[Intel-gfx] [drm-intel:drm-intel-gt-next] [drm/i915] [confidence: ] 9bb66c179f: assertion_failure

[Oliver Sang]

hi, Jani,

On Fri, Dec 08, 2023 at 11:59:43AM +0200, Jani Nikula wrote:
> On Thu, 07 Dec 2023, kernel test robot <oliver.sang at intel.com> wrote:
> > Hello,
> >
> > kernel test robot noticed "assertion_failure" on:
> >
> > commit: 9bb66c179f50e61df20ba13c9b34ca17d00b05fb ("drm/i915: Reserve some kernel space per vm")
> > git://anongit.freedesktop.org/drm-intel drm-intel-gt-next
> >
> > in testcase: igt
> 
> So the kernel test robot runs igt. I've seen a handful of reports over
> the years, but not a whole lot. If you run it even semi-regularly, I
> would have expected more. What's the deal here?

our team merges lots of linux kernel repo/tree in so-called hourly kernels,
tests these hourly kernels by 80+ tools (including igt) on various platforms,
if any failure found comparing to some good bases, we will trigger auto-bisect
to try to catpure fbc (first bad commit) and report. this is the reason you
see this report.

we cannot gurantee coverage since resource constraints, possible failures
in different stage such like merging, building, auto-bisect, and so on. this
means we could fail to capture some issues.

> 
> There's clearly overlap with what our CI is doing. Maybe better
> coordination would be useful? Especially wrt reporting. I'm not sure if
> anyone's going to track these mails.
> 
> Cc: Ewelina

for this case we just capture a assertion_failure upon
branch: drm-intel/drm-intel-gt-next
on a Commet Lake (with 16G memory). not sure if this is valuable to you,
say, for some legacy platform regression check?

we want to seek advices from you:
(1) if this is still useful, do you want us to limit the receivers of this kind
    of report?
(2) or would you suggest there is no need for us to test below repo at all?
    git://anongit.freedesktop.org/drm-intel

> 
> > version: igt-x86_64-0f075441-1_20230520
> 
> That's six months old and more than 1k commits behind. The results are
> going to be useless, I'm afraid.

got it. if you still want us to keep the test upon the repo, we would upgrade
igt to latest version.

> 
> 
> BR,
> Jani.
> 
>