[Intel-gfx] ✗ Fi.CI.IGT: failure for drm/i915: SAGV fixes (rev2)

[Ville Syrjälä]

On Thu, Feb 17, 2022 at 08:03:41PM -0000, Patchwork wrote:
> == Series Details ==
> 
> Series: drm/i915: SAGV fixes (rev2)
> URL   : https://patchwork.freedesktop.org/series/100091/
> State : failure
> 
> == Summary ==
> 
> CI Bug Log - changes from CI_DRM_11239_full -> Patchwork_22302_full
> ====================================================
> 
> Summary
> -------
> 
>   **FAILURE**
> 
>   Serious unknown changes coming with Patchwork_22302_full absolutely need to be
>   verified manually.
>   
>   If you think the reported changes have nothing to do with the changes
>   introduced in Patchwork_22302_full, please notify your bug team to allow them
>   to document this new failure mode, which will reduce false positives in CI.
> 
>   
> 
> Participating hosts (11 -> 11)
> ------------------------------
> 
>   No changes in participating hosts
> 
> Possible new issues
> -------------------
> 
>   Here are the unknown changes that may have been introduced in Patchwork_22302_full:
> 
> ### IGT changes ###
> 
> #### Possible regressions ####
> 
>   * igt at kms_cursor_legacy@all-pipes-forked-bo:
>     - shard-iclb:         [PASS][1] -> [INCOMPLETE][2] +10 similar issues

Argh. These are due to the extra debug spam from
intel_bw_atomic_check() since that now goes throug the full function
a lot more.

So either we just nuke a bunch of those debugs, or I guess we do it a
bit more like what Stan suggested and try to check more carefully if the
inputs to to the QGV calculation actually changed.

I guess I'll try the latter approach, in case those debugs are actually
useful. The challenge will be doing that and keeping the patch looking
reasonable for stable...

-- 
Ville Syrjälä
Intel