Quoting Chris Wilson (2019-05-03 14:52:29)
> When we want to wait for a request to be executed, we first ask if it is
> not on the GPU as if it's on the gpu, there's no need to wait. However,
> we have to take into account that a request may not be on the GPU
> because it has already completed!
>
> The window is small due to the numerous preceding checks that our target
> has not yet completed, yet there is still a very small window across the
> kmalloc.
Ok, there's a second part to this problem as this only happens under
preempt-to-busy as it requires the request running in the background and
completing after unsubmission.
> Fixes: e88619646971 ("drm/i915: Use HW semaphores for inter-engine
> synchronisation on gen8+")
So not required.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
