netfilter: ecache: always set events bits, filter them later
authorPablo Neira Ayuso <pablo@netfilter.org>
Tue, 1 Feb 2011 15:06:30 +0000 (16:06 +0100)
committerPatrick McHardy <kaber@trash.net>
Tue, 1 Feb 2011 15:06:30 +0000 (16:06 +0100)
For the following rule:

iptables -I PREROUTING -t raw -j CT --ctevents assured

The event delivered looks like the following:

 [UPDATE] tcp      6 src=192.168.0.2 dst=192.168.1.2 sport=37041 dport=80 src=192.168.1.2 dst=192.168.1.100 sport=80 dport=37041 [ASSURED]

Note that the TCP protocol state is not included. For that reason
the CT event filtering is not very useful for conntrackd.

To resolve this issue, instead of conditionally setting the CT events
bits based on the ctmask, we always set them and perform the filtering
in the late stage, just before the delivery.

Thus, the event delivered looks like the following:

 [UPDATE] tcp      6 432000 ESTABLISHED src=192.168.0.2 dst=192.168.1.2 sport=37041 dport=80 src=192.168.1.2 dst=192.168.1.100 sport=80 dport=37041 [ASSURED]

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Patrick McHardy <kaber@trash.net>
include/net/netfilter/nf_conntrack_ecache.h
net/netfilter/nf_conntrack_ecache.c

index 96ba5f7dcab60ad26618fa028e5968ecb83310e8..349cefedc9f3c869627211b95bbc57e5ecfde341 100644 (file)
@@ -77,9 +77,6 @@ nf_conntrack_event_cache(enum ip_conntrack_events event, struct nf_conn *ct)
        if (e == NULL)
                return;
 
-       if (!(e->ctmask & (1 << event)))
-               return;
-
        set_bit(event, &e->cache);
 }
 
index 5702de35e2bb327ea0e9bd346f57bf8ebf36c951..63a1b915a7e40e4871f14dad9170b542aaf9a75a 100644 (file)
@@ -63,6 +63,9 @@ void nf_ct_deliver_cached_events(struct nf_conn *ct)
                 * this does not harm and it happens very rarely. */
                unsigned long missed = e->missed;
 
+               if (!((events | missed) & e->ctmask))
+                       goto out_unlock;
+
                ret = notify->fcn(events | missed, &item);
                if (unlikely(ret < 0 || missed)) {
                        spin_lock_bh(&ct->lock);