drm/ttm: flip the switch for driver allocated resources v2

Instead of both driver and TTM allocating memory finalize embedding the
ttm_resource object as base into the driver backends.

v2: fix typo in vmwgfx grid mgr and double init in amdgpu_vram_mgr.c

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210602100914.46246-10-christian.koenig@amd.com

1 files changed, 27 insertions, 33 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c
index 5ebfaed37e47..9a6df02477ce 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c
@@ -219,19 +219,20 @@ static u64 amdgpu_vram_mgr_vis_size(struct amdgpu_device *adev,
 u64 amdgpu_vram_mgr_bo_visible_size(struct amdgpu_bo *bo)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
-	struct ttm_resource *mem = bo->tbo.resource;
-	struct drm_mm_node *nodes = mem->mm_node;
-	unsigned pages = mem->num_pages;
+	struct ttm_resource *res = bo->tbo.resource;
+	unsigned pages = res->num_pages;
+	struct drm_mm_node *mm;
 	u64 usage;
 
 	if (amdgpu_gmc_vram_full_visible(&adev->gmc))
 		return amdgpu_bo_size(bo);
 
-	if (mem->start >= adev->gmc.visible_vram_size >> PAGE_SHIFT)
+	if (res->start >= adev->gmc.visible_vram_size >> PAGE_SHIFT)
 		return 0;
 
-	for (usage = 0; nodes && pages; pages -= nodes->size, nodes++)
-		usage += amdgpu_vram_mgr_vis_size(adev, nodes);
+	mm = &container_of(res, struct ttm_range_mgr_node, base)->mm_nodes[0];
+	for (usage = 0; pages; pages -= mm->size, mm++)
+		usage += amdgpu_vram_mgr_vis_size(adev, mm);
 
 	return usage;
 }
@@ -367,7 +368,7 @@ static void amdgpu_vram_mgr_virt_start(struct ttm_resource *mem,
 static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 			       struct ttm_buffer_object *tbo,
 			       const struct ttm_place *place,
-			       struct ttm_resource *mem)
+			       struct ttm_resource **res)
 {
 	unsigned long lpfn, num_nodes, pages_per_node, pages_left, pages;
 	struct amdgpu_vram_mgr *mgr = to_vram_mgr(man);
@@ -388,7 +389,7 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 		max_bytes -= AMDGPU_VM_RESERVED_VRAM;
 
 	/* bail out quickly if there's likely not enough VRAM for this BO */
-	mem_bytes = (u64)mem->num_pages << PAGE_SHIFT;
+	mem_bytes = tbo->base.size;
 	if (atomic64_add_return(mem_bytes, &mgr->usage) > max_bytes) {
 		r = -ENOSPC;
 		goto error_sub;
@@ -406,7 +407,7 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 #endif
 		pages_per_node = max_t(uint32_t, pages_per_node,
 				       tbo->page_alignment);
-		num_nodes = DIV_ROUND_UP(mem->num_pages, pages_per_node);
+		num_nodes = DIV_ROUND_UP(PFN_UP(mem_bytes), pages_per_node);
 	}
 
 	node = kvmalloc(struct_size(node, mm_nodes, num_nodes),
@@ -422,8 +423,7 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 	if (place->flags & TTM_PL_FLAG_TOPDOWN)
 		mode = DRM_MM_INSERT_HIGH;
 
-	mem->start = 0;
-	pages_left = mem->num_pages;
+	pages_left = node->base.num_pages;
 
 	/* Limit maximum size to 2GB due to SG table limitations */
 	pages = min(pages_left, 2UL << (30 - PAGE_SHIFT));
@@ -451,7 +451,7 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 		}
 
 		vis_usage += amdgpu_vram_mgr_vis_size(adev, &node->mm_nodes[i]);
-		amdgpu_vram_mgr_virt_start(mem, &node->mm_nodes[i]);
+		amdgpu_vram_mgr_virt_start(&node->base, &node->mm_nodes[i]);
 		pages_left -= pages;
 		++i;
 
@@ -461,10 +461,10 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 	spin_unlock(&mgr->lock);
 
 	if (i == 1)
-		mem->placement |= TTM_PL_FLAG_CONTIGUOUS;
+		node->base.placement |= TTM_PL_FLAG_CONTIGUOUS;
 
 	atomic64_add(vis_usage, &mgr->vis_usage);
-	mem->mm_node = &node->mm_nodes[0];
+	*res = &node->base;
 	return 0;
 
 error_free:
@@ -487,28 +487,22 @@ error_sub:
  * Free the allocated VRAM again.
  */
 static void amdgpu_vram_mgr_del(struct ttm_resource_manager *man,
-				struct ttm_resource *mem)
+				struct ttm_resource *res)
 {
+	struct ttm_range_mgr_node *node = to_ttm_range_mgr_node(res);
 	struct amdgpu_vram_mgr *mgr = to_vram_mgr(man);
 	struct amdgpu_device *adev = to_amdgpu_device(mgr);
-	struct ttm_range_mgr_node *node;
 	uint64_t usage = 0, vis_usage = 0;
-	unsigned pages = mem->num_pages;
-	struct drm_mm_node *nodes;
-
-	if (!mem->mm_node)
-		return;
-
-	node = to_ttm_range_mgr_node(mem);
-	nodes = &node->mm_nodes[0];
+	unsigned i, pages;
 
 	spin_lock(&mgr->lock);
-	while (pages) {
-		pages -= nodes->size;
-		drm_mm_remove_node(nodes);
-		usage += nodes->size << PAGE_SHIFT;
-		vis_usage += amdgpu_vram_mgr_vis_size(adev, nodes);
-		++nodes;
+	for (i = 0, pages = res->num_pages; pages;
+	     pages -= node->mm_nodes[i].size, ++i) {
+		struct drm_mm_node *mm = &node->mm_nodes[i];
+
+		drm_mm_remove_node(mm);
+		usage += mm->size << PAGE_SHIFT;
+		vis_usage += amdgpu_vram_mgr_vis_size(adev, mm);
 	}
 	amdgpu_vram_mgr_do_reserve(man);
 	spin_unlock(&mgr->lock);
@@ -533,7 +527,7 @@ static void amdgpu_vram_mgr_del(struct ttm_resource_manager *man,
  * Allocate and fill a sg table from a VRAM allocation.
  */
 int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev,
-			      struct ttm_resource *mem,
+			      struct ttm_resource *res,
 			      u64 offset, u64 length,
 			      struct device *dev,
 			      enum dma_data_direction dir,
@@ -549,7 +543,7 @@ int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev,
 		return -ENOMEM;
 
 	/* Determine the number of DRM_MM nodes to export */
-	amdgpu_res_first(mem, offset, length, &cursor);
+	amdgpu_res_first(res, offset, length, &cursor);
 	while (cursor.remaining) {
 		num_entries++;
 		amdgpu_res_next(&cursor, cursor.size);
@@ -569,7 +563,7 @@ int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev,
 	 * and the number of bytes from it. Access the following
 	 * DRM_MM node(s) if more buffer needs to exported
 	 */
-	amdgpu_res_first(mem, offset, length, &cursor);
+	amdgpu_res_first(res, offset, length, &cursor);
 	for_each_sgtable_sg((*sgt), sg, i) {
 		phys_addr_t phys = cursor.start + adev->gmc.aper_base;
 		size_t size = cursor.size;