Re: Linux regressions report for mainline [2022-03-06]

[Thorsten Leemhuis]

On 06.03.22 22:33, Linus Torvalds wrote:
> On Sun, Mar 6, 2022 at 11:58 AM Regzbot (on behalf of Thorsten
> Leemhuis) <regressions@xxxxxxxxxxxxx> wrote:
>>
>> ========================================================
>> current cycle (v5.16.. aka v5.17-rc), culprit identified
>> ========================================================
>>
>> Follow-up error for the commit fixing "PCIe regression on APM Merlin (aarch64 dev platform) preventing NVME initialization"
>> ---------------------------------------------------------------------------------------------------------------------------
>> https://linux-regtracking.leemhuis.info/regzbot/regression/Yf2wTLjmcRj+AbDv@xps13.dannf/
>> https://lore.kernel.org/stable/Yf2wTLjmcRj%2BAbDv@xps13.dannf/
>>
>> By dann frazier, 29 days ago; 7 activities, latest 23 days ago; poked 13 days ago.
>> Introduced in c7a75d07827a (v5.17-rc1)
> 
> Hmm. The culprit may be identified, but it looks like we don't have a
> fix for it, so this may be one of those "left for later" things. It
> being Xgene, there's a limited number of people who care, I'm afraid.
> 
> Alternatively, maybe 6dce5aa59e0b ("PCI: xgene: Use inbound resources
> for setup") should just be reverted as broken?

I don't care much, I just hope someone once again will look into this,
as this (and the previous) regression are on my list for quite a while
already and process once again seems to have slowed down. :-/

>> ====================================================
>> current cycle (v5.16.. aka v5.17-rc), unknown culprit
>> ====================================================
>>
>>
>> net: bluetooth: qualcom and intel adapters, unable to reliably connect to bluetooth devices
>> -------------------------------------------------------------------------------------------
>> https://linux-regtracking.leemhuis.info/regzbot/regression/CAJCQCtSeUtHCgsHXLGrSTWKmyjaQDbDNpP4rb0i+RE+L2FTXSA@xxxxxxxxxxxxxx/
>> https://lore.kernel.org/linux-bluetooth/CAJCQCtSeUtHCgsHXLGrSTWKmyjaQDbDNpP4rb0i%2BRE%2BL2FTXSA@xxxxxxxxxxxxxx/
>>
>> By Chris Murphy, 23 days ago; 47 activities, latest 3 days ago.
>> Introduced in v5.16..f1baf68e1383 (v5.16..v5.17-rc4)
>>
>> Fix incoming:
>> * https://lore.kernel.org/regressions/1686eb5f-7484-8ec2-8564-84fe04bf6a70@xxxxxxxxxxxxx/
> 
> That's a recent fix, it seems to be only in the bluetooth tree right
> now, and won't be in rc7. I'm hoping that I'll get it in next week's
> networking dump.

You will, it afaics already got merged into the right branch already:
https://lore.kernel.org/all/164637060982.7623.2431046217416492091.git-patchwork-notify@xxxxxxxxxx/

It just came slightly too late for the last pull into the net. The same
had happened a week earlier with regression fixes, that's why I already
brought this up:

https://lore.kernel.org/netdev/37349299-c47b-1f67-2229-78ae9b9b4488@xxxxxxxxxxxxx/

Short story: things should hopefully improve soon.

BTW: @bluetooth maintainers, I noticed your -next branch is in next, but
your -fixes branch
git://git.kernel.org/pub/scm/linux/kernel/git/bluetooth/bluetooth.git is
not. Is that on purpose? Some (many?) of the other trees downstream to
the net tree have both their -fixes and their -next branches included
there regularly.

> Cc'ing the right people just to prod them, since we've had much too
> many "Oh, I didn't even realize it was a regression" issues this time
> around.

Yeah, since I started regression tracking again I noticed that quite a
few regressions fixes are merged quite slowly because maintainers put
them in their -next branches instead of sending them towards mainline
within a few days. I always try to make the maintainers aware of this
when I noticed it. That already helped a little afaics, but there is
more work ahead of me afaics...

Ciao, Thorsten