On Mon, May 21, 2018 at 2:50 PM, Jan Kara [off-list ref] wrote:

On Sun 20-05-18 23:07:02, syzbot wrote:

quoted

syzbot has found a reproducer for the following crash on:

HEAD commit:    771c577c23ba Linux 4.17-rc6
git tree:       upstream
console output: https://syzkaller.appspot.com/x/log.txt?x=108c9f37800000
kernel config:  https://syzkaller.appspot.com/x/.config?x=982e2df1b9e60b02
dashboard link: https://syzkaller.appspot.com/bug?extid=9a44753ac327c557796a
compiler:       gcc (GCC) 8.0.1 20180413 (experimental)
syzkaller repro:https://syzkaller.appspot.com/x/repro.syz?x=1655584f800000
C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=15134357800000

IMPORTANT: if you fix the bug, please add the following tag to the commit:
Reported-by: syzbot+9a44753ac327c557796a@syzkaller.appspotmail.com

WARNING: CPU: 0 PID: 4577 at fs/ext4/inode.c:3894
ext4_set_page_dirty+0x28d/0x330 fs/ext4/inode.c:3894
Kernel panic - not syncing: panic_on_warn set ...

Looking at the reproducer, this is the issue we have been speaking about at
LSF/MM - direct IO read dirties pages after writeback has written them out
and so we don't expect them to be dirtied. If someone told me I would not
believe this is possible to hit by fuzzing ;) Good job from the syzkaller
guys!

Thanks, Jan! Very nice to hear.
I believe we are not yet at the limit of what's possible :)