I was doing some testing with the latest ebtables and I think I've found a bug in ebt_deliver_counters that was introduced in the following commit: http://ebtables.cvs.sourceforge.net/viewvc/ebtables/ebtables2/userspace/ebtables2/communication.c?r1=1.40&r2=1.41 It seems that the chainnr++ on line 308 is only reached when entries is NULL, causing the code to repeatedly loop over the rules for the first non-empty chain. This manifests as every chain having its counters copied from the first non-empty chain instead of getting the counters assigned with -c: ebtables -t nat -N CHAIN1 ebtables -t nat -A CHAIN1 -s 0:0:0:0:1:1 -j ACCEPT -c 101 101 ebtables -t nat -A CHAIN1 -s 0:0:0:0:1:2 -j ACCEPT -c 102 102 ebtables -t nat -N CHAIN2 ebtables -t nat -A CHAIN2 -s 0:0:0:0:2:1 -j ACCEPT -c 201 201 ebtables -t nat -A CHAIN2 -s 0:0:0:0:2:2 -j ACCEPT -c 202 202 ebtables -t nat -N CHAIN3 ebtables -t nat -A CHAIN3 -s 0:0:0:0:3:1 -j ACCEPT -c 302 302 ebtables -t nat -A CHAIN3 -s 0:0:0:0:3:2 -j ACCEPT -c 303 303 ebtables -t nat -L --Lc Bridge table: nat Bridge chain: PREROUTING, entries: 0, policy: ACCEPT Bridge chain: OUTPUT, entries: 0, policy: ACCEPT Bridge chain: POSTROUTING, entries: 0, policy: ACCEPT Bridge chain: CHAIN1, entries: 2, policy: ACCEPT -s 0:0:0:0:1:1 -j ACCEPT , pcnt = 101 -- bcnt = 101 -s 0:0:0:0:1:2 -j ACCEPT , pcnt = 102 -- bcnt = 102 Bridge chain: CHAIN2, entries: 2, policy: ACCEPT -s 0:0:0:0:2:1 -j ACCEPT , pcnt = 101 -- bcnt = 101 -s 0:0:0:0:2:2 -j ACCEPT , pcnt = 102 -- bcnt = 102 Bridge chain: CHAIN3, entries: 2, policy: ACCEPT -s 0:0:0:0:3:1 -j ACCEPT , pcnt = 101 -- bcnt = 101 -s 0:0:0:0:3:2 -j ACCEPT , pcnt = 102 -- bcnt = 102 I've attempted to fix the bug, and my patch is included below. I tried to account for all of the edge cases, but I don't have a solid enough understanding of the data structures used to claim that I've been successful. My code is on GitHub as well if that's easier: https://github.com/irgeek/ebtables/commit/29221fea0021795a7005d17288b656bf21519e84 diff --git a/communication.c b/communication.c index 0917f6e..f1a6f08 100644 --- a/communication.c +++ b/communication.c @@ -308,13 +308,14 @@ void ebt_deliver_counters(struct ebt_u_replace *u_repl) old = u_repl->counters; new = newcounters; while (cc != u_repl->cc) { - if (!next || next == entries->entries) { - while (chainnr < u_repl->num_chains && (!(entries = u_repl->chains[chainnr]) || - (next = entries->entries->next) == entries->entries)) - chainnr++; - if (chainnr == u_repl->num_chains) - break; + while (!next || (next == entries->entries && chainnr < u_repl->num_chains)) { + next = NULL; + if ((entries = u_repl->chains[chainnr++])) { + next = entries->entries->next; + } } + if (chainnr >= u_repl->num_chains && (!entries || next == entries->entries)) + break; if (next == NULL) ebt_print_bug("next == NULL"); if (cc->type == CNT_NORM) { James Sinclair Linode, LLC -- To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
