drm/vmwgfx: Add support for streamoutput with mob commands

With SM5 capability a new version of streamoutput is supported by device
which need backing mob and a new field. With this change the new command
is supported in command buffer.

v2: Also track streamoutput context binding in binding manager.

v3: Track only one streamoutput as only one can be set to context.
v4: Fix comment typos

Signed-off-by: Deepak Rawat <drawat.floss@gmail.com>
Signed-off-by: Neha Bhende <bhenden@vmware.com>
Reviewed-by: Thomas Hellström (VMware) <thomas_os@shipmail.org>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Signed-off-by: Roland Scheidegger <sroland@vmware.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/vmwgfx_binding.c

@@ -78,6 +78,7 @@
  * @index_buffer: Index buffer binding.
  * @per_shader: Per shader-type bindings.
  * @ua_views: UAV bindings.
+ * @so_state: StreamOutput bindings.
  * @dirty: Bitmap tracking per binding-type changes that have not yet
  * been emitted to the device.
  * @dirty_vb: Bitmap tracking individual vertex buffer binding changes that
@@ -103,6 +104,7 @@ struct vmw_ctx_binding_state {
 	struct vmw_ctx_bindinfo_ib index_buffer;
 	struct vmw_dx_shader_bindings per_shader[SVGA3D_NUM_SHADERTYPE];
 	struct vmw_ctx_bindinfo_uav ua_views[VMW_MAX_UAV_BIND_TYPE];
+	struct vmw_ctx_bindinfo_so so_state;
 
 	unsigned long dirty;
 	DECLARE_BITMAP(dirty_vb, SVGA3D_DX_MAX_VERTEXBUFFERS);
@@ -127,6 +129,7 @@ static int vmw_binding_scrub_ib(struct vmw_ctx_bindinfo *bi, bool rebind);
 static int vmw_binding_scrub_vb(struct vmw_ctx_bindinfo *bi, bool rebind);
 static int vmw_binding_scrub_uav(struct vmw_ctx_bindinfo *bi, bool rebind);
 static int vmw_binding_scrub_cs_uav(struct vmw_ctx_bindinfo *bi, bool rebind);
+static int vmw_binding_scrub_so(struct vmw_ctx_bindinfo *bi, bool rebind);
 
 static void vmw_binding_build_asserts(void) __attribute__ ((unused));
 
@@ -202,6 +205,9 @@ static const size_t vmw_binding_uav_offsets[] = {
 static const size_t vmw_binding_cs_uav_offsets[] = {
 	offsetof(struct vmw_ctx_binding_state, ua_views[1].views),
 };
+static const size_t vmw_binding_so_offsets[] = {
+	offsetof(struct vmw_ctx_binding_state, so_state),
+};
 
 static const struct vmw_binding_info vmw_binding_infos[] = {
 	[vmw_ctx_binding_shader] = {
@@ -256,6 +262,10 @@ static const struct vmw_binding_info vmw_binding_infos[] = {
 		.size = sizeof(struct vmw_ctx_bindinfo_view),
 		.offsets = vmw_binding_cs_uav_offsets,
 		.scrub_func = vmw_binding_scrub_cs_uav},
+	[vmw_ctx_binding_so] = {
+		.size = sizeof(struct vmw_ctx_bindinfo_so),
+		.offsets = vmw_binding_so_offsets,
+		.scrub_func = vmw_binding_scrub_so},
 };
 
 /**
@@ -1290,6 +1300,33 @@ static int vmw_binding_scrub_cs_uav(struct vmw_ctx_bindinfo *bi, bool rebind)
 	return 0;
 }
 
+/**
+ * vmw_binding_scrub_so - Scrub a streamoutput binding from context.
+ * @bi: Single binding information.
+ * @rebind: Whether to issue a bind instead of scrub command.
+ */
+static int vmw_binding_scrub_so(struct vmw_ctx_bindinfo *bi, bool rebind)
+{
+	struct vmw_ctx_bindinfo_so *binding =
+		container_of(bi, typeof(*binding), bi);
+	struct vmw_private *dev_priv = bi->ctx->dev_priv;
+	struct {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdDXSetStreamOutput body;
+	} *cmd;
+
+	cmd = VMW_FIFO_RESERVE_DX(dev_priv, sizeof(*cmd), bi->ctx->id);
+	if (!cmd)
+		return -ENOMEM;
+
+	cmd->header.id = SVGA_3D_CMD_DX_SET_STREAMOUTPUT;
+	cmd->header.size = sizeof(cmd->body);
+	cmd->body.soid = rebind ? bi->res->id : SVGA3D_INVALID_ID;
+	vmw_fifo_commit(dev_priv, sizeof(*cmd));
+
+	return 0;
+}
+
 /**
  * vmw_binding_state_alloc - Allocate a struct vmw_ctx_binding_state with
  * memory accounting.
@@ -1393,7 +1430,7 @@ u32 vmw_binding_dirtying(enum vmw_ctx_binding_type binding_type)
 	};
 
 	/* Review this function as new bindings are added. */
-	BUILD_BUG_ON(vmw_ctx_binding_max != 13);
+	BUILD_BUG_ON(vmw_ctx_binding_max != 14);
 	return is_binding_dirtying[binding_type];
 }
 

drivers/gpu/drm/vmwgfx/vmwgfx_binding.h

@@ -55,6 +55,7 @@ enum vmw_ctx_binding_type {
 	vmw_ctx_binding_ib,
 	vmw_ctx_binding_uav,
 	vmw_ctx_binding_cs_uav,
+	vmw_ctx_binding_so,
 	vmw_ctx_binding_max
 };
 
@@ -203,6 +204,16 @@ struct vmw_ctx_bindinfo_uav {
 	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);

drivers/gpu/drm/vmwgfx/vmwgfx_cotable.c

@@ -80,7 +80,7 @@ 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(SVGACOTableDXUAViewEntry), &vmw_view_cotable_list_destroy}

drivers/gpu/drm/vmwgfx/vmwgfx_drv.h

@@ -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;
@@ -1307,6 +1309,24 @@ 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_execbuf.c

@@ -2948,6 +2948,169 @@ static int vmw_cmd_set_cs_uav(struct vmw_private *dev_priv,
 	return ret;
 }
 
+static int vmw_cmd_dx_define_streamoutput(struct vmw_private *dev_priv,
+					  struct vmw_sw_context *sw_context,
+					  SVGA3dCmdHeader *header)
+{
+	struct vmw_ctx_validation_info *ctx_node = sw_context->dx_ctx_node;
+	struct vmw_resource *res;
+	struct {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdDXDefineStreamOutputWithMob body;
+	} *cmd = container_of(header, typeof(*cmd), header);
+	int ret;
+
+	if (!has_sm5_context(dev_priv))
+		return -EINVAL;
+
+	if (!ctx_node) {
+		DRM_ERROR("DX Context not set.\n");
+		return -EINVAL;
+	}
+
+	res = vmw_context_cotable(ctx_node->ctx, SVGA_COTABLE_STREAMOUTPUT);
+	ret = vmw_cotable_notify(res, cmd->body.soid);
+	if (ret)
+		return ret;
+
+	return vmw_dx_streamoutput_add(sw_context->man, ctx_node->ctx,
+				       cmd->body.soid,
+				       &sw_context->staged_cmd_res);
+}
+
+static int vmw_cmd_dx_destroy_streamoutput(struct vmw_private *dev_priv,
+					   struct vmw_sw_context *sw_context,
+					   SVGA3dCmdHeader *header)
+{
+	struct vmw_ctx_validation_info *ctx_node = sw_context->dx_ctx_node;
+	struct vmw_resource *res;
+	struct {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdDXDestroyStreamOutput body;
+	} *cmd = container_of(header, typeof(*cmd), header);
+
+	if (!ctx_node) {
+		DRM_ERROR("DX Context not set.\n");
+		return -EINVAL;
+	}
+
+	/*
+	 * When device does not support SM5 then streamoutput with mob command is
+	 * not available to user-space. Simply return in this case.
+	 */
+	if (!has_sm5_context(dev_priv))
+		return 0;
+
+	/*
+	 * With SM5 capable device if lookup fails then user-space probably used
+	 * old streamoutput define command. Return without an error.
+	 */
+	res = vmw_dx_streamoutput_lookup(vmw_context_res_man(ctx_node->ctx),
+					 cmd->body.soid);
+	if (IS_ERR(res))
+		return 0;
+
+	return vmw_dx_streamoutput_remove(sw_context->man, cmd->body.soid,
+					  &sw_context->staged_cmd_res);
+}
+
+static int vmw_cmd_dx_bind_streamoutput(struct vmw_private *dev_priv,
+					struct vmw_sw_context *sw_context,
+					SVGA3dCmdHeader *header)
+{
+	struct vmw_ctx_validation_info *ctx_node = sw_context->dx_ctx_node;
+	struct vmw_resource *res;
+	struct {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdDXBindStreamOutput body;
+	} *cmd = container_of(header, typeof(*cmd), header);
+	int ret;
+
+	if (!has_sm5_context(dev_priv))
+		return -EINVAL;
+
+	if (!ctx_node) {
+		DRM_ERROR("DX Context not set.\n");
+		return -EINVAL;
+	}
+
+	res = vmw_dx_streamoutput_lookup(vmw_context_res_man(ctx_node->ctx),
+					 cmd->body.soid);
+	if (IS_ERR(res)) {
+		DRM_ERROR("Cound not find streamoutput to bind.\n");
+		return PTR_ERR(res);
+	}
+
+	vmw_dx_streamoutput_set_size(res, cmd->body.sizeInBytes);
+
+	ret = vmw_execbuf_res_noctx_val_add(sw_context, res,
+					    VMW_RES_DIRTY_NONE);
+	if (ret) {
+		DRM_ERROR("Error creating resource validation node.\n");
+		return ret;
+	}
+
+	return vmw_cmd_res_switch_backup(dev_priv, sw_context, res,
+					 &cmd->body.mobid,
+					 cmd->body.offsetInBytes);
+}
+
+static int vmw_cmd_dx_set_streamoutput(struct vmw_private *dev_priv,
+				       struct vmw_sw_context *sw_context,
+				       SVGA3dCmdHeader *header)
+{
+	struct vmw_ctx_validation_info *ctx_node = sw_context->dx_ctx_node;
+	struct vmw_resource *res;
+	struct vmw_ctx_bindinfo_so binding;
+	struct {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdDXSetStreamOutput body;
+	} *cmd = container_of(header, typeof(*cmd), header);
+	int ret;
+
+	if (!ctx_node) {
+		DRM_ERROR("DX Context not set.\n");
+		return -EINVAL;
+	}
+
+	if (cmd->body.soid == SVGA3D_INVALID_ID)
+		return 0;
+
+	/*
+	 * When device does not support SM5 then streamoutput with mob command is
+	 * not available to user-space. Simply return in this case.
+	 */
+	if (!has_sm5_context(dev_priv))
+		return 0;
+
+	/*
+	 * With SM5 capable device if lookup fails then user-space probably used
+	 * old streamoutput define command. Return without an error.
+	 */
+	res = vmw_dx_streamoutput_lookup(vmw_context_res_man(ctx_node->ctx),
+					 cmd->body.soid);
+	if (IS_ERR(res)) {
+		return 0;
+	}
+
+	ret = vmw_execbuf_res_noctx_val_add(sw_context, res,
+					    VMW_RES_DIRTY_NONE);
+	if (ret) {
+		DRM_ERROR("Error creating resource validation node.\n");
+		return ret;
+	}
+
+	binding.bi.ctx = ctx_node->ctx;
+	binding.bi.res = res;
+	binding.bi.bt = vmw_ctx_binding_so;
+	binding.slot = 0; /* Only one SO set to context at a time. */
+
+	vmw_binding_add(sw_context->dx_ctx_node->staged, &binding.bi, 0,
+			binding.slot);
+
+	return ret;
+}
+
 static int vmw_cmd_indexed_instanced_indirect(struct vmw_private *dev_priv,
 					      struct vmw_sw_context *sw_context,
 					      SVGA3dCmdHeader *header)
@@ -3330,9 +3493,9 @@ static const struct vmw_cmd_entry vmw_cmd_entries[SVGA_3D_CMD_MAX] = {
 	VMW_CMD_DEF(SVGA_3D_CMD_DX_DEFINE_STREAMOUTPUT,
 		    &vmw_cmd_dx_so_define, true, false, true),
 	VMW_CMD_DEF(SVGA_3D_CMD_DX_DESTROY_STREAMOUTPUT,
-		    &vmw_cmd_dx_cid_check, true, false, true),
-	VMW_CMD_DEF(SVGA_3D_CMD_DX_SET_STREAMOUTPUT, &vmw_cmd_dx_cid_check,
-		    true, false, true),
+		    &vmw_cmd_dx_destroy_streamoutput, true, false, true),
+	VMW_CMD_DEF(SVGA_3D_CMD_DX_SET_STREAMOUTPUT,
+		    &vmw_cmd_dx_set_streamoutput, true, false, true),
 	VMW_CMD_DEF(SVGA_3D_CMD_DX_SET_SOTARGETS,
 		    &vmw_cmd_dx_set_so_targets, true, false, true),
 	VMW_CMD_DEF(SVGA_3D_CMD_DX_SET_INPUT_LAYOUT,
@@ -3375,6 +3538,10 @@ static const struct vmw_cmd_entry vmw_cmd_entries[SVGA_3D_CMD_MAX] = {
 		    false, true),
 	VMW_CMD_DEF(SVGA_3D_CMD_DX_DEFINE_DEPTHSTENCIL_VIEW_V2,
 		    &vmw_cmd_sm5_view_define, true, false, true),
+	VMW_CMD_DEF(SVGA_3D_CMD_DX_DEFINE_STREAMOUTPUT_WITH_MOB,
+		    &vmw_cmd_dx_define_streamoutput, true, false, true),
+	VMW_CMD_DEF(SVGA_3D_CMD_DX_BIND_STREAMOUTPUT,
+		    &vmw_cmd_dx_bind_streamoutput, true, false, true),
 };
 
 bool vmw_cmd_describe(const void *buf, u32 *size, char const **cmd)

drivers/gpu/drm/vmwgfx/vmwgfx_so.h

@@ -129,6 +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: