[Intel-gfx] [PATCH 2/2] drm/i915: Initialise userptr mmu_notifier serial to 1

[Tvrtko Ursulin]

On 07/11/2014 12:35 PM, Chris Wilson wrote:
> On Fri, Jul 11, 2014 at 12:31:12PM +0100, Tvrtko Ursulin wrote:
>>
>> On 07/11/2014 12:06 PM, Chris Wilson wrote:
>>> On Fri, Jul 11, 2014 at 12:00:26PM +0100, Tvrtko Ursulin wrote:
>>>> But it will be interesting to know what code managed to trigger this
>>>> race, because as we discussed on IRC it would indicate some pretty
>>>> wild userspace behaviour. Or lack of imagination on our part?
>>>
>>> A threaded client. One thread using userptr, the other doing munmap or
>>> free. Given enough embarrassment, it will happen every time.
>>
>> Yes fine, but I struggle to imagine what would be the intention of
>> such code or how did it manage to fail in such way. I hope the only
>> difference is not that userptr "upgraded" the failure mode for heap
>> corruption or memory management races in general.
>
> The mmu notifier is called everytime a process sneezes. It does not
> imply that our object is being invalidated, just that some portion of
> the current->mm is being modified.

Ah yes, I did not see the big picture.

Tvrtko