Re: ✗ Xe.CI.Full: failure for tests/intel/xe_wedged: Manage verbosity of GT reset fault injection

[Matt Roper <matthew.d.roper@intel.com>]

On Wed, Nov 27, 2024 at 12:30:49AM +0000, Patchwork wrote:
> == Series Details ==
> 
> Series: tests/intel/xe_wedged: Manage verbosity of GT reset fault injection
> URL   : https://patchwork.freedesktop.org/series/141810/
> State : failure
> 
> == Summary ==
> 
> CI Bug Log - changes from XEIGT_8126_full -> XEIGTPW_12203_full
> ====================================================
> 
> Summary
> -------
> 
>   **FAILURE**
> 
>   Serious unknown changes coming with XEIGTPW_12203_full absolutely need to be
>   verified manually.
>   
>   If you think the reported changes have nothing to do with the changes
>   introduced in XEIGTPW_12203_full, please notify your bug team (I915-ci-infra@lists.freedesktop.org) to allow them
>   to document this new failure mode, which will reduce false positives in CI.
> 
>   
> 
> Participating hosts (4 -> 4)
> ------------------------------
> 
>   No changes in participating hosts
> 
> Possible new issues
> -------------------
> 
>   Here are the unknown changes that may have been introduced in XEIGTPW_12203_full:
> 
> ### IGT changes ###
> 
> #### Possible regressions ####
> 
>   * igt@core_hotunplug@hotreplug:
>     - shard-lnl:          [PASS][1] -> [DMESG-WARN][2]
>    [1]: https://intel-gfx-ci.01.org/tree/intel-xe/IGT_8126/shard-lnl-5/igt@core_hotunplug@hotreplug.html
>    [2]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_12203/shard-lnl-6/igt@core_hotunplug@hotreplug.html

This change to the xe_wedged test would not impact the behavior of the
core_hotunplug test.

> Known issues
> ------------
<snip>
>   * igt@xe_wedged@basic-wedged:
>     - shard-dg2-set2:     [SKIP][800] ([Intel XE#1130]) -> [DMESG-WARN][801] ([Intel XE#2919])
>    [800]: https://intel-gfx-ci.01.org/tree/intel-xe/IGT_8126/shard-dg2-466/igt@xe_wedged@basic-wedged.html
>    [801]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_12203/shard-dg2-435/igt@xe_wedged@basic-wedged.html

Although this patch completely fixed the issue for me on the LNL system
I was using to test locally, it looks like we do still need to tell
igt_runner to ignore one more non-stackdump message to completely fix
the bug on all platforms (which I think the other patch on the list from
Kamil should take care of).

Applied to master with Kamil's r-b from the gitlab comment.  Since it
looks like this wasn't a full fix (despite being a move in the right
direction), I re-opened the gitlab issue that got auto-closed as well
until we have the rest of the fix.


Matt

-- 
Matt Roper
Graphics Software Engineer
Linux GPU Platform Enablement
Intel Corporation