[PATCH v5] drm/i915/gvt: save RING_HEAD into vreg when vgpu switched out

[Zhenyu Wang]

On 2019.06.03 02:18:38 +0000, Zhang, Xiaolin wrote:
> >>>
> >>>
> >> Just update the vreg during "update_guest_context()" is not enough, thinking there is one workload was preempted out by host many times.
> >>
> > For preemption case which need extra track of host vs. guest ring
> > to reflect guest state, this one is always needed for normal finish.
> > Yeah, need further change to update guest state in time.
> per talk with Weinan, it looks like if we put these code change to
> switch_mmio, it can handle preempt case very well.

If you wanna always put ring state tracking there, which would always
try to calculate guest ring update according with host ring state, then
that'll also work. But that will need more careful tracking. I'm ok to
first go with this one for current regression, then with more refined one
to calculate interval updates for preemption.

-- 
Open Source Technology Center, Intel ltd.

$gpg --keyserver wwwkeys.pgp.net --recv-keys 4D781827
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 195 bytes
Desc: not available
URL: <https://lists.freedesktop.org/archives/intel-gvt-dev/attachments/20190603/37d9c121/attachment.sig>