Re: [syzbot] [exfat?] kernel BUG in folio_set_bh

[Luis Chamberlain <mcgrof@kernel.org>]

On Thu, Jul 03, 2025 at 01:07:02AM -0700, syzbot wrote:
> syzbot has bisected this issue to:
> 
> commit 47dd67532303803a87f43195e088b3b4bcf0454d
> Author: Luis Chamberlain <mcgrof@kernel.org>
> Date:   Fri Feb 21 22:38:22 2025 +0000
> 
>     block/bdev: lift block size restrictions to 64k
> 
> bisection log:  https://syzkaller.appspot.com/x/bisect.txt?x=15ec33d4580000
> start commit:   50c8770a42fa Add linux-next specific files for 20250702
> git tree:       linux-next
> final oops:     https://syzkaller.appspot.com/x/report.txt?x=17ec33d4580000
> console output: https://syzkaller.appspot.com/x/log.txt?x=13ec33d4580000
> kernel config:  https://syzkaller.appspot.com/x/.config?x=d831c9dfe03f77ec
> dashboard link: https://syzkaller.appspot.com/bug?extid=f4f84b57a01d6b8364ad
> syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=15c93770580000
> C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=1001aebc580000
> 
> Reported-by: syzbot+f4f84b57a01d6b8364ad@syzkaller.appspotmail.com
> Fixes: 47dd67532303 ("block/bdev: lift block size restrictions to 64k")
> 
> For information about bisection process see: https://goo.gl/tpsmEJ#bisection

Odd, I can't see where the null pointer comes from.

bdev_getblk() --> __getblk_slow() properly returns NULL and doesn't use
the data. But neither does fat_fill_super() on failure. My only
suspicion was on fat_msg() but that sb usage seems fine and the goto out_fail
seems fine as iput() also doesn't process null inodes and unload_nls()
is fine. The return value is also set to -EIO correctly so we don't return NULL
actually. I jus tdon't see anything odd on _fat_msg() either.

Hrm..

  Luis