On 15.02.24 10:13, Greg KH wrote:
> On Thu, Feb 15, 2024 at 09:45:01AM +0100, Thorsten Leemhuis wrote:
>> Linus, what...
>>
>> On 14.02.24 18:22, Vincent Guittot wrote:
>>> On Wed, 14 Feb 2024 at 18:20, Linus Torvalds
>>> <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
>>>> On Wed, 14 Feb 2024 at 09:12, Jon Hunter <jonathanh@xxxxxxxxxx> wrote:
>>>>> We have also observed a performance degradation on our Tegra platforms
>>>>> with v6.8-rc1. Unfortunately, the above change does not fix the problem
>>>>> for us and we are still seeing a performance issue with v6.8-rc4. For
>>>>> example, running Dhrystone on Tegra234 I am seeing the following ...
>>>>> [...]
>>>>> If I revert this change and the following ...
>>>>> b3edde44e5d4 ("cpufreq/schedutil: Use a fixed reference frequency")
>>>>> f12560779f9d ("sched/cpufreq: Rework iowait boost")
>>>>> 9c0b4bb7f630 ("sched/cpufreq: Rework schedutil governor
>>>>> ... then the perf is similar to where it was ...
>>>>
>>>> Ok, guys, this whole scheduler / cpufreq rewrite seems to have been
>>>> completely buggered.
>>>> [...]
>>> This should fix it:
>>> https://lore.kernel.org/lkml/20240117190545.596057-1-vincent.guittot@xxxxxxxxxx/
>>
>> ...do you want me to do in situations like this? I'm asking, as I see
>> situations like this frequently -- e.g. people reporting problems a
>> second, third, or fourth time while the fix is already sitting in -next
>> for a few days.
>>
>> Want me to list them in the weekly reports so that you can cherry-pick
>> them from -next if you want?
>
> Poke the maintainer to get off their butt and submit the pull request to
> Linus
Well, I did that sometimes and will continue to do so. But some
maintainers then feel pestered and become annoyed by my efforts -- which
in the long-term is counter productive, as regression tracking will only
work well if maintainers and I work well together. That's why I'm a bit
careful with such things (side note: don't worry, I know that some
conflict is inevitable -- but I don't have your or Linus standing, so I
have to choose my fights carefully...).
I sometimes also got replies along the lines of "we are only at -rc2,
this can wait till -rc5 or -rc6" -- and I have no quote from Linus at
hand I can point maintainers to that says something along the lines of
"if a regression fix was in -next for at least two days, submit it to
mainline before the next -rc, unless there is a strong reason why that
particular fix needs more testing" (or whatever he actually wants).
> (note, this is me in this case...)
> I'll get it into the next -rc, sorry for the delay, other things got in
> the way, my fault.
Happens, I don't care too much about this specific event, more about the
general problem. Especially the -mm tree bugs me sometimes, as I noticed
that Andrew often lets regression fixes linger in -next for round about
a week; he furthermore sometimes sends this stuff to Linus on Mondays or
Tuesdays. Due to that the fixes often miss at least one, sometimes two
-rcs. That is especially hard to watch if the regression made it to a
stable kernel and you are waiting for the fix to get mainlined.
Ciao, Thorsten
