linux-kernel - Re: [PATCH v9 6/6] drm/amdgpu: Make use of drm_wedge_task

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <a0f508fd-3277-4839-a4b6-e6bc56546f6c@amd.com>
Date: Wed, 18 Jun 2025 09:29:04 +0200
From: Christian König <christian.koenig@....com>
To: André Almeida <andrealmeid@...lia.com>,
 Alex Deucher <alexander.deucher@....com>, siqueira@...lia.com,
 airlied@...il.com, simona@...ll.ch, Raag Jadav <raag.jadav@...el.com>,
 rodrigo.vivi@...el.com, jani.nikula@...ux.intel.com,
 Xaver Hugl <xaver.hugl@...il.com>,
 Krzysztof Karas <krzysztof.karas@...el.com>
Cc: dri-devel@...ts.freedesktop.org, linux-kernel@...r.kernel.org,
 kernel-dev@...lia.com, amd-gfx@...ts.freedesktop.org,
 intel-xe@...ts.freedesktop.org, intel-gfx@...ts.freedesktop.org
Subject: Re: [PATCH v9 6/6] drm/amdgpu: Make use of drm_wedge_task_info

On 6/17/25 15:22, André Almeida wrote:
> Em 17/06/2025 10:07, Christian König escreveu:
>> On 6/17/25 14:49, André Almeida wrote:
>>> To notify userspace about which task (if any) made the device get in a
>>> wedge state, make use of drm_wedge_task_info parameter, filling it with
>>> the task PID and name.
>>>
>>> Signed-off-by: André Almeida <andrealmeid@...lia.com>
>>
>> Reviewed-by: Christian König <christian.koenig@....com>
>>
>> Do you have commit right for drm-misc-next?
>>
> 
> Thanks for the reviews!
> 
> I do have access, but if you don't mind, can you push this one?

Sure, but give me till the end of today.

(And maybe ping me next week should I forget about it).

Regards,
Christian.

> 
>> Regards,
>> Christian.
>>
>>> ---
>>> v8:
>>>   - Drop check before calling amdgpu_vm_put_task_info()
>>>   - Drop local variable `info`
>>> v7:
>>>   - Remove struct cast, now we can use `info = &ti->task`
>>>   - Fix struct lifetime, move amdgpu_vm_put_task_info() after
>>>     drm_dev_wedged_event() call
>>> ---
>>>   drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 13 +++++++++++--
>>>   drivers/gpu/drm/amd/amdgpu/amdgpu_job.c    |  7 +++++--
>>>   2 files changed, 16 insertions(+), 4 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>>> index 8a0f36f33f13..a59f194e3360 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>>> @@ -6363,8 +6363,17 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
>>>         atomic_set(&adev->reset_domain->reset_res, r);
>>>   -    if (!r)
>>> -        drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE, NULL);
>>> +    if (!r) {
>>> +        struct amdgpu_task_info *ti = NULL;
>>> +
>>> +        if (job)
>>> +            ti = amdgpu_vm_get_task_info_pasid(adev, job->pasid);
>>> +
>>> +        drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE,
>>> +                     ti ? &ti->task : NULL);
>>> +
>>> +        amdgpu_vm_put_task_info(ti);
>>> +    }
>>>         return r;
>>>   }
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
>>> index 0c1381b527fe..1e24590ae144 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
>>> @@ -89,6 +89,7 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
>>>   {
>>>       struct amdgpu_ring *ring = to_amdgpu_ring(s_job->sched);
>>>       struct amdgpu_job *job = to_amdgpu_job(s_job);
>>> +    struct drm_wedge_task_info *info = NULL;
>>>       struct amdgpu_task_info *ti;
>>>       struct amdgpu_device *adev = ring->adev;
>>>       int idx;
>>> @@ -125,7 +126,7 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
>>>       ti = amdgpu_vm_get_task_info_pasid(ring->adev, job->pasid);
>>>       if (ti) {
>>>           amdgpu_vm_print_task_info(adev, ti);
>>> -        amdgpu_vm_put_task_info(ti);
>>> +        info = &ti->task;
>>>       }
>>>         /* attempt a per ring reset */
>>> @@ -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));
>>
>