netdev - [PATCH net-next] net: ipv6: Fix refcnt on host routes

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-Id: <1456947007-32444-1-git-send-email-dsa@cumulusnetworks.com>
Date:	Wed,  2 Mar 2016 11:30:07 -0800
From:	David Ahern <dsa@...ulusnetworks.com>
To:	netdev@...r.kernel.org
Cc:	David Ahern <dsa@...ulusnetworks.com>,
	Andrey Wagin <avagin@...il.com>,
	Ying Huang <ying.huang@...ux.intel.com>
Subject: [PATCH net-next] net: ipv6: Fix refcnt on host routes

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>
---
 net/ipv6/addrconf.c | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/net/ipv6/addrconf.c b/net/ipv6/addrconf.c
index b1169d511ab4..8c0dab2de5c9 100644
--- a/net/ipv6/addrconf.c
+++ b/net/ipv6/addrconf.c
@@ -5319,12 +5319,10 @@ static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
 			if (rt)
 				ip6_del_rt(rt);
 		}
-		if (ifp->rt) {
-			dst_hold(&ifp->rt->dst);
+		dst_hold(&ifp->rt->dst);
+
+		ip6_del_rt(ifp->rt);
 
-			ip6_del_rt(ifp->rt);
-			ifp->rt = NULL;
-		}
 		rt_genid_bump_ipv6(net);
 		break;
 	}
-- 
2.1.4