diff --git a/bgpd/bgp_mplsvpn.c b/bgpd/bgp_mplsvpn.c
index 6b93737282..0ac76c48b4 100644
--- a/bgpd/bgp_mplsvpn.c
+++ b/bgpd/bgp_mplsvpn.c
@@ -39,6 +39,9 @@
 #include "bgpd/rfapi/rfapi_backend.h"
 #endif
 
+DEFINE_MTYPE_STATIC(BGPD, MPLSVPN_NH_LABEL_BIND_CACHE,
+		    "BGP MPLSVPN nexthop label bind cache");
+
 /*
  * Definitions and external declarations.
  */
@@ -3860,3 +3863,61 @@ void bgp_vpn_leak_export(struct bgp *from_bgp)
 		}
 	}
 }
+
+/* The nexthops values are compared to
+ * find in the tree the appropriate cache entry
+ */
+int bgp_mplsvpn_nh_label_bind_cmp(
+	const struct bgp_mplsvpn_nh_label_bind_cache *a,
+	const struct bgp_mplsvpn_nh_label_bind_cache *b)
+{
+	if (prefix_cmp(&a->nexthop, &b->nexthop))
+		return 1;
+	if (a->orig_label > b->orig_label)
+		return 1;
+	if (a->orig_label < b->orig_label)
+		return -1;
+	return 0;
+}
+
+void bgp_mplsvpn_nh_label_bind_free(
+	struct bgp_mplsvpn_nh_label_bind_cache *bmnc)
+{
+	if (bmnc->new_label != MPLS_INVALID_LABEL)
+		bgp_lp_release(LP_TYPE_BGP_L3VPN_BIND, bmnc, bmnc->new_label);
+	bgp_mplsvpn_nh_label_bind_cache_del(
+		&bmnc->bgp_vpn->mplsvpn_nh_label_bind, bmnc);
+	XFREE(MTYPE_MPLSVPN_NH_LABEL_BIND_CACHE, bmnc);
+}
+
+struct bgp_mplsvpn_nh_label_bind_cache *
+bgp_mplsvpn_nh_label_bind_new(struct bgp_mplsvpn_nh_label_bind_cache_head *tree,
+			      struct prefix *p, mpls_label_t orig_label)
+{
+	struct bgp_mplsvpn_nh_label_bind_cache *bmnc;
+
+	bmnc = XCALLOC(MTYPE_MPLSVPN_NH_LABEL_BIND_CACHE,
+		       sizeof(struct bgp_mplsvpn_nh_label_bind_cache));
+	bmnc->new_label = MPLS_INVALID_LABEL;
+	prefix_copy(&bmnc->nexthop, p);
+	bmnc->orig_label = orig_label;
+
+	LIST_INIT(&(bmnc->paths));
+	bgp_mplsvpn_nh_label_bind_cache_add(tree, bmnc);
+
+	return bmnc;
+}
+
+struct bgp_mplsvpn_nh_label_bind_cache *bgp_mplsvpn_nh_label_bind_find(
+	struct bgp_mplsvpn_nh_label_bind_cache_head *tree, struct prefix *p,
+	mpls_label_t orig_label)
+{
+	struct bgp_mplsvpn_nh_label_bind_cache bmnc = {0};
+
+	if (!tree)
+		return NULL;
+	prefix_copy(&bmnc.nexthop, p);
+	bmnc.orig_label = orig_label;
+
+	return bgp_mplsvpn_nh_label_bind_cache_find(tree, &bmnc);
+}
diff --git a/bgpd/bgp_mplsvpn.h b/bgpd/bgp_mplsvpn.h
index 75758edcc2..5fe64ade48 100644
--- a/bgpd/bgp_mplsvpn.h
+++ b/bgpd/bgp_mplsvpn.h
@@ -325,4 +325,60 @@ extern void vpn_handle_router_id_update(struct bgp *bgp, bool withdraw,
 extern void bgp_vpn_leak_unimport(struct bgp *from_bgp);
 extern void bgp_vpn_leak_export(struct bgp *from_bgp);
 
+extern int
+bgp_mplsvpn_nh_label_bind_cmp(const struct bgp_mplsvpn_nh_label_bind_cache *a,
+			      const struct bgp_mplsvpn_nh_label_bind_cache *b);
+
+/* used to bind a local label to the (label, nexthop) values
+ * from an incoming BGP mplsvpn update
+ */
+struct bgp_mplsvpn_nh_label_bind_cache {
+
+	/* RB-tree entry. */
+	struct bgp_mplsvpn_nh_label_bind_cache_item entry;
+
+	/* The nexthop and the vpn label are the key of the list.
+	 * Only received BGP MPLSVPN updates may use that structure.
+	 * orig_label is the original label received from the BGP Update.
+	 */
+	struct prefix nexthop;
+	mpls_label_t orig_label;
+
+	/* number of mplsvpn path */
+	unsigned int path_count;
+
+	/* back pointer to bgp instance */
+	struct bgp *bgp_vpn;
+
+	/* MPLS label allocated value.
+	 * When the next-hop is changed because of 'next-hop-self' or
+	 * because it is an eBGP peer, the redistributed orig_label value
+	 * is unmodified, unless the 'l3vpn-multi-domain-switching'
+	 * is enabled: a new_label value is allocated:
+	 * - The new_label value is sent in the advertised BGP update,
+	 * instead of the label value.
+	 * - An MPLS entry is set to swap <new_label> with <orig_label>.
+	 */
+	mpls_label_t new_label;
+
+	/* list of path_vrfs using it */
+	LIST_HEAD(mplsvpn_nh_label_bind_path_lists, bgp_path_info) paths;
+
+	time_t last_update;
+};
+
+DECLARE_RBTREE_UNIQ(bgp_mplsvpn_nh_label_bind_cache,
+		    struct bgp_mplsvpn_nh_label_bind_cache, entry,
+		    bgp_mplsvpn_nh_label_bind_cmp);
+
+void bgp_mplsvpn_nh_label_bind_free(
+	struct bgp_mplsvpn_nh_label_bind_cache *bmnc);
+
+struct bgp_mplsvpn_nh_label_bind_cache *
+bgp_mplsvpn_nh_label_bind_new(struct bgp_mplsvpn_nh_label_bind_cache_head *tree,
+			      struct prefix *p, mpls_label_t orig_label);
+struct bgp_mplsvpn_nh_label_bind_cache *bgp_mplsvpn_nh_label_bind_find(
+	struct bgp_mplsvpn_nh_label_bind_cache_head *tree, struct prefix *p,
+	mpls_label_t orig_label);
+
 #endif /* _QUAGGA_BGP_MPLSVPN_H */
diff --git a/bgpd/bgpd.c b/bgpd/bgpd.c
index ba2985d304..22f48114d4 100644
--- a/bgpd/bgpd.c
+++ b/bgpd/bgpd.c
@@ -3361,6 +3361,8 @@ static struct bgp *bgp_create(as_t *as, const char *name,
 		bgp_label_per_nexthop_cache_init(
 			&bgp->mpls_labels_per_nexthop[afi]);
 
+	bgp_mplsvpn_nh_label_bind_cache_init(&bgp->mplsvpn_nh_label_bind);
+
 	if (name)
 		bgp->name = XSTRDUP(MTYPE_BGP, name);
 
diff --git a/bgpd/bgpd.h b/bgpd/bgpd.h
index 0eb58773d0..d7caf840fd 100644
--- a/bgpd/bgpd.h
+++ b/bgpd/bgpd.h
@@ -331,6 +331,9 @@ struct as_confed {
 	char *as_pretty;
 };
 
+struct bgp_mplsvpn_nh_label_bind_cache;
+PREDECL_RBTREE_UNIQ(bgp_mplsvpn_nh_label_bind_cache);
+
 /* BGP instance structure.  */
 struct bgp {
 	/* AS number of this BGP instance.  */
@@ -578,6 +581,9 @@ struct bgp {
 	struct bgp_label_per_nexthop_cache_head
 		mpls_labels_per_nexthop[AFI_MAX];
 
+	/* Tree for mplsvpn next-hop label bind cache */
+	struct bgp_mplsvpn_nh_label_bind_cache_head mplsvpn_nh_label_bind;
+
 	/* Allocate hash entries to store policy routing information
 	 * The hash are used to host pbr rules somewhere.
 	 * Actually, pbr will only be used by flowspec