>
>> + int tag = 0;
>> + ofp_port_t ofport;
>> + if (binding->parent_port) {
>> + ofport = u16_to_ofp(simap_get(&lport_to_ofport,
>> + binding->parent_port));
>> + if (ofport) {
>
> I don't think that it is safe to rely on 'tag' being nonnull here. It
> should be, but the database constraints can't ensure it:
>
>> + tag = *binding->tag;
I will fix it.
>> @@ -122,10 +136,20 @@ physical_run(struct controller_ctx *ctx)
>> *
>> * For packets that arrive from a vif: set MFF_LOG_INPORT to the
>> * logical input port, MFF_METADATA to the logical datapath, and
>> - * resubmit into the logical pipeline starting at table 16. */
>> + * resubmit into the logical pipeline starting at table 16.
>> + *
>> + * Containers sitting behind a vif, come with VLAN tags.
>> + * Match on the tags and then strip it before resubmitting to
>> + * the next tables. */
>> match_init_catchall(&match);
>> ofpbuf_clear(&ofpacts);
>> match_set_in_port(&match, ofport);
>> + if (tag) {
>> + match_set_dl_vlan(&match, htons(tag));
>> + } else {
>> + match_set_dl_vlan(&match, htons(OFP10_VLAN_NONE));
>> + }
>
> Above, I would rather avoid matching on the VLAN at all in the case
> where there is no tag. That gives the logical pipeline a chance to use
> the VLAN tag in such a case. (Currently, the logical pipeline just
> discards packets that have VLAN tags, but I'd like to leave room for
> expansion in the future, especially since there seems to be some ongoing
> work on QinQ support.)
If I have 2 flows:
1. inport=10, actions=foo
2. inport=10,dl_vlan=20, actions=foo
Then is it not true that a packet coming from inport=10 with vlan 20
can match either of the above flows?
If the above is correct, I think I should match on a 'no vlan' because
I need to let the mgmt IP of the VM that hosts containers to belong to
a different logical network. (I guess, I can change the priorities of
the vlan matched packets to get the behavior that you desire too.)
>
> Acked-by: Ben Pfaff <b...@nicira.com>
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
