[Intel-gfx] ✗ Fi.CI.BAT: failure for drm/i915: CCS prep stuff

[Chris Wilson]

On Wed, Mar 08, 2017 at 04:08:05PM +0200, Ville Syrjälä wrote:
> On Wed, Mar 08, 2017 at 01:52:09PM -0000, Patchwork wrote:
> > == Series Details ==
> > 
> > Series: drm/i915: CCS prep stuff
> > URL   : https://patchwork.freedesktop.org/series/20851/
> > State : failure
> > 
> > == Summary ==
> > 
> > Series 20851v1 drm/i915: CCS prep stuff
> > https://patchwork.freedesktop.org/api/1.0/series/20851/revisions/1/mbox/
> > 
> > Test gem_exec_flush:
> >         Subgroup basic-uc-set-default:
> >                 pass       -> INCOMPLETE (fi-skl-6770hq)
> 
> Going to chuck this into the same bucket as
> https://bugs.freedesktop.org/show_bug.cgi?id=100112
> 
> I also see a few in the skl-600hq long term results. The common
> pattern seems to be skl+gem_exec_flush (the specific subtest
> varies between runs).

Afaict, it is not a i915.ko bug. It seems to be that the exit_handler in
the test running is not killing off its children, and that causes piglit
to wait forever. We have a workaround for that that should be hitting CI
the next time it refreshes igt.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre