drm/nouveau: create temp vmas for both src and dst of bo moves

Greatly simplifies a number of things, particularly once per-client GPU
address spaces are involved.

May add this back later once I know what things'll look like.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drivers/gpu/drm/nouveau/nouveau_bo.c

@@ -496,19 +496,12 @@ static int
 nvc0_bo_move_m2mf(struct nouveau_channel *chan, struct ttm_buffer_object *bo,
 		  struct ttm_mem_reg *old_mem, struct ttm_mem_reg *new_mem)
 {
-	struct nouveau_mem *old_node = old_mem->mm_node;
-	struct nouveau_mem *new_node = new_mem->mm_node;
-	struct nouveau_bo *nvbo = nouveau_bo(bo);
+	struct nouveau_mem *node = old_mem->mm_node;
+	u64 src_offset = node->vma[0].offset;
+	u64 dst_offset = node->vma[1].offset;
 	u32 page_count = new_mem->num_pages;
-	u64 src_offset, dst_offset;
 	int ret;
 
-	src_offset = old_node->tmp_vma.offset;
-	if (new_node->tmp_vma.node)
-		dst_offset = new_node->tmp_vma.offset;
-	else
-		dst_offset = nvbo->vma.offset;
-
 	page_count = new_mem->num_pages;
 	while (page_count) {
 		int line_count = (page_count > 2047) ? 2047 : page_count;
@@ -542,19 +535,13 @@ static int
 nv50_bo_move_m2mf(struct nouveau_channel *chan, struct ttm_buffer_object *bo,
 		  struct ttm_mem_reg *old_mem, struct ttm_mem_reg *new_mem)
 {
-	struct nouveau_mem *old_node = old_mem->mm_node;
-	struct nouveau_mem *new_node = new_mem->mm_node;
+	struct nouveau_mem *node = old_mem->mm_node;
 	struct nouveau_bo *nvbo = nouveau_bo(bo);
 	u64 length = (new_mem->num_pages << PAGE_SHIFT);
-	u64 src_offset, dst_offset;
+	u64 src_offset = node->vma[0].offset;
+	u64 dst_offset = node->vma[1].offset;
 	int ret;
 
-	src_offset = old_node->tmp_vma.offset;
-	if (new_node->tmp_vma.node)
-		dst_offset = new_node->tmp_vma.offset;
-	else
-		dst_offset = nvbo->vma.offset;
-
 	while (length) {
 		u32 amount, stride, height;
 
@@ -689,6 +676,27 @@ nv04_bo_move_m2mf(struct nouveau_channel *chan, struct ttm_buffer_object *bo,
 	return 0;
 }
 
+static int
+nouveau_vma_getmap(struct nouveau_channel *chan, struct nouveau_bo *nvbo,
+		   struct ttm_mem_reg *mem, struct nouveau_vma *vma)
+{
+	struct nouveau_mem *node = mem->mm_node;
+	int ret;
+
+	ret = nouveau_vm_get(chan->vm, mem->num_pages << PAGE_SHIFT,
+			     node->page_shift, NV_MEM_ACCESS_RO, vma);
+	if (ret)
+		return ret;
+
+	if (mem->mem_type == TTM_PL_VRAM)
+		nouveau_vm_map(vma, node);
+	else
+		nouveau_vm_map_sg(vma, 0, mem->num_pages << PAGE_SHIFT,
+				  node, node->pages);
+
+	return 0;
+}
+
 static int
 nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int evict, bool intr,
 		     bool no_wait_reserve, bool no_wait_gpu,
@@ -706,31 +714,20 @@ nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int evict, bool intr,
 		mutex_lock_nested(&chan->mutex, NOUVEAU_KCHANNEL_MUTEX);
 	}
 
-	/* create temporary vma for old memory, this will get cleaned
-	 * up after ttm destroys the ttm_mem_reg
+	/* create temporary vmas for the transfer and attach them to the
+	 * old nouveau_mem node, these will get cleaned up after ttm has
+	 * destroyed the ttm_mem_reg
 	 */
 	if (dev_priv->card_type >= NV_50) {
 		struct nouveau_mem *node = old_mem->mm_node;
-		if (!node->tmp_vma.node) {
-			u32 page_shift = nvbo->vma.node->type;
-			if (old_mem->mem_type == TTM_PL_TT)
-				page_shift = nvbo->vma.vm->spg_shift;
-
-			ret = nouveau_vm_get(chan->vm,
-					     old_mem->num_pages << PAGE_SHIFT,
-					     page_shift, NV_MEM_ACCESS_RO,
-					     &node->tmp_vma);
-			if (ret)
-				goto out;
-		}
 
-		if (old_mem->mem_type == TTM_PL_VRAM)
-			nouveau_vm_map(&node->tmp_vma, node);
-		else {
-			nouveau_vm_map_sg(&node->tmp_vma, 0,
-					  old_mem->num_pages << PAGE_SHIFT,
-					  node, node->pages);
-		}
+		ret = nouveau_vma_getmap(chan, nvbo, old_mem, &node->vma[0]);
+		if (ret)
+			goto out;
+
+		ret = nouveau_vma_getmap(chan, nvbo, new_mem, &node->vma[1]);
+		if (ret)
+			goto out;
 	}
 
 	if (dev_priv->card_type < NV_50)
@@ -757,7 +754,6 @@ nouveau_bo_move_flipd(struct ttm_buffer_object *bo, bool evict, bool intr,
 		      bool no_wait_reserve, bool no_wait_gpu,
 		      struct ttm_mem_reg *new_mem)
 {
-	struct drm_nouveau_private *dev_priv = nouveau_bdev(bo->bdev);
 	u32 placement_memtype = TTM_PL_FLAG_TT | TTM_PL_MASK_CACHING;
 	struct ttm_placement placement;
 	struct ttm_mem_reg tmp_mem;
@@ -777,23 +773,7 @@ nouveau_bo_move_flipd(struct ttm_buffer_object *bo, bool evict, bool intr,
 	if (ret)
 		goto out;
 
-	if (dev_priv->card_type >= NV_50) {
-		struct nouveau_bo *nvbo = nouveau_bo(bo);
-		struct nouveau_mem *node = tmp_mem.mm_node;
-		struct nouveau_vma *vma = &nvbo->vma;
-		if (vma->node->type != vma->vm->spg_shift)
-			vma = &node->tmp_vma;
-		nouveau_vm_map_sg(vma, 0, tmp_mem.num_pages << PAGE_SHIFT,
-				  node, node->pages);
-	}
-
 	ret = nouveau_bo_move_m2mf(bo, true, intr, no_wait_reserve, no_wait_gpu, &tmp_mem);
-
-	if (dev_priv->card_type >= NV_50) {
-		struct nouveau_bo *nvbo = nouveau_bo(bo);
-		nouveau_vm_unmap(&nvbo->vma);
-	}
-
 	if (ret)
 		goto out;
 
@@ -846,21 +826,15 @@ nouveau_bo_move_ntfy(struct ttm_buffer_object *bo, struct ttm_mem_reg *new_mem)
 	if (!vma->vm)
 		return;
 
-	switch (new_mem->mem_type) {
-	case TTM_PL_VRAM:
-		nouveau_vm_map(vma, node);
-		break;
-	case TTM_PL_TT:
-		if (vma->node->type != vma->vm->spg_shift) {
-			nouveau_vm_unmap(vma);
-			vma = &node->tmp_vma;
-		}
-		nouveau_vm_map_sg(vma, 0, new_mem->num_pages << PAGE_SHIFT,
-				  node, node->pages);
-		break;
-	default:
+	if (new_mem->mem_type == TTM_PL_VRAM) {
+		nouveau_vm_map(&nvbo->vma, new_mem->mm_node);
+	} else
+	if (new_mem->mem_type == TTM_PL_TT &&
+	    nvbo->page_shift == nvbo->vma.vm->spg_shift) {
+		nouveau_vm_map_sg(&nvbo->vma, 0, new_mem->
+				  num_pages << PAGE_SHIFT, node, node->pages);
+	} else {
 		nouveau_vm_unmap(&nvbo->vma);
-		break;
 	}
 }
 

drivers/gpu/drm/nouveau/nouveau_drv.h

@@ -77,7 +77,7 @@ struct nouveau_mem {
 	struct drm_device *dev;
 
 	struct nouveau_vma bar_vma;
-	struct nouveau_vma tmp_vma;
+	struct nouveau_vma vma[2];
 	u8  page_shift;
 
 	struct drm_mm_node *tag;

drivers/gpu/drm/nouveau/nouveau_mem.c

@@ -762,20 +762,29 @@ nouveau_vram_manager_fini(struct ttm_mem_type_manager *man)
 	return 0;
 }
 
+static inline void
+nouveau_mem_node_cleanup(struct nouveau_mem *node)
+{
+	if (node->vma[0].node) {
+		nouveau_vm_unmap(&node->vma[0]);
+		nouveau_vm_put(&node->vma[0]);
+	}
+
+	if (node->vma[1].node) {
+		nouveau_vm_unmap(&node->vma[1]);
+		nouveau_vm_put(&node->vma[1]);
+	}
+}
+
 static void
 nouveau_vram_manager_del(struct ttm_mem_type_manager *man,
 			 struct ttm_mem_reg *mem)
 {
 	struct drm_nouveau_private *dev_priv = nouveau_bdev(man->bdev);
 	struct nouveau_vram_engine *vram = &dev_priv->engine.vram;
-	struct nouveau_mem *node = mem->mm_node;
 	struct drm_device *dev = dev_priv->dev;
 
-	if (node->tmp_vma.node) {
-		nouveau_vm_unmap(&node->tmp_vma);
-		nouveau_vm_put(&node->tmp_vma);
-	}
-
+	nouveau_mem_node_cleanup(mem->mm_node);
 	vram->put(dev, (struct nouveau_mem **)&mem->mm_node);
 }
 
@@ -860,15 +869,9 @@ static void
 nouveau_gart_manager_del(struct ttm_mem_type_manager *man,
 			 struct ttm_mem_reg *mem)
 {
-	struct nouveau_mem *node = mem->mm_node;
-
-	if (node->tmp_vma.node) {
-		nouveau_vm_unmap(&node->tmp_vma);
-		nouveau_vm_put(&node->tmp_vma);
-	}
-
+	nouveau_mem_node_cleanup(mem->mm_node);
 	mem->mm_node = NULL;
-	kfree(node);
+	kfree(mem->mm_node);
 }
 
 static int
@@ -878,11 +881,7 @@ nouveau_gart_manager_new(struct ttm_mem_type_manager *man,
 			 struct ttm_mem_reg *mem)
 {
 	struct drm_nouveau_private *dev_priv = nouveau_bdev(bo->bdev);
-	struct nouveau_bo *nvbo = nouveau_bo(bo);
-	struct nouveau_vma *vma = &nvbo->vma;
-	struct nouveau_vm *vm = vma->vm;
 	struct nouveau_mem *node;
-	int ret;
 
 	if (unlikely((mem->num_pages << PAGE_SHIFT) >=
 		     dev_priv->gart_info.aper_size))
@@ -891,24 +890,8 @@ nouveau_gart_manager_new(struct ttm_mem_type_manager *man,
 	node = kzalloc(sizeof(*node), GFP_KERNEL);
 	if (!node)
 		return -ENOMEM;
+	node->page_shift = 12;
 
-	/* This node must be for evicting large-paged VRAM
-	 * to system memory.  Due to a nv50 limitation of
-	 * not being able to mix large/small pages within
-	 * the same PDE, we need to create a temporary
-	 * small-paged VMA for the eviction.
-	 */
-	if (vma->node->type != vm->spg_shift) {
-		ret = nouveau_vm_get(vm, (u64)vma->node->length << 12,
-				     vm->spg_shift, NV_MEM_ACCESS_RW,
-				     &node->tmp_vma);
-		if (ret) {
-			kfree(node);
-			return ret;
-		}
-	}
-
-	node->page_shift = nvbo->vma.node->type;
 	mem->mm_node = node;
 	mem->start   = 0;
 	return 0;