diff options
| -rw-r--r-- | bgpd/bgp_route.c | 17 |
1 files changed, 16 insertions, 1 deletions
diff --git a/bgpd/bgp_route.c b/bgpd/bgp_route.c index 94c21e1861..e1cc62ef00 100644 --- a/bgpd/bgp_route.c +++ b/bgpd/bgp_route.c @@ -4638,7 +4638,22 @@ void bgp_update(struct peer *peer, const struct prefix *p, uint32_t addpath_id, if (aspath_get_last_as(attr->aspath) == bgp->as) do_loop_check = 0; - if (CHECK_FLAG(peer->af_flags[afi][safi], PEER_FLAG_REFLECTOR_CLIENT)) + /* When using bgp ipv4 labeled session, the local prefix is + * received by a peer, and finds out that the proposed prefix + * and its next-hop are the same. To avoid a route loop locally, + * no nexthop entry is referenced for that prefix, and the route + * will not be selected. + * + * As it has been done for ipv4-unicast, apply the following fix + * for labeled address families: when the received peer is + * a route reflector, the prefix has to be selected, even if the + * route can not be installed locally. + */ + if (CHECK_FLAG(peer->af_flags[afi][safi], PEER_FLAG_REFLECTOR_CLIENT) || + (safi == SAFI_UNICAST && !peer->afc[afi][safi] && + peer->afc[afi][SAFI_LABELED_UNICAST] && + CHECK_FLAG(peer->af_flags[afi][SAFI_LABELED_UNICAST], + PEER_FLAG_REFLECTOR_CLIENT))) bgp_nht_param_prefix = NULL; else bgp_nht_param_prefix = p; |