Re: ✗ i915.CI.BAT: failure for drm/{i915, xe}: display (runtime) suspend/resume/shutdown unification, part 2

[Jani Nikula <jani.nikula@intel.com>]

On Wed, 03 Jun 2026, Imre Deak <imre.deak@intel.com> wrote:
> On Wed, Jun 03, 2026 at 04:29:19PM +0300, Jani Nikula wrote:
>> On Tue, 02 Jun 2026, Jani Nikula <jani.nikula@intel.com> wrote:
>> > On Tue, 02 Jun 2026, Patchwork <patchwork@emeril.freedesktop.org> wrote:
>> >> #### Possible regressions ####
>> >>
>> >>   * igt@i915_selftest@live:
>> >>     - fi-bsw-n3050:       [PASS][1] -> [ABORT][2] +1 other test abort
>> >>    [1]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_18602/fi-bsw-n3050/igt@i915_selftest@live.html
>> >>    [2]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_167725v1/fi-bsw-n3050/igt@i915_selftest@live.html
>> >>
>> >>   * igt@i915_selftest@live@hangcheck:
>> >>     - bat-arlh-2:         [PASS][3] -> [INCOMPLETE][4]
>> >>    [3]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_18602/bat-arlh-2/igt@i915_selftest@live@hangcheck.html
>> >>    [4]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_167725v1/bat-arlh-2/igt@i915_selftest@live@hangcheck.html
>> >
>> > Imre, Ville, any idea what in this series could cause this? Feels like
>> > it could be a real regression, but I can't figure it out. :(
>> 
>> Passed on the next try. I'll optimistically think the first try was a
>> fluke. ;)
>
> The Oops in __drm_mm_interval_first() on fi-bsw-n3050 looks like an
> actual driver problem, unless it's a random memory corruption due to HW
> errors. I couldn't find any previous instance of this Oops in CI runs,
> but I also can't see how it would relate to the changes in this
> patchset. Since these selftests involve a lot unloading/reloading of the
> driver one theory was that it's due to unproper unloading of the driver
> after the test preceeding the above igt@i915_selftest@live test
> (gem_migrate), but couldn't see any proof for that either and not sure
> how such an issue would cause a memory corruption in the reloaded driver
> like the above.

Thanks a lot for looking. I didn't intend for you to go into a sort of
wild goose chase with this, but to see if you could figure something out
quickly when I couldn't.

Anyway, this just reaffirms I'm not going to let this block merging if
retry passes.

BR,
Jani.



>
>> BR,
>> Jani.
>> 
>> -- 
>> Jani Nikula, Intel

-- 
Jani Nikula, Intel