Re: [syzbot] Monthly exfat report (Nov 2023)

[Aleksandr Nogikh]

Hi,

Thank you very much for taking a look at this report!

On Wed, Nov 29, 2023 at 3:10 PM Namjae Jeon <linkinjeon@xxxxxxxxxx> wrote:
>
> 2023-11-28 6:03 GMT+09:00, syzbot
> <syzbot+listb51f932ee38ecac1b05d@xxxxxxxxxxxxxxxxxxxxxxxxx>:
> > Hello exfat maintainers/developers,
> Hi,
> >
> > This is a 31-day syzbot report for the exfat subsystem.
> > All related reports/information can be found at:
> > https://syzkaller.appspot.com/upstream/s/exfat
> >
> > During the period, 1 new issues were detected and 0 were fixed.
> > In total, 7 issues are still open and 12 have been fixed so far.
> >
> > Some of the still happening issues:
> >
> > Ref Crashes Repro Title
> > <1> 225     No    INFO: task hung in path_openat (7)
> There is no reproducer and I couldn't find where it is connected to exfat.
> >
> > https://syzkaller.appspot.com/bug?extid=950a0cdaa2fdd14f5bdc

Syzbot assigned exfat because of this report:
https://syzkaller.appspot.com/text?tag=CrashReport&x=10e02711680000

But it indeed does look like a false positive, let's update the subsystem.

#syz set <1> subsystems: fs

> > <2> 150     No    INFO: task hung in exfat_sync_fs
> There is no reproducer... Let me know how to reproduce it.

Syzbot has not yet found a reproducer for it, even though it does not
seem to be very rare (there've been already >150 such crashes).
There's a command to hide the finding from monthly reports if you
believe this is not actionable for you anyway.

> This can happen when disk speed is very low and a lot of writes occur.
> Is that so?
>
> >
> > https://syzkaller.appspot.com/bug?extid=205c2644abdff9d3f9fc
> > <3> 119     Yes   WARNING in drop_nlink (2)
> I'm curious how this issue was classified as an exfat issue.
> When looking at the backtrace, this issue appears to be an hfsplus issue.
>
>  hfsplus_unlink+0x3fe/0x790 fs/hfsplus/dir.c:381
>  hfsplus_rename+0xc8/0x1c0 fs/hfsplus/dir.c:547
>  vfs_rename+0xaba/0xde0 fs/namei.c:4844
> >
> > https://syzkaller.appspot.com/bug?extid=651ca866e5e2b4b5095b

This is triggered by both hfsplus and exfat mounts. See e.g. the crash
at "2023/06/20 03:21":
https://syzkaller.appspot.com/text?tag=CrashReport&x=1248deff280000

> > <4> 38      Yes   INFO: task hung in exfat_write_inode
> I could not reproduce this by using C reproducer provided by you.
> Can you confirm that this really cannot be reproduced using it?
>
> >
> > https://syzkaller.appspot.com/bug?extid=2f73ed585f115e98aee8

I've followed the guide at [1] and the C reproducer crashed the VM in
~420 seconds with exactly the same output as on sykaller.appspot.com.

[1] https://github.com/google/syzkaller/blob/master/docs/syzbot_assets.md#run-a-c-reproducer

> > <5> 1       No    UBSAN: shift-out-of-bounds in exfat_fill_super (2)
> This is false alarm. ->sect_size_bits could not be greater than 12 by
> the check below.
>
>         /*
>          * sect_size_bits could be at least 9 and at most 12.
>          */
>         if (p_boot->sect_size_bits < EXFAT_MIN_SECT_SIZE_BITS ||
>             p_boot->sect_size_bits > EXFAT_MAX_SECT_SIZE_BITS) {
>                 exfat_err(sb, "bogus sector size bits : %u",
>                                 p_boot->sect_size_bits);
>                 return -EINVAL;
>         }
> >
> > https://syzkaller.appspot.com/bug?extid=d33808a177641a02213e

Maybe something corrupts it in between the check and the use?
Syzkaller may generate a lot of situations when mount and other
operations on the same block device happen simultaneously.

There's a patch series[2] that helps prohibit this, but it has not yet
reached all trees we fuzz.

[2] https://lore.kernel.org/all/20231101173542.23597-1-jack@xxxxxxx/

-- 
Aleksandr