Re: possible deadlock in ipv6_sock_mc_close

[Paolo Abeni <pabeni@redhat.com>]

Hello,

On Mon, 2021-03-01 at 14:52 +0000, Chuck Lever wrote:
> > On Mar 1, 2021, at 8:49 AM, syzbot <syzbot+e2fa57709a385e6db10f@syzkaller.appspotmail.com> wrote:
> > 
> > Hello,
> > 
> > syzbot found the following issue on:
> > 
> > HEAD commit:    eee7ede6 Merge branch 'bnxt_en-error-recovery-bug-fixes'
> > git tree:       net
> > console output: https://syzkaller.appspot.com/x/log.txt?x=123ad632d00000
> > kernel config:  https://syzkaller.appspot.com/x/.config?x=e2d5ba72abae4f14
> > dashboard link: https://syzkaller.appspot.com/bug?extid=e2fa57709a385e6db10f
> > syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=109d89b6d00000
> > C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=12e9e0dad00000
> > 
> > The issue was bisected to:
> > 
> > commit c8e88e3aa73889421461f878cd569ef84f231ceb
> > Author: Chuck Lever <chuck.lever@oracle.com>
> > Date:   Tue Nov 3 20:06:04 2020 +0000
> > 
> >    NFSD: Replace READ* macros in nfsd4_decode_layoutget()
> > 
> > bisection log:  https://syzkaller.appspot.com/x/bisect.txt?x=13bef9ccd00000
> > final oops:     https://syzkaller.appspot.com/x/report.txt?x=107ef9ccd00000
> > console output: https://syzkaller.appspot.com/x/log.txt?x=17bef9ccd00000
> > 
> > IMPORTANT: if you fix the issue, please add the following tag to the commit:
> > Reported-by: syzbot+e2fa57709a385e6db10f@syzkaller.appspotmail.com
> > Fixes: c8e88e3aa738 ("NFSD: Replace READ* macros in nfsd4_decode_layoutget()")
> > 
> > ======================================================
> > WARNING: possible circular locking dependency detected
> > 5.11.0-syzkaller #0 Not tainted
> > ------------------------------------------------------
> > syz-executor905/8822 is trying to acquire lock:
> > ffffffff8d678fe8 (rtnl_mutex){+.+.}-{3:3}, at: ipv6_sock_mc_close+0xd7/0x110 net/ipv6/mcast.c:323
> > 
> > but task is already holding lock:
> > ffff888024390120 (sk_lock-AF_INET6){+.+.}-{0:0}, at: lock_sock include/net/sock.h:1600 [inline]
> > ffff888024390120 (sk_lock-AF_INET6){+.+.}-{0:0}, at: mptcp6_release+0x57/0x130 net/mptcp/protocol.c:3507
> > 
> > which lock already depends on the new lock.
> 
> Hi, thanks for the report.
> 
> Initial analysis:
> 
> c8e88e3aa738 ("NFSD: Replace READ* macros in nfsd4_decode_layoutget()"
> changes code several layers above the network layer. In addition,
> neither of the stack traces contain NFSD functions. And, repro.c does
> not appear to exercise any filesystem code.
> 
> Therefore the bisect result looks implausible to me. I don't see any
> obvious connection between the lockdep splat and c8e88e3aa738. (If
> someone else does, please let me know where to look).

I agree the bisect result is unexpected.

This looks really as an MPTCP-specific issue, likely introduced by:

32fcc880e0a9 ("mptcp: provide subflow aware release function")

and should be fixed inside MPTCP.

Cheers,

Paolo