[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20221122174807.GA9658@linuxonhyperv3.guj3yctzbm1etfxqx2vob5hsef.xx.internal.cloudapp.net>
Date: Tue, 22 Nov 2022 09:48:07 -0800
From: Jeremi Piotrowski <jpiotrowski@...ux.microsoft.com>
To: Jan Kara <jack@...e.cz>
Cc: Thilo Fromm <t-lo@...ux.microsoft.com>,
Ye Bin <yebin10@...wei.com>, jack@...e.com, tytso@....edu,
linux-ext4@...r.kernel.org, regressions@...ts.linux.dev
Subject: Re: [syzbot] possible deadlock in jbd2_journal_lock_updates
On Tue, Nov 22, 2022 at 12:57:15PM +0100, Jan Kara wrote:
> On Mon 21-11-22 10:15:58, Jeremi Piotrowski wrote:
> > On Mon, Nov 21, 2022 at 04:00:18PM +0100, Jan Kara wrote:
> > >
> > > OK, attached patch fixes the deadlock for me. Can you test whether it fixes
> > > the problem for you as well? Thanks!
> >
> > I'll test the fix tomorrow, but I've noticed it doesn't apply cleanly to
> > 5.15.78, which seems to be missing:
> >
> > - 5fc4cbd9fde5d4630494fd6ffc884148fb618087 mbcache: Avoid nesting of cache->c_list_lock under bit locks
> > (this one is marked for stable but not in 5.15?)
> > - 307af6c879377c1c63e71cbdd978201f9c7ee8df mbcache: automatically delete entries from cache on freeing
> > (this one is not marked for stable)
> >
> > So either a special backport is needed or these two would need to be applied as
> > well.
>
> Right. The fix is against current mainline kernel. Stable tree backports
> are a second step once the fix is confirmed to work :). Let me know in case
> you need help with porting to the kernel version you need for testing.
>
I confirm that this patch fixes things with the more complicated reproducer :)
Attached is my tweaked version of the patch that applies against 5.15.
View attachment "0001-ext4-Fix-deadlock-due-to-mbcache-en.patch" of type "text/plain" (4665 bytes)
Powered by blists - more mailing lists