From mboxrd@z Thu Jan 1 00:00:00 1970 From: Cong Wang Subject: Re: [Patch net-next v2] netpoll: fix a rtnl lock assertion failure Date: Mon, 14 Jan 2013 18:42:48 +0800 Message-ID: <1358160168.338.12.camel@cr0> References: <1358159054-21951-1-git-send-email-amwang@redhat.com> <20130114103201.GA1633@minipsycho.brq.redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org, "David S. Miller" To: Jiri Pirko Return-path: Received: from mx1.redhat.com ([209.132.183.28]:65216 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755954Ab3ANKm4 (ORCPT ); Mon, 14 Jan 2013 05:42:56 -0500 In-Reply-To: <20130114103201.GA1633@minipsycho.brq.redhat.com> Sender: netdev-owner@vger.kernel.org List-ID: On Mon, 2013-01-14 at 11:32 +0100, Jiri Pirko wrote: > Mon, Jan 14, 2013 at 11:24:14AM CET, amwang@redhat.com wrote: > >From: Cong Wang > > > >This patch fixes the following warning: > > > >[ 72.013864] RTNL: assertion failed at net/core/dev.c (4955) > >[ 72.017758] Pid: 668, comm: netpoll-prep-v6 Not tainted 3.8.0-rc1+ #474 > >[ 72.019582] Call Trace: > >[ 72.020295] [] netdev_master_upper_dev_get+0x35/0x58 > >[ 72.022545] [] netpoll_setup+0x61/0x340 > >[ 72.024846] [] store_enabled+0x82/0xc3 > >[ 72.027466] [] netconsole_target_attr_store+0x35/0x37 > >[ 72.029348] [] configfs_write_file+0xe2/0x10c > >[ 72.030959] [] vfs_write+0xaf/0xf6 > >[ 72.032359] [] ? sysret_check+0x22/0x5d > >[ 72.033824] [] sys_write+0x5c/0x84 > >[ 72.035328] [] system_call_fastpath+0x16/0x1b > > > >Just hold RCU read lock and call netdev_master_upper_dev_get_rcu(), > >as suggested by Jiri. > > > >Cc: Jiri Pirko > >Cc: David S. Miller > >Signed-off-by: Cong Wang > > > >--- > >diff --git a/net/core/netpoll.c b/net/core/netpoll.c > >index 9f05067..6d56b57 100644 > >--- a/net/core/netpoll.c > >+++ b/net/core/netpoll.c > >@@ -1055,11 +1055,15 @@ int netpoll_setup(struct netpoll *np) > > return -ENODEV; > > } > > > >- if (netdev_master_upper_dev_get(ndev)) { > >+ rcu_read_lock(); > >+ /* We permit devices with VLAN attached to */ > > I think that this comment might be misleading. Master upper is not set > not only for VLAN but for macvlan (and in future possibly for others) as well. > OK, I will remove it.