[Intel-gfx] ✗ Fi.CI.BAT: failure for drm/i915: Skip PIPESTAT reads from irq handler on VLV/CHV when power well is down

[Ville Syrjälä]

On Fri, Feb 19, 2016 at 08:20:27AM -0000, Patchwork wrote:
> == Summary ==
> 
> Series 3599v1 drm/i915: Skip PIPESTAT reads from irq handler on VLV/CHV when power well is down
> http://patchwork.freedesktop.org/api/1.0/series/3599/revisions/1/mbox/
> 
> Test core_auth:
>         Subgroup basic-auth:
>                 pass       -> INCOMPLETE (ilk-hp8440p)
> Test core_prop_blob:
>         Subgroup basic:
>                 pass       -> INCOMPLETE (ilk-hp8440p)

The ilk seems very unahppy lately. Not sure if this is just some
igt/piglit/etc. change that broke things or if we really managed to break
the driver.

It looks like this mess started from CI_DRM_1077, so I tried to dig out
the kernel shas for CI_DRM_1076 and CI_DRM_1077. Assuming I looked in the
right place the diff consists of two parts:
dd75619853e4 ("drm/i915: Extract intel_encoder_has_connectors()")
67ec1072b053 ("ALSA: pcm: Fix rwsem deadlock for non-atomic PCM

I can't convince myself that there's anything wrong with my
intel_encoder_has_connectors(), so it's either the sound locking
stuff (which also feels a bit unlikely to me), or just totally
unrelated to the kernel.

-- 
Ville Syrjälä
Intel OTC