Merge branch 'for-upstream/mali-dp' of git://linux-arm.org/linux-ld into drm-next

misc mali-dp updates.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Liviu Dudau <Liviu.Dudau@arm.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181003105009.GD1156@e110455-lin.cambridge.arm.com

7 files changed, 520 insertions, 73 deletions

diff --git a/drivers/gpu/drm/arm/malidp_crtc.c b/drivers/gpu/drm/arm/malidp_crtc.c
index ef44202fb43f..e1b72782848c 100644
--- a/drivers/gpu/drm/arm/malidp_crtc.c
+++ b/drivers/gpu/drm/arm/malidp_crtc.c
@@ -348,19 +348,20 @@ static int malidp_crtc_atomic_check(struct drm_crtc *crtc,
 
 	/*
 	 * check if there is enough rotation memory available for planes
-	 * that need 90° and 270° rotation. Each plane has set its required
-	 * memory size in the ->plane_check() callback, here we only make
-	 * sure that the sums are less that the total usable memory.
+	 * that need 90° and 270° rotion or planes that are compressed.
+	 * Each plane has set its required memory size in the ->plane_check()
+	 * callback, here we only make sure that the sums are less that the
+	 * total usable memory.
 	 *
 	 * The rotation memory allocation algorithm (for each plane):
-	 *  a. If no more rotated planes exist, all remaining rotate
-	 *     memory in the bank is available for use by the plane.
-	 *  b. If other rotated planes exist, and plane's layer ID is
-	 *     DE_VIDEO1, it can use all the memory from first bank if
-	 *     secondary rotation memory bank is available, otherwise it can
+	 *  a. If no more rotated or compressed planes exist, all remaining
+	 *     rotate memory in the bank is available for use by the plane.
+	 *  b. If other rotated or compressed planes exist, and plane's
+	 *     layer ID is DE_VIDEO1, it can use all the memory from first bank
+	 *     if secondary rotation memory bank is available, otherwise it can
 	 *     use up to half the bank's memory.
-	 *  c. If other rotated planes exist, and plane's layer ID is not
-	 *     DE_VIDEO1, it can use half of the available memory
+	 *  c. If other rotated or compressed planes exist, and plane's layer ID
+	 *     is not DE_VIDEO1, it can use half of the available memory.
 	 *
 	 * Note: this algorithm assumes that the order in which the planes are
 	 * checked always has DE_VIDEO1 plane first in the list if it is
@@ -372,7 +373,9 @@ static int malidp_crtc_atomic_check(struct drm_crtc *crtc,
 
 	/* first count the number of rotated planes */
 	drm_atomic_crtc_state_for_each_plane_state(plane, pstate, state) {
-		if (pstate->rotation & MALIDP_ROTATED_MASK)
+		struct drm_framebuffer *fb = pstate->fb;
+
+		if ((pstate->rotation & MALIDP_ROTATED_MASK) || fb->modifier)
 			rotated_planes++;
 	}
 
@@ -388,8 +391,9 @@ static int malidp_crtc_atomic_check(struct drm_crtc *crtc,
 	drm_atomic_crtc_state_for_each_plane_state(plane, pstate, state) {
 		struct malidp_plane *mp = to_malidp_plane(plane);
 		struct malidp_plane_state *ms = to_malidp_plane_state(pstate);
+		struct drm_framebuffer *fb = pstate->fb;
 
-		if (pstate->rotation & MALIDP_ROTATED_MASK) {
+		if ((pstate->rotation & MALIDP_ROTATED_MASK) || fb->modifier) {
 			/* process current plane */
 			rotated_planes--;
 
diff --git a/drivers/gpu/drm/arm/malidp_drv.c b/drivers/gpu/drm/arm/malidp_drv.c
index 34eec1a22428..90214851637f 100644
--- a/drivers/gpu/drm/arm/malidp_drv.c
+++ b/drivers/gpu/drm/arm/malidp_drv.c
@@ -37,6 +37,7 @@
 #include "malidp_hw.h"
 
 #define MALIDP_CONF_VALID_TIMEOUT	250
+#define AFBC_HEADER_SIZE		16
 
 static void malidp_write_gamma_table(struct malidp_hw_device *hwdev,
 				     u32 data[MALIDP_COEFFTAB_NUM_COEFFS])
@@ -258,8 +259,133 @@ static const struct drm_mode_config_helper_funcs malidp_mode_config_helpers = {
 	.atomic_commit_tail = malidp_atomic_commit_tail,
 };
 
+static bool
+malidp_verify_afbc_framebuffer_caps(struct drm_device *dev,
+				    const struct drm_mode_fb_cmd2 *mode_cmd)
+{
+	const struct drm_format_info *info;
+
+	if ((mode_cmd->modifier[0] >> 56) != DRM_FORMAT_MOD_VENDOR_ARM) {
+		DRM_DEBUG_KMS("Unknown modifier (not Arm)\n");
+		return false;
+	}
+
+	if (mode_cmd->modifier[0] &
+	    ~DRM_FORMAT_MOD_ARM_AFBC(AFBC_MOD_VALID_BITS)) {
+		DRM_DEBUG_KMS("Unsupported modifiers\n");
+		return false;
+	}
+
+	info = drm_get_format_info(dev, mode_cmd);
+	if (!info) {
+		DRM_DEBUG_KMS("Unable to get the format information\n");
+		return false;
+	}
+
+	if (info->num_planes != 1) {
+		DRM_DEBUG_KMS("AFBC buffers expect one plane\n");
+		return false;
+	}
+
+	if (mode_cmd->offsets[0] != 0) {
+		DRM_DEBUG_KMS("AFBC buffers' plane offset should be 0\n");
+		return false;
+	}
+
+	switch (mode_cmd->modifier[0] & AFBC_FORMAT_MOD_BLOCK_SIZE_MASK) {
+	case AFBC_FORMAT_MOD_BLOCK_SIZE_16x16:
+		if ((mode_cmd->width % 16) || (mode_cmd->height % 16)) {
+			DRM_DEBUG_KMS("AFBC buffers must be aligned to 16 pixels\n");
+			return false;
+		}
+		break;
+	default:
+		DRM_DEBUG_KMS("Unsupported AFBC block size\n");
+		return false;
+	}
+
+	return true;
+}
+
+static bool
+malidp_verify_afbc_framebuffer_size(struct drm_device *dev,
+				    struct drm_file *file,
+				    const struct drm_mode_fb_cmd2 *mode_cmd)
+{
+	int n_superblocks = 0;
+	const struct drm_format_info *info;
+	struct drm_gem_object *objs = NULL;
+	u32 afbc_superblock_size = 0, afbc_superblock_height = 0;
+	u32 afbc_superblock_width = 0, afbc_size = 0;
+
+	switch (mode_cmd->modifier[0] & AFBC_FORMAT_MOD_BLOCK_SIZE_MASK) {
+	case AFBC_FORMAT_MOD_BLOCK_SIZE_16x16:
+		afbc_superblock_height = 16;
+		afbc_superblock_width = 16;
+		break;
+	default:
+		DRM_DEBUG_KMS("AFBC superblock size is not supported\n");
+		return false;
+	}
+
+	info = drm_get_format_info(dev, mode_cmd);
+
+	n_superblocks = (mode_cmd->width / afbc_superblock_width) *
+		(mode_cmd->height / afbc_superblock_height);
+
+	afbc_superblock_size = info->cpp[0] * afbc_superblock_width *
+		afbc_superblock_height;
+
+	afbc_size = ALIGN(n_superblocks * AFBC_HEADER_SIZE, 128);
+
+	if (mode_cmd->width * info->cpp[0] != mode_cmd->pitches[0]) {
+		DRM_DEBUG_KMS("Invalid value of pitch (=%u) should be same as width (=%u) * cpp (=%u)\n",
+			      mode_cmd->pitches[0], mode_cmd->width, info->cpp[0]);
+		return false;
+	}
+
+	objs = drm_gem_object_lookup(file, mode_cmd->handles[0]);
+	if (!objs) {
+		DRM_DEBUG_KMS("Failed to lookup GEM object\n");
+		return false;
+	}
+
+	if (objs->size < afbc_size) {
+		DRM_DEBUG_KMS("buffer size (%zu) too small for AFBC buffer size = %u\n",
+			      objs->size, afbc_size);
+		drm_gem_object_put_unlocked(objs);
+		return false;
+	}
+
+	drm_gem_object_put_unlocked(objs);
+
+	return true;
+}
+
+static bool
+malidp_verify_afbc_framebuffer(struct drm_device *dev, struct drm_file *file,
+			       const struct drm_mode_fb_cmd2 *mode_cmd)
+{
+	if (malidp_verify_afbc_framebuffer_caps(dev, mode_cmd))
+		return malidp_verify_afbc_framebuffer_size(dev, file, mode_cmd);
+
+	return false;
+}
+
+struct drm_framebuffer *
+malidp_fb_create(struct drm_device *dev, struct drm_file *file,
+		 const struct drm_mode_fb_cmd2 *mode_cmd)
+{
+	if (mode_cmd->modifier[0]) {
+		if (!malidp_verify_afbc_framebuffer(dev, file, mode_cmd))
+			return ERR_PTR(-EINVAL);
+	}
+
+	return drm_gem_fb_create(dev, file, mode_cmd);
+}
+
 static const struct drm_mode_config_funcs malidp_mode_config_funcs = {
-	.fb_create = drm_gem_fb_create,
+	.fb_create = malidp_fb_create,
 	.atomic_check = drm_atomic_helper_check,
 	.atomic_commit = drm_atomic_helper_commit,
 };
diff --git a/drivers/gpu/drm/arm/malidp_drv.h b/drivers/gpu/drm/arm/malidp_drv.h
index e3eb0cb1f385..b76c86f18a56 100644
--- a/drivers/gpu/drm/arm/malidp_drv.h
+++ b/drivers/gpu/drm/arm/malidp_drv.h
@@ -55,6 +55,12 @@ struct malidp_plane {
 	const struct malidp_layer *layer;
 };
 
+enum mmu_prefetch_mode {
+	MALIDP_PREFETCH_MODE_NONE,
+	MALIDP_PREFETCH_MODE_PARTIAL,
+	MALIDP_PREFETCH_MODE_FULL,
+};
+
 struct malidp_plane_state {
 	struct drm_plane_state base;
 
@@ -63,6 +69,8 @@ struct malidp_plane_state {
 	/* internal format ID */
 	u8 format;
 	u8 n_planes;
+	enum mmu_prefetch_mode mmu_prefetch_mode;
+	u32 mmu_prefetch_pgsize;
 };
 
 #define to_malidp_plane(x) container_of(x, struct malidp_plane, base)
diff --git a/drivers/gpu/drm/arm/malidp_hw.c b/drivers/gpu/drm/arm/malidp_hw.c
index 2781e462c1ed..7aad7dd80d8c 100644
--- a/drivers/gpu/drm/arm/malidp_hw.c
+++ b/drivers/gpu/drm/arm/malidp_hw.c
@@ -84,16 +84,48 @@ static const struct malidp_format_id malidp550_de_formats[] = {
 };
 
 static const struct malidp_layer malidp500_layers[] = {
-	{ DE_VIDEO1, MALIDP500_DE_LV_BASE, MALIDP500_DE_LV_PTR_BASE, MALIDP_DE_LV_STRIDE0, MALIDP500_LV_YUV2RGB },
-	{ DE_GRAPHICS1, MALIDP500_DE_LG1_BASE, MALIDP500_DE_LG1_PTR_BASE, MALIDP_DE_LG_STRIDE, 0 },
-	{ DE_GRAPHICS2, MALIDP500_DE_LG2_BASE, MALIDP500_DE_LG2_PTR_BASE, MALIDP_DE_LG_STRIDE, 0 },
+	/* id, base address, fb pointer address base, stride offset,
+	 *	yuv2rgb matrix offset, mmu control register offset, rotation_features
+	 */
+	{ DE_VIDEO1, MALIDP500_DE_LV_BASE, MALIDP500_DE_LV_PTR_BASE,
+		MALIDP_DE_LV_STRIDE0, MALIDP500_LV_YUV2RGB, 0, ROTATE_ANY },
+	{ DE_GRAPHICS1, MALIDP500_DE_LG1_BASE, MALIDP500_DE_LG1_PTR_BASE,
+		MALIDP_DE_LG_STRIDE, 0, 0, ROTATE_ANY },
+	{ DE_GRAPHICS2, MALIDP500_DE_LG2_BASE, MALIDP500_DE_LG2_PTR_BASE,
+		MALIDP_DE_LG_STRIDE, 0, 0, ROTATE_ANY },
 };
 
 static const struct malidp_layer malidp550_layers[] = {
-	{ DE_VIDEO1, MALIDP550_DE_LV1_BASE, MALIDP550_DE_LV1_PTR_BASE, MALIDP_DE_LV_STRIDE0, MALIDP550_LV_YUV2RGB },
-	{ DE_GRAPHICS1, MALIDP550_DE_LG_BASE, MALIDP550_DE_LG_PTR_BASE, MALIDP_DE_LG_STRIDE, 0 },
-	{ DE_VIDEO2, MALIDP550_DE_LV2_BASE, MALIDP550_DE_LV2_PTR_BASE, MALIDP_DE_LV_STRIDE0, MALIDP550_LV_YUV2RGB },
-	{ DE_SMART, MALIDP550_DE_LS_BASE, MALIDP550_DE_LS_PTR_BASE, MALIDP550_DE_LS_R1_STRIDE, 0 },
+	/* id, base address, fb pointer address base, stride offset,
+	 *	yuv2rgb matrix offset, mmu control register offset, rotation_features
+	 */
+	{ DE_VIDEO1, MALIDP550_DE_LV1_BASE, MALIDP550_DE_LV1_PTR_BASE,
+		MALIDP_DE_LV_STRIDE0, MALIDP550_LV_YUV2RGB, 0, ROTATE_ANY },
+	{ DE_GRAPHICS1, MALIDP550_DE_LG_BASE, MALIDP550_DE_LG_PTR_BASE,
+		MALIDP_DE_LG_STRIDE, 0, 0, ROTATE_ANY },
+	{ DE_VIDEO2, MALIDP550_DE_LV2_BASE, MALIDP550_DE_LV2_PTR_BASE,
+		MALIDP_DE_LV_STRIDE0, MALIDP550_LV_YUV2RGB, 0, ROTATE_ANY },
+	{ DE_SMART, MALIDP550_DE_LS_BASE, MALIDP550_DE_LS_PTR_BASE,
+		MALIDP550_DE_LS_R1_STRIDE, 0, 0, ROTATE_NONE },
+};
+
+static const struct malidp_layer malidp650_layers[] = {
+	/* id, base address, fb pointer address base, stride offset,
+	 *	yuv2rgb matrix offset, mmu control register offset,
+	 *	rotation_features
+	 */
+	{ DE_VIDEO1, MALIDP550_DE_LV1_BASE, MALIDP550_DE_LV1_PTR_BASE,
+		MALIDP_DE_LV_STRIDE0, MALIDP550_LV_YUV2RGB,
+		MALIDP650_DE_LV_MMU_CTRL, ROTATE_ANY },
+	{ DE_GRAPHICS1, MALIDP550_DE_LG_BASE, MALIDP550_DE_LG_PTR_BASE,
+		MALIDP_DE_LG_STRIDE, 0, MALIDP650_DE_LG_MMU_CTRL,
+		ROTATE_COMPRESSED },
+	{ DE_VIDEO2, MALIDP550_DE_LV2_BASE, MALIDP550_DE_LV2_PTR_BASE,
+		MALIDP_DE_LV_STRIDE0, MALIDP550_LV_YUV2RGB,
+		MALIDP650_DE_LV_MMU_CTRL, ROTATE_ANY },
+	{ DE_SMART, MALIDP550_DE_LS_BASE, MALIDP550_DE_LS_PTR_BASE,
+		MALIDP550_DE_LS_R1_STRIDE, 0, MALIDP650_DE_LS_MMU_CTRL,
+		ROTATE_NONE },
 };
 
 #define SE_N_SCALING_COEFFS	96
@@ -288,10 +320,6 @@ static void malidp500_modeset(struct malidp_hw_device *hwdev, struct videomode *
 
 static int malidp500_rotmem_required(struct malidp_hw_device *hwdev, u16 w, u16 h, u32 fmt)
 {
-	/* RGB888 or BGR888 can't be rotated */
-	if ((fmt == DRM_FORMAT_RGB888) || (fmt == DRM_FORMAT_BGR888))
-		return -EINVAL;
-
 	/*
 	 * Each layer needs enough rotation memory to fit 8 lines
 	 * worth of pixel data. Required size is then:
@@ -579,10 +607,6 @@ static int malidp550_rotmem_required(struct malidp_hw_device *hwdev, u16 w, u16
 {
 	u32 bytes_per_col;
 
-	/* raw RGB888 or BGR888 can't be rotated */
-	if ((fmt == DRM_FORMAT_RGB888) || (fmt == DRM_FORMAT_BGR888))
-		return -EINVAL;
-
 	switch (fmt) {
 	/* 8 lines at 4 bytes per pixel */
 	case DRM_FORMAT_ARGB2101010:
@@ -853,8 +877,8 @@ const struct malidp_hw malidp_device[MALIDP_MAX_DEVICES] = {
 			.dc_base = MALIDP550_DC_BASE,
 			.out_depth_base = MALIDP550_DE_OUTPUT_DEPTH,
 			.features = MALIDP_REGMAP_HAS_CLEARIRQ,
-			.n_layers = ARRAY_SIZE(malidp550_layers),
-			.layers = malidp550_layers,
+			.n_layers = ARRAY_SIZE(malidp650_layers),
+			.layers = malidp650_layers,
 			.de_irq_map = {
 				.irq_mask = MALIDP_DE_IRQ_UNDERRUN |
 					    MALIDP650_DE_IRQ_DRIFT |
diff --git a/drivers/gpu/drm/arm/malidp_hw.h b/drivers/gpu/drm/arm/malidp_hw.h
index 9fc94c08190f..40155e2ea9d9 100644
--- a/drivers/gpu/drm/arm/malidp_hw.h
+++ b/drivers/gpu/drm/arm/malidp_hw.h
@@ -36,6 +36,12 @@ enum {
 	SE_MEMWRITE = BIT(5),
 };
 
+enum rotation_features {
+	ROTATE_NONE,		/* does not support rotation at all */
+	ROTATE_ANY,		/* supports rotation on any buffers */
+	ROTATE_COMPRESSED,	/* supports rotation only on compressed buffers */
+};
+
 struct malidp_format_id {
 	u32 format;		/* DRM fourcc */
 	u8 layer;		/* bitmask of layers supporting it */
@@ -62,6 +68,8 @@ struct malidp_layer {
 	u16 ptr;		/* address offset for the pointer register */
 	u16 stride_offset;	/* offset to the first stride register. */
 	s16 yuv2rgb_offset;	/* offset to the YUV->RGB matrix entries */
+	u16 mmu_ctrl_offset;    /* offset to the MMU control register */
+	enum rotation_features rot;	/* type of rotation supported */
 };
 
 enum malidp_scaling_coeff_set {
@@ -380,4 +388,9 @@ static inline void malidp_se_set_enh_coeffs(struct malidp_hw_device *hwdev)
 
 #define MALIDP_GAMMA_LUT_SIZE		4096
 
+#define AFBC_MOD_VALID_BITS (AFBC_FORMAT_MOD_BLOCK_SIZE_MASK | \
+			AFBC_FORMAT_MOD_YTR | AFBC_FORMAT_MOD_SPLIT | \
+			AFBC_FORMAT_MOD_SPARSE | AFBC_FORMAT_MOD_CBR | \
+			AFBC_FORMAT_MOD_TILED | AFBC_FORMAT_MOD_SC)
+
 #endif  /* __MALIDP_HW_H__ */
diff --git a/drivers/gpu/drm/arm/malidp_planes.c b/drivers/gpu/drm/arm/malidp_planes.c
index 49c37f6dd63e..837a24d56675 100644
--- a/drivers/gpu/drm/arm/malidp_planes.c
+++ b/drivers/gpu/drm/arm/malidp_planes.c
@@ -10,11 +10,14 @@
  * ARM Mali DP plane manipulation routines.
  */
 
+#include <linux/iommu.h>
+
 #include <drm/drmP.h>
 #include <drm/drm_atomic.h>
 #include <drm/drm_atomic_helper.h>
 #include <drm/drm_fb_cma_helper.h>
 #include <drm/drm_gem_cma_helper.h>
+#include <drm/drm_gem_framebuffer_helper.h>
 #include <drm/drm_plane_helper.h>
 #include <drm/drm_print.h>
 
@@ -36,6 +39,7 @@
 #define   LAYER_COMP_MASK		(0x3 << 12)
 #define   LAYER_COMP_PIXEL		(0x3 << 12)
 #define   LAYER_COMP_PLANE		(0x2 << 12)
+#define   LAYER_PMUL_ENABLE		(0x1 << 14)
 #define   LAYER_ALPHA_OFFSET		(16)
 #define   LAYER_ALPHA_MASK		(0xff)
 #define   LAYER_ALPHA(x)		(((x) & LAYER_ALPHA_MASK) << LAYER_ALPHA_OFFSET)
@@ -56,6 +60,13 @@
  */
 #define MALIDP_ALPHA_LUT 0xffaa5500
 
+/* page sizes the MMU prefetcher can support */
+#define MALIDP_MMU_PREFETCH_PARTIAL_PGSIZES	(SZ_4K | SZ_64K)
+#define MALIDP_MMU_PREFETCH_FULL_PGSIZES	(SZ_1M | SZ_2M)
+
+/* readahead for partial-frame prefetch */
+#define MALIDP_MMU_PREFETCH_READAHEAD		8
+
 static void malidp_de_plane_destroy(struct drm_plane *plane)
 {
 	struct malidp_plane *mp = to_malidp_plane(plane);
@@ -100,6 +111,9 @@ drm_plane_state *malidp_duplicate_plane_state(struct drm_plane *plane)
 	state->format = m_state->format;
 	state->n_planes = m_state->n_planes;
 
+	state->mmu_prefetch_mode = m_state->mmu_prefetch_mode;
+	state->mmu_prefetch_pgsize = m_state->mmu_prefetch_pgsize;
+
 	return &state->base;
 }
 
@@ -112,6 +126,12 @@ static void malidp_destroy_plane_state(struct drm_plane *plane,
 	kfree(m_state);
 }
 
+static const char * const prefetch_mode_names[] = {
+	[MALIDP_PREFETCH_MODE_NONE] = "MMU_PREFETCH_NONE",
+	[MALIDP_PREFETCH_MODE_PARTIAL] = "MMU_PREFETCH_PARTIAL",
+	[MALIDP_PREFETCH_MODE_FULL] = "MMU_PREFETCH_FULL",
+};
+
 static void malidp_plane_atomic_print_state(struct drm_printer *p,
 					    const struct drm_plane_state *state)
 {
@@ -120,6 +140,9 @@ static void malidp_plane_atomic_print_state(struct drm_printer *p,
 	drm_printf(p, "\trotmem_size=%u\n", ms->rotmem_size);
 	drm_printf(p, "\tformat_id=%u\n", ms->format);
 	drm_printf(p, "\tn_planes=%u\n", ms->n_planes);
+	drm_printf(p, "\tmmu_prefetch_mode=%s\n",
+		   prefetch_mode_names[ms->mmu_prefetch_mode]);
+	drm_printf(p, "\tmmu_prefetch_pgsize=%d\n", ms->mmu_prefetch_pgsize);
 }
 
 static const struct drm_plane_funcs malidp_de_plane_funcs = {
@@ -173,6 +196,199 @@ static int malidp_se_check_scaling(struct malidp_plane *mp,
 	return 0;
 }
 
+static u32 malidp_get_pgsize_bitmap(struct malidp_plane *mp)
+{
+	u32 pgsize_bitmap = 0;
+
+	if (iommu_present(&platform_bus_type)) {
+		struct iommu_domain *mmu_dom =
+			iommu_get_domain_for_dev(mp->base.dev->dev);
+
+		if (mmu_dom)
+			pgsize_bitmap = mmu_dom->pgsize_bitmap;
+	}
+
+	return pgsize_bitmap;
+}
+
+/*
+ * Check if the framebuffer is entirely made up of pages at least pgsize in
+ * size. Only a heuristic: assumes that each scatterlist entry has been aligned
+ * to the largest page size smaller than its length and that the MMU maps to
+ * the largest page size possible.
+ */
+static bool malidp_check_pages_threshold(struct malidp_plane_state *ms,
+					 u32 pgsize)
+{
+	int i;
+
+	for (i = 0; i < ms->n_planes; i++) {
+		struct drm_gem_object *obj;
+		struct drm_gem_cma_object *cma_obj;
+		struct sg_table *sgt;
+		struct scatterlist *sgl;
+
+		obj = drm_gem_fb_get_obj(ms->base.fb, i);
+		cma_obj = to_drm_gem_cma_obj(obj);
+
+		if (cma_obj->sgt)
+			sgt = cma_obj->sgt;
+		else
+			sgt = obj->dev->driver->gem_prime_get_sg_table(obj);
+
+		if (!sgt)
+			return false;
+
+		sgl = sgt->sgl;
+
+		while (sgl) {
+			if (sgl->length < pgsize) {
+				if (!cma_obj->sgt)
+					kfree(sgt);
+				return false;
+			}
+
+			sgl = sg_next(sgl);
+		}
+		if (!cma_obj->sgt)
+			kfree(sgt);
+	}
+
+	return true;
+}
+
+/*
+ * Check if it is possible to enable partial-frame MMU prefetch given the
+ * current format, AFBC state and rotation.
+ */
+static bool malidp_partial_prefetch_supported(u32 format, u64 modifier,
+					      unsigned int rotation)
+{
+	bool afbc, sparse;
+
+	/* rotation and horizontal flip not supported for partial prefetch */
+	if (rotation & (DRM_MODE_ROTATE_90 | DRM_MODE_ROTATE_180 |
+			DRM_MODE_ROTATE_270 | DRM_MODE_REFLECT_X))
+		return false;
+
+	afbc = modifier & DRM_FORMAT_MOD_ARM_AFBC(0);
+	sparse = modifier & AFBC_FORMAT_MOD_SPARSE;
+
+	switch (format) {
+	case DRM_FORMAT_ARGB2101010:
+	case DRM_FORMAT_RGBA1010102:
+	case DRM_FORMAT_BGRA1010102:
+	case DRM_FORMAT_ARGB8888:
+	case DRM_FORMAT_RGBA8888:
+	case DRM_FORMAT_BGRA8888:
+	case DRM_FORMAT_XRGB8888:
+	case DRM_FORMAT_XBGR8888:
+	case DRM_FORMAT_RGBX8888:
+	case DRM_FORMAT_BGRX8888:
+	case DRM_FORMAT_RGB888:
+	case DRM_FORMAT_RGBA5551:
+	case DRM_FORMAT_RGB565:
+		/* always supported */
+		return true;
+
+	case DRM_FORMAT_ABGR2101010:
+	case DRM_FORMAT_ABGR8888:
+	case DRM_FORMAT_ABGR1555:
+	case DRM_FORMAT_BGR565:
+		/* supported, but if AFBC then must be sparse mode */
+		return (!afbc) || (afbc && sparse);
+
+	case DRM_FORMAT_BGR888:
+		/* supported, but not for AFBC */
+		return !afbc;
+
+	case DRM_FORMAT_YUYV:
+	case DRM_FORMAT_UYVY:
+	case DRM_FORMAT_NV12:
+	case DRM_FORMAT_YUV420:
+		/* not supported */
+		return false;
+
+	default:
+		return false;
+	}
+}
+
+/*
+ * Select the preferred MMU prefetch mode. Full-frame prefetch is preferred as
+ * long as the framebuffer is all large pages. Otherwise partial-frame prefetch
+ * is selected as long as it is supported for the current format. The selected
+ * page size for prefetch is returned in pgsize_bitmap.
+ */
+static enum mmu_prefetch_mode malidp_mmu_prefetch_select_mode
+		(struct malidp_plane_state *ms,	u32 *pgsize_bitmap)
+{
+	u32 pgsizes;
+
+	/* get the full-frame prefetch page size(s) supported by the MMU */
+	pgsizes = *pgsize_bitmap & MALIDP_MMU_PREFETCH_FULL_PGSIZES;
+
+	while (pgsizes) {
+		u32 largest_pgsize = 1 << __fls(pgsizes);
+
+		if (malidp_check_pages_threshold(ms, largest_pgsize)) {
+			*pgsize_bitmap = largest_pgsize;
+			return MALIDP_PREFETCH_MODE_FULL;
+		}
+
+		pgsizes -= largest_pgsize;
+	}
+
+	/* get the partial-frame prefetch page size(s) supported by the MMU */
+	pgsizes = *pgsize_bitmap & MALIDP_MMU_PREFETCH_PARTIAL_PGSIZES;
+
+	if (malidp_partial_prefetch_supported(ms->base.fb->format->format,
+					      ms->base.fb->modifier,
+					      ms->base.rotation)) {
+		/* partial prefetch using the smallest page size */
+		*pgsize_bitmap = 1 << __ffs(pgsizes);
+		return MALIDP_PREFETCH_MODE_PARTIAL;
+	}
+	*pgsize_bitmap = 0;
+	return MALIDP_PREFETCH_MODE_NONE;
+}
+
+static u32 malidp_calc_mmu_control_value(enum mmu_prefetch_mode mode,
+					 u8 readahead, u8 n_planes, u32 pgsize)
+{
+	u32 mmu_ctrl = 0;
+
+	if (mode != MALIDP_PREFETCH_MODE_NONE) {
+		mmu_ctrl |= MALIDP_MMU_CTRL_EN;
+
+		if (mode == MALIDP_PREFETCH_MODE_PARTIAL) {
+			mmu_ctrl |= MALIDP_MMU_CTRL_MODE;
+			mmu_ctrl |= MALIDP_MMU_CTRL_PP_NUM_REQ(readahead);
+		}
+
+		if (pgsize == SZ_64K || pgsize == SZ_2M) {
+			int i;
+
+			for (i = 0; i < n_planes; i++)
+				mmu_ctrl |= MALIDP_MMU_CTRL_PX_PS(i);
+		}
+	}
+
+	return mmu_ctrl;
+}
+
+static void malidp_de_prefetch_settings(struct malidp_plane *mp,
+					struct malidp_plane_state *ms)
+{
+	if (!mp->layer->mmu_ctrl_offset)
+		return;
+
+	/* get the page sizes supported by the MMU */
+	ms->mmu_prefetch_pgsize = malidp_get_pgsize_bitmap(mp);
+	ms->mmu_prefetch_mode  =
+		malidp_mmu_prefetch_select_mode(ms, &ms->mmu_prefetch_pgsize);
+}
+
 static int malidp_de_plane_check(struct drm_plane *plane,
 				 struct drm_plane_state *state)
 {
@@ -180,6 +396,7 @@ static int malidp_de_plane_check(struct drm_plane *plane,
 	struct malidp_plane_state *ms = to_malidp_plane_state(state);
 	bool rotated = state->rotation & MALIDP_ROTATED_MASK;
 	struct drm_framebuffer *fb;
+	u16 pixel_alpha = state->pixel_blend_mode;
 	int i, ret;
 
 	if (!state->crtc || !state->fb)
@@ -223,11 +440,20 @@ static int malidp_de_plane_check(struct drm_plane *plane,
 	if (ret)
 		return ret;
 
-	/* packed RGB888 / BGR888 can't be rotated or flipped */
-	if (state->rotation != DRM_MODE_ROTATE_0 &&
-	    (fb->format->format == DRM_FORMAT_RGB888 ||
-	     fb->format->format == DRM_FORMAT_BGR888))
-		return -EINVAL;
+	/* validate the rotation constraints for each layer */
+	if (state->rotation != DRM_MODE_ROTATE_0) {
+		if (mp->layer->rot == ROTATE_NONE)
+			return -EINVAL;
+		if ((mp->layer->rot == ROTATE_COMPRESSED) && !(fb->modifier))
+			return -EINVAL;
+		/*
+		 * packed RGB888 / BGR888 can't be rotated or flipped
+		 * unless they are stored in a compressed way
+		 */
+		if ((fb->format->format == DRM_FORMAT_RGB888 ||
+		     fb->format->format == DRM_FORMAT_BGR888) && !(fb->modifier))
+			return -EINVAL;
+	}
 
 	ms->rotmem_size = 0;
 	if (state->rotation & MALIDP_ROTATED_MASK) {
@@ -242,6 +468,14 @@ static int malidp_de_plane_check(struct drm_plane *plane,
 		ms->rotmem_size = val;
 	}
 
+	/* HW can't support plane + pixel blending */
+	if ((state->alpha != DRM_BLEND_ALPHA_OPAQUE) &&
+	    (pixel_alpha != DRM_MODE_BLEND_PIXEL_NONE) &&
+	    fb->format->has_alpha)
+		return -EINVAL;
+
+	malidp_de_prefetch_settings(mp, ms);
+
 	return 0;
 }
 
@@ -318,22 +552,42 @@ static void malidp_de_set_color_encoding(struct malidp_plane *plane,
 	}
 }
 
+static void malidp_de_set_mmu_control(struct malidp_plane *mp,
+				      struct malidp_plane_state *ms)
+{
+	u32 mmu_ctrl;
+
+	/* check hardware supports MMU prefetch */
+	if (!mp->layer->mmu_ctrl_offset)
+		return;
+
+	mmu_ctrl = malidp_calc_mmu_control_value(ms->mmu_prefetch_mode,
+						 MALIDP_MMU_PREFETCH_READAHEAD,
+						 ms->n_planes,
+						 ms->mmu_prefetch_pgsize);
+
+	malidp_hw_write(mp->hwdev, mmu_ctrl,
+			mp->layer->base + mp->layer->mmu_ctrl_offset);
+}
+
 static void malidp_de_plane_update(struct drm_plane *plane,
 				   struct drm_plane_state *old_state)
 {
 	struct malidp_plane *mp;
 	struct malidp_plane_state *ms = to_malidp_plane_state(plane->state);
+	struct drm_plane_state *state = plane->state;
+	u16 pixel_alpha = state->pixel_blend_mode;
+	u8 plane_alpha = state->alpha >> 8;
 	u32 src_w, src_h, dest_w, dest_h, val;
 	int i;
-	bool format_has_alpha = plane->state->fb->format->has_alpha;
 
 	mp = to_malidp_plane(plane);
 
 	/* convert src values from Q16 fixed point to integer */
-	src_w = plane->state->src_w >> 16;
-	src_h = plane->state->src_h >> 16;
-	dest_w = plane->state->crtc_w;
-	dest_h = plane->state->crtc_h;
+	src_w = state->src_w >> 16;
+	src_h = state->src_h >> 16;
+	dest_w = state->crtc_w;
+	dest_h = state->crtc_h;
 
 	val = malidp_hw_read(mp->hwdev, mp->layer->base);
 	val = (val & ~LAYER_FORMAT_MASK) | ms->format;
@@ -342,14 +596,17 @@ static void malidp_de_plane_update(struct drm_plane *plane,
 	for (i = 0; i < ms->n_planes; i++) {
 		/* calculate the offset for the layer's plane registers */
 		u16 ptr = mp->layer->ptr + (i << 4);
-		dma_addr_t fb_addr = drm_fb_cma_get_gem_addr(plane->state->fb,
-							     plane->state, i);
+		dma_addr_t fb_addr = drm_fb_cma_get_gem_addr(state->fb,
+							     state, i);
 
 		malidp_hw_write(mp->hwdev, lower_32_bits(fb_addr), ptr);
 		malidp_hw_write(mp->hwdev, upper_32_bits(fb_addr), ptr + 4);
 	}
+
+	malidp_de_set_mmu_control(mp, ms);
+
 	malidp_de_set_plane_pitches(mp, ms->n_planes,
-				    plane->state->fb->pitches);
+				    state->fb->pitches);
 
 	if ((plane->state->color_encoding != old_state->color_encoding) ||
 	    (plane->state->color_range != old_state->color_range))
@@ -362,52 +619,56 @@ static void malidp_de_plane_update(struct drm_plane *plane,
 	malidp_hw_write(mp->hwdev, LAYER_H_VAL(dest_w) | LAYER_V_VAL(dest_h),
 			mp->layer->base + MALIDP_LAYER_COMP_SIZE);
 
-	malidp_hw_write(mp->hwdev, LAYER_H_VAL(plane->state->crtc_x) |
-			LAYER_V_VAL(plane->state->crtc_y),
+	malidp_hw_write(mp->hwdev, LAYER_H_VAL(state->crtc_x) |
+			LAYER_V_VAL(state->crtc_y),
 			mp->layer->base + MALIDP_LAYER_OFFSET);
 
-	if (mp->layer->id == DE_SMART)
+	if (mp->layer->id == DE_SMART) {
+		/*
+		 * Enable the first rectangle in the SMART layer to be
+		 * able to use it as a drm plane.
+		 */
+		malidp_hw_write(mp->hwdev, 1,
+				mp->layer->base + MALIDP550_LS_ENABLE);
 		malidp_hw_write(mp->hwdev,
 				LAYER_H_VAL(src_w) | LAYER_V_VAL(src_h),
 				mp->layer->base + MALIDP550_LS_R1_IN_SIZE);
+	}
 
 	/* first clear the rotation bits */
 	val = malidp_hw_read(mp->hwdev, mp->layer->base + MALIDP_LAYER_CONTROL);
 	val &= ~LAYER_ROT_MASK;
 
 	/* setup the rotation and axis flip bits */
-	if (plane->state->rotation & DRM_MODE_ROTATE_MASK)
+	if (state->rotation & DRM_MODE_ROTATE_MASK)
 		val |= ilog2(plane->state->rotation & DRM_MODE_ROTATE_MASK) <<
 		       LAYER_ROT_OFFSET;
-	if (plane->state->rotation & DRM_MODE_REFLECT_X)
+	if (state->rotation & DRM_MODE_REFLECT_X)
 		val |= LAYER_H_FLIP;
-	if (plane->state->rotation & DRM_MODE_REFLECT_Y)
+	if (state->rotation & DRM_MODE_REFLECT_Y)
 		val |= LAYER_V_FLIP;
 
-	val &= ~LAYER_COMP_MASK;
-	if (format_has_alpha) {
-
-		/*
-		 * always enable pixel alpha blending until we have a way
-		 * to change blend modes
-		 */
-		val |= LAYER_COMP_PIXEL;
-	} else {
+	val &= ~(LAYER_COMP_MASK | LAYER_PMUL_ENABLE | LAYER_ALPHA(0xff));
 
-		/*
-		 * do not enable pixel alpha blending as the color channel
-		 * does not have any alpha information
-		 */
+	if (state->alpha != DRM_BLEND_ALPHA_OPAQUE) {
 		val |= LAYER_COMP_PLANE;
-
-		/* Set layer alpha coefficient to 0xff ie fully opaque */
-		val |= LAYER_ALPHA(0xff);
+	} else if (state->fb->format->has_alpha) {
+		/* We only care about blend mode if the format has alpha */
+		switch (pixel_alpha) {
+		case DRM_MODE_BLEND_PREMULTI:
+			val |= LAYER_COMP_PIXEL | LAYER_PMUL_ENABLE;
+			break;
+		case DRM_MODE_BLEND_COVERAGE:
+			val |= LAYER_COMP_PIXEL;
+			break;
+		}
 	}
+	val |= LAYER_ALPHA(plane_alpha);
 
 	val &= ~LAYER_FLOWCFG(LAYER_FLOWCFG_MASK);
-	if (plane->state->crtc) {
+	if (state->crtc) {
 		struct malidp_crtc_state *m =
-			to_malidp_crtc_state(plane->state->crtc->state);
+			to_malidp_crtc_state(state->crtc->state);
 
 		if (m->scaler_config.scale_enable &&
 		    m->scaler_config.plane_src_id == mp->layer->id)
@@ -446,6 +707,9 @@ int malidp_de_planes_init(struct drm_device *drm)
 	unsigned long crtcs = 1 << drm->mode_config.num_crtc;
 	unsigned long flags = DRM_MODE_ROTATE_0 | DRM_MODE_ROTATE_90 | DRM_MODE_ROTATE_180 |
 			      DRM_MODE_ROTATE_270 | DRM_MODE_REFLECT_X | DRM_MODE_REFLECT_Y;
+	unsigned int blend_caps = BIT(DRM_MODE_BLEND_PIXEL_NONE) |
+				  BIT(DRM_MODE_BLEND_PREMULTI)   |
+				  BIT(DRM_MODE_BLEND_COVERAGE);
 	u32 *formats;
 	int ret, i, j, n;
 
@@ -483,13 +747,10 @@ int malidp_de_planes_init(struct drm_device *drm)
 		plane->hwdev = malidp->dev;
 		plane->layer = &map->layers[i];
 
+		drm_plane_create_alpha_property(&plane->base);
+		drm_plane_create_blend_mode_property(&plane->base, blend_caps);
+
 		if (id == DE_SMART) {
-			/*
-			 * Enable the first rectangle in the SMART layer to be
-			 * able to use it as a drm plane.
-			 */
-			malidp_hw_write(malidp->dev, 1,
-					plane->layer->base + MALIDP550_LS_ENABLE);
 			/* Skip the features which the SMART layer doesn't have. */
 			continue;
 		}
diff --git a/drivers/gpu/drm/arm/malidp_regs.h b/drivers/gpu/drm/arm/malidp_regs.h
index 6ffe849774f2..7ce3e141464d 100644
--- a/drivers/gpu/drm/arm/malidp_regs.h
+++ b/drivers/gpu/drm/arm/malidp_regs.h
@@ -247,6 +247,17 @@
 #define MALIDP550_CONFIG_VALID		0x0c014
 #define MALIDP550_CONFIG_ID		0x0ffd4
 
+/* register offsets specific to DP650 */
+#define MALIDP650_DE_LV_MMU_CTRL	0x000D0
+#define MALIDP650_DE_LG_MMU_CTRL	0x00048
+#define MALIDP650_DE_LS_MMU_CTRL	0x00078
+
+/* bit masks to set the MMU control register */
+#define MALIDP_MMU_CTRL_EN		(1 << 0)
+#define MALIDP_MMU_CTRL_MODE		(1 << 4)
+#define MALIDP_MMU_CTRL_PX_PS(x)	(1 << (8 + (x)))
+#define MALIDP_MMU_CTRL_PP_NUM_REQ(x)	(((x) & 0x7f) << 12)
+
 /*
  * Starting with DP550 the register map blocks has been standardised to the
  * following layout: