drm/ttm: simplify memory accounting for ttm user v2
Provide helper function to compute the kernel memory size needed
for each buffer object. Move all the accounting inside ttm, simplifying
driver and avoiding code duplication accross them.
v2 fix accounting of ghost object, one would have thought that i
would have run into the issue since a longtime but it seems
ghost object are rare when you have plenty of vram ;)
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index d6326af..7ac7bc3 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -93,6 +93,7 @@
{
struct drm_nouveau_private *dev_priv = dev->dev_private;
struct nouveau_bo *nvbo;
+ size_t acc_size;
int ret;
nvbo = kzalloc(sizeof(struct nouveau_bo), GFP_KERNEL);
@@ -115,9 +116,12 @@
nvbo->bo.mem.num_pages = size >> PAGE_SHIFT;
nouveau_bo_placement_set(nvbo, flags, 0);
+ acc_size = ttm_bo_dma_acc_size(&dev_priv->ttm.bdev, size,
+ sizeof(struct nouveau_bo));
+
ret = ttm_bo_init(&dev_priv->ttm.bdev, &nvbo->bo, size,
ttm_bo_type_device, &nvbo->placement,
- align >> PAGE_SHIFT, 0, false, NULL, size,
+ align >> PAGE_SHIFT, 0, false, NULL, acc_size,
nouveau_bo_del_ttm);
if (ret) {
/* ttm will call nouveau_bo_del_ttm if it fails.. */
diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c
index 1c85152..695b480 100644
--- a/drivers/gpu/drm/radeon/radeon_object.c
+++ b/drivers/gpu/drm/radeon/radeon_object.c
@@ -95,6 +95,7 @@
enum ttm_bo_type type;
unsigned long page_align = roundup(byte_align, PAGE_SIZE) >> PAGE_SHIFT;
unsigned long max_size = 0;
+ size_t acc_size;
int r;
size = ALIGN(size, PAGE_SIZE);
@@ -117,6 +118,9 @@
return -ENOMEM;
}
+ acc_size = ttm_bo_dma_acc_size(&rdev->mman.bdev, size,
+ sizeof(struct radeon_bo));
+
retry:
bo = kzalloc(sizeof(struct radeon_bo), GFP_KERNEL);
if (bo == NULL)
@@ -134,8 +138,8 @@
/* Kernel allocation are uninterruptible */
mutex_lock(&rdev->vram_mutex);
r = ttm_bo_init(&rdev->mman.bdev, &bo->tbo, size, type,
- &bo->placement, page_align, 0, !kernel, NULL, size,
- &radeon_ttm_bo_destroy);
+ &bo->placement, page_align, 0, !kernel, NULL,
+ acc_size, &radeon_ttm_bo_destroy);
mutex_unlock(&rdev->vram_mutex);
if (unlikely(r != 0)) {
if (r != -ERESTARTSYS) {
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index cb73527..de7ad99 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -137,6 +137,7 @@
struct ttm_buffer_object *bo =
container_of(list_kref, struct ttm_buffer_object, list_kref);
struct ttm_bo_device *bdev = bo->bdev;
+ size_t acc_size = bo->acc_size;
BUG_ON(atomic_read(&bo->list_kref.refcount));
BUG_ON(atomic_read(&bo->kref.refcount));
@@ -152,9 +153,9 @@
if (bo->destroy)
bo->destroy(bo);
else {
- ttm_mem_global_free(bdev->glob->mem_glob, bo->acc_size);
kfree(bo);
}
+ ttm_mem_global_free(bdev->glob->mem_glob, acc_size);
}
int ttm_bo_wait_unreserved(struct ttm_buffer_object *bo, bool interruptible)
@@ -1157,6 +1158,17 @@
{
int ret = 0;
unsigned long num_pages;
+ struct ttm_mem_global *mem_glob = bdev->glob->mem_glob;
+
+ ret = ttm_mem_global_alloc(mem_glob, acc_size, false, false);
+ if (ret) {
+ printk(KERN_ERR TTM_PFX "Out of kernel memory.\n");
+ if (destroy)
+ (*destroy)(bo);
+ else
+ kfree(bo);
+ return -ENOMEM;
+ }
size += buffer_start & ~PAGE_MASK;
num_pages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT;
@@ -1227,14 +1239,34 @@
}
EXPORT_SYMBOL(ttm_bo_init);
-static inline size_t ttm_bo_size(struct ttm_bo_global *glob,
- unsigned long num_pages)
+size_t ttm_bo_acc_size(struct ttm_bo_device *bdev,
+ unsigned long bo_size,
+ unsigned struct_size)
{
- size_t page_array_size = (num_pages * sizeof(void *) + PAGE_SIZE - 1) &
- PAGE_MASK;
+ unsigned npages = (PAGE_ALIGN(bo_size)) >> PAGE_SHIFT;
+ size_t size = 0;
- return glob->ttm_bo_size + 2 * page_array_size;
+ size += ttm_round_pot(struct_size);
+ size += PAGE_ALIGN(npages * sizeof(void *));
+ size += ttm_round_pot(sizeof(struct ttm_tt));
+ return size;
}
+EXPORT_SYMBOL(ttm_bo_acc_size);
+
+size_t ttm_bo_dma_acc_size(struct ttm_bo_device *bdev,
+ unsigned long bo_size,
+ unsigned struct_size)
+{
+ unsigned npages = (PAGE_ALIGN(bo_size)) >> PAGE_SHIFT;
+ size_t size = 0;
+
+ size += ttm_round_pot(struct_size);
+ size += PAGE_ALIGN(npages * sizeof(void *));
+ size += PAGE_ALIGN(npages * sizeof(dma_addr_t));
+ size += ttm_round_pot(sizeof(struct ttm_dma_tt));
+ return size;
+}
+EXPORT_SYMBOL(ttm_bo_dma_acc_size);
int ttm_bo_create(struct ttm_bo_device *bdev,
unsigned long size,
@@ -1248,10 +1280,10 @@
{
struct ttm_buffer_object *bo;
struct ttm_mem_global *mem_glob = bdev->glob->mem_glob;
+ size_t acc_size;
int ret;
- size_t acc_size =
- ttm_bo_size(bdev->glob, (size + PAGE_SIZE - 1) >> PAGE_SHIFT);
+ acc_size = ttm_bo_acc_size(bdev, size, sizeof(struct ttm_buffer_object));
ret = ttm_mem_global_alloc(mem_glob, acc_size, false, false);
if (unlikely(ret != 0))
return ret;
@@ -1437,10 +1469,6 @@
goto out_no_shrink;
}
- glob->ttm_bo_extra_size = ttm_round_pot(sizeof(struct ttm_tt));
- glob->ttm_bo_size = glob->ttm_bo_extra_size +
- ttm_round_pot(sizeof(struct ttm_buffer_object));
-
atomic_set(&glob->bo_count, 0);
ret = kobject_init_and_add(
diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c
index 60f204d..f8187ea 100644
--- a/drivers/gpu/drm/ttm/ttm_bo_util.c
+++ b/drivers/gpu/drm/ttm/ttm_bo_util.c
@@ -445,6 +445,7 @@
kref_init(&fbo->list_kref);
kref_init(&fbo->kref);
fbo->destroy = &ttm_transfered_destroy;
+ fbo->acc_size = 0;
*new_obj = fbo;
return 0;
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
index 86c5e4c..2eb84a5 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
@@ -1517,29 +1517,10 @@
/**
* Buffer management.
*/
-
-static size_t vmw_dmabuf_acc_size(struct ttm_bo_global *glob,
- unsigned long num_pages)
-{
- static size_t bo_user_size = ~0;
-
- size_t page_array_size =
- (num_pages * sizeof(void *) + PAGE_SIZE - 1) & PAGE_MASK;
-
- if (unlikely(bo_user_size == ~0)) {
- bo_user_size = glob->ttm_bo_extra_size +
- ttm_round_pot(sizeof(struct vmw_dma_buffer));
- }
-
- return bo_user_size + page_array_size;
-}
-
void vmw_dmabuf_bo_free(struct ttm_buffer_object *bo)
{
struct vmw_dma_buffer *vmw_bo = vmw_dma_buffer(bo);
- struct ttm_bo_global *glob = bo->glob;
- ttm_mem_global_free(glob->mem_glob, bo->acc_size);
kfree(vmw_bo);
}
@@ -1550,24 +1531,12 @@
void (*bo_free) (struct ttm_buffer_object *bo))
{
struct ttm_bo_device *bdev = &dev_priv->bdev;
- struct ttm_mem_global *mem_glob = bdev->glob->mem_glob;
size_t acc_size;
int ret;
BUG_ON(!bo_free);
- acc_size =
- vmw_dmabuf_acc_size(bdev->glob,
- (size + PAGE_SIZE - 1) >> PAGE_SHIFT);
-
- ret = ttm_mem_global_alloc(mem_glob, acc_size, false, false);
- if (unlikely(ret != 0)) {
- /* we must free the bo here as
- * ttm_buffer_object_init does so as well */
- bo_free(&vmw_bo->base);
- return ret;
- }
-
+ acc_size = ttm_bo_acc_size(bdev, size, sizeof(struct vmw_dma_buffer));
memset(vmw_bo, 0, sizeof(*vmw_bo));
INIT_LIST_HEAD(&vmw_bo->validate_list);
@@ -1582,9 +1551,7 @@
static void vmw_user_dmabuf_destroy(struct ttm_buffer_object *bo)
{
struct vmw_user_dma_buffer *vmw_user_bo = vmw_user_dma_buffer(bo);
- struct ttm_bo_global *glob = bo->glob;
- ttm_mem_global_free(glob->mem_glob, bo->acc_size);
kfree(vmw_user_bo);
}