Re: [PATH 5.10 0/4] xfs stable candidate patches for 5.10.y (part 1)

[Dave Chinner <david@fromorbit.com>]

On Fri, May 27, 2022 at 10:01:48AM +0300, Amir Goldstein wrote:
> On Fri, May 27, 2022 at 9:06 AM Christoph Hellwig <hch@infradead.org> wrote:
> >
> > FYI, below is the 5.10-stable backport I have been testing earlier this
> > week that fixes a bugzilla reported hang:
> >
> > https://bugzilla.kernel.org/show_bug.cgi?id=214767
> >
> > I was just going to submit it to the stable maintaines today after
> > beeing out of the holiday, but if you want to add it to this queue
> > that is fine with me as well.
> >
> 
> Let me take it for a short spin in out xfs stable test environment, since
> this env has caught one regression with an allegedly safe fix.
> This env has also VMs with old xfsprogs, which is kind of important to
> test since those LTS patches may end up in distros with old xfsprogs.
> 
> If all is well, I'll send your patch later today to stable maintainers
> with this first for-5.10 series.
> 
> > ---
> > From 8e0464752b24f2b3919e8e92298759d116b283eb Mon Sep 17 00:00:00 2001
> > From: Dave Chinner <dchinner@redhat.com>
> > Date: Fri, 18 Jun 2021 08:21:51 -0700
> > Subject: xfs: Fix CIL throttle hang when CIL space used going backwards
> >
> 
> Damn! this patch slipped through my process (even though I did see
> the correspondence on the list).
> 
> My (human) process has eliminated the entire 38 patch series
> https://lore.kernel.org/linux-xfs/20210603052240.171998-1-david@fromorbit.com/
> without noticing the fix that was inside it.

The first two times it was in much smaller patch series (5 and 8
patches total).


Also, you probably need to search for commit IDs on the list, too,
because this discussion was held in November about backporting the
fix to 5.10 stable kernels:

Subject: Help deciding about backported patch (kernel bug 214767, 19f4e7cc8197 xfs: Fix CIL throttle hang when CIL space used going backwards)
https://lore.kernel.org/linux-xfs/C1EC87A2-15B4-45B1-ACE2-F225E9E30DA9@flyingcircus.io/

> In this case, I guess Dave was not aware of the severity of the bug fixed

I was very aware of the severity of the problem, and I don't need
anyone trying to tell me what I should have been doing 18 months
ago.

It simply wasn't a severe bug. We had one user reporting it, and the
when I found the bug I realised that it was a zero-day thinko in
delayed logging accounting I made back in 2010 (~2.6.38 timeframe,
IIRC).  IOWs, it took 10 years before we got the first indication
there was a deep, dark corner case bug lurking in that code.

The time between first post of the bug fix and merge was about 6
months, so it also wasn't considered serious by anyone at the time
as it missed 2 whole kernel releases before it was reviewed and
merged...

There's been a small handful of user reports of this bug since (e.g
the bz above and the backport discussions), but it's pretty clear
that this bug is not (and never has been) a widespread issue.  It
just doesn't fit any of the criteria for a severe bug.

Backport candidate: yes. Severe: absolutely not.

-Dave.
-- 
Dave Chinner
david@fromorbit.com