[igt-dev] ✗ Fi.CI.IGT: failure for lib/debugfs: Sanity check even discarded CRCs (rev3)

[Ville Syrjälä]

On Fri, May 15, 2020 at 06:14:30PM -0000, Patchwork wrote:
> == Series Details ==
> 
> Series: lib/debugfs: Sanity check even discarded CRCs (rev3)
> URL   : https://patchwork.freedesktop.org/series/53982/
> State : failure
> 
> == Summary ==
> 
> CI Bug Log - changes from CI_DRM_8488_full -> IGTPW_4574_full
> ====================================================
> 
> Summary
> -------
> 
>   **FAILURE**
> 
>   Serious unknown changes coming with IGTPW_4574_full absolutely need to be
>   verified manually.
>   
>   If you think the reported changes have nothing to do with the changes
>   introduced in IGTPW_4574_full, please notify your bug team to allow them
>   to document this new failure mode, which will reduce false positives in CI.
> 
>   External URL: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_4574/index.html
> 
> Possible new issues
> -------------------
> 
>   Here are the unknown changes that may have been introduced in IGTPW_4574_full:
> 
> ### IGT changes ###
> 
> #### Possible regressions ####
> 
>   * igt at gem_ctx_ringsize@idle at rcs0:
>     - shard-iclb:         [PASS][1] -> [INCOMPLETE][2]
>    [1]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8488/shard-iclb8/igt@gem_ctx_ringsize@idle@rcs0.html
>    [2]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_4574/shard-iclb5/igt@gem_ctx_ringsize@idle@rcs0.html
> 
>   * igt at kms_cursor_edge_walk@pipe-a-256x256-bottom-edge:
>     - shard-iclb:         [PASS][3] -> [WARN][4] +33 similar issues
>    [3]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8488/shard-iclb8/igt@kms_cursor_edge_walk@pipe-a-256x256-bottom-edge.html
>    [4]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_4574/shard-iclb8/igt@kms_cursor_edge_walk@pipe-a-256x256-bottom-edge.html

Somewhat surprising result. kms_cursor_edge_walk sees zeroed
crcs all the time, but no other test ever sees them. Wasn't
quite expecting that.

Hmm. Now that I think about it IIRC there was some problem
with CRCs when all planes are turned off. I think the spec
even says the dmux crc source shouldn't be used with all planes
turned off, but I seem to recall the problem only manifesting
for a single frame when transitioning between no-planes vs. some
planes.

However I don't think kms_cursor_edge_walk should even be turning
off all the planes. This would seem to need a deeper investigation.

-- 
Ville Syrjälä
Intel