On 7/7/23 05:53, Dmitry Osipenko wrote:
> On 7/7/23 05:49, Dmitry Osipenko wrote:
>> On 6/28/23 18:58, Gurchetan Singh wrote:
>>> @@ -168,9 +168,13 @@ static int virtio_gpu_init_submit(struct 
>>> virtio_gpu_submit *submit,
>>>  
>>>     memset(submit, 0, sizeof(*submit));
>>>  
>>> -   out_fence = virtio_gpu_fence_alloc(vgdev, fence_ctx, ring_idx);
>>> -   if (!out_fence)
>>> -           return -ENOMEM;
>>> +   if ((exbuf->flags & VIRTGPU_EXECBUF_FENCE_FD_OUT) ||
>>> +       ((exbuf->flags & VIRTGPU_EXECBUF_RING_IDX) &&
>>> +       (vfpriv->ring_idx_mask & BIT_ULL(ring_idx))) ||
>>
>> Looks like there is a problem here. The virtio_gpu_fence_event_create()
>> doesn't check whether VIRTGPU_EXECBUF_RING_IDX flag is set, so it's
>> possible to trigger NULL-deref in that function if userspace will set
>> ring_idx_mask=1. Perhaps virtio_gpu_fence_event_create() need to be
>> changed to check the flag presence.
> 
> Or check whether fence is NULL

Actually, maybe this code shouldn't check VIRTGPU_EXECBUF_RING_IDX flag
at all. This flag tells which ring to use fo submission, but not which
ring to poll. Please check and correct it in v3.

-- 
Best regards,
Dmitry

Reply via email to