bgpd: unimport evpn routes when implicit withdraw

When bgp update is received for EVPN prefix
where for an existing path's nexthop becomes unreachable,
the path is marked as not VALID but the routes
were not unimported from tenant vrfs, which lead to
stale unicast route(s) and nexthop(s).

In Multipath scenario only a specific path may have marked as
not VALID, then specific path info for the EVPN prefix required to be
unimported from tenant vrf.

Ticket: #3671288

Signed-off-by: Chirag Shah <chirag@nvidia.com>
Signed-off-by: Donald Sharp <sharpd@nvidia.com>

diff --git a/bgpd/bgp_route.c b/bgpd/bgp_route.c
index 08fdb40a08525e0f5d2c49bb8dccfa8d62d6ed09..2899b94d2ffdabb56ec9b38793436165136316be 100644 (file)
--- a/bgpd/bgp_route.c
+++ b/bgpd/bgp_route.c
@@ -4777,9 +4777,10 @@ void bgp_update(struct peer *peer, const struct prefix *p, uint32_t addpath_id,
                                                       BGP_PATH_VALID);
                        } else {
                                if (BGP_DEBUG(nht, NHT)) {
-                                       zlog_debug("%s(%pI4): NH unresolved",
+                                       zlog_debug("%s(%pI4): NH unresolved for existing %pFX pi %p flags 0x%x",
                                                   __func__,
-                                                  (in_addr_t *)&attr_new->nexthop);
+                                                  (in_addr_t *)&attr_new->nexthop,
+                                                  p, pi, pi->flags);
                                }
                                bgp_path_info_unset_flag(dest, pi,
                                                         BGP_PATH_VALID);
@@ -4824,10 +4825,23 @@ void bgp_update(struct peer *peer, const struct prefix *p, uint32_t addpath_id,
                 * updating
                 * the attributes for the route in the VNI(s).
                 */
-               if (safi == SAFI_EVPN &&
-                   (!same_attr || force_evpn_import) &&
-                   CHECK_FLAG(pi->flags, BGP_PATH_VALID))
-                       bgp_evpn_import_route(bgp, afi, safi, p, pi);
+               if (safi == SAFI_EVPN) {
+                       if ((!same_attr || force_evpn_import) &&
+                           CHECK_FLAG(pi->flags, BGP_PATH_VALID))
+                               bgp_evpn_import_route(bgp, afi, safi, p, pi);
+
+                       /* If existing path is marked invalid then unimport the
+                        * path from EVPN prefix. This will ensure EVPN route
+                        * has only valid paths and path refcount maintained in
+                        * EVPN nexthop is decremented appropriately.
+                        */
+                       else if (!CHECK_FLAG(pi->flags, BGP_PATH_VALID)) {
+                               if (BGP_DEBUG(nht, NHT))
+                                       zlog_debug("%s unimport EVPN %pFX as pi %p is not VALID",
+                                                  __func__, p, pi);
+                               bgp_evpn_unimport_route(bgp, afi, safi, p, pi);
+                       }
+               }
 
                /* Process change. */
                bgp_aggregate_increment(bgp, p, pi, afi, safi);