ospf6d: Fix ECMP inter-area route nexthop update

Fixes #9720. When updating an ECMP inter-area route, we compute
a new route and check whether that already exists. If so, we keep the old
route and only update its nexthops. Previously, we merged the new route's
nexthops into the old one's, but this way, it's impossible to remove
nexthops from the old route, resulting in stale nexthops.

This commit fixes this by first removing all nexthops from the old route and
then copying all nexthops from the new route into it. If the new route has
fewer nexthops, the old one will have as well afterwards.

Signed-off-by: Martin Buck <mb-tmp-tvguho.pbz@gromit.dyndns.org>

diff --git a/ospf6d/ospf6_abr.c b/ospf6d/ospf6_abr.c
index 443759344e963b9fc5376146fe4e9f2847c30771..5af1139d9b484fb56b243e2873c2c8778e700282 100644 (file)
--- a/ospf6d/ospf6_abr.c
+++ b/ospf6d/ospf6_abr.c
@@ -1276,7 +1276,8 @@ void ospf6_abr_examin_summary(struct ospf6_lsa *lsa, struct ospf6_area *oa)
                        continue;
                }
 
-               ospf6_route_merge_nexthops(old_route, route);
+               list_delete_all_node(old_route->nh_list);
+               ospf6_route_copy_nexthops(old_route, route);
                old_entry_updated = true;
 
                for (ALL_LIST_ELEMENTS_RO(old_route->paths, anode,