Quoting Maarten Lankhorst (2026-09-18 17:24:39)
> Hello,
> 
> 
> On 9/18/26 12:12, Joonas Lahtinen wrote:
> > Quoting Tvrtko Ursulin (2026-09-18 12:44:43)
> >>
> >> On 18/09/2026 09:31, Sebastian Andrzej Siewior wrote:
> >>> On 2026-07-09 13:36:57 [+0200], Maarten Lankhorst wrote:
> >>>> Hey,
> >>> Hi Maarten,
> >>>
> >>>> Yeah the problem was similar, I believe neither check should exist,
> >>>> instead the caller should give this information like the comment 
> >>>> mentions.
> >>>
> >>> Any update here? Maybe I have been left out on reposts. Any chance this
> >>> gets in before the LTS of this year gets out?
> >>
> >> Side question, there was a bunch of sashiko issues against v4 - have 
> >> those been addressed?
> >>
> >> As today it is the first time I see patch 7/7 which is a bit of a scary 
> >> one. I have to ask how much does Intel commit to fixing out any 
> >> potential breakage given the status of i915?
> > 
> > I'd say there is very close to zero chance of accepting patch 7. And
> > that would take massive functional and performance validation efforts
> > done by the person wanting the patch to get merged.
> > 
> 
> I don't mind taking the responsibility to debug and troubleshoot if things 
> break.
> Patch 7/7 fixes a real bug that is hard to trigger otherwise, but can most 
> definitely
> happen.
> 
> I believe it should be treated it as any other fix, merge it and when there 
> are
> bug reports, solve those.
>
> Most of my energy and work is going to other projects now, so I won't be able 
> to do
> the entire validation in advance.

For a patch with such potential to cause breakage (functional and
performance), we really need the validation efforts up front.

> I will definitely ensure CI is clean and passing,
> it's how I came across this bugfix in the first place.
> 
> If bug reports come in and point to this patch, I will be able to take a look 
> and
> solve the reported issues, but I won't be able to hunt for them in advance.

I think somebody will actually need to run real workloads on such kernel
across a few HW generations.

Couple of browser based GPU benchmarks and video transcode benchmarks
would be a good start.

> It's only a few weeks until the -rc cutoff, so we can make it part of the 
> next cycle.

Definitely not a fan of sending such a patch to mainline under the holiday
season without any testing.

Regards, Joonas

> > And yes, there would also need to be commitment to fix any fallout by
> > that same person in weekly -rc cadence. So please raise your hand, whoever
> > wants to volunteer.
> > 
> > There's already the earlier much smaller patch about vblank that needs
> > to get reviewes and acceptance and it hasn't been able to be driven
> > to completion.
> > 
> > So projecting from that, I don't exactly expect patch like 7 in this
> > series to get merged in any near future without somebody really stepping
> > forward and signing to do all the validation and and perf regression
> > testing.
> > 
> > Regards, Joonas
> Kind regards,
> ~Maarten Lankhorst

Reply via email to