linux-kernel - Re: possible deadlock in ovl_copy_up

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <CAOQ4uxhw5ewvH6_x3oSU67Z+zLB28BUWijq7kM_h1+Jt66T1LA@mail.gmail.com>
Date:   Thu, 18 Oct 2018 13:40:14 +0300
From:   Amir Goldstein <amir73il@...il.com>
To:     Miklos Szeredi <miklos@...redi.hu>
Cc:     syzbot+3ef5c0d1a5cb0b21e6be@...kaller.appspotmail.com,
        linux-kernel <linux-kernel@...r.kernel.org>,
        overlayfs <linux-unionfs@...r.kernel.org>,
        syzkaller-bugs@...glegroups.com
Subject: Re: possible deadlock in ovl_copy_up_start

On Thu, Oct 18, 2018 at 11:41 AM Miklos Szeredi <miklos@...redi.hu> wrote:
>
> On Thu, Oct 18, 2018 at 8:26 AM, Amir Goldstein <amir73il@...il.com> wrote:
>
> > Can someone tell me what the expected behavior of a nested
> > mutex_lock_interruptible(&lock); ?
> >
> > Why does the reproducer only warn and not really deadlock.
> > It is because that is considered the lesser evil?
> > and obviously, then inner unlock releases the outer lock?
>
> No, it's not the same lock, just the same lock class (first one is
> OVL_I(d_inode(old))->lock, the other is
> OVL_I(d_inode(new->d_parent)))->lock).

Doh! of course.

>
> So we could possibly get away with annotating with
> mutex_lock_nested().  Is this the only place that ovl_i_lock is
> nested?
>

As far as I can see it is.
But how would we annotate it for consistent and clear locking order?
NLINK/COPYUP?
if we want this annotation to maintain locking order we need to patch
I posted (does copyup of new->parent prior to nlink_start), so then
we don't need nested annotations anymore.

Nah, I don't think we need to add nested annotations.

FYI, I am working on a small cleanup series for ovl_nlink_start/end
and ovl_inode_lock/unlock, but it is independent of the fix patch I posted
for this bug.

Thanks,
Amir.