Re: Known and unfixed active data loss bug in MM + XFS with large folios since Dec 2021 (any kernel from 6.1 upwards)
From: Kairui Song
Date: Tue Oct 01 2024 - 17:11:00 EST
On Fri, Sep 27, 2024 at 10:58 PM Jens Axboe <axboe@xxxxxxxxx> wrote:
>
> On 9/27/24 8:51 AM, Sam James wrote:
> > Kairui Song <ryncsn@xxxxxxxxx> writes:
> >
> >> On Wed, Sep 25, 2024 at 1:16?AM Sam James <sam@xxxxxxxxxx> wrote:
> >>>
> >>> Kairui, could you send them to the stable ML to be queued if Willy is
> >>> fine with it?
> >>>
> >>
> >> Hi Sam,
> >
> > Hi Kairui,
> >
> >>
> >> Thanks for adding me to the discussion.
> >>
> >> Yes I'd like to, just not sure if people are still testing and
> >> checking the commits.
> >>
> >> And I haven't sent seperate fix just for stable fix before, so can
> >> anyone teach me, should I send only two patches for a minimal change,
> >> or send a whole series (with some minor clean up patch as dependency)
> >> for minimal conflicts? Or the stable team can just pick these up?
> >
> > Please see https://www.kernel.org/doc/html/v6.11/process/stable-kernel-rules.html.
> >
> > If Option 2 can't work (because of conflicts), please follow Option 3
> > (https://www.kernel.org/doc/html/v6.11/process/stable-kernel-rules.html#option-3).
> >
> > Just explain the background and link to this thread in a cover letter
> > and mention it's your first time. Greg didn't bite me when I fumbled my
> > way around it :)y
> >
> > (greg, please correct me if I'm talking rubbish)
>
> It needs two cherry picks, one of them won't pick cleanly. So I suggest
> whoever submits this to stable does:
>
> 1) Cherry pick the two commits, fixup the simple issue with one of them.
> I forget what it was since it's been a week and a half since I did
> it, but it's trivial to fixup.
>
> Don't forget to add the "commit XXX upstream" to the commit message.
>
> 2) Test that it compiles and boots and send an email to
> stable@xxxxxxxxxxxxxxx with the patches attached and CC the folks in
> this thread, to help spot if there are mistakes.
>
> and that should be it. Worst case, we'll need a few different patches
> since this affects anything back to 5.19, and each currently maintained
> stable kernel version will need it.
>
Hi Sam, Jens,
Thanks very much, currently maintained upstream kernels are
6.10, 6.6, 6.1, 5.15, 5.10, 5.4, 4.19.
I think only 6.6 and 6.1 need backport, I've sent a fix for these two,
it's three checkpicks from the one 6.10 series so the conflict is
minimal. The stable series can be applied without conflict for both.