BGPD: Prevent IPv6-LL being fwd to IBGP peers not on same link.

Prevent IPv6 Link-local address being forward to IBGP peer,
which are not directly connected.

R1----IPV6-unnumbered-EBGP-------R2-----IPV6-IBGP-----R3

Configure route-map to set preferred global address on and apply
route-map-IN on R2 for R1-R2 session. Now check on R3's BGP and
RIB table has route nexthop as R1 link-local address, which is
not correct.

As of now we clear link-local address info from mp_nexthop_global,
only if mp_nexthop_global is populated with link-local address.

We should do it even if route-map is configured boz forwarding
link-local address from one link scope to another is violation of
the standards.

Signed-off-by: Biswajit Sadhu sadhub@vmware.com
This commit is contained in:
bisdhdh 2019-11-05 10:59:58 +05:30
parent ba3afa4c2c
commit 63696f1d89

View File

@ -1941,19 +1941,19 @@ int subgroup_announce_check(struct bgp_node *rn, struct bgp_path_info *pi,
__func__, family2str(family));
subgroup_announce_reset_nhop(family, attr);
}
}
/* If IPv6/MP and nexthop does not have any override and happens
* to
* be a link-local address, reset it so that we don't pass along
* the
* source's link-local IPv6 address to recipients who may not be
* on
* the same interface.
*/
if (p->family == AF_INET6 || peer_cap_enhe(peer, afi, safi)) {
if (IN6_IS_ADDR_LINKLOCAL(&attr->mp_nexthop_global))
subgroup_announce_reset_nhop(AF_INET6, attr);
}
/* If IPv6/MP and nexthop does not have any override and happens
* to
* be a link-local address, reset it so that we don't pass along
* the
* source's link-local IPv6 address to recipients who may not be
* on
* the same interface.
*/
if (p->family == AF_INET6 || peer_cap_enhe(peer, afi, safi)) {
if (IN6_IS_ADDR_LINKLOCAL(&attr->mp_nexthop_global))
subgroup_announce_reset_nhop(AF_INET6, attr);
}
return 1;