The rationale for locking mac learning table entires wrt. gratuitous
ARP packets and bond interfaces was too cryptic for me to understand.
After reading vswitchd/INTERNALS the issue is understandable, but we
can still improve the comment to prevent such confusion in future.
Signed-off-by: Jarno Rajahalme <ja...@ovn.org>
---
ofproto/ofproto-dpif-xlate.c | 9 +++++++--
1 file changed, 7 insertions(+), 2 deletions(-)
diff --git a/ofproto/ofproto-dpif-xlate.c b/ofproto/ofproto-dpif-xlate.c
index 1e2957c..74e3387 100644
--- a/ofproto/ofproto-dpif-xlate.c
+++ b/ofproto/ofproto-dpif-xlate.c
@@ -2077,8 +2077,13 @@ OVS_REQ_RDLOCK(ml->rwlock)
}
if (is_gratuitous_arp(flow, wc)) {
- /* We don't want to learn from gratuitous ARP packets that are
- * reflected back over bond slaves so we lock the learning table. */
+ /* Gratuitous ARP packets received over non-bond interfaces could be
+ * reflected back over bond slaves. We don't want to learn from these
+ * reflected packets, so we lock each entry for which a gratuitous ARP
+ * packet was received over a non-bond interface and refrain from
+ * learning from gratuitous ARP packets that arrive over bond
+ * interfaces for this entry while the lock is in effect. See
+ * vswitchd/INTERNALS for more in-depth discussion on this topic. */
if (!in_xbundle->bond) {
return true;
} else if (mac_entry_is_grat_arp_locked(mac)) {
--
2.1.4
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
