[Intel-gfx] [PATCH v4] drm/i915/guc: Clean up locks in GuC

[Chris Wilson]

On Tue, Jan 26, 2016 at 05:54:21PM +0100, Daniel Vetter wrote:
> On Tue, Jan 26, 2016 at 12:08:37PM +0000, Chris Wilson wrote:
> > On Tue, Jan 26, 2016 at 11:53:45AM +0000, Tvrtko Ursulin wrote:
> > > Chris Harris has liaised with Tomi and apparently GuC firmware is
> > > already on the lab SKLs from ~two weeks ago.
> > > 
> > > Since the firmware is on all machines, presumably we would also want
> > > to have CI runs without the GuC enabled to ensure patches are not
> > > breaking the execlists mode?
> > > 
> > > How would you go about that? Submit patch series twice? Once with
> > > the final "enable GuC" patch and one without?
> > 
> > After the enable GuC patch, execlists will still be triggered on
> > machines without the right firmware. We either remove the firmware from
> > the CI image for that run, or reboot with i915.enable_guc=0
> > 
> > Yes, we should be checking that both modes "still" function.
> 
> Since guc mode will be for skl only bdw will still do execlist tests for
> us. I think that should be good enough. But I've heard also that the
> private patch submission is coming along, so you could submit a 2nd time
> without the "enable guc" patch.

Since skl doesn't pass igt, I think it is presumptuous that will be
fixed first if the results are totally ignored.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre