Re: [syzbot] [udf?] possible deadlock in udf_free_blocks

[Jan Kara]

On Tue 03-12-24 04:41:03, syzbot wrote:
> syzbot has bisected this issue to:
> 
> commit 7908632f2927b65f7486ae6b67c24071666ba43f
> Author: Maíra Canal <mcanal@xxxxxxxxxx>
> Date:   Thu Sep 14 10:19:02 2023 +0000
> 
>     Revert "drm/vkms: Fix race-condition between the hrtimer and the atomic commit"

This is obviously bogus. Apparently the reproducer is not quite reliable...
But I can see where's the problem in UDF code.

								Honza

> bisection log:  https://syzkaller.appspot.com/x/bisect.txt?x=13dbe80f980000
> start commit:   cdd30ebb1b9f module: Convert symbol namespace to string li..
> git tree:       upstream
> final oops:     https://syzkaller.appspot.com/x/report.txt?x=103be80f980000
> console output: https://syzkaller.appspot.com/x/log.txt?x=17dbe80f980000
> kernel config:  https://syzkaller.appspot.com/x/.config?x=91c852e3d1d7c1a6
> dashboard link: https://syzkaller.appspot.com/bug?extid=d472c32c5dd4cd2fb5c5
> syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=117440f8580000
> C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=1659b5e8580000
> 
> Reported-by: syzbot+d472c32c5dd4cd2fb5c5@xxxxxxxxxxxxxxxxxxxxxxxxx
> Fixes: 7908632f2927 ("Revert "drm/vkms: Fix race-condition between the hrtimer and the atomic commit"")
> 
> For information about bisection process see: https://goo.gl/tpsmEJ#bisection
-- 
Jan Kara <jack@xxxxxxxx>
SUSE Labs, CR