pimd: Only use paths that we have nbrs on

When looking up nexthops for a path, we
should only allow nexthops that we
actually have neighbors formed for.  Otherwise
when we send join/prune messages they will
do nothing.

Ticket:CM-12754
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

diff --git a/pimd/pim_zlookup.c b/pimd/pim_zlookup.c
index f81c6e87812ec74308484dc1b4aab4c82d645352..d269c04254fe23a145efb7108e399073a2f1a500 100644 (file)
--- a/pimd/pim_zlookup.c
+++ b/pimd/pim_zlookup.c
@@ -332,6 +332,7 @@ zclient_lookup_nexthop (struct pim_zlookup_nexthop nexthop_tab[],
   int lookup;
   uint32_t route_metric = 0xFFFFFFFF;
   uint8_t  protocol_distance = 0xFF;
+  int i;
 
   qpim_nexthop_lookups++;
 
@@ -360,9 +361,13 @@ zclient_lookup_nexthop (struct pim_zlookup_nexthop nexthop_tab[],
     }
     
     /*
-      FIXME: Non-recursive nexthop ensured only for first ifindex.
-      However, recursive route lookup should really be fixed in zebra daemon.
-      See also TODO T24.
+     * FIXME: Non-recursive nexthop ensured only for first ifindex.
+     * However, recursive route lookup should really be fixed in zebra daemon.
+     * See also TODO T24.
+     *
+     * So Zebra for NEXTHOP_TYPE_IPV4 returns the ifindex now since
+     * it was being stored.  This Doesn't solve all cases of
+     * recursive lookup but for the most common types it does.
      */
     first_ifindex = nexthop_tab[0].ifindex;
     nexthop_addr = nexthop_tab[0].nexthop_addr;
@@ -389,6 +394,27 @@ zclient_lookup_nexthop (struct pim_zlookup_nexthop nexthop_tab[],
        nexthop_tab[0].protocol_distance = protocol_distance;
       }
 
+      /*
+       * Let's see if any of the nexthops can actually be used.
+       * We need to check them against the neighbors that we
+       * have formed.  As that we shouldn't be sending
+       * j/p messages upstream towards non-neighbors
+       */
+      for (i = 0; i < num_ifindex ; i++)
+       {
+         struct interface *ifp;
+         struct pim_neighbor *nbr;
+
+         ifp = if_lookup_by_index_vrf (nexthop_tab[i].ifindex, VRF_DEFAULT);
+         nbr = pim_neighbor_find (ifp, nexthop_tab[i].nexthop_addr.u.prefix4);
+         if (!nbr)
+           {
+             num_ifindex--;
+             if (i != num_ifindex)
+               memcpy (&nexthop_tab[i], &nexthop_tab[i+1], sizeof (nexthop_tab[i]));
+             i--;
+           }
+       }
       return num_ifindex;
     }