netfilter: restart search if moved to other chain
authorFlorian Westphal <fw@strlen.de>
Thu, 25 Aug 2016 13:33:29 +0000 (15:33 +0200)
committerBen Hutchings <ben@decadent.org.uk>
Thu, 23 Feb 2017 03:50:48 +0000 (03:50 +0000)
commit 95a8d19f28e6b29377a880c6264391a62e07fccc upstream.

In case nf_conntrack_tuple_taken did not find a conflicting entry
check that all entries in this hash slot were tested and restart
in case an entry was moved to another chain.

Reported-by: Eric Dumazet <edumazet@google.com>
Fixes: ea781f197d6a ("netfilter: nf_conntrack: use SLAB_DESTROY_BY_RCU and get rid of call_rcu()")
Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
[bwh: Backported to 3.2:
 - Adjust context
 - Use NF_CT_STAT_INC(), not the _ATOMIC variant, since we disable BHs]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
net/netfilter/nf_conntrack_core.c

index c855673..69ce23f 100644 (file)
@@ -550,6 +550,7 @@ nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple,
         * least once for the stats anyway.
         */
        rcu_read_lock_bh();
+ begin:
        hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], hnnode) {
                ct = nf_ct_tuplehash_to_ctrack(h);
                if (ct != ignored_conntrack &&
@@ -561,6 +562,12 @@ nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple,
                }
                NF_CT_STAT_INC(net, searched);
        }
+
+       if (get_nulls_value(n) != hash) {
+               NF_CT_STAT_INC(net, search_restart);
+               goto begin;
+       }
+
        rcu_read_unlock_bh();
 
        return 0;