drm/ttm: allow reserving more than one shared slot v3

Let's support simultaneous submissions to multiple engines.

v2: rename the field to num_shared and fix up all users
v3: rebased

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c

@@ -535,7 +535,7 @@ static void add_kgd_mem_to_kfd_bo_list(struct kgd_mem *mem,
 	struct amdgpu_bo *bo = mem->bo;
 
 	INIT_LIST_HEAD(&entry->head);
-	entry->shared = true;
+	entry->num_shared = 1;
 	entry->bo = &bo->tbo;
 	mutex_lock(&process_info->lock);
 	if (userptr)
@@ -676,7 +676,7 @@ static int reserve_bo_and_vm(struct kgd_mem *mem,
 
 	ctx->kfd_bo.priority = 0;
 	ctx->kfd_bo.tv.bo = &bo->tbo;
-	ctx->kfd_bo.tv.shared = true;
+	ctx->kfd_bo.tv.num_shared = 1;
 	ctx->kfd_bo.user_pages = NULL;
 	list_add(&ctx->kfd_bo.tv.head, &ctx->list);
 
@@ -740,7 +740,7 @@ static int reserve_bo_and_cond_vms(struct kgd_mem *mem,
 
 	ctx->kfd_bo.priority = 0;
 	ctx->kfd_bo.tv.bo = &bo->tbo;
-	ctx->kfd_bo.tv.shared = true;
+	ctx->kfd_bo.tv.num_shared = 1;
 	ctx->kfd_bo.user_pages = NULL;
 	list_add(&ctx->kfd_bo.tv.head, &ctx->list);
 
@@ -1830,7 +1830,7 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info)
 			    validate_list.head) {
 		list_add_tail(&mem->resv_list.head, &resv_list);
 		mem->resv_list.bo = mem->validate_list.bo;
-		mem->resv_list.shared = mem->validate_list.shared;
+		mem->resv_list.num_shared = mem->validate_list.num_shared;
 	}
 
 	/* Reserve all BOs and page tables for validation */
@@ -2049,7 +2049,7 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef)
 
 		list_add_tail(&mem->resv_list.head, &ctx.list);
 		mem->resv_list.bo = mem->validate_list.bo;
-		mem->resv_list.shared = mem->validate_list.shared;
+		mem->resv_list.num_shared = mem->validate_list.num_shared;
 	}
 
 	ret = ttm_eu_reserve_buffers(&ctx.ticket, &ctx.list,

drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c

@@ -118,7 +118,7 @@ int amdgpu_bo_list_create(struct amdgpu_device *adev, struct drm_file *filp,
 		entry->priority = min(info[i].bo_priority,
 				      AMDGPU_BO_LIST_MAX_PRIORITY);
 		entry->tv.bo = &bo->tbo;
-		entry->tv.shared = !bo->prime_shared_count;
+		entry->tv.num_shared = !bo->prime_shared_count;
 
 		if (bo->preferred_domains == AMDGPU_GEM_DOMAIN_GDS)
 			list->gds_obj = bo;

drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c

@@ -50,7 +50,7 @@ static int amdgpu_cs_user_fence_chunk(struct amdgpu_cs_parser *p,
 	bo = amdgpu_bo_ref(gem_to_amdgpu_bo(gobj));
 	p->uf_entry.priority = 0;
 	p->uf_entry.tv.bo = &bo->tbo;
-	p->uf_entry.tv.shared = true;
+	p->uf_entry.tv.num_shared = 1;
 	p->uf_entry.user_pages = NULL;
 
 	drm_gem_object_put_unlocked(gobj);

drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c

@@ -74,7 +74,7 @@ int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	INIT_LIST_HEAD(&list);
 	INIT_LIST_HEAD(&csa_tv.head);
 	csa_tv.bo = &bo->tbo;
-	csa_tv.shared = true;
+	csa_tv.num_shared = 1;
 
 	list_add(&csa_tv.head, &list);
 	amdgpu_vm_get_pd_bo(vm, &list, &pd);

drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c

@@ -169,7 +169,7 @@ void amdgpu_gem_object_close(struct drm_gem_object *obj,
 	INIT_LIST_HEAD(&duplicates);
 
 	tv.bo = &bo->tbo;
-	tv.shared = true;
+	tv.num_shared = 1;
 	list_add(&tv.head, &list);
 
 	amdgpu_vm_get_pd_bo(vm, &list, &vm_pd);
@@ -604,7 +604,10 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 			return -ENOENT;
 		abo = gem_to_amdgpu_bo(gobj);
 		tv.bo = &abo->tbo;
-		tv.shared = !!(abo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID);
+		if (abo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID)
+			tv.num_shared = 1;
+		else
+			tv.num_shared = 0;
 		list_add(&tv.head, &list);
 	} else {
 		gobj = NULL;

drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c

@@ -617,7 +617,7 @@ void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm,
 {
 	entry->priority = 0;
 	entry->tv.bo = &vm->root.base.bo->tbo;
-	entry->tv.shared = true;
+	entry->tv.num_shared = 1;
 	entry->user_pages = NULL;
 	list_add(&entry->tv.head, validated);
 }

drivers/gpu/drm/qxl/qxl_release.c

@@ -217,7 +217,7 @@ int qxl_release_list_add(struct qxl_release *release, struct qxl_bo *bo)
 
 	qxl_bo_ref(bo);
 	entry->tv.bo = &bo->tbo;
-	entry->tv.shared = false;
+	entry->tv.num_shared = 0;
 	list_add_tail(&entry->tv.head, &release->bos);
 	return 0;
 }

drivers/gpu/drm/radeon/radeon_cs.c

@@ -178,7 +178,7 @@ static int radeon_cs_parser_relocs(struct radeon_cs_parser *p)
 		}
 
 		p->relocs[i].tv.bo = &p->relocs[i].robj->tbo;
-		p->relocs[i].tv.shared = !r->write_domain;
+		p->relocs[i].tv.num_shared = !r->write_domain;
 
 		radeon_cs_buckets_add(&buckets, &p->relocs[i].tv.head,
 				      priority);
@@ -253,7 +253,7 @@ static int radeon_cs_sync_rings(struct radeon_cs_parser *p)
 
 		resv = reloc->robj->tbo.resv;
 		r = radeon_sync_resv(p->rdev, &p->ib.sync, resv,
-				     reloc->tv.shared);
+				     reloc->tv.num_shared);
 		if (r)
 			return r;
 	}

drivers/gpu/drm/radeon/radeon_gem.c

@@ -552,7 +552,7 @@ static void radeon_gem_va_update_vm(struct radeon_device *rdev,
 	INIT_LIST_HEAD(&list);
 
 	tv.bo = &bo_va->bo->tbo;
-	tv.shared = true;
+	tv.num_shared = 1;
 	list_add(&tv.head, &list);
 
 	vm_bos = radeon_vm_get_bos(rdev, bo_va->vm, &list);

drivers/gpu/drm/radeon/radeon_vm.c

@@ -142,7 +142,7 @@ struct radeon_bo_list *radeon_vm_get_bos(struct radeon_device *rdev,
 	list[0].preferred_domains = RADEON_GEM_DOMAIN_VRAM;
 	list[0].allowed_domains = RADEON_GEM_DOMAIN_VRAM;
 	list[0].tv.bo = &vm->page_directory->tbo;
-	list[0].tv.shared = true;
+	list[0].tv.num_shared = 1;
 	list[0].tiling_flags = 0;
 	list_add(&list[0].tv.head, head);
 
@@ -154,7 +154,7 @@ struct radeon_bo_list *radeon_vm_get_bos(struct radeon_device *rdev,
 		list[idx].preferred_domains = RADEON_GEM_DOMAIN_VRAM;
 		list[idx].allowed_domains = RADEON_GEM_DOMAIN_VRAM;
 		list[idx].tv.bo = &list[idx].robj->tbo;
-		list[idx].tv.shared = true;
+		list[idx].tv.num_shared = 1;
 		list[idx].tiling_flags = 0;
 		list_add(&list[idx++].tv.head, head);
 	}

drivers/gpu/drm/ttm/ttm_execbuf_util.c

@@ -126,10 +126,11 @@ int ttm_eu_reserve_buffers(struct ww_acquire_ctx *ticket,
 		}
 
 		if (!ret) {
-			if (!entry->shared)
+			if (!entry->num_shared)
 				continue;
 
-			ret = reservation_object_reserve_shared(bo->resv, 1);
+			ret = reservation_object_reserve_shared(bo->resv,
+								entry->num_shared);
 			if (!ret)
 				continue;
 		}
@@ -150,8 +151,9 @@ int ttm_eu_reserve_buffers(struct ww_acquire_ctx *ticket,
 			}
 		}
 
-		if (!ret && entry->shared)
+		if (!ret && entry->num_shared)
-			ret = reservation_object_reserve_shared(bo->resv, 1);
+			ret = reservation_object_reserve_shared(bo->resv,
+								entry->num_shared);
 
 		if (unlikely(ret != 0)) {
 			if (ret == -EINTR)
@@ -199,7 +201,7 @@ void ttm_eu_fence_buffer_objects(struct ww_acquire_ctx *ticket,
 
 	list_for_each_entry(entry, list, head) {
 		bo = entry->bo;
-		if (entry->shared)
+		if (entry->num_shared)
 			reservation_object_add_shared_fence(bo->resv, fence);
 		else
 			reservation_object_add_excl_fence(bo->resv, fence);

drivers/gpu/drm/vmwgfx/vmwgfx_resource.c

@@ -85,7 +85,7 @@ static void vmw_resource_release(struct kref *kref)
 			struct ttm_validate_buffer val_buf;
 
 			val_buf.bo = bo;
-			val_buf.shared = false;
+			val_buf.num_shared = 0;
 			res->func->unbind(res, false, &val_buf);
 		}
 		res->backup_dirty = false;
@@ -462,7 +462,7 @@ vmw_resource_check_buffer(struct ww_acquire_ctx *ticket,
 
 	INIT_LIST_HEAD(&val_list);
 	val_buf->bo = ttm_bo_reference(&res->backup->base);
-	val_buf->shared = false;
+	val_buf->num_shared = 0;
 	list_add_tail(&val_buf->head, &val_list);
 	ret = ttm_eu_reserve_buffers(ticket, &val_list, interruptible, NULL);
 	if (unlikely(ret != 0))
@@ -565,7 +565,7 @@ static int vmw_resource_do_evict(struct ww_acquire_ctx *ticket,
 	BUG_ON(!func->may_evict);
 
 	val_buf.bo = NULL;
-	val_buf.shared = false;
+	val_buf.num_shared = 0;
 	ret = vmw_resource_check_buffer(ticket, res, interruptible, &val_buf);
 	if (unlikely(ret != 0))
 		return ret;
@@ -614,7 +614,7 @@ int vmw_resource_validate(struct vmw_resource *res, bool intr)
 		return 0;
 
 	val_buf.bo = NULL;
-	val_buf.shared = false;
+	val_buf.num_shared = 0;
 	if (res->backup)
 		val_buf.bo = &res->backup->base;
 	do {
@@ -685,7 +685,7 @@ void vmw_resource_unbind_list(struct vmw_buffer_object *vbo)
 	struct vmw_resource *res, *next;
 	struct ttm_validate_buffer val_buf = {
 		.bo = &vbo->base,
-		.shared = false
+		.num_shared = 0
 	};
 
 	lockdep_assert_held(&vbo->base.resv->lock.base);

drivers/gpu/drm/vmwgfx/vmwgfx_validation.c

@@ -266,7 +266,7 @@ int vmw_validation_add_bo(struct vmw_validation_context *ctx,
 		val_buf->bo = ttm_bo_get_unless_zero(&vbo->base);
 		if (!val_buf->bo)
 			return -ESRCH;
-		val_buf->shared = false;
+		val_buf->num_shared = 0;
 		list_add_tail(&val_buf->head, &ctx->bo_list);
 		bo_node->as_mob = as_mob;
 		bo_node->cpu_blit = cpu_blit;

include/drm/ttm/ttm_execbuf_util.h

@@ -40,13 +40,13 @@
  *
  * @head:           list head for thread-private list.
  * @bo:             refcounted buffer object pointer.
- * @shared:         should the fence be added shared?
+ * @num_shared:     How many shared fences we want to add.
  */
 
 struct ttm_validate_buffer {
 	struct list_head head;
 	struct ttm_buffer_object *bo;
-	bool shared;
+	unsigned int num_shared;
 };
 
 /**