Re: [PATCH 4.14 108/159] kvm, mm: account kvm related kmem slabs to kmemcg

[alexander . levin]

On Fri, Dec 22, 2017 at 07:22:32PM +0100, Michal Hocko wrote:
>On Fri 22-12-17 18:07:23, Sasha Levin wrote:
>> I don't try and override maintainers, I mostly try to get fixes out
>> of subsystems where maintainers/authors partially (or just don't)
>> mark their commits for stable.
>
>Well, I have see quite some MM patches and I believe we are quite good
>at marking patches for stable trees... I also think we we (as the whole
>kernel) are much better are using Fixes tag (although it is over used
>sometimes).

Indeed, mm/ is probably as good as it gets in the kernel.

>Moreover it makes more sense to push on those maintainers than try to
>substitude them without being so closely familiar with the subsystem. If
>missing backports result in bug reports then this just increase the
>pressure on those maintainers /me think.

Both is happening, but it's difficult to force maintainers into doing
anything, as you might have guessed...

I'm hoping that one result of this work is a tool we can stick into
scripts/ (maybe glue it to checkpatch) that'll alert when the patch
is -stable material and suggest adding tags.

>> These patches also go through a much longer review process than
>> commits that are marked for stable (there are at least 3 emails issued
>> for each such commit, and at least 1 week (usually much more) is
>> given for reviews).
>
>Does any of the maintainers read those emails? How many acks/reviewes do
>you get for those patches for the stable tree? To be honest I tend to

I get a fair amount of reviews which seems to be slightly above what
-stable tagged patches get, which is good.

Acks are not expected, and are not happening too often.

>ignore those patchbombs most of the time because it is simply impossible
>to handle them for me. I try to help backporting obvious fixes but
>reviewing seemingly randomly selected patch which applies and changelog
>looks reasonaly is simply out of my time budget. Not to mention that
>this is not just about the patch itself but also the tree it is applied
>to and other patches that are in the same pile.

I'd hope that these patches aren't "random" :)

For some background, this is based on Julia Lawall's work (and paper
https://soarsmu.github.io/papers/icse12-patch.pdf).

-- 

Thanks,
Sasha