[Intel-gfx] ✗ Fi.CI.IGT: failure for drm/i915: Add relocation exceptions for two other platforms

[Ville Syrjälä]

On Thu, Apr 29, 2021 at 10:16:31AM +0200, Zbigniew Kempczyński wrote:
> On Wed, Apr 28, 2021 at 07:43:42PM +0000, Patchwork wrote:
> >    Patch Details
> > 
> >    Series:  drm/i915: Add relocation exceptions for two other platforms         
> >    URL:     https://patchwork.freedesktop.org/series/89594/                     
> >    State:   failure                                                             
> >    Details: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20017/index.html 
> > 
> >       CI Bug Log - changes from CI_DRM_10023_full -> Patchwork_20017_full
> > 
> > Summary
> > 
> >    FAILURE
> > 
> >    Serious unknown changes coming with Patchwork_20017_full absolutely need
> >    to be
> >    verified manually.
> > 
> >    If you think the reported changes have nothing to do with the changes
> >    introduced in Patchwork_20017_full, please notify your bug team to allow
> >    them
> >    to document this new failure mode, which will reduce false positives in
> >    CI.
> > 
> > Possible new issues
> > 
> >    Here are the unknown changes that may have been introduced in
> >    Patchwork_20017_full:
> > 
> >   IGT changes
> > 
> >     Possible regressions
> > 
> >      * igt at kms_flip_tiling@flip-changes-tiling-yf at edp-1-pipe-b:
> >           * shard-skl: PASS -> FAIL
> 
> I don't think enabling relocations on rkl/adl has the influence to skl
> but if anyone knows I'm wrong please share the details. Otherwise
> I would rerun the series to ensure this is incidental failure.

I suspect this is due to a bug in igt kms legacy commits. At least
I see a failing alpha blend test on pipe B in the log before this one,
which is how the bug hits us usually.

I'm trying to fix it atm, but unfortunately the fix is revealing other
harder to solve bugs which need to be fixed first :/

-- 
Ville Syrjälä
Intel