Re: KASAN: use-after-free Read in btrfs_scan_one_device

[Rustam Kovhaev]

On Thu, Oct 01, 2020 at 03:35:46PM +0200, David Sterba wrote:
> On Thu, Oct 01, 2020 at 03:08:34PM +0200, Dmitry Vyukov wrote:
> > On Thu, Oct 1, 2020 at 3:05 PM Dmitry Vyukov <dvyukov@xxxxxxxxxx> wrote:
> > >
> > > On Wed, Sep 30, 2020 at 8:06 PM David Sterba <dsterba@xxxxxxx> wrote:
> > > >
> > > > On Wed, Sep 30, 2020 at 06:57:56PM +0200, David Sterba wrote:
> > > > > On Sun, Sep 20, 2020 at 07:12:14AM -0700, syzbot wrote:
> > > > > > Hello,
> > > > > >
> > > > > > syzbot found the following issue on:
> > > > > >
> > > > > > HEAD commit:    eb5f95f1 Merge tag 's390-5.9-6' of git://git.kernel.org/pu..
> > > > > > git tree:       upstream
> > > > > > console output: https://syzkaller.appspot.com/x/log.txt?x=10a0a8bb900000
> > > > > > kernel config:  https://syzkaller.appspot.com/x/.config?x=ffe85b197a57c180
> > > > > > dashboard link: https://syzkaller.appspot.com/bug?extid=582e66e5edf36a22c7b0
> > > > > > compiler:       gcc (GCC) 10.1.0-syz 20200507
> > > > > >
> > > > > > Unfortunately, I don't have any reproducer for this issue yet.
> > > > > >
> > > > > > IMPORTANT: if you fix the issue, please add the following tag to the commit:
> > > > > > Reported-by: syzbot+582e66e5edf36a22c7b0@xxxxxxxxxxxxxxxxxxxxxxxxx
> > > > >
> > > > > #syz fix: btrfs: fix overflow when copying corrupt csums for a message
> > > >
> > > > Johannes spotted that this is not the right fix for this report, I don't
> > > > know how to tell syzbot to revert the 'fix:' command, there isn't
> > > > 'unfix' (like there's 'undup').
> > >
> > > Hi David,
> > >
> > > I've added "unfix" command:
> > > https://github.com/google/syzkaller/pull/2156
> > >
> > > Let's give it a try:
> > > #syz unfix
> > >
> > > Thanks
> > 
> > Voilà ! Unfixed:
> > https://syzkaller.appspot.com/bug?extid=582e66e5edf36a22c7b0
> 
> Thanks!

the problem is that btrfs_kill_super() frees *fs_info while it is still
being referenced by btrfs_scan_one_device() on behalf of another
concurrent mount syscall
a very simple and dumb fix is to remove that printk that references
*fs_info:
https://syzkaller.appspot.com/text?tag=Patch&x=123537fb900000
but instead, i think proper synchronization is needed here
any advice or pointers would be highly appreciated
tyvm!