[PATCH] drm/amdgpu: Don't move pinned BOs

Christian König deathsimple at vodafone.de
Mon Mar 28 14:33:17 UTC 2016


Am 28.03.2016 um 05:53 schrieb Michel Dänzer:
> From: Michel Dänzer <michel.daenzer at amd.com>
>
> The purpose of pinning is to prevent a buffer from moving.
>
> Reviewed-by: Christian König <christian.koenig at amd.com>
> Tested-by: Rex Zhu <Rex.Zhu at amd.com>
> Signed-off-by: Michel Dänzer <michel.daenzer at amd.com>

Well, that should not happen (at least when all the code for disallowing 
pinning into invisible VRAM works correctly).

But I would say better save than sorry, so this patch and the radeon 
version is Reviewed-by: Christian König <christian.koenig at amd.com>

Regards,
Christian.

> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 4 ++++
>   drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c    | 6 ++++++
>   2 files changed, 10 insertions(+)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> index 151a2d4..56d1458 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> @@ -608,6 +608,10 @@ int amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo)
>   	if ((offset + size) <= adev->mc.visible_vram_size)
>   		return 0;
>   
> +	/* Can't move a pinned BO to visible VRAM */
> +	if (abo->pin_count > 0)
> +		return -EINVAL;
> +
>   	/* hurrah the memory is not visible ! */
>   	amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_VRAM);
>   	lpfn =	adev->mc.visible_vram_size >> PAGE_SHIFT;
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> index 0f42b1a..9d3341d 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> @@ -384,9 +384,15 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo,
>   			struct ttm_mem_reg *new_mem)
>   {
>   	struct amdgpu_device *adev;
> +	struct amdgpu_bo *abo;
>   	struct ttm_mem_reg *old_mem = &bo->mem;
>   	int r;
>   
> +	/* Can't move a pinned BO */
> +	abo = container_of(bo, struct amdgpu_bo, tbo);
> +	if (WARN_ON_ONCE(abo->pin_count > 0))
> +		return -EINVAL;
> +
>   	adev = amdgpu_get_adev(bo->bdev);
>   	if (old_mem->mem_type == TTM_PL_SYSTEM && bo->ttm == NULL) {
>   		amdgpu_move_null(bo, new_mem);



More information about the dri-devel mailing list