From: Jiri Olsa <jolsa@redhat.com>
Date: Fri, 17 Sep 2010 16:40:53 -0400
Subject: [net] ipv4/defrag: check socket type before reference
Message-id: <1284741653-14539-1-git-send-email-jolsa@redhat.com>
Patchwork-id: 28290
O-Subject: [PATCH RHEL5] BZ 633450 ipv4 - nodefrag: check proper socket type
	before reference
Bugzilla: 632266
RH-Acked-by: Jarod Wilson <jarod@redhat.com>
RH-Acked-by: Neil Horman <nhorman@redhat.com>
RH-Acked-by: David S. Miller <davem@redhat.com>

Bugzilla: 632266
[RHEL5] connectathon test hangs from KVM guests with virtio network driver
https://bugzilla.redhat.com/show_bug.cgi?id=632266

Description:
============
Need to check proper socket type within ip_conntrack_defrag function
before referencing the nodefrag flag.
The tun driver receive path produces skbs with AF_UNSPEC socket type,
causing unwanted fragmented packets going out.

Brew:
=====
https://brewweb.devel.redhat.com/taskinfo?taskID=2768184

Tested:
=======
The connectathon tests all finish correctly now.

Signed-off-by: Jarod Wilson <jarod@redhat.com>

diff --git a/net/ipv4/netfilter/ip_conntrack_standalone.c b/net/ipv4/netfilter/ip_conntrack_standalone.c
index 453d2aa..60e14cf 100644
--- a/net/ipv4/netfilter/ip_conntrack_standalone.c
+++ b/net/ipv4/netfilter/ip_conntrack_standalone.c
@@ -451,9 +451,11 @@ static unsigned int ip_conntrack_defrag(unsigned int hooknum,
 				        const struct net_device *out,
 				        int (*okfn)(struct sk_buff *))
 {
+	struct sock *sk = (*pskb)->sk;
 	struct inet_sock *inet = inet_sk((*pskb)->sk);
 
-	if (inet && inet->nodefrag)
+	if (sk && (sk->sk_family == PF_INET) &&
+	    inet->nodefrag)
 		return NF_ACCEPT;
 
 #if !defined(CONFIG_IP_NF_NAT) && !defined(CONFIG_IP_NF_NAT_MODULE)