[PATCH 3/4] drm/amdgpu: Optimization of mutex usage

[Christian König]

Am 12.06.2017 um 22:31 schrieb Alex Xie:
> There is no need to lock the bo_list mutex,
> because there is no other task can access
> the newly created BO list yet.

Again NAK, this can race as well when a task guess the next ID number or 
incorrectly uses an old one.

Christian.

>
> Signed-off-by: Alex Xie <AlexBin.Xie at amd.com>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c | 3 ---
>   1 file changed, 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c
> index c994a04..4363f28 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c
> @@ -60,8 +60,6 @@ static int amdgpu_bo_list_create(struct amdgpu_fpriv *fpriv,
>   	(*result)->num_entries = 0;
>   	(*result)->array = NULL;
>   
> -	mutex_lock(&(*result)->lock);
> -
>   	return 0;
>   }
>   
> @@ -282,7 +280,6 @@ int amdgpu_bo_list_ioctl(struct drm_device *dev, void *data,
>   
>   		r = amdgpu_bo_list_set(adev, filp, list, info,
>   					      args->in.bo_number);
> -		amdgpu_bo_list_put(list);
>   		if (r)
>   			goto error_free;
>