From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Dumazet Subject: [PATCH] netfilter: Use hlist_add_head_rcu() in nf_conntrack_set_hashsize() Date: Tue, 24 Mar 2009 20:54:53 +0100 Message-ID: <49C93A8D.8000603@cosmosbay.com> References: <49C77D71.8090709@trash.net> <49C780AD.70704@trash.net> <49C7CB9B.1040409@trash.net> <49C8A415.1090606@cosmosbay.com> <49C8CCF4.5050104@cosmosbay.com> <1237907850.12351.80.camel@sakura.staff.proxad.net> <49C8FBCA.40402@cosmosbay.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: mbizon@freebox.fr, "Paul E. McKenney" , Joakim Tjernlund , avorontsov@ru.mvista.com, netdev@vger.kernel.org To: Patrick McHardy Return-path: Received: from gw1.cosmosbay.com ([212.99.114.194]:50662 "EHLO gw1.cosmosbay.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752331AbZCXTzK convert rfc822-to-8bit (ORCPT ); Tue, 24 Mar 2009 15:55:10 -0400 In-Reply-To: <49C8FBCA.40402@cosmosbay.com> Sender: netdev-owner@vger.kernel.org List-ID: Eric Dumazet a =E9crit : >=20 > We are working on a SLAB_DESTROY_BY_RCU implementation so that > conntrack wont use call_rcu() anymore, give us a couple of days :) >=20 While working on this stuff, I found one suspect use of hlist_add_head(= ) Its not a hot path, I believe following patch would make sure nothing wrong happens. If a chain contains element A and B, then we might build a new table with a new chain containing B and A (in this reverse order), and a cpu could see A->next =3D B (new pointer), B->next =3D A (old pointe= r) Thanks [PATCH] netfilter: Use hlist_add_head_rcu() in nf_conntrack_set_hashsiz= e() Using hlist_add_head() in nf_conntrack_set_hashsize() is quite dangerou= s. Without any barrier, one CPU could see a loop while doing its lookup. Its true new table cannot be seen by another cpu, but previous table is= still readable. Signed-off-by: Eric Dumazet diff --git a/net/netfilter/nf_conntrack_core.c b/net/netfilter/nf_connt= rack_core.c index 55befe5..54e983f 100644 --- a/net/netfilter/nf_conntrack_core.c +++ b/net/netfilter/nf_conntrack_core.c @@ -1121,7 +1121,7 @@ int nf_conntrack_set_hashsize(const char *val, st= ruct kernel_param *kp) struct nf_conntrack_tuple_hash, hnode); hlist_del_rcu(&h->hnode); bucket =3D __hash_conntrack(&h->tuple, hashsize, rnd); - hlist_add_head(&h->hnode, &hash[bucket]); + hlist_add_head_rcu(&h->hnode, &hash[bucket]); } } old_size =3D nf_conntrack_htable_size;