提交 85a4b579 编写于 作者: C Christian König 提交者: Alex Deucher

drm/amdgpu: free userptrs even if GTT isn't bound

This fixes a memory leak since binding GTT only on demand.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
上级 3d5f4d47
......@@ -757,6 +757,9 @@ static int amdgpu_ttm_backend_unbind(struct ttm_tt *ttm)
{
struct amdgpu_ttm_tt *gtt = (void *)ttm;
if (gtt->userptr)
amdgpu_ttm_tt_unpin_userptr(ttm);
if (!amdgpu_ttm_is_bound(ttm))
return 0;
......@@ -764,9 +767,6 @@ static int amdgpu_ttm_backend_unbind(struct ttm_tt *ttm)
if (gtt->adev->gart.ready)
amdgpu_gart_unbind(gtt->adev, gtt->offset, ttm->num_pages);
if (gtt->userptr)
amdgpu_ttm_tt_unpin_userptr(ttm);
spin_lock(&gtt->adev->gtt_list_lock);
list_del_init(&gtt->list);
spin_unlock(&gtt->adev->gtt_list_lock);
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册