From mboxrd@z Thu Jan  1 00:00:00 1970
From: Eric Dumazet <eric.dumazet@gmail.com>
Subject: Re: recvmsg sleeping from invalid context
Date: Sat, 14 Jan 2012 23:16:02 +0100
Message-ID: <1326579362.5287.45.camel@edumazet-laptop>
References: <20120113182401.GA25885@redhat.com>
	 <20120113.102712.600185688659703015.davem@davemloft.net>
	 <20120114214334.GA28423@redhat.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: QUOTED-PRINTABLE
Cc: David Miller <davem@davemloft.net>, netdev@vger.kernel.org,
	linux-kernel@vger.kernel.org, glommer@parallels.com
To: Dave Jones <davej@redhat.com>
Return-path: <linux-kernel-owner@vger.kernel.org>
In-Reply-To: <20120114214334.GA28423@redhat.com>
Sender: linux-kernel-owner@vger.kernel.org
List-Id: netdev.vger.kernel.org

Le samedi 14 janvier 2012 =C3=A0 16:43 -0500, Dave Jones a =C3=A9crit :
> On Fri, Jan 13, 2012 at 10:27:12AM -0800, David Miller wrote:
>  > From: Dave Jones <davej@redhat.com>
>  > Date: Fri, 13 Jan 2012 13:24:01 -0500
>  >=20
>  > > getting a ton of these on the latest head (099469502f62fbe0d7e4f=
0b83a2f22538367f734)
>  > >=20
>  > > BUG: sleeping function called from invalid context at mm/memory.=
c:3905
>  > > in_atomic(): 0, irqs_disabled(): 0, pid: 1067, name: NetworkMana=
ger
>  > > INFO: lockdep is turned off.
>  > > Pid: 1067, comm: NetworkManager Not tainted 3.2.0+ #22
>  > > Call Trace:
>  > >  [<ffffffff81099415>] __might_sleep+0x145/0x200
>  > >  [<ffffffff811752a4>] might_fault+0x34/0xb0
>  > >  [<ffffffff81551555>] ? sock_def_readable+0x25/0x1a0
>  > >  [<ffffffff8155c387>] put_cmsg+0x77/0x120
>  > >  [<ffffffff8159379c>] netlink_recvmsg+0x35c/0x480
>  > >  [<ffffffff8155201a>] ? sock_update_classid+0x9a/0x260
>  > >  [<ffffffff81552052>] ? sock_update_classid+0xd2/0x260
>  > >  [<ffffffff81549fbd>] sock_recvmsg+0x11d/0x140
>  > >  [<ffffffff811752c3>] ? might_fault+0x53/0xb0
>  > >  [<ffffffff8117530c>] ? might_fault+0x9c/0xb0
>  > >  [<ffffffff811752c3>] ? might_fault+0x53/0xb0
>  > >  [<ffffffff8154b1b3>] __sys_recvmsg+0x153/0x2d0
>  > >  [<ffffffff811bc39a>] ? fget_light+0x5a/0x470
>  > >  [<ffffffff8109fd11>] ? get_parent_ip+0x11/0x50
>  > >  [<ffffffff816ac23d>] ? sub_preempt_count+0x9d/0xd0
>  > >  [<ffffffff811bc43b>] ? fget_light+0xfb/0x470
>  > >  [<ffffffff811bc39a>] ? fget_light+0x5a/0x470
>  > >  [<ffffffff8154e2e9>] sys_recvmsg+0x49/0x90
>  > >  [<ffffffff816b00e9>] system_call_fastpath+0x16/0x1b
>  >=20
>  > Sigh, I suspect the new socket memcg code, which I didn't want to
>  > even apply in the first place. :-/
>  >=20
>  > Glauber, please fix this.
>=20
> How new is 'new' ?
>=20
> interestingly, I now started getting these in 3.1.9 where I never not=
iced them before.
>=20
>=20
> BUG: sleeping function called from invalid context at mm/memory.c:390=
5
> in_atomic(): 1, irqs_disabled(): 0, pid: 962, name: NetworkManager
> 1 lock held by NetworkManager/962:
>  #0:  (rcu_read_lock){.+.+..}, at: [<ffffffff815f610d>] inet6_dump_fi=
b+0x3d/0x3d0
> Pid: 962, comm: NetworkManager Not tainted 3.1.9-1.fc16.x86_64.debug =
#1
> Call Trace:
>  [<ffffffff8105fbd0>] __might_sleep+0xf0/0x120
>  [<ffffffff81160f68>] might_fault+0x38/0xb0
>  [<ffffffff81519070>] ? sock_def_error_report+0x120/0x120
>  [<ffffffff81523877>] put_cmsg+0x77/0x120
>  [<ffffffff815587ec>] netlink_recvmsg+0x35c/0x480
>  [<ffffffff81518d2e>] ? sock_update_classid+0x8e/0x190
>  [<ffffffff81518d68>] ? sock_update_classid+0xc8/0x190
>  [<ffffffff815122ad>] sock_recvmsg+0x11d/0x140
>  [<ffffffff81160f8c>] ? might_fault+0x5c/0xb0
>  [<ffffffff81160fd5>] ? might_fault+0xa5/0xb0
>  [<ffffffff81160f8c>] ? might_fault+0x5c/0xb0
>  [<ffffffff81513373>] __sys_recvmsg+0x153/0x2d0
>  [<ffffffff810a8f28>] ? sched_clock_cpu+0xa8/0x110
>  [<ffffffff810b6bfd>] ? trace_hardirqs_off+0xd/0x10
>  [<ffffffff810a8fff>] ? local_clock+0x6f/0x80
>  [<ffffffff810b74a5>] ? lock_release_holdtime.part.9+0x15/0x1a0
>  [<ffffffff811a4bef>] ? fget_light+0xcf/0x3b0
>  [<ffffffff811a4c07>] ? fget_light+0xe7/0x3b0
>  [<ffffffff811a4b68>] ? fget_light+0x48/0x3b0
>  [<ffffffff81516359>] sys_recvmsg+0x49/0x90
>  [<ffffffff81669dc2>] system_call_fastpath+0x16/0x1b
>=20

I dont understand this trace.

inet6_dump_fib() does indeed an rcu_read_lock(), but pair it by
rcu_read_unlock().

fib6_dump_table() is called with rcu_read_lock(), but certainly cannot
end calling pu_cmsg().