Re: [REGRESSION] 9pfs issues on 6.12-rc1

[Christian Brauner]

On Thu, Oct 17, 2024 at 08:00:35PM +0200, Antony Antony wrote:
> Hi,
> 
> On Thu, Oct 03, 2024 at 03:12:15AM +0200, Sedat Dilek wrote:
> > On Wed, Oct 2, 2024 at 11:58 PM Maximilian Bosch <maximilian@xxxxxxxxx> wrote:
> > >
> > > Good evening,
> > >
> > > thanks a lot for the quick reply!
> > >
> > > > A fix for it is already pending in the vfs.fixes branch and -next:
> > > > https://lore.kernel.org/all/cbaf141ba6c0e2e209717d02746584072844841a.1727722269.git.osandov@xxxxxx/
> > >
> > > I applied the patch on top of Linux 6.12-rc1 locally and I can confirm
> > > that this resolves the issue, thanks!
> 
> Maximilian, would you like to re-run the test a few times? I wonder if there 
> is another intermittend bug related to the same commit.
> 
> > >
> > > With best regards
> > >
> > > Maximilian
> > >
> > 
> > Thanks for testing.
> > 
> > For the records:
> > 
> > iov_iter: fix advancing slot in iter_folioq_get_pages()
> > https://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs.git/commit/?h=vfs.fixes&id=0d24852bd71ec85ca0016b6d6fc997e6a3381552
> > 
> > https://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs.git/log/?h=vfs.fixes
> 
> I’m still seeing a kernel oops after the fix in 6.12-rc3, but I’ve noticed 
> that the issue is no longer 100% reproducible. Most of the time, the system 
> crashes. Before this fix it was 100% reproducible.
> 
> When using the nix testing, I have to force the test to re-run.
> 
> result=$(readlink -f ./result); rm ./result && nix-store --delete $result
> 
> nix-build -v nixos/tests/kernel-generic.nix -A linux_testing
> 
> So may be there is a new bug showing up after the fix. I have reported it.
> 
> https://lore.kernel.org/regressions/ZxFEi1Tod43pD6JC@xxxxxxxxxxxxxxx/T/#u

I've pinged David again so hopefully we'll have a fix soon.