Merge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf

Pablo Neira Ayuso says:

====================
Netfilter/IPVS/OVS fixes for net

The following patchset contains a rather large batch of Netfilter, IPVS
and OVS fixes for your net tree. This includes fixes for ctnetlink, the
userspace conntrack helper infrastructure, conntrack OVS support,
ebtables DNAT target, several leaks in error path among other. More
specifically, they are:

1) Fix reference count leak in the CT target error path, from Gao Feng.

2) Remove conntrack entry clashing with a matching expectation, patch
   from Jarno Rajahalme.

3) Fix bogus EEXIST when registering two different userspace helpers,
   from Liping Zhang.

4) Don't leak dummy elements in the new bitmap set type in nf_tables,
   from Liping Zhang.

5) Get rid of module autoload from conntrack update path in ctnetlink,
   we don't need autoload at this late stage and it is happening with
   rcu read lock held which is not good. From Liping Zhang.

6) Fix deadlock due to double-acquire of the expect_lock from conntrack
   update path, this fixes a bug that was introduced when the central
   spinlock got removed. Again from Liping Zhang.

7) Safe ct->status update from ctnetlink path, from Liping. The expect_lock
   protection that was selected when the central spinlock was removed was
   not really protecting anything at all.

8) Protect sequence adjustment under ct->lock.

9) Missing socket match with IPv6, from Peter Tirsek.

10) Adjust skb->pkt_type of DNAT'ed frames from ebtables, from
    Linus Luessing.

11) Don't give up on evaluating the expression on new entries added via
    dynset expression in nf_tables, from Liping Zhang.

12) Use skb_checksum() when mangling icmpv6 in IPv6 NAT as this deals
    with non-linear skbuffs.

13) Don't allow IPv6 service in IPVS if no IPv6 support is available,
    from Paolo Abeni.

14) Missing mutex release in error path of xt_find_table_lock(), from
    Dan Carpenter.

15) Update maintainers files, Netfilter section. Add Florian to the
    file, refer to nftables.org and change project status from Supported
    to Maintained.

16) Bail out on mismatching extensions in element updates in nf_tables.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 29 insertions, 1 deletions

diff --git a/net/openvswitch/conntrack.c b/net/openvswitch/conntrack.c
index 42a9591..bf602e3 100644
--- a/net/openvswitch/conntrack.c
+++ b/net/openvswitch/conntrack.c
@@ -516,10 +516,38 @@ ovs_ct_expect_find(struct net *net, const struct nf_conntrack_zone *zone,
 		   u16 proto, const struct sk_buff *skb)
 {
 	struct nf_conntrack_tuple tuple;
+	struct nf_conntrack_expect *exp;
 
 	if (!nf_ct_get_tuplepr(skb, skb_network_offset(skb), proto, net, &tuple))
 		return NULL;
-	return __nf_ct_expect_find(net, zone, &tuple);
+
+	exp = __nf_ct_expect_find(net, zone, &tuple);
+	if (exp) {
+		struct nf_conntrack_tuple_hash *h;
+
+		/* Delete existing conntrack entry, if it clashes with the
+		 * expectation.  This can happen since conntrack ALGs do not
+		 * check for clashes between (new) expectations and existing
+		 * conntrack entries.  nf_conntrack_in() will check the
+		 * expectations only if a conntrack entry can not be found,
+		 * which can lead to OVS finding the expectation (here) in the
+		 * init direction, but which will not be removed by the
+		 * nf_conntrack_in() call, if a matching conntrack entry is
+		 * found instead.  In this case all init direction packets
+		 * would be reported as new related packets, while reply
+		 * direction packets would be reported as un-related
+		 * established packets.
+		 */
+		h = nf_conntrack_find_get(net, zone, &tuple);
+		if (h) {
+			struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
+
+			nf_ct_delete(ct, 0, 0);
+			nf_conntrack_put(&ct->ct_general);
+		}
+	}
+
+	return exp;
 }
 
 /* This replicates logic from nf_conntrack_core.c that is not exported. */