ldpd: update local labels when necessary

ldpd allocates null labels for directly connected routes. If a connected
route is removed (interface goes down) and an IGP learned route takes its
place in the RIB, ldpd must update the local label of the associated FEC
entry with a non-null label. The same applies for the other way around
(an interface goes up and a connected route is selected in favour of an
IGP route). Labels should be dynamic and change when necessary.

Additionally, this patch fixes the processing of route delete messages
from zebra. Route delete messages don't contain any nexthop, meaning that
whenever we receive such messages we must delete all nexthop previously
received.

Based on a patch from Bingen Eguzkitza <bingen@voltanet.io>.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>

6 files changed, 121 insertions, 100 deletions

diff --git a/ldpd/l2vpn.c b/ldpd/l2vpn.c
index c1d0437fb..792608d42 100644
--- a/ldpd/l2vpn.c
+++ b/ldpd/l2vpn.c
@@ -211,6 +211,7 @@ l2vpn_pw_init(struct l2vpn_pw *pw)
 	l2vpn_pw_fec(pw, &fec);
 	lde_kernel_insert(&fec, AF_INET, (union ldpd_addr*)&pw->lsr_id, 0, 0,
 	    0, (void *)pw);
+	lde_kernel_update(&fec);
 }
 
 void
@@ -220,6 +221,7 @@ l2vpn_pw_exit(struct l2vpn_pw *pw)
 
 	l2vpn_pw_fec(pw, &fec);
 	lde_kernel_remove(&fec, AF_INET, (union ldpd_addr*)&pw->lsr_id, 0, 0);
+	lde_kernel_update(&fec);
 }
 
 static void
diff --git a/ldpd/lde.c b/ldpd/lde.c
index 05267a82c..1323ba3d0 100644
--- a/ldpd/lde.c
+++ b/ldpd/lde.c
@@ -414,8 +414,7 @@ lde_dispatch_parent(struct thread *thread)
 
 		switch (imsg.hdr.type) {
 		case IMSG_NETWORK_ADD:
-		case IMSG_NETWORK_ADD_END:
-		case IMSG_NETWORK_DEL:
+		case IMSG_NETWORK_UPDATE:
 			if (imsg.hdr.len != IMSG_HEADER_SIZE + sizeof(kr)) {
 				log_warnx("%s: wrong imsg len", __func__);
 				break;
@@ -443,12 +442,8 @@ lde_dispatch_parent(struct thread *thread)
 				    kr.ifindex, kr.priority,
 				    kr.flags & F_CONNECTED, NULL);
 				break;
-			case IMSG_NETWORK_ADD_END:
-				lde_kernel_reevaluate(&fec);
-				break;
-			case IMSG_NETWORK_DEL:
-				lde_kernel_remove(&fec, kr.af, &kr.nexthop,
-				    kr.ifindex, kr.priority);
+			case IMSG_NETWORK_UPDATE:
+				lde_kernel_update(&fec);
 				break;
 			}
 			break;
@@ -584,28 +579,37 @@ lde_acl_check(char *acl_name, int af, union ldpd_addr *addr, uint8_t prefixlen)
 }
 
 uint32_t
-lde_assign_label(struct fec *fec, int connected)
+lde_update_label(struct fec_node *fn)
 {
-	static uint32_t label = MPLS_LABEL_RESERVED_MAX;
+	static uint32_t	 label = MPLS_LABEL_RESERVED_MAX;
+	struct fec_nh	*fnh;
+	int		 connected = 0;
+
+	LIST_FOREACH(fnh, &fn->nexthops, entry) {
+		if (fnh->flags & F_FEC_NH_CONNECTED) {
+			connected = 1;
+			break;
+		}
+	}
 
 	/* should we allocate a label for this fec? */
-	switch (fec->type) {
+	switch (fn->fec.type) {
 	case FEC_TYPE_IPV4:
 		if ((ldeconf->ipv4.flags & F_LDPD_AF_ALLOCHOSTONLY) &&
-		    fec->u.ipv4.prefixlen != 32)
+		    fn->fec.u.ipv4.prefixlen != 32)
 			return (NO_LABEL);
 		if (lde_acl_check(ldeconf->ipv4.acl_label_allocate_for,
-		    AF_INET, (union ldpd_addr *)&fec->u.ipv4.prefix,
-		    fec->u.ipv4.prefixlen) != FILTER_PERMIT)
+		    AF_INET, (union ldpd_addr *)&fn->fec.u.ipv4.prefix,
+		    fn->fec.u.ipv4.prefixlen) != FILTER_PERMIT)
 			return (NO_LABEL);
 		break;
 	case FEC_TYPE_IPV6:
 		if ((ldeconf->ipv6.flags & F_LDPD_AF_ALLOCHOSTONLY) &&
-		    fec->u.ipv6.prefixlen != 128)
+		    fn->fec.u.ipv6.prefixlen != 128)
 			return (NO_LABEL);
 		if (lde_acl_check(ldeconf->ipv6.acl_label_allocate_for,
-		    AF_INET6, (union ldpd_addr *)&fec->u.ipv6.prefix,
-		    fec->u.ipv6.prefixlen) != FILTER_PERMIT)
+		    AF_INET6, (union ldpd_addr *)&fn->fec.u.ipv6.prefix,
+		    fn->fec.u.ipv6.prefixlen) != FILTER_PERMIT)
 			return (NO_LABEL);
 		break;
 	default:
@@ -614,29 +618,34 @@ lde_assign_label(struct fec *fec, int connected)
 
 	if (connected) {
 		/* choose implicit or explicit-null depending on configuration */
-		switch (fec->type) {
+		switch (fn->fec.type) {
 		case FEC_TYPE_IPV4:
 			if (!(ldeconf->ipv4.flags & F_LDPD_AF_EXPNULL))
 				return (MPLS_LABEL_IMPLNULL);
 			if (lde_acl_check(ldeconf->ipv4.acl_label_expnull_for,
-			    AF_INET, (union ldpd_addr *)&fec->u.ipv4.prefix,
-			    fec->u.ipv4.prefixlen) != FILTER_PERMIT)
+			    AF_INET, (union ldpd_addr *)&fn->fec.u.ipv4.prefix,
+			    fn->fec.u.ipv4.prefixlen) != FILTER_PERMIT)
 				return (MPLS_LABEL_IMPLNULL);
 			return (MPLS_LABEL_IPV4NULL);
 		case FEC_TYPE_IPV6:
 			if (!(ldeconf->ipv6.flags & F_LDPD_AF_EXPNULL))
 				return (MPLS_LABEL_IMPLNULL);
 			if (lde_acl_check(ldeconf->ipv6.acl_label_expnull_for,
-			    AF_INET6, (union ldpd_addr *)&fec->u.ipv6.prefix,
-			    fec->u.ipv6.prefixlen) != FILTER_PERMIT)
+			    AF_INET6, (union ldpd_addr *)&fn->fec.u.ipv6.prefix,
+			    fn->fec.u.ipv6.prefixlen) != FILTER_PERMIT)
 				return (MPLS_LABEL_IMPLNULL);
 			return (MPLS_LABEL_IPV6NULL);
 		default:
-			fatalx("lde_assign_label: unexpected fec type");
+			fatalx("lde_update_label: unexpected fec type");
 			break;
 		}
 	}
 
+	/* preserve current label if there's no need to update it */
+	if (fn->local_label != NO_LABEL &&
+	    fn->local_label > MPLS_LABEL_RESERVED_MAX)
+		return (fn->local_label);
+
 	/*
 	 * TODO: request label to zebra or define a range of labels for ldpd.
 	 */
@@ -1371,7 +1380,7 @@ lde_change_egress_label(int af)
 			fatalx("lde_change_egress_label: unknown af");
 		}
 
-		fn->local_label = lde_assign_label(&fn->fec, 1);
+		fn->local_label = lde_update_label(fn);
 		if (fn->local_label != NO_LABEL)
 			RB_FOREACH(ln, nbr_tree, &lde_nbrs)
 				lde_send_labelmapping(ln, fn, 0);
diff --git a/ldpd/lde.h b/ldpd/lde.h
index 7fa5219b1..e0e9873d5 100644
--- a/ldpd/lde.h
+++ b/ldpd/lde.h
@@ -110,6 +110,7 @@ struct fec_nh {
 	uint8_t			 flags;
 };
 #define F_FEC_NH_NEW		0x01
+#define F_FEC_NH_CONNECTED	0x02
 
 struct fec_node {
 	struct fec		 fec;
@@ -134,7 +135,7 @@ void		 lde(const char *, const char *);
 int		 lde_imsg_compose_parent(int, pid_t, void *, uint16_t);
 int		 lde_imsg_compose_ldpe(int, uint32_t, pid_t, void *, uint16_t);
 int		 lde_acl_check(char *, int, union ldpd_addr *, uint8_t);
-uint32_t	 lde_assign_label(struct fec *, int);
+uint32_t	 lde_update_label(struct fec_node *);
 void		 lde_send_change_klabel(struct fec_node *, struct fec_nh *);
 void		 lde_send_delete_klabel(struct fec_node *, struct fec_nh *);
 void		 lde_fec2map(struct fec *, struct map *);
@@ -174,7 +175,7 @@ void		 lde_kernel_insert(struct fec *, int, union ldpd_addr *,
 		    ifindex_t, uint8_t, int, void *);
 void		 lde_kernel_remove(struct fec *, int, union ldpd_addr *,
 		    ifindex_t, uint8_t);
-void		 lde_kernel_reevaluate(struct fec *);
+void		 lde_kernel_update(struct fec *);
 void		 lde_check_mapping(struct map *, struct lde_nbr *);
 void		 lde_check_request(struct map *, struct lde_nbr *);
 void		 lde_check_release(struct map *, struct lde_nbr *);
diff --git a/ldpd/lde_lib.c b/ldpd/lde_lib.c
index 7a4cb760f..234d373fb 100644
--- a/ldpd/lde_lib.c
+++ b/ldpd/lde_lib.c
@@ -311,55 +311,19 @@ lde_kernel_insert(struct fec *fec, int af, union ldpd_addr *nexthop,
 {
 	struct fec_node		*fn;
 	struct fec_nh		*fnh;
-	struct lde_map		*me;
-	struct lde_nbr		*ln;
 
 	fn = (struct fec_node *)fec_find(&ft, fec);
 	if (fn == NULL)
 		fn = fec_add(fec);
-	fnh = fec_nh_find(fn, af, nexthop, ifindex, priority);
-	if (fnh != NULL) {
-		lde_send_change_klabel(fn, fnh);
-		fnh->flags |= F_FEC_NH_NEW;
-		return;
-	}
-
-	if (fn->fec.type == FEC_TYPE_PWID)
+	if (data)
 		fn->data = data;
 
-	if (fn->local_label == NO_LABEL) {
-		fn->local_label = lde_assign_label(&fn->fec, connected);
-
-		/* FEC.1: perform lsr label distribution procedure */
-		if (fn->local_label != NO_LABEL)
-			RB_FOREACH(ln, nbr_tree, &lde_nbrs)
-				lde_send_labelmapping(ln, fn, 1);
-	}
-
-	fnh = fec_nh_add(fn, af, nexthop, ifindex, priority);
+	fnh = fec_nh_find(fn, af, nexthop, ifindex, priority);
+	if (fnh == NULL)
+		fnh = fec_nh_add(fn, af, nexthop, ifindex, priority);
 	fnh->flags |= F_FEC_NH_NEW;
-	lde_send_change_klabel(fn, fnh);
-
-	switch (fn->fec.type) {
-	case FEC_TYPE_IPV4:
-	case FEC_TYPE_IPV6:
-		ln = lde_nbr_find_by_addr(af, &fnh->nexthop);
-		break;
-	case FEC_TYPE_PWID:
-		ln = lde_nbr_find_by_lsrid(fn->fec.u.pwid.lsr_id);
-		break;
-	default:
-		ln = NULL;
-		break;
-	}
-
-	if (ln) {
-		/* FEC.2  */
-		me = (struct lde_map *)fec_find(&ln->recv_map, &fn->fec);
-		if (me)
-			/* FEC.5 */
-			lde_check_mapping(&me->map, ln);
-	}
+	if (connected)
+		fnh->flags |= F_FEC_NH_CONNECTED;
 }
 
 void
@@ -380,12 +344,6 @@ lde_kernel_remove(struct fec *fec, int af, union ldpd_addr *nexthop,
 
 	lde_send_delete_klabel(fn, fnh);
 	fec_nh_del(fnh);
-	if (LIST_EMPTY(&fn->nexthops)) {
-		lde_send_labelwithdraw_all(fn, NO_LABEL);
-		fn->local_label = NO_LABEL;
-		if (fn->fec.type == FEC_TYPE_PWID)
-			fn->data = NULL;
-	}
 }
 
 /*
@@ -395,10 +353,12 @@ lde_kernel_remove(struct fec *fec, int af, union ldpd_addr *nexthop,
  * them (if any), withdraw the associated labels from zebra.
  */
 void
-lde_kernel_reevaluate(struct fec *fec)
+lde_kernel_update(struct fec *fec)
 {
 	struct fec_node		*fn;
 	struct fec_nh		*fnh, *safe;
+	struct lde_nbr		*ln;
+	struct lde_map		*me;
 
 	fn = (struct fec_node *)fec_find(&ft, fec);
 	if (fn == NULL)
@@ -407,9 +367,53 @@ lde_kernel_reevaluate(struct fec *fec)
 	LIST_FOREACH_SAFE(fnh, &fn->nexthops, entry, safe) {
 		if (fnh->flags & F_FEC_NH_NEW)
 			fnh->flags &= ~F_FEC_NH_NEW;
-		else
-			lde_kernel_remove(fec, fnh->af, &fnh->nexthop,
-			    fnh->ifindex, fnh->priority);
+		else {
+			lde_send_delete_klabel(fn, fnh);
+			fec_nh_del(fnh);
+		}
+	}
+
+	if (LIST_EMPTY(&fn->nexthops)) {
+		lde_send_labelwithdraw_all(fn, NO_LABEL);
+		fn->local_label = NO_LABEL;
+		fn->data = NULL;
+	} else {
+		uint32_t	 previous_label;
+
+		previous_label = fn->local_label;
+		fn->local_label = lde_update_label(fn);
+
+		if (fn->local_label != NO_LABEL &&
+		    fn->local_label != previous_label) {
+			/* FEC.1: perform lsr label distribution procedure */
+			RB_FOREACH(ln, nbr_tree, &lde_nbrs)
+				lde_send_labelmapping(ln, fn, 1);
+		}
+	}
+
+	LIST_FOREACH(fnh, &fn->nexthops, entry) {
+		lde_send_change_klabel(fn, fnh);
+
+		switch (fn->fec.type) {
+		case FEC_TYPE_IPV4:
+		case FEC_TYPE_IPV6:
+			ln = lde_nbr_find_by_addr(fnh->af, &fnh->nexthop);
+			break;
+		case FEC_TYPE_PWID:
+			ln = lde_nbr_find_by_lsrid(fn->fec.u.pwid.lsr_id);
+			break;
+		default:
+			ln = NULL;
+			break;
+		}
+
+		if (ln) {
+			/* FEC.2  */
+			me = (struct lde_map *)fec_find(&ln->recv_map, &fn->fec);
+			if (me)
+				/* FEC.5 */
+				lde_check_mapping(&me->map, ln);
+		}
 	}
 }
 
diff --git a/ldpd/ldp_zebra.c b/ldpd/ldp_zebra.c
index 79c4f5b37..12954b91a 100644
--- a/ldpd/ldp_zebra.c
+++ b/ldpd/ldp_zebra.c
@@ -353,7 +353,7 @@ ldp_zebra_read_route(int command, struct zclient *zclient, zebra_size_t length,
 	u_char			 type;
 	u_char			 message_flags;
 	struct kroute		 kr;
-	int			 nhnum, nhlen;
+	int			 nhnum = 0, nhlen;
 	size_t			 nhmark;
 
 	memset(&kr, 0, sizeof(kr));
@@ -374,8 +374,6 @@ ldp_zebra_read_route(int command, struct zclient *zclient, zebra_size_t length,
 	stream_getl(s); /* flags, unused */
 	stream_getw(s); /* instance, unused */
 	message_flags = stream_getc(s);
-	if (!CHECK_FLAG(message_flags, ZAPI_MESSAGE_NEXTHOP))
-		return (0);
 
 	switch (command) {
 	case ZEBRA_REDISTRIBUTE_IPV4_ADD:
@@ -409,16 +407,35 @@ ldp_zebra_read_route(int command, struct zclient *zclient, zebra_size_t length,
 			return (0);
 	}
 
-	nhnum = stream_getc(s);
-	nhmark = stream_get_getp(s);
-	stream_set_getp(s, nhmark + nhnum * (nhlen + 5));
+	if (CHECK_FLAG(message_flags, ZAPI_MESSAGE_NEXTHOP)) {
+		nhnum = stream_getc(s);
+		nhmark = stream_get_getp(s);
+		stream_set_getp(s, nhmark + nhnum * (nhlen + 5));
+	}
 
 	if (CHECK_FLAG(message_flags, ZAPI_MESSAGE_DISTANCE))
 		kr.priority = stream_getc(s);
 	if (CHECK_FLAG(message_flags, ZAPI_MESSAGE_METRIC))
 		stream_getl(s);	/* metric, not used */
 
-	stream_set_getp(s, nhmark);
+	if (CHECK_FLAG(message_flags, ZAPI_MESSAGE_NEXTHOP))
+		stream_set_getp(s, nhmark);
+
+	if (nhnum == 0) {
+		switch (command) {
+		case ZEBRA_REDISTRIBUTE_IPV4_ADD:
+		case ZEBRA_REDISTRIBUTE_IPV6_ADD:
+			return (0);
+		case ZEBRA_REDISTRIBUTE_IPV4_DEL:
+		case ZEBRA_REDISTRIBUTE_IPV6_DEL:
+			debug_zebra_in("route delete %s/%d (%s)",
+			    log_addr(kr.af, &kr.prefix), kr.prefixlen,
+			    zebra_route_string(type));
+			break;
+		default:
+			fatalx("ldp_zebra_read_route: unknown command");
+		}
+	}
 
 	/* loop through all the nexthops */
 	for (; nhnum > 0; nhnum--) {
@@ -445,23 +462,12 @@ ldp_zebra_read_route(int command, struct zclient *zclient, zebra_size_t length,
 			main_imsg_compose_lde(IMSG_NETWORK_ADD, 0, &kr,
 			    sizeof(kr));
 			break;
-		case ZEBRA_REDISTRIBUTE_IPV4_DEL:
-		case ZEBRA_REDISTRIBUTE_IPV6_DEL:
-			debug_zebra_in("route delete %s/%d nexthop %s "
-			    "ifindex %u (%s)", log_addr(kr.af, &kr.prefix),
-			    kr.prefixlen, log_addr(kr.af, &kr.nexthop),
-			    kr.ifindex, zebra_route_string(type));
-			main_imsg_compose_lde(IMSG_NETWORK_DEL, 0, &kr,
-			    sizeof(kr));
-			break;
 		default:
-			fatalx("ldp_zebra_read_route: unknown command");
+			break;
 		}
 	}
 
-	if (command == ZEBRA_REDISTRIBUTE_IPV4_ADD ||
-	    command == ZEBRA_REDISTRIBUTE_IPV6_ADD)
-		main_imsg_compose_lde(IMSG_NETWORK_ADD_END, 0, &kr, sizeof(kr));
+	main_imsg_compose_lde(IMSG_NETWORK_UPDATE, 0, &kr, sizeof(kr));
 
 	return (0);
 }
diff --git a/ldpd/ldpd.h b/ldpd/ldpd.h
index fa3789a83..ff3af19db 100644
--- a/ldpd/ldpd.h
+++ b/ldpd/ldpd.h
@@ -122,8 +122,7 @@ enum imsg_type {
 	IMSG_NEIGHBOR_UP,
 	IMSG_NEIGHBOR_DOWN,
 	IMSG_NETWORK_ADD,
-	IMSG_NETWORK_ADD_END,
-	IMSG_NETWORK_DEL,
+	IMSG_NETWORK_UPDATE,
 	IMSG_SOCKET_IPC,
 	IMSG_SOCKET_NET,
 	IMSG_CLOSE_SOCKETS,