[PATCH v1 1/2] drm/ci: uprev mesa

[Dmitry Baryshkov]

On Sat, Nov 15, 2025 at 05:55:17AM +0200, Dmitry Baryshkov wrote:
> On Sat, Nov 15, 2025 at 02:44:56AM +0200, Dmitry Baryshkov wrote:
> > On Fri, Nov 14, 2025 at 08:30:50AM +0530, Vignesh Raman wrote:
> > > Uprev mesa to adapt to the latest changes in Mesa CI, including support
> > > for firmware via LAVA overlays, removal of the python-artifacts job,
> > > split container and build rules, use lava-job-submitter container,
> > > and various misc fixes.
> > > 
> > > Co-developed-by: Dmitry Baryshkov <dmitry.baryshkov at oss.qualcomm.com>
> > > Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov at oss.qualcomm.com>
> > > Signed-off-by: Vignesh Raman <vignesh.raman at collabora.com>
> > > ---
> > >  drivers/gpu/drm/ci/build.yml         |  22 +++---
> > >  drivers/gpu/drm/ci/container.yml     |  28 ++++++--
> > >  drivers/gpu/drm/ci/gitlab-ci.yml     |  95 ++++++++++++++++++++-----
> > >  drivers/gpu/drm/ci/igt_runner.sh     |   4 +-
> > >  drivers/gpu/drm/ci/image-tags.yml    |  22 +++---
> > >  drivers/gpu/drm/ci/lava-submit.sh    | 101 +++++++++++++--------------
> > >  drivers/gpu/drm/ci/static-checks.yml |   1 +
> > >  drivers/gpu/drm/ci/test.yml          |  18 +++--
> > >  8 files changed, 186 insertions(+), 105 deletions(-)
> > > 
> > 
> > Thanks a lot for taking care of it!
> > 
> > 
> > Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov at oss.qualcomm.com>
> 
> Hmm, this causes a lot of failures, see [1] (note, the run is not
> finished yet):
> - On some platforms (msm, rk3399, mt8173) the
>   core_setmaster at master-drop-set-root test fails. On other it seems to
>   work correctly
> - sm8350 seems to have some issues with this uprev, it cant' finish the
>   jobs at all
> - mt8173 and rockchip:rk3399 seem to have other test failures too

I have fixed most of the fallouts, see [2]. But ideally this should be
split into two parts: fix before mesa uprev (I think some of them are
already there), uprev mesa, applying necessary fixes and then enable apq
jobs.

At this point I have no idea what's wrong with MT8183. Most likely it
needs either a DT fix or a driver fix. The driver fails to probe because
it can't build a sensible component chain.

On i915 several tests kill the hardware, making it timeout all remaining
tests in a bad way.

BTW: I see that for some of the drivers we skip the absolute majority of
the tests. What are we actually testing on those?

> 
> Could you please update fails / flakes tor all the affected platforms?
> Also I'm really interested in the master-drop-set tests. Why do they
> fail on some of the platforms, but not on all? This seems to be some
> infrastructure-related issue, since exactly the same kernel with the
> same IGT passes those tests (at least on msm hardware).
> 
> [1] https://gitlab.freedesktop.org/drm/msm/-/pipelines/1547684

[2] https://gitlab.freedesktop.org/lumag/msm/-/commits/msm-next-lumag-db820c


-- 
With best wishes
Dmitry