[Intel-gfx] ✗ Fi.CI.BAT: failure for series starting with [1/5] drm/i915/selftests: Force bonded submission to overlap

[Chris Wilson]

Quoting Patchwork (2019-11-22 13:30:26)
> == Series Details ==
> 
> Series: series starting with [1/5] drm/i915/selftests: Force bonded submission to overlap
> URL   : https://patchwork.freedesktop.org/series/69887/
> State : failure
> 
> == Summary ==
> 
> CI Bug Log - changes from CI_DRM_7404 -> Patchwork_15394
> ====================================================
> 
> Summary
> -------
> 
>   **FAILURE**
> 
>   Serious unknown changes coming with Patchwork_15394 absolutely need to be
>   verified manually.
>   
>   If you think the reported changes have nothing to do with the changes
>   introduced in Patchwork_15394, 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/Patchwork_15394/index.html
> 
> Possible new issues
> -------------------
> 
>   Here are the unknown changes that may have been introduced in Patchwork_15394:
> 
> ### IGT changes ###
> 
> #### Possible regressions ####
> 
>   * igt at i915_selftest@live_gem_contexts:
>     - fi-skl-lmem:        [PASS][1] -> [TIMEOUT][2]
>    [1]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_7404/fi-skl-lmem/igt@i915_selftest@live_gem_contexts.html
>    [2]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_15394/fi-skl-lmem/igt@i915_selftest@live_gem_contexts.html

Well at least now I know that randomly moving code around
i915_request_retire() is not the solution.

If it's the same as before, and I can't see any reason why it wouldn't
be, somewhere between engine_retire(), intel_gt_retire_requests() and
__engine_park() we drop the kernel_context from the
gt->timelines.active_list.
-Chris