qib_init(dd, 1);
+	else
+		ret = -EAGAIN;
+	if (ret)
+		qib_dev_err(dd, "Reinitialize unit %u after "
+			    "reset failed with %d\n", unit, ret);
+	else
+		qib_devinfo(dd->pcidev, "Reinitialized unit %u after "
+			    "resetting\n", unit);
+
+bail:
+	return ret;
+}
diff --git a/drivers/infiniband/hw/qib/qib_eeprom.c b/drivers/infiniband/hw/qib/qib_eeprom.c
new file mode 100644
index 000000000000..92d9cfe98a68
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_eeprom.c
@@ -0,0 +1,451 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/delay.h>
+#include <linux/pci.h>
+#include <linux/vmalloc.h>
+
+#include "qib.h"
+
+/*
+ * Functions specific to the serial EEPROM on cards handled by ib_qib.
+ * The actual serail interface code is in qib_twsi.c. This file is a client
+ */
+
+/**
+ * qib_eeprom_read - receives bytes from the eeprom via I2C
+ * @dd: the qlogic_ib device
+ * @eeprom_offset: address to read from
+ * @buffer: where to store result
+ * @len: number of bytes to receive
+ */
+int qib_eeprom_read(struct qib_devdata *dd, u8 eeprom_offset,
+		    void *buff, int len)
+{
+	int ret;
+
+	ret = mutex_lock_interruptible(&dd->eep_lock);
+	if (!ret) {
+		ret = qib_twsi_reset(dd);
+		if (ret)
+			qib_dev_err(dd, "EEPROM Reset for read failed\n");
+		else
+			ret = qib_twsi_blk_rd(dd, dd->twsi_eeprom_dev,
+					      eeprom_offset, buff, len);
+		mutex_unlock(&dd->eep_lock);
+	}
+
+	return ret;
+}
+
+/*
+ * Actually update the eeprom, first doing write enable if
+ * needed, then restoring write enable state.
+ * Must be called with eep_lock held
+ */
+static int eeprom_write_with_enable(struct qib_devdata *dd, u8 offset,
+		     const void *buf, int len)
+{
+	int ret, pwen;
+
+	pwen = dd->f_eeprom_wen(dd, 1);
+	ret = qib_twsi_reset(dd);
+	if (ret)
+		qib_dev_err(dd, "EEPROM Reset for write failed\n");
+	else
+		ret = qib_twsi_blk_wr(dd, dd->twsi_eeprom_dev,
+				      offset, buf, len);
+	dd->f_eeprom_wen(dd, pwen);
+	return ret;
+}
+
+/**
+ * qib_eeprom_write - writes data to the eeprom via I2C
+ * @dd: the qlogic_ib device
+ * @eeprom_offset: where to place data
+ * @buffer: data to write
+ * @len: number of bytes to write
+ */
+int qib_eeprom_write(struct qib_devdata *dd, u8 eeprom_offset,
+		     const void *buff, int len)
+{
+	int ret;
+
+	ret = mutex_lock_interruptible(&dd->eep_lock);
+	if (!ret) {
+		ret = eeprom_write_with_enable(dd, eeprom_offset, buff, len);
+		mutex_unlock(&dd->eep_lock);
+	}
+
+	return ret;
+}
+
+static u8 flash_csum(struct qib_flash *ifp, int adjust)
+{
+	u8 *ip = (u8 *) ifp;
+	u8 csum = 0, len;
+
+	/*
+	 * Limit length checksummed to max length of actual data.
+	 * Checksum of erased eeprom will still be bad, but we avoid
+	 * reading past the end of the buffer we were passed.
+	 */
+	len = ifp->if_length;
+	if (len > sizeof(struct qib_flash))
+		len = sizeof(struct qib_flash);
+	while (len--)
+		csum += *ip++;
+	csum -= ifp->if_csum;
+	csum = ~csum;
+	if (adjust)
+		ifp->if_csum = csum;
+
+	return csum;
+}
+
+/**
+ * qib_get_eeprom_info- get the GUID et al. from the TSWI EEPROM device
+ * @dd: the qlogic_ib device
+ *
+ * We have the capability to use the nguid field, and get
+ * the guid from the first chip's flash, to use for all of them.
+ */
+void qib_get_eeprom_info(struct qib_devdata *dd)
+{
+	void *buf;
+	struct qib_flash *ifp;
+	__be64 guid;
+	int len, eep_stat;
+	u8 csum, *bguid;
+	int t = dd->unit;
+	struct qib_devdata *dd0 = qib_lookup(0);
+
+	if (t && dd0->nguid > 1 && t <= dd0->nguid) {
+		u8 oguid;
+		dd->base_guid = dd0->base_guid;
+		bguid = (u8 *) &dd->base_guid;
+
+		oguid = bguid[7];
+		bguid[7] += t;
+		if (oguid > bguid[7]) {
+			if (bguid[6] == 0xff) {
+				if (bguid[5] == 0xff) {
+					qib_dev_err(dd, "Can't set %s GUID"
+						    " from base, wraps to"
+						    " OUI!\n",
+						    qib_get_unit_name(t));
+					dd->base_guid = 0;
+					goto bail;
+				}
+				bguid[5]++;
+			}
+			bguid[6]++;
+		}
+		dd->nguid = 1;
+		goto bail;
+	}
+
+	/*
+	 * Read full flash, not just currently used part, since it may have
+	 * been written with a newer definition.
+	 * */
+	len = sizeof(struct qib_flash);
+	buf = vmalloc(len);
+	if (!buf) {
+		qib_dev_err(dd, "Couldn't allocate memory to read %u "
+			    "bytes from eeprom for GUID\n", len);
+		goto bail;
+	}
+
+	/*
+	 * Use "public" eeprom read function, which does locking and
+	 * figures out device. This will migrate to chip-specific.
+	 */
+	eep_stat = qib_eeprom_read(dd, 0, buf, len);
+
+	if (eep_stat) {
+		qib_dev_err(dd, "Failed reading GUID from eeprom\n");
+		goto done;
+	}
+	ifp = (struct qib_flash *)buf;
+
+	csum = flash_csum(ifp, 0);
+	if (csum != ifp->if_csum) {
+		qib_devinfo(dd->pcidev, "Bad I2C flash checksum: "
+			 "0x%x, not 0x%x\n", csum, ifp->if_csum);
+		goto done;
+	}
+	if (*(__be64 *) ifp->if_guid == cpu_to_be64(0) ||
+	    *(__be64 *) ifp->if_guid == ~cpu_to_be64(0)) {
+		qib_dev_err(dd, "Invalid GUID %llx from flash; ignoring\n",
+			    *(unsigned long long *) ifp->if_guid);
+		/* don't allow GUID if all 0 or all 1's */
+		goto done;
+	}
+
+	/* complain, but allow it */
+	if (*(u64 *) ifp->if_guid == 0x100007511000000ULL)
+		qib_devinfo(dd->pcidev, "Warning, GUID %llx is "
+			 "default, probably not correct!\n",
+			 *(unsigned long long *) ifp->if_guid);
+
+	bguid = ifp->if_guid;
+	if (!bguid[0] && !bguid[1] && !bguid[2]) {
+		/*
+		 * Original incorrect GUID format in flash; fix in
+		 * core copy, by shifting up 2 octets; don't need to
+		 * change top octet, since both it and shifted are 0.
+		 */
+		bguid[1] = bguid[3];
+		bguid[2] = bguid[4];
+		bguid[3] = 0;
+		bguid[4] = 0;
+		guid = *(__be64 *) ifp->if_guid;
+	} else
+		guid = *(__be64 *) ifp->if_guid;
+	dd->base_guid = guid;
+	dd->nguid = ifp->if_numguid;
+	/*
+	 * Things are slightly complicated by the desire to transparently
+	 * support both the Pathscale 10-digit serial number and the QLogic
+	 * 13-character version.
+	 */
+	if ((ifp->if_fversion > 1) && ifp->if_sprefix[0] &&
+	    ((u8 *) ifp->if_sprefix)[0] != 0xFF) {
+		char *snp = dd->serial;
+
+		/*
+		 * This board has a Serial-prefix, which is stored
+		 * elsewhere for backward-compatibility.
+		 */
+		memcpy(snp, ifp->if_sprefix, sizeof ifp->if_sprefix);
+		snp[sizeof ifp->if_sprefix] = '\0';
+		len = strlen(snp);
+		snp += len;
+		len = (sizeof dd->serial) - len;
+		if (len > sizeof ifp->if_serial)
+			len = sizeof ifp->if_serial;
+		memcpy(snp, ifp->if_serial, len);
+	} else
+		memcpy(dd->serial, ifp->if_serial,
+		       sizeof ifp->if_serial);
+	if (!strstr(ifp->if_comment, "Tested successfully"))
+		qib_dev_err(dd, "Board SN %s did not pass functional "
+			    "test: %s\n", dd->serial, ifp->if_comment);
+
+	memcpy(&dd->eep_st_errs, &ifp->if_errcntp, QIB_EEP_LOG_CNT);
+	/*
+	 * Power-on (actually "active") hours are kept as little-endian value
+	 * in EEPROM, but as seconds in a (possibly as small as 24-bit)
+	 * atomic_t while running.
+	 */
+	atomic_set(&dd->active_time, 0);
+	dd->eep_hrs = ifp->if_powerhour[0] | (ifp->if_powerhour[1] << 8);
+
+done:
+	vfree(buf);
+
+bail:;
+}
+
+/**
+ * qib_update_eeprom_log - copy active-time and error counters to eeprom
+ * @dd: the qlogic_ib device
+ *
+ * Although the time is kept as seconds in the qib_devdata struct, it is
+ * rounded to hours for re-write, as we have only 16 bits in EEPROM.
+ * First-cut code reads whole (expected) struct qib_flash, modifies,
+ * re-writes. Future direction: read/write only what we need, assuming
+ * that the EEPROM had to have been "good enough" for driver init, and
+ * if not, we aren't making it worse.
+ *
+ */
+int qib_update_eeprom_log(struct qib_devdata *dd)
+{
+	void *buf;
+	struct qib_flash *ifp;
+	int len, hi_water;
+	uint32_t new_time, new_hrs;
+	u8 csum;
+	int ret, idx;
+	unsigned long flags;
+
+	/* first, check if we actually need to do anything. */
+	ret = 0;
+	for (idx = 0; idx < QIB_EEP_LOG_CNT; ++idx) {
+		if (dd->eep_st_new_errs[idx]) {
+			ret = 1;
+			break;
+		}
+	}
+	new_time = atomic_read(&dd->active_time);
+
+	if (ret == 0 && new_time < 3600)
+		goto bail;
+
+	/*
+	 * The quick-check above determined that there is something worthy
+	 * of logging, so get current contents and do a more detailed idea.
+	 * read full flash, not just currently used part, since it may have
+	 * been written with a newer definition
+	 */
+	len = sizeof(struct qib_flash);
+	buf = vmalloc(len);
+	ret = 1;
+	if (!buf) {
+		qib_dev_err(dd, "Couldn't allocate memory to read %u "
+			    "bytes from eeprom for logging\n", len);
+		goto bail;
+	}
+
+	/* Grab semaphore and read current EEPROM. If we get an
+	 * error, let go, but if not, keep it until we finish write.
+	 */
+	ret = mutex_lock_interruptible(&dd->eep_lock);
+	if (ret) {
+		qib_dev_err(dd, "Unable to acquire EEPROM for logging\n");
+		goto free_bail;
+	}
+	ret = qib_twsi_blk_rd(dd, dd->twsi_eeprom_dev, 0, buf, len);
+	if (ret) {
+		mutex_unlock(&dd->eep_lock);
+		qib_dev_err(dd, "Unable read EEPROM for logging\n");
+		goto free_bail;
+	}
+	ifp = (struct qib_flash *)buf;
+
+	csum = flash_csum(ifp, 0);
+	if (csum != ifp->if_csum) {
+		mutex_unlock(&dd->eep_lock);
+		qib_dev_err(dd, "EEPROM cks err (0x%02X, S/B 0x%02X)\n",
+			    csum, ifp->if_csum);
+		ret = 1;
+		goto free_bail;
+	}
+	hi_water = 0;
+	spin_lock_irqsave(&dd->eep_st_lock, flags);
+	for (idx = 0; idx < QIB_EEP_LOG_CNT; ++idx) {
+		int new_val = dd->eep_st_new_errs[idx];
+		if (new_val) {
+			/*
+			 * If we have seen any errors, add to EEPROM values
+			 * We need to saturate at 0xFF (255) and we also
+			 * would need to adjust the checksum if we were
+			 * trying to minimize EEPROM traffic
+			 * Note that we add to actual current count in EEPROM,
+			 * in case it was altered while we were running.
+			 */
+			new_val += ifp->if_errcntp[idx];
+			if (new_val > 0xFF)
+				new_val = 0xFF;
+			if (ifp->if_errcntp[idx] != new_val) {
+				ifp->if_errcntp[idx] = new_val;
+				hi_water = offsetof(struct qib_flash,
+						    if_errcntp) + idx;
+			}
+			/*
+			 * update our shadow (used to minimize EEPROM
+			 * traffic), to match what we are about to write.
+			 */
+			dd->eep_st_errs[idx] = new_val;
+			dd->eep_st_new_errs[idx] = 0;
+		}
+	}
+	/*
+	 * Now update active-time. We would like to round to the nearest hour
+	 * but unless atomic_t are sure to be proper signed ints we cannot,
+	 * because we need to account for what we "transfer" to EEPROM and
+	 * if we log an hour at 31 minutes, then we would need to set
+	 * active_time to -29 to accurately count the _next_ hour.
+	 */
+	if (new_time >= 3600) {
+		new_hrs = new_time / 3600;
+		atomic_sub((new_hrs * 3600), &dd->active_time);
+		new_hrs += dd->eep_hrs;
+		if (new_hrs > 0xFFFF)
+			new_hrs = 0xFFFF;
+		dd->eep_hrs = new_hrs;
+		if ((new_hrs & 0xFF) != ifp->if_powerhour[0]) {
+			ifp->if_powerhour[0] = new_hrs & 0xFF;
+			hi_water = offsetof(struct qib_flash, if_powerhour);
+		}
+		if ((new_hrs >> 8) != ifp->if_powerhour[1]) {
+			ifp->if_powerhour[1] = new_hrs >> 8;
+			hi_water = offsetof(struct qib_flash, if_powerhour) + 1;
+		}
+	}
+	/*
+	 * There is a tiny possibility that we could somehow fail to write
+	 * the EEPROM after updating our shadows, but problems from holding
+	 * the spinlock too long are a much bigger issue.
+	 */
+	spin_unlock_irqrestore(&dd->eep_st_lock, flags);
+	if (hi_water) {
+		/* we made some change to the data, uopdate cksum and write */
+		csum = flash_csum(ifp, 1);
+		ret = eeprom_write_with_enable(dd, 0, buf, hi_water + 1);
+	}
+	mutex_unlock(&dd->eep_lock);
+	if (ret)
+		qib_dev_err(dd, "Failed updating EEPROM\n");
+
+free_bail:
+	vfree(buf);
+bail:
+	return ret;
+}
+
+/**
+ * qib_inc_eeprom_err - increment one of the four error counters
+ * that are logged to EEPROM.
+ * @dd: the qlogic_ib device
+ * @eidx: 0..3, the counter to increment
+ * @incr: how much to add
+ *
+ * Each counter is 8-bits, and saturates at 255 (0xFF). They
+ * are copied to the EEPROM (aka flash) whenever qib_update_eeprom_log()
+ * is called, but it can only be called in a context that allows sleep.
+ * This function can be called even at interrupt level.
+ */
+void qib_inc_eeprom_err(struct qib_devdata *dd, u32 eidx, u32 incr)
+{
+	uint new_val;
+	unsigned long flags;
+
+	spin_lock_irqsave(&dd->eep_st_lock, flags);
+	new_val = dd->eep_st_new_errs[eidx] + incr;
+	if (new_val > 255)
+		new_val = 255;
+	dd->eep_st_new_errs[eidx] = new_val;
+	spin_unlock_irqrestore(&dd->eep_st_lock, flags);
+}
diff --git a/drivers/infiniband/hw/qib/qib_file_ops.c b/drivers/infiniband/hw/qib/qib_file_ops.c
new file mode 100644
index 000000000000..a142a9eb5226
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_file_ops.c
@@ -0,0 +1,2317 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009, 2010 QLogic Corporation.
+ * All rights reserved.
+ * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/pci.h>
+#include <linux/poll.h>
+#include <linux/cdev.h>
+#include <linux/swap.h>
+#include <linux/vmalloc.h>
+#include <linux/highmem.h>
+#include <linux/io.h>
+#include <linux/uio.h>
+#include <linux/jiffies.h>
+#include <asm/pgtable.h>
+#include <linux/delay.h>
+
+#include "qib.h"
+#include "qib_common.h"
+#include "qib_user_sdma.h"
+
+static int qib_open(struct inode *, struct file *);
+static int qib_close(struct inode *, struct file *);
+static ssize_t qib_write(struct file *, const char __user *, size_t, loff_t *);
+static ssize_t qib_aio_write(struct kiocb *, const struct iovec *,
+			     unsigned long, loff_t);
+static unsigned int qib_poll(struct file *, struct poll_table_struct *);
+static int qib_mmapf(struct file *, struct vm_area_struct *);
+
+static const struct file_operations qib_file_ops = {
+	.owner = THIS_MODULE,
+	.write = qib_write,
+	.aio_write = qib_aio_write,
+	.open = qib_open,
+	.release = qib_close,
+	.poll = qib_poll,
+	.mmap = qib_mmapf
+};
+
+/*
+ * Convert kernel virtual addresses to physical addresses so they don't
+ * potentially conflict with the chip addresses used as mmap offsets.
+ * It doesn't really matter what mmap offset we use as long as we can
+ * interpret it correctly.
+ */
+static u64 cvt_kvaddr(void *p)
+{
+	struct page *page;
+	u64 paddr = 0;
+
+	page = vmalloc_to_page(p);
+	if (page)
+		paddr = page_to_pfn(page) << PAGE_SHIFT;
+
+	return paddr;
+}
+
+static int qib_get_base_info(struct file *fp, void __user *ubase,
+			     size_t ubase_size)
+{
+	struct qib_ctxtdata *rcd = ctxt_fp(fp);
+	int ret = 0;
+	struct qib_base_info *kinfo = NULL;
+	struct qib_devdata *dd = rcd->dd;
+	struct qib_pportdata *ppd = rcd->ppd;
+	unsigned subctxt_cnt;
+	int shared, master;
+	size_t sz;
+
+	subctxt_cnt = rcd->subctxt_cnt;
+	if (!subctxt_cnt) {
+		shared = 0;
+		master = 0;
+		subctxt_cnt = 1;
+	} else {
+		shared = 1;
+		master = !subctxt_fp(fp);
+	}
+
+	sz = sizeof(*kinfo);
+	/* If context sharing is not requested, allow the old size structure */
+	if (!shared)
+		sz -= 7 * sizeof(u64);
+	if (ubase_size < sz) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	kinfo = kzalloc(sizeof(*kinfo), GFP_KERNEL);
+	if (kinfo == NULL) {
+		ret = -ENOMEM;
+		goto bail;
+	}
+
+	ret = dd->f_get_base_info(rcd, kinfo);
+	if (ret < 0)
+		goto bail;
+
+	kinfo->spi_rcvhdr_cnt = dd->rcvhdrcnt;
+	kinfo->spi_rcvhdrent_size = dd->rcvhdrentsize;
+	kinfo->spi_tidegrcnt = rcd->rcvegrcnt;
+	kinfo->spi_rcv_egrbufsize = dd->rcvegrbufsize;
+	/*
+	 * have to mmap whole thing
+	 */
+	kinfo->spi_rcv_egrbuftotlen =
+		rcd->rcvegrbuf_chunks * rcd->rcvegrbuf_size;
+	kinfo->spi_rcv_egrperchunk = rcd->rcvegrbufs_perchunk;
+	kinfo->spi_rcv_egrchunksize = kinfo->spi_rcv_egrbuftotlen /
+		rcd->rcvegrbuf_chunks;
+	kinfo->spi_tidcnt = dd->rcvtidcnt / subctxt_cnt;
+	if (master)
+		kinfo->spi_tidcnt += dd->rcvtidcnt % subctxt_cnt;
+	/*
+	 * for this use, may be cfgctxts summed over all chips that
+	 * are are configured and present
+	 */
+	kinfo->spi_nctxts = dd->cfgctxts;
+	/* unit (chip/board) our context is on */
+	kinfo->spi_unit = dd->unit;
+	kinfo->spi_port = ppd->port;
+	/* for now, only a single page */
+	kinfo->spi_tid_maxsize = PAGE_SIZE;
+
+	/*
+	 * Doing this per context, and based on the skip value, etc.  This has
+	 * to be the actual buffer size, since the protocol code treats it
+	 * as an array.
+	 *
+	 * These have to be set to user addresses in the user code via mmap.
+	 * These values are used on return to user code for the mmap target
+	 * addresses only.  For 32 bit, same 44 bit address problem, so use
+	 * the physical address, not virtual.  Before 2.6.11, using the
+	 * page_address() macro worked, but in 2.6.11, even that returns the
+	 * full 64 bit address (upper bits all 1's).  So far, using the
+	 * physical addresses (or chip offsets, for chip mapping) works, but
+	 * no doubt some future kernel release will change that, and we'll be
+	 * on to yet another method of dealing with this.
+	 * Normally only one of rcvhdr_tailaddr or rhf_offset is useful
+	 * since the chips with non-zero rhf_offset don't normally
+	 * enable tail register updates to host memory, but for testing,
+	 * both can be enabled and used.
+	 */
+	kinfo->spi_rcvhdr_base = (u64) rcd->rcvhdrq_phys;
+	kinfo->spi_rcvhdr_tailaddr = (u64) rcd->rcvhdrqtailaddr_phys;
+	kinfo->spi_rhf_offset = dd->rhf_offset;
+	kinfo->spi_rcv_egrbufs = (u64) rcd->rcvegr_phys;
+	kinfo->spi_pioavailaddr = (u64) dd->pioavailregs_phys;
+	/* setup per-unit (not port) status area for user programs */
+	kinfo->spi_status = (u64) kinfo->spi_pioavailaddr +
+		(char *) ppd->statusp -
+		(char *) dd->pioavailregs_dma;
+	kinfo->spi_uregbase = (u64) dd->uregbase + dd->ureg_align * rcd->ctxt;
+	if (!shared) {
+		kinfo->spi_piocnt = rcd->piocnt;
+		kinfo->spi_piobufbase = (u64) rcd->piobufs;
+		kinfo->spi_sendbuf_status = cvt_kvaddr(rcd->user_event_mask);
+	} else if (master) {
+		kinfo->spi_piocnt = (rcd->piocnt / subctxt_cnt) +
+				    (rcd->piocnt % subctxt_cnt);
+		/* Master's PIO buffers are after all the slave's */
+		kinfo->spi_piobufbase = (u64) rcd->piobufs +
+			dd->palign *
+			(rcd->piocnt - kinfo->spi_piocnt);
+	} else {
+		unsigned slave = subctxt_fp(fp) - 1;
+
+		kinfo->spi_piocnt = rcd->piocnt / subctxt_cnt;
+		kinfo->spi_piobufbase = (u64) rcd->piobufs +
+			dd->palign * kinfo->spi_piocnt * slave;
+	}
+
+	if (shared) {
+		kinfo->spi_sendbuf_status =
+			cvt_kvaddr(&rcd->user_event_mask[subctxt_fp(fp)]);
+		/* only spi_subctxt_* fields should be set in this block! */
+		kinfo->spi_subctxt_uregbase = cvt_kvaddr(rcd->subctxt_uregbase);
+
+		kinfo->spi_subctxt_rcvegrbuf =
+			cvt_kvaddr(rcd->subctxt_rcvegrbuf);
+		kinfo->spi_subctxt_rcvhdr_base =
+			cvt_kvaddr(rcd->subctxt_rcvhdr_base);
+	}
+
+	/*
+	 * All user buffers are 2KB buffers.  If we ever support
+	 * giving 4KB buffers to user processes, this will need some
+	 * work.  Can't use piobufbase directly, because it has
+	 * both 2K and 4K buffer base values.
+	 */
+	kinfo->spi_pioindex = (kinfo->spi_piobufbase - dd->pio2k_bufbase) /
+		dd->palign;
+	kinfo->spi_pioalign = dd->palign;
+	kinfo->spi_qpair = QIB_KD_QP;
+	/*
+	 * user mode PIO buffers are always 2KB, even when 4KB can
+	 * be received, and sent via the kernel; this is ibmaxlen
+	 * for 2K MTU.
+	 */
+	kinfo->spi_piosize = dd->piosize2k - 2 * sizeof(u32);
+	kinfo->spi_mtu = ppd->ibmaxlen; /* maxlen, not ibmtu */
+	kinfo->spi_ctxt = rcd->ctxt;
+	kinfo->spi_subctxt = subctxt_fp(fp);
+	kinfo->spi_sw_version = QIB_KERN_SWVERSION;
+	kinfo->spi_sw_version |= 1U << 31; /* QLogic-built, not kernel.org */
+	kinfo->spi_hw_version = dd->revision;
+
+	if (master)
+		kinfo->spi_runtime_flags |= QIB_RUNTIME_MASTER;
+
+	sz = (ubase_size < sizeof(*kinfo)) ? ubase_size : sizeof(*kinfo);
+	if (copy_to_user(ubase, kinfo, sz))
+		ret = -EFAULT;
+bail:
+	kfree(kinfo);
+	return ret;
+}
+
+/**
+ * qib_tid_update - update a context TID
+ * @rcd: the context
+ * @fp: the qib device file
+ * @ti: the TID information
+ *
+ * The new implementation as of Oct 2004 is that the driver assigns
+ * the tid and returns it to the caller.   To reduce search time, we
+ * keep a cursor for each context, walking the shadow tid array to find
+ * one that's not in use.
+ *
+ * For now, if we can't allocate the full list, we fail, although
+ * in the long run, we'll allocate as many as we can, and the
+ * caller will deal with that by trying the remaining pages later.
+ * That means that when we fail, we have to mark the tids as not in
+ * use again, in our shadow copy.
+ *
+ * It's up to the caller to free the tids when they are done.
+ * We'll unlock the pages as they free them.
+ *
+ * Also, right now we are locking one page at a time, but since
+ * the intended use of this routine is for a single group of
+ * virtually contiguous pages, that should change to improve
+ * performance.
+ */
+static int qib_tid_update(struct qib_ctxtdata *rcd, struct file *fp,
+			  const struct qib_tid_info *ti)
+{
+	int ret = 0, ntids;
+	u32 tid, ctxttid, cnt, i, tidcnt, tidoff;
+	u16 *tidlist;
+	struct qib_devdata *dd = rcd->dd;
+	u64 physaddr;
+	unsigned long vaddr;
+	u64 __iomem *tidbase;
+	unsigned long tidmap[8];
+	struct page **pagep = NULL;
+	unsigned subctxt = subctxt_fp(fp);
+
+	if (!dd->pageshadow) {
+		ret = -ENOMEM;
+		goto done;
+	}
+
+	cnt = ti->tidcnt;
+	if (!cnt) {
+		ret = -EFAULT;
+		goto done;
+	}
+	ctxttid = rcd->ctxt * dd->rcvtidcnt;
+	if (!rcd->subctxt_cnt) {
+		tidcnt = dd->rcvtidcnt;
+		tid = rcd->tidcursor;
+		tidoff = 0;
+	} else if (!subctxt) {
+		tidcnt = (dd->rcvtidcnt / rcd->subctxt_cnt) +
+			 (dd->rcvtidcnt % rcd->subctxt_cnt);
+		tidoff = dd->rcvtidcnt - tidcnt;
+		ctxttid += tidoff;
+		tid = tidcursor_fp(fp);
+	} else {
+		tidcnt = dd->rcvtidcnt / rcd->subctxt_cnt;
+		tidoff = tidcnt * (subctxt - 1);
+		ctxttid += tidoff;
+		tid = tidcursor_fp(fp);
+	}
+	if (cnt > tidcnt) {
+		/* make sure it all fits in tid_pg_list */
+		qib_devinfo(dd->pcidev, "Process tried to allocate %u "
+			 "TIDs, only trying max (%u)\n", cnt, tidcnt);
+		cnt = tidcnt;
+	}
+	pagep = (struct page **) rcd->tid_pg_list;
+	tidlist = (u16 *) &pagep[dd->rcvtidcnt];
+	pagep += tidoff;
+	tidlist += tidoff;
+
+	memset(tidmap, 0, sizeof(tidmap));
+	/* before decrement; chip actual # */
+	ntids = tidcnt;
+	tidbase = (u64 __iomem *) (((char __iomem *) dd->kregbase) +
+				   dd->rcvtidbase +
+				   ctxttid * sizeof(*tidbase));
+
+	/* virtual address of first page in transfer */
+	vaddr = ti->tidvaddr;
+	if (!access_ok(VERIFY_WRITE, (void __user *) vaddr,
+		       cnt * PAGE_SIZE)) {
+		ret = -EFAULT;
+		goto done;
+	}
+	ret = qib_get_user_pages(vaddr, cnt, pagep);
+	if (ret) {
+		/*
+		 * if (ret == -EBUSY)
+		 * We can't continue because the pagep array won't be
+		 * initialized. This should never happen,
+		 * unless perhaps the user has mpin'ed the pages
+		 * themselves.
+		 */
+		qib_devinfo(dd->pcidev,
+			 "Failed to lock addr %p, %u pages: "
+			 "errno %d\n", (void *) vaddr, cnt, -ret);
+		goto done;
+	}
+	for (i = 0; i < cnt; i++, vaddr += PAGE_SIZE) {
+		for (; ntids--; tid++) {
+			if (tid == tidcnt)
+				tid = 0;
+			if (!dd->pageshadow[ctxttid + tid])
+				break;
+		}
+		if (ntids < 0) {
+			/*
+			 * Oops, wrapped all the way through their TIDs,
+			 * and didn't have enough free; see comments at
+			 * start of routine
+			 */
+			i--;    /* last tidlist[i] not filled in */
+			ret = -ENOMEM;
+			break;
+		}
+		tidlist[i] = tid + tidoff;
+		/* we "know" system pages and TID pages are same size */
+		dd->pageshadow[ctxttid + tid] = pagep[i];
+		dd->physshadow[ctxttid + tid] =
+			qib_map_page(dd->pcidev, pagep[i], 0, PAGE_SIZE,
+				     PCI_DMA_FROMDEVICE);
+		/*
+		 * don't need atomic or it's overhead
+		 */
+		__set_bit(tid, tidmap);
+		physaddr = dd->physshadow[ctxttid + tid];
+		/* PERFORMANCE: below should almost certainly be cached */
+		dd->f_put_tid(dd, &tidbase[tid],
+				  RCVHQ_RCV_TYPE_EXPECTED, physaddr);
+		/*
+		 * don't check this tid in qib_ctxtshadow, since we
+		 * just filled it in; start with the next one.
+		 */
+		tid++;
+	}
+
+	if (ret) {
+		u32 limit;
+cleanup:
+		/* jump here if copy out of updated info failed... */
+		/* same code that's in qib_free_tid() */
+		limit = sizeof(tidmap) * BITS_PER_BYTE;
+		if (limit > tidcnt)
+			/* just in case size changes in future */
+			limit = tidcnt;
+		tid = find_first_bit((const unsigned long *)tidmap, limit);
+		for (; tid < limit; tid++) {
+			if (!test_bit(tid, tidmap))
+				continue;
+			if (dd->pageshadow[ctxttid + tid]) {
+				dma_addr_t phys;
+
+				phys = dd->physshadow[ctxttid + tid];
+				dd->physshadow[ctxttid + tid] = dd->tidinvalid;
+				/* PERFORMANCE: below should almost certainly
+				 * be cached
+				 */
+				dd->f_put_tid(dd, &tidbase[tid],
+					      RCVHQ_RCV_TYPE_EXPECTED,
+					      dd->tidinvalid);
+				pci_unmap_page(dd->pcidev, phys, PAGE_SIZE,
+					       PCI_DMA_FROMDEVICE);
+				dd->pageshadow[ctxttid + tid] = NULL;
+			}
+		}
+		qib_release_user_pages(pagep, cnt);
+	} else {
+		/*
+		 * Copy the updated array, with qib_tid's filled in, back
+		 * to user.  Since we did the copy in already, this "should
+		 * never fail" If it does, we have to clean up...
+		 */
+		if (copy_to_user((void __user *)
+				 (unsigned long) ti->tidlist,
+				 tidlist, cnt * sizeof(*tidlist))) {
+			ret = -EFAULT;
+			goto cleanup;
+		}
+		if (copy_to_user((void __user *) (unsigned long) ti->tidmap,
+				 tidmap, sizeof tidmap)) {
+			ret = -EFAULT;
+			goto cleanup;
+		}
+		if (tid == tidcnt)
+			tid = 0;
+		if (!rcd->subctxt_cnt)
+			rcd->tidcursor = tid;
+		else
+			tidcursor_fp(fp) = tid;
+	}
+
+done:
+	return ret;
+}
+
+/**
+ * qib_tid_free - free a context TID
+ * @rcd: the context
+ * @subctxt: the subcontext
+ * @ti: the TID info
+ *
+ * right now we are unlocking one page at a time, but since
+ * the intended use of this routine is for a single group of
+ * virtually contiguous pages, that should change to improve
+ * performance.  We check that the TID is in range for this context
+ * but otherwise don't check validity; if user has an error and
+ * frees the wrong tid, it's only their own data that can thereby
+ * be corrupted.  We do check that the TID was in use, for sanity
+ * We always use our idea of the saved address, not the address that
+ * they pass in to us.
+ */
+static int qib_tid_free(struct qib_ctxtdata *rcd, unsigned subctxt,
+			const struct qib_tid_info *ti)
+{
+	int ret = 0;
+	u32 tid, ctxttid, cnt, limit, tidcnt;
+	struct qib_devdata *dd = rcd->dd;
+	u64 __iomem *tidbase;
+	unsigned long tidmap[8];
+
+	if (!dd->pageshadow) {
+		ret = -ENOMEM;
+		goto done;
+	}
+
+	if (copy_from_user(tidmap, (void __user *)(unsigned long)ti->tidmap,
+			   sizeof tidmap)) {
+		ret = -EFAULT;
+		goto done;
+	}
+
+	ctxttid = rcd->ctxt * dd->rcvtidcnt;
+	if (!rcd->subctxt_cnt)
+		tidcnt = dd->rcvtidcnt;
+	else if (!subctxt) {
+		tidcnt = (dd->rcvtidcnt / rcd->subctxt_cnt) +
+			 (dd->rcvtidcnt % rcd->subctxt_cnt);
+		ctxttid += dd->rcvtidcnt - tidcnt;
+	} else {
+		tidcnt = dd->rcvtidcnt / rcd->subctxt_cnt;
+		ctxttid += tidcnt * (subctxt - 1);
+	}
+	tidbase = (u64 __iomem *) ((char __iomem *)(dd->kregbase) +
+				   dd->rcvtidbase +
+				   ctxttid * sizeof(*tidbase));
+
+	limit = sizeof(tidmap) * BITS_PER_BYTE;
+	if (limit > tidcnt)
+		/* just in case size changes in future */
+		limit = tidcnt;
+	tid = find_first_bit(tidmap, limit);
+	for (cnt = 0; tid < limit; tid++) {
+		/*
+		 * small optimization; if we detect a run of 3 or so without
+		 * any set, use find_first_bit again.  That's mainly to
+		 * accelerate the case where we wrapped, so we have some at
+		 * the beginning, and some at the end, and a big gap
+		 * in the middle.
+		 */
+		if (!test_bit(tid, tidmap))
+			continue;
+		cnt++;
+		if (dd->pageshadow[ctxttid + tid]) {
+			struct page *p;
+			dma_addr_t phys;
+
+			p = dd->pageshadow[ctxttid + tid];
+			dd->pageshadow[ctxttid + tid] = NULL;
+			phys = dd->physshadow[ctxttid + tid];
+			dd->physshadow[ctxttid + tid] = dd->tidinvalid;
+			/* PERFORMANCE: below should almost certainly be
+			 * cached
+			 */
+			dd->f_put_tid(dd, &tidbase[tid],
+				      RCVHQ_RCV_TYPE_EXPECTED, dd->tidinvalid);
+			pci_unmap_page(dd->pcidev, phys, PAGE_SIZE,
+				       PCI_DMA_FROMDEVICE);
+			qib_release_user_pages(&p, 1);
+		}
+	}
+done:
+	return ret;
+}
+
+/**
+ * qib_set_part_key - set a partition key
+ * @rcd: the context
+ * @key: the key
+ *
+ * We can have up to 4 active at a time (other than the default, which is
+ * always allowed).  This is somewhat tricky, since multiple contexts may set
+ * the same key, so we reference count them, and clean up at exit.  All 4
+ * partition keys are packed into a single qlogic_ib register.  It's an
+ * error for a process to set the same pkey multiple times.  We provide no
+ * mechanism to de-allocate a pkey at this time, we may eventually need to
+ * do that.  I've used the atomic operations, and no locking, and only make
+ * a single pass through what's available.  This should be more than
+ * adequate for some time. I'll think about spinlocks or the like if and as
+ * it's necessary.
+ */
+static int qib_set_part_key(struct qib_ctxtdata *rcd, u16 key)
+{
+	struct qib_pportdata *ppd = rcd->ppd;
+	int i, any = 0, pidx = -1;
+	u16 lkey = key & 0x7FFF;
+	int ret;
+
+	if (lkey == (QIB_DEFAULT_P_KEY & 0x7FFF)) {
+		/* nothing to do; this key always valid */
+		ret = 0;
+		goto bail;
+	}
+
+	if (!lkey) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	/*
+	 * Set the full membership bit, because it has to be
+	 * set in the register or the packet, and it seems
+	 * cleaner to set in the register than to force all
+	 * callers to set it.
+	 */
+	key |= 0x8000;
+
+	for (i = 0; i < ARRAY_SIZE(rcd->pkeys); i++) {
+		if (!rcd->pkeys[i] && pidx == -1)
+			pidx = i;
+		if (rcd->pkeys[i] == key) {
+			ret = -EEXIST;
+			goto bail;
+		}
+	}
+	if (pidx == -1) {
+		ret = -EBUSY;
+		goto bail;
+	}
+	for (any = i = 0; i < ARRAY_SIZE(ppd->pkeys); i++) {
+		if (!ppd->pkeys[i]) {
+			any++;
+			continue;
+		}
+		if (ppd->pkeys[i] == key) {
+			atomic_t *pkrefs = &ppd->pkeyrefs[i];
+
+			if (atomic_inc_return(pkrefs) > 1) {
+				rcd->pkeys[pidx] = key;
+				ret = 0;
+				goto bail;
+			} else {
+				/*
+				 * lost race, decrement count, catch below
+				 */
+				atomic_dec(pkrefs);
+				any++;
+			}
+		}
+		if ((ppd->pkeys[i] & 0x7FFF) == lkey) {
+			/*
+			 * It makes no sense to have both the limited and
+			 * full membership PKEY set at the same time since
+			 * the unlimited one will disable the limited one.
+			 */
+			ret = -EEXIST;
+			goto bail;
+		}
+	}
+	if (!any) {
+		ret = -EBUSY;
+		goto bail;
+	}
+	for (any = i = 0; i < ARRAY_SIZE(ppd->pkeys); i++) {
+		if (!ppd->pkeys[i] &&
+		    atomic_inc_return(&ppd->pkeyrefs[i]) == 1) {
+			rcd->pkeys[pidx] = key;
+			ppd->pkeys[i] = key;
+			(void) ppd->dd->f_set_ib_cfg(ppd, QIB_IB_CFG_PKEYS, 0);
+			ret = 0;
+			goto bail;
+		}
+	}
+	ret = -EBUSY;
+
+bail:
+	return ret;
+}
+
+/**
+ * qib_manage_rcvq - manage a context's receive queue
+ * @rcd: the context
+ * @subctxt: the subcontext
+ * @start_stop: action to carry out
+ *
+ * start_stop == 0 disables receive on the context, for use in queue
+ * overflow conditions.  start_stop==1 re-enables, to be used to
+ * re-init the software copy of the head register
+ */
+static int qib_manage_rcvq(struct qib_ctxtdata *rcd, unsigned subctxt,
+			   int start_stop)
+{
+	struct qib_devdata *dd = rcd->dd;
+	unsigned int rcvctrl_op;
+
+	if (subctxt)
+		goto bail;
+	/* atomically clear receive enable ctxt. */
+	if (start_stop) {
+		/*
+		 * On enable, force in-memory copy of the tail register to
+		 * 0, so that protocol code doesn't have to worry about
+		 * whether or not the chip has yet updated the in-memory
+		 * copy or not on return from the system call. The chip
+		 * always resets it's tail register back to 0 on a
+		 * transition from disabled to enabled.
+		 */
+		if (rcd->rcvhdrtail_kvaddr)
+			qib_clear_rcvhdrtail(rcd);
+		rcvctrl_op = QIB_RCVCTRL_CTXT_ENB;
+	} else
+		rcvctrl_op = QIB_RCVCTRL_CTXT_DIS;
+	dd->f_rcvctrl(rcd->ppd, rcvctrl_op, rcd->ctxt);
+	/* always; new head should be equal to new tail; see above */
+bail:
+	return 0;
+}
+
+static void qib_clean_part_key(struct qib_ctxtdata *rcd,
+			       struct qib_devdata *dd)
+{
+	int i, j, pchanged = 0;
+	u64 oldpkey;
+	struct qib_pportdata *ppd = rcd->ppd;
+
+	/* for debugging only */
+	oldpkey = (u64) ppd->pkeys[0] |
+		((u64) ppd->pkeys[1] << 16) |
+		((u64) ppd->pkeys[2] << 32) |
+		((u64) ppd->pkeys[3] << 48);
+
+	for (i = 0; i < ARRAY_SIZE(rcd->pkeys); i++) {
+		if (!rcd->pkeys[i])
+			continue;
+		for (j = 0; j < ARRAY_SIZE(ppd->pkeys); j++) {
+			/* check for match independent of the global bit */
+			if ((ppd->pkeys[j] & 0x7fff) !=
+			    (rcd->pkeys[i] & 0x7fff))
+				continue;
+			if (atomic_dec_and_test(&ppd->pkeyrefs[j])) {
+				ppd->pkeys[j] = 0;
+				pchanged++;
+			}
+			break;
+		}
+		rcd->pkeys[i] = 0;
+	}
+	if (pchanged)
+		(void) ppd->dd->f_set_ib_cfg(ppd, QIB_IB_CFG_PKEYS, 0);
+}
+
+/* common code for the mappings on dma_alloc_coherent mem */
+static int qib_mmap_mem(struct vm_area_struct *vma, struct qib_ctxtdata *rcd,
+			unsigned len, void *kvaddr, u32 write_ok, char *what)
+{
+	struct qib_devdata *dd = rcd->dd;
+	unsigned long pfn;
+	int ret;
+
+	if ((vma->vm_end - vma->vm_start) > len) {
+		qib_devinfo(dd->pcidev,
+			 "FAIL on %s: len %lx > %x\n", what,
+			 vma->vm_end - vma->vm_start, len);
+		ret = -EFAULT;
+		goto bail;
+	}
+
+	/*
+	 * shared context user code requires rcvhdrq mapped r/w, others
+	 * only allowed readonly mapping.
+	 */
+	if (!write_ok) {
+		if (vma->vm_flags & VM_WRITE) {
+			qib_devinfo(dd->pcidev,
+				 "%s must be mapped readonly\n", what);
+			ret = -EPERM;
+			goto bail;
+		}
+
+		/* don't allow them to later change with mprotect */
+		vma->vm_flags &= ~VM_MAYWRITE;
+	}
+
+	pfn = virt_to_phys(kvaddr) >> PAGE_SHIFT;
+	ret = remap_pfn_range(vma, vma->vm_start, pfn,
+			      len, vma->vm_page_prot);
+	if (ret)
+		qib_devinfo(dd->pcidev, "%s ctxt%u mmap of %lx, %x "
+			 "bytes failed: %d\n", what, rcd->ctxt,
+			 pfn, len, ret);
+bail:
+	return ret;
+}
+
+static int mmap_ureg(struct vm_area_struct *vma, struct qib_devdata *dd,
+		     u64 ureg)
+{
+	unsigned long phys;
+	unsigned long sz;
+	int ret;
+
+	/*
+	 * This is real hardware, so use io_remap.  This is the mechanism
+	 * for the user process to update the head registers for their ctxt
+	 * in the chip.
+	 */
+	sz = dd->flags & QIB_HAS_HDRSUPP ? 2 * PAGE_SIZE : PAGE_SIZE;
+	if ((vma->vm_end - vma->vm_start) > sz) {
+		qib_devinfo(dd->pcidev, "FAIL mmap userreg: reqlen "
+			 "%lx > PAGE\n", vma->vm_end - vma->vm_start);
+		ret = -EFAULT;
+	} else {
+		phys = dd->physaddr + ureg;
+		vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
+
+		vma->vm_flags |= VM_DONTCOPY | VM_DONTEXPAND;
+		ret = io_remap_pfn_range(vma, vma->vm_start,
+					 phys >> PAGE_SHIFT,
+					 vma->vm_end - vma->vm_start,
+					 vma->vm_page_prot);
+	}
+	return ret;
+}
+
+static int mmap_piobufs(struct vm_area_struct *vma,
+			struct qib_devdata *dd,
+			struct qib_ctxtdata *rcd,
+			unsigned piobufs, unsigned piocnt)
+{
+	unsigned long phys;
+	int ret;
+
+	/*
+	 * When we map the PIO buffers in the chip, we want to map them as
+	 * writeonly, no read possible; unfortunately, x86 doesn't allow
+	 * for this in hardware, but we still prevent users from asking
+	 * for it.
+	 */
+	if ((vma->vm_end - vma->vm_start) > (piocnt * dd->palign)) {
+		qib_devinfo(dd->pcidev, "FAIL mmap piobufs: "
+			 "reqlen %lx > PAGE\n",
+			 vma->vm_end - vma->vm_start);
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	phys = dd->physaddr + piobufs;
+
+#if defined(__powerpc__)
+	/* There isn't a generic way to specify writethrough mappings */
+	pgprot_val(vma->vm_page_prot) |= _PAGE_NO_CACHE;
+	pgprot_val(vma->vm_page_prot) |= _PAGE_WRITETHRU;
+	pgprot_val(vma->vm_page_prot) &= ~_PAGE_GUARDED;
+#endif
+
+	/*
+	 * don't allow them to later change to readable with mprotect (for when
+	 * not initially mapped readable, as is normally the case)
+	 */
+	vma->vm_flags &= ~VM_MAYREAD;
+	vma->vm_flags |= VM_DONTCOPY | VM_DONTEXPAND;
+
+	if (qib_wc_pat)
+		vma->vm_page_prot = pgprot_writecombine(vma->vm_page_prot);
+
+	ret = io_remap_pfn_range(vma, vma->vm_start, phys >> PAGE_SHIFT,
+				 vma->vm_end - vma->vm_start,
+				 vma->vm_page_prot);
+bail:
+	return ret;
+}
+
+static int mmap_rcvegrbufs(struct vm_area_struct *vma,
+			   struct qib_ctxtdata *rcd)
+{
+	struct qib_devdata *dd = rcd->dd;
+	unsigned long start, size;
+	size_t total_size, i;
+	unsigned long pfn;
+	int ret;
+
+	size = rcd->rcvegrbuf_size;
+	total_size = rcd->rcvegrbuf_chunks * size;
+	if ((vma->vm_end - vma->vm_start) > total_size) {
+		qib_devinfo(dd->pcidev, "FAIL on egr bufs: "
+			 "reqlen %lx > actual %lx\n",
+			 vma->vm_end - vma->vm_start,
+			 (unsigned long) total_size);
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	if (vma->vm_flags & VM_WRITE) {
+		qib_devinfo(dd->pcidev, "Can't map eager buffers as "
+			 "writable (flags=%lx)\n", vma->vm_flags);
+		ret = -EPERM;
+		goto bail;
+	}
+	/* don't allow them to later change to writeable with mprotect */
+	vma->vm_flags &= ~VM_MAYWRITE;
+
+	start = vma->vm_start;
+
+	for (i = 0; i < rcd->rcvegrbuf_chunks; i++, start += size) {
+		pfn = virt_to_phys(rcd->rcvegrbuf[i]) >> PAGE_SHIFT;
+		ret = remap_pfn_range(vma, start, pfn, size,
+				      vma->vm_page_prot);
+		if (ret < 0)
+			goto bail;
+	}
+	ret = 0;
+
+bail:
+	return ret;
+}
+
+/*
+ * qib_file_vma_fault - handle a VMA page fault.
+ */
+static int qib_file_vma_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
+{
+	struct page *page;
+
+	page = vmalloc_to_page((void *)(vmf->pgoff << PAGE_SHIFT));
+	if (!page)
+		return VM_FAULT_SIGBUS;
+
+	get_page(page);
+	vmf->page = page;
+
+	return 0;
+}
+
+static struct vm_operations_struct qib_file_vm_ops = {
+	.fault = qib_file_vma_fault,
+};
+
+static int mmap_kvaddr(struct vm_area_struct *vma, u64 pgaddr,
+		       struct qib_ctxtdata *rcd, unsigned subctxt)
+{
+	struct qib_devdata *dd = rcd->dd;
+	unsigned subctxt_cnt;
+	unsigned long len;
+	void *addr;
+	size_t size;
+	int ret = 0;
+
+	subctxt_cnt = rcd->subctxt_cnt;
+	size = rcd->rcvegrbuf_chunks * rcd->rcvegrbuf_size;
+
+	/*
+	 * Each process has all the subctxt uregbase, rcvhdrq, and
+	 * rcvegrbufs mmapped - as an array for all the processes,
+	 * and also separately for this process.
+	 */
+	if (pgaddr == cvt_kvaddr(rcd->subctxt_uregbase)) {
+		addr = rcd->subctxt_uregbase;
+		size = PAGE_SIZE * subctxt_cnt;
+	} else if (pgaddr == cvt_kvaddr(rcd->subctxt_rcvhdr_base)) {
+		addr = rcd->subctxt_rcvhdr_base;
+		size = rcd->rcvhdrq_size * subctxt_cnt;
+	} else if (pgaddr == cvt_kvaddr(rcd->subctxt_rcvegrbuf)) {
+		addr = rcd->subctxt_rcvegrbuf;
+		size *= subctxt_cnt;
+	} else if (pgaddr == cvt_kvaddr(rcd->subctxt_uregbase +
+					PAGE_SIZE * subctxt)) {
+		addr = rcd->subctxt_uregbase + PAGE_SIZE * subctxt;
+		size = PAGE_SIZE;
+	} else if (pgaddr == cvt_kvaddr(rcd->subctxt_rcvhdr_base +
+					rcd->rcvhdrq_size * subctxt)) {
+		addr = rcd->subctxt_rcvhdr_base +
+			rcd->rcvhdrq_size * subctxt;
+		size = rcd->rcvhdrq_size;
+	} else if (pgaddr == cvt_kvaddr(&rcd->user_event_mask[subctxt])) {
+		addr = rcd->user_event_mask;
+		size = PAGE_SIZE;
+	} else if (pgaddr == cvt_kvaddr(rcd->subctxt_rcvegrbuf +
+					size * subctxt)) {
+		addr = rcd->subctxt_rcvegrbuf + size * subctxt;
+		/* rcvegrbufs are read-only on the slave */
+		if (vma->vm_flags & VM_WRITE) {
+			qib_devinfo(dd->pcidev,
+				 "Can't map eager buffers as "
+				 "writable (flags=%lx)\n", vma->vm_flags);
+			ret = -EPERM;
+			goto bail;
+		}
+		/*
+		 * Don't allow permission to later change to writeable
+		 * with mprotect.
+		 */
+		vma->vm_flags &= ~VM_MAYWRITE;
+	} else
+		goto bail;
+	len = vma->vm_end - vma->vm_start;
+	if (len > size) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	vma->vm_pgoff = (unsigned long) addr >> PAGE_SHIFT;
+	vma->vm_ops = &qib_file_vm_ops;
+	vma->vm_flags |= VM_RESERVED | VM_DONTEXPAND;
+	ret = 1;
+
+bail:
+	return ret;
+}
+
+/**
+ * qib_mmapf - mmap various structures into user space
+ * @fp: the file pointer
+ * @vma: the VM area
+ *
+ * We use this to have a shared buffer between the kernel and the user code
+ * for the rcvhdr queue, egr buffers, and the per-context user regs and pio
+ * buffers in the chip.  We have the open and close entries so we can bump
+ * the ref count and keep the driver from being unloaded while still mapped.
+ */
+static int qib_mmapf(struct file *fp, struct vm_area_struct *vma)
+{
+	struct qib_ctxtdata *rcd;
+	struct qib_devdata *dd;
+	u64 pgaddr, ureg;
+	unsigned piobufs, piocnt;
+	int ret, match = 1;
+
+	rcd = ctxt_fp(fp);
+	if (!rcd || !(vma->vm_flags & VM_SHARED)) {
+		ret = -EINVAL;
+		goto bail;
+	}
+	dd = rcd->dd;
+
+	/*
+	 * This is the qib_do_user_init() code, mapping the shared buffers
+	 * and per-context user registers into the user process. The address
+	 * referred to by vm_pgoff is the file offset passed via mmap().
+	 * For shared contexts, this is the kernel vmalloc() address of the
+	 * pages to share with the master.
+	 * For non-shared or master ctxts, this is a physical address.
+	 * We only do one mmap for each space mapped.
+	 */
+	pgaddr = vma->vm_pgoff << PAGE_SHIFT;
+
+	/*
+	 * Check for 0 in case one of the allocations failed, but user
+	 * called mmap anyway.
+	 */
+	if (!pgaddr)  {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	/*
+	 * Physical addresses must fit in 40 bits for our hardware.
+	 * Check for kernel virtual addresses first, anything else must
+	 * match a HW or memory address.
+	 */
+	ret = mmap_kvaddr(vma, pgaddr, rcd, subctxt_fp(fp));
+	if (ret) {
+		if (ret > 0)
+			ret = 0;
+		goto bail;
+	}
+
+	ureg = dd->uregbase + dd->ureg_align * rcd->ctxt;
+	if (!rcd->subctxt_cnt) {
+		/* ctxt is not shared */
+		piocnt = rcd->piocnt;
+		piobufs = rcd->piobufs;
+	} else if (!subctxt_fp(fp)) {
+		/* caller is the master */
+		piocnt = (rcd->piocnt / rcd->subctxt_cnt) +
+			 (rcd->piocnt % rcd->subctxt_cnt);
+		piobufs = rcd->piobufs +
+			dd->palign * (rcd->piocnt - piocnt);
+	} else {
+		unsigned slave = subctxt_fp(fp) - 1;
+
+		/* caller is a slave */
+		piocnt = rcd->piocnt / rcd->subctxt_cnt;
+		piobufs = rcd->piobufs + dd->palign * piocnt * slave;
+	}
+
+	if (pgaddr == ureg)
+		ret = mmap_ureg(vma, dd, ureg);
+	else if (pgaddr == piobufs)
+		ret = mmap_piobufs(vma, dd, rcd, piobufs, piocnt);
+	else if (pgaddr == dd->pioavailregs_phys)
+		/* in-memory copy of pioavail registers */
+		ret = qib_mmap_mem(vma, rcd, PAGE_SIZE,
+				   (void *) dd->pioavailregs_dma, 0,
+				   "pioavail registers");
+	else if (pgaddr == rcd->rcvegr_phys)
+		ret = mmap_rcvegrbufs(vma, rcd);
+	else if (pgaddr == (u64) rcd->rcvhdrq_phys)
+		/*
+		 * The rcvhdrq itself; multiple pages, contiguous
+		 * from an i/o perspective.  Shared contexts need
+		 * to map r/w, so we allow writing.
+		 */
+		ret = qib_mmap_mem(vma, rcd, rcd->rcvhdrq_size,
+				   rcd->rcvhdrq, 1, "rcvhdrq");
+	else if (pgaddr == (u64) rcd->rcvhdrqtailaddr_phys)
+		/* in-memory copy of rcvhdrq tail register */
+		ret = qib_mmap_mem(vma, rcd, PAGE_SIZE,
+				   rcd->rcvhdrtail_kvaddr, 0,
+				   "rcvhdrq tail");
+	else
+		match = 0;
+	if (!match)
+		ret = -EINVAL;
+
+	vma->vm_private_data = NULL;
+
+	if (ret < 0)
+		qib_devinfo(dd->pcidev,
+			 "mmap Failure %d: off %llx len %lx\n",
+			 -ret, (unsigned long long)pgaddr,
+			 vma->vm_end - vma->vm_start);
+bail:
+	return ret;
+}
+
+static unsigned int qib_poll_urgent(struct qib_ctxtdata *rcd,
+				    struct file *fp,
+				    struct poll_table_struct *pt)
+{
+	struct qib_devdata *dd = rcd->dd;
+	unsigned pollflag;
+
+	poll_wait(fp, &rcd->wait, pt);
+
+	spin_lock_irq(&dd->uctxt_lock);
+	if (rcd->urgent != rcd->urgent_poll) {
+		pollflag = POLLIN | POLLRDNORM;
+		rcd->urgent_poll = rcd->urgent;
+	} else {
+		pollflag = 0;
+		set_bit(QIB_CTXT_WAITING_URG, &rcd->flag);
+	}
+	spin_unlock_irq(&dd->uctxt_lock);
+
+	return pollflag;
+}
+
+static unsigned int qib_poll_next(struct qib_ctxtdata *rcd,
+				  struct file *fp,
+				  struct poll_table_struct *pt)
+{
+	struct qib_devdata *dd = rcd->dd;
+	unsigned pollflag;
+
+	poll_wait(fp, &rcd->wait, pt);
+
+	spin_lock_irq(&dd->uctxt_lock);
+	if (dd->f_hdrqempty(rcd)) {
+		set_bit(QIB_CTXT_WAITING_RCV, &rcd->flag);
+		dd->f_rcvctrl(rcd->ppd, QIB_RCVCTRL_INTRAVAIL_ENB, rcd->ctxt);
+		pollflag = 0;
+	} else
+		pollflag = POLLIN | POLLRDNORM;
+	spin_unlock_irq(&dd->uctxt_lock);
+
+	return pollflag;
+}
+
+static unsigned int qib_poll(struct file *fp, struct poll_table_struct *pt)
+{
+	struct qib_ctxtdata *rcd;
+	unsigned pollflag;
+
+	rcd = ctxt_fp(fp);
+	if (!rcd)
+		pollflag = POLLERR;
+	else if (rcd->poll_type == QIB_POLL_TYPE_URGENT)
+		pollflag = qib_poll_urgent(rcd, fp, pt);
+	else  if (rcd->poll_type == QIB_POLL_TYPE_ANYRCV)
+		pollflag = qib_poll_next(rcd, fp, pt);
+	else /* invalid */
+		pollflag = POLLERR;
+
+	return pollflag;
+}
+
+/*
+ * Check that userland and driver are compatible for subcontexts.
+ */
+static int qib_compatible_subctxts(int user_swmajor, int user_swminor)
+{
+	/* this code is written long-hand for clarity */
+	if (QIB_USER_SWMAJOR != user_swmajor) {
+		/* no promise of compatibility if major mismatch */
+		return 0;
+	}
+	if (QIB_USER_SWMAJOR == 1) {
+		switch (QIB_USER_SWMINOR) {
+		case 0:
+		case 1:
+		case 2:
+			/* no subctxt implementation so cannot be compatible */
+			return 0;
+		case 3:
+			/* 3 is only compatible with itself */
+			return user_swminor == 3;
+		default:
+			/* >= 4 are compatible (or are expected to be) */
+			return user_swminor >= 4;
+		}
+	}
+	/* make no promises yet for future major versions */
+	return 0;
+}
+
+static int init_subctxts(struct qib_devdata *dd,
+			 struct qib_ctxtdata *rcd,
+			 const struct qib_user_info *uinfo)
+{
+	int ret = 0;
+	unsigned num_subctxts;
+	size_t size;
+
+	/*
+	 * If the user is requesting zero subctxts,
+	 * skip the subctxt allocation.
+	 */
+	if (uinfo->spu_subctxt_cnt <= 0)
+		goto bail;
+	num_subctxts = uinfo->spu_subctxt_cnt;
+
+	/* Check for subctxt compatibility */
+	if (!qib_compatible_subctxts(uinfo->spu_userversion >> 16,
+		uinfo->spu_userversion & 0xffff)) {
+		qib_devinfo(dd->pcidev,
+			 "Mismatched user version (%d.%d) and driver "
+			 "version (%d.%d) while context sharing. Ensure "
+			 "that driver and library are from the same "
+			 "release.\n",
+			 (int) (uinfo->spu_userversion >> 16),
+			 (int) (uinfo->spu_userversion & 0xffff),
+			 QIB_USER_SWMAJOR, QIB_USER_SWMINOR);
+		goto bail;
+	}
+	if (num_subctxts > QLOGIC_IB_MAX_SUBCTXT) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	rcd->subctxt_uregbase = vmalloc_user(PAGE_SIZE * num_subctxts);
+	if (!rcd->subctxt_uregbase) {
+		ret = -ENOMEM;
+		goto bail;
+	}
+	/* Note: rcd->rcvhdrq_size isn't initialized yet. */
+	size = ALIGN(dd->rcvhdrcnt * dd->rcvhdrentsize *
+		     sizeof(u32), PAGE_SIZE) * num_subctxts;
+	rcd->subctxt_rcvhdr_base = vmalloc_user(size);
+	if (!rcd->subctxt_rcvhdr_base) {
+		ret = -ENOMEM;
+		goto bail_ureg;
+	}
+
+	rcd->subctxt_rcvegrbuf = vmalloc_user(rcd->rcvegrbuf_chunks *
+					      rcd->rcvegrbuf_size *
+					      num_subctxts);
+	if (!rcd->subctxt_rcvegrbuf) {
+		ret = -ENOMEM;
+		goto bail_rhdr;
+	}
+
+	rcd->subctxt_cnt = uinfo->spu_subctxt_cnt;
+	rcd->subctxt_id = uinfo->spu_subctxt_id;
+	rcd->active_slaves = 1;
+	rcd->redirect_seq_cnt = 1;
+	set_bit(QIB_CTXT_MASTER_UNINIT, &rcd->flag);
+	goto bail;
+
+bail_rhdr:
+	vfree(rcd->subctxt_rcvhdr_base);
+bail_ureg:
+	vfree(rcd->subctxt_uregbase);
+	rcd->subctxt_uregbase = NULL;
+bail:
+	return ret;
+}
+
+static int setup_ctxt(struct qib_pportdata *ppd, int ctxt,
+		      struct file *fp, const struct qib_user_info *uinfo)
+{
+	struct qib_devdata *dd = ppd->dd;
+	struct qib_ctxtdata *rcd;
+	void *ptmp = NULL;
+	int ret;
+
+	rcd = qib_create_ctxtdata(ppd, ctxt);
+
+	/*
+	 * Allocate memory for use in qib_tid_update() at open to
+	 * reduce cost of expected send setup per message segment
+	 */
+	if (rcd)
+		ptmp = kmalloc(dd->rcvtidcnt * sizeof(u16) +
+			       dd->rcvtidcnt * sizeof(struct page **),
+			       GFP_KERNEL);
+
+	if (!rcd || !ptmp) {
+		qib_dev_err(dd, "Unable to allocate ctxtdata "
+			    "memory, failing open\n");
+		ret = -ENOMEM;
+		goto bailerr;
+	}
+	rcd->userversion = uinfo->spu_userversion;
+	ret = init_subctxts(dd, rcd, uinfo);
+	if (ret)
+		goto bailerr;
+	rcd->tid_pg_list = ptmp;
+	rcd->pid = current->pid;
+	init_waitqueue_head(&dd->rcd[ctxt]->wait);
+	strlcpy(rcd->comm, current->comm, sizeof(rcd->comm));
+	ctxt_fp(fp) = rcd;
+	qib_stats.sps_ctxts++;
+	ret = 0;
+	goto bail;
+
+bailerr:
+	dd->rcd[ctxt] = NULL;
+	kfree(rcd);
+	kfree(ptmp);
+bail:
+	return ret;
+}
+
+static inline int usable(struct qib_pportdata *ppd, int active_only)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u32 linkok = active_only ? QIBL_LINKACTIVE :
+		 (QIBL_LINKINIT | QIBL_LINKARMED | QIBL_LINKACTIVE);
+
+	return dd && (dd->flags & QIB_PRESENT) && dd->kregbase && ppd->lid &&
+		(ppd->lflags & linkok);
+}
+
+static int find_free_ctxt(int unit, struct file *fp,
+			  const struct qib_user_info *uinfo)
+{
+	struct qib_devdata *dd = qib_lookup(unit);
+	struct qib_pportdata *ppd = NULL;
+	int ret;
+	u32 ctxt;
+
+	if (!dd || (uinfo->spu_port && uinfo->spu_port > dd->num_pports)) {
+		ret = -ENODEV;
+		goto bail;
+	}
+
+	/*
+	 * If users requests specific port, only try that one port, else
+	 * select "best" port below, based on context.
+	 */
+	if (uinfo->spu_port) {
+		ppd = dd->pport + uinfo->spu_port - 1;
+		if (!usable(ppd, 0)) {
+			ret = -ENETDOWN;
+			goto bail;
+		}
+	}
+
+	for (ctxt = dd->first_user_ctxt; ctxt < dd->cfgctxts; ctxt++) {
+		if (dd->rcd[ctxt])
+			continue;
+		/*
+		 * The setting and clearing of user context rcd[x] protected
+		 * by the qib_mutex
+		 */
+		if (!ppd) {
+			/* choose port based on ctxt, if up, else 1st up */
+			ppd = dd->pport + (ctxt % dd->num_pports);
+			if (!usable(ppd, 0)) {
+				int i;
+				for (i = 0; i < dd->num_pports; i++) {
+					ppd = dd->pport + i;
+					if (usable(ppd, 0))
+						break;
+				}
+				if (i == dd->num_pports) {
+					ret = -ENETDOWN;
+					goto bail;
+				}
+			}
+		}
+		ret = setup_ctxt(ppd, ctxt, fp, uinfo);
+		goto bail;
+	}
+	ret = -EBUSY;
+
+bail:
+	return ret;
+}
+
+static int get_a_ctxt(struct file *fp, const struct qib_user_info *uinfo)
+{
+	struct qib_pportdata *ppd;
+	int ret = 0, devmax;
+	int npresent, nup;
+	int ndev;
+	u32 port = uinfo->spu_port, ctxt;
+
+	devmax = qib_count_units(&npresent, &nup);
+
+	for (ndev = 0; ndev < devmax; ndev++) {
+		struct qib_devdata *dd = qib_lookup(ndev);
+
+		/* device portion of usable() */
+		if (!(dd && (dd->flags & QIB_PRESENT) && dd->kregbase))
+			continue;
+		for (ctxt = dd->first_user_ctxt; ctxt < dd->cfgctxts; ctxt++) {
+			if (dd->rcd[ctxt])
+				continue;
+			if (port) {
+				if (port > dd->num_pports)
+					continue;
+				ppd = dd->pport + port - 1;
+				if (!usable(ppd, 0))
+					continue;
+			} else {
+				/*
+				 * choose port based on ctxt, if up, else
+				 * first port that's up for multi-port HCA
+				 */
+				ppd = dd->pport + (ctxt % dd->num_pports);
+				if (!usable(ppd, 0)) {
+					int j;
+
+					ppd = NULL;
+					for (j = 0; j < dd->num_pports &&
+						!ppd; j++)
+						if (usable(dd->pport + j, 0))
+							ppd = dd->pport + j;
+					if (!ppd)
+						continue; /* to next unit */
+				}
+			}
+			ret = setup_ctxt(ppd, ctxt, fp, uinfo);
+			goto done;
+		}
+	}
+
+	if (npresent) {
+		if (nup == 0)
+			ret = -ENETDOWN;
+		else
+			ret = -EBUSY;
+	} else
+		ret = -ENXIO;
+
+done:
+	return ret;
+}
+
+static int find_shared_ctxt(struct file *fp,
+			    const struct qib_user_info *uinfo)
+{
+	int devmax, ndev, i;
+	int ret = 0;
+
+	devmax = qib_count_units(NULL, NULL);
+
+	for (ndev = 0; ndev < devmax; ndev++) {
+		struct qib_devdata *dd = qib_lookup(ndev);
+
+		/* device portion of usable() */
+		if (!(dd && (dd->flags & QIB_PRESENT) && dd->kregbase))
+			continue;
+		for (i = dd->first_user_ctxt; i < dd->cfgctxts; i++) {
+			struct qib_ctxtdata *rcd = dd->rcd[i];
+
+			/* Skip ctxts which are not yet open */
+			if (!rcd || !rcd->cnt)
+				continue;
+			/* Skip ctxt if it doesn't match the requested one */
+			if (rcd->subctxt_id != uinfo->spu_subctxt_id)
+				continue;
+			/* Verify the sharing process matches the master */
+			if (rcd->subctxt_cnt != uinfo->spu_subctxt_cnt ||
+			    rcd->userversion != uinfo->spu_userversion ||
+			    rcd->cnt >= rcd->subctxt_cnt) {
+				ret = -EINVAL;
+				goto done;
+			}
+			ctxt_fp(fp) = rcd;
+			subctxt_fp(fp) = rcd->cnt++;
+			rcd->subpid[subctxt_fp(fp)] = current->pid;
+			tidcursor_fp(fp) = 0;
+			rcd->active_slaves |= 1 << subctxt_fp(fp);
+			ret = 1;
+			goto done;
+		}
+	}
+
+done:
+	return ret;
+}
+
+static int qib_open(struct inode *in, struct file *fp)
+{
+	/* The real work is performed later in qib_assign_ctxt() */
+	fp->private_data = kzalloc(sizeof(struct qib_filedata), GFP_KERNEL);
+	if (fp->private_data) /* no cpu affinity by default */
+		((struct qib_filedata *)fp->private_data)->rec_cpu_num = -1;
+	return fp->private_data ? 0 : -ENOMEM;
+}
+
+/*
+ * Get ctxt early, so can set affinity prior to memory allocation.
+ */
+static int qib_assign_ctxt(struct file *fp, const struct qib_user_info *uinfo)
+{
+	int ret;
+	int i_minor;
+	unsigned swmajor, swminor;
+
+	/* Check to be sure we haven't already initialized this file */
+	if (ctxt_fp(fp)) {
+		ret = -EINVAL;
+		goto done;
+	}
+
+	/* for now, if major version is different, bail */
+	swmajor = uinfo->spu_userversion >> 16;
+	if (swmajor != QIB_USER_SWMAJOR) {
+		ret = -ENODEV;
+		goto done;
+	}
+
+	swminor = uinfo->spu_userversion & 0xffff;
+
+	mutex_lock(&qib_mutex);
+
+	if (qib_compatible_subctxts(swmajor, swminor) &&
+	    uinfo->spu_subctxt_cnt) {
+		ret = find_shared_ctxt(fp, uinfo);
+		if (ret) {
+			if (ret > 0)
+				ret = 0;
+			goto done_chk_sdma;
+		}
+	}
+
+	i_minor = iminor(fp->f_dentry->d_inode) - QIB_USER_MINOR_BASE;
+	if (i_minor)
+		ret = find_free_ctxt(i_minor - 1, fp, uinfo);
+	else
+		ret = get_a_ctxt(fp, uinfo);
+
+done_chk_sdma:
+	if (!ret) {
+		struct qib_filedata *fd = fp->private_data;
+		const struct qib_ctxtdata *rcd = fd->rcd;
+		const struct qib_devdata *dd = rcd->dd;
+
+		if (dd->flags & QIB_HAS_SEND_DMA) {
+			fd->pq = qib_user_sdma_queue_create(&dd->pcidev->dev,
+							    dd->unit,
+							    rcd->ctxt,
+							    fd->subctxt);
+			if (!fd->pq)
+				ret = -ENOMEM;
+		}
+
+		/*
+		 * If process has NOT already set it's affinity, select and
+		 * reserve a processor for it, as a rendevous for all
+		 * users of the driver.  If they don't actually later
+		 * set affinity to this cpu, or set it to some other cpu,
+		 * it just means that sooner or later we don't recommend
+		 * a cpu, and let the scheduler do it's best.
+		 */
+		if (!ret && cpus_weight(current->cpus_allowed) >=
+		    qib_cpulist_count) {
+			int cpu;
+			cpu = find_first_zero_bit(qib_cpulist,
+						  qib_cpulist_count);
+			if (cpu != qib_cpulist_count) {
+				__set_bit(cpu, qib_cpulist);
+				fd->rec_cpu_num = cpu;
+			}
+		} else if (cpus_weight(current->cpus_allowed) == 1 &&
+			test_bit(first_cpu(current->cpus_allowed),
+				 qib_cpulist))
+			qib_devinfo(dd->pcidev, "%s PID %u affinity "
+				    "set to cpu %d; already allocated\n",
+				    current->comm, current->pid,
+				    first_cpu(current->cpus_allowed));
+	}
+
+	mutex_unlock(&qib_mutex);
+
+done:
+	return ret;
+}
+
+
+static int qib_do_user_init(struct file *fp,
+			    const struct qib_user_info *uinfo)
+{
+	int ret;
+	struct qib_ctxtdata *rcd = ctxt_fp(fp);
+	struct qib_devdata *dd;
+	unsigned uctxt;
+
+	/* Subctxts don't need to initialize anything since master did it. */
+	if (subctxt_fp(fp)) {
+		ret = wait_event_interruptible(rcd->wait,
+			!test_bit(QIB_CTXT_MASTER_UNINIT, &rcd->flag));
+		goto bail;
+	}
+
+	dd = rcd->dd;
+
+	/* some ctxts may get extra buffers, calculate that here */
+	uctxt = rcd->ctxt - dd->first_user_ctxt;
+	if (uctxt < dd->ctxts_extrabuf) {
+		rcd->piocnt = dd->pbufsctxt + 1;
+		rcd->pio_base = rcd->piocnt * uctxt;
+	} else {
+		rcd->piocnt = dd->pbufsctxt;
+		rcd->pio_base = rcd->piocnt * uctxt +
+			dd->ctxts_extrabuf;
+	}
+
+	/*
+	 * All user buffers are 2KB buffers.  If we ever support
+	 * giving 4KB buffers to user processes, this will need some
+	 * work.  Can't use piobufbase directly, because it has
+	 * both 2K and 4K buffer base values.  So check and handle.
+	 */
+	if ((rcd->pio_base + rcd->piocnt) > dd->piobcnt2k) {
+		if (rcd->pio_base >= dd->piobcnt2k) {
+			qib_dev_err(dd,
+				    "%u:ctxt%u: no 2KB buffers available\n",
+				    dd->unit, rcd->ctxt);
+			ret = -ENOBUFS;
+			goto bail;
+		}
+		rcd->piocnt = dd->piobcnt2k - rcd->pio_base;
+		qib_dev_err(dd, "Ctxt%u: would use 4KB bufs, using %u\n",
+			    rcd->ctxt, rcd->piocnt);
+	}
+
+	rcd->piobufs = dd->pio2k_bufbase + rcd->pio_base * dd->palign;
+	qib_chg_pioavailkernel(dd, rcd->pio_base, rcd->piocnt,
+			       TXCHK_CHG_TYPE_USER, rcd);
+	/*
+	 * try to ensure that processes start up with consistent avail update
+	 * for their own range, at least.   If system very quiet, it might
+	 * have the in-memory copy out of date at startup for this range of
+	 * buffers, when a context gets re-used.  Do after the chg_pioavail
+	 * and before the rest of setup, so it's "almost certain" the dma
+	 * will have occurred (can't 100% guarantee, but should be many
+	 * decimals of 9s, with this ordering), given how much else happens
+	 * after this.
+	 */
+	dd->f_sendctrl(dd->pport, QIB_SENDCTRL_AVAIL_BLIP);
+
+	/*
+	 * Now allocate the rcvhdr Q and eager TIDs; skip the TID
+	 * array for time being.  If rcd->ctxt > chip-supported,
+	 * we need to do extra stuff here to handle by handling overflow
+	 * through ctxt 0, someday
+	 */
+	ret = qib_create_rcvhdrq(dd, rcd);
+	if (!ret)
+		ret = qib_setup_eagerbufs(rcd);
+	if (ret)
+		goto bail_pio;
+
+	rcd->tidcursor = 0; /* start at beginning after open */
+
+	/* initialize poll variables... */
+	rcd->urgent = 0;
+	rcd->urgent_poll = 0;
+
+	/*
+	 * Now enable the ctxt for receive.
+	 * For chips that are set to DMA the tail register to memory
+	 * when they change (and when the update bit transitions from
+	 * 0 to 1.  So for those chips, we turn it off and then back on.
+	 * This will (very briefly) affect any other open ctxts, but the
+	 * duration is very short, and therefore isn't an issue.  We
+	 * explictly set the in-memory tail copy to 0 beforehand, so we
+	 * don't have to wait to be sure the DMA update has happened
+	 * (chip resets head/tail to 0 on transition to enable).
+	 */
+	if (rcd->rcvhdrtail_kvaddr)
+		qib_clear_rcvhdrtail(rcd);
+
+	dd->f_rcvctrl(rcd->ppd, QIB_RCVCTRL_CTXT_ENB | QIB_RCVCTRL_TIDFLOW_ENB,
+		      rcd->ctxt);
+
+	/* Notify any waiting slaves */
+	if (rcd->subctxt_cnt) {
+		clear_bit(QIB_CTXT_MASTER_UNINIT, &rcd->flag);
+		wake_up(&rcd->wait);
+	}
+	return 0;
+
+bail_pio:
+	qib_chg_pioavailkernel(dd, rcd->pio_base, rcd->piocnt,
+			       TXCHK_CHG_TYPE_KERN, rcd);
+bail:
+	return ret;
+}
+
+/**
+ * unlock_exptid - unlock any expected TID entries context still had in use
+ * @rcd: ctxt
+ *
+ * We don't actually update the chip here, because we do a bulk update
+ * below, using f_clear_tids.
+ */
+static void unlock_expected_tids(struct qib_ctxtdata *rcd)
+{
+	struct qib_devdata *dd = rcd->dd;
+	int ctxt_tidbase = rcd->ctxt * dd->rcvtidcnt;
+	int i, cnt = 0, maxtid = ctxt_tidbase + dd->rcvtidcnt;
+
+	for (i = ctxt_tidbase; i < maxtid; i++) {
+		struct page *p = dd->pageshadow[i];
+		dma_addr_t phys;
+
+		if (!p)
+			continue;
+
+		phys = dd->physshadow[i];
+		dd->physshadow[i] = dd->tidinvalid;
+		dd->pageshadow[i] = NULL;
+		pci_unmap_page(dd->pcidev, phys, PAGE_SIZE,
+			       PCI_DMA_FROMDEVICE);
+		qib_release_user_pages(&p, 1);
+		cnt++;
+	}
+}
+
+static int qib_close(struct inode *in, struct file *fp)
+{
+	int ret = 0;
+	struct qib_filedata *fd;
+	struct qib_ctxtdata *rcd;
+	struct qib_devdata *dd;
+	unsigned long flags;
+	unsigned ctxt;
+	pid_t pid;
+
+	mutex_lock(&qib_mutex);
+
+	fd = (struct qib_filedata *) fp->private_data;
+	fp->private_data = NULL;
+	rcd = fd->rcd;
+	if (!rcd) {
+		mutex_unlock(&qib_mutex);
+		goto bail;
+	}
+
+	dd = rcd->dd;
+
+	/* ensure all pio buffer writes in progress are flushed */
+	qib_flush_wc();
+
+	/* drain user sdma queue */
+	if (fd->pq) {
+		qib_user_sdma_queue_drain(rcd->ppd, fd->pq);
+		qib_user_sdma_queue_destroy(fd->pq);
+	}
+
+	if (fd->rec_cpu_num != -1)
+		__clear_bit(fd->rec_cpu_num, qib_cpulist);
+
+	if (--rcd->cnt) {
+		/*
+		 * XXX If the master closes the context before the slave(s),
+		 * revoke the mmap for the eager receive queue so
+		 * the slave(s) don't wait for receive data forever.
+		 */
+		rcd->active_slaves &= ~(1 << fd->subctxt);
+		rcd->subpid[fd->subctxt] = 0;
+		mutex_unlock(&qib_mutex);
+		goto bail;
+	}
+
+	/* early; no interrupt users after this */
+	spin_lock_irqsave(&dd->uctxt_lock, flags);
+	ctxt = rcd->ctxt;
+	dd->rcd[ctxt] = NULL;
+	pid = rcd->pid;
+	rcd->pid = 0;
+	spin_unlock_irqrestore(&dd->uctxt_lock, flags);
+
+	if (rcd->rcvwait_to || rcd->piowait_to ||
+	    rcd->rcvnowait || rcd->pionowait) {
+		rcd->rcvwait_to = 0;
+		rcd->piowait_to = 0;
+		rcd->rcvnowait = 0;
+		rcd->pionowait = 0;
+	}
+	if (rcd->flag)
+		rcd->flag = 0;
+
+	if (dd->kregbase) {
+		/* atomically clear receive enable ctxt and intr avail. */
+		dd->f_rcvctrl(rcd->ppd, QIB_RCVCTRL_CTXT_DIS |
+				  QIB_RCVCTRL_INTRAVAIL_DIS, ctxt);
+
+		/* clean up the pkeys for this ctxt user */
+		qib_clean_part_key(rcd, dd);
+		qib_disarm_piobufs(dd, rcd->pio_base, rcd->piocnt);
+		qib_chg_pioavailkernel(dd, rcd->pio_base,
+				       rcd->piocnt, TXCHK_CHG_TYPE_KERN, NULL);
+
+		dd->f_clear_tids(dd, rcd);
+
+		if (dd->pageshadow)
+			unlock_expected_tids(rcd);
+		qib_stats.sps_ctxts--;
+	}
+
+	mutex_unlock(&qib_mutex);
+	qib_free_ctxtdata(dd, rcd); /* after releasing the mutex */
+
+bail:
+	kfree(fd);
+	return ret;
+}
+
+static int qib_ctxt_info(struct file *fp, struct qib_ctxt_info __user *uinfo)
+{
+	struct qib_ctxt_info info;
+	int ret;
+	size_t sz;
+	struct qib_ctxtdata *rcd = ctxt_fp(fp);
+	struct qib_filedata *fd;
+
+	fd = (struct qib_filedata *) fp->private_data;
+
+	info.num_active = qib_count_active_units();
+	info.unit = rcd->dd->unit;
+	info.port = rcd->ppd->port;
+	info.ctxt = rcd->ctxt;
+	info.subctxt =  subctxt_fp(fp);
+	/* Number of user ctxts available for this device. */
+	info.num_ctxts = rcd->dd->cfgctxts - rcd->dd->first_user_ctxt;
+	info.num_subctxts = rcd->subctxt_cnt;
+	info.rec_cpu = fd->rec_cpu_num;
+	sz = sizeof(info);
+
+	if (copy_to_user(uinfo, &info, sz)) {
+		ret = -EFAULT;
+		goto bail;
+	}
+	ret = 0;
+
+bail:
+	return ret;
+}
+
+static int qib_sdma_get_inflight(struct qib_user_sdma_queue *pq,
+				 u32 __user *inflightp)
+{
+	const u32 val = qib_user_sdma_inflight_counter(pq);
+
+	if (put_user(val, inflightp))
+		return -EFAULT;
+
+	return 0;
+}
+
+static int qib_sdma_get_complete(struct qib_pportdata *ppd,
+				 struct qib_user_sdma_queue *pq,
+				 u32 __user *completep)
+{
+	u32 val;
+	int err;
+
+	if (!pq)
+		return -EINVAL;
+
+	err = qib_user_sdma_make_progress(ppd, pq);
+	if (err < 0)
+		return err;
+
+	val = qib_user_sdma_complete_counter(pq);
+	if (put_user(val, completep))
+		return -EFAULT;
+
+	return 0;
+}
+
+static int disarm_req_delay(struct qib_ctxtdata *rcd)
+{
+	int ret = 0;
+
+	if (!usable(rcd->ppd, 1)) {
+		int i;
+		/*
+		 * if link is down, or otherwise not usable, delay
+		 * the caller up to 30 seconds, so we don't thrash
+		 * in trying to get the chip back to ACTIVE, and
+		 * set flag so they make the call again.
+		 */
+		if (rcd->user_event_mask) {
+			/*
+			 * subctxt_cnt is 0 if not shared, so do base
+			 * separately, first, then remaining subctxt, if any
+			 */
+			set_bit(_QIB_EVENT_DISARM_BUFS_BIT,
+				&rcd->user_event_mask[0]);
+			for (i = 1; i < rcd->subctxt_cnt; i++)
+				set_bit(_QIB_EVENT_DISARM_BUFS_BIT,
+					&rcd->user_event_mask[i]);
+		}
+		for (i = 0; !usable(rcd->ppd, 1) && i < 300; i++)
+			msleep(100);
+		ret = -ENETDOWN;
+	}
+	return ret;
+}
+
+/*
+ * Find all user contexts in use, and set the specified bit in their
+ * event mask.
+ * See also find_ctxt() for a similar use, that is specific to send buffers.
+ */
+int qib_set_uevent_bits(struct qib_pportdata *ppd, const int evtbit)
+{
+	struct qib_ctxtdata *rcd;
+	unsigned ctxt;
+	int ret = 0;
+
+	spin_lock(&ppd->dd->uctxt_lock);
+	for (ctxt = ppd->dd->first_user_ctxt; ctxt < ppd->dd->cfgctxts;
+	     ctxt++) {
+		rcd = ppd->dd->rcd[ctxt];
+		if (!rcd)
+			continue;
+		if (rcd->user_event_mask) {
+			int i;
+			/*
+			 * subctxt_cnt is 0 if not shared, so do base
+			 * separately, first, then remaining subctxt, if any
+			 */
+			set_bit(evtbit, &rcd->user_event_mask[0]);
+			for (i = 1; i < rcd->subctxt_cnt; i++)
+				set_bit(evtbit, &rcd->user_event_mask[i]);
+		}
+		ret = 1;
+		break;
+	}
+	spin_unlock(&ppd->dd->uctxt_lock);
+
+	return ret;
+}
+
+/*
+ * clear the event notifier events for this context.
+ * For the DISARM_BUFS case, we also take action (this obsoletes
+ * the older QIB_CMD_DISARM_BUFS, but we keep it for backwards
+ * compatibility.
+ * Other bits don't currently require actions, just atomically clear.
+ * User process then performs actions appropriate to bit having been
+ * set, if desired, and checks again in future.
+ */
+static int qib_user_event_ack(struct qib_ctxtdata *rcd, int subctxt,
+			      unsigned long events)
+{
+	int ret = 0, i;
+
+	for (i = 0; i <= _QIB_MAX_EVENT_BIT; i++) {
+		if (!test_bit(i, &events))
+			continue;
+		if (i == _QIB_EVENT_DISARM_BUFS_BIT) {
+			(void)qib_disarm_piobufs_ifneeded(rcd);
+			ret = disarm_req_delay(rcd);
+		} else
+			clear_bit(i, &rcd->user_event_mask[subctxt]);
+	}
+	return ret;
+}
+
+static ssize_t qib_write(struct file *fp, const char __user *data,
+			 size_t count, loff_t *off)
+{
+	const struct qib_cmd __user *ucmd;
+	struct qib_ctxtdata *rcd;
+	const void __user *src;
+	size_t consumed, copy = 0;
+	struct qib_cmd cmd;
+	ssize_t ret = 0;
+	void *dest;
+
+	if (count < sizeof(cmd.type)) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	ucmd = (const struct qib_cmd __user *) data;
+
+	if (copy_from_user(&cmd.type, &ucmd->type, sizeof(cmd.type))) {
+		ret = -EFAULT;
+		goto bail;
+	}
+
+	consumed = sizeof(cmd.type);
+
+	switch (cmd.type) {
+	case QIB_CMD_ASSIGN_CTXT:
+	case QIB_CMD_USER_INIT:
+		copy = sizeof(cmd.cmd.user_info);
+		dest = &cmd.cmd.user_info;
+		src = &ucmd->cmd.user_info;
+		break;
+
+	case QIB_CMD_RECV_CTRL:
+		copy = sizeof(cmd.cmd.recv_ctrl);
+		dest = &cmd.cmd.recv_ctrl;
+		src = &ucmd->cmd.recv_ctrl;
+		break;
+
+	case QIB_CMD_CTXT_INFO:
+		copy = sizeof(cmd.cmd.ctxt_info);
+		dest = &cmd.cmd.ctxt_info;
+		src = &ucmd->cmd.ctxt_info;
+		break;
+
+	case QIB_CMD_TID_UPDATE:
+	case QIB_CMD_TID_FREE:
+		copy = sizeof(cmd.cmd.tid_info);
+		dest = &cmd.cmd.tid_info;
+		src = &ucmd->cmd.tid_info;
+		break;
+
+	case QIB_CMD_SET_PART_KEY:
+		copy = sizeof(cmd.cmd.part_key);
+		dest = &cmd.cmd.part_key;
+		src = &ucmd->cmd.part_key;
+		break;
+
+	case QIB_CMD_DISARM_BUFS:
+	case QIB_CMD_PIOAVAILUPD: /* force an update of PIOAvail reg */
+		copy = 0;
+		src = NULL;
+		dest = NULL;
+		break;
+
+	case QIB_CMD_POLL_TYPE:
+		copy = sizeof(cmd.cmd.poll_type);
+		dest = &cmd.cmd.poll_type;
+		src = &ucmd->cmd.poll_type;
+		break;
+
+	case QIB_CMD_ARMLAUNCH_CTRL:
+		copy = sizeof(cmd.cmd.armlaunch_ctrl);
+		dest = &cmd.cmd.armlaunch_ctrl;
+		src = &ucmd->cmd.armlaunch_ctrl;
+		break;
+
+	case QIB_CMD_SDMA_INFLIGHT:
+		copy = sizeof(cmd.cmd.sdma_inflight);
+		dest = &cmd.cmd.sdma_inflight;
+		src = &ucmd->cmd.sdma_inflight;
+		break;
+
+	case QIB_CMD_SDMA_COMPLETE:
+		copy = sizeof(cmd.cmd.sdma_complete);
+		dest = &cmd.cmd.sdma_complete;
+		src = &ucmd->cmd.sdma_complete;
+		break;
+
+	case QIB_CMD_ACK_EVENT:
+		copy = sizeof(cmd.cmd.event_mask);
+		dest = &cmd.cmd.event_mask;
+		src = &ucmd->cmd.event_mask;
+		break;
+
+	default:
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	if (copy) {
+		if ((count - consumed) < copy) {
+			ret = -EINVAL;
+			goto bail;
+		}
+		if (copy_from_user(dest, src, copy)) {
+			ret = -EFAULT;
+			goto bail;
+		}
+		consumed += copy;
+	}
+
+	rcd = ctxt_fp(fp);
+	if (!rcd && cmd.type != QIB_CMD_ASSIGN_CTXT) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	switch (cmd.type) {
+	case QIB_CMD_ASSIGN_CTXT:
+		ret = qib_assign_ctxt(fp, &cmd.cmd.user_info);
+		if (ret)
+			goto bail;
+		break;
+
+	case QIB_CMD_USER_INIT:
+		ret = qib_do_user_init(fp, &cmd.cmd.user_info);
+		if (ret)
+			goto bail;
+		ret = qib_get_base_info(fp, (void __user *) (unsigned long)
+					cmd.cmd.user_info.spu_base_info,
+					cmd.cmd.user_info.spu_base_info_size);
+		break;
+
+	case QIB_CMD_RECV_CTRL:
+		ret = qib_manage_rcvq(rcd, subctxt_fp(fp), cmd.cmd.recv_ctrl);
+		break;
+
+	case QIB_CMD_CTXT_INFO:
+		ret = qib_ctxt_info(fp, (struct qib_ctxt_info __user *)
+				    (unsigned long) cmd.cmd.ctxt_info);
+		break;
+
+	case QIB_CMD_TID_UPDATE:
+		ret = qib_tid_update(rcd, fp, &cmd.cmd.tid_info);
+		break;
+
+	case QIB_CMD_TID_FREE:
+		ret = qib_tid_free(rcd, subctxt_fp(fp), &cmd.cmd.tid_info);
+		break;
+
+	case QIB_CMD_SET_PART_KEY:
+		ret = qib_set_part_key(rcd, cmd.cmd.part_key);
+		break;
+
+	case QIB_CMD_DISARM_BUFS:
+		(void)qib_disarm_piobufs_ifneeded(rcd);
+		ret = disarm_req_delay(rcd);
+		break;
+
+	case QIB_CMD_PIOAVAILUPD:
+		qib_force_pio_avail_update(rcd->dd);
+		break;
+
+	case QIB_CMD_POLL_TYPE:
+		rcd->poll_type = cmd.cmd.poll_type;
+		break;
+
+	case QIB_CMD_ARMLAUNCH_CTRL:
+		rcd->dd->f_set_armlaunch(rcd->dd, cmd.cmd.armlaunch_ctrl);
+		break;
+
+	case QIB_CMD_SDMA_INFLIGHT:
+		ret = qib_sdma_get_inflight(user_sdma_queue_fp(fp),
+					    (u32 __user *) (unsigned long)
+					    cmd.cmd.sdma_inflight);
+		break;
+
+	case QIB_CMD_SDMA_COMPLETE:
+		ret = qib_sdma_get_complete(rcd->ppd,
+					    user_sdma_queue_fp(fp),
+					    (u32 __user *) (unsigned long)
+					    cmd.cmd.sdma_complete);
+		break;
+
+	case QIB_CMD_ACK_EVENT:
+		ret = qib_user_event_ack(rcd, subctxt_fp(fp),
+					 cmd.cmd.event_mask);
+		break;
+	}
+
+	if (ret >= 0)
+		ret = consumed;
+
+bail:
+	return ret;
+}
+
+static ssize_t qib_aio_write(struct kiocb *iocb, const struct iovec *iov,
+			     unsigned long dim, loff_t off)
+{
+	struct qib_filedata *fp = iocb->ki_filp->private_data;
+	struct qib_ctxtdata *rcd = ctxt_fp(iocb->ki_filp);
+	struct qib_user_sdma_queue *pq = fp->pq;
+
+	if (!dim || !pq)
+		return -EINVAL;
+
+	return qib_user_sdma_writev(rcd, pq, iov, dim);
+}
+
+static struct class *qib_class;
+static dev_t qib_dev;
+
+int qib_cdev_init(int minor, const char *name,
+		  const struct file_operations *fops,
+		  struct cdev **cdevp, struct device **devp)
+{
+	const dev_t dev = MKDEV(MAJOR(qib_dev), minor);
+	struct cdev *cdev;
+	struct device *device = NULL;
+	int ret;
+
+	cdev = cdev_alloc();
+	if (!cdev) {
+		printk(KERN_ERR QIB_DRV_NAME
+		       ": Could not allocate cdev for minor %d, %s\n",
+		       minor, name);
+		ret = -ENOMEM;
+		goto done;
+	}
+
+	cdev->owner = THIS_MODULE;
+	cdev->ops = fops;
+	kobject_set_name(&cdev->kobj, name);
+
+	ret = cdev_add(cdev, dev, 1);
+	if (ret < 0) {
+		printk(KERN_ERR QIB_DRV_NAME
+		       ": Could not add cdev for minor %d, %s (err %d)\n",
+		       minor, name, -ret);
+		goto err_cdev;
+	}
+
+	device = device_create(qib_class, NULL, dev, NULL, name);
+	if (!IS_ERR(device))
+		goto done;
+	ret = PTR_ERR(device);
+	device = NULL;
+	printk(KERN_ERR QIB_DRV_NAME ": Could not create "
+	       "device for minor %d, %s (err %d)\n",
+	       minor, name, -ret);
+err_cdev:
+	cdev_del(cdev);
+	cdev = NULL;
+done:
+	*cdevp = cdev;
+	*devp = device;
+	return ret;
+}
+
+void qib_cdev_cleanup(struct cdev **cdevp, struct device **devp)
+{
+	struct device *device = *devp;
+
+	if (device) {
+		device_unregister(device);
+		*devp = NULL;
+	}
+
+	if (*cdevp) {
+		cdev_del(*cdevp);
+		*cdevp = NULL;
+	}
+}
+
+static struct cdev *wildcard_cdev;
+static struct device *wildcard_device;
+
+int __init qib_dev_init(void)
+{
+	int ret;
+
+	ret = alloc_chrdev_region(&qib_dev, 0, QIB_NMINORS, QIB_DRV_NAME);
+	if (ret < 0) {
+		printk(KERN_ERR QIB_DRV_NAME ": Could not allocate "
+		       "chrdev region (err %d)\n", -ret);
+		goto done;
+	}
+
+	qib_class = class_create(THIS_MODULE, "ipath");
+	if (IS_ERR(qib_class)) {
+		ret = PTR_ERR(qib_class);
+		printk(KERN_ERR QIB_DRV_NAME ": Could not create "
+		       "device class (err %d)\n", -ret);
+		unregister_chrdev_region(qib_dev, QIB_NMINORS);
+	}
+
+done:
+	return ret;
+}
+
+void qib_dev_cleanup(void)
+{
+	if (qib_class) {
+		class_destroy(qib_class);
+		qib_class = NULL;
+	}
+
+	unregister_chrdev_region(qib_dev, QIB_NMINORS);
+}
+
+static atomic_t user_count = ATOMIC_INIT(0);
+
+static void qib_user_remove(struct qib_devdata *dd)
+{
+	if (atomic_dec_return(&user_count) == 0)
+		qib_cdev_cleanup(&wildcard_cdev, &wildcard_device);
+
+	qib_cdev_cleanup(&dd->user_cdev, &dd->user_device);
+}
+
+static int qib_user_add(struct qib_devdata *dd)
+{
+	char name[10];
+	int ret;
+
+	if (atomic_inc_return(&user_count) == 1) {
+		ret = qib_cdev_init(0, "ipath", &qib_file_ops,
+				    &wildcard_cdev, &wildcard_device);
+		if (ret)
+			goto done;
+	}
+
+	snprintf(name, sizeof(name), "ipath%d", dd->unit);
+	ret = qib_cdev_init(dd->unit + 1, name, &qib_file_ops,
+			    &dd->user_cdev, &dd->user_device);
+	if (ret)
+		qib_user_remove(dd);
+done:
+	return ret;
+}
+
+/*
+ * Create per-unit files in /dev
+ */
+int qib_device_create(struct qib_devdata *dd)
+{
+	int r, ret;
+
+	r = qib_user_add(dd);
+	ret = qib_diag_add(dd);
+	if (r && !ret)
+		ret = r;
+	return ret;
+}
+
+/*
+ * Remove per-unit files in /dev
+ * void, core kernel returns no errors for this stuff
+ */
+void qib_device_remove(struct qib_devdata *dd)
+{
+	qib_user_remove(dd);
+	qib_diag_remove(dd);
+}
diff --git a/drivers/infiniband/hw/qib/qib_fs.c b/drivers/infiniband/hw/qib/qib_fs.c
new file mode 100644
index 000000000000..844954bf417b
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_fs.c
@@ -0,0 +1,616 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/module.h>
+#include <linux/fs.h>
+#include <linux/mount.h>
+#include <linux/pagemap.h>
+#include <linux/init.h>
+#include <linux/namei.h>
+
+#include "qib.h"
+
+#define QIBFS_MAGIC 0x726a77
+
+static struct super_block *qib_super;
+
+#define private2dd(file) ((file)->f_dentry->d_inode->i_private)
+
+static int qibfs_mknod(struct inode *dir, struct dentry *dentry,
+		       int mode, const struct file_operations *fops,
+		       void *data)
+{
+	int error;
+	struct inode *inode = new_inode(dir->i_sb);
+
+	if (!inode) {
+		error = -EPERM;
+		goto bail;
+	}
+
+	inode->i_mode = mode;
+	inode->i_uid = 0;
+	inode->i_gid = 0;
+	inode->i_blocks = 0;
+	inode->i_atime = CURRENT_TIME;
+	inode->i_mtime = inode->i_atime;
+	inode->i_ctime = inode->i_atime;
+	inode->i_private = data;
+	if ((mode & S_IFMT) == S_IFDIR) {
+		inode->i_op = &simple_dir_inode_operations;
+		inc_nlink(inode);
+		inc_nlink(dir);
+	}
+
+	inode->i_fop = fops;
+
+	d_instantiate(dentry, inode);
+	error = 0;
+
+bail:
+	return error;
+}
+
+static int create_file(const char *name, mode_t mode,
+		       struct dentry *parent, struct dentry **dentry,
+		       const struct file_operations *fops, void *data)
+{
+	int error;
+
+	*dentry = NULL;
+	mutex_lock(&parent->d_inode->i_mutex);
+	*dentry = lookup_one_len(name, parent, strlen(name));
+	if (!IS_ERR(*dentry))
+		error = qibfs_mknod(parent->d_inode, *dentry,
+				    mode, fops, data);
+	else
+		error = PTR_ERR(*dentry);
+	mutex_unlock(&parent->d_inode->i_mutex);
+
+	return error;
+}
+
+static ssize_t driver_stats_read(struct file *file, char __user *buf,
+				 size_t count, loff_t *ppos)
+{
+	return simple_read_from_buffer(buf, count, ppos, &qib_stats,
+				       sizeof qib_stats);
+}
+
+/*
+ * driver stats field names, one line per stat, single string.  Used by
+ * programs like ipathstats to print the stats in a way which works for
+ * different versions of drivers, without changing program source.
+ * if qlogic_ib_stats changes, this needs to change.  Names need to be
+ * 12 chars or less (w/o newline), for proper display by ipathstats utility.
+ */
+static const char qib_statnames[] =
+	"KernIntr\n"
+	"ErrorIntr\n"
+	"Tx_Errs\n"
+	"Rcv_Errs\n"
+	"H/W_Errs\n"
+	"NoPIOBufs\n"
+	"CtxtsOpen\n"
+	"RcvLen_Errs\n"
+	"EgrBufFull\n"
+	"EgrHdrFull\n"
+	;
+
+static ssize_t driver_names_read(struct file *file, char __user *buf,
+				 size_t count, loff_t *ppos)
+{
+	return simple_read_from_buffer(buf, count, ppos, qib_statnames,
+		sizeof qib_statnames - 1); /* no null */
+}
+
+static const struct file_operations driver_ops[] = {
+	{ .read = driver_stats_read, },
+	{ .read = driver_names_read, },
+};
+
+/* read the per-device counters */
+static ssize_t dev_counters_read(struct file *file, char __user *buf,
+				 size_t count, loff_t *ppos)
+{
+	u64 *counters;
+	size_t avail;
+	struct qib_devdata *dd = private2dd(file);
+
+	avail = dd->f_read_cntrs(dd, *ppos, NULL, &counters);
+	return simple_read_from_buffer(buf, count, ppos, counters, avail);
+}
+
+/* read the per-device counters */
+static ssize_t dev_names_read(struct file *file, char __user *buf,
+			      size_t count, loff_t *ppos)
+{
+	char *names;
+	size_t avail;
+	struct qib_devdata *dd = private2dd(file);
+
+	avail = dd->f_read_cntrs(dd, *ppos, &names, NULL);
+	return simple_read_from_buffer(buf, count, ppos, names, avail);
+}
+
+static const struct file_operations cntr_ops[] = {
+	{ .read = dev_counters_read, },
+	{ .read = dev_names_read, },
+};
+
+/*
+ * Could use file->f_dentry->d_inode->i_ino to figure out which file,
+ * instead of separate routine for each, but for now, this works...
+ */
+
+/* read the per-port names (same for each port) */
+static ssize_t portnames_read(struct file *file, char __user *buf,
+			      size_t count, loff_t *ppos)
+{
+	char *names;
+	size_t avail;
+	struct qib_devdata *dd = private2dd(file);
+
+	avail = dd->f_read_portcntrs(dd, *ppos, 0, &names, NULL);
+	return simple_read_from_buffer(buf, count, ppos, names, avail);
+}
+
+/* read the per-port counters for port 1 (pidx 0) */
+static ssize_t portcntrs_1_read(struct file *file, char __user *buf,
+				size_t count, loff_t *ppos)
+{
+	u64 *counters;
+	size_t avail;
+	struct qib_devdata *dd = private2dd(file);
+
+	avail = dd->f_read_portcntrs(dd, *ppos, 0, NULL, &counters);
+	return simple_read_from_buffer(buf, count, ppos, counters, avail);
+}
+
+/* read the per-port counters for port 2 (pidx 1) */
+static ssize_t portcntrs_2_read(struct file *file, char __user *buf,
+				size_t count, loff_t *ppos)
+{
+	u64 *counters;
+	size_t avail;
+	struct qib_devdata *dd = private2dd(file);
+
+	avail = dd->f_read_portcntrs(dd, *ppos, 1, NULL, &counters);
+	return simple_read_from_buffer(buf, count, ppos, counters, avail);
+}
+
+static const struct file_operations portcntr_ops[] = {
+	{ .read = portnames_read, },
+	{ .read = portcntrs_1_read, },
+	{ .read = portcntrs_2_read, },
+};
+
+/*
+ * read the per-port QSFP data for port 1 (pidx 0)
+ */
+static ssize_t qsfp_1_read(struct file *file, char __user *buf,
+			   size_t count, loff_t *ppos)
+{
+	struct qib_devdata *dd = private2dd(file);
+	char *tmp;
+	int ret;
+
+	tmp = kmalloc(PAGE_SIZE, GFP_KERNEL);
+	if (!tmp)
+		return -ENOMEM;
+
+	ret = qib_qsfp_dump(dd->pport, tmp, PAGE_SIZE);
+	if (ret > 0)
+		ret = simple_read_from_buffer(buf, count, ppos, tmp, ret);
+	kfree(tmp);
+	return ret;
+}
+
+/*
+ * read the per-port QSFP data for port 2 (pidx 1)
+ */
+static ssize_t qsfp_2_read(struct file *file, char __user *buf,
+			   size_t count, loff_t *ppos)
+{
+	struct qib_devdata *dd = private2dd(file);
+	char *tmp;
+	int ret;
+
+	if (dd->num_pports < 2)
+		return -ENODEV;
+
+	tmp = kmalloc(PAGE_SIZE, GFP_KERNEL);
+	if (!tmp)
+		return -ENOMEM;
+
+	ret = qib_qsfp_dump(dd->pport + 1, tmp, PAGE_SIZE);
+	if (ret > 0)
+		ret = simple_read_from_buffer(buf, count, ppos, tmp, ret);
+	kfree(tmp);
+	return ret;
+}
+
+static const struct file_operations qsfp_ops[] = {
+	{ .read = qsfp_1_read, },
+	{ .read = qsfp_2_read, },
+};
+
+static ssize_t flash_read(struct file *file, char __user *buf,
+			  size_t count, loff_t *ppos)
+{
+	struct qib_devdata *dd;
+	ssize_t ret;
+	loff_t pos;
+	char *tmp;
+
+	pos = *ppos;
+
+	if (pos < 0) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	if (pos >= sizeof(struct qib_flash)) {
+		ret = 0;
+		goto bail;
+	}
+
+	if (count > sizeof(struct qib_flash) - pos)
+		count = sizeof(struct qib_flash) - pos;
+
+	tmp = kmalloc(count, GFP_KERNEL);
+	if (!tmp) {
+		ret = -ENOMEM;
+		goto bail;
+	}
+
+	dd = private2dd(file);
+	if (qib_eeprom_read(dd, pos, tmp, count)) {
+		qib_dev_err(dd, "failed to read from flash\n");
+		ret = -ENXIO;
+		goto bail_tmp;
+	}
+
+	if (copy_to_user(buf, tmp, count)) {
+		ret = -EFAULT;
+		goto bail_tmp;
+	}
+
+	*ppos = pos + count;
+	ret = count;
+
+bail_tmp:
+	kfree(tmp);
+
+bail:
+	return ret;
+}
+
+static ssize_t flash_write(struct file *file, const char __user *buf,
+			   size_t count, loff_t *ppos)
+{
+	struct qib_devdata *dd;
+	ssize_t ret;
+	loff_t pos;
+	char *tmp;
+
+	pos = *ppos;
+
+	if (pos != 0) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	if (count != sizeof(struct qib_flash)) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	tmp = kmalloc(count, GFP_KERNEL);
+	if (!tmp) {
+		ret = -ENOMEM;
+		goto bail;
+	}
+
+	if (copy_from_user(tmp, buf, count)) {
+		ret = -EFAULT;
+		goto bail_tmp;
+	}
+
+	dd = private2dd(file);
+	if (qib_eeprom_write(dd, pos, tmp, count)) {
+		ret = -ENXIO;
+		qib_dev_err(dd, "failed to write to flash\n");
+		goto bail_tmp;
+	}
+
+	*ppos = pos + count;
+	ret = count;
+
+bail_tmp:
+	kfree(tmp);
+
+bail:
+	return ret;
+}
+
+static const struct file_operations flash_ops = {
+	.read = flash_read,
+	.write = flash_write,
+};
+
+static int add_cntr_files(struct super_block *sb, struct qib_devdata *dd)
+{
+	struct dentry *dir, *tmp;
+	char unit[10];
+	int ret, i;
+
+	/* create the per-unit directory */
+	snprintf(unit, sizeof unit, "%u", dd->unit);
+	ret = create_file(unit, S_IFDIR|S_IRUGO|S_IXUGO, sb->s_root, &dir,
+			  &simple_dir_operations, dd);
+	if (ret) {
+		printk(KERN_ERR "create_file(%s) failed: %d\n", unit, ret);
+		goto bail;
+	}
+
+	/* create the files in the new directory */
+	ret = create_file("counters", S_IFREG|S_IRUGO, dir, &tmp,
+			  &cntr_ops[0], dd);
+	if (ret) {
+		printk(KERN_ERR "create_file(%s/counters) failed: %d\n",
+		       unit, ret);
+		goto bail;
+	}
+	ret = create_file("counter_names", S_IFREG|S_IRUGO, dir, &tmp,
+			  &cntr_ops[1], dd);
+	if (ret) {
+		printk(KERN_ERR "create_file(%s/counter_names) failed: %d\n",
+		       unit, ret);
+		goto bail;
+	}
+	ret = create_file("portcounter_names", S_IFREG|S_IRUGO, dir, &tmp,
+			  &portcntr_ops[0], dd);
+	if (ret) {
+		printk(KERN_ERR "create_file(%s/%s) failed: %d\n",
+		       unit, "portcounter_names", ret);
+		goto bail;
+	}
+	for (i = 1; i <= dd->num_pports; i++) {
+		char fname[24];
+
+		sprintf(fname, "port%dcounters", i);
+		/* create the files in the new directory */
+		ret = create_file(fname, S_IFREG|S_IRUGO, dir, &tmp,
+				  &portcntr_ops[i], dd);
+		if (ret) {
+			printk(KERN_ERR "create_file(%s/%s) failed: %d\n",
+				unit, fname, ret);
+			goto bail;
+		}
+		if (!(dd->flags & QIB_HAS_QSFP))
+			continue;
+		sprintf(fname, "qsfp%d", i);
+		ret = create_file(fname, S_IFREG|S_IRUGO, dir, &tmp,
+				  &qsfp_ops[i - 1], dd);
+		if (ret) {
+			printk(KERN_ERR "create_file(%s/%s) failed: %d\n",
+				unit, fname, ret);
+			goto bail;
+		}
+	}
+
+	ret = create_file("flash", S_IFREG|S_IWUSR|S_IRUGO, dir, &tmp,
+			  &flash_ops, dd);
+	if (ret)
+		printk(KERN_ERR "create_file(%s/flash) failed: %d\n",
+			unit, ret);
+bail:
+	return ret;
+}
+
+static int remove_file(struct dentry *parent, char *name)
+{
+	struct dentry *tmp;
+	int ret;
+
+	tmp = lookup_one_len(name, parent, strlen(name));
+
+	if (IS_ERR(tmp)) {
+		ret = PTR_ERR(tmp);
+		goto bail;
+	}
+
+	spin_lock(&dcache_lock);
+	spin_lock(&tmp->d_lock);
+	if (!(d_unhashed(tmp) && tmp->d_inode)) {
+		dget_locked(tmp);
+		__d_drop(tmp);
+		spin_unlock(&tmp->d_lock);
+		spin_unlock(&dcache_lock);
+		simple_unlink(parent->d_inode, tmp);
+	} else {
+		spin_unlock(&tmp->d_lock);
+		spin_unlock(&dcache_lock);
+	}
+
+	ret = 0;
+bail:
+	/*
+	 * We don't expect clients to care about the return value, but
+	 * it's there if they need it.
+	 */
+	return ret;
+}
+
+static int remove_device_files(struct super_block *sb,
+			       struct qib_devdata *dd)
+{
+	struct dentry *dir, *root;
+	char unit[10];
+	int ret, i;
+
+	root = dget(sb->s_root);
+	mutex_lock(&root->d_inode->i_mutex);
+	snprintf(unit, sizeof unit, "%u", dd->unit);
+	dir = lookup_one_len(unit, root, strlen(unit));
+
+	if (IS_ERR(dir)) {
+		ret = PTR_ERR(dir);
+		printk(KERN_ERR "Lookup of %s failed\n", unit);
+		goto bail;
+	}
+
+	remove_file(dir, "counters");
+	remove_file(dir, "counter_names");
+	remove_file(dir, "portcounter_names");
+	for (i = 0; i < dd->num_pports; i++) {
+		char fname[24];
+
+		sprintf(fname, "port%dcounters", i + 1);
+		remove_file(dir, fname);
+		if (dd->flags & QIB_HAS_QSFP) {
+			sprintf(fname, "qsfp%d", i + 1);
+			remove_file(dir, fname);
+		}
+	}
+	remove_file(dir, "flash");
+	d_delete(dir);
+	ret = simple_rmdir(root->d_inode, dir);
+
+bail:
+	mutex_unlock(&root->d_inode->i_mutex);
+	dput(root);
+	return ret;
+}
+
+/*
+ * This fills everything in when the fs is mounted, to handle umount/mount
+ * after device init.  The direct add_cntr_files() call handles adding
+ * them from the init code, when the fs is already mounted.
+ */
+static int qibfs_fill_super(struct super_block *sb, void *data, int silent)
+{
+	struct qib_devdata *dd, *tmp;
+	unsigned long flags;
+	int ret;
+
+	static struct tree_descr files[] = {
+		[2] = {"driver_stats", &driver_ops[0], S_IRUGO},
+		[3] = {"driver_stats_names", &driver_ops[1], S_IRUGO},
+		{""},
+	};
+
+	ret = simple_fill_super(sb, QIBFS_MAGIC, files);
+	if (ret) {
+		printk(KERN_ERR "simple_fill_super failed: %d\n", ret);
+		goto bail;
+	}
+
+	spin_lock_irqsave(&qib_devs_lock, flags);
+
+	list_for_each_entry_safe(dd, tmp, &qib_dev_list, list) {
+		spin_unlock_irqrestore(&qib_devs_lock, flags);
+		ret = add_cntr_files(sb, dd);
+		if (ret)
+			goto bail;
+		spin_lock_irqsave(&qib_devs_lock, flags);
+	}
+
+	spin_unlock_irqrestore(&qib_devs_lock, flags);
+
+bail:
+	return ret;
+}
+
+static int qibfs_get_sb(struct file_system_type *fs_type, int flags,
+			const char *dev_name, void *data, struct vfsmount *mnt)
+{
+	int ret = get_sb_single(fs_type, flags, data,
+				qibfs_fill_super, mnt);
+	if (ret >= 0)
+		qib_super = mnt->mnt_sb;
+	return ret;
+}
+
+static void qibfs_kill_super(struct super_block *s)
+{
+	kill_litter_super(s);
+	qib_super = NULL;
+}
+
+int qibfs_add(struct qib_devdata *dd)
+{
+	int ret;
+
+	/*
+	 * On first unit initialized, qib_super will not yet exist
+	 * because nobody has yet tried to mount the filesystem, so
+	 * we can't consider that to be an error; if an error occurs
+	 * during the mount, that will get a complaint, so this is OK.
+	 * add_cntr_files() for all units is done at mount from
+	 * qibfs_fill_super(), so one way or another, everything works.
+	 */
+	if (qib_super == NULL)
+		ret = 0;
+	else
+		ret = add_cntr_files(qib_super, dd);
+	return ret;
+}
+
+int qibfs_remove(struct qib_devdata *dd)
+{
+	int ret = 0;
+
+	if (qib_super)
+		ret = remove_device_files(qib_super, dd);
+
+	return ret;
+}
+
+static struct file_system_type qibfs_fs_type = {
+	.owner =        THIS_MODULE,
+	.name =         "ipathfs",
+	.get_sb =       qibfs_get_sb,
+	.kill_sb =      qibfs_kill_super,
+};
+
+int __init qib_init_qibfs(void)
+{
+	return register_filesystem(&qibfs_fs_type);
+}
+
+int __exit qib_exit_qibfs(void)
+{
+	return unregister_filesystem(&qibfs_fs_type);
+}
diff --git a/drivers/infiniband/hw/qib/qib_iba6120.c b/drivers/infiniband/hw/qib/qib_iba6120.c
new file mode 100644
index 000000000000..1eadadc13da8
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_iba6120.c
@@ -0,0 +1,3576 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009, 2010 QLogic Corporation.
+ * All rights reserved.
+ * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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.
+ */
+/*
+ * This file contains all of the code that is specific to the
+ * QLogic_IB 6120 PCIe chip.
+ */
+
+#include <linux/interrupt.h>
+#include <linux/pci.h>
+#include <linux/delay.h>
+#include <rdma/ib_verbs.h>
+
+#include "qib.h"
+#include "qib_6120_regs.h"
+
+static void qib_6120_setup_setextled(struct qib_pportdata *, u32);
+static void sendctrl_6120_mod(struct qib_pportdata *ppd, u32 op);
+static u8 qib_6120_phys_portstate(u64);
+static u32 qib_6120_iblink_state(u64);
+
+/*
+ * This file contains all the chip-specific register information and
+ * access functions for the QLogic QLogic_IB PCI-Express chip.
+ *
+ */
+
+/* KREG_IDX uses machine-generated #defines */
+#define KREG_IDX(regname) (QIB_6120_##regname##_OFFS / sizeof(u64))
+
+/* Use defines to tie machine-generated names to lower-case names */
+#define kr_extctrl KREG_IDX(EXTCtrl)
+#define kr_extstatus KREG_IDX(EXTStatus)
+#define kr_gpio_clear KREG_IDX(GPIOClear)
+#define kr_gpio_mask KREG_IDX(GPIOMask)
+#define kr_gpio_out KREG_IDX(GPIOOut)
+#define kr_gpio_status KREG_IDX(GPIOStatus)
+#define kr_rcvctrl KREG_IDX(RcvCtrl)
+#define kr_sendctrl KREG_IDX(SendCtrl)
+#define kr_partitionkey KREG_IDX(RcvPartitionKey)
+#define kr_hwdiagctrl KREG_IDX(HwDiagCtrl)
+#define kr_ibcstatus KREG_IDX(IBCStatus)
+#define kr_ibcctrl KREG_IDX(IBCCtrl)
+#define kr_sendbuffererror KREG_IDX(SendBufErr0)
+#define kr_rcvbthqp KREG_IDX(RcvBTHQP)
+#define kr_counterregbase KREG_IDX(CntrRegBase)
+#define kr_palign KREG_IDX(PageAlign)
+#define kr_rcvegrbase KREG_IDX(RcvEgrBase)
+#define kr_rcvegrcnt KREG_IDX(RcvEgrCnt)
+#define kr_rcvhdrcnt KREG_IDX(RcvHdrCnt)
+#define kr_rcvhdrentsize KREG_IDX(RcvHdrEntSize)
+#define kr_rcvhdrsize KREG_IDX(RcvHdrSize)
+#define kr_rcvtidbase KREG_IDX(RcvTIDBase)
+#define kr_rcvtidcnt KREG_IDX(RcvTIDCnt)
+#define kr_scratch KREG_IDX(Scratch)
+#define kr_sendctrl KREG_IDX(SendCtrl)
+#define kr_sendpioavailaddr KREG_IDX(SendPIOAvailAddr)
+#define kr_sendpiobufbase KREG_IDX(SendPIOBufBase)
+#define kr_sendpiobufcnt KREG_IDX(SendPIOBufCnt)
+#define kr_sendpiosize KREG_IDX(SendPIOSize)
+#define kr_sendregbase KREG_IDX(SendRegBase)
+#define kr_userregbase KREG_IDX(UserRegBase)
+#define kr_control KREG_IDX(Control)
+#define kr_intclear KREG_IDX(IntClear)
+#define kr_intmask KREG_IDX(IntMask)
+#define kr_intstatus KREG_IDX(IntStatus)
+#define kr_errclear KREG_IDX(ErrClear)
+#define kr_errmask KREG_IDX(ErrMask)
+#define kr_errstatus KREG_IDX(ErrStatus)
+#define kr_hwerrclear KREG_IDX(HwErrClear)
+#define kr_hwerrmask KREG_IDX(HwErrMask)
+#define kr_hwerrstatus KREG_IDX(HwErrStatus)
+#define kr_revision KREG_IDX(Revision)
+#define kr_portcnt KREG_IDX(PortCnt)
+#define kr_serdes_cfg0 KREG_IDX(SerdesCfg0)
+#define kr_serdes_cfg1 (kr_serdes_cfg0 + 1)
+#define kr_serdes_stat KREG_IDX(SerdesStat)
+#define kr_xgxs_cfg KREG_IDX(XGXSCfg)
+
+/* These must only be written via qib_write_kreg_ctxt() */
+#define kr_rcvhdraddr KREG_IDX(RcvHdrAddr0)
+#define kr_rcvhdrtailaddr KREG_IDX(RcvHdrTailAddr0)
+
+#define CREG_IDX(regname) ((QIB_6120_##regname##_OFFS - \
+			QIB_6120_LBIntCnt_OFFS) / sizeof(u64))
+
+#define cr_badformat CREG_IDX(RxBadFormatCnt)
+#define cr_erricrc CREG_IDX(RxICRCErrCnt)
+#define cr_errlink CREG_IDX(RxLinkProblemCnt)
+#define cr_errlpcrc CREG_IDX(RxLPCRCErrCnt)
+#define cr_errpkey CREG_IDX(RxPKeyMismatchCnt)
+#define cr_rcvflowctrl_err CREG_IDX(RxFlowCtrlErrCnt)
+#define cr_err_rlen CREG_IDX(RxLenErrCnt)
+#define cr_errslen CREG_IDX(TxLenErrCnt)
+#define cr_errtidfull CREG_IDX(RxTIDFullErrCnt)
+#define cr_errtidvalid CREG_IDX(RxTIDValidErrCnt)
+#define cr_errvcrc CREG_IDX(RxVCRCErrCnt)
+#define cr_ibstatuschange CREG_IDX(IBStatusChangeCnt)
+#define cr_lbint CREG_IDX(LBIntCnt)
+#define cr_invalidrlen CREG_IDX(RxMaxMinLenErrCnt)
+#define cr_invalidslen CREG_IDX(TxMaxMinLenErrCnt)
+#define cr_lbflowstall CREG_IDX(LBFlowStallCnt)
+#define cr_pktrcv CREG_IDX(RxDataPktCnt)
+#define cr_pktrcvflowctrl CREG_IDX(RxFlowPktCnt)
+#define cr_pktsend CREG_IDX(TxDataPktCnt)
+#define cr_pktsendflow CREG_IDX(TxFlowPktCnt)
+#define cr_portovfl CREG_IDX(RxP0HdrEgrOvflCnt)
+#define cr_rcvebp CREG_IDX(RxEBPCnt)
+#define cr_rcvovfl CREG_IDX(RxBufOvflCnt)
+#define cr_senddropped CREG_IDX(TxDroppedPktCnt)
+#define cr_sendstall CREG_IDX(TxFlowStallCnt)
+#define cr_sendunderrun CREG_IDX(TxUnderrunCnt)
+#define cr_wordrcv CREG_IDX(RxDwordCnt)
+#define cr_wordsend CREG_IDX(TxDwordCnt)
+#define cr_txunsupvl CREG_IDX(TxUnsupVLErrCnt)
+#define cr_rxdroppkt CREG_IDX(RxDroppedPktCnt)
+#define cr_iblinkerrrecov CREG_IDX(IBLinkErrRecoveryCnt)
+#define cr_iblinkdown CREG_IDX(IBLinkDownedCnt)
+#define cr_ibsymbolerr CREG_IDX(IBSymbolErrCnt)
+
+#define SYM_RMASK(regname, fldname) ((u64)              \
+	QIB_6120_##regname##_##fldname##_RMASK)
+#define SYM_MASK(regname, fldname) ((u64)               \
+	QIB_6120_##regname##_##fldname##_RMASK <<       \
+	 QIB_6120_##regname##_##fldname##_LSB)
+#define SYM_LSB(regname, fldname) (QIB_6120_##regname##_##fldname##_LSB)
+
+#define SYM_FIELD(value, regname, fldname) ((u64) \
+	(((value) >> SYM_LSB(regname, fldname)) & \
+	 SYM_RMASK(regname, fldname)))
+#define ERR_MASK(fldname) SYM_MASK(ErrMask, fldname##Mask)
+#define HWE_MASK(fldname) SYM_MASK(HwErrMask, fldname##Mask)
+
+/* link training states, from IBC */
+#define IB_6120_LT_STATE_DISABLED        0x00
+#define IB_6120_LT_STATE_LINKUP          0x01
+#define IB_6120_LT_STATE_POLLACTIVE      0x02
+#define IB_6120_LT_STATE_POLLQUIET       0x03
+#define IB_6120_LT_STATE_SLEEPDELAY      0x04
+#define IB_6120_LT_STATE_SLEEPQUIET      0x05
+#define IB_6120_LT_STATE_CFGDEBOUNCE     0x08
+#define IB_6120_LT_STATE_CFGRCVFCFG      0x09
+#define IB_6120_LT_STATE_CFGWAITRMT      0x0a
+#define IB_6120_LT_STATE_CFGIDLE 0x0b
+#define IB_6120_LT_STATE_RECOVERRETRAIN  0x0c
+#define IB_6120_LT_STATE_RECOVERWAITRMT  0x0e
+#define IB_6120_LT_STATE_RECOVERIDLE     0x0f
+
+/* link state machine states from IBC */
+#define IB_6120_L_STATE_DOWN             0x0
+#define IB_6120_L_STATE_INIT             0x1
+#define IB_6120_L_STATE_ARM              0x2
+#define IB_6120_L_STATE_ACTIVE           0x3
+#define IB_6120_L_STATE_ACT_DEFER        0x4
+
+static const u8 qib_6120_physportstate[0x20] = {
+	[IB_6120_LT_STATE_DISABLED] = IB_PHYSPORTSTATE_DISABLED,
+	[IB_6120_LT_STATE_LINKUP] = IB_PHYSPORTSTATE_LINKUP,
+	[IB_6120_LT_STATE_POLLACTIVE] = IB_PHYSPORTSTATE_POLL,
+	[IB_6120_LT_STATE_POLLQUIET] = IB_PHYSPORTSTATE_POLL,
+	[IB_6120_LT_STATE_SLEEPDELAY] = IB_PHYSPORTSTATE_SLEEP,
+	[IB_6120_LT_STATE_SLEEPQUIET] = IB_PHYSPORTSTATE_SLEEP,
+	[IB_6120_LT_STATE_CFGDEBOUNCE] =
+		IB_PHYSPORTSTATE_CFG_TRAIN,
+	[IB_6120_LT_STATE_CFGRCVFCFG] =
+		IB_PHYSPORTSTATE_CFG_TRAIN,
+	[IB_6120_LT_STATE_CFGWAITRMT] =
+		IB_PHYSPORTSTATE_CFG_TRAIN,
+	[IB_6120_LT_STATE_CFGIDLE] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[IB_6120_LT_STATE_RECOVERRETRAIN] =
+		IB_PHYSPORTSTATE_LINK_ERR_RECOVER,
+	[IB_6120_LT_STATE_RECOVERWAITRMT] =
+		IB_PHYSPORTSTATE_LINK_ERR_RECOVER,
+	[IB_6120_LT_STATE_RECOVERIDLE] =
+		IB_PHYSPORTSTATE_LINK_ERR_RECOVER,
+	[0x10] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x11] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x12] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x13] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x14] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x15] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x16] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x17] = IB_PHYSPORTSTATE_CFG_TRAIN
+};
+
+
+struct qib_chip_specific {
+	u64 __iomem *cregbase;
+	u64 *cntrs;
+	u64 *portcntrs;
+	void *dummy_hdrq;   /* used after ctxt close */
+	dma_addr_t dummy_hdrq_phys;
+	spinlock_t kernel_tid_lock; /* no back to back kernel TID writes */
+	spinlock_t user_tid_lock; /* no back to back user TID writes */
+	spinlock_t rcvmod_lock; /* protect rcvctrl shadow changes */
+	spinlock_t gpio_lock; /* RMW of shadows/regs for ExtCtrl and GPIO */
+	u64 hwerrmask;
+	u64 errormask;
+	u64 gpio_out; /* shadow of kr_gpio_out, for rmw ops */
+	u64 gpio_mask; /* shadow the gpio mask register */
+	u64 extctrl; /* shadow the gpio output enable, etc... */
+	/*
+	 * these 5 fields are used to establish deltas for IB symbol
+	 * errors and linkrecovery errors.  They can be reported on
+	 * some chips during link negotiation prior to INIT, and with
+	 * DDR when faking DDR negotiations with non-IBTA switches.
+	 * The chip counters are adjusted at driver unload if there is
+	 * a non-zero delta.
+	 */
+	u64 ibdeltainprog;
+	u64 ibsymdelta;
+	u64 ibsymsnap;
+	u64 iblnkerrdelta;
+	u64 iblnkerrsnap;
+	u64 ibcctrl; /* shadow for kr_ibcctrl */
+	u32 lastlinkrecov; /* link recovery issue */
+	int irq;
+	u32 cntrnamelen;
+	u32 portcntrnamelen;
+	u32 ncntrs;
+	u32 nportcntrs;
+	/* used with gpio interrupts to implement IB counters */
+	u32 rxfc_unsupvl_errs;
+	u32 overrun_thresh_errs;
+	/*
+	 * these count only cases where _successive_ LocalLinkIntegrity
+	 * errors were seen in the receive headers of IB standard packets
+	 */
+	u32 lli_errs;
+	u32 lli_counter;
+	u64 lli_thresh;
+	u64 sword; /* total dwords sent (sample result) */
+	u64 rword; /* total dwords received (sample result) */
+	u64 spkts; /* total packets sent (sample result) */
+	u64 rpkts; /* total packets received (sample result) */
+	u64 xmit_wait; /* # of ticks no data sent (sample result) */
+	struct timer_list pma_timer;
+	char emsgbuf[128];
+	char bitsmsgbuf[64];
+	u8 pma_sample_status;
+};
+
+/* ibcctrl bits */
+#define QLOGIC_IB_IBCC_LINKINITCMD_DISABLE 1
+/* cycle through TS1/TS2 till OK */
+#define QLOGIC_IB_IBCC_LINKINITCMD_POLL 2
+/* wait for TS1, then go on */
+#define QLOGIC_IB_IBCC_LINKINITCMD_SLEEP 3
+#define QLOGIC_IB_IBCC_LINKINITCMD_SHIFT 16
+
+#define QLOGIC_IB_IBCC_LINKCMD_DOWN 1           /* move to 0x11 */
+#define QLOGIC_IB_IBCC_LINKCMD_ARMED 2          /* move to 0x21 */
+#define QLOGIC_IB_IBCC_LINKCMD_ACTIVE 3 /* move to 0x31 */
+#define QLOGIC_IB_IBCC_LINKCMD_SHIFT 18
+
+/*
+ * We could have a single register get/put routine, that takes a group type,
+ * but this is somewhat clearer and cleaner.  It also gives us some error
+ * checking.  64 bit register reads should always work, but are inefficient
+ * on opteron (the northbridge always generates 2 separate HT 32 bit reads),
+ * so we use kreg32 wherever possible.  User register and counter register
+ * reads are always 32 bit reads, so only one form of those routines.
+ */
+
+/**
+ * qib_read_ureg32 - read 32-bit virtualized per-context register
+ * @dd: device
+ * @regno: register number
+ * @ctxt: context number
+ *
+ * Return the contents of a register that is virtualized to be per context.
+ * Returns -1 on errors (not distinguishable from valid contents at
+ * runtime; we may add a separate error variable at some point).
+ */
+static inline u32 qib_read_ureg32(const struct qib_devdata *dd,
+				  enum qib_ureg regno, int ctxt)
+{
+	if (!dd->kregbase || !(dd->flags & QIB_PRESENT))
+		return 0;
+
+	if (dd->userbase)
+		return readl(regno + (u64 __iomem *)
+			     ((char __iomem *)dd->userbase +
+			      dd->ureg_align * ctxt));
+	else
+		return readl(regno + (u64 __iomem *)
+			     (dd->uregbase +
+			      (char __iomem *)dd->kregbase +
+			      dd->ureg_align * ctxt));
+}
+
+/**
+ * qib_write_ureg - write 32-bit virtualized per-context register
+ * @dd: device
+ * @regno: register number
+ * @value: value
+ * @ctxt: context
+ *
+ * Write the contents of a register that is virtualized to be per context.
+ */
+static inline void qib_write_ureg(const struct qib_devdata *dd,
+				  enum qib_ureg regno, u64 value, int ctxt)
+{
+	u64 __iomem *ubase;
+	if (dd->userbase)
+		ubase = (u64 __iomem *)
+			((char __iomem *) dd->userbase +
+			 dd->ureg_align * ctxt);
+	else
+		ubase = (u64 __iomem *)
+			(dd->uregbase +
+			 (char __iomem *) dd->kregbase +
+			 dd->ureg_align * ctxt);
+
+	if (dd->kregbase && (dd->flags & QIB_PRESENT))
+		writeq(value, &ubase[regno]);
+}
+
+static inline u32 qib_read_kreg32(const struct qib_devdata *dd,
+				  const u16 regno)
+{
+	if (!dd->kregbase || !(dd->flags & QIB_PRESENT))
+		return -1;
+	return readl((u32 __iomem *)&dd->kregbase[regno]);
+}
+
+static inline u64 qib_read_kreg64(const struct qib_devdata *dd,
+				  const u16 regno)
+{
+	if (!dd->kregbase || !(dd->flags & QIB_PRESENT))
+		return -1;
+
+	return readq(&dd->kregbase[regno]);
+}
+
+static inline void qib_write_kreg(const struct qib_devdata *dd,
+				  const u16 regno, u64 value)
+{
+	if (dd->kregbase && (dd->flags & QIB_PRESENT))
+		writeq(value, &dd->kregbase[regno]);
+}
+
+/**
+ * qib_write_kreg_ctxt - write a device's per-ctxt 64-bit kernel register
+ * @dd: the qlogic_ib device
+ * @regno: the register number to write
+ * @ctxt: the context containing the register
+ * @value: the value to write
+ */
+static inline void qib_write_kreg_ctxt(const struct qib_devdata *dd,
+				       const u16 regno, unsigned ctxt,
+				       u64 value)
+{
+	qib_write_kreg(dd, regno + ctxt, value);
+}
+
+static inline void write_6120_creg(const struct qib_devdata *dd,
+				   u16 regno, u64 value)
+{
+	if (dd->cspec->cregbase && (dd->flags & QIB_PRESENT))
+		writeq(value, &dd->cspec->cregbase[regno]);
+}
+
+static inline u64 read_6120_creg(const struct qib_devdata *dd, u16 regno)
+{
+	if (!dd->cspec->cregbase || !(dd->flags & QIB_PRESENT))
+		return 0;
+	return readq(&dd->cspec->cregbase[regno]);
+}
+
+static inline u32 read_6120_creg32(const struct qib_devdata *dd, u16 regno)
+{
+	if (!dd->cspec->cregbase || !(dd->flags & QIB_PRESENT))
+		return 0;
+	return readl(&dd->cspec->cregbase[regno]);
+}
+
+/* kr_control bits */
+#define QLOGIC_IB_C_RESET 1U
+
+/* kr_intstatus, kr_intclear, kr_intmask bits */
+#define QLOGIC_IB_I_RCVURG_MASK ((1U << 5) - 1)
+#define QLOGIC_IB_I_RCVURG_SHIFT 0
+#define QLOGIC_IB_I_RCVAVAIL_MASK ((1U << 5) - 1)
+#define QLOGIC_IB_I_RCVAVAIL_SHIFT 12
+
+#define QLOGIC_IB_C_FREEZEMODE 0x00000002
+#define QLOGIC_IB_C_LINKENABLE 0x00000004
+#define QLOGIC_IB_I_ERROR               0x0000000080000000ULL
+#define QLOGIC_IB_I_SPIOSENT            0x0000000040000000ULL
+#define QLOGIC_IB_I_SPIOBUFAVAIL        0x0000000020000000ULL
+#define QLOGIC_IB_I_GPIO                0x0000000010000000ULL
+#define QLOGIC_IB_I_BITSEXTANT \
+		((QLOGIC_IB_I_RCVURG_MASK << QLOGIC_IB_I_RCVURG_SHIFT) | \
+		(QLOGIC_IB_I_RCVAVAIL_MASK << \
+		 QLOGIC_IB_I_RCVAVAIL_SHIFT) | \
+		QLOGIC_IB_I_ERROR | QLOGIC_IB_I_SPIOSENT | \
+		QLOGIC_IB_I_SPIOBUFAVAIL | QLOGIC_IB_I_GPIO)
+
+/* kr_hwerrclear, kr_hwerrmask, kr_hwerrstatus, bits */
+#define QLOGIC_IB_HWE_PCIEMEMPARITYERR_MASK  0x000000000000003fULL
+#define QLOGIC_IB_HWE_PCIEMEMPARITYERR_SHIFT 0
+#define QLOGIC_IB_HWE_PCIEPOISONEDTLP      0x0000000010000000ULL
+#define QLOGIC_IB_HWE_PCIECPLTIMEOUT       0x0000000020000000ULL
+#define QLOGIC_IB_HWE_PCIEBUSPARITYXTLH    0x0000000040000000ULL
+#define QLOGIC_IB_HWE_PCIEBUSPARITYXADM    0x0000000080000000ULL
+#define QLOGIC_IB_HWE_PCIEBUSPARITYRADM    0x0000000100000000ULL
+#define QLOGIC_IB_HWE_COREPLL_FBSLIP       0x0080000000000000ULL
+#define QLOGIC_IB_HWE_COREPLL_RFSLIP       0x0100000000000000ULL
+#define QLOGIC_IB_HWE_PCIE1PLLFAILED       0x0400000000000000ULL
+#define QLOGIC_IB_HWE_PCIE0PLLFAILED       0x0800000000000000ULL
+#define QLOGIC_IB_HWE_SERDESPLLFAILED      0x1000000000000000ULL
+
+
+/* kr_extstatus bits */
+#define QLOGIC_IB_EXTS_FREQSEL 0x2
+#define QLOGIC_IB_EXTS_SERDESSEL 0x4
+#define QLOGIC_IB_EXTS_MEMBIST_ENDTEST     0x0000000000004000
+#define QLOGIC_IB_EXTS_MEMBIST_FOUND       0x0000000000008000
+
+/* kr_xgxsconfig bits */
+#define QLOGIC_IB_XGXS_RESET          0x5ULL
+
+#define _QIB_GPIO_SDA_NUM 1
+#define _QIB_GPIO_SCL_NUM 0
+
+/* Bits in GPIO for the added IB link interrupts */
+#define GPIO_RXUVL_BIT 3
+#define GPIO_OVRUN_BIT 4
+#define GPIO_LLI_BIT 5
+#define GPIO_ERRINTR_MASK 0x38
+
+
+#define QLOGIC_IB_RT_BUFSIZE_MASK 0xe0000000ULL
+#define QLOGIC_IB_RT_BUFSIZE_SHIFTVAL(tid) \
+	((((tid) & QLOGIC_IB_RT_BUFSIZE_MASK) >> 29) + 11 - 1)
+#define QLOGIC_IB_RT_BUFSIZE(tid) (1 << QLOGIC_IB_RT_BUFSIZE_SHIFTVAL(tid))
+#define QLOGIC_IB_RT_IS_VALID(tid) \
+	(((tid) & QLOGIC_IB_RT_BUFSIZE_MASK) && \
+	 ((((tid) & QLOGIC_IB_RT_BUFSIZE_MASK) != QLOGIC_IB_RT_BUFSIZE_MASK)))
+#define QLOGIC_IB_RT_ADDR_MASK 0x1FFFFFFFULL /* 29 bits valid */
+#define QLOGIC_IB_RT_ADDR_SHIFT 10
+
+#define QLOGIC_IB_R_INTRAVAIL_SHIFT 16
+#define QLOGIC_IB_R_TAILUPD_SHIFT 31
+#define IBA6120_R_PKEY_DIS_SHIFT 30
+
+#define PBC_6120_VL15_SEND_CTRL (1ULL << 31) /* pbc; VL15; link_buf only */
+
+#define IBCBUSFRSPCPARITYERR HWE_MASK(IBCBusFromSPCParityErr)
+#define IBCBUSTOSPCPARITYERR HWE_MASK(IBCBusToSPCParityErr)
+
+#define SYM_MASK_BIT(regname, fldname, bit) ((u64) \
+	((1ULL << (SYM_LSB(regname, fldname) + (bit)))))
+
+#define TXEMEMPARITYERR_PIOBUF \
+	SYM_MASK_BIT(HwErrMask, TXEMemParityErrMask, 0)
+#define TXEMEMPARITYERR_PIOPBC \
+	SYM_MASK_BIT(HwErrMask, TXEMemParityErrMask, 1)
+#define TXEMEMPARITYERR_PIOLAUNCHFIFO \
+	SYM_MASK_BIT(HwErrMask, TXEMemParityErrMask, 2)
+
+#define RXEMEMPARITYERR_RCVBUF \
+	SYM_MASK_BIT(HwErrMask, RXEMemParityErrMask, 0)
+#define RXEMEMPARITYERR_LOOKUPQ \
+	SYM_MASK_BIT(HwErrMask, RXEMemParityErrMask, 1)
+#define RXEMEMPARITYERR_EXPTID \
+	SYM_MASK_BIT(HwErrMask, RXEMemParityErrMask, 2)
+#define RXEMEMPARITYERR_EAGERTID \
+	SYM_MASK_BIT(HwErrMask, RXEMemParityErrMask, 3)
+#define RXEMEMPARITYERR_FLAGBUF \
+	SYM_MASK_BIT(HwErrMask, RXEMemParityErrMask, 4)
+#define RXEMEMPARITYERR_DATAINFO \
+	SYM_MASK_BIT(HwErrMask, RXEMemParityErrMask, 5)
+#define RXEMEMPARITYERR_HDRINFO \
+	SYM_MASK_BIT(HwErrMask, RXEMemParityErrMask, 6)
+
+/* 6120 specific hardware errors... */
+static const struct qib_hwerror_msgs qib_6120_hwerror_msgs[] = {
+	/* generic hardware errors */
+	QLOGIC_IB_HWE_MSG(IBCBUSFRSPCPARITYERR, "QIB2IB Parity"),
+	QLOGIC_IB_HWE_MSG(IBCBUSTOSPCPARITYERR, "IB2QIB Parity"),
+
+	QLOGIC_IB_HWE_MSG(TXEMEMPARITYERR_PIOBUF,
+			  "TXE PIOBUF Memory Parity"),
+	QLOGIC_IB_HWE_MSG(TXEMEMPARITYERR_PIOPBC,
+			  "TXE PIOPBC Memory Parity"),
+	QLOGIC_IB_HWE_MSG(TXEMEMPARITYERR_PIOLAUNCHFIFO,
+			  "TXE PIOLAUNCHFIFO Memory Parity"),
+
+	QLOGIC_IB_HWE_MSG(RXEMEMPARITYERR_RCVBUF,
+			  "RXE RCVBUF Memory Parity"),
+	QLOGIC_IB_HWE_MSG(RXEMEMPARITYERR_LOOKUPQ,
+			  "RXE LOOKUPQ Memory Parity"),
+	QLOGIC_IB_HWE_MSG(RXEMEMPARITYERR_EAGERTID,
+			  "RXE EAGERTID Memory Parity"),
+	QLOGIC_IB_HWE_MSG(RXEMEMPARITYERR_EXPTID,
+			  "RXE EXPTID Memory Parity"),
+	QLOGIC_IB_HWE_MSG(RXEMEMPARITYERR_FLAGBUF,
+			  "RXE FLAGBUF Memory Parity"),
+	QLOGIC_IB_HWE_MSG(RXEMEMPARITYERR_DATAINFO,
+			  "RXE DATAINFO Memory Parity"),
+	QLOGIC_IB_HWE_MSG(RXEMEMPARITYERR_HDRINFO,
+			  "RXE HDRINFO Memory Parity"),
+
+	/* chip-specific hardware errors */
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIEPOISONEDTLP,
+			  "PCIe Poisoned TLP"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIECPLTIMEOUT,
+			  "PCIe completion timeout"),
+	/*
+	 * In practice, it's unlikely wthat we'll see PCIe PLL, or bus
+	 * parity or memory parity error failures, because most likely we
+	 * won't be able to talk to the core of the chip.  Nonetheless, we
+	 * might see them, if they are in parts of the PCIe core that aren't
+	 * essential.
+	 */
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIE1PLLFAILED,
+			  "PCIePLL1"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIE0PLLFAILED,
+			  "PCIePLL0"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIEBUSPARITYXTLH,
+			  "PCIe XTLH core parity"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIEBUSPARITYXADM,
+			  "PCIe ADM TX core parity"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIEBUSPARITYRADM,
+			  "PCIe ADM RX core parity"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_SERDESPLLFAILED,
+			  "SerDes PLL"),
+};
+
+#define TXE_PIO_PARITY (TXEMEMPARITYERR_PIOBUF | TXEMEMPARITYERR_PIOPBC)
+#define _QIB_PLL_FAIL (QLOGIC_IB_HWE_COREPLL_FBSLIP |   \
+		QLOGIC_IB_HWE_COREPLL_RFSLIP)
+
+	/* variables for sanity checking interrupt and errors */
+#define IB_HWE_BITSEXTANT \
+	(HWE_MASK(RXEMemParityErr) |					\
+	 HWE_MASK(TXEMemParityErr) |					\
+	 (QLOGIC_IB_HWE_PCIEMEMPARITYERR_MASK <<			\
+	  QLOGIC_IB_HWE_PCIEMEMPARITYERR_SHIFT) |			\
+	 QLOGIC_IB_HWE_PCIE1PLLFAILED |					\
+	 QLOGIC_IB_HWE_PCIE0PLLFAILED |					\
+	 QLOGIC_IB_HWE_PCIEPOISONEDTLP |				\
+	 QLOGIC_IB_HWE_PCIECPLTIMEOUT |					\
+	 QLOGIC_IB_HWE_PCIEBUSPARITYXTLH |				\
+	 QLOGIC_IB_HWE_PCIEBUSPARITYXADM |				\
+	 QLOGIC_IB_HWE_PCIEBUSPARITYRADM |				\
+	 HWE_MASK(PowerOnBISTFailed) |					\
+	 QLOGIC_IB_HWE_COREPLL_FBSLIP |					\
+	 QLOGIC_IB_HWE_COREPLL_RFSLIP |					\
+	 QLOGIC_IB_HWE_SERDESPLLFAILED |				\
+	 HWE_MASK(IBCBusToSPCParityErr) |				\
+	 HWE_MASK(IBCBusFromSPCParityErr))
+
+#define IB_E_BITSEXTANT \
+	(ERR_MASK(RcvFormatErr) | ERR_MASK(RcvVCRCErr) |		\
+	 ERR_MASK(RcvICRCErr) | ERR_MASK(RcvMinPktLenErr) |		\
+	 ERR_MASK(RcvMaxPktLenErr) | ERR_MASK(RcvLongPktLenErr) |	\
+	 ERR_MASK(RcvShortPktLenErr) | ERR_MASK(RcvUnexpectedCharErr) | \
+	 ERR_MASK(RcvUnsupportedVLErr) | ERR_MASK(RcvEBPErr) |		\
+	 ERR_MASK(RcvIBFlowErr) | ERR_MASK(RcvBadVersionErr) |		\
+	 ERR_MASK(RcvEgrFullErr) | ERR_MASK(RcvHdrFullErr) |		\
+	 ERR_MASK(RcvBadTidErr) | ERR_MASK(RcvHdrLenErr) |		\
+	 ERR_MASK(RcvHdrErr) | ERR_MASK(RcvIBLostLinkErr) |		\
+	 ERR_MASK(SendMinPktLenErr) | ERR_MASK(SendMaxPktLenErr) |	\
+	 ERR_MASK(SendUnderRunErr) | ERR_MASK(SendPktLenErr) |		\
+	 ERR_MASK(SendDroppedSmpPktErr) |				\
+	 ERR_MASK(SendDroppedDataPktErr) |				\
+	 ERR_MASK(SendPioArmLaunchErr) |				\
+	 ERR_MASK(SendUnexpectedPktNumErr) |				\
+	 ERR_MASK(SendUnsupportedVLErr) | ERR_MASK(IBStatusChanged) |	\
+	 ERR_MASK(InvalidAddrErr) | ERR_MASK(ResetNegated) |		\
+	 ERR_MASK(HardwareErr))
+
+#define QLOGIC_IB_E_PKTERRS ( \
+		ERR_MASK(SendPktLenErr) |				\
+		ERR_MASK(SendDroppedDataPktErr) |			\
+		ERR_MASK(RcvVCRCErr) |					\
+		ERR_MASK(RcvICRCErr) |					\
+		ERR_MASK(RcvShortPktLenErr) |				\
+		ERR_MASK(RcvEBPErr))
+
+/* These are all rcv-related errors which we want to count for stats */
+#define E_SUM_PKTERRS						\
+	(ERR_MASK(RcvHdrLenErr) | ERR_MASK(RcvBadTidErr) |		\
+	 ERR_MASK(RcvBadVersionErr) | ERR_MASK(RcvHdrErr) |		\
+	 ERR_MASK(RcvLongPktLenErr) | ERR_MASK(RcvShortPktLenErr) |	\
+	 ERR_MASK(RcvMaxPktLenErr) | ERR_MASK(RcvMinPktLenErr) |	\
+	 ERR_MASK(RcvFormatErr) | ERR_MASK(RcvUnsupportedVLErr) |	\
+	 ERR_MASK(RcvUnexpectedCharErr) | ERR_MASK(RcvEBPErr))
+
+/* These are all send-related errors which we want to count for stats */
+#define E_SUM_ERRS							\
+	(ERR_MASK(SendPioArmLaunchErr) |				\
+	 ERR_MASK(SendUnexpectedPktNumErr) |				\
+	 ERR_MASK(SendDroppedDataPktErr) |				\
+	 ERR_MASK(SendDroppedSmpPktErr) |				\
+	 ERR_MASK(SendMaxPktLenErr) | ERR_MASK(SendUnsupportedVLErr) |	\
+	 ERR_MASK(SendMinPktLenErr) | ERR_MASK(SendPktLenErr) |		\
+	 ERR_MASK(InvalidAddrErr))
+
+/*
+ * this is similar to E_SUM_ERRS, but can't ignore armlaunch, don't ignore
+ * errors not related to freeze and cancelling buffers.  Can't ignore
+ * armlaunch because could get more while still cleaning up, and need
+ * to cancel those as they happen.
+ */
+#define E_SPKT_ERRS_IGNORE \
+	(ERR_MASK(SendDroppedDataPktErr) |				\
+	 ERR_MASK(SendDroppedSmpPktErr) |				\
+	 ERR_MASK(SendMaxPktLenErr) | ERR_MASK(SendMinPktLenErr) |	\
+	 ERR_MASK(SendPktLenErr))
+
+/*
+ * these are errors that can occur when the link changes state while
+ * a packet is being sent or received.  This doesn't cover things
+ * like EBP or VCRC that can be the result of a sending having the
+ * link change state, so we receive a "known bad" packet.
+ */
+#define E_SUM_LINK_PKTERRS		\
+	(ERR_MASK(SendDroppedDataPktErr) |				\
+	 ERR_MASK(SendDroppedSmpPktErr) |				\
+	 ERR_MASK(SendMinPktLenErr) | ERR_MASK(SendPktLenErr) |		\
+	 ERR_MASK(RcvShortPktLenErr) | ERR_MASK(RcvMinPktLenErr) |	\
+	 ERR_MASK(RcvUnexpectedCharErr))
+
+static void qib_6120_put_tid_2(struct qib_devdata *, u64 __iomem *,
+			       u32, unsigned long);
+
+/*
+ * On platforms using this chip, and not having ordered WC stores, we
+ * can get TXE parity errors due to speculative reads to the PIO buffers,
+ * and this, due to a chip issue can result in (many) false parity error
+ * reports.  So it's a debug print on those, and an info print on systems
+ * where the speculative reads don't occur.
+ */
+static void qib_6120_txe_recover(struct qib_devdata *dd)
+{
+	if (!qib_unordered_wc())
+		qib_devinfo(dd->pcidev,
+			    "Recovering from TXE PIO parity error\n");
+}
+
+/* enable/disable chip from delivering interrupts */
+static void qib_6120_set_intr_state(struct qib_devdata *dd, u32 enable)
+{
+	if (enable) {
+		if (dd->flags & QIB_BADINTR)
+			return;
+		qib_write_kreg(dd, kr_intmask, ~0ULL);
+		/* force re-interrupt of any pending interrupts. */
+		qib_write_kreg(dd, kr_intclear, 0ULL);
+	} else
+		qib_write_kreg(dd, kr_intmask, 0ULL);
+}
+
+/*
+ * Try to cleanup as much as possible for anything that might have gone
+ * wrong while in freeze mode, such as pio buffers being written by user
+ * processes (causing armlaunch), send errors due to going into freeze mode,
+ * etc., and try to avoid causing extra interrupts while doing so.
+ * Forcibly update the in-memory pioavail register copies after cleanup
+ * because the chip won't do it while in freeze mode (the register values
+ * themselves are kept correct).
+ * Make sure that we don't lose any important interrupts by using the chip
+ * feature that says that writing 0 to a bit in *clear that is set in
+ * *status will cause an interrupt to be generated again (if allowed by
+ * the *mask value).
+ * This is in chip-specific code because of all of the register accesses,
+ * even though the details are similar on most chips
+ */
+static void qib_6120_clear_freeze(struct qib_devdata *dd)
+{
+	/* disable error interrupts, to avoid confusion */
+	qib_write_kreg(dd, kr_errmask, 0ULL);
+
+	/* also disable interrupts; errormask is sometimes overwriten */
+	qib_6120_set_intr_state(dd, 0);
+
+	qib_cancel_sends(dd->pport);
+
+	/* clear the freeze, and be sure chip saw it */
+	qib_write_kreg(dd, kr_control, dd->control);
+	qib_read_kreg32(dd, kr_scratch);
+
+	/* force in-memory update now we are out of freeze */
+	qib_force_pio_avail_update(dd);
+
+	/*
+	 * force new interrupt if any hwerr, error or interrupt bits are
+	 * still set, and clear "safe" send packet errors related to freeze
+	 * and cancelling sends.  Re-enable error interrupts before possible
+	 * force of re-interrupt on pending interrupts.
+	 */
+	qib_write_kreg(dd, kr_hwerrclear, 0ULL);
+	qib_write_kreg(dd, kr_errclear, E_SPKT_ERRS_IGNORE);
+	qib_write_kreg(dd, kr_errmask, dd->cspec->errormask);
+	qib_6120_set_intr_state(dd, 1);
+}
+
+/**
+ * qib_handle_6120_hwerrors - display hardware errors.
+ * @dd: the qlogic_ib device
+ * @msg: the output buffer
+ * @msgl: the size of the output buffer
+ *
+ * Use same msg buffer as regular errors to avoid excessive stack
+ * use.  Most hardware errors are catastrophic, but for right now,
+ * we'll print them and continue.  Reuse the same message buffer as
+ * handle_6120_errors() to avoid excessive stack usage.
+ */
+static void qib_handle_6120_hwerrors(struct qib_devdata *dd, char *msg,
+				     size_t msgl)
+{
+	u64 hwerrs;
+	u32 bits, ctrl;
+	int isfatal = 0;
+	char *bitsmsg;
+	int log_idx;
+
+	hwerrs = qib_read_kreg64(dd, kr_hwerrstatus);
+	if (!hwerrs)
+		return;
+	if (hwerrs == ~0ULL) {
+		qib_dev_err(dd, "Read of hardware error status failed "
+			    "(all bits set); ignoring\n");
+		return;
+	}
+	qib_stats.sps_hwerrs++;
+
+	/* Always clear the error status register, except MEMBISTFAIL,
+	 * regardless of whether we continue or stop using the chip.
+	 * We want that set so we know it failed, even across driver reload.
+	 * We'll still ignore it in the hwerrmask.  We do this partly for
+	 * diagnostics, but also for support */
+	qib_write_kreg(dd, kr_hwerrclear,
+		       hwerrs & ~HWE_MASK(PowerOnBISTFailed));
+
+	hwerrs &= dd->cspec->hwerrmask;
+
+	/* We log some errors to EEPROM, check if we have any of those. */
+	for (log_idx = 0; log_idx < QIB_EEP_LOG_CNT; ++log_idx)
+		if (hwerrs & dd->eep_st_masks[log_idx].hwerrs_to_log)
+			qib_inc_eeprom_err(dd, log_idx, 1);
+
+	/*
+	 * Make sure we get this much out, unless told to be quiet,
+	 * or it's occurred within the last 5 seconds.
+	 */
+	if (hwerrs & ~(TXE_PIO_PARITY | RXEMEMPARITYERR_EAGERTID))
+		qib_devinfo(dd->pcidev, "Hardware error: hwerr=0x%llx "
+			 "(cleared)\n", (unsigned long long) hwerrs);
+
+	if (hwerrs & ~IB_HWE_BITSEXTANT)
+		qib_dev_err(dd, "hwerror interrupt with unknown errors "
+			    "%llx set\n", (unsigned long long)
+			    (hwerrs & ~IB_HWE_BITSEXTANT));
+
+	ctrl = qib_read_kreg32(dd, kr_control);
+	if ((ctrl & QLOGIC_IB_C_FREEZEMODE) && !dd->diag_client) {
+		/*
+		 * Parity errors in send memory are recoverable,
+		 * just cancel the send (if indicated in * sendbuffererror),
+		 * count the occurrence, unfreeze (if no other handled
+		 * hardware error bits are set), and continue. They can
+		 * occur if a processor speculative read is done to the PIO
+		 * buffer while we are sending a packet, for example.
+		 */
+		if (hwerrs & TXE_PIO_PARITY) {
+			qib_6120_txe_recover(dd);
+			hwerrs &= ~TXE_PIO_PARITY;
+		}
+
+		if (!hwerrs) {
+			static u32 freeze_cnt;
+
+			freeze_cnt++;
+			qib_6120_clear_freeze(dd);
+		} else
+			isfatal = 1;
+	}
+
+	*msg = '\0';
+
+	if (hwerrs & HWE_MASK(PowerOnBISTFailed)) {
+		isfatal = 1;
+		strlcat(msg, "[Memory BIST test failed, InfiniPath hardware"
+			" unusable]", msgl);
+		/* ignore from now on, so disable until driver reloaded */
+		dd->cspec->hwerrmask &= ~HWE_MASK(PowerOnBISTFailed);
+		qib_write_kreg(dd, kr_hwerrmask, dd->cspec->hwerrmask);
+	}
+
+	qib_format_hwerrors(hwerrs, qib_6120_hwerror_msgs,
+			    ARRAY_SIZE(qib_6120_hwerror_msgs), msg, msgl);
+
+	bitsmsg = dd->cspec->bitsmsgbuf;
+	if (hwerrs & (QLOGIC_IB_HWE_PCIEMEMPARITYERR_MASK <<
+		      QLOGIC_IB_HWE_PCIEMEMPARITYERR_SHIFT)) {
+		bits = (u32) ((hwerrs >>
+			       QLOGIC_IB_HWE_PCIEMEMPARITYERR_SHIFT) &
+			      QLOGIC_IB_HWE_PCIEMEMPARITYERR_MASK);
+		snprintf(bitsmsg, sizeof dd->cspec->bitsmsgbuf,
+			 "[PCIe Mem Parity Errs %x] ", bits);
+		strlcat(msg, bitsmsg, msgl);
+	}
+
+	if (hwerrs & _QIB_PLL_FAIL) {
+		isfatal = 1;
+		snprintf(bitsmsg, sizeof dd->cspec->bitsmsgbuf,
+			 "[PLL failed (%llx), InfiniPath hardware unusable]",
+			 (unsigned long long) hwerrs & _QIB_PLL_FAIL);
+		strlcat(msg, bitsmsg, msgl);
+		/* ignore from now on, so disable until driver reloaded */
+		dd->cspec->hwerrmask &= ~(hwerrs & _QIB_PLL_FAIL);
+		qib_write_kreg(dd, kr_hwerrmask, dd->cspec->hwerrmask);
+	}
+
+	if (hwerrs & QLOGIC_IB_HWE_SERDESPLLFAILED) {
+		/*
+		 * If it occurs, it is left masked since the external
+		 * interface is unused
+		 */
+		dd->cspec->hwerrmask &= ~QLOGIC_IB_HWE_SERDESPLLFAILED;
+		qib_write_kreg(dd, kr_hwerrmask, dd->cspec->hwerrmask);
+	}
+
+	if (hwerrs)
+		/*
+		 * if any set that we aren't ignoring; only
+		 * make the complaint once, in case it's stuck
+		 * or recurring, and we get here multiple
+		 * times.
+		 */
+		qib_dev_err(dd, "%s hardware error\n", msg);
+	else
+		*msg = 0; /* recovered from all of them */
+
+	if (isfatal && !dd->diag_client) {
+		qib_dev_err(dd, "Fatal Hardware Error, no longer"
+			    " usable, SN %.16s\n", dd->serial);
+		/*
+		 * for /sys status file and user programs to print; if no
+		 * trailing brace is copied, we'll know it was truncated.
+		 */
+		if (dd->freezemsg)
+			snprintf(dd->freezemsg, dd->freezelen,
+				 "{%s}", msg);
+		qib_disable_after_error(dd);
+	}
+}
+
+/*
+ * Decode the error status into strings, deciding whether to always
+ * print * it or not depending on "normal packet errors" vs everything
+ * else.   Return 1 if "real" errors, otherwise 0 if only packet
+ * errors, so caller can decide what to print with the string.
+ */
+static int qib_decode_6120_err(struct qib_devdata *dd, char *buf, size_t blen,
+			       u64 err)
+{
+	int iserr = 1;
+
+	*buf = '\0';
+	if (err & QLOGIC_IB_E_PKTERRS) {
+		if (!(err & ~QLOGIC_IB_E_PKTERRS))
+			iserr = 0;
+		if ((err & ERR_MASK(RcvICRCErr)) &&
+		    !(err&(ERR_MASK(RcvVCRCErr)|ERR_MASK(RcvEBPErr))))
+			strlcat(buf, "CRC ", blen);
+		if (!iserr)
+			goto done;
+	}
+	if (err & ERR_MASK(RcvHdrLenErr))
+		strlcat(buf, "rhdrlen ", blen);
+	if (err & ERR_MASK(RcvBadTidErr))
+		strlcat(buf, "rbadtid ", blen);
+	if (err & ERR_MASK(RcvBadVersionErr))
+		strlcat(buf, "rbadversion ", blen);
+	if (err & ERR_MASK(RcvHdrErr))
+		strlcat(buf, "rhdr ", blen);
+	if (err & ERR_MASK(RcvLongPktLenErr))
+		strlcat(buf, "rlongpktlen ", blen);
+	if (err & ERR_MASK(RcvMaxPktLenErr))
+		strlcat(buf, "rmaxpktlen ", blen);
+	if (err & ERR_MASK(RcvMinPktLenErr))
+		strlcat(buf, "rminpktlen ", blen);
+	if (err & ERR_MASK(SendMinPktLenErr))
+		strlcat(buf, "sminpktlen ", blen);
+	if (err & ERR_MASK(RcvFormatErr))
+		strlcat(buf, "rformaterr ", blen);
+	if (err & ERR_MASK(RcvUnsupportedVLErr))
+		strlcat(buf, "runsupvl ", blen);
+	if (err & ERR_MASK(RcvUnexpectedCharErr))
+		strlcat(buf, "runexpchar ", blen);
+	if (err & ERR_MASK(RcvIBFlowErr))
+		strlcat(buf, "ribflow ", blen);
+	if (err & ERR_MASK(SendUnderRunErr))
+		strlcat(buf, "sunderrun ", blen);
+	if (err & ERR_MASK(SendPioArmLaunchErr))
+		strlcat(buf, "spioarmlaunch ", blen);
+	if (err & ERR_MASK(SendUnexpectedPktNumErr))
+		strlcat(buf, "sunexperrpktnum ", blen);
+	if (err & ERR_MASK(SendDroppedSmpPktErr))
+		strlcat(buf, "sdroppedsmppkt ", blen);
+	if (err & ERR_MASK(SendMaxPktLenErr))
+		strlcat(buf, "smaxpktlen ", blen);
+	if (err & ERR_MASK(SendUnsupportedVLErr))
+		strlcat(buf, "sunsupVL ", blen);
+	if (err & ERR_MASK(InvalidAddrErr))
+		strlcat(buf, "invalidaddr ", blen);
+	if (err & ERR_MASK(RcvEgrFullErr))
+		strlcat(buf, "rcvegrfull ", blen);
+	if (err & ERR_MASK(RcvHdrFullErr))
+		strlcat(buf, "rcvhdrfull ", blen);
+	if (err & ERR_MASK(IBStatusChanged))
+		strlcat(buf, "ibcstatuschg ", blen);
+	if (err & ERR_MASK(RcvIBLostLinkErr))
+		strlcat(buf, "riblostlink ", blen);
+	if (err & ERR_MASK(HardwareErr))
+		strlcat(buf, "hardware ", blen);
+	if (err & ERR_MASK(ResetNegated))
+		strlcat(buf, "reset ", blen);
+done:
+	return iserr;
+}
+
+/*
+ * Called when we might have an error that is specific to a particular
+ * PIO buffer, and may need to cancel that buffer, so it can be re-used.
+ */
+static void qib_disarm_6120_senderrbufs(struct qib_pportdata *ppd)
+{
+	unsigned long sbuf[2];
+	struct qib_devdata *dd = ppd->dd;
+
+	/*
+	 * It's possible that sendbuffererror could have bits set; might
+	 * have already done this as a result of hardware error handling.
+	 */
+	sbuf[0] = qib_read_kreg64(dd, kr_sendbuffererror);
+	sbuf[1] = qib_read_kreg64(dd, kr_sendbuffererror + 1);
+
+	if (sbuf[0] || sbuf[1])
+		qib_disarm_piobufs_set(dd, sbuf,
+				       dd->piobcnt2k + dd->piobcnt4k);
+}
+
+static int chk_6120_linkrecovery(struct qib_devdata *dd, u64 ibcs)
+{
+	int ret = 1;
+	u32 ibstate = qib_6120_iblink_state(ibcs);
+	u32 linkrecov = read_6120_creg32(dd, cr_iblinkerrrecov);
+
+	if (linkrecov != dd->cspec->lastlinkrecov) {
+		/* and no more until active again */
+		dd->cspec->lastlinkrecov = 0;
+		qib_set_linkstate(dd->pport, QIB_IB_LINKDOWN);
+		ret = 0;
+	}
+	if (ibstate == IB_PORT_ACTIVE)
+		dd->cspec->lastlinkrecov =
+			read_6120_creg32(dd, cr_iblinkerrrecov);
+	return ret;
+}
+
+static void handle_6120_errors(struct qib_devdata *dd, u64 errs)
+{
+	char *msg;
+	u64 ignore_this_time = 0;
+	u64 iserr = 0;
+	int log_idx;
+	struct qib_pportdata *ppd = dd->pport;
+	u64 mask;
+
+	/* don't report errors that are masked */
+	errs &= dd->cspec->errormask;
+	msg = dd->cspec->emsgbuf;
+
+	/* do these first, they are most important */
+	if (errs & ERR_MASK(HardwareErr))
+		qib_handle_6120_hwerrors(dd, msg, sizeof dd->cspec->emsgbuf);
+	else
+		for (log_idx = 0; log_idx < QIB_EEP_LOG_CNT; ++log_idx)
+			if (errs & dd->eep_st_masks[log_idx].errs_to_log)
+				qib_inc_eeprom_err(dd, log_idx, 1);
+
+	if (errs & ~IB_E_BITSEXTANT)
+		qib_dev_err(dd, "error interrupt with unknown errors "
+			    "%llx set\n",
+			    (unsigned long long) (errs & ~IB_E_BITSEXTANT));
+
+	if (errs & E_SUM_ERRS) {
+		qib_disarm_6120_senderrbufs(ppd);
+		if ((errs & E_SUM_LINK_PKTERRS) &&
+		    !(ppd->lflags & QIBL_LINKACTIVE)) {
+			/*
+			 * This can happen when trying to bring the link
+			 * up, but the IB link changes state at the "wrong"
+			 * time. The IB logic then complains that the packet
+			 * isn't valid.  We don't want to confuse people, so
+			 * we just don't print them, except at debug
+			 */
+			ignore_this_time = errs & E_SUM_LINK_PKTERRS;
+		}
+	} else if ((errs & E_SUM_LINK_PKTERRS) &&
+		   !(ppd->lflags & QIBL_LINKACTIVE)) {
+		/*
+		 * This can happen when SMA is trying to bring the link
+		 * up, but the IB link changes state at the "wrong" time.
+		 * The IB logic then complains that the packet isn't
+		 * valid.  We don't want to confuse people, so we just
+		 * don't print them, except at debug
+		 */
+		ignore_this_time = errs & E_SUM_LINK_PKTERRS;
+	}
+
+	qib_write_kreg(dd, kr_errclear, errs);
+
+	errs &= ~ignore_this_time;
+	if (!errs)
+		goto done;
+
+	/*
+	 * The ones we mask off are handled specially below
+	 * or above.
+	 */
+	mask = ERR_MASK(IBStatusChanged) | ERR_MASK(RcvEgrFullErr) |
+		ERR_MASK(RcvHdrFullErr) | ERR_MASK(HardwareErr);
+	qib_decode_6120_err(dd, msg, sizeof dd->cspec->emsgbuf, errs & ~mask);
+
+	if (errs & E_SUM_PKTERRS)
+		qib_stats.sps_rcverrs++;
+	if (errs & E_SUM_ERRS)
+		qib_stats.sps_txerrs++;
+
+	iserr = errs & ~(E_SUM_PKTERRS | QLOGIC_IB_E_PKTERRS);
+
+	if (errs & ERR_MASK(IBStatusChanged)) {
+		u64 ibcs = qib_read_kreg64(dd, kr_ibcstatus);
+		u32 ibstate = qib_6120_iblink_state(ibcs);
+		int handle = 1;
+
+		if (ibstate != IB_PORT_INIT && dd->cspec->lastlinkrecov)
+			handle = chk_6120_linkrecovery(dd, ibcs);
+		/*
+		 * Since going into a recovery state causes the link state
+		 * to go down and since recovery is transitory, it is better
+		 * if we "miss" ever seeing the link training state go into
+		 * recovery (i.e., ignore this transition for link state
+		 * special handling purposes) without updating lastibcstat.
+		 */
+		if (handle && qib_6120_phys_portstate(ibcs) ==
+					    IB_PHYSPORTSTATE_LINK_ERR_RECOVER)
+			handle = 0;
+		if (handle)
+			qib_handle_e_ibstatuschanged(ppd, ibcs);
+	}
+
+	if (errs & ERR_MASK(ResetNegated)) {
+		qib_dev_err(dd, "Got reset, requires re-init "
+			      "(unload and reload driver)\n");
+		dd->flags &= ~QIB_INITTED;  /* needs re-init */
+		/* mark as having had error */
+		*dd->devstatusp |= QIB_STATUS_HWERROR;
+		*dd->pport->statusp &= ~QIB_STATUS_IB_CONF;
+	}
+
+	if (*msg && iserr)
+		qib_dev_porterr(dd, ppd->port, "%s error\n", msg);
+
+	if (ppd->state_wanted & ppd->lflags)
+		wake_up_interruptible(&ppd->state_wait);
+
+	/*
+	 * If there were hdrq or egrfull errors, wake up any processes
+	 * waiting in poll.  We used to try to check which contexts had
+	 * the overflow, but given the cost of that and the chip reads
+	 * to support it, it's better to just wake everybody up if we
+	 * get an overflow; waiters can poll again if it's not them.
+	 */
+	if (errs & (ERR_MASK(RcvEgrFullErr) | ERR_MASK(RcvHdrFullErr))) {
+		qib_handle_urcv(dd, ~0U);
+		if (errs & ERR_MASK(RcvEgrFullErr))
+			qib_stats.sps_buffull++;
+		else
+			qib_stats.sps_hdrfull++;
+	}
+done:
+	return;
+}
+
+/**
+ * qib_6120_init_hwerrors - enable hardware errors
+ * @dd: the qlogic_ib device
+ *
+ * now that we have finished initializing everything that might reasonably
+ * cause a hardware error, and cleared those errors bits as they occur,
+ * we can enable hardware errors in the mask (potentially enabling
+ * freeze mode), and enable hardware errors as errors (along with
+ * everything else) in errormask
+ */
+static void qib_6120_init_hwerrors(struct qib_devdata *dd)
+{
+	u64 val;
+	u64 extsval;
+
+	extsval = qib_read_kreg64(dd, kr_extstatus);
+
+	if (!(extsval & QLOGIC_IB_EXTS_MEMBIST_ENDTEST))
+		qib_dev_err(dd, "MemBIST did not complete!\n");
+
+	/* init so all hwerrors interrupt, and enter freeze, ajdust below */
+	val = ~0ULL;
+	if (dd->minrev < 2) {
+		/*
+		 * Avoid problem with internal interface bus parity
+		 * checking. Fixed in Rev2.
+		 */
+		val &= ~QLOGIC_IB_HWE_PCIEBUSPARITYRADM;
+	}
+	/* avoid some intel cpu's speculative read freeze mode issue */
+	val &= ~TXEMEMPARITYERR_PIOBUF;
+
+	dd->cspec->hwerrmask = val;
+
+	qib_write_kreg(dd, kr_hwerrclear, ~HWE_MASK(PowerOnBISTFailed));
+	qib_write_kreg(dd, kr_hwerrmask, dd->cspec->hwerrmask);
+
+	/* clear all */
+	qib_write_kreg(dd, kr_errclear, ~0ULL);
+	/* enable errors that are masked, at least this first time. */
+	qib_write_kreg(dd, kr_errmask, ~0ULL);
+	dd->cspec->errormask = qib_read_kreg64(dd, kr_errmask);
+	/* clear any interrupts up to this point (ints still not enabled) */
+	qib_write_kreg(dd, kr_intclear, ~0ULL);
+
+	qib_write_kreg(dd, kr_rcvbthqp,
+		       dd->qpn_mask << (QIB_6120_RcvBTHQP_BTHQP_Mask_LSB - 1) |
+		       QIB_KD_QP);
+}
+
+/*
+ * Disable and enable the armlaunch error.  Used for PIO bandwidth testing
+ * on chips that are count-based, rather than trigger-based.  There is no
+ * reference counting, but that's also fine, given the intended use.
+ * Only chip-specific because it's all register accesses
+ */
+static void qib_set_6120_armlaunch(struct qib_devdata *dd, u32 enable)
+{
+	if (enable) {
+		qib_write_kreg(dd, kr_errclear,
+			       ERR_MASK(SendPioArmLaunchErr));
+		dd->cspec->errormask |= ERR_MASK(SendPioArmLaunchErr);
+	} else
+		dd->cspec->errormask &= ~ERR_MASK(SendPioArmLaunchErr);
+	qib_write_kreg(dd, kr_errmask, dd->cspec->errormask);
+}
+
+/*
+ * Formerly took parameter <which> in pre-shifted,
+ * pre-merged form with LinkCmd and LinkInitCmd
+ * together, and assuming the zero was NOP.
+ */
+static void qib_set_ib_6120_lstate(struct qib_pportdata *ppd, u16 linkcmd,
+				   u16 linitcmd)
+{
+	u64 mod_wd;
+	struct qib_devdata *dd = ppd->dd;
+	unsigned long flags;
+
+	if (linitcmd == QLOGIC_IB_IBCC_LINKINITCMD_DISABLE) {
+		/*
+		 * If we are told to disable, note that so link-recovery
+		 * code does not attempt to bring us back up.
+		 */
+		spin_lock_irqsave(&ppd->lflags_lock, flags);
+		ppd->lflags |= QIBL_IB_LINK_DISABLED;
+		spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+	} else if (linitcmd || linkcmd == QLOGIC_IB_IBCC_LINKCMD_DOWN) {
+		/*
+		 * Any other linkinitcmd will lead to LINKDOWN and then
+		 * to INIT (if all is well), so clear flag to let
+		 * link-recovery code attempt to bring us back up.
+		 */
+		spin_lock_irqsave(&ppd->lflags_lock, flags);
+		ppd->lflags &= ~QIBL_IB_LINK_DISABLED;
+		spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+	}
+
+	mod_wd = (linkcmd << QLOGIC_IB_IBCC_LINKCMD_SHIFT) |
+		(linitcmd << QLOGIC_IB_IBCC_LINKINITCMD_SHIFT);
+
+	qib_write_kreg(dd, kr_ibcctrl, dd->cspec->ibcctrl | mod_wd);
+	/* write to chip to prevent back-to-back writes of control reg */
+	qib_write_kreg(dd, kr_scratch, 0);
+}
+
+/**
+ * qib_6120_bringup_serdes - bring up the serdes
+ * @dd: the qlogic_ib device
+ */
+static int qib_6120_bringup_serdes(struct qib_pportdata *ppd)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u64 val, config1, prev_val, hwstat, ibc;
+
+	/* Put IBC in reset, sends disabled */
+	dd->control &= ~QLOGIC_IB_C_LINKENABLE;
+	qib_write_kreg(dd, kr_control, 0ULL);
+
+	dd->cspec->ibdeltainprog = 1;
+	dd->cspec->ibsymsnap = read_6120_creg32(dd, cr_ibsymbolerr);
+	dd->cspec->iblnkerrsnap = read_6120_creg32(dd, cr_iblinkerrrecov);
+
+	/* flowcontrolwatermark is in units of KBytes */
+	ibc = 0x5ULL << SYM_LSB(IBCCtrl, FlowCtrlWaterMark);
+	/*
+	 * How often flowctrl sent.  More or less in usecs; balance against
+	 * watermark value, so that in theory senders always get a flow
+	 * control update in time to not let the IB link go idle.
+	 */
+	ibc |= 0x3ULL << SYM_LSB(IBCCtrl, FlowCtrlPeriod);
+	/* max error tolerance */
+	dd->cspec->lli_thresh = 0xf;
+	ibc |= (u64) dd->cspec->lli_thresh << SYM_LSB(IBCCtrl, PhyerrThreshold);
+	/* use "real" buffer space for */
+	ibc |= 4ULL << SYM_LSB(IBCCtrl, CreditScale);
+	/* IB credit flow control. */
+	ibc |= 0xfULL << SYM_LSB(IBCCtrl, OverrunThreshold);
+	/*
+	 * set initial max size pkt IBC will send, including ICRC; it's the
+	 * PIO buffer size in dwords, less 1; also see qib_set_mtu()
+	 */
+	ibc |= ((u64)(ppd->ibmaxlen >> 2) + 1) << SYM_LSB(IBCCtrl, MaxPktLen);
+	dd->cspec->ibcctrl = ibc; /* without linkcmd or linkinitcmd! */
+
+	/* initially come up waiting for TS1, without sending anything. */
+	val = dd->cspec->ibcctrl | (QLOGIC_IB_IBCC_LINKINITCMD_DISABLE <<
+		QLOGIC_IB_IBCC_LINKINITCMD_SHIFT);
+	qib_write_kreg(dd, kr_ibcctrl, val);
+
+	val = qib_read_kreg64(dd, kr_serdes_cfg0);
+	config1 = qib_read_kreg64(dd, kr_serdes_cfg1);
+
+	/*
+	 * Force reset on, also set rxdetect enable.  Must do before reading
+	 * serdesstatus at least for simulation, or some of the bits in
+	 * serdes status will come back as undefined and cause simulation
+	 * failures
+	 */
+	val |= SYM_MASK(SerdesCfg0, ResetPLL) |
+		SYM_MASK(SerdesCfg0, RxDetEnX) |
+		(SYM_MASK(SerdesCfg0, L1PwrDnA) |
+		 SYM_MASK(SerdesCfg0, L1PwrDnB) |
+		 SYM_MASK(SerdesCfg0, L1PwrDnC) |
+		 SYM_MASK(SerdesCfg0, L1PwrDnD));
+	qib_write_kreg(dd, kr_serdes_cfg0, val);
+	/* be sure chip saw it */
+	qib_read_kreg64(dd, kr_scratch);
+	udelay(5);              /* need pll reset set at least for a bit */
+	/*
+	 * after PLL is reset, set the per-lane Resets and TxIdle and
+	 * clear the PLL reset and rxdetect (to get falling edge).
+	 * Leave L1PWR bits set (permanently)
+	 */
+	val &= ~(SYM_MASK(SerdesCfg0, RxDetEnX) |
+		 SYM_MASK(SerdesCfg0, ResetPLL) |
+		 (SYM_MASK(SerdesCfg0, L1PwrDnA) |
+		  SYM_MASK(SerdesCfg0, L1PwrDnB) |
+		  SYM_MASK(SerdesCfg0, L1PwrDnC) |
+		  SYM_MASK(SerdesCfg0, L1PwrDnD)));
+	val |= (SYM_MASK(SerdesCfg0, ResetA) |
+		SYM_MASK(SerdesCfg0, ResetB) |
+		SYM_MASK(SerdesCfg0, ResetC) |
+		SYM_MASK(SerdesCfg0, ResetD)) |
+		SYM_MASK(SerdesCfg0, TxIdeEnX);
+	qib_write_kreg(dd, kr_serdes_cfg0, val);
+	/* be sure chip saw it */
+	(void) qib_read_kreg64(dd, kr_scratch);
+	/* need PLL reset clear for at least 11 usec before lane
+	 * resets cleared; give it a few more to be sure */
+	udelay(15);
+	val &= ~((SYM_MASK(SerdesCfg0, ResetA) |
+		  SYM_MASK(SerdesCfg0, ResetB) |
+		  SYM_MASK(SerdesCfg0, ResetC) |
+		  SYM_MASK(SerdesCfg0, ResetD)) |
+		 SYM_MASK(SerdesCfg0, TxIdeEnX));
+
+	qib_write_kreg(dd, kr_serdes_cfg0, val);
+	/* be sure chip saw it */
+	(void) qib_read_kreg64(dd, kr_scratch);
+
+	val = qib_read_kreg64(dd, kr_xgxs_cfg);
+	prev_val = val;
+	if (val & QLOGIC_IB_XGXS_RESET)
+		val &= ~QLOGIC_IB_XGXS_RESET;
+	if (SYM_FIELD(val, XGXSCfg, polarity_inv) != ppd->rx_pol_inv) {
+		/* need to compensate for Tx inversion in partner */
+		val &= ~SYM_MASK(XGXSCfg, polarity_inv);
+		val |= (u64)ppd->rx_pol_inv << SYM_LSB(XGXSCfg, polarity_inv);
+	}
+	if (val != prev_val)
+		qib_write_kreg(dd, kr_xgxs_cfg, val);
+
+	val = qib_read_kreg64(dd, kr_serdes_cfg0);
+
+	/* clear current and de-emphasis bits */
+	config1 &= ~0x0ffffffff00ULL;
+	/* set current to 20ma */
+	config1 |= 0x00000000000ULL;
+	/* set de-emphasis to -5.68dB */
+	config1 |= 0x0cccc000000ULL;
+	qib_write_kreg(dd, kr_serdes_cfg1, config1);
+
+	/* base and port guid same for single port */
+	ppd->guid = dd->base_guid;
+
+	/*
+	 * the process of setting and un-resetting the serdes normally
+	 * causes a serdes PLL error, so check for that and clear it
+	 * here.  Also clearr hwerr bit in errstatus, but not others.
+	 */
+	hwstat = qib_read_kreg64(dd, kr_hwerrstatus);
+	if (hwstat) {
+		/* should just have PLL, clear all set, in an case */
+		if (hwstat & ~QLOGIC_IB_HWE_SERDESPLLFAILED)
+			qib_write_kreg(dd, kr_hwerrclear, hwstat);
+		qib_write_kreg(dd, kr_errclear, ERR_MASK(HardwareErr));
+	}
+
+	dd->control |= QLOGIC_IB_C_LINKENABLE;
+	dd->control &= ~QLOGIC_IB_C_FREEZEMODE;
+	qib_write_kreg(dd, kr_control, dd->control);
+
+	return 0;
+}
+
+/**
+ * qib_6120_quiet_serdes - set serdes to txidle
+ * @ppd: physical port of the qlogic_ib device
+ * Called when driver is being unloaded
+ */
+static void qib_6120_quiet_serdes(struct qib_pportdata *ppd)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u64 val;
+
+	qib_set_ib_6120_lstate(ppd, 0, QLOGIC_IB_IBCC_LINKINITCMD_DISABLE);
+
+	/* disable IBC */
+	dd->control &= ~QLOGIC_IB_C_LINKENABLE;
+	qib_write_kreg(dd, kr_control,
+		       dd->control | QLOGIC_IB_C_FREEZEMODE);
+
+	if (dd->cspec->ibsymdelta || dd->cspec->iblnkerrdelta ||
+	    dd->cspec->ibdeltainprog) {
+		u64 diagc;
+
+		/* enable counter writes */
+		diagc = qib_read_kreg64(dd, kr_hwdiagctrl);
+		qib_write_kreg(dd, kr_hwdiagctrl,
+			       diagc | SYM_MASK(HwDiagCtrl, CounterWrEnable));
+
+		if (dd->cspec->ibsymdelta || dd->cspec->ibdeltainprog) {
+			val = read_6120_creg32(dd, cr_ibsymbolerr);
+			if (dd->cspec->ibdeltainprog)
+				val -= val - dd->cspec->ibsymsnap;
+			val -= dd->cspec->ibsymdelta;
+			write_6120_creg(dd, cr_ibsymbolerr, val);
+		}
+		if (dd->cspec->iblnkerrdelta || dd->cspec->ibdeltainprog) {
+			val = read_6120_creg32(dd, cr_iblinkerrrecov);
+			if (dd->cspec->ibdeltainprog)
+				val -= val - dd->cspec->iblnkerrsnap;
+			val -= dd->cspec->iblnkerrdelta;
+			write_6120_creg(dd, cr_iblinkerrrecov, val);
+		}
+
+		/* and disable counter writes */
+		qib_write_kreg(dd, kr_hwdiagctrl, diagc);
+	}
+
+	val = qib_read_kreg64(dd, kr_serdes_cfg0);
+	val |= SYM_MASK(SerdesCfg0, TxIdeEnX);
+	qib_write_kreg(dd, kr_serdes_cfg0, val);
+}
+
+/**
+ * qib_6120_setup_setextled - set the state of the two external LEDs
+ * @dd: the qlogic_ib device
+ * @on: whether the link is up or not
+ *
+ * The exact combo of LEDs if on is true is determined by looking
+ * at the ibcstatus.
+
+ * These LEDs indicate the physical and logical state of IB link.
+ * For this chip (at least with recommended board pinouts), LED1
+ * is Yellow (logical state) and LED2 is Green (physical state),
+ *
+ * Note:  We try to match the Mellanox HCA LED behavior as best
+ * we can.  Green indicates physical link state is OK (something is
+ * plugged in, and we can train).
+ * Amber indicates the link is logically up (ACTIVE).
+ * Mellanox further blinks the amber LED to indicate data packet
+ * activity, but we have no hardware support for that, so it would
+ * require waking up every 10-20 msecs and checking the counters
+ * on the chip, and then turning the LED off if appropriate.  That's
+ * visible overhead, so not something we will do.
+ *
+ */
+static void qib_6120_setup_setextled(struct qib_pportdata *ppd, u32 on)
+{
+	u64 extctl, val, lst, ltst;
+	unsigned long flags;
+	struct qib_devdata *dd = ppd->dd;
+
+	/*
+	 * The diags use the LED to indicate diag info, so we leave
+	 * the external LED alone when the diags are running.
+	 */
+	if (dd->diag_client)
+		return;
+
+	/* Allow override of LED display for, e.g. Locating system in rack */
+	if (ppd->led_override) {
+		ltst = (ppd->led_override & QIB_LED_PHYS) ?
+			IB_PHYSPORTSTATE_LINKUP : IB_PHYSPORTSTATE_DISABLED,
+		lst = (ppd->led_override & QIB_LED_LOG) ?
+			IB_PORT_ACTIVE : IB_PORT_DOWN;
+	} else if (on) {
+		val = qib_read_kreg64(dd, kr_ibcstatus);
+		ltst = qib_6120_phys_portstate(val);
+		lst = qib_6120_iblink_state(val);
+	} else {
+		ltst = 0;
+		lst = 0;
+	}
+
+	spin_lock_irqsave(&dd->cspec->gpio_lock, flags);
+	extctl = dd->cspec->extctrl & ~(SYM_MASK(EXTCtrl, LEDPriPortGreenOn) |
+				 SYM_MASK(EXTCtrl, LEDPriPortYellowOn));
+
+	if (ltst == IB_PHYSPORTSTATE_LINKUP)
+		extctl |= SYM_MASK(EXTCtrl, LEDPriPortYellowOn);
+	if (lst == IB_PORT_ACTIVE)
+		extctl |= SYM_MASK(EXTCtrl, LEDPriPortGreenOn);
+	dd->cspec->extctrl = extctl;
+	qib_write_kreg(dd, kr_extctrl, extctl);
+	spin_unlock_irqrestore(&dd->cspec->gpio_lock, flags);
+}
+
+static void qib_6120_free_irq(struct qib_devdata *dd)
+{
+	if (dd->cspec->irq) {
+		free_irq(dd->cspec->irq, dd);
+		dd->cspec->irq = 0;
+	}
+	qib_nomsi(dd);
+}
+
+/**
+ * qib_6120_setup_cleanup - clean up any per-chip chip-specific stuff
+ * @dd: the qlogic_ib device
+ *
+ * This is called during driver unload.
+*/
+static void qib_6120_setup_cleanup(struct qib_devdata *dd)
+{
+	qib_6120_free_irq(dd);
+	kfree(dd->cspec->cntrs);
+	kfree(dd->cspec->portcntrs);
+	if (dd->cspec->dummy_hdrq) {
+		dma_free_coherent(&dd->pcidev->dev,
+				  ALIGN(dd->rcvhdrcnt *
+					dd->rcvhdrentsize *
+					sizeof(u32), PAGE_SIZE),
+				  dd->cspec->dummy_hdrq,
+				  dd->cspec->dummy_hdrq_phys);
+		dd->cspec->dummy_hdrq = NULL;
+	}
+}
+
+static void qib_wantpiobuf_6120_intr(struct qib_devdata *dd, u32 needint)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&dd->sendctrl_lock, flags);
+	if (needint)
+		dd->sendctrl |= SYM_MASK(SendCtrl, PIOIntBufAvail);
+	else
+		dd->sendctrl &= ~SYM_MASK(SendCtrl, PIOIntBufAvail);
+	qib_write_kreg(dd, kr_sendctrl, dd->sendctrl);
+	qib_write_kreg(dd, kr_scratch, 0ULL);
+	spin_unlock_irqrestore(&dd->sendctrl_lock, flags);
+}
+
+/*
+ * handle errors and unusual events first, separate function
+ * to improve cache hits for fast path interrupt handling
+ */
+static noinline void unlikely_6120_intr(struct qib_devdata *dd, u64 istat)
+{
+	if (unlikely(istat & ~QLOGIC_IB_I_BITSEXTANT))
+		qib_dev_err(dd, "interrupt with unknown interrupts %Lx set\n",
+			    istat & ~QLOGIC_IB_I_BITSEXTANT);
+
+	if (istat & QLOGIC_IB_I_ERROR) {
+		u64 estat = 0;
+
+		qib_stats.sps_errints++;
+		estat = qib_read_kreg64(dd, kr_errstatus);
+		if (!estat)
+			qib_devinfo(dd->pcidev, "error interrupt (%Lx), "
+				 "but no error bits set!\n", istat);
+		handle_6120_errors(dd, estat);
+	}
+
+	if (istat & QLOGIC_IB_I_GPIO) {
+		u32 gpiostatus;
+		u32 to_clear = 0;
+
+		/*
+		 * GPIO_3..5 on IBA6120 Rev2 chips indicate
+		 * errors that we need to count.
+		 */
+		gpiostatus = qib_read_kreg32(dd, kr_gpio_status);
+		/* First the error-counter case. */
+		if (gpiostatus & GPIO_ERRINTR_MASK) {
+			/* want to clear the bits we see asserted. */
+			to_clear |= (gpiostatus & GPIO_ERRINTR_MASK);
+
+			/*
+			 * Count appropriately, clear bits out of our copy,
+			 * as they have been "handled".
+			 */
+			if (gpiostatus & (1 << GPIO_RXUVL_BIT))
+				dd->cspec->rxfc_unsupvl_errs++;
+			if (gpiostatus & (1 << GPIO_OVRUN_BIT))
+				dd->cspec->overrun_thresh_errs++;
+			if (gpiostatus & (1 << GPIO_LLI_BIT))
+				dd->cspec->lli_errs++;
+			gpiostatus &= ~GPIO_ERRINTR_MASK;
+		}
+		if (gpiostatus) {
+			/*
+			 * Some unexpected bits remain. If they could have
+			 * caused the interrupt, complain and clear.
+			 * To avoid repetition of this condition, also clear
+			 * the mask. It is almost certainly due to error.
+			 */
+			const u32 mask = qib_read_kreg32(dd, kr_gpio_mask);
+
+			/*
+			 * Also check that the chip reflects our shadow,
+			 * and report issues, If they caused the interrupt.
+			 * we will suppress by refreshing from the shadow.
+			 */
+			if (mask & gpiostatus) {
+				to_clear |= (gpiostatus & mask);
+				dd->cspec->gpio_mask &= ~(gpiostatus & mask);
+				qib_write_kreg(dd, kr_gpio_mask,
+					       dd->cspec->gpio_mask);
+			}
+		}
+		if (to_clear)
+			qib_write_kreg(dd, kr_gpio_clear, (u64) to_clear);
+	}
+}
+
+static irqreturn_t qib_6120intr(int irq, void *data)
+{
+	struct qib_devdata *dd = data;
+	irqreturn_t ret;
+	u32 istat, ctxtrbits, rmask, crcs = 0;
+	unsigned i;
+
+	if ((dd->flags & (QIB_PRESENT | QIB_BADINTR)) != QIB_PRESENT) {
+		/*
+		 * This return value is not great, but we do not want the
+		 * interrupt core code to remove our interrupt handler
+		 * because we don't appear to be handling an interrupt
+		 * during a chip reset.
+		 */
+		ret = IRQ_HANDLED;
+		goto bail;
+	}
+
+	istat = qib_read_kreg32(dd, kr_intstatus);
+
+	if (unlikely(!istat)) {
+		ret = IRQ_NONE; /* not our interrupt, or already handled */
+		goto bail;
+	}
+	if (unlikely(istat == -1)) {
+		qib_bad_intrstatus(dd);
+		/* don't know if it was our interrupt or not */
+		ret = IRQ_NONE;
+		goto bail;
+	}
+
+	qib_stats.sps_ints++;
+	if (dd->int_counter != (u32) -1)
+		dd->int_counter++;
+
+	if (unlikely(istat & (~QLOGIC_IB_I_BITSEXTANT |
+			      QLOGIC_IB_I_GPIO | QLOGIC_IB_I_ERROR)))
+		unlikely_6120_intr(dd, istat);
+
+	/*
+	 * Clear the interrupt bits we found set, relatively early, so we
+	 * "know" know the chip will have seen this by the time we process
+	 * the queue, and will re-interrupt if necessary.  The processor
+	 * itself won't take the interrupt again until we return.
+	 */
+	qib_write_kreg(dd, kr_intclear, istat);
+
+	/*
+	 * Handle kernel receive queues before checking for pio buffers
+	 * available since receives can overflow; piobuf waiters can afford
+	 * a few extra cycles, since they were waiting anyway.
+	 */
+	ctxtrbits = istat &
+		((QLOGIC_IB_I_RCVAVAIL_MASK << QLOGIC_IB_I_RCVAVAIL_SHIFT) |
+		 (QLOGIC_IB_I_RCVURG_MASK << QLOGIC_IB_I_RCVURG_SHIFT));
+	if (ctxtrbits) {
+		rmask = (1U << QLOGIC_IB_I_RCVAVAIL_SHIFT) |
+			(1U << QLOGIC_IB_I_RCVURG_SHIFT);
+		for (i = 0; i < dd->first_user_ctxt; i++) {
+			if (ctxtrbits & rmask) {
+				ctxtrbits &= ~rmask;
+				crcs += qib_kreceive(dd->rcd[i],
+						     &dd->cspec->lli_counter,
+						     NULL);
+			}
+			rmask <<= 1;
+		}
+		if (crcs) {
+			u32 cntr = dd->cspec->lli_counter;
+			cntr += crcs;
+			if (cntr) {
+				if (cntr > dd->cspec->lli_thresh) {
+					dd->cspec->lli_counter = 0;
+					dd->cspec->lli_errs++;
+				} else
+					dd->cspec->lli_counter += cntr;
+			}
+		}
+
+
+		if (ctxtrbits) {
+			ctxtrbits =
+				(ctxtrbits >> QLOGIC_IB_I_RCVAVAIL_SHIFT) |
+				(ctxtrbits >> QLOGIC_IB_I_RCVURG_SHIFT);
+			qib_handle_urcv(dd, ctxtrbits);
+		}
+	}
+
+	if ((istat & QLOGIC_IB_I_SPIOBUFAVAIL) && (dd->flags & QIB_INITTED))
+		qib_ib_piobufavail(dd);
+
+	ret = IRQ_HANDLED;
+bail:
+	return ret;
+}
+
+/*
+ * Set up our chip-specific interrupt handler
+ * The interrupt type has already been setup, so
+ * we just need to do the registration and error checking.
+ */
+static void qib_setup_6120_interrupt(struct qib_devdata *dd)
+{
+	/*
+	 * If the chip supports added error indication via GPIO pins,
+	 * enable interrupts on those bits so the interrupt routine
+	 * can count the events. Also set flag so interrupt routine
+	 * can know they are expected.
+	 */
+	if (SYM_FIELD(dd->revision, Revision_R,
+		      ChipRevMinor) > 1) {
+		/* Rev2+ reports extra errors via internal GPIO pins */
+		dd->cspec->gpio_mask |= GPIO_ERRINTR_MASK;
+		qib_write_kreg(dd, kr_gpio_mask, dd->cspec->gpio_mask);
+	}
+
+	if (!dd->cspec->irq)
+		qib_dev_err(dd, "irq is 0, BIOS error?  Interrupts won't "
+			    "work\n");
+	else {
+		int ret;
+		ret = request_irq(dd->cspec->irq, qib_6120intr, 0,
+				  QIB_DRV_NAME, dd);
+		if (ret)
+			qib_dev_err(dd, "Couldn't setup interrupt "
+				    "(irq=%d): %d\n", dd->cspec->irq,
+				    ret);
+	}
+}
+
+/**
+ * pe_boardname - fill in the board name
+ * @dd: the qlogic_ib device
+ *
+ * info is based on the board revision register
+ */
+static void pe_boardname(struct qib_devdata *dd)
+{
+	char *n;
+	u32 boardid, namelen;
+
+	boardid = SYM_FIELD(dd->revision, Revision,
+			    BoardID);
+
+	switch (boardid) {
+	case 2:
+		n = "InfiniPath_QLE7140";
+		break;
+	default:
+		qib_dev_err(dd, "Unknown 6120 board with ID %u\n", boardid);
+		n = "Unknown_InfiniPath_6120";
+		break;
+	}
+	namelen = strlen(n) + 1;
+	dd->boardname = kmalloc(namelen, GFP_KERNEL);
+	if (!dd->boardname)
+		qib_dev_err(dd, "Failed allocation for board name: %s\n", n);
+	else
+		snprintf(dd->boardname, namelen, "%s", n);
+
+	if (dd->majrev != 4 || !dd->minrev || dd->minrev > 2)
+		qib_dev_err(dd, "Unsupported InfiniPath hardware revision "
+			    "%u.%u!\n", dd->majrev, dd->minrev);
+
+	snprintf(dd->boardversion, sizeof(dd->boardversion),
+		 "ChipABI %u.%u, %s, InfiniPath%u %u.%u, SW Compat %u\n",
+		 QIB_CHIP_VERS_MAJ, QIB_CHIP_VERS_MIN, dd->boardname,
+		 (unsigned)SYM_FIELD(dd->revision, Revision_R, Arch),
+		 dd->majrev, dd->minrev,
+		 (unsigned)SYM_FIELD(dd->revision, Revision_R, SW));
+
+}
+
+/*
+ * This routine sleeps, so it can only be called from user context, not
+ * from interrupt context.  If we need interrupt context, we can split
+ * it into two routines.
+ */
+static int qib_6120_setup_reset(struct qib_devdata *dd)
+{
+	u64 val;
+	int i;
+	int ret;
+	u16 cmdval;
+	u8 int_line, clinesz;
+
+	qib_pcie_getcmd(dd, &cmdval, &int_line, &clinesz);
+
+	/* Use ERROR so it shows up in logs, etc. */
+	qib_dev_err(dd, "Resetting InfiniPath unit %u\n", dd->unit);
+
+	/* no interrupts till re-initted */
+	qib_6120_set_intr_state(dd, 0);
+
+	dd->cspec->ibdeltainprog = 0;
+	dd->cspec->ibsymdelta = 0;
+	dd->cspec->iblnkerrdelta = 0;
+
+	/*
+	 * Keep chip from being accessed until we are ready.  Use
+	 * writeq() directly, to allow the write even though QIB_PRESENT
+	 * isnt' set.
+	 */
+	dd->flags &= ~(QIB_INITTED | QIB_PRESENT);
+	dd->int_counter = 0; /* so we check interrupts work again */
+	val = dd->control | QLOGIC_IB_C_RESET;
+	writeq(val, &dd->kregbase[kr_control]);
+	mb(); /* prevent compiler re-ordering around actual reset */
+
+	for (i = 1; i <= 5; i++) {
+		/*
+		 * Allow MBIST, etc. to complete; longer on each retry.
+		 * We sometimes get machine checks from bus timeout if no
+		 * response, so for now, make it *really* long.
+		 */
+		msleep(1000 + (1 + i) * 2000);
+
+		qib_pcie_reenable(dd, cmdval, int_line, clinesz);
+
+		/*
+		 * Use readq directly, so we don't need to mark it as PRESENT
+		 * until we get a successful indication that all is well.
+		 */
+		val = readq(&dd->kregbase[kr_revision]);
+		if (val == dd->revision) {
+			dd->flags |= QIB_PRESENT; /* it's back */
+			ret = qib_reinit_intr(dd);
+			goto bail;
+		}
+	}
+	ret = 0; /* failed */
+
+bail:
+	if (ret) {
+		if (qib_pcie_params(dd, dd->lbus_width, NULL, NULL))
+			qib_dev_err(dd, "Reset failed to setup PCIe or "
+				    "interrupts; continuing anyway\n");
+		/* clear the reset error, init error/hwerror mask */
+		qib_6120_init_hwerrors(dd);
+		/* for Rev2 error interrupts; nop for rev 1 */
+		qib_write_kreg(dd, kr_gpio_mask, dd->cspec->gpio_mask);
+		/* clear the reset error, init error/hwerror mask */
+		qib_6120_init_hwerrors(dd);
+	}
+	return ret;
+}
+
+/**
+ * qib_6120_put_tid - write a TID in chip
+ * @dd: the qlogic_ib device
+ * @tidptr: pointer to the expected TID (in chip) to update
+ * @tidtype: RCVHQ_RCV_TYPE_EAGER (1) for eager, RCVHQ_RCV_TYPE_EXPECTED (0)
+ * for expected
+ * @pa: physical address of in memory buffer; tidinvalid if freeing
+ *
+ * This exists as a separate routine to allow for special locking etc.
+ * It's used for both the full cleanup on exit, as well as the normal
+ * setup and teardown.
+ */
+static void qib_6120_put_tid(struct qib_devdata *dd, u64 __iomem *tidptr,
+			     u32 type, unsigned long pa)
+{
+	u32 __iomem *tidp32 = (u32 __iomem *)tidptr;
+	unsigned long flags;
+	int tidx;
+	spinlock_t *tidlockp; /* select appropriate spinlock */
+
+	if (!dd->kregbase)
+		return;
+
+	if (pa != dd->tidinvalid) {
+		if (pa & ((1U << 11) - 1)) {
+			qib_dev_err(dd, "Physaddr %lx not 2KB aligned!\n",
+				    pa);
+			return;
+		}
+		pa >>= 11;
+		if (pa & ~QLOGIC_IB_RT_ADDR_MASK) {
+			qib_dev_err(dd, "Physical page address 0x%lx "
+				    "larger than supported\n", pa);
+			return;
+		}
+
+		if (type == RCVHQ_RCV_TYPE_EAGER)
+			pa |= dd->tidtemplate;
+		else /* for now, always full 4KB page */
+			pa |= 2 << 29;
+	}
+
+	/*
+	 * Avoid chip issue by writing the scratch register
+	 * before and after the TID, and with an io write barrier.
+	 * We use a spinlock around the writes, so they can't intermix
+	 * with other TID (eager or expected) writes (the chip problem
+	 * is triggered by back to back TID writes). Unfortunately, this
+	 * call can be done from interrupt level for the ctxt 0 eager TIDs,
+	 * so we have to use irqsave locks.
+	 */
+	/*
+	 * Assumes tidptr always > egrtidbase
+	 * if type == RCVHQ_RCV_TYPE_EAGER.
+	 */
+	tidx = tidptr - dd->egrtidbase;
+
+	tidlockp = (type == RCVHQ_RCV_TYPE_EAGER && tidx < dd->rcvhdrcnt)
+		? &dd->cspec->kernel_tid_lock : &dd->cspec->user_tid_lock;
+	spin_lock_irqsave(tidlockp, flags);
+	qib_write_kreg(dd, kr_scratch, 0xfeeddeaf);
+	writel(pa, tidp32);
+	qib_write_kreg(dd, kr_scratch, 0xdeadbeef);
+	mmiowb();
+	spin_unlock_irqrestore(tidlockp, flags);
+}
+
+/**
+ * qib_6120_put_tid_2 - write a TID in chip, Revision 2 or higher
+ * @dd: the qlogic_ib device
+ * @tidptr: pointer to the expected TID (in chip) to update
+ * @tidtype: RCVHQ_RCV_TYPE_EAGER (1) for eager, RCVHQ_RCV_TYPE_EXPECTED (0)
+ * for expected
+ * @pa: physical address of in memory buffer; tidinvalid if freeing
+ *
+ * This exists as a separate routine to allow for selection of the
+ * appropriate "flavor". The static calls in cleanup just use the
+ * revision-agnostic form, as they are not performance critical.
+ */
+static void qib_6120_put_tid_2(struct qib_devdata *dd, u64 __iomem *tidptr,
+			       u32 type, unsigned long pa)
+{
+	u32 __iomem *tidp32 = (u32 __iomem *)tidptr;
+	u32 tidx;
+
+	if (!dd->kregbase)
+		return;
+
+	if (pa != dd->tidinvalid) {
+		if (pa & ((1U << 11) - 1)) {
+			qib_dev_err(dd, "Physaddr %lx not 2KB aligned!\n",
+				    pa);
+			return;
+		}
+		pa >>= 11;
+		if (pa & ~QLOGIC_IB_RT_ADDR_MASK) {
+			qib_dev_err(dd, "Physical page address 0x%lx "
+				    "larger than supported\n", pa);
+			return;
+		}
+
+		if (type == RCVHQ_RCV_TYPE_EAGER)
+			pa |= dd->tidtemplate;
+		else /* for now, always full 4KB page */
+			pa |= 2 << 29;
+	}
+	tidx = tidptr - dd->egrtidbase;
+	writel(pa, tidp32);
+	mmiowb();
+}
+
+
+/**
+ * qib_6120_clear_tids - clear all TID entries for a context, expected and eager
+ * @dd: the qlogic_ib device
+ * @ctxt: the context
+ *
+ * clear all TID entries for a context, expected and eager.
+ * Used from qib_close().  On this chip, TIDs are only 32 bits,
+ * not 64, but they are still on 64 bit boundaries, so tidbase
+ * is declared as u64 * for the pointer math, even though we write 32 bits
+ */
+static void qib_6120_clear_tids(struct qib_devdata *dd,
+				struct qib_ctxtdata *rcd)
+{
+	u64 __iomem *tidbase;
+	unsigned long tidinv;
+	u32 ctxt;
+	int i;
+
+	if (!dd->kregbase || !rcd)
+		return;
+
+	ctxt = rcd->ctxt;
+
+	tidinv = dd->tidinvalid;
+	tidbase = (u64 __iomem *)
+		((char __iomem *)(dd->kregbase) +
+		 dd->rcvtidbase +
+		 ctxt * dd->rcvtidcnt * sizeof(*tidbase));
+
+	for (i = 0; i < dd->rcvtidcnt; i++)
+		/* use func pointer because could be one of two funcs */
+		dd->f_put_tid(dd, &tidbase[i], RCVHQ_RCV_TYPE_EXPECTED,
+				  tidinv);
+
+	tidbase = (u64 __iomem *)
+		((char __iomem *)(dd->kregbase) +
+		 dd->rcvegrbase +
+		 rcd->rcvegr_tid_base * sizeof(*tidbase));
+
+	for (i = 0; i < rcd->rcvegrcnt; i++)
+		/* use func pointer because could be one of two funcs */
+		dd->f_put_tid(dd, &tidbase[i], RCVHQ_RCV_TYPE_EAGER,
+				  tidinv);
+}
+
+/**
+ * qib_6120_tidtemplate - setup constants for TID updates
+ * @dd: the qlogic_ib device
+ *
+ * We setup stuff that we use a lot, to avoid calculating each time
+ */
+static void qib_6120_tidtemplate(struct qib_devdata *dd)
+{
+	u32 egrsize = dd->rcvegrbufsize;
+
+	/*
+	 * For now, we always allocate 4KB buffers (at init) so we can
+	 * receive max size packets.  We may want a module parameter to
+	 * specify 2KB or 4KB and/or make be per ctxt instead of per device
+	 * for those who want to reduce memory footprint.  Note that the
+	 * rcvhdrentsize size must be large enough to hold the largest
+	 * IB header (currently 96 bytes) that we expect to handle (plus of
+	 * course the 2 dwords of RHF).
+	 */
+	if (egrsize == 2048)
+		dd->tidtemplate = 1U << 29;
+	else if (egrsize == 4096)
+		dd->tidtemplate = 2U << 29;
+	dd->tidinvalid = 0;
+}
+
+int __attribute__((weak)) qib_unordered_wc(void)
+{
+	return 0;
+}
+
+/**
+ * qib_6120_get_base_info - set chip-specific flags for user code
+ * @rcd: the qlogic_ib ctxt
+ * @kbase: qib_base_info pointer
+ *
+ * We set the PCIE flag because the lower bandwidth on PCIe vs
+ * HyperTransport can affect some user packet algorithms.
+ */
+static int qib_6120_get_base_info(struct qib_ctxtdata *rcd,
+				  struct qib_base_info *kinfo)
+{
+	if (qib_unordered_wc())
+		kinfo->spi_runtime_flags |= QIB_RUNTIME_FORCE_WC_ORDER;
+
+	kinfo->spi_runtime_flags |= QIB_RUNTIME_PCIE |
+		QIB_RUNTIME_FORCE_PIOAVAIL | QIB_RUNTIME_PIO_REGSWAPPED;
+	return 0;
+}
+
+
+static struct qib_message_header *
+qib_6120_get_msgheader(struct qib_devdata *dd, __le32 *rhf_addr)
+{
+	return (struct qib_message_header *)
+		&rhf_addr[sizeof(u64) / sizeof(u32)];
+}
+
+static void qib_6120_config_ctxts(struct qib_devdata *dd)
+{
+	dd->ctxtcnt = qib_read_kreg32(dd, kr_portcnt);
+	if (qib_n_krcv_queues > 1) {
+		dd->first_user_ctxt = qib_n_krcv_queues * dd->num_pports;
+		if (dd->first_user_ctxt > dd->ctxtcnt)
+			dd->first_user_ctxt = dd->ctxtcnt;
+		dd->qpn_mask = dd->first_user_ctxt <= 2 ? 2 : 6;
+	} else
+		dd->first_user_ctxt = dd->num_pports;
+	dd->n_krcv_queues = dd->first_user_ctxt;
+}
+
+static void qib_update_6120_usrhead(struct qib_ctxtdata *rcd, u64 hd,
+				    u32 updegr, u32 egrhd)
+{
+	qib_write_ureg(rcd->dd, ur_rcvhdrhead, hd, rcd->ctxt);
+	if (updegr)
+		qib_write_ureg(rcd->dd, ur_rcvegrindexhead, egrhd, rcd->ctxt);
+}
+
+static u32 qib_6120_hdrqempty(struct qib_ctxtdata *rcd)
+{
+	u32 head, tail;
+
+	head = qib_read_ureg32(rcd->dd, ur_rcvhdrhead, rcd->ctxt);
+	if (rcd->rcvhdrtail_kvaddr)
+		tail = qib_get_rcvhdrtail(rcd);
+	else
+		tail = qib_read_ureg32(rcd->dd, ur_rcvhdrtail, rcd->ctxt);
+	return head == tail;
+}
+
+/*
+ * Used when we close any ctxt, for DMA already in flight
+ * at close.  Can't be done until we know hdrq size, so not
+ * early in chip init.
+ */
+static void alloc_dummy_hdrq(struct qib_devdata *dd)
+{
+	dd->cspec->dummy_hdrq = dma_alloc_coherent(&dd->pcidev->dev,
+					dd->rcd[0]->rcvhdrq_size,
+					&dd->cspec->dummy_hdrq_phys,
+					GFP_KERNEL | __GFP_COMP);
+	if (!dd->cspec->dummy_hdrq) {
+		qib_devinfo(dd->pcidev, "Couldn't allocate dummy hdrq\n");
+		/* fallback to just 0'ing */
+		dd->cspec->dummy_hdrq_phys = 0UL;
+	}
+}
+
+/*
+ * Modify the RCVCTRL register in chip-specific way. This
+ * is a function because bit positions and (future) register
+ * location is chip-specific, but the needed operations are
+ * generic. <op> is a bit-mask because we often want to
+ * do multiple modifications.
+ */
+static void rcvctrl_6120_mod(struct qib_pportdata *ppd, unsigned int op,
+			     int ctxt)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u64 mask, val;
+	unsigned long flags;
+
+	spin_lock_irqsave(&dd->cspec->rcvmod_lock, flags);
+
+	if (op & QIB_RCVCTRL_TAILUPD_ENB)
+		dd->rcvctrl |= (1ULL << QLOGIC_IB_R_TAILUPD_SHIFT);
+	if (op & QIB_RCVCTRL_TAILUPD_DIS)
+		dd->rcvctrl &= ~(1ULL << QLOGIC_IB_R_TAILUPD_SHIFT);
+	if (op & QIB_RCVCTRL_PKEY_ENB)
+		dd->rcvctrl &= ~(1ULL << IBA6120_R_PKEY_DIS_SHIFT);
+	if (op & QIB_RCVCTRL_PKEY_DIS)
+		dd->rcvctrl |= (1ULL << IBA6120_R_PKEY_DIS_SHIFT);
+	if (ctxt < 0)
+		mask = (1ULL << dd->ctxtcnt) - 1;
+	else
+		mask = (1ULL << ctxt);
+	if (op & QIB_RCVCTRL_CTXT_ENB) {
+		/* always done for specific ctxt */
+		dd->rcvctrl |= (mask << SYM_LSB(RcvCtrl, PortEnable));
+		if (!(dd->flags & QIB_NODMA_RTAIL))
+			dd->rcvctrl |= 1ULL << QLOGIC_IB_R_TAILUPD_SHIFT;
+		/* Write these registers before the context is enabled. */
+		qib_write_kreg_ctxt(dd, kr_rcvhdrtailaddr, ctxt,
+			dd->rcd[ctxt]->rcvhdrqtailaddr_phys);
+		qib_write_kreg_ctxt(dd, kr_rcvhdraddr, ctxt,
+			dd->rcd[ctxt]->rcvhdrq_phys);
+
+		if (ctxt == 0 && !dd->cspec->dummy_hdrq)
+			alloc_dummy_hdrq(dd);
+	}
+	if (op & QIB_RCVCTRL_CTXT_DIS)
+		dd->rcvctrl &= ~(mask << SYM_LSB(RcvCtrl, PortEnable));
+	if (op & QIB_RCVCTRL_INTRAVAIL_ENB)
+		dd->rcvctrl |= (mask << QLOGIC_IB_R_INTRAVAIL_SHIFT);
+	if (op & QIB_RCVCTRL_INTRAVAIL_DIS)
+		dd->rcvctrl &= ~(mask << QLOGIC_IB_R_INTRAVAIL_SHIFT);
+	qib_write_kreg(dd, kr_rcvctrl, dd->rcvctrl);
+	if ((op & QIB_RCVCTRL_INTRAVAIL_ENB) && dd->rhdrhead_intr_off) {
+		/* arm rcv interrupt */
+		val = qib_read_ureg32(dd, ur_rcvhdrhead, ctxt) |
+			dd->rhdrhead_intr_off;
+		qib_write_ureg(dd, ur_rcvhdrhead, val, ctxt);
+	}
+	if (op & QIB_RCVCTRL_CTXT_ENB) {
+		/*
+		 * Init the context registers also; if we were
+		 * disabled, tail and head should both be zero
+		 * already from the enable, but since we don't
+		 * know, we have to do it explictly.
+		 */
+		val = qib_read_ureg32(dd, ur_rcvegrindextail, ctxt);
+		qib_write_ureg(dd, ur_rcvegrindexhead, val, ctxt);
+
+		val = qib_read_ureg32(dd, ur_rcvhdrtail, ctxt);
+		dd->rcd[ctxt]->head = val;
+		/* If kctxt, interrupt on next receive. */
+		if (ctxt < dd->first_user_ctxt)
+			val |= dd->rhdrhead_intr_off;
+		qib_write_ureg(dd, ur_rcvhdrhead, val, ctxt);
+	}
+	if (op & QIB_RCVCTRL_CTXT_DIS) {
+		/*
+		 * Be paranoid, and never write 0's to these, just use an
+		 * unused page.  Of course,
+		 * rcvhdraddr points to a large chunk of memory, so this
+		 * could still trash things, but at least it won't trash
+		 * page 0, and by disabling the ctxt, it should stop "soon",
+		 * even if a packet or two is in already in flight after we
+		 * disabled the ctxt.  Only 6120 has this issue.
+		 */
+		if (ctxt >= 0) {
+			qib_write_kreg_ctxt(dd, kr_rcvhdrtailaddr, ctxt,
+					    dd->cspec->dummy_hdrq_phys);
+			qib_write_kreg_ctxt(dd, kr_rcvhdraddr, ctxt,
+					    dd->cspec->dummy_hdrq_phys);
+		} else {
+			unsigned i;
+
+			for (i = 0; i < dd->cfgctxts; i++) {
+				qib_write_kreg_ctxt(dd, kr_rcvhdrtailaddr,
+					    i, dd->cspec->dummy_hdrq_phys);
+				qib_write_kreg_ctxt(dd, kr_rcvhdraddr,
+					    i, dd->cspec->dummy_hdrq_phys);
+			}
+		}
+	}
+	spin_unlock_irqrestore(&dd->cspec->rcvmod_lock, flags);
+}
+
+/*
+ * Modify the SENDCTRL register in chip-specific way. This
+ * is a function there may be multiple such registers with
+ * slightly different layouts. Only operations actually used
+ * are implemented yet.
+ * Chip requires no back-back sendctrl writes, so write
+ * scratch register after writing sendctrl
+ */
+static void sendctrl_6120_mod(struct qib_pportdata *ppd, u32 op)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u64 tmp_dd_sendctrl;
+	unsigned long flags;
+
+	spin_lock_irqsave(&dd->sendctrl_lock, flags);
+
+	/* First the ones that are "sticky", saved in shadow */
+	if (op & QIB_SENDCTRL_CLEAR)
+		dd->sendctrl = 0;
+	if (op & QIB_SENDCTRL_SEND_DIS)
+		dd->sendctrl &= ~SYM_MASK(SendCtrl, PIOEnable);
+	else if (op & QIB_SENDCTRL_SEND_ENB)
+		dd->sendctrl |= SYM_MASK(SendCtrl, PIOEnable);
+	if (op & QIB_SENDCTRL_AVAIL_DIS)
+		dd->sendctrl &= ~SYM_MASK(SendCtrl, PIOBufAvailUpd);
+	else if (op & QIB_SENDCTRL_AVAIL_ENB)
+		dd->sendctrl |= SYM_MASK(SendCtrl, PIOBufAvailUpd);
+
+	if (op & QIB_SENDCTRL_DISARM_ALL) {
+		u32 i, last;
+
+		tmp_dd_sendctrl = dd->sendctrl;
+		/*
+		 * disarm any that are not yet launched, disabling sends
+		 * and updates until done.
+		 */
+		last = dd->piobcnt2k + dd->piobcnt4k;
+		tmp_dd_sendctrl &=
+			~(SYM_MASK(SendCtrl, PIOEnable) |
+			  SYM_MASK(SendCtrl, PIOBufAvailUpd));
+		for (i = 0; i < last; i++) {
+			qib_write_kreg(dd, kr_sendctrl, tmp_dd_sendctrl |
+				       SYM_MASK(SendCtrl, Disarm) | i);
+			qib_write_kreg(dd, kr_scratch, 0);
+		}
+	}
+
+	tmp_dd_sendctrl = dd->sendctrl;
+
+	if (op & QIB_SENDCTRL_FLUSH)
+		tmp_dd_sendctrl |= SYM_MASK(SendCtrl, Abort);
+	if (op & QIB_SENDCTRL_DISARM)
+		tmp_dd_sendctrl |= SYM_MASK(SendCtrl, Disarm) |
+			((op & QIB_6120_SendCtrl_DisarmPIOBuf_RMASK) <<
+			 SYM_LSB(SendCtrl, DisarmPIOBuf));
+	if (op & QIB_SENDCTRL_AVAIL_BLIP)
+		tmp_dd_sendctrl &= ~SYM_MASK(SendCtrl, PIOBufAvailUpd);
+
+	qib_write_kreg(dd, kr_sendctrl, tmp_dd_sendctrl);
+	qib_write_kreg(dd, kr_scratch, 0);
+
+	if (op & QIB_SENDCTRL_AVAIL_BLIP) {
+		qib_write_kreg(dd, kr_sendctrl, dd->sendctrl);
+		qib_write_kreg(dd, kr_scratch, 0);
+	}
+
+	spin_unlock_irqrestore(&dd->sendctrl_lock, flags);
+
+	if (op & QIB_SENDCTRL_FLUSH) {
+		u32 v;
+		/*
+		 * ensure writes have hit chip, then do a few
+		 * more reads, to allow DMA of pioavail registers
+		 * to occur, so in-memory copy is in sync with
+		 * the chip.  Not always safe to sleep.
+		 */
+		v = qib_read_kreg32(dd, kr_scratch);
+		qib_write_kreg(dd, kr_scratch, v);
+		v = qib_read_kreg32(dd, kr_scratch);
+		qib_write_kreg(dd, kr_scratch, v);
+		qib_read_kreg32(dd, kr_scratch);
+	}
+}
+
+/**
+ * qib_portcntr_6120 - read a per-port counter
+ * @dd: the qlogic_ib device
+ * @creg: the counter to snapshot
+ */
+static u64 qib_portcntr_6120(struct qib_pportdata *ppd, u32 reg)
+{
+	u64 ret = 0ULL;
+	struct qib_devdata *dd = ppd->dd;
+	u16 creg;
+	/* 0xffff for unimplemented or synthesized counters */
+	static const u16 xlator[] = {
+		[QIBPORTCNTR_PKTSEND] = cr_pktsend,
+		[QIBPORTCNTR_WORDSEND] = cr_wordsend,
+		[QIBPORTCNTR_PSXMITDATA] = 0xffff,
+		[QIBPORTCNTR_PSXMITPKTS] = 0xffff,
+		[QIBPORTCNTR_PSXMITWAIT] = 0xffff,
+		[QIBPORTCNTR_SENDSTALL] = cr_sendstall,
+		[QIBPORTCNTR_PKTRCV] = cr_pktrcv,
+		[QIBPORTCNTR_PSRCVDATA] = 0xffff,
+		[QIBPORTCNTR_PSRCVPKTS] = 0xffff,
+		[QIBPORTCNTR_RCVEBP] = cr_rcvebp,
+		[QIBPORTCNTR_RCVOVFL] = cr_rcvovfl,
+		[QIBPORTCNTR_WORDRCV] = cr_wordrcv,
+		[QIBPORTCNTR_RXDROPPKT] = cr_rxdroppkt,
+		[QIBPORTCNTR_RXLOCALPHYERR] = 0xffff,
+		[QIBPORTCNTR_RXVLERR] = 0xffff,
+		[QIBPORTCNTR_ERRICRC] = cr_erricrc,
+		[QIBPORTCNTR_ERRVCRC] = cr_errvcrc,
+		[QIBPORTCNTR_ERRLPCRC] = cr_errlpcrc,
+		[QIBPORTCNTR_BADFORMAT] = cr_badformat,
+		[QIBPORTCNTR_ERR_RLEN] = cr_err_rlen,
+		[QIBPORTCNTR_IBSYMBOLERR] = cr_ibsymbolerr,
+		[QIBPORTCNTR_INVALIDRLEN] = cr_invalidrlen,
+		[QIBPORTCNTR_UNSUPVL] = cr_txunsupvl,
+		[QIBPORTCNTR_EXCESSBUFOVFL] = 0xffff,
+		[QIBPORTCNTR_ERRLINK] = cr_errlink,
+		[QIBPORTCNTR_IBLINKDOWN] = cr_iblinkdown,
+		[QIBPORTCNTR_IBLINKERRRECOV] = cr_iblinkerrrecov,
+		[QIBPORTCNTR_LLI] = 0xffff,
+		[QIBPORTCNTR_PSINTERVAL] = 0xffff,
+		[QIBPORTCNTR_PSSTART] = 0xffff,
+		[QIBPORTCNTR_PSSTAT] = 0xffff,
+		[QIBPORTCNTR_VL15PKTDROP] = 0xffff,
+		[QIBPORTCNTR_ERRPKEY] = cr_errpkey,
+		[QIBPORTCNTR_KHDROVFL] = 0xffff,
+	};
+
+	if (reg >= ARRAY_SIZE(xlator)) {
+		qib_devinfo(ppd->dd->pcidev,
+			 "Unimplemented portcounter %u\n", reg);
+		goto done;
+	}
+	creg = xlator[reg];
+
+	/* handle counters requests not implemented as chip counters */
+	if (reg == QIBPORTCNTR_LLI)
+		ret = dd->cspec->lli_errs;
+	else if (reg == QIBPORTCNTR_EXCESSBUFOVFL)
+		ret = dd->cspec->overrun_thresh_errs;
+	else if (reg == QIBPORTCNTR_KHDROVFL) {
+		int i;
+
+		/* sum over all kernel contexts */
+		for (i = 0; i < dd->first_user_ctxt; i++)
+			ret += read_6120_creg32(dd, cr_portovfl + i);
+	} else if (reg == QIBPORTCNTR_PSSTAT)
+		ret = dd->cspec->pma_sample_status;
+	if (creg == 0xffff)
+		goto done;
+
+	/*
+	 * only fast incrementing counters are 64bit; use 32 bit reads to
+	 * avoid two independent reads when on opteron
+	 */
+	if (creg == cr_wordsend || creg == cr_wordrcv ||
+	    creg == cr_pktsend || creg == cr_pktrcv)
+		ret = read_6120_creg(dd, creg);
+	else
+		ret = read_6120_creg32(dd, creg);
+	if (creg == cr_ibsymbolerr) {
+		if (dd->cspec->ibdeltainprog)
+			ret -= ret - dd->cspec->ibsymsnap;
+		ret -= dd->cspec->ibsymdelta;
+	} else if (creg == cr_iblinkerrrecov) {
+		if (dd->cspec->ibdeltainprog)
+			ret -= ret - dd->cspec->iblnkerrsnap;
+		ret -= dd->cspec->iblnkerrdelta;
+	}
+	if (reg == QIBPORTCNTR_RXDROPPKT) /* add special cased count */
+		ret += dd->cspec->rxfc_unsupvl_errs;
+
+done:
+	return ret;
+}
+
+/*
+ * Device counter names (not port-specific), one line per stat,
+ * single string.  Used by utilities like ipathstats to print the stats
+ * in a way which works for different versions of drivers, without changing
+ * the utility.  Names need to be 12 chars or less (w/o newline), for proper
+ * display by utility.
+ * Non-error counters are first.
+ * Start of "error" conters is indicated by a leading "E " on the first
+ * "error" counter, and doesn't count in label length.
+ * The EgrOvfl list needs to be last so we truncate them at the configured
+ * context count for the device.
+ * cntr6120indices contains the corresponding register indices.
+ */
+static const char cntr6120names[] =
+	"Interrupts\n"
+	"HostBusStall\n"
+	"E RxTIDFull\n"
+	"RxTIDInvalid\n"
+	"Ctxt0EgrOvfl\n"
+	"Ctxt1EgrOvfl\n"
+	"Ctxt2EgrOvfl\n"
+	"Ctxt3EgrOvfl\n"
+	"Ctxt4EgrOvfl\n";
+
+static const size_t cntr6120indices[] = {
+	cr_lbint,
+	cr_lbflowstall,
+	cr_errtidfull,
+	cr_errtidvalid,
+	cr_portovfl + 0,
+	cr_portovfl + 1,
+	cr_portovfl + 2,
+	cr_portovfl + 3,
+	cr_portovfl + 4,
+};
+
+/*
+ * same as cntr6120names and cntr6120indices, but for port-specific counters.
+ * portcntr6120indices is somewhat complicated by some registers needing
+ * adjustments of various kinds, and those are ORed with _PORT_VIRT_FLAG
+ */
+static const char portcntr6120names[] =
+	"TxPkt\n"
+	"TxFlowPkt\n"
+	"TxWords\n"
+	"RxPkt\n"
+	"RxFlowPkt\n"
+	"RxWords\n"
+	"TxFlowStall\n"
+	"E IBStatusChng\n"
+	"IBLinkDown\n"
+	"IBLnkRecov\n"
+	"IBRxLinkErr\n"
+	"IBSymbolErr\n"
+	"RxLLIErr\n"
+	"RxBadFormat\n"
+	"RxBadLen\n"
+	"RxBufOvrfl\n"
+	"RxEBP\n"
+	"RxFlowCtlErr\n"
+	"RxICRCerr\n"
+	"RxLPCRCerr\n"
+	"RxVCRCerr\n"
+	"RxInvalLen\n"
+	"RxInvalPKey\n"
+	"RxPktDropped\n"
+	"TxBadLength\n"
+	"TxDropped\n"
+	"TxInvalLen\n"
+	"TxUnderrun\n"
+	"TxUnsupVL\n"
+	;
+
+#define _PORT_VIRT_FLAG 0x8000 /* "virtual", need adjustments */
+static const size_t portcntr6120indices[] = {
+	QIBPORTCNTR_PKTSEND | _PORT_VIRT_FLAG,
+	cr_pktsendflow,
+	QIBPORTCNTR_WORDSEND | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_PKTRCV | _PORT_VIRT_FLAG,
+	cr_pktrcvflowctrl,
+	QIBPORTCNTR_WORDRCV | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_SENDSTALL | _PORT_VIRT_FLAG,
+	cr_ibstatuschange,
+	QIBPORTCNTR_IBLINKDOWN | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_IBLINKERRRECOV | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_ERRLINK | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_IBSYMBOLERR | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_LLI | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_BADFORMAT | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_ERR_RLEN | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_RCVOVFL | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_RCVEBP | _PORT_VIRT_FLAG,
+	cr_rcvflowctrl_err,
+	QIBPORTCNTR_ERRICRC | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_ERRLPCRC | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_ERRVCRC | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_INVALIDRLEN | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_ERRPKEY | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_RXDROPPKT | _PORT_VIRT_FLAG,
+	cr_invalidslen,
+	cr_senddropped,
+	cr_errslen,
+	cr_sendunderrun,
+	cr_txunsupvl,
+};
+
+/* do all the setup to make the counter reads efficient later */
+static void init_6120_cntrnames(struct qib_devdata *dd)
+{
+	int i, j = 0;
+	char *s;
+
+	for (i = 0, s = (char *)cntr6120names; s && j <= dd->cfgctxts;
+	     i++) {
+		/* we always have at least one counter before the egrovfl */
+		if (!j && !strncmp("Ctxt0EgrOvfl", s + 1, 12))
+			j = 1;
+		s = strchr(s + 1, '\n');
+		if (s && j)
+			j++;
+	}
+	dd->cspec->ncntrs = i;
+	if (!s)
+		/* full list; size is without terminating null */
+		dd->cspec->cntrnamelen = sizeof(cntr6120names) - 1;
+	else
+		dd->cspec->cntrnamelen = 1 + s - cntr6120names;
+	dd->cspec->cntrs = kmalloc(dd->cspec->ncntrs
+		* sizeof(u64), GFP_KERNEL);
+	if (!dd->cspec->cntrs)
+		qib_dev_err(dd, "Failed allocation for counters\n");
+
+	for (i = 0, s = (char *)portcntr6120names; s; i++)
+		s = strchr(s + 1, '\n');
+	dd->cspec->nportcntrs = i - 1;
+	dd->cspec->portcntrnamelen = sizeof(portcntr6120names) - 1;
+	dd->cspec->portcntrs = kmalloc(dd->cspec->nportcntrs
+		* sizeof(u64), GFP_KERNEL);
+	if (!dd->cspec->portcntrs)
+		qib_dev_err(dd, "Failed allocation for portcounters\n");
+}
+
+static u32 qib_read_6120cntrs(struct qib_devdata *dd, loff_t pos, char **namep,
+			      u64 **cntrp)
+{
+	u32 ret;
+
+	if (namep) {
+		ret = dd->cspec->cntrnamelen;
+		if (pos >= ret)
+			ret = 0; /* final read after getting everything */
+		else
+			*namep = (char *)cntr6120names;
+	} else {
+		u64 *cntr = dd->cspec->cntrs;
+		int i;
+
+		ret = dd->cspec->ncntrs * sizeof(u64);
+		if (!cntr || pos >= ret) {
+			/* everything read, or couldn't get memory */
+			ret = 0;
+			goto done;
+		}
+		if (pos >= ret) {
+			ret = 0; /* final read after getting everything */
+			goto done;
+		}
+		*cntrp = cntr;
+		for (i = 0; i < dd->cspec->ncntrs; i++)
+			*cntr++ = read_6120_creg32(dd, cntr6120indices[i]);
+	}
+done:
+	return ret;
+}
+
+static u32 qib_read_6120portcntrs(struct qib_devdata *dd, loff_t pos, u32 port,
+				  char **namep, u64 **cntrp)
+{
+	u32 ret;
+
+	if (namep) {
+		ret = dd->cspec->portcntrnamelen;
+		if (pos >= ret)
+			ret = 0; /* final read after getting everything */
+		else
+			*namep = (char *)portcntr6120names;
+	} else {
+		u64 *cntr = dd->cspec->portcntrs;
+		struct qib_pportdata *ppd = &dd->pport[port];
+		int i;
+
+		ret = dd->cspec->nportcntrs * sizeof(u64);
+		if (!cntr || pos >= ret) {
+			/* everything read, or couldn't get memory */
+			ret = 0;
+			goto done;
+		}
+		*cntrp = cntr;
+		for (i = 0; i < dd->cspec->nportcntrs; i++) {
+			if (portcntr6120indices[i] & _PORT_VIRT_FLAG)
+				*cntr++ = qib_portcntr_6120(ppd,
+					portcntr6120indices[i] &
+					~_PORT_VIRT_FLAG);
+			else
+				*cntr++ = read_6120_creg32(dd,
+					   portcntr6120indices[i]);
+		}
+	}
+done:
+	return ret;
+}
+
+static void qib_chk_6120_errormask(struct qib_devdata *dd)
+{
+	static u32 fixed;
+	u32 ctrl;
+	unsigned long errormask;
+	unsigned long hwerrs;
+
+	if (!dd->cspec->errormask || !(dd->flags & QIB_INITTED))
+		return;
+
+	errormask = qib_read_kreg64(dd, kr_errmask);
+
+	if (errormask == dd->cspec->errormask)
+		return;
+	fixed++;
+
+	hwerrs = qib_read_kreg64(dd, kr_hwerrstatus);
+	ctrl = qib_read_kreg32(dd, kr_control);
+
+	qib_write_kreg(dd, kr_errmask,
+		dd->cspec->errormask);
+
+	if ((hwerrs & dd->cspec->hwerrmask) ||
+	    (ctrl & QLOGIC_IB_C_FREEZEMODE)) {
+		qib_write_kreg(dd, kr_hwerrclear, 0ULL);
+		qib_write_kreg(dd, kr_errclear, 0ULL);
+		/* force re-interrupt of pending events, just in case */
+		qib_write_kreg(dd, kr_intclear, 0ULL);
+		qib_devinfo(dd->pcidev,
+			 "errormask fixed(%u) %lx->%lx, ctrl %x hwerr %lx\n",
+			 fixed, errormask, (unsigned long)dd->cspec->errormask,
+			 ctrl, hwerrs);
+	}
+}
+
+/**
+ * qib_get_faststats - get word counters from chip before they overflow
+ * @opaque - contains a pointer to the qlogic_ib device qib_devdata
+ *
+ * This needs more work; in particular, decision on whether we really
+ * need traffic_wds done the way it is
+ * called from add_timer
+ */
+static void qib_get_6120_faststats(unsigned long opaque)
+{
+	struct qib_devdata *dd = (struct qib_devdata *) opaque;
+	struct qib_pportdata *ppd = dd->pport;
+	unsigned long flags;
+	u64 traffic_wds;
+
+	/*
+	 * don't access the chip while running diags, or memory diags can
+	 * fail
+	 */
+	if (!(dd->flags & QIB_INITTED) || dd->diag_client)
+		/* but re-arm the timer, for diags case; won't hurt other */
+		goto done;
+
+	/*
+	 * We now try to maintain an activity timer, based on traffic
+	 * exceeding a threshold, so we need to check the word-counts
+	 * even if they are 64-bit.
+	 */
+	traffic_wds = qib_portcntr_6120(ppd, cr_wordsend) +
+		qib_portcntr_6120(ppd, cr_wordrcv);
+	spin_lock_irqsave(&dd->eep_st_lock, flags);
+	traffic_wds -= dd->traffic_wds;
+	dd->traffic_wds += traffic_wds;
+	if (traffic_wds  >= QIB_TRAFFIC_ACTIVE_THRESHOLD)
+		atomic_add(5, &dd->active_time); /* S/B #define */
+	spin_unlock_irqrestore(&dd->eep_st_lock, flags);
+
+	qib_chk_6120_errormask(dd);
+done:
+	mod_timer(&dd->stats_timer, jiffies + HZ * ACTIVITY_TIMER);
+}
+
+/* no interrupt fallback for these chips */
+static int qib_6120_nointr_fallback(struct qib_devdata *dd)
+{
+	return 0;
+}
+
+/*
+ * reset the XGXS (between serdes and IBC).  Slightly less intrusive
+ * than resetting the IBC or external link state, and useful in some
+ * cases to cause some retraining.  To do this right, we reset IBC
+ * as well.
+ */
+static void qib_6120_xgxs_reset(struct qib_pportdata *ppd)
+{
+	u64 val, prev_val;
+	struct qib_devdata *dd = ppd->dd;
+
+	prev_val = qib_read_kreg64(dd, kr_xgxs_cfg);
+	val = prev_val | QLOGIC_IB_XGXS_RESET;
+	prev_val &= ~QLOGIC_IB_XGXS_RESET; /* be sure */
+	qib_write_kreg(dd, kr_control,
+		       dd->control & ~QLOGIC_IB_C_LINKENABLE);
+	qib_write_kreg(dd, kr_xgxs_cfg, val);
+	qib_read_kreg32(dd, kr_scratch);
+	qib_write_kreg(dd, kr_xgxs_cfg, prev_val);
+	qib_write_kreg(dd, kr_control, dd->control);
+}
+
+static int qib_6120_get_ib_cfg(struct qib_pportdata *ppd, int which)
+{
+	int ret;
+
+	switch (which) {
+	case QIB_IB_CFG_LWID:
+		ret = ppd->link_width_active;
+		break;
+
+	case QIB_IB_CFG_SPD:
+		ret = ppd->link_speed_active;
+		break;
+
+	case QIB_IB_CFG_LWID_ENB:
+		ret = ppd->link_width_enabled;
+		break;
+
+	case QIB_IB_CFG_SPD_ENB:
+		ret = ppd->link_speed_enabled;
+		break;
+
+	case QIB_IB_CFG_OP_VLS:
+		ret = ppd->vls_operational;
+		break;
+
+	case QIB_IB_CFG_VL_HIGH_CAP:
+		ret = 0;
+		break;
+
+	case QIB_IB_CFG_VL_LOW_CAP:
+		ret = 0;
+		break;
+
+	case QIB_IB_CFG_OVERRUN_THRESH: /* IB overrun threshold */
+		ret = SYM_FIELD(ppd->dd->cspec->ibcctrl, IBCCtrl,
+				OverrunThreshold);
+		break;
+
+	case QIB_IB_CFG_PHYERR_THRESH: /* IB PHY error threshold */
+		ret = SYM_FIELD(ppd->dd->cspec->ibcctrl, IBCCtrl,
+				PhyerrThreshold);
+		break;
+
+	case QIB_IB_CFG_LINKDEFAULT: /* IB link default (sleep/poll) */
+		/* will only take effect when the link state changes */
+		ret = (ppd->dd->cspec->ibcctrl &
+		       SYM_MASK(IBCCtrl, LinkDownDefaultState)) ?
+			IB_LINKINITCMD_SLEEP : IB_LINKINITCMD_POLL;
+		break;
+
+	case QIB_IB_CFG_HRTBT: /* Get Heartbeat off/enable/auto */
+		ret = 0; /* no heartbeat on this chip */
+		break;
+
+	case QIB_IB_CFG_PMA_TICKS:
+		ret = 250; /* 1 usec. */
+		break;
+
+	default:
+		ret =  -EINVAL;
+		break;
+	}
+	return ret;
+}
+
+/*
+ * We assume range checking is already done, if needed.
+ */
+static int qib_6120_set_ib_cfg(struct qib_pportdata *ppd, int which, u32 val)
+{
+	struct qib_devdata *dd = ppd->dd;
+	int ret = 0;
+	u64 val64;
+	u16 lcmd, licmd;
+
+	switch (which) {
+	case QIB_IB_CFG_LWID_ENB:
+		ppd->link_width_enabled = val;
+		break;
+
+	case QIB_IB_CFG_SPD_ENB:
+		ppd->link_speed_enabled = val;
+		break;
+
+	case QIB_IB_CFG_OVERRUN_THRESH: /* IB overrun threshold */
+		val64 = SYM_FIELD(dd->cspec->ibcctrl, IBCCtrl,
+				  OverrunThreshold);
+		if (val64 != val) {
+			dd->cspec->ibcctrl &=
+				~SYM_MASK(IBCCtrl, OverrunThreshold);
+			dd->cspec->ibcctrl |= (u64) val <<
+				SYM_LSB(IBCCtrl, OverrunThreshold);
+			qib_write_kreg(dd, kr_ibcctrl, dd->cspec->ibcctrl);
+			qib_write_kreg(dd, kr_scratch, 0);
+		}
+		break;
+
+	case QIB_IB_CFG_PHYERR_THRESH: /* IB PHY error threshold */
+		val64 = SYM_FIELD(dd->cspec->ibcctrl, IBCCtrl,
+				  PhyerrThreshold);
+		if (val64 != val) {
+			dd->cspec->ibcctrl &=
+				~SYM_MASK(IBCCtrl, PhyerrThreshold);
+			dd->cspec->ibcctrl |= (u64) val <<
+				SYM_LSB(IBCCtrl, PhyerrThreshold);
+			qib_write_kreg(dd, kr_ibcctrl, dd->cspec->ibcctrl);
+			qib_write_kreg(dd, kr_scratch, 0);
+		}
+		break;
+
+	case QIB_IB_CFG_PKEYS: /* update pkeys */
+		val64 = (u64) ppd->pkeys[0] | ((u64) ppd->pkeys[1] << 16) |
+			((u64) ppd->pkeys[2] << 32) |
+			((u64) ppd->pkeys[3] << 48);
+		qib_write_kreg(dd, kr_partitionkey, val64);
+		break;
+
+	case QIB_IB_CFG_LINKDEFAULT: /* IB link default (sleep/poll) */
+		/* will only take effect when the link state changes */
+		if (val == IB_LINKINITCMD_POLL)
+			dd->cspec->ibcctrl &=
+				~SYM_MASK(IBCCtrl, LinkDownDefaultState);
+		else /* SLEEP */
+			dd->cspec->ibcctrl |=
+				SYM_MASK(IBCCtrl, LinkDownDefaultState);
+		qib_write_kreg(dd, kr_ibcctrl, dd->cspec->ibcctrl);
+		qib_write_kreg(dd, kr_scratch, 0);
+		break;
+
+	case QIB_IB_CFG_MTU: /* update the MTU in IBC */
+		/*
+		 * Update our housekeeping variables, and set IBC max
+		 * size, same as init code; max IBC is max we allow in
+		 * buffer, less the qword pbc, plus 1 for ICRC, in dwords
+		 * Set even if it's unchanged, print debug message only
+		 * on changes.
+		 */
+		val = (ppd->ibmaxlen >> 2) + 1;
+		dd->cspec->ibcctrl &= ~SYM_MASK(IBCCtrl, MaxPktLen);
+		dd->cspec->ibcctrl |= (u64)val <<
+			SYM_LSB(IBCCtrl, MaxPktLen);
+		qib_write_kreg(dd, kr_ibcctrl, dd->cspec->ibcctrl);
+		qib_write_kreg(dd, kr_scratch, 0);
+		break;
+
+	case QIB_IB_CFG_LSTATE: /* set the IB link state */
+		switch (val & 0xffff0000) {
+		case IB_LINKCMD_DOWN:
+			lcmd = QLOGIC_IB_IBCC_LINKCMD_DOWN;
+			if (!dd->cspec->ibdeltainprog) {
+				dd->cspec->ibdeltainprog = 1;
+				dd->cspec->ibsymsnap =
+					read_6120_creg32(dd, cr_ibsymbolerr);
+				dd->cspec->iblnkerrsnap =
+					read_6120_creg32(dd, cr_iblinkerrrecov);
+			}
+			break;
+
+		case IB_LINKCMD_ARMED:
+			lcmd = QLOGIC_IB_IBCC_LINKCMD_ARMED;
+			break;
+
+		case IB_LINKCMD_ACTIVE:
+			lcmd = QLOGIC_IB_IBCC_LINKCMD_ACTIVE;
+			break;
+
+		default:
+			ret = -EINVAL;
+			qib_dev_err(dd, "bad linkcmd req 0x%x\n", val >> 16);
+			goto bail;
+		}
+		switch (val & 0xffff) {
+		case IB_LINKINITCMD_NOP:
+			licmd = 0;
+			break;
+
+		case IB_LINKINITCMD_POLL:
+			licmd = QLOGIC_IB_IBCC_LINKINITCMD_POLL;
+			break;
+
+		case IB_LINKINITCMD_SLEEP:
+			licmd = QLOGIC_IB_IBCC_LINKINITCMD_SLEEP;
+			break;
+
+		case IB_LINKINITCMD_DISABLE:
+			licmd = QLOGIC_IB_IBCC_LINKINITCMD_DISABLE;
+			break;
+
+		default:
+			ret = -EINVAL;
+			qib_dev_err(dd, "bad linkinitcmd req 0x%x\n",
+				    val & 0xffff);
+			goto bail;
+		}
+		qib_set_ib_6120_lstate(ppd, lcmd, licmd);
+		goto bail;
+
+	case QIB_IB_CFG_HRTBT:
+		ret = -EINVAL;
+		break;
+
+	default:
+		ret = -EINVAL;
+	}
+bail:
+	return ret;
+}
+
+static int qib_6120_set_loopback(struct qib_pportdata *ppd, const char *what)
+{
+	int ret = 0;
+	if (!strncmp(what, "ibc", 3)) {
+		ppd->dd->cspec->ibcctrl |= SYM_MASK(IBCCtrl, Loopback);
+		qib_devinfo(ppd->dd->pcidev, "Enabling IB%u:%u IBC loopback\n",
+			 ppd->dd->unit, ppd->port);
+	} else if (!strncmp(what, "off", 3)) {
+		ppd->dd->cspec->ibcctrl &= ~SYM_MASK(IBCCtrl, Loopback);
+		qib_devinfo(ppd->dd->pcidev, "Disabling IB%u:%u IBC loopback "
+			    "(normal)\n", ppd->dd->unit, ppd->port);
+	} else
+		ret = -EINVAL;
+	if (!ret) {
+		qib_write_kreg(ppd->dd, kr_ibcctrl, ppd->dd->cspec->ibcctrl);
+		qib_write_kreg(ppd->dd, kr_scratch, 0);
+	}
+	return ret;
+}
+
+static void pma_6120_timer(unsigned long data)
+{
+	struct qib_pportdata *ppd = (struct qib_pportdata *)data;
+	struct qib_chip_specific *cs = ppd->dd->cspec;
+	struct qib_ibport *ibp = &ppd->ibport_data;
+	unsigned long flags;
+
+	spin_lock_irqsave(&ibp->lock, flags);
+	if (cs->pma_sample_status == IB_PMA_SAMPLE_STATUS_STARTED) {
+		cs->pma_sample_status = IB_PMA_SAMPLE_STATUS_RUNNING;
+		qib_snapshot_counters(ppd, &cs->sword, &cs->rword,
+				      &cs->spkts, &cs->rpkts, &cs->xmit_wait);
+		mod_timer(&cs->pma_timer,
+			  jiffies + usecs_to_jiffies(ibp->pma_sample_interval));
+	} else if (cs->pma_sample_status == IB_PMA_SAMPLE_STATUS_RUNNING) {
+		u64 ta, tb, tc, td, te;
+
+		cs->pma_sample_status = IB_PMA_SAMPLE_STATUS_DONE;
+		qib_snapshot_counters(ppd, &ta, &tb, &tc, &td, &te);
+
+		cs->sword = ta - cs->sword;
+		cs->rword = tb - cs->rword;
+		cs->spkts = tc - cs->spkts;
+		cs->rpkts = td - cs->rpkts;
+		cs->xmit_wait = te - cs->xmit_wait;
+	}
+	spin_unlock_irqrestore(&ibp->lock, flags);
+}
+
+/*
+ * Note that the caller has the ibp->lock held.
+ */
+static void qib_set_cntr_6120_sample(struct qib_pportdata *ppd, u32 intv,
+				     u32 start)
+{
+	struct qib_chip_specific *cs = ppd->dd->cspec;
+
+	if (start && intv) {
+		cs->pma_sample_status = IB_PMA_SAMPLE_STATUS_STARTED;
+		mod_timer(&cs->pma_timer, jiffies + usecs_to_jiffies(start));
+	} else if (intv) {
+		cs->pma_sample_status = IB_PMA_SAMPLE_STATUS_RUNNING;
+		qib_snapshot_counters(ppd, &cs->sword, &cs->rword,
+				      &cs->spkts, &cs->rpkts, &cs->xmit_wait);
+		mod_timer(&cs->pma_timer, jiffies + usecs_to_jiffies(intv));
+	} else {
+		cs->pma_sample_status = IB_PMA_SAMPLE_STATUS_DONE;
+		cs->sword = 0;
+		cs->rword = 0;
+		cs->spkts = 0;
+		cs->rpkts = 0;
+		cs->xmit_wait = 0;
+	}
+}
+
+static u32 qib_6120_iblink_state(u64 ibcs)
+{
+	u32 state = (u32)SYM_FIELD(ibcs, IBCStatus, LinkState);
+
+	switch (state) {
+	case IB_6120_L_STATE_INIT:
+		state = IB_PORT_INIT;
+		break;
+	case IB_6120_L_STATE_ARM:
+		state = IB_PORT_ARMED;
+		break;
+	case IB_6120_L_STATE_ACTIVE:
+		/* fall through */
+	case IB_6120_L_STATE_ACT_DEFER:
+		state = IB_PORT_ACTIVE;
+		break;
+	default: /* fall through */
+	case IB_6120_L_STATE_DOWN:
+		state = IB_PORT_DOWN;
+		break;
+	}
+	return state;
+}
+
+/* returns the IBTA port state, rather than the IBC link training state */
+static u8 qib_6120_phys_portstate(u64 ibcs)
+{
+	u8 state = (u8)SYM_FIELD(ibcs, IBCStatus, LinkTrainingState);
+	return qib_6120_physportstate[state];
+}
+
+static int qib_6120_ib_updown(struct qib_pportdata *ppd, int ibup, u64 ibcs)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&ppd->lflags_lock, flags);
+	ppd->lflags &= ~QIBL_IB_FORCE_NOTIFY;
+	spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+
+	if (ibup) {
+		if (ppd->dd->cspec->ibdeltainprog) {
+			ppd->dd->cspec->ibdeltainprog = 0;
+			ppd->dd->cspec->ibsymdelta +=
+				read_6120_creg32(ppd->dd, cr_ibsymbolerr) -
+					ppd->dd->cspec->ibsymsnap;
+			ppd->dd->cspec->iblnkerrdelta +=
+				read_6120_creg32(ppd->dd, cr_iblinkerrrecov) -
+					ppd->dd->cspec->iblnkerrsnap;
+		}
+		qib_hol_init(ppd);
+	} else {
+		ppd->dd->cspec->lli_counter = 0;
+		if (!ppd->dd->cspec->ibdeltainprog) {
+			ppd->dd->cspec->ibdeltainprog = 1;
+			ppd->dd->cspec->ibsymsnap =
+				read_6120_creg32(ppd->dd, cr_ibsymbolerr);
+			ppd->dd->cspec->iblnkerrsnap =
+				read_6120_creg32(ppd->dd, cr_iblinkerrrecov);
+		}
+		qib_hol_down(ppd);
+	}
+
+	qib_6120_setup_setextled(ppd, ibup);
+
+	return 0;
+}
+
+/* Does read/modify/write to appropriate registers to
+ * set output and direction bits selected by mask.
+ * these are in their canonical postions (e.g. lsb of
+ * dir will end up in D48 of extctrl on existing chips).
+ * returns contents of GP Inputs.
+ */
+static int gpio_6120_mod(struct qib_devdata *dd, u32 out, u32 dir, u32 mask)
+{
+	u64 read_val, new_out;
+	unsigned long flags;
+
+	if (mask) {
+		/* some bits being written, lock access to GPIO */
+		dir &= mask;
+		out &= mask;
+		spin_lock_irqsave(&dd->cspec->gpio_lock, flags);
+		dd->cspec->extctrl &= ~((u64)mask << SYM_LSB(EXTCtrl, GPIOOe));
+		dd->cspec->extctrl |= ((u64) dir << SYM_LSB(EXTCtrl, GPIOOe));
+		new_out = (dd->cspec->gpio_out & ~mask) | out;
+
+		qib_write_kreg(dd, kr_extctrl, dd->cspec->extctrl);
+		qib_write_kreg(dd, kr_gpio_out, new_out);
+		dd->cspec->gpio_out = new_out;
+		spin_unlock_irqrestore(&dd->cspec->gpio_lock, flags);
+	}
+	/*
+	 * It is unlikely that a read at this time would get valid
+	 * data on a pin whose direction line was set in the same
+	 * call to this function. We include the read here because
+	 * that allows us to potentially combine a change on one pin with
+	 * a read on another, and because the old code did something like
+	 * this.
+	 */
+	read_val = qib_read_kreg64(dd, kr_extstatus);
+	return SYM_FIELD(read_val, EXTStatus, GPIOIn);
+}
+
+/*
+ * Read fundamental info we need to use the chip.  These are
+ * the registers that describe chip capabilities, and are
+ * saved in shadow registers.
+ */
+static void get_6120_chip_params(struct qib_devdata *dd)
+{
+	u64 val;
+	u32 piobufs;
+	int mtu;
+
+	dd->uregbase = qib_read_kreg32(dd, kr_userregbase);
+
+	dd->rcvtidcnt = qib_read_kreg32(dd, kr_rcvtidcnt);
+	dd->rcvtidbase = qib_read_kreg32(dd, kr_rcvtidbase);
+	dd->rcvegrbase = qib_read_kreg32(dd, kr_rcvegrbase);
+	dd->palign = qib_read_kreg32(dd, kr_palign);
+	dd->piobufbase = qib_read_kreg64(dd, kr_sendpiobufbase);
+	dd->pio2k_bufbase = dd->piobufbase & 0xffffffff;
+
+	dd->rcvhdrcnt = qib_read_kreg32(dd, kr_rcvegrcnt);
+
+	val = qib_read_kreg64(dd, kr_sendpiosize);
+	dd->piosize2k = val & ~0U;
+	dd->piosize4k = val >> 32;
+
+	mtu = ib_mtu_enum_to_int(qib_ibmtu);
+	if (mtu == -1)
+		mtu = QIB_DEFAULT_MTU;
+	dd->pport->ibmtu = (u32)mtu;
+
+	val = qib_read_kreg64(dd, kr_sendpiobufcnt);
+	dd->piobcnt2k = val & ~0U;
+	dd->piobcnt4k = val >> 32;
+	/* these may be adjusted in init_chip_wc_pat() */
+	dd->pio2kbase = (u32 __iomem *)
+		(((char __iomem *)dd->kregbase) + dd->pio2k_bufbase);
+	if (dd->piobcnt4k) {
+		dd->pio4kbase = (u32 __iomem *)
+			(((char __iomem *) dd->kregbase) +
+			 (dd->piobufbase >> 32));
+		/*
+		 * 4K buffers take 2 pages; we use roundup just to be
+		 * paranoid; we calculate it once here, rather than on
+		 * ever buf allocate
+		 */
+		dd->align4k = ALIGN(dd->piosize4k, dd->palign);
+	}
+
+	piobufs = dd->piobcnt4k + dd->piobcnt2k;
+
+	dd->pioavregs = ALIGN(piobufs, sizeof(u64) * BITS_PER_BYTE / 2) /
+		(sizeof(u64) * BITS_PER_BYTE / 2);
+}
+
+/*
+ * The chip base addresses in cspec and cpspec have to be set
+ * after possible init_chip_wc_pat(), rather than in
+ * get_6120_chip_params(), so split out as separate function
+ */
+static void set_6120_baseaddrs(struct qib_devdata *dd)
+{
+	u32 cregbase;
+	cregbase = qib_read_kreg32(dd, kr_counterregbase);
+	dd->cspec->cregbase = (u64 __iomem *)
+		((char __iomem *) dd->kregbase + cregbase);
+
+	dd->egrtidbase = (u64 __iomem *)
+		((char __iomem *) dd->kregbase + dd->rcvegrbase);
+}
+
+/*
+ * Write the final few registers that depend on some of the
+ * init setup.  Done late in init, just before bringing up
+ * the serdes.
+ */
+static int qib_late_6120_initreg(struct qib_devdata *dd)
+{
+	int ret = 0;
+	u64 val;
+
+	qib_write_kreg(dd, kr_rcvhdrentsize, dd->rcvhdrentsize);
+	qib_write_kreg(dd, kr_rcvhdrsize, dd->rcvhdrsize);
+	qib_write_kreg(dd, kr_rcvhdrcnt, dd->rcvhdrcnt);
+	qib_write_kreg(dd, kr_sendpioavailaddr, dd->pioavailregs_phys);
+	val = qib_read_kreg64(dd, kr_sendpioavailaddr);
+	if (val != dd->pioavailregs_phys) {
+		qib_dev_err(dd, "Catastrophic software error, "
+			    "SendPIOAvailAddr written as %lx, "
+			    "read back as %llx\n",
+			    (unsigned long) dd->pioavailregs_phys,
+			    (unsigned long long) val);
+		ret = -EINVAL;
+	}
+	return ret;
+}
+
+static int init_6120_variables(struct qib_devdata *dd)
+{
+	int ret = 0;
+	struct qib_pportdata *ppd;
+	u32 sbufs;
+
+	ppd = (struct qib_pportdata *)(dd + 1);
+	dd->pport = ppd;
+	dd->num_pports = 1;
+
+	dd->cspec = (struct qib_chip_specific *)(ppd + dd->num_pports);
+	ppd->cpspec = NULL; /* not used in this chip */
+
+	spin_lock_init(&dd->cspec->kernel_tid_lock);
+	spin_lock_init(&dd->cspec->user_tid_lock);
+	spin_lock_init(&dd->cspec->rcvmod_lock);
+	spin_lock_init(&dd->cspec->gpio_lock);
+
+	/* we haven't yet set QIB_PRESENT, so use read directly */
+	dd->revision = readq(&dd->kregbase[kr_revision]);
+
+	if ((dd->revision & 0xffffffffU) == 0xffffffffU) {
+		qib_dev_err(dd, "Revision register read failure, "
+			    "giving up initialization\n");
+		ret = -ENODEV;
+		goto bail;
+	}
+	dd->flags |= QIB_PRESENT;  /* now register routines work */
+
+	dd->majrev = (u8) SYM_FIELD(dd->revision, Revision_R,
+				    ChipRevMajor);
+	dd->minrev = (u8) SYM_FIELD(dd->revision, Revision_R,
+				    ChipRevMinor);
+
+	get_6120_chip_params(dd);
+	pe_boardname(dd); /* fill in boardname */
+
+	/*
+	 * GPIO bits for TWSI data and clock,
+	 * used for serial EEPROM.
+	 */
+	dd->gpio_sda_num = _QIB_GPIO_SDA_NUM;
+	dd->gpio_scl_num = _QIB_GPIO_SCL_NUM;
+	dd->twsi_eeprom_dev = QIB_TWSI_NO_DEV;
+
+	if (qib_unordered_wc())
+		dd->flags |= QIB_PIO_FLUSH_WC;
+
+	/*
+	 * EEPROM error log 0 is TXE Parity errors. 1 is RXE Parity.
+	 * 2 is Some Misc, 3 is reserved for future.
+	 */
+	dd->eep_st_masks[0].hwerrs_to_log = HWE_MASK(TXEMemParityErr);
+
+	/* Ignore errors in PIO/PBC on systems with unordered write-combining */
+	if (qib_unordered_wc())
+		dd->eep_st_masks[0].hwerrs_to_log &= ~TXE_PIO_PARITY;
+
+	dd->eep_st_masks[1].hwerrs_to_log = HWE_MASK(RXEMemParityErr);
+
+	dd->eep_st_masks[2].errs_to_log = ERR_MASK(ResetNegated);
+
+	qib_init_pportdata(ppd, dd, 0, 1);
+	ppd->link_width_supported = IB_WIDTH_1X | IB_WIDTH_4X;
+	ppd->link_speed_supported = QIB_IB_SDR;
+	ppd->link_width_enabled = IB_WIDTH_4X;
+	ppd->link_speed_enabled = ppd->link_speed_supported;
+	/* these can't change for this chip, so set once */
+	ppd->link_width_active = ppd->link_width_enabled;
+	ppd->link_speed_active = ppd->link_speed_enabled;
+	ppd->vls_supported = IB_VL_VL0;
+	ppd->vls_operational = ppd->vls_supported;
+
+	dd->rcvhdrentsize = QIB_RCVHDR_ENTSIZE;
+	dd->rcvhdrsize = QIB_DFLT_RCVHDRSIZE;
+	dd->rhf_offset = 0;
+
+	/* we always allocate at least 2048 bytes for eager buffers */
+	ret = ib_mtu_enum_to_int(qib_ibmtu);
+	dd->rcvegrbufsize = ret != -1 ? max(ret, 2048) : QIB_DEFAULT_MTU;
+
+	qib_6120_tidtemplate(dd);
+
+	/*
+	 * We can request a receive interrupt for 1 or
+	 * more packets from current offset.  For now, we set this
+	 * up for a single packet.
+	 */
+	dd->rhdrhead_intr_off = 1ULL << 32;
+
+	/* setup the stats timer; the add_timer is done at end of init */
+	init_timer(&dd->stats_timer);
+	dd->stats_timer.function = qib_get_6120_faststats;
+	dd->stats_timer.data = (unsigned long) dd;
+
+	init_timer(&dd->cspec->pma_timer);
+	dd->cspec->pma_timer.function = pma_6120_timer;
+	dd->cspec->pma_timer.data = (unsigned long) ppd;
+
+	dd->ureg_align = qib_read_kreg32(dd, kr_palign);
+
+	dd->piosize2kmax_dwords = dd->piosize2k >> 2;
+	qib_6120_config_ctxts(dd);
+	qib_set_ctxtcnt(dd);
+
+	if (qib_wc_pat) {
+		ret = init_chip_wc_pat(dd, 0);
+		if (ret)
+			goto bail;
+	}
+	set_6120_baseaddrs(dd); /* set chip access pointers now */
+
+	ret = 0;
+	if (qib_mini_init)
+		goto bail;
+
+	qib_num_cfg_vls = 1; /* if any 6120's, only one VL */
+
+	ret = qib_create_ctxts(dd);
+	init_6120_cntrnames(dd);
+
+	/* use all of 4KB buffers for the kernel, otherwise 16 */
+	sbufs = dd->piobcnt4k ?  dd->piobcnt4k : 16;
+
+	dd->lastctxt_piobuf = dd->piobcnt2k + dd->piobcnt4k - sbufs;
+	dd->pbufsctxt = dd->lastctxt_piobuf /
+		(dd->cfgctxts - dd->first_user_ctxt);
+
+	if (ret)
+		goto bail;
+bail:
+	return ret;
+}
+
+/*
+ * For this chip, we want to use the same buffer every time
+ * when we are trying to bring the link up (they are always VL15
+ * packets).  At that link state the packet should always go out immediately
+ * (or at least be discarded at the tx interface if the link is down).
+ * If it doesn't, and the buffer isn't available, that means some other
+ * sender has gotten ahead of us, and is preventing our packet from going
+ * out.  In that case, we flush all packets, and try again.  If that still
+ * fails, we fail the request, and hope things work the next time around.
+ *
+ * We don't need very complicated heuristics on whether the packet had
+ * time to go out or not, since even at SDR 1X, it goes out in very short
+ * time periods, covered by the chip reads done here and as part of the
+ * flush.
+ */
+static u32 __iomem *get_6120_link_buf(struct qib_pportdata *ppd, u32 *bnum)
+{
+	u32 __iomem *buf;
+	u32 lbuf = ppd->dd->piobcnt2k + ppd->dd->piobcnt4k - 1;
+
+	/*
+	 * always blip to get avail list updated, since it's almost
+	 * always needed, and is fairly cheap.
+	 */
+	sendctrl_6120_mod(ppd->dd->pport, QIB_SENDCTRL_AVAIL_BLIP);
+	qib_read_kreg64(ppd->dd, kr_scratch); /* extra chip flush */
+	buf = qib_getsendbuf_range(ppd->dd, bnum, lbuf, lbuf);
+	if (buf)
+		goto done;
+
+	sendctrl_6120_mod(ppd, QIB_SENDCTRL_DISARM_ALL | QIB_SENDCTRL_FLUSH |
+			  QIB_SENDCTRL_AVAIL_BLIP);
+	ppd->dd->upd_pio_shadow  = 1; /* update our idea of what's busy */
+	qib_read_kreg64(ppd->dd, kr_scratch); /* extra chip flush */
+	buf = qib_getsendbuf_range(ppd->dd, bnum, lbuf, lbuf);
+done:
+	return buf;
+}
+
+static u32 __iomem *qib_6120_getsendbuf(struct qib_pportdata *ppd, u64 pbc,
+					u32 *pbufnum)
+{
+	u32 first, last, plen = pbc & QIB_PBC_LENGTH_MASK;
+	struct qib_devdata *dd = ppd->dd;
+	u32 __iomem *buf;
+
+	if (((pbc >> 32) & PBC_6120_VL15_SEND_CTRL) &&
+		!(ppd->lflags & (QIBL_IB_AUTONEG_INPROG | QIBL_LINKACTIVE)))
+		buf = get_6120_link_buf(ppd, pbufnum);
+	else {
+
+		if ((plen + 1) > dd->piosize2kmax_dwords)
+			first = dd->piobcnt2k;
+		else
+			first = 0;
+		/* try 4k if all 2k busy, so same last for both sizes */
+		last = dd->piobcnt2k + dd->piobcnt4k - 1;
+		buf = qib_getsendbuf_range(dd, pbufnum, first, last);
+	}
+	return buf;
+}
+
+static int init_sdma_6120_regs(struct qib_pportdata *ppd)
+{
+	return -ENODEV;
+}
+
+static u16 qib_sdma_6120_gethead(struct qib_pportdata *ppd)
+{
+	return 0;
+}
+
+static int qib_sdma_6120_busy(struct qib_pportdata *ppd)
+{
+	return 0;
+}
+
+static void qib_sdma_update_6120_tail(struct qib_pportdata *ppd, u16 tail)
+{
+}
+
+static void qib_6120_sdma_sendctrl(struct qib_pportdata *ppd, unsigned op)
+{
+}
+
+static void qib_sdma_set_6120_desc_cnt(struct qib_pportdata *ppd, unsigned cnt)
+{
+}
+
+/*
+ * the pbc doesn't need a VL15 indicator, but we need it for link_buf.
+ * The chip ignores the bit if set.
+ */
+static u32 qib_6120_setpbc_control(struct qib_pportdata *ppd, u32 plen,
+				   u8 srate, u8 vl)
+{
+	return vl == 15 ? PBC_6120_VL15_SEND_CTRL : 0;
+}
+
+static void qib_6120_initvl15_bufs(struct qib_devdata *dd)
+{
+}
+
+static void qib_6120_init_ctxt(struct qib_ctxtdata *rcd)
+{
+	rcd->rcvegrcnt = rcd->dd->rcvhdrcnt;
+	rcd->rcvegr_tid_base = rcd->ctxt * rcd->rcvegrcnt;
+}
+
+static void qib_6120_txchk_change(struct qib_devdata *dd, u32 start,
+	u32 len, u32 avail, struct qib_ctxtdata *rcd)
+{
+}
+
+static void writescratch(struct qib_devdata *dd, u32 val)
+{
+	(void) qib_write_kreg(dd, kr_scratch, val);
+}
+
+static int qib_6120_tempsense_rd(struct qib_devdata *dd, int regnum)
+{
+	return -ENXIO;
+}
+
+/* Dummy function, as 6120 boards never disable EEPROM Write */
+static int qib_6120_eeprom_wen(struct qib_devdata *dd, int wen)
+{
+	return 1;
+}
+
+/**
+ * qib_init_iba6120_funcs - set up the chip-specific function pointers
+ * @pdev: pci_dev of the qlogic_ib device
+ * @ent: pci_device_id matching this chip
+ *
+ * This is global, and is called directly at init to set up the
+ * chip-specific function pointers for later use.
+ *
+ * It also allocates/partially-inits the qib_devdata struct for
+ * this device.
+ */
+struct qib_devdata *qib_init_iba6120_funcs(struct pci_dev *pdev,
+					   const struct pci_device_id *ent)
+{
+	struct qib_devdata *dd;
+	int ret;
+
+	dd = qib_alloc_devdata(pdev, sizeof(struct qib_pportdata) +
+			       sizeof(struct qib_chip_specific));
+	if (IS_ERR(dd))
+		goto bail;
+
+	dd->f_bringup_serdes    = qib_6120_bringup_serdes;
+	dd->f_cleanup           = qib_6120_setup_cleanup;
+	dd->f_clear_tids        = qib_6120_clear_tids;
+	dd->f_free_irq          = qib_6120_free_irq;
+	dd->f_get_base_info     = qib_6120_get_base_info;
+	dd->f_get_msgheader     = qib_6120_get_msgheader;
+	dd->f_getsendbuf        = qib_6120_getsendbuf;
+	dd->f_gpio_mod          = gpio_6120_mod;
+	dd->f_eeprom_wen	= qib_6120_eeprom_wen;
+	dd->f_hdrqempty         = qib_6120_hdrqempty;
+	dd->f_ib_updown         = qib_6120_ib_updown;
+	dd->f_init_ctxt         = qib_6120_init_ctxt;
+	dd->f_initvl15_bufs     = qib_6120_initvl15_bufs;
+	dd->f_intr_fallback     = qib_6120_nointr_fallback;
+	dd->f_late_initreg      = qib_late_6120_initreg;
+	dd->f_setpbc_control    = qib_6120_setpbc_control;
+	dd->f_portcntr          = qib_portcntr_6120;
+	dd->f_put_tid           = (dd->minrev >= 2) ?
+				      qib_6120_put_tid_2 :
+				      qib_6120_put_tid;
+	dd->f_quiet_serdes      = qib_6120_quiet_serdes;
+	dd->f_rcvctrl           = rcvctrl_6120_mod;
+	dd->f_read_cntrs        = qib_read_6120cntrs;
+	dd->f_read_portcntrs    = qib_read_6120portcntrs;
+	dd->f_reset             = qib_6120_setup_reset;
+	dd->f_init_sdma_regs    = init_sdma_6120_regs;
+	dd->f_sdma_busy         = qib_sdma_6120_busy;
+	dd->f_sdma_gethead      = qib_sdma_6120_gethead;
+	dd->f_sdma_sendctrl     = qib_6120_sdma_sendctrl;
+	dd->f_sdma_set_desc_cnt = qib_sdma_set_6120_desc_cnt;
+	dd->f_sdma_update_tail  = qib_sdma_update_6120_tail;
+	dd->f_sendctrl          = sendctrl_6120_mod;
+	dd->f_set_armlaunch     = qib_set_6120_armlaunch;
+	dd->f_set_cntr_sample   = qib_set_cntr_6120_sample;
+	dd->f_iblink_state      = qib_6120_iblink_state;
+	dd->f_ibphys_portstate  = qib_6120_phys_portstate;
+	dd->f_get_ib_cfg        = qib_6120_get_ib_cfg;
+	dd->f_set_ib_cfg        = qib_6120_set_ib_cfg;
+	dd->f_set_ib_loopback   = qib_6120_set_loopback;
+	dd->f_set_intr_state    = qib_6120_set_intr_state;
+	dd->f_setextled         = qib_6120_setup_setextled;
+	dd->f_txchk_change      = qib_6120_txchk_change;
+	dd->f_update_usrhead    = qib_update_6120_usrhead;
+	dd->f_wantpiobuf_intr   = qib_wantpiobuf_6120_intr;
+	dd->f_xgxs_reset        = qib_6120_xgxs_reset;
+	dd->f_writescratch      = writescratch;
+	dd->f_tempsense_rd	= qib_6120_tempsense_rd;
+	/*
+	 * Do remaining pcie setup and save pcie values in dd.
+	 * Any error printing is already done by the init code.
+	 * On return, we have the chip mapped and accessible,
+	 * but chip registers are not set up until start of
+	 * init_6120_variables.
+	 */
+	ret = qib_pcie_ddinit(dd, pdev, ent);
+	if (ret < 0)
+		goto bail_free;
+
+	/* initialize chip-specific variables */
+	ret = init_6120_variables(dd);
+	if (ret)
+		goto bail_cleanup;
+
+	if (qib_mini_init)
+		goto bail;
+
+	if (qib_pcie_params(dd, 8, NULL, NULL))
+		qib_dev_err(dd, "Failed to setup PCIe or interrupts; "
+			    "continuing anyway\n");
+	dd->cspec->irq = pdev->irq; /* save IRQ */
+
+	/* clear diagctrl register, in case diags were running and crashed */
+	qib_write_kreg(dd, kr_hwdiagctrl, 0);
+
+	if (qib_read_kreg64(dd, kr_hwerrstatus) &
+	    QLOGIC_IB_HWE_SERDESPLLFAILED)
+		qib_write_kreg(dd, kr_hwerrclear,
+			       QLOGIC_IB_HWE_SERDESPLLFAILED);
+
+	/* setup interrupt handler (interrupt type handled above) */
+	qib_setup_6120_interrupt(dd);
+	/* Note that qpn_mask is set by qib_6120_config_ctxts() first */
+	qib_6120_init_hwerrors(dd);
+
+	goto bail;
+
+bail_cleanup:
+	qib_pcie_ddcleanup(dd);
+bail_free:
+	qib_free_devdata(dd);
+	dd = ERR_PTR(ret);
+bail:
+	return dd;
+}
diff --git a/drivers/infiniband/hw/qib/qib_iba7220.c b/drivers/infiniband/hw/qib/qib_iba7220.c
new file mode 100644
index 000000000000..6fd8d74e7392
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_iba7220.c
@@ -0,0 +1,4618 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009, 2010 QLogic Corporation.
+ * All rights reserved.
+ * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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.
+ */
+/*
+ * This file contains all of the code that is specific to the
+ * QLogic_IB 7220 chip (except that specific to the SerDes)
+ */
+
+#include <linux/interrupt.h>
+#include <linux/pci.h>
+#include <linux/delay.h>
+#include <linux/io.h>
+#include <rdma/ib_verbs.h>
+
+#include "qib.h"
+#include "qib_7220.h"
+
+static void qib_setup_7220_setextled(struct qib_pportdata *, u32);
+static void qib_7220_handle_hwerrors(struct qib_devdata *, char *, size_t);
+static void sendctrl_7220_mod(struct qib_pportdata *ppd, u32 op);
+static u32 qib_7220_iblink_state(u64);
+static u8 qib_7220_phys_portstate(u64);
+static void qib_sdma_update_7220_tail(struct qib_pportdata *, u16);
+static void qib_set_ib_7220_lstate(struct qib_pportdata *, u16, u16);
+
+/*
+ * This file contains almost all the chip-specific register information and
+ * access functions for the QLogic QLogic_IB 7220 PCI-Express chip, with the
+ * exception of SerDes support, which in in qib_sd7220.c.
+ */
+
+/* Below uses machine-generated qib_chipnum_regs.h file */
+#define KREG_IDX(regname) (QIB_7220_##regname##_OFFS / sizeof(u64))
+
+/* Use defines to tie machine-generated names to lower-case names */
+#define kr_control KREG_IDX(Control)
+#define kr_counterregbase KREG_IDX(CntrRegBase)
+#define kr_errclear KREG_IDX(ErrClear)
+#define kr_errmask KREG_IDX(ErrMask)
+#define kr_errstatus KREG_IDX(ErrStatus)
+#define kr_extctrl KREG_IDX(EXTCtrl)
+#define kr_extstatus KREG_IDX(EXTStatus)
+#define kr_gpio_clear KREG_IDX(GPIOClear)
+#define kr_gpio_mask KREG_IDX(GPIOMask)
+#define kr_gpio_out KREG_IDX(GPIOOut)
+#define kr_gpio_status KREG_IDX(GPIOStatus)
+#define kr_hrtbt_guid KREG_IDX(HRTBT_GUID)
+#define kr_hwdiagctrl KREG_IDX(HwDiagCtrl)
+#define kr_hwerrclear KREG_IDX(HwErrClear)
+#define kr_hwerrmask KREG_IDX(HwErrMask)
+#define kr_hwerrstatus KREG_IDX(HwErrStatus)
+#define kr_ibcctrl KREG_IDX(IBCCtrl)
+#define kr_ibcddrctrl KREG_IDX(IBCDDRCtrl)
+#define kr_ibcddrstatus KREG_IDX(IBCDDRStatus)
+#define kr_ibcstatus KREG_IDX(IBCStatus)
+#define kr_ibserdesctrl KREG_IDX(IBSerDesCtrl)
+#define kr_intclear KREG_IDX(IntClear)
+#define kr_intmask KREG_IDX(IntMask)
+#define kr_intstatus KREG_IDX(IntStatus)
+#define kr_ncmodectrl KREG_IDX(IBNCModeCtrl)
+#define kr_palign KREG_IDX(PageAlign)
+#define kr_partitionkey KREG_IDX(RcvPartitionKey)
+#define kr_portcnt KREG_IDX(PortCnt)
+#define kr_rcvbthqp KREG_IDX(RcvBTHQP)
+#define kr_rcvctrl KREG_IDX(RcvCtrl)
+#define kr_rcvegrbase KREG_IDX(RcvEgrBase)
+#define kr_rcvegrcnt KREG_IDX(RcvEgrCnt)
+#define kr_rcvhdrcnt KREG_IDX(RcvHdrCnt)
+#define kr_rcvhdrentsize KREG_IDX(RcvHdrEntSize)
+#define kr_rcvhdrsize KREG_IDX(RcvHdrSize)
+#define kr_rcvpktledcnt KREG_IDX(RcvPktLEDCnt)
+#define kr_rcvtidbase KREG_IDX(RcvTIDBase)
+#define kr_rcvtidcnt KREG_IDX(RcvTIDCnt)
+#define kr_revision KREG_IDX(Revision)
+#define kr_scratch KREG_IDX(Scratch)
+#define kr_sendbuffererror KREG_IDX(SendBufErr0)
+#define kr_sendctrl KREG_IDX(SendCtrl)
+#define kr_senddmabase KREG_IDX(SendDmaBase)
+#define kr_senddmabufmask0 KREG_IDX(SendDmaBufMask0)
+#define kr_senddmabufmask1 (KREG_IDX(SendDmaBufMask0) + 1)
+#define kr_senddmabufmask2 (KREG_IDX(SendDmaBufMask0) + 2)
+#define kr_senddmahead KREG_IDX(SendDmaHead)
+#define kr_senddmaheadaddr KREG_IDX(SendDmaHeadAddr)
+#define kr_senddmalengen KREG_IDX(SendDmaLenGen)
+#define kr_senddmastatus KREG_IDX(SendDmaStatus)
+#define kr_senddmatail KREG_IDX(SendDmaTail)
+#define kr_sendpioavailaddr KREG_IDX(SendBufAvailAddr)
+#define kr_sendpiobufbase KREG_IDX(SendBufBase)
+#define kr_sendpiobufcnt KREG_IDX(SendBufCnt)
+#define kr_sendpiosize KREG_IDX(SendBufSize)
+#define kr_sendregbase KREG_IDX(SendRegBase)
+#define kr_userregbase KREG_IDX(UserRegBase)
+#define kr_xgxs_cfg KREG_IDX(XGXSCfg)
+
+/* These must only be written via qib_write_kreg_ctxt() */
+#define kr_rcvhdraddr KREG_IDX(RcvHdrAddr0)
+#define kr_rcvhdrtailaddr KREG_IDX(RcvHdrTailAddr0)
+
+
+#define CREG_IDX(regname) ((QIB_7220_##regname##_OFFS - \
+			QIB_7220_LBIntCnt_OFFS) / sizeof(u64))
+
+#define cr_badformat CREG_IDX(RxVersionErrCnt)
+#define cr_erricrc CREG_IDX(RxICRCErrCnt)
+#define cr_errlink CREG_IDX(RxLinkMalformCnt)
+#define cr_errlpcrc CREG_IDX(RxLPCRCErrCnt)
+#define cr_errpkey CREG_IDX(RxPKeyMismatchCnt)
+#define cr_rcvflowctrl_err CREG_IDX(RxFlowCtrlViolCnt)
+#define cr_err_rlen CREG_IDX(RxLenErrCnt)
+#define cr_errslen CREG_IDX(TxLenErrCnt)
+#define cr_errtidfull CREG_IDX(RxTIDFullErrCnt)
+#define cr_errtidvalid CREG_IDX(RxTIDValidErrCnt)
+#define cr_errvcrc CREG_IDX(RxVCRCErrCnt)
+#define cr_ibstatuschange CREG_IDX(IBStatusChangeCnt)
+#define cr_lbint CREG_IDX(LBIntCnt)
+#define cr_invalidrlen CREG_IDX(RxMaxMinLenErrCnt)
+#define cr_invalidslen CREG_IDX(TxMaxMinLenErrCnt)
+#define cr_lbflowstall CREG_IDX(LBFlowStallCnt)
+#define cr_pktrcv CREG_IDX(RxDataPktCnt)
+#define cr_pktrcvflowctrl CREG_IDX(RxFlowPktCnt)
+#define cr_pktsend CREG_IDX(TxDataPktCnt)
+#define cr_pktsendflow CREG_IDX(TxFlowPktCnt)
+#define cr_portovfl CREG_IDX(RxP0HdrEgrOvflCnt)
+#define cr_rcvebp CREG_IDX(RxEBPCnt)
+#define cr_rcvovfl CREG_IDX(RxBufOvflCnt)
+#define cr_senddropped CREG_IDX(TxDroppedPktCnt)
+#define cr_sendstall CREG_IDX(TxFlowStallCnt)
+#define cr_sendunderrun CREG_IDX(TxUnderrunCnt)
+#define cr_wordrcv CREG_IDX(RxDwordCnt)
+#define cr_wordsend CREG_IDX(TxDwordCnt)
+#define cr_txunsupvl CREG_IDX(TxUnsupVLErrCnt)
+#define cr_rxdroppkt CREG_IDX(RxDroppedPktCnt)
+#define cr_iblinkerrrecov CREG_IDX(IBLinkErrRecoveryCnt)
+#define cr_iblinkdown CREG_IDX(IBLinkDownedCnt)
+#define cr_ibsymbolerr CREG_IDX(IBSymbolErrCnt)
+#define cr_vl15droppedpkt CREG_IDX(RxVL15DroppedPktCnt)
+#define cr_rxotherlocalphyerr CREG_IDX(RxOtherLocalPhyErrCnt)
+#define cr_excessbufferovfl CREG_IDX(ExcessBufferOvflCnt)
+#define cr_locallinkintegrityerr CREG_IDX(LocalLinkIntegrityErrCnt)
+#define cr_rxvlerr CREG_IDX(RxVlErrCnt)
+#define cr_rxdlidfltr CREG_IDX(RxDlidFltrCnt)
+#define cr_psstat CREG_IDX(PSStat)
+#define cr_psstart CREG_IDX(PSStart)
+#define cr_psinterval CREG_IDX(PSInterval)
+#define cr_psrcvdatacount CREG_IDX(PSRcvDataCount)
+#define cr_psrcvpktscount CREG_IDX(PSRcvPktsCount)
+#define cr_psxmitdatacount CREG_IDX(PSXmitDataCount)
+#define cr_psxmitpktscount CREG_IDX(PSXmitPktsCount)
+#define cr_psxmitwaitcount CREG_IDX(PSXmitWaitCount)
+#define cr_txsdmadesc CREG_IDX(TxSDmaDescCnt)
+#define cr_pcieretrydiag CREG_IDX(PcieRetryBufDiagQwordCnt)
+
+#define SYM_RMASK(regname, fldname) ((u64)              \
+	QIB_7220_##regname##_##fldname##_RMASK)
+#define SYM_MASK(regname, fldname) ((u64)               \
+	QIB_7220_##regname##_##fldname##_RMASK <<       \
+	 QIB_7220_##regname##_##fldname##_LSB)
+#define SYM_LSB(regname, fldname) (QIB_7220_##regname##_##fldname##_LSB)
+#define SYM_FIELD(value, regname, fldname) ((u64) \
+	(((value) >> SYM_LSB(regname, fldname)) & \
+	 SYM_RMASK(regname, fldname)))
+#define ERR_MASK(fldname) SYM_MASK(ErrMask, fldname##Mask)
+#define HWE_MASK(fldname) SYM_MASK(HwErrMask, fldname##Mask)
+
+/* ibcctrl bits */
+#define QLOGIC_IB_IBCC_LINKINITCMD_DISABLE 1
+/* cycle through TS1/TS2 till OK */
+#define QLOGIC_IB_IBCC_LINKINITCMD_POLL 2
+/* wait for TS1, then go on */
+#define QLOGIC_IB_IBCC_LINKINITCMD_SLEEP 3
+#define QLOGIC_IB_IBCC_LINKINITCMD_SHIFT 16
+
+#define QLOGIC_IB_IBCC_LINKCMD_DOWN 1           /* move to 0x11 */
+#define QLOGIC_IB_IBCC_LINKCMD_ARMED 2          /* move to 0x21 */
+#define QLOGIC_IB_IBCC_LINKCMD_ACTIVE 3 /* move to 0x31 */
+
+#define BLOB_7220_IBCHG 0x81
+
+/*
+ * We could have a single register get/put routine, that takes a group type,
+ * but this is somewhat clearer and cleaner.  It also gives us some error
+ * checking.  64 bit register reads should always work, but are inefficient
+ * on opteron (the northbridge always generates 2 separate HT 32 bit reads),
+ * so we use kreg32 wherever possible.  User register and counter register
+ * reads are always 32 bit reads, so only one form of those routines.
+ */
+
+/**
+ * qib_read_ureg32 - read 32-bit virtualized per-context register
+ * @dd: device
+ * @regno: register number
+ * @ctxt: context number
+ *
+ * Return the contents of a register that is virtualized to be per context.
+ * Returns -1 on errors (not distinguishable from valid contents at
+ * runtime; we may add a separate error variable at some point).
+ */
+static inline u32 qib_read_ureg32(const struct qib_devdata *dd,
+				  enum qib_ureg regno, int ctxt)
+{
+	if (!dd->kregbase || !(dd->flags & QIB_PRESENT))
+		return 0;
+
+	if (dd->userbase)
+		return readl(regno + (u64 __iomem *)
+			     ((char __iomem *)dd->userbase +
+			      dd->ureg_align * ctxt));
+	else
+		return readl(regno + (u64 __iomem *)
+			     (dd->uregbase +
+			      (char __iomem *)dd->kregbase +
+			      dd->ureg_align * ctxt));
+}
+
+/**
+ * qib_write_ureg - write 32-bit virtualized per-context register
+ * @dd: device
+ * @regno: register number
+ * @value: value
+ * @ctxt: context
+ *
+ * Write the contents of a register that is virtualized to be per context.
+ */
+static inline void qib_write_ureg(const struct qib_devdata *dd,
+				  enum qib_ureg regno, u64 value, int ctxt)
+{
+	u64 __iomem *ubase;
+
+	if (dd->userbase)
+		ubase = (u64 __iomem *)
+			((char __iomem *) dd->userbase +
+			 dd->ureg_align * ctxt);
+	else
+		ubase = (u64 __iomem *)
+			(dd->uregbase +
+			 (char __iomem *) dd->kregbase +
+			 dd->ureg_align * ctxt);
+
+	if (dd->kregbase && (dd->flags & QIB_PRESENT))
+		writeq(value, &ubase[regno]);
+}
+
+/**
+ * qib_write_kreg_ctxt - write a device's per-ctxt 64-bit kernel register
+ * @dd: the qlogic_ib device
+ * @regno: the register number to write
+ * @ctxt: the context containing the register
+ * @value: the value to write
+ */
+static inline void qib_write_kreg_ctxt(const struct qib_devdata *dd,
+				       const u16 regno, unsigned ctxt,
+				       u64 value)
+{
+	qib_write_kreg(dd, regno + ctxt, value);
+}
+
+static inline void write_7220_creg(const struct qib_devdata *dd,
+				   u16 regno, u64 value)
+{
+	if (dd->cspec->cregbase && (dd->flags & QIB_PRESENT))
+		writeq(value, &dd->cspec->cregbase[regno]);
+}
+
+static inline u64 read_7220_creg(const struct qib_devdata *dd, u16 regno)
+{
+	if (!dd->cspec->cregbase || !(dd->flags & QIB_PRESENT))
+		return 0;
+	return readq(&dd->cspec->cregbase[regno]);
+}
+
+static inline u32 read_7220_creg32(const struct qib_devdata *dd, u16 regno)
+{
+	if (!dd->cspec->cregbase || !(dd->flags & QIB_PRESENT))
+		return 0;
+	return readl(&dd->cspec->cregbase[regno]);
+}
+
+/* kr_revision bits */
+#define QLOGIC_IB_R_EMULATORREV_MASK ((1ULL << 22) - 1)
+#define QLOGIC_IB_R_EMULATORREV_SHIFT 40
+
+/* kr_control bits */
+#define QLOGIC_IB_C_RESET (1U << 7)
+
+/* kr_intstatus, kr_intclear, kr_intmask bits */
+#define QLOGIC_IB_I_RCVURG_MASK ((1ULL << 17) - 1)
+#define QLOGIC_IB_I_RCVURG_SHIFT 32
+#define QLOGIC_IB_I_RCVAVAIL_MASK ((1ULL << 17) - 1)
+#define QLOGIC_IB_I_RCVAVAIL_SHIFT 0
+#define QLOGIC_IB_I_SERDESTRIMDONE (1ULL << 27)
+
+#define QLOGIC_IB_C_FREEZEMODE 0x00000002
+#define QLOGIC_IB_C_LINKENABLE 0x00000004
+
+#define QLOGIC_IB_I_SDMAINT             0x8000000000000000ULL
+#define QLOGIC_IB_I_SDMADISABLED        0x4000000000000000ULL
+#define QLOGIC_IB_I_ERROR               0x0000000080000000ULL
+#define QLOGIC_IB_I_SPIOSENT            0x0000000040000000ULL
+#define QLOGIC_IB_I_SPIOBUFAVAIL        0x0000000020000000ULL
+#define QLOGIC_IB_I_GPIO                0x0000000010000000ULL
+
+/* variables for sanity checking interrupt and errors */
+#define QLOGIC_IB_I_BITSEXTANT \
+		(QLOGIC_IB_I_SDMAINT | QLOGIC_IB_I_SDMADISABLED | \
+		(QLOGIC_IB_I_RCVURG_MASK << QLOGIC_IB_I_RCVURG_SHIFT) | \
+		(QLOGIC_IB_I_RCVAVAIL_MASK << \
+		 QLOGIC_IB_I_RCVAVAIL_SHIFT) | \
+		QLOGIC_IB_I_ERROR | QLOGIC_IB_I_SPIOSENT | \
+		QLOGIC_IB_I_SPIOBUFAVAIL | QLOGIC_IB_I_GPIO | \
+		QLOGIC_IB_I_SERDESTRIMDONE)
+
+#define IB_HWE_BITSEXTANT \
+	       (HWE_MASK(RXEMemParityErr) | \
+		HWE_MASK(TXEMemParityErr) | \
+		(QLOGIC_IB_HWE_PCIEMEMPARITYERR_MASK <<	 \
+		 QLOGIC_IB_HWE_PCIEMEMPARITYERR_SHIFT) | \
+		QLOGIC_IB_HWE_PCIE1PLLFAILED | \
+		QLOGIC_IB_HWE_PCIE0PLLFAILED | \
+		QLOGIC_IB_HWE_PCIEPOISONEDTLP | \
+		QLOGIC_IB_HWE_PCIECPLTIMEOUT | \
+		QLOGIC_IB_HWE_PCIEBUSPARITYXTLH | \
+		QLOGIC_IB_HWE_PCIEBUSPARITYXADM | \
+		QLOGIC_IB_HWE_PCIEBUSPARITYRADM | \
+		HWE_MASK(PowerOnBISTFailed) |	  \
+		QLOGIC_IB_HWE_COREPLL_FBSLIP | \
+		QLOGIC_IB_HWE_COREPLL_RFSLIP | \
+		QLOGIC_IB_HWE_SERDESPLLFAILED | \
+		HWE_MASK(IBCBusToSPCParityErr) | \
+		HWE_MASK(IBCBusFromSPCParityErr) | \
+		QLOGIC_IB_HWE_PCIECPLDATAQUEUEERR | \
+		QLOGIC_IB_HWE_PCIECPLHDRQUEUEERR | \
+		QLOGIC_IB_HWE_SDMAMEMREADERR | \
+		QLOGIC_IB_HWE_CLK_UC_PLLNOTLOCKED | \
+		QLOGIC_IB_HWE_PCIESERDESQ0PCLKNOTDETECT | \
+		QLOGIC_IB_HWE_PCIESERDESQ1PCLKNOTDETECT | \
+		QLOGIC_IB_HWE_PCIESERDESQ2PCLKNOTDETECT | \
+		QLOGIC_IB_HWE_PCIESERDESQ3PCLKNOTDETECT | \
+		QLOGIC_IB_HWE_DDSRXEQMEMORYPARITYERR | \
+		QLOGIC_IB_HWE_IB_UC_MEMORYPARITYERR | \
+		QLOGIC_IB_HWE_PCIE_UC_OCT0MEMORYPARITYERR | \
+		QLOGIC_IB_HWE_PCIE_UC_OCT1MEMORYPARITYERR)
+
+#define IB_E_BITSEXTANT							\
+	(ERR_MASK(RcvFormatErr) | ERR_MASK(RcvVCRCErr) |		\
+	 ERR_MASK(RcvICRCErr) | ERR_MASK(RcvMinPktLenErr) |		\
+	 ERR_MASK(RcvMaxPktLenErr) | ERR_MASK(RcvLongPktLenErr) |	\
+	 ERR_MASK(RcvShortPktLenErr) | ERR_MASK(RcvUnexpectedCharErr) | \
+	 ERR_MASK(RcvUnsupportedVLErr) | ERR_MASK(RcvEBPErr) |		\
+	 ERR_MASK(RcvIBFlowErr) | ERR_MASK(RcvBadVersionErr) |		\
+	 ERR_MASK(RcvEgrFullErr) | ERR_MASK(RcvHdrFullErr) |		\
+	 ERR_MASK(RcvBadTidErr) | ERR_MASK(RcvHdrLenErr) |		\
+	 ERR_MASK(RcvHdrErr) | ERR_MASK(RcvIBLostLinkErr) |		\
+	 ERR_MASK(SendSpecialTriggerErr) |				\
+	 ERR_MASK(SDmaDisabledErr) | ERR_MASK(SendMinPktLenErr) |	\
+	 ERR_MASK(SendMaxPktLenErr) | ERR_MASK(SendUnderRunErr) |	\
+	 ERR_MASK(SendPktLenErr) | ERR_MASK(SendDroppedSmpPktErr) |	\
+	 ERR_MASK(SendDroppedDataPktErr) |				\
+	 ERR_MASK(SendPioArmLaunchErr) |				\
+	 ERR_MASK(SendUnexpectedPktNumErr) |				\
+	 ERR_MASK(SendUnsupportedVLErr) | ERR_MASK(SendBufMisuseErr) |	\
+	 ERR_MASK(SDmaGenMismatchErr) | ERR_MASK(SDmaOutOfBoundErr) |	\
+	 ERR_MASK(SDmaTailOutOfBoundErr) | ERR_MASK(SDmaBaseErr) |	\
+	 ERR_MASK(SDma1stDescErr) | ERR_MASK(SDmaRpyTagErr) |		\
+	 ERR_MASK(SDmaDwEnErr) | ERR_MASK(SDmaMissingDwErr) |		\
+	 ERR_MASK(SDmaUnexpDataErr) |					\
+	 ERR_MASK(IBStatusChanged) | ERR_MASK(InvalidAddrErr) |		\
+	 ERR_MASK(ResetNegated) | ERR_MASK(HardwareErr) |		\
+	 ERR_MASK(SDmaDescAddrMisalignErr) |				\
+	 ERR_MASK(InvalidEEPCmd))
+
+/* kr_hwerrclear, kr_hwerrmask, kr_hwerrstatus, bits */
+#define QLOGIC_IB_HWE_PCIEMEMPARITYERR_MASK  0x00000000000000ffULL
+#define QLOGIC_IB_HWE_PCIEMEMPARITYERR_SHIFT 0
+#define QLOGIC_IB_HWE_PCIEPOISONEDTLP      0x0000000010000000ULL
+#define QLOGIC_IB_HWE_PCIECPLTIMEOUT       0x0000000020000000ULL
+#define QLOGIC_IB_HWE_PCIEBUSPARITYXTLH    0x0000000040000000ULL
+#define QLOGIC_IB_HWE_PCIEBUSPARITYXADM    0x0000000080000000ULL
+#define QLOGIC_IB_HWE_PCIEBUSPARITYRADM    0x0000000100000000ULL
+#define QLOGIC_IB_HWE_COREPLL_FBSLIP       0x0080000000000000ULL
+#define QLOGIC_IB_HWE_COREPLL_RFSLIP       0x0100000000000000ULL
+#define QLOGIC_IB_HWE_PCIE1PLLFAILED       0x0400000000000000ULL
+#define QLOGIC_IB_HWE_PCIE0PLLFAILED       0x0800000000000000ULL
+#define QLOGIC_IB_HWE_SERDESPLLFAILED      0x1000000000000000ULL
+/* specific to this chip */
+#define QLOGIC_IB_HWE_PCIECPLDATAQUEUEERR         0x0000000000000040ULL
+#define QLOGIC_IB_HWE_PCIECPLHDRQUEUEERR          0x0000000000000080ULL
+#define QLOGIC_IB_HWE_SDMAMEMREADERR              0x0000000010000000ULL
+#define QLOGIC_IB_HWE_CLK_UC_PLLNOTLOCKED          0x2000000000000000ULL
+#define QLOGIC_IB_HWE_PCIESERDESQ0PCLKNOTDETECT   0x0100000000000000ULL
+#define QLOGIC_IB_HWE_PCIESERDESQ1PCLKNOTDETECT   0x0200000000000000ULL
+#define QLOGIC_IB_HWE_PCIESERDESQ2PCLKNOTDETECT   0x0400000000000000ULL
+#define QLOGIC_IB_HWE_PCIESERDESQ3PCLKNOTDETECT   0x0800000000000000ULL
+#define QLOGIC_IB_HWE_DDSRXEQMEMORYPARITYERR       0x0000008000000000ULL
+#define QLOGIC_IB_HWE_IB_UC_MEMORYPARITYERR        0x0000004000000000ULL
+#define QLOGIC_IB_HWE_PCIE_UC_OCT0MEMORYPARITYERR 0x0000001000000000ULL
+#define QLOGIC_IB_HWE_PCIE_UC_OCT1MEMORYPARITYERR 0x0000002000000000ULL
+
+#define IBA7220_IBCC_LINKCMD_SHIFT 19
+
+/* kr_ibcddrctrl bits */
+#define IBA7220_IBC_DLIDLMC_MASK        0xFFFFFFFFUL
+#define IBA7220_IBC_DLIDLMC_SHIFT       32
+
+#define IBA7220_IBC_HRTBT_MASK  (SYM_RMASK(IBCDDRCtrl, HRTBT_AUTO) | \
+				 SYM_RMASK(IBCDDRCtrl, HRTBT_ENB))
+#define IBA7220_IBC_HRTBT_SHIFT SYM_LSB(IBCDDRCtrl, HRTBT_ENB)
+
+#define IBA7220_IBC_LANE_REV_SUPPORTED (1<<8)
+#define IBA7220_IBC_LREV_MASK   1
+#define IBA7220_IBC_LREV_SHIFT  8
+#define IBA7220_IBC_RXPOL_MASK  1
+#define IBA7220_IBC_RXPOL_SHIFT 7
+#define IBA7220_IBC_WIDTH_SHIFT 5
+#define IBA7220_IBC_WIDTH_MASK  0x3
+#define IBA7220_IBC_WIDTH_1X_ONLY       (0 << IBA7220_IBC_WIDTH_SHIFT)
+#define IBA7220_IBC_WIDTH_4X_ONLY       (1 << IBA7220_IBC_WIDTH_SHIFT)
+#define IBA7220_IBC_WIDTH_AUTONEG       (2 << IBA7220_IBC_WIDTH_SHIFT)
+#define IBA7220_IBC_SPEED_AUTONEG       (1 << 1)
+#define IBA7220_IBC_SPEED_SDR           (1 << 2)
+#define IBA7220_IBC_SPEED_DDR           (1 << 3)
+#define IBA7220_IBC_SPEED_AUTONEG_MASK  (0x7 << 1)
+#define IBA7220_IBC_IBTA_1_2_MASK       (1)
+
+/* kr_ibcddrstatus */
+/* link latency shift is 0, don't bother defining */
+#define IBA7220_DDRSTAT_LINKLAT_MASK    0x3ffffff
+
+/* kr_extstatus bits */
+#define QLOGIC_IB_EXTS_FREQSEL 0x2
+#define QLOGIC_IB_EXTS_SERDESSEL 0x4
+#define QLOGIC_IB_EXTS_MEMBIST_ENDTEST     0x0000000000004000
+#define QLOGIC_IB_EXTS_MEMBIST_DISABLED    0x0000000000008000
+
+/* kr_xgxsconfig bits */
+#define QLOGIC_IB_XGXS_RESET          0x5ULL
+#define QLOGIC_IB_XGXS_FC_SAFE        (1ULL << 63)
+
+/* kr_rcvpktledcnt */
+#define IBA7220_LEDBLINK_ON_SHIFT 32 /* 4ns period on after packet */
+#define IBA7220_LEDBLINK_OFF_SHIFT 0 /* 4ns period off before next on */
+
+#define _QIB_GPIO_SDA_NUM 1
+#define _QIB_GPIO_SCL_NUM 0
+#define QIB_TWSI_EEPROM_DEV 0xA2 /* All Production 7220 cards. */
+#define QIB_TWSI_TEMP_DEV 0x98
+
+/* HW counter clock is at 4nsec */
+#define QIB_7220_PSXMITWAIT_CHECK_RATE 4000
+
+#define IBA7220_R_INTRAVAIL_SHIFT 17
+#define IBA7220_R_PKEY_DIS_SHIFT 34
+#define IBA7220_R_TAILUPD_SHIFT 35
+#define IBA7220_R_CTXTCFG_SHIFT 36
+
+#define IBA7220_HDRHEAD_PKTINT_SHIFT 32 /* interrupt cnt in upper 32 bits */
+
+/*
+ * the size bits give us 2^N, in KB units.  0 marks as invalid,
+ * and 7 is reserved.  We currently use only 2KB and 4KB
+ */
+#define IBA7220_TID_SZ_SHIFT 37 /* shift to 3bit size selector */
+#define IBA7220_TID_SZ_2K (1UL << IBA7220_TID_SZ_SHIFT) /* 2KB */
+#define IBA7220_TID_SZ_4K (2UL << IBA7220_TID_SZ_SHIFT) /* 4KB */
+#define IBA7220_TID_PA_SHIFT 11U /* TID addr in chip stored w/o low bits */
+#define PBC_7220_VL15_SEND (1ULL << 63) /* pbc; VL15, no credit check */
+#define PBC_7220_VL15_SEND_CTRL (1ULL << 31) /* control version of same */
+
+#define AUTONEG_TRIES 5 /* sequential retries to negotiate DDR */
+
+/* packet rate matching delay multiplier */
+static u8 rate_to_delay[2][2] = {
+	/* 1x, 4x */
+	{   8, 2 }, /* SDR */
+	{   4, 1 }  /* DDR */
+};
+
+static u8 ib_rate_to_delay[IB_RATE_120_GBPS + 1] = {
+	[IB_RATE_2_5_GBPS] = 8,
+	[IB_RATE_5_GBPS] = 4,
+	[IB_RATE_10_GBPS] = 2,
+	[IB_RATE_20_GBPS] = 1
+};
+
+#define IBA7220_LINKSPEED_SHIFT SYM_LSB(IBCStatus, LinkSpeedActive)
+#define IBA7220_LINKWIDTH_SHIFT SYM_LSB(IBCStatus, LinkWidthActive)
+
+/* link training states, from IBC */
+#define IB_7220_LT_STATE_DISABLED        0x00
+#define IB_7220_LT_STATE_LINKUP          0x01
+#define IB_7220_LT_STATE_POLLACTIVE      0x02
+#define IB_7220_LT_STATE_POLLQUIET       0x03
+#define IB_7220_LT_STATE_SLEEPDELAY      0x04
+#define IB_7220_LT_STATE_SLEEPQUIET      0x05
+#define IB_7220_LT_STATE_CFGDEBOUNCE     0x08
+#define IB_7220_LT_STATE_CFGRCVFCFG      0x09
+#define IB_7220_LT_STATE_CFGWAITRMT      0x0a
+#define IB_7220_LT_STATE_CFGIDLE 0x0b
+#define IB_7220_LT_STATE_RECOVERRETRAIN  0x0c
+#define IB_7220_LT_STATE_RECOVERWAITRMT  0x0e
+#define IB_7220_LT_STATE_RECOVERIDLE     0x0f
+
+/* link state machine states from IBC */
+#define IB_7220_L_STATE_DOWN             0x0
+#define IB_7220_L_STATE_INIT             0x1
+#define IB_7220_L_STATE_ARM              0x2
+#define IB_7220_L_STATE_ACTIVE           0x3
+#define IB_7220_L_STATE_ACT_DEFER        0x4
+
+static const u8 qib_7220_physportstate[0x20] = {
+	[IB_7220_LT_STATE_DISABLED] = IB_PHYSPORTSTATE_DISABLED,
+	[IB_7220_LT_STATE_LINKUP] = IB_PHYSPORTSTATE_LINKUP,
+	[IB_7220_LT_STATE_POLLACTIVE] = IB_PHYSPORTSTATE_POLL,
+	[IB_7220_LT_STATE_POLLQUIET] = IB_PHYSPORTSTATE_POLL,
+	[IB_7220_LT_STATE_SLEEPDELAY] = IB_PHYSPORTSTATE_SLEEP,
+	[IB_7220_LT_STATE_SLEEPQUIET] = IB_PHYSPORTSTATE_SLEEP,
+	[IB_7220_LT_STATE_CFGDEBOUNCE] =
+		IB_PHYSPORTSTATE_CFG_TRAIN,
+	[IB_7220_LT_STATE_CFGRCVFCFG] =
+		IB_PHYSPORTSTATE_CFG_TRAIN,
+	[IB_7220_LT_STATE_CFGWAITRMT] =
+		IB_PHYSPORTSTATE_CFG_TRAIN,
+	[IB_7220_LT_STATE_CFGIDLE] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[IB_7220_LT_STATE_RECOVERRETRAIN] =
+		IB_PHYSPORTSTATE_LINK_ERR_RECOVER,
+	[IB_7220_LT_STATE_RECOVERWAITRMT] =
+		IB_PHYSPORTSTATE_LINK_ERR_RECOVER,
+	[IB_7220_LT_STATE_RECOVERIDLE] =
+		IB_PHYSPORTSTATE_LINK_ERR_RECOVER,
+	[0x10] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x11] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x12] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x13] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x14] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x15] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x16] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x17] = IB_PHYSPORTSTATE_CFG_TRAIN
+};
+
+int qib_special_trigger;
+module_param_named(special_trigger, qib_special_trigger, int, S_IRUGO);
+MODULE_PARM_DESC(special_trigger, "Enable SpecialTrigger arm/launch");
+
+#define IBCBUSFRSPCPARITYERR HWE_MASK(IBCBusFromSPCParityErr)
+#define IBCBUSTOSPCPARITYERR HWE_MASK(IBCBusToSPCParityErr)
+
+#define SYM_MASK_BIT(regname, fldname, bit) ((u64) \
+	(1ULL << (SYM_LSB(regname, fldname) + (bit))))
+
+#define TXEMEMPARITYERR_PIOBUF \
+	SYM_MASK_BIT(HwErrMask, TXEMemParityErrMask, 0)
+#define TXEMEMPARITYERR_PIOPBC \
+	SYM_MASK_BIT(HwErrMask, TXEMemParityErrMask, 1)
+#define TXEMEMPARITYERR_PIOLAUNCHFIFO \
+	SYM_MASK_BIT(HwErrMask, TXEMemParityErrMask, 2)
+
+#define RXEMEMPARITYERR_RCVBUF \
+	SYM_MASK_BIT(HwErrMask, RXEMemParityErrMask, 0)
+#define RXEMEMPARITYERR_LOOKUPQ \
+	SYM_MASK_BIT(HwErrMask, RXEMemParityErrMask, 1)
+#define RXEMEMPARITYERR_EXPTID \
+	SYM_MASK_BIT(HwErrMask, RXEMemParityErrMask, 2)
+#define RXEMEMPARITYERR_EAGERTID \
+	SYM_MASK_BIT(HwErrMask, RXEMemParityErrMask, 3)
+#define RXEMEMPARITYERR_FLAGBUF \
+	SYM_MASK_BIT(HwErrMask, RXEMemParityErrMask, 4)
+#define RXEMEMPARITYERR_DATAINFO \
+	SYM_MASK_BIT(HwErrMask, RXEMemParityErrMask, 5)
+#define RXEMEMPARITYERR_HDRINFO \
+	SYM_MASK_BIT(HwErrMask, RXEMemParityErrMask, 6)
+
+/* 7220 specific hardware errors... */
+static const struct qib_hwerror_msgs qib_7220_hwerror_msgs[] = {
+	/* generic hardware errors */
+	QLOGIC_IB_HWE_MSG(IBCBUSFRSPCPARITYERR, "QIB2IB Parity"),
+	QLOGIC_IB_HWE_MSG(IBCBUSTOSPCPARITYERR, "IB2QIB Parity"),
+
+	QLOGIC_IB_HWE_MSG(TXEMEMPARITYERR_PIOBUF,
+			  "TXE PIOBUF Memory Parity"),
+	QLOGIC_IB_HWE_MSG(TXEMEMPARITYERR_PIOPBC,
+			  "TXE PIOPBC Memory Parity"),
+	QLOGIC_IB_HWE_MSG(TXEMEMPARITYERR_PIOLAUNCHFIFO,
+			  "TXE PIOLAUNCHFIFO Memory Parity"),
+
+	QLOGIC_IB_HWE_MSG(RXEMEMPARITYERR_RCVBUF,
+			  "RXE RCVBUF Memory Parity"),
+	QLOGIC_IB_HWE_MSG(RXEMEMPARITYERR_LOOKUPQ,
+			  "RXE LOOKUPQ Memory Parity"),
+	QLOGIC_IB_HWE_MSG(RXEMEMPARITYERR_EAGERTID,
+			  "RXE EAGERTID Memory Parity"),
+	QLOGIC_IB_HWE_MSG(RXEMEMPARITYERR_EXPTID,
+			  "RXE EXPTID Memory Parity"),
+	QLOGIC_IB_HWE_MSG(RXEMEMPARITYERR_FLAGBUF,
+			  "RXE FLAGBUF Memory Parity"),
+	QLOGIC_IB_HWE_MSG(RXEMEMPARITYERR_DATAINFO,
+			  "RXE DATAINFO Memory Parity"),
+	QLOGIC_IB_HWE_MSG(RXEMEMPARITYERR_HDRINFO,
+			  "RXE HDRINFO Memory Parity"),
+
+	/* chip-specific hardware errors */
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIEPOISONEDTLP,
+			  "PCIe Poisoned TLP"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIECPLTIMEOUT,
+			  "PCIe completion timeout"),
+	/*
+	 * In practice, it's unlikely wthat we'll see PCIe PLL, or bus
+	 * parity or memory parity error failures, because most likely we
+	 * won't be able to talk to the core of the chip.  Nonetheless, we
+	 * might see them, if they are in parts of the PCIe core that aren't
+	 * essential.
+	 */
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIE1PLLFAILED,
+			  "PCIePLL1"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIE0PLLFAILED,
+			  "PCIePLL0"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIEBUSPARITYXTLH,
+			  "PCIe XTLH core parity"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIEBUSPARITYXADM,
+			  "PCIe ADM TX core parity"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIEBUSPARITYRADM,
+			  "PCIe ADM RX core parity"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_SERDESPLLFAILED,
+			  "SerDes PLL"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIECPLDATAQUEUEERR,
+			  "PCIe cpl header queue"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIECPLHDRQUEUEERR,
+			  "PCIe cpl data queue"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_SDMAMEMREADERR,
+			  "Send DMA memory read"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_CLK_UC_PLLNOTLOCKED,
+			  "uC PLL clock not locked"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIESERDESQ0PCLKNOTDETECT,
+			  "PCIe serdes Q0 no clock"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIESERDESQ1PCLKNOTDETECT,
+			  "PCIe serdes Q1 no clock"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIESERDESQ2PCLKNOTDETECT,
+			  "PCIe serdes Q2 no clock"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIESERDESQ3PCLKNOTDETECT,
+			  "PCIe serdes Q3 no clock"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_DDSRXEQMEMORYPARITYERR,
+			  "DDS RXEQ memory parity"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_IB_UC_MEMORYPARITYERR,
+			  "IB uC memory parity"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIE_UC_OCT0MEMORYPARITYERR,
+			  "PCIe uC oct0 memory parity"),
+	QLOGIC_IB_HWE_MSG(QLOGIC_IB_HWE_PCIE_UC_OCT1MEMORYPARITYERR,
+			  "PCIe uC oct1 memory parity"),
+};
+
+#define RXE_PARITY (RXEMEMPARITYERR_EAGERTID|RXEMEMPARITYERR_EXPTID)
+
+#define QLOGIC_IB_E_PKTERRS (\
+		ERR_MASK(SendPktLenErr) |				\
+		ERR_MASK(SendDroppedDataPktErr) |			\
+		ERR_MASK(RcvVCRCErr) |					\
+		ERR_MASK(RcvICRCErr) |					\
+		ERR_MASK(RcvShortPktLenErr) |				\
+		ERR_MASK(RcvEBPErr))
+
+/* Convenience for decoding Send DMA errors */
+#define QLOGIC_IB_E_SDMAERRS ( \
+		ERR_MASK(SDmaGenMismatchErr) |				\
+		ERR_MASK(SDmaOutOfBoundErr) |				\
+		ERR_MASK(SDmaTailOutOfBoundErr) | ERR_MASK(SDmaBaseErr) | \
+		ERR_MASK(SDma1stDescErr) | ERR_MASK(SDmaRpyTagErr) |	\
+		ERR_MASK(SDmaDwEnErr) | ERR_MASK(SDmaMissingDwErr) |	\
+		ERR_MASK(SDmaUnexpDataErr) |				\
+		ERR_MASK(SDmaDescAddrMisalignErr) |			\
+		ERR_MASK(SDmaDisabledErr) |				\
+		ERR_MASK(SendBufMisuseErr))
+
+/* These are all rcv-related errors which we want to count for stats */
+#define E_SUM_PKTERRS \
+	(ERR_MASK(RcvHdrLenErr) | ERR_MASK(RcvBadTidErr) |		\
+	 ERR_MASK(RcvBadVersionErr) | ERR_MASK(RcvHdrErr) |		\
+	 ERR_MASK(RcvLongPktLenErr) | ERR_MASK(RcvShortPktLenErr) |	\
+	 ERR_MASK(RcvMaxPktLenErr) | ERR_MASK(RcvMinPktLenErr) |	\
+	 ERR_MASK(RcvFormatErr) | ERR_MASK(RcvUnsupportedVLErr) |	\
+	 ERR_MASK(RcvUnexpectedCharErr) | ERR_MASK(RcvEBPErr))
+
+/* These are all send-related errors which we want to count for stats */
+#define E_SUM_ERRS \
+	(ERR_MASK(SendPioArmLaunchErr) | ERR_MASK(SendUnexpectedPktNumErr) | \
+	 ERR_MASK(SendDroppedDataPktErr) | ERR_MASK(SendDroppedSmpPktErr) | \
+	 ERR_MASK(SendMaxPktLenErr) | ERR_MASK(SendUnsupportedVLErr) |	\
+	 ERR_MASK(SendMinPktLenErr) | ERR_MASK(SendPktLenErr) |		\
+	 ERR_MASK(InvalidAddrErr))
+
+/*
+ * this is similar to E_SUM_ERRS, but can't ignore armlaunch, don't ignore
+ * errors not related to freeze and cancelling buffers.  Can't ignore
+ * armlaunch because could get more while still cleaning up, and need
+ * to cancel those as they happen.
+ */
+#define E_SPKT_ERRS_IGNORE \
+	(ERR_MASK(SendDroppedDataPktErr) | ERR_MASK(SendDroppedSmpPktErr) | \
+	 ERR_MASK(SendMaxPktLenErr) | ERR_MASK(SendMinPktLenErr) |	\
+	 ERR_MASK(SendPktLenErr))
+
+/*
+ * these are errors that can occur when the link changes state while
+ * a packet is being sent or received.  This doesn't cover things
+ * like EBP or VCRC that can be the result of a sending having the
+ * link change state, so we receive a "known bad" packet.
+ */
+#define E_SUM_LINK_PKTERRS \
+	(ERR_MASK(SendDroppedDataPktErr) | ERR_MASK(SendDroppedSmpPktErr) | \
+	 ERR_MASK(SendMinPktLenErr) | ERR_MASK(SendPktLenErr) |		\
+	 ERR_MASK(RcvShortPktLenErr) | ERR_MASK(RcvMinPktLenErr) |	\
+	 ERR_MASK(RcvUnexpectedCharErr))
+
+static void autoneg_7220_work(struct work_struct *);
+static u32 __iomem *qib_7220_getsendbuf(struct qib_pportdata *, u64, u32 *);
+
+/*
+ * Called when we might have an error that is specific to a particular
+ * PIO buffer, and may need to cancel that buffer, so it can be re-used.
+ * because we don't need to force the update of pioavail.
+ */
+static void qib_disarm_7220_senderrbufs(struct qib_pportdata *ppd)
+{
+	unsigned long sbuf[3];
+	struct qib_devdata *dd = ppd->dd;
+
+	/*
+	 * It's possible that sendbuffererror could have bits set; might
+	 * have already done this as a result of hardware error handling.
+	 */
+	/* read these before writing errorclear */
+	sbuf[0] = qib_read_kreg64(dd, kr_sendbuffererror);
+	sbuf[1] = qib_read_kreg64(dd, kr_sendbuffererror + 1);
+	sbuf[2] = qib_read_kreg64(dd, kr_sendbuffererror + 2);
+
+	if (sbuf[0] || sbuf[1] || sbuf[2])
+		qib_disarm_piobufs_set(dd, sbuf,
+				       dd->piobcnt2k + dd->piobcnt4k);
+}
+
+static void qib_7220_txe_recover(struct qib_devdata *dd)
+{
+	qib_devinfo(dd->pcidev, "Recovering from TXE PIO parity error\n");
+	qib_disarm_7220_senderrbufs(dd->pport);
+}
+
+/*
+ * This is called with interrupts disabled and sdma_lock held.
+ */
+static void qib_7220_sdma_sendctrl(struct qib_pportdata *ppd, unsigned op)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u64 set_sendctrl = 0;
+	u64 clr_sendctrl = 0;
+
+	if (op & QIB_SDMA_SENDCTRL_OP_ENABLE)
+		set_sendctrl |= SYM_MASK(SendCtrl, SDmaEnable);
+	else
+		clr_sendctrl |= SYM_MASK(SendCtrl, SDmaEnable);
+
+	if (op & QIB_SDMA_SENDCTRL_OP_INTENABLE)
+		set_sendctrl |= SYM_MASK(SendCtrl, SDmaIntEnable);
+	else
+		clr_sendctrl |= SYM_MASK(SendCtrl, SDmaIntEnable);
+
+	if (op & QIB_SDMA_SENDCTRL_OP_HALT)
+		set_sendctrl |= SYM_MASK(SendCtrl, SDmaHalt);
+	else
+		clr_sendctrl |= SYM_MASK(SendCtrl, SDmaHalt);
+
+	spin_lock(&dd->sendctrl_lock);
+
+	dd->sendctrl |= set_sendctrl;
+	dd->sendctrl &= ~clr_sendctrl;
+
+	qib_write_kreg(dd, kr_sendctrl, dd->sendctrl);
+	qib_write_kreg(dd, kr_scratch, 0);
+
+	spin_unlock(&dd->sendctrl_lock);
+}
+
+static void qib_decode_7220_sdma_errs(struct qib_pportdata *ppd,
+				      u64 err, char *buf, size_t blen)
+{
+	static const struct {
+		u64 err;
+		const char *msg;
+	} errs[] = {
+		{ ERR_MASK(SDmaGenMismatchErr),
+		  "SDmaGenMismatch" },
+		{ ERR_MASK(SDmaOutOfBoundErr),
+		  "SDmaOutOfBound" },
+		{ ERR_MASK(SDmaTailOutOfBoundErr),
+		  "SDmaTailOutOfBound" },
+		{ ERR_MASK(SDmaBaseErr),
+		  "SDmaBase" },
+		{ ERR_MASK(SDma1stDescErr),
+		  "SDma1stDesc" },
+		{ ERR_MASK(SDmaRpyTagErr),
+		  "SDmaRpyTag" },
+		{ ERR_MASK(SDmaDwEnErr),
+		  "SDmaDwEn" },
+		{ ERR_MASK(SDmaMissingDwErr),
+		  "SDmaMissingDw" },
+		{ ERR_MASK(SDmaUnexpDataErr),
+		  "SDmaUnexpData" },
+		{ ERR_MASK(SDmaDescAddrMisalignErr),
+		  "SDmaDescAddrMisalign" },
+		{ ERR_MASK(SendBufMisuseErr),
+		  "SendBufMisuse" },
+		{ ERR_MASK(SDmaDisabledErr),
+		  "SDmaDisabled" },
+	};
+	int i;
+	size_t bidx = 0;
+
+	for (i = 0; i < ARRAY_SIZE(errs); i++) {
+		if (err & errs[i].err)
+			bidx += scnprintf(buf + bidx, blen - bidx,
+					 "%s ", errs[i].msg);
+	}
+}
+
+/*
+ * This is called as part of link down clean up so disarm and flush
+ * all send buffers so that SMP packets can be sent.
+ */
+static void qib_7220_sdma_hw_clean_up(struct qib_pportdata *ppd)
+{
+	/* This will trigger the Abort interrupt */
+	sendctrl_7220_mod(ppd, QIB_SENDCTRL_DISARM_ALL | QIB_SENDCTRL_FLUSH |
+			  QIB_SENDCTRL_AVAIL_BLIP);
+	ppd->dd->upd_pio_shadow  = 1; /* update our idea of what's busy */
+}
+
+static void qib_sdma_7220_setlengen(struct qib_pportdata *ppd)
+{
+	/*
+	 * Set SendDmaLenGen and clear and set
+	 * the MSB of the generation count to enable generation checking
+	 * and load the internal generation counter.
+	 */
+	qib_write_kreg(ppd->dd, kr_senddmalengen, ppd->sdma_descq_cnt);
+	qib_write_kreg(ppd->dd, kr_senddmalengen,
+		       ppd->sdma_descq_cnt |
+		       (1ULL << QIB_7220_SendDmaLenGen_Generation_MSB));
+}
+
+static void qib_7220_sdma_hw_start_up(struct qib_pportdata *ppd)
+{
+	qib_sdma_7220_setlengen(ppd);
+	qib_sdma_update_7220_tail(ppd, 0); /* Set SendDmaTail */
+	ppd->sdma_head_dma[0] = 0;
+}
+
+#define DISABLES_SDMA (							\
+		ERR_MASK(SDmaDisabledErr) |				\
+		ERR_MASK(SDmaBaseErr) |					\
+		ERR_MASK(SDmaTailOutOfBoundErr) |			\
+		ERR_MASK(SDmaOutOfBoundErr) |				\
+		ERR_MASK(SDma1stDescErr) |				\
+		ERR_MASK(SDmaRpyTagErr) |				\
+		ERR_MASK(SDmaGenMismatchErr) |				\
+		ERR_MASK(SDmaDescAddrMisalignErr) |			\
+		ERR_MASK(SDmaMissingDwErr) |				\
+		ERR_MASK(SDmaDwEnErr))
+
+static void sdma_7220_errors(struct qib_pportdata *ppd, u64 errs)
+{
+	unsigned long flags;
+	struct qib_devdata *dd = ppd->dd;
+	char *msg;
+
+	errs &= QLOGIC_IB_E_SDMAERRS;
+
+	msg = dd->cspec->sdmamsgbuf;
+	qib_decode_7220_sdma_errs(ppd, errs, msg, sizeof dd->cspec->sdmamsgbuf);
+	spin_lock_irqsave(&ppd->sdma_lock, flags);
+
+	if (errs & ERR_MASK(SendBufMisuseErr)) {
+		unsigned long sbuf[3];
+
+		sbuf[0] = qib_read_kreg64(dd, kr_sendbuffererror);
+		sbuf[1] = qib_read_kreg64(dd, kr_sendbuffererror + 1);
+		sbuf[2] = qib_read_kreg64(dd, kr_sendbuffererror + 2);
+
+		qib_dev_err(ppd->dd,
+			    "IB%u:%u SendBufMisuse: %04lx %016lx %016lx\n",
+			    ppd->dd->unit, ppd->port, sbuf[2], sbuf[1],
+			    sbuf[0]);
+	}
+
+	if (errs & ERR_MASK(SDmaUnexpDataErr))
+		qib_dev_err(dd, "IB%u:%u SDmaUnexpData\n", ppd->dd->unit,
+			    ppd->port);
+
+	switch (ppd->sdma_state.current_state) {
+	case qib_sdma_state_s00_hw_down:
+		/* not expecting any interrupts */
+		break;
+
+	case qib_sdma_state_s10_hw_start_up_wait:
+		/* handled in intr path */
+		break;
+
+	case qib_sdma_state_s20_idle:
+		/* not expecting any interrupts */
+		break;
+
+	case qib_sdma_state_s30_sw_clean_up_wait:
+		/* not expecting any interrupts */
+		break;
+
+	case qib_sdma_state_s40_hw_clean_up_wait:
+		if (errs & ERR_MASK(SDmaDisabledErr))
+			__qib_sdma_process_event(ppd,
+				qib_sdma_event_e50_hw_cleaned);
+		break;
+
+	case qib_sdma_state_s50_hw_halt_wait:
+		/* handled in intr path */
+		break;
+
+	case qib_sdma_state_s99_running:
+		if (errs & DISABLES_SDMA)
+			__qib_sdma_process_event(ppd,
+				qib_sdma_event_e7220_err_halted);
+		break;
+	}
+
+	spin_unlock_irqrestore(&ppd->sdma_lock, flags);
+}
+
+/*
+ * Decode the error status into strings, deciding whether to always
+ * print * it or not depending on "normal packet errors" vs everything
+ * else.   Return 1 if "real" errors, otherwise 0 if only packet
+ * errors, so caller can decide what to print with the string.
+ */
+static int qib_decode_7220_err(struct qib_devdata *dd, char *buf, size_t blen,
+			       u64 err)
+{
+	int iserr = 1;
+
+	*buf = '\0';
+	if (err & QLOGIC_IB_E_PKTERRS) {
+		if (!(err & ~QLOGIC_IB_E_PKTERRS))
+			iserr = 0;
+		if ((err & ERR_MASK(RcvICRCErr)) &&
+		    !(err & (ERR_MASK(RcvVCRCErr) | ERR_MASK(RcvEBPErr))))
+			strlcat(buf, "CRC ", blen);
+		if (!iserr)
+			goto done;
+	}
+	if (err & ERR_MASK(RcvHdrLenErr))
+		strlcat(buf, "rhdrlen ", blen);
+	if (err & ERR_MASK(RcvBadTidErr))
+		strlcat(buf, "rbadtid ", blen);
+	if (err & ERR_MASK(RcvBadVersionErr))
+		strlcat(buf, "rbadversion ", blen);
+	if (err & ERR_MASK(RcvHdrErr))
+		strlcat(buf, "rhdr ", blen);
+	if (err & ERR_MASK(SendSpecialTriggerErr))
+		strlcat(buf, "sendspecialtrigger ", blen);
+	if (err & ERR_MASK(RcvLongPktLenErr))
+		strlcat(buf, "rlongpktlen ", blen);
+	if (err & ERR_MASK(RcvMaxPktLenErr))
+		strlcat(buf, "rmaxpktlen ", blen);
+	if (err & ERR_MASK(RcvMinPktLenErr))
+		strlcat(buf, "rminpktlen ", blen);
+	if (err & ERR_MASK(SendMinPktLenErr))
+		strlcat(buf, "sminpktlen ", blen);
+	if (err & ERR_MASK(RcvFormatErr))
+		strlcat(buf, "rformaterr ", blen);
+	if (err & ERR_MASK(RcvUnsupportedVLErr))
+		strlcat(buf, "runsupvl ", blen);
+	if (err & ERR_MASK(RcvUnexpectedCharErr))
+		strlcat(buf, "runexpchar ", blen);
+	if (err & ERR_MASK(RcvIBFlowErr))
+		strlcat(buf, "ribflow ", blen);
+	if (err & ERR_MASK(SendUnderRunErr))
+		strlcat(buf, "sunderrun ", blen);
+	if (err & ERR_MASK(SendPioArmLaunchErr))
+		strlcat(buf, "spioarmlaunch ", blen);
+	if (err & ERR_MASK(SendUnexpectedPktNumErr))
+		strlcat(buf, "sunexperrpktnum ", blen);
+	if (err & ERR_MASK(SendDroppedSmpPktErr))
+		strlcat(buf, "sdroppedsmppkt ", blen);
+	if (err & ERR_MASK(SendMaxPktLenErr))
+		strlcat(buf, "smaxpktlen ", blen);
+	if (err & ERR_MASK(SendUnsupportedVLErr))
+		strlcat(buf, "sunsupVL ", blen);
+	if (err & ERR_MASK(InvalidAddrErr))
+		strlcat(buf, "invalidaddr ", blen);
+	if (err & ERR_MASK(RcvEgrFullErr))
+		strlcat(buf, "rcvegrfull ", blen);
+	if (err & ERR_MASK(RcvHdrFullErr))
+		strlcat(buf, "rcvhdrfull ", blen);
+	if (err & ERR_MASK(IBStatusChanged))
+		strlcat(buf, "ibcstatuschg ", blen);
+	if (err & ERR_MASK(RcvIBLostLinkErr))
+		strlcat(buf, "riblostlink ", blen);
+	if (err & ERR_MASK(HardwareErr))
+		strlcat(buf, "hardware ", blen);
+	if (err & ERR_MASK(ResetNegated))
+		strlcat(buf, "reset ", blen);
+	if (err & QLOGIC_IB_E_SDMAERRS)
+		qib_decode_7220_sdma_errs(dd->pport, err, buf, blen);
+	if (err & ERR_MASK(InvalidEEPCmd))
+		strlcat(buf, "invalideepromcmd ", blen);
+done:
+	return iserr;
+}
+
+static void reenable_7220_chase(unsigned long opaque)
+{
+	struct qib_pportdata *ppd = (struct qib_pportdata *)opaque;
+	ppd->cpspec->chase_timer.expires = 0;
+	qib_set_ib_7220_lstate(ppd, QLOGIC_IB_IBCC_LINKCMD_DOWN,
+		QLOGIC_IB_IBCC_LINKINITCMD_POLL);
+}
+
+static void handle_7220_chase(struct qib_pportdata *ppd, u64 ibcst)
+{
+	u8 ibclt;
+	u64 tnow;
+
+	ibclt = (u8)SYM_FIELD(ibcst, IBCStatus, LinkTrainingState);
+
+	/*
+	 * Detect and handle the state chase issue, where we can
+	 * get stuck if we are unlucky on timing on both sides of
+	 * the link.   If we are, we disable, set a timer, and
+	 * then re-enable.
+	 */
+	switch (ibclt) {
+	case IB_7220_LT_STATE_CFGRCVFCFG:
+	case IB_7220_LT_STATE_CFGWAITRMT:
+	case IB_7220_LT_STATE_TXREVLANES:
+	case IB_7220_LT_STATE_CFGENH:
+		tnow = get_jiffies_64();
+		if (ppd->cpspec->chase_end &&
+		    time_after64(tnow, ppd->cpspec->chase_end)) {
+			ppd->cpspec->chase_end = 0;
+			qib_set_ib_7220_lstate(ppd,
+				QLOGIC_IB_IBCC_LINKCMD_DOWN,
+				QLOGIC_IB_IBCC_LINKINITCMD_DISABLE);
+			ppd->cpspec->chase_timer.expires = jiffies +
+				QIB_CHASE_DIS_TIME;
+			add_timer(&ppd->cpspec->chase_timer);
+		} else if (!ppd->cpspec->chase_end)
+			ppd->cpspec->chase_end = tnow + QIB_CHASE_TIME;
+		break;
+
+	default:
+		ppd->cpspec->chase_end = 0;
+		break;
+	}
+}
+
+static void handle_7220_errors(struct qib_devdata *dd, u64 errs)
+{
+	char *msg;
+	u64 ignore_this_time = 0;
+	u64 iserr = 0;
+	int log_idx;
+	struct qib_pportdata *ppd = dd->pport;
+	u64 mask;
+
+	/* don't report errors that are masked */
+	errs &= dd->cspec->errormask;
+	msg = dd->cspec->emsgbuf;
+
+	/* do these first, they are most important */
+	if (errs & ERR_MASK(HardwareErr))
+		qib_7220_handle_hwerrors(dd, msg, sizeof dd->cspec->emsgbuf);
+	else
+		for (log_idx = 0; log_idx < QIB_EEP_LOG_CNT; ++log_idx)
+			if (errs & dd->eep_st_masks[log_idx].errs_to_log)
+				qib_inc_eeprom_err(dd, log_idx, 1);
+
+	if (errs & QLOGIC_IB_E_SDMAERRS)
+		sdma_7220_errors(ppd, errs);
+
+	if (errs & ~IB_E_BITSEXTANT)
+		qib_dev_err(dd, "error interrupt with unknown errors "
+			    "%llx set\n", (unsigned long long)
+			    (errs & ~IB_E_BITSEXTANT));
+
+	if (errs & E_SUM_ERRS) {
+		qib_disarm_7220_senderrbufs(ppd);
+		if ((errs & E_SUM_LINK_PKTERRS) &&
+		    !(ppd->lflags & QIBL_LINKACTIVE)) {
+			/*
+			 * This can happen when trying to bring the link
+			 * up, but the IB link changes state at the "wrong"
+			 * time. The IB logic then complains that the packet
+			 * isn't valid.  We don't want to confuse people, so
+			 * we just don't print them, except at debug
+			 */
+			ignore_this_time = errs & E_SUM_LINK_PKTERRS;
+		}
+	} else if ((errs & E_SUM_LINK_PKTERRS) &&
+		   !(ppd->lflags & QIBL_LINKACTIVE)) {
+		/*
+		 * This can happen when SMA is trying to bring the link
+		 * up, but the IB link changes state at the "wrong" time.
+		 * The IB logic then complains that the packet isn't
+		 * valid.  We don't want to confuse people, so we just
+		 * don't print them, except at debug
+		 */
+		ignore_this_time = errs & E_SUM_LINK_PKTERRS;
+	}
+
+	qib_write_kreg(dd, kr_errclear, errs);
+
+	errs &= ~ignore_this_time;
+	if (!errs)
+		goto done;
+
+	/*
+	 * The ones we mask off are handled specially below
+	 * or above.  Also mask SDMADISABLED by default as it
+	 * is too chatty.
+	 */
+	mask = ERR_MASK(IBStatusChanged) |
+		ERR_MASK(RcvEgrFullErr) | ERR_MASK(RcvHdrFullErr) |
+		ERR_MASK(HardwareErr) | ERR_MASK(SDmaDisabledErr);
+
+	qib_decode_7220_err(dd, msg, sizeof dd->cspec->emsgbuf, errs & ~mask);
+
+	if (errs & E_SUM_PKTERRS)
+		qib_stats.sps_rcverrs++;
+	if (errs & E_SUM_ERRS)
+		qib_stats.sps_txerrs++;
+	iserr = errs & ~(E_SUM_PKTERRS | QLOGIC_IB_E_PKTERRS |
+			 ERR_MASK(SDmaDisabledErr));
+
+	if (errs & ERR_MASK(IBStatusChanged)) {
+		u64 ibcs;
+
+		ibcs = qib_read_kreg64(dd, kr_ibcstatus);
+		if (!(ppd->lflags & QIBL_IB_AUTONEG_INPROG))
+			handle_7220_chase(ppd, ibcs);
+
+		/* Update our picture of width and speed from chip */
+		ppd->link_width_active =
+			((ibcs >> IBA7220_LINKWIDTH_SHIFT) & 1) ?
+			    IB_WIDTH_4X : IB_WIDTH_1X;
+		ppd->link_speed_active =
+			((ibcs >> IBA7220_LINKSPEED_SHIFT) & 1) ?
+			    QIB_IB_DDR : QIB_IB_SDR;
+
+		/*
+		 * Since going into a recovery state causes the link state
+		 * to go down and since recovery is transitory, it is better
+		 * if we "miss" ever seeing the link training state go into
+		 * recovery (i.e., ignore this transition for link state
+		 * special handling purposes) without updating lastibcstat.
+		 */
+		if (qib_7220_phys_portstate(ibcs) !=
+					    IB_PHYSPORTSTATE_LINK_ERR_RECOVER)
+			qib_handle_e_ibstatuschanged(ppd, ibcs);
+	}
+
+	if (errs & ERR_MASK(ResetNegated)) {
+		qib_dev_err(dd, "Got reset, requires re-init "
+			    "(unload and reload driver)\n");
+		dd->flags &= ~QIB_INITTED;  /* needs re-init */
+		/* mark as having had error */
+		*dd->devstatusp |= QIB_STATUS_HWERROR;
+		*dd->pport->statusp &= ~QIB_STATUS_IB_CONF;
+	}
+
+	if (*msg && iserr)
+		qib_dev_porterr(dd, ppd->port, "%s error\n", msg);
+
+	if (ppd->state_wanted & ppd->lflags)
+		wake_up_interruptible(&ppd->state_wait);
+
+	/*
+	 * If there were hdrq or egrfull errors, wake up any processes
+	 * waiting in poll.  We used to try to check which contexts had
+	 * the overflow, but given the cost of that and the chip reads
+	 * to support it, it's better to just wake everybody up if we
+	 * get an overflow; waiters can poll again if it's not them.
+	 */
+	if (errs & (ERR_MASK(RcvEgrFullErr) | ERR_MASK(RcvHdrFullErr))) {
+		qib_handle_urcv(dd, ~0U);
+		if (errs & ERR_MASK(RcvEgrFullErr))
+			qib_stats.sps_buffull++;
+		else
+			qib_stats.sps_hdrfull++;
+	}
+done:
+	return;
+}
+
+/* enable/disable chip from delivering interrupts */
+static void qib_7220_set_intr_state(struct qib_devdata *dd, u32 enable)
+{
+	if (enable) {
+		if (dd->flags & QIB_BADINTR)
+			return;
+		qib_write_kreg(dd, kr_intmask, ~0ULL);
+		/* force re-interrupt of any pending interrupts. */
+		qib_write_kreg(dd, kr_intclear, 0ULL);
+	} else
+		qib_write_kreg(dd, kr_intmask, 0ULL);
+}
+
+/*
+ * Try to cleanup as much as possible for anything that might have gone
+ * wrong while in freeze mode, such as pio buffers being written by user
+ * processes (causing armlaunch), send errors due to going into freeze mode,
+ * etc., and try to avoid causing extra interrupts while doing so.
+ * Forcibly update the in-memory pioavail register copies after cleanup
+ * because the chip won't do it while in freeze mode (the register values
+ * themselves are kept correct).
+ * Make sure that we don't lose any important interrupts by using the chip
+ * feature that says that writing 0 to a bit in *clear that is set in
+ * *status will cause an interrupt to be generated again (if allowed by
+ * the *mask value).
+ * This is in chip-specific code because of all of the register accesses,
+ * even though the details are similar on most chips.
+ */
+static void qib_7220_clear_freeze(struct qib_devdata *dd)
+{
+	/* disable error interrupts, to avoid confusion */
+	qib_write_kreg(dd, kr_errmask, 0ULL);
+
+	/* also disable interrupts; errormask is sometimes overwriten */
+	qib_7220_set_intr_state(dd, 0);
+
+	qib_cancel_sends(dd->pport);
+
+	/* clear the freeze, and be sure chip saw it */
+	qib_write_kreg(dd, kr_control, dd->control);
+	qib_read_kreg32(dd, kr_scratch);
+
+	/* force in-memory update now we are out of freeze */
+	qib_force_pio_avail_update(dd);
+
+	/*
+	 * force new interrupt if any hwerr, error or interrupt bits are
+	 * still set, and clear "safe" send packet errors related to freeze
+	 * and cancelling sends.  Re-enable error interrupts before possible
+	 * force of re-interrupt on pending interrupts.
+	 */
+	qib_write_kreg(dd, kr_hwerrclear, 0ULL);
+	qib_write_kreg(dd, kr_errclear, E_SPKT_ERRS_IGNORE);
+	qib_write_kreg(dd, kr_errmask, dd->cspec->errormask);
+	qib_7220_set_intr_state(dd, 1);
+}
+
+/**
+ * qib_7220_handle_hwerrors - display hardware errors.
+ * @dd: the qlogic_ib device
+ * @msg: the output buffer
+ * @msgl: the size of the output buffer
+ *
+ * Use same msg buffer as regular errors to avoid excessive stack
+ * use.  Most hardware errors are catastrophic, but for right now,
+ * we'll print them and continue.  We reuse the same message buffer as
+ * handle_7220_errors() to avoid excessive stack usage.
+ */
+static void qib_7220_handle_hwerrors(struct qib_devdata *dd, char *msg,
+				     size_t msgl)
+{
+	u64 hwerrs;
+	u32 bits, ctrl;
+	int isfatal = 0;
+	char *bitsmsg;
+	int log_idx;
+
+	hwerrs = qib_read_kreg64(dd, kr_hwerrstatus);
+	if (!hwerrs)
+		goto bail;
+	if (hwerrs == ~0ULL) {
+		qib_dev_err(dd, "Read of hardware error status failed "
+			    "(all bits set); ignoring\n");
+		goto bail;
+	}
+	qib_stats.sps_hwerrs++;
+
+	/*
+	 * Always clear the error status register, except MEMBISTFAIL,
+	 * regardless of whether we continue or stop using the chip.
+	 * We want that set so we know it failed, even across driver reload.
+	 * We'll still ignore it in the hwerrmask.  We do this partly for
+	 * diagnostics, but also for support.
+	 */
+	qib_write_kreg(dd, kr_hwerrclear,
+		       hwerrs & ~HWE_MASK(PowerOnBISTFailed));
+
+	hwerrs &= dd->cspec->hwerrmask;
+
+	/* We log some errors to EEPROM, check if we have any of those. */
+	for (log_idx = 0; log_idx < QIB_EEP_LOG_CNT; ++log_idx)
+		if (hwerrs & dd->eep_st_masks[log_idx].hwerrs_to_log)
+			qib_inc_eeprom_err(dd, log_idx, 1);
+	if (hwerrs & ~(TXEMEMPARITYERR_PIOBUF | TXEMEMPARITYERR_PIOPBC |
+		       RXE_PARITY))
+		qib_devinfo(dd->pcidev, "Hardware error: hwerr=0x%llx "
+			 "(cleared)\n", (unsigned long long) hwerrs);
+
+	if (hwerrs & ~IB_HWE_BITSEXTANT)
+		qib_dev_err(dd, "hwerror interrupt with unknown errors "
+			    "%llx set\n", (unsigned long long)
+			    (hwerrs & ~IB_HWE_BITSEXTANT));
+
+	if (hwerrs & QLOGIC_IB_HWE_IB_UC_MEMORYPARITYERR)
+		qib_sd7220_clr_ibpar(dd);
+
+	ctrl = qib_read_kreg32(dd, kr_control);
+	if ((ctrl & QLOGIC_IB_C_FREEZEMODE) && !dd->diag_client) {
+		/*
+		 * Parity errors in send memory are recoverable by h/w
+		 * just do housekeeping, exit freeze mode and continue.
+		 */
+		if (hwerrs & (TXEMEMPARITYERR_PIOBUF |
+			      TXEMEMPARITYERR_PIOPBC)) {
+			qib_7220_txe_recover(dd);
+			hwerrs &= ~(TXEMEMPARITYERR_PIOBUF |
+				    TXEMEMPARITYERR_PIOPBC);
+		}
+		if (hwerrs)
+			isfatal = 1;
+		else
+			qib_7220_clear_freeze(dd);
+	}
+
+	*msg = '\0';
+
+	if (hwerrs & HWE_MASK(PowerOnBISTFailed)) {
+		isfatal = 1;
+		strlcat(msg, "[Memory BIST test failed, "
+			"InfiniPath hardware unusable]", msgl);
+		/* ignore from now on, so disable until driver reloaded */
+		dd->cspec->hwerrmask &= ~HWE_MASK(PowerOnBISTFailed);
+		qib_write_kreg(dd, kr_hwerrmask, dd->cspec->hwerrmask);
+	}
+
+	qib_format_hwerrors(hwerrs, qib_7220_hwerror_msgs,
+			    ARRAY_SIZE(qib_7220_hwerror_msgs), msg, msgl);
+
+	bitsmsg = dd->cspec->bitsmsgbuf;
+	if (hwerrs & (QLOGIC_IB_HWE_PCIEMEMPARITYERR_MASK <<
+		      QLOGIC_IB_HWE_PCIEMEMPARITYERR_SHIFT)) {
+		bits = (u32) ((hwerrs >>
+			       QLOGIC_IB_HWE_PCIEMEMPARITYERR_SHIFT) &
+			      QLOGIC_IB_HWE_PCIEMEMPARITYERR_MASK);
+		snprintf(bitsmsg, sizeof dd->cspec->bitsmsgbuf,
+			 "[PCIe Mem Parity Errs %x] ", bits);
+		strlcat(msg, bitsmsg, msgl);
+	}
+
+#define _QIB_PLL_FAIL (QLOGIC_IB_HWE_COREPLL_FBSLIP |   \
+			 QLOGIC_IB_HWE_COREPLL_RFSLIP)
+
+	if (hwerrs & _QIB_PLL_FAIL) {
+		isfatal = 1;
+		snprintf(bitsmsg, sizeof dd->cspec->bitsmsgbuf,
+			 "[PLL failed (%llx), InfiniPath hardware unusable]",
+			 (unsigned long long) hwerrs & _QIB_PLL_FAIL);
+		strlcat(msg, bitsmsg, msgl);
+		/* ignore from now on, so disable until driver reloaded */
+		dd->cspec->hwerrmask &= ~(hwerrs & _QIB_PLL_FAIL);
+		qib_write_kreg(dd, kr_hwerrmask, dd->cspec->hwerrmask);
+	}
+
+	if (hwerrs & QLOGIC_IB_HWE_SERDESPLLFAILED) {
+		/*
+		 * If it occurs, it is left masked since the eternal
+		 * interface is unused.
+		 */
+		dd->cspec->hwerrmask &= ~QLOGIC_IB_HWE_SERDESPLLFAILED;
+		qib_write_kreg(dd, kr_hwerrmask, dd->cspec->hwerrmask);
+	}
+
+	qib_dev_err(dd, "%s hardware error\n", msg);
+
+	if (isfatal && !dd->diag_client) {
+		qib_dev_err(dd, "Fatal Hardware Error, no longer"
+			    " usable, SN %.16s\n", dd->serial);
+		/*
+		 * For /sys status file and user programs to print; if no
+		 * trailing brace is copied, we'll know it was truncated.
+		 */
+		if (dd->freezemsg)
+			snprintf(dd->freezemsg, dd->freezelen,
+				 "{%s}", msg);
+		qib_disable_after_error(dd);
+	}
+bail:;
+}
+
+/**
+ * qib_7220_init_hwerrors - enable hardware errors
+ * @dd: the qlogic_ib device
+ *
+ * now that we have finished initializing everything that might reasonably
+ * cause a hardware error, and cleared those errors bits as they occur,
+ * we can enable hardware errors in the mask (potentially enabling
+ * freeze mode), and enable hardware errors as errors (along with
+ * everything else) in errormask
+ */
+static void qib_7220_init_hwerrors(struct qib_devdata *dd)
+{
+	u64 val;
+	u64 extsval;
+
+	extsval = qib_read_kreg64(dd, kr_extstatus);
+
+	if (!(extsval & (QLOGIC_IB_EXTS_MEMBIST_ENDTEST |
+			 QLOGIC_IB_EXTS_MEMBIST_DISABLED)))
+		qib_dev_err(dd, "MemBIST did not complete!\n");
+	if (extsval & QLOGIC_IB_EXTS_MEMBIST_DISABLED)
+		qib_devinfo(dd->pcidev, "MemBIST is disabled.\n");
+
+	val = ~0ULL;    /* default to all hwerrors become interrupts, */
+
+	val &= ~QLOGIC_IB_HWE_IB_UC_MEMORYPARITYERR;
+	dd->cspec->hwerrmask = val;
+
+	qib_write_kreg(dd, kr_hwerrclear, ~HWE_MASK(PowerOnBISTFailed));
+	qib_write_kreg(dd, kr_hwerrmask, dd->cspec->hwerrmask);
+
+	/* clear all */
+	qib_write_kreg(dd, kr_errclear, ~0ULL);
+	/* enable errors that are masked, at least this first time. */
+	qib_write_kreg(dd, kr_errmask, ~0ULL);
+	dd->cspec->errormask = qib_read_kreg64(dd, kr_errmask);
+	/* clear any interrupts up to this point (ints still not enabled) */
+	qib_write_kreg(dd, kr_intclear, ~0ULL);
+}
+
+/*
+ * Disable and enable the armlaunch error.  Used for PIO bandwidth testing
+ * on chips that are count-based, rather than trigger-based.  There is no
+ * reference counting, but that's also fine, given the intended use.
+ * Only chip-specific because it's all register accesses
+ */
+static void qib_set_7220_armlaunch(struct qib_devdata *dd, u32 enable)
+{
+	if (enable) {
+		qib_write_kreg(dd, kr_errclear, ERR_MASK(SendPioArmLaunchErr));
+		dd->cspec->errormask |= ERR_MASK(SendPioArmLaunchErr);
+	} else
+		dd->cspec->errormask &= ~ERR_MASK(SendPioArmLaunchErr);
+	qib_write_kreg(dd, kr_errmask, dd->cspec->errormask);
+}
+
+/*
+ * Formerly took parameter <which> in pre-shifted,
+ * pre-merged form with LinkCmd and LinkInitCmd
+ * together, and assuming the zero was NOP.
+ */
+static void qib_set_ib_7220_lstate(struct qib_pportdata *ppd, u16 linkcmd,
+				   u16 linitcmd)
+{
+	u64 mod_wd;
+	struct qib_devdata *dd = ppd->dd;
+	unsigned long flags;
+
+	if (linitcmd == QLOGIC_IB_IBCC_LINKINITCMD_DISABLE) {
+		/*
+		 * If we are told to disable, note that so link-recovery
+		 * code does not attempt to bring us back up.
+		 */
+		spin_lock_irqsave(&ppd->lflags_lock, flags);
+		ppd->lflags |= QIBL_IB_LINK_DISABLED;
+		spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+	} else if (linitcmd || linkcmd == QLOGIC_IB_IBCC_LINKCMD_DOWN) {
+		/*
+		 * Any other linkinitcmd will lead to LINKDOWN and then
+		 * to INIT (if all is well), so clear flag to let
+		 * link-recovery code attempt to bring us back up.
+		 */
+		spin_lock_irqsave(&ppd->lflags_lock, flags);
+		ppd->lflags &= ~QIBL_IB_LINK_DISABLED;
+		spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+	}
+
+	mod_wd = (linkcmd << IBA7220_IBCC_LINKCMD_SHIFT) |
+		(linitcmd << QLOGIC_IB_IBCC_LINKINITCMD_SHIFT);
+
+	qib_write_kreg(dd, kr_ibcctrl, ppd->cpspec->ibcctrl | mod_wd);
+	/* write to chip to prevent back-to-back writes of ibc reg */
+	qib_write_kreg(dd, kr_scratch, 0);
+}
+
+/*
+ * All detailed interaction with the SerDes has been moved to qib_sd7220.c
+ *
+ * The portion of IBA7220-specific bringup_serdes() that actually deals with
+ * registers and memory within the SerDes itself is qib_sd7220_init().
+ */
+
+/**
+ * qib_7220_bringup_serdes - bring up the serdes
+ * @ppd: physical port on the qlogic_ib device
+ */
+static int qib_7220_bringup_serdes(struct qib_pportdata *ppd)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u64 val, prev_val, guid, ibc;
+	int ret = 0;
+
+	/* Put IBC in reset, sends disabled */
+	dd->control &= ~QLOGIC_IB_C_LINKENABLE;
+	qib_write_kreg(dd, kr_control, 0ULL);
+
+	if (qib_compat_ddr_negotiate) {
+		ppd->cpspec->ibdeltainprog = 1;
+		ppd->cpspec->ibsymsnap = read_7220_creg32(dd, cr_ibsymbolerr);
+		ppd->cpspec->iblnkerrsnap =
+			read_7220_creg32(dd, cr_iblinkerrrecov);
+	}
+
+	/* flowcontrolwatermark is in units of KBytes */
+	ibc = 0x5ULL << SYM_LSB(IBCCtrl, FlowCtrlWaterMark);
+	/*
+	 * How often flowctrl sent.  More or less in usecs; balance against
+	 * watermark value, so that in theory senders always get a flow
+	 * control update in time to not let the IB link go idle.
+	 */
+	ibc |= 0x3ULL << SYM_LSB(IBCCtrl, FlowCtrlPeriod);
+	/* max error tolerance */
+	ibc |= 0xfULL << SYM_LSB(IBCCtrl, PhyerrThreshold);
+	/* use "real" buffer space for */
+	ibc |= 4ULL << SYM_LSB(IBCCtrl, CreditScale);
+	/* IB credit flow control. */
+	ibc |= 0xfULL << SYM_LSB(IBCCtrl, OverrunThreshold);
+	/*
+	 * set initial max size pkt IBC will send, including ICRC; it's the
+	 * PIO buffer size in dwords, less 1; also see qib_set_mtu()
+	 */
+	ibc |= ((u64)(ppd->ibmaxlen >> 2) + 1) << SYM_LSB(IBCCtrl, MaxPktLen);
+	ppd->cpspec->ibcctrl = ibc; /* without linkcmd or linkinitcmd! */
+
+	/* initially come up waiting for TS1, without sending anything. */
+	val = ppd->cpspec->ibcctrl | (QLOGIC_IB_IBCC_LINKINITCMD_DISABLE <<
+		QLOGIC_IB_IBCC_LINKINITCMD_SHIFT);
+	qib_write_kreg(dd, kr_ibcctrl, val);
+
+	if (!ppd->cpspec->ibcddrctrl) {
+		/* not on re-init after reset */
+		ppd->cpspec->ibcddrctrl = qib_read_kreg64(dd, kr_ibcddrctrl);
+
+		if (ppd->link_speed_enabled == (QIB_IB_SDR | QIB_IB_DDR))
+			ppd->cpspec->ibcddrctrl |=
+				IBA7220_IBC_SPEED_AUTONEG_MASK |
+				IBA7220_IBC_IBTA_1_2_MASK;
+		else
+			ppd->cpspec->ibcddrctrl |=
+				ppd->link_speed_enabled == QIB_IB_DDR ?
+				IBA7220_IBC_SPEED_DDR : IBA7220_IBC_SPEED_SDR;
+		if ((ppd->link_width_enabled & (IB_WIDTH_1X | IB_WIDTH_4X)) ==
+		    (IB_WIDTH_1X | IB_WIDTH_4X))
+			ppd->cpspec->ibcddrctrl |= IBA7220_IBC_WIDTH_AUTONEG;
+		else
+			ppd->cpspec->ibcddrctrl |=
+				ppd->link_width_enabled == IB_WIDTH_4X ?
+				IBA7220_IBC_WIDTH_4X_ONLY :
+				IBA7220_IBC_WIDTH_1X_ONLY;
+
+		/* always enable these on driver reload, not sticky */
+		ppd->cpspec->ibcddrctrl |=
+			IBA7220_IBC_RXPOL_MASK << IBA7220_IBC_RXPOL_SHIFT;
+		ppd->cpspec->ibcddrctrl |=
+			IBA7220_IBC_HRTBT_MASK << IBA7220_IBC_HRTBT_SHIFT;
+
+		/* enable automatic lane reversal detection for receive */
+		ppd->cpspec->ibcddrctrl |= IBA7220_IBC_LANE_REV_SUPPORTED;
+	} else
+		/* write to chip to prevent back-to-back writes of ibc reg */
+		qib_write_kreg(dd, kr_scratch, 0);
+
+	qib_write_kreg(dd, kr_ibcddrctrl, ppd->cpspec->ibcddrctrl);
+	qib_write_kreg(dd, kr_scratch, 0);
+
+	qib_write_kreg(dd, kr_ncmodectrl, 0Ull);
+	qib_write_kreg(dd, kr_scratch, 0);
+
+	ret = qib_sd7220_init(dd);
+
+	val = qib_read_kreg64(dd, kr_xgxs_cfg);
+	prev_val = val;
+	val |= QLOGIC_IB_XGXS_FC_SAFE;
+	if (val != prev_val) {
+		qib_write_kreg(dd, kr_xgxs_cfg, val);
+		qib_read_kreg32(dd, kr_scratch);
+	}
+	if (val & QLOGIC_IB_XGXS_RESET)
+		val &= ~QLOGIC_IB_XGXS_RESET;
+	if (val != prev_val)
+		qib_write_kreg(dd, kr_xgxs_cfg, val);
+
+	/* first time through, set port guid */
+	if (!ppd->guid)
+		ppd->guid = dd->base_guid;
+	guid = be64_to_cpu(ppd->guid);
+
+	qib_write_kreg(dd, kr_hrtbt_guid, guid);
+	if (!ret) {
+		dd->control |= QLOGIC_IB_C_LINKENABLE;
+		qib_write_kreg(dd, kr_control, dd->control);
+	} else
+		/* write to chip to prevent back-to-back writes of ibc reg */
+		qib_write_kreg(dd, kr_scratch, 0);
+	return ret;
+}
+
+/**
+ * qib_7220_quiet_serdes - set serdes to txidle
+ * @ppd: physical port of the qlogic_ib device
+ * Called when driver is being unloaded
+ */
+static void qib_7220_quiet_serdes(struct qib_pportdata *ppd)
+{
+	u64 val;
+	struct qib_devdata *dd = ppd->dd;
+	unsigned long flags;
+
+	/* disable IBC */
+	dd->control &= ~QLOGIC_IB_C_LINKENABLE;
+	qib_write_kreg(dd, kr_control,
+		       dd->control | QLOGIC_IB_C_FREEZEMODE);
+
+	ppd->cpspec->chase_end = 0;
+	if (ppd->cpspec->chase_timer.data) /* if initted */
+		del_timer_sync(&ppd->cpspec->chase_timer);
+
+	if (ppd->cpspec->ibsymdelta || ppd->cpspec->iblnkerrdelta ||
+	    ppd->cpspec->ibdeltainprog) {
+		u64 diagc;
+
+		/* enable counter writes */
+		diagc = qib_read_kreg64(dd, kr_hwdiagctrl);
+		qib_write_kreg(dd, kr_hwdiagctrl,
+			       diagc | SYM_MASK(HwDiagCtrl, CounterWrEnable));
+
+		if (ppd->cpspec->ibsymdelta || ppd->cpspec->ibdeltainprog) {
+			val = read_7220_creg32(dd, cr_ibsymbolerr);
+			if (ppd->cpspec->ibdeltainprog)
+				val -= val - ppd->cpspec->ibsymsnap;
+			val -= ppd->cpspec->ibsymdelta;
+			write_7220_creg(dd, cr_ibsymbolerr, val);
+		}
+		if (ppd->cpspec->iblnkerrdelta || ppd->cpspec->ibdeltainprog) {
+			val = read_7220_creg32(dd, cr_iblinkerrrecov);
+			if (ppd->cpspec->ibdeltainprog)
+				val -= val - ppd->cpspec->iblnkerrsnap;
+			val -= ppd->cpspec->iblnkerrdelta;
+			write_7220_creg(dd, cr_iblinkerrrecov, val);
+		}
+
+		/* and disable counter writes */
+		qib_write_kreg(dd, kr_hwdiagctrl, diagc);
+	}
+	qib_set_ib_7220_lstate(ppd, 0, QLOGIC_IB_IBCC_LINKINITCMD_DISABLE);
+
+	spin_lock_irqsave(&ppd->lflags_lock, flags);
+	ppd->lflags &= ~QIBL_IB_AUTONEG_INPROG;
+	spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+	wake_up(&ppd->cpspec->autoneg_wait);
+	cancel_delayed_work(&ppd->cpspec->autoneg_work);
+	flush_scheduled_work();
+
+	shutdown_7220_relock_poll(ppd->dd);
+	val = qib_read_kreg64(ppd->dd, kr_xgxs_cfg);
+	val |= QLOGIC_IB_XGXS_RESET;
+	qib_write_kreg(ppd->dd, kr_xgxs_cfg, val);
+}
+
+/**
+ * qib_setup_7220_setextled - set the state of the two external LEDs
+ * @dd: the qlogic_ib device
+ * @on: whether the link is up or not
+ *
+ * The exact combo of LEDs if on is true is determined by looking
+ * at the ibcstatus.
+ *
+ * These LEDs indicate the physical and logical state of IB link.
+ * For this chip (at least with recommended board pinouts), LED1
+ * is Yellow (logical state) and LED2 is Green (physical state),
+ *
+ * Note:  We try to match the Mellanox HCA LED behavior as best
+ * we can.  Green indicates physical link state is OK (something is
+ * plugged in, and we can train).
+ * Amber indicates the link is logically up (ACTIVE).
+ * Mellanox further blinks the amber LED to indicate data packet
+ * activity, but we have no hardware support for that, so it would
+ * require waking up every 10-20 msecs and checking the counters
+ * on the chip, and then turning the LED off if appropriate.  That's
+ * visible overhead, so not something we will do.
+ *
+ */
+static void qib_setup_7220_setextled(struct qib_pportdata *ppd, u32 on)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u64 extctl, ledblink = 0, val, lst, ltst;
+	unsigned long flags;
+
+	/*
+	 * The diags use the LED to indicate diag info, so we leave
+	 * the external LED alone when the diags are running.
+	 */
+	if (dd->diag_client)
+		return;
+
+	if (ppd->led_override) {
+		ltst = (ppd->led_override & QIB_LED_PHYS) ?
+			IB_PHYSPORTSTATE_LINKUP : IB_PHYSPORTSTATE_DISABLED,
+		lst = (ppd->led_override & QIB_LED_LOG) ?
+			IB_PORT_ACTIVE : IB_PORT_DOWN;
+	} else if (on) {
+		val = qib_read_kreg64(dd, kr_ibcstatus);
+		ltst = qib_7220_phys_portstate(val);
+		lst = qib_7220_iblink_state(val);
+	} else {
+		ltst = 0;
+		lst = 0;
+	}
+
+	spin_lock_irqsave(&dd->cspec->gpio_lock, flags);
+	extctl = dd->cspec->extctrl & ~(SYM_MASK(EXTCtrl, LEDPriPortGreenOn) |
+				 SYM_MASK(EXTCtrl, LEDPriPortYellowOn));
+	if (ltst == IB_PHYSPORTSTATE_LINKUP) {
+		extctl |= SYM_MASK(EXTCtrl, LEDPriPortGreenOn);
+		/*
+		 * counts are in chip clock (4ns) periods.
+		 * This is 1/16 sec (66.6ms) on,
+		 * 3/16 sec (187.5 ms) off, with packets rcvd
+		 */
+		ledblink = ((66600 * 1000UL / 4) << IBA7220_LEDBLINK_ON_SHIFT)
+			| ((187500 * 1000UL / 4) << IBA7220_LEDBLINK_OFF_SHIFT);
+	}
+	if (lst == IB_PORT_ACTIVE)
+		extctl |= SYM_MASK(EXTCtrl, LEDPriPortYellowOn);
+	dd->cspec->extctrl = extctl;
+	qib_write_kreg(dd, kr_extctrl, extctl);
+	spin_unlock_irqrestore(&dd->cspec->gpio_lock, flags);
+
+	if (ledblink) /* blink the LED on packet receive */
+		qib_write_kreg(dd, kr_rcvpktledcnt, ledblink);
+}
+
+static void qib_7220_free_irq(struct qib_devdata *dd)
+{
+	if (dd->cspec->irq) {
+		free_irq(dd->cspec->irq, dd);
+		dd->cspec->irq = 0;
+	}
+	qib_nomsi(dd);
+}
+
+/*
+ * qib_setup_7220_cleanup - clean up any per-chip chip-specific stuff
+ * @dd: the qlogic_ib device
+ *
+ * This is called during driver unload.
+ *
+ */
+static void qib_setup_7220_cleanup(struct qib_devdata *dd)
+{
+	qib_7220_free_irq(dd);
+	kfree(dd->cspec->cntrs);
+	kfree(dd->cspec->portcntrs);
+}
+
+/*
+ * This is only called for SDmaInt.
+ * SDmaDisabled is handled on the error path.
+ */
+static void sdma_7220_intr(struct qib_pportdata *ppd, u64 istat)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&ppd->sdma_lock, flags);
+
+	switch (ppd->sdma_state.current_state) {
+	case qib_sdma_state_s00_hw_down:
+		break;
+
+	case qib_sdma_state_s10_hw_start_up_wait:
+		__qib_sdma_process_event(ppd, qib_sdma_event_e20_hw_started);
+		break;
+
+	case qib_sdma_state_s20_idle:
+		break;
+
+	case qib_sdma_state_s30_sw_clean_up_wait:
+		break;
+
+	case qib_sdma_state_s40_hw_clean_up_wait:
+		break;
+
+	case qib_sdma_state_s50_hw_halt_wait:
+		__qib_sdma_process_event(ppd, qib_sdma_event_e60_hw_halted);
+		break;
+
+	case qib_sdma_state_s99_running:
+		/* too chatty to print here */
+		__qib_sdma_intr(ppd);
+		break;
+	}
+	spin_unlock_irqrestore(&ppd->sdma_lock, flags);
+}
+
+static void qib_wantpiobuf_7220_intr(struct qib_devdata *dd, u32 needint)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&dd->sendctrl_lock, flags);
+	if (needint) {
+		if (!(dd->sendctrl & SYM_MASK(SendCtrl, SendBufAvailUpd)))
+			goto done;
+		/*
+		 * blip the availupd off, next write will be on, so
+		 * we ensure an avail update, regardless of threshold or
+		 * buffers becoming free, whenever we want an interrupt
+		 */
+		qib_write_kreg(dd, kr_sendctrl, dd->sendctrl &
+			~SYM_MASK(SendCtrl, SendBufAvailUpd));
+		qib_write_kreg(dd, kr_scratch, 0ULL);
+		dd->sendctrl |= SYM_MASK(SendCtrl, SendIntBufAvail);
+	} else
+		dd->sendctrl &= ~SYM_MASK(SendCtrl, SendIntBufAvail);
+	qib_write_kreg(dd, kr_sendctrl, dd->sendctrl);
+	qib_write_kreg(dd, kr_scratch, 0ULL);
+done:
+	spin_unlock_irqrestore(&dd->sendctrl_lock, flags);
+}
+
+/*
+ * Handle errors and unusual events first, separate function
+ * to improve cache hits for fast path interrupt handling.
+ */
+static noinline void unlikely_7220_intr(struct qib_devdata *dd, u64 istat)
+{
+	if (unlikely(istat & ~QLOGIC_IB_I_BITSEXTANT))
+		qib_dev_err(dd,
+			    "interrupt with unknown interrupts %Lx set\n",
+			    istat & ~QLOGIC_IB_I_BITSEXTANT);
+
+	if (istat & QLOGIC_IB_I_GPIO) {
+		u32 gpiostatus;
+
+		/*
+		 * Boards for this chip currently don't use GPIO interrupts,
+		 * so clear by writing GPIOstatus to GPIOclear, and complain
+		 * to alert developer. To avoid endless repeats, clear
+		 * the bits in the mask, since there is some kind of
+		 * programming error or chip problem.
+		 */
+		gpiostatus = qib_read_kreg32(dd, kr_gpio_status);
+		/*
+		 * In theory, writing GPIOstatus to GPIOclear could
+		 * have a bad side-effect on some diagnostic that wanted
+		 * to poll for a status-change, but the various shadows
+		 * make that problematic at best. Diags will just suppress
+		 * all GPIO interrupts during such tests.
+		 */
+		qib_write_kreg(dd, kr_gpio_clear, gpiostatus);
+
+		if (gpiostatus) {
+			const u32 mask = qib_read_kreg32(dd, kr_gpio_mask);
+			u32 gpio_irq = mask & gpiostatus;
+
+			/*
+			 * A bit set in status and (chip) Mask register
+			 * would cause an interrupt. Since we are not
+			 * expecting any, report it. Also check that the
+			 * chip reflects our shadow, report issues,
+			 * and refresh from the shadow.
+			 */
+			/*
+			 * Clear any troublemakers, and update chip
+			 * from shadow
+			 */
+			dd->cspec->gpio_mask &= ~gpio_irq;
+			qib_write_kreg(dd, kr_gpio_mask, dd->cspec->gpio_mask);
+		}
+	}
+
+	if (istat & QLOGIC_IB_I_ERROR) {
+		u64 estat;
+
+		qib_stats.sps_errints++;
+		estat = qib_read_kreg64(dd, kr_errstatus);
+		if (!estat)
+			qib_devinfo(dd->pcidev, "error interrupt (%Lx), "
+				 "but no error bits set!\n", istat);
+		else
+			handle_7220_errors(dd, estat);
+	}
+}
+
+static irqreturn_t qib_7220intr(int irq, void *data)
+{
+	struct qib_devdata *dd = data;
+	irqreturn_t ret;
+	u64 istat;
+	u64 ctxtrbits;
+	u64 rmask;
+	unsigned i;
+
+	if ((dd->flags & (QIB_PRESENT | QIB_BADINTR)) != QIB_PRESENT) {
+		/*
+		 * This return value is not great, but we do not want the
+		 * interrupt core code to remove our interrupt handler
+		 * because we don't appear to be handling an interrupt
+		 * during a chip reset.
+		 */
+		ret = IRQ_HANDLED;
+		goto bail;
+	}
+
+	istat = qib_read_kreg64(dd, kr_intstatus);
+
+	if (unlikely(!istat)) {
+		ret = IRQ_NONE; /* not our interrupt, or already handled */
+		goto bail;
+	}
+	if (unlikely(istat == -1)) {
+		qib_bad_intrstatus(dd);
+		/* don't know if it was our interrupt or not */
+		ret = IRQ_NONE;
+		goto bail;
+	}
+
+	qib_stats.sps_ints++;
+	if (dd->int_counter != (u32) -1)
+		dd->int_counter++;
+
+	if (unlikely(istat & (~QLOGIC_IB_I_BITSEXTANT |
+			      QLOGIC_IB_I_GPIO | QLOGIC_IB_I_ERROR)))
+		unlikely_7220_intr(dd, istat);
+
+	/*
+	 * Clear the interrupt bits we found set, relatively early, so we
+	 * "know" know the chip will have seen this by the time we process
+	 * the queue, and will re-interrupt if necessary.  The processor
+	 * itself won't take the interrupt again until we return.
+	 */
+	qib_write_kreg(dd, kr_intclear, istat);
+
+	/*
+	 * Handle kernel receive queues before checking for pio buffers
+	 * available since receives can overflow; piobuf waiters can afford
+	 * a few extra cycles, since they were waiting anyway.
+	 */
+	ctxtrbits = istat &
+		((QLOGIC_IB_I_RCVAVAIL_MASK << QLOGIC_IB_I_RCVAVAIL_SHIFT) |
+		 (QLOGIC_IB_I_RCVURG_MASK << QLOGIC_IB_I_RCVURG_SHIFT));
+	if (ctxtrbits) {
+		rmask = (1ULL << QLOGIC_IB_I_RCVAVAIL_SHIFT) |
+			(1ULL << QLOGIC_IB_I_RCVURG_SHIFT);
+		for (i = 0; i < dd->first_user_ctxt; i++) {
+			if (ctxtrbits & rmask) {
+				ctxtrbits &= ~rmask;
+				qib_kreceive(dd->rcd[i], NULL, NULL);
+			}
+			rmask <<= 1;
+		}
+		if (ctxtrbits) {
+			ctxtrbits =
+				(ctxtrbits >> QLOGIC_IB_I_RCVAVAIL_SHIFT) |
+				(ctxtrbits >> QLOGIC_IB_I_RCVURG_SHIFT);
+			qib_handle_urcv(dd, ctxtrbits);
+		}
+	}
+
+	/* only call for SDmaInt */
+	if (istat & QLOGIC_IB_I_SDMAINT)
+		sdma_7220_intr(dd->pport, istat);
+
+	if ((istat & QLOGIC_IB_I_SPIOBUFAVAIL) && (dd->flags & QIB_INITTED))
+		qib_ib_piobufavail(dd);
+
+	ret = IRQ_HANDLED;
+bail:
+	return ret;
+}
+
+/*
+ * Set up our chip-specific interrupt handler.
+ * The interrupt type has already been setup, so
+ * we just need to do the registration and error checking.
+ * If we are using MSI interrupts, we may fall back to
+ * INTx later, if the interrupt handler doesn't get called
+ * within 1/2 second (see verify_interrupt()).
+ */
+static void qib_setup_7220_interrupt(struct qib_devdata *dd)
+{
+	if (!dd->cspec->irq)
+		qib_dev_err(dd, "irq is 0, BIOS error?  Interrupts won't "
+			    "work\n");
+	else {
+		int ret = request_irq(dd->cspec->irq, qib_7220intr,
+			dd->msi_lo ? 0 : IRQF_SHARED,
+			QIB_DRV_NAME, dd);
+
+		if (ret)
+			qib_dev_err(dd, "Couldn't setup %s interrupt "
+				    "(irq=%d): %d\n", dd->msi_lo ?
+				    "MSI" : "INTx", dd->cspec->irq, ret);
+	}
+}
+
+/**
+ * qib_7220_boardname - fill in the board name
+ * @dd: the qlogic_ib device
+ *
+ * info is based on the board revision register
+ */
+static void qib_7220_boardname(struct qib_devdata *dd)
+{
+	char *n;
+	u32 boardid, namelen;
+
+	boardid = SYM_FIELD(dd->revision, Revision,
+			    BoardID);
+
+	switch (boardid) {
+	case 1:
+		n = "InfiniPath_QLE7240";
+		break;
+	case 2:
+		n = "InfiniPath_QLE7280";
+		break;
+	default:
+		qib_dev_err(dd, "Unknown 7220 board with ID %u\n", boardid);
+		n = "Unknown_InfiniPath_7220";
+		break;
+	}
+
+	namelen = strlen(n) + 1;
+	dd->boardname = kmalloc(namelen, GFP_KERNEL);
+	if (!dd->boardname)
+		qib_dev_err(dd, "Failed allocation for board name: %s\n", n);
+	else
+		snprintf(dd->boardname, namelen, "%s", n);
+
+	if (dd->majrev != 5 || !dd->minrev || dd->minrev > 2)
+		qib_dev_err(dd, "Unsupported InfiniPath hardware "
+			    "revision %u.%u!\n",
+			    dd->majrev, dd->minrev);
+
+	snprintf(dd->boardversion, sizeof(dd->boardversion),
+		 "ChipABI %u.%u, %s, InfiniPath%u %u.%u, SW Compat %u\n",
+		 QIB_CHIP_VERS_MAJ, QIB_CHIP_VERS_MIN, dd->boardname,
+		 (unsigned)SYM_FIELD(dd->revision, Revision_R, Arch),
+		 dd->majrev, dd->minrev,
+		 (unsigned)SYM_FIELD(dd->revision, Revision_R, SW));
+}
+
+/*
+ * This routine sleeps, so it can only be called from user context, not
+ * from interrupt context.
+ */
+static int qib_setup_7220_reset(struct qib_devdata *dd)
+{
+	u64 val;
+	int i;
+	int ret;
+	u16 cmdval;
+	u8 int_line, clinesz;
+	unsigned long flags;
+
+	qib_pcie_getcmd(dd, &cmdval, &int_line, &clinesz);
+
+	/* Use dev_err so it shows up in logs, etc. */
+	qib_dev_err(dd, "Resetting InfiniPath unit %u\n", dd->unit);
+
+	/* no interrupts till re-initted */
+	qib_7220_set_intr_state(dd, 0);
+
+	dd->pport->cpspec->ibdeltainprog = 0;
+	dd->pport->cpspec->ibsymdelta = 0;
+	dd->pport->cpspec->iblnkerrdelta = 0;
+
+	/*
+	 * Keep chip from being accessed until we are ready.  Use
+	 * writeq() directly, to allow the write even though QIB_PRESENT
+	 * isnt' set.
+	 */
+	dd->flags &= ~(QIB_INITTED | QIB_PRESENT);
+	dd->int_counter = 0; /* so we check interrupts work again */
+	val = dd->control | QLOGIC_IB_C_RESET;
+	writeq(val, &dd->kregbase[kr_control]);
+	mb(); /* prevent compiler reordering around actual reset */
+
+	for (i = 1; i <= 5; i++) {
+		/*
+		 * Allow MBIST, etc. to complete; longer on each retry.
+		 * We sometimes get machine checks from bus timeout if no
+		 * response, so for now, make it *really* long.
+		 */
+		msleep(1000 + (1 + i) * 2000);
+
+		qib_pcie_reenable(dd, cmdval, int_line, clinesz);
+
+		/*
+		 * Use readq directly, so we don't need to mark it as PRESENT
+		 * until we get a successful indication that all is well.
+		 */
+		val = readq(&dd->kregbase[kr_revision]);
+		if (val == dd->revision) {
+			dd->flags |= QIB_PRESENT; /* it's back */
+			ret = qib_reinit_intr(dd);
+			goto bail;
+		}
+	}
+	ret = 0; /* failed */
+
+bail:
+	if (ret) {
+		if (qib_pcie_params(dd, dd->lbus_width, NULL, NULL))
+			qib_dev_err(dd, "Reset failed to setup PCIe or "
+				    "interrupts; continuing anyway\n");
+
+		/* hold IBC in reset, no sends, etc till later */
+		qib_write_kreg(dd, kr_control, 0ULL);
+
+		/* clear the reset error, init error/hwerror mask */
+		qib_7220_init_hwerrors(dd);
+
+		/* do setup similar to speed or link-width changes */
+		if (dd->pport->cpspec->ibcddrctrl & IBA7220_IBC_IBTA_1_2_MASK)
+			dd->cspec->presets_needed = 1;
+		spin_lock_irqsave(&dd->pport->lflags_lock, flags);
+		dd->pport->lflags |= QIBL_IB_FORCE_NOTIFY;
+		dd->pport->lflags &= ~QIBL_IB_AUTONEG_FAILED;
+		spin_unlock_irqrestore(&dd->pport->lflags_lock, flags);
+	}
+
+	return ret;
+}
+
+/**
+ * qib_7220_put_tid - write a TID to the chip
+ * @dd: the qlogic_ib device
+ * @tidptr: pointer to the expected TID (in chip) to update
+ * @tidtype: 0 for eager, 1 for expected
+ * @pa: physical address of in memory buffer; tidinvalid if freeing
+ */
+static void qib_7220_put_tid(struct qib_devdata *dd, u64 __iomem *tidptr,
+			     u32 type, unsigned long pa)
+{
+	if (pa != dd->tidinvalid) {
+		u64 chippa = pa >> IBA7220_TID_PA_SHIFT;
+
+		/* paranoia checks */
+		if (pa != (chippa << IBA7220_TID_PA_SHIFT)) {
+			qib_dev_err(dd, "Physaddr %lx not 2KB aligned!\n",
+				    pa);
+			return;
+		}
+		if (chippa >= (1UL << IBA7220_TID_SZ_SHIFT)) {
+			qib_dev_err(dd, "Physical page address 0x%lx "
+				"larger than supported\n", pa);
+			return;
+		}
+
+		if (type == RCVHQ_RCV_TYPE_EAGER)
+			chippa |= dd->tidtemplate;
+		else /* for now, always full 4KB page */
+			chippa |= IBA7220_TID_SZ_4K;
+		pa = chippa;
+	}
+	writeq(pa, tidptr);
+	mmiowb();
+}
+
+/**
+ * qib_7220_clear_tids - clear all TID entries for a ctxt, expected and eager
+ * @dd: the qlogic_ib device
+ * @ctxt: the ctxt
+ *
+ * clear all TID entries for a ctxt, expected and eager.
+ * Used from qib_close().  On this chip, TIDs are only 32 bits,
+ * not 64, but they are still on 64 bit boundaries, so tidbase
+ * is declared as u64 * for the pointer math, even though we write 32 bits
+ */
+static void qib_7220_clear_tids(struct qib_devdata *dd,
+				struct qib_ctxtdata *rcd)
+{
+	u64 __iomem *tidbase;
+	unsigned long tidinv;
+	u32 ctxt;
+	int i;
+
+	if (!dd->kregbase || !rcd)
+		return;
+
+	ctxt = rcd->ctxt;
+
+	tidinv = dd->tidinvalid;
+	tidbase = (u64 __iomem *)
+		((char __iomem *)(dd->kregbase) +
+		 dd->rcvtidbase +
+		 ctxt * dd->rcvtidcnt * sizeof(*tidbase));
+
+	for (i = 0; i < dd->rcvtidcnt; i++)
+		qib_7220_put_tid(dd, &tidbase[i], RCVHQ_RCV_TYPE_EXPECTED,
+				 tidinv);
+
+	tidbase = (u64 __iomem *)
+		((char __iomem *)(dd->kregbase) +
+		 dd->rcvegrbase +
+		 rcd->rcvegr_tid_base * sizeof(*tidbase));
+
+	for (i = 0; i < rcd->rcvegrcnt; i++)
+		qib_7220_put_tid(dd, &tidbase[i], RCVHQ_RCV_TYPE_EAGER,
+				 tidinv);
+}
+
+/**
+ * qib_7220_tidtemplate - setup constants for TID updates
+ * @dd: the qlogic_ib device
+ *
+ * We setup stuff that we use a lot, to avoid calculating each time
+ */
+static void qib_7220_tidtemplate(struct qib_devdata *dd)
+{
+	if (dd->rcvegrbufsize == 2048)
+		dd->tidtemplate = IBA7220_TID_SZ_2K;
+	else if (dd->rcvegrbufsize == 4096)
+		dd->tidtemplate = IBA7220_TID_SZ_4K;
+	dd->tidinvalid = 0;
+}
+
+/**
+ * qib_init_7220_get_base_info - set chip-specific flags for user code
+ * @rcd: the qlogic_ib ctxt
+ * @kbase: qib_base_info pointer
+ *
+ * We set the PCIE flag because the lower bandwidth on PCIe vs
+ * HyperTransport can affect some user packet algorithims.
+ */
+static int qib_7220_get_base_info(struct qib_ctxtdata *rcd,
+				  struct qib_base_info *kinfo)
+{
+	kinfo->spi_runtime_flags |= QIB_RUNTIME_PCIE |
+		QIB_RUNTIME_NODMA_RTAIL | QIB_RUNTIME_SDMA;
+
+	if (rcd->dd->flags & QIB_USE_SPCL_TRIG)
+		kinfo->spi_runtime_flags |= QIB_RUNTIME_SPECIAL_TRIGGER;
+
+	return 0;
+}
+
+static struct qib_message_header *
+qib_7220_get_msgheader(struct qib_devdata *dd, __le32 *rhf_addr)
+{
+	u32 offset = qib_hdrget_offset(rhf_addr);
+
+	return (struct qib_message_header *)
+		(rhf_addr - dd->rhf_offset + offset);
+}
+
+static void qib_7220_config_ctxts(struct qib_devdata *dd)
+{
+	unsigned long flags;
+	u32 nchipctxts;
+
+	nchipctxts = qib_read_kreg32(dd, kr_portcnt);
+	dd->cspec->numctxts = nchipctxts;
+	if (qib_n_krcv_queues > 1) {
+		dd->qpn_mask = 0x3f;
+		dd->first_user_ctxt = qib_n_krcv_queues * dd->num_pports;
+		if (dd->first_user_ctxt > nchipctxts)
+			dd->first_user_ctxt = nchipctxts;
+	} else
+		dd->first_user_ctxt = dd->num_pports;
+	dd->n_krcv_queues = dd->first_user_ctxt;
+
+	if (!qib_cfgctxts) {
+		int nctxts = dd->first_user_ctxt + num_online_cpus();
+
+		if (nctxts <= 5)
+			dd->ctxtcnt = 5;
+		else if (nctxts <= 9)
+			dd->ctxtcnt = 9;
+		else if (nctxts <= nchipctxts)
+			dd->ctxtcnt = nchipctxts;
+	} else if (qib_cfgctxts <= nchipctxts)
+		dd->ctxtcnt = qib_cfgctxts;
+	if (!dd->ctxtcnt) /* none of the above, set to max */
+		dd->ctxtcnt = nchipctxts;
+
+	/*
+	 * Chip can be configured for 5, 9, or 17 ctxts, and choice
+	 * affects number of eager TIDs per ctxt (1K, 2K, 4K).
+	 * Lock to be paranoid about later motion, etc.
+	 */
+	spin_lock_irqsave(&dd->cspec->rcvmod_lock, flags);
+	if (dd->ctxtcnt > 9)
+		dd->rcvctrl |= 2ULL << IBA7220_R_CTXTCFG_SHIFT;
+	else if (dd->ctxtcnt > 5)
+		dd->rcvctrl |= 1ULL << IBA7220_R_CTXTCFG_SHIFT;
+	/* else configure for default 5 receive ctxts */
+	if (dd->qpn_mask)
+		dd->rcvctrl |= 1ULL << QIB_7220_RcvCtrl_RcvQPMapEnable_LSB;
+	qib_write_kreg(dd, kr_rcvctrl, dd->rcvctrl);
+	spin_unlock_irqrestore(&dd->cspec->rcvmod_lock, flags);
+
+	/* kr_rcvegrcnt changes based on the number of contexts enabled */
+	dd->cspec->rcvegrcnt = qib_read_kreg32(dd, kr_rcvegrcnt);
+	dd->rcvhdrcnt = max(dd->cspec->rcvegrcnt, IBA7220_KRCVEGRCNT);
+}
+
+static int qib_7220_get_ib_cfg(struct qib_pportdata *ppd, int which)
+{
+	int lsb, ret = 0;
+	u64 maskr; /* right-justified mask */
+
+	switch (which) {
+	case QIB_IB_CFG_LWID_ENB: /* Get allowed Link-width */
+		ret = ppd->link_width_enabled;
+		goto done;
+
+	case QIB_IB_CFG_LWID: /* Get currently active Link-width */
+		ret = ppd->link_width_active;
+		goto done;
+
+	case QIB_IB_CFG_SPD_ENB: /* Get allowed Link speeds */
+		ret = ppd->link_speed_enabled;
+		goto done;
+
+	case QIB_IB_CFG_SPD: /* Get current Link spd */
+		ret = ppd->link_speed_active;
+		goto done;
+
+	case QIB_IB_CFG_RXPOL_ENB: /* Get Auto-RX-polarity enable */
+		lsb = IBA7220_IBC_RXPOL_SHIFT;
+		maskr = IBA7220_IBC_RXPOL_MASK;
+		break;
+
+	case QIB_IB_CFG_LREV_ENB: /* Get Auto-Lane-reversal enable */
+		lsb = IBA7220_IBC_LREV_SHIFT;
+		maskr = IBA7220_IBC_LREV_MASK;
+		break;
+
+	case QIB_IB_CFG_LINKLATENCY:
+		ret = qib_read_kreg64(ppd->dd, kr_ibcddrstatus)
+			& IBA7220_DDRSTAT_LINKLAT_MASK;
+		goto done;
+
+	case QIB_IB_CFG_OP_VLS:
+		ret = ppd->vls_operational;
+		goto done;
+
+	case QIB_IB_CFG_VL_HIGH_CAP:
+		ret = 0;
+		goto done;
+
+	case QIB_IB_CFG_VL_LOW_CAP:
+		ret = 0;
+		goto done;
+
+	case QIB_IB_CFG_OVERRUN_THRESH: /* IB overrun threshold */
+		ret = SYM_FIELD(ppd->cpspec->ibcctrl, IBCCtrl,
+				OverrunThreshold);
+		goto done;
+
+	case QIB_IB_CFG_PHYERR_THRESH: /* IB PHY error threshold */
+		ret = SYM_FIELD(ppd->cpspec->ibcctrl, IBCCtrl,
+				PhyerrThreshold);
+		goto done;
+
+	case QIB_IB_CFG_LINKDEFAULT: /* IB link default (sleep/poll) */
+		/* will only take effect when the link state changes */
+		ret = (ppd->cpspec->ibcctrl &
+		       SYM_MASK(IBCCtrl, LinkDownDefaultState)) ?
+			IB_LINKINITCMD_SLEEP : IB_LINKINITCMD_POLL;
+		goto done;
+
+	case QIB_IB_CFG_HRTBT: /* Get Heartbeat off/enable/auto */
+		lsb = IBA7220_IBC_HRTBT_SHIFT;
+		maskr = IBA7220_IBC_HRTBT_MASK;
+		break;
+
+	case QIB_IB_CFG_PMA_TICKS:
+		/*
+		 * 0x00 = 10x link transfer rate or 4 nsec. for 2.5Gbs
+		 * Since the clock is always 250MHz, the value is 1 or 0.
+		 */
+		ret = (ppd->link_speed_active == QIB_IB_DDR);
+		goto done;
+
+	default:
+		ret = -EINVAL;
+		goto done;
+	}
+	ret = (int)((ppd->cpspec->ibcddrctrl >> lsb) & maskr);
+done:
+	return ret;
+}
+
+static int qib_7220_set_ib_cfg(struct qib_pportdata *ppd, int which, u32 val)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u64 maskr; /* right-justified mask */
+	int lsb, ret = 0, setforce = 0;
+	u16 lcmd, licmd;
+	unsigned long flags;
+
+	switch (which) {
+	case QIB_IB_CFG_LIDLMC:
+		/*
+		 * Set LID and LMC. Combined to avoid possible hazard
+		 * caller puts LMC in 16MSbits, DLID in 16LSbits of val
+		 */
+		lsb = IBA7220_IBC_DLIDLMC_SHIFT;
+		maskr = IBA7220_IBC_DLIDLMC_MASK;
+		break;
+
+	case QIB_IB_CFG_LWID_ENB: /* set allowed Link-width */
+		/*
+		 * As with speed, only write the actual register if
+		 * the link is currently down, otherwise takes effect
+		 * on next link change.
+		 */
+		ppd->link_width_enabled = val;
+		if (!(ppd->lflags & QIBL_LINKDOWN))
+			goto bail;
+		/*
+		 * We set the QIBL_IB_FORCE_NOTIFY bit so updown
+		 * will get called because we want update
+		 * link_width_active, and the change may not take
+		 * effect for some time (if we are in POLL), so this
+		 * flag will force the updown routine to be called
+		 * on the next ibstatuschange down interrupt, even
+		 * if it's not an down->up transition.
+		 */
+		val--; /* convert from IB to chip */
+		maskr = IBA7220_IBC_WIDTH_MASK;
+		lsb = IBA7220_IBC_WIDTH_SHIFT;
+		setforce = 1;
+		spin_lock_irqsave(&ppd->lflags_lock, flags);
+		ppd->lflags |= QIBL_IB_FORCE_NOTIFY;
+		spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+		break;
+
+	case QIB_IB_CFG_SPD_ENB: /* set allowed Link speeds */
+		/*
+		 * If we turn off IB1.2, need to preset SerDes defaults,
+		 * but not right now. Set a flag for the next time
+		 * we command the link down.  As with width, only write the
+		 * actual register if the link is currently down, otherwise
+		 * takes effect on next link change.  Since setting is being
+		 * explictly requested (via MAD or sysfs), clear autoneg
+		 * failure status if speed autoneg is enabled.
+		 */
+		ppd->link_speed_enabled = val;
+		if ((ppd->cpspec->ibcddrctrl & IBA7220_IBC_IBTA_1_2_MASK) &&
+		    !(val & (val - 1)))
+			dd->cspec->presets_needed = 1;
+		if (!(ppd->lflags & QIBL_LINKDOWN))
+			goto bail;
+		/*
+		 * We set the QIBL_IB_FORCE_NOTIFY bit so updown
+		 * will get called because we want update
+		 * link_speed_active, and the change may not take
+		 * effect for some time (if we are in POLL), so this
+		 * flag will force the updown routine to be called
+		 * on the next ibstatuschange down interrupt, even
+		 * if it's not an down->up transition.
+		 */
+		if (val == (QIB_IB_SDR | QIB_IB_DDR)) {
+			val = IBA7220_IBC_SPEED_AUTONEG_MASK |
+				IBA7220_IBC_IBTA_1_2_MASK;
+			spin_lock_irqsave(&ppd->lflags_lock, flags);
+			ppd->lflags &= ~QIBL_IB_AUTONEG_FAILED;
+			spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+		} else
+			val = val == QIB_IB_DDR ?
+				IBA7220_IBC_SPEED_DDR : IBA7220_IBC_SPEED_SDR;
+		maskr = IBA7220_IBC_SPEED_AUTONEG_MASK |
+			IBA7220_IBC_IBTA_1_2_MASK;
+		/* IBTA 1.2 mode + speed bits are contiguous */
+		lsb = SYM_LSB(IBCDDRCtrl, IB_ENHANCED_MODE);
+		setforce = 1;
+		break;
+
+	case QIB_IB_CFG_RXPOL_ENB: /* set Auto-RX-polarity enable */
+		lsb = IBA7220_IBC_RXPOL_SHIFT;
+		maskr = IBA7220_IBC_RXPOL_MASK;
+		break;
+
+	case QIB_IB_CFG_LREV_ENB: /* set Auto-Lane-reversal enable */
+		lsb = IBA7220_IBC_LREV_SHIFT;
+		maskr = IBA7220_IBC_LREV_MASK;
+		break;
+
+	case QIB_IB_CFG_OVERRUN_THRESH: /* IB overrun threshold */
+		maskr = SYM_FIELD(ppd->cpspec->ibcctrl, IBCCtrl,
+				  OverrunThreshold);
+		if (maskr != val) {
+			ppd->cpspec->ibcctrl &=
+				~SYM_MASK(IBCCtrl, OverrunThreshold);
+			ppd->cpspec->ibcctrl |= (u64) val <<
+				SYM_LSB(IBCCtrl, OverrunThreshold);
+			qib_write_kreg(dd, kr_ibcctrl, ppd->cpspec->ibcctrl);
+			qib_write_kreg(dd, kr_scratch, 0);
+		}
+		goto bail;
+
+	case QIB_IB_CFG_PHYERR_THRESH: /* IB PHY error threshold */
+		maskr = SYM_FIELD(ppd->cpspec->ibcctrl, IBCCtrl,
+				  PhyerrThreshold);
+		if (maskr != val) {
+			ppd->cpspec->ibcctrl &=
+				~SYM_MASK(IBCCtrl, PhyerrThreshold);
+			ppd->cpspec->ibcctrl |= (u64) val <<
+				SYM_LSB(IBCCtrl, PhyerrThreshold);
+			qib_write_kreg(dd, kr_ibcctrl, ppd->cpspec->ibcctrl);
+			qib_write_kreg(dd, kr_scratch, 0);
+		}
+		goto bail;
+
+	case QIB_IB_CFG_PKEYS: /* update pkeys */
+		maskr = (u64) ppd->pkeys[0] | ((u64) ppd->pkeys[1] << 16) |
+			((u64) ppd->pkeys[2] << 32) |
+			((u64) ppd->pkeys[3] << 48);
+		qib_write_kreg(dd, kr_partitionkey, maskr);
+		goto bail;
+
+	case QIB_IB_CFG_LINKDEFAULT: /* IB link default (sleep/poll) */
+		/* will only take effect when the link state changes */
+		if (val == IB_LINKINITCMD_POLL)
+			ppd->cpspec->ibcctrl &=
+				~SYM_MASK(IBCCtrl, LinkDownDefaultState);
+		else /* SLEEP */
+			ppd->cpspec->ibcctrl |=
+				SYM_MASK(IBCCtrl, LinkDownDefaultState);
+		qib_write_kreg(dd, kr_ibcctrl, ppd->cpspec->ibcctrl);
+		qib_write_kreg(dd, kr_scratch, 0);
+		goto bail;
+
+	case QIB_IB_CFG_MTU: /* update the MTU in IBC */
+		/*
+		 * Update our housekeeping variables, and set IBC max
+		 * size, same as init code; max IBC is max we allow in
+		 * buffer, less the qword pbc, plus 1 for ICRC, in dwords
+		 * Set even if it's unchanged, print debug message only
+		 * on changes.
+		 */
+		val = (ppd->ibmaxlen >> 2) + 1;
+		ppd->cpspec->ibcctrl &= ~SYM_MASK(IBCCtrl, MaxPktLen);
+		ppd->cpspec->ibcctrl |= (u64)val << SYM_LSB(IBCCtrl, MaxPktLen);
+		qib_write_kreg(dd, kr_ibcctrl, ppd->cpspec->ibcctrl);
+		qib_write_kreg(dd, kr_scratch, 0);
+		goto bail;
+
+	case QIB_IB_CFG_LSTATE: /* set the IB link state */
+		switch (val & 0xffff0000) {
+		case IB_LINKCMD_DOWN:
+			lcmd = QLOGIC_IB_IBCC_LINKCMD_DOWN;
+			if (!ppd->cpspec->ibdeltainprog &&
+			    qib_compat_ddr_negotiate) {
+				ppd->cpspec->ibdeltainprog = 1;
+				ppd->cpspec->ibsymsnap =
+					read_7220_creg32(dd, cr_ibsymbolerr);
+				ppd->cpspec->iblnkerrsnap =
+					read_7220_creg32(dd, cr_iblinkerrrecov);
+			}
+			break;
+
+		case IB_LINKCMD_ARMED:
+			lcmd = QLOGIC_IB_IBCC_LINKCMD_ARMED;
+			break;
+
+		case IB_LINKCMD_ACTIVE:
+			lcmd = QLOGIC_IB_IBCC_LINKCMD_ACTIVE;
+			break;
+
+		default:
+			ret = -EINVAL;
+			qib_dev_err(dd, "bad linkcmd req 0x%x\n", val >> 16);
+			goto bail;
+		}
+		switch (val & 0xffff) {
+		case IB_LINKINITCMD_NOP:
+			licmd = 0;
+			break;
+
+		case IB_LINKINITCMD_POLL:
+			licmd = QLOGIC_IB_IBCC_LINKINITCMD_POLL;
+			break;
+
+		case IB_LINKINITCMD_SLEEP:
+			licmd = QLOGIC_IB_IBCC_LINKINITCMD_SLEEP;
+			break;
+
+		case IB_LINKINITCMD_DISABLE:
+			licmd = QLOGIC_IB_IBCC_LINKINITCMD_DISABLE;
+			ppd->cpspec->chase_end = 0;
+			/*
+			 * stop state chase counter and timer, if running.
+			 * wait forpending timer, but don't clear .data (ppd)!
+			 */
+			if (ppd->cpspec->chase_timer.expires) {
+				del_timer_sync(&ppd->cpspec->chase_timer);
+				ppd->cpspec->chase_timer.expires = 0;
+			}
+			break;
+
+		default:
+			ret = -EINVAL;
+			qib_dev_err(dd, "bad linkinitcmd req 0x%x\n",
+				    val & 0xffff);
+			goto bail;
+		}
+		qib_set_ib_7220_lstate(ppd, lcmd, licmd);
+		goto bail;
+
+	case QIB_IB_CFG_HRTBT: /* set Heartbeat off/enable/auto */
+		if (val > IBA7220_IBC_HRTBT_MASK) {
+			ret = -EINVAL;
+			goto bail;
+		}
+		lsb = IBA7220_IBC_HRTBT_SHIFT;
+		maskr = IBA7220_IBC_HRTBT_MASK;
+		break;
+
+	default:
+		ret = -EINVAL;
+		goto bail;
+	}
+	ppd->cpspec->ibcddrctrl &= ~(maskr << lsb);
+	ppd->cpspec->ibcddrctrl |= (((u64) val & maskr) << lsb);
+	qib_write_kreg(dd, kr_ibcddrctrl, ppd->cpspec->ibcddrctrl);
+	qib_write_kreg(dd, kr_scratch, 0);
+	if (setforce) {
+		spin_lock_irqsave(&ppd->lflags_lock, flags);
+		ppd->lflags |= QIBL_IB_FORCE_NOTIFY;
+		spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+	}
+bail:
+	return ret;
+}
+
+static int qib_7220_set_loopback(struct qib_pportdata *ppd, const char *what)
+{
+	int ret = 0;
+	u64 val, ddr;
+
+	if (!strncmp(what, "ibc", 3)) {
+		ppd->cpspec->ibcctrl |= SYM_MASK(IBCCtrl, Loopback);
+		val = 0; /* disable heart beat, so link will come up */
+		qib_devinfo(ppd->dd->pcidev, "Enabling IB%u:%u IBC loopback\n",
+			 ppd->dd->unit, ppd->port);
+	} else if (!strncmp(what, "off", 3)) {
+		ppd->cpspec->ibcctrl &= ~SYM_MASK(IBCCtrl, Loopback);
+		/* enable heart beat again */
+		val = IBA7220_IBC_HRTBT_MASK << IBA7220_IBC_HRTBT_SHIFT;
+		qib_devinfo(ppd->dd->pcidev, "Disabling IB%u:%u IBC loopback "
+			    "(normal)\n", ppd->dd->unit, ppd->port);
+	} else
+		ret = -EINVAL;
+	if (!ret) {
+		qib_write_kreg(ppd->dd, kr_ibcctrl, ppd->cpspec->ibcctrl);
+		ddr = ppd->cpspec->ibcddrctrl & ~(IBA7220_IBC_HRTBT_MASK
+					     << IBA7220_IBC_HRTBT_SHIFT);
+		ppd->cpspec->ibcddrctrl = ddr | val;
+		qib_write_kreg(ppd->dd, kr_ibcddrctrl,
+			       ppd->cpspec->ibcddrctrl);
+		qib_write_kreg(ppd->dd, kr_scratch, 0);
+	}
+	return ret;
+}
+
+static void qib_update_7220_usrhead(struct qib_ctxtdata *rcd, u64 hd,
+				    u32 updegr, u32 egrhd)
+{
+	qib_write_ureg(rcd->dd, ur_rcvhdrhead, hd, rcd->ctxt);
+	if (updegr)
+		qib_write_ureg(rcd->dd, ur_rcvegrindexhead, egrhd, rcd->ctxt);
+}
+
+static u32 qib_7220_hdrqempty(struct qib_ctxtdata *rcd)
+{
+	u32 head, tail;
+
+	head = qib_read_ureg32(rcd->dd, ur_rcvhdrhead, rcd->ctxt);
+	if (rcd->rcvhdrtail_kvaddr)
+		tail = qib_get_rcvhdrtail(rcd);
+	else
+		tail = qib_read_ureg32(rcd->dd, ur_rcvhdrtail, rcd->ctxt);
+	return head == tail;
+}
+
+/*
+ * Modify the RCVCTRL register in chip-specific way. This
+ * is a function because bit positions and (future) register
+ * location is chip-specifc, but the needed operations are
+ * generic. <op> is a bit-mask because we often want to
+ * do multiple modifications.
+ */
+static void rcvctrl_7220_mod(struct qib_pportdata *ppd, unsigned int op,
+			     int ctxt)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u64 mask, val;
+	unsigned long flags;
+
+	spin_lock_irqsave(&dd->cspec->rcvmod_lock, flags);
+	if (op & QIB_RCVCTRL_TAILUPD_ENB)
+		dd->rcvctrl |= (1ULL << IBA7220_R_TAILUPD_SHIFT);
+	if (op & QIB_RCVCTRL_TAILUPD_DIS)
+		dd->rcvctrl &= ~(1ULL << IBA7220_R_TAILUPD_SHIFT);
+	if (op & QIB_RCVCTRL_PKEY_ENB)
+		dd->rcvctrl &= ~(1ULL << IBA7220_R_PKEY_DIS_SHIFT);
+	if (op & QIB_RCVCTRL_PKEY_DIS)
+		dd->rcvctrl |= (1ULL << IBA7220_R_PKEY_DIS_SHIFT);
+	if (ctxt < 0)
+		mask = (1ULL << dd->ctxtcnt) - 1;
+	else
+		mask = (1ULL << ctxt);
+	if (op & QIB_RCVCTRL_CTXT_ENB) {
+		/* always done for specific ctxt */
+		dd->rcvctrl |= (mask << SYM_LSB(RcvCtrl, PortEnable));
+		if (!(dd->flags & QIB_NODMA_RTAIL))
+			dd->rcvctrl |= 1ULL << IBA7220_R_TAILUPD_SHIFT;
+		/* Write these registers before the context is enabled. */
+		qib_write_kreg_ctxt(dd, kr_rcvhdrtailaddr, ctxt,
+			dd->rcd[ctxt]->rcvhdrqtailaddr_phys);
+		qib_write_kreg_ctxt(dd, kr_rcvhdraddr, ctxt,
+			dd->rcd[ctxt]->rcvhdrq_phys);
+		dd->rcd[ctxt]->seq_cnt = 1;
+	}
+	if (op & QIB_RCVCTRL_CTXT_DIS)
+		dd->rcvctrl &= ~(mask << SYM_LSB(RcvCtrl, PortEnable));
+	if (op & QIB_RCVCTRL_INTRAVAIL_ENB)
+		dd->rcvctrl |= (mask << IBA7220_R_INTRAVAIL_SHIFT);
+	if (op & QIB_RCVCTRL_INTRAVAIL_DIS)
+		dd->rcvctrl &= ~(mask << IBA7220_R_INTRAVAIL_SHIFT);
+	qib_write_kreg(dd, kr_rcvctrl, dd->rcvctrl);
+	if ((op & QIB_RCVCTRL_INTRAVAIL_ENB) && dd->rhdrhead_intr_off) {
+		/* arm rcv interrupt */
+		val = qib_read_ureg32(dd, ur_rcvhdrhead, ctxt) |
+			dd->rhdrhead_intr_off;
+		qib_write_ureg(dd, ur_rcvhdrhead, val, ctxt);
+	}
+	if (op & QIB_RCVCTRL_CTXT_ENB) {
+		/*
+		 * Init the context registers also; if we were
+		 * disabled, tail and head should both be zero
+		 * already from the enable, but since we don't
+		 * know, we have to do it explictly.
+		 */
+		val = qib_read_ureg32(dd, ur_rcvegrindextail, ctxt);
+		qib_write_ureg(dd, ur_rcvegrindexhead, val, ctxt);
+
+		val = qib_read_ureg32(dd, ur_rcvhdrtail, ctxt);
+		dd->rcd[ctxt]->head = val;
+		/* If kctxt, interrupt on next receive. */
+		if (ctxt < dd->first_user_ctxt)
+			val |= dd->rhdrhead_intr_off;
+		qib_write_ureg(dd, ur_rcvhdrhead, val, ctxt);
+	}
+	if (op & QIB_RCVCTRL_CTXT_DIS) {
+		if (ctxt >= 0) {
+			qib_write_kreg_ctxt(dd, kr_rcvhdrtailaddr, ctxt, 0);
+			qib_write_kreg_ctxt(dd, kr_rcvhdraddr, ctxt, 0);
+		} else {
+			unsigned i;
+
+			for (i = 0; i < dd->cfgctxts; i++) {
+				qib_write_kreg_ctxt(dd, kr_rcvhdrtailaddr,
+						    i, 0);
+				qib_write_kreg_ctxt(dd, kr_rcvhdraddr, i, 0);
+			}
+		}
+	}
+	spin_unlock_irqrestore(&dd->cspec->rcvmod_lock, flags);
+}
+
+/*
+ * Modify the SENDCTRL register in chip-specific way. This
+ * is a function there may be multiple such registers with
+ * slightly different layouts. To start, we assume the
+ * "canonical" register layout of the first chips.
+ * Chip requires no back-back sendctrl writes, so write
+ * scratch register after writing sendctrl
+ */
+static void sendctrl_7220_mod(struct qib_pportdata *ppd, u32 op)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u64 tmp_dd_sendctrl;
+	unsigned long flags;
+
+	spin_lock_irqsave(&dd->sendctrl_lock, flags);
+
+	/* First the ones that are "sticky", saved in shadow */
+	if (op & QIB_SENDCTRL_CLEAR)
+		dd->sendctrl = 0;
+	if (op & QIB_SENDCTRL_SEND_DIS)
+		dd->sendctrl &= ~SYM_MASK(SendCtrl, SPioEnable);
+	else if (op & QIB_SENDCTRL_SEND_ENB) {
+		dd->sendctrl |= SYM_MASK(SendCtrl, SPioEnable);
+		if (dd->flags & QIB_USE_SPCL_TRIG)
+			dd->sendctrl |= SYM_MASK(SendCtrl,
+						 SSpecialTriggerEn);
+	}
+	if (op & QIB_SENDCTRL_AVAIL_DIS)
+		dd->sendctrl &= ~SYM_MASK(SendCtrl, SendBufAvailUpd);
+	else if (op & QIB_SENDCTRL_AVAIL_ENB)
+		dd->sendctrl |= SYM_MASK(SendCtrl, SendBufAvailUpd);
+
+	if (op & QIB_SENDCTRL_DISARM_ALL) {
+		u32 i, last;
+
+		tmp_dd_sendctrl = dd->sendctrl;
+		/*
+		 * disarm any that are not yet launched, disabling sends
+		 * and updates until done.
+		 */
+		last = dd->piobcnt2k + dd->piobcnt4k;
+		tmp_dd_sendctrl &=
+			~(SYM_MASK(SendCtrl, SPioEnable) |
+			  SYM_MASK(SendCtrl, SendBufAvailUpd));
+		for (i = 0; i < last; i++) {
+			qib_write_kreg(dd, kr_sendctrl,
+				       tmp_dd_sendctrl |
+				       SYM_MASK(SendCtrl, Disarm) | i);
+			qib_write_kreg(dd, kr_scratch, 0);
+		}
+	}
+
+	tmp_dd_sendctrl = dd->sendctrl;
+
+	if (op & QIB_SENDCTRL_FLUSH)
+		tmp_dd_sendctrl |= SYM_MASK(SendCtrl, Abort);
+	if (op & QIB_SENDCTRL_DISARM)
+		tmp_dd_sendctrl |= SYM_MASK(SendCtrl, Disarm) |
+			((op & QIB_7220_SendCtrl_DisarmPIOBuf_RMASK) <<
+			 SYM_LSB(SendCtrl, DisarmPIOBuf));
+	if ((op & QIB_SENDCTRL_AVAIL_BLIP) &&
+	    (dd->sendctrl & SYM_MASK(SendCtrl, SendBufAvailUpd)))
+		tmp_dd_sendctrl &= ~SYM_MASK(SendCtrl, SendBufAvailUpd);
+
+	qib_write_kreg(dd, kr_sendctrl, tmp_dd_sendctrl);
+	qib_write_kreg(dd, kr_scratch, 0);
+
+	if (op & QIB_SENDCTRL_AVAIL_BLIP) {
+		qib_write_kreg(dd, kr_sendctrl, dd->sendctrl);
+		qib_write_kreg(dd, kr_scratch, 0);
+	}
+
+	spin_unlock_irqrestore(&dd->sendctrl_lock, flags);
+
+	if (op & QIB_SENDCTRL_FLUSH) {
+		u32 v;
+		/*
+		 * ensure writes have hit chip, then do a few
+		 * more reads, to allow DMA of pioavail registers
+		 * to occur, so in-memory copy is in sync with
+		 * the chip.  Not always safe to sleep.
+		 */
+		v = qib_read_kreg32(dd, kr_scratch);
+		qib_write_kreg(dd, kr_scratch, v);
+		v = qib_read_kreg32(dd, kr_scratch);
+		qib_write_kreg(dd, kr_scratch, v);
+		qib_read_kreg32(dd, kr_scratch);
+	}
+}
+
+/**
+ * qib_portcntr_7220 - read a per-port counter
+ * @dd: the qlogic_ib device
+ * @creg: the counter to snapshot
+ */
+static u64 qib_portcntr_7220(struct qib_pportdata *ppd, u32 reg)
+{
+	u64 ret = 0ULL;
+	struct qib_devdata *dd = ppd->dd;
+	u16 creg;
+	/* 0xffff for unimplemented or synthesized counters */
+	static const u16 xlator[] = {
+		[QIBPORTCNTR_PKTSEND] = cr_pktsend,
+		[QIBPORTCNTR_WORDSEND] = cr_wordsend,
+		[QIBPORTCNTR_PSXMITDATA] = cr_psxmitdatacount,
+		[QIBPORTCNTR_PSXMITPKTS] = cr_psxmitpktscount,
+		[QIBPORTCNTR_PSXMITWAIT] = cr_psxmitwaitcount,
+		[QIBPORTCNTR_SENDSTALL] = cr_sendstall,
+		[QIBPORTCNTR_PKTRCV] = cr_pktrcv,
+		[QIBPORTCNTR_PSRCVDATA] = cr_psrcvdatacount,
+		[QIBPORTCNTR_PSRCVPKTS] = cr_psrcvpktscount,
+		[QIBPORTCNTR_RCVEBP] = cr_rcvebp,
+		[QIBPORTCNTR_RCVOVFL] = cr_rcvovfl,
+		[QIBPORTCNTR_WORDRCV] = cr_wordrcv,
+		[QIBPORTCNTR_RXDROPPKT] = cr_rxdroppkt,
+		[QIBPORTCNTR_RXLOCALPHYERR] = cr_rxotherlocalphyerr,
+		[QIBPORTCNTR_RXVLERR] = cr_rxvlerr,
+		[QIBPORTCNTR_ERRICRC] = cr_erricrc,
+		[QIBPORTCNTR_ERRVCRC] = cr_errvcrc,
+		[QIBPORTCNTR_ERRLPCRC] = cr_errlpcrc,
+		[QIBPORTCNTR_BADFORMAT] = cr_badformat,
+		[QIBPORTCNTR_ERR_RLEN] = cr_err_rlen,
+		[QIBPORTCNTR_IBSYMBOLERR] = cr_ibsymbolerr,
+		[QIBPORTCNTR_INVALIDRLEN] = cr_invalidrlen,
+		[QIBPORTCNTR_UNSUPVL] = cr_txunsupvl,
+		[QIBPORTCNTR_EXCESSBUFOVFL] = cr_excessbufferovfl,
+		[QIBPORTCNTR_ERRLINK] = cr_errlink,
+		[QIBPORTCNTR_IBLINKDOWN] = cr_iblinkdown,
+		[QIBPORTCNTR_IBLINKERRRECOV] = cr_iblinkerrrecov,
+		[QIBPORTCNTR_LLI] = cr_locallinkintegrityerr,
+		[QIBPORTCNTR_PSINTERVAL] = cr_psinterval,
+		[QIBPORTCNTR_PSSTART] = cr_psstart,
+		[QIBPORTCNTR_PSSTAT] = cr_psstat,
+		[QIBPORTCNTR_VL15PKTDROP] = cr_vl15droppedpkt,
+		[QIBPORTCNTR_ERRPKEY] = cr_errpkey,
+		[QIBPORTCNTR_KHDROVFL] = 0xffff,
+	};
+
+	if (reg >= ARRAY_SIZE(xlator)) {
+		qib_devinfo(ppd->dd->pcidev,
+			 "Unimplemented portcounter %u\n", reg);
+		goto done;
+	}
+	creg = xlator[reg];
+
+	if (reg == QIBPORTCNTR_KHDROVFL) {
+		int i;
+
+		/* sum over all kernel contexts */
+		for (i = 0; i < dd->first_user_ctxt; i++)
+			ret += read_7220_creg32(dd, cr_portovfl + i);
+	}
+	if (creg == 0xffff)
+		goto done;
+
+	/*
+	 * only fast incrementing counters are 64bit; use 32 bit reads to
+	 * avoid two independent reads when on opteron
+	 */
+	if ((creg == cr_wordsend || creg == cr_wordrcv ||
+	     creg == cr_pktsend || creg == cr_pktrcv))
+		ret = read_7220_creg(dd, creg);
+	else
+		ret = read_7220_creg32(dd, creg);
+	if (creg == cr_ibsymbolerr) {
+		if (dd->pport->cpspec->ibdeltainprog)
+			ret -= ret - ppd->cpspec->ibsymsnap;
+		ret -= dd->pport->cpspec->ibsymdelta;
+	} else if (creg == cr_iblinkerrrecov) {
+		if (dd->pport->cpspec->ibdeltainprog)
+			ret -= ret - ppd->cpspec->iblnkerrsnap;
+		ret -= dd->pport->cpspec->iblnkerrdelta;
+	}
+done:
+	return ret;
+}
+
+/*
+ * Device counter names (not port-specific), one line per stat,
+ * single string.  Used by utilities like ipathstats to print the stats
+ * in a way which works for different versions of drivers, without changing
+ * the utility.  Names need to be 12 chars or less (w/o newline), for proper
+ * display by utility.
+ * Non-error counters are first.
+ * Start of "error" conters is indicated by a leading "E " on the first
+ * "error" counter, and doesn't count in label length.
+ * The EgrOvfl list needs to be last so we truncate them at the configured
+ * context count for the device.
+ * cntr7220indices contains the corresponding register indices.
+ */
+static const char cntr7220names[] =
+	"Interrupts\n"
+	"HostBusStall\n"
+	"E RxTIDFull\n"
+	"RxTIDInvalid\n"
+	"Ctxt0EgrOvfl\n"
+	"Ctxt1EgrOvfl\n"
+	"Ctxt2EgrOvfl\n"
+	"Ctxt3EgrOvfl\n"
+	"Ctxt4EgrOvfl\n"
+	"Ctxt5EgrOvfl\n"
+	"Ctxt6EgrOvfl\n"
+	"Ctxt7EgrOvfl\n"
+	"Ctxt8EgrOvfl\n"
+	"Ctxt9EgrOvfl\n"
+	"Ctx10EgrOvfl\n"
+	"Ctx11EgrOvfl\n"
+	"Ctx12EgrOvfl\n"
+	"Ctx13EgrOvfl\n"
+	"Ctx14EgrOvfl\n"
+	"Ctx15EgrOvfl\n"
+	"Ctx16EgrOvfl\n";
+
+static const size_t cntr7220indices[] = {
+	cr_lbint,
+	cr_lbflowstall,
+	cr_errtidfull,
+	cr_errtidvalid,
+	cr_portovfl + 0,
+	cr_portovfl + 1,
+	cr_portovfl + 2,
+	cr_portovfl + 3,
+	cr_portovfl + 4,
+	cr_portovfl + 5,
+	cr_portovfl + 6,
+	cr_portovfl + 7,
+	cr_portovfl + 8,
+	cr_portovfl + 9,
+	cr_portovfl + 10,
+	cr_portovfl + 11,
+	cr_portovfl + 12,
+	cr_portovfl + 13,
+	cr_portovfl + 14,
+	cr_portovfl + 15,
+	cr_portovfl + 16,
+};
+
+/*
+ * same as cntr7220names and cntr7220indices, but for port-specific counters.
+ * portcntr7220indices is somewhat complicated by some registers needing
+ * adjustments of various kinds, and those are ORed with _PORT_VIRT_FLAG
+ */
+static const char portcntr7220names[] =
+	"TxPkt\n"
+	"TxFlowPkt\n"
+	"TxWords\n"
+	"RxPkt\n"
+	"RxFlowPkt\n"
+	"RxWords\n"
+	"TxFlowStall\n"
+	"TxDmaDesc\n"  /* 7220 and 7322-only */
+	"E RxDlidFltr\n"  /* 7220 and 7322-only */
+	"IBStatusChng\n"
+	"IBLinkDown\n"
+	"IBLnkRecov\n"
+	"IBRxLinkErr\n"
+	"IBSymbolErr\n"
+	"RxLLIErr\n"
+	"RxBadFormat\n"
+	"RxBadLen\n"
+	"RxBufOvrfl\n"
+	"RxEBP\n"
+	"RxFlowCtlErr\n"
+	"RxICRCerr\n"
+	"RxLPCRCerr\n"
+	"RxVCRCerr\n"
+	"RxInvalLen\n"
+	"RxInvalPKey\n"
+	"RxPktDropped\n"
+	"TxBadLength\n"
+	"TxDropped\n"
+	"TxInvalLen\n"
+	"TxUnderrun\n"
+	"TxUnsupVL\n"
+	"RxLclPhyErr\n" /* 7220 and 7322-only */
+	"RxVL15Drop\n" /* 7220 and 7322-only */
+	"RxVlErr\n" /* 7220 and 7322-only */
+	"XcessBufOvfl\n" /* 7220 and 7322-only */
+	;
+
+#define _PORT_VIRT_FLAG 0x8000 /* "virtual", need adjustments */
+static const size_t portcntr7220indices[] = {
+	QIBPORTCNTR_PKTSEND | _PORT_VIRT_FLAG,
+	cr_pktsendflow,
+	QIBPORTCNTR_WORDSEND | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_PKTRCV | _PORT_VIRT_FLAG,
+	cr_pktrcvflowctrl,
+	QIBPORTCNTR_WORDRCV | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_SENDSTALL | _PORT_VIRT_FLAG,
+	cr_txsdmadesc,
+	cr_rxdlidfltr,
+	cr_ibstatuschange,
+	QIBPORTCNTR_IBLINKDOWN | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_IBLINKERRRECOV | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_ERRLINK | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_IBSYMBOLERR | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_LLI | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_BADFORMAT | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_ERR_RLEN | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_RCVOVFL | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_RCVEBP | _PORT_VIRT_FLAG,
+	cr_rcvflowctrl_err,
+	QIBPORTCNTR_ERRICRC | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_ERRLPCRC | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_ERRVCRC | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_INVALIDRLEN | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_ERRPKEY | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_RXDROPPKT | _PORT_VIRT_FLAG,
+	cr_invalidslen,
+	cr_senddropped,
+	cr_errslen,
+	cr_sendunderrun,
+	cr_txunsupvl,
+	QIBPORTCNTR_RXLOCALPHYERR | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_VL15PKTDROP | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_RXVLERR | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_EXCESSBUFOVFL | _PORT_VIRT_FLAG,
+};
+
+/* do all the setup to make the counter reads efficient later */
+static void init_7220_cntrnames(struct qib_devdata *dd)
+{
+	int i, j = 0;
+	char *s;
+
+	for (i = 0, s = (char *)cntr7220names; s && j <= dd->cfgctxts;
+	     i++) {
+		/* we always have at least one counter before the egrovfl */
+		if (!j && !strncmp("Ctxt0EgrOvfl", s + 1, 12))
+			j = 1;
+		s = strchr(s + 1, '\n');
+		if (s && j)
+			j++;
+	}
+	dd->cspec->ncntrs = i;
+	if (!s)
+		/* full list; size is without terminating null */
+		dd->cspec->cntrnamelen = sizeof(cntr7220names) - 1;
+	else
+		dd->cspec->cntrnamelen = 1 + s - cntr7220names;
+	dd->cspec->cntrs = kmalloc(dd->cspec->ncntrs
+		* sizeof(u64), GFP_KERNEL);
+	if (!dd->cspec->cntrs)
+		qib_dev_err(dd, "Failed allocation for counters\n");
+
+	for (i = 0, s = (char *)portcntr7220names; s; i++)
+		s = strchr(s + 1, '\n');
+	dd->cspec->nportcntrs = i - 1;
+	dd->cspec->portcntrnamelen = sizeof(portcntr7220names) - 1;
+	dd->cspec->portcntrs = kmalloc(dd->cspec->nportcntrs
+		* sizeof(u64), GFP_KERNEL);
+	if (!dd->cspec->portcntrs)
+		qib_dev_err(dd, "Failed allocation for portcounters\n");
+}
+
+static u32 qib_read_7220cntrs(struct qib_devdata *dd, loff_t pos, char **namep,
+			      u64 **cntrp)
+{
+	u32 ret;
+
+	if (!dd->cspec->cntrs) {
+		ret = 0;
+		goto done;
+	}
+
+	if (namep) {
+		*namep = (char *)cntr7220names;
+		ret = dd->cspec->cntrnamelen;
+		if (pos >= ret)
+			ret = 0; /* final read after getting everything */
+	} else {
+		u64 *cntr = dd->cspec->cntrs;
+		int i;
+
+		ret = dd->cspec->ncntrs * sizeof(u64);
+		if (!cntr || pos >= ret) {
+			/* everything read, or couldn't get memory */
+			ret = 0;
+			goto done;
+		}
+
+		*cntrp = cntr;
+		for (i = 0; i < dd->cspec->ncntrs; i++)
+			*cntr++ = read_7220_creg32(dd, cntr7220indices[i]);
+	}
+done:
+	return ret;
+}
+
+static u32 qib_read_7220portcntrs(struct qib_devdata *dd, loff_t pos, u32 port,
+				  char **namep, u64 **cntrp)
+{
+	u32 ret;
+
+	if (!dd->cspec->portcntrs) {
+		ret = 0;
+		goto done;
+	}
+	if (namep) {
+		*namep = (char *)portcntr7220names;
+		ret = dd->cspec->portcntrnamelen;
+		if (pos >= ret)
+			ret = 0; /* final read after getting everything */
+	} else {
+		u64 *cntr = dd->cspec->portcntrs;
+		struct qib_pportdata *ppd = &dd->pport[port];
+		int i;
+
+		ret = dd->cspec->nportcntrs * sizeof(u64);
+		if (!cntr || pos >= ret) {
+			/* everything read, or couldn't get memory */
+			ret = 0;
+			goto done;
+		}
+		*cntrp = cntr;
+		for (i = 0; i < dd->cspec->nportcntrs; i++) {
+			if (portcntr7220indices[i] & _PORT_VIRT_FLAG)
+				*cntr++ = qib_portcntr_7220(ppd,
+					portcntr7220indices[i] &
+					~_PORT_VIRT_FLAG);
+			else
+				*cntr++ = read_7220_creg32(dd,
+					   portcntr7220indices[i]);
+		}
+	}
+done:
+	return ret;
+}
+
+/**
+ * qib_get_7220_faststats - get word counters from chip before they overflow
+ * @opaque - contains a pointer to the qlogic_ib device qib_devdata
+ *
+ * This needs more work; in particular, decision on whether we really
+ * need traffic_wds done the way it is
+ * called from add_timer
+ */
+static void qib_get_7220_faststats(unsigned long opaque)
+{
+	struct qib_devdata *dd = (struct qib_devdata *) opaque;
+	struct qib_pportdata *ppd = dd->pport;
+	unsigned long flags;
+	u64 traffic_wds;
+
+	/*
+	 * don't access the chip while running diags, or memory diags can
+	 * fail
+	 */
+	if (!(dd->flags & QIB_INITTED) || dd->diag_client)
+		/* but re-arm the timer, for diags case; won't hurt other */
+		goto done;
+
+	/*
+	 * We now try to maintain an activity timer, based on traffic
+	 * exceeding a threshold, so we need to check the word-counts
+	 * even if they are 64-bit.
+	 */
+	traffic_wds = qib_portcntr_7220(ppd, cr_wordsend) +
+		qib_portcntr_7220(ppd, cr_wordrcv);
+	spin_lock_irqsave(&dd->eep_st_lock, flags);
+	traffic_wds -= dd->traffic_wds;
+	dd->traffic_wds += traffic_wds;
+	if (traffic_wds  >= QIB_TRAFFIC_ACTIVE_THRESHOLD)
+		atomic_add(5, &dd->active_time); /* S/B #define */
+	spin_unlock_irqrestore(&dd->eep_st_lock, flags);
+done:
+	mod_timer(&dd->stats_timer, jiffies + HZ * ACTIVITY_TIMER);
+}
+
+/*
+ * If we are using MSI, try to fallback to INTx.
+ */
+static int qib_7220_intr_fallback(struct qib_devdata *dd)
+{
+	if (!dd->msi_lo)
+		return 0;
+
+	qib_devinfo(dd->pcidev, "MSI interrupt not detected,"
+		 " trying INTx interrupts\n");
+	qib_7220_free_irq(dd);
+	qib_enable_intx(dd->pcidev);
+	/*
+	 * Some newer kernels require free_irq before disable_msi,
+	 * and irq can be changed during disable and INTx enable
+	 * and we need to therefore use the pcidev->irq value,
+	 * not our saved MSI value.
+	 */
+	dd->cspec->irq = dd->pcidev->irq;
+	qib_setup_7220_interrupt(dd);
+	return 1;
+}
+
+/*
+ * Reset the XGXS (between serdes and IBC).  Slightly less intrusive
+ * than resetting the IBC or external link state, and useful in some
+ * cases to cause some retraining.  To do this right, we reset IBC
+ * as well.
+ */
+static void qib_7220_xgxs_reset(struct qib_pportdata *ppd)
+{
+	u64 val, prev_val;
+	struct qib_devdata *dd = ppd->dd;
+
+	prev_val = qib_read_kreg64(dd, kr_xgxs_cfg);
+	val = prev_val | QLOGIC_IB_XGXS_RESET;
+	prev_val &= ~QLOGIC_IB_XGXS_RESET; /* be sure */
+	qib_write_kreg(dd, kr_control,
+		       dd->control & ~QLOGIC_IB_C_LINKENABLE);
+	qib_write_kreg(dd, kr_xgxs_cfg, val);
+	qib_read_kreg32(dd, kr_scratch);
+	qib_write_kreg(dd, kr_xgxs_cfg, prev_val);
+	qib_write_kreg(dd, kr_control, dd->control);
+}
+
+/*
+ * For this chip, we want to use the same buffer every time
+ * when we are trying to bring the link up (they are always VL15
+ * packets).  At that link state the packet should always go out immediately
+ * (or at least be discarded at the tx interface if the link is down).
+ * If it doesn't, and the buffer isn't available, that means some other
+ * sender has gotten ahead of us, and is preventing our packet from going
+ * out.  In that case, we flush all packets, and try again.  If that still
+ * fails, we fail the request, and hope things work the next time around.
+ *
+ * We don't need very complicated heuristics on whether the packet had
+ * time to go out or not, since even at SDR 1X, it goes out in very short
+ * time periods, covered by the chip reads done here and as part of the
+ * flush.
+ */
+static u32 __iomem *get_7220_link_buf(struct qib_pportdata *ppd, u32 *bnum)
+{
+	u32 __iomem *buf;
+	u32 lbuf = ppd->dd->cspec->lastbuf_for_pio;
+	int do_cleanup;
+	unsigned long flags;
+
+	/*
+	 * always blip to get avail list updated, since it's almost
+	 * always needed, and is fairly cheap.
+	 */
+	sendctrl_7220_mod(ppd->dd->pport, QIB_SENDCTRL_AVAIL_BLIP);
+	qib_read_kreg64(ppd->dd, kr_scratch); /* extra chip flush */
+	buf = qib_getsendbuf_range(ppd->dd, bnum, lbuf, lbuf);
+	if (buf)
+		goto done;
+
+	spin_lock_irqsave(&ppd->sdma_lock, flags);
+	if (ppd->sdma_state.current_state == qib_sdma_state_s20_idle &&
+	    ppd->sdma_state.current_state != qib_sdma_state_s00_hw_down) {
+		__qib_sdma_process_event(ppd, qib_sdma_event_e00_go_hw_down);
+		do_cleanup = 0;
+	} else {
+		do_cleanup = 1;
+		qib_7220_sdma_hw_clean_up(ppd);
+	}
+	spin_unlock_irqrestore(&ppd->sdma_lock, flags);
+
+	if (do_cleanup) {
+		qib_read_kreg64(ppd->dd, kr_scratch); /* extra chip flush */
+		buf = qib_getsendbuf_range(ppd->dd, bnum, lbuf, lbuf);
+	}
+done:
+	return buf;
+}
+
+/*
+ * This code for non-IBTA-compliant IB speed negotiation is only known to
+ * work for the SDR to DDR transition, and only between an HCA and a switch
+ * with recent firmware.  It is based on observed heuristics, rather than
+ * actual knowledge of the non-compliant speed negotiation.
+ * It has a number of hard-coded fields, since the hope is to rewrite this
+ * when a spec is available on how the negoation is intended to work.
+ */
+static void autoneg_7220_sendpkt(struct qib_pportdata *ppd, u32 *hdr,
+				 u32 dcnt, u32 *data)
+{
+	int i;
+	u64 pbc;
+	u32 __iomem *piobuf;
+	u32 pnum;
+	struct qib_devdata *dd = ppd->dd;
+
+	i = 0;
+	pbc = 7 + dcnt + 1; /* 7 dword header, dword data, icrc */
+	pbc |= PBC_7220_VL15_SEND;
+	while (!(piobuf = get_7220_link_buf(ppd, &pnum))) {
+		if (i++ > 5)
+			return;
+		udelay(2);
+	}
+	sendctrl_7220_mod(dd->pport, QIB_SENDCTRL_DISARM_BUF(pnum));
+	writeq(pbc, piobuf);
+	qib_flush_wc();
+	qib_pio_copy(piobuf + 2, hdr, 7);
+	qib_pio_copy(piobuf + 9, data, dcnt);
+	if (dd->flags & QIB_USE_SPCL_TRIG) {
+		u32 spcl_off = (pnum >= dd->piobcnt2k) ? 2047 : 1023;
+
+		qib_flush_wc();
+		__raw_writel(0xaebecede, piobuf + spcl_off);
+	}
+	qib_flush_wc();
+	qib_sendbuf_done(dd, pnum);
+}
+
+/*
+ * _start packet gets sent twice at start, _done gets sent twice at end
+ */
+static void autoneg_7220_send(struct qib_pportdata *ppd, int which)
+{
+	struct qib_devdata *dd = ppd->dd;
+	static u32 swapped;
+	u32 dw, i, hcnt, dcnt, *data;
+	static u32 hdr[7] = { 0xf002ffff, 0x48ffff, 0x6400abba };
+	static u32 madpayload_start[0x40] = {
+		0x1810103, 0x1, 0x0, 0x0, 0x2c90000, 0x2c9, 0x0, 0x0,
+		0xffffffff, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0,
+		0x1, 0x1388, 0x15e, 0x1, /* rest 0's */
+		};
+	static u32 madpayload_done[0x40] = {
+		0x1810103, 0x1, 0x0, 0x0, 0x2c90000, 0x2c9, 0x0, 0x0,
+		0xffffffff, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0,
+		0x40000001, 0x1388, 0x15e, /* rest 0's */
+		};
+
+	dcnt = ARRAY_SIZE(madpayload_start);
+	hcnt = ARRAY_SIZE(hdr);
+	if (!swapped) {
+		/* for maintainability, do it at runtime */
+		for (i = 0; i < hcnt; i++) {
+			dw = (__force u32) cpu_to_be32(hdr[i]);
+			hdr[i] = dw;
+		}
+		for (i = 0; i < dcnt; i++) {
+			dw = (__force u32) cpu_to_be32(madpayload_start[i]);
+			madpayload_start[i] = dw;
+			dw = (__force u32) cpu_to_be32(madpayload_done[i]);
+			madpayload_done[i] = dw;
+		}
+		swapped = 1;
+	}
+
+	data = which ? madpayload_done : madpayload_start;
+
+	autoneg_7220_sendpkt(ppd, hdr, dcnt, data);
+	qib_read_kreg64(dd, kr_scratch);
+	udelay(2);
+	autoneg_7220_sendpkt(ppd, hdr, dcnt, data);
+	qib_read_kreg64(dd, kr_scratch);
+	udelay(2);
+}
+
+/*
+ * Do the absolute minimum to cause an IB speed change, and make it
+ * ready, but don't actually trigger the change.   The caller will
+ * do that when ready (if link is in Polling training state, it will
+ * happen immediately, otherwise when link next goes down)
+ *
+ * This routine should only be used as part of the DDR autonegotation
+ * code for devices that are not compliant with IB 1.2 (or code that
+ * fixes things up for same).
+ *
+ * When link has gone down, and autoneg enabled, or autoneg has
+ * failed and we give up until next time we set both speeds, and
+ * then we want IBTA enabled as well as "use max enabled speed.
+ */
+static void set_7220_ibspeed_fast(struct qib_pportdata *ppd, u32 speed)
+{
+	ppd->cpspec->ibcddrctrl &= ~(IBA7220_IBC_SPEED_AUTONEG_MASK |
+		IBA7220_IBC_IBTA_1_2_MASK);
+
+	if (speed == (QIB_IB_SDR | QIB_IB_DDR))
+		ppd->cpspec->ibcddrctrl |= IBA7220_IBC_SPEED_AUTONEG_MASK |
+			IBA7220_IBC_IBTA_1_2_MASK;
+	else
+		ppd->cpspec->ibcddrctrl |= speed == QIB_IB_DDR ?
+			IBA7220_IBC_SPEED_DDR : IBA7220_IBC_SPEED_SDR;
+
+	qib_write_kreg(ppd->dd, kr_ibcddrctrl, ppd->cpspec->ibcddrctrl);
+	qib_write_kreg(ppd->dd, kr_scratch, 0);
+}
+
+/*
+ * This routine is only used when we are not talking to another
+ * IB 1.2-compliant device that we think can do DDR.
+ * (This includes all existing switch chips as of Oct 2007.)
+ * 1.2-compliant devices go directly to DDR prior to reaching INIT
+ */
+static void try_7220_autoneg(struct qib_pportdata *ppd)
+{
+	unsigned long flags;
+
+	/*
+	 * Required for older non-IB1.2 DDR switches.  Newer
+	 * non-IB-compliant switches don't need it, but so far,
+	 * aren't bothered by it either.  "Magic constant"
+	 */
+	qib_write_kreg(ppd->dd, kr_ncmodectrl, 0x3b9dc07);
+
+	spin_lock_irqsave(&ppd->lflags_lock, flags);
+	ppd->lflags |= QIBL_IB_AUTONEG_INPROG;
+	spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+	autoneg_7220_send(ppd, 0);
+	set_7220_ibspeed_fast(ppd, QIB_IB_DDR);
+
+	toggle_7220_rclkrls(ppd->dd);
+	/* 2 msec is minimum length of a poll cycle */
+	schedule_delayed_work(&ppd->cpspec->autoneg_work,
+			      msecs_to_jiffies(2));
+}
+
+/*
+ * Handle the empirically determined mechanism for auto-negotiation
+ * of DDR speed with switches.
+ */
+static void autoneg_7220_work(struct work_struct *work)
+{
+	struct qib_pportdata *ppd;
+	struct qib_devdata *dd;
+	u64 startms;
+	u32 i;
+	unsigned long flags;
+
+	ppd = &container_of(work, struct qib_chippport_specific,
+			    autoneg_work.work)->pportdata;
+	dd = ppd->dd;
+
+	startms = jiffies_to_msecs(jiffies);
+
+	/*
+	 * Busy wait for this first part, it should be at most a
+	 * few hundred usec, since we scheduled ourselves for 2msec.
+	 */
+	for (i = 0; i < 25; i++) {
+		if (SYM_FIELD(ppd->lastibcstat, IBCStatus, LinkTrainingState)
+		     == IB_7220_LT_STATE_POLLQUIET) {
+			qib_set_linkstate(ppd, QIB_IB_LINKDOWN_DISABLE);
+			break;
+		}
+		udelay(100);
+	}
+
+	if (!(ppd->lflags & QIBL_IB_AUTONEG_INPROG))
+		goto done; /* we got there early or told to stop */
+
+	/* we expect this to timeout */
+	if (wait_event_timeout(ppd->cpspec->autoneg_wait,
+			       !(ppd->lflags & QIBL_IB_AUTONEG_INPROG),
+			       msecs_to_jiffies(90)))
+		goto done;
+
+	toggle_7220_rclkrls(dd);
+
+	/* we expect this to timeout */
+	if (wait_event_timeout(ppd->cpspec->autoneg_wait,
+			       !(ppd->lflags & QIBL_IB_AUTONEG_INPROG),
+			       msecs_to_jiffies(1700)))
+		goto done;
+
+	set_7220_ibspeed_fast(ppd, QIB_IB_SDR);
+	toggle_7220_rclkrls(dd);
+
+	/*
+	 * Wait up to 250 msec for link to train and get to INIT at DDR;
+	 * this should terminate early.
+	 */
+	wait_event_timeout(ppd->cpspec->autoneg_wait,
+		!(ppd->lflags & QIBL_IB_AUTONEG_INPROG),
+		msecs_to_jiffies(250));
+done:
+	if (ppd->lflags & QIBL_IB_AUTONEG_INPROG) {
+		spin_lock_irqsave(&ppd->lflags_lock, flags);
+		ppd->lflags &= ~QIBL_IB_AUTONEG_INPROG;
+		if (dd->cspec->autoneg_tries == AUTONEG_TRIES) {
+			ppd->lflags |= QIBL_IB_AUTONEG_FAILED;
+			dd->cspec->autoneg_tries = 0;
+		}
+		spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+		set_7220_ibspeed_fast(ppd, ppd->link_speed_enabled);
+	}
+}
+
+static u32 qib_7220_iblink_state(u64 ibcs)
+{
+	u32 state = (u32)SYM_FIELD(ibcs, IBCStatus, LinkState);
+
+	switch (state) {
+	case IB_7220_L_STATE_INIT:
+		state = IB_PORT_INIT;
+		break;
+	case IB_7220_L_STATE_ARM:
+		state = IB_PORT_ARMED;
+		break;
+	case IB_7220_L_STATE_ACTIVE:
+		/* fall through */
+	case IB_7220_L_STATE_ACT_DEFER:
+		state = IB_PORT_ACTIVE;
+		break;
+	default: /* fall through */
+	case IB_7220_L_STATE_DOWN:
+		state = IB_PORT_DOWN;
+		break;
+	}
+	return state;
+}
+
+/* returns the IBTA port state, rather than the IBC link training state */
+static u8 qib_7220_phys_portstate(u64 ibcs)
+{
+	u8 state = (u8)SYM_FIELD(ibcs, IBCStatus, LinkTrainingState);
+	return qib_7220_physportstate[state];
+}
+
+static int qib_7220_ib_updown(struct qib_pportdata *ppd, int ibup, u64 ibcs)
+{
+	int ret = 0, symadj = 0;
+	struct qib_devdata *dd = ppd->dd;
+	unsigned long flags;
+
+	spin_lock_irqsave(&ppd->lflags_lock, flags);
+	ppd->lflags &= ~QIBL_IB_FORCE_NOTIFY;
+	spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+
+	if (!ibup) {
+		/*
+		 * When the link goes down we don't want AEQ running, so it
+		 * won't interfere with IBC training, etc., and we need
+		 * to go back to the static SerDes preset values.
+		 */
+		if (!(ppd->lflags & (QIBL_IB_AUTONEG_FAILED |
+				     QIBL_IB_AUTONEG_INPROG)))
+			set_7220_ibspeed_fast(ppd, ppd->link_speed_enabled);
+		if (!(ppd->lflags & QIBL_IB_AUTONEG_INPROG)) {
+			qib_sd7220_presets(dd);
+			qib_cancel_sends(ppd); /* initial disarm, etc. */
+			spin_lock_irqsave(&ppd->sdma_lock, flags);
+			if (__qib_sdma_running(ppd))
+				__qib_sdma_process_event(ppd,
+					qib_sdma_event_e70_go_idle);
+			spin_unlock_irqrestore(&ppd->sdma_lock, flags);
+		}
+		/* this might better in qib_sd7220_presets() */
+		set_7220_relock_poll(dd, ibup);
+	} else {
+		if (qib_compat_ddr_negotiate &&
+		    !(ppd->lflags & (QIBL_IB_AUTONEG_FAILED |
+				     QIBL_IB_AUTONEG_INPROG)) &&
+		    ppd->link_speed_active == QIB_IB_SDR &&
+		    (ppd->link_speed_enabled & (QIB_IB_DDR | QIB_IB_SDR)) ==
+		    (QIB_IB_DDR | QIB_IB_SDR) &&
+		    dd->cspec->autoneg_tries < AUTONEG_TRIES) {
+			/* we are SDR, and DDR auto-negotiation enabled */
+			++dd->cspec->autoneg_tries;
+			if (!ppd->cpspec->ibdeltainprog) {
+				ppd->cpspec->ibdeltainprog = 1;
+				ppd->cpspec->ibsymsnap = read_7220_creg32(dd,
+					cr_ibsymbolerr);
+				ppd->cpspec->iblnkerrsnap = read_7220_creg32(dd,
+					cr_iblinkerrrecov);
+			}
+			try_7220_autoneg(ppd);
+			ret = 1; /* no other IB status change processing */
+		} else if ((ppd->lflags & QIBL_IB_AUTONEG_INPROG) &&
+			   ppd->link_speed_active == QIB_IB_SDR) {
+			autoneg_7220_send(ppd, 1);
+			set_7220_ibspeed_fast(ppd, QIB_IB_DDR);
+			udelay(2);
+			toggle_7220_rclkrls(dd);
+			ret = 1; /* no other IB status change processing */
+		} else {
+			if ((ppd->lflags & QIBL_IB_AUTONEG_INPROG) &&
+			    (ppd->link_speed_active & QIB_IB_DDR)) {
+				spin_lock_irqsave(&ppd->lflags_lock, flags);
+				ppd->lflags &= ~(QIBL_IB_AUTONEG_INPROG |
+						 QIBL_IB_AUTONEG_FAILED);
+				spin_unlock_irqrestore(&ppd->lflags_lock,
+						       flags);
+				dd->cspec->autoneg_tries = 0;
+				/* re-enable SDR, for next link down */
+				set_7220_ibspeed_fast(ppd,
+						      ppd->link_speed_enabled);
+				wake_up(&ppd->cpspec->autoneg_wait);
+				symadj = 1;
+			} else if (ppd->lflags & QIBL_IB_AUTONEG_FAILED) {
+				/*
+				 * Clear autoneg failure flag, and do setup
+				 * so we'll try next time link goes down and
+				 * back to INIT (possibly connected to a
+				 * different device).
+				 */
+				spin_lock_irqsave(&ppd->lflags_lock, flags);
+				ppd->lflags &= ~QIBL_IB_AUTONEG_FAILED;
+				spin_unlock_irqrestore(&ppd->lflags_lock,
+						       flags);
+				ppd->cpspec->ibcddrctrl |=
+					IBA7220_IBC_IBTA_1_2_MASK;
+				qib_write_kreg(dd, kr_ncmodectrl, 0);
+				symadj = 1;
+			}
+		}
+
+		if (!(ppd->lflags & QIBL_IB_AUTONEG_INPROG))
+			symadj = 1;
+
+		if (!ret) {
+			ppd->delay_mult = rate_to_delay
+			    [(ibcs >> IBA7220_LINKSPEED_SHIFT) & 1]
+			    [(ibcs >> IBA7220_LINKWIDTH_SHIFT) & 1];
+
+			set_7220_relock_poll(dd, ibup);
+			spin_lock_irqsave(&ppd->sdma_lock, flags);
+			/*
+			 * Unlike 7322, the 7220 needs this, due to lack of
+			 * interrupt in some cases when we have sdma active
+			 * when the link goes down.
+			 */
+			if (ppd->sdma_state.current_state !=
+			    qib_sdma_state_s20_idle)
+				__qib_sdma_process_event(ppd,
+					qib_sdma_event_e00_go_hw_down);
+			spin_unlock_irqrestore(&ppd->sdma_lock, flags);
+		}
+	}
+
+	if (symadj) {
+		if (ppd->cpspec->ibdeltainprog) {
+			ppd->cpspec->ibdeltainprog = 0;
+			ppd->cpspec->ibsymdelta += read_7220_creg32(ppd->dd,
+				cr_ibsymbolerr) - ppd->cpspec->ibsymsnap;
+			ppd->cpspec->iblnkerrdelta += read_7220_creg32(ppd->dd,
+				cr_iblinkerrrecov) - ppd->cpspec->iblnkerrsnap;
+		}
+	} else if (!ibup && qib_compat_ddr_negotiate &&
+		   !ppd->cpspec->ibdeltainprog &&
+			!(ppd->lflags & QIBL_IB_AUTONEG_INPROG)) {
+		ppd->cpspec->ibdeltainprog = 1;
+		ppd->cpspec->ibsymsnap = read_7220_creg32(ppd->dd,
+							  cr_ibsymbolerr);
+		ppd->cpspec->iblnkerrsnap = read_7220_creg32(ppd->dd,
+						     cr_iblinkerrrecov);
+	}
+
+	if (!ret)
+		qib_setup_7220_setextled(ppd, ibup);
+	return ret;
+}
+
+/*
+ * Does read/modify/write to appropriate registers to
+ * set output and direction bits selected by mask.
+ * these are in their canonical postions (e.g. lsb of
+ * dir will end up in D48 of extctrl on existing chips).
+ * returns contents of GP Inputs.
+ */
+static int gpio_7220_mod(struct qib_devdata *dd, u32 out, u32 dir, u32 mask)
+{
+	u64 read_val, new_out;
+	unsigned long flags;
+
+	if (mask) {
+		/* some bits being written, lock access to GPIO */
+		dir &= mask;
+		out &= mask;
+		spin_lock_irqsave(&dd->cspec->gpio_lock, flags);
+		dd->cspec->extctrl &= ~((u64)mask << SYM_LSB(EXTCtrl, GPIOOe));
+		dd->cspec->extctrl |= ((u64) dir << SYM_LSB(EXTCtrl, GPIOOe));
+		new_out = (dd->cspec->gpio_out & ~mask) | out;
+
+		qib_write_kreg(dd, kr_extctrl, dd->cspec->extctrl);
+		qib_write_kreg(dd, kr_gpio_out, new_out);
+		dd->cspec->gpio_out = new_out;
+		spin_unlock_irqrestore(&dd->cspec->gpio_lock, flags);
+	}
+	/*
+	 * It is unlikely that a read at this time would get valid
+	 * data on a pin whose direction line was set in the same
+	 * call to this function. We include the read here because
+	 * that allows us to potentially combine a change on one pin with
+	 * a read on another, and because the old code did something like
+	 * this.
+	 */
+	read_val = qib_read_kreg64(dd, kr_extstatus);
+	return SYM_FIELD(read_val, EXTStatus, GPIOIn);
+}
+
+/*
+ * Read fundamental info we need to use the chip.  These are
+ * the registers that describe chip capabilities, and are
+ * saved in shadow registers.
+ */
+static void get_7220_chip_params(struct qib_devdata *dd)
+{
+	u64 val;
+	u32 piobufs;
+	int mtu;
+
+	dd->uregbase = qib_read_kreg32(dd, kr_userregbase);
+
+	dd->rcvtidcnt = qib_read_kreg32(dd, kr_rcvtidcnt);
+	dd->rcvtidbase = qib_read_kreg32(dd, kr_rcvtidbase);
+	dd->rcvegrbase = qib_read_kreg32(dd, kr_rcvegrbase);
+	dd->palign = qib_read_kreg32(dd, kr_palign);
+	dd->piobufbase = qib_read_kreg64(dd, kr_sendpiobufbase);
+	dd->pio2k_bufbase = dd->piobufbase & 0xffffffff;
+
+	val = qib_read_kreg64(dd, kr_sendpiosize);
+	dd->piosize2k = val & ~0U;
+	dd->piosize4k = val >> 32;
+
+	mtu = ib_mtu_enum_to_int(qib_ibmtu);
+	if (mtu == -1)
+		mtu = QIB_DEFAULT_MTU;
+	dd->pport->ibmtu = (u32)mtu;
+
+	val = qib_read_kreg64(dd, kr_sendpiobufcnt);
+	dd->piobcnt2k = val & ~0U;
+	dd->piobcnt4k = val >> 32;
+	/* these may be adjusted in init_chip_wc_pat() */
+	dd->pio2kbase = (u32 __iomem *)
+		((char __iomem *) dd->kregbase + dd->pio2k_bufbase);
+	if (dd->piobcnt4k) {
+		dd->pio4kbase = (u32 __iomem *)
+			((char __iomem *) dd->kregbase +
+			 (dd->piobufbase >> 32));
+		/*
+		 * 4K buffers take 2 pages; we use roundup just to be
+		 * paranoid; we calculate it once here, rather than on
+		 * ever buf allocate
+		 */
+		dd->align4k = ALIGN(dd->piosize4k, dd->palign);
+	}
+
+	piobufs = dd->piobcnt4k + dd->piobcnt2k;
+
+	dd->pioavregs = ALIGN(piobufs, sizeof(u64) * BITS_PER_BYTE / 2) /
+		(sizeof(u64) * BITS_PER_BYTE / 2);
+}
+
+/*
+ * The chip base addresses in cspec and cpspec have to be set
+ * after possible init_chip_wc_pat(), rather than in
+ * qib_get_7220_chip_params(), so split out as separate function
+ */
+static void set_7220_baseaddrs(struct qib_devdata *dd)
+{
+	u32 cregbase;
+	/* init after possible re-map in init_chip_wc_pat() */
+	cregbase = qib_read_kreg32(dd, kr_counterregbase);
+	dd->cspec->cregbase = (u64 __iomem *)
+		((char __iomem *) dd->kregbase + cregbase);
+
+	dd->egrtidbase = (u64 __iomem *)
+		((char __iomem *) dd->kregbase + dd->rcvegrbase);
+}
+
+
+#define SENDCTRL_SHADOWED (SYM_MASK(SendCtrl, SendIntBufAvail) |	\
+			   SYM_MASK(SendCtrl, SPioEnable) |		\
+			   SYM_MASK(SendCtrl, SSpecialTriggerEn) |	\
+			   SYM_MASK(SendCtrl, SendBufAvailUpd) |	\
+			   SYM_MASK(SendCtrl, AvailUpdThld) |		\
+			   SYM_MASK(SendCtrl, SDmaEnable) |		\
+			   SYM_MASK(SendCtrl, SDmaIntEnable) |		\
+			   SYM_MASK(SendCtrl, SDmaHalt) |		\
+			   SYM_MASK(SendCtrl, SDmaSingleDescriptor))
+
+static int sendctrl_hook(struct qib_devdata *dd,
+			 const struct diag_observer *op,
+			 u32 offs, u64 *data, u64 mask, int only_32)
+{
+	unsigned long flags;
+	unsigned idx = offs / sizeof(u64);
+	u64 local_data, all_bits;
+
+	if (idx != kr_sendctrl) {
+		qib_dev_err(dd, "SendCtrl Hook called with offs %X, %s-bit\n",
+			    offs, only_32 ? "32" : "64");
+		return 0;
+	}
+
+	all_bits = ~0ULL;
+	if (only_32)
+		all_bits >>= 32;
+	spin_lock_irqsave(&dd->sendctrl_lock, flags);
+	if ((mask & all_bits) != all_bits) {
+		/*
+		 * At least some mask bits are zero, so we need
+		 * to read. The judgement call is whether from
+		 * reg or shadow. First-cut: read reg, and complain
+		 * if any bits which should be shadowed are different
+		 * from their shadowed value.
+		 */
+		if (only_32)
+			local_data = (u64)qib_read_kreg32(dd, idx);
+		else
+			local_data = qib_read_kreg64(dd, idx);
+		qib_dev_err(dd, "Sendctrl -> %X, Shad -> %X\n",
+			    (u32)local_data, (u32)dd->sendctrl);
+		if ((local_data & SENDCTRL_SHADOWED) !=
+		    (dd->sendctrl & SENDCTRL_SHADOWED))
+			qib_dev_err(dd, "Sendctrl read: %X shadow is %X\n",
+				(u32)local_data, (u32) dd->sendctrl);
+		*data = (local_data & ~mask) | (*data & mask);
+	}
+	if (mask) {
+		/*
+		 * At least some mask bits are one, so we need
+		 * to write, but only shadow some bits.
+		 */
+		u64 sval, tval; /* Shadowed, transient */
+
+		/*
+		 * New shadow val is bits we don't want to touch,
+		 * ORed with bits we do, that are intended for shadow.
+		 */
+		sval = (dd->sendctrl & ~mask);
+		sval |= *data & SENDCTRL_SHADOWED & mask;
+		dd->sendctrl = sval;
+		tval = sval | (*data & ~SENDCTRL_SHADOWED & mask);
+		qib_dev_err(dd, "Sendctrl <- %X, Shad <- %X\n",
+			    (u32)tval, (u32)sval);
+		qib_write_kreg(dd, kr_sendctrl, tval);
+		qib_write_kreg(dd, kr_scratch, 0Ull);
+	}
+	spin_unlock_irqrestore(&dd->sendctrl_lock, flags);
+
+	return only_32 ? 4 : 8;
+}
+
+static const struct diag_observer sendctrl_observer = {
+	sendctrl_hook, kr_sendctrl * sizeof(u64),
+	kr_sendctrl * sizeof(u64)
+};
+
+/*
+ * write the final few registers that depend on some of the
+ * init setup.  Done late in init, just before bringing up
+ * the serdes.
+ */
+static int qib_late_7220_initreg(struct qib_devdata *dd)
+{
+	int ret = 0;
+	u64 val;
+
+	qib_write_kreg(dd, kr_rcvhdrentsize, dd->rcvhdrentsize);
+	qib_write_kreg(dd, kr_rcvhdrsize, dd->rcvhdrsize);
+	qib_write_kreg(dd, kr_rcvhdrcnt, dd->rcvhdrcnt);
+	qib_write_kreg(dd, kr_sendpioavailaddr, dd->pioavailregs_phys);
+	val = qib_read_kreg64(dd, kr_sendpioavailaddr);
+	if (val != dd->pioavailregs_phys) {
+		qib_dev_err(dd, "Catastrophic software error, "
+			    "SendPIOAvailAddr written as %lx, "
+			    "read back as %llx\n",
+			    (unsigned long) dd->pioavailregs_phys,
+			    (unsigned long long) val);
+		ret = -EINVAL;
+	}
+	qib_register_observer(dd, &sendctrl_observer);
+	return ret;
+}
+
+static int qib_init_7220_variables(struct qib_devdata *dd)
+{
+	struct qib_chippport_specific *cpspec;
+	struct qib_pportdata *ppd;
+	int ret = 0;
+	u32 sbufs, updthresh;
+
+	cpspec = (struct qib_chippport_specific *)(dd + 1);
+	ppd = &cpspec->pportdata;
+	dd->pport = ppd;
+	dd->num_pports = 1;
+
+	dd->cspec = (struct qib_chip_specific *)(cpspec + dd->num_pports);
+	ppd->cpspec = cpspec;
+
+	spin_lock_init(&dd->cspec->sdepb_lock);
+	spin_lock_init(&dd->cspec->rcvmod_lock);
+	spin_lock_init(&dd->cspec->gpio_lock);
+
+	/* we haven't yet set QIB_PRESENT, so use read directly */
+	dd->revision = readq(&dd->kregbase[kr_revision]);
+
+	if ((dd->revision & 0xffffffffU) == 0xffffffffU) {
+		qib_dev_err(dd, "Revision register read failure, "
+			    "giving up initialization\n");
+		ret = -ENODEV;
+		goto bail;
+	}
+	dd->flags |= QIB_PRESENT;  /* now register routines work */
+
+	dd->majrev = (u8) SYM_FIELD(dd->revision, Revision_R,
+				    ChipRevMajor);
+	dd->minrev = (u8) SYM_FIELD(dd->revision, Revision_R,
+				    ChipRevMinor);
+
+	get_7220_chip_params(dd);
+	qib_7220_boardname(dd);
+
+	/*
+	 * GPIO bits for TWSI data and clock,
+	 * used for serial EEPROM.
+	 */
+	dd->gpio_sda_num = _QIB_GPIO_SDA_NUM;
+	dd->gpio_scl_num = _QIB_GPIO_SCL_NUM;
+	dd->twsi_eeprom_dev = QIB_TWSI_EEPROM_DEV;
+
+	dd->flags |= QIB_HAS_INTX | QIB_HAS_LINK_LATENCY |
+		QIB_NODMA_RTAIL | QIB_HAS_THRESH_UPDATE;
+	dd->flags |= qib_special_trigger ?
+		QIB_USE_SPCL_TRIG : QIB_HAS_SEND_DMA;
+
+	/*
+	 * EEPROM error log 0 is TXE Parity errors. 1 is RXE Parity.
+	 * 2 is Some Misc, 3 is reserved for future.
+	 */
+	dd->eep_st_masks[0].hwerrs_to_log = HWE_MASK(TXEMemParityErr);
+
+	dd->eep_st_masks[1].hwerrs_to_log = HWE_MASK(RXEMemParityErr);
+
+	dd->eep_st_masks[2].errs_to_log = ERR_MASK(ResetNegated);
+
+	init_waitqueue_head(&cpspec->autoneg_wait);
+	INIT_DELAYED_WORK(&cpspec->autoneg_work, autoneg_7220_work);
+
+	qib_init_pportdata(ppd, dd, 0, 1);
+	ppd->link_width_supported = IB_WIDTH_1X | IB_WIDTH_4X;
+	ppd->link_speed_supported = QIB_IB_SDR | QIB_IB_DDR;
+
+	ppd->link_width_enabled = ppd->link_width_supported;
+	ppd->link_speed_enabled = ppd->link_speed_supported;
+	/*
+	 * Set the initial values to reasonable default, will be set
+	 * for real when link is up.
+	 */
+	ppd->link_width_active = IB_WIDTH_4X;
+	ppd->link_speed_active = QIB_IB_SDR;
+	ppd->delay_mult = rate_to_delay[0][1];
+	ppd->vls_supported = IB_VL_VL0;
+	ppd->vls_operational = ppd->vls_supported;
+
+	if (!qib_mini_init)
+		qib_write_kreg(dd, kr_rcvbthqp, QIB_KD_QP);
+
+	init_timer(&ppd->cpspec->chase_timer);
+	ppd->cpspec->chase_timer.function = reenable_7220_chase;
+	ppd->cpspec->chase_timer.data = (unsigned long)ppd;
+
+	qib_num_cfg_vls = 1; /* if any 7220's, only one VL */
+
+	dd->rcvhdrentsize = QIB_RCVHDR_ENTSIZE;
+	dd->rcvhdrsize = QIB_DFLT_RCVHDRSIZE;
+	dd->rhf_offset =
+		dd->rcvhdrentsize - sizeof(u64) / sizeof(u32);
+
+	/* we always allocate at least 2048 bytes for eager buffers */
+	ret = ib_mtu_enum_to_int(qib_ibmtu);
+	dd->rcvegrbufsize = ret != -1 ? max(ret, 2048) : QIB_DEFAULT_MTU;
+
+	qib_7220_tidtemplate(dd);
+
+	/*
+	 * We can request a receive interrupt for 1 or
+	 * more packets from current offset.  For now, we set this
+	 * up for a single packet.
+	 */
+	dd->rhdrhead_intr_off = 1ULL << 32;
+
+	/* setup the stats timer; the add_timer is done at end of init */
+	init_timer(&dd->stats_timer);
+	dd->stats_timer.function = qib_get_7220_faststats;
+	dd->stats_timer.data = (unsigned long) dd;
+	dd->stats_timer.expires = jiffies + ACTIVITY_TIMER * HZ;
+
+	/*
+	 * Control[4] has been added to change the arbitration within
+	 * the SDMA engine between favoring data fetches over descriptor
+	 * fetches.  qib_sdma_fetch_arb==0 gives data fetches priority.
+	 */
+	if (qib_sdma_fetch_arb)
+		dd->control |= 1 << 4;
+
+	dd->ureg_align = 0x10000;  /* 64KB alignment */
+
+	dd->piosize2kmax_dwords = (dd->piosize2k >> 2)-1;
+	qib_7220_config_ctxts(dd);
+	qib_set_ctxtcnt(dd);  /* needed for PAT setup */
+
+	if (qib_wc_pat) {
+		ret = init_chip_wc_pat(dd, 0);
+		if (ret)
+			goto bail;
+	}
+	set_7220_baseaddrs(dd); /* set chip access pointers now */
+
+	ret = 0;
+	if (qib_mini_init)
+		goto bail;
+
+	ret = qib_create_ctxts(dd);
+	init_7220_cntrnames(dd);
+
+	/* use all of 4KB buffers for the kernel SDMA, zero if !SDMA.
+	 * reserve the update threshold amount for other kernel use, such
+	 * as sending SMI, MAD, and ACKs, or 3, whichever is greater,
+	 * unless we aren't enabling SDMA, in which case we want to use
+	 * all the 4k bufs for the kernel.
+	 * if this was less than the update threshold, we could wait
+	 * a long time for an update.  Coded this way because we
+	 * sometimes change the update threshold for various reasons,
+	 * and we want this to remain robust.
+	 */
+	updthresh = 8U; /* update threshold */
+	if (dd->flags & QIB_HAS_SEND_DMA) {
+		dd->cspec->sdmabufcnt =  dd->piobcnt4k;
+		sbufs = updthresh > 3 ? updthresh : 3;
+	} else {
+		dd->cspec->sdmabufcnt = 0;
+		sbufs = dd->piobcnt4k;
+	}
+
+	dd->cspec->lastbuf_for_pio = dd->piobcnt2k + dd->piobcnt4k -
+		dd->cspec->sdmabufcnt;
+	dd->lastctxt_piobuf = dd->cspec->lastbuf_for_pio - sbufs;
+	dd->cspec->lastbuf_for_pio--; /* range is <= , not < */
+	dd->pbufsctxt = dd->lastctxt_piobuf /
+		(dd->cfgctxts - dd->first_user_ctxt);
+
+	/*
+	 * if we are at 16 user contexts, we will have one 7 sbufs
+	 * per context, so drop the update threshold to match.  We
+	 * want to update before we actually run out, at low pbufs/ctxt
+	 * so give ourselves some margin
+	 */
+	if ((dd->pbufsctxt - 2) < updthresh)
+		updthresh = dd->pbufsctxt - 2;
+
+	dd->cspec->updthresh_dflt = updthresh;
+	dd->cspec->updthresh = updthresh;
+
+	/* before full enable, no interrupts, no locking needed */
+	dd->sendctrl |= (updthresh & SYM_RMASK(SendCtrl, AvailUpdThld))
+			     << SYM_LSB(SendCtrl, AvailUpdThld);
+
+	dd->psxmitwait_supported = 1;
+	dd->psxmitwait_check_rate = QIB_7220_PSXMITWAIT_CHECK_RATE;
+bail:
+	return ret;
+}
+
+static u32 __iomem *qib_7220_getsendbuf(struct qib_pportdata *ppd, u64 pbc,
+					u32 *pbufnum)
+{
+	u32 first, last, plen = pbc & QIB_PBC_LENGTH_MASK;
+	struct qib_devdata *dd = ppd->dd;
+	u32 __iomem *buf;
+
+	if (((pbc >> 32) & PBC_7220_VL15_SEND_CTRL) &&
+		!(ppd->lflags & (QIBL_IB_AUTONEG_INPROG | QIBL_LINKACTIVE)))
+		buf = get_7220_link_buf(ppd, pbufnum);
+	else {
+		if ((plen + 1) > dd->piosize2kmax_dwords)
+			first = dd->piobcnt2k;
+		else
+			first = 0;
+		/* try 4k if all 2k busy, so same last for both sizes */
+		last = dd->cspec->lastbuf_for_pio;
+		buf = qib_getsendbuf_range(dd, pbufnum, first, last);
+	}
+	return buf;
+}
+
+/* these 2 "counters" are really control registers, and are always RW */
+static void qib_set_cntr_7220_sample(struct qib_pportdata *ppd, u32 intv,
+				     u32 start)
+{
+	write_7220_creg(ppd->dd, cr_psinterval, intv);
+	write_7220_creg(ppd->dd, cr_psstart, start);
+}
+
+/*
+ * NOTE: no real attempt is made to generalize the SDMA stuff.
+ * At some point "soon" we will have a new more generalized
+ * set of sdma interface, and then we'll clean this up.
+ */
+
+/* Must be called with sdma_lock held, or before init finished */
+static void qib_sdma_update_7220_tail(struct qib_pportdata *ppd, u16 tail)
+{
+	/* Commit writes to memory and advance the tail on the chip */
+	wmb();
+	ppd->sdma_descq_tail = tail;
+	qib_write_kreg(ppd->dd, kr_senddmatail, tail);
+}
+
+static void qib_sdma_set_7220_desc_cnt(struct qib_pportdata *ppd, unsigned cnt)
+{
+}
+
+static struct sdma_set_state_action sdma_7220_action_table[] = {
+	[qib_sdma_state_s00_hw_down] = {
+		.op_enable = 0,
+		.op_intenable = 0,
+		.op_halt = 0,
+		.go_s99_running_tofalse = 1,
+	},
+	[qib_sdma_state_s10_hw_start_up_wait] = {
+		.op_enable = 1,
+		.op_intenable = 1,
+		.op_halt = 1,
+	},
+	[qib_sdma_state_s20_idle] = {
+		.op_enable = 1,
+		.op_intenable = 1,
+		.op_halt = 1,
+	},
+	[qib_sdma_state_s30_sw_clean_up_wait] = {
+		.op_enable = 0,
+		.op_intenable = 1,
+		.op_halt = 0,
+	},
+	[qib_sdma_state_s40_hw_clean_up_wait] = {
+		.op_enable = 1,
+		.op_intenable = 1,
+		.op_halt = 1,
+	},
+	[qib_sdma_state_s50_hw_halt_wait] = {
+		.op_enable = 1,
+		.op_intenable = 1,
+		.op_halt = 1,
+	},
+	[qib_sdma_state_s99_running] = {
+		.op_enable = 1,
+		.op_intenable = 1,
+		.op_halt = 0,
+		.go_s99_running_totrue = 1,
+	},
+};
+
+static void qib_7220_sdma_init_early(struct qib_pportdata *ppd)
+{
+	ppd->sdma_state.set_state_action = sdma_7220_action_table;
+}
+
+static int init_sdma_7220_regs(struct qib_pportdata *ppd)
+{
+	struct qib_devdata *dd = ppd->dd;
+	unsigned i, n;
+	u64 senddmabufmask[3] = { 0 };
+
+	/* Set SendDmaBase */
+	qib_write_kreg(dd, kr_senddmabase, ppd->sdma_descq_phys);
+	qib_sdma_7220_setlengen(ppd);
+	qib_sdma_update_7220_tail(ppd, 0); /* Set SendDmaTail */
+	/* Set SendDmaHeadAddr */
+	qib_write_kreg(dd, kr_senddmaheadaddr, ppd->sdma_head_phys);
+
+	/*
+	 * Reserve all the former "kernel" piobufs, using high number range
+	 * so we get as many 4K buffers as possible
+	 */
+	n = dd->piobcnt2k + dd->piobcnt4k;
+	i = n - dd->cspec->sdmabufcnt;
+
+	for (; i < n; ++i) {
+		unsigned word = i / 64;
+		unsigned bit = i & 63;
+
+		BUG_ON(word >= 3);
+		senddmabufmask[word] |= 1ULL << bit;
+	}
+	qib_write_kreg(dd, kr_senddmabufmask0, senddmabufmask[0]);
+	qib_write_kreg(dd, kr_senddmabufmask1, senddmabufmask[1]);
+	qib_write_kreg(dd, kr_senddmabufmask2, senddmabufmask[2]);
+
+	ppd->sdma_state.first_sendbuf = i;
+	ppd->sdma_state.last_sendbuf = n;
+
+	return 0;
+}
+
+/* sdma_lock must be held */
+static u16 qib_sdma_7220_gethead(struct qib_pportdata *ppd)
+{
+	struct qib_devdata *dd = ppd->dd;
+	int sane;
+	int use_dmahead;
+	u16 swhead;
+	u16 swtail;
+	u16 cnt;
+	u16 hwhead;
+
+	use_dmahead = __qib_sdma_running(ppd) &&
+		(dd->flags & QIB_HAS_SDMA_TIMEOUT);
+retry:
+	hwhead = use_dmahead ?
+		(u16)le64_to_cpu(*ppd->sdma_head_dma) :
+		(u16)qib_read_kreg32(dd, kr_senddmahead);
+
+	swhead = ppd->sdma_descq_head;
+	swtail = ppd->sdma_descq_tail;
+	cnt = ppd->sdma_descq_cnt;
+
+	if (swhead < swtail) {
+		/* not wrapped */
+		sane = (hwhead >= swhead) & (hwhead <= swtail);
+	} else if (swhead > swtail) {
+		/* wrapped around */
+		sane = ((hwhead >= swhead) && (hwhead < cnt)) ||
+			(hwhead <= swtail);
+	} else {
+		/* empty */
+		sane = (hwhead == swhead);
+	}
+
+	if (unlikely(!sane)) {
+		if (use_dmahead) {
+			/* try one more time, directly from the register */
+			use_dmahead = 0;
+			goto retry;
+		}
+		/* assume no progress */
+		hwhead = swhead;
+	}
+
+	return hwhead;
+}
+
+static int qib_sdma_7220_busy(struct qib_pportdata *ppd)
+{
+	u64 hwstatus = qib_read_kreg64(ppd->dd, kr_senddmastatus);
+
+	return (hwstatus & SYM_MASK(SendDmaStatus, ScoreBoardDrainInProg)) ||
+	       (hwstatus & SYM_MASK(SendDmaStatus, AbortInProg)) ||
+	       (hwstatus & SYM_MASK(SendDmaStatus, InternalSDmaEnable)) ||
+	       !(hwstatus & SYM_MASK(SendDmaStatus, ScbEmpty));
+}
+
+/*
+ * Compute the amount of delay before sending the next packet if the
+ * port's send rate differs from the static rate set for the QP.
+ * Since the delay affects this packet but the amount of the delay is
+ * based on the length of the previous packet, use the last delay computed
+ * and save the delay count for this packet to be used next time
+ * we get here.
+ */
+static u32 qib_7220_setpbc_control(struct qib_pportdata *ppd, u32 plen,
+				   u8 srate, u8 vl)
+{
+	u8 snd_mult = ppd->delay_mult;
+	u8 rcv_mult = ib_rate_to_delay[srate];
+	u32 ret = ppd->cpspec->last_delay_mult;
+
+	ppd->cpspec->last_delay_mult = (rcv_mult > snd_mult) ?
+		(plen * (rcv_mult - snd_mult) + 1) >> 1 : 0;
+
+	/* Indicate VL15, if necessary */
+	if (vl == 15)
+		ret |= PBC_7220_VL15_SEND_CTRL;
+	return ret;
+}
+
+static void qib_7220_initvl15_bufs(struct qib_devdata *dd)
+{
+}
+
+static void qib_7220_init_ctxt(struct qib_ctxtdata *rcd)
+{
+	if (!rcd->ctxt) {
+		rcd->rcvegrcnt = IBA7220_KRCVEGRCNT;
+		rcd->rcvegr_tid_base = 0;
+	} else {
+		rcd->rcvegrcnt = rcd->dd->cspec->rcvegrcnt;
+		rcd->rcvegr_tid_base = IBA7220_KRCVEGRCNT +
+			(rcd->ctxt - 1) * rcd->rcvegrcnt;
+	}
+}
+
+static void qib_7220_txchk_change(struct qib_devdata *dd, u32 start,
+				  u32 len, u32 which, struct qib_ctxtdata *rcd)
+{
+	int i;
+	unsigned long flags;
+
+	switch (which) {
+	case TXCHK_CHG_TYPE_KERN:
+		/* see if we need to raise avail update threshold */
+		spin_lock_irqsave(&dd->uctxt_lock, flags);
+		for (i = dd->first_user_ctxt;
+		     dd->cspec->updthresh != dd->cspec->updthresh_dflt
+		     && i < dd->cfgctxts; i++)
+			if (dd->rcd[i] && dd->rcd[i]->subctxt_cnt &&
+			   ((dd->rcd[i]->piocnt / dd->rcd[i]->subctxt_cnt) - 1)
+			   < dd->cspec->updthresh_dflt)
+				break;
+		spin_unlock_irqrestore(&dd->uctxt_lock, flags);
+		if (i == dd->cfgctxts) {
+			spin_lock_irqsave(&dd->sendctrl_lock, flags);
+			dd->cspec->updthresh = dd->cspec->updthresh_dflt;
+			dd->sendctrl &= ~SYM_MASK(SendCtrl, AvailUpdThld);
+			dd->sendctrl |= (dd->cspec->updthresh &
+					 SYM_RMASK(SendCtrl, AvailUpdThld)) <<
+					   SYM_LSB(SendCtrl, AvailUpdThld);
+			spin_unlock_irqrestore(&dd->sendctrl_lock, flags);
+			sendctrl_7220_mod(dd->pport, QIB_SENDCTRL_AVAIL_BLIP);
+		}
+		break;
+	case TXCHK_CHG_TYPE_USER:
+		spin_lock_irqsave(&dd->sendctrl_lock, flags);
+		if (rcd && rcd->subctxt_cnt && ((rcd->piocnt
+			/ rcd->subctxt_cnt) - 1) < dd->cspec->updthresh) {
+			dd->cspec->updthresh = (rcd->piocnt /
+						rcd->subctxt_cnt) - 1;
+			dd->sendctrl &= ~SYM_MASK(SendCtrl, AvailUpdThld);
+			dd->sendctrl |= (dd->cspec->updthresh &
+					SYM_RMASK(SendCtrl, AvailUpdThld))
+					<< SYM_LSB(SendCtrl, AvailUpdThld);
+			spin_unlock_irqrestore(&dd->sendctrl_lock, flags);
+			sendctrl_7220_mod(dd->pport, QIB_SENDCTRL_AVAIL_BLIP);
+		} else
+			spin_unlock_irqrestore(&dd->sendctrl_lock, flags);
+		break;
+	}
+}
+
+static void writescratch(struct qib_devdata *dd, u32 val)
+{
+	qib_write_kreg(dd, kr_scratch, val);
+}
+
+#define VALID_TS_RD_REG_MASK 0xBF
+/**
+ * qib_7220_tempsense_read - read register of temp sensor via TWSI
+ * @dd: the qlogic_ib device
+ * @regnum: register to read from
+ *
+ * returns reg contents (0..255) or < 0 for error
+ */
+static int qib_7220_tempsense_rd(struct qib_devdata *dd, int regnum)
+{
+	int ret;
+	u8 rdata;
+
+	if (regnum > 7) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	/* return a bogus value for (the one) register we do not have */
+	if (!((1 << regnum) & VALID_TS_RD_REG_MASK)) {
+		ret = 0;
+		goto bail;
+	}
+
+	ret = mutex_lock_interruptible(&dd->eep_lock);
+	if (ret)
+		goto bail;
+
+	ret = qib_twsi_blk_rd(dd, QIB_TWSI_TEMP_DEV, regnum, &rdata, 1);
+	if (!ret)
+		ret = rdata;
+
+	mutex_unlock(&dd->eep_lock);
+
+	/*
+	 * There are three possibilities here:
+	 * ret is actual value (0..255)
+	 * ret is -ENXIO or -EINVAL from twsi code or this file
+	 * ret is -EINTR from mutex_lock_interruptible.
+	 */
+bail:
+	return ret;
+}
+
+/* Dummy function, as 7220 boards never disable EEPROM Write */
+static int qib_7220_eeprom_wen(struct qib_devdata *dd, int wen)
+{
+	return 1;
+}
+
+/**
+ * qib_init_iba7220_funcs - set up the chip-specific function pointers
+ * @dev: the pci_dev for qlogic_ib device
+ * @ent: pci_device_id struct for this dev
+ *
+ * This is global, and is called directly at init to set up the
+ * chip-specific function pointers for later use.
+ */
+struct qib_devdata *qib_init_iba7220_funcs(struct pci_dev *pdev,
+					   const struct pci_device_id *ent)
+{
+	struct qib_devdata *dd;
+	int ret;
+	u32 boardid, minwidth;
+
+	dd = qib_alloc_devdata(pdev, sizeof(struct qib_chip_specific) +
+		sizeof(struct qib_chippport_specific));
+	if (IS_ERR(dd))
+		goto bail;
+
+	dd->f_bringup_serdes    = qib_7220_bringup_serdes;
+	dd->f_cleanup           = qib_setup_7220_cleanup;
+	dd->f_clear_tids        = qib_7220_clear_tids;
+	dd->f_free_irq          = qib_7220_free_irq;
+	dd->f_get_base_info     = qib_7220_get_base_info;
+	dd->f_get_msgheader     = qib_7220_get_msgheader;
+	dd->f_getsendbuf        = qib_7220_getsendbuf;
+	dd->f_gpio_mod          = gpio_7220_mod;
+	dd->f_eeprom_wen        = qib_7220_eeprom_wen;
+	dd->f_hdrqempty         = qib_7220_hdrqempty;
+	dd->f_ib_updown         = qib_7220_ib_updown;
+	dd->f_init_ctxt         = qib_7220_init_ctxt;
+	dd->f_initvl15_bufs     = qib_7220_initvl15_bufs;
+	dd->f_intr_fallback     = qib_7220_intr_fallback;
+	dd->f_late_initreg      = qib_late_7220_initreg;
+	dd->f_setpbc_control    = qib_7220_setpbc_control;
+	dd->f_portcntr          = qib_portcntr_7220;
+	dd->f_put_tid           = qib_7220_put_tid;
+	dd->f_quiet_serdes      = qib_7220_quiet_serdes;
+	dd->f_rcvctrl           = rcvctrl_7220_mod;
+	dd->f_read_cntrs        = qib_read_7220cntrs;
+	dd->f_read_portcntrs    = qib_read_7220portcntrs;
+	dd->f_reset             = qib_setup_7220_reset;
+	dd->f_init_sdma_regs    = init_sdma_7220_regs;
+	dd->f_sdma_busy         = qib_sdma_7220_busy;
+	dd->f_sdma_gethead      = qib_sdma_7220_gethead;
+	dd->f_sdma_sendctrl     = qib_7220_sdma_sendctrl;
+	dd->f_sdma_set_desc_cnt = qib_sdma_set_7220_desc_cnt;
+	dd->f_sdma_update_tail  = qib_sdma_update_7220_tail;
+	dd->f_sdma_hw_clean_up  = qib_7220_sdma_hw_clean_up;
+	dd->f_sdma_hw_start_up  = qib_7220_sdma_hw_start_up;
+	dd->f_sdma_init_early   = qib_7220_sdma_init_early;
+	dd->f_sendctrl          = sendctrl_7220_mod;
+	dd->f_set_armlaunch     = qib_set_7220_armlaunch;
+	dd->f_set_cntr_sample   = qib_set_cntr_7220_sample;
+	dd->f_iblink_state      = qib_7220_iblink_state;
+	dd->f_ibphys_portstate  = qib_7220_phys_portstate;
+	dd->f_get_ib_cfg        = qib_7220_get_ib_cfg;
+	dd->f_set_ib_cfg        = qib_7220_set_ib_cfg;
+	dd->f_set_ib_loopback   = qib_7220_set_loopback;
+	dd->f_set_intr_state    = qib_7220_set_intr_state;
+	dd->f_setextled         = qib_setup_7220_setextled;
+	dd->f_txchk_change      = qib_7220_txchk_change;
+	dd->f_update_usrhead    = qib_update_7220_usrhead;
+	dd->f_wantpiobuf_intr   = qib_wantpiobuf_7220_intr;
+	dd->f_xgxs_reset        = qib_7220_xgxs_reset;
+	dd->f_writescratch      = writescratch;
+	dd->f_tempsense_rd	= qib_7220_tempsense_rd;
+	/*
+	 * Do remaining pcie setup and save pcie values in dd.
+	 * Any error printing is already done by the init code.
+	 * On return, we have the chip mapped, but chip registers
+	 * are not set up until start of qib_init_7220_variables.
+	 */
+	ret = qib_pcie_ddinit(dd, pdev, ent);
+	if (ret < 0)
+		goto bail_free;
+
+	/* initialize chip-specific variables */
+	ret = qib_init_7220_variables(dd);
+	if (ret)
+		goto bail_cleanup;
+
+	if (qib_mini_init)
+		goto bail;
+
+	boardid = SYM_FIELD(dd->revision, Revision,
+			    BoardID);
+	switch (boardid) {
+	case 0:
+	case 2:
+	case 10:
+	case 12:
+		minwidth = 16; /* x16 capable boards */
+		break;
+	default:
+		minwidth = 8; /* x8 capable boards */
+		break;
+	}
+	if (qib_pcie_params(dd, minwidth, NULL, NULL))
+		qib_dev_err(dd, "Failed to setup PCIe or interrupts; "
+			    "continuing anyway\n");
+
+	/* save IRQ for possible later use */
+	dd->cspec->irq = pdev->irq;
+
+	if (qib_read_kreg64(dd, kr_hwerrstatus) &
+	    QLOGIC_IB_HWE_SERDESPLLFAILED)
+		qib_write_kreg(dd, kr_hwerrclear,
+			       QLOGIC_IB_HWE_SERDESPLLFAILED);
+
+	/* setup interrupt handler (interrupt type handled above) */
+	qib_setup_7220_interrupt(dd);
+	qib_7220_init_hwerrors(dd);
+
+	/* clear diagctrl register, in case diags were running and crashed */
+	qib_write_kreg(dd, kr_hwdiagctrl, 0);
+
+	goto bail;
+
+bail_cleanup:
+	qib_pcie_ddcleanup(dd);
+bail_free:
+	qib_free_devdata(dd);
+	dd = ERR_PTR(ret);
+bail:
+	return dd;
+}
diff --git a/drivers/infiniband/hw/qib/qib_iba7322.c b/drivers/infiniband/hw/qib/qib_iba7322.c
new file mode 100644
index 000000000000..503992d9c5ce
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_iba7322.c
@@ -0,0 +1,7645 @@
+/*
+ * Copyright (c) 2008, 2009, 2010 QLogic Corporation. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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.
+ */
+
+/*
+ * This file contains all of the code that is specific to the
+ * InfiniPath 7322 chip
+ */
+
+#include <linux/interrupt.h>
+#include <linux/pci.h>
+#include <linux/delay.h>
+#include <linux/io.h>
+#include <linux/jiffies.h>
+#include <rdma/ib_verbs.h>
+#include <rdma/ib_smi.h>
+
+#include "qib.h"
+#include "qib_7322_regs.h"
+#include "qib_qsfp.h"
+
+#include "qib_mad.h"
+
+static void qib_setup_7322_setextled(struct qib_pportdata *, u32);
+static void qib_7322_handle_hwerrors(struct qib_devdata *, char *, size_t);
+static void sendctrl_7322_mod(struct qib_pportdata *ppd, u32 op);
+static irqreturn_t qib_7322intr(int irq, void *data);
+static irqreturn_t qib_7322bufavail(int irq, void *data);
+static irqreturn_t sdma_intr(int irq, void *data);
+static irqreturn_t sdma_idle_intr(int irq, void *data);
+static irqreturn_t sdma_progress_intr(int irq, void *data);
+static irqreturn_t sdma_cleanup_intr(int irq, void *data);
+static void qib_7322_txchk_change(struct qib_devdata *, u32, u32, u32,
+				  struct qib_ctxtdata *rcd);
+static u8 qib_7322_phys_portstate(u64);
+static u32 qib_7322_iblink_state(u64);
+static void qib_set_ib_7322_lstate(struct qib_pportdata *ppd, u16 linkcmd,
+				   u16 linitcmd);
+static void force_h1(struct qib_pportdata *);
+static void adj_tx_serdes(struct qib_pportdata *);
+static u32 qib_7322_setpbc_control(struct qib_pportdata *, u32, u8, u8);
+static void qib_7322_mini_pcs_reset(struct qib_pportdata *);
+
+static u32 ahb_mod(struct qib_devdata *, int, int, int, u32, u32);
+static void ibsd_wr_allchans(struct qib_pportdata *, int, unsigned, unsigned);
+
+#define BMASK(msb, lsb) (((1 << ((msb) + 1 - (lsb))) - 1) << (lsb))
+
+/* LE2 serdes values for different cases */
+#define LE2_DEFAULT 5
+#define LE2_5m 4
+#define LE2_QME 0
+
+/* Below is special-purpose, so only really works for the IB SerDes blocks. */
+#define IBSD(hw_pidx) (hw_pidx + 2)
+
+/* these are variables for documentation and experimentation purposes */
+static const unsigned rcv_int_timeout = 375;
+static const unsigned rcv_int_count = 16;
+static const unsigned sdma_idle_cnt = 64;
+
+/* Time to stop altering Rx Equalization parameters, after link up. */
+#define RXEQ_DISABLE_MSECS 2500
+
+/*
+ * Number of VLs we are configured to use (to allow for more
+ * credits per vl, etc.)
+ */
+ushort qib_num_cfg_vls = 2;
+module_param_named(num_vls, qib_num_cfg_vls, ushort, S_IRUGO);
+MODULE_PARM_DESC(num_vls, "Set number of Virtual Lanes to use (1-8)");
+
+static ushort qib_chase = 1;
+module_param_named(chase, qib_chase, ushort, S_IRUGO);
+MODULE_PARM_DESC(chase, "Enable state chase handling");
+
+static ushort qib_long_atten = 10; /* 10 dB ~= 5m length */
+module_param_named(long_attenuation, qib_long_atten, ushort, S_IRUGO);
+MODULE_PARM_DESC(long_attenuation, \
+		 "attenuation cutoff (dB) for long copper cable setup");
+
+static ushort qib_singleport;
+module_param_named(singleport, qib_singleport, ushort, S_IRUGO);
+MODULE_PARM_DESC(singleport, "Use only IB port 1; more per-port buffer space");
+
+#define MAX_ATTEN_LEN 64 /* plenty for any real system */
+/* for read back, default index is ~5m copper cable */
+static char txselect_list[MAX_ATTEN_LEN] = "10";
+static struct kparam_string kp_txselect = {
+	.string = txselect_list,
+	.maxlen = MAX_ATTEN_LEN
+};
+static int  setup_txselect(const char *, struct kernel_param *);
+module_param_call(txselect, setup_txselect, param_get_string,
+		  &kp_txselect, S_IWUSR | S_IRUGO);
+MODULE_PARM_DESC(txselect, \
+		 "Tx serdes indices (for no QSFP or invalid QSFP data)");
+
+#define BOARD_QME7342 5
+#define BOARD_QMH7342 6
+#define IS_QMH(dd) (SYM_FIELD((dd)->revision, Revision, BoardID) == \
+		    BOARD_QMH7342)
+#define IS_QME(dd) (SYM_FIELD((dd)->revision, Revision, BoardID) == \
+		    BOARD_QME7342)
+
+#define KREG_IDX(regname)     (QIB_7322_##regname##_OFFS / sizeof(u64))
+
+#define KREG_IBPORT_IDX(regname) ((QIB_7322_##regname##_0_OFFS / sizeof(u64)))
+
+#define MASK_ACROSS(lsb, msb) \
+	(((1ULL << ((msb) + 1 - (lsb))) - 1) << (lsb))
+
+#define SYM_RMASK(regname, fldname) ((u64)              \
+	QIB_7322_##regname##_##fldname##_RMASK)
+
+#define SYM_MASK(regname, fldname) ((u64)               \
+	QIB_7322_##regname##_##fldname##_RMASK <<       \
+	 QIB_7322_##regname##_##fldname##_LSB)
+
+#define SYM_FIELD(value, regname, fldname) ((u64)	\
+	(((value) >> SYM_LSB(regname, fldname)) &	\
+	 SYM_RMASK(regname, fldname)))
+
+/* useful for things like LaFifoEmpty_0...7, TxCreditOK_0...7, etc. */
+#define SYM_FIELD_ACROSS(value, regname, fldname, nbits) \
+	(((value) >> SYM_LSB(regname, fldname)) & MASK_ACROSS(0, nbits))
+
+#define HWE_MASK(fldname) SYM_MASK(HwErrMask, fldname##Mask)
+#define ERR_MASK(fldname) SYM_MASK(ErrMask, fldname##Mask)
+#define ERR_MASK_N(fldname) SYM_MASK(ErrMask_0, fldname##Mask)
+#define INT_MASK(fldname) SYM_MASK(IntMask, fldname##IntMask)
+#define INT_MASK_P(fldname, port) SYM_MASK(IntMask, fldname##IntMask##_##port)
+/* Below because most, but not all, fields of IntMask have that full suffix */
+#define INT_MASK_PM(fldname, port) SYM_MASK(IntMask, fldname##Mask##_##port)
+
+
+#define SYM_LSB(regname, fldname) (QIB_7322_##regname##_##fldname##_LSB)
+
+/*
+ * the size bits give us 2^N, in KB units.  0 marks as invalid,
+ * and 7 is reserved.  We currently use only 2KB and 4KB
+ */
+#define IBA7322_TID_SZ_SHIFT QIB_7322_RcvTIDArray0_RT_BufSize_LSB
+#define IBA7322_TID_SZ_2K (1UL<<IBA7322_TID_SZ_SHIFT) /* 2KB */
+#define IBA7322_TID_SZ_4K (2UL<<IBA7322_TID_SZ_SHIFT) /* 4KB */
+#define IBA7322_TID_PA_SHIFT 11U /* TID addr in chip stored w/o low bits */
+
+#define SendIBSLIDAssignMask \
+	QIB_7322_SendIBSLIDAssign_0_SendIBSLIDAssign_15_0_RMASK
+#define SendIBSLMCMask \
+	QIB_7322_SendIBSLIDMask_0_SendIBSLIDMask_15_0_RMASK
+
+#define ExtLED_IB1_YEL SYM_MASK(EXTCtrl, LEDPort0YellowOn)
+#define ExtLED_IB1_GRN SYM_MASK(EXTCtrl, LEDPort0GreenOn)
+#define ExtLED_IB2_YEL SYM_MASK(EXTCtrl, LEDPort1YellowOn)
+#define ExtLED_IB2_GRN SYM_MASK(EXTCtrl, LEDPort1GreenOn)
+#define ExtLED_IB1_MASK (ExtLED_IB1_YEL | ExtLED_IB1_GRN)
+#define ExtLED_IB2_MASK (ExtLED_IB2_YEL | ExtLED_IB2_GRN)
+
+#define _QIB_GPIO_SDA_NUM 1
+#define _QIB_GPIO_SCL_NUM 0
+#define QIB_EEPROM_WEN_NUM 14
+#define QIB_TWSI_EEPROM_DEV 0xA2 /* All Production 7322 cards. */
+
+/* HW counter clock is at 4nsec */
+#define QIB_7322_PSXMITWAIT_CHECK_RATE 4000
+
+/* full speed IB port 1 only */
+#define PORT_SPD_CAP (QIB_IB_SDR | QIB_IB_DDR | QIB_IB_QDR)
+#define PORT_SPD_CAP_SHIFT 3
+
+/* full speed featuremask, both ports */
+#define DUAL_PORT_CAP (PORT_SPD_CAP | (PORT_SPD_CAP << PORT_SPD_CAP_SHIFT))
+
+/*
+ * This file contains almost all the chip-specific register information and
+ * access functions for the FAKED QLogic InfiniPath 7322 PCI-Express chip.
+ */
+
+/* Use defines to tie machine-generated names to lower-case names */
+#define kr_contextcnt KREG_IDX(ContextCnt)
+#define kr_control KREG_IDX(Control)
+#define kr_counterregbase KREG_IDX(CntrRegBase)
+#define kr_errclear KREG_IDX(ErrClear)
+#define kr_errmask KREG_IDX(ErrMask)
+#define kr_errstatus KREG_IDX(ErrStatus)
+#define kr_extctrl KREG_IDX(EXTCtrl)
+#define kr_extstatus KREG_IDX(EXTStatus)
+#define kr_gpio_clear KREG_IDX(GPIOClear)
+#define kr_gpio_mask KREG_IDX(GPIOMask)
+#define kr_gpio_out KREG_IDX(GPIOOut)
+#define kr_gpio_status KREG_IDX(GPIOStatus)
+#define kr_hwdiagctrl KREG_IDX(HwDiagCtrl)
+#define kr_debugportval KREG_IDX(DebugPortValueReg)
+#define kr_fmask KREG_IDX(feature_mask)
+#define kr_act_fmask KREG_IDX(active_feature_mask)
+#define kr_hwerrclear KREG_IDX(HwErrClear)
+#define kr_hwerrmask KREG_IDX(HwErrMask)
+#define kr_hwerrstatus KREG_IDX(HwErrStatus)
+#define kr_intclear KREG_IDX(IntClear)
+#define kr_intmask KREG_IDX(IntMask)
+#define kr_intredirect KREG_IDX(IntRedirect0)
+#define kr_intstatus KREG_IDX(IntStatus)
+#define kr_pagealign KREG_IDX(PageAlign)
+#define kr_rcvavailtimeout KREG_IDX(RcvAvailTimeOut0)
+#define kr_rcvctrl KREG_IDX(RcvCtrl) /* Common, but chip also has per-port */
+#define kr_rcvegrbase KREG_IDX(RcvEgrBase)
+#define kr_rcvegrcnt KREG_IDX(RcvEgrCnt)
+#define kr_rcvhdrcnt KREG_IDX(RcvHdrCnt)
+#define kr_rcvhdrentsize KREG_IDX(RcvHdrEntSize)
+#define kr_rcvhdrsize KREG_IDX(RcvHdrSize)
+#define kr_rcvtidbase KREG_IDX(RcvTIDBase)
+#define kr_rcvtidcnt KREG_IDX(RcvTIDCnt)
+#define kr_revision KREG_IDX(Revision)
+#define kr_scratch KREG_IDX(Scratch)
+#define kr_sendbuffererror KREG_IDX(SendBufErr0) /* and base for 1 and 2 */
+#define kr_sendcheckmask KREG_IDX(SendCheckMask0) /* and 1, 2 */
+#define kr_sendctrl KREG_IDX(SendCtrl)
+#define kr_sendgrhcheckmask KREG_IDX(SendGRHCheckMask0) /* and 1, 2 */
+#define kr_sendibpktmask KREG_IDX(SendIBPacketMask0) /* and 1, 2 */
+#define kr_sendpioavailaddr KREG_IDX(SendBufAvailAddr)
+#define kr_sendpiobufbase KREG_IDX(SendBufBase)
+#define kr_sendpiobufcnt KREG_IDX(SendBufCnt)
+#define kr_sendpiosize KREG_IDX(SendBufSize)
+#define kr_sendregbase KREG_IDX(SendRegBase)
+#define kr_sendbufavail0 KREG_IDX(SendBufAvail0)
+#define kr_userregbase KREG_IDX(UserRegBase)
+#define kr_intgranted KREG_IDX(Int_Granted)
+#define kr_vecclr_wo_int KREG_IDX(vec_clr_without_int)
+#define kr_intblocked KREG_IDX(IntBlocked)
+#define kr_r_access KREG_IDX(SPC_JTAG_ACCESS_REG)
+
+/*
+ * per-port kernel registers.  Access only with qib_read_kreg_port()
+ * or qib_write_kreg_port()
+ */
+#define krp_errclear KREG_IBPORT_IDX(ErrClear)
+#define krp_errmask KREG_IBPORT_IDX(ErrMask)
+#define krp_errstatus KREG_IBPORT_IDX(ErrStatus)
+#define krp_highprio_0 KREG_IBPORT_IDX(HighPriority0)
+#define krp_highprio_limit KREG_IBPORT_IDX(HighPriorityLimit)
+#define krp_hrtbt_guid KREG_IBPORT_IDX(HRTBT_GUID)
+#define krp_ib_pcsconfig KREG_IBPORT_IDX(IBPCSConfig)
+#define krp_ibcctrl_a KREG_IBPORT_IDX(IBCCtrlA)
+#define krp_ibcctrl_b KREG_IBPORT_IDX(IBCCtrlB)
+#define krp_ibcctrl_c KREG_IBPORT_IDX(IBCCtrlC)
+#define krp_ibcstatus_a KREG_IBPORT_IDX(IBCStatusA)
+#define krp_ibcstatus_b KREG_IBPORT_IDX(IBCStatusB)
+#define krp_txestatus KREG_IBPORT_IDX(TXEStatus)
+#define krp_lowprio_0 KREG_IBPORT_IDX(LowPriority0)
+#define krp_ncmodectrl KREG_IBPORT_IDX(IBNCModeCtrl)
+#define krp_partitionkey KREG_IBPORT_IDX(RcvPartitionKey)
+#define krp_psinterval KREG_IBPORT_IDX(PSInterval)
+#define krp_psstart KREG_IBPORT_IDX(PSStart)
+#define krp_psstat KREG_IBPORT_IDX(PSStat)
+#define krp_rcvbthqp KREG_IBPORT_IDX(RcvBTHQP)
+#define krp_rcvctrl KREG_IBPORT_IDX(RcvCtrl)
+#define krp_rcvpktledcnt KREG_IBPORT_IDX(RcvPktLEDCnt)
+#define krp_rcvqpmaptable KREG_IBPORT_IDX(RcvQPMapTableA)
+#define krp_rxcreditvl0 KREG_IBPORT_IDX(RxCreditVL0)
+#define krp_rxcreditvl15 (KREG_IBPORT_IDX(RxCreditVL0)+15)
+#define krp_sendcheckcontrol KREG_IBPORT_IDX(SendCheckControl)
+#define krp_sendctrl KREG_IBPORT_IDX(SendCtrl)
+#define krp_senddmabase KREG_IBPORT_IDX(SendDmaBase)
+#define krp_senddmabufmask0 KREG_IBPORT_IDX(SendDmaBufMask0)
+#define krp_senddmabufmask1 (KREG_IBPORT_IDX(SendDmaBufMask0) + 1)
+#define krp_senddmabufmask2 (KREG_IBPORT_IDX(SendDmaBufMask0) + 2)
+#define krp_senddmabuf_use0 KREG_IBPORT_IDX(SendDmaBufUsed0)
+#define krp_senddmabuf_use1 (KREG_IBPORT_IDX(SendDmaBufUsed0) + 1)
+#define krp_senddmabuf_use2 (KREG_IBPORT_IDX(SendDmaBufUsed0) + 2)
+#define krp_senddmadesccnt KREG_IBPORT_IDX(SendDmaDescCnt)
+#define krp_senddmahead KREG_IBPORT_IDX(SendDmaHead)
+#define krp_senddmaheadaddr KREG_IBPORT_IDX(SendDmaHeadAddr)
+#define krp_senddmaidlecnt KREG_IBPORT_IDX(SendDmaIdleCnt)
+#define krp_senddmalengen KREG_IBPORT_IDX(SendDmaLenGen)
+#define krp_senddmaprioritythld KREG_IBPORT_IDX(SendDmaPriorityThld)
+#define krp_senddmareloadcnt KREG_IBPORT_IDX(SendDmaReloadCnt)
+#define krp_senddmastatus KREG_IBPORT_IDX(SendDmaStatus)
+#define krp_senddmatail KREG_IBPORT_IDX(SendDmaTail)
+#define krp_sendhdrsymptom KREG_IBPORT_IDX(SendHdrErrSymptom)
+#define krp_sendslid KREG_IBPORT_IDX(SendIBSLIDAssign)
+#define krp_sendslidmask KREG_IBPORT_IDX(SendIBSLIDMask)
+#define krp_ibsdtestiftx KREG_IBPORT_IDX(IB_SDTEST_IF_TX)
+#define krp_adapt_dis_timer KREG_IBPORT_IDX(ADAPT_DISABLE_TIMER_THRESHOLD)
+#define krp_tx_deemph_override KREG_IBPORT_IDX(IBSD_TX_DEEMPHASIS_OVERRIDE)
+#define krp_serdesctrl KREG_IBPORT_IDX(IBSerdesCtrl)
+
+/*
+ * Per-context kernel registers.  Acess only with qib_read_kreg_ctxt()
+ * or qib_write_kreg_ctxt()
+ */
+#define krc_rcvhdraddr KREG_IDX(RcvHdrAddr0)
+#define krc_rcvhdrtailaddr KREG_IDX(RcvHdrTailAddr0)
+
+/*
+ * TID Flow table, per context.  Reduces
+ * number of hdrq updates to one per flow (or on errors).
+ * context 0 and 1 share same memory, but have distinct
+ * addresses.  Since for now, we never use expected sends
+ * on kernel contexts, we don't worry about that (we initialize
+ * those entries for ctxt 0/1 on driver load twice, for example).
+ */
+#define NUM_TIDFLOWS_CTXT 0x20 /* 0x20 per context; have to hardcode */
+#define ur_rcvflowtable (KREG_IDX(RcvTIDFlowTable0) - KREG_IDX(RcvHdrTail0))
+
+/* these are the error bits in the tid flows, and are W1C */
+#define TIDFLOW_ERRBITS  ( \
+	(SYM_MASK(RcvTIDFlowTable0, GenMismatch) << \
+	SYM_LSB(RcvTIDFlowTable0, GenMismatch)) | \
+	(SYM_MASK(RcvTIDFlowTable0, SeqMismatch) << \
+	SYM_LSB(RcvTIDFlowTable0, SeqMismatch)))
+
+/* Most (not all) Counters are per-IBport.
+ * Requires LBIntCnt is at offset 0 in the group
+ */
+#define CREG_IDX(regname) \
+((QIB_7322_##regname##_0_OFFS - QIB_7322_LBIntCnt_OFFS) / sizeof(u64))
+
+#define crp_badformat CREG_IDX(RxVersionErrCnt)
+#define crp_err_rlen CREG_IDX(RxLenErrCnt)
+#define crp_erricrc CREG_IDX(RxICRCErrCnt)
+#define crp_errlink CREG_IDX(RxLinkMalformCnt)
+#define crp_errlpcrc CREG_IDX(RxLPCRCErrCnt)
+#define crp_errpkey CREG_IDX(RxPKeyMismatchCnt)
+#define crp_errvcrc CREG_IDX(RxVCRCErrCnt)
+#define crp_excessbufferovfl CREG_IDX(ExcessBufferOvflCnt)
+#define crp_iblinkdown CREG_IDX(IBLinkDownedCnt)
+#define crp_iblinkerrrecov CREG_IDX(IBLinkErrRecoveryCnt)
+#define crp_ibstatuschange CREG_IDX(IBStatusChangeCnt)
+#define crp_ibsymbolerr CREG_IDX(IBSymbolErrCnt)
+#define crp_invalidrlen CREG_IDX(RxMaxMinLenErrCnt)
+#define crp_locallinkintegrityerr CREG_IDX(LocalLinkIntegrityErrCnt)
+#define crp_pktrcv CREG_IDX(RxDataPktCnt)
+#define crp_pktrcvflowctrl CREG_IDX(RxFlowPktCnt)
+#define crp_pktsend CREG_IDX(TxDataPktCnt)
+#define crp_pktsendflow CREG_IDX(TxFlowPktCnt)
+#define crp_psrcvdatacount CREG_IDX(PSRcvDataCount)
+#define crp_psrcvpktscount CREG_IDX(PSRcvPktsCount)
+#define crp_psxmitdatacount CREG_IDX(PSXmitDataCount)
+#define crp_psxmitpktscount CREG_IDX(PSXmitPktsCount)
+#define crp_psxmitwaitcount CREG_IDX(PSXmitWaitCount)
+#define crp_rcvebp CREG_IDX(RxEBPCnt)
+#define crp_rcvflowctrlviol CREG_IDX(RxFlowCtrlViolCnt)
+#define crp_rcvovfl CREG_IDX(RxBufOvflCnt)
+#define crp_rxdlidfltr CREG_IDX(RxDlidFltrCnt)
+#define crp_rxdroppkt CREG_IDX(RxDroppedPktCnt)
+#define crp_rxotherlocalphyerr CREG_IDX(RxOtherLocalPhyErrCnt)
+#define crp_rxqpinvalidctxt CREG_IDX(RxQPInvalidContextCnt)
+#define crp_rxvlerr CREG_IDX(RxVlErrCnt)
+#define crp_sendstall CREG_IDX(TxFlowStallCnt)
+#define crp_txdroppedpkt CREG_IDX(TxDroppedPktCnt)
+#define crp_txhdrerr CREG_IDX(TxHeadersErrCnt)
+#define crp_txlenerr CREG_IDX(TxLenErrCnt)
+#define crp_txlenerr CREG_IDX(TxLenErrCnt)
+#define crp_txminmaxlenerr CREG_IDX(TxMaxMinLenErrCnt)
+#define crp_txsdmadesc CREG_IDX(TxSDmaDescCnt)
+#define crp_txunderrun CREG_IDX(TxUnderrunCnt)
+#define crp_txunsupvl CREG_IDX(TxUnsupVLErrCnt)
+#define crp_vl15droppedpkt CREG_IDX(RxVL15DroppedPktCnt)
+#define crp_wordrcv CREG_IDX(RxDwordCnt)
+#define crp_wordsend CREG_IDX(TxDwordCnt)
+#define crp_tx_creditstalls CREG_IDX(TxCreditUpToDateTimeOut)
+
+/* these are the (few) counters that are not port-specific */
+#define CREG_DEVIDX(regname) ((QIB_7322_##regname##_OFFS - \
+			QIB_7322_LBIntCnt_OFFS) / sizeof(u64))
+#define cr_base_egrovfl CREG_DEVIDX(RxP0HdrEgrOvflCnt)
+#define cr_lbint CREG_DEVIDX(LBIntCnt)
+#define cr_lbstall CREG_DEVIDX(LBFlowStallCnt)
+#define cr_pcieretrydiag CREG_DEVIDX(PcieRetryBufDiagQwordCnt)
+#define cr_rxtidflowdrop CREG_DEVIDX(RxTidFlowDropCnt)
+#define cr_tidfull CREG_DEVIDX(RxTIDFullErrCnt)
+#define cr_tidinvalid CREG_DEVIDX(RxTIDValidErrCnt)
+
+/* no chip register for # of IB ports supported, so define */
+#define NUM_IB_PORTS 2
+
+/* 1 VL15 buffer per hardware IB port, no register for this, so define */
+#define NUM_VL15_BUFS NUM_IB_PORTS
+
+/*
+ * context 0 and 1 are special, and there is no chip register that
+ * defines this value, so we have to define it here.
+ * These are all allocated to either 0 or 1 for single port
+ * hardware configuration, otherwise each gets half
+ */
+#define KCTXT0_EGRCNT 2048
+
+/* values for vl and port fields in PBC, 7322-specific */
+#define PBC_PORT_SEL_LSB 26
+#define PBC_PORT_SEL_RMASK 1
+#define PBC_VL_NUM_LSB 27
+#define PBC_VL_NUM_RMASK 7
+#define PBC_7322_VL15_SEND (1ULL << 63) /* pbc; VL15, no credit check */
+#define PBC_7322_VL15_SEND_CTRL (1ULL << 31) /* control version of same */
+
+static u8 ib_rate_to_delay[IB_RATE_120_GBPS + 1] = {
+	[IB_RATE_2_5_GBPS] = 16,
+	[IB_RATE_5_GBPS] = 8,
+	[IB_RATE_10_GBPS] = 4,
+	[IB_RATE_20_GBPS] = 2,
+	[IB_RATE_30_GBPS] = 2,
+	[IB_RATE_40_GBPS] = 1
+};
+
+#define IBA7322_LINKSPEED_SHIFT SYM_LSB(IBCStatusA_0, LinkSpeedActive)
+#define IBA7322_LINKWIDTH_SHIFT SYM_LSB(IBCStatusA_0, LinkWidthActive)
+
+/* link training states, from IBC */
+#define IB_7322_LT_STATE_DISABLED        0x00
+#define IB_7322_LT_STATE_LINKUP          0x01
+#define IB_7322_LT_STATE_POLLACTIVE      0x02
+#define IB_7322_LT_STATE_POLLQUIET       0x03
+#define IB_7322_LT_STATE_SLEEPDELAY      0x04
+#define IB_7322_LT_STATE_SLEEPQUIET      0x05
+#define IB_7322_LT_STATE_CFGDEBOUNCE     0x08
+#define IB_7322_LT_STATE_CFGRCVFCFG      0x09
+#define IB_7322_LT_STATE_CFGWAITRMT      0x0a
+#define IB_7322_LT_STATE_CFGIDLE         0x0b
+#define IB_7322_LT_STATE_RECOVERRETRAIN  0x0c
+#define IB_7322_LT_STATE_TXREVLANES      0x0d
+#define IB_7322_LT_STATE_RECOVERWAITRMT  0x0e
+#define IB_7322_LT_STATE_RECOVERIDLE     0x0f
+#define IB_7322_LT_STATE_CFGENH          0x10
+#define IB_7322_LT_STATE_CFGTEST         0x11
+
+/* link state machine states from IBC */
+#define IB_7322_L_STATE_DOWN             0x0
+#define IB_7322_L_STATE_INIT             0x1
+#define IB_7322_L_STATE_ARM              0x2
+#define IB_7322_L_STATE_ACTIVE           0x3
+#define IB_7322_L_STATE_ACT_DEFER        0x4
+
+static const u8 qib_7322_physportstate[0x20] = {
+	[IB_7322_LT_STATE_DISABLED] = IB_PHYSPORTSTATE_DISABLED,
+	[IB_7322_LT_STATE_LINKUP] = IB_PHYSPORTSTATE_LINKUP,
+	[IB_7322_LT_STATE_POLLACTIVE] = IB_PHYSPORTSTATE_POLL,
+	[IB_7322_LT_STATE_POLLQUIET] = IB_PHYSPORTSTATE_POLL,
+	[IB_7322_LT_STATE_SLEEPDELAY] = IB_PHYSPORTSTATE_SLEEP,
+	[IB_7322_LT_STATE_SLEEPQUIET] = IB_PHYSPORTSTATE_SLEEP,
+	[IB_7322_LT_STATE_CFGDEBOUNCE] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[IB_7322_LT_STATE_CFGRCVFCFG] =
+		IB_PHYSPORTSTATE_CFG_TRAIN,
+	[IB_7322_LT_STATE_CFGWAITRMT] =
+		IB_PHYSPORTSTATE_CFG_TRAIN,
+	[IB_7322_LT_STATE_CFGIDLE] = IB_PHYSPORTSTATE_CFG_IDLE,
+	[IB_7322_LT_STATE_RECOVERRETRAIN] =
+		IB_PHYSPORTSTATE_LINK_ERR_RECOVER,
+	[IB_7322_LT_STATE_RECOVERWAITRMT] =
+		IB_PHYSPORTSTATE_LINK_ERR_RECOVER,
+	[IB_7322_LT_STATE_RECOVERIDLE] =
+		IB_PHYSPORTSTATE_LINK_ERR_RECOVER,
+	[IB_7322_LT_STATE_CFGENH] = IB_PHYSPORTSTATE_CFG_ENH,
+	[IB_7322_LT_STATE_CFGTEST] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x12] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x13] = IB_PHYSPORTSTATE_CFG_WAIT_ENH,
+	[0x14] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x15] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x16] = IB_PHYSPORTSTATE_CFG_TRAIN,
+	[0x17] = IB_PHYSPORTSTATE_CFG_TRAIN
+};
+
+struct qib_chip_specific {
+	u64 __iomem *cregbase;
+	u64 *cntrs;
+	spinlock_t rcvmod_lock; /* protect rcvctrl shadow changes */
+	spinlock_t gpio_lock; /* RMW of shadows/regs for ExtCtrl and GPIO */
+	u64 main_int_mask;      /* clear bits which have dedicated handlers */
+	u64 int_enable_mask;  /* for per port interrupts in single port mode */
+	u64 errormask;
+	u64 hwerrmask;
+	u64 gpio_out; /* shadow of kr_gpio_out, for rmw ops */
+	u64 gpio_mask; /* shadow the gpio mask register */
+	u64 extctrl; /* shadow the gpio output enable, etc... */
+	u32 ncntrs;
+	u32 nportcntrs;
+	u32 cntrnamelen;
+	u32 portcntrnamelen;
+	u32 numctxts;
+	u32 rcvegrcnt;
+	u32 updthresh; /* current AvailUpdThld */
+	u32 updthresh_dflt; /* default AvailUpdThld */
+	u32 r1;
+	int irq;
+	u32 num_msix_entries;
+	u32 sdmabufcnt;
+	u32 lastbuf_for_pio;
+	u32 stay_in_freeze;
+	u32 recovery_ports_initted;
+	struct msix_entry *msix_entries;
+	void  **msix_arg;
+	unsigned long *sendchkenable;
+	unsigned long *sendgrhchk;
+	unsigned long *sendibchk;
+	u32 rcvavail_timeout[18];
+	char emsgbuf[128]; /* for device error interrupt msg buffer */
+};
+
+/* Table of entries in "human readable" form Tx Emphasis. */
+struct txdds_ent {
+	u8 amp;
+	u8 pre;
+	u8 main;
+	u8 post;
+};
+
+struct vendor_txdds_ent {
+	u8 oui[QSFP_VOUI_LEN];
+	u8 *partnum;
+	struct txdds_ent sdr;
+	struct txdds_ent ddr;
+	struct txdds_ent qdr;
+};
+
+static void write_tx_serdes_param(struct qib_pportdata *, struct txdds_ent *);
+
+#define TXDDS_TABLE_SZ 16 /* number of entries per speed in onchip table */
+#define TXDDS_EXTRA_SZ 11 /* number of extra tx settings entries */
+#define SERDES_CHANS 4 /* yes, it's obvious, but one less magic number */
+
+#define H1_FORCE_VAL 8
+#define H1_FORCE_QME 1 /*  may be overridden via setup_txselect() */
+#define H1_FORCE_QMH 7 /*  may be overridden via setup_txselect() */
+
+/* The static and dynamic registers are paired, and the pairs indexed by spd */
+#define krp_static_adapt_dis(spd) (KREG_IBPORT_IDX(ADAPT_DISABLE_STATIC_SDR) \
+	+ ((spd) * 2))
+
+#define QDR_DFE_DISABLE_DELAY 4000 /* msec after LINKUP */
+#define QDR_STATIC_ADAPT_DOWN 0xf0f0f0f0ULL /* link down, H1-H4 QDR adapts */
+#define QDR_STATIC_ADAPT_DOWN_R1 0ULL /* r1 link down, H1-H4 QDR adapts */
+#define QDR_STATIC_ADAPT_INIT 0xffffffffffULL /* up, disable H0,H1-8, LE */
+#define QDR_STATIC_ADAPT_INIT_R1 0xf0ffffffffULL /* r1 up, disable H0,H1-8 */
+
+struct qib_chippport_specific {
+	u64 __iomem *kpregbase;
+	u64 __iomem *cpregbase;
+	u64 *portcntrs;
+	struct qib_pportdata *ppd;
+	wait_queue_head_t autoneg_wait;
+	struct delayed_work autoneg_work;
+	struct delayed_work ipg_work;
+	struct timer_list chase_timer;
+	/*
+	 * these 5 fields are used to establish deltas for IB symbol
+	 * errors and linkrecovery errors.  They can be reported on
+	 * some chips during link negotiation prior to INIT, and with
+	 * DDR when faking DDR negotiations with non-IBTA switches.
+	 * The chip counters are adjusted at driver unload if there is
+	 * a non-zero delta.
+	 */
+	u64 ibdeltainprog;
+	u64 ibsymdelta;
+	u64 ibsymsnap;
+	u64 iblnkerrdelta;
+	u64 iblnkerrsnap;
+	u64 iblnkdownsnap;
+	u64 iblnkdowndelta;
+	u64 ibmalfdelta;
+	u64 ibmalfsnap;
+	u64 ibcctrl_a; /* krp_ibcctrl_a shadow */
+	u64 ibcctrl_b; /* krp_ibcctrl_b shadow */
+	u64 qdr_dfe_time;
+	u64 chase_end;
+	u32 autoneg_tries;
+	u32 recovery_init;
+	u32 qdr_dfe_on;
+	u32 qdr_reforce;
+	/*
+	 * Per-bay per-channel rcv QMH H1 values and Tx values for QDR.
+	 * entry zero is unused, to simplify indexing
+	 */
+	u8 h1_val;
+	u8 no_eep;  /* txselect table index to use if no qsfp info */
+	u8 ipg_tries;
+	u8 ibmalfusesnap;
+	struct qib_qsfp_data qsfp_data;
+	char epmsgbuf[192]; /* for port error interrupt msg buffer */
+};
+
+static struct {
+	const char *name;
+	irq_handler_t handler;
+	int lsb;
+	int port; /* 0 if not port-specific, else port # */
+} irq_table[] = {
+	{ QIB_DRV_NAME, qib_7322intr, -1, 0 },
+	{ QIB_DRV_NAME " (buf avail)", qib_7322bufavail,
+		SYM_LSB(IntStatus, SendBufAvail), 0 },
+	{ QIB_DRV_NAME " (sdma 0)", sdma_intr,
+		SYM_LSB(IntStatus, SDmaInt_0), 1 },
+	{ QIB_DRV_NAME " (sdma 1)", sdma_intr,
+		SYM_LSB(IntStatus, SDmaInt_1), 2 },
+	{ QIB_DRV_NAME " (sdmaI 0)", sdma_idle_intr,
+		SYM_LSB(IntStatus, SDmaIdleInt_0), 1 },
+	{ QIB_DRV_NAME " (sdmaI 1)", sdma_idle_intr,
+		SYM_LSB(IntStatus, SDmaIdleInt_1), 2 },
+	{ QIB_DRV_NAME " (sdmaP 0)", sdma_progress_intr,
+		SYM_LSB(IntStatus, SDmaProgressInt_0), 1 },
+	{ QIB_DRV_NAME " (sdmaP 1)", sdma_progress_intr,
+		SYM_LSB(IntStatus, SDmaProgressInt_1), 2 },
+	{ QIB_DRV_NAME " (sdmaC 0)", sdma_cleanup_intr,
+		SYM_LSB(IntStatus, SDmaCleanupDone_0), 1 },
+	{ QIB_DRV_NAME " (sdmaC 1)", sdma_cleanup_intr,
+		SYM_LSB(IntStatus, SDmaCleanupDone_1), 2 },
+};
+
+/* ibcctrl bits */
+#define QLOGIC_IB_IBCC_LINKINITCMD_DISABLE 1
+/* cycle through TS1/TS2 till OK */
+#define QLOGIC_IB_IBCC_LINKINITCMD_POLL 2
+/* wait for TS1, then go on */
+#define QLOGIC_IB_IBCC_LINKINITCMD_SLEEP 3
+#define QLOGIC_IB_IBCC_LINKINITCMD_SHIFT 16
+
+#define QLOGIC_IB_IBCC_LINKCMD_DOWN 1           /* move to 0x11 */
+#define QLOGIC_IB_IBCC_LINKCMD_ARMED 2          /* move to 0x21 */
+#define QLOGIC_IB_IBCC_LINKCMD_ACTIVE 3 /* move to 0x31 */
+
+#define BLOB_7322_IBCHG 0x101
+
+static inline void qib_write_kreg(const struct qib_devdata *dd,
+				  const u32 regno, u64 value);
+static inline u32 qib_read_kreg32(const struct qib_devdata *, const u32);
+static void write_7322_initregs(struct qib_devdata *);
+static void write_7322_init_portregs(struct qib_pportdata *);
+static void setup_7322_link_recovery(struct qib_pportdata *, u32);
+static void check_7322_rxe_status(struct qib_pportdata *);
+static u32 __iomem *qib_7322_getsendbuf(struct qib_pportdata *, u64, u32 *);
+
+/**
+ * qib_read_ureg32 - read 32-bit virtualized per-context register
+ * @dd: device
+ * @regno: register number
+ * @ctxt: context number
+ *
+ * Return the contents of a register that is virtualized to be per context.
+ * Returns -1 on errors (not distinguishable from valid contents at
+ * runtime; we may add a separate error variable at some point).
+ */
+static inline u32 qib_read_ureg32(const struct qib_devdata *dd,
+				  enum qib_ureg regno, int ctxt)
+{
+	if (!dd->kregbase || !(dd->flags & QIB_PRESENT))
+		return 0;
+	return readl(regno + (u64 __iomem *)(
+		(dd->ureg_align * ctxt) + (dd->userbase ?
+		 (char __iomem *)dd->userbase :
+		 (char __iomem *)dd->kregbase + dd->uregbase)));
+}
+
+/**
+ * qib_read_ureg - read virtualized per-context register
+ * @dd: device
+ * @regno: register number
+ * @ctxt: context number
+ *
+ * Return the contents of a register that is virtualized to be per context.
+ * Returns -1 on errors (not distinguishable from valid contents at
+ * runtime; we may add a separate error variable at some point).
+ */
+static inline u64 qib_read_ureg(const struct qib_devdata *dd,
+				enum qib_ureg regno, int ctxt)
+{
+
+	if (!dd->kregbase || !(dd->flags & QIB_PRESENT))
+		return 0;
+	return readq(regno + (u64 __iomem *)(
+		(dd->ureg_align * ctxt) + (dd->userbase ?
+		 (char __iomem *)dd->userbase :
+		 (char __iomem *)dd->kregbase + dd->uregbase)));
+}
+
+/**
+ * qib_write_ureg - write virtualized per-context register
+ * @dd: device
+ * @regno: register number
+ * @value: value
+ * @ctxt: context
+ *
+ * Write the contents of a register that is virtualized to be per context.
+ */
+static inline void qib_write_ureg(const struct qib_devdata *dd,
+				  enum qib_ureg regno, u64 value, int ctxt)
+{
+	u64 __iomem *ubase;
+	if (dd->userbase)
+		ubase = (u64 __iomem *)
+			((char __iomem *) dd->userbase +
+			 dd->ureg_align * ctxt);
+	else
+		ubase = (u64 __iomem *)
+			(dd->uregbase +
+			 (char __iomem *) dd->kregbase +
+			 dd->ureg_align * ctxt);
+
+	if (dd->kregbase && (dd->flags & QIB_PRESENT))
+		writeq(value, &ubase[regno]);
+}
+
+static inline u32 qib_read_kreg32(const struct qib_devdata *dd,
+				  const u32 regno)
+{
+	if (!dd->kregbase || !(dd->flags & QIB_PRESENT))
+		return -1;
+	return readl((u32 __iomem *) &dd->kregbase[regno]);
+}
+
+static inline u64 qib_read_kreg64(const struct qib_devdata *dd,
+				  const u32 regno)
+{
+	if (!dd->kregbase || !(dd->flags & QIB_PRESENT))
+		return -1;
+	return readq(&dd->kregbase[regno]);
+}
+
+static inline void qib_write_kreg(const struct qib_devdata *dd,
+				  const u32 regno, u64 value)
+{
+	if (dd->kregbase && (dd->flags & QIB_PRESENT))
+		writeq(value, &dd->kregbase[regno]);
+}
+
+/*
+ * not many sanity checks for the port-specific kernel register routines,
+ * since they are only used when it's known to be safe.
+*/
+static inline u64 qib_read_kreg_port(const struct qib_pportdata *ppd,
+				     const u16 regno)
+{
+	if (!ppd->cpspec->kpregbase || !(ppd->dd->flags & QIB_PRESENT))
+		return 0ULL;
+	return readq(&ppd->cpspec->kpregbase[regno]);
+}
+
+static inline void qib_write_kreg_port(const struct qib_pportdata *ppd,
+				       const u16 regno, u64 value)
+{
+	if (ppd->cpspec && ppd->dd && ppd->cpspec->kpregbase &&
+	    (ppd->dd->flags & QIB_PRESENT))
+		writeq(value, &ppd->cpspec->kpregbase[regno]);
+}
+
+/**
+ * qib_write_kreg_ctxt - write a device's per-ctxt 64-bit kernel register
+ * @dd: the qlogic_ib device
+ * @regno: the register number to write
+ * @ctxt: the context containing the register
+ * @value: the value to write
+ */
+static inline void qib_write_kreg_ctxt(const struct qib_devdata *dd,
+				       const u16 regno, unsigned ctxt,
+				       u64 value)
+{
+	qib_write_kreg(dd, regno + ctxt, value);
+}
+
+static inline u64 read_7322_creg(const struct qib_devdata *dd, u16 regno)
+{
+	if (!dd->cspec->cregbase || !(dd->flags & QIB_PRESENT))
+		return 0;
+	return readq(&dd->cspec->cregbase[regno]);
+
+
+}
+
+static inline u32 read_7322_creg32(const struct qib_devdata *dd, u16 regno)
+{
+	if (!dd->cspec->cregbase || !(dd->flags & QIB_PRESENT))
+		return 0;
+	return readl(&dd->cspec->cregbase[regno]);
+
+
+}
+
+static inline void write_7322_creg_port(const struct qib_pportdata *ppd,
+					u16 regno, u64 value)
+{
+	if (ppd->cpspec && ppd->cpspec->cpregbase &&
+	    (ppd->dd->flags & QIB_PRESENT))
+		writeq(value, &ppd->cpspec->cpregbase[regno]);
+}
+
+static inline u64 read_7322_creg_port(const struct qib_pportdata *ppd,
+				      u16 regno)
+{
+	if (!ppd->cpspec || !ppd->cpspec->cpregbase ||
+	    !(ppd->dd->flags & QIB_PRESENT))
+		return 0;
+	return readq(&ppd->cpspec->cpregbase[regno]);
+}
+
+static inline u32 read_7322_creg32_port(const struct qib_pportdata *ppd,
+					u16 regno)
+{
+	if (!ppd->cpspec || !ppd->cpspec->cpregbase ||
+	    !(ppd->dd->flags & QIB_PRESENT))
+		return 0;
+	return readl(&ppd->cpspec->cpregbase[regno]);
+}
+
+/* bits in Control register */
+#define QLOGIC_IB_C_RESET SYM_MASK(Control, SyncReset)
+#define QLOGIC_IB_C_SDMAFETCHPRIOEN SYM_MASK(Control, SDmaDescFetchPriorityEn)
+
+/* bits in general interrupt regs */
+#define QIB_I_RCVURG_LSB SYM_LSB(IntMask, RcvUrg0IntMask)
+#define QIB_I_RCVURG_RMASK MASK_ACROSS(0, 17)
+#define QIB_I_RCVURG_MASK (QIB_I_RCVURG_RMASK << QIB_I_RCVURG_LSB)
+#define QIB_I_RCVAVAIL_LSB SYM_LSB(IntMask, RcvAvail0IntMask)
+#define QIB_I_RCVAVAIL_RMASK MASK_ACROSS(0, 17)
+#define QIB_I_RCVAVAIL_MASK (QIB_I_RCVAVAIL_RMASK << QIB_I_RCVAVAIL_LSB)
+#define QIB_I_C_ERROR INT_MASK(Err)
+
+#define QIB_I_SPIOSENT (INT_MASK_P(SendDone, 0) | INT_MASK_P(SendDone, 1))
+#define QIB_I_SPIOBUFAVAIL INT_MASK(SendBufAvail)
+#define QIB_I_GPIO INT_MASK(AssertGPIO)
+#define QIB_I_P_SDMAINT(pidx) \
+	(INT_MASK_P(SDma, pidx) | INT_MASK_P(SDmaIdle, pidx) | \
+	 INT_MASK_P(SDmaProgress, pidx) | \
+	 INT_MASK_PM(SDmaCleanupDone, pidx))
+
+/* Interrupt bits that are "per port" */
+#define QIB_I_P_BITSEXTANT(pidx) \
+	(INT_MASK_P(Err, pidx) | INT_MASK_P(SendDone, pidx) | \
+	INT_MASK_P(SDma, pidx) | INT_MASK_P(SDmaIdle, pidx) | \
+	INT_MASK_P(SDmaProgress, pidx) | \
+	INT_MASK_PM(SDmaCleanupDone, pidx))
+
+/* Interrupt bits that are common to a device */
+/* currently unused: QIB_I_SPIOSENT */
+#define QIB_I_C_BITSEXTANT \
+	(QIB_I_RCVURG_MASK | QIB_I_RCVAVAIL_MASK | \
+	QIB_I_SPIOSENT | \
+	QIB_I_C_ERROR | QIB_I_SPIOBUFAVAIL | QIB_I_GPIO)
+
+#define QIB_I_BITSEXTANT (QIB_I_C_BITSEXTANT | \
+	QIB_I_P_BITSEXTANT(0) | QIB_I_P_BITSEXTANT(1))
+
+/*
+ * Error bits that are "per port".
+ */
+#define QIB_E_P_IBSTATUSCHANGED ERR_MASK_N(IBStatusChanged)
+#define QIB_E_P_SHDR ERR_MASK_N(SHeadersErr)
+#define QIB_E_P_VL15_BUF_MISUSE ERR_MASK_N(VL15BufMisuseErr)
+#define QIB_E_P_SND_BUF_MISUSE ERR_MASK_N(SendBufMisuseErr)
+#define QIB_E_P_SUNSUPVL ERR_MASK_N(SendUnsupportedVLErr)
+#define QIB_E_P_SUNEXP_PKTNUM ERR_MASK_N(SendUnexpectedPktNumErr)
+#define QIB_E_P_SDROP_DATA ERR_MASK_N(SendDroppedDataPktErr)
+#define QIB_E_P_SDROP_SMP ERR_MASK_N(SendDroppedSmpPktErr)
+#define QIB_E_P_SPKTLEN ERR_MASK_N(SendPktLenErr)
+#define QIB_E_P_SUNDERRUN ERR_MASK_N(SendUnderRunErr)
+#define QIB_E_P_SMAXPKTLEN ERR_MASK_N(SendMaxPktLenErr)
+#define QIB_E_P_SMINPKTLEN ERR_MASK_N(SendMinPktLenErr)
+#define QIB_E_P_RIBLOSTLINK ERR_MASK_N(RcvIBLostLinkErr)
+#define QIB_E_P_RHDR ERR_MASK_N(RcvHdrErr)
+#define QIB_E_P_RHDRLEN ERR_MASK_N(RcvHdrLenErr)
+#define QIB_E_P_RBADTID ERR_MASK_N(RcvBadTidErr)
+#define QIB_E_P_RBADVERSION ERR_MASK_N(RcvBadVersionErr)
+#define QIB_E_P_RIBFLOW ERR_MASK_N(RcvIBFlowErr)
+#define QIB_E_P_REBP ERR_MASK_N(RcvEBPErr)
+#define QIB_E_P_RUNSUPVL ERR_MASK_N(RcvUnsupportedVLErr)
+#define QIB_E_P_RUNEXPCHAR ERR_MASK_N(RcvUnexpectedCharErr)
+#define QIB_E_P_RSHORTPKTLEN ERR_MASK_N(RcvShortPktLenErr)
+#define QIB_E_P_RLONGPKTLEN ERR_MASK_N(RcvLongPktLenErr)
+#define QIB_E_P_RMAXPKTLEN ERR_MASK_N(RcvMaxPktLenErr)
+#define QIB_E_P_RMINPKTLEN ERR_MASK_N(RcvMinPktLenErr)
+#define QIB_E_P_RICRC ERR_MASK_N(RcvICRCErr)
+#define QIB_E_P_RVCRC ERR_MASK_N(RcvVCRCErr)
+#define QIB_E_P_RFORMATERR ERR_MASK_N(RcvFormatErr)
+
+#define QIB_E_P_SDMA1STDESC ERR_MASK_N(SDma1stDescErr)
+#define QIB_E_P_SDMABASE ERR_MASK_N(SDmaBaseErr)
+#define QIB_E_P_SDMADESCADDRMISALIGN ERR_MASK_N(SDmaDescAddrMisalignErr)
+#define QIB_E_P_SDMADWEN ERR_MASK_N(SDmaDwEnErr)
+#define QIB_E_P_SDMAGENMISMATCH ERR_MASK_N(SDmaGenMismatchErr)
+#define QIB_E_P_SDMAHALT ERR_MASK_N(SDmaHaltErr)
+#define QIB_E_P_SDMAMISSINGDW ERR_MASK_N(SDmaMissingDwErr)
+#define QIB_E_P_SDMAOUTOFBOUND ERR_MASK_N(SDmaOutOfBoundErr)
+#define QIB_E_P_SDMARPYTAG ERR_MASK_N(SDmaRpyTagErr)
+#define QIB_E_P_SDMATAILOUTOFBOUND ERR_MASK_N(SDmaTailOutOfBoundErr)
+#define QIB_E_P_SDMAUNEXPDATA ERR_MASK_N(SDmaUnexpDataErr)
+
+/* Error bits that are common to a device */
+#define QIB_E_RESET ERR_MASK(ResetNegated)
+#define QIB_E_HARDWARE ERR_MASK(HardwareErr)
+#define QIB_E_INVALIDADDR ERR_MASK(InvalidAddrErr)
+
+
+/*
+ * Per chip (rather than per-port) errors.  Most either do
+ * nothing but trigger a print (because they self-recover, or
+ * always occur in tandem with other errors that handle the
+ * issue), or because they indicate errors with no recovery,
+ * but we want to know that they happened.
+ */
+#define QIB_E_SBUF_VL15_MISUSE ERR_MASK(SBufVL15MisUseErr)
+#define QIB_E_BADEEP ERR_MASK(InvalidEEPCmd)
+#define QIB_E_VLMISMATCH ERR_MASK(SendVLMismatchErr)
+#define QIB_E_ARMLAUNCH ERR_MASK(SendArmLaunchErr)
+#define QIB_E_SPCLTRIG ERR_MASK(SendSpecialTriggerErr)
+#define QIB_E_RRCVHDRFULL ERR_MASK(RcvHdrFullErr)
+#define QIB_E_RRCVEGRFULL ERR_MASK(RcvEgrFullErr)
+#define QIB_E_RCVCTXTSHARE ERR_MASK(RcvContextShareErr)
+
+/* SDMA chip errors (not per port)
+ * QIB_E_SDMA_BUF_DUP needs no special handling, because we will also get
+ * the SDMAHALT error immediately, so we just print the dup error via the
+ * E_AUTO mechanism.  This is true of most of the per-port fatal errors
+ * as well, but since this is port-independent, by definition, it's
+ * handled a bit differently.  SDMA_VL15 and SDMA_WRONG_PORT are per
+ * packet send errors, and so are handled in the same manner as other
+ * per-packet errors.
+ */
+#define QIB_E_SDMA_VL15 ERR_MASK(SDmaVL15Err)
+#define QIB_E_SDMA_WRONG_PORT ERR_MASK(SDmaWrongPortErr)
+#define QIB_E_SDMA_BUF_DUP ERR_MASK(SDmaBufMaskDuplicateErr)
+
+/*
+ * Below functionally equivalent to legacy QLOGIC_IB_E_PKTERRS
+ * it is used to print "common" packet errors.
+ */
+#define QIB_E_P_PKTERRS (QIB_E_P_SPKTLEN |\
+	QIB_E_P_SDROP_DATA | QIB_E_P_RVCRC |\
+	QIB_E_P_RICRC | QIB_E_P_RSHORTPKTLEN |\
+	QIB_E_P_VL15_BUF_MISUSE | QIB_E_P_SHDR | \
+	QIB_E_P_REBP)
+
+/* Error Bits that Packet-related (Receive, per-port) */
+#define QIB_E_P_RPKTERRS (\
+	QIB_E_P_RHDRLEN | QIB_E_P_RBADTID | \
+	QIB_E_P_RBADVERSION | QIB_E_P_RHDR | \
+	QIB_E_P_RLONGPKTLEN | QIB_E_P_RSHORTPKTLEN |\
+	QIB_E_P_RMAXPKTLEN | QIB_E_P_RMINPKTLEN | \
+	QIB_E_P_RFORMATERR | QIB_E_P_RUNSUPVL | \
+	QIB_E_P_RUNEXPCHAR | QIB_E_P_RIBFLOW | QIB_E_P_REBP)
+
+/*
+ * Error bits that are Send-related (per port)
+ * (ARMLAUNCH excluded from E_SPKTERRS because it gets special handling).
+ * All of these potentially need to have a buffer disarmed
+ */
+#define QIB_E_P_SPKTERRS (\
+	QIB_E_P_SUNEXP_PKTNUM |\
+	QIB_E_P_SDROP_DATA | QIB_E_P_SDROP_SMP |\
+	QIB_E_P_SMAXPKTLEN |\
+	QIB_E_P_VL15_BUF_MISUSE | QIB_E_P_SHDR | \
+	QIB_E_P_SMINPKTLEN | QIB_E_P_SPKTLEN | \
+	QIB_E_P_SND_BUF_MISUSE | QIB_E_P_SUNSUPVL)
+
+#define QIB_E_SPKTERRS ( \
+		QIB_E_SBUF_VL15_MISUSE | QIB_E_VLMISMATCH | \
+		ERR_MASK_N(SendUnsupportedVLErr) |			\
+		QIB_E_SPCLTRIG | QIB_E_SDMA_VL15 | QIB_E_SDMA_WRONG_PORT)
+
+#define QIB_E_P_SDMAERRS ( \
+	QIB_E_P_SDMAHALT | \
+	QIB_E_P_SDMADESCADDRMISALIGN | \
+	QIB_E_P_SDMAUNEXPDATA | \
+	QIB_E_P_SDMAMISSINGDW | \
+	QIB_E_P_SDMADWEN | \
+	QIB_E_P_SDMARPYTAG | \
+	QIB_E_P_SDMA1STDESC | \
+	QIB_E_P_SDMABASE | \
+	QIB_E_P_SDMATAILOUTOFBOUND | \
+	QIB_E_P_SDMAOUTOFBOUND | \
+	QIB_E_P_SDMAGENMISMATCH)
+
+/*
+ * This sets some bits more than once, but makes it more obvious which
+ * bits are not handled under other categories, and the repeat definition
+ * is not a problem.
+ */
+#define QIB_E_P_BITSEXTANT ( \
+	QIB_E_P_SPKTERRS | QIB_E_P_PKTERRS | QIB_E_P_RPKTERRS | \
+	QIB_E_P_RIBLOSTLINK | QIB_E_P_IBSTATUSCHANGED | \
+	QIB_E_P_SND_BUF_MISUSE | QIB_E_P_SUNDERRUN | \
+	QIB_E_P_SHDR | QIB_E_P_VL15_BUF_MISUSE | QIB_E_P_SDMAERRS \
+	)
+
+/*
+ * These are errors that can occur when the link
+ * changes state while a packet is being sent or received.  This doesn't
+ * cover things like EBP or VCRC that can be the result of a sending
+ * having the link change state, so we receive a "known bad" packet.
+ * All of these are "per port", so renamed:
+ */
+#define QIB_E_P_LINK_PKTERRS (\
+	QIB_E_P_SDROP_DATA | QIB_E_P_SDROP_SMP |\
+	QIB_E_P_SMINPKTLEN | QIB_E_P_SPKTLEN |\
+	QIB_E_P_RSHORTPKTLEN | QIB_E_P_RMINPKTLEN |\
+	QIB_E_P_RUNEXPCHAR)
+
+/*
+ * This sets some bits more than once, but makes it more obvious which
+ * bits are not handled under other categories (such as QIB_E_SPKTERRS),
+ * and the repeat definition is not a problem.
+ */
+#define QIB_E_C_BITSEXTANT (\
+	QIB_E_HARDWARE | QIB_E_INVALIDADDR | QIB_E_BADEEP |\
+	QIB_E_ARMLAUNCH | QIB_E_VLMISMATCH | QIB_E_RRCVHDRFULL |\
+	QIB_E_RRCVEGRFULL | QIB_E_RESET | QIB_E_SBUF_VL15_MISUSE)
+
+/* Likewise Neuter E_SPKT_ERRS_IGNORE */
+#define E_SPKT_ERRS_IGNORE 0
+
+#define QIB_EXTS_MEMBIST_DISABLED \
+	SYM_MASK(EXTStatus, MemBISTDisabled)
+#define QIB_EXTS_MEMBIST_ENDTEST \
+	SYM_MASK(EXTStatus, MemBISTEndTest)
+
+#define QIB_E_SPIOARMLAUNCH \
+	ERR_MASK(SendArmLaunchErr)
+
+#define IBA7322_IBCC_LINKINITCMD_MASK SYM_RMASK(IBCCtrlA_0, LinkInitCmd)
+#define IBA7322_IBCC_LINKCMD_SHIFT SYM_LSB(IBCCtrlA_0, LinkCmd)
+
+/*
+ * IBTA_1_2 is set when multiple speeds are enabled (normal),
+ * and also if forced QDR (only QDR enabled).  It's enabled for the
+ * forced QDR case so that scrambling will be enabled by the TS3
+ * exchange, when supported by both sides of the link.
+ */
+#define IBA7322_IBC_IBTA_1_2_MASK SYM_MASK(IBCCtrlB_0, IB_ENHANCED_MODE)
+#define IBA7322_IBC_MAX_SPEED_MASK SYM_MASK(IBCCtrlB_0, SD_SPEED)
+#define IBA7322_IBC_SPEED_QDR SYM_MASK(IBCCtrlB_0, SD_SPEED_QDR)
+#define IBA7322_IBC_SPEED_DDR SYM_MASK(IBCCtrlB_0, SD_SPEED_DDR)
+#define IBA7322_IBC_SPEED_SDR SYM_MASK(IBCCtrlB_0, SD_SPEED_SDR)
+#define IBA7322_IBC_SPEED_MASK (SYM_MASK(IBCCtrlB_0, SD_SPEED_SDR) | \
+	SYM_MASK(IBCCtrlB_0, SD_SPEED_DDR) | SYM_MASK(IBCCtrlB_0, SD_SPEED_QDR))
+#define IBA7322_IBC_SPEED_LSB SYM_LSB(IBCCtrlB_0, SD_SPEED_SDR)
+
+#define IBA7322_LEDBLINK_OFF_SHIFT SYM_LSB(RcvPktLEDCnt_0, OFFperiod)
+#define IBA7322_LEDBLINK_ON_SHIFT SYM_LSB(RcvPktLEDCnt_0, ONperiod)
+
+#define IBA7322_IBC_WIDTH_AUTONEG SYM_MASK(IBCCtrlB_0, IB_NUM_CHANNELS)
+#define IBA7322_IBC_WIDTH_4X_ONLY (1<<SYM_LSB(IBCCtrlB_0, IB_NUM_CHANNELS))
+#define IBA7322_IBC_WIDTH_1X_ONLY (0<<SYM_LSB(IBCCtrlB_0, IB_NUM_CHANNELS))
+
+#define IBA7322_IBC_RXPOL_MASK SYM_MASK(IBCCtrlB_0, IB_POLARITY_REV_SUPP)
+#define IBA7322_IBC_RXPOL_LSB SYM_LSB(IBCCtrlB_0, IB_POLARITY_REV_SUPP)
+#define IBA7322_IBC_HRTBT_MASK (SYM_MASK(IBCCtrlB_0, HRTBT_AUTO) | \
+	SYM_MASK(IBCCtrlB_0, HRTBT_ENB))
+#define IBA7322_IBC_HRTBT_RMASK (IBA7322_IBC_HRTBT_MASK >> \
+	SYM_LSB(IBCCtrlB_0, HRTBT_ENB))
+#define IBA7322_IBC_HRTBT_LSB SYM_LSB(IBCCtrlB_0, HRTBT_ENB)
+
+#define IBA7322_REDIRECT_VEC_PER_REG 12
+
+#define IBA7322_SENDCHK_PKEY SYM_MASK(SendCheckControl_0, PKey_En)
+#define IBA7322_SENDCHK_BTHQP SYM_MASK(SendCheckControl_0, BTHQP_En)
+#define IBA7322_SENDCHK_SLID SYM_MASK(SendCheckControl_0, SLID_En)
+#define IBA7322_SENDCHK_RAW_IPV6 SYM_MASK(SendCheckControl_0, RawIPV6_En)
+#define IBA7322_SENDCHK_MINSZ SYM_MASK(SendCheckControl_0, PacketTooSmall_En)
+
+#define AUTONEG_TRIES 3 /* sequential retries to negotiate DDR */
+
+#define HWE_AUTO(fldname) { .mask = SYM_MASK(HwErrMask, fldname##Mask), \
+	.msg = #fldname }
+#define HWE_AUTO_P(fldname, port) { .mask = SYM_MASK(HwErrMask, \
+	fldname##Mask##_##port), .msg = #fldname }
+static const struct qib_hwerror_msgs qib_7322_hwerror_msgs[] = {
+	HWE_AUTO_P(IBSerdesPClkNotDetect, 1),
+	HWE_AUTO_P(IBSerdesPClkNotDetect, 0),
+	HWE_AUTO(PCIESerdesPClkNotDetect),
+	HWE_AUTO(PowerOnBISTFailed),
+	HWE_AUTO(TempsenseTholdReached),
+	HWE_AUTO(MemoryErr),
+	HWE_AUTO(PCIeBusParityErr),
+	HWE_AUTO(PcieCplTimeout),
+	HWE_AUTO(PciePoisonedTLP),
+	HWE_AUTO_P(SDmaMemReadErr, 1),
+	HWE_AUTO_P(SDmaMemReadErr, 0),
+	HWE_AUTO_P(IBCBusFromSPCParityErr, 1),
+	HWE_AUTO_P(IBCBusFromSPCParityErr, 0),
+	HWE_AUTO_P(statusValidNoEop, 1),
+	HWE_AUTO_P(statusValidNoEop, 0),
+	HWE_AUTO(LATriggered),
+	{ .mask = 0 }
+};
+
+#define E_AUTO(fldname) { .mask = SYM_MASK(ErrMask, fldname##Mask), \
+	.msg = #fldname }
+#define E_P_AUTO(fldname) { .mask = SYM_MASK(ErrMask_0, fldname##Mask), \
+	.msg = #fldname }
+static const struct qib_hwerror_msgs qib_7322error_msgs[] = {
+	E_AUTO(ResetNegated),
+	E_AUTO(HardwareErr),
+	E_AUTO(InvalidAddrErr),
+	E_AUTO(SDmaVL15Err),
+	E_AUTO(SBufVL15MisUseErr),
+	E_AUTO(InvalidEEPCmd),
+	E_AUTO(RcvContextShareErr),
+	E_AUTO(SendVLMismatchErr),
+	E_AUTO(SendArmLaunchErr),
+	E_AUTO(SendSpecialTriggerErr),
+	E_AUTO(SDmaWrongPortErr),
+	E_AUTO(SDmaBufMaskDuplicateErr),
+	E_AUTO(RcvHdrFullErr),
+	E_AUTO(RcvEgrFullErr),
+	{ .mask = 0 }
+};
+
+static const struct  qib_hwerror_msgs qib_7322p_error_msgs[] = {
+	E_P_AUTO(IBStatusChanged),
+	E_P_AUTO(SHeadersErr),
+	E_P_AUTO(VL15BufMisuseErr),
+	/*
+	 * SDmaHaltErr is not really an error, make it clearer;
+	 */
+	{.mask = SYM_MASK(ErrMask_0, SDmaHaltErrMask), .msg = "SDmaHalted"},
+	E_P_AUTO(SDmaDescAddrMisalignErr),
+	E_P_AUTO(SDmaUnexpDataErr),
+	E_P_AUTO(SDmaMissingDwErr),
+	E_P_AUTO(SDmaDwEnErr),
+	E_P_AUTO(SDmaRpyTagErr),
+	E_P_AUTO(SDma1stDescErr),
+	E_P_AUTO(SDmaBaseErr),
+	E_P_AUTO(SDmaTailOutOfBoundErr),
+	E_P_AUTO(SDmaOutOfBoundErr),
+	E_P_AUTO(SDmaGenMismatchErr),
+	E_P_AUTO(SendBufMisuseErr),
+	E_P_AUTO(SendUnsupportedVLErr),
+	E_P_AUTO(SendUnexpectedPktNumErr),
+	E_P_AUTO(SendDroppedDataPktErr),
+	E_P_AUTO(SendDroppedSmpPktErr),
+	E_P_AUTO(SendPktLenErr),
+	E_P_AUTO(SendUnderRunErr),
+	E_P_AUTO(SendMaxPktLenErr),
+	E_P_AUTO(SendMinPktLenErr),
+	E_P_AUTO(RcvIBLostLinkErr),
+	E_P_AUTO(RcvHdrErr),
+	E_P_AUTO(RcvHdrLenErr),
+	E_P_AUTO(RcvBadTidErr),
+	E_P_AUTO(RcvBadVersionErr),
+	E_P_AUTO(RcvIBFlowErr),
+	E_P_AUTO(RcvEBPErr),
+	E_P_AUTO(RcvUnsupportedVLErr),
+	E_P_AUTO(RcvUnexpectedCharErr),
+	E_P_AUTO(RcvShortPktLenErr),
+	E_P_AUTO(RcvLongPktLenErr),
+	E_P_AUTO(RcvMaxPktLenErr),
+	E_P_AUTO(RcvMinPktLenErr),
+	E_P_AUTO(RcvICRCErr),
+	E_P_AUTO(RcvVCRCErr),
+	E_P_AUTO(RcvFormatErr),
+	{ .mask = 0 }
+};
+
+/*
+ * Below generates "auto-message" for interrupts not specific to any port or
+ * context
+ */
+#define INTR_AUTO(fldname) { .mask = SYM_MASK(IntMask, fldname##Mask), \
+	.msg = #fldname }
+/* Below generates "auto-message" for interrupts specific to a port */
+#define INTR_AUTO_P(fldname) { .mask = MASK_ACROSS(\
+	SYM_LSB(IntMask, fldname##Mask##_0), \
+	SYM_LSB(IntMask, fldname##Mask##_1)), \
+	.msg = #fldname "_P" }
+/* For some reason, the SerDesTrimDone bits are reversed */
+#define INTR_AUTO_PI(fldname) { .mask = MASK_ACROSS(\
+	SYM_LSB(IntMask, fldname##Mask##_1), \
+	SYM_LSB(IntMask, fldname##Mask##_0)), \
+	.msg = #fldname "_P" }
+/*
+ * Below generates "auto-message" for interrupts specific to a context,
+ * with ctxt-number appended
+ */
+#define INTR_AUTO_C(fldname) { .mask = MASK_ACROSS(\
+	SYM_LSB(IntMask, fldname##0IntMask), \
+	SYM_LSB(IntMask, fldname##17IntMask)), \
+	.msg = #fldname "_C"}
+
+static const struct  qib_hwerror_msgs qib_7322_intr_msgs[] = {
+	INTR_AUTO_P(SDmaInt),
+	INTR_AUTO_P(SDmaProgressInt),
+	INTR_AUTO_P(SDmaIdleInt),
+	INTR_AUTO_P(SDmaCleanupDone),
+	INTR_AUTO_C(RcvUrg),
+	INTR_AUTO_P(ErrInt),
+	INTR_AUTO(ErrInt),      /* non-port-specific errs */
+	INTR_AUTO(AssertGPIOInt),
+	INTR_AUTO_P(SendDoneInt),
+	INTR_AUTO(SendBufAvailInt),
+	INTR_AUTO_C(RcvAvail),
+	{ .mask = 0 }
+};
+
+#define TXSYMPTOM_AUTO_P(fldname) \
+	{ .mask = SYM_MASK(SendHdrErrSymptom_0, fldname), .msg = #fldname }
+static const struct  qib_hwerror_msgs hdrchk_msgs[] = {
+	TXSYMPTOM_AUTO_P(NonKeyPacket),
+	TXSYMPTOM_AUTO_P(GRHFail),
+	TXSYMPTOM_AUTO_P(PkeyFail),
+	TXSYMPTOM_AUTO_P(QPFail),
+	TXSYMPTOM_AUTO_P(SLIDFail),
+	TXSYMPTOM_AUTO_P(RawIPV6),
+	TXSYMPTOM_AUTO_P(PacketTooSmall),
+	{ .mask = 0 }
+};
+
+#define IBA7322_HDRHEAD_PKTINT_SHIFT 32 /* interrupt cnt in upper 32 bits */
+
+/*
+ * Called when we might have an error that is specific to a particular
+ * PIO buffer, and may need to cancel that buffer, so it can be re-used,
+ * because we don't need to force the update of pioavail
+ */
+static void qib_disarm_7322_senderrbufs(struct qib_pportdata *ppd)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u32 i;
+	int any;
+	u32 piobcnt = dd->piobcnt2k + dd->piobcnt4k + NUM_VL15_BUFS;
+	u32 regcnt = (piobcnt + BITS_PER_LONG - 1) / BITS_PER_LONG;
+	unsigned long sbuf[4];
+
+	/*
+	 * It's possible that sendbuffererror could have bits set; might
+	 * have already done this as a result of hardware error handling.
+	 */
+	any = 0;
+	for (i = 0; i < regcnt; ++i) {
+		sbuf[i] = qib_read_kreg64(dd, kr_sendbuffererror + i);
+		if (sbuf[i]) {
+			any = 1;
+			qib_write_kreg(dd, kr_sendbuffererror + i, sbuf[i]);
+		}
+	}
+
+	if (any)
+		qib_disarm_piobufs_set(dd, sbuf, piobcnt);
+}
+
+/* No txe_recover yet, if ever */
+
+/* No decode__errors yet */
+static void err_decode(char *msg, size_t len, u64 errs,
+		       const struct qib_hwerror_msgs *msp)
+{
+	u64 these, lmask;
+	int took, multi, n = 0;
+
+	while (msp && msp->mask) {
+		multi = (msp->mask & (msp->mask - 1));
+		while (errs & msp->mask) {
+			these = (errs & msp->mask);
+			lmask = (these & (these - 1)) ^ these;
+			if (len) {
+				if (n++) {
+					/* separate the strings */
+					*msg++ = ',';
+					len--;
+				}
+				took = scnprintf(msg, len, "%s", msp->msg);
+				len -= took;
+				msg += took;
+			}
+			errs &= ~lmask;
+			if (len && multi) {
+				/* More than one bit this mask */
+				int idx = -1;
+
+				while (lmask & msp->mask) {
+					++idx;
+					lmask >>= 1;
+				}
+				took = scnprintf(msg, len, "_%d", idx);
+				len -= took;
+				msg += took;
+			}
+		}
+		++msp;
+	}
+	/* If some bits are left, show in hex. */
+	if (len && errs)
+		snprintf(msg, len, "%sMORE:%llX", n ? "," : "",
+			(unsigned long long) errs);
+}
+
+/* only called if r1 set */
+static void flush_fifo(struct qib_pportdata *ppd)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u32 __iomem *piobuf;
+	u32 bufn;
+	u32 *hdr;
+	u64 pbc;
+	const unsigned hdrwords = 7;
+	static struct qib_ib_header ibhdr = {
+		.lrh[0] = cpu_to_be16(0xF000 | QIB_LRH_BTH),
+		.lrh[1] = IB_LID_PERMISSIVE,
+		.lrh[2] = cpu_to_be16(hdrwords + SIZE_OF_CRC),
+		.lrh[3] = IB_LID_PERMISSIVE,
+		.u.oth.bth[0] = cpu_to_be32(
+			(IB_OPCODE_UD_SEND_ONLY << 24) | QIB_DEFAULT_P_KEY),
+		.u.oth.bth[1] = cpu_to_be32(0),
+		.u.oth.bth[2] = cpu_to_be32(0),
+		.u.oth.u.ud.deth[0] = cpu_to_be32(0),
+		.u.oth.u.ud.deth[1] = cpu_to_be32(0),
+	};
+
+	/*
+	 * Send a dummy VL15 packet to flush the launch FIFO.
+	 * This will not actually be sent since the TxeBypassIbc bit is set.
+	 */
+	pbc = PBC_7322_VL15_SEND |
+		(((u64)ppd->hw_pidx) << (PBC_PORT_SEL_LSB + 32)) |
+		(hdrwords + SIZE_OF_CRC);
+	piobuf = qib_7322_getsendbuf(ppd, pbc, &bufn);
+	if (!piobuf)
+		return;
+	writeq(pbc, piobuf);
+	hdr = (u32 *) &ibhdr;
+	if (dd->flags & QIB_PIO_FLUSH_WC) {
+		qib_flush_wc();
+		qib_pio_copy(piobuf + 2, hdr, hdrwords - 1);
+		qib_flush_wc();
+		__raw_writel(hdr[hdrwords - 1], piobuf + hdrwords + 1);
+		qib_flush_wc();
+	} else
+		qib_pio_copy(piobuf + 2, hdr, hdrwords);
+	qib_sendbuf_done(dd, bufn);
+}
+
+/*
+ * This is called with interrupts disabled and sdma_lock held.
+ */
+static void qib_7322_sdma_sendctrl(struct qib_pportdata *ppd, unsigned op)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u64 set_sendctrl = 0;
+	u64 clr_sendctrl = 0;
+
+	if (op & QIB_SDMA_SENDCTRL_OP_ENABLE)
+		set_sendctrl |= SYM_MASK(SendCtrl_0, SDmaEnable);
+	else
+		clr_sendctrl |= SYM_MASK(SendCtrl_0, SDmaEnable);
+
+	if (op & QIB_SDMA_SENDCTRL_OP_INTENABLE)
+		set_sendctrl |= SYM_MASK(SendCtrl_0, SDmaIntEnable);
+	else
+		clr_sendctrl |= SYM_MASK(SendCtrl_0, SDmaIntEnable);
+
+	if (op & QIB_SDMA_SENDCTRL_OP_HALT)
+		set_sendctrl |= SYM_MASK(SendCtrl_0, SDmaHalt);
+	else
+		clr_sendctrl |= SYM_MASK(SendCtrl_0, SDmaHalt);
+
+	if (op & QIB_SDMA_SENDCTRL_OP_DRAIN)
+		set_sendctrl |= SYM_MASK(SendCtrl_0, TxeBypassIbc) |
+				SYM_MASK(SendCtrl_0, TxeAbortIbc) |
+				SYM_MASK(SendCtrl_0, TxeDrainRmFifo);
+	else
+		clr_sendctrl |= SYM_MASK(SendCtrl_0, TxeBypassIbc) |
+				SYM_MASK(SendCtrl_0, TxeAbortIbc) |
+				SYM_MASK(SendCtrl_0, TxeDrainRmFifo);
+
+	spin_lock(&dd->sendctrl_lock);
+
+	/* If we are draining everything, block sends first */
+	if (op & QIB_SDMA_SENDCTRL_OP_DRAIN) {
+		ppd->p_sendctrl &= ~SYM_MASK(SendCtrl_0, SendEnable);
+		qib_write_kreg_port(ppd, krp_sendctrl, ppd->p_sendctrl);
+		qib_write_kreg(dd, kr_scratch, 0);
+	}
+
+	ppd->p_sendctrl |= set_sendctrl;
+	ppd->p_sendctrl &= ~clr_sendctrl;
+
+	if (op & QIB_SDMA_SENDCTRL_OP_CLEANUP)
+		qib_write_kreg_port(ppd, krp_sendctrl,
+				    ppd->p_sendctrl |
+				    SYM_MASK(SendCtrl_0, SDmaCleanup));
+	else
+		qib_write_kreg_port(ppd, krp_sendctrl, ppd->p_sendctrl);
+	qib_write_kreg(dd, kr_scratch, 0);
+
+	if (op & QIB_SDMA_SENDCTRL_OP_DRAIN) {
+		ppd->p_sendctrl |= SYM_MASK(SendCtrl_0, SendEnable);
+		qib_write_kreg_port(ppd, krp_sendctrl, ppd->p_sendctrl);
+		qib_write_kreg(dd, kr_scratch, 0);
+	}
+
+	spin_unlock(&dd->sendctrl_lock);
+
+	if ((op & QIB_SDMA_SENDCTRL_OP_DRAIN) && ppd->dd->cspec->r1)
+		flush_fifo(ppd);
+}
+
+static void qib_7322_sdma_hw_clean_up(struct qib_pportdata *ppd)
+{
+	__qib_sdma_process_event(ppd, qib_sdma_event_e50_hw_cleaned);
+}
+
+static void qib_sdma_7322_setlengen(struct qib_pportdata *ppd)
+{
+	/*
+	 * Set SendDmaLenGen and clear and set
+	 * the MSB of the generation count to enable generation checking
+	 * and load the internal generation counter.
+	 */
+	qib_write_kreg_port(ppd, krp_senddmalengen, ppd->sdma_descq_cnt);
+	qib_write_kreg_port(ppd, krp_senddmalengen,
+			    ppd->sdma_descq_cnt |
+			    (1ULL << QIB_7322_SendDmaLenGen_0_Generation_MSB));
+}
+
+/*
+ * Must be called with sdma_lock held, or before init finished.
+ */
+static void qib_sdma_update_7322_tail(struct qib_pportdata *ppd, u16 tail)
+{
+	/* Commit writes to memory and advance the tail on the chip */
+	wmb();
+	ppd->sdma_descq_tail = tail;
+	qib_write_kreg_port(ppd, krp_senddmatail, tail);
+}
+
+/*
+ * This is called with interrupts disabled and sdma_lock held.
+ */
+static void qib_7322_sdma_hw_start_up(struct qib_pportdata *ppd)
+{
+	/*
+	 * Drain all FIFOs.
+	 * The hardware doesn't require this but we do it so that verbs
+	 * and user applications don't wait for link active to send stale
+	 * data.
+	 */
+	sendctrl_7322_mod(ppd, QIB_SENDCTRL_FLUSH);
+
+	qib_sdma_7322_setlengen(ppd);
+	qib_sdma_update_7322_tail(ppd, 0); /* Set SendDmaTail */
+	ppd->sdma_head_dma[0] = 0;
+	qib_7322_sdma_sendctrl(ppd,
+		ppd->sdma_state.current_op | QIB_SDMA_SENDCTRL_OP_CLEANUP);
+}
+
+#define DISABLES_SDMA ( \
+	QIB_E_P_SDMAHALT | \
+	QIB_E_P_SDMADESCADDRMISALIGN | \
+	QIB_E_P_SDMAMISSINGDW | \
+	QIB_E_P_SDMADWEN | \
+	QIB_E_P_SDMARPYTAG | \
+	QIB_E_P_SDMA1STDESC | \
+	QIB_E_P_SDMABASE | \
+	QIB_E_P_SDMATAILOUTOFBOUND | \
+	QIB_E_P_SDMAOUTOFBOUND | \
+	QIB_E_P_SDMAGENMISMATCH)
+
+static void sdma_7322_p_errors(struct qib_pportdata *ppd, u64 errs)
+{
+	unsigned long flags;
+	struct qib_devdata *dd = ppd->dd;
+
+	errs &= QIB_E_P_SDMAERRS;
+
+	if (errs & QIB_E_P_SDMAUNEXPDATA)
+		qib_dev_err(dd, "IB%u:%u SDmaUnexpData\n", dd->unit,
+			    ppd->port);
+
+	spin_lock_irqsave(&ppd->sdma_lock, flags);
+
+	switch (ppd->sdma_state.current_state) {
+	case qib_sdma_state_s00_hw_down:
+		break;
+
+	case qib_sdma_state_s10_hw_start_up_wait:
+		if (errs & QIB_E_P_SDMAHALT)
+			__qib_sdma_process_event(ppd,
+				qib_sdma_event_e20_hw_started);
+		break;
+
+	case qib_sdma_state_s20_idle:
+		break;
+
+	case qib_sdma_state_s30_sw_clean_up_wait:
+		break;
+
+	case qib_sdma_state_s40_hw_clean_up_wait:
+		if (errs & QIB_E_P_SDMAHALT)
+			__qib_sdma_process_event(ppd,
+				qib_sdma_event_e50_hw_cleaned);
+		break;
+
+	case qib_sdma_state_s50_hw_halt_wait:
+		if (errs & QIB_E_P_SDMAHALT)
+			__qib_sdma_process_event(ppd,
+				qib_sdma_event_e60_hw_halted);
+		break;
+
+	case qib_sdma_state_s99_running:
+		__qib_sdma_process_event(ppd, qib_sdma_event_e7322_err_halted);
+		__qib_sdma_process_event(ppd, qib_sdma_event_e60_hw_halted);
+		break;
+	}
+
+	spin_unlock_irqrestore(&ppd->sdma_lock, flags);
+}
+
+/*
+ * handle per-device errors (not per-port errors)
+ */
+static noinline void handle_7322_errors(struct qib_devdata *dd)
+{
+	char *msg;
+	u64 iserr = 0;
+	u64 errs;
+	u64 mask;
+	int log_idx;
+
+	qib_stats.sps_errints++;
+	errs = qib_read_kreg64(dd, kr_errstatus);
+	if (!errs) {
+		qib_devinfo(dd->pcidev, "device error interrupt, "
+			 "but no error bits set!\n");
+		goto done;
+	}
+
+	/* don't report errors that are masked */
+	errs &= dd->cspec->errormask;
+	msg = dd->cspec->emsgbuf;
+
+	/* do these first, they are most important */
+	if (errs & QIB_E_HARDWARE) {
+		*msg = '\0';
+		qib_7322_handle_hwerrors(dd, msg, sizeof dd->cspec->emsgbuf);
+	} else
+		for (log_idx = 0; log_idx < QIB_EEP_LOG_CNT; ++log_idx)
+			if (errs & dd->eep_st_masks[log_idx].errs_to_log)
+				qib_inc_eeprom_err(dd, log_idx, 1);
+
+	if (errs & QIB_E_SPKTERRS) {
+		qib_disarm_7322_senderrbufs(dd->pport);
+		qib_stats.sps_txerrs++;
+	} else if (errs & QIB_E_INVALIDADDR)
+		qib_stats.sps_txerrs++;
+	else if (errs & QIB_E_ARMLAUNCH) {
+		qib_stats.sps_txerrs++;
+		qib_disarm_7322_senderrbufs(dd->pport);
+	}
+	qib_write_kreg(dd, kr_errclear, errs);
+
+	/*
+	 * The ones we mask off are handled specially below
+	 * or above.  Also mask SDMADISABLED by default as it
+	 * is too chatty.
+	 */
+	mask = QIB_E_HARDWARE;
+	*msg = '\0';
+
+	err_decode(msg, sizeof dd->cspec->emsgbuf, errs & ~mask,
+		   qib_7322error_msgs);
+
+	/*
+	 * Getting reset is a tragedy for all ports. Mark the device
+	 * _and_ the ports as "offline" in way meaningful to each.
+	 */
+	if (errs & QIB_E_RESET) {
+		int pidx;
+
+		qib_dev_err(dd, "Got reset, requires re-init "
+			    "(unload and reload driver)\n");
+		dd->flags &= ~QIB_INITTED;  /* needs re-init */
+		/* mark as having had error */
+		*dd->devstatusp |= QIB_STATUS_HWERROR;
+		for (pidx = 0; pidx < dd->num_pports; ++pidx)
+			if (dd->pport[pidx].link_speed_supported)
+				*dd->pport[pidx].statusp &= ~QIB_STATUS_IB_CONF;
+	}
+
+	if (*msg && iserr)
+		qib_dev_err(dd, "%s error\n", msg);
+
+	/*
+	 * If there were hdrq or egrfull errors, wake up any processes
+	 * waiting in poll.  We used to try to check which contexts had
+	 * the overflow, but given the cost of that and the chip reads
+	 * to support it, it's better to just wake everybody up if we
+	 * get an overflow; waiters can poll again if it's not them.
+	 */
+	if (errs & (ERR_MASK(RcvEgrFullErr) | ERR_MASK(RcvHdrFullErr))) {
+		qib_handle_urcv(dd, ~0U);
+		if (errs & ERR_MASK(RcvEgrFullErr))
+			qib_stats.sps_buffull++;
+		else
+			qib_stats.sps_hdrfull++;
+	}
+
+done:
+	return;
+}
+
+static void reenable_chase(unsigned long opaque)
+{
+	struct qib_pportdata *ppd = (struct qib_pportdata *)opaque;
+
+	ppd->cpspec->chase_timer.expires = 0;
+	qib_set_ib_7322_lstate(ppd, QLOGIC_IB_IBCC_LINKCMD_DOWN,
+		QLOGIC_IB_IBCC_LINKINITCMD_POLL);
+}
+
+static void disable_chase(struct qib_pportdata *ppd, u64 tnow, u8 ibclt)
+{
+	ppd->cpspec->chase_end = 0;
+
+	if (!qib_chase)
+		return;
+
+	qib_set_ib_7322_lstate(ppd, QLOGIC_IB_IBCC_LINKCMD_DOWN,
+		QLOGIC_IB_IBCC_LINKINITCMD_DISABLE);
+	ppd->cpspec->chase_timer.expires = jiffies + QIB_CHASE_DIS_TIME;
+	add_timer(&ppd->cpspec->chase_timer);
+}
+
+static void handle_serdes_issues(struct qib_pportdata *ppd, u64 ibcst)
+{
+	u8 ibclt;
+	u64 tnow;
+
+	ibclt = (u8)SYM_FIELD(ibcst, IBCStatusA_0, LinkTrainingState);
+
+	/*
+	 * Detect and handle the state chase issue, where we can
+	 * get stuck if we are unlucky on timing on both sides of
+	 * the link.   If we are, we disable, set a timer, and
+	 * then re-enable.
+	 */
+	switch (ibclt) {
+	case IB_7322_LT_STATE_CFGRCVFCFG:
+	case IB_7322_LT_STATE_CFGWAITRMT:
+	case IB_7322_LT_STATE_TXREVLANES:
+	case IB_7322_LT_STATE_CFGENH:
+		tnow = get_jiffies_64();
+		if (ppd->cpspec->chase_end &&
+		     time_after64(tnow, ppd->cpspec->chase_end))
+			disable_chase(ppd, tnow, ibclt);
+		else if (!ppd->cpspec->chase_end)
+			ppd->cpspec->chase_end = tnow + QIB_CHASE_TIME;
+		break;
+	default:
+		ppd->cpspec->chase_end = 0;
+		break;
+	}
+
+	if (ibclt == IB_7322_LT_STATE_CFGTEST &&
+	    (ibcst & SYM_MASK(IBCStatusA_0, LinkSpeedQDR))) {
+		force_h1(ppd);
+		ppd->cpspec->qdr_reforce = 1;
+	} else if (ppd->cpspec->qdr_reforce &&
+		(ibcst & SYM_MASK(IBCStatusA_0, LinkSpeedQDR)) &&
+		 (ibclt == IB_7322_LT_STATE_CFGENH ||
+		ibclt == IB_7322_LT_STATE_CFGIDLE ||
+		ibclt == IB_7322_LT_STATE_LINKUP))
+		force_h1(ppd);
+
+	if ((IS_QMH(ppd->dd) || IS_QME(ppd->dd)) &&
+	    ppd->link_speed_enabled == QIB_IB_QDR &&
+	    (ibclt == IB_7322_LT_STATE_CFGTEST ||
+	     ibclt == IB_7322_LT_STATE_CFGENH ||
+	     (ibclt >= IB_7322_LT_STATE_POLLACTIVE &&
+	      ibclt <= IB_7322_LT_STATE_SLEEPQUIET)))
+		adj_tx_serdes(ppd);
+
+	if (!ppd->cpspec->qdr_dfe_on && ibclt != IB_7322_LT_STATE_LINKUP &&
+	    ibclt <= IB_7322_LT_STATE_SLEEPQUIET) {
+		ppd->cpspec->qdr_dfe_on = 1;
+		ppd->cpspec->qdr_dfe_time = 0;
+		/* On link down, reenable QDR adaptation */
+		qib_write_kreg_port(ppd, krp_static_adapt_dis(2),
+			ppd->dd->cspec->r1 ?
+				    QDR_STATIC_ADAPT_DOWN_R1 :
+				    QDR_STATIC_ADAPT_DOWN);
+	}
+}
+
+/*
+ * This is per-pport error handling.
+ * will likely get it's own MSIx interrupt (one for each port,
+ * although just a single handler).
+ */
+static noinline void handle_7322_p_errors(struct qib_pportdata *ppd)
+{
+	char *msg;
+	u64 ignore_this_time = 0, iserr = 0, errs, fmask;
+	struct qib_devdata *dd = ppd->dd;
+
+	/* do this as soon as possible */
+	fmask = qib_read_kreg64(dd, kr_act_fmask);
+	if (!fmask)
+		check_7322_rxe_status(ppd);
+
+	errs = qib_read_kreg_port(ppd, krp_errstatus);
+	if (!errs)
+		qib_devinfo(dd->pcidev,
+			 "Port%d error interrupt, but no error bits set!\n",
+			 ppd->port);
+	if (!fmask)
+		errs &= ~QIB_E_P_IBSTATUSCHANGED;
+	if (!errs)
+		goto done;
+
+	msg = ppd->cpspec->epmsgbuf;
+	*msg = '\0';
+
+	if (errs & ~QIB_E_P_BITSEXTANT) {
+		err_decode(msg, sizeof ppd->cpspec->epmsgbuf,
+			   errs & ~QIB_E_P_BITSEXTANT, qib_7322p_error_msgs);
+		if (!*msg)
+			snprintf(msg, sizeof ppd->cpspec->epmsgbuf,
+				 "no others");
+		qib_dev_porterr(dd, ppd->port, "error interrupt with unknown"
+				" errors 0x%016Lx set (and %s)\n",
+				(errs & ~QIB_E_P_BITSEXTANT), msg);
+		*msg = '\0';
+	}
+
+	if (errs & QIB_E_P_SHDR) {
+		u64 symptom;
+
+		/* determine cause, then write to clear */
+		symptom = qib_read_kreg_port(ppd, krp_sendhdrsymptom);
+		qib_write_kreg_port(ppd, krp_sendhdrsymptom, 0);
+		err_decode(msg, sizeof ppd->cpspec->epmsgbuf, symptom,
+			   hdrchk_msgs);
+		*msg = '\0';
+		/* senderrbuf cleared in SPKTERRS below */
+	}
+
+	if (errs & QIB_E_P_SPKTERRS) {
+		if ((errs & QIB_E_P_LINK_PKTERRS) &&
+		    !(ppd->lflags & QIBL_LINKACTIVE)) {
+			/*
+			 * This can happen when trying to bring the link
+			 * up, but the IB link changes state at the "wrong"
+			 * time. The IB logic then complains that the packet
+			 * isn't valid.  We don't want to confuse people, so
+			 * we just don't print them, except at debug
+			 */
+			err_decode(msg, sizeof ppd->cpspec->epmsgbuf,
+				   (errs & QIB_E_P_LINK_PKTERRS),
+				   qib_7322p_error_msgs);
+			*msg = '\0';
+			ignore_this_time = errs & QIB_E_P_LINK_PKTERRS;
+		}
+		qib_disarm_7322_senderrbufs(ppd);
+	} else if ((errs & QIB_E_P_LINK_PKTERRS) &&
+		   !(ppd->lflags & QIBL_LINKACTIVE)) {
+		/*
+		 * This can happen when SMA is trying to bring the link
+		 * up, but the IB link changes state at the "wrong" time.
+		 * The IB logic then complains that the packet isn't
+		 * valid.  We don't want to confuse people, so we just
+		 * don't print them, except at debug
+		 */
+		err_decode(msg, sizeof ppd->cpspec->epmsgbuf, errs,
+			   qib_7322p_error_msgs);
+		ignore_this_time = errs & QIB_E_P_LINK_PKTERRS;
+		*msg = '\0';
+	}
+
+	qib_write_kreg_port(ppd, krp_errclear, errs);
+
+	errs &= ~ignore_this_time;
+	if (!errs)
+		goto done;
+
+	if (errs & QIB_E_P_RPKTERRS)
+		qib_stats.sps_rcverrs++;
+	if (errs & QIB_E_P_SPKTERRS)
+		qib_stats.sps_txerrs++;
+
+	iserr = errs & ~(QIB_E_P_RPKTERRS | QIB_E_P_PKTERRS);
+
+	if (errs & QIB_E_P_SDMAERRS)
+		sdma_7322_p_errors(ppd, errs);
+
+	if (errs & QIB_E_P_IBSTATUSCHANGED) {
+		u64 ibcs;
+		u8 ltstate;
+
+		ibcs = qib_read_kreg_port(ppd, krp_ibcstatus_a);
+		ltstate = qib_7322_phys_portstate(ibcs);
+
+		if (!(ppd->lflags & QIBL_IB_AUTONEG_INPROG))
+			handle_serdes_issues(ppd, ibcs);
+		if (!(ppd->cpspec->ibcctrl_a &
+		      SYM_MASK(IBCCtrlA_0, IBStatIntReductionEn))) {
+			/*
+			 * We got our interrupt, so init code should be
+			 * happy and not try alternatives. Now squelch
+			 * other "chatter" from link-negotiation (pre Init)
+			 */
+			ppd->cpspec->ibcctrl_a |=
+				SYM_MASK(IBCCtrlA_0, IBStatIntReductionEn);
+			qib_write_kreg_port(ppd, krp_ibcctrl_a,
+					    ppd->cpspec->ibcctrl_a);
+		}
+
+		/* Update our picture of width and speed from chip */
+		ppd->link_width_active =
+			(ibcs & SYM_MASK(IBCStatusA_0, LinkWidthActive)) ?
+			    IB_WIDTH_4X : IB_WIDTH_1X;
+		ppd->link_speed_active = (ibcs & SYM_MASK(IBCStatusA_0,
+			LinkSpeedQDR)) ? QIB_IB_QDR : (ibcs &
+			  SYM_MASK(IBCStatusA_0, LinkSpeedActive)) ?
+				   QIB_IB_DDR : QIB_IB_SDR;
+
+		if ((ppd->lflags & QIBL_IB_LINK_DISABLED) && ltstate !=
+		    IB_PHYSPORTSTATE_DISABLED)
+			qib_set_ib_7322_lstate(ppd, 0,
+			       QLOGIC_IB_IBCC_LINKINITCMD_DISABLE);
+		else
+			/*
+			 * Since going into a recovery state causes the link
+			 * state to go down and since recovery is transitory,
+			 * it is better if we "miss" ever seeing the link
+			 * training state go into recovery (i.e., ignore this
+			 * transition for link state special handling purposes)
+			 * without updating lastibcstat.
+			 */
+			if (ltstate != IB_PHYSPORTSTATE_LINK_ERR_RECOVER &&
+			    ltstate != IB_PHYSPORTSTATE_RECOVERY_RETRAIN &&
+			    ltstate != IB_PHYSPORTSTATE_RECOVERY_WAITRMT &&
+			    ltstate != IB_PHYSPORTSTATE_RECOVERY_IDLE)
+				qib_handle_e_ibstatuschanged(ppd, ibcs);
+	}
+	if (*msg && iserr)
+		qib_dev_porterr(dd, ppd->port, "%s error\n", msg);
+
+	if (ppd->state_wanted & ppd->lflags)
+		wake_up_interruptible(&ppd->state_wait);
+done:
+	return;
+}
+
+/* enable/disable chip from delivering interrupts */
+static void qib_7322_set_intr_state(struct qib_devdata *dd, u32 enable)
+{
+	if (enable) {
+		if (dd->flags & QIB_BADINTR)
+			return;
+		qib_write_kreg(dd, kr_intmask, dd->cspec->int_enable_mask);
+		/* cause any pending enabled interrupts to be re-delivered */
+		qib_write_kreg(dd, kr_intclear, 0ULL);
+		if (dd->cspec->num_msix_entries) {
+			/* and same for MSIx */
+			u64 val = qib_read_kreg64(dd, kr_intgranted);
+			if (val)
+				qib_write_kreg(dd, kr_intgranted, val);
+		}
+	} else
+		qib_write_kreg(dd, kr_intmask, 0ULL);
+}
+
+/*
+ * Try to cleanup as much as possible for anything that might have gone
+ * wrong while in freeze mode, such as pio buffers being written by user
+ * processes (causing armlaunch), send errors due to going into freeze mode,
+ * etc., and try to avoid causing extra interrupts while doing so.
+ * Forcibly update the in-memory pioavail register copies after cleanup
+ * because the chip won't do it while in freeze mode (the register values
+ * themselves are kept correct).
+ * Make sure that we don't lose any important interrupts by using the chip
+ * feature that says that writing 0 to a bit in *clear that is set in
+ * *status will cause an interrupt to be generated again (if allowed by
+ * the *mask value).
+ * This is in chip-specific code because of all of the register accesses,
+ * even though the details are similar on most chips.
+ */
+static void qib_7322_clear_freeze(struct qib_devdata *dd)
+{
+	int pidx;
+
+	/* disable error interrupts, to avoid confusion */
+	qib_write_kreg(dd, kr_errmask, 0ULL);
+
+	for (pidx = 0; pidx < dd->num_pports; ++pidx)
+		if (dd->pport[pidx].link_speed_supported)
+			qib_write_kreg_port(dd->pport + pidx, krp_errmask,
+					    0ULL);
+
+	/* also disable interrupts; errormask is sometimes overwriten */
+	qib_7322_set_intr_state(dd, 0);
+
+	/* clear the freeze, and be sure chip saw it */
+	qib_write_kreg(dd, kr_control, dd->control);
+	qib_read_kreg32(dd, kr_scratch);
+
+	/*
+	 * Force new interrupt if any hwerr, error or interrupt bits are
+	 * still set, and clear "safe" send packet errors related to freeze
+	 * and cancelling sends.  Re-enable error interrupts before possible
+	 * force of re-interrupt on pending interrupts.
+	 */
+	qib_write_kreg(dd, kr_hwerrclear, 0ULL);
+	qib_write_kreg(dd, kr_errclear, E_SPKT_ERRS_IGNORE);
+	qib_write_kreg(dd, kr_errmask, dd->cspec->errormask);
+	/* We need to purge per-port errs and reset mask, too */
+	for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+		if (!dd->pport[pidx].link_speed_supported)
+			continue;
+		qib_write_kreg_port(dd->pport + pidx, krp_errclear, ~0Ull);
+		qib_write_kreg_port(dd->pport + pidx, krp_errmask, ~0Ull);
+	}
+	qib_7322_set_intr_state(dd, 1);
+}
+
+/* no error handling to speak of */
+/**
+ * qib_7322_handle_hwerrors - display hardware errors.
+ * @dd: the qlogic_ib device
+ * @msg: the output buffer
+ * @msgl: the size of the output buffer
+ *
+ * Use same msg buffer as regular errors to avoid excessive stack
+ * use.  Most hardware errors are catastrophic, but for right now,
+ * we'll print them and continue.  We reuse the same message buffer as
+ * qib_handle_errors() to avoid excessive stack usage.
+ */
+static void qib_7322_handle_hwerrors(struct qib_devdata *dd, char *msg,
+				     size_t msgl)
+{
+	u64 hwerrs;
+	u32 ctrl;
+	int isfatal = 0;
+
+	hwerrs = qib_read_kreg64(dd, kr_hwerrstatus);
+	if (!hwerrs)
+		goto bail;
+	if (hwerrs == ~0ULL) {
+		qib_dev_err(dd, "Read of hardware error status failed "
+			    "(all bits set); ignoring\n");
+		goto bail;
+	}
+	qib_stats.sps_hwerrs++;
+
+	/* Always clear the error status register, except BIST fail */
+	qib_write_kreg(dd, kr_hwerrclear, hwerrs &
+		       ~HWE_MASK(PowerOnBISTFailed));
+
+	hwerrs &= dd->cspec->hwerrmask;
+
+	/* no EEPROM logging, yet */
+
+	if (hwerrs)
+		qib_devinfo(dd->pcidev, "Hardware error: hwerr=0x%llx "
+			    "(cleared)\n", (unsigned long long) hwerrs);
+
+	ctrl = qib_read_kreg32(dd, kr_control);
+	if ((ctrl & SYM_MASK(Control, FreezeMode)) && !dd->diag_client) {
+		/*
+		 * No recovery yet...
+		 */
+		if ((hwerrs & ~HWE_MASK(LATriggered)) ||
+		    dd->cspec->stay_in_freeze) {
+			/*
+			 * If any set that we aren't ignoring only make the
+			 * complaint once, in case it's stuck or recurring,
+			 * and we get here multiple times
+			 * Force link down, so switch knows, and
+			 * LEDs are turned off.
+			 */
+			if (dd->flags & QIB_INITTED)
+				isfatal = 1;
+		} else
+			qib_7322_clear_freeze(dd);
+	}
+
+	if (hwerrs & HWE_MASK(PowerOnBISTFailed)) {
+		isfatal = 1;
+		strlcpy(msg, "[Memory BIST test failed, "
+			"InfiniPath hardware unusable]", msgl);
+		/* ignore from now on, so disable until driver reloaded */
+		dd->cspec->hwerrmask &= ~HWE_MASK(PowerOnBISTFailed);
+		qib_write_kreg(dd, kr_hwerrmask, dd->cspec->hwerrmask);
+	}
+
+	err_decode(msg, msgl, hwerrs, qib_7322_hwerror_msgs);
+
+	/* Ignore esoteric PLL failures et al. */
+
+	qib_dev_err(dd, "%s hardware error\n", msg);
+
+	if (isfatal && !dd->diag_client) {
+		qib_dev_err(dd, "Fatal Hardware Error, no longer"
+			    " usable, SN %.16s\n", dd->serial);
+		/*
+		 * for /sys status file and user programs to print; if no
+		 * trailing brace is copied, we'll know it was truncated.
+		 */
+		if (dd->freezemsg)
+			snprintf(dd->freezemsg, dd->freezelen,
+				 "{%s}", msg);
+		qib_disable_after_error(dd);
+	}
+bail:;
+}
+
+/**
+ * qib_7322_init_hwerrors - enable hardware errors
+ * @dd: the qlogic_ib device
+ *
+ * now that we have finished initializing everything that might reasonably
+ * cause a hardware error, and cleared those errors bits as they occur,
+ * we can enable hardware errors in the mask (potentially enabling
+ * freeze mode), and enable hardware errors as errors (along with
+ * everything else) in errormask
+ */
+static void qib_7322_init_hwerrors(struct qib_devdata *dd)
+{
+	int pidx;
+	u64 extsval;
+
+	extsval = qib_read_kreg64(dd, kr_extstatus);
+	if (!(extsval & (QIB_EXTS_MEMBIST_DISABLED |
+			 QIB_EXTS_MEMBIST_ENDTEST)))
+		qib_dev_err(dd, "MemBIST did not complete!\n");
+
+	/* never clear BIST failure, so reported on each driver load */
+	qib_write_kreg(dd, kr_hwerrclear, ~HWE_MASK(PowerOnBISTFailed));
+	qib_write_kreg(dd, kr_hwerrmask, dd->cspec->hwerrmask);
+
+	/* clear all */
+	qib_write_kreg(dd, kr_errclear, ~0ULL);
+	/* enable errors that are masked, at least this first time. */
+	qib_write_kreg(dd, kr_errmask, ~0ULL);
+	dd->cspec->errormask = qib_read_kreg64(dd, kr_errmask);
+	for (pidx = 0; pidx < dd->num_pports; ++pidx)
+		if (dd->pport[pidx].link_speed_supported)
+			qib_write_kreg_port(dd->pport + pidx, krp_errmask,
+					    ~0ULL);
+}
+
+/*
+ * Disable and enable the armlaunch error.  Used for PIO bandwidth testing
+ * on chips that are count-based, rather than trigger-based.  There is no
+ * reference counting, but that's also fine, given the intended use.
+ * Only chip-specific because it's all register accesses
+ */
+static void qib_set_7322_armlaunch(struct qib_devdata *dd, u32 enable)
+{
+	if (enable) {
+		qib_write_kreg(dd, kr_errclear, QIB_E_SPIOARMLAUNCH);
+		dd->cspec->errormask |= QIB_E_SPIOARMLAUNCH;
+	} else
+		dd->cspec->errormask &= ~QIB_E_SPIOARMLAUNCH;
+	qib_write_kreg(dd, kr_errmask, dd->cspec->errormask);
+}
+
+/*
+ * Formerly took parameter <which> in pre-shifted,
+ * pre-merged form with LinkCmd and LinkInitCmd
+ * together, and assuming the zero was NOP.
+ */
+static void qib_set_ib_7322_lstate(struct qib_pportdata *ppd, u16 linkcmd,
+				   u16 linitcmd)
+{
+	u64 mod_wd;
+	struct qib_devdata *dd = ppd->dd;
+	unsigned long flags;
+
+	if (linitcmd == QLOGIC_IB_IBCC_LINKINITCMD_DISABLE) {
+		/*
+		 * If we are told to disable, note that so link-recovery
+		 * code does not attempt to bring us back up.
+		 * Also reset everything that we can, so we start
+		 * completely clean when re-enabled (before we
+		 * actually issue the disable to the IBC)
+		 */
+		qib_7322_mini_pcs_reset(ppd);
+		spin_lock_irqsave(&ppd->lflags_lock, flags);
+		ppd->lflags |= QIBL_IB_LINK_DISABLED;
+		spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+	} else if (linitcmd || linkcmd == QLOGIC_IB_IBCC_LINKCMD_DOWN) {
+		/*
+		 * Any other linkinitcmd will lead to LINKDOWN and then
+		 * to INIT (if all is well), so clear flag to let
+		 * link-recovery code attempt to bring us back up.
+		 */
+		spin_lock_irqsave(&ppd->lflags_lock, flags);
+		ppd->lflags &= ~QIBL_IB_LINK_DISABLED;
+		spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+		/*
+		 * Clear status change interrupt reduction so the
+		 * new state is seen.
+		 */
+		ppd->cpspec->ibcctrl_a &=
+			~SYM_MASK(IBCCtrlA_0, IBStatIntReductionEn);
+	}
+
+	mod_wd = (linkcmd << IBA7322_IBCC_LINKCMD_SHIFT) |
+		(linitcmd << QLOGIC_IB_IBCC_LINKINITCMD_SHIFT);
+
+	qib_write_kreg_port(ppd, krp_ibcctrl_a, ppd->cpspec->ibcctrl_a |
+			    mod_wd);
+	/* write to chip to prevent back-to-back writes of ibc reg */
+	qib_write_kreg(dd, kr_scratch, 0);
+
+}
+
+/*
+ * The total RCV buffer memory is 64KB, used for both ports, and is
+ * in units of 64 bytes (same as IB flow control credit unit).
+ * The consumedVL unit in the same registers are in 32 byte units!
+ * So, a VL15 packet needs 4.50 IB credits, and 9 rx buffer chunks,
+ * and we can therefore allocate just 9 IB credits for 2 VL15 packets
+ * in krp_rxcreditvl15, rather than 10.
+ */
+#define RCV_BUF_UNITSZ 64
+#define NUM_RCV_BUF_UNITS(dd) ((64 * 1024) / (RCV_BUF_UNITSZ * dd->num_pports))
+
+static void set_vls(struct qib_pportdata *ppd)
+{
+	int i, numvls, totcred, cred_vl, vl0extra;
+	struct qib_devdata *dd = ppd->dd;
+	u64 val;
+
+	numvls = qib_num_vls(ppd->vls_operational);
+
+	/*
+	 * Set up per-VL credits. Below is kluge based on these assumptions:
+	 * 1) port is disabled at the time early_init is called.
+	 * 2) give VL15 17 credits, for two max-plausible packets.
+	 * 3) Give VL0-N the rest, with any rounding excess used for VL0
+	 */
+	/* 2 VL15 packets @ 288 bytes each (including IB headers) */
+	totcred = NUM_RCV_BUF_UNITS(dd);
+	cred_vl = (2 * 288 + RCV_BUF_UNITSZ - 1) / RCV_BUF_UNITSZ;
+	totcred -= cred_vl;
+	qib_write_kreg_port(ppd, krp_rxcreditvl15, (u64) cred_vl);
+	cred_vl = totcred / numvls;
+	vl0extra = totcred - cred_vl * numvls;
+	qib_write_kreg_port(ppd, krp_rxcreditvl0, cred_vl + vl0extra);
+	for (i = 1; i < numvls; i++)
+		qib_write_kreg_port(ppd, krp_rxcreditvl0 + i, cred_vl);
+	for (; i < 8; i++) /* no buffer space for other VLs */
+		qib_write_kreg_port(ppd, krp_rxcreditvl0 + i, 0);
+
+	/* Notify IBC that credits need to be recalculated */
+	val = qib_read_kreg_port(ppd, krp_ibsdtestiftx);
+	val |= SYM_MASK(IB_SDTEST_IF_TX_0, CREDIT_CHANGE);
+	qib_write_kreg_port(ppd, krp_ibsdtestiftx, val);
+	qib_write_kreg(dd, kr_scratch, 0ULL);
+	val &= ~SYM_MASK(IB_SDTEST_IF_TX_0, CREDIT_CHANGE);
+	qib_write_kreg_port(ppd, krp_ibsdtestiftx, val);
+
+	for (i = 0; i < numvls; i++)
+		val = qib_read_kreg_port(ppd, krp_rxcreditvl0 + i);
+	val = qib_read_kreg_port(ppd, krp_rxcreditvl15);
+
+	/* Change the number of operational VLs */
+	ppd->cpspec->ibcctrl_a = (ppd->cpspec->ibcctrl_a &
+				~SYM_MASK(IBCCtrlA_0, NumVLane)) |
+		((u64)(numvls - 1) << SYM_LSB(IBCCtrlA_0, NumVLane));
+	qib_write_kreg_port(ppd, krp_ibcctrl_a, ppd->cpspec->ibcctrl_a);
+	qib_write_kreg(dd, kr_scratch, 0ULL);
+}
+
+/*
+ * The code that deals with actual SerDes is in serdes_7322_init().
+ * Compared to the code for iba7220, it is minimal.
+ */
+static int serdes_7322_init(struct qib_pportdata *ppd);
+
+/**
+ * qib_7322_bringup_serdes - bring up the serdes
+ * @ppd: physical port on the qlogic_ib device
+ */
+static int qib_7322_bringup_serdes(struct qib_pportdata *ppd)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u64 val, guid, ibc;
+	unsigned long flags;
+	int ret = 0;
+
+	/*
+	 * SerDes model not in Pd, but still need to
+	 * set up much of IBCCtrl and IBCDDRCtrl; move elsewhere
+	 * eventually.
+	 */
+	/* Put IBC in reset, sends disabled (should be in reset already) */
+	ppd->cpspec->ibcctrl_a &= ~SYM_MASK(IBCCtrlA_0, IBLinkEn);
+	qib_write_kreg_port(ppd, krp_ibcctrl_a, ppd->cpspec->ibcctrl_a);
+	qib_write_kreg(dd, kr_scratch, 0ULL);
+
+	if (qib_compat_ddr_negotiate) {
+		ppd->cpspec->ibdeltainprog = 1;
+		ppd->cpspec->ibsymsnap = read_7322_creg32_port(ppd,
+						crp_ibsymbolerr);
+		ppd->cpspec->iblnkerrsnap = read_7322_creg32_port(ppd,
+						crp_iblinkerrrecov);
+	}
+
+	/* flowcontrolwatermark is in units of KBytes */
+	ibc = 0x5ULL << SYM_LSB(IBCCtrlA_0, FlowCtrlWaterMark);
+	/*
+	 * Flow control is sent this often, even if no changes in
+	 * buffer space occur.  Units are 128ns for this chip.
+	 * Set to 3usec.
+	 */
+	ibc |= 24ULL << SYM_LSB(IBCCtrlA_0, FlowCtrlPeriod);
+	/* max error tolerance */
+	ibc |= 0xfULL << SYM_LSB(IBCCtrlA_0, PhyerrThreshold);
+	/* IB credit flow control. */
+	ibc |= 0xfULL << SYM_LSB(IBCCtrlA_0, OverrunThreshold);
+	/*
+	 * set initial max size pkt IBC will send, including ICRC; it's the
+	 * PIO buffer size in dwords, less 1; also see qib_set_mtu()
+	 */
+	ibc |= ((u64)(ppd->ibmaxlen >> 2) + 1) <<
+		SYM_LSB(IBCCtrlA_0, MaxPktLen);
+	ppd->cpspec->ibcctrl_a = ibc; /* without linkcmd or linkinitcmd! */
+
+	/* initially come up waiting for TS1, without sending anything. */
+	val = ppd->cpspec->ibcctrl_a | (QLOGIC_IB_IBCC_LINKINITCMD_DISABLE <<
+		QLOGIC_IB_IBCC_LINKINITCMD_SHIFT);
+
+	/*
+	 * Reset the PCS interface to the serdes (and also ibc, which is still
+	 * in reset from above).  Writes new value of ibcctrl_a as last step.
+	 */
+	qib_7322_mini_pcs_reset(ppd);
+	qib_write_kreg(dd, kr_scratch, 0ULL);
+
+	if (!ppd->cpspec->ibcctrl_b) {
+		unsigned lse = ppd->link_speed_enabled;
+
+		/*
+		 * Not on re-init after reset, establish shadow
+		 * and force initial config.
+		 */
+		ppd->cpspec->ibcctrl_b = qib_read_kreg_port(ppd,
+							     krp_ibcctrl_b);
+		ppd->cpspec->ibcctrl_b &= ~(IBA7322_IBC_SPEED_QDR |
+				IBA7322_IBC_SPEED_DDR |
+				IBA7322_IBC_SPEED_SDR |
+				IBA7322_IBC_WIDTH_AUTONEG |
+				SYM_MASK(IBCCtrlB_0, IB_LANE_REV_SUPPORTED));
+		if (lse & (lse - 1)) /* Muliple speeds enabled */
+			ppd->cpspec->ibcctrl_b |=
+				(lse << IBA7322_IBC_SPEED_LSB) |
+				IBA7322_IBC_IBTA_1_2_MASK |
+				IBA7322_IBC_MAX_SPEED_MASK;
+		else
+			ppd->cpspec->ibcctrl_b |= (lse == QIB_IB_QDR) ?
+				IBA7322_IBC_SPEED_QDR |
+				 IBA7322_IBC_IBTA_1_2_MASK :
+				(lse == QIB_IB_DDR) ?
+					IBA7322_IBC_SPEED_DDR :
+					IBA7322_IBC_SPEED_SDR;
+		if ((ppd->link_width_enabled & (IB_WIDTH_1X | IB_WIDTH_4X)) ==
+		    (IB_WIDTH_1X | IB_WIDTH_4X))
+			ppd->cpspec->ibcctrl_b |= IBA7322_IBC_WIDTH_AUTONEG;
+		else
+			ppd->cpspec->ibcctrl_b |=
+				ppd->link_width_enabled == IB_WIDTH_4X ?
+				IBA7322_IBC_WIDTH_4X_ONLY :
+				IBA7322_IBC_WIDTH_1X_ONLY;
+
+		/* always enable these on driver reload, not sticky */
+		ppd->cpspec->ibcctrl_b |= (IBA7322_IBC_RXPOL_MASK |
+			IBA7322_IBC_HRTBT_MASK);
+	}
+	qib_write_kreg_port(ppd, krp_ibcctrl_b, ppd->cpspec->ibcctrl_b);
+
+	/* setup so we have more time at CFGTEST to change H1 */
+	val = qib_read_kreg_port(ppd, krp_ibcctrl_c);
+	val &= ~SYM_MASK(IBCCtrlC_0, IB_FRONT_PORCH);
+	val |= 0xfULL << SYM_LSB(IBCCtrlC_0, IB_FRONT_PORCH);
+	qib_write_kreg_port(ppd, krp_ibcctrl_c, val);
+
+	serdes_7322_init(ppd);
+
+	guid = be64_to_cpu(ppd->guid);
+	if (!guid) {
+		if (dd->base_guid)
+			guid = be64_to_cpu(dd->base_guid) + ppd->port - 1;
+		ppd->guid = cpu_to_be64(guid);
+	}
+
+	qib_write_kreg_port(ppd, krp_hrtbt_guid, guid);
+	/* write to chip to prevent back-to-back writes of ibc reg */
+	qib_write_kreg(dd, kr_scratch, 0);
+
+	/* Enable port */
+	ppd->cpspec->ibcctrl_a |= SYM_MASK(IBCCtrlA_0, IBLinkEn);
+	set_vls(ppd);
+
+	/* be paranoid against later code motion, etc. */
+	spin_lock_irqsave(&dd->cspec->rcvmod_lock, flags);
+	ppd->p_rcvctrl |= SYM_MASK(RcvCtrl_0, RcvIBPortEnable);
+	qib_write_kreg_port(ppd, krp_rcvctrl, ppd->p_rcvctrl);
+	spin_unlock_irqrestore(&dd->cspec->rcvmod_lock, flags);
+
+	/* Also enable IBSTATUSCHG interrupt.  */
+	val = qib_read_kreg_port(ppd, krp_errmask);
+	qib_write_kreg_port(ppd, krp_errmask,
+		val | ERR_MASK_N(IBStatusChanged));
+
+	/* Always zero until we start messing with SerDes for real */
+	return ret;
+}
+
+/**
+ * qib_7322_quiet_serdes - set serdes to txidle
+ * @dd: the qlogic_ib device
+ * Called when driver is being unloaded
+ */
+static void qib_7322_mini_quiet_serdes(struct qib_pportdata *ppd)
+{
+	u64 val;
+	unsigned long flags;
+
+	qib_set_ib_7322_lstate(ppd, 0, QLOGIC_IB_IBCC_LINKINITCMD_DISABLE);
+
+	spin_lock_irqsave(&ppd->lflags_lock, flags);
+	ppd->lflags &= ~QIBL_IB_AUTONEG_INPROG;
+	spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+	wake_up(&ppd->cpspec->autoneg_wait);
+	cancel_delayed_work(&ppd->cpspec->autoneg_work);
+	if (ppd->dd->cspec->r1)
+		cancel_delayed_work(&ppd->cpspec->ipg_work);
+	flush_scheduled_work();
+
+	ppd->cpspec->chase_end = 0;
+	if (ppd->cpspec->chase_timer.data) /* if initted */
+		del_timer_sync(&ppd->cpspec->chase_timer);
+
+	/*
+	 * Despite the name, actually disables IBC as well. Do it when
+	 * we are as sure as possible that no more packets can be
+	 * received, following the down and the PCS reset.
+	 * The actual disabling happens in qib_7322_mini_pci_reset(),
+	 * along with the PCS being reset.
+	 */
+	ppd->cpspec->ibcctrl_a &= ~SYM_MASK(IBCCtrlA_0, IBLinkEn);
+	qib_7322_mini_pcs_reset(ppd);
+
+	/*
+	 * Update the adjusted counters so the adjustment persists
+	 * across driver reload.
+	 */
+	if (ppd->cpspec->ibsymdelta || ppd->cpspec->iblnkerrdelta ||
+	    ppd->cpspec->ibdeltainprog || ppd->cpspec->iblnkdowndelta) {
+		struct qib_devdata *dd = ppd->dd;
+		u64 diagc;
+
+		/* enable counter writes */
+		diagc = qib_read_kreg64(dd, kr_hwdiagctrl);
+		qib_write_kreg(dd, kr_hwdiagctrl,
+			       diagc | SYM_MASK(HwDiagCtrl, CounterWrEnable));
+
+		if (ppd->cpspec->ibsymdelta || ppd->cpspec->ibdeltainprog) {
+			val = read_7322_creg32_port(ppd, crp_ibsymbolerr);
+			if (ppd->cpspec->ibdeltainprog)
+				val -= val - ppd->cpspec->ibsymsnap;
+			val -= ppd->cpspec->ibsymdelta;
+			write_7322_creg_port(ppd, crp_ibsymbolerr, val);
+		}
+		if (ppd->cpspec->iblnkerrdelta || ppd->cpspec->ibdeltainprog) {
+			val = read_7322_creg32_port(ppd, crp_iblinkerrrecov);
+			if (ppd->cpspec->ibdeltainprog)
+				val -= val - ppd->cpspec->iblnkerrsnap;
+			val -= ppd->cpspec->iblnkerrdelta;
+			write_7322_creg_port(ppd, crp_iblinkerrrecov, val);
+		}
+		if (ppd->cpspec->iblnkdowndelta) {
+			val = read_7322_creg32_port(ppd, crp_iblinkdown);
+			val += ppd->cpspec->iblnkdowndelta;
+			write_7322_creg_port(ppd, crp_iblinkdown, val);
+		}
+		/*
+		 * No need to save ibmalfdelta since IB perfcounters
+		 * are cleared on driver reload.
+		 */
+
+		/* and disable counter writes */
+		qib_write_kreg(dd, kr_hwdiagctrl, diagc);
+	}
+}
+
+/**
+ * qib_setup_7322_setextled - set the state of the two external LEDs
+ * @ppd: physical port on the qlogic_ib device
+ * @on: whether the link is up or not
+ *
+ * The exact combo of LEDs if on is true is determined by looking
+ * at the ibcstatus.
+ *
+ * These LEDs indicate the physical and logical state of IB link.
+ * For this chip (at least with recommended board pinouts), LED1
+ * is Yellow (logical state) and LED2 is Green (physical state),
+ *
+ * Note:  We try to match the Mellanox HCA LED behavior as best
+ * we can.  Green indicates physical link state is OK (something is
+ * plugged in, and we can train).
+ * Amber indicates the link is logically up (ACTIVE).
+ * Mellanox further blinks the amber LED to indicate data packet
+ * activity, but we have no hardware support for that, so it would
+ * require waking up every 10-20 msecs and checking the counters
+ * on the chip, and then turning the LED off if appropriate.  That's
+ * visible overhead, so not something we will do.
+ */
+static void qib_setup_7322_setextled(struct qib_pportdata *ppd, u32 on)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u64 extctl, ledblink = 0, val;
+	unsigned long flags;
+	int yel, grn;
+
+	/*
+	 * The diags use the LED to indicate diag info, so we leave
+	 * the external LED alone when the diags are running.
+	 */
+	if (dd->diag_client)
+		return;
+
+	/* Allow override of LED display for, e.g. Locating system in rack */
+	if (ppd->led_override) {
+		grn = (ppd->led_override & QIB_LED_PHYS);
+		yel = (ppd->led_override & QIB_LED_LOG);
+	} else if (on) {
+		val = qib_read_kreg_port(ppd, krp_ibcstatus_a);
+		grn = qib_7322_phys_portstate(val) ==
+			IB_PHYSPORTSTATE_LINKUP;
+		yel = qib_7322_iblink_state(val) == IB_PORT_ACTIVE;
+	} else {
+		grn = 0;
+		yel = 0;
+	}
+
+	spin_lock_irqsave(&dd->cspec->gpio_lock, flags);
+	extctl = dd->cspec->extctrl & (ppd->port == 1 ?
+		~ExtLED_IB1_MASK : ~ExtLED_IB2_MASK);
+	if (grn) {
+		extctl |= ppd->port == 1 ? ExtLED_IB1_GRN : ExtLED_IB2_GRN;
+		/*
+		 * Counts are in chip clock (4ns) periods.
+		 * This is 1/16 sec (66.6ms) on,
+		 * 3/16 sec (187.5 ms) off, with packets rcvd.
+		 */
+		ledblink = ((66600 * 1000UL / 4) << IBA7322_LEDBLINK_ON_SHIFT) |
+			((187500 * 1000UL / 4) << IBA7322_LEDBLINK_OFF_SHIFT);
+	}
+	if (yel)
+		extctl |= ppd->port == 1 ? ExtLED_IB1_YEL : ExtLED_IB2_YEL;
+	dd->cspec->extctrl = extctl;
+	qib_write_kreg(dd, kr_extctrl, dd->cspec->extctrl);
+	spin_unlock_irqrestore(&dd->cspec->gpio_lock, flags);
+
+	if (ledblink) /* blink the LED on packet receive */
+		qib_write_kreg_port(ppd, krp_rcvpktledcnt, ledblink);
+}
+
+/*
+ * Disable MSIx interrupt if enabled, call generic MSIx code
+ * to cleanup, and clear pending MSIx interrupts.
+ * Used for fallback to INTx, after reset, and when MSIx setup fails.
+ */
+static void qib_7322_nomsix(struct qib_devdata *dd)
+{
+	u64 intgranted;
+	int n;
+
+	dd->cspec->main_int_mask = ~0ULL;
+	n = dd->cspec->num_msix_entries;
+	if (n) {
+		int i;
+
+		dd->cspec->num_msix_entries = 0;
+		for (i = 0; i < n; i++)
+			free_irq(dd->cspec->msix_entries[i].vector,
+				 dd->cspec->msix_arg[i]);
+		qib_nomsix(dd);
+	}
+	/* make sure no MSIx interrupts are left pending */
+	intgranted = qib_read_kreg64(dd, kr_intgranted);
+	if (intgranted)
+		qib_write_kreg(dd, kr_intgranted, intgranted);
+}
+
+static void qib_7322_free_irq(struct qib_devdata *dd)
+{
+	if (dd->cspec->irq) {
+		free_irq(dd->cspec->irq, dd);
+		dd->cspec->irq = 0;
+	}
+	qib_7322_nomsix(dd);
+}
+
+static void qib_setup_7322_cleanup(struct qib_devdata *dd)
+{
+	int i;
+
+	qib_7322_free_irq(dd);
+	kfree(dd->cspec->cntrs);
+	kfree(dd->cspec->sendchkenable);
+	kfree(dd->cspec->sendgrhchk);
+	kfree(dd->cspec->sendibchk);
+	kfree(dd->cspec->msix_entries);
+	kfree(dd->cspec->msix_arg);
+	for (i = 0; i < dd->num_pports; i++) {
+		unsigned long flags;
+		u32 mask = QSFP_GPIO_MOD_PRS_N |
+			(QSFP_GPIO_MOD_PRS_N << QSFP_GPIO_PORT2_SHIFT);
+
+		kfree(dd->pport[i].cpspec->portcntrs);
+		if (dd->flags & QIB_HAS_QSFP) {
+			spin_lock_irqsave(&dd->cspec->gpio_lock, flags);
+			dd->cspec->gpio_mask &= ~mask;
+			qib_write_kreg(dd, kr_gpio_mask, dd->cspec->gpio_mask);
+			spin_unlock_irqrestore(&dd->cspec->gpio_lock, flags);
+			qib_qsfp_deinit(&dd->pport[i].cpspec->qsfp_data);
+		}
+		if (dd->pport[i].ibport_data.smi_ah)
+			ib_destroy_ah(&dd->pport[i].ibport_data.smi_ah->ibah);
+	}
+}
+
+/* handle SDMA interrupts */
+static void sdma_7322_intr(struct qib_devdata *dd, u64 istat)
+{
+	struct qib_pportdata *ppd0 = &dd->pport[0];
+	struct qib_pportdata *ppd1 = &dd->pport[1];
+	u64 intr0 = istat & (INT_MASK_P(SDma, 0) |
+		INT_MASK_P(SDmaIdle, 0) | INT_MASK_P(SDmaProgress, 0));
+	u64 intr1 = istat & (INT_MASK_P(SDma, 1) |
+		INT_MASK_P(SDmaIdle, 1) | INT_MASK_P(SDmaProgress, 1));
+
+	if (intr0)
+		qib_sdma_intr(ppd0);
+	if (intr1)
+		qib_sdma_intr(ppd1);
+
+	if (istat & INT_MASK_PM(SDmaCleanupDone, 0))
+		qib_sdma_process_event(ppd0, qib_sdma_event_e20_hw_started);
+	if (istat & INT_MASK_PM(SDmaCleanupDone, 1))
+		qib_sdma_process_event(ppd1, qib_sdma_event_e20_hw_started);
+}
+
+/*
+ * Set or clear the Send buffer available interrupt enable bit.
+ */
+static void qib_wantpiobuf_7322_intr(struct qib_devdata *dd, u32 needint)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&dd->sendctrl_lock, flags);
+	if (needint)
+		dd->sendctrl |= SYM_MASK(SendCtrl, SendIntBufAvail);
+	else
+		dd->sendctrl &= ~SYM_MASK(SendCtrl, SendIntBufAvail);
+	qib_write_kreg(dd, kr_sendctrl, dd->sendctrl);
+	qib_write_kreg(dd, kr_scratch, 0ULL);
+	spin_unlock_irqrestore(&dd->sendctrl_lock, flags);
+}
+
+/*
+ * Somehow got an interrupt with reserved bits set in interrupt status.
+ * Print a message so we know it happened, then clear them.
+ * keep mainline interrupt handler cache-friendly
+ */
+static noinline void unknown_7322_ibits(struct qib_devdata *dd, u64 istat)
+{
+	u64 kills;
+	char msg[128];
+
+	kills = istat & ~QIB_I_BITSEXTANT;
+	qib_dev_err(dd, "Clearing reserved interrupt(s) 0x%016llx:"
+		    " %s\n", (unsigned long long) kills, msg);
+	qib_write_kreg(dd, kr_intmask, (dd->cspec->int_enable_mask & ~kills));
+}
+
+/* keep mainline interrupt handler cache-friendly */
+static noinline void unknown_7322_gpio_intr(struct qib_devdata *dd)
+{
+	u32 gpiostatus;
+	int handled = 0;
+	int pidx;
+
+	/*
+	 * Boards for this chip currently don't use GPIO interrupts,
+	 * so clear by writing GPIOstatus to GPIOclear, and complain
+	 * to developer.  To avoid endless repeats, clear
+	 * the bits in the mask, since there is some kind of
+	 * programming error or chip problem.
+	 */
+	gpiostatus = qib_read_kreg32(dd, kr_gpio_status);
+	/*
+	 * In theory, writing GPIOstatus to GPIOclear could
+	 * have a bad side-effect on some diagnostic that wanted
+	 * to poll for a status-change, but the various shadows
+	 * make that problematic at best. Diags will just suppress
+	 * all GPIO interrupts during such tests.
+	 */
+	qib_write_kreg(dd, kr_gpio_clear, gpiostatus);
+	/*
+	 * Check for QSFP MOD_PRS changes
+	 * only works for single port if IB1 != pidx1
+	 */
+	for (pidx = 0; pidx < dd->num_pports && (dd->flags & QIB_HAS_QSFP);
+	     ++pidx) {
+		struct qib_pportdata *ppd;
+		struct qib_qsfp_data *qd;
+		u32 mask;
+		if (!dd->pport[pidx].link_speed_supported)
+			continue;
+		mask = QSFP_GPIO_MOD_PRS_N;
+		ppd = dd->pport + pidx;
+		mask <<= (QSFP_GPIO_PORT2_SHIFT * ppd->hw_pidx);
+		if (gpiostatus & dd->cspec->gpio_mask & mask) {
+			u64 pins;
+			qd = &ppd->cpspec->qsfp_data;
+			gpiostatus &= ~mask;
+			pins = qib_read_kreg64(dd, kr_extstatus);
+			pins >>= SYM_LSB(EXTStatus, GPIOIn);
+			if (!(pins & mask)) {
+				++handled;
+				qd->t_insert = get_jiffies_64();
+				schedule_work(&qd->work);
+			}
+		}
+	}
+
+	if (gpiostatus && !handled) {
+		const u32 mask = qib_read_kreg32(dd, kr_gpio_mask);
+		u32 gpio_irq = mask & gpiostatus;
+
+		/*
+		 * Clear any troublemakers, and update chip from shadow
+		 */
+		dd->cspec->gpio_mask &= ~gpio_irq;
+		qib_write_kreg(dd, kr_gpio_mask, dd->cspec->gpio_mask);
+	}
+}
+
+/*
+ * Handle errors and unusual events first, separate function
+ * to improve cache hits for fast path interrupt handling.
+ */
+static noinline void unlikely_7322_intr(struct qib_devdata *dd, u64 istat)
+{
+	if (istat & ~QIB_I_BITSEXTANT)
+		unknown_7322_ibits(dd, istat);
+	if (istat & QIB_I_GPIO)
+		unknown_7322_gpio_intr(dd);
+	if (istat & QIB_I_C_ERROR)
+		handle_7322_errors(dd);
+	if (istat & INT_MASK_P(Err, 0) && dd->rcd[0])
+		handle_7322_p_errors(dd->rcd[0]->ppd);
+	if (istat & INT_MASK_P(Err, 1) && dd->rcd[1])
+		handle_7322_p_errors(dd->rcd[1]->ppd);
+}
+
+/*
+ * Dynamically adjust the rcv int timeout for a context based on incoming
+ * packet rate.
+ */
+static void adjust_rcv_timeout(struct qib_ctxtdata *rcd, int npkts)
+{
+	struct qib_devdata *dd = rcd->dd;
+	u32 timeout = dd->cspec->rcvavail_timeout[rcd->ctxt];
+
+	/*
+	 * Dynamically adjust idle timeout on chip
+	 * based on number of packets processed.
+	 */
+	if (npkts < rcv_int_count && timeout > 2)
+		timeout >>= 1;
+	else if (npkts >= rcv_int_count && timeout < rcv_int_timeout)
+		timeout = min(timeout << 1, rcv_int_timeout);
+	else
+		return;
+
+	dd->cspec->rcvavail_timeout[rcd->ctxt] = timeout;
+	qib_write_kreg(dd, kr_rcvavailtimeout + rcd->ctxt, timeout);
+}
+
+/*
+ * This is the main interrupt handler.
+ * It will normally only be used for low frequency interrupts but may
+ * have to handle all interrupts if INTx is enabled or fewer than normal
+ * MSIx interrupts were allocated.
+ * This routine should ignore the interrupt bits for any of the
+ * dedicated MSIx handlers.
+ */
+static irqreturn_t qib_7322intr(int irq, void *data)
+{
+	struct qib_devdata *dd = data;
+	irqreturn_t ret;
+	u64 istat;
+	u64 ctxtrbits;
+	u64 rmask;
+	unsigned i;
+	u32 npkts;
+
+	if ((dd->flags & (QIB_PRESENT | QIB_BADINTR)) != QIB_PRESENT) {
+		/*
+		 * This return value is not great, but we do not want the
+		 * interrupt core code to remove our interrupt handler
+		 * because we don't appear to be handling an interrupt
+		 * during a chip reset.
+		 */
+		ret = IRQ_HANDLED;
+		goto bail;
+	}
+
+	istat = qib_read_kreg64(dd, kr_intstatus);
+
+	if (unlikely(istat == ~0ULL)) {
+		qib_bad_intrstatus(dd);
+		qib_dev_err(dd, "Interrupt status all f's, skipping\n");
+		/* don't know if it was our interrupt or not */
+		ret = IRQ_NONE;
+		goto bail;
+	}
+
+	istat &= dd->cspec->main_int_mask;
+	if (unlikely(!istat)) {
+		/* already handled, or shared and not us */
+		ret = IRQ_NONE;
+		goto bail;
+	}
+
+	qib_stats.sps_ints++;
+	if (dd->int_counter != (u32) -1)
+		dd->int_counter++;
+
+	/* handle "errors" of various kinds first, device ahead of port */
+	if (unlikely(istat & (~QIB_I_BITSEXTANT | QIB_I_GPIO |
+			      QIB_I_C_ERROR | INT_MASK_P(Err, 0) |
+			      INT_MASK_P(Err, 1))))
+		unlikely_7322_intr(dd, istat);
+
+	/*
+	 * Clear the interrupt bits we found set, relatively early, so we
+	 * "know" know the chip will have seen this by the time we process
+	 * the queue, and will re-interrupt if necessary.  The processor
+	 * itself won't take the interrupt again until we return.
+	 */
+	qib_write_kreg(dd, kr_intclear, istat);
+
+	/*
+	 * Handle kernel receive queues before checking for pio buffers
+	 * available since receives can overflow; piobuf waiters can afford
+	 * a few extra cycles, since they were waiting anyway.
+	 */
+	ctxtrbits = istat & (QIB_I_RCVAVAIL_MASK | QIB_I_RCVURG_MASK);
+	if (ctxtrbits) {
+		rmask = (1ULL << QIB_I_RCVAVAIL_LSB) |
+			(1ULL << QIB_I_RCVURG_LSB);
+		for (i = 0; i < dd->first_user_ctxt; i++) {
+			if (ctxtrbits & rmask) {
+				ctxtrbits &= ~rmask;
+				if (dd->rcd[i]) {
+					qib_kreceive(dd->rcd[i], NULL, &npkts);
+					adjust_rcv_timeout(dd->rcd[i], npkts);
+				}
+			}
+			rmask <<= 1;
+		}
+		if (ctxtrbits) {
+			ctxtrbits = (ctxtrbits >> QIB_I_RCVAVAIL_LSB) |
+				(ctxtrbits >> QIB_I_RCVURG_LSB);
+			qib_handle_urcv(dd, ctxtrbits);
+		}
+	}
+
+	if (istat & (QIB_I_P_SDMAINT(0) | QIB_I_P_SDMAINT(1)))
+		sdma_7322_intr(dd, istat);
+
+	if ((istat & QIB_I_SPIOBUFAVAIL) && (dd->flags & QIB_INITTED))
+		qib_ib_piobufavail(dd);
+
+	ret = IRQ_HANDLED;
+bail:
+	return ret;
+}
+
+/*
+ * Dedicated receive packet available interrupt handler.
+ */
+static irqreturn_t qib_7322pintr(int irq, void *data)
+{
+	struct qib_ctxtdata *rcd = data;
+	struct qib_devdata *dd = rcd->dd;
+	u32 npkts;
+
+	if ((dd->flags & (QIB_PRESENT | QIB_BADINTR)) != QIB_PRESENT)
+		/*
+		 * This return value is not great, but we do not want the
+		 * interrupt core code to remove our interrupt handler
+		 * because we don't appear to be handling an interrupt
+		 * during a chip reset.
+		 */
+		return IRQ_HANDLED;
+
+	qib_stats.sps_ints++;
+	if (dd->int_counter != (u32) -1)
+		dd->int_counter++;
+
+	/* Clear the interrupt bit we expect to be set. */
+	qib_write_kreg(dd, kr_intclear, ((1ULL << QIB_I_RCVAVAIL_LSB) |
+		       (1ULL << QIB_I_RCVURG_LSB)) << rcd->ctxt);
+
+	qib_kreceive(rcd, NULL, &npkts);
+	adjust_rcv_timeout(rcd, npkts);
+
+	return IRQ_HANDLED;
+}
+
+/*
+ * Dedicated Send buffer available interrupt handler.
+ */
+static irqreturn_t qib_7322bufavail(int irq, void *data)
+{
+	struct qib_devdata *dd = data;
+
+	if ((dd->flags & (QIB_PRESENT | QIB_BADINTR)) != QIB_PRESENT)
+		/*
+		 * This return value is not great, but we do not want the
+		 * interrupt core code to remove our interrupt handler
+		 * because we don't appear to be handling an interrupt
+		 * during a chip reset.
+		 */
+		return IRQ_HANDLED;
+
+	qib_stats.sps_ints++;
+	if (dd->int_counter != (u32) -1)
+		dd->int_counter++;
+
+	/* Clear the interrupt bit we expect to be set. */
+	qib_write_kreg(dd, kr_intclear, QIB_I_SPIOBUFAVAIL);
+
+	/* qib_ib_piobufavail() will clear the want PIO interrupt if needed */
+	if (dd->flags & QIB_INITTED)
+		qib_ib_piobufavail(dd);
+	else
+		qib_wantpiobuf_7322_intr(dd, 0);
+
+	return IRQ_HANDLED;
+}
+
+/*
+ * Dedicated Send DMA interrupt handler.
+ */
+static irqreturn_t sdma_intr(int irq, void *data)
+{
+	struct qib_pportdata *ppd = data;
+	struct qib_devdata *dd = ppd->dd;
+
+	if ((dd->flags & (QIB_PRESENT | QIB_BADINTR)) != QIB_PRESENT)
+		/*
+		 * This return value is not great, but we do not want the
+		 * interrupt core code to remove our interrupt handler
+		 * because we don't appear to be handling an interrupt
+		 * during a chip reset.
+		 */
+		return IRQ_HANDLED;
+
+	qib_stats.sps_ints++;
+	if (dd->int_counter != (u32) -1)
+		dd->int_counter++;
+
+	/* Clear the interrupt bit we expect to be set. */
+	qib_write_kreg(dd, kr_intclear, ppd->hw_pidx ?
+		       INT_MASK_P(SDma, 1) : INT_MASK_P(SDma, 0));
+	qib_sdma_intr(ppd);
+
+	return IRQ_HANDLED;
+}
+
+/*
+ * Dedicated Send DMA idle interrupt handler.
+ */
+static irqreturn_t sdma_idle_intr(int irq, void *data)
+{
+	struct qib_pportdata *ppd = data;
+	struct qib_devdata *dd = ppd->dd;
+
+	if ((dd->flags & (QIB_PRESENT | QIB_BADINTR)) != QIB_PRESENT)
+		/*
+		 * This return value is not great, but we do not want the
+		 * interrupt core code to remove our interrupt handler
+		 * because we don't appear to be handling an interrupt
+		 * during a chip reset.
+		 */
+		return IRQ_HANDLED;
+
+	qib_stats.sps_ints++;
+	if (dd->int_counter != (u32) -1)
+		dd->int_counter++;
+
+	/* Clear the interrupt bit we expect to be set. */
+	qib_write_kreg(dd, kr_intclear, ppd->hw_pidx ?
+		       INT_MASK_P(SDmaIdle, 1) : INT_MASK_P(SDmaIdle, 0));
+	qib_sdma_intr(ppd);
+
+	return IRQ_HANDLED;
+}
+
+/*
+ * Dedicated Send DMA progress interrupt handler.
+ */
+static irqreturn_t sdma_progress_intr(int irq, void *data)
+{
+	struct qib_pportdata *ppd = data;
+	struct qib_devdata *dd = ppd->dd;
+
+	if ((dd->flags & (QIB_PRESENT | QIB_BADINTR)) != QIB_PRESENT)
+		/*
+		 * This return value is not great, but we do not want the
+		 * interrupt core code to remove our interrupt handler
+		 * because we don't appear to be handling an interrupt
+		 * during a chip reset.
+		 */
+		return IRQ_HANDLED;
+
+	qib_stats.sps_ints++;
+	if (dd->int_counter != (u32) -1)
+		dd->int_counter++;
+
+	/* Clear the interrupt bit we expect to be set. */
+	qib_write_kreg(dd, kr_intclear, ppd->hw_pidx ?
+		       INT_MASK_P(SDmaProgress, 1) :
+		       INT_MASK_P(SDmaProgress, 0));
+	qib_sdma_intr(ppd);
+
+	return IRQ_HANDLED;
+}
+
+/*
+ * Dedicated Send DMA cleanup interrupt handler.
+ */
+static irqreturn_t sdma_cleanup_intr(int irq, void *data)
+{
+	struct qib_pportdata *ppd = data;
+	struct qib_devdata *dd = ppd->dd;
+
+	if ((dd->flags & (QIB_PRESENT | QIB_BADINTR)) != QIB_PRESENT)
+		/*
+		 * This return value is not great, but we do not want the
+		 * interrupt core code to remove our interrupt handler
+		 * because we don't appear to be handling an interrupt
+		 * during a chip reset.
+		 */
+		return IRQ_HANDLED;
+
+	qib_stats.sps_ints++;
+	if (dd->int_counter != (u32) -1)
+		dd->int_counter++;
+
+	/* Clear the interrupt bit we expect to be set. */
+	qib_write_kreg(dd, kr_intclear, ppd->hw_pidx ?
+		       INT_MASK_PM(SDmaCleanupDone, 1) :
+		       INT_MASK_PM(SDmaCleanupDone, 0));
+	qib_sdma_process_event(ppd, qib_sdma_event_e20_hw_started);
+
+	return IRQ_HANDLED;
+}
+
+/*
+ * Set up our chip-specific interrupt handler.
+ * The interrupt type has already been setup, so
+ * we just need to do the registration and error checking.
+ * If we are using MSIx interrupts, we may fall back to
+ * INTx later, if the interrupt handler doesn't get called
+ * within 1/2 second (see verify_interrupt()).
+ */
+static void qib_setup_7322_interrupt(struct qib_devdata *dd, int clearpend)
+{
+	int ret, i, msixnum;
+	u64 redirect[6];
+	u64 mask;
+
+	if (!dd->num_pports)
+		return;
+
+	if (clearpend) {
+		/*
+		 * if not switching interrupt types, be sure interrupts are
+		 * disabled, and then clear anything pending at this point,
+		 * because we are starting clean.
+		 */
+		qib_7322_set_intr_state(dd, 0);
+
+		/* clear the reset error, init error/hwerror mask */
+		qib_7322_init_hwerrors(dd);
+
+		/* clear any interrupt bits that might be set */
+		qib_write_kreg(dd, kr_intclear, ~0ULL);
+
+		/* make sure no pending MSIx intr, and clear diag reg */
+		qib_write_kreg(dd, kr_intgranted, ~0ULL);
+		qib_write_kreg(dd, kr_vecclr_wo_int, ~0ULL);
+	}
+
+	if (!dd->cspec->num_msix_entries) {
+		/* Try to get INTx interrupt */
+try_intx:
+		if (!dd->pcidev->irq) {
+			qib_dev_err(dd, "irq is 0, BIOS error?  "
+				    "Interrupts won't work\n");
+			goto bail;
+		}
+		ret = request_irq(dd->pcidev->irq, qib_7322intr,
+				  IRQF_SHARED, QIB_DRV_NAME, dd);
+		if (ret) {
+			qib_dev_err(dd, "Couldn't setup INTx "
+				    "interrupt (irq=%d): %d\n",
+				    dd->pcidev->irq, ret);
+			goto bail;
+		}
+		dd->cspec->irq = dd->pcidev->irq;
+		dd->cspec->main_int_mask = ~0ULL;
+		goto bail;
+	}
+
+	/* Try to get MSIx interrupts */
+	memset(redirect, 0, sizeof redirect);
+	mask = ~0ULL;
+	msixnum = 0;
+	for (i = 0; msixnum < dd->cspec->num_msix_entries; i++) {
+		irq_handler_t handler;
+		const char *name;
+		void *arg;
+		u64 val;
+		int lsb, reg, sh;
+
+		if (i < ARRAY_SIZE(irq_table)) {
+			if (irq_table[i].port) {
+				/* skip if for a non-configured port */
+				if (irq_table[i].port > dd->num_pports)
+					continue;
+				arg = dd->pport + irq_table[i].port - 1;
+			} else
+				arg = dd;
+			lsb = irq_table[i].lsb;
+			handler = irq_table[i].handler;
+			name = irq_table[i].name;
+		} else {
+			unsigned ctxt;
+
+			ctxt = i - ARRAY_SIZE(irq_table);
+			/* per krcvq context receive interrupt */
+			arg = dd->rcd[ctxt];
+			if (!arg)
+				continue;
+			lsb = QIB_I_RCVAVAIL_LSB + ctxt;
+			handler = qib_7322pintr;
+			name = QIB_DRV_NAME " (kctx)";
+		}
+		ret = request_irq(dd->cspec->msix_entries[msixnum].vector,
+				  handler, 0, name, arg);
+		if (ret) {
+			/*
+			 * Shouldn't happen since the enable said we could
+			 * have as many as we are trying to setup here.
+			 */
+			qib_dev_err(dd, "Couldn't setup MSIx "
+				    "interrupt (vec=%d, irq=%d): %d\n", msixnum,
+				    dd->cspec->msix_entries[msixnum].vector,
+				    ret);
+			qib_7322_nomsix(dd);
+			goto try_intx;
+		}
+		dd->cspec->msix_arg[msixnum] = arg;
+		if (lsb >= 0) {
+			reg = lsb / IBA7322_REDIRECT_VEC_PER_REG;
+			sh = (lsb % IBA7322_REDIRECT_VEC_PER_REG) *
+				SYM_LSB(IntRedirect0, vec1);
+			mask &= ~(1ULL << lsb);
+			redirect[reg] |= ((u64) msixnum) << sh;
+		}
+		val = qib_read_kreg64(dd, 2 * msixnum + 1 +
+			(QIB_7322_MsixTable_OFFS / sizeof(u64)));
+		msixnum++;
+	}
+	/* Initialize the vector mapping */
+	for (i = 0; i < ARRAY_SIZE(redirect); i++)
+		qib_write_kreg(dd, kr_intredirect + i, redirect[i]);
+	dd->cspec->main_int_mask = mask;
+bail:;
+}
+
+/**
+ * qib_7322_boardname - fill in the board name and note features
+ * @dd: the qlogic_ib device
+ *
+ * info will be based on the board revision register
+ */
+static unsigned qib_7322_boardname(struct qib_devdata *dd)
+{
+	/* Will need enumeration of board-types here */
+	char *n;
+	u32 boardid, namelen;
+	unsigned features = DUAL_PORT_CAP;
+
+	boardid = SYM_FIELD(dd->revision, Revision, BoardID);
+
+	switch (boardid) {
+	case 0:
+		n = "InfiniPath_QLE7342_Emulation";
+		break;
+	case 1:
+		n = "InfiniPath_QLE7340";
+		dd->flags |= QIB_HAS_QSFP;
+		features = PORT_SPD_CAP;
+		break;
+	case 2:
+		n = "InfiniPath_QLE7342";
+		dd->flags |= QIB_HAS_QSFP;
+		break;
+	case 3:
+		n = "InfiniPath_QMI7342";
+		break;
+	case 4:
+		n = "InfiniPath_Unsupported7342";
+		qib_dev_err(dd, "Unsupported version of QMH7342\n");
+		features = 0;
+		break;
+	case BOARD_QMH7342:
+		n = "InfiniPath_QMH7342";
+		features = 0x24;
+		break;
+	case BOARD_QME7342:
+		n = "InfiniPath_QME7342";
+		break;
+	case 15:
+		n = "InfiniPath_QLE7342_TEST";
+		dd->flags |= QIB_HAS_QSFP;
+		break;
+	default:
+		n = "InfiniPath_QLE73xy_UNKNOWN";
+		qib_dev_err(dd, "Unknown 7322 board type %u\n", boardid);
+		break;
+	}
+	dd->board_atten = 1; /* index into txdds_Xdr */
+
+	namelen = strlen(n) + 1;
+	dd->boardname = kmalloc(namelen, GFP_KERNEL);
+	if (!dd->boardname)
+		qib_dev_err(dd, "Failed allocation for board name: %s\n", n);
+	else
+		snprintf(dd->boardname, namelen, "%s", n);
+
+	snprintf(dd->boardversion, sizeof(dd->boardversion),
+		 "ChipABI %u.%u, %s, InfiniPath%u %u.%u, SW Compat %u\n",
+		 QIB_CHIP_VERS_MAJ, QIB_CHIP_VERS_MIN, dd->boardname,
+		 (unsigned)SYM_FIELD(dd->revision, Revision_R, Arch),
+		 dd->majrev, dd->minrev,
+		 (unsigned)SYM_FIELD(dd->revision, Revision_R, SW));
+
+	if (qib_singleport && (features >> PORT_SPD_CAP_SHIFT) & PORT_SPD_CAP) {
+		qib_devinfo(dd->pcidev, "IB%u: Forced to single port mode"
+			    " by module parameter\n", dd->unit);
+		features &= PORT_SPD_CAP;
+	}
+
+	return features;
+}
+
+/*
+ * This routine sleeps, so it can only be called from user context, not
+ * from interrupt context.
+ */
+static int qib_do_7322_reset(struct qib_devdata *dd)
+{
+	u64 val;
+	u64 *msix_vecsave;
+	int i, msix_entries, ret = 1;
+	u16 cmdval;
+	u8 int_line, clinesz;
+	unsigned long flags;
+
+	/* Use dev_err so it shows up in logs, etc. */
+	qib_dev_err(dd, "Resetting InfiniPath unit %u\n", dd->unit);
+
+	qib_pcie_getcmd(dd, &cmdval, &int_line, &clinesz);
+
+	msix_entries = dd->cspec->num_msix_entries;
+
+	/* no interrupts till re-initted */
+	qib_7322_set_intr_state(dd, 0);
+
+	if (msix_entries) {
+		qib_7322_nomsix(dd);
+		/* can be up to 512 bytes, too big for stack */
+		msix_vecsave = kmalloc(2 * dd->cspec->num_msix_entries *
+			sizeof(u64), GFP_KERNEL);
+		if (!msix_vecsave)
+			qib_dev_err(dd, "No mem to save MSIx data\n");
+	} else
+		msix_vecsave = NULL;
+
+	/*
+	 * Core PCI (as of 2.6.18) doesn't save or rewrite the full vector
+	 * info that is set up by the BIOS, so we have to save and restore
+	 * it ourselves.   There is some risk something could change it,
+	 * after we save it, but since we have disabled the MSIx, it
+	 * shouldn't be touched...
+	 */
+	for (i = 0; i < msix_entries; i++) {
+		u64 vecaddr, vecdata;
+		vecaddr = qib_read_kreg64(dd, 2 * i +
+				  (QIB_7322_MsixTable_OFFS / sizeof(u64)));
+		vecdata = qib_read_kreg64(dd, 1 + 2 * i +
+				  (QIB_7322_MsixTable_OFFS / sizeof(u64)));
+		if (msix_vecsave) {
+			msix_vecsave[2 * i] = vecaddr;
+			/* save it without the masked bit set */
+			msix_vecsave[1 + 2 * i] = vecdata & ~0x100000000ULL;
+		}
+	}
+
+	dd->pport->cpspec->ibdeltainprog = 0;
+	dd->pport->cpspec->ibsymdelta = 0;
+	dd->pport->cpspec->iblnkerrdelta = 0;
+	dd->pport->cpspec->ibmalfdelta = 0;
+	dd->int_counter = 0; /* so we check interrupts work again */
+
+	/*
+	 * Keep chip from being accessed until we are ready.  Use
+	 * writeq() directly, to allow the write even though QIB_PRESENT
+	 * isnt' set.
+	 */
+	dd->flags &= ~(QIB_INITTED | QIB_PRESENT | QIB_BADINTR);
+	dd->flags |= QIB_DOING_RESET;
+	val = dd->control | QLOGIC_IB_C_RESET;
+	writeq(val, &dd->kregbase[kr_control]);
+
+	for (i = 1; i <= 5; i++) {
+		/*
+		 * Allow MBIST, etc. to complete; longer on each retry.
+		 * We sometimes get machine checks from bus timeout if no
+		 * response, so for now, make it *really* long.
+		 */
+		msleep(1000 + (1 + i) * 3000);
+
+		qib_pcie_reenable(dd, cmdval, int_line, clinesz);
+
+		/*
+		 * Use readq directly, so we don't need to mark it as PRESENT
+		 * until we get a successful indication that all is well.
+		 */
+		val = readq(&dd->kregbase[kr_revision]);
+		if (val == dd->revision)
+			break;
+		if (i == 5) {
+			qib_dev_err(dd, "Failed to initialize after reset, "
+				    "unusable\n");
+			ret = 0;
+			goto  bail;
+		}
+	}
+
+	dd->flags |= QIB_PRESENT; /* it's back */
+
+	if (msix_entries) {
+		/* restore the MSIx vector address and data if saved above */
+		for (i = 0; i < msix_entries; i++) {
+			dd->cspec->msix_entries[i].entry = i;
+			if (!msix_vecsave || !msix_vecsave[2 * i])
+				continue;
+			qib_write_kreg(dd, 2 * i +
+				(QIB_7322_MsixTable_OFFS / sizeof(u64)),
+				msix_vecsave[2 * i]);
+			qib_write_kreg(dd, 1 + 2 * i +
+				(QIB_7322_MsixTable_OFFS / sizeof(u64)),
+				msix_vecsave[1 + 2 * i]);
+		}
+	}
+
+	/* initialize the remaining registers.  */
+	for (i = 0; i < dd->num_pports; ++i)
+		write_7322_init_portregs(&dd->pport[i]);
+	write_7322_initregs(dd);
+
+	if (qib_pcie_params(dd, dd->lbus_width,
+			    &dd->cspec->num_msix_entries,
+			    dd->cspec->msix_entries))
+		qib_dev_err(dd, "Reset failed to setup PCIe or interrupts; "
+				"continuing anyway\n");
+
+	qib_setup_7322_interrupt(dd, 1);
+
+	for (i = 0; i < dd->num_pports; ++i) {
+		struct qib_pportdata *ppd = &dd->pport[i];
+
+		spin_lock_irqsave(&ppd->lflags_lock, flags);
+		ppd->lflags |= QIBL_IB_FORCE_NOTIFY;
+		ppd->lflags &= ~QIBL_IB_AUTONEG_FAILED;
+		spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+	}
+
+bail:
+	dd->flags &= ~QIB_DOING_RESET; /* OK or not, no longer resetting */
+	kfree(msix_vecsave);
+	return ret;
+}
+
+/**
+ * qib_7322_put_tid - write a TID to the chip
+ * @dd: the qlogic_ib device
+ * @tidptr: pointer to the expected TID (in chip) to update
+ * @tidtype: 0 for eager, 1 for expected
+ * @pa: physical address of in memory buffer; tidinvalid if freeing
+ */
+static void qib_7322_put_tid(struct qib_devdata *dd, u64 __iomem *tidptr,
+			     u32 type, unsigned long pa)
+{
+	if (!(dd->flags & QIB_PRESENT))
+		return;
+	if (pa != dd->tidinvalid) {
+		u64 chippa = pa >> IBA7322_TID_PA_SHIFT;
+
+		/* paranoia checks */
+		if (pa != (chippa << IBA7322_TID_PA_SHIFT)) {
+			qib_dev_err(dd, "Physaddr %lx not 2KB aligned!\n",
+				    pa);
+			return;
+		}
+		if (chippa >= (1UL << IBA7322_TID_SZ_SHIFT)) {
+			qib_dev_err(dd, "Physical page address 0x%lx "
+				"larger than supported\n", pa);
+			return;
+		}
+
+		if (type == RCVHQ_RCV_TYPE_EAGER)
+			chippa |= dd->tidtemplate;
+		else /* for now, always full 4KB page */
+			chippa |= IBA7322_TID_SZ_4K;
+		pa = chippa;
+	}
+	writeq(pa, tidptr);
+	mmiowb();
+}
+
+/**
+ * qib_7322_clear_tids - clear all TID entries for a ctxt, expected and eager
+ * @dd: the qlogic_ib device
+ * @ctxt: the ctxt
+ *
+ * clear all TID entries for a ctxt, expected and eager.
+ * Used from qib_close().
+ */
+static void qib_7322_clear_tids(struct qib_devdata *dd,
+				struct qib_ctxtdata *rcd)
+{
+	u64 __iomem *tidbase;
+	unsigned long tidinv;
+	u32 ctxt;
+	int i;
+
+	if (!dd->kregbase || !rcd)
+		return;
+
+	ctxt = rcd->ctxt;
+
+	tidinv = dd->tidinvalid;
+	tidbase = (u64 __iomem *)
+		((char __iomem *) dd->kregbase +
+		 dd->rcvtidbase +
+		 ctxt * dd->rcvtidcnt * sizeof(*tidbase));
+
+	for (i = 0; i < dd->rcvtidcnt; i++)
+		qib_7322_put_tid(dd, &tidbase[i], RCVHQ_RCV_TYPE_EXPECTED,
+				 tidinv);
+
+	tidbase = (u64 __iomem *)
+		((char __iomem *) dd->kregbase +
+		 dd->rcvegrbase +
+		 rcd->rcvegr_tid_base * sizeof(*tidbase));
+
+	for (i = 0; i < rcd->rcvegrcnt; i++)
+		qib_7322_put_tid(dd, &tidbase[i], RCVHQ_RCV_TYPE_EAGER,
+				 tidinv);
+}
+
+/**
+ * qib_7322_tidtemplate - setup constants for TID updates
+ * @dd: the qlogic_ib device
+ *
+ * We setup stuff that we use a lot, to avoid calculating each time
+ */
+static void qib_7322_tidtemplate(struct qib_devdata *dd)
+{
+	/*
+	 * For now, we always allocate 4KB buffers (at init) so we can
+	 * receive max size packets.  We may want a module parameter to
+	 * specify 2KB or 4KB and/or make it per port instead of per device
+	 * for those who want to reduce memory footprint.  Note that the
+	 * rcvhdrentsize size must be large enough to hold the largest
+	 * IB header (currently 96 bytes) that we expect to handle (plus of
+	 * course the 2 dwords of RHF).
+	 */
+	if (dd->rcvegrbufsize == 2048)
+		dd->tidtemplate = IBA7322_TID_SZ_2K;
+	else if (dd->rcvegrbufsize == 4096)
+		dd->tidtemplate = IBA7322_TID_SZ_4K;
+	dd->tidinvalid = 0;
+}
+
+/**
+ * qib_init_7322_get_base_info - set chip-specific flags for user code
+ * @rcd: the qlogic_ib ctxt
+ * @kbase: qib_base_info pointer
+ *
+ * We set the PCIE flag because the lower bandwidth on PCIe vs
+ * HyperTransport can affect some user packet algorithims.
+ */
+
+static int qib_7322_get_base_info(struct qib_ctxtdata *rcd,
+				  struct qib_base_info *kinfo)
+{
+	kinfo->spi_runtime_flags |= QIB_RUNTIME_CTXT_MSB_IN_QP |
+		QIB_RUNTIME_PCIE | QIB_RUNTIME_NODMA_RTAIL |
+		QIB_RUNTIME_HDRSUPP | QIB_RUNTIME_SDMA;
+	if (rcd->dd->cspec->r1)
+		kinfo->spi_runtime_flags |= QIB_RUNTIME_RCHK;
+	if (rcd->dd->flags & QIB_USE_SPCL_TRIG)
+		kinfo->spi_runtime_flags |= QIB_RUNTIME_SPECIAL_TRIGGER;
+
+	return 0;
+}
+
+static struct qib_message_header *
+qib_7322_get_msgheader(struct qib_devdata *dd, __le32 *rhf_addr)
+{
+	u32 offset = qib_hdrget_offset(rhf_addr);
+
+	return (struct qib_message_header *)
+		(rhf_addr - dd->rhf_offset + offset);
+}
+
+/*
+ * Configure number of contexts.
+ */
+static void qib_7322_config_ctxts(struct qib_devdata *dd)
+{
+	unsigned long flags;
+	u32 nchipctxts;
+
+	nchipctxts = qib_read_kreg32(dd, kr_contextcnt);
+	dd->cspec->numctxts = nchipctxts;
+	if (qib_n_krcv_queues > 1 && dd->num_pports) {
+		/*
+		 * Set the mask for which bits from the QPN are used
+		 * to select a context number.
+		 */
+		dd->qpn_mask = 0x3f;
+		dd->first_user_ctxt = NUM_IB_PORTS +
+			(qib_n_krcv_queues - 1) * dd->num_pports;
+		if (dd->first_user_ctxt > nchipctxts)
+			dd->first_user_ctxt = nchipctxts;
+		dd->n_krcv_queues = dd->first_user_ctxt / dd->num_pports;
+	} else {
+		dd->first_user_ctxt = NUM_IB_PORTS;
+		dd->n_krcv_queues = 1;
+	}
+
+	if (!qib_cfgctxts) {
+		int nctxts = dd->first_user_ctxt + num_online_cpus();
+
+		if (nctxts <= 6)
+			dd->ctxtcnt = 6;
+		else if (nctxts <= 10)
+			dd->ctxtcnt = 10;
+		else if (nctxts <= nchipctxts)
+			dd->ctxtcnt = nchipctxts;
+	} else if (qib_cfgctxts < dd->num_pports)
+		dd->ctxtcnt = dd->num_pports;
+	else if (qib_cfgctxts <= nchipctxts)
+		dd->ctxtcnt = qib_cfgctxts;
+	if (!dd->ctxtcnt) /* none of the above, set to max */
+		dd->ctxtcnt = nchipctxts;
+
+	/*
+	 * Chip can be configured for 6, 10, or 18 ctxts, and choice
+	 * affects number of eager TIDs per ctxt (1K, 2K, 4K).
+	 * Lock to be paranoid about later motion, etc.
+	 */
+	spin_lock_irqsave(&dd->cspec->rcvmod_lock, flags);
+	if (dd->ctxtcnt > 10)
+		dd->rcvctrl |= 2ULL << SYM_LSB(RcvCtrl, ContextCfg);
+	else if (dd->ctxtcnt > 6)
+		dd->rcvctrl |= 1ULL << SYM_LSB(RcvCtrl, ContextCfg);
+	/* else configure for default 6 receive ctxts */
+
+	/* The XRC opcode is 5. */
+	dd->rcvctrl |= 5ULL << SYM_LSB(RcvCtrl, XrcTypeCode);
+
+	/*
+	 * RcvCtrl *must* be written here so that the
+	 * chip understands how to change rcvegrcnt below.
+	 */
+	qib_write_kreg(dd, kr_rcvctrl, dd->rcvctrl);
+	spin_unlock_irqrestore(&dd->cspec->rcvmod_lock, flags);
+
+	/* kr_rcvegrcnt changes based on the number of contexts enabled */
+	dd->cspec->rcvegrcnt = qib_read_kreg32(dd, kr_rcvegrcnt);
+	dd->rcvhdrcnt = max(dd->cspec->rcvegrcnt,
+				dd->num_pports > 1 ? 1024U : 2048U);
+}
+
+static int qib_7322_get_ib_cfg(struct qib_pportdata *ppd, int which)
+{
+
+	int lsb, ret = 0;
+	u64 maskr; /* right-justified mask */
+
+	switch (which) {
+
+	case QIB_IB_CFG_LWID_ENB: /* Get allowed Link-width */
+		ret = ppd->link_width_enabled;
+		goto done;
+
+	case QIB_IB_CFG_LWID: /* Get currently active Link-width */
+		ret = ppd->link_width_active;
+		goto done;
+
+	case QIB_IB_CFG_SPD_ENB: /* Get allowed Link speeds */
+		ret = ppd->link_speed_enabled;
+		goto done;
+
+	case QIB_IB_CFG_SPD: /* Get current Link spd */
+		ret = ppd->link_speed_active;
+		goto done;
+
+	case QIB_IB_CFG_RXPOL_ENB: /* Get Auto-RX-polarity enable */
+		lsb = SYM_LSB(IBCCtrlB_0, IB_POLARITY_REV_SUPP);
+		maskr = SYM_RMASK(IBCCtrlB_0, IB_POLARITY_REV_SUPP);
+		break;
+
+	case QIB_IB_CFG_LREV_ENB: /* Get Auto-Lane-reversal enable */
+		lsb = SYM_LSB(IBCCtrlB_0, IB_LANE_REV_SUPPORTED);
+		maskr = SYM_RMASK(IBCCtrlB_0, IB_LANE_REV_SUPPORTED);
+		break;
+
+	case QIB_IB_CFG_LINKLATENCY:
+		ret = qib_read_kreg_port(ppd, krp_ibcstatus_b) &
+			SYM_MASK(IBCStatusB_0, LinkRoundTripLatency);
+		goto done;
+
+	case QIB_IB_CFG_OP_VLS:
+		ret = ppd->vls_operational;
+		goto done;
+
+	case QIB_IB_CFG_VL_HIGH_CAP:
+		ret = 16;
+		goto done;
+
+	case QIB_IB_CFG_VL_LOW_CAP:
+		ret = 16;
+		goto done;
+
+	case QIB_IB_CFG_OVERRUN_THRESH: /* IB overrun threshold */
+		ret = SYM_FIELD(ppd->cpspec->ibcctrl_a, IBCCtrlA_0,
+				OverrunThreshold);
+		goto done;
+
+	case QIB_IB_CFG_PHYERR_THRESH: /* IB PHY error threshold */
+		ret = SYM_FIELD(ppd->cpspec->ibcctrl_a, IBCCtrlA_0,
+				PhyerrThreshold);
+		goto done;
+
+	case QIB_IB_CFG_LINKDEFAULT: /* IB link default (sleep/poll) */
+		/* will only take effect when the link state changes */
+		ret = (ppd->cpspec->ibcctrl_a &
+		       SYM_MASK(IBCCtrlA_0, LinkDownDefaultState)) ?
+			IB_LINKINITCMD_SLEEP : IB_LINKINITCMD_POLL;
+		goto done;
+
+	case QIB_IB_CFG_HRTBT: /* Get Heartbeat off/enable/auto */
+		lsb = IBA7322_IBC_HRTBT_LSB;
+		maskr = IBA7322_IBC_HRTBT_RMASK; /* OR of AUTO and ENB */
+		break;
+
+	case QIB_IB_CFG_PMA_TICKS:
+		/*
+		 * 0x00 = 10x link transfer rate or 4 nsec. for 2.5Gbs
+		 * Since the clock is always 250MHz, the value is 3, 1 or 0.
+		 */
+		if (ppd->link_speed_active == QIB_IB_QDR)
+			ret = 3;
+		else if (ppd->link_speed_active == QIB_IB_DDR)
+			ret = 1;
+		else
+			ret = 0;
+		goto done;
+
+	default:
+		ret = -EINVAL;
+		goto done;
+	}
+	ret = (int)((ppd->cpspec->ibcctrl_b >> lsb) & maskr);
+done:
+	return ret;
+}
+
+/*
+ * Below again cribbed liberally from older version. Do not lean
+ * heavily on it.
+ */
+#define IBA7322_IBC_DLIDLMC_SHIFT QIB_7322_IBCCtrlB_0_IB_DLID_LSB
+#define IBA7322_IBC_DLIDLMC_MASK (QIB_7322_IBCCtrlB_0_IB_DLID_RMASK \
+	| (QIB_7322_IBCCtrlB_0_IB_DLID_MASK_RMASK << 16))
+
+static int qib_7322_set_ib_cfg(struct qib_pportdata *ppd, int which, u32 val)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u64 maskr; /* right-justified mask */
+	int lsb, ret = 0;
+	u16 lcmd, licmd;
+	unsigned long flags;
+
+	switch (which) {
+	case QIB_IB_CFG_LIDLMC:
+		/*
+		 * Set LID and LMC. Combined to avoid possible hazard
+		 * caller puts LMC in 16MSbits, DLID in 16LSbits of val
+		 */
+		lsb = IBA7322_IBC_DLIDLMC_SHIFT;
+		maskr = IBA7322_IBC_DLIDLMC_MASK;
+		/*
+		 * For header-checking, the SLID in the packet will
+		 * be masked with SendIBSLMCMask, and compared
+		 * with SendIBSLIDAssignMask. Make sure we do not
+		 * set any bits not covered by the mask, or we get
+		 * false-positives.
+		 */
+		qib_write_kreg_port(ppd, krp_sendslid,
+				    val & (val >> 16) & SendIBSLIDAssignMask);
+		qib_write_kreg_port(ppd, krp_sendslidmask,
+				    (val >> 16) & SendIBSLMCMask);
+		break;
+
+	case QIB_IB_CFG_LWID_ENB: /* set allowed Link-width */
+		ppd->link_width_enabled = val;
+		/* convert IB value to chip register value */
+		if (val == IB_WIDTH_1X)
+			val = 0;
+		else if (val == IB_WIDTH_4X)
+			val = 1;
+		else
+			val = 3;
+		maskr = SYM_RMASK(IBCCtrlB_0, IB_NUM_CHANNELS);
+		lsb = SYM_LSB(IBCCtrlB_0, IB_NUM_CHANNELS);
+		break;
+
+	case QIB_IB_CFG_SPD_ENB: /* set allowed Link speeds */
+		/*
+		 * As with width, only write the actual register if the
+		 * link is currently down, otherwise takes effect on next
+		 * link change.  Since setting is being explictly requested
+		 * (via MAD or sysfs), clear autoneg failure status if speed
+		 * autoneg is enabled.
+		 */
+		ppd->link_speed_enabled = val;
+		val <<= IBA7322_IBC_SPEED_LSB;
+		maskr = IBA7322_IBC_SPEED_MASK | IBA7322_IBC_IBTA_1_2_MASK |
+			IBA7322_IBC_MAX_SPEED_MASK;
+		if (val & (val - 1)) {
+			/* Muliple speeds enabled */
+			val |= IBA7322_IBC_IBTA_1_2_MASK |
+				IBA7322_IBC_MAX_SPEED_MASK;
+			spin_lock_irqsave(&ppd->lflags_lock, flags);
+			ppd->lflags &= ~QIBL_IB_AUTONEG_FAILED;
+			spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+		} else if (val & IBA7322_IBC_SPEED_QDR)
+			val |= IBA7322_IBC_IBTA_1_2_MASK;
+		/* IBTA 1.2 mode + min/max + speed bits are contiguous */
+		lsb = SYM_LSB(IBCCtrlB_0, IB_ENHANCED_MODE);
+		break;
+
+	case QIB_IB_CFG_RXPOL_ENB: /* set Auto-RX-polarity enable */
+		lsb = SYM_LSB(IBCCtrlB_0, IB_POLARITY_REV_SUPP);
+		maskr = SYM_RMASK(IBCCtrlB_0, IB_POLARITY_REV_SUPP);
+		break;
+
+	case QIB_IB_CFG_LREV_ENB: /* set Auto-Lane-reversal enable */
+		lsb = SYM_LSB(IBCCtrlB_0, IB_LANE_REV_SUPPORTED);
+		maskr = SYM_RMASK(IBCCtrlB_0, IB_LANE_REV_SUPPORTED);
+		break;
+
+	case QIB_IB_CFG_OVERRUN_THRESH: /* IB overrun threshold */
+		maskr = SYM_FIELD(ppd->cpspec->ibcctrl_a, IBCCtrlA_0,
+				  OverrunThreshold);
+		if (maskr != val) {
+			ppd->cpspec->ibcctrl_a &=
+				~SYM_MASK(IBCCtrlA_0, OverrunThreshold);
+			ppd->cpspec->ibcctrl_a |= (u64) val <<
+				SYM_LSB(IBCCtrlA_0, OverrunThreshold);
+			qib_write_kreg_port(ppd, krp_ibcctrl_a,
+					    ppd->cpspec->ibcctrl_a);
+			qib_write_kreg(dd, kr_scratch, 0ULL);
+		}
+		goto bail;
+
+	case QIB_IB_CFG_PHYERR_THRESH: /* IB PHY error threshold */
+		maskr = SYM_FIELD(ppd->cpspec->ibcctrl_a, IBCCtrlA_0,
+				  PhyerrThreshold);
+		if (maskr != val) {
+			ppd->cpspec->ibcctrl_a &=
+				~SYM_MASK(IBCCtrlA_0, PhyerrThreshold);
+			ppd->cpspec->ibcctrl_a |= (u64) val <<
+				SYM_LSB(IBCCtrlA_0, PhyerrThreshold);
+			qib_write_kreg_port(ppd, krp_ibcctrl_a,
+					    ppd->cpspec->ibcctrl_a);
+			qib_write_kreg(dd, kr_scratch, 0ULL);
+		}
+		goto bail;
+
+	case QIB_IB_CFG_PKEYS: /* update pkeys */
+		maskr = (u64) ppd->pkeys[0] | ((u64) ppd->pkeys[1] << 16) |
+			((u64) ppd->pkeys[2] << 32) |
+			((u64) ppd->pkeys[3] << 48);
+		qib_write_kreg_port(ppd, krp_partitionkey, maskr);
+		goto bail;
+
+	case QIB_IB_CFG_LINKDEFAULT: /* IB link default (sleep/poll) */
+		/* will only take effect when the link state changes */
+		if (val == IB_LINKINITCMD_POLL)
+			ppd->cpspec->ibcctrl_a &=
+				~SYM_MASK(IBCCtrlA_0, LinkDownDefaultState);
+		else /* SLEEP */
+			ppd->cpspec->ibcctrl_a |=
+				SYM_MASK(IBCCtrlA_0, LinkDownDefaultState);
+		qib_write_kreg_port(ppd, krp_ibcctrl_a, ppd->cpspec->ibcctrl_a);
+		qib_write_kreg(dd, kr_scratch, 0ULL);
+		goto bail;
+
+	case QIB_IB_CFG_MTU: /* update the MTU in IBC */
+		/*
+		 * Update our housekeeping variables, and set IBC max
+		 * size, same as init code; max IBC is max we allow in
+		 * buffer, less the qword pbc, plus 1 for ICRC, in dwords
+		 * Set even if it's unchanged, print debug message only
+		 * on changes.
+		 */
+		val = (ppd->ibmaxlen >> 2) + 1;
+		ppd->cpspec->ibcctrl_a &= ~SYM_MASK(IBCCtrlA_0, MaxPktLen);
+		ppd->cpspec->ibcctrl_a |= (u64)val <<
+			SYM_LSB(IBCCtrlA_0, MaxPktLen);
+		qib_write_kreg_port(ppd, krp_ibcctrl_a,
+				    ppd->cpspec->ibcctrl_a);
+		qib_write_kreg(dd, kr_scratch, 0ULL);
+		goto bail;
+
+	case QIB_IB_CFG_LSTATE: /* set the IB link state */
+		switch (val & 0xffff0000) {
+		case IB_LINKCMD_DOWN:
+			lcmd = QLOGIC_IB_IBCC_LINKCMD_DOWN;
+			ppd->cpspec->ibmalfusesnap = 1;
+			ppd->cpspec->ibmalfsnap = read_7322_creg32_port(ppd,
+				crp_errlink);
+			if (!ppd->cpspec->ibdeltainprog &&
+			    qib_compat_ddr_negotiate) {
+				ppd->cpspec->ibdeltainprog = 1;
+				ppd->cpspec->ibsymsnap =
+					read_7322_creg32_port(ppd,
+							      crp_ibsymbolerr);
+				ppd->cpspec->iblnkerrsnap =
+					read_7322_creg32_port(ppd,
+						      crp_iblinkerrrecov);
+			}
+			break;
+
+		case IB_LINKCMD_ARMED:
+			lcmd = QLOGIC_IB_IBCC_LINKCMD_ARMED;
+			if (ppd->cpspec->ibmalfusesnap) {
+				ppd->cpspec->ibmalfusesnap = 0;
+				ppd->cpspec->ibmalfdelta +=
+					read_7322_creg32_port(ppd,
+							      crp_errlink) -
+					ppd->cpspec->ibmalfsnap;
+			}
+			break;
+
+		case IB_LINKCMD_ACTIVE:
+			lcmd = QLOGIC_IB_IBCC_LINKCMD_ACTIVE;
+			break;
+
+		default:
+			ret = -EINVAL;
+			qib_dev_err(dd, "bad linkcmd req 0x%x\n", val >> 16);
+			goto bail;
+		}
+		switch (val & 0xffff) {
+		case IB_LINKINITCMD_NOP:
+			licmd = 0;
+			break;
+
+		case IB_LINKINITCMD_POLL:
+			licmd = QLOGIC_IB_IBCC_LINKINITCMD_POLL;
+			break;
+
+		case IB_LINKINITCMD_SLEEP:
+			licmd = QLOGIC_IB_IBCC_LINKINITCMD_SLEEP;
+			break;
+
+		case IB_LINKINITCMD_DISABLE:
+			licmd = QLOGIC_IB_IBCC_LINKINITCMD_DISABLE;
+			ppd->cpspec->chase_end = 0;
+			/*
+			 * stop state chase counter and timer, if running.
+			 * wait forpending timer, but don't clear .data (ppd)!
+			 */
+			if (ppd->cpspec->chase_timer.expires) {
+				del_timer_sync(&ppd->cpspec->chase_timer);
+				ppd->cpspec->chase_timer.expires = 0;
+			}
+			break;
+
+		default:
+			ret = -EINVAL;
+			qib_dev_err(dd, "bad linkinitcmd req 0x%x\n",
+				    val & 0xffff);
+			goto bail;
+		}
+		qib_set_ib_7322_lstate(ppd, lcmd, licmd);
+		goto bail;
+
+	case QIB_IB_CFG_OP_VLS:
+		if (ppd->vls_operational != val) {
+			ppd->vls_operational = val;
+			set_vls(ppd);
+		}
+		goto bail;
+
+	case QIB_IB_CFG_VL_HIGH_LIMIT:
+		qib_write_kreg_port(ppd, krp_highprio_limit, val);
+		goto bail;
+
+	case QIB_IB_CFG_HRTBT: /* set Heartbeat off/enable/auto */
+		if (val > 3) {
+			ret = -EINVAL;
+			goto bail;
+		}
+		lsb = IBA7322_IBC_HRTBT_LSB;
+		maskr = IBA7322_IBC_HRTBT_RMASK; /* OR of AUTO and ENB */
+		break;
+
+	case QIB_IB_CFG_PORT:
+		/* val is the port number of the switch we are connected to. */
+		if (ppd->dd->cspec->r1) {
+			cancel_delayed_work(&ppd->cpspec->ipg_work);
+			ppd->cpspec->ipg_tries = 0;
+		}
+		goto bail;
+
+	default:
+		ret = -EINVAL;
+		goto bail;
+	}
+	ppd->cpspec->ibcctrl_b &= ~(maskr << lsb);
+	ppd->cpspec->ibcctrl_b |= (((u64) val & maskr) << lsb);
+	qib_write_kreg_port(ppd, krp_ibcctrl_b, ppd->cpspec->ibcctrl_b);
+	qib_write_kreg(dd, kr_scratch, 0);
+bail:
+	return ret;
+}
+
+static int qib_7322_set_loopback(struct qib_pportdata *ppd, const char *what)
+{
+	int ret = 0;
+	u64 val, ctrlb;
+
+	/* only IBC loopback, may add serdes and xgxs loopbacks later */
+	if (!strncmp(what, "ibc", 3)) {
+		ppd->cpspec->ibcctrl_a |= SYM_MASK(IBCCtrlA_0,
+						       Loopback);
+		val = 0; /* disable heart beat, so link will come up */
+		qib_devinfo(ppd->dd->pcidev, "Enabling IB%u:%u IBC loopback\n",
+			 ppd->dd->unit, ppd->port);
+	} else if (!strncmp(what, "off", 3)) {
+		ppd->cpspec->ibcctrl_a &= ~SYM_MASK(IBCCtrlA_0,
+							Loopback);
+		/* enable heart beat again */
+		val = IBA7322_IBC_HRTBT_RMASK << IBA7322_IBC_HRTBT_LSB;
+		qib_devinfo(ppd->dd->pcidev, "Disabling IB%u:%u IBC loopback "
+			    "(normal)\n", ppd->dd->unit, ppd->port);
+	} else
+		ret = -EINVAL;
+	if (!ret) {
+		qib_write_kreg_port(ppd, krp_ibcctrl_a,
+				    ppd->cpspec->ibcctrl_a);
+		ctrlb = ppd->cpspec->ibcctrl_b & ~(IBA7322_IBC_HRTBT_MASK
+					     << IBA7322_IBC_HRTBT_LSB);
+		ppd->cpspec->ibcctrl_b = ctrlb | val;
+		qib_write_kreg_port(ppd, krp_ibcctrl_b,
+				    ppd->cpspec->ibcctrl_b);
+		qib_write_kreg(ppd->dd, kr_scratch, 0);
+	}
+	return ret;
+}
+
+static void get_vl_weights(struct qib_pportdata *ppd, unsigned regno,
+			   struct ib_vl_weight_elem *vl)
+{
+	unsigned i;
+
+	for (i = 0; i < 16; i++, regno++, vl++) {
+		u32 val = qib_read_kreg_port(ppd, regno);
+
+		vl->vl = (val >> SYM_LSB(LowPriority0_0, VirtualLane)) &
+			SYM_RMASK(LowPriority0_0, VirtualLane);
+		vl->weight = (val >> SYM_LSB(LowPriority0_0, Weight)) &
+			SYM_RMASK(LowPriority0_0, Weight);
+	}
+}
+
+static void set_vl_weights(struct qib_pportdata *ppd, unsigned regno,
+			   struct ib_vl_weight_elem *vl)
+{
+	unsigned i;
+
+	for (i = 0; i < 16; i++, regno++, vl++) {
+		u64 val;
+
+		val = ((vl->vl & SYM_RMASK(LowPriority0_0, VirtualLane)) <<
+			SYM_LSB(LowPriority0_0, VirtualLane)) |
+		      ((vl->weight & SYM_RMASK(LowPriority0_0, Weight)) <<
+			SYM_LSB(LowPriority0_0, Weight));
+		qib_write_kreg_port(ppd, regno, val);
+	}
+	if (!(ppd->p_sendctrl & SYM_MASK(SendCtrl_0, IBVLArbiterEn))) {
+		struct qib_devdata *dd = ppd->dd;
+		unsigned long flags;
+
+		spin_lock_irqsave(&dd->sendctrl_lock, flags);
+		ppd->p_sendctrl |= SYM_MASK(SendCtrl_0, IBVLArbiterEn);
+		qib_write_kreg_port(ppd, krp_sendctrl, ppd->p_sendctrl);
+		qib_write_kreg(dd, kr_scratch, 0);
+		spin_unlock_irqrestore(&dd->sendctrl_lock, flags);
+	}
+}
+
+static int qib_7322_get_ib_table(struct qib_pportdata *ppd, int which, void *t)
+{
+	switch (which) {
+	case QIB_IB_TBL_VL_HIGH_ARB:
+		get_vl_weights(ppd, krp_highprio_0, t);
+		break;
+
+	case QIB_IB_TBL_VL_LOW_ARB:
+		get_vl_weights(ppd, krp_lowprio_0, t);
+		break;
+
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static int qib_7322_set_ib_table(struct qib_pportdata *ppd, int which, void *t)
+{
+	switch (which) {
+	case QIB_IB_TBL_VL_HIGH_ARB:
+		set_vl_weights(ppd, krp_highprio_0, t);
+		break;
+
+	case QIB_IB_TBL_VL_LOW_ARB:
+		set_vl_weights(ppd, krp_lowprio_0, t);
+		break;
+
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static void qib_update_7322_usrhead(struct qib_ctxtdata *rcd, u64 hd,
+				    u32 updegr, u32 egrhd)
+{
+	qib_write_ureg(rcd->dd, ur_rcvhdrhead, hd, rcd->ctxt);
+	qib_write_ureg(rcd->dd, ur_rcvhdrhead, hd, rcd->ctxt);
+	if (updegr)
+		qib_write_ureg(rcd->dd, ur_rcvegrindexhead, egrhd, rcd->ctxt);
+}
+
+static u32 qib_7322_hdrqempty(struct qib_ctxtdata *rcd)
+{
+	u32 head, tail;
+
+	head = qib_read_ureg32(rcd->dd, ur_rcvhdrhead, rcd->ctxt);
+	if (rcd->rcvhdrtail_kvaddr)
+		tail = qib_get_rcvhdrtail(rcd);
+	else
+		tail = qib_read_ureg32(rcd->dd, ur_rcvhdrtail, rcd->ctxt);
+	return head == tail;
+}
+
+#define RCVCTRL_COMMON_MODS (QIB_RCVCTRL_CTXT_ENB | \
+	QIB_RCVCTRL_CTXT_DIS | \
+	QIB_RCVCTRL_TIDFLOW_ENB | \
+	QIB_RCVCTRL_TIDFLOW_DIS | \
+	QIB_RCVCTRL_TAILUPD_ENB | \
+	QIB_RCVCTRL_TAILUPD_DIS | \
+	QIB_RCVCTRL_INTRAVAIL_ENB | \
+	QIB_RCVCTRL_INTRAVAIL_DIS | \
+	QIB_RCVCTRL_BP_ENB | \
+	QIB_RCVCTRL_BP_DIS)
+
+#define RCVCTRL_PORT_MODS (QIB_RCVCTRL_CTXT_ENB | \
+	QIB_RCVCTRL_CTXT_DIS | \
+	QIB_RCVCTRL_PKEY_DIS | \
+	QIB_RCVCTRL_PKEY_ENB)
+
+/*
+ * Modify the RCVCTRL register in chip-specific way. This
+ * is a function because bit positions and (future) register
+ * location is chip-specifc, but the needed operations are
+ * generic. <op> is a bit-mask because we often want to
+ * do multiple modifications.
+ */
+static void rcvctrl_7322_mod(struct qib_pportdata *ppd, unsigned int op,
+			     int ctxt)
+{
+	struct qib_devdata *dd = ppd->dd;
+	struct qib_ctxtdata *rcd;
+	u64 mask, val;
+	unsigned long flags;
+
+	spin_lock_irqsave(&dd->cspec->rcvmod_lock, flags);
+
+	if (op & QIB_RCVCTRL_TIDFLOW_ENB)
+		dd->rcvctrl |= SYM_MASK(RcvCtrl, TidFlowEnable);
+	if (op & QIB_RCVCTRL_TIDFLOW_DIS)
+		dd->rcvctrl &= ~SYM_MASK(RcvCtrl, TidFlowEnable);
+	if (op & QIB_RCVCTRL_TAILUPD_ENB)
+		dd->rcvctrl |= SYM_MASK(RcvCtrl, TailUpd);
+	if (op & QIB_RCVCTRL_TAILUPD_DIS)
+		dd->rcvctrl &= ~SYM_MASK(RcvCtrl, TailUpd);
+	if (op & QIB_RCVCTRL_PKEY_ENB)
+		ppd->p_rcvctrl &= ~SYM_MASK(RcvCtrl_0, RcvPartitionKeyDisable);
+	if (op & QIB_RCVCTRL_PKEY_DIS)
+		ppd->p_rcvctrl |= SYM_MASK(RcvCtrl_0, RcvPartitionKeyDisable);
+	if (ctxt < 0) {
+		mask = (1ULL << dd->ctxtcnt) - 1;
+		rcd = NULL;
+	} else {
+		mask = (1ULL << ctxt);
+		rcd = dd->rcd[ctxt];
+	}
+	if ((op & QIB_RCVCTRL_CTXT_ENB) && rcd) {
+		ppd->p_rcvctrl |=
+			(mask << SYM_LSB(RcvCtrl_0, ContextEnableKernel));
+		if (!(dd->flags & QIB_NODMA_RTAIL)) {
+			op |= QIB_RCVCTRL_TAILUPD_ENB; /* need reg write */
+			dd->rcvctrl |= SYM_MASK(RcvCtrl, TailUpd);
+		}
+		/* Write these registers before the context is enabled. */
+		qib_write_kreg_ctxt(dd, krc_rcvhdrtailaddr, ctxt,
+				    rcd->rcvhdrqtailaddr_phys);
+		qib_write_kreg_ctxt(dd, krc_rcvhdraddr, ctxt,
+				    rcd->rcvhdrq_phys);
+		rcd->seq_cnt = 1;
+	}
+	if (op & QIB_RCVCTRL_CTXT_DIS)
+		ppd->p_rcvctrl &=
+			~(mask << SYM_LSB(RcvCtrl_0, ContextEnableKernel));
+	if (op & QIB_RCVCTRL_BP_ENB)
+		dd->rcvctrl |= mask << SYM_LSB(RcvCtrl, dontDropRHQFull);
+	if (op & QIB_RCVCTRL_BP_DIS)
+		dd->rcvctrl &= ~(mask << SYM_LSB(RcvCtrl, dontDropRHQFull));
+	if (op & QIB_RCVCTRL_INTRAVAIL_ENB)
+		dd->rcvctrl |= (mask << SYM_LSB(RcvCtrl, IntrAvail));
+	if (op & QIB_RCVCTRL_INTRAVAIL_DIS)
+		dd->rcvctrl &= ~(mask << SYM_LSB(RcvCtrl, IntrAvail));
+	/*
+	 * Decide which registers to write depending on the ops enabled.
+	 * Special case is "flush" (no bits set at all)
+	 * which needs to write both.
+	 */
+	if (op == 0 || (op & RCVCTRL_COMMON_MODS))
+		qib_write_kreg(dd, kr_rcvctrl, dd->rcvctrl);
+	if (op == 0 || (op & RCVCTRL_PORT_MODS))
+		qib_write_kreg_port(ppd, krp_rcvctrl, ppd->p_rcvctrl);
+	if ((op & QIB_RCVCTRL_CTXT_ENB) && dd->rcd[ctxt]) {
+		/*
+		 * Init the context registers also; if we were
+		 * disabled, tail and head should both be zero
+		 * already from the enable, but since we don't
+		 * know, we have to do it explictly.
+		 */
+		val = qib_read_ureg32(dd, ur_rcvegrindextail, ctxt);
+		qib_write_ureg(dd, ur_rcvegrindexhead, val, ctxt);
+
+		/* be sure enabling write seen; hd/tl should be 0 */
+		(void) qib_read_kreg32(dd, kr_scratch);
+		val = qib_read_ureg32(dd, ur_rcvhdrtail, ctxt);
+		dd->rcd[ctxt]->head = val;
+		/* If kctxt, interrupt on next receive. */
+		if (ctxt < dd->first_user_ctxt)
+			val |= dd->rhdrhead_intr_off;
+		qib_write_ureg(dd, ur_rcvhdrhead, val, ctxt);
+	} else if ((op & QIB_RCVCTRL_INTRAVAIL_ENB) &&
+		dd->rcd[ctxt] && dd->rhdrhead_intr_off) {
+		/* arm rcv interrupt */
+		val = dd->rcd[ctxt]->head | dd->rhdrhead_intr_off;
+		qib_write_ureg(dd, ur_rcvhdrhead, val, ctxt);
+	}
+	if (op & QIB_RCVCTRL_CTXT_DIS) {
+		unsigned f;
+
+		/* Now that the context is disabled, clear these registers. */
+		if (ctxt >= 0) {
+			qib_write_kreg_ctxt(dd, krc_rcvhdrtailaddr, ctxt, 0);
+			qib_write_kreg_ctxt(dd, krc_rcvhdraddr, ctxt, 0);
+			for (f = 0; f < NUM_TIDFLOWS_CTXT; f++)
+				qib_write_ureg(dd, ur_rcvflowtable + f,
+					       TIDFLOW_ERRBITS, ctxt);
+		} else {
+			unsigned i;
+
+			for (i = 0; i < dd->cfgctxts; i++) {
+				qib_write_kreg_ctxt(dd, krc_rcvhdrtailaddr,
+						    i, 0);
+				qib_write_kreg_ctxt(dd, krc_rcvhdraddr, i, 0);
+				for (f = 0; f < NUM_TIDFLOWS_CTXT; f++)
+					qib_write_ureg(dd, ur_rcvflowtable + f,
+						       TIDFLOW_ERRBITS, i);
+			}
+		}
+	}
+	spin_unlock_irqrestore(&dd->cspec->rcvmod_lock, flags);
+}
+
+/*
+ * Modify the SENDCTRL register in chip-specific way. This
+ * is a function where there are multiple such registers with
+ * slightly different layouts.
+ * The chip doesn't allow back-to-back sendctrl writes, so write
+ * the scratch register after writing sendctrl.
+ *
+ * Which register is written depends on the operation.
+ * Most operate on the common register, while
+ * SEND_ENB and SEND_DIS operate on the per-port ones.
+ * SEND_ENB is included in common because it can change SPCL_TRIG
+ */
+#define SENDCTRL_COMMON_MODS (\
+	QIB_SENDCTRL_CLEAR | \
+	QIB_SENDCTRL_AVAIL_DIS | \
+	QIB_SENDCTRL_AVAIL_ENB | \
+	QIB_SENDCTRL_AVAIL_BLIP | \
+	QIB_SENDCTRL_DISARM | \
+	QIB_SENDCTRL_DISARM_ALL | \
+	QIB_SENDCTRL_SEND_ENB)
+
+#define SENDCTRL_PORT_MODS (\
+	QIB_SENDCTRL_CLEAR | \
+	QIB_SENDCTRL_SEND_ENB | \
+	QIB_SENDCTRL_SEND_DIS | \
+	QIB_SENDCTRL_FLUSH)
+
+static void sendctrl_7322_mod(struct qib_pportdata *ppd, u32 op)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u64 tmp_dd_sendctrl;
+	unsigned long flags;
+
+	spin_lock_irqsave(&dd->sendctrl_lock, flags);
+
+	/* First the dd ones that are "sticky", saved in shadow */
+	if (op & QIB_SENDCTRL_CLEAR)
+		dd->sendctrl = 0;
+	if (op & QIB_SENDCTRL_AVAIL_DIS)
+		dd->sendctrl &= ~SYM_MASK(SendCtrl, SendBufAvailUpd);
+	else if (op & QIB_SENDCTRL_AVAIL_ENB) {
+		dd->sendctrl |= SYM_MASK(SendCtrl, SendBufAvailUpd);
+		if (dd->flags & QIB_USE_SPCL_TRIG)
+			dd->sendctrl |= SYM_MASK(SendCtrl, SpecialTriggerEn);
+	}
+
+	/* Then the ppd ones that are "sticky", saved in shadow */
+	if (op & QIB_SENDCTRL_SEND_DIS)
+		ppd->p_sendctrl &= ~SYM_MASK(SendCtrl_0, SendEnable);
+	else if (op & QIB_SENDCTRL_SEND_ENB)
+		ppd->p_sendctrl |= SYM_MASK(SendCtrl_0, SendEnable);
+
+	if (op & QIB_SENDCTRL_DISARM_ALL) {
+		u32 i, last;
+
+		tmp_dd_sendctrl = dd->sendctrl;
+		last = dd->piobcnt2k + dd->piobcnt4k + NUM_VL15_BUFS;
+		/*
+		 * Disarm any buffers that are not yet launched,
+		 * disabling updates until done.
+		 */
+		tmp_dd_sendctrl &= ~SYM_MASK(SendCtrl, SendBufAvailUpd);
+		for (i = 0; i < last; i++) {
+			qib_write_kreg(dd, kr_sendctrl,
+				       tmp_dd_sendctrl |
+				       SYM_MASK(SendCtrl, Disarm) | i);
+			qib_write_kreg(dd, kr_scratch, 0);
+		}
+	}
+
+	if (op & QIB_SENDCTRL_FLUSH) {
+		u64 tmp_ppd_sendctrl = ppd->p_sendctrl;
+
+		/*
+		 * Now drain all the fifos.  The Abort bit should never be
+		 * needed, so for now, at least, we don't use it.
+		 */
+		tmp_ppd_sendctrl |=
+			SYM_MASK(SendCtrl_0, TxeDrainRmFifo) |
+			SYM_MASK(SendCtrl_0, TxeDrainLaFifo) |
+			SYM_MASK(SendCtrl_0, TxeBypassIbc);
+		qib_write_kreg_port(ppd, krp_sendctrl, tmp_ppd_sendctrl);
+		qib_write_kreg(dd, kr_scratch, 0);
+	}
+
+	tmp_dd_sendctrl = dd->sendctrl;
+
+	if (op & QIB_SENDCTRL_DISARM)
+		tmp_dd_sendctrl |= SYM_MASK(SendCtrl, Disarm) |
+			((op & QIB_7322_SendCtrl_DisarmSendBuf_RMASK) <<
+			 SYM_LSB(SendCtrl, DisarmSendBuf));
+	if ((op & QIB_SENDCTRL_AVAIL_BLIP) &&
+	    (dd->sendctrl & SYM_MASK(SendCtrl, SendBufAvailUpd)))
+		tmp_dd_sendctrl &= ~SYM_MASK(SendCtrl, SendBufAvailUpd);
+
+	if (op == 0 || (op & SENDCTRL_COMMON_MODS)) {
+		qib_write_kreg(dd, kr_sendctrl, tmp_dd_sendctrl);
+		qib_write_kreg(dd, kr_scratch, 0);
+	}
+
+	if (op == 0 || (op & SENDCTRL_PORT_MODS)) {
+		qib_write_kreg_port(ppd, krp_sendctrl, ppd->p_sendctrl);
+		qib_write_kreg(dd, kr_scratch, 0);
+	}
+
+	if (op & QIB_SENDCTRL_AVAIL_BLIP) {
+		qib_write_kreg(dd, kr_sendctrl, dd->sendctrl);
+		qib_write_kreg(dd, kr_scratch, 0);
+	}
+
+	spin_unlock_irqrestore(&dd->sendctrl_lock, flags);
+
+	if (op & QIB_SENDCTRL_FLUSH) {
+		u32 v;
+		/*
+		 * ensure writes have hit chip, then do a few
+		 * more reads, to allow DMA of pioavail registers
+		 * to occur, so in-memory copy is in sync with
+		 * the chip.  Not always safe to sleep.
+		 */
+		v = qib_read_kreg32(dd, kr_scratch);
+		qib_write_kreg(dd, kr_scratch, v);
+		v = qib_read_kreg32(dd, kr_scratch);
+		qib_write_kreg(dd, kr_scratch, v);
+		qib_read_kreg32(dd, kr_scratch);
+	}
+}
+
+#define _PORT_VIRT_FLAG 0x8000U /* "virtual", need adjustments */
+#define _PORT_64BIT_FLAG 0x10000U /* not "virtual", but 64bit */
+#define _PORT_CNTR_IDXMASK 0x7fffU /* mask off flags above */
+
+/**
+ * qib_portcntr_7322 - read a per-port chip counter
+ * @ppd: the qlogic_ib pport
+ * @creg: the counter to read (not a chip offset)
+ */
+static u64 qib_portcntr_7322(struct qib_pportdata *ppd, u32 reg)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u64 ret = 0ULL;
+	u16 creg;
+	/* 0xffff for unimplemented or synthesized counters */
+	static const u32 xlator[] = {
+		[QIBPORTCNTR_PKTSEND] = crp_pktsend | _PORT_64BIT_FLAG,
+		[QIBPORTCNTR_WORDSEND] = crp_wordsend | _PORT_64BIT_FLAG,
+		[QIBPORTCNTR_PSXMITDATA] = crp_psxmitdatacount,
+		[QIBPORTCNTR_PSXMITPKTS] = crp_psxmitpktscount,
+		[QIBPORTCNTR_PSXMITWAIT] = crp_psxmitwaitcount,
+		[QIBPORTCNTR_SENDSTALL] = crp_sendstall,
+		[QIBPORTCNTR_PKTRCV] = crp_pktrcv | _PORT_64BIT_FLAG,
+		[QIBPORTCNTR_PSRCVDATA] = crp_psrcvdatacount,
+		[QIBPORTCNTR_PSRCVPKTS] = crp_psrcvpktscount,
+		[QIBPORTCNTR_RCVEBP] = crp_rcvebp,
+		[QIBPORTCNTR_RCVOVFL] = crp_rcvovfl,
+		[QIBPORTCNTR_WORDRCV] = crp_wordrcv | _PORT_64BIT_FLAG,
+		[QIBPORTCNTR_RXDROPPKT] = 0xffff, /* not needed  for 7322 */
+		[QIBPORTCNTR_RXLOCALPHYERR] = crp_rxotherlocalphyerr,
+		[QIBPORTCNTR_RXVLERR] = crp_rxvlerr,
+		[QIBPORTCNTR_ERRICRC] = crp_erricrc,
+		[QIBPORTCNTR_ERRVCRC] = crp_errvcrc,
+		[QIBPORTCNTR_ERRLPCRC] = crp_errlpcrc,
+		[QIBPORTCNTR_BADFORMAT] = crp_badformat,
+		[QIBPORTCNTR_ERR_RLEN] = crp_err_rlen,
+		[QIBPORTCNTR_IBSYMBOLERR] = crp_ibsymbolerr,
+		[QIBPORTCNTR_INVALIDRLEN] = crp_invalidrlen,
+		[QIBPORTCNTR_UNSUPVL] = crp_txunsupvl,
+		[QIBPORTCNTR_EXCESSBUFOVFL] = crp_excessbufferovfl,
+		[QIBPORTCNTR_ERRLINK] = crp_errlink,
+		[QIBPORTCNTR_IBLINKDOWN] = crp_iblinkdown,
+		[QIBPORTCNTR_IBLINKERRRECOV] = crp_iblinkerrrecov,
+		[QIBPORTCNTR_LLI] = crp_locallinkintegrityerr,
+		[QIBPORTCNTR_VL15PKTDROP] = crp_vl15droppedpkt,
+		[QIBPORTCNTR_ERRPKEY] = crp_errpkey,
+		/*
+		 * the next 3 aren't really counters, but were implemented
+		 * as counters in older chips, so still get accessed as
+		 * though they were counters from this code.
+		 */
+		[QIBPORTCNTR_PSINTERVAL] = krp_psinterval,
+		[QIBPORTCNTR_PSSTART] = krp_psstart,
+		[QIBPORTCNTR_PSSTAT] = krp_psstat,
+		/* pseudo-counter, summed for all ports */
+		[QIBPORTCNTR_KHDROVFL] = 0xffff,
+	};
+
+	if (reg >= ARRAY_SIZE(xlator)) {
+		qib_devinfo(ppd->dd->pcidev,
+			 "Unimplemented portcounter %u\n", reg);
+		goto done;
+	}
+	creg = xlator[reg] & _PORT_CNTR_IDXMASK;
+
+	/* handle non-counters and special cases first */
+	if (reg == QIBPORTCNTR_KHDROVFL) {
+		int i;
+
+		/* sum over all kernel contexts (skip if mini_init) */
+		for (i = 0; dd->rcd && i < dd->first_user_ctxt; i++) {
+			struct qib_ctxtdata *rcd = dd->rcd[i];
+
+			if (!rcd || rcd->ppd != ppd)
+				continue;
+			ret += read_7322_creg32(dd, cr_base_egrovfl + i);
+		}
+		goto done;
+	} else if (reg == QIBPORTCNTR_RXDROPPKT) {
+		/*
+		 * Used as part of the synthesis of port_rcv_errors
+		 * in the verbs code for IBTA counters.  Not needed for 7322,
+		 * because all the errors are already counted by other cntrs.
+		 */
+		goto done;
+	} else if (reg == QIBPORTCNTR_PSINTERVAL ||
+		   reg == QIBPORTCNTR_PSSTART || reg == QIBPORTCNTR_PSSTAT) {
+		/* were counters in older chips, now per-port kernel regs */
+		ret = qib_read_kreg_port(ppd, creg);
+		goto done;
+	}
+
+	/*
+	 * Only fast increment counters are 64 bits; use 32 bit reads to
+	 * avoid two independent reads when on Opteron.
+	 */
+	if (xlator[reg] & _PORT_64BIT_FLAG)
+		ret = read_7322_creg_port(ppd, creg);
+	else
+		ret = read_7322_creg32_port(ppd, creg);
+	if (creg == crp_ibsymbolerr) {
+		if (ppd->cpspec->ibdeltainprog)
+			ret -= ret - ppd->cpspec->ibsymsnap;
+		ret -= ppd->cpspec->ibsymdelta;
+	} else if (creg == crp_iblinkerrrecov) {
+		if (ppd->cpspec->ibdeltainprog)
+			ret -= ret - ppd->cpspec->iblnkerrsnap;
+		ret -= ppd->cpspec->iblnkerrdelta;
+	} else if (creg == crp_errlink)
+		ret -= ppd->cpspec->ibmalfdelta;
+	else if (creg == crp_iblinkdown)
+		ret += ppd->cpspec->iblnkdowndelta;
+done:
+	return ret;
+}
+
+/*
+ * Device counter names (not port-specific), one line per stat,
+ * single string.  Used by utilities like ipathstats to print the stats
+ * in a way which works for different versions of drivers, without changing
+ * the utility.  Names need to be 12 chars or less (w/o newline), for proper
+ * display by utility.
+ * Non-error counters are first.
+ * Start of "error" conters is indicated by a leading "E " on the first
+ * "error" counter, and doesn't count in label length.
+ * The EgrOvfl list needs to be last so we truncate them at the configured
+ * context count for the device.
+ * cntr7322indices contains the corresponding register indices.
+ */
+static const char cntr7322names[] =
+	"Interrupts\n"
+	"HostBusStall\n"
+	"E RxTIDFull\n"
+	"RxTIDInvalid\n"
+	"RxTIDFloDrop\n" /* 7322 only */
+	"Ctxt0EgrOvfl\n"
+	"Ctxt1EgrOvfl\n"
+	"Ctxt2EgrOvfl\n"
+	"Ctxt3EgrOvfl\n"
+	"Ctxt4EgrOvfl\n"
+	"Ctxt5EgrOvfl\n"
+	"Ctxt6EgrOvfl\n"
+	"Ctxt7EgrOvfl\n"
+	"Ctxt8EgrOvfl\n"
+	"Ctxt9EgrOvfl\n"
+	"Ctx10EgrOvfl\n"
+	"Ctx11EgrOvfl\n"
+	"Ctx12EgrOvfl\n"
+	"Ctx13EgrOvfl\n"
+	"Ctx14EgrOvfl\n"
+	"Ctx15EgrOvfl\n"
+	"Ctx16EgrOvfl\n"
+	"Ctx17EgrOvfl\n"
+	;
+
+static const u32 cntr7322indices[] = {
+	cr_lbint | _PORT_64BIT_FLAG,
+	cr_lbstall | _PORT_64BIT_FLAG,
+	cr_tidfull,
+	cr_tidinvalid,
+	cr_rxtidflowdrop,
+	cr_base_egrovfl + 0,
+	cr_base_egrovfl + 1,
+	cr_base_egrovfl + 2,
+	cr_base_egrovfl + 3,
+	cr_base_egrovfl + 4,
+	cr_base_egrovfl + 5,
+	cr_base_egrovfl + 6,
+	cr_base_egrovfl + 7,
+	cr_base_egrovfl + 8,
+	cr_base_egrovfl + 9,
+	cr_base_egrovfl + 10,
+	cr_base_egrovfl + 11,
+	cr_base_egrovfl + 12,
+	cr_base_egrovfl + 13,
+	cr_base_egrovfl + 14,
+	cr_base_egrovfl + 15,
+	cr_base_egrovfl + 16,
+	cr_base_egrovfl + 17,
+};
+
+/*
+ * same as cntr7322names and cntr7322indices, but for port-specific counters.
+ * portcntr7322indices is somewhat complicated by some registers needing
+ * adjustments of various kinds, and those are ORed with _PORT_VIRT_FLAG
+ */
+static const char portcntr7322names[] =
+	"TxPkt\n"
+	"TxFlowPkt\n"
+	"TxWords\n"
+	"RxPkt\n"
+	"RxFlowPkt\n"
+	"RxWords\n"
+	"TxFlowStall\n"
+	"TxDmaDesc\n"  /* 7220 and 7322-only */
+	"E RxDlidFltr\n"  /* 7220 and 7322-only */
+	"IBStatusChng\n"
+	"IBLinkDown\n"
+	"IBLnkRecov\n"
+	"IBRxLinkErr\n"
+	"IBSymbolErr\n"
+	"RxLLIErr\n"
+	"RxBadFormat\n"
+	"RxBadLen\n"
+	"RxBufOvrfl\n"
+	"RxEBP\n"
+	"RxFlowCtlErr\n"
+	"RxICRCerr\n"
+	"RxLPCRCerr\n"
+	"RxVCRCerr\n"
+	"RxInvalLen\n"
+	"RxInvalPKey\n"
+	"RxPktDropped\n"
+	"TxBadLength\n"
+	"TxDropped\n"
+	"TxInvalLen\n"
+	"TxUnderrun\n"
+	"TxUnsupVL\n"
+	"RxLclPhyErr\n" /* 7220 and 7322-only from here down */
+	"RxVL15Drop\n"
+	"RxVlErr\n"
+	"XcessBufOvfl\n"
+	"RxQPBadCtxt\n" /* 7322-only from here down */
+	"TXBadHeader\n"
+	;
+
+static const u32 portcntr7322indices[] = {
+	QIBPORTCNTR_PKTSEND | _PORT_VIRT_FLAG,
+	crp_pktsendflow,
+	QIBPORTCNTR_WORDSEND | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_PKTRCV | _PORT_VIRT_FLAG,
+	crp_pktrcvflowctrl,
+	QIBPORTCNTR_WORDRCV | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_SENDSTALL | _PORT_VIRT_FLAG,
+	crp_txsdmadesc | _PORT_64BIT_FLAG,
+	crp_rxdlidfltr,
+	crp_ibstatuschange,
+	QIBPORTCNTR_IBLINKDOWN | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_IBLINKERRRECOV | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_ERRLINK | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_IBSYMBOLERR | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_LLI | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_BADFORMAT | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_ERR_RLEN | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_RCVOVFL | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_RCVEBP | _PORT_VIRT_FLAG,
+	crp_rcvflowctrlviol,
+	QIBPORTCNTR_ERRICRC | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_ERRLPCRC | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_ERRVCRC | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_INVALIDRLEN | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_ERRPKEY | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_RXDROPPKT | _PORT_VIRT_FLAG,
+	crp_txminmaxlenerr,
+	crp_txdroppedpkt,
+	crp_txlenerr,
+	crp_txunderrun,
+	crp_txunsupvl,
+	QIBPORTCNTR_RXLOCALPHYERR | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_VL15PKTDROP | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_RXVLERR | _PORT_VIRT_FLAG,
+	QIBPORTCNTR_EXCESSBUFOVFL | _PORT_VIRT_FLAG,
+	crp_rxqpinvalidctxt,
+	crp_txhdrerr,
+};
+
+/* do all the setup to make the counter reads efficient later */
+static void init_7322_cntrnames(struct qib_devdata *dd)
+{
+	int i, j = 0;
+	char *s;
+
+	for (i = 0, s = (char *)cntr7322names; s && j <= dd->cfgctxts;
+	     i++) {
+		/* we always have at least one counter before the egrovfl */
+		if (!j && !strncmp("Ctxt0EgrOvfl", s + 1, 12))
+			j = 1;
+		s = strchr(s + 1, '\n');
+		if (s && j)
+			j++;
+	}
+	dd->cspec->ncntrs = i;
+	if (!s)
+		/* full list; size is without terminating null */
+		dd->cspec->cntrnamelen = sizeof(cntr7322names) - 1;
+	else
+		dd->cspec->cntrnamelen = 1 + s - cntr7322names;
+	dd->cspec->cntrs = kmalloc(dd->cspec->ncntrs
+		* sizeof(u64), GFP_KERNEL);
+	if (!dd->cspec->cntrs)
+		qib_dev_err(dd, "Failed allocation for counters\n");
+
+	for (i = 0, s = (char *)portcntr7322names; s; i++)
+		s = strchr(s + 1, '\n');
+	dd->cspec->nportcntrs = i - 1;
+	dd->cspec->portcntrnamelen = sizeof(portcntr7322names) - 1;
+	for (i = 0; i < dd->num_pports; ++i) {
+		dd->pport[i].cpspec->portcntrs = kmalloc(dd->cspec->nportcntrs
+			* sizeof(u64), GFP_KERNEL);
+		if (!dd->pport[i].cpspec->portcntrs)
+			qib_dev_err(dd, "Failed allocation for"
+				    " portcounters\n");
+	}
+}
+
+static u32 qib_read_7322cntrs(struct qib_devdata *dd, loff_t pos, char **namep,
+			      u64 **cntrp)
+{
+	u32 ret;
+
+	if (namep) {
+		ret = dd->cspec->cntrnamelen;
+		if (pos >= ret)
+			ret = 0; /* final read after getting everything */
+		else
+			*namep = (char *) cntr7322names;
+	} else {
+		u64 *cntr = dd->cspec->cntrs;
+		int i;
+
+		ret = dd->cspec->ncntrs * sizeof(u64);
+		if (!cntr || pos >= ret) {
+			/* everything read, or couldn't get memory */
+			ret = 0;
+			goto done;
+		}
+		*cntrp = cntr;
+		for (i = 0; i < dd->cspec->ncntrs; i++)
+			if (cntr7322indices[i] & _PORT_64BIT_FLAG)
+				*cntr++ = read_7322_creg(dd,
+							 cntr7322indices[i] &
+							 _PORT_CNTR_IDXMASK);
+			else
+				*cntr++ = read_7322_creg32(dd,
+							   cntr7322indices[i]);
+	}
+done:
+	return ret;
+}
+
+static u32 qib_read_7322portcntrs(struct qib_devdata *dd, loff_t pos, u32 port,
+				  char **namep, u64 **cntrp)
+{
+	u32 ret;
+
+	if (namep) {
+		ret = dd->cspec->portcntrnamelen;
+		if (pos >= ret)
+			ret = 0; /* final read after getting everything */
+		else
+			*namep = (char *)portcntr7322names;
+	} else {
+		struct qib_pportdata *ppd = &dd->pport[port];
+		u64 *cntr = ppd->cpspec->portcntrs;
+		int i;
+
+		ret = dd->cspec->nportcntrs * sizeof(u64);
+		if (!cntr || pos >= ret) {
+			/* everything read, or couldn't get memory */
+			ret = 0;
+			goto done;
+		}
+		*cntrp = cntr;
+		for (i = 0; i < dd->cspec->nportcntrs; i++) {
+			if (portcntr7322indices[i] & _PORT_VIRT_FLAG)
+				*cntr++ = qib_portcntr_7322(ppd,
+					portcntr7322indices[i] &
+					_PORT_CNTR_IDXMASK);
+			else if (portcntr7322indices[i] & _PORT_64BIT_FLAG)
+				*cntr++ = read_7322_creg_port(ppd,
+					   portcntr7322indices[i] &
+					    _PORT_CNTR_IDXMASK);
+			else
+				*cntr++ = read_7322_creg32_port(ppd,
+					   portcntr7322indices[i]);
+		}
+	}
+done:
+	return ret;
+}
+
+/**
+ * qib_get_7322_faststats - get word counters from chip before they overflow
+ * @opaque - contains a pointer to the qlogic_ib device qib_devdata
+ *
+ * VESTIGIAL IBA7322 has no "small fast counters", so the only
+ * real purpose of this function is to maintain the notion of
+ * "active time", which in turn is only logged into the eeprom,
+ * which we don;t have, yet, for 7322-based boards.
+ *
+ * called from add_timer
+ */
+static void qib_get_7322_faststats(unsigned long opaque)
+{
+	struct qib_devdata *dd = (struct qib_devdata *) opaque;
+	struct qib_pportdata *ppd;
+	unsigned long flags;
+	u64 traffic_wds;
+	int pidx;
+
+	for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+		ppd = dd->pport + pidx;
+
+		/*
+		 * If port isn't enabled or not operational ports, or
+		 * diags is running (can cause memory diags to fail)
+		 * skip this port this time.
+		 */
+		if (!ppd->link_speed_supported || !(dd->flags & QIB_INITTED)
+		    || dd->diag_client)
+			continue;
+
+		/*
+		 * Maintain an activity timer, based on traffic
+		 * exceeding a threshold, so we need to check the word-counts
+		 * even if they are 64-bit.
+		 */
+		traffic_wds = qib_portcntr_7322(ppd, QIBPORTCNTR_WORDRCV) +
+			qib_portcntr_7322(ppd, QIBPORTCNTR_WORDSEND);
+		spin_lock_irqsave(&ppd->dd->eep_st_lock, flags);
+		traffic_wds -= ppd->dd->traffic_wds;
+		ppd->dd->traffic_wds += traffic_wds;
+		if (traffic_wds >= QIB_TRAFFIC_ACTIVE_THRESHOLD)
+			atomic_add(ACTIVITY_TIMER, &ppd->dd->active_time);
+		spin_unlock_irqrestore(&ppd->dd->eep_st_lock, flags);
+		if (ppd->cpspec->qdr_dfe_on && (ppd->link_speed_active &
+						QIB_IB_QDR) &&
+		    (ppd->lflags & (QIBL_LINKINIT | QIBL_LINKARMED |
+				    QIBL_LINKACTIVE)) &&
+		    ppd->cpspec->qdr_dfe_time &&
+		    time_after64(get_jiffies_64(), ppd->cpspec->qdr_dfe_time)) {
+			ppd->cpspec->qdr_dfe_on = 0;
+
+			qib_write_kreg_port(ppd, krp_static_adapt_dis(2),
+					    ppd->dd->cspec->r1 ?
+					    QDR_STATIC_ADAPT_INIT_R1 :
+					    QDR_STATIC_ADAPT_INIT);
+			force_h1(ppd);
+		}
+	}
+	mod_timer(&dd->stats_timer, jiffies + HZ * ACTIVITY_TIMER);
+}
+
+/*
+ * If we were using MSIx, try to fallback to INTx.
+ */
+static int qib_7322_intr_fallback(struct qib_devdata *dd)
+{
+	if (!dd->cspec->num_msix_entries)
+		return 0; /* already using INTx */
+
+	qib_devinfo(dd->pcidev, "MSIx interrupt not detected,"
+		 " trying INTx interrupts\n");
+	qib_7322_nomsix(dd);
+	qib_enable_intx(dd->pcidev);
+	qib_setup_7322_interrupt(dd, 0);
+	return 1;
+}
+
+/*
+ * Reset the XGXS (between serdes and IBC).  Slightly less intrusive
+ * than resetting the IBC or external link state, and useful in some
+ * cases to cause some retraining.  To do this right, we reset IBC
+ * as well, then return to previous state (which may be still in reset)
+ * NOTE: some callers of this "know" this writes the current value
+ * of cpspec->ibcctrl_a as part of it's operation, so if that changes,
+ * check all callers.
+ */
+static void qib_7322_mini_pcs_reset(struct qib_pportdata *ppd)
+{
+	u64 val;
+	struct qib_devdata *dd = ppd->dd;
+	const u64 reset_bits = SYM_MASK(IBPCSConfig_0, xcv_rreset) |
+		SYM_MASK(IBPCSConfig_0, xcv_treset) |
+		SYM_MASK(IBPCSConfig_0, tx_rx_reset);
+
+	val = qib_read_kreg_port(ppd, krp_ib_pcsconfig);
+	qib_write_kreg_port(ppd, krp_ibcctrl_a,
+			    ppd->cpspec->ibcctrl_a &
+			    ~SYM_MASK(IBCCtrlA_0, IBLinkEn));
+
+	qib_write_kreg_port(ppd, krp_ib_pcsconfig, val | reset_bits);
+	qib_read_kreg32(dd, kr_scratch);
+	qib_write_kreg_port(ppd, krp_ib_pcsconfig, val & ~reset_bits);
+	qib_write_kreg_port(ppd, krp_ibcctrl_a, ppd->cpspec->ibcctrl_a);
+	qib_write_kreg(dd, kr_scratch, 0ULL);
+}
+
+/*
+ * This code for non-IBTA-compliant IB speed negotiation is only known to
+ * work for the SDR to DDR transition, and only between an HCA and a switch
+ * with recent firmware.  It is based on observed heuristics, rather than
+ * actual knowledge of the non-compliant speed negotiation.
+ * It has a number of hard-coded fields, since the hope is to rewrite this
+ * when a spec is available on how the negoation is intended to work.
+ */
+static void autoneg_7322_sendpkt(struct qib_pportdata *ppd, u32 *hdr,
+				 u32 dcnt, u32 *data)
+{
+	int i;
+	u64 pbc;
+	u32 __iomem *piobuf;
+	u32 pnum, control, len;
+	struct qib_devdata *dd = ppd->dd;
+
+	i = 0;
+	len = 7 + dcnt + 1; /* 7 dword header, dword data, icrc */
+	control = qib_7322_setpbc_control(ppd, len, 0, 15);
+	pbc = ((u64) control << 32) | len;
+	while (!(piobuf = qib_7322_getsendbuf(ppd, pbc, &pnum))) {
+		if (i++ > 15)
+			return;
+		udelay(2);
+	}
+	/* disable header check on this packet, since it can't be valid */
+	dd->f_txchk_change(dd, pnum, 1, TXCHK_CHG_TYPE_DIS1, NULL);
+	writeq(pbc, piobuf);
+	qib_flush_wc();
+	qib_pio_copy(piobuf + 2, hdr, 7);
+	qib_pio_copy(piobuf + 9, data, dcnt);
+	if (dd->flags & QIB_USE_SPCL_TRIG) {
+		u32 spcl_off = (pnum >= dd->piobcnt2k) ? 2047 : 1023;
+
+		qib_flush_wc();
+		__raw_writel(0xaebecede, piobuf + spcl_off);
+	}
+	qib_flush_wc();
+	qib_sendbuf_done(dd, pnum);
+	/* and re-enable hdr check */
+	dd->f_txchk_change(dd, pnum, 1, TXCHK_CHG_TYPE_ENAB1, NULL);
+}
+
+/*
+ * _start packet gets sent twice at start, _done gets sent twice at end
+ */
+static void qib_autoneg_7322_send(struct qib_pportdata *ppd, int which)
+{
+	struct qib_devdata *dd = ppd->dd;
+	static u32 swapped;
+	u32 dw, i, hcnt, dcnt, *data;
+	static u32 hdr[7] = { 0xf002ffff, 0x48ffff, 0x6400abba };
+	static u32 madpayload_start[0x40] = {
+		0x1810103, 0x1, 0x0, 0x0, 0x2c90000, 0x2c9, 0x0, 0x0,
+		0xffffffff, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0,
+		0x1, 0x1388, 0x15e, 0x1, /* rest 0's */
+		};
+	static u32 madpayload_done[0x40] = {
+		0x1810103, 0x1, 0x0, 0x0, 0x2c90000, 0x2c9, 0x0, 0x0,
+		0xffffffff, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0,
+		0x40000001, 0x1388, 0x15e, /* rest 0's */
+		};
+
+	dcnt = ARRAY_SIZE(madpayload_start);
+	hcnt = ARRAY_SIZE(hdr);
+	if (!swapped) {
+		/* for maintainability, do it at runtime */
+		for (i = 0; i < hcnt; i++) {
+			dw = (__force u32) cpu_to_be32(hdr[i]);
+			hdr[i] = dw;
+		}
+		for (i = 0; i < dcnt; i++) {
+			dw = (__force u32) cpu_to_be32(madpayload_start[i]);
+			madpayload_start[i] = dw;
+			dw = (__force u32) cpu_to_be32(madpayload_done[i]);
+			madpayload_done[i] = dw;
+		}
+		swapped = 1;
+	}
+
+	data = which ? madpayload_done : madpayload_start;
+
+	autoneg_7322_sendpkt(ppd, hdr, dcnt, data);
+	qib_read_kreg64(dd, kr_scratch);
+	udelay(2);
+	autoneg_7322_sendpkt(ppd, hdr, dcnt, data);
+	qib_read_kreg64(dd, kr_scratch);
+	udelay(2);
+}
+
+/*
+ * Do the absolute minimum to cause an IB speed change, and make it
+ * ready, but don't actually trigger the change.   The caller will
+ * do that when ready (if link is in Polling training state, it will
+ * happen immediately, otherwise when link next goes down)
+ *
+ * This routine should only be used as part of the DDR autonegotation
+ * code for devices that are not compliant with IB 1.2 (or code that
+ * fixes things up for same).
+ *
+ * When link has gone down, and autoneg enabled, or autoneg has
+ * failed and we give up until next time we set both speeds, and
+ * then we want IBTA enabled as well as "use max enabled speed.
+ */
+static void set_7322_ibspeed_fast(struct qib_pportdata *ppd, u32 speed)
+{
+	u64 newctrlb;
+	newctrlb = ppd->cpspec->ibcctrl_b & ~(IBA7322_IBC_SPEED_MASK |
+				    IBA7322_IBC_IBTA_1_2_MASK |
+				    IBA7322_IBC_MAX_SPEED_MASK);
+
+	if (speed & (speed - 1)) /* multiple speeds */
+		newctrlb |= (speed << IBA7322_IBC_SPEED_LSB) |
+				    IBA7322_IBC_IBTA_1_2_MASK |
+				    IBA7322_IBC_MAX_SPEED_MASK;
+	else
+		newctrlb |= speed == QIB_IB_QDR ?
+			IBA7322_IBC_SPEED_QDR | IBA7322_IBC_IBTA_1_2_MASK :
+			((speed == QIB_IB_DDR ?
+			  IBA7322_IBC_SPEED_DDR : IBA7322_IBC_SPEED_SDR));
+
+	if (newctrlb == ppd->cpspec->ibcctrl_b)
+		return;
+
+	ppd->cpspec->ibcctrl_b = newctrlb;
+	qib_write_kreg_port(ppd, krp_ibcctrl_b, ppd->cpspec->ibcctrl_b);
+	qib_write_kreg(ppd->dd, kr_scratch, 0);
+}
+
+/*
+ * This routine is only used when we are not talking to another
+ * IB 1.2-compliant device that we think can do DDR.
+ * (This includes all existing switch chips as of Oct 2007.)
+ * 1.2-compliant devices go directly to DDR prior to reaching INIT
+ */
+static void try_7322_autoneg(struct qib_pportdata *ppd)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&ppd->lflags_lock, flags);
+	ppd->lflags |= QIBL_IB_AUTONEG_INPROG;
+	spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+	qib_autoneg_7322_send(ppd, 0);
+	set_7322_ibspeed_fast(ppd, QIB_IB_DDR);
+	qib_7322_mini_pcs_reset(ppd);
+	/* 2 msec is minimum length of a poll cycle */
+	schedule_delayed_work(&ppd->cpspec->autoneg_work,
+			      msecs_to_jiffies(2));
+}
+
+/*
+ * Handle the empirically determined mechanism for auto-negotiation
+ * of DDR speed with switches.
+ */
+static void autoneg_7322_work(struct work_struct *work)
+{
+	struct qib_pportdata *ppd;
+	struct qib_devdata *dd;
+	u64 startms;
+	u32 i;
+	unsigned long flags;
+
+	ppd = container_of(work, struct qib_chippport_specific,
+			    autoneg_work.work)->ppd;
+	dd = ppd->dd;
+
+	startms = jiffies_to_msecs(jiffies);
+
+	/*
+	 * Busy wait for this first part, it should be at most a
+	 * few hundred usec, since we scheduled ourselves for 2msec.
+	 */
+	for (i = 0; i < 25; i++) {
+		if (SYM_FIELD(ppd->lastibcstat, IBCStatusA_0, LinkState)
+		     == IB_7322_LT_STATE_POLLQUIET) {
+			qib_set_linkstate(ppd, QIB_IB_LINKDOWN_DISABLE);
+			break;
+		}
+		udelay(100);
+	}
+
+	if (!(ppd->lflags & QIBL_IB_AUTONEG_INPROG))
+		goto done; /* we got there early or told to stop */
+
+	/* we expect this to timeout */
+	if (wait_event_timeout(ppd->cpspec->autoneg_wait,
+			       !(ppd->lflags & QIBL_IB_AUTONEG_INPROG),
+			       msecs_to_jiffies(90)))
+		goto done;
+	qib_7322_mini_pcs_reset(ppd);
+
+	/* we expect this to timeout */
+	if (wait_event_timeout(ppd->cpspec->autoneg_wait,
+			       !(ppd->lflags & QIBL_IB_AUTONEG_INPROG),
+			       msecs_to_jiffies(1700)))
+		goto done;
+	qib_7322_mini_pcs_reset(ppd);
+
+	set_7322_ibspeed_fast(ppd, QIB_IB_SDR);
+
+	/*
+	 * Wait up to 250 msec for link to train and get to INIT at DDR;
+	 * this should terminate early.
+	 */
+	wait_event_timeout(ppd->cpspec->autoneg_wait,
+		!(ppd->lflags & QIBL_IB_AUTONEG_INPROG),
+		msecs_to_jiffies(250));
+done:
+	if (ppd->lflags & QIBL_IB_AUTONEG_INPROG) {
+		spin_lock_irqsave(&ppd->lflags_lock, flags);
+		ppd->lflags &= ~QIBL_IB_AUTONEG_INPROG;
+		if (ppd->cpspec->autoneg_tries == AUTONEG_TRIES) {
+			ppd->lflags |= QIBL_IB_AUTONEG_FAILED;
+			ppd->cpspec->autoneg_tries = 0;
+		}
+		spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+		set_7322_ibspeed_fast(ppd, ppd->link_speed_enabled);
+	}
+}
+
+/*
+ * This routine is used to request IPG set in the QLogic switch.
+ * Only called if r1.
+ */
+static void try_7322_ipg(struct qib_pportdata *ppd)
+{
+	struct qib_ibport *ibp = &ppd->ibport_data;
+	struct ib_mad_send_buf *send_buf;
+	struct ib_mad_agent *agent;
+	struct ib_smp *smp;
+	unsigned delay;
+	int ret;
+
+	agent = ibp->send_agent;
+	if (!agent)
+		goto retry;
+
+	send_buf = ib_create_send_mad(agent, 0, 0, 0, IB_MGMT_MAD_HDR,
+				      IB_MGMT_MAD_DATA, GFP_ATOMIC);
+	if (IS_ERR(send_buf))
+		goto retry;
+
+	if (!ibp->smi_ah) {
+		struct ib_ah_attr attr;
+		struct ib_ah *ah;
+
+		memset(&attr, 0, sizeof attr);
+		attr.dlid = be16_to_cpu(IB_LID_PERMISSIVE);
+		attr.port_num = ppd->port;
+		ah = ib_create_ah(ibp->qp0->ibqp.pd, &attr);
+		if (IS_ERR(ah))
+			ret = -EINVAL;
+		else {
+			send_buf->ah = ah;
+			ibp->smi_ah = to_iah(ah);
+			ret = 0;
+		}
+	} else {
+		send_buf->ah = &ibp->smi_ah->ibah;
+		ret = 0;
+	}
+
+	smp = send_buf->mad;
+	smp->base_version = IB_MGMT_BASE_VERSION;
+	smp->mgmt_class = IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE;
+	smp->class_version = 1;
+	smp->method = IB_MGMT_METHOD_SEND;
+	smp->hop_cnt = 1;
+	smp->attr_id = QIB_VENDOR_IPG;
+	smp->attr_mod = 0;
+
+	if (!ret)
+		ret = ib_post_send_mad(send_buf, NULL);
+	if (ret)
+		ib_free_send_mad(send_buf);
+retry:
+	delay = 2 << ppd->cpspec->ipg_tries;
+	schedule_delayed_work(&ppd->cpspec->ipg_work, msecs_to_jiffies(delay));
+}
+
+/*
+ * Timeout handler for setting IPG.
+ * Only called if r1.
+ */
+static void ipg_7322_work(struct work_struct *work)
+{
+	struct qib_pportdata *ppd;
+
+	ppd = container_of(work, struct qib_chippport_specific,
+			   ipg_work.work)->ppd;
+	if ((ppd->lflags & (QIBL_LINKINIT | QIBL_LINKARMED | QIBL_LINKACTIVE))
+	    && ++ppd->cpspec->ipg_tries <= 10)
+		try_7322_ipg(ppd);
+}
+
+static u32 qib_7322_iblink_state(u64 ibcs)
+{
+	u32 state = (u32)SYM_FIELD(ibcs, IBCStatusA_0, LinkState);
+
+	switch (state) {
+	case IB_7322_L_STATE_INIT:
+		state = IB_PORT_INIT;
+		break;
+	case IB_7322_L_STATE_ARM:
+		state = IB_PORT_ARMED;
+		break;
+	case IB_7322_L_STATE_ACTIVE:
+		/* fall through */
+	case IB_7322_L_STATE_ACT_DEFER:
+		state = IB_PORT_ACTIVE;
+		break;
+	default: /* fall through */
+	case IB_7322_L_STATE_DOWN:
+		state = IB_PORT_DOWN;
+		break;
+	}
+	return state;
+}
+
+/* returns the IBTA port state, rather than the IBC link training state */
+static u8 qib_7322_phys_portstate(u64 ibcs)
+{
+	u8 state = (u8)SYM_FIELD(ibcs, IBCStatusA_0, LinkTrainingState);
+	return qib_7322_physportstate[state];
+}
+
+static int qib_7322_ib_updown(struct qib_pportdata *ppd, int ibup, u64 ibcs)
+{
+	int ret = 0, symadj = 0;
+	unsigned long flags;
+	int mult;
+
+	spin_lock_irqsave(&ppd->lflags_lock, flags);
+	ppd->lflags &= ~QIBL_IB_FORCE_NOTIFY;
+	spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+
+	/* Update our picture of width and speed from chip */
+	if (ibcs & SYM_MASK(IBCStatusA_0, LinkSpeedQDR)) {
+		ppd->link_speed_active = QIB_IB_QDR;
+		mult = 4;
+	} else if (ibcs & SYM_MASK(IBCStatusA_0, LinkSpeedActive)) {
+		ppd->link_speed_active = QIB_IB_DDR;
+		mult = 2;
+	} else {
+		ppd->link_speed_active = QIB_IB_SDR;
+		mult = 1;
+	}
+	if (ibcs & SYM_MASK(IBCStatusA_0, LinkWidthActive)) {
+		ppd->link_width_active = IB_WIDTH_4X;
+		mult *= 4;
+	} else
+		ppd->link_width_active = IB_WIDTH_1X;
+	ppd->delay_mult = ib_rate_to_delay[mult_to_ib_rate(mult)];
+
+	if (!ibup) {
+		u64 clr;
+
+		/* Link went down. */
+		/* do IPG MAD again after linkdown, even if last time failed */
+		ppd->cpspec->ipg_tries = 0;
+		clr = qib_read_kreg_port(ppd, krp_ibcstatus_b) &
+			(SYM_MASK(IBCStatusB_0, heartbeat_timed_out) |
+			 SYM_MASK(IBCStatusB_0, heartbeat_crosstalk));
+		if (clr)
+			qib_write_kreg_port(ppd, krp_ibcstatus_b, clr);
+		if (!(ppd->lflags & (QIBL_IB_AUTONEG_FAILED |
+				     QIBL_IB_AUTONEG_INPROG)))
+			set_7322_ibspeed_fast(ppd, ppd->link_speed_enabled);
+		if (!(ppd->lflags & QIBL_IB_AUTONEG_INPROG)) {
+			/* unlock the Tx settings, speed may change */
+			qib_write_kreg_port(ppd, krp_tx_deemph_override,
+				SYM_MASK(IBSD_TX_DEEMPHASIS_OVERRIDE_0,
+				reset_tx_deemphasis_override));
+			qib_cancel_sends(ppd);
+			/* on link down, ensure sane pcs state */
+			qib_7322_mini_pcs_reset(ppd);
+			spin_lock_irqsave(&ppd->sdma_lock, flags);
+			if (__qib_sdma_running(ppd))
+				__qib_sdma_process_event(ppd,
+					qib_sdma_event_e70_go_idle);
+			spin_unlock_irqrestore(&ppd->sdma_lock, flags);
+		}
+		clr = read_7322_creg32_port(ppd, crp_iblinkdown);
+		if (clr == ppd->cpspec->iblnkdownsnap)
+			ppd->cpspec->iblnkdowndelta++;
+	} else {
+		if (qib_compat_ddr_negotiate &&
+		    !(ppd->lflags & (QIBL_IB_AUTONEG_FAILED |
+				     QIBL_IB_AUTONEG_INPROG)) &&
+		    ppd->link_speed_active == QIB_IB_SDR &&
+		    (ppd->link_speed_enabled & QIB_IB_DDR)
+		    && ppd->cpspec->autoneg_tries < AUTONEG_TRIES) {
+			/* we are SDR, and auto-negotiation enabled */
+			++ppd->cpspec->autoneg_tries;
+			if (!ppd->cpspec->ibdeltainprog) {
+				ppd->cpspec->ibdeltainprog = 1;
+				ppd->cpspec->ibsymdelta +=
+					read_7322_creg32_port(ppd,
+						crp_ibsymbolerr) -
+						ppd->cpspec->ibsymsnap;
+				ppd->cpspec->iblnkerrdelta +=
+					read_7322_creg32_port(ppd,
+						crp_iblinkerrrecov) -
+						ppd->cpspec->iblnkerrsnap;
+			}
+			try_7322_autoneg(ppd);
+			ret = 1; /* no other IB status change processing */
+		} else if ((ppd->lflags & QIBL_IB_AUTONEG_INPROG) &&
+			   ppd->link_speed_active == QIB_IB_SDR) {
+			qib_autoneg_7322_send(ppd, 1);
+			set_7322_ibspeed_fast(ppd, QIB_IB_DDR);
+			qib_7322_mini_pcs_reset(ppd);
+			udelay(2);
+			ret = 1; /* no other IB status change processing */
+		} else if ((ppd->lflags & QIBL_IB_AUTONEG_INPROG) &&
+			   (ppd->link_speed_active & QIB_IB_DDR)) {
+			spin_lock_irqsave(&ppd->lflags_lock, flags);
+			ppd->lflags &= ~(QIBL_IB_AUTONEG_INPROG |
+					 QIBL_IB_AUTONEG_FAILED);
+			spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+			ppd->cpspec->autoneg_tries = 0;
+			/* re-enable SDR, for next link down */
+			set_7322_ibspeed_fast(ppd, ppd->link_speed_enabled);
+			wake_up(&ppd->cpspec->autoneg_wait);
+			symadj = 1;
+		} else if (ppd->lflags & QIBL_IB_AUTONEG_FAILED) {
+			/*
+			 * Clear autoneg failure flag, and do setup
+			 * so we'll try next time link goes down and
+			 * back to INIT (possibly connected to a
+			 * different device).
+			 */
+			spin_lock_irqsave(&ppd->lflags_lock, flags);
+			ppd->lflags &= ~QIBL_IB_AUTONEG_FAILED;
+			spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+			ppd->cpspec->ibcctrl_b |= IBA7322_IBC_IBTA_1_2_MASK;
+			symadj = 1;
+		}
+		if (!(ppd->lflags & QIBL_IB_AUTONEG_INPROG)) {
+			symadj = 1;
+			if (ppd->dd->cspec->r1 && ppd->cpspec->ipg_tries <= 10)
+				try_7322_ipg(ppd);
+			if (!ppd->cpspec->recovery_init)
+				setup_7322_link_recovery(ppd, 0);
+			ppd->cpspec->qdr_dfe_time = jiffies +
+				msecs_to_jiffies(QDR_DFE_DISABLE_DELAY);
+		}
+		ppd->cpspec->ibmalfusesnap = 0;
+		ppd->cpspec->ibmalfsnap = read_7322_creg32_port(ppd,
+			crp_errlink);
+	}
+	if (symadj) {
+		ppd->cpspec->iblnkdownsnap =
+			read_7322_creg32_port(ppd, crp_iblinkdown);
+		if (ppd->cpspec->ibdeltainprog) {
+			ppd->cpspec->ibdeltainprog = 0;
+			ppd->cpspec->ibsymdelta += read_7322_creg32_port(ppd,
+				crp_ibsymbolerr) - ppd->cpspec->ibsymsnap;
+			ppd->cpspec->iblnkerrdelta += read_7322_creg32_port(ppd,
+				crp_iblinkerrrecov) - ppd->cpspec->iblnkerrsnap;
+		}
+	} else if (!ibup && qib_compat_ddr_negotiate &&
+		   !ppd->cpspec->ibdeltainprog &&
+			!(ppd->lflags & QIBL_IB_AUTONEG_INPROG)) {
+		ppd->cpspec->ibdeltainprog = 1;
+		ppd->cpspec->ibsymsnap = read_7322_creg32_port(ppd,
+			crp_ibsymbolerr);
+		ppd->cpspec->iblnkerrsnap = read_7322_creg32_port(ppd,
+			crp_iblinkerrrecov);
+	}
+
+	if (!ret)
+		qib_setup_7322_setextled(ppd, ibup);
+	return ret;
+}
+
+/*
+ * Does read/modify/write to appropriate registers to
+ * set output and direction bits selected by mask.
+ * these are in their canonical postions (e.g. lsb of
+ * dir will end up in D48 of extctrl on existing chips).
+ * returns contents of GP Inputs.
+ */
+static int gpio_7322_mod(struct qib_devdata *dd, u32 out, u32 dir, u32 mask)
+{
+	u64 read_val, new_out;
+	unsigned long flags;
+
+	if (mask) {
+		/* some bits being written, lock access to GPIO */
+		dir &= mask;
+		out &= mask;
+		spin_lock_irqsave(&dd->cspec->gpio_lock, flags);
+		dd->cspec->extctrl &= ~((u64)mask << SYM_LSB(EXTCtrl, GPIOOe));
+		dd->cspec->extctrl |= ((u64) dir << SYM_LSB(EXTCtrl, GPIOOe));
+		new_out = (dd->cspec->gpio_out & ~mask) | out;
+
+		qib_write_kreg(dd, kr_extctrl, dd->cspec->extctrl);
+		qib_write_kreg(dd, kr_gpio_out, new_out);
+		dd->cspec->gpio_out = new_out;
+		spin_unlock_irqrestore(&dd->cspec->gpio_lock, flags);
+	}
+	/*
+	 * It is unlikely that a read at this time would get valid
+	 * data on a pin whose direction line was set in the same
+	 * call to this function. We include the read here because
+	 * that allows us to potentially combine a change on one pin with
+	 * a read on another, and because the old code did something like
+	 * this.
+	 */
+	read_val = qib_read_kreg64(dd, kr_extstatus);
+	return SYM_FIELD(read_val, EXTStatus, GPIOIn);
+}
+
+/* Enable writes to config EEPROM, if possible. Returns previous state */
+static int qib_7322_eeprom_wen(struct qib_devdata *dd, int wen)
+{
+	int prev_wen;
+	u32 mask;
+
+	mask = 1 << QIB_EEPROM_WEN_NUM;
+	prev_wen = ~gpio_7322_mod(dd, 0, 0, 0) >> QIB_EEPROM_WEN_NUM;
+	gpio_7322_mod(dd, wen ? 0 : mask, mask, mask);
+
+	return prev_wen & 1;
+}
+
+/*
+ * Read fundamental info we need to use the chip.  These are
+ * the registers that describe chip capabilities, and are
+ * saved in shadow registers.
+ */
+static void get_7322_chip_params(struct qib_devdata *dd)
+{
+	u64 val;
+	u32 piobufs;
+	int mtu;
+
+	dd->palign = qib_read_kreg32(dd, kr_pagealign);
+
+	dd->uregbase = qib_read_kreg32(dd, kr_userregbase);
+
+	dd->rcvtidcnt = qib_read_kreg32(dd, kr_rcvtidcnt);
+	dd->rcvtidbase = qib_read_kreg32(dd, kr_rcvtidbase);
+	dd->rcvegrbase = qib_read_kreg32(dd, kr_rcvegrbase);
+	dd->piobufbase = qib_read_kreg64(dd, kr_sendpiobufbase);
+	dd->pio2k_bufbase = dd->piobufbase & 0xffffffff;
+
+	val = qib_read_kreg64(dd, kr_sendpiobufcnt);
+	dd->piobcnt2k = val & ~0U;
+	dd->piobcnt4k = val >> 32;
+	val = qib_read_kreg64(dd, kr_sendpiosize);
+	dd->piosize2k = val & ~0U;
+	dd->piosize4k = val >> 32;
+
+	mtu = ib_mtu_enum_to_int(qib_ibmtu);
+	if (mtu == -1)
+		mtu = QIB_DEFAULT_MTU;
+	dd->pport[0].ibmtu = (u32)mtu;
+	dd->pport[1].ibmtu = (u32)mtu;
+
+	/* these may be adjusted in init_chip_wc_pat() */
+	dd->pio2kbase = (u32 __iomem *)
+		((char __iomem *) dd->kregbase + dd->pio2k_bufbase);
+	dd->pio4kbase = (u32 __iomem *)
+		((char __iomem *) dd->kregbase +
+		 (dd->piobufbase >> 32));
+	/*
+	 * 4K buffers take 2 pages; we use roundup just to be
+	 * paranoid; we calculate it once here, rather than on
+	 * ever buf allocate
+	 */
+	dd->align4k = ALIGN(dd->piosize4k, dd->palign);
+
+	piobufs = dd->piobcnt4k + dd->piobcnt2k + NUM_VL15_BUFS;
+
+	dd->pioavregs = ALIGN(piobufs, sizeof(u64) * BITS_PER_BYTE / 2) /
+		(sizeof(u64) * BITS_PER_BYTE / 2);
+}
+
+/*
+ * The chip base addresses in cspec and cpspec have to be set
+ * after possible init_chip_wc_pat(), rather than in
+ * get_7322_chip_params(), so split out as separate function
+ */
+static void qib_7322_set_baseaddrs(struct qib_devdata *dd)
+{
+	u32 cregbase;
+	cregbase = qib_read_kreg32(dd, kr_counterregbase);
+
+	dd->cspec->cregbase = (u64 __iomem *)(cregbase +
+		(char __iomem *)dd->kregbase);
+
+	dd->egrtidbase = (u64 __iomem *)
+		((char __iomem *) dd->kregbase + dd->rcvegrbase);
+
+	/* port registers are defined as relative to base of chip */
+	dd->pport[0].cpspec->kpregbase =
+		(u64 __iomem *)((char __iomem *)dd->kregbase);
+	dd->pport[1].cpspec->kpregbase =
+		(u64 __iomem *)(dd->palign +
+		(char __iomem *)dd->kregbase);
+	dd->pport[0].cpspec->cpregbase =
+		(u64 __iomem *)(qib_read_kreg_port(&dd->pport[0],
+		kr_counterregbase) + (char __iomem *)dd->kregbase);
+	dd->pport[1].cpspec->cpregbase =
+		(u64 __iomem *)(qib_read_kreg_port(&dd->pport[1],
+		kr_counterregbase) + (char __iomem *)dd->kregbase);
+}
+
+/*
+ * This is a fairly special-purpose observer, so we only support
+ * the port-specific parts of SendCtrl
+ */
+
+#define SENDCTRL_SHADOWED (SYM_MASK(SendCtrl_0, SendEnable) |		\
+			   SYM_MASK(SendCtrl_0, SDmaEnable) |		\
+			   SYM_MASK(SendCtrl_0, SDmaIntEnable) |	\
+			   SYM_MASK(SendCtrl_0, SDmaSingleDescriptor) | \
+			   SYM_MASK(SendCtrl_0, SDmaHalt) |		\
+			   SYM_MASK(SendCtrl_0, IBVLArbiterEn) |	\
+			   SYM_MASK(SendCtrl_0, ForceCreditUpToDate))
+
+static int sendctrl_hook(struct qib_devdata *dd,
+			 const struct diag_observer *op, u32 offs,
+			 u64 *data, u64 mask, int only_32)
+{
+	unsigned long flags;
+	unsigned idx;
+	unsigned pidx;
+	struct qib_pportdata *ppd = NULL;
+	u64 local_data, all_bits;
+
+	/*
+	 * The fixed correspondence between Physical ports and pports is
+	 * severed. We need to hunt for the ppd that corresponds
+	 * to the offset we got. And we have to do that without admitting
+	 * we know the stride, apparently.
+	 */
+	for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+		u64 __iomem *psptr;
+		u32 psoffs;
+
+		ppd = dd->pport + pidx;
+		if (!ppd->cpspec->kpregbase)
+			continue;
+
+		psptr = ppd->cpspec->kpregbase + krp_sendctrl;
+		psoffs = (u32) (psptr - dd->kregbase) * sizeof(*psptr);
+		if (psoffs == offs)
+			break;
+	}
+
+	/* If pport is not being managed by driver, just avoid shadows. */
+	if (pidx >= dd->num_pports)
+		ppd = NULL;
+
+	/* In any case, "idx" is flat index in kreg space */
+	idx = offs / sizeof(u64);
+
+	all_bits = ~0ULL;
+	if (only_32)
+		all_bits >>= 32;
+
+	spin_lock_irqsave(&dd->sendctrl_lock, flags);
+	if (!ppd || (mask & all_bits) != all_bits) {
+		/*
+		 * At least some mask bits are zero, so we need
+		 * to read. The judgement call is whether from
+		 * reg or shadow. First-cut: read reg, and complain
+		 * if any bits which should be shadowed are different
+		 * from their shadowed value.
+		 */
+		if (only_32)
+			local_data = (u64)qib_read_kreg32(dd, idx);
+		else
+			local_data = qib_read_kreg64(dd, idx);
+		*data = (local_data & ~mask) | (*data & mask);
+	}
+	if (mask) {
+		/*
+		 * At least some mask bits are one, so we need
+		 * to write, but only shadow some bits.
+		 */
+		u64 sval, tval; /* Shadowed, transient */
+
+		/*
+		 * New shadow val is bits we don't want to touch,
+		 * ORed with bits we do, that are intended for shadow.
+		 */
+		if (ppd) {
+			sval = ppd->p_sendctrl & ~mask;
+			sval |= *data & SENDCTRL_SHADOWED & mask;
+			ppd->p_sendctrl = sval;
+		} else
+			sval = *data & SENDCTRL_SHADOWED & mask;
+		tval = sval | (*data & ~SENDCTRL_SHADOWED & mask);
+		qib_write_kreg(dd, idx, tval);
+		qib_write_kreg(dd, kr_scratch, 0Ull);
+	}
+	spin_unlock_irqrestore(&dd->sendctrl_lock, flags);
+	return only_32 ? 4 : 8;
+}
+
+static const struct diag_observer sendctrl_0_observer = {
+	sendctrl_hook, KREG_IDX(SendCtrl_0) * sizeof(u64),
+	KREG_IDX(SendCtrl_0) * sizeof(u64)
+};
+
+static const struct diag_observer sendctrl_1_observer = {
+	sendctrl_hook, KREG_IDX(SendCtrl_1) * sizeof(u64),
+	KREG_IDX(SendCtrl_1) * sizeof(u64)
+};
+
+static ushort sdma_fetch_prio = 8;
+module_param_named(sdma_fetch_prio, sdma_fetch_prio, ushort, S_IRUGO);
+MODULE_PARM_DESC(sdma_fetch_prio, "SDMA descriptor fetch priority");
+
+/* Besides logging QSFP events, we set appropriate TxDDS values */
+static void init_txdds_table(struct qib_pportdata *ppd, int override);
+
+static void qsfp_7322_event(struct work_struct *work)
+{
+	struct qib_qsfp_data *qd;
+	struct qib_pportdata *ppd;
+	u64 pwrup;
+	int ret;
+	u32 le2;
+
+	qd = container_of(work, struct qib_qsfp_data, work);
+	ppd = qd->ppd;
+	pwrup = qd->t_insert + msecs_to_jiffies(QSFP_PWR_LAG_MSEC);
+
+	/*
+	 * Some QSFP's not only do not respond until the full power-up
+	 * time, but may behave badly if we try. So hold off responding
+	 * to insertion.
+	 */
+	while (1) {
+		u64 now = get_jiffies_64();
+		if (time_after64(now, pwrup))
+			break;
+		msleep(1);
+	}
+	ret = qib_refresh_qsfp_cache(ppd, &qd->cache);
+	/*
+	 * Need to change LE2 back to defaults if we couldn't
+	 * read the cable type (to handle cable swaps), so do this
+	 * even on failure to read cable information.  We don't
+	 * get here for QME, so IS_QME check not needed here.
+	 */
+	le2 = (!ret && qd->cache.atten[1] >= qib_long_atten &&
+	       !ppd->dd->cspec->r1 && QSFP_IS_CU(qd->cache.tech)) ?
+		LE2_5m : LE2_DEFAULT;
+	ibsd_wr_allchans(ppd, 13, (le2 << 7), BMASK(9, 7));
+	init_txdds_table(ppd, 0);
+}
+
+/*
+ * There is little we can do but complain to the user if QSFP
+ * initialization fails.
+ */
+static void qib_init_7322_qsfp(struct qib_pportdata *ppd)
+{
+	unsigned long flags;
+	struct qib_qsfp_data *qd = &ppd->cpspec->qsfp_data;
+	struct qib_devdata *dd = ppd->dd;
+	u64 mod_prs_bit = QSFP_GPIO_MOD_PRS_N;
+
+	mod_prs_bit <<= (QSFP_GPIO_PORT2_SHIFT * ppd->hw_pidx);
+	qd->ppd = ppd;
+	qib_qsfp_init(qd, qsfp_7322_event);
+	spin_lock_irqsave(&dd->cspec->gpio_lock, flags);
+	dd->cspec->extctrl |= (mod_prs_bit << SYM_LSB(EXTCtrl, GPIOInvert));
+	dd->cspec->gpio_mask |= mod_prs_bit;
+	qib_write_kreg(dd, kr_extctrl, dd->cspec->extctrl);
+	qib_write_kreg(dd, kr_gpio_mask, dd->cspec->gpio_mask);
+	spin_unlock_irqrestore(&dd->cspec->gpio_lock, flags);
+}
+
+/*
+ * called at device initialization time, and also if the txselect
+ * module parameter is changed.  This is used for cables that don't
+ * have valid QSFP EEPROMs (not present, or attenuation is zero).
+ * We initialize to the default, then if there is a specific
+ * unit,port match, we use that (and set it immediately, for the
+ * current speed, if the link is at INIT or better).
+ * String format is "default# unit#,port#=# ... u,p=#", separators must
+ * be a SPACE character.  A newline terminates.  The u,p=# tuples may
+ * optionally have "u,p=#,#", where the final # is the H1 value
+ * The last specific match is used (actually, all are used, but last
+ * one is the one that winds up set); if none at all, fall back on default.
+ */
+static void set_no_qsfp_atten(struct qib_devdata *dd, int change)
+{
+	char *nxt, *str;
+	u32 pidx, unit, port, deflt, h1;
+	unsigned long val;
+	int any = 0, seth1;
+
+	str = txselect_list;
+
+	/* default number is validated in setup_txselect() */
+	deflt = simple_strtoul(str, &nxt, 0);
+	for (pidx = 0; pidx < dd->num_pports; ++pidx)
+		dd->pport[pidx].cpspec->no_eep = deflt;
+
+	while (*nxt && nxt[1]) {
+		str = ++nxt;
+		unit = simple_strtoul(str, &nxt, 0);
+		if (nxt == str || !*nxt || *nxt != ',') {
+			while (*nxt && *nxt++ != ' ') /* skip to next, if any */
+				;
+			continue;
+		}
+		str = ++nxt;
+		port = simple_strtoul(str, &nxt, 0);
+		if (nxt == str || *nxt != '=') {
+			while (*nxt && *nxt++ != ' ') /* skip to next, if any */
+				;
+			continue;
+		}
+		str = ++nxt;
+		val = simple_strtoul(str, &nxt, 0);
+		if (nxt == str) {
+			while (*nxt && *nxt++ != ' ') /* skip to next, if any */
+				;
+			continue;
+		}
+		if (val >= TXDDS_TABLE_SZ + TXDDS_EXTRA_SZ)
+			continue;
+		seth1 = 0;
+		h1 = 0; /* gcc thinks it might be used uninitted */
+		if (*nxt == ',' && nxt[1]) {
+			str = ++nxt;
+			h1 = (u32)simple_strtoul(str, &nxt, 0);
+			if (nxt == str)
+				while (*nxt && *nxt++ != ' ') /* skip */
+					;
+			else
+				seth1 = 1;
+		}
+		for (pidx = 0; dd->unit == unit && pidx < dd->num_pports;
+		     ++pidx) {
+			struct qib_pportdata *ppd = &dd->pport[pidx];
+
+			if (ppd->port != port || !ppd->link_speed_supported)
+				continue;
+			ppd->cpspec->no_eep = val;
+			/* now change the IBC and serdes, overriding generic */
+			init_txdds_table(ppd, 1);
+			any++;
+		}
+		if (*nxt == '\n')
+			break; /* done */
+	}
+	if (change && !any) {
+		/* no specific setting, use the default.
+		 * Change the IBC and serdes, but since it's
+		 * general, don't override specific settings.
+		 */
+		for (pidx = 0; pidx < dd->num_pports; ++pidx)
+			if (dd->pport[pidx].link_speed_supported)
+				init_txdds_table(&dd->pport[pidx], 0);
+	}
+}
+
+/* handle the txselect parameter changing */
+static int setup_txselect(const char *str, struct kernel_param *kp)
+{
+	struct qib_devdata *dd;
+	unsigned long val;
+	char *n;
+	if (strlen(str) >= MAX_ATTEN_LEN) {
+		printk(KERN_INFO QIB_DRV_NAME " txselect_values string "
+		       "too long\n");
+		return -ENOSPC;
+	}
+	val = simple_strtoul(str, &n, 0);
+	if (n == str || val >= (TXDDS_TABLE_SZ + TXDDS_EXTRA_SZ)) {
+		printk(KERN_INFO QIB_DRV_NAME
+		       "txselect_values must start with a number < %d\n",
+			TXDDS_TABLE_SZ + TXDDS_EXTRA_SZ);
+		return -EINVAL;
+	}
+	strcpy(txselect_list, str);
+
+	list_for_each_entry(dd, &qib_dev_list, list)
+		if (dd->deviceid == PCI_DEVICE_ID_QLOGIC_IB_7322)
+			set_no_qsfp_atten(dd, 1);
+	return 0;
+}
+
+/*
+ * Write the final few registers that depend on some of the
+ * init setup.  Done late in init, just before bringing up
+ * the serdes.
+ */
+static int qib_late_7322_initreg(struct qib_devdata *dd)
+{
+	int ret = 0, n;
+	u64 val;
+
+	qib_write_kreg(dd, kr_rcvhdrentsize, dd->rcvhdrentsize);
+	qib_write_kreg(dd, kr_rcvhdrsize, dd->rcvhdrsize);
+	qib_write_kreg(dd, kr_rcvhdrcnt, dd->rcvhdrcnt);
+	qib_write_kreg(dd, kr_sendpioavailaddr, dd->pioavailregs_phys);
+	val = qib_read_kreg64(dd, kr_sendpioavailaddr);
+	if (val != dd->pioavailregs_phys) {
+		qib_dev_err(dd, "Catastrophic software error, "
+			    "SendPIOAvailAddr written as %lx, "
+			    "read back as %llx\n",
+			    (unsigned long) dd->pioavailregs_phys,
+			    (unsigned long long) val);
+		ret = -EINVAL;
+	}
+
+	n = dd->piobcnt2k + dd->piobcnt4k + NUM_VL15_BUFS;
+	qib_7322_txchk_change(dd, 0, n, TXCHK_CHG_TYPE_KERN, NULL);
+	/* driver sends get pkey, lid, etc. checking also, to catch bugs */
+	qib_7322_txchk_change(dd, 0, n, TXCHK_CHG_TYPE_ENAB1, NULL);
+
+	qib_register_observer(dd, &sendctrl_0_observer);
+	qib_register_observer(dd, &sendctrl_1_observer);
+
+	dd->control &= ~QLOGIC_IB_C_SDMAFETCHPRIOEN;
+	qib_write_kreg(dd, kr_control, dd->control);
+	/*
+	 * Set SendDmaFetchPriority and init Tx params, including
+	 * QSFP handler on boards that have QSFP.
+	 * First set our default attenuation entry for cables that
+	 * don't have valid attenuation.
+	 */
+	set_no_qsfp_atten(dd, 0);
+	for (n = 0; n < dd->num_pports; ++n) {
+		struct qib_pportdata *ppd = dd->pport + n;
+
+		qib_write_kreg_port(ppd, krp_senddmaprioritythld,
+				    sdma_fetch_prio & 0xf);
+		/* Initialize qsfp if present on board. */
+		if (dd->flags & QIB_HAS_QSFP)
+			qib_init_7322_qsfp(ppd);
+	}
+	dd->control |= QLOGIC_IB_C_SDMAFETCHPRIOEN;
+	qib_write_kreg(dd, kr_control, dd->control);
+
+	return ret;
+}
+
+/* per IB port errors.  */
+#define SENDCTRL_PIBP (MASK_ACROSS(0, 1) | MASK_ACROSS(3, 3) | \
+	MASK_ACROSS(8, 15))
+#define RCVCTRL_PIBP (MASK_ACROSS(0, 17) | MASK_ACROSS(39, 41))
+#define ERRS_PIBP (MASK_ACROSS(57, 58) | MASK_ACROSS(54, 54) | \
+	MASK_ACROSS(36, 49) | MASK_ACROSS(29, 34) | MASK_ACROSS(14, 17) | \
+	MASK_ACROSS(0, 11))
+
+/*
+ * Write the initialization per-port registers that need to be done at
+ * driver load and after reset completes (i.e., that aren't done as part
+ * of other init procedures called from qib_init.c).
+ * Some of these should be redundant on reset, but play safe.
+ */
+static void write_7322_init_portregs(struct qib_pportdata *ppd)
+{
+	u64 val;
+	int i;
+
+	if (!ppd->link_speed_supported) {
+		/* no buffer credits for this port */
+		for (i = 1; i < 8; i++)
+			qib_write_kreg_port(ppd, krp_rxcreditvl0 + i, 0);
+		qib_write_kreg_port(ppd, krp_ibcctrl_b, 0);
+		qib_write_kreg(ppd->dd, kr_scratch, 0);
+		return;
+	}
+
+	/*
+	 * Set the number of supported virtual lanes in IBC,
+	 * for flow control packet handling on unsupported VLs
+	 */
+	val = qib_read_kreg_port(ppd, krp_ibsdtestiftx);
+	val &= ~SYM_MASK(IB_SDTEST_IF_TX_0, VL_CAP);
+	val |= (u64)(ppd->vls_supported - 1) <<
+		SYM_LSB(IB_SDTEST_IF_TX_0, VL_CAP);
+	qib_write_kreg_port(ppd, krp_ibsdtestiftx, val);
+
+	qib_write_kreg_port(ppd, krp_rcvbthqp, QIB_KD_QP);
+
+	/* enable tx header checking */
+	qib_write_kreg_port(ppd, krp_sendcheckcontrol, IBA7322_SENDCHK_PKEY |
+			    IBA7322_SENDCHK_BTHQP | IBA7322_SENDCHK_SLID |
+			    IBA7322_SENDCHK_RAW_IPV6 | IBA7322_SENDCHK_MINSZ);
+
+	qib_write_kreg_port(ppd, krp_ncmodectrl,
+		SYM_MASK(IBNCModeCtrl_0, ScrambleCapLocal));
+
+	/*
+	 * Unconditionally clear the bufmask bits.  If SDMA is
+	 * enabled, we'll set them appropriately later.
+	 */
+	qib_write_kreg_port(ppd, krp_senddmabufmask0, 0);
+	qib_write_kreg_port(ppd, krp_senddmabufmask1, 0);
+	qib_write_kreg_port(ppd, krp_senddmabufmask2, 0);
+	if (ppd->dd->cspec->r1)
+		ppd->p_sendctrl |= SYM_MASK(SendCtrl_0, ForceCreditUpToDate);
+}
+
+/*
+ * Write the initialization per-device registers that need to be done at
+ * driver load and after reset completes (i.e., that aren't done as part
+ * of other init procedures called from qib_init.c).  Also write per-port
+ * registers that are affected by overall device config, such as QP mapping
+ * Some of these should be redundant on reset, but play safe.
+ */
+static void write_7322_initregs(struct qib_devdata *dd)
+{
+	struct qib_pportdata *ppd;
+	int i, pidx;
+	u64 val;
+
+	/* Set Multicast QPs received by port 2 to map to context one. */
+	qib_write_kreg(dd, KREG_IDX(RcvQPMulticastContext_1), 1);
+
+	for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+		unsigned n, regno;
+		unsigned long flags;
+
+		if (!dd->qpn_mask || !dd->pport[pidx].link_speed_supported)
+			continue;
+
+		ppd = &dd->pport[pidx];
+
+		/* be paranoid against later code motion, etc. */
+		spin_lock_irqsave(&dd->cspec->rcvmod_lock, flags);
+		ppd->p_rcvctrl |= SYM_MASK(RcvCtrl_0, RcvQPMapEnable);
+		spin_unlock_irqrestore(&dd->cspec->rcvmod_lock, flags);
+
+		/* Initialize QP to context mapping */
+		regno = krp_rcvqpmaptable;
+		val = 0;
+		if (dd->num_pports > 1)
+			n = dd->first_user_ctxt / dd->num_pports;
+		else
+			n = dd->first_user_ctxt - 1;
+		for (i = 0; i < 32; ) {
+			unsigned ctxt;
+
+			if (dd->num_pports > 1)
+				ctxt = (i % n) * dd->num_pports + pidx;
+			else if (i % n)
+				ctxt = (i % n) + 1;
+			else
+				ctxt = ppd->hw_pidx;
+			val |= ctxt << (5 * (i % 6));
+			i++;
+			if (i % 6 == 0) {
+				qib_write_kreg_port(ppd, regno, val);
+				val = 0;
+				regno++;
+			}
+		}
+		qib_write_kreg_port(ppd, regno, val);
+	}
+
+	/*
+	 * Setup up interrupt mitigation for kernel contexts, but
+	 * not user contexts (user contexts use interrupts when
+	 * stalled waiting for any packet, so want those interrupts
+	 * right away).
+	 */
+	for (i = 0; i < dd->first_user_ctxt; i++) {
+		dd->cspec->rcvavail_timeout[i] = rcv_int_timeout;
+		qib_write_kreg(dd, kr_rcvavailtimeout + i, rcv_int_timeout);
+	}
+
+	/*
+	 * Initialize  as (disabled) rcvflow tables.  Application code
+	 * will setup each flow as it uses the flow.
+	 * Doesn't clear any of the error bits that might be set.
+	 */
+	val = TIDFLOW_ERRBITS; /* these are W1C */
+	for (i = 0; i < dd->ctxtcnt; i++) {
+		int flow;
+		for (flow = 0; flow < NUM_TIDFLOWS_CTXT; flow++)
+			qib_write_ureg(dd, ur_rcvflowtable+flow, val, i);
+	}
+
+	/*
+	 * dual cards init to dual port recovery, single port cards to
+	 * the one port.  Dual port cards may later adjust to 1 port,
+	 * and then back to dual port if both ports are connected
+	 * */
+	if (dd->num_pports)
+		setup_7322_link_recovery(dd->pport, dd->num_pports > 1);
+}
+
+static int qib_init_7322_variables(struct qib_devdata *dd)
+{
+	struct qib_pportdata *ppd;
+	unsigned features, pidx, sbufcnt;
+	int ret, mtu;
+	u32 sbufs, updthresh;
+
+	/* pport structs are contiguous, allocated after devdata */
+	ppd = (struct qib_pportdata *)(dd + 1);
+	dd->pport = ppd;
+	ppd[0].dd = dd;
+	ppd[1].dd = dd;
+
+	dd->cspec = (struct qib_chip_specific *)(ppd + 2);
+
+	ppd[0].cpspec = (struct qib_chippport_specific *)(dd->cspec + 1);
+	ppd[1].cpspec = &ppd[0].cpspec[1];
+	ppd[0].cpspec->ppd = &ppd[0]; /* for autoneg_7322_work() */
+	ppd[1].cpspec->ppd = &ppd[1]; /* for autoneg_7322_work() */
+
+	spin_lock_init(&dd->cspec->rcvmod_lock);
+	spin_lock_init(&dd->cspec->gpio_lock);
+
+	/* we haven't yet set QIB_PRESENT, so use read directly */
+	dd->revision = readq(&dd->kregbase[kr_revision]);
+
+	if ((dd->revision & 0xffffffffU) == 0xffffffffU) {
+		qib_dev_err(dd, "Revision register read failure, "
+			    "giving up initialization\n");
+		ret = -ENODEV;
+		goto bail;
+	}
+	dd->flags |= QIB_PRESENT;  /* now register routines work */
+
+	dd->majrev = (u8) SYM_FIELD(dd->revision, Revision_R, ChipRevMajor);
+	dd->minrev = (u8) SYM_FIELD(dd->revision, Revision_R, ChipRevMinor);
+	dd->cspec->r1 = dd->minrev == 1;
+
+	get_7322_chip_params(dd);
+	features = qib_7322_boardname(dd);
+
+	/* now that piobcnt2k and 4k set, we can allocate these */
+	sbufcnt = dd->piobcnt2k + dd->piobcnt4k +
+		NUM_VL15_BUFS + BITS_PER_LONG - 1;
+	sbufcnt /= BITS_PER_LONG;
+	dd->cspec->sendchkenable = kmalloc(sbufcnt *
+		sizeof(*dd->cspec->sendchkenable), GFP_KERNEL);
+	dd->cspec->sendgrhchk = kmalloc(sbufcnt *
+		sizeof(*dd->cspec->sendgrhchk), GFP_KERNEL);
+	dd->cspec->sendibchk = kmalloc(sbufcnt *
+		sizeof(*dd->cspec->sendibchk), GFP_KERNEL);
+	if (!dd->cspec->sendchkenable || !dd->cspec->sendgrhchk ||
+		!dd->cspec->sendibchk) {
+		qib_dev_err(dd, "Failed allocation for hdrchk bitmaps\n");
+		ret = -ENOMEM;
+		goto bail;
+	}
+
+	ppd = dd->pport;
+
+	/*
+	 * GPIO bits for TWSI data and clock,
+	 * used for serial EEPROM.
+	 */
+	dd->gpio_sda_num = _QIB_GPIO_SDA_NUM;
+	dd->gpio_scl_num = _QIB_GPIO_SCL_NUM;
+	dd->twsi_eeprom_dev = QIB_TWSI_EEPROM_DEV;
+
+	dd->flags |= QIB_HAS_INTX | QIB_HAS_LINK_LATENCY |
+		QIB_NODMA_RTAIL | QIB_HAS_VLSUPP | QIB_HAS_HDRSUPP |
+		QIB_HAS_THRESH_UPDATE |
+		(sdma_idle_cnt ? QIB_HAS_SDMA_TIMEOUT : 0);
+	dd->flags |= qib_special_trigger ?
+		QIB_USE_SPCL_TRIG : QIB_HAS_SEND_DMA;
+
+	/*
+	 * Setup initial values.  These may change when PAT is enabled, but
+	 * we need these to do initial chip register accesses.
+	 */
+	qib_7322_set_baseaddrs(dd);
+
+	mtu = ib_mtu_enum_to_int(qib_ibmtu);
+	if (mtu == -1)
+		mtu = QIB_DEFAULT_MTU;
+
+	dd->cspec->int_enable_mask = QIB_I_BITSEXTANT;
+	/* all hwerrors become interrupts, unless special purposed */
+	dd->cspec->hwerrmask = ~0ULL;
+	/*  link_recovery setup causes these errors, so ignore them,
+	 *  other than clearing them when they occur */
+	dd->cspec->hwerrmask &=
+		~(SYM_MASK(HwErrMask, IBSerdesPClkNotDetectMask_0) |
+		  SYM_MASK(HwErrMask, IBSerdesPClkNotDetectMask_1) |
+		  HWE_MASK(LATriggered));
+
+	for (pidx = 0; pidx < NUM_IB_PORTS; ++pidx) {
+		struct qib_chippport_specific *cp = ppd->cpspec;
+		ppd->link_speed_supported = features & PORT_SPD_CAP;
+		features >>=  PORT_SPD_CAP_SHIFT;
+		if (!ppd->link_speed_supported) {
+			/* single port mode (7340, or configured) */
+			dd->skip_kctxt_mask |= 1 << pidx;
+			if (pidx == 0) {
+				/* Make sure port is disabled. */
+				qib_write_kreg_port(ppd, krp_rcvctrl, 0);
+				qib_write_kreg_port(ppd, krp_ibcctrl_a, 0);
+				ppd[0] = ppd[1];
+				dd->cspec->hwerrmask &= ~(SYM_MASK(HwErrMask,
+						  IBSerdesPClkNotDetectMask_0)
+						  | SYM_MASK(HwErrMask,
+						  SDmaMemReadErrMask_0));
+				dd->cspec->int_enable_mask &= ~(
+				     SYM_MASK(IntMask, SDmaCleanupDoneMask_0) |
+				     SYM_MASK(IntMask, SDmaIdleIntMask_0) |
+				     SYM_MASK(IntMask, SDmaProgressIntMask_0) |
+				     SYM_MASK(IntMask, SDmaIntMask_0) |
+				     SYM_MASK(IntMask, ErrIntMask_0) |
+				     SYM_MASK(IntMask, SendDoneIntMask_0));
+			} else {
+				/* Make sure port is disabled. */
+				qib_write_kreg_port(ppd, krp_rcvctrl, 0);
+				qib_write_kreg_port(ppd, krp_ibcctrl_a, 0);
+				dd->cspec->hwerrmask &= ~(SYM_MASK(HwErrMask,
+						  IBSerdesPClkNotDetectMask_1)
+						  | SYM_MASK(HwErrMask,
+						  SDmaMemReadErrMask_1));
+				dd->cspec->int_enable_mask &= ~(
+				     SYM_MASK(IntMask, SDmaCleanupDoneMask_1) |
+				     SYM_MASK(IntMask, SDmaIdleIntMask_1) |
+				     SYM_MASK(IntMask, SDmaProgressIntMask_1) |
+				     SYM_MASK(IntMask, SDmaIntMask_1) |
+				     SYM_MASK(IntMask, ErrIntMask_1) |
+				     SYM_MASK(IntMask, SendDoneIntMask_1));
+			}
+			continue;
+		}
+
+		dd->num_pports++;
+		qib_init_pportdata(ppd, dd, pidx, dd->num_pports);
+
+		ppd->link_width_supported = IB_WIDTH_1X | IB_WIDTH_4X;
+		ppd->link_width_enabled = IB_WIDTH_4X;
+		ppd->link_speed_enabled = ppd->link_speed_supported;
+		/*
+		 * Set the initial values to reasonable default, will be set
+		 * for real when link is up.
+		 */
+		ppd->link_width_active = IB_WIDTH_4X;
+		ppd->link_speed_active = QIB_IB_SDR;
+		ppd->delay_mult = ib_rate_to_delay[IB_RATE_10_GBPS];
+		switch (qib_num_cfg_vls) {
+		case 1:
+			ppd->vls_supported = IB_VL_VL0;
+			break;
+		case 2:
+			ppd->vls_supported = IB_VL_VL0_1;
+			break;
+		default:
+			qib_devinfo(dd->pcidev,
+				    "Invalid num_vls %u, using 4 VLs\n",
+				    qib_num_cfg_vls);
+			qib_num_cfg_vls = 4;
+			/* fall through */
+		case 4:
+			ppd->vls_supported = IB_VL_VL0_3;
+			break;
+		case 8:
+			if (mtu <= 2048)
+				ppd->vls_supported = IB_VL_VL0_7;
+			else {
+				qib_devinfo(dd->pcidev,
+					    "Invalid num_vls %u for MTU %d "
+					    ", using 4 VLs\n",
+					    qib_num_cfg_vls, mtu);
+				ppd->vls_supported = IB_VL_VL0_3;
+				qib_num_cfg_vls = 4;
+			}
+			break;
+		}
+		ppd->vls_operational = ppd->vls_supported;
+
+		init_waitqueue_head(&cp->autoneg_wait);
+		INIT_DELAYED_WORK(&cp->autoneg_work,
+				  autoneg_7322_work);
+		if (ppd->dd->cspec->r1)
+			INIT_DELAYED_WORK(&cp->ipg_work, ipg_7322_work);
+
+		/*
+		 * For Mez and similar cards, no qsfp info, so do
+		 * the "cable info" setup here.  Can be overridden
+		 * in adapter-specific routines.
+		 */
+		if (!(ppd->dd->flags & QIB_HAS_QSFP)) {
+			if (!IS_QMH(ppd->dd) && !IS_QME(ppd->dd))
+				qib_devinfo(ppd->dd->pcidev, "IB%u:%u: "
+					    "Unknown mezzanine card type\n",
+					    dd->unit, ppd->port);
+			cp->h1_val = IS_QMH(dd) ? H1_FORCE_QMH : H1_FORCE_QME;
+			/*
+			 * Choose center value as default tx serdes setting
+			 * until changed through module parameter.
+			 */
+			ppd->cpspec->no_eep = IS_QMH(dd) ?
+				TXDDS_TABLE_SZ + 2 : TXDDS_TABLE_SZ + 4;
+		} else
+			cp->h1_val = H1_FORCE_VAL;
+
+		/* Avoid writes to chip for mini_init */
+		if (!qib_mini_init)
+			write_7322_init_portregs(ppd);
+
+		init_timer(&cp->chase_timer);
+		cp->chase_timer.function = reenable_chase;
+		cp->chase_timer.data = (unsigned long)ppd;
+
+		ppd++;
+	}
+
+	dd->rcvhdrentsize = QIB_RCVHDR_ENTSIZE;
+	dd->rcvhdrsize = QIB_DFLT_RCVHDRSIZE;
+	dd->rhf_offset = dd->rcvhdrentsize - sizeof(u64) / sizeof(u32);
+
+	/* we always allocate at least 2048 bytes for eager buffers */
+	dd->rcvegrbufsize = max(mtu, 2048);
+
+	qib_7322_tidtemplate(dd);
+
+	/*
+	 * We can request a receive interrupt for 1 or
+	 * more packets from current offset.
+	 */
+	dd->rhdrhead_intr_off =
+		(u64) rcv_int_count << IBA7322_HDRHEAD_PKTINT_SHIFT;
+
+	/* setup the stats timer; the add_timer is done at end of init */
+	init_timer(&dd->stats_timer);
+	dd->stats_timer.function = qib_get_7322_faststats;
+	dd->stats_timer.data = (unsigned long) dd;
+
+	dd->ureg_align = 0x10000;  /* 64KB alignment */
+
+	dd->piosize2kmax_dwords = dd->piosize2k >> 2;
+
+	qib_7322_config_ctxts(dd);
+	qib_set_ctxtcnt(dd);
+
+	if (qib_wc_pat) {
+		ret = init_chip_wc_pat(dd, NUM_VL15_BUFS * dd->align4k);
+		if (ret)
+			goto bail;
+	}
+	qib_7322_set_baseaddrs(dd); /* set chip access pointers now */
+
+	ret = 0;
+	if (qib_mini_init)
+		goto bail;
+	if (!dd->num_pports) {
+		qib_dev_err(dd, "No ports enabled, giving up initialization\n");
+		goto bail; /* no error, so can still figure out why err */
+	}
+
+	write_7322_initregs(dd);
+	ret = qib_create_ctxts(dd);
+	init_7322_cntrnames(dd);
+
+	updthresh = 8U; /* update threshold */
+
+	/* use all of 4KB buffers for the kernel SDMA, zero if !SDMA.
+	 * reserve the update threshold amount for other kernel use, such
+	 * as sending SMI, MAD, and ACKs, or 3, whichever is greater,
+	 * unless we aren't enabling SDMA, in which case we want to use
+	 * all the 4k bufs for the kernel.
+	 * if this was less than the update threshold, we could wait
+	 * a long time for an update.  Coded this way because we
+	 * sometimes change the update threshold for various reasons,
+	 * and we want this to remain robust.
+	 */
+	if (dd->flags & QIB_HAS_SEND_DMA) {
+		dd->cspec->sdmabufcnt = dd->piobcnt4k;
+		sbufs = updthresh > 3 ? updthresh : 3;
+	} else {
+		dd->cspec->sdmabufcnt = 0;
+		sbufs = dd->piobcnt4k;
+	}
+	dd->cspec->lastbuf_for_pio = dd->piobcnt2k + dd->piobcnt4k -
+		dd->cspec->sdmabufcnt;
+	dd->lastctxt_piobuf = dd->cspec->lastbuf_for_pio - sbufs;
+	dd->cspec->lastbuf_for_pio--; /* range is <= , not < */
+	dd->pbufsctxt = (dd->cfgctxts > dd->first_user_ctxt) ?
+		dd->lastctxt_piobuf / (dd->cfgctxts - dd->first_user_ctxt) : 0;
+
+	/*
+	 * If we have 16 user contexts, we will have 7 sbufs
+	 * per context, so reduce the update threshold to match.  We
+	 * want to update before we actually run out, at low pbufs/ctxt
+	 * so give ourselves some margin.
+	 */
+	if (dd->pbufsctxt >= 2 && dd->pbufsctxt - 2 < updthresh)
+		updthresh = dd->pbufsctxt - 2;
+	dd->cspec->updthresh_dflt = updthresh;
+	dd->cspec->updthresh = updthresh;
+
+	/* before full enable, no interrupts, no locking needed */
+	dd->sendctrl |= ((updthresh & SYM_RMASK(SendCtrl, AvailUpdThld))
+			     << SYM_LSB(SendCtrl, AvailUpdThld)) |
+			SYM_MASK(SendCtrl, SendBufAvailPad64Byte);
+
+	dd->psxmitwait_supported = 1;
+	dd->psxmitwait_check_rate = QIB_7322_PSXMITWAIT_CHECK_RATE;
+bail:
+	if (!dd->ctxtcnt)
+		dd->ctxtcnt = 1; /* for other initialization code */
+
+	return ret;
+}
+
+static u32 __iomem *qib_7322_getsendbuf(struct qib_pportdata *ppd, u64 pbc,
+					u32 *pbufnum)
+{
+	u32 first, last, plen = pbc & QIB_PBC_LENGTH_MASK;
+	struct qib_devdata *dd = ppd->dd;
+
+	/* last is same for 2k and 4k, because we use 4k if all 2k busy */
+	if (pbc & PBC_7322_VL15_SEND) {
+		first = dd->piobcnt2k + dd->piobcnt4k + ppd->hw_pidx;
+		last = first;
+	} else {
+		if ((plen + 1) > dd->piosize2kmax_dwords)
+			first = dd->piobcnt2k;
+		else
+			first = 0;
+		last = dd->cspec->lastbuf_for_pio;
+	}
+	return qib_getsendbuf_range(dd, pbufnum, first, last);
+}
+
+static void qib_set_cntr_7322_sample(struct qib_pportdata *ppd, u32 intv,
+				     u32 start)
+{
+	qib_write_kreg_port(ppd, krp_psinterval, intv);
+	qib_write_kreg_port(ppd, krp_psstart, start);
+}
+
+/*
+ * Must be called with sdma_lock held, or before init finished.
+ */
+static void qib_sdma_set_7322_desc_cnt(struct qib_pportdata *ppd, unsigned cnt)
+{
+	qib_write_kreg_port(ppd, krp_senddmadesccnt, cnt);
+}
+
+static struct sdma_set_state_action sdma_7322_action_table[] = {
+	[qib_sdma_state_s00_hw_down] = {
+		.go_s99_running_tofalse = 1,
+		.op_enable = 0,
+		.op_intenable = 0,
+		.op_halt = 0,
+		.op_drain = 0,
+	},
+	[qib_sdma_state_s10_hw_start_up_wait] = {
+		.op_enable = 0,
+		.op_intenable = 1,
+		.op_halt = 1,
+		.op_drain = 0,
+	},
+	[qib_sdma_state_s20_idle] = {
+		.op_enable = 1,
+		.op_intenable = 1,
+		.op_halt = 1,
+		.op_drain = 0,
+	},
+	[qib_sdma_state_s30_sw_clean_up_wait] = {
+		.op_enable = 0,
+		.op_intenable = 1,
+		.op_halt = 1,
+		.op_drain = 0,
+	},
+	[qib_sdma_state_s40_hw_clean_up_wait] = {
+		.op_enable = 1,
+		.op_intenable = 1,
+		.op_halt = 1,
+		.op_drain = 0,
+	},
+	[qib_sdma_state_s50_hw_halt_wait] = {
+		.op_enable = 1,
+		.op_intenable = 1,
+		.op_halt = 1,
+		.op_drain = 1,
+	},
+	[qib_sdma_state_s99_running] = {
+		.op_enable = 1,
+		.op_intenable = 1,
+		.op_halt = 0,
+		.op_drain = 0,
+		.go_s99_running_totrue = 1,
+	},
+};
+
+static void qib_7322_sdma_init_early(struct qib_pportdata *ppd)
+{
+	ppd->sdma_state.set_state_action = sdma_7322_action_table;
+}
+
+static int init_sdma_7322_regs(struct qib_pportdata *ppd)
+{
+	struct qib_devdata *dd = ppd->dd;
+	unsigned lastbuf, erstbuf;
+	u64 senddmabufmask[3] = { 0 };
+	int n, ret = 0;
+
+	qib_write_kreg_port(ppd, krp_senddmabase, ppd->sdma_descq_phys);
+	qib_sdma_7322_setlengen(ppd);
+	qib_sdma_update_7322_tail(ppd, 0); /* Set SendDmaTail */
+	qib_write_kreg_port(ppd, krp_senddmareloadcnt, sdma_idle_cnt);
+	qib_write_kreg_port(ppd, krp_senddmadesccnt, 0);
+	qib_write_kreg_port(ppd, krp_senddmaheadaddr, ppd->sdma_head_phys);
+
+	if (dd->num_pports)
+		n = dd->cspec->sdmabufcnt / dd->num_pports; /* no remainder */
+	else
+		n = dd->cspec->sdmabufcnt; /* failsafe for init */
+	erstbuf = (dd->piobcnt2k + dd->piobcnt4k) -
+		((dd->num_pports == 1 || ppd->port == 2) ? n :
+		dd->cspec->sdmabufcnt);
+	lastbuf = erstbuf + n;
+
+	ppd->sdma_state.first_sendbuf = erstbuf;
+	ppd->sdma_state.last_sendbuf = lastbuf;
+	for (; erstbuf < lastbuf; ++erstbuf) {
+		unsigned word = erstbuf / BITS_PER_LONG;
+		unsigned bit = erstbuf & (BITS_PER_LONG - 1);
+
+		BUG_ON(word >= 3);
+		senddmabufmask[word] |= 1ULL << bit;
+	}
+	qib_write_kreg_port(ppd, krp_senddmabufmask0, senddmabufmask[0]);
+	qib_write_kreg_port(ppd, krp_senddmabufmask1, senddmabufmask[1]);
+	qib_write_kreg_port(ppd, krp_senddmabufmask2, senddmabufmask[2]);
+	return ret;
+}
+
+/* sdma_lock must be held */
+static u16 qib_sdma_7322_gethead(struct qib_pportdata *ppd)
+{
+	struct qib_devdata *dd = ppd->dd;
+	int sane;
+	int use_dmahead;
+	u16 swhead;
+	u16 swtail;
+	u16 cnt;
+	u16 hwhead;
+
+	use_dmahead = __qib_sdma_running(ppd) &&
+		(dd->flags & QIB_HAS_SDMA_TIMEOUT);
+retry:
+	hwhead = use_dmahead ?
+		(u16) le64_to_cpu(*ppd->sdma_head_dma) :
+		(u16) qib_read_kreg_port(ppd, krp_senddmahead);
+
+	swhead = ppd->sdma_descq_head;
+	swtail = ppd->sdma_descq_tail;
+	cnt = ppd->sdma_descq_cnt;
+
+	if (swhead < swtail)
+		/* not wrapped */
+		sane = (hwhead >= swhead) & (hwhead <= swtail);
+	else if (swhead > swtail)
+		/* wrapped around */
+		sane = ((hwhead >= swhead) && (hwhead < cnt)) ||
+			(hwhead <= swtail);
+	else
+		/* empty */
+		sane = (hwhead == swhead);
+
+	if (unlikely(!sane)) {
+		if (use_dmahead) {
+			/* try one more time, directly from the register */
+			use_dmahead = 0;
+			goto retry;
+		}
+		/* proceed as if no progress */
+		hwhead = swhead;
+	}
+
+	return hwhead;
+}
+
+static int qib_sdma_7322_busy(struct qib_pportdata *ppd)
+{
+	u64 hwstatus = qib_read_kreg_port(ppd, krp_senddmastatus);
+
+	return (hwstatus & SYM_MASK(SendDmaStatus_0, ScoreBoardDrainInProg)) ||
+	       (hwstatus & SYM_MASK(SendDmaStatus_0, HaltInProg)) ||
+	       !(hwstatus & SYM_MASK(SendDmaStatus_0, InternalSDmaHalt)) ||
+	       !(hwstatus & SYM_MASK(SendDmaStatus_0, ScbEmpty));
+}
+
+/*
+ * Compute the amount of delay before sending the next packet if the
+ * port's send rate differs from the static rate set for the QP.
+ * The delay affects the next packet and the amount of the delay is
+ * based on the length of the this packet.
+ */
+static u32 qib_7322_setpbc_control(struct qib_pportdata *ppd, u32 plen,
+				   u8 srate, u8 vl)
+{
+	u8 snd_mult = ppd->delay_mult;
+	u8 rcv_mult = ib_rate_to_delay[srate];
+	u32 ret;
+
+	ret = rcv_mult > snd_mult ? ((plen + 1) >> 1) * snd_mult : 0;
+
+	/* Indicate VL15, else set the VL in the control word */
+	if (vl == 15)
+		ret |= PBC_7322_VL15_SEND_CTRL;
+	else
+		ret |= vl << PBC_VL_NUM_LSB;
+	ret |= ((u32)(ppd->hw_pidx)) << PBC_PORT_SEL_LSB;
+
+	return ret;
+}
+
+/*
+ * Enable the per-port VL15 send buffers for use.
+ * They follow the rest of the buffers, without a config parameter.
+ * This was in initregs, but that is done before the shadow
+ * is set up, and this has to be done after the shadow is
+ * set up.
+ */
+static void qib_7322_initvl15_bufs(struct qib_devdata *dd)
+{
+	unsigned vl15bufs;
+
+	vl15bufs = dd->piobcnt2k + dd->piobcnt4k;
+	qib_chg_pioavailkernel(dd, vl15bufs, NUM_VL15_BUFS,
+			       TXCHK_CHG_TYPE_KERN, NULL);
+}
+
+static void qib_7322_init_ctxt(struct qib_ctxtdata *rcd)
+{
+	if (rcd->ctxt < NUM_IB_PORTS) {
+		if (rcd->dd->num_pports > 1) {
+			rcd->rcvegrcnt = KCTXT0_EGRCNT / 2;
+			rcd->rcvegr_tid_base = rcd->ctxt ? rcd->rcvegrcnt : 0;
+		} else {
+			rcd->rcvegrcnt = KCTXT0_EGRCNT;
+			rcd->rcvegr_tid_base = 0;
+		}
+	} else {
+		rcd->rcvegrcnt = rcd->dd->cspec->rcvegrcnt;
+		rcd->rcvegr_tid_base = KCTXT0_EGRCNT +
+			(rcd->ctxt - NUM_IB_PORTS) * rcd->rcvegrcnt;
+	}
+}
+
+#define QTXSLEEPS 5000
+static void qib_7322_txchk_change(struct qib_devdata *dd, u32 start,
+				  u32 len, u32 which, struct qib_ctxtdata *rcd)
+{
+	int i;
+	const int last = start + len - 1;
+	const int lastr = last / BITS_PER_LONG;
+	u32 sleeps = 0;
+	int wait = rcd != NULL;
+	unsigned long flags;
+
+	while (wait) {
+		unsigned long shadow;
+		int cstart, previ = -1;
+
+		/*
+		 * when flipping from kernel to user, we can't change
+		 * the checking type if the buffer is allocated to the
+		 * driver.   It's OK the other direction, because it's
+		 * from close, and we have just disarm'ed all the
+		 * buffers.  All the kernel to kernel changes are also
+		 * OK.
+		 */
+		for (cstart = start; cstart <= last; cstart++) {
+			i = ((2 * cstart) + QLOGIC_IB_SENDPIOAVAIL_BUSY_SHIFT)
+				/ BITS_PER_LONG;
+			if (i != previ) {
+				shadow = (unsigned long)
+					le64_to_cpu(dd->pioavailregs_dma[i]);
+				previ = i;
+			}
+			if (test_bit(((2 * cstart) +
+				      QLOGIC_IB_SENDPIOAVAIL_BUSY_SHIFT)
+				     % BITS_PER_LONG, &shadow))
+				break;
+		}
+
+		if (cstart > last)
+			break;
+
+		if (sleeps == QTXSLEEPS)
+			break;
+		/* make sure we see an updated copy next time around */
+		sendctrl_7322_mod(dd->pport, QIB_SENDCTRL_AVAIL_BLIP);
+		sleeps++;
+		msleep(1);
+	}
+
+	switch (which) {
+	case TXCHK_CHG_TYPE_DIS1:
+		/*
+		 * disable checking on a range; used by diags; just
+		 * one buffer, but still written generically
+		 */
+		for (i = start; i <= last; i++)
+			clear_bit(i, dd->cspec->sendchkenable);
+		break;
+
+	case TXCHK_CHG_TYPE_ENAB1:
+		/*
+		 * (re)enable checking on a range; used by diags; just
+		 * one buffer, but still written generically; read
+		 * scratch to be sure buffer actually triggered, not
+		 * just flushed from processor.
+		 */
+		qib_read_kreg32(dd, kr_scratch);
+		for (i = start; i <= last; i++)
+			set_bit(i, dd->cspec->sendchkenable);
+		break;
+
+	case TXCHK_CHG_TYPE_KERN:
+		/* usable by kernel */
+		for (i = start; i <= last; i++) {
+			set_bit(i, dd->cspec->sendibchk);
+			clear_bit(i, dd->cspec->sendgrhchk);
+		}
+		spin_lock_irqsave(&dd->uctxt_lock, flags);
+		/* see if we need to raise avail update threshold */
+		for (i = dd->first_user_ctxt;
+		     dd->cspec->updthresh != dd->cspec->updthresh_dflt
+		     && i < dd->cfgctxts; i++)
+			if (dd->rcd[i] && dd->rcd[i]->subctxt_cnt &&
+			   ((dd->rcd[i]->piocnt / dd->rcd[i]->subctxt_cnt) - 1)
+			   < dd->cspec->updthresh_dflt)
+				break;
+		spin_unlock_irqrestore(&dd->uctxt_lock, flags);
+		if (i == dd->cfgctxts) {
+			spin_lock_irqsave(&dd->sendctrl_lock, flags);
+			dd->cspec->updthresh = dd->cspec->updthresh_dflt;
+			dd->sendctrl &= ~SYM_MASK(SendCtrl, AvailUpdThld);
+			dd->sendctrl |= (dd->cspec->updthresh &
+					 SYM_RMASK(SendCtrl, AvailUpdThld)) <<
+					   SYM_LSB(SendCtrl, AvailUpdThld);
+			spin_unlock_irqrestore(&dd->sendctrl_lock, flags);
+			sendctrl_7322_mod(dd->pport, QIB_SENDCTRL_AVAIL_BLIP);
+		}
+		break;
+
+	case TXCHK_CHG_TYPE_USER:
+		/* for user process */
+		for (i = start; i <= last; i++) {
+			clear_bit(i, dd->cspec->sendibchk);
+			set_bit(i, dd->cspec->sendgrhchk);
+		}
+		spin_lock_irqsave(&dd->sendctrl_lock, flags);
+		if (rcd && rcd->subctxt_cnt && ((rcd->piocnt
+			/ rcd->subctxt_cnt) - 1) < dd->cspec->updthresh) {
+			dd->cspec->updthresh = (rcd->piocnt /
+						rcd->subctxt_cnt) - 1;
+			dd->sendctrl &= ~SYM_MASK(SendCtrl, AvailUpdThld);
+			dd->sendctrl |= (dd->cspec->updthresh &
+					SYM_RMASK(SendCtrl, AvailUpdThld))
+					<< SYM_LSB(SendCtrl, AvailUpdThld);
+			spin_unlock_irqrestore(&dd->sendctrl_lock, flags);
+			sendctrl_7322_mod(dd->pport, QIB_SENDCTRL_AVAIL_BLIP);
+		} else
+			spin_unlock_irqrestore(&dd->sendctrl_lock, flags);
+		break;
+
+	default:
+		break;
+	}
+
+	for (i = start / BITS_PER_LONG; which >= 2 && i <= lastr; ++i)
+		qib_write_kreg(dd, kr_sendcheckmask + i,
+			       dd->cspec->sendchkenable[i]);
+
+	for (i = start / BITS_PER_LONG; which < 2 && i <= lastr; ++i) {
+		qib_write_kreg(dd, kr_sendgrhcheckmask + i,
+			       dd->cspec->sendgrhchk[i]);
+		qib_write_kreg(dd, kr_sendibpktmask + i,
+			       dd->cspec->sendibchk[i]);
+	}
+
+	/*
+	 * Be sure whatever we did was seen by the chip and acted upon,
+	 * before we return.  Mostly important for which >= 2.
+	 */
+	qib_read_kreg32(dd, kr_scratch);
+}
+
+
+/* useful for trigger analyzers, etc. */
+static void writescratch(struct qib_devdata *dd, u32 val)
+{
+	qib_write_kreg(dd, kr_scratch, val);
+}
+
+/* Dummy for now, use chip regs soon */
+static int qib_7322_tempsense_rd(struct qib_devdata *dd, int regnum)
+{
+	return -ENXIO;
+}
+
+/**
+ * qib_init_iba7322_funcs - set up the chip-specific function pointers
+ * @dev: the pci_dev for qlogic_ib device
+ * @ent: pci_device_id struct for this dev
+ *
+ * Also allocates, inits, and returns the devdata struct for this
+ * device instance
+ *
+ * This is global, and is called directly at init to set up the
+ * chip-specific function pointers for later use.
+ */
+struct qib_devdata *qib_init_iba7322_funcs(struct pci_dev *pdev,
+					   const struct pci_device_id *ent)
+{
+	struct qib_devdata *dd;
+	int ret, i;
+	u32 tabsize, actual_cnt = 0;
+
+	dd = qib_alloc_devdata(pdev,
+		NUM_IB_PORTS * sizeof(struct qib_pportdata) +
+		sizeof(struct qib_chip_specific) +
+		NUM_IB_PORTS * sizeof(struct qib_chippport_specific));
+	if (IS_ERR(dd))
+		goto bail;
+
+	dd->f_bringup_serdes    = qib_7322_bringup_serdes;
+	dd->f_cleanup           = qib_setup_7322_cleanup;
+	dd->f_clear_tids        = qib_7322_clear_tids;
+	dd->f_free_irq          = qib_7322_free_irq;
+	dd->f_get_base_info     = qib_7322_get_base_info;
+	dd->f_get_msgheader     = qib_7322_get_msgheader;
+	dd->f_getsendbuf        = qib_7322_getsendbuf;
+	dd->f_gpio_mod          = gpio_7322_mod;
+	dd->f_eeprom_wen        = qib_7322_eeprom_wen;
+	dd->f_hdrqempty         = qib_7322_hdrqempty;
+	dd->f_ib_updown         = qib_7322_ib_updown;
+	dd->f_init_ctxt         = qib_7322_init_ctxt;
+	dd->f_initvl15_bufs     = qib_7322_initvl15_bufs;
+	dd->f_intr_fallback     = qib_7322_intr_fallback;
+	dd->f_late_initreg      = qib_late_7322_initreg;
+	dd->f_setpbc_control    = qib_7322_setpbc_control;
+	dd->f_portcntr          = qib_portcntr_7322;
+	dd->f_put_tid           = qib_7322_put_tid;
+	dd->f_quiet_serdes      = qib_7322_mini_quiet_serdes;
+	dd->f_rcvctrl           = rcvctrl_7322_mod;
+	dd->f_read_cntrs        = qib_read_7322cntrs;
+	dd->f_read_portcntrs    = qib_read_7322portcntrs;
+	dd->f_reset             = qib_do_7322_reset;
+	dd->f_init_sdma_regs    = init_sdma_7322_regs;
+	dd->f_sdma_busy         = qib_sdma_7322_busy;
+	dd->f_sdma_gethead      = qib_sdma_7322_gethead;
+	dd->f_sdma_sendctrl     = qib_7322_sdma_sendctrl;
+	dd->f_sdma_set_desc_cnt = qib_sdma_set_7322_desc_cnt;
+	dd->f_sdma_update_tail  = qib_sdma_update_7322_tail;
+	dd->f_sendctrl          = sendctrl_7322_mod;
+	dd->f_set_armlaunch     = qib_set_7322_armlaunch;
+	dd->f_set_cntr_sample   = qib_set_cntr_7322_sample;
+	dd->f_iblink_state      = qib_7322_iblink_state;
+	dd->f_ibphys_portstate  = qib_7322_phys_portstate;
+	dd->f_get_ib_cfg        = qib_7322_get_ib_cfg;
+	dd->f_set_ib_cfg        = qib_7322_set_ib_cfg;
+	dd->f_set_ib_loopback   = qib_7322_set_loopback;
+	dd->f_get_ib_table      = qib_7322_get_ib_table;
+	dd->f_set_ib_table      = qib_7322_set_ib_table;
+	dd->f_set_intr_state    = qib_7322_set_intr_state;
+	dd->f_setextled         = qib_setup_7322_setextled;
+	dd->f_txchk_change      = qib_7322_txchk_change;
+	dd->f_update_usrhead    = qib_update_7322_usrhead;
+	dd->f_wantpiobuf_intr   = qib_wantpiobuf_7322_intr;
+	dd->f_xgxs_reset        = qib_7322_mini_pcs_reset;
+	dd->f_sdma_hw_clean_up  = qib_7322_sdma_hw_clean_up;
+	dd->f_sdma_hw_start_up  = qib_7322_sdma_hw_start_up;
+	dd->f_sdma_init_early   = qib_7322_sdma_init_early;
+	dd->f_writescratch      = writescratch;
+	dd->f_tempsense_rd	= qib_7322_tempsense_rd;
+	/*
+	 * Do remaining PCIe setup and save PCIe values in dd.
+	 * Any error printing is already done by the init code.
+	 * On return, we have the chip mapped, but chip registers
+	 * are not set up until start of qib_init_7322_variables.
+	 */
+	ret = qib_pcie_ddinit(dd, pdev, ent);
+	if (ret < 0)
+		goto bail_free;
+
+	/* initialize chip-specific variables */
+	ret = qib_init_7322_variables(dd);
+	if (ret)
+		goto bail_cleanup;
+
+	if (qib_mini_init || !dd->num_pports)
+		goto bail;
+
+	/*
+	 * Determine number of vectors we want; depends on port count
+	 * and number of configured kernel receive queues actually used.
+	 * Should also depend on whether sdma is enabled or not, but
+	 * that's such a rare testing case it's not worth worrying about.
+	 */
+	tabsize = dd->first_user_ctxt + ARRAY_SIZE(irq_table);
+	for (i = 0; i < tabsize; i++)
+		if ((i < ARRAY_SIZE(irq_table) &&
+		     irq_table[i].port <= dd->num_pports) ||
+		    (i >= ARRAY_SIZE(irq_table) &&
+		     dd->rcd[i - ARRAY_SIZE(irq_table)]))
+			actual_cnt++;
+	tabsize = actual_cnt;
+	dd->cspec->msix_entries = kmalloc(tabsize *
+			sizeof(struct msix_entry), GFP_KERNEL);
+	dd->cspec->msix_arg = kmalloc(tabsize *
+			sizeof(void *), GFP_KERNEL);
+	if (!dd->cspec->msix_entries || !dd->cspec->msix_arg) {
+		qib_dev_err(dd, "No memory for MSIx table\n");
+		tabsize = 0;
+	}
+	for (i = 0; i < tabsize; i++)
+		dd->cspec->msix_entries[i].entry = i;
+
+	if (qib_pcie_params(dd, 8, &tabsize, dd->cspec->msix_entries))
+		qib_dev_err(dd, "Failed to setup PCIe or interrupts; "
+			    "continuing anyway\n");
+	/* may be less than we wanted, if not enough available */
+	dd->cspec->num_msix_entries = tabsize;
+
+	/* setup interrupt handler */
+	qib_setup_7322_interrupt(dd, 1);
+
+	/* clear diagctrl register, in case diags were running and crashed */
+	qib_write_kreg(dd, kr_hwdiagctrl, 0);
+
+	goto bail;
+
+bail_cleanup:
+	qib_pcie_ddcleanup(dd);
+bail_free:
+	qib_free_devdata(dd);
+	dd = ERR_PTR(ret);
+bail:
+	return dd;
+}
+
+/*
+ * Set the table entry at the specified index from the table specifed.
+ * There are 3 * TXDDS_TABLE_SZ entries in all per port, with the first
+ * TXDDS_TABLE_SZ for SDR, the next for DDR, and the last for QDR.
+ * 'idx' below addresses the correct entry, while its 4 LSBs select the
+ * corresponding entry (one of TXDDS_TABLE_SZ) from the selected table.
+ */
+#define DDS_ENT_AMP_LSB 14
+#define DDS_ENT_MAIN_LSB 9
+#define DDS_ENT_POST_LSB 5
+#define DDS_ENT_PRE_XTRA_LSB 3
+#define DDS_ENT_PRE_LSB 0
+
+/*
+ * Set one entry in the TxDDS table for spec'd port
+ * ridx picks one of the entries, while tp points
+ * to the appropriate table entry.
+ */
+static void set_txdds(struct qib_pportdata *ppd, int ridx,
+		      const struct txdds_ent *tp)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u32 pack_ent;
+	int regidx;
+
+	/* Get correct offset in chip-space, and in source table */
+	regidx = KREG_IBPORT_IDX(IBSD_DDS_MAP_TABLE) + ridx;
+	/*
+	 * We do not use qib_write_kreg_port() because it was intended
+	 * only for registers in the lower "port specific" pages.
+	 * So do index calculation  by hand.
+	 */
+	if (ppd->hw_pidx)
+		regidx += (dd->palign / sizeof(u64));
+
+	pack_ent = tp->amp << DDS_ENT_AMP_LSB;
+	pack_ent |= tp->main << DDS_ENT_MAIN_LSB;
+	pack_ent |= tp->pre << DDS_ENT_PRE_LSB;
+	pack_ent |= tp->post << DDS_ENT_POST_LSB;
+	qib_write_kreg(dd, regidx, pack_ent);
+	/* Prevent back-to-back writes by hitting scratch */
+	qib_write_kreg(ppd->dd, kr_scratch, 0);
+}
+
+static const struct vendor_txdds_ent vendor_txdds[] = {
+	{ /* Amphenol 1m 30awg NoEq */
+		{ 0x41, 0x50, 0x48 }, "584470002       ",
+		{ 10,  0,  0,  5 }, { 10,  0,  0,  9 }, {  7,  1,  0, 13 },
+	},
+	{ /* Amphenol 3m 28awg NoEq */
+		{ 0x41, 0x50, 0x48 }, "584470004       ",
+		{  0,  0,  0,  8 }, {  0,  0,  0, 11 }, {  0,  1,  7, 15 },
+	},
+	{ /* Finisar 3m OM2 Optical */
+		{ 0x00, 0x90, 0x65 }, "FCBG410QB1C03-QL",
+		{  0,  0,  0,  3 }, {  0,  0,  0,  4 }, {  0,  0,  0, 13 },
+	},
+	{ /* Finisar 30m OM2 Optical */
+		{ 0x00, 0x90, 0x65 }, "FCBG410QB1C30-QL",
+		{  0,  0,  0,  1 }, {  0,  0,  0,  5 }, {  0,  0,  0, 11 },
+	},
+	{ /* Finisar Default OM2 Optical */
+		{ 0x00, 0x90, 0x65 }, NULL,
+		{  0,  0,  0,  2 }, {  0,  0,  0,  5 }, {  0,  0,  0, 12 },
+	},
+	{ /* Gore 1m 30awg NoEq */
+		{ 0x00, 0x21, 0x77 }, "QSN3300-1       ",
+		{  0,  0,  0,  6 }, {  0,  0,  0,  9 }, {  0,  1,  0, 15 },
+	},
+	{ /* Gore 2m 30awg NoEq */
+		{ 0x00, 0x21, 0x77 }, "QSN3300-2       ",
+		{  0,  0,  0,  8 }, {  0,  0,  0, 10 }, {  0,  1,  7, 15 },
+	},
+	{ /* Gore 1m 28awg NoEq */
+		{ 0x00, 0x21, 0x77 }, "QSN3800-1       ",
+		{  0,  0,  0,  6 }, {  0,  0,  0,  8 }, {  0,  1,  0, 15 },
+	},
+	{ /* Gore 3m 28awg NoEq */
+		{ 0x00, 0x21, 0x77 }, "QSN3800-3       ",
+		{  0,  0,  0,  9 }, {  0,  0,  0, 13 }, {  0,  1,  7, 15 },
+	},
+	{ /* Gore 5m 24awg Eq */
+		{ 0x00, 0x21, 0x77 }, "QSN7000-5       ",
+		{  0,  0,  0,  7 }, {  0,  0,  0,  9 }, {  0,  1,  3, 15 },
+	},
+	{ /* Gore 7m 24awg Eq */
+		{ 0x00, 0x21, 0x77 }, "QSN7000-7       ",
+		{  0,  0,  0,  9 }, {  0,  0,  0, 11 }, {  0,  2,  6, 15 },
+	},
+	{ /* Gore 5m 26awg Eq */
+		{ 0x00, 0x21, 0x77 }, "QSN7600-5       ",
+		{  0,  0,  0,  8 }, {  0,  0,  0, 11 }, {  0,  1,  9, 13 },
+	},
+	{ /* Gore 7m 26awg Eq */
+		{ 0x00, 0x21, 0x77 }, "QSN7600-7       ",
+		{  0,  0,  0,  8 }, {  0,  0,  0, 11 }, {  10,  1,  8, 15 },
+	},
+	{ /* Intersil 12m 24awg Active */
+		{ 0x00, 0x30, 0xB4 }, "QLX4000CQSFP1224",
+		{  0,  0,  0,  2 }, {  0,  0,  0,  5 }, {  0,  3,  0,  9 },
+	},
+	{ /* Intersil 10m 28awg Active */
+		{ 0x00, 0x30, 0xB4 }, "QLX4000CQSFP1028",
+		{  0,  0,  0,  6 }, {  0,  0,  0,  4 }, {  0,  2,  0,  2 },
+	},
+	{ /* Intersil 7m 30awg Active */
+		{ 0x00, 0x30, 0xB4 }, "QLX4000CQSFP0730",
+		{  0,  0,  0,  6 }, {  0,  0,  0,  4 }, {  0,  1,  0,  3 },
+	},
+	{ /* Intersil 5m 32awg Active */
+		{ 0x00, 0x30, 0xB4 }, "QLX4000CQSFP0532",
+		{  0,  0,  0,  6 }, {  0,  0,  0,  6 }, {  0,  2,  0,  8 },
+	},
+	{ /* Intersil Default Active */
+		{ 0x00, 0x30, 0xB4 }, NULL,
+		{  0,  0,  0,  6 }, {  0,  0,  0,  5 }, {  0,  2,  0,  5 },
+	},
+	{ /* Luxtera 20m Active Optical */
+		{ 0x00, 0x25, 0x63 }, NULL,
+		{  0,  0,  0,  5 }, {  0,  0,  0,  8 }, {  0,  2,  0,  12 },
+	},
+	{ /* Molex 1M Cu loopback */
+		{ 0x00, 0x09, 0x3A }, "74763-0025      ",
+		{  2,  2,  6, 15 }, {  2,  2,  6, 15 }, {  2,  2,  6, 15 },
+	},
+	{ /* Molex 2m 28awg NoEq */
+		{ 0x00, 0x09, 0x3A }, "74757-2201      ",
+		{  0,  0,  0,  6 }, {  0,  0,  0,  9 }, {  0,  1,  1, 15 },
+	},
+};
+
+static const struct txdds_ent txdds_sdr[TXDDS_TABLE_SZ] = {
+	/* amp, pre, main, post */
+	{  2, 2, 15,  6 },	/* Loopback */
+	{  0, 0,  0,  1 },	/*  2 dB */
+	{  0, 0,  0,  2 },	/*  3 dB */
+	{  0, 0,  0,  3 },	/*  4 dB */
+	{  0, 0,  0,  4 },	/*  5 dB */
+	{  0, 0,  0,  5 },	/*  6 dB */
+	{  0, 0,  0,  6 },	/*  7 dB */
+	{  0, 0,  0,  7 },	/*  8 dB */
+	{  0, 0,  0,  8 },	/*  9 dB */
+	{  0, 0,  0,  9 },	/* 10 dB */
+	{  0, 0,  0, 10 },	/* 11 dB */
+	{  0, 0,  0, 11 },	/* 12 dB */
+	{  0, 0,  0, 12 },	/* 13 dB */
+	{  0, 0,  0, 13 },	/* 14 dB */
+	{  0, 0,  0, 14 },	/* 15 dB */
+	{  0, 0,  0, 15 },	/* 16 dB */
+};
+
+static const struct txdds_ent txdds_ddr[TXDDS_TABLE_SZ] = {
+	/* amp, pre, main, post */
+	{  2, 2, 15,  6 },	/* Loopback */
+	{  0, 0,  0,  8 },	/*  2 dB */
+	{  0, 0,  0,  8 },	/*  3 dB */
+	{  0, 0,  0,  9 },	/*  4 dB */
+	{  0, 0,  0,  9 },	/*  5 dB */
+	{  0, 0,  0, 10 },	/*  6 dB */
+	{  0, 0,  0, 10 },	/*  7 dB */
+	{  0, 0,  0, 11 },	/*  8 dB */
+	{  0, 0,  0, 11 },	/*  9 dB */
+	{  0, 0,  0, 12 },	/* 10 dB */
+	{  0, 0,  0, 12 },	/* 11 dB */
+	{  0, 0,  0, 13 },	/* 12 dB */
+	{  0, 0,  0, 13 },	/* 13 dB */
+	{  0, 0,  0, 14 },	/* 14 dB */
+	{  0, 0,  0, 14 },	/* 15 dB */
+	{  0, 0,  0, 15 },	/* 16 dB */
+};
+
+static const struct txdds_ent txdds_qdr[TXDDS_TABLE_SZ] = {
+	/* amp, pre, main, post */
+	{  2, 2, 15,  6 },	/* Loopback */
+	{  0, 1,  0,  7 },	/*  2 dB (also QMH7342) */
+	{  0, 1,  0,  9 },	/*  3 dB (also QMH7342) */
+	{  0, 1,  0, 11 },	/*  4 dB */
+	{  0, 1,  0, 13 },	/*  5 dB */
+	{  0, 1,  0, 15 },	/*  6 dB */
+	{  0, 1,  3, 15 },	/*  7 dB */
+	{  0, 1,  7, 15 },	/*  8 dB */
+	{  0, 1,  7, 15 },	/*  9 dB */
+	{  0, 1,  8, 15 },	/* 10 dB */
+	{  0, 1,  9, 15 },	/* 11 dB */
+	{  0, 1, 10, 15 },	/* 12 dB */
+	{  0, 2,  6, 15 },	/* 13 dB */
+	{  0, 2,  7, 15 },	/* 14 dB */
+	{  0, 2,  8, 15 },	/* 15 dB */
+	{  0, 2,  9, 15 },	/* 16 dB */
+};
+
+/*
+ * extra entries for use with txselect, for indices >= TXDDS_TABLE_SZ.
+ * These are mostly used for mez cards going through connectors
+ * and backplane traces, but can be used to add other "unusual"
+ * table values as well.
+ */
+static const struct txdds_ent txdds_extra_sdr[TXDDS_EXTRA_SZ] = {
+	/* amp, pre, main, post */
+	{  0, 0, 0,  1 },	/* QMH7342 backplane settings */
+	{  0, 0, 0,  1 },	/* QMH7342 backplane settings */
+	{  0, 0, 0,  2 },	/* QMH7342 backplane settings */
+	{  0, 0, 0,  2 },	/* QMH7342 backplane settings */
+	{  0, 0, 0, 11 },	/* QME7342 backplane settings */
+	{  0, 0, 0, 11 },	/* QME7342 backplane settings */
+	{  0, 0, 0, 11 },	/* QME7342 backplane settings */
+	{  0, 0, 0, 11 },	/* QME7342 backplane settings */
+	{  0, 0, 0, 11 },	/* QME7342 backplane settings */
+	{  0, 0, 0, 11 },	/* QME7342 backplane settings */
+	{  0, 0, 0, 11 },	/* QME7342 backplane settings */
+};
+
+static const struct txdds_ent txdds_extra_ddr[TXDDS_EXTRA_SZ] = {
+	/* amp, pre, main, post */
+	{  0, 0, 0,  7 },	/* QMH7342 backplane settings */
+	{  0, 0, 0,  7 },	/* QMH7342 backplane settings */
+	{  0, 0, 0,  8 },	/* QMH7342 backplane settings */
+	{  0, 0, 0,  8 },	/* QMH7342 backplane settings */
+	{  0, 0, 0, 13 },	/* QME7342 backplane settings */
+	{  0, 0, 0, 13 },	/* QME7342 backplane settings */
+	{  0, 0, 0, 13 },	/* QME7342 backplane settings */
+	{  0, 0, 0, 13 },	/* QME7342 backplane settings */
+	{  0, 0, 0, 13 },	/* QME7342 backplane settings */
+	{  0, 0, 0, 13 },	/* QME7342 backplane settings */
+	{  0, 0, 0, 13 },	/* QME7342 backplane settings */
+};
+
+static const struct txdds_ent txdds_extra_qdr[TXDDS_EXTRA_SZ] = {
+	/* amp, pre, main, post */
+	{  0, 1,  0,  4 },	/* QMH7342 backplane settings */
+	{  0, 1,  0,  5 },	/* QMH7342 backplane settings */
+	{  0, 1,  0,  6 },	/* QMH7342 backplane settings */
+	{  0, 1,  0,  8 },	/* QMH7342 backplane settings */
+	{  0, 1, 12, 10 },	/* QME7342 backplane setting */
+	{  0, 1, 12, 11 },	/* QME7342 backplane setting */
+	{  0, 1, 12, 12 },	/* QME7342 backplane setting */
+	{  0, 1, 12, 14 },	/* QME7342 backplane setting */
+	{  0, 1, 12,  6 },	/* QME7342 backplane setting */
+	{  0, 1, 12,  7 },	/* QME7342 backplane setting */
+	{  0, 1, 12,  8 },	/* QME7342 backplane setting */
+};
+
+static const struct txdds_ent *get_atten_table(const struct txdds_ent *txdds,
+					       unsigned atten)
+{
+	/*
+	 * The attenuation table starts at 2dB for entry 1,
+	 * with entry 0 being the loopback entry.
+	 */
+	if (atten <= 2)
+		atten = 1;
+	else if (atten > TXDDS_TABLE_SZ)
+		atten = TXDDS_TABLE_SZ - 1;
+	else
+		atten--;
+	return txdds + atten;
+}
+
+/*
+ * if override is set, the module parameter txselect has a value
+ * for this specific port, so use it, rather than our normal mechanism.
+ */
+static void find_best_ent(struct qib_pportdata *ppd,
+			  const struct txdds_ent **sdr_dds,
+			  const struct txdds_ent **ddr_dds,
+			  const struct txdds_ent **qdr_dds, int override)
+{
+	struct qib_qsfp_cache *qd = &ppd->cpspec->qsfp_data.cache;
+	int idx;
+
+	/* Search table of known cables */
+	for (idx = 0; !override && idx < ARRAY_SIZE(vendor_txdds); ++idx) {
+		const struct vendor_txdds_ent *v = vendor_txdds + idx;
+
+		if (!memcmp(v->oui, qd->oui, QSFP_VOUI_LEN) &&
+		    (!v->partnum ||
+		     !memcmp(v->partnum, qd->partnum, QSFP_PN_LEN))) {
+			*sdr_dds = &v->sdr;
+			*ddr_dds = &v->ddr;
+			*qdr_dds = &v->qdr;
+			return;
+		}
+	}
+
+	/* Lookup serdes setting by cable type and attenuation */
+	if (!override && QSFP_IS_ACTIVE(qd->tech)) {
+		*sdr_dds = txdds_sdr + ppd->dd->board_atten;
+		*ddr_dds = txdds_ddr + ppd->dd->board_atten;
+		*qdr_dds = txdds_qdr + ppd->dd->board_atten;
+		return;
+	}
+
+	if (!override && QSFP_HAS_ATTEN(qd->tech) && (qd->atten[0] ||
+						      qd->atten[1])) {
+		*sdr_dds = get_atten_table(txdds_sdr, qd->atten[0]);
+		*ddr_dds = get_atten_table(txdds_ddr, qd->atten[0]);
+		*qdr_dds = get_atten_table(txdds_qdr, qd->atten[1]);
+		return;
+	} else if (ppd->cpspec->no_eep < TXDDS_TABLE_SZ) {
+		/*
+		 * If we have no (or incomplete) data from the cable
+		 * EEPROM, or no QSFP, or override is set, use the
+		 * module parameter value to index into the attentuation
+		 * table.
+		 */
+		idx = ppd->cpspec->no_eep;
+		*sdr_dds = &txdds_sdr[idx];
+		*ddr_dds = &txdds_ddr[idx];
+		*qdr_dds = &txdds_qdr[idx];
+	} else if (ppd->cpspec->no_eep < (TXDDS_TABLE_SZ + TXDDS_EXTRA_SZ)) {
+		/* similar to above, but index into the "extra" table. */
+		idx = ppd->cpspec->no_eep - TXDDS_TABLE_SZ;
+		*sdr_dds = &txdds_extra_sdr[idx];
+		*ddr_dds = &txdds_extra_ddr[idx];
+		*qdr_dds = &txdds_extra_qdr[idx];
+	} else {
+		/* this shouldn't happen, it's range checked */
+		*sdr_dds = txdds_sdr + qib_long_atten;
+		*ddr_dds = txdds_ddr + qib_long_atten;
+		*qdr_dds = txdds_qdr + qib_long_atten;
+	}
+}
+
+static void init_txdds_table(struct qib_pportdata *ppd, int override)
+{
+	const struct txdds_ent *sdr_dds, *ddr_dds, *qdr_dds;
+	struct txdds_ent *dds;
+	int idx;
+	int single_ent = 0;
+
+	find_best_ent(ppd, &sdr_dds, &ddr_dds, &qdr_dds, override);
+
+	/* for mez cards or override, use the selected value for all entries */
+	if (!(ppd->dd->flags & QIB_HAS_QSFP) || override)
+		single_ent = 1;
+
+	/* Fill in the first entry with the best entry found. */
+	set_txdds(ppd, 0, sdr_dds);
+	set_txdds(ppd, TXDDS_TABLE_SZ, ddr_dds);
+	set_txdds(ppd, 2 * TXDDS_TABLE_SZ, qdr_dds);
+	if (ppd->lflags & (QIBL_LINKINIT | QIBL_LINKARMED |
+		QIBL_LINKACTIVE)) {
+		dds = (struct txdds_ent *)(ppd->link_speed_active ==
+					   QIB_IB_QDR ?  qdr_dds :
+					   (ppd->link_speed_active ==
+					    QIB_IB_DDR ? ddr_dds : sdr_dds));
+		write_tx_serdes_param(ppd, dds);
+	}
+
+	/* Fill in the remaining entries with the default table values. */
+	for (idx = 1; idx < ARRAY_SIZE(txdds_sdr); ++idx) {
+		set_txdds(ppd, idx, single_ent ? sdr_dds : txdds_sdr + idx);
+		set_txdds(ppd, idx + TXDDS_TABLE_SZ,
+			  single_ent ? ddr_dds : txdds_ddr + idx);
+		set_txdds(ppd, idx + 2 * TXDDS_TABLE_SZ,
+			  single_ent ? qdr_dds : txdds_qdr + idx);
+	}
+}
+
+#define KR_AHB_ACC KREG_IDX(ahb_access_ctrl)
+#define KR_AHB_TRANS KREG_IDX(ahb_transaction_reg)
+#define AHB_TRANS_RDY SYM_MASK(ahb_transaction_reg, ahb_rdy)
+#define AHB_ADDR_LSB SYM_LSB(ahb_transaction_reg, ahb_address)
+#define AHB_DATA_LSB SYM_LSB(ahb_transaction_reg, ahb_data)
+#define AHB_WR SYM_MASK(ahb_transaction_reg, write_not_read)
+#define AHB_TRANS_TRIES 10
+
+/*
+ * The chan argument is 0=chan0, 1=chan1, 2=pll, 3=chan2, 4=chan4,
+ * 5=subsystem which is why most calls have "chan + chan >> 1"
+ * for the channel argument.
+ */
+static u32 ahb_mod(struct qib_devdata *dd, int quad, int chan, int addr,
+		    u32 data, u32 mask)
+{
+	u32 rd_data, wr_data, sz_mask;
+	u64 trans, acc, prev_acc;
+	u32 ret = 0xBAD0BAD;
+	int tries;
+
+	prev_acc = qib_read_kreg64(dd, KR_AHB_ACC);
+	/* From this point on, make sure we return access */
+	acc = (quad << 1) | 1;
+	qib_write_kreg(dd, KR_AHB_ACC, acc);
+
+	for (tries = 1; tries < AHB_TRANS_TRIES; ++tries) {
+		trans = qib_read_kreg64(dd, KR_AHB_TRANS);
+		if (trans & AHB_TRANS_RDY)
+			break;
+	}
+	if (tries >= AHB_TRANS_TRIES) {
+		qib_dev_err(dd, "No ahb_rdy in %d tries\n", AHB_TRANS_TRIES);
+		goto bail;
+	}
+
+	/* If mask is not all 1s, we need to read, but different SerDes
+	 * entities have different sizes
+	 */
+	sz_mask = (1UL << ((quad == 1) ? 32 : 16)) - 1;
+	wr_data = data & mask & sz_mask;
+	if ((~mask & sz_mask) != 0) {
+		trans = ((chan << 6) | addr) << (AHB_ADDR_LSB + 1);
+		qib_write_kreg(dd, KR_AHB_TRANS, trans);
+
+		for (tries = 1; tries < AHB_TRANS_TRIES; ++tries) {
+			trans = qib_read_kreg64(dd, KR_AHB_TRANS);
+			if (trans & AHB_TRANS_RDY)
+				break;
+		}
+		if (tries >= AHB_TRANS_TRIES) {
+			qib_dev_err(dd, "No Rd ahb_rdy in %d tries\n",
+				    AHB_TRANS_TRIES);
+			goto bail;
+		}
+		/* Re-read in case host split reads and read data first */
+		trans = qib_read_kreg64(dd, KR_AHB_TRANS);
+		rd_data = (uint32_t)(trans >> AHB_DATA_LSB);
+		wr_data |= (rd_data & ~mask & sz_mask);
+	}
+
+	/* If mask is not zero, we need to write. */
+	if (mask & sz_mask) {
+		trans = ((chan << 6) | addr) << (AHB_ADDR_LSB + 1);
+		trans |= ((uint64_t)wr_data << AHB_DATA_LSB);
+		trans |= AHB_WR;
+		qib_write_kreg(dd, KR_AHB_TRANS, trans);
+
+		for (tries = 1; tries < AHB_TRANS_TRIES; ++tries) {
+			trans = qib_read_kreg64(dd, KR_AHB_TRANS);
+			if (trans & AHB_TRANS_RDY)
+				break;
+		}
+		if (tries >= AHB_TRANS_TRIES) {
+			qib_dev_err(dd, "No Wr ahb_rdy in %d tries\n",
+				    AHB_TRANS_TRIES);
+			goto bail;
+		}
+	}
+	ret = wr_data;
+bail:
+	qib_write_kreg(dd, KR_AHB_ACC, prev_acc);
+	return ret;
+}
+
+static void ibsd_wr_allchans(struct qib_pportdata *ppd, int addr, unsigned data,
+			     unsigned mask)
+{
+	struct qib_devdata *dd = ppd->dd;
+	int chan;
+	u32 rbc;
+
+	for (chan = 0; chan < SERDES_CHANS; ++chan) {
+		ahb_mod(dd, IBSD(ppd->hw_pidx), (chan + (chan >> 1)), addr,
+			data, mask);
+		rbc = ahb_mod(dd, IBSD(ppd->hw_pidx), (chan + (chan >> 1)),
+			      addr, 0, 0);
+	}
+}
+
+static int serdes_7322_init(struct qib_pportdata *ppd)
+{
+	u64 data;
+	u32 le_val;
+
+	/*
+	 * Initialize the Tx DDS tables.  Also done every QSFP event,
+	 * for adapters with QSFP
+	 */
+	init_txdds_table(ppd, 0);
+
+	/* ensure no tx overrides from earlier driver loads */
+	qib_write_kreg_port(ppd, krp_tx_deemph_override,
+		SYM_MASK(IBSD_TX_DEEMPHASIS_OVERRIDE_0,
+		reset_tx_deemphasis_override));
+
+	/* Patch some SerDes defaults to "Better for IB" */
+	/* Timing Loop Bandwidth: cdr_timing[11:9] = 0 */
+	ibsd_wr_allchans(ppd, 2, 0, BMASK(11, 9));
+
+	/* Termination: rxtermctrl_r2d addr 11 bits [12:11] = 1 */
+	ibsd_wr_allchans(ppd, 11, (1 << 11), BMASK(12, 11));
+	/* Enable LE2: rxle2en_r2a addr 13 bit [6] = 1 */
+	ibsd_wr_allchans(ppd, 13, (1 << 6), (1 << 6));
+
+	/* May be overridden in qsfp_7322_event */
+	le_val = IS_QME(ppd->dd) ? LE2_QME : LE2_DEFAULT;
+	ibsd_wr_allchans(ppd, 13, (le_val << 7), BMASK(9, 7));
+
+	/* enable LE1 adaptation for all but QME, which is disabled */
+	le_val = IS_QME(ppd->dd) ? 0 : 1;
+	ibsd_wr_allchans(ppd, 13, (le_val << 5), (1 << 5));
+
+	/* Clear cmode-override, may be set from older driver */
+	ahb_mod(ppd->dd, IBSD(ppd->hw_pidx), 5, 10, 0 << 14, 1 << 14);
+
+	/* Timing Recovery: rxtapsel addr 5 bits [9:8] = 0 */
+	ibsd_wr_allchans(ppd, 5, (0 << 8), BMASK(9, 8));
+
+	/* setup LoS params; these are subsystem, so chan == 5 */
+	/* LoS filter threshold_count on, ch 0-3, set to 8 */
+	ahb_mod(ppd->dd, IBSD(ppd->hw_pidx), 5, 5, 8 << 11, BMASK(14, 11));
+	ahb_mod(ppd->dd, IBSD(ppd->hw_pidx), 5, 7, 8 << 4, BMASK(7, 4));
+	ahb_mod(ppd->dd, IBSD(ppd->hw_pidx), 5, 8, 8 << 11, BMASK(14, 11));
+	ahb_mod(ppd->dd, IBSD(ppd->hw_pidx), 5, 10, 8 << 4, BMASK(7, 4));
+
+	/* LoS filter threshold_count off, ch 0-3, set to 4 */
+	ahb_mod(ppd->dd, IBSD(ppd->hw_pidx), 5, 6, 4 << 0, BMASK(3, 0));
+	ahb_mod(ppd->dd, IBSD(ppd->hw_pidx), 5, 7, 4 << 8, BMASK(11, 8));
+	ahb_mod(ppd->dd, IBSD(ppd->hw_pidx), 5, 9, 4 << 0, BMASK(3, 0));
+	ahb_mod(ppd->dd, IBSD(ppd->hw_pidx), 5, 10, 4 << 8, BMASK(11, 8));
+
+	/* LoS filter select enabled */
+	ahb_mod(ppd->dd, IBSD(ppd->hw_pidx), 5, 9, 1 << 15, 1 << 15);
+
+	/* LoS target data:  SDR=4, DDR=2, QDR=1 */
+	ibsd_wr_allchans(ppd, 14, (1 << 3), BMASK(5, 3)); /* QDR */
+	ibsd_wr_allchans(ppd, 20, (2 << 10), BMASK(12, 10)); /* DDR */
+	ibsd_wr_allchans(ppd, 20, (4 << 13), BMASK(15, 13)); /* SDR */
+
+	data = qib_read_kreg_port(ppd, krp_serdesctrl);
+	qib_write_kreg_port(ppd, krp_serdesctrl, data |
+		SYM_MASK(IBSerdesCtrl_0, RXLOSEN));
+
+	/* rxbistena; set 0 to avoid effects of it switch later */
+	ibsd_wr_allchans(ppd, 9, 0 << 15, 1 << 15);
+
+	/* Configure 4 DFE taps, and only they adapt */
+	ibsd_wr_allchans(ppd, 16, 0 << 0, BMASK(1, 0));
+
+	/* gain hi stop 32 (22) (6:1) lo stop 7 (10:7) target 22 (13) (15:11) */
+	le_val = (ppd->dd->cspec->r1 || IS_QME(ppd->dd)) ? 0xb6c0 : 0x6bac;
+	ibsd_wr_allchans(ppd, 21, le_val, 0xfffe);
+
+	/*
+	 * Set receive adaptation mode.  SDR and DDR adaptation are
+	 * always on, and QDR is initially enabled; later disabled.
+	 */
+	qib_write_kreg_port(ppd, krp_static_adapt_dis(0), 0ULL);
+	qib_write_kreg_port(ppd, krp_static_adapt_dis(1), 0ULL);
+	qib_write_kreg_port(ppd, krp_static_adapt_dis(2),
+			    ppd->dd->cspec->r1 ?
+			    QDR_STATIC_ADAPT_DOWN_R1 : QDR_STATIC_ADAPT_DOWN);
+	ppd->cpspec->qdr_dfe_on = 1;
+
+	/* FLoop LOS gate: PPM filter  enabled */
+	ibsd_wr_allchans(ppd, 38, 0 << 10, 1 << 10);
+
+	/* rx offset center enabled */
+	ibsd_wr_allchans(ppd, 12, 1 << 4, 1 << 4);
+
+	if (!ppd->dd->cspec->r1) {
+		ibsd_wr_allchans(ppd, 12, 1 << 12, 1 << 12);
+		ibsd_wr_allchans(ppd, 12, 2 << 8, 0x0f << 8);
+	}
+
+	/* Set the frequency loop bandwidth to 15 */
+	ibsd_wr_allchans(ppd, 2, 15 << 5, BMASK(8, 5));
+
+	return 0;
+}
+
+/* start adjust QMH serdes parameters */
+
+static void set_man_code(struct qib_pportdata *ppd, int chan, int code)
+{
+	ahb_mod(ppd->dd, IBSD(ppd->hw_pidx), (chan + (chan >> 1)),
+		9, code << 9, 0x3f << 9);
+}
+
+static void set_man_mode_h1(struct qib_pportdata *ppd, int chan,
+	int enable, u32 tapenable)
+{
+	if (enable)
+		ahb_mod(ppd->dd, IBSD(ppd->hw_pidx), (chan + (chan >> 1)),
+			1, 3 << 10, 0x1f << 10);
+	else
+		ahb_mod(ppd->dd, IBSD(ppd->hw_pidx), (chan + (chan >> 1)),
+			1, 0, 0x1f << 10);
+}
+
+/* Set clock to 1, 0, 1, 0 */
+static void clock_man(struct qib_pportdata *ppd, int chan)
+{
+	ahb_mod(ppd->dd, IBSD(ppd->hw_pidx), (chan + (chan >> 1)),
+		4, 0x4000, 0x4000);
+	ahb_mod(ppd->dd, IBSD(ppd->hw_pidx), (chan + (chan >> 1)),
+		4, 0, 0x4000);
+	ahb_mod(ppd->dd, IBSD(ppd->hw_pidx), (chan + (chan >> 1)),
+		4, 0x4000, 0x4000);
+	ahb_mod(ppd->dd, IBSD(ppd->hw_pidx), (chan + (chan >> 1)),
+		4, 0, 0x4000);
+}
+
+/*
+ * write the current Tx serdes pre,post,main,amp settings into the serdes.
+ * The caller must pass the settings appropriate for the current speed,
+ * or not care if they are correct for the current speed.
+ */
+static void write_tx_serdes_param(struct qib_pportdata *ppd,
+				  struct txdds_ent *txdds)
+{
+	u64 deemph;
+
+	deemph = qib_read_kreg_port(ppd, krp_tx_deemph_override);
+	/* field names for amp, main, post, pre, respectively */
+	deemph &= ~(SYM_MASK(IBSD_TX_DEEMPHASIS_OVERRIDE_0, txampcntl_d2a) |
+		    SYM_MASK(IBSD_TX_DEEMPHASIS_OVERRIDE_0, txc0_ena) |
+		    SYM_MASK(IBSD_TX_DEEMPHASIS_OVERRIDE_0, txcp1_ena) |
+		    SYM_MASK(IBSD_TX_DEEMPHASIS_OVERRIDE_0, txcn1_ena));
+
+	deemph |= SYM_MASK(IBSD_TX_DEEMPHASIS_OVERRIDE_0,
+			   tx_override_deemphasis_select);
+	deemph |= (txdds->amp & SYM_RMASK(IBSD_TX_DEEMPHASIS_OVERRIDE_0,
+		    txampcntl_d2a)) << SYM_LSB(IBSD_TX_DEEMPHASIS_OVERRIDE_0,
+				       txampcntl_d2a);
+	deemph |= (txdds->main & SYM_RMASK(IBSD_TX_DEEMPHASIS_OVERRIDE_0,
+		     txc0_ena)) << SYM_LSB(IBSD_TX_DEEMPHASIS_OVERRIDE_0,
+				   txc0_ena);
+	deemph |= (txdds->post & SYM_RMASK(IBSD_TX_DEEMPHASIS_OVERRIDE_0,
+		     txcp1_ena)) << SYM_LSB(IBSD_TX_DEEMPHASIS_OVERRIDE_0,
+				    txcp1_ena);
+	deemph |= (txdds->pre & SYM_RMASK(IBSD_TX_DEEMPHASIS_OVERRIDE_0,
+		     txcn1_ena)) << SYM_LSB(IBSD_TX_DEEMPHASIS_OVERRIDE_0,
+				    txcn1_ena);
+	qib_write_kreg_port(ppd, krp_tx_deemph_override, deemph);
+}
+
+/*
+ * Set the parameters for mez cards on link bounce, so they are
+ * always exactly what was requested.  Similar logic to init_txdds
+ * but does just the serdes.
+ */
+static void adj_tx_serdes(struct qib_pportdata *ppd)
+{
+	const struct txdds_ent *sdr_dds, *ddr_dds, *qdr_dds;
+	struct txdds_ent *dds;
+
+	find_best_ent(ppd, &sdr_dds, &ddr_dds, &qdr_dds, 1);
+	dds = (struct txdds_ent *)(ppd->link_speed_active == QIB_IB_QDR ?
+		qdr_dds : (ppd->link_speed_active == QIB_IB_DDR ?
+				ddr_dds : sdr_dds));
+	write_tx_serdes_param(ppd, dds);
+}
+
+/* set QDR forced value for H1, if needed */
+static void force_h1(struct qib_pportdata *ppd)
+{
+	int chan;
+
+	ppd->cpspec->qdr_reforce = 0;
+	if (!ppd->dd->cspec->r1)
+		return;
+
+	for (chan = 0; chan < SERDES_CHANS; chan++) {
+		set_man_mode_h1(ppd, chan, 1, 0);
+		set_man_code(ppd, chan, ppd->cpspec->h1_val);
+		clock_man(ppd, chan);
+		set_man_mode_h1(ppd, chan, 0, 0);
+	}
+}
+
+#define SJA_EN SYM_MASK(SPC_JTAG_ACCESS_REG, SPC_JTAG_ACCESS_EN)
+#define BISTEN_LSB SYM_LSB(SPC_JTAG_ACCESS_REG, bist_en)
+
+#define R_OPCODE_LSB 3
+#define R_OP_NOP 0
+#define R_OP_SHIFT 2
+#define R_OP_UPDATE 3
+#define R_TDI_LSB 2
+#define R_TDO_LSB 1
+#define R_RDY 1
+
+static int qib_r_grab(struct qib_devdata *dd)
+{
+	u64 val;
+	val = SJA_EN;
+	qib_write_kreg(dd, kr_r_access, val);
+	qib_read_kreg32(dd, kr_scratch);
+	return 0;
+}
+
+/* qib_r_wait_for_rdy() not only waits for the ready bit, it
+ * returns the current state of R_TDO
+ */
+static int qib_r_wait_for_rdy(struct qib_devdata *dd)
+{
+	u64 val;
+	int timeout;
+	for (timeout = 0; timeout < 100 ; ++timeout) {
+		val = qib_read_kreg32(dd, kr_r_access);
+		if (val & R_RDY)
+			return (val >> R_TDO_LSB) & 1;
+	}
+	return -1;
+}
+
+static int qib_r_shift(struct qib_devdata *dd, int bisten,
+		       int len, u8 *inp, u8 *outp)
+{
+	u64 valbase, val;
+	int ret, pos;
+
+	valbase = SJA_EN | (bisten << BISTEN_LSB) |
+		(R_OP_SHIFT << R_OPCODE_LSB);
+	ret = qib_r_wait_for_rdy(dd);
+	if (ret < 0)
+		goto bail;
+	for (pos = 0; pos < len; ++pos) {
+		val = valbase;
+		if (outp) {
+			outp[pos >> 3] &= ~(1 << (pos & 7));
+			outp[pos >> 3] |= (ret << (pos & 7));
+		}
+		if (inp) {
+			int tdi = inp[pos >> 3] >> (pos & 7);
+			val |= ((tdi & 1) << R_TDI_LSB);
+		}
+		qib_write_kreg(dd, kr_r_access, val);
+		qib_read_kreg32(dd, kr_scratch);
+		ret = qib_r_wait_for_rdy(dd);
+		if (ret < 0)
+			break;
+	}
+	/* Restore to NOP between operations. */
+	val =  SJA_EN | (bisten << BISTEN_LSB);
+	qib_write_kreg(dd, kr_r_access, val);
+	qib_read_kreg32(dd, kr_scratch);
+	ret = qib_r_wait_for_rdy(dd);
+
+	if (ret >= 0)
+		ret = pos;
+bail:
+	return ret;
+}
+
+static int qib_r_update(struct qib_devdata *dd, int bisten)
+{
+	u64 val;
+	int ret;
+
+	val = SJA_EN | (bisten << BISTEN_LSB) | (R_OP_UPDATE << R_OPCODE_LSB);
+	ret = qib_r_wait_for_rdy(dd);
+	if (ret >= 0) {
+		qib_write_kreg(dd, kr_r_access, val);
+		qib_read_kreg32(dd, kr_scratch);
+	}
+	return ret;
+}
+
+#define BISTEN_PORT_SEL 15
+#define LEN_PORT_SEL 625
+#define BISTEN_AT 17
+#define LEN_AT 156
+#define BISTEN_ETM 16
+#define LEN_ETM 632
+
+#define BIT2BYTE(x) (((x) +  BITS_PER_BYTE - 1) / BITS_PER_BYTE)
+
+/* these are common for all IB port use cases. */
+static u8 reset_at[BIT2BYTE(LEN_AT)] = {
+	0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+	0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x20, 0x00,
+};
+static u8 reset_atetm[BIT2BYTE(LEN_ETM)] = {
+	0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+	0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+	0x00, 0x00, 0x00, 0x80, 0xe3, 0x81, 0x73, 0x3c, 0x70, 0x8e,
+	0x07, 0xce, 0xf1, 0xc0, 0x39, 0x1e, 0x38, 0xc7, 0x03, 0xe7,
+	0x78, 0xe0, 0x1c, 0x0f, 0x9c, 0x7f, 0x80, 0x73, 0x0f, 0x70,
+	0xde, 0x01, 0xce, 0x39, 0xc0, 0xf9, 0x06, 0x38, 0xd7, 0x00,
+	0xe7, 0x19, 0xe0, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+	0x00, 0xff, 0xff, 0xff, 0xff, 0x00, 0x00, 0x00, 0x00,
+};
+static u8 at[BIT2BYTE(LEN_AT)] = {
+	0x00, 0x00, 0x18, 0x00, 0x00, 0x00, 0x18, 0x00, 0x00, 0x00,
+	0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x20, 0x00,
+};
+
+/* used for IB1 or IB2, only one in use */
+static u8 atetm_1port[BIT2BYTE(LEN_ETM)] = {
+	0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+	0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+	0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+	0x00, 0x10, 0xf2, 0x80, 0x83, 0x1e, 0x38, 0x00, 0x00, 0x00,
+	0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+	0x00, 0x00, 0x50, 0xf4, 0x41, 0x00, 0x18, 0x78, 0xc8, 0x03,
+	0x07, 0x7b, 0xa0, 0x3e, 0x00, 0x02, 0x00, 0x00, 0x18, 0x00,
+	0x18, 0x00, 0x00, 0x00, 0x00, 0x4b, 0x00, 0x00, 0x00,
+};
+
+/* used when both IB1 and IB2 are in use */
+static u8 atetm_2port[BIT2BYTE(LEN_ETM)] = {
+	0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+	0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x79,
+	0xc0, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+	0x00, 0x00, 0xf8, 0x80, 0x83, 0x1e, 0x38, 0xe0, 0x03, 0x05,
+	0x7b, 0xa0, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x80,
+	0xa2, 0x0f, 0x50, 0xf4, 0x41, 0x00, 0x18, 0x78, 0xd1, 0x07,
+	0x02, 0x7c, 0x80, 0x3e, 0x00, 0x02, 0x00, 0x00, 0x3e, 0x00,
+	0x02, 0x00, 0x00, 0x00, 0x00, 0x64, 0x00, 0x00, 0x00,
+};
+
+/* used when only IB1 is in use */
+static u8 portsel_port1[BIT2BYTE(LEN_PORT_SEL)] = {
+	0x32, 0x65, 0xa4, 0x7b, 0x10, 0x98, 0xdc, 0xfe, 0x13, 0x13,
+	0x13, 0x13, 0x13, 0x13, 0x13, 0x13, 0x73, 0x0c, 0x0c, 0x0c,
+	0x0c, 0x0c, 0x13, 0x13, 0x13, 0x13, 0x13, 0x13, 0x13, 0x13,
+	0x13, 0x78, 0x78, 0x13, 0x13, 0x13, 0x13, 0x13, 0x13, 0x13,
+	0x14, 0x14, 0x14, 0x14, 0x14, 0x14, 0x14, 0x14, 0x74, 0x32,
+	0x32, 0x32, 0x32, 0x32, 0x14, 0x14, 0x14, 0x14, 0x14, 0x14,
+	0x14, 0x14, 0x14, 0x14, 0x14, 0x14, 0x14, 0x14, 0x14, 0x14,
+	0x14, 0x14, 0x9f, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00,
+};
+
+/* used when only IB2 is in use */
+static u8 portsel_port2[BIT2BYTE(LEN_PORT_SEL)] = {
+	0x32, 0x65, 0xa4, 0x7b, 0x10, 0x98, 0xdc, 0xfe, 0x39, 0x39,
+	0x39, 0x39, 0x39, 0x39, 0x39, 0x39, 0x73, 0x32, 0x32, 0x32,
+	0x32, 0x32, 0x39, 0x39, 0x39, 0x39, 0x39, 0x39, 0x39, 0x39,
+	0x39, 0x78, 0x78, 0x39, 0x39, 0x39, 0x39, 0x39, 0x39, 0x39,
+	0x3a, 0x3a, 0x3a, 0x3a, 0x3a, 0x3a, 0x3a, 0x3a, 0x74, 0x32,
+	0x32, 0x32, 0x32, 0x32, 0x3a, 0x3a, 0x3a, 0x3a, 0x3a, 0x3a,
+	0x3a, 0x3a, 0x3a, 0x3a, 0x3a, 0x3a, 0x3a, 0x3a, 0x3a, 0x3a,
+	0x3a, 0x3a, 0x9f, 0x01, 0x00, 0x00, 0x00, 0x00, 0x01,
+};
+
+/* used when both IB1 and IB2 are in use */
+static u8 portsel_2port[BIT2BYTE(LEN_PORT_SEL)] = {
+	0x32, 0xba, 0x54, 0x76, 0x10, 0x98, 0xdc, 0xfe, 0x13, 0x13,
+	0x13, 0x13, 0x13, 0x13, 0x13, 0x13, 0x73, 0x0c, 0x0c, 0x0c,
+	0x0c, 0x0c, 0x13, 0x13, 0x13, 0x13, 0x13, 0x13, 0x13, 0x13,
+	0x13, 0x13, 0x13, 0x13, 0x13, 0x13, 0x13, 0x13, 0x13, 0x13,
+	0x14, 0x14, 0x14, 0x14, 0x14, 0x14, 0x14, 0x14, 0x74, 0x32,
+	0x32, 0x32, 0x32, 0x32, 0x14, 0x14, 0x14, 0x14, 0x14, 0x3a,
+	0x3a, 0x14, 0x14, 0x14, 0x14, 0x14, 0x14, 0x14, 0x14, 0x14,
+	0x14, 0x14, 0x9f, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00,
+};
+
+/*
+ * Do setup to properly handle IB link recovery; if port is zero, we
+ * are initializing to cover both ports; otherwise we are initializing
+ * to cover a single port card, or the port has reached INIT and we may
+ * need to switch coverage types.
+ */
+static void setup_7322_link_recovery(struct qib_pportdata *ppd, u32 both)
+{
+	u8 *portsel, *etm;
+	struct qib_devdata *dd = ppd->dd;
+
+	if (!ppd->dd->cspec->r1)
+		return;
+	if (!both) {
+		dd->cspec->recovery_ports_initted++;
+		ppd->cpspec->recovery_init = 1;
+	}
+	if (!both && dd->cspec->recovery_ports_initted == 1) {
+		portsel = ppd->port == 1 ? portsel_port1 : portsel_port2;
+		etm = atetm_1port;
+	} else {
+		portsel = portsel_2port;
+		etm = atetm_2port;
+	}
+
+	if (qib_r_grab(dd) < 0 ||
+		qib_r_shift(dd, BISTEN_ETM, LEN_ETM, reset_atetm, NULL) < 0 ||
+		qib_r_update(dd, BISTEN_ETM) < 0 ||
+		qib_r_shift(dd, BISTEN_AT, LEN_AT, reset_at, NULL) < 0 ||
+		qib_r_update(dd, BISTEN_AT) < 0 ||
+		qib_r_shift(dd, BISTEN_PORT_SEL, LEN_PORT_SEL,
+			    portsel, NULL) < 0 ||
+		qib_r_update(dd, BISTEN_PORT_SEL) < 0 ||
+		qib_r_shift(dd, BISTEN_AT, LEN_AT, at, NULL) < 0 ||
+		qib_r_update(dd, BISTEN_AT) < 0 ||
+		qib_r_shift(dd, BISTEN_ETM, LEN_ETM, etm, NULL) < 0 ||
+		qib_r_update(dd, BISTEN_ETM) < 0)
+		qib_dev_err(dd, "Failed IB link recovery setup\n");
+}
+
+static void check_7322_rxe_status(struct qib_pportdata *ppd)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u64 fmask;
+
+	if (dd->cspec->recovery_ports_initted != 1)
+		return; /* rest doesn't apply to dualport */
+	qib_write_kreg(dd, kr_control, dd->control |
+		       SYM_MASK(Control, FreezeMode));
+	(void)qib_read_kreg64(dd, kr_scratch);
+	udelay(3); /* ibcreset asserted 400ns, be sure that's over */
+	fmask = qib_read_kreg64(dd, kr_act_fmask);
+	if (!fmask) {
+		/*
+		 * require a powercycle before we'll work again, and make
+		 * sure we get no more interrupts, and don't turn off
+		 * freeze.
+		 */
+		ppd->dd->cspec->stay_in_freeze = 1;
+		qib_7322_set_intr_state(ppd->dd, 0);
+		qib_write_kreg(dd, kr_fmask, 0ULL);
+		qib_dev_err(dd, "HCA unusable until powercycled\n");
+		return; /* eventually reset */
+	}
+
+	qib_write_kreg(ppd->dd, kr_hwerrclear,
+	    SYM_MASK(HwErrClear, IBSerdesPClkNotDetectClear_1));
+
+	/* don't do the full clear_freeze(), not needed for this */
+	qib_write_kreg(dd, kr_control, dd->control);
+	qib_read_kreg32(dd, kr_scratch);
+	/* take IBC out of reset */
+	if (ppd->link_speed_supported) {
+		ppd->cpspec->ibcctrl_a &=
+			~SYM_MASK(IBCCtrlA_0, IBStatIntReductionEn);
+		qib_write_kreg_port(ppd, krp_ibcctrl_a,
+				    ppd->cpspec->ibcctrl_a);
+		qib_read_kreg32(dd, kr_scratch);
+		if (ppd->lflags & QIBL_IB_LINK_DISABLED)
+			qib_set_ib_7322_lstate(ppd, 0,
+				QLOGIC_IB_IBCC_LINKINITCMD_DISABLE);
+	}
+}
diff --git a/drivers/infiniband/hw/qib/qib_init.c b/drivers/infiniband/hw/qib/qib_init.c
new file mode 100644
index 000000000000..9b40f345ac3f
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_init.c
@@ -0,0 +1,1586 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009, 2010 QLogic Corporation.
+ * All rights reserved.
+ * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/pci.h>
+#include <linux/netdevice.h>
+#include <linux/vmalloc.h>
+#include <linux/delay.h>
+#include <linux/idr.h>
+
+#include "qib.h"
+#include "qib_common.h"
+
+/*
+ * min buffers we want to have per context, after driver
+ */
+#define QIB_MIN_USER_CTXT_BUFCNT 7
+
+#define QLOGIC_IB_R_SOFTWARE_MASK 0xFF
+#define QLOGIC_IB_R_SOFTWARE_SHIFT 24
+#define QLOGIC_IB_R_EMULATOR_MASK (1ULL<<62)
+
+/*
+ * Number of ctxts we are configured to use (to allow for more pio
+ * buffers per ctxt, etc.)  Zero means use chip value.
+ */
+ushort qib_cfgctxts;
+module_param_named(cfgctxts, qib_cfgctxts, ushort, S_IRUGO);
+MODULE_PARM_DESC(cfgctxts, "Set max number of contexts to use");
+
+/*
+ * If set, do not write to any regs if avoidable, hack to allow
+ * check for deranged default register values.
+ */
+ushort qib_mini_init;
+module_param_named(mini_init, qib_mini_init, ushort, S_IRUGO);
+MODULE_PARM_DESC(mini_init, "If set, do minimal diag init");
+
+unsigned qib_n_krcv_queues;
+module_param_named(krcvqs, qib_n_krcv_queues, uint, S_IRUGO);
+MODULE_PARM_DESC(krcvqs, "number of kernel receive queues per IB port");
+
+/*
+ * qib_wc_pat parameter:
+ *      0 is WC via MTRR
+ *      1 is WC via PAT
+ *      If PAT initialization fails, code reverts back to MTRR
+ */
+unsigned qib_wc_pat = 1; /* default (1) is to use PAT, not MTRR */
+module_param_named(wc_pat, qib_wc_pat, uint, S_IRUGO);
+MODULE_PARM_DESC(wc_pat, "enable write-combining via PAT mechanism");
+
+struct workqueue_struct *qib_wq;
+struct workqueue_struct *qib_cq_wq;
+
+static void verify_interrupt(unsigned long);
+
+static struct idr qib_unit_table;
+u32 qib_cpulist_count;
+unsigned long *qib_cpulist;
+
+/* set number of contexts we'll actually use */
+void qib_set_ctxtcnt(struct qib_devdata *dd)
+{
+	if (!qib_cfgctxts)
+		dd->cfgctxts = dd->ctxtcnt;
+	else if (qib_cfgctxts < dd->num_pports)
+		dd->cfgctxts = dd->ctxtcnt;
+	else if (qib_cfgctxts <= dd->ctxtcnt)
+		dd->cfgctxts = qib_cfgctxts;
+	else
+		dd->cfgctxts = dd->ctxtcnt;
+}
+
+/*
+ * Common code for creating the receive context array.
+ */
+int qib_create_ctxts(struct qib_devdata *dd)
+{
+	unsigned i;
+	int ret;
+
+	/*
+	 * Allocate full ctxtcnt array, rather than just cfgctxts, because
+	 * cleanup iterates across all possible ctxts.
+	 */
+	dd->rcd = kzalloc(sizeof(*dd->rcd) * dd->ctxtcnt, GFP_KERNEL);
+	if (!dd->rcd) {
+		qib_dev_err(dd, "Unable to allocate ctxtdata array, "
+			    "failing\n");
+		ret = -ENOMEM;
+		goto done;
+	}
+
+	/* create (one or more) kctxt */
+	for (i = 0; i < dd->first_user_ctxt; ++i) {
+		struct qib_pportdata *ppd;
+		struct qib_ctxtdata *rcd;
+
+		if (dd->skip_kctxt_mask & (1 << i))
+			continue;
+
+		ppd = dd->pport + (i % dd->num_pports);
+		rcd = qib_create_ctxtdata(ppd, i);
+		if (!rcd) {
+			qib_dev_err(dd, "Unable to allocate ctxtdata"
+				    " for Kernel ctxt, failing\n");
+			ret = -ENOMEM;
+			goto done;
+		}
+		rcd->pkeys[0] = QIB_DEFAULT_P_KEY;
+		rcd->seq_cnt = 1;
+	}
+	ret = 0;
+done:
+	return ret;
+}
+
+/*
+ * Common code for user and kernel context setup.
+ */
+struct qib_ctxtdata *qib_create_ctxtdata(struct qib_pportdata *ppd, u32 ctxt)
+{
+	struct qib_devdata *dd = ppd->dd;
+	struct qib_ctxtdata *rcd;
+
+	rcd = kzalloc(sizeof(*rcd), GFP_KERNEL);
+	if (rcd) {
+		INIT_LIST_HEAD(&rcd->qp_wait_list);
+		rcd->ppd = ppd;
+		rcd->dd = dd;
+		rcd->cnt = 1;
+		rcd->ctxt = ctxt;
+		dd->rcd[ctxt] = rcd;
+
+		dd->f_init_ctxt(rcd);
+
+		/*
+		 * To avoid wasting a lot of memory, we allocate 32KB chunks
+		 * of physically contiguous memory, advance through it until
+		 * used up and then allocate more.  Of course, we need
+		 * memory to store those extra pointers, now.  32KB seems to
+		 * be the most that is "safe" under memory pressure
+		 * (creating large files and then copying them over
+		 * NFS while doing lots of MPI jobs).  The OOM killer can
+		 * get invoked, even though we say we can sleep and this can
+		 * cause significant system problems....
+		 */
+		rcd->rcvegrbuf_size = 0x8000;
+		rcd->rcvegrbufs_perchunk =
+			rcd->rcvegrbuf_size / dd->rcvegrbufsize;
+		rcd->rcvegrbuf_chunks = (rcd->rcvegrcnt +
+			rcd->rcvegrbufs_perchunk - 1) /
+			rcd->rcvegrbufs_perchunk;
+	}
+	return rcd;
+}
+
+/*
+ * Common code for initializing the physical port structure.
+ */
+void qib_init_pportdata(struct qib_pportdata *ppd, struct qib_devdata *dd,
+			u8 hw_pidx, u8 port)
+{
+	ppd->dd = dd;
+	ppd->hw_pidx = hw_pidx;
+	ppd->port = port; /* IB port number, not index */
+
+	spin_lock_init(&ppd->sdma_lock);
+	spin_lock_init(&ppd->lflags_lock);
+	init_waitqueue_head(&ppd->state_wait);
+
+	init_timer(&ppd->symerr_clear_timer);
+	ppd->symerr_clear_timer.function = qib_clear_symerror_on_linkup;
+	ppd->symerr_clear_timer.data = (unsigned long)ppd;
+}
+
+static int init_pioavailregs(struct qib_devdata *dd)
+{
+	int ret, pidx;
+	u64 *status_page;
+
+	dd->pioavailregs_dma = dma_alloc_coherent(
+		&dd->pcidev->dev, PAGE_SIZE, &dd->pioavailregs_phys,
+		GFP_KERNEL);
+	if (!dd->pioavailregs_dma) {
+		qib_dev_err(dd, "failed to allocate PIOavail reg area "
+			    "in memory\n");
+		ret = -ENOMEM;
+		goto done;
+	}
+
+	/*
+	 * We really want L2 cache aligned, but for current CPUs of
+	 * interest, they are the same.
+	 */
+	status_page = (u64 *)
+		((char *) dd->pioavailregs_dma +
+		 ((2 * L1_CACHE_BYTES +
+		   dd->pioavregs * sizeof(u64)) & ~L1_CACHE_BYTES));
+	/* device status comes first, for backwards compatibility */
+	dd->devstatusp = status_page;
+	*status_page++ = 0;
+	for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+		dd->pport[pidx].statusp = status_page;
+		*status_page++ = 0;
+	}
+
+	/*
+	 * Setup buffer to hold freeze and other messages, accessible to
+	 * apps, following statusp.  This is per-unit, not per port.
+	 */
+	dd->freezemsg = (char *) status_page;
+	*dd->freezemsg = 0;
+	/* length of msg buffer is "whatever is left" */
+	ret = (char *) status_page - (char *) dd->pioavailregs_dma;
+	dd->freezelen = PAGE_SIZE - ret;
+
+	ret = 0;
+
+done:
+	return ret;
+}
+
+/**
+ * init_shadow_tids - allocate the shadow TID array
+ * @dd: the qlogic_ib device
+ *
+ * allocate the shadow TID array, so we can qib_munlock previous
+ * entries.  It may make more sense to move the pageshadow to the
+ * ctxt data structure, so we only allocate memory for ctxts actually
+ * in use, since we at 8k per ctxt, now.
+ * We don't want failures here to prevent use of the driver/chip,
+ * so no return value.
+ */
+static void init_shadow_tids(struct qib_devdata *dd)
+{
+	struct page **pages;
+	dma_addr_t *addrs;
+
+	pages = vmalloc(dd->cfgctxts * dd->rcvtidcnt * sizeof(struct page *));
+	if (!pages) {
+		qib_dev_err(dd, "failed to allocate shadow page * "
+			    "array, no expected sends!\n");
+		goto bail;
+	}
+
+	addrs = vmalloc(dd->cfgctxts * dd->rcvtidcnt * sizeof(dma_addr_t));
+	if (!addrs) {
+		qib_dev_err(dd, "failed to allocate shadow dma handle "
+			    "array, no expected sends!\n");
+		goto bail_free;
+	}
+
+	memset(pages, 0, dd->cfgctxts * dd->rcvtidcnt * sizeof(struct page *));
+	memset(addrs, 0, dd->cfgctxts * dd->rcvtidcnt * sizeof(dma_addr_t));
+
+	dd->pageshadow = pages;
+	dd->physshadow = addrs;
+	return;
+
+bail_free:
+	vfree(pages);
+bail:
+	dd->pageshadow = NULL;
+}
+
+/*
+ * Do initialization for device that is only needed on
+ * first detect, not on resets.
+ */
+static int loadtime_init(struct qib_devdata *dd)
+{
+	int ret = 0;
+
+	if (((dd->revision >> QLOGIC_IB_R_SOFTWARE_SHIFT) &
+	     QLOGIC_IB_R_SOFTWARE_MASK) != QIB_CHIP_SWVERSION) {
+		qib_dev_err(dd, "Driver only handles version %d, "
+			    "chip swversion is %d (%llx), failng\n",
+			    QIB_CHIP_SWVERSION,
+			    (int)(dd->revision >>
+				QLOGIC_IB_R_SOFTWARE_SHIFT) &
+			    QLOGIC_IB_R_SOFTWARE_MASK,
+			    (unsigned long long) dd->revision);
+		ret = -ENOSYS;
+		goto done;
+	}
+
+	if (dd->revision & QLOGIC_IB_R_EMULATOR_MASK)
+		qib_devinfo(dd->pcidev, "%s", dd->boardversion);
+
+	spin_lock_init(&dd->pioavail_lock);
+	spin_lock_init(&dd->sendctrl_lock);
+	spin_lock_init(&dd->uctxt_lock);
+	spin_lock_init(&dd->qib_diag_trans_lock);
+	spin_lock_init(&dd->eep_st_lock);
+	mutex_init(&dd->eep_lock);
+
+	if (qib_mini_init)
+		goto done;
+
+	ret = init_pioavailregs(dd);
+	init_shadow_tids(dd);
+
+	qib_get_eeprom_info(dd);
+
+	/* setup time (don't start yet) to verify we got interrupt */
+	init_timer(&dd->intrchk_timer);
+	dd->intrchk_timer.function = verify_interrupt;
+	dd->intrchk_timer.data = (unsigned long) dd;
+
+done:
+	return ret;
+}
+
+/**
+ * init_after_reset - re-initialize after a reset
+ * @dd: the qlogic_ib device
+ *
+ * sanity check at least some of the values after reset, and
+ * ensure no receive or transmit (explictly, in case reset
+ * failed
+ */
+static int init_after_reset(struct qib_devdata *dd)
+{
+	int i;
+
+	/*
+	 * Ensure chip does no sends or receives, tail updates, or
+	 * pioavail updates while we re-initialize.  This is mostly
+	 * for the driver data structures, not chip registers.
+	 */
+	for (i = 0; i < dd->num_pports; ++i) {
+		/*
+		 * ctxt == -1 means "all contexts". Only really safe for
+		 * _dis_abling things, as here.
+		 */
+		dd->f_rcvctrl(dd->pport + i, QIB_RCVCTRL_CTXT_DIS |
+				  QIB_RCVCTRL_INTRAVAIL_DIS |
+				  QIB_RCVCTRL_TAILUPD_DIS, -1);
+		/* Redundant across ports for some, but no big deal.  */
+		dd->f_sendctrl(dd->pport + i, QIB_SENDCTRL_SEND_DIS |
+			QIB_SENDCTRL_AVAIL_DIS);
+	}
+
+	return 0;
+}
+
+static void enable_chip(struct qib_devdata *dd)
+{
+	u64 rcvmask;
+	int i;
+
+	/*
+	 * Enable PIO send, and update of PIOavail regs to memory.
+	 */
+	for (i = 0; i < dd->num_pports; ++i)
+		dd->f_sendctrl(dd->pport + i, QIB_SENDCTRL_SEND_ENB |
+			QIB_SENDCTRL_AVAIL_ENB);
+	/*
+	 * Enable kernel ctxts' receive and receive interrupt.
+	 * Other ctxts done as user opens and inits them.
+	 */
+	rcvmask = QIB_RCVCTRL_CTXT_ENB | QIB_RCVCTRL_INTRAVAIL_ENB;
+	rcvmask |= (dd->flags & QIB_NODMA_RTAIL) ?
+		  QIB_RCVCTRL_TAILUPD_DIS : QIB_RCVCTRL_TAILUPD_ENB;
+	for (i = 0; dd->rcd && i < dd->first_user_ctxt; ++i) {
+		struct qib_ctxtdata *rcd = dd->rcd[i];
+
+		if (rcd)
+			dd->f_rcvctrl(rcd->ppd, rcvmask, i);
+	}
+}
+
+static void verify_interrupt(unsigned long opaque)
+{
+	struct qib_devdata *dd = (struct qib_devdata *) opaque;
+
+	if (!dd)
+		return; /* being torn down */
+
+	/*
+	 * If we don't have a lid or any interrupts, let the user know and
+	 * don't bother checking again.
+	 */
+	if (dd->int_counter == 0) {
+		if (!dd->f_intr_fallback(dd))
+			dev_err(&dd->pcidev->dev, "No interrupts detected, "
+				"not usable.\n");
+		else /* re-arm the timer to see if fallback works */
+			mod_timer(&dd->intrchk_timer, jiffies + HZ/2);
+	}
+}
+
+static void init_piobuf_state(struct qib_devdata *dd)
+{
+	int i, pidx;
+	u32 uctxts;
+
+	/*
+	 * Ensure all buffers are free, and fifos empty.  Buffers
+	 * are common, so only do once for port 0.
+	 *
+	 * After enable and qib_chg_pioavailkernel so we can safely
+	 * enable pioavail updates and PIOENABLE.  After this, packets
+	 * are ready and able to go out.
+	 */
+	dd->f_sendctrl(dd->pport, QIB_SENDCTRL_DISARM_ALL);
+	for (pidx = 0; pidx < dd->num_pports; ++pidx)
+		dd->f_sendctrl(dd->pport + pidx, QIB_SENDCTRL_FLUSH);
+
+	/*
+	 * If not all sendbufs are used, add the one to each of the lower
+	 * numbered contexts.  pbufsctxt and lastctxt_piobuf are
+	 * calculated in chip-specific code because it may cause some
+	 * chip-specific adjustments to be made.
+	 */
+	uctxts = dd->cfgctxts - dd->first_user_ctxt;
+	dd->ctxts_extrabuf = dd->pbufsctxt ?
+		dd->lastctxt_piobuf - (dd->pbufsctxt * uctxts) : 0;
+
+	/*
+	 * Set up the shadow copies of the piobufavail registers,
+	 * which we compare against the chip registers for now, and
+	 * the in memory DMA'ed copies of the registers.
+	 * By now pioavail updates to memory should have occurred, so
+	 * copy them into our working/shadow registers; this is in
+	 * case something went wrong with abort, but mostly to get the
+	 * initial values of the generation bit correct.
+	 */
+	for (i = 0; i < dd->pioavregs; i++) {
+		__le64 tmp;
+
+		tmp = dd->pioavailregs_dma[i];
+		/*
+		 * Don't need to worry about pioavailkernel here
+		 * because we will call qib_chg_pioavailkernel() later
+		 * in initialization, to busy out buffers as needed.
+		 */
+		dd->pioavailshadow[i] = le64_to_cpu(tmp);
+	}
+	while (i < ARRAY_SIZE(dd->pioavailshadow))
+		dd->pioavailshadow[i++] = 0; /* for debugging sanity */
+
+	/* after pioavailshadow is setup */
+	qib_chg_pioavailkernel(dd, 0, dd->piobcnt2k + dd->piobcnt4k,
+			       TXCHK_CHG_TYPE_KERN, NULL);
+	dd->f_initvl15_bufs(dd);
+}
+
+/**
+ * qib_init - do the actual initialization sequence on the chip
+ * @dd: the qlogic_ib device
+ * @reinit: reinitializing, so don't allocate new memory
+ *
+ * Do the actual initialization sequence on the chip.  This is done
+ * both from the init routine called from the PCI infrastructure, and
+ * when we reset the chip, or detect that it was reset internally,
+ * or it's administratively re-enabled.
+ *
+ * Memory allocation here and in called routines is only done in
+ * the first case (reinit == 0).  We have to be careful, because even
+ * without memory allocation, we need to re-write all the chip registers
+ * TIDs, etc. after the reset or enable has completed.
+ */
+int qib_init(struct qib_devdata *dd, int reinit)
+{
+	int ret = 0, pidx, lastfail = 0;
+	u32 portok = 0;
+	unsigned i;
+	struct qib_ctxtdata *rcd;
+	struct qib_pportdata *ppd;
+	unsigned long flags;
+
+	/* Set linkstate to unknown, so we can watch for a transition. */
+	for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+		ppd = dd->pport + pidx;
+		spin_lock_irqsave(&ppd->lflags_lock, flags);
+		ppd->lflags &= ~(QIBL_LINKACTIVE | QIBL_LINKARMED |
+				 QIBL_LINKDOWN | QIBL_LINKINIT |
+				 QIBL_LINKV);
+		spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+	}
+
+	if (reinit)
+		ret = init_after_reset(dd);
+	else
+		ret = loadtime_init(dd);
+	if (ret)
+		goto done;
+
+	/* Bypass most chip-init, to get to device creation */
+	if (qib_mini_init)
+		return 0;
+
+	ret = dd->f_late_initreg(dd);
+	if (ret)
+		goto done;
+
+	/* dd->rcd can be NULL if early init failed */
+	for (i = 0; dd->rcd && i < dd->first_user_ctxt; ++i) {
+		/*
+		 * Set up the (kernel) rcvhdr queue and egr TIDs.  If doing
+		 * re-init, the simplest way to handle this is to free
+		 * existing, and re-allocate.
+		 * Need to re-create rest of ctxt 0 ctxtdata as well.
+		 */
+		rcd = dd->rcd[i];
+		if (!rcd)
+			continue;
+
+		lastfail = qib_create_rcvhdrq(dd, rcd);
+		if (!lastfail)
+			lastfail = qib_setup_eagerbufs(rcd);
+		if (lastfail) {
+			qib_dev_err(dd, "failed to allocate kernel ctxt's "
+				    "rcvhdrq and/or egr bufs\n");
+			continue;
+		}
+	}
+
+	for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+		int mtu;
+		if (lastfail)
+			ret = lastfail;
+		ppd = dd->pport + pidx;
+		mtu = ib_mtu_enum_to_int(qib_ibmtu);
+		if (mtu == -1) {
+			mtu = QIB_DEFAULT_MTU;
+			qib_ibmtu = 0; /* don't leave invalid value */
+		}
+		/* set max we can ever have for this driver load */
+		ppd->init_ibmaxlen = min(mtu > 2048 ?
+					 dd->piosize4k : dd->piosize2k,
+					 dd->rcvegrbufsize +
+					 (dd->rcvhdrentsize << 2));
+		/*
+		 * Have to initialize ibmaxlen, but this will normally
+		 * change immediately in qib_set_mtu().
+		 */
+		ppd->ibmaxlen = ppd->init_ibmaxlen;
+		qib_set_mtu(ppd, mtu);
+
+		spin_lock_irqsave(&ppd->lflags_lock, flags);
+		ppd->lflags |= QIBL_IB_LINK_DISABLED;
+		spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+
+		lastfail = dd->f_bringup_serdes(ppd);
+		if (lastfail) {
+			qib_devinfo(dd->pcidev,
+				 "Failed to bringup IB port %u\n", ppd->port);
+			lastfail = -ENETDOWN;
+			continue;
+		}
+
+		/* let link come up, and enable IBC */
+		spin_lock_irqsave(&ppd->lflags_lock, flags);
+		ppd->lflags &= ~QIBL_IB_LINK_DISABLED;
+		spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+		portok++;
+	}
+
+	if (!portok) {
+		/* none of the ports initialized */
+		if (!ret && lastfail)
+			ret = lastfail;
+		else if (!ret)
+			ret = -ENETDOWN;
+		/* but continue on, so we can debug cause */
+	}
+
+	enable_chip(dd);
+
+	init_piobuf_state(dd);
+
+done:
+	if (!ret) {
+		/* chip is OK for user apps; mark it as initialized */
+		for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+			ppd = dd->pport + pidx;
+			/*
+			 * Set status even if port serdes is not initialized
+			 * so that diags will work.
+			 */
+			*ppd->statusp |= QIB_STATUS_CHIP_PRESENT |
+				QIB_STATUS_INITTED;
+			if (!ppd->link_speed_enabled)
+				continue;
+			if (dd->flags & QIB_HAS_SEND_DMA)
+				ret = qib_setup_sdma(ppd);
+			init_timer(&ppd->hol_timer);
+			ppd->hol_timer.function = qib_hol_event;
+			ppd->hol_timer.data = (unsigned long)ppd;
+			ppd->hol_state = QIB_HOL_UP;
+		}
+
+		/* now we can enable all interrupts from the chip */
+		dd->f_set_intr_state(dd, 1);
+
+		/*
+		 * Setup to verify we get an interrupt, and fallback
+		 * to an alternate if necessary and possible.
+		 */
+		mod_timer(&dd->intrchk_timer, jiffies + HZ/2);
+		/* start stats retrieval timer */
+		mod_timer(&dd->stats_timer, jiffies + HZ * ACTIVITY_TIMER);
+	}
+
+	/* if ret is non-zero, we probably should do some cleanup here... */
+	return ret;
+}
+
+/*
+ * These next two routines are placeholders in case we don't have per-arch
+ * code for controlling write combining.  If explicit control of write
+ * combining is not available, performance will probably be awful.
+ */
+
+int __attribute__((weak)) qib_enable_wc(struct qib_devdata *dd)
+{
+	return -EOPNOTSUPP;
+}
+
+void __attribute__((weak)) qib_disable_wc(struct qib_devdata *dd)
+{
+}
+
+static inline struct qib_devdata *__qib_lookup(int unit)
+{
+	return idr_find(&qib_unit_table, unit);
+}
+
+struct qib_devdata *qib_lookup(int unit)
+{
+	struct qib_devdata *dd;
+	unsigned long flags;
+
+	spin_lock_irqsave(&qib_devs_lock, flags);
+	dd = __qib_lookup(unit);
+	spin_unlock_irqrestore(&qib_devs_lock, flags);
+
+	return dd;
+}
+
+/*
+ * Stop the timers during unit shutdown, or after an error late
+ * in initialization.
+ */
+static void qib_stop_timers(struct qib_devdata *dd)
+{
+	struct qib_pportdata *ppd;
+	int pidx;
+
+	if (dd->stats_timer.data) {
+		del_timer_sync(&dd->stats_timer);
+		dd->stats_timer.data = 0;
+	}
+	if (dd->intrchk_timer.data) {
+		del_timer_sync(&dd->intrchk_timer);
+		dd->intrchk_timer.data = 0;
+	}
+	for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+		ppd = dd->pport + pidx;
+		if (ppd->hol_timer.data)
+			del_timer_sync(&ppd->hol_timer);
+		if (ppd->led_override_timer.data) {
+			del_timer_sync(&ppd->led_override_timer);
+			atomic_set(&ppd->led_override_timer_active, 0);
+		}
+		if (ppd->symerr_clear_timer.data)
+			del_timer_sync(&ppd->symerr_clear_timer);
+	}
+}
+
+/**
+ * qib_shutdown_device - shut down a device
+ * @dd: the qlogic_ib device
+ *
+ * This is called to make the device quiet when we are about to
+ * unload the driver, and also when the device is administratively
+ * disabled.   It does not free any data structures.
+ * Everything it does has to be setup again by qib_init(dd, 1)
+ */
+static void qib_shutdown_device(struct qib_devdata *dd)
+{
+	struct qib_pportdata *ppd;
+	unsigned pidx;
+
+	for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+		ppd = dd->pport + pidx;
+
+		spin_lock_irq(&ppd->lflags_lock);
+		ppd->lflags &= ~(QIBL_LINKDOWN | QIBL_LINKINIT |
+				 QIBL_LINKARMED | QIBL_LINKACTIVE |
+				 QIBL_LINKV);
+		spin_unlock_irq(&ppd->lflags_lock);
+		*ppd->statusp &= ~(QIB_STATUS_IB_CONF | QIB_STATUS_IB_READY);
+	}
+	dd->flags &= ~QIB_INITTED;
+
+	/* mask interrupts, but not errors */
+	dd->f_set_intr_state(dd, 0);
+
+	for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+		ppd = dd->pport + pidx;
+		dd->f_rcvctrl(ppd, QIB_RCVCTRL_TAILUPD_DIS |
+				   QIB_RCVCTRL_CTXT_DIS |
+				   QIB_RCVCTRL_INTRAVAIL_DIS |
+				   QIB_RCVCTRL_PKEY_ENB, -1);
+		/*
+		 * Gracefully stop all sends allowing any in progress to
+		 * trickle out first.
+		 */
+		dd->f_sendctrl(ppd, QIB_SENDCTRL_CLEAR);
+	}
+
+	/*
+	 * Enough for anything that's going to trickle out to have actually
+	 * done so.
+	 */
+	udelay(20);
+
+	for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+		ppd = dd->pport + pidx;
+		dd->f_setextled(ppd, 0); /* make sure LEDs are off */
+
+		if (dd->flags & QIB_HAS_SEND_DMA)
+			qib_teardown_sdma(ppd);
+
+		dd->f_sendctrl(ppd, QIB_SENDCTRL_AVAIL_DIS |
+				    QIB_SENDCTRL_SEND_DIS);
+		/*
+		 * Clear SerdesEnable.
+		 * We can't count on interrupts since we are stopping.
+		 */
+		dd->f_quiet_serdes(ppd);
+	}
+
+	qib_update_eeprom_log(dd);
+}
+
+/**
+ * qib_free_ctxtdata - free a context's allocated data
+ * @dd: the qlogic_ib device
+ * @rcd: the ctxtdata structure
+ *
+ * free up any allocated data for a context
+ * This should not touch anything that would affect a simultaneous
+ * re-allocation of context data, because it is called after qib_mutex
+ * is released (and can be called from reinit as well).
+ * It should never change any chip state, or global driver state.
+ */
+void qib_free_ctxtdata(struct qib_devdata *dd, struct qib_ctxtdata *rcd)
+{
+	if (!rcd)
+		return;
+
+	if (rcd->rcvhdrq) {
+		dma_free_coherent(&dd->pcidev->dev, rcd->rcvhdrq_size,
+				  rcd->rcvhdrq, rcd->rcvhdrq_phys);
+		rcd->rcvhdrq = NULL;
+		if (rcd->rcvhdrtail_kvaddr) {
+			dma_free_coherent(&dd->pcidev->dev, PAGE_SIZE,
+					  rcd->rcvhdrtail_kvaddr,
+					  rcd->rcvhdrqtailaddr_phys);
+			rcd->rcvhdrtail_kvaddr = NULL;
+		}
+	}
+	if (rcd->rcvegrbuf) {
+		unsigned e;
+
+		for (e = 0; e < rcd->rcvegrbuf_chunks; e++) {
+			void *base = rcd->rcvegrbuf[e];
+			size_t size = rcd->rcvegrbuf_size;
+
+			dma_free_coherent(&dd->pcidev->dev, size,
+					  base, rcd->rcvegrbuf_phys[e]);
+		}
+		kfree(rcd->rcvegrbuf);
+		rcd->rcvegrbuf = NULL;
+		kfree(rcd->rcvegrbuf_phys);
+		rcd->rcvegrbuf_phys = NULL;
+		rcd->rcvegrbuf_chunks = 0;
+	}
+
+	kfree(rcd->tid_pg_list);
+	vfree(rcd->user_event_mask);
+	vfree(rcd->subctxt_uregbase);
+	vfree(rcd->subctxt_rcvegrbuf);
+	vfree(rcd->subctxt_rcvhdr_base);
+	kfree(rcd);
+}
+
+/*
+ * Perform a PIO buffer bandwidth write test, to verify proper system
+ * configuration.  Even when all the setup calls work, occasionally
+ * BIOS or other issues can prevent write combining from working, or
+ * can cause other bandwidth problems to the chip.
+ *
+ * This test simply writes the same buffer over and over again, and
+ * measures close to the peak bandwidth to the chip (not testing
+ * data bandwidth to the wire).   On chips that use an address-based
+ * trigger to send packets to the wire, this is easy.  On chips that
+ * use a count to trigger, we want to make sure that the packet doesn't
+ * go out on the wire, or trigger flow control checks.
+ */
+static void qib_verify_pioperf(struct qib_devdata *dd)
+{
+	u32 pbnum, cnt, lcnt;
+	u32 __iomem *piobuf;
+	u32 *addr;
+	u64 msecs, emsecs;
+
+	piobuf = dd->f_getsendbuf(dd->pport, 0ULL, &pbnum);
+	if (!piobuf) {
+		qib_devinfo(dd->pcidev,
+			 "No PIObufs for checking perf, skipping\n");
+		return;
+	}
+
+	/*
+	 * Enough to give us a reasonable test, less than piobuf size, and
+	 * likely multiple of store buffer length.
+	 */
+	cnt = 1024;
+
+	addr = vmalloc(cnt);
+	if (!addr) {
+		qib_devinfo(dd->pcidev,
+			 "Couldn't get memory for checking PIO perf,"
+			 " skipping\n");
+		goto done;
+	}
+
+	preempt_disable();  /* we want reasonably accurate elapsed time */
+	msecs = 1 + jiffies_to_msecs(jiffies);
+	for (lcnt = 0; lcnt < 10000U; lcnt++) {
+		/* wait until we cross msec boundary */
+		if (jiffies_to_msecs(jiffies) >= msecs)
+			break;
+		udelay(1);
+	}
+
+	dd->f_set_armlaunch(dd, 0);
+
+	/*
+	 * length 0, no dwords actually sent
+	 */
+	writeq(0, piobuf);
+	qib_flush_wc();
+
+	/*
+	 * This is only roughly accurate, since even with preempt we
+	 * still take interrupts that could take a while.   Running for
+	 * >= 5 msec seems to get us "close enough" to accurate values.
+	 */
+	msecs = jiffies_to_msecs(jiffies);
+	for (emsecs = lcnt = 0; emsecs <= 5UL; lcnt++) {
+		qib_pio_copy(piobuf + 64, addr, cnt >> 2);
+		emsecs = jiffies_to_msecs(jiffies) - msecs;
+	}
+
+	/* 1 GiB/sec, slightly over IB SDR line rate */
+	if (lcnt < (emsecs * 1024U))
+		qib_dev_err(dd,
+			    "Performance problem: bandwidth to PIO buffers is "
+			    "only %u MiB/sec\n",
+			    lcnt / (u32) emsecs);
+
+	preempt_enable();
+
+	vfree(addr);
+
+done:
+	/* disarm piobuf, so it's available again */
+	dd->f_sendctrl(dd->pport, QIB_SENDCTRL_DISARM_BUF(pbnum));
+	qib_sendbuf_done(dd, pbnum);
+	dd->f_set_armlaunch(dd, 1);
+}
+
+
+void qib_free_devdata(struct qib_devdata *dd)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&qib_devs_lock, flags);
+	idr_remove(&qib_unit_table, dd->unit);
+	list_del(&dd->list);
+	spin_unlock_irqrestore(&qib_devs_lock, flags);
+
+	ib_dealloc_device(&dd->verbs_dev.ibdev);
+}
+
+/*
+ * Allocate our primary per-unit data structure.  Must be done via verbs
+ * allocator, because the verbs cleanup process both does cleanup and
+ * free of the data structure.
+ * "extra" is for chip-specific data.
+ *
+ * Use the idr mechanism to get a unit number for this unit.
+ */
+struct qib_devdata *qib_alloc_devdata(struct pci_dev *pdev, size_t extra)
+{
+	unsigned long flags;
+	struct qib_devdata *dd;
+	int ret;
+
+	if (!idr_pre_get(&qib_unit_table, GFP_KERNEL)) {
+		dd = ERR_PTR(-ENOMEM);
+		goto bail;
+	}
+
+	dd = (struct qib_devdata *) ib_alloc_device(sizeof(*dd) + extra);
+	if (!dd) {
+		dd = ERR_PTR(-ENOMEM);
+		goto bail;
+	}
+
+	spin_lock_irqsave(&qib_devs_lock, flags);
+	ret = idr_get_new(&qib_unit_table, dd, &dd->unit);
+	if (ret >= 0)
+		list_add(&dd->list, &qib_dev_list);
+	spin_unlock_irqrestore(&qib_devs_lock, flags);
+
+	if (ret < 0) {
+		qib_early_err(&pdev->dev,
+			      "Could not allocate unit ID: error %d\n", -ret);
+		ib_dealloc_device(&dd->verbs_dev.ibdev);
+		dd = ERR_PTR(ret);
+		goto bail;
+	}
+
+	if (!qib_cpulist_count) {
+		u32 count = num_online_cpus();
+		qib_cpulist = kzalloc(BITS_TO_LONGS(count) *
+				      sizeof(long), GFP_KERNEL);
+		if (qib_cpulist)
+			qib_cpulist_count = count;
+		else
+			qib_early_err(&pdev->dev, "Could not alloc cpulist "
+				      "info, cpu affinity might be wrong\n");
+	}
+
+bail:
+	return dd;
+}
+
+/*
+ * Called from freeze mode handlers, and from PCI error
+ * reporting code.  Should be paranoid about state of
+ * system and data structures.
+ */
+void qib_disable_after_error(struct qib_devdata *dd)
+{
+	if (dd->flags & QIB_INITTED) {
+		u32 pidx;
+
+		dd->flags &= ~QIB_INITTED;
+		if (dd->pport)
+			for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+				struct qib_pportdata *ppd;
+
+				ppd = dd->pport + pidx;
+				if (dd->flags & QIB_PRESENT) {
+					qib_set_linkstate(ppd,
+						QIB_IB_LINKDOWN_DISABLE);
+					dd->f_setextled(ppd, 0);
+				}
+				*ppd->statusp &= ~QIB_STATUS_IB_READY;
+			}
+	}
+
+	/*
+	 * Mark as having had an error for driver, and also
+	 * for /sys and status word mapped to user programs.
+	 * This marks unit as not usable, until reset.
+	 */
+	if (dd->devstatusp)
+		*dd->devstatusp |= QIB_STATUS_HWERROR;
+}
+
+static void __devexit qib_remove_one(struct pci_dev *);
+static int __devinit qib_init_one(struct pci_dev *,
+				  const struct pci_device_id *);
+
+#define DRIVER_LOAD_MSG "QLogic " QIB_DRV_NAME " loaded: "
+#define PFX QIB_DRV_NAME ": "
+
+static const struct pci_device_id qib_pci_tbl[] = {
+	{ PCI_DEVICE(PCI_VENDOR_ID_PATHSCALE, PCI_DEVICE_ID_QLOGIC_IB_6120) },
+	{ PCI_DEVICE(PCI_VENDOR_ID_QLOGIC, PCI_DEVICE_ID_QLOGIC_IB_7220) },
+	{ PCI_DEVICE(PCI_VENDOR_ID_QLOGIC, PCI_DEVICE_ID_QLOGIC_IB_7322) },
+	{ 0, }
+};
+
+MODULE_DEVICE_TABLE(pci, qib_pci_tbl);
+
+struct pci_driver qib_driver = {
+	.name = QIB_DRV_NAME,
+	.probe = qib_init_one,
+	.remove = __devexit_p(qib_remove_one),
+	.id_table = qib_pci_tbl,
+	.err_handler = &qib_pci_err_handler,
+};
+
+/*
+ * Do all the generic driver unit- and chip-independent memory
+ * allocation and initialization.
+ */
+static int __init qlogic_ib_init(void)
+{
+	int ret;
+
+	ret = qib_dev_init();
+	if (ret)
+		goto bail;
+
+	/*
+	 * We create our own workqueue mainly because we want to be
+	 * able to flush it when devices are being removed.  We can't
+	 * use schedule_work()/flush_scheduled_work() because both
+	 * unregister_netdev() and linkwatch_event take the rtnl lock,
+	 * so flush_scheduled_work() can deadlock during device
+	 * removal.
+	 */
+	qib_wq = create_workqueue("qib");
+	if (!qib_wq) {
+		ret = -ENOMEM;
+		goto bail_dev;
+	}
+
+	qib_cq_wq = create_workqueue("qib_cq");
+	if (!qib_cq_wq) {
+		ret = -ENOMEM;
+		goto bail_wq;
+	}
+
+	/*
+	 * These must be called before the driver is registered with
+	 * the PCI subsystem.
+	 */
+	idr_init(&qib_unit_table);
+	if (!idr_pre_get(&qib_unit_table, GFP_KERNEL)) {
+		printk(KERN_ERR QIB_DRV_NAME ": idr_pre_get() failed\n");
+		ret = -ENOMEM;
+		goto bail_cq_wq;
+	}
+
+	ret = pci_register_driver(&qib_driver);
+	if (ret < 0) {
+		printk(KERN_ERR QIB_DRV_NAME
+		       ": Unable to register driver: error %d\n", -ret);
+		goto bail_unit;
+	}
+
+	/* not fatal if it doesn't work */
+	if (qib_init_qibfs())
+		printk(KERN_ERR QIB_DRV_NAME ": Unable to register ipathfs\n");
+	goto bail; /* all OK */
+
+bail_unit:
+	idr_destroy(&qib_unit_table);
+bail_cq_wq:
+	destroy_workqueue(qib_cq_wq);
+bail_wq:
+	destroy_workqueue(qib_wq);
+bail_dev:
+	qib_dev_cleanup();
+bail:
+	return ret;
+}
+
+module_init(qlogic_ib_init);
+
+/*
+ * Do the non-unit driver cleanup, memory free, etc. at unload.
+ */
+static void __exit qlogic_ib_cleanup(void)
+{
+	int ret;
+
+	ret = qib_exit_qibfs();
+	if (ret)
+		printk(KERN_ERR QIB_DRV_NAME ": "
+			"Unable to cleanup counter filesystem: "
+			"error %d\n", -ret);
+
+	pci_unregister_driver(&qib_driver);
+
+	destroy_workqueue(qib_wq);
+	destroy_workqueue(qib_cq_wq);
+
+	qib_cpulist_count = 0;
+	kfree(qib_cpulist);
+
+	idr_destroy(&qib_unit_table);
+	qib_dev_cleanup();
+}
+
+module_exit(qlogic_ib_cleanup);
+
+/* this can only be called after a successful initialization */
+static void cleanup_device_data(struct qib_devdata *dd)
+{
+	int ctxt;
+	int pidx;
+	struct qib_ctxtdata **tmp;
+	unsigned long flags;
+
+	/* users can't do anything more with chip */
+	for (pidx = 0; pidx < dd->num_pports; ++pidx)
+		if (dd->pport[pidx].statusp)
+			*dd->pport[pidx].statusp &= ~QIB_STATUS_CHIP_PRESENT;
+
+	if (!qib_wc_pat)
+		qib_disable_wc(dd);
+
+	if (dd->pioavailregs_dma) {
+		dma_free_coherent(&dd->pcidev->dev, PAGE_SIZE,
+				  (void *) dd->pioavailregs_dma,
+				  dd->pioavailregs_phys);
+		dd->pioavailregs_dma = NULL;
+	}
+
+	if (dd->pageshadow) {
+		struct page **tmpp = dd->pageshadow;
+		dma_addr_t *tmpd = dd->physshadow;
+		int i, cnt = 0;
+
+		for (ctxt = 0; ctxt < dd->cfgctxts; ctxt++) {
+			int ctxt_tidbase = ctxt * dd->rcvtidcnt;
+			int maxtid = ctxt_tidbase + dd->rcvtidcnt;
+
+			for (i = ctxt_tidbase; i < maxtid; i++) {
+				if (!tmpp[i])
+					continue;
+				pci_unmap_page(dd->pcidev, tmpd[i],
+					       PAGE_SIZE, PCI_DMA_FROMDEVICE);
+				qib_release_user_pages(&tmpp[i], 1);
+				tmpp[i] = NULL;
+				cnt++;
+			}
+		}
+
+		tmpp = dd->pageshadow;
+		dd->pageshadow = NULL;
+		vfree(tmpp);
+	}
+
+	/*
+	 * Free any resources still in use (usually just kernel contexts)
+	 * at unload; we do for ctxtcnt, because that's what we allocate.
+	 * We acquire lock to be really paranoid that rcd isn't being
+	 * accessed from some interrupt-related code (that should not happen,
+	 * but best to be sure).
+	 */
+	spin_lock_irqsave(&dd->uctxt_lock, flags);
+	tmp = dd->rcd;
+	dd->rcd = NULL;
+	spin_unlock_irqrestore(&dd->uctxt_lock, flags);
+	for (ctxt = 0; tmp && ctxt < dd->ctxtcnt; ctxt++) {
+		struct qib_ctxtdata *rcd = tmp[ctxt];
+
+		tmp[ctxt] = NULL; /* debugging paranoia */
+		qib_free_ctxtdata(dd, rcd);
+	}
+	kfree(tmp);
+	kfree(dd->boardname);
+}
+
+/*
+ * Clean up on unit shutdown, or error during unit load after
+ * successful initialization.
+ */
+static void qib_postinit_cleanup(struct qib_devdata *dd)
+{
+	/*
+	 * Clean up chip-specific stuff.
+	 * We check for NULL here, because it's outside
+	 * the kregbase check, and we need to call it
+	 * after the free_irq.  Thus it's possible that
+	 * the function pointers were never initialized.
+	 */
+	if (dd->f_cleanup)
+		dd->f_cleanup(dd);
+
+	qib_pcie_ddcleanup(dd);
+
+	cleanup_device_data(dd);
+
+	qib_free_devdata(dd);
+}
+
+static int __devinit qib_init_one(struct pci_dev *pdev,
+				  const struct pci_device_id *ent)
+{
+	int ret, j, pidx, initfail;
+	struct qib_devdata *dd = NULL;
+
+	ret = qib_pcie_init(pdev, ent);
+	if (ret)
+		goto bail;
+
+	/*
+	 * Do device-specific initialiation, function table setup, dd
+	 * allocation, etc.
+	 */
+	switch (ent->device) {
+	case PCI_DEVICE_ID_QLOGIC_IB_6120:
+#ifdef CONFIG_PCI_MSI
+		dd = qib_init_iba6120_funcs(pdev, ent);
+#else
+		qib_early_err(&pdev->dev, "QLogic PCIE device 0x%x cannot "
+		      "work if CONFIG_PCI_MSI is not enabled\n",
+		      ent->device);
+#endif
+		break;
+
+	case PCI_DEVICE_ID_QLOGIC_IB_7220:
+		dd = qib_init_iba7220_funcs(pdev, ent);
+		break;
+
+	case PCI_DEVICE_ID_QLOGIC_IB_7322:
+		dd = qib_init_iba7322_funcs(pdev, ent);
+		break;
+
+	default:
+		qib_early_err(&pdev->dev, "Failing on unknown QLogic "
+			      "deviceid 0x%x\n", ent->device);
+		ret = -ENODEV;
+	}
+
+	if (IS_ERR(dd))
+		ret = PTR_ERR(dd);
+	if (ret)
+		goto bail; /* error already printed */
+
+	/* do the generic initialization */
+	initfail = qib_init(dd, 0);
+
+	ret = qib_register_ib_device(dd);
+
+	/*
+	 * Now ready for use.  this should be cleared whenever we
+	 * detect a reset, or initiate one.  If earlier failure,
+	 * we still create devices, so diags, etc. can be used
+	 * to determine cause of problem.
+	 */
+	if (!qib_mini_init && !initfail && !ret)
+		dd->flags |= QIB_INITTED;
+
+	j = qib_device_create(dd);
+	if (j)
+		qib_dev_err(dd, "Failed to create /dev devices: %d\n", -j);
+	j = qibfs_add(dd);
+	if (j)
+		qib_dev_err(dd, "Failed filesystem setup for counters: %d\n",
+			    -j);
+
+	if (qib_mini_init || initfail || ret) {
+		qib_stop_timers(dd);
+		for (pidx = 0; pidx < dd->num_pports; ++pidx)
+			dd->f_quiet_serdes(dd->pport + pidx);
+		if (initfail)
+			ret = initfail;
+		goto bail;
+	}
+
+	if (!qib_wc_pat) {
+		ret = qib_enable_wc(dd);
+		if (ret) {
+			qib_dev_err(dd, "Write combining not enabled "
+				    "(err %d): performance may be poor\n",
+				    -ret);
+			ret = 0;
+		}
+	}
+
+	qib_verify_pioperf(dd);
+bail:
+	return ret;
+}
+
+static void __devexit qib_remove_one(struct pci_dev *pdev)
+{
+	struct qib_devdata *dd = pci_get_drvdata(pdev);
+	int ret;
+
+	/* unregister from IB core */
+	qib_unregister_ib_device(dd);
+
+	/*
+	 * Disable the IB link, disable interrupts on the device,
+	 * clear dma engines, etc.
+	 */
+	if (!qib_mini_init)
+		qib_shutdown_device(dd);
+
+	qib_stop_timers(dd);
+
+	/* wait until all of our (qsfp) schedule_work() calls complete */
+	flush_scheduled_work();
+
+	ret = qibfs_remove(dd);
+	if (ret)
+		qib_dev_err(dd, "Failed counters filesystem cleanup: %d\n",
+			    -ret);
+
+	qib_device_remove(dd);
+
+	qib_postinit_cleanup(dd);
+}
+
+/**
+ * qib_create_rcvhdrq - create a receive header queue
+ * @dd: the qlogic_ib device
+ * @rcd: the context data
+ *
+ * This must be contiguous memory (from an i/o perspective), and must be
+ * DMA'able (which means for some systems, it will go through an IOMMU,
+ * or be forced into a low address range).
+ */
+int qib_create_rcvhdrq(struct qib_devdata *dd, struct qib_ctxtdata *rcd)
+{
+	unsigned amt;
+
+	if (!rcd->rcvhdrq) {
+		dma_addr_t phys_hdrqtail;
+		gfp_t gfp_flags;
+
+		amt = ALIGN(dd->rcvhdrcnt * dd->rcvhdrentsize *
+			    sizeof(u32), PAGE_SIZE);
+		gfp_flags = (rcd->ctxt >= dd->first_user_ctxt) ?
+			GFP_USER : GFP_KERNEL;
+		rcd->rcvhdrq = dma_alloc_coherent(
+			&dd->pcidev->dev, amt, &rcd->rcvhdrq_phys,
+			gfp_flags | __GFP_COMP);
+
+		if (!rcd->rcvhdrq) {
+			qib_dev_err(dd, "attempt to allocate %d bytes "
+				    "for ctxt %u rcvhdrq failed\n",
+				    amt, rcd->ctxt);
+			goto bail;
+		}
+
+		if (rcd->ctxt >= dd->first_user_ctxt) {
+			rcd->user_event_mask = vmalloc_user(PAGE_SIZE);
+			if (!rcd->user_event_mask)
+				goto bail_free_hdrq;
+		}
+
+		if (!(dd->flags & QIB_NODMA_RTAIL)) {
+			rcd->rcvhdrtail_kvaddr = dma_alloc_coherent(
+				&dd->pcidev->dev, PAGE_SIZE, &phys_hdrqtail,
+				gfp_flags);
+			if (!rcd->rcvhdrtail_kvaddr)
+				goto bail_free;
+			rcd->rcvhdrqtailaddr_phys = phys_hdrqtail;
+		}
+
+		rcd->rcvhdrq_size = amt;
+	}
+
+	/* clear for security and sanity on each use */
+	memset(rcd->rcvhdrq, 0, rcd->rcvhdrq_size);
+	if (rcd->rcvhdrtail_kvaddr)
+		memset(rcd->rcvhdrtail_kvaddr, 0, PAGE_SIZE);
+	return 0;
+
+bail_free:
+	qib_dev_err(dd, "attempt to allocate 1 page for ctxt %u "
+		    "rcvhdrqtailaddr failed\n", rcd->ctxt);
+	vfree(rcd->user_event_mask);
+	rcd->user_event_mask = NULL;
+bail_free_hdrq:
+	dma_free_coherent(&dd->pcidev->dev, amt, rcd->rcvhdrq,
+			  rcd->rcvhdrq_phys);
+	rcd->rcvhdrq = NULL;
+bail:
+	return -ENOMEM;
+}
+
+/**
+ * allocate eager buffers, both kernel and user contexts.
+ * @rcd: the context we are setting up.
+ *
+ * Allocate the eager TID buffers and program them into hip.
+ * They are no longer completely contiguous, we do multiple allocation
+ * calls.  Otherwise we get the OOM code involved, by asking for too
+ * much per call, with disastrous results on some kernels.
+ */
+int qib_setup_eagerbufs(struct qib_ctxtdata *rcd)
+{
+	struct qib_devdata *dd = rcd->dd;
+	unsigned e, egrcnt, egrperchunk, chunk, egrsize, egroff;
+	size_t size;
+	gfp_t gfp_flags;
+
+	/*
+	 * GFP_USER, but without GFP_FS, so buffer cache can be
+	 * coalesced (we hope); otherwise, even at order 4,
+	 * heavy filesystem activity makes these fail, and we can
+	 * use compound pages.
+	 */
+	gfp_flags = __GFP_WAIT | __GFP_IO | __GFP_COMP;
+
+	egrcnt = rcd->rcvegrcnt;
+	egroff = rcd->rcvegr_tid_base;
+	egrsize = dd->rcvegrbufsize;
+
+	chunk = rcd->rcvegrbuf_chunks;
+	egrperchunk = rcd->rcvegrbufs_perchunk;
+	size = rcd->rcvegrbuf_size;
+	if (!rcd->rcvegrbuf) {
+		rcd->rcvegrbuf =
+			kzalloc(chunk * sizeof(rcd->rcvegrbuf[0]),
+				GFP_KERNEL);
+		if (!rcd->rcvegrbuf)
+			goto bail;
+	}
+	if (!rcd->rcvegrbuf_phys) {
+		rcd->rcvegrbuf_phys =
+			kmalloc(chunk * sizeof(rcd->rcvegrbuf_phys[0]),
+				GFP_KERNEL);
+		if (!rcd->rcvegrbuf_phys)
+			goto bail_rcvegrbuf;
+	}
+	for (e = 0; e < rcd->rcvegrbuf_chunks; e++) {
+		if (rcd->rcvegrbuf[e])
+			continue;
+		rcd->rcvegrbuf[e] =
+			dma_alloc_coherent(&dd->pcidev->dev, size,
+					   &rcd->rcvegrbuf_phys[e],
+					   gfp_flags);
+		if (!rcd->rcvegrbuf[e])
+			goto bail_rcvegrbuf_phys;
+	}
+
+	rcd->rcvegr_phys = rcd->rcvegrbuf_phys[0];
+
+	for (e = chunk = 0; chunk < rcd->rcvegrbuf_chunks; chunk++) {
+		dma_addr_t pa = rcd->rcvegrbuf_phys[chunk];
+		unsigned i;
+
+		for (i = 0; e < egrcnt && i < egrperchunk; e++, i++) {
+			dd->f_put_tid(dd, e + egroff +
+					  (u64 __iomem *)
+					  ((char __iomem *)
+					   dd->kregbase +
+					   dd->rcvegrbase),
+					  RCVHQ_RCV_TYPE_EAGER, pa);
+			pa += egrsize;
+		}
+		cond_resched(); /* don't hog the cpu */
+	}
+
+	return 0;
+
+bail_rcvegrbuf_phys:
+	for (e = 0; e < rcd->rcvegrbuf_chunks && rcd->rcvegrbuf[e]; e++)
+		dma_free_coherent(&dd->pcidev->dev, size,
+				  rcd->rcvegrbuf[e], rcd->rcvegrbuf_phys[e]);
+	kfree(rcd->rcvegrbuf_phys);
+	rcd->rcvegrbuf_phys = NULL;
+bail_rcvegrbuf:
+	kfree(rcd->rcvegrbuf);
+	rcd->rcvegrbuf = NULL;
+bail:
+	return -ENOMEM;
+}
+
+int init_chip_wc_pat(struct qib_devdata *dd, u32 vl15buflen)
+{
+	u64 __iomem *qib_kregbase = NULL;
+	void __iomem *qib_piobase = NULL;
+	u64 __iomem *qib_userbase = NULL;
+	u64 qib_kreglen;
+	u64 qib_pio2koffset = dd->piobufbase & 0xffffffff;
+	u64 qib_pio4koffset = dd->piobufbase >> 32;
+	u64 qib_pio2klen = dd->piobcnt2k * dd->palign;
+	u64 qib_pio4klen = dd->piobcnt4k * dd->align4k;
+	u64 qib_physaddr = dd->physaddr;
+	u64 qib_piolen;
+	u64 qib_userlen = 0;
+
+	/*
+	 * Free the old mapping because the kernel will try to reuse the
+	 * old mapping and not create a new mapping with the
+	 * write combining attribute.
+	 */
+	iounmap(dd->kregbase);
+	dd->kregbase = NULL;
+
+	/*
+	 * Assumes chip address space looks like:
+	 *	- kregs + sregs + cregs + uregs (in any order)
+	 *	- piobufs (2K and 4K bufs in either order)
+	 * or:
+	 *	- kregs + sregs + cregs (in any order)
+	 *	- piobufs (2K and 4K bufs in either order)
+	 *	- uregs
+	 */
+	if (dd->piobcnt4k == 0) {
+		qib_kreglen = qib_pio2koffset;
+		qib_piolen = qib_pio2klen;
+	} else if (qib_pio2koffset < qib_pio4koffset) {
+		qib_kreglen = qib_pio2koffset;
+		qib_piolen = qib_pio4koffset + qib_pio4klen - qib_kreglen;
+	} else {
+		qib_kreglen = qib_pio4koffset;
+		qib_piolen = qib_pio2koffset + qib_pio2klen - qib_kreglen;
+	}
+	qib_piolen += vl15buflen;
+	/* Map just the configured ports (not all hw ports) */
+	if (dd->uregbase > qib_kreglen)
+		qib_userlen = dd->ureg_align * dd->cfgctxts;
+
+	/* Sanity checks passed, now create the new mappings */
+	qib_kregbase = ioremap_nocache(qib_physaddr, qib_kreglen);
+	if (!qib_kregbase)
+		goto bail;
+
+	qib_piobase = ioremap_wc(qib_physaddr + qib_kreglen, qib_piolen);
+	if (!qib_piobase)
+		goto bail_kregbase;
+
+	if (qib_userlen) {
+		qib_userbase = ioremap_nocache(qib_physaddr + dd->uregbase,
+					       qib_userlen);
+		if (!qib_userbase)
+			goto bail_piobase;
+	}
+
+	dd->kregbase = qib_kregbase;
+	dd->kregend = (u64 __iomem *)
+		((char __iomem *) qib_kregbase + qib_kreglen);
+	dd->piobase = qib_piobase;
+	dd->pio2kbase = (void __iomem *)
+		(((char __iomem *) dd->piobase) +
+		 qib_pio2koffset - qib_kreglen);
+	if (dd->piobcnt4k)
+		dd->pio4kbase = (void __iomem *)
+			(((char __iomem *) dd->piobase) +
+			 qib_pio4koffset - qib_kreglen);
+	if (qib_userlen)
+		/* ureg will now be accessed relative to dd->userbase */
+		dd->userbase = qib_userbase;
+	return 0;
+
+bail_piobase:
+	iounmap(qib_piobase);
+bail_kregbase:
+	iounmap(qib_kregbase);
+bail:
+	return -ENOMEM;
+}
diff --git a/drivers/infiniband/hw/qib/qib_intr.c b/drivers/infiniband/hw/qib/qib_intr.c
new file mode 100644
index 000000000000..54a40828a106
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_intr.c
@@ -0,0 +1,236 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009, 2010 QLogic Corporation.
+ * All rights reserved.
+ * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/pci.h>
+#include <linux/delay.h>
+
+#include "qib.h"
+#include "qib_common.h"
+
+/**
+ * qib_format_hwmsg - format a single hwerror message
+ * @msg message buffer
+ * @msgl length of message buffer
+ * @hwmsg message to add to message buffer
+ */
+static void qib_format_hwmsg(char *msg, size_t msgl, const char *hwmsg)
+{
+	strlcat(msg, "[", msgl);
+	strlcat(msg, hwmsg, msgl);
+	strlcat(msg, "]", msgl);
+}
+
+/**
+ * qib_format_hwerrors - format hardware error messages for display
+ * @hwerrs hardware errors bit vector
+ * @hwerrmsgs hardware error descriptions
+ * @nhwerrmsgs number of hwerrmsgs
+ * @msg message buffer
+ * @msgl message buffer length
+ */
+void qib_format_hwerrors(u64 hwerrs, const struct qib_hwerror_msgs *hwerrmsgs,
+			 size_t nhwerrmsgs, char *msg, size_t msgl)
+{
+	int i;
+
+	for (i = 0; i < nhwerrmsgs; i++)
+		if (hwerrs & hwerrmsgs[i].mask)
+			qib_format_hwmsg(msg, msgl, hwerrmsgs[i].msg);
+}
+
+static void signal_ib_event(struct qib_pportdata *ppd, enum ib_event_type ev)
+{
+	struct ib_event event;
+	struct qib_devdata *dd = ppd->dd;
+
+	event.device = &dd->verbs_dev.ibdev;
+	event.element.port_num = ppd->port;
+	event.event = ev;
+	ib_dispatch_event(&event);
+}
+
+void qib_handle_e_ibstatuschanged(struct qib_pportdata *ppd, u64 ibcs)
+{
+	struct qib_devdata *dd = ppd->dd;
+	unsigned long flags;
+	u32 lstate;
+	u8 ltstate;
+	enum ib_event_type ev = 0;
+
+	lstate = dd->f_iblink_state(ibcs); /* linkstate */
+	ltstate = dd->f_ibphys_portstate(ibcs);
+
+	/*
+	 * If linkstate transitions into INIT from any of the various down
+	 * states, or if it transitions from any of the up (INIT or better)
+	 * states into any of the down states (except link recovery), then
+	 * call the chip-specific code to take appropriate actions.
+	 */
+	if (lstate >= IB_PORT_INIT && (ppd->lflags & QIBL_LINKDOWN) &&
+	    ltstate == IB_PHYSPORTSTATE_LINKUP) {
+		/* transitioned to UP */
+		if (dd->f_ib_updown(ppd, 1, ibcs))
+			goto skip_ibchange; /* chip-code handled */
+	} else if (ppd->lflags & (QIBL_LINKINIT | QIBL_LINKARMED |
+		   QIBL_LINKACTIVE | QIBL_IB_FORCE_NOTIFY)) {
+		if (ltstate != IB_PHYSPORTSTATE_LINKUP &&
+		    ltstate <= IB_PHYSPORTSTATE_CFG_TRAIN &&
+		    dd->f_ib_updown(ppd, 0, ibcs))
+			goto skip_ibchange; /* chip-code handled */
+		qib_set_uevent_bits(ppd, _QIB_EVENT_LINKDOWN_BIT);
+	}
+
+	if (lstate != IB_PORT_DOWN) {
+		/* lstate is INIT, ARMED, or ACTIVE */
+		if (lstate != IB_PORT_ACTIVE) {
+			*ppd->statusp &= ~QIB_STATUS_IB_READY;
+			if (ppd->lflags & QIBL_LINKACTIVE)
+				ev = IB_EVENT_PORT_ERR;
+			spin_lock_irqsave(&ppd->lflags_lock, flags);
+			if (lstate == IB_PORT_ARMED) {
+				ppd->lflags |= QIBL_LINKARMED | QIBL_LINKV;
+				ppd->lflags &= ~(QIBL_LINKINIT |
+					QIBL_LINKDOWN | QIBL_LINKACTIVE);
+			} else {
+				ppd->lflags |= QIBL_LINKINIT | QIBL_LINKV;
+				ppd->lflags &= ~(QIBL_LINKARMED |
+					QIBL_LINKDOWN | QIBL_LINKACTIVE);
+			}
+			spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+			/* start a 75msec timer to clear symbol errors */
+			mod_timer(&ppd->symerr_clear_timer,
+				  msecs_to_jiffies(75));
+		} else if (ltstate == IB_PHYSPORTSTATE_LINKUP) {
+			/* active, but not active defered */
+			qib_hol_up(ppd); /* useful only for 6120 now */
+			*ppd->statusp |=
+				QIB_STATUS_IB_READY | QIB_STATUS_IB_CONF;
+			qib_clear_symerror_on_linkup((unsigned long)ppd);
+			spin_lock_irqsave(&ppd->lflags_lock, flags);
+			ppd->lflags |= QIBL_LINKACTIVE | QIBL_LINKV;
+			ppd->lflags &= ~(QIBL_LINKINIT |
+				QIBL_LINKDOWN | QIBL_LINKARMED);
+			spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+			if (dd->flags & QIB_HAS_SEND_DMA)
+				qib_sdma_process_event(ppd,
+					qib_sdma_event_e30_go_running);
+			ev = IB_EVENT_PORT_ACTIVE;
+			dd->f_setextled(ppd, 1);
+		}
+	} else { /* down */
+		if (ppd->lflags & QIBL_LINKACTIVE)
+			ev = IB_EVENT_PORT_ERR;
+		spin_lock_irqsave(&ppd->lflags_lock, flags);
+		ppd->lflags |= QIBL_LINKDOWN | QIBL_LINKV;
+		ppd->lflags &= ~(QIBL_LINKINIT |
+				 QIBL_LINKACTIVE | QIBL_LINKARMED);
+		spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+		*ppd->statusp &= ~QIB_STATUS_IB_READY;
+	}
+
+skip_ibchange:
+	ppd->lastibcstat = ibcs;
+	if (ev)
+		signal_ib_event(ppd, ev);
+	return;
+}
+
+void qib_clear_symerror_on_linkup(unsigned long opaque)
+{
+	struct qib_pportdata *ppd = (struct qib_pportdata *)opaque;
+
+	if (ppd->lflags & QIBL_LINKACTIVE)
+		return;
+
+	ppd->ibport_data.z_symbol_error_counter =
+		ppd->dd->f_portcntr(ppd, QIBPORTCNTR_IBSYMBOLERR);
+}
+
+/*
+ * Handle receive interrupts for user ctxts; this means a user
+ * process was waiting for a packet to arrive, and didn't want
+ * to poll.
+ */
+void qib_handle_urcv(struct qib_devdata *dd, u64 ctxtr)
+{
+	struct qib_ctxtdata *rcd;
+	unsigned long flags;
+	int i;
+
+	spin_lock_irqsave(&dd->uctxt_lock, flags);
+	for (i = dd->first_user_ctxt; dd->rcd && i < dd->cfgctxts; i++) {
+		if (!(ctxtr & (1ULL << i)))
+			continue;
+		rcd = dd->rcd[i];
+		if (!rcd || !rcd->cnt)
+			continue;
+
+		if (test_and_clear_bit(QIB_CTXT_WAITING_RCV, &rcd->flag)) {
+			wake_up_interruptible(&rcd->wait);
+			dd->f_rcvctrl(rcd->ppd, QIB_RCVCTRL_INTRAVAIL_DIS,
+				      rcd->ctxt);
+		} else if (test_and_clear_bit(QIB_CTXT_WAITING_URG,
+					      &rcd->flag)) {
+			rcd->urgent++;
+			wake_up_interruptible(&rcd->wait);
+		}
+	}
+	spin_unlock_irqrestore(&dd->uctxt_lock, flags);
+}
+
+void qib_bad_intrstatus(struct qib_devdata *dd)
+{
+	static int allbits;
+
+	/* separate routine, for better optimization of qib_intr() */
+
+	/*
+	 * We print the message and disable interrupts, in hope of
+	 * having a better chance of debugging the problem.
+	 */
+	qib_dev_err(dd, "Read of chip interrupt status failed"
+		    " disabling interrupts\n");
+	if (allbits++) {
+		/* disable interrupt delivery, something is very wrong */
+		if (allbits == 2)
+			dd->f_set_intr_state(dd, 0);
+		if (allbits == 3) {
+			qib_dev_err(dd, "2nd bad interrupt status, "
+				    "unregistering interrupts\n");
+			dd->flags |= QIB_BADINTR;
+			dd->flags &= ~QIB_INITTED;
+			dd->f_free_irq(dd);
+		}
+	}
+}
diff --git a/drivers/infiniband/hw/qib/qib_keys.c b/drivers/infiniband/hw/qib/qib_keys.c
new file mode 100644
index 000000000000..4b80eb153d57
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_keys.c
@@ -0,0 +1,328 @@
+/*
+ * Copyright (c) 2006, 2007, 2009 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 "qib.h"
+
+/**
+ * qib_alloc_lkey - allocate an lkey
+ * @rkt: lkey table in which to allocate the lkey
+ * @mr: memory region that this lkey protects
+ *
+ * Returns 1 if successful, otherwise returns 0.
+ */
+
+int qib_alloc_lkey(struct qib_lkey_table *rkt, struct qib_mregion *mr)
+{
+	unsigned long flags;
+	u32 r;
+	u32 n;
+	int ret;
+
+	spin_lock_irqsave(&rkt->lock, flags);
+
+	/* Find the next available LKEY */
+	r = rkt->next;
+	n = r;
+	for (;;) {
+		if (rkt->table[r] == NULL)
+			break;
+		r = (r + 1) & (rkt->max - 1);
+		if (r == n) {
+			spin_unlock_irqrestore(&rkt->lock, flags);
+			ret = 0;
+			goto bail;
+		}
+	}
+	rkt->next = (r + 1) & (rkt->max - 1);
+	/*
+	 * Make sure lkey is never zero which is reserved to indicate an
+	 * unrestricted LKEY.
+	 */
+	rkt->gen++;
+	mr->lkey = (r << (32 - ib_qib_lkey_table_size)) |
+		((((1 << (24 - ib_qib_lkey_table_size)) - 1) & rkt->gen)
+		 << 8);
+	if (mr->lkey == 0) {
+		mr->lkey |= 1 << 8;
+		rkt->gen++;
+	}
+	rkt->table[r] = mr;
+	spin_unlock_irqrestore(&rkt->lock, flags);
+
+	ret = 1;
+
+bail:
+	return ret;
+}
+
+/**
+ * qib_free_lkey - free an lkey
+ * @rkt: table from which to free the lkey
+ * @lkey: lkey id to free
+ */
+int qib_free_lkey(struct qib_ibdev *dev, struct qib_mregion *mr)
+{
+	unsigned long flags;
+	u32 lkey = mr->lkey;
+	u32 r;
+	int ret;
+
+	spin_lock_irqsave(&dev->lk_table.lock, flags);
+	if (lkey == 0) {
+		if (dev->dma_mr && dev->dma_mr == mr) {
+			ret = atomic_read(&dev->dma_mr->refcount);
+			if (!ret)
+				dev->dma_mr = NULL;
+		} else
+			ret = 0;
+	} else {
+		r = lkey >> (32 - ib_qib_lkey_table_size);
+		ret = atomic_read(&dev->lk_table.table[r]->refcount);
+		if (!ret)
+			dev->lk_table.table[r] = NULL;
+	}
+	spin_unlock_irqrestore(&dev->lk_table.lock, flags);
+
+	if (ret)
+		ret = -EBUSY;
+	return ret;
+}
+
+/**
+ * qib_lkey_ok - check IB SGE for validity and initialize
+ * @rkt: table containing lkey to check SGE against
+ * @isge: outgoing internal SGE
+ * @sge: SGE to check
+ * @acc: access flags
+ *
+ * Return 1 if valid and successful, otherwise returns 0.
+ *
+ * Check the IB SGE for validity and initialize our internal version
+ * of it.
+ */
+int qib_lkey_ok(struct qib_lkey_table *rkt, struct qib_pd *pd,
+		struct qib_sge *isge, struct ib_sge *sge, int acc)
+{
+	struct qib_mregion *mr;
+	unsigned n, m;
+	size_t off;
+	int ret = 0;
+	unsigned long flags;
+
+	/*
+	 * We use LKEY == zero for kernel virtual addresses
+	 * (see qib_get_dma_mr and qib_dma.c).
+	 */
+	spin_lock_irqsave(&rkt->lock, flags);
+	if (sge->lkey == 0) {
+		struct qib_ibdev *dev = to_idev(pd->ibpd.device);
+
+		if (pd->user)
+			goto bail;
+		if (!dev->dma_mr)
+			goto bail;
+		atomic_inc(&dev->dma_mr->refcount);
+		isge->mr = dev->dma_mr;
+		isge->vaddr = (void *) sge->addr;
+		isge->length = sge->length;
+		isge->sge_length = sge->length;
+		isge->m = 0;
+		isge->n = 0;
+		goto ok;
+	}
+	mr = rkt->table[(sge->lkey >> (32 - ib_qib_lkey_table_size))];
+	if (unlikely(mr == NULL || mr->lkey != sge->lkey ||
+		     mr->pd != &pd->ibpd))
+		goto bail;
+
+	off = sge->addr - mr->user_base;
+	if (unlikely(sge->addr < mr->user_base ||
+		     off + sge->length > mr->length ||
+		     (mr->access_flags & acc) != acc))
+		goto bail;
+
+	off += mr->offset;
+	m = 0;
+	n = 0;
+	while (off >= mr->map[m]->segs[n].length) {
+		off -= mr->map[m]->segs[n].length;
+		n++;
+		if (n >= QIB_SEGSZ) {
+			m++;
+			n = 0;
+		}
+	}
+	atomic_inc(&mr->refcount);
+	isge->mr = mr;
+	isge->vaddr = mr->map[m]->segs[n].vaddr + off;
+	isge->length = mr->map[m]->segs[n].length - off;
+	isge->sge_length = sge->length;
+	isge->m = m;
+	isge->n = n;
+ok:
+	ret = 1;
+bail:
+	spin_unlock_irqrestore(&rkt->lock, flags);
+	return ret;
+}
+
+/**
+ * qib_rkey_ok - check the IB virtual address, length, and RKEY
+ * @dev: infiniband device
+ * @ss: SGE state
+ * @len: length of data
+ * @vaddr: virtual address to place data
+ * @rkey: rkey to check
+ * @acc: access flags
+ *
+ * Return 1 if successful, otherwise 0.
+ */
+int qib_rkey_ok(struct qib_qp *qp, struct qib_sge *sge,
+		u32 len, u64 vaddr, u32 rkey, int acc)
+{
+	struct qib_lkey_table *rkt = &to_idev(qp->ibqp.device)->lk_table;
+	struct qib_mregion *mr;
+	unsigned n, m;
+	size_t off;
+	int ret = 0;
+	unsigned long flags;
+
+	/*
+	 * We use RKEY == zero for kernel virtual addresses
+	 * (see qib_get_dma_mr and qib_dma.c).
+	 */
+	spin_lock_irqsave(&rkt->lock, flags);
+	if (rkey == 0) {
+		struct qib_pd *pd = to_ipd(qp->ibqp.pd);
+		struct qib_ibdev *dev = to_idev(pd->ibpd.device);
+
+		if (pd->user)
+			goto bail;
+		if (!dev->dma_mr)
+			goto bail;
+		atomic_inc(&dev->dma_mr->refcount);
+		sge->mr = dev->dma_mr;
+		sge->vaddr = (void *) vaddr;
+		sge->length = len;
+		sge->sge_length = len;
+		sge->m = 0;
+		sge->n = 0;
+		goto ok;
+	}
+
+	mr = rkt->table[(rkey >> (32 - ib_qib_lkey_table_size))];
+	if (unlikely(mr == NULL || mr->lkey != rkey || qp->ibqp.pd != mr->pd))
+		goto bail;
+
+	off = vaddr - mr->iova;
+	if (unlikely(vaddr < mr->iova || off + len > mr->length ||
+		     (mr->access_flags & acc) == 0))
+		goto bail;
+
+	off += mr->offset;
+	m = 0;
+	n = 0;
+	while (off >= mr->map[m]->segs[n].length) {
+		off -= mr->map[m]->segs[n].length;
+		n++;
+		if (n >= QIB_SEGSZ) {
+			m++;
+			n = 0;
+		}
+	}
+	atomic_inc(&mr->refcount);
+	sge->mr = mr;
+	sge->vaddr = mr->map[m]->segs[n].vaddr + off;
+	sge->length = mr->map[m]->segs[n].length - off;
+	sge->sge_length = len;
+	sge->m = m;
+	sge->n = n;
+ok:
+	ret = 1;
+bail:
+	spin_unlock_irqrestore(&rkt->lock, flags);
+	return ret;
+}
+
+/*
+ * Initialize the memory region specified by the work reqeust.
+ */
+int qib_fast_reg_mr(struct qib_qp *qp, struct ib_send_wr *wr)
+{
+	struct qib_lkey_table *rkt = &to_idev(qp->ibqp.device)->lk_table;
+	struct qib_pd *pd = to_ipd(qp->ibqp.pd);
+	struct qib_mregion *mr;
+	u32 rkey = wr->wr.fast_reg.rkey;
+	unsigned i, n, m;
+	int ret = -EINVAL;
+	unsigned long flags;
+	u64 *page_list;
+	size_t ps;
+
+	spin_lock_irqsave(&rkt->lock, flags);
+	if (pd->user || rkey == 0)
+		goto bail;
+
+	mr = rkt->table[(rkey >> (32 - ib_qib_lkey_table_size))];
+	if (unlikely(mr == NULL || qp->ibqp.pd != mr->pd))
+		goto bail;
+
+	if (wr->wr.fast_reg.page_list_len > mr->max_segs)
+		goto bail;
+
+	ps = 1UL << wr->wr.fast_reg.page_shift;
+	if (wr->wr.fast_reg.length > ps * wr->wr.fast_reg.page_list_len)
+		goto bail;
+
+	mr->user_base = wr->wr.fast_reg.iova_start;
+	mr->iova = wr->wr.fast_reg.iova_start;
+	mr->lkey = rkey;
+	mr->length = wr->wr.fast_reg.length;
+	mr->access_flags = wr->wr.fast_reg.access_flags;
+	page_list = wr->wr.fast_reg.page_list->page_list;
+	m = 0;
+	n = 0;
+	for (i = 0; i < wr->wr.fast_reg.page_list_len; i++) {
+		mr->map[m]->segs[n].vaddr = (void *) page_list[i];
+		mr->map[m]->segs[n].length = ps;
+		if (++n == QIB_SEGSZ) {
+			m++;
+			n = 0;
+		}
+	}
+
+	ret = 0;
+bail:
+	spin_unlock_irqrestore(&rkt->lock, flags);
+	return ret;
+}
diff --git a/drivers/infiniband/hw/qib/qib_mad.c b/drivers/infiniband/hw/qib/qib_mad.c
new file mode 100644
index 000000000000..94b0d1f3a8f0
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_mad.c
@@ -0,0 +1,2173 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009, 2010 QLogic Corporation.
+ * All rights reserved.
+ * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <rdma/ib_smi.h>
+
+#include "qib.h"
+#include "qib_mad.h"
+
+static int reply(struct ib_smp *smp)
+{
+	/*
+	 * The verbs framework will handle the directed/LID route
+	 * packet changes.
+	 */
+	smp->method = IB_MGMT_METHOD_GET_RESP;
+	if (smp->mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE)
+		smp->status |= IB_SMP_DIRECTION;
+	return IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_REPLY;
+}
+
+static void qib_send_trap(struct qib_ibport *ibp, void *data, unsigned len)
+{
+	struct ib_mad_send_buf *send_buf;
+	struct ib_mad_agent *agent;
+	struct ib_smp *smp;
+	int ret;
+	unsigned long flags;
+	unsigned long timeout;
+
+	agent = ibp->send_agent;
+	if (!agent)
+		return;
+
+	/* o14-3.2.1 */
+	if (!(ppd_from_ibp(ibp)->lflags & QIBL_LINKACTIVE))
+		return;
+
+	/* o14-2 */
+	if (ibp->trap_timeout && time_before(jiffies, ibp->trap_timeout))
+		return;
+
+	send_buf = ib_create_send_mad(agent, 0, 0, 0, IB_MGMT_MAD_HDR,
+				      IB_MGMT_MAD_DATA, GFP_ATOMIC);
+	if (IS_ERR(send_buf))
+		return;
+
+	smp = send_buf->mad;
+	smp->base_version = IB_MGMT_BASE_VERSION;
+	smp->mgmt_class = IB_MGMT_CLASS_SUBN_LID_ROUTED;
+	smp->class_version = 1;
+	smp->method = IB_MGMT_METHOD_TRAP;
+	ibp->tid++;
+	smp->tid = cpu_to_be64(ibp->tid);
+	smp->attr_id = IB_SMP_ATTR_NOTICE;
+	/* o14-1: smp->mkey = 0; */
+	memcpy(smp->data, data, len);
+
+	spin_lock_irqsave(&ibp->lock, flags);
+	if (!ibp->sm_ah) {
+		if (ibp->sm_lid != be16_to_cpu(IB_LID_PERMISSIVE)) {
+			struct ib_ah *ah;
+			struct ib_ah_attr attr;
+
+			memset(&attr, 0, sizeof attr);
+			attr.dlid = ibp->sm_lid;
+			attr.port_num = ppd_from_ibp(ibp)->port;
+			ah = ib_create_ah(ibp->qp0->ibqp.pd, &attr);
+			if (IS_ERR(ah))
+				ret = -EINVAL;
+			else {
+				send_buf->ah = ah;
+				ibp->sm_ah = to_iah(ah);
+				ret = 0;
+			}
+		} else
+			ret = -EINVAL;
+	} else {
+		send_buf->ah = &ibp->sm_ah->ibah;
+		ret = 0;
+	}
+	spin_unlock_irqrestore(&ibp->lock, flags);
+
+	if (!ret)
+		ret = ib_post_send_mad(send_buf, NULL);
+	if (!ret) {
+		/* 4.096 usec. */
+		timeout = (4096 * (1UL << ibp->subnet_timeout)) / 1000;
+		ibp->trap_timeout = jiffies + usecs_to_jiffies(timeout);
+	} else {
+		ib_free_send_mad(send_buf);
+		ibp->trap_timeout = 0;
+	}
+}
+
+/*
+ * Send a bad [PQ]_Key trap (ch. 14.3.8).
+ */
+void qib_bad_pqkey(struct qib_ibport *ibp, __be16 trap_num, u32 key, u32 sl,
+		   u32 qp1, u32 qp2, __be16 lid1, __be16 lid2)
+{
+	struct ib_mad_notice_attr data;
+
+	if (trap_num == IB_NOTICE_TRAP_BAD_PKEY)
+		ibp->pkey_violations++;
+	else
+		ibp->qkey_violations++;
+	ibp->n_pkt_drops++;
+
+	/* Send violation trap */
+	data.generic_type = IB_NOTICE_TYPE_SECURITY;
+	data.prod_type_msb = 0;
+	data.prod_type_lsb = IB_NOTICE_PROD_CA;
+	data.trap_num = trap_num;
+	data.issuer_lid = cpu_to_be16(ppd_from_ibp(ibp)->lid);
+	data.toggle_count = 0;
+	memset(&data.details, 0, sizeof data.details);
+	data.details.ntc_257_258.lid1 = lid1;
+	data.details.ntc_257_258.lid2 = lid2;
+	data.details.ntc_257_258.key = cpu_to_be32(key);
+	data.details.ntc_257_258.sl_qp1 = cpu_to_be32((sl << 28) | qp1);
+	data.details.ntc_257_258.qp2 = cpu_to_be32(qp2);
+
+	qib_send_trap(ibp, &data, sizeof data);
+}
+
+/*
+ * Send a bad M_Key trap (ch. 14.3.9).
+ */
+static void qib_bad_mkey(struct qib_ibport *ibp, struct ib_smp *smp)
+{
+	struct ib_mad_notice_attr data;
+
+	/* Send violation trap */
+	data.generic_type = IB_NOTICE_TYPE_SECURITY;
+	data.prod_type_msb = 0;
+	data.prod_type_lsb = IB_NOTICE_PROD_CA;
+	data.trap_num = IB_NOTICE_TRAP_BAD_MKEY;
+	data.issuer_lid = cpu_to_be16(ppd_from_ibp(ibp)->lid);
+	data.toggle_count = 0;
+	memset(&data.details, 0, sizeof data.details);
+	data.details.ntc_256.lid = data.issuer_lid;
+	data.details.ntc_256.method = smp->method;
+	data.details.ntc_256.attr_id = smp->attr_id;
+	data.details.ntc_256.attr_mod = smp->attr_mod;
+	data.details.ntc_256.mkey = smp->mkey;
+	if (smp->mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) {
+		u8 hop_cnt;
+
+		data.details.ntc_256.dr_slid = smp->dr_slid;
+		data.details.ntc_256.dr_trunc_hop = IB_NOTICE_TRAP_DR_NOTICE;
+		hop_cnt = smp->hop_cnt;
+		if (hop_cnt > ARRAY_SIZE(data.details.ntc_256.dr_rtn_path)) {
+			data.details.ntc_256.dr_trunc_hop |=
+				IB_NOTICE_TRAP_DR_TRUNC;
+			hop_cnt = ARRAY_SIZE(data.details.ntc_256.dr_rtn_path);
+		}
+		data.details.ntc_256.dr_trunc_hop |= hop_cnt;
+		memcpy(data.details.ntc_256.dr_rtn_path, smp->return_path,
+		       hop_cnt);
+	}
+
+	qib_send_trap(ibp, &data, sizeof data);
+}
+
+/*
+ * Send a Port Capability Mask Changed trap (ch. 14.3.11).
+ */
+void qib_cap_mask_chg(struct qib_ibport *ibp)
+{
+	struct ib_mad_notice_attr data;
+
+	data.generic_type = IB_NOTICE_TYPE_INFO;
+	data.prod_type_msb = 0;
+	data.prod_type_lsb = IB_NOTICE_PROD_CA;
+	data.trap_num = IB_NOTICE_TRAP_CAP_MASK_CHG;
+	data.issuer_lid = cpu_to_be16(ppd_from_ibp(ibp)->lid);
+	data.toggle_count = 0;
+	memset(&data.details, 0, sizeof data.details);
+	data.details.ntc_144.lid = data.issuer_lid;
+	data.details.ntc_144.new_cap_mask = cpu_to_be32(ibp->port_cap_flags);
+
+	qib_send_trap(ibp, &data, sizeof data);
+}
+
+/*
+ * Send a System Image GUID Changed trap (ch. 14.3.12).
+ */
+void qib_sys_guid_chg(struct qib_ibport *ibp)
+{
+	struct ib_mad_notice_attr data;
+
+	data.generic_type = IB_NOTICE_TYPE_INFO;
+	data.prod_type_msb = 0;
+	data.prod_type_lsb = IB_NOTICE_PROD_CA;
+	data.trap_num = IB_NOTICE_TRAP_SYS_GUID_CHG;
+	data.issuer_lid = cpu_to_be16(ppd_from_ibp(ibp)->lid);
+	data.toggle_count = 0;
+	memset(&data.details, 0, sizeof data.details);
+	data.details.ntc_145.lid = data.issuer_lid;
+	data.details.ntc_145.new_sys_guid = ib_qib_sys_image_guid;
+
+	qib_send_trap(ibp, &data, sizeof data);
+}
+
+/*
+ * Send a Node Description Changed trap (ch. 14.3.13).
+ */
+void qib_node_desc_chg(struct qib_ibport *ibp)
+{
+	struct ib_mad_notice_attr data;
+
+	data.generic_type = IB_NOTICE_TYPE_INFO;
+	data.prod_type_msb = 0;
+	data.prod_type_lsb = IB_NOTICE_PROD_CA;
+	data.trap_num = IB_NOTICE_TRAP_CAP_MASK_CHG;
+	data.issuer_lid = cpu_to_be16(ppd_from_ibp(ibp)->lid);
+	data.toggle_count = 0;
+	memset(&data.details, 0, sizeof data.details);
+	data.details.ntc_144.lid = data.issuer_lid;
+	data.details.ntc_144.local_changes = 1;
+	data.details.ntc_144.change_flags = IB_NOTICE_TRAP_NODE_DESC_CHG;
+
+	qib_send_trap(ibp, &data, sizeof data);
+}
+
+static int subn_get_nodedescription(struct ib_smp *smp,
+				    struct ib_device *ibdev)
+{
+	if (smp->attr_mod)
+		smp->status |= IB_SMP_INVALID_FIELD;
+
+	memcpy(smp->data, ibdev->node_desc, sizeof(smp->data));
+
+	return reply(smp);
+}
+
+static int subn_get_nodeinfo(struct ib_smp *smp, struct ib_device *ibdev,
+			     u8 port)
+{
+	struct ib_node_info *nip = (struct ib_node_info *)&smp->data;
+	struct qib_devdata *dd = dd_from_ibdev(ibdev);
+	u32 vendor, majrev, minrev;
+	unsigned pidx = port - 1; /* IB number port from 1, hdw from 0 */
+
+	/* GUID 0 is illegal */
+	if (smp->attr_mod || pidx >= dd->num_pports ||
+	    dd->pport[pidx].guid == 0)
+		smp->status |= IB_SMP_INVALID_FIELD;
+	else
+		nip->port_guid = dd->pport[pidx].guid;
+
+	nip->base_version = 1;
+	nip->class_version = 1;
+	nip->node_type = 1;     /* channel adapter */
+	nip->num_ports = ibdev->phys_port_cnt;
+	/* This is already in network order */
+	nip->sys_guid = ib_qib_sys_image_guid;
+	nip->node_guid = dd->pport->guid; /* Use first-port GUID as node */
+	nip->partition_cap = cpu_to_be16(qib_get_npkeys(dd));
+	nip->device_id = cpu_to_be16(dd->deviceid);
+	majrev = dd->majrev;
+	minrev = dd->minrev;
+	nip->revision = cpu_to_be32((majrev << 16) | minrev);
+	nip->local_port_num = port;
+	vendor = dd->vendorid;
+	nip->vendor_id[0] = QIB_SRC_OUI_1;
+	nip->vendor_id[1] = QIB_SRC_OUI_2;
+	nip->vendor_id[2] = QIB_SRC_OUI_3;
+
+	return reply(smp);
+}
+
+static int subn_get_guidinfo(struct ib_smp *smp, struct ib_device *ibdev,
+			     u8 port)
+{
+	struct qib_devdata *dd = dd_from_ibdev(ibdev);
+	u32 startgx = 8 * be32_to_cpu(smp->attr_mod);
+	__be64 *p = (__be64 *) smp->data;
+	unsigned pidx = port - 1; /* IB number port from 1, hdw from 0 */
+
+	/* 32 blocks of 8 64-bit GUIDs per block */
+
+	memset(smp->data, 0, sizeof(smp->data));
+
+	if (startgx == 0 && pidx < dd->num_pports) {
+		struct qib_pportdata *ppd = dd->pport + pidx;
+		struct qib_ibport *ibp = &ppd->ibport_data;
+		__be64 g = ppd->guid;
+		unsigned i;
+
+		/* GUID 0 is illegal */
+		if (g == 0)
+			smp->status |= IB_SMP_INVALID_FIELD;
+		else {
+			/* The first is a copy of the read-only HW GUID. */
+			p[0] = g;
+			for (i = 1; i < QIB_GUIDS_PER_PORT; i++)
+				p[i] = ibp->guids[i - 1];
+		}
+	} else
+		smp->status |= IB_SMP_INVALID_FIELD;
+
+	return reply(smp);
+}
+
+static void set_link_width_enabled(struct qib_pportdata *ppd, u32 w)
+{
+	(void) ppd->dd->f_set_ib_cfg(ppd, QIB_IB_CFG_LWID_ENB, w);
+}
+
+static void set_link_speed_enabled(struct qib_pportdata *ppd, u32 s)
+{
+	(void) ppd->dd->f_set_ib_cfg(ppd, QIB_IB_CFG_SPD_ENB, s);
+}
+
+static int get_overrunthreshold(struct qib_pportdata *ppd)
+{
+	return ppd->dd->f_get_ib_cfg(ppd, QIB_IB_CFG_OVERRUN_THRESH);
+}
+
+/**
+ * set_overrunthreshold - set the overrun threshold
+ * @ppd: the physical port data
+ * @n: the new threshold
+ *
+ * Note that this will only take effect when the link state changes.
+ */
+static int set_overrunthreshold(struct qib_pportdata *ppd, unsigned n)
+{
+	(void) ppd->dd->f_set_ib_cfg(ppd, QIB_IB_CFG_OVERRUN_THRESH,
+					 (u32)n);
+	return 0;
+}
+
+static int get_phyerrthreshold(struct qib_pportdata *ppd)
+{
+	return ppd->dd->f_get_ib_cfg(ppd, QIB_IB_CFG_PHYERR_THRESH);
+}
+
+/**
+ * set_phyerrthreshold - set the physical error threshold
+ * @ppd: the physical port data
+ * @n: the new threshold
+ *
+ * Note that this will only take effect when the link state changes.
+ */
+static int set_phyerrthreshold(struct qib_pportdata *ppd, unsigned n)
+{
+	(void) ppd->dd->f_set_ib_cfg(ppd, QIB_IB_CFG_PHYERR_THRESH,
+					 (u32)n);
+	return 0;
+}
+
+/**
+ * get_linkdowndefaultstate - get the default linkdown state
+ * @ppd: the physical port data
+ *
+ * Returns zero if the default is POLL, 1 if the default is SLEEP.
+ */
+static int get_linkdowndefaultstate(struct qib_pportdata *ppd)
+{
+	return ppd->dd->f_get_ib_cfg(ppd, QIB_IB_CFG_LINKDEFAULT) ==
+		IB_LINKINITCMD_SLEEP;
+}
+
+static int check_mkey(struct qib_ibport *ibp, struct ib_smp *smp, int mad_flags)
+{
+	int ret = 0;
+
+	/* Is the mkey in the process of expiring? */
+	if (ibp->mkey_lease_timeout &&
+	    time_after_eq(jiffies, ibp->mkey_lease_timeout)) {
+		/* Clear timeout and mkey protection field. */
+		ibp->mkey_lease_timeout = 0;
+		ibp->mkeyprot = 0;
+	}
+
+	/* M_Key checking depends on Portinfo:M_Key_protect_bits */
+	if ((mad_flags & IB_MAD_IGNORE_MKEY) == 0 && ibp->mkey != 0 &&
+	    ibp->mkey != smp->mkey &&
+	    (smp->method == IB_MGMT_METHOD_SET ||
+	     smp->method == IB_MGMT_METHOD_TRAP_REPRESS ||
+	     (smp->method == IB_MGMT_METHOD_GET && ibp->mkeyprot >= 2))) {
+		if (ibp->mkey_violations != 0xFFFF)
+			++ibp->mkey_violations;
+		if (!ibp->mkey_lease_timeout && ibp->mkey_lease_period)
+			ibp->mkey_lease_timeout = jiffies +
+				ibp->mkey_lease_period * HZ;
+		/* Generate a trap notice. */
+		qib_bad_mkey(ibp, smp);
+		ret = IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_CONSUMED;
+	} else if (ibp->mkey_lease_timeout)
+		ibp->mkey_lease_timeout = 0;
+
+	return ret;
+}
+
+static int subn_get_portinfo(struct ib_smp *smp, struct ib_device *ibdev,
+			     u8 port)
+{
+	struct qib_devdata *dd;
+	struct qib_pportdata *ppd;
+	struct qib_ibport *ibp;
+	struct ib_port_info *pip = (struct ib_port_info *)smp->data;
+	u16 lid;
+	u8 mtu;
+	int ret;
+	u32 state;
+	u32 port_num = be32_to_cpu(smp->attr_mod);
+
+	if (port_num == 0)
+		port_num = port;
+	else {
+		if (port_num > ibdev->phys_port_cnt) {
+			smp->status |= IB_SMP_INVALID_FIELD;
+			ret = reply(smp);
+			goto bail;
+		}
+		if (port_num != port) {
+			ibp = to_iport(ibdev, port_num);
+			ret = check_mkey(ibp, smp, 0);
+			if (ret)
+				goto bail;
+		}
+	}
+
+	dd = dd_from_ibdev(ibdev);
+	/* IB numbers ports from 1, hdw from 0 */
+	ppd = dd->pport + (port_num - 1);
+	ibp = &ppd->ibport_data;
+
+	/* Clear all fields.  Only set the non-zero fields. */
+	memset(smp->data, 0, sizeof(smp->data));
+
+	/* Only return the mkey if the protection field allows it. */
+	if (smp->method == IB_MGMT_METHOD_SET || ibp->mkey == smp->mkey ||
+	    ibp->mkeyprot == 0)
+		pip->mkey = ibp->mkey;
+	pip->gid_prefix = ibp->gid_prefix;
+	lid = ppd->lid;
+	pip->lid = lid ? cpu_to_be16(lid) : IB_LID_PERMISSIVE;
+	pip->sm_lid = cpu_to_be16(ibp->sm_lid);
+	pip->cap_mask = cpu_to_be32(ibp->port_cap_flags);
+	/* pip->diag_code; */
+	pip->mkey_lease_period = cpu_to_be16(ibp->mkey_lease_period);
+	pip->local_port_num = port;
+	pip->link_width_enabled = ppd->link_width_enabled;
+	pip->link_width_supported = ppd->link_width_supported;
+	pip->link_width_active = ppd->link_width_active;
+	state = dd->f_iblink_state(ppd->lastibcstat);
+	pip->linkspeed_portstate = ppd->link_speed_supported << 4 | state;
+
+	pip->portphysstate_linkdown =
+		(dd->f_ibphys_portstate(ppd->lastibcstat) << 4) |
+		(get_linkdowndefaultstate(ppd) ? 1 : 2);
+	pip->mkeyprot_resv_lmc = (ibp->mkeyprot << 6) | ppd->lmc;
+	pip->linkspeedactive_enabled = (ppd->link_speed_active << 4) |
+		ppd->link_speed_enabled;
+	switch (ppd->ibmtu) {
+	default: /* something is wrong; fall through */
+	case 4096:
+		mtu = IB_MTU_4096;
+		break;
+	case 2048:
+		mtu = IB_MTU_2048;
+		break;
+	case 1024:
+		mtu = IB_MTU_1024;
+		break;
+	case 512:
+		mtu = IB_MTU_512;
+		break;
+	case 256:
+		mtu = IB_MTU_256;
+		break;
+	}
+	pip->neighbormtu_mastersmsl = (mtu << 4) | ibp->sm_sl;
+	pip->vlcap_inittype = ppd->vls_supported << 4;  /* InitType = 0 */
+	pip->vl_high_limit = ibp->vl_high_limit;
+	pip->vl_arb_high_cap =
+		dd->f_get_ib_cfg(ppd, QIB_IB_CFG_VL_HIGH_CAP);
+	pip->vl_arb_low_cap =
+		dd->f_get_ib_cfg(ppd, QIB_IB_CFG_VL_LOW_CAP);
+	/* InitTypeReply = 0 */
+	pip->inittypereply_mtucap = qib_ibmtu ? qib_ibmtu : IB_MTU_4096;
+	/* HCAs ignore VLStallCount and HOQLife */
+	/* pip->vlstallcnt_hoqlife; */
+	pip->operationalvl_pei_peo_fpi_fpo =
+		dd->f_get_ib_cfg(ppd, QIB_IB_CFG_OP_VLS) << 4;
+	pip->mkey_violations = cpu_to_be16(ibp->mkey_violations);
+	/* P_KeyViolations are counted by hardware. */
+	pip->pkey_violations = cpu_to_be16(ibp->pkey_violations);
+	pip->qkey_violations = cpu_to_be16(ibp->qkey_violations);
+	/* Only the hardware GUID is supported for now */
+	pip->guid_cap = QIB_GUIDS_PER_PORT;
+	pip->clientrereg_resv_subnetto = ibp->subnet_timeout;
+	/* 32.768 usec. response time (guessing) */
+	pip->resv_resptimevalue = 3;
+	pip->localphyerrors_overrunerrors =
+		(get_phyerrthreshold(ppd) << 4) |
+		get_overrunthreshold(ppd);
+	/* pip->max_credit_hint; */
+	if (ibp->port_cap_flags & IB_PORT_LINK_LATENCY_SUP) {
+		u32 v;
+
+		v = dd->f_get_ib_cfg(ppd, QIB_IB_CFG_LINKLATENCY);
+		pip->link_roundtrip_latency[0] = v >> 16;
+		pip->link_roundtrip_latency[1] = v >> 8;
+		pip->link_roundtrip_latency[2] = v;
+	}
+
+	ret = reply(smp);
+
+bail:
+	return ret;
+}
+
+/**
+ * get_pkeys - return the PKEY table
+ * @dd: the qlogic_ib device
+ * @port: the IB port number
+ * @pkeys: the pkey table is placed here
+ */
+static int get_pkeys(struct qib_devdata *dd, u8 port, u16 *pkeys)
+{
+	struct qib_pportdata *ppd = dd->pport + port - 1;
+	/*
+	 * always a kernel context, no locking needed.
+	 * If we get here with ppd setup, no need to check
+	 * that pd is valid.
+	 */
+	struct qib_ctxtdata *rcd = dd->rcd[ppd->hw_pidx];
+
+	memcpy(pkeys, rcd->pkeys, sizeof(rcd->pkeys));
+
+	return 0;
+}
+
+static int subn_get_pkeytable(struct ib_smp *smp, struct ib_device *ibdev,
+			      u8 port)
+{
+	u32 startpx = 32 * (be32_to_cpu(smp->attr_mod) & 0xffff);
+	u16 *p = (u16 *) smp->data;
+	__be16 *q = (__be16 *) smp->data;
+
+	/* 64 blocks of 32 16-bit P_Key entries */
+
+	memset(smp->data, 0, sizeof(smp->data));
+	if (startpx == 0) {
+		struct qib_devdata *dd = dd_from_ibdev(ibdev);
+		unsigned i, n = qib_get_npkeys(dd);
+
+		get_pkeys(dd, port, p);
+
+		for (i = 0; i < n; i++)
+			q[i] = cpu_to_be16(p[i]);
+	} else
+		smp->status |= IB_SMP_INVALID_FIELD;
+
+	return reply(smp);
+}
+
+static int subn_set_guidinfo(struct ib_smp *smp, struct ib_device *ibdev,
+			     u8 port)
+{
+	struct qib_devdata *dd = dd_from_ibdev(ibdev);
+	u32 startgx = 8 * be32_to_cpu(smp->attr_mod);
+	__be64 *p = (__be64 *) smp->data;
+	unsigned pidx = port - 1; /* IB number port from 1, hdw from 0 */
+
+	/* 32 blocks of 8 64-bit GUIDs per block */
+
+	if (startgx == 0 && pidx < dd->num_pports) {
+		struct qib_pportdata *ppd = dd->pport + pidx;
+		struct qib_ibport *ibp = &ppd->ibport_data;
+		unsigned i;
+
+		/* The first entry is read-only. */
+		for (i = 1; i < QIB_GUIDS_PER_PORT; i++)
+			ibp->guids[i - 1] = p[i];
+	} else
+		smp->status |= IB_SMP_INVALID_FIELD;
+
+	/* The only GUID we support is the first read-only entry. */
+	return subn_get_guidinfo(smp, ibdev, port);
+}
+
+/**
+ * subn_set_portinfo - set port information
+ * @smp: the incoming SM packet
+ * @ibdev: the infiniband device
+ * @port: the port on the device
+ *
+ * Set Portinfo (see ch. 14.2.5.6).
+ */
+static int subn_set_portinfo(struct ib_smp *smp, struct ib_device *ibdev,
+			     u8 port)
+{
+	struct ib_port_info *pip = (struct ib_port_info *)smp->data;
+	struct ib_event event;
+	struct qib_devdata *dd;
+	struct qib_pportdata *ppd;
+	struct qib_ibport *ibp;
+	char clientrereg = 0;
+	unsigned long flags;
+	u16 lid, smlid;
+	u8 lwe;
+	u8 lse;
+	u8 state;
+	u8 vls;
+	u8 msl;
+	u16 lstate;
+	int ret, ore, mtu;
+	u32 port_num = be32_to_cpu(smp->attr_mod);
+
+	if (port_num == 0)
+		port_num = port;
+	else {
+		if (port_num > ibdev->phys_port_cnt)
+			goto err;
+		/* Port attributes can only be set on the receiving port */
+		if (port_num != port)
+			goto get_only;
+	}
+
+	dd = dd_from_ibdev(ibdev);
+	/* IB numbers ports from 1, hdw from 0 */
+	ppd = dd->pport + (port_num - 1);
+	ibp = &ppd->ibport_data;
+	event.device = ibdev;
+	event.element.port_num = port;
+
+	ibp->mkey = pip->mkey;
+	ibp->gid_prefix = pip->gid_prefix;
+	ibp->mkey_lease_period = be16_to_cpu(pip->mkey_lease_period);
+
+	lid = be16_to_cpu(pip->lid);
+	/* Must be a valid unicast LID address. */
+	if (lid == 0 || lid >= QIB_MULTICAST_LID_BASE)
+		goto err;
+	if (ppd->lid != lid || ppd->lmc != (pip->mkeyprot_resv_lmc & 7)) {
+		if (ppd->lid != lid)
+			qib_set_uevent_bits(ppd, _QIB_EVENT_LID_CHANGE_BIT);
+		if (ppd->lmc != (pip->mkeyprot_resv_lmc & 7))
+			qib_set_uevent_bits(ppd, _QIB_EVENT_LMC_CHANGE_BIT);
+		qib_set_lid(ppd, lid, pip->mkeyprot_resv_lmc & 7);
+		event.event = IB_EVENT_LID_CHANGE;
+		ib_dispatch_event(&event);
+	}
+
+	smlid = be16_to_cpu(pip->sm_lid);
+	msl = pip->neighbormtu_mastersmsl & 0xF;
+	/* Must be a valid unicast LID address. */
+	if (smlid == 0 || smlid >= QIB_MULTICAST_LID_BASE)
+		goto err;
+	if (smlid != ibp->sm_lid || msl != ibp->sm_sl) {
+		spin_lock_irqsave(&ibp->lock, flags);
+		if (ibp->sm_ah) {
+			if (smlid != ibp->sm_lid)
+				ibp->sm_ah->attr.dlid = smlid;
+			if (msl != ibp->sm_sl)
+				ibp->sm_ah->attr.sl = msl;
+		}
+		spin_unlock_irqrestore(&ibp->lock, flags);
+		if (smlid != ibp->sm_lid)
+			ibp->sm_lid = smlid;
+		if (msl != ibp->sm_sl)
+			ibp->sm_sl = msl;
+		event.event = IB_EVENT_SM_CHANGE;
+		ib_dispatch_event(&event);
+	}
+
+	/* Allow 1x or 4x to be set (see 14.2.6.6). */
+	lwe = pip->link_width_enabled;
+	if (lwe) {
+		if (lwe == 0xFF)
+			lwe = ppd->link_width_supported;
+		else if (lwe >= 16 || (lwe & ~ppd->link_width_supported))
+			goto err;
+		set_link_width_enabled(ppd, lwe);
+	}
+
+	lse = pip->linkspeedactive_enabled & 0xF;
+	if (lse) {
+		/*
+		 * The IB 1.2 spec. only allows link speed values
+		 * 1, 3, 5, 7, 15.  1.2.1 extended to allow specific
+		 * speeds.
+		 */
+		if (lse == 15)
+			lse = ppd->link_speed_supported;
+		else if (lse >= 8 || (lse & ~ppd->link_speed_supported))
+			goto err;
+		set_link_speed_enabled(ppd, lse);
+	}
+
+	/* Set link down default state. */
+	switch (pip->portphysstate_linkdown & 0xF) {
+	case 0: /* NOP */
+		break;
+	case 1: /* SLEEP */
+		(void) dd->f_set_ib_cfg(ppd, QIB_IB_CFG_LINKDEFAULT,
+					IB_LINKINITCMD_SLEEP);
+		break;
+	case 2: /* POLL */
+		(void) dd->f_set_ib_cfg(ppd, QIB_IB_CFG_LINKDEFAULT,
+					IB_LINKINITCMD_POLL);
+		break;
+	default:
+		goto err;
+	}
+
+	ibp->mkeyprot = pip->mkeyprot_resv_lmc >> 6;
+	ibp->vl_high_limit = pip->vl_high_limit;
+	(void) dd->f_set_ib_cfg(ppd, QIB_IB_CFG_VL_HIGH_LIMIT,
+				    ibp->vl_high_limit);
+
+	mtu = ib_mtu_enum_to_int((pip->neighbormtu_mastersmsl >> 4) & 0xF);
+	if (mtu == -1)
+		goto err;
+	qib_set_mtu(ppd, mtu);
+
+	/* Set operational VLs */
+	vls = (pip->operationalvl_pei_peo_fpi_fpo >> 4) & 0xF;
+	if (vls) {
+		if (vls > ppd->vls_supported)
+			goto err;
+		(void) dd->f_set_ib_cfg(ppd, QIB_IB_CFG_OP_VLS, vls);
+	}
+
+	if (pip->mkey_violations == 0)
+		ibp->mkey_violations = 0;
+
+	if (pip->pkey_violations == 0)
+		ibp->pkey_violations = 0;
+
+	if (pip->qkey_violations == 0)
+		ibp->qkey_violations = 0;
+
+	ore = pip->localphyerrors_overrunerrors;
+	if (set_phyerrthreshold(ppd, (ore >> 4) & 0xF))
+		goto err;
+
+	if (set_overrunthreshold(ppd, (ore & 0xF)))
+		goto err;
+
+	ibp->subnet_timeout = pip->clientrereg_resv_subnetto & 0x1F;
+
+	if (pip->clientrereg_resv_subnetto & 0x80) {
+		clientrereg = 1;
+		event.event = IB_EVENT_CLIENT_REREGISTER;
+		ib_dispatch_event(&event);
+	}
+
+	/*
+	 * Do the port state change now that the other link parameters
+	 * have been set.
+	 * Changing the port physical state only makes sense if the link
+	 * is down or is being set to down.
+	 */
+	state = pip->linkspeed_portstate & 0xF;
+	lstate = (pip->portphysstate_linkdown >> 4) & 0xF;
+	if (lstate && !(state == IB_PORT_DOWN || state == IB_PORT_NOP))
+		goto err;
+
+	/*
+	 * Only state changes of DOWN, ARM, and ACTIVE are valid
+	 * and must be in the correct state to take effect (see 7.2.6).
+	 */
+	switch (state) {
+	case IB_PORT_NOP:
+		if (lstate == 0)
+			break;
+		/* FALLTHROUGH */
+	case IB_PORT_DOWN:
+		if (lstate == 0)
+			lstate = QIB_IB_LINKDOWN_ONLY;
+		else if (lstate == 1)
+			lstate = QIB_IB_LINKDOWN_SLEEP;
+		else if (lstate == 2)
+			lstate = QIB_IB_LINKDOWN;
+		else if (lstate == 3)
+			lstate = QIB_IB_LINKDOWN_DISABLE;
+		else
+			goto err;
+		spin_lock_irqsave(&ppd->lflags_lock, flags);
+		ppd->lflags &= ~QIBL_LINKV;
+		spin_unlock_irqrestore(&ppd->lflags_lock, flags);
+		qib_set_linkstate(ppd, lstate);
+		/*
+		 * Don't send a reply if the response would be sent
+		 * through the disabled port.
+		 */
+		if (lstate == QIB_IB_LINKDOWN_DISABLE && smp->hop_cnt) {
+			ret = IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_CONSUMED;
+			goto done;
+		}
+		qib_wait_linkstate(ppd, QIBL_LINKV, 10);
+		break;
+	case IB_PORT_ARMED:
+		qib_set_linkstate(ppd, QIB_IB_LINKARM);
+		break;
+	case IB_PORT_ACTIVE:
+		qib_set_linkstate(ppd, QIB_IB_LINKACTIVE);
+		break;
+	default:
+		/* XXX We have already partially updated our state! */
+		goto err;
+	}
+
+	ret = subn_get_portinfo(smp, ibdev, port);
+
+	if (clientrereg)
+		pip->clientrereg_resv_subnetto |= 0x80;
+
+	goto done;
+
+err:
+	smp->status |= IB_SMP_INVALID_FIELD;
+get_only:
+	ret = subn_get_portinfo(smp, ibdev, port);
+done:
+	return ret;
+}
+
+/**
+ * rm_pkey - decrecment the reference count for the given PKEY
+ * @dd: the qlogic_ib device
+ * @key: the PKEY index
+ *
+ * Return true if this was the last reference and the hardware table entry
+ * needs to be changed.
+ */
+static int rm_pkey(struct qib_pportdata *ppd, u16 key)
+{
+	int i;
+	int ret;
+
+	for (i = 0; i < ARRAY_SIZE(ppd->pkeys); i++) {
+		if (ppd->pkeys[i] != key)
+			continue;
+		if (atomic_dec_and_test(&ppd->pkeyrefs[i])) {
+			ppd->pkeys[i] = 0;
+			ret = 1;
+			goto bail;
+		}
+		break;
+	}
+
+	ret = 0;
+
+bail:
+	return ret;
+}
+
+/**
+ * add_pkey - add the given PKEY to the hardware table
+ * @dd: the qlogic_ib device
+ * @key: the PKEY
+ *
+ * Return an error code if unable to add the entry, zero if no change,
+ * or 1 if the hardware PKEY register needs to be updated.
+ */
+static int add_pkey(struct qib_pportdata *ppd, u16 key)
+{
+	int i;
+	u16 lkey = key & 0x7FFF;
+	int any = 0;
+	int ret;
+
+	if (lkey == 0x7FFF) {
+		ret = 0;
+		goto bail;
+	}
+
+	/* Look for an empty slot or a matching PKEY. */
+	for (i = 0; i < ARRAY_SIZE(ppd->pkeys); i++) {
+		if (!ppd->pkeys[i]) {
+			any++;
+			continue;
+		}
+		/* If it matches exactly, try to increment the ref count */
+		if (ppd->pkeys[i] == key) {
+			if (atomic_inc_return(&ppd->pkeyrefs[i]) > 1) {
+				ret = 0;
+				goto bail;
+			}
+			/* Lost the race. Look for an empty slot below. */
+			atomic_dec(&ppd->pkeyrefs[i]);
+			any++;
+		}
+		/*
+		 * It makes no sense to have both the limited and unlimited
+		 * PKEY set at the same time since the unlimited one will
+		 * disable the limited one.
+		 */
+		if ((ppd->pkeys[i] & 0x7FFF) == lkey) {
+			ret = -EEXIST;
+			goto bail;
+		}
+	}
+	if (!any) {
+		ret = -EBUSY;
+		goto bail;
+	}
+	for (i = 0; i < ARRAY_SIZE(ppd->pkeys); i++) {
+		if (!ppd->pkeys[i] &&
+		    atomic_inc_return(&ppd->pkeyrefs[i]) == 1) {
+			/* for qibstats, etc. */
+			ppd->pkeys[i] = key;
+			ret = 1;
+			goto bail;
+		}
+	}
+	ret = -EBUSY;
+
+bail:
+	return ret;
+}
+
+/**
+ * set_pkeys - set the PKEY table for ctxt 0
+ * @dd: the qlogic_ib device
+ * @port: the IB port number
+ * @pkeys: the PKEY table
+ */
+static int set_pkeys(struct qib_devdata *dd, u8 port, u16 *pkeys)
+{
+	struct qib_pportdata *ppd;
+	struct qib_ctxtdata *rcd;
+	int i;
+	int changed = 0;
+
+	/*
+	 * IB port one/two always maps to context zero/one,
+	 * always a kernel context, no locking needed
+	 * If we get here with ppd setup, no need to check
+	 * that rcd is valid.
+	 */
+	ppd = dd->pport + (port - 1);
+	rcd = dd->rcd[ppd->hw_pidx];
+
+	for (i = 0; i < ARRAY_SIZE(rcd->pkeys); i++) {
+		u16 key = pkeys[i];
+		u16 okey = rcd->pkeys[i];
+
+		if (key == okey)
+			continue;
+		/*
+		 * The value of this PKEY table entry is changing.
+		 * Remove the old entry in the hardware's array of PKEYs.
+		 */
+		if (okey & 0x7FFF)
+			changed |= rm_pkey(ppd, okey);
+		if (key & 0x7FFF) {
+			int ret = add_pkey(ppd, key);
+
+			if (ret < 0)
+				key = 0;
+			else
+				changed |= ret;
+		}
+		rcd->pkeys[i] = key;
+	}
+	if (changed) {
+		struct ib_event event;
+
+		(void) dd->f_set_ib_cfg(ppd, QIB_IB_CFG_PKEYS, 0);
+
+		event.event = IB_EVENT_PKEY_CHANGE;
+		event.device = &dd->verbs_dev.ibdev;
+		event.element.port_num = 1;
+		ib_dispatch_event(&event);
+	}
+	return 0;
+}
+
+static int subn_set_pkeytable(struct ib_smp *smp, struct ib_device *ibdev,
+			      u8 port)
+{
+	u32 startpx = 32 * (be32_to_cpu(smp->attr_mod) & 0xffff);
+	__be16 *p = (__be16 *) smp->data;
+	u16 *q = (u16 *) smp->data;
+	struct qib_devdata *dd = dd_from_ibdev(ibdev);
+	unsigned i, n = qib_get_npkeys(dd);
+
+	for (i = 0; i < n; i++)
+		q[i] = be16_to_cpu(p[i]);
+
+	if (startpx != 0 || set_pkeys(dd, port, q) != 0)
+		smp->status |= IB_SMP_INVALID_FIELD;
+
+	return subn_get_pkeytable(smp, ibdev, port);
+}
+
+static int subn_get_sl_to_vl(struct ib_smp *smp, struct ib_device *ibdev,
+			     u8 port)
+{
+	struct qib_ibport *ibp = to_iport(ibdev, port);
+	u8 *p = (u8 *) smp->data;
+	unsigned i;
+
+	memset(smp->data, 0, sizeof(smp->data));
+
+	if (!(ibp->port_cap_flags & IB_PORT_SL_MAP_SUP))
+		smp->status |= IB_SMP_UNSUP_METHOD;
+	else
+		for (i = 0; i < ARRAY_SIZE(ibp->sl_to_vl); i += 2)
+			*p++ = (ibp->sl_to_vl[i] << 4) | ibp->sl_to_vl[i + 1];
+
+	return reply(smp);
+}
+
+static int subn_set_sl_to_vl(struct ib_smp *smp, struct ib_device *ibdev,
+			     u8 port)
+{
+	struct qib_ibport *ibp = to_iport(ibdev, port);
+	u8 *p = (u8 *) smp->data;
+	unsigned i;
+
+	if (!(ibp->port_cap_flags & IB_PORT_SL_MAP_SUP)) {
+		smp->status |= IB_SMP_UNSUP_METHOD;
+		return reply(smp);
+	}
+
+	for (i = 0; i < ARRAY_SIZE(ibp->sl_to_vl); i += 2, p++) {
+		ibp->sl_to_vl[i] = *p >> 4;
+		ibp->sl_to_vl[i + 1] = *p & 0xF;
+	}
+	qib_set_uevent_bits(ppd_from_ibp(to_iport(ibdev, port)),
+			    _QIB_EVENT_SL2VL_CHANGE_BIT);
+
+	return subn_get_sl_to_vl(smp, ibdev, port);
+}
+
+static int subn_get_vl_arb(struct ib_smp *smp, struct ib_device *ibdev,
+			   u8 port)
+{
+	unsigned which = be32_to_cpu(smp->attr_mod) >> 16;
+	struct qib_pportdata *ppd = ppd_from_ibp(to_iport(ibdev, port));
+
+	memset(smp->data, 0, sizeof(smp->data));
+
+	if (ppd->vls_supported == IB_VL_VL0)
+		smp->status |= IB_SMP_UNSUP_METHOD;
+	else if (which == IB_VLARB_LOWPRI_0_31)
+		(void) ppd->dd->f_get_ib_table(ppd, QIB_IB_TBL_VL_LOW_ARB,
+						   smp->data);
+	else if (which == IB_VLARB_HIGHPRI_0_31)
+		(void) ppd->dd->f_get_ib_table(ppd, QIB_IB_TBL_VL_HIGH_ARB,
+						   smp->data);
+	else
+		smp->status |= IB_SMP_INVALID_FIELD;
+
+	return reply(smp);
+}
+
+static int subn_set_vl_arb(struct ib_smp *smp, struct ib_device *ibdev,
+			   u8 port)
+{
+	unsigned which = be32_to_cpu(smp->attr_mod) >> 16;
+	struct qib_pportdata *ppd = ppd_from_ibp(to_iport(ibdev, port));
+
+	if (ppd->vls_supported == IB_VL_VL0)
+		smp->status |= IB_SMP_UNSUP_METHOD;
+	else if (which == IB_VLARB_LOWPRI_0_31)
+		(void) ppd->dd->f_set_ib_table(ppd, QIB_IB_TBL_VL_LOW_ARB,
+						   smp->data);
+	else if (which == IB_VLARB_HIGHPRI_0_31)
+		(void) ppd->dd->f_set_ib_table(ppd, QIB_IB_TBL_VL_HIGH_ARB,
+						   smp->data);
+	else
+		smp->status |= IB_SMP_INVALID_FIELD;
+
+	return subn_get_vl_arb(smp, ibdev, port);
+}
+
+static int subn_trap_repress(struct ib_smp *smp, struct ib_device *ibdev,
+			     u8 port)
+{
+	/*
+	 * For now, we only send the trap once so no need to process this.
+	 * o13-6, o13-7,
+	 * o14-3.a4 The SMA shall not send any message in response to a valid
+	 * SubnTrapRepress() message.
+	 */
+	return IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_CONSUMED;
+}
+
+static int pma_get_classportinfo(struct ib_perf *pmp,
+				 struct ib_device *ibdev)
+{
+	struct ib_pma_classportinfo *p =
+		(struct ib_pma_classportinfo *)pmp->data;
+	struct qib_devdata *dd = dd_from_ibdev(ibdev);
+
+	memset(pmp->data, 0, sizeof(pmp->data));
+
+	if (pmp->attr_mod != 0)
+		pmp->status |= IB_SMP_INVALID_FIELD;
+
+	/* Note that AllPortSelect is not valid */
+	p->base_version = 1;
+	p->class_version = 1;
+	p->cap_mask = IB_PMA_CLASS_CAP_EXT_WIDTH;
+	/*
+	 * Set the most significant bit of CM2 to indicate support for
+	 * congestion statistics
+	 */
+	p->reserved[0] = dd->psxmitwait_supported << 7;
+	/*
+	 * Expected response time is 4.096 usec. * 2^18 == 1.073741824 sec.
+	 */
+	p->resp_time_value = 18;
+
+	return reply((struct ib_smp *) pmp);
+}
+
+static int pma_get_portsamplescontrol(struct ib_perf *pmp,
+				      struct ib_device *ibdev, u8 port)
+{
+	struct ib_pma_portsamplescontrol *p =
+		(struct ib_pma_portsamplescontrol *)pmp->data;
+	struct qib_ibdev *dev = to_idev(ibdev);
+	struct qib_devdata *dd = dd_from_dev(dev);
+	struct qib_ibport *ibp = to_iport(ibdev, port);
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	unsigned long flags;
+	u8 port_select = p->port_select;
+
+	memset(pmp->data, 0, sizeof(pmp->data));
+
+	p->port_select = port_select;
+	if (pmp->attr_mod != 0 || port_select != port) {
+		pmp->status |= IB_SMP_INVALID_FIELD;
+		goto bail;
+	}
+	spin_lock_irqsave(&ibp->lock, flags);
+	p->tick = dd->f_get_ib_cfg(ppd, QIB_IB_CFG_PMA_TICKS);
+	p->sample_status = dd->f_portcntr(ppd, QIBPORTCNTR_PSSTAT);
+	p->counter_width = 4;   /* 32 bit counters */
+	p->counter_mask0_9 = COUNTER_MASK0_9;
+	p->sample_start = cpu_to_be32(ibp->pma_sample_start);
+	p->sample_interval = cpu_to_be32(ibp->pma_sample_interval);
+	p->tag = cpu_to_be16(ibp->pma_tag);
+	p->counter_select[0] = ibp->pma_counter_select[0];
+	p->counter_select[1] = ibp->pma_counter_select[1];
+	p->counter_select[2] = ibp->pma_counter_select[2];
+	p->counter_select[3] = ibp->pma_counter_select[3];
+	p->counter_select[4] = ibp->pma_counter_select[4];
+	spin_unlock_irqrestore(&ibp->lock, flags);
+
+bail:
+	return reply((struct ib_smp *) pmp);
+}
+
+static int pma_set_portsamplescontrol(struct ib_perf *pmp,
+				      struct ib_device *ibdev, u8 port)
+{
+	struct ib_pma_portsamplescontrol *p =
+		(struct ib_pma_portsamplescontrol *)pmp->data;
+	struct qib_ibdev *dev = to_idev(ibdev);
+	struct qib_devdata *dd = dd_from_dev(dev);
+	struct qib_ibport *ibp = to_iport(ibdev, port);
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	unsigned long flags;
+	u8 status, xmit_flags;
+	int ret;
+
+	if (pmp->attr_mod != 0 || p->port_select != port) {
+		pmp->status |= IB_SMP_INVALID_FIELD;
+		ret = reply((struct ib_smp *) pmp);
+		goto bail;
+	}
+
+	spin_lock_irqsave(&ibp->lock, flags);
+
+	/* Port Sampling code owns the PS* HW counters */
+	xmit_flags = ppd->cong_stats.flags;
+	ppd->cong_stats.flags = IB_PMA_CONG_HW_CONTROL_SAMPLE;
+	status = dd->f_portcntr(ppd, QIBPORTCNTR_PSSTAT);
+	if (status == IB_PMA_SAMPLE_STATUS_DONE ||
+	    (status == IB_PMA_SAMPLE_STATUS_RUNNING &&
+	     xmit_flags == IB_PMA_CONG_HW_CONTROL_TIMER)) {
+		ibp->pma_sample_start = be32_to_cpu(p->sample_start);
+		ibp->pma_sample_interval = be32_to_cpu(p->sample_interval);
+		ibp->pma_tag = be16_to_cpu(p->tag);
+		ibp->pma_counter_select[0] = p->counter_select[0];
+		ibp->pma_counter_select[1] = p->counter_select[1];
+		ibp->pma_counter_select[2] = p->counter_select[2];
+		ibp->pma_counter_select[3] = p->counter_select[3];
+		ibp->pma_counter_select[4] = p->counter_select[4];
+		dd->f_set_cntr_sample(ppd, ibp->pma_sample_interval,
+				      ibp->pma_sample_start);
+	}
+	spin_unlock_irqrestore(&ibp->lock, flags);
+
+	ret = pma_get_portsamplescontrol(pmp, ibdev, port);
+
+bail:
+	return ret;
+}
+
+static u64 get_counter(struct qib_ibport *ibp, struct qib_pportdata *ppd,
+		       __be16 sel)
+{
+	u64 ret;
+
+	switch (sel) {
+	case IB_PMA_PORT_XMIT_DATA:
+		ret = ppd->dd->f_portcntr(ppd, QIBPORTCNTR_PSXMITDATA);
+		break;
+	case IB_PMA_PORT_RCV_DATA:
+		ret = ppd->dd->f_portcntr(ppd, QIBPORTCNTR_PSRCVDATA);
+		break;
+	case IB_PMA_PORT_XMIT_PKTS:
+		ret = ppd->dd->f_portcntr(ppd, QIBPORTCNTR_PSXMITPKTS);
+		break;
+	case IB_PMA_PORT_RCV_PKTS:
+		ret = ppd->dd->f_portcntr(ppd, QIBPORTCNTR_PSRCVPKTS);
+		break;
+	case IB_PMA_PORT_XMIT_WAIT:
+		ret = ppd->dd->f_portcntr(ppd, QIBPORTCNTR_PSXMITWAIT);
+		break;
+	default:
+		ret = 0;
+	}
+
+	return ret;
+}
+
+/* This function assumes that the xmit_wait lock is already held */
+static u64 xmit_wait_get_value_delta(struct qib_pportdata *ppd)
+{
+	u32 delta;
+
+	delta = get_counter(&ppd->ibport_data, ppd,
+			    IB_PMA_PORT_XMIT_WAIT);
+	return ppd->cong_stats.counter + delta;
+}
+
+static void cache_hw_sample_counters(struct qib_pportdata *ppd)
+{
+	struct qib_ibport *ibp = &ppd->ibport_data;
+
+	ppd->cong_stats.counter_cache.psxmitdata =
+		get_counter(ibp, ppd, IB_PMA_PORT_XMIT_DATA);
+	ppd->cong_stats.counter_cache.psrcvdata =
+		get_counter(ibp, ppd, IB_PMA_PORT_RCV_DATA);
+	ppd->cong_stats.counter_cache.psxmitpkts =
+		get_counter(ibp, ppd, IB_PMA_PORT_XMIT_PKTS);
+	ppd->cong_stats.counter_cache.psrcvpkts =
+		get_counter(ibp, ppd, IB_PMA_PORT_RCV_PKTS);
+	ppd->cong_stats.counter_cache.psxmitwait =
+		get_counter(ibp, ppd, IB_PMA_PORT_XMIT_WAIT);
+}
+
+static u64 get_cache_hw_sample_counters(struct qib_pportdata *ppd,
+					__be16 sel)
+{
+	u64 ret;
+
+	switch (sel) {
+	case IB_PMA_PORT_XMIT_DATA:
+		ret = ppd->cong_stats.counter_cache.psxmitdata;
+		break;
+	case IB_PMA_PORT_RCV_DATA:
+		ret = ppd->cong_stats.counter_cache.psrcvdata;
+		break;
+	case IB_PMA_PORT_XMIT_PKTS:
+		ret = ppd->cong_stats.counter_cache.psxmitpkts;
+		break;
+	case IB_PMA_PORT_RCV_PKTS:
+		ret = ppd->cong_stats.counter_cache.psrcvpkts;
+		break;
+	case IB_PMA_PORT_XMIT_WAIT:
+		ret = ppd->cong_stats.counter_cache.psxmitwait;
+		break;
+	default:
+		ret = 0;
+	}
+
+	return ret;
+}
+
+static int pma_get_portsamplesresult(struct ib_perf *pmp,
+				     struct ib_device *ibdev, u8 port)
+{
+	struct ib_pma_portsamplesresult *p =
+		(struct ib_pma_portsamplesresult *)pmp->data;
+	struct qib_ibdev *dev = to_idev(ibdev);
+	struct qib_devdata *dd = dd_from_dev(dev);
+	struct qib_ibport *ibp = to_iport(ibdev, port);
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	unsigned long flags;
+	u8 status;
+	int i;
+
+	memset(pmp->data, 0, sizeof(pmp->data));
+	spin_lock_irqsave(&ibp->lock, flags);
+	p->tag = cpu_to_be16(ibp->pma_tag);
+	if (ppd->cong_stats.flags == IB_PMA_CONG_HW_CONTROL_TIMER)
+		p->sample_status = IB_PMA_SAMPLE_STATUS_DONE;
+	else {
+		status = dd->f_portcntr(ppd, QIBPORTCNTR_PSSTAT);
+		p->sample_status = cpu_to_be16(status);
+		if (status == IB_PMA_SAMPLE_STATUS_DONE) {
+			cache_hw_sample_counters(ppd);
+			ppd->cong_stats.counter =
+				xmit_wait_get_value_delta(ppd);
+			dd->f_set_cntr_sample(ppd,
+					      QIB_CONG_TIMER_PSINTERVAL, 0);
+			ppd->cong_stats.flags = IB_PMA_CONG_HW_CONTROL_TIMER;
+		}
+	}
+	for (i = 0; i < ARRAY_SIZE(ibp->pma_counter_select); i++)
+		p->counter[i] = cpu_to_be32(
+			get_cache_hw_sample_counters(
+				ppd, ibp->pma_counter_select[i]));
+	spin_unlock_irqrestore(&ibp->lock, flags);
+
+	return reply((struct ib_smp *) pmp);
+}
+
+static int pma_get_portsamplesresult_ext(struct ib_perf *pmp,
+					 struct ib_device *ibdev, u8 port)
+{
+	struct ib_pma_portsamplesresult_ext *p =
+		(struct ib_pma_portsamplesresult_ext *)pmp->data;
+	struct qib_ibdev *dev = to_idev(ibdev);
+	struct qib_devdata *dd = dd_from_dev(dev);
+	struct qib_ibport *ibp = to_iport(ibdev, port);
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	unsigned long flags;
+	u8 status;
+	int i;
+
+	/* Port Sampling code owns the PS* HW counters */
+	memset(pmp->data, 0, sizeof(pmp->data));
+	spin_lock_irqsave(&ibp->lock, flags);
+	p->tag = cpu_to_be16(ibp->pma_tag);
+	if (ppd->cong_stats.flags == IB_PMA_CONG_HW_CONTROL_TIMER)
+		p->sample_status = IB_PMA_SAMPLE_STATUS_DONE;
+	else {
+		status = dd->f_portcntr(ppd, QIBPORTCNTR_PSSTAT);
+		p->sample_status = cpu_to_be16(status);
+		/* 64 bits */
+		p->extended_width = cpu_to_be32(0x80000000);
+		if (status == IB_PMA_SAMPLE_STATUS_DONE) {
+			cache_hw_sample_counters(ppd);
+			ppd->cong_stats.counter =
+				xmit_wait_get_value_delta(ppd);
+			dd->f_set_cntr_sample(ppd,
+					      QIB_CONG_TIMER_PSINTERVAL, 0);
+			ppd->cong_stats.flags = IB_PMA_CONG_HW_CONTROL_TIMER;
+		}
+	}
+	for (i = 0; i < ARRAY_SIZE(ibp->pma_counter_select); i++)
+		p->counter[i] = cpu_to_be64(
+			get_cache_hw_sample_counters(
+				ppd, ibp->pma_counter_select[i]));
+	spin_unlock_irqrestore(&ibp->lock, flags);
+
+	return reply((struct ib_smp *) pmp);
+}
+
+static int pma_get_portcounters(struct ib_perf *pmp,
+				struct ib_device *ibdev, u8 port)
+{
+	struct ib_pma_portcounters *p = (struct ib_pma_portcounters *)
+		pmp->data;
+	struct qib_ibport *ibp = to_iport(ibdev, port);
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	struct qib_verbs_counters cntrs;
+	u8 port_select = p->port_select;
+
+	qib_get_counters(ppd, &cntrs);
+
+	/* Adjust counters for any resets done. */
+	cntrs.symbol_error_counter -= ibp->z_symbol_error_counter;
+	cntrs.link_error_recovery_counter -=
+		ibp->z_link_error_recovery_counter;
+	cntrs.link_downed_counter -= ibp->z_link_downed_counter;
+	cntrs.port_rcv_errors -= ibp->z_port_rcv_errors;
+	cntrs.port_rcv_remphys_errors -= ibp->z_port_rcv_remphys_errors;
+	cntrs.port_xmit_discards -= ibp->z_port_xmit_discards;
+	cntrs.port_xmit_data -= ibp->z_port_xmit_data;
+	cntrs.port_rcv_data -= ibp->z_port_rcv_data;
+	cntrs.port_xmit_packets -= ibp->z_port_xmit_packets;
+	cntrs.port_rcv_packets -= ibp->z_port_rcv_packets;
+	cntrs.local_link_integrity_errors -=
+		ibp->z_local_link_integrity_errors;
+	cntrs.excessive_buffer_overrun_errors -=
+		ibp->z_excessive_buffer_overrun_errors;
+	cntrs.vl15_dropped -= ibp->z_vl15_dropped;
+	cntrs.vl15_dropped += ibp->n_vl15_dropped;
+
+	memset(pmp->data, 0, sizeof(pmp->data));
+
+	p->port_select = port_select;
+	if (pmp->attr_mod != 0 || port_select != port)
+		pmp->status |= IB_SMP_INVALID_FIELD;
+
+	if (cntrs.symbol_error_counter > 0xFFFFUL)
+		p->symbol_error_counter = cpu_to_be16(0xFFFF);
+	else
+		p->symbol_error_counter =
+			cpu_to_be16((u16)cntrs.symbol_error_counter);
+	if (cntrs.link_error_recovery_counter > 0xFFUL)
+		p->link_error_recovery_counter = 0xFF;
+	else
+		p->link_error_recovery_counter =
+			(u8)cntrs.link_error_recovery_counter;
+	if (cntrs.link_downed_counter > 0xFFUL)
+		p->link_downed_counter = 0xFF;
+	else
+		p->link_downed_counter = (u8)cntrs.link_downed_counter;
+	if (cntrs.port_rcv_errors > 0xFFFFUL)
+		p->port_rcv_errors = cpu_to_be16(0xFFFF);
+	else
+		p->port_rcv_errors =
+			cpu_to_be16((u16) cntrs.port_rcv_errors);
+	if (cntrs.port_rcv_remphys_errors > 0xFFFFUL)
+		p->port_rcv_remphys_errors = cpu_to_be16(0xFFFF);
+	else
+		p->port_rcv_remphys_errors =
+			cpu_to_be16((u16)cntrs.port_rcv_remphys_errors);
+	if (cntrs.port_xmit_discards > 0xFFFFUL)
+		p->port_xmit_discards = cpu_to_be16(0xFFFF);
+	else
+		p->port_xmit_discards =
+			cpu_to_be16((u16)cntrs.port_xmit_discards);
+	if (cntrs.local_link_integrity_errors > 0xFUL)
+		cntrs.local_link_integrity_errors = 0xFUL;
+	if (cntrs.excessive_buffer_overrun_errors > 0xFUL)
+		cntrs.excessive_buffer_overrun_errors = 0xFUL;
+	p->lli_ebor_errors = (cntrs.local_link_integrity_errors << 4) |
+		cntrs.excessive_buffer_overrun_errors;
+	if (cntrs.vl15_dropped > 0xFFFFUL)
+		p->vl15_dropped = cpu_to_be16(0xFFFF);
+	else
+		p->vl15_dropped = cpu_to_be16((u16)cntrs.vl15_dropped);
+	if (cntrs.port_xmit_data > 0xFFFFFFFFUL)
+		p->port_xmit_data = cpu_to_be32(0xFFFFFFFF);
+	else
+		p->port_xmit_data = cpu_to_be32((u32)cntrs.port_xmit_data);
+	if (cntrs.port_rcv_data > 0xFFFFFFFFUL)
+		p->port_rcv_data = cpu_to_be32(0xFFFFFFFF);
+	else
+		p->port_rcv_data = cpu_to_be32((u32)cntrs.port_rcv_data);
+	if (cntrs.port_xmit_packets > 0xFFFFFFFFUL)
+		p->port_xmit_packets = cpu_to_be32(0xFFFFFFFF);
+	else
+		p->port_xmit_packets =
+			cpu_to_be32((u32)cntrs.port_xmit_packets);
+	if (cntrs.port_rcv_packets > 0xFFFFFFFFUL)
+		p->port_rcv_packets = cpu_to_be32(0xFFFFFFFF);
+	else
+		p->port_rcv_packets =
+			cpu_to_be32((u32) cntrs.port_rcv_packets);
+
+	return reply((struct ib_smp *) pmp);
+}
+
+static int pma_get_portcounters_cong(struct ib_perf *pmp,
+				     struct ib_device *ibdev, u8 port)
+{
+	/* Congestion PMA packets start at offset 24 not 64 */
+	struct ib_pma_portcounters_cong *p =
+		(struct ib_pma_portcounters_cong *)pmp->reserved;
+	struct qib_verbs_counters cntrs;
+	struct qib_ibport *ibp = to_iport(ibdev, port);
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	struct qib_devdata *dd = dd_from_ppd(ppd);
+	u32 port_select = be32_to_cpu(pmp->attr_mod) & 0xFF;
+	u64 xmit_wait_counter;
+	unsigned long flags;
+
+	/*
+	 * This check is performed only in the GET method because the
+	 * SET method ends up calling this anyway.
+	 */
+	if (!dd->psxmitwait_supported)
+		pmp->status |= IB_SMP_UNSUP_METH_ATTR;
+	if (port_select != port)
+		pmp->status |= IB_SMP_INVALID_FIELD;
+
+	qib_get_counters(ppd, &cntrs);
+	spin_lock_irqsave(&ppd->ibport_data.lock, flags);
+	xmit_wait_counter = xmit_wait_get_value_delta(ppd);
+	spin_unlock_irqrestore(&ppd->ibport_data.lock, flags);
+
+	/* Adjust counters for any resets done. */
+	cntrs.symbol_error_counter -= ibp->z_symbol_error_counter;
+	cntrs.link_error_recovery_counter -=
+		ibp->z_link_error_recovery_counter;
+	cntrs.link_downed_counter -= ibp->z_link_downed_counter;
+	cntrs.port_rcv_errors -= ibp->z_port_rcv_errors;
+	cntrs.port_rcv_remphys_errors -=
+		ibp->z_port_rcv_remphys_errors;
+	cntrs.port_xmit_discards -= ibp->z_port_xmit_discards;
+	cntrs.local_link_integrity_errors -=
+		ibp->z_local_link_integrity_errors;
+	cntrs.excessive_buffer_overrun_errors -=
+		ibp->z_excessive_buffer_overrun_errors;
+	cntrs.vl15_dropped -= ibp->z_vl15_dropped;
+	cntrs.vl15_dropped += ibp->n_vl15_dropped;
+	cntrs.port_xmit_data -= ibp->z_port_xmit_data;
+	cntrs.port_rcv_data -= ibp->z_port_rcv_data;
+	cntrs.port_xmit_packets -= ibp->z_port_xmit_packets;
+	cntrs.port_rcv_packets -= ibp->z_port_rcv_packets;
+
+	memset(pmp->reserved, 0, sizeof(pmp->reserved) +
+	       sizeof(pmp->data));
+
+	/*
+	 * Set top 3 bits to indicate interval in picoseconds in
+	 * remaining bits.
+	 */
+	p->port_check_rate =
+		cpu_to_be16((QIB_XMIT_RATE_PICO << 13) |
+			    (dd->psxmitwait_check_rate &
+			     ~(QIB_XMIT_RATE_PICO << 13)));
+	p->port_adr_events = cpu_to_be64(0);
+	p->port_xmit_wait = cpu_to_be64(xmit_wait_counter);
+	p->port_xmit_data = cpu_to_be64(cntrs.port_xmit_data);
+	p->port_rcv_data = cpu_to_be64(cntrs.port_rcv_data);
+	p->port_xmit_packets =
+		cpu_to_be64(cntrs.port_xmit_packets);
+	p->port_rcv_packets =
+		cpu_to_be64(cntrs.port_rcv_packets);
+	if (cntrs.symbol_error_counter > 0xFFFFUL)
+		p->symbol_error_counter = cpu_to_be16(0xFFFF);
+	else
+		p->symbol_error_counter =
+			cpu_to_be16(
+				(u16)cntrs.symbol_error_counter);
+	if (cntrs.link_error_recovery_counter > 0xFFUL)
+		p->link_error_recovery_counter = 0xFF;
+	else
+		p->link_error_recovery_counter =
+			(u8)cntrs.link_error_recovery_counter;
+	if (cntrs.link_downed_counter > 0xFFUL)
+		p->link_downed_counter = 0xFF;
+	else
+		p->link_downed_counter =
+			(u8)cntrs.link_downed_counter;
+	if (cntrs.port_rcv_errors > 0xFFFFUL)
+		p->port_rcv_errors = cpu_to_be16(0xFFFF);
+	else
+		p->port_rcv_errors =
+			cpu_to_be16((u16) cntrs.port_rcv_errors);
+	if (cntrs.port_rcv_remphys_errors > 0xFFFFUL)
+		p->port_rcv_remphys_errors = cpu_to_be16(0xFFFF);
+	else
+		p->port_rcv_remphys_errors =
+			cpu_to_be16(
+				(u16)cntrs.port_rcv_remphys_errors);
+	if (cntrs.port_xmit_discards > 0xFFFFUL)
+		p->port_xmit_discards = cpu_to_be16(0xFFFF);
+	else
+		p->port_xmit_discards =
+			cpu_to_be16((u16)cntrs.port_xmit_discards);
+	if (cntrs.local_link_integrity_errors > 0xFUL)
+		cntrs.local_link_integrity_errors = 0xFUL;
+	if (cntrs.excessive_buffer_overrun_errors > 0xFUL)
+		cntrs.excessive_buffer_overrun_errors = 0xFUL;
+	p->lli_ebor_errors = (cntrs.local_link_integrity_errors << 4) |
+		cntrs.excessive_buffer_overrun_errors;
+	if (cntrs.vl15_dropped > 0xFFFFUL)
+		p->vl15_dropped = cpu_to_be16(0xFFFF);
+	else
+		p->vl15_dropped = cpu_to_be16((u16)cntrs.vl15_dropped);
+
+	return reply((struct ib_smp *)pmp);
+}
+
+static int pma_get_portcounters_ext(struct ib_perf *pmp,
+				    struct ib_device *ibdev, u8 port)
+{
+	struct ib_pma_portcounters_ext *p =
+		(struct ib_pma_portcounters_ext *)pmp->data;
+	struct qib_ibport *ibp = to_iport(ibdev, port);
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	u64 swords, rwords, spkts, rpkts, xwait;
+	u8 port_select = p->port_select;
+
+	memset(pmp->data, 0, sizeof(pmp->data));
+
+	p->port_select = port_select;
+	if (pmp->attr_mod != 0 || port_select != port) {
+		pmp->status |= IB_SMP_INVALID_FIELD;
+		goto bail;
+	}
+
+	qib_snapshot_counters(ppd, &swords, &rwords, &spkts, &rpkts, &xwait);
+
+	/* Adjust counters for any resets done. */
+	swords -= ibp->z_port_xmit_data;
+	rwords -= ibp->z_port_rcv_data;
+	spkts -= ibp->z_port_xmit_packets;
+	rpkts -= ibp->z_port_rcv_packets;
+
+	p->port_xmit_data = cpu_to_be64(swords);
+	p->port_rcv_data = cpu_to_be64(rwords);
+	p->port_xmit_packets = cpu_to_be64(spkts);
+	p->port_rcv_packets = cpu_to_be64(rpkts);
+	p->port_unicast_xmit_packets = cpu_to_be64(ibp->n_unicast_xmit);
+	p->port_unicast_rcv_packets = cpu_to_be64(ibp->n_unicast_rcv);
+	p->port_multicast_xmit_packets = cpu_to_be64(ibp->n_multicast_xmit);
+	p->port_multicast_rcv_packets = cpu_to_be64(ibp->n_multicast_rcv);
+
+bail:
+	return reply((struct ib_smp *) pmp);
+}
+
+static int pma_set_portcounters(struct ib_perf *pmp,
+				struct ib_device *ibdev, u8 port)
+{
+	struct ib_pma_portcounters *p = (struct ib_pma_portcounters *)
+		pmp->data;
+	struct qib_ibport *ibp = to_iport(ibdev, port);
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	struct qib_verbs_counters cntrs;
+
+	/*
+	 * Since the HW doesn't support clearing counters, we save the
+	 * current count and subtract it from future responses.
+	 */
+	qib_get_counters(ppd, &cntrs);
+
+	if (p->counter_select & IB_PMA_SEL_SYMBOL_ERROR)
+		ibp->z_symbol_error_counter = cntrs.symbol_error_counter;
+
+	if (p->counter_select & IB_PMA_SEL_LINK_ERROR_RECOVERY)
+		ibp->z_link_error_recovery_counter =
+			cntrs.link_error_recovery_counter;
+
+	if (p->counter_select & IB_PMA_SEL_LINK_DOWNED)
+		ibp->z_link_downed_counter = cntrs.link_downed_counter;
+
+	if (p->counter_select & IB_PMA_SEL_PORT_RCV_ERRORS)
+		ibp->z_port_rcv_errors = cntrs.port_rcv_errors;
+
+	if (p->counter_select & IB_PMA_SEL_PORT_RCV_REMPHYS_ERRORS)
+		ibp->z_port_rcv_remphys_errors =
+			cntrs.port_rcv_remphys_errors;
+
+	if (p->counter_select & IB_PMA_SEL_PORT_XMIT_DISCARDS)
+		ibp->z_port_xmit_discards = cntrs.port_xmit_discards;
+
+	if (p->counter_select & IB_PMA_SEL_LOCAL_LINK_INTEGRITY_ERRORS)
+		ibp->z_local_link_integrity_errors =
+			cntrs.local_link_integrity_errors;
+
+	if (p->counter_select & IB_PMA_SEL_EXCESSIVE_BUFFER_OVERRUNS)
+		ibp->z_excessive_buffer_overrun_errors =
+			cntrs.excessive_buffer_overrun_errors;
+
+	if (p->counter_select & IB_PMA_SEL_PORT_VL15_DROPPED) {
+		ibp->n_vl15_dropped = 0;
+		ibp->z_vl15_dropped = cntrs.vl15_dropped;
+	}
+
+	if (p->counter_select & IB_PMA_SEL_PORT_XMIT_DATA)
+		ibp->z_port_xmit_data = cntrs.port_xmit_data;
+
+	if (p->counter_select & IB_PMA_SEL_PORT_RCV_DATA)
+		ibp->z_port_rcv_data = cntrs.port_rcv_data;
+
+	if (p->counter_select & IB_PMA_SEL_PORT_XMIT_PACKETS)
+		ibp->z_port_xmit_packets = cntrs.port_xmit_packets;
+
+	if (p->counter_select & IB_PMA_SEL_PORT_RCV_PACKETS)
+		ibp->z_port_rcv_packets = cntrs.port_rcv_packets;
+
+	return pma_get_portcounters(pmp, ibdev, port);
+}
+
+static int pma_set_portcounters_cong(struct ib_perf *pmp,
+				     struct ib_device *ibdev, u8 port)
+{
+	struct qib_ibport *ibp = to_iport(ibdev, port);
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	struct qib_devdata *dd = dd_from_ppd(ppd);
+	struct qib_verbs_counters cntrs;
+	u32 counter_select = (be32_to_cpu(pmp->attr_mod) >> 24) & 0xFF;
+	int ret = 0;
+	unsigned long flags;
+
+	qib_get_counters(ppd, &cntrs);
+	/* Get counter values before we save them */
+	ret = pma_get_portcounters_cong(pmp, ibdev, port);
+
+	if (counter_select & IB_PMA_SEL_CONG_XMIT) {
+		spin_lock_irqsave(&ppd->ibport_data.lock, flags);
+		ppd->cong_stats.counter = 0;
+		dd->f_set_cntr_sample(ppd, QIB_CONG_TIMER_PSINTERVAL,
+				      0x0);
+		spin_unlock_irqrestore(&ppd->ibport_data.lock, flags);
+	}
+	if (counter_select & IB_PMA_SEL_CONG_PORT_DATA) {
+		ibp->z_port_xmit_data = cntrs.port_xmit_data;
+		ibp->z_port_rcv_data = cntrs.port_rcv_data;
+		ibp->z_port_xmit_packets = cntrs.port_xmit_packets;
+		ibp->z_port_rcv_packets = cntrs.port_rcv_packets;
+	}
+	if (counter_select & IB_PMA_SEL_CONG_ALL) {
+		ibp->z_symbol_error_counter =
+			cntrs.symbol_error_counter;
+		ibp->z_link_error_recovery_counter =
+			cntrs.link_error_recovery_counter;
+		ibp->z_link_downed_counter =
+			cntrs.link_downed_counter;
+		ibp->z_port_rcv_errors = cntrs.port_rcv_errors;
+		ibp->z_port_rcv_remphys_errors =
+			cntrs.port_rcv_remphys_errors;
+		ibp->z_port_xmit_discards =
+			cntrs.port_xmit_discards;
+		ibp->z_local_link_integrity_errors =
+			cntrs.local_link_integrity_errors;
+		ibp->z_excessive_buffer_overrun_errors =
+			cntrs.excessive_buffer_overrun_errors;
+		ibp->n_vl15_dropped = 0;
+		ibp->z_vl15_dropped = cntrs.vl15_dropped;
+	}
+
+	return ret;
+}
+
+static int pma_set_portcounters_ext(struct ib_perf *pmp,
+				    struct ib_device *ibdev, u8 port)
+{
+	struct ib_pma_portcounters *p = (struct ib_pma_portcounters *)
+		pmp->data;
+	struct qib_ibport *ibp = to_iport(ibdev, port);
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	u64 swords, rwords, spkts, rpkts, xwait;
+
+	qib_snapshot_counters(ppd, &swords, &rwords, &spkts, &rpkts, &xwait);
+
+	if (p->counter_select & IB_PMA_SELX_PORT_XMIT_DATA)
+		ibp->z_port_xmit_data = swords;
+
+	if (p->counter_select & IB_PMA_SELX_PORT_RCV_DATA)
+		ibp->z_port_rcv_data = rwords;
+
+	if (p->counter_select & IB_PMA_SELX_PORT_XMIT_PACKETS)
+		ibp->z_port_xmit_packets = spkts;
+
+	if (p->counter_select & IB_PMA_SELX_PORT_RCV_PACKETS)
+		ibp->z_port_rcv_packets = rpkts;
+
+	if (p->counter_select & IB_PMA_SELX_PORT_UNI_XMIT_PACKETS)
+		ibp->n_unicast_xmit = 0;
+
+	if (p->counter_select & IB_PMA_SELX_PORT_UNI_RCV_PACKETS)
+		ibp->n_unicast_rcv = 0;
+
+	if (p->counter_select & IB_PMA_SELX_PORT_MULTI_XMIT_PACKETS)
+		ibp->n_multicast_xmit = 0;
+
+	if (p->counter_select & IB_PMA_SELX_PORT_MULTI_RCV_PACKETS)
+		ibp->n_multicast_rcv = 0;
+
+	return pma_get_portcounters_ext(pmp, ibdev, port);
+}
+
+static int process_subn(struct ib_device *ibdev, int mad_flags,
+			u8 port, struct ib_mad *in_mad,
+			struct ib_mad *out_mad)
+{
+	struct ib_smp *smp = (struct ib_smp *)out_mad;
+	struct qib_ibport *ibp = to_iport(ibdev, port);
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	int ret;
+
+	*out_mad = *in_mad;
+	if (smp->class_version != 1) {
+		smp->status |= IB_SMP_UNSUP_VERSION;
+		ret = reply(smp);
+		goto bail;
+	}
+
+	ret = check_mkey(ibp, smp, mad_flags);
+	if (ret) {
+		u32 port_num = be32_to_cpu(smp->attr_mod);
+
+		/*
+		 * If this is a get/set portinfo, we already check the
+		 * M_Key if the MAD is for another port and the M_Key
+		 * is OK on the receiving port. This check is needed
+		 * to increment the error counters when the M_Key
+		 * fails to match on *both* ports.
+		 */
+		if (in_mad->mad_hdr.attr_id == IB_SMP_ATTR_PORT_INFO &&
+		    (smp->method == IB_MGMT_METHOD_GET ||
+		     smp->method == IB_MGMT_METHOD_SET) &&
+		    port_num && port_num <= ibdev->phys_port_cnt &&
+		    port != port_num)
+			(void) check_mkey(to_iport(ibdev, port_num), smp, 0);
+		goto bail;
+	}
+
+	switch (smp->method) {
+	case IB_MGMT_METHOD_GET:
+		switch (smp->attr_id) {
+		case IB_SMP_ATTR_NODE_DESC:
+			ret = subn_get_nodedescription(smp, ibdev);
+			goto bail;
+		case IB_SMP_ATTR_NODE_INFO:
+			ret = subn_get_nodeinfo(smp, ibdev, port);
+			goto bail;
+		case IB_SMP_ATTR_GUID_INFO:
+			ret = subn_get_guidinfo(smp, ibdev, port);
+			goto bail;
+		case IB_SMP_ATTR_PORT_INFO:
+			ret = subn_get_portinfo(smp, ibdev, port);
+			goto bail;
+		case IB_SMP_ATTR_PKEY_TABLE:
+			ret = subn_get_pkeytable(smp, ibdev, port);
+			goto bail;
+		case IB_SMP_ATTR_SL_TO_VL_TABLE:
+			ret = subn_get_sl_to_vl(smp, ibdev, port);
+			goto bail;
+		case IB_SMP_ATTR_VL_ARB_TABLE:
+			ret = subn_get_vl_arb(smp, ibdev, port);
+			goto bail;
+		case IB_SMP_ATTR_SM_INFO:
+			if (ibp->port_cap_flags & IB_PORT_SM_DISABLED) {
+				ret = IB_MAD_RESULT_SUCCESS |
+					IB_MAD_RESULT_CONSUMED;
+				goto bail;
+			}
+			if (ibp->port_cap_flags & IB_PORT_SM) {
+				ret = IB_MAD_RESULT_SUCCESS;
+				goto bail;
+			}
+			/* FALLTHROUGH */
+		default:
+			smp->status |= IB_SMP_UNSUP_METH_ATTR;
+			ret = reply(smp);
+			goto bail;
+		}
+
+	case IB_MGMT_METHOD_SET:
+		switch (smp->attr_id) {
+		case IB_SMP_ATTR_GUID_INFO:
+			ret = subn_set_guidinfo(smp, ibdev, port);
+			goto bail;
+		case IB_SMP_ATTR_PORT_INFO:
+			ret = subn_set_portinfo(smp, ibdev, port);
+			goto bail;
+		case IB_SMP_ATTR_PKEY_TABLE:
+			ret = subn_set_pkeytable(smp, ibdev, port);
+			goto bail;
+		case IB_SMP_ATTR_SL_TO_VL_TABLE:
+			ret = subn_set_sl_to_vl(smp, ibdev, port);
+			goto bail;
+		case IB_SMP_ATTR_VL_ARB_TABLE:
+			ret = subn_set_vl_arb(smp, ibdev, port);
+			goto bail;
+		case IB_SMP_ATTR_SM_INFO:
+			if (ibp->port_cap_flags & IB_PORT_SM_DISABLED) {
+				ret = IB_MAD_RESULT_SUCCESS |
+					IB_MAD_RESULT_CONSUMED;
+				goto bail;
+			}
+			if (ibp->port_cap_flags & IB_PORT_SM) {
+				ret = IB_MAD_RESULT_SUCCESS;
+				goto bail;
+			}
+			/* FALLTHROUGH */
+		default:
+			smp->status |= IB_SMP_UNSUP_METH_ATTR;
+			ret = reply(smp);
+			goto bail;
+		}
+
+	case IB_MGMT_METHOD_TRAP_REPRESS:
+		if (smp->attr_id == IB_SMP_ATTR_NOTICE)
+			ret = subn_trap_repress(smp, ibdev, port);
+		else {
+			smp->status |= IB_SMP_UNSUP_METH_ATTR;
+			ret = reply(smp);
+		}
+		goto bail;
+
+	case IB_MGMT_METHOD_TRAP:
+	case IB_MGMT_METHOD_REPORT:
+	case IB_MGMT_METHOD_REPORT_RESP:
+	case IB_MGMT_METHOD_GET_RESP:
+		/*
+		 * The ib_mad module will call us to process responses
+		 * before checking for other consumers.
+		 * Just tell the caller to process it normally.
+		 */
+		ret = IB_MAD_RESULT_SUCCESS;
+		goto bail;
+
+	case IB_MGMT_METHOD_SEND:
+		if (ib_get_smp_direction(smp) &&
+		    smp->attr_id == QIB_VENDOR_IPG) {
+			ppd->dd->f_set_ib_cfg(ppd, QIB_IB_CFG_PORT,
+					      smp->data[0]);
+			ret = IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_CONSUMED;
+		} else
+			ret = IB_MAD_RESULT_SUCCESS;
+		goto bail;
+
+	default:
+		smp->status |= IB_SMP_UNSUP_METHOD;
+		ret = reply(smp);
+	}
+
+bail:
+	return ret;
+}
+
+static int process_perf(struct ib_device *ibdev, u8 port,
+			struct ib_mad *in_mad,
+			struct ib_mad *out_mad)
+{
+	struct ib_perf *pmp = (struct ib_perf *)out_mad;
+	int ret;
+
+	*out_mad = *in_mad;
+	if (pmp->class_version != 1) {
+		pmp->status |= IB_SMP_UNSUP_VERSION;
+		ret = reply((struct ib_smp *) pmp);
+		goto bail;
+	}
+
+	switch (pmp->method) {
+	case IB_MGMT_METHOD_GET:
+		switch (pmp->attr_id) {
+		case IB_PMA_CLASS_PORT_INFO:
+			ret = pma_get_classportinfo(pmp, ibdev);
+			goto bail;
+		case IB_PMA_PORT_SAMPLES_CONTROL:
+			ret = pma_get_portsamplescontrol(pmp, ibdev, port);
+			goto bail;
+		case IB_PMA_PORT_SAMPLES_RESULT:
+			ret = pma_get_portsamplesresult(pmp, ibdev, port);
+			goto bail;
+		case IB_PMA_PORT_SAMPLES_RESULT_EXT:
+			ret = pma_get_portsamplesresult_ext(pmp, ibdev, port);
+			goto bail;
+		case IB_PMA_PORT_COUNTERS:
+			ret = pma_get_portcounters(pmp, ibdev, port);
+			goto bail;
+		case IB_PMA_PORT_COUNTERS_EXT:
+			ret = pma_get_portcounters_ext(pmp, ibdev, port);
+			goto bail;
+		case IB_PMA_PORT_COUNTERS_CONG:
+			ret = pma_get_portcounters_cong(pmp, ibdev, port);
+			goto bail;
+		default:
+			pmp->status |= IB_SMP_UNSUP_METH_ATTR;
+			ret = reply((struct ib_smp *) pmp);
+			goto bail;
+		}
+
+	case IB_MGMT_METHOD_SET:
+		switch (pmp->attr_id) {
+		case IB_PMA_PORT_SAMPLES_CONTROL:
+			ret = pma_set_portsamplescontrol(pmp, ibdev, port);
+			goto bail;
+		case IB_PMA_PORT_COUNTERS:
+			ret = pma_set_portcounters(pmp, ibdev, port);
+			goto bail;
+		case IB_PMA_PORT_COUNTERS_EXT:
+			ret = pma_set_portcounters_ext(pmp, ibdev, port);
+			goto bail;
+		case IB_PMA_PORT_COUNTERS_CONG:
+			ret = pma_set_portcounters_cong(pmp, ibdev, port);
+			goto bail;
+		default:
+			pmp->status |= IB_SMP_UNSUP_METH_ATTR;
+			ret = reply((struct ib_smp *) pmp);
+			goto bail;
+		}
+
+	case IB_MGMT_METHOD_TRAP:
+	case IB_MGMT_METHOD_GET_RESP:
+		/*
+		 * The ib_mad module will call us to process responses
+		 * before checking for other consumers.
+		 * Just tell the caller to process it normally.
+		 */
+		ret = IB_MAD_RESULT_SUCCESS;
+		goto bail;
+
+	default:
+		pmp->status |= IB_SMP_UNSUP_METHOD;
+		ret = reply((struct ib_smp *) pmp);
+	}
+
+bail:
+	return ret;
+}
+
+/**
+ * qib_process_mad - process an incoming MAD packet
+ * @ibdev: the infiniband device this packet came in on
+ * @mad_flags: MAD flags
+ * @port: the port number this packet came in on
+ * @in_wc: the work completion entry for this packet
+ * @in_grh: the global route header for this packet
+ * @in_mad: the incoming MAD
+ * @out_mad: any outgoing MAD reply
+ *
+ * Returns IB_MAD_RESULT_SUCCESS if this is a MAD that we are not
+ * interested in processing.
+ *
+ * Note that the verbs framework has already done the MAD sanity checks,
+ * and hop count/pointer updating for IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE
+ * MADs.
+ *
+ * This is called by the ib_mad module.
+ */
+int qib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port,
+		    struct ib_wc *in_wc, struct ib_grh *in_grh,
+		    struct ib_mad *in_mad, struct ib_mad *out_mad)
+{
+	int ret;
+
+	switch (in_mad->mad_hdr.mgmt_class) {
+	case IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE:
+	case IB_MGMT_CLASS_SUBN_LID_ROUTED:
+		ret = process_subn(ibdev, mad_flags, port, in_mad, out_mad);
+		goto bail;
+
+	case IB_MGMT_CLASS_PERF_MGMT:
+		ret = process_perf(ibdev, port, in_mad, out_mad);
+		goto bail;
+
+	default:
+		ret = IB_MAD_RESULT_SUCCESS;
+	}
+
+bail:
+	return ret;
+}
+
+static void send_handler(struct ib_mad_agent *agent,
+			 struct ib_mad_send_wc *mad_send_wc)
+{
+	ib_free_send_mad(mad_send_wc->send_buf);
+}
+
+static void xmit_wait_timer_func(unsigned long opaque)
+{
+	struct qib_pportdata *ppd = (struct qib_pportdata *)opaque;
+	struct qib_devdata *dd = dd_from_ppd(ppd);
+	unsigned long flags;
+	u8 status;
+
+	spin_lock_irqsave(&ppd->ibport_data.lock, flags);
+	if (ppd->cong_stats.flags == IB_PMA_CONG_HW_CONTROL_SAMPLE) {
+		status = dd->f_portcntr(ppd, QIBPORTCNTR_PSSTAT);
+		if (status == IB_PMA_SAMPLE_STATUS_DONE) {
+			/* save counter cache */
+			cache_hw_sample_counters(ppd);
+			ppd->cong_stats.flags = IB_PMA_CONG_HW_CONTROL_TIMER;
+		} else
+			goto done;
+	}
+	ppd->cong_stats.counter = xmit_wait_get_value_delta(ppd);
+	dd->f_set_cntr_sample(ppd, QIB_CONG_TIMER_PSINTERVAL, 0x0);
+done:
+	spin_unlock_irqrestore(&ppd->ibport_data.lock, flags);
+	mod_timer(&ppd->cong_stats.timer, jiffies + HZ);
+}
+
+int qib_create_agents(struct qib_ibdev *dev)
+{
+	struct qib_devdata *dd = dd_from_dev(dev);
+	struct ib_mad_agent *agent;
+	struct qib_ibport *ibp;
+	int p;
+	int ret;
+
+	for (p = 0; p < dd->num_pports; p++) {
+		ibp = &dd->pport[p].ibport_data;
+		agent = ib_register_mad_agent(&dev->ibdev, p + 1, IB_QPT_SMI,
+					      NULL, 0, send_handler,
+					      NULL, NULL);
+		if (IS_ERR(agent)) {
+			ret = PTR_ERR(agent);
+			goto err;
+		}
+
+		/* Initialize xmit_wait structure */
+		dd->pport[p].cong_stats.counter = 0;
+		init_timer(&dd->pport[p].cong_stats.timer);
+		dd->pport[p].cong_stats.timer.function = xmit_wait_timer_func;
+		dd->pport[p].cong_stats.timer.data =
+			(unsigned long)(&dd->pport[p]);
+		dd->pport[p].cong_stats.timer.expires = 0;
+		add_timer(&dd->pport[p].cong_stats.timer);
+
+		ibp->send_agent = agent;
+	}
+
+	return 0;
+
+err:
+	for (p = 0; p < dd->num_pports; p++) {
+		ibp = &dd->pport[p].ibport_data;
+		if (ibp->send_agent) {
+			agent = ibp->send_agent;
+			ibp->send_agent = NULL;
+			ib_unregister_mad_agent(agent);
+		}
+	}
+
+	return ret;
+}
+
+void qib_free_agents(struct qib_ibdev *dev)
+{
+	struct qib_devdata *dd = dd_from_dev(dev);
+	struct ib_mad_agent *agent;
+	struct qib_ibport *ibp;
+	int p;
+
+	for (p = 0; p < dd->num_pports; p++) {
+		ibp = &dd->pport[p].ibport_data;
+		if (ibp->send_agent) {
+			agent = ibp->send_agent;
+			ibp->send_agent = NULL;
+			ib_unregister_mad_agent(agent);
+		}
+		if (ibp->sm_ah) {
+			ib_destroy_ah(&ibp->sm_ah->ibah);
+			ibp->sm_ah = NULL;
+		}
+		if (dd->pport[p].cong_stats.timer.data)
+			del_timer_sync(&dd->pport[p].cong_stats.timer);
+	}
+}
diff --git a/drivers/infiniband/hw/qib/qib_mad.h b/drivers/infiniband/hw/qib/qib_mad.h
new file mode 100644
index 000000000000..147aff9117d7
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_mad.h
@@ -0,0 +1,373 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009, 2010 QLogic Corporation.
+ * All rights reserved.
+ * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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.
+ */
+
+#define IB_SMP_UNSUP_VERSION    cpu_to_be16(0x0004)
+#define IB_SMP_UNSUP_METHOD     cpu_to_be16(0x0008)
+#define IB_SMP_UNSUP_METH_ATTR  cpu_to_be16(0x000C)
+#define IB_SMP_INVALID_FIELD    cpu_to_be16(0x001C)
+
+struct ib_node_info {
+	u8 base_version;
+	u8 class_version;
+	u8 node_type;
+	u8 num_ports;
+	__be64 sys_guid;
+	__be64 node_guid;
+	__be64 port_guid;
+	__be16 partition_cap;
+	__be16 device_id;
+	__be32 revision;
+	u8 local_port_num;
+	u8 vendor_id[3];
+} __attribute__ ((packed));
+
+struct ib_mad_notice_attr {
+	u8 generic_type;
+	u8 prod_type_msb;
+	__be16 prod_type_lsb;
+	__be16 trap_num;
+	__be16 issuer_lid;
+	__be16 toggle_count;
+
+	union {
+		struct {
+			u8	details[54];
+		} raw_data;
+
+		struct {
+			__be16	reserved;
+			__be16	lid;		/* where violation happened */
+			u8	port_num;	/* where violation happened */
+		} __attribute__ ((packed)) ntc_129_131;
+
+		struct {
+			__be16	reserved;
+			__be16	lid;		/* LID where change occured */
+			u8	reserved2;
+			u8	local_changes;	/* low bit - local changes */
+			__be32	new_cap_mask;	/* new capability mask */
+			u8	reserved3;
+			u8	change_flags;	/* low 3 bits only */
+		} __attribute__ ((packed)) ntc_144;
+
+		struct {
+			__be16	reserved;
+			__be16	lid;		/* lid where sys guid changed */
+			__be16	reserved2;
+			__be64	new_sys_guid;
+		} __attribute__ ((packed)) ntc_145;
+
+		struct {
+			__be16	reserved;
+			__be16	lid;
+			__be16	dr_slid;
+			u8	method;
+			u8	reserved2;
+			__be16	attr_id;
+			__be32	attr_mod;
+			__be64	mkey;
+			u8	reserved3;
+			u8	dr_trunc_hop;
+			u8	dr_rtn_path[30];
+		} __attribute__ ((packed)) ntc_256;
+
+		struct {
+			__be16		reserved;
+			__be16		lid1;
+			__be16		lid2;
+			__be32		key;
+			__be32		sl_qp1;	/* SL: high 4 bits */
+			__be32		qp2;	/* high 8 bits reserved */
+			union ib_gid	gid1;
+			union ib_gid	gid2;
+		} __attribute__ ((packed)) ntc_257_258;
+
+	} details;
+};
+
+/*
+ * Generic trap/notice types
+ */
+#define IB_NOTICE_TYPE_FATAL	0x80
+#define IB_NOTICE_TYPE_URGENT	0x81
+#define IB_NOTICE_TYPE_SECURITY	0x82
+#define IB_NOTICE_TYPE_SM	0x83
+#define IB_NOTICE_TYPE_INFO	0x84
+
+/*
+ * Generic trap/notice producers
+ */
+#define IB_NOTICE_PROD_CA		cpu_to_be16(1)
+#define IB_NOTICE_PROD_SWITCH		cpu_to_be16(2)
+#define IB_NOTICE_PROD_ROUTER		cpu_to_be16(3)
+#define IB_NOTICE_PROD_CLASS_MGR	cpu_to_be16(4)
+
+/*
+ * Generic trap/notice numbers
+ */
+#define IB_NOTICE_TRAP_LLI_THRESH	cpu_to_be16(129)
+#define IB_NOTICE_TRAP_EBO_THRESH	cpu_to_be16(130)
+#define IB_NOTICE_TRAP_FLOW_UPDATE	cpu_to_be16(131)
+#define IB_NOTICE_TRAP_CAP_MASK_CHG	cpu_to_be16(144)
+#define IB_NOTICE_TRAP_SYS_GUID_CHG	cpu_to_be16(145)
+#define IB_NOTICE_TRAP_BAD_MKEY		cpu_to_be16(256)
+#define IB_NOTICE_TRAP_BAD_PKEY		cpu_to_be16(257)
+#define IB_NOTICE_TRAP_BAD_QKEY		cpu_to_be16(258)
+
+/*
+ * Repress trap/notice flags
+ */
+#define IB_NOTICE_REPRESS_LLI_THRESH	(1 << 0)
+#define IB_NOTICE_REPRESS_EBO_THRESH	(1 << 1)
+#define IB_NOTICE_REPRESS_FLOW_UPDATE	(1 << 2)
+#define IB_NOTICE_REPRESS_CAP_MASK_CHG	(1 << 3)
+#define IB_NOTICE_REPRESS_SYS_GUID_CHG	(1 << 4)
+#define IB_NOTICE_REPRESS_BAD_MKEY	(1 << 5)
+#define IB_NOTICE_REPRESS_BAD_PKEY	(1 << 6)
+#define IB_NOTICE_REPRESS_BAD_QKEY	(1 << 7)
+
+/*
+ * Generic trap/notice other local changes flags (trap 144).
+ */
+#define IB_NOTICE_TRAP_LSE_CHG		0x04	/* Link Speed Enable changed */
+#define IB_NOTICE_TRAP_LWE_CHG		0x02	/* Link Width Enable changed */
+#define IB_NOTICE_TRAP_NODE_DESC_CHG	0x01
+
+/*
+ * Generic trap/notice M_Key volation flags in dr_trunc_hop (trap 256).
+ */
+#define IB_NOTICE_TRAP_DR_NOTICE	0x80
+#define IB_NOTICE_TRAP_DR_TRUNC		0x40
+
+struct ib_vl_weight_elem {
+	u8      vl;     /* Only low 4 bits, upper 4 bits reserved */
+	u8      weight;
+};
+
+#define IB_VLARB_LOWPRI_0_31    1
+#define IB_VLARB_LOWPRI_32_63   2
+#define IB_VLARB_HIGHPRI_0_31   3
+#define IB_VLARB_HIGHPRI_32_63  4
+
+/*
+ * PMA class portinfo capability mask bits
+ */
+#define IB_PMA_CLASS_CAP_ALLPORTSELECT  cpu_to_be16(1 << 8)
+#define IB_PMA_CLASS_CAP_EXT_WIDTH      cpu_to_be16(1 << 9)
+#define IB_PMA_CLASS_CAP_XMIT_WAIT      cpu_to_be16(1 << 12)
+
+#define IB_PMA_CLASS_PORT_INFO          cpu_to_be16(0x0001)
+#define IB_PMA_PORT_SAMPLES_CONTROL     cpu_to_be16(0x0010)
+#define IB_PMA_PORT_SAMPLES_RESULT      cpu_to_be16(0x0011)
+#define IB_PMA_PORT_COUNTERS            cpu_to_be16(0x0012)
+#define IB_PMA_PORT_COUNTERS_EXT        cpu_to_be16(0x001D)
+#define IB_PMA_PORT_SAMPLES_RESULT_EXT  cpu_to_be16(0x001E)
+#define IB_PMA_PORT_COUNTERS_CONG       cpu_to_be16(0xFF00)
+
+struct ib_perf {
+	u8 base_version;
+	u8 mgmt_class;
+	u8 class_version;
+	u8 method;
+	__be16 status;
+	__be16 unused;
+	__be64 tid;
+	__be16 attr_id;
+	__be16 resv;
+	__be32 attr_mod;
+	u8 reserved[40];
+	u8 data[192];
+} __attribute__ ((packed));
+
+struct ib_pma_classportinfo {
+	u8 base_version;
+	u8 class_version;
+	__be16 cap_mask;
+	u8 reserved[3];
+	u8 resp_time_value;     /* only lower 5 bits */
+	union ib_gid redirect_gid;
+	__be32 redirect_tc_sl_fl;       /* 8, 4, 20 bits respectively */
+	__be16 redirect_lid;
+	__be16 redirect_pkey;
+	__be32 redirect_qp;     /* only lower 24 bits */
+	__be32 redirect_qkey;
+	union ib_gid trap_gid;
+	__be32 trap_tc_sl_fl;   /* 8, 4, 20 bits respectively */
+	__be16 trap_lid;
+	__be16 trap_pkey;
+	__be32 trap_hl_qp;      /* 8, 24 bits respectively */
+	__be32 trap_qkey;
+} __attribute__ ((packed));
+
+struct ib_pma_portsamplescontrol {
+	u8 opcode;
+	u8 port_select;
+	u8 tick;
+	u8 counter_width;       /* only lower 3 bits */
+	__be32 counter_mask0_9; /* 2, 10 * 3, bits */
+	__be16 counter_mask10_14;       /* 1, 5 * 3, bits */
+	u8 sample_mechanisms;
+	u8 sample_status;       /* only lower 2 bits */
+	__be64 option_mask;
+	__be64 vendor_mask;
+	__be32 sample_start;
+	__be32 sample_interval;
+	__be16 tag;
+	__be16 counter_select[15];
+} __attribute__ ((packed));
+
+struct ib_pma_portsamplesresult {
+	__be16 tag;
+	__be16 sample_status;   /* only lower 2 bits */
+	__be32 counter[15];
+} __attribute__ ((packed));
+
+struct ib_pma_portsamplesresult_ext {
+	__be16 tag;
+	__be16 sample_status;   /* only lower 2 bits */
+	__be32 extended_width;  /* only upper 2 bits */
+	__be64 counter[15];
+} __attribute__ ((packed));
+
+struct ib_pma_portcounters {
+	u8 reserved;
+	u8 port_select;
+	__be16 counter_select;
+	__be16 symbol_error_counter;
+	u8 link_error_recovery_counter;
+	u8 link_downed_counter;
+	__be16 port_rcv_errors;
+	__be16 port_rcv_remphys_errors;
+	__be16 port_rcv_switch_relay_errors;
+	__be16 port_xmit_discards;
+	u8 port_xmit_constraint_errors;
+	u8 port_rcv_constraint_errors;
+	u8 reserved1;
+	u8 lli_ebor_errors;     /* 4, 4, bits */
+	__be16 reserved2;
+	__be16 vl15_dropped;
+	__be32 port_xmit_data;
+	__be32 port_rcv_data;
+	__be32 port_xmit_packets;
+	__be32 port_rcv_packets;
+} __attribute__ ((packed));
+
+struct ib_pma_portcounters_cong {
+	u8 reserved;
+	u8 reserved1;
+	__be16 port_check_rate;
+	__be16 symbol_error_counter;
+	u8 link_error_recovery_counter;
+	u8 link_downed_counter;
+	__be16 port_rcv_errors;
+	__be16 port_rcv_remphys_errors;
+	__be16 port_rcv_switch_relay_errors;
+	__be16 port_xmit_discards;
+	u8 port_xmit_constraint_errors;
+	u8 port_rcv_constraint_errors;
+	u8 reserved2;
+	u8 lli_ebor_errors;    /* 4, 4, bits */
+	__be16 reserved3;
+	__be16 vl15_dropped;
+	__be64 port_xmit_data;
+	__be64 port_rcv_data;
+	__be64 port_xmit_packets;
+	__be64 port_rcv_packets;
+	__be64 port_xmit_wait;
+	__be64 port_adr_events;
+} __attribute__ ((packed));
+
+#define IB_PMA_CONG_HW_CONTROL_TIMER            0x00
+#define IB_PMA_CONG_HW_CONTROL_SAMPLE           0x01
+
+#define QIB_XMIT_RATE_UNSUPPORTED               0x0
+#define QIB_XMIT_RATE_PICO                      0x7
+/* number of 4nsec cycles equaling 2secs */
+#define QIB_CONG_TIMER_PSINTERVAL               0x1DCD64EC
+
+#define IB_PMA_SEL_SYMBOL_ERROR                 cpu_to_be16(0x0001)
+#define IB_PMA_SEL_LINK_ERROR_RECOVERY          cpu_to_be16(0x0002)
+#define IB_PMA_SEL_LINK_DOWNED                  cpu_to_be16(0x0004)
+#define IB_PMA_SEL_PORT_RCV_ERRORS              cpu_to_be16(0x0008)
+#define IB_PMA_SEL_PORT_RCV_REMPHYS_ERRORS      cpu_to_be16(0x0010)
+#define IB_PMA_SEL_PORT_XMIT_DISCARDS           cpu_to_be16(0x0040)
+#define IB_PMA_SEL_LOCAL_LINK_INTEGRITY_ERRORS  cpu_to_be16(0x0200)
+#define IB_PMA_SEL_EXCESSIVE_BUFFER_OVERRUNS    cpu_to_be16(0x0400)
+#define IB_PMA_SEL_PORT_VL15_DROPPED            cpu_to_be16(0x0800)
+#define IB_PMA_SEL_PORT_XMIT_DATA               cpu_to_be16(0x1000)
+#define IB_PMA_SEL_PORT_RCV_DATA                cpu_to_be16(0x2000)
+#define IB_PMA_SEL_PORT_XMIT_PACKETS            cpu_to_be16(0x4000)
+#define IB_PMA_SEL_PORT_RCV_PACKETS             cpu_to_be16(0x8000)
+
+#define IB_PMA_SEL_CONG_ALL                     0x01
+#define IB_PMA_SEL_CONG_PORT_DATA               0x02
+#define IB_PMA_SEL_CONG_XMIT                    0x04
+#define IB_PMA_SEL_CONG_ROUTING                 0x08
+
+struct ib_pma_portcounters_ext {
+	u8 reserved;
+	u8 port_select;
+	__be16 counter_select;
+	__be32 reserved1;
+	__be64 port_xmit_data;
+	__be64 port_rcv_data;
+	__be64 port_xmit_packets;
+	__be64 port_rcv_packets;
+	__be64 port_unicast_xmit_packets;
+	__be64 port_unicast_rcv_packets;
+	__be64 port_multicast_xmit_packets;
+	__be64 port_multicast_rcv_packets;
+} __attribute__ ((packed));
+
+#define IB_PMA_SELX_PORT_XMIT_DATA              cpu_to_be16(0x0001)
+#define IB_PMA_SELX_PORT_RCV_DATA               cpu_to_be16(0x0002)
+#define IB_PMA_SELX_PORT_XMIT_PACKETS           cpu_to_be16(0x0004)
+#define IB_PMA_SELX_PORT_RCV_PACKETS            cpu_to_be16(0x0008)
+#define IB_PMA_SELX_PORT_UNI_XMIT_PACKETS       cpu_to_be16(0x0010)
+#define IB_PMA_SELX_PORT_UNI_RCV_PACKETS        cpu_to_be16(0x0020)
+#define IB_PMA_SELX_PORT_MULTI_XMIT_PACKETS     cpu_to_be16(0x0040)
+#define IB_PMA_SELX_PORT_MULTI_RCV_PACKETS      cpu_to_be16(0x0080)
+
+/*
+ * The PortSamplesControl.CounterMasks field is an array of 3 bit fields
+ * which specify the N'th counter's capabilities. See ch. 16.1.3.2.
+ * We support 5 counters which only count the mandatory quantities.
+ */
+#define COUNTER_MASK(q, n) (q << ((9 - n) * 3))
+#define COUNTER_MASK0_9 \
+	cpu_to_be32(COUNTER_MASK(1, 0) | \
+		    COUNTER_MASK(1, 1) | \
+		    COUNTER_MASK(1, 2) | \
+		    COUNTER_MASK(1, 3) | \
+		    COUNTER_MASK(1, 4))
diff --git a/drivers/infiniband/hw/qib/qib_mmap.c b/drivers/infiniband/hw/qib/qib_mmap.c
new file mode 100644
index 000000000000..8b73a11d571c
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_mmap.c
@@ -0,0 +1,174 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/module.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+#include <linux/mm.h>
+#include <linux/errno.h>
+#include <asm/pgtable.h>
+
+#include "qib_verbs.h"
+
+/**
+ * qib_release_mmap_info - free mmap info structure
+ * @ref: a pointer to the kref within struct qib_mmap_info
+ */
+void qib_release_mmap_info(struct kref *ref)
+{
+	struct qib_mmap_info *ip =
+		container_of(ref, struct qib_mmap_info, ref);
+	struct qib_ibdev *dev = to_idev(ip->context->device);
+
+	spin_lock_irq(&dev->pending_lock);
+	list_del(&ip->pending_mmaps);
+	spin_unlock_irq(&dev->pending_lock);
+
+	vfree(ip->obj);
+	kfree(ip);
+}
+
+/*
+ * open and close keep track of how many times the CQ is mapped,
+ * to avoid releasing it.
+ */
+static void qib_vma_open(struct vm_area_struct *vma)
+{
+	struct qib_mmap_info *ip = vma->vm_private_data;
+
+	kref_get(&ip->ref);
+}
+
+static void qib_vma_close(struct vm_area_struct *vma)
+{
+	struct qib_mmap_info *ip = vma->vm_private_data;
+
+	kref_put(&ip->ref, qib_release_mmap_info);
+}
+
+static struct vm_operations_struct qib_vm_ops = {
+	.open =     qib_vma_open,
+	.close =    qib_vma_close,
+};
+
+/**
+ * qib_mmap - create a new mmap region
+ * @context: the IB user context of the process making the mmap() call
+ * @vma: the VMA to be initialized
+ * Return zero if the mmap is OK. Otherwise, return an errno.
+ */
+int qib_mmap(struct ib_ucontext *context, struct vm_area_struct *vma)
+{
+	struct qib_ibdev *dev = to_idev(context->device);
+	unsigned long offset = vma->vm_pgoff << PAGE_SHIFT;
+	unsigned long size = vma->vm_end - vma->vm_start;
+	struct qib_mmap_info *ip, *pp;
+	int ret = -EINVAL;
+
+	/*
+	 * Search the device's list of objects waiting for a mmap call.
+	 * Normally, this list is very short since a call to create a
+	 * CQ, QP, or SRQ is soon followed by a call to mmap().
+	 */
+	spin_lock_irq(&dev->pending_lock);
+	list_for_each_entry_safe(ip, pp, &dev->pending_mmaps,
+				 pending_mmaps) {
+		/* Only the creator is allowed to mmap the object */
+		if (context != ip->context || (__u64) offset != ip->offset)
+			continue;
+		/* Don't allow a mmap larger than the object. */
+		if (size > ip->size)
+			break;
+
+		list_del_init(&ip->pending_mmaps);
+		spin_unlock_irq(&dev->pending_lock);
+
+		ret = remap_vmalloc_range(vma, ip->obj, 0);
+		if (ret)
+			goto done;
+		vma->vm_ops = &qib_vm_ops;
+		vma->vm_private_data = ip;
+		qib_vma_open(vma);
+		goto done;
+	}
+	spin_unlock_irq(&dev->pending_lock);
+done:
+	return ret;
+}
+
+/*
+ * Allocate information for qib_mmap
+ */
+struct qib_mmap_info *qib_create_mmap_info(struct qib_ibdev *dev,
+					   u32 size,
+					   struct ib_ucontext *context,
+					   void *obj) {
+	struct qib_mmap_info *ip;
+
+	ip = kmalloc(sizeof *ip, GFP_KERNEL);
+	if (!ip)
+		goto bail;
+
+	size = PAGE_ALIGN(size);
+
+	spin_lock_irq(&dev->mmap_offset_lock);
+	if (dev->mmap_offset == 0)
+		dev->mmap_offset = PAGE_SIZE;
+	ip->offset = dev->mmap_offset;
+	dev->mmap_offset += size;
+	spin_unlock_irq(&dev->mmap_offset_lock);
+
+	INIT_LIST_HEAD(&ip->pending_mmaps);
+	ip->size = size;
+	ip->context = context;
+	ip->obj = obj;
+	kref_init(&ip->ref);
+
+bail:
+	return ip;
+}
+
+void qib_update_mmap_info(struct qib_ibdev *dev, struct qib_mmap_info *ip,
+			  u32 size, void *obj)
+{
+	size = PAGE_ALIGN(size);
+
+	spin_lock_irq(&dev->mmap_offset_lock);
+	if (dev->mmap_offset == 0)
+		dev->mmap_offset = PAGE_SIZE;
+	ip->offset = dev->mmap_offset;
+	dev->mmap_offset += size;
+	spin_unlock_irq(&dev->mmap_offset_lock);
+
+	ip->size = size;
+	ip->obj = obj;
+}
diff --git a/drivers/infiniband/hw/qib/qib_mr.c b/drivers/infiniband/hw/qib/qib_mr.c
new file mode 100644
index 000000000000..5f95f0f6385d
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_mr.c
@@ -0,0 +1,503 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <rdma/ib_umem.h>
+#include <rdma/ib_smi.h>
+
+#include "qib.h"
+
+/* Fast memory region */
+struct qib_fmr {
+	struct ib_fmr ibfmr;
+	u8 page_shift;
+	struct qib_mregion mr;        /* must be last */
+};
+
+static inline struct qib_fmr *to_ifmr(struct ib_fmr *ibfmr)
+{
+	return container_of(ibfmr, struct qib_fmr, ibfmr);
+}
+
+/**
+ * qib_get_dma_mr - get a DMA memory region
+ * @pd: protection domain for this memory region
+ * @acc: access flags
+ *
+ * Returns the memory region on success, otherwise returns an errno.
+ * Note that all DMA addresses should be created via the
+ * struct ib_dma_mapping_ops functions (see qib_dma.c).
+ */
+struct ib_mr *qib_get_dma_mr(struct ib_pd *pd, int acc)
+{
+	struct qib_ibdev *dev = to_idev(pd->device);
+	struct qib_mr *mr;
+	struct ib_mr *ret;
+	unsigned long flags;
+
+	if (to_ipd(pd)->user) {
+		ret = ERR_PTR(-EPERM);
+		goto bail;
+	}
+
+	mr = kzalloc(sizeof *mr, GFP_KERNEL);
+	if (!mr) {
+		ret = ERR_PTR(-ENOMEM);
+		goto bail;
+	}
+
+	mr->mr.access_flags = acc;
+	atomic_set(&mr->mr.refcount, 0);
+
+	spin_lock_irqsave(&dev->lk_table.lock, flags);
+	if (!dev->dma_mr)
+		dev->dma_mr = &mr->mr;
+	spin_unlock_irqrestore(&dev->lk_table.lock, flags);
+
+	ret = &mr->ibmr;
+
+bail:
+	return ret;
+}
+
+static struct qib_mr *alloc_mr(int count, struct qib_lkey_table *lk_table)
+{
+	struct qib_mr *mr;
+	int m, i = 0;
+
+	/* Allocate struct plus pointers to first level page tables. */
+	m = (count + QIB_SEGSZ - 1) / QIB_SEGSZ;
+	mr = kmalloc(sizeof *mr + m * sizeof mr->mr.map[0], GFP_KERNEL);
+	if (!mr)
+		goto done;
+
+	/* Allocate first level page tables. */
+	for (; i < m; i++) {
+		mr->mr.map[i] = kmalloc(sizeof *mr->mr.map[0], GFP_KERNEL);
+		if (!mr->mr.map[i])
+			goto bail;
+	}
+	mr->mr.mapsz = m;
+	mr->mr.max_segs = count;
+
+	/*
+	 * ib_reg_phys_mr() will initialize mr->ibmr except for
+	 * lkey and rkey.
+	 */
+	if (!qib_alloc_lkey(lk_table, &mr->mr))
+		goto bail;
+	mr->ibmr.lkey = mr->mr.lkey;
+	mr->ibmr.rkey = mr->mr.lkey;
+
+	atomic_set(&mr->mr.refcount, 0);
+	goto done;
+
+bail:
+	while (i)
+		kfree(mr->mr.map[--i]);
+	kfree(mr);
+	mr = NULL;
+
+done:
+	return mr;
+}
+
+/**
+ * qib_reg_phys_mr - register a physical memory region
+ * @pd: protection domain for this memory region
+ * @buffer_list: pointer to the list of physical buffers to register
+ * @num_phys_buf: the number of physical buffers to register
+ * @iova_start: the starting address passed over IB which maps to this MR
+ *
+ * Returns the memory region on success, otherwise returns an errno.
+ */
+struct ib_mr *qib_reg_phys_mr(struct ib_pd *pd,
+			      struct ib_phys_buf *buffer_list,
+			      int num_phys_buf, int acc, u64 *iova_start)
+{
+	struct qib_mr *mr;
+	int n, m, i;
+	struct ib_mr *ret;
+
+	mr = alloc_mr(num_phys_buf, &to_idev(pd->device)->lk_table);
+	if (mr == NULL) {
+		ret = ERR_PTR(-ENOMEM);
+		goto bail;
+	}
+
+	mr->mr.pd = pd;
+	mr->mr.user_base = *iova_start;
+	mr->mr.iova = *iova_start;
+	mr->mr.length = 0;
+	mr->mr.offset = 0;
+	mr->mr.access_flags = acc;
+	mr->umem = NULL;
+
+	m = 0;
+	n = 0;
+	for (i = 0; i < num_phys_buf; i++) {
+		mr->mr.map[m]->segs[n].vaddr = (void *) buffer_list[i].addr;
+		mr->mr.map[m]->segs[n].length = buffer_list[i].size;
+		mr->mr.length += buffer_list[i].size;
+		n++;
+		if (n == QIB_SEGSZ) {
+			m++;
+			n = 0;
+		}
+	}
+
+	ret = &mr->ibmr;
+
+bail:
+	return ret;
+}
+
+/**
+ * qib_reg_user_mr - register a userspace memory region
+ * @pd: protection domain for this memory region
+ * @start: starting userspace address
+ * @length: length of region to register
+ * @virt_addr: virtual address to use (from HCA's point of view)
+ * @mr_access_flags: access flags for this memory region
+ * @udata: unused by the QLogic_IB driver
+ *
+ * Returns the memory region on success, otherwise returns an errno.
+ */
+struct ib_mr *qib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
+			      u64 virt_addr, int mr_access_flags,
+			      struct ib_udata *udata)
+{
+	struct qib_mr *mr;
+	struct ib_umem *umem;
+	struct ib_umem_chunk *chunk;
+	int n, m, i;
+	struct ib_mr *ret;
+
+	if (length == 0) {
+		ret = ERR_PTR(-EINVAL);
+		goto bail;
+	}
+
+	umem = ib_umem_get(pd->uobject->context, start, length,
+			   mr_access_flags, 0);
+	if (IS_ERR(umem))
+		return (void *) umem;
+
+	n = 0;
+	list_for_each_entry(chunk, &umem->chunk_list, list)
+		n += chunk->nents;
+
+	mr = alloc_mr(n, &to_idev(pd->device)->lk_table);
+	if (!mr) {
+		ret = ERR_PTR(-ENOMEM);
+		ib_umem_release(umem);
+		goto bail;
+	}
+
+	mr->mr.pd = pd;
+	mr->mr.user_base = start;
+	mr->mr.iova = virt_addr;
+	mr->mr.length = length;
+	mr->mr.offset = umem->offset;
+	mr->mr.access_flags = mr_access_flags;
+	mr->umem = umem;
+
+	m = 0;
+	n = 0;
+	list_for_each_entry(chunk, &umem->chunk_list, list) {
+		for (i = 0; i < chunk->nents; i++) {
+			void *vaddr;
+
+			vaddr = page_address(sg_page(&chunk->page_list[i]));
+			if (!vaddr) {
+				ret = ERR_PTR(-EINVAL);
+				goto bail;
+			}
+			mr->mr.map[m]->segs[n].vaddr = vaddr;
+			mr->mr.map[m]->segs[n].length = umem->page_size;
+			n++;
+			if (n == QIB_SEGSZ) {
+				m++;
+				n = 0;
+			}
+		}
+	}
+	ret = &mr->ibmr;
+
+bail:
+	return ret;
+}
+
+/**
+ * qib_dereg_mr - unregister and free a memory region
+ * @ibmr: the memory region to free
+ *
+ * Returns 0 on success.
+ *
+ * Note that this is called to free MRs created by qib_get_dma_mr()
+ * or qib_reg_user_mr().
+ */
+int qib_dereg_mr(struct ib_mr *ibmr)
+{
+	struct qib_mr *mr = to_imr(ibmr);
+	struct qib_ibdev *dev = to_idev(ibmr->device);
+	int ret;
+	int i;
+
+	ret = qib_free_lkey(dev, &mr->mr);
+	if (ret)
+		return ret;
+
+	i = mr->mr.mapsz;
+	while (i)
+		kfree(mr->mr.map[--i]);
+	if (mr->umem)
+		ib_umem_release(mr->umem);
+	kfree(mr);
+	return 0;
+}
+
+/*
+ * Allocate a memory region usable with the
+ * IB_WR_FAST_REG_MR send work request.
+ *
+ * Return the memory region on success, otherwise return an errno.
+ */
+struct ib_mr *qib_alloc_fast_reg_mr(struct ib_pd *pd, int max_page_list_len)
+{
+	struct qib_mr *mr;
+
+	mr = alloc_mr(max_page_list_len, &to_idev(pd->device)->lk_table);
+	if (mr == NULL)
+		return ERR_PTR(-ENOMEM);
+
+	mr->mr.pd = pd;
+	mr->mr.user_base = 0;
+	mr->mr.iova = 0;
+	mr->mr.length = 0;
+	mr->mr.offset = 0;
+	mr->mr.access_flags = 0;
+	mr->umem = NULL;
+
+	return &mr->ibmr;
+}
+
+struct ib_fast_reg_page_list *
+qib_alloc_fast_reg_page_list(struct ib_device *ibdev, int page_list_len)
+{
+	unsigned size = page_list_len * sizeof(u64);
+	struct ib_fast_reg_page_list *pl;
+
+	if (size > PAGE_SIZE)
+		return ERR_PTR(-EINVAL);
+
+	pl = kmalloc(sizeof *pl, GFP_KERNEL);
+	if (!pl)
+		return ERR_PTR(-ENOMEM);
+
+	pl->page_list = kmalloc(size, GFP_KERNEL);
+	if (!pl->page_list)
+		goto err_free;
+
+	return pl;
+
+err_free:
+	kfree(pl);
+	return ERR_PTR(-ENOMEM);
+}
+
+void qib_free_fast_reg_page_list(struct ib_fast_reg_page_list *pl)
+{
+	kfree(pl->page_list);
+	kfree(pl);
+}
+
+/**
+ * qib_alloc_fmr - allocate a fast memory region
+ * @pd: the protection domain for this memory region
+ * @mr_access_flags: access flags for this memory region
+ * @fmr_attr: fast memory region attributes
+ *
+ * Returns the memory region on success, otherwise returns an errno.
+ */
+struct ib_fmr *qib_alloc_fmr(struct ib_pd *pd, int mr_access_flags,
+			     struct ib_fmr_attr *fmr_attr)
+{
+	struct qib_fmr *fmr;
+	int m, i = 0;
+	struct ib_fmr *ret;
+
+	/* Allocate struct plus pointers to first level page tables. */
+	m = (fmr_attr->max_pages + QIB_SEGSZ - 1) / QIB_SEGSZ;
+	fmr = kmalloc(sizeof *fmr + m * sizeof fmr->mr.map[0], GFP_KERNEL);
+	if (!fmr)
+		goto bail;
+
+	/* Allocate first level page tables. */
+	for (; i < m; i++) {
+		fmr->mr.map[i] = kmalloc(sizeof *fmr->mr.map[0],
+					 GFP_KERNEL);
+		if (!fmr->mr.map[i])
+			goto bail;
+	}
+	fmr->mr.mapsz = m;
+
+	/*
+	 * ib_alloc_fmr() will initialize fmr->ibfmr except for lkey &
+	 * rkey.
+	 */
+	if (!qib_alloc_lkey(&to_idev(pd->device)->lk_table, &fmr->mr))
+		goto bail;
+	fmr->ibfmr.rkey = fmr->mr.lkey;
+	fmr->ibfmr.lkey = fmr->mr.lkey;
+	/*
+	 * Resources are allocated but no valid mapping (RKEY can't be
+	 * used).
+	 */
+	fmr->mr.pd = pd;
+	fmr->mr.user_base = 0;
+	fmr->mr.iova = 0;
+	fmr->mr.length = 0;
+	fmr->mr.offset = 0;
+	fmr->mr.access_flags = mr_access_flags;
+	fmr->mr.max_segs = fmr_attr->max_pages;
+	fmr->page_shift = fmr_attr->page_shift;
+
+	atomic_set(&fmr->mr.refcount, 0);
+	ret = &fmr->ibfmr;
+	goto done;
+
+bail:
+	while (i)
+		kfree(fmr->mr.map[--i]);
+	kfree(fmr);
+	ret = ERR_PTR(-ENOMEM);
+
+done:
+	return ret;
+}
+
+/**
+ * qib_map_phys_fmr - set up a fast memory region
+ * @ibmfr: the fast memory region to set up
+ * @page_list: the list of pages to associate with the fast memory region
+ * @list_len: the number of pages to associate with the fast memory region
+ * @iova: the virtual address of the start of the fast memory region
+ *
+ * This may be called from interrupt context.
+ */
+
+int qib_map_phys_fmr(struct ib_fmr *ibfmr, u64 *page_list,
+		     int list_len, u64 iova)
+{
+	struct qib_fmr *fmr = to_ifmr(ibfmr);
+	struct qib_lkey_table *rkt;
+	unsigned long flags;
+	int m, n, i;
+	u32 ps;
+	int ret;
+
+	if (atomic_read(&fmr->mr.refcount))
+		return -EBUSY;
+
+	if (list_len > fmr->mr.max_segs) {
+		ret = -EINVAL;
+		goto bail;
+	}
+	rkt = &to_idev(ibfmr->device)->lk_table;
+	spin_lock_irqsave(&rkt->lock, flags);
+	fmr->mr.user_base = iova;
+	fmr->mr.iova = iova;
+	ps = 1 << fmr->page_shift;
+	fmr->mr.length = list_len * ps;
+	m = 0;
+	n = 0;
+	for (i = 0; i < list_len; i++) {
+		fmr->mr.map[m]->segs[n].vaddr = (void *) page_list[i];
+		fmr->mr.map[m]->segs[n].length = ps;
+		if (++n == QIB_SEGSZ) {
+			m++;
+			n = 0;
+		}
+	}
+	spin_unlock_irqrestore(&rkt->lock, flags);
+	ret = 0;
+
+bail:
+	return ret;
+}
+
+/**
+ * qib_unmap_fmr - unmap fast memory regions
+ * @fmr_list: the list of fast memory regions to unmap
+ *
+ * Returns 0 on success.
+ */
+int qib_unmap_fmr(struct list_head *fmr_list)
+{
+	struct qib_fmr *fmr;
+	struct qib_lkey_table *rkt;
+	unsigned long flags;
+
+	list_for_each_entry(fmr, fmr_list, ibfmr.list) {
+		rkt = &to_idev(fmr->ibfmr.device)->lk_table;
+		spin_lock_irqsave(&rkt->lock, flags);
+		fmr->mr.user_base = 0;
+		fmr->mr.iova = 0;
+		fmr->mr.length = 0;
+		spin_unlock_irqrestore(&rkt->lock, flags);
+	}
+	return 0;
+}
+
+/**
+ * qib_dealloc_fmr - deallocate a fast memory region
+ * @ibfmr: the fast memory region to deallocate
+ *
+ * Returns 0 on success.
+ */
+int qib_dealloc_fmr(struct ib_fmr *ibfmr)
+{
+	struct qib_fmr *fmr = to_ifmr(ibfmr);
+	int ret;
+	int i;
+
+	ret = qib_free_lkey(to_idev(ibfmr->device), &fmr->mr);
+	if (ret)
+		return ret;
+
+	i = fmr->mr.mapsz;
+	while (i)
+		kfree(fmr->mr.map[--i]);
+	kfree(fmr);
+	return 0;
+}
diff --git a/drivers/infiniband/hw/qib/qib_pcie.c b/drivers/infiniband/hw/qib/qib_pcie.c
new file mode 100644
index 000000000000..c926bf4541df
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_pcie.c
@@ -0,0 +1,738 @@
+/*
+ * Copyright (c) 2008, 2009 QLogic Corporation. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/pci.h>
+#include <linux/io.h>
+#include <linux/delay.h>
+#include <linux/vmalloc.h>
+#include <linux/aer.h>
+
+#include "qib.h"
+
+/*
+ * This file contains PCIe utility routines that are common to the
+ * various QLogic InfiniPath adapters
+ */
+
+/*
+ * Code to adjust PCIe capabilities.
+ * To minimize the change footprint, we call it
+ * from qib_pcie_params, which every chip-specific
+ * file calls, even though this violates some
+ * expectations of harmlessness.
+ */
+static int qib_tune_pcie_caps(struct qib_devdata *);
+static int qib_tune_pcie_coalesce(struct qib_devdata *);
+
+/*
+ * Do all the common PCIe setup and initialization.
+ * devdata is not yet allocated, and is not allocated until after this
+ * routine returns success.  Therefore qib_dev_err() can't be used for error
+ * printing.
+ */
+int qib_pcie_init(struct pci_dev *pdev, const struct pci_device_id *ent)
+{
+	int ret;
+
+	ret = pci_enable_device(pdev);
+	if (ret) {
+		/*
+		 * This can happen (in theory) iff:
+		 * We did a chip reset, and then failed to reprogram the
+		 * BAR, or the chip reset due to an internal error.  We then
+		 * unloaded the driver and reloaded it.
+		 *
+		 * Both reset cases set the BAR back to initial state.  For
+		 * the latter case, the AER sticky error bit at offset 0x718
+		 * should be set, but the Linux kernel doesn't yet know
+		 * about that, it appears.  If the original BAR was retained
+		 * in the kernel data structures, this may be OK.
+		 */
+		qib_early_err(&pdev->dev, "pci enable failed: error %d\n",
+			      -ret);
+		goto done;
+	}
+
+	ret = pci_request_regions(pdev, QIB_DRV_NAME);
+	if (ret) {
+		qib_devinfo(pdev, "pci_request_regions fails: err %d\n", -ret);
+		goto bail;
+	}
+
+	ret = pci_set_dma_mask(pdev, DMA_BIT_MASK(64));
+	if (ret) {
+		/*
+		 * If the 64 bit setup fails, try 32 bit.  Some systems
+		 * do not setup 64 bit maps on systems with 2GB or less
+		 * memory installed.
+		 */
+		ret = pci_set_dma_mask(pdev, DMA_BIT_MASK(32));
+		if (ret) {
+			qib_devinfo(pdev, "Unable to set DMA mask: %d\n", ret);
+			goto bail;
+		}
+		ret = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32));
+	} else
+		ret = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(64));
+	if (ret)
+		qib_early_err(&pdev->dev,
+			      "Unable to set DMA consistent mask: %d\n", ret);
+
+	pci_set_master(pdev);
+	ret = pci_enable_pcie_error_reporting(pdev);
+	if (ret)
+		qib_early_err(&pdev->dev,
+			      "Unable to enable pcie error reporting: %d\n",
+			      ret);
+	goto done;
+
+bail:
+	pci_disable_device(pdev);
+	pci_release_regions(pdev);
+done:
+	return ret;
+}
+
+/*
+ * Do remaining PCIe setup, once dd is allocated, and save away
+ * fields required to re-initialize after a chip reset, or for
+ * various other purposes
+ */
+int qib_pcie_ddinit(struct qib_devdata *dd, struct pci_dev *pdev,
+		    const struct pci_device_id *ent)
+{
+	unsigned long len;
+	resource_size_t addr;
+
+	dd->pcidev = pdev;
+	pci_set_drvdata(pdev, dd);
+
+	addr = pci_resource_start(pdev, 0);
+	len = pci_resource_len(pdev, 0);
+
+#if defined(__powerpc__)
+	/* There isn't a generic way to specify writethrough mappings */
+	dd->kregbase = __ioremap(addr, len, _PAGE_NO_CACHE | _PAGE_WRITETHRU);
+#else
+	dd->kregbase = ioremap_nocache(addr, len);
+#endif
+
+	if (!dd->kregbase)
+		return -ENOMEM;
+
+	dd->kregend = (u64 __iomem *)((void __iomem *) dd->kregbase + len);
+	dd->physaddr = addr;        /* used for io_remap, etc. */
+
+	/*
+	 * Save BARs to rewrite after device reset.  Save all 64 bits of
+	 * BAR, just in case.
+	 */
+	dd->pcibar0 = addr;
+	dd->pcibar1 = addr >> 32;
+	dd->deviceid = ent->device; /* save for later use */
+	dd->vendorid = ent->vendor;
+
+	return 0;
+}
+
+/*
+ * Do PCIe cleanup, after chip-specific cleanup, etc.  Just prior
+ * to releasing the dd memory.
+ * void because none of the core pcie cleanup returns are void
+ */
+void qib_pcie_ddcleanup(struct qib_devdata *dd)
+{
+	u64 __iomem *base = (void __iomem *) dd->kregbase;
+
+	dd->kregbase = NULL;
+	iounmap(base);
+	if (dd->piobase)
+		iounmap(dd->piobase);
+	if (dd->userbase)
+		iounmap(dd->userbase);
+
+	pci_disable_device(dd->pcidev);
+	pci_release_regions(dd->pcidev);
+
+	pci_set_drvdata(dd->pcidev, NULL);
+}
+
+static void qib_msix_setup(struct qib_devdata *dd, int pos, u32 *msixcnt,
+			   struct msix_entry *msix_entry)
+{
+	int ret;
+	u32 tabsize = 0;
+	u16 msix_flags;
+
+	pci_read_config_word(dd->pcidev, pos + PCI_MSIX_FLAGS, &msix_flags);
+	tabsize = 1 + (msix_flags & PCI_MSIX_FLAGS_QSIZE);
+	if (tabsize > *msixcnt)
+		tabsize = *msixcnt;
+	ret = pci_enable_msix(dd->pcidev, msix_entry, tabsize);
+	if (ret > 0) {
+		tabsize = ret;
+		ret = pci_enable_msix(dd->pcidev, msix_entry, tabsize);
+	}
+	if (ret) {
+		qib_dev_err(dd, "pci_enable_msix %d vectors failed: %d, "
+			    "falling back to INTx\n", tabsize, ret);
+		tabsize = 0;
+	}
+	*msixcnt = tabsize;
+
+	if (ret)
+		qib_enable_intx(dd->pcidev);
+
+}
+
+/**
+ * We save the msi lo and hi values, so we can restore them after
+ * chip reset (the kernel PCI infrastructure doesn't yet handle that
+ * correctly.
+ */
+static int qib_msi_setup(struct qib_devdata *dd, int pos)
+{
+	struct pci_dev *pdev = dd->pcidev;
+	u16 control;
+	int ret;
+
+	ret = pci_enable_msi(pdev);
+	if (ret)
+		qib_dev_err(dd, "pci_enable_msi failed: %d, "
+			    "interrupts may not work\n", ret);
+	/* continue even if it fails, we may still be OK... */
+
+	pci_read_config_dword(pdev, pos + PCI_MSI_ADDRESS_LO,
+			      &dd->msi_lo);
+	pci_read_config_dword(pdev, pos + PCI_MSI_ADDRESS_HI,
+			      &dd->msi_hi);
+	pci_read_config_word(pdev, pos + PCI_MSI_FLAGS, &control);
+	/* now save the data (vector) info */
+	pci_read_config_word(pdev, pos + ((control & PCI_MSI_FLAGS_64BIT)
+				    ? 12 : 8),
+			     &dd->msi_data);
+	return ret;
+}
+
+int qib_pcie_params(struct qib_devdata *dd, u32 minw, u32 *nent,
+		    struct msix_entry *entry)
+{
+	u16 linkstat, speed;
+	int pos = 0, pose, ret = 1;
+
+	pose = pci_find_capability(dd->pcidev, PCI_CAP_ID_EXP);
+	if (!pose) {
+		qib_dev_err(dd, "Can't find PCI Express capability!\n");
+		/* set up something... */
+		dd->lbus_width = 1;
+		dd->lbus_speed = 2500; /* Gen1, 2.5GHz */
+		goto bail;
+	}
+
+	pos = pci_find_capability(dd->pcidev, PCI_CAP_ID_MSIX);
+	if (nent && *nent && pos) {
+		qib_msix_setup(dd, pos, nent, entry);
+		ret = 0; /* did it, either MSIx or INTx */
+	} else {
+		pos = pci_find_capability(dd->pcidev, PCI_CAP_ID_MSI);
+		if (pos)
+			ret = qib_msi_setup(dd, pos);
+		else
+			qib_dev_err(dd, "No PCI MSI or MSIx capability!\n");
+	}
+	if (!pos)
+		qib_enable_intx(dd->pcidev);
+
+	pci_read_config_word(dd->pcidev, pose + PCI_EXP_LNKSTA, &linkstat);
+	/*
+	 * speed is bits 0-3, linkwidth is bits 4-8
+	 * no defines for them in headers
+	 */
+	speed = linkstat & 0xf;
+	linkstat >>= 4;
+	linkstat &= 0x1f;
+	dd->lbus_width = linkstat;
+
+	switch (speed) {
+	case 1:
+		dd->lbus_speed = 2500; /* Gen1, 2.5GHz */
+		break;
+	case 2:
+		dd->lbus_speed = 5000; /* Gen1, 5GHz */
+		break;
+	default: /* not defined, assume gen1 */
+		dd->lbus_speed = 2500;
+		break;
+	}
+
+	/*
+	 * Check against expected pcie width and complain if "wrong"
+	 * on first initialization, not afterwards (i.e., reset).
+	 */
+	if (minw && linkstat < minw)
+		qib_dev_err(dd,
+			    "PCIe width %u (x%u HCA), performance reduced\n",
+			    linkstat, minw);
+
+	qib_tune_pcie_caps(dd);
+
+	qib_tune_pcie_coalesce(dd);
+
+bail:
+	/* fill in string, even on errors */
+	snprintf(dd->lbus_info, sizeof(dd->lbus_info),
+		 "PCIe,%uMHz,x%u\n", dd->lbus_speed, dd->lbus_width);
+	return ret;
+}
+
+/*
+ * Setup pcie interrupt stuff again after a reset.  I'd like to just call
+ * pci_enable_msi() again for msi, but when I do that,
+ * the MSI enable bit doesn't get set in the command word, and
+ * we switch to to a different interrupt vector, which is confusing,
+ * so I instead just do it all inline.  Perhaps somehow can tie this
+ * into the PCIe hotplug support at some point
+ */
+int qib_reinit_intr(struct qib_devdata *dd)
+{
+	int pos;
+	u16 control;
+	int ret = 0;
+
+	/* If we aren't using MSI, don't restore it */
+	if (!dd->msi_lo)
+		goto bail;
+
+	pos = pci_find_capability(dd->pcidev, PCI_CAP_ID_MSI);
+	if (!pos) {
+		qib_dev_err(dd, "Can't find MSI capability, "
+			    "can't restore MSI settings\n");
+		ret = 0;
+		/* nothing special for MSIx, just MSI */
+		goto bail;
+	}
+	pci_write_config_dword(dd->pcidev, pos + PCI_MSI_ADDRESS_LO,
+			       dd->msi_lo);
+	pci_write_config_dword(dd->pcidev, pos + PCI_MSI_ADDRESS_HI,
+			       dd->msi_hi);
+	pci_read_config_word(dd->pcidev, pos + PCI_MSI_FLAGS, &control);
+	if (!(control & PCI_MSI_FLAGS_ENABLE)) {
+		control |= PCI_MSI_FLAGS_ENABLE;
+		pci_write_config_word(dd->pcidev, pos + PCI_MSI_FLAGS,
+				      control);
+	}
+	/* now rewrite the data (vector) info */
+	pci_write_config_word(dd->pcidev, pos +
+			      ((control & PCI_MSI_FLAGS_64BIT) ? 12 : 8),
+			      dd->msi_data);
+	ret = 1;
+bail:
+	if (!ret && (dd->flags & QIB_HAS_INTX)) {
+		qib_enable_intx(dd->pcidev);
+		ret = 1;
+	}
+
+	/* and now set the pci master bit again */
+	pci_set_master(dd->pcidev);
+
+	return ret;
+}
+
+/*
+ * Disable msi interrupt if enabled, and clear msi_lo.
+ * This is used primarily for the fallback to INTx, but
+ * is also used in reinit after reset, and during cleanup.
+ */
+void qib_nomsi(struct qib_devdata *dd)
+{
+	dd->msi_lo = 0;
+	pci_disable_msi(dd->pcidev);
+}
+
+/*
+ * Same as qib_nosmi, but for MSIx.
+ */
+void qib_nomsix(struct qib_devdata *dd)
+{
+	pci_disable_msix(dd->pcidev);
+}
+
+/*
+ * Similar to pci_intx(pdev, 1), except that we make sure
+ * msi(x) is off.
+ */
+void qib_enable_intx(struct pci_dev *pdev)
+{
+	u16 cw, new;
+	int pos;
+
+	/* first, turn on INTx */
+	pci_read_config_word(pdev, PCI_COMMAND, &cw);
+	new = cw & ~PCI_COMMAND_INTX_DISABLE;
+	if (new != cw)
+		pci_write_config_word(pdev, PCI_COMMAND, new);
+
+	pos = pci_find_capability(pdev, PCI_CAP_ID_MSI);
+	if (pos) {
+		/* then turn off MSI */
+		pci_read_config_word(pdev, pos + PCI_MSI_FLAGS, &cw);
+		new = cw & ~PCI_MSI_FLAGS_ENABLE;
+		if (new != cw)
+			pci_write_config_word(pdev, pos + PCI_MSI_FLAGS, new);
+	}
+	pos = pci_find_capability(pdev, PCI_CAP_ID_MSIX);
+	if (pos) {
+		/* then turn off MSIx */
+		pci_read_config_word(pdev, pos + PCI_MSIX_FLAGS, &cw);
+		new = cw & ~PCI_MSIX_FLAGS_ENABLE;
+		if (new != cw)
+			pci_write_config_word(pdev, pos + PCI_MSIX_FLAGS, new);
+	}
+}
+
+/*
+ * These two routines are helper routines for the device reset code
+ * to move all the pcie code out of the chip-specific driver code.
+ */
+void qib_pcie_getcmd(struct qib_devdata *dd, u16 *cmd, u8 *iline, u8 *cline)
+{
+	pci_read_config_word(dd->pcidev, PCI_COMMAND, cmd);
+	pci_read_config_byte(dd->pcidev, PCI_INTERRUPT_LINE, iline);
+	pci_read_config_byte(dd->pcidev, PCI_CACHE_LINE_SIZE, cline);
+}
+
+void qib_pcie_reenable(struct qib_devdata *dd, u16 cmd, u8 iline, u8 cline)
+{
+	int r;
+	r = pci_write_config_dword(dd->pcidev, PCI_BASE_ADDRESS_0,
+				   dd->pcibar0);
+	if (r)
+		qib_dev_err(dd, "rewrite of BAR0 failed: %d\n", r);
+	r = pci_write_config_dword(dd->pcidev, PCI_BASE_ADDRESS_1,
+				   dd->pcibar1);
+	if (r)
+		qib_dev_err(dd, "rewrite of BAR1 failed: %d\n", r);
+	/* now re-enable memory access, and restore cosmetic settings */
+	pci_write_config_word(dd->pcidev, PCI_COMMAND, cmd);
+	pci_write_config_byte(dd->pcidev, PCI_INTERRUPT_LINE, iline);
+	pci_write_config_byte(dd->pcidev, PCI_CACHE_LINE_SIZE, cline);
+	r = pci_enable_device(dd->pcidev);
+	if (r)
+		qib_dev_err(dd, "pci_enable_device failed after "
+			    "reset: %d\n", r);
+}
+
+/* code to adjust PCIe capabilities. */
+
+static int fld2val(int wd, int mask)
+{
+	int lsbmask;
+
+	if (!mask)
+		return 0;
+	wd &= mask;
+	lsbmask = mask ^ (mask & (mask - 1));
+	wd /= lsbmask;
+	return wd;
+}
+
+static int val2fld(int wd, int mask)
+{
+	int lsbmask;
+
+	if (!mask)
+		return 0;
+	lsbmask = mask ^ (mask & (mask - 1));
+	wd *= lsbmask;
+	return wd;
+}
+
+static int qib_pcie_coalesce;
+module_param_named(pcie_coalesce, qib_pcie_coalesce, int, S_IRUGO);
+MODULE_PARM_DESC(pcie_coalesce, "tune PCIe colescing on some Intel chipsets");
+
+/*
+ * Enable PCIe completion and data coalescing, on Intel 5x00 and 7300
+ * chipsets.   This is known to be unsafe for some revisions of some
+ * of these chipsets, with some BIOS settings, and enabling it on those
+ * systems may result in the system crashing, and/or data corruption.
+ */
+static int qib_tune_pcie_coalesce(struct qib_devdata *dd)
+{
+	int r;
+	struct pci_dev *parent;
+	int ppos;
+	u16 devid;
+	u32 mask, bits, val;
+
+	if (!qib_pcie_coalesce)
+		return 0;
+
+	/* Find out supported and configured values for parent (root) */
+	parent = dd->pcidev->bus->self;
+	if (parent->bus->parent) {
+		qib_devinfo(dd->pcidev, "Parent not root\n");
+		return 1;
+	}
+	ppos = pci_find_capability(parent, PCI_CAP_ID_EXP);
+	if (!ppos)
+		return 1;
+	if (parent->vendor != 0x8086)
+		return 1;
+
+	/*
+	 *  - bit 12: Max_rdcmp_Imt_EN: need to set to 1
+	 *  - bit 11: COALESCE_FORCE: need to set to 0
+	 *  - bit 10: COALESCE_EN: need to set to 1
+	 *  (but limitations on some on some chipsets)
+	 *
+	 *  On the Intel 5000, 5100, and 7300 chipsets, there is
+	 *  also: - bit 25:24: COALESCE_MODE, need to set to 0
+	 */
+	devid = parent->device;
+	if (devid >= 0x25e2 && devid <= 0x25fa) {
+		u8 rev;
+
+		/* 5000 P/V/X/Z */
+		pci_read_config_byte(parent, PCI_REVISION_ID, &rev);
+		if (rev <= 0xb2)
+			bits = 1U << 10;
+		else
+			bits = 7U << 10;
+		mask = (3U << 24) | (7U << 10);
+	} else if (devid >= 0x65e2 && devid <= 0x65fa) {
+		/* 5100 */
+		bits = 1U << 10;
+		mask = (3U << 24) | (7U << 10);
+	} else if (devid >= 0x4021 && devid <= 0x402e) {
+		/* 5400 */
+		bits = 7U << 10;
+		mask = 7U << 10;
+	} else if (devid >= 0x3604 && devid <= 0x360a) {
+		/* 7300 */
+		bits = 7U << 10;
+		mask = (3U << 24) | (7U << 10);
+	} else {
+		/* not one of the chipsets that we know about */
+		return 1;
+	}
+	pci_read_config_dword(parent, 0x48, &val);
+	val &= ~mask;
+	val |= bits;
+	r = pci_write_config_dword(parent, 0x48, val);
+	return 0;
+}
+
+/*
+ * BIOS may not set PCIe bus-utilization parameters for best performance.
+ * Check and optionally adjust them to maximize our throughput.
+ */
+static int qib_pcie_caps;
+module_param_named(pcie_caps, qib_pcie_caps, int, S_IRUGO);
+MODULE_PARM_DESC(pcie_caps, "Max PCIe tuning: Payload (4lsb), ReadReq (D4..7)");
+
+static int qib_tune_pcie_caps(struct qib_devdata *dd)
+{
+	int ret = 1; /* Assume the worst */
+	struct pci_dev *parent;
+	int ppos, epos;
+	u16 pcaps, pctl, ecaps, ectl;
+	int rc_sup, ep_sup;
+	int rc_cur, ep_cur;
+
+	/* Find out supported and configured values for parent (root) */
+	parent = dd->pcidev->bus->self;
+	if (parent->bus->parent) {
+		qib_devinfo(dd->pcidev, "Parent not root\n");
+		goto bail;
+	}
+	ppos = pci_find_capability(parent, PCI_CAP_ID_EXP);
+	if (ppos) {
+		pci_read_config_word(parent, ppos + PCI_EXP_DEVCAP, &pcaps);
+		pci_read_config_word(parent, ppos + PCI_EXP_DEVCTL, &pctl);
+	} else
+		goto bail;
+	/* Find out supported and configured values for endpoint (us) */
+	epos = pci_find_capability(dd->pcidev, PCI_CAP_ID_EXP);
+	if (epos) {
+		pci_read_config_word(dd->pcidev, epos + PCI_EXP_DEVCAP, &ecaps);
+		pci_read_config_word(dd->pcidev, epos + PCI_EXP_DEVCTL, &ectl);
+	} else
+		goto bail;
+	ret = 0;
+	/* Find max payload supported by root, endpoint */
+	rc_sup = fld2val(pcaps, PCI_EXP_DEVCAP_PAYLOAD);
+	ep_sup = fld2val(ecaps, PCI_EXP_DEVCAP_PAYLOAD);
+	if (rc_sup > ep_sup)
+		rc_sup = ep_sup;
+
+	rc_cur = fld2val(pctl, PCI_EXP_DEVCTL_PAYLOAD);
+	ep_cur = fld2val(ectl, PCI_EXP_DEVCTL_PAYLOAD);
+
+	/* If Supported greater than limit in module param, limit it */
+	if (rc_sup > (qib_pcie_caps & 7))
+		rc_sup = qib_pcie_caps & 7;
+	/* If less than (allowed, supported), bump root payload */
+	if (rc_sup > rc_cur) {
+		rc_cur = rc_sup;
+		pctl = (pctl & ~PCI_EXP_DEVCTL_PAYLOAD) |
+			val2fld(rc_cur, PCI_EXP_DEVCTL_PAYLOAD);
+		pci_write_config_word(parent, ppos + PCI_EXP_DEVCTL, pctl);
+	}
+	/* If less than (allowed, supported), bump endpoint payload */
+	if (rc_sup > ep_cur) {
+		ep_cur = rc_sup;
+		ectl = (ectl & ~PCI_EXP_DEVCTL_PAYLOAD) |
+			val2fld(ep_cur, PCI_EXP_DEVCTL_PAYLOAD);
+		pci_write_config_word(dd->pcidev, epos + PCI_EXP_DEVCTL, ectl);
+	}
+
+	/*
+	 * Now the Read Request size.
+	 * No field for max supported, but PCIe spec limits it to 4096,
+	 * which is code '5' (log2(4096) - 7)
+	 */
+	rc_sup = 5;
+	if (rc_sup > ((qib_pcie_caps >> 4) & 7))
+		rc_sup = (qib_pcie_caps >> 4) & 7;
+	rc_cur = fld2val(pctl, PCI_EXP_DEVCTL_READRQ);
+	ep_cur = fld2val(ectl, PCI_EXP_DEVCTL_READRQ);
+
+	if (rc_sup > rc_cur) {
+		rc_cur = rc_sup;
+		pctl = (pctl & ~PCI_EXP_DEVCTL_READRQ) |
+			val2fld(rc_cur, PCI_EXP_DEVCTL_READRQ);
+		pci_write_config_word(parent, ppos + PCI_EXP_DEVCTL, pctl);
+	}
+	if (rc_sup > ep_cur) {
+		ep_cur = rc_sup;
+		ectl = (ectl & ~PCI_EXP_DEVCTL_READRQ) |
+			val2fld(ep_cur, PCI_EXP_DEVCTL_READRQ);
+		pci_write_config_word(dd->pcidev, epos + PCI_EXP_DEVCTL, ectl);
+	}
+bail:
+	return ret;
+}
+/* End of PCIe capability tuning */
+
+/*
+ * From here through qib_pci_err_handler definition is invoked via
+ * PCI error infrastructure, registered via pci
+ */
+static pci_ers_result_t
+qib_pci_error_detected(struct pci_dev *pdev, pci_channel_state_t state)
+{
+	struct qib_devdata *dd = pci_get_drvdata(pdev);
+	pci_ers_result_t ret = PCI_ERS_RESULT_RECOVERED;
+
+	switch (state) {
+	case pci_channel_io_normal:
+		qib_devinfo(pdev, "State Normal, ignoring\n");
+		break;
+
+	case pci_channel_io_frozen:
+		qib_devinfo(pdev, "State Frozen, requesting reset\n");
+		pci_disable_device(pdev);
+		ret = PCI_ERS_RESULT_NEED_RESET;
+		break;
+
+	case pci_channel_io_perm_failure:
+		qib_devinfo(pdev, "State Permanent Failure, disabling\n");
+		if (dd) {
+			/* no more register accesses! */
+			dd->flags &= ~QIB_PRESENT;
+			qib_disable_after_error(dd);
+		}
+		 /* else early, or other problem */
+		ret =  PCI_ERS_RESULT_DISCONNECT;
+		break;
+
+	default: /* shouldn't happen */
+		qib_devinfo(pdev, "QIB PCI errors detected (state %d)\n",
+			state);
+		break;
+	}
+	return ret;
+}
+
+static pci_ers_result_t
+qib_pci_mmio_enabled(struct pci_dev *pdev)
+{
+	u64 words = 0U;
+	struct qib_devdata *dd = pci_get_drvdata(pdev);
+	pci_ers_result_t ret = PCI_ERS_RESULT_RECOVERED;
+
+	if (dd && dd->pport) {
+		words = dd->f_portcntr(dd->pport, QIBPORTCNTR_WORDRCV);
+		if (words == ~0ULL)
+			ret = PCI_ERS_RESULT_NEED_RESET;
+	}
+	qib_devinfo(pdev, "QIB mmio_enabled function called, "
+		 "read wordscntr %Lx, returning %d\n", words, ret);
+	return  ret;
+}
+
+static pci_ers_result_t
+qib_pci_slot_reset(struct pci_dev *pdev)
+{
+	qib_devinfo(pdev, "QIB link_reset function called, ignored\n");
+	return PCI_ERS_RESULT_CAN_RECOVER;
+}
+
+static pci_ers_result_t
+qib_pci_link_reset(struct pci_dev *pdev)
+{
+	qib_devinfo(pdev, "QIB link_reset function called, ignored\n");
+	return PCI_ERS_RESULT_CAN_RECOVER;
+}
+
+static void
+qib_pci_resume(struct pci_dev *pdev)
+{
+	struct qib_devdata *dd = pci_get_drvdata(pdev);
+	qib_devinfo(pdev, "QIB resume function called\n");
+	pci_cleanup_aer_uncorrect_error_status(pdev);
+	/*
+	 * Running jobs will fail, since it's asynchronous
+	 * unlike sysfs-requested reset.   Better than
+	 * doing nothing.
+	 */
+	qib_init(dd, 1); /* same as re-init after reset */
+}
+
+struct pci_error_handlers qib_pci_err_handler = {
+	.error_detected = qib_pci_error_detected,
+	.mmio_enabled = qib_pci_mmio_enabled,
+	.link_reset = qib_pci_link_reset,
+	.slot_reset = qib_pci_slot_reset,
+	.resume = qib_pci_resume,
+};
diff --git a/drivers/infiniband/hw/qib/qib_pio_copy.c b/drivers/infiniband/hw/qib/qib_pio_copy.c
new file mode 100644
index 000000000000..10b8c444dd31
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_pio_copy.c
@@ -0,0 +1,64 @@
+/*
+ * Copyright (c) 2009 QLogic Corporation. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 "qib.h"
+
+/**
+ * qib_pio_copy - copy data to MMIO space, in multiples of 32-bits
+ * @to: destination, in MMIO space (must be 64-bit aligned)
+ * @from: source (must be 64-bit aligned)
+ * @count: number of 32-bit quantities to copy
+ *
+ * Copy data from kernel space to MMIO space, in multiples of 32 bits at a
+ * time.  Order of access is not guaranteed, nor is a memory barrier
+ * performed afterwards.
+ */
+void qib_pio_copy(void __iomem *to, const void *from, size_t count)
+{
+#ifdef CONFIG_64BIT
+	u64 __iomem *dst = to;
+	const u64 *src = from;
+	const u64 *end = src + (count >> 1);
+
+	while (src < end)
+		__raw_writeq(*src++, dst++);
+	if (count & 1)
+		__raw_writel(*(const u32 *)src, dst);
+#else
+	u32 __iomem *dst = to;
+	const u32 *src = from;
+	const u32 *end = src + count;
+
+	while (src < end)
+		__raw_writel(*src++, dst++);
+#endif
+}
diff --git a/drivers/infiniband/hw/qib/qib_qp.c b/drivers/infiniband/hw/qib/qib_qp.c
new file mode 100644
index 000000000000..e0f65e39076b
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_qp.c
@@ -0,0 +1,1255 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009, 2010 QLogic Corporation.
+ * All rights reserved.
+ * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/err.h>
+#include <linux/vmalloc.h>
+
+#include "qib.h"
+
+#define BITS_PER_PAGE           (PAGE_SIZE*BITS_PER_BYTE)
+#define BITS_PER_PAGE_MASK      (BITS_PER_PAGE-1)
+
+static inline unsigned mk_qpn(struct qib_qpn_table *qpt,
+			      struct qpn_map *map, unsigned off)
+{
+	return (map - qpt->map) * BITS_PER_PAGE + off;
+}
+
+static inline unsigned find_next_offset(struct qib_qpn_table *qpt,
+					struct qpn_map *map, unsigned off,
+					unsigned r)
+{
+	if (qpt->mask) {
+		off++;
+		if ((off & qpt->mask) >> 1 != r)
+			off = ((off & qpt->mask) ?
+				(off | qpt->mask) + 1 : off) | (r << 1);
+	} else
+		off = find_next_zero_bit(map->page, BITS_PER_PAGE, off);
+	return off;
+}
+
+/*
+ * Convert the AETH credit code into the number of credits.
+ */
+static u32 credit_table[31] = {
+	0,                      /* 0 */
+	1,                      /* 1 */
+	2,                      /* 2 */
+	3,                      /* 3 */
+	4,                      /* 4 */
+	6,                      /* 5 */
+	8,                      /* 6 */
+	12,                     /* 7 */
+	16,                     /* 8 */
+	24,                     /* 9 */
+	32,                     /* A */
+	48,                     /* B */
+	64,                     /* C */
+	96,                     /* D */
+	128,                    /* E */
+	192,                    /* F */
+	256,                    /* 10 */
+	384,                    /* 11 */
+	512,                    /* 12 */
+	768,                    /* 13 */
+	1024,                   /* 14 */
+	1536,                   /* 15 */
+	2048,                   /* 16 */
+	3072,                   /* 17 */
+	4096,                   /* 18 */
+	6144,                   /* 19 */
+	8192,                   /* 1A */
+	12288,                  /* 1B */
+	16384,                  /* 1C */
+	24576,                  /* 1D */
+	32768                   /* 1E */
+};
+
+static void get_map_page(struct qib_qpn_table *qpt, struct qpn_map *map)
+{
+	unsigned long page = get_zeroed_page(GFP_KERNEL);
+
+	/*
+	 * Free the page if someone raced with us installing it.
+	 */
+
+	spin_lock(&qpt->lock);
+	if (map->page)
+		free_page(page);
+	else
+		map->page = (void *)page;
+	spin_unlock(&qpt->lock);
+}
+
+/*
+ * Allocate the next available QPN or
+ * zero/one for QP type IB_QPT_SMI/IB_QPT_GSI.
+ */
+static int alloc_qpn(struct qib_devdata *dd, struct qib_qpn_table *qpt,
+		     enum ib_qp_type type, u8 port)
+{
+	u32 i, offset, max_scan, qpn;
+	struct qpn_map *map;
+	u32 ret;
+	int r;
+
+	if (type == IB_QPT_SMI || type == IB_QPT_GSI) {
+		unsigned n;
+
+		ret = type == IB_QPT_GSI;
+		n = 1 << (ret + 2 * (port - 1));
+		spin_lock(&qpt->lock);
+		if (qpt->flags & n)
+			ret = -EINVAL;
+		else
+			qpt->flags |= n;
+		spin_unlock(&qpt->lock);
+		goto bail;
+	}
+
+	r = smp_processor_id();
+	if (r >= dd->n_krcv_queues)
+		r %= dd->n_krcv_queues;
+	qpn = qpt->last + 1;
+	if (qpn >= QPN_MAX)
+		qpn = 2;
+	if (qpt->mask && ((qpn & qpt->mask) >> 1) != r)
+		qpn = ((qpn & qpt->mask) ? (qpn | qpt->mask) + 1 : qpn) |
+			(r << 1);
+	offset = qpn & BITS_PER_PAGE_MASK;
+	map = &qpt->map[qpn / BITS_PER_PAGE];
+	max_scan = qpt->nmaps - !offset;
+	for (i = 0;;) {
+		if (unlikely(!map->page)) {
+			get_map_page(qpt, map);
+			if (unlikely(!map->page))
+				break;
+		}
+		do {
+			if (!test_and_set_bit(offset, map->page)) {
+				qpt->last = qpn;
+				ret = qpn;
+				goto bail;
+			}
+			offset = find_next_offset(qpt, map, offset, r);
+			qpn = mk_qpn(qpt, map, offset);
+			/*
+			 * This test differs from alloc_pidmap().
+			 * If find_next_offset() does find a zero
+			 * bit, we don't need to check for QPN
+			 * wrapping around past our starting QPN.
+			 * We just need to be sure we don't loop
+			 * forever.
+			 */
+		} while (offset < BITS_PER_PAGE && qpn < QPN_MAX);
+		/*
+		 * In order to keep the number of pages allocated to a
+		 * minimum, we scan the all existing pages before increasing
+		 * the size of the bitmap table.
+		 */
+		if (++i > max_scan) {
+			if (qpt->nmaps == QPNMAP_ENTRIES)
+				break;
+			map = &qpt->map[qpt->nmaps++];
+			offset = qpt->mask ? (r << 1) : 0;
+		} else if (map < &qpt->map[qpt->nmaps]) {
+			++map;
+			offset = qpt->mask ? (r << 1) : 0;
+		} else {
+			map = &qpt->map[0];
+			offset = qpt->mask ? (r << 1) : 2;
+		}
+		qpn = mk_qpn(qpt, map, offset);
+	}
+
+	ret = -ENOMEM;
+
+bail:
+	return ret;
+}
+
+static void free_qpn(struct qib_qpn_table *qpt, u32 qpn)
+{
+	struct qpn_map *map;
+
+	map = qpt->map + qpn / BITS_PER_PAGE;
+	if (map->page)
+		clear_bit(qpn & BITS_PER_PAGE_MASK, map->page);
+}
+
+/*
+ * Put the QP into the hash table.
+ * The hash table holds a reference to the QP.
+ */
+static void insert_qp(struct qib_ibdev *dev, struct qib_qp *qp)
+{
+	struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
+	unsigned n = qp->ibqp.qp_num % dev->qp_table_size;
+	unsigned long flags;
+
+	spin_lock_irqsave(&dev->qpt_lock, flags);
+
+	if (qp->ibqp.qp_num == 0)
+		ibp->qp0 = qp;
+	else if (qp->ibqp.qp_num == 1)
+		ibp->qp1 = qp;
+	else {
+		qp->next = dev->qp_table[n];
+		dev->qp_table[n] = qp;
+	}
+	atomic_inc(&qp->refcount);
+
+	spin_unlock_irqrestore(&dev->qpt_lock, flags);
+}
+
+/*
+ * Remove the QP from the table so it can't be found asynchronously by
+ * the receive interrupt routine.
+ */
+static void remove_qp(struct qib_ibdev *dev, struct qib_qp *qp)
+{
+	struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
+	struct qib_qp *q, **qpp;
+	unsigned long flags;
+
+	qpp = &dev->qp_table[qp->ibqp.qp_num % dev->qp_table_size];
+
+	spin_lock_irqsave(&dev->qpt_lock, flags);
+
+	if (ibp->qp0 == qp) {
+		ibp->qp0 = NULL;
+		atomic_dec(&qp->refcount);
+	} else if (ibp->qp1 == qp) {
+		ibp->qp1 = NULL;
+		atomic_dec(&qp->refcount);
+	} else
+		for (; (q = *qpp) != NULL; qpp = &q->next)
+			if (q == qp) {
+				*qpp = qp->next;
+				qp->next = NULL;
+				atomic_dec(&qp->refcount);
+				break;
+			}
+
+	spin_unlock_irqrestore(&dev->qpt_lock, flags);
+}
+
+/**
+ * qib_free_all_qps - check for QPs still in use
+ * @qpt: the QP table to empty
+ *
+ * There should not be any QPs still in use.
+ * Free memory for table.
+ */
+unsigned qib_free_all_qps(struct qib_devdata *dd)
+{
+	struct qib_ibdev *dev = &dd->verbs_dev;
+	unsigned long flags;
+	struct qib_qp *qp;
+	unsigned n, qp_inuse = 0;
+
+	for (n = 0; n < dd->num_pports; n++) {
+		struct qib_ibport *ibp = &dd->pport[n].ibport_data;
+
+		if (!qib_mcast_tree_empty(ibp))
+			qp_inuse++;
+		if (ibp->qp0)
+			qp_inuse++;
+		if (ibp->qp1)
+			qp_inuse++;
+	}
+
+	spin_lock_irqsave(&dev->qpt_lock, flags);
+	for (n = 0; n < dev->qp_table_size; n++) {
+		qp = dev->qp_table[n];
+		dev->qp_table[n] = NULL;
+
+		for (; qp; qp = qp->next)
+			qp_inuse++;
+	}
+	spin_unlock_irqrestore(&dev->qpt_lock, flags);
+
+	return qp_inuse;
+}
+
+/**
+ * qib_lookup_qpn - return the QP with the given QPN
+ * @qpt: the QP table
+ * @qpn: the QP number to look up
+ *
+ * The caller is responsible for decrementing the QP reference count
+ * when done.
+ */
+struct qib_qp *qib_lookup_qpn(struct qib_ibport *ibp, u32 qpn)
+{
+	struct qib_ibdev *dev = &ppd_from_ibp(ibp)->dd->verbs_dev;
+	unsigned long flags;
+	struct qib_qp *qp;
+
+	spin_lock_irqsave(&dev->qpt_lock, flags);
+
+	if (qpn == 0)
+		qp = ibp->qp0;
+	else if (qpn == 1)
+		qp = ibp->qp1;
+	else
+		for (qp = dev->qp_table[qpn % dev->qp_table_size]; qp;
+		     qp = qp->next)
+			if (qp->ibqp.qp_num == qpn)
+				break;
+	if (qp)
+		atomic_inc(&qp->refcount);
+
+	spin_unlock_irqrestore(&dev->qpt_lock, flags);
+	return qp;
+}
+
+/**
+ * qib_reset_qp - initialize the QP state to the reset state
+ * @qp: the QP to reset
+ * @type: the QP type
+ */
+static void qib_reset_qp(struct qib_qp *qp, enum ib_qp_type type)
+{
+	qp->remote_qpn = 0;
+	qp->qkey = 0;
+	qp->qp_access_flags = 0;
+	atomic_set(&qp->s_dma_busy, 0);
+	qp->s_flags &= QIB_S_SIGNAL_REQ_WR;
+	qp->s_hdrwords = 0;
+	qp->s_wqe = NULL;
+	qp->s_draining = 0;
+	qp->s_next_psn = 0;
+	qp->s_last_psn = 0;
+	qp->s_sending_psn = 0;
+	qp->s_sending_hpsn = 0;
+	qp->s_psn = 0;
+	qp->r_psn = 0;
+	qp->r_msn = 0;
+	if (type == IB_QPT_RC) {
+		qp->s_state = IB_OPCODE_RC_SEND_LAST;
+		qp->r_state = IB_OPCODE_RC_SEND_LAST;
+	} else {
+		qp->s_state = IB_OPCODE_UC_SEND_LAST;
+		qp->r_state = IB_OPCODE_UC_SEND_LAST;
+	}
+	qp->s_ack_state = IB_OPCODE_RC_ACKNOWLEDGE;
+	qp->r_nak_state = 0;
+	qp->r_aflags = 0;
+	qp->r_flags = 0;
+	qp->s_head = 0;
+	qp->s_tail = 0;
+	qp->s_cur = 0;
+	qp->s_acked = 0;
+	qp->s_last = 0;
+	qp->s_ssn = 1;
+	qp->s_lsn = 0;
+	qp->s_mig_state = IB_MIG_MIGRATED;
+	memset(qp->s_ack_queue, 0, sizeof(qp->s_ack_queue));
+	qp->r_head_ack_queue = 0;
+	qp->s_tail_ack_queue = 0;
+	qp->s_num_rd_atomic = 0;
+	if (qp->r_rq.wq) {
+		qp->r_rq.wq->head = 0;
+		qp->r_rq.wq->tail = 0;
+	}
+	qp->r_sge.num_sge = 0;
+}
+
+static void clear_mr_refs(struct qib_qp *qp, int clr_sends)
+{
+	unsigned n;
+
+	if (test_and_clear_bit(QIB_R_REWIND_SGE, &qp->r_aflags))
+		while (qp->s_rdma_read_sge.num_sge) {
+			atomic_dec(&qp->s_rdma_read_sge.sge.mr->refcount);
+			if (--qp->s_rdma_read_sge.num_sge)
+				qp->s_rdma_read_sge.sge =
+					*qp->s_rdma_read_sge.sg_list++;
+		}
+
+	while (qp->r_sge.num_sge) {
+		atomic_dec(&qp->r_sge.sge.mr->refcount);
+		if (--qp->r_sge.num_sge)
+			qp->r_sge.sge = *qp->r_sge.sg_list++;
+	}
+
+	if (clr_sends) {
+		while (qp->s_last != qp->s_head) {
+			struct qib_swqe *wqe = get_swqe_ptr(qp, qp->s_last);
+			unsigned i;
+
+			for (i = 0; i < wqe->wr.num_sge; i++) {
+				struct qib_sge *sge = &wqe->sg_list[i];
+
+				atomic_dec(&sge->mr->refcount);
+			}
+			if (qp->ibqp.qp_type == IB_QPT_UD ||
+			    qp->ibqp.qp_type == IB_QPT_SMI ||
+			    qp->ibqp.qp_type == IB_QPT_GSI)
+				atomic_dec(&to_iah(wqe->wr.wr.ud.ah)->refcount);
+			if (++qp->s_last >= qp->s_size)
+				qp->s_last = 0;
+		}
+		if (qp->s_rdma_mr) {
+			atomic_dec(&qp->s_rdma_mr->refcount);
+			qp->s_rdma_mr = NULL;
+		}
+	}
+
+	if (qp->ibqp.qp_type != IB_QPT_RC)
+		return;
+
+	for (n = 0; n < ARRAY_SIZE(qp->s_ack_queue); n++) {
+		struct qib_ack_entry *e = &qp->s_ack_queue[n];
+
+		if (e->opcode == IB_OPCODE_RC_RDMA_READ_REQUEST &&
+		    e->rdma_sge.mr) {
+			atomic_dec(&e->rdma_sge.mr->refcount);
+			e->rdma_sge.mr = NULL;
+		}
+	}
+}
+
+/**
+ * qib_error_qp - put a QP into the error state
+ * @qp: the QP to put into the error state
+ * @err: the receive completion error to signal if a RWQE is active
+ *
+ * Flushes both send and receive work queues.
+ * Returns true if last WQE event should be generated.
+ * The QP s_lock should be held and interrupts disabled.
+ * If we are already in error state, just return.
+ */
+int qib_error_qp(struct qib_qp *qp, enum ib_wc_status err)
+{
+	struct qib_ibdev *dev = to_idev(qp->ibqp.device);
+	struct ib_wc wc;
+	int ret = 0;
+
+	if (qp->state == IB_QPS_ERR || qp->state == IB_QPS_RESET)
+		goto bail;
+
+	qp->state = IB_QPS_ERR;
+
+	if (qp->s_flags & (QIB_S_TIMER | QIB_S_WAIT_RNR)) {
+		qp->s_flags &= ~(QIB_S_TIMER | QIB_S_WAIT_RNR);
+		del_timer(&qp->s_timer);
+	}
+	spin_lock(&dev->pending_lock);
+	if (!list_empty(&qp->iowait) && !(qp->s_flags & QIB_S_BUSY)) {
+		qp->s_flags &= ~QIB_S_ANY_WAIT_IO;
+		list_del_init(&qp->iowait);
+	}
+	spin_unlock(&dev->pending_lock);
+
+	if (!(qp->s_flags & QIB_S_BUSY)) {
+		qp->s_hdrwords = 0;
+		if (qp->s_rdma_mr) {
+			atomic_dec(&qp->s_rdma_mr->refcount);
+			qp->s_rdma_mr = NULL;
+		}
+		if (qp->s_tx) {
+			qib_put_txreq(qp->s_tx);
+			qp->s_tx = NULL;
+		}
+	}
+
+	/* Schedule the sending tasklet to drain the send work queue. */
+	if (qp->s_last != qp->s_head)
+		qib_schedule_send(qp);
+
+	clear_mr_refs(qp, 0);
+
+	memset(&wc, 0, sizeof(wc));
+	wc.qp = &qp->ibqp;
+	wc.opcode = IB_WC_RECV;
+
+	if (test_and_clear_bit(QIB_R_WRID_VALID, &qp->r_aflags)) {
+		wc.wr_id = qp->r_wr_id;
+		wc.status = err;
+		qib_cq_enter(to_icq(qp->ibqp.recv_cq), &wc, 1);
+	}
+	wc.status = IB_WC_WR_FLUSH_ERR;
+
+	if (qp->r_rq.wq) {
+		struct qib_rwq *wq;
+		u32 head;
+		u32 tail;
+
+		spin_lock(&qp->r_rq.lock);
+
+		/* sanity check pointers before trusting them */
+		wq = qp->r_rq.wq;
+		head = wq->head;
+		if (head >= qp->r_rq.size)
+			head = 0;
+		tail = wq->tail;
+		if (tail >= qp->r_rq.size)
+			tail = 0;
+		while (tail != head) {
+			wc.wr_id = get_rwqe_ptr(&qp->r_rq, tail)->wr_id;
+			if (++tail >= qp->r_rq.size)
+				tail = 0;
+			qib_cq_enter(to_icq(qp->ibqp.recv_cq), &wc, 1);
+		}
+		wq->tail = tail;
+
+		spin_unlock(&qp->r_rq.lock);
+	} else if (qp->ibqp.event_handler)
+		ret = 1;
+
+bail:
+	return ret;
+}
+
+/**
+ * qib_modify_qp - modify the attributes of a queue pair
+ * @ibqp: the queue pair who's attributes we're modifying
+ * @attr: the new attributes
+ * @attr_mask: the mask of attributes to modify
+ * @udata: user data for libibverbs.so
+ *
+ * Returns 0 on success, otherwise returns an errno.
+ */
+int qib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
+		  int attr_mask, struct ib_udata *udata)
+{
+	struct qib_ibdev *dev = to_idev(ibqp->device);
+	struct qib_qp *qp = to_iqp(ibqp);
+	enum ib_qp_state cur_state, new_state;
+	struct ib_event ev;
+	int lastwqe = 0;
+	int mig = 0;
+	int ret;
+	u32 pmtu = 0; /* for gcc warning only */
+
+	spin_lock_irq(&qp->r_lock);
+	spin_lock(&qp->s_lock);
+
+	cur_state = attr_mask & IB_QP_CUR_STATE ?
+		attr->cur_qp_state : qp->state;
+	new_state = attr_mask & IB_QP_STATE ? attr->qp_state : cur_state;
+
+	if (!ib_modify_qp_is_ok(cur_state, new_state, ibqp->qp_type,
+				attr_mask))
+		goto inval;
+
+	if (attr_mask & IB_QP_AV) {
+		if (attr->ah_attr.dlid >= QIB_MULTICAST_LID_BASE)
+			goto inval;
+		if (qib_check_ah(qp->ibqp.device, &attr->ah_attr))
+			goto inval;
+	}
+
+	if (attr_mask & IB_QP_ALT_PATH) {
+		if (attr->alt_ah_attr.dlid >= QIB_MULTICAST_LID_BASE)
+			goto inval;
+		if (qib_check_ah(qp->ibqp.device, &attr->alt_ah_attr))
+			goto inval;
+		if (attr->alt_pkey_index >= qib_get_npkeys(dd_from_dev(dev)))
+			goto inval;
+	}
+
+	if (attr_mask & IB_QP_PKEY_INDEX)
+		if (attr->pkey_index >= qib_get_npkeys(dd_from_dev(dev)))
+			goto inval;
+
+	if (attr_mask & IB_QP_MIN_RNR_TIMER)
+		if (attr->min_rnr_timer > 31)
+			goto inval;
+
+	if (attr_mask & IB_QP_PORT)
+		if (qp->ibqp.qp_type == IB_QPT_SMI ||
+		    qp->ibqp.qp_type == IB_QPT_GSI ||
+		    attr->port_num == 0 ||
+		    attr->port_num > ibqp->device->phys_port_cnt)
+			goto inval;
+
+	if (attr_mask & IB_QP_DEST_QPN)
+		if (attr->dest_qp_num > QIB_QPN_MASK)
+			goto inval;
+
+	if (attr_mask & IB_QP_RETRY_CNT)
+		if (attr->retry_cnt > 7)
+			goto inval;
+
+	if (attr_mask & IB_QP_RNR_RETRY)
+		if (attr->rnr_retry > 7)
+			goto inval;
+
+	/*
+	 * Don't allow invalid path_mtu values.  OK to set greater
+	 * than the active mtu (or even the max_cap, if we have tuned
+	 * that to a small mtu.  We'll set qp->path_mtu
+	 * to the lesser of requested attribute mtu and active,
+	 * for packetizing messages.
+	 * Note that the QP port has to be set in INIT and MTU in RTR.
+	 */
+	if (attr_mask & IB_QP_PATH_MTU) {
+		struct qib_devdata *dd = dd_from_dev(dev);
+		int mtu, pidx = qp->port_num - 1;
+
+		mtu = ib_mtu_enum_to_int(attr->path_mtu);
+		if (mtu == -1)
+			goto inval;
+		if (mtu > dd->pport[pidx].ibmtu) {
+			switch (dd->pport[pidx].ibmtu) {
+			case 4096:
+				pmtu = IB_MTU_4096;
+				break;
+			case 2048:
+				pmtu = IB_MTU_2048;
+				break;
+			case 1024:
+				pmtu = IB_MTU_1024;
+				break;
+			case 512:
+				pmtu = IB_MTU_512;
+				break;
+			case 256:
+				pmtu = IB_MTU_256;
+				break;
+			default:
+				pmtu = IB_MTU_2048;
+			}
+		} else
+			pmtu = attr->path_mtu;
+	}
+
+	if (attr_mask & IB_QP_PATH_MIG_STATE) {
+		if (attr->path_mig_state == IB_MIG_REARM) {
+			if (qp->s_mig_state == IB_MIG_ARMED)
+				goto inval;
+			if (new_state != IB_QPS_RTS)
+				goto inval;
+		} else if (attr->path_mig_state == IB_MIG_MIGRATED) {
+			if (qp->s_mig_state == IB_MIG_REARM)
+				goto inval;
+			if (new_state != IB_QPS_RTS && new_state != IB_QPS_SQD)
+				goto inval;
+			if (qp->s_mig_state == IB_MIG_ARMED)
+				mig = 1;
+		} else
+			goto inval;
+	}
+
+	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC)
+		if (attr->max_dest_rd_atomic > QIB_MAX_RDMA_ATOMIC)
+			goto inval;
+
+	switch (new_state) {
+	case IB_QPS_RESET:
+		if (qp->state != IB_QPS_RESET) {
+			qp->state = IB_QPS_RESET;
+			spin_lock(&dev->pending_lock);
+			if (!list_empty(&qp->iowait))
+				list_del_init(&qp->iowait);
+			spin_unlock(&dev->pending_lock);
+			qp->s_flags &= ~(QIB_S_TIMER | QIB_S_ANY_WAIT);
+			spin_unlock(&qp->s_lock);
+			spin_unlock_irq(&qp->r_lock);
+			/* Stop the sending work queue and retry timer */
+			cancel_work_sync(&qp->s_work);
+			del_timer_sync(&qp->s_timer);
+			wait_event(qp->wait_dma, !atomic_read(&qp->s_dma_busy));
+			if (qp->s_tx) {
+				qib_put_txreq(qp->s_tx);
+				qp->s_tx = NULL;
+			}
+			remove_qp(dev, qp);
+			wait_event(qp->wait, !atomic_read(&qp->refcount));
+			spin_lock_irq(&qp->r_lock);
+			spin_lock(&qp->s_lock);
+			clear_mr_refs(qp, 1);
+			qib_reset_qp(qp, ibqp->qp_type);
+		}
+		break;
+
+	case IB_QPS_RTR:
+		/* Allow event to retrigger if QP set to RTR more than once */
+		qp->r_flags &= ~QIB_R_COMM_EST;
+		qp->state = new_state;
+		break;
+
+	case IB_QPS_SQD:
+		qp->s_draining = qp->s_last != qp->s_cur;
+		qp->state = new_state;
+		break;
+
+	case IB_QPS_SQE:
+		if (qp->ibqp.qp_type == IB_QPT_RC)
+			goto inval;
+		qp->state = new_state;
+		break;
+
+	case IB_QPS_ERR:
+		lastwqe = qib_error_qp(qp, IB_WC_WR_FLUSH_ERR);
+		break;
+
+	default:
+		qp->state = new_state;
+		break;
+	}
+
+	if (attr_mask & IB_QP_PKEY_INDEX)
+		qp->s_pkey_index = attr->pkey_index;
+
+	if (attr_mask & IB_QP_PORT)
+		qp->port_num = attr->port_num;
+
+	if (attr_mask & IB_QP_DEST_QPN)
+		qp->remote_qpn = attr->dest_qp_num;
+
+	if (attr_mask & IB_QP_SQ_PSN) {
+		qp->s_next_psn = attr->sq_psn & QIB_PSN_MASK;
+		qp->s_psn = qp->s_next_psn;
+		qp->s_sending_psn = qp->s_next_psn;
+		qp->s_last_psn = qp->s_next_psn - 1;
+		qp->s_sending_hpsn = qp->s_last_psn;
+	}
+
+	if (attr_mask & IB_QP_RQ_PSN)
+		qp->r_psn = attr->rq_psn & QIB_PSN_MASK;
+
+	if (attr_mask & IB_QP_ACCESS_FLAGS)
+		qp->qp_access_flags = attr->qp_access_flags;
+
+	if (attr_mask & IB_QP_AV) {
+		qp->remote_ah_attr = attr->ah_attr;
+		qp->s_srate = attr->ah_attr.static_rate;
+	}
+
+	if (attr_mask & IB_QP_ALT_PATH) {
+		qp->alt_ah_attr = attr->alt_ah_attr;
+		qp->s_alt_pkey_index = attr->alt_pkey_index;
+	}
+
+	if (attr_mask & IB_QP_PATH_MIG_STATE) {
+		qp->s_mig_state = attr->path_mig_state;
+		if (mig) {
+			qp->remote_ah_attr = qp->alt_ah_attr;
+			qp->port_num = qp->alt_ah_attr.port_num;
+			qp->s_pkey_index = qp->s_alt_pkey_index;
+		}
+	}
+
+	if (attr_mask & IB_QP_PATH_MTU)
+		qp->path_mtu = pmtu;
+
+	if (attr_mask & IB_QP_RETRY_CNT) {
+		qp->s_retry_cnt = attr->retry_cnt;
+		qp->s_retry = attr->retry_cnt;
+	}
+
+	if (attr_mask & IB_QP_RNR_RETRY) {
+		qp->s_rnr_retry_cnt = attr->rnr_retry;
+		qp->s_rnr_retry = attr->rnr_retry;
+	}
+
+	if (attr_mask & IB_QP_MIN_RNR_TIMER)
+		qp->r_min_rnr_timer = attr->min_rnr_timer;
+
+	if (attr_mask & IB_QP_TIMEOUT)
+		qp->timeout = attr->timeout;
+
+	if (attr_mask & IB_QP_QKEY)
+		qp->qkey = attr->qkey;
+
+	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC)
+		qp->r_max_rd_atomic = attr->max_dest_rd_atomic;
+
+	if (attr_mask & IB_QP_MAX_QP_RD_ATOMIC)
+		qp->s_max_rd_atomic = attr->max_rd_atomic;
+
+	spin_unlock(&qp->s_lock);
+	spin_unlock_irq(&qp->r_lock);
+
+	if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT)
+		insert_qp(dev, qp);
+
+	if (lastwqe) {
+		ev.device = qp->ibqp.device;
+		ev.element.qp = &qp->ibqp;
+		ev.event = IB_EVENT_QP_LAST_WQE_REACHED;
+		qp->ibqp.event_handler(&ev, qp->ibqp.qp_context);
+	}
+	if (mig) {
+		ev.device = qp->ibqp.device;
+		ev.element.qp = &qp->ibqp;
+		ev.event = IB_EVENT_PATH_MIG;
+		qp->ibqp.event_handler(&ev, qp->ibqp.qp_context);
+	}
+	ret = 0;
+	goto bail;
+
+inval:
+	spin_unlock(&qp->s_lock);
+	spin_unlock_irq(&qp->r_lock);
+	ret = -EINVAL;
+
+bail:
+	return ret;
+}
+
+int qib_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
+		 int attr_mask, struct ib_qp_init_attr *init_attr)
+{
+	struct qib_qp *qp = to_iqp(ibqp);
+
+	attr->qp_state = qp->state;
+	attr->cur_qp_state = attr->qp_state;
+	attr->path_mtu = qp->path_mtu;
+	attr->path_mig_state = qp->s_mig_state;
+	attr->qkey = qp->qkey;
+	attr->rq_psn = qp->r_psn & QIB_PSN_MASK;
+	attr->sq_psn = qp->s_next_psn & QIB_PSN_MASK;
+	attr->dest_qp_num = qp->remote_qpn;
+	attr->qp_access_flags = qp->qp_access_flags;
+	attr->cap.max_send_wr = qp->s_size - 1;
+	attr->cap.max_recv_wr = qp->ibqp.srq ? 0 : qp->r_rq.size - 1;
+	attr->cap.max_send_sge = qp->s_max_sge;
+	attr->cap.max_recv_sge = qp->r_rq.max_sge;
+	attr->cap.max_inline_data = 0;
+	attr->ah_attr = qp->remote_ah_attr;
+	attr->alt_ah_attr = qp->alt_ah_attr;
+	attr->pkey_index = qp->s_pkey_index;
+	attr->alt_pkey_index = qp->s_alt_pkey_index;
+	attr->en_sqd_async_notify = 0;
+	attr->sq_draining = qp->s_draining;
+	attr->max_rd_atomic = qp->s_max_rd_atomic;
+	attr->max_dest_rd_atomic = qp->r_max_rd_atomic;
+	attr->min_rnr_timer = qp->r_min_rnr_timer;
+	attr->port_num = qp->port_num;
+	attr->timeout = qp->timeout;
+	attr->retry_cnt = qp->s_retry_cnt;
+	attr->rnr_retry = qp->s_rnr_retry_cnt;
+	attr->alt_port_num = qp->alt_ah_attr.port_num;
+	attr->alt_timeout = qp->alt_timeout;
+
+	init_attr->event_handler = qp->ibqp.event_handler;
+	init_attr->qp_context = qp->ibqp.qp_context;
+	init_attr->send_cq = qp->ibqp.send_cq;
+	init_attr->recv_cq = qp->ibqp.recv_cq;
+	init_attr->srq = qp->ibqp.srq;
+	init_attr->cap = attr->cap;
+	if (qp->s_flags & QIB_S_SIGNAL_REQ_WR)
+		init_attr->sq_sig_type = IB_SIGNAL_REQ_WR;
+	else
+		init_attr->sq_sig_type = IB_SIGNAL_ALL_WR;
+	init_attr->qp_type = qp->ibqp.qp_type;
+	init_attr->port_num = qp->port_num;
+	return 0;
+}
+
+/**
+ * qib_compute_aeth - compute the AETH (syndrome + MSN)
+ * @qp: the queue pair to compute the AETH for
+ *
+ * Returns the AETH.
+ */
+__be32 qib_compute_aeth(struct qib_qp *qp)
+{
+	u32 aeth = qp->r_msn & QIB_MSN_MASK;
+
+	if (qp->ibqp.srq) {
+		/*
+		 * Shared receive queues don't generate credits.
+		 * Set the credit field to the invalid value.
+		 */
+		aeth |= QIB_AETH_CREDIT_INVAL << QIB_AETH_CREDIT_SHIFT;
+	} else {
+		u32 min, max, x;
+		u32 credits;
+		struct qib_rwq *wq = qp->r_rq.wq;
+		u32 head;
+		u32 tail;
+
+		/* sanity check pointers before trusting them */
+		head = wq->head;
+		if (head >= qp->r_rq.size)
+			head = 0;
+		tail = wq->tail;
+		if (tail >= qp->r_rq.size)
+			tail = 0;
+		/*
+		 * Compute the number of credits available (RWQEs).
+		 * XXX Not holding the r_rq.lock here so there is a small
+		 * chance that the pair of reads are not atomic.
+		 */
+		credits = head - tail;
+		if ((int)credits < 0)
+			credits += qp->r_rq.size;
+		/*
+		 * Binary search the credit table to find the code to
+		 * use.
+		 */
+		min = 0;
+		max = 31;
+		for (;;) {
+			x = (min + max) / 2;
+			if (credit_table[x] == credits)
+				break;
+			if (credit_table[x] > credits)
+				max = x;
+			else if (min == x)
+				break;
+			else
+				min = x;
+		}
+		aeth |= x << QIB_AETH_CREDIT_SHIFT;
+	}
+	return cpu_to_be32(aeth);
+}
+
+/**
+ * qib_create_qp - create a queue pair for a device
+ * @ibpd: the protection domain who's device we create the queue pair for
+ * @init_attr: the attributes of the queue pair
+ * @udata: user data for libibverbs.so
+ *
+ * Returns the queue pair on success, otherwise returns an errno.
+ *
+ * Called by the ib_create_qp() core verbs function.
+ */
+struct ib_qp *qib_create_qp(struct ib_pd *ibpd,
+			    struct ib_qp_init_attr *init_attr,
+			    struct ib_udata *udata)
+{
+	struct qib_qp *qp;
+	int err;
+	struct qib_swqe *swq = NULL;
+	struct qib_ibdev *dev;
+	struct qib_devdata *dd;
+	size_t sz;
+	size_t sg_list_sz;
+	struct ib_qp *ret;
+
+	if (init_attr->cap.max_send_sge > ib_qib_max_sges ||
+	    init_attr->cap.max_send_wr > ib_qib_max_qp_wrs) {
+		ret = ERR_PTR(-EINVAL);
+		goto bail;
+	}
+
+	/* Check receive queue parameters if no SRQ is specified. */
+	if (!init_attr->srq) {
+		if (init_attr->cap.max_recv_sge > ib_qib_max_sges ||
+		    init_attr->cap.max_recv_wr > ib_qib_max_qp_wrs) {
+			ret = ERR_PTR(-EINVAL);
+			goto bail;
+		}
+		if (init_attr->cap.max_send_sge +
+		    init_attr->cap.max_send_wr +
+		    init_attr->cap.max_recv_sge +
+		    init_attr->cap.max_recv_wr == 0) {
+			ret = ERR_PTR(-EINVAL);
+			goto bail;
+		}
+	}
+
+	switch (init_attr->qp_type) {
+	case IB_QPT_SMI:
+	case IB_QPT_GSI:
+		if (init_attr->port_num == 0 ||
+		    init_attr->port_num > ibpd->device->phys_port_cnt) {
+			ret = ERR_PTR(-EINVAL);
+			goto bail;
+		}
+	case IB_QPT_UC:
+	case IB_QPT_RC:
+	case IB_QPT_UD:
+		sz = sizeof(struct qib_sge) *
+			init_attr->cap.max_send_sge +
+			sizeof(struct qib_swqe);
+		swq = vmalloc((init_attr->cap.max_send_wr + 1) * sz);
+		if (swq == NULL) {
+			ret = ERR_PTR(-ENOMEM);
+			goto bail;
+		}
+		sz = sizeof(*qp);
+		sg_list_sz = 0;
+		if (init_attr->srq) {
+			struct qib_srq *srq = to_isrq(init_attr->srq);
+
+			if (srq->rq.max_sge > 1)
+				sg_list_sz = sizeof(*qp->r_sg_list) *
+					(srq->rq.max_sge - 1);
+		} else if (init_attr->cap.max_recv_sge > 1)
+			sg_list_sz = sizeof(*qp->r_sg_list) *
+				(init_attr->cap.max_recv_sge - 1);
+		qp = kzalloc(sz + sg_list_sz, GFP_KERNEL);
+		if (!qp) {
+			ret = ERR_PTR(-ENOMEM);
+			goto bail_swq;
+		}
+		if (init_attr->srq)
+			sz = 0;
+		else {
+			qp->r_rq.size = init_attr->cap.max_recv_wr + 1;
+			qp->r_rq.max_sge = init_attr->cap.max_recv_sge;
+			sz = (sizeof(struct ib_sge) * qp->r_rq.max_sge) +
+				sizeof(struct qib_rwqe);
+			qp->r_rq.wq = vmalloc_user(sizeof(struct qib_rwq) +
+						   qp->r_rq.size * sz);
+			if (!qp->r_rq.wq) {
+				ret = ERR_PTR(-ENOMEM);
+				goto bail_qp;
+			}
+		}
+
+		/*
+		 * ib_create_qp() will initialize qp->ibqp
+		 * except for qp->ibqp.qp_num.
+		 */
+		spin_lock_init(&qp->r_lock);
+		spin_lock_init(&qp->s_lock);
+		spin_lock_init(&qp->r_rq.lock);
+		atomic_set(&qp->refcount, 0);
+		init_waitqueue_head(&qp->wait);
+		init_waitqueue_head(&qp->wait_dma);
+		init_timer(&qp->s_timer);
+		qp->s_timer.data = (unsigned long)qp;
+		INIT_WORK(&qp->s_work, qib_do_send);
+		INIT_LIST_HEAD(&qp->iowait);
+		INIT_LIST_HEAD(&qp->rspwait);
+		qp->state = IB_QPS_RESET;
+		qp->s_wq = swq;
+		qp->s_size = init_attr->cap.max_send_wr + 1;
+		qp->s_max_sge = init_attr->cap.max_send_sge;
+		if (init_attr->sq_sig_type == IB_SIGNAL_REQ_WR)
+			qp->s_flags = QIB_S_SIGNAL_REQ_WR;
+		dev = to_idev(ibpd->device);
+		dd = dd_from_dev(dev);
+		err = alloc_qpn(dd, &dev->qpn_table, init_attr->qp_type,
+				init_attr->port_num);
+		if (err < 0) {
+			ret = ERR_PTR(err);
+			vfree(qp->r_rq.wq);
+			goto bail_qp;
+		}
+		qp->ibqp.qp_num = err;
+		qp->port_num = init_attr->port_num;
+		qp->processor_id = smp_processor_id();
+		qib_reset_qp(qp, init_attr->qp_type);
+		break;
+
+	default:
+		/* Don't support raw QPs */
+		ret = ERR_PTR(-ENOSYS);
+		goto bail;
+	}
+
+	init_attr->cap.max_inline_data = 0;
+
+	/*
+	 * Return the address of the RWQ as the offset to mmap.
+	 * See qib_mmap() for details.
+	 */
+	if (udata && udata->outlen >= sizeof(__u64)) {
+		if (!qp->r_rq.wq) {
+			__u64 offset = 0;
+
+			err = ib_copy_to_udata(udata, &offset,
+					       sizeof(offset));
+			if (err) {
+				ret = ERR_PTR(err);
+				goto bail_ip;
+			}
+		} else {
+			u32 s = sizeof(struct qib_rwq) + qp->r_rq.size * sz;
+
+			qp->ip = qib_create_mmap_info(dev, s,
+						      ibpd->uobject->context,
+						      qp->r_rq.wq);
+			if (!qp->ip) {
+				ret = ERR_PTR(-ENOMEM);
+				goto bail_ip;
+			}
+
+			err = ib_copy_to_udata(udata, &(qp->ip->offset),
+					       sizeof(qp->ip->offset));
+			if (err) {
+				ret = ERR_PTR(err);
+				goto bail_ip;
+			}
+		}
+	}
+
+	spin_lock(&dev->n_qps_lock);
+	if (dev->n_qps_allocated == ib_qib_max_qps) {
+		spin_unlock(&dev->n_qps_lock);
+		ret = ERR_PTR(-ENOMEM);
+		goto bail_ip;
+	}
+
+	dev->n_qps_allocated++;
+	spin_unlock(&dev->n_qps_lock);
+
+	if (qp->ip) {
+		spin_lock_irq(&dev->pending_lock);
+		list_add(&qp->ip->pending_mmaps, &dev->pending_mmaps);
+		spin_unlock_irq(&dev->pending_lock);
+	}
+
+	ret = &qp->ibqp;
+	goto bail;
+
+bail_ip:
+	if (qp->ip)
+		kref_put(&qp->ip->ref, qib_release_mmap_info);
+	else
+		vfree(qp->r_rq.wq);
+	free_qpn(&dev->qpn_table, qp->ibqp.qp_num);
+bail_qp:
+	kfree(qp);
+bail_swq:
+	vfree(swq);
+bail:
+	return ret;
+}
+
+/**
+ * qib_destroy_qp - destroy a queue pair
+ * @ibqp: the queue pair to destroy
+ *
+ * Returns 0 on success.
+ *
+ * Note that this can be called while the QP is actively sending or
+ * receiving!
+ */
+int qib_destroy_qp(struct ib_qp *ibqp)
+{
+	struct qib_qp *qp = to_iqp(ibqp);
+	struct qib_ibdev *dev = to_idev(ibqp->device);
+
+	/* Make sure HW and driver activity is stopped. */
+	spin_lock_irq(&qp->s_lock);
+	if (qp->state != IB_QPS_RESET) {
+		qp->state = IB_QPS_RESET;
+		spin_lock(&dev->pending_lock);
+		if (!list_empty(&qp->iowait))
+			list_del_init(&qp->iowait);
+		spin_unlock(&dev->pending_lock);
+		qp->s_flags &= ~(QIB_S_TIMER | QIB_S_ANY_WAIT);
+		spin_unlock_irq(&qp->s_lock);
+		cancel_work_sync(&qp->s_work);
+		del_timer_sync(&qp->s_timer);
+		wait_event(qp->wait_dma, !atomic_read(&qp->s_dma_busy));
+		if (qp->s_tx) {
+			qib_put_txreq(qp->s_tx);
+			qp->s_tx = NULL;
+		}
+		remove_qp(dev, qp);
+		wait_event(qp->wait, !atomic_read(&qp->refcount));
+		clear_mr_refs(qp, 1);
+	} else
+		spin_unlock_irq(&qp->s_lock);
+
+	/* all user's cleaned up, mark it available */
+	free_qpn(&dev->qpn_table, qp->ibqp.qp_num);
+	spin_lock(&dev->n_qps_lock);
+	dev->n_qps_allocated--;
+	spin_unlock(&dev->n_qps_lock);
+
+	if (qp->ip)
+		kref_put(&qp->ip->ref, qib_release_mmap_info);
+	else
+		vfree(qp->r_rq.wq);
+	vfree(qp->s_wq);
+	kfree(qp);
+	return 0;
+}
+
+/**
+ * qib_init_qpn_table - initialize the QP number table for a device
+ * @qpt: the QPN table
+ */
+void qib_init_qpn_table(struct qib_devdata *dd, struct qib_qpn_table *qpt)
+{
+	spin_lock_init(&qpt->lock);
+	qpt->last = 1;          /* start with QPN 2 */
+	qpt->nmaps = 1;
+	qpt->mask = dd->qpn_mask;
+}
+
+/**
+ * qib_free_qpn_table - free the QP number table for a device
+ * @qpt: the QPN table
+ */
+void qib_free_qpn_table(struct qib_qpn_table *qpt)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(qpt->map); i++)
+		if (qpt->map[i].page)
+			free_page((unsigned long) qpt->map[i].page);
+}
+
+/**
+ * qib_get_credit - flush the send work queue of a QP
+ * @qp: the qp who's send work queue to flush
+ * @aeth: the Acknowledge Extended Transport Header
+ *
+ * The QP s_lock should be held.
+ */
+void qib_get_credit(struct qib_qp *qp, u32 aeth)
+{
+	u32 credit = (aeth >> QIB_AETH_CREDIT_SHIFT) & QIB_AETH_CREDIT_MASK;
+
+	/*
+	 * If the credit is invalid, we can send
+	 * as many packets as we like.  Otherwise, we have to
+	 * honor the credit field.
+	 */
+	if (credit == QIB_AETH_CREDIT_INVAL) {
+		if (!(qp->s_flags & QIB_S_UNLIMITED_CREDIT)) {
+			qp->s_flags |= QIB_S_UNLIMITED_CREDIT;
+			if (qp->s_flags & QIB_S_WAIT_SSN_CREDIT) {
+				qp->s_flags &= ~QIB_S_WAIT_SSN_CREDIT;
+				qib_schedule_send(qp);
+			}
+		}
+	} else if (!(qp->s_flags & QIB_S_UNLIMITED_CREDIT)) {
+		/* Compute new LSN (i.e., MSN + credit) */
+		credit = (aeth + credit_table[credit]) & QIB_MSN_MASK;
+		if (qib_cmp24(credit, qp->s_lsn) > 0) {
+			qp->s_lsn = credit;
+			if (qp->s_flags & QIB_S_WAIT_SSN_CREDIT) {
+				qp->s_flags &= ~QIB_S_WAIT_SSN_CREDIT;
+				qib_schedule_send(qp);
+			}
+		}
+	}
+}
diff --git a/drivers/infiniband/hw/qib/qib_qsfp.c b/drivers/infiniband/hw/qib/qib_qsfp.c
new file mode 100644
index 000000000000..35b3604b691d
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_qsfp.c
@@ -0,0 +1,564 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/delay.h>
+#include <linux/pci.h>
+#include <linux/vmalloc.h>
+
+#include "qib.h"
+#include "qib_qsfp.h"
+
+/*
+ * QSFP support for ib_qib driver, using "Two Wire Serial Interface" driver
+ * in qib_twsi.c
+ */
+#define QSFP_MAX_RETRY 4
+
+static int qsfp_read(struct qib_pportdata *ppd, int addr, void *bp, int len)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u32 out, mask;
+	int ret, cnt, pass = 0;
+	int stuck = 0;
+	u8 *buff = bp;
+
+	ret = mutex_lock_interruptible(&dd->eep_lock);
+	if (ret)
+		goto no_unlock;
+
+	if (dd->twsi_eeprom_dev == QIB_TWSI_NO_DEV) {
+		ret = -ENXIO;
+		goto bail;
+	}
+
+	/*
+	 * We presume, if we are called at all, that this board has
+	 * QSFP. This is on the same i2c chain as the legacy parts,
+	 * but only responds if the module is selected via GPIO pins.
+	 * Further, there are very long setup and hold requirements
+	 * on MODSEL.
+	 */
+	mask = QSFP_GPIO_MOD_SEL_N | QSFP_GPIO_MOD_RST_N | QSFP_GPIO_LP_MODE;
+	out = QSFP_GPIO_MOD_RST_N | QSFP_GPIO_LP_MODE;
+	if (ppd->hw_pidx) {
+		mask <<= QSFP_GPIO_PORT2_SHIFT;
+		out <<= QSFP_GPIO_PORT2_SHIFT;
+	}
+
+	dd->f_gpio_mod(dd, out, mask, mask);
+
+	/*
+	 * Module could take up to 2 Msec to respond to MOD_SEL, and there
+	 * is no way to tell if it is ready, so we must wait.
+	 */
+	msleep(2);
+
+	/* Make sure TWSI bus is in sane state. */
+	ret = qib_twsi_reset(dd);
+	if (ret) {
+		qib_dev_porterr(dd, ppd->port,
+				"QSFP interface Reset for read failed\n");
+		ret = -EIO;
+		stuck = 1;
+		goto deselect;
+	}
+
+	/* All QSFP modules are at A0 */
+
+	cnt = 0;
+	while (cnt < len) {
+		unsigned in_page;
+		int wlen = len - cnt;
+		in_page = addr % QSFP_PAGESIZE;
+		if ((in_page + wlen) > QSFP_PAGESIZE)
+			wlen = QSFP_PAGESIZE - in_page;
+		ret = qib_twsi_blk_rd(dd, QSFP_DEV, addr, buff + cnt, wlen);
+		/* Some QSFP's fail first try. Retry as experiment */
+		if (ret && cnt == 0 && ++pass < QSFP_MAX_RETRY)
+			continue;
+		if (ret) {
+			/* qib_twsi_blk_rd() 1 for error, else 0 */
+			ret = -EIO;
+			goto deselect;
+		}
+		addr += wlen;
+		cnt += wlen;
+	}
+	ret = cnt;
+
+deselect:
+	/*
+	 * Module could take up to 10 uSec after transfer before
+	 * ready to respond to MOD_SEL negation, and there is no way
+	 * to tell if it is ready, so we must wait.
+	 */
+	udelay(10);
+	/* set QSFP MODSEL, RST. LP all high */
+	dd->f_gpio_mod(dd, mask, mask, mask);
+
+	/*
+	 * Module could take up to 2 Msec to respond to MOD_SEL
+	 * going away, and there is no way to tell if it is ready.
+	 * so we must wait.
+	 */
+	if (stuck)
+		qib_dev_err(dd, "QSFP interface bus stuck non-idle\n");
+
+	if (pass >= QSFP_MAX_RETRY && ret)
+		qib_dev_porterr(dd, ppd->port, "QSFP failed even retrying\n");
+	else if (pass)
+		qib_dev_porterr(dd, ppd->port, "QSFP retries: %d\n", pass);
+
+	msleep(2);
+
+bail:
+	mutex_unlock(&dd->eep_lock);
+
+no_unlock:
+	return ret;
+}
+
+/*
+ * qsfp_write
+ * We do not ordinarily write the QSFP, but this is needed to select
+ * the page on non-flat QSFPs, and possibly later unusual cases
+ */
+static int qib_qsfp_write(struct qib_pportdata *ppd, int addr, void *bp,
+			  int len)
+{
+	struct qib_devdata *dd = ppd->dd;
+	u32 out, mask;
+	int ret, cnt;
+	u8 *buff = bp;
+
+	ret = mutex_lock_interruptible(&dd->eep_lock);
+	if (ret)
+		goto no_unlock;
+
+	if (dd->twsi_eeprom_dev == QIB_TWSI_NO_DEV) {
+		ret = -ENXIO;
+		goto bail;
+	}
+
+	/*
+	 * We presume, if we are called at all, that this board has
+	 * QSFP. This is on the same i2c chain as the legacy parts,
+	 * but only responds if the module is selected via GPIO pins.
+	 * Further, there are very long setup and hold requirements
+	 * on MODSEL.
+	 */
+	mask = QSFP_GPIO_MOD_SEL_N | QSFP_GPIO_MOD_RST_N | QSFP_GPIO_LP_MODE;
+	out = QSFP_GPIO_MOD_RST_N | QSFP_GPIO_LP_MODE;
+	if (ppd->hw_pidx) {
+		mask <<= QSFP_GPIO_PORT2_SHIFT;
+		out <<= QSFP_GPIO_PORT2_SHIFT;
+	}
+	dd->f_gpio_mod(dd, out, mask, mask);
+
+	/*
+	 * Module could take up to 2 Msec to respond to MOD_SEL,
+	 * and there is no way to tell if it is ready, so we must wait.
+	 */
+	msleep(2);
+
+	/* Make sure TWSI bus is in sane state. */
+	ret = qib_twsi_reset(dd);
+	if (ret) {
+		qib_dev_porterr(dd, ppd->port,
+				"QSFP interface Reset for write failed\n");
+		ret = -EIO;
+		goto deselect;
+	}
+
+	/* All QSFP modules are at A0 */
+
+	cnt = 0;
+	while (cnt < len) {
+		unsigned in_page;
+		int wlen = len - cnt;
+		in_page = addr % QSFP_PAGESIZE;
+		if ((in_page + wlen) > QSFP_PAGESIZE)
+			wlen = QSFP_PAGESIZE - in_page;
+		ret = qib_twsi_blk_wr(dd, QSFP_DEV, addr, buff + cnt, wlen);
+		if (ret) {
+			/* qib_twsi_blk_wr() 1 for error, else 0 */
+			ret = -EIO;
+			goto deselect;
+		}
+		addr += wlen;
+		cnt += wlen;
+	}
+	ret = cnt;
+
+deselect:
+	/*
+	 * Module could take up to 10 uSec after transfer before
+	 * ready to respond to MOD_SEL negation, and there is no way
+	 * to tell if it is ready, so we must wait.
+	 */
+	udelay(10);
+	/* set QSFP MODSEL, RST, LP high */
+	dd->f_gpio_mod(dd, mask, mask, mask);
+	/*
+	 * Module could take up to 2 Msec to respond to MOD_SEL
+	 * going away, and there is no way to tell if it is ready.
+	 * so we must wait.
+	 */
+	msleep(2);
+
+bail:
+	mutex_unlock(&dd->eep_lock);
+
+no_unlock:
+	return ret;
+}
+
+/*
+ * For validation, we want to check the checksums, even of the
+ * fields we do not otherwise use. This function reads the bytes from
+ * <first> to <next-1> and returns the 8lsbs of the sum, or <0 for errors
+ */
+static int qsfp_cks(struct qib_pportdata *ppd, int first, int next)
+{
+	int ret;
+	u16 cks;
+	u8 bval;
+
+	cks = 0;
+	while (first < next) {
+		ret = qsfp_read(ppd, first, &bval, 1);
+		if (ret < 0)
+			goto bail;
+		cks += bval;
+		++first;
+	}
+	ret = cks & 0xFF;
+bail:
+	return ret;
+
+}
+
+int qib_refresh_qsfp_cache(struct qib_pportdata *ppd, struct qib_qsfp_cache *cp)
+{
+	int ret;
+	int idx;
+	u16 cks;
+	u32 mask;
+	u8 peek[4];
+
+	/* ensure sane contents on invalid reads, for cable swaps */
+	memset(cp, 0, sizeof(*cp));
+
+	mask = QSFP_GPIO_MOD_PRS_N;
+	if (ppd->hw_pidx)
+		mask <<= QSFP_GPIO_PORT2_SHIFT;
+
+	ret = ppd->dd->f_gpio_mod(ppd->dd, 0, 0, 0);
+	if (ret & mask) {
+		ret = -ENODEV;
+		goto bail;
+	}
+
+	ret = qsfp_read(ppd, 0, peek, 3);
+	if (ret < 0)
+		goto bail;
+	if ((peek[0] & 0xFE) != 0x0C)
+		qib_dev_porterr(ppd->dd, ppd->port,
+				"QSFP byte0 is 0x%02X, S/B 0x0C/D\n", peek[0]);
+
+	if ((peek[2] & 2) == 0) {
+		/*
+		 * If cable is paged, rather than "flat memory", we need to
+		 * set the page to zero, Even if it already appears to be zero.
+		 */
+		u8 poke = 0;
+		ret = qib_qsfp_write(ppd, 127, &poke, 1);
+		udelay(50);
+		if (ret != 1) {
+			qib_dev_porterr(ppd->dd, ppd->port,
+					"Failed QSFP Page set\n");
+			goto bail;
+		}
+	}
+
+	ret = qsfp_read(ppd, QSFP_MOD_ID_OFFS, &cp->id, 1);
+	if (ret < 0)
+		goto bail;
+	if ((cp->id & 0xFE) != 0x0C)
+		qib_dev_porterr(ppd->dd, ppd->port,
+				"QSFP ID byte is 0x%02X, S/B 0x0C/D\n", cp->id);
+	cks = cp->id;
+
+	ret = qsfp_read(ppd, QSFP_MOD_PWR_OFFS, &cp->pwr, 1);
+	if (ret < 0)
+		goto bail;
+	cks += cp->pwr;
+
+	ret = qsfp_cks(ppd, QSFP_MOD_PWR_OFFS + 1, QSFP_MOD_LEN_OFFS);
+	if (ret < 0)
+		goto bail;
+	cks += ret;
+
+	ret = qsfp_read(ppd, QSFP_MOD_LEN_OFFS, &cp->len, 1);
+	if (ret < 0)
+		goto bail;
+	cks += cp->len;
+
+	ret = qsfp_read(ppd, QSFP_MOD_TECH_OFFS, &cp->tech, 1);
+	if (ret < 0)
+		goto bail;
+	cks += cp->tech;
+
+	ret = qsfp_read(ppd, QSFP_VEND_OFFS, &cp->vendor, QSFP_VEND_LEN);
+	if (ret < 0)
+		goto bail;
+	for (idx = 0; idx < QSFP_VEND_LEN; ++idx)
+		cks += cp->vendor[idx];
+
+	ret = qsfp_read(ppd, QSFP_IBXCV_OFFS, &cp->xt_xcv, 1);
+	if (ret < 0)
+		goto bail;
+	cks += cp->xt_xcv;
+
+	ret = qsfp_read(ppd, QSFP_VOUI_OFFS, &cp->oui, QSFP_VOUI_LEN);
+	if (ret < 0)
+		goto bail;
+	for (idx = 0; idx < QSFP_VOUI_LEN; ++idx)
+		cks += cp->oui[idx];
+
+	ret = qsfp_read(ppd, QSFP_PN_OFFS, &cp->partnum, QSFP_PN_LEN);
+	if (ret < 0)
+		goto bail;
+	for (idx = 0; idx < QSFP_PN_LEN; ++idx)
+		cks += cp->partnum[idx];
+
+	ret = qsfp_read(ppd, QSFP_REV_OFFS, &cp->rev, QSFP_REV_LEN);
+	if (ret < 0)
+		goto bail;
+	for (idx = 0; idx < QSFP_REV_LEN; ++idx)
+		cks += cp->rev[idx];
+
+	ret = qsfp_read(ppd, QSFP_ATTEN_OFFS, &cp->atten, QSFP_ATTEN_LEN);
+	if (ret < 0)
+		goto bail;
+	for (idx = 0; idx < QSFP_ATTEN_LEN; ++idx)
+		cks += cp->atten[idx];
+
+	ret = qsfp_cks(ppd, QSFP_ATTEN_OFFS + QSFP_ATTEN_LEN, QSFP_CC_OFFS);
+	if (ret < 0)
+		goto bail;
+	cks += ret;
+
+	cks &= 0xFF;
+	ret = qsfp_read(ppd, QSFP_CC_OFFS, &cp->cks1, 1);
+	if (ret < 0)
+		goto bail;
+	if (cks != cp->cks1)
+		qib_dev_porterr(ppd->dd, ppd->port,
+				"QSFP cks1 is %02X, computed %02X\n", cp->cks1,
+				cks);
+
+	/* Second checksum covers 192 to (serial, date, lot) */
+	ret = qsfp_cks(ppd, QSFP_CC_OFFS + 1, QSFP_SN_OFFS);
+	if (ret < 0)
+		goto bail;
+	cks = ret;
+
+	ret = qsfp_read(ppd, QSFP_SN_OFFS, &cp->serial, QSFP_SN_LEN);
+	if (ret < 0)
+		goto bail;
+	for (idx = 0; idx < QSFP_SN_LEN; ++idx)
+		cks += cp->serial[idx];
+
+	ret = qsfp_read(ppd, QSFP_DATE_OFFS, &cp->date, QSFP_DATE_LEN);
+	if (ret < 0)
+		goto bail;
+	for (idx = 0; idx < QSFP_DATE_LEN; ++idx)
+		cks += cp->date[idx];
+
+	ret = qsfp_read(ppd, QSFP_LOT_OFFS, &cp->lot, QSFP_LOT_LEN);
+	if (ret < 0)
+		goto bail;
+	for (idx = 0; idx < QSFP_LOT_LEN; ++idx)
+		cks += cp->lot[idx];
+
+	ret = qsfp_cks(ppd, QSFP_LOT_OFFS + QSFP_LOT_LEN, QSFP_CC_EXT_OFFS);
+	if (ret < 0)
+		goto bail;
+	cks += ret;
+
+	ret = qsfp_read(ppd, QSFP_CC_EXT_OFFS, &cp->cks2, 1);
+	if (ret < 0)
+		goto bail;
+	cks &= 0xFF;
+	if (cks != cp->cks2)
+		qib_dev_porterr(ppd->dd, ppd->port,
+				"QSFP cks2 is %02X, computed %02X\n", cp->cks2,
+				cks);
+	return 0;
+
+bail:
+	cp->id = 0;
+	return ret;
+}
+
+const char * const qib_qsfp_devtech[16] = {
+	"850nm VCSEL", "1310nm VCSEL", "1550nm VCSEL", "1310nm FP",
+	"1310nm DFB", "1550nm DFB", "1310nm EML", "1550nm EML",
+	"Cu Misc", "1490nm DFB", "Cu NoEq", "Cu Eq",
+	"Undef", "Cu Active BothEq", "Cu FarEq", "Cu NearEq"
+};
+
+#define QSFP_DUMP_CHUNK 16 /* Holds longest string */
+#define QSFP_DEFAULT_HDR_CNT 224
+
+static const char *pwr_codes = "1.5W2.0W2.5W3.5W";
+
+/*
+ * Initialize structures that control access to QSFP. Called once per port
+ * on cards that support QSFP.
+ */
+void qib_qsfp_init(struct qib_qsfp_data *qd,
+		   void (*fevent)(struct work_struct *))
+{
+	u32 mask, highs;
+	int pins;
+
+	struct qib_devdata *dd = qd->ppd->dd;
+
+	/* Initialize work struct for later QSFP events */
+	INIT_WORK(&qd->work, fevent);
+
+	/*
+	 * Later, we may want more validation. For now, just set up pins and
+	 * blip reset. If module is present, call qib_refresh_qsfp_cache(),
+	 * to do further init.
+	 */
+	mask = QSFP_GPIO_MOD_SEL_N | QSFP_GPIO_MOD_RST_N | QSFP_GPIO_LP_MODE;
+	highs = mask - QSFP_GPIO_MOD_RST_N;
+	if (qd->ppd->hw_pidx) {
+		mask <<= QSFP_GPIO_PORT2_SHIFT;
+		highs <<= QSFP_GPIO_PORT2_SHIFT;
+	}
+	dd->f_gpio_mod(dd, highs, mask, mask);
+	udelay(20); /* Generous RST dwell */
+
+	dd->f_gpio_mod(dd, mask, mask, mask);
+	/* Spec says module can take up to two seconds! */
+	mask = QSFP_GPIO_MOD_PRS_N;
+	if (qd->ppd->hw_pidx)
+		mask <<= QSFP_GPIO_PORT2_SHIFT;
+
+	/* Do not try to wait here. Better to let event handle it */
+	pins = dd->f_gpio_mod(dd, 0, 0, 0);
+	if (pins & mask)
+		goto bail;
+	/* We see a module, but it may be unwise to look yet. Just schedule */
+	qd->t_insert = get_jiffies_64();
+	schedule_work(&qd->work);
+bail:
+	return;
+}
+
+void qib_qsfp_deinit(struct qib_qsfp_data *qd)
+{
+	/*
+	 * There is nothing to do here for now.  our
+	 * work is scheduled with schedule_work(), and
+	 * flush_scheduled_work() from remove_one will
+	 * block until all work ssetup with schedule_work()
+	 * completes.
+	 */
+}
+
+int qib_qsfp_dump(struct qib_pportdata *ppd, char *buf, int len)
+{
+	struct qib_qsfp_cache cd;
+	u8 bin_buff[QSFP_DUMP_CHUNK];
+	char lenstr[6];
+	int sofar, ret;
+	int bidx = 0;
+
+	sofar = 0;
+	ret = qib_refresh_qsfp_cache(ppd, &cd);
+	if (ret < 0)
+		goto bail;
+
+	lenstr[0] = ' ';
+	lenstr[1] = '\0';
+	if (QSFP_IS_CU(cd.tech))
+		sprintf(lenstr, "%dM ", cd.len);
+
+	sofar += scnprintf(buf + sofar, len - sofar, "PWR:%.3sW\n", pwr_codes +
+			   (QSFP_PWR(cd.pwr) * 4));
+
+	sofar += scnprintf(buf + sofar, len - sofar, "TECH:%s%s\n", lenstr,
+			   qib_qsfp_devtech[cd.tech >> 4]);
+
+	sofar += scnprintf(buf + sofar, len - sofar, "Vendor:%.*s\n",
+			   QSFP_VEND_LEN, cd.vendor);
+
+	sofar += scnprintf(buf + sofar, len - sofar, "OUI:%06X\n",
+			   QSFP_OUI(cd.oui));
+
+	sofar += scnprintf(buf + sofar, len - sofar, "Part#:%.*s\n",
+			   QSFP_PN_LEN, cd.partnum);
+	sofar += scnprintf(buf + sofar, len - sofar, "Rev:%.*s\n",
+			   QSFP_REV_LEN, cd.rev);
+	if (QSFP_IS_CU(cd.tech))
+		sofar += scnprintf(buf + sofar, len - sofar, "Atten:%d, %d\n",
+				   QSFP_ATTEN_SDR(cd.atten),
+				   QSFP_ATTEN_DDR(cd.atten));
+	sofar += scnprintf(buf + sofar, len - sofar, "Serial:%.*s\n",
+			   QSFP_SN_LEN, cd.serial);
+	sofar += scnprintf(buf + sofar, len - sofar, "Date:%.*s\n",
+			   QSFP_DATE_LEN, cd.date);
+	sofar += scnprintf(buf + sofar, len - sofar, "Lot:%.*s\n",
+			   QSFP_LOT_LEN, cd.date);
+
+	while (bidx < QSFP_DEFAULT_HDR_CNT) {
+		int iidx;
+		ret = qsfp_read(ppd, bidx, bin_buff, QSFP_DUMP_CHUNK);
+		if (ret < 0)
+			goto bail;
+		for (iidx = 0; iidx < ret; ++iidx) {
+			sofar += scnprintf(buf + sofar, len-sofar, " %02X",
+				bin_buff[iidx]);
+		}
+		sofar += scnprintf(buf + sofar, len - sofar, "\n");
+		bidx += QSFP_DUMP_CHUNK;
+	}
+	ret = sofar;
+bail:
+	return ret;
+}
diff --git a/drivers/infiniband/hw/qib/qib_qsfp.h b/drivers/infiniband/hw/qib/qib_qsfp.h
new file mode 100644
index 000000000000..19b527bafd57
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_qsfp.h
@@ -0,0 +1,184 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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.
+ */
+/* QSFP support common definitions, for ib_qib driver */
+
+#define QSFP_DEV 0xA0
+#define QSFP_PWR_LAG_MSEC 2000
+
+/*
+ * Below are masks for various QSFP signals, for Port 1.
+ * Port2 equivalents are shifted by QSFP_GPIO_PORT2_SHIFT.
+ * _N means asserted low
+ */
+#define QSFP_GPIO_MOD_SEL_N (4)
+#define QSFP_GPIO_MOD_PRS_N (8)
+#define QSFP_GPIO_INT_N (0x10)
+#define QSFP_GPIO_MOD_RST_N (0x20)
+#define QSFP_GPIO_LP_MODE (0x40)
+#define QSFP_GPIO_PORT2_SHIFT 5
+
+#define QSFP_PAGESIZE 128
+/* Defined fields that QLogic requires of qualified cables */
+/* Byte 0 is Identifier, not checked */
+/* Byte 1 is reserved "status MSB" */
+/* Byte 2 is "status LSB" We only care that D2 "Flat Mem" is set. */
+/*
+ * Rest of first 128 not used, although 127 is reserved for page select
+ * if module is not "Flat memory".
+ */
+/* Byte 128 is Identifier: must be 0x0c for QSFP, or 0x0d for QSFP+ */
+#define QSFP_MOD_ID_OFFS 128
+/*
+ * Byte 129 is "Extended Identifier". We only care about D7,D6: Power class
+ *  0:1.5W, 1:2.0W, 2:2.5W, 3:3.5W
+ */
+#define QSFP_MOD_PWR_OFFS 129
+/* Byte 130 is Connector type. Not QLogic req'd */
+/* Bytes 131..138 are Transceiver types, bit maps for various tech, none IB */
+/* Byte 139 is encoding. code 0x01 is 8b10b. Not QLogic req'd */
+/* byte 140 is nominal bit-rate, in units of 100Mbits/sec Not QLogic req'd */
+/* Byte 141 is Extended Rate Select. Not QLogic req'd */
+/* Bytes 142..145 are lengths for various fiber types. Not QLogic req'd */
+/* Byte 146 is length for Copper. Units of 1 meter */
+#define QSFP_MOD_LEN_OFFS 146
+/*
+ * Byte 147 is Device technology. D0..3 not Qlogc req'd
+ * D4..7 select from 15 choices, translated by table:
+ */
+#define QSFP_MOD_TECH_OFFS 147
+extern const char *const qib_qsfp_devtech[16];
+/* Active Equalization includes fiber, copper full EQ, and copper near Eq */
+#define QSFP_IS_ACTIVE(tech) ((0xA2FF >> ((tech) >> 4)) & 1)
+/* Attenuation should be valid for copper other than full/near Eq */
+#define QSFP_HAS_ATTEN(tech) ((0x4D00 >> ((tech) >> 4)) & 1)
+/* Length is only valid if technology is "copper" */
+#define QSFP_IS_CU(tech) ((0xED00 >> ((tech) >> 4)) & 1)
+#define QSFP_TECH_1490 9
+
+#define QSFP_OUI(oui) (((unsigned)oui[0] << 16) | ((unsigned)oui[1] << 8) | \
+			oui[2])
+#define QSFP_OUI_AMPHENOL 0x415048
+#define QSFP_OUI_FINISAR  0x009065
+#define QSFP_OUI_GORE     0x002177
+
+/* Bytes 148..163 are Vendor Name, Left-justified Blank-filled */
+#define QSFP_VEND_OFFS 148
+#define QSFP_VEND_LEN 16
+/* Byte 164 is IB Extended tranceiver codes Bits D0..3 are SDR,DDR,QDR,EDR */
+#define QSFP_IBXCV_OFFS 164
+/* Bytes 165..167 are Vendor OUI number */
+#define QSFP_VOUI_OFFS 165
+#define QSFP_VOUI_LEN 3
+/* Bytes 168..183 are Vendor Part Number, string */
+#define QSFP_PN_OFFS 168
+#define QSFP_PN_LEN 16
+/* Bytes 184,185 are Vendor Rev. Left Justified, Blank-filled */
+#define QSFP_REV_OFFS 184
+#define QSFP_REV_LEN 2
+/*
+ * Bytes 186,187 are Wavelength, if Optical. Not Qlogic req'd
+ *  If copper, they are attenuation in dB:
+ * Byte 186 is at 2.5Gb/sec (SDR), Byte 187 at 5.0Gb/sec (DDR)
+ */
+#define QSFP_ATTEN_OFFS 186
+#define QSFP_ATTEN_LEN 2
+/* Bytes 188,189 are Wavelength tolerance, not QLogic req'd */
+/* Byte 190 is Max Case Temp. Not QLogic req'd */
+/* Byte 191 is LSB of sum of bytes 128..190. Not QLogic req'd */
+#define QSFP_CC_OFFS 191
+/* Bytes 192..195 are Options implemented in qsfp. Not Qlogic req'd */
+/* Bytes 196..211 are Serial Number, String */
+#define QSFP_SN_OFFS 196
+#define QSFP_SN_LEN 16
+/* Bytes 212..219 are date-code YYMMDD (MM==1 for Jan) */
+#define QSFP_DATE_OFFS 212
+#define QSFP_DATE_LEN 6
+/* Bytes 218,219 are optional lot-code, string */
+#define QSFP_LOT_OFFS 218
+#define QSFP_LOT_LEN 2
+/* Bytes 220, 221 indicate monitoring options, Not QLogic req'd */
+/* Byte 223 is LSB of sum of bytes 192..222 */
+#define QSFP_CC_EXT_OFFS 223
+
+/*
+ * struct qib_qsfp_data encapsulates state of QSFP device for one port.
+ * it will be part of port-chip-specific data if a board supports QSFP.
+ *
+ * Since multiple board-types use QSFP, and their pport_data structs
+ * differ (in the chip-specific section), we need a pointer to its head.
+ *
+ * Avoiding premature optimization, we will have one work_struct per port,
+ * and let the (increasingly inaccurately named) eep_lock arbitrate
+ * access to common resources.
+ *
+ */
+
+/*
+ * Hold the parts of the onboard EEPROM that we care about, so we aren't
+ * coonstantly bit-boffing
+ */
+struct qib_qsfp_cache {
+	u8 id;	/* must be 0x0C or 0x0D; 0 indicates invalid EEPROM read */
+	u8 pwr; /* in D6,7 */
+	u8 len;	/* in meters, Cu only */
+	u8 tech;
+	char vendor[QSFP_VEND_LEN];
+	u8 xt_xcv; /* Ext. tranceiver codes, 4 lsbs are IB speed supported */
+	u8 oui[QSFP_VOUI_LEN];
+	u8 partnum[QSFP_PN_LEN];
+	u8 rev[QSFP_REV_LEN];
+	u8 atten[QSFP_ATTEN_LEN];
+	u8 cks1;	/* Checksum of bytes 128..190 */
+	u8 serial[QSFP_SN_LEN];
+	u8 date[QSFP_DATE_LEN];
+	u8 lot[QSFP_LOT_LEN];
+	u8 cks2;	/* Checsum of bytes 192..222 */
+};
+
+#define QSFP_PWR(pbyte) (((pbyte) >> 6) & 3)
+#define QSFP_ATTEN_SDR(attenarray) (attenarray[0])
+#define QSFP_ATTEN_DDR(attenarray) (attenarray[1])
+
+struct qib_qsfp_data {
+	/* Helps to find our way */
+	struct qib_pportdata *ppd;
+	struct work_struct work;
+	struct qib_qsfp_cache cache;
+	u64 t_insert;
+};
+
+extern int qib_refresh_qsfp_cache(struct qib_pportdata *ppd,
+				  struct qib_qsfp_cache *cp);
+extern void qib_qsfp_init(struct qib_qsfp_data *qd,
+			  void (*fevent)(struct work_struct *));
+extern void qib_qsfp_deinit(struct qib_qsfp_data *qd);
diff --git a/drivers/infiniband/hw/qib/qib_rc.c b/drivers/infiniband/hw/qib/qib_rc.c
new file mode 100644
index 000000000000..40c0a373719c
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_rc.c
@@ -0,0 +1,2288 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/io.h>
+
+#include "qib.h"
+
+/* cut down ridiculously long IB macro names */
+#define OP(x) IB_OPCODE_RC_##x
+
+static void rc_timeout(unsigned long arg);
+
+static u32 restart_sge(struct qib_sge_state *ss, struct qib_swqe *wqe,
+		       u32 psn, u32 pmtu)
+{
+	u32 len;
+
+	len = ((psn - wqe->psn) & QIB_PSN_MASK) * pmtu;
+	ss->sge = wqe->sg_list[0];
+	ss->sg_list = wqe->sg_list + 1;
+	ss->num_sge = wqe->wr.num_sge;
+	ss->total_len = wqe->length;
+	qib_skip_sge(ss, len, 0);
+	return wqe->length - len;
+}
+
+static void start_timer(struct qib_qp *qp)
+{
+	qp->s_flags |= QIB_S_TIMER;
+	qp->s_timer.function = rc_timeout;
+	/* 4.096 usec. * (1 << qp->timeout) */
+	qp->s_timer.expires = jiffies +
+		usecs_to_jiffies((4096UL * (1UL << qp->timeout)) / 1000UL);
+	add_timer(&qp->s_timer);
+}
+
+/**
+ * qib_make_rc_ack - construct a response packet (ACK, NAK, or RDMA read)
+ * @dev: the device for this QP
+ * @qp: a pointer to the QP
+ * @ohdr: a pointer to the IB header being constructed
+ * @pmtu: the path MTU
+ *
+ * Return 1 if constructed; otherwise, return 0.
+ * Note that we are in the responder's side of the QP context.
+ * Note the QP s_lock must be held.
+ */
+static int qib_make_rc_ack(struct qib_ibdev *dev, struct qib_qp *qp,
+			   struct qib_other_headers *ohdr, u32 pmtu)
+{
+	struct qib_ack_entry *e;
+	u32 hwords;
+	u32 len;
+	u32 bth0;
+	u32 bth2;
+
+	/* Don't send an ACK if we aren't supposed to. */
+	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK))
+		goto bail;
+
+	/* header size in 32-bit words LRH+BTH = (8+12)/4. */
+	hwords = 5;
+
+	switch (qp->s_ack_state) {
+	case OP(RDMA_READ_RESPONSE_LAST):
+	case OP(RDMA_READ_RESPONSE_ONLY):
+		e = &qp->s_ack_queue[qp->s_tail_ack_queue];
+		if (e->rdma_sge.mr) {
+			atomic_dec(&e->rdma_sge.mr->refcount);
+			e->rdma_sge.mr = NULL;
+		}
+		/* FALLTHROUGH */
+	case OP(ATOMIC_ACKNOWLEDGE):
+		/*
+		 * We can increment the tail pointer now that the last
+		 * response has been sent instead of only being
+		 * constructed.
+		 */
+		if (++qp->s_tail_ack_queue > QIB_MAX_RDMA_ATOMIC)
+			qp->s_tail_ack_queue = 0;
+		/* FALLTHROUGH */
+	case OP(SEND_ONLY):
+	case OP(ACKNOWLEDGE):
+		/* Check for no next entry in the queue. */
+		if (qp->r_head_ack_queue == qp->s_tail_ack_queue) {
+			if (qp->s_flags & QIB_S_ACK_PENDING)
+				goto normal;
+			goto bail;
+		}
+
+		e = &qp->s_ack_queue[qp->s_tail_ack_queue];
+		if (e->opcode == OP(RDMA_READ_REQUEST)) {
+			/*
+			 * If a RDMA read response is being resent and
+			 * we haven't seen the duplicate request yet,
+			 * then stop sending the remaining responses the
+			 * responder has seen until the requester resends it.
+			 */
+			len = e->rdma_sge.sge_length;
+			if (len && !e->rdma_sge.mr) {
+				qp->s_tail_ack_queue = qp->r_head_ack_queue;
+				goto bail;
+			}
+			/* Copy SGE state in case we need to resend */
+			qp->s_rdma_mr = e->rdma_sge.mr;
+			if (qp->s_rdma_mr)
+				atomic_inc(&qp->s_rdma_mr->refcount);
+			qp->s_ack_rdma_sge.sge = e->rdma_sge;
+			qp->s_ack_rdma_sge.num_sge = 1;
+			qp->s_cur_sge = &qp->s_ack_rdma_sge;
+			if (len > pmtu) {
+				len = pmtu;
+				qp->s_ack_state = OP(RDMA_READ_RESPONSE_FIRST);
+			} else {
+				qp->s_ack_state = OP(RDMA_READ_RESPONSE_ONLY);
+				e->sent = 1;
+			}
+			ohdr->u.aeth = qib_compute_aeth(qp);
+			hwords++;
+			qp->s_ack_rdma_psn = e->psn;
+			bth2 = qp->s_ack_rdma_psn++ & QIB_PSN_MASK;
+		} else {
+			/* COMPARE_SWAP or FETCH_ADD */
+			qp->s_cur_sge = NULL;
+			len = 0;
+			qp->s_ack_state = OP(ATOMIC_ACKNOWLEDGE);
+			ohdr->u.at.aeth = qib_compute_aeth(qp);
+			ohdr->u.at.atomic_ack_eth[0] =
+				cpu_to_be32(e->atomic_data >> 32);
+			ohdr->u.at.atomic_ack_eth[1] =
+				cpu_to_be32(e->atomic_data);
+			hwords += sizeof(ohdr->u.at) / sizeof(u32);
+			bth2 = e->psn & QIB_PSN_MASK;
+			e->sent = 1;
+		}
+		bth0 = qp->s_ack_state << 24;
+		break;
+
+	case OP(RDMA_READ_RESPONSE_FIRST):
+		qp->s_ack_state = OP(RDMA_READ_RESPONSE_MIDDLE);
+		/* FALLTHROUGH */
+	case OP(RDMA_READ_RESPONSE_MIDDLE):
+		qp->s_cur_sge = &qp->s_ack_rdma_sge;
+		qp->s_rdma_mr = qp->s_ack_rdma_sge.sge.mr;
+		if (qp->s_rdma_mr)
+			atomic_inc(&qp->s_rdma_mr->refcount);
+		len = qp->s_ack_rdma_sge.sge.sge_length;
+		if (len > pmtu)
+			len = pmtu;
+		else {
+			ohdr->u.aeth = qib_compute_aeth(qp);
+			hwords++;
+			qp->s_ack_state = OP(RDMA_READ_RESPONSE_LAST);
+			e = &qp->s_ack_queue[qp->s_tail_ack_queue];
+			e->sent = 1;
+		}
+		bth0 = qp->s_ack_state << 24;
+		bth2 = qp->s_ack_rdma_psn++ & QIB_PSN_MASK;
+		break;
+
+	default:
+normal:
+		/*
+		 * Send a regular ACK.
+		 * Set the s_ack_state so we wait until after sending
+		 * the ACK before setting s_ack_state to ACKNOWLEDGE
+		 * (see above).
+		 */
+		qp->s_ack_state = OP(SEND_ONLY);
+		qp->s_flags &= ~QIB_S_ACK_PENDING;
+		qp->s_cur_sge = NULL;
+		if (qp->s_nak_state)
+			ohdr->u.aeth =
+				cpu_to_be32((qp->r_msn & QIB_MSN_MASK) |
+					    (qp->s_nak_state <<
+					     QIB_AETH_CREDIT_SHIFT));
+		else
+			ohdr->u.aeth = qib_compute_aeth(qp);
+		hwords++;
+		len = 0;
+		bth0 = OP(ACKNOWLEDGE) << 24;
+		bth2 = qp->s_ack_psn & QIB_PSN_MASK;
+	}
+	qp->s_rdma_ack_cnt++;
+	qp->s_hdrwords = hwords;
+	qp->s_cur_size = len;
+	qib_make_ruc_header(qp, ohdr, bth0, bth2);
+	return 1;
+
+bail:
+	qp->s_ack_state = OP(ACKNOWLEDGE);
+	qp->s_flags &= ~(QIB_S_RESP_PENDING | QIB_S_ACK_PENDING);
+	return 0;
+}
+
+/**
+ * qib_make_rc_req - construct a request packet (SEND, RDMA r/w, ATOMIC)
+ * @qp: a pointer to the QP
+ *
+ * Return 1 if constructed; otherwise, return 0.
+ */
+int qib_make_rc_req(struct qib_qp *qp)
+{
+	struct qib_ibdev *dev = to_idev(qp->ibqp.device);
+	struct qib_other_headers *ohdr;
+	struct qib_sge_state *ss;
+	struct qib_swqe *wqe;
+	u32 hwords;
+	u32 len;
+	u32 bth0;
+	u32 bth2;
+	u32 pmtu = ib_mtu_enum_to_int(qp->path_mtu);
+	char newreq;
+	unsigned long flags;
+	int ret = 0;
+	int delta;
+
+	ohdr = &qp->s_hdr.u.oth;
+	if (qp->remote_ah_attr.ah_flags & IB_AH_GRH)
+		ohdr = &qp->s_hdr.u.l.oth;
+
+	/*
+	 * The lock is needed to synchronize between the sending tasklet,
+	 * the receive interrupt handler, and timeout resends.
+	 */
+	spin_lock_irqsave(&qp->s_lock, flags);
+
+	/* Sending responses has higher priority over sending requests. */
+	if ((qp->s_flags & QIB_S_RESP_PENDING) &&
+	    qib_make_rc_ack(dev, qp, ohdr, pmtu))
+		goto done;
+
+	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_SEND_OK)) {
+		if (!(ib_qib_state_ops[qp->state] & QIB_FLUSH_SEND))
+			goto bail;
+		/* We are in the error state, flush the work request. */
+		if (qp->s_last == qp->s_head)
+			goto bail;
+		/* If DMAs are in progress, we can't flush immediately. */
+		if (atomic_read(&qp->s_dma_busy)) {
+			qp->s_flags |= QIB_S_WAIT_DMA;
+			goto bail;
+		}
+		wqe = get_swqe_ptr(qp, qp->s_last);
+		while (qp->s_last != qp->s_acked) {
+			qib_send_complete(qp, wqe, IB_WC_SUCCESS);
+			if (++qp->s_last >= qp->s_size)
+				qp->s_last = 0;
+			wqe = get_swqe_ptr(qp, qp->s_last);
+		}
+		qib_send_complete(qp, wqe, IB_WC_WR_FLUSH_ERR);
+		goto done;
+	}
+
+	if (qp->s_flags & (QIB_S_WAIT_RNR | QIB_S_WAIT_ACK))
+		goto bail;
+
+	if (qib_cmp24(qp->s_psn, qp->s_sending_hpsn) <= 0) {
+		if (qib_cmp24(qp->s_sending_psn, qp->s_sending_hpsn) <= 0) {
+			qp->s_flags |= QIB_S_WAIT_PSN;
+			goto bail;
+		}
+		qp->s_sending_psn = qp->s_psn;
+		qp->s_sending_hpsn = qp->s_psn - 1;
+	}
+
+	/* header size in 32-bit words LRH+BTH = (8+12)/4. */
+	hwords = 5;
+	bth0 = 0;
+
+	/* Send a request. */
+	wqe = get_swqe_ptr(qp, qp->s_cur);
+	switch (qp->s_state) {
+	default:
+		if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_NEXT_SEND_OK))
+			goto bail;
+		/*
+		 * Resend an old request or start a new one.
+		 *
+		 * We keep track of the current SWQE so that
+		 * we don't reset the "furthest progress" state
+		 * if we need to back up.
+		 */
+		newreq = 0;
+		if (qp->s_cur == qp->s_tail) {
+			/* Check if send work queue is empty. */
+			if (qp->s_tail == qp->s_head)
+				goto bail;
+			/*
+			 * If a fence is requested, wait for previous
+			 * RDMA read and atomic operations to finish.
+			 */
+			if ((wqe->wr.send_flags & IB_SEND_FENCE) &&
+			    qp->s_num_rd_atomic) {
+				qp->s_flags |= QIB_S_WAIT_FENCE;
+				goto bail;
+			}
+			wqe->psn = qp->s_next_psn;
+			newreq = 1;
+		}
+		/*
+		 * Note that we have to be careful not to modify the
+		 * original work request since we may need to resend
+		 * it.
+		 */
+		len = wqe->length;
+		ss = &qp->s_sge;
+		bth2 = qp->s_psn & QIB_PSN_MASK;
+		switch (wqe->wr.opcode) {
+		case IB_WR_SEND:
+		case IB_WR_SEND_WITH_IMM:
+			/* If no credit, return. */
+			if (!(qp->s_flags & QIB_S_UNLIMITED_CREDIT) &&
+			    qib_cmp24(wqe->ssn, qp->s_lsn + 1) > 0) {
+				qp->s_flags |= QIB_S_WAIT_SSN_CREDIT;
+				goto bail;
+			}
+			wqe->lpsn = wqe->psn;
+			if (len > pmtu) {
+				wqe->lpsn += (len - 1) / pmtu;
+				qp->s_state = OP(SEND_FIRST);
+				len = pmtu;
+				break;
+			}
+			if (wqe->wr.opcode == IB_WR_SEND)
+				qp->s_state = OP(SEND_ONLY);
+			else {
+				qp->s_state = OP(SEND_ONLY_WITH_IMMEDIATE);
+				/* Immediate data comes after the BTH */
+				ohdr->u.imm_data = wqe->wr.ex.imm_data;
+				hwords += 1;
+			}
+			if (wqe->wr.send_flags & IB_SEND_SOLICITED)
+				bth0 |= IB_BTH_SOLICITED;
+			bth2 |= IB_BTH_REQ_ACK;
+			if (++qp->s_cur == qp->s_size)
+				qp->s_cur = 0;
+			break;
+
+		case IB_WR_RDMA_WRITE:
+			if (newreq && !(qp->s_flags & QIB_S_UNLIMITED_CREDIT))
+				qp->s_lsn++;
+			/* FALLTHROUGH */
+		case IB_WR_RDMA_WRITE_WITH_IMM:
+			/* If no credit, return. */
+			if (!(qp->s_flags & QIB_S_UNLIMITED_CREDIT) &&
+			    qib_cmp24(wqe->ssn, qp->s_lsn + 1) > 0) {
+				qp->s_flags |= QIB_S_WAIT_SSN_CREDIT;
+				goto bail;
+			}
+			ohdr->u.rc.reth.vaddr =
+				cpu_to_be64(wqe->wr.wr.rdma.remote_addr);
+			ohdr->u.rc.reth.rkey =
+				cpu_to_be32(wqe->wr.wr.rdma.rkey);
+			ohdr->u.rc.reth.length = cpu_to_be32(len);
+			hwords += sizeof(struct ib_reth) / sizeof(u32);
+			wqe->lpsn = wqe->psn;
+			if (len > pmtu) {
+				wqe->lpsn += (len - 1) / pmtu;
+				qp->s_state = OP(RDMA_WRITE_FIRST);
+				len = pmtu;
+				break;
+			}
+			if (wqe->wr.opcode == IB_WR_RDMA_WRITE)
+				qp->s_state = OP(RDMA_WRITE_ONLY);
+			else {
+				qp->s_state =
+					OP(RDMA_WRITE_ONLY_WITH_IMMEDIATE);
+				/* Immediate data comes after RETH */
+				ohdr->u.rc.imm_data = wqe->wr.ex.imm_data;
+				hwords += 1;
+				if (wqe->wr.send_flags & IB_SEND_SOLICITED)
+					bth0 |= IB_BTH_SOLICITED;
+			}
+			bth2 |= IB_BTH_REQ_ACK;
+			if (++qp->s_cur == qp->s_size)
+				qp->s_cur = 0;
+			break;
+
+		case IB_WR_RDMA_READ:
+			/*
+			 * Don't allow more operations to be started
+			 * than the QP limits allow.
+			 */
+			if (newreq) {
+				if (qp->s_num_rd_atomic >=
+				    qp->s_max_rd_atomic) {
+					qp->s_flags |= QIB_S_WAIT_RDMAR;
+					goto bail;
+				}
+				qp->s_num_rd_atomic++;
+				if (!(qp->s_flags & QIB_S_UNLIMITED_CREDIT))
+					qp->s_lsn++;
+				/*
+				 * Adjust s_next_psn to count the
+				 * expected number of responses.
+				 */
+				if (len > pmtu)
+					qp->s_next_psn += (len - 1) / pmtu;
+				wqe->lpsn = qp->s_next_psn++;
+			}
+			ohdr->u.rc.reth.vaddr =
+				cpu_to_be64(wqe->wr.wr.rdma.remote_addr);
+			ohdr->u.rc.reth.rkey =
+				cpu_to_be32(wqe->wr.wr.rdma.rkey);
+			ohdr->u.rc.reth.length = cpu_to_be32(len);
+			qp->s_state = OP(RDMA_READ_REQUEST);
+			hwords += sizeof(ohdr->u.rc.reth) / sizeof(u32);
+			ss = NULL;
+			len = 0;
+			bth2 |= IB_BTH_REQ_ACK;
+			if (++qp->s_cur == qp->s_size)
+				qp->s_cur = 0;
+			break;
+
+		case IB_WR_ATOMIC_CMP_AND_SWP:
+		case IB_WR_ATOMIC_FETCH_AND_ADD:
+			/*
+			 * Don't allow more operations to be started
+			 * than the QP limits allow.
+			 */
+			if (newreq) {
+				if (qp->s_num_rd_atomic >=
+				    qp->s_max_rd_atomic) {
+					qp->s_flags |= QIB_S_WAIT_RDMAR;
+					goto bail;
+				}
+				qp->s_num_rd_atomic++;
+				if (!(qp->s_flags & QIB_S_UNLIMITED_CREDIT))
+					qp->s_lsn++;
+				wqe->lpsn = wqe->psn;
+			}
+			if (wqe->wr.opcode == IB_WR_ATOMIC_CMP_AND_SWP) {
+				qp->s_state = OP(COMPARE_SWAP);
+				ohdr->u.atomic_eth.swap_data = cpu_to_be64(
+					wqe->wr.wr.atomic.swap);
+				ohdr->u.atomic_eth.compare_data = cpu_to_be64(
+					wqe->wr.wr.atomic.compare_add);
+			} else {
+				qp->s_state = OP(FETCH_ADD);
+				ohdr->u.atomic_eth.swap_data = cpu_to_be64(
+					wqe->wr.wr.atomic.compare_add);
+				ohdr->u.atomic_eth.compare_data = 0;
+			}
+			ohdr->u.atomic_eth.vaddr[0] = cpu_to_be32(
+				wqe->wr.wr.atomic.remote_addr >> 32);
+			ohdr->u.atomic_eth.vaddr[1] = cpu_to_be32(
+				wqe->wr.wr.atomic.remote_addr);
+			ohdr->u.atomic_eth.rkey = cpu_to_be32(
+				wqe->wr.wr.atomic.rkey);
+			hwords += sizeof(struct ib_atomic_eth) / sizeof(u32);
+			ss = NULL;
+			len = 0;
+			bth2 |= IB_BTH_REQ_ACK;
+			if (++qp->s_cur == qp->s_size)
+				qp->s_cur = 0;
+			break;
+
+		default:
+			goto bail;
+		}
+		qp->s_sge.sge = wqe->sg_list[0];
+		qp->s_sge.sg_list = wqe->sg_list + 1;
+		qp->s_sge.num_sge = wqe->wr.num_sge;
+		qp->s_sge.total_len = wqe->length;
+		qp->s_len = wqe->length;
+		if (newreq) {
+			qp->s_tail++;
+			if (qp->s_tail >= qp->s_size)
+				qp->s_tail = 0;
+		}
+		if (wqe->wr.opcode == IB_WR_RDMA_READ)
+			qp->s_psn = wqe->lpsn + 1;
+		else {
+			qp->s_psn++;
+			if (qib_cmp24(qp->s_psn, qp->s_next_psn) > 0)
+				qp->s_next_psn = qp->s_psn;
+		}
+		break;
+
+	case OP(RDMA_READ_RESPONSE_FIRST):
+		/*
+		 * qp->s_state is normally set to the opcode of the
+		 * last packet constructed for new requests and therefore
+		 * is never set to RDMA read response.
+		 * RDMA_READ_RESPONSE_FIRST is used by the ACK processing
+		 * thread to indicate a SEND needs to be restarted from an
+		 * earlier PSN without interferring with the sending thread.
+		 * See qib_restart_rc().
+		 */
+		qp->s_len = restart_sge(&qp->s_sge, wqe, qp->s_psn, pmtu);
+		/* FALLTHROUGH */
+	case OP(SEND_FIRST):
+		qp->s_state = OP(SEND_MIDDLE);
+		/* FALLTHROUGH */
+	case OP(SEND_MIDDLE):
+		bth2 = qp->s_psn++ & QIB_PSN_MASK;
+		if (qib_cmp24(qp->s_psn, qp->s_next_psn) > 0)
+			qp->s_next_psn = qp->s_psn;
+		ss = &qp->s_sge;
+		len = qp->s_len;
+		if (len > pmtu) {
+			len = pmtu;
+			break;
+		}
+		if (wqe->wr.opcode == IB_WR_SEND)
+			qp->s_state = OP(SEND_LAST);
+		else {
+			qp->s_state = OP(SEND_LAST_WITH_IMMEDIATE);
+			/* Immediate data comes after the BTH */
+			ohdr->u.imm_data = wqe->wr.ex.imm_data;
+			hwords += 1;
+		}
+		if (wqe->wr.send_flags & IB_SEND_SOLICITED)
+			bth0 |= IB_BTH_SOLICITED;
+		bth2 |= IB_BTH_REQ_ACK;
+		qp->s_cur++;
+		if (qp->s_cur >= qp->s_size)
+			qp->s_cur = 0;
+		break;
+
+	case OP(RDMA_READ_RESPONSE_LAST):
+		/*
+		 * qp->s_state is normally set to the opcode of the
+		 * last packet constructed for new requests and therefore
+		 * is never set to RDMA read response.
+		 * RDMA_READ_RESPONSE_LAST is used by the ACK processing
+		 * thread to indicate a RDMA write needs to be restarted from
+		 * an earlier PSN without interferring with the sending thread.
+		 * See qib_restart_rc().
+		 */
+		qp->s_len = restart_sge(&qp->s_sge, wqe, qp->s_psn, pmtu);
+		/* FALLTHROUGH */
+	case OP(RDMA_WRITE_FIRST):
+		qp->s_state = OP(RDMA_WRITE_MIDDLE);
+		/* FALLTHROUGH */
+	case OP(RDMA_WRITE_MIDDLE):
+		bth2 = qp->s_psn++ & QIB_PSN_MASK;
+		if (qib_cmp24(qp->s_psn, qp->s_next_psn) > 0)
+			qp->s_next_psn = qp->s_psn;
+		ss = &qp->s_sge;
+		len = qp->s_len;
+		if (len > pmtu) {
+			len = pmtu;
+			break;
+		}
+		if (wqe->wr.opcode == IB_WR_RDMA_WRITE)
+			qp->s_state = OP(RDMA_WRITE_LAST);
+		else {
+			qp->s_state = OP(RDMA_WRITE_LAST_WITH_IMMEDIATE);
+			/* Immediate data comes after the BTH */
+			ohdr->u.imm_data = wqe->wr.ex.imm_data;
+			hwords += 1;
+			if (wqe->wr.send_flags & IB_SEND_SOLICITED)
+				bth0 |= IB_BTH_SOLICITED;
+		}
+		bth2 |= IB_BTH_REQ_ACK;
+		qp->s_cur++;
+		if (qp->s_cur >= qp->s_size)
+			qp->s_cur = 0;
+		break;
+
+	case OP(RDMA_READ_RESPONSE_MIDDLE):
+		/*
+		 * qp->s_state is normally set to the opcode of the
+		 * last packet constructed for new requests and therefore
+		 * is never set to RDMA read response.
+		 * RDMA_READ_RESPONSE_MIDDLE is used by the ACK processing
+		 * thread to indicate a RDMA read needs to be restarted from
+		 * an earlier PSN without interferring with the sending thread.
+		 * See qib_restart_rc().
+		 */
+		len = ((qp->s_psn - wqe->psn) & QIB_PSN_MASK) * pmtu;
+		ohdr->u.rc.reth.vaddr =
+			cpu_to_be64(wqe->wr.wr.rdma.remote_addr + len);
+		ohdr->u.rc.reth.rkey =
+			cpu_to_be32(wqe->wr.wr.rdma.rkey);
+		ohdr->u.rc.reth.length = cpu_to_be32(wqe->length - len);
+		qp->s_state = OP(RDMA_READ_REQUEST);
+		hwords += sizeof(ohdr->u.rc.reth) / sizeof(u32);
+		bth2 = (qp->s_psn & QIB_PSN_MASK) | IB_BTH_REQ_ACK;
+		qp->s_psn = wqe->lpsn + 1;
+		ss = NULL;
+		len = 0;
+		qp->s_cur++;
+		if (qp->s_cur == qp->s_size)
+			qp->s_cur = 0;
+		break;
+	}
+	qp->s_sending_hpsn = bth2;
+	delta = (((int) bth2 - (int) wqe->psn) << 8) >> 8;
+	if (delta && delta % QIB_PSN_CREDIT == 0)
+		bth2 |= IB_BTH_REQ_ACK;
+	if (qp->s_flags & QIB_S_SEND_ONE) {
+		qp->s_flags &= ~QIB_S_SEND_ONE;
+		qp->s_flags |= QIB_S_WAIT_ACK;
+		bth2 |= IB_BTH_REQ_ACK;
+	}
+	qp->s_len -= len;
+	qp->s_hdrwords = hwords;
+	qp->s_cur_sge = ss;
+	qp->s_cur_size = len;
+	qib_make_ruc_header(qp, ohdr, bth0 | (qp->s_state << 24), bth2);
+done:
+	ret = 1;
+	goto unlock;
+
+bail:
+	qp->s_flags &= ~QIB_S_BUSY;
+unlock:
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+	return ret;
+}
+
+/**
+ * qib_send_rc_ack - Construct an ACK packet and send it
+ * @qp: a pointer to the QP
+ *
+ * This is called from qib_rc_rcv() and qib_kreceive().
+ * Note that RDMA reads and atomics are handled in the
+ * send side QP state and tasklet.
+ */
+void qib_send_rc_ack(struct qib_qp *qp)
+{
+	struct qib_devdata *dd = dd_from_ibdev(qp->ibqp.device);
+	struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	u64 pbc;
+	u16 lrh0;
+	u32 bth0;
+	u32 hwords;
+	u32 pbufn;
+	u32 __iomem *piobuf;
+	struct qib_ib_header hdr;
+	struct qib_other_headers *ohdr;
+	u32 control;
+	unsigned long flags;
+
+	spin_lock_irqsave(&qp->s_lock, flags);
+
+	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK))
+		goto unlock;
+
+	/* Don't send ACK or NAK if a RDMA read or atomic is pending. */
+	if ((qp->s_flags & QIB_S_RESP_PENDING) || qp->s_rdma_ack_cnt)
+		goto queue_ack;
+
+	/* Construct the header with s_lock held so APM doesn't change it. */
+	ohdr = &hdr.u.oth;
+	lrh0 = QIB_LRH_BTH;
+	/* header size in 32-bit words LRH+BTH+AETH = (8+12+4)/4. */
+	hwords = 6;
+	if (unlikely(qp->remote_ah_attr.ah_flags & IB_AH_GRH)) {
+		hwords += qib_make_grh(ibp, &hdr.u.l.grh,
+				       &qp->remote_ah_attr.grh, hwords, 0);
+		ohdr = &hdr.u.l.oth;
+		lrh0 = QIB_LRH_GRH;
+	}
+	/* read pkey_index w/o lock (its atomic) */
+	bth0 = qib_get_pkey(ibp, qp->s_pkey_index) | (OP(ACKNOWLEDGE) << 24);
+	if (qp->s_mig_state == IB_MIG_MIGRATED)
+		bth0 |= IB_BTH_MIG_REQ;
+	if (qp->r_nak_state)
+		ohdr->u.aeth = cpu_to_be32((qp->r_msn & QIB_MSN_MASK) |
+					    (qp->r_nak_state <<
+					     QIB_AETH_CREDIT_SHIFT));
+	else
+		ohdr->u.aeth = qib_compute_aeth(qp);
+	lrh0 |= ibp->sl_to_vl[qp->remote_ah_attr.sl] << 12 |
+		qp->remote_ah_attr.sl << 4;
+	hdr.lrh[0] = cpu_to_be16(lrh0);
+	hdr.lrh[1] = cpu_to_be16(qp->remote_ah_attr.dlid);
+	hdr.lrh[2] = cpu_to_be16(hwords + SIZE_OF_CRC);
+	hdr.lrh[3] = cpu_to_be16(ppd->lid | qp->remote_ah_attr.src_path_bits);
+	ohdr->bth[0] = cpu_to_be32(bth0);
+	ohdr->bth[1] = cpu_to_be32(qp->remote_qpn);
+	ohdr->bth[2] = cpu_to_be32(qp->r_ack_psn & QIB_PSN_MASK);
+
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+
+	/* Don't try to send ACKs if the link isn't ACTIVE */
+	if (!(ppd->lflags & QIBL_LINKACTIVE))
+		goto done;
+
+	control = dd->f_setpbc_control(ppd, hwords + SIZE_OF_CRC,
+				       qp->s_srate, lrh0 >> 12);
+	/* length is + 1 for the control dword */
+	pbc = ((u64) control << 32) | (hwords + 1);
+
+	piobuf = dd->f_getsendbuf(ppd, pbc, &pbufn);
+	if (!piobuf) {
+		/*
+		 * We are out of PIO buffers at the moment.
+		 * Pass responsibility for sending the ACK to the
+		 * send tasklet so that when a PIO buffer becomes
+		 * available, the ACK is sent ahead of other outgoing
+		 * packets.
+		 */
+		spin_lock_irqsave(&qp->s_lock, flags);
+		goto queue_ack;
+	}
+
+	/*
+	 * Write the pbc.
+	 * We have to flush after the PBC for correctness
+	 * on some cpus or WC buffer can be written out of order.
+	 */
+	writeq(pbc, piobuf);
+
+	if (dd->flags & QIB_PIO_FLUSH_WC) {
+		u32 *hdrp = (u32 *) &hdr;
+
+		qib_flush_wc();
+		qib_pio_copy(piobuf + 2, hdrp, hwords - 1);
+		qib_flush_wc();
+		__raw_writel(hdrp[hwords - 1], piobuf + hwords + 1);
+	} else
+		qib_pio_copy(piobuf + 2, (u32 *) &hdr, hwords);
+
+	if (dd->flags & QIB_USE_SPCL_TRIG) {
+		u32 spcl_off = (pbufn >= dd->piobcnt2k) ? 2047 : 1023;
+
+		qib_flush_wc();
+		__raw_writel(0xaebecede, piobuf + spcl_off);
+	}
+
+	qib_flush_wc();
+	qib_sendbuf_done(dd, pbufn);
+
+	ibp->n_unicast_xmit++;
+	goto done;
+
+queue_ack:
+	if (ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK) {
+		ibp->n_rc_qacks++;
+		qp->s_flags |= QIB_S_ACK_PENDING | QIB_S_RESP_PENDING;
+		qp->s_nak_state = qp->r_nak_state;
+		qp->s_ack_psn = qp->r_ack_psn;
+
+		/* Schedule the send tasklet. */
+		qib_schedule_send(qp);
+	}
+unlock:
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+done:
+	return;
+}
+
+/**
+ * reset_psn - reset the QP state to send starting from PSN
+ * @qp: the QP
+ * @psn: the packet sequence number to restart at
+ *
+ * This is called from qib_rc_rcv() to process an incoming RC ACK
+ * for the given QP.
+ * Called at interrupt level with the QP s_lock held.
+ */
+static void reset_psn(struct qib_qp *qp, u32 psn)
+{
+	u32 n = qp->s_acked;
+	struct qib_swqe *wqe = get_swqe_ptr(qp, n);
+	u32 opcode;
+
+	qp->s_cur = n;
+
+	/*
+	 * If we are starting the request from the beginning,
+	 * let the normal send code handle initialization.
+	 */
+	if (qib_cmp24(psn, wqe->psn) <= 0) {
+		qp->s_state = OP(SEND_LAST);
+		goto done;
+	}
+
+	/* Find the work request opcode corresponding to the given PSN. */
+	opcode = wqe->wr.opcode;
+	for (;;) {
+		int diff;
+
+		if (++n == qp->s_size)
+			n = 0;
+		if (n == qp->s_tail)
+			break;
+		wqe = get_swqe_ptr(qp, n);
+		diff = qib_cmp24(psn, wqe->psn);
+		if (diff < 0)
+			break;
+		qp->s_cur = n;
+		/*
+		 * If we are starting the request from the beginning,
+		 * let the normal send code handle initialization.
+		 */
+		if (diff == 0) {
+			qp->s_state = OP(SEND_LAST);
+			goto done;
+		}
+		opcode = wqe->wr.opcode;
+	}
+
+	/*
+	 * Set the state to restart in the middle of a request.
+	 * Don't change the s_sge, s_cur_sge, or s_cur_size.
+	 * See qib_make_rc_req().
+	 */
+	switch (opcode) {
+	case IB_WR_SEND:
+	case IB_WR_SEND_WITH_IMM:
+		qp->s_state = OP(RDMA_READ_RESPONSE_FIRST);
+		break;
+
+	case IB_WR_RDMA_WRITE:
+	case IB_WR_RDMA_WRITE_WITH_IMM:
+		qp->s_state = OP(RDMA_READ_RESPONSE_LAST);
+		break;
+
+	case IB_WR_RDMA_READ:
+		qp->s_state = OP(RDMA_READ_RESPONSE_MIDDLE);
+		break;
+
+	default:
+		/*
+		 * This case shouldn't happen since its only
+		 * one PSN per req.
+		 */
+		qp->s_state = OP(SEND_LAST);
+	}
+done:
+	qp->s_psn = psn;
+	/*
+	 * Set QIB_S_WAIT_PSN as qib_rc_complete() may start the timer
+	 * asynchronously before the send tasklet can get scheduled.
+	 * Doing it in qib_make_rc_req() is too late.
+	 */
+	if ((qib_cmp24(qp->s_psn, qp->s_sending_hpsn) <= 0) &&
+	    (qib_cmp24(qp->s_sending_psn, qp->s_sending_hpsn) <= 0))
+		qp->s_flags |= QIB_S_WAIT_PSN;
+}
+
+/*
+ * Back up requester to resend the last un-ACKed request.
+ * The QP s_lock should be held and interrupts disabled.
+ */
+static void qib_restart_rc(struct qib_qp *qp, u32 psn, int wait)
+{
+	struct qib_swqe *wqe = get_swqe_ptr(qp, qp->s_acked);
+	struct qib_ibport *ibp;
+
+	if (qp->s_retry == 0) {
+		if (qp->s_mig_state == IB_MIG_ARMED) {
+			qib_migrate_qp(qp);
+			qp->s_retry = qp->s_retry_cnt;
+		} else if (qp->s_last == qp->s_acked) {
+			qib_send_complete(qp, wqe, IB_WC_RETRY_EXC_ERR);
+			qib_error_qp(qp, IB_WC_WR_FLUSH_ERR);
+			return;
+		} else /* XXX need to handle delayed completion */
+			return;
+	} else
+		qp->s_retry--;
+
+	ibp = to_iport(qp->ibqp.device, qp->port_num);
+	if (wqe->wr.opcode == IB_WR_RDMA_READ)
+		ibp->n_rc_resends++;
+	else
+		ibp->n_rc_resends += (qp->s_psn - psn) & QIB_PSN_MASK;
+
+	qp->s_flags &= ~(QIB_S_WAIT_FENCE | QIB_S_WAIT_RDMAR |
+			 QIB_S_WAIT_SSN_CREDIT | QIB_S_WAIT_PSN |
+			 QIB_S_WAIT_ACK);
+	if (wait)
+		qp->s_flags |= QIB_S_SEND_ONE;
+	reset_psn(qp, psn);
+}
+
+/*
+ * This is called from s_timer for missing responses.
+ */
+static void rc_timeout(unsigned long arg)
+{
+	struct qib_qp *qp = (struct qib_qp *)arg;
+	struct qib_ibport *ibp;
+	unsigned long flags;
+
+	spin_lock_irqsave(&qp->s_lock, flags);
+	if (qp->s_flags & QIB_S_TIMER) {
+		ibp = to_iport(qp->ibqp.device, qp->port_num);
+		ibp->n_rc_timeouts++;
+		qp->s_flags &= ~QIB_S_TIMER;
+		del_timer(&qp->s_timer);
+		qib_restart_rc(qp, qp->s_last_psn + 1, 1);
+		qib_schedule_send(qp);
+	}
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+}
+
+/*
+ * This is called from s_timer for RNR timeouts.
+ */
+void qib_rc_rnr_retry(unsigned long arg)
+{
+	struct qib_qp *qp = (struct qib_qp *)arg;
+	unsigned long flags;
+
+	spin_lock_irqsave(&qp->s_lock, flags);
+	if (qp->s_flags & QIB_S_WAIT_RNR) {
+		qp->s_flags &= ~QIB_S_WAIT_RNR;
+		del_timer(&qp->s_timer);
+		qib_schedule_send(qp);
+	}
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+}
+
+/*
+ * Set qp->s_sending_psn to the next PSN after the given one.
+ * This would be psn+1 except when RDMA reads are present.
+ */
+static void reset_sending_psn(struct qib_qp *qp, u32 psn)
+{
+	struct qib_swqe *wqe;
+	u32 n = qp->s_last;
+
+	/* Find the work request corresponding to the given PSN. */
+	for (;;) {
+		wqe = get_swqe_ptr(qp, n);
+		if (qib_cmp24(psn, wqe->lpsn) <= 0) {
+			if (wqe->wr.opcode == IB_WR_RDMA_READ)
+				qp->s_sending_psn = wqe->lpsn + 1;
+			else
+				qp->s_sending_psn = psn + 1;
+			break;
+		}
+		if (++n == qp->s_size)
+			n = 0;
+		if (n == qp->s_tail)
+			break;
+	}
+}
+
+/*
+ * This should be called with the QP s_lock held and interrupts disabled.
+ */
+void qib_rc_send_complete(struct qib_qp *qp, struct qib_ib_header *hdr)
+{
+	struct qib_other_headers *ohdr;
+	struct qib_swqe *wqe;
+	struct ib_wc wc;
+	unsigned i;
+	u32 opcode;
+	u32 psn;
+
+	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_OR_FLUSH_SEND))
+		return;
+
+	/* Find out where the BTH is */
+	if ((be16_to_cpu(hdr->lrh[0]) & 3) == QIB_LRH_BTH)
+		ohdr = &hdr->u.oth;
+	else
+		ohdr = &hdr->u.l.oth;
+
+	opcode = be32_to_cpu(ohdr->bth[0]) >> 24;
+	if (opcode >= OP(RDMA_READ_RESPONSE_FIRST) &&
+	    opcode <= OP(ATOMIC_ACKNOWLEDGE)) {
+		WARN_ON(!qp->s_rdma_ack_cnt);
+		qp->s_rdma_ack_cnt--;
+		return;
+	}
+
+	psn = be32_to_cpu(ohdr->bth[2]);
+	reset_sending_psn(qp, psn);
+
+	/*
+	 * Start timer after a packet requesting an ACK has been sent and
+	 * there are still requests that haven't been acked.
+	 */
+	if ((psn & IB_BTH_REQ_ACK) && qp->s_acked != qp->s_tail &&
+	    !(qp->s_flags & (QIB_S_TIMER | QIB_S_WAIT_RNR | QIB_S_WAIT_PSN)))
+		start_timer(qp);
+
+	while (qp->s_last != qp->s_acked) {
+		wqe = get_swqe_ptr(qp, qp->s_last);
+		if (qib_cmp24(wqe->lpsn, qp->s_sending_psn) >= 0 &&
+		    qib_cmp24(qp->s_sending_psn, qp->s_sending_hpsn) <= 0)
+			break;
+		for (i = 0; i < wqe->wr.num_sge; i++) {
+			struct qib_sge *sge = &wqe->sg_list[i];
+
+			atomic_dec(&sge->mr->refcount);
+		}
+		/* Post a send completion queue entry if requested. */
+		if (!(qp->s_flags & QIB_S_SIGNAL_REQ_WR) ||
+		    (wqe->wr.send_flags & IB_SEND_SIGNALED)) {
+			memset(&wc, 0, sizeof wc);
+			wc.wr_id = wqe->wr.wr_id;
+			wc.status = IB_WC_SUCCESS;
+			wc.opcode = ib_qib_wc_opcode[wqe->wr.opcode];
+			wc.byte_len = wqe->length;
+			wc.qp = &qp->ibqp;
+			qib_cq_enter(to_icq(qp->ibqp.send_cq), &wc, 0);
+		}
+		if (++qp->s_last >= qp->s_size)
+			qp->s_last = 0;
+	}
+	/*
+	 * If we were waiting for sends to complete before resending,
+	 * and they are now complete, restart sending.
+	 */
+	if (qp->s_flags & QIB_S_WAIT_PSN &&
+	    qib_cmp24(qp->s_sending_psn, qp->s_sending_hpsn) > 0) {
+		qp->s_flags &= ~QIB_S_WAIT_PSN;
+		qp->s_sending_psn = qp->s_psn;
+		qp->s_sending_hpsn = qp->s_psn - 1;
+		qib_schedule_send(qp);
+	}
+}
+
+static inline void update_last_psn(struct qib_qp *qp, u32 psn)
+{
+	qp->s_last_psn = psn;
+}
+
+/*
+ * Generate a SWQE completion.
+ * This is similar to qib_send_complete but has to check to be sure
+ * that the SGEs are not being referenced if the SWQE is being resent.
+ */
+static struct qib_swqe *do_rc_completion(struct qib_qp *qp,
+					 struct qib_swqe *wqe,
+					 struct qib_ibport *ibp)
+{
+	struct ib_wc wc;
+	unsigned i;
+
+	/*
+	 * Don't decrement refcount and don't generate a
+	 * completion if the SWQE is being resent until the send
+	 * is finished.
+	 */
+	if (qib_cmp24(wqe->lpsn, qp->s_sending_psn) < 0 ||
+	    qib_cmp24(qp->s_sending_psn, qp->s_sending_hpsn) > 0) {
+		for (i = 0; i < wqe->wr.num_sge; i++) {
+			struct qib_sge *sge = &wqe->sg_list[i];
+
+			atomic_dec(&sge->mr->refcount);
+		}
+		/* Post a send completion queue entry if requested. */
+		if (!(qp->s_flags & QIB_S_SIGNAL_REQ_WR) ||
+		    (wqe->wr.send_flags & IB_SEND_SIGNALED)) {
+			memset(&wc, 0, sizeof wc);
+			wc.wr_id = wqe->wr.wr_id;
+			wc.status = IB_WC_SUCCESS;
+			wc.opcode = ib_qib_wc_opcode[wqe->wr.opcode];
+			wc.byte_len = wqe->length;
+			wc.qp = &qp->ibqp;
+			qib_cq_enter(to_icq(qp->ibqp.send_cq), &wc, 0);
+		}
+		if (++qp->s_last >= qp->s_size)
+			qp->s_last = 0;
+	} else
+		ibp->n_rc_delayed_comp++;
+
+	qp->s_retry = qp->s_retry_cnt;
+	update_last_psn(qp, wqe->lpsn);
+
+	/*
+	 * If we are completing a request which is in the process of
+	 * being resent, we can stop resending it since we know the
+	 * responder has already seen it.
+	 */
+	if (qp->s_acked == qp->s_cur) {
+		if (++qp->s_cur >= qp->s_size)
+			qp->s_cur = 0;
+		qp->s_acked = qp->s_cur;
+		wqe = get_swqe_ptr(qp, qp->s_cur);
+		if (qp->s_acked != qp->s_tail) {
+			qp->s_state = OP(SEND_LAST);
+			qp->s_psn = wqe->psn;
+		}
+	} else {
+		if (++qp->s_acked >= qp->s_size)
+			qp->s_acked = 0;
+		if (qp->state == IB_QPS_SQD && qp->s_acked == qp->s_cur)
+			qp->s_draining = 0;
+		wqe = get_swqe_ptr(qp, qp->s_acked);
+	}
+	return wqe;
+}
+
+/**
+ * do_rc_ack - process an incoming RC ACK
+ * @qp: the QP the ACK came in on
+ * @psn: the packet sequence number of the ACK
+ * @opcode: the opcode of the request that resulted in the ACK
+ *
+ * This is called from qib_rc_rcv_resp() to process an incoming RC ACK
+ * for the given QP.
+ * Called at interrupt level with the QP s_lock held.
+ * Returns 1 if OK, 0 if current operation should be aborted (NAK).
+ */
+static int do_rc_ack(struct qib_qp *qp, u32 aeth, u32 psn, int opcode,
+		     u64 val, struct qib_ctxtdata *rcd)
+{
+	struct qib_ibport *ibp;
+	enum ib_wc_status status;
+	struct qib_swqe *wqe;
+	int ret = 0;
+	u32 ack_psn;
+	int diff;
+
+	/* Remove QP from retry timer */
+	if (qp->s_flags & (QIB_S_TIMER | QIB_S_WAIT_RNR)) {
+		qp->s_flags &= ~(QIB_S_TIMER | QIB_S_WAIT_RNR);
+		del_timer(&qp->s_timer);
+	}
+
+	/*
+	 * Note that NAKs implicitly ACK outstanding SEND and RDMA write
+	 * requests and implicitly NAK RDMA read and atomic requests issued
+	 * before the NAK'ed request.  The MSN won't include the NAK'ed
+	 * request but will include an ACK'ed request(s).
+	 */
+	ack_psn = psn;
+	if (aeth >> 29)
+		ack_psn--;
+	wqe = get_swqe_ptr(qp, qp->s_acked);
+	ibp = to_iport(qp->ibqp.device, qp->port_num);
+
+	/*
+	 * The MSN might be for a later WQE than the PSN indicates so
+	 * only complete WQEs that the PSN finishes.
+	 */
+	while ((diff = qib_cmp24(ack_psn, wqe->lpsn)) >= 0) {
+		/*
+		 * RDMA_READ_RESPONSE_ONLY is a special case since
+		 * we want to generate completion events for everything
+		 * before the RDMA read, copy the data, then generate
+		 * the completion for the read.
+		 */
+		if (wqe->wr.opcode == IB_WR_RDMA_READ &&
+		    opcode == OP(RDMA_READ_RESPONSE_ONLY) &&
+		    diff == 0) {
+			ret = 1;
+			goto bail;
+		}
+		/*
+		 * If this request is a RDMA read or atomic, and the ACK is
+		 * for a later operation, this ACK NAKs the RDMA read or
+		 * atomic.  In other words, only a RDMA_READ_LAST or ONLY
+		 * can ACK a RDMA read and likewise for atomic ops.  Note
+		 * that the NAK case can only happen if relaxed ordering is
+		 * used and requests are sent after an RDMA read or atomic
+		 * is sent but before the response is received.
+		 */
+		if ((wqe->wr.opcode == IB_WR_RDMA_READ &&
+		     (opcode != OP(RDMA_READ_RESPONSE_LAST) || diff != 0)) ||
+		    ((wqe->wr.opcode == IB_WR_ATOMIC_CMP_AND_SWP ||
+		      wqe->wr.opcode == IB_WR_ATOMIC_FETCH_AND_ADD) &&
+		     (opcode != OP(ATOMIC_ACKNOWLEDGE) || diff != 0))) {
+			/* Retry this request. */
+			if (!(qp->r_flags & QIB_R_RDMAR_SEQ)) {
+				qp->r_flags |= QIB_R_RDMAR_SEQ;
+				qib_restart_rc(qp, qp->s_last_psn + 1, 0);
+				if (list_empty(&qp->rspwait)) {
+					qp->r_flags |= QIB_R_RSP_SEND;
+					atomic_inc(&qp->refcount);
+					list_add_tail(&qp->rspwait,
+						      &rcd->qp_wait_list);
+				}
+			}
+			/*
+			 * No need to process the ACK/NAK since we are
+			 * restarting an earlier request.
+			 */
+			goto bail;
+		}
+		if (wqe->wr.opcode == IB_WR_ATOMIC_CMP_AND_SWP ||
+		    wqe->wr.opcode == IB_WR_ATOMIC_FETCH_AND_ADD) {
+			u64 *vaddr = wqe->sg_list[0].vaddr;
+			*vaddr = val;
+		}
+		if (qp->s_num_rd_atomic &&
+		    (wqe->wr.opcode == IB_WR_RDMA_READ ||
+		     wqe->wr.opcode == IB_WR_ATOMIC_CMP_AND_SWP ||
+		     wqe->wr.opcode == IB_WR_ATOMIC_FETCH_AND_ADD)) {
+			qp->s_num_rd_atomic--;
+			/* Restart sending task if fence is complete */
+			if ((qp->s_flags & QIB_S_WAIT_FENCE) &&
+			    !qp->s_num_rd_atomic) {
+				qp->s_flags &= ~(QIB_S_WAIT_FENCE |
+						 QIB_S_WAIT_ACK);
+				qib_schedule_send(qp);
+			} else if (qp->s_flags & QIB_S_WAIT_RDMAR) {
+				qp->s_flags &= ~(QIB_S_WAIT_RDMAR |
+						 QIB_S_WAIT_ACK);
+				qib_schedule_send(qp);
+			}
+		}
+		wqe = do_rc_completion(qp, wqe, ibp);
+		if (qp->s_acked == qp->s_tail)
+			break;
+	}
+
+	switch (aeth >> 29) {
+	case 0:         /* ACK */
+		ibp->n_rc_acks++;
+		if (qp->s_acked != qp->s_tail) {
+			/*
+			 * We are expecting more ACKs so
+			 * reset the retransmit timer.
+			 */
+			start_timer(qp);
+			/*
+			 * We can stop resending the earlier packets and
+			 * continue with the next packet the receiver wants.
+			 */
+			if (qib_cmp24(qp->s_psn, psn) <= 0)
+				reset_psn(qp, psn + 1);
+		} else if (qib_cmp24(qp->s_psn, psn) <= 0) {
+			qp->s_state = OP(SEND_LAST);
+			qp->s_psn = psn + 1;
+		}
+		if (qp->s_flags & QIB_S_WAIT_ACK) {
+			qp->s_flags &= ~QIB_S_WAIT_ACK;
+			qib_schedule_send(qp);
+		}
+		qib_get_credit(qp, aeth);
+		qp->s_rnr_retry = qp->s_rnr_retry_cnt;
+		qp->s_retry = qp->s_retry_cnt;
+		update_last_psn(qp, psn);
+		ret = 1;
+		goto bail;
+
+	case 1:         /* RNR NAK */
+		ibp->n_rnr_naks++;
+		if (qp->s_acked == qp->s_tail)
+			goto bail;
+		if (qp->s_flags & QIB_S_WAIT_RNR)
+			goto bail;
+		if (qp->s_rnr_retry == 0) {
+			status = IB_WC_RNR_RETRY_EXC_ERR;
+			goto class_b;
+		}
+		if (qp->s_rnr_retry_cnt < 7)
+			qp->s_rnr_retry--;
+
+		/* The last valid PSN is the previous PSN. */
+		update_last_psn(qp, psn - 1);
+
+		ibp->n_rc_resends += (qp->s_psn - psn) & QIB_PSN_MASK;
+
+		reset_psn(qp, psn);
+
+		qp->s_flags &= ~(QIB_S_WAIT_SSN_CREDIT | QIB_S_WAIT_ACK);
+		qp->s_flags |= QIB_S_WAIT_RNR;
+		qp->s_timer.function = qib_rc_rnr_retry;
+		qp->s_timer.expires = jiffies + usecs_to_jiffies(
+			ib_qib_rnr_table[(aeth >> QIB_AETH_CREDIT_SHIFT) &
+					   QIB_AETH_CREDIT_MASK]);
+		add_timer(&qp->s_timer);
+		goto bail;
+
+	case 3:         /* NAK */
+		if (qp->s_acked == qp->s_tail)
+			goto bail;
+		/* The last valid PSN is the previous PSN. */
+		update_last_psn(qp, psn - 1);
+		switch ((aeth >> QIB_AETH_CREDIT_SHIFT) &
+			QIB_AETH_CREDIT_MASK) {
+		case 0: /* PSN sequence error */
+			ibp->n_seq_naks++;
+			/*
+			 * Back up to the responder's expected PSN.
+			 * Note that we might get a NAK in the middle of an
+			 * RDMA READ response which terminates the RDMA
+			 * READ.
+			 */
+			qib_restart_rc(qp, psn, 0);
+			qib_schedule_send(qp);
+			break;
+
+		case 1: /* Invalid Request */
+			status = IB_WC_REM_INV_REQ_ERR;
+			ibp->n_other_naks++;
+			goto class_b;
+
+		case 2: /* Remote Access Error */
+			status = IB_WC_REM_ACCESS_ERR;
+			ibp->n_other_naks++;
+			goto class_b;
+
+		case 3: /* Remote Operation Error */
+			status = IB_WC_REM_OP_ERR;
+			ibp->n_other_naks++;
+class_b:
+			if (qp->s_last == qp->s_acked) {
+				qib_send_complete(qp, wqe, status);
+				qib_error_qp(qp, IB_WC_WR_FLUSH_ERR);
+			}
+			break;
+
+		default:
+			/* Ignore other reserved NAK error codes */
+			goto reserved;
+		}
+		qp->s_retry = qp->s_retry_cnt;
+		qp->s_rnr_retry = qp->s_rnr_retry_cnt;
+		goto bail;
+
+	default:                /* 2: reserved */
+reserved:
+		/* Ignore reserved NAK codes. */
+		goto bail;
+	}
+
+bail:
+	return ret;
+}
+
+/*
+ * We have seen an out of sequence RDMA read middle or last packet.
+ * This ACKs SENDs and RDMA writes up to the first RDMA read or atomic SWQE.
+ */
+static void rdma_seq_err(struct qib_qp *qp, struct qib_ibport *ibp, u32 psn,
+			 struct qib_ctxtdata *rcd)
+{
+	struct qib_swqe *wqe;
+
+	/* Remove QP from retry timer */
+	if (qp->s_flags & (QIB_S_TIMER | QIB_S_WAIT_RNR)) {
+		qp->s_flags &= ~(QIB_S_TIMER | QIB_S_WAIT_RNR);
+		del_timer(&qp->s_timer);
+	}
+
+	wqe = get_swqe_ptr(qp, qp->s_acked);
+
+	while (qib_cmp24(psn, wqe->lpsn) > 0) {
+		if (wqe->wr.opcode == IB_WR_RDMA_READ ||
+		    wqe->wr.opcode == IB_WR_ATOMIC_CMP_AND_SWP ||
+		    wqe->wr.opcode == IB_WR_ATOMIC_FETCH_AND_ADD)
+			break;
+		wqe = do_rc_completion(qp, wqe, ibp);
+	}
+
+	ibp->n_rdma_seq++;
+	qp->r_flags |= QIB_R_RDMAR_SEQ;
+	qib_restart_rc(qp, qp->s_last_psn + 1, 0);
+	if (list_empty(&qp->rspwait)) {
+		qp->r_flags |= QIB_R_RSP_SEND;
+		atomic_inc(&qp->refcount);
+		list_add_tail(&qp->rspwait, &rcd->qp_wait_list);
+	}
+}
+
+/**
+ * qib_rc_rcv_resp - process an incoming RC response packet
+ * @ibp: the port this packet came in on
+ * @ohdr: the other headers for this packet
+ * @data: the packet data
+ * @tlen: the packet length
+ * @qp: the QP for this packet
+ * @opcode: the opcode for this packet
+ * @psn: the packet sequence number for this packet
+ * @hdrsize: the header length
+ * @pmtu: the path MTU
+ *
+ * This is called from qib_rc_rcv() to process an incoming RC response
+ * packet for the given QP.
+ * Called at interrupt level.
+ */
+static void qib_rc_rcv_resp(struct qib_ibport *ibp,
+			    struct qib_other_headers *ohdr,
+			    void *data, u32 tlen,
+			    struct qib_qp *qp,
+			    u32 opcode,
+			    u32 psn, u32 hdrsize, u32 pmtu,
+			    struct qib_ctxtdata *rcd)
+{
+	struct qib_swqe *wqe;
+	enum ib_wc_status status;
+	unsigned long flags;
+	int diff;
+	u32 pad;
+	u32 aeth;
+	u64 val;
+
+	spin_lock_irqsave(&qp->s_lock, flags);
+
+	/* Double check we can process this now that we hold the s_lock. */
+	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK))
+		goto ack_done;
+
+	/* Ignore invalid responses. */
+	if (qib_cmp24(psn, qp->s_next_psn) >= 0)
+		goto ack_done;
+
+	/* Ignore duplicate responses. */
+	diff = qib_cmp24(psn, qp->s_last_psn);
+	if (unlikely(diff <= 0)) {
+		/* Update credits for "ghost" ACKs */
+		if (diff == 0 && opcode == OP(ACKNOWLEDGE)) {
+			aeth = be32_to_cpu(ohdr->u.aeth);
+			if ((aeth >> 29) == 0)
+				qib_get_credit(qp, aeth);
+		}
+		goto ack_done;
+	}
+
+	/*
+	 * Skip everything other than the PSN we expect, if we are waiting
+	 * for a reply to a restarted RDMA read or atomic op.
+	 */
+	if (qp->r_flags & QIB_R_RDMAR_SEQ) {
+		if (qib_cmp24(psn, qp->s_last_psn + 1) != 0)
+			goto ack_done;
+		qp->r_flags &= ~QIB_R_RDMAR_SEQ;
+	}
+
+	if (unlikely(qp->s_acked == qp->s_tail))
+		goto ack_done;
+	wqe = get_swqe_ptr(qp, qp->s_acked);
+	status = IB_WC_SUCCESS;
+
+	switch (opcode) {
+	case OP(ACKNOWLEDGE):
+	case OP(ATOMIC_ACKNOWLEDGE):
+	case OP(RDMA_READ_RESPONSE_FIRST):
+		aeth = be32_to_cpu(ohdr->u.aeth);
+		if (opcode == OP(ATOMIC_ACKNOWLEDGE)) {
+			__be32 *p = ohdr->u.at.atomic_ack_eth;
+
+			val = ((u64) be32_to_cpu(p[0]) << 32) |
+				be32_to_cpu(p[1]);
+		} else
+			val = 0;
+		if (!do_rc_ack(qp, aeth, psn, opcode, val, rcd) ||
+		    opcode != OP(RDMA_READ_RESPONSE_FIRST))
+			goto ack_done;
+		hdrsize += 4;
+		wqe = get_swqe_ptr(qp, qp->s_acked);
+		if (unlikely(wqe->wr.opcode != IB_WR_RDMA_READ))
+			goto ack_op_err;
+		/*
+		 * If this is a response to a resent RDMA read, we
+		 * have to be careful to copy the data to the right
+		 * location.
+		 */
+		qp->s_rdma_read_len = restart_sge(&qp->s_rdma_read_sge,
+						  wqe, psn, pmtu);
+		goto read_middle;
+
+	case OP(RDMA_READ_RESPONSE_MIDDLE):
+		/* no AETH, no ACK */
+		if (unlikely(qib_cmp24(psn, qp->s_last_psn + 1)))
+			goto ack_seq_err;
+		if (unlikely(wqe->wr.opcode != IB_WR_RDMA_READ))
+			goto ack_op_err;
+read_middle:
+		if (unlikely(tlen != (hdrsize + pmtu + 4)))
+			goto ack_len_err;
+		if (unlikely(pmtu >= qp->s_rdma_read_len))
+			goto ack_len_err;
+
+		/*
+		 * We got a response so update the timeout.
+		 * 4.096 usec. * (1 << qp->timeout)
+		 */
+		qp->s_flags |= QIB_S_TIMER;
+		mod_timer(&qp->s_timer, jiffies +
+			usecs_to_jiffies((4096UL * (1UL << qp->timeout)) /
+					 1000UL));
+		if (qp->s_flags & QIB_S_WAIT_ACK) {
+			qp->s_flags &= ~QIB_S_WAIT_ACK;
+			qib_schedule_send(qp);
+		}
+
+		if (opcode == OP(RDMA_READ_RESPONSE_MIDDLE))
+			qp->s_retry = qp->s_retry_cnt;
+
+		/*
+		 * Update the RDMA receive state but do the copy w/o
+		 * holding the locks and blocking interrupts.
+		 */
+		qp->s_rdma_read_len -= pmtu;
+		update_last_psn(qp, psn);
+		spin_unlock_irqrestore(&qp->s_lock, flags);
+		qib_copy_sge(&qp->s_rdma_read_sge, data, pmtu, 0);
+		goto bail;
+
+	case OP(RDMA_READ_RESPONSE_ONLY):
+		aeth = be32_to_cpu(ohdr->u.aeth);
+		if (!do_rc_ack(qp, aeth, psn, opcode, 0, rcd))
+			goto ack_done;
+		/* Get the number of bytes the message was padded by. */
+		pad = (be32_to_cpu(ohdr->bth[0]) >> 20) & 3;
+		/*
+		 * Check that the data size is >= 0 && <= pmtu.
+		 * Remember to account for the AETH header (4) and
+		 * ICRC (4).
+		 */
+		if (unlikely(tlen < (hdrsize + pad + 8)))
+			goto ack_len_err;
+		/*
+		 * If this is a response to a resent RDMA read, we
+		 * have to be careful to copy the data to the right
+		 * location.
+		 */
+		wqe = get_swqe_ptr(qp, qp->s_acked);
+		qp->s_rdma_read_len = restart_sge(&qp->s_rdma_read_sge,
+						  wqe, psn, pmtu);
+		goto read_last;
+
+	case OP(RDMA_READ_RESPONSE_LAST):
+		/* ACKs READ req. */
+		if (unlikely(qib_cmp24(psn, qp->s_last_psn + 1)))
+			goto ack_seq_err;
+		if (unlikely(wqe->wr.opcode != IB_WR_RDMA_READ))
+			goto ack_op_err;
+		/* Get the number of bytes the message was padded by. */
+		pad = (be32_to_cpu(ohdr->bth[0]) >> 20) & 3;
+		/*
+		 * Check that the data size is >= 1 && <= pmtu.
+		 * Remember to account for the AETH header (4) and
+		 * ICRC (4).
+		 */
+		if (unlikely(tlen <= (hdrsize + pad + 8)))
+			goto ack_len_err;
+read_last:
+		tlen -= hdrsize + pad + 8;
+		if (unlikely(tlen != qp->s_rdma_read_len))
+			goto ack_len_err;
+		aeth = be32_to_cpu(ohdr->u.aeth);
+		qib_copy_sge(&qp->s_rdma_read_sge, data, tlen, 0);
+		WARN_ON(qp->s_rdma_read_sge.num_sge);
+		(void) do_rc_ack(qp, aeth, psn,
+				 OP(RDMA_READ_RESPONSE_LAST), 0, rcd);
+		goto ack_done;
+	}
+
+ack_op_err:
+	status = IB_WC_LOC_QP_OP_ERR;
+	goto ack_err;
+
+ack_seq_err:
+	rdma_seq_err(qp, ibp, psn, rcd);
+	goto ack_done;
+
+ack_len_err:
+	status = IB_WC_LOC_LEN_ERR;
+ack_err:
+	if (qp->s_last == qp->s_acked) {
+		qib_send_complete(qp, wqe, status);
+		qib_error_qp(qp, IB_WC_WR_FLUSH_ERR);
+	}
+ack_done:
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+bail:
+	return;
+}
+
+/**
+ * qib_rc_rcv_error - process an incoming duplicate or error RC packet
+ * @ohdr: the other headers for this packet
+ * @data: the packet data
+ * @qp: the QP for this packet
+ * @opcode: the opcode for this packet
+ * @psn: the packet sequence number for this packet
+ * @diff: the difference between the PSN and the expected PSN
+ *
+ * This is called from qib_rc_rcv() to process an unexpected
+ * incoming RC packet for the given QP.
+ * Called at interrupt level.
+ * Return 1 if no more processing is needed; otherwise return 0 to
+ * schedule a response to be sent.
+ */
+static int qib_rc_rcv_error(struct qib_other_headers *ohdr,
+			    void *data,
+			    struct qib_qp *qp,
+			    u32 opcode,
+			    u32 psn,
+			    int diff,
+			    struct qib_ctxtdata *rcd)
+{
+	struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
+	struct qib_ack_entry *e;
+	unsigned long flags;
+	u8 i, prev;
+	int old_req;
+
+	if (diff > 0) {
+		/*
+		 * Packet sequence error.
+		 * A NAK will ACK earlier sends and RDMA writes.
+		 * Don't queue the NAK if we already sent one.
+		 */
+		if (!qp->r_nak_state) {
+			ibp->n_rc_seqnak++;
+			qp->r_nak_state = IB_NAK_PSN_ERROR;
+			/* Use the expected PSN. */
+			qp->r_ack_psn = qp->r_psn;
+			/*
+			 * Wait to send the sequence NAK until all packets
+			 * in the receive queue have been processed.
+			 * Otherwise, we end up propagating congestion.
+			 */
+			if (list_empty(&qp->rspwait)) {
+				qp->r_flags |= QIB_R_RSP_NAK;
+				atomic_inc(&qp->refcount);
+				list_add_tail(&qp->rspwait, &rcd->qp_wait_list);
+			}
+		}
+		goto done;
+	}
+
+	/*
+	 * Handle a duplicate request.  Don't re-execute SEND, RDMA
+	 * write or atomic op.  Don't NAK errors, just silently drop
+	 * the duplicate request.  Note that r_sge, r_len, and
+	 * r_rcv_len may be in use so don't modify them.
+	 *
+	 * We are supposed to ACK the earliest duplicate PSN but we
+	 * can coalesce an outstanding duplicate ACK.  We have to
+	 * send the earliest so that RDMA reads can be restarted at
+	 * the requester's expected PSN.
+	 *
+	 * First, find where this duplicate PSN falls within the
+	 * ACKs previously sent.
+	 * old_req is true if there is an older response that is scheduled
+	 * to be sent before sending this one.
+	 */
+	e = NULL;
+	old_req = 1;
+	ibp->n_rc_dupreq++;
+
+	spin_lock_irqsave(&qp->s_lock, flags);
+	/* Double check we can process this now that we hold the s_lock. */
+	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK))
+		goto unlock_done;
+
+	for (i = qp->r_head_ack_queue; ; i = prev) {
+		if (i == qp->s_tail_ack_queue)
+			old_req = 0;
+		if (i)
+			prev = i - 1;
+		else
+			prev = QIB_MAX_RDMA_ATOMIC;
+		if (prev == qp->r_head_ack_queue) {
+			e = NULL;
+			break;
+		}
+		e = &qp->s_ack_queue[prev];
+		if (!e->opcode) {
+			e = NULL;
+			break;
+		}
+		if (qib_cmp24(psn, e->psn) >= 0) {
+			if (prev == qp->s_tail_ack_queue &&
+			    qib_cmp24(psn, e->lpsn) <= 0)
+				old_req = 0;
+			break;
+		}
+	}
+	switch (opcode) {
+	case OP(RDMA_READ_REQUEST): {
+		struct ib_reth *reth;
+		u32 offset;
+		u32 len;
+
+		/*
+		 * If we didn't find the RDMA read request in the ack queue,
+		 * we can ignore this request.
+		 */
+		if (!e || e->opcode != OP(RDMA_READ_REQUEST))
+			goto unlock_done;
+		/* RETH comes after BTH */
+		reth = &ohdr->u.rc.reth;
+		/*
+		 * Address range must be a subset of the original
+		 * request and start on pmtu boundaries.
+		 * We reuse the old ack_queue slot since the requester
+		 * should not back up and request an earlier PSN for the
+		 * same request.
+		 */
+		offset = ((psn - e->psn) & QIB_PSN_MASK) *
+			ib_mtu_enum_to_int(qp->path_mtu);
+		len = be32_to_cpu(reth->length);
+		if (unlikely(offset + len != e->rdma_sge.sge_length))
+			goto unlock_done;
+		if (e->rdma_sge.mr) {
+			atomic_dec(&e->rdma_sge.mr->refcount);
+			e->rdma_sge.mr = NULL;
+		}
+		if (len != 0) {
+			u32 rkey = be32_to_cpu(reth->rkey);
+			u64 vaddr = be64_to_cpu(reth->vaddr);
+			int ok;
+
+			ok = qib_rkey_ok(qp, &e->rdma_sge, len, vaddr, rkey,
+					 IB_ACCESS_REMOTE_READ);
+			if (unlikely(!ok))
+				goto unlock_done;
+		} else {
+			e->rdma_sge.vaddr = NULL;
+			e->rdma_sge.length = 0;
+			e->rdma_sge.sge_length = 0;
+		}
+		e->psn = psn;
+		if (old_req)
+			goto unlock_done;
+		qp->s_tail_ack_queue = prev;
+		break;
+	}
+
+	case OP(COMPARE_SWAP):
+	case OP(FETCH_ADD): {
+		/*
+		 * If we didn't find the atomic request in the ack queue
+		 * or the send tasklet is already backed up to send an
+		 * earlier entry, we can ignore this request.
+		 */
+		if (!e || e->opcode != (u8) opcode || old_req)
+			goto unlock_done;
+		qp->s_tail_ack_queue = prev;
+		break;
+	}
+
+	default:
+		/*
+		 * Ignore this operation if it doesn't request an ACK
+		 * or an earlier RDMA read or atomic is going to be resent.
+		 */
+		if (!(psn & IB_BTH_REQ_ACK) || old_req)
+			goto unlock_done;
+		/*
+		 * Resend the most recent ACK if this request is
+		 * after all the previous RDMA reads and atomics.
+		 */
+		if (i == qp->r_head_ack_queue) {
+			spin_unlock_irqrestore(&qp->s_lock, flags);
+			qp->r_nak_state = 0;
+			qp->r_ack_psn = qp->r_psn - 1;
+			goto send_ack;
+		}
+		/*
+		 * Try to send a simple ACK to work around a Mellanox bug
+		 * which doesn't accept a RDMA read response or atomic
+		 * response as an ACK for earlier SENDs or RDMA writes.
+		 */
+		if (!(qp->s_flags & QIB_S_RESP_PENDING)) {
+			spin_unlock_irqrestore(&qp->s_lock, flags);
+			qp->r_nak_state = 0;
+			qp->r_ack_psn = qp->s_ack_queue[i].psn - 1;
+			goto send_ack;
+		}
+		/*
+		 * Resend the RDMA read or atomic op which
+		 * ACKs this duplicate request.
+		 */
+		qp->s_tail_ack_queue = i;
+		break;
+	}
+	qp->s_ack_state = OP(ACKNOWLEDGE);
+	qp->s_flags |= QIB_S_RESP_PENDING;
+	qp->r_nak_state = 0;
+	qib_schedule_send(qp);
+
+unlock_done:
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+done:
+	return 1;
+
+send_ack:
+	return 0;
+}
+
+void qib_rc_error(struct qib_qp *qp, enum ib_wc_status err)
+{
+	unsigned long flags;
+	int lastwqe;
+
+	spin_lock_irqsave(&qp->s_lock, flags);
+	lastwqe = qib_error_qp(qp, err);
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+
+	if (lastwqe) {
+		struct ib_event ev;
+
+		ev.device = qp->ibqp.device;
+		ev.element.qp = &qp->ibqp;
+		ev.event = IB_EVENT_QP_LAST_WQE_REACHED;
+		qp->ibqp.event_handler(&ev, qp->ibqp.qp_context);
+	}
+}
+
+static inline void qib_update_ack_queue(struct qib_qp *qp, unsigned n)
+{
+	unsigned next;
+
+	next = n + 1;
+	if (next > QIB_MAX_RDMA_ATOMIC)
+		next = 0;
+	qp->s_tail_ack_queue = next;
+	qp->s_ack_state = OP(ACKNOWLEDGE);
+}
+
+/**
+ * qib_rc_rcv - process an incoming RC packet
+ * @rcd: the context pointer
+ * @hdr: the header of this packet
+ * @has_grh: true if the header has a GRH
+ * @data: the packet data
+ * @tlen: the packet length
+ * @qp: the QP for this packet
+ *
+ * This is called from qib_qp_rcv() to process an incoming RC packet
+ * for the given QP.
+ * Called at interrupt level.
+ */
+void qib_rc_rcv(struct qib_ctxtdata *rcd, struct qib_ib_header *hdr,
+		int has_grh, void *data, u32 tlen, struct qib_qp *qp)
+{
+	struct qib_ibport *ibp = &rcd->ppd->ibport_data;
+	struct qib_other_headers *ohdr;
+	u32 opcode;
+	u32 hdrsize;
+	u32 psn;
+	u32 pad;
+	struct ib_wc wc;
+	u32 pmtu = ib_mtu_enum_to_int(qp->path_mtu);
+	int diff;
+	struct ib_reth *reth;
+	unsigned long flags;
+	int ret;
+
+	/* Check for GRH */
+	if (!has_grh) {
+		ohdr = &hdr->u.oth;
+		hdrsize = 8 + 12;       /* LRH + BTH */
+	} else {
+		ohdr = &hdr->u.l.oth;
+		hdrsize = 8 + 40 + 12;  /* LRH + GRH + BTH */
+	}
+
+	opcode = be32_to_cpu(ohdr->bth[0]);
+	spin_lock_irqsave(&qp->s_lock, flags);
+	if (qib_ruc_check_hdr(ibp, hdr, has_grh, qp, opcode))
+		goto sunlock;
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+
+	psn = be32_to_cpu(ohdr->bth[2]);
+	opcode >>= 24;
+
+	/* Prevent simultaneous processing after APM on different CPUs */
+	spin_lock(&qp->r_lock);
+
+	/*
+	 * Process responses (ACKs) before anything else.  Note that the
+	 * packet sequence number will be for something in the send work
+	 * queue rather than the expected receive packet sequence number.
+	 * In other words, this QP is the requester.
+	 */
+	if (opcode >= OP(RDMA_READ_RESPONSE_FIRST) &&
+	    opcode <= OP(ATOMIC_ACKNOWLEDGE)) {
+		qib_rc_rcv_resp(ibp, ohdr, data, tlen, qp, opcode, psn,
+				hdrsize, pmtu, rcd);
+		goto runlock;
+	}
+
+	/* Compute 24 bits worth of difference. */
+	diff = qib_cmp24(psn, qp->r_psn);
+	if (unlikely(diff)) {
+		if (qib_rc_rcv_error(ohdr, data, qp, opcode, psn, diff, rcd))
+			goto runlock;
+		goto send_ack;
+	}
+
+	/* Check for opcode sequence errors. */
+	switch (qp->r_state) {
+	case OP(SEND_FIRST):
+	case OP(SEND_MIDDLE):
+		if (opcode == OP(SEND_MIDDLE) ||
+		    opcode == OP(SEND_LAST) ||
+		    opcode == OP(SEND_LAST_WITH_IMMEDIATE))
+			break;
+		goto nack_inv;
+
+	case OP(RDMA_WRITE_FIRST):
+	case OP(RDMA_WRITE_MIDDLE):
+		if (opcode == OP(RDMA_WRITE_MIDDLE) ||
+		    opcode == OP(RDMA_WRITE_LAST) ||
+		    opcode == OP(RDMA_WRITE_LAST_WITH_IMMEDIATE))
+			break;
+		goto nack_inv;
+
+	default:
+		if (opcode == OP(SEND_MIDDLE) ||
+		    opcode == OP(SEND_LAST) ||
+		    opcode == OP(SEND_LAST_WITH_IMMEDIATE) ||
+		    opcode == OP(RDMA_WRITE_MIDDLE) ||
+		    opcode == OP(RDMA_WRITE_LAST) ||
+		    opcode == OP(RDMA_WRITE_LAST_WITH_IMMEDIATE))
+			goto nack_inv;
+		/*
+		 * Note that it is up to the requester to not send a new
+		 * RDMA read or atomic operation before receiving an ACK
+		 * for the previous operation.
+		 */
+		break;
+	}
+
+	memset(&wc, 0, sizeof wc);
+
+	if (qp->state == IB_QPS_RTR && !(qp->r_flags & QIB_R_COMM_EST)) {
+		qp->r_flags |= QIB_R_COMM_EST;
+		if (qp->ibqp.event_handler) {
+			struct ib_event ev;
+
+			ev.device = qp->ibqp.device;
+			ev.element.qp = &qp->ibqp;
+			ev.event = IB_EVENT_COMM_EST;
+			qp->ibqp.event_handler(&ev, qp->ibqp.qp_context);
+		}
+	}
+
+	/* OK, process the packet. */
+	switch (opcode) {
+	case OP(SEND_FIRST):
+		ret = qib_get_rwqe(qp, 0);
+		if (ret < 0)
+			goto nack_op_err;
+		if (!ret)
+			goto rnr_nak;
+		qp->r_rcv_len = 0;
+		/* FALLTHROUGH */
+	case OP(SEND_MIDDLE):
+	case OP(RDMA_WRITE_MIDDLE):
+send_middle:
+		/* Check for invalid length PMTU or posted rwqe len. */
+		if (unlikely(tlen != (hdrsize + pmtu + 4)))
+			goto nack_inv;
+		qp->r_rcv_len += pmtu;
+		if (unlikely(qp->r_rcv_len > qp->r_len))
+			goto nack_inv;
+		qib_copy_sge(&qp->r_sge, data, pmtu, 1);
+		break;
+
+	case OP(RDMA_WRITE_LAST_WITH_IMMEDIATE):
+		/* consume RWQE */
+		ret = qib_get_rwqe(qp, 1);
+		if (ret < 0)
+			goto nack_op_err;
+		if (!ret)
+			goto rnr_nak;
+		goto send_last_imm;
+
+	case OP(SEND_ONLY):
+	case OP(SEND_ONLY_WITH_IMMEDIATE):
+		ret = qib_get_rwqe(qp, 0);
+		if (ret < 0)
+			goto nack_op_err;
+		if (!ret)
+			goto rnr_nak;
+		qp->r_rcv_len = 0;
+		if (opcode == OP(SEND_ONLY))
+			goto send_last;
+		/* FALLTHROUGH */
+	case OP(SEND_LAST_WITH_IMMEDIATE):
+send_last_imm:
+		wc.ex.imm_data = ohdr->u.imm_data;
+		hdrsize += 4;
+		wc.wc_flags = IB_WC_WITH_IMM;
+		/* FALLTHROUGH */
+	case OP(SEND_LAST):
+	case OP(RDMA_WRITE_LAST):
+send_last:
+		/* Get the number of bytes the message was padded by. */
+		pad = (be32_to_cpu(ohdr->bth[0]) >> 20) & 3;
+		/* Check for invalid length. */
+		/* XXX LAST len should be >= 1 */
+		if (unlikely(tlen < (hdrsize + pad + 4)))
+			goto nack_inv;
+		/* Don't count the CRC. */
+		tlen -= (hdrsize + pad + 4);
+		wc.byte_len = tlen + qp->r_rcv_len;
+		if (unlikely(wc.byte_len > qp->r_len))
+			goto nack_inv;
+		qib_copy_sge(&qp->r_sge, data, tlen, 1);
+		while (qp->r_sge.num_sge) {
+			atomic_dec(&qp->r_sge.sge.mr->refcount);
+			if (--qp->r_sge.num_sge)
+				qp->r_sge.sge = *qp->r_sge.sg_list++;
+		}
+		qp->r_msn++;
+		if (!test_and_clear_bit(QIB_R_WRID_VALID, &qp->r_aflags))
+			break;
+		wc.wr_id = qp->r_wr_id;
+		wc.status = IB_WC_SUCCESS;
+		if (opcode == OP(RDMA_WRITE_LAST_WITH_IMMEDIATE) ||
+		    opcode == OP(RDMA_WRITE_ONLY_WITH_IMMEDIATE))
+			wc.opcode = IB_WC_RECV_RDMA_WITH_IMM;
+		else
+			wc.opcode = IB_WC_RECV;
+		wc.qp = &qp->ibqp;
+		wc.src_qp = qp->remote_qpn;
+		wc.slid = qp->remote_ah_attr.dlid;
+		wc.sl = qp->remote_ah_attr.sl;
+		/* Signal completion event if the solicited bit is set. */
+		qib_cq_enter(to_icq(qp->ibqp.recv_cq), &wc,
+			     (ohdr->bth[0] &
+			      cpu_to_be32(IB_BTH_SOLICITED)) != 0);
+		break;
+
+	case OP(RDMA_WRITE_FIRST):
+	case OP(RDMA_WRITE_ONLY):
+	case OP(RDMA_WRITE_ONLY_WITH_IMMEDIATE):
+		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_WRITE)))
+			goto nack_inv;
+		/* consume RWQE */
+		reth = &ohdr->u.rc.reth;
+		hdrsize += sizeof(*reth);
+		qp->r_len = be32_to_cpu(reth->length);
+		qp->r_rcv_len = 0;
+		qp->r_sge.sg_list = NULL;
+		if (qp->r_len != 0) {
+			u32 rkey = be32_to_cpu(reth->rkey);
+			u64 vaddr = be64_to_cpu(reth->vaddr);
+			int ok;
+
+			/* Check rkey & NAK */
+			ok = qib_rkey_ok(qp, &qp->r_sge.sge, qp->r_len, vaddr,
+					 rkey, IB_ACCESS_REMOTE_WRITE);
+			if (unlikely(!ok))
+				goto nack_acc;
+			qp->r_sge.num_sge = 1;
+		} else {
+			qp->r_sge.num_sge = 0;
+			qp->r_sge.sge.mr = NULL;
+			qp->r_sge.sge.vaddr = NULL;
+			qp->r_sge.sge.length = 0;
+			qp->r_sge.sge.sge_length = 0;
+		}
+		if (opcode == OP(RDMA_WRITE_FIRST))
+			goto send_middle;
+		else if (opcode == OP(RDMA_WRITE_ONLY))
+			goto send_last;
+		ret = qib_get_rwqe(qp, 1);
+		if (ret < 0)
+			goto nack_op_err;
+		if (!ret)
+			goto rnr_nak;
+		goto send_last_imm;
+
+	case OP(RDMA_READ_REQUEST): {
+		struct qib_ack_entry *e;
+		u32 len;
+		u8 next;
+
+		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_READ)))
+			goto nack_inv;
+		next = qp->r_head_ack_queue + 1;
+		/* s_ack_queue is size QIB_MAX_RDMA_ATOMIC+1 so use > not >= */
+		if (next > QIB_MAX_RDMA_ATOMIC)
+			next = 0;
+		spin_lock_irqsave(&qp->s_lock, flags);
+		/* Double check we can process this while holding the s_lock. */
+		if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK))
+			goto srunlock;
+		if (unlikely(next == qp->s_tail_ack_queue)) {
+			if (!qp->s_ack_queue[next].sent)
+				goto nack_inv_unlck;
+			qib_update_ack_queue(qp, next);
+		}
+		e = &qp->s_ack_queue[qp->r_head_ack_queue];
+		if (e->opcode == OP(RDMA_READ_REQUEST) && e->rdma_sge.mr) {
+			atomic_dec(&e->rdma_sge.mr->refcount);
+			e->rdma_sge.mr = NULL;
+		}
+		reth = &ohdr->u.rc.reth;
+		len = be32_to_cpu(reth->length);
+		if (len) {
+			u32 rkey = be32_to_cpu(reth->rkey);
+			u64 vaddr = be64_to_cpu(reth->vaddr);
+			int ok;
+
+			/* Check rkey & NAK */
+			ok = qib_rkey_ok(qp, &e->rdma_sge, len, vaddr,
+					 rkey, IB_ACCESS_REMOTE_READ);
+			if (unlikely(!ok))
+				goto nack_acc_unlck;
+			/*
+			 * Update the next expected PSN.  We add 1 later
+			 * below, so only add the remainder here.
+			 */
+			if (len > pmtu)
+				qp->r_psn += (len - 1) / pmtu;
+		} else {
+			e->rdma_sge.mr = NULL;
+			e->rdma_sge.vaddr = NULL;
+			e->rdma_sge.length = 0;
+			e->rdma_sge.sge_length = 0;
+		}
+		e->opcode = opcode;
+		e->sent = 0;
+		e->psn = psn;
+		e->lpsn = qp->r_psn;
+		/*
+		 * We need to increment the MSN here instead of when we
+		 * finish sending the result since a duplicate request would
+		 * increment it more than once.
+		 */
+		qp->r_msn++;
+		qp->r_psn++;
+		qp->r_state = opcode;
+		qp->r_nak_state = 0;
+		qp->r_head_ack_queue = next;
+
+		/* Schedule the send tasklet. */
+		qp->s_flags |= QIB_S_RESP_PENDING;
+		qib_schedule_send(qp);
+
+		goto srunlock;
+	}
+
+	case OP(COMPARE_SWAP):
+	case OP(FETCH_ADD): {
+		struct ib_atomic_eth *ateth;
+		struct qib_ack_entry *e;
+		u64 vaddr;
+		atomic64_t *maddr;
+		u64 sdata;
+		u32 rkey;
+		u8 next;
+
+		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_ATOMIC)))
+			goto nack_inv;
+		next = qp->r_head_ack_queue + 1;
+		if (next > QIB_MAX_RDMA_ATOMIC)
+			next = 0;
+		spin_lock_irqsave(&qp->s_lock, flags);
+		/* Double check we can process this while holding the s_lock. */
+		if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK))
+			goto srunlock;
+		if (unlikely(next == qp->s_tail_ack_queue)) {
+			if (!qp->s_ack_queue[next].sent)
+				goto nack_inv_unlck;
+			qib_update_ack_queue(qp, next);
+		}
+		e = &qp->s_ack_queue[qp->r_head_ack_queue];
+		if (e->opcode == OP(RDMA_READ_REQUEST) && e->rdma_sge.mr) {
+			atomic_dec(&e->rdma_sge.mr->refcount);
+			e->rdma_sge.mr = NULL;
+		}
+		ateth = &ohdr->u.atomic_eth;
+		vaddr = ((u64) be32_to_cpu(ateth->vaddr[0]) << 32) |
+			be32_to_cpu(ateth->vaddr[1]);
+		if (unlikely(vaddr & (sizeof(u64) - 1)))
+			goto nack_inv_unlck;
+		rkey = be32_to_cpu(ateth->rkey);
+		/* Check rkey & NAK */
+		if (unlikely(!qib_rkey_ok(qp, &qp->r_sge.sge, sizeof(u64),
+					  vaddr, rkey,
+					  IB_ACCESS_REMOTE_ATOMIC)))
+			goto nack_acc_unlck;
+		/* Perform atomic OP and save result. */
+		maddr = (atomic64_t *) qp->r_sge.sge.vaddr;
+		sdata = be64_to_cpu(ateth->swap_data);
+		e->atomic_data = (opcode == OP(FETCH_ADD)) ?
+			(u64) atomic64_add_return(sdata, maddr) - sdata :
+			(u64) cmpxchg((u64 *) qp->r_sge.sge.vaddr,
+				      be64_to_cpu(ateth->compare_data),
+				      sdata);
+		atomic_dec(&qp->r_sge.sge.mr->refcount);
+		qp->r_sge.num_sge = 0;
+		e->opcode = opcode;
+		e->sent = 0;
+		e->psn = psn;
+		e->lpsn = psn;
+		qp->r_msn++;
+		qp->r_psn++;
+		qp->r_state = opcode;
+		qp->r_nak_state = 0;
+		qp->r_head_ack_queue = next;
+
+		/* Schedule the send tasklet. */
+		qp->s_flags |= QIB_S_RESP_PENDING;
+		qib_schedule_send(qp);
+
+		goto srunlock;
+	}
+
+	default:
+		/* NAK unknown opcodes. */
+		goto nack_inv;
+	}
+	qp->r_psn++;
+	qp->r_state = opcode;
+	qp->r_ack_psn = psn;
+	qp->r_nak_state = 0;
+	/* Send an ACK if requested or required. */
+	if (psn & (1 << 31))
+		goto send_ack;
+	goto runlock;
+
+rnr_nak:
+	qp->r_nak_state = IB_RNR_NAK | qp->r_min_rnr_timer;
+	qp->r_ack_psn = qp->r_psn;
+	/* Queue RNR NAK for later */
+	if (list_empty(&qp->rspwait)) {
+		qp->r_flags |= QIB_R_RSP_NAK;
+		atomic_inc(&qp->refcount);
+		list_add_tail(&qp->rspwait, &rcd->qp_wait_list);
+	}
+	goto runlock;
+
+nack_op_err:
+	qib_rc_error(qp, IB_WC_LOC_QP_OP_ERR);
+	qp->r_nak_state = IB_NAK_REMOTE_OPERATIONAL_ERROR;
+	qp->r_ack_psn = qp->r_psn;
+	/* Queue NAK for later */
+	if (list_empty(&qp->rspwait)) {
+		qp->r_flags |= QIB_R_RSP_NAK;
+		atomic_inc(&qp->refcount);
+		list_add_tail(&qp->rspwait, &rcd->qp_wait_list);
+	}
+	goto runlock;
+
+nack_inv_unlck:
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+nack_inv:
+	qib_rc_error(qp, IB_WC_LOC_QP_OP_ERR);
+	qp->r_nak_state = IB_NAK_INVALID_REQUEST;
+	qp->r_ack_psn = qp->r_psn;
+	/* Queue NAK for later */
+	if (list_empty(&qp->rspwait)) {
+		qp->r_flags |= QIB_R_RSP_NAK;
+		atomic_inc(&qp->refcount);
+		list_add_tail(&qp->rspwait, &rcd->qp_wait_list);
+	}
+	goto runlock;
+
+nack_acc_unlck:
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+nack_acc:
+	qib_rc_error(qp, IB_WC_LOC_PROT_ERR);
+	qp->r_nak_state = IB_NAK_REMOTE_ACCESS_ERROR;
+	qp->r_ack_psn = qp->r_psn;
+send_ack:
+	qib_send_rc_ack(qp);
+runlock:
+	spin_unlock(&qp->r_lock);
+	return;
+
+srunlock:
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+	spin_unlock(&qp->r_lock);
+	return;
+
+sunlock:
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+}
diff --git a/drivers/infiniband/hw/qib/qib_ruc.c b/drivers/infiniband/hw/qib/qib_ruc.c
new file mode 100644
index 000000000000..eb78d9367f06
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_ruc.c
@@ -0,0 +1,817 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/spinlock.h>
+
+#include "qib.h"
+#include "qib_mad.h"
+
+/*
+ * Convert the AETH RNR timeout code into the number of microseconds.
+ */
+const u32 ib_qib_rnr_table[32] = {
+	655360,	/* 00: 655.36 */
+	10,	/* 01:    .01 */
+	20,	/* 02     .02 */
+	30,	/* 03:    .03 */
+	40,	/* 04:    .04 */
+	60,	/* 05:    .06 */
+	80,	/* 06:    .08 */
+	120,	/* 07:    .12 */
+	160,	/* 08:    .16 */
+	240,	/* 09:    .24 */
+	320,	/* 0A:    .32 */
+	480,	/* 0B:    .48 */
+	640,	/* 0C:    .64 */
+	960,	/* 0D:    .96 */
+	1280,	/* 0E:   1.28 */
+	1920,	/* 0F:   1.92 */
+	2560,	/* 10:   2.56 */
+	3840,	/* 11:   3.84 */
+	5120,	/* 12:   5.12 */
+	7680,	/* 13:   7.68 */
+	10240,	/* 14:  10.24 */
+	15360,	/* 15:  15.36 */
+	20480,	/* 16:  20.48 */
+	30720,	/* 17:  30.72 */
+	40960,	/* 18:  40.96 */
+	61440,	/* 19:  61.44 */
+	81920,	/* 1A:  81.92 */
+	122880,	/* 1B: 122.88 */
+	163840,	/* 1C: 163.84 */
+	245760,	/* 1D: 245.76 */
+	327680,	/* 1E: 327.68 */
+	491520	/* 1F: 491.52 */
+};
+
+/*
+ * Validate a RWQE and fill in the SGE state.
+ * Return 1 if OK.
+ */
+static int qib_init_sge(struct qib_qp *qp, struct qib_rwqe *wqe)
+{
+	int i, j, ret;
+	struct ib_wc wc;
+	struct qib_lkey_table *rkt;
+	struct qib_pd *pd;
+	struct qib_sge_state *ss;
+
+	rkt = &to_idev(qp->ibqp.device)->lk_table;
+	pd = to_ipd(qp->ibqp.srq ? qp->ibqp.srq->pd : qp->ibqp.pd);
+	ss = &qp->r_sge;
+	ss->sg_list = qp->r_sg_list;
+	qp->r_len = 0;
+	for (i = j = 0; i < wqe->num_sge; i++) {
+		if (wqe->sg_list[i].length == 0)
+			continue;
+		/* Check LKEY */
+		if (!qib_lkey_ok(rkt, pd, j ? &ss->sg_list[j - 1] : &ss->sge,
+				 &wqe->sg_list[i], IB_ACCESS_LOCAL_WRITE))
+			goto bad_lkey;
+		qp->r_len += wqe->sg_list[i].length;
+		j++;
+	}
+	ss->num_sge = j;
+	ss->total_len = qp->r_len;
+	ret = 1;
+	goto bail;
+
+bad_lkey:
+	while (j) {
+		struct qib_sge *sge = --j ? &ss->sg_list[j - 1] : &ss->sge;
+
+		atomic_dec(&sge->mr->refcount);
+	}
+	ss->num_sge = 0;
+	memset(&wc, 0, sizeof(wc));
+	wc.wr_id = wqe->wr_id;
+	wc.status = IB_WC_LOC_PROT_ERR;
+	wc.opcode = IB_WC_RECV;
+	wc.qp = &qp->ibqp;
+	/* Signal solicited completion event. */
+	qib_cq_enter(to_icq(qp->ibqp.recv_cq), &wc, 1);
+	ret = 0;
+bail:
+	return ret;
+}
+
+/**
+ * qib_get_rwqe - copy the next RWQE into the QP's RWQE
+ * @qp: the QP
+ * @wr_id_only: update qp->r_wr_id only, not qp->r_sge
+ *
+ * Return -1 if there is a local error, 0 if no RWQE is available,
+ * otherwise return 1.
+ *
+ * Can be called from interrupt level.
+ */
+int qib_get_rwqe(struct qib_qp *qp, int wr_id_only)
+{
+	unsigned long flags;
+	struct qib_rq *rq;
+	struct qib_rwq *wq;
+	struct qib_srq *srq;
+	struct qib_rwqe *wqe;
+	void (*handler)(struct ib_event *, void *);
+	u32 tail;
+	int ret;
+
+	if (qp->ibqp.srq) {
+		srq = to_isrq(qp->ibqp.srq);
+		handler = srq->ibsrq.event_handler;
+		rq = &srq->rq;
+	} else {
+		srq = NULL;
+		handler = NULL;
+		rq = &qp->r_rq;
+	}
+
+	spin_lock_irqsave(&rq->lock, flags);
+	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK)) {
+		ret = 0;
+		goto unlock;
+	}
+
+	wq = rq->wq;
+	tail = wq->tail;
+	/* Validate tail before using it since it is user writable. */
+	if (tail >= rq->size)
+		tail = 0;
+	if (unlikely(tail == wq->head)) {
+		ret = 0;
+		goto unlock;
+	}
+	/* Make sure entry is read after head index is read. */
+	smp_rmb();
+	wqe = get_rwqe_ptr(rq, tail);
+	/*
+	 * Even though we update the tail index in memory, the verbs
+	 * consumer is not supposed to post more entries until a
+	 * completion is generated.
+	 */
+	if (++tail >= rq->size)
+		tail = 0;
+	wq->tail = tail;
+	if (!wr_id_only && !qib_init_sge(qp, wqe)) {
+		ret = -1;
+		goto unlock;
+	}
+	qp->r_wr_id = wqe->wr_id;
+
+	ret = 1;
+	set_bit(QIB_R_WRID_VALID, &qp->r_aflags);
+	if (handler) {
+		u32 n;
+
+		/*
+		 * Validate head pointer value and compute
+		 * the number of remaining WQEs.
+		 */
+		n = wq->head;
+		if (n >= rq->size)
+			n = 0;
+		if (n < tail)
+			n += rq->size - tail;
+		else
+			n -= tail;
+		if (n < srq->limit) {
+			struct ib_event ev;
+
+			srq->limit = 0;
+			spin_unlock_irqrestore(&rq->lock, flags);
+			ev.device = qp->ibqp.device;
+			ev.element.srq = qp->ibqp.srq;
+			ev.event = IB_EVENT_SRQ_LIMIT_REACHED;
+			handler(&ev, srq->ibsrq.srq_context);
+			goto bail;
+		}
+	}
+unlock:
+	spin_unlock_irqrestore(&rq->lock, flags);
+bail:
+	return ret;
+}
+
+/*
+ * Switch to alternate path.
+ * The QP s_lock should be held and interrupts disabled.
+ */
+void qib_migrate_qp(struct qib_qp *qp)
+{
+	struct ib_event ev;
+
+	qp->s_mig_state = IB_MIG_MIGRATED;
+	qp->remote_ah_attr = qp->alt_ah_attr;
+	qp->port_num = qp->alt_ah_attr.port_num;
+	qp->s_pkey_index = qp->s_alt_pkey_index;
+
+	ev.device = qp->ibqp.device;
+	ev.element.qp = &qp->ibqp;
+	ev.event = IB_EVENT_PATH_MIG;
+	qp->ibqp.event_handler(&ev, qp->ibqp.qp_context);
+}
+
+static __be64 get_sguid(struct qib_ibport *ibp, unsigned index)
+{
+	if (!index) {
+		struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+
+		return ppd->guid;
+	} else
+		return ibp->guids[index - 1];
+}
+
+static int gid_ok(union ib_gid *gid, __be64 gid_prefix, __be64 id)
+{
+	return (gid->global.interface_id == id &&
+		(gid->global.subnet_prefix == gid_prefix ||
+		 gid->global.subnet_prefix == IB_DEFAULT_GID_PREFIX));
+}
+
+/*
+ *
+ * This should be called with the QP s_lock held.
+ */
+int qib_ruc_check_hdr(struct qib_ibport *ibp, struct qib_ib_header *hdr,
+		      int has_grh, struct qib_qp *qp, u32 bth0)
+{
+	__be64 guid;
+
+	if (qp->s_mig_state == IB_MIG_ARMED && (bth0 & IB_BTH_MIG_REQ)) {
+		if (!has_grh) {
+			if (qp->alt_ah_attr.ah_flags & IB_AH_GRH)
+				goto err;
+		} else {
+			if (!(qp->alt_ah_attr.ah_flags & IB_AH_GRH))
+				goto err;
+			guid = get_sguid(ibp, qp->alt_ah_attr.grh.sgid_index);
+			if (!gid_ok(&hdr->u.l.grh.dgid, ibp->gid_prefix, guid))
+				goto err;
+			if (!gid_ok(&hdr->u.l.grh.sgid,
+			    qp->alt_ah_attr.grh.dgid.global.subnet_prefix,
+			    qp->alt_ah_attr.grh.dgid.global.interface_id))
+				goto err;
+		}
+		if (!qib_pkey_ok((u16)bth0,
+				 qib_get_pkey(ibp, qp->s_alt_pkey_index))) {
+			qib_bad_pqkey(ibp, IB_NOTICE_TRAP_BAD_PKEY,
+				      (u16)bth0,
+				      (be16_to_cpu(hdr->lrh[0]) >> 4) & 0xF,
+				      0, qp->ibqp.qp_num,
+				      hdr->lrh[3], hdr->lrh[1]);
+			goto err;
+		}
+		/* Validate the SLID. See Ch. 9.6.1.5 and 17.2.8 */
+		if (be16_to_cpu(hdr->lrh[3]) != qp->alt_ah_attr.dlid ||
+		    ppd_from_ibp(ibp)->port != qp->alt_ah_attr.port_num)
+			goto err;
+		qib_migrate_qp(qp);
+	} else {
+		if (!has_grh) {
+			if (qp->remote_ah_attr.ah_flags & IB_AH_GRH)
+				goto err;
+		} else {
+			if (!(qp->remote_ah_attr.ah_flags & IB_AH_GRH))
+				goto err;
+			guid = get_sguid(ibp,
+					 qp->remote_ah_attr.grh.sgid_index);
+			if (!gid_ok(&hdr->u.l.grh.dgid, ibp->gid_prefix, guid))
+				goto err;
+			if (!gid_ok(&hdr->u.l.grh.sgid,
+			    qp->remote_ah_attr.grh.dgid.global.subnet_prefix,
+			    qp->remote_ah_attr.grh.dgid.global.interface_id))
+				goto err;
+		}
+		if (!qib_pkey_ok((u16)bth0,
+				 qib_get_pkey(ibp, qp->s_pkey_index))) {
+			qib_bad_pqkey(ibp, IB_NOTICE_TRAP_BAD_PKEY,
+				      (u16)bth0,
+				      (be16_to_cpu(hdr->lrh[0]) >> 4) & 0xF,
+				      0, qp->ibqp.qp_num,
+				      hdr->lrh[3], hdr->lrh[1]);
+			goto err;
+		}
+		/* Validate the SLID. See Ch. 9.6.1.5 */
+		if (be16_to_cpu(hdr->lrh[3]) != qp->remote_ah_attr.dlid ||
+		    ppd_from_ibp(ibp)->port != qp->port_num)
+			goto err;
+		if (qp->s_mig_state == IB_MIG_REARM &&
+		    !(bth0 & IB_BTH_MIG_REQ))
+			qp->s_mig_state = IB_MIG_ARMED;
+	}
+
+	return 0;
+
+err:
+	return 1;
+}
+
+/**
+ * qib_ruc_loopback - handle UC and RC lookback requests
+ * @sqp: the sending QP
+ *
+ * This is called from qib_do_send() to
+ * forward a WQE addressed to the same HCA.
+ * Note that although we are single threaded due to the tasklet, we still
+ * have to protect against post_send().  We don't have to worry about
+ * receive interrupts since this is a connected protocol and all packets
+ * will pass through here.
+ */
+static void qib_ruc_loopback(struct qib_qp *sqp)
+{
+	struct qib_ibport *ibp = to_iport(sqp->ibqp.device, sqp->port_num);
+	struct qib_qp *qp;
+	struct qib_swqe *wqe;
+	struct qib_sge *sge;
+	unsigned long flags;
+	struct ib_wc wc;
+	u64 sdata;
+	atomic64_t *maddr;
+	enum ib_wc_status send_status;
+	int release;
+	int ret;
+
+	/*
+	 * Note that we check the responder QP state after
+	 * checking the requester's state.
+	 */
+	qp = qib_lookup_qpn(ibp, sqp->remote_qpn);
+
+	spin_lock_irqsave(&sqp->s_lock, flags);
+
+	/* Return if we are already busy processing a work request. */
+	if ((sqp->s_flags & (QIB_S_BUSY | QIB_S_ANY_WAIT)) ||
+	    !(ib_qib_state_ops[sqp->state] & QIB_PROCESS_OR_FLUSH_SEND))
+		goto unlock;
+
+	sqp->s_flags |= QIB_S_BUSY;
+
+again:
+	if (sqp->s_last == sqp->s_head)
+		goto clr_busy;
+	wqe = get_swqe_ptr(sqp, sqp->s_last);
+
+	/* Return if it is not OK to start a new work reqeust. */
+	if (!(ib_qib_state_ops[sqp->state] & QIB_PROCESS_NEXT_SEND_OK)) {
+		if (!(ib_qib_state_ops[sqp->state] & QIB_FLUSH_SEND))
+			goto clr_busy;
+		/* We are in the error state, flush the work request. */
+		send_status = IB_WC_WR_FLUSH_ERR;
+		goto flush_send;
+	}
+
+	/*
+	 * We can rely on the entry not changing without the s_lock
+	 * being held until we update s_last.
+	 * We increment s_cur to indicate s_last is in progress.
+	 */
+	if (sqp->s_last == sqp->s_cur) {
+		if (++sqp->s_cur >= sqp->s_size)
+			sqp->s_cur = 0;
+	}
+	spin_unlock_irqrestore(&sqp->s_lock, flags);
+
+	if (!qp || !(ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK) ||
+	    qp->ibqp.qp_type != sqp->ibqp.qp_type) {
+		ibp->n_pkt_drops++;
+		/*
+		 * For RC, the requester would timeout and retry so
+		 * shortcut the timeouts and just signal too many retries.
+		 */
+		if (sqp->ibqp.qp_type == IB_QPT_RC)
+			send_status = IB_WC_RETRY_EXC_ERR;
+		else
+			send_status = IB_WC_SUCCESS;
+		goto serr;
+	}
+
+	memset(&wc, 0, sizeof wc);
+	send_status = IB_WC_SUCCESS;
+
+	release = 1;
+	sqp->s_sge.sge = wqe->sg_list[0];
+	sqp->s_sge.sg_list = wqe->sg_list + 1;
+	sqp->s_sge.num_sge = wqe->wr.num_sge;
+	sqp->s_len = wqe->length;
+	switch (wqe->wr.opcode) {
+	case IB_WR_SEND_WITH_IMM:
+		wc.wc_flags = IB_WC_WITH_IMM;
+		wc.ex.imm_data = wqe->wr.ex.imm_data;
+		/* FALLTHROUGH */
+	case IB_WR_SEND:
+		ret = qib_get_rwqe(qp, 0);
+		if (ret < 0)
+			goto op_err;
+		if (!ret)
+			goto rnr_nak;
+		break;
+
+	case IB_WR_RDMA_WRITE_WITH_IMM:
+		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_WRITE)))
+			goto inv_err;
+		wc.wc_flags = IB_WC_WITH_IMM;
+		wc.ex.imm_data = wqe->wr.ex.imm_data;
+		ret = qib_get_rwqe(qp, 1);
+		if (ret < 0)
+			goto op_err;
+		if (!ret)
+			goto rnr_nak;
+		/* FALLTHROUGH */
+	case IB_WR_RDMA_WRITE:
+		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_WRITE)))
+			goto inv_err;
+		if (wqe->length == 0)
+			break;
+		if (unlikely(!qib_rkey_ok(qp, &qp->r_sge.sge, wqe->length,
+					  wqe->wr.wr.rdma.remote_addr,
+					  wqe->wr.wr.rdma.rkey,
+					  IB_ACCESS_REMOTE_WRITE)))
+			goto acc_err;
+		qp->r_sge.sg_list = NULL;
+		qp->r_sge.num_sge = 1;
+		qp->r_sge.total_len = wqe->length;
+		break;
+
+	case IB_WR_RDMA_READ:
+		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_READ)))
+			goto inv_err;
+		if (unlikely(!qib_rkey_ok(qp, &sqp->s_sge.sge, wqe->length,
+					  wqe->wr.wr.rdma.remote_addr,
+					  wqe->wr.wr.rdma.rkey,
+					  IB_ACCESS_REMOTE_READ)))
+			goto acc_err;
+		release = 0;
+		sqp->s_sge.sg_list = NULL;
+		sqp->s_sge.num_sge = 1;
+		qp->r_sge.sge = wqe->sg_list[0];
+		qp->r_sge.sg_list = wqe->sg_list + 1;
+		qp->r_sge.num_sge = wqe->wr.num_sge;
+		qp->r_sge.total_len = wqe->length;
+		break;
+
+	case IB_WR_ATOMIC_CMP_AND_SWP:
+	case IB_WR_ATOMIC_FETCH_AND_ADD:
+		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_ATOMIC)))
+			goto inv_err;
+		if (unlikely(!qib_rkey_ok(qp, &qp->r_sge.sge, sizeof(u64),
+					  wqe->wr.wr.atomic.remote_addr,
+					  wqe->wr.wr.atomic.rkey,
+					  IB_ACCESS_REMOTE_ATOMIC)))
+			goto acc_err;
+		/* Perform atomic OP and save result. */
+		maddr = (atomic64_t *) qp->r_sge.sge.vaddr;
+		sdata = wqe->wr.wr.atomic.compare_add;
+		*(u64 *) sqp->s_sge.sge.vaddr =
+			(wqe->wr.opcode == IB_WR_ATOMIC_FETCH_AND_ADD) ?
+			(u64) atomic64_add_return(sdata, maddr) - sdata :
+			(u64) cmpxchg((u64 *) qp->r_sge.sge.vaddr,
+				      sdata, wqe->wr.wr.atomic.swap);
+		atomic_dec(&qp->r_sge.sge.mr->refcount);
+		qp->r_sge.num_sge = 0;
+		goto send_comp;
+
+	default:
+		send_status = IB_WC_LOC_QP_OP_ERR;
+		goto serr;
+	}
+
+	sge = &sqp->s_sge.sge;
+	while (sqp->s_len) {
+		u32 len = sqp->s_len;
+
+		if (len > sge->length)
+			len = sge->length;
+		if (len > sge->sge_length)
+			len = sge->sge_length;
+		BUG_ON(len == 0);
+		qib_copy_sge(&qp->r_sge, sge->vaddr, len, release);
+		sge->vaddr += len;
+		sge->length -= len;
+		sge->sge_length -= len;
+		if (sge->sge_length == 0) {
+			if (!release)
+				atomic_dec(&sge->mr->refcount);
+			if (--sqp->s_sge.num_sge)
+				*sge = *sqp->s_sge.sg_list++;
+		} else if (sge->length == 0 && sge->mr->lkey) {
+			if (++sge->n >= QIB_SEGSZ) {
+				if (++sge->m >= sge->mr->mapsz)
+					break;
+				sge->n = 0;
+			}
+			sge->vaddr =
+				sge->mr->map[sge->m]->segs[sge->n].vaddr;
+			sge->length =
+				sge->mr->map[sge->m]->segs[sge->n].length;
+		}
+		sqp->s_len -= len;
+	}
+	if (release)
+		while (qp->r_sge.num_sge) {
+			atomic_dec(&qp->r_sge.sge.mr->refcount);
+			if (--qp->r_sge.num_sge)
+				qp->r_sge.sge = *qp->r_sge.sg_list++;
+		}
+
+	if (!test_and_clear_bit(QIB_R_WRID_VALID, &qp->r_aflags))
+		goto send_comp;
+
+	if (wqe->wr.opcode == IB_WR_RDMA_WRITE_WITH_IMM)
+		wc.opcode = IB_WC_RECV_RDMA_WITH_IMM;
+	else
+		wc.opcode = IB_WC_RECV;
+	wc.wr_id = qp->r_wr_id;
+	wc.status = IB_WC_SUCCESS;
+	wc.byte_len = wqe->length;
+	wc.qp = &qp->ibqp;
+	wc.src_qp = qp->remote_qpn;
+	wc.slid = qp->remote_ah_attr.dlid;
+	wc.sl = qp->remote_ah_attr.sl;
+	wc.port_num = 1;
+	/* Signal completion event if the solicited bit is set. */
+	qib_cq_enter(to_icq(qp->ibqp.recv_cq), &wc,
+		       wqe->wr.send_flags & IB_SEND_SOLICITED);
+
+send_comp:
+	spin_lock_irqsave(&sqp->s_lock, flags);
+	ibp->n_loop_pkts++;
+flush_send:
+	sqp->s_rnr_retry = sqp->s_rnr_retry_cnt;
+	qib_send_complete(sqp, wqe, send_status);
+	goto again;
+
+rnr_nak:
+	/* Handle RNR NAK */
+	if (qp->ibqp.qp_type == IB_QPT_UC)
+		goto send_comp;
+	ibp->n_rnr_naks++;
+	/*
+	 * Note: we don't need the s_lock held since the BUSY flag
+	 * makes this single threaded.
+	 */
+	if (sqp->s_rnr_retry == 0) {
+		send_status = IB_WC_RNR_RETRY_EXC_ERR;
+		goto serr;
+	}
+	if (sqp->s_rnr_retry_cnt < 7)
+		sqp->s_rnr_retry--;
+	spin_lock_irqsave(&sqp->s_lock, flags);
+	if (!(ib_qib_state_ops[sqp->state] & QIB_PROCESS_RECV_OK))
+		goto clr_busy;
+	sqp->s_flags |= QIB_S_WAIT_RNR;
+	sqp->s_timer.function = qib_rc_rnr_retry;
+	sqp->s_timer.expires = jiffies +
+		usecs_to_jiffies(ib_qib_rnr_table[qp->r_min_rnr_timer]);
+	add_timer(&sqp->s_timer);
+	goto clr_busy;
+
+op_err:
+	send_status = IB_WC_REM_OP_ERR;
+	wc.status = IB_WC_LOC_QP_OP_ERR;
+	goto err;
+
+inv_err:
+	send_status = IB_WC_REM_INV_REQ_ERR;
+	wc.status = IB_WC_LOC_QP_OP_ERR;
+	goto err;
+
+acc_err:
+	send_status = IB_WC_REM_ACCESS_ERR;
+	wc.status = IB_WC_LOC_PROT_ERR;
+err:
+	/* responder goes to error state */
+	qib_rc_error(qp, wc.status);
+
+serr:
+	spin_lock_irqsave(&sqp->s_lock, flags);
+	qib_send_complete(sqp, wqe, send_status);
+	if (sqp->ibqp.qp_type == IB_QPT_RC) {
+		int lastwqe = qib_error_qp(sqp, IB_WC_WR_FLUSH_ERR);
+
+		sqp->s_flags &= ~QIB_S_BUSY;
+		spin_unlock_irqrestore(&sqp->s_lock, flags);
+		if (lastwqe) {
+			struct ib_event ev;
+
+			ev.device = sqp->ibqp.device;
+			ev.element.qp = &sqp->ibqp;
+			ev.event = IB_EVENT_QP_LAST_WQE_REACHED;
+			sqp->ibqp.event_handler(&ev, sqp->ibqp.qp_context);
+		}
+		goto done;
+	}
+clr_busy:
+	sqp->s_flags &= ~QIB_S_BUSY;
+unlock:
+	spin_unlock_irqrestore(&sqp->s_lock, flags);
+done:
+	if (qp && atomic_dec_and_test(&qp->refcount))
+		wake_up(&qp->wait);
+}
+
+/**
+ * qib_make_grh - construct a GRH header
+ * @ibp: a pointer to the IB port
+ * @hdr: a pointer to the GRH header being constructed
+ * @grh: the global route address to send to
+ * @hwords: the number of 32 bit words of header being sent
+ * @nwords: the number of 32 bit words of data being sent
+ *
+ * Return the size of the header in 32 bit words.
+ */
+u32 qib_make_grh(struct qib_ibport *ibp, struct ib_grh *hdr,
+		 struct ib_global_route *grh, u32 hwords, u32 nwords)
+{
+	hdr->version_tclass_flow =
+		cpu_to_be32((IB_GRH_VERSION << IB_GRH_VERSION_SHIFT) |
+			    (grh->traffic_class << IB_GRH_TCLASS_SHIFT) |
+			    (grh->flow_label << IB_GRH_FLOW_SHIFT));
+	hdr->paylen = cpu_to_be16((hwords - 2 + nwords + SIZE_OF_CRC) << 2);
+	/* next_hdr is defined by C8-7 in ch. 8.4.1 */
+	hdr->next_hdr = IB_GRH_NEXT_HDR;
+	hdr->hop_limit = grh->hop_limit;
+	/* The SGID is 32-bit aligned. */
+	hdr->sgid.global.subnet_prefix = ibp->gid_prefix;
+	hdr->sgid.global.interface_id = grh->sgid_index ?
+		ibp->guids[grh->sgid_index - 1] : ppd_from_ibp(ibp)->guid;
+	hdr->dgid = grh->dgid;
+
+	/* GRH header size in 32-bit words. */
+	return sizeof(struct ib_grh) / sizeof(u32);
+}
+
+void qib_make_ruc_header(struct qib_qp *qp, struct qib_other_headers *ohdr,
+			 u32 bth0, u32 bth2)
+{
+	struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
+	u16 lrh0;
+	u32 nwords;
+	u32 extra_bytes;
+
+	/* Construct the header. */
+	extra_bytes = -qp->s_cur_size & 3;
+	nwords = (qp->s_cur_size + extra_bytes) >> 2;
+	lrh0 = QIB_LRH_BTH;
+	if (unlikely(qp->remote_ah_attr.ah_flags & IB_AH_GRH)) {
+		qp->s_hdrwords += qib_make_grh(ibp, &qp->s_hdr.u.l.grh,
+					       &qp->remote_ah_attr.grh,
+					       qp->s_hdrwords, nwords);
+		lrh0 = QIB_LRH_GRH;
+	}
+	lrh0 |= ibp->sl_to_vl[qp->remote_ah_attr.sl] << 12 |
+		qp->remote_ah_attr.sl << 4;
+	qp->s_hdr.lrh[0] = cpu_to_be16(lrh0);
+	qp->s_hdr.lrh[1] = cpu_to_be16(qp->remote_ah_attr.dlid);
+	qp->s_hdr.lrh[2] = cpu_to_be16(qp->s_hdrwords + nwords + SIZE_OF_CRC);
+	qp->s_hdr.lrh[3] = cpu_to_be16(ppd_from_ibp(ibp)->lid |
+				       qp->remote_ah_attr.src_path_bits);
+	bth0 |= qib_get_pkey(ibp, qp->s_pkey_index);
+	bth0 |= extra_bytes << 20;
+	if (qp->s_mig_state == IB_MIG_MIGRATED)
+		bth0 |= IB_BTH_MIG_REQ;
+	ohdr->bth[0] = cpu_to_be32(bth0);
+	ohdr->bth[1] = cpu_to_be32(qp->remote_qpn);
+	ohdr->bth[2] = cpu_to_be32(bth2);
+}
+
+/**
+ * qib_do_send - perform a send on a QP
+ * @work: contains a pointer to the QP
+ *
+ * Process entries in the send work queue until credit or queue is
+ * exhausted.  Only allow one CPU to send a packet per QP (tasklet).
+ * Otherwise, two threads could send packets out of order.
+ */
+void qib_do_send(struct work_struct *work)
+{
+	struct qib_qp *qp = container_of(work, struct qib_qp, s_work);
+	struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	int (*make_req)(struct qib_qp *qp);
+	unsigned long flags;
+
+	if ((qp->ibqp.qp_type == IB_QPT_RC ||
+	     qp->ibqp.qp_type == IB_QPT_UC) &&
+	    (qp->remote_ah_attr.dlid & ~((1 << ppd->lmc) - 1)) == ppd->lid) {
+		qib_ruc_loopback(qp);
+		return;
+	}
+
+	if (qp->ibqp.qp_type == IB_QPT_RC)
+		make_req = qib_make_rc_req;
+	else if (qp->ibqp.qp_type == IB_QPT_UC)
+		make_req = qib_make_uc_req;
+	else
+		make_req = qib_make_ud_req;
+
+	spin_lock_irqsave(&qp->s_lock, flags);
+
+	/* Return if we are already busy processing a work request. */
+	if (!qib_send_ok(qp)) {
+		spin_unlock_irqrestore(&qp->s_lock, flags);
+		return;
+	}
+
+	qp->s_flags |= QIB_S_BUSY;
+
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+
+	do {
+		/* Check for a constructed packet to be sent. */
+		if (qp->s_hdrwords != 0) {
+			/*
+			 * If the packet cannot be sent now, return and
+			 * the send tasklet will be woken up later.
+			 */
+			if (qib_verbs_send(qp, &qp->s_hdr, qp->s_hdrwords,
+					   qp->s_cur_sge, qp->s_cur_size))
+				break;
+			/* Record that s_hdr is empty. */
+			qp->s_hdrwords = 0;
+		}
+	} while (make_req(qp));
+}
+
+/*
+ * This should be called with s_lock held.
+ */
+void qib_send_complete(struct qib_qp *qp, struct qib_swqe *wqe,
+		       enum ib_wc_status status)
+{
+	u32 old_last, last;
+	unsigned i;
+
+	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_OR_FLUSH_SEND))
+		return;
+
+	for (i = 0; i < wqe->wr.num_sge; i++) {
+		struct qib_sge *sge = &wqe->sg_list[i];
+
+		atomic_dec(&sge->mr->refcount);
+	}
+	if (qp->ibqp.qp_type == IB_QPT_UD ||
+	    qp->ibqp.qp_type == IB_QPT_SMI ||
+	    qp->ibqp.qp_type == IB_QPT_GSI)
+		atomic_dec(&to_iah(wqe->wr.wr.ud.ah)->refcount);
+
+	/* See ch. 11.2.4.1 and 10.7.3.1 */
+	if (!(qp->s_flags & QIB_S_SIGNAL_REQ_WR) ||
+	    (wqe->wr.send_flags & IB_SEND_SIGNALED) ||
+	    status != IB_WC_SUCCESS) {
+		struct ib_wc wc;
+
+		memset(&wc, 0, sizeof wc);
+		wc.wr_id = wqe->wr.wr_id;
+		wc.status = status;
+		wc.opcode = ib_qib_wc_opcode[wqe->wr.opcode];
+		wc.qp = &qp->ibqp;
+		if (status == IB_WC_SUCCESS)
+			wc.byte_len = wqe->length;
+		qib_cq_enter(to_icq(qp->ibqp.send_cq), &wc,
+			     status != IB_WC_SUCCESS);
+	}
+
+	last = qp->s_last;
+	old_last = last;
+	if (++last >= qp->s_size)
+		last = 0;
+	qp->s_last = last;
+	if (qp->s_acked == old_last)
+		qp->s_acked = last;
+	if (qp->s_cur == old_last)
+		qp->s_cur = last;
+	if (qp->s_tail == old_last)
+		qp->s_tail = last;
+	if (qp->state == IB_QPS_SQD && last == qp->s_cur)
+		qp->s_draining = 0;
+}
diff --git a/drivers/infiniband/hw/ipath/ipath_sd7220.c b/drivers/infiniband/hw/qib/qib_sd7220.c
index 2a68d9f624dd..0aeed0e74cb6 100644
--- a/drivers/infiniband/hw/ipath/ipath_sd7220.c
+++ b/drivers/infiniband/hw/qib/qib_sd7220.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2006, 2007, 2008 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
  * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
  *
  * This software is available to you under a choice of one of two
@@ -32,22 +32,40 @@
  */
 /*
  * This file contains all of the code that is specific to the SerDes
- * on the InfiniPath 7220 chip.
+ * on the QLogic_IB 7220 chip.
  */
 
 #include <linux/pci.h>
 #include <linux/delay.h>
 
-#include "ipath_kernel.h"
-#include "ipath_registers.h"
-#include "ipath_7220.h"
+#include "qib.h"
+#include "qib_7220.h"
+
+/*
+ * Same as in qib_iba7220.c, but just the registers needed here.
+ * Could move whole set to qib_7220.h, but decided better to keep
+ * local.
+ */
+#define KREG_IDX(regname) (QIB_7220_##regname##_OFFS / sizeof(u64))
+#define kr_hwerrclear KREG_IDX(HwErrClear)
+#define kr_hwerrmask KREG_IDX(HwErrMask)
+#define kr_hwerrstatus KREG_IDX(HwErrStatus)
+#define kr_ibcstatus KREG_IDX(IBCStatus)
+#define kr_ibserdesctrl KREG_IDX(IBSerDesCtrl)
+#define kr_scratch KREG_IDX(Scratch)
+#define kr_xgxs_cfg KREG_IDX(XGXSCfg)
+/* these are used only here, not in qib_iba7220.c */
+#define kr_ibsd_epb_access_ctrl KREG_IDX(ibsd_epb_access_ctrl)
+#define kr_ibsd_epb_transaction_reg KREG_IDX(ibsd_epb_transaction_reg)
+#define kr_pciesd_epb_transaction_reg KREG_IDX(pciesd_epb_transaction_reg)
+#define kr_pciesd_epb_access_ctrl KREG_IDX(pciesd_epb_access_ctrl)
+#define kr_serdes_ddsrxeq0 KREG_IDX(SerDes_DDSRXEQ0)
 
 /*
  * The IBSerDesMappTable is a memory that holds values to be stored in
- * various SerDes registers by IBC. It is not part of the normal kregs
- * map and is used in exactly one place, hence the #define below.
+ * various SerDes registers by IBC.
  */
-#define KR_IBSerDesMappTable (0x94000 / (sizeof(uint64_t)))
+#define kr_serdes_maptable KREG_IDX(IBSerDesMappTable)
 
 /*
  * Below used for sdnum parameter, selecting one of the two sections
@@ -71,42 +89,37 @@
 #define EPB_GLOBAL_WR (1U << (EPB_ADDR_SHF + 8))
 
 /* Forward declarations. */
-static int ipath_sd7220_reg_mod(struct ipath_devdata *dd, int sdnum, u32 loc,
-				u32 data, u32 mask);
-static int ibsd_mod_allchnls(struct ipath_devdata *dd, int loc, int val,
+static int qib_sd7220_reg_mod(struct qib_devdata *dd, int sdnum, u32 loc,
+			      u32 data, u32 mask);
+static int ibsd_mod_allchnls(struct qib_devdata *dd, int loc, int val,
 			     int mask);
-static int ipath_sd_trimdone_poll(struct ipath_devdata *dd);
-static void ipath_sd_trimdone_monitor(struct ipath_devdata *dd,
-				      const char *where);
-static int ipath_sd_setvals(struct ipath_devdata *dd);
-static int ipath_sd_early(struct ipath_devdata *dd);
-static int ipath_sd_dactrim(struct ipath_devdata *dd);
-/* Set the registers that IBC may muck with to their default "preset" values */
-int ipath_sd7220_presets(struct ipath_devdata *dd);
-static int ipath_internal_presets(struct ipath_devdata *dd);
+static int qib_sd_trimdone_poll(struct qib_devdata *dd);
+static void qib_sd_trimdone_monitor(struct qib_devdata *dd, const char *where);
+static int qib_sd_setvals(struct qib_devdata *dd);
+static int qib_sd_early(struct qib_devdata *dd);
+static int qib_sd_dactrim(struct qib_devdata *dd);
+static int qib_internal_presets(struct qib_devdata *dd);
 /* Tweak the register (CMUCTRL5) that contains the TRIMSELF controls */
-static int ipath_sd_trimself(struct ipath_devdata *dd, int val);
-static int epb_access(struct ipath_devdata *dd, int sdnum, int claim);
-
-void ipath_set_relock_poll(struct ipath_devdata *dd, int ibup);
+static int qib_sd_trimself(struct qib_devdata *dd, int val);
+static int epb_access(struct qib_devdata *dd, int sdnum, int claim);
 
 /*
  * Below keeps track of whether the "once per power-on" initialization has
  * been done, because uC code Version 1.32.17 or higher allows the uC to
  * be reset at will, and Automatic Equalization may require it. So the
- * state of the reset "pin", as reflected in was_reset parameter to
- * ipath_sd7220_init() is no longer valid. Instead, we check for the
+ * state of the reset "pin", is no longer valid. Instead, we check for the
  * actual uC code having been loaded.
  */
-static int ipath_ibsd_ucode_loaded(struct ipath_devdata *dd)
+static int qib_ibsd_ucode_loaded(struct qib_pportdata *ppd)
 {
-	if (!dd->serdes_first_init_done && (ipath_sd7220_ib_vfy(dd) > 0))
-		dd->serdes_first_init_done = 1;
-	return dd->serdes_first_init_done;
+	struct qib_devdata *dd = ppd->dd;
+	if (!dd->cspec->serdes_first_init_done && (qib_sd7220_ib_vfy(dd) > 0))
+		dd->cspec->serdes_first_init_done = 1;
+	return dd->cspec->serdes_first_init_done;
 }
 
-/* repeat #define for local use. "Real" #define is in ipath_iba7220.c */
-#define INFINIPATH_HWE_IB_UC_MEMORYPARITYERR      0x0000004000000000ULL
+/* repeat #define for local use. "Real" #define is in qib_iba7220.c */
+#define QLOGIC_IB_HWE_IB_UC_MEMORYPARITYERR      0x0000004000000000ULL
 #define IB_MPREG5 (EPB_LOC(6, 0, 0xE) | (1L << EPB_IB_UC_CS_SHF))
 #define IB_MPREG6 (EPB_LOC(6, 0, 0xF) | (1U << EPB_IB_UC_CS_SHF))
 #define UC_PAR_CLR_D 8
@@ -114,25 +127,25 @@ static int ipath_ibsd_ucode_loaded(struct ipath_devdata *dd)
 #define IB_CTRL2(chn) (EPB_LOC(chn, 7, 3) | EPB_IB_QUAD0_CS)
 #define START_EQ1(chan) EPB_LOC(chan, 7, 0x27)
 
-void ipath_sd7220_clr_ibpar(struct ipath_devdata *dd)
+void qib_sd7220_clr_ibpar(struct qib_devdata *dd)
 {
 	int ret;
 
 	/* clear, then re-enable parity errs */
-	ret = ipath_sd7220_reg_mod(dd, IB_7220_SERDES, IB_MPREG6,
+	ret = qib_sd7220_reg_mod(dd, IB_7220_SERDES, IB_MPREG6,
 		UC_PAR_CLR_D, UC_PAR_CLR_M);
 	if (ret < 0) {
-		ipath_dev_err(dd, "Failed clearing IBSerDes Parity err\n");
+		qib_dev_err(dd, "Failed clearing IBSerDes Parity err\n");
 		goto bail;
 	}
-	ret = ipath_sd7220_reg_mod(dd, IB_7220_SERDES, IB_MPREG6, 0,
+	ret = qib_sd7220_reg_mod(dd, IB_7220_SERDES, IB_MPREG6, 0,
 		UC_PAR_CLR_M);
 
-	ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
+	qib_read_kreg32(dd, kr_scratch);
 	udelay(4);
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_hwerrclear,
-		INFINIPATH_HWE_IB_UC_MEMORYPARITYERR);
-	ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
+	qib_write_kreg(dd, kr_hwerrclear,
+		QLOGIC_IB_HWE_IB_UC_MEMORYPARITYERR);
+	qib_read_kreg32(dd, kr_scratch);
 bail:
 	return;
 }
@@ -146,7 +159,7 @@ bail:
 #define IB_PGUDP(chn) (EPB_LOC((chn), 2, 1) | EPB_IB_QUAD0_CS)
 #define IB_CMUDONE(chn) (EPB_LOC((chn), 7, 0xF) | EPB_IB_QUAD0_CS)
 
-static int ipath_resync_ibepb(struct ipath_devdata *dd)
+static int qib_resync_ibepb(struct qib_devdata *dd)
 {
 	int ret, pat, tries, chn;
 	u32 loc;
@@ -155,43 +168,42 @@ static int ipath_resync_ibepb(struct ipath_devdata *dd)
 	chn = 0;
 	for (tries = 0; tries < (4 * IBSD_RESYNC_TRIES); ++tries) {
 		loc = IB_PGUDP(chn);
-		ret = ipath_sd7220_reg_mod(dd, IB_7220_SERDES, loc, 0, 0);
+		ret = qib_sd7220_reg_mod(dd, IB_7220_SERDES, loc, 0, 0);
 		if (ret < 0) {
-			ipath_dev_err(dd, "Failed read in resync\n");
+			qib_dev_err(dd, "Failed read in resync\n");
 			continue;
 		}
 		if (ret != 0xF0 && ret != 0x55 && tries == 0)
-			ipath_dev_err(dd, "unexpected pattern in resync\n");
+			qib_dev_err(dd, "unexpected pattern in resync\n");
 		pat = ret ^ 0xA5; /* alternate F0 and 55 */
-		ret = ipath_sd7220_reg_mod(dd, IB_7220_SERDES, loc, pat, 0xFF);
+		ret = qib_sd7220_reg_mod(dd, IB_7220_SERDES, loc, pat, 0xFF);
 		if (ret < 0) {
-			ipath_dev_err(dd, "Failed write in resync\n");
+			qib_dev_err(dd, "Failed write in resync\n");
 			continue;
 		}
-		ret = ipath_sd7220_reg_mod(dd, IB_7220_SERDES, loc, 0, 0);
+		ret = qib_sd7220_reg_mod(dd, IB_7220_SERDES, loc, 0, 0);
 		if (ret < 0) {
-			ipath_dev_err(dd, "Failed re-read in resync\n");
+			qib_dev_err(dd, "Failed re-read in resync\n");
 			continue;
 		}
 		if (ret != pat) {
-			ipath_dev_err(dd, "Failed compare1 in resync\n");
+			qib_dev_err(dd, "Failed compare1 in resync\n");
 			continue;
 		}
 		loc = IB_CMUDONE(chn);
-		ret = ipath_sd7220_reg_mod(dd, IB_7220_SERDES, loc, 0, 0);
+		ret = qib_sd7220_reg_mod(dd, IB_7220_SERDES, loc, 0, 0);
 		if (ret < 0) {
-			ipath_dev_err(dd, "Failed CMUDONE rd in resync\n");
+			qib_dev_err(dd, "Failed CMUDONE rd in resync\n");
 			continue;
 		}
 		if ((ret & 0x70) != ((chn << 4) | 0x40)) {
-			ipath_dev_err(dd, "Bad CMUDONE value %02X, chn %d\n",
-				ret, chn);
+			qib_dev_err(dd, "Bad CMUDONE value %02X, chn %d\n",
+				    ret, chn);
 			continue;
 		}
 		if (++chn == 4)
 			break;  /* Success */
 	}
-	ipath_cdbg(VERBOSE, "Resync in %d tries\n", tries);
 	return (ret > 0) ? 0 : ret;
 }
 
@@ -199,32 +211,32 @@ static int ipath_resync_ibepb(struct ipath_devdata *dd)
  * Localize the stuff that should be done to change IB uC reset
  * returns <0 for errors.
  */
-static int ipath_ibsd_reset(struct ipath_devdata *dd, int assert_rst)
+static int qib_ibsd_reset(struct qib_devdata *dd, int assert_rst)
 {
 	u64 rst_val;
 	int ret = 0;
 	unsigned long flags;
 
-	rst_val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_ibserdesctrl);
+	rst_val = qib_read_kreg64(dd, kr_ibserdesctrl);
 	if (assert_rst) {
 		/*
 		 * Vendor recommends "interrupting" uC before reset, to
 		 * minimize possible glitches.
 		 */
-		spin_lock_irqsave(&dd->ipath_sdepb_lock, flags);
+		spin_lock_irqsave(&dd->cspec->sdepb_lock, flags);
 		epb_access(dd, IB_7220_SERDES, 1);
 		rst_val |= 1ULL;
 		/* Squelch possible parity error from _asserting_ reset */
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_hwerrmask,
-			dd->ipath_hwerrmask &
-			~INFINIPATH_HWE_IB_UC_MEMORYPARITYERR);
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_ibserdesctrl, rst_val);
+		qib_write_kreg(dd, kr_hwerrmask,
+			       dd->cspec->hwerrmask &
+			       ~QLOGIC_IB_HWE_IB_UC_MEMORYPARITYERR);
+		qib_write_kreg(dd, kr_ibserdesctrl, rst_val);
 		/* flush write, delay to ensure it took effect */
-		ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
+		qib_read_kreg32(dd, kr_scratch);
 		udelay(2);
 		/* once it's reset, can remove interrupt */
 		epb_access(dd, IB_7220_SERDES, -1);
-		spin_unlock_irqrestore(&dd->ipath_sdepb_lock, flags);
+		spin_unlock_irqrestore(&dd->cspec->sdepb_lock, flags);
 	} else {
 		/*
 		 * Before we de-assert reset, we need to deal with
@@ -235,46 +247,46 @@ static int ipath_ibsd_reset(struct ipath_devdata *dd, int assert_rst)
 		 */
 		u64 val;
 		rst_val &= ~(1ULL);
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_hwerrmask,
-			dd->ipath_hwerrmask &
-			~INFINIPATH_HWE_IB_UC_MEMORYPARITYERR);
+		qib_write_kreg(dd, kr_hwerrmask,
+			       dd->cspec->hwerrmask &
+			       ~QLOGIC_IB_HWE_IB_UC_MEMORYPARITYERR);
 
-		ret = ipath_resync_ibepb(dd);
+		ret = qib_resync_ibepb(dd);
 		if (ret < 0)
-			ipath_dev_err(dd, "unable to re-sync IB EPB\n");
+			qib_dev_err(dd, "unable to re-sync IB EPB\n");
 
 		/* set uC control regs to suppress parity errs */
-		ret = ipath_sd7220_reg_mod(dd, IB_7220_SERDES, IB_MPREG5, 1, 1);
+		ret = qib_sd7220_reg_mod(dd, IB_7220_SERDES, IB_MPREG5, 1, 1);
 		if (ret < 0)
 			goto bail;
 		/* IB uC code past Version 1.32.17 allow suppression of wdog */
-		ret = ipath_sd7220_reg_mod(dd, IB_7220_SERDES, IB_MPREG6, 0x80,
+		ret = qib_sd7220_reg_mod(dd, IB_7220_SERDES, IB_MPREG6, 0x80,
 			0x80);
 		if (ret < 0) {
-			ipath_dev_err(dd, "Failed to set WDOG disable\n");
+			qib_dev_err(dd, "Failed to set WDOG disable\n");
 			goto bail;
 		}
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_ibserdesctrl, rst_val);
+		qib_write_kreg(dd, kr_ibserdesctrl, rst_val);
 		/* flush write, delay for startup */
-		ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
+		qib_read_kreg32(dd, kr_scratch);
 		udelay(1);
 		/* clear, then re-enable parity errs */
-		ipath_sd7220_clr_ibpar(dd);
-		val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_hwerrstatus);
-		if (val & INFINIPATH_HWE_IB_UC_MEMORYPARITYERR) {
-			ipath_dev_err(dd, "IBUC Parity still set after RST\n");
-			dd->ipath_hwerrmask &=
-				~INFINIPATH_HWE_IB_UC_MEMORYPARITYERR;
+		qib_sd7220_clr_ibpar(dd);
+		val = qib_read_kreg64(dd, kr_hwerrstatus);
+		if (val & QLOGIC_IB_HWE_IB_UC_MEMORYPARITYERR) {
+			qib_dev_err(dd, "IBUC Parity still set after RST\n");
+			dd->cspec->hwerrmask &=
+				~QLOGIC_IB_HWE_IB_UC_MEMORYPARITYERR;
 		}
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_hwerrmask,
-			dd->ipath_hwerrmask);
+		qib_write_kreg(dd, kr_hwerrmask,
+			dd->cspec->hwerrmask);
 	}
 
 bail:
 	return ret;
 }
 
-static void ipath_sd_trimdone_monitor(struct ipath_devdata *dd,
+static void qib_sd_trimdone_monitor(struct qib_devdata *dd,
        const char *where)
 {
 	int ret, chn, baduns;
@@ -286,69 +298,71 @@ static void ipath_sd_trimdone_monitor(struct ipath_devdata *dd,
 	/* give time for reset to settle out in EPB */
 	udelay(2);
 
-	ret = ipath_resync_ibepb(dd);
+	ret = qib_resync_ibepb(dd);
 	if (ret < 0)
-		ipath_dev_err(dd, "not able to re-sync IB EPB (%s)\n", where);
+		qib_dev_err(dd, "not able to re-sync IB EPB (%s)\n", where);
 
 	/* Do "sacrificial read" to get EPB in sane state after reset */
-	ret = ipath_sd7220_reg_mod(dd, IB_7220_SERDES, IB_CTRL2(0), 0, 0);
+	ret = qib_sd7220_reg_mod(dd, IB_7220_SERDES, IB_CTRL2(0), 0, 0);
 	if (ret < 0)
-		ipath_dev_err(dd, "Failed TRIMDONE 1st read, (%s)\n", where);
+		qib_dev_err(dd, "Failed TRIMDONE 1st read, (%s)\n", where);
 
 	/* Check/show "summary" Trim-done bit in IBCStatus */
-	val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_ibcstatus);
-	if (val & (1ULL << 11))
-		ipath_cdbg(VERBOSE, "IBCS TRIMDONE set (%s)\n", where);
-	else
-		ipath_dev_err(dd, "IBCS TRIMDONE clear (%s)\n", where);
-
+	val = qib_read_kreg64(dd, kr_ibcstatus);
+	if (!(val & (1ULL << 11)))
+		qib_dev_err(dd, "IBCS TRIMDONE clear (%s)\n", where);
+	/*
+	 * Do "dummy read/mod/wr" to get EPB in sane state after reset
+	 * The default value for MPREG6 is 0.
+	 */
 	udelay(2);
 
-	ret = ipath_sd7220_reg_mod(dd, IB_7220_SERDES, IB_MPREG6, 0x80, 0x80);
+	ret = qib_sd7220_reg_mod(dd, IB_7220_SERDES, IB_MPREG6, 0x80, 0x80);
 	if (ret < 0)
-		ipath_dev_err(dd, "Failed Dummy RMW, (%s)\n", where);
+		qib_dev_err(dd, "Failed Dummy RMW, (%s)\n", where);
 	udelay(10);
 
 	baduns = 0;
 
 	for (chn = 3; chn >= 0; --chn) {
 		/* Read CTRL reg for each channel to check TRIMDONE */
-		ret = ipath_sd7220_reg_mod(dd, IB_7220_SERDES,
+		ret = qib_sd7220_reg_mod(dd, IB_7220_SERDES,
 			IB_CTRL2(chn), 0, 0);
 		if (ret < 0)
-			ipath_dev_err(dd, "Failed checking TRIMDONE, chn %d"
-				" (%s)\n", chn, where);
+			qib_dev_err(dd, "Failed checking TRIMDONE, chn %d"
+				    " (%s)\n", chn, where);
 
 		if (!(ret & 0x10)) {
 			int probe;
+
 			baduns |= (1 << chn);
-			ipath_dev_err(dd, "TRIMDONE cleared on chn %d (%02X)."
+			qib_dev_err(dd, "TRIMDONE cleared on chn %d (%02X)."
 				" (%s)\n", chn, ret, where);
-			probe = ipath_sd7220_reg_mod(dd, IB_7220_SERDES,
+			probe = qib_sd7220_reg_mod(dd, IB_7220_SERDES,
 				IB_PGUDP(0), 0, 0);
-			ipath_dev_err(dd, "probe is %d (%02X)\n",
+			qib_dev_err(dd, "probe is %d (%02X)\n",
 				probe, probe);
-			probe = ipath_sd7220_reg_mod(dd, IB_7220_SERDES,
+			probe = qib_sd7220_reg_mod(dd, IB_7220_SERDES,
 				IB_CTRL2(chn), 0, 0);
-			ipath_dev_err(dd, "re-read: %d (%02X)\n",
+			qib_dev_err(dd, "re-read: %d (%02X)\n",
 				probe, probe);
-			ret = ipath_sd7220_reg_mod(dd, IB_7220_SERDES,
+			ret = qib_sd7220_reg_mod(dd, IB_7220_SERDES,
 				IB_CTRL2(chn), 0x10, 0x10);
 			if (ret < 0)
-				ipath_dev_err(dd,
+				qib_dev_err(dd,
 					"Err on TRIMDONE rewrite1\n");
 		}
 	}
 	for (chn = 3; chn >= 0; --chn) {
 		/* Read CTRL reg for each channel to check TRIMDONE */
 		if (baduns & (1 << chn)) {
-			ipath_dev_err(dd,
+			qib_dev_err(dd,
 				"Reseting TRIMDONE on chn %d (%s)\n",
 				chn, where);
-			ret = ipath_sd7220_reg_mod(dd, IB_7220_SERDES,
+			ret = qib_sd7220_reg_mod(dd, IB_7220_SERDES,
 				IB_CTRL2(chn), 0x10, 0x10);
 			if (ret < 0)
-				ipath_dev_err(dd, "Failed re-setting "
+				qib_dev_err(dd, "Failed re-setting "
 					"TRIMDONE, chn %d (%s)\n",
 					chn, where);
 		}
@@ -361,96 +375,86 @@ static void ipath_sd_trimdone_monitor(struct ipath_devdata *dd,
  * Post IB uC code version 1.32.17, was_reset being 1 is not really
  * informative, so we double-check.
  */
-int ipath_sd7220_init(struct ipath_devdata *dd, int was_reset)
+int qib_sd7220_init(struct qib_devdata *dd)
 {
 	int ret = 1; /* default to failure */
-	int first_reset;
-	int val_stat;
+	int first_reset, was_reset;
 
+	/* SERDES MPU reset recorded in D0 */
+	was_reset = (qib_read_kreg64(dd, kr_ibserdesctrl) & 1);
 	if (!was_reset) {
 		/* entered with reset not asserted, we need to do it */
-		ipath_ibsd_reset(dd, 1);
-		ipath_sd_trimdone_monitor(dd, "Driver-reload");
+		qib_ibsd_reset(dd, 1);
+		qib_sd_trimdone_monitor(dd, "Driver-reload");
 	}
-
 	/* Substitute our deduced value for was_reset */
-	ret = ipath_ibsd_ucode_loaded(dd);
-	if (ret < 0) {
-		ret = 1;
-		goto done;
-	}
-	first_reset = !ret; /* First reset if IBSD uCode not yet loaded */
+	ret = qib_ibsd_ucode_loaded(dd->pport);
+	if (ret < 0)
+		goto bail;
 
+	first_reset = !ret; /* First reset if IBSD uCode not yet loaded */
 	/*
 	 * Alter some regs per vendor latest doc, reset-defaults
 	 * are not right for IB.
 	 */
-	ret = ipath_sd_early(dd);
+	ret = qib_sd_early(dd);
 	if (ret < 0) {
-		ipath_dev_err(dd, "Failed to set IB SERDES early defaults\n");
-		ret = 1;
-		goto done;
+		qib_dev_err(dd, "Failed to set IB SERDES early defaults\n");
+		goto bail;
 	}
-
 	/*
 	 * Set DAC manual trim IB.
 	 * We only do this once after chip has been reset (usually
 	 * same as once per system boot).
 	 */
 	if (first_reset) {
-		ret = ipath_sd_dactrim(dd);
+		ret = qib_sd_dactrim(dd);
 		if (ret < 0) {
-			ipath_dev_err(dd, "Failed IB SERDES DAC trim\n");
-			ret = 1;
-			goto done;
+			qib_dev_err(dd, "Failed IB SERDES DAC trim\n");
+			goto bail;
 		}
 	}
-
 	/*
 	 * Set various registers (DDS and RXEQ) that will be
 	 * controlled by IBC (in 1.2 mode) to reasonable preset values
 	 * Calling the "internal" version avoids the "check for needed"
 	 * and "trimdone monitor" that might be counter-productive.
 	 */
-	ret = ipath_internal_presets(dd);
+	ret = qib_internal_presets(dd);
 	if (ret < 0) {
-		ipath_dev_err(dd, "Failed to set IB SERDES presets\n");
-		ret = 1;
-		goto done;
+		qib_dev_err(dd, "Failed to set IB SERDES presets\n");
+		goto bail;
 	}
-	ret = ipath_sd_trimself(dd, 0x80);
+	ret = qib_sd_trimself(dd, 0x80);
 	if (ret < 0) {
-		ipath_dev_err(dd, "Failed to set IB SERDES TRIMSELF\n");
-		ret = 1;
-		goto done;
+		qib_dev_err(dd, "Failed to set IB SERDES TRIMSELF\n");
+		goto bail;
 	}
 
 	/* Load image, then try to verify */
-	ret = 0;	/* Assume success */
+	ret = 0;        /* Assume success */
 	if (first_reset) {
 		int vfy;
 		int trim_done;
-		ipath_dbg("SerDes uC was reset, reloading PRAM\n");
-		ret = ipath_sd7220_ib_load(dd);
+
+		ret = qib_sd7220_ib_load(dd);
 		if (ret < 0) {
-			ipath_dev_err(dd, "Failed to load IB SERDES image\n");
-			ret = 1;
-			goto done;
-		}
+			qib_dev_err(dd, "Failed to load IB SERDES image\n");
+			goto bail;
+		} else {
+			/* Loaded image, try to verify */
+			vfy = qib_sd7220_ib_vfy(dd);
+			if (vfy != ret) {
+				qib_dev_err(dd, "SERDES PRAM VFY failed\n");
+				goto bail;
+			} /* end if verified */
+		} /* end if loaded */
 
-		/* Loaded image, try to verify */
-		vfy = ipath_sd7220_ib_vfy(dd);
-		if (vfy != ret) {
-			ipath_dev_err(dd, "SERDES PRAM VFY failed\n");
-			ret = 1;
-			goto done;
-		}
 		/*
 		 * Loaded and verified. Almost good...
 		 * hold "success" in ret
 		 */
 		ret = 0;
-
 		/*
 		 * Prev steps all worked, continue bringup
 		 * De-assert RESET to uC, only in first reset, to allow
@@ -461,45 +465,47 @@ int ipath_sd7220_init(struct ipath_devdata *dd, int was_reset)
 		 */
 		ret = ibsd_mod_allchnls(dd, START_EQ1(0), 0, 0x38);
 		if (ret < 0) {
-			ipath_dev_err(dd, "Failed clearing START_EQ1\n");
-			ret = 1;
-			goto done;
+			qib_dev_err(dd, "Failed clearing START_EQ1\n");
+			goto bail;
 		}
 
-		ipath_ibsd_reset(dd, 0);
+		qib_ibsd_reset(dd, 0);
 		/*
 		 * If this is not the first reset, trimdone should be set
-		 * already.
+		 * already. We may need to check about this.
 		 */
-		trim_done = ipath_sd_trimdone_poll(dd);
+		trim_done = qib_sd_trimdone_poll(dd);
 		/*
 		 * Whether or not trimdone succeeded, we need to put the
 		 * uC back into reset to avoid a possible fight with the
 		 * IBC state-machine.
 		 */
-		ipath_ibsd_reset(dd, 1);
+		qib_ibsd_reset(dd, 1);
 
 		if (!trim_done) {
-			ipath_dev_err(dd, "No TRIMDONE seen\n");
-			ret = 1;
-			goto done;
+			qib_dev_err(dd, "No TRIMDONE seen\n");
+			goto bail;
 		}
-
-		ipath_sd_trimdone_monitor(dd, "First-reset");
+		/*
+		 * DEBUG: check each time we reset if trimdone bits have
+		 * gotten cleared, and re-set them.
+		 */
+		qib_sd_trimdone_monitor(dd, "First-reset");
 		/* Remember so we do not re-do the load, dactrim, etc. */
-		dd->serdes_first_init_done = 1;
+		dd->cspec->serdes_first_init_done = 1;
 	}
 	/*
-	 * Setup for channel training and load values for
+	 * setup for channel training and load values for
 	 * RxEq and DDS in tables used by IBC in IB1.2 mode
 	 */
-
-	val_stat = ipath_sd_setvals(dd);
-	if (val_stat < 0)
-		ret = 1;
+	ret = 0;
+	if (qib_sd_setvals(dd) >= 0)
+		goto done;
+bail:
+	ret = 1;
 done:
 	/* start relock timer regardless, but start at 1 second */
-	ipath_set_relock_poll(dd, -1);
+	set_7220_relock_poll(dd, -1);
 	return ret;
 }
 
@@ -517,7 +523,7 @@ done:
  * the "claim" parameter is >0 to claim, <0 to release, 0 to query.
  * Returns <0 for errors, >0 if we had ownership, else 0.
  */
-static int epb_access(struct ipath_devdata *dd, int sdnum, int claim)
+static int epb_access(struct qib_devdata *dd, int sdnum, int claim)
 {
 	u16 acc;
 	u64 accval;
@@ -525,28 +531,30 @@ static int epb_access(struct ipath_devdata *dd, int sdnum, int claim)
 	u64 oct_sel = 0;
 
 	switch (sdnum) {
-	case IB_7220_SERDES :
+	case IB_7220_SERDES:
 		/*
 		 * The IB SERDES "ownership" is fairly simple. A single each
 		 * request/grant.
 		 */
-		acc = dd->ipath_kregs->kr_ib_epbacc;
+		acc = kr_ibsd_epb_access_ctrl;
 		break;
-	case PCIE_SERDES0 :
-	case PCIE_SERDES1 :
+
+	case PCIE_SERDES0:
+	case PCIE_SERDES1:
 		/* PCIe SERDES has two "octants", need to select which */
-		acc = dd->ipath_kregs->kr_pcie_epbacc;
+		acc = kr_pciesd_epb_access_ctrl;
 		oct_sel = (2 << (sdnum - PCIE_SERDES0));
 		break;
-	default :
+
+	default:
 		return 0;
 	}
 
 	/* Make sure any outstanding transaction was seen */
-	ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
+	qib_read_kreg32(dd, kr_scratch);
 	udelay(15);
 
-	accval = ipath_read_kreg32(dd, acc);
+	accval = qib_read_kreg32(dd, acc);
 
 	owned = !!(accval & EPB_ACC_GNT);
 	if (claim < 0) {
@@ -557,22 +565,22 @@ static int epb_access(struct ipath_devdata *dd, int sdnum, int claim)
 		 * Both should be clear
 		 */
 		u64 newval = 0;
-		ipath_write_kreg(dd, acc, newval);
+		qib_write_kreg(dd, acc, newval);
 		/* First read after write is not trustworthy */
-		pollval = ipath_read_kreg32(dd, acc);
+		pollval = qib_read_kreg32(dd, acc);
 		udelay(5);
-		pollval = ipath_read_kreg32(dd, acc);
+		pollval = qib_read_kreg32(dd, acc);
 		if (pollval & EPB_ACC_GNT)
 			owned = -1;
 	} else if (claim > 0) {
 		/* Need to claim */
 		u64 pollval;
 		u64 newval = EPB_ACC_REQ | oct_sel;
-		ipath_write_kreg(dd, acc, newval);
+		qib_write_kreg(dd, acc, newval);
 		/* First read after write is not trustworthy */
-		pollval = ipath_read_kreg32(dd, acc);
+		pollval = qib_read_kreg32(dd, acc);
 		udelay(5);
-		pollval = ipath_read_kreg32(dd, acc);
+		pollval = qib_read_kreg32(dd, acc);
 		if (!(pollval & EPB_ACC_GNT))
 			owned = -1;
 	}
@@ -582,18 +590,17 @@ static int epb_access(struct ipath_devdata *dd, int sdnum, int claim)
 /*
  * Lemma to deal with race condition of write..read to epb regs
  */
-static int epb_trans(struct ipath_devdata *dd, u16 reg, u64 i_val, u64 *o_vp)
+static int epb_trans(struct qib_devdata *dd, u16 reg, u64 i_val, u64 *o_vp)
 {
 	int tries;
 	u64 transval;
 
-
-	ipath_write_kreg(dd, reg, i_val);
+	qib_write_kreg(dd, reg, i_val);
 	/* Throw away first read, as RDY bit may be stale */
-	transval = ipath_read_kreg64(dd, reg);
+	transval = qib_read_kreg64(dd, reg);
 
 	for (tries = EPB_TRANS_TRIES; tries; --tries) {
-		transval = ipath_read_kreg32(dd, reg);
+		transval = qib_read_kreg32(dd, reg);
 		if (transval & EPB_TRANS_RDY)
 			break;
 		udelay(5);
@@ -606,21 +613,20 @@ static int epb_trans(struct ipath_devdata *dd, u16 reg, u64 i_val, u64 *o_vp)
 }
 
 /**
- *
- * ipath_sd7220_reg_mod - modify SERDES register
- * @dd: the infinipath device
+ * qib_sd7220_reg_mod - modify SERDES register
+ * @dd: the qlogic_ib device
  * @sdnum: which SERDES to access
  * @loc: location - channel, element, register, as packed by EPB_LOC() macro.
  * @wd: Write Data - value to set in register
  * @mask: ones where data should be spliced into reg.
  *
- * Basic register read/modify/write, with un-needed accesses elided. That is,
+ * Basic register read/modify/write, with un-needed acesses elided. That is,
  * a mask of zero will prevent write, while a mask of 0xFF will prevent read.
  * returns current (presumed, if a write was done) contents of selected
  * register, or <0 if errors.
  */
-static int ipath_sd7220_reg_mod(struct ipath_devdata *dd, int sdnum, u32 loc,
-				u32 wd, u32 mask)
+static int qib_sd7220_reg_mod(struct qib_devdata *dd, int sdnum, u32 loc,
+			      u32 wd, u32 mask)
 {
 	u16 trans;
 	u64 transval;
@@ -629,14 +635,16 @@ static int ipath_sd7220_reg_mod(struct ipath_devdata *dd, int sdnum, u32 loc,
 	unsigned long flags;
 
 	switch (sdnum) {
-	case IB_7220_SERDES :
-		trans = dd->ipath_kregs->kr_ib_epbtrans;
+	case IB_7220_SERDES:
+		trans = kr_ibsd_epb_transaction_reg;
 		break;
-	case PCIE_SERDES0 :
-	case PCIE_SERDES1 :
-		trans = dd->ipath_kregs->kr_pcie_epbtrans;
+
+	case PCIE_SERDES0:
+	case PCIE_SERDES1:
+		trans = kr_pciesd_epb_transaction_reg;
 		break;
-	default :
+
+	default:
 		return -1;
 	}
 
@@ -644,23 +652,23 @@ static int ipath_sd7220_reg_mod(struct ipath_devdata *dd, int sdnum, u32 loc,
 	 * All access is locked in software (vs other host threads) and
 	 * hardware (vs uC access).
 	 */
-	spin_lock_irqsave(&dd->ipath_sdepb_lock, flags);
+	spin_lock_irqsave(&dd->cspec->sdepb_lock, flags);
 
 	owned = epb_access(dd, sdnum, 1);
 	if (owned < 0) {
-		spin_unlock_irqrestore(&dd->ipath_sdepb_lock, flags);
+		spin_unlock_irqrestore(&dd->cspec->sdepb_lock, flags);
 		return -1;
 	}
 	ret = 0;
 	for (tries = EPB_TRANS_TRIES; tries; --tries) {
-		transval = ipath_read_kreg32(dd, trans);
+		transval = qib_read_kreg32(dd, trans);
 		if (transval & EPB_TRANS_RDY)
 			break;
 		udelay(5);
 	}
 
 	if (tries > 0) {
-		tries = 1;	/* to make read-skip work */
+		tries = 1;      /* to make read-skip work */
 		if (mask != 0xFF) {
 			/*
 			 * Not a pure write, so need to read.
@@ -688,7 +696,7 @@ static int ipath_sd7220_reg_mod(struct ipath_devdata *dd, int sdnum, u32 loc,
 	else
 		ret = transval & EPB_DATA_MASK;
 
-	spin_unlock_irqrestore(&dd->ipath_sdepb_lock, flags);
+	spin_unlock_irqrestore(&dd->cspec->sdepb_lock, flags);
 	if (tries <= 0)
 		ret = -1;
 	return ret;
@@ -707,7 +715,7 @@ static int ipath_sd7220_reg_mod(struct ipath_devdata *dd, int sdnum, u32 loc,
 #define EPB_RAMDATA EPB_LOC(6, 0, 5)
 
 /* Transfer date to/from uC Program RAM of IB or PCIe SerDes */
-static int ipath_sd7220_ram_xfer(struct ipath_devdata *dd, int sdnum, u32 loc,
+static int qib_sd7220_ram_xfer(struct qib_devdata *dd, int sdnum, u32 loc,
 			       u8 *buf, int cnt, int rd_notwr)
 {
 	u16 trans;
@@ -723,29 +731,28 @@ static int ipath_sd7220_ram_xfer(struct ipath_devdata *dd, int sdnum, u32 loc,
 
 	/* Pick appropriate transaction reg and "Chip select" for this serdes */
 	switch (sdnum) {
-	case IB_7220_SERDES :
+	case IB_7220_SERDES:
 		csbit = 1ULL << EPB_IB_UC_CS_SHF;
-		trans = dd->ipath_kregs->kr_ib_epbtrans;
+		trans = kr_ibsd_epb_transaction_reg;
 		break;
-	case PCIE_SERDES0 :
-	case PCIE_SERDES1 :
+
+	case PCIE_SERDES0:
+	case PCIE_SERDES1:
 		/* PCIe SERDES has uC "chip select" in different bit, too */
 		csbit = 1ULL << EPB_PCIE_UC_CS_SHF;
-		trans = dd->ipath_kregs->kr_pcie_epbtrans;
+		trans = kr_pciesd_epb_transaction_reg;
 		break;
-	default :
+
+	default:
 		return -1;
 	}
 
 	op = rd_notwr ? "Rd" : "Wr";
-	spin_lock_irqsave(&dd->ipath_sdepb_lock, flags);
+	spin_lock_irqsave(&dd->cspec->sdepb_lock, flags);
 
 	owned = epb_access(dd, sdnum, 1);
 	if (owned < 0) {
-		spin_unlock_irqrestore(&dd->ipath_sdepb_lock, flags);
-		ipath_dbg("Could not get %s access to %s EPB: %X, loc %X\n",
-			op, (sdnum == IB_7220_SERDES) ? "IB" : "PCIe",
-			owned, loc);
+		spin_unlock_irqrestore(&dd->cspec->sdepb_lock, flags);
 		return -1;
 	}
 
@@ -758,16 +765,14 @@ static int ipath_sd7220_ram_xfer(struct ipath_devdata *dd, int sdnum, u32 loc,
 	 */
 	addr = loc & 0x1FFF;
 	for (tries = EPB_TRANS_TRIES; tries; --tries) {
-		transval = ipath_read_kreg32(dd, trans);
+		transval = qib_read_kreg32(dd, trans);
 		if (transval & EPB_TRANS_RDY)
 			break;
 		udelay(5);
 	}
 
 	sofar = 0;
-	if (tries <= 0)
-		ipath_dbg("No initial RDY on EPB access request\n");
-	else {
+	if (tries > 0) {
 		/*
 		 * Every "memory" access is doubly-indirect.
 		 * We set two bytes of address, then read/write
@@ -778,8 +783,6 @@ static int ipath_sd7220_ram_xfer(struct ipath_devdata *dd, int sdnum, u32 loc,
 		transval = csbit | EPB_UC_CTL |
 			(rd_notwr ? EPB_ROM_R : EPB_ROM_W);
 		tries = epb_trans(dd, trans, transval, &transval);
-		if (tries <= 0)
-			ipath_dbg("No EPB response to uC %s cmd\n", op);
 		while (tries > 0 && sofar < cnt) {
 			if (!sofar) {
 				/* Only set address at start of chunk */
@@ -787,18 +790,14 @@ static int ipath_sd7220_ram_xfer(struct ipath_devdata *dd, int sdnum, u32 loc,
 				transval = csbit | EPB_MADDRH | addrbyte;
 				tries = epb_trans(dd, trans, transval,
 						  &transval);
-				if (tries <= 0) {
-					ipath_dbg("No EPB response ADDRH\n");
+				if (tries <= 0)
 					break;
-				}
 				addrbyte = (addr + sofar) & 0xFF;
 				transval = csbit | EPB_MADDRL | addrbyte;
 				tries = epb_trans(dd, trans, transval,
 						 &transval);
-				if (tries <= 0) {
-					ipath_dbg("No EPB response ADDRL\n");
+				if (tries <= 0)
 					break;
-				}
 			}
 
 			if (rd_notwr)
@@ -806,10 +805,8 @@ static int ipath_sd7220_ram_xfer(struct ipath_devdata *dd, int sdnum, u32 loc,
 			else
 				transval = csbit | EPB_ROMDATA | buf[sofar];
 			tries = epb_trans(dd, trans, transval, &transval);
-			if (tries <= 0) {
-				ipath_dbg("No EPB response DATA\n");
+			if (tries <= 0)
 				break;
-			}
 			if (rd_notwr)
 				buf[sofar] = transval & EPB_DATA_MASK;
 			++sofar;
@@ -817,8 +814,6 @@ static int ipath_sd7220_ram_xfer(struct ipath_devdata *dd, int sdnum, u32 loc,
 		/* Finally, clear control-bit for Read or Write */
 		transval = csbit | EPB_UC_CTL;
 		tries = epb_trans(dd, trans, transval, &transval);
-		if (tries <= 0)
-			ipath_dbg("No EPB response to drop of uC %s cmd\n", op);
 	}
 
 	ret = sofar;
@@ -826,18 +821,16 @@ static int ipath_sd7220_ram_xfer(struct ipath_devdata *dd, int sdnum, u32 loc,
 	if (epb_access(dd, sdnum, -1) < 0)
 		ret = -1;
 
-	spin_unlock_irqrestore(&dd->ipath_sdepb_lock, flags);
-	if (tries <= 0) {
-		ipath_dbg("SERDES PRAM %s failed after %d bytes\n", op, sofar);
+	spin_unlock_irqrestore(&dd->cspec->sdepb_lock, flags);
+	if (tries <= 0)
 		ret = -1;
-	}
 	return ret;
 }
 
 #define PROG_CHUNK 64
 
-int ipath_sd7220_prog_ld(struct ipath_devdata *dd, int sdnum,
-	u8 *img, int len, int offset)
+int qib_sd7220_prog_ld(struct qib_devdata *dd, int sdnum,
+		       u8 *img, int len, int offset)
 {
 	int cnt, sofar, req;
 
@@ -846,7 +839,7 @@ int ipath_sd7220_prog_ld(struct ipath_devdata *dd, int sdnum,
 		req = len - sofar;
 		if (req > PROG_CHUNK)
 			req = PROG_CHUNK;
-		cnt = ipath_sd7220_ram_xfer(dd, sdnum, offset + sofar,
+		cnt = qib_sd7220_ram_xfer(dd, sdnum, offset + sofar,
 					  img + sofar, req, 0);
 		if (cnt < req) {
 			sofar = -1;
@@ -860,8 +853,8 @@ int ipath_sd7220_prog_ld(struct ipath_devdata *dd, int sdnum,
 #define VFY_CHUNK 64
 #define SD_PRAM_ERROR_LIMIT 42
 
-int ipath_sd7220_prog_vfy(struct ipath_devdata *dd, int sdnum,
-	const u8 *img, int len, int offset)
+int qib_sd7220_prog_vfy(struct qib_devdata *dd, int sdnum,
+			const u8 *img, int len, int offset)
 {
 	int cnt, sofar, req, idx, errors;
 	unsigned char readback[VFY_CHUNK];
@@ -872,7 +865,7 @@ int ipath_sd7220_prog_vfy(struct ipath_devdata *dd, int sdnum,
 		req = len - sofar;
 		if (req > VFY_CHUNK)
 			req = VFY_CHUNK;
-		cnt = ipath_sd7220_ram_xfer(dd, sdnum, sofar + offset,
+		cnt = qib_sd7220_ram_xfer(dd, sdnum, sofar + offset,
 					  readback, req, 1);
 		if (cnt < req) {
 			/* failed in read itself */
@@ -888,11 +881,13 @@ int ipath_sd7220_prog_vfy(struct ipath_devdata *dd, int sdnum,
 	return errors ? -errors : sofar;
 }
 
-/* IRQ not set up at this point in init, so we poll. */
+/*
+ * IRQ not set up at this point in init, so we poll.
+ */
 #define IB_SERDES_TRIM_DONE (1ULL << 11)
 #define TRIM_TMO (30)
 
-static int ipath_sd_trimdone_poll(struct ipath_devdata *dd)
+static int qib_sd_trimdone_poll(struct qib_devdata *dd)
 {
 	int trim_tmo, ret;
 	uint64_t val;
@@ -903,16 +898,15 @@ static int ipath_sd_trimdone_poll(struct ipath_devdata *dd)
 	 */
 	ret = 0;
 	for (trim_tmo = 0; trim_tmo < TRIM_TMO; ++trim_tmo) {
-		val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_ibcstatus);
+		val = qib_read_kreg64(dd, kr_ibcstatus);
 		if (val & IB_SERDES_TRIM_DONE) {
-			ipath_cdbg(VERBOSE, "TRIMDONE after %d\n", trim_tmo);
 			ret = 1;
 			break;
 		}
 		msleep(10);
 	}
 	if (trim_tmo >= TRIM_TMO) {
-		ipath_dev_err(dd, "No TRIMDONE in %d tries\n", trim_tmo);
+		qib_dev_err(dd, "No TRIMDONE in %d tries\n", trim_tmo);
 		ret = 0;
 	}
 	return ret;
@@ -964,8 +958,7 @@ static struct dds_init {
 };
 
 /*
- * Next, values related to Receive Equalization.
- * In comments, FDR (Full) is IB DDR, HDR (Half) is IB SDR
+ * Now the RXEQ section of the table.
  */
 /* Hardware packs an element number and register address thus: */
 #define RXEQ_INIT_RDESC(elt, addr) (((elt) & 0xF) | ((addr) << 4))
@@ -981,23 +974,23 @@ static struct dds_init {
 #define RXEQ_SDR_ZCNT 23
 
 static struct rxeq_init {
-	u16 rdesc;	/* in form used in SerDesDDSRXEQ */
+	u16 rdesc;      /* in form used in SerDesDDSRXEQ */
 	u8  rdata[4];
 } rxeq_init_vals[] = {
 	/* Set Rcv Eq. to Preset node */
 	RXEQ_VAL_ALL(7, 0x27, 0x10),
 	/* Set DFELTHFDR/HDR thresholds */
-	RXEQ_VAL(7, 8,    0, 0, 0, 0), /* FDR */
+	RXEQ_VAL(7, 8,    0, 0, 0, 0), /* FDR, was 0, 1, 2, 3 */
 	RXEQ_VAL(7, 0x21, 0, 0, 0, 0), /* HDR */
-	/* Set TLTHFDR/HDR threshold */
-	RXEQ_VAL(7, 9,    2, 2, 2, 2), /* FDR */
-	RXEQ_VAL(7, 0x23, 2, 2, 2, 2), /* HDR */
+	/* Set TLTHFDR/HDR theshold */
+	RXEQ_VAL(7, 9,    2, 2, 2, 2), /* FDR, was 0, 2, 4, 6 */
+	RXEQ_VAL(7, 0x23, 2, 2, 2, 2), /* HDR, was  0, 1, 2, 3 */
 	/* Set Preamp setting 2 (ZFR/ZCNT) */
-	RXEQ_VAL(7, 0x1B, 12, 12, 12, 12), /* FDR */
-	RXEQ_VAL(7, 0x1C, 12, 12, 12, 12), /* HDR */
+	RXEQ_VAL(7, 0x1B, 12, 12, 12, 12), /* FDR, was 12, 16, 20, 24 */
+	RXEQ_VAL(7, 0x1C, 12, 12, 12, 12), /* HDR, was 12, 16, 20, 24 */
 	/* Set Preamp DC gain and Setting 1 (GFR/GHR) */
-	RXEQ_VAL(7, 0x1E, 0x10, 0x10, 0x10, 0x10), /* FDR */
-	RXEQ_VAL(7, 0x1F, 0x10, 0x10, 0x10, 0x10), /* HDR */
+	RXEQ_VAL(7, 0x1E, 16, 16, 16, 16), /* FDR, was 16, 17, 18, 20 */
+	RXEQ_VAL(7, 0x1F, 16, 16, 16, 16), /* HDR, was 16, 17, 18, 20 */
 	/* Toggle RELOCK (in VCDL_CTRL0) to lock to data */
 	RXEQ_VAL_ALL(6, 6, 0x20), /* Set D5 High */
 	RXEQ_VAL_ALL(6, 6, 0), /* Set D5 Low */
@@ -1007,27 +1000,27 @@ static struct rxeq_init {
 #define DDS_ROWS (16)
 #define RXEQ_ROWS ARRAY_SIZE(rxeq_init_vals)
 
-static int ipath_sd_setvals(struct ipath_devdata *dd)
+static int qib_sd_setvals(struct qib_devdata *dd)
 {
 	int idx, midx;
-	int min_idx;	 /* Minimum index for this portion of table */
+	int min_idx;     /* Minimum index for this portion of table */
 	uint32_t dds_reg_map;
 	u64 __iomem *taddr, *iaddr;
 	uint64_t data;
 	uint64_t sdctl;
 
-	taddr = dd->ipath_kregbase + KR_IBSerDesMappTable;
-	iaddr = dd->ipath_kregbase + dd->ipath_kregs->kr_ib_ddsrxeq;
+	taddr = dd->kregbase + kr_serdes_maptable;
+	iaddr = dd->kregbase + kr_serdes_ddsrxeq0;
 
 	/*
 	 * Init the DDS section of the table.
 	 * Each "row" of the table provokes NUM_DDS_REG writes, to the
 	 * registers indicated in DDS_REG_MAP.
 	 */
-	sdctl = ipath_read_kreg64(dd, dd->ipath_kregs->kr_ibserdesctrl);
+	sdctl = qib_read_kreg64(dd, kr_ibserdesctrl);
 	sdctl = (sdctl & ~(0x1f << 8)) | (NUM_DDS_REGS << 8);
 	sdctl = (sdctl & ~(0x1f << 13)) | (RXEQ_ROWS << 13);
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_ibserdesctrl, sdctl);
+	qib_write_kreg(dd, kr_ibserdesctrl, sdctl);
 
 	/*
 	 * Iterate down table within loop for each register to store.
@@ -1037,21 +1030,21 @@ static int ipath_sd_setvals(struct ipath_devdata *dd)
 		data = ((dds_reg_map & 0xF) << 4) | TX_FAST_ELT;
 		writeq(data, iaddr + idx);
 		mmiowb();
-		ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
+		qib_read_kreg32(dd, kr_scratch);
 		dds_reg_map >>= 4;
 		for (midx = 0; midx < DDS_ROWS; ++midx) {
 			u64 __iomem *daddr = taddr + ((midx << 4) + idx);
 			data = dds_init_vals[midx].reg_vals[idx];
 			writeq(data, daddr);
 			mmiowb();
-			ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
+			qib_read_kreg32(dd, kr_scratch);
 		} /* End inner for (vals for this reg, each row) */
 	} /* end outer for (regs to be stored) */
 
 	/*
-	 * Init the RXEQ section of the table. As explained above the table
-	 * rxeq_init_vals[], this runs in a different order, as the pattern
-	 * of register references is more complex, but there are only
+	 * Init the RXEQ section of the table.
+	 * This runs in a different order, as the pattern of
+	 * register references is more complex, but there are only
 	 * four "data" values per register.
 	 */
 	min_idx = idx; /* RXEQ indices pick up where DDS left off */
@@ -1066,13 +1059,13 @@ static int ipath_sd_setvals(struct ipath_devdata *dd)
 		/* Store the next RXEQ register address */
 		writeq(rxeq_init_vals[idx].rdesc, iaddr + didx);
 		mmiowb();
-		ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
+		qib_read_kreg32(dd, kr_scratch);
 		/* Iterate through RXEQ values */
 		for (vidx = 0; vidx < 4; vidx++) {
 			data = rxeq_init_vals[idx].rdata[vidx];
 			writeq(data, taddr + (vidx << 6) + idx);
 			mmiowb();
-			ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
+			qib_read_kreg32(dd, kr_scratch);
 		}
 	} /* end outer for (Reg-writes for RXEQ) */
 	return 0;
@@ -1085,33 +1078,18 @@ static int ipath_sd_setvals(struct ipath_devdata *dd)
 #define VCDL_CTRL2(chan) EPB_LOC(chan, 6, 8)
 #define START_EQ2(chan) EPB_LOC(chan, 7, 0x28)
 
-static int ibsd_sto_noisy(struct ipath_devdata *dd, int loc, int val, int mask)
-{
-	int ret = -1;
-	int sloc; /* shifted loc, for messages */
-
-	loc |= (1U << EPB_IB_QUAD0_CS_SHF);
-	sloc = loc >> EPB_ADDR_SHF;
-
-	ret = ipath_sd7220_reg_mod(dd, IB_7220_SERDES, loc, val, mask);
-	if (ret < 0)
-		ipath_dev_err(dd, "Write failed: elt %d,"
-			" addr 0x%X, chnl %d, val 0x%02X, mask 0x%02X\n",
-			(sloc & 0xF), (sloc >> 9) & 0x3f, (sloc >> 4) & 7,
-			val & 0xFF, mask & 0xFF);
-	return ret;
-}
-
 /*
  * Repeat a "store" across all channels of the IB SerDes.
  * Although nominally it inherits the "read value" of the last
  * channel it modified, the only really useful return is <0 for
  * failure, >= 0 for success. The parameter 'loc' is assumed to
- * be the location for the channel-0 copy of the register to
- * be modified.
+ * be the location in some channel of the register to be modified
+ * The caller can specify use of the "gang write" option of EPB,
+ * in which case we use the specified channel data for any fields
+ * not explicitely written.
  */
-static int ibsd_mod_allchnls(struct ipath_devdata *dd, int loc, int val,
-	int mask)
+static int ibsd_mod_allchnls(struct qib_devdata *dd, int loc, int val,
+			     int mask)
 {
 	int ret = -1;
 	int chnl;
@@ -1126,24 +1104,27 @@ static int ibsd_mod_allchnls(struct ipath_devdata *dd, int loc, int val,
 		loc |= (1U << EPB_IB_QUAD0_CS_SHF);
 		chnl = (loc >> (4 + EPB_ADDR_SHF)) & 7;
 		if (mask != 0xFF) {
-			ret = ipath_sd7220_reg_mod(dd, IB_7220_SERDES,
-				loc & ~EPB_GLOBAL_WR, 0, 0);
+			ret = qib_sd7220_reg_mod(dd, IB_7220_SERDES,
+						 loc & ~EPB_GLOBAL_WR, 0, 0);
 			if (ret < 0) {
 				int sloc = loc >> EPB_ADDR_SHF;
-				ipath_dev_err(dd, "pre-read failed: elt %d,"
-					" addr 0x%X, chnl %d\n", (sloc & 0xF),
-					(sloc >> 9) & 0x3f, chnl);
+
+				qib_dev_err(dd, "pre-read failed: elt %d,"
+					    " addr 0x%X, chnl %d\n",
+					    (sloc & 0xF),
+					    (sloc >> 9) & 0x3f, chnl);
 				return ret;
 			}
 			val = (ret & ~mask) | (val & mask);
 		}
 		loc &=  ~(7 << (4+EPB_ADDR_SHF));
-		ret = ipath_sd7220_reg_mod(dd, IB_7220_SERDES, loc, val, 0xFF);
+		ret = qib_sd7220_reg_mod(dd, IB_7220_SERDES, loc, val, 0xFF);
 		if (ret < 0) {
 			int sloc = loc >> EPB_ADDR_SHF;
-			ipath_dev_err(dd, "Global WR failed: elt %d,"
-				" addr 0x%X, val %02X\n",
-				(sloc & 0xF), (sloc >> 9) & 0x3f, val);
+
+			qib_dev_err(dd, "Global WR failed: elt %d,"
+				    " addr 0x%X, val %02X\n",
+				    (sloc & 0xF), (sloc >> 9) & 0x3f, val);
 		}
 		return ret;
 	}
@@ -1151,16 +1132,17 @@ static int ibsd_mod_allchnls(struct ipath_devdata *dd, int loc, int val,
 	loc &=  ~(7 << (4+EPB_ADDR_SHF));
 	loc |= (1U << EPB_IB_QUAD0_CS_SHF);
 	for (chnl = 0; chnl < 4; ++chnl) {
-		int cloc;
-		cloc = loc | (chnl << (4+EPB_ADDR_SHF));
-		ret = ipath_sd7220_reg_mod(dd, IB_7220_SERDES, cloc, val, mask);
+		int cloc = loc | (chnl << (4+EPB_ADDR_SHF));
+
+		ret = qib_sd7220_reg_mod(dd, IB_7220_SERDES, cloc, val, mask);
 		if (ret < 0) {
 			int sloc = loc >> EPB_ADDR_SHF;
-			ipath_dev_err(dd, "Write failed: elt %d,"
-				" addr 0x%X, chnl %d, val 0x%02X,"
-				" mask 0x%02X\n",
-				(sloc & 0xF), (sloc >> 9) & 0x3f, chnl,
-				val & 0xFF, mask & 0xFF);
+
+			qib_dev_err(dd, "Write failed: elt %d,"
+				    " addr 0x%X, chnl %d, val 0x%02X,"
+				    " mask 0x%02X\n",
+				    (sloc & 0xF), (sloc >> 9) & 0x3f, chnl,
+				    val & 0xFF, mask & 0xFF);
 			break;
 		}
 	}
@@ -1171,7 +1153,7 @@ static int ibsd_mod_allchnls(struct ipath_devdata *dd, int loc, int val,
  * Set the Tx values normally modified by IBC in IB1.2 mode to default
  * values, as gotten from first row of init table.
  */
-static int set_dds_vals(struct ipath_devdata *dd, struct dds_init *ddi)
+static int set_dds_vals(struct qib_devdata *dd, struct dds_init *ddi)
 {
 	int ret;
 	int idx, reg, data;
@@ -1194,7 +1176,7 @@ static int set_dds_vals(struct ipath_devdata *dd, struct dds_init *ddi)
  * Set the Rx values normally modified by IBC in IB1.2 mode to default
  * values, as gotten from selected column of init table.
  */
-static int set_rxeq_vals(struct ipath_devdata *dd, int vsel)
+static int set_rxeq_vals(struct qib_devdata *dd, int vsel)
 {
 	int ret;
 	int ridx;
@@ -1202,6 +1184,7 @@ static int set_rxeq_vals(struct ipath_devdata *dd, int vsel)
 
 	for (ridx = 0; ridx < cnt; ++ridx) {
 		int elt, reg, val, loc;
+
 		elt = rxeq_init_vals[ridx].rdesc & 0xF;
 		reg = rxeq_init_vals[ridx].rdesc >> 4;
 		loc = EPB_LOC(0, elt, reg);
@@ -1217,83 +1200,66 @@ static int set_rxeq_vals(struct ipath_devdata *dd, int vsel)
 /*
  * Set the default values (row 0) for DDR Driver Demphasis.
  * we do this initially and whenever we turn off IB-1.2
+ *
  * The "default" values for Rx equalization are also stored to
  * SerDes registers. Formerly (and still default), we used set 2.
  * For experimenting with cables and link-partners, we allow changing
  * that via a module parameter.
  */
-static unsigned ipath_rxeq_set = 2;
-module_param_named(rxeq_default_set, ipath_rxeq_set, uint,
-	S_IWUSR | S_IRUGO);
+static unsigned qib_rxeq_set = 2;
+module_param_named(rxeq_default_set, qib_rxeq_set, uint,
+		   S_IWUSR | S_IRUGO);
 MODULE_PARM_DESC(rxeq_default_set,
-	"Which set [0..3] of Rx Equalization values is default");
+		 "Which set [0..3] of Rx Equalization values is default");
 
-static int ipath_internal_presets(struct ipath_devdata *dd)
+static int qib_internal_presets(struct qib_devdata *dd)
 {
 	int ret = 0;
 
 	ret = set_dds_vals(dd, dds_init_vals + DDS_3M);
 
 	if (ret < 0)
-		ipath_dev_err(dd, "Failed to set default DDS values\n");
-	ret = set_rxeq_vals(dd, ipath_rxeq_set & 3);
+		qib_dev_err(dd, "Failed to set default DDS values\n");
+	ret = set_rxeq_vals(dd, qib_rxeq_set & 3);
 	if (ret < 0)
-		ipath_dev_err(dd, "Failed to set default RXEQ values\n");
+		qib_dev_err(dd, "Failed to set default RXEQ values\n");
 	return ret;
 }
 
-int ipath_sd7220_presets(struct ipath_devdata *dd)
+int qib_sd7220_presets(struct qib_devdata *dd)
 {
 	int ret = 0;
 
-	if (!dd->ipath_presets_needed)
+	if (!dd->cspec->presets_needed)
 		return ret;
-	dd->ipath_presets_needed = 0;
+	dd->cspec->presets_needed = 0;
 	/* Assert uC reset, so we don't clash with it. */
-	ipath_ibsd_reset(dd, 1);
+	qib_ibsd_reset(dd, 1);
 	udelay(2);
-	ipath_sd_trimdone_monitor(dd, "link-down");
+	qib_sd_trimdone_monitor(dd, "link-down");
 
-	ret = ipath_internal_presets(dd);
-return ret;
+	ret = qib_internal_presets(dd);
+	return ret;
 }
 
-static int ipath_sd_trimself(struct ipath_devdata *dd, int val)
+static int qib_sd_trimself(struct qib_devdata *dd, int val)
 {
-	return ibsd_sto_noisy(dd, CMUCTRL5, val, 0xFF);
+	int loc = CMUCTRL5 | (1U << EPB_IB_QUAD0_CS_SHF);
+
+	return qib_sd7220_reg_mod(dd, IB_7220_SERDES, loc, val, 0xFF);
 }
 
-static int ipath_sd_early(struct ipath_devdata *dd)
+static int qib_sd_early(struct qib_devdata *dd)
 {
-	int ret = -1; /* Default failed */
-	int chnl;
+	int ret;
 
-	for (chnl = 0; chnl < 4; ++chnl) {
-		ret = ibsd_sto_noisy(dd, RXHSCTRL0(chnl), 0xD4, 0xFF);
-		if (ret < 0)
-			goto bail;
-	}
-	for (chnl = 0; chnl < 4; ++chnl) {
-		ret = ibsd_sto_noisy(dd, VCDL_DAC2(chnl), 0x2D, 0xFF);
-		if (ret < 0)
-			goto bail;
-	}
-	/* more fine-tuning of what will be default */
-	for (chnl = 0; chnl < 4; ++chnl) {
-		ret = ibsd_sto_noisy(dd, VCDL_CTRL2(chnl), 3, 0xF);
-		if (ret < 0)
-			goto bail;
-	}
-	for (chnl = 0; chnl < 4; ++chnl) {
-		ret = ibsd_sto_noisy(dd, START_EQ1(chnl), 0x10, 0xFF);
-		if (ret < 0)
-			goto bail;
-	}
-	for (chnl = 0; chnl < 4; ++chnl) {
-		ret = ibsd_sto_noisy(dd, START_EQ2(chnl), 0x30, 0xFF);
-		if (ret < 0)
-			goto bail;
-	}
+	ret = ibsd_mod_allchnls(dd, RXHSCTRL0(0) | EPB_GLOBAL_WR, 0xD4, 0xFF);
+	if (ret < 0)
+		goto bail;
+	ret = ibsd_mod_allchnls(dd, START_EQ1(0) | EPB_GLOBAL_WR, 0x10, 0xFF);
+	if (ret < 0)
+		goto bail;
+	ret = ibsd_mod_allchnls(dd, START_EQ2(0) | EPB_GLOBAL_WR, 0x30, 0xFF);
 bail:
 	return ret;
 }
@@ -1302,50 +1268,53 @@ bail:
 #define LDOUTCTRL1(chnl) EPB_LOC(chnl, 7, 6)
 #define RXHSSTATUS(chnl) EPB_LOC(chnl, 6, 0xF)
 
-static int ipath_sd_dactrim(struct ipath_devdata *dd)
+static int qib_sd_dactrim(struct qib_devdata *dd)
 {
-	int ret = -1; /* Default failed */
-	int chnl;
+	int ret;
+
+	ret = ibsd_mod_allchnls(dd, VCDL_DAC2(0) | EPB_GLOBAL_WR, 0x2D, 0xFF);
+	if (ret < 0)
+		goto bail;
+
+	/* more fine-tuning of what will be default */
+	ret = ibsd_mod_allchnls(dd, VCDL_CTRL2(0), 3, 0xF);
+	if (ret < 0)
+		goto bail;
+
+	ret = ibsd_mod_allchnls(dd, BACTRL(0) | EPB_GLOBAL_WR, 0x40, 0xFF);
+	if (ret < 0)
+		goto bail;
+
+	ret = ibsd_mod_allchnls(dd, LDOUTCTRL1(0) | EPB_GLOBAL_WR, 0x04, 0xFF);
+	if (ret < 0)
+		goto bail;
+
+	ret = ibsd_mod_allchnls(dd, RXHSSTATUS(0) | EPB_GLOBAL_WR, 0x04, 0xFF);
+	if (ret < 0)
+		goto bail;
 
-	for (chnl = 0; chnl < 4; ++chnl) {
-		ret = ibsd_sto_noisy(dd, BACTRL(chnl), 0x40, 0xFF);
-		if (ret < 0)
-			goto bail;
-	}
-	for (chnl = 0; chnl < 4; ++chnl) {
-		ret = ibsd_sto_noisy(dd, LDOUTCTRL1(chnl), 0x04, 0xFF);
-		if (ret < 0)
-			goto bail;
-	}
-	for (chnl = 0; chnl < 4; ++chnl) {
-		ret = ibsd_sto_noisy(dd, RXHSSTATUS(chnl), 0x04, 0xFF);
-		if (ret < 0)
-			goto bail;
-	}
 	/*
-	 * delay for max possible number of steps, with slop.
+	 * Delay for max possible number of steps, with slop.
 	 * Each step is about 4usec.
 	 */
 	udelay(415);
-	for (chnl = 0; chnl < 4; ++chnl) {
-		ret = ibsd_sto_noisy(dd, LDOUTCTRL1(chnl), 0x00, 0xFF);
-		if (ret < 0)
-			goto bail;
-	}
+
+	ret = ibsd_mod_allchnls(dd, LDOUTCTRL1(0) | EPB_GLOBAL_WR, 0x00, 0xFF);
+
 bail:
 	return ret;
 }
 
 #define RELOCK_FIRST_MS 3
 #define RXLSPPM(chan) EPB_LOC(chan, 0, 2)
-void ipath_toggle_rclkrls(struct ipath_devdata *dd)
+void toggle_7220_rclkrls(struct qib_devdata *dd)
 {
 	int loc = RXLSPPM(0) | EPB_GLOBAL_WR;
 	int ret;
 
 	ret = ibsd_mod_allchnls(dd, loc, 0, 0x80);
 	if (ret < 0)
-		ipath_dev_err(dd, "RCLKRLS failed to clear D7\n");
+		qib_dev_err(dd, "RCLKRLS failed to clear D7\n");
 	else {
 		udelay(1);
 		ibsd_mod_allchnls(dd, loc, 0x80, 0x80);
@@ -1354,109 +1323,91 @@ void ipath_toggle_rclkrls(struct ipath_devdata *dd)
 	udelay(1);
 	ret = ibsd_mod_allchnls(dd, loc, 0, 0x80);
 	if (ret < 0)
-		ipath_dev_err(dd, "RCLKRLS failed to clear D7\n");
+		qib_dev_err(dd, "RCLKRLS failed to clear D7\n");
 	else {
 		udelay(1);
 		ibsd_mod_allchnls(dd, loc, 0x80, 0x80);
 	}
 	/* Now reset xgxs and IBC to complete the recovery */
-	dd->ipath_f_xgxs_reset(dd);
+	dd->f_xgxs_reset(dd->pport);
 }
 
 /*
  * Shut down the timer that polls for relock occasions, if needed
- * this is "hooked" from ipath_7220_quiet_serdes(), which is called
- * just before ipath_shutdown_device() in ipath_driver.c shuts down all
+ * this is "hooked" from qib_7220_quiet_serdes(), which is called
+ * just before qib_shutdown_device() in qib_driver.c shuts down all
  * the other timers
  */
-void ipath_shutdown_relock_poll(struct ipath_devdata *dd)
+void shutdown_7220_relock_poll(struct qib_devdata *dd)
 {
-	struct ipath_relock *irp = &dd->ipath_relock_singleton;
-	if (atomic_read(&irp->ipath_relock_timer_active)) {
-		del_timer_sync(&irp->ipath_relock_timer);
-		atomic_set(&irp->ipath_relock_timer_active, 0);
-	}
+	if (dd->cspec->relock_timer_active)
+		del_timer_sync(&dd->cspec->relock_timer);
 }
 
-static unsigned ipath_relock_by_timer = 1;
-module_param_named(relock_by_timer, ipath_relock_by_timer, uint,
-	S_IWUSR | S_IRUGO);
+static unsigned qib_relock_by_timer = 1;
+module_param_named(relock_by_timer, qib_relock_by_timer, uint,
+		   S_IWUSR | S_IRUGO);
 MODULE_PARM_DESC(relock_by_timer, "Allow relock attempt if link not up");
 
-static void ipath_run_relock(unsigned long opaque)
+static void qib_run_relock(unsigned long opaque)
 {
-	struct ipath_devdata *dd = (struct ipath_devdata *)opaque;
-	struct ipath_relock *irp = &dd->ipath_relock_singleton;
-	u64 val, ltstate;
-
-	if (!(dd->ipath_flags & IPATH_INITTED)) {
-		/* Not yet up, just reenable the timer for later */
-		irp->ipath_relock_interval = HZ;
-		mod_timer(&irp->ipath_relock_timer, jiffies + HZ);
-		return;
-	}
+	struct qib_devdata *dd = (struct qib_devdata *)opaque;
+	struct qib_pportdata *ppd = dd->pport;
+	struct qib_chip_specific *cs = dd->cspec;
+	int timeoff;
 
 	/*
-	 * Check link-training state for "stuck" state.
+	 * Check link-training state for "stuck" state, when down.
 	 * if found, try relock and schedule another try at
 	 * exponentially growing delay, maxed at one second.
 	 * if not stuck, our work is done.
 	 */
-	val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_ibcstatus);
-	ltstate = ipath_ib_linktrstate(dd, val);
-
-	if (ltstate <= INFINIPATH_IBCS_LT_STATE_CFGWAITRMT
-		&& ltstate != INFINIPATH_IBCS_LT_STATE_LINKUP) {
-		int timeoff;
-		/* Not up yet. Try again, if allowed by module-param */
-		if (ipath_relock_by_timer) {
-			if (dd->ipath_flags & IPATH_IB_AUTONEG_INPROG)
-				ipath_cdbg(VERBOSE, "Skip RELOCK in AUTONEG\n");
-			else if (!(dd->ipath_flags & IPATH_IB_LINK_DISABLED)) {
-				ipath_cdbg(VERBOSE, "RELOCK\n");
-				ipath_toggle_rclkrls(dd);
-			}
+	if ((dd->flags & QIB_INITTED) && !(ppd->lflags &
+	    (QIBL_IB_AUTONEG_INPROG | QIBL_LINKINIT | QIBL_LINKARMED |
+	     QIBL_LINKACTIVE))) {
+		if (qib_relock_by_timer) {
+			if (!(ppd->lflags & QIBL_IB_LINK_DISABLED))
+				toggle_7220_rclkrls(dd);
 		}
 		/* re-set timer for next check */
-		timeoff = irp->ipath_relock_interval << 1;
+		timeoff = cs->relock_interval << 1;
 		if (timeoff > HZ)
 			timeoff = HZ;
-		irp->ipath_relock_interval = timeoff;
-
-		mod_timer(&irp->ipath_relock_timer, jiffies + timeoff);
-	} else {
-		/* Up, so no more need to check so often */
-		mod_timer(&irp->ipath_relock_timer, jiffies + HZ);
-	}
+		cs->relock_interval = timeoff;
+	} else
+		timeoff = HZ;
+	mod_timer(&cs->relock_timer, jiffies + timeoff);
 }
 
-void ipath_set_relock_poll(struct ipath_devdata *dd, int ibup)
+void set_7220_relock_poll(struct qib_devdata *dd, int ibup)
 {
-	struct ipath_relock *irp = &dd->ipath_relock_singleton;
+	struct qib_chip_specific *cs = dd->cspec;
 
-	if (ibup > 0) {
-		/* we are now up, so relax timer to 1 second interval */
-		if (atomic_read(&irp->ipath_relock_timer_active))
-			mod_timer(&irp->ipath_relock_timer, jiffies + HZ);
+	if (ibup) {
+		/* We are now up, relax timer to 1 second interval */
+		if (cs->relock_timer_active) {
+			cs->relock_interval = HZ;
+			mod_timer(&cs->relock_timer, jiffies + HZ);
+		}
 	} else {
 		/* Transition to down, (re-)set timer to short interval. */
-		int timeout;
-		timeout = (HZ * ((ibup == -1) ? 1000 : RELOCK_FIRST_MS))/1000;
+		unsigned int timeout;
+
+		timeout = msecs_to_jiffies(RELOCK_FIRST_MS);
 		if (timeout == 0)
 			timeout = 1;
 		/* If timer has not yet been started, do so. */
-		if (atomic_inc_return(&irp->ipath_relock_timer_active) == 1) {
-			init_timer(&irp->ipath_relock_timer);
-			irp->ipath_relock_timer.function = ipath_run_relock;
-			irp->ipath_relock_timer.data = (unsigned long) dd;
-			irp->ipath_relock_interval = timeout;
-			irp->ipath_relock_timer.expires = jiffies + timeout;
-			add_timer(&irp->ipath_relock_timer);
+		if (!cs->relock_timer_active) {
+			cs->relock_timer_active = 1;
+			init_timer(&cs->relock_timer);
+			cs->relock_timer.function = qib_run_relock;
+			cs->relock_timer.data = (unsigned long) dd;
+			cs->relock_interval = timeout;
+			cs->relock_timer.expires = jiffies + timeout;
+			add_timer(&cs->relock_timer);
 		} else {
-			irp->ipath_relock_interval = timeout;
-			mod_timer(&irp->ipath_relock_timer, jiffies + timeout);
-			atomic_dec(&irp->ipath_relock_timer_active);
+			cs->relock_interval = timeout;
+			mod_timer(&cs->relock_timer, jiffies + timeout);
 		}
 	}
 }
-
diff --git a/drivers/infiniband/hw/ipath/ipath_sd7220_img.c b/drivers/infiniband/hw/qib/qib_sd7220_img.c
index 5ef59da9270a..a1118fbd2370 100644
--- a/drivers/infiniband/hw/ipath/ipath_sd7220_img.c
+++ b/drivers/infiniband/hw/qib/qib_sd7220_img.c
@@ -38,11 +38,10 @@
 #include <linux/pci.h>
 #include <linux/delay.h>
 
-#include "ipath_kernel.h"
-#include "ipath_registers.h"
-#include "ipath_7220.h"
+#include "qib.h"
+#include "qib_7220.h"
 
-static unsigned char ipath_sd7220_ib_img[] = {
+static unsigned char qib_sd7220_ib_img[] = {
 /*0000*/0x02, 0x0A, 0x29, 0x02, 0x0A, 0x87, 0xE5, 0xE6,
 	0x30, 0xE6, 0x04, 0x7F, 0x01, 0x80, 0x02, 0x7F,
 /*0010*/0x00, 0xE5, 0xE2, 0x30, 0xE4, 0x04, 0x7E, 0x01,
@@ -1069,14 +1068,14 @@ static unsigned char ipath_sd7220_ib_img[] = {
 	0x01, 0x20, 0x11, 0x00, 0x04, 0x20, 0x00, 0x81
 };
 
-int ipath_sd7220_ib_load(struct ipath_devdata *dd)
+int qib_sd7220_ib_load(struct qib_devdata *dd)
 {
-	return ipath_sd7220_prog_ld(dd, IB_7220_SERDES, ipath_sd7220_ib_img,
-		sizeof(ipath_sd7220_ib_img), 0);
+	return qib_sd7220_prog_ld(dd, IB_7220_SERDES, qib_sd7220_ib_img,
+		sizeof(qib_sd7220_ib_img), 0);
 }
 
-int ipath_sd7220_ib_vfy(struct ipath_devdata *dd)
+int qib_sd7220_ib_vfy(struct qib_devdata *dd)
 {
-	return ipath_sd7220_prog_vfy(dd, IB_7220_SERDES, ipath_sd7220_ib_img,
-		sizeof(ipath_sd7220_ib_img), 0);
+	return qib_sd7220_prog_vfy(dd, IB_7220_SERDES, qib_sd7220_ib_img,
+		sizeof(qib_sd7220_ib_img), 0);
 }
diff --git a/drivers/infiniband/hw/qib/qib_sdma.c b/drivers/infiniband/hw/qib/qib_sdma.c
new file mode 100644
index 000000000000..b8456881f7f6
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_sdma.c
@@ -0,0 +1,973 @@
+/*
+ * Copyright (c) 2007, 2008, 2009, 2010 QLogic Corporation. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/spinlock.h>
+#include <linux/netdevice.h>
+
+#include "qib.h"
+#include "qib_common.h"
+
+/* default pio off, sdma on */
+static ushort sdma_descq_cnt = 256;
+module_param_named(sdma_descq_cnt, sdma_descq_cnt, ushort, S_IRUGO);
+MODULE_PARM_DESC(sdma_descq_cnt, "Number of SDMA descq entries");
+
+/*
+ * Bits defined in the send DMA descriptor.
+ */
+#define SDMA_DESC_LAST          (1ULL << 11)
+#define SDMA_DESC_FIRST         (1ULL << 12)
+#define SDMA_DESC_DMA_HEAD      (1ULL << 13)
+#define SDMA_DESC_USE_LARGE_BUF (1ULL << 14)
+#define SDMA_DESC_INTR          (1ULL << 15)
+#define SDMA_DESC_COUNT_LSB     16
+#define SDMA_DESC_GEN_LSB       30
+
+char *qib_sdma_state_names[] = {
+	[qib_sdma_state_s00_hw_down]          = "s00_HwDown",
+	[qib_sdma_state_s10_hw_start_up_wait] = "s10_HwStartUpWait",
+	[qib_sdma_state_s20_idle]             = "s20_Idle",
+	[qib_sdma_state_s30_sw_clean_up_wait] = "s30_SwCleanUpWait",
+	[qib_sdma_state_s40_hw_clean_up_wait] = "s40_HwCleanUpWait",
+	[qib_sdma_state_s50_hw_halt_wait]     = "s50_HwHaltWait",
+	[qib_sdma_state_s99_running]          = "s99_Running",
+};
+
+char *qib_sdma_event_names[] = {
+	[qib_sdma_event_e00_go_hw_down]   = "e00_GoHwDown",
+	[qib_sdma_event_e10_go_hw_start]  = "e10_GoHwStart",
+	[qib_sdma_event_e20_hw_started]   = "e20_HwStarted",
+	[qib_sdma_event_e30_go_running]   = "e30_GoRunning",
+	[qib_sdma_event_e40_sw_cleaned]   = "e40_SwCleaned",
+	[qib_sdma_event_e50_hw_cleaned]   = "e50_HwCleaned",
+	[qib_sdma_event_e60_hw_halted]    = "e60_HwHalted",
+	[qib_sdma_event_e70_go_idle]      = "e70_GoIdle",
+	[qib_sdma_event_e7220_err_halted] = "e7220_ErrHalted",
+	[qib_sdma_event_e7322_err_halted] = "e7322_ErrHalted",
+	[qib_sdma_event_e90_timer_tick]   = "e90_TimerTick",
+};
+
+/* declare all statics here rather than keep sorting */
+static int alloc_sdma(struct qib_pportdata *);
+static void sdma_complete(struct kref *);
+static void sdma_finalput(struct qib_sdma_state *);
+static void sdma_get(struct qib_sdma_state *);
+static void sdma_put(struct qib_sdma_state *);
+static void sdma_set_state(struct qib_pportdata *, enum qib_sdma_states);
+static void sdma_start_sw_clean_up(struct qib_pportdata *);
+static void sdma_sw_clean_up_task(unsigned long);
+static void unmap_desc(struct qib_pportdata *, unsigned);
+
+static void sdma_get(struct qib_sdma_state *ss)
+{
+	kref_get(&ss->kref);
+}
+
+static void sdma_complete(struct kref *kref)
+{
+	struct qib_sdma_state *ss =
+		container_of(kref, struct qib_sdma_state, kref);
+
+	complete(&ss->comp);
+}
+
+static void sdma_put(struct qib_sdma_state *ss)
+{
+	kref_put(&ss->kref, sdma_complete);
+}
+
+static void sdma_finalput(struct qib_sdma_state *ss)
+{
+	sdma_put(ss);
+	wait_for_completion(&ss->comp);
+}
+
+/*
+ * Complete all the sdma requests on the active list, in the correct
+ * order, and with appropriate processing.   Called when cleaning up
+ * after sdma shutdown, and when new sdma requests are submitted for
+ * a link that is down.   This matches what is done for requests
+ * that complete normally, it's just the full list.
+ *
+ * Must be called with sdma_lock held
+ */
+static void clear_sdma_activelist(struct qib_pportdata *ppd)
+{
+	struct qib_sdma_txreq *txp, *txp_next;
+
+	list_for_each_entry_safe(txp, txp_next, &ppd->sdma_activelist, list) {
+		list_del_init(&txp->list);
+		if (txp->flags & QIB_SDMA_TXREQ_F_FREEDESC) {
+			unsigned idx;
+
+			idx = txp->start_idx;
+			while (idx != txp->next_descq_idx) {
+				unmap_desc(ppd, idx);
+				if (++idx == ppd->sdma_descq_cnt)
+					idx = 0;
+			}
+		}
+		if (txp->callback)
+			(*txp->callback)(txp, QIB_SDMA_TXREQ_S_ABORTED);
+	}
+}
+
+static void sdma_sw_clean_up_task(unsigned long opaque)
+{
+	struct qib_pportdata *ppd = (struct qib_pportdata *) opaque;
+	unsigned long flags;
+
+	spin_lock_irqsave(&ppd->sdma_lock, flags);
+
+	/*
+	 * At this point, the following should always be true:
+	 * - We are halted, so no more descriptors are getting retired.
+	 * - We are not running, so no one is submitting new work.
+	 * - Only we can send the e40_sw_cleaned, so we can't start
+	 *   running again until we say so.  So, the active list and
+	 *   descq are ours to play with.
+	 */
+
+	/* Process all retired requests. */
+	qib_sdma_make_progress(ppd);
+
+	clear_sdma_activelist(ppd);
+
+	/*
+	 * Resync count of added and removed.  It is VERY important that
+	 * sdma_descq_removed NEVER decrement - user_sdma depends on it.
+	 */
+	ppd->sdma_descq_removed = ppd->sdma_descq_added;
+
+	/*
+	 * Reset our notion of head and tail.
+	 * Note that the HW registers will be reset when switching states
+	 * due to calling __qib_sdma_process_event() below.
+	 */
+	ppd->sdma_descq_tail = 0;
+	ppd->sdma_descq_head = 0;
+	ppd->sdma_head_dma[0] = 0;
+	ppd->sdma_generation = 0;
+
+	__qib_sdma_process_event(ppd, qib_sdma_event_e40_sw_cleaned);
+
+	spin_unlock_irqrestore(&ppd->sdma_lock, flags);
+}
+
+/*
+ * This is called when changing to state qib_sdma_state_s10_hw_start_up_wait
+ * as a result of send buffer errors or send DMA descriptor errors.
+ * We want to disarm the buffers in these cases.
+ */
+static void sdma_hw_start_up(struct qib_pportdata *ppd)
+{
+	struct qib_sdma_state *ss = &ppd->sdma_state;
+	unsigned bufno;
+
+	for (bufno = ss->first_sendbuf; bufno < ss->last_sendbuf; ++bufno)
+		ppd->dd->f_sendctrl(ppd, QIB_SENDCTRL_DISARM_BUF(bufno));
+
+	ppd->dd->f_sdma_hw_start_up(ppd);
+}
+
+static void sdma_sw_tear_down(struct qib_pportdata *ppd)
+{
+	struct qib_sdma_state *ss = &ppd->sdma_state;
+
+	/* Releasing this reference means the state machine has stopped. */
+	sdma_put(ss);
+}
+
+static void sdma_start_sw_clean_up(struct qib_pportdata *ppd)
+{
+	tasklet_hi_schedule(&ppd->sdma_sw_clean_up_task);
+}
+
+static void sdma_set_state(struct qib_pportdata *ppd,
+	enum qib_sdma_states next_state)
+{
+	struct qib_sdma_state *ss = &ppd->sdma_state;
+	struct sdma_set_state_action *action = ss->set_state_action;
+	unsigned op = 0;
+
+	/* debugging bookkeeping */
+	ss->previous_state = ss->current_state;
+	ss->previous_op = ss->current_op;
+
+	ss->current_state = next_state;
+
+	if (action[next_state].op_enable)
+		op |= QIB_SDMA_SENDCTRL_OP_ENABLE;
+
+	if (action[next_state].op_intenable)
+		op |= QIB_SDMA_SENDCTRL_OP_INTENABLE;
+
+	if (action[next_state].op_halt)
+		op |= QIB_SDMA_SENDCTRL_OP_HALT;
+
+	if (action[next_state].op_drain)
+		op |= QIB_SDMA_SENDCTRL_OP_DRAIN;
+
+	if (action[next_state].go_s99_running_tofalse)
+		ss->go_s99_running = 0;
+
+	if (action[next_state].go_s99_running_totrue)
+		ss->go_s99_running = 1;
+
+	ss->current_op = op;
+
+	ppd->dd->f_sdma_sendctrl(ppd, ss->current_op);
+}
+
+static void unmap_desc(struct qib_pportdata *ppd, unsigned head)
+{
+	__le64 *descqp = &ppd->sdma_descq[head].qw[0];
+	u64 desc[2];
+	dma_addr_t addr;
+	size_t len;
+
+	desc[0] = le64_to_cpu(descqp[0]);
+	desc[1] = le64_to_cpu(descqp[1]);
+
+	addr = (desc[1] << 32) | (desc[0] >> 32);
+	len = (desc[0] >> 14) & (0x7ffULL << 2);
+	dma_unmap_single(&ppd->dd->pcidev->dev, addr, len, DMA_TO_DEVICE);
+}
+
+static int alloc_sdma(struct qib_pportdata *ppd)
+{
+	ppd->sdma_descq_cnt = sdma_descq_cnt;
+	if (!ppd->sdma_descq_cnt)
+		ppd->sdma_descq_cnt = 256;
+
+	/* Allocate memory for SendDMA descriptor FIFO */
+	ppd->sdma_descq = dma_alloc_coherent(&ppd->dd->pcidev->dev,
+		ppd->sdma_descq_cnt * sizeof(u64[2]), &ppd->sdma_descq_phys,
+		GFP_KERNEL);
+
+	if (!ppd->sdma_descq) {
+		qib_dev_err(ppd->dd, "failed to allocate SendDMA descriptor "
+			    "FIFO memory\n");
+		goto bail;
+	}
+
+	/* Allocate memory for DMA of head register to memory */
+	ppd->sdma_head_dma = dma_alloc_coherent(&ppd->dd->pcidev->dev,
+		PAGE_SIZE, &ppd->sdma_head_phys, GFP_KERNEL);
+	if (!ppd->sdma_head_dma) {
+		qib_dev_err(ppd->dd, "failed to allocate SendDMA "
+			    "head memory\n");
+		goto cleanup_descq;
+	}
+	ppd->sdma_head_dma[0] = 0;
+	return 0;
+
+cleanup_descq:
+	dma_free_coherent(&ppd->dd->pcidev->dev,
+		ppd->sdma_descq_cnt * sizeof(u64[2]), (void *)ppd->sdma_descq,
+		ppd->sdma_descq_phys);
+	ppd->sdma_descq = NULL;
+	ppd->sdma_descq_phys = 0;
+bail:
+	ppd->sdma_descq_cnt = 0;
+	return -ENOMEM;
+}
+
+static void free_sdma(struct qib_pportdata *ppd)
+{
+	struct qib_devdata *dd = ppd->dd;
+
+	if (ppd->sdma_head_dma) {
+		dma_free_coherent(&dd->pcidev->dev, PAGE_SIZE,
+				  (void *)ppd->sdma_head_dma,
+				  ppd->sdma_head_phys);
+		ppd->sdma_head_dma = NULL;
+		ppd->sdma_head_phys = 0;
+	}
+
+	if (ppd->sdma_descq) {
+		dma_free_coherent(&dd->pcidev->dev,
+				  ppd->sdma_descq_cnt * sizeof(u64[2]),
+				  ppd->sdma_descq, ppd->sdma_descq_phys);
+		ppd->sdma_descq = NULL;
+		ppd->sdma_descq_phys = 0;
+	}
+}
+
+static inline void make_sdma_desc(struct qib_pportdata *ppd,
+				  u64 *sdmadesc, u64 addr, u64 dwlen,
+				  u64 dwoffset)
+{
+
+	WARN_ON(addr & 3);
+	/* SDmaPhyAddr[47:32] */
+	sdmadesc[1] = addr >> 32;
+	/* SDmaPhyAddr[31:0] */
+	sdmadesc[0] = (addr & 0xfffffffcULL) << 32;
+	/* SDmaGeneration[1:0] */
+	sdmadesc[0] |= (ppd->sdma_generation & 3ULL) <<
+		SDMA_DESC_GEN_LSB;
+	/* SDmaDwordCount[10:0] */
+	sdmadesc[0] |= (dwlen & 0x7ffULL) << SDMA_DESC_COUNT_LSB;
+	/* SDmaBufOffset[12:2] */
+	sdmadesc[0] |= dwoffset & 0x7ffULL;
+}
+
+/* sdma_lock must be held */
+int qib_sdma_make_progress(struct qib_pportdata *ppd)
+{
+	struct list_head *lp = NULL;
+	struct qib_sdma_txreq *txp = NULL;
+	struct qib_devdata *dd = ppd->dd;
+	int progress = 0;
+	u16 hwhead;
+	u16 idx = 0;
+
+	hwhead = dd->f_sdma_gethead(ppd);
+
+	/* The reason for some of the complexity of this code is that
+	 * not all descriptors have corresponding txps.  So, we have to
+	 * be able to skip over descs until we wander into the range of
+	 * the next txp on the list.
+	 */
+
+	if (!list_empty(&ppd->sdma_activelist)) {
+		lp = ppd->sdma_activelist.next;
+		txp = list_entry(lp, struct qib_sdma_txreq, list);
+		idx = txp->start_idx;
+	}
+
+	while (ppd->sdma_descq_head != hwhead) {
+		/* if desc is part of this txp, unmap if needed */
+		if (txp && (txp->flags & QIB_SDMA_TXREQ_F_FREEDESC) &&
+		    (idx == ppd->sdma_descq_head)) {
+			unmap_desc(ppd, ppd->sdma_descq_head);
+			if (++idx == ppd->sdma_descq_cnt)
+				idx = 0;
+		}
+
+		/* increment dequed desc count */
+		ppd->sdma_descq_removed++;
+
+		/* advance head, wrap if needed */
+		if (++ppd->sdma_descq_head == ppd->sdma_descq_cnt)
+			ppd->sdma_descq_head = 0;
+
+		/* if now past this txp's descs, do the callback */
+		if (txp && txp->next_descq_idx == ppd->sdma_descq_head) {
+			/* remove from active list */
+			list_del_init(&txp->list);
+			if (txp->callback)
+				(*txp->callback)(txp, QIB_SDMA_TXREQ_S_OK);
+			/* see if there is another txp */
+			if (list_empty(&ppd->sdma_activelist))
+				txp = NULL;
+			else {
+				lp = ppd->sdma_activelist.next;
+				txp = list_entry(lp, struct qib_sdma_txreq,
+					list);
+				idx = txp->start_idx;
+			}
+		}
+		progress = 1;
+	}
+	if (progress)
+		qib_verbs_sdma_desc_avail(ppd, qib_sdma_descq_freecnt(ppd));
+	return progress;
+}
+
+/*
+ * This is called from interrupt context.
+ */
+void qib_sdma_intr(struct qib_pportdata *ppd)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&ppd->sdma_lock, flags);
+
+	__qib_sdma_intr(ppd);
+
+	spin_unlock_irqrestore(&ppd->sdma_lock, flags);
+}
+
+void __qib_sdma_intr(struct qib_pportdata *ppd)
+{
+	if (__qib_sdma_running(ppd))
+		qib_sdma_make_progress(ppd);
+}
+
+int qib_setup_sdma(struct qib_pportdata *ppd)
+{
+	struct qib_devdata *dd = ppd->dd;
+	unsigned long flags;
+	int ret = 0;
+
+	ret = alloc_sdma(ppd);
+	if (ret)
+		goto bail;
+
+	/* set consistent sdma state */
+	ppd->dd->f_sdma_init_early(ppd);
+	spin_lock_irqsave(&ppd->sdma_lock, flags);
+	sdma_set_state(ppd, qib_sdma_state_s00_hw_down);
+	spin_unlock_irqrestore(&ppd->sdma_lock, flags);
+
+	/* set up reference counting */
+	kref_init(&ppd->sdma_state.kref);
+	init_completion(&ppd->sdma_state.comp);
+
+	ppd->sdma_generation = 0;
+	ppd->sdma_descq_head = 0;
+	ppd->sdma_descq_removed = 0;
+	ppd->sdma_descq_added = 0;
+
+	INIT_LIST_HEAD(&ppd->sdma_activelist);
+
+	tasklet_init(&ppd->sdma_sw_clean_up_task, sdma_sw_clean_up_task,
+		(unsigned long)ppd);
+
+	ret = dd->f_init_sdma_regs(ppd);
+	if (ret)
+		goto bail_alloc;
+
+	qib_sdma_process_event(ppd, qib_sdma_event_e10_go_hw_start);
+
+	return 0;
+
+bail_alloc:
+	qib_teardown_sdma(ppd);
+bail:
+	return ret;
+}
+
+void qib_teardown_sdma(struct qib_pportdata *ppd)
+{
+	qib_sdma_process_event(ppd, qib_sdma_event_e00_go_hw_down);
+
+	/*
+	 * This waits for the state machine to exit so it is not
+	 * necessary to kill the sdma_sw_clean_up_task to make sure
+	 * it is not running.
+	 */
+	sdma_finalput(&ppd->sdma_state);
+
+	free_sdma(ppd);
+}
+
+int qib_sdma_running(struct qib_pportdata *ppd)
+{
+	unsigned long flags;
+	int ret;
+
+	spin_lock_irqsave(&ppd->sdma_lock, flags);
+	ret = __qib_sdma_running(ppd);
+	spin_unlock_irqrestore(&ppd->sdma_lock, flags);
+
+	return ret;
+}
+
+/*
+ * Complete a request when sdma not running; likely only request
+ * but to simplify the code, always queue it, then process the full
+ * activelist.  We process the entire list to ensure that this particular
+ * request does get it's callback, but in the correct order.
+ * Must be called with sdma_lock held
+ */
+static void complete_sdma_err_req(struct qib_pportdata *ppd,
+				  struct qib_verbs_txreq *tx)
+{
+	atomic_inc(&tx->qp->s_dma_busy);
+	/* no sdma descriptors, so no unmap_desc */
+	tx->txreq.start_idx = 0;
+	tx->txreq.next_descq_idx = 0;
+	list_add_tail(&tx->txreq.list, &ppd->sdma_activelist);
+	clear_sdma_activelist(ppd);
+}
+
+/*
+ * This function queues one IB packet onto the send DMA queue per call.
+ * The caller is responsible for checking:
+ * 1) The number of send DMA descriptor entries is less than the size of
+ *    the descriptor queue.
+ * 2) The IB SGE addresses and lengths are 32-bit aligned
+ *    (except possibly the last SGE's length)
+ * 3) The SGE addresses are suitable for passing to dma_map_single().
+ */
+int qib_sdma_verbs_send(struct qib_pportdata *ppd,
+			struct qib_sge_state *ss, u32 dwords,
+			struct qib_verbs_txreq *tx)
+{
+	unsigned long flags;
+	struct qib_sge *sge;
+	struct qib_qp *qp;
+	int ret = 0;
+	u16 tail;
+	__le64 *descqp;
+	u64 sdmadesc[2];
+	u32 dwoffset;
+	dma_addr_t addr;
+
+	spin_lock_irqsave(&ppd->sdma_lock, flags);
+
+retry:
+	if (unlikely(!__qib_sdma_running(ppd))) {
+		complete_sdma_err_req(ppd, tx);
+		goto unlock;
+	}
+
+	if (tx->txreq.sg_count > qib_sdma_descq_freecnt(ppd)) {
+		if (qib_sdma_make_progress(ppd))
+			goto retry;
+		if (ppd->dd->flags & QIB_HAS_SDMA_TIMEOUT)
+			ppd->dd->f_sdma_set_desc_cnt(ppd,
+					ppd->sdma_descq_cnt / 2);
+		goto busy;
+	}
+
+	dwoffset = tx->hdr_dwords;
+	make_sdma_desc(ppd, sdmadesc, (u64) tx->txreq.addr, dwoffset, 0);
+
+	sdmadesc[0] |= SDMA_DESC_FIRST;
+	if (tx->txreq.flags & QIB_SDMA_TXREQ_F_USELARGEBUF)
+		sdmadesc[0] |= SDMA_DESC_USE_LARGE_BUF;
+
+	/* write to the descq */
+	tail = ppd->sdma_descq_tail;
+	descqp = &ppd->sdma_descq[tail].qw[0];
+	*descqp++ = cpu_to_le64(sdmadesc[0]);
+	*descqp++ = cpu_to_le64(sdmadesc[1]);
+
+	/* increment the tail */
+	if (++tail == ppd->sdma_descq_cnt) {
+		tail = 0;
+		descqp = &ppd->sdma_descq[0].qw[0];
+		++ppd->sdma_generation;
+	}
+
+	tx->txreq.start_idx = tail;
+
+	sge = &ss->sge;
+	while (dwords) {
+		u32 dw;
+		u32 len;
+
+		len = dwords << 2;
+		if (len > sge->length)
+			len = sge->length;
+		if (len > sge->sge_length)
+			len = sge->sge_length;
+		BUG_ON(len == 0);
+		dw = (len + 3) >> 2;
+		addr = dma_map_single(&ppd->dd->pcidev->dev, sge->vaddr,
+				      dw << 2, DMA_TO_DEVICE);
+		if (dma_mapping_error(&ppd->dd->pcidev->dev, addr))
+			goto unmap;
+		sdmadesc[0] = 0;
+		make_sdma_desc(ppd, sdmadesc, (u64) addr, dw, dwoffset);
+		/* SDmaUseLargeBuf has to be set in every descriptor */
+		if (tx->txreq.flags & QIB_SDMA_TXREQ_F_USELARGEBUF)
+			sdmadesc[0] |= SDMA_DESC_USE_LARGE_BUF;
+		/* write to the descq */
+		*descqp++ = cpu_to_le64(sdmadesc[0]);
+		*descqp++ = cpu_to_le64(sdmadesc[1]);
+
+		/* increment the tail */
+		if (++tail == ppd->sdma_descq_cnt) {
+			tail = 0;
+			descqp = &ppd->sdma_descq[0].qw[0];
+			++ppd->sdma_generation;
+		}
+		sge->vaddr += len;
+		sge->length -= len;
+		sge->sge_length -= len;
+		if (sge->sge_length == 0) {
+			if (--ss->num_sge)
+				*sge = *ss->sg_list++;
+		} else if (sge->length == 0 && sge->mr->lkey) {
+			if (++sge->n >= QIB_SEGSZ) {
+				if (++sge->m >= sge->mr->mapsz)
+					break;
+				sge->n = 0;
+			}
+			sge->vaddr =
+				sge->mr->map[sge->m]->segs[sge->n].vaddr;
+			sge->length =
+				sge->mr->map[sge->m]->segs[sge->n].length;
+		}
+
+		dwoffset += dw;
+		dwords -= dw;
+	}
+
+	if (!tail)
+		descqp = &ppd->sdma_descq[ppd->sdma_descq_cnt].qw[0];
+	descqp -= 2;
+	descqp[0] |= cpu_to_le64(SDMA_DESC_LAST);
+	if (tx->txreq.flags & QIB_SDMA_TXREQ_F_HEADTOHOST)
+		descqp[0] |= cpu_to_le64(SDMA_DESC_DMA_HEAD);
+	if (tx->txreq.flags & QIB_SDMA_TXREQ_F_INTREQ)
+		descqp[0] |= cpu_to_le64(SDMA_DESC_INTR);
+
+	atomic_inc(&tx->qp->s_dma_busy);
+	tx->txreq.next_descq_idx = tail;
+	ppd->dd->f_sdma_update_tail(ppd, tail);
+	ppd->sdma_descq_added += tx->txreq.sg_count;
+	list_add_tail(&tx->txreq.list, &ppd->sdma_activelist);
+	goto unlock;
+
+unmap:
+	for (;;) {
+		if (!tail)
+			tail = ppd->sdma_descq_cnt - 1;
+		else
+			tail--;
+		if (tail == ppd->sdma_descq_tail)
+			break;
+		unmap_desc(ppd, tail);
+	}
+	qp = tx->qp;
+	qib_put_txreq(tx);
+	spin_lock(&qp->s_lock);
+	if (qp->ibqp.qp_type == IB_QPT_RC) {
+		/* XXX what about error sending RDMA read responses? */
+		if (ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK)
+			qib_error_qp(qp, IB_WC_GENERAL_ERR);
+	} else if (qp->s_wqe)
+		qib_send_complete(qp, qp->s_wqe, IB_WC_GENERAL_ERR);
+	spin_unlock(&qp->s_lock);
+	/* return zero to process the next send work request */
+	goto unlock;
+
+busy:
+	qp = tx->qp;
+	spin_lock(&qp->s_lock);
+	if (ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK) {
+		struct qib_ibdev *dev;
+
+		/*
+		 * If we couldn't queue the DMA request, save the info
+		 * and try again later rather than destroying the
+		 * buffer and undoing the side effects of the copy.
+		 */
+		tx->ss = ss;
+		tx->dwords = dwords;
+		qp->s_tx = tx;
+		dev = &ppd->dd->verbs_dev;
+		spin_lock(&dev->pending_lock);
+		if (list_empty(&qp->iowait)) {
+			struct qib_ibport *ibp;
+
+			ibp = &ppd->ibport_data;
+			ibp->n_dmawait++;
+			qp->s_flags |= QIB_S_WAIT_DMA_DESC;
+			list_add_tail(&qp->iowait, &dev->dmawait);
+		}
+		spin_unlock(&dev->pending_lock);
+		qp->s_flags &= ~QIB_S_BUSY;
+		spin_unlock(&qp->s_lock);
+		ret = -EBUSY;
+	} else {
+		spin_unlock(&qp->s_lock);
+		qib_put_txreq(tx);
+	}
+unlock:
+	spin_unlock_irqrestore(&ppd->sdma_lock, flags);
+	return ret;
+}
+
+void qib_sdma_process_event(struct qib_pportdata *ppd,
+	enum qib_sdma_events event)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&ppd->sdma_lock, flags);
+
+	__qib_sdma_process_event(ppd, event);
+
+	if (ppd->sdma_state.current_state == qib_sdma_state_s99_running)
+		qib_verbs_sdma_desc_avail(ppd, qib_sdma_descq_freecnt(ppd));
+
+	spin_unlock_irqrestore(&ppd->sdma_lock, flags);
+}
+
+void __qib_sdma_process_event(struct qib_pportdata *ppd,
+	enum qib_sdma_events event)
+{
+	struct qib_sdma_state *ss = &ppd->sdma_state;
+
+	switch (ss->current_state) {
+	case qib_sdma_state_s00_hw_down:
+		switch (event) {
+		case qib_sdma_event_e00_go_hw_down:
+			break;
+		case qib_sdma_event_e30_go_running:
+			/*
+			 * If down, but running requested (usually result
+			 * of link up, then we need to start up.
+			 * This can happen when hw down is requested while
+			 * bringing the link up with traffic active on
+			 * 7220, e.g. */
+			ss->go_s99_running = 1;
+			/* fall through and start dma engine */
+		case qib_sdma_event_e10_go_hw_start:
+			/* This reference means the state machine is started */
+			sdma_get(&ppd->sdma_state);
+			sdma_set_state(ppd,
+				       qib_sdma_state_s10_hw_start_up_wait);
+			break;
+		case qib_sdma_event_e20_hw_started:
+			break;
+		case qib_sdma_event_e40_sw_cleaned:
+			sdma_sw_tear_down(ppd);
+			break;
+		case qib_sdma_event_e50_hw_cleaned:
+			break;
+		case qib_sdma_event_e60_hw_halted:
+			break;
+		case qib_sdma_event_e70_go_idle:
+			break;
+		case qib_sdma_event_e7220_err_halted:
+			break;
+		case qib_sdma_event_e7322_err_halted:
+			break;
+		case qib_sdma_event_e90_timer_tick:
+			break;
+		}
+		break;
+
+	case qib_sdma_state_s10_hw_start_up_wait:
+		switch (event) {
+		case qib_sdma_event_e00_go_hw_down:
+			sdma_set_state(ppd, qib_sdma_state_s00_hw_down);
+			sdma_sw_tear_down(ppd);
+			break;
+		case qib_sdma_event_e10_go_hw_start:
+			break;
+		case qib_sdma_event_e20_hw_started:
+			sdma_set_state(ppd, ss->go_s99_running ?
+				       qib_sdma_state_s99_running :
+				       qib_sdma_state_s20_idle);
+			break;
+		case qib_sdma_event_e30_go_running:
+			ss->go_s99_running = 1;
+			break;
+		case qib_sdma_event_e40_sw_cleaned:
+			break;
+		case qib_sdma_event_e50_hw_cleaned:
+			break;
+		case qib_sdma_event_e60_hw_halted:
+			break;
+		case qib_sdma_event_e70_go_idle:
+			ss->go_s99_running = 0;
+			break;
+		case qib_sdma_event_e7220_err_halted:
+			break;
+		case qib_sdma_event_e7322_err_halted:
+			break;
+		case qib_sdma_event_e90_timer_tick:
+			break;
+		}
+		break;
+
+	case qib_sdma_state_s20_idle:
+		switch (event) {
+		case qib_sdma_event_e00_go_hw_down:
+			sdma_set_state(ppd, qib_sdma_state_s00_hw_down);
+			sdma_sw_tear_down(ppd);
+			break;
+		case qib_sdma_event_e10_go_hw_start:
+			break;
+		case qib_sdma_event_e20_hw_started:
+			break;
+		case qib_sdma_event_e30_go_running:
+			sdma_set_state(ppd, qib_sdma_state_s99_running);
+			ss->go_s99_running = 1;
+			break;
+		case qib_sdma_event_e40_sw_cleaned:
+			break;
+		case qib_sdma_event_e50_hw_cleaned:
+			break;
+		case qib_sdma_event_e60_hw_halted:
+			break;
+		case qib_sdma_event_e70_go_idle:
+			break;
+		case qib_sdma_event_e7220_err_halted:
+			break;
+		case qib_sdma_event_e7322_err_halted:
+			break;
+		case qib_sdma_event_e90_timer_tick:
+			break;
+		}
+		break;
+
+	case qib_sdma_state_s30_sw_clean_up_wait:
+		switch (event) {
+		case qib_sdma_event_e00_go_hw_down:
+			sdma_set_state(ppd, qib_sdma_state_s00_hw_down);
+			break;
+		case qib_sdma_event_e10_go_hw_start:
+			break;
+		case qib_sdma_event_e20_hw_started:
+			break;
+		case qib_sdma_event_e30_go_running:
+			ss->go_s99_running = 1;
+			break;
+		case qib_sdma_event_e40_sw_cleaned:
+			sdma_set_state(ppd,
+				       qib_sdma_state_s10_hw_start_up_wait);
+			sdma_hw_start_up(ppd);
+			break;
+		case qib_sdma_event_e50_hw_cleaned:
+			break;
+		case qib_sdma_event_e60_hw_halted:
+			break;
+		case qib_sdma_event_e70_go_idle:
+			ss->go_s99_running = 0;
+			break;
+		case qib_sdma_event_e7220_err_halted:
+			break;
+		case qib_sdma_event_e7322_err_halted:
+			break;
+		case qib_sdma_event_e90_timer_tick:
+			break;
+		}
+		break;
+
+	case qib_sdma_state_s40_hw_clean_up_wait:
+		switch (event) {
+		case qib_sdma_event_e00_go_hw_down:
+			sdma_set_state(ppd, qib_sdma_state_s00_hw_down);
+			sdma_start_sw_clean_up(ppd);
+			break;
+		case qib_sdma_event_e10_go_hw_start:
+			break;
+		case qib_sdma_event_e20_hw_started:
+			break;
+		case qib_sdma_event_e30_go_running:
+			ss->go_s99_running = 1;
+			break;
+		case qib_sdma_event_e40_sw_cleaned:
+			break;
+		case qib_sdma_event_e50_hw_cleaned:
+			sdma_set_state(ppd,
+				       qib_sdma_state_s30_sw_clean_up_wait);
+			sdma_start_sw_clean_up(ppd);
+			break;
+		case qib_sdma_event_e60_hw_halted:
+			break;
+		case qib_sdma_event_e70_go_idle:
+			ss->go_s99_running = 0;
+			break;
+		case qib_sdma_event_e7220_err_halted:
+			break;
+		case qib_sdma_event_e7322_err_halted:
+			break;
+		case qib_sdma_event_e90_timer_tick:
+			break;
+		}
+		break;
+
+	case qib_sdma_state_s50_hw_halt_wait:
+		switch (event) {
+		case qib_sdma_event_e00_go_hw_down:
+			sdma_set_state(ppd, qib_sdma_state_s00_hw_down);
+			sdma_start_sw_clean_up(ppd);
+			break;
+		case qib_sdma_event_e10_go_hw_start:
+			break;
+		case qib_sdma_event_e20_hw_started:
+			break;
+		case qib_sdma_event_e30_go_running:
+			ss->go_s99_running = 1;
+			break;
+		case qib_sdma_event_e40_sw_cleaned:
+			break;
+		case qib_sdma_event_e50_hw_cleaned:
+			break;
+		case qib_sdma_event_e60_hw_halted:
+			sdma_set_state(ppd,
+				       qib_sdma_state_s40_hw_clean_up_wait);
+			ppd->dd->f_sdma_hw_clean_up(ppd);
+			break;
+		case qib_sdma_event_e70_go_idle:
+			ss->go_s99_running = 0;
+			break;
+		case qib_sdma_event_e7220_err_halted:
+			break;
+		case qib_sdma_event_e7322_err_halted:
+			break;
+		case qib_sdma_event_e90_timer_tick:
+			break;
+		}
+		break;
+
+	case qib_sdma_state_s99_running:
+		switch (event) {
+		case qib_sdma_event_e00_go_hw_down:
+			sdma_set_state(ppd, qib_sdma_state_s00_hw_down);
+			sdma_start_sw_clean_up(ppd);
+			break;
+		case qib_sdma_event_e10_go_hw_start:
+			break;
+		case qib_sdma_event_e20_hw_started:
+			break;
+		case qib_sdma_event_e30_go_running:
+			break;
+		case qib_sdma_event_e40_sw_cleaned:
+			break;
+		case qib_sdma_event_e50_hw_cleaned:
+			break;
+		case qib_sdma_event_e60_hw_halted:
+			sdma_set_state(ppd,
+				       qib_sdma_state_s30_sw_clean_up_wait);
+			sdma_start_sw_clean_up(ppd);
+			break;
+		case qib_sdma_event_e70_go_idle:
+			sdma_set_state(ppd, qib_sdma_state_s50_hw_halt_wait);
+			ss->go_s99_running = 0;
+			break;
+		case qib_sdma_event_e7220_err_halted:
+			sdma_set_state(ppd,
+				       qib_sdma_state_s30_sw_clean_up_wait);
+			sdma_start_sw_clean_up(ppd);
+			break;
+		case qib_sdma_event_e7322_err_halted:
+			sdma_set_state(ppd, qib_sdma_state_s50_hw_halt_wait);
+			break;
+		case qib_sdma_event_e90_timer_tick:
+			break;
+		}
+		break;
+	}
+
+	ss->last_event = event;
+}
diff --git a/drivers/infiniband/hw/qib/qib_srq.c b/drivers/infiniband/hw/qib/qib_srq.c
new file mode 100644
index 000000000000..c3ec8efc2ed8
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_srq.c
@@ -0,0 +1,375 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/err.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+
+#include "qib_verbs.h"
+
+/**
+ * qib_post_srq_receive - post a receive on a shared receive queue
+ * @ibsrq: the SRQ to post the receive on
+ * @wr: the list of work requests to post
+ * @bad_wr: A pointer to the first WR to cause a problem is put here
+ *
+ * This may be called from interrupt context.
+ */
+int qib_post_srq_receive(struct ib_srq *ibsrq, struct ib_recv_wr *wr,
+			 struct ib_recv_wr **bad_wr)
+{
+	struct qib_srq *srq = to_isrq(ibsrq);
+	struct qib_rwq *wq;
+	unsigned long flags;
+	int ret;
+
+	for (; wr; wr = wr->next) {
+		struct qib_rwqe *wqe;
+		u32 next;
+		int i;
+
+		if ((unsigned) wr->num_sge > srq->rq.max_sge) {
+			*bad_wr = wr;
+			ret = -EINVAL;
+			goto bail;
+		}
+
+		spin_lock_irqsave(&srq->rq.lock, flags);
+		wq = srq->rq.wq;
+		next = wq->head + 1;
+		if (next >= srq->rq.size)
+			next = 0;
+		if (next == wq->tail) {
+			spin_unlock_irqrestore(&srq->rq.lock, flags);
+			*bad_wr = wr;
+			ret = -ENOMEM;
+			goto bail;
+		}
+
+		wqe = get_rwqe_ptr(&srq->rq, wq->head);
+		wqe->wr_id = wr->wr_id;
+		wqe->num_sge = wr->num_sge;
+		for (i = 0; i < wr->num_sge; i++)
+			wqe->sg_list[i] = wr->sg_list[i];
+		/* Make sure queue entry is written before the head index. */
+		smp_wmb();
+		wq->head = next;
+		spin_unlock_irqrestore(&srq->rq.lock, flags);
+	}
+	ret = 0;
+
+bail:
+	return ret;
+}
+
+/**
+ * qib_create_srq - create a shared receive queue
+ * @ibpd: the protection domain of the SRQ to create
+ * @srq_init_attr: the attributes of the SRQ
+ * @udata: data from libibverbs when creating a user SRQ
+ */
+struct ib_srq *qib_create_srq(struct ib_pd *ibpd,
+			      struct ib_srq_init_attr *srq_init_attr,
+			      struct ib_udata *udata)
+{
+	struct qib_ibdev *dev = to_idev(ibpd->device);
+	struct qib_srq *srq;
+	u32 sz;
+	struct ib_srq *ret;
+
+	if (srq_init_attr->attr.max_sge == 0 ||
+	    srq_init_attr->attr.max_sge > ib_qib_max_srq_sges ||
+	    srq_init_attr->attr.max_wr == 0 ||
+	    srq_init_attr->attr.max_wr > ib_qib_max_srq_wrs) {
+		ret = ERR_PTR(-EINVAL);
+		goto done;
+	}
+
+	srq = kmalloc(sizeof(*srq), GFP_KERNEL);
+	if (!srq) {
+		ret = ERR_PTR(-ENOMEM);
+		goto done;
+	}
+
+	/*
+	 * Need to use vmalloc() if we want to support large #s of entries.
+	 */
+	srq->rq.size = srq_init_attr->attr.max_wr + 1;
+	srq->rq.max_sge = srq_init_attr->attr.max_sge;
+	sz = sizeof(struct ib_sge) * srq->rq.max_sge +
+		sizeof(struct qib_rwqe);
+	srq->rq.wq = vmalloc_user(sizeof(struct qib_rwq) + srq->rq.size * sz);
+	if (!srq->rq.wq) {
+		ret = ERR_PTR(-ENOMEM);
+		goto bail_srq;
+	}
+
+	/*
+	 * Return the address of the RWQ as the offset to mmap.
+	 * See qib_mmap() for details.
+	 */
+	if (udata && udata->outlen >= sizeof(__u64)) {
+		int err;
+		u32 s = sizeof(struct qib_rwq) + srq->rq.size * sz;
+
+		srq->ip =
+		    qib_create_mmap_info(dev, s, ibpd->uobject->context,
+					 srq->rq.wq);
+		if (!srq->ip) {
+			ret = ERR_PTR(-ENOMEM);
+			goto bail_wq;
+		}
+
+		err = ib_copy_to_udata(udata, &srq->ip->offset,
+				       sizeof(srq->ip->offset));
+		if (err) {
+			ret = ERR_PTR(err);
+			goto bail_ip;
+		}
+	} else
+		srq->ip = NULL;
+
+	/*
+	 * ib_create_srq() will initialize srq->ibsrq.
+	 */
+	spin_lock_init(&srq->rq.lock);
+	srq->rq.wq->head = 0;
+	srq->rq.wq->tail = 0;
+	srq->limit = srq_init_attr->attr.srq_limit;
+
+	spin_lock(&dev->n_srqs_lock);
+	if (dev->n_srqs_allocated == ib_qib_max_srqs) {
+		spin_unlock(&dev->n_srqs_lock);
+		ret = ERR_PTR(-ENOMEM);
+		goto bail_ip;
+	}
+
+	dev->n_srqs_allocated++;
+	spin_unlock(&dev->n_srqs_lock);
+
+	if (srq->ip) {
+		spin_lock_irq(&dev->pending_lock);
+		list_add(&srq->ip->pending_mmaps, &dev->pending_mmaps);
+		spin_unlock_irq(&dev->pending_lock);
+	}
+
+	ret = &srq->ibsrq;
+	goto done;
+
+bail_ip:
+	kfree(srq->ip);
+bail_wq:
+	vfree(srq->rq.wq);
+bail_srq:
+	kfree(srq);
+done:
+	return ret;
+}
+
+/**
+ * qib_modify_srq - modify a shared receive queue
+ * @ibsrq: the SRQ to modify
+ * @attr: the new attributes of the SRQ
+ * @attr_mask: indicates which attributes to modify
+ * @udata: user data for libibverbs.so
+ */
+int qib_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
+		   enum ib_srq_attr_mask attr_mask,
+		   struct ib_udata *udata)
+{
+	struct qib_srq *srq = to_isrq(ibsrq);
+	struct qib_rwq *wq;
+	int ret = 0;
+
+	if (attr_mask & IB_SRQ_MAX_WR) {
+		struct qib_rwq *owq;
+		struct qib_rwqe *p;
+		u32 sz, size, n, head, tail;
+
+		/* Check that the requested sizes are below the limits. */
+		if ((attr->max_wr > ib_qib_max_srq_wrs) ||
+		    ((attr_mask & IB_SRQ_LIMIT) ?
+		     attr->srq_limit : srq->limit) > attr->max_wr) {
+			ret = -EINVAL;
+			goto bail;
+		}
+
+		sz = sizeof(struct qib_rwqe) +
+			srq->rq.max_sge * sizeof(struct ib_sge);
+		size = attr->max_wr + 1;
+		wq = vmalloc_user(sizeof(struct qib_rwq) + size * sz);
+		if (!wq) {
+			ret = -ENOMEM;
+			goto bail;
+		}
+
+		/* Check that we can write the offset to mmap. */
+		if (udata && udata->inlen >= sizeof(__u64)) {
+			__u64 offset_addr;
+			__u64 offset = 0;
+
+			ret = ib_copy_from_udata(&offset_addr, udata,
+						 sizeof(offset_addr));
+			if (ret)
+				goto bail_free;
+			udata->outbuf =
+				(void __user *) (unsigned long) offset_addr;
+			ret = ib_copy_to_udata(udata, &offset,
+					       sizeof(offset));
+			if (ret)
+				goto bail_free;
+		}
+
+		spin_lock_irq(&srq->rq.lock);
+		/*
+		 * validate head and tail pointer values and compute
+		 * the number of remaining WQEs.
+		 */
+		owq = srq->rq.wq;
+		head = owq->head;
+		tail = owq->tail;
+		if (head >= srq->rq.size || tail >= srq->rq.size) {
+			ret = -EINVAL;
+			goto bail_unlock;
+		}
+		n = head;
+		if (n < tail)
+			n += srq->rq.size - tail;
+		else
+			n -= tail;
+		if (size <= n) {
+			ret = -EINVAL;
+			goto bail_unlock;
+		}
+		n = 0;
+		p = wq->wq;
+		while (tail != head) {
+			struct qib_rwqe *wqe;
+			int i;
+
+			wqe = get_rwqe_ptr(&srq->rq, tail);
+			p->wr_id = wqe->wr_id;
+			p->num_sge = wqe->num_sge;
+			for (i = 0; i < wqe->num_sge; i++)
+				p->sg_list[i] = wqe->sg_list[i];
+			n++;
+			p = (struct qib_rwqe *)((char *) p + sz);
+			if (++tail >= srq->rq.size)
+				tail = 0;
+		}
+		srq->rq.wq = wq;
+		srq->rq.size = size;
+		wq->head = n;
+		wq->tail = 0;
+		if (attr_mask & IB_SRQ_LIMIT)
+			srq->limit = attr->srq_limit;
+		spin_unlock_irq(&srq->rq.lock);
+
+		vfree(owq);
+
+		if (srq->ip) {
+			struct qib_mmap_info *ip = srq->ip;
+			struct qib_ibdev *dev = to_idev(srq->ibsrq.device);
+			u32 s = sizeof(struct qib_rwq) + size * sz;
+
+			qib_update_mmap_info(dev, ip, s, wq);
+
+			/*
+			 * Return the offset to mmap.
+			 * See qib_mmap() for details.
+			 */
+			if (udata && udata->inlen >= sizeof(__u64)) {
+				ret = ib_copy_to_udata(udata, &ip->offset,
+						       sizeof(ip->offset));
+				if (ret)
+					goto bail;
+			}
+
+			/*
+			 * Put user mapping info onto the pending list
+			 * unless it already is on the list.
+			 */
+			spin_lock_irq(&dev->pending_lock);
+			if (list_empty(&ip->pending_mmaps))
+				list_add(&ip->pending_mmaps,
+					 &dev->pending_mmaps);
+			spin_unlock_irq(&dev->pending_lock);
+		}
+	} else if (attr_mask & IB_SRQ_LIMIT) {
+		spin_lock_irq(&srq->rq.lock);
+		if (attr->srq_limit >= srq->rq.size)
+			ret = -EINVAL;
+		else
+			srq->limit = attr->srq_limit;
+		spin_unlock_irq(&srq->rq.lock);
+	}
+	goto bail;
+
+bail_unlock:
+	spin_unlock_irq(&srq->rq.lock);
+bail_free:
+	vfree(wq);
+bail:
+	return ret;
+}
+
+int qib_query_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr)
+{
+	struct qib_srq *srq = to_isrq(ibsrq);
+
+	attr->max_wr = srq->rq.size - 1;
+	attr->max_sge = srq->rq.max_sge;
+	attr->srq_limit = srq->limit;
+	return 0;
+}
+
+/**
+ * qib_destroy_srq - destroy a shared receive queue
+ * @ibsrq: the SRQ to destroy
+ */
+int qib_destroy_srq(struct ib_srq *ibsrq)
+{
+	struct qib_srq *srq = to_isrq(ibsrq);
+	struct qib_ibdev *dev = to_idev(ibsrq->device);
+
+	spin_lock(&dev->n_srqs_lock);
+	dev->n_srqs_allocated--;
+	spin_unlock(&dev->n_srqs_lock);
+	if (srq->ip)
+		kref_put(&srq->ip->ref, qib_release_mmap_info);
+	else
+		vfree(srq->rq.wq);
+	kfree(srq);
+
+	return 0;
+}
diff --git a/drivers/infiniband/hw/qib/qib_sysfs.c b/drivers/infiniband/hw/qib/qib_sysfs.c
new file mode 100644
index 000000000000..dab4d9f4a2cc
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_sysfs.c
@@ -0,0 +1,691 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/ctype.h>
+
+#include "qib.h"
+
+/**
+ * qib_parse_ushort - parse an unsigned short value in an arbitrary base
+ * @str: the string containing the number
+ * @valp: where to put the result
+ *
+ * Returns the number of bytes consumed, or negative value on error.
+ */
+static int qib_parse_ushort(const char *str, unsigned short *valp)
+{
+	unsigned long val;
+	char *end;
+	int ret;
+
+	if (!isdigit(str[0])) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	val = simple_strtoul(str, &end, 0);
+
+	if (val > 0xffff) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	*valp = val;
+
+	ret = end + 1 - str;
+	if (ret == 0)
+		ret = -EINVAL;
+
+bail:
+	return ret;
+}
+
+/* start of per-port functions */
+/*
+ * Get/Set heartbeat enable. OR of 1=enabled, 2=auto
+ */
+static ssize_t show_hrtbt_enb(struct qib_pportdata *ppd, char *buf)
+{
+	struct qib_devdata *dd = ppd->dd;
+	int ret;
+
+	ret = dd->f_get_ib_cfg(ppd, QIB_IB_CFG_HRTBT);
+	ret = scnprintf(buf, PAGE_SIZE, "%d\n", ret);
+	return ret;
+}
+
+static ssize_t store_hrtbt_enb(struct qib_pportdata *ppd, const char *buf,
+			       size_t count)
+{
+	struct qib_devdata *dd = ppd->dd;
+	int ret;
+	u16 val;
+
+	ret = qib_parse_ushort(buf, &val);
+
+	/*
+	 * Set the "intentional" heartbeat enable per either of
+	 * "Enable" and "Auto", as these are normally set together.
+	 * This bit is consulted when leaving loopback mode,
+	 * because entering loopback mode overrides it and automatically
+	 * disables heartbeat.
+	 */
+	if (ret >= 0)
+		ret = dd->f_set_ib_cfg(ppd, QIB_IB_CFG_HRTBT, val);
+	if (ret < 0)
+		qib_dev_err(dd, "attempt to set invalid Heartbeat enable\n");
+	return ret < 0 ? ret : count;
+}
+
+static ssize_t store_loopback(struct qib_pportdata *ppd, const char *buf,
+			      size_t count)
+{
+	struct qib_devdata *dd = ppd->dd;
+	int ret = count, r;
+
+	r = dd->f_set_ib_loopback(ppd, buf);
+	if (r < 0)
+		ret = r;
+
+	return ret;
+}
+
+static ssize_t store_led_override(struct qib_pportdata *ppd, const char *buf,
+				  size_t count)
+{
+	struct qib_devdata *dd = ppd->dd;
+	int ret;
+	u16 val;
+
+	ret = qib_parse_ushort(buf, &val);
+	if (ret > 0)
+		qib_set_led_override(ppd, val);
+	else
+		qib_dev_err(dd, "attempt to set invalid LED override\n");
+	return ret < 0 ? ret : count;
+}
+
+static ssize_t show_status(struct qib_pportdata *ppd, char *buf)
+{
+	ssize_t ret;
+
+	if (!ppd->statusp)
+		ret = -EINVAL;
+	else
+		ret = scnprintf(buf, PAGE_SIZE, "0x%llx\n",
+				(unsigned long long) *(ppd->statusp));
+	return ret;
+}
+
+/*
+ * For userland compatibility, these offsets must remain fixed.
+ * They are strings for QIB_STATUS_*
+ */
+static const char *qib_status_str[] = {
+	"Initted",
+	"",
+	"",
+	"",
+	"",
+	"Present",
+	"IB_link_up",
+	"IB_configured",
+	"",
+	"Fatal_Hardware_Error",
+	NULL,
+};
+
+static ssize_t show_status_str(struct qib_pportdata *ppd, char *buf)
+{
+	int i, any;
+	u64 s;
+	ssize_t ret;
+
+	if (!ppd->statusp) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	s = *(ppd->statusp);
+	*buf = '\0';
+	for (any = i = 0; s && qib_status_str[i]; i++) {
+		if (s & 1) {
+			/* if overflow */
+			if (any && strlcat(buf, " ", PAGE_SIZE) >= PAGE_SIZE)
+				break;
+			if (strlcat(buf, qib_status_str[i], PAGE_SIZE) >=
+					PAGE_SIZE)
+				break;
+			any = 1;
+		}
+		s >>= 1;
+	}
+	if (any)
+		strlcat(buf, "\n", PAGE_SIZE);
+
+	ret = strlen(buf);
+
+bail:
+	return ret;
+}
+
+/* end of per-port functions */
+
+/*
+ * Start of per-port file structures and support code
+ * Because we are fitting into other infrastructure, we have to supply the
+ * full set of kobject/sysfs_ops structures and routines.
+ */
+#define QIB_PORT_ATTR(name, mode, show, store) \
+	static struct qib_port_attr qib_port_attr_##name = \
+		__ATTR(name, mode, show, store)
+
+struct qib_port_attr {
+	struct attribute attr;
+	ssize_t (*show)(struct qib_pportdata *, char *);
+	ssize_t (*store)(struct qib_pportdata *, const char *, size_t);
+};
+
+QIB_PORT_ATTR(loopback, S_IWUSR, NULL, store_loopback);
+QIB_PORT_ATTR(led_override, S_IWUSR, NULL, store_led_override);
+QIB_PORT_ATTR(hrtbt_enable, S_IWUSR | S_IRUGO, show_hrtbt_enb,
+	      store_hrtbt_enb);
+QIB_PORT_ATTR(status, S_IRUGO, show_status, NULL);
+QIB_PORT_ATTR(status_str, S_IRUGO, show_status_str, NULL);
+
+static struct attribute *port_default_attributes[] = {
+	&qib_port_attr_loopback.attr,
+	&qib_port_attr_led_override.attr,
+	&qib_port_attr_hrtbt_enable.attr,
+	&qib_port_attr_status.attr,
+	&qib_port_attr_status_str.attr,
+	NULL
+};
+
+static ssize_t qib_portattr_show(struct kobject *kobj,
+	struct attribute *attr, char *buf)
+{
+	struct qib_port_attr *pattr =
+		container_of(attr, struct qib_port_attr, attr);
+	struct qib_pportdata *ppd =
+		container_of(kobj, struct qib_pportdata, pport_kobj);
+
+	return pattr->show(ppd, buf);
+}
+
+static ssize_t qib_portattr_store(struct kobject *kobj,
+	struct attribute *attr, const char *buf, size_t len)
+{
+	struct qib_port_attr *pattr =
+		container_of(attr, struct qib_port_attr, attr);
+	struct qib_pportdata *ppd =
+		container_of(kobj, struct qib_pportdata, pport_kobj);
+
+	return pattr->store(ppd, buf, len);
+}
+
+static void qib_port_release(struct kobject *kobj)
+{
+	/* nothing to do since memory is freed by qib_free_devdata() */
+}
+
+static const struct sysfs_ops qib_port_ops = {
+	.show = qib_portattr_show,
+	.store = qib_portattr_store,
+};
+
+static struct kobj_type qib_port_ktype = {
+	.release = qib_port_release,
+	.sysfs_ops = &qib_port_ops,
+	.default_attrs = port_default_attributes
+};
+
+/* Start sl2vl */
+
+#define QIB_SL2VL_ATTR(N) \
+	static struct qib_sl2vl_attr qib_sl2vl_attr_##N = { \
+		.attr = { .name = __stringify(N), .mode = 0444 }, \
+		.sl = N \
+	}
+
+struct qib_sl2vl_attr {
+	struct attribute attr;
+	int sl;
+};
+
+QIB_SL2VL_ATTR(0);
+QIB_SL2VL_ATTR(1);
+QIB_SL2VL_ATTR(2);
+QIB_SL2VL_ATTR(3);
+QIB_SL2VL_ATTR(4);
+QIB_SL2VL_ATTR(5);
+QIB_SL2VL_ATTR(6);
+QIB_SL2VL_ATTR(7);
+QIB_SL2VL_ATTR(8);
+QIB_SL2VL_ATTR(9);
+QIB_SL2VL_ATTR(10);
+QIB_SL2VL_ATTR(11);
+QIB_SL2VL_ATTR(12);
+QIB_SL2VL_ATTR(13);
+QIB_SL2VL_ATTR(14);
+QIB_SL2VL_ATTR(15);
+
+static struct attribute *sl2vl_default_attributes[] = {
+	&qib_sl2vl_attr_0.attr,
+	&qib_sl2vl_attr_1.attr,
+	&qib_sl2vl_attr_2.attr,
+	&qib_sl2vl_attr_3.attr,
+	&qib_sl2vl_attr_4.attr,
+	&qib_sl2vl_attr_5.attr,
+	&qib_sl2vl_attr_6.attr,
+	&qib_sl2vl_attr_7.attr,
+	&qib_sl2vl_attr_8.attr,
+	&qib_sl2vl_attr_9.attr,
+	&qib_sl2vl_attr_10.attr,
+	&qib_sl2vl_attr_11.attr,
+	&qib_sl2vl_attr_12.attr,
+	&qib_sl2vl_attr_13.attr,
+	&qib_sl2vl_attr_14.attr,
+	&qib_sl2vl_attr_15.attr,
+	NULL
+};
+
+static ssize_t sl2vl_attr_show(struct kobject *kobj, struct attribute *attr,
+			       char *buf)
+{
+	struct qib_sl2vl_attr *sattr =
+		container_of(attr, struct qib_sl2vl_attr, attr);
+	struct qib_pportdata *ppd =
+		container_of(kobj, struct qib_pportdata, sl2vl_kobj);
+	struct qib_ibport *qibp = &ppd->ibport_data;
+
+	return sprintf(buf, "%u\n", qibp->sl_to_vl[sattr->sl]);
+}
+
+static const struct sysfs_ops qib_sl2vl_ops = {
+	.show = sl2vl_attr_show,
+};
+
+static struct kobj_type qib_sl2vl_ktype = {
+	.release = qib_port_release,
+	.sysfs_ops = &qib_sl2vl_ops,
+	.default_attrs = sl2vl_default_attributes
+};
+
+/* End sl2vl */
+
+/* Start diag_counters */
+
+#define QIB_DIAGC_ATTR(N) \
+	static struct qib_diagc_attr qib_diagc_attr_##N = { \
+		.attr = { .name = __stringify(N), .mode = 0444 }, \
+		.counter = offsetof(struct qib_ibport, n_##N) \
+	}
+
+struct qib_diagc_attr {
+	struct attribute attr;
+	size_t counter;
+};
+
+QIB_DIAGC_ATTR(rc_resends);
+QIB_DIAGC_ATTR(rc_acks);
+QIB_DIAGC_ATTR(rc_qacks);
+QIB_DIAGC_ATTR(rc_delayed_comp);
+QIB_DIAGC_ATTR(seq_naks);
+QIB_DIAGC_ATTR(rdma_seq);
+QIB_DIAGC_ATTR(rnr_naks);
+QIB_DIAGC_ATTR(other_naks);
+QIB_DIAGC_ATTR(rc_timeouts);
+QIB_DIAGC_ATTR(loop_pkts);
+QIB_DIAGC_ATTR(pkt_drops);
+QIB_DIAGC_ATTR(dmawait);
+QIB_DIAGC_ATTR(unaligned);
+QIB_DIAGC_ATTR(rc_dupreq);
+QIB_DIAGC_ATTR(rc_seqnak);
+
+static struct attribute *diagc_default_attributes[] = {
+	&qib_diagc_attr_rc_resends.attr,
+	&qib_diagc_attr_rc_acks.attr,
+	&qib_diagc_attr_rc_qacks.attr,
+	&qib_diagc_attr_rc_delayed_comp.attr,
+	&qib_diagc_attr_seq_naks.attr,
+	&qib_diagc_attr_rdma_seq.attr,
+	&qib_diagc_attr_rnr_naks.attr,
+	&qib_diagc_attr_other_naks.attr,
+	&qib_diagc_attr_rc_timeouts.attr,
+	&qib_diagc_attr_loop_pkts.attr,
+	&qib_diagc_attr_pkt_drops.attr,
+	&qib_diagc_attr_dmawait.attr,
+	&qib_diagc_attr_unaligned.attr,
+	&qib_diagc_attr_rc_dupreq.attr,
+	&qib_diagc_attr_rc_seqnak.attr,
+	NULL
+};
+
+static ssize_t diagc_attr_show(struct kobject *kobj, struct attribute *attr,
+			       char *buf)
+{
+	struct qib_diagc_attr *dattr =
+		container_of(attr, struct qib_diagc_attr, attr);
+	struct qib_pportdata *ppd =
+		container_of(kobj, struct qib_pportdata, diagc_kobj);
+	struct qib_ibport *qibp = &ppd->ibport_data;
+
+	return sprintf(buf, "%u\n", *(u32 *)((char *)qibp + dattr->counter));
+}
+
+static const struct sysfs_ops qib_diagc_ops = {
+	.show = diagc_attr_show,
+};
+
+static struct kobj_type qib_diagc_ktype = {
+	.release = qib_port_release,
+	.sysfs_ops = &qib_diagc_ops,
+	.default_attrs = diagc_default_attributes
+};
+
+/* End diag_counters */
+
+/* end of per-port file structures and support code */
+
+/*
+ * Start of per-unit (or driver, in some cases, but replicated
+ * per unit) functions (these get a device *)
+ */
+static ssize_t show_rev(struct device *device, struct device_attribute *attr,
+			char *buf)
+{
+	struct qib_ibdev *dev =
+		container_of(device, struct qib_ibdev, ibdev.dev);
+
+	return sprintf(buf, "%x\n", dd_from_dev(dev)->minrev);
+}
+
+static ssize_t show_hca(struct device *device, struct device_attribute *attr,
+			char *buf)
+{
+	struct qib_ibdev *dev =
+		container_of(device, struct qib_ibdev, ibdev.dev);
+	struct qib_devdata *dd = dd_from_dev(dev);
+	int ret;
+
+	if (!dd->boardname)
+		ret = -EINVAL;
+	else
+		ret = scnprintf(buf, PAGE_SIZE, "%s\n", dd->boardname);
+	return ret;
+}
+
+static ssize_t show_version(struct device *device,
+			    struct device_attribute *attr, char *buf)
+{
+	/* The string printed here is already newline-terminated. */
+	return scnprintf(buf, PAGE_SIZE, "%s", (char *)ib_qib_version);
+}
+
+static ssize_t show_boardversion(struct device *device,
+				 struct device_attribute *attr, char *buf)
+{
+	struct qib_ibdev *dev =
+		container_of(device, struct qib_ibdev, ibdev.dev);
+	struct qib_devdata *dd = dd_from_dev(dev);
+
+	/* The string printed here is already newline-terminated. */
+	return scnprintf(buf, PAGE_SIZE, "%s", dd->boardversion);
+}
+
+
+static ssize_t show_localbus_info(struct device *device,
+				  struct device_attribute *attr, char *buf)
+{
+	struct qib_ibdev *dev =
+		container_of(device, struct qib_ibdev, ibdev.dev);
+	struct qib_devdata *dd = dd_from_dev(dev);
+
+	/* The string printed here is already newline-terminated. */
+	return scnprintf(buf, PAGE_SIZE, "%s", dd->lbus_info);
+}
+
+
+static ssize_t show_nctxts(struct device *device,
+			   struct device_attribute *attr, char *buf)
+{
+	struct qib_ibdev *dev =
+		container_of(device, struct qib_ibdev, ibdev.dev);
+	struct qib_devdata *dd = dd_from_dev(dev);
+
+	/* Return the number of user ports (contexts) available. */
+	return scnprintf(buf, PAGE_SIZE, "%u\n", dd->cfgctxts -
+		dd->first_user_ctxt);
+}
+
+static ssize_t show_serial(struct device *device,
+			   struct device_attribute *attr, char *buf)
+{
+	struct qib_ibdev *dev =
+		container_of(device, struct qib_ibdev, ibdev.dev);
+	struct qib_devdata *dd = dd_from_dev(dev);
+
+	buf[sizeof dd->serial] = '\0';
+	memcpy(buf, dd->serial, sizeof dd->serial);
+	strcat(buf, "\n");
+	return strlen(buf);
+}
+
+static ssize_t store_chip_reset(struct device *device,
+				struct device_attribute *attr, const char *buf,
+				size_t count)
+{
+	struct qib_ibdev *dev =
+		container_of(device, struct qib_ibdev, ibdev.dev);
+	struct qib_devdata *dd = dd_from_dev(dev);
+	int ret;
+
+	if (count < 5 || memcmp(buf, "reset", 5) || !dd->diag_client) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	ret = qib_reset_device(dd->unit);
+bail:
+	return ret < 0 ? ret : count;
+}
+
+static ssize_t show_logged_errs(struct device *device,
+				struct device_attribute *attr, char *buf)
+{
+	struct qib_ibdev *dev =
+		container_of(device, struct qib_ibdev, ibdev.dev);
+	struct qib_devdata *dd = dd_from_dev(dev);
+	int idx, count;
+
+	/* force consistency with actual EEPROM */
+	if (qib_update_eeprom_log(dd) != 0)
+		return -ENXIO;
+
+	count = 0;
+	for (idx = 0; idx < QIB_EEP_LOG_CNT; ++idx) {
+		count += scnprintf(buf + count, PAGE_SIZE - count, "%d%c",
+				   dd->eep_st_errs[idx],
+				   idx == (QIB_EEP_LOG_CNT - 1) ? '\n' : ' ');
+	}
+
+	return count;
+}
+
+/*
+ * Dump tempsense regs. in decimal, to ease shell-scripts.
+ */
+static ssize_t show_tempsense(struct device *device,
+			      struct device_attribute *attr, char *buf)
+{
+	struct qib_ibdev *dev =
+		container_of(device, struct qib_ibdev, ibdev.dev);
+	struct qib_devdata *dd = dd_from_dev(dev);
+	int ret;
+	int idx;
+	u8 regvals[8];
+
+	ret = -ENXIO;
+	for (idx = 0; idx < 8; ++idx) {
+		if (idx == 6)
+			continue;
+		ret = dd->f_tempsense_rd(dd, idx);
+		if (ret < 0)
+			break;
+		regvals[idx] = ret;
+	}
+	if (idx == 8)
+		ret = scnprintf(buf, PAGE_SIZE, "%d %d %02X %02X %d %d\n",
+				*(signed char *)(regvals),
+				*(signed char *)(regvals + 1),
+				regvals[2], regvals[3],
+				*(signed char *)(regvals + 5),
+				*(signed char *)(regvals + 7));
+	return ret;
+}
+
+/*
+ * end of per-unit (or driver, in some cases, but replicated
+ * per unit) functions
+ */
+
+/* start of per-unit file structures and support code */
+static DEVICE_ATTR(hw_rev, S_IRUGO, show_rev, NULL);
+static DEVICE_ATTR(hca_type, S_IRUGO, show_hca, NULL);
+static DEVICE_ATTR(board_id, S_IRUGO, show_hca, NULL);
+static DEVICE_ATTR(version, S_IRUGO, show_version, NULL);
+static DEVICE_ATTR(nctxts, S_IRUGO, show_nctxts, NULL);
+static DEVICE_ATTR(serial, S_IRUGO, show_serial, NULL);
+static DEVICE_ATTR(boardversion, S_IRUGO, show_boardversion, NULL);
+static DEVICE_ATTR(logged_errors, S_IRUGO, show_logged_errs, NULL);
+static DEVICE_ATTR(tempsense, S_IRUGO, show_tempsense, NULL);
+static DEVICE_ATTR(localbus_info, S_IRUGO, show_localbus_info, NULL);
+static DEVICE_ATTR(chip_reset, S_IWUSR, NULL, store_chip_reset);
+
+static struct device_attribute *qib_attributes[] = {
+	&dev_attr_hw_rev,
+	&dev_attr_hca_type,
+	&dev_attr_board_id,
+	&dev_attr_version,
+	&dev_attr_nctxts,
+	&dev_attr_serial,
+	&dev_attr_boardversion,
+	&dev_attr_logged_errors,
+	&dev_attr_tempsense,
+	&dev_attr_localbus_info,
+	&dev_attr_chip_reset,
+};
+
+int qib_create_port_files(struct ib_device *ibdev, u8 port_num,
+			  struct kobject *kobj)
+{
+	struct qib_pportdata *ppd;
+	struct qib_devdata *dd = dd_from_ibdev(ibdev);
+	int ret;
+
+	if (!port_num || port_num > dd->num_pports) {
+		qib_dev_err(dd, "Skipping infiniband class with "
+			    "invalid port %u\n", port_num);
+		ret = -ENODEV;
+		goto bail;
+	}
+	ppd = &dd->pport[port_num - 1];
+
+	ret = kobject_init_and_add(&ppd->pport_kobj, &qib_port_ktype, kobj,
+				   "linkcontrol");
+	if (ret) {
+		qib_dev_err(dd, "Skipping linkcontrol sysfs info, "
+			    "(err %d) port %u\n", ret, port_num);
+		goto bail;
+	}
+	kobject_uevent(&ppd->pport_kobj, KOBJ_ADD);
+
+	ret = kobject_init_and_add(&ppd->sl2vl_kobj, &qib_sl2vl_ktype, kobj,
+				   "sl2vl");
+	if (ret) {
+		qib_dev_err(dd, "Skipping sl2vl sysfs info, "
+			    "(err %d) port %u\n", ret, port_num);
+		goto bail_sl;
+	}
+	kobject_uevent(&ppd->sl2vl_kobj, KOBJ_ADD);
+
+	ret = kobject_init_and_add(&ppd->diagc_kobj, &qib_diagc_ktype, kobj,
+				   "diag_counters");
+	if (ret) {
+		qib_dev_err(dd, "Skipping diag_counters sysfs info, "
+			    "(err %d) port %u\n", ret, port_num);
+		goto bail_diagc;
+	}
+	kobject_uevent(&ppd->diagc_kobj, KOBJ_ADD);
+
+	return 0;
+
+bail_diagc:
+	kobject_put(&ppd->sl2vl_kobj);
+bail_sl:
+	kobject_put(&ppd->pport_kobj);
+bail:
+	return ret;
+}
+
+/*
+ * Register and create our files in /sys/class/infiniband.
+ */
+int qib_verbs_register_sysfs(struct qib_devdata *dd)
+{
+	struct ib_device *dev = &dd->verbs_dev.ibdev;
+	int i, ret;
+
+	for (i = 0; i < ARRAY_SIZE(qib_attributes); ++i) {
+		ret = device_create_file(&dev->dev, qib_attributes[i]);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
+/*
+ * Unregister and remove our files in /sys/class/infiniband.
+ */
+void qib_verbs_unregister_sysfs(struct qib_devdata *dd)
+{
+	struct qib_pportdata *ppd;
+	int i;
+
+	for (i = 0; i < dd->num_pports; i++) {
+		ppd = &dd->pport[i];
+		kobject_put(&ppd->pport_kobj);
+		kobject_put(&ppd->sl2vl_kobj);
+	}
+}
diff --git a/drivers/infiniband/hw/qib/qib_twsi.c b/drivers/infiniband/hw/qib/qib_twsi.c
new file mode 100644
index 000000000000..6f31ca5039db
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_twsi.c
@@ -0,0 +1,498 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/delay.h>
+#include <linux/pci.h>
+#include <linux/vmalloc.h>
+
+#include "qib.h"
+
+/*
+ * QLogic_IB "Two Wire Serial Interface" driver.
+ * Originally written for a not-quite-i2c serial eeprom, which is
+ * still used on some supported boards. Later boards have added a
+ * variety of other uses, most board-specific, so teh bit-boffing
+ * part has been split off to this file, while the other parts
+ * have been moved to chip-specific files.
+ *
+ * We have also dropped all pretense of fully generic (e.g. pretend
+ * we don't know whether '1' is the higher voltage) interface, as
+ * the restrictions of the generic i2c interface (e.g. no access from
+ * driver itself) make it unsuitable for this use.
+ */
+
+#define READ_CMD 1
+#define WRITE_CMD 0
+
+/**
+ * i2c_wait_for_writes - wait for a write
+ * @dd: the qlogic_ib device
+ *
+ * We use this instead of udelay directly, so we can make sure
+ * that previous register writes have been flushed all the way
+ * to the chip.  Since we are delaying anyway, the cost doesn't
+ * hurt, and makes the bit twiddling more regular
+ */
+static void i2c_wait_for_writes(struct qib_devdata *dd)
+{
+	/*
+	 * implicit read of EXTStatus is as good as explicit
+	 * read of scratch, if all we want to do is flush
+	 * writes.
+	 */
+	dd->f_gpio_mod(dd, 0, 0, 0);
+	rmb(); /* inlined, so prevent compiler reordering */
+}
+
+/*
+ * QSFP modules are allowed to hold SCL low for 500uSec. Allow twice that
+ * for "almost compliant" modules
+ */
+#define SCL_WAIT_USEC 1000
+
+/* BUF_WAIT is time bus must be free between STOP or ACK and to next START.
+ * Should be 20, but some chips need more.
+ */
+#define TWSI_BUF_WAIT_USEC 60
+
+static void scl_out(struct qib_devdata *dd, u8 bit)
+{
+	u32 mask;
+
+	udelay(1);
+
+	mask = 1UL << dd->gpio_scl_num;
+
+	/* SCL is meant to be bare-drain, so never set "OUT", just DIR */
+	dd->f_gpio_mod(dd, 0, bit ? 0 : mask, mask);
+
+	/*
+	 * Allow for slow slaves by simple
+	 * delay for falling edge, sampling on rise.
+	 */
+	if (!bit)
+		udelay(2);
+	else {
+		int rise_usec;
+		for (rise_usec = SCL_WAIT_USEC; rise_usec > 0; rise_usec -= 2) {
+			if (mask & dd->f_gpio_mod(dd, 0, 0, 0))
+				break;
+			udelay(2);
+		}
+		if (rise_usec <= 0)
+			qib_dev_err(dd, "SCL interface stuck low > %d uSec\n",
+				    SCL_WAIT_USEC);
+	}
+	i2c_wait_for_writes(dd);
+}
+
+static void sda_out(struct qib_devdata *dd, u8 bit)
+{
+	u32 mask;
+
+	mask = 1UL << dd->gpio_sda_num;
+
+	/* SDA is meant to be bare-drain, so never set "OUT", just DIR */
+	dd->f_gpio_mod(dd, 0, bit ? 0 : mask, mask);
+
+	i2c_wait_for_writes(dd);
+	udelay(2);
+}
+
+static u8 sda_in(struct qib_devdata *dd, int wait)
+{
+	int bnum;
+	u32 read_val, mask;
+
+	bnum = dd->gpio_sda_num;
+	mask = (1UL << bnum);
+	/* SDA is meant to be bare-drain, so never set "OUT", just DIR */
+	dd->f_gpio_mod(dd, 0, 0, mask);
+	read_val = dd->f_gpio_mod(dd, 0, 0, 0);
+	if (wait)
+		i2c_wait_for_writes(dd);
+	return (read_val & mask) >> bnum;
+}
+
+/**
+ * i2c_ackrcv - see if ack following write is true
+ * @dd: the qlogic_ib device
+ */
+static int i2c_ackrcv(struct qib_devdata *dd)
+{
+	u8 ack_received;
+
+	/* AT ENTRY SCL = LOW */
+	/* change direction, ignore data */
+	ack_received = sda_in(dd, 1);
+	scl_out(dd, 1);
+	ack_received = sda_in(dd, 1) == 0;
+	scl_out(dd, 0);
+	return ack_received;
+}
+
+static void stop_cmd(struct qib_devdata *dd);
+
+/**
+ * rd_byte - read a byte, sending STOP on last, else ACK
+ * @dd: the qlogic_ib device
+ *
+ * Returns byte shifted out of device
+ */
+static int rd_byte(struct qib_devdata *dd, int last)
+{
+	int bit_cntr, data;
+
+	data = 0;
+
+	for (bit_cntr = 7; bit_cntr >= 0; --bit_cntr) {
+		data <<= 1;
+		scl_out(dd, 1);
+		data |= sda_in(dd, 0);
+		scl_out(dd, 0);
+	}
+	if (last) {
+		scl_out(dd, 1);
+		stop_cmd(dd);
+	} else {
+		sda_out(dd, 0);
+		scl_out(dd, 1);
+		scl_out(dd, 0);
+		sda_out(dd, 1);
+	}
+	return data;
+}
+
+/**
+ * wr_byte - write a byte, one bit at a time
+ * @dd: the qlogic_ib device
+ * @data: the byte to write
+ *
+ * Returns 0 if we got the following ack, otherwise 1
+ */
+static int wr_byte(struct qib_devdata *dd, u8 data)
+{
+	int bit_cntr;
+	u8 bit;
+
+	for (bit_cntr = 7; bit_cntr >= 0; bit_cntr--) {
+		bit = (data >> bit_cntr) & 1;
+		sda_out(dd, bit);
+		scl_out(dd, 1);
+		scl_out(dd, 0);
+	}
+	return (!i2c_ackrcv(dd)) ? 1 : 0;
+}
+
+/*
+ * issue TWSI start sequence:
+ * (both clock/data high, clock high, data low while clock is high)
+ */
+static void start_seq(struct qib_devdata *dd)
+{
+	sda_out(dd, 1);
+	scl_out(dd, 1);
+	sda_out(dd, 0);
+	udelay(1);
+	scl_out(dd, 0);
+}
+
+/**
+ * stop_seq - transmit the stop sequence
+ * @dd: the qlogic_ib device
+ *
+ * (both clock/data low, clock high, data high while clock is high)
+ */
+static void stop_seq(struct qib_devdata *dd)
+{
+	scl_out(dd, 0);
+	sda_out(dd, 0);
+	scl_out(dd, 1);
+	sda_out(dd, 1);
+}
+
+/**
+ * stop_cmd - transmit the stop condition
+ * @dd: the qlogic_ib device
+ *
+ * (both clock/data low, clock high, data high while clock is high)
+ */
+static void stop_cmd(struct qib_devdata *dd)
+{
+	stop_seq(dd);
+	udelay(TWSI_BUF_WAIT_USEC);
+}
+
+/**
+ * qib_twsi_reset - reset I2C communication
+ * @dd: the qlogic_ib device
+ */
+
+int qib_twsi_reset(struct qib_devdata *dd)
+{
+	int clock_cycles_left = 9;
+	int was_high = 0;
+	u32 pins, mask;
+
+	/* Both SCL and SDA should be high. If not, there
+	 * is something wrong.
+	 */
+	mask = (1UL << dd->gpio_scl_num) | (1UL << dd->gpio_sda_num);
+
+	/*
+	 * Force pins to desired innocuous state.
+	 * This is the default power-on state with out=0 and dir=0,
+	 * So tri-stated and should be floating high (barring HW problems)
+	 */
+	dd->f_gpio_mod(dd, 0, 0, mask);
+
+	/*
+	 * Clock nine times to get all listeners into a sane state.
+	 * If SDA does not go high at any point, we are wedged.
+	 * One vendor recommends then issuing START followed by STOP.
+	 * we cannot use our "normal" functions to do that, because
+	 * if SCL drops between them, another vendor's part will
+	 * wedge, dropping SDA and keeping it low forever, at the end of
+	 * the next transaction (even if it was not the device addressed).
+	 * So our START and STOP take place with SCL held high.
+	 */
+	while (clock_cycles_left--) {
+		scl_out(dd, 0);
+		scl_out(dd, 1);
+		/* Note if SDA is high, but keep clocking to sync slave */
+		was_high |= sda_in(dd, 0);
+	}
+
+	if (was_high) {
+		/*
+		 * We saw a high, which we hope means the slave is sync'd.
+		 * Issue START, STOP, pause for T_BUF.
+		 */
+
+		pins = dd->f_gpio_mod(dd, 0, 0, 0);
+		if ((pins & mask) != mask)
+			qib_dev_err(dd, "GPIO pins not at rest: %d\n",
+				    pins & mask);
+		/* Drop SDA to issue START */
+		udelay(1); /* Guarantee .6 uSec setup */
+		sda_out(dd, 0);
+		udelay(1); /* Guarantee .6 uSec hold */
+		/* At this point, SCL is high, SDA low. Raise SDA for STOP */
+		sda_out(dd, 1);
+		udelay(TWSI_BUF_WAIT_USEC);
+	}
+
+	return !was_high;
+}
+
+#define QIB_TWSI_START 0x100
+#define QIB_TWSI_STOP 0x200
+
+/* Write byte to TWSI, optionally prefixed with START or suffixed with
+ * STOP.
+ * returns 0 if OK (ACK received), else != 0
+ */
+static int qib_twsi_wr(struct qib_devdata *dd, int data, int flags)
+{
+	int ret = 1;
+	if (flags & QIB_TWSI_START)
+		start_seq(dd);
+
+	ret = wr_byte(dd, data); /* Leaves SCL low (from i2c_ackrcv()) */
+
+	if (flags & QIB_TWSI_STOP)
+		stop_cmd(dd);
+	return ret;
+}
+
+/* Added functionality for IBA7220-based cards */
+#define QIB_TEMP_DEV 0x98
+
+/*
+ * qib_twsi_blk_rd
+ * Formerly called qib_eeprom_internal_read, and only used for eeprom,
+ * but now the general interface for data transfer from twsi devices.
+ * One vestige of its former role is that it recognizes a device
+ * QIB_TWSI_NO_DEV and does the correct operation for the legacy part,
+ * which responded to all TWSI device codes, interpreting them as
+ * address within device. On all other devices found on board handled by
+ * this driver, the device is followed by a one-byte "address" which selects
+ * the "register" or "offset" within the device from which data should
+ * be read.
+ */
+int qib_twsi_blk_rd(struct qib_devdata *dd, int dev, int addr,
+		    void *buffer, int len)
+{
+	int ret;
+	u8 *bp = buffer;
+
+	ret = 1;
+
+	if (dev == QIB_TWSI_NO_DEV) {
+		/* legacy not-really-I2C */
+		addr = (addr << 1) | READ_CMD;
+		ret = qib_twsi_wr(dd, addr, QIB_TWSI_START);
+	} else {
+		/* Actual I2C */
+		ret = qib_twsi_wr(dd, dev | WRITE_CMD, QIB_TWSI_START);
+		if (ret) {
+			stop_cmd(dd);
+			ret = 1;
+			goto bail;
+		}
+		/*
+		 * SFF spec claims we do _not_ stop after the addr
+		 * but simply issue a start with the "read" dev-addr.
+		 * Since we are implicitely waiting for ACK here,
+		 * we need t_buf (nominally 20uSec) before that start,
+		 * and cannot rely on the delay built in to the STOP
+		 */
+		ret = qib_twsi_wr(dd, addr, 0);
+		udelay(TWSI_BUF_WAIT_USEC);
+
+		if (ret) {
+			qib_dev_err(dd,
+				"Failed to write interface read addr %02X\n",
+				addr);
+			ret = 1;
+			goto bail;
+		}
+		ret = qib_twsi_wr(dd, dev | READ_CMD, QIB_TWSI_START);
+	}
+	if (ret) {
+		stop_cmd(dd);
+		ret = 1;
+		goto bail;
+	}
+
+	/*
+	 * block devices keeps clocking data out as long as we ack,
+	 * automatically incrementing the address. Some have "pages"
+	 * whose boundaries will not be crossed, but the handling
+	 * of these is left to the caller, who is in a better
+	 * position to know.
+	 */
+	while (len-- > 0) {
+		/*
+		 * Get and store data, sending ACK if length remaining,
+		 * else STOP
+		 */
+		*bp++ = rd_byte(dd, !len);
+	}
+
+	ret = 0;
+
+bail:
+	return ret;
+}
+
+/*
+ * qib_twsi_blk_wr
+ * Formerly called qib_eeprom_internal_write, and only used for eeprom,
+ * but now the general interface for data transfer to twsi devices.
+ * One vestige of its former role is that it recognizes a device
+ * QIB_TWSI_NO_DEV and does the correct operation for the legacy part,
+ * which responded to all TWSI device codes, interpreting them as
+ * address within device. On all other devices found on board handled by
+ * this driver, the device is followed by a one-byte "address" which selects
+ * the "register" or "offset" within the device to which data should
+ * be written.
+ */
+int qib_twsi_blk_wr(struct qib_devdata *dd, int dev, int addr,
+		    const void *buffer, int len)
+{
+	int sub_len;
+	const u8 *bp = buffer;
+	int max_wait_time, i;
+	int ret;
+	ret = 1;
+
+	while (len > 0) {
+		if (dev == QIB_TWSI_NO_DEV) {
+			if (qib_twsi_wr(dd, (addr << 1) | WRITE_CMD,
+					QIB_TWSI_START)) {
+				goto failed_write;
+			}
+		} else {
+			/* Real I2C */
+			if (qib_twsi_wr(dd, dev | WRITE_CMD, QIB_TWSI_START))
+				goto failed_write;
+			ret = qib_twsi_wr(dd, addr, 0);
+			if (ret) {
+				qib_dev_err(dd, "Failed to write interface"
+					    " write addr %02X\n", addr);
+				goto failed_write;
+			}
+		}
+
+		sub_len = min(len, 4);
+		addr += sub_len;
+		len -= sub_len;
+
+		for (i = 0; i < sub_len; i++)
+			if (qib_twsi_wr(dd, *bp++, 0))
+				goto failed_write;
+
+		stop_cmd(dd);
+
+		/*
+		 * Wait for write complete by waiting for a successful
+		 * read (the chip replies with a zero after the write
+		 * cmd completes, and before it writes to the eeprom.
+		 * The startcmd for the read will fail the ack until
+		 * the writes have completed.   We do this inline to avoid
+		 * the debug prints that are in the real read routine
+		 * if the startcmd fails.
+		 * We also use the proper device address, so it doesn't matter
+		 * whether we have real eeprom_dev. Legacy likes any address.
+		 */
+		max_wait_time = 100;
+		while (qib_twsi_wr(dd, dev | READ_CMD, QIB_TWSI_START)) {
+			stop_cmd(dd);
+			if (!--max_wait_time)
+				goto failed_write;
+		}
+		/* now read (and ignore) the resulting byte */
+		rd_byte(dd, 1);
+	}
+
+	ret = 0;
+	goto bail;
+
+failed_write:
+	stop_cmd(dd);
+	ret = 1;
+
+bail:
+	return ret;
+}
diff --git a/drivers/infiniband/hw/qib/qib_tx.c b/drivers/infiniband/hw/qib/qib_tx.c
new file mode 100644
index 000000000000..f7eb1ddff5f3
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_tx.c
@@ -0,0 +1,557 @@
+/*
+ * Copyright (c) 2008, 2009, 2010 QLogic Corporation. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/spinlock.h>
+#include <linux/pci.h>
+#include <linux/io.h>
+#include <linux/delay.h>
+#include <linux/netdevice.h>
+#include <linux/vmalloc.h>
+
+#include "qib.h"
+
+static unsigned qib_hol_timeout_ms = 3000;
+module_param_named(hol_timeout_ms, qib_hol_timeout_ms, uint, S_IRUGO);
+MODULE_PARM_DESC(hol_timeout_ms,
+		 "duration of user app suspension after link failure");
+
+unsigned qib_sdma_fetch_arb = 1;
+module_param_named(fetch_arb, qib_sdma_fetch_arb, uint, S_IRUGO);
+MODULE_PARM_DESC(fetch_arb, "IBA7220: change SDMA descriptor arbitration");
+
+/**
+ * qib_disarm_piobufs - cancel a range of PIO buffers
+ * @dd: the qlogic_ib device
+ * @first: the first PIO buffer to cancel
+ * @cnt: the number of PIO buffers to cancel
+ *
+ * Cancel a range of PIO buffers. Used at user process close,
+ * in case it died while writing to a PIO buffer.
+ */
+void qib_disarm_piobufs(struct qib_devdata *dd, unsigned first, unsigned cnt)
+{
+	unsigned long flags;
+	unsigned i;
+	unsigned last;
+
+	last = first + cnt;
+	spin_lock_irqsave(&dd->pioavail_lock, flags);
+	for (i = first; i < last; i++) {
+		__clear_bit(i, dd->pio_need_disarm);
+		dd->f_sendctrl(dd->pport, QIB_SENDCTRL_DISARM_BUF(i));
+	}
+	spin_unlock_irqrestore(&dd->pioavail_lock, flags);
+}
+
+/*
+ * This is called by a user process when it sees the DISARM_BUFS event
+ * bit is set.
+ */
+int qib_disarm_piobufs_ifneeded(struct qib_ctxtdata *rcd)
+{
+	struct qib_devdata *dd = rcd->dd;
+	unsigned i;
+	unsigned last;
+	unsigned n = 0;
+
+	last = rcd->pio_base + rcd->piocnt;
+	/*
+	 * Don't need uctxt_lock here, since user has called in to us.
+	 * Clear at start in case more interrupts set bits while we
+	 * are disarming
+	 */
+	if (rcd->user_event_mask) {
+		/*
+		 * subctxt_cnt is 0 if not shared, so do base
+		 * separately, first, then remaining subctxt, if any
+		 */
+		clear_bit(_QIB_EVENT_DISARM_BUFS_BIT, &rcd->user_event_mask[0]);
+		for (i = 1; i < rcd->subctxt_cnt; i++)
+			clear_bit(_QIB_EVENT_DISARM_BUFS_BIT,
+				  &rcd->user_event_mask[i]);
+	}
+	spin_lock_irq(&dd->pioavail_lock);
+	for (i = rcd->pio_base; i < last; i++) {
+		if (__test_and_clear_bit(i, dd->pio_need_disarm)) {
+			n++;
+			dd->f_sendctrl(rcd->ppd, QIB_SENDCTRL_DISARM_BUF(i));
+		}
+	}
+	spin_unlock_irq(&dd->pioavail_lock);
+	return 0;
+}
+
+static struct qib_pportdata *is_sdma_buf(struct qib_devdata *dd, unsigned i)
+{
+	struct qib_pportdata *ppd;
+	unsigned pidx;
+
+	for (pidx = 0; pidx < dd->num_pports; pidx++) {
+		ppd = dd->pport + pidx;
+		if (i >= ppd->sdma_state.first_sendbuf &&
+		    i < ppd->sdma_state.last_sendbuf)
+			return ppd;
+	}
+	return NULL;
+}
+
+/*
+ * Return true if send buffer is being used by a user context.
+ * Sets  _QIB_EVENT_DISARM_BUFS_BIT in user_event_mask as a side effect
+ */
+static int find_ctxt(struct qib_devdata *dd, unsigned bufn)
+{
+	struct qib_ctxtdata *rcd;
+	unsigned ctxt;
+	int ret = 0;
+
+	spin_lock(&dd->uctxt_lock);
+	for (ctxt = dd->first_user_ctxt; ctxt < dd->cfgctxts; ctxt++) {
+		rcd = dd->rcd[ctxt];
+		if (!rcd || bufn < rcd->pio_base ||
+		    bufn >= rcd->pio_base + rcd->piocnt)
+			continue;
+		if (rcd->user_event_mask) {
+			int i;
+			/*
+			 * subctxt_cnt is 0 if not shared, so do base
+			 * separately, first, then remaining subctxt, if any
+			 */
+			set_bit(_QIB_EVENT_DISARM_BUFS_BIT,
+				&rcd->user_event_mask[0]);
+			for (i = 1; i < rcd->subctxt_cnt; i++)
+				set_bit(_QIB_EVENT_DISARM_BUFS_BIT,
+					&rcd->user_event_mask[i]);
+		}
+		ret = 1;
+		break;
+	}
+	spin_unlock(&dd->uctxt_lock);
+
+	return ret;
+}
+
+/*
+ * Disarm a set of send buffers.  If the buffer might be actively being
+ * written to, mark the buffer to be disarmed later when it is not being
+ * written to.
+ *
+ * This should only be called from the IRQ error handler.
+ */
+void qib_disarm_piobufs_set(struct qib_devdata *dd, unsigned long *mask,
+			    unsigned cnt)
+{
+	struct qib_pportdata *ppd, *pppd[dd->num_pports];
+	unsigned i;
+	unsigned long flags;
+
+	for (i = 0; i < dd->num_pports; i++)
+		pppd[i] = NULL;
+
+	for (i = 0; i < cnt; i++) {
+		int which;
+		if (!test_bit(i, mask))
+			continue;
+		/*
+		 * If the buffer is owned by the DMA hardware,
+		 * reset the DMA engine.
+		 */
+		ppd = is_sdma_buf(dd, i);
+		if (ppd) {
+			pppd[ppd->port] = ppd;
+			continue;
+		}
+		/*
+		 * If the kernel is writing the buffer or the buffer is
+		 * owned by a user process, we can't clear it yet.
+		 */
+		spin_lock_irqsave(&dd->pioavail_lock, flags);
+		if (test_bit(i, dd->pio_writing) ||
+		    (!test_bit(i << 1, dd->pioavailkernel) &&
+		     find_ctxt(dd, i))) {
+			__set_bit(i, dd->pio_need_disarm);
+			which = 0;
+		} else {
+			which = 1;
+			dd->f_sendctrl(dd->pport, QIB_SENDCTRL_DISARM_BUF(i));
+		}
+		spin_unlock_irqrestore(&dd->pioavail_lock, flags);
+	}
+
+	/* do cancel_sends once per port that had sdma piobufs in error */
+	for (i = 0; i < dd->num_pports; i++)
+		if (pppd[i])
+			qib_cancel_sends(pppd[i]);
+}
+
+/**
+ * update_send_bufs - update shadow copy of the PIO availability map
+ * @dd: the qlogic_ib device
+ *
+ * called whenever our local copy indicates we have run out of send buffers
+ */
+static void update_send_bufs(struct qib_devdata *dd)
+{
+	unsigned long flags;
+	unsigned i;
+	const unsigned piobregs = dd->pioavregs;
+
+	/*
+	 * If the generation (check) bits have changed, then we update the
+	 * busy bit for the corresponding PIO buffer.  This algorithm will
+	 * modify positions to the value they already have in some cases
+	 * (i.e., no change), but it's faster than changing only the bits
+	 * that have changed.
+	 *
+	 * We would like to do this atomicly, to avoid spinlocks in the
+	 * critical send path, but that's not really possible, given the
+	 * type of changes, and that this routine could be called on
+	 * multiple cpu's simultaneously, so we lock in this routine only,
+	 * to avoid conflicting updates; all we change is the shadow, and
+	 * it's a single 64 bit memory location, so by definition the update
+	 * is atomic in terms of what other cpu's can see in testing the
+	 * bits.  The spin_lock overhead isn't too bad, since it only
+	 * happens when all buffers are in use, so only cpu overhead, not
+	 * latency or bandwidth is affected.
+	 */
+	if (!dd->pioavailregs_dma)
+		return;
+	spin_lock_irqsave(&dd->pioavail_lock, flags);
+	for (i = 0; i < piobregs; i++) {
+		u64 pchbusy, pchg, piov, pnew;
+
+		piov = le64_to_cpu(dd->pioavailregs_dma[i]);
+		pchg = dd->pioavailkernel[i] &
+			~(dd->pioavailshadow[i] ^ piov);
+		pchbusy = pchg << QLOGIC_IB_SENDPIOAVAIL_BUSY_SHIFT;
+		if (pchg && (pchbusy & dd->pioavailshadow[i])) {
+			pnew = dd->pioavailshadow[i] & ~pchbusy;
+			pnew |= piov & pchbusy;
+			dd->pioavailshadow[i] = pnew;
+		}
+	}
+	spin_unlock_irqrestore(&dd->pioavail_lock, flags);
+}
+
+/*
+ * Debugging code and stats updates if no pio buffers available.
+ */
+static noinline void no_send_bufs(struct qib_devdata *dd)
+{
+	dd->upd_pio_shadow = 1;
+
+	/* not atomic, but if we lose a stat count in a while, that's OK */
+	qib_stats.sps_nopiobufs++;
+}
+
+/*
+ * Common code for normal driver send buffer allocation, and reserved
+ * allocation.
+ *
+ * Do appropriate marking as busy, etc.
+ * Returns buffer pointer if one is found, otherwise NULL.
+ */
+u32 __iomem *qib_getsendbuf_range(struct qib_devdata *dd, u32 *pbufnum,
+				  u32 first, u32 last)
+{
+	unsigned i, j, updated = 0;
+	unsigned nbufs;
+	unsigned long flags;
+	unsigned long *shadow = dd->pioavailshadow;
+	u32 __iomem *buf;
+
+	if (!(dd->flags & QIB_PRESENT))
+		return NULL;
+
+	nbufs = last - first + 1; /* number in range to check */
+	if (dd->upd_pio_shadow) {
+		/*
+		 * Minor optimization.  If we had no buffers on last call,
+		 * start out by doing the update; continue and do scan even
+		 * if no buffers were updated, to be paranoid.
+		 */
+		update_send_bufs(dd);
+		updated++;
+	}
+	i = first;
+rescan:
+	/*
+	 * While test_and_set_bit() is atomic, we do that and then the
+	 * change_bit(), and the pair is not.  See if this is the cause
+	 * of the remaining armlaunch errors.
+	 */
+	spin_lock_irqsave(&dd->pioavail_lock, flags);
+	for (j = 0; j < nbufs; j++, i++) {
+		if (i > last)
+			i = first;
+		if (__test_and_set_bit((2 * i) + 1, shadow))
+			continue;
+		/* flip generation bit */
+		__change_bit(2 * i, shadow);
+		/* remember that the buffer can be written to now */
+		__set_bit(i, dd->pio_writing);
+		break;
+	}
+	spin_unlock_irqrestore(&dd->pioavail_lock, flags);
+
+	if (j == nbufs) {
+		if (!updated) {
+			/*
+			 * First time through; shadow exhausted, but may be
+			 * buffers available, try an update and then rescan.
+			 */
+			update_send_bufs(dd);
+			updated++;
+			i = first;
+			goto rescan;
+		}
+		no_send_bufs(dd);
+		buf = NULL;
+	} else {
+		if (i < dd->piobcnt2k)
+			buf = (u32 __iomem *)(dd->pio2kbase +
+				i * dd->palign);
+		else
+			buf = (u32 __iomem *)(dd->pio4kbase +
+				(i - dd->piobcnt2k) * dd->align4k);
+		if (pbufnum)
+			*pbufnum = i;
+		dd->upd_pio_shadow = 0;
+	}
+
+	return buf;
+}
+
+/*
+ * Record that the caller is finished writing to the buffer so we don't
+ * disarm it while it is being written and disarm it now if needed.
+ */
+void qib_sendbuf_done(struct qib_devdata *dd, unsigned n)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&dd->pioavail_lock, flags);
+	__clear_bit(n, dd->pio_writing);
+	if (__test_and_clear_bit(n, dd->pio_need_disarm))
+		dd->f_sendctrl(dd->pport, QIB_SENDCTRL_DISARM_BUF(n));
+	spin_unlock_irqrestore(&dd->pioavail_lock, flags);
+}
+
+/**
+ * qib_chg_pioavailkernel - change which send buffers are available for kernel
+ * @dd: the qlogic_ib device
+ * @start: the starting send buffer number
+ * @len: the number of send buffers
+ * @avail: true if the buffers are available for kernel use, false otherwise
+ */
+void qib_chg_pioavailkernel(struct qib_devdata *dd, unsigned start,
+	unsigned len, u32 avail, struct qib_ctxtdata *rcd)
+{
+	unsigned long flags;
+	unsigned end;
+	unsigned ostart = start;
+
+	/* There are two bits per send buffer (busy and generation) */
+	start *= 2;
+	end = start + len * 2;
+
+	spin_lock_irqsave(&dd->pioavail_lock, flags);
+	/* Set or clear the busy bit in the shadow. */
+	while (start < end) {
+		if (avail) {
+			unsigned long dma;
+			int i;
+
+			/*
+			 * The BUSY bit will never be set, because we disarm
+			 * the user buffers before we hand them back to the
+			 * kernel.  We do have to make sure the generation
+			 * bit is set correctly in shadow, since it could
+			 * have changed many times while allocated to user.
+			 * We can't use the bitmap functions on the full
+			 * dma array because it is always little-endian, so
+			 * we have to flip to host-order first.
+			 * BITS_PER_LONG is slightly wrong, since it's
+			 * always 64 bits per register in chip...
+			 * We only work on 64 bit kernels, so that's OK.
+			 */
+			i = start / BITS_PER_LONG;
+			__clear_bit(QLOGIC_IB_SENDPIOAVAIL_BUSY_SHIFT + start,
+				    dd->pioavailshadow);
+			dma = (unsigned long)
+				le64_to_cpu(dd->pioavailregs_dma[i]);
+			if (test_bit((QLOGIC_IB_SENDPIOAVAIL_CHECK_SHIFT +
+				      start) % BITS_PER_LONG, &dma))
+				__set_bit(QLOGIC_IB_SENDPIOAVAIL_CHECK_SHIFT +
+					  start, dd->pioavailshadow);
+			else
+				__clear_bit(QLOGIC_IB_SENDPIOAVAIL_CHECK_SHIFT
+					    + start, dd->pioavailshadow);
+			__set_bit(start, dd->pioavailkernel);
+		} else {
+			__set_bit(start + QLOGIC_IB_SENDPIOAVAIL_BUSY_SHIFT,
+				  dd->pioavailshadow);
+			__clear_bit(start, dd->pioavailkernel);
+		}
+		start += 2;
+	}
+
+	spin_unlock_irqrestore(&dd->pioavail_lock, flags);
+
+	dd->f_txchk_change(dd, ostart, len, avail, rcd);
+}
+
+/*
+ * Flush all sends that might be in the ready to send state, as well as any
+ * that are in the process of being sent.  Used whenever we need to be
+ * sure the send side is idle.  Cleans up all buffer state by canceling
+ * all pio buffers, and issuing an abort, which cleans up anything in the
+ * launch fifo.  The cancel is superfluous on some chip versions, but
+ * it's safer to always do it.
+ * PIOAvail bits are updated by the chip as if a normal send had happened.
+ */
+void qib_cancel_sends(struct qib_pportdata *ppd)
+{
+	struct qib_devdata *dd = ppd->dd;
+	struct qib_ctxtdata *rcd;
+	unsigned long flags;
+	unsigned ctxt;
+	unsigned i;
+	unsigned last;
+
+	/*
+	 * Tell PSM to disarm buffers again before trying to reuse them.
+	 * We need to be sure the rcd doesn't change out from under us
+	 * while we do so.  We hold the two locks sequentially.  We might
+	 * needlessly set some need_disarm bits as a result, if the
+	 * context is closed after we release the uctxt_lock, but that's
+	 * fairly benign, and safer than nesting the locks.
+	 */
+	for (ctxt = dd->first_user_ctxt; ctxt < dd->cfgctxts; ctxt++) {
+		spin_lock_irqsave(&dd->uctxt_lock, flags);
+		rcd = dd->rcd[ctxt];
+		if (rcd && rcd->ppd == ppd) {
+			last = rcd->pio_base + rcd->piocnt;
+			if (rcd->user_event_mask) {
+				/*
+				 * subctxt_cnt is 0 if not shared, so do base
+				 * separately, first, then remaining subctxt,
+				 * if any
+				 */
+				set_bit(_QIB_EVENT_DISARM_BUFS_BIT,
+					&rcd->user_event_mask[0]);
+				for (i = 1; i < rcd->subctxt_cnt; i++)
+					set_bit(_QIB_EVENT_DISARM_BUFS_BIT,
+						&rcd->user_event_mask[i]);
+			}
+			i = rcd->pio_base;
+			spin_unlock_irqrestore(&dd->uctxt_lock, flags);
+			spin_lock_irqsave(&dd->pioavail_lock, flags);
+			for (; i < last; i++)
+				__set_bit(i, dd->pio_need_disarm);
+			spin_unlock_irqrestore(&dd->pioavail_lock, flags);
+		} else
+			spin_unlock_irqrestore(&dd->uctxt_lock, flags);
+	}
+
+	if (!(dd->flags & QIB_HAS_SEND_DMA))
+		dd->f_sendctrl(ppd, QIB_SENDCTRL_DISARM_ALL |
+				    QIB_SENDCTRL_FLUSH);
+}
+
+/*
+ * Force an update of in-memory copy of the pioavail registers, when
+ * needed for any of a variety of reasons.
+ * If already off, this routine is a nop, on the assumption that the
+ * caller (or set of callers) will "do the right thing".
+ * This is a per-device operation, so just the first port.
+ */
+void qib_force_pio_avail_update(struct qib_devdata *dd)
+{
+	dd->f_sendctrl(dd->pport, QIB_SENDCTRL_AVAIL_BLIP);
+}
+
+void qib_hol_down(struct qib_pportdata *ppd)
+{
+	/*
+	 * Cancel sends when the link goes DOWN so that we aren't doing it
+	 * at INIT when we might be trying to send SMI packets.
+	 */
+	if (!(ppd->lflags & QIBL_IB_AUTONEG_INPROG))
+		qib_cancel_sends(ppd);
+}
+
+/*
+ * Link is at INIT.
+ * We start the HoL timer so we can detect stuck packets blocking SMP replies.
+ * Timer may already be running, so use mod_timer, not add_timer.
+ */
+void qib_hol_init(struct qib_pportdata *ppd)
+{
+	if (ppd->hol_state != QIB_HOL_INIT) {
+		ppd->hol_state = QIB_HOL_INIT;
+		mod_timer(&ppd->hol_timer,
+			  jiffies + msecs_to_jiffies(qib_hol_timeout_ms));
+	}
+}
+
+/*
+ * Link is up, continue any user processes, and ensure timer
+ * is a nop, if running.  Let timer keep running, if set; it
+ * will nop when it sees the link is up.
+ */
+void qib_hol_up(struct qib_pportdata *ppd)
+{
+	ppd->hol_state = QIB_HOL_UP;
+}
+
+/*
+ * This is only called via the timer.
+ */
+void qib_hol_event(unsigned long opaque)
+{
+	struct qib_pportdata *ppd = (struct qib_pportdata *)opaque;
+
+	/* If hardware error, etc, skip. */
+	if (!(ppd->dd->flags & QIB_INITTED))
+		return;
+
+	if (ppd->hol_state != QIB_HOL_UP) {
+		/*
+		 * Try to flush sends in case a stuck packet is blocking
+		 * SMP replies.
+		 */
+		qib_hol_down(ppd);
+		mod_timer(&ppd->hol_timer,
+			  jiffies + msecs_to_jiffies(qib_hol_timeout_ms));
+	}
+}
diff --git a/drivers/infiniband/hw/qib/qib_uc.c b/drivers/infiniband/hw/qib/qib_uc.c
new file mode 100644
index 000000000000..6c7fe78cca64
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_uc.c
@@ -0,0 +1,555 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009, 2010 QLogic Corporation.
+ * All rights reserved.
+ * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 "qib.h"
+
+/* cut down ridiculously long IB macro names */
+#define OP(x) IB_OPCODE_UC_##x
+
+/**
+ * qib_make_uc_req - construct a request packet (SEND, RDMA write)
+ * @qp: a pointer to the QP
+ *
+ * Return 1 if constructed; otherwise, return 0.
+ */
+int qib_make_uc_req(struct qib_qp *qp)
+{
+	struct qib_other_headers *ohdr;
+	struct qib_swqe *wqe;
+	unsigned long flags;
+	u32 hwords;
+	u32 bth0;
+	u32 len;
+	u32 pmtu = ib_mtu_enum_to_int(qp->path_mtu);
+	int ret = 0;
+
+	spin_lock_irqsave(&qp->s_lock, flags);
+
+	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_SEND_OK)) {
+		if (!(ib_qib_state_ops[qp->state] & QIB_FLUSH_SEND))
+			goto bail;
+		/* We are in the error state, flush the work request. */
+		if (qp->s_last == qp->s_head)
+			goto bail;
+		/* If DMAs are in progress, we can't flush immediately. */
+		if (atomic_read(&qp->s_dma_busy)) {
+			qp->s_flags |= QIB_S_WAIT_DMA;
+			goto bail;
+		}
+		wqe = get_swqe_ptr(qp, qp->s_last);
+		qib_send_complete(qp, wqe, IB_WC_WR_FLUSH_ERR);
+		goto done;
+	}
+
+	ohdr = &qp->s_hdr.u.oth;
+	if (qp->remote_ah_attr.ah_flags & IB_AH_GRH)
+		ohdr = &qp->s_hdr.u.l.oth;
+
+	/* header size in 32-bit words LRH+BTH = (8+12)/4. */
+	hwords = 5;
+	bth0 = 0;
+
+	/* Get the next send request. */
+	wqe = get_swqe_ptr(qp, qp->s_cur);
+	qp->s_wqe = NULL;
+	switch (qp->s_state) {
+	default:
+		if (!(ib_qib_state_ops[qp->state] &
+		    QIB_PROCESS_NEXT_SEND_OK))
+			goto bail;
+		/* Check if send work queue is empty. */
+		if (qp->s_cur == qp->s_head)
+			goto bail;
+		/*
+		 * Start a new request.
+		 */
+		wqe->psn = qp->s_next_psn;
+		qp->s_psn = qp->s_next_psn;
+		qp->s_sge.sge = wqe->sg_list[0];
+		qp->s_sge.sg_list = wqe->sg_list + 1;
+		qp->s_sge.num_sge = wqe->wr.num_sge;
+		qp->s_sge.total_len = wqe->length;
+		len = wqe->length;
+		qp->s_len = len;
+		switch (wqe->wr.opcode) {
+		case IB_WR_SEND:
+		case IB_WR_SEND_WITH_IMM:
+			if (len > pmtu) {
+				qp->s_state = OP(SEND_FIRST);
+				len = pmtu;
+				break;
+			}
+			if (wqe->wr.opcode == IB_WR_SEND)
+				qp->s_state = OP(SEND_ONLY);
+			else {
+				qp->s_state =
+					OP(SEND_ONLY_WITH_IMMEDIATE);
+				/* Immediate data comes after the BTH */
+				ohdr->u.imm_data = wqe->wr.ex.imm_data;
+				hwords += 1;
+			}
+			if (wqe->wr.send_flags & IB_SEND_SOLICITED)
+				bth0 |= IB_BTH_SOLICITED;
+			qp->s_wqe = wqe;
+			if (++qp->s_cur >= qp->s_size)
+				qp->s_cur = 0;
+			break;
+
+		case IB_WR_RDMA_WRITE:
+		case IB_WR_RDMA_WRITE_WITH_IMM:
+			ohdr->u.rc.reth.vaddr =
+				cpu_to_be64(wqe->wr.wr.rdma.remote_addr);
+			ohdr->u.rc.reth.rkey =
+				cpu_to_be32(wqe->wr.wr.rdma.rkey);
+			ohdr->u.rc.reth.length = cpu_to_be32(len);
+			hwords += sizeof(struct ib_reth) / 4;
+			if (len > pmtu) {
+				qp->s_state = OP(RDMA_WRITE_FIRST);
+				len = pmtu;
+				break;
+			}
+			if (wqe->wr.opcode == IB_WR_RDMA_WRITE)
+				qp->s_state = OP(RDMA_WRITE_ONLY);
+			else {
+				qp->s_state =
+					OP(RDMA_WRITE_ONLY_WITH_IMMEDIATE);
+				/* Immediate data comes after the RETH */
+				ohdr->u.rc.imm_data = wqe->wr.ex.imm_data;
+				hwords += 1;
+				if (wqe->wr.send_flags & IB_SEND_SOLICITED)
+					bth0 |= IB_BTH_SOLICITED;
+			}
+			qp->s_wqe = wqe;
+			if (++qp->s_cur >= qp->s_size)
+				qp->s_cur = 0;
+			break;
+
+		default:
+			goto bail;
+		}
+		break;
+
+	case OP(SEND_FIRST):
+		qp->s_state = OP(SEND_MIDDLE);
+		/* FALLTHROUGH */
+	case OP(SEND_MIDDLE):
+		len = qp->s_len;
+		if (len > pmtu) {
+			len = pmtu;
+			break;
+		}
+		if (wqe->wr.opcode == IB_WR_SEND)
+			qp->s_state = OP(SEND_LAST);
+		else {
+			qp->s_state = OP(SEND_LAST_WITH_IMMEDIATE);
+			/* Immediate data comes after the BTH */
+			ohdr->u.imm_data = wqe->wr.ex.imm_data;
+			hwords += 1;
+		}
+		if (wqe->wr.send_flags & IB_SEND_SOLICITED)
+			bth0 |= IB_BTH_SOLICITED;
+		qp->s_wqe = wqe;
+		if (++qp->s_cur >= qp->s_size)
+			qp->s_cur = 0;
+		break;
+
+	case OP(RDMA_WRITE_FIRST):
+		qp->s_state = OP(RDMA_WRITE_MIDDLE);
+		/* FALLTHROUGH */
+	case OP(RDMA_WRITE_MIDDLE):
+		len = qp->s_len;
+		if (len > pmtu) {
+			len = pmtu;
+			break;
+		}
+		if (wqe->wr.opcode == IB_WR_RDMA_WRITE)
+			qp->s_state = OP(RDMA_WRITE_LAST);
+		else {
+			qp->s_state =
+				OP(RDMA_WRITE_LAST_WITH_IMMEDIATE);
+			/* Immediate data comes after the BTH */
+			ohdr->u.imm_data = wqe->wr.ex.imm_data;
+			hwords += 1;
+			if (wqe->wr.send_flags & IB_SEND_SOLICITED)
+				bth0 |= IB_BTH_SOLICITED;
+		}
+		qp->s_wqe = wqe;
+		if (++qp->s_cur >= qp->s_size)
+			qp->s_cur = 0;
+		break;
+	}
+	qp->s_len -= len;
+	qp->s_hdrwords = hwords;
+	qp->s_cur_sge = &qp->s_sge;
+	qp->s_cur_size = len;
+	qib_make_ruc_header(qp, ohdr, bth0 | (qp->s_state << 24),
+			    qp->s_next_psn++ & QIB_PSN_MASK);
+done:
+	ret = 1;
+	goto unlock;
+
+bail:
+	qp->s_flags &= ~QIB_S_BUSY;
+unlock:
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+	return ret;
+}
+
+/**
+ * qib_uc_rcv - handle an incoming UC packet
+ * @ibp: the port the packet came in on
+ * @hdr: the header of the packet
+ * @has_grh: true if the packet has a GRH
+ * @data: the packet data
+ * @tlen: the length of the packet
+ * @qp: the QP for this packet.
+ *
+ * This is called from qib_qp_rcv() to process an incoming UC packet
+ * for the given QP.
+ * Called at interrupt level.
+ */
+void qib_uc_rcv(struct qib_ibport *ibp, struct qib_ib_header *hdr,
+		int has_grh, void *data, u32 tlen, struct qib_qp *qp)
+{
+	struct qib_other_headers *ohdr;
+	unsigned long flags;
+	u32 opcode;
+	u32 hdrsize;
+	u32 psn;
+	u32 pad;
+	struct ib_wc wc;
+	u32 pmtu = ib_mtu_enum_to_int(qp->path_mtu);
+	struct ib_reth *reth;
+	int ret;
+
+	/* Check for GRH */
+	if (!has_grh) {
+		ohdr = &hdr->u.oth;
+		hdrsize = 8 + 12;       /* LRH + BTH */
+	} else {
+		ohdr = &hdr->u.l.oth;
+		hdrsize = 8 + 40 + 12;  /* LRH + GRH + BTH */
+	}
+
+	opcode = be32_to_cpu(ohdr->bth[0]);
+	spin_lock_irqsave(&qp->s_lock, flags);
+	if (qib_ruc_check_hdr(ibp, hdr, has_grh, qp, opcode))
+		goto sunlock;
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+
+	psn = be32_to_cpu(ohdr->bth[2]);
+	opcode >>= 24;
+	memset(&wc, 0, sizeof wc);
+
+	/* Prevent simultaneous processing after APM on different CPUs */
+	spin_lock(&qp->r_lock);
+
+	/* Compare the PSN verses the expected PSN. */
+	if (unlikely(qib_cmp24(psn, qp->r_psn) != 0)) {
+		/*
+		 * Handle a sequence error.
+		 * Silently drop any current message.
+		 */
+		qp->r_psn = psn;
+inv:
+		if (qp->r_state == OP(SEND_FIRST) ||
+		    qp->r_state == OP(SEND_MIDDLE)) {
+			set_bit(QIB_R_REWIND_SGE, &qp->r_aflags);
+			qp->r_sge.num_sge = 0;
+		} else
+			while (qp->r_sge.num_sge) {
+				atomic_dec(&qp->r_sge.sge.mr->refcount);
+				if (--qp->r_sge.num_sge)
+					qp->r_sge.sge = *qp->r_sge.sg_list++;
+			}
+		qp->r_state = OP(SEND_LAST);
+		switch (opcode) {
+		case OP(SEND_FIRST):
+		case OP(SEND_ONLY):
+		case OP(SEND_ONLY_WITH_IMMEDIATE):
+			goto send_first;
+
+		case OP(RDMA_WRITE_FIRST):
+		case OP(RDMA_WRITE_ONLY):
+		case OP(RDMA_WRITE_ONLY_WITH_IMMEDIATE):
+			goto rdma_first;
+
+		default:
+			goto drop;
+		}
+	}
+
+	/* Check for opcode sequence errors. */
+	switch (qp->r_state) {
+	case OP(SEND_FIRST):
+	case OP(SEND_MIDDLE):
+		if (opcode == OP(SEND_MIDDLE) ||
+		    opcode == OP(SEND_LAST) ||
+		    opcode == OP(SEND_LAST_WITH_IMMEDIATE))
+			break;
+		goto inv;
+
+	case OP(RDMA_WRITE_FIRST):
+	case OP(RDMA_WRITE_MIDDLE):
+		if (opcode == OP(RDMA_WRITE_MIDDLE) ||
+		    opcode == OP(RDMA_WRITE_LAST) ||
+		    opcode == OP(RDMA_WRITE_LAST_WITH_IMMEDIATE))
+			break;
+		goto inv;
+
+	default:
+		if (opcode == OP(SEND_FIRST) ||
+		    opcode == OP(SEND_ONLY) ||
+		    opcode == OP(SEND_ONLY_WITH_IMMEDIATE) ||
+		    opcode == OP(RDMA_WRITE_FIRST) ||
+		    opcode == OP(RDMA_WRITE_ONLY) ||
+		    opcode == OP(RDMA_WRITE_ONLY_WITH_IMMEDIATE))
+			break;
+		goto inv;
+	}
+
+	if (qp->state == IB_QPS_RTR && !(qp->r_flags & QIB_R_COMM_EST)) {
+		qp->r_flags |= QIB_R_COMM_EST;
+		if (qp->ibqp.event_handler) {
+			struct ib_event ev;
+
+			ev.device = qp->ibqp.device;
+			ev.element.qp = &qp->ibqp;
+			ev.event = IB_EVENT_COMM_EST;
+			qp->ibqp.event_handler(&ev, qp->ibqp.qp_context);
+		}
+	}
+
+	/* OK, process the packet. */
+	switch (opcode) {
+	case OP(SEND_FIRST):
+	case OP(SEND_ONLY):
+	case OP(SEND_ONLY_WITH_IMMEDIATE):
+send_first:
+		if (test_and_clear_bit(QIB_R_REWIND_SGE, &qp->r_aflags))
+			qp->r_sge = qp->s_rdma_read_sge;
+		else {
+			ret = qib_get_rwqe(qp, 0);
+			if (ret < 0)
+				goto op_err;
+			if (!ret)
+				goto drop;
+			/*
+			 * qp->s_rdma_read_sge will be the owner
+			 * of the mr references.
+			 */
+			qp->s_rdma_read_sge = qp->r_sge;
+		}
+		qp->r_rcv_len = 0;
+		if (opcode == OP(SEND_ONLY))
+			goto send_last;
+		else if (opcode == OP(SEND_ONLY_WITH_IMMEDIATE))
+			goto send_last_imm;
+		/* FALLTHROUGH */
+	case OP(SEND_MIDDLE):
+		/* Check for invalid length PMTU or posted rwqe len. */
+		if (unlikely(tlen != (hdrsize + pmtu + 4)))
+			goto rewind;
+		qp->r_rcv_len += pmtu;
+		if (unlikely(qp->r_rcv_len > qp->r_len))
+			goto rewind;
+		qib_copy_sge(&qp->r_sge, data, pmtu, 0);
+		break;
+
+	case OP(SEND_LAST_WITH_IMMEDIATE):
+send_last_imm:
+		wc.ex.imm_data = ohdr->u.imm_data;
+		hdrsize += 4;
+		wc.wc_flags = IB_WC_WITH_IMM;
+		/* FALLTHROUGH */
+	case OP(SEND_LAST):
+send_last:
+		/* Get the number of bytes the message was padded by. */
+		pad = (be32_to_cpu(ohdr->bth[0]) >> 20) & 3;
+		/* Check for invalid length. */
+		/* XXX LAST len should be >= 1 */
+		if (unlikely(tlen < (hdrsize + pad + 4)))
+			goto rewind;
+		/* Don't count the CRC. */
+		tlen -= (hdrsize + pad + 4);
+		wc.byte_len = tlen + qp->r_rcv_len;
+		if (unlikely(wc.byte_len > qp->r_len))
+			goto rewind;
+		wc.opcode = IB_WC_RECV;
+last_imm:
+		qib_copy_sge(&qp->r_sge, data, tlen, 0);
+		while (qp->s_rdma_read_sge.num_sge) {
+			atomic_dec(&qp->s_rdma_read_sge.sge.mr->refcount);
+			if (--qp->s_rdma_read_sge.num_sge)
+				qp->s_rdma_read_sge.sge =
+					*qp->s_rdma_read_sge.sg_list++;
+		}
+		wc.wr_id = qp->r_wr_id;
+		wc.status = IB_WC_SUCCESS;
+		wc.qp = &qp->ibqp;
+		wc.src_qp = qp->remote_qpn;
+		wc.slid = qp->remote_ah_attr.dlid;
+		wc.sl = qp->remote_ah_attr.sl;
+		/* Signal completion event if the solicited bit is set. */
+		qib_cq_enter(to_icq(qp->ibqp.recv_cq), &wc,
+			     (ohdr->bth[0] &
+				cpu_to_be32(IB_BTH_SOLICITED)) != 0);
+		break;
+
+	case OP(RDMA_WRITE_FIRST):
+	case OP(RDMA_WRITE_ONLY):
+	case OP(RDMA_WRITE_ONLY_WITH_IMMEDIATE): /* consume RWQE */
+rdma_first:
+		if (unlikely(!(qp->qp_access_flags &
+			       IB_ACCESS_REMOTE_WRITE))) {
+			goto drop;
+		}
+		reth = &ohdr->u.rc.reth;
+		hdrsize += sizeof(*reth);
+		qp->r_len = be32_to_cpu(reth->length);
+		qp->r_rcv_len = 0;
+		qp->r_sge.sg_list = NULL;
+		if (qp->r_len != 0) {
+			u32 rkey = be32_to_cpu(reth->rkey);
+			u64 vaddr = be64_to_cpu(reth->vaddr);
+			int ok;
+
+			/* Check rkey */
+			ok = qib_rkey_ok(qp, &qp->r_sge.sge, qp->r_len,
+					 vaddr, rkey, IB_ACCESS_REMOTE_WRITE);
+			if (unlikely(!ok))
+				goto drop;
+			qp->r_sge.num_sge = 1;
+		} else {
+			qp->r_sge.num_sge = 0;
+			qp->r_sge.sge.mr = NULL;
+			qp->r_sge.sge.vaddr = NULL;
+			qp->r_sge.sge.length = 0;
+			qp->r_sge.sge.sge_length = 0;
+		}
+		if (opcode == OP(RDMA_WRITE_ONLY))
+			goto rdma_last;
+		else if (opcode == OP(RDMA_WRITE_ONLY_WITH_IMMEDIATE))
+			goto rdma_last_imm;
+		/* FALLTHROUGH */
+	case OP(RDMA_WRITE_MIDDLE):
+		/* Check for invalid length PMTU or posted rwqe len. */
+		if (unlikely(tlen != (hdrsize + pmtu + 4)))
+			goto drop;
+		qp->r_rcv_len += pmtu;
+		if (unlikely(qp->r_rcv_len > qp->r_len))
+			goto drop;
+		qib_copy_sge(&qp->r_sge, data, pmtu, 1);
+		break;
+
+	case OP(RDMA_WRITE_LAST_WITH_IMMEDIATE):
+rdma_last_imm:
+		wc.ex.imm_data = ohdr->u.imm_data;
+		hdrsize += 4;
+		wc.wc_flags = IB_WC_WITH_IMM;
+
+		/* Get the number of bytes the message was padded by. */
+		pad = (be32_to_cpu(ohdr->bth[0]) >> 20) & 3;
+		/* Check for invalid length. */
+		/* XXX LAST len should be >= 1 */
+		if (unlikely(tlen < (hdrsize + pad + 4)))
+			goto drop;
+		/* Don't count the CRC. */
+		tlen -= (hdrsize + pad + 4);
+		if (unlikely(tlen + qp->r_rcv_len != qp->r_len))
+			goto drop;
+		if (test_and_clear_bit(QIB_R_REWIND_SGE, &qp->r_aflags))
+			while (qp->s_rdma_read_sge.num_sge) {
+				atomic_dec(&qp->s_rdma_read_sge.sge.mr->
+					   refcount);
+				if (--qp->s_rdma_read_sge.num_sge)
+					qp->s_rdma_read_sge.sge =
+						*qp->s_rdma_read_sge.sg_list++;
+			}
+		else {
+			ret = qib_get_rwqe(qp, 1);
+			if (ret < 0)
+				goto op_err;
+			if (!ret)
+				goto drop;
+		}
+		wc.byte_len = qp->r_len;
+		wc.opcode = IB_WC_RECV_RDMA_WITH_IMM;
+		goto last_imm;
+
+	case OP(RDMA_WRITE_LAST):
+rdma_last:
+		/* Get the number of bytes the message was padded by. */
+		pad = (be32_to_cpu(ohdr->bth[0]) >> 20) & 3;
+		/* Check for invalid length. */
+		/* XXX LAST len should be >= 1 */
+		if (unlikely(tlen < (hdrsize + pad + 4)))
+			goto drop;
+		/* Don't count the CRC. */
+		tlen -= (hdrsize + pad + 4);
+		if (unlikely(tlen + qp->r_rcv_len != qp->r_len))
+			goto drop;
+		qib_copy_sge(&qp->r_sge, data, tlen, 1);
+		while (qp->r_sge.num_sge) {
+			atomic_dec(&qp->r_sge.sge.mr->refcount);
+			if (--qp->r_sge.num_sge)
+				qp->r_sge.sge = *qp->r_sge.sg_list++;
+		}
+		break;
+
+	default:
+		/* Drop packet for unknown opcodes. */
+		goto drop;
+	}
+	qp->r_psn++;
+	qp->r_state = opcode;
+	spin_unlock(&qp->r_lock);
+	return;
+
+rewind:
+	set_bit(QIB_R_REWIND_SGE, &qp->r_aflags);
+	qp->r_sge.num_sge = 0;
+drop:
+	ibp->n_pkt_drops++;
+	spin_unlock(&qp->r_lock);
+	return;
+
+op_err:
+	qib_rc_error(qp, IB_WC_LOC_QP_OP_ERR);
+	spin_unlock(&qp->r_lock);
+	return;
+
+sunlock:
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+}
diff --git a/drivers/infiniband/hw/qib/qib_ud.c b/drivers/infiniband/hw/qib/qib_ud.c
new file mode 100644
index 000000000000..c838cda73347
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_ud.c
@@ -0,0 +1,607 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <rdma/ib_smi.h>
+
+#include "qib.h"
+#include "qib_mad.h"
+
+/**
+ * qib_ud_loopback - handle send on loopback QPs
+ * @sqp: the sending QP
+ * @swqe: the send work request
+ *
+ * This is called from qib_make_ud_req() to forward a WQE addressed
+ * to the same HCA.
+ * Note that the receive interrupt handler may be calling qib_ud_rcv()
+ * while this is being called.
+ */
+static void qib_ud_loopback(struct qib_qp *sqp, struct qib_swqe *swqe)
+{
+	struct qib_ibport *ibp = to_iport(sqp->ibqp.device, sqp->port_num);
+	struct qib_pportdata *ppd;
+	struct qib_qp *qp;
+	struct ib_ah_attr *ah_attr;
+	unsigned long flags;
+	struct qib_sge_state ssge;
+	struct qib_sge *sge;
+	struct ib_wc wc;
+	u32 length;
+
+	qp = qib_lookup_qpn(ibp, swqe->wr.wr.ud.remote_qpn);
+	if (!qp) {
+		ibp->n_pkt_drops++;
+		return;
+	}
+	if (qp->ibqp.qp_type != sqp->ibqp.qp_type ||
+	    !(ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK)) {
+		ibp->n_pkt_drops++;
+		goto drop;
+	}
+
+	ah_attr = &to_iah(swqe->wr.wr.ud.ah)->attr;
+	ppd = ppd_from_ibp(ibp);
+
+	if (qp->ibqp.qp_num > 1) {
+		u16 pkey1;
+		u16 pkey2;
+		u16 lid;
+
+		pkey1 = qib_get_pkey(ibp, sqp->s_pkey_index);
+		pkey2 = qib_get_pkey(ibp, qp->s_pkey_index);
+		if (unlikely(!qib_pkey_ok(pkey1, pkey2))) {
+			lid = ppd->lid | (ah_attr->src_path_bits &
+					  ((1 << ppd->lmc) - 1));
+			qib_bad_pqkey(ibp, IB_NOTICE_TRAP_BAD_PKEY, pkey1,
+				      ah_attr->sl,
+				      sqp->ibqp.qp_num, qp->ibqp.qp_num,
+				      cpu_to_be16(lid),
+				      cpu_to_be16(ah_attr->dlid));
+			goto drop;
+		}
+	}
+
+	/*
+	 * Check that the qkey matches (except for QP0, see 9.6.1.4.1).
+	 * Qkeys with the high order bit set mean use the
+	 * qkey from the QP context instead of the WR (see 10.2.5).
+	 */
+	if (qp->ibqp.qp_num) {
+		u32 qkey;
+
+		qkey = (int)swqe->wr.wr.ud.remote_qkey < 0 ?
+			sqp->qkey : swqe->wr.wr.ud.remote_qkey;
+		if (unlikely(qkey != qp->qkey)) {
+			u16 lid;
+
+			lid = ppd->lid | (ah_attr->src_path_bits &
+					  ((1 << ppd->lmc) - 1));
+			qib_bad_pqkey(ibp, IB_NOTICE_TRAP_BAD_QKEY, qkey,
+				      ah_attr->sl,
+				      sqp->ibqp.qp_num, qp->ibqp.qp_num,
+				      cpu_to_be16(lid),
+				      cpu_to_be16(ah_attr->dlid));
+			goto drop;
+		}
+	}
+
+	/*
+	 * A GRH is expected to preceed the data even if not
+	 * present on the wire.
+	 */
+	length = swqe->length;
+	memset(&wc, 0, sizeof wc);
+	wc.byte_len = length + sizeof(struct ib_grh);
+
+	if (swqe->wr.opcode == IB_WR_SEND_WITH_IMM) {
+		wc.wc_flags = IB_WC_WITH_IMM;
+		wc.ex.imm_data = swqe->wr.ex.imm_data;
+	}
+
+	spin_lock_irqsave(&qp->r_lock, flags);
+
+	/*
+	 * Get the next work request entry to find where to put the data.
+	 */
+	if (qp->r_flags & QIB_R_REUSE_SGE)
+		qp->r_flags &= ~QIB_R_REUSE_SGE;
+	else {
+		int ret;
+
+		ret = qib_get_rwqe(qp, 0);
+		if (ret < 0) {
+			qib_rc_error(qp, IB_WC_LOC_QP_OP_ERR);
+			goto bail_unlock;
+		}
+		if (!ret) {
+			if (qp->ibqp.qp_num == 0)
+				ibp->n_vl15_dropped++;
+			goto bail_unlock;
+		}
+	}
+	/* Silently drop packets which are too big. */
+	if (unlikely(wc.byte_len > qp->r_len)) {
+		qp->r_flags |= QIB_R_REUSE_SGE;
+		ibp->n_pkt_drops++;
+		goto bail_unlock;
+	}
+
+	if (ah_attr->ah_flags & IB_AH_GRH) {
+		qib_copy_sge(&qp->r_sge, &ah_attr->grh,
+			     sizeof(struct ib_grh), 1);
+		wc.wc_flags |= IB_WC_GRH;
+	} else
+		qib_skip_sge(&qp->r_sge, sizeof(struct ib_grh), 1);
+	ssge.sg_list = swqe->sg_list + 1;
+	ssge.sge = *swqe->sg_list;
+	ssge.num_sge = swqe->wr.num_sge;
+	sge = &ssge.sge;
+	while (length) {
+		u32 len = sge->length;
+
+		if (len > length)
+			len = length;
+		if (len > sge->sge_length)
+			len = sge->sge_length;
+		BUG_ON(len == 0);
+		qib_copy_sge(&qp->r_sge, sge->vaddr, len, 1);
+		sge->vaddr += len;
+		sge->length -= len;
+		sge->sge_length -= len;
+		if (sge->sge_length == 0) {
+			if (--ssge.num_sge)
+				*sge = *ssge.sg_list++;
+		} else if (sge->length == 0 && sge->mr->lkey) {
+			if (++sge->n >= QIB_SEGSZ) {
+				if (++sge->m >= sge->mr->mapsz)
+					break;
+				sge->n = 0;
+			}
+			sge->vaddr =
+				sge->mr->map[sge->m]->segs[sge->n].vaddr;
+			sge->length =
+				sge->mr->map[sge->m]->segs[sge->n].length;
+		}
+		length -= len;
+	}
+	while (qp->r_sge.num_sge) {
+		atomic_dec(&qp->r_sge.sge.mr->refcount);
+		if (--qp->r_sge.num_sge)
+			qp->r_sge.sge = *qp->r_sge.sg_list++;
+	}
+	if (!test_and_clear_bit(QIB_R_WRID_VALID, &qp->r_aflags))
+		goto bail_unlock;
+	wc.wr_id = qp->r_wr_id;
+	wc.status = IB_WC_SUCCESS;
+	wc.opcode = IB_WC_RECV;
+	wc.qp = &qp->ibqp;
+	wc.src_qp = sqp->ibqp.qp_num;
+	wc.pkey_index = qp->ibqp.qp_type == IB_QPT_GSI ?
+		swqe->wr.wr.ud.pkey_index : 0;
+	wc.slid = ppd->lid | (ah_attr->src_path_bits & ((1 << ppd->lmc) - 1));
+	wc.sl = ah_attr->sl;
+	wc.dlid_path_bits = ah_attr->dlid & ((1 << ppd->lmc) - 1);
+	wc.port_num = qp->port_num;
+	/* Signal completion event if the solicited bit is set. */
+	qib_cq_enter(to_icq(qp->ibqp.recv_cq), &wc,
+		     swqe->wr.send_flags & IB_SEND_SOLICITED);
+	ibp->n_loop_pkts++;
+bail_unlock:
+	spin_unlock_irqrestore(&qp->r_lock, flags);
+drop:
+	if (atomic_dec_and_test(&qp->refcount))
+		wake_up(&qp->wait);
+}
+
+/**
+ * qib_make_ud_req - construct a UD request packet
+ * @qp: the QP
+ *
+ * Return 1 if constructed; otherwise, return 0.
+ */
+int qib_make_ud_req(struct qib_qp *qp)
+{
+	struct qib_other_headers *ohdr;
+	struct ib_ah_attr *ah_attr;
+	struct qib_pportdata *ppd;
+	struct qib_ibport *ibp;
+	struct qib_swqe *wqe;
+	unsigned long flags;
+	u32 nwords;
+	u32 extra_bytes;
+	u32 bth0;
+	u16 lrh0;
+	u16 lid;
+	int ret = 0;
+	int next_cur;
+
+	spin_lock_irqsave(&qp->s_lock, flags);
+
+	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_NEXT_SEND_OK)) {
+		if (!(ib_qib_state_ops[qp->state] & QIB_FLUSH_SEND))
+			goto bail;
+		/* We are in the error state, flush the work request. */
+		if (qp->s_last == qp->s_head)
+			goto bail;
+		/* If DMAs are in progress, we can't flush immediately. */
+		if (atomic_read(&qp->s_dma_busy)) {
+			qp->s_flags |= QIB_S_WAIT_DMA;
+			goto bail;
+		}
+		wqe = get_swqe_ptr(qp, qp->s_last);
+		qib_send_complete(qp, wqe, IB_WC_WR_FLUSH_ERR);
+		goto done;
+	}
+
+	if (qp->s_cur == qp->s_head)
+		goto bail;
+
+	wqe = get_swqe_ptr(qp, qp->s_cur);
+	next_cur = qp->s_cur + 1;
+	if (next_cur >= qp->s_size)
+		next_cur = 0;
+
+	/* Construct the header. */
+	ibp = to_iport(qp->ibqp.device, qp->port_num);
+	ppd = ppd_from_ibp(ibp);
+	ah_attr = &to_iah(wqe->wr.wr.ud.ah)->attr;
+	if (ah_attr->dlid >= QIB_MULTICAST_LID_BASE) {
+		if (ah_attr->dlid != QIB_PERMISSIVE_LID)
+			ibp->n_multicast_xmit++;
+		else
+			ibp->n_unicast_xmit++;
+	} else {
+		ibp->n_unicast_xmit++;
+		lid = ah_attr->dlid & ~((1 << ppd->lmc) - 1);
+		if (unlikely(lid == ppd->lid)) {
+			/*
+			 * If DMAs are in progress, we can't generate
+			 * a completion for the loopback packet since
+			 * it would be out of order.
+			 * XXX Instead of waiting, we could queue a
+			 * zero length descriptor so we get a callback.
+			 */
+			if (atomic_read(&qp->s_dma_busy)) {
+				qp->s_flags |= QIB_S_WAIT_DMA;
+				goto bail;
+			}
+			qp->s_cur = next_cur;
+			spin_unlock_irqrestore(&qp->s_lock, flags);
+			qib_ud_loopback(qp, wqe);
+			spin_lock_irqsave(&qp->s_lock, flags);
+			qib_send_complete(qp, wqe, IB_WC_SUCCESS);
+			goto done;
+		}
+	}
+
+	qp->s_cur = next_cur;
+	extra_bytes = -wqe->length & 3;
+	nwords = (wqe->length + extra_bytes) >> 2;
+
+	/* header size in 32-bit words LRH+BTH+DETH = (8+12+8)/4. */
+	qp->s_hdrwords = 7;
+	qp->s_cur_size = wqe->length;
+	qp->s_cur_sge = &qp->s_sge;
+	qp->s_srate = ah_attr->static_rate;
+	qp->s_wqe = wqe;
+	qp->s_sge.sge = wqe->sg_list[0];
+	qp->s_sge.sg_list = wqe->sg_list + 1;
+	qp->s_sge.num_sge = wqe->wr.num_sge;
+	qp->s_sge.total_len = wqe->length;
+
+	if (ah_attr->ah_flags & IB_AH_GRH) {
+		/* Header size in 32-bit words. */
+		qp->s_hdrwords += qib_make_grh(ibp, &qp->s_hdr.u.l.grh,
+					       &ah_attr->grh,
+					       qp->s_hdrwords, nwords);
+		lrh0 = QIB_LRH_GRH;
+		ohdr = &qp->s_hdr.u.l.oth;
+		/*
+		 * Don't worry about sending to locally attached multicast
+		 * QPs.  It is unspecified by the spec. what happens.
+		 */
+	} else {
+		/* Header size in 32-bit words. */
+		lrh0 = QIB_LRH_BTH;
+		ohdr = &qp->s_hdr.u.oth;
+	}
+	if (wqe->wr.opcode == IB_WR_SEND_WITH_IMM) {
+		qp->s_hdrwords++;
+		ohdr->u.ud.imm_data = wqe->wr.ex.imm_data;
+		bth0 = IB_OPCODE_UD_SEND_ONLY_WITH_IMMEDIATE << 24;
+	} else
+		bth0 = IB_OPCODE_UD_SEND_ONLY << 24;
+	lrh0 |= ah_attr->sl << 4;
+	if (qp->ibqp.qp_type == IB_QPT_SMI)
+		lrh0 |= 0xF000; /* Set VL (see ch. 13.5.3.1) */
+	else
+		lrh0 |= ibp->sl_to_vl[ah_attr->sl] << 12;
+	qp->s_hdr.lrh[0] = cpu_to_be16(lrh0);
+	qp->s_hdr.lrh[1] = cpu_to_be16(ah_attr->dlid);  /* DEST LID */
+	qp->s_hdr.lrh[2] = cpu_to_be16(qp->s_hdrwords + nwords + SIZE_OF_CRC);
+	lid = ppd->lid;
+	if (lid) {
+		lid |= ah_attr->src_path_bits & ((1 << ppd->lmc) - 1);
+		qp->s_hdr.lrh[3] = cpu_to_be16(lid);
+	} else
+		qp->s_hdr.lrh[3] = IB_LID_PERMISSIVE;
+	if (wqe->wr.send_flags & IB_SEND_SOLICITED)
+		bth0 |= IB_BTH_SOLICITED;
+	bth0 |= extra_bytes << 20;
+	bth0 |= qp->ibqp.qp_type == IB_QPT_SMI ? QIB_DEFAULT_P_KEY :
+		qib_get_pkey(ibp, qp->ibqp.qp_type == IB_QPT_GSI ?
+			     wqe->wr.wr.ud.pkey_index : qp->s_pkey_index);
+	ohdr->bth[0] = cpu_to_be32(bth0);
+	/*
+	 * Use the multicast QP if the destination LID is a multicast LID.
+	 */
+	ohdr->bth[1] = ah_attr->dlid >= QIB_MULTICAST_LID_BASE &&
+		ah_attr->dlid != QIB_PERMISSIVE_LID ?
+		cpu_to_be32(QIB_MULTICAST_QPN) :
+		cpu_to_be32(wqe->wr.wr.ud.remote_qpn);
+	ohdr->bth[2] = cpu_to_be32(qp->s_next_psn++ & QIB_PSN_MASK);
+	/*
+	 * Qkeys with the high order bit set mean use the
+	 * qkey from the QP context instead of the WR (see 10.2.5).
+	 */
+	ohdr->u.ud.deth[0] = cpu_to_be32((int)wqe->wr.wr.ud.remote_qkey < 0 ?
+					 qp->qkey : wqe->wr.wr.ud.remote_qkey);
+	ohdr->u.ud.deth[1] = cpu_to_be32(qp->ibqp.qp_num);
+
+done:
+	ret = 1;
+	goto unlock;
+
+bail:
+	qp->s_flags &= ~QIB_S_BUSY;
+unlock:
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+	return ret;
+}
+
+static unsigned qib_lookup_pkey(struct qib_ibport *ibp, u16 pkey)
+{
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	struct qib_devdata *dd = ppd->dd;
+	unsigned ctxt = ppd->hw_pidx;
+	unsigned i;
+
+	pkey &= 0x7fff;	/* remove limited/full membership bit */
+
+	for (i = 0; i < ARRAY_SIZE(dd->rcd[ctxt]->pkeys); ++i)
+		if ((dd->rcd[ctxt]->pkeys[i] & 0x7fff) == pkey)
+			return i;
+
+	/*
+	 * Should not get here, this means hardware failed to validate pkeys.
+	 * Punt and return index 0.
+	 */
+	return 0;
+}
+
+/**
+ * qib_ud_rcv - receive an incoming UD packet
+ * @ibp: the port the packet came in on
+ * @hdr: the packet header
+ * @has_grh: true if the packet has a GRH
+ * @data: the packet data
+ * @tlen: the packet length
+ * @qp: the QP the packet came on
+ *
+ * This is called from qib_qp_rcv() to process an incoming UD packet
+ * for the given QP.
+ * Called at interrupt level.
+ */
+void qib_ud_rcv(struct qib_ibport *ibp, struct qib_ib_header *hdr,
+		int has_grh, void *data, u32 tlen, struct qib_qp *qp)
+{
+	struct qib_other_headers *ohdr;
+	int opcode;
+	u32 hdrsize;
+	u32 pad;
+	struct ib_wc wc;
+	u32 qkey;
+	u32 src_qp;
+	u16 dlid;
+
+	/* Check for GRH */
+	if (!has_grh) {
+		ohdr = &hdr->u.oth;
+		hdrsize = 8 + 12 + 8;   /* LRH + BTH + DETH */
+	} else {
+		ohdr = &hdr->u.l.oth;
+		hdrsize = 8 + 40 + 12 + 8; /* LRH + GRH + BTH + DETH */
+	}
+	qkey = be32_to_cpu(ohdr->u.ud.deth[0]);
+	src_qp = be32_to_cpu(ohdr->u.ud.deth[1]) & QIB_QPN_MASK;
+
+	/* Get the number of bytes the message was padded by. */
+	pad = (be32_to_cpu(ohdr->bth[0]) >> 20) & 3;
+	if (unlikely(tlen < (hdrsize + pad + 4))) {
+		/* Drop incomplete packets. */
+		ibp->n_pkt_drops++;
+		goto bail;
+	}
+	tlen -= hdrsize + pad + 4;
+
+	/*
+	 * Check that the permissive LID is only used on QP0
+	 * and the QKEY matches (see 9.6.1.4.1 and 9.6.1.5.1).
+	 */
+	if (qp->ibqp.qp_num) {
+		if (unlikely(hdr->lrh[1] == IB_LID_PERMISSIVE ||
+			     hdr->lrh[3] == IB_LID_PERMISSIVE)) {
+			ibp->n_pkt_drops++;
+			goto bail;
+		}
+		if (qp->ibqp.qp_num > 1) {
+			u16 pkey1, pkey2;
+
+			pkey1 = be32_to_cpu(ohdr->bth[0]);
+			pkey2 = qib_get_pkey(ibp, qp->s_pkey_index);
+			if (unlikely(!qib_pkey_ok(pkey1, pkey2))) {
+				qib_bad_pqkey(ibp, IB_NOTICE_TRAP_BAD_PKEY,
+					      pkey1,
+					      (be16_to_cpu(hdr->lrh[0]) >> 4) &
+						0xF,
+					      src_qp, qp->ibqp.qp_num,
+					      hdr->lrh[3], hdr->lrh[1]);
+				goto bail;
+			}
+		}
+		if (unlikely(qkey != qp->qkey)) {
+			qib_bad_pqkey(ibp, IB_NOTICE_TRAP_BAD_QKEY, qkey,
+				      (be16_to_cpu(hdr->lrh[0]) >> 4) & 0xF,
+				      src_qp, qp->ibqp.qp_num,
+				      hdr->lrh[3], hdr->lrh[1]);
+			goto bail;
+		}
+		/* Drop invalid MAD packets (see 13.5.3.1). */
+		if (unlikely(qp->ibqp.qp_num == 1 &&
+			     (tlen != 256 ||
+			      (be16_to_cpu(hdr->lrh[0]) >> 12) == 15))) {
+			ibp->n_pkt_drops++;
+			goto bail;
+		}
+	} else {
+		struct ib_smp *smp;
+
+		/* Drop invalid MAD packets (see 13.5.3.1). */
+		if (tlen != 256 || (be16_to_cpu(hdr->lrh[0]) >> 12) != 15) {
+			ibp->n_pkt_drops++;
+			goto bail;
+		}
+		smp = (struct ib_smp *) data;
+		if ((hdr->lrh[1] == IB_LID_PERMISSIVE ||
+		     hdr->lrh[3] == IB_LID_PERMISSIVE) &&
+		    smp->mgmt_class != IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) {
+			ibp->n_pkt_drops++;
+			goto bail;
+		}
+	}
+
+	/*
+	 * The opcode is in the low byte when its in network order
+	 * (top byte when in host order).
+	 */
+	opcode = be32_to_cpu(ohdr->bth[0]) >> 24;
+	if (qp->ibqp.qp_num > 1 &&
+	    opcode == IB_OPCODE_UD_SEND_ONLY_WITH_IMMEDIATE) {
+		wc.ex.imm_data = ohdr->u.ud.imm_data;
+		wc.wc_flags = IB_WC_WITH_IMM;
+		hdrsize += sizeof(u32);
+	} else if (opcode == IB_OPCODE_UD_SEND_ONLY) {
+		wc.ex.imm_data = 0;
+		wc.wc_flags = 0;
+	} else {
+		ibp->n_pkt_drops++;
+		goto bail;
+	}
+
+	/*
+	 * A GRH is expected to preceed the data even if not
+	 * present on the wire.
+	 */
+	wc.byte_len = tlen + sizeof(struct ib_grh);
+
+	/*
+	 * We need to serialize getting a receive work queue entry and
+	 * generating a completion for it against QPs sending to this QP
+	 * locally.
+	 */
+	spin_lock(&qp->r_lock);
+
+	/*
+	 * Get the next work request entry to find where to put the data.
+	 */
+	if (qp->r_flags & QIB_R_REUSE_SGE)
+		qp->r_flags &= ~QIB_R_REUSE_SGE;
+	else {
+		int ret;
+
+		ret = qib_get_rwqe(qp, 0);
+		if (ret < 0) {
+			qib_rc_error(qp, IB_WC_LOC_QP_OP_ERR);
+			goto bail_unlock;
+		}
+		if (!ret) {
+			if (qp->ibqp.qp_num == 0)
+				ibp->n_vl15_dropped++;
+			goto bail_unlock;
+		}
+	}
+	/* Silently drop packets which are too big. */
+	if (unlikely(wc.byte_len > qp->r_len)) {
+		qp->r_flags |= QIB_R_REUSE_SGE;
+		ibp->n_pkt_drops++;
+		goto bail_unlock;
+	}
+	if (has_grh) {
+		qib_copy_sge(&qp->r_sge, &hdr->u.l.grh,
+			     sizeof(struct ib_grh), 1);
+		wc.wc_flags |= IB_WC_GRH;
+	} else
+		qib_skip_sge(&qp->r_sge, sizeof(struct ib_grh), 1);
+	qib_copy_sge(&qp->r_sge, data, wc.byte_len - sizeof(struct ib_grh), 1);
+	while (qp->r_sge.num_sge) {
+		atomic_dec(&qp->r_sge.sge.mr->refcount);
+		if (--qp->r_sge.num_sge)
+			qp->r_sge.sge = *qp->r_sge.sg_list++;
+	}
+	if (!test_and_clear_bit(QIB_R_WRID_VALID, &qp->r_aflags))
+		goto bail_unlock;
+	wc.wr_id = qp->r_wr_id;
+	wc.status = IB_WC_SUCCESS;
+	wc.opcode = IB_WC_RECV;
+	wc.vendor_err = 0;
+	wc.qp = &qp->ibqp;
+	wc.src_qp = src_qp;
+	wc.pkey_index = qp->ibqp.qp_type == IB_QPT_GSI ?
+		qib_lookup_pkey(ibp, be32_to_cpu(ohdr->bth[0])) : 0;
+	wc.slid = be16_to_cpu(hdr->lrh[3]);
+	wc.sl = (be16_to_cpu(hdr->lrh[0]) >> 4) & 0xF;
+	dlid = be16_to_cpu(hdr->lrh[1]);
+	/*
+	 * Save the LMC lower bits if the destination LID is a unicast LID.
+	 */
+	wc.dlid_path_bits = dlid >= QIB_MULTICAST_LID_BASE ? 0 :
+		dlid & ((1 << ppd_from_ibp(ibp)->lmc) - 1);
+	wc.port_num = qp->port_num;
+	/* Signal completion event if the solicited bit is set. */
+	qib_cq_enter(to_icq(qp->ibqp.recv_cq), &wc,
+		     (ohdr->bth[0] &
+			cpu_to_be32(IB_BTH_SOLICITED)) != 0);
+bail_unlock:
+	spin_unlock(&qp->r_lock);
+bail:;
+}
diff --git a/drivers/infiniband/hw/qib/qib_user_pages.c b/drivers/infiniband/hw/qib/qib_user_pages.c
new file mode 100644
index 000000000000..d7a26c1d4f37
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_user_pages.c
@@ -0,0 +1,157 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/mm.h>
+#include <linux/device.h>
+
+#include "qib.h"
+
+static void __qib_release_user_pages(struct page **p, size_t num_pages,
+				     int dirty)
+{
+	size_t i;
+
+	for (i = 0; i < num_pages; i++) {
+		if (dirty)
+			set_page_dirty_lock(p[i]);
+		put_page(p[i]);
+	}
+}
+
+/*
+ * Call with current->mm->mmap_sem held.
+ */
+static int __get_user_pages(unsigned long start_page, size_t num_pages,
+			    struct page **p, struct vm_area_struct **vma)
+{
+	unsigned long lock_limit;
+	size_t got;
+	int ret;
+
+	lock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
+
+	if (num_pages > lock_limit && !capable(CAP_IPC_LOCK)) {
+		ret = -ENOMEM;
+		goto bail;
+	}
+
+	for (got = 0; got < num_pages; got += ret) {
+		ret = get_user_pages(current, current->mm,
+				     start_page + got * PAGE_SIZE,
+				     num_pages - got, 1, 1,
+				     p + got, vma);
+		if (ret < 0)
+			goto bail_release;
+	}
+
+	current->mm->locked_vm += num_pages;
+
+	ret = 0;
+	goto bail;
+
+bail_release:
+	__qib_release_user_pages(p, got, 0);
+bail:
+	return ret;
+}
+
+/**
+ * qib_map_page - a safety wrapper around pci_map_page()
+ *
+ * A dma_addr of all 0's is interpreted by the chip as "disabled".
+ * Unfortunately, it can also be a valid dma_addr returned on some
+ * architectures.
+ *
+ * The powerpc iommu assigns dma_addrs in ascending order, so we don't
+ * have to bother with retries or mapping a dummy page to insure we
+ * don't just get the same mapping again.
+ *
+ * I'm sure we won't be so lucky with other iommu's, so FIXME.
+ */
+dma_addr_t qib_map_page(struct pci_dev *hwdev, struct page *page,
+			unsigned long offset, size_t size, int direction)
+{
+	dma_addr_t phys;
+
+	phys = pci_map_page(hwdev, page, offset, size, direction);
+
+	if (phys == 0) {
+		pci_unmap_page(hwdev, phys, size, direction);
+		phys = pci_map_page(hwdev, page, offset, size, direction);
+		/*
+		 * FIXME: If we get 0 again, we should keep this page,
+		 * map another, then free the 0 page.
+		 */
+	}
+
+	return phys;
+}
+
+/**
+ * qib_get_user_pages - lock user pages into memory
+ * @start_page: the start page
+ * @num_pages: the number of pages
+ * @p: the output page structures
+ *
+ * This function takes a given start page (page aligned user virtual
+ * address) and pins it and the following specified number of pages.  For
+ * now, num_pages is always 1, but that will probably change at some point
+ * (because caller is doing expected sends on a single virtually contiguous
+ * buffer, so we can do all pages at once).
+ */
+int qib_get_user_pages(unsigned long start_page, size_t num_pages,
+		       struct page **p)
+{
+	int ret;
+
+	down_write(&current->mm->mmap_sem);
+
+	ret = __get_user_pages(start_page, num_pages, p, NULL);
+
+	up_write(&current->mm->mmap_sem);
+
+	return ret;
+}
+
+void qib_release_user_pages(struct page **p, size_t num_pages)
+{
+	if (current->mm) /* during close after signal, mm can be NULL */
+		down_write(&current->mm->mmap_sem);
+
+	__qib_release_user_pages(p, num_pages, 1);
+
+	if (current->mm) {
+		current->mm->locked_vm -= num_pages;
+		up_write(&current->mm->mmap_sem);
+	}
+}
diff --git a/drivers/infiniband/hw/qib/qib_user_sdma.c b/drivers/infiniband/hw/qib/qib_user_sdma.c
new file mode 100644
index 000000000000..4c19e06b5e85
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_user_sdma.c
@@ -0,0 +1,897 @@
+/*
+ * Copyright (c) 2007, 2008, 2009 QLogic Corporation. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/mm.h>
+#include <linux/types.h>
+#include <linux/device.h>
+#include <linux/dmapool.h>
+#include <linux/slab.h>
+#include <linux/list.h>
+#include <linux/highmem.h>
+#include <linux/io.h>
+#include <linux/uio.h>
+#include <linux/rbtree.h>
+#include <linux/spinlock.h>
+#include <linux/delay.h>
+
+#include "qib.h"
+#include "qib_user_sdma.h"
+
+/* minimum size of header */
+#define QIB_USER_SDMA_MIN_HEADER_LENGTH 64
+/* expected size of headers (for dma_pool) */
+#define QIB_USER_SDMA_EXP_HEADER_LENGTH 64
+/* attempt to drain the queue for 5secs */
+#define QIB_USER_SDMA_DRAIN_TIMEOUT 500
+
+struct qib_user_sdma_pkt {
+	u8 naddr;               /* dimension of addr (1..3) ... */
+	u32 counter;            /* sdma pkts queued counter for this entry */
+	u64 added;              /* global descq number of entries */
+
+	struct {
+		u32 offset;                     /* offset for kvaddr, addr */
+		u32 length;                     /* length in page */
+		u8  put_page;                   /* should we put_page? */
+		u8  dma_mapped;                 /* is page dma_mapped? */
+		struct page *page;              /* may be NULL (coherent mem) */
+		void *kvaddr;                   /* FIXME: only for pio hack */
+		dma_addr_t addr;
+	} addr[4];   /* max pages, any more and we coalesce */
+	struct list_head list;  /* list element */
+};
+
+struct qib_user_sdma_queue {
+	/*
+	 * pkts sent to dma engine are queued on this
+	 * list head.  the type of the elements of this
+	 * list are struct qib_user_sdma_pkt...
+	 */
+	struct list_head sent;
+
+	/* headers with expected length are allocated from here... */
+	char header_cache_name[64];
+	struct dma_pool *header_cache;
+
+	/* packets are allocated from the slab cache... */
+	char pkt_slab_name[64];
+	struct kmem_cache *pkt_slab;
+
+	/* as packets go on the queued queue, they are counted... */
+	u32 counter;
+	u32 sent_counter;
+
+	/* dma page table */
+	struct rb_root dma_pages_root;
+
+	/* protect everything above... */
+	struct mutex lock;
+};
+
+struct qib_user_sdma_queue *
+qib_user_sdma_queue_create(struct device *dev, int unit, int ctxt, int sctxt)
+{
+	struct qib_user_sdma_queue *pq =
+		kmalloc(sizeof(struct qib_user_sdma_queue), GFP_KERNEL);
+
+	if (!pq)
+		goto done;
+
+	pq->counter = 0;
+	pq->sent_counter = 0;
+	INIT_LIST_HEAD(&pq->sent);
+
+	mutex_init(&pq->lock);
+
+	snprintf(pq->pkt_slab_name, sizeof(pq->pkt_slab_name),
+		 "qib-user-sdma-pkts-%u-%02u.%02u", unit, ctxt, sctxt);
+	pq->pkt_slab = kmem_cache_create(pq->pkt_slab_name,
+					 sizeof(struct qib_user_sdma_pkt),
+					 0, 0, NULL);
+
+	if (!pq->pkt_slab)
+		goto err_kfree;
+
+	snprintf(pq->header_cache_name, sizeof(pq->header_cache_name),
+		 "qib-user-sdma-headers-%u-%02u.%02u", unit, ctxt, sctxt);
+	pq->header_cache = dma_pool_create(pq->header_cache_name,
+					   dev,
+					   QIB_USER_SDMA_EXP_HEADER_LENGTH,
+					   4, 0);
+	if (!pq->header_cache)
+		goto err_slab;
+
+	pq->dma_pages_root = RB_ROOT;
+
+	goto done;
+
+err_slab:
+	kmem_cache_destroy(pq->pkt_slab);
+err_kfree:
+	kfree(pq);
+	pq = NULL;
+
+done:
+	return pq;
+}
+
+static void qib_user_sdma_init_frag(struct qib_user_sdma_pkt *pkt,
+				    int i, size_t offset, size_t len,
+				    int put_page, int dma_mapped,
+				    struct page *page,
+				    void *kvaddr, dma_addr_t dma_addr)
+{
+	pkt->addr[i].offset = offset;
+	pkt->addr[i].length = len;
+	pkt->addr[i].put_page = put_page;
+	pkt->addr[i].dma_mapped = dma_mapped;
+	pkt->addr[i].page = page;
+	pkt->addr[i].kvaddr = kvaddr;
+	pkt->addr[i].addr = dma_addr;
+}
+
+static void qib_user_sdma_init_header(struct qib_user_sdma_pkt *pkt,
+				      u32 counter, size_t offset,
+				      size_t len, int dma_mapped,
+				      struct page *page,
+				      void *kvaddr, dma_addr_t dma_addr)
+{
+	pkt->naddr = 1;
+	pkt->counter = counter;
+	qib_user_sdma_init_frag(pkt, 0, offset, len, 0, dma_mapped, page,
+				kvaddr, dma_addr);
+}
+
+/* we've too many pages in the iovec, coalesce to a single page */
+static int qib_user_sdma_coalesce(const struct qib_devdata *dd,
+				  struct qib_user_sdma_pkt *pkt,
+				  const struct iovec *iov,
+				  unsigned long niov)
+{
+	int ret = 0;
+	struct page *page = alloc_page(GFP_KERNEL);
+	void *mpage_save;
+	char *mpage;
+	int i;
+	int len = 0;
+	dma_addr_t dma_addr;
+
+	if (!page) {
+		ret = -ENOMEM;
+		goto done;
+	}
+
+	mpage = kmap(page);
+	mpage_save = mpage;
+	for (i = 0; i < niov; i++) {
+		int cfur;
+
+		cfur = copy_from_user(mpage,
+				      iov[i].iov_base, iov[i].iov_len);
+		if (cfur) {
+			ret = -EFAULT;
+			goto free_unmap;
+		}
+
+		mpage += iov[i].iov_len;
+		len += iov[i].iov_len;
+	}
+
+	dma_addr = dma_map_page(&dd->pcidev->dev, page, 0, len,
+				DMA_TO_DEVICE);
+	if (dma_mapping_error(&dd->pcidev->dev, dma_addr)) {
+		ret = -ENOMEM;
+		goto free_unmap;
+	}
+
+	qib_user_sdma_init_frag(pkt, 1, 0, len, 0, 1, page, mpage_save,
+				dma_addr);
+	pkt->naddr = 2;
+
+	goto done;
+
+free_unmap:
+	kunmap(page);
+	__free_page(page);
+done:
+	return ret;
+}
+
+/*
+ * How many pages in this iovec element?
+ */
+static int qib_user_sdma_num_pages(const struct iovec *iov)
+{
+	const unsigned long addr  = (unsigned long) iov->iov_base;
+	const unsigned long  len  = iov->iov_len;
+	const unsigned long spage = addr & PAGE_MASK;
+	const unsigned long epage = (addr + len - 1) & PAGE_MASK;
+
+	return 1 + ((epage - spage) >> PAGE_SHIFT);
+}
+
+/*
+ * Truncate length to page boundry.
+ */
+static int qib_user_sdma_page_length(unsigned long addr, unsigned long len)
+{
+	const unsigned long offset = addr & ~PAGE_MASK;
+
+	return ((offset + len) > PAGE_SIZE) ? (PAGE_SIZE - offset) : len;
+}
+
+static void qib_user_sdma_free_pkt_frag(struct device *dev,
+					struct qib_user_sdma_queue *pq,
+					struct qib_user_sdma_pkt *pkt,
+					int frag)
+{
+	const int i = frag;
+
+	if (pkt->addr[i].page) {
+		if (pkt->addr[i].dma_mapped)
+			dma_unmap_page(dev,
+				       pkt->addr[i].addr,
+				       pkt->addr[i].length,
+				       DMA_TO_DEVICE);
+
+		if (pkt->addr[i].kvaddr)
+			kunmap(pkt->addr[i].page);
+
+		if (pkt->addr[i].put_page)
+			put_page(pkt->addr[i].page);
+		else
+			__free_page(pkt->addr[i].page);
+	} else if (pkt->addr[i].kvaddr)
+		/* free coherent mem from cache... */
+		dma_pool_free(pq->header_cache,
+			      pkt->addr[i].kvaddr, pkt->addr[i].addr);
+}
+
+/* return number of pages pinned... */
+static int qib_user_sdma_pin_pages(const struct qib_devdata *dd,
+				   struct qib_user_sdma_pkt *pkt,
+				   unsigned long addr, int tlen, int npages)
+{
+	struct page *pages[2];
+	int j;
+	int ret;
+
+	ret = get_user_pages(current, current->mm, addr,
+			     npages, 0, 1, pages, NULL);
+
+	if (ret != npages) {
+		int i;
+
+		for (i = 0; i < ret; i++)
+			put_page(pages[i]);
+
+		ret = -ENOMEM;
+		goto done;
+	}
+
+	for (j = 0; j < npages; j++) {
+		/* map the pages... */
+		const int flen = qib_user_sdma_page_length(addr, tlen);
+		dma_addr_t dma_addr =
+			dma_map_page(&dd->pcidev->dev,
+				     pages[j], 0, flen, DMA_TO_DEVICE);
+		unsigned long fofs = addr & ~PAGE_MASK;
+
+		if (dma_mapping_error(&dd->pcidev->dev, dma_addr)) {
+			ret = -ENOMEM;
+			goto done;
+		}
+
+		qib_user_sdma_init_frag(pkt, pkt->naddr, fofs, flen, 1, 1,
+					pages[j], kmap(pages[j]), dma_addr);
+
+		pkt->naddr++;
+		addr += flen;
+		tlen -= flen;
+	}
+
+done:
+	return ret;
+}
+
+static int qib_user_sdma_pin_pkt(const struct qib_devdata *dd,
+				 struct qib_user_sdma_queue *pq,
+				 struct qib_user_sdma_pkt *pkt,
+				 const struct iovec *iov,
+				 unsigned long niov)
+{
+	int ret = 0;
+	unsigned long idx;
+
+	for (idx = 0; idx < niov; idx++) {
+		const int npages = qib_user_sdma_num_pages(iov + idx);
+		const unsigned long addr = (unsigned long) iov[idx].iov_base;
+
+		ret = qib_user_sdma_pin_pages(dd, pkt, addr,
+					      iov[idx].iov_len, npages);
+		if (ret < 0)
+			goto free_pkt;
+	}
+
+	goto done;
+
+free_pkt:
+	for (idx = 0; idx < pkt->naddr; idx++)
+		qib_user_sdma_free_pkt_frag(&dd->pcidev->dev, pq, pkt, idx);
+
+done:
+	return ret;
+}
+
+static int qib_user_sdma_init_payload(const struct qib_devdata *dd,
+				      struct qib_user_sdma_queue *pq,
+				      struct qib_user_sdma_pkt *pkt,
+				      const struct iovec *iov,
+				      unsigned long niov, int npages)
+{
+	int ret = 0;
+
+	if (npages >= ARRAY_SIZE(pkt->addr))
+		ret = qib_user_sdma_coalesce(dd, pkt, iov, niov);
+	else
+		ret = qib_user_sdma_pin_pkt(dd, pq, pkt, iov, niov);
+
+	return ret;
+}
+
+/* free a packet list -- return counter value of last packet */
+static void qib_user_sdma_free_pkt_list(struct device *dev,
+					struct qib_user_sdma_queue *pq,
+					struct list_head *list)
+{
+	struct qib_user_sdma_pkt *pkt, *pkt_next;
+
+	list_for_each_entry_safe(pkt, pkt_next, list, list) {
+		int i;
+
+		for (i = 0; i < pkt->naddr; i++)
+			qib_user_sdma_free_pkt_frag(dev, pq, pkt, i);
+
+		kmem_cache_free(pq->pkt_slab, pkt);
+	}
+}
+
+/*
+ * copy headers, coalesce etc -- pq->lock must be held
+ *
+ * we queue all the packets to list, returning the
+ * number of bytes total.  list must be empty initially,
+ * as, if there is an error we clean it...
+ */
+static int qib_user_sdma_queue_pkts(const struct qib_devdata *dd,
+				    struct qib_user_sdma_queue *pq,
+				    struct list_head *list,
+				    const struct iovec *iov,
+				    unsigned long niov,
+				    int maxpkts)
+{
+	unsigned long idx = 0;
+	int ret = 0;
+	int npkts = 0;
+	struct page *page = NULL;
+	__le32 *pbc;
+	dma_addr_t dma_addr;
+	struct qib_user_sdma_pkt *pkt = NULL;
+	size_t len;
+	size_t nw;
+	u32 counter = pq->counter;
+	int dma_mapped = 0;
+
+	while (idx < niov && npkts < maxpkts) {
+		const unsigned long addr = (unsigned long) iov[idx].iov_base;
+		const unsigned long idx_save = idx;
+		unsigned pktnw;
+		unsigned pktnwc;
+		int nfrags = 0;
+		int npages = 0;
+		int cfur;
+
+		dma_mapped = 0;
+		len = iov[idx].iov_len;
+		nw = len >> 2;
+		page = NULL;
+
+		pkt = kmem_cache_alloc(pq->pkt_slab, GFP_KERNEL);
+		if (!pkt) {
+			ret = -ENOMEM;
+			goto free_list;
+		}
+
+		if (len < QIB_USER_SDMA_MIN_HEADER_LENGTH ||
+		    len > PAGE_SIZE || len & 3 || addr & 3) {
+			ret = -EINVAL;
+			goto free_pkt;
+		}
+
+		if (len == QIB_USER_SDMA_EXP_HEADER_LENGTH)
+			pbc = dma_pool_alloc(pq->header_cache, GFP_KERNEL,
+					     &dma_addr);
+		else
+			pbc = NULL;
+
+		if (!pbc) {
+			page = alloc_page(GFP_KERNEL);
+			if (!page) {
+				ret = -ENOMEM;
+				goto free_pkt;
+			}
+			pbc = kmap(page);
+		}
+
+		cfur = copy_from_user(pbc, iov[idx].iov_base, len);
+		if (cfur) {
+			ret = -EFAULT;
+			goto free_pbc;
+		}
+
+		/*
+		 * This assignment is a bit strange.  it's because the
+		 * the pbc counts the number of 32 bit words in the full
+		 * packet _except_ the first word of the pbc itself...
+		 */
+		pktnwc = nw - 1;
+
+		/*
+		 * pktnw computation yields the number of 32 bit words
+		 * that the caller has indicated in the PBC.  note that
+		 * this is one less than the total number of words that
+		 * goes to the send DMA engine as the first 32 bit word
+		 * of the PBC itself is not counted.  Armed with this count,
+		 * we can verify that the packet is consistent with the
+		 * iovec lengths.
+		 */
+		pktnw = le32_to_cpu(*pbc) & QIB_PBC_LENGTH_MASK;
+		if (pktnw < pktnwc || pktnw > pktnwc + (PAGE_SIZE >> 2)) {
+			ret = -EINVAL;
+			goto free_pbc;
+		}
+
+		idx++;
+		while (pktnwc < pktnw && idx < niov) {
+			const size_t slen = iov[idx].iov_len;
+			const unsigned long faddr =
+				(unsigned long) iov[idx].iov_base;
+
+			if (slen & 3 || faddr & 3 || !slen ||
+			    slen > PAGE_SIZE) {
+				ret = -EINVAL;
+				goto free_pbc;
+			}
+
+			npages++;
+			if ((faddr & PAGE_MASK) !=
+			    ((faddr + slen - 1) & PAGE_MASK))
+				npages++;
+
+			pktnwc += slen >> 2;
+			idx++;
+			nfrags++;
+		}
+
+		if (pktnwc != pktnw) {
+			ret = -EINVAL;
+			goto free_pbc;
+		}
+
+		if (page) {
+			dma_addr = dma_map_page(&dd->pcidev->dev,
+						page, 0, len, DMA_TO_DEVICE);
+			if (dma_mapping_error(&dd->pcidev->dev, dma_addr)) {
+				ret = -ENOMEM;
+				goto free_pbc;
+			}
+
+			dma_mapped = 1;
+		}
+
+		qib_user_sdma_init_header(pkt, counter, 0, len, dma_mapped,
+					  page, pbc, dma_addr);
+
+		if (nfrags) {
+			ret = qib_user_sdma_init_payload(dd, pq, pkt,
+							 iov + idx_save + 1,
+							 nfrags, npages);
+			if (ret < 0)
+				goto free_pbc_dma;
+		}
+
+		counter++;
+		npkts++;
+
+		list_add_tail(&pkt->list, list);
+	}
+
+	ret = idx;
+	goto done;
+
+free_pbc_dma:
+	if (dma_mapped)
+		dma_unmap_page(&dd->pcidev->dev, dma_addr, len, DMA_TO_DEVICE);
+free_pbc:
+	if (page) {
+		kunmap(page);
+		__free_page(page);
+	} else
+		dma_pool_free(pq->header_cache, pbc, dma_addr);
+free_pkt:
+	kmem_cache_free(pq->pkt_slab, pkt);
+free_list:
+	qib_user_sdma_free_pkt_list(&dd->pcidev->dev, pq, list);
+done:
+	return ret;
+}
+
+static void qib_user_sdma_set_complete_counter(struct qib_user_sdma_queue *pq,
+					       u32 c)
+{
+	pq->sent_counter = c;
+}
+
+/* try to clean out queue -- needs pq->lock */
+static int qib_user_sdma_queue_clean(struct qib_pportdata *ppd,
+				     struct qib_user_sdma_queue *pq)
+{
+	struct qib_devdata *dd = ppd->dd;
+	struct list_head free_list;
+	struct qib_user_sdma_pkt *pkt;
+	struct qib_user_sdma_pkt *pkt_prev;
+	int ret = 0;
+
+	INIT_LIST_HEAD(&free_list);
+
+	list_for_each_entry_safe(pkt, pkt_prev, &pq->sent, list) {
+		s64 descd = ppd->sdma_descq_removed - pkt->added;
+
+		if (descd < 0)
+			break;
+
+		list_move_tail(&pkt->list, &free_list);
+
+		/* one more packet cleaned */
+		ret++;
+	}
+
+	if (!list_empty(&free_list)) {
+		u32 counter;
+
+		pkt = list_entry(free_list.prev,
+				 struct qib_user_sdma_pkt, list);
+		counter = pkt->counter;
+
+		qib_user_sdma_free_pkt_list(&dd->pcidev->dev, pq, &free_list);
+		qib_user_sdma_set_complete_counter(pq, counter);
+	}
+
+	return ret;
+}
+
+void qib_user_sdma_queue_destroy(struct qib_user_sdma_queue *pq)
+{
+	if (!pq)
+		return;
+
+	kmem_cache_destroy(pq->pkt_slab);
+	dma_pool_destroy(pq->header_cache);
+	kfree(pq);
+}
+
+/* clean descriptor queue, returns > 0 if some elements cleaned */
+static int qib_user_sdma_hwqueue_clean(struct qib_pportdata *ppd)
+{
+	int ret;
+	unsigned long flags;
+
+	spin_lock_irqsave(&ppd->sdma_lock, flags);
+	ret = qib_sdma_make_progress(ppd);
+	spin_unlock_irqrestore(&ppd->sdma_lock, flags);
+
+	return ret;
+}
+
+/* we're in close, drain packets so that we can cleanup successfully... */
+void qib_user_sdma_queue_drain(struct qib_pportdata *ppd,
+			       struct qib_user_sdma_queue *pq)
+{
+	struct qib_devdata *dd = ppd->dd;
+	int i;
+
+	if (!pq)
+		return;
+
+	for (i = 0; i < QIB_USER_SDMA_DRAIN_TIMEOUT; i++) {
+		mutex_lock(&pq->lock);
+		if (list_empty(&pq->sent)) {
+			mutex_unlock(&pq->lock);
+			break;
+		}
+		qib_user_sdma_hwqueue_clean(ppd);
+		qib_user_sdma_queue_clean(ppd, pq);
+		mutex_unlock(&pq->lock);
+		msleep(10);
+	}
+
+	if (!list_empty(&pq->sent)) {
+		struct list_head free_list;
+
+		qib_dev_err(dd, "user sdma lists not empty: forcing!\n");
+		INIT_LIST_HEAD(&free_list);
+		mutex_lock(&pq->lock);
+		list_splice_init(&pq->sent, &free_list);
+		qib_user_sdma_free_pkt_list(&dd->pcidev->dev, pq, &free_list);
+		mutex_unlock(&pq->lock);
+	}
+}
+
+static inline __le64 qib_sdma_make_desc0(struct qib_pportdata *ppd,
+					 u64 addr, u64 dwlen, u64 dwoffset)
+{
+	u8 tmpgen;
+
+	tmpgen = ppd->sdma_generation;
+
+	return cpu_to_le64(/* SDmaPhyAddr[31:0] */
+			   ((addr & 0xfffffffcULL) << 32) |
+			   /* SDmaGeneration[1:0] */
+			   ((tmpgen & 3ULL) << 30) |
+			   /* SDmaDwordCount[10:0] */
+			   ((dwlen & 0x7ffULL) << 16) |
+			   /* SDmaBufOffset[12:2] */
+			   (dwoffset & 0x7ffULL));
+}
+
+static inline __le64 qib_sdma_make_first_desc0(__le64 descq)
+{
+	return descq | cpu_to_le64(1ULL << 12);
+}
+
+static inline __le64 qib_sdma_make_last_desc0(__le64 descq)
+{
+					      /* last */  /* dma head */
+	return descq | cpu_to_le64(1ULL << 11 | 1ULL << 13);
+}
+
+static inline __le64 qib_sdma_make_desc1(u64 addr)
+{
+	/* SDmaPhyAddr[47:32] */
+	return cpu_to_le64(addr >> 32);
+}
+
+static void qib_user_sdma_send_frag(struct qib_pportdata *ppd,
+				    struct qib_user_sdma_pkt *pkt, int idx,
+				    unsigned ofs, u16 tail)
+{
+	const u64 addr = (u64) pkt->addr[idx].addr +
+		(u64) pkt->addr[idx].offset;
+	const u64 dwlen = (u64) pkt->addr[idx].length / 4;
+	__le64 *descqp;
+	__le64 descq0;
+
+	descqp = &ppd->sdma_descq[tail].qw[0];
+
+	descq0 = qib_sdma_make_desc0(ppd, addr, dwlen, ofs);
+	if (idx == 0)
+		descq0 = qib_sdma_make_first_desc0(descq0);
+	if (idx == pkt->naddr - 1)
+		descq0 = qib_sdma_make_last_desc0(descq0);
+
+	descqp[0] = descq0;
+	descqp[1] = qib_sdma_make_desc1(addr);
+}
+
+/* pq->lock must be held, get packets on the wire... */
+static int qib_user_sdma_push_pkts(struct qib_pportdata *ppd,
+				   struct qib_user_sdma_queue *pq,
+				   struct list_head *pktlist)
+{
+	struct qib_devdata *dd = ppd->dd;
+	int ret = 0;
+	unsigned long flags;
+	u16 tail;
+	u8 generation;
+	u64 descq_added;
+
+	if (list_empty(pktlist))
+		return 0;
+
+	if (unlikely(!(ppd->lflags & QIBL_LINKACTIVE)))
+		return -ECOMM;
+
+	spin_lock_irqsave(&ppd->sdma_lock, flags);
+
+	/* keep a copy for restoring purposes in case of problems */
+	generation = ppd->sdma_generation;
+	descq_added = ppd->sdma_descq_added;
+
+	if (unlikely(!__qib_sdma_running(ppd))) {
+		ret = -ECOMM;
+		goto unlock;
+	}
+
+	tail = ppd->sdma_descq_tail;
+	while (!list_empty(pktlist)) {
+		struct qib_user_sdma_pkt *pkt =
+			list_entry(pktlist->next, struct qib_user_sdma_pkt,
+				   list);
+		int i;
+		unsigned ofs = 0;
+		u16 dtail = tail;
+
+		if (pkt->naddr > qib_sdma_descq_freecnt(ppd))
+			goto unlock_check_tail;
+
+		for (i = 0; i < pkt->naddr; i++) {
+			qib_user_sdma_send_frag(ppd, pkt, i, ofs, tail);
+			ofs += pkt->addr[i].length >> 2;
+
+			if (++tail == ppd->sdma_descq_cnt) {
+				tail = 0;
+				++ppd->sdma_generation;
+			}
+		}
+
+		if ((ofs << 2) > ppd->ibmaxlen) {
+			ret = -EMSGSIZE;
+			goto unlock;
+		}
+
+		/*
+		 * If the packet is >= 2KB mtu equivalent, we have to use
+		 * the large buffers, and have to mark each descriptor as
+		 * part of a large buffer packet.
+		 */
+		if (ofs > dd->piosize2kmax_dwords) {
+			for (i = 0; i < pkt->naddr; i++) {
+				ppd->sdma_descq[dtail].qw[0] |=
+					cpu_to_le64(1ULL << 14);
+				if (++dtail == ppd->sdma_descq_cnt)
+					dtail = 0;
+			}
+		}
+
+		ppd->sdma_descq_added += pkt->naddr;
+		pkt->added = ppd->sdma_descq_added;
+		list_move_tail(&pkt->list, &pq->sent);
+		ret++;
+	}
+
+unlock_check_tail:
+	/* advance the tail on the chip if necessary */
+	if (ppd->sdma_descq_tail != tail)
+		dd->f_sdma_update_tail(ppd, tail);
+
+unlock:
+	if (unlikely(ret < 0)) {
+		ppd->sdma_generation = generation;
+		ppd->sdma_descq_added = descq_added;
+	}
+	spin_unlock_irqrestore(&ppd->sdma_lock, flags);
+
+	return ret;
+}
+
+int qib_user_sdma_writev(struct qib_ctxtdata *rcd,
+			 struct qib_user_sdma_queue *pq,
+			 const struct iovec *iov,
+			 unsigned long dim)
+{
+	struct qib_devdata *dd = rcd->dd;
+	struct qib_pportdata *ppd = rcd->ppd;
+	int ret = 0;
+	struct list_head list;
+	int npkts = 0;
+
+	INIT_LIST_HEAD(&list);
+
+	mutex_lock(&pq->lock);
+
+	/* why not -ECOMM like qib_user_sdma_push_pkts() below? */
+	if (!qib_sdma_running(ppd))
+		goto done_unlock;
+
+	if (ppd->sdma_descq_added != ppd->sdma_descq_removed) {
+		qib_user_sdma_hwqueue_clean(ppd);
+		qib_user_sdma_queue_clean(ppd, pq);
+	}
+
+	while (dim) {
+		const int mxp = 8;
+
+		down_write(&current->mm->mmap_sem);
+		ret = qib_user_sdma_queue_pkts(dd, pq, &list, iov, dim, mxp);
+		up_write(&current->mm->mmap_sem);
+
+		if (ret <= 0)
+			goto done_unlock;
+		else {
+			dim -= ret;
+			iov += ret;
+		}
+
+		/* force packets onto the sdma hw queue... */
+		if (!list_empty(&list)) {
+			/*
+			 * Lazily clean hw queue.  the 4 is a guess of about
+			 * how many sdma descriptors a packet will take (it
+			 * doesn't have to be perfect).
+			 */
+			if (qib_sdma_descq_freecnt(ppd) < ret * 4) {
+				qib_user_sdma_hwqueue_clean(ppd);
+				qib_user_sdma_queue_clean(ppd, pq);
+			}
+
+			ret = qib_user_sdma_push_pkts(ppd, pq, &list);
+			if (ret < 0)
+				goto done_unlock;
+			else {
+				npkts += ret;
+				pq->counter += ret;
+
+				if (!list_empty(&list))
+					goto done_unlock;
+			}
+		}
+	}
+
+done_unlock:
+	if (!list_empty(&list))
+		qib_user_sdma_free_pkt_list(&dd->pcidev->dev, pq, &list);
+	mutex_unlock(&pq->lock);
+
+	return (ret < 0) ? ret : npkts;
+}
+
+int qib_user_sdma_make_progress(struct qib_pportdata *ppd,
+				struct qib_user_sdma_queue *pq)
+{
+	int ret = 0;
+
+	mutex_lock(&pq->lock);
+	qib_user_sdma_hwqueue_clean(ppd);
+	ret = qib_user_sdma_queue_clean(ppd, pq);
+	mutex_unlock(&pq->lock);
+
+	return ret;
+}
+
+u32 qib_user_sdma_complete_counter(const struct qib_user_sdma_queue *pq)
+{
+	return pq ? pq->sent_counter : 0;
+}
+
+u32 qib_user_sdma_inflight_counter(struct qib_user_sdma_queue *pq)
+{
+	return pq ? pq->counter : 0;
+}
diff --git a/drivers/infiniband/hw/qib/qib_user_sdma.h b/drivers/infiniband/hw/qib/qib_user_sdma.h
new file mode 100644
index 000000000000..ce8cbaf6a5c2
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_user_sdma.h
@@ -0,0 +1,52 @@
+/*
+ * Copyright (c) 2007, 2008 QLogic Corporation. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/device.h>
+
+struct qib_user_sdma_queue;
+
+struct qib_user_sdma_queue *
+qib_user_sdma_queue_create(struct device *dev, int unit, int port, int sport);
+void qib_user_sdma_queue_destroy(struct qib_user_sdma_queue *pq);
+
+int qib_user_sdma_writev(struct qib_ctxtdata *pd,
+			 struct qib_user_sdma_queue *pq,
+			 const struct iovec *iov,
+			 unsigned long dim);
+
+int qib_user_sdma_make_progress(struct qib_pportdata *ppd,
+				struct qib_user_sdma_queue *pq);
+
+void qib_user_sdma_queue_drain(struct qib_pportdata *ppd,
+			       struct qib_user_sdma_queue *pq);
+
+u32 qib_user_sdma_complete_counter(const struct qib_user_sdma_queue *pq);
+u32 qib_user_sdma_inflight_counter(struct qib_user_sdma_queue *pq);
diff --git a/drivers/infiniband/hw/qib/qib_verbs.c b/drivers/infiniband/hw/qib/qib_verbs.c
new file mode 100644
index 000000000000..cda8f4173d23
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_verbs.c
@@ -0,0 +1,2248 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009, 2010 QLogic Corporation.
+ * All rights reserved.
+ * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <rdma/ib_mad.h>
+#include <rdma/ib_user_verbs.h>
+#include <linux/io.h>
+#include <linux/utsname.h>
+#include <linux/rculist.h>
+#include <linux/mm.h>
+
+#include "qib.h"
+#include "qib_common.h"
+
+static unsigned int ib_qib_qp_table_size = 251;
+module_param_named(qp_table_size, ib_qib_qp_table_size, uint, S_IRUGO);
+MODULE_PARM_DESC(qp_table_size, "QP table size");
+
+unsigned int ib_qib_lkey_table_size = 16;
+module_param_named(lkey_table_size, ib_qib_lkey_table_size, uint,
+		   S_IRUGO);
+MODULE_PARM_DESC(lkey_table_size,
+		 "LKEY table size in bits (2^n, 1 <= n <= 23)");
+
+static unsigned int ib_qib_max_pds = 0xFFFF;
+module_param_named(max_pds, ib_qib_max_pds, uint, S_IRUGO);
+MODULE_PARM_DESC(max_pds,
+		 "Maximum number of protection domains to support");
+
+static unsigned int ib_qib_max_ahs = 0xFFFF;
+module_param_named(max_ahs, ib_qib_max_ahs, uint, S_IRUGO);
+MODULE_PARM_DESC(max_ahs, "Maximum number of address handles to support");
+
+unsigned int ib_qib_max_cqes = 0x2FFFF;
+module_param_named(max_cqes, ib_qib_max_cqes, uint, S_IRUGO);
+MODULE_PARM_DESC(max_cqes,
+		 "Maximum number of completion queue entries to support");
+
+unsigned int ib_qib_max_cqs = 0x1FFFF;
+module_param_named(max_cqs, ib_qib_max_cqs, uint, S_IRUGO);
+MODULE_PARM_DESC(max_cqs, "Maximum number of completion queues to support");
+
+unsigned int ib_qib_max_qp_wrs = 0x3FFF;
+module_param_named(max_qp_wrs, ib_qib_max_qp_wrs, uint, S_IRUGO);
+MODULE_PARM_DESC(max_qp_wrs, "Maximum number of QP WRs to support");
+
+unsigned int ib_qib_max_qps = 16384;
+module_param_named(max_qps, ib_qib_max_qps, uint, S_IRUGO);
+MODULE_PARM_DESC(max_qps, "Maximum number of QPs to support");
+
+unsigned int ib_qib_max_sges = 0x60;
+module_param_named(max_sges, ib_qib_max_sges, uint, S_IRUGO);
+MODULE_PARM_DESC(max_sges, "Maximum number of SGEs to support");
+
+unsigned int ib_qib_max_mcast_grps = 16384;
+module_param_named(max_mcast_grps, ib_qib_max_mcast_grps, uint, S_IRUGO);
+MODULE_PARM_DESC(max_mcast_grps,
+		 "Maximum number of multicast groups to support");
+
+unsigned int ib_qib_max_mcast_qp_attached = 16;
+module_param_named(max_mcast_qp_attached, ib_qib_max_mcast_qp_attached,
+		   uint, S_IRUGO);
+MODULE_PARM_DESC(max_mcast_qp_attached,
+		 "Maximum number of attached QPs to support");
+
+unsigned int ib_qib_max_srqs = 1024;
+module_param_named(max_srqs, ib_qib_max_srqs, uint, S_IRUGO);
+MODULE_PARM_DESC(max_srqs, "Maximum number of SRQs to support");
+
+unsigned int ib_qib_max_srq_sges = 128;
+module_param_named(max_srq_sges, ib_qib_max_srq_sges, uint, S_IRUGO);
+MODULE_PARM_DESC(max_srq_sges, "Maximum number of SRQ SGEs to support");
+
+unsigned int ib_qib_max_srq_wrs = 0x1FFFF;
+module_param_named(max_srq_wrs, ib_qib_max_srq_wrs, uint, S_IRUGO);
+MODULE_PARM_DESC(max_srq_wrs, "Maximum number of SRQ WRs support");
+
+static unsigned int ib_qib_disable_sma;
+module_param_named(disable_sma, ib_qib_disable_sma, uint, S_IWUSR | S_IRUGO);
+MODULE_PARM_DESC(disable_sma, "Disable the SMA");
+
+/*
+ * Note that it is OK to post send work requests in the SQE and ERR
+ * states; qib_do_send() will process them and generate error
+ * completions as per IB 1.2 C10-96.
+ */
+const int ib_qib_state_ops[IB_QPS_ERR + 1] = {
+	[IB_QPS_RESET] = 0,
+	[IB_QPS_INIT] = QIB_POST_RECV_OK,
+	[IB_QPS_RTR] = QIB_POST_RECV_OK | QIB_PROCESS_RECV_OK,
+	[IB_QPS_RTS] = QIB_POST_RECV_OK | QIB_PROCESS_RECV_OK |
+	    QIB_POST_SEND_OK | QIB_PROCESS_SEND_OK |
+	    QIB_PROCESS_NEXT_SEND_OK,
+	[IB_QPS_SQD] = QIB_POST_RECV_OK | QIB_PROCESS_RECV_OK |
+	    QIB_POST_SEND_OK | QIB_PROCESS_SEND_OK,
+	[IB_QPS_SQE] = QIB_POST_RECV_OK | QIB_PROCESS_RECV_OK |
+	    QIB_POST_SEND_OK | QIB_FLUSH_SEND,
+	[IB_QPS_ERR] = QIB_POST_RECV_OK | QIB_FLUSH_RECV |
+	    QIB_POST_SEND_OK | QIB_FLUSH_SEND,
+};
+
+struct qib_ucontext {
+	struct ib_ucontext ibucontext;
+};
+
+static inline struct qib_ucontext *to_iucontext(struct ib_ucontext
+						  *ibucontext)
+{
+	return container_of(ibucontext, struct qib_ucontext, ibucontext);
+}
+
+/*
+ * Translate ib_wr_opcode into ib_wc_opcode.
+ */
+const enum ib_wc_opcode ib_qib_wc_opcode[] = {
+	[IB_WR_RDMA_WRITE] = IB_WC_RDMA_WRITE,
+	[IB_WR_RDMA_WRITE_WITH_IMM] = IB_WC_RDMA_WRITE,
+	[IB_WR_SEND] = IB_WC_SEND,
+	[IB_WR_SEND_WITH_IMM] = IB_WC_SEND,
+	[IB_WR_RDMA_READ] = IB_WC_RDMA_READ,
+	[IB_WR_ATOMIC_CMP_AND_SWP] = IB_WC_COMP_SWAP,
+	[IB_WR_ATOMIC_FETCH_AND_ADD] = IB_WC_FETCH_ADD
+};
+
+/*
+ * System image GUID.
+ */
+__be64 ib_qib_sys_image_guid;
+
+/**
+ * qib_copy_sge - copy data to SGE memory
+ * @ss: the SGE state
+ * @data: the data to copy
+ * @length: the length of the data
+ */
+void qib_copy_sge(struct qib_sge_state *ss, void *data, u32 length, int release)
+{
+	struct qib_sge *sge = &ss->sge;
+
+	while (length) {
+		u32 len = sge->length;
+
+		if (len > length)
+			len = length;
+		if (len > sge->sge_length)
+			len = sge->sge_length;
+		BUG_ON(len == 0);
+		memcpy(sge->vaddr, data, len);
+		sge->vaddr += len;
+		sge->length -= len;
+		sge->sge_length -= len;
+		if (sge->sge_length == 0) {
+			if (release)
+				atomic_dec(&sge->mr->refcount);
+			if (--ss->num_sge)
+				*sge = *ss->sg_list++;
+		} else if (sge->length == 0 && sge->mr->lkey) {
+			if (++sge->n >= QIB_SEGSZ) {
+				if (++sge->m >= sge->mr->mapsz)
+					break;
+				sge->n = 0;
+			}
+			sge->vaddr =
+				sge->mr->map[sge->m]->segs[sge->n].vaddr;
+			sge->length =
+				sge->mr->map[sge->m]->segs[sge->n].length;
+		}
+		data += len;
+		length -= len;
+	}
+}
+
+/**
+ * qib_skip_sge - skip over SGE memory - XXX almost dup of prev func
+ * @ss: the SGE state
+ * @length: the number of bytes to skip
+ */
+void qib_skip_sge(struct qib_sge_state *ss, u32 length, int release)
+{
+	struct qib_sge *sge = &ss->sge;
+
+	while (length) {
+		u32 len = sge->length;
+
+		if (len > length)
+			len = length;
+		if (len > sge->sge_length)
+			len = sge->sge_length;
+		BUG_ON(len == 0);
+		sge->vaddr += len;
+		sge->length -= len;
+		sge->sge_length -= len;
+		if (sge->sge_length == 0) {
+			if (release)
+				atomic_dec(&sge->mr->refcount);
+			if (--ss->num_sge)
+				*sge = *ss->sg_list++;
+		} else if (sge->length == 0 && sge->mr->lkey) {
+			if (++sge->n >= QIB_SEGSZ) {
+				if (++sge->m >= sge->mr->mapsz)
+					break;
+				sge->n = 0;
+			}
+			sge->vaddr =
+				sge->mr->map[sge->m]->segs[sge->n].vaddr;
+			sge->length =
+				sge->mr->map[sge->m]->segs[sge->n].length;
+		}
+		length -= len;
+	}
+}
+
+/*
+ * Count the number of DMA descriptors needed to send length bytes of data.
+ * Don't modify the qib_sge_state to get the count.
+ * Return zero if any of the segments is not aligned.
+ */
+static u32 qib_count_sge(struct qib_sge_state *ss, u32 length)
+{
+	struct qib_sge *sg_list = ss->sg_list;
+	struct qib_sge sge = ss->sge;
+	u8 num_sge = ss->num_sge;
+	u32 ndesc = 1;  /* count the header */
+
+	while (length) {
+		u32 len = sge.length;
+
+		if (len > length)
+			len = length;
+		if (len > sge.sge_length)
+			len = sge.sge_length;
+		BUG_ON(len == 0);
+		if (((long) sge.vaddr & (sizeof(u32) - 1)) ||
+		    (len != length && (len & (sizeof(u32) - 1)))) {
+			ndesc = 0;
+			break;
+		}
+		ndesc++;
+		sge.vaddr += len;
+		sge.length -= len;
+		sge.sge_length -= len;
+		if (sge.sge_length == 0) {
+			if (--num_sge)
+				sge = *sg_list++;
+		} else if (sge.length == 0 && sge.mr->lkey) {
+			if (++sge.n >= QIB_SEGSZ) {
+				if (++sge.m >= sge.mr->mapsz)
+					break;
+				sge.n = 0;
+			}
+			sge.vaddr =
+				sge.mr->map[sge.m]->segs[sge.n].vaddr;
+			sge.length =
+				sge.mr->map[sge.m]->segs[sge.n].length;
+		}
+		length -= len;
+	}
+	return ndesc;
+}
+
+/*
+ * Copy from the SGEs to the data buffer.
+ */
+static void qib_copy_from_sge(void *data, struct qib_sge_state *ss, u32 length)
+{
+	struct qib_sge *sge = &ss->sge;
+
+	while (length) {
+		u32 len = sge->length;
+
+		if (len > length)
+			len = length;
+		if (len > sge->sge_length)
+			len = sge->sge_length;
+		BUG_ON(len == 0);
+		memcpy(data, sge->vaddr, len);
+		sge->vaddr += len;
+		sge->length -= len;
+		sge->sge_length -= len;
+		if (sge->sge_length == 0) {
+			if (--ss->num_sge)
+				*sge = *ss->sg_list++;
+		} else if (sge->length == 0 && sge->mr->lkey) {
+			if (++sge->n >= QIB_SEGSZ) {
+				if (++sge->m >= sge->mr->mapsz)
+					break;
+				sge->n = 0;
+			}
+			sge->vaddr =
+				sge->mr->map[sge->m]->segs[sge->n].vaddr;
+			sge->length =
+				sge->mr->map[sge->m]->segs[sge->n].length;
+		}
+		data += len;
+		length -= len;
+	}
+}
+
+/**
+ * qib_post_one_send - post one RC, UC, or UD send work request
+ * @qp: the QP to post on
+ * @wr: the work request to send
+ */
+static int qib_post_one_send(struct qib_qp *qp, struct ib_send_wr *wr)
+{
+	struct qib_swqe *wqe;
+	u32 next;
+	int i;
+	int j;
+	int acc;
+	int ret;
+	unsigned long flags;
+	struct qib_lkey_table *rkt;
+	struct qib_pd *pd;
+
+	spin_lock_irqsave(&qp->s_lock, flags);
+
+	/* Check that state is OK to post send. */
+	if (unlikely(!(ib_qib_state_ops[qp->state] & QIB_POST_SEND_OK)))
+		goto bail_inval;
+
+	/* IB spec says that num_sge == 0 is OK. */
+	if (wr->num_sge > qp->s_max_sge)
+		goto bail_inval;
+
+	/*
+	 * Don't allow RDMA reads or atomic operations on UC or
+	 * undefined operations.
+	 * Make sure buffer is large enough to hold the result for atomics.
+	 */
+	if (wr->opcode == IB_WR_FAST_REG_MR) {
+		if (qib_fast_reg_mr(qp, wr))
+			goto bail_inval;
+	} else if (qp->ibqp.qp_type == IB_QPT_UC) {
+		if ((unsigned) wr->opcode >= IB_WR_RDMA_READ)
+			goto bail_inval;
+	} else if (qp->ibqp.qp_type != IB_QPT_RC) {
+		/* Check IB_QPT_SMI, IB_QPT_GSI, IB_QPT_UD opcode */
+		if (wr->opcode != IB_WR_SEND &&
+		    wr->opcode != IB_WR_SEND_WITH_IMM)
+			goto bail_inval;
+		/* Check UD destination address PD */
+		if (qp->ibqp.pd != wr->wr.ud.ah->pd)
+			goto bail_inval;
+	} else if ((unsigned) wr->opcode > IB_WR_ATOMIC_FETCH_AND_ADD)
+		goto bail_inval;
+	else if (wr->opcode >= IB_WR_ATOMIC_CMP_AND_SWP &&
+		   (wr->num_sge == 0 ||
+		    wr->sg_list[0].length < sizeof(u64) ||
+		    wr->sg_list[0].addr & (sizeof(u64) - 1)))
+		goto bail_inval;
+	else if (wr->opcode >= IB_WR_RDMA_READ && !qp->s_max_rd_atomic)
+		goto bail_inval;
+
+	next = qp->s_head + 1;
+	if (next >= qp->s_size)
+		next = 0;
+	if (next == qp->s_last) {
+		ret = -ENOMEM;
+		goto bail;
+	}
+
+	rkt = &to_idev(qp->ibqp.device)->lk_table;
+	pd = to_ipd(qp->ibqp.pd);
+	wqe = get_swqe_ptr(qp, qp->s_head);
+	wqe->wr = *wr;
+	wqe->length = 0;
+	j = 0;
+	if (wr->num_sge) {
+		acc = wr->opcode >= IB_WR_RDMA_READ ?
+			IB_ACCESS_LOCAL_WRITE : 0;
+		for (i = 0; i < wr->num_sge; i++) {
+			u32 length = wr->sg_list[i].length;
+			int ok;
+
+			if (length == 0)
+				continue;
+			ok = qib_lkey_ok(rkt, pd, &wqe->sg_list[j],
+					 &wr->sg_list[i], acc);
+			if (!ok)
+				goto bail_inval_free;
+			wqe->length += length;
+			j++;
+		}
+		wqe->wr.num_sge = j;
+	}
+	if (qp->ibqp.qp_type == IB_QPT_UC ||
+	    qp->ibqp.qp_type == IB_QPT_RC) {
+		if (wqe->length > 0x80000000U)
+			goto bail_inval_free;
+	} else if (wqe->length > (dd_from_ibdev(qp->ibqp.device)->pport +
+				  qp->port_num - 1)->ibmtu)
+		goto bail_inval_free;
+	else
+		atomic_inc(&to_iah(wr->wr.ud.ah)->refcount);
+	wqe->ssn = qp->s_ssn++;
+	qp->s_head = next;
+
+	ret = 0;
+	goto bail;
+
+bail_inval_free:
+	while (j) {
+		struct qib_sge *sge = &wqe->sg_list[--j];
+
+		atomic_dec(&sge->mr->refcount);
+	}
+bail_inval:
+	ret = -EINVAL;
+bail:
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+	return ret;
+}
+
+/**
+ * qib_post_send - post a send on a QP
+ * @ibqp: the QP to post the send on
+ * @wr: the list of work requests to post
+ * @bad_wr: the first bad WR is put here
+ *
+ * This may be called from interrupt context.
+ */
+static int qib_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
+			 struct ib_send_wr **bad_wr)
+{
+	struct qib_qp *qp = to_iqp(ibqp);
+	int err = 0;
+
+	for (; wr; wr = wr->next) {
+		err = qib_post_one_send(qp, wr);
+		if (err) {
+			*bad_wr = wr;
+			goto bail;
+		}
+	}
+
+	/* Try to do the send work in the caller's context. */
+	qib_do_send(&qp->s_work);
+
+bail:
+	return err;
+}
+
+/**
+ * qib_post_receive - post a receive on a QP
+ * @ibqp: the QP to post the receive on
+ * @wr: the WR to post
+ * @bad_wr: the first bad WR is put here
+ *
+ * This may be called from interrupt context.
+ */
+static int qib_post_receive(struct ib_qp *ibqp, struct ib_recv_wr *wr,
+			    struct ib_recv_wr **bad_wr)
+{
+	struct qib_qp *qp = to_iqp(ibqp);
+	struct qib_rwq *wq = qp->r_rq.wq;
+	unsigned long flags;
+	int ret;
+
+	/* Check that state is OK to post receive. */
+	if (!(ib_qib_state_ops[qp->state] & QIB_POST_RECV_OK) || !wq) {
+		*bad_wr = wr;
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	for (; wr; wr = wr->next) {
+		struct qib_rwqe *wqe;
+		u32 next;
+		int i;
+
+		if ((unsigned) wr->num_sge > qp->r_rq.max_sge) {
+			*bad_wr = wr;
+			ret = -EINVAL;
+			goto bail;
+		}
+
+		spin_lock_irqsave(&qp->r_rq.lock, flags);
+		next = wq->head + 1;
+		if (next >= qp->r_rq.size)
+			next = 0;
+		if (next == wq->tail) {
+			spin_unlock_irqrestore(&qp->r_rq.lock, flags);
+			*bad_wr = wr;
+			ret = -ENOMEM;
+			goto bail;
+		}
+
+		wqe = get_rwqe_ptr(&qp->r_rq, wq->head);
+		wqe->wr_id = wr->wr_id;
+		wqe->num_sge = wr->num_sge;
+		for (i = 0; i < wr->num_sge; i++)
+			wqe->sg_list[i] = wr->sg_list[i];
+		/* Make sure queue entry is written before the head index. */
+		smp_wmb();
+		wq->head = next;
+		spin_unlock_irqrestore(&qp->r_rq.lock, flags);
+	}
+	ret = 0;
+
+bail:
+	return ret;
+}
+
+/**
+ * qib_qp_rcv - processing an incoming packet on a QP
+ * @rcd: the context pointer
+ * @hdr: the packet header
+ * @has_grh: true if the packet has a GRH
+ * @data: the packet data
+ * @tlen: the packet length
+ * @qp: the QP the packet came on
+ *
+ * This is called from qib_ib_rcv() to process an incoming packet
+ * for the given QP.
+ * Called at interrupt level.
+ */
+static void qib_qp_rcv(struct qib_ctxtdata *rcd, struct qib_ib_header *hdr,
+		       int has_grh, void *data, u32 tlen, struct qib_qp *qp)
+{
+	struct qib_ibport *ibp = &rcd->ppd->ibport_data;
+
+	/* Check for valid receive state. */
+	if (!(ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK)) {
+		ibp->n_pkt_drops++;
+		return;
+	}
+
+	switch (qp->ibqp.qp_type) {
+	case IB_QPT_SMI:
+	case IB_QPT_GSI:
+		if (ib_qib_disable_sma)
+			break;
+		/* FALLTHROUGH */
+	case IB_QPT_UD:
+		qib_ud_rcv(ibp, hdr, has_grh, data, tlen, qp);
+		break;
+
+	case IB_QPT_RC:
+		qib_rc_rcv(rcd, hdr, has_grh, data, tlen, qp);
+		break;
+
+	case IB_QPT_UC:
+		qib_uc_rcv(ibp, hdr, has_grh, data, tlen, qp);
+		break;
+
+	default:
+		break;
+	}
+}
+
+/**
+ * qib_ib_rcv - process an incoming packet
+ * @rcd: the context pointer
+ * @rhdr: the header of the packet
+ * @data: the packet payload
+ * @tlen: the packet length
+ *
+ * This is called from qib_kreceive() to process an incoming packet at
+ * interrupt level. Tlen is the length of the header + data + CRC in bytes.
+ */
+void qib_ib_rcv(struct qib_ctxtdata *rcd, void *rhdr, void *data, u32 tlen)
+{
+	struct qib_pportdata *ppd = rcd->ppd;
+	struct qib_ibport *ibp = &ppd->ibport_data;
+	struct qib_ib_header *hdr = rhdr;
+	struct qib_other_headers *ohdr;
+	struct qib_qp *qp;
+	u32 qp_num;
+	int lnh;
+	u8 opcode;
+	u16 lid;
+
+	/* 24 == LRH+BTH+CRC */
+	if (unlikely(tlen < 24))
+		goto drop;
+
+	/* Check for a valid destination LID (see ch. 7.11.1). */
+	lid = be16_to_cpu(hdr->lrh[1]);
+	if (lid < QIB_MULTICAST_LID_BASE) {
+		lid &= ~((1 << ppd->lmc) - 1);
+		if (unlikely(lid != ppd->lid))
+			goto drop;
+	}
+
+	/* Check for GRH */
+	lnh = be16_to_cpu(hdr->lrh[0]) & 3;
+	if (lnh == QIB_LRH_BTH)
+		ohdr = &hdr->u.oth;
+	else if (lnh == QIB_LRH_GRH) {
+		u32 vtf;
+
+		ohdr = &hdr->u.l.oth;
+		if (hdr->u.l.grh.next_hdr != IB_GRH_NEXT_HDR)
+			goto drop;
+		vtf = be32_to_cpu(hdr->u.l.grh.version_tclass_flow);
+		if ((vtf >> IB_GRH_VERSION_SHIFT) != IB_GRH_VERSION)
+			goto drop;
+	} else
+		goto drop;
+
+	opcode = be32_to_cpu(ohdr->bth[0]) >> 24;
+	ibp->opstats[opcode & 0x7f].n_bytes += tlen;
+	ibp->opstats[opcode & 0x7f].n_packets++;
+
+	/* Get the destination QP number. */
+	qp_num = be32_to_cpu(ohdr->bth[1]) & QIB_QPN_MASK;
+	if (qp_num == QIB_MULTICAST_QPN) {
+		struct qib_mcast *mcast;
+		struct qib_mcast_qp *p;
+
+		if (lnh != QIB_LRH_GRH)
+			goto drop;
+		mcast = qib_mcast_find(ibp, &hdr->u.l.grh.dgid);
+		if (mcast == NULL)
+			goto drop;
+		ibp->n_multicast_rcv++;
+		list_for_each_entry_rcu(p, &mcast->qp_list, list)
+			qib_qp_rcv(rcd, hdr, 1, data, tlen, p->qp);
+		/*
+		 * Notify qib_multicast_detach() if it is waiting for us
+		 * to finish.
+		 */
+		if (atomic_dec_return(&mcast->refcount) <= 1)
+			wake_up(&mcast->wait);
+	} else {
+		qp = qib_lookup_qpn(ibp, qp_num);
+		if (!qp)
+			goto drop;
+		ibp->n_unicast_rcv++;
+		qib_qp_rcv(rcd, hdr, lnh == QIB_LRH_GRH, data, tlen, qp);
+		/*
+		 * Notify qib_destroy_qp() if it is waiting
+		 * for us to finish.
+		 */
+		if (atomic_dec_and_test(&qp->refcount))
+			wake_up(&qp->wait);
+	}
+	return;
+
+drop:
+	ibp->n_pkt_drops++;
+}
+
+/*
+ * This is called from a timer to check for QPs
+ * which need kernel memory in order to send a packet.
+ */
+static void mem_timer(unsigned long data)
+{
+	struct qib_ibdev *dev = (struct qib_ibdev *) data;
+	struct list_head *list = &dev->memwait;
+	struct qib_qp *qp = NULL;
+	unsigned long flags;
+
+	spin_lock_irqsave(&dev->pending_lock, flags);
+	if (!list_empty(list)) {
+		qp = list_entry(list->next, struct qib_qp, iowait);
+		list_del_init(&qp->iowait);
+		atomic_inc(&qp->refcount);
+		if (!list_empty(list))
+			mod_timer(&dev->mem_timer, jiffies + 1);
+	}
+	spin_unlock_irqrestore(&dev->pending_lock, flags);
+
+	if (qp) {
+		spin_lock_irqsave(&qp->s_lock, flags);
+		if (qp->s_flags & QIB_S_WAIT_KMEM) {
+			qp->s_flags &= ~QIB_S_WAIT_KMEM;
+			qib_schedule_send(qp);
+		}
+		spin_unlock_irqrestore(&qp->s_lock, flags);
+		if (atomic_dec_and_test(&qp->refcount))
+			wake_up(&qp->wait);
+	}
+}
+
+static void update_sge(struct qib_sge_state *ss, u32 length)
+{
+	struct qib_sge *sge = &ss->sge;
+
+	sge->vaddr += length;
+	sge->length -= length;
+	sge->sge_length -= length;
+	if (sge->sge_length == 0) {
+		if (--ss->num_sge)
+			*sge = *ss->sg_list++;
+	} else if (sge->length == 0 && sge->mr->lkey) {
+		if (++sge->n >= QIB_SEGSZ) {
+			if (++sge->m >= sge->mr->mapsz)
+				return;
+			sge->n = 0;
+		}
+		sge->vaddr = sge->mr->map[sge->m]->segs[sge->n].vaddr;
+		sge->length = sge->mr->map[sge->m]->segs[sge->n].length;
+	}
+}
+
+#ifdef __LITTLE_ENDIAN
+static inline u32 get_upper_bits(u32 data, u32 shift)
+{
+	return data >> shift;
+}
+
+static inline u32 set_upper_bits(u32 data, u32 shift)
+{
+	return data << shift;
+}
+
+static inline u32 clear_upper_bytes(u32 data, u32 n, u32 off)
+{
+	data <<= ((sizeof(u32) - n) * BITS_PER_BYTE);
+	data >>= ((sizeof(u32) - n - off) * BITS_PER_BYTE);
+	return data;
+}
+#else
+static inline u32 get_upper_bits(u32 data, u32 shift)
+{
+	return data << shift;
+}
+
+static inline u32 set_upper_bits(u32 data, u32 shift)
+{
+	return data >> shift;
+}
+
+static inline u32 clear_upper_bytes(u32 data, u32 n, u32 off)
+{
+	data >>= ((sizeof(u32) - n) * BITS_PER_BYTE);
+	data <<= ((sizeof(u32) - n - off) * BITS_PER_BYTE);
+	return data;
+}
+#endif
+
+static void copy_io(u32 __iomem *piobuf, struct qib_sge_state *ss,
+		    u32 length, unsigned flush_wc)
+{
+	u32 extra = 0;
+	u32 data = 0;
+	u32 last;
+
+	while (1) {
+		u32 len = ss->sge.length;
+		u32 off;
+
+		if (len > length)
+			len = length;
+		if (len > ss->sge.sge_length)
+			len = ss->sge.sge_length;
+		BUG_ON(len == 0);
+		/* If the source address is not aligned, try to align it. */
+		off = (unsigned long)ss->sge.vaddr & (sizeof(u32) - 1);
+		if (off) {
+			u32 *addr = (u32 *)((unsigned long)ss->sge.vaddr &
+					    ~(sizeof(u32) - 1));
+			u32 v = get_upper_bits(*addr, off * BITS_PER_BYTE);
+			u32 y;
+
+			y = sizeof(u32) - off;
+			if (len > y)
+				len = y;
+			if (len + extra >= sizeof(u32)) {
+				data |= set_upper_bits(v, extra *
+						       BITS_PER_BYTE);
+				len = sizeof(u32) - extra;
+				if (len == length) {
+					last = data;
+					break;
+				}
+				__raw_writel(data, piobuf);
+				piobuf++;
+				extra = 0;
+				data = 0;
+			} else {
+				/* Clear unused upper bytes */
+				data |= clear_upper_bytes(v, len, extra);
+				if (len == length) {
+					last = data;
+					break;
+				}
+				extra += len;
+			}
+		} else if (extra) {
+			/* Source address is aligned. */
+			u32 *addr = (u32 *) ss->sge.vaddr;
+			int shift = extra * BITS_PER_BYTE;
+			int ushift = 32 - shift;
+			u32 l = len;
+
+			while (l >= sizeof(u32)) {
+				u32 v = *addr;
+
+				data |= set_upper_bits(v, shift);
+				__raw_writel(data, piobuf);
+				data = get_upper_bits(v, ushift);
+				piobuf++;
+				addr++;
+				l -= sizeof(u32);
+			}
+			/*
+			 * We still have 'extra' number of bytes leftover.
+			 */
+			if (l) {
+				u32 v = *addr;
+
+				if (l + extra >= sizeof(u32)) {
+					data |= set_upper_bits(v, shift);
+					len -= l + extra - sizeof(u32);
+					if (len == length) {
+						last = data;
+						break;
+					}
+					__raw_writel(data, piobuf);
+					piobuf++;
+					extra = 0;
+					data = 0;
+				} else {
+					/* Clear unused upper bytes */
+					data |= clear_upper_bytes(v, l, extra);
+					if (len == length) {
+						last = data;
+						break;
+					}
+					extra += l;
+				}
+			} else if (len == length) {
+				last = data;
+				break;
+			}
+		} else if (len == length) {
+			u32 w;
+
+			/*
+			 * Need to round up for the last dword in the
+			 * packet.
+			 */
+			w = (len + 3) >> 2;
+			qib_pio_copy(piobuf, ss->sge.vaddr, w - 1);
+			piobuf += w - 1;
+			last = ((u32 *) ss->sge.vaddr)[w - 1];
+			break;
+		} else {
+			u32 w = len >> 2;
+
+			qib_pio_copy(piobuf, ss->sge.vaddr, w);
+			piobuf += w;
+
+			extra = len & (sizeof(u32) - 1);
+			if (extra) {
+				u32 v = ((u32 *) ss->sge.vaddr)[w];
+
+				/* Clear unused upper bytes */
+				data = clear_upper_bytes(v, extra, 0);
+			}
+		}
+		update_sge(ss, len);
+		length -= len;
+	}
+	/* Update address before sending packet. */
+	update_sge(ss, length);
+	if (flush_wc) {
+		/* must flush early everything before trigger word */
+		qib_flush_wc();
+		__raw_writel(last, piobuf);
+		/* be sure trigger word is written */
+		qib_flush_wc();
+	} else
+		__raw_writel(last, piobuf);
+}
+
+static struct qib_verbs_txreq *get_txreq(struct qib_ibdev *dev,
+					 struct qib_qp *qp, int *retp)
+{
+	struct qib_verbs_txreq *tx;
+	unsigned long flags;
+
+	spin_lock_irqsave(&qp->s_lock, flags);
+	spin_lock(&dev->pending_lock);
+
+	if (!list_empty(&dev->txreq_free)) {
+		struct list_head *l = dev->txreq_free.next;
+
+		list_del(l);
+		tx = list_entry(l, struct qib_verbs_txreq, txreq.list);
+		*retp = 0;
+	} else {
+		if (ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK &&
+		    list_empty(&qp->iowait)) {
+			dev->n_txwait++;
+			qp->s_flags |= QIB_S_WAIT_TX;
+			list_add_tail(&qp->iowait, &dev->txwait);
+		}
+		tx = NULL;
+		qp->s_flags &= ~QIB_S_BUSY;
+		*retp = -EBUSY;
+	}
+
+	spin_unlock(&dev->pending_lock);
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+
+	return tx;
+}
+
+void qib_put_txreq(struct qib_verbs_txreq *tx)
+{
+	struct qib_ibdev *dev;
+	struct qib_qp *qp;
+	unsigned long flags;
+
+	qp = tx->qp;
+	dev = to_idev(qp->ibqp.device);
+
+	if (atomic_dec_and_test(&qp->refcount))
+		wake_up(&qp->wait);
+	if (tx->mr) {
+		atomic_dec(&tx->mr->refcount);
+		tx->mr = NULL;
+	}
+	if (tx->txreq.flags & QIB_SDMA_TXREQ_F_FREEBUF) {
+		tx->txreq.flags &= ~QIB_SDMA_TXREQ_F_FREEBUF;
+		dma_unmap_single(&dd_from_dev(dev)->pcidev->dev,
+				 tx->txreq.addr, tx->hdr_dwords << 2,
+				 DMA_TO_DEVICE);
+		kfree(tx->align_buf);
+	}
+
+	spin_lock_irqsave(&dev->pending_lock, flags);
+
+	/* Put struct back on free list */
+	list_add(&tx->txreq.list, &dev->txreq_free);
+
+	if (!list_empty(&dev->txwait)) {
+		/* Wake up first QP wanting a free struct */
+		qp = list_entry(dev->txwait.next, struct qib_qp, iowait);
+		list_del_init(&qp->iowait);
+		atomic_inc(&qp->refcount);
+		spin_unlock_irqrestore(&dev->pending_lock, flags);
+
+		spin_lock_irqsave(&qp->s_lock, flags);
+		if (qp->s_flags & QIB_S_WAIT_TX) {
+			qp->s_flags &= ~QIB_S_WAIT_TX;
+			qib_schedule_send(qp);
+		}
+		spin_unlock_irqrestore(&qp->s_lock, flags);
+
+		if (atomic_dec_and_test(&qp->refcount))
+			wake_up(&qp->wait);
+	} else
+		spin_unlock_irqrestore(&dev->pending_lock, flags);
+}
+
+/*
+ * This is called when there are send DMA descriptors that might be
+ * available.
+ *
+ * This is called with ppd->sdma_lock held.
+ */
+void qib_verbs_sdma_desc_avail(struct qib_pportdata *ppd, unsigned avail)
+{
+	struct qib_qp *qp, *nqp;
+	struct qib_qp *qps[20];
+	struct qib_ibdev *dev;
+	unsigned i, n;
+
+	n = 0;
+	dev = &ppd->dd->verbs_dev;
+	spin_lock(&dev->pending_lock);
+
+	/* Search wait list for first QP wanting DMA descriptors. */
+	list_for_each_entry_safe(qp, nqp, &dev->dmawait, iowait) {
+		if (qp->port_num != ppd->port)
+			continue;
+		if (n == ARRAY_SIZE(qps))
+			break;
+		if (qp->s_tx->txreq.sg_count > avail)
+			break;
+		avail -= qp->s_tx->txreq.sg_count;
+		list_del_init(&qp->iowait);
+		atomic_inc(&qp->refcount);
+		qps[n++] = qp;
+	}
+
+	spin_unlock(&dev->pending_lock);
+
+	for (i = 0; i < n; i++) {
+		qp = qps[i];
+		spin_lock(&qp->s_lock);
+		if (qp->s_flags & QIB_S_WAIT_DMA_DESC) {
+			qp->s_flags &= ~QIB_S_WAIT_DMA_DESC;
+			qib_schedule_send(qp);
+		}
+		spin_unlock(&qp->s_lock);
+		if (atomic_dec_and_test(&qp->refcount))
+			wake_up(&qp->wait);
+	}
+}
+
+/*
+ * This is called with ppd->sdma_lock held.
+ */
+static void sdma_complete(struct qib_sdma_txreq *cookie, int status)
+{
+	struct qib_verbs_txreq *tx =
+		container_of(cookie, struct qib_verbs_txreq, txreq);
+	struct qib_qp *qp = tx->qp;
+
+	spin_lock(&qp->s_lock);
+	if (tx->wqe)
+		qib_send_complete(qp, tx->wqe, IB_WC_SUCCESS);
+	else if (qp->ibqp.qp_type == IB_QPT_RC) {
+		struct qib_ib_header *hdr;
+
+		if (tx->txreq.flags & QIB_SDMA_TXREQ_F_FREEBUF)
+			hdr = &tx->align_buf->hdr;
+		else {
+			struct qib_ibdev *dev = to_idev(qp->ibqp.device);
+
+			hdr = &dev->pio_hdrs[tx->hdr_inx].hdr;
+		}
+		qib_rc_send_complete(qp, hdr);
+	}
+	if (atomic_dec_and_test(&qp->s_dma_busy)) {
+		if (qp->state == IB_QPS_RESET)
+			wake_up(&qp->wait_dma);
+		else if (qp->s_flags & QIB_S_WAIT_DMA) {
+			qp->s_flags &= ~QIB_S_WAIT_DMA;
+			qib_schedule_send(qp);
+		}
+	}
+	spin_unlock(&qp->s_lock);
+
+	qib_put_txreq(tx);
+}
+
+static int wait_kmem(struct qib_ibdev *dev, struct qib_qp *qp)
+{
+	unsigned long flags;
+	int ret = 0;
+
+	spin_lock_irqsave(&qp->s_lock, flags);
+	if (ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK) {
+		spin_lock(&dev->pending_lock);
+		if (list_empty(&qp->iowait)) {
+			if (list_empty(&dev->memwait))
+				mod_timer(&dev->mem_timer, jiffies + 1);
+			qp->s_flags |= QIB_S_WAIT_KMEM;
+			list_add_tail(&qp->iowait, &dev->memwait);
+		}
+		spin_unlock(&dev->pending_lock);
+		qp->s_flags &= ~QIB_S_BUSY;
+		ret = -EBUSY;
+	}
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+
+	return ret;
+}
+
+static int qib_verbs_send_dma(struct qib_qp *qp, struct qib_ib_header *hdr,
+			      u32 hdrwords, struct qib_sge_state *ss, u32 len,
+			      u32 plen, u32 dwords)
+{
+	struct qib_ibdev *dev = to_idev(qp->ibqp.device);
+	struct qib_devdata *dd = dd_from_dev(dev);
+	struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	struct qib_verbs_txreq *tx;
+	struct qib_pio_header *phdr;
+	u32 control;
+	u32 ndesc;
+	int ret;
+
+	tx = qp->s_tx;
+	if (tx) {
+		qp->s_tx = NULL;
+		/* resend previously constructed packet */
+		ret = qib_sdma_verbs_send(ppd, tx->ss, tx->dwords, tx);
+		goto bail;
+	}
+
+	tx = get_txreq(dev, qp, &ret);
+	if (!tx)
+		goto bail;
+
+	control = dd->f_setpbc_control(ppd, plen, qp->s_srate,
+				       be16_to_cpu(hdr->lrh[0]) >> 12);
+	tx->qp = qp;
+	atomic_inc(&qp->refcount);
+	tx->wqe = qp->s_wqe;
+	tx->mr = qp->s_rdma_mr;
+	if (qp->s_rdma_mr)
+		qp->s_rdma_mr = NULL;
+	tx->txreq.callback = sdma_complete;
+	if (dd->flags & QIB_HAS_SDMA_TIMEOUT)
+		tx->txreq.flags = QIB_SDMA_TXREQ_F_HEADTOHOST;
+	else
+		tx->txreq.flags = QIB_SDMA_TXREQ_F_INTREQ;
+	if (plen + 1 > dd->piosize2kmax_dwords)
+		tx->txreq.flags |= QIB_SDMA_TXREQ_F_USELARGEBUF;
+
+	if (len) {
+		/*
+		 * Don't try to DMA if it takes more descriptors than
+		 * the queue holds.
+		 */
+		ndesc = qib_count_sge(ss, len);
+		if (ndesc >= ppd->sdma_descq_cnt)
+			ndesc = 0;
+	} else
+		ndesc = 1;
+	if (ndesc) {
+		phdr = &dev->pio_hdrs[tx->hdr_inx];
+		phdr->pbc[0] = cpu_to_le32(plen);
+		phdr->pbc[1] = cpu_to_le32(control);
+		memcpy(&phdr->hdr, hdr, hdrwords << 2);
+		tx->txreq.flags |= QIB_SDMA_TXREQ_F_FREEDESC;
+		tx->txreq.sg_count = ndesc;
+		tx->txreq.addr = dev->pio_hdrs_phys +
+			tx->hdr_inx * sizeof(struct qib_pio_header);
+		tx->hdr_dwords = hdrwords + 2; /* add PBC length */
+		ret = qib_sdma_verbs_send(ppd, ss, dwords, tx);
+		goto bail;
+	}
+
+	/* Allocate a buffer and copy the header and payload to it. */
+	tx->hdr_dwords = plen + 1;
+	phdr = kmalloc(tx->hdr_dwords << 2, GFP_ATOMIC);
+	if (!phdr)
+		goto err_tx;
+	phdr->pbc[0] = cpu_to_le32(plen);
+	phdr->pbc[1] = cpu_to_le32(control);
+	memcpy(&phdr->hdr, hdr, hdrwords << 2);
+	qib_copy_from_sge((u32 *) &phdr->hdr + hdrwords, ss, len);
+
+	tx->txreq.addr = dma_map_single(&dd->pcidev->dev, phdr,
+					tx->hdr_dwords << 2, DMA_TO_DEVICE);
+	if (dma_mapping_error(&dd->pcidev->dev, tx->txreq.addr))
+		goto map_err;
+	tx->align_buf = phdr;
+	tx->txreq.flags |= QIB_SDMA_TXREQ_F_FREEBUF;
+	tx->txreq.sg_count = 1;
+	ret = qib_sdma_verbs_send(ppd, NULL, 0, tx);
+	goto unaligned;
+
+map_err:
+	kfree(phdr);
+err_tx:
+	qib_put_txreq(tx);
+	ret = wait_kmem(dev, qp);
+unaligned:
+	ibp->n_unaligned++;
+bail:
+	return ret;
+}
+
+/*
+ * If we are now in the error state, return zero to flush the
+ * send work request.
+ */
+static int no_bufs_available(struct qib_qp *qp)
+{
+	struct qib_ibdev *dev = to_idev(qp->ibqp.device);
+	struct qib_devdata *dd;
+	unsigned long flags;
+	int ret = 0;
+
+	/*
+	 * Note that as soon as want_buffer() is called and
+	 * possibly before it returns, qib_ib_piobufavail()
+	 * could be called. Therefore, put QP on the I/O wait list before
+	 * enabling the PIO avail interrupt.
+	 */
+	spin_lock_irqsave(&qp->s_lock, flags);
+	if (ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK) {
+		spin_lock(&dev->pending_lock);
+		if (list_empty(&qp->iowait)) {
+			dev->n_piowait++;
+			qp->s_flags |= QIB_S_WAIT_PIO;
+			list_add_tail(&qp->iowait, &dev->piowait);
+			dd = dd_from_dev(dev);
+			dd->f_wantpiobuf_intr(dd, 1);
+		}
+		spin_unlock(&dev->pending_lock);
+		qp->s_flags &= ~QIB_S_BUSY;
+		ret = -EBUSY;
+	}
+	spin_unlock_irqrestore(&qp->s_lock, flags);
+	return ret;
+}
+
+static int qib_verbs_send_pio(struct qib_qp *qp, struct qib_ib_header *ibhdr,
+			      u32 hdrwords, struct qib_sge_state *ss, u32 len,
+			      u32 plen, u32 dwords)
+{
+	struct qib_devdata *dd = dd_from_ibdev(qp->ibqp.device);
+	struct qib_pportdata *ppd = dd->pport + qp->port_num - 1;
+	u32 *hdr = (u32 *) ibhdr;
+	u32 __iomem *piobuf_orig;
+	u32 __iomem *piobuf;
+	u64 pbc;
+	unsigned long flags;
+	unsigned flush_wc;
+	u32 control;
+	u32 pbufn;
+
+	control = dd->f_setpbc_control(ppd, plen, qp->s_srate,
+		be16_to_cpu(ibhdr->lrh[0]) >> 12);
+	pbc = ((u64) control << 32) | plen;
+	piobuf = dd->f_getsendbuf(ppd, pbc, &pbufn);
+	if (unlikely(piobuf == NULL))
+		return no_bufs_available(qp);
+
+	/*
+	 * Write the pbc.
+	 * We have to flush after the PBC for correctness on some cpus
+	 * or WC buffer can be written out of order.
+	 */
+	writeq(pbc, piobuf);
+	piobuf_orig = piobuf;
+	piobuf += 2;
+
+	flush_wc = dd->flags & QIB_PIO_FLUSH_WC;
+	if (len == 0) {
+		/*
+		 * If there is just the header portion, must flush before
+		 * writing last word of header for correctness, and after
+		 * the last header word (trigger word).
+		 */
+		if (flush_wc) {
+			qib_flush_wc();
+			qib_pio_copy(piobuf, hdr, hdrwords - 1);
+			qib_flush_wc();
+			__raw_writel(hdr[hdrwords - 1], piobuf + hdrwords - 1);
+			qib_flush_wc();
+		} else
+			qib_pio_copy(piobuf, hdr, hdrwords);
+		goto done;
+	}
+
+	if (flush_wc)
+		qib_flush_wc();
+	qib_pio_copy(piobuf, hdr, hdrwords);
+	piobuf += hdrwords;
+
+	/* The common case is aligned and contained in one segment. */
+	if (likely(ss->num_sge == 1 && len <= ss->sge.length &&
+		   !((unsigned long)ss->sge.vaddr & (sizeof(u32) - 1)))) {
+		u32 *addr = (u32 *) ss->sge.vaddr;
+
+		/* Update address before sending packet. */
+		update_sge(ss, len);
+		if (flush_wc) {
+			qib_pio_copy(piobuf, addr, dwords - 1);
+			/* must flush early everything before trigger word */
+			qib_flush_wc();
+			__raw_writel(addr[dwords - 1], piobuf + dwords - 1);
+			/* be sure trigger word is written */
+			qib_flush_wc();
+		} else
+			qib_pio_copy(piobuf, addr, dwords);
+		goto done;
+	}
+	copy_io(piobuf, ss, len, flush_wc);
+done:
+	if (dd->flags & QIB_USE_SPCL_TRIG) {
+		u32 spcl_off = (pbufn >= dd->piobcnt2k) ? 2047 : 1023;
+		qib_flush_wc();
+		__raw_writel(0xaebecede, piobuf_orig + spcl_off);
+	}
+	qib_sendbuf_done(dd, pbufn);
+	if (qp->s_rdma_mr) {
+		atomic_dec(&qp->s_rdma_mr->refcount);
+		qp->s_rdma_mr = NULL;
+	}
+	if (qp->s_wqe) {
+		spin_lock_irqsave(&qp->s_lock, flags);
+		qib_send_complete(qp, qp->s_wqe, IB_WC_SUCCESS);
+		spin_unlock_irqrestore(&qp->s_lock, flags);
+	} else if (qp->ibqp.qp_type == IB_QPT_RC) {
+		spin_lock_irqsave(&qp->s_lock, flags);
+		qib_rc_send_complete(qp, ibhdr);
+		spin_unlock_irqrestore(&qp->s_lock, flags);
+	}
+	return 0;
+}
+
+/**
+ * qib_verbs_send - send a packet
+ * @qp: the QP to send on
+ * @hdr: the packet header
+ * @hdrwords: the number of 32-bit words in the header
+ * @ss: the SGE to send
+ * @len: the length of the packet in bytes
+ *
+ * Return zero if packet is sent or queued OK.
+ * Return non-zero and clear qp->s_flags QIB_S_BUSY otherwise.
+ */
+int qib_verbs_send(struct qib_qp *qp, struct qib_ib_header *hdr,
+		   u32 hdrwords, struct qib_sge_state *ss, u32 len)
+{
+	struct qib_devdata *dd = dd_from_ibdev(qp->ibqp.device);
+	u32 plen;
+	int ret;
+	u32 dwords = (len + 3) >> 2;
+
+	/*
+	 * Calculate the send buffer trigger address.
+	 * The +1 counts for the pbc control dword following the pbc length.
+	 */
+	plen = hdrwords + dwords + 1;
+
+	/*
+	 * VL15 packets (IB_QPT_SMI) will always use PIO, so we
+	 * can defer SDMA restart until link goes ACTIVE without
+	 * worrying about just how we got there.
+	 */
+	if (qp->ibqp.qp_type == IB_QPT_SMI ||
+	    !(dd->flags & QIB_HAS_SEND_DMA))
+		ret = qib_verbs_send_pio(qp, hdr, hdrwords, ss, len,
+					 plen, dwords);
+	else
+		ret = qib_verbs_send_dma(qp, hdr, hdrwords, ss, len,
+					 plen, dwords);
+
+	return ret;
+}
+
+int qib_snapshot_counters(struct qib_pportdata *ppd, u64 *swords,
+			  u64 *rwords, u64 *spkts, u64 *rpkts,
+			  u64 *xmit_wait)
+{
+	int ret;
+	struct qib_devdata *dd = ppd->dd;
+
+	if (!(dd->flags & QIB_PRESENT)) {
+		/* no hardware, freeze, etc. */
+		ret = -EINVAL;
+		goto bail;
+	}
+	*swords = dd->f_portcntr(ppd, QIBPORTCNTR_WORDSEND);
+	*rwords = dd->f_portcntr(ppd, QIBPORTCNTR_WORDRCV);
+	*spkts = dd->f_portcntr(ppd, QIBPORTCNTR_PKTSEND);
+	*rpkts = dd->f_portcntr(ppd, QIBPORTCNTR_PKTRCV);
+	*xmit_wait = dd->f_portcntr(ppd, QIBPORTCNTR_SENDSTALL);
+
+	ret = 0;
+
+bail:
+	return ret;
+}
+
+/**
+ * qib_get_counters - get various chip counters
+ * @dd: the qlogic_ib device
+ * @cntrs: counters are placed here
+ *
+ * Return the counters needed by recv_pma_get_portcounters().
+ */
+int qib_get_counters(struct qib_pportdata *ppd,
+		     struct qib_verbs_counters *cntrs)
+{
+	int ret;
+
+	if (!(ppd->dd->flags & QIB_PRESENT)) {
+		/* no hardware, freeze, etc. */
+		ret = -EINVAL;
+		goto bail;
+	}
+	cntrs->symbol_error_counter =
+		ppd->dd->f_portcntr(ppd, QIBPORTCNTR_IBSYMBOLERR);
+	cntrs->link_error_recovery_counter =
+		ppd->dd->f_portcntr(ppd, QIBPORTCNTR_IBLINKERRRECOV);
+	/*
+	 * The link downed counter counts when the other side downs the
+	 * connection.  We add in the number of times we downed the link
+	 * due to local link integrity errors to compensate.
+	 */
+	cntrs->link_downed_counter =
+		ppd->dd->f_portcntr(ppd, QIBPORTCNTR_IBLINKDOWN);
+	cntrs->port_rcv_errors =
+		ppd->dd->f_portcntr(ppd, QIBPORTCNTR_RXDROPPKT) +
+		ppd->dd->f_portcntr(ppd, QIBPORTCNTR_RCVOVFL) +
+		ppd->dd->f_portcntr(ppd, QIBPORTCNTR_ERR_RLEN) +
+		ppd->dd->f_portcntr(ppd, QIBPORTCNTR_INVALIDRLEN) +
+		ppd->dd->f_portcntr(ppd, QIBPORTCNTR_ERRLINK) +
+		ppd->dd->f_portcntr(ppd, QIBPORTCNTR_ERRICRC) +
+		ppd->dd->f_portcntr(ppd, QIBPORTCNTR_ERRVCRC) +
+		ppd->dd->f_portcntr(ppd, QIBPORTCNTR_ERRLPCRC) +
+		ppd->dd->f_portcntr(ppd, QIBPORTCNTR_BADFORMAT);
+	cntrs->port_rcv_errors +=
+		ppd->dd->f_portcntr(ppd, QIBPORTCNTR_RXLOCALPHYERR);
+	cntrs->port_rcv_errors +=
+		ppd->dd->f_portcntr(ppd, QIBPORTCNTR_RXVLERR);
+	cntrs->port_rcv_remphys_errors =
+		ppd->dd->f_portcntr(ppd, QIBPORTCNTR_RCVEBP);
+	cntrs->port_xmit_discards =
+		ppd->dd->f_portcntr(ppd, QIBPORTCNTR_UNSUPVL);
+	cntrs->port_xmit_data = ppd->dd->f_portcntr(ppd,
+			QIBPORTCNTR_WORDSEND);
+	cntrs->port_rcv_data = ppd->dd->f_portcntr(ppd,
+			QIBPORTCNTR_WORDRCV);
+	cntrs->port_xmit_packets = ppd->dd->f_portcntr(ppd,
+			QIBPORTCNTR_PKTSEND);
+	cntrs->port_rcv_packets = ppd->dd->f_portcntr(ppd,
+			QIBPORTCNTR_PKTRCV);
+	cntrs->local_link_integrity_errors =
+		ppd->dd->f_portcntr(ppd, QIBPORTCNTR_LLI);
+	cntrs->excessive_buffer_overrun_errors =
+		ppd->dd->f_portcntr(ppd, QIBPORTCNTR_EXCESSBUFOVFL);
+	cntrs->vl15_dropped =
+		ppd->dd->f_portcntr(ppd, QIBPORTCNTR_VL15PKTDROP);
+
+	ret = 0;
+
+bail:
+	return ret;
+}
+
+/**
+ * qib_ib_piobufavail - callback when a PIO buffer is available
+ * @dd: the device pointer
+ *
+ * This is called from qib_intr() at interrupt level when a PIO buffer is
+ * available after qib_verbs_send() returned an error that no buffers were
+ * available. Disable the interrupt if there are no more QPs waiting.
+ */
+void qib_ib_piobufavail(struct qib_devdata *dd)
+{
+	struct qib_ibdev *dev = &dd->verbs_dev;
+	struct list_head *list;
+	struct qib_qp *qps[5];
+	struct qib_qp *qp;
+	unsigned long flags;
+	unsigned i, n;
+
+	list = &dev->piowait;
+	n = 0;
+
+	/*
+	 * Note: checking that the piowait list is empty and clearing
+	 * the buffer available interrupt needs to be atomic or we
+	 * could end up with QPs on the wait list with the interrupt
+	 * disabled.
+	 */
+	spin_lock_irqsave(&dev->pending_lock, flags);
+	while (!list_empty(list)) {
+		if (n == ARRAY_SIZE(qps))
+			goto full;
+		qp = list_entry(list->next, struct qib_qp, iowait);
+		list_del_init(&qp->iowait);
+		atomic_inc(&qp->refcount);
+		qps[n++] = qp;
+	}
+	dd->f_wantpiobuf_intr(dd, 0);
+full:
+	spin_unlock_irqrestore(&dev->pending_lock, flags);
+
+	for (i = 0; i < n; i++) {
+		qp = qps[i];
+
+		spin_lock_irqsave(&qp->s_lock, flags);
+		if (qp->s_flags & QIB_S_WAIT_PIO) {
+			qp->s_flags &= ~QIB_S_WAIT_PIO;
+			qib_schedule_send(qp);
+		}
+		spin_unlock_irqrestore(&qp->s_lock, flags);
+
+		/* Notify qib_destroy_qp() if it is waiting. */
+		if (atomic_dec_and_test(&qp->refcount))
+			wake_up(&qp->wait);
+	}
+}
+
+static int qib_query_device(struct ib_device *ibdev,
+			    struct ib_device_attr *props)
+{
+	struct qib_devdata *dd = dd_from_ibdev(ibdev);
+	struct qib_ibdev *dev = to_idev(ibdev);
+
+	memset(props, 0, sizeof(*props));
+
+	props->device_cap_flags = IB_DEVICE_BAD_PKEY_CNTR |
+		IB_DEVICE_BAD_QKEY_CNTR | IB_DEVICE_SHUTDOWN_PORT |
+		IB_DEVICE_SYS_IMAGE_GUID | IB_DEVICE_RC_RNR_NAK_GEN |
+		IB_DEVICE_PORT_ACTIVE_EVENT | IB_DEVICE_SRQ_RESIZE;
+	props->page_size_cap = PAGE_SIZE;
+	props->vendor_id =
+		QIB_SRC_OUI_1 << 16 | QIB_SRC_OUI_2 << 8 | QIB_SRC_OUI_3;
+	props->vendor_part_id = dd->deviceid;
+	props->hw_ver = dd->minrev;
+	props->sys_image_guid = ib_qib_sys_image_guid;
+	props->max_mr_size = ~0ULL;
+	props->max_qp = ib_qib_max_qps;
+	props->max_qp_wr = ib_qib_max_qp_wrs;
+	props->max_sge = ib_qib_max_sges;
+	props->max_cq = ib_qib_max_cqs;
+	props->max_ah = ib_qib_max_ahs;
+	props->max_cqe = ib_qib_max_cqes;
+	props->max_mr = dev->lk_table.max;
+	props->max_fmr = dev->lk_table.max;
+	props->max_map_per_fmr = 32767;
+	props->max_pd = ib_qib_max_pds;
+	props->max_qp_rd_atom = QIB_MAX_RDMA_ATOMIC;
+	props->max_qp_init_rd_atom = 255;
+	/* props->max_res_rd_atom */
+	props->max_srq = ib_qib_max_srqs;
+	props->max_srq_wr = ib_qib_max_srq_wrs;
+	props->max_srq_sge = ib_qib_max_srq_sges;
+	/* props->local_ca_ack_delay */
+	props->atomic_cap = IB_ATOMIC_GLOB;
+	props->max_pkeys = qib_get_npkeys(dd);
+	props->max_mcast_grp = ib_qib_max_mcast_grps;
+	props->max_mcast_qp_attach = ib_qib_max_mcast_qp_attached;
+	props->max_total_mcast_qp_attach = props->max_mcast_qp_attach *
+		props->max_mcast_grp;
+
+	return 0;
+}
+
+static int qib_query_port(struct ib_device *ibdev, u8 port,
+			  struct ib_port_attr *props)
+{
+	struct qib_devdata *dd = dd_from_ibdev(ibdev);
+	struct qib_ibport *ibp = to_iport(ibdev, port);
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	enum ib_mtu mtu;
+	u16 lid = ppd->lid;
+
+	memset(props, 0, sizeof(*props));
+	props->lid = lid ? lid : be16_to_cpu(IB_LID_PERMISSIVE);
+	props->lmc = ppd->lmc;
+	props->sm_lid = ibp->sm_lid;
+	props->sm_sl = ibp->sm_sl;
+	props->state = dd->f_iblink_state(ppd->lastibcstat);
+	props->phys_state = dd->f_ibphys_portstate(ppd->lastibcstat);
+	props->port_cap_flags = ibp->port_cap_flags;
+	props->gid_tbl_len = QIB_GUIDS_PER_PORT;
+	props->max_msg_sz = 0x80000000;
+	props->pkey_tbl_len = qib_get_npkeys(dd);
+	props->bad_pkey_cntr = ibp->pkey_violations;
+	props->qkey_viol_cntr = ibp->qkey_violations;
+	props->active_width = ppd->link_width_active;
+	/* See rate_show() */
+	props->active_speed = ppd->link_speed_active;
+	props->max_vl_num = qib_num_vls(ppd->vls_supported);
+	props->init_type_reply = 0;
+
+	props->max_mtu = qib_ibmtu ? qib_ibmtu : IB_MTU_4096;
+	switch (ppd->ibmtu) {
+	case 4096:
+		mtu = IB_MTU_4096;
+		break;
+	case 2048:
+		mtu = IB_MTU_2048;
+		break;
+	case 1024:
+		mtu = IB_MTU_1024;
+		break;
+	case 512:
+		mtu = IB_MTU_512;
+		break;
+	case 256:
+		mtu = IB_MTU_256;
+		break;
+	default:
+		mtu = IB_MTU_2048;
+	}
+	props->active_mtu = mtu;
+	props->subnet_timeout = ibp->subnet_timeout;
+
+	return 0;
+}
+
+static int qib_modify_device(struct ib_device *device,
+			     int device_modify_mask,
+			     struct ib_device_modify *device_modify)
+{
+	struct qib_devdata *dd = dd_from_ibdev(device);
+	unsigned i;
+	int ret;
+
+	if (device_modify_mask & ~(IB_DEVICE_MODIFY_SYS_IMAGE_GUID |
+				   IB_DEVICE_MODIFY_NODE_DESC)) {
+		ret = -EOPNOTSUPP;
+		goto bail;
+	}
+
+	if (device_modify_mask & IB_DEVICE_MODIFY_NODE_DESC) {
+		memcpy(device->node_desc, device_modify->node_desc, 64);
+		for (i = 0; i < dd->num_pports; i++) {
+			struct qib_ibport *ibp = &dd->pport[i].ibport_data;
+
+			qib_node_desc_chg(ibp);
+		}
+	}
+
+	if (device_modify_mask & IB_DEVICE_MODIFY_SYS_IMAGE_GUID) {
+		ib_qib_sys_image_guid =
+			cpu_to_be64(device_modify->sys_image_guid);
+		for (i = 0; i < dd->num_pports; i++) {
+			struct qib_ibport *ibp = &dd->pport[i].ibport_data;
+
+			qib_sys_guid_chg(ibp);
+		}
+	}
+
+	ret = 0;
+
+bail:
+	return ret;
+}
+
+static int qib_modify_port(struct ib_device *ibdev, u8 port,
+			   int port_modify_mask, struct ib_port_modify *props)
+{
+	struct qib_ibport *ibp = to_iport(ibdev, port);
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+
+	ibp->port_cap_flags |= props->set_port_cap_mask;
+	ibp->port_cap_flags &= ~props->clr_port_cap_mask;
+	if (props->set_port_cap_mask || props->clr_port_cap_mask)
+		qib_cap_mask_chg(ibp);
+	if (port_modify_mask & IB_PORT_SHUTDOWN)
+		qib_set_linkstate(ppd, QIB_IB_LINKDOWN);
+	if (port_modify_mask & IB_PORT_RESET_QKEY_CNTR)
+		ibp->qkey_violations = 0;
+	return 0;
+}
+
+static int qib_query_gid(struct ib_device *ibdev, u8 port,
+			 int index, union ib_gid *gid)
+{
+	struct qib_devdata *dd = dd_from_ibdev(ibdev);
+	int ret = 0;
+
+	if (!port || port > dd->num_pports)
+		ret = -EINVAL;
+	else {
+		struct qib_ibport *ibp = to_iport(ibdev, port);
+		struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+
+		gid->global.subnet_prefix = ibp->gid_prefix;
+		if (index == 0)
+			gid->global.interface_id = ppd->guid;
+		else if (index < QIB_GUIDS_PER_PORT)
+			gid->global.interface_id = ibp->guids[index - 1];
+		else
+			ret = -EINVAL;
+	}
+
+	return ret;
+}
+
+static struct ib_pd *qib_alloc_pd(struct ib_device *ibdev,
+				  struct ib_ucontext *context,
+				  struct ib_udata *udata)
+{
+	struct qib_ibdev *dev = to_idev(ibdev);
+	struct qib_pd *pd;
+	struct ib_pd *ret;
+
+	/*
+	 * This is actually totally arbitrary.  Some correctness tests
+	 * assume there's a maximum number of PDs that can be allocated.
+	 * We don't actually have this limit, but we fail the test if
+	 * we allow allocations of more than we report for this value.
+	 */
+
+	pd = kmalloc(sizeof *pd, GFP_KERNEL);
+	if (!pd) {
+		ret = ERR_PTR(-ENOMEM);
+		goto bail;
+	}
+
+	spin_lock(&dev->n_pds_lock);
+	if (dev->n_pds_allocated == ib_qib_max_pds) {
+		spin_unlock(&dev->n_pds_lock);
+		kfree(pd);
+		ret = ERR_PTR(-ENOMEM);
+		goto bail;
+	}
+
+	dev->n_pds_allocated++;
+	spin_unlock(&dev->n_pds_lock);
+
+	/* ib_alloc_pd() will initialize pd->ibpd. */
+	pd->user = udata != NULL;
+
+	ret = &pd->ibpd;
+
+bail:
+	return ret;
+}
+
+static int qib_dealloc_pd(struct ib_pd *ibpd)
+{
+	struct qib_pd *pd = to_ipd(ibpd);
+	struct qib_ibdev *dev = to_idev(ibpd->device);
+
+	spin_lock(&dev->n_pds_lock);
+	dev->n_pds_allocated--;
+	spin_unlock(&dev->n_pds_lock);
+
+	kfree(pd);
+
+	return 0;
+}
+
+int qib_check_ah(struct ib_device *ibdev, struct ib_ah_attr *ah_attr)
+{
+	/* A multicast address requires a GRH (see ch. 8.4.1). */
+	if (ah_attr->dlid >= QIB_MULTICAST_LID_BASE &&
+	    ah_attr->dlid != QIB_PERMISSIVE_LID &&
+	    !(ah_attr->ah_flags & IB_AH_GRH))
+		goto bail;
+	if ((ah_attr->ah_flags & IB_AH_GRH) &&
+	    ah_attr->grh.sgid_index >= QIB_GUIDS_PER_PORT)
+		goto bail;
+	if (ah_attr->dlid == 0)
+		goto bail;
+	if (ah_attr->port_num < 1 ||
+	    ah_attr->port_num > ibdev->phys_port_cnt)
+		goto bail;
+	if (ah_attr->static_rate != IB_RATE_PORT_CURRENT &&
+	    ib_rate_to_mult(ah_attr->static_rate) < 0)
+		goto bail;
+	if (ah_attr->sl > 15)
+		goto bail;
+	return 0;
+bail:
+	return -EINVAL;
+}
+
+/**
+ * qib_create_ah - create an address handle
+ * @pd: the protection domain
+ * @ah_attr: the attributes of the AH
+ *
+ * This may be called from interrupt context.
+ */
+static struct ib_ah *qib_create_ah(struct ib_pd *pd,
+				   struct ib_ah_attr *ah_attr)
+{
+	struct qib_ah *ah;
+	struct ib_ah *ret;
+	struct qib_ibdev *dev = to_idev(pd->device);
+	unsigned long flags;
+
+	if (qib_check_ah(pd->device, ah_attr)) {
+		ret = ERR_PTR(-EINVAL);
+		goto bail;
+	}
+
+	ah = kmalloc(sizeof *ah, GFP_ATOMIC);
+	if (!ah) {
+		ret = ERR_PTR(-ENOMEM);
+		goto bail;
+	}
+
+	spin_lock_irqsave(&dev->n_ahs_lock, flags);
+	if (dev->n_ahs_allocated == ib_qib_max_ahs) {
+		spin_unlock_irqrestore(&dev->n_ahs_lock, flags);
+		kfree(ah);
+		ret = ERR_PTR(-ENOMEM);
+		goto bail;
+	}
+
+	dev->n_ahs_allocated++;
+	spin_unlock_irqrestore(&dev->n_ahs_lock, flags);
+
+	/* ib_create_ah() will initialize ah->ibah. */
+	ah->attr = *ah_attr;
+	atomic_set(&ah->refcount, 0);
+
+	ret = &ah->ibah;
+
+bail:
+	return ret;
+}
+
+/**
+ * qib_destroy_ah - destroy an address handle
+ * @ibah: the AH to destroy
+ *
+ * This may be called from interrupt context.
+ */
+static int qib_destroy_ah(struct ib_ah *ibah)
+{
+	struct qib_ibdev *dev = to_idev(ibah->device);
+	struct qib_ah *ah = to_iah(ibah);
+	unsigned long flags;
+
+	if (atomic_read(&ah->refcount) != 0)
+		return -EBUSY;
+
+	spin_lock_irqsave(&dev->n_ahs_lock, flags);
+	dev->n_ahs_allocated--;
+	spin_unlock_irqrestore(&dev->n_ahs_lock, flags);
+
+	kfree(ah);
+
+	return 0;
+}
+
+static int qib_modify_ah(struct ib_ah *ibah, struct ib_ah_attr *ah_attr)
+{
+	struct qib_ah *ah = to_iah(ibah);
+
+	if (qib_check_ah(ibah->device, ah_attr))
+		return -EINVAL;
+
+	ah->attr = *ah_attr;
+
+	return 0;
+}
+
+static int qib_query_ah(struct ib_ah *ibah, struct ib_ah_attr *ah_attr)
+{
+	struct qib_ah *ah = to_iah(ibah);
+
+	*ah_attr = ah->attr;
+
+	return 0;
+}
+
+/**
+ * qib_get_npkeys - return the size of the PKEY table for context 0
+ * @dd: the qlogic_ib device
+ */
+unsigned qib_get_npkeys(struct qib_devdata *dd)
+{
+	return ARRAY_SIZE(dd->rcd[0]->pkeys);
+}
+
+/*
+ * Return the indexed PKEY from the port PKEY table.
+ * No need to validate rcd[ctxt]; the port is setup if we are here.
+ */
+unsigned qib_get_pkey(struct qib_ibport *ibp, unsigned index)
+{
+	struct qib_pportdata *ppd = ppd_from_ibp(ibp);
+	struct qib_devdata *dd = ppd->dd;
+	unsigned ctxt = ppd->hw_pidx;
+	unsigned ret;
+
+	/* dd->rcd null if mini_init or some init failures */
+	if (!dd->rcd || index >= ARRAY_SIZE(dd->rcd[ctxt]->pkeys))
+		ret = 0;
+	else
+		ret = dd->rcd[ctxt]->pkeys[index];
+
+	return ret;
+}
+
+static int qib_query_pkey(struct ib_device *ibdev, u8 port, u16 index,
+			  u16 *pkey)
+{
+	struct qib_devdata *dd = dd_from_ibdev(ibdev);
+	int ret;
+
+	if (index >= qib_get_npkeys(dd)) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	*pkey = qib_get_pkey(to_iport(ibdev, port), index);
+	ret = 0;
+
+bail:
+	return ret;
+}
+
+/**
+ * qib_alloc_ucontext - allocate a ucontest
+ * @ibdev: the infiniband device
+ * @udata: not used by the QLogic_IB driver
+ */
+
+static struct ib_ucontext *qib_alloc_ucontext(struct ib_device *ibdev,
+					      struct ib_udata *udata)
+{
+	struct qib_ucontext *context;
+	struct ib_ucontext *ret;
+
+	context = kmalloc(sizeof *context, GFP_KERNEL);
+	if (!context) {
+		ret = ERR_PTR(-ENOMEM);
+		goto bail;
+	}
+
+	ret = &context->ibucontext;
+
+bail:
+	return ret;
+}
+
+static int qib_dealloc_ucontext(struct ib_ucontext *context)
+{
+	kfree(to_iucontext(context));
+	return 0;
+}
+
+static void init_ibport(struct qib_pportdata *ppd)
+{
+	struct qib_verbs_counters cntrs;
+	struct qib_ibport *ibp = &ppd->ibport_data;
+
+	spin_lock_init(&ibp->lock);
+	/* Set the prefix to the default value (see ch. 4.1.1) */
+	ibp->gid_prefix = IB_DEFAULT_GID_PREFIX;
+	ibp->sm_lid = be16_to_cpu(IB_LID_PERMISSIVE);
+	ibp->port_cap_flags = IB_PORT_SYS_IMAGE_GUID_SUP |
+		IB_PORT_CLIENT_REG_SUP | IB_PORT_SL_MAP_SUP |
+		IB_PORT_TRAP_SUP | IB_PORT_AUTO_MIGR_SUP |
+		IB_PORT_DR_NOTICE_SUP | IB_PORT_CAP_MASK_NOTICE_SUP |
+		IB_PORT_OTHER_LOCAL_CHANGES_SUP;
+	if (ppd->dd->flags & QIB_HAS_LINK_LATENCY)
+		ibp->port_cap_flags |= IB_PORT_LINK_LATENCY_SUP;
+	ibp->pma_counter_select[0] = IB_PMA_PORT_XMIT_DATA;
+	ibp->pma_counter_select[1] = IB_PMA_PORT_RCV_DATA;
+	ibp->pma_counter_select[2] = IB_PMA_PORT_XMIT_PKTS;
+	ibp->pma_counter_select[3] = IB_PMA_PORT_RCV_PKTS;
+	ibp->pma_counter_select[4] = IB_PMA_PORT_XMIT_WAIT;
+
+	/* Snapshot current HW counters to "clear" them. */
+	qib_get_counters(ppd, &cntrs);
+	ibp->z_symbol_error_counter = cntrs.symbol_error_counter;
+	ibp->z_link_error_recovery_counter =
+		cntrs.link_error_recovery_counter;
+	ibp->z_link_downed_counter = cntrs.link_downed_counter;
+	ibp->z_port_rcv_errors = cntrs.port_rcv_errors;
+	ibp->z_port_rcv_remphys_errors = cntrs.port_rcv_remphys_errors;
+	ibp->z_port_xmit_discards = cntrs.port_xmit_discards;
+	ibp->z_port_xmit_data = cntrs.port_xmit_data;
+	ibp->z_port_rcv_data = cntrs.port_rcv_data;
+	ibp->z_port_xmit_packets = cntrs.port_xmit_packets;
+	ibp->z_port_rcv_packets = cntrs.port_rcv_packets;
+	ibp->z_local_link_integrity_errors =
+		cntrs.local_link_integrity_errors;
+	ibp->z_excessive_buffer_overrun_errors =
+		cntrs.excessive_buffer_overrun_errors;
+	ibp->z_vl15_dropped = cntrs.vl15_dropped;
+}
+
+/**
+ * qib_register_ib_device - register our device with the infiniband core
+ * @dd: the device data structure
+ * Return the allocated qib_ibdev pointer or NULL on error.
+ */
+int qib_register_ib_device(struct qib_devdata *dd)
+{
+	struct qib_ibdev *dev = &dd->verbs_dev;
+	struct ib_device *ibdev = &dev->ibdev;
+	struct qib_pportdata *ppd = dd->pport;
+	unsigned i, lk_tab_size;
+	int ret;
+
+	dev->qp_table_size = ib_qib_qp_table_size;
+	dev->qp_table = kzalloc(dev->qp_table_size * sizeof *dev->qp_table,
+				GFP_KERNEL);
+	if (!dev->qp_table) {
+		ret = -ENOMEM;
+		goto err_qpt;
+	}
+
+	for (i = 0; i < dd->num_pports; i++)
+		init_ibport(ppd + i);
+
+	/* Only need to initialize non-zero fields. */
+	spin_lock_init(&dev->qpt_lock);
+	spin_lock_init(&dev->n_pds_lock);
+	spin_lock_init(&dev->n_ahs_lock);
+	spin_lock_init(&dev->n_cqs_lock);
+	spin_lock_init(&dev->n_qps_lock);
+	spin_lock_init(&dev->n_srqs_lock);
+	spin_lock_init(&dev->n_mcast_grps_lock);
+	init_timer(&dev->mem_timer);
+	dev->mem_timer.function = mem_timer;
+	dev->mem_timer.data = (unsigned long) dev;
+
+	qib_init_qpn_table(dd, &dev->qpn_table);
+
+	/*
+	 * The top ib_qib_lkey_table_size bits are used to index the
+	 * table.  The lower 8 bits can be owned by the user (copied from
+	 * the LKEY).  The remaining bits act as a generation number or tag.
+	 */
+	spin_lock_init(&dev->lk_table.lock);
+	dev->lk_table.max = 1 << ib_qib_lkey_table_size;
+	lk_tab_size = dev->lk_table.max * sizeof(*dev->lk_table.table);
+	dev->lk_table.table = (struct qib_mregion **)
+		__get_free_pages(GFP_KERNEL, get_order(lk_tab_size));
+	if (dev->lk_table.table == NULL) {
+		ret = -ENOMEM;
+		goto err_lk;
+	}
+	memset(dev->lk_table.table, 0, lk_tab_size);
+	INIT_LIST_HEAD(&dev->pending_mmaps);
+	spin_lock_init(&dev->pending_lock);
+	dev->mmap_offset = PAGE_SIZE;
+	spin_lock_init(&dev->mmap_offset_lock);
+	INIT_LIST_HEAD(&dev->piowait);
+	INIT_LIST_HEAD(&dev->dmawait);
+	INIT_LIST_HEAD(&dev->txwait);
+	INIT_LIST_HEAD(&dev->memwait);
+	INIT_LIST_HEAD(&dev->txreq_free);
+
+	if (ppd->sdma_descq_cnt) {
+		dev->pio_hdrs = dma_alloc_coherent(&dd->pcidev->dev,
+						ppd->sdma_descq_cnt *
+						sizeof(struct qib_pio_header),
+						&dev->pio_hdrs_phys,
+						GFP_KERNEL);
+		if (!dev->pio_hdrs) {
+			ret = -ENOMEM;
+			goto err_hdrs;
+		}
+	}
+
+	for (i = 0; i < ppd->sdma_descq_cnt; i++) {
+		struct qib_verbs_txreq *tx;
+
+		tx = kzalloc(sizeof *tx, GFP_KERNEL);
+		if (!tx) {
+			ret = -ENOMEM;
+			goto err_tx;
+		}
+		tx->hdr_inx = i;
+		list_add(&tx->txreq.list, &dev->txreq_free);
+	}
+
+	/*
+	 * The system image GUID is supposed to be the same for all
+	 * IB HCAs in a single system but since there can be other
+	 * device types in the system, we can't be sure this is unique.
+	 */
+	if (!ib_qib_sys_image_guid)
+		ib_qib_sys_image_guid = ppd->guid;
+
+	strlcpy(ibdev->name, "qib%d", IB_DEVICE_NAME_MAX);
+	ibdev->owner = THIS_MODULE;
+	ibdev->node_guid = ppd->guid;
+	ibdev->uverbs_abi_ver = QIB_UVERBS_ABI_VERSION;
+	ibdev->uverbs_cmd_mask =
+		(1ull << IB_USER_VERBS_CMD_GET_CONTEXT)         |
+		(1ull << IB_USER_VERBS_CMD_QUERY_DEVICE)        |
+		(1ull << IB_USER_VERBS_CMD_QUERY_PORT)          |
+		(1ull << IB_USER_VERBS_CMD_ALLOC_PD)            |
+		(1ull << IB_USER_VERBS_CMD_DEALLOC_PD)          |
+		(1ull << IB_USER_VERBS_CMD_CREATE_AH)           |
+		(1ull << IB_USER_VERBS_CMD_MODIFY_AH)           |
+		(1ull << IB_USER_VERBS_CMD_QUERY_AH)            |
+		(1ull << IB_USER_VERBS_CMD_DESTROY_AH)          |
+		(1ull << IB_USER_VERBS_CMD_REG_MR)              |
+		(1ull << IB_USER_VERBS_CMD_DEREG_MR)            |
+		(1ull << IB_USER_VERBS_CMD_CREATE_COMP_CHANNEL) |
+		(1ull << IB_USER_VERBS_CMD_CREATE_CQ)           |
+		(1ull << IB_USER_VERBS_CMD_RESIZE_CQ)           |
+		(1ull << IB_USER_VERBS_CMD_DESTROY_CQ)          |
+		(1ull << IB_USER_VERBS_CMD_POLL_CQ)             |
+		(1ull << IB_USER_VERBS_CMD_REQ_NOTIFY_CQ)       |
+		(1ull << IB_USER_VERBS_CMD_CREATE_QP)           |
+		(1ull << IB_USER_VERBS_CMD_QUERY_QP)            |
+		(1ull << IB_USER_VERBS_CMD_MODIFY_QP)           |
+		(1ull << IB_USER_VERBS_CMD_DESTROY_QP)          |
+		(1ull << IB_USER_VERBS_CMD_POST_SEND)           |
+		(1ull << IB_USER_VERBS_CMD_POST_RECV)           |
+		(1ull << IB_USER_VERBS_CMD_ATTACH_MCAST)        |
+		(1ull << IB_USER_VERBS_CMD_DETACH_MCAST)        |
+		(1ull << IB_USER_VERBS_CMD_CREATE_SRQ)          |
+		(1ull << IB_USER_VERBS_CMD_MODIFY_SRQ)          |
+		(1ull << IB_USER_VERBS_CMD_QUERY_SRQ)           |
+		(1ull << IB_USER_VERBS_CMD_DESTROY_SRQ)         |
+		(1ull << IB_USER_VERBS_CMD_POST_SRQ_RECV);
+	ibdev->node_type = RDMA_NODE_IB_CA;
+	ibdev->phys_port_cnt = dd->num_pports;
+	ibdev->num_comp_vectors = 1;
+	ibdev->dma_device = &dd->pcidev->dev;
+	ibdev->query_device = qib_query_device;
+	ibdev->modify_device = qib_modify_device;
+	ibdev->query_port = qib_query_port;
+	ibdev->modify_port = qib_modify_port;
+	ibdev->query_pkey = qib_query_pkey;
+	ibdev->query_gid = qib_query_gid;
+	ibdev->alloc_ucontext = qib_alloc_ucontext;
+	ibdev->dealloc_ucontext = qib_dealloc_ucontext;
+	ibdev->alloc_pd = qib_alloc_pd;
+	ibdev->dealloc_pd = qib_dealloc_pd;
+	ibdev->create_ah = qib_create_ah;
+	ibdev->destroy_ah = qib_destroy_ah;
+	ibdev->modify_ah = qib_modify_ah;
+	ibdev->query_ah = qib_query_ah;
+	ibdev->create_srq = qib_create_srq;
+	ibdev->modify_srq = qib_modify_srq;
+	ibdev->query_srq = qib_query_srq;
+	ibdev->destroy_srq = qib_destroy_srq;
+	ibdev->create_qp = qib_create_qp;
+	ibdev->modify_qp = qib_modify_qp;
+	ibdev->query_qp = qib_query_qp;
+	ibdev->destroy_qp = qib_destroy_qp;
+	ibdev->post_send = qib_post_send;
+	ibdev->post_recv = qib_post_receive;
+	ibdev->post_srq_recv = qib_post_srq_receive;
+	ibdev->create_cq = qib_create_cq;
+	ibdev->destroy_cq = qib_destroy_cq;
+	ibdev->resize_cq = qib_resize_cq;
+	ibdev->poll_cq = qib_poll_cq;
+	ibdev->req_notify_cq = qib_req_notify_cq;
+	ibdev->get_dma_mr = qib_get_dma_mr;
+	ibdev->reg_phys_mr = qib_reg_phys_mr;
+	ibdev->reg_user_mr = qib_reg_user_mr;
+	ibdev->dereg_mr = qib_dereg_mr;
+	ibdev->alloc_fast_reg_mr = qib_alloc_fast_reg_mr;
+	ibdev->alloc_fast_reg_page_list = qib_alloc_fast_reg_page_list;
+	ibdev->free_fast_reg_page_list = qib_free_fast_reg_page_list;
+	ibdev->alloc_fmr = qib_alloc_fmr;
+	ibdev->map_phys_fmr = qib_map_phys_fmr;
+	ibdev->unmap_fmr = qib_unmap_fmr;
+	ibdev->dealloc_fmr = qib_dealloc_fmr;
+	ibdev->attach_mcast = qib_multicast_attach;
+	ibdev->detach_mcast = qib_multicast_detach;
+	ibdev->process_mad = qib_process_mad;
+	ibdev->mmap = qib_mmap;
+	ibdev->dma_ops = &qib_dma_mapping_ops;
+
+	snprintf(ibdev->node_desc, sizeof(ibdev->node_desc),
+		 QIB_IDSTR " %s", init_utsname()->nodename);
+
+	ret = ib_register_device(ibdev, qib_create_port_files);
+	if (ret)
+		goto err_reg;
+
+	ret = qib_create_agents(dev);
+	if (ret)
+		goto err_agents;
+
+	if (qib_verbs_register_sysfs(dd))
+		goto err_class;
+
+	goto bail;
+
+err_class:
+	qib_free_agents(dev);
+err_agents:
+	ib_unregister_device(ibdev);
+err_reg:
+err_tx:
+	while (!list_empty(&dev->txreq_free)) {
+		struct list_head *l = dev->txreq_free.next;
+		struct qib_verbs_txreq *tx;
+
+		list_del(l);
+		tx = list_entry(l, struct qib_verbs_txreq, txreq.list);
+		kfree(tx);
+	}
+	if (ppd->sdma_descq_cnt)
+		dma_free_coherent(&dd->pcidev->dev,
+				  ppd->sdma_descq_cnt *
+					sizeof(struct qib_pio_header),
+				  dev->pio_hdrs, dev->pio_hdrs_phys);
+err_hdrs:
+	free_pages((unsigned long) dev->lk_table.table, get_order(lk_tab_size));
+err_lk:
+	kfree(dev->qp_table);
+err_qpt:
+	qib_dev_err(dd, "cannot register verbs: %d!\n", -ret);
+bail:
+	return ret;
+}
+
+void qib_unregister_ib_device(struct qib_devdata *dd)
+{
+	struct qib_ibdev *dev = &dd->verbs_dev;
+	struct ib_device *ibdev = &dev->ibdev;
+	u32 qps_inuse;
+	unsigned lk_tab_size;
+
+	qib_verbs_unregister_sysfs(dd);
+
+	qib_free_agents(dev);
+
+	ib_unregister_device(ibdev);
+
+	if (!list_empty(&dev->piowait))
+		qib_dev_err(dd, "piowait list not empty!\n");
+	if (!list_empty(&dev->dmawait))
+		qib_dev_err(dd, "dmawait list not empty!\n");
+	if (!list_empty(&dev->txwait))
+		qib_dev_err(dd, "txwait list not empty!\n");
+	if (!list_empty(&dev->memwait))
+		qib_dev_err(dd, "memwait list not empty!\n");
+	if (dev->dma_mr)
+		qib_dev_err(dd, "DMA MR not NULL!\n");
+
+	qps_inuse = qib_free_all_qps(dd);
+	if (qps_inuse)
+		qib_dev_err(dd, "QP memory leak! %u still in use\n",
+			    qps_inuse);
+
+	del_timer_sync(&dev->mem_timer);
+	qib_free_qpn_table(&dev->qpn_table);
+	while (!list_empty(&dev->txreq_free)) {
+		struct list_head *l = dev->txreq_free.next;
+		struct qib_verbs_txreq *tx;
+
+		list_del(l);
+		tx = list_entry(l, struct qib_verbs_txreq, txreq.list);
+		kfree(tx);
+	}
+	if (dd->pport->sdma_descq_cnt)
+		dma_free_coherent(&dd->pcidev->dev,
+				  dd->pport->sdma_descq_cnt *
+					sizeof(struct qib_pio_header),
+				  dev->pio_hdrs, dev->pio_hdrs_phys);
+	lk_tab_size = dev->lk_table.max * sizeof(*dev->lk_table.table);
+	free_pages((unsigned long) dev->lk_table.table,
+		   get_order(lk_tab_size));
+	kfree(dev->qp_table);
+}
diff --git a/drivers/infiniband/hw/qib/qib_verbs.h b/drivers/infiniband/hw/qib/qib_verbs.h
new file mode 100644
index 000000000000..bd57c1273225
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_verbs.h
@@ -0,0 +1,1100 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009, 2010 QLogic Corporation.
+ * All rights reserved.
+ * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 QIB_VERBS_H
+#define QIB_VERBS_H
+
+#include <linux/types.h>
+#include <linux/spinlock.h>
+#include <linux/kernel.h>
+#include <linux/interrupt.h>
+#include <linux/kref.h>
+#include <linux/workqueue.h>
+#include <rdma/ib_pack.h>
+#include <rdma/ib_user_verbs.h>
+
+struct qib_ctxtdata;
+struct qib_pportdata;
+struct qib_devdata;
+struct qib_verbs_txreq;
+
+#define QIB_MAX_RDMA_ATOMIC     16
+#define QIB_GUIDS_PER_PORT	5
+
+#define QPN_MAX                 (1 << 24)
+#define QPNMAP_ENTRIES          (QPN_MAX / PAGE_SIZE / BITS_PER_BYTE)
+
+/*
+ * Increment this value if any changes that break userspace ABI
+ * compatibility are made.
+ */
+#define QIB_UVERBS_ABI_VERSION       2
+
+/*
+ * Define an ib_cq_notify value that is not valid so we know when CQ
+ * notifications are armed.
+ */
+#define IB_CQ_NONE      (IB_CQ_NEXT_COMP + 1)
+
+#define IB_SEQ_NAK	(3 << 29)
+
+/* AETH NAK opcode values */
+#define IB_RNR_NAK                      0x20
+#define IB_NAK_PSN_ERROR                0x60
+#define IB_NAK_INVALID_REQUEST          0x61
+#define IB_NAK_REMOTE_ACCESS_ERROR      0x62
+#define IB_NAK_REMOTE_OPERATIONAL_ERROR 0x63
+#define IB_NAK_INVALID_RD_REQUEST       0x64
+
+/* Flags for checking QP state (see ib_qib_state_ops[]) */
+#define QIB_POST_SEND_OK                0x01
+#define QIB_POST_RECV_OK                0x02
+#define QIB_PROCESS_RECV_OK             0x04
+#define QIB_PROCESS_SEND_OK             0x08
+#define QIB_PROCESS_NEXT_SEND_OK        0x10
+#define QIB_FLUSH_SEND			0x20
+#define QIB_FLUSH_RECV			0x40
+#define QIB_PROCESS_OR_FLUSH_SEND \
+	(QIB_PROCESS_SEND_OK | QIB_FLUSH_SEND)
+
+/* IB Performance Manager status values */
+#define IB_PMA_SAMPLE_STATUS_DONE       0x00
+#define IB_PMA_SAMPLE_STATUS_STARTED    0x01
+#define IB_PMA_SAMPLE_STATUS_RUNNING    0x02
+
+/* Mandatory IB performance counter select values. */
+#define IB_PMA_PORT_XMIT_DATA   cpu_to_be16(0x0001)
+#define IB_PMA_PORT_RCV_DATA    cpu_to_be16(0x0002)
+#define IB_PMA_PORT_XMIT_PKTS   cpu_to_be16(0x0003)
+#define IB_PMA_PORT_RCV_PKTS    cpu_to_be16(0x0004)
+#define IB_PMA_PORT_XMIT_WAIT   cpu_to_be16(0x0005)
+
+#define QIB_VENDOR_IPG		cpu_to_be16(0xFFA0)
+
+#define IB_BTH_REQ_ACK		(1 << 31)
+#define IB_BTH_SOLICITED	(1 << 23)
+#define IB_BTH_MIG_REQ		(1 << 22)
+
+/* XXX Should be defined in ib_verbs.h enum ib_port_cap_flags */
+#define IB_PORT_OTHER_LOCAL_CHANGES_SUP (1 << 26)
+
+#define IB_GRH_VERSION		6
+#define IB_GRH_VERSION_MASK	0xF
+#define IB_GRH_VERSION_SHIFT	28
+#define IB_GRH_TCLASS_MASK	0xFF
+#define IB_GRH_TCLASS_SHIFT	20
+#define IB_GRH_FLOW_MASK	0xFFFFF
+#define IB_GRH_FLOW_SHIFT	0
+#define IB_GRH_NEXT_HDR		0x1B
+
+#define IB_DEFAULT_GID_PREFIX	cpu_to_be64(0xfe80000000000000ULL)
+
+/* Values for set/get portinfo VLCap OperationalVLs */
+#define IB_VL_VL0       1
+#define IB_VL_VL0_1     2
+#define IB_VL_VL0_3     3
+#define IB_VL_VL0_7     4
+#define IB_VL_VL0_14    5
+
+static inline int qib_num_vls(int vls)
+{
+	switch (vls) {
+	default:
+	case IB_VL_VL0:
+		return 1;
+	case IB_VL_VL0_1:
+		return 2;
+	case IB_VL_VL0_3:
+		return 4;
+	case IB_VL_VL0_7:
+		return 8;
+	case IB_VL_VL0_14:
+		return 15;
+	}
+}
+
+struct ib_reth {
+	__be64 vaddr;
+	__be32 rkey;
+	__be32 length;
+} __attribute__ ((packed));
+
+struct ib_atomic_eth {
+	__be32 vaddr[2];        /* unaligned so access as 2 32-bit words */
+	__be32 rkey;
+	__be64 swap_data;
+	__be64 compare_data;
+} __attribute__ ((packed));
+
+struct qib_other_headers {
+	__be32 bth[3];
+	union {
+		struct {
+			__be32 deth[2];
+			__be32 imm_data;
+		} ud;
+		struct {
+			struct ib_reth reth;
+			__be32 imm_data;
+		} rc;
+		struct {
+			__be32 aeth;
+			__be32 atomic_ack_eth[2];
+		} at;
+		__be32 imm_data;
+		__be32 aeth;
+		struct ib_atomic_eth atomic_eth;
+	} u;
+} __attribute__ ((packed));
+
+/*
+ * Note that UD packets with a GRH header are 8+40+12+8 = 68 bytes
+ * long (72 w/ imm_data).  Only the first 56 bytes of the IB header
+ * will be in the eager header buffer.  The remaining 12 or 16 bytes
+ * are in the data buffer.
+ */
+struct qib_ib_header {
+	__be16 lrh[4];
+	union {
+		struct {
+			struct ib_grh grh;
+			struct qib_other_headers oth;
+		} l;
+		struct qib_other_headers oth;
+	} u;
+} __attribute__ ((packed));
+
+struct qib_pio_header {
+	__le32 pbc[2];
+	struct qib_ib_header hdr;
+} __attribute__ ((packed));
+
+/*
+ * There is one struct qib_mcast for each multicast GID.
+ * All attached QPs are then stored as a list of
+ * struct qib_mcast_qp.
+ */
+struct qib_mcast_qp {
+	struct list_head list;
+	struct qib_qp *qp;
+};
+
+struct qib_mcast {
+	struct rb_node rb_node;
+	union ib_gid mgid;
+	struct list_head qp_list;
+	wait_queue_head_t wait;
+	atomic_t refcount;
+	int n_attached;
+};
+
+/* Protection domain */
+struct qib_pd {
+	struct ib_pd ibpd;
+	int user;               /* non-zero if created from user space */
+};
+
+/* Address Handle */
+struct qib_ah {
+	struct ib_ah ibah;
+	struct ib_ah_attr attr;
+	atomic_t refcount;
+};
+
+/*
+ * This structure is used by qib_mmap() to validate an offset
+ * when an mmap() request is made.  The vm_area_struct then uses
+ * this as its vm_private_data.
+ */
+struct qib_mmap_info {
+	struct list_head pending_mmaps;
+	struct ib_ucontext *context;
+	void *obj;
+	__u64 offset;
+	struct kref ref;
+	unsigned size;
+};
+
+/*
+ * This structure is used to contain the head pointer, tail pointer,
+ * and completion queue entries as a single memory allocation so
+ * it can be mmap'ed into user space.
+ */
+struct qib_cq_wc {
+	u32 head;               /* index of next entry to fill */
+	u32 tail;               /* index of next ib_poll_cq() entry */
+	union {
+		/* these are actually size ibcq.cqe + 1 */
+		struct ib_uverbs_wc uqueue[0];
+		struct ib_wc kqueue[0];
+	};
+};
+
+/*
+ * The completion queue structure.
+ */
+struct qib_cq {
+	struct ib_cq ibcq;
+	struct work_struct comptask;
+	spinlock_t lock; /* protect changes in this struct */
+	u8 notify;
+	u8 triggered;
+	struct qib_cq_wc *queue;
+	struct qib_mmap_info *ip;
+};
+
+/*
+ * A segment is a linear region of low physical memory.
+ * XXX Maybe we should use phys addr here and kmap()/kunmap().
+ * Used by the verbs layer.
+ */
+struct qib_seg {
+	void *vaddr;
+	size_t length;
+};
+
+/* The number of qib_segs that fit in a page. */
+#define QIB_SEGSZ     (PAGE_SIZE / sizeof(struct qib_seg))
+
+struct qib_segarray {
+	struct qib_seg segs[QIB_SEGSZ];
+};
+
+struct qib_mregion {
+	struct ib_pd *pd;       /* shares refcnt of ibmr.pd */
+	u64 user_base;          /* User's address for this region */
+	u64 iova;               /* IB start address of this region */
+	size_t length;
+	u32 lkey;
+	u32 offset;             /* offset (bytes) to start of region */
+	int access_flags;
+	u32 max_segs;           /* number of qib_segs in all the arrays */
+	u32 mapsz;              /* size of the map array */
+	atomic_t refcount;
+	struct qib_segarray *map[0];    /* the segments */
+};
+
+/*
+ * These keep track of the copy progress within a memory region.
+ * Used by the verbs layer.
+ */
+struct qib_sge {
+	struct qib_mregion *mr;
+	void *vaddr;            /* kernel virtual address of segment */
+	u32 sge_length;         /* length of the SGE */
+	u32 length;             /* remaining length of the segment */
+	u16 m;                  /* current index: mr->map[m] */
+	u16 n;                  /* current index: mr->map[m]->segs[n] */
+};
+
+/* Memory region */
+struct qib_mr {
+	struct ib_mr ibmr;
+	struct ib_umem *umem;
+	struct qib_mregion mr;  /* must be last */
+};
+
+/*
+ * Send work request queue entry.
+ * The size of the sg_list is determined when the QP is created and stored
+ * in qp->s_max_sge.
+ */
+struct qib_swqe {
+	struct ib_send_wr wr;   /* don't use wr.sg_list */
+	u32 psn;                /* first packet sequence number */
+	u32 lpsn;               /* last packet sequence number */
+	u32 ssn;                /* send sequence number */
+	u32 length;             /* total length of data in sg_list */
+	struct qib_sge sg_list[0];
+};
+
+/*
+ * Receive work request queue entry.
+ * The size of the sg_list is determined when the QP (or SRQ) is created
+ * and stored in qp->r_rq.max_sge (or srq->rq.max_sge).
+ */
+struct qib_rwqe {
+	u64 wr_id;
+	u8 num_sge;
+	struct ib_sge sg_list[0];
+};
+
+/*
+ * This structure is used to contain the head pointer, tail pointer,
+ * and receive work queue entries as a single memory allocation so
+ * it can be mmap'ed into user space.
+ * Note that the wq array elements are variable size so you can't
+ * just index into the array to get the N'th element;
+ * use get_rwqe_ptr() instead.
+ */
+struct qib_rwq {
+	u32 head;               /* new work requests posted to the head */
+	u32 tail;               /* receives pull requests from here. */
+	struct qib_rwqe wq[0];
+};
+
+struct qib_rq {
+	struct qib_rwq *wq;
+	spinlock_t lock; /* protect changes in this struct */
+	u32 size;               /* size of RWQE array */
+	u8 max_sge;
+};
+
+struct qib_srq {
+	struct ib_srq ibsrq;
+	struct qib_rq rq;
+	struct qib_mmap_info *ip;
+	/* send signal when number of RWQEs < limit */
+	u32 limit;
+};
+
+struct qib_sge_state {
+	struct qib_sge *sg_list;      /* next SGE to be used if any */
+	struct qib_sge sge;   /* progress state for the current SGE */
+	u32 total_len;
+	u8 num_sge;
+};
+
+/*
+ * This structure holds the information that the send tasklet needs
+ * to send a RDMA read response or atomic operation.
+ */
+struct qib_ack_entry {
+	u8 opcode;
+	u8 sent;
+	u32 psn;
+	u32 lpsn;
+	union {
+		struct qib_sge rdma_sge;
+		u64 atomic_data;
+	};
+};
+
+/*
+ * Variables prefixed with s_ are for the requester (sender).
+ * Variables prefixed with r_ are for the responder (receiver).
+ * Variables prefixed with ack_ are for responder replies.
+ *
+ * Common variables are protected by both r_rq.lock and s_lock in that order
+ * which only happens in modify_qp() or changing the QP 'state'.
+ */
+struct qib_qp {
+	struct ib_qp ibqp;
+	struct qib_qp *next;            /* link list for QPN hash table */
+	struct qib_qp *timer_next;      /* link list for qib_ib_timer() */
+	struct list_head iowait;        /* link for wait PIO buf */
+	struct list_head rspwait;       /* link for waititing to respond */
+	struct ib_ah_attr remote_ah_attr;
+	struct ib_ah_attr alt_ah_attr;
+	struct qib_ib_header s_hdr;     /* next packet header to send */
+	atomic_t refcount;
+	wait_queue_head_t wait;
+	wait_queue_head_t wait_dma;
+	struct timer_list s_timer;
+	struct work_struct s_work;
+	struct qib_mmap_info *ip;
+	struct qib_sge_state *s_cur_sge;
+	struct qib_verbs_txreq *s_tx;
+	struct qib_mregion *s_rdma_mr;
+	struct qib_sge_state s_sge;     /* current send request data */
+	struct qib_ack_entry s_ack_queue[QIB_MAX_RDMA_ATOMIC + 1];
+	struct qib_sge_state s_ack_rdma_sge;
+	struct qib_sge_state s_rdma_read_sge;
+	struct qib_sge_state r_sge;     /* current receive data */
+	spinlock_t r_lock;      /* used for APM */
+	spinlock_t s_lock;
+	atomic_t s_dma_busy;
+	unsigned processor_id;	/* Processor ID QP is bound to */
+	u32 s_flags;
+	u32 s_cur_size;         /* size of send packet in bytes */
+	u32 s_len;              /* total length of s_sge */
+	u32 s_rdma_read_len;    /* total length of s_rdma_read_sge */
+	u32 s_next_psn;         /* PSN for next request */
+	u32 s_last_psn;         /* last response PSN processed */
+	u32 s_sending_psn;      /* lowest PSN that is being sent */
+	u32 s_sending_hpsn;     /* highest PSN that is being sent */
+	u32 s_psn;              /* current packet sequence number */
+	u32 s_ack_rdma_psn;     /* PSN for sending RDMA read responses */
+	u32 s_ack_psn;          /* PSN for acking sends and RDMA writes */
+	u32 s_rnr_timeout;      /* number of milliseconds for RNR timeout */
+	u32 r_ack_psn;          /* PSN for next ACK or atomic ACK */
+	u64 r_wr_id;            /* ID for current receive WQE */
+	unsigned long r_aflags;
+	u32 r_len;              /* total length of r_sge */
+	u32 r_rcv_len;          /* receive data len processed */
+	u32 r_psn;              /* expected rcv packet sequence number */
+	u32 r_msn;              /* message sequence number */
+	u16 s_hdrwords;         /* size of s_hdr in 32 bit words */
+	u16 s_rdma_ack_cnt;
+	u8 state;               /* QP state */
+	u8 s_state;             /* opcode of last packet sent */
+	u8 s_ack_state;         /* opcode of packet to ACK */
+	u8 s_nak_state;         /* non-zero if NAK is pending */
+	u8 r_state;             /* opcode of last packet received */
+	u8 r_nak_state;         /* non-zero if NAK is pending */
+	u8 r_min_rnr_timer;     /* retry timeout value for RNR NAKs */
+	u8 r_flags;
+	u8 r_max_rd_atomic;     /* max number of RDMA read/atomic to receive */
+	u8 r_head_ack_queue;    /* index into s_ack_queue[] */
+	u8 qp_access_flags;
+	u8 s_max_sge;           /* size of s_wq->sg_list */
+	u8 s_retry_cnt;         /* number of times to retry */
+	u8 s_rnr_retry_cnt;
+	u8 s_retry;             /* requester retry counter */
+	u8 s_rnr_retry;         /* requester RNR retry counter */
+	u8 s_pkey_index;        /* PKEY index to use */
+	u8 s_alt_pkey_index;    /* Alternate path PKEY index to use */
+	u8 s_max_rd_atomic;     /* max number of RDMA read/atomic to send */
+	u8 s_num_rd_atomic;     /* number of RDMA read/atomic pending */
+	u8 s_tail_ack_queue;    /* index into s_ack_queue[] */
+	u8 s_srate;
+	u8 s_draining;
+	u8 s_mig_state;
+	u8 timeout;             /* Timeout for this QP */
+	u8 alt_timeout;         /* Alternate path timeout for this QP */
+	u8 port_num;
+	enum ib_mtu path_mtu;
+	u32 remote_qpn;
+	u32 qkey;               /* QKEY for this QP (for UD or RD) */
+	u32 s_size;             /* send work queue size */
+	u32 s_head;             /* new entries added here */
+	u32 s_tail;             /* next entry to process */
+	u32 s_cur;              /* current work queue entry */
+	u32 s_acked;            /* last un-ACK'ed entry */
+	u32 s_last;             /* last completed entry */
+	u32 s_ssn;              /* SSN of tail entry */
+	u32 s_lsn;              /* limit sequence number (credit) */
+	struct qib_swqe *s_wq;  /* send work queue */
+	struct qib_swqe *s_wqe;
+	struct qib_rq r_rq;             /* receive work queue */
+	struct qib_sge r_sg_list[0];    /* verified SGEs */
+};
+
+/*
+ * Atomic bit definitions for r_aflags.
+ */
+#define QIB_R_WRID_VALID        0
+#define QIB_R_REWIND_SGE        1
+
+/*
+ * Bit definitions for r_flags.
+ */
+#define QIB_R_REUSE_SGE 0x01
+#define QIB_R_RDMAR_SEQ 0x02
+#define QIB_R_RSP_NAK   0x04
+#define QIB_R_RSP_SEND  0x08
+#define QIB_R_COMM_EST  0x10
+
+/*
+ * Bit definitions for s_flags.
+ *
+ * QIB_S_SIGNAL_REQ_WR - set if QP send WRs contain completion signaled
+ * QIB_S_BUSY - send tasklet is processing the QP
+ * QIB_S_TIMER - the RC retry timer is active
+ * QIB_S_ACK_PENDING - an ACK is waiting to be sent after RDMA read/atomics
+ * QIB_S_WAIT_FENCE - waiting for all prior RDMA read or atomic SWQEs
+ *                         before processing the next SWQE
+ * QIB_S_WAIT_RDMAR - waiting for a RDMA read or atomic SWQE to complete
+ *                         before processing the next SWQE
+ * QIB_S_WAIT_RNR - waiting for RNR timeout
+ * QIB_S_WAIT_SSN_CREDIT - waiting for RC credits to process next SWQE
+ * QIB_S_WAIT_DMA - waiting for send DMA queue to drain before generating
+ *                  next send completion entry not via send DMA
+ * QIB_S_WAIT_PIO - waiting for a send buffer to be available
+ * QIB_S_WAIT_TX - waiting for a struct qib_verbs_txreq to be available
+ * QIB_S_WAIT_DMA_DESC - waiting for DMA descriptors to be available
+ * QIB_S_WAIT_KMEM - waiting for kernel memory to be available
+ * QIB_S_WAIT_PSN - waiting for a packet to exit the send DMA queue
+ * QIB_S_WAIT_ACK - waiting for an ACK packet before sending more requests
+ * QIB_S_SEND_ONE - send one packet, request ACK, then wait for ACK
+ */
+#define QIB_S_SIGNAL_REQ_WR	0x0001
+#define QIB_S_BUSY		0x0002
+#define QIB_S_TIMER		0x0004
+#define QIB_S_RESP_PENDING	0x0008
+#define QIB_S_ACK_PENDING	0x0010
+#define QIB_S_WAIT_FENCE	0x0020
+#define QIB_S_WAIT_RDMAR	0x0040
+#define QIB_S_WAIT_RNR		0x0080
+#define QIB_S_WAIT_SSN_CREDIT	0x0100
+#define QIB_S_WAIT_DMA		0x0200
+#define QIB_S_WAIT_PIO		0x0400
+#define QIB_S_WAIT_TX		0x0800
+#define QIB_S_WAIT_DMA_DESC	0x1000
+#define QIB_S_WAIT_KMEM		0x2000
+#define QIB_S_WAIT_PSN		0x4000
+#define QIB_S_WAIT_ACK		0x8000
+#define QIB_S_SEND_ONE		0x10000
+#define QIB_S_UNLIMITED_CREDIT	0x20000
+
+/*
+ * Wait flags that would prevent any packet type from being sent.
+ */
+#define QIB_S_ANY_WAIT_IO (QIB_S_WAIT_PIO | QIB_S_WAIT_TX | \
+	QIB_S_WAIT_DMA_DESC | QIB_S_WAIT_KMEM)
+
+/*
+ * Wait flags that would prevent send work requests from making progress.
+ */
+#define QIB_S_ANY_WAIT_SEND (QIB_S_WAIT_FENCE | QIB_S_WAIT_RDMAR | \
+	QIB_S_WAIT_RNR | QIB_S_WAIT_SSN_CREDIT | QIB_S_WAIT_DMA | \
+	QIB_S_WAIT_PSN | QIB_S_WAIT_ACK)
+
+#define QIB_S_ANY_WAIT (QIB_S_ANY_WAIT_IO | QIB_S_ANY_WAIT_SEND)
+
+#define QIB_PSN_CREDIT  16
+
+/*
+ * Since struct qib_swqe is not a fixed size, we can't simply index into
+ * struct qib_qp.s_wq.  This function does the array index computation.
+ */
+static inline struct qib_swqe *get_swqe_ptr(struct qib_qp *qp,
+					      unsigned n)
+{
+	return (struct qib_swqe *)((char *)qp->s_wq +
+				     (sizeof(struct qib_swqe) +
+				      qp->s_max_sge *
+				      sizeof(struct qib_sge)) * n);
+}
+
+/*
+ * Since struct qib_rwqe is not a fixed size, we can't simply index into
+ * struct qib_rwq.wq.  This function does the array index computation.
+ */
+static inline struct qib_rwqe *get_rwqe_ptr(struct qib_rq *rq, unsigned n)
+{
+	return (struct qib_rwqe *)
+		((char *) rq->wq->wq +
+		 (sizeof(struct qib_rwqe) +
+		  rq->max_sge * sizeof(struct ib_sge)) * n);
+}
+
+/*
+ * QPN-map pages start out as NULL, they get allocated upon
+ * first use and are never deallocated. This way,
+ * large bitmaps are not allocated unless large numbers of QPs are used.
+ */
+struct qpn_map {
+	void *page;
+};
+
+struct qib_qpn_table {
+	spinlock_t lock; /* protect changes in this struct */
+	unsigned flags;         /* flags for QP0/1 allocated for each port */
+	u32 last;               /* last QP number allocated */
+	u32 nmaps;              /* size of the map table */
+	u16 limit;
+	u16 mask;
+	/* bit map of free QP numbers other than 0/1 */
+	struct qpn_map map[QPNMAP_ENTRIES];
+};
+
+struct qib_lkey_table {
+	spinlock_t lock; /* protect changes in this struct */
+	u32 next;               /* next unused index (speeds search) */
+	u32 gen;                /* generation count */
+	u32 max;                /* size of the table */
+	struct qib_mregion **table;
+};
+
+struct qib_opcode_stats {
+	u64 n_packets;          /* number of packets */
+	u64 n_bytes;            /* total number of bytes */
+};
+
+struct qib_ibport {
+	struct qib_qp *qp0;
+	struct qib_qp *qp1;
+	struct ib_mad_agent *send_agent;	/* agent for SMI (traps) */
+	struct qib_ah *sm_ah;
+	struct qib_ah *smi_ah;
+	struct rb_root mcast_tree;
+	spinlock_t lock;		/* protect changes in this struct */
+
+	/* non-zero when timer is set */
+	unsigned long mkey_lease_timeout;
+	unsigned long trap_timeout;
+	__be64 gid_prefix;      /* in network order */
+	__be64 mkey;
+	__be64 guids[QIB_GUIDS_PER_PORT	- 1];	/* writable GUIDs */
+	u64 tid;		/* TID for traps */
+	u64 n_unicast_xmit;     /* total unicast packets sent */
+	u64 n_unicast_rcv;      /* total unicast packets received */
+	u64 n_multicast_xmit;   /* total multicast packets sent */
+	u64 n_multicast_rcv;    /* total multicast packets received */
+	u64 z_symbol_error_counter;             /* starting count for PMA */
+	u64 z_link_error_recovery_counter;      /* starting count for PMA */
+	u64 z_link_downed_counter;              /* starting count for PMA */
+	u64 z_port_rcv_errors;                  /* starting count for PMA */
+	u64 z_port_rcv_remphys_errors;          /* starting count for PMA */
+	u64 z_port_xmit_discards;               /* starting count for PMA */
+	u64 z_port_xmit_data;                   /* starting count for PMA */
+	u64 z_port_rcv_data;                    /* starting count for PMA */
+	u64 z_port_xmit_packets;                /* starting count for PMA */
+	u64 z_port_rcv_packets;                 /* starting count for PMA */
+	u32 z_local_link_integrity_errors;      /* starting count for PMA */
+	u32 z_excessive_buffer_overrun_errors;  /* starting count for PMA */
+	u32 z_vl15_dropped;                     /* starting count for PMA */
+	u32 n_rc_resends;
+	u32 n_rc_acks;
+	u32 n_rc_qacks;
+	u32 n_rc_delayed_comp;
+	u32 n_seq_naks;
+	u32 n_rdma_seq;
+	u32 n_rnr_naks;
+	u32 n_other_naks;
+	u32 n_loop_pkts;
+	u32 n_pkt_drops;
+	u32 n_vl15_dropped;
+	u32 n_rc_timeouts;
+	u32 n_dmawait;
+	u32 n_unaligned;
+	u32 n_rc_dupreq;
+	u32 n_rc_seqnak;
+	u32 port_cap_flags;
+	u32 pma_sample_start;
+	u32 pma_sample_interval;
+	__be16 pma_counter_select[5];
+	u16 pma_tag;
+	u16 pkey_violations;
+	u16 qkey_violations;
+	u16 mkey_violations;
+	u16 mkey_lease_period;
+	u16 sm_lid;
+	u16 repress_traps;
+	u8 sm_sl;
+	u8 mkeyprot;
+	u8 subnet_timeout;
+	u8 vl_high_limit;
+	u8 sl_to_vl[16];
+
+	struct qib_opcode_stats opstats[128];
+};
+
+struct qib_ibdev {
+	struct ib_device ibdev;
+	struct list_head pending_mmaps;
+	spinlock_t mmap_offset_lock; /* protect mmap_offset */
+	u32 mmap_offset;
+	struct qib_mregion *dma_mr;
+
+	/* QP numbers are shared by all IB ports */
+	struct qib_qpn_table qpn_table;
+	struct qib_lkey_table lk_table;
+	struct list_head piowait;       /* list for wait PIO buf */
+	struct list_head dmawait;	/* list for wait DMA */
+	struct list_head txwait;        /* list for wait qib_verbs_txreq */
+	struct list_head memwait;       /* list for wait kernel memory */
+	struct list_head txreq_free;
+	struct timer_list mem_timer;
+	struct qib_qp **qp_table;
+	struct qib_pio_header *pio_hdrs;
+	dma_addr_t pio_hdrs_phys;
+	/* list of QPs waiting for RNR timer */
+	spinlock_t pending_lock; /* protect wait lists, PMA counters, etc. */
+	unsigned qp_table_size; /* size of the hash table */
+	spinlock_t qpt_lock;
+
+	u32 n_piowait;
+	u32 n_txwait;
+
+	u32 n_pds_allocated;    /* number of PDs allocated for device */
+	spinlock_t n_pds_lock;
+	u32 n_ahs_allocated;    /* number of AHs allocated for device */
+	spinlock_t n_ahs_lock;
+	u32 n_cqs_allocated;    /* number of CQs allocated for device */
+	spinlock_t n_cqs_lock;
+	u32 n_qps_allocated;    /* number of QPs allocated for device */
+	spinlock_t n_qps_lock;
+	u32 n_srqs_allocated;   /* number of SRQs allocated for device */
+	spinlock_t n_srqs_lock;
+	u32 n_mcast_grps_allocated; /* number of mcast groups allocated */
+	spinlock_t n_mcast_grps_lock;
+};
+
+struct qib_verbs_counters {
+	u64 symbol_error_counter;
+	u64 link_error_recovery_counter;
+	u64 link_downed_counter;
+	u64 port_rcv_errors;
+	u64 port_rcv_remphys_errors;
+	u64 port_xmit_discards;
+	u64 port_xmit_data;
+	u64 port_rcv_data;
+	u64 port_xmit_packets;
+	u64 port_rcv_packets;
+	u32 local_link_integrity_errors;
+	u32 excessive_buffer_overrun_errors;
+	u32 vl15_dropped;
+};
+
+static inline struct qib_mr *to_imr(struct ib_mr *ibmr)
+{
+	return container_of(ibmr, struct qib_mr, ibmr);
+}
+
+static inline struct qib_pd *to_ipd(struct ib_pd *ibpd)
+{
+	return container_of(ibpd, struct qib_pd, ibpd);
+}
+
+static inline struct qib_ah *to_iah(struct ib_ah *ibah)
+{
+	return container_of(ibah, struct qib_ah, ibah);
+}
+
+static inline struct qib_cq *to_icq(struct ib_cq *ibcq)
+{
+	return container_of(ibcq, struct qib_cq, ibcq);
+}
+
+static inline struct qib_srq *to_isrq(struct ib_srq *ibsrq)
+{
+	return container_of(ibsrq, struct qib_srq, ibsrq);
+}
+
+static inline struct qib_qp *to_iqp(struct ib_qp *ibqp)
+{
+	return container_of(ibqp, struct qib_qp, ibqp);
+}
+
+static inline struct qib_ibdev *to_idev(struct ib_device *ibdev)
+{
+	return container_of(ibdev, struct qib_ibdev, ibdev);
+}
+
+/*
+ * Send if not busy or waiting for I/O and either
+ * a RC response is pending or we can process send work requests.
+ */
+static inline int qib_send_ok(struct qib_qp *qp)
+{
+	return !(qp->s_flags & (QIB_S_BUSY | QIB_S_ANY_WAIT_IO)) &&
+		(qp->s_hdrwords || (qp->s_flags & QIB_S_RESP_PENDING) ||
+		 !(qp->s_flags & QIB_S_ANY_WAIT_SEND));
+}
+
+extern struct workqueue_struct *qib_wq;
+extern struct workqueue_struct *qib_cq_wq;
+
+/*
+ * This must be called with s_lock held.
+ */
+static inline void qib_schedule_send(struct qib_qp *qp)
+{
+	if (qib_send_ok(qp)) {
+		if (qp->processor_id == smp_processor_id())
+			queue_work(qib_wq, &qp->s_work);
+		else
+			queue_work_on(qp->processor_id,
+				      qib_wq, &qp->s_work);
+	}
+}
+
+static inline int qib_pkey_ok(u16 pkey1, u16 pkey2)
+{
+	u16 p1 = pkey1 & 0x7FFF;
+	u16 p2 = pkey2 & 0x7FFF;
+
+	/*
+	 * Low 15 bits must be non-zero and match, and
+	 * one of the two must be a full member.
+	 */
+	return p1 && p1 == p2 && ((__s16)pkey1 < 0 || (__s16)pkey2 < 0);
+}
+
+void qib_bad_pqkey(struct qib_ibport *ibp, __be16 trap_num, u32 key, u32 sl,
+		   u32 qp1, u32 qp2, __be16 lid1, __be16 lid2);
+void qib_cap_mask_chg(struct qib_ibport *ibp);
+void qib_sys_guid_chg(struct qib_ibport *ibp);
+void qib_node_desc_chg(struct qib_ibport *ibp);
+int qib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
+		    struct ib_wc *in_wc, struct ib_grh *in_grh,
+		    struct ib_mad *in_mad, struct ib_mad *out_mad);
+int qib_create_agents(struct qib_ibdev *dev);
+void qib_free_agents(struct qib_ibdev *dev);
+
+/*
+ * Compare the lower 24 bits of the two values.
+ * Returns an integer <, ==, or > than zero.
+ */
+static inline int qib_cmp24(u32 a, u32 b)
+{
+	return (((int) a) - ((int) b)) << 8;
+}
+
+struct qib_mcast *qib_mcast_find(struct qib_ibport *ibp, union ib_gid *mgid);
+
+int qib_snapshot_counters(struct qib_pportdata *ppd, u64 *swords,
+			  u64 *rwords, u64 *spkts, u64 *rpkts,
+			  u64 *xmit_wait);
+
+int qib_get_counters(struct qib_pportdata *ppd,
+		     struct qib_verbs_counters *cntrs);
+
+int qib_multicast_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid);
+
+int qib_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid);
+
+int qib_mcast_tree_empty(struct qib_ibport *ibp);
+
+__be32 qib_compute_aeth(struct qib_qp *qp);
+
+struct qib_qp *qib_lookup_qpn(struct qib_ibport *ibp, u32 qpn);
+
+struct ib_qp *qib_create_qp(struct ib_pd *ibpd,
+			    struct ib_qp_init_attr *init_attr,
+			    struct ib_udata *udata);
+
+int qib_destroy_qp(struct ib_qp *ibqp);
+
+int qib_error_qp(struct qib_qp *qp, enum ib_wc_status err);
+
+int qib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
+		  int attr_mask, struct ib_udata *udata);
+
+int qib_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
+		 int attr_mask, struct ib_qp_init_attr *init_attr);
+
+unsigned qib_free_all_qps(struct qib_devdata *dd);
+
+void qib_init_qpn_table(struct qib_devdata *dd, struct qib_qpn_table *qpt);
+
+void qib_free_qpn_table(struct qib_qpn_table *qpt);
+
+void qib_get_credit(struct qib_qp *qp, u32 aeth);
+
+unsigned qib_pkt_delay(u32 plen, u8 snd_mult, u8 rcv_mult);
+
+void qib_verbs_sdma_desc_avail(struct qib_pportdata *ppd, unsigned avail);
+
+void qib_put_txreq(struct qib_verbs_txreq *tx);
+
+int qib_verbs_send(struct qib_qp *qp, struct qib_ib_header *hdr,
+		   u32 hdrwords, struct qib_sge_state *ss, u32 len);
+
+void qib_copy_sge(struct qib_sge_state *ss, void *data, u32 length,
+		  int release);
+
+void qib_skip_sge(struct qib_sge_state *ss, u32 length, int release);
+
+void qib_uc_rcv(struct qib_ibport *ibp, struct qib_ib_header *hdr,
+		int has_grh, void *data, u32 tlen, struct qib_qp *qp);
+
+void qib_rc_rcv(struct qib_ctxtdata *rcd, struct qib_ib_header *hdr,
+		int has_grh, void *data, u32 tlen, struct qib_qp *qp);
+
+int qib_check_ah(struct ib_device *ibdev, struct ib_ah_attr *ah_attr);
+
+void qib_rc_rnr_retry(unsigned long arg);
+
+void qib_rc_send_complete(struct qib_qp *qp, struct qib_ib_header *hdr);
+
+void qib_rc_error(struct qib_qp *qp, enum ib_wc_status err);
+
+int qib_post_ud_send(struct qib_qp *qp, struct ib_send_wr *wr);
+
+void qib_ud_rcv(struct qib_ibport *ibp, struct qib_ib_header *hdr,
+		int has_grh, void *data, u32 tlen, struct qib_qp *qp);
+
+int qib_alloc_lkey(struct qib_lkey_table *rkt, struct qib_mregion *mr);
+
+int qib_free_lkey(struct qib_ibdev *dev, struct qib_mregion *mr);
+
+int qib_lkey_ok(struct qib_lkey_table *rkt, struct qib_pd *pd,
+		struct qib_sge *isge, struct ib_sge *sge, int acc);
+
+int qib_rkey_ok(struct qib_qp *qp, struct qib_sge *sge,
+		u32 len, u64 vaddr, u32 rkey, int acc);
+
+int qib_post_srq_receive(struct ib_srq *ibsrq, struct ib_recv_wr *wr,
+			 struct ib_recv_wr **bad_wr);
+
+struct ib_srq *qib_create_srq(struct ib_pd *ibpd,
+			      struct ib_srq_init_attr *srq_init_attr,
+			      struct ib_udata *udata);
+
+int qib_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
+		   enum ib_srq_attr_mask attr_mask,
+		   struct ib_udata *udata);
+
+int qib_query_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr);
+
+int qib_destroy_srq(struct ib_srq *ibsrq);
+
+void qib_cq_enter(struct qib_cq *cq, struct ib_wc *entry, int sig);
+
+int qib_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry);
+
+struct ib_cq *qib_create_cq(struct ib_device *ibdev, int entries,
+			    int comp_vector, struct ib_ucontext *context,
+			    struct ib_udata *udata);
+
+int qib_destroy_cq(struct ib_cq *ibcq);
+
+int qib_req_notify_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags notify_flags);
+
+int qib_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata);
+
+struct ib_mr *qib_get_dma_mr(struct ib_pd *pd, int acc);
+
+struct ib_mr *qib_reg_phys_mr(struct ib_pd *pd,
+			      struct ib_phys_buf *buffer_list,
+			      int num_phys_buf, int acc, u64 *iova_start);
+
+struct ib_mr *qib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
+			      u64 virt_addr, int mr_access_flags,
+			      struct ib_udata *udata);
+
+int qib_dereg_mr(struct ib_mr *ibmr);
+
+struct ib_mr *qib_alloc_fast_reg_mr(struct ib_pd *pd, int max_page_list_len);
+
+struct ib_fast_reg_page_list *qib_alloc_fast_reg_page_list(
+				struct ib_device *ibdev, int page_list_len);
+
+void qib_free_fast_reg_page_list(struct ib_fast_reg_page_list *pl);
+
+int qib_fast_reg_mr(struct qib_qp *qp, struct ib_send_wr *wr);
+
+struct ib_fmr *qib_alloc_fmr(struct ib_pd *pd, int mr_access_flags,
+			     struct ib_fmr_attr *fmr_attr);
+
+int qib_map_phys_fmr(struct ib_fmr *ibfmr, u64 *page_list,
+		     int list_len, u64 iova);
+
+int qib_unmap_fmr(struct list_head *fmr_list);
+
+int qib_dealloc_fmr(struct ib_fmr *ibfmr);
+
+void qib_release_mmap_info(struct kref *ref);
+
+struct qib_mmap_info *qib_create_mmap_info(struct qib_ibdev *dev, u32 size,
+					   struct ib_ucontext *context,
+					   void *obj);
+
+void qib_update_mmap_info(struct qib_ibdev *dev, struct qib_mmap_info *ip,
+			  u32 size, void *obj);
+
+int qib_mmap(struct ib_ucontext *context, struct vm_area_struct *vma);
+
+int qib_get_rwqe(struct qib_qp *qp, int wr_id_only);
+
+void qib_migrate_qp(struct qib_qp *qp);
+
+int qib_ruc_check_hdr(struct qib_ibport *ibp, struct qib_ib_header *hdr,
+		      int has_grh, struct qib_qp *qp, u32 bth0);
+
+u32 qib_make_grh(struct qib_ibport *ibp, struct ib_grh *hdr,
+		 struct ib_global_route *grh, u32 hwords, u32 nwords);
+
+void qib_make_ruc_header(struct qib_qp *qp, struct qib_other_headers *ohdr,
+			 u32 bth0, u32 bth2);
+
+void qib_do_send(struct work_struct *work);
+
+void qib_send_complete(struct qib_qp *qp, struct qib_swqe *wqe,
+		       enum ib_wc_status status);
+
+void qib_send_rc_ack(struct qib_qp *qp);
+
+int qib_make_rc_req(struct qib_qp *qp);
+
+int qib_make_uc_req(struct qib_qp *qp);
+
+int qib_make_ud_req(struct qib_qp *qp);
+
+int qib_register_ib_device(struct qib_devdata *);
+
+void qib_unregister_ib_device(struct qib_devdata *);
+
+void qib_ib_rcv(struct qib_ctxtdata *, void *, void *, u32);
+
+void qib_ib_piobufavail(struct qib_devdata *);
+
+unsigned qib_get_npkeys(struct qib_devdata *);
+
+unsigned qib_get_pkey(struct qib_ibport *, unsigned);
+
+extern const enum ib_wc_opcode ib_qib_wc_opcode[];
+
+/*
+ * Below  HCA-independent IB PhysPortState values, returned
+ * by the f_ibphys_portstate() routine.
+ */
+#define IB_PHYSPORTSTATE_SLEEP 1
+#define IB_PHYSPORTSTATE_POLL 2
+#define IB_PHYSPORTSTATE_DISABLED 3
+#define IB_PHYSPORTSTATE_CFG_TRAIN 4
+#define IB_PHYSPORTSTATE_LINKUP 5
+#define IB_PHYSPORTSTATE_LINK_ERR_RECOVER 6
+#define IB_PHYSPORTSTATE_CFG_DEBOUNCE 8
+#define IB_PHYSPORTSTATE_CFG_IDLE 0xB
+#define IB_PHYSPORTSTATE_RECOVERY_RETRAIN 0xC
+#define IB_PHYSPORTSTATE_RECOVERY_WAITRMT 0xE
+#define IB_PHYSPORTSTATE_RECOVERY_IDLE 0xF
+#define IB_PHYSPORTSTATE_CFG_ENH 0x10
+#define IB_PHYSPORTSTATE_CFG_WAIT_ENH 0x13
+
+extern const int ib_qib_state_ops[];
+
+extern __be64 ib_qib_sys_image_guid;    /* in network order */
+
+extern unsigned int ib_qib_lkey_table_size;
+
+extern unsigned int ib_qib_max_cqes;
+
+extern unsigned int ib_qib_max_cqs;
+
+extern unsigned int ib_qib_max_qp_wrs;
+
+extern unsigned int ib_qib_max_qps;
+
+extern unsigned int ib_qib_max_sges;
+
+extern unsigned int ib_qib_max_mcast_grps;
+
+extern unsigned int ib_qib_max_mcast_qp_attached;
+
+extern unsigned int ib_qib_max_srqs;
+
+extern unsigned int ib_qib_max_srq_sges;
+
+extern unsigned int ib_qib_max_srq_wrs;
+
+extern const u32 ib_qib_rnr_table[];
+
+extern struct ib_dma_mapping_ops qib_dma_mapping_ops;
+
+#endif                          /* QIB_VERBS_H */
diff --git a/drivers/infiniband/hw/qib/qib_verbs_mcast.c b/drivers/infiniband/hw/qib/qib_verbs_mcast.c
new file mode 100644
index 000000000000..dabb697b1c2a
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_verbs_mcast.c
@@ -0,0 +1,368 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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 <linux/rculist.h>
+
+#include "qib.h"
+
+/**
+ * qib_mcast_qp_alloc - alloc a struct to link a QP to mcast GID struct
+ * @qp: the QP to link
+ */
+static struct qib_mcast_qp *qib_mcast_qp_alloc(struct qib_qp *qp)
+{
+	struct qib_mcast_qp *mqp;
+
+	mqp = kmalloc(sizeof *mqp, GFP_KERNEL);
+	if (!mqp)
+		goto bail;
+
+	mqp->qp = qp;
+	atomic_inc(&qp->refcount);
+
+bail:
+	return mqp;
+}
+
+static void qib_mcast_qp_free(struct qib_mcast_qp *mqp)
+{
+	struct qib_qp *qp = mqp->qp;
+
+	/* Notify qib_destroy_qp() if it is waiting. */
+	if (atomic_dec_and_test(&qp->refcount))
+		wake_up(&qp->wait);
+
+	kfree(mqp);
+}
+
+/**
+ * qib_mcast_alloc - allocate the multicast GID structure
+ * @mgid: the multicast GID
+ *
+ * A list of QPs will be attached to this structure.
+ */
+static struct qib_mcast *qib_mcast_alloc(union ib_gid *mgid)
+{
+	struct qib_mcast *mcast;
+
+	mcast = kmalloc(sizeof *mcast, GFP_KERNEL);
+	if (!mcast)
+		goto bail;
+
+	mcast->mgid = *mgid;
+	INIT_LIST_HEAD(&mcast->qp_list);
+	init_waitqueue_head(&mcast->wait);
+	atomic_set(&mcast->refcount, 0);
+	mcast->n_attached = 0;
+
+bail:
+	return mcast;
+}
+
+static void qib_mcast_free(struct qib_mcast *mcast)
+{
+	struct qib_mcast_qp *p, *tmp;
+
+	list_for_each_entry_safe(p, tmp, &mcast->qp_list, list)
+		qib_mcast_qp_free(p);
+
+	kfree(mcast);
+}
+
+/**
+ * qib_mcast_find - search the global table for the given multicast GID
+ * @ibp: the IB port structure
+ * @mgid: the multicast GID to search for
+ *
+ * Returns NULL if not found.
+ *
+ * The caller is responsible for decrementing the reference count if found.
+ */
+struct qib_mcast *qib_mcast_find(struct qib_ibport *ibp, union ib_gid *mgid)
+{
+	struct rb_node *n;
+	unsigned long flags;
+	struct qib_mcast *mcast;
+
+	spin_lock_irqsave(&ibp->lock, flags);
+	n = ibp->mcast_tree.rb_node;
+	while (n) {
+		int ret;
+
+		mcast = rb_entry(n, struct qib_mcast, rb_node);
+
+		ret = memcmp(mgid->raw, mcast->mgid.raw,
+			     sizeof(union ib_gid));
+		if (ret < 0)
+			n = n->rb_left;
+		else if (ret > 0)
+			n = n->rb_right;
+		else {
+			atomic_inc(&mcast->refcount);
+			spin_unlock_irqrestore(&ibp->lock, flags);
+			goto bail;
+		}
+	}
+	spin_unlock_irqrestore(&ibp->lock, flags);
+
+	mcast = NULL;
+
+bail:
+	return mcast;
+}
+
+/**
+ * qib_mcast_add - insert mcast GID into table and attach QP struct
+ * @mcast: the mcast GID table
+ * @mqp: the QP to attach
+ *
+ * Return zero if both were added.  Return EEXIST if the GID was already in
+ * the table but the QP was added.  Return ESRCH if the QP was already
+ * attached and neither structure was added.
+ */
+static int qib_mcast_add(struct qib_ibdev *dev, struct qib_ibport *ibp,
+			 struct qib_mcast *mcast, struct qib_mcast_qp *mqp)
+{
+	struct rb_node **n = &ibp->mcast_tree.rb_node;
+	struct rb_node *pn = NULL;
+	int ret;
+
+	spin_lock_irq(&ibp->lock);
+
+	while (*n) {
+		struct qib_mcast *tmcast;
+		struct qib_mcast_qp *p;
+
+		pn = *n;
+		tmcast = rb_entry(pn, struct qib_mcast, rb_node);
+
+		ret = memcmp(mcast->mgid.raw, tmcast->mgid.raw,
+			     sizeof(union ib_gid));
+		if (ret < 0) {
+			n = &pn->rb_left;
+			continue;
+		}
+		if (ret > 0) {
+			n = &pn->rb_right;
+			continue;
+		}
+
+		/* Search the QP list to see if this is already there. */
+		list_for_each_entry_rcu(p, &tmcast->qp_list, list) {
+			if (p->qp == mqp->qp) {
+				ret = ESRCH;
+				goto bail;
+			}
+		}
+		if (tmcast->n_attached == ib_qib_max_mcast_qp_attached) {
+			ret = ENOMEM;
+			goto bail;
+		}
+
+		tmcast->n_attached++;
+
+		list_add_tail_rcu(&mqp->list, &tmcast->qp_list);
+		ret = EEXIST;
+		goto bail;
+	}
+
+	spin_lock(&dev->n_mcast_grps_lock);
+	if (dev->n_mcast_grps_allocated == ib_qib_max_mcast_grps) {
+		spin_unlock(&dev->n_mcast_grps_lock);
+		ret = ENOMEM;
+		goto bail;
+	}
+
+	dev->n_mcast_grps_allocated++;
+	spin_unlock(&dev->n_mcast_grps_lock);
+
+	mcast->n_attached++;
+
+	list_add_tail_rcu(&mqp->list, &mcast->qp_list);
+
+	atomic_inc(&mcast->refcount);
+	rb_link_node(&mcast->rb_node, pn, n);
+	rb_insert_color(&mcast->rb_node, &ibp->mcast_tree);
+
+	ret = 0;
+
+bail:
+	spin_unlock_irq(&ibp->lock);
+
+	return ret;
+}
+
+int qib_multicast_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
+{
+	struct qib_qp *qp = to_iqp(ibqp);
+	struct qib_ibdev *dev = to_idev(ibqp->device);
+	struct qib_ibport *ibp;
+	struct qib_mcast *mcast;
+	struct qib_mcast_qp *mqp;
+	int ret;
+
+	if (ibqp->qp_num <= 1 || qp->state == IB_QPS_RESET) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	/*
+	 * Allocate data structures since its better to do this outside of
+	 * spin locks and it will most likely be needed.
+	 */
+	mcast = qib_mcast_alloc(gid);
+	if (mcast == NULL) {
+		ret = -ENOMEM;
+		goto bail;
+	}
+	mqp = qib_mcast_qp_alloc(qp);
+	if (mqp == NULL) {
+		qib_mcast_free(mcast);
+		ret = -ENOMEM;
+		goto bail;
+	}
+	ibp = to_iport(ibqp->device, qp->port_num);
+	switch (qib_mcast_add(dev, ibp, mcast, mqp)) {
+	case ESRCH:
+		/* Neither was used: OK to attach the same QP twice. */
+		qib_mcast_qp_free(mqp);
+		qib_mcast_free(mcast);
+		break;
+
+	case EEXIST:            /* The mcast wasn't used */
+		qib_mcast_free(mcast);
+		break;
+
+	case ENOMEM:
+		/* Exceeded the maximum number of mcast groups. */
+		qib_mcast_qp_free(mqp);
+		qib_mcast_free(mcast);
+		ret = -ENOMEM;
+		goto bail;
+
+	default:
+		break;
+	}
+
+	ret = 0;
+
+bail:
+	return ret;
+}
+
+int qib_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
+{
+	struct qib_qp *qp = to_iqp(ibqp);
+	struct qib_ibdev *dev = to_idev(ibqp->device);
+	struct qib_ibport *ibp = to_iport(ibqp->device, qp->port_num);
+	struct qib_mcast *mcast = NULL;
+	struct qib_mcast_qp *p, *tmp;
+	struct rb_node *n;
+	int last = 0;
+	int ret;
+
+	if (ibqp->qp_num <= 1 || qp->state == IB_QPS_RESET) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	spin_lock_irq(&ibp->lock);
+
+	/* Find the GID in the mcast table. */
+	n = ibp->mcast_tree.rb_node;
+	while (1) {
+		if (n == NULL) {
+			spin_unlock_irq(&ibp->lock);
+			ret = -EINVAL;
+			goto bail;
+		}
+
+		mcast = rb_entry(n, struct qib_mcast, rb_node);
+		ret = memcmp(gid->raw, mcast->mgid.raw,
+			     sizeof(union ib_gid));
+		if (ret < 0)
+			n = n->rb_left;
+		else if (ret > 0)
+			n = n->rb_right;
+		else
+			break;
+	}
+
+	/* Search the QP list. */
+	list_for_each_entry_safe(p, tmp, &mcast->qp_list, list) {
+		if (p->qp != qp)
+			continue;
+		/*
+		 * We found it, so remove it, but don't poison the forward
+		 * link until we are sure there are no list walkers.
+		 */
+		list_del_rcu(&p->list);
+		mcast->n_attached--;
+
+		/* If this was the last attached QP, remove the GID too. */
+		if (list_empty(&mcast->qp_list)) {
+			rb_erase(&mcast->rb_node, &ibp->mcast_tree);
+			last = 1;
+		}
+		break;
+	}
+
+	spin_unlock_irq(&ibp->lock);
+
+	if (p) {
+		/*
+		 * Wait for any list walkers to finish before freeing the
+		 * list element.
+		 */
+		wait_event(mcast->wait, atomic_read(&mcast->refcount) <= 1);
+		qib_mcast_qp_free(p);
+	}
+	if (last) {
+		atomic_dec(&mcast->refcount);
+		wait_event(mcast->wait, !atomic_read(&mcast->refcount));
+		qib_mcast_free(mcast);
+		spin_lock_irq(&dev->n_mcast_grps_lock);
+		dev->n_mcast_grps_allocated--;
+		spin_unlock_irq(&dev->n_mcast_grps_lock);
+	}
+
+	ret = 0;
+
+bail:
+	return ret;
+}
+
+int qib_mcast_tree_empty(struct qib_ibport *ibp)
+{
+	return ibp->mcast_tree.rb_node == NULL;
+}
diff --git a/drivers/infiniband/hw/ipath/ipath_7220.h b/drivers/infiniband/hw/qib/qib_wc_ppc64.c
index 74fa5cc5131d..673cf4c22ebd 100644
--- a/drivers/infiniband/hw/ipath/ipath_7220.h
+++ b/drivers/infiniband/hw/qib/qib_wc_ppc64.c
@@ -1,7 +1,5 @@
-#ifndef _IPATH_7220_H
-#define _IPATH_7220_H
 /*
- * Copyright (c) 2007 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2006, 2007, 2008 QLogic Corporation. All rights reserved.
  *
  * This software is available to you under a choice of one of two
  * licenses.  You may choose to be licensed under the terms of the GNU
@@ -33,25 +31,32 @@
  */
 
 /*
- * This header file provides the declarations and common definitions
- * for (mostly) manipulation of the SerDes blocks within the IBA7220.
- * the functions declared should only be called from within other
- * 7220-related files such as ipath_iba7220.c or ipath_sd7220.c.
+ * This file is conditionally built on PowerPC only.  Otherwise weak symbol
+ * versions of the functions exported from here are used.
  */
-int ipath_sd7220_presets(struct ipath_devdata *dd);
-int ipath_sd7220_init(struct ipath_devdata *dd, int was_reset);
-int ipath_sd7220_prog_ld(struct ipath_devdata *dd, int sdnum, u8 *img,
-	int len, int offset);
-int ipath_sd7220_prog_vfy(struct ipath_devdata *dd, int sdnum, const u8 *img,
-	int len, int offset);
-/*
- * Below used for sdnum parameter, selecting one of the two sections
- * used for PCIe, or the single SerDes used for IB, which is the
- * only one currently used
- */
-#define IB_7220_SERDES 2
 
-int ipath_sd7220_ib_load(struct ipath_devdata *dd);
-int ipath_sd7220_ib_vfy(struct ipath_devdata *dd);
+#include "qib.h"
 
-#endif /* _IPATH_7220_H */
+/**
+ * qib_enable_wc - enable write combining for MMIO writes to the device
+ * @dd: qlogic_ib device
+ *
+ * Nothing to do on PowerPC, so just return without error.
+ */
+int qib_enable_wc(struct qib_devdata *dd)
+{
+	return 0;
+}
+
+/**
+ * qib_unordered_wc - indicate whether write combining is unordered
+ *
+ * Because our performance depends on our ability to do write
+ * combining mmio writes in the most efficient way, we need to
+ * know if we are on a processor that may reorder stores when
+ * write combining.
+ */
+int qib_unordered_wc(void)
+{
+	return 1;
+}
diff --git a/drivers/infiniband/hw/qib/qib_wc_x86_64.c b/drivers/infiniband/hw/qib/qib_wc_x86_64.c
new file mode 100644
index 000000000000..561b8bca4060
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_wc_x86_64.c
@@ -0,0 +1,171 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * 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.
+ */
+
+/*
+ * This file is conditionally built on x86_64 only.  Otherwise weak symbol
+ * versions of the functions exported from here are used.
+ */
+
+#include <linux/pci.h>
+#include <asm/mtrr.h>
+#include <asm/processor.h>
+
+#include "qib.h"
+
+/**
+ * qib_enable_wc - enable write combining for MMIO writes to the device
+ * @dd: qlogic_ib device
+ *
+ * This routine is x86_64-specific; it twiddles the CPU's MTRRs to enable
+ * write combining.
+ */
+int qib_enable_wc(struct qib_devdata *dd)
+{
+	int ret = 0;
+	u64 pioaddr, piolen;
+	unsigned bits;
+	const unsigned long addr = pci_resource_start(dd->pcidev, 0);
+	const size_t len = pci_resource_len(dd->pcidev, 0);
+
+	/*
+	 * Set the PIO buffers to be WCCOMB, so we get HT bursts to the
+	 * chip.  Linux (possibly the hardware) requires it to be on a power
+	 * of 2 address matching the length (which has to be a power of 2).
+	 * For rev1, that means the base address, for rev2, it will be just
+	 * the PIO buffers themselves.
+	 * For chips with two sets of buffers, the calculations are
+	 * somewhat more complicated; we need to sum, and the piobufbase
+	 * register has both offsets, 2K in low 32 bits, 4K in high 32 bits.
+	 * The buffers are still packed, so a single range covers both.
+	 */
+	if (dd->piobcnt2k && dd->piobcnt4k) {
+		/* 2 sizes for chip */
+		unsigned long pio2kbase, pio4kbase;
+		pio2kbase = dd->piobufbase & 0xffffffffUL;
+		pio4kbase = (dd->piobufbase >> 32) & 0xffffffffUL;
+		if (pio2kbase < pio4kbase) {
+			/* all current chips */
+			pioaddr = addr + pio2kbase;
+			piolen = pio4kbase - pio2kbase +
+				dd->piobcnt4k * dd->align4k;
+		} else {
+			pioaddr = addr + pio4kbase;
+			piolen = pio2kbase - pio4kbase +
+				dd->piobcnt2k * dd->palign;
+		}
+	} else {  /* single buffer size (2K, currently) */
+		pioaddr = addr + dd->piobufbase;
+		piolen = dd->piobcnt2k * dd->palign +
+			dd->piobcnt4k * dd->align4k;
+	}
+
+	for (bits = 0; !(piolen & (1ULL << bits)); bits++)
+		/* do nothing */ ;
+
+	if (piolen != (1ULL << bits)) {
+		piolen >>= bits;
+		while (piolen >>= 1)
+			bits++;
+		piolen = 1ULL << (bits + 1);
+	}
+	if (pioaddr & (piolen - 1)) {
+		u64 atmp;
+		atmp = pioaddr & ~(piolen - 1);
+		if (atmp < addr || (atmp + piolen) > (addr + len)) {
+			qib_dev_err(dd, "No way to align address/size "
+				    "(%llx/%llx), no WC mtrr\n",
+				    (unsigned long long) atmp,
+				    (unsigned long long) piolen << 1);
+			ret = -ENODEV;
+		} else {
+			pioaddr = atmp;
+			piolen <<= 1;
+		}
+	}
+
+	if (!ret) {
+		int cookie;
+
+		cookie = mtrr_add(pioaddr, piolen, MTRR_TYPE_WRCOMB, 0);
+		if (cookie < 0) {
+			{
+				qib_devinfo(dd->pcidev,
+					 "mtrr_add()  WC for PIO bufs "
+					 "failed (%d)\n",
+					 cookie);
+				ret = -EINVAL;
+			}
+		} else {
+			dd->wc_cookie = cookie;
+			dd->wc_base = (unsigned long) pioaddr;
+			dd->wc_len = (unsigned long) piolen;
+		}
+	}
+
+	return ret;
+}
+
+/**
+ * qib_disable_wc - disable write combining for MMIO writes to the device
+ * @dd: qlogic_ib device
+ */
+void qib_disable_wc(struct qib_devdata *dd)
+{
+	if (dd->wc_cookie) {
+		int r;
+
+		r = mtrr_del(dd->wc_cookie, dd->wc_base,
+			     dd->wc_len);
+		if (r < 0)
+			qib_devinfo(dd->pcidev,
+				 "mtrr_del(%lx, %lx, %lx) failed: %d\n",
+				 dd->wc_cookie, dd->wc_base,
+				 dd->wc_len, r);
+		dd->wc_cookie = 0; /* even on failure */
+	}
+}
+
+/**
+ * qib_unordered_wc - indicate whether write combining is ordered
+ *
+ * Because our performance depends on our ability to do write combining mmio
+ * writes in the most efficient way, we need to know if we are on an Intel
+ * or AMD x86_64 processor.  AMD x86_64 processors flush WC buffers out in
+ * the order completed, and so no special flushing is required to get
+ * correct ordering.  Intel processors, however, will flush write buffers
+ * out in "random" orders, and so explicit ordering is needed at times.
+ */
+int qib_unordered_wc(void)
+{
+	return boot_cpu_data.x86_vendor != X86_VENDOR_AMD;
+}
diff --git a/drivers/input/joydev.c b/drivers/input/joydev.c
index 423e0e6031ab..34157bb97ed6 100644
--- a/drivers/input/joydev.c
+++ b/drivers/input/joydev.c
@@ -47,15 +47,15 @@ struct joydev {
 	struct mutex mutex;
 	struct device dev;
 
-	struct js_corr corr[ABS_MAX + 1];
+	struct js_corr corr[ABS_CNT];
 	struct JS_DATA_SAVE_TYPE glue;
 	int nabs;
 	int nkey;
 	__u16 keymap[KEY_MAX - BTN_MISC + 1];
 	__u16 keypam[KEY_MAX - BTN_MISC + 1];
-	__u8 absmap[ABS_MAX + 1];
-	__u8 abspam[ABS_MAX + 1];
-	__s16 abs[ABS_MAX + 1];
+	__u8 absmap[ABS_CNT];
+	__u8 abspam[ABS_CNT];
+	__s16 abs[ABS_CNT];
 };
 
 struct joydev_client {
@@ -826,7 +826,7 @@ static int joydev_connect(struct input_handler *handler, struct input_dev *dev,
 	joydev->handle.handler = handler;
 	joydev->handle.private = joydev;
 
-	for (i = 0; i < ABS_MAX + 1; i++)
+	for (i = 0; i < ABS_CNT; i++)
 		if (test_bit(i, dev->absbit)) {
 			joydev->absmap[i] = joydev->nabs;
 			joydev->abspam[joydev->nabs] = i;
diff --git a/drivers/input/keyboard/adp5588-keys.c b/drivers/input/keyboard/adp5588-keys.c
index 4771ab172b59..744600eff222 100644
--- a/drivers/input/keyboard/adp5588-keys.c
+++ b/drivers/input/keyboard/adp5588-keys.c
@@ -287,7 +287,6 @@ static int __devexit adp5588_remove(struct i2c_client *client)
 	free_irq(client->irq, kpad);
 	cancel_delayed_work_sync(&kpad->work);
 	input_unregister_device(kpad->input);
-	i2c_set_clientdata(client, NULL);
 	kfree(kpad);
 
 	return 0;
diff --git a/drivers/input/keyboard/amikbd.c b/drivers/input/keyboard/amikbd.c
index 35149ec455a9..79172af164f2 100644
--- a/drivers/input/keyboard/amikbd.c
+++ b/drivers/input/keyboard/amikbd.c
@@ -35,6 +35,7 @@
 #include <linux/delay.h>
 #include <linux/interrupt.h>
 #include <linux/keyboard.h>
+#include <linux/platform_device.h>
 
 #include <asm/amigaints.h>
 #include <asm/amigahw.h>
@@ -154,10 +155,9 @@ static const char *amikbd_messages[8] = {
 	[7] = KERN_WARNING "amikbd: keyboard interrupt\n"
 };
 
-static struct input_dev *amikbd_dev;
-
-static irqreturn_t amikbd_interrupt(int irq, void *dummy)
+static irqreturn_t amikbd_interrupt(int irq, void *data)
 {
+	struct input_dev *dev = data;
 	unsigned char scancode, down;
 
 	scancode = ~ciaa.sdr;		/* get and invert scancode (keyboard is active low) */
@@ -170,47 +170,42 @@ static irqreturn_t amikbd_interrupt(int irq, void *dummy)
 
 	if (scancode < 0x78) {		/* scancodes < 0x78 are keys */
 		if (scancode == 98) {	/* CapsLock is a toggle switch key on Amiga */
-			input_report_key(amikbd_dev, scancode, 1);
-			input_report_key(amikbd_dev, scancode, 0);
+			input_report_key(dev, scancode, 1);
+			input_report_key(dev, scancode, 0);
 		} else {
-			input_report_key(amikbd_dev, scancode, down);
+			input_report_key(dev, scancode, down);
 		}
 
-		input_sync(amikbd_dev);
+		input_sync(dev);
 	} else				/* scancodes >= 0x78 are error codes */
 		printk(amikbd_messages[scancode - 0x78]);
 
 	return IRQ_HANDLED;
 }
 
-static int __init amikbd_init(void)
+static int __init amikbd_probe(struct platform_device *pdev)
 {
+	struct input_dev *dev;
 	int i, j, err;
 
-	if (!AMIGAHW_PRESENT(AMI_KEYBOARD))
-		return -ENODEV;
-
-	if (!request_mem_region(CIAA_PHYSADDR-1+0xb00, 0x100, "amikeyb"))
-		return -EBUSY;
-
-	amikbd_dev = input_allocate_device();
-	if (!amikbd_dev) {
-		printk(KERN_ERR "amikbd: not enough memory for input device\n");
-		err = -ENOMEM;
-		goto fail1;
+	dev = input_allocate_device();
+	if (!dev) {
+		dev_err(&pdev->dev, "Not enough memory for input device\n");
+		return -ENOMEM;
 	}
 
-	amikbd_dev->name = "Amiga Keyboard";
-	amikbd_dev->phys = "amikbd/input0";
-	amikbd_dev->id.bustype = BUS_AMIGA;
-	amikbd_dev->id.vendor = 0x0001;
-	amikbd_dev->id.product = 0x0001;
-	amikbd_dev->id.version = 0x0100;
+	dev->name = pdev->name;
+	dev->phys = "amikbd/input0";
+	dev->id.bustype = BUS_AMIGA;
+	dev->id.vendor = 0x0001;
+	dev->id.product = 0x0001;
+	dev->id.version = 0x0100;
+	dev->dev.parent = &pdev->dev;
 
-	amikbd_dev->evbit[0] = BIT_MASK(EV_KEY) | BIT_MASK(EV_REP);
+	dev->evbit[0] = BIT_MASK(EV_KEY) | BIT_MASK(EV_REP);
 
 	for (i = 0; i < 0x78; i++)
-		set_bit(i, amikbd_dev->keybit);
+		set_bit(i, dev->keybit);
 
 	for (i = 0; i < MAX_NR_KEYMAPS; i++) {
 		static u_short temp_map[NR_KEYS] __initdata;
@@ -229,30 +224,54 @@ static int __init amikbd_init(void)
 		memcpy(key_maps[i], temp_map, sizeof(temp_map));
 	}
 	ciaa.cra &= ~0x41;	 /* serial data in, turn off TA */
-	if (request_irq(IRQ_AMIGA_CIAA_SP, amikbd_interrupt, 0, "amikbd",
-			amikbd_interrupt)) {
-		err = -EBUSY;
+	err = request_irq(IRQ_AMIGA_CIAA_SP, amikbd_interrupt, 0, "amikbd",
+			  dev);
+	if (err)
 		goto fail2;
-	}
 
-	err = input_register_device(amikbd_dev);
+	err = input_register_device(dev);
 	if (err)
 		goto fail3;
 
+	platform_set_drvdata(pdev, dev);
+
 	return 0;
 
- fail3:	free_irq(IRQ_AMIGA_CIAA_SP, amikbd_interrupt);
- fail2:	input_free_device(amikbd_dev);
- fail1:	release_mem_region(CIAA_PHYSADDR - 1 + 0xb00, 0x100);
+ fail3:	free_irq(IRQ_AMIGA_CIAA_SP, dev);
+ fail2:	input_free_device(dev);
 	return err;
 }
 
-static void __exit amikbd_exit(void)
+static int __exit amikbd_remove(struct platform_device *pdev)
+{
+	struct input_dev *dev = platform_get_drvdata(pdev);
+
+	platform_set_drvdata(pdev, NULL);
+	free_irq(IRQ_AMIGA_CIAA_SP, dev);
+	input_unregister_device(dev);
+	return 0;
+}
+
+static struct platform_driver amikbd_driver = {
+	.remove = __exit_p(amikbd_remove),
+	.driver   = {
+		.name	= "amiga-keyboard",
+		.owner	= THIS_MODULE,
+	},
+};
+
+static int __init amikbd_init(void)
 {
-	free_irq(IRQ_AMIGA_CIAA_SP, amikbd_interrupt);
-	input_unregister_device(amikbd_dev);
-	release_mem_region(CIAA_PHYSADDR - 1 + 0xb00, 0x100);
+	return platform_driver_probe(&amikbd_driver, amikbd_probe);
 }
 
 module_init(amikbd_init);
+
+static void __exit amikbd_exit(void)
+{
+	platform_driver_unregister(&amikbd_driver);
+}
+
 module_exit(amikbd_exit);
+
+MODULE_ALIAS("platform:amiga-keyboard");
diff --git a/drivers/input/keyboard/lm8323.c b/drivers/input/keyboard/lm8323.c
index bc696931fed7..40b032f0e32c 100644
--- a/drivers/input/keyboard/lm8323.c
+++ b/drivers/input/keyboard/lm8323.c
@@ -778,8 +778,6 @@ static int __devexit lm8323_remove(struct i2c_client *client)
 	struct lm8323_chip *lm = i2c_get_clientdata(client);
 	int i;
 
-	i2c_set_clientdata(client, NULL);
-
 	disable_irq_wake(client->irq);
 	free_irq(client->irq, lm);
 	cancel_work_sync(&lm->work);
diff --git a/drivers/input/keyboard/max7359_keypad.c b/drivers/input/keyboard/max7359_keypad.c
index 7fc8185e5c1b..9091ff5ea808 100644
--- a/drivers/input/keyboard/max7359_keypad.c
+++ b/drivers/input/keyboard/max7359_keypad.c
@@ -265,7 +265,6 @@ static int __devexit max7359_remove(struct i2c_client *client)
 
 	free_irq(client->irq, keypad);
 	input_unregister_device(keypad->input_dev);
-	i2c_set_clientdata(client, NULL);
 	kfree(keypad);
 
 	return 0;
diff --git a/drivers/input/keyboard/qt2160.c b/drivers/input/keyboard/qt2160.c
index 31f30087b591..fac695157e8a 100644
--- a/drivers/input/keyboard/qt2160.c
+++ b/drivers/input/keyboard/qt2160.c
@@ -358,7 +358,6 @@ static int __devexit qt2160_remove(struct i2c_client *client)
 	input_unregister_device(qt2160->input);
 	kfree(qt2160);
 
-	i2c_set_clientdata(client, NULL);
 	return 0;
 }
 
diff --git a/drivers/input/keyboard/tca6416-keypad.c b/drivers/input/keyboard/tca6416-keypad.c
index 493c93f25e2a..00137bebcf97 100644
--- a/drivers/input/keyboard/tca6416-keypad.c
+++ b/drivers/input/keyboard/tca6416-keypad.c
@@ -316,8 +316,6 @@ static int __devexit tca6416_keypad_remove(struct i2c_client *client)
 	input_unregister_device(chip->input);
 	kfree(chip);
 
-	i2c_set_clientdata(client, NULL);
-
 	return 0;
 }
 
diff --git a/drivers/input/misc/Kconfig b/drivers/input/misc/Kconfig
index 48cdabec372a..c44b9eafc556 100644
--- a/drivers/input/misc/Kconfig
+++ b/drivers/input/misc/Kconfig
@@ -80,6 +80,16 @@ config INPUT_M68K_BEEP
 	tristate "M68k Beeper support"
 	depends on M68K
 
+config INPUT_MAX8925_ONKEY
+	tristate "MAX8925 ONKEY support"
+	depends on MFD_MAX8925
+	help
+	  Support the ONKEY of MAX8925 PMICs as an input device
+	  reporting power button status.
+
+	  To compile this driver as a module, choose M here: the module
+	  will be called max8925_onkey.
+
 config INPUT_APANEL
 	tristate "Fujitsu Lifebook Application Panel buttons"
 	depends on X86 && I2C && LEDS_CLASS
diff --git a/drivers/input/misc/Makefile b/drivers/input/misc/Makefile
index f9f577031e06..71fe57d8023f 100644
--- a/drivers/input/misc/Makefile
+++ b/drivers/input/misc/Makefile
@@ -20,6 +20,7 @@ obj-$(CONFIG_HP_SDC_RTC)		+= hp_sdc_rtc.o
 obj-$(CONFIG_INPUT_IXP4XX_BEEPER)	+= ixp4xx-beeper.o
 obj-$(CONFIG_INPUT_KEYSPAN_REMOTE)	+= keyspan_remote.o
 obj-$(CONFIG_INPUT_M68K_BEEP)		+= m68kspkr.o
+obj-$(CONFIG_INPUT_MAX8925_ONKEY)	+= max8925_onkey.o
 obj-$(CONFIG_INPUT_PCAP)		+= pcap_keys.o
 obj-$(CONFIG_INPUT_PCF50633_PMU)	+= pcf50633-input.o
 obj-$(CONFIG_INPUT_PCF8574)		+= pcf8574_keypad.o
diff --git a/drivers/input/misc/ad714x-i2c.c b/drivers/input/misc/ad714x-i2c.c
index e9adbe49f6a4..2bef8fa56c94 100644
--- a/drivers/input/misc/ad714x-i2c.c
+++ b/drivers/input/misc/ad714x-i2c.c
@@ -97,7 +97,6 @@ static int __devexit ad714x_i2c_remove(struct i2c_client *client)
 	struct ad714x_chip *chip = i2c_get_clientdata(client);
 
 	ad714x_remove(chip);
-	i2c_set_clientdata(client, NULL);
 
 	return 0;
 }
diff --git a/drivers/input/misc/hp_sdc_rtc.c b/drivers/input/misc/hp_sdc_rtc.c
index ad730e15afc0..c19066479057 100644
--- a/drivers/input/misc/hp_sdc_rtc.c
+++ b/drivers/input/misc/hp_sdc_rtc.c
@@ -43,6 +43,7 @@
 #include <linux/proc_fs.h>
 #include <linux/poll.h>
 #include <linux/rtc.h>
+#include <linux/smp_lock.h>
 #include <linux/semaphore.h>
 
 MODULE_AUTHOR("Brian S. Julin <bri@calyx.com>");
@@ -64,8 +65,8 @@ static DECLARE_WAIT_QUEUE_HEAD(hp_sdc_rtc_wait);
 static ssize_t hp_sdc_rtc_read(struct file *file, char __user *buf,
 			       size_t count, loff_t *ppos);
 
-static int hp_sdc_rtc_ioctl(struct inode *inode, struct file *file,
-			    unsigned int cmd, unsigned long arg);
+static long hp_sdc_rtc_unlocked_ioctl(struct file *file,
+				      unsigned int cmd, unsigned long arg);
 
 static unsigned int hp_sdc_rtc_poll(struct file *file, poll_table *wait);
 
@@ -512,7 +513,7 @@ static int hp_sdc_rtc_read_proc(char *page, char **start, off_t off,
         return len;
 }
 
-static int hp_sdc_rtc_ioctl(struct inode *inode, struct file *file, 
+static int hp_sdc_rtc_ioctl(struct file *file, 
 			    unsigned int cmd, unsigned long arg)
 {
 #if 1
@@ -659,14 +660,27 @@ static int hp_sdc_rtc_ioctl(struct inode *inode, struct file *file,
 #endif
 }
 
+static long hp_sdc_rtc_unlocked_ioctl(struct file *file,
+				      unsigned int cmd, unsigned long arg)
+{
+	int ret;
+
+	lock_kernel();
+	ret = hp_sdc_rtc_ioctl(file, cmd, arg);
+	unlock_kernel();
+
+	return ret;
+}
+
+
 static const struct file_operations hp_sdc_rtc_fops = {
-        .owner =	THIS_MODULE,
-        .llseek =	no_llseek,
-        .read =		hp_sdc_rtc_read,
-        .poll =		hp_sdc_rtc_poll,
-        .ioctl =	hp_sdc_rtc_ioctl,
-        .open =		hp_sdc_rtc_open,
-        .fasync =	hp_sdc_rtc_fasync,
+        .owner =		THIS_MODULE,
+        .llseek =		no_llseek,
+        .read =			hp_sdc_rtc_read,
+        .poll =			hp_sdc_rtc_poll,
+        .unlocked_ioctl =	hp_sdc_rtc_unlocked_ioctl,
+        .open =			hp_sdc_rtc_open,
+        .fasync =		hp_sdc_rtc_fasync,
 };
 
 static struct miscdevice hp_sdc_rtc_dev = {
diff --git a/drivers/input/misc/max8925_onkey.c b/drivers/input/misc/max8925_onkey.c
new file mode 100644
index 000000000000..80af44608018
--- /dev/null
+++ b/drivers/input/misc/max8925_onkey.c
@@ -0,0 +1,148 @@
+/**
+ * max8925_onkey.c - MAX8925 ONKEY driver
+ *
+ * Copyright (C) 2009 Marvell International Ltd.
+ *      Haojian Zhuang <haojian.zhuang@marvell.com>
+ *
+ * This file is subject to the terms and conditions of the GNU General
+ * Public License. See the file "COPYING" in the main directory of this
+ * archive for more details.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/i2c.h>
+#include <linux/input.h>
+#include <linux/interrupt.h>
+#include <linux/mfd/max8925.h>
+#include <linux/slab.h>
+
+#define HARDRESET_EN		(1 << 7)
+#define PWREN_EN		(1 << 7)
+
+struct max8925_onkey_info {
+	struct input_dev	*idev;
+	struct i2c_client	*i2c;
+	int			irq;
+};
+
+/*
+ * MAX8925 gives us an interrupt when ONKEY is held for 3 seconds.
+ * max8925_set_bits() operates I2C bus and may sleep. So implement
+ * it in thread IRQ handler.
+ */
+static irqreturn_t max8925_onkey_handler(int irq, void *data)
+{
+	struct max8925_onkey_info *info = data;
+
+	input_report_key(info->idev, KEY_POWER, 1);
+	input_sync(info->idev);
+
+	/* Enable hardreset to halt if system isn't shutdown on time */
+	max8925_set_bits(info->i2c, MAX8925_SYSENSEL,
+			 HARDRESET_EN, HARDRESET_EN);
+
+	return IRQ_HANDLED;
+}
+
+static int __devinit max8925_onkey_probe(struct platform_device *pdev)
+{
+	struct max8925_chip *chip = dev_get_drvdata(pdev->dev.parent);
+	struct max8925_onkey_info *info;
+	int error;
+
+	info = kzalloc(sizeof(struct max8925_onkey_info), GFP_KERNEL);
+	if (!info)
+		return -ENOMEM;
+
+	info->i2c = chip->i2c;
+	info->irq = chip->irq_base + MAX8925_IRQ_GPM_SW_3SEC;
+
+	info->idev = input_allocate_device();
+	if (!info->idev) {
+		dev_err(chip->dev, "Failed to allocate input dev\n");
+		error = -ENOMEM;
+		goto out_input;
+	}
+
+	info->idev->name = "max8925_on";
+	info->idev->phys = "max8925_on/input0";
+	info->idev->id.bustype = BUS_I2C;
+	info->idev->dev.parent = &pdev->dev;
+	info->idev->evbit[0] = BIT_MASK(EV_KEY);
+	info->idev->keybit[BIT_WORD(KEY_POWER)] = BIT_MASK(KEY_POWER);
+
+	error = request_threaded_irq(info->irq, NULL, max8925_onkey_handler,
+				     IRQF_ONESHOT, "onkey", info);
+	if (error < 0) {
+		dev_err(chip->dev, "Failed to request IRQ: #%d: %d\n",
+			info->irq, error);
+		goto out_irq;
+	}
+
+	error = input_register_device(info->idev);
+	if (error) {
+		dev_err(chip->dev, "Can't register input device: %d\n", error);
+		goto out;
+	}
+
+	platform_set_drvdata(pdev, info);
+
+	return 0;
+
+out:
+	free_irq(info->irq, info);
+out_irq:
+	input_free_device(info->idev);
+out_input:
+	kfree(info);
+	return error;
+}
+
+static int __devexit max8925_onkey_remove(struct platform_device *pdev)
+{
+	struct max8925_onkey_info *info = platform_get_drvdata(pdev);
+
+	free_irq(info->irq, info);
+	input_unregister_device(info->idev);
+	kfree(info);
+
+	platform_set_drvdata(pdev, NULL);
+
+	return 0;
+}
+
+static struct platform_driver max8925_onkey_driver = {
+	.driver		= {
+		.name	= "max8925-onkey",
+		.owner	= THIS_MODULE,
+	},
+	.probe		= max8925_onkey_probe,
+	.remove		= __devexit_p(max8925_onkey_remove),
+};
+
+static int __init max8925_onkey_init(void)
+{
+	return platform_driver_register(&max8925_onkey_driver);
+}
+module_init(max8925_onkey_init);
+
+static void __exit max8925_onkey_exit(void)
+{
+	platform_driver_unregister(&max8925_onkey_driver);
+}
+module_exit(max8925_onkey_exit);
+
+MODULE_DESCRIPTION("Maxim MAX8925 ONKEY driver");
+MODULE_AUTHOR("Haojian Zhuang <haojian.zhuang@marvell.com>");
+MODULE_LICENSE("GPL");
diff --git a/drivers/input/misc/pcf8574_keypad.c b/drivers/input/misc/pcf8574_keypad.c
index 5c3ac4e0b055..0ac47d2898ec 100644
--- a/drivers/input/misc/pcf8574_keypad.c
+++ b/drivers/input/misc/pcf8574_keypad.c
@@ -168,8 +168,6 @@ static int __devexit pcf8574_kp_remove(struct i2c_client *client)
 	input_unregister_device(lp->idev);
 	kfree(lp);
 
-	i2c_set_clientdata(client, NULL);
-
 	return 0;
 }
 
diff --git a/drivers/input/misc/sparcspkr.c b/drivers/input/misc/sparcspkr.c
index 0d45422f8095..1dacae4b43f0 100644
--- a/drivers/input/misc/sparcspkr.c
+++ b/drivers/input/misc/sparcspkr.c
@@ -259,8 +259,11 @@ static const struct of_device_id bbc_beep_match[] = {
 };
 
 static struct of_platform_driver bbc_beep_driver = {
-	.name		= "bbcbeep",
-	.match_table	= bbc_beep_match,
+	.driver = {
+		.name = "bbcbeep",
+		.owner = THIS_MODULE,
+		.of_match_table = bbc_beep_match,
+	},
 	.probe		= bbc_beep_probe,
 	.remove		= __devexit_p(bbc_remove),
 	.shutdown	= sparcspkr_shutdown,
@@ -338,8 +341,11 @@ static const struct of_device_id grover_beep_match[] = {
 };
 
 static struct of_platform_driver grover_beep_driver = {
-	.name		= "groverbeep",
-	.match_table	= grover_beep_match,
+	.driver = {
+		.name = "groverbeep",
+		.owner = THIS_MODULE,
+		.of_match_table = grover_beep_match,
+	},
 	.probe		= grover_beep_probe,
 	.remove		= __devexit_p(grover_remove),
 	.shutdown	= sparcspkr_shutdown,
diff --git a/drivers/input/misc/twl4030-vibra.c b/drivers/input/misc/twl4030-vibra.c
index fee9eac8e04a..4f9b2afc24e8 100644
--- a/drivers/input/misc/twl4030-vibra.c
+++ b/drivers/input/misc/twl4030-vibra.c
@@ -90,8 +90,8 @@ static void vibra_disable(struct vibra_info *info)
 	twl_i2c_write_u8(TWL4030_MODULE_AUDIO_VOICE,
 			 (reg & ~TWL4030_VIBRA_EN), TWL4030_REG_VIBRA_CTL);
 
-	twl4030_codec_disable_resource(TWL4030_CODEC_RES_POWER);
 	twl4030_codec_disable_resource(TWL4030_CODEC_RES_APLL);
+	twl4030_codec_disable_resource(TWL4030_CODEC_RES_POWER);
 
 	info->enabled = false;
 }
diff --git a/drivers/input/misc/uinput.c b/drivers/input/misc/uinput.c
index 1477466076ad..b71eb55f2dbc 100644
--- a/drivers/input/misc/uinput.c
+++ b/drivers/input/misc/uinput.c
@@ -300,7 +300,7 @@ static int uinput_validate_absbits(struct input_dev *dev)
 	unsigned int cnt;
 	int retval = 0;
 
-	for (cnt = 0; cnt < ABS_MAX + 1; cnt++) {
+	for (cnt = 0; cnt < ABS_CNT; cnt++) {
 		if (!test_bit(cnt, dev->absbit))
 			continue;
 
@@ -387,7 +387,7 @@ static int uinput_setup_device(struct uinput_device *udev, const char __user *bu
 	dev->id.product	= user_dev->id.product;
 	dev->id.version	= user_dev->id.version;
 
-	size = sizeof(int) * (ABS_MAX + 1);
+	size = sizeof(int) * ABS_CNT;
 	memcpy(dev->absmax, user_dev->absmax, size);
 	memcpy(dev->absmin, user_dev->absmin, size);
 	memcpy(dev->absfuzz, user_dev->absfuzz, size);
diff --git a/drivers/input/mouse/amimouse.c b/drivers/input/mouse/amimouse.c
index a185ac78a42c..ff5f61a0fd3a 100644
--- a/drivers/input/mouse/amimouse.c
+++ b/drivers/input/mouse/amimouse.c
@@ -21,6 +21,7 @@
 #include <linux/init.h>
 #include <linux/input.h>
 #include <linux/interrupt.h>
+#include <linux/platform_device.h>
 
 #include <asm/irq.h>
 #include <asm/setup.h>
@@ -34,10 +35,10 @@ MODULE_DESCRIPTION("Amiga mouse driver");
 MODULE_LICENSE("GPL");
 
 static int amimouse_lastx, amimouse_lasty;
-static struct input_dev *amimouse_dev;
 
-static irqreturn_t amimouse_interrupt(int irq, void *dummy)
+static irqreturn_t amimouse_interrupt(int irq, void *data)
 {
+	struct input_dev *dev = data;
 	unsigned short joy0dat, potgor;
 	int nx, ny, dx, dy;
 
@@ -59,14 +60,14 @@ static irqreturn_t amimouse_interrupt(int irq, void *dummy)
 
 	potgor = amiga_custom.potgor;
 
-	input_report_rel(amimouse_dev, REL_X, dx);
-	input_report_rel(amimouse_dev, REL_Y, dy);
+	input_report_rel(dev, REL_X, dx);
+	input_report_rel(dev, REL_Y, dy);
 
-	input_report_key(amimouse_dev, BTN_LEFT,   ciaa.pra & 0x40);
-	input_report_key(amimouse_dev, BTN_MIDDLE, potgor & 0x0100);
-	input_report_key(amimouse_dev, BTN_RIGHT,  potgor & 0x0400);
+	input_report_key(dev, BTN_LEFT,   ciaa.pra & 0x40);
+	input_report_key(dev, BTN_MIDDLE, potgor & 0x0100);
+	input_report_key(dev, BTN_RIGHT,  potgor & 0x0400);
 
-	input_sync(amimouse_dev);
+	input_sync(dev);
 
 	return IRQ_HANDLED;
 }
@@ -74,63 +75,90 @@ static irqreturn_t amimouse_interrupt(int irq, void *dummy)
 static int amimouse_open(struct input_dev *dev)
 {
 	unsigned short joy0dat;
+	int error;
 
 	joy0dat = amiga_custom.joy0dat;
 
 	amimouse_lastx = joy0dat & 0xff;
 	amimouse_lasty = joy0dat >> 8;
 
-	if (request_irq(IRQ_AMIGA_VERTB, amimouse_interrupt, 0, "amimouse", amimouse_interrupt)) {
-                printk(KERN_ERR "amimouse.c: Can't allocate irq %d\n", IRQ_AMIGA_VERTB);
-                return -EBUSY;
-        }
+	error = request_irq(IRQ_AMIGA_VERTB, amimouse_interrupt, 0, "amimouse",
+			    dev);
+	if (error)
+		dev_err(&dev->dev, "Can't allocate irq %d\n", IRQ_AMIGA_VERTB);
 
-        return 0;
+	return error;
 }
 
 static void amimouse_close(struct input_dev *dev)
 {
-	free_irq(IRQ_AMIGA_VERTB, amimouse_interrupt);
+	free_irq(IRQ_AMIGA_VERTB, dev);
 }
 
-static int __init amimouse_init(void)
+static int __init amimouse_probe(struct platform_device *pdev)
 {
 	int err;
+	struct input_dev *dev;
 
-	if (!MACH_IS_AMIGA || !AMIGAHW_PRESENT(AMI_MOUSE))
-		return -ENODEV;
-
-	amimouse_dev = input_allocate_device();
-	if (!amimouse_dev)
+	dev = input_allocate_device();
+	if (!dev)
 		return -ENOMEM;
 
-	amimouse_dev->name = "Amiga mouse";
-	amimouse_dev->phys = "amimouse/input0";
-	amimouse_dev->id.bustype = BUS_AMIGA;
-	amimouse_dev->id.vendor = 0x0001;
-	amimouse_dev->id.product = 0x0002;
-	amimouse_dev->id.version = 0x0100;
+	dev->name = pdev->name;
+	dev->phys = "amimouse/input0";
+	dev->id.bustype = BUS_AMIGA;
+	dev->id.vendor = 0x0001;
+	dev->id.product = 0x0002;
+	dev->id.version = 0x0100;
 
-	amimouse_dev->evbit[0] = BIT_MASK(EV_KEY) | BIT_MASK(EV_REL);
-	amimouse_dev->relbit[0] = BIT_MASK(REL_X) | BIT_MASK(REL_Y);
-	amimouse_dev->keybit[BIT_WORD(BTN_LEFT)] = BIT_MASK(BTN_LEFT) |
+	dev->evbit[0] = BIT_MASK(EV_KEY) | BIT_MASK(EV_REL);
+	dev->relbit[0] = BIT_MASK(REL_X) | BIT_MASK(REL_Y);
+	dev->keybit[BIT_WORD(BTN_LEFT)] = BIT_MASK(BTN_LEFT) |
 		BIT_MASK(BTN_MIDDLE) | BIT_MASK(BTN_RIGHT);
-	amimouse_dev->open = amimouse_open;
-	amimouse_dev->close = amimouse_close;
+	dev->open = amimouse_open;
+	dev->close = amimouse_close;
+	dev->dev.parent = &pdev->dev;
 
-	err = input_register_device(amimouse_dev);
+	err = input_register_device(dev);
 	if (err) {
-		input_free_device(amimouse_dev);
+		input_free_device(dev);
 		return err;
 	}
 
+	platform_set_drvdata(pdev, dev);
+
 	return 0;
 }
 
-static void __exit amimouse_exit(void)
+static int __exit amimouse_remove(struct platform_device *pdev)
 {
-        input_unregister_device(amimouse_dev);
+	struct input_dev *dev = platform_get_drvdata(pdev);
+
+	platform_set_drvdata(pdev, NULL);
+	input_unregister_device(dev);
+	return 0;
+}
+
+static struct platform_driver amimouse_driver = {
+	.remove = __exit_p(amimouse_remove),
+	.driver   = {
+		.name	= "amiga-mouse",
+		.owner	= THIS_MODULE,
+	},
+};
+
+static int __init amimouse_init(void)
+{
+	return platform_driver_probe(&amimouse_driver, amimouse_probe);
 }
 
 module_init(amimouse_init);
+
+static void __exit amimouse_exit(void)
+{
+	platform_driver_unregister(&amimouse_driver);
+}
+
 module_exit(amimouse_exit);
+
+MODULE_ALIAS("platform:amiga-mouse");
diff --git a/drivers/input/mouse/synaptics_i2c.c b/drivers/input/mouse/synaptics_i2c.c
index 8291e7399ffa..0ae62f0bcb32 100644
--- a/drivers/input/mouse/synaptics_i2c.c
+++ b/drivers/input/mouse/synaptics_i2c.c
@@ -613,7 +613,6 @@ static int __devexit synaptics_i2c_remove(struct i2c_client *client)
 		free_irq(client->irq, touch);
 
 	input_unregister_device(touch->input);
-	i2c_set_clientdata(client, NULL);
 	kfree(touch);
 
 	return 0;
diff --git a/drivers/input/serio/Kconfig b/drivers/input/serio/Kconfig
index f34f1dbeb577..3bfe8fafc6ad 100644
--- a/drivers/input/serio/Kconfig
+++ b/drivers/input/serio/Kconfig
@@ -21,7 +21,8 @@ if SERIO
 config SERIO_I8042
 	tristate "i8042 PC Keyboard controller" if EMBEDDED || !X86
 	default y
-	depends on !PARISC && (!ARM || ARCH_SHARK || FOOTBRIDGE_HOST) && !M68K && !BLACKFIN
+	depends on !PARISC && (!ARM || ARCH_SHARK || FOOTBRIDGE_HOST) && \
+		   (!SUPERH || SH_CAYMAN) && !M68K && !BLACKFIN
 	help
 	  i8042 is the chip over which the standard AT keyboard and PS/2
 	  mouse are connected to the computer. If you use these devices,
diff --git a/drivers/input/serio/i8042-sparcio.h b/drivers/input/serio/i8042-sparcio.h
index 5071af2c0604..04e32f2d1241 100644
--- a/drivers/input/serio/i8042-sparcio.h
+++ b/drivers/input/serio/i8042-sparcio.h
@@ -51,7 +51,7 @@ static inline void i8042_write_command(int val)
 
 static int __devinit sparc_i8042_probe(struct of_device *op, const struct of_device_id *match)
 {
-	struct device_node *dp = op->node;
+	struct device_node *dp = op->dev.of_node;
 
 	dp = dp->child;
 	while (dp) {
@@ -96,8 +96,11 @@ static const struct of_device_id sparc_i8042_match[] = {
 MODULE_DEVICE_TABLE(of, sparc_i8042_match);
 
 static struct of_platform_driver sparc_i8042_driver = {
-	.name		= "i8042",
-	.match_table	= sparc_i8042_match,
+	.driver = {
+		.name = "i8042",
+		.owner = THIS_MODULE,
+		.of_match_table = sparc_i8042_match,
+	},
 	.probe		= sparc_i8042_probe,
 	.remove		= __devexit_p(sparc_i8042_remove),
 };
diff --git a/drivers/input/serio/xilinx_ps2.c b/drivers/input/serio/xilinx_ps2.c
index f84f8e32e3f1..e2c028d2638f 100644
--- a/drivers/input/serio/xilinx_ps2.c
+++ b/drivers/input/serio/xilinx_ps2.c
@@ -244,17 +244,17 @@ static int __devinit xps2_of_probe(struct of_device *ofdev,
 	int error;
 
 	dev_info(dev, "Device Tree Probing \'%s\'\n",
-			ofdev->node->name);
+			ofdev->dev.of_node->name);
 
 	/* Get iospace for the device */
-	error = of_address_to_resource(ofdev->node, 0, &r_mem);
+	error = of_address_to_resource(ofdev->dev.of_node, 0, &r_mem);
 	if (error) {
 		dev_err(dev, "invalid address\n");
 		return error;
 	}
 
 	/* Get IRQ for the device */
-	if (of_irq_to_resource(ofdev->node, 0, &r_irq) == NO_IRQ) {
+	if (of_irq_to_resource(ofdev->dev.of_node, 0, &r_irq) == NO_IRQ) {
 		dev_err(dev, "no IRQ found\n");
 		return -ENODEV;
 	}
@@ -342,7 +342,7 @@ static int __devexit xps2_of_remove(struct of_device *of_dev)
 	iounmap(drvdata->base_address);
 
 	/* Get iospace of the device */
-	if (of_address_to_resource(of_dev->node, 0, &r_mem))
+	if (of_address_to_resource(of_dev->dev.of_node, 0, &r_mem))
 		dev_err(dev, "invalid address\n");
 	else
 		release_mem_region(r_mem.start, resource_size(&r_mem));
@@ -362,8 +362,11 @@ static const struct of_device_id xps2_of_match[] __devinitconst = {
 MODULE_DEVICE_TABLE(of, xps2_of_match);
 
 static struct of_platform_driver xps2_of_driver = {
-	.name		= DRIVER_NAME,
-	.match_table	= xps2_of_match,
+	.driver = {
+		.name = DRIVER_NAME,
+		.owner = THIS_MODULE,
+		.of_match_table = xps2_of_match,
+	},
 	.probe		= xps2_of_probe,
 	.remove		= __devexit_p(xps2_of_remove),
 };
diff --git a/drivers/input/tablet/wacom_sys.c b/drivers/input/tablet/wacom_sys.c
index 2dc0c07c0469..42ba3691d908 100644
--- a/drivers/input/tablet/wacom_sys.c
+++ b/drivers/input/tablet/wacom_sys.c
@@ -508,7 +508,6 @@ static int wacom_probe(struct usb_interface *intf, const struct usb_device_id *i
 	}
 
 	input_dev->name = wacom_wac->name;
-	input_dev->name = wacom_wac->name;
 	input_dev->dev.parent = &intf->dev;
 	input_dev->open = wacom_open;
 	input_dev->close = wacom_close;
diff --git a/drivers/input/tablet/wacom_wac.c b/drivers/input/tablet/wacom_wac.c
index 847fd0135bcf..d564af58175c 100644
--- a/drivers/input/tablet/wacom_wac.c
+++ b/drivers/input/tablet/wacom_wac.c
@@ -300,7 +300,7 @@ static int wacom_intuos_inout(struct wacom_wac *wacom)
 		case 0x823: /* Intuos3 Grip Pen */
 		case 0x813: /* Intuos3 Classic Pen */
 		case 0x885: /* Intuos3 Marker Pen */
-		case 0x802: /* Intuos4 Grip Pen Eraser */
+		case 0x802: /* Intuos4 General Pen */
 		case 0x804: /* Intuos4 Marker Pen */
 		case 0x40802: /* Intuos4 Classic Pen */
 		case 0x022:
@@ -335,7 +335,7 @@ static int wacom_intuos_inout(struct wacom_wac *wacom)
 		case 0x81b: /* Intuos3 Classic Pen Eraser */
 		case 0x91b: /* Intuos3 Airbrush Eraser */
 		case 0x80c: /* Intuos4 Marker Pen Eraser */
-		case 0x80a: /* Intuos4 Grip Pen Eraser */
+		case 0x80a: /* Intuos4 General Pen Eraser */
 		case 0x4080a: /* Intuos4 Classic Pen Eraser */
 		case 0x90a: /* Intuos4 Airbrush Eraser */
 			wacom->tool[idx] = BTN_TOOL_RUBBER;
@@ -356,6 +356,11 @@ static int wacom_intuos_inout(struct wacom_wac *wacom)
 		return 1;
 	}
 
+	/* older I4 styli don't work with new Cintiqs */
+	if (!((wacom->id[idx] >> 20) & 0x01) &&
+			(features->type == WACOM_21UX2))
+		return 1;
+
 	/* Exit report */
 	if ((data[1] & 0xfe) == 0x80) {
 		/*
@@ -474,21 +479,43 @@ static int wacom_intuos_irq(struct wacom_wac *wacom)
 				input_report_abs(input, ABS_MISC, 0);
 			}
 		} else {
-			input_report_key(input, BTN_0, (data[5] & 0x01));
-			input_report_key(input, BTN_1, (data[5] & 0x02));
-			input_report_key(input, BTN_2, (data[5] & 0x04));
-			input_report_key(input, BTN_3, (data[5] & 0x08));
-			input_report_key(input, BTN_4, (data[6] & 0x01));
-			input_report_key(input, BTN_5, (data[6] & 0x02));
-			input_report_key(input, BTN_6, (data[6] & 0x04));
-			input_report_key(input, BTN_7, (data[6] & 0x08));
-			input_report_key(input, BTN_8, (data[5] & 0x10));
-			input_report_key(input, BTN_9, (data[6] & 0x10));
+			if (features->type == WACOM_21UX2) {
+				input_report_key(input, BTN_0, (data[5] & 0x01));
+				input_report_key(input, BTN_1, (data[6] & 0x01));
+				input_report_key(input, BTN_2, (data[6] & 0x02));
+				input_report_key(input, BTN_3, (data[6] & 0x04));
+				input_report_key(input, BTN_4, (data[6] & 0x08));
+				input_report_key(input, BTN_5, (data[6] & 0x10));
+				input_report_key(input, BTN_6, (data[6] & 0x20));
+				input_report_key(input, BTN_7, (data[6] & 0x40));
+				input_report_key(input, BTN_8, (data[6] & 0x80));
+				input_report_key(input, BTN_9, (data[7] & 0x01));
+				input_report_key(input, BTN_A, (data[8] & 0x01));
+				input_report_key(input, BTN_B, (data[8] & 0x02));
+				input_report_key(input, BTN_C, (data[8] & 0x04));
+				input_report_key(input, BTN_X, (data[8] & 0x08));
+				input_report_key(input, BTN_Y, (data[8] & 0x10));
+				input_report_key(input, BTN_Z, (data[8] & 0x20));
+				input_report_key(input, BTN_BASE, (data[8] & 0x40));
+				input_report_key(input, BTN_BASE2, (data[8] & 0x80));
+			} else {
+				input_report_key(input, BTN_0, (data[5] & 0x01));
+				input_report_key(input, BTN_1, (data[5] & 0x02));
+				input_report_key(input, BTN_2, (data[5] & 0x04));
+				input_report_key(input, BTN_3, (data[5] & 0x08));
+				input_report_key(input, BTN_4, (data[6] & 0x01));
+				input_report_key(input, BTN_5, (data[6] & 0x02));
+				input_report_key(input, BTN_6, (data[6] & 0x04));
+				input_report_key(input, BTN_7, (data[6] & 0x08));
+				input_report_key(input, BTN_8, (data[5] & 0x10));
+				input_report_key(input, BTN_9, (data[6] & 0x10));
+			}
 			input_report_abs(input, ABS_RX, ((data[1] & 0x1f) << 8) | data[2]);
 			input_report_abs(input, ABS_RY, ((data[3] & 0x1f) << 8) | data[4]);
 
 			if ((data[5] & 0x1f) | (data[6] & 0x1f) | (data[1] & 0x1f) |
-				data[2] | (data[3] & 0x1f) | data[4]) {
+				data[2] | (data[3] & 0x1f) | data[4] | data[8] |
+				(data[7] & 0x01)) {
 				input_report_key(input, wacom->tool[1], 1);
 				input_report_abs(input, ABS_MISC, PAD_DEVICE_ID);
 			} else {
@@ -640,7 +667,7 @@ static void wacom_tpc_finger_in(struct wacom_wac *wacom, char *data, int idx)
 	if (!idx)
 		input_report_key(input, BTN_TOUCH, 1);
 	input_event(input, EV_MSC, MSC_SERIAL, finger);
-	input_sync(wacom->input);
+	input_sync(input);
 
 	wacom->last_finger = finger;
 }
@@ -826,6 +853,7 @@ void wacom_wac_irq(struct wacom_wac *wacom_wac, size_t len)
 	case INTUOS4L:
 	case CINTIQ:
 	case WACOM_BEE:
+	case WACOM_21UX2:
 		sync = wacom_intuos_irq(wacom_wac);
 		break;
 
@@ -921,6 +949,17 @@ void wacom_setup_input_capabilities(struct input_dev *input_dev,
 		__set_bit(BTN_STYLUS2, input_dev->keybit);
 		break;
 
+	case WACOM_21UX2:
+		__set_bit(BTN_A, input_dev->keybit);
+		__set_bit(BTN_B, input_dev->keybit);
+		__set_bit(BTN_C, input_dev->keybit);
+		__set_bit(BTN_X, input_dev->keybit);
+		__set_bit(BTN_Y, input_dev->keybit);
+		__set_bit(BTN_Z, input_dev->keybit);
+		__set_bit(BTN_BASE, input_dev->keybit);
+		__set_bit(BTN_BASE2, input_dev->keybit);
+		/* fall through */
+
 	case WACOM_BEE:
 		__set_bit(BTN_8, input_dev->keybit);
 		__set_bit(BTN_9, input_dev->keybit);
@@ -1105,6 +1144,8 @@ static const struct wacom_features wacom_features_0xBA =
 	{ "Wacom Intuos4 8x13",   WACOM_PKGLEN_INTUOS,    65024, 40640, 2047, 63, INTUOS4L };
 static const struct wacom_features wacom_features_0xBB =
 	{ "Wacom Intuos4 12x19",  WACOM_PKGLEN_INTUOS,    97536, 60960, 2047, 63, INTUOS4L };
+static const struct wacom_features wacom_features_0xBC =
+	{ "Wacom Intuos4 WL",     WACOM_PKGLEN_INTUOS,    40840, 25400, 2047, 63, INTUOS4 };
 static const struct wacom_features wacom_features_0x3F =
 	{ "Wacom Cintiq 21UX",    WACOM_PKGLEN_INTUOS,    87200, 65600, 1023, 63, CINTIQ };
 static const struct wacom_features wacom_features_0xC5 =
@@ -1113,6 +1154,8 @@ static const struct wacom_features wacom_features_0xC6 =
 	{ "Wacom Cintiq 12WX",    WACOM_PKGLEN_INTUOS,    53020, 33440, 1023, 63, WACOM_BEE };
 static const struct wacom_features wacom_features_0xC7 =
 	{ "Wacom DTU1931",        WACOM_PKGLEN_GRAPHIRE,  37832, 30305,  511,  0, PL };
+static const struct wacom_features wacom_features_0xCC =
+	{ "Wacom Cintiq 21UX2",   WACOM_PKGLEN_INTUOS,    87200, 65600, 2047, 63, WACOM_21UX2 };
 static const struct wacom_features wacom_features_0x90 =
 	{ "Wacom ISDv4 90",       WACOM_PKGLEN_GRAPHIRE,  26202, 16325,  255,  0, TABLETPC };
 static const struct wacom_features wacom_features_0x93 =
@@ -1185,10 +1228,12 @@ const struct usb_device_id wacom_ids[] = {
 	{ USB_DEVICE_WACOM(0xB9) },
 	{ USB_DEVICE_WACOM(0xBA) },
 	{ USB_DEVICE_WACOM(0xBB) },
+	{ USB_DEVICE_WACOM(0xBC) },
 	{ USB_DEVICE_WACOM(0x3F) },
 	{ USB_DEVICE_WACOM(0xC5) },
 	{ USB_DEVICE_WACOM(0xC6) },
 	{ USB_DEVICE_WACOM(0xC7) },
+	{ USB_DEVICE_WACOM(0xCC) },
 	{ USB_DEVICE_WACOM(0x90) },
 	{ USB_DEVICE_WACOM(0x93) },
 	{ USB_DEVICE_WACOM(0x9A) },
diff --git a/drivers/input/tablet/wacom_wac.h b/drivers/input/tablet/wacom_wac.h
index 063f1af3204f..854b92092dfc 100644
--- a/drivers/input/tablet/wacom_wac.h
+++ b/drivers/input/tablet/wacom_wac.h
@@ -50,6 +50,7 @@ enum {
 	INTUOS4S,
 	INTUOS4,
 	INTUOS4L,
+	WACOM_21UX2,
 	CINTIQ,
 	WACOM_BEE,
 	WACOM_MO,
diff --git a/drivers/input/touchscreen/Kconfig b/drivers/input/touchscreen/Kconfig
index b9f58ca82fd1..3b9d5e2105d7 100644
--- a/drivers/input/touchscreen/Kconfig
+++ b/drivers/input/touchscreen/Kconfig
@@ -156,7 +156,7 @@ config TOUCHSCREEN_FUJITSU
 config TOUCHSCREEN_S3C2410
 	tristate "Samsung S3C2410/generic touchscreen input driver"
 	depends on ARCH_S3C2410 || SAMSUNG_DEV_TS
-	select S3C24XX_ADC
+	select S3C_ADC
 	help
 	  Say Y here if you have the s3c2410 touchscreen.
 
@@ -590,4 +590,17 @@ config TOUCHSCREEN_PCAP
 
 	  To compile this driver as a module, choose M here: the
 	  module will be called pcap_ts.
+
+config TOUCHSCREEN_TPS6507X
+	tristate "TPS6507x based touchscreens"
+	depends on I2C
+	help
+	  Say Y here if you have a TPS6507x based touchscreen
+	  controller.
+
+	  If unsure, say N.
+
+	  To compile this driver as a module, choose M here: the
+	  module will be called tps6507x_ts.
+
 endif
diff --git a/drivers/input/touchscreen/Makefile b/drivers/input/touchscreen/Makefile
index 8ad36eef90a2..497964a7a214 100644
--- a/drivers/input/touchscreen/Makefile
+++ b/drivers/input/touchscreen/Makefile
@@ -46,3 +46,4 @@ obj-$(CONFIG_TOUCHSCREEN_WM97XX_ATMEL)	+= atmel-wm97xx.o
 obj-$(CONFIG_TOUCHSCREEN_WM97XX_MAINSTONE)	+= mainstone-wm97xx.o
 obj-$(CONFIG_TOUCHSCREEN_WM97XX_ZYLONITE)	+= zylonite-wm97xx.o
 obj-$(CONFIG_TOUCHSCREEN_W90X900)	+= w90p910_ts.o
+obj-$(CONFIG_TOUCHSCREEN_TPS6507X)	+= tps6507x-ts.o
diff --git a/drivers/input/touchscreen/ad7879.c b/drivers/input/touchscreen/ad7879.c
index 794d070c6900..4b32fb4704cd 100644
--- a/drivers/input/touchscreen/ad7879.c
+++ b/drivers/input/touchscreen/ad7879.c
@@ -812,10 +812,8 @@ static int __devinit ad7879_probe(struct i2c_client *client,
 	ts->bus = client;
 
 	error = ad7879_construct(client, ts);
-	if (error) {
-		i2c_set_clientdata(client, NULL);
+	if (error)
 		kfree(ts);
-	}
 
 	return error;
 }
@@ -825,7 +823,6 @@ static int __devexit ad7879_remove(struct i2c_client *client)
 	struct ad7879 *ts = dev_get_drvdata(&client->dev);
 
 	ad7879_destroy(client, ts);
-	i2c_set_clientdata(client, NULL);
 	kfree(ts);
 
 	return 0;
diff --git a/drivers/input/touchscreen/ads7846.c b/drivers/input/touchscreen/ads7846.c
index 532279cda0e4..a9fdf55c0238 100644
--- a/drivers/input/touchscreen/ads7846.c
+++ b/drivers/input/touchscreen/ads7846.c
@@ -1163,8 +1163,8 @@ static int __devinit ads7846_probe(struct spi_device *spi)
 
 	ts->reg = regulator_get(&spi->dev, "vcc");
 	if (IS_ERR(ts->reg)) {
-		dev_err(&spi->dev, "unable to get regulator: %ld\n",
-			PTR_ERR(ts->reg));
+		err = PTR_ERR(ts->reg);
+		dev_err(&spi->dev, "unable to get regulator: %d\n", err);
 		goto err_free_gpio;
 	}
 
diff --git a/drivers/input/touchscreen/eeti_ts.c b/drivers/input/touchscreen/eeti_ts.c
index 75f8b73010fa..7a3a916f84a8 100644
--- a/drivers/input/touchscreen/eeti_ts.c
+++ b/drivers/input/touchscreen/eeti_ts.c
@@ -238,7 +238,6 @@ err2:
 	input = NULL; /* so we dont try to free it below */
 err1:
 	input_free_device(input);
-	i2c_set_clientdata(client, NULL);
 	kfree(priv);
 err0:
 	return err;
@@ -256,7 +255,6 @@ static int __devexit eeti_ts_remove(struct i2c_client *client)
 	enable_irq(priv->irq);
 
 	input_unregister_device(priv->input);
-	i2c_set_clientdata(client, NULL);
 	kfree(priv);
 
 	return 0;
diff --git a/drivers/input/touchscreen/mcs5000_ts.c b/drivers/input/touchscreen/mcs5000_ts.c
index ce8ab0269f6f..1fb0c2f06a44 100644
--- a/drivers/input/touchscreen/mcs5000_ts.c
+++ b/drivers/input/touchscreen/mcs5000_ts.c
@@ -256,7 +256,6 @@ static int __devexit mcs5000_ts_remove(struct i2c_client *client)
 	free_irq(client->irq, data);
 	input_unregister_device(data->input_dev);
 	kfree(data);
-	i2c_set_clientdata(client, NULL);
 
 	return 0;
 }
diff --git a/drivers/input/touchscreen/s3c2410_ts.c b/drivers/input/touchscreen/s3c2410_ts.c
index e0b7c834111d..6085d12fd561 100644
--- a/drivers/input/touchscreen/s3c2410_ts.c
+++ b/drivers/input/touchscreen/s3c2410_ts.c
@@ -173,7 +173,7 @@ static irqreturn_t stylus_irq(int irq, void *dev_id)
 	if (down)
 		s3c_adc_start(ts.client, 0, 1 << ts.shift);
 	else
-		dev_info(ts.dev, "%s: count=%d\n", __func__, ts.count);
+		dev_dbg(ts.dev, "%s: count=%d\n", __func__, ts.count);
 
 	if (ts.features & FEAT_PEN_IRQ) {
 		/* Clear pen down/up interrupt */
@@ -413,6 +413,8 @@ static struct dev_pm_ops s3c_ts_pmops = {
 #endif
 
 static struct platform_device_id s3cts_driver_ids[] = {
+	{ "s3c2410-ts", 0 },
+	{ "s3c2440-ts", 0 },
 	{ "s3c64xx-ts", FEAT_PEN_IRQ },
 	{ }
 };
diff --git a/drivers/input/touchscreen/tps6507x-ts.c b/drivers/input/touchscreen/tps6507x-ts.c
new file mode 100644
index 000000000000..5b70a1419b4d
--- /dev/null
+++ b/drivers/input/touchscreen/tps6507x-ts.c
@@ -0,0 +1,396 @@
+/*
+ * drivers/input/touchscreen/tps6507x_ts.c
+ *
+ * Touchscreen driver for the tps6507x chip.
+ *
+ * Copyright (c) 2009 RidgeRun (todd.fischer@ridgerun.com)
+ *
+ * Credits:
+ *
+ *    Using code from tsc2007, MtekVision Co., Ltd.
+ *
+ * For licencing details see kernel-base/COPYING
+ *
+ * TPS65070, TPS65073, TPS650731, and TPS650732 support
+ * 10 bit touch screen interface.
+ */
+
+#include <linux/module.h>
+#include <linux/workqueue.h>
+#include <linux/slab.h>
+#include <linux/input.h>
+#include <linux/platform_device.h>
+#include <linux/mfd/tps6507x.h>
+#include <linux/input/tps6507x-ts.h>
+#include <linux/delay.h>
+
+#define TSC_DEFAULT_POLL_PERIOD 30 /* ms */
+#define TPS_DEFAULT_MIN_PRESSURE 0x30
+#define MAX_10BIT ((1 << 10) - 1)
+
+#define	TPS6507X_ADCONFIG_CONVERT_TS (TPS6507X_ADCONFIG_AD_ENABLE | \
+					 TPS6507X_ADCONFIG_START_CONVERSION | \
+					 TPS6507X_ADCONFIG_INPUT_REAL_TSC)
+#define	TPS6507X_ADCONFIG_POWER_DOWN_TS (TPS6507X_ADCONFIG_INPUT_REAL_TSC)
+
+struct ts_event {
+	u16	x;
+	u16	y;
+	u16	pressure;
+};
+
+struct tps6507x_ts {
+	struct input_dev	*input_dev;
+	struct device		*dev;
+	char			phys[32];
+	struct workqueue_struct *wq;
+	struct delayed_work	work;
+	unsigned		polling;	/* polling is active */
+	struct ts_event		tc;
+	struct tps6507x_dev	*mfd;
+	u16			model;
+	unsigned		pendown;
+	int			irq;
+	void			(*clear_penirq)(void);
+	unsigned long		poll_period;	/* ms */
+	u16			min_pressure;
+	int			vref;		/* non-zero to leave vref on */
+};
+
+static int tps6507x_read_u8(struct tps6507x_ts *tsc, u8 reg, u8 *data)
+{
+	int err;
+
+	err = tsc->mfd->read_dev(tsc->mfd, reg, 1, data);
+
+	if (err)
+		return err;
+
+	return 0;
+}
+
+static int tps6507x_write_u8(struct tps6507x_ts *tsc, u8 reg, u8 data)
+{
+	return tsc->mfd->write_dev(tsc->mfd, reg, 1, &data);
+}
+
+static s32 tps6507x_adc_conversion(struct tps6507x_ts *tsc,
+				   u8 tsc_mode, u16 *value)
+{
+	s32 ret;
+	u8 adc_status;
+	u8 result;
+
+	/* Route input signal to A/D converter */
+
+	ret = tps6507x_write_u8(tsc, TPS6507X_REG_TSCMODE, tsc_mode);
+	if (ret) {
+		dev_err(tsc->dev, "TSC mode read failed\n");
+		goto err;
+	}
+
+	/* Start A/D conversion */
+
+	ret = tps6507x_write_u8(tsc, TPS6507X_REG_ADCONFIG,
+				TPS6507X_ADCONFIG_CONVERT_TS);
+	if (ret) {
+		dev_err(tsc->dev, "ADC config write failed\n");
+		return ret;
+	}
+
+	do {
+		ret = tps6507x_read_u8(tsc, TPS6507X_REG_ADCONFIG,
+				       &adc_status);
+		if (ret) {
+			dev_err(tsc->dev, "ADC config read failed\n");
+			goto err;
+		}
+	} while (adc_status & TPS6507X_ADCONFIG_START_CONVERSION);
+
+	ret = tps6507x_read_u8(tsc, TPS6507X_REG_ADRESULT_2, &result);
+	if (ret) {
+		dev_err(tsc->dev, "ADC result 2 read failed\n");
+		goto err;
+	}
+
+	*value = (result & TPS6507X_REG_ADRESULT_2_MASK) << 8;
+
+	ret = tps6507x_read_u8(tsc, TPS6507X_REG_ADRESULT_1, &result);
+	if (ret) {
+		dev_err(tsc->dev, "ADC result 1 read failed\n");
+		goto err;
+	}
+
+	*value |= result;
+
+	dev_dbg(tsc->dev, "TSC channel %d = 0x%X\n", tsc_mode, *value);
+
+err:
+	return ret;
+}
+
+/* Need to call tps6507x_adc_standby() after using A/D converter for the
+ * touch screen interrupt to work properly.
+ */
+
+static s32 tps6507x_adc_standby(struct tps6507x_ts *tsc)
+{
+	s32 ret;
+	s32 loops = 0;
+	u8 val;
+
+	ret = tps6507x_write_u8(tsc,  TPS6507X_REG_ADCONFIG,
+				TPS6507X_ADCONFIG_INPUT_TSC);
+	if (ret)
+		return ret;
+
+	ret = tps6507x_write_u8(tsc, TPS6507X_REG_TSCMODE,
+				TPS6507X_TSCMODE_STANDBY);
+	if (ret)
+		return ret;
+
+	ret = tps6507x_read_u8(tsc, TPS6507X_REG_INT, &val);
+	if (ret)
+		return ret;
+
+	while (val & TPS6507X_REG_TSC_INT) {
+		mdelay(10);
+		ret = tps6507x_read_u8(tsc, TPS6507X_REG_INT, &val);
+		if (ret)
+			return ret;
+		loops++;
+	}
+
+	return ret;
+}
+
+static void tps6507x_ts_handler(struct work_struct *work)
+{
+	struct tps6507x_ts *tsc =  container_of(work,
+				struct tps6507x_ts, work.work);
+	struct input_dev *input_dev = tsc->input_dev;
+	int pendown;
+	int schd;
+	int poll = 0;
+	s32 ret;
+
+	ret =  tps6507x_adc_conversion(tsc, TPS6507X_TSCMODE_PRESSURE,
+				       &tsc->tc.pressure);
+	if (ret)
+		goto done;
+
+	pendown = tsc->tc.pressure > tsc->min_pressure;
+
+	if (unlikely(!pendown && tsc->pendown)) {
+		dev_dbg(tsc->dev, "UP\n");
+		input_report_key(input_dev, BTN_TOUCH, 0);
+		input_report_abs(input_dev, ABS_PRESSURE, 0);
+		input_sync(input_dev);
+		tsc->pendown = 0;
+	}
+
+	if (pendown) {
+
+		if (!tsc->pendown) {
+			dev_dbg(tsc->dev, "DOWN\n");
+			input_report_key(input_dev, BTN_TOUCH, 1);
+		} else
+			dev_dbg(tsc->dev, "still down\n");
+
+		ret =  tps6507x_adc_conversion(tsc, TPS6507X_TSCMODE_X_POSITION,
+					       &tsc->tc.x);
+		if (ret)
+			goto done;
+
+		ret =  tps6507x_adc_conversion(tsc, TPS6507X_TSCMODE_Y_POSITION,
+					       &tsc->tc.y);
+		if (ret)
+			goto done;
+
+		input_report_abs(input_dev, ABS_X, tsc->tc.x);
+		input_report_abs(input_dev, ABS_Y, tsc->tc.y);
+		input_report_abs(input_dev, ABS_PRESSURE, tsc->tc.pressure);
+		input_sync(input_dev);
+		tsc->pendown = 1;
+		poll = 1;
+	}
+
+done:
+	/* always poll if not using interrupts */
+	poll = 1;
+
+	if (poll) {
+		schd = queue_delayed_work(tsc->wq, &tsc->work,
+					  msecs_to_jiffies(tsc->poll_period));
+		if (schd)
+			tsc->polling = 1;
+		else {
+			tsc->polling = 0;
+			dev_err(tsc->dev, "re-schedule failed");
+		}
+	} else
+		tsc->polling = 0;
+
+	ret = tps6507x_adc_standby(tsc);
+}
+
+static int tps6507x_ts_probe(struct platform_device *pdev)
+{
+	int error;
+	struct tps6507x_ts *tsc;
+	struct tps6507x_dev *tps6507x_dev = dev_get_drvdata(pdev->dev.parent);
+	struct touchscreen_init_data *init_data;
+	struct input_dev *input_dev;
+	struct tps6507x_board *tps_board;
+	int schd;
+
+	/**
+	 * tps_board points to pmic related constants
+	 * coming from the board-evm file.
+	 */
+
+	tps_board = (struct tps6507x_board *)tps6507x_dev->dev->platform_data;
+
+	if (!tps_board) {
+		dev_err(tps6507x_dev->dev,
+			"Could not find tps6507x platform data\n");
+		return -EIO;
+	}
+
+	/**
+	 * init_data points to array of regulator_init structures
+	 * coming from the board-evm file.
+	 */
+
+	init_data = tps_board->tps6507x_ts_init_data;
+
+	tsc = kzalloc(sizeof(struct tps6507x_ts), GFP_KERNEL);
+	if (!tsc) {
+		dev_err(tps6507x_dev->dev, "failed to allocate driver data\n");
+		error = -ENOMEM;
+		goto err0;
+	}
+
+	tps6507x_dev->ts = tsc;
+	tsc->mfd = tps6507x_dev;
+	tsc->dev = tps6507x_dev->dev;
+	input_dev = input_allocate_device();
+	if (!input_dev) {
+		dev_err(tsc->dev, "Failed to allocate input device.\n");
+		error = -ENOMEM;
+		goto err1;
+	}
+
+	input_dev->evbit[0] = BIT_MASK(EV_KEY) | BIT_MASK(EV_ABS);
+	input_dev->keybit[BIT_WORD(BTN_TOUCH)] = BIT_MASK(BTN_TOUCH);
+
+	input_set_abs_params(input_dev, ABS_X, 0, MAX_10BIT, 0, 0);
+	input_set_abs_params(input_dev, ABS_Y, 0, MAX_10BIT, 0, 0);
+	input_set_abs_params(input_dev, ABS_PRESSURE, 0, MAX_10BIT, 0, 0);
+
+	input_dev->name = "TPS6507x Touchscreen";
+	input_dev->id.bustype = BUS_I2C;
+	input_dev->dev.parent = tsc->dev;
+
+	snprintf(tsc->phys, sizeof(tsc->phys),
+		 "%s/input0", dev_name(tsc->dev));
+	input_dev->phys = tsc->phys;
+
+	dev_dbg(tsc->dev, "device: %s\n", input_dev->phys);
+
+	input_set_drvdata(input_dev, tsc);
+
+	tsc->input_dev = input_dev;
+
+	INIT_DELAYED_WORK(&tsc->work, tps6507x_ts_handler);
+	tsc->wq = create_workqueue("TPS6507x Touchscreen");
+
+	if (init_data) {
+		tsc->poll_period = init_data->poll_period;
+		tsc->vref = init_data->vref;
+		tsc->min_pressure = init_data->min_pressure;
+		input_dev->id.vendor = init_data->vendor;
+		input_dev->id.product = init_data->product;
+		input_dev->id.version = init_data->version;
+	} else {
+		tsc->poll_period = TSC_DEFAULT_POLL_PERIOD;
+		tsc->min_pressure = TPS_DEFAULT_MIN_PRESSURE;
+	}
+
+	error = tps6507x_adc_standby(tsc);
+	if (error)
+		goto err2;
+
+	error = input_register_device(input_dev);
+	if (error)
+		goto err2;
+
+	schd = queue_delayed_work(tsc->wq, &tsc->work,
+				  msecs_to_jiffies(tsc->poll_period));
+
+	if (schd)
+		tsc->polling = 1;
+	else {
+		tsc->polling = 0;
+		dev_err(tsc->dev, "schedule failed");
+		goto err2;
+	 }
+
+	return 0;
+
+err2:
+	cancel_delayed_work_sync(&tsc->work);
+	destroy_workqueue(tsc->wq);
+	input_free_device(input_dev);
+err1:
+	kfree(tsc);
+	tps6507x_dev->ts = NULL;
+err0:
+	return error;
+}
+
+static int __devexit tps6507x_ts_remove(struct platform_device *pdev)
+{
+	struct tps6507x_dev *tps6507x_dev = platform_get_drvdata(pdev);
+	struct tps6507x_ts *tsc = tps6507x_dev->ts;
+	struct input_dev *input_dev = tsc->input_dev;
+
+	if (!tsc)
+		return 0;
+
+	cancel_delayed_work_sync(&tsc->work);
+	destroy_workqueue(tsc->wq);
+
+	input_free_device(input_dev);
+
+	tps6507x_dev->ts = NULL;
+	kfree(tsc);
+
+	return 0;
+}
+
+static struct platform_driver tps6507x_ts_driver = {
+	.driver = {
+		.name = "tps6507x-ts",
+		.owner = THIS_MODULE,
+	},
+	.probe = tps6507x_ts_probe,
+	.remove = __devexit_p(tps6507x_ts_remove),
+};
+
+static int __init tps6507x_ts_init(void)
+{
+	return platform_driver_register(&tps6507x_ts_driver);
+}
+module_init(tps6507x_ts_init);
+
+static void __exit tps6507x_ts_exit(void)
+{
+	platform_driver_unregister(&tps6507x_ts_driver);
+}
+module_exit(tps6507x_ts_exit);
+
+MODULE_AUTHOR("Todd Fischer <todd.fischer@ridgerun.com>");
+MODULE_DESCRIPTION("TPS6507x - TouchScreen driver");
+MODULE_LICENSE("GPL v2");
+MODULE_ALIAS("platform:tps6507x-tsc");
diff --git a/drivers/input/touchscreen/tsc2007.c b/drivers/input/touchscreen/tsc2007.c
index 769b479fcaa6..be23780e8a3e 100644
--- a/drivers/input/touchscreen/tsc2007.c
+++ b/drivers/input/touchscreen/tsc2007.c
@@ -347,8 +347,6 @@ static int __devexit tsc2007_remove(struct i2c_client *client)
 	struct tsc2007	*ts = i2c_get_clientdata(client);
 	struct tsc2007_platform_data *pdata = client->dev.platform_data;
 
-	i2c_set_clientdata(client, NULL);
-
 	tsc2007_free_irq(ts);
 
 	if (pdata->exit_platform_hw)
diff --git a/drivers/input/touchscreen/usbtouchscreen.c b/drivers/input/touchscreen/usbtouchscreen.c
index 29a8bbf3f086..567d57215c28 100644
--- a/drivers/input/touchscreen/usbtouchscreen.c
+++ b/drivers/input/touchscreen/usbtouchscreen.c
@@ -857,6 +857,11 @@ static int nexio_read_data(struct usbtouch_usb *usbtouch, unsigned char *pkt)
 	if ((pkt[0] & 0xe0) != 0xe0)
 		return 0;
 
+	if (be16_to_cpu(packet->data_len) > 0xff)
+		packet->data_len = cpu_to_be16(be16_to_cpu(packet->data_len) - 0x100);
+	if (be16_to_cpu(packet->x_len) > 0xff)
+		packet->x_len = cpu_to_be16(be16_to_cpu(packet->x_len) - 0x80);
+
 	/* send ACK */
 	ret = usb_submit_urb(priv->ack, GFP_ATOMIC);
 
@@ -1112,7 +1117,7 @@ static struct usbtouch_device_info usbtouch_dev_info[] = {
 
 #ifdef CONFIG_TOUCHSCREEN_USB_NEXIO
 	[DEVTYPE_NEXIO] = {
-		.rept_size	= 128,
+		.rept_size	= 1024,
 		.irq_always	= true,
 		.read_data	= nexio_read_data,
 		.init		= nexio_init,
diff --git a/drivers/isdn/capi/capi.c b/drivers/isdn/capi/capi.c
index ee5837522f5a..0cabe31f26df 100644
--- a/drivers/isdn/capi/capi.c
+++ b/drivers/isdn/capi/capi.c
@@ -787,8 +787,7 @@ capi_poll(struct file *file, poll_table * wait)
 }
 
 static int
-capi_ioctl(struct inode *inode, struct file *file,
-	   unsigned int cmd, unsigned long arg)
+capi_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
 {
 	struct capidev *cdev = file->private_data;
 	capi_ioctl_struct data;
@@ -981,6 +980,18 @@ register_out:
 	}
 }
 
+static long
+capi_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
+{
+	int ret;
+
+	lock_kernel();
+	ret = capi_ioctl(file, cmd, arg);
+	unlock_kernel();
+
+	return ret;
+}
+
 static int capi_open(struct inode *inode, struct file *file)
 {
 	struct capidev *cdev;
@@ -1026,7 +1037,7 @@ static const struct file_operations capi_fops =
 	.read		= capi_read,
 	.write		= capi_write,
 	.poll		= capi_poll,
-	.ioctl		= capi_ioctl,
+	.unlocked_ioctl	= capi_unlocked_ioctl,
 	.open		= capi_open,
 	.release	= capi_release,
 };
diff --git a/drivers/isdn/capi/kcapi.c b/drivers/isdn/capi/kcapi.c
index bd00dceacaf0..b054494df846 100644
--- a/drivers/isdn/capi/kcapi.c
+++ b/drivers/isdn/capi/kcapi.c
@@ -1020,12 +1020,12 @@ static int old_capi_manufacturer(unsigned int cmd, void __user *data)
 		if (cmd == AVMB1_ADDCARD) {
 		   if ((retval = copy_from_user(&cdef, data,
 					    sizeof(avmb1_carddef))))
-			   return retval;
+			   return -EFAULT;
 		   cdef.cardtype = AVM_CARDTYPE_B1;
 		} else {
 		   if ((retval = copy_from_user(&cdef, data,
 					    sizeof(avmb1_extcarddef))))
-			   return retval;
+			   return -EFAULT;
 		}
 		cparams.port = cdef.port;
 		cparams.irq = cdef.irq;
@@ -1147,6 +1147,12 @@ load_unlock_out:
 		if (ctr->state == CAPI_CTR_DETECTED)
 			goto reset_unlock_out;
 
+		if (ctr->reset_ctr == NULL) {
+			printk(KERN_DEBUG "kcapi: reset: no reset function\n");
+			retval = -ESRCH;
+			goto reset_unlock_out;
+		}
+
 		ctr->reset_ctr(ctr);
 
 		retval = wait_on_ctr_state(ctr, CAPI_CTR_DETECTED);
@@ -1212,7 +1218,7 @@ int capi20_manufacturer(unsigned int cmd, void __user *data)
 		kcapi_carddef cdef;
 
 		if ((retval = copy_from_user(&cdef, data, sizeof(cdef))))
-			return retval;
+			return -EFAULT;
 
 		cparams.port = cdef.port;
 		cparams.irq = cdef.irq;
diff --git a/drivers/isdn/gigaset/capi.c b/drivers/isdn/gigaset/capi.c
index 964a55fb1486..8f78f15c8ef7 100644
--- a/drivers/isdn/gigaset/capi.c
+++ b/drivers/isdn/gigaset/capi.c
@@ -170,17 +170,6 @@ static inline void ignore_cstruct_param(struct cardstate *cs, _cstruct param,
 }
 
 /*
- * convert hex to binary
- */
-static inline u8 hex2bin(char c)
-{
-	int result = c & 0x0f;
-	if (c & 0x40)
-		result += 9;
-	return result;
-}
-
-/*
  * convert an IE from Gigaset hex string to ETSI binary representation
  * including length byte
  * return value: result length, -1 on error
@@ -191,7 +180,7 @@ static int encode_ie(char *in, u8 *out, int maxlen)
 	while (*in) {
 		if (!isxdigit(in[0]) || !isxdigit(in[1]) || l >= maxlen)
 			return -1;
-		out[++l] = (hex2bin(in[0]) << 4) + hex2bin(in[1]);
+		out[++l] = (hex_to_bin(in[0]) << 4) + hex_to_bin(in[1]);
 		in += 2;
 	}
 	out[0] = l;
@@ -933,30 +922,6 @@ void gigaset_isdn_stop(struct cardstate *cs)
  */
 
 /*
- * load firmware
- */
-static int gigaset_load_firmware(struct capi_ctr *ctr, capiloaddata *data)
-{
-	struct cardstate *cs = ctr->driverdata;
-
-	/* AVM specific operation, not needed for Gigaset -- ignore */
-	dev_notice(cs->dev, "load_firmware ignored\n");
-
-	return 0;
-}
-
-/*
- * reset (deactivate) controller
- */
-static void gigaset_reset_ctr(struct capi_ctr *ctr)
-{
-	struct cardstate *cs = ctr->driverdata;
-
-	/* AVM specific operation, not needed for Gigaset -- ignore */
-	dev_notice(cs->dev, "reset_ctr ignored\n");
-}
-
-/*
  * register CAPI application
  */
 static void gigaset_register_appl(struct capi_ctr *ctr, u16 appl,
@@ -2213,8 +2178,8 @@ int gigaset_isdn_regdev(struct cardstate *cs, const char *isdnid)
 	iif->ctr.driverdata    = cs;
 	strncpy(iif->ctr.name, isdnid, sizeof(iif->ctr.name));
 	iif->ctr.driver_name   = "gigaset";
-	iif->ctr.load_firmware = gigaset_load_firmware;
-	iif->ctr.reset_ctr     = gigaset_reset_ctr;
+	iif->ctr.load_firmware = NULL;
+	iif->ctr.reset_ctr     = NULL;
 	iif->ctr.register_appl = gigaset_register_appl;
 	iif->ctr.release_appl  = gigaset_release_appl;
 	iif->ctr.send_message  = gigaset_send_message;
diff --git a/drivers/isdn/hardware/mISDN/hfcsusb.c b/drivers/isdn/hardware/mISDN/hfcsusb.c
index b3b7e2879bac..8700474747e8 100644
--- a/drivers/isdn/hardware/mISDN/hfcsusb.c
+++ b/drivers/isdn/hardware/mISDN/hfcsusb.c
@@ -97,8 +97,10 @@ static int write_reg(struct hfcsusb *hw, __u8 reg, __u8 val)
 			hw->name, __func__, reg, val);
 
 	spin_lock(&hw->ctrl_lock);
-	if (hw->ctrl_cnt >= HFC_CTRL_BUFSIZE)
+	if (hw->ctrl_cnt >= HFC_CTRL_BUFSIZE) {
+		spin_unlock(&hw->ctrl_lock);
 		return 1;
+	}
 	buf = &hw->ctrl_buff[hw->ctrl_in_idx];
 	buf->hfcs_reg = reg;
 	buf->reg_val = val;
diff --git a/drivers/isdn/hardware/mISDN/netjet.c b/drivers/isdn/hardware/mISDN/netjet.c
index 0a3553df065f..54ae71a907f9 100644
--- a/drivers/isdn/hardware/mISDN/netjet.c
+++ b/drivers/isdn/hardware/mISDN/netjet.c
@@ -320,12 +320,12 @@ inittiger(struct tiger_hw *card)
 		return -ENOMEM;
 	}
 	for (i = 0; i < 2; i++) {
-		card->bc[i].hsbuf = kmalloc(NJ_DMA_TXSIZE, GFP_KERNEL);
+		card->bc[i].hsbuf = kmalloc(NJ_DMA_TXSIZE, GFP_ATOMIC);
 		if (!card->bc[i].hsbuf) {
 			pr_info("%s: no B%d send buffer\n", card->name, i + 1);
 			return -ENOMEM;
 		}
-		card->bc[i].hrbuf = kmalloc(NJ_DMA_RXSIZE, GFP_KERNEL);
+		card->bc[i].hrbuf = kmalloc(NJ_DMA_RXSIZE, GFP_ATOMIC);
 		if (!card->bc[i].hrbuf) {
 			pr_info("%s: no B%d recv buffer\n", card->name, i + 1);
 			return -ENOMEM;
diff --git a/drivers/isdn/hisax/hisax_fcpcipnp.c b/drivers/isdn/hisax/hisax_fcpcipnp.c
index 1925118122f8..8b0a7d86b30f 100644
--- a/drivers/isdn/hisax/hisax_fcpcipnp.c
+++ b/drivers/isdn/hisax/hisax_fcpcipnp.c
@@ -74,9 +74,10 @@ static struct pnp_device_id fcpnp_ids[] __devinitdata = {
 		.id		= "AVM0900",
 		.driver_data	= (unsigned long) "Fritz!Card PnP",
 	},
+	{ .id = "" }
 };
 
-MODULE_DEVICE_TABLE(isapnp, fcpnp_ids);
+MODULE_DEVICE_TABLE(pnp, fcpnp_ids);
 #endif
 
 static int protocol = 2;       /* EURO-ISDN Default */
diff --git a/drivers/isdn/i4l/isdn_common.c b/drivers/isdn/i4l/isdn_common.c
index 70044ee4b228..a44cdb492ea9 100644
--- a/drivers/isdn/i4l/isdn_common.c
+++ b/drivers/isdn/i4l/isdn_common.c
@@ -1272,9 +1272,9 @@ isdn_poll(struct file *file, poll_table * wait)
 
 
 static int
-isdn_ioctl(struct inode *inode, struct file *file, uint cmd, ulong arg)
+isdn_ioctl(struct file *file, uint cmd, ulong arg)
 {
-	uint minor = iminor(inode);
+	uint minor = iminor(file->f_path.dentry->d_inode);
 	isdn_ctrl c;
 	int drvidx;
 	int chidx;
@@ -1722,6 +1722,18 @@ isdn_ioctl(struct inode *inode, struct file *file, uint cmd, ulong arg)
 #undef cfg
 }
 
+static long
+isdn_unlocked_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
+{
+	int ret;
+
+	lock_kernel();
+	ret = isdn_ioctl(file, cmd, arg);
+	unlock_kernel();
+
+	return ret;
+}
+
 /*
  * Open the device code.
  */
@@ -1838,7 +1850,7 @@ static const struct file_operations isdn_fops =
 	.read		= isdn_read,
 	.write		= isdn_write,
 	.poll		= isdn_poll,
-	.ioctl		= isdn_ioctl,
+	.unlocked_ioctl	= isdn_unlocked_ioctl,
 	.open		= isdn_open,
 	.release	= isdn_close,
 };
diff --git a/drivers/isdn/mISDN/timerdev.c b/drivers/isdn/mISDN/timerdev.c
index 8785004e85e0..81048b8ed8ad 100644
--- a/drivers/isdn/mISDN/timerdev.c
+++ b/drivers/isdn/mISDN/timerdev.c
@@ -24,6 +24,7 @@
 #include <linux/miscdevice.h>
 #include <linux/module.h>
 #include <linux/mISDNif.h>
+#include <linux/smp_lock.h>
 #include "core.h"
 
 static u_int	*debug;
@@ -97,8 +98,6 @@ mISDN_read(struct file *filep, char __user *buf, size_t count, loff_t *off)
 	if (*debug & DEBUG_TIMER)
 		printk(KERN_DEBUG "%s(%p, %p, %d, %p)\n", __func__,
 			filep, buf, (int)count, off);
-	if (*off != filep->f_pos)
-		return -ESPIPE;
 
 	if (list_empty(&dev->expired) && (dev->work == 0)) {
 		if (filep->f_flags & O_NONBLOCK)
@@ -215,9 +214,8 @@ unlock:
 	return ret;
 }
 
-static int
-mISDN_ioctl(struct inode *inode, struct file *filep, unsigned int cmd,
-    unsigned long arg)
+static long
+mISDN_ioctl(struct file *filep, unsigned int cmd, unsigned long arg)
 {
 	struct mISDNtimerdev	*dev = filep->private_data;
 	int			id, tout, ret = 0;
@@ -226,6 +224,7 @@ mISDN_ioctl(struct inode *inode, struct file *filep, unsigned int cmd,
 	if (*debug & DEBUG_TIMER)
 		printk(KERN_DEBUG "%s(%p, %x, %lx)\n", __func__,
 		    filep, cmd, arg);
+	lock_kernel();
 	switch (cmd) {
 	case IMADDTIMER:
 		if (get_user(tout, (int __user *)arg)) {
@@ -257,13 +256,14 @@ mISDN_ioctl(struct inode *inode, struct file *filep, unsigned int cmd,
 	default:
 		ret = -EINVAL;
 	}
+	unlock_kernel();
 	return ret;
 }
 
 static const struct file_operations mISDN_fops = {
 	.read		= mISDN_read,
 	.poll		= mISDN_poll,
-	.ioctl		= mISDN_ioctl,
+	.unlocked_ioctl	= mISDN_ioctl,
 	.open		= mISDN_open,
 	.release	= mISDN_close,
 };
diff --git a/drivers/leds/Kconfig b/drivers/leds/Kconfig
index 505eb64c329c..81bf25e67ce1 100644
--- a/drivers/leds/Kconfig
+++ b/drivers/leds/Kconfig
@@ -21,7 +21,7 @@ comment "LED drivers"
 
 config LEDS_88PM860X
 	tristate "LED Support for Marvell 88PM860x PMIC"
-	depends on LEDS_CLASS && MFD_88PM860X
+	depends on MFD_88PM860X
 	help
 	  This option enables support for on-chip LED drivers found on Marvell
 	  Semiconductor 88PM8606 PMIC.
@@ -67,6 +67,16 @@ config LEDS_NET48XX
 	  This option enables support for the Soekris net4801 and net4826 error
 	  LED.
 
+config LEDS_NET5501
+	tristate "LED Support for Soekris net5501 series Error LED"
+	depends on LEDS_TRIGGERS
+	depends on X86 && LEDS_GPIO_PLATFORM && GPIO_CS5535
+	select LEDS_TRIGGER_DEFAULT_ON
+	default n
+	help
+	  Add support for the Soekris net5501 board (detection, error led
+	  and GPIO).
+
 config LEDS_FSG
 	tristate "LED Support for the Freecom FSG-3"
 	depends on MACH_FSG
@@ -285,6 +295,13 @@ config LEDS_DELL_NETBOOKS
 	  This adds support for the Latitude 2100 and similar
 	  notebooks that have an external LED.
 
+config LEDS_MC13783
+	tristate "LED Support for MC13783 PMIC"
+	depends on MFD_MC13783
+	help
+	  This option enable support for on-chip LED drivers found
+	  on Freescale Semiconductor MC13783 PMIC.
+
 config LEDS_TRIGGERS
 	bool "LED Trigger support"
 	help
diff --git a/drivers/leds/Makefile b/drivers/leds/Makefile
index 0cd8b9957380..2493de499374 100644
--- a/drivers/leds/Makefile
+++ b/drivers/leds/Makefile
@@ -13,6 +13,7 @@ obj-$(CONFIG_LEDS_MIKROTIK_RB532)	+= leds-rb532.o
 obj-$(CONFIG_LEDS_S3C24XX)		+= leds-s3c24xx.o
 obj-$(CONFIG_LEDS_AMS_DELTA)		+= leds-ams-delta.o
 obj-$(CONFIG_LEDS_NET48XX)		+= leds-net48xx.o
+obj-$(CONFIG_LEDS_NET5501)		+= leds-net5501.o
 obj-$(CONFIG_LEDS_WRAP)			+= leds-wrap.o
 obj-$(CONFIG_LEDS_ALIX2)		+= leds-alix2.o
 obj-$(CONFIG_LEDS_H1940)		+= leds-h1940.o
@@ -35,6 +36,7 @@ obj-$(CONFIG_LEDS_INTEL_SS4200)		+= leds-ss4200.o
 obj-$(CONFIG_LEDS_LT3593)		+= leds-lt3593.o
 obj-$(CONFIG_LEDS_ADP5520)		+= leds-adp5520.o
 obj-$(CONFIG_LEDS_DELL_NETBOOKS)	+= dell-led.o
+obj-$(CONFIG_LEDS_MC13783)		+= leds-mc13783.o
 
 # LED SPI Drivers
 obj-$(CONFIG_LEDS_DAC124S085)		+= leds-dac124s085.o
diff --git a/drivers/leds/led-class.c b/drivers/leds/led-class.c
index 69e7d86a5143..260660076507 100644
--- a/drivers/leds/led-class.c
+++ b/drivers/leds/led-class.c
@@ -74,7 +74,7 @@ static ssize_t led_max_brightness_show(struct device *dev,
 
 static struct device_attribute led_class_attrs[] = {
 	__ATTR(brightness, 0644, led_brightness_show, led_brightness_store),
-	__ATTR(max_brightness, 0644, led_max_brightness_show, NULL),
+	__ATTR(max_brightness, 0444, led_max_brightness_show, NULL),
 #ifdef CONFIG_LEDS_TRIGGERS
 	__ATTR(trigger, 0644, led_trigger_show, led_trigger_store),
 #endif
diff --git a/drivers/leds/leds-88pm860x.c b/drivers/leds/leds-88pm860x.c
index 16a60c06c96c..b7677106cff8 100644
--- a/drivers/leds/leds-88pm860x.c
+++ b/drivers/leds/leds-88pm860x.c
@@ -256,8 +256,10 @@ static int pm860x_led_probe(struct platform_device *pdev)
 	if (pdev->dev.parent->platform_data) {
 		pm860x_pdata = pdev->dev.parent->platform_data;
 		pdata = pm860x_pdata->led;
-	} else
-		pdata = NULL;
+	} else {
+		dev_err(&pdev->dev, "missing platform data\n");
+		return -EINVAL;
+	}
 
 	data = kzalloc(sizeof(struct pm860x_led), GFP_KERNEL);
 	if (data == NULL)
@@ -268,8 +270,11 @@ static int pm860x_led_probe(struct platform_device *pdev)
 	data->i2c = (chip->id == CHIP_PM8606) ? chip->client : chip->companion;
 	data->iset = pdata->iset;
 	data->port = __check_device(pdata, data->name);
-	if (data->port < 0)
+	if (data->port < 0) {
+		dev_err(&pdev->dev, "check device failed\n");
+		kfree(data);
 		return -EINVAL;
+	}
 
 	data->current_brightness = 0;
 	data->cdev.name = data->name;
diff --git a/drivers/leds/leds-bd2802.c b/drivers/leds/leds-bd2802.c
index 286b501a3573..5dcdf9d69b3a 100644
--- a/drivers/leds/leds-bd2802.c
+++ b/drivers/leds/leds-bd2802.c
@@ -742,7 +742,6 @@ failed_unregister_dev_file:
 	for (i--; i >= 0; i--)
 		device_remove_file(&led->client->dev, bd2802_attributes[i]);
 failed_free:
-	i2c_set_clientdata(client, NULL);
 	kfree(led);
 
 	return ret;
@@ -759,7 +758,6 @@ static int __exit bd2802_remove(struct i2c_client *client)
 		bd2802_disable_adv_conf(led);
 	for (i = 0; i < ARRAY_SIZE(bd2802_attributes); i++)
 		device_remove_file(&led->client->dev, bd2802_attributes[i]);
-	i2c_set_clientdata(client, NULL);
 	kfree(led);
 
 	return 0;
diff --git a/drivers/leds/leds-gpio.c b/drivers/leds/leds-gpio.c
index c6e4b772b757..cc22eeefa10b 100644
--- a/drivers/leds/leds-gpio.c
+++ b/drivers/leds/leds-gpio.c
@@ -26,7 +26,8 @@ struct gpio_led_data {
 	u8 new_level;
 	u8 can_sleep;
 	u8 active_low;
-	int (*platform_gpio_blink_set)(unsigned gpio,
+	u8 blinking;
+	int (*platform_gpio_blink_set)(unsigned gpio, int state,
 			unsigned long *delay_on, unsigned long *delay_off);
 };
 
@@ -35,7 +36,13 @@ static void gpio_led_work(struct work_struct *work)
 	struct gpio_led_data	*led_dat =
 		container_of(work, struct gpio_led_data, work);
 
-	gpio_set_value_cansleep(led_dat->gpio, led_dat->new_level);
+	if (led_dat->blinking) {
+		led_dat->platform_gpio_blink_set(led_dat->gpio,
+						 led_dat->new_level,
+						 NULL, NULL);
+		led_dat->blinking = 0;
+	} else
+		gpio_set_value_cansleep(led_dat->gpio, led_dat->new_level);
 }
 
 static void gpio_led_set(struct led_classdev *led_cdev,
@@ -60,8 +67,14 @@ static void gpio_led_set(struct led_classdev *led_cdev,
 	if (led_dat->can_sleep) {
 		led_dat->new_level = level;
 		schedule_work(&led_dat->work);
-	} else
-		gpio_set_value(led_dat->gpio, level);
+	} else {
+		if (led_dat->blinking) {
+			led_dat->platform_gpio_blink_set(led_dat->gpio, level,
+							 NULL, NULL);
+			led_dat->blinking = 0;
+		} else
+			gpio_set_value(led_dat->gpio, level);
+	}
 }
 
 static int gpio_blink_set(struct led_classdev *led_cdev,
@@ -70,12 +83,14 @@ static int gpio_blink_set(struct led_classdev *led_cdev,
 	struct gpio_led_data *led_dat =
 		container_of(led_cdev, struct gpio_led_data, cdev);
 
-	return led_dat->platform_gpio_blink_set(led_dat->gpio, delay_on, delay_off);
+	led_dat->blinking = 1;
+	return led_dat->platform_gpio_blink_set(led_dat->gpio, GPIO_LED_BLINK,
+						delay_on, delay_off);
 }
 
 static int __devinit create_gpio_led(const struct gpio_led *template,
 	struct gpio_led_data *led_dat, struct device *parent,
-	int (*blink_set)(unsigned, unsigned long *, unsigned long *))
+	int (*blink_set)(unsigned, int, unsigned long *, unsigned long *))
 {
 	int ret, state;
 
@@ -97,6 +112,7 @@ static int __devinit create_gpio_led(const struct gpio_led *template,
 	led_dat->gpio = template->gpio;
 	led_dat->can_sleep = gpio_cansleep(template->gpio);
 	led_dat->active_low = template->active_low;
+	led_dat->blinking = 0;
 	if (blink_set) {
 		led_dat->platform_gpio_blink_set = blink_set;
 		led_dat->cdev.blink_set = gpio_blink_set;
@@ -113,7 +129,7 @@ static int __devinit create_gpio_led(const struct gpio_led *template,
 	ret = gpio_direction_output(led_dat->gpio, led_dat->active_low ^ state);
 	if (ret < 0)
 		goto err;
-
+		
 	INIT_WORK(&led_dat->work, gpio_led_work);
 
 	ret = led_classdev_register(parent, &led_dat->cdev);
@@ -211,7 +227,7 @@ struct gpio_led_of_platform_data {
 static int __devinit of_gpio_leds_probe(struct of_device *ofdev,
 					const struct of_device_id *match)
 {
-	struct device_node *np = ofdev->node, *child;
+	struct device_node *np = ofdev->dev.of_node, *child;
 	struct gpio_led_of_platform_data *pdata;
 	int count = 0, ret;
 
@@ -291,8 +307,8 @@ static struct of_platform_driver of_gpio_leds_driver = {
 	.driver = {
 		.name = "of_gpio_leds",
 		.owner = THIS_MODULE,
+		.of_match_table = of_gpio_leds_match,
 	},
-	.match_table = of_gpio_leds_match,
 	.probe = of_gpio_leds_probe,
 	.remove = __devexit_p(of_gpio_leds_remove),
 };
diff --git a/drivers/leds/leds-lp3944.c b/drivers/leds/leds-lp3944.c
index 8d5ecceba181..9010c054615e 100644
--- a/drivers/leds/leds-lp3944.c
+++ b/drivers/leds/leds-lp3944.c
@@ -379,6 +379,7 @@ static int __devinit lp3944_probe(struct i2c_client *client,
 {
 	struct lp3944_platform_data *lp3944_pdata = client->dev.platform_data;
 	struct lp3944_data *data;
+	int err;
 
 	if (lp3944_pdata == NULL) {
 		dev_err(&client->dev, "no platform data\n");
@@ -401,9 +402,13 @@ static int __devinit lp3944_probe(struct i2c_client *client,
 
 	mutex_init(&data->lock);
 
-	dev_info(&client->dev, "lp3944 enabled\n");
+	err = lp3944_configure(client, data, lp3944_pdata);
+	if (err < 0) {
+		kfree(data);
+		return err;
+	}
 
-	lp3944_configure(client, data, lp3944_pdata);
+	dev_info(&client->dev, "lp3944 enabled\n");
 	return 0;
 }
 
@@ -427,7 +432,6 @@ static int __devexit lp3944_remove(struct i2c_client *client)
 		}
 
 	kfree(data);
-	i2c_set_clientdata(client, NULL);
 
 	return 0;
 }
diff --git a/drivers/leds/leds-mc13783.c b/drivers/leds/leds-mc13783.c
new file mode 100644
index 000000000000..f05bb08d0f09
--- /dev/null
+++ b/drivers/leds/leds-mc13783.c
@@ -0,0 +1,403 @@
+/*
+ * LEDs driver for Freescale MC13783
+ *
+ * Copyright (C) 2010 Philippe Rétornaz
+ *
+ * Based on leds-da903x:
+ * Copyright (C) 2008 Compulab, Ltd.
+ *      Mike Rapoport <mike@compulab.co.il>
+ *
+ * Copyright (C) 2006-2008 Marvell International Ltd.
+ *      Eric Miao <eric.miao@marvell.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/platform_device.h>
+#include <linux/leds.h>
+#include <linux/workqueue.h>
+#include <linux/mfd/mc13783.h>
+#include <linux/slab.h>
+
+struct mc13783_led {
+	struct led_classdev	cdev;
+	struct work_struct	work;
+	struct mc13783		*master;
+	enum led_brightness	new_brightness;
+	int			id;
+};
+
+#define MC13783_REG_LED_CONTROL_0	51
+#define MC13783_LED_C0_ENABLE_BIT	(1 << 0)
+#define MC13783_LED_C0_TRIODE_MD_BIT	(1 << 7)
+#define MC13783_LED_C0_TRIODE_AD_BIT	(1 << 8)
+#define MC13783_LED_C0_TRIODE_KP_BIT	(1 << 9)
+#define MC13783_LED_C0_BOOST_BIT	(1 << 10)
+#define MC13783_LED_C0_ABMODE_MASK	0x7
+#define MC13783_LED_C0_ABMODE		11
+#define MC13783_LED_C0_ABREF_MASK	0x3
+#define MC13783_LED_C0_ABREF		14
+
+#define MC13783_REG_LED_CONTROL_1	52
+#define MC13783_LED_C1_TC1HALF_BIT	(1 << 18)
+
+#define MC13783_REG_LED_CONTROL_2	53
+#define MC13783_LED_C2_BL_P_MASK	0xf
+#define MC13783_LED_C2_MD_P		9
+#define MC13783_LED_C2_AD_P		13
+#define MC13783_LED_C2_KP_P		17
+#define MC13783_LED_C2_BL_C_MASK	0x7
+#define MC13783_LED_C2_MD_C		0
+#define MC13783_LED_C2_AD_C		3
+#define MC13783_LED_C2_KP_C		6
+
+#define MC13783_REG_LED_CONTROL_3	54
+#define MC13783_LED_C3_TC_P		6
+#define MC13783_LED_C3_TC_P_MASK	0x1f
+
+#define MC13783_REG_LED_CONTROL_4	55
+#define MC13783_REG_LED_CONTROL_5	56
+
+#define MC13783_LED_Cx_PERIOD		21
+#define MC13783_LED_Cx_PERIOD_MASK	0x3
+#define MC13783_LED_Cx_SLEWLIM_BIT      (1 << 23)
+#define MC13783_LED_Cx_TRIODE_TC_BIT	(1 << 23)
+#define MC13783_LED_Cx_TC_C_MASK	0x3
+
+static void mc13783_led_work(struct work_struct *work)
+{
+	struct mc13783_led *led = container_of(work, struct mc13783_led, work);
+	int reg = 0;
+	int mask = 0;
+	int value = 0;
+	int bank, off, shift;
+
+	switch (led->id) {
+	case MC13783_LED_MD:
+		reg = MC13783_REG_LED_CONTROL_2;
+		mask = MC13783_LED_C2_BL_P_MASK << MC13783_LED_C2_MD_P;
+		value = (led->new_brightness >> 4) << MC13783_LED_C2_MD_P;
+		break;
+	case MC13783_LED_AD:
+		reg = MC13783_REG_LED_CONTROL_2;
+		mask = MC13783_LED_C2_BL_P_MASK << MC13783_LED_C2_AD_P;
+		value = (led->new_brightness >> 4) << MC13783_LED_C2_AD_P;
+		break;
+	case MC13783_LED_KP:
+		reg = MC13783_REG_LED_CONTROL_2;
+		mask = MC13783_LED_C2_BL_P_MASK << MC13783_LED_C2_KP_P;
+		value = (led->new_brightness >> 4) << MC13783_LED_C2_KP_P;
+		break;
+	case MC13783_LED_R1:
+	case MC13783_LED_G1:
+	case MC13783_LED_B1:
+	case MC13783_LED_R2:
+	case MC13783_LED_G2:
+	case MC13783_LED_B2:
+	case MC13783_LED_R3:
+	case MC13783_LED_G3:
+	case MC13783_LED_B3:
+		off = led->id - MC13783_LED_R1;
+		bank = off/3;
+		reg = MC13783_REG_LED_CONTROL_3 + off/3;
+		shift = (off - bank * 3) * 5 + MC13783_LED_C3_TC_P;
+		value = (led->new_brightness >> 3) << shift;
+		mask = MC13783_LED_C3_TC_P_MASK << shift;
+		break;
+	}
+
+	mc13783_lock(led->master);
+
+	mc13783_reg_rmw(led->master, reg, mask, value);
+
+	mc13783_unlock(led->master);
+}
+
+static void mc13783_led_set(struct led_classdev *led_cdev,
+			   enum led_brightness value)
+{
+	struct mc13783_led *led;
+
+	led = container_of(led_cdev, struct mc13783_led, cdev);
+	led->new_brightness = value;
+	schedule_work(&led->work);
+}
+
+static int __devinit mc13783_led_setup(struct mc13783_led *led, int max_current)
+{
+	int shift = 0;
+	int mask = 0;
+	int value = 0;
+	int reg = 0;
+	int ret, bank;
+
+	switch (led->id) {
+	case MC13783_LED_MD:
+		shift = MC13783_LED_C2_MD_C;
+		mask = MC13783_LED_C2_BL_C_MASK;
+		value = max_current & MC13783_LED_C2_BL_C_MASK;
+		reg = MC13783_REG_LED_CONTROL_2;
+		break;
+	case MC13783_LED_AD:
+		shift = MC13783_LED_C2_AD_C;
+		mask = MC13783_LED_C2_BL_C_MASK;
+		value = max_current & MC13783_LED_C2_BL_C_MASK;
+		reg = MC13783_REG_LED_CONTROL_2;
+		break;
+	case MC13783_LED_KP:
+		shift = MC13783_LED_C2_KP_C;
+		mask = MC13783_LED_C2_BL_C_MASK;
+		value = max_current & MC13783_LED_C2_BL_C_MASK;
+		reg = MC13783_REG_LED_CONTROL_2;
+		break;
+	case MC13783_LED_R1:
+	case MC13783_LED_G1:
+	case MC13783_LED_B1:
+	case MC13783_LED_R2:
+	case MC13783_LED_G2:
+	case MC13783_LED_B2:
+	case MC13783_LED_R3:
+	case MC13783_LED_G3:
+	case MC13783_LED_B3:
+		bank = (led->id - MC13783_LED_R1)/3;
+		reg = MC13783_REG_LED_CONTROL_3 + bank;
+		shift = ((led->id - MC13783_LED_R1) - bank * 3) * 2;
+		mask = MC13783_LED_Cx_TC_C_MASK;
+		value = max_current & MC13783_LED_Cx_TC_C_MASK;
+		break;
+	}
+
+	mc13783_lock(led->master);
+
+	ret = mc13783_reg_rmw(led->master, reg, mask << shift,
+						value << shift);
+
+	mc13783_unlock(led->master);
+	return ret;
+}
+
+static int __devinit mc13783_leds_prepare(struct platform_device *pdev)
+{
+	struct mc13783_leds_platform_data *pdata = dev_get_platdata(&pdev->dev);
+	struct mc13783 *dev = dev_get_drvdata(pdev->dev.parent);
+	int ret = 0;
+	int reg = 0;
+
+	mc13783_lock(dev);
+
+	if (pdata->flags & MC13783_LED_TC1HALF)
+		reg |= MC13783_LED_C1_TC1HALF_BIT;
+
+	if (pdata->flags & MC13783_LED_SLEWLIMTC)
+		reg |= MC13783_LED_Cx_SLEWLIM_BIT;
+
+	ret = mc13783_reg_write(dev, MC13783_REG_LED_CONTROL_1, reg);
+	if (ret)
+		goto out;
+
+	reg = (pdata->bl_period & MC13783_LED_Cx_PERIOD_MASK) <<
+							MC13783_LED_Cx_PERIOD;
+
+	if (pdata->flags & MC13783_LED_SLEWLIMBL)
+		reg |= MC13783_LED_Cx_SLEWLIM_BIT;
+
+	ret = mc13783_reg_write(dev, MC13783_REG_LED_CONTROL_2, reg);
+	if (ret)
+		goto out;
+
+	reg = (pdata->tc1_period & MC13783_LED_Cx_PERIOD_MASK) <<
+							MC13783_LED_Cx_PERIOD;
+
+	if (pdata->flags & MC13783_LED_TRIODE_TC1)
+		reg |= MC13783_LED_Cx_TRIODE_TC_BIT;
+
+	ret = mc13783_reg_write(dev, MC13783_REG_LED_CONTROL_3, reg);
+	if (ret)
+		goto out;
+
+	reg = (pdata->tc2_period & MC13783_LED_Cx_PERIOD_MASK) <<
+							MC13783_LED_Cx_PERIOD;
+
+	if (pdata->flags & MC13783_LED_TRIODE_TC2)
+		reg |= MC13783_LED_Cx_TRIODE_TC_BIT;
+
+	ret = mc13783_reg_write(dev, MC13783_REG_LED_CONTROL_4, reg);
+	if (ret)
+		goto out;
+
+	reg = (pdata->tc3_period & MC13783_LED_Cx_PERIOD_MASK) <<
+							MC13783_LED_Cx_PERIOD;
+
+	if (pdata->flags & MC13783_LED_TRIODE_TC3)
+		reg |= MC13783_LED_Cx_TRIODE_TC_BIT;;
+
+	ret = mc13783_reg_write(dev, MC13783_REG_LED_CONTROL_5, reg);
+	if (ret)
+		goto out;
+
+	reg = MC13783_LED_C0_ENABLE_BIT;
+	if (pdata->flags & MC13783_LED_TRIODE_MD)
+		reg |= MC13783_LED_C0_TRIODE_MD_BIT;
+	if (pdata->flags & MC13783_LED_TRIODE_AD)
+		reg |= MC13783_LED_C0_TRIODE_AD_BIT;
+	if (pdata->flags & MC13783_LED_TRIODE_KP)
+		reg |= MC13783_LED_C0_TRIODE_KP_BIT;
+	if (pdata->flags & MC13783_LED_BOOST_EN)
+		reg |= MC13783_LED_C0_BOOST_BIT;
+
+	reg |= (pdata->abmode & MC13783_LED_C0_ABMODE_MASK) <<
+							MC13783_LED_C0_ABMODE;
+	reg |= (pdata->abref & MC13783_LED_C0_ABREF_MASK) <<
+							MC13783_LED_C0_ABREF;
+
+	ret = mc13783_reg_write(dev, MC13783_REG_LED_CONTROL_0, reg);
+
+out:
+	mc13783_unlock(dev);
+	return ret;
+}
+
+static int __devinit mc13783_led_probe(struct platform_device *pdev)
+{
+	struct mc13783_leds_platform_data *pdata = dev_get_platdata(&pdev->dev);
+	struct mc13783_led_platform_data *led_cur;
+	struct mc13783_led *led, *led_dat;
+	int ret, i;
+	int init_led = 0;
+
+	if (pdata == NULL) {
+		dev_err(&pdev->dev, "missing platform data\n");
+		return -ENODEV;
+	}
+
+	if (pdata->num_leds < 1 || pdata->num_leds > MC13783_LED_MAX) {
+		dev_err(&pdev->dev, "Invalid led count %d\n", pdata->num_leds);
+		return -EINVAL;
+	}
+
+	led = kzalloc(sizeof(*led) * pdata->num_leds, GFP_KERNEL);
+	if (led == NULL) {
+		dev_err(&pdev->dev, "failed to alloc memory\n");
+		return -ENOMEM;
+	}
+
+	ret = mc13783_leds_prepare(pdev);
+	if (ret) {
+		dev_err(&pdev->dev, "unable to init led driver\n");
+		goto err_free;
+	}
+
+	for (i = 0; i < pdata->num_leds; i++) {
+		led_dat = &led[i];
+		led_cur = &pdata->led[i];
+
+		if (led_cur->id > MC13783_LED_MAX || led_cur->id < 0) {
+			dev_err(&pdev->dev, "invalid id %d\n", led_cur->id);
+			ret = -EINVAL;
+			goto err_register;
+		}
+
+		if (init_led & (1 << led_cur->id)) {
+			dev_err(&pdev->dev, "led %d already initialized\n",
+					led_cur->id);
+			ret = -EINVAL;
+			goto err_register;
+		}
+
+		init_led |= 1 << led_cur->id;
+		led_dat->cdev.name = led_cur->name;
+		led_dat->cdev.default_trigger = led_cur->default_trigger;
+		led_dat->cdev.brightness_set = mc13783_led_set;
+		led_dat->cdev.brightness = LED_OFF;
+		led_dat->id = led_cur->id;
+		led_dat->master = dev_get_drvdata(pdev->dev.parent);
+
+		INIT_WORK(&led_dat->work, mc13783_led_work);
+
+		ret = led_classdev_register(pdev->dev.parent, &led_dat->cdev);
+		if (ret) {
+			dev_err(&pdev->dev, "failed to register led %d\n",
+					led_dat->id);
+			goto err_register;
+		}
+
+		ret = mc13783_led_setup(led_dat, led_cur->max_current);
+		if (ret) {
+			dev_err(&pdev->dev, "unable to init led %d\n",
+					led_dat->id);
+			i++;
+			goto err_register;
+		}
+	}
+
+	platform_set_drvdata(pdev, led);
+	return 0;
+
+err_register:
+	for (i = i - 1; i >= 0; i--) {
+		led_classdev_unregister(&led[i].cdev);
+		cancel_work_sync(&led[i].work);
+	}
+
+err_free:
+	kfree(led);
+	return ret;
+}
+
+static int __devexit mc13783_led_remove(struct platform_device *pdev)
+{
+	struct mc13783_leds_platform_data *pdata = dev_get_platdata(&pdev->dev);
+	struct mc13783_led *led = platform_get_drvdata(pdev);
+	struct mc13783 *dev = dev_get_drvdata(pdev->dev.parent);
+	int i;
+
+	for (i = 0; i < pdata->num_leds; i++) {
+		led_classdev_unregister(&led[i].cdev);
+		cancel_work_sync(&led[i].work);
+	}
+
+	mc13783_lock(dev);
+
+	mc13783_reg_write(dev, MC13783_REG_LED_CONTROL_0, 0);
+	mc13783_reg_write(dev, MC13783_REG_LED_CONTROL_1, 0);
+	mc13783_reg_write(dev, MC13783_REG_LED_CONTROL_2, 0);
+	mc13783_reg_write(dev, MC13783_REG_LED_CONTROL_3, 0);
+	mc13783_reg_write(dev, MC13783_REG_LED_CONTROL_4, 0);
+	mc13783_reg_write(dev, MC13783_REG_LED_CONTROL_5, 0);
+
+	mc13783_unlock(dev);
+
+	kfree(led);
+	return 0;
+}
+
+static struct platform_driver mc13783_led_driver = {
+	.driver	= {
+		.name	= "mc13783-led",
+		.owner	= THIS_MODULE,
+	},
+	.probe		= mc13783_led_probe,
+	.remove		= __devexit_p(mc13783_led_remove),
+};
+
+static int __init mc13783_led_init(void)
+{
+	return platform_driver_register(&mc13783_led_driver);
+}
+module_init(mc13783_led_init);
+
+static void __exit mc13783_led_exit(void)
+{
+	platform_driver_unregister(&mc13783_led_driver);
+}
+module_exit(mc13783_led_exit);
+
+MODULE_DESCRIPTION("LEDs driver for Freescale MC13783 PMIC");
+MODULE_AUTHOR("Philippe Retornaz <philippe.retornaz@epfl.ch>");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS("platform:mc13783-led");
diff --git a/drivers/leds/leds-net5501.c b/drivers/leds/leds-net5501.c
new file mode 100644
index 000000000000..3063f591f0dc
--- /dev/null
+++ b/drivers/leds/leds-net5501.c
@@ -0,0 +1,94 @@
+/*
+ * Soekris board support code
+ *
+ * Copyright (C) 2008-2009 Tower Technologies
+ * Written by Alessandro Zummo <a.zummo@towertech.it>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2
+ * as published by the Free Software Foundation.
+ */
+
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/io.h>
+#include <linux/string.h>
+#include <linux/leds.h>
+#include <linux/platform_device.h>
+#include <linux/gpio.h>
+
+#include <asm/geode.h>
+
+static struct gpio_led net5501_leds[] = {
+	{
+		.name = "error",
+		.gpio = 6,
+		.default_trigger = "default-on",
+	},
+};
+
+static struct gpio_led_platform_data net5501_leds_data = {
+	.num_leds = ARRAY_SIZE(net5501_leds),
+	.leds = net5501_leds,
+};
+
+static struct platform_device net5501_leds_dev = {
+	.name = "leds-gpio",
+	.id = -1,
+	.dev.platform_data = &net5501_leds_data,
+};
+
+static void __init init_net5501(void)
+{
+	platform_device_register(&net5501_leds_dev);
+}
+
+struct soekris_board {
+	u16	offset;
+	char	*sig;
+	u8	len;
+	void	(*init)(void);
+};
+
+static struct soekris_board __initdata boards[] = {
+	{ 0xb7b, "net5501", 7, init_net5501 },	/* net5501 v1.33/1.33c */
+	{ 0xb1f, "net5501", 7, init_net5501 },	/* net5501 v1.32i */
+};
+
+static int __init soekris_init(void)
+{
+	int i;
+	unsigned char *rombase, *bios;
+
+	if (!is_geode())
+		return 0;
+
+	rombase = ioremap(0xffff0000, 0xffff);
+	if (!rombase) {
+		printk(KERN_INFO "Soekris net5501 LED driver failed to get rombase");
+		return 0;
+	}
+
+	bios = rombase + 0x20;	/* null terminated */
+
+	if (strncmp(bios, "comBIOS", 7))
+		goto unmap;
+
+	for (i = 0; i < ARRAY_SIZE(boards); i++) {
+		unsigned char *model = rombase + boards[i].offset;
+
+		if (strncmp(model, boards[i].sig, boards[i].len) == 0) {
+			printk(KERN_INFO "Soekris %s: %s\n", model, bios);
+
+			if (boards[i].init)
+				boards[i].init();
+			break;
+		}
+	}
+
+unmap:
+	iounmap(rombase);
+	return 0;
+}
+
+arch_initcall(soekris_init);
diff --git a/drivers/leds/leds-pca9532.c b/drivers/leds/leds-pca9532.c
index 6682175fa9f7..43d08756d823 100644
--- a/drivers/leds/leds-pca9532.c
+++ b/drivers/leds/leds-pca9532.c
@@ -320,10 +320,8 @@ static int pca9532_probe(struct i2c_client *client,
 	mutex_init(&data->update_lock);
 
 	err = pca9532_configure(client, data, pca9532_pdata);
-	if (err) {
+	if (err)
 		kfree(data);
-		i2c_set_clientdata(client, NULL);
-	}
 
 	return err;
 }
@@ -351,7 +349,6 @@ static int pca9532_remove(struct i2c_client *client)
 		}
 
 	kfree(data);
-	i2c_set_clientdata(client, NULL);
 	return 0;
 }
 
diff --git a/drivers/leds/leds-pca955x.c b/drivers/leds/leds-pca955x.c
index 8ff50f234190..66aa3e8e786f 100644
--- a/drivers/leds/leds-pca955x.c
+++ b/drivers/leds/leds-pca955x.c
@@ -342,7 +342,6 @@ exit:
 	}
 
 	kfree(pca955x);
-	i2c_set_clientdata(client, NULL);
 
 	return err;
 }
@@ -358,7 +357,6 @@ static int __devexit pca955x_remove(struct i2c_client *client)
 	}
 
 	kfree(pca955x);
-	i2c_set_clientdata(client, NULL);
 
 	return 0;
 }
diff --git a/drivers/leds/leds-ss4200.c b/drivers/leds/leds-ss4200.c
index 51477ec71391..a688293abd0b 100644
--- a/drivers/leds/leds-ss4200.c
+++ b/drivers/leds/leds-ss4200.c
@@ -534,7 +534,7 @@ static int __init nas_gpio_init(void)
 	set_power_light_amber_noblink();
 	return 0;
 out_err:
-	for (; i >= 0; i--)
+	for (i--; i >= 0; i--)
 		unregister_nasgpio_led(i);
 	pci_unregister_driver(&nas_gpio_pci_driver);
 	return ret;
diff --git a/drivers/macintosh/macio_asic.c b/drivers/macintosh/macio_asic.c
index 26a303a1d1ab..b6e7ddc09d76 100644
--- a/drivers/macintosh/macio_asic.c
+++ b/drivers/macintosh/macio_asic.c
@@ -39,14 +39,12 @@ static struct macio_chip      *macio_on_hold;
 
 static int macio_bus_match(struct device *dev, struct device_driver *drv) 
 {
-	struct macio_dev * macio_dev = to_macio_device(dev);
-	struct macio_driver * macio_drv = to_macio_driver(drv);
-	const struct of_device_id * matches = macio_drv->match_table;
+	const struct of_device_id * matches = drv->of_match_table;
 
 	if (!matches) 
 		return 0;
 
-	return of_match_device(matches, &macio_dev->ofdev) != NULL;
+	return of_match_device(matches, dev) != NULL;
 }
 
 struct macio_dev *macio_dev_get(struct macio_dev *dev)
@@ -84,7 +82,7 @@ static int macio_device_probe(struct device *dev)
 
 	macio_dev_get(macio_dev);
 
-	match = of_match_device(drv->match_table, &macio_dev->ofdev);
+	match = of_match_device(drv->driver.of_match_table, dev);
 	if (match)
 		error = drv->probe(macio_dev, match);
 	if (error)
@@ -248,7 +246,7 @@ static void macio_create_fixup_irq(struct macio_dev *dev, int index,
 
 static void macio_add_missing_resources(struct macio_dev *dev)
 {
-	struct device_node *np = dev->ofdev.node;
+	struct device_node *np = dev->ofdev.dev.of_node;
 	unsigned int irq_base;
 
 	/* Gatwick has some missing interrupts on child nodes */
@@ -289,7 +287,7 @@ static void macio_add_missing_resources(struct macio_dev *dev)
 
 static void macio_setup_interrupts(struct macio_dev *dev)
 {
-	struct device_node *np = dev->ofdev.node;
+	struct device_node *np = dev->ofdev.dev.of_node;
 	unsigned int irq;
 	int i = 0, j = 0;
 
@@ -317,7 +315,7 @@ static void macio_setup_interrupts(struct macio_dev *dev)
 static void macio_setup_resources(struct macio_dev *dev,
 				  struct resource *parent_res)
 {
-	struct device_node *np = dev->ofdev.node;
+	struct device_node *np = dev->ofdev.dev.of_node;
 	struct resource r;
 	int index;
 
@@ -373,9 +371,9 @@ static struct macio_dev * macio_add_one_device(struct macio_chip *chip,
 
 	dev->bus = &chip->lbus;
 	dev->media_bay = in_bay;
-	dev->ofdev.node = np;
-	dev->ofdev.dma_mask = 0xffffffffUL;
-	dev->ofdev.dev.dma_mask = &dev->ofdev.dma_mask;
+	dev->ofdev.dev.of_node = np;
+	dev->ofdev.archdata.dma_mask = 0xffffffffUL;
+	dev->ofdev.dev.dma_mask = &dev->ofdev.archdata.dma_mask;
 	dev->ofdev.dev.parent = parent;
 	dev->ofdev.dev.bus = &macio_bus_type;
 	dev->ofdev.dev.release = macio_release_dev;
@@ -494,9 +492,9 @@ static void macio_pci_add_devices(struct macio_chip *chip)
 	}
 
 	/* Add media bay devices if any */
-	if (mbdev)
-		for (np = NULL; (np = of_get_next_child(mbdev->ofdev.node, np))
-			     != NULL;) {
+	if (mbdev) {
+		pnode = mbdev->ofdev.dev.of_node;
+		for (np = NULL; (np = of_get_next_child(pnode, np)) != NULL;) {
 			if (macio_skip_device(np))
 				continue;
 			of_node_get(np);
@@ -504,11 +502,12 @@ static void macio_pci_add_devices(struct macio_chip *chip)
 						 mbdev,  root_res) == NULL)
 				of_node_put(np);
 		}
+	}
 
 	/* Add serial ports if any */
 	if (sdev) {
-		for (np = NULL; (np = of_get_next_child(sdev->ofdev.node, np))
-			     != NULL;) {
+		pnode = sdev->ofdev.dev.of_node;
+		for (np = NULL; (np = of_get_next_child(pnode, np)) != NULL;) {
 			if (macio_skip_device(np))
 				continue;
 			of_node_get(np);
@@ -527,7 +526,6 @@ static void macio_pci_add_devices(struct macio_chip *chip)
 int macio_register_driver(struct macio_driver *drv)
 {
 	/* initialize common driver fields */
-	drv->driver.name = drv->name;
 	drv->driver.bus = &macio_bus_type;
 
 	/* register with core */
diff --git a/drivers/macintosh/macio_sysfs.c b/drivers/macintosh/macio_sysfs.c
index 9e9453b58425..6999ce59fd10 100644
--- a/drivers/macintosh/macio_sysfs.c
+++ b/drivers/macintosh/macio_sysfs.c
@@ -9,7 +9,7 @@ field##_show (struct device *dev, struct device_attribute *attr,	\
               char *buf)						\
 {									\
 	struct macio_dev *mdev = to_macio_device (dev);			\
-	return sprintf (buf, format_string, mdev->ofdev.node->field);	\
+	return sprintf (buf, format_string, mdev->ofdev.dev.of_node->field); \
 }
 
 static ssize_t
@@ -21,7 +21,7 @@ compatible_show (struct device *dev, struct device_attribute *attr, char *buf)
 	int length = 0;
 
 	of = &to_macio_device (dev)->ofdev;
-	compat = of_get_property(of->node, "compatible", &cplen);
+	compat = of_get_property(of->dev.of_node, "compatible", &cplen);
 	if (!compat) {
 		*buf = '\0';
 		return 0;
@@ -58,7 +58,7 @@ static ssize_t devspec_show(struct device *dev,
 	struct of_device *ofdev;
 
 	ofdev = to_of_device(dev);
-	return sprintf(buf, "%s\n", ofdev->node->full_name);
+	return sprintf(buf, "%s\n", ofdev->dev.of_node->full_name);
 }
 
 macio_config_of_attr (name, "%s\n");
diff --git a/drivers/macintosh/mediabay.c b/drivers/macintosh/mediabay.c
index 08002b88f342..2fd435bc542e 100644
--- a/drivers/macintosh/mediabay.c
+++ b/drivers/macintosh/mediabay.c
@@ -564,7 +564,7 @@ static int __devinit media_bay_attach(struct macio_dev *mdev, const struct of_de
 	unsigned long base;
 	int i;
 
-	ofnode = mdev->ofdev.node;
+	ofnode = mdev->ofdev.dev.of_node;
 
 	if (macio_resource_count(mdev) < 1)
 		return -ENODEV;
@@ -728,8 +728,10 @@ static struct of_device_id media_bay_match[] =
 
 static struct macio_driver media_bay_driver =
 {
-	.name		= "media-bay",
-	.match_table	= media_bay_match,
+	.driver = {
+		.name		= "media-bay",
+		.of_match_table	= media_bay_match,
+	},
 	.probe		= media_bay_attach,
 	.suspend	= media_bay_suspend,
 	.resume		= media_bay_resume
diff --git a/drivers/macintosh/nvram.c b/drivers/macintosh/nvram.c
index c876349c32de..a271c8218d82 100644
--- a/drivers/macintosh/nvram.c
+++ b/drivers/macintosh/nvram.c
@@ -100,7 +100,7 @@ const struct file_operations nvram_fops = {
 	.llseek		= nvram_llseek,
 	.read		= read_nvram,
 	.write		= write_nvram,
-	.ioctl		= nvram_ioctl,
+	.unlocked_ioctl	= nvram_ioctl,
 };
 
 static struct miscdevice nvram_dev = {
diff --git a/drivers/macintosh/rack-meter.c b/drivers/macintosh/rack-meter.c
index 7c54d80c4fb2..53cce3a5da23 100644
--- a/drivers/macintosh/rack-meter.c
+++ b/drivers/macintosh/rack-meter.c
@@ -375,7 +375,7 @@ static int __devinit rackmeter_probe(struct macio_dev* mdev,
 	pr_debug("rackmeter_probe()\n");
 
 	/* Get i2s-a node */
-	while ((i2s = of_get_next_child(mdev->ofdev.node, i2s)) != NULL)
+	while ((i2s = of_get_next_child(mdev->ofdev.dev.of_node, i2s)) != NULL)
 	       if (strcmp(i2s->name, "i2s-a") == 0)
 		       break;
 	if (i2s == NULL) {
@@ -431,7 +431,7 @@ static int __devinit rackmeter_probe(struct macio_dev* mdev,
 	    of_address_to_resource(i2s, 1, &rdma)) {
 		printk(KERN_ERR
 		       "rackmeter: found match but lacks resources: %s",
-		       mdev->ofdev.node->full_name);
+		       mdev->ofdev.dev.of_node->full_name);
 		rc = -ENXIO;
 		goto bail_free;
 	}
@@ -584,9 +584,11 @@ static struct of_device_id rackmeter_match[] = {
 };
 
 static struct macio_driver rackmeter_driver = {
-	.name = "rackmeter",
-	.owner = THIS_MODULE,
-	.match_table = rackmeter_match,
+	.driver = {
+		.name = "rackmeter",
+		.owner = THIS_MODULE,
+		.of_match_table = rackmeter_match,
+	},
 	.probe = rackmeter_probe,
 	.remove = __devexit_p(rackmeter_remove),
 	.shutdown = rackmeter_shutdown,
diff --git a/drivers/macintosh/smu.c b/drivers/macintosh/smu.c
index c9da5c4c167d..2506c957712e 100644
--- a/drivers/macintosh/smu.c
+++ b/drivers/macintosh/smu.c
@@ -671,8 +671,11 @@ static const struct of_device_id smu_platform_match[] =
 
 static struct of_platform_driver smu_of_platform_driver =
 {
-	.name 		= "smu",
-	.match_table	= smu_platform_match,
+	.driver = {
+		.name = "smu",
+		.owner = THIS_MODULE,
+		.of_match_table = smu_platform_match,
+	},
 	.probe		= smu_platform_probe,
 };
 
diff --git a/drivers/macintosh/therm_adt746x.c b/drivers/macintosh/therm_adt746x.c
index 16d82f17ae82..c42eeb43042d 100644
--- a/drivers/macintosh/therm_adt746x.c
+++ b/drivers/macintosh/therm_adt746x.c
@@ -182,7 +182,6 @@ remove_thermostat(struct i2c_client *client)
 
 	thermostat = NULL;
 
-	i2c_set_clientdata(client, NULL);
 	kfree(th);
 
 	return 0;
@@ -400,7 +399,6 @@ static int probe_thermostat(struct i2c_client *client,
 	rc = read_reg(th, CONFIG_REG);
 	if (rc < 0) {
 		dev_err(&client->dev, "Thermostat failed to read config!\n");
-		i2c_set_clientdata(client, NULL);
 		kfree(th);
 		return -ENODEV;
 	}
diff --git a/drivers/macintosh/therm_pm72.c b/drivers/macintosh/therm_pm72.c
index b18fa948f3d1..e60605bd0ea9 100644
--- a/drivers/macintosh/therm_pm72.c
+++ b/drivers/macintosh/therm_pm72.c
@@ -2215,7 +2215,7 @@ static int fcu_of_probe(struct of_device* dev, const struct of_device_id *match)
 	state = state_detached;
 
 	/* Lookup the fans in the device tree */
-	fcu_lookup_fans(dev->node);
+	fcu_lookup_fans(dev->dev.of_node);
 
 	/* Add the driver */
 	return i2c_add_driver(&therm_pm72_driver);
@@ -2238,8 +2238,11 @@ static const struct of_device_id fcu_match[] =
 
 static struct of_platform_driver fcu_of_platform_driver = 
 {
-	.name 		= "temperature",
-	.match_table	= fcu_match,
+	.driver = {
+		.name = "temperature",
+		.owner = THIS_MODULE,
+		.of_match_table = fcu_match,
+	},
 	.probe		= fcu_of_probe,
 	.remove		= fcu_of_remove
 };
diff --git a/drivers/macintosh/therm_windtunnel.c b/drivers/macintosh/therm_windtunnel.c
index 0839770e4ec5..5c9367acf0cf 100644
--- a/drivers/macintosh/therm_windtunnel.c
+++ b/drivers/macintosh/therm_windtunnel.c
@@ -463,8 +463,11 @@ static const struct of_device_id therm_of_match[] = {{
 };
 
 static struct of_platform_driver therm_of_driver = {
-	.name		= "temperature",
-	.match_table	= therm_of_match,
+	.driver = {
+		.name = "temperature",
+		.owner = THIS_MODULE,
+		.of_match_table = therm_of_match,
+	},
 	.probe		= therm_of_probe,
 	.remove		= therm_of_remove,
 };
diff --git a/drivers/macintosh/via-pmu.c b/drivers/macintosh/via-pmu.c
index 42764849eb78..3d4fc0f7b00b 100644
--- a/drivers/macintosh/via-pmu.c
+++ b/drivers/macintosh/via-pmu.c
@@ -2273,8 +2273,7 @@ static int register_pmu_pm_ops(void)
 device_initcall(register_pmu_pm_ops);
 #endif
 
-static int
-pmu_ioctl(struct inode * inode, struct file *filp,
+static int pmu_ioctl(struct file *filp,
 		     u_int cmd, u_long arg)
 {
 	__u32 __user *argp = (__u32 __user *)arg;
@@ -2337,11 +2336,23 @@ pmu_ioctl(struct inode * inode, struct file *filp,
 	return error;
 }
 
+static long pmu_unlocked_ioctl(struct file *filp,
+			       u_int cmd, u_long arg)
+{
+	int ret;
+
+	lock_kernel();
+	ret = pmu_ioctl(filp, cmd, arg);
+	unlock_kernel();
+
+	return ret;
+}
+
 static const struct file_operations pmu_device_fops = {
 	.read		= pmu_read,
 	.write		= pmu_write,
 	.poll		= pmu_fpoll,
-	.ioctl		= pmu_ioctl,
+	.unlocked_ioctl	= pmu_unlocked_ioctl,
 	.open		= pmu_open,
 	.release	= pmu_release,
 };
diff --git a/drivers/macintosh/windfarm_lm75_sensor.c b/drivers/macintosh/windfarm_lm75_sensor.c
index d8257d35afde..647c6add2193 100644
--- a/drivers/macintosh/windfarm_lm75_sensor.c
+++ b/drivers/macintosh/windfarm_lm75_sensor.c
@@ -107,10 +107,8 @@ static int wf_lm75_probe(struct i2c_client *client,
 	i2c_set_clientdata(client, lm);
 
 	rc = wf_register_sensor(&lm->sens);
-	if (rc) {
-		i2c_set_clientdata(client, NULL);
+	if (rc)
 		kfree(lm);
-	}
 
 	return rc;
 }
@@ -216,7 +214,6 @@ static int wf_lm75_remove(struct i2c_client *client)
 	/* release sensor */
 	wf_unregister_sensor(&lm->sens);
 
-	i2c_set_clientdata(client, NULL);
 	return 0;
 }
 
diff --git a/drivers/macintosh/windfarm_max6690_sensor.c b/drivers/macintosh/windfarm_max6690_sensor.c
index b486eb929fde..8204113268f4 100644
--- a/drivers/macintosh/windfarm_max6690_sensor.c
+++ b/drivers/macintosh/windfarm_max6690_sensor.c
@@ -81,7 +81,6 @@ static int wf_max6690_probe(struct i2c_client *client,
 
 	rc = wf_register_sensor(&max->sens);
 	if (rc) {
-		i2c_set_clientdata(client, NULL);
 		kfree(max);
 	}
 
diff --git a/drivers/macintosh/windfarm_smu_sat.c b/drivers/macintosh/windfarm_smu_sat.c
index e20330a28959..65a8ff3e1f8e 100644
--- a/drivers/macintosh/windfarm_smu_sat.c
+++ b/drivers/macintosh/windfarm_smu_sat.c
@@ -376,7 +376,6 @@ static int wf_sat_remove(struct i2c_client *client)
 	/* XXX TODO */
 
 	sat->i2c = NULL;
-	i2c_set_clientdata(client, NULL);
 	return 0;
 }
 
diff --git a/drivers/md/Kconfig b/drivers/md/Kconfig
index acb3a4e404ff..4a6feac8c94a 100644
--- a/drivers/md/Kconfig
+++ b/drivers/md/Kconfig
@@ -100,8 +100,8 @@ config MD_RAID1
 	  If unsure, say Y.
 
 config MD_RAID10
-	tristate "RAID-10 (mirrored striping) mode (EXPERIMENTAL)"
-	depends on BLK_DEV_MD && EXPERIMENTAL
+	tristate "RAID-10 (mirrored striping) mode"
+	depends on BLK_DEV_MD
 	---help---
 	  RAID-10 provides a combination of striping (RAID-0) and
 	  mirroring (RAID-1) with easier configuration and more flexible
diff --git a/drivers/md/bitmap.c b/drivers/md/bitmap.c
index f084249295d9..1742435ce3ae 100644
--- a/drivers/md/bitmap.c
+++ b/drivers/md/bitmap.c
@@ -505,7 +505,7 @@ void bitmap_update_sb(struct bitmap *bitmap)
 		return;
 	}
 	spin_unlock_irqrestore(&bitmap->lock, flags);
-	sb = (bitmap_super_t *)kmap_atomic(bitmap->sb_page, KM_USER0);
+	sb = kmap_atomic(bitmap->sb_page, KM_USER0);
 	sb->events = cpu_to_le64(bitmap->mddev->events);
 	if (bitmap->mddev->events < bitmap->events_cleared) {
 		/* rocking back to read-only */
@@ -526,7 +526,7 @@ void bitmap_print_sb(struct bitmap *bitmap)
 
 	if (!bitmap || !bitmap->sb_page)
 		return;
-	sb = (bitmap_super_t *)kmap_atomic(bitmap->sb_page, KM_USER0);
+	sb = kmap_atomic(bitmap->sb_page, KM_USER0);
 	printk(KERN_DEBUG "%s: bitmap file superblock:\n", bmname(bitmap));
 	printk(KERN_DEBUG "         magic: %08x\n", le32_to_cpu(sb->magic));
 	printk(KERN_DEBUG "       version: %d\n", le32_to_cpu(sb->version));
@@ -575,7 +575,7 @@ static int bitmap_read_sb(struct bitmap *bitmap)
 		return err;
 	}
 
-	sb = (bitmap_super_t *)kmap_atomic(bitmap->sb_page, KM_USER0);
+	sb = kmap_atomic(bitmap->sb_page, KM_USER0);
 
 	chunksize = le32_to_cpu(sb->chunksize);
 	daemon_sleep = le32_to_cpu(sb->daemon_sleep) * HZ;
@@ -661,7 +661,7 @@ static int bitmap_mask_state(struct bitmap *bitmap, enum bitmap_state bits,
 		return 0;
 	}
 	spin_unlock_irqrestore(&bitmap->lock, flags);
-	sb = (bitmap_super_t *)kmap_atomic(bitmap->sb_page, KM_USER0);
+	sb = kmap_atomic(bitmap->sb_page, KM_USER0);
 	old = le32_to_cpu(sb->state) & bits;
 	switch (op) {
 		case MASK_SET: sb->state |= cpu_to_le32(bits);
@@ -1292,9 +1292,14 @@ int bitmap_startwrite(struct bitmap *bitmap, sector_t offset, unsigned long sect
 	if (!bitmap) return 0;
 
 	if (behind) {
+		int bw;
 		atomic_inc(&bitmap->behind_writes);
+		bw = atomic_read(&bitmap->behind_writes);
+		if (bw > bitmap->behind_writes_used)
+			bitmap->behind_writes_used = bw;
+
 		PRINTK(KERN_DEBUG "inc write-behind count %d/%d\n",
-		  atomic_read(&bitmap->behind_writes), bitmap->max_write_behind);
+		       bw, bitmap->max_write_behind);
 	}
 
 	while (sectors) {
@@ -1351,7 +1356,8 @@ void bitmap_endwrite(struct bitmap *bitmap, sector_t offset, unsigned long secto
 {
 	if (!bitmap) return;
 	if (behind) {
-		atomic_dec(&bitmap->behind_writes);
+		if (atomic_dec_and_test(&bitmap->behind_writes))
+			wake_up(&bitmap->behind_wait);
 		PRINTK(KERN_DEBUG "dec write-behind count %d/%d\n",
 		  atomic_read(&bitmap->behind_writes), bitmap->max_write_behind);
 	}
@@ -1675,6 +1681,7 @@ int bitmap_create(mddev_t *mddev)
 	atomic_set(&bitmap->pending_writes, 0);
 	init_waitqueue_head(&bitmap->write_wait);
 	init_waitqueue_head(&bitmap->overflow_wait);
+	init_waitqueue_head(&bitmap->behind_wait);
 
 	bitmap->mddev = mddev;
 
@@ -1692,7 +1699,7 @@ int bitmap_create(mddev_t *mddev)
 		 * and bypass the page cache, we must sync the file
 		 * first.
 		 */
-		vfs_fsync(file, file->f_dentry, 1);
+		vfs_fsync(file, 1);
 	}
 	/* read superblock from bitmap file (this sets mddev->bitmap_info.chunksize) */
 	if (!mddev->bitmap_info.external)
@@ -2006,6 +2013,27 @@ static ssize_t can_clear_store(mddev_t *mddev, const char *buf, size_t len)
 static struct md_sysfs_entry bitmap_can_clear =
 __ATTR(can_clear, S_IRUGO|S_IWUSR, can_clear_show, can_clear_store);
 
+static ssize_t
+behind_writes_used_show(mddev_t *mddev, char *page)
+{
+	if (mddev->bitmap == NULL)
+		return sprintf(page, "0\n");
+	return sprintf(page, "%lu\n",
+		       mddev->bitmap->behind_writes_used);
+}
+
+static ssize_t
+behind_writes_used_reset(mddev_t *mddev, const char *buf, size_t len)
+{
+	if (mddev->bitmap)
+		mddev->bitmap->behind_writes_used = 0;
+	return len;
+}
+
+static struct md_sysfs_entry max_backlog_used =
+__ATTR(max_backlog_used, S_IRUGO | S_IWUSR,
+       behind_writes_used_show, behind_writes_used_reset);
+
 static struct attribute *md_bitmap_attrs[] = {
 	&bitmap_location.attr,
 	&bitmap_timeout.attr,
@@ -2013,6 +2041,7 @@ static struct attribute *md_bitmap_attrs[] = {
 	&bitmap_chunksize.attr,
 	&bitmap_metadata.attr,
 	&bitmap_can_clear.attr,
+	&max_backlog_used.attr,
 	NULL
 };
 struct attribute_group md_bitmap_group = {
diff --git a/drivers/md/bitmap.h b/drivers/md/bitmap.h
index cb821d76d1b4..3797dea4723a 100644
--- a/drivers/md/bitmap.h
+++ b/drivers/md/bitmap.h
@@ -227,6 +227,7 @@ struct bitmap {
 	int allclean;
 
 	atomic_t behind_writes;
+	unsigned long behind_writes_used; /* highest actual value at runtime */
 
 	/*
 	 * the bitmap daemon - periodically wakes up and sweeps the bitmap
@@ -239,6 +240,7 @@ struct bitmap {
 	atomic_t pending_writes; /* pending writes to the bitmap file */
 	wait_queue_head_t write_wait;
 	wait_queue_head_t overflow_wait;
+	wait_queue_head_t behind_wait;
 
 	struct sysfs_dirent *sysfs_can_clear;
 };
diff --git a/drivers/md/faulty.c b/drivers/md/faulty.c
index 8e3850b98cca..1a8987884614 100644
--- a/drivers/md/faulty.c
+++ b/drivers/md/faulty.c
@@ -169,10 +169,9 @@ static void add_sector(conf_t *conf, sector_t start, int mode)
 		conf->nfaults = n+1;
 }
 
-static int make_request(struct request_queue *q, struct bio *bio)
+static int make_request(mddev_t *mddev, struct bio *bio)
 {
-	mddev_t *mddev = q->queuedata;
-	conf_t *conf = (conf_t*)mddev->private;
+	conf_t *conf = mddev->private;
 	int failit = 0;
 
 	if (bio_data_dir(bio) == WRITE) {
@@ -225,7 +224,7 @@ static int make_request(struct request_queue *q, struct bio *bio)
 
 static void status(struct seq_file *seq, mddev_t *mddev)
 {
-	conf_t *conf = (conf_t*)mddev->private;
+	conf_t *conf = mddev->private;
 	int n;
 
 	if ((n=atomic_read(&conf->counters[WriteTransient])) != 0)
@@ -328,7 +327,7 @@ static int run(mddev_t *mddev)
 
 static int stop(mddev_t *mddev)
 {
-	conf_t *conf = (conf_t *)mddev->private;
+	conf_t *conf = mddev->private;
 
 	kfree(conf);
 	mddev->private = NULL;
diff --git a/drivers/md/linear.c b/drivers/md/linear.c
index 09437e958235..7e0e057db9a7 100644
--- a/drivers/md/linear.c
+++ b/drivers/md/linear.c
@@ -159,7 +159,8 @@ static linear_conf_t *linear_conf(mddev_t *mddev, int raid_disks)
 		sector_t sectors;
 
 		if (j < 0 || j >= raid_disks || disk->rdev) {
-			printk("linear: disk numbering problem. Aborting!\n");
+			printk(KERN_ERR "md/linear:%s: disk numbering problem. Aborting!\n",
+			       mdname(mddev));
 			goto out;
 		}
 
@@ -187,7 +188,8 @@ static linear_conf_t *linear_conf(mddev_t *mddev, int raid_disks)
 
 	}
 	if (cnt != raid_disks) {
-		printk("linear: not enough drives present. Aborting!\n");
+		printk(KERN_ERR "md/linear:%s: not enough drives present. Aborting!\n",
+		       mdname(mddev));
 		goto out;
 	}
 
@@ -282,29 +284,21 @@ static int linear_stop (mddev_t *mddev)
 	rcu_barrier();
 	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
 	kfree(conf);
+	mddev->private = NULL;
 
 	return 0;
 }
 
-static int linear_make_request (struct request_queue *q, struct bio *bio)
+static int linear_make_request (mddev_t *mddev, struct bio *bio)
 {
-	const int rw = bio_data_dir(bio);
-	mddev_t *mddev = q->queuedata;
 	dev_info_t *tmp_dev;
 	sector_t start_sector;
-	int cpu;
 
 	if (unlikely(bio_rw_flagged(bio, BIO_RW_BARRIER))) {
 		md_barrier_request(mddev, bio);
 		return 0;
 	}
 
-	cpu = part_stat_lock();
-	part_stat_inc(cpu, &mddev->gendisk->part0, ios[rw]);
-	part_stat_add(cpu, &mddev->gendisk->part0, sectors[rw],
-		      bio_sectors(bio));
-	part_stat_unlock();
-
 	rcu_read_lock();
 	tmp_dev = which_dev(mddev, bio->bi_sector);
 	start_sector = tmp_dev->end_sector - tmp_dev->rdev->sectors;
@@ -314,12 +308,14 @@ static int linear_make_request (struct request_queue *q, struct bio *bio)
 		     || (bio->bi_sector < start_sector))) {
 		char b[BDEVNAME_SIZE];
 
-		printk("linear_make_request: Sector %llu out of bounds on "
-			"dev %s: %llu sectors, offset %llu\n",
-			(unsigned long long)bio->bi_sector,
-			bdevname(tmp_dev->rdev->bdev, b),
-			(unsigned long long)tmp_dev->rdev->sectors,
-			(unsigned long long)start_sector);
+		printk(KERN_ERR
+		       "md/linear:%s: make_request: Sector %llu out of bounds on "
+		       "dev %s: %llu sectors, offset %llu\n",
+		       mdname(mddev),
+		       (unsigned long long)bio->bi_sector,
+		       bdevname(tmp_dev->rdev->bdev, b),
+		       (unsigned long long)tmp_dev->rdev->sectors,
+		       (unsigned long long)start_sector);
 		rcu_read_unlock();
 		bio_io_error(bio);
 		return 0;
@@ -336,9 +332,9 @@ static int linear_make_request (struct request_queue *q, struct bio *bio)
 
 		bp = bio_split(bio, end_sector - bio->bi_sector);
 
-		if (linear_make_request(q, &bp->bio1))
+		if (linear_make_request(mddev, &bp->bio1))
 			generic_make_request(&bp->bio1);
-		if (linear_make_request(q, &bp->bio2))
+		if (linear_make_request(mddev, &bp->bio2))
 			generic_make_request(&bp->bio2);
 		bio_pair_release(bp);
 		return 0;
diff --git a/drivers/md/md.c b/drivers/md/md.c
index a9fd491796ac..46b3a044eadf 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -215,8 +215,11 @@ static DEFINE_SPINLOCK(all_mddevs_lock);
  */
 static int md_make_request(struct request_queue *q, struct bio *bio)
 {
+	const int rw = bio_data_dir(bio);
 	mddev_t *mddev = q->queuedata;
 	int rv;
+	int cpu;
+
 	if (mddev == NULL || mddev->pers == NULL) {
 		bio_io_error(bio);
 		return 0;
@@ -237,13 +240,27 @@ static int md_make_request(struct request_queue *q, struct bio *bio)
 	}
 	atomic_inc(&mddev->active_io);
 	rcu_read_unlock();
-	rv = mddev->pers->make_request(q, bio);
+
+	rv = mddev->pers->make_request(mddev, bio);
+
+	cpu = part_stat_lock();
+	part_stat_inc(cpu, &mddev->gendisk->part0, ios[rw]);
+	part_stat_add(cpu, &mddev->gendisk->part0, sectors[rw],
+		      bio_sectors(bio));
+	part_stat_unlock();
+
 	if (atomic_dec_and_test(&mddev->active_io) && mddev->suspended)
 		wake_up(&mddev->sb_wait);
 
 	return rv;
 }
 
+/* mddev_suspend makes sure no new requests are submitted
+ * to the device, and that any requests that have been submitted
+ * are completely handled.
+ * Once ->stop is called and completes, the module will be completely
+ * unused.
+ */
 static void mddev_suspend(mddev_t *mddev)
 {
 	BUG_ON(mddev->suspended);
@@ -251,13 +268,6 @@ static void mddev_suspend(mddev_t *mddev)
 	synchronize_rcu();
 	wait_event(mddev->sb_wait, atomic_read(&mddev->active_io) == 0);
 	mddev->pers->quiesce(mddev, 1);
-	md_unregister_thread(mddev->thread);
-	mddev->thread = NULL;
-	/* we now know that no code is executing in the personality module,
-	 * except possibly the tail end of a ->bi_end_io function, but that
-	 * is certain to complete before the module has a chance to get
-	 * unloaded
-	 */
 }
 
 static void mddev_resume(mddev_t *mddev)
@@ -344,7 +354,7 @@ static void md_submit_barrier(struct work_struct *ws)
 		bio_endio(bio, 0);
 	else {
 		bio->bi_rw &= ~(1<<BIO_RW_BARRIER);
-		if (mddev->pers->make_request(mddev->queue, bio))
+		if (mddev->pers->make_request(mddev, bio))
 			generic_make_request(bio);
 		mddev->barrier = POST_REQUEST_BARRIER;
 		submit_barriers(mddev);
@@ -406,6 +416,27 @@ static void mddev_put(mddev_t *mddev)
 	spin_unlock(&all_mddevs_lock);
 }
 
+static void mddev_init(mddev_t *mddev)
+{
+	mutex_init(&mddev->open_mutex);
+	mutex_init(&mddev->reconfig_mutex);
+	mutex_init(&mddev->bitmap_info.mutex);
+	INIT_LIST_HEAD(&mddev->disks);
+	INIT_LIST_HEAD(&mddev->all_mddevs);
+	init_timer(&mddev->safemode_timer);
+	atomic_set(&mddev->active, 1);
+	atomic_set(&mddev->openers, 0);
+	atomic_set(&mddev->active_io, 0);
+	spin_lock_init(&mddev->write_lock);
+	atomic_set(&mddev->flush_pending, 0);
+	init_waitqueue_head(&mddev->sb_wait);
+	init_waitqueue_head(&mddev->recovery_wait);
+	mddev->reshape_position = MaxSector;
+	mddev->resync_min = 0;
+	mddev->resync_max = MaxSector;
+	mddev->level = LEVEL_NONE;
+}
+
 static mddev_t * mddev_find(dev_t unit)
 {
 	mddev_t *mddev, *new = NULL;
@@ -472,23 +503,7 @@ static mddev_t * mddev_find(dev_t unit)
 	else
 		new->md_minor = MINOR(unit) >> MdpMinorShift;
 
-	mutex_init(&new->open_mutex);
-	mutex_init(&new->reconfig_mutex);
-	mutex_init(&new->bitmap_info.mutex);
-	INIT_LIST_HEAD(&new->disks);
-	INIT_LIST_HEAD(&new->all_mddevs);
-	init_timer(&new->safemode_timer);
-	atomic_set(&new->active, 1);
-	atomic_set(&new->openers, 0);
-	atomic_set(&new->active_io, 0);
-	spin_lock_init(&new->write_lock);
-	atomic_set(&new->flush_pending, 0);
-	init_waitqueue_head(&new->sb_wait);
-	init_waitqueue_head(&new->recovery_wait);
-	new->reshape_position = MaxSector;
-	new->resync_min = 0;
-	new->resync_max = MaxSector;
-	new->level = LEVEL_NONE;
+	mddev_init(new);
 
 	goto retry;
 }
@@ -508,9 +523,36 @@ static inline int mddev_trylock(mddev_t * mddev)
 	return mutex_trylock(&mddev->reconfig_mutex);
 }
 
-static inline void mddev_unlock(mddev_t * mddev)
+static struct attribute_group md_redundancy_group;
+
+static void mddev_unlock(mddev_t * mddev)
 {
-	mutex_unlock(&mddev->reconfig_mutex);
+	if (mddev->to_remove) {
+		/* These cannot be removed under reconfig_mutex as
+		 * an access to the files will try to take reconfig_mutex
+		 * while holding the file unremovable, which leads to
+		 * a deadlock.
+		 * So hold open_mutex instead - we are allowed to take
+		 * it while holding reconfig_mutex, and md_run can
+		 * use it to wait for the remove to complete.
+		 */
+		struct attribute_group *to_remove = mddev->to_remove;
+		mddev->to_remove = NULL;
+		mutex_lock(&mddev->open_mutex);
+		mutex_unlock(&mddev->reconfig_mutex);
+
+		if (to_remove != &md_redundancy_group)
+			sysfs_remove_group(&mddev->kobj, to_remove);
+		if (mddev->pers == NULL ||
+		    mddev->pers->sync_request == NULL) {
+			sysfs_remove_group(&mddev->kobj, &md_redundancy_group);
+			if (mddev->sysfs_action)
+				sysfs_put(mddev->sysfs_action);
+			mddev->sysfs_action = NULL;
+		}
+		mutex_unlock(&mddev->open_mutex);
+	} else
+		mutex_unlock(&mddev->reconfig_mutex);
 
 	md_wakeup_thread(mddev->thread);
 }
@@ -1029,10 +1071,13 @@ static int super_90_validate(mddev_t *mddev, mdk_rdev_t *rdev)
 				mddev->bitmap_info.default_offset;
 
 	} else if (mddev->pers == NULL) {
-		/* Insist on good event counter while assembling */
+		/* Insist on good event counter while assembling, except
+		 * for spares (which don't need an event count) */
 		++ev1;
-		if (ev1 < mddev->events) 
-			return -EINVAL;
+		if (sb->disks[rdev->desc_nr].state & (
+			    (1<<MD_DISK_SYNC) | (1 << MD_DISK_ACTIVE)))
+			if (ev1 < mddev->events) 
+				return -EINVAL;
 	} else if (mddev->bitmap) {
 		/* if adding to array with a bitmap, then we can accept an
 		 * older device ... but not too old.
@@ -1428,10 +1473,14 @@ static int super_1_validate(mddev_t *mddev, mdk_rdev_t *rdev)
 		}
 
 	} else if (mddev->pers == NULL) {
-		/* Insist of good event counter while assembling */
+		/* Insist of good event counter while assembling, except for
+		 * spares (which don't need an event count) */
 		++ev1;
-		if (ev1 < mddev->events)
-			return -EINVAL;
+		if (rdev->desc_nr >= 0 &&
+		    rdev->desc_nr < le32_to_cpu(sb->max_dev) &&
+		    le16_to_cpu(sb->dev_roles[rdev->desc_nr]) < 0xfffe)
+			if (ev1 < mddev->events)
+				return -EINVAL;
 	} else if (mddev->bitmap) {
 		/* If adding to array with a bitmap, then we can accept an
 		 * older device, but not too old.
@@ -2047,7 +2096,6 @@ static void sync_sbs(mddev_t * mddev, int nospares)
 		if (rdev->sb_events == mddev->events ||
 		    (nospares &&
 		     rdev->raid_disk < 0 &&
-		     (rdev->sb_events&1)==0 &&
 		     rdev->sb_events+1 == mddev->events)) {
 			/* Don't update this superblock */
 			rdev->sb_loaded = 2;
@@ -2100,28 +2148,14 @@ repeat:
 	 * and 'events' is odd, we can roll back to the previous clean state */
 	if (nospares
 	    && (mddev->in_sync && mddev->recovery_cp == MaxSector)
-	    && (mddev->events & 1)
-	    && mddev->events != 1)
+	    && mddev->can_decrease_events
+	    && mddev->events != 1) {
 		mddev->events--;
-	else {
+		mddev->can_decrease_events = 0;
+	} else {
 		/* otherwise we have to go forward and ... */
 		mddev->events ++;
-		if (!mddev->in_sync || mddev->recovery_cp != MaxSector) { /* not clean */
-			/* .. if the array isn't clean, an 'even' event must also go
-			 * to spares. */
-			if ((mddev->events&1)==0) {
-				nospares = 0;
-				sync_req = 2; /* force a second update to get the
-					       * even/odd in sync */
-			}
-		} else {
-			/* otherwise an 'odd' event must go to spares */
-			if ((mddev->events&1)) {
-				nospares = 0;
-				sync_req = 2; /* force a second update to get the
-					       * even/odd in sync */
-			}
-		}
+		mddev->can_decrease_events = nospares;
 	}
 
 	if (!mddev->events) {
@@ -2365,6 +2399,7 @@ slot_store(mdk_rdev_t *rdev, const char *buf, size_t len)
 			return err;
 		sprintf(nm, "rd%d", rdev->raid_disk);
 		sysfs_remove_link(&rdev->mddev->kobj, nm);
+		rdev->raid_disk = -1;
 		set_bit(MD_RECOVERY_NEEDED, &rdev->mddev->recovery);
 		md_wakeup_thread(rdev->mddev->thread);
 	} else if (rdev->mddev->pers) {
@@ -2780,8 +2815,9 @@ static void analyze_sbs(mddev_t * mddev)
 
 	i = 0;
 	rdev_for_each(rdev, tmp, mddev) {
-		if (rdev->desc_nr >= mddev->max_disks ||
-		    i > mddev->max_disks) {
+		if (mddev->max_disks &&
+		    (rdev->desc_nr >= mddev->max_disks ||
+		     i > mddev->max_disks)) {
 			printk(KERN_WARNING
 			       "md: %s: %s: only %d devices permitted\n",
 			       mdname(mddev), bdevname(rdev->bdev, b),
@@ -2897,9 +2933,10 @@ level_show(mddev_t *mddev, char *page)
 static ssize_t
 level_store(mddev_t *mddev, const char *buf, size_t len)
 {
-	char level[16];
+	char clevel[16];
 	ssize_t rv = len;
 	struct mdk_personality *pers;
+	long level;
 	void *priv;
 	mdk_rdev_t *rdev;
 
@@ -2932,19 +2969,22 @@ level_store(mddev_t *mddev, const char *buf, size_t len)
 	}
 
 	/* Now find the new personality */
-	if (len == 0 || len >= sizeof(level))
+	if (len == 0 || len >= sizeof(clevel))
 		return -EINVAL;
-	strncpy(level, buf, len);
-	if (level[len-1] == '\n')
+	strncpy(clevel, buf, len);
+	if (clevel[len-1] == '\n')
 		len--;
-	level[len] = 0;
+	clevel[len] = 0;
+	if (strict_strtol(clevel, 10, &level))
+		level = LEVEL_NONE;
 
-	request_module("md-%s", level);
+	if (request_module("md-%s", clevel) != 0)
+		request_module("md-level-%s", clevel);
 	spin_lock(&pers_lock);
-	pers = find_pers(LEVEL_NONE, level);
+	pers = find_pers(level, clevel);
 	if (!pers || !try_module_get(pers->owner)) {
 		spin_unlock(&pers_lock);
-		printk(KERN_WARNING "md: personality %s not loaded\n", level);
+		printk(KERN_WARNING "md: personality %s not loaded\n", clevel);
 		return -EINVAL;
 	}
 	spin_unlock(&pers_lock);
@@ -2957,7 +2997,7 @@ level_store(mddev_t *mddev, const char *buf, size_t len)
 	if (!pers->takeover) {
 		module_put(pers->owner);
 		printk(KERN_WARNING "md: %s: %s does not support personality takeover\n",
-		       mdname(mddev), level);
+		       mdname(mddev), clevel);
 		return -EINVAL;
 	}
 
@@ -2973,13 +3013,44 @@ level_store(mddev_t *mddev, const char *buf, size_t len)
 		mddev->delta_disks = 0;
 		module_put(pers->owner);
 		printk(KERN_WARNING "md: %s: %s would not accept array\n",
-		       mdname(mddev), level);
+		       mdname(mddev), clevel);
 		return PTR_ERR(priv);
 	}
 
 	/* Looks like we have a winner */
 	mddev_suspend(mddev);
 	mddev->pers->stop(mddev);
+	
+	if (mddev->pers->sync_request == NULL &&
+	    pers->sync_request != NULL) {
+		/* need to add the md_redundancy_group */
+		if (sysfs_create_group(&mddev->kobj, &md_redundancy_group))
+			printk(KERN_WARNING
+			       "md: cannot register extra attributes for %s\n",
+			       mdname(mddev));
+		mddev->sysfs_action = sysfs_get_dirent(mddev->kobj.sd, NULL, "sync_action");
+	}		
+	if (mddev->pers->sync_request != NULL &&
+	    pers->sync_request == NULL) {
+		/* need to remove the md_redundancy_group */
+		if (mddev->to_remove == NULL)
+			mddev->to_remove = &md_redundancy_group;
+	}
+
+	if (mddev->pers->sync_request == NULL &&
+	    mddev->external) {
+		/* We are converting from a no-redundancy array
+		 * to a redundancy array and metadata is managed
+		 * externally so we need to be sure that writes
+		 * won't block due to a need to transition
+		 *      clean->dirty
+		 * until external management is started.
+		 */
+		mddev->in_sync = 0;
+		mddev->safemode_delay = 0;
+		mddev->safemode = 0;
+	}
+
 	module_put(mddev->pers->owner);
 	/* Invalidate devices that are now superfluous */
 	list_for_each_entry(rdev, &mddev->disks, same_set)
@@ -2994,11 +3065,20 @@ level_store(mddev_t *mddev, const char *buf, size_t len)
 	mddev->layout = mddev->new_layout;
 	mddev->chunk_sectors = mddev->new_chunk_sectors;
 	mddev->delta_disks = 0;
+	if (mddev->pers->sync_request == NULL) {
+		/* this is now an array without redundancy, so
+		 * it must always be in_sync
+		 */
+		mddev->in_sync = 1;
+		del_timer_sync(&mddev->safemode_timer);
+	}
 	pers->run(mddev);
 	mddev_resume(mddev);
 	set_bit(MD_CHANGE_DEVS, &mddev->flags);
 	set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 	md_wakeup_thread(mddev->thread);
+	sysfs_notify(&mddev->kobj, NULL, "level");
+	md_new_event(mddev);
 	return rv;
 }
 
@@ -3237,6 +3317,7 @@ array_state_show(mddev_t *mddev, char *page)
 }
 
 static int do_md_stop(mddev_t * mddev, int ro, int is_open);
+static int md_set_readonly(mddev_t * mddev, int is_open);
 static int do_md_run(mddev_t * mddev);
 static int restart_array(mddev_t *mddev);
 
@@ -3267,7 +3348,7 @@ array_state_store(mddev_t *mddev, const char *buf, size_t len)
 		break; /* not supported yet */
 	case readonly:
 		if (mddev->pers)
-			err = do_md_stop(mddev, 1, 0);
+			err = md_set_readonly(mddev, 0);
 		else {
 			mddev->ro = 1;
 			set_disk_ro(mddev->gendisk, 1);
@@ -3277,7 +3358,7 @@ array_state_store(mddev_t *mddev, const char *buf, size_t len)
 	case read_auto:
 		if (mddev->pers) {
 			if (mddev->ro == 0)
-				err = do_md_stop(mddev, 1, 0);
+				err = md_set_readonly(mddev, 0);
 			else if (mddev->ro == 1)
 				err = restart_array(mddev);
 			if (err == 0) {
@@ -4082,15 +4163,6 @@ static void mddev_delayed_delete(struct work_struct *ws)
 {
 	mddev_t *mddev = container_of(ws, mddev_t, del_work);
 
-	if (mddev->private) {
-		sysfs_remove_group(&mddev->kobj, &md_redundancy_group);
-		if (mddev->private != (void*)1)
-			sysfs_remove_group(&mddev->kobj, mddev->private);
-		if (mddev->sysfs_action)
-			sysfs_put(mddev->sysfs_action);
-		mddev->sysfs_action = NULL;
-		mddev->private = NULL;
-	}
 	sysfs_remove_group(&mddev->kobj, &md_bitmap_group);
 	kobject_del(&mddev->kobj);
 	kobject_put(&mddev->kobj);
@@ -4234,11 +4306,10 @@ static void md_safemode_timeout(unsigned long data)
 
 static int start_dirty_degraded;
 
-static int do_md_run(mddev_t * mddev)
+static int md_run(mddev_t *mddev)
 {
 	int err;
 	mdk_rdev_t *rdev;
-	struct gendisk *disk;
 	struct mdk_personality *pers;
 
 	if (list_empty(&mddev->disks))
@@ -4248,6 +4319,13 @@ static int do_md_run(mddev_t * mddev)
 	if (mddev->pers)
 		return -EBUSY;
 
+	/* These two calls synchronise us with the
+	 * sysfs_remove_group calls in mddev_unlock,
+	 * so they must have completed.
+	 */
+	mutex_lock(&mddev->open_mutex);
+	mutex_unlock(&mddev->open_mutex);
+
 	/*
 	 * Analyze all RAID superblock(s)
 	 */
@@ -4296,8 +4374,6 @@ static int do_md_run(mddev_t * mddev)
 		sysfs_notify_dirent(rdev->sysfs_state);
 	}
 
-	disk = mddev->gendisk;
-
 	spin_lock(&pers_lock);
 	pers = find_pers(mddev->level, mddev->clevel);
 	if (!pers || !try_module_get(pers->owner)) {
@@ -4425,22 +4501,32 @@ static int do_md_run(mddev_t * mddev)
 	if (mddev->flags)
 		md_update_sb(mddev, 0);
 
-	set_capacity(disk, mddev->array_sectors);
-
 	md_wakeup_thread(mddev->thread);
 	md_wakeup_thread(mddev->sync_thread); /* possibly kick off a reshape */
 
-	revalidate_disk(mddev->gendisk);
-	mddev->changed = 1;
 	md_new_event(mddev);
 	sysfs_notify_dirent(mddev->sysfs_state);
 	if (mddev->sysfs_action)
 		sysfs_notify_dirent(mddev->sysfs_action);
 	sysfs_notify(&mddev->kobj, NULL, "degraded");
-	kobject_uevent(&disk_to_dev(mddev->gendisk)->kobj, KOBJ_CHANGE);
 	return 0;
 }
 
+static int do_md_run(mddev_t *mddev)
+{
+	int err;
+
+	err = md_run(mddev);
+	if (err)
+		goto out;
+
+	set_capacity(mddev->gendisk, mddev->array_sectors);
+	revalidate_disk(mddev->gendisk);
+	kobject_uevent(&disk_to_dev(mddev->gendisk)->kobj, KOBJ_CHANGE);
+out:
+	return err;
+}
+
 static int restart_array(mddev_t *mddev)
 {
 	struct gendisk *disk = mddev->gendisk;
@@ -4491,9 +4577,110 @@ void restore_bitmap_write_access(struct file *file)
 	spin_unlock(&inode->i_lock);
 }
 
+static void md_clean(mddev_t *mddev)
+{
+	mddev->array_sectors = 0;
+	mddev->external_size = 0;
+	mddev->dev_sectors = 0;
+	mddev->raid_disks = 0;
+	mddev->recovery_cp = 0;
+	mddev->resync_min = 0;
+	mddev->resync_max = MaxSector;
+	mddev->reshape_position = MaxSector;
+	mddev->external = 0;
+	mddev->persistent = 0;
+	mddev->level = LEVEL_NONE;
+	mddev->clevel[0] = 0;
+	mddev->flags = 0;
+	mddev->ro = 0;
+	mddev->metadata_type[0] = 0;
+	mddev->chunk_sectors = 0;
+	mddev->ctime = mddev->utime = 0;
+	mddev->layout = 0;
+	mddev->max_disks = 0;
+	mddev->events = 0;
+	mddev->can_decrease_events = 0;
+	mddev->delta_disks = 0;
+	mddev->new_level = LEVEL_NONE;
+	mddev->new_layout = 0;
+	mddev->new_chunk_sectors = 0;
+	mddev->curr_resync = 0;
+	mddev->resync_mismatches = 0;
+	mddev->suspend_lo = mddev->suspend_hi = 0;
+	mddev->sync_speed_min = mddev->sync_speed_max = 0;
+	mddev->recovery = 0;
+	mddev->in_sync = 0;
+	mddev->degraded = 0;
+	mddev->barriers_work = 0;
+	mddev->safemode = 0;
+	mddev->bitmap_info.offset = 0;
+	mddev->bitmap_info.default_offset = 0;
+	mddev->bitmap_info.chunksize = 0;
+	mddev->bitmap_info.daemon_sleep = 0;
+	mddev->bitmap_info.max_write_behind = 0;
+}
+
+static void md_stop_writes(mddev_t *mddev)
+{
+	if (mddev->sync_thread) {
+		set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+		set_bit(MD_RECOVERY_INTR, &mddev->recovery);
+		md_unregister_thread(mddev->sync_thread);
+		mddev->sync_thread = NULL;
+	}
+
+	del_timer_sync(&mddev->safemode_timer);
+
+	bitmap_flush(mddev);
+	md_super_wait(mddev);
+
+	if (!mddev->in_sync || mddev->flags) {
+		/* mark array as shutdown cleanly */
+		mddev->in_sync = 1;
+		md_update_sb(mddev, 1);
+	}
+}
+
+static void md_stop(mddev_t *mddev)
+{
+	md_stop_writes(mddev);
+
+	mddev->pers->stop(mddev);
+	if (mddev->pers->sync_request && mddev->to_remove == NULL)
+		mddev->to_remove = &md_redundancy_group;
+	module_put(mddev->pers->owner);
+	mddev->pers = NULL;
+	clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+}
+
+static int md_set_readonly(mddev_t *mddev, int is_open)
+{
+	int err = 0;
+	mutex_lock(&mddev->open_mutex);
+	if (atomic_read(&mddev->openers) > is_open) {
+		printk("md: %s still in use.\n",mdname(mddev));
+		err = -EBUSY;
+		goto out;
+	}
+	if (mddev->pers) {
+		md_stop_writes(mddev);
+
+		err  = -ENXIO;
+		if (mddev->ro==1)
+			goto out;
+		mddev->ro = 1;
+		set_disk_ro(mddev->gendisk, 1);
+		clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+		sysfs_notify_dirent(mddev->sysfs_state);
+		err = 0;	
+	}
+out:
+	mutex_unlock(&mddev->open_mutex);
+	return err;
+}
+
 /* mode:
  *   0 - completely stop and dis-assemble array
- *   1 - switch to readonly
  *   2 - stop but do not disassemble array
  */
 static int do_md_stop(mddev_t * mddev, int mode, int is_open)
@@ -4508,64 +4695,32 @@ static int do_md_stop(mddev_t * mddev, int mode, int is_open)
 		err = -EBUSY;
 	} else if (mddev->pers) {
 
-		if (mddev->sync_thread) {
-			set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
-			set_bit(MD_RECOVERY_INTR, &mddev->recovery);
-			md_unregister_thread(mddev->sync_thread);
-			mddev->sync_thread = NULL;
-		}
-
-		del_timer_sync(&mddev->safemode_timer);
+		if (mddev->ro)
+			set_disk_ro(disk, 0);
 
-		switch(mode) {
-		case 1: /* readonly */
-			err  = -ENXIO;
-			if (mddev->ro==1)
-				goto out;
-			mddev->ro = 1;
-			break;
-		case 0: /* disassemble */
-		case 2: /* stop */
-			bitmap_flush(mddev);
-			md_super_wait(mddev);
-			if (mddev->ro)
-				set_disk_ro(disk, 0);
+		md_stop(mddev);
+		mddev->queue->merge_bvec_fn = NULL;
+		mddev->queue->unplug_fn = NULL;
+		mddev->queue->backing_dev_info.congested_fn = NULL;
 
-			mddev->pers->stop(mddev);
-			mddev->queue->merge_bvec_fn = NULL;
-			mddev->queue->unplug_fn = NULL;
-			mddev->queue->backing_dev_info.congested_fn = NULL;
-			module_put(mddev->pers->owner);
-			if (mddev->pers->sync_request && mddev->private == NULL)
-				mddev->private = (void*)1;
-			mddev->pers = NULL;
-			/* tell userspace to handle 'inactive' */
-			sysfs_notify_dirent(mddev->sysfs_state);
+		/* tell userspace to handle 'inactive' */
+		sysfs_notify_dirent(mddev->sysfs_state);
 
-			list_for_each_entry(rdev, &mddev->disks, same_set)
-				if (rdev->raid_disk >= 0) {
-					char nm[20];
-					sprintf(nm, "rd%d", rdev->raid_disk);
-					sysfs_remove_link(&mddev->kobj, nm);
-				}
+		list_for_each_entry(rdev, &mddev->disks, same_set)
+			if (rdev->raid_disk >= 0) {
+				char nm[20];
+				sprintf(nm, "rd%d", rdev->raid_disk);
+				sysfs_remove_link(&mddev->kobj, nm);
+			}
 
-			set_capacity(disk, 0);
-			mddev->changed = 1;
+		set_capacity(disk, 0);
+		revalidate_disk(disk);
 
-			if (mddev->ro)
-				mddev->ro = 0;
-		}
-		if (!mddev->in_sync || mddev->flags) {
-			/* mark array as shutdown cleanly */
-			mddev->in_sync = 1;
-			md_update_sb(mddev, 1);
-		}
-		if (mode == 1)
-			set_disk_ro(disk, 1);
-		clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+		if (mddev->ro)
+			mddev->ro = 0;
+		
 		err = 0;
 	}
-out:
 	mutex_unlock(&mddev->open_mutex);
 	if (err)
 		return err;
@@ -4586,52 +4741,12 @@ out:
 
 		export_array(mddev);
 
-		mddev->array_sectors = 0;
-		mddev->external_size = 0;
-		mddev->dev_sectors = 0;
-		mddev->raid_disks = 0;
-		mddev->recovery_cp = 0;
-		mddev->resync_min = 0;
-		mddev->resync_max = MaxSector;
-		mddev->reshape_position = MaxSector;
-		mddev->external = 0;
-		mddev->persistent = 0;
-		mddev->level = LEVEL_NONE;
-		mddev->clevel[0] = 0;
-		mddev->flags = 0;
-		mddev->ro = 0;
-		mddev->metadata_type[0] = 0;
-		mddev->chunk_sectors = 0;
-		mddev->ctime = mddev->utime = 0;
-		mddev->layout = 0;
-		mddev->max_disks = 0;
-		mddev->events = 0;
-		mddev->delta_disks = 0;
-		mddev->new_level = LEVEL_NONE;
-		mddev->new_layout = 0;
-		mddev->new_chunk_sectors = 0;
-		mddev->curr_resync = 0;
-		mddev->resync_mismatches = 0;
-		mddev->suspend_lo = mddev->suspend_hi = 0;
-		mddev->sync_speed_min = mddev->sync_speed_max = 0;
-		mddev->recovery = 0;
-		mddev->in_sync = 0;
-		mddev->changed = 0;
-		mddev->degraded = 0;
-		mddev->barriers_work = 0;
-		mddev->safemode = 0;
-		mddev->bitmap_info.offset = 0;
-		mddev->bitmap_info.default_offset = 0;
-		mddev->bitmap_info.chunksize = 0;
-		mddev->bitmap_info.daemon_sleep = 0;
-		mddev->bitmap_info.max_write_behind = 0;
+		md_clean(mddev);
 		kobject_uevent(&disk_to_dev(mddev->gendisk)->kobj, KOBJ_CHANGE);
 		if (mddev->hold_active == UNTIL_STOP)
 			mddev->hold_active = 0;
 
-	} else if (mddev->pers)
-		printk(KERN_INFO "md: %s switched to read-only mode.\n",
-			mdname(mddev));
+	}
 	err = 0;
 	blk_integrity_unregister(disk);
 	md_new_event(mddev);
@@ -5349,7 +5464,7 @@ static int update_raid_disks(mddev_t *mddev, int raid_disks)
 	if (mddev->pers->check_reshape == NULL)
 		return -EINVAL;
 	if (raid_disks <= 0 ||
-	    raid_disks >= mddev->max_disks)
+	    (mddev->max_disks && raid_disks >= mddev->max_disks))
 		return -EINVAL;
 	if (mddev->sync_thread || mddev->reshape_position != MaxSector)
 		return -EBUSY;
@@ -5486,7 +5601,7 @@ static int md_getgeo(struct block_device *bdev, struct hd_geometry *geo)
 
 	geo->heads = 2;
 	geo->sectors = 4;
-	geo->cylinders = get_capacity(mddev->gendisk) / 8;
+	geo->cylinders = mddev->array_sectors / 8;
 	return 0;
 }
 
@@ -5496,6 +5611,7 @@ static int md_ioctl(struct block_device *bdev, fmode_t mode,
 	int err = 0;
 	void __user *argp = (void __user *)arg;
 	mddev_t *mddev = NULL;
+	int ro;
 
 	if (!capable(CAP_SYS_ADMIN))
 		return -EACCES;
@@ -5628,9 +5744,37 @@ static int md_ioctl(struct block_device *bdev, fmode_t mode,
 			goto done_unlock;
 
 		case STOP_ARRAY_RO:
-			err = do_md_stop(mddev, 1, 1);
+			err = md_set_readonly(mddev, 1);
 			goto done_unlock;
 
+		case BLKROSET:
+			if (get_user(ro, (int __user *)(arg))) {
+				err = -EFAULT;
+				goto done_unlock;
+			}
+			err = -EINVAL;
+
+			/* if the bdev is going readonly the value of mddev->ro
+			 * does not matter, no writes are coming
+			 */
+			if (ro)
+				goto done_unlock;
+
+			/* are we are already prepared for writes? */
+			if (mddev->ro != 1)
+				goto done_unlock;
+
+			/* transitioning to readauto need only happen for
+			 * arrays that call md_write_start
+			 */
+			if (mddev->pers) {
+				err = restart_array(mddev);
+				if (err == 0) {
+					mddev->ro = 2;
+					set_disk_ro(mddev->gendisk, 0);
+				}
+			}
+			goto done_unlock;
 	}
 
 	/*
@@ -5751,7 +5895,6 @@ static int md_open(struct block_device *bdev, fmode_t mode)
 	atomic_inc(&mddev->openers);
 	mutex_unlock(&mddev->open_mutex);
 
-	check_disk_change(bdev);
  out:
 	return err;
 }
@@ -5766,21 +5909,6 @@ static int md_release(struct gendisk *disk, fmode_t mode)
 
 	return 0;
 }
-
-static int md_media_changed(struct gendisk *disk)
-{
-	mddev_t *mddev = disk->private_data;
-
-	return mddev->changed;
-}
-
-static int md_revalidate(struct gendisk *disk)
-{
-	mddev_t *mddev = disk->private_data;
-
-	mddev->changed = 0;
-	return 0;
-}
 static const struct block_device_operations md_fops =
 {
 	.owner		= THIS_MODULE,
@@ -5791,8 +5919,6 @@ static const struct block_device_operations md_fops =
 	.compat_ioctl	= md_compat_ioctl,
 #endif
 	.getgeo		= md_getgeo,
-	.media_changed	= md_media_changed,
-	.revalidate_disk= md_revalidate,
 };
 
 static int md_thread(void * arg)
@@ -5906,7 +6032,7 @@ void md_error(mddev_t *mddev, mdk_rdev_t *rdev)
 	mddev->pers->error_handler(mddev,rdev);
 	if (mddev->degraded)
 		set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
-	set_bit(StateChanged, &rdev->flags);
+	sysfs_notify_dirent(rdev->sysfs_state);
 	set_bit(MD_RECOVERY_INTR, &mddev->recovery);
 	set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 	md_wakeup_thread(mddev->thread);
@@ -6898,11 +7024,6 @@ void md_check_recovery(mddev_t *mddev)
 		if (mddev->flags)
 			md_update_sb(mddev, 0);
 
-		list_for_each_entry(rdev, &mddev->disks, same_set)
-			if (test_and_clear_bit(StateChanged, &rdev->flags))
-				sysfs_notify_dirent(rdev->sysfs_state);
-
-
 		if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) &&
 		    !test_bit(MD_RECOVERY_DONE, &mddev->recovery)) {
 			/* resync/recovery still happening */
@@ -7039,7 +7160,7 @@ static int md_notify_reboot(struct notifier_block *this,
 				 * appears to still be in use.  Hence
 				 * the '100'.
 				 */
-				do_md_stop(mddev, 1, 100);
+				md_set_readonly(mddev, 100);
 				mddev_unlock(mddev);
 			}
 		/*
diff --git a/drivers/md/md.h b/drivers/md/md.h
index 8e4c75c00d46..7ab5ea155452 100644
--- a/drivers/md/md.h
+++ b/drivers/md/md.h
@@ -74,9 +74,6 @@ struct mdk_rdev_s
 #define Blocked		8		/* An error occured on an externally
 					 * managed array, don't allow writes
 					 * until it is cleared */
-#define StateChanged	9		/* Faulty or Blocked has changed during
-					 * interrupt, so it needs to be
-					 * notified by the thread */
 	wait_queue_head_t blocked_wait;
 
 	int desc_nr;			/* descriptor index in the superblock */
@@ -153,6 +150,12 @@ struct mddev_s
 	int				external_size; /* size managed
 							* externally */
 	__u64				events;
+	/* If the last 'event' was simply a clean->dirty transition, and
+	 * we didn't write it to the spares, then it is safe and simple
+	 * to just decrement the event count on a dirty->clean transition.
+	 * So we record that possibility here.
+	 */
+	int				can_decrease_events;
 
 	char				uuid[16];
 
@@ -240,7 +243,6 @@ struct mddev_s
 	atomic_t			active;		/* general refcount */
 	atomic_t			openers;	/* number of active opens */
 
-	int				changed;	/* true if we might need to reread partition info */
 	int				degraded;	/* whether md should consider
 							 * adding a spare
 							 */
@@ -279,9 +281,6 @@ struct mddev_s
 	atomic_t			writes_pending; 
 	struct request_queue		*queue;	/* for plugging ... */
 
-	atomic_t                        write_behind; /* outstanding async IO */
-	unsigned int                    max_write_behind; /* 0 = sync */
-
 	struct bitmap                   *bitmap; /* the bitmap for the device */
 	struct {
 		struct file		*file; /* the bitmap file */
@@ -305,6 +304,7 @@ struct mddev_s
 	atomic_t 			max_corr_read_errors; /* max read retries */
 	struct list_head		all_mddevs;
 
+	struct attribute_group		*to_remove;
 	/* Generic barrier handling.
 	 * If there is a pending barrier request, all other
 	 * writes are blocked while the devices are flushed.
@@ -336,7 +336,7 @@ struct mdk_personality
 	int level;
 	struct list_head list;
 	struct module *owner;
-	int (*make_request)(struct request_queue *q, struct bio *bio);
+	int (*make_request)(mddev_t *mddev, struct bio *bio);
 	int (*run)(mddev_t *mddev);
 	int (*stop)(mddev_t *mddev);
 	void (*status)(struct seq_file *seq, mddev_t *mddev);
diff --git a/drivers/md/multipath.c b/drivers/md/multipath.c
index 789bf535d29c..410fb60699ac 100644
--- a/drivers/md/multipath.c
+++ b/drivers/md/multipath.c
@@ -85,7 +85,7 @@ static void multipath_end_bh_io (struct multipath_bh *mp_bh, int err)
 static void multipath_end_request(struct bio *bio, int error)
 {
 	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
-	struct multipath_bh * mp_bh = (struct multipath_bh *)(bio->bi_private);
+	struct multipath_bh *mp_bh = bio->bi_private;
 	multipath_conf_t *conf = mp_bh->mddev->private;
 	mdk_rdev_t *rdev = conf->multipaths[mp_bh->path].rdev;
 
@@ -136,14 +136,11 @@ static void multipath_unplug(struct request_queue *q)
 }
 
 
-static int multipath_make_request (struct request_queue *q, struct bio * bio)
+static int multipath_make_request(mddev_t *mddev, struct bio * bio)
 {
-	mddev_t *mddev = q->queuedata;
 	multipath_conf_t *conf = mddev->private;
 	struct multipath_bh * mp_bh;
 	struct multipath_info *multipath;
-	const int rw = bio_data_dir(bio);
-	int cpu;
 
 	if (unlikely(bio_rw_flagged(bio, BIO_RW_BARRIER))) {
 		md_barrier_request(mddev, bio);
@@ -155,12 +152,6 @@ static int multipath_make_request (struct request_queue *q, struct bio * bio)
 	mp_bh->master_bio = bio;
 	mp_bh->mddev = mddev;
 
-	cpu = part_stat_lock();
-	part_stat_inc(cpu, &mddev->gendisk->part0, ios[rw]);
-	part_stat_add(cpu, &mddev->gendisk->part0, sectors[rw],
-		      bio_sectors(bio));
-	part_stat_unlock();
-
 	mp_bh->path = multipath_map(conf);
 	if (mp_bh->path < 0) {
 		bio_endio(bio, -EIO);
diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c
index c3bec024612e..e70f004c99e8 100644
--- a/drivers/md/raid0.c
+++ b/drivers/md/raid0.c
@@ -23,15 +23,17 @@
 #include <linux/slab.h>
 #include "md.h"
 #include "raid0.h"
+#include "raid5.h"
 
 static void raid0_unplug(struct request_queue *q)
 {
 	mddev_t *mddev = q->queuedata;
 	raid0_conf_t *conf = mddev->private;
 	mdk_rdev_t **devlist = conf->devlist;
+	int raid_disks = conf->strip_zone[0].nb_dev;
 	int i;
 
-	for (i=0; i<mddev->raid_disks; i++) {
+	for (i=0; i < raid_disks; i++) {
 		struct request_queue *r_queue = bdev_get_queue(devlist[i]->bdev);
 
 		blk_unplug(r_queue);
@@ -43,12 +45,13 @@ static int raid0_congested(void *data, int bits)
 	mddev_t *mddev = data;
 	raid0_conf_t *conf = mddev->private;
 	mdk_rdev_t **devlist = conf->devlist;
+	int raid_disks = conf->strip_zone[0].nb_dev;
 	int i, ret = 0;
 
 	if (mddev_congested(mddev, bits))
 		return 1;
 
-	for (i = 0; i < mddev->raid_disks && !ret ; i++) {
+	for (i = 0; i < raid_disks && !ret ; i++) {
 		struct request_queue *q = bdev_get_queue(devlist[i]->bdev);
 
 		ret |= bdi_congested(&q->backing_dev_info, bits);
@@ -66,16 +69,17 @@ static void dump_zones(mddev_t *mddev)
 	sector_t zone_start = 0;
 	char b[BDEVNAME_SIZE];
 	raid0_conf_t *conf = mddev->private;
+	int raid_disks = conf->strip_zone[0].nb_dev;
 	printk(KERN_INFO "******* %s configuration *********\n",
 		mdname(mddev));
 	h = 0;
 	for (j = 0; j < conf->nr_strip_zones; j++) {
 		printk(KERN_INFO "zone%d=[", j);
 		for (k = 0; k < conf->strip_zone[j].nb_dev; k++)
-			printk("%s/",
-			bdevname(conf->devlist[j*mddev->raid_disks
+			printk(KERN_CONT "%s/",
+			bdevname(conf->devlist[j*raid_disks
 						+ k]->bdev, b));
-		printk("]\n");
+		printk(KERN_CONT "]\n");
 
 		zone_size  = conf->strip_zone[j].zone_end - zone_start;
 		printk(KERN_INFO "        zone offset=%llukb "
@@ -88,7 +92,7 @@ static void dump_zones(mddev_t *mddev)
 	printk(KERN_INFO "**********************************\n\n");
 }
 
-static int create_strip_zones(mddev_t *mddev)
+static int create_strip_zones(mddev_t *mddev, raid0_conf_t **private_conf)
 {
 	int i, c, err;
 	sector_t curr_zone_end, sectors;
@@ -101,8 +105,9 @@ static int create_strip_zones(mddev_t *mddev)
 	if (!conf)
 		return -ENOMEM;
 	list_for_each_entry(rdev1, &mddev->disks, same_set) {
-		printk(KERN_INFO "raid0: looking at %s\n",
-			bdevname(rdev1->bdev,b));
+		printk(KERN_INFO "md/raid0:%s: looking at %s\n",
+		       mdname(mddev),
+		       bdevname(rdev1->bdev, b));
 		c = 0;
 
 		/* round size to chunk_size */
@@ -111,14 +116,16 @@ static int create_strip_zones(mddev_t *mddev)
 		rdev1->sectors = sectors * mddev->chunk_sectors;
 
 		list_for_each_entry(rdev2, &mddev->disks, same_set) {
-			printk(KERN_INFO "raid0:   comparing %s(%llu)",
+			printk(KERN_INFO "md/raid0:%s:   comparing %s(%llu)",
+			       mdname(mddev),
 			       bdevname(rdev1->bdev,b),
 			       (unsigned long long)rdev1->sectors);
-			printk(KERN_INFO " with %s(%llu)\n",
+			printk(KERN_CONT " with %s(%llu)\n",
 			       bdevname(rdev2->bdev,b),
 			       (unsigned long long)rdev2->sectors);
 			if (rdev2 == rdev1) {
-				printk(KERN_INFO "raid0:   END\n");
+				printk(KERN_INFO "md/raid0:%s:   END\n",
+				       mdname(mddev));
 				break;
 			}
 			if (rdev2->sectors == rdev1->sectors) {
@@ -126,20 +133,24 @@ static int create_strip_zones(mddev_t *mddev)
 				 * Not unique, don't count it as a new
 				 * group
 				 */
-				printk(KERN_INFO "raid0:   EQUAL\n");
+				printk(KERN_INFO "md/raid0:%s:   EQUAL\n",
+				       mdname(mddev));
 				c = 1;
 				break;
 			}
-			printk(KERN_INFO "raid0:   NOT EQUAL\n");
+			printk(KERN_INFO "md/raid0:%s:   NOT EQUAL\n",
+			       mdname(mddev));
 		}
 		if (!c) {
-			printk(KERN_INFO "raid0:   ==> UNIQUE\n");
+			printk(KERN_INFO "md/raid0:%s:   ==> UNIQUE\n",
+			       mdname(mddev));
 			conf->nr_strip_zones++;
-			printk(KERN_INFO "raid0: %d zones\n",
-				conf->nr_strip_zones);
+			printk(KERN_INFO "md/raid0:%s: %d zones\n",
+			       mdname(mddev), conf->nr_strip_zones);
 		}
 	}
-	printk(KERN_INFO "raid0: FINAL %d zones\n", conf->nr_strip_zones);
+	printk(KERN_INFO "md/raid0:%s: FINAL %d zones\n",
+	       mdname(mddev), conf->nr_strip_zones);
 	err = -ENOMEM;
 	conf->strip_zone = kzalloc(sizeof(struct strip_zone)*
 				conf->nr_strip_zones, GFP_KERNEL);
@@ -162,14 +173,18 @@ static int create_strip_zones(mddev_t *mddev)
 	list_for_each_entry(rdev1, &mddev->disks, same_set) {
 		int j = rdev1->raid_disk;
 
+		if (mddev->level == 10)
+			/* taking over a raid10-n2 array */
+			j /= 2;
+
 		if (j < 0 || j >= mddev->raid_disks) {
-			printk(KERN_ERR "raid0: bad disk number %d - "
-				"aborting!\n", j);
+			printk(KERN_ERR "md/raid0:%s: bad disk number %d - "
+			       "aborting!\n", mdname(mddev), j);
 			goto abort;
 		}
 		if (dev[j]) {
-			printk(KERN_ERR "raid0: multiple devices for %d - "
-				"aborting!\n", j);
+			printk(KERN_ERR "md/raid0:%s: multiple devices for %d - "
+			       "aborting!\n", mdname(mddev), j);
 			goto abort;
 		}
 		dev[j] = rdev1;
@@ -191,8 +206,8 @@ static int create_strip_zones(mddev_t *mddev)
 		cnt++;
 	}
 	if (cnt != mddev->raid_disks) {
-		printk(KERN_ERR "raid0: too few disks (%d of %d) - "
-			"aborting!\n", cnt, mddev->raid_disks);
+		printk(KERN_ERR "md/raid0:%s: too few disks (%d of %d) - "
+		       "aborting!\n", mdname(mddev), cnt, mddev->raid_disks);
 		goto abort;
 	}
 	zone->nb_dev = cnt;
@@ -208,39 +223,44 @@ static int create_strip_zones(mddev_t *mddev)
 		zone = conf->strip_zone + i;
 		dev = conf->devlist + i * mddev->raid_disks;
 
-		printk(KERN_INFO "raid0: zone %d\n", i);
+		printk(KERN_INFO "md/raid0:%s: zone %d\n",
+		       mdname(mddev), i);
 		zone->dev_start = smallest->sectors;
 		smallest = NULL;
 		c = 0;
 
 		for (j=0; j<cnt; j++) {
 			rdev = conf->devlist[j];
-			printk(KERN_INFO "raid0: checking %s ...",
-				bdevname(rdev->bdev, b));
+			printk(KERN_INFO "md/raid0:%s: checking %s ...",
+			       mdname(mddev),
+			       bdevname(rdev->bdev, b));
 			if (rdev->sectors <= zone->dev_start) {
-				printk(KERN_INFO " nope.\n");
+				printk(KERN_CONT " nope.\n");
 				continue;
 			}
-			printk(KERN_INFO " contained as device %d\n", c);
+			printk(KERN_CONT " contained as device %d\n", c);
 			dev[c] = rdev;
 			c++;
 			if (!smallest || rdev->sectors < smallest->sectors) {
 				smallest = rdev;
-				printk(KERN_INFO "  (%llu) is smallest!.\n",
-					(unsigned long long)rdev->sectors);
+				printk(KERN_INFO "md/raid0:%s:  (%llu) is smallest!.\n",
+				       mdname(mddev),
+				       (unsigned long long)rdev->sectors);
 			}
 		}
 
 		zone->nb_dev = c;
 		sectors = (smallest->sectors - zone->dev_start) * c;
-		printk(KERN_INFO "raid0: zone->nb_dev: %d, sectors: %llu\n",
-			zone->nb_dev, (unsigned long long)sectors);
+		printk(KERN_INFO "md/raid0:%s: zone->nb_dev: %d, sectors: %llu\n",
+		       mdname(mddev),
+		       zone->nb_dev, (unsigned long long)sectors);
 
 		curr_zone_end += sectors;
 		zone->zone_end = curr_zone_end;
 
-		printk(KERN_INFO "raid0: current zone start: %llu\n",
-			(unsigned long long)smallest->sectors);
+		printk(KERN_INFO "md/raid0:%s: current zone start: %llu\n",
+		       mdname(mddev),
+		       (unsigned long long)smallest->sectors);
 	}
 	mddev->queue->unplug_fn = raid0_unplug;
 	mddev->queue->backing_dev_info.congested_fn = raid0_congested;
@@ -251,7 +271,7 @@ static int create_strip_zones(mddev_t *mddev)
 	 * chunk size is a multiple of that sector size
 	 */
 	if ((mddev->chunk_sectors << 9) % queue_logical_block_size(mddev->queue)) {
-		printk(KERN_ERR "%s chunk_size of %d not valid\n",
+		printk(KERN_ERR "md/raid0:%s: chunk_size of %d not valid\n",
 		       mdname(mddev),
 		       mddev->chunk_sectors << 9);
 		goto abort;
@@ -261,14 +281,15 @@ static int create_strip_zones(mddev_t *mddev)
 	blk_queue_io_opt(mddev->queue,
 			 (mddev->chunk_sectors << 9) * mddev->raid_disks);
 
-	printk(KERN_INFO "raid0: done.\n");
-	mddev->private = conf;
+	printk(KERN_INFO "md/raid0:%s: done.\n", mdname(mddev));
+	*private_conf = conf;
+
 	return 0;
 abort:
 	kfree(conf->strip_zone);
 	kfree(conf->devlist);
 	kfree(conf);
-	mddev->private = NULL;
+	*private_conf = NULL;
 	return err;
 }
 
@@ -319,10 +340,12 @@ static sector_t raid0_size(mddev_t *mddev, sector_t sectors, int raid_disks)
 
 static int raid0_run(mddev_t *mddev)
 {
+	raid0_conf_t *conf;
 	int ret;
 
 	if (mddev->chunk_sectors == 0) {
-		printk(KERN_ERR "md/raid0: chunk size must be set.\n");
+		printk(KERN_ERR "md/raid0:%s: chunk size must be set.\n",
+		       mdname(mddev));
 		return -EINVAL;
 	}
 	if (md_check_no_bitmap(mddev))
@@ -330,15 +353,27 @@ static int raid0_run(mddev_t *mddev)
 	blk_queue_max_hw_sectors(mddev->queue, mddev->chunk_sectors);
 	mddev->queue->queue_lock = &mddev->queue->__queue_lock;
 
-	ret = create_strip_zones(mddev);
-	if (ret < 0)
-		return ret;
+	/* if private is not null, we are here after takeover */
+	if (mddev->private == NULL) {
+		ret = create_strip_zones(mddev, &conf);
+		if (ret < 0)
+			return ret;
+		mddev->private = conf;
+	}
+	conf = mddev->private;
+	if (conf->scale_raid_disks) {
+		int i;
+		for (i=0; i < conf->strip_zone[0].nb_dev; i++)
+			conf->devlist[i]->raid_disk /= conf->scale_raid_disks;
+		/* FIXME update sysfs rd links */
+	}
 
 	/* calculate array device size */
 	md_set_array_sectors(mddev, raid0_size(mddev, 0, 0));
 
-	printk(KERN_INFO "raid0 : md_size is %llu sectors.\n",
-		(unsigned long long)mddev->array_sectors);
+	printk(KERN_INFO "md/raid0:%s: md_size is %llu sectors.\n",
+	       mdname(mddev),
+	       (unsigned long long)mddev->array_sectors);
 	/* calculate the max read-ahead size.
 	 * For read-ahead of large files to be effective, we need to
 	 * readahead at least twice a whole stripe. i.e. number of devices
@@ -402,6 +437,7 @@ static mdk_rdev_t *map_sector(mddev_t *mddev, struct strip_zone *zone,
 	unsigned int sect_in_chunk;
 	sector_t chunk;
 	raid0_conf_t *conf = mddev->private;
+	int raid_disks = conf->strip_zone[0].nb_dev;
 	unsigned int chunk_sects = mddev->chunk_sectors;
 
 	if (is_power_of_2(chunk_sects)) {
@@ -424,7 +460,7 @@ static mdk_rdev_t *map_sector(mddev_t *mddev, struct strip_zone *zone,
 	*	+ the position in the chunk
 	*/
 	*sector_offset = (chunk * chunk_sects) + sect_in_chunk;
-	return conf->devlist[(zone - conf->strip_zone)*mddev->raid_disks
+	return conf->devlist[(zone - conf->strip_zone)*raid_disks
 			     + sector_div(sector, zone->nb_dev)];
 }
 
@@ -444,27 +480,18 @@ static inline int is_io_in_chunk_boundary(mddev_t *mddev,
 	}
 }
 
-static int raid0_make_request(struct request_queue *q, struct bio *bio)
+static int raid0_make_request(mddev_t *mddev, struct bio *bio)
 {
-	mddev_t *mddev = q->queuedata;
 	unsigned int chunk_sects;
 	sector_t sector_offset;
 	struct strip_zone *zone;
 	mdk_rdev_t *tmp_dev;
-	const int rw = bio_data_dir(bio);
-	int cpu;
 
 	if (unlikely(bio_rw_flagged(bio, BIO_RW_BARRIER))) {
 		md_barrier_request(mddev, bio);
 		return 0;
 	}
 
-	cpu = part_stat_lock();
-	part_stat_inc(cpu, &mddev->gendisk->part0, ios[rw]);
-	part_stat_add(cpu, &mddev->gendisk->part0, sectors[rw],
-		      bio_sectors(bio));
-	part_stat_unlock();
-
 	chunk_sects = mddev->chunk_sectors;
 	if (unlikely(!is_io_in_chunk_boundary(mddev, chunk_sects, bio))) {
 		sector_t sector = bio->bi_sector;
@@ -482,9 +509,9 @@ static int raid0_make_request(struct request_queue *q, struct bio *bio)
 		else
 			bp = bio_split(bio, chunk_sects -
 				       sector_div(sector, chunk_sects));
-		if (raid0_make_request(q, &bp->bio1))
+		if (raid0_make_request(mddev, &bp->bio1))
 			generic_make_request(&bp->bio1);
-		if (raid0_make_request(q, &bp->bio2))
+		if (raid0_make_request(mddev, &bp->bio2))
 			generic_make_request(&bp->bio2);
 
 		bio_pair_release(bp);
@@ -504,9 +531,10 @@ static int raid0_make_request(struct request_queue *q, struct bio *bio)
 	return 1;
 
 bad_map:
-	printk("raid0_make_request bug: can't convert block across chunks"
-		" or bigger than %dk %llu %d\n", chunk_sects / 2,
-		(unsigned long long)bio->bi_sector, bio->bi_size >> 10);
+	printk("md/raid0:%s: make_request bug: can't convert block across chunks"
+	       " or bigger than %dk %llu %d\n",
+	       mdname(mddev), chunk_sects / 2,
+	       (unsigned long long)bio->bi_sector, bio->bi_size >> 10);
 
 	bio_io_error(bio);
 	return 0;
@@ -519,6 +547,7 @@ static void raid0_status(struct seq_file *seq, mddev_t *mddev)
 	int j, k, h;
 	char b[BDEVNAME_SIZE];
 	raid0_conf_t *conf = mddev->private;
+	int raid_disks = conf->strip_zone[0].nb_dev;
 
 	sector_t zone_size;
 	sector_t zone_start = 0;
@@ -529,7 +558,7 @@ static void raid0_status(struct seq_file *seq, mddev_t *mddev)
 		seq_printf(seq, "=[");
 		for (k = 0; k < conf->strip_zone[j].nb_dev; k++)
 			seq_printf(seq, "%s/", bdevname(
-				conf->devlist[j*mddev->raid_disks + k]
+				conf->devlist[j*raid_disks + k]
 						->bdev, b));
 
 		zone_size  = conf->strip_zone[j].zone_end - zone_start;
@@ -544,6 +573,104 @@ static void raid0_status(struct seq_file *seq, mddev_t *mddev)
 	return;
 }
 
+static void *raid0_takeover_raid5(mddev_t *mddev)
+{
+	mdk_rdev_t *rdev;
+	raid0_conf_t *priv_conf;
+
+	if (mddev->degraded != 1) {
+		printk(KERN_ERR "md/raid0:%s: raid5 must be degraded! Degraded disks: %d\n",
+		       mdname(mddev),
+		       mddev->degraded);
+		return ERR_PTR(-EINVAL);
+	}
+
+	list_for_each_entry(rdev, &mddev->disks, same_set) {
+		/* check slot number for a disk */
+		if (rdev->raid_disk == mddev->raid_disks-1) {
+			printk(KERN_ERR "md/raid0:%s: raid5 must have missing parity disk!\n",
+			       mdname(mddev));
+			return ERR_PTR(-EINVAL);
+		}
+	}
+
+	/* Set new parameters */
+	mddev->new_level = 0;
+	mddev->new_chunk_sectors = mddev->chunk_sectors;
+	mddev->raid_disks--;
+	mddev->delta_disks = -1;
+	/* make sure it will be not marked as dirty */
+	mddev->recovery_cp = MaxSector;
+
+	create_strip_zones(mddev, &priv_conf);
+	return priv_conf;
+}
+
+static void *raid0_takeover_raid10(mddev_t *mddev)
+{
+	raid0_conf_t *priv_conf;
+
+	/* Check layout:
+	 *  - far_copies must be 1
+	 *  - near_copies must be 2
+	 *  - disks number must be even
+	 *  - all mirrors must be already degraded
+	 */
+	if (mddev->layout != ((1 << 8) + 2)) {
+		printk(KERN_ERR "md/raid0:%s:: Raid0 cannot takover layout: 0x%x\n",
+		       mdname(mddev),
+		       mddev->layout);
+		return ERR_PTR(-EINVAL);
+	}
+	if (mddev->raid_disks & 1) {
+		printk(KERN_ERR "md/raid0:%s: Raid0 cannot takover Raid10 with odd disk number.\n",
+		       mdname(mddev));
+		return ERR_PTR(-EINVAL);
+	}
+	if (mddev->degraded != (mddev->raid_disks>>1)) {
+		printk(KERN_ERR "md/raid0:%s: All mirrors must be already degraded!\n",
+		       mdname(mddev));
+		return ERR_PTR(-EINVAL);
+	}
+
+	/* Set new parameters */
+	mddev->new_level = 0;
+	mddev->new_chunk_sectors = mddev->chunk_sectors;
+	mddev->delta_disks = - mddev->raid_disks / 2;
+	mddev->raid_disks += mddev->delta_disks;
+	mddev->degraded = 0;
+	/* make sure it will be not marked as dirty */
+	mddev->recovery_cp = MaxSector;
+
+	create_strip_zones(mddev, &priv_conf);
+	priv_conf->scale_raid_disks = 2;
+	return priv_conf;
+}
+
+static void *raid0_takeover(mddev_t *mddev)
+{
+	/* raid0 can take over:
+	 *  raid5 - providing it is Raid4 layout and one disk is faulty
+	 *  raid10 - assuming we have all necessary active disks
+	 */
+	if (mddev->level == 5) {
+		if (mddev->layout == ALGORITHM_PARITY_N)
+			return raid0_takeover_raid5(mddev);
+
+		printk(KERN_ERR "md/raid0:%s: Raid can only takeover Raid5 with layout: %d\n",
+		       mdname(mddev), ALGORITHM_PARITY_N);
+	}
+
+	if (mddev->level == 10)
+		return raid0_takeover_raid10(mddev);
+
+	return ERR_PTR(-EINVAL);
+}
+
+static void raid0_quiesce(mddev_t *mddev, int state)
+{
+}
+
 static struct mdk_personality raid0_personality=
 {
 	.name		= "raid0",
@@ -554,6 +681,8 @@ static struct mdk_personality raid0_personality=
 	.stop		= raid0_stop,
 	.status		= raid0_status,
 	.size		= raid0_size,
+	.takeover	= raid0_takeover,
+	.quiesce	= raid0_quiesce,
 };
 
 static int __init raid0_init (void)
diff --git a/drivers/md/raid0.h b/drivers/md/raid0.h
index 91f8e876ee64..d724e664ca4d 100644
--- a/drivers/md/raid0.h
+++ b/drivers/md/raid0.h
@@ -13,6 +13,9 @@ struct raid0_private_data
 	struct strip_zone *strip_zone;
 	mdk_rdev_t **devlist; /* lists of rdevs, pointed to by strip_zone->dev */
 	int nr_strip_zones;
+	int scale_raid_disks; /* divide rdev->raid_disks by this in run()
+			       * to handle conversion from raid10
+			       */
 };
 
 typedef struct raid0_private_data raid0_conf_t;
diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
index e59b10e66edb..a948da8012de 100644
--- a/drivers/md/raid1.c
+++ b/drivers/md/raid1.c
@@ -263,7 +263,7 @@ static inline void update_head_pos(int disk, r1bio_t *r1_bio)
 static void raid1_end_read_request(struct bio *bio, int error)
 {
 	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
-	r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
+	r1bio_t *r1_bio = bio->bi_private;
 	int mirror;
 	conf_t *conf = r1_bio->mddev->private;
 
@@ -297,7 +297,8 @@ static void raid1_end_read_request(struct bio *bio, int error)
 		 */
 		char b[BDEVNAME_SIZE];
 		if (printk_ratelimit())
-			printk(KERN_ERR "raid1: %s: rescheduling sector %llu\n",
+			printk(KERN_ERR "md/raid1:%s: %s: rescheduling sector %llu\n",
+			       mdname(conf->mddev),
 			       bdevname(conf->mirrors[mirror].rdev->bdev,b), (unsigned long long)r1_bio->sector);
 		reschedule_retry(r1_bio);
 	}
@@ -308,7 +309,7 @@ static void raid1_end_read_request(struct bio *bio, int error)
 static void raid1_end_write_request(struct bio *bio, int error)
 {
 	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
-	r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
+	r1bio_t *r1_bio = bio->bi_private;
 	int mirror, behind = test_bit(R1BIO_BehindIO, &r1_bio->state);
 	conf_t *conf = r1_bio->mddev->private;
 	struct bio *to_put = NULL;
@@ -418,7 +419,7 @@ static void raid1_end_write_request(struct bio *bio, int error)
  */
 static int read_balance(conf_t *conf, r1bio_t *r1_bio)
 {
-	const unsigned long this_sector = r1_bio->sector;
+	const sector_t this_sector = r1_bio->sector;
 	int new_disk = conf->last_used, disk = new_disk;
 	int wonly_disk = -1;
 	const int sectors = r1_bio->sectors;
@@ -434,7 +435,7 @@ static int read_balance(conf_t *conf, r1bio_t *r1_bio)
  retry:
 	if (conf->mddev->recovery_cp < MaxSector &&
 	    (this_sector + sectors >= conf->next_resync)) {
-		/* Choose the first operation device, for consistancy */
+		/* Choose the first operational device, for consistancy */
 		new_disk = 0;
 
 		for (rdev = rcu_dereference(conf->mirrors[new_disk].rdev);
@@ -774,9 +775,8 @@ do_sync_io:
 	return NULL;
 }
 
-static int make_request(struct request_queue *q, struct bio * bio)
+static int make_request(mddev_t *mddev, struct bio * bio)
 {
-	mddev_t *mddev = q->queuedata;
 	conf_t *conf = mddev->private;
 	mirror_info_t *mirror;
 	r1bio_t *r1_bio;
@@ -788,7 +788,6 @@ static int make_request(struct request_queue *q, struct bio * bio)
 	struct page **behind_pages = NULL;
 	const int rw = bio_data_dir(bio);
 	const bool do_sync = bio_rw_flagged(bio, BIO_RW_SYNCIO);
-	int cpu;
 	bool do_barriers;
 	mdk_rdev_t *blocked_rdev;
 
@@ -834,12 +833,6 @@ static int make_request(struct request_queue *q, struct bio * bio)
 
 	bitmap = mddev->bitmap;
 
-	cpu = part_stat_lock();
-	part_stat_inc(cpu, &mddev->gendisk->part0, ios[rw]);
-	part_stat_add(cpu, &mddev->gendisk->part0, sectors[rw],
-		      bio_sectors(bio));
-	part_stat_unlock();
-
 	/*
 	 * make_request() can abort the operation when READA is being
 	 * used and no empty request is available.
@@ -866,6 +859,15 @@ static int make_request(struct request_queue *q, struct bio * bio)
 		}
 		mirror = conf->mirrors + rdisk;
 
+		if (test_bit(WriteMostly, &mirror->rdev->flags) &&
+		    bitmap) {
+			/* Reading from a write-mostly device must
+			 * take care not to over-take any writes
+			 * that are 'behind'
+			 */
+			wait_event(bitmap->behind_wait,
+				   atomic_read(&bitmap->behind_writes) == 0);
+		}
 		r1_bio->read_disk = rdisk;
 
 		read_bio = bio_clone(bio, GFP_NOIO);
@@ -912,9 +914,10 @@ static int make_request(struct request_queue *q, struct bio * bio)
 			if (test_bit(Faulty, &rdev->flags)) {
 				rdev_dec_pending(rdev, mddev);
 				r1_bio->bios[i] = NULL;
-			} else
+			} else {
 				r1_bio->bios[i] = bio;
-			targets++;
+				targets++;
+			}
 		} else
 			r1_bio->bios[i] = NULL;
 	}
@@ -942,10 +945,14 @@ static int make_request(struct request_queue *q, struct bio * bio)
 		set_bit(R1BIO_Degraded, &r1_bio->state);
 	}
 
-	/* do behind I/O ? */
+	/* do behind I/O ?
+	 * Not if there are too many, or cannot allocate memory,
+	 * or a reader on WriteMostly is waiting for behind writes 
+	 * to flush */
 	if (bitmap &&
 	    (atomic_read(&bitmap->behind_writes)
 	     < mddev->bitmap_info.max_write_behind) &&
+	    !waitqueue_active(&bitmap->behind_wait) &&
 	    (behind_pages = alloc_behind_pages(bio)) != NULL)
 		set_bit(R1BIO_BehindIO, &r1_bio->state);
 
@@ -1070,21 +1077,22 @@ static void error(mddev_t *mddev, mdk_rdev_t *rdev)
 	} else
 		set_bit(Faulty, &rdev->flags);
 	set_bit(MD_CHANGE_DEVS, &mddev->flags);
-	printk(KERN_ALERT "raid1: Disk failure on %s, disabling device.\n"
-		"raid1: Operation continuing on %d devices.\n",
-		bdevname(rdev->bdev,b), conf->raid_disks - mddev->degraded);
+	printk(KERN_ALERT "md/raid1:%s: Disk failure on %s, disabling device.\n"
+	       KERN_ALERT "md/raid1:%s: Operation continuing on %d devices.\n",
+	       mdname(mddev), bdevname(rdev->bdev, b),
+	       mdname(mddev), conf->raid_disks - mddev->degraded);
 }
 
 static void print_conf(conf_t *conf)
 {
 	int i;
 
-	printk("RAID1 conf printout:\n");
+	printk(KERN_DEBUG "RAID1 conf printout:\n");
 	if (!conf) {
-		printk("(!conf)\n");
+		printk(KERN_DEBUG "(!conf)\n");
 		return;
 	}
-	printk(" --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
+	printk(KERN_DEBUG " --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
 		conf->raid_disks);
 
 	rcu_read_lock();
@@ -1092,7 +1100,7 @@ static void print_conf(conf_t *conf)
 		char b[BDEVNAME_SIZE];
 		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
 		if (rdev)
-			printk(" disk %d, wo:%d, o:%d, dev:%s\n",
+			printk(KERN_DEBUG " disk %d, wo:%d, o:%d, dev:%s\n",
 			       i, !test_bit(In_sync, &rdev->flags),
 			       !test_bit(Faulty, &rdev->flags),
 			       bdevname(rdev->bdev,b));
@@ -1223,7 +1231,7 @@ abort:
 
 static void end_sync_read(struct bio *bio, int error)
 {
-	r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
+	r1bio_t *r1_bio = bio->bi_private;
 	int i;
 
 	for (i=r1_bio->mddev->raid_disks; i--; )
@@ -1246,7 +1254,7 @@ static void end_sync_read(struct bio *bio, int error)
 static void end_sync_write(struct bio *bio, int error)
 {
 	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
-	r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
+	r1bio_t *r1_bio = bio->bi_private;
 	mddev_t *mddev = r1_bio->mddev;
 	conf_t *conf = mddev->private;
 	int i;
@@ -1453,9 +1461,10 @@ static void sync_request_write(mddev_t *mddev, r1bio_t *r1_bio)
 				char b[BDEVNAME_SIZE];
 				/* Cannot read from anywhere, array is toast */
 				md_error(mddev, conf->mirrors[r1_bio->read_disk].rdev);
-				printk(KERN_ALERT "raid1: %s: unrecoverable I/O read error"
+				printk(KERN_ALERT "md/raid1:%s: %s: unrecoverable I/O read error"
 				       " for block %llu\n",
-				       bdevname(bio->bi_bdev,b),
+				       mdname(mddev),
+				       bdevname(bio->bi_bdev, b),
 				       (unsigned long long)r1_bio->sector);
 				md_done_sync(mddev, r1_bio->sectors, 0);
 				put_buf(r1_bio);
@@ -1577,7 +1586,7 @@ static void fix_read_error(conf_t *conf, int read_disk,
 				else {
 					atomic_add(s, &rdev->corrected_errors);
 					printk(KERN_INFO
-					       "raid1:%s: read error corrected "
+					       "md/raid1:%s: read error corrected "
 					       "(%d sectors at %llu on %s)\n",
 					       mdname(mddev), s,
 					       (unsigned long long)(sect +
@@ -1682,8 +1691,9 @@ static void raid1d(mddev_t *mddev)
 
 			bio = r1_bio->bios[r1_bio->read_disk];
 			if ((disk=read_balance(conf, r1_bio)) == -1) {
-				printk(KERN_ALERT "raid1: %s: unrecoverable I/O"
+				printk(KERN_ALERT "md/raid1:%s: %s: unrecoverable I/O"
 				       " read error for block %llu\n",
+				       mdname(mddev),
 				       bdevname(bio->bi_bdev,b),
 				       (unsigned long long)r1_bio->sector);
 				raid_end_bio_io(r1_bio);
@@ -1697,10 +1707,11 @@ static void raid1d(mddev_t *mddev)
 				r1_bio->bios[r1_bio->read_disk] = bio;
 				rdev = conf->mirrors[disk].rdev;
 				if (printk_ratelimit())
-					printk(KERN_ERR "raid1: %s: redirecting sector %llu to"
-					       " another mirror\n",
-					       bdevname(rdev->bdev,b),
-					       (unsigned long long)r1_bio->sector);
+					printk(KERN_ERR "md/raid1:%s: redirecting sector %llu to"
+					       " other mirror: %s\n",
+					       mdname(mddev),
+					       (unsigned long long)r1_bio->sector,
+					       bdevname(rdev->bdev,b));
 				bio->bi_sector = r1_bio->sector + rdev->data_offset;
 				bio->bi_bdev = rdev->bdev;
 				bio->bi_end_io = raid1_end_read_request;
@@ -1755,13 +1766,8 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 	int still_degraded = 0;
 
 	if (!conf->r1buf_pool)
-	{
-/*
-		printk("sync start - bitmap %p\n", mddev->bitmap);
-*/
 		if (init_resync(conf))
 			return 0;
-	}
 
 	max_sector = mddev->dev_sectors;
 	if (sector_nr >= max_sector) {
@@ -2042,7 +2048,7 @@ static conf_t *setup_conf(mddev_t *mddev)
 
 	err = -EIO;
 	if (conf->last_used < 0) {
-		printk(KERN_ERR "raid1: no operational mirrors for %s\n",
+		printk(KERN_ERR "md/raid1:%s: no operational mirrors\n",
 		       mdname(mddev));
 		goto abort;
 	}
@@ -2050,7 +2056,7 @@ static conf_t *setup_conf(mddev_t *mddev)
 	conf->thread = md_register_thread(raid1d, mddev, NULL);
 	if (!conf->thread) {
 		printk(KERN_ERR
-		       "raid1: couldn't allocate thread for %s\n",
+		       "md/raid1:%s: couldn't allocate thread\n",
 		       mdname(mddev));
 		goto abort;
 	}
@@ -2076,12 +2082,12 @@ static int run(mddev_t *mddev)
 	mdk_rdev_t *rdev;
 
 	if (mddev->level != 1) {
-		printk("raid1: %s: raid level not set to mirroring (%d)\n",
+		printk(KERN_ERR "md/raid1:%s: raid level not set to mirroring (%d)\n",
 		       mdname(mddev), mddev->level);
 		return -EIO;
 	}
 	if (mddev->reshape_position != MaxSector) {
-		printk("raid1: %s: reshape_position set but not supported\n",
+		printk(KERN_ERR "md/raid1:%s: reshape_position set but not supported\n",
 		       mdname(mddev));
 		return -EIO;
 	}
@@ -2124,11 +2130,11 @@ static int run(mddev_t *mddev)
 		mddev->recovery_cp = MaxSector;
 
 	if (mddev->recovery_cp != MaxSector)
-		printk(KERN_NOTICE "raid1: %s is not clean"
+		printk(KERN_NOTICE "md/raid1:%s: not clean"
 		       " -- starting background reconstruction\n",
 		       mdname(mddev));
 	printk(KERN_INFO 
-		"raid1: raid set %s active with %d out of %d mirrors\n",
+		"md/raid1:%s: active with %d out of %d mirrors\n",
 		mdname(mddev), mddev->raid_disks - mddev->degraded, 
 		mddev->raid_disks);
 
@@ -2152,15 +2158,14 @@ static int stop(mddev_t *mddev)
 {
 	conf_t *conf = mddev->private;
 	struct bitmap *bitmap = mddev->bitmap;
-	int behind_wait = 0;
 
 	/* wait for behind writes to complete */
-	while (bitmap && atomic_read(&bitmap->behind_writes) > 0) {
-		behind_wait++;
-		printk(KERN_INFO "raid1: behind writes in progress on device %s, waiting to stop (%d)\n", mdname(mddev), behind_wait);
-		set_current_state(TASK_UNINTERRUPTIBLE);
-		schedule_timeout(HZ); /* wait a second */
+	if (bitmap && atomic_read(&bitmap->behind_writes) > 0) {
+		printk(KERN_INFO "md/raid1:%s: behind writes in progress - waiting to stop.\n",
+		       mdname(mddev));
 		/* need to kick something here to make sure I/O goes? */
+		wait_event(bitmap->behind_wait,
+			   atomic_read(&bitmap->behind_writes) == 0);
 	}
 
 	raise_barrier(conf);
@@ -2191,7 +2196,6 @@ static int raid1_resize(mddev_t *mddev, sector_t sectors)
 	if (mddev->array_sectors > raid1_size(mddev, sectors, 0))
 		return -EINVAL;
 	set_capacity(mddev->gendisk, mddev->array_sectors);
-	mddev->changed = 1;
 	revalidate_disk(mddev->gendisk);
 	if (sectors > mddev->dev_sectors &&
 	    mddev->recovery_cp == MaxSector) {
@@ -2286,9 +2290,9 @@ static int raid1_reshape(mddev_t *mddev)
 			if (sysfs_create_link(&mddev->kobj,
 					      &rdev->kobj, nm))
 				printk(KERN_WARNING
-				       "md/raid1: cannot register "
-				       "%s for %s\n",
-				       nm, mdname(mddev));
+				       "md/raid1:%s: cannot register "
+				       "%s\n",
+				       mdname(mddev), nm);
 		}
 		if (rdev)
 			newmirrors[d2++].rdev = rdev;
diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
index e2766d8251a1..03724992cdf2 100644
--- a/drivers/md/raid10.c
+++ b/drivers/md/raid10.c
@@ -24,6 +24,7 @@
 #include <linux/seq_file.h>
 #include "md.h"
 #include "raid10.h"
+#include "raid0.h"
 #include "bitmap.h"
 
 /*
@@ -255,7 +256,7 @@ static inline void update_head_pos(int slot, r10bio_t *r10_bio)
 static void raid10_end_read_request(struct bio *bio, int error)
 {
 	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
-	r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private);
+	r10bio_t *r10_bio = bio->bi_private;
 	int slot, dev;
 	conf_t *conf = r10_bio->mddev->private;
 
@@ -285,7 +286,8 @@ static void raid10_end_read_request(struct bio *bio, int error)
 		 */
 		char b[BDEVNAME_SIZE];
 		if (printk_ratelimit())
-			printk(KERN_ERR "raid10: %s: rescheduling sector %llu\n",
+			printk(KERN_ERR "md/raid10:%s: %s: rescheduling sector %llu\n",
+			       mdname(conf->mddev),
 			       bdevname(conf->mirrors[dev].rdev->bdev,b), (unsigned long long)r10_bio->sector);
 		reschedule_retry(r10_bio);
 	}
@@ -296,7 +298,7 @@ static void raid10_end_read_request(struct bio *bio, int error)
 static void raid10_end_write_request(struct bio *bio, int error)
 {
 	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
-	r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private);
+	r10bio_t *r10_bio = bio->bi_private;
 	int slot, dev;
 	conf_t *conf = r10_bio->mddev->private;
 
@@ -494,7 +496,7 @@ static int raid10_mergeable_bvec(struct request_queue *q,
  */
 static int read_balance(conf_t *conf, r10bio_t *r10_bio)
 {
-	const unsigned long this_sector = r10_bio->sector;
+	const sector_t this_sector = r10_bio->sector;
 	int disk, slot, nslot;
 	const int sectors = r10_bio->sectors;
 	sector_t new_distance, current_distance;
@@ -601,7 +603,7 @@ static void unplug_slaves(mddev_t *mddev)
 	int i;
 
 	rcu_read_lock();
-	for (i=0; i<mddev->raid_disks; i++) {
+	for (i=0; i < conf->raid_disks; i++) {
 		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
 		if (rdev && !test_bit(Faulty, &rdev->flags) && atomic_read(&rdev->nr_pending)) {
 			struct request_queue *r_queue = bdev_get_queue(rdev->bdev);
@@ -635,7 +637,7 @@ static int raid10_congested(void *data, int bits)
 	if (mddev_congested(mddev, bits))
 		return 1;
 	rcu_read_lock();
-	for (i = 0; i < mddev->raid_disks && ret == 0; i++) {
+	for (i = 0; i < conf->raid_disks && ret == 0; i++) {
 		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
 		if (rdev && !test_bit(Faulty, &rdev->flags)) {
 			struct request_queue *q = bdev_get_queue(rdev->bdev);
@@ -788,14 +790,12 @@ static void unfreeze_array(conf_t *conf)
 	spin_unlock_irq(&conf->resync_lock);
 }
 
-static int make_request(struct request_queue *q, struct bio * bio)
+static int make_request(mddev_t *mddev, struct bio * bio)
 {
-	mddev_t *mddev = q->queuedata;
 	conf_t *conf = mddev->private;
 	mirror_info_t *mirror;
 	r10bio_t *r10_bio;
 	struct bio *read_bio;
-	int cpu;
 	int i;
 	int chunk_sects = conf->chunk_mask + 1;
 	const int rw = bio_data_dir(bio);
@@ -825,16 +825,16 @@ static int make_request(struct request_queue *q, struct bio * bio)
 		 */
 		bp = bio_split(bio,
 			       chunk_sects - (bio->bi_sector & (chunk_sects - 1)) );
-		if (make_request(q, &bp->bio1))
+		if (make_request(mddev, &bp->bio1))
 			generic_make_request(&bp->bio1);
-		if (make_request(q, &bp->bio2))
+		if (make_request(mddev, &bp->bio2))
 			generic_make_request(&bp->bio2);
 
 		bio_pair_release(bp);
 		return 0;
 	bad_map:
-		printk("raid10_make_request bug: can't convert block across chunks"
-		       " or bigger than %dk %llu %d\n", chunk_sects/2,
+		printk("md/raid10:%s: make_request bug: can't convert block across chunks"
+		       " or bigger than %dk %llu %d\n", mdname(mddev), chunk_sects/2,
 		       (unsigned long long)bio->bi_sector, bio->bi_size >> 10);
 
 		bio_io_error(bio);
@@ -850,12 +850,6 @@ static int make_request(struct request_queue *q, struct bio * bio)
 	 */
 	wait_barrier(conf);
 
-	cpu = part_stat_lock();
-	part_stat_inc(cpu, &mddev->gendisk->part0, ios[rw]);
-	part_stat_add(cpu, &mddev->gendisk->part0, sectors[rw],
-		      bio_sectors(bio));
-	part_stat_unlock();
-
 	r10_bio = mempool_alloc(conf->r10bio_pool, GFP_NOIO);
 
 	r10_bio->master_bio = bio;
@@ -1039,9 +1033,10 @@ static void error(mddev_t *mddev, mdk_rdev_t *rdev)
 	}
 	set_bit(Faulty, &rdev->flags);
 	set_bit(MD_CHANGE_DEVS, &mddev->flags);
-	printk(KERN_ALERT "raid10: Disk failure on %s, disabling device.\n"
-		"raid10: Operation continuing on %d devices.\n",
-		bdevname(rdev->bdev,b), conf->raid_disks - mddev->degraded);
+	printk(KERN_ALERT "md/raid10:%s: Disk failure on %s, disabling device.\n"
+	       KERN_ALERT "md/raid10:%s: Operation continuing on %d devices.\n",
+	       mdname(mddev), bdevname(rdev->bdev, b),
+	       mdname(mddev), conf->raid_disks - mddev->degraded);
 }
 
 static void print_conf(conf_t *conf)
@@ -1049,19 +1044,19 @@ static void print_conf(conf_t *conf)
 	int i;
 	mirror_info_t *tmp;
 
-	printk("RAID10 conf printout:\n");
+	printk(KERN_DEBUG "RAID10 conf printout:\n");
 	if (!conf) {
-		printk("(!conf)\n");
+		printk(KERN_DEBUG "(!conf)\n");
 		return;
 	}
-	printk(" --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
+	printk(KERN_DEBUG " --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
 		conf->raid_disks);
 
 	for (i = 0; i < conf->raid_disks; i++) {
 		char b[BDEVNAME_SIZE];
 		tmp = conf->mirrors + i;
 		if (tmp->rdev)
-			printk(" disk %d, wo:%d, o:%d, dev:%s\n",
+			printk(KERN_DEBUG " disk %d, wo:%d, o:%d, dev:%s\n",
 				i, !test_bit(In_sync, &tmp->rdev->flags),
 			        !test_bit(Faulty, &tmp->rdev->flags),
 				bdevname(tmp->rdev->bdev,b));
@@ -1132,7 +1127,7 @@ static int raid10_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
 	int mirror;
 	mirror_info_t *p;
 	int first = 0;
-	int last = mddev->raid_disks - 1;
+	int last = conf->raid_disks - 1;
 
 	if (mddev->recovery_cp < MaxSector)
 		/* only hot-add to in-sync arrays, as recovery is
@@ -1224,7 +1219,7 @@ abort:
 
 static void end_sync_read(struct bio *bio, int error)
 {
-	r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private);
+	r10bio_t *r10_bio = bio->bi_private;
 	conf_t *conf = r10_bio->mddev->private;
 	int i,d;
 
@@ -1261,7 +1256,7 @@ static void end_sync_read(struct bio *bio, int error)
 static void end_sync_write(struct bio *bio, int error)
 {
 	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
-	r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private);
+	r10bio_t *r10_bio = bio->bi_private;
 	mddev_t *mddev = r10_bio->mddev;
 	conf_t *conf = mddev->private;
 	int i,d;
@@ -1510,13 +1505,14 @@ static void fix_read_error(conf_t *conf, mddev_t *mddev, r10bio_t *r10_bio)
 		if (cur_read_error_count > max_read_errors) {
 			rcu_read_unlock();
 			printk(KERN_NOTICE
-			       "raid10: %s: Raid device exceeded "
+			       "md/raid10:%s: %s: Raid device exceeded "
 			       "read_error threshold "
 			       "[cur %d:max %d]\n",
+			       mdname(mddev),
 			       b, cur_read_error_count, max_read_errors);
 			printk(KERN_NOTICE
-			       "raid10: %s: Failing raid "
-			       "device\n", b);
+			       "md/raid10:%s: %s: Failing raid "
+			       "device\n", mdname(mddev), b);
 			md_error(mddev, conf->mirrors[d].rdev);
 			return;
 		}
@@ -1586,15 +1582,16 @@ static void fix_read_error(conf_t *conf, mddev_t *mddev, r10bio_t *r10_bio)
 				    == 0) {
 					/* Well, this device is dead */
 					printk(KERN_NOTICE
-					       "raid10:%s: read correction "
+					       "md/raid10:%s: read correction "
 					       "write failed"
 					       " (%d sectors at %llu on %s)\n",
 					       mdname(mddev), s,
 					       (unsigned long long)(sect+
 					       rdev->data_offset),
 					       bdevname(rdev->bdev, b));
-					printk(KERN_NOTICE "raid10:%s: failing "
+					printk(KERN_NOTICE "md/raid10:%s: %s: failing "
 					       "drive\n",
+					       mdname(mddev),
 					       bdevname(rdev->bdev, b));
 					md_error(mddev, rdev);
 				}
@@ -1622,20 +1619,21 @@ static void fix_read_error(conf_t *conf, mddev_t *mddev, r10bio_t *r10_bio)
 						 READ) == 0) {
 					/* Well, this device is dead */
 					printk(KERN_NOTICE
-					       "raid10:%s: unable to read back "
+					       "md/raid10:%s: unable to read back "
 					       "corrected sectors"
 					       " (%d sectors at %llu on %s)\n",
 					       mdname(mddev), s,
 					       (unsigned long long)(sect+
 						    rdev->data_offset),
 					       bdevname(rdev->bdev, b));
-					printk(KERN_NOTICE "raid10:%s: failing drive\n",
+					printk(KERN_NOTICE "md/raid10:%s: %s: failing drive\n",
+					       mdname(mddev),
 					       bdevname(rdev->bdev, b));
 
 					md_error(mddev, rdev);
 				} else {
 					printk(KERN_INFO
-					       "raid10:%s: read error corrected"
+					       "md/raid10:%s: read error corrected"
 					       " (%d sectors at %llu on %s)\n",
 					       mdname(mddev), s,
 					       (unsigned long long)(sect+
@@ -1710,8 +1708,9 @@ static void raid10d(mddev_t *mddev)
 				mddev->ro ? IO_BLOCKED : NULL;
 			mirror = read_balance(conf, r10_bio);
 			if (mirror == -1) {
-				printk(KERN_ALERT "raid10: %s: unrecoverable I/O"
+				printk(KERN_ALERT "md/raid10:%s: %s: unrecoverable I/O"
 				       " read error for block %llu\n",
+				       mdname(mddev),
 				       bdevname(bio->bi_bdev,b),
 				       (unsigned long long)r10_bio->sector);
 				raid_end_bio_io(r10_bio);
@@ -1721,8 +1720,9 @@ static void raid10d(mddev_t *mddev)
 				bio_put(bio);
 				rdev = conf->mirrors[mirror].rdev;
 				if (printk_ratelimit())
-					printk(KERN_ERR "raid10: %s: redirecting sector %llu to"
+					printk(KERN_ERR "md/raid10:%s: %s: redirecting sector %llu to"
 					       " another mirror\n",
+					       mdname(mddev),
 					       bdevname(rdev->bdev,b),
 					       (unsigned long long)r10_bio->sector);
 				bio = bio_clone(r10_bio->master_bio, GFP_NOIO);
@@ -1980,7 +1980,8 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 					r10_bio = rb2;
 					if (!test_and_set_bit(MD_RECOVERY_INTR,
 							      &mddev->recovery))
-						printk(KERN_INFO "raid10: %s: insufficient working devices for recovery.\n",
+						printk(KERN_INFO "md/raid10:%s: insufficient "
+						       "working devices for recovery.\n",
 						       mdname(mddev));
 					break;
 				}
@@ -2140,9 +2141,9 @@ raid10_size(mddev_t *mddev, sector_t sectors, int raid_disks)
 	conf_t *conf = mddev->private;
 
 	if (!raid_disks)
-		raid_disks = mddev->raid_disks;
+		raid_disks = conf->raid_disks;
 	if (!sectors)
-		sectors = mddev->dev_sectors;
+		sectors = conf->dev_sectors;
 
 	size = sectors >> conf->chunk_shift;
 	sector_div(size, conf->far_copies);
@@ -2152,62 +2153,61 @@ raid10_size(mddev_t *mddev, sector_t sectors, int raid_disks)
 	return size << conf->chunk_shift;
 }
 
-static int run(mddev_t *mddev)
+
+static conf_t *setup_conf(mddev_t *mddev)
 {
-	conf_t *conf;
-	int i, disk_idx, chunk_size;
-	mirror_info_t *disk;
-	mdk_rdev_t *rdev;
+	conf_t *conf = NULL;
 	int nc, fc, fo;
 	sector_t stride, size;
+	int err = -EINVAL;
 
 	if (mddev->chunk_sectors < (PAGE_SIZE >> 9) ||
 	    !is_power_of_2(mddev->chunk_sectors)) {
-		printk(KERN_ERR "md/raid10: chunk size must be "
-		       "at least PAGE_SIZE(%ld) and be a power of 2.\n", PAGE_SIZE);
-		return -EINVAL;
+		printk(KERN_ERR "md/raid10:%s: chunk size must be "
+		       "at least PAGE_SIZE(%ld) and be a power of 2.\n",
+		       mdname(mddev), PAGE_SIZE);
+		goto out;
 	}
 
 	nc = mddev->layout & 255;
 	fc = (mddev->layout >> 8) & 255;
 	fo = mddev->layout & (1<<16);
+
 	if ((nc*fc) <2 || (nc*fc) > mddev->raid_disks ||
 	    (mddev->layout >> 17)) {
-		printk(KERN_ERR "raid10: %s: unsupported raid10 layout: 0x%8x\n",
+		printk(KERN_ERR "md/raid10:%s: unsupported raid10 layout: 0x%8x\n",
 		       mdname(mddev), mddev->layout);
 		goto out;
 	}
-	/*
-	 * copy the already verified devices into our private RAID10
-	 * bookkeeping area. [whatever we allocate in run(),
-	 * should be freed in stop()]
-	 */
+
+	err = -ENOMEM;
 	conf = kzalloc(sizeof(conf_t), GFP_KERNEL);
-	mddev->private = conf;
-	if (!conf) {
-		printk(KERN_ERR "raid10: couldn't allocate memory for %s\n",
-			mdname(mddev));
+	if (!conf)
 		goto out;
-	}
+
 	conf->mirrors = kzalloc(sizeof(struct mirror_info)*mddev->raid_disks,
-				 GFP_KERNEL);
-	if (!conf->mirrors) {
-		printk(KERN_ERR "raid10: couldn't allocate memory for %s\n",
-		       mdname(mddev));
-		goto out_free_conf;
-	}
+				GFP_KERNEL);
+	if (!conf->mirrors)
+		goto out;
 
 	conf->tmppage = alloc_page(GFP_KERNEL);
 	if (!conf->tmppage)
-		goto out_free_conf;
+		goto out;
+
 
 	conf->raid_disks = mddev->raid_disks;
 	conf->near_copies = nc;
 	conf->far_copies = fc;
 	conf->copies = nc*fc;
 	conf->far_offset = fo;
-	conf->chunk_mask = mddev->chunk_sectors - 1;
-	conf->chunk_shift = ffz(~mddev->chunk_sectors);
+	conf->chunk_mask = mddev->new_chunk_sectors - 1;
+	conf->chunk_shift = ffz(~mddev->new_chunk_sectors);
+
+	conf->r10bio_pool = mempool_create(NR_RAID10_BIOS, r10bio_pool_alloc,
+					   r10bio_pool_free, conf);
+	if (!conf->r10bio_pool)
+		goto out;
+
 	size = mddev->dev_sectors >> conf->chunk_shift;
 	sector_div(size, fc);
 	size = size * conf->raid_disks;
@@ -2221,7 +2221,8 @@ static int run(mddev_t *mddev)
 	 */
 	stride += conf->raid_disks - 1;
 	sector_div(stride, conf->raid_disks);
-	mddev->dev_sectors = stride << conf->chunk_shift;
+
+	conf->dev_sectors = stride << conf->chunk_shift;
 
 	if (fo)
 		stride = 1;
@@ -2229,18 +2230,63 @@ static int run(mddev_t *mddev)
 		sector_div(stride, fc);
 	conf->stride = stride << conf->chunk_shift;
 
-	conf->r10bio_pool = mempool_create(NR_RAID10_BIOS, r10bio_pool_alloc,
-						r10bio_pool_free, conf);
-	if (!conf->r10bio_pool) {
-		printk(KERN_ERR "raid10: couldn't allocate memory for %s\n",
-			mdname(mddev));
-		goto out_free_conf;
-	}
 
-	conf->mddev = mddev;
 	spin_lock_init(&conf->device_lock);
+	INIT_LIST_HEAD(&conf->retry_list);
+
+	spin_lock_init(&conf->resync_lock);
+	init_waitqueue_head(&conf->wait_barrier);
+
+	conf->thread = md_register_thread(raid10d, mddev, NULL);
+	if (!conf->thread)
+		goto out;
+
+	conf->scale_disks = 0;
+	conf->mddev = mddev;
+	return conf;
+
+ out:
+	printk(KERN_ERR "md/raid10:%s: couldn't allocate memory.\n",
+	       mdname(mddev));
+	if (conf) {
+		if (conf->r10bio_pool)
+			mempool_destroy(conf->r10bio_pool);
+		kfree(conf->mirrors);
+		safe_put_page(conf->tmppage);
+		kfree(conf);
+	}
+	return ERR_PTR(err);
+}
+
+static int run(mddev_t *mddev)
+{
+	conf_t *conf;
+	int i, disk_idx, chunk_size;
+	mirror_info_t *disk;
+	mdk_rdev_t *rdev;
+	sector_t size;
+
+	/*
+	 * copy the already verified devices into our private RAID10
+	 * bookkeeping area. [whatever we allocate in run(),
+	 * should be freed in stop()]
+	 */
+
+	if (mddev->private == NULL) {
+		conf = setup_conf(mddev);
+		if (IS_ERR(conf))
+			return PTR_ERR(conf);
+		mddev->private = conf;
+	}
+	conf = mddev->private;
+	if (!conf)
+		goto out;
+
 	mddev->queue->queue_lock = &conf->device_lock;
 
+	mddev->thread = conf->thread;
+	conf->thread = NULL;
+
 	chunk_size = mddev->chunk_sectors << 9;
 	blk_queue_io_min(mddev->queue, chunk_size);
 	if (conf->raid_disks % conf->near_copies)
@@ -2251,9 +2297,14 @@ static int run(mddev_t *mddev)
 
 	list_for_each_entry(rdev, &mddev->disks, same_set) {
 		disk_idx = rdev->raid_disk;
-		if (disk_idx >= mddev->raid_disks
+		if (disk_idx >= conf->raid_disks
 		    || disk_idx < 0)
 			continue;
+		if (conf->scale_disks) {
+			disk_idx *= conf->scale_disks;
+			rdev->raid_disk = disk_idx;
+			/* MOVE 'rd%d' link !! */
+		}
 		disk = conf->mirrors + disk_idx;
 
 		disk->rdev = rdev;
@@ -2271,14 +2322,9 @@ static int run(mddev_t *mddev)
 
 		disk->head_position = 0;
 	}
-	INIT_LIST_HEAD(&conf->retry_list);
-
-	spin_lock_init(&conf->resync_lock);
-	init_waitqueue_head(&conf->wait_barrier);
-
 	/* need to check that every block has at least one working mirror */
 	if (!enough(conf)) {
-		printk(KERN_ERR "raid10: not enough operational mirrors for %s\n",
+		printk(KERN_ERR "md/raid10:%s: not enough operational mirrors.\n",
 		       mdname(mddev));
 		goto out_free_conf;
 	}
@@ -2297,28 +2343,21 @@ static int run(mddev_t *mddev)
 		}
 	}
 
-
-	mddev->thread = md_register_thread(raid10d, mddev, NULL);
-	if (!mddev->thread) {
-		printk(KERN_ERR
-		       "raid10: couldn't allocate thread for %s\n",
-		       mdname(mddev));
-		goto out_free_conf;
-	}
-
 	if (mddev->recovery_cp != MaxSector)
-		printk(KERN_NOTICE "raid10: %s is not clean"
+		printk(KERN_NOTICE "md/raid10:%s: not clean"
 		       " -- starting background reconstruction\n",
 		       mdname(mddev));
 	printk(KERN_INFO
-		"raid10: raid set %s active with %d out of %d devices\n",
-		mdname(mddev), mddev->raid_disks - mddev->degraded,
-		mddev->raid_disks);
+		"md/raid10:%s: active with %d out of %d devices\n",
+		mdname(mddev), conf->raid_disks - mddev->degraded,
+		conf->raid_disks);
 	/*
 	 * Ok, everything is just fine now
 	 */
-	md_set_array_sectors(mddev, raid10_size(mddev, 0, 0));
-	mddev->resync_max_sectors = raid10_size(mddev, 0, 0);
+	mddev->dev_sectors = conf->dev_sectors;
+	size = raid10_size(mddev, 0, 0);
+	md_set_array_sectors(mddev, size);
+	mddev->resync_max_sectors = size;
 
 	mddev->queue->unplug_fn = raid10_unplug;
 	mddev->queue->backing_dev_info.congested_fn = raid10_congested;
@@ -2336,7 +2375,7 @@ static int run(mddev_t *mddev)
 			mddev->queue->backing_dev_info.ra_pages = 2* stripe;
 	}
 
-	if (conf->near_copies < mddev->raid_disks)
+	if (conf->near_copies < conf->raid_disks)
 		blk_queue_merge_bvec(mddev->queue, raid10_mergeable_bvec);
 	md_integrity_register(mddev);
 	return 0;
@@ -2348,6 +2387,7 @@ out_free_conf:
 	kfree(conf->mirrors);
 	kfree(conf);
 	mddev->private = NULL;
+	md_unregister_thread(mddev->thread);
 out:
 	return -EIO;
 }
@@ -2384,6 +2424,61 @@ static void raid10_quiesce(mddev_t *mddev, int state)
 	}
 }
 
+static void *raid10_takeover_raid0(mddev_t *mddev)
+{
+	mdk_rdev_t *rdev;
+	conf_t *conf;
+
+	if (mddev->degraded > 0) {
+		printk(KERN_ERR "md/raid10:%s: Error: degraded raid0!\n",
+		       mdname(mddev));
+		return ERR_PTR(-EINVAL);
+	}
+
+	/* Update slot numbers to obtain
+	 * degraded raid10 with missing mirrors
+	 */
+	list_for_each_entry(rdev, &mddev->disks, same_set) {
+		rdev->raid_disk *= 2;
+	}
+
+	/* Set new parameters */
+	mddev->new_level = 10;
+	/* new layout: far_copies = 1, near_copies = 2 */
+	mddev->new_layout = (1<<8) + 2;
+	mddev->new_chunk_sectors = mddev->chunk_sectors;
+	mddev->delta_disks = mddev->raid_disks;
+	mddev->degraded = mddev->raid_disks;
+	mddev->raid_disks *= 2;
+	/* make sure it will be not marked as dirty */
+	mddev->recovery_cp = MaxSector;
+
+	conf = setup_conf(mddev);
+	conf->scale_disks = 2;
+	return conf;
+}
+
+static void *raid10_takeover(mddev_t *mddev)
+{
+	struct raid0_private_data *raid0_priv;
+
+	/* raid10 can take over:
+	 *  raid0 - providing it has only two drives
+	 */
+	if (mddev->level == 0) {
+		/* for raid0 takeover only one zone is supported */
+		raid0_priv = mddev->private;
+		if (raid0_priv->nr_strip_zones > 1) {
+			printk(KERN_ERR "md/raid10:%s: cannot takeover raid 0"
+			       " with more than one zone.\n",
+			       mdname(mddev));
+			return ERR_PTR(-EINVAL);
+		}
+		return raid10_takeover_raid0(mddev);
+	}
+	return ERR_PTR(-EINVAL);
+}
+
 static struct mdk_personality raid10_personality =
 {
 	.name		= "raid10",
@@ -2400,6 +2495,7 @@ static struct mdk_personality raid10_personality =
 	.sync_request	= sync_request,
 	.quiesce	= raid10_quiesce,
 	.size		= raid10_size,
+	.takeover	= raid10_takeover,
 };
 
 static int __init raid_init(void)
diff --git a/drivers/md/raid10.h b/drivers/md/raid10.h
index 59cd1efb8d30..3824a087e17c 100644
--- a/drivers/md/raid10.h
+++ b/drivers/md/raid10.h
@@ -33,9 +33,16 @@ struct r10_private_data_s {
 					       * 1 stripe.
 					       */
 
+	sector_t		dev_sectors;  /* temp copy of mddev->dev_sectors */
+
 	int chunk_shift; /* shift from chunks to sectors */
 	sector_t chunk_mask;
 
+	int			scale_disks;  /* When starting array, multiply
+					       * each ->raid_disk by this.
+					       * Need for raid0->raid10 migration
+					       */
+
 	struct list_head	retry_list;
 	/* queue pending writes and submit them on unplug */
 	struct bio_list		pending_bio_list;
@@ -57,6 +64,11 @@ struct r10_private_data_s {
 	mempool_t *r10bio_pool;
 	mempool_t *r10buf_pool;
 	struct page		*tmppage;
+
+	/* When taking over an array from a different personality, we store
+	 * the new thread here until we fully activate the array.
+	 */
+	struct mdk_thread_s	*thread;
 };
 
 typedef struct r10_private_data_s conf_t;
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 15348c393b5d..d2c0f94fa37d 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -53,6 +53,7 @@
 #include <linux/slab.h>
 #include "md.h"
 #include "raid5.h"
+#include "raid0.h"
 #include "bitmap.h"
 
 /*
@@ -1509,7 +1510,7 @@ static void raid5_end_read_request(struct bio * bi, int error)
 		set_bit(R5_UPTODATE, &sh->dev[i].flags);
 		if (test_bit(R5_ReadError, &sh->dev[i].flags)) {
 			rdev = conf->disks[i].rdev;
-			printk_rl(KERN_INFO "raid5:%s: read error corrected"
+			printk_rl(KERN_INFO "md/raid:%s: read error corrected"
 				  " (%lu sectors at %llu on %s)\n",
 				  mdname(conf->mddev), STRIPE_SECTORS,
 				  (unsigned long long)(sh->sector
@@ -1529,7 +1530,7 @@ static void raid5_end_read_request(struct bio * bi, int error)
 		atomic_inc(&rdev->read_errors);
 		if (conf->mddev->degraded >= conf->max_degraded)
 			printk_rl(KERN_WARNING
-				  "raid5:%s: read error not correctable "
+				  "md/raid:%s: read error not correctable "
 				  "(sector %llu on %s).\n",
 				  mdname(conf->mddev),
 				  (unsigned long long)(sh->sector
@@ -1538,7 +1539,7 @@ static void raid5_end_read_request(struct bio * bi, int error)
 		else if (test_bit(R5_ReWrite, &sh->dev[i].flags))
 			/* Oh, no!!! */
 			printk_rl(KERN_WARNING
-				  "raid5:%s: read error NOT corrected!! "
+				  "md/raid:%s: read error NOT corrected!! "
 				  "(sector %llu on %s).\n",
 				  mdname(conf->mddev),
 				  (unsigned long long)(sh->sector
@@ -1547,7 +1548,7 @@ static void raid5_end_read_request(struct bio * bi, int error)
 		else if (atomic_read(&rdev->read_errors)
 			 > conf->max_nr_stripes)
 			printk(KERN_WARNING
-			       "raid5:%s: Too many read errors, failing device %s.\n",
+			       "md/raid:%s: Too many read errors, failing device %s.\n",
 			       mdname(conf->mddev), bdn);
 		else
 			retry = 1;
@@ -1619,8 +1620,8 @@ static void raid5_build_block(struct stripe_head *sh, int i, int previous)
 static void error(mddev_t *mddev, mdk_rdev_t *rdev)
 {
 	char b[BDEVNAME_SIZE];
-	raid5_conf_t *conf = (raid5_conf_t *) mddev->private;
-	pr_debug("raid5: error called\n");
+	raid5_conf_t *conf = mddev->private;
+	pr_debug("raid456: error called\n");
 
 	if (!test_bit(Faulty, &rdev->flags)) {
 		set_bit(MD_CHANGE_DEVS, &mddev->flags);
@@ -1636,9 +1637,13 @@ static void error(mddev_t *mddev, mdk_rdev_t *rdev)
 		}
 		set_bit(Faulty, &rdev->flags);
 		printk(KERN_ALERT
-		       "raid5: Disk failure on %s, disabling device.\n"
-		       "raid5: Operation continuing on %d devices.\n",
-		       bdevname(rdev->bdev,b), conf->raid_disks - mddev->degraded);
+		       "md/raid:%s: Disk failure on %s, disabling device.\n"
+		       KERN_ALERT
+		       "md/raid:%s: Operation continuing on %d devices.\n",
+		       mdname(mddev),
+		       bdevname(rdev->bdev, b),
+		       mdname(mddev),
+		       conf->raid_disks - mddev->degraded);
 	}
 }
 
@@ -1714,8 +1719,6 @@ static sector_t raid5_compute_sector(raid5_conf_t *conf, sector_t r_sector,
 			pd_idx = data_disks;
 			break;
 		default:
-			printk(KERN_ERR "raid5: unsupported algorithm %d\n",
-				algorithm);
 			BUG();
 		}
 		break;
@@ -1832,10 +1835,7 @@ static sector_t raid5_compute_sector(raid5_conf_t *conf, sector_t r_sector,
 			qd_idx = raid_disks - 1;
 			break;
 
-
 		default:
-			printk(KERN_CRIT "raid6: unsupported algorithm %d\n",
-			       algorithm);
 			BUG();
 		}
 		break;
@@ -1898,8 +1898,6 @@ static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous)
 		case ALGORITHM_PARITY_N:
 			break;
 		default:
-			printk(KERN_ERR "raid5: unsupported algorithm %d\n",
-			       algorithm);
 			BUG();
 		}
 		break;
@@ -1958,8 +1956,6 @@ static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous)
 			i -= 1;
 			break;
 		default:
-			printk(KERN_CRIT "raid6: unsupported algorithm %d\n",
-			       algorithm);
 			BUG();
 		}
 		break;
@@ -1972,7 +1968,8 @@ static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous)
 				     previous, &dummy1, &sh2);
 	if (check != sh->sector || dummy1 != dd_idx || sh2.pd_idx != sh->pd_idx
 		|| sh2.qd_idx != sh->qd_idx) {
-		printk(KERN_ERR "compute_blocknr: map not correct\n");
+		printk(KERN_ERR "md/raid:%s: compute_blocknr: map not correct\n",
+		       mdname(conf->mddev));
 		return 0;
 	}
 	return r_sector;
@@ -3709,10 +3706,10 @@ static void raid5_align_endio(struct bio *bi, int error)
 
 	bio_put(bi);
 
-	mddev = raid_bi->bi_bdev->bd_disk->queue->queuedata;
-	conf = mddev->private;
 	rdev = (void*)raid_bi->bi_next;
 	raid_bi->bi_next = NULL;
+	mddev = rdev->mddev;
+	conf = mddev->private;
 
 	rdev_dec_pending(rdev, conf->mddev);
 
@@ -3749,9 +3746,8 @@ static int bio_fits_rdev(struct bio *bi)
 }
 
 
-static int chunk_aligned_read(struct request_queue *q, struct bio * raid_bio)
+static int chunk_aligned_read(mddev_t *mddev, struct bio * raid_bio)
 {
-	mddev_t *mddev = q->queuedata;
 	raid5_conf_t *conf = mddev->private;
 	int dd_idx;
 	struct bio* align_bi;
@@ -3866,16 +3862,15 @@ static struct stripe_head *__get_priority_stripe(raid5_conf_t *conf)
 	return sh;
 }
 
-static int make_request(struct request_queue *q, struct bio * bi)
+static int make_request(mddev_t *mddev, struct bio * bi)
 {
-	mddev_t *mddev = q->queuedata;
 	raid5_conf_t *conf = mddev->private;
 	int dd_idx;
 	sector_t new_sector;
 	sector_t logical_sector, last_sector;
 	struct stripe_head *sh;
 	const int rw = bio_data_dir(bi);
-	int cpu, remaining;
+	int remaining;
 
 	if (unlikely(bio_rw_flagged(bi, BIO_RW_BARRIER))) {
 		/* Drain all pending writes.  We only really need
@@ -3890,15 +3885,9 @@ static int make_request(struct request_queue *q, struct bio * bi)
 
 	md_write_start(mddev, bi);
 
-	cpu = part_stat_lock();
-	part_stat_inc(cpu, &mddev->gendisk->part0, ios[rw]);
-	part_stat_add(cpu, &mddev->gendisk->part0, sectors[rw],
-		      bio_sectors(bi));
-	part_stat_unlock();
-
 	if (rw == READ &&
 	     mddev->reshape_position == MaxSector &&
-	     chunk_aligned_read(q,bi))
+	     chunk_aligned_read(mddev,bi))
 		return 0;
 
 	logical_sector = bi->bi_sector & ~((sector_t)STRIPE_SECTORS-1);
@@ -3946,7 +3935,7 @@ static int make_request(struct request_queue *q, struct bio * bi)
 		new_sector = raid5_compute_sector(conf, logical_sector,
 						  previous,
 						  &dd_idx, NULL);
-		pr_debug("raid5: make_request, sector %llu logical %llu\n",
+		pr_debug("raid456: make_request, sector %llu logical %llu\n",
 			(unsigned long long)new_sector, 
 			(unsigned long long)logical_sector);
 
@@ -4054,7 +4043,7 @@ static sector_t reshape_request(mddev_t *mddev, sector_t sector_nr, int *skipped
 	 * As the reads complete, handle_stripe will copy the data
 	 * into the destination stripe and release that stripe.
 	 */
-	raid5_conf_t *conf = (raid5_conf_t *) mddev->private;
+	raid5_conf_t *conf = mddev->private;
 	struct stripe_head *sh;
 	sector_t first_sector, last_sector;
 	int raid_disks = conf->previous_raid_disks;
@@ -4263,7 +4252,7 @@ static sector_t reshape_request(mddev_t *mddev, sector_t sector_nr, int *skipped
 /* FIXME go_faster isn't used */
 static inline sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster)
 {
-	raid5_conf_t *conf = (raid5_conf_t *) mddev->private;
+	raid5_conf_t *conf = mddev->private;
 	struct stripe_head *sh;
 	sector_t max_sector = mddev->dev_sectors;
 	int sync_blocks;
@@ -4656,7 +4645,7 @@ static int raid456_cpu_notify(struct notifier_block *nfb, unsigned long action,
 			kfree(percpu->scribble);
 			pr_err("%s: failed memory allocation for cpu%ld\n",
 			       __func__, cpu);
-			return NOTIFY_BAD;
+			return notifier_from_errno(-ENOMEM);
 		}
 		break;
 	case CPU_DEAD:
@@ -4725,7 +4714,7 @@ static raid5_conf_t *setup_conf(mddev_t *mddev)
 	if (mddev->new_level != 5
 	    && mddev->new_level != 4
 	    && mddev->new_level != 6) {
-		printk(KERN_ERR "raid5: %s: raid level not set to 4/5/6 (%d)\n",
+		printk(KERN_ERR "md/raid:%s: raid level not set to 4/5/6 (%d)\n",
 		       mdname(mddev), mddev->new_level);
 		return ERR_PTR(-EIO);
 	}
@@ -4733,12 +4722,12 @@ static raid5_conf_t *setup_conf(mddev_t *mddev)
 	     && !algorithm_valid_raid5(mddev->new_layout)) ||
 	    (mddev->new_level == 6
 	     && !algorithm_valid_raid6(mddev->new_layout))) {
-		printk(KERN_ERR "raid5: %s: layout %d not supported\n",
+		printk(KERN_ERR "md/raid:%s: layout %d not supported\n",
 		       mdname(mddev), mddev->new_layout);
 		return ERR_PTR(-EIO);
 	}
 	if (mddev->new_level == 6 && mddev->raid_disks < 4) {
-		printk(KERN_ERR "raid6: not enough configured devices for %s (%d, minimum 4)\n",
+		printk(KERN_ERR "md/raid:%s: not enough configured devices (%d, minimum 4)\n",
 		       mdname(mddev), mddev->raid_disks);
 		return ERR_PTR(-EINVAL);
 	}
@@ -4746,8 +4735,8 @@ static raid5_conf_t *setup_conf(mddev_t *mddev)
 	if (!mddev->new_chunk_sectors ||
 	    (mddev->new_chunk_sectors << 9) % PAGE_SIZE ||
 	    !is_power_of_2(mddev->new_chunk_sectors)) {
-		printk(KERN_ERR "raid5: invalid chunk size %d for %s\n",
-		       mddev->new_chunk_sectors << 9, mdname(mddev));
+		printk(KERN_ERR "md/raid:%s: invalid chunk size %d\n",
+		       mdname(mddev), mddev->new_chunk_sectors << 9);
 		return ERR_PTR(-EINVAL);
 	}
 
@@ -4789,7 +4778,7 @@ static raid5_conf_t *setup_conf(mddev_t *mddev)
 	if (raid5_alloc_percpu(conf) != 0)
 		goto abort;
 
-	pr_debug("raid5: run(%s) called.\n", mdname(mddev));
+	pr_debug("raid456: run(%s) called.\n", mdname(mddev));
 
 	list_for_each_entry(rdev, &mddev->disks, same_set) {
 		raid_disk = rdev->raid_disk;
@@ -4802,9 +4791,9 @@ static raid5_conf_t *setup_conf(mddev_t *mddev)
 
 		if (test_bit(In_sync, &rdev->flags)) {
 			char b[BDEVNAME_SIZE];
-			printk(KERN_INFO "raid5: device %s operational as raid"
-				" disk %d\n", bdevname(rdev->bdev,b),
-				raid_disk);
+			printk(KERN_INFO "md/raid:%s: device %s operational as raid"
+			       " disk %d\n",
+			       mdname(mddev), bdevname(rdev->bdev, b), raid_disk);
 		} else
 			/* Cannot rely on bitmap to complete recovery */
 			conf->fullsync = 1;
@@ -4828,16 +4817,17 @@ static raid5_conf_t *setup_conf(mddev_t *mddev)
 		 max_disks * ((sizeof(struct bio) + PAGE_SIZE))) / 1024;
 	if (grow_stripes(conf, conf->max_nr_stripes)) {
 		printk(KERN_ERR
-			"raid5: couldn't allocate %dkB for buffers\n", memory);
+		       "md/raid:%s: couldn't allocate %dkB for buffers\n",
+		       mdname(mddev), memory);
 		goto abort;
 	} else
-		printk(KERN_INFO "raid5: allocated %dkB for %s\n",
-			memory, mdname(mddev));
+		printk(KERN_INFO "md/raid:%s: allocated %dkB\n",
+		       mdname(mddev), memory);
 
 	conf->thread = md_register_thread(raid5d, mddev, NULL);
 	if (!conf->thread) {
 		printk(KERN_ERR
-		       "raid5: couldn't allocate thread for %s\n",
+		       "md/raid:%s: couldn't allocate thread.\n",
 		       mdname(mddev));
 		goto abort;
 	}
@@ -4888,7 +4878,7 @@ static int run(mddev_t *mddev)
 	sector_t reshape_offset = 0;
 
 	if (mddev->recovery_cp != MaxSector)
-		printk(KERN_NOTICE "raid5: %s is not clean"
+		printk(KERN_NOTICE "md/raid:%s: not clean"
 		       " -- starting background reconstruction\n",
 		       mdname(mddev));
 	if (mddev->reshape_position != MaxSector) {
@@ -4902,7 +4892,7 @@ static int run(mddev_t *mddev)
 		int max_degraded = (mddev->level == 6 ? 2 : 1);
 
 		if (mddev->new_level != mddev->level) {
-			printk(KERN_ERR "raid5: %s: unsupported reshape "
+			printk(KERN_ERR "md/raid:%s: unsupported reshape "
 			       "required - aborting.\n",
 			       mdname(mddev));
 			return -EINVAL;
@@ -4915,8 +4905,8 @@ static int run(mddev_t *mddev)
 		here_new = mddev->reshape_position;
 		if (sector_div(here_new, mddev->new_chunk_sectors *
 			       (mddev->raid_disks - max_degraded))) {
-			printk(KERN_ERR "raid5: reshape_position not "
-			       "on a stripe boundary\n");
+			printk(KERN_ERR "md/raid:%s: reshape_position not "
+			       "on a stripe boundary\n", mdname(mddev));
 			return -EINVAL;
 		}
 		reshape_offset = here_new * mddev->new_chunk_sectors;
@@ -4937,8 +4927,9 @@ static int run(mddev_t *mddev)
 			if ((here_new * mddev->new_chunk_sectors != 
 			     here_old * mddev->chunk_sectors) ||
 			    mddev->ro == 0) {
-				printk(KERN_ERR "raid5: in-place reshape must be started"
-				       " in read-only mode - aborting\n");
+				printk(KERN_ERR "md/raid:%s: in-place reshape must be started"
+				       " in read-only mode - aborting\n",
+				       mdname(mddev));
 				return -EINVAL;
 			}
 		} else if (mddev->delta_disks < 0
@@ -4947,11 +4938,13 @@ static int run(mddev_t *mddev)
 		    : (here_new * mddev->new_chunk_sectors >=
 		       here_old * mddev->chunk_sectors)) {
 			/* Reading from the same stripe as writing to - bad */
-			printk(KERN_ERR "raid5: reshape_position too early for "
-			       "auto-recovery - aborting.\n");
+			printk(KERN_ERR "md/raid:%s: reshape_position too early for "
+			       "auto-recovery - aborting.\n",
+			       mdname(mddev));
 			return -EINVAL;
 		}
-		printk(KERN_INFO "raid5: reshape will continue\n");
+		printk(KERN_INFO "md/raid:%s: reshape will continue\n",
+		       mdname(mddev));
 		/* OK, we should be able to continue; */
 	} else {
 		BUG_ON(mddev->level != mddev->new_level);
@@ -4993,18 +4986,6 @@ static int run(mddev_t *mddev)
 		    mddev->minor_version > 90)
 			rdev->recovery_offset = reshape_offset;
 			
-		printk("%d: w=%d pa=%d pr=%d m=%d a=%d r=%d op1=%d op2=%d\n",
-		       rdev->raid_disk, working_disks, conf->prev_algo,
-		       conf->previous_raid_disks, conf->max_degraded,
-		       conf->algorithm, conf->raid_disks, 
-		       only_parity(rdev->raid_disk,
-				   conf->prev_algo,
-				   conf->previous_raid_disks,
-				   conf->max_degraded),
-		       only_parity(rdev->raid_disk,
-				   conf->algorithm,
-				   conf->raid_disks,
-				   conf->max_degraded));
 		if (rdev->recovery_offset < reshape_offset) {
 			/* We need to check old and new layout */
 			if (!only_parity(rdev->raid_disk,
@@ -5025,7 +5006,7 @@ static int run(mddev_t *mddev)
 			   - working_disks);
 
 	if (mddev->degraded > conf->max_degraded) {
-		printk(KERN_ERR "raid5: not enough operational devices for %s"
+		printk(KERN_ERR "md/raid:%s: not enough operational devices"
 			" (%d/%d failed)\n",
 			mdname(mddev), mddev->degraded, conf->raid_disks);
 		goto abort;
@@ -5039,32 +5020,32 @@ static int run(mddev_t *mddev)
 	    mddev->recovery_cp != MaxSector) {
 		if (mddev->ok_start_degraded)
 			printk(KERN_WARNING
-			       "raid5: starting dirty degraded array: %s"
-			       "- data corruption possible.\n",
+			       "md/raid:%s: starting dirty degraded array"
+			       " - data corruption possible.\n",
 			       mdname(mddev));
 		else {
 			printk(KERN_ERR
-			       "raid5: cannot start dirty degraded array for %s\n",
+			       "md/raid:%s: cannot start dirty degraded array.\n",
 			       mdname(mddev));
 			goto abort;
 		}
 	}
 
 	if (mddev->degraded == 0)
-		printk("raid5: raid level %d set %s active with %d out of %d"
-		       " devices, algorithm %d\n", conf->level, mdname(mddev),
+		printk(KERN_INFO "md/raid:%s: raid level %d active with %d out of %d"
+		       " devices, algorithm %d\n", mdname(mddev), conf->level,
 		       mddev->raid_disks-mddev->degraded, mddev->raid_disks,
 		       mddev->new_layout);
 	else
-		printk(KERN_ALERT "raid5: raid level %d set %s active with %d"
-			" out of %d devices, algorithm %d\n", conf->level,
-			mdname(mddev), mddev->raid_disks - mddev->degraded,
-			mddev->raid_disks, mddev->new_layout);
+		printk(KERN_ALERT "md/raid:%s: raid level %d active with %d"
+		       " out of %d devices, algorithm %d\n",
+		       mdname(mddev), conf->level,
+		       mddev->raid_disks - mddev->degraded,
+		       mddev->raid_disks, mddev->new_layout);
 
 	print_raid5_conf(conf);
 
 	if (conf->reshape_progress != MaxSector) {
-		printk("...ok start reshape thread\n");
 		conf->reshape_safe = conf->reshape_progress;
 		atomic_set(&conf->reshape_stripes, 0);
 		clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
@@ -5087,9 +5068,11 @@ static int run(mddev_t *mddev)
 	}
 
 	/* Ok, everything is just fine now */
-	if (sysfs_create_group(&mddev->kobj, &raid5_attrs_group))
+	if (mddev->to_remove == &raid5_attrs_group)
+		mddev->to_remove = NULL;
+	else if (sysfs_create_group(&mddev->kobj, &raid5_attrs_group))
 		printk(KERN_WARNING
-		       "raid5: failed to create sysfs attributes for %s\n",
+		       "md/raid:%s: failed to create sysfs attributes.\n",
 		       mdname(mddev));
 
 	mddev->queue->queue_lock = &conf->device_lock;
@@ -5119,22 +5102,21 @@ abort:
 		free_conf(conf);
 	}
 	mddev->private = NULL;
-	printk(KERN_ALERT "raid5: failed to run raid set %s\n", mdname(mddev));
+	printk(KERN_ALERT "md/raid:%s: failed to run raid set.\n", mdname(mddev));
 	return -EIO;
 }
 
-
-
 static int stop(mddev_t *mddev)
 {
-	raid5_conf_t *conf = (raid5_conf_t *) mddev->private;
+	raid5_conf_t *conf = mddev->private;
 
 	md_unregister_thread(mddev->thread);
 	mddev->thread = NULL;
 	mddev->queue->backing_dev_info.congested_fn = NULL;
 	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
 	free_conf(conf);
-	mddev->private = &raid5_attrs_group;
+	mddev->private = NULL;
+	mddev->to_remove = &raid5_attrs_group;
 	return 0;
 }
 
@@ -5175,7 +5157,7 @@ static void printall(struct seq_file *seq, raid5_conf_t *conf)
 
 static void status(struct seq_file *seq, mddev_t *mddev)
 {
-	raid5_conf_t *conf = (raid5_conf_t *) mddev->private;
+	raid5_conf_t *conf = mddev->private;
 	int i;
 
 	seq_printf(seq, " level %d, %dk chunk, algorithm %d", mddev->level,
@@ -5197,21 +5179,22 @@ static void print_raid5_conf (raid5_conf_t *conf)
 	int i;
 	struct disk_info *tmp;
 
-	printk("RAID5 conf printout:\n");
+	printk(KERN_DEBUG "RAID conf printout:\n");
 	if (!conf) {
 		printk("(conf==NULL)\n");
 		return;
 	}
-	printk(" --- rd:%d wd:%d\n", conf->raid_disks,
-		 conf->raid_disks - conf->mddev->degraded);
+	printk(KERN_DEBUG " --- level:%d rd:%d wd:%d\n", conf->level,
+	       conf->raid_disks,
+	       conf->raid_disks - conf->mddev->degraded);
 
 	for (i = 0; i < conf->raid_disks; i++) {
 		char b[BDEVNAME_SIZE];
 		tmp = conf->disks + i;
 		if (tmp->rdev)
-		printk(" disk %d, o:%d, dev:%s\n",
-			i, !test_bit(Faulty, &tmp->rdev->flags),
-			bdevname(tmp->rdev->bdev,b));
+			printk(KERN_DEBUG " disk %d, o:%d, dev:%s\n",
+			       i, !test_bit(Faulty, &tmp->rdev->flags),
+			       bdevname(tmp->rdev->bdev, b));
 	}
 }
 
@@ -5334,7 +5317,6 @@ static int raid5_resize(mddev_t *mddev, sector_t sectors)
 	    raid5_size(mddev, sectors, mddev->raid_disks))
 		return -EINVAL;
 	set_capacity(mddev->gendisk, mddev->array_sectors);
-	mddev->changed = 1;
 	revalidate_disk(mddev->gendisk);
 	if (sectors > mddev->dev_sectors && mddev->recovery_cp == MaxSector) {
 		mddev->recovery_cp = mddev->dev_sectors;
@@ -5360,7 +5342,8 @@ static int check_stripe_cache(mddev_t *mddev)
 	    > conf->max_nr_stripes ||
 	    ((mddev->new_chunk_sectors << 9) / STRIPE_SIZE) * 4
 	    > conf->max_nr_stripes) {
-		printk(KERN_WARNING "raid5: reshape: not enough stripes.  Needed %lu\n",
+		printk(KERN_WARNING "md/raid:%s: reshape: not enough stripes.  Needed %lu\n",
+		       mdname(mddev),
 		       ((max(mddev->chunk_sectors, mddev->new_chunk_sectors) << 9)
 			/ STRIPE_SIZE)*4);
 		return 0;
@@ -5431,7 +5414,7 @@ static int raid5_start_reshape(mddev_t *mddev)
 	 */
 	if (raid5_size(mddev, 0, conf->raid_disks + mddev->delta_disks)
 	    < mddev->array_sectors) {
-		printk(KERN_ERR "md: %s: array size must be reduced "
+		printk(KERN_ERR "md/raid:%s: array size must be reduced "
 		       "before number of disks\n", mdname(mddev));
 		return -EINVAL;
 	}
@@ -5469,9 +5452,9 @@ static int raid5_start_reshape(mddev_t *mddev)
 				if (sysfs_create_link(&mddev->kobj,
 						      &rdev->kobj, nm))
 					printk(KERN_WARNING
-					       "raid5: failed to create "
-					       " link %s for %s\n",
-					       nm, mdname(mddev));
+					       "md/raid:%s: failed to create "
+					       " link %s\n",
+					       mdname(mddev), nm);
 			} else
 				break;
 		}
@@ -5548,7 +5531,6 @@ static void raid5_finish_reshape(mddev_t *mddev)
 		if (mddev->delta_disks > 0) {
 			md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
 			set_capacity(mddev->gendisk, mddev->array_sectors);
-			mddev->changed = 1;
 			revalidate_disk(mddev->gendisk);
 		} else {
 			int d;
@@ -5613,6 +5595,29 @@ static void raid5_quiesce(mddev_t *mddev, int state)
 }
 
 
+static void *raid45_takeover_raid0(mddev_t *mddev, int level)
+{
+	struct raid0_private_data *raid0_priv = mddev->private;
+
+	/* for raid0 takeover only one zone is supported */
+	if (raid0_priv->nr_strip_zones > 1) {
+		printk(KERN_ERR "md/raid:%s: cannot takeover raid0 with more than one zone.\n",
+		       mdname(mddev));
+		return ERR_PTR(-EINVAL);
+	}
+
+	mddev->new_level = level;
+	mddev->new_layout = ALGORITHM_PARITY_N;
+	mddev->new_chunk_sectors = mddev->chunk_sectors;
+	mddev->raid_disks += 1;
+	mddev->delta_disks = 1;
+	/* make sure it will be not marked as dirty */
+	mddev->recovery_cp = MaxSector;
+
+	return setup_conf(mddev);
+}
+
+
 static void *raid5_takeover_raid1(mddev_t *mddev)
 {
 	int chunksect;
@@ -5737,12 +5742,13 @@ static int raid6_check_reshape(mddev_t *mddev)
 static void *raid5_takeover(mddev_t *mddev)
 {
 	/* raid5 can take over:
-	 *  raid0 - if all devices are the same - make it a raid4 layout
+	 *  raid0 - if there is only one strip zone - make it a raid4 layout
 	 *  raid1 - if there are two drives.  We need to know the chunk size
 	 *  raid4 - trivial - just use a raid4 layout.
 	 *  raid6 - Providing it is a *_6 layout
 	 */
-
+	if (mddev->level == 0)
+		return raid45_takeover_raid0(mddev, 5);
 	if (mddev->level == 1)
 		return raid5_takeover_raid1(mddev);
 	if (mddev->level == 4) {
@@ -5756,6 +5762,22 @@ static void *raid5_takeover(mddev_t *mddev)
 	return ERR_PTR(-EINVAL);
 }
 
+static void *raid4_takeover(mddev_t *mddev)
+{
+	/* raid4 can take over:
+	 *  raid0 - if there is only one strip zone
+	 *  raid5 - if layout is right
+	 */
+	if (mddev->level == 0)
+		return raid45_takeover_raid0(mddev, 4);
+	if (mddev->level == 5 &&
+	    mddev->layout == ALGORITHM_PARITY_N) {
+		mddev->new_layout = 0;
+		mddev->new_level = 4;
+		return setup_conf(mddev);
+	}
+	return ERR_PTR(-EINVAL);
+}
 
 static struct mdk_personality raid5_personality;
 
@@ -5871,6 +5893,7 @@ static struct mdk_personality raid4_personality =
 	.start_reshape  = raid5_start_reshape,
 	.finish_reshape = raid5_finish_reshape,
 	.quiesce	= raid5_quiesce,
+	.takeover	= raid4_takeover,
 };
 
 static int __init raid5_init(void)
diff --git a/drivers/media/IR/Kconfig b/drivers/media/IR/Kconfig
index 195c6cf359f6..d22a8ec523fc 100644
--- a/drivers/media/IR/Kconfig
+++ b/drivers/media/IR/Kconfig
@@ -13,6 +13,7 @@ source "drivers/media/IR/keymaps/Kconfig"
 config IR_NEC_DECODER
 	tristate "Enable IR raw decoder for the NEC protocol"
 	depends on IR_CORE
+	select BITREVERSE
 	default y
 
 	---help---
@@ -22,6 +23,7 @@ config IR_NEC_DECODER
 config IR_RC5_DECODER
 	tristate "Enable IR raw decoder for the RC-5 protocol"
 	depends on IR_CORE
+	select BITREVERSE
 	default y
 
 	---help---
diff --git a/drivers/media/IR/imon.c b/drivers/media/IR/imon.c
index 5e2045670004..4bbd45f4284c 100644
--- a/drivers/media/IR/imon.c
+++ b/drivers/media/IR/imon.c
@@ -94,6 +94,7 @@ struct imon_context {
 
 	bool display_supported;		/* not all controllers do */
 	bool display_isopen;		/* display port has been opened */
+	bool rf_device;			/* true if iMON 2.4G LT/DT RF device */
 	bool rf_isassociating;		/* RF remote associating */
 	bool dev_present_intf0;		/* USB device presence, interface 0 */
 	bool dev_present_intf1;		/* USB device presence, interface 1 */
@@ -385,7 +386,7 @@ static int display_open(struct inode *inode, struct file *file)
 		err("%s: display port is already open", __func__);
 		retval = -EBUSY;
 	} else {
-		ictx->display_isopen = 1;
+		ictx->display_isopen = true;
 		file->private_data = ictx;
 		dev_dbg(ictx->dev, "display port opened\n");
 	}
@@ -422,7 +423,7 @@ static int display_close(struct inode *inode, struct file *file)
 		err("%s: display is not open", __func__);
 		retval = -EIO;
 	} else {
-		ictx->display_isopen = 0;
+		ictx->display_isopen = false;
 		dev_dbg(ictx->dev, "display port closed\n");
 		if (!ictx->dev_present_intf0) {
 			/*
@@ -491,12 +492,12 @@ static int send_packet(struct imon_context *ictx)
 	}
 
 	init_completion(&ictx->tx.finished);
-	ictx->tx.busy = 1;
+	ictx->tx.busy = true;
 	smp_rmb(); /* ensure later readers know we're busy */
 
 	retval = usb_submit_urb(ictx->tx_urb, GFP_KERNEL);
 	if (retval) {
-		ictx->tx.busy = 0;
+		ictx->tx.busy = false;
 		smp_rmb(); /* ensure later readers know we're not busy */
 		err("%s: error submitting urb(%d)", __func__, retval);
 	} else {
@@ -682,7 +683,7 @@ static ssize_t store_associate_remote(struct device *d,
 		return -ENODEV;
 
 	mutex_lock(&ictx->lock);
-	ictx->rf_isassociating = 1;
+	ictx->rf_isassociating = true;
 	send_associate_24g(ictx);
 	mutex_unlock(&ictx->lock);
 
@@ -950,7 +951,7 @@ static void usb_tx_callback(struct urb *urb)
 	ictx->tx.status = urb->status;
 
 	/* notify waiters that write has finished */
-	ictx->tx.busy = 0;
+	ictx->tx.busy = false;
 	smp_rmb(); /* ensure later readers know we're not busy */
 	complete(&ictx->tx.finished);
 }
@@ -1215,7 +1216,7 @@ static bool imon_mouse_event(struct imon_context *ictx,
 {
 	char rel_x = 0x00, rel_y = 0x00;
 	u8 right_shift = 1;
-	bool mouse_input = 1;
+	bool mouse_input = true;
 	int dir = 0;
 
 	/* newer iMON device PAD or mouse button */
@@ -1246,7 +1247,7 @@ static bool imon_mouse_event(struct imon_context *ictx,
 	} else if (ictx->kc == KEY_CHANNELDOWN && (buf[2] & 0x40) != 0x40) {
 		dir = -1;
 	} else
-		mouse_input = 0;
+		mouse_input = false;
 
 	if (mouse_input) {
 		dev_dbg(ictx->dev, "sending mouse data via input subsystem\n");
@@ -1450,7 +1451,7 @@ static void imon_incoming_packet(struct imon_context *ictx,
 	unsigned char *buf = urb->transfer_buffer;
 	struct device *dev = ictx->dev;
 	u32 kc;
-	bool norelease = 0;
+	bool norelease = false;
 	int i;
 	u64 temp_key;
 	u64 panel_key = 0;
@@ -1465,7 +1466,7 @@ static void imon_incoming_packet(struct imon_context *ictx,
 	idev = ictx->idev;
 
 	/* filter out junk data on the older 0xffdc imon devices */
-	if ((buf[0] == 0xff) && (buf[7] == 0xff))
+	if ((buf[0] == 0xff) && (buf[1] == 0xff) && (buf[2] == 0xff))
 		return;
 
 	/* Figure out what key was pressed */
@@ -1517,7 +1518,7 @@ static void imon_incoming_packet(struct imon_context *ictx,
 	     !(buf[1] & 0x1 || buf[1] >> 2 & 0x1))) {
 		len = 8;
 		imon_pad_to_keys(ictx, buf);
-		norelease = 1;
+		norelease = true;
 	}
 
 	if (debug) {
@@ -1580,7 +1581,7 @@ not_input_data:
 	    (buf[6] == 0x5E && buf[7] == 0xDF))) {	/* DT */
 		dev_warn(dev, "%s: remote associated refid=%02X\n",
 			 __func__, buf[1]);
-		ictx->rf_isassociating = 0;
+		ictx->rf_isassociating = false;
 	}
 }
 
@@ -1790,9 +1791,9 @@ static bool imon_find_endpoints(struct imon_context *ictx,
 	int ifnum = iface_desc->desc.bInterfaceNumber;
 	int num_endpts = iface_desc->desc.bNumEndpoints;
 	int i, ep_dir, ep_type;
-	bool ir_ep_found = 0;
-	bool display_ep_found = 0;
-	bool tx_control = 0;
+	bool ir_ep_found = false;
+	bool display_ep_found = false;
+	bool tx_control = false;
 
 	/*
 	 * Scan the endpoint list and set:
@@ -1808,13 +1809,13 @@ static bool imon_find_endpoints(struct imon_context *ictx,
 		    ep_type == USB_ENDPOINT_XFER_INT) {
 
 			rx_endpoint = ep;
-			ir_ep_found = 1;
+			ir_ep_found = true;
 			dev_dbg(ictx->dev, "%s: found IR endpoint\n", __func__);
 
 		} else if (!display_ep_found && ep_dir == USB_DIR_OUT &&
 			   ep_type == USB_ENDPOINT_XFER_INT) {
 			tx_endpoint = ep;
-			display_ep_found = 1;
+			display_ep_found = true;
 			dev_dbg(ictx->dev, "%s: found display endpoint\n", __func__);
 		}
 	}
@@ -1835,8 +1836,8 @@ static bool imon_find_endpoints(struct imon_context *ictx,
 	 * newer iMON devices that use control urb instead of interrupt
 	 */
 	if (!display_ep_found) {
-		tx_control = 1;
-		display_ep_found = 1;
+		tx_control = true;
+		display_ep_found = true;
 		dev_dbg(ictx->dev, "%s: device uses control endpoint, not "
 			"interface OUT endpoint\n", __func__);
 	}
@@ -1847,7 +1848,7 @@ static bool imon_find_endpoints(struct imon_context *ictx,
 	 * and without... :\
 	 */
 	if (ictx->display_type == IMON_DISPLAY_TYPE_NONE) {
-		display_ep_found = 0;
+		display_ep_found = false;
 		dev_dbg(ictx->dev, "%s: device has no display\n", __func__);
 	}
 
@@ -1856,7 +1857,7 @@ static bool imon_find_endpoints(struct imon_context *ictx,
 	 * that refers to e.g. /dev/lcd0 (a character device LCD or VFD).
 	 */
 	if (ictx->display_type == IMON_DISPLAY_TYPE_VGA) {
-		display_ep_found = 0;
+		display_ep_found = false;
 		dev_dbg(ictx->dev, "%s: iMON Touch device found\n", __func__);
 	}
 
@@ -1905,9 +1906,10 @@ static struct imon_context *imon_init_intf0(struct usb_interface *intf)
 
 	ictx->dev = dev;
 	ictx->usbdev_intf0 = usb_get_dev(interface_to_usbdev(intf));
-	ictx->dev_present_intf0 = 1;
+	ictx->dev_present_intf0 = true;
 	ictx->rx_urb_intf0 = rx_urb;
 	ictx->tx_urb = tx_urb;
+	ictx->rf_device = false;
 
 	ictx->vendor  = le16_to_cpu(ictx->usbdev_intf0->descriptor.idVendor);
 	ictx->product = le16_to_cpu(ictx->usbdev_intf0->descriptor.idProduct);
@@ -1979,7 +1981,7 @@ static struct imon_context *imon_init_intf1(struct usb_interface *intf,
 	}
 
 	ictx->usbdev_intf1 = usb_get_dev(interface_to_usbdev(intf));
-	ictx->dev_present_intf1 = 1;
+	ictx->dev_present_intf1 = true;
 	ictx->rx_urb_intf1 = rx_urb;
 
 	ret = -ENODEV;
@@ -2047,6 +2049,12 @@ static void imon_get_ffdc_type(struct imon_context *ictx)
 		dev_info(ictx->dev, "0xffdc iMON Knob, iMON IR");
 		ictx->display_supported = false;
 		break;
+	/* iMON 2.4G LT (usb stick), no display, iMON RF */
+	case 0x4e:
+		dev_info(ictx->dev, "0xffdc iMON 2.4G LT, iMON RF");
+		ictx->display_supported = false;
+		ictx->rf_device = true;
+		break;
 	/* iMON VFD, no IR (does have vol knob tho) */
 	case 0x35:
 		dev_info(ictx->dev, "0xffdc iMON VFD + knob, no IR");
@@ -2197,15 +2205,6 @@ static int __devinit imon_probe(struct usb_interface *interface,
 			goto fail;
 		}
 
-		if (product == 0xffdc) {
-			/* RF products *also* use 0xffdc... sigh... */
-			sysfs_err = sysfs_create_group(&interface->dev.kobj,
-						       &imon_rf_attribute_group);
-			if (sysfs_err)
-				err("%s: Could not create RF sysfs entries(%d)",
-				    __func__, sysfs_err);
-		}
-
 	} else {
 	/* this is the secondary interface on the device */
 		ictx = imon_init_intf1(interface, first_if_ctx);
@@ -2233,6 +2232,14 @@ static int __devinit imon_probe(struct usb_interface *interface,
 
 		imon_set_display_type(ictx, interface);
 
+		if (product == 0xffdc && ictx->rf_device) {
+			sysfs_err = sysfs_create_group(&interface->dev.kobj,
+						       &imon_rf_attribute_group);
+			if (sysfs_err)
+				err("%s: Could not create RF sysfs entries(%d)",
+				    __func__, sysfs_err);
+		}
+
 		if (ictx->display_supported)
 			imon_init_display(ictx, interface);
 	}
@@ -2297,7 +2304,7 @@ static void __devexit imon_disconnect(struct usb_interface *interface)
 	}
 
 	if (ifnum == 0) {
-		ictx->dev_present_intf0 = 0;
+		ictx->dev_present_intf0 = false;
 		usb_kill_urb(ictx->rx_urb_intf0);
 		input_unregister_device(ictx->idev);
 		if (ictx->display_supported) {
@@ -2307,7 +2314,7 @@ static void __devexit imon_disconnect(struct usb_interface *interface)
 				usb_deregister_dev(interface, &imon_vfd_class);
 		}
 	} else {
-		ictx->dev_present_intf1 = 0;
+		ictx->dev_present_intf1 = false;
 		usb_kill_urb(ictx->rx_urb_intf1);
 		if (ictx->display_type == IMON_DISPLAY_TYPE_VGA)
 			input_unregister_device(ictx->touch);
diff --git a/drivers/media/IR/ir-keytable.c b/drivers/media/IR/ir-keytable.c
index 9374a006f43d..94a8577e72eb 100644
--- a/drivers/media/IR/ir-keytable.c
+++ b/drivers/media/IR/ir-keytable.c
@@ -490,11 +490,12 @@ int __ir_input_register(struct input_dev *input_dev,
 	if (rc < 0)
 		goto out_table;
 
-	if (ir_dev->props->driver_type == RC_DRIVER_IR_RAW) {
-		rc = ir_raw_event_register(input_dev);
-		if (rc < 0)
-			goto out_event;
-	}
+	if (ir_dev->props)
+		if (ir_dev->props->driver_type == RC_DRIVER_IR_RAW) {
+			rc = ir_raw_event_register(input_dev);
+			if (rc < 0)
+				goto out_event;
+		}
 
 	IR_dprintk(1, "Registered input device on %s for %s remote.\n",
 		   driver_name, rc_tab->name);
@@ -530,8 +531,10 @@ void ir_input_unregister(struct input_dev *input_dev)
 	IR_dprintk(1, "Freed keycode table\n");
 
 	del_timer_sync(&ir_dev->timer_keyup);
-	if (ir_dev->props->driver_type == RC_DRIVER_IR_RAW)
-		ir_raw_event_unregister(input_dev);
+	if (ir_dev->props)
+		if (ir_dev->props->driver_type == RC_DRIVER_IR_RAW)
+			ir_raw_event_unregister(input_dev);
+
 	rc_tab = &ir_dev->rc_tab;
 	rc_tab->size = 0;
 	kfree(rc_tab->scan);
diff --git a/drivers/media/IR/ir-sysfs.c b/drivers/media/IR/ir-sysfs.c
index d7da63e16c92..2098dd1488e0 100644
--- a/drivers/media/IR/ir-sysfs.c
+++ b/drivers/media/IR/ir-sysfs.c
@@ -221,9 +221,10 @@ int ir_register_class(struct input_dev *input_dev)
 	if (unlikely(devno < 0))
 		return devno;
 
-	if (ir_dev->props->driver_type == RC_DRIVER_SCANCODE)
-		ir_dev->dev.type = &rc_dev_type;
-	else
+	if (ir_dev->props) {
+		if (ir_dev->props->driver_type == RC_DRIVER_SCANCODE)
+			ir_dev->dev.type = &rc_dev_type;
+	} else
 		ir_dev->dev.type = &ir_raw_dev_type;
 
 	ir_dev->dev.class = &ir_input_class;
diff --git a/drivers/media/IR/keymaps/Makefile b/drivers/media/IR/keymaps/Makefile
index ec25258a955f..aea649fbcf5a 100644
--- a/drivers/media/IR/keymaps/Makefile
+++ b/drivers/media/IR/keymaps/Makefile
@@ -6,7 +6,8 @@ obj-$(CONFIG_RC_MAP) += rc-adstech-dvb-t-pci.o \
 			rc-avermedia.o \
 			rc-avermedia-cardbus.o \
 			rc-avermedia-dvbt.o \
-			rc-avermedia-m135a-rm-jx.o \
+			rc-avermedia-m135a.o \
+			rc-avermedia-m733a-rm-k6.o \
 			rc-avertv-303.o \
 			rc-behold.o \
 			rc-behold-columbus.o \
diff --git a/drivers/media/IR/keymaps/rc-avermedia-m135a-rm-jx.c b/drivers/media/IR/keymaps/rc-avermedia-m135a-rm-jx.c
deleted file mode 100644
index 101e7ea85941..000000000000
--- a/drivers/media/IR/keymaps/rc-avermedia-m135a-rm-jx.c
+++ /dev/null
@@ -1,90 +0,0 @@
-/* avermedia-m135a-rm-jx.h - Keytable for avermedia_m135a_rm_jx Remote Controller
- *
- * keymap imported from ir-keymaps.c
- *
- * Copyright (c) 2010 by Mauro Carvalho Chehab <mchehab@redhat.com>
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License as published by
- * the Free Software Foundation; either version 2 of the License, or
- * (at your option) any later version.
- */
-
-#include <media/rc-map.h>
-
-/*
- * Avermedia M135A with IR model RM-JX
- * The same codes exist on both Positivo (BR) and original IR
- * Mauro Carvalho Chehab <mchehab@infradead.org>
- */
-
-static struct ir_scancode avermedia_m135a_rm_jx[] = {
-	{ 0x0200, KEY_POWER2 },
-	{ 0x022e, KEY_DOT },		/* '.' */
-	{ 0x0201, KEY_MODE },		/* TV/FM or SOURCE */
-
-	{ 0x0205, KEY_1 },
-	{ 0x0206, KEY_2 },
-	{ 0x0207, KEY_3 },
-	{ 0x0209, KEY_4 },
-	{ 0x020a, KEY_5 },
-	{ 0x020b, KEY_6 },
-	{ 0x020d, KEY_7 },
-	{ 0x020e, KEY_8 },
-	{ 0x020f, KEY_9 },
-	{ 0x0211, KEY_0 },
-
-	{ 0x0213, KEY_RIGHT },		/* -> or L */
-	{ 0x0212, KEY_LEFT },		/* <- or R */
-
-	{ 0x0217, KEY_SLEEP },		/* Capturar Imagem or Snapshot */
-	{ 0x0210, KEY_SHUFFLE },	/* Amostra or 16 chan prev */
-
-	{ 0x0303, KEY_CHANNELUP },
-	{ 0x0302, KEY_CHANNELDOWN },
-	{ 0x021f, KEY_VOLUMEUP },
-	{ 0x021e, KEY_VOLUMEDOWN },
-	{ 0x020c, KEY_ENTER },		/* Full Screen */
-
-	{ 0x0214, KEY_MUTE },
-	{ 0x0208, KEY_AUDIO },
-
-	{ 0x0203, KEY_TEXT },		/* Teletext */
-	{ 0x0204, KEY_EPG },
-	{ 0x022b, KEY_TV2 },		/* TV2 or PIP */
-
-	{ 0x021d, KEY_RED },
-	{ 0x021c, KEY_YELLOW },
-	{ 0x0301, KEY_GREEN },
-	{ 0x0300, KEY_BLUE },
-
-	{ 0x021a, KEY_PLAYPAUSE },
-	{ 0x0219, KEY_RECORD },
-	{ 0x0218, KEY_PLAY },
-	{ 0x021b, KEY_STOP },
-};
-
-static struct rc_keymap avermedia_m135a_rm_jx_map = {
-	.map = {
-		.scan    = avermedia_m135a_rm_jx,
-		.size    = ARRAY_SIZE(avermedia_m135a_rm_jx),
-		.ir_type = IR_TYPE_NEC,
-		.name    = RC_MAP_AVERMEDIA_M135A_RM_JX,
-	}
-};
-
-static int __init init_rc_map_avermedia_m135a_rm_jx(void)
-{
-	return ir_register_map(&avermedia_m135a_rm_jx_map);
-}
-
-static void __exit exit_rc_map_avermedia_m135a_rm_jx(void)
-{
-	ir_unregister_map(&avermedia_m135a_rm_jx_map);
-}
-
-module_init(init_rc_map_avermedia_m135a_rm_jx)
-module_exit(exit_rc_map_avermedia_m135a_rm_jx)
-
-MODULE_LICENSE("GPL");
-MODULE_AUTHOR("Mauro Carvalho Chehab <mchehab@redhat.com>");
diff --git a/drivers/media/IR/keymaps/rc-avermedia-m135a.c b/drivers/media/IR/keymaps/rc-avermedia-m135a.c
new file mode 100644
index 000000000000..e4471fb2ad1e
--- /dev/null
+++ b/drivers/media/IR/keymaps/rc-avermedia-m135a.c
@@ -0,0 +1,147 @@
+/* avermedia-m135a.c - Keytable for Avermedia M135A Remote Controllers
+ *
+ * Copyright (c) 2010 by Mauro Carvalho Chehab <mchehab@redhat.com>
+ * Copyright (c) 2010 by Herton Ronaldo Krzesinski <herton@mandriva.com.br>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ */
+
+#include <media/rc-map.h>
+
+/*
+ * Avermedia M135A with RM-JX and RM-K6 remote controls
+ *
+ * On Avermedia M135A with IR model RM-JX, the same codes exist on both
+ * Positivo (BR) and original IR, initial version and remote control codes
+ * added by Mauro Carvalho Chehab <mchehab@infradead.org>
+ *
+ * Positivo also ships Avermedia M135A with model RM-K6, extra control
+ * codes added by Herton Ronaldo Krzesinski <herton@mandriva.com.br>
+ */
+
+static struct ir_scancode avermedia_m135a[] = {
+	/* RM-JX */
+	{ 0x0200, KEY_POWER2 },
+	{ 0x022e, KEY_DOT },		/* '.' */
+	{ 0x0201, KEY_MODE },		/* TV/FM or SOURCE */
+
+	{ 0x0205, KEY_1 },
+	{ 0x0206, KEY_2 },
+	{ 0x0207, KEY_3 },
+	{ 0x0209, KEY_4 },
+	{ 0x020a, KEY_5 },
+	{ 0x020b, KEY_6 },
+	{ 0x020d, KEY_7 },
+	{ 0x020e, KEY_8 },
+	{ 0x020f, KEY_9 },
+	{ 0x0211, KEY_0 },
+
+	{ 0x0213, KEY_RIGHT },		/* -> or L */
+	{ 0x0212, KEY_LEFT },		/* <- or R */
+
+	{ 0x0217, KEY_SLEEP },		/* Capturar Imagem or Snapshot */
+	{ 0x0210, KEY_SHUFFLE },	/* Amostra or 16 chan prev */
+
+	{ 0x0303, KEY_CHANNELUP },
+	{ 0x0302, KEY_CHANNELDOWN },
+	{ 0x021f, KEY_VOLUMEUP },
+	{ 0x021e, KEY_VOLUMEDOWN },
+	{ 0x020c, KEY_ENTER },		/* Full Screen */
+
+	{ 0x0214, KEY_MUTE },
+	{ 0x0208, KEY_AUDIO },
+
+	{ 0x0203, KEY_TEXT },		/* Teletext */
+	{ 0x0204, KEY_EPG },
+	{ 0x022b, KEY_TV2 },		/* TV2 or PIP */
+
+	{ 0x021d, KEY_RED },
+	{ 0x021c, KEY_YELLOW },
+	{ 0x0301, KEY_GREEN },
+	{ 0x0300, KEY_BLUE },
+
+	{ 0x021a, KEY_PLAYPAUSE },
+	{ 0x0219, KEY_RECORD },
+	{ 0x0218, KEY_PLAY },
+	{ 0x021b, KEY_STOP },
+
+	/* RM-K6 */
+	{ 0x0401, KEY_POWER2 },
+	{ 0x0406, KEY_MUTE },
+	{ 0x0408, KEY_MODE },     /* TV/FM */
+
+	{ 0x0409, KEY_1 },
+	{ 0x040a, KEY_2 },
+	{ 0x040b, KEY_3 },
+	{ 0x040c, KEY_4 },
+	{ 0x040d, KEY_5 },
+	{ 0x040e, KEY_6 },
+	{ 0x040f, KEY_7 },
+	{ 0x0410, KEY_8 },
+	{ 0x0411, KEY_9 },
+	{ 0x044c, KEY_DOT },      /* '.' */
+	{ 0x0412, KEY_0 },
+	{ 0x0407, KEY_REFRESH },  /* Refresh/Reload */
+
+	{ 0x0413, KEY_AUDIO },
+	{ 0x0440, KEY_SCREEN },   /* Full Screen toggle */
+	{ 0x0441, KEY_HOME },
+	{ 0x0442, KEY_BACK },
+	{ 0x0447, KEY_UP },
+	{ 0x0448, KEY_DOWN },
+	{ 0x0449, KEY_LEFT },
+	{ 0x044a, KEY_RIGHT },
+	{ 0x044b, KEY_OK },
+	{ 0x0404, KEY_VOLUMEUP },
+	{ 0x0405, KEY_VOLUMEDOWN },
+	{ 0x0402, KEY_CHANNELUP },
+	{ 0x0403, KEY_CHANNELDOWN },
+
+	{ 0x0443, KEY_RED },
+	{ 0x0444, KEY_GREEN },
+	{ 0x0445, KEY_YELLOW },
+	{ 0x0446, KEY_BLUE },
+
+	{ 0x0414, KEY_TEXT },
+	{ 0x0415, KEY_EPG },
+	{ 0x041a, KEY_TV2 },      /* PIP */
+	{ 0x041b, KEY_MHP },      /* Snapshot */
+
+	{ 0x0417, KEY_RECORD },
+	{ 0x0416, KEY_PLAYPAUSE },
+	{ 0x0418, KEY_STOP },
+	{ 0x0419, KEY_PAUSE },
+
+	{ 0x041f, KEY_PREVIOUS },
+	{ 0x041c, KEY_REWIND },
+	{ 0x041d, KEY_FORWARD },
+	{ 0x041e, KEY_NEXT },
+};
+
+static struct rc_keymap avermedia_m135a_map = {
+	.map = {
+		.scan    = avermedia_m135a,
+		.size    = ARRAY_SIZE(avermedia_m135a),
+		.ir_type = IR_TYPE_NEC,
+		.name    = RC_MAP_AVERMEDIA_M135A,
+	}
+};
+
+static int __init init_rc_map_avermedia_m135a(void)
+{
+	return ir_register_map(&avermedia_m135a_map);
+}
+
+static void __exit exit_rc_map_avermedia_m135a(void)
+{
+	ir_unregister_map(&avermedia_m135a_map);
+}
+
+module_init(init_rc_map_avermedia_m135a)
+module_exit(exit_rc_map_avermedia_m135a)
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Mauro Carvalho Chehab <mchehab@redhat.com>");
diff --git a/drivers/media/IR/keymaps/rc-avermedia-m733a-rm-k6.c b/drivers/media/IR/keymaps/rc-avermedia-m733a-rm-k6.c
new file mode 100644
index 000000000000..cf8d45717cb3
--- /dev/null
+++ b/drivers/media/IR/keymaps/rc-avermedia-m733a-rm-k6.c
@@ -0,0 +1,95 @@
+/* avermedia-m733a-rm-k6.h - Keytable for avermedia_m733a_rm_k6 Remote Controller
+ *
+ * Copyright (c) 2010 by Herton Ronaldo Krzesinski <herton@mandriva.com.br>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ */
+
+#include <media/rc-map.h>
+
+/*
+ * Avermedia M733A with IR model RM-K6
+ * This is the stock remote controller used with Positivo machines with M733A
+ * Herton Ronaldo Krzesinski <herton@mandriva.com.br>
+ */
+
+static struct ir_scancode avermedia_m733a_rm_k6[] = {
+	{ 0x0401, KEY_POWER2 },
+	{ 0x0406, KEY_MUTE },
+	{ 0x0408, KEY_MODE },     /* TV/FM */
+
+	{ 0x0409, KEY_1 },
+	{ 0x040a, KEY_2 },
+	{ 0x040b, KEY_3 },
+	{ 0x040c, KEY_4 },
+	{ 0x040d, KEY_5 },
+	{ 0x040e, KEY_6 },
+	{ 0x040f, KEY_7 },
+	{ 0x0410, KEY_8 },
+	{ 0x0411, KEY_9 },
+	{ 0x044c, KEY_DOT },      /* '.' */
+	{ 0x0412, KEY_0 },
+	{ 0x0407, KEY_REFRESH },  /* Refresh/Reload */
+
+	{ 0x0413, KEY_AUDIO },
+	{ 0x0440, KEY_SCREEN },   /* Full Screen toggle */
+	{ 0x0441, KEY_HOME },
+	{ 0x0442, KEY_BACK },
+	{ 0x0447, KEY_UP },
+	{ 0x0448, KEY_DOWN },
+	{ 0x0449, KEY_LEFT },
+	{ 0x044a, KEY_RIGHT },
+	{ 0x044b, KEY_OK },
+	{ 0x0404, KEY_VOLUMEUP },
+	{ 0x0405, KEY_VOLUMEDOWN },
+	{ 0x0402, KEY_CHANNELUP },
+	{ 0x0403, KEY_CHANNELDOWN },
+
+	{ 0x0443, KEY_RED },
+	{ 0x0444, KEY_GREEN },
+	{ 0x0445, KEY_YELLOW },
+	{ 0x0446, KEY_BLUE },
+
+	{ 0x0414, KEY_TEXT },
+	{ 0x0415, KEY_EPG },
+	{ 0x041a, KEY_TV2 },      /* PIP */
+	{ 0x041b, KEY_MHP },      /* Snapshot */
+
+	{ 0x0417, KEY_RECORD },
+	{ 0x0416, KEY_PLAYPAUSE },
+	{ 0x0418, KEY_STOP },
+	{ 0x0419, KEY_PAUSE },
+
+	{ 0x041f, KEY_PREVIOUS },
+	{ 0x041c, KEY_REWIND },
+	{ 0x041d, KEY_FORWARD },
+	{ 0x041e, KEY_NEXT },
+};
+
+static struct rc_keymap avermedia_m733a_rm_k6_map = {
+	.map = {
+		.scan    = avermedia_m733a_rm_k6,
+		.size    = ARRAY_SIZE(avermedia_m733a_rm_k6),
+		.ir_type = IR_TYPE_NEC,
+		.name    = RC_MAP_AVERMEDIA_M733A_RM_K6,
+	}
+};
+
+static int __init init_rc_map_avermedia_m733a_rm_k6(void)
+{
+	return ir_register_map(&avermedia_m733a_rm_k6_map);
+}
+
+static void __exit exit_rc_map_avermedia_m733a_rm_k6(void)
+{
+	ir_unregister_map(&avermedia_m733a_rm_k6_map);
+}
+
+module_init(init_rc_map_avermedia_m733a_rm_k6)
+module_exit(exit_rc_map_avermedia_m733a_rm_k6)
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Mauro Carvalho Chehab <mchehab@redhat.com>");
diff --git a/drivers/media/dvb/dm1105/dm1105.c b/drivers/media/dvb/dm1105/dm1105.c
index b762e561a6d5..bca07c0bcd01 100644
--- a/drivers/media/dvb/dm1105/dm1105.c
+++ b/drivers/media/dvb/dm1105/dm1105.c
@@ -594,7 +594,7 @@ static irqreturn_t dm1105_irq(int irq, void *dev_id)
 int __devinit dm1105_ir_init(struct dm1105_dev *dm1105)
 {
 	struct input_dev *input_dev;
-	char *ir_codes = NULL;
+	char *ir_codes = RC_MAP_DM1105_NEC;
 	int err = -ENOMEM;
 
 	input_dev = input_allocate_device();
diff --git a/drivers/media/dvb/dvb-core/dmxdev.c b/drivers/media/dvb/dvb-core/dmxdev.c
index 9ddc57909d49..425862ffb285 100644
--- a/drivers/media/dvb/dvb-core/dmxdev.c
+++ b/drivers/media/dvb/dvb-core/dmxdev.c
@@ -25,6 +25,7 @@
 #include <linux/slab.h>
 #include <linux/vmalloc.h>
 #include <linux/module.h>
+#include <linux/smp_lock.h>
 #include <linux/poll.h>
 #include <linux/ioctl.h>
 #include <linux/wait.h>
@@ -963,7 +964,7 @@ dvb_demux_read(struct file *file, char __user *buf, size_t count,
 	return ret;
 }
 
-static int dvb_demux_do_ioctl(struct inode *inode, struct file *file,
+static int dvb_demux_do_ioctl(struct file *file,
 			      unsigned int cmd, void *parg)
 {
 	struct dmxdev_filter *dmxdevfilter = file->private_data;
@@ -1084,10 +1085,16 @@ static int dvb_demux_do_ioctl(struct inode *inode, struct file *file,
 	return ret;
 }
 
-static int dvb_demux_ioctl(struct inode *inode, struct file *file,
-			   unsigned int cmd, unsigned long arg)
+static long dvb_demux_ioctl(struct file *file, unsigned int cmd,
+			    unsigned long arg)
 {
-	return dvb_usercopy(inode, file, cmd, arg, dvb_demux_do_ioctl);
+	int ret;
+
+	lock_kernel();
+	ret = dvb_usercopy(file, cmd, arg, dvb_demux_do_ioctl);
+	unlock_kernel();
+
+	return ret;
 }
 
 static unsigned int dvb_demux_poll(struct file *file, poll_table *wait)
@@ -1139,7 +1146,7 @@ static int dvb_demux_release(struct inode *inode, struct file *file)
 static const struct file_operations dvb_demux_fops = {
 	.owner = THIS_MODULE,
 	.read = dvb_demux_read,
-	.ioctl = dvb_demux_ioctl,
+	.unlocked_ioctl = dvb_demux_ioctl,
 	.open = dvb_demux_open,
 	.release = dvb_demux_release,
 	.poll = dvb_demux_poll,
@@ -1152,7 +1159,7 @@ static struct dvb_device dvbdev_demux = {
 	.fops = &dvb_demux_fops
 };
 
-static int dvb_dvr_do_ioctl(struct inode *inode, struct file *file,
+static int dvb_dvr_do_ioctl(struct file *file,
 			    unsigned int cmd, void *parg)
 {
 	struct dvb_device *dvbdev = file->private_data;
@@ -1176,10 +1183,16 @@ static int dvb_dvr_do_ioctl(struct inode *inode, struct file *file,
 	return ret;
 }
 
-static int dvb_dvr_ioctl(struct inode *inode, struct file *file,
+static long dvb_dvr_ioctl(struct file *file,
 			 unsigned int cmd, unsigned long arg)
 {
-	return dvb_usercopy(inode, file, cmd, arg, dvb_dvr_do_ioctl);
+	int ret;
+
+	lock_kernel();
+	ret = dvb_usercopy(file, cmd, arg, dvb_dvr_do_ioctl);
+	unlock_kernel();
+
+	return ret;
 }
 
 static unsigned int dvb_dvr_poll(struct file *file, poll_table *wait)
@@ -1208,7 +1221,7 @@ static const struct file_operations dvb_dvr_fops = {
 	.owner = THIS_MODULE,
 	.read = dvb_dvr_read,
 	.write = dvb_dvr_write,
-	.ioctl = dvb_dvr_ioctl,
+	.unlocked_ioctl = dvb_dvr_ioctl,
 	.open = dvb_dvr_open,
 	.release = dvb_dvr_release,
 	.poll = dvb_dvr_poll,
diff --git a/drivers/media/dvb/dvb-core/dvb_ca_en50221.c b/drivers/media/dvb/dvb-core/dvb_ca_en50221.c
index cb22da53bfb0..ef259a0718ac 100644
--- a/drivers/media/dvb/dvb-core/dvb_ca_en50221.c
+++ b/drivers/media/dvb/dvb-core/dvb_ca_en50221.c
@@ -36,6 +36,7 @@
 #include <linux/delay.h>
 #include <linux/spinlock.h>
 #include <linux/sched.h>
+#include <linux/smp_lock.h>
 #include <linux/kthread.h>
 
 #include "dvb_ca_en50221.h"
@@ -1181,7 +1182,7 @@ static int dvb_ca_en50221_thread(void *data)
  *
  * @return 0 on success, <0 on error.
  */
-static int dvb_ca_en50221_io_do_ioctl(struct inode *inode, struct file *file,
+static int dvb_ca_en50221_io_do_ioctl(struct file *file,
 				      unsigned int cmd, void *parg)
 {
 	struct dvb_device *dvbdev = file->private_data;
@@ -1255,10 +1256,16 @@ static int dvb_ca_en50221_io_do_ioctl(struct inode *inode, struct file *file,
  *
  * @return 0 on success, <0 on error.
  */
-static int dvb_ca_en50221_io_ioctl(struct inode *inode, struct file *file,
-				   unsigned int cmd, unsigned long arg)
+static long dvb_ca_en50221_io_ioctl(struct file *file,
+				    unsigned int cmd, unsigned long arg)
 {
-	return dvb_usercopy(inode, file, cmd, arg, dvb_ca_en50221_io_do_ioctl);
+	int ret;
+
+	lock_kernel();
+	ret = dvb_usercopy(file, cmd, arg, dvb_ca_en50221_io_do_ioctl);
+	unlock_kernel();
+
+	return ret;
 }
 
 
@@ -1611,7 +1618,7 @@ static const struct file_operations dvb_ca_fops = {
 	.owner = THIS_MODULE,
 	.read = dvb_ca_en50221_io_read,
 	.write = dvb_ca_en50221_io_write,
-	.ioctl = dvb_ca_en50221_io_ioctl,
+	.unlocked_ioctl = dvb_ca_en50221_io_ioctl,
 	.open = dvb_ca_en50221_io_open,
 	.release = dvb_ca_en50221_io_release,
 	.poll = dvb_ca_en50221_io_poll,
diff --git a/drivers/media/dvb/dvb-core/dvb_frontend.c b/drivers/media/dvb/dvb-core/dvb_frontend.c
index 6932def4d266..44ae89ecef94 100644
--- a/drivers/media/dvb/dvb-core/dvb_frontend.c
+++ b/drivers/media/dvb/dvb-core/dvb_frontend.c
@@ -36,6 +36,7 @@
 #include <linux/list.h>
 #include <linux/freezer.h>
 #include <linux/jiffies.h>
+#include <linux/smp_lock.h>
 #include <linux/kthread.h>
 #include <asm/processor.h>
 
@@ -1195,14 +1196,14 @@ static void dtv_property_cache_submit(struct dvb_frontend *fe)
 	}
 }
 
-static int dvb_frontend_ioctl_legacy(struct inode *inode, struct file *file,
+static int dvb_frontend_ioctl_legacy(struct file *file,
 			unsigned int cmd, void *parg);
-static int dvb_frontend_ioctl_properties(struct inode *inode, struct file *file,
+static int dvb_frontend_ioctl_properties(struct file *file,
 			unsigned int cmd, void *parg);
 
 static int dtv_property_process_get(struct dvb_frontend *fe,
 				    struct dtv_property *tvp,
-				    struct inode *inode, struct file *file)
+				    struct file *file)
 {
 	int r = 0;
 
@@ -1335,7 +1336,6 @@ static int dtv_property_process_get(struct dvb_frontend *fe,
 
 static int dtv_property_process_set(struct dvb_frontend *fe,
 				    struct dtv_property *tvp,
-				    struct inode *inode,
 				    struct file *file)
 {
 	int r = 0;
@@ -1366,7 +1366,7 @@ static int dtv_property_process_set(struct dvb_frontend *fe,
 		dprintk("%s() Finalised property cache\n", __func__);
 		dtv_property_cache_submit(fe);
 
-		r |= dvb_frontend_ioctl_legacy(inode, file, FE_SET_FRONTEND,
+		r |= dvb_frontend_ioctl_legacy(file, FE_SET_FRONTEND,
 			&fepriv->parameters);
 		break;
 	case DTV_FREQUENCY:
@@ -1398,12 +1398,12 @@ static int dtv_property_process_set(struct dvb_frontend *fe,
 		break;
 	case DTV_VOLTAGE:
 		fe->dtv_property_cache.voltage = tvp->u.data;
-		r = dvb_frontend_ioctl_legacy(inode, file, FE_SET_VOLTAGE,
+		r = dvb_frontend_ioctl_legacy(file, FE_SET_VOLTAGE,
 			(void *)fe->dtv_property_cache.voltage);
 		break;
 	case DTV_TONE:
 		fe->dtv_property_cache.sectone = tvp->u.data;
-		r = dvb_frontend_ioctl_legacy(inode, file, FE_SET_TONE,
+		r = dvb_frontend_ioctl_legacy(file, FE_SET_TONE,
 			(void *)fe->dtv_property_cache.sectone);
 		break;
 	case DTV_CODE_RATE_HP:
@@ -1487,7 +1487,7 @@ static int dtv_property_process_set(struct dvb_frontend *fe,
 	return r;
 }
 
-static int dvb_frontend_ioctl(struct inode *inode, struct file *file,
+static int dvb_frontend_ioctl(struct file *file,
 			unsigned int cmd, void *parg)
 {
 	struct dvb_device *dvbdev = file->private_data;
@@ -1509,17 +1509,17 @@ static int dvb_frontend_ioctl(struct inode *inode, struct file *file,
 		return -ERESTARTSYS;
 
 	if ((cmd == FE_SET_PROPERTY) || (cmd == FE_GET_PROPERTY))
-		err = dvb_frontend_ioctl_properties(inode, file, cmd, parg);
+		err = dvb_frontend_ioctl_properties(file, cmd, parg);
 	else {
 		fe->dtv_property_cache.state = DTV_UNDEFINED;
-		err = dvb_frontend_ioctl_legacy(inode, file, cmd, parg);
+		err = dvb_frontend_ioctl_legacy(file, cmd, parg);
 	}
 
 	up(&fepriv->sem);
 	return err;
 }
 
-static int dvb_frontend_ioctl_properties(struct inode *inode, struct file *file,
+static int dvb_frontend_ioctl_properties(struct file *file,
 			unsigned int cmd, void *parg)
 {
 	struct dvb_device *dvbdev = file->private_data;
@@ -1555,7 +1555,7 @@ static int dvb_frontend_ioctl_properties(struct inode *inode, struct file *file,
 		}
 
 		for (i = 0; i < tvps->num; i++) {
-			(tvp + i)->result = dtv_property_process_set(fe, tvp + i, inode, file);
+			(tvp + i)->result = dtv_property_process_set(fe, tvp + i, file);
 			err |= (tvp + i)->result;
 		}
 
@@ -1587,7 +1587,7 @@ static int dvb_frontend_ioctl_properties(struct inode *inode, struct file *file,
 		}
 
 		for (i = 0; i < tvps->num; i++) {
-			(tvp + i)->result = dtv_property_process_get(fe, tvp + i, inode, file);
+			(tvp + i)->result = dtv_property_process_get(fe, tvp + i, file);
 			err |= (tvp + i)->result;
 		}
 
@@ -1604,7 +1604,7 @@ out:
 	return err;
 }
 
-static int dvb_frontend_ioctl_legacy(struct inode *inode, struct file *file,
+static int dvb_frontend_ioctl_legacy(struct file *file,
 			unsigned int cmd, void *parg)
 {
 	struct dvb_device *dvbdev = file->private_data;
@@ -2031,7 +2031,7 @@ static int dvb_frontend_release(struct inode *inode, struct file *file)
 
 static const struct file_operations dvb_frontend_fops = {
 	.owner		= THIS_MODULE,
-	.ioctl		= dvb_generic_ioctl,
+	.unlocked_ioctl	= dvb_generic_ioctl,
 	.poll		= dvb_frontend_poll,
 	.open		= dvb_frontend_open,
 	.release	= dvb_frontend_release
diff --git a/drivers/media/dvb/dvb-core/dvb_net.c b/drivers/media/dvb/dvb-core/dvb_net.c
index cccea412088b..6c3a8a06ccab 100644
--- a/drivers/media/dvb/dvb-core/dvb_net.c
+++ b/drivers/media/dvb/dvb-core/dvb_net.c
@@ -59,6 +59,7 @@
 #include <linux/netdevice.h>
 #include <linux/etherdevice.h>
 #include <linux/dvb/net.h>
+#include <linux/smp_lock.h>
 #include <linux/uio.h>
 #include <asm/uaccess.h>
 #include <linux/crc32.h>
@@ -350,6 +351,7 @@ static void dvb_net_ule( struct net_device *dev, const u8 *buf, size_t buf_len )
 	const u8 *ts, *ts_end, *from_where = NULL;
 	u8 ts_remain = 0, how_much = 0, new_ts = 1;
 	struct ethhdr *ethh = NULL;
+	bool error = false;
 
 #ifdef ULE_DEBUG
 	/* The code inside ULE_DEBUG keeps a history of the last 100 TS cells processed. */
@@ -459,10 +461,16 @@ static void dvb_net_ule( struct net_device *dev, const u8 *buf, size_t buf_len )
 
 						/* Drop partly decoded SNDU, reset state, resync on PUSI. */
 						if (priv->ule_skb) {
-							dev_kfree_skb( priv->ule_skb );
+							error = true;
+							dev_kfree_skb(priv->ule_skb);
+						}
+
+						if (error || priv->ule_sndu_remain) {
 							dev->stats.rx_errors++;
 							dev->stats.rx_frame_errors++;
+							error = false;
 						}
+
 						reset_ule(priv);
 						priv->need_pusi = 1;
 						continue;
@@ -534,6 +542,7 @@ static void dvb_net_ule( struct net_device *dev, const u8 *buf, size_t buf_len )
 				from_where += 2;
 			}
 
+			priv->ule_sndu_remain = priv->ule_sndu_len + 2;
 			/*
 			 * State of current TS:
 			 *   ts_remain (remaining bytes in the current TS cell)
@@ -543,6 +552,7 @@ static void dvb_net_ule( struct net_device *dev, const u8 *buf, size_t buf_len )
 			 */
 			switch (ts_remain) {
 				case 1:
+					priv->ule_sndu_remain--;
 					priv->ule_sndu_type = from_where[0] << 8;
 					priv->ule_sndu_type_1 = 1; /* first byte of ule_type is set. */
 					ts_remain -= 1; from_where += 1;
@@ -556,6 +566,7 @@ static void dvb_net_ule( struct net_device *dev, const u8 *buf, size_t buf_len )
 				default: /* complete ULE header is present in current TS. */
 					/* Extract ULE type field. */
 					if (priv->ule_sndu_type_1) {
+						priv->ule_sndu_type_1 = 0;
 						priv->ule_sndu_type |= from_where[0];
 						from_where += 1; /* points to payload start. */
 						ts_remain -= 1;
@@ -1329,7 +1340,7 @@ static int dvb_net_remove_if(struct dvb_net *dvbnet, unsigned long num)
 	return 0;
 }
 
-static int dvb_net_do_ioctl(struct inode *inode, struct file *file,
+static int dvb_net_do_ioctl(struct file *file,
 		  unsigned int cmd, void *parg)
 {
 	struct dvb_device *dvbdev = file->private_data;
@@ -1431,10 +1442,16 @@ static int dvb_net_do_ioctl(struct inode *inode, struct file *file,
 	return 0;
 }
 
-static int dvb_net_ioctl(struct inode *inode, struct file *file,
+static long dvb_net_ioctl(struct file *file,
 	      unsigned int cmd, unsigned long arg)
 {
-	return dvb_usercopy(inode, file, cmd, arg, dvb_net_do_ioctl);
+	int ret;
+
+	lock_kernel();
+	ret = dvb_usercopy(file, cmd, arg, dvb_net_do_ioctl);
+	unlock_kernel();
+
+	return ret;
 }
 
 static int dvb_net_close(struct inode *inode, struct file *file)
@@ -1455,7 +1472,7 @@ static int dvb_net_close(struct inode *inode, struct file *file)
 
 static const struct file_operations dvb_net_fops = {
 	.owner = THIS_MODULE,
-	.ioctl = dvb_net_ioctl,
+	.unlocked_ioctl = dvb_net_ioctl,
 	.open =	dvb_generic_open,
 	.release = dvb_net_close,
 };
diff --git a/drivers/media/dvb/dvb-core/dvbdev.c b/drivers/media/dvb/dvb-core/dvbdev.c
index 94159b90f733..b915c39d782f 100644
--- a/drivers/media/dvb/dvb-core/dvbdev.c
+++ b/drivers/media/dvb/dvb-core/dvbdev.c
@@ -154,10 +154,11 @@ int dvb_generic_release(struct inode *inode, struct file *file)
 EXPORT_SYMBOL(dvb_generic_release);
 
 
-int dvb_generic_ioctl(struct inode *inode, struct file *file,
-		      unsigned int cmd, unsigned long arg)
+long dvb_generic_ioctl(struct file *file,
+		       unsigned int cmd, unsigned long arg)
 {
 	struct dvb_device *dvbdev = file->private_data;
+	int ret;
 
 	if (!dvbdev)
 		return -ENODEV;
@@ -165,7 +166,11 @@ int dvb_generic_ioctl(struct inode *inode, struct file *file,
 	if (!dvbdev->kernel_ioctl)
 		return -EINVAL;
 
-	return dvb_usercopy (inode, file, cmd, arg, dvbdev->kernel_ioctl);
+	lock_kernel();
+	ret = dvb_usercopy(file, cmd, arg, dvbdev->kernel_ioctl);
+	unlock_kernel();
+
+	return ret;
 }
 EXPORT_SYMBOL(dvb_generic_ioctl);
 
@@ -377,9 +382,9 @@ EXPORT_SYMBOL(dvb_unregister_adapter);
    define this as video_usercopy(). this will introduce a dependecy
    to the v4l "videodev.o" module, which is unnecessary for some
    cards (ie. the budget dvb-cards don't need the v4l module...) */
-int dvb_usercopy(struct inode *inode, struct file *file,
+int dvb_usercopy(struct file *file,
 		     unsigned int cmd, unsigned long arg,
-		     int (*func)(struct inode *inode, struct file *file,
+		     int (*func)(struct file *file,
 		     unsigned int cmd, void *arg))
 {
 	char    sbuf[128];
@@ -416,7 +421,7 @@ int dvb_usercopy(struct inode *inode, struct file *file,
 	}
 
 	/* call driver */
-	if ((err = func(inode, file, cmd, parg)) == -ENOIOCTLCMD)
+	if ((err = func(file, cmd, parg)) == -ENOIOCTLCMD)
 		err = -EINVAL;
 
 	if (err < 0)
diff --git a/drivers/media/dvb/dvb-core/dvbdev.h b/drivers/media/dvb/dvb-core/dvbdev.h
index f7b499d4a3c0..fcc6ae98745e 100644
--- a/drivers/media/dvb/dvb-core/dvbdev.h
+++ b/drivers/media/dvb/dvb-core/dvbdev.h
@@ -116,8 +116,7 @@ struct dvb_device {
 
 	wait_queue_head_t	  wait_queue;
 	/* don't really need those !? -- FIXME: use video_usercopy  */
-	int (*kernel_ioctl)(struct inode *inode, struct file *file,
-			    unsigned int cmd, void *arg);
+	int (*kernel_ioctl)(struct file *file, unsigned int cmd, void *arg);
 
 	void *priv;
 };
@@ -138,17 +137,15 @@ extern void dvb_unregister_device (struct dvb_device *dvbdev);
 
 extern int dvb_generic_open (struct inode *inode, struct file *file);
 extern int dvb_generic_release (struct inode *inode, struct file *file);
-extern int dvb_generic_ioctl (struct inode *inode, struct file *file,
+extern long dvb_generic_ioctl (struct file *file,
 			      unsigned int cmd, unsigned long arg);
 
 /* we don't mess with video_usercopy() any more,
 we simply define out own dvb_usercopy(), which will hopefully become
 generic_usercopy()  someday... */
 
-extern int dvb_usercopy(struct inode *inode, struct file *file,
-			    unsigned int cmd, unsigned long arg,
-			    int (*func)(struct inode *inode, struct file *file,
-			    unsigned int cmd, void *arg));
+extern int dvb_usercopy(struct file *file, unsigned int cmd, unsigned long arg,
+			    int (*func)(struct file *file, unsigned int cmd, void *arg));
 
 /** generic DVB attach function. */
 #ifdef CONFIG_MEDIA_ATTACH
diff --git a/drivers/media/dvb/dvb-usb/Kconfig b/drivers/media/dvb/dvb-usb/Kconfig
index e5f91f16ffa4..553b48ac1919 100644
--- a/drivers/media/dvb/dvb-usb/Kconfig
+++ b/drivers/media/dvb/dvb-usb/Kconfig
@@ -76,6 +76,7 @@ config DVB_USB_DIB0700
 	select DVB_S5H1411 if !DVB_FE_CUSTOMISE
 	select DVB_LGDT3305 if !DVB_FE_CUSTOMISE
 	select DVB_TUNER_DIB0070 if !DVB_FE_CUSTOMISE
+	select DVB_TUNER_DIB0090 if !DVB_FE_CUSTOMISE
 	select MEDIA_TUNER_MT2060 if !MEDIA_TUNER_CUSTOMISE
 	select MEDIA_TUNER_MT2266 if !MEDIA_TUNER_CUSTOMISE
 	select MEDIA_TUNER_XC2028 if !MEDIA_TUNER_CUSTOMISE
@@ -134,6 +135,7 @@ config DVB_USB_M920X
 	select DVB_TDA1004X if !DVB_FE_CUSTOMISE
 	select MEDIA_TUNER_QT1010 if !MEDIA_TUNER_CUSTOMISE
 	select MEDIA_TUNER_TDA827X if !MEDIA_TUNER_CUSTOMISE
+	select MEDIA_TUNER_SIMPLE if !MEDIA_TUNER_CUSTOMISE
 	help
 	  Say Y here to support the MSI Mega Sky 580 USB2.0 DVB-T receiver.
 	  Currently, only devices with a product id of
@@ -264,7 +266,7 @@ config DVB_USB_DW2102
 	select DVB_STB6000 if !DVB_FE_CUSTOMISE
 	select DVB_CX24116 if !DVB_FE_CUSTOMISE
 	select DVB_SI21XX if !DVB_FE_CUSTOMISE
-	select DVB_TDA10021 if !DVB_FE_CUSTOMISE
+	select DVB_TDA10023 if !DVB_FE_CUSTOMISE
 	select DVB_MT312 if !DVB_FE_CUSTOMISE
 	select DVB_ZL10039 if !DVB_FE_CUSTOMISE
 	select DVB_DS3000 if !DVB_FE_CUSTOMISE
diff --git a/drivers/media/dvb/dvb-usb/cxusb.c b/drivers/media/dvb/dvb-usb/cxusb.c
index 0eb490889162..11e9e85dac86 100644
--- a/drivers/media/dvb/dvb-usb/cxusb.c
+++ b/drivers/media/dvb/dvb-usb/cxusb.c
@@ -1026,8 +1026,10 @@ static int cxusb_dualdig4_rev2_frontend_attach(struct dvb_usb_adapter *adap)
 	cxusb_bluebird_gpio_pulse(adap->dev, 0x02, 1);
 
 	if (dib7000p_i2c_enumeration(&adap->dev->i2c_adap, 1, 18,
-				 &cxusb_dualdig4_rev2_config) < 0)
+				     &cxusb_dualdig4_rev2_config) < 0) {
+		printk(KERN_WARNING "Unable to enumerate dib7000p\n");
 		return -ENODEV;
+	}
 
 	adap->fe = dvb_attach(dib7000p_attach, &adap->dev->i2c_adap, 0x80,
 			      &cxusb_dualdig4_rev2_config);
diff --git a/drivers/media/dvb/dvb-usb/dvb-usb-ids.h b/drivers/media/dvb/dvb-usb/dvb-usb-ids.h
index 085c4e457e0e..b4afe6f8ed19 100644
--- a/drivers/media/dvb/dvb-usb/dvb-usb-ids.h
+++ b/drivers/media/dvb/dvb-usb/dvb-usb-ids.h
@@ -198,6 +198,7 @@
 #define USB_PID_AVERMEDIA_A850				0x850a
 #define USB_PID_AVERMEDIA_A805				0xa805
 #define USB_PID_TECHNOTREND_CONNECT_S2400               0x3006
+#define USB_PID_TECHNOTREND_CONNECT_CT3650		0x300d
 #define USB_PID_TERRATEC_CINERGY_DT_XS_DIVERSITY	0x005a
 #define USB_PID_TERRATEC_CINERGY_DT_XS_DIVERSITY_2	0x0081
 #define USB_PID_TERRATEC_CINERGY_HT_USB_XE		0x0058
diff --git a/drivers/media/dvb/dvb-usb/ttusb2.c b/drivers/media/dvb/dvb-usb/ttusb2.c
index 20ca9d9ee99b..a6de489a6a39 100644
--- a/drivers/media/dvb/dvb-usb/ttusb2.c
+++ b/drivers/media/dvb/dvb-usb/ttusb2.c
@@ -29,6 +29,8 @@
 
 #include "tda826x.h"
 #include "tda10086.h"
+#include "tda1002x.h"
+#include "tda827x.h"
 #include "lnbp21.h"
 
 /* debug */
@@ -150,7 +152,17 @@ static struct tda10086_config tda10086_config = {
 	.xtal_freq = TDA10086_XTAL_16M,
 };
 
-static int ttusb2_frontend_attach(struct dvb_usb_adapter *adap)
+static struct tda10023_config tda10023_config = {
+	.demod_address = 0x0c,
+	.invert = 0,
+	.xtal = 16000000,
+	.pll_m = 11,
+	.pll_p = 3,
+	.pll_n = 1,
+	.deltaf = 0xa511,
+};
+
+static int ttusb2_frontend_tda10086_attach(struct dvb_usb_adapter *adap)
 {
 	if (usb_set_interface(adap->dev->udev,0,3) < 0)
 		err("set interface to alts=3 failed");
@@ -163,7 +175,27 @@ static int ttusb2_frontend_attach(struct dvb_usb_adapter *adap)
 	return 0;
 }
 
-static int ttusb2_tuner_attach(struct dvb_usb_adapter *adap)
+static int ttusb2_frontend_tda10023_attach(struct dvb_usb_adapter *adap)
+{
+	if (usb_set_interface(adap->dev->udev, 0, 3) < 0)
+		err("set interface to alts=3 failed");
+	if ((adap->fe = dvb_attach(tda10023_attach, &tda10023_config, &adap->dev->i2c_adap, 0x48)) == NULL) {
+		deb_info("TDA10023 attach failed\n");
+		return -ENODEV;
+	}
+	return 0;
+}
+
+static int ttusb2_tuner_tda827x_attach(struct dvb_usb_adapter *adap)
+{
+	if (dvb_attach(tda827x_attach, adap->fe, 0x61, &adap->dev->i2c_adap, NULL) == NULL) {
+		printk(KERN_ERR "%s: No tda827x found!\n", __func__);
+		return -ENODEV;
+	}
+	return 0;
+}
+
+static int ttusb2_tuner_tda826x_attach(struct dvb_usb_adapter *adap)
 {
 	if (dvb_attach(tda826x_attach, adap->fe, 0x60, &adap->dev->i2c_adap, 0) == NULL) {
 		deb_info("TDA8263 attach failed\n");
@@ -180,6 +212,7 @@ static int ttusb2_tuner_attach(struct dvb_usb_adapter *adap)
 /* DVB USB Driver stuff */
 static struct dvb_usb_device_properties ttusb2_properties;
 static struct dvb_usb_device_properties ttusb2_properties_s2400;
+static struct dvb_usb_device_properties ttusb2_properties_ct3650;
 
 static int ttusb2_probe(struct usb_interface *intf,
 		const struct usb_device_id *id)
@@ -187,6 +220,8 @@ static int ttusb2_probe(struct usb_interface *intf,
 	if (0 == dvb_usb_device_init(intf, &ttusb2_properties,
 				     THIS_MODULE, NULL, adapter_nr) ||
 	    0 == dvb_usb_device_init(intf, &ttusb2_properties_s2400,
+				     THIS_MODULE, NULL, adapter_nr) ||
+	    0 == dvb_usb_device_init(intf, &ttusb2_properties_ct3650,
 				     THIS_MODULE, NULL, adapter_nr))
 		return 0;
 	return -ENODEV;
@@ -197,6 +232,8 @@ static struct usb_device_id ttusb2_table [] = {
 	{ USB_DEVICE(USB_VID_PINNACLE, USB_PID_PCTV_450E) },
 	{ USB_DEVICE(USB_VID_TECHNOTREND,
 		USB_PID_TECHNOTREND_CONNECT_S2400) },
+	{ USB_DEVICE(USB_VID_TECHNOTREND,
+		USB_PID_TECHNOTREND_CONNECT_CT3650) },
 	{}		/* Terminating entry */
 };
 MODULE_DEVICE_TABLE (usb, ttusb2_table);
@@ -214,8 +251,8 @@ static struct dvb_usb_device_properties ttusb2_properties = {
 		{
 			.streaming_ctrl   = NULL, // ttusb2_streaming_ctrl,
 
-			.frontend_attach  = ttusb2_frontend_attach,
-			.tuner_attach     = ttusb2_tuner_attach,
+			.frontend_attach  = ttusb2_frontend_tda10086_attach,
+			.tuner_attach     = ttusb2_tuner_tda826x_attach,
 
 			/* parameter for the MPEG2-data transfer */
 			.stream = {
@@ -266,8 +303,8 @@ static struct dvb_usb_device_properties ttusb2_properties_s2400 = {
 		{
 			.streaming_ctrl   = NULL,
 
-			.frontend_attach  = ttusb2_frontend_attach,
-			.tuner_attach     = ttusb2_tuner_attach,
+			.frontend_attach  = ttusb2_frontend_tda10086_attach,
+			.tuner_attach     = ttusb2_tuner_tda826x_attach,
 
 			/* parameter for the MPEG2-data transfer */
 			.stream = {
@@ -301,6 +338,52 @@ static struct dvb_usb_device_properties ttusb2_properties_s2400 = {
 	}
 };
 
+static struct dvb_usb_device_properties ttusb2_properties_ct3650 = {
+	.caps = DVB_USB_IS_AN_I2C_ADAPTER,
+
+	.usb_ctrl = CYPRESS_FX2,
+
+	.size_of_priv = sizeof(struct ttusb2_state),
+
+	.num_adapters = 1,
+	.adapter = {
+		{
+			.streaming_ctrl   = NULL,
+
+			.frontend_attach  = ttusb2_frontend_tda10023_attach,
+			.tuner_attach = ttusb2_tuner_tda827x_attach,
+
+			/* parameter for the MPEG2-data transfer */
+			.stream = {
+				.type = USB_ISOC,
+				.count = 5,
+				.endpoint = 0x02,
+				.u = {
+					.isoc = {
+						.framesperurb = 4,
+						.framesize = 940,
+						.interval = 1,
+					}
+				}
+			}
+		},
+	},
+
+	.power_ctrl       = ttusb2_power_ctrl,
+	.identify_state   = ttusb2_identify_state,
+
+	.i2c_algo         = &ttusb2_i2c_algo,
+
+	.generic_bulk_ctrl_endpoint = 0x01,
+
+	.num_device_descs = 1,
+	.devices = {
+		{   "Technotrend TT-connect CT-3650",
+			.warm_ids = { &ttusb2_table[3], NULL },
+		},
+	}
+};
+
 static struct usb_driver ttusb2_driver = {
 	.name		= "dvb_usb_ttusb2",
 	.probe		= ttusb2_probe,
diff --git a/drivers/media/dvb/firewire/firedtv-1394.c b/drivers/media/dvb/firewire/firedtv-1394.c
index 26333b4f4d3e..b34ca7afb0e6 100644
--- a/drivers/media/dvb/firewire/firedtv-1394.c
+++ b/drivers/media/dvb/firewire/firedtv-1394.c
@@ -58,7 +58,7 @@ static void rawiso_activity_cb(struct hpsb_iso *iso)
 	num = hpsb_iso_n_ready(iso);
 
 	if (!fdtv) {
-		dev_err(fdtv->device, "received at unknown iso channel\n");
+		pr_err("received at unknown iso channel\n");
 		goto out;
 	}
 
diff --git a/drivers/media/dvb/firewire/firedtv-ci.c b/drivers/media/dvb/firewire/firedtv-ci.c
index 853e04b7cb36..d3c2cf60de76 100644
--- a/drivers/media/dvb/firewire/firedtv-ci.c
+++ b/drivers/media/dvb/firewire/firedtv-ci.c
@@ -175,8 +175,7 @@ static int fdtv_ca_send_msg(struct firedtv *fdtv, void *arg)
 	return err;
 }
 
-static int fdtv_ca_ioctl(struct inode *inode, struct file *file,
-			    unsigned int cmd, void *arg)
+static int fdtv_ca_ioctl(struct file *file, unsigned int cmd, void *arg)
 {
 	struct dvb_device *dvbdev = file->private_data;
 	struct firedtv *fdtv = dvbdev->priv;
@@ -217,7 +216,7 @@ static unsigned int fdtv_ca_io_poll(struct file *file, poll_table *wait)
 
 static const struct file_operations fdtv_ca_fops = {
 	.owner		= THIS_MODULE,
-	.ioctl		= dvb_generic_ioctl,
+	.unlocked_ioctl	= dvb_generic_ioctl,
 	.open		= dvb_generic_open,
 	.release	= dvb_generic_release,
 	.poll		= fdtv_ca_io_poll,
diff --git a/drivers/media/dvb/frontends/au8522_decoder.c b/drivers/media/dvb/frontends/au8522_decoder.c
index 68dba3a4b4da..29cdbfe36852 100644
--- a/drivers/media/dvb/frontends/au8522_decoder.c
+++ b/drivers/media/dvb/frontends/au8522_decoder.c
@@ -567,30 +567,6 @@ static int au8522_g_ctrl(struct v4l2_subdev *sd, struct v4l2_control *ctrl)
 
 /* ----------------------------------------------------------------------- */
 
-static int au8522_g_fmt(struct v4l2_subdev *sd, struct v4l2_format *fmt)
-{
-	switch (fmt->type) {
-	default:
-		return -EINVAL;
-	}
-	return 0;
-}
-
-static int au8522_s_fmt(struct v4l2_subdev *sd, struct v4l2_format *fmt)
-{
-	switch (fmt->type) {
-	case V4L2_BUF_TYPE_VIDEO_CAPTURE:
-		/* Not yet implemented */
-		break;
-	default:
-		return -EINVAL;
-	}
-
-	return 0;
-}
-
-/* ----------------------------------------------------------------------- */
-
 #ifdef CONFIG_VIDEO_ADV_DEBUG
 static int au8522_g_register(struct v4l2_subdev *sd,
 			     struct v4l2_dbg_register *reg)
@@ -772,8 +748,6 @@ static const struct v4l2_subdev_audio_ops au8522_audio_ops = {
 
 static const struct v4l2_subdev_video_ops au8522_video_ops = {
 	.s_routing = au8522_s_video_routing,
-	.g_fmt = au8522_g_fmt,
-	.s_fmt = au8522_s_fmt,
 	.s_stream = au8522_s_stream,
 };
 
diff --git a/drivers/media/dvb/frontends/ds3000.c b/drivers/media/dvb/frontends/ds3000.c
index 78001e8bcdb7..fc61d9230db8 100644
--- a/drivers/media/dvb/frontends/ds3000.c
+++ b/drivers/media/dvb/frontends/ds3000.c
@@ -969,15 +969,12 @@ struct dvb_frontend *ds3000_attach(const struct ds3000_config *config,
 	dprintk("%s\n", __func__);
 
 	/* allocate memory for the internal state */
-	state = kmalloc(sizeof(struct ds3000_state), GFP_KERNEL);
+	state = kzalloc(sizeof(struct ds3000_state), GFP_KERNEL);
 	if (state == NULL) {
 		printk(KERN_ERR "Unable to kmalloc\n");
 		goto error2;
 	}
 
-	/* setup the state */
-	memset(state, 0, sizeof(struct ds3000_state));
-
 	state->config = config;
 	state->i2c = i2c;
 	state->prevUCBS2 = 0;
diff --git a/drivers/media/dvb/frontends/stv6110x.c b/drivers/media/dvb/frontends/stv6110x.c
index 42591ce1aaad..f36cab12bdc7 100644
--- a/drivers/media/dvb/frontends/stv6110x.c
+++ b/drivers/media/dvb/frontends/stv6110x.c
@@ -303,7 +303,10 @@ static int stv6110x_set_mode(struct dvb_frontend *fe, enum tuner_mode mode)
 
 static int stv6110x_sleep(struct dvb_frontend *fe)
 {
-	return stv6110x_set_mode(fe, TUNER_SLEEP);
+	if (fe->tuner_priv)
+		return stv6110x_set_mode(fe, TUNER_SLEEP);
+
+	return 0;
 }
 
 static int stv6110x_get_status(struct dvb_frontend *fe, u32 *status)
diff --git a/drivers/media/dvb/ngene/ngene-cards.c b/drivers/media/dvb/ngene/ngene-cards.c
index 692c3e226e83..4692a41ad95b 100644
--- a/drivers/media/dvb/ngene/ngene-cards.c
+++ b/drivers/media/dvb/ngene/ngene-cards.c
@@ -217,6 +217,19 @@ static struct ngene_info ngene_info_cineS2v5 = {
 	.fw_version	= 15,
 };
 
+static struct ngene_info ngene_info_duoFlexS2 = {
+	.type           = NGENE_SIDEWINDER,
+	.name           = "Digital Devices DuoFlex S2 miniPCIe",
+	.io_type        = {NGENE_IO_TSIN, NGENE_IO_TSIN},
+	.demod_attach   = {demod_attach_stv0900, demod_attach_stv0900},
+	.tuner_attach   = {tuner_attach_stv6110, tuner_attach_stv6110},
+	.fe_config      = {&fe_cineS2, &fe_cineS2},
+	.tuner_config   = {&tuner_cineS2_0, &tuner_cineS2_1},
+	.lnb            = {0x0a, 0x08},
+	.tsf            = {3, 3},
+	.fw_version     = 15,
+};
+
 static struct ngene_info ngene_info_m780 = {
 	.type           = NGENE_APP,
 	.name           = "Aver M780 ATSC/QAM-B",
@@ -256,6 +269,8 @@ static const struct pci_device_id ngene_id_tbl[] __devinitdata = {
 	NGENE_ID(0x18c3, 0xdb01, ngene_info_satixS2),
 	NGENE_ID(0x18c3, 0xdb02, ngene_info_satixS2v2),
 	NGENE_ID(0x18c3, 0xdd00, ngene_info_cineS2v5),
+	NGENE_ID(0x18c3, 0xdd10, ngene_info_duoFlexS2),
+	NGENE_ID(0x18c3, 0xdd20, ngene_info_duoFlexS2),
 	NGENE_ID(0x1461, 0x062e, ngene_info_m780),
 	{0}
 };
diff --git a/drivers/media/dvb/ngene/ngene-core.c b/drivers/media/dvb/ngene/ngene-core.c
index c8b4dfa0ab5f..4caeb163a666 100644
--- a/drivers/media/dvb/ngene/ngene-core.c
+++ b/drivers/media/dvb/ngene/ngene-core.c
@@ -53,8 +53,6 @@ MODULE_PARM_DESC(debug, "Print debugging information.");
 
 DVB_DEFINE_MOD_OPT_ADAPTER_NR(adapter_nr);
 
-#define COMMAND_TIMEOUT_WORKAROUND
-
 #define dprintk	if (debug) printk
 
 #define ngwriteb(dat, adr)         writeb((dat), (char *)(dev->iomem + (adr)))
@@ -147,24 +145,24 @@ static void demux_tasklet(unsigned long data)
 		} else {
 			if (chan->HWState == HWSTATE_RUN) {
 				u32 Flags = 0;
+				IBufferExchange *exch1 = chan->pBufferExchange;
+				IBufferExchange *exch2 = chan->pBufferExchange2;
 				if (Cur->ngeneBuffer.SR.Flags & 0x01)
 					Flags |= BEF_EVEN_FIELD;
 				if (Cur->ngeneBuffer.SR.Flags & 0x20)
 					Flags |= BEF_OVERFLOW;
-				if (chan->pBufferExchange)
-					chan->pBufferExchange(chan,
-							      Cur->Buffer1,
-							      chan->
-							      Capture1Length,
-							      Cur->ngeneBuffer.
-							      SR.Clock, Flags);
-				if (chan->pBufferExchange2)
-					chan->pBufferExchange2(chan,
-							       Cur->Buffer2,
-							       chan->
-							       Capture2Length,
-							       Cur->ngeneBuffer.
-							       SR.Clock, Flags);
+				spin_unlock_irq(&chan->state_lock);
+				if (exch1)
+					exch1(chan, Cur->Buffer1,
+						chan->Capture1Length,
+						Cur->ngeneBuffer.SR.Clock,
+						Flags);
+				if (exch2)
+					exch2(chan, Cur->Buffer2,
+						chan->Capture2Length,
+						Cur->ngeneBuffer.SR.Clock,
+						Flags);
+				spin_lock_irq(&chan->state_lock);
 			} else if (chan->HWState != HWSTATE_STOP)
 				chan->HWState = HWSTATE_RUN;
 		}
@@ -572,11 +570,7 @@ static int ngene_command_stream_control(struct ngene *dev, u8 stream,
 	u16 BsSPI = ((stream & 1) ? 0x9800 : 0x9700);
 	u16 BsSDO = 0x9B00;
 
-	/* down(&dev->stream_mutex); */
-	while (down_trylock(&dev->stream_mutex)) {
-		printk(KERN_INFO DEVICE_NAME ": SC locked\n");
-		msleep(1);
-	}
+	down(&dev->stream_mutex);
 	memset(&com, 0, sizeof(com));
 	com.cmd.hdr.Opcode = CMD_CONTROL;
 	com.cmd.hdr.Length = sizeof(struct FW_STREAM_CONTROL) - 2;
@@ -1252,14 +1246,17 @@ static int ngene_load_firm(struct ngene *dev)
 		version = 15;
 		size = 23466;
 		fw_name = "ngene_15.fw";
+		dev->cmd_timeout_workaround = true;
 		break;
 	case 16:
 		size = 23498;
 		fw_name = "ngene_16.fw";
+		dev->cmd_timeout_workaround = true;
 		break;
 	case 17:
 		size = 24446;
 		fw_name = "ngene_17.fw";
+		dev->cmd_timeout_workaround = true;
 		break;
 	}
 
@@ -1299,11 +1296,16 @@ static void ngene_stop(struct ngene *dev)
 	ngwritel(0, NGENE_EVENT);
 	ngwritel(0, NGENE_EVENT_HI);
 	free_irq(dev->pci_dev->irq, dev);
+#ifdef CONFIG_PCI_MSI
+	if (dev->msi_enabled)
+		pci_disable_msi(dev->pci_dev);
+#endif
 }
 
 static int ngene_start(struct ngene *dev)
 {
 	int stat;
+	unsigned long flags;
 	int i;
 
 	pci_set_master(dev->pci_dev);
@@ -1333,6 +1335,28 @@ static int ngene_start(struct ngene *dev)
 	if (stat < 0)
 		goto fail;
 
+#ifdef CONFIG_PCI_MSI
+	/* enable MSI if kernel and card support it */
+	if (pci_msi_enabled() && dev->card_info->msi_supported) {
+		ngwritel(0, NGENE_INT_ENABLE);
+		free_irq(dev->pci_dev->irq, dev);
+		stat = pci_enable_msi(dev->pci_dev);
+		if (stat) {
+			printk(KERN_INFO DEVICE_NAME
+				": MSI not available\n");
+			flags = IRQF_SHARED;
+		} else {
+			flags = 0;
+			dev->msi_enabled = true;
+		}
+		stat = request_irq(dev->pci_dev->irq, irq_handler,
+					flags, "nGene", dev);
+		if (stat < 0)
+			goto fail2;
+		ngwritel(1, NGENE_INT_ENABLE);
+	}
+#endif
+
 	stat = ngene_i2c_init(dev, 0);
 	if (stat < 0)
 		goto fail;
@@ -1358,10 +1382,18 @@ static int ngene_start(struct ngene *dev)
 			bconf = BUFFER_CONFIG_3333;
 		stat = ngene_command_config_buf(dev, bconf);
 	}
-	return stat;
+	if (!stat)
+		return stat;
+
+	/* otherwise error: fall through */
 fail:
 	ngwritel(0, NGENE_INT_ENABLE);
 	free_irq(dev->pci_dev->irq, dev);
+#ifdef CONFIG_PCI_MSI
+fail2:
+	if (dev->msi_enabled)
+		pci_disable_msi(dev->pci_dev);
+#endif
 	return stat;
 }
 
@@ -1379,10 +1411,8 @@ static void release_channel(struct ngene_channel *chan)
 	struct ngene_info *ni = dev->card_info;
 	int io = ni->io_type[chan->number];
 
-#ifdef COMMAND_TIMEOUT_WORKAROUND
-	if (chan->running)
+	if (chan->dev->cmd_timeout_workaround && chan->running)
 		set_transfer(chan, 0);
-#endif
 
 	tasklet_kill(&chan->demux_tasklet);
 
diff --git a/drivers/media/dvb/ngene/ngene-dvb.c b/drivers/media/dvb/ngene/ngene-dvb.c
index 96013eb353cd..48f980b21d66 100644
--- a/drivers/media/dvb/ngene/ngene-dvb.c
+++ b/drivers/media/dvb/ngene/ngene-dvb.c
@@ -37,15 +37,12 @@
 #include <linux/pci.h>
 #include <linux/smp_lock.h>
 #include <linux/timer.h>
-#include <linux/version.h>
 #include <linux/byteorder/generic.h>
 #include <linux/firmware.h>
 #include <linux/vmalloc.h>
 
 #include "ngene.h"
 
-#define COMMAND_TIMEOUT_WORKAROUND
-
 
 /****************************************************************************/
 /* COMMAND API interface ****************************************************/
@@ -69,9 +66,7 @@ void *tsin_exchange(void *priv, void *buf, u32 len, u32 clock, u32 flags)
 	struct ngene_channel *chan = priv;
 
 
-#ifdef COMMAND_TIMEOUT_WORKAROUND
 	if (chan->users > 0)
-#endif
 		dvb_dmx_swfilter(&chan->demux, buf, len);
 	return NULL;
 }
@@ -106,11 +101,8 @@ int ngene_start_feed(struct dvb_demux_feed *dvbdmxfeed)
 	struct ngene_channel *chan = dvbdmx->priv;
 
 	if (chan->users == 0) {
-#ifdef COMMAND_TIMEOUT_WORKAROUND
-		if (!chan->running)
-#endif
+		if (!chan->dev->cmd_timeout_workaround || !chan->running)
 			set_transfer(chan, 1);
-		/* msleep(10); */
 	}
 
 	return ++chan->users;
@@ -124,9 +116,8 @@ int ngene_stop_feed(struct dvb_demux_feed *dvbdmxfeed)
 	if (--chan->users)
 		return chan->users;
 
-#ifndef COMMAND_TIMEOUT_WORKAROUND
-	set_transfer(chan, 0);
-#endif
+	if (!chan->dev->cmd_timeout_workaround)
+		set_transfer(chan, 0);
 
 	return 0;
 }
diff --git a/drivers/media/dvb/ngene/ngene-i2c.c b/drivers/media/dvb/ngene/ngene-i2c.c
index 2ef54ca6badd..477fe0aade86 100644
--- a/drivers/media/dvb/ngene/ngene-i2c.c
+++ b/drivers/media/dvb/ngene/ngene-i2c.c
@@ -39,7 +39,6 @@
 #include <linux/pci_ids.h>
 #include <linux/smp_lock.h>
 #include <linux/timer.h>
-#include <linux/version.h>
 #include <linux/byteorder/generic.h>
 #include <linux/firmware.h>
 #include <linux/vmalloc.h>
diff --git a/drivers/media/dvb/ngene/ngene.h b/drivers/media/dvb/ngene/ngene.h
index 676fcbb79026..8fb4200f83f8 100644
--- a/drivers/media/dvb/ngene/ngene.h
+++ b/drivers/media/dvb/ngene/ngene.h
@@ -725,6 +725,8 @@ struct ngene {
 	u32                   device_version;
 	u32                   fw_interface_version;
 	u32                   icounts;
+	bool                  msi_enabled;
+	bool                  cmd_timeout_workaround;
 
 	u8                   *CmdDoneByte;
 	int                   BootFirmware;
@@ -797,6 +799,7 @@ struct ngene_info {
 #define NGENE_VBOX_V2	 7
 
 	int   fw_version;
+	bool  msi_supported;
 	char *name;
 
 	int   io_type[MAX_STREAM];
diff --git a/drivers/media/dvb/ttpci/Kconfig b/drivers/media/dvb/ttpci/Kconfig
index d8d4214fd65f..32a7ec65ec42 100644
--- a/drivers/media/dvb/ttpci/Kconfig
+++ b/drivers/media/dvb/ttpci/Kconfig
@@ -68,13 +68,14 @@ config DVB_BUDGET
 	select DVB_VES1820 if !DVB_FE_CUSTOMISE
 	select DVB_L64781 if !DVB_FE_CUSTOMISE
 	select DVB_TDA8083 if !DVB_FE_CUSTOMISE
-	select DVB_TDA10021 if !DVB_FE_CUSTOMISE
-	select DVB_TDA10023 if !DVB_FE_CUSTOMISE
 	select DVB_S5H1420 if !DVB_FE_CUSTOMISE
 	select DVB_TDA10086 if !DVB_FE_CUSTOMISE
 	select DVB_TDA826X if !DVB_FE_CUSTOMISE
 	select DVB_LNBP21 if !DVB_FE_CUSTOMISE
 	select DVB_TDA1004X if !DVB_FE_CUSTOMISE
+	select DVB_ISL6423 if !DVB_FE_CUSTOMISE
+	select DVB_STV090x if !DVB_FE_CUSTOMISE
+	select DVB_STV6110x if !DVB_FE_CUSTOMISE
 	help
 	  Support for simple SAA7146 based DVB cards (so called Budget-
 	  or Nova-PCI cards) without onboard MPEG2 decoder, and without
diff --git a/drivers/media/dvb/ttpci/av7110.c b/drivers/media/dvb/ttpci/av7110.c
index 38915591c6e5..a6be529eec5c 100644
--- a/drivers/media/dvb/ttpci/av7110.c
+++ b/drivers/media/dvb/ttpci/av7110.c
@@ -708,7 +708,7 @@ static void gpioirq(unsigned long cookie)
 
 
 #ifdef CONFIG_DVB_AV7110_OSD
-static int dvb_osd_ioctl(struct inode *inode, struct file *file,
+static int dvb_osd_ioctl(struct file *file,
 			 unsigned int cmd, void *parg)
 {
 	struct dvb_device *dvbdev = file->private_data;
@@ -727,7 +727,7 @@ static int dvb_osd_ioctl(struct inode *inode, struct file *file,
 
 static const struct file_operations dvb_osd_fops = {
 	.owner		= THIS_MODULE,
-	.ioctl		= dvb_generic_ioctl,
+	.unlocked_ioctl	= dvb_generic_ioctl,
 	.open		= dvb_generic_open,
 	.release	= dvb_generic_release,
 };
diff --git a/drivers/media/dvb/ttpci/av7110_av.c b/drivers/media/dvb/ttpci/av7110_av.c
index 53884814161c..13efba942dac 100644
--- a/drivers/media/dvb/ttpci/av7110_av.c
+++ b/drivers/media/dvb/ttpci/av7110_av.c
@@ -1089,7 +1089,7 @@ static int play_iframe(struct av7110 *av7110, char __user *buf, unsigned int len
 }
 
 
-static int dvb_video_ioctl(struct inode *inode, struct file *file,
+static int dvb_video_ioctl(struct file *file,
 			   unsigned int cmd, void *parg)
 {
 	struct dvb_device *dvbdev = file->private_data;
@@ -1297,7 +1297,7 @@ static int dvb_video_ioctl(struct inode *inode, struct file *file,
 	return ret;
 }
 
-static int dvb_audio_ioctl(struct inode *inode, struct file *file,
+static int dvb_audio_ioctl(struct file *file,
 			   unsigned int cmd, void *parg)
 {
 	struct dvb_device *dvbdev = file->private_data;
@@ -1517,7 +1517,7 @@ static int dvb_audio_release(struct inode *inode, struct file *file)
 static const struct file_operations dvb_video_fops = {
 	.owner		= THIS_MODULE,
 	.write		= dvb_video_write,
-	.ioctl		= dvb_generic_ioctl,
+	.unlocked_ioctl	= dvb_generic_ioctl,
 	.open		= dvb_video_open,
 	.release	= dvb_video_release,
 	.poll		= dvb_video_poll,
@@ -1535,7 +1535,7 @@ static struct dvb_device dvbdev_video = {
 static const struct file_operations dvb_audio_fops = {
 	.owner		= THIS_MODULE,
 	.write		= dvb_audio_write,
-	.ioctl		= dvb_generic_ioctl,
+	.unlocked_ioctl	= dvb_generic_ioctl,
 	.open		= dvb_audio_open,
 	.release	= dvb_audio_release,
 	.poll		= dvb_audio_poll,
diff --git a/drivers/media/dvb/ttpci/av7110_ca.c b/drivers/media/dvb/ttpci/av7110_ca.c
index ac7779c45c5b..4eba35a018e3 100644
--- a/drivers/media/dvb/ttpci/av7110_ca.c
+++ b/drivers/media/dvb/ttpci/av7110_ca.c
@@ -248,8 +248,7 @@ static unsigned int dvb_ca_poll (struct file *file, poll_table *wait)
 	return mask;
 }
 
-static int dvb_ca_ioctl(struct inode *inode, struct file *file,
-		 unsigned int cmd, void *parg)
+static int dvb_ca_ioctl(struct file *file, unsigned int cmd, void *parg)
 {
 	struct dvb_device *dvbdev = file->private_data;
 	struct av7110 *av7110 = dvbdev->priv;
@@ -350,7 +349,7 @@ static const struct file_operations dvb_ca_fops = {
 	.owner		= THIS_MODULE,
 	.read		= dvb_ca_read,
 	.write		= dvb_ca_write,
-	.ioctl		= dvb_generic_ioctl,
+	.unlocked_ioctl	= dvb_generic_ioctl,
 	.open		= dvb_ca_open,
 	.release	= dvb_generic_release,
 	.poll		= dvb_ca_poll,
diff --git a/drivers/media/dvb/ttpci/budget-ci.c b/drivers/media/dvb/ttpci/budget-ci.c
index 461714396331..13ac9e3ab121 100644
--- a/drivers/media/dvb/ttpci/budget-ci.c
+++ b/drivers/media/dvb/ttpci/budget-ci.c
@@ -215,6 +215,7 @@ static int msp430_ir_init(struct budget_ci *budget_ci)
 		break;
 	case 0x1010:
 	case 0x1017:
+	case 0x1019:
 	case 0x101a:
 		/* for the Technotrend 1500 bundled remote */
 		ir_codes = RC_MAP_TT_1500;
diff --git a/drivers/media/radio/si470x/radio-si470x-i2c.c b/drivers/media/radio/si470x/radio-si470x-i2c.c
index a5844d08d8b7..67a4ec8768a6 100644
--- a/drivers/media/radio/si470x/radio-si470x-i2c.c
+++ b/drivers/media/radio/si470x/radio-si470x-i2c.c
@@ -482,7 +482,6 @@ static __devexit int si470x_i2c_remove(struct i2c_client *client)
 	cancel_work_sync(&radio->radio_work);
 	video_unregister_device(radio->videodev);
 	kfree(radio);
-	i2c_set_clientdata(client, NULL);
 
 	return 0;
 }
diff --git a/drivers/media/video/Kconfig b/drivers/media/video/Kconfig
index ad9e6f9c22e9..bdbc9d305419 100644
--- a/drivers/media/video/Kconfig
+++ b/drivers/media/video/Kconfig
@@ -646,7 +646,7 @@ config VIDEO_PMS
 
 config VIDEO_BWQCAM
 	tristate "Quickcam BW Video For Linux"
-	depends on PARPORT && VIDEO_V4L1
+	depends on PARPORT && VIDEO_V4L2
 	help
 	  Say Y have if you the black and white version of the QuickCam
 	  camera. See the next option for the color version.
@@ -656,7 +656,7 @@ config VIDEO_BWQCAM
 
 config VIDEO_CQCAM
 	tristate "QuickCam Colour Video For Linux (EXPERIMENTAL)"
-	depends on EXPERIMENTAL && PARPORT && VIDEO_V4L1
+	depends on EXPERIMENTAL && PARPORT && VIDEO_V4L2
 	help
 	  This is the video4linux driver for the colour version of the
 	  Connectix QuickCam.  If you have one of these cameras, say Y here,
diff --git a/drivers/media/video/ak881x.c b/drivers/media/video/ak881x.c
index 35390d4717b9..1573392f74bd 100644
--- a/drivers/media/video/ak881x.c
+++ b/drivers/media/video/ak881x.c
@@ -11,6 +11,7 @@
 #include <linux/i2c.h>
 #include <linux/init.h>
 #include <linux/platform_device.h>
+#include <linux/slab.h>
 #include <linux/videodev2.h>
 
 #include <media/ak881x.h>
@@ -141,7 +142,7 @@ static int ak881x_s_mbus_fmt(struct v4l2_subdev *sd,
 	return ak881x_try_g_mbus_fmt(sd, mf);
 }
 
-static int ak881x_enum_mbus_fmt(struct v4l2_subdev *sd, int index,
+static int ak881x_enum_mbus_fmt(struct v4l2_subdev *sd, unsigned int index,
 				enum v4l2_mbus_pixelcode *code)
 {
 	if (index)
diff --git a/drivers/media/video/bw-qcam.c b/drivers/media/video/bw-qcam.c
index 3c9e754d73a0..935e0c9a9674 100644
--- a/drivers/media/video/bw-qcam.c
+++ b/drivers/media/video/bw-qcam.c
@@ -66,19 +66,58 @@ OTHER DEALINGS IN THE SOFTWARE.
 #include <linux/delay.h>
 #include <linux/errno.h>
 #include <linux/fs.h>
-#include <linux/init.h>
 #include <linux/kernel.h>
 #include <linux/slab.h>
 #include <linux/mm.h>
 #include <linux/parport.h>
 #include <linux/sched.h>
-#include <linux/videodev.h>
-#include <media/v4l2-common.h>
-#include <media/v4l2-ioctl.h>
+#include <linux/version.h>
+#include <linux/videodev2.h>
 #include <linux/mutex.h>
 #include <asm/uaccess.h>
-
-#include "bw-qcam.h"
+#include <media/v4l2-common.h>
+#include <media/v4l2-ioctl.h>
+#include <media/v4l2-device.h>
+
+/* One from column A... */
+#define QC_NOTSET 0
+#define QC_UNIDIR 1
+#define QC_BIDIR  2
+#define QC_SERIAL 3
+
+/* ... and one from column B */
+#define QC_ANY          0x00
+#define QC_FORCE_UNIDIR 0x10
+#define QC_FORCE_BIDIR  0x20
+#define QC_FORCE_SERIAL 0x30
+/* in the port_mode member */
+
+#define QC_MODE_MASK    0x07
+#define QC_FORCE_MASK   0x70
+
+#define MAX_HEIGHT 243
+#define MAX_WIDTH 336
+
+/* Bit fields for status flags */
+#define QC_PARAM_CHANGE	0x01 /* Camera status change has occurred */
+
+struct qcam {
+	struct v4l2_device v4l2_dev;
+	struct video_device vdev;
+	struct pardevice *pdev;
+	struct parport *pport;
+	struct mutex lock;
+	int width, height;
+	int bpp;
+	int mode;
+	int contrast, brightness, whitebal;
+	int port_mode;
+	int transfer_scale;
+	int top, left;
+	int status;
+	unsigned int saved_bits;
+	unsigned long in_use;
+};
 
 static unsigned int maxpoll = 250;   /* Maximum busy-loop count for qcam I/O */
 static unsigned int yieldlines = 4;  /* Yield after this many during capture */
@@ -93,22 +132,26 @@ module_param(video_nr, int, 0);
  * immediately attempt to initialize qcam */
 module_param(force_init, int, 0);
 
-static inline int read_lpstatus(struct qcam_device *q)
+#define MAX_CAMS 4
+static struct qcam *qcams[MAX_CAMS];
+static unsigned int num_cams;
+
+static inline int read_lpstatus(struct qcam *q)
 {
 	return parport_read_status(q->pport);
 }
 
-static inline int read_lpdata(struct qcam_device *q)
+static inline int read_lpdata(struct qcam *q)
 {
 	return parport_read_data(q->pport);
 }
 
-static inline void write_lpdata(struct qcam_device *q, int d)
+static inline void write_lpdata(struct qcam *q, int d)
 {
 	parport_write_data(q->pport, d);
 }
 
-static inline void write_lpcontrol(struct qcam_device *q, int d)
+static void write_lpcontrol(struct qcam *q, int d)
 {
 	if (d & 0x20) {
 		/* Set bidirectional mode to reverse (data in) */
@@ -124,126 +167,11 @@ static inline void write_lpcontrol(struct qcam_device *q, int d)
 	parport_write_control(q->pport, d);
 }
 
-static int qc_waithand(struct qcam_device *q, int val);
-static int qc_command(struct qcam_device *q, int command);
-static int qc_readparam(struct qcam_device *q);
-static int qc_setscanmode(struct qcam_device *q);
-static int qc_readbytes(struct qcam_device *q, char buffer[]);
-
-static struct video_device qcam_template;
-
-static int qc_calibrate(struct qcam_device *q)
-{
-	/*
-	 *	Bugfix by Hanno Mueller hmueller@kabel.de, Mai 21 96
-	 *	The white balance is an individiual value for each
-	 *	quickcam.
-	 */
-
-	int value;
-	int count = 0;
-
-	qc_command(q, 27);	/* AutoAdjustOffset */
-	qc_command(q, 0);	/* Dummy Parameter, ignored by the camera */
-
-	/* GetOffset (33) will read 255 until autocalibration */
-	/* is finished. After that, a value of 1-254 will be */
-	/* returned. */
-
-	do {
-		qc_command(q, 33);
-		value = qc_readparam(q);
-		mdelay(1);
-		schedule();
-		count++;
-	} while (value == 0xff && count < 2048);
-
-	q->whitebal = value;
-	return value;
-}
-
-/* Initialize the QuickCam driver control structure.  This is where
- * defaults are set for people who don't have a config file.*/
-
-static struct qcam_device *qcam_init(struct parport *port)
-{
-	struct qcam_device *q;
-
-	q = kmalloc(sizeof(struct qcam_device), GFP_KERNEL);
-	if (q == NULL)
-		return NULL;
-
-	q->pport = port;
-	q->pdev = parport_register_device(port, "bw-qcam", NULL, NULL,
-			NULL, 0, NULL);
-	if (q->pdev == NULL) {
-		printk(KERN_ERR "bw-qcam: couldn't register for %s.\n",
-				port->name);
-		kfree(q);
-		return NULL;
-	}
-
-	memcpy(&q->vdev, &qcam_template, sizeof(qcam_template));
-
-	mutex_init(&q->lock);
-
-	q->port_mode = (QC_ANY | QC_NOTSET);
-	q->width = 320;
-	q->height = 240;
-	q->bpp = 4;
-	q->transfer_scale = 2;
-	q->contrast = 192;
-	q->brightness = 180;
-	q->whitebal = 105;
-	q->top = 1;
-	q->left = 14;
-	q->mode = -1;
-	q->status = QC_PARAM_CHANGE;
-	return q;
-}
-
-
-/* qc_command is probably a bit of a misnomer -- it's used to send
- * bytes *to* the camera.  Generally, these bytes are either commands
- * or arguments to commands, so the name fits, but it still bugs me a
- * bit.  See the documentation for a list of commands. */
-
-static int qc_command(struct qcam_device *q, int command)
-{
-	int n1, n2;
-	int cmd;
-
-	write_lpdata(q, command);
-	write_lpcontrol(q, 6);
-
-	n1 = qc_waithand(q, 1);
-
-	write_lpcontrol(q, 0xe);
-	n2 = qc_waithand(q, 0);
-
-	cmd = (n1 & 0xf0) | ((n2 & 0xf0) >> 4);
-	return cmd;
-}
-
-static int qc_readparam(struct qcam_device *q)
-{
-	int n1, n2;
-	int cmd;
-
-	write_lpcontrol(q, 6);
-	n1 = qc_waithand(q, 1);
-
-	write_lpcontrol(q, 0xe);
-	n2 = qc_waithand(q, 0);
-
-	cmd = (n1 & 0xf0) | ((n2 & 0xf0) >> 4);
-	return cmd;
-}
 
 /* qc_waithand busy-waits for a handshake signal from the QuickCam.
  * Almost all communication with the camera requires handshaking. */
 
-static int qc_waithand(struct qcam_device *q, int val)
+static int qc_waithand(struct qcam *q, int val)
 {
 	int status;
 	int runs = 0;
@@ -286,7 +214,7 @@ static int qc_waithand(struct qcam_device *q, int val)
  * (bit 3 of status register).  It also returns the last value read,
  * since this data is useful. */
 
-static unsigned int qc_waithand2(struct qcam_device *q, int val)
+static unsigned int qc_waithand2(struct qcam *q, int val)
 {
 	unsigned int status;
 	int runs = 0;
@@ -309,6 +237,43 @@ static unsigned int qc_waithand2(struct qcam_device *q, int val)
 	return status;
 }
 
+/* qc_command is probably a bit of a misnomer -- it's used to send
+ * bytes *to* the camera.  Generally, these bytes are either commands
+ * or arguments to commands, so the name fits, but it still bugs me a
+ * bit.  See the documentation for a list of commands. */
+
+static int qc_command(struct qcam *q, int command)
+{
+	int n1, n2;
+	int cmd;
+
+	write_lpdata(q, command);
+	write_lpcontrol(q, 6);
+
+	n1 = qc_waithand(q, 1);
+
+	write_lpcontrol(q, 0xe);
+	n2 = qc_waithand(q, 0);
+
+	cmd = (n1 & 0xf0) | ((n2 & 0xf0) >> 4);
+	return cmd;
+}
+
+static int qc_readparam(struct qcam *q)
+{
+	int n1, n2;
+	int cmd;
+
+	write_lpcontrol(q, 6);
+	n1 = qc_waithand(q, 1);
+
+	write_lpcontrol(q, 0xe);
+	n2 = qc_waithand(q, 0);
+
+	cmd = (n1 & 0xf0) | ((n2 & 0xf0) >> 4);
+	return cmd;
+}
+
 
 /* Try to detect a QuickCam.  It appears to flash the upper 4 bits of
    the status register at 5-10 Hz.  This is only used in the autoprobe
@@ -317,7 +282,7 @@ static unsigned int qc_waithand2(struct qcam_device *q, int val)
    almost completely safe, while their method screws up my printer if
    I plug it in before the camera. */
 
-static int qc_detect(struct qcam_device *q)
+static int qc_detect(struct qcam *q)
 {
 	int reg, lastreg;
 	int count = 0;
@@ -358,41 +323,6 @@ static int qc_detect(struct qcam_device *q)
 	}
 }
 
-
-/* Reset the QuickCam.  This uses the same sequence the Windows
- * QuickPic program uses.  Someone with a bi-directional port should
- * check that bi-directional mode is detected right, and then
- * implement bi-directional mode in qc_readbyte(). */
-
-static void qc_reset(struct qcam_device *q)
-{
-	switch (q->port_mode & QC_FORCE_MASK) {
-	case QC_FORCE_UNIDIR:
-		q->port_mode = (q->port_mode & ~QC_MODE_MASK) | QC_UNIDIR;
-		break;
-
-	case QC_FORCE_BIDIR:
-		q->port_mode = (q->port_mode & ~QC_MODE_MASK) | QC_BIDIR;
-		break;
-
-	case QC_ANY:
-		write_lpcontrol(q, 0x20);
-		write_lpdata(q, 0x75);
-
-		if (read_lpdata(q) != 0x75)
-			q->port_mode = (q->port_mode & ~QC_MODE_MASK) | QC_BIDIR;
-		else
-			q->port_mode = (q->port_mode & ~QC_MODE_MASK) | QC_UNIDIR;
-		break;
-	}
-
-	write_lpcontrol(q, 0xb);
-	udelay(250);
-	write_lpcontrol(q, 0xe);
-	qc_setscanmode(q);		/* in case port_mode changed */
-}
-
-
 /* Decide which scan mode to use.  There's no real requirement that
  * the scanmode match the resolution in q->height and q-> width -- the
  * camera takes the picture at the resolution specified in the
@@ -402,7 +332,7 @@ static void qc_reset(struct qcam_device *q)
  * returned.  If the scan is smaller, then the rest of the image
  * returned contains garbage. */
 
-static int qc_setscanmode(struct qcam_device *q)
+static int qc_setscanmode(struct qcam *q)
 {
 	int old_mode = q->mode;
 
@@ -442,10 +372,45 @@ static int qc_setscanmode(struct qcam_device *q)
 }
 
 
+/* Reset the QuickCam.  This uses the same sequence the Windows
+ * QuickPic program uses.  Someone with a bi-directional port should
+ * check that bi-directional mode is detected right, and then
+ * implement bi-directional mode in qc_readbyte(). */
+
+static void qc_reset(struct qcam *q)
+{
+	switch (q->port_mode & QC_FORCE_MASK) {
+	case QC_FORCE_UNIDIR:
+		q->port_mode = (q->port_mode & ~QC_MODE_MASK) | QC_UNIDIR;
+		break;
+
+	case QC_FORCE_BIDIR:
+		q->port_mode = (q->port_mode & ~QC_MODE_MASK) | QC_BIDIR;
+		break;
+
+	case QC_ANY:
+		write_lpcontrol(q, 0x20);
+		write_lpdata(q, 0x75);
+
+		if (read_lpdata(q) != 0x75)
+			q->port_mode = (q->port_mode & ~QC_MODE_MASK) | QC_BIDIR;
+		else
+			q->port_mode = (q->port_mode & ~QC_MODE_MASK) | QC_UNIDIR;
+		break;
+	}
+
+	write_lpcontrol(q, 0xb);
+	udelay(250);
+	write_lpcontrol(q, 0xe);
+	qc_setscanmode(q);		/* in case port_mode changed */
+}
+
+
+
 /* Reset the QuickCam and program for brightness, contrast,
  * white-balance, and resolution. */
 
-static void qc_set(struct qcam_device *q)
+static void qc_set(struct qcam *q)
 {
 	int val;
 	int val2;
@@ -499,7 +464,7 @@ static void qc_set(struct qcam_device *q)
    the supplied buffer.  It returns the number of bytes read,
    or -1 on error. */
 
-static inline int qc_readbytes(struct qcam_device *q, char buffer[])
+static inline int qc_readbytes(struct qcam *q, char buffer[])
 {
 	int ret = 1;
 	unsigned int hi, lo;
@@ -590,7 +555,7 @@ static inline int qc_readbytes(struct qcam_device *q, char buffer[])
  * n=2^(bit depth)-1.  Ask me for more details if you don't understand
  * this. */
 
-static long qc_capture(struct qcam_device *q, char __user *buf, unsigned long len)
+static long qc_capture(struct qcam *q, char __user *buf, unsigned long len)
 {
 	int i, j, k, yield;
 	int bytes;
@@ -674,171 +639,206 @@ static long qc_capture(struct qcam_device *q, char __user *buf, unsigned long le
  *	Video4linux interfacing
  */
 
-static long qcam_do_ioctl(struct file *file, unsigned int cmd, void *arg)
+static int qcam_querycap(struct file *file, void  *priv,
+					struct v4l2_capability *vcap)
 {
-	struct video_device *dev = video_devdata(file);
-	struct qcam_device *qcam = (struct qcam_device *)dev;
-
-	switch (cmd) {
-	case VIDIOCGCAP:
-		{
-			struct video_capability *b = arg;
-			strcpy(b->name, "Quickcam");
-			b->type = VID_TYPE_CAPTURE|VID_TYPE_SCALES|VID_TYPE_MONOCHROME;
-			b->channels = 1;
-			b->audios = 0;
-			b->maxwidth = 320;
-			b->maxheight = 240;
-			b->minwidth = 80;
-			b->minheight = 60;
-			return 0;
-		}
-	case VIDIOCGCHAN:
-		{
-			struct video_channel *v = arg;
-			if (v->channel != 0)
-				return -EINVAL;
-			v->flags = 0;
-			v->tuners = 0;
-			/* Good question.. its composite or SVHS so.. */
-			v->type = VIDEO_TYPE_CAMERA;
-			strcpy(v->name, "Camera");
-			return 0;
-		}
-	case VIDIOCSCHAN:
-		{
-			struct video_channel *v = arg;
-			if (v->channel != 0)
-				return -EINVAL;
-			return 0;
-		}
-	case VIDIOCGTUNER:
-		{
-			struct video_tuner *v = arg;
-			if (v->tuner)
-				return -EINVAL;
-			strcpy(v->name, "Format");
-			v->rangelow = 0;
-			v->rangehigh = 0;
-			v->flags = 0;
-			v->mode = VIDEO_MODE_AUTO;
-			return 0;
-		}
-	case VIDIOCSTUNER:
-		{
-			struct video_tuner *v = arg;
-			if (v->tuner)
-				return -EINVAL;
-			if (v->mode != VIDEO_MODE_AUTO)
-				return -EINVAL;
-			return 0;
-		}
-	case VIDIOCGPICT:
-		{
-			struct video_picture *p = arg;
-			p->colour = 0x8000;
-			p->hue = 0x8000;
-			p->brightness = qcam->brightness << 8;
-			p->contrast = qcam->contrast << 8;
-			p->whiteness = qcam->whitebal << 8;
-			p->depth = qcam->bpp;
-			p->palette = VIDEO_PALETTE_GREY;
-			return 0;
-		}
-	case VIDIOCSPICT:
-		{
-			struct video_picture *p = arg;
-			if (p->palette != VIDEO_PALETTE_GREY)
-				return -EINVAL;
-			if (p->depth != 4 && p->depth != 6)
-				return -EINVAL;
-
-			/*
-			 *	Now load the camera.
-			 */
-
-			qcam->brightness = p->brightness >> 8;
-			qcam->contrast = p->contrast >> 8;
-			qcam->whitebal = p->whiteness >> 8;
-			qcam->bpp = p->depth;
-
-			mutex_lock(&qcam->lock);
-			qc_setscanmode(qcam);
-			mutex_unlock(&qcam->lock);
-			qcam->status |= QC_PARAM_CHANGE;
+	struct qcam *qcam = video_drvdata(file);
 
-			return 0;
-		}
-	case VIDIOCSWIN:
-		{
-			struct video_window *vw = arg;
-			if (vw->flags)
-				return -EINVAL;
-			if (vw->clipcount)
-				return -EINVAL;
-			if (vw->height < 60 || vw->height > 240)
-				return -EINVAL;
-			if (vw->width < 80 || vw->width > 320)
-				return -EINVAL;
-
-			qcam->width = 320;
-			qcam->height = 240;
-			qcam->transfer_scale = 4;
-
-			if (vw->width >= 160 && vw->height >= 120)
-				qcam->transfer_scale = 2;
-			if (vw->width >= 320 && vw->height >= 240) {
-				qcam->width = 320;
-				qcam->height = 240;
-				qcam->transfer_scale = 1;
-			}
-			mutex_lock(&qcam->lock);
-			qc_setscanmode(qcam);
-			mutex_unlock(&qcam->lock);
+	strlcpy(vcap->driver, qcam->v4l2_dev.name, sizeof(vcap->driver));
+	strlcpy(vcap->card, "B&W Quickcam", sizeof(vcap->card));
+	strlcpy(vcap->bus_info, "parport", sizeof(vcap->bus_info));
+	vcap->version = KERNEL_VERSION(0, 0, 2);
+	vcap->capabilities = V4L2_CAP_VIDEO_CAPTURE | V4L2_CAP_READWRITE;
+	return 0;
+}
 
-			/* We must update the camera before we grab. We could
-			   just have changed the grab size */
-			qcam->status |= QC_PARAM_CHANGE;
+static int qcam_enum_input(struct file *file, void *fh, struct v4l2_input *vin)
+{
+	if (vin->index > 0)
+		return -EINVAL;
+	strlcpy(vin->name, "Camera", sizeof(vin->name));
+	vin->type = V4L2_INPUT_TYPE_CAMERA;
+	vin->audioset = 0;
+	vin->tuner = 0;
+	vin->std = 0;
+	vin->status = 0;
+	return 0;
+}
 
-			/* Ok we figured out what to use from our wide choice */
-			return 0;
-		}
-	case VIDIOCGWIN:
-		{
-			struct video_window *vw = arg;
+static int qcam_g_input(struct file *file, void *fh, unsigned int *inp)
+{
+	*inp = 0;
+	return 0;
+}
 
-			memset(vw, 0, sizeof(*vw));
-			vw->width = qcam->width / qcam->transfer_scale;
-			vw->height = qcam->height / qcam->transfer_scale;
-			return 0;
-		}
-	case VIDIOCKEY:
-		return 0;
-	case VIDIOCCAPTURE:
-	case VIDIOCGFBUF:
-	case VIDIOCSFBUF:
-	case VIDIOCGFREQ:
-	case VIDIOCSFREQ:
-	case VIDIOCGAUDIO:
-	case VIDIOCSAUDIO:
-		return -EINVAL;
+static int qcam_s_input(struct file *file, void *fh, unsigned int inp)
+{
+	return (inp > 0) ? -EINVAL : 0;
+}
+
+static int qcam_queryctrl(struct file *file, void *priv,
+					struct v4l2_queryctrl *qc)
+{
+	switch (qc->id) {
+	case V4L2_CID_BRIGHTNESS:
+		return v4l2_ctrl_query_fill(qc, 0, 255, 1, 180);
+	case V4L2_CID_CONTRAST:
+		return v4l2_ctrl_query_fill(qc, 0, 255, 1, 192);
+	case V4L2_CID_GAMMA:
+		return v4l2_ctrl_query_fill(qc, 0, 255, 1, 105);
+	}
+	return -EINVAL;
+}
+
+static int qcam_g_ctrl(struct file *file, void *priv,
+					struct v4l2_control *ctrl)
+{
+	struct qcam *qcam = video_drvdata(file);
+	int ret = 0;
+
+	switch (ctrl->id) {
+	case V4L2_CID_BRIGHTNESS:
+		ctrl->value = qcam->brightness;
+		break;
+	case V4L2_CID_CONTRAST:
+		ctrl->value = qcam->contrast;
+		break;
+	case V4L2_CID_GAMMA:
+		ctrl->value = qcam->whitebal;
+		break;
 	default:
-		return -ENOIOCTLCMD;
+		ret = -EINVAL;
+		break;
 	}
+	return ret;
+}
+
+static int qcam_s_ctrl(struct file *file, void *priv,
+					struct v4l2_control *ctrl)
+{
+	struct qcam *qcam = video_drvdata(file);
+	int ret = 0;
+
+	mutex_lock(&qcam->lock);
+	switch (ctrl->id) {
+	case V4L2_CID_BRIGHTNESS:
+		qcam->brightness = ctrl->value;
+		break;
+	case V4L2_CID_CONTRAST:
+		qcam->contrast = ctrl->value;
+		break;
+	case V4L2_CID_GAMMA:
+		qcam->whitebal = ctrl->value;
+		break;
+	default:
+		ret = -EINVAL;
+		break;
+	}
+	if (ret == 0) {
+		qc_setscanmode(qcam);
+		qcam->status |= QC_PARAM_CHANGE;
+	}
+	mutex_unlock(&qcam->lock);
+	return ret;
+}
+
+static int qcam_g_fmt_vid_cap(struct file *file, void *fh, struct v4l2_format *fmt)
+{
+	struct qcam *qcam = video_drvdata(file);
+	struct v4l2_pix_format *pix = &fmt->fmt.pix;
+
+	pix->width = qcam->width / qcam->transfer_scale;
+	pix->height = qcam->height / qcam->transfer_scale;
+	pix->pixelformat = (qcam->bpp == 4) ? V4L2_PIX_FMT_Y4 : V4L2_PIX_FMT_Y6;
+	pix->field = V4L2_FIELD_NONE;
+	pix->bytesperline = qcam->width;
+	pix->sizeimage = qcam->width * qcam->height;
+	/* Just a guess */
+	pix->colorspace = V4L2_COLORSPACE_SRGB;
+	return 0;
+}
+
+static int qcam_try_fmt_vid_cap(struct file *file, void *fh, struct v4l2_format *fmt)
+{
+	struct v4l2_pix_format *pix = &fmt->fmt.pix;
+
+	if (pix->height <= 60 || pix->width <= 80) {
+		pix->height = 60;
+		pix->width = 80;
+	} else if (pix->height <= 120 || pix->width <= 160) {
+		pix->height = 120;
+		pix->width = 160;
+	} else {
+		pix->height = 240;
+		pix->width = 320;
+	}
+	if (pix->pixelformat != V4L2_PIX_FMT_Y4 &&
+	    pix->pixelformat != V4L2_PIX_FMT_Y6)
+		pix->pixelformat = V4L2_PIX_FMT_Y4;
+	pix->field = V4L2_FIELD_NONE;
+	pix->bytesperline = pix->width;
+	pix->sizeimage = pix->width * pix->height;
+	/* Just a guess */
+	pix->colorspace = V4L2_COLORSPACE_SRGB;
+	return 0;
+}
+
+static int qcam_s_fmt_vid_cap(struct file *file, void *fh, struct v4l2_format *fmt)
+{
+	struct qcam *qcam = video_drvdata(file);
+	struct v4l2_pix_format *pix = &fmt->fmt.pix;
+	int ret = qcam_try_fmt_vid_cap(file, fh, fmt);
+
+	if (ret)
+		return ret;
+	qcam->width = 320;
+	qcam->height = 240;
+	if (pix->height == 60)
+		qcam->transfer_scale = 4;
+	else if (pix->height == 120)
+		qcam->transfer_scale = 2;
+	else
+		qcam->transfer_scale = 1;
+	if (pix->pixelformat == V4L2_PIX_FMT_Y6)
+		qcam->bpp = 6;
+	else
+		qcam->bpp = 4;
+
+	mutex_lock(&qcam->lock);
+	qc_setscanmode(qcam);
+	/* We must update the camera before we grab. We could
+	   just have changed the grab size */
+	qcam->status |= QC_PARAM_CHANGE;
+	mutex_unlock(&qcam->lock);
 	return 0;
 }
 
-static long qcam_ioctl(struct file *file,
-		unsigned int cmd, unsigned long arg)
+static int qcam_enum_fmt_vid_cap(struct file *file, void *fh, struct v4l2_fmtdesc *fmt)
 {
-	return video_usercopy(file, cmd, arg, qcam_do_ioctl);
+	static struct v4l2_fmtdesc formats[] = {
+		{ 0, 0, 0,
+		  "4-Bit Monochrome", V4L2_PIX_FMT_Y4,
+		  { 0, 0, 0, 0 }
+		},
+		{ 0, 0, 0,
+		  "6-Bit Monochrome", V4L2_PIX_FMT_Y6,
+		  { 0, 0, 0, 0 }
+		},
+	};
+	enum v4l2_buf_type type = fmt->type;
+
+	if (fmt->index > 1)
+		return -EINVAL;
+
+	*fmt = formats[fmt->index];
+	fmt->type = type;
+	return 0;
 }
 
 static ssize_t qcam_read(struct file *file, char __user *buf,
 		size_t count, loff_t *ppos)
 {
-	struct video_device *v = video_devdata(file);
-	struct qcam_device *qcam = (struct qcam_device *)v;
+	struct qcam *qcam = video_drvdata(file);
 	int len;
 	parport_claim_or_block(qcam->pdev);
 
@@ -858,43 +858,112 @@ static ssize_t qcam_read(struct file *file, char __user *buf,
 	return len;
 }
 
-static int qcam_exclusive_open(struct file *file)
+static const struct v4l2_file_operations qcam_fops = {
+	.owner		= THIS_MODULE,
+	.ioctl          = video_ioctl2,
+	.read		= qcam_read,
+};
+
+static const struct v4l2_ioctl_ops qcam_ioctl_ops = {
+	.vidioc_querycap    		    = qcam_querycap,
+	.vidioc_g_input      		    = qcam_g_input,
+	.vidioc_s_input      		    = qcam_s_input,
+	.vidioc_enum_input   		    = qcam_enum_input,
+	.vidioc_queryctrl 		    = qcam_queryctrl,
+	.vidioc_g_ctrl  		    = qcam_g_ctrl,
+	.vidioc_s_ctrl 			    = qcam_s_ctrl,
+	.vidioc_enum_fmt_vid_cap 	    = qcam_enum_fmt_vid_cap,
+	.vidioc_g_fmt_vid_cap 		    = qcam_g_fmt_vid_cap,
+	.vidioc_s_fmt_vid_cap  		    = qcam_s_fmt_vid_cap,
+	.vidioc_try_fmt_vid_cap  	    = qcam_try_fmt_vid_cap,
+};
+
+/* Initialize the QuickCam driver control structure.  This is where
+ * defaults are set for people who don't have a config file.*/
+
+static struct qcam *qcam_init(struct parport *port)
 {
-	struct video_device *dev = video_devdata(file);
-	struct qcam_device *qcam = (struct qcam_device *)dev;
+	struct qcam *qcam;
+	struct v4l2_device *v4l2_dev;
+
+	qcam = kzalloc(sizeof(struct qcam), GFP_KERNEL);
+	if (qcam == NULL)
+		return NULL;
+
+	v4l2_dev = &qcam->v4l2_dev;
+	strlcpy(v4l2_dev->name, "bw-qcam", sizeof(v4l2_dev->name));
 
-	return test_and_set_bit(0, &qcam->in_use) ? -EBUSY : 0;
+	if (v4l2_device_register(NULL, v4l2_dev) < 0) {
+		v4l2_err(v4l2_dev, "Could not register v4l2_device\n");
+		return NULL;
+	}
+
+	qcam->pport = port;
+	qcam->pdev = parport_register_device(port, "bw-qcam", NULL, NULL,
+			NULL, 0, NULL);
+	if (qcam->pdev == NULL) {
+		v4l2_err(v4l2_dev, "couldn't register for %s.\n", port->name);
+		kfree(qcam);
+		return NULL;
+	}
+
+	strlcpy(qcam->vdev.name, "Connectix QuickCam", sizeof(qcam->vdev.name));
+	qcam->vdev.v4l2_dev = v4l2_dev;
+	qcam->vdev.fops = &qcam_fops;
+	qcam->vdev.ioctl_ops = &qcam_ioctl_ops;
+	qcam->vdev.release = video_device_release_empty;
+	video_set_drvdata(&qcam->vdev, qcam);
+
+	mutex_init(&qcam->lock);
+
+	qcam->port_mode = (QC_ANY | QC_NOTSET);
+	qcam->width = 320;
+	qcam->height = 240;
+	qcam->bpp = 4;
+	qcam->transfer_scale = 2;
+	qcam->contrast = 192;
+	qcam->brightness = 180;
+	qcam->whitebal = 105;
+	qcam->top = 1;
+	qcam->left = 14;
+	qcam->mode = -1;
+	qcam->status = QC_PARAM_CHANGE;
+	return qcam;
 }
 
-static int qcam_exclusive_release(struct file *file)
+static int qc_calibrate(struct qcam *q)
 {
-	struct video_device *dev = video_devdata(file);
-	struct qcam_device *qcam = (struct qcam_device *)dev;
+	/*
+	 *	Bugfix by Hanno Mueller hmueller@kabel.de, Mai 21 96
+	 *	The white balance is an individual value for each
+	 *	quickcam.
+	 */
 
-	clear_bit(0, &qcam->in_use);
-	return 0;
-}
+	int value;
+	int count = 0;
 
-static const struct v4l2_file_operations qcam_fops = {
-	.owner		= THIS_MODULE,
-	.open           = qcam_exclusive_open,
-	.release        = qcam_exclusive_release,
-	.ioctl          = qcam_ioctl,
-	.read		= qcam_read,
-};
-static struct video_device qcam_template = {
-	.name		= "Connectix Quickcam",
-	.fops           = &qcam_fops,
-	.release 	= video_device_release_empty,
-};
+	qc_command(q, 27);	/* AutoAdjustOffset */
+	qc_command(q, 0);	/* Dummy Parameter, ignored by the camera */
 
-#define MAX_CAMS 4
-static struct qcam_device *qcams[MAX_CAMS];
-static unsigned int num_cams;
+	/* GetOffset (33) will read 255 until autocalibration */
+	/* is finished. After that, a value of 1-254 will be */
+	/* returned. */
+
+	do {
+		qc_command(q, 33);
+		value = qc_readparam(q);
+		mdelay(1);
+		schedule();
+		count++;
+	} while (value == 0xff && count < 2048);
+
+	q->whitebal = value;
+	return value;
+}
 
 static int init_bwqcam(struct parport *port)
 {
-	struct qcam_device *qcam;
+	struct qcam *qcam;
 
 	if (num_cams == MAX_CAMS) {
 		printk(KERN_ERR "Too many Quickcams (max %d)\n", MAX_CAMS);
@@ -919,7 +988,7 @@ static int init_bwqcam(struct parport *port)
 
 	parport_release(qcam->pdev);
 
-	printk(KERN_INFO "Connectix Quickcam on %s\n", qcam->pport->name);
+	v4l2_info(&qcam->v4l2_dev, "Connectix Quickcam on %s\n", qcam->pport->name);
 
 	if (video_register_device(&qcam->vdev, VFL_TYPE_GRABBER, video_nr) < 0) {
 		parport_unregister_device(qcam->pdev);
@@ -932,7 +1001,7 @@ static int init_bwqcam(struct parport *port)
 	return 0;
 }
 
-static void close_bwqcam(struct qcam_device *qcam)
+static void close_bwqcam(struct qcam *qcam)
 {
 	video_unregister_device(&qcam->vdev);
 	parport_unregister_device(qcam->pdev);
@@ -983,7 +1052,7 @@ static void bwqcam_detach(struct parport *port)
 {
 	int i;
 	for (i = 0; i < num_cams; i++) {
-		struct qcam_device *qcam = qcams[i];
+		struct qcam *qcam = qcams[i];
 		if (qcam && qcam->pdev->port == port) {
 			qcams[i] = NULL;
 			close_bwqcam(qcam);
diff --git a/drivers/media/video/bw-qcam.h b/drivers/media/video/bw-qcam.h
deleted file mode 100644
index 8a60c5de0935..000000000000
--- a/drivers/media/video/bw-qcam.h
+++ /dev/null
@@ -1,69 +0,0 @@
-/*
- *	Video4Linux bw-qcam driver
- *
- *	Derived from code..
- */
-
-/******************************************************************
-
-Copyright (C) 1996 by Scott Laird
-
-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 SCOTT LAIRD 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.
-
-******************************************************************/
-
-/* One from column A... */
-#define QC_NOTSET 0
-#define QC_UNIDIR 1
-#define QC_BIDIR  2
-#define QC_SERIAL 3
-
-/* ... and one from column B */
-#define QC_ANY          0x00
-#define QC_FORCE_UNIDIR 0x10
-#define QC_FORCE_BIDIR  0x20
-#define QC_FORCE_SERIAL 0x30
-/* in the port_mode member */
-
-#define QC_MODE_MASK    0x07
-#define QC_FORCE_MASK   0x70
-
-#define MAX_HEIGHT 243
-#define MAX_WIDTH 336
-
-/* Bit fields for status flags */
-#define QC_PARAM_CHANGE	0x01 /* Camera status change has occurred */
-
-struct qcam_device {
-	struct video_device vdev;
-	struct pardevice *pdev;
-	struct parport *pport;
-	struct mutex lock;
-	int width, height;
-	int bpp;
-	int mode;
-	int contrast, brightness, whitebal;
-	int port_mode;
-	int transfer_scale;
-	int top, left;
-	int status;
-	unsigned int saved_bits;
-	unsigned long in_use;
-};
diff --git a/drivers/media/video/c-qcam.c b/drivers/media/video/c-qcam.c
index 8f1dd88b32a6..6e4b19698c13 100644
--- a/drivers/media/video/c-qcam.c
+++ b/drivers/media/video/c-qcam.c
@@ -33,15 +33,17 @@
 #include <linux/mm.h>
 #include <linux/parport.h>
 #include <linux/sched.h>
-#include <linux/videodev.h>
-#include <media/v4l2-common.h>
-#include <media/v4l2-ioctl.h>
 #include <linux/mutex.h>
 #include <linux/jiffies.h>
-
+#include <linux/version.h>
+#include <linux/videodev2.h>
 #include <asm/uaccess.h>
+#include <media/v4l2-device.h>
+#include <media/v4l2-common.h>
+#include <media/v4l2-ioctl.h>
 
-struct qcam_device {
+struct qcam {
+	struct v4l2_device v4l2_dev;
 	struct video_device vdev;
 	struct pardevice *pdev;
 	struct parport *pport;
@@ -51,7 +53,6 @@ struct qcam_device {
 	int contrast, brightness, whitebal;
 	int top, left;
 	unsigned int bidirectional;
-	unsigned long in_use;
 	struct mutex lock;
 };
 
@@ -68,33 +69,45 @@ struct qcam_device {
 #define QC_DECIMATION_2		2
 #define QC_DECIMATION_4		4
 
-#define BANNER "Colour QuickCam for Video4Linux v0.05"
+#define BANNER "Colour QuickCam for Video4Linux v0.06"
 
 static int parport[MAX_CAMS] = { [1 ... MAX_CAMS-1] = -1 };
 static int probe = 2;
 static int force_rgb;
 static int video_nr = -1;
 
-static inline void qcam_set_ack(struct qcam_device *qcam, unsigned int i)
+/* FIXME: parport=auto would never have worked, surely? --RR */
+MODULE_PARM_DESC(parport, "parport=<auto|n[,n]...> for port detection method\n"
+			  "probe=<0|1|2> for camera detection method\n"
+			  "force_rgb=<0|1> for RGB data format (default BGR)");
+module_param_array(parport, int, NULL, 0);
+module_param(probe, int, 0);
+module_param(force_rgb, bool, 0);
+module_param(video_nr, int, 0);
+
+static struct qcam *qcams[MAX_CAMS];
+static unsigned int num_cams;
+
+static inline void qcam_set_ack(struct qcam *qcam, unsigned int i)
 {
 	/* note: the QC specs refer to the PCAck pin by voltage, not
 	   software level.  PC ports have builtin inverters. */
 	parport_frob_control(qcam->pport, 8, i ? 8 : 0);
 }
 
-static inline unsigned int qcam_ready1(struct qcam_device *qcam)
+static inline unsigned int qcam_ready1(struct qcam *qcam)
 {
 	return (parport_read_status(qcam->pport) & 0x8) ? 1 : 0;
 }
 
-static inline unsigned int qcam_ready2(struct qcam_device *qcam)
+static inline unsigned int qcam_ready2(struct qcam *qcam)
 {
 	return (parport_read_data(qcam->pport) & 0x1) ? 1 : 0;
 }
 
-static unsigned int qcam_await_ready1(struct qcam_device *qcam,
-					     int value)
+static unsigned int qcam_await_ready1(struct qcam *qcam, int value)
 {
+	struct v4l2_device *v4l2_dev = &qcam->v4l2_dev;
 	unsigned long oldjiffies = jiffies;
 	unsigned int i;
 
@@ -112,14 +125,15 @@ static unsigned int qcam_await_ready1(struct qcam_device *qcam,
 	}
 
 	/* Probably somebody pulled the plug out.  Not much we can do. */
-	printk(KERN_ERR "c-qcam: ready1 timeout (%d) %x %x\n", value,
+	v4l2_err(v4l2_dev, "ready1 timeout (%d) %x %x\n", value,
 	       parport_read_status(qcam->pport),
 	       parport_read_control(qcam->pport));
 	return 1;
 }
 
-static unsigned int qcam_await_ready2(struct qcam_device *qcam, int value)
+static unsigned int qcam_await_ready2(struct qcam *qcam, int value)
 {
+	struct v4l2_device *v4l2_dev = &qcam->v4l2_dev;
 	unsigned long oldjiffies = jiffies;
 	unsigned int i;
 
@@ -137,14 +151,14 @@ static unsigned int qcam_await_ready2(struct qcam_device *qcam, int value)
 	}
 
 	/* Probably somebody pulled the plug out.  Not much we can do. */
-	printk(KERN_ERR "c-qcam: ready2 timeout (%d) %x %x %x\n", value,
+	v4l2_err(v4l2_dev, "ready2 timeout (%d) %x %x %x\n", value,
 	       parport_read_status(qcam->pport),
 	       parport_read_control(qcam->pport),
 	       parport_read_data(qcam->pport));
 	return 1;
 }
 
-static int qcam_read_data(struct qcam_device *qcam)
+static int qcam_read_data(struct qcam *qcam)
 {
 	unsigned int idata;
 
@@ -159,21 +173,22 @@ static int qcam_read_data(struct qcam_device *qcam)
 	return idata;
 }
 
-static int qcam_write_data(struct qcam_device *qcam, unsigned int data)
+static int qcam_write_data(struct qcam *qcam, unsigned int data)
 {
+	struct v4l2_device *v4l2_dev = &qcam->v4l2_dev;
 	unsigned int idata;
 
 	parport_write_data(qcam->pport, data);
 	idata = qcam_read_data(qcam);
 	if (data != idata) {
-		printk(KERN_WARNING "cqcam: sent %x but received %x\n", data,
+		v4l2_warn(v4l2_dev, "sent %x but received %x\n", data,
 		       idata);
 		return 1;
 	}
 	return 0;
 }
 
-static inline int qcam_set(struct qcam_device *qcam, unsigned int cmd, unsigned int data)
+static inline int qcam_set(struct qcam *qcam, unsigned int cmd, unsigned int data)
 {
 	if (qcam_write_data(qcam, cmd))
 		return -1;
@@ -182,14 +197,14 @@ static inline int qcam_set(struct qcam_device *qcam, unsigned int cmd, unsigned
 	return 0;
 }
 
-static inline int qcam_get(struct qcam_device *qcam, unsigned int cmd)
+static inline int qcam_get(struct qcam *qcam, unsigned int cmd)
 {
 	if (qcam_write_data(qcam, cmd))
 		return -1;
 	return qcam_read_data(qcam);
 }
 
-static int qc_detect(struct qcam_device *qcam)
+static int qc_detect(struct qcam *qcam)
 {
 	unsigned int stat, ostat, i, count = 0;
 
@@ -246,7 +261,7 @@ static int qc_detect(struct qcam_device *qcam)
 	return 0;
 }
 
-static void qc_reset(struct qcam_device *qcam)
+static void qc_reset(struct qcam *qcam)
 {
 	parport_write_control(qcam->pport, 0xc);
 	parport_write_control(qcam->pport, 0x8);
@@ -258,55 +273,55 @@ static void qc_reset(struct qcam_device *qcam)
 /* Reset the QuickCam and program for brightness, contrast,
  * white-balance, and resolution. */
 
-static void qc_setup(struct qcam_device *q)
+static void qc_setup(struct qcam *qcam)
 {
-	qc_reset(q);
+	qc_reset(qcam);
 
 	/* Set the brightness. */
-	qcam_set(q, 11, q->brightness);
+	qcam_set(qcam, 11, qcam->brightness);
 
 	/* Set the height and width.  These refer to the actual
 	   CCD area *before* applying the selected decimation.  */
-	qcam_set(q, 17, q->ccd_height);
-	qcam_set(q, 19, q->ccd_width / 2);
+	qcam_set(qcam, 17, qcam->ccd_height);
+	qcam_set(qcam, 19, qcam->ccd_width / 2);
 
 	/* Set top and left.  */
-	qcam_set(q, 0xd, q->top);
-	qcam_set(q, 0xf, q->left);
+	qcam_set(qcam, 0xd, qcam->top);
+	qcam_set(qcam, 0xf, qcam->left);
 
 	/* Set contrast and white balance.  */
-	qcam_set(q, 0x19, q->contrast);
-	qcam_set(q, 0x1f, q->whitebal);
+	qcam_set(qcam, 0x19, qcam->contrast);
+	qcam_set(qcam, 0x1f, qcam->whitebal);
 
 	/* Set the speed.  */
-	qcam_set(q, 45, 2);
+	qcam_set(qcam, 45, 2);
 }
 
 /* Read some bytes from the camera and put them in the buffer.
    nbytes should be a multiple of 3, because bidirectional mode gives
    us three bytes at a time.  */
 
-static unsigned int qcam_read_bytes(struct qcam_device *q, unsigned char *buf, unsigned int nbytes)
+static unsigned int qcam_read_bytes(struct qcam *qcam, unsigned char *buf, unsigned int nbytes)
 {
 	unsigned int bytes = 0;
 
-	qcam_set_ack(q, 0);
-	if (q->bidirectional) {
+	qcam_set_ack(qcam, 0);
+	if (qcam->bidirectional) {
 		/* It's a bidirectional port */
 		while (bytes < nbytes) {
 			unsigned int lo1, hi1, lo2, hi2;
 			unsigned char r, g, b;
 
-			if (qcam_await_ready2(q, 1))
+			if (qcam_await_ready2(qcam, 1))
 				return bytes;
-			lo1 = parport_read_data(q->pport) >> 1;
-			hi1 = ((parport_read_status(q->pport) >> 3) & 0x1f) ^ 0x10;
-			qcam_set_ack(q, 1);
-			if (qcam_await_ready2(q, 0))
+			lo1 = parport_read_data(qcam->pport) >> 1;
+			hi1 = ((parport_read_status(qcam->pport) >> 3) & 0x1f) ^ 0x10;
+			qcam_set_ack(qcam, 1);
+			if (qcam_await_ready2(qcam, 0))
 				return bytes;
-			lo2 = parport_read_data(q->pport) >> 1;
-			hi2 = ((parport_read_status(q->pport) >> 3) & 0x1f) ^ 0x10;
-			qcam_set_ack(q, 0);
+			lo2 = parport_read_data(qcam->pport) >> 1;
+			hi2 = ((parport_read_status(qcam->pport) >> 3) & 0x1f) ^ 0x10;
+			qcam_set_ack(qcam, 0);
 			r = lo1 | ((hi1 & 1) << 7);
 			g = ((hi1 & 0x1e) << 3) | ((hi2 & 0x1e) >> 1);
 			b = lo2 | ((hi2 & 1) << 7);
@@ -328,14 +343,14 @@ static unsigned int qcam_read_bytes(struct qcam_device *q, unsigned char *buf, u
 		while (bytes < nbytes) {
 			unsigned int hi, lo;
 
-			if (qcam_await_ready1(q, 1))
+			if (qcam_await_ready1(qcam, 1))
 				return bytes;
-			hi = (parport_read_status(q->pport) & 0xf0);
-			qcam_set_ack(q, 1);
-			if (qcam_await_ready1(q, 0))
+			hi = (parport_read_status(qcam->pport) & 0xf0);
+			qcam_set_ack(qcam, 1);
+			if (qcam_await_ready1(qcam, 0))
 				return bytes;
-			lo = (parport_read_status(q->pport) & 0xf0);
-			qcam_set_ack(q, 0);
+			lo = (parport_read_status(qcam->pport) & 0xf0);
+			qcam_set_ack(qcam, 0);
 			/* flip some bits */
 			rgb[(i = bytes++ % 3)] = (hi | (lo >> 4)) ^ 0x88;
 			if (i >= 2) {
@@ -361,10 +376,11 @@ get_fragment:
 
 #define BUFSZ	150
 
-static long qc_capture(struct qcam_device *q, char __user *buf, unsigned long len)
+static long qc_capture(struct qcam *qcam, char __user *buf, unsigned long len)
 {
+	struct v4l2_device *v4l2_dev = &qcam->v4l2_dev;
 	unsigned lines, pixelsperline, bitsperxfer;
-	unsigned int is_bi_dir = q->bidirectional;
+	unsigned int is_bi_dir = qcam->bidirectional;
 	size_t wantlen, outptr = 0;
 	char tmpbuf[BUFSZ];
 
@@ -373,10 +389,10 @@ static long qc_capture(struct qcam_device *q, char __user *buf, unsigned long le
 
 	/* Wait for camera to become ready */
 	for (;;) {
-		int i = qcam_get(q, 41);
+		int i = qcam_get(qcam, 41);
 
 		if (i == -1) {
-			qc_setup(q);
+			qc_setup(qcam);
 			return -EIO;
 		}
 		if ((i & 0x80) == 0)
@@ -384,25 +400,25 @@ static long qc_capture(struct qcam_device *q, char __user *buf, unsigned long le
 		schedule();
 	}
 
-	if (qcam_set(q, 7, (q->mode | (is_bi_dir ? 1 : 0)) + 1))
+	if (qcam_set(qcam, 7, (qcam->mode | (is_bi_dir ? 1 : 0)) + 1))
 		return -EIO;
 
-	lines = q->height;
-	pixelsperline = q->width;
+	lines = qcam->height;
+	pixelsperline = qcam->width;
 	bitsperxfer = (is_bi_dir) ? 24 : 8;
 
 	if (is_bi_dir) {
 		/* Turn the port around */
-		parport_data_reverse(q->pport);
+		parport_data_reverse(qcam->pport);
 		mdelay(3);
-		qcam_set_ack(q, 0);
-		if (qcam_await_ready1(q, 1)) {
-			qc_setup(q);
+		qcam_set_ack(qcam, 0);
+		if (qcam_await_ready1(qcam, 1)) {
+			qc_setup(qcam);
 			return -EIO;
 		}
-		qcam_set_ack(q, 1);
-		if (qcam_await_ready1(q, 0)) {
-			qc_setup(q);
+		qcam_set_ack(qcam, 1);
+		if (qcam_await_ready1(qcam, 0)) {
+			qc_setup(qcam);
 			return -EIO;
 		}
 	}
@@ -413,7 +429,7 @@ static long qc_capture(struct qcam_device *q, char __user *buf, unsigned long le
 		size_t t, s;
 
 		s = (wantlen > BUFSZ) ? BUFSZ : wantlen;
-		t = qcam_read_bytes(q, tmpbuf, s);
+		t = qcam_read_bytes(qcam, tmpbuf, s);
 		if (outptr < len) {
 			size_t sz = len - outptr;
 
@@ -432,10 +448,10 @@ static long qc_capture(struct qcam_device *q, char __user *buf, unsigned long le
 	len = outptr;
 
 	if (wantlen) {
-		printk(KERN_ERR "qcam: short read.\n");
+		v4l2_err(v4l2_dev, "short read.\n");
 		if (is_bi_dir)
-			parport_data_forward(q->pport);
-		qc_setup(q);
+			parport_data_forward(qcam->pport);
+		qc_setup(qcam);
 		return len;
 	}
 
@@ -443,49 +459,49 @@ static long qc_capture(struct qcam_device *q, char __user *buf, unsigned long le
 		int l;
 
 		do {
-			l = qcam_read_bytes(q, tmpbuf, 3);
+			l = qcam_read_bytes(qcam, tmpbuf, 3);
 			cond_resched();
 		} while (l && (tmpbuf[0] == 0x7e || tmpbuf[1] == 0x7e || tmpbuf[2] == 0x7e));
 		if (force_rgb) {
 			if (tmpbuf[0] != 0xe || tmpbuf[1] != 0x0 || tmpbuf[2] != 0xf)
-				printk(KERN_ERR "qcam: bad EOF\n");
+				v4l2_err(v4l2_dev, "bad EOF\n");
 		} else {
 			if (tmpbuf[0] != 0xf || tmpbuf[1] != 0x0 || tmpbuf[2] != 0xe)
-				printk(KERN_ERR "qcam: bad EOF\n");
+				v4l2_err(v4l2_dev, "bad EOF\n");
 		}
-		qcam_set_ack(q, 0);
-		if (qcam_await_ready1(q, 1)) {
-			printk(KERN_ERR "qcam: no ack after EOF\n");
-			parport_data_forward(q->pport);
-			qc_setup(q);
+		qcam_set_ack(qcam, 0);
+		if (qcam_await_ready1(qcam, 1)) {
+			v4l2_err(v4l2_dev, "no ack after EOF\n");
+			parport_data_forward(qcam->pport);
+			qc_setup(qcam);
 			return len;
 		}
-		parport_data_forward(q->pport);
+		parport_data_forward(qcam->pport);
 		mdelay(3);
-		qcam_set_ack(q, 1);
-		if (qcam_await_ready1(q, 0)) {
-			printk(KERN_ERR "qcam: no ack to port turnaround\n");
-			qc_setup(q);
+		qcam_set_ack(qcam, 1);
+		if (qcam_await_ready1(qcam, 0)) {
+			v4l2_err(v4l2_dev, "no ack to port turnaround\n");
+			qc_setup(qcam);
 			return len;
 		}
 	} else {
 		int l;
 
 		do {
-			l = qcam_read_bytes(q, tmpbuf, 1);
+			l = qcam_read_bytes(qcam, tmpbuf, 1);
 			cond_resched();
 		} while (l && tmpbuf[0] == 0x7e);
-		l = qcam_read_bytes(q, tmpbuf + 1, 2);
+		l = qcam_read_bytes(qcam, tmpbuf + 1, 2);
 		if (force_rgb) {
 			if (tmpbuf[0] != 0xe || tmpbuf[1] != 0x0 || tmpbuf[2] != 0xf)
-				printk(KERN_ERR "qcam: bad EOF\n");
+				v4l2_err(v4l2_dev, "bad EOF\n");
 		} else {
 			if (tmpbuf[0] != 0xf || tmpbuf[1] != 0x0 || tmpbuf[2] != 0xe)
-				printk(KERN_ERR "qcam: bad EOF\n");
+				v4l2_err(v4l2_dev, "bad EOF\n");
 		}
 	}
 
-	qcam_write_data(q, 0);
+	qcam_write_data(qcam, 0);
 	return len;
 }
 
@@ -493,184 +509,202 @@ static long qc_capture(struct qcam_device *q, char __user *buf, unsigned long le
  *	Video4linux interfacing
  */
 
-static long qcam_do_ioctl(struct file *file, unsigned int cmd, void *arg)
+static int qcam_querycap(struct file *file, void  *priv,
+					struct v4l2_capability *vcap)
 {
-	struct video_device *dev = video_devdata(file);
-	struct qcam_device *qcam = (struct qcam_device *)dev;
+	struct qcam *qcam = video_drvdata(file);
 
-	switch (cmd) {
-	case VIDIOCGCAP:
-	{
-		struct video_capability *b = arg;
+	strlcpy(vcap->driver, qcam->v4l2_dev.name, sizeof(vcap->driver));
+	strlcpy(vcap->card, "Color Quickcam", sizeof(vcap->card));
+	strlcpy(vcap->bus_info, "parport", sizeof(vcap->bus_info));
+	vcap->version = KERNEL_VERSION(0, 0, 3);
+	vcap->capabilities = V4L2_CAP_VIDEO_CAPTURE | V4L2_CAP_READWRITE;
+	return 0;
+}
 
-		strcpy(b->name, "Quickcam");
-		b->type = VID_TYPE_CAPTURE | VID_TYPE_SCALES;
-		b->channels = 1;
-		b->audios = 0;
-		b->maxwidth = 320;
-		b->maxheight = 240;
-		b->minwidth = 80;
-		b->minheight = 60;
-		return 0;
-	}
-	case VIDIOCGCHAN:
-	{
-		struct video_channel *v = arg;
-
-		if (v->channel != 0)
-			return -EINVAL;
-		v->flags = 0;
-		v->tuners = 0;
-		/* Good question.. its composite or SVHS so.. */
-		v->type = VIDEO_TYPE_CAMERA;
-		strcpy(v->name, "Camera");
-		return 0;
-	}
-	case VIDIOCSCHAN:
-	{
-		struct video_channel *v = arg;
+static int qcam_enum_input(struct file *file, void *fh, struct v4l2_input *vin)
+{
+	if (vin->index > 0)
+		return -EINVAL;
+	strlcpy(vin->name, "Camera", sizeof(vin->name));
+	vin->type = V4L2_INPUT_TYPE_CAMERA;
+	vin->audioset = 0;
+	vin->tuner = 0;
+	vin->std = 0;
+	vin->status = 0;
+	return 0;
+}
 
-		if (v->channel != 0)
-			return -EINVAL;
-		return 0;
-	}
-	case VIDIOCGTUNER:
-	{
-		struct video_tuner *v = arg;
-
-		if (v->tuner)
-			return -EINVAL;
-		memset(v, 0, sizeof(*v));
-		strcpy(v->name, "Format");
-		v->mode = VIDEO_MODE_AUTO;
-		return 0;
+static int qcam_g_input(struct file *file, void *fh, unsigned int *inp)
+{
+	*inp = 0;
+	return 0;
+}
+
+static int qcam_s_input(struct file *file, void *fh, unsigned int inp)
+{
+	return (inp > 0) ? -EINVAL : 0;
+}
+
+static int qcam_queryctrl(struct file *file, void *priv,
+					struct v4l2_queryctrl *qc)
+{
+	switch (qc->id) {
+	case V4L2_CID_BRIGHTNESS:
+		return v4l2_ctrl_query_fill(qc, 0, 255, 1, 240);
+	case V4L2_CID_CONTRAST:
+		return v4l2_ctrl_query_fill(qc, 0, 255, 1, 192);
+	case V4L2_CID_GAMMA:
+		return v4l2_ctrl_query_fill(qc, 0, 255, 1, 128);
 	}
-	case VIDIOCSTUNER:
-	{
-		struct video_tuner *v = arg;
-
-		if (v->tuner)
-			return -EINVAL;
-		if (v->mode != VIDEO_MODE_AUTO)
-			return -EINVAL;
-		return 0;
+	return -EINVAL;
+}
+
+static int qcam_g_ctrl(struct file *file, void *priv,
+					struct v4l2_control *ctrl)
+{
+	struct qcam *qcam = video_drvdata(file);
+	int ret = 0;
+
+	switch (ctrl->id) {
+	case V4L2_CID_BRIGHTNESS:
+		ctrl->value = qcam->brightness;
+		break;
+	case V4L2_CID_CONTRAST:
+		ctrl->value = qcam->contrast;
+		break;
+	case V4L2_CID_GAMMA:
+		ctrl->value = qcam->whitebal;
+		break;
+	default:
+		ret = -EINVAL;
+		break;
 	}
-	case VIDIOCGPICT:
-	{
-		struct video_picture *p = arg;
-
-		p->colour = 0x8000;
-		p->hue = 0x8000;
-		p->brightness = qcam->brightness << 8;
-		p->contrast = qcam->contrast << 8;
-		p->whiteness = qcam->whitebal << 8;
-		p->depth = 24;
-		p->palette = VIDEO_PALETTE_RGB24;
-		return 0;
+	return ret;
+}
+
+static int qcam_s_ctrl(struct file *file, void *priv,
+					struct v4l2_control *ctrl)
+{
+	struct qcam *qcam = video_drvdata(file);
+	int ret = 0;
+
+	mutex_lock(&qcam->lock);
+	switch (ctrl->id) {
+	case V4L2_CID_BRIGHTNESS:
+		qcam->brightness = ctrl->value;
+		break;
+	case V4L2_CID_CONTRAST:
+		qcam->contrast = ctrl->value;
+		break;
+	case V4L2_CID_GAMMA:
+		qcam->whitebal = ctrl->value;
+		break;
+	default:
+		ret = -EINVAL;
+		break;
 	}
-	case VIDIOCSPICT:
-	{
-		struct video_picture *p = arg;
-
-		/*
-		 *	Sanity check args
-		 */
-		if (p->depth != 24 || p->palette != VIDEO_PALETTE_RGB24)
-			return -EINVAL;
-
-		/*
-		 *	Now load the camera.
-		 */
-		qcam->brightness = p->brightness >> 8;
-		qcam->contrast = p->contrast >> 8;
-		qcam->whitebal = p->whiteness >> 8;
-
-		mutex_lock(&qcam->lock);
+	if (ret == 0) {
 		parport_claim_or_block(qcam->pdev);
 		qc_setup(qcam);
 		parport_release(qcam->pdev);
-		mutex_unlock(&qcam->lock);
-		return 0;
 	}
-	case VIDIOCSWIN:
-	{
-		struct video_window *vw = arg;
-
-		if (vw->flags)
-			return -EINVAL;
-		if (vw->clipcount)
-			return -EINVAL;
-		if (vw->height < 60 || vw->height > 240)
-			return -EINVAL;
-		if (vw->width < 80 || vw->width > 320)
-			return -EINVAL;
-
-		qcam->width = 80;
-		qcam->height = 60;
-		qcam->mode = QC_DECIMATION_4;
+	mutex_unlock(&qcam->lock);
+	return ret;
+}
 
-		if (vw->width >= 160 && vw->height >= 120) {
-			qcam->width = 160;
-			qcam->height = 120;
-			qcam->mode = QC_DECIMATION_2;
-		}
-		if (vw->width >= 320 && vw->height >= 240) {
-			qcam->width = 320;
-			qcam->height = 240;
-			qcam->mode = QC_DECIMATION_1;
-		}
-		qcam->mode |= QC_MILLIONS;
-#if 0
-		if (vw->width >= 640 && vw->height >= 480) {
-			qcam->width = 640;
-			qcam->height = 480;
-			qcam->mode = QC_BILLIONS | QC_DECIMATION_1;
-		}
-#endif
-		/* Ok we figured out what to use from our
-		   wide choice */
-		mutex_lock(&qcam->lock);
-		parport_claim_or_block(qcam->pdev);
-		qc_setup(qcam);
-		parport_release(qcam->pdev);
-		mutex_unlock(&qcam->lock);
-		return 0;
-	}
-	case VIDIOCGWIN:
-	{
-		struct video_window *vw = arg;
-		memset(vw, 0, sizeof(*vw));
-		vw->width = qcam->width;
-		vw->height = qcam->height;
-		return 0;
+static int qcam_g_fmt_vid_cap(struct file *file, void *fh, struct v4l2_format *fmt)
+{
+	struct qcam *qcam = video_drvdata(file);
+	struct v4l2_pix_format *pix = &fmt->fmt.pix;
+
+	pix->width = qcam->width;
+	pix->height = qcam->height;
+	pix->pixelformat = V4L2_PIX_FMT_RGB24;
+	pix->field = V4L2_FIELD_NONE;
+	pix->bytesperline = 3 * qcam->width;
+	pix->sizeimage = 3 * qcam->width * qcam->height;
+	/* Just a guess */
+	pix->colorspace = V4L2_COLORSPACE_SRGB;
+	return 0;
+}
+
+static int qcam_try_fmt_vid_cap(struct file *file, void *fh, struct v4l2_format *fmt)
+{
+	struct v4l2_pix_format *pix = &fmt->fmt.pix;
+
+	if (pix->height < 60 || pix->width < 80) {
+		pix->height = 60;
+		pix->width = 80;
+	} else if (pix->height < 120 || pix->width < 160) {
+		pix->height = 120;
+		pix->width = 160;
+	} else {
+		pix->height = 240;
+		pix->width = 320;
 	}
-	case VIDIOCKEY:
-		return 0;
-	case VIDIOCCAPTURE:
-	case VIDIOCGFBUF:
-	case VIDIOCSFBUF:
-	case VIDIOCGFREQ:
-	case VIDIOCSFREQ:
-	case VIDIOCGAUDIO:
-	case VIDIOCSAUDIO:
-		return -EINVAL;
+	pix->pixelformat = V4L2_PIX_FMT_RGB24;
+	pix->field = V4L2_FIELD_NONE;
+	pix->bytesperline = 3 * pix->width;
+	pix->sizeimage = 3 * pix->width * pix->height;
+	/* Just a guess */
+	pix->colorspace = V4L2_COLORSPACE_SRGB;
+	return 0;
+}
+
+static int qcam_s_fmt_vid_cap(struct file *file, void *fh, struct v4l2_format *fmt)
+{
+	struct qcam *qcam = video_drvdata(file);
+	struct v4l2_pix_format *pix = &fmt->fmt.pix;
+	int ret = qcam_try_fmt_vid_cap(file, fh, fmt);
+
+	if (ret)
+		return ret;
+	switch (pix->height) {
+	case 60:
+		qcam->mode = QC_DECIMATION_4;
+		break;
+	case 120:
+		qcam->mode = QC_DECIMATION_2;
+		break;
 	default:
-		return -ENOIOCTLCMD;
+		qcam->mode = QC_DECIMATION_1;
+		break;
 	}
+
+	mutex_lock(&qcam->lock);
+	qcam->mode |= QC_MILLIONS;
+	qcam->height = pix->height;
+	qcam->width = pix->width;
+	parport_claim_or_block(qcam->pdev);
+	qc_setup(qcam);
+	parport_release(qcam->pdev);
+	mutex_unlock(&qcam->lock);
 	return 0;
 }
 
-static long qcam_ioctl(struct file *file,
-		      unsigned int cmd, unsigned long arg)
+static int qcam_enum_fmt_vid_cap(struct file *file, void *fh, struct v4l2_fmtdesc *fmt)
 {
-	return video_usercopy(file, cmd, arg, qcam_do_ioctl);
+	static struct v4l2_fmtdesc formats[] = {
+		{ 0, 0, 0,
+		  "RGB 8:8:8", V4L2_PIX_FMT_RGB24,
+		  { 0, 0, 0, 0 }
+		},
+	};
+	enum v4l2_buf_type type = fmt->type;
+
+	if (fmt->index > 0)
+		return -EINVAL;
+
+	*fmt = formats[fmt->index];
+	fmt->type = type;
+	return 0;
 }
 
 static ssize_t qcam_read(struct file *file, char __user *buf,
 			 size_t count, loff_t *ppos)
 {
-	struct video_device *v = video_devdata(file);
-	struct qcam_device *qcam = (struct qcam_device *)v;
+	struct qcam *qcam = video_drvdata(file);
 	int len;
 
 	mutex_lock(&qcam->lock);
@@ -682,81 +716,80 @@ static ssize_t qcam_read(struct file *file, char __user *buf,
 	return len;
 }
 
-static int qcam_exclusive_open(struct file *file)
-{
-	struct video_device *dev = video_devdata(file);
-	struct qcam_device *qcam = (struct qcam_device *)dev;
-
-	return test_and_set_bit(0, &qcam->in_use) ? -EBUSY : 0;
-}
-
-static int qcam_exclusive_release(struct file *file)
-{
-	struct video_device *dev = video_devdata(file);
-	struct qcam_device *qcam = (struct qcam_device *)dev;
-
-	clear_bit(0, &qcam->in_use);
-	return 0;
-}
-
-/* video device template */
 static const struct v4l2_file_operations qcam_fops = {
 	.owner		= THIS_MODULE,
-	.open           = qcam_exclusive_open,
-	.release        = qcam_exclusive_release,
-	.ioctl          = qcam_ioctl,
+	.ioctl		= video_ioctl2,
 	.read		= qcam_read,
 };
 
-static struct video_device qcam_template = {
-	.name		= "Colour QuickCam",
-	.fops           = &qcam_fops,
-	.release 	= video_device_release_empty,
+static const struct v4l2_ioctl_ops qcam_ioctl_ops = {
+	.vidioc_querycap    		    = qcam_querycap,
+	.vidioc_g_input      		    = qcam_g_input,
+	.vidioc_s_input      		    = qcam_s_input,
+	.vidioc_enum_input   		    = qcam_enum_input,
+	.vidioc_queryctrl 		    = qcam_queryctrl,
+	.vidioc_g_ctrl  		    = qcam_g_ctrl,
+	.vidioc_s_ctrl 			    = qcam_s_ctrl,
+	.vidioc_enum_fmt_vid_cap 	    = qcam_enum_fmt_vid_cap,
+	.vidioc_g_fmt_vid_cap 		    = qcam_g_fmt_vid_cap,
+	.vidioc_s_fmt_vid_cap  		    = qcam_s_fmt_vid_cap,
+	.vidioc_try_fmt_vid_cap  	    = qcam_try_fmt_vid_cap,
 };
 
 /* Initialize the QuickCam driver control structure. */
 
-static struct qcam_device *qcam_init(struct parport *port)
+static struct qcam *qcam_init(struct parport *port)
 {
-	struct qcam_device *q;
+	struct qcam *qcam;
+	struct v4l2_device *v4l2_dev;
 
-	q = kmalloc(sizeof(struct qcam_device), GFP_KERNEL);
-	if (q == NULL)
+	qcam = kzalloc(sizeof(*qcam), GFP_KERNEL);
+	if (qcam == NULL)
 		return NULL;
 
-	q->pport = port;
-	q->pdev = parport_register_device(port, "c-qcam", NULL, NULL,
+	v4l2_dev = &qcam->v4l2_dev;
+	strlcpy(v4l2_dev->name, "c-qcam", sizeof(v4l2_dev->name));
+
+	if (v4l2_device_register(NULL, v4l2_dev) < 0) {
+		v4l2_err(v4l2_dev, "Could not register v4l2_device\n");
+		return NULL;
+	}
+
+	qcam->pport = port;
+	qcam->pdev = parport_register_device(port, "c-qcam", NULL, NULL,
 					  NULL, 0, NULL);
 
-	q->bidirectional = (q->pport->modes & PARPORT_MODE_TRISTATE) ? 1 : 0;
+	qcam->bidirectional = (qcam->pport->modes & PARPORT_MODE_TRISTATE) ? 1 : 0;
 
-	if (q->pdev == NULL) {
-		printk(KERN_ERR "c-qcam: couldn't register for %s.\n",
-		       port->name);
-		kfree(q);
+	if (qcam->pdev == NULL) {
+		v4l2_err(v4l2_dev, "couldn't register for %s.\n", port->name);
+		kfree(qcam);
 		return NULL;
 	}
 
-	memcpy(&q->vdev, &qcam_template, sizeof(qcam_template));
-
-	mutex_init(&q->lock);
-	q->width = q->ccd_width = 320;
-	q->height = q->ccd_height = 240;
-	q->mode = QC_MILLIONS | QC_DECIMATION_1;
-	q->contrast = 192;
-	q->brightness = 240;
-	q->whitebal = 128;
-	q->top = 1;
-	q->left = 14;
-	return q;
+	strlcpy(qcam->vdev.name, "Colour QuickCam", sizeof(qcam->vdev.name));
+	qcam->vdev.v4l2_dev = v4l2_dev;
+	qcam->vdev.fops = &qcam_fops;
+	qcam->vdev.ioctl_ops = &qcam_ioctl_ops;
+	qcam->vdev.release = video_device_release_empty;
+	video_set_drvdata(&qcam->vdev, qcam);
+
+	mutex_init(&qcam->lock);
+	qcam->width = qcam->ccd_width = 320;
+	qcam->height = qcam->ccd_height = 240;
+	qcam->mode = QC_MILLIONS | QC_DECIMATION_1;
+	qcam->contrast = 192;
+	qcam->brightness = 240;
+	qcam->whitebal = 128;
+	qcam->top = 1;
+	qcam->left = 14;
+	return qcam;
 }
 
-static struct qcam_device *qcams[MAX_CAMS];
-static unsigned int num_cams;
-
 static int init_cqcam(struct parport *port)
 {
-	struct qcam_device *qcam;
+	struct qcam *qcam;
+	struct v4l2_device *v4l2_dev;
 
 	if (parport[0] != -1) {
 		/* The user gave specific instructions */
@@ -777,6 +810,8 @@ static int init_cqcam(struct parport *port)
 	if (qcam == NULL)
 		return -ENODEV;
 
+	v4l2_dev = &qcam->v4l2_dev;
+
 	parport_claim_or_block(qcam->pdev);
 
 	qc_reset(qcam);
@@ -793,14 +828,14 @@ static int init_cqcam(struct parport *port)
 	parport_release(qcam->pdev);
 
 	if (video_register_device(&qcam->vdev, VFL_TYPE_GRABBER, video_nr) < 0) {
-		printk(KERN_ERR "Unable to register Colour QuickCam on %s\n",
+		v4l2_err(v4l2_dev, "Unable to register Colour QuickCam on %s\n",
 		       qcam->pport->name);
 		parport_unregister_device(qcam->pdev);
 		kfree(qcam);
 		return -ENODEV;
 	}
 
-	printk(KERN_INFO "%s: Colour QuickCam found on %s\n",
+	v4l2_info(v4l2_dev, "%s: Colour QuickCam found on %s\n",
 	       video_device_node_name(&qcam->vdev), qcam->pport->name);
 
 	qcams[num_cams++] = qcam;
@@ -808,7 +843,7 @@ static int init_cqcam(struct parport *port)
 	return 0;
 }
 
-static void close_cqcam(struct qcam_device *qcam)
+static void close_cqcam(struct qcam *qcam)
 {
 	video_unregister_device(&qcam->vdev);
 	parport_unregister_device(qcam->pdev);
@@ -833,7 +868,7 @@ static struct parport_driver cqcam_driver = {
 
 static int __init cqcam_init(void)
 {
-	printk(BANNER "\n");
+	printk(KERN_INFO BANNER "\n");
 
 	return parport_register_driver(&cqcam_driver);
 }
@@ -852,14 +887,5 @@ MODULE_AUTHOR("Philip Blundell <philb@gnu.org>");
 MODULE_DESCRIPTION(BANNER);
 MODULE_LICENSE("GPL");
 
-/* FIXME: parport=auto would never have worked, surely? --RR */
-MODULE_PARM_DESC(parport, "parport=<auto|n[,n]...> for port detection method\n"
-			  "probe=<0|1|2> for camera detection method\n"
-			  "force_rgb=<0|1> for RGB data format (default BGR)");
-module_param_array(parport, int, NULL, 0);
-module_param(probe, int, 0);
-module_param(force_rgb, bool, 0);
-module_param(video_nr, int, 0);
-
 module_init(cqcam_init);
 module_exit(cqcam_cleanup);
diff --git a/drivers/media/video/cx18/cx18-alsa-main.c b/drivers/media/video/cx18/cx18-alsa-main.c
index b5d7cbf4528a..d50d69da387b 100644
--- a/drivers/media/video/cx18/cx18-alsa-main.c
+++ b/drivers/media/video/cx18/cx18-alsa-main.c
@@ -1,7 +1,7 @@
 /*
  *  ALSA interface to cx18 PCM capture streams
  *
- *  Copyright (C) 2009  Andy Walls <awalls@radix.net>
+ *  Copyright (C) 2009  Andy Walls <awalls@md.metrocast.net>
  *  Copyright (C) 2009  Devin Heitmueller <dheitmueller@kernellabs.com>
  *
  *  Portions of this work were sponsored by ONELAN Limited.
diff --git a/drivers/media/video/cx18/cx18-alsa-mixer.c b/drivers/media/video/cx18/cx18-alsa-mixer.c
index ef21114309fe..341bddc00b77 100644
--- a/drivers/media/video/cx18/cx18-alsa-mixer.c
+++ b/drivers/media/video/cx18/cx18-alsa-mixer.c
@@ -2,7 +2,7 @@
  *  ALSA mixer controls for the
  *  ALSA interface to cx18 PCM capture streams
  *
- *  Copyright (C) 2009  Andy Walls <awalls@radix.net>
+ *  Copyright (C) 2009  Andy Walls <awalls@md.metrocast.net>
  *
  *  This program is free software; you can redistribute it and/or modify
  *  it under the terms of the GNU General Public License as published by
diff --git a/drivers/media/video/cx18/cx18-alsa-mixer.h b/drivers/media/video/cx18/cx18-alsa-mixer.h
index 2d418db000fe..ec9238793f6f 100644
--- a/drivers/media/video/cx18/cx18-alsa-mixer.h
+++ b/drivers/media/video/cx18/cx18-alsa-mixer.h
@@ -2,7 +2,7 @@
  *  ALSA mixer controls for the
  *  ALSA interface to cx18 PCM capture streams
  *
- *  Copyright (C) 2009  Andy Walls <awalls@radix.net>
+ *  Copyright (C) 2009  Andy Walls <awalls@md.metrocast.net>
  *
  *  This program is free software; you can redistribute it and/or modify
  *  it under the terms of the GNU General Public License as published by
diff --git a/drivers/media/video/cx18/cx18-alsa-pcm.c b/drivers/media/video/cx18/cx18-alsa-pcm.c
index 2bd312daeb1e..8f55692db36d 100644
--- a/drivers/media/video/cx18/cx18-alsa-pcm.c
+++ b/drivers/media/video/cx18/cx18-alsa-pcm.c
@@ -2,7 +2,7 @@
  *  ALSA PCM device for the
  *  ALSA interface to cx18 PCM capture streams
  *
- *  Copyright (C) 2009  Andy Walls <awalls@radix.net>
+ *  Copyright (C) 2009  Andy Walls <awalls@md.metrocast.net>
  *  Copyright (C) 2009  Devin Heitmueller <dheitmueller@kernellabs.com>
  *
  *  Portions of this work were sponsored by ONELAN Limited.
diff --git a/drivers/media/video/cx18/cx18-alsa-pcm.h b/drivers/media/video/cx18/cx18-alsa-pcm.h

path: root/tests/virtio-blk-test.c

/*
 * QTest testcase for VirtIO Block Device
 *
 * Copyright (c) 2014 SUSE LINUX Products GmbH
 * Copyright (c) 2014 Marc Marí
 *
 * This work is licensed under the terms of the GNU GPL, version 2 or later.
 * See the COPYING file in the top-level directory.
 */

#include "qemu/osdep.h"
#include "libqtest.h"
#include "libqos/libqos-pc.h"
#include "libqos/libqos-spapr.h"
#include "libqos/virtio.h"
#include "libqos/virtio-pci.h"
#include "libqos/virtio-mmio.h"
#include "libqos/malloc-generic.h"
#include "qapi/qmp/qdict.h"
#include "qemu/bswap.h"
#include "standard-headers/linux/virtio_ids.h"
#include "standard-headers/linux/virtio_config.h"
#include "standard-headers/linux/virtio_ring.h"
#include "standard-headers/linux/virtio_blk.h"
#include "standard-headers/linux/virtio_pci.h"

/* TODO actually test the results and get rid of this */
#define qmp_discard_response(...) qobject_unref(qmp(__VA_ARGS__))

#define TEST_IMAGE_SIZE         (64 * 1024 * 1024)
#define QVIRTIO_BLK_TIMEOUT_US  (30 * 1000 * 1000)
#define PCI_SLOT_HP             0x06
#define PCI_SLOT                0x04
#define PCI_FN                  0x00

#define MMIO_PAGE_SIZE          4096
#define MMIO_DEV_BASE_ADDR      0x0A003E00
#define MMIO_RAM_ADDR           0x40000000
#define MMIO_RAM_SIZE           0x20000000

typedef struct QVirtioBlkReq {
    uint32_t type;
    uint32_t ioprio;
    uint64_t sector;
    char *data;
    uint8_t status;
} QVirtioBlkReq;

static char *drive_create(void)
{
    int fd, ret;
    char *tmp_path = g_strdup("/tmp/qtest.XXXXXX");

    /* Create a temporary raw image */
    fd = mkstemp(tmp_path);
    g_assert_cmpint(fd, >=, 0);
    ret = ftruncate(fd, TEST_IMAGE_SIZE);
    g_assert_cmpint(ret, ==, 0);
    close(fd);

    return tmp_path;
}

static QOSState *pci_test_start(void)
{
    QOSState *qs;
    const char *arch = qtest_get_arch();
    char *tmp_path;
    const char *cmd = "-drive if=none,id=drive0,file=%s,format=raw "
                      "-drive if=none,id=drive1,file=null-co://,format=raw "
                      "-device virtio-blk-pci,id=drv0,drive=drive0,"
                      "addr=%x.%x";

    tmp_path = drive_create();

    if (strcmp(arch, "i386") == 0 || strcmp(arch, "x86_64") == 0) {
        qs = qtest_pc_boot(cmd, tmp_path, PCI_SLOT, PCI_FN);
    } else if (strcmp(arch, "ppc64") == 0) {
        qs = qtest_spapr_boot(cmd, tmp_path, PCI_SLOT, PCI_FN);
    } else {
        g_printerr("virtio-blk tests are only available on x86 or ppc64\n");
        exit(EXIT_FAILURE);
    }
    global_qtest = qs->qts;
    unlink(tmp_path);
    g_free(tmp_path);
    return qs;
}

static void arm_test_start(void)
{
    char *tmp_path;

    tmp_path = drive_create();

    global_qtest = qtest_initf("-machine virt "
                               "-drive if=none,id=drive0,file=%s,format=raw "
                               "-device virtio-blk-device,drive=drive0",
                               tmp_path);
    unlink(tmp_path);
    g_free(tmp_path);
}

static void test_end(void)
{
    qtest_end();
}

static QVirtioPCIDevice *virtio_blk_pci_init(QPCIBus *bus, int slot)
{
    QVirtioPCIDevice *dev;

    dev = qvirtio_pci_device_find_slot(bus, VIRTIO_ID_BLOCK, slot);
    g_assert(dev != NULL);
    g_assert_cmphex(dev->vdev.device_type, ==, VIRTIO_ID_BLOCK);
    g_assert_cmphex(dev->pdev->devfn, ==, ((slot << 3) | PCI_FN));

    qvirtio_pci_device_enable(dev);
    qvirtio_reset(&dev->vdev);
    qvirtio_set_acknowledge(&dev->vdev);
    qvirtio_set_driver(&dev->vdev);

    return dev;
}

static inline void virtio_blk_fix_request(QVirtioDevice *d, QVirtioBlkReq *req)
{
#ifdef HOST_WORDS_BIGENDIAN
    const bool host_is_big_endian = true;
#else
    const bool host_is_big_endian = false;
#endif

    if (qvirtio_is_big_endian(d) != host_is_big_endian) {
        req->type = bswap32(req->type);
        req->ioprio = bswap32(req->ioprio);
        req->sector = bswap64(req->sector);
    }
}

static uint64_t virtio_blk_request(QGuestAllocator *alloc, QVirtioDevice *d,
                                   QVirtioBlkReq *req, uint64_t data_size)
{
    uint64_t addr;
    uint8_t status = 0xFF;

    g_assert_cmpuint(data_size % 512, ==, 0);
    addr = guest_alloc(alloc, sizeof(*req) + data_size);

    virtio_blk_fix_request(d, req);

    memwrite(addr, req, 16);
    memwrite(addr + 16, req->data, data_size);
    memwrite(addr + 16 + data_size, &status, sizeof(status));

    return addr;
}

static void test_basic(QVirtioDevice *dev, QGuestAllocator *alloc,
                       QVirtQueue *vq)
{
    QVirtioBlkReq req;
    uint64_t req_addr;
    uint64_t capacity;
    uint32_t features;
    uint32_t free_head;
    uint8_t status;
    char *data;

    capacity = qvirtio_config_readq(dev, 0);

    g_assert_cmpint(capacity, ==, TEST_IMAGE_SIZE / 512);

    features = qvirtio_get_features(dev);
    features = features & ~(QVIRTIO_F_BAD_FEATURE |
                    (1u << VIRTIO_RING_F_INDIRECT_DESC) |
                    (1u << VIRTIO_RING_F_EVENT_IDX) |
                    (1u << VIRTIO_BLK_F_SCSI));
    qvirtio_set_features(dev, features);

    qvirtio_set_driver_ok(dev);

    /* Write and read with 3 descriptor layout */
    /* Write request */
    req.type = VIRTIO_BLK_T_OUT;
    req.ioprio = 1;
    req.sector = 0;
    req.data = g_malloc0(512);
    strcpy(req.data, "TEST");

    req_addr = virtio_blk_request(alloc, dev, &req, 512);

    g_free(req.data);

    free_head = qvirtqueue_add(vq, req_addr, 16, false, true);
    qvirtqueue_add(vq, req_addr + 16, 512, false, true);
    qvirtqueue_add(vq, req_addr + 528, 1, true, false);

    qvirtqueue_kick(dev, vq, free_head);

    qvirtio_wait_used_elem(dev, vq, free_head, NULL, QVIRTIO_BLK_TIMEOUT_US);
    status = readb(req_addr + 528);
    g_assert_cmpint(status, ==, 0);

    guest_free(alloc, req_addr);

    /* Read request */
    req.type = VIRTIO_BLK_T_IN;
    req.ioprio = 1;
    req.sector = 0;
    req.data = g_malloc0(512);

    req_addr = virtio_blk_request(alloc, dev, &req, 512);

    g_free(req.data);

    free_head = qvirtqueue_add(vq, req_addr, 16, false, true);
    qvirtqueue_add(vq, req_addr + 16, 512, true, true);
    qvirtqueue_add(vq, req_addr + 528, 1, true, false);

    qvirtqueue_kick(dev, vq, free_head);

    qvirtio_wait_used_elem(dev, vq, free_head, NULL, QVIRTIO_BLK_TIMEOUT_US);
    status = readb(req_addr + 528);
    g_assert_cmpint(status, ==, 0);

    data = g_malloc0(512);
    memread(req_addr + 16, data, 512);
    g_assert_cmpstr(data, ==, "TEST");
    g_free(data);

    guest_free(alloc, req_addr);

    if (features & (1u << VIRTIO_F_ANY_LAYOUT)) {
        /* Write and read with 2 descriptor layout */
        /* Write request */
        req.type = VIRTIO_BLK_T_OUT;
        req.ioprio = 1;
        req.sector = 1;
        req.data = g_malloc0(512);
        strcpy(req.data, "TEST");

        req_addr = virtio_blk_request(alloc, dev, &req, 512);

        g_free(req.data);

        free_head = qvirtqueue_add(vq, req_addr, 528, false, true);
        qvirtqueue_add(vq, req_addr + 528, 1, true, false);
        qvirtqueue_kick(dev, vq, free_head);

        qvirtio_wait_used_elem(dev, vq, free_head, NULL,
                               QVIRTIO_BLK_TIMEOUT_US);
        status = readb(req_addr + 528);
        g_assert_cmpint(status, ==, 0);

        guest_free(alloc, req_addr);

        /* Read request */
        req.type = VIRTIO_BLK_T_IN;
        req.ioprio = 1;
        req.sector = 1;
        req.data = g_malloc0(512);

        req_addr = virtio_blk_request(alloc, dev, &req, 512);

        g_free(req.data);

        free_head = qvirtqueue_add(vq, req_addr, 16, false, true);
        qvirtqueue_add(vq, req_addr + 16, 513, true, false);

        qvirtqueue_kick(dev, vq, free_head);

        qvirtio_wait_used_elem(dev, vq, free_head, NULL,
                               QVIRTIO_BLK_TIMEOUT_US);
        status = readb(req_addr + 528);
        g_assert_cmpint(status, ==, 0);

        data = g_malloc0(512);
        memread(req_addr + 16, data, 512);
        g_assert_cmpstr(data, ==, "TEST");
        g_free(data);

        guest_free(alloc, req_addr);
    }
}

static void pci_basic(void)
{
    QVirtioPCIDevice *dev;
    QOSState *qs;
    QVirtQueuePCI *vqpci;

    qs = pci_test_start();
    dev = virtio_blk_pci_init(qs->pcibus, PCI_SLOT);

    vqpci = (QVirtQueuePCI *)qvirtqueue_setup(&dev->vdev, qs->alloc, 0);

    test_basic(&dev->vdev, qs->alloc, &vqpci->vq);

    /* End test */
    qvirtqueue_cleanup(dev->vdev.bus, &vqpci->vq, qs->alloc);
    qvirtio_pci_device_disable(dev);
    qvirtio_pci_device_free(dev);
    qtest_shutdown(qs);
}

static void pci_indirect(void)
{
    QVirtioPCIDevice *dev;
    QVirtQueuePCI *vqpci;
    QOSState *qs;
    QVirtioBlkReq req;
    QVRingIndirectDesc *indirect;
    uint64_t req_addr;
    uint64_t capacity;
    uint32_t features;
    uint32_t free_head;
    uint8_t status;
    char *data;

    qs = pci_test_start();

    dev = virtio_blk_pci_init(qs->pcibus, PCI_SLOT);

    capacity = qvirtio_config_readq(&dev->vdev, 0);
    g_assert_cmpint(capacity, ==, TEST_IMAGE_SIZE / 512);

    features = qvirtio_get_features(&dev->vdev);
    g_assert_cmphex(features & (1u << VIRTIO_RING_F_INDIRECT_DESC), !=, 0);
    features = features & ~(QVIRTIO_F_BAD_FEATURE |
                            (1u << VIRTIO_RING_F_EVENT_IDX) |
                            (1u << VIRTIO_BLK_F_SCSI));
    qvirtio_set_features(&dev->vdev, features);

    vqpci = (QVirtQueuePCI *)qvirtqueue_setup(&dev->vdev, qs->alloc, 0);
    qvirtio_set_driver_ok(&dev->vdev);

    /* Write request */
    req.type = VIRTIO_BLK_T_OUT;
    req.ioprio = 1;
    req.sector = 0;
    req.data = g_malloc0(512);
    strcpy(req.data, "TEST");

    req_addr = virtio_blk_request(qs->alloc, &dev->vdev, &req, 512);

    g_free(req.data);

    indirect = qvring_indirect_desc_setup(&dev->vdev, qs->alloc, 2);
    qvring_indirect_desc_add(indirect, req_addr, 528, false);
    qvring_indirect_desc_add(indirect, req_addr + 528, 1, true);
    free_head = qvirtqueue_add_indirect(&vqpci->vq, indirect);
    qvirtqueue_kick(&dev->vdev, &vqpci->vq, free_head);

    qvirtio_wait_used_elem(&dev->vdev, &vqpci->vq, free_head, NULL,
                           QVIRTIO_BLK_TIMEOUT_US);
    status = readb(req_addr + 528);
    g_assert_cmpint(status, ==, 0);

    g_free(indirect);
    guest_free(qs->alloc, req_addr);

    /* Read request */
    req.type = VIRTIO_BLK_T_IN;
    req.ioprio = 1;
    req.sector = 0;
    req.data = g_malloc0(512);
    strcpy(req.data, "TEST");

    req_addr = virtio_blk_request(qs->alloc, &dev->vdev, &req, 512);

    g_free(req.data);

    indirect = qvring_indirect_desc_setup(&dev->vdev, qs->alloc, 2);
    qvring_indirect_desc_add(indirect, req_addr, 16, false);
    qvring_indirect_desc_add(indirect, req_addr + 16, 513, true);
    free_head = qvirtqueue_add_indirect(&vqpci->vq, indirect);
    qvirtqueue_kick(&dev->vdev, &vqpci->vq, free_head);

    qvirtio_wait_used_elem(&dev->vdev, &vqpci->vq, free_head, NULL,
                           QVIRTIO_BLK_TIMEOUT_US);
    status = readb(req_addr + 528);
    g_assert_cmpint(status, ==, 0);

    data = g_malloc0(512);
    memread(req_addr + 16, data, 512);
    g_assert_cmpstr(data, ==, "TEST");
    g_free(data);

    g_free(indirect);
    guest_free(qs->alloc, req_addr);

    /* End test */
    qvirtqueue_cleanup(dev->vdev.bus, &vqpci->vq, qs->alloc);
    qvirtio_pci_device_disable(dev);
    qvirtio_pci_device_free(dev);
    qtest_shutdown(qs);
}

static void pci_config(void)
{
    QVirtioPCIDevice *dev;
    QOSState *qs;
    int n_size = TEST_IMAGE_SIZE / 2;
    uint64_t capacity;

    qs = pci_test_start();

    dev = virtio_blk_pci_init(qs->pcibus, PCI_SLOT);

    capacity = qvirtio_config_readq(&dev->vdev, 0);
    g_assert_cmpint(capacity, ==, TEST_IMAGE_SIZE / 512);

    qvirtio_set_driver_ok(&dev->vdev);

    qmp_discard_response("{ 'execute': 'block_resize', "
                         " 'arguments': { 'device': 'drive0', "
                         " 'size': %d } }", n_size);
    qvirtio_wait_config_isr(&dev->vdev, QVIRTIO_BLK_TIMEOUT_US);

    capacity = qvirtio_config_readq(&dev->vdev, 0);
    g_assert_cmpint(capacity, ==, n_size / 512);

    qvirtio_pci_device_disable(dev);
    qvirtio_pci_device_free(dev);

    qtest_shutdown(qs);
}

static void pci_msix(void)
{
    QVirtioPCIDevice *dev;
    QOSState *qs;
    QVirtQueuePCI *vqpci;
    QVirtioBlkReq req;
    int n_size = TEST_IMAGE_SIZE / 2;
    uint64_t req_addr;
    uint64_t capacity;
    uint32_t features;
    uint32_t free_head;
    uint8_t status;
    char *data;

    qs = pci_test_start();

    dev = virtio_blk_pci_init(qs->pcibus, PCI_SLOT);
    qpci_msix_enable(dev->pdev);

    qvirtio_pci_set_msix_configuration_vector(dev, qs->alloc, 0);

    capacity = qvirtio_config_readq(&dev->vdev, 0);
    g_assert_cmpint(capacity, ==, TEST_IMAGE_SIZE / 512);

    features = qvirtio_get_features(&dev->vdev);
    features = features & ~(QVIRTIO_F_BAD_FEATURE |
                            (1u << VIRTIO_RING_F_INDIRECT_DESC) |
                            (1u << VIRTIO_RING_F_EVENT_IDX) |
                            (1u << VIRTIO_BLK_F_SCSI));
    qvirtio_set_features(&dev->vdev, features);

    vqpci = (QVirtQueuePCI *)qvirtqueue_setup(&dev->vdev, qs->alloc, 0);
    qvirtqueue_pci_msix_setup(dev, vqpci, qs->alloc, 1);

    qvirtio_set_driver_ok(&dev->vdev);

    qmp_discard_response("{ 'execute': 'block_resize', "
                         " 'arguments': { 'device': 'drive0', "
                         " 'size': %d } }", n_size);

    qvirtio_wait_config_isr(&dev->vdev, QVIRTIO_BLK_TIMEOUT_US);

    capacity = qvirtio_config_readq(&dev->vdev, 0);
    g_assert_cmpint(capacity, ==, n_size / 512);

    /* Write request */
    req.type = VIRTIO_BLK_T_OUT;
    req.ioprio = 1;
    req.sector = 0;
    req.data = g_malloc0(512);
    strcpy(req.data, "TEST");

    req_addr = virtio_blk_request(qs->alloc, &dev->vdev, &req, 512);

    g_free(req.data);

    free_head = qvirtqueue_add(&vqpci->vq, req_addr, 16, false, true);
    qvirtqueue_add(&vqpci->vq, req_addr + 16, 512, false, true);
    qvirtqueue_add(&vqpci->vq, req_addr + 528, 1, true, false);
    qvirtqueue_kick(&dev->vdev, &vqpci->vq, free_head);

    qvirtio_wait_used_elem(&dev->vdev, &vqpci->vq, free_head, NULL,
                           QVIRTIO_BLK_TIMEOUT_US);

    status = readb(req_addr + 528);
    g_assert_cmpint(status, ==, 0);

    guest_free(qs->alloc, req_addr);

    /* Read request */
    req.type = VIRTIO_BLK_T_IN;
    req.ioprio = 1;
    req.sector = 0;
    req.data = g_malloc0(512);

    req_addr = virtio_blk_request(qs->alloc, &dev->vdev, &req, 512);

    g_free(req.data);

    free_head = qvirtqueue_add(&vqpci->vq, req_addr, 16, false, true);
    qvirtqueue_add(&vqpci->vq, req_addr + 16, 512, true, true);
    qvirtqueue_add(&vqpci->vq, req_addr + 528, 1, true, false);

    qvirtqueue_kick(&dev->vdev, &vqpci->vq, free_head);


    qvirtio_wait_used_elem(&dev->vdev, &vqpci->vq, free_head, NULL,
                           QVIRTIO_BLK_TIMEOUT_US);

    status = readb(req_addr + 528);
    g_assert_cmpint(status, ==, 0);

    data = g_malloc0(512);
    memread(req_addr + 16, data, 512);
    g_assert_cmpstr(data, ==, "TEST");
    g_free(data);

    guest_free(qs->alloc, req_addr);

    /* End test */
    qvirtqueue_cleanup(dev->vdev.bus, &vqpci->vq, qs->alloc);
    qpci_msix_disable(dev->pdev);
    qvirtio_pci_device_disable(dev);
    qvirtio_pci_device_free(dev);
    qtest_shutdown(qs);
}

static void pci_idx(void)
{
    QVirtioPCIDevice *dev;
    QOSState *qs;
    QVirtQueuePCI *vqpci;
    QVirtioBlkReq req;
    uint64_t req_addr;
    uint64_t capacity;
    uint32_t features;
    uint32_t free_head;
    uint32_t write_head;
    uint32_t desc_idx;
    uint8_t status;
    char *data;

    qs = pci_test_start();

    dev = virtio_blk_pci_init(qs->pcibus, PCI_SLOT);
    qpci_msix_enable(dev->pdev);

    qvirtio_pci_set_msix_configuration_vector(dev, qs->alloc, 0);

    capacity = qvirtio_config_readq(&dev->vdev, 0);
    g_assert_cmpint(capacity, ==, TEST_IMAGE_SIZE / 512);

    features = qvirtio_get_features(&dev->vdev);
    features = features & ~(QVIRTIO_F_BAD_FEATURE |
                            (1u << VIRTIO_RING_F_INDIRECT_DESC) |
                            (1u << VIRTIO_F_NOTIFY_ON_EMPTY) |
                            (1u << VIRTIO_BLK_F_SCSI));
    qvirtio_set_features(&dev->vdev, features);

    vqpci = (QVirtQueuePCI *)qvirtqueue_setup(&dev->vdev, qs->alloc, 0);
    qvirtqueue_pci_msix_setup(dev, vqpci, qs->alloc, 1);

    qvirtio_set_driver_ok(&dev->vdev);

    /* Write request */
    req.type = VIRTIO_BLK_T_OUT;
    req.ioprio = 1;
    req.sector = 0;
    req.data = g_malloc0(512);
    strcpy(req.data, "TEST");

    req_addr = virtio_blk_request(qs->alloc, &dev->vdev, &req, 512);

    g_free(req.data);

    free_head = qvirtqueue_add(&vqpci->vq, req_addr, 16, false, true);
    qvirtqueue_add(&vqpci->vq, req_addr + 16, 512, false, true);
    qvirtqueue_add(&vqpci->vq, req_addr + 528, 1, true, false);
    qvirtqueue_kick(&dev->vdev, &vqpci->vq, free_head);

    qvirtio_wait_used_elem(&dev->vdev, &vqpci->vq, free_head, NULL,
                           QVIRTIO_BLK_TIMEOUT_US);

    /* Write request */
    req.type = VIRTIO_BLK_T_OUT;
    req.ioprio = 1;
    req.sector = 1;
    req.data = g_malloc0(512);
    strcpy(req.data, "TEST");

    req_addr = virtio_blk_request(qs->alloc, &dev->vdev, &req, 512);

    g_free(req.data);

    /* Notify after processing the third request */
    qvirtqueue_set_used_event(&vqpci->vq, 2);
    free_head = qvirtqueue_add(&vqpci->vq, req_addr, 16, false, true);
    qvirtqueue_add(&vqpci->vq, req_addr + 16, 512, false, true);
    qvirtqueue_add(&vqpci->vq, req_addr + 528, 1, true, false);
    qvirtqueue_kick(&dev->vdev, &vqpci->vq, free_head);
    write_head = free_head;

    /* No notification expected */
    status = qvirtio_wait_status_byte_no_isr(&dev->vdev,
                                             &vqpci->vq, req_addr + 528,
                                             QVIRTIO_BLK_TIMEOUT_US);
    g_assert_cmpint(status, ==, 0);

    guest_free(qs->alloc, req_addr);

    /* Read request */
    req.type = VIRTIO_BLK_T_IN;
    req.ioprio = 1;
    req.sector = 1;
    req.data = g_malloc0(512);

    req_addr = virtio_blk_request(qs->alloc, &dev->vdev, &req, 512);

    g_free(req.data);

    free_head = qvirtqueue_add(&vqpci->vq, req_addr, 16, false, true);
    qvirtqueue_add(&vqpci->vq, req_addr + 16, 512, true, true);
    qvirtqueue_add(&vqpci->vq, req_addr + 528, 1, true, false);

    qvirtqueue_kick(&dev->vdev, &vqpci->vq, free_head);

    /* We get just one notification for both requests */
    qvirtio_wait_used_elem(&dev->vdev, &vqpci->vq, write_head, NULL,
                           QVIRTIO_BLK_TIMEOUT_US);
    g_assert(qvirtqueue_get_buf(&vqpci->vq, &desc_idx, NULL));
    g_assert_cmpint(desc_idx, ==, free_head);

    status = readb(req_addr + 528);
    g_assert_cmpint(status, ==, 0);

    data = g_malloc0(512);
    memread(req_addr + 16, data, 512);
    g_assert_cmpstr(data, ==, "TEST");
    g_free(data);

    guest_free(qs->alloc, req_addr);

    /* End test */
    qvirtqueue_cleanup(dev->vdev.bus, &vqpci->vq, qs->alloc);
    qpci_msix_disable(dev->pdev);
    qvirtio_pci_device_disable(dev);
    qvirtio_pci_device_free(dev);
    qtest_shutdown(qs);
}

static void pci_hotplug(void)
{
    QVirtioPCIDevice *dev;
    QOSState *qs;
    const char *arch = qtest_get_arch();

    qs = pci_test_start();

    /* plug secondary disk */
    qtest_qmp_device_add("virtio-blk-pci", "drv1",
                         "{'addr': %s, 'drive': 'drive1'}",
                         stringify(PCI_SLOT_HP));

    dev = virtio_blk_pci_init(qs->pcibus, PCI_SLOT_HP);
    g_assert(dev);
    qvirtio_pci_device_disable(dev);
    qvirtio_pci_device_free(dev);

    /* unplug secondary disk */
    if (strcmp(arch, "i386") == 0 || strcmp(arch, "x86_64") == 0) {
        qpci_unplug_acpi_device_test("drv1", PCI_SLOT_HP);
    }
    qtest_shutdown(qs);
}

/*
 * Check that setting the vring addr on a non-existent virtqueue does
 * not crash.
 */
static void test_nonexistent_virtqueue(void)
{
    QPCIBar bar0;
    QOSState *qs;
    QPCIDevice *dev;

    qs = pci_test_start();
    dev = qpci_device_find(qs->pcibus, QPCI_DEVFN(4, 0));
    g_assert(dev != NULL);

    qpci_device_enable(dev);
    bar0 = qpci_iomap(dev, 0, NULL);

    qpci_io_writeb(dev, bar0, VIRTIO_PCI_QUEUE_SEL, 2);
    qpci_io_writel(dev, bar0, VIRTIO_PCI_QUEUE_PFN, 1);

    g_free(dev);
    qtest_shutdown(qs);
}

static void mmio_basic(void)
{
    QVirtioMMIODevice *dev;
    QVirtQueue *vq;
    QGuestAllocator *alloc;
    int n_size = TEST_IMAGE_SIZE / 2;
    uint64_t capacity;

    arm_test_start();

    dev = qvirtio_mmio_init_device(MMIO_DEV_BASE_ADDR, MMIO_PAGE_SIZE);
    g_assert(dev != NULL);
    g_assert_cmphex(dev->vdev.device_type, ==, VIRTIO_ID_BLOCK);

    qvirtio_reset(&dev->vdev);
    qvirtio_set_acknowledge(&dev->vdev);
    qvirtio_set_driver(&dev->vdev);

    alloc = generic_alloc_init(MMIO_RAM_ADDR, MMIO_RAM_SIZE, MMIO_PAGE_SIZE);
    vq = qvirtqueue_setup(&dev->vdev, alloc, 0);

    test_basic(&dev->vdev, alloc, vq);

    qmp_discard_response("{ 'execute': 'block_resize', "
                         " 'arguments': { 'device': 'drive0', "
                         " 'size': %d } }", n_size);

    qvirtio_wait_queue_isr(&dev->vdev, vq, QVIRTIO_BLK_TIMEOUT_US);

    capacity = qvirtio_config_readq(&dev->vdev, 0);
    g_assert_cmpint(capacity, ==, n_size / 512);

    /* End test */
    qvirtqueue_cleanup(dev->vdev.bus, vq, alloc);
    g_free(dev);
    generic_alloc_uninit(alloc);
    test_end();
}

int main(int argc, char **argv)
{
    const char *arch = qtest_get_arch();

    g_test_init(&argc, &argv, NULL);

    if (strcmp(arch, "i386") == 0 || strcmp(arch, "x86_64") == 0 ||
        strcmp(arch, "ppc64") == 0) {
        qtest_add_func("/virtio/blk/pci/basic", pci_basic);
        qtest_add_func("/virtio/blk/pci/indirect", pci_indirect);
        qtest_add_func("/virtio/blk/pci/config", pci_config);
        qtest_add_func("/virtio/blk/pci/nxvirtq", test_nonexistent_virtqueue);
        if (strcmp(arch, "i386") == 0 || strcmp(arch, "x86_64") == 0) {
            qtest_add_func("/virtio/blk/pci/msix", pci_msix);
            qtest_add_func("/virtio/blk/pci/idx", pci_idx);
        }
        qtest_add_func("/virtio/blk/pci/hotplug", pci_hotplug);
    } else if (strcmp(arch, "arm") == 0) {
        qtest_add_func("/virtio/blk/mmio/basic", mmio_basic);
    }

    return g_test_run();
}