[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Thu, 03 Mar 2016 17:26:45 -0500 (EST)
From: David Miller <davem@...emloft.net>
To: dsa@...ulusnetworks.com
Cc: netdev@...r.kernel.org, avagin@...il.com,
ying.huang@...ux.intel.com
Subject: Re: [PATCH net-next] net: ipv6: Fix refcnt on host routes
From: David Ahern <dsa@...ulusnetworks.com>
Date: Wed, 2 Mar 2016 11:30:07 -0800
> Andrew and Ying Huang's test robot both reported usage count problems that
> trace back to the 'keep address on ifdown' patch.
>
> From Andrew:
> We execute CRIU test on linux-next. On the current linux-next kernel
> they hangs on creating a network namespace.
>
> The kernel log contains many massages like this:
> [ 1036.122108] unregister_netdevice: waiting for lo to become free.
> Usage count = 2
> [ 1046.165156] unregister_netdevice: waiting for lo to become free.
> Usage count = 2
> [ 1056.210287] unregister_netdevice: waiting for lo to become free.
> Usage count = 2
>
> I tried to revert this patch and the bug disappeared.
>
> Here is a set of commands to reproduce this bug:
>
> [root@...ux-next-test linux-next]# uname -a
> Linux linux-next-test 4.5.0-rc6-next-20160301+ #3 SMP Wed Mar 2
> 17:32:18 UTC 2016 x86_64 x86_64 x86_64 GNU/Linux
>
> [root@...ux-next-test ~]# unshare -n
> [root@...ux-next-test ~]# ip link set up dev lo
> [root@...ux-next-test ~]# ip a
> 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN
> group default qlen 1
> link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
> inet 127.0.0.1/8 scope host lo
> valid_lft forever preferred_lft forever
> inet6 ::1/128 scope host
> valid_lft forever preferred_lft forever
> [root@...ux-next-test ~]# logout
> [root@...ux-next-test ~]# unshare -n
>
> -----
>
> The problem is a change made to RTM_DELADDR case in __ipv6_ifa_notify that
> was added in an early version of the offending patch and is no longer
> needed.
>
> Fixes: f1705ec197e7 ("net: ipv6: Make address flushing on ifdown optional")
> Cc: Andrey Wagin <avagin@...il.com>
> Cc: Ying Huang <ying.huang@...ux.intel.com>
> Signed-off-by: David Ahern <dsa@...ulusnetworks.com>
Applied, thanks David.
Powered by blists - more mailing lists