bgpd: program mac-ip routes in matching vrfs

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd/bgp_evpn.c

@@ -782,6 +782,95 @@ static void evpn_delete_old_local_route(struct bgp *bgp, struct bgpevpn *vpn,
 	bgp_info_delete(rn, old_local);
 }
 
+/*
+ * Calculate the best path for an EVPN route. Install/update best path in zebra,
+ * if appropriate.
+ */
+static int evpn_vrf_route_select_install(struct bgp *bgp_vrf,
+					 struct bgp_node *rn,
+					 struct prefix *p,
+					 afi_t afi,
+					 safi_t safi)
+{
+	struct bgp_info *old_select, *new_select;
+	struct bgp_info_pair old_and_new;
+	int ret = 0;
+
+	/* Compute the best path. */
+	bgp_best_selection(bgp_vrf, rn, &bgp_vrf->maxpaths[afi][safi],
+			   &old_and_new, afi, safi);
+	old_select = old_and_new.old;
+	new_select = old_and_new.new;
+
+	/* If the best path hasn't changed - see if there is still something to
+	 * update
+	 * to zebra RIB.
+	 */
+	if (old_select && old_select == new_select
+	    && old_select->type == ZEBRA_ROUTE_BGP
+	    && old_select->sub_type == BGP_ROUTE_NORMAL
+	    && !CHECK_FLAG(rn->flags, BGP_NODE_USER_CLEAR)
+	    && !CHECK_FLAG(old_select->flags, BGP_INFO_ATTR_CHANGED)
+	    && !bgp_vrf->addpath_tx_used[afi][safi]) {
+		if (bgp_zebra_has_route_changed(rn, old_select))
+			bgp_zebra_announce(rn, p, old_select, bgp_vrf,
+					   afi, safi);
+		UNSET_FLAG(old_select->flags, BGP_INFO_MULTIPATH_CHG);
+		bgp_zebra_clear_route_change_flags(rn);
+		return ret;
+	}
+
+	/* If the user did a "clear" this flag will be set */
+	UNSET_FLAG(rn->flags, BGP_NODE_USER_CLEAR);
+
+	/* bestpath has changed; update relevant fields and install or uninstall
+	 * into the zebra RIB.
+	 */
+	if (old_select || new_select)
+		bgp_bump_version(rn);
+
+	if (old_select)
+		bgp_info_unset_flag(rn, old_select, BGP_INFO_SELECTED);
+	if (new_select) {
+		bgp_info_set_flag(rn, new_select, BGP_INFO_SELECTED);
+		bgp_info_unset_flag(rn, new_select, BGP_INFO_ATTR_CHANGED);
+		UNSET_FLAG(new_select->flags, BGP_INFO_MULTIPATH_CHG);
+	}
+
+	if (new_select && new_select->type == ZEBRA_ROUTE_BGP
+	    && new_select->sub_type == BGP_ROUTE_NORMAL) {
+		bgp_zebra_announce(rn, p, new_select, bgp_vrf,
+				   afi, safi);
+
+		/* If an old best existed and it was a "local" route, the only
+		 * reason
+		 * it would be supplanted is due to MAC mobility procedures. So,
+		 * we
+		 * need to do an implicit delete and withdraw that route from
+		 * peers.
+		 */
+		/*if (old_select && old_select->peer == bgp_vrf->peer_self
+		    && old_select->type == ZEBRA_ROUTE_BGP
+		    && old_select->sub_type == BGP_ROUTE_STATIC) {
+			//evpn_delete_old_local_route(bgp, vpn, rn, old_select);
+		}*/ //TODO_MITESH: probably not needed for vrf routes, think!!
+	} else {
+		if (old_select && old_select->type == ZEBRA_ROUTE_BGP
+		    && old_select->sub_type == BGP_ROUTE_NORMAL) {
+			bgp_zebra_withdraw(p, old_select, safi);
+		}
+	}
+
+	/* Clear any route change flags. */
+	bgp_zebra_clear_route_change_flags(rn);
+
+	/* Reap old select bgp_info, if it has been removed */
+	if (old_select && CHECK_FLAG(old_select->flags, BGP_INFO_REMOVED))
+		bgp_info_reap(rn, old_select);
+
+	return ret;
+}
+
 /*
  * Calculate the best path for an EVPN route. Install/update best path in zebra,
  * if appropriate.
@@ -1418,6 +1507,88 @@ static int handle_tunnel_ip_change(struct bgp *bgp, struct bgpevpn *vpn,
 	return update_routes_for_vni(bgp, vpn);
 }
 
+/*
+ * Install route entry into the VRF routing table and invoke route selection.
+ */
+static int install_evpn_route_entry_in_vrf(struct bgp *bgp_vrf,
+					   struct prefix_evpn *evp,
+					   struct bgp_info *parent_ri)
+{
+	struct bgp_node *rn;
+	struct bgp_info *ri;
+	struct attr *attr_new;
+	int ret;
+	struct prefix p;
+	struct prefix *pp = &p;
+	afi_t afi = 0;
+	safi_t safi = 0;
+
+	memset(pp, 0, sizeof(struct prefix));
+	ip_prefix_from_type2_prefix(evp, pp);
+
+	/* Create (or fetch) route within the VRF. */
+	/* NOTE: There is no RD here. */
+	if (IS_EVPN_PREFIX_IPADDR_V4(evp)) {
+		afi = AFI_IP;
+		safi = SAFI_UNICAST;
+		rn = bgp_node_get(bgp_vrf->rib[afi][safi], pp);
+	} else if (IS_EVPN_PREFIX_IPADDR_V6(evp)) {
+		afi = AFI_IP6;
+		safi = SAFI_UNICAST;
+		rn = bgp_node_get(bgp_vrf->rib[afi][safi], pp);
+	} else
+		return 0;
+
+	/* Check if route entry is already present. */
+	for (ri = rn->info; ri; ri = ri->next)
+		if (ri->extra
+		    && (struct bgp_info *)ri->extra->parent == parent_ri)
+			break;
+
+	if (!ri) {
+		/* Add (or update) attribute to hash. */
+		attr_new = bgp_attr_intern(parent_ri->attr);
+
+		/* Create new route with its attribute. */
+		ri = info_make(parent_ri->type, parent_ri->sub_type, 0,
+			       parent_ri->peer, attr_new, rn);
+		SET_FLAG(ri->flags, BGP_INFO_VALID);
+		bgp_info_extra_get(ri);
+		ri->extra->parent = parent_ri;
+		if (parent_ri->extra)
+			memcpy(&ri->extra->label, &parent_ri->extra->label,
+			       BGP_LABEL_BYTES);
+		bgp_info_add(rn, ri);
+	} else {
+		if (attrhash_cmp(ri->attr, parent_ri->attr)
+		    && !CHECK_FLAG(ri->flags, BGP_INFO_REMOVED)) {
+			bgp_unlock_node(rn);
+			return 0;
+		}
+		/* The attribute has changed. */
+		/* Add (or update) attribute to hash. */
+		attr_new = bgp_attr_intern(parent_ri->attr);
+
+		/* Restore route, if needed. */
+		if (CHECK_FLAG(ri->flags, BGP_INFO_REMOVED))
+			bgp_info_restore(rn, ri);
+
+		/* Mark if nexthop has changed. */
+		if (!IPV4_ADDR_SAME(&ri->attr->nexthop, &attr_new->nexthop))
+			SET_FLAG(ri->flags, BGP_INFO_IGP_CHANGED);
+
+		/* Unintern existing, set to new. */
+		bgp_attr_unintern(&ri->attr);
+		ri->attr = attr_new;
+		ri->uptime = bgp_clock();
+	}
+
+	/* Perform route selection and update zebra, if required. */
+	ret = evpn_vrf_route_select_install(bgp_vrf, rn, pp, afi, safi);
+
+	return ret;
+}
+
 /*
  * Install route entry into the VNI routing table and invoke route selection.
  */
@@ -1484,6 +1655,61 @@ static int install_evpn_route_entry(struct bgp *bgp, struct bgpevpn *vpn,
 	return ret;
 }
 
+/*
+ * Uninstall route entry from the VRF routing table and send message
+ * to zebra, if appropriate.
+ */
+static int uninstall_evpn_route_entry_in_vrf(struct bgp *bgp_vrf,
+					     struct prefix_evpn *evp,
+					     struct bgp_info *parent_ri)
+{
+	struct bgp_node *rn;
+	struct bgp_info *ri;
+	int ret;
+	struct prefix p;
+	struct prefix *pp = &p;
+	afi_t afi = 0;
+	safi_t safi = 0;
+
+	memset(pp, 0, sizeof(struct prefix));
+	ip_prefix_from_type2_prefix(evp, pp);
+
+	/* Locate route within the VRF. */
+	/* NOTE: There is no RD here. */
+	if (IS_EVPN_PREFIX_IPADDR_V4(evp)) {
+		afi = AFI_IP;
+		safi = SAFI_UNICAST;
+		rn = bgp_node_lookup(bgp_vrf->rib[afi][safi], pp);
+	} else {
+		afi = AFI_IP6;
+		safi = SAFI_UNICAST;
+		rn = bgp_node_lookup(bgp_vrf->rib[afi][safi], pp);
+	}
+
+	if (!rn)
+		return 0;
+
+	/* Find matching route entry. */
+	for (ri = rn->info; ri; ri = ri->next)
+		if (ri->extra
+		    && (struct bgp_info *)ri->extra->parent == parent_ri)
+			break;
+
+	if (!ri)
+		return 0;
+
+	/* Mark entry for deletion */
+	bgp_info_delete(rn, ri);
+
+	/* Perform route selection and update zebra, if required. */
+	ret = evpn_vrf_route_select_install(bgp_vrf, rn, pp, afi, safi);
+
+	/* Unlock route node. */
+	bgp_unlock_node(rn);
+
+	return ret;
+}
+
 /*
  * Uninstall route entry from the VNI routing table and send message
  * to zebra, if appropriate.
@@ -1706,6 +1932,45 @@ static int uninstall_routes_for_vni(struct bgp *bgp, struct bgpevpn *vpn)
 						0);
 }
 
+/*
+ * Install or uninstall route in matching VRFs (list).
+ */
+static int install_uninstall_route_in_vrfs(struct bgp *bgp_def, afi_t afi,
+					   safi_t safi, struct prefix_evpn *evp,
+					   struct bgp_info *ri,
+					   struct list *vrfs, int install)
+{
+	char buf[PREFIX2STR_BUFFER];
+	struct bgp *bgp_vrf;
+	struct listnode *node, *nnode;
+
+	/* Only type-2 routes go into a VRF */
+	if (!(evp->prefix.route_type == BGP_EVPN_MAC_IP_ROUTE))
+		return 0;
+
+	for (ALL_LIST_ELEMENTS(vrfs, node, nnode, bgp_vrf)) {
+		int ret;
+
+		if (install)
+			ret = install_evpn_route_entry_in_vrf(bgp_vrf,
+							      evp, ri);
+		else
+			ret = uninstall_evpn_route_entry_in_vrf(bgp_vrf,
+								evp, ri);
+
+		if (ret) {
+			zlog_err("%u: Failed to %s prefix %s in VRF %s",
+				 bgp_def->vrf_id,
+				 install ? "install" : "uninstall",
+				 prefix2str(evp, buf, sizeof(buf)),
+				 vrf_id_to_name(bgp_vrf->vrf_id));
+			return ret;
+		}
+	}
+
+	return 0;
+}
+
 /*
  * Install or uninstall route in matching VNIs (list).
  */
@@ -1777,7 +2042,8 @@ static int install_uninstall_evpn_route(struct bgp *bgp, afi_t afi, safi_t safi,
 		u_char type, sub_type;
 		struct ecommunity_val *eval;
 		struct ecommunity_val eval_tmp;
-		struct irt_node *irt;
+		struct irt_node *irt; /* import rt for l2vni */
+		struct vrf_irt_node *vrf_irt; /* import rt for l3vni */
 
 		/* Only deal with RTs */
 		pnt = (ecom->val + (i * ECOMMUNITY_SIZE));
@@ -1788,12 +2054,18 @@ static int install_uninstall_evpn_route(struct bgp *bgp, afi_t afi, safi_t safi,
 		if (sub_type != ECOMMUNITY_ROUTE_TARGET)
 			continue;
 
-		/* Are we interested in this RT? */
+		/* Import route into matching l2-vnis */
 		irt = lookup_import_rt(bgp, eval);
 		if (irt && irt->vnis)
 			install_uninstall_route_in_vnis(bgp, afi, safi, evp, ri,
 							irt->vnis, import);
 
+		/* Import route into matching l3-vnis (vrfs) */
+		vrf_irt = lookup_vrf_import_rt(eval);
+		if (vrf_irt && vrf_irt->vrfs)
+			install_uninstall_route_in_vrfs(bgp, afi, safi, evp, ri,
+							vrf_irt->vrfs, import);
+
 		/* Also check for non-exact match. In this, we mask out the AS
 		 * and
 		 * only check on the local-admin sub-field. This is to
@@ -1801,16 +2073,22 @@ static int install_uninstall_evpn_route(struct bgp *bgp, afi_t afi, safi_t safi,
 		 * VNI as the RT for EBGP peering too.
 		 */
 		irt = NULL;
+		vrf_irt = NULL;
 		if (type == ECOMMUNITY_ENCODE_AS
 		    || type == ECOMMUNITY_ENCODE_AS4
 		    || type == ECOMMUNITY_ENCODE_IP) {
 			memcpy(&eval_tmp, eval, ECOMMUNITY_SIZE);
 			mask_ecom_global_admin(&eval_tmp, eval);
 			irt = lookup_import_rt(bgp, &eval_tmp);
+			vrf_irt = lookup_vrf_import_rt(&eval_tmp);
 		}
 		if (irt && irt->vnis)
 			install_uninstall_route_in_vnis(bgp, afi, safi, evp, ri,
 							irt->vnis, import);
+		if (vrf_irt && vrf_irt->vrfs)
+			install_uninstall_route_in_vrfs(bgp, afi, safi, evp,
+							ri, vrf_irt->vrfs,
+							import);
 	}
 
 	return 0;

bgpd/bgp_evpn_private.h

@@ -258,6 +258,25 @@ static inline void encode_mac_mobility_extcomm(int static_mac, u_int32_t seq,
 	eval->val[7] = seq & 0xff;
 }
 
+static inline void ip_prefix_from_type2_prefix(struct prefix_evpn *evp,
+					       struct prefix *ip)
+{
+	memset(ip, 0, sizeof(struct prefix));
+	if (IS_EVPN_PREFIX_IPADDR_V4(evp)) {
+		ip->family = AF_INET;
+		ip->prefixlen = IPV4_MAX_BITLEN;
+		memcpy(&(ip->u.prefix4),
+		       &(evp->prefix.ip.ip),
+		       IPV4_MAX_BYTELEN);
+	} else if (IS_EVPN_PREFIX_IPADDR_V6(evp)) {
+		ip->family = AF_INET6;
+		ip->prefixlen = IPV6_MAX_BITLEN;
+		memcpy(&(ip->u.prefix6),
+		       &(evp->prefix.ip.ip),
+		       IPV6_MAX_BYTELEN);
+	}
+}
+
 static inline void build_evpn_type2_prefix(struct prefix_evpn *p,
 					   struct ethaddr *mac,
 					   struct ipaddr *ip)

bgpd/bgp_evpn_vty.c

@@ -411,6 +411,52 @@ static void display_vni(struct vty *vty, struct bgpevpn *vpn, json_object *json)
 		json_object_object_add(json, "exportRts", json_export_rtl);
 }
 
+static void evpn_show_vrf_routes(struct vty *vty,
+				 struct bgp *bgp_vrf)
+{
+	struct bgp *bgp_def = NULL;
+	struct bgp_node *rn;
+	struct bgp_info *ri;
+	int header = 1;
+	u_int32_t prefix_cnt, path_cnt;
+
+	prefix_cnt = path_cnt = 0;
+	bgp_def = bgp_get_default();
+	if (!bgp_def)
+		return;
+
+	for (rn = bgp_table_top(bgp_vrf->rib[AFI_L2VPN][SAFI_EVPN]); rn;
+	     rn = bgp_route_next(rn)) {
+		char prefix_str[BUFSIZ];
+
+		bgp_evpn_route2str((struct prefix_evpn *)&rn->p, prefix_str,
+				   sizeof(prefix_str));
+
+		if (rn->info) {
+			/* Overall header/legend displayed once. */
+			if (header) {
+				bgp_evpn_show_route_header(vty, bgp_def, NULL);
+				header = 0;
+			}
+			prefix_cnt++;
+		}
+
+		/* For EVPN, the prefix is displayed for each path (to fit in
+		 * with code that already exists).
+		 */
+		for (ri = rn->info; ri; ri = ri->next) {
+			route_vty_out(vty, &rn->p, ri, 0, SAFI_EVPN, NULL);
+			path_cnt++;
+		}
+	}
+
+	if (prefix_cnt == 0)
+		vty_out(vty, "No EVPN prefixes exist for this VRF");
+	else
+		vty_out(vty, "\nDisplayed %u prefixes (%u paths)",
+			prefix_cnt, path_cnt);
+}
+
 static void show_vni_routes(struct bgp *bgp, struct bgpevpn *vpn, int type,
 			    struct vty *vty, struct in_addr vtep_ip,
 			    json_object *json)
@@ -2640,6 +2686,33 @@ DEFUN(show_bgp_l2vpn_evpn_route_rd_macip,
 	return CMD_SUCCESS;
 }
 
+/*
+ * Display per-VRF EVPN routing table.
+ */
+DEFUN(show_bgp_l2vpn_evpn_route_vrf, show_bgp_l2vpn_evpn_route_vrf_cmd,
+      "show bgp l2vpn evpn route vrf VRFNAME",
+      SHOW_STR
+      BGP_STR
+      L2VPN_HELP_STR
+      EVPN_HELP_STR
+      "EVPN route information\n"
+      "VRF\n"
+      "VRF Name\n")
+{
+	int vrf_idx = 6;
+	char *vrf_name = NULL;
+	struct bgp *bgp_vrf = NULL;
+
+	vrf_name = argv[vrf_idx]->arg;
+	bgp_vrf = bgp_lookup_by_name(vrf_name);
+	if (!bgp_vrf)
+		return CMD_WARNING;
+
+	evpn_show_vrf_routes(vty, bgp_vrf);
+
+	return CMD_SUCCESS;
+}
+
 /*
  * Display per-VNI EVPN routing table.
  */
@@ -3697,6 +3770,7 @@ void bgp_ethernetvpn_init(void)
 	install_element(VIEW_NODE, &show_bgp_l2vpn_evpn_route_rd_cmd);
 	install_element(VIEW_NODE, &show_bgp_l2vpn_evpn_route_rd_macip_cmd);
 	install_element(VIEW_NODE, &show_bgp_l2vpn_evpn_route_vni_cmd);
+	install_element(VIEW_NODE, &show_bgp_l2vpn_evpn_route_vrf_cmd);
 	install_element(VIEW_NODE,
 			&show_bgp_l2vpn_evpn_route_vni_multicast_cmd);
 	install_element(VIEW_NODE, &show_bgp_l2vpn_evpn_route_vni_macip_cmd);

bgpd/bgp_zebra.c

@@ -1015,6 +1015,12 @@ void bgp_zebra_announce(struct bgp_node *rn, struct prefix *p,
 	if (info->sub_type == BGP_ROUTE_AGGREGATE)
 		zapi_route_set_blackhole(&api, BLACKHOLE_NULL);
 
+	/* If it is an EVPN route mark as such.
+	 * Currently presence of rmac in attr denotes
+	 * this is an EVPN type-2 route */
+	if (!is_zero_mac(&(info->attr->rmac)))
+		SET_FLAG(api.flags, ZEBRA_FLAG_EVPN_TYPE2_ROUTE);
+
 	if (peer->sort == BGP_PEER_IBGP || peer->sort == BGP_PEER_CONFED
 	    || info->sub_type == BGP_ROUTE_AGGREGATE) {
 		SET_FLAG(api.flags, ZEBRA_FLAG_IBGP);
@@ -1072,7 +1078,13 @@ void bgp_zebra_announce(struct bgp_node *rn, struct prefix *p,
 
 			api_nh = &api.nexthops[valid_nh_count];
 			api_nh->gate.ipv4 = *nexthop;
-			api_nh->type = NEXTHOP_TYPE_IPV4;
+
+			/* EVPN type-2 routes are
+			   programmed as onlink on l3-vni SVI */
+			if (CHECK_FLAG(api.flags, ZEBRA_FLAG_EVPN_TYPE2_ROUTE))
+				api_nh->type = NEXTHOP_TYPE_IPV4_IFINDEX;
+			else
+				api_nh->type = NEXTHOP_TYPE_IPV4;
 		} else {
 			ifindex_t ifindex;
 			struct in6_addr *nexthop;

lib/zebra.h

@@ -409,6 +409,7 @@ extern const char *zserv_command_string(unsigned int command);
 #define ZEBRA_FLAG_STATIC             0x40
 #define ZEBRA_FLAG_SCOPE_LINK         0x100
 #define ZEBRA_FLAG_FIB_OVERRIDE       0x200
+#define ZEBRA_FLAG_EVPN_TYPE2_ROUTE   0x400
 /* ZEBRA_FLAG_BLACKHOLE was 0x04 */
 /* ZEBRA_FLAG_REJECT was 0x80 */
 

zebra/zebra_rib.c

@@ -259,7 +259,8 @@ struct nexthop *route_entry_nexthop_ipv4_ifindex_add(struct route_entry *re,
 	/*Pending: need to think if null ifp here is ok during bootup?
 	  There was a crash because ifp here was coming to be NULL */
 	if (ifp)
-		if (connected_is_unnumbered(ifp)) {
+		if (connected_is_unnumbered(ifp) ||
+		    CHECK_FLAG(re->flags, ZEBRA_FLAG_EVPN_TYPE2_ROUTE)) {
 			SET_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK);
 		}
 

zebra/zebra_vxlan.c

@@ -6084,3 +6084,15 @@ void zebra_vxlan_ns_disable(struct zebra_ns *zns)
 {
 	hash_free(zns->l3vni_table);
 }
+
+/* get the l3vni svi ifindex */
+ifindex_t get_l3vni_svi_ifindex(vrf_id_t vrf_id)
+{
+	zebra_l3vni_t *zl3vni = NULL;
+
+	zl3vni = zl3vni_from_vrf(vrf_id);
+	if (!zl3vni || !is_l3vni_oper_up(zl3vni))
+		return 0;
+
+	return zl3vni->svi_if->ifindex;
+}

zebra/zebra_vxlan.h

@@ -51,6 +51,7 @@ is_evpn_enabled()
 
 #define VNI_STR_LEN 32
 
+extern ifindex_t get_l3vni_svi_ifindex(vrf_id_t vrf_id);
 extern int zebra_vxlan_vrf_delete(struct zebra_vrf *);
 extern void zebra_vxlan_print_macs_vni(struct vty *vty, struct zebra_vrf *zvrf,
 				       vni_t vni, u_char use_json);

zebra/zserv.c

@@ -1149,6 +1149,7 @@ static int zread_route_add(struct zserv *client, u_short length,
 	if (CHECK_FLAG(api.message, ZAPI_MESSAGE_NEXTHOP)) {
 		for (i = 0; i < api.nexthop_num; i++) {
 			api_nh = &api.nexthops[i];
+			ifindex_t ifindex = 0;
 
 			switch (api_nh->type) {
 			case NEXTHOP_TYPE_IFINDEX:
@@ -1160,9 +1161,16 @@ static int zread_route_add(struct zserv *client, u_short length,
 					re, &api_nh->gate.ipv4, NULL);
 				break;
 			case NEXTHOP_TYPE_IPV4_IFINDEX:
+				if (CHECK_FLAG(api.flags,
+					       ZEBRA_FLAG_EVPN_TYPE2_ROUTE)) {
+					ifindex =
+						get_l3vni_svi_ifindex(zvrf_id(zvrf));
+				} else {
+					ifindex = api_nh->ifindex;
+				}
 				nexthop = route_entry_nexthop_ipv4_ifindex_add(
 					re, &api_nh->gate.ipv4, NULL,
-					api_nh->ifindex);
+					ifindex);
 				break;
 			case NEXTHOP_TYPE_IPV6:
 				nexthop = route_entry_nexthop_ipv6_add(