Re: possible amdgpu_task_info reference leak

[André Almeida]

Hi Dave,

Em 03/07/2025 20:59, Dave Airlie escreveu:

In this commit :
a72002cb181f350734108228b24c5d10d358f95a
Author: André Almeida <andrealm...@igalia.com>
Date:   Tue Jun 17 09:49:49 2025 -0300

     drm/amdgpu: Make use of drm_wedge_task_info



@@ -164,13 +165,15 @@ static enum drm_gpu_sched_stat
amdgpu_job_timedout(struct drm_sched_job *s_job)
                         if (amdgpu_ring_sched_ready(ring))
                                 drm_sched_start(&ring->sched, 0);
                         dev_err(adev->dev, "Ring %s reset
succeeded\n", ring->sched.name);
-                       drm_dev_wedged_event(adev_to_drm(adev),
DRM_WEDGE_RECOVERY_NONE, NULL);
+                       drm_dev_wedged_event(adev_to_drm(adev),
DRM_WEDGE_RECOVERY_NONE, info);
                         goto exit;
                 }
                 dev_err(adev->dev, "Ring %s reset failure\n", 
ring->sched.name);
         }
         dma_fence_set_error(&s_job->s_fence->finished, -ETIME);

+       amdgpu_vm_put_task_info(ti);
+
         if (amdgpu_device_should_recover_gpu(ring->adev)) {
                 struct amdgpu_reset_context reset_context;
                 memset(&reset_context, 0, sizeof(reset_context));


Doesn't the goto_exit bypass the amdgpu_vm_put_task_info? and cause a
reference leak at least?

Yep, I think you are right. Let me send a fix for this.

Dave.