Re: refcount_warn_saturate in __btrfs_release_delayed_node for 6.18.2-arch2-1

[Leo Martins <loemra.dev@gmail.com>]

On Wed, 7 Jan 2026 14:15:56 +0100 Christian Heusel <christian@heusel.eu> wrote:

> On 26/01/07 02:52AM, Christian Heusel wrote:
> > On 26/01/06 05:10PM, Leo Martins wrote:
> > > On Tue, 06 Jan 2026 22:01:58 +0000 "Gideon Farrell" <gideon@solnickfarrell.co.uk> wrote:
> > > 
> > > > Hi there,
> > > > 
> > > > I recently experienced a type of crash I haven't seen before on this system which seems to originate in __btrfs_release_delayed_node on Kernel 6.18.2-arch2-1.
> > > 
> > > Hey, thanks for the report. This looks very similar to a different
> > > report that has been fixed in 6.19-rc5.
> > > Report: https://lore.kernel.org/oe-lkp/202511262228.6dda231e-lkp@intel.com/
> > 
> > I think this is the issue we have been seeing on the Arch Linux
> > infrastructure aswell, see this issue for reference:
> > 
> > https://gitlab.archlinux.org/archlinux/infrastructure/-/issues/788#note_385420
> > 
> > This has caused us to downgrade the kernel on all hosts in our
> > infrastructure due to the crashes caused by this.
> > 
> > > Fix: https://lore.kernel.org/linux-btrfs/7c89417ac3352ce3cb0a6373a1746155c1e2754d.1765588168.git.loemra.dev@gmail.com/
> > > 
> > > Please let me know if this fixes your issue.
> > 
> > So far we have not found a good lab-setting to reproduce the issue
> > outside of the production machines workloads (but we have also not
> > looked into the issue a lot).
> > 
> > Do you have a good reproducer already that we could use to verify the
> > fix?

I believe the lkp-test from the original report is able to consistently
reproduce this.

> > 
> > Also the fix is not yet even scheduled for inclusion in the stable
> > trees, since it is still not in linus tree right?
> 
> Sorry, I messed up here! I got fooled by the output of `git log --grep`
> and therefore missed the commit! As pointed out by heftig on gitlab the
> commit already is present in 6.18.3 which I'll try to roll out to our
> systems shortly! :)

Let me know if you're still seeing any issues after rolling out the fix.

Thanks,
Leo Martins.

> 
> > 
> > > 
> > > > 
> > > > Here's the stack trace:
> > > > 
> > > > ```
> > > 
> > > Are these the first refcount_t: warnings in your dmesg? I would
> > > expect an earlier warning that looks like
> > > refcount_t: addition on 0; use-after-free.
> > > 
> > 
> > This is what our stacktrace looks like:
> > 
> > [...]
> > 
> > Cheers,
> > Chris
> 
> 
>