Re: [PATCH] drm/virtio: Destroy obj_restore_lock on the final drm_dev_put()

[Dmitry Osipenko <dmitry.osipenko@collabora.com>]

On 10/14/26 22:53, Abhin Parekadan Jose wrote:
> On Wed, Oct 14, 2026 at 01:16:57AM +0300, Dmitry Osipenko wrote:
>> On 10/12/26 08:44, Abhin Parekadan Jose wrote:
>>> Gentle ping on this fix for the syzbot report.
>>>
>>> Dmitry, Gerd, do you have a preference between moving mutex_destroy()
>>> to virtio_gpu_release() (this patch) and using drmm_mutex_init()? I can
>>> send a v2 with the other option if you prefer it.
>>>
>>> Thanks,
>>> Abhin
>>>
>>> On Tue, Sep 29, 2026 at 07:07:14AM +0000, Abhin Parekadan Jose wrote:
>>>> virtio_gpu_deinit() calls mutex_destroy() on obj_restore_lock, but
>>>> deinit runs from virtio_gpu_remove() at unbind or PCI remove, and GEM
>>>> objects can outlive that. An open /dev/fb0 keeps the fbdev buffer
>>>> alive, and the fbdev DRM client holds a drm_device reference. When the
>>>> fd is finally closed, the buffer is freed and takes the destroyed lock:
>>>>
>>>>   fb_release()
>>>>     drm_fbdev_shmem_fb_destroy()
>>>>       drm_client_buffer_delete()
>>>>         virtio_gpu_free_object()
>>>>           virtio_gpu_remove_from_restore_list()
>>>>             mutex_lock(&vgdev->obj_restore_lock)
>>>>
>>>>   DEBUG_LOCKS_WARN_ON(lock->magic != lock)
>>>>   WARNING: kernel/locking/mutex.c:625 at __mutex_lock+0xf2c/0x12ec
>>>>
>>>> virtio_gpu_deinit() should only do hardware teardown in this case
>>>> related virtio, its queues and others. Software state that GEM callbacks
>>>> still use belongs in virtio_gpu_release(), which runs on the final
>>>> drm_dev_put().
>>>>
>>>> There are two equivalent ways to tie the mutex lifetime to the
>>>> drm_device:
>>>>
>>>>  1) Move mutex_destroy() from virtio_gpu_deinit() to the end of
>>>>     virtio_gpu_release(). This is what this patch does. It is the
>>>>     smallest change and keeps the teardown next to the rest of the
>>>>     final-put software cleanup.
>>>>
>>>>  2) Initialise the lock with drmm_mutex_init() (checking its return
>>>>     value) and drop mutex_destroy() from virtio_gpu_deinit(), so DRM's
>>>>     managed release destroys it. This also covers the virtio_gpu_init()
>>>>     error path, where release returns early because dev_private is
>>>>     NULL. Skipping mutex_destroy() there is harmless, since it only
>>>>     matters for mutex debugging.
>>>>
>>>> Either fix removes the WARN. Option 1 is sent as the smaller change,
>>>> and option 2 can be done instead if preferred.
>>>>
>>>> Reproduced on arm64 QEMU (-device virtio-gpu-pci) with DEBUG_MUTEXES
>>>> and PROVE_LOCKING: open /dev/fb0, write 1 to
>>>> /sys/bus/pci/devices/0000:00:01.0/remove, then close the fd. With this
>>>> patch the WARN is gone, and rebinding brings back /dev/fb0 and
>>>> /dev/dri/card0.
>>>>
>>>> Fixes: 54a970048296 ("drm/virtio: Add support for saving and restoring virtio_gpu_objects")
>>>> Reported-by: syzbot+3590d97d8a586fa955c2@syzkaller.appspotmail.com
>>>> Closes: https://syzkaller.appspot.com/bug?extid=3590d97d8a586fa955c2
>>>> Assisted-by: LLM
>>>> Signed-off-by: Abhin Parekadan Jose <abhinjoses@gmail.com>
>>>> ---
>>>>  drivers/gpu/drm/virtio/virtgpu_kms.c | 3 ++-
>>>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/drivers/gpu/drm/virtio/virtgpu_kms.c b/drivers/gpu/drm/virtio/virtgpu_kms.c
>>>> index 1d4d3bf46a20..0ec755c35050 100644
>>>> --- a/drivers/gpu/drm/virtio/virtgpu_kms.c
>>>> +++ b/drivers/gpu/drm/virtio/virtgpu_kms.c
>>>> @@ -354,7 +354,6 @@ void virtio_gpu_deinit(struct drm_device *dev)
>>>>  	virtio_reset_device(vgdev->vdev);
>>>>  	virtio_gpu_reclaim_vbufs(vgdev);
>>>>  	vgdev->vdev->config->del_vqs(vgdev->vdev);
>>>> -	mutex_destroy(&vgdev->obj_restore_lock);
>>>>  }
>>>>  
>>>>  void virtio_gpu_release(struct drm_device *dev)
>>>> @@ -370,6 +369,8 @@ void virtio_gpu_release(struct drm_device *dev)
>>>>  
>>>>  	if (vgdev->has_host_visible)
>>>>  		drm_mm_takedown(&vgdev->host_visible_mm);
>>>> +
>>>> +	mutex_destroy(&vgdev->obj_restore_lock);
>>>>  }
>>>>  
>>>>  int virtio_gpu_driver_open(struct drm_device *dev, struct drm_file *file)
>> Calling virtio_gpu_deinit() from virtio_gpu_release() might be a better
>> option
> 
> Wouldn't that end up freeing the virtqueues twice, or using them after
> they are freed? virtio_gpu_deinit() would then run from both the
> virtio_gpu_remove() path and the virtio_gpu_release() path.
> or do you mean removing it from virtio_gpu_remove() and calling it only
> from virtio_gpu_release()?
> This is only from a high-level look so far. I'll look into it more
> closely and get back to you.

Remove from virtio_gpu_remove() and call only from virtio_gpu_release().

-- 
Best regards,
Dmitry