[PATCH v3] drm/amdgpu: fix scheduler entity leak in cleaner shader job

[PATCH v3] drm/amdgpu: fix scheduler entity leak in cleaner shader job

[Wentao Liang]

In amdgpu_gfx_run_cleaner_shader_job(), if amdgpu_job_alloc_with_ib()
fails, the function returns without destroying the scheduler entity,
causing a resource leak.

Fix this by adding drm_sched_entity_destroy() to the error path.

Also remove the unnecessary error check for dma_fence_wait() since
it never fails with intr=false and infinite timeout.

Cc: stable@vger.kernel.org
Fixes: 559a285816af ("drm/amdgpu: Replace 'amdgpu_job_submit_direct' with 'drm_sched_entity' in cleaner shader")
Signed-off-by: Wentao Liang <vulab@iscas.ac.cn>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
index b8ca876694ff..523b681d0da9 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
@@ -1658,7 +1658,7 @@ static int amdgpu_gfx_run_cleaner_shader_job(struct amdgpu_ring *ring)
 				  &sched, 1, NULL);
 	if (r) {
 		dev_err(adev->dev, "Failed setting up GFX kernel entity.\n");
-		goto err;
+		return r;
 	}
 
 	/*
@@ -1685,9 +1685,7 @@ static int amdgpu_gfx_run_cleaner_shader_job(struct amdgpu_ring *ring)
 
 	f = amdgpu_job_submit(job);
 
-	r = dma_fence_wait(f, false);
-	if (r)
-		goto err;
+	dma_fence_wait(f, false);
 
 	dma_fence_put(f);
 
@@ -1696,6 +1694,8 @@ static int amdgpu_gfx_run_cleaner_shader_job(struct amdgpu_ring *ring)
 	return 0;
 
 err:
+	/* Clean up the scheduler entity */
+	drm_sched_entity_destroy(&entity);
 	return r;
 }
 
-- 
2.39.5 (Apple Git-154)

Re: [PATCH v3] drm/amdgpu: fix scheduler entity leak in cleaner shader job

[Christian König]

On 6/22/26 17:35, Wentao Liang wrote:
> In amdgpu_gfx_run_cleaner_shader_job(), if amdgpu_job_alloc_with_ib()
> fails, the function returns without destroying the scheduler entity,
> causing a resource leak.
> 
> Fix this by adding drm_sched_entity_destroy() to the error path.
> 
> Also remove the unnecessary error check for dma_fence_wait() since
> it never fails with intr=false and infinite timeout.
> 
> Cc: stable@vger.kernel.org
> Fixes: 559a285816af ("drm/amdgpu: Replace 'amdgpu_job_submit_direct' with 'drm_sched_entity' in cleaner shader")
> Signed-off-by: Wentao Liang <vulab@iscas.ac.cn>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 8 ++++----
>  1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
> index b8ca876694ff..523b681d0da9 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
> @@ -1658,7 +1658,7 @@ static int amdgpu_gfx_run_cleaner_shader_job(struct amdgpu_ring *ring)
>  				  &sched, 1, NULL);
>  	if (r) {
>  		dev_err(adev->dev, "Failed setting up GFX kernel entity.\n");
> -		goto err;
> +		return r;
>  	}
>  
>  	/*
> @@ -1685,9 +1685,7 @@ static int amdgpu_gfx_run_cleaner_shader_job(struct amdgpu_ring *ring)
>  
>  	f = amdgpu_job_submit(job);
>  
> -	r = dma_fence_wait(f, false);
> -	if (r)
> -		goto err;
> +	dma_fence_wait(f, false);
>  
>  	dma_fence_put(f);
>  
> @@ -1696,6 +1694,8 @@ static int amdgpu_gfx_run_cleaner_shader_job(struct amdgpu_ring *ring)
>  	return 0;
>  
>  err:
> +	/* Clean up the scheduler entity */
> +	drm_sched_entity_destroy(&entity);

That is actually redundant. The err label should just be a few more lines up and the "return 0;" changed to "return r;".

Regards,
Christian.

>  	return r;
>  }
>