[PATCH] drm/syncobj: return meaningful value to user space

[Lionel Landwerlin]

On 18/07/2019 14:13, Chunming Zhou wrote:
> if WAIT_FOR_SUBMIT isn't set and in the meanwhile no underlying fence on syncobj,
> then return non-block error code to user sapce.
>
> Signed-off-by: Chunming Zhou <david1.zhou at amd.com>
> ---
>   drivers/gpu/drm/drm_syncobj.c | 4 ++--
>   1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/drm_syncobj.c b/drivers/gpu/drm/drm_syncobj.c
> index 361a01a08c18..929f7c64f9a2 100644
> --- a/drivers/gpu/drm/drm_syncobj.c
> +++ b/drivers/gpu/drm/drm_syncobj.c
> @@ -252,7 +252,7 @@ int drm_syncobj_find_fence(struct drm_file *file_private,
>   			return 0;
>   		dma_fence_put(*fence);
>   	} else {
> -		ret = -EINVAL;
> +		ret = -ENOTBLK;
>   	}
>   
>   	if (!(flags & DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT))
> @@ -832,7 +832,7 @@ static signed long drm_syncobj_array_wait_timeout(struct drm_syncobj **syncobjs,
>   			if (flags & DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT) {
>   				continue;
>   			} else {
> -				timeout = -EINVAL;
> +				timeout = -ENOTBLK;
>   				goto cleanup_entries;
>   			}
>   		}


This would break existing tests for binary syncobjs.

Is this really what we want?


-Lionel