Merge branch 'feature/staging_sm5' of git://people.freedesktop.org/~sroland/linux into drm-next

vmwgfx pull for for 5.7. Needed for GL4 functionality.
Sync up device headers, add support for new commands, code
refactoring around surface definition.
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: "Roland Scheidegger (VMware)" <rscheidegger.oss@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200323235434.11780-1-rscheidegger.oss@gmail.com

drivers/gpu/drm/vmwgfx/Makefile

@@ -8,7 +8,7 @@ vmwgfx-y := vmwgfx_execbuf.o vmwgfx_gmr.o vmwgfx_kms.o vmwgfx_drv.o \
 	    vmwgfx_cmdbuf_res.o vmwgfx_cmdbuf.o vmwgfx_stdu.o \
 	    vmwgfx_cotable.o vmwgfx_so.o vmwgfx_binding.o vmwgfx_msg.o \
 	    vmwgfx_simple_resource.o vmwgfx_va.o vmwgfx_blit.o \
-	    vmwgfx_validation.o vmwgfx_page_dirty.o \
+	    vmwgfx_validation.o vmwgfx_page_dirty.o vmwgfx_streamoutput.o \
 	    ttm_object.o ttm_lock.o
 
 obj-$(CONFIG_DRM_VMWGFX) := vmwgfx.o

drivers/gpu/drm/vmwgfx/device_include/svga3d_limits.h

 /* SPDX-License-Identifier: GPL-2.0 OR MIT */
 /**********************************************************
- * Copyright 2007-2015 VMware, Inc.
+ * Copyright 2007-2019 VMware, Inc.
  *
  * Permission is hereby granted, free of charge, to any person
  * obtaining a copy of this software and associated documentation
@@ -40,11 +40,25 @@
 #include "includeCheck.h"
 
 #define SVGA3D_NUM_CLIPPLANES                   6
+#define SVGA3D_MAX_CONTEXT_IDS                  256
+#define SVGA3D_MAX_SURFACE_IDS                  (32 * 1024)
+
+/*
+ * While there are separate bind-points for RenderTargetViews and
+ * UnorderedAccessViews in a DXContext, there is in fact one shared
+ * semantic space that the guest-driver can use on any given draw call.
+ * So there are really only 8 slots that can be spilt up between them, with the
+ * spliceIndex controlling where the UAV's sit in the collapsed array.
+ */
 #define SVGA3D_MAX_RENDER_TARGETS               8
 #define SVGA3D_MAX_SIMULTANEOUS_RENDER_TARGETS  (SVGA3D_MAX_RENDER_TARGETS)
 #define SVGA3D_MAX_UAVIEWS                      8
-#define SVGA3D_MAX_CONTEXT_IDS                  256
-#define SVGA3D_MAX_SURFACE_IDS                  (32 * 1024)
+#define SVGA3D_DX11_1_MAX_UAVIEWS               64
+
+/*
+ * Maximum canonical size of a surface in host-backed mode (pre-GBObjects).
+ */
+#define SVGA3D_HB_MAX_SURFACE_SIZE MBYTES_2_BYTES(128)
 
 /*
  * Maximum ID a shader can be assigned on a given context.
@@ -59,6 +73,8 @@
 #define SVGA3D_NUM_TEXTURE_UNITS                32
 #define SVGA3D_NUM_LIGHTS                       8
 
+#define SVGA3D_MAX_VIDEOPROCESSOR_SAMPLERS      32
+
 /*
  * Maximum size in dwords of shader text the SVGA device will allow.
  * Currently 8 MB.
@@ -67,6 +83,11 @@
 #define SVGA3D_MAX_SHADER_MEMORY  (SVGA3D_MAX_SHADER_MEMORY_BYTES / \
                                    sizeof(uint32))
 
+/*
+ * The maximum value of threadGroupCount in each dimension
+ */
+#define SVGA3D_MAX_SHADER_THREAD_GROUPS 65535
+
 #define SVGA3D_MAX_CLIP_PLANES    6
 
 /*
@@ -85,7 +106,9 @@
 /*
  * Maximum number of array indexes in a GB surface (with DX enabled).
  */
-#define SVGA3D_MAX_SURFACE_ARRAYSIZE 512
+#define SVGA3D_SM4_MAX_SURFACE_ARRAYSIZE 512
+#define SVGA3D_SM5_MAX_SURFACE_ARRAYSIZE 2048
+#define SVGA3D_MAX_SURFACE_ARRAYSIZE SVGA3D_SM5_MAX_SURFACE_ARRAYSIZE
 
 /*
  * The maximum number of vertex arrays we're guaranteed to support in
@@ -99,4 +122,9 @@
  */
 #define SVGA3D_MAX_DRAW_PRIMITIVE_RANGES 32
 
+/*
+ * The maximum number of samples that can be contained in a surface.
+ */
+#define SVGA3D_MAX_SAMPLES 8
+
 #endif /* _SVGA3D_LIMITS_H_ */

drivers/gpu/drm/vmwgfx/device_include/svga3d_surfacedefs.h

@@ -131,6 +131,8 @@ enum svga3d_block_desc {
 	SVGA3DBLOCKDESC_BC3         = 1 << 26,
 	SVGA3DBLOCKDESC_BC4         = 1 << 27,
 	SVGA3DBLOCKDESC_BC5         = 1 << 28,
+	SVGA3DBLOCKDESC_BC6H        = 1 << 29,
+	SVGA3DBLOCKDESC_BC7         = 1 << 30,
 
 	SVGA3DBLOCKDESC_A_UINT    = SVGA3DBLOCKDESC_ALPHA |
 				    SVGA3DBLOCKDESC_UINT |
@@ -290,6 +292,18 @@ enum svga3d_block_desc {
 					 SVGA3DBLOCKDESC_COMP_UNORM,
 	SVGA3DBLOCKDESC_BC5_COMP_SNORM = SVGA3DBLOCKDESC_BC5 |
 					 SVGA3DBLOCKDESC_COMP_SNORM,
+	SVGA3DBLOCKDESC_BC6H_COMP_TYPELESS = SVGA3DBLOCKDESC_BC6H |
+					     SVGA3DBLOCKDESC_COMP_TYPELESS,
+	SVGA3DBLOCKDESC_BC6H_COMP_UF16 = SVGA3DBLOCKDESC_BC6H |
+					 SVGA3DBLOCKDESC_COMPRESSED,
+	SVGA3DBLOCKDESC_BC6H_COMP_SF16 = SVGA3DBLOCKDESC_BC6H |
+					 SVGA3DBLOCKDESC_COMPRESSED,
+	SVGA3DBLOCKDESC_BC7_COMP_TYPELESS = SVGA3DBLOCKDESC_BC7 |
+					    SVGA3DBLOCKDESC_COMP_TYPELESS,
+	SVGA3DBLOCKDESC_BC7_COMP_UNORM = SVGA3DBLOCKDESC_BC7 |
+					 SVGA3DBLOCKDESC_COMP_UNORM,
+	SVGA3DBLOCKDESC_BC7_COMP_UNORM_SRGB = SVGA3DBLOCKDESC_BC7_COMP_UNORM |
+					      SVGA3DBLOCKDESC_SRGB,
 
 	SVGA3DBLOCKDESC_NV12       = SVGA3DBLOCKDESC_YUV_VIDEO |
 				     SVGA3DBLOCKDESC_PLANAR_YUV |
@@ -494,7 +508,7 @@ static const struct svga3d_surface_desc svga3d_surface_descs[] = {
       {{8}, {8}, {8}, {0}},
       {{16}, {8}, {0}, {0}}},
 
-   {SVGA3D_FORMAT_DEAD1, SVGA3DBLOCKDESC_UVL,
+   {SVGA3D_FORMAT_DEAD1, SVGA3DBLOCKDESC_NONE,
       {1, 1, 1},  3, 3,
       {{8}, {8}, {8}, {0}},
       {{16}, {8}, {0}, {0}}},
@@ -604,7 +618,7 @@ static const struct svga3d_surface_desc svga3d_surface_descs[] = {
       {{0}, {0}, {48}, {0}},
       {{0}, {0}, {0}, {0}}},
 
-   {SVGA3D_AYUV, SVGA3DBLOCKDESC_AYUV,
+   {SVGA3D_FORMAT_DEAD2, SVGA3DBLOCKDESC_NONE,
       {1, 1, 1},  4, 4,
       {{8}, {8}, {8}, {8}},
       {{0}, {8}, {16}, {24}}},
@@ -1103,6 +1117,46 @@ static const struct svga3d_surface_desc svga3d_surface_descs[] = {
       {4, 4, 1},  16, 16,
       {{0}, {0}, {128}, {0}},
       {{0}, {0}, {0}, {0}}},
+
+   {SVGA3D_B4G4R4A4_UNORM, SVGA3DBLOCKDESC_RGBA_UNORM,
+      {1, 1, 1},  2, 2,
+      {{4}, {4}, {4}, {4}},
+      {{0}, {4}, {8}, {12}}},
+
+   {SVGA3D_BC6H_TYPELESS, SVGA3DBLOCKDESC_BC6H_COMP_TYPELESS,
+      {4, 4, 1},  16, 16,
+      {{0}, {0}, {128}, {0}},
+      {{0}, {0}, {0}, {0}}},
+
+   {SVGA3D_BC6H_UF16, SVGA3DBLOCKDESC_BC6H_COMP_UF16,
+      {4, 4, 1},  16, 16,
+      {{0}, {0}, {128}, {0}},
+      {{0}, {0}, {0}, {0}}},
+
+   {SVGA3D_BC6H_SF16, SVGA3DBLOCKDESC_BC6H_COMP_SF16,
+      {4, 4, 1},  16, 16,
+      {{0}, {0}, {128}, {0}},
+      {{0}, {0}, {0}, {0}}},
+
+   {SVGA3D_BC7_TYPELESS, SVGA3DBLOCKDESC_BC7_COMP_TYPELESS,
+      {4, 4, 1},  16, 16,
+      {{0}, {0}, {128}, {0}},
+      {{0}, {0}, {0}, {0}}},
+
+   {SVGA3D_BC7_UNORM, SVGA3DBLOCKDESC_BC7_COMP_UNORM,
+      {4, 4, 1},  16, 16,
+      {{0}, {0}, {128}, {0}},
+      {{0}, {0}, {0}, {0}}},
+
+   {SVGA3D_BC7_UNORM_SRGB, SVGA3DBLOCKDESC_BC7_COMP_UNORM_SRGB,
+      {4, 4, 1},  16, 16,
+      {{0}, {0}, {128}, {0}},
+      {{0}, {0}, {0}, {0}}},
+
+   {SVGA3D_AYUV, SVGA3DBLOCKDESC_AYUV,
+      {1, 1, 1},  4, 4,
+      {{8}, {8}, {8}, {8}},
+      {{0}, {8}, {16}, {24}}},
 };
 
 static inline u32 clamped_umul32(u32 a, u32 b)

drivers/gpu/drm/vmwgfx/device_include/svga_types.h

@@ -37,6 +37,7 @@ typedef s8  int8;
 
 typedef uint64 PA;
 typedef uint32 PPN;
+typedef uint32 PPN32;
 typedef uint64 PPN64;
 
 typedef bool Bool;

drivers/gpu/drm/vmwgfx/vmwgfx_binding.h

@@ -33,6 +33,8 @@
 
 #define VMW_MAX_VIEW_BINDINGS 128
 
+#define VMW_MAX_UAV_BIND_TYPE 2
+
 struct vmw_private;
 struct vmw_ctx_binding_state;
 
@@ -48,9 +50,12 @@ enum vmw_ctx_binding_type {
 	vmw_ctx_binding_dx_rt,
 	vmw_ctx_binding_sr,
 	vmw_ctx_binding_ds,
-	vmw_ctx_binding_so,
+	vmw_ctx_binding_so_target,
 	vmw_ctx_binding_vb,
 	vmw_ctx_binding_ib,
+	vmw_ctx_binding_uav,
+	vmw_ctx_binding_cs_uav,
+	vmw_ctx_binding_so,
 	vmw_ctx_binding_max
 };
 
@@ -128,14 +133,14 @@ struct vmw_ctx_bindinfo_view {
 };
 
 /**
- * struct vmw_ctx_bindinfo_so - StreamOutput binding metadata
+ * struct vmw_ctx_bindinfo_so_target - StreamOutput binding metadata
  *
  * @bi: struct vmw_ctx_bindinfo we derive from.
  * @offset: Device data used to reconstruct binding command.
  * @size: Device data used to reconstruct binding command.
  * @slot: Device data used to reconstruct binding command.
  */
-struct vmw_ctx_bindinfo_so {
+struct vmw_ctx_bindinfo_so_target {
 	struct vmw_ctx_bindinfo bi;
 	uint32 offset;
 	uint32 size;
@@ -189,9 +194,31 @@ struct vmw_dx_shader_bindings {
 	unsigned long dirty;
 };
 
+/**
+ * struct vmw_ctx_bindinfo_uav - UAV context binding state.
+ * @views: UAV view bindings.
+ * @splice_index: The device splice index set by user-space.
+ */
+struct vmw_ctx_bindinfo_uav {
+	struct vmw_ctx_bindinfo_view views[SVGA3D_MAX_UAVIEWS];
+	uint32 index;
+};
+
+/**
+ * struct vmw_ctx_bindinfo_so - Stream output binding metadata.
+ * @bi: struct vmw_ctx_bindinfo we derive from.
+ * @slot: Device data used to reconstruct binding command.
+ */
+struct vmw_ctx_bindinfo_so {
+	struct vmw_ctx_bindinfo bi;
+	uint32 slot;
+};
+
 extern void vmw_binding_add(struct vmw_ctx_binding_state *cbs,
 			    const struct vmw_ctx_bindinfo *ci,
 			    u32 shader_slot, u32 slot);
+extern void vmw_binding_add_uav_index(struct vmw_ctx_binding_state *cbs,
+				      uint32 slot, uint32 splice_index);
 extern void
 vmw_binding_state_commit(struct vmw_ctx_binding_state *to,
 			 struct vmw_ctx_binding_state *from);

drivers/gpu/drm/vmwgfx/vmwgfx_context.c

@@ -36,7 +36,7 @@ struct vmw_user_context {
 	struct vmw_resource res;
 	struct vmw_ctx_binding_state *cbs;
 	struct vmw_cmdbuf_res_manager *man;
-	struct vmw_resource *cotables[SVGA_COTABLE_DX10_MAX];
+	struct vmw_resource *cotables[SVGA_COTABLE_MAX];
 	spinlock_t cotable_lock;
 	struct vmw_buffer_object *dx_query_mob;
 };
@@ -116,12 +116,15 @@ static const struct vmw_res_func vmw_dx_context_func = {
  * Context management:
  */
 
-static void vmw_context_cotables_unref(struct vmw_user_context *uctx)
+static void vmw_context_cotables_unref(struct vmw_private *dev_priv,
+				       struct vmw_user_context *uctx)
 {
 	struct vmw_resource *res;
 	int i;
+	u32 cotable_max = has_sm5_context(dev_priv) ?
+		SVGA_COTABLE_MAX : SVGA_COTABLE_DX10_MAX;
 
-	for (i = 0; i < SVGA_COTABLE_DX10_MAX; ++i) {
+	for (i = 0; i < cotable_max; ++i) {
 		spin_lock(&uctx->cotable_lock);
 		res = uctx->cotables[i];
 		uctx->cotables[i] = NULL;
@@ -155,7 +158,7 @@ static void vmw_hw_context_destroy(struct vmw_resource *res)
 		    !dev_priv->query_cid_valid)
 			__vmw_execbuf_release_pinned_bo(dev_priv, NULL);
 		mutex_unlock(&dev_priv->cmdbuf_mutex);
-		vmw_context_cotables_unref(uctx);
+		vmw_context_cotables_unref(dev_priv, uctx);
 		return;
 	}
 
@@ -208,7 +211,9 @@ static int vmw_gb_context_init(struct vmw_private *dev_priv,
 	spin_lock_init(&uctx->cotable_lock);
 
 	if (dx) {
-		for (i = 0; i < SVGA_COTABLE_DX10_MAX; ++i) {
+		u32 cotable_max = has_sm5_context(dev_priv) ?
+			SVGA_COTABLE_MAX : SVGA_COTABLE_DX10_MAX;
+		for (i = 0; i < cotable_max; ++i) {
 			uctx->cotables[i] = vmw_cotable_alloc(dev_priv,
 							      &uctx->res, i);
 			if (IS_ERR(uctx->cotables[i])) {
@@ -222,7 +227,7 @@ static int vmw_gb_context_init(struct vmw_private *dev_priv,
 	return 0;
 
 out_cotables:
-	vmw_context_cotables_unref(uctx);
+	vmw_context_cotables_unref(dev_priv, uctx);
 out_err:
 	if (res_free)
 		res_free(res);
@@ -545,10 +550,12 @@ void vmw_dx_context_scrub_cotables(struct vmw_resource *ctx,
 {
 	struct vmw_user_context *uctx =
 		container_of(ctx, struct vmw_user_context, res);
+	u32 cotable_max = has_sm5_context(ctx->dev_priv) ?
+		SVGA_COTABLE_MAX : SVGA_COTABLE_DX10_MAX;
 	int i;
 
 	vmw_binding_state_scrub(uctx->cbs);
-	for (i = 0; i < SVGA_COTABLE_DX10_MAX; ++i) {
+	for (i = 0; i < cotable_max; ++i) {
 		struct vmw_resource *res;
 
 		/* Avoid racing with ongoing cotable destruction. */
@@ -731,7 +738,7 @@ static int vmw_context_define(struct drm_device *dev, void *data,
 	};
 	int ret;
 
-	if (!dev_priv->has_dx && dx) {
+	if (!has_sm4_context(dev_priv) && dx) {
 		VMW_DEBUG_USER("DX contexts not supported by device.\n");
 		return -EINVAL;
 	}
@@ -839,7 +846,10 @@ struct vmw_cmdbuf_res_manager *vmw_context_res_man(struct vmw_resource *ctx)
 struct vmw_resource *vmw_context_cotable(struct vmw_resource *ctx,
 					 SVGACOTableType cotable_type)
 {
-	if (cotable_type >= SVGA_COTABLE_DX10_MAX)
+	u32 cotable_max = has_sm5_context(ctx->dev_priv) ?
+		SVGA_COTABLE_MAX : SVGA_COTABLE_DX10_MAX;
+
+	if (cotable_type >= cotable_max)
 		return ERR_PTR(-EINVAL);
 
 	return container_of(ctx, struct vmw_user_context, res)->

drivers/gpu/drm/vmwgfx/vmwgfx_cotable.c

@@ -80,9 +80,10 @@ static const struct vmw_cotable_info co_info[] = {
 	{1, sizeof(SVGACOTableDXDepthStencilEntry), NULL},
 	{1, sizeof(SVGACOTableDXRasterizerStateEntry), NULL},
 	{1, sizeof(SVGACOTableDXSamplerEntry), NULL},
-	{1, sizeof(SVGACOTableDXStreamOutputEntry), NULL},
+	{1, sizeof(SVGACOTableDXStreamOutputEntry), &vmw_dx_streamoutput_cotable_list_scrub},
 	{1, sizeof(SVGACOTableDXQueryEntry), NULL},
-	{1, sizeof(SVGACOTableDXShaderEntry), &vmw_dx_shader_cotable_list_scrub}
+	{1, sizeof(SVGACOTableDXShaderEntry), &vmw_dx_shader_cotable_list_scrub},
+	{1, sizeof(SVGACOTableDXUAViewEntry), &vmw_view_cotable_list_destroy}
 };
 
 /*
@@ -102,6 +103,7 @@ const SVGACOTableType vmw_cotable_scrub_order[] = {
 	SVGA_COTABLE_SAMPLER,
 	SVGA_COTABLE_STREAMOUTPUT,
 	SVGA_COTABLE_DXQUERY,
+	SVGA_COTABLE_UAVIEW,
 };
 
 static int vmw_cotable_bind(struct vmw_resource *res,

drivers/gpu/drm/vmwgfx/vmwgfx_drv.c

@@ -290,6 +290,8 @@ static void vmw_print_capabilities2(uint32_t capabilities2)
 		DRM_INFO("  Grow oTable.\n");
 	if (capabilities2 & SVGA_CAP2_INTRA_SURFACE_COPY)
 		DRM_INFO("  IntraSurface copy.\n");
+	if (capabilities2 & SVGA_CAP2_DX3)
+		DRM_INFO("  DX3.\n");
 }
 
 static void vmw_print_capabilities(uint32_t capabilities)
@@ -449,7 +451,7 @@ static int vmw_request_device(struct vmw_private *dev_priv)
 	dev_priv->cman = vmw_cmdbuf_man_create(dev_priv);
 	if (IS_ERR(dev_priv->cman)) {
 		dev_priv->cman = NULL;
-		dev_priv->has_dx = false;
+		dev_priv->sm_type = VMW_SM_LEGACY;
 	}
 
 	ret = vmw_request_device_late(dev_priv);
@@ -718,9 +720,15 @@ static int vmw_driver_load(struct drm_device *dev, unsigned long chipset)
 	dev_priv->max_mob_pages = 0;
 	dev_priv->max_mob_size = 0;
 	if (dev_priv->capabilities & SVGA_CAP_GBOBJECTS) {
-		uint64_t mem_size =
-			vmw_read(dev_priv,
-				 SVGA_REG_SUGGESTED_GBOBJECT_MEM_SIZE_KB);
+		uint64_t mem_size;
+
+		if (dev_priv->capabilities2 & SVGA_CAP2_GB_MEMSIZE_2)
+			mem_size = vmw_read(dev_priv,
+					    SVGA_REG_GBOBJECT_MEM_SIZE_KB);
+		else
+			mem_size =
+				vmw_read(dev_priv,
+					 SVGA_REG_SUGGESTED_GBOBJECT_MEM_SIZE_KB);
 
 		/*
 		 * Workaround for low memory 2D VMs to compensate for the
@@ -883,14 +891,32 @@ static int vmw_driver_load(struct drm_device *dev, unsigned long chipset)
 		}
 	}
 
-	if (dev_priv->has_mob) {
+	if (dev_priv->has_mob && (dev_priv->capabilities & SVGA_CAP_DX)) {
 		spin_lock(&dev_priv->cap_lock);
 		vmw_write(dev_priv, SVGA_REG_DEV_CAP, SVGA3D_DEVCAP_DXCONTEXT);
-		dev_priv->has_dx = !!vmw_read(dev_priv, SVGA_REG_DEV_CAP);
+		if (vmw_read(dev_priv, SVGA_REG_DEV_CAP))
+			dev_priv->sm_type = VMW_SM_4;
 		spin_unlock(&dev_priv->cap_lock);
 	}
 
 	vmw_validation_mem_init_ttm(dev_priv, VMWGFX_VALIDATION_MEM_GRAN);
+
+	/* SVGA_CAP2_DX2 (DefineGBSurface_v3) is needed for SM4_1 support */
+	if (has_sm4_context(dev_priv) &&
+	    (dev_priv->capabilities2 & SVGA_CAP2_DX2)) {
+		vmw_write(dev_priv, SVGA_REG_DEV_CAP, SVGA3D_DEVCAP_SM41);
+
+		if (vmw_read(dev_priv, SVGA_REG_DEV_CAP))
+			dev_priv->sm_type = VMW_SM_4_1;
+
+		if (has_sm4_1_context(dev_priv) &&
+		    (dev_priv->capabilities2 & SVGA_CAP2_DX3)) {
+			vmw_write(dev_priv, SVGA_REG_DEV_CAP, SVGA3D_DEVCAP_SM5);
+			if (vmw_read(dev_priv, SVGA_REG_DEV_CAP))
+				dev_priv->sm_type = VMW_SM_5;
+		}
+	}
+
 	ret = vmw_kms_init(dev_priv);
 	if (unlikely(ret != 0))
 		goto out_no_kms;
@@ -900,23 +926,14 @@ static int vmw_driver_load(struct drm_device *dev, unsigned long chipset)
 	if (ret)
 		goto out_no_fifo;
 
-	if (dev_priv->has_dx) {
-		/*
-		 * SVGA_CAP2_DX2 (DefineGBSurface_v3) is needed for SM4_1
-		 * support
-		 */
-		if ((dev_priv->capabilities2 & SVGA_CAP2_DX2) != 0) {
-			vmw_write(dev_priv, SVGA_REG_DEV_CAP,
-					SVGA3D_DEVCAP_SM41);
-			dev_priv->has_sm4_1 = vmw_read(dev_priv,
-							SVGA_REG_DEV_CAP);
-		}
-	}
-
-	DRM_INFO("DX: %s\n", dev_priv->has_dx ? "yes." : "no.");
 	DRM_INFO("Atomic: %s\n", (dev->driver->driver_features & DRIVER_ATOMIC)
 		 ? "yes." : "no.");
-	DRM_INFO("SM4_1: %s\n", dev_priv->has_sm4_1 ? "yes." : "no.");
+	if (dev_priv->sm_type == VMW_SM_5)
+		DRM_INFO("SM5 support available.\n");
+	if (dev_priv->sm_type == VMW_SM_4_1)
+		DRM_INFO("SM4_1 support available.\n");
+	if (dev_priv->sm_type == VMW_SM_4)
+		DRM_INFO("SM4 support available.\n");
 
 	snprintf(host_log, sizeof(host_log), "vmwgfx: %s-%s",
 		VMWGFX_REPO, VMWGFX_GIT_VERSION);

drivers/gpu/drm/vmwgfx/vmwgfx_drv.h

@@ -58,7 +58,7 @@
 #define VMWGFX_DRIVER_NAME "vmwgfx"
 #define VMWGFX_DRIVER_DATE "20200114"
 #define VMWGFX_DRIVER_MAJOR 2
-#define VMWGFX_DRIVER_MINOR 17
+#define VMWGFX_DRIVER_MINOR 18
 #define VMWGFX_DRIVER_PATCHLEVEL 0
 #define VMWGFX_FIFO_STATIC_SIZE (1024*1024)
 #define VMWGFX_MAX_RELOCATIONS 2048
@@ -202,6 +202,7 @@ enum vmw_res_type {
 	vmw_res_dx_context,
 	vmw_res_cotable,
 	vmw_res_view,
+	vmw_res_streamoutput,
 	vmw_res_max
 };
 
@@ -210,7 +211,8 @@ enum vmw_res_type {
  */
 enum vmw_cmdbuf_res_type {
 	vmw_cmdbuf_res_shader,
-	vmw_cmdbuf_res_view
+	vmw_cmdbuf_res_view,
+	vmw_cmdbuf_res_streamoutput
 };
 
 struct vmw_cmdbuf_res_manager;
@@ -223,24 +225,58 @@ struct vmw_cursor_snooper {
 struct vmw_framebuffer;
 struct vmw_surface_offset;
 
-struct vmw_surface {
-	struct vmw_resource res;
-	SVGA3dSurfaceAllFlags flags;
-	uint32_t format;
-	uint32_t mip_levels[DRM_VMW_MAX_SURFACE_FACES];
+/**
+ * struct vmw_surface_metadata - Metadata describing a surface.
+ *
+ * @flags: Device flags.
+ * @format: Surface SVGA3D_x format.
+ * @mip_levels: Mip level for each face. For GB first index is used only.
+ * @multisample_count: Sample count.
+ * @multisample_pattern: Sample patterns.
+ * @quality_level: Quality level.
+ * @autogen_filter: Filter for automatically generated mipmaps.
+ * @array_size: Number of array elements for a 1D/2D texture. For cubemap
+                texture number of faces * array_size. This should be 0 for pre
+		SM4 device.
+ * @buffer_byte_stride: Buffer byte stride.
+ * @num_sizes: Size of @sizes. For GB surface this should always be 1.
+ * @base_size: Surface dimension.
+ * @sizes: Array representing mip sizes. Legacy only.
+ * @scanout: Whether this surface will be used for scanout.
+ *
+ * This tracks metadata for both legacy and guest backed surface.
+ */
+struct vmw_surface_metadata {
+	u64 flags;
+	u32 format;
+	u32 mip_levels[DRM_VMW_MAX_SURFACE_FACES];
+	u32 multisample_count;
+	u32 multisample_pattern;
+	u32 quality_level;
+	u32 autogen_filter;
+	u32 array_size;
+	u32 num_sizes;
+	u32 buffer_byte_stride;
 	struct drm_vmw_size base_size;
 	struct drm_vmw_size *sizes;
-	uint32_t num_sizes;
 	bool scanout;
-	uint32_t array_size;
-	/* TODO so far just a extra pointer */
+};
+
+/**
+ * struct vmw_surface: Resource structure for a surface.
+ *
+ * @res: The base resource for this surface.
+ * @metadata: Metadata for this surface resource.
+ * @snooper: Cursor data. Legacy surface only.
+ * @offsets: Legacy surface only.
+ * @view_list: List of views bound to this surface.
+ */
+struct vmw_surface {
+	struct vmw_resource res;
+	struct vmw_surface_metadata metadata;
 	struct vmw_cursor_snooper snooper;
 	struct vmw_surface_offset *offsets;
-	SVGA3dTextureFilter autogen_filter;
-	uint32_t multisample_count;
 	struct list_head view_list;
-	SVGA3dMSPattern multisample_pattern;
-	SVGA3dMSQualityLevel quality_level;
 };
 
 struct vmw_marker_queue {
@@ -441,6 +477,22 @@ enum {
 	VMW_IRQTHREAD_MAX
 };
 
+/**
+ * enum vmw_sm_type - Graphics context capability supported by device.
+ * @VMW_SM_LEGACY: Pre DX context.
+ * @VMW_SM_4: Context support upto SM4.
+ * @VMW_SM_4_1: Context support upto SM4_1.
+ * @VMW_SM_5: Context support up to SM5.
+ * @VMW_SM_MAX: Should be the last.
+ */
+enum vmw_sm_type {
+	VMW_SM_LEGACY = 0,
+	VMW_SM_4,
+	VMW_SM_4_1,
+	VMW_SM_5,
+	VMW_SM_MAX
+};
+
 struct vmw_private {
 	struct ttm_bo_device bdev;
 
@@ -475,9 +527,9 @@ struct vmw_private {
 	bool has_mob;
 	spinlock_t hw_lock;
 	spinlock_t cap_lock;
-	bool has_dx;
 	bool assume_16bpp;
-	bool has_sm4_1;
+
+	enum vmw_sm_type sm_type;
 
 	/*
 	 * Framebuffer info.
@@ -648,6 +700,39 @@ static inline uint32_t vmw_read(struct vmw_private *dev_priv,
 	return val;
 }
 
+/**
+ * has_sm4_context - Does the device support SM4 context.
+ * @dev_priv: Device private.
+ *
+ * Return: Bool value if device support SM4 context or not.
+ */
+static inline bool has_sm4_context(const struct vmw_private *dev_priv)
+{
+	return (dev_priv->sm_type >= VMW_SM_4);
+}
+
+/**
+ * has_sm4_1_context - Does the device support SM4_1 context.
+ * @dev_priv: Device private.
+ *
+ * Return: Bool value if device support SM4_1 context or not.
+ */
+static inline bool has_sm4_1_context(const struct vmw_private *dev_priv)
+{
+	return (dev_priv->sm_type >= VMW_SM_4_1);
+}
+
+/**
+ * has_sm5_context - Does the device support SM5 context.
+ * @dev_priv: Device private.
+ *
+ * Return: Bool value if device support SM5 context or not.
+ */
+static inline bool has_sm5_context(const struct vmw_private *dev_priv)
+{
+	return (dev_priv->sm_type >= VMW_SM_5);
+}
+
 extern void vmw_svga_enable(struct vmw_private *dev_priv);
 extern void vmw_svga_disable(struct vmw_private *dev_priv);
 
@@ -1226,6 +1311,11 @@ extern int vmw_gb_surface_reference_ext_ioctl(struct drm_device *dev,
 					      void *data,
 					      struct drm_file *file_priv);
 
+int vmw_gb_surface_define(struct vmw_private *dev_priv,
+			  uint32_t user_accounting_size,
+			  const struct vmw_surface_metadata *req,
+			  struct vmw_surface **srf_out);
+
 /*
  * Shader management - vmwgfx_shader.c
  */
@@ -1258,6 +1348,24 @@ extern struct vmw_resource *
 vmw_shader_lookup(struct vmw_cmdbuf_res_manager *man,
 		  u32 user_key, SVGA3dShaderType shader_type);
 
+/*
+ * Streamoutput management
+ */
+struct vmw_resource *
+vmw_dx_streamoutput_lookup(struct vmw_cmdbuf_res_manager *man,
+			   u32 user_key);
+int vmw_dx_streamoutput_add(struct vmw_cmdbuf_res_manager *man,
+			    struct vmw_resource *ctx,
+			    SVGA3dStreamOutputId user_key,
+			    struct list_head *list);
+void vmw_dx_streamoutput_set_size(struct vmw_resource *res, u32 size);
+int vmw_dx_streamoutput_remove(struct vmw_cmdbuf_res_manager *man,
+			       SVGA3dStreamOutputId user_key,
+			       struct list_head *list);
+void vmw_dx_streamoutput_cotable_list_scrub(struct vmw_private *dev_priv,
+					    struct list_head *list,
+					    bool readback);
+
 /*
  * Command buffer managed resources - vmwgfx_cmdbuf_res.c
  */

drivers/gpu/drm/vmwgfx/vmwgfx_ioctl.c

@@ -114,10 +114,13 @@ int vmw_getparam_ioctl(struct drm_device *dev, void *data,
 			(dev_priv->active_display_unit == vmw_du_screen_target);
 		break;
 	case DRM_VMW_PARAM_DX:
-		param->value = dev_priv->has_dx;
+		param->value = has_sm4_context(dev_priv);
 		break;
 	case DRM_VMW_PARAM_SM4_1:
-		param->value = dev_priv->has_sm4_1;
+		param->value = has_sm4_1_context(dev_priv);
+		break;
+	case DRM_VMW_PARAM_SM5:
+		param->value = has_sm5_context(dev_priv);
 		break;
 	default:
 		return -EINVAL;
@@ -126,14 +129,17 @@ int vmw_getparam_ioctl(struct drm_device *dev, void *data,
 	return 0;
 }
 
-static u32 vmw_mask_multisample(unsigned int cap, u32 fmt_value)
+static u32 vmw_mask_legacy_multisample(unsigned int cap, u32 fmt_value)
 {
 	/*
 	 * A version of user-space exists which use MULTISAMPLE_MASKABLESAMPLES
 	 * to check the sample count supported by virtual device. Since there
 	 * never was support for multisample count for backing MOB return 0.
+	 *
+	 * MULTISAMPLE_MASKABLESAMPLES devcap is marked as deprecated by virtual
+	 * device.
 	 */
-	if (cap == SVGA3D_DEVCAP_MULTISAMPLE_MASKABLESAMPLES)
+	if (cap == SVGA3D_DEVCAP_DEAD5)
 		return 0;
 
 	return fmt_value;
@@ -164,7 +170,7 @@ static int vmw_fill_compat_cap(struct vmw_private *dev_priv, void *bounce,
 	for (i = 0; i < max_size; ++i) {
 		vmw_write(dev_priv, SVGA_REG_DEV_CAP, i);
 		compat_cap->pairs[i][0] = i;
-		compat_cap->pairs[i][1] = vmw_mask_multisample
+		compat_cap->pairs[i][1] = vmw_mask_legacy_multisample
 			(i, vmw_read(dev_priv, SVGA_REG_DEV_CAP));
 	}
 	spin_unlock(&dev_priv->cap_lock);
@@ -220,7 +226,7 @@ int vmw_get_cap_3d_ioctl(struct drm_device *dev, void *data,
 		spin_lock(&dev_priv->cap_lock);
 		for (i = 0; i < num; ++i) {
 			vmw_write(dev_priv, SVGA_REG_DEV_CAP, i);
-			*bounce32++ = vmw_mask_multisample
+			*bounce32++ = vmw_mask_legacy_multisample
 				(i, vmw_read(dev_priv, SVGA_REG_DEV_CAP));
 		}
 		spin_unlock(&dev_priv->cap_lock);

drivers/gpu/drm/vmwgfx/vmwgfx_kms.c

@@ -905,14 +905,14 @@ static int vmw_kms_new_framebuffer_surface(struct vmw_private *dev_priv,
 	 */
 
 	/* Surface must be marked as a scanout. */
-	if (unlikely(!surface->scanout))
+	if (unlikely(!surface->metadata.scanout))
 		return -EINVAL;
 
-	if (unlikely(surface->mip_levels[0] != 1 ||
-		     surface->num_sizes != 1 ||
-		     surface->base_size.width < mode_cmd->width ||
-		     surface->base_size.height < mode_cmd->height ||
-		     surface->base_size.depth != 1)) {
+	if (unlikely(surface->metadata.mip_levels[0] != 1 ||
+		     surface->metadata.num_sizes != 1 ||
+		     surface->metadata.base_size.width < mode_cmd->width ||
+		     surface->metadata.base_size.height < mode_cmd->height ||
+		     surface->metadata.base_size.depth != 1)) {
 		DRM_ERROR("Incompatible surface dimensions "
 			  "for requested mode.\n");
 		return -EINVAL;
@@ -941,7 +941,7 @@ static int vmw_kms_new_framebuffer_surface(struct vmw_private *dev_priv,
 	 * For DX, surface format validation is done when surface->scanout
 	 * is set.
 	 */
-	if (!dev_priv->has_dx && format != surface->format) {
+	if (!has_sm4_context(dev_priv) && format != surface->metadata.format) {
 		DRM_ERROR("Invalid surface format for requested mode.\n");
 		return -EINVAL;
 	}
@@ -1144,8 +1144,8 @@ static int vmw_create_bo_proxy(struct drm_device *dev,
 			       struct vmw_buffer_object *bo_mob,
 			       struct vmw_surface **srf_out)
 {
+	struct vmw_surface_metadata metadata = {0};
 	uint32_t format;
-	struct drm_vmw_size content_base_size = {0};
 	struct vmw_resource *res;
 	unsigned int bytes_pp;
 	struct drm_format_name_buf format_name;
@@ -1175,22 +1175,15 @@ static int vmw_create_bo_proxy(struct drm_device *dev,
 		return -EINVAL;
 	}
 
-	content_base_size.width  = mode_cmd->pitches[0] / bytes_pp;
-	content_base_size.height = mode_cmd->height;
-	content_base_size.depth  = 1;
-
-	ret = vmw_surface_gb_priv_define(dev,
-					 0, /* kernel visible only */
-					 0, /* flags */
-					 format,
-					 true, /* can be a scanout buffer */
-					 1, /* num of mip levels */
-					 0,
-					 0,
-					 content_base_size,
-					 SVGA3D_MS_PATTERN_NONE,
-					 SVGA3D_MS_QUALITY_NONE,
-					 srf_out);
+	metadata.format = format;
+	metadata.mip_levels[0] = 1;
+	metadata.num_sizes = 1;
+	metadata.base_size.width = mode_cmd->pitches[0] / bytes_pp;
+	metadata.base_size.height =  mode_cmd->height;
+	metadata.base_size.depth = 1;
+	metadata.scanout = true;
+
+	ret = vmw_gb_surface_define(vmw_priv(dev), 0, &metadata, srf_out);
 	if (ret) {
 		DRM_ERROR("Failed to allocate proxy content buffer\n");
 		return ret;
@@ -2516,7 +2509,7 @@ int vmw_kms_update_proxy(struct vmw_resource *res,
 			 int increment)
 {
 	struct vmw_private *dev_priv = res->dev_priv;
-	struct drm_vmw_size *size = &vmw_res_to_srf(res)->base_size;
+	struct drm_vmw_size *size = &vmw_res_to_srf(res)->metadata.base_size;
 	struct {
 		SVGA3dCmdHeader header;
 		SVGA3dCmdUpdateGBImage body;

drivers/gpu/drm/vmwgfx/vmwgfx_mob.c

@@ -320,7 +320,7 @@ int vmw_otables_setup(struct vmw_private *dev_priv)
 	struct vmw_otable **otables = &dev_priv->otable_batch.otables;
 	int ret;
 
-	if (dev_priv->has_dx) {
+	if (has_sm4_context(dev_priv)) {
 		*otables = kmemdup(dx_tables, sizeof(dx_tables), GFP_KERNEL);
 		if (!(*otables))
 			return -ENOMEM;

drivers/gpu/drm/vmwgfx/vmwgfx_so.c

@@ -319,7 +319,8 @@ int vmw_view_add(struct vmw_cmdbuf_res_manager *man,
 	static const size_t vmw_view_define_sizes[] = {
 		[vmw_view_sr] = sizeof(SVGA3dCmdDXDefineShaderResourceView),
 		[vmw_view_rt] = sizeof(SVGA3dCmdDXDefineRenderTargetView),
-		[vmw_view_ds] = sizeof(SVGA3dCmdDXDefineDepthStencilView)
+		[vmw_view_ds] = sizeof(SVGA3dCmdDXDefineDepthStencilView),
+		[vmw_view_ua] = sizeof(SVGA3dCmdDXDefineUAView)
 	};
 
 	struct vmw_private *dev_priv = ctx->dev_priv;
@@ -499,8 +500,8 @@ struct vmw_resource *vmw_view_lookup(struct vmw_cmdbuf_res_manager *man,
  * Each time a resource is put on the validation list as the result of a
  * view pointing to it, we need to determine whether that resource will
  * be dirtied (written to by the GPU) as a result of the corresponding
- * GPU operation. Currently only rendertarget- and depth-stencil views are
- * capable of dirtying its resource.
+ * GPU operation. Currently only rendertarget-, depth-stencil and unordered
+ * access views are capable of dirtying its resource.
  *
  * Return: Whether the view type of @res dirties the resource it points to.
  */
@@ -509,10 +510,11 @@ u32 vmw_view_dirtying(struct vmw_resource *res)
 	static u32 view_is_dirtying[vmw_view_max] = {
 		[vmw_view_rt] = VMW_RES_DIRTY_SET,
 		[vmw_view_ds] = VMW_RES_DIRTY_SET,
+		[vmw_view_ua] = VMW_RES_DIRTY_SET,
 	};
 
 	/* Update this function as we add more view types */
-	BUILD_BUG_ON(vmw_view_max != 3);
+	BUILD_BUG_ON(vmw_view_max != 4);
 	return view_is_dirtying[vmw_view(res)->view_type];
 }
 
@@ -520,12 +522,14 @@ const u32 vmw_view_destroy_cmds[] = {
 	[vmw_view_sr] = SVGA_3D_CMD_DX_DESTROY_SHADERRESOURCE_VIEW,
 	[vmw_view_rt] = SVGA_3D_CMD_DX_DESTROY_RENDERTARGET_VIEW,
 	[vmw_view_ds] = SVGA_3D_CMD_DX_DESTROY_DEPTHSTENCIL_VIEW,
+	[vmw_view_ua] = SVGA_3D_CMD_DX_DESTROY_UA_VIEW,
 };
 
 const SVGACOTableType vmw_view_cotables[] = {
 	[vmw_view_sr] = SVGA_COTABLE_SRVIEW,
 	[vmw_view_rt] = SVGA_COTABLE_RTVIEW,
 	[vmw_view_ds] = SVGA_COTABLE_DSVIEW,
+	[vmw_view_ua] = SVGA_COTABLE_UAVIEW,
 };
 
 const SVGACOTableType vmw_so_cotables[] = {

drivers/gpu/drm/vmwgfx/vmwgfx_so.h

@@ -30,6 +30,7 @@ enum vmw_view_type {
 	vmw_view_sr,
 	vmw_view_rt,
 	vmw_view_ds,
+	vmw_view_ua,
 	vmw_view_max,
 };
 
@@ -61,6 +62,7 @@ union vmw_view_destroy {
 	struct SVGA3dCmdDXDestroyRenderTargetView rtv;
 	struct SVGA3dCmdDXDestroyShaderResourceView srv;
 	struct SVGA3dCmdDXDestroyDepthStencilView dsv;
+	struct SVGA3dCmdDXDestroyUAView uav;
 	u32 view_id;
 };
 
@@ -87,6 +89,10 @@ static inline enum vmw_view_type vmw_view_cmd_to_type(u32 id)
 {
 	u32 tmp = (id - SVGA_3D_CMD_DX_DEFINE_SHADERRESOURCE_VIEW) / 2;
 
+	if (id == SVGA_3D_CMD_DX_DEFINE_UA_VIEW ||
+	    id == SVGA_3D_CMD_DX_DESTROY_UA_VIEW)
+		return vmw_view_ua;
+
 	if (tmp > (u32)vmw_view_max)
 		return vmw_view_max;
 
@@ -123,6 +129,7 @@ static inline enum vmw_so_type vmw_so_cmd_to_type(u32 id)
 	case SVGA_3D_CMD_DX_DESTROY_SAMPLER_STATE:
 		return vmw_so_ss;
 	case SVGA_3D_CMD_DX_DEFINE_STREAMOUTPUT:
+	case SVGA_3D_CMD_DX_DEFINE_STREAMOUTPUT_WITH_MOB:
 	case SVGA_3D_CMD_DX_DESTROY_STREAMOUTPUT:
 		return vmw_so_so;
 	default:

drivers/gpu/drm/vmwgfx/vmwgfx_stdu.c

@@ -590,7 +590,7 @@ static void vmw_stdu_bo_cpu_commit(struct vmw_kms_dirty *dirty)
 		return;
 
 	/* Assume we are blitting from Guest (bo) to Host (display_srf) */
-	dst_pitch = stdu->display_srf->base_size.width * stdu->cpp;
+	dst_pitch = stdu->display_srf->metadata.base_size.width * stdu->cpp;
 	dst_bo = &stdu->display_srf->res.backup->base;
 	dst_offset = ddirty->top * dst_pitch + ddirty->left * stdu->cpp;
 
@@ -1041,7 +1041,6 @@ vmw_stdu_primary_plane_prepare_fb(struct drm_plane *plane,
 	struct vmw_plane_state *vps = vmw_plane_state_to_vps(new_state);
 	enum stdu_content_type new_content_type;
 	struct vmw_framebuffer_surface *new_vfbs;
-	struct drm_crtc *crtc = new_state->crtc;
 	uint32_t hdisplay = new_state->crtc_w, vdisplay = new_state->crtc_h;
 	int ret;
 
@@ -1058,8 +1057,9 @@ vmw_stdu_primary_plane_prepare_fb(struct drm_plane *plane,
 	vfb = vmw_framebuffer_to_vfb(new_fb);
 	new_vfbs = (vfb->bo) ? NULL : vmw_framebuffer_to_vfbs(new_fb);
 
-	if (new_vfbs && new_vfbs->surface->base_size.width == hdisplay &&
-	    new_vfbs->surface->base_size.height == vdisplay)
+	if (new_vfbs &&
+	    new_vfbs->surface->metadata.base_size.width == hdisplay &&
+	    new_vfbs->surface->metadata.base_size.height == vdisplay)
 		new_content_type = SAME_AS_DISPLAY;
 	else if (vfb->bo)
 		new_content_type = SEPARATE_BO;
@@ -1067,12 +1067,11 @@ vmw_stdu_primary_plane_prepare_fb(struct drm_plane *plane,
 		new_content_type = SEPARATE_SURFACE;
 
 	if (new_content_type != SAME_AS_DISPLAY) {
-		struct vmw_surface content_srf;
-		struct drm_vmw_size display_base_size = {0};
+		struct vmw_surface_metadata metadata = {0};
 
-		display_base_size.width  = hdisplay;
-		display_base_size.height = vdisplay;
-		display_base_size.depth  = 1;
+		metadata.base_size.width = hdisplay;
+		metadata.base_size.height = vdisplay;
+		metadata.base_size.depth = 1;
 
 		/*
 		 * If content buffer is a buffer object, then we have to
@@ -1082,15 +1081,15 @@ vmw_stdu_primary_plane_prepare_fb(struct drm_plane *plane,
 
 			switch (new_fb->format->cpp[0]*8) {
 			case 32:
-				content_srf.format = SVGA3D_X8R8G8B8;
+				metadata.format = SVGA3D_X8R8G8B8;
 				break;
 
 			case 16:
-				content_srf.format = SVGA3D_R5G6B5;
+				metadata.format = SVGA3D_R5G6B5;
 				break;
 
 			case 8:
-				content_srf.format = SVGA3D_P8;
+				metadata.format = SVGA3D_P8;
 				break;
 
 			default:
@@ -1098,22 +1097,20 @@ vmw_stdu_primary_plane_prepare_fb(struct drm_plane *plane,
 				return -EINVAL;
 			}
 
-			content_srf.flags             = 0;
-			content_srf.mip_levels[0]     = 1;
-			content_srf.multisample_count = 0;
-			content_srf.multisample_pattern =
-				SVGA3D_MS_PATTERN_NONE;
-			content_srf.quality_level = SVGA3D_MS_QUALITY_NONE;
+			metadata.mip_levels[0] = 1;
+			metadata.num_sizes = 1;
+			metadata.scanout = true;
 		} else {
-			content_srf = *new_vfbs->surface;
+			metadata = new_vfbs->surface->metadata;
 		}
 
 		if (vps->surf) {
-			struct drm_vmw_size cur_base_size = vps->surf->base_size;
+			struct drm_vmw_size cur_base_size =
+				vps->surf->metadata.base_size;
 
-			if (cur_base_size.width != display_base_size.width ||
-			    cur_base_size.height != display_base_size.height ||
-			    vps->surf->format != content_srf.format) {
+			if (cur_base_size.width != metadata.base_size.width ||
+			    cur_base_size.height != metadata.base_size.height ||
+			    vps->surf->metadata.format != metadata.format) {
 				WARN_ON(vps->pinned != 0);
 				vmw_surface_unreference(&vps->surf);
 			}
@@ -1121,20 +1118,8 @@ vmw_stdu_primary_plane_prepare_fb(struct drm_plane *plane,
 		}
 
 		if (!vps->surf) {
-			ret = vmw_surface_gb_priv_define
-				(crtc->dev,
-				 /* Kernel visible only */
-				 0,
-				 content_srf.flags,
-				 content_srf.format,
-				 true,  /* a scanout buffer */
-				 content_srf.mip_levels[0],
-				 content_srf.multisample_count,
-				 0,
-				 display_base_size,
-				 content_srf.multisample_pattern,
-				 content_srf.quality_level,
-				 &vps->surf);
+			ret = vmw_gb_surface_define(dev_priv, 0, &metadata,
+						    &vps->surf);
 			if (ret != 0) {
 				DRM_ERROR("Couldn't allocate STDU surface.\n");
 				return ret;
@@ -1311,7 +1296,7 @@ vmw_stdu_bo_populate_update_cpu(struct vmw_du_update_plane  *update, void *cmd,
 	diff.cpp = stdu->cpp;
 
 	dst_bo = &stdu->display_srf->res.backup->base;
-	dst_pitch = stdu->display_srf->base_size.width * stdu->cpp;
+	dst_pitch = stdu->display_srf->metadata.base_size.width * stdu->cpp;
 	dst_offset = bb->y1 * dst_pitch + bb->x1 * stdu->cpp;
 
 	src_bo = &vfbbo->buffer->base;

include/uapi/drm/vmwgfx_drm.h

@@ -86,6 +86,9 @@ extern "C" {
  *
  * DRM_VMW_PARAM_SM4_1
  * SM4_1 support is enabled.
+ *
+ * DRM_VMW_PARAM_SM5
+ * SM5 support is enabled.
  */
 
 #define DRM_VMW_PARAM_NUM_STREAMS      0
@@ -103,6 +106,7 @@ extern "C" {
 #define DRM_VMW_PARAM_DX               12
 #define DRM_VMW_PARAM_HW_CAPS2         13
 #define DRM_VMW_PARAM_SM4_1            14
+#define DRM_VMW_PARAM_SM5              15
 
 /**
  * enum drm_vmw_handle_type - handle type for ref ioctls
@@ -1133,7 +1137,7 @@ struct drm_vmw_handle_close_arg {
  * svga3d surface flags split into 2, upper half and lower half.
  */
 enum drm_vmw_surface_version {
-	drm_vmw_gb_surface_v1
+	drm_vmw_gb_surface_v1,
 };
 
 /**
@@ -1144,6 +1148,7 @@ enum drm_vmw_surface_version {
  * @svga3d_flags_upper_32_bits: Upper 32 bits of svga3d flags.
  * @multisample_pattern: Multisampling pattern when msaa is supported.
  * @quality_level: Precision settings for each sample.
+ * @buffer_byte_stride: Buffer byte stride.
  * @must_be_zero: Reserved for future usage.
  *
  * Input argument to the  DRM_VMW_GB_SURFACE_CREATE_EXT Ioctl.
@@ -1152,10 +1157,11 @@ enum drm_vmw_surface_version {
 struct drm_vmw_gb_surface_create_ext_req {
 	struct drm_vmw_gb_surface_create_req base;
 	enum drm_vmw_surface_version version;
-	uint32_t svga3d_flags_upper_32_bits;
-	SVGA3dMSPattern multisample_pattern;
-	SVGA3dMSQualityLevel quality_level;
-	uint64_t must_be_zero;
+	__u32 svga3d_flags_upper_32_bits;
+	__u32 multisample_pattern;
+	__u32 quality_level;
+	__u32 buffer_byte_stride;
+	__u32 must_be_zero;
 };
 
 /**