✗ i915.CI.Full: failure for drm/i915: Precompute plane SURF address/etc.

[Ville Syrjälä]

On Wed, Apr 02, 2025 at 04:33:20AM -0000, Patchwork wrote:
> == Series Details ==
> 
> Series: drm/i915: Precompute plane SURF address/etc.
> URL   : https://patchwork.freedesktop.org/series/147097/
> State : failure
> 
> == Summary ==
> 
> CI Bug Log - changes from CI_DRM_16361_full -> Patchwork_147097v1_full
> ====================================================
> 
> Summary
> -------
> 
>   **FAILURE**
> 
>   Serious unknown changes coming with Patchwork_147097v1_full absolutely need to be
>   verified manually.
>   
>   If you think the reported changes have nothing to do with the changes
>   introduced in Patchwork_147097v1_full, please notify your bug team (I915-ci-infra at lists.freedesktop.org) to allow them
>   to document this new failure mode, which will reduce false positives in CI.
> 
>   
> 
> Participating hosts (10 -> 10)
> ------------------------------
> 
>   No changes in participating hosts
> 
> Possible new issues
> -------------------
> 
>   Here are the unknown changes that may have been introduced in Patchwork_147097v1_full:
> 
> ### IGT changes ###
> 
> #### Possible regressions ####
> 
>   * igt at kms_color@ctm-max at pipe-a-hdmi-a-1:
>     - shard-snb:          [PASS][1] -> [ABORT][2] +17 other tests abort
>    [1]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_16361/shard-snb2/igt@kms_color@ctm-max@pipe-a-hdmi-a-1.html
>    [2]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_147097v1/shard-snb5/igt@kms_color@ctm-max@pipe-a-hdmi-a-1.html

Apparently I forgot that we can't do the *_ctl_crtc() stuff
so early as crtc_state->gamma_enable & co. haven't been computed
yet.

I think I still want to clean up the ->ctl stuff, but I probably
need to add a new plane vfunc for this. And I think I might take
that opportunity to move some of plane related details from
intel_color_add_affected_planes() into the plane code proper.

For now I'll probably repost the first six patches as a
separate series since all that should be completely fine as is.

What's also slightly troubling about these CI results is that
GLK didn't catch anything, despite the potential to use a stale
PLANE_COLOR_CTL. And presumable SNB only caught stuff becasue the
state checker noticed the stale DSPCNTR, but the tests didn't.
So seems like we are somehow missing tests that would change
gamma_mode & co. without touching the plane...

> 
>   * igt at kms_color@ctm-max at pipe-b-hdmi-a-1:
>     - shard-snb:          [PASS][3] -> [DMESG-WARN][4] +3 other tests dmesg-warn
>    [3]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_16361/shard-snb2/igt@kms_color@ctm-max@pipe-b-hdmi-a-1.html
>    [4]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_147097v1/shard-snb5/igt@kms_color@ctm-max@pipe-b-hdmi-a-1.html

CI folks, why is this reported as 'dmesg-warn' vs. the previous test was
'abort' even though both tests failed in the same exact way (state checker
WARN)?

-- 
Ville Syrjälä
Intel