[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <1345784182.5904.1920.camel@edumazet-glaptop>
Date: Fri, 24 Aug 2012 06:56:22 +0200
From: Eric Dumazet <eric.dumazet@...il.com>
To: Gao feng <gaofeng@...fujitsu.com>
Cc: davem@...emloft.net, netdev@...r.kernel.org,
"Eric W. Biederman" <ebiederm@...ssion.com>
Subject: Re: [PATCH net-next v2] net: dev: fix the incorrect hold of net
namespace's lo device
On Fri, 2012-08-24 at 09:36 +0800, Gao feng wrote:
> When moving a net device from one net namespace to another
> net namespace,dev_change_net_namespace calls NETDEV_DOWN
> event,so the original net namespace's dst entries which
> beloned to this net device will be put into dst_garbage
> list.
>
> then dev_change_net_namespace will set this net device's
> net to the new net namespace.
>
> If we unregister this net device's driver, this will trigger
> the NETDEV_UNREGISTER_FINAL event, dst_ifdown will be called,
> and get this net device's dst entries from dst_garbage list,
> put these entries' dev to the new net namespace's lo device.
>
> It's not what we want,actually we need these dst entries hold
> the original net namespace's lo device,this incorrect device
> holding will trigger emg message like below.
> unregister_netdevice: waiting for lo to become free. Usage count = 1
>
> so we should call NETDEV_UNREGISTER_FINAL event in
> dev_change_net_namespace too,in order to make sure dst entries
> already in the dst_garbage list, we need rcu_barrier before we
> call NETDEV_UNREGISTER_FINAL event.
>
> With help form Eric Dumazet.
>
> Signed-off-by: Gao feng <gaofeng@...fujitsu.com>
> Cc: Eric Dumazet <eric.dumazet@...il.com>
> Cc: "Eric W. Biederman" <ebiederm@...ssion.com>
> ---
> net/core/dev.c | 2 ++
> 1 files changed, 2 insertions(+), 0 deletions(-)
>
> diff --git a/net/core/dev.c b/net/core/dev.c
> index bc857fe..6956d4c 100644
> --- a/net/core/dev.c
> +++ b/net/core/dev.c
> @@ -6256,6 +6256,8 @@ int dev_change_net_namespace(struct net_device *dev, struct net *net, const char
> the device is just moving and can keep their slaves up.
> */
> call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
> + rcu_barrier();
> + call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
> rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
>
> /*
Acked-by: Eric Dumazet <edumazet@...gle.com>
Thanks !
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists