[Intel-gfx] ✗ Fi.CI.BAT: failure for drm: drm_plane_helper_check_state() related stuff

[Ville Syrjälä]

On Wed, Nov 01, 2017 at 07:46:18PM -0000, Patchwork wrote:
> == Series Details ==
> 
> Series: drm: drm_plane_helper_check_state() related stuff
> URL   : https://patchwork.freedesktop.org/series/33002/
> State : failure
> 
> == Summary ==
> 
> Series 33002v1 drm: drm_plane_helper_check_state() related stuff
> https://patchwork.freedesktop.org/api/1.0/series/33002/revisions/1/mbox/
> 
> Test chamelium:
>         Subgroup dp-edid-read:
>                 pass       -> DMESG-WARN (fi-kbl-7500u) fdo#102672
>         Subgroup dp-crc-fast:
>                 pass       -> DMESG-WARN (fi-kbl-7500u) fdo#102514
>         Subgroup hdmi-edid-read:
>                 pass       -> DMESG-WARN (fi-skl-6700k)
>         Subgroup hdmi-crc-fast:
>                 pass       -> DMESG-WARN (fi-skl-6700k)
>         Subgroup common-hpd-after-suspend:
>                 pass       -> DMESG-WARN (fi-skl-6700k)

Thinko in my WARNs.

I guess it needs to be something like:
-       WARN_ON(!crtc_state != !plane_state->crtc);
-       WARN_ON(crtc_state && crtc_state->crtc != plane_state->crtc);
+       WARN_ON(plane_state->crtc && plane_state->crtc != crtc_state->crtc);

since we grab the crtc from the old plane state if the new plane state
doesn't have one. And thus the crtc state we pass around may refer to
the old crtc, not the new one (well, there won't be a new one in
those cases).

Although I'm not sure how sensible it is to use the crtc state of the
old crtc for state computation/checks in general. Feels a bit iffy
at least.

-- 
Ville Syrjälä
Intel OTC