On Tue, Dec 23, 2025 at 08:54:34AM -0800, Andrew Morton wrote:
On Tue, 23 Dec 2025 06:43:45 -0500 Sasha Levin <[email protected]> wrote:
>> > Yes, -stable maintainers have been asked to only backport patches where
>> > the MM developers asked for that, with cc:stable. There may be
>> > slipups, but as far as I know this is working.
>> >
>> > I don't actually know how they determine which patches need this
>> > special treatment. Pathname? Signed-off-by:akpm?
>>
>> I guess it is pathname, based on ignore_list file [1] of stable-queue repo.
>>
>> [1]
https://git.kernel.org/pub/scm/linux/kernel/git/stable/stable-queue.git/tree/ignore_list#n16
>>
>
>Oh, that's a bit sad.
>
>- other trees sometimes mess with mm/ and they probably aren't aware
> that they need an explicit cc:stable.
>
>- misses drivers/block/zram and probably various other things that
> the MM team maintains.
>
>Oh well, I guess simple mm/* coverage is good enough. But I do worry a
>little that useful fixes coming into mm/ via other trees without
>cc:stable will get missed.
How should we improve the filter? mm/ AND signed off by akpm?
I think just signed-off-by:akpm please. That way, mm fixes which come
in via other trees without cc:stable get backported.
Ack
Obviously we'd prefer that such patches get appropriate consideration
by the MM developers but sometimes other-tree people aren't that
cooperative. In this case it's better to backport the thing rather
than missing a fix?
I tend to err on the side of taking one extra thing rather than missing
something :)
Would you be interested in trying out AUTOSEL for mm/ again? In the current
workflow I give about 2-3 weeks for review, and all it takes to have a patch
dropped is to just reply with a "no".
I can generate a series with mm/ patches from v6.18..v6.19-rc2 that don't have
a Fixes/stable tag but the LLM thinks that it should be backported as a way for
you and other mm/ folk to gauge the current state of AUTOSEL, if that helps?
--
Thanks,
Sasha
