Merge pull request #3020 from donaldsharp/global_5549

Allow v6 global addresses to be nexthops for v4 addresses in bgp

bgpd/bgp_nht.c

@@ -415,6 +415,8 @@ void bgp_parse_nexthop_update(int command, vrf_id_t vrf_id)
 		bnc->change_flags |= BGP_NEXTHOP_CHANGED;
 
 	if (nhr.nexthop_num) {
+		struct peer *peer = bnc->nht_info;
+
 		/* notify bgp fsm if nbr ip goes from invalid->valid */
 		if (!bnc->nexthop_num)
 			UNSET_FLAG(bnc->flags, BGP_NEXTHOP_PEER_NOTIFIED);
@@ -430,6 +432,22 @@ void bgp_parse_nexthop_update(int command, vrf_id_t vrf_id)
 
 			nexthop = nexthop_from_zapi_nexthop(&nhr.nexthops[i]);
 
+			/*
+			 * Turn on RA for the v6 nexthops
+			 * we receive from bgp.  This is to allow us
+			 * to work with v4 routing over v6 nexthops
+			 */
+			if (peer &&
+			    CHECK_FLAG(peer->flags, PEER_FLAG_CAPABILITY_ENHE)
+			    && nhr.prefix.family == AF_INET6) {
+				struct interface *ifp;
+
+				ifp = if_lookup_by_index(nexthop->ifindex,
+							 nexthop->vrf_id);
+				zclient_send_interface_radv_req(
+					zclient, nexthop->vrf_id, ifp, true,
+					BGP_UNNUM_DEFAULT_RA_INTERVAL);
+			}
 			/* There is at least one label-switched path */
 			if (nexthop->nh_label &&
 				nexthop->nh_label->num_labels) {

bgpd/bgp_open.c

@@ -1331,7 +1331,6 @@ void bgp_open_capability(struct stream *s, struct peer *peer)
 			 */
 			if (CHECK_FLAG(peer->flags, PEER_FLAG_CAPABILITY_ENHE)
 			    && peer->su.sa.sa_family == AF_INET6
-			    && IN6_IS_ADDR_LINKLOCAL(&peer->su.sin6.sin6_addr)
 			    && afi == AFI_IP
 			    && (safi == SAFI_UNICAST
 				|| safi == SAFI_LABELED_UNICAST)) {

doc/user/bgp.rst

@@ -845,6 +845,15 @@ Configuring Peers
    specified number of hops away will be allowed to become neighbors. This
    command is mutually exclusive with *ebgp-multihop*.
 
+.. index:: [no] neighbor PEER capability extended-nexthop
+.. clicmd:: [no] neighbor PEER capability extended-nexthop
+
+   Allow bgp to negotiate the extended-nexthop capability with it's peer.
+   If you are peering over a v6 LL address then this capability is turned
+   on automatically.  If you are peering over a v6 Global Address then
+   turning on this command will allow BGP to install v4 routes with
+   v6 nexthops if you do not have v4 configured on interfaces.
+
 .. index:: [no] bgp fast-external-failover
 .. clicmd:: [no] bgp fast-external-failover
 

zebra/interface.c

@@ -51,6 +51,8 @@
 #include "zebra/zebra_vxlan.h"
 #include "zebra/zebra_errors.h"
 
+DEFINE_MTYPE_STATIC(ZEBRA, ZINFO, "Zebra Interface Information")
+
 #define ZEBRA_PTM_SUPPORT
 
 DEFINE_HOOK(zebra_if_extra_info, (struct vty * vty, struct interface *ifp),
@@ -99,7 +101,7 @@ static int if_zebra_new_hook(struct interface *ifp)
 {
 	struct zebra_if *zebra_if;
 
-	zebra_if = XCALLOC(MTYPE_TMP, sizeof(struct zebra_if));
+	zebra_if = XCALLOC(MTYPE_ZINFO, sizeof(struct zebra_if));
 
 	zebra_if->multicast = IF_ZEBRA_MULTICAST_UNSPEC;
 	zebra_if->shutdown = IF_ZEBRA_SHUTDOWN_OFF;
@@ -136,6 +138,8 @@ static int if_zebra_new_hook(struct interface *ifp)
 	}
 #endif /* HAVE_RTADV */
 
+	memset(&zebra_if->neigh_mac[0], 0, 6);
+
 	/* Initialize installed address chains tree. */
 	zebra_if->ipv4_subnets =
 		route_table_init_with_delegate(&zebra_if_table_delegate);
@@ -175,7 +179,7 @@ static int if_zebra_delete_hook(struct interface *ifp)
 
 		THREAD_OFF(zebra_if->speed_update);
 
-		XFREE(MTYPE_TMP, zebra_if);
+		XFREE(MTYPE_ZINFO, zebra_if);
 	}
 
 	return 0;
@@ -802,19 +806,32 @@ static void ipv6_ll_address_to_mac(struct in6_addr *address, uint8_t *mac)
 	mac[5] = address->s6_addr[15];
 }
 
-void if_nbr_ipv6ll_to_ipv4ll_neigh_update(struct interface *ifp,
+static bool mac_is_same(char *mac1, char *mac2)
-					  struct in6_addr *address, int add)
+{
+	if (mac1[0] == mac2[0] &&
+	    mac1[1] == mac2[1] &&
+	    mac1[2] == mac2[2] &&
+	    mac1[3] == mac2[3] &&
+	    mac1[4] == mac2[4] &&
+	    mac1[5] == mac2[5])
+		return true;
+	else
+		return false;
+}
+
+void if_nbr_mac_to_ipv4ll_neigh_update(struct interface *ifp,
+				       char mac[6],
+				       struct in6_addr *address,
+				       int add)
 {
 	struct zebra_vrf *zvrf = vrf_info_lookup(ifp->vrf_id);
 	struct zebra_if *zif = ifp->info;
 	char buf[16] = "169.254.0.1";
 	struct in_addr ipv4_ll;
-	char mac[6];
 	ns_id_t ns_id;
 
 	inet_pton(AF_INET, buf, &ipv4_ll);
 
-	ipv6_ll_address_to_mac(address, (uint8_t *)mac);
 	ns_id = zvrf->zns->ns_id;
 
 	/*
@@ -823,10 +840,16 @@ void if_nbr_ipv6ll_to_ipv4ll_neigh_update(struct interface *ifp,
 	 *
 	 * supported message types are RTM_NEWNEIGH and RTM_DELNEIGH
 	 */
-	kernel_neigh_update(0, ifp->ifindex, ipv4_ll.s_addr, mac, 6, ns_id);
+	if (!mac_is_same(zif->neigh_mac, mac)) {
+		kernel_neigh_update(0, ifp->ifindex, ipv4_ll.s_addr,
+				    mac, 6, ns_id);
 
 		/* Add arp record */
-	kernel_neigh_update(add, ifp->ifindex, ipv4_ll.s_addr, mac, 6, ns_id);
+		kernel_neigh_update(add, ifp->ifindex, ipv4_ll.s_addr,
+				    mac, 6, ns_id);
+	}
+
+	memcpy(&zif->neigh_mac[0], &mac[0], 6);
 
 	/*
 	 * We need to note whether or not we originated a v6
@@ -840,6 +863,16 @@ void if_nbr_ipv6ll_to_ipv4ll_neigh_update(struct interface *ifp,
 	zvrf->neigh_updates++;
 }
 
+void if_nbr_ipv6ll_to_ipv4ll_neigh_update(struct interface *ifp,
+					  struct in6_addr *address, int add)
+{
+
+	char mac[6];
+
+	ipv6_ll_address_to_mac(address, (uint8_t *)mac);
+	if_nbr_mac_to_ipv4ll_neigh_update(ifp, mac, address, add);
+}
+
 static void if_nbr_ipv6ll_to_ipv4ll_neigh_add_all(struct interface *ifp)
 {
 	if (listhead(ifp->nbr_connected)) {

zebra/interface.h

@@ -279,6 +279,7 @@ struct zebra_if {
 	 * for bgp unnumbered?
 	 */
 	bool v6_2_v4_ll_neigh_entry;
+	char neigh_mac[6];
 	struct in6_addr v6_2_v4_ll_addr6;
 };
 
@@ -332,6 +333,10 @@ extern struct interface *if_link_per_ns(struct zebra_ns *, struct interface *);
 extern const char *ifindex2ifname_per_ns(struct zebra_ns *, unsigned int);
 
 extern void if_unlink_per_ns(struct interface *);
+extern void if_nbr_mac_to_ipv4ll_neigh_update(struct interface *fip,
+					      char mac[6],
+					      struct in6_addr *address,
+					      int add);
 extern void if_nbr_ipv6ll_to_ipv4ll_neigh_update(struct interface *ifp,
 						 struct in6_addr *address,
 						 int add);

zebra/rtadv.c

@@ -455,6 +455,38 @@ static void rtadv_process_solicit(struct interface *ifp)
 	rtadv_send_packet(zns->rtadv.sock, ifp);
 }
 
+/*
+ * This function processes optional attributes off of
+ * end of a RA packet received.  At this point in
+ * time we only care about this in one situation
+ * which is when a interface does not have a LL
+ * v6 address.  We still need to be able to install
+ * the mac address for v4 to v6 resolution
+ */
+static void rtadv_process_optional(uint8_t *optional, unsigned int len,
+				   struct interface *ifp,
+				   struct sockaddr_in6 *addr)
+{
+	char *mac;
+
+	while (len > 0) {
+		struct nd_opt_hdr *opt_hdr = (struct nd_opt_hdr *)optional;
+
+		switch(opt_hdr->nd_opt_type) {
+		case ND_OPT_SOURCE_LINKADDR:
+			mac = (char *)(optional+2);
+			if_nbr_mac_to_ipv4ll_neigh_update(ifp, mac,
+							  &addr->sin6_addr, 1);
+			break;
+		default:
+			break;
+		}
+
+		len -= 8 * opt_hdr->nd_opt_len;
+		optional += 8 * opt_hdr->nd_opt_len;
+	}
+}
+
 static void rtadv_process_advert(uint8_t *msg, unsigned int len,
 				 struct interface *ifp,
 				 struct sockaddr_in6 *addr)
@@ -469,13 +501,18 @@ static void rtadv_process_advert(uint8_t *msg, unsigned int len,
 	inet_ntop(AF_INET6, &addr->sin6_addr, addr_str, INET6_ADDRSTRLEN);
 
 	if (len < sizeof(struct nd_router_advert)) {
+		if (IS_ZEBRA_DEBUG_PACKET)
 			zlog_debug("%s(%u): Rx RA with invalid length %d from %s",
 				   ifp->name, ifp->ifindex, len, addr_str);
 		return;
 	}
+
 	if (!IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
-		zlog_debug(
+		rtadv_process_optional(msg + sizeof(struct nd_router_advert),
-			"%s(%u): Rx RA with non-linklocal source address from %s",
+				       len - sizeof(struct nd_router_advert),
+				       ifp, addr);
+		if (IS_ZEBRA_DEBUG_PACKET)
+			zlog_debug("%s(%u): Rx RA with non-linklocal source address from %s",
 				   ifp->name, ifp->ifindex, addr_str);
 		return;
 	}