Re: [PATCH v2 00/13] mm: batch of (urgentish?) fixes in oomk, cachestat, dev-migrate, percpu

From: Lorenzo Stoakes (ARM)

Date: Thu Nov 19 2026 - 16:58:36 EST


On Wed, Nov 18, 2026 at 08:47:49PM +0100, David Hildenbrand (Arm) wrote:
> On 11/18/26 18:21, Lorenzo Stoakes (ARM) wrote:
> > On Wed, Nov 18, 2026 at 04:35:32PM +0100, David Hildenbrand (Arm) wrote:
> >> On 11/13/26 22:45, Lorenzo Stoakes (ARM) wrote:
> >>> A bunch of fixes for oomk, cachestat, device migration and the per-CPU
> >>> allocator.
> >>>
> >>> These patches were originally AI-generated, tamed into upstreamable state.
> >>>
> >>> Some are more reworked than others. Those which needed a significant
> >>> rethink are submitted as mine and marked as Reported-by the relevant
> >>> original author, others are smaller edits which are annotated as such in
> >>> the usual fashion.
> >>>
> >>> Please check for correctness, sanity, and whether the cc: stable is
> >>> warranted (I have not included it for 3 where it seemed unwarranted.)
> >>>
> >>> Signed-off-by: Lorenzo Stoakes (ARM) <ljs@xxxxxxxxxx>
> >>> ---
> >>
> >> Lorenzo, I would pick up the migrate_device patches now.
> >>
> >> Which one should be fast-tracked upstream and which can wait for the next cycle?
> >>
> >> Use-after-free, shadow stack and out-of-bounds write ... look rather bad.
> >
> > I think given Gregory was able to repro nearly all of them, I'd say most. They
> > are mostly small. So patches 4-9 I think should be.
>
> Just to clarify: The later we are in the cycle the less non-hotfixes Linus wants
> to see. AFAIU, really urgent stuff is the exception.
>
> So the question is less "can this be reproduced" but rather "is is that bad that
> we want to fix it asap".

Yeah I guess I'm kinda saying it's hard to differentiate between them :) or
maybe I just want the migrate-device horrors to stop :'(

Given the ones I reworked fully were more involved maybe those are higher
priority than the rest, as a litmus test?

>
> >
> > Actually the shadow stack one (10) is unrepro'd and super borderine and requires
> > a broken driver and etc. etc. so I'd say that one is less important.
>
> Ack.
>
> --
> Cheers,
>
> David

--
Cheers, Lorenzo