When amdgpu_job_alloc_with_ib() fails in amdgpu_gfx_run_cleaner_shader_job(), the function returns directly without destroying the scheduler entity, causing a resource leak.
Fix this by moving the entity cleanup to a common error path. Set r = 0 on success and use a single cleanup point at the err label to ensure the entity is always destroyed regardless of whether the function succeeds or fails. Also remove the unnecessary error check for dma_fence_wait() since it never fails with intr=false and infinite timeout. Cc: [email protected] Fixes: 559a285816af ("drm/amdgpu: Replace 'amdgpu_job_submit_direct' with 'drm_sched_entity' in cleaner shader") Signed-off-by: Wentao Liang <[email protected]> --- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c index 523b681d0da9..29a07af6f5f4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c @@ -1689,9 +1689,7 @@ static int amdgpu_gfx_run_cleaner_shader_job(struct amdgpu_ring *ring) dma_fence_put(f); - /* Clean up the scheduler entity */ - drm_sched_entity_destroy(&entity); - return 0; + r = 0; err: /* Clean up the scheduler entity */ -- 2.39.5 (Apple Git-154)
