diff --git a/bgpd/bgp_addpath.c b/bgpd/bgp_addpath.c
new file mode 100644
index 0000000000..22401f0017
--- /dev/null
+++ b/bgpd/bgp_addpath.c
@@ -0,0 +1,422 @@
+/*
+ * Addpath TX ID selection, and related utilities
+ * Copyright (C) 2018  Amazon.com, Inc. or its affiliates
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the Free
+ * Software Foundation; either version 2 of the License, or (at your option)
+ * any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; see the file COPYING; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "bgp_addpath.h"
+#include "bgp_route.h"
+
+static struct bgp_addpath_strategy_names strat_names[BGP_ADDPATH_MAX] = {
+	{
+		.config_name = "addpath-tx-all-paths",
+		.human_name = "All",
+		.human_description = "Advertise all paths via addpath",
+		.type_json_name = "addpathTxAllPaths",
+		.id_json_name = "addpathTxIdAll"
+	},
+	{
+		.config_name = "addpath-tx-bestpath-per-AS",
+		.human_name = "Best-Per-AS",
+		.human_description = "Advertise bestpath per AS via addpath",
+		.type_json_name = "addpathTxBestpathPerAS",
+		.id_json_name = "addpathTxIdBestPerAS"
+	}
+};
+
+static struct bgp_addpath_strategy_names unknown_names = {
+	.config_name = "addpath-tx-unknown",
+	.human_name = "Unknown-Addpath-Strategy",
+	.human_description = "Unknown Addpath Strategy",
+	.type_json_name = "addpathTxUnknown",
+	.id_json_name = "addpathTxIdUnknown"
+};
+
+/*
+ * Returns a structure full of strings associated with an addpath type. Will
+ * never return null.
+ */
+struct bgp_addpath_strategy_names *
+bgp_addpath_names(enum bgp_addpath_strat strat)
+{
+	if (strat < BGP_ADDPATH_MAX)
+		return &(strat_names[strat]);
+	else
+		return &unknown_names;
+};
+
+/*
+ * Returns if any peer is transmitting addpaths for a given afi/safi.
+ */
+int bgp_addpath_is_addpath_used(struct bgp_addpath_bgp_data *d, afi_t afi,
+			      safi_t safi)
+{
+	return d->total_peercount[afi][safi] > 0;
+}
+
+/*
+ * Initialize the BGP instance level data for addpath.
+ */
+void bgp_addpath_init_bgp_data(struct bgp_addpath_bgp_data *d)
+{
+	safi_t safi;
+	afi_t afi;
+	int i;
+
+	for (afi = AFI_IP; afi < AFI_MAX; afi++) {
+		for (safi = SAFI_UNICAST; safi < SAFI_MAX; safi++) {
+			for (i = 0; i < BGP_ADDPATH_MAX; i++) {
+				d->id_allocators[afi][safi][i] = NULL;
+				d->peercount[afi][safi][i] = 0;
+			}
+			d->total_peercount[afi][safi] = 0;
+		}
+	}
+}
+
+/*
+ * Free up resources associated with BGP route info structures.
+ */
+void bgp_addpath_free_info_data(struct bgp_addpath_info_data *d,
+			      struct bgp_addpath_node_data *nd)
+{
+	int i;
+
+	for (i = 0; i < BGP_ADDPATH_MAX; i++) {
+		if (d->addpath_tx_id[i] != IDALLOC_INVALID)
+			idalloc_free_to_pool(&nd->free_ids[i],
+					     d->addpath_tx_id[i]);
+	}
+}
+
+/*
+ * Return the addpath ID used to send a particular route, to a particular peer,
+ * in a particular AFI/SAFI.
+ */
+uint32_t bgp_addpath_id_for_peer(struct peer *peer, afi_t afi, safi_t safi,
+				struct bgp_addpath_info_data *d)
+{
+	if (peer->addpath_type[afi][safi] < BGP_ADDPATH_MAX)
+		return d->addpath_tx_id[peer->addpath_type[afi][safi]];
+	else
+		return IDALLOC_INVALID;
+}
+
+/*
+ * Returns true if the path has an assigned addpath ID for any of the addpath
+ * strategies.
+ */
+int bgp_addpath_info_has_ids(struct bgp_addpath_info_data *d)
+{
+	int i;
+
+	for (i = 0; i < BGP_ADDPATH_MAX; i++)
+		if (d->addpath_tx_id[i] != 0)
+			return 1;
+
+	return 0;
+}
+
+/*
+ * Releases any ID's associated with the BGP prefix.
+ */
+void bgp_addpath_free_node_data(struct bgp_addpath_bgp_data *bd,
+			      struct bgp_addpath_node_data *nd, afi_t afi,
+			      safi_t safi)
+{
+	int i;
+
+	for (i = 0; i < BGP_ADDPATH_MAX; i++) {
+		idalloc_drain_pool(bd->id_allocators[afi][safi][i],
+				   &(nd->free_ids[i]));
+	}
+}
+
+/*
+ * Check to see if the addpath strategy requires DMED to be configured to work.
+ */
+int bgp_addpath_dmed_required(int strategy)
+{
+	return strategy == BGP_ADDPATH_BEST_PER_AS;
+}
+
+/*
+ * Return true if this is a path we should advertise due to a
+ * configured addpath-tx knob
+ */
+int bgp_addpath_tx_path(enum bgp_addpath_strat strat,
+			    struct bgp_path_info *pi)
+{
+	switch (strat) {
+	case BGP_ADDPATH_NONE:
+		return 0;
+	case BGP_ADDPATH_ALL:
+		return 1;
+	case BGP_ADDPATH_BEST_PER_AS:
+		if (CHECK_FLAG(pi->flags, BGP_PATH_DMED_SELECTED))
+			return 1;
+		else
+			return 0;
+	default:
+		return 0;
+	}
+}
+
+/*
+ * Purge all addpath ID's on a BGP instance associated with the addpath
+ * strategy, and afi/safi combination. This lets us let go of all memory held to
+ * track ID numbers associated with an addpath type not in use. Since
+ * post-bestpath ID processing is skipped for types not used, this is the only
+ * chance to free this data.
+ */
+static void bgp_addpath_flush_type(struct bgp *bgp, afi_t afi, safi_t safi,
+				   enum bgp_addpath_strat addpath_type)
+{
+	struct bgp_node *rn;
+	struct bgp_path_info *pi;
+
+	for (rn = bgp_table_top(bgp->rib[afi][safi]); rn;
+	     rn = bgp_route_next(rn)) {
+		idalloc_drain_pool(
+			bgp->tx_addpath.id_allocators[afi][safi][addpath_type],
+			&(rn->tx_addpath.free_ids[addpath_type]));
+		for (pi = rn->info; pi; pi = pi->next) {
+			if (pi->tx_addpath.addpath_tx_id[addpath_type]
+			    != IDALLOC_INVALID) {
+				idalloc_free(
+					bgp->tx_addpath
+						.id_allocators[afi][safi]
+							      [addpath_type],
+					pi->tx_addpath
+						.addpath_tx_id[addpath_type]);
+				pi->tx_addpath.addpath_tx_id[addpath_type] =
+					IDALLOC_INVALID;
+			}
+		}
+	}
+
+	idalloc_destroy(bgp->tx_addpath.id_allocators[afi][safi][addpath_type]);
+	bgp->tx_addpath.id_allocators[afi][safi][addpath_type] = NULL;
+}
+
+/*
+ * Allocate an Addpath ID for the given type on a path, if necessary.
+ */
+static void bgp_addpath_populate_path(struct id_alloc *allocator,
+				      struct bgp_path_info *path,
+				      enum bgp_addpath_strat addpath_type)
+{
+	if (bgp_addpath_tx_path(addpath_type, path)) {
+		path->tx_addpath.addpath_tx_id[addpath_type] =
+			idalloc_allocate(allocator);
+	}
+}
+
+/*
+ * Compute addpath ID's on a BGP instance associated with the addpath strategy,
+ * and afi/safi combination. Since we won't waste the time computing addpath IDs
+ * for unused strategies, the first time a peer is configured to use a strategy,
+ * we have to backfill the data.
+ */
+static void bgp_addpath_populate_type(struct bgp *bgp, afi_t afi, safi_t safi,
+				    enum bgp_addpath_strat addpath_type)
+{
+	struct bgp_node *rn;
+	struct bgp_path_info *bi;
+	char buf[200];
+	struct id_alloc *allocator;
+
+	snprintf(buf, sizeof(buf), "Addpath ID Allocator %s:%d/%d",
+		 bgp_addpath_names(addpath_type)->config_name, (int)afi,
+		 (int)safi);
+	buf[sizeof(buf) - 1] = '\0';
+	zlog_info("Computing addpath IDs for addpath type %s",
+		bgp_addpath_names(addpath_type)->human_name);
+
+	bgp->tx_addpath.id_allocators[afi][safi][addpath_type] =
+		idalloc_new(buf);
+
+	idalloc_reserve(bgp->tx_addpath.id_allocators[afi][safi][addpath_type],
+		BGP_ADDPATH_TX_ID_FOR_DEFAULT_ORIGINATE);
+
+	allocator = bgp->tx_addpath.id_allocators[afi][safi][addpath_type];
+
+	for (rn = bgp_table_top(bgp->rib[afi][safi]); rn;
+	     rn = bgp_route_next(rn))
+		for (bi = rn->info; bi; bi = bi->next)
+			bgp_addpath_populate_path(allocator, bi, addpath_type);
+}
+
+/*
+ * Handle updates to a peer or group's addpath strategy. If after adjusting
+ * counts a addpath strategy is in use for the first time, or no longer in use,
+ * the IDs for that strategy will be populated or flushed.
+ */
+void bgp_addpath_type_changed(struct bgp *bgp)
+{
+	afi_t afi;
+	safi_t safi;
+	struct listnode *node, *nnode;
+	struct peer *peer;
+	int peer_count[AFI_MAX][SAFI_MAX][BGP_ADDPATH_MAX];
+	enum bgp_addpath_strat type;
+
+	FOREACH_AFI_SAFI(afi, safi) {
+		for (type=0; type<BGP_ADDPATH_MAX; type++) {
+			peer_count[afi][safi][type] = 0;
+		}
+	}
+
+	for (ALL_LIST_ELEMENTS(bgp->peer, node, nnode, peer)) {
+		FOREACH_AFI_SAFI(afi, safi) {
+			type = peer->addpath_type[afi][safi];
+			if (type != BGP_ADDPATH_NONE) {
+				peer_count[afi][safi][type] += 1;
+			}
+		}
+	}
+
+	FOREACH_AFI_SAFI(afi, safi) {
+		for (type=0; type<BGP_ADDPATH_MAX; type++) {
+			int old = bgp->tx_addpath.peercount[afi][safi][type];
+			int new = peer_count[afi][safi][type];
+
+			bgp->tx_addpath.peercount[afi][safi][type] = new;
+
+			if (old == 0 && new != 0) {
+				bgp_addpath_populate_type(bgp, afi, safi,
+					type);
+			} else if (old != 0 && new == 0) {
+				bgp_addpath_flush_type(bgp, afi, safi, type);
+			}
+		}
+	}
+}
+
+/*
+ * Change the addpath type assigned to a peer, or peer group. In addition to
+ * adjusting the counts, peer sessions will be reset as needed to make the
+ * change take effect.
+ */
+void bgp_addpath_set_peer_type(struct peer *peer, afi_t afi, safi_t safi,
+			      enum bgp_addpath_strat addpath_type)
+{
+	struct bgp *bgp = peer->bgp;
+	enum bgp_addpath_strat old_type = peer->addpath_type[afi][safi];
+	struct listnode *node, *nnode;
+	struct peer *tmp_peer;
+	struct peer_group *group;
+
+	if (addpath_type == old_type)
+		return;
+
+	if (addpath_type == BGP_ADDPATH_NONE && peer->group &&
+	    !CHECK_FLAG(peer->sflags, PEER_STATUS_GROUP)) {
+		/* A "no" config on a group member inherits group */
+		addpath_type = peer->group->conf->addpath_type[afi][safi];
+	}
+
+	peer->addpath_type[afi][safi] = addpath_type;
+
+	bgp_addpath_type_changed(bgp);
+
+	if (addpath_type != BGP_ADDPATH_NONE) {
+		if (bgp_addpath_dmed_required(addpath_type)) {
+			if (!bgp_flag_check(bgp, BGP_FLAG_DETERMINISTIC_MED)) {
+				zlog_warn(
+					"%s: enabling bgp deterministic-med, this is required for addpath-tx-bestpath-per-AS",
+					peer->host);
+				bgp_flag_set(bgp, BGP_FLAG_DETERMINISTIC_MED);
+				bgp_recalculate_all_bestpaths(bgp);
+			}
+		}
+	}
+
+	zlog_info("Resetting peer %s%s due to change in addpath config\n",
+		  CHECK_FLAG(peer->sflags, PEER_STATUS_GROUP) ? "group " : "",
+		  peer->host);
+
+	if (CHECK_FLAG(peer->sflags, PEER_STATUS_GROUP)) {
+		group = peer->group;
+
+		/* group will be null as peer_group_delete calls peer_delete on
+		 * group->conf. That peer_delete will eventuallly end up here
+		 * if the group was configured to tx addpaths.
+		 */
+		if (group != NULL) {
+			for (ALL_LIST_ELEMENTS(group->peer, node, nnode,
+			     tmp_peer)) {
+				if (tmp_peer->addpath_type[afi][safi] ==
+				    old_type) {
+					bgp_addpath_set_peer_type(tmp_peer,
+								 afi,
+								 safi,
+								 addpath_type);
+				}
+			}
+		}
+	} else {
+		peer_change_action(peer, afi, safi, peer_change_reset);
+	}
+
+}
+
+/*
+ * Intended to run after bestpath. This function will take TX IDs from paths
+ * that no longer need them, and give them to paths that do. This prevents
+ * best-per-as updates from needing to do a separate withdraw and update just to
+ * swap out which path is sent.
+ */
+void bgp_addpath_update_ids(struct bgp *bgp, struct bgp_node *bn, afi_t afi,
+			  safi_t safi)
+{
+	int i;
+	struct bgp_path_info *pi;
+	struct id_alloc_pool **pool_ptr;
+
+	for (i = 0; i < BGP_ADDPATH_MAX; i++) {
+		struct id_alloc *alloc =
+			bgp->tx_addpath.id_allocators[afi][safi][i];
+		pool_ptr = &(bn->tx_addpath.free_ids[i]);
+
+		if (bgp->tx_addpath.peercount[afi][safi][i] == 0)
+			continue;
+
+		/* Free Unused IDs back to the pool.*/
+		for (pi = bn->info; pi; pi = pi->next) {
+			if (pi->tx_addpath.addpath_tx_id[i] != IDALLOC_INVALID
+			    && !bgp_addpath_tx_path(i, pi)) {
+				idalloc_free_to_pool(pool_ptr,
+					pi->tx_addpath.addpath_tx_id[i]);
+				pi->tx_addpath.addpath_tx_id[i] =
+					IDALLOC_INVALID;
+			}
+		}
+
+		/* Give IDs to paths that need them (pulling from the pool) */
+		for (pi = bn->info; pi; pi = pi->next) {
+			if (pi->tx_addpath.addpath_tx_id[i] == IDALLOC_INVALID
+			    && bgp_addpath_tx_path(i, pi)) {
+				pi->tx_addpath.addpath_tx_id[i] =
+					idalloc_allocate_prefer_pool(
+						alloc, pool_ptr);
+			}
+		}
+
+		/* Free any IDs left in the pool to the main allocator */
+		idalloc_drain_pool(alloc, pool_ptr);
+	}
+}
diff --git a/bgpd/bgp_addpath.h b/bgpd/bgp_addpath.h
new file mode 100644
index 0000000000..c0c182791b
--- /dev/null
+++ b/bgpd/bgp_addpath.h
@@ -0,0 +1,72 @@
+/*
+ * Addpath TX ID selection, and related utilities
+ * Copyright (C) 2018  Amazon.com, Inc. or its affiliates
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the Free
+ * Software Foundation; either version 2 of the License, or (at your option)
+ * any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; see the file COPYING; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef _QUAGGA_BGPD_TX_ADDPATH_H
+#define _QUAGGA_BGPD_TX_ADDPATH_H
+
+#include <stdint.h>
+#include <zebra.h>
+
+#include "bgpd/bgp_addpath_types.h"
+#include "bgpd/bgp_route.h"
+#include "bgpd/bgp_table.h"
+#include "lib/json.h"
+
+#define BGP_ADDPATH_TX_ID_FOR_DEFAULT_ORIGINATE 1
+
+void bgp_addpath_init_bgp_data(struct bgp_addpath_bgp_data *d);
+
+int bgp_addpath_is_addpath_used(struct bgp_addpath_bgp_data *d, afi_t afi,
+			      safi_t safi);
+
+void bgp_addpath_free_node_data(struct bgp_addpath_bgp_data *bd,
+			      struct bgp_addpath_node_data *nd,
+			      afi_t afi, safi_t safi);
+
+void bgp_addpath_free_info_data(struct bgp_addpath_info_data *d,
+			      struct bgp_addpath_node_data *nd);
+
+
+int bgp_addpath_info_has_ids(struct bgp_addpath_info_data *d);
+
+uint32_t bgp_addpath_id_for_peer(struct peer *peer, afi_t afi, safi_t safi,
+				struct bgp_addpath_info_data *d);
+
+struct bgp_addpath_strategy_names *
+bgp_addpath_names(enum bgp_addpath_strat strat);
+
+int bgp_addpath_dmed_required(int strategy);
+
+/*
+ * Return true if this is a path we should advertise due to a configured
+ * addpath-tx knob
+ */
+int bgp_addpath_tx_path(enum bgp_addpath_strat strat,
+			    struct bgp_path_info *pi);
+/*
+ * Change the type of addpath used for a peer.
+ */
+void bgp_addpath_set_peer_type(struct peer *peer, afi_t afi, safi_t safi,
+			      enum bgp_addpath_strat addpath_type);
+
+void bgp_addpath_update_ids(struct bgp *bgp, struct bgp_node *bn, afi_t afi,
+			    safi_t safi);
+
+void bgp_addpath_type_changed(struct bgp *bgp);
+#endif
diff --git a/bgpd/bgp_addpath_types.h b/bgpd/bgp_addpath_types.h
new file mode 100644
index 0000000000..b0b3302979
--- /dev/null
+++ b/bgpd/bgp_addpath_types.h
@@ -0,0 +1,55 @@
+/*
+ * Addpath TX ID selection, and related utilities
+ * Copyright (C) 2018  Amazon.com, Inc. or its affiliates
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the Free
+ * Software Foundation; either version 2 of the License, or (at your option)
+ * any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; see the file COPYING; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef _QUAGGA_BGPD_TX_ADDPATH_DATA_H
+#define _QUAGGA_BGPD_TX_ADDPATH_DATA_H
+#include "lib/id_alloc.h"
+#include <stdint.h>
+
+enum bgp_addpath_strat {
+	BGP_ADDPATH_ALL = 0,
+	BGP_ADDPATH_BEST_PER_AS,
+	BGP_ADDPATH_MAX,
+	BGP_ADDPATH_NONE,
+};
+
+/* TX Addpath structures */
+struct bgp_addpath_bgp_data {
+	unsigned int peercount[AFI_MAX][SAFI_MAX][BGP_ADDPATH_MAX];
+	unsigned int total_peercount[AFI_MAX][SAFI_MAX];
+	struct id_alloc *id_allocators[AFI_MAX][SAFI_MAX][BGP_ADDPATH_MAX];
+};
+
+struct bgp_addpath_node_data {
+	struct id_alloc_pool *free_ids[BGP_ADDPATH_MAX];
+};
+
+struct bgp_addpath_info_data {
+	uint32_t addpath_tx_id[BGP_ADDPATH_MAX];
+};
+
+struct bgp_addpath_strategy_names {
+	const char *config_name;
+	const char *human_name;	       /* path detail non-json */
+	const char *human_description; /* non-json peer descriptions */
+	const char *type_json_name;    /* json peer listings */
+	const char *id_json_name;      /* path json output for tx ID# */
+};
+
+#endif
diff --git a/bgpd/bgp_evpn.c b/bgpd/bgp_evpn.c
index cff050a9ef..fc3ac28723 100644
--- a/bgpd/bgp_evpn.c
+++ b/bgpd/bgp_evpn.c
@@ -47,6 +47,7 @@
 #include "bgpd/bgp_aspath.h"
 #include "bgpd/bgp_zebra.h"
 #include "bgpd/bgp_nexthop.h"
+#include "bgpd/bgp_addpath.h"
 
 /*
  * Definitions and external declarations.
@@ -1059,7 +1060,7 @@ static int evpn_es_route_select_install(struct bgp *bgp,
 	    && old_select->sub_type == BGP_ROUTE_IMPORTED
 	    && !CHECK_FLAG(rn->flags, BGP_NODE_USER_CLEAR)
 	    && !CHECK_FLAG(old_select->flags, BGP_PATH_ATTR_CHANGED)
-	    && !bgp->addpath_tx_used[afi][safi]) {
+	    && !bgp_addpath_is_addpath_used(&bgp->tx_addpath, afi, safi)) {
 		if (bgp_zebra_has_route_changed(rn, old_select)) {
 			ret = evpn_es_install_vtep(bgp, es,
 						   (struct prefix_evpn *)&rn->p,
@@ -1142,7 +1143,7 @@ static int evpn_route_select_install(struct bgp *bgp, struct bgpevpn *vpn,
 	    && old_select->sub_type == BGP_ROUTE_IMPORTED
 	    && !CHECK_FLAG(rn->flags, BGP_NODE_USER_CLEAR)
 	    && !CHECK_FLAG(old_select->flags, BGP_PATH_ATTR_CHANGED)
-	    && !bgp->addpath_tx_used[afi][safi]) {
+	    && !bgp_addpath_is_addpath_used(&bgp->tx_addpath, afi, safi)) {
 		if (bgp_zebra_has_route_changed(rn, old_select)) {
 			if (old_select->attr->sticky)
 				SET_FLAG(flags, ZEBRA_MACIP_TYPE_STICKY);
diff --git a/bgpd/bgp_open.c b/bgpd/bgp_open.c
index cf5901df5a..3018124f45 100644
--- a/bgpd/bgp_open.c
+++ b/bgpd/bgp_open.c
@@ -1387,10 +1387,7 @@ void bgp_open_capability(struct stream *s, struct peer *peer)
 			/* Only advertise addpath TX if a feature that
 			 * will use it is
 			 * configured */
-			if (CHECK_FLAG(peer->af_flags[afi][safi],
-				       PEER_FLAG_ADDPATH_TX_ALL_PATHS)
-			    || CHECK_FLAG(peer->af_flags[afi][safi],
-					  PEER_FLAG_ADDPATH_TX_BESTPATH_PER_AS))
+			if (peer->addpath_type[afi][safi] != BGP_ADDPATH_NONE)
 				adv_addpath_tx = 1;
 		}
 	}
diff --git a/bgpd/bgp_route.c b/bgpd/bgp_route.c
index 8ae74a008c..bb8c5d8f6e 100644
--- a/bgpd/bgp_route.c
+++ b/bgpd/bgp_route.c
@@ -65,6 +65,7 @@
 #include "bgpd/bgp_nht.h"
 #include "bgpd/bgp_updgrp.h"
 #include "bgpd/bgp_label.h"
+#include "bgpd/bgp_addpath.h"
 
 #if ENABLE_BGP_VNC
 #include "bgpd/rfapi/rfapi_backend.h"
@@ -247,6 +248,8 @@ static void bgp_path_info_free(struct bgp_path_info *path)
 	bgp_unlink_nexthop(path);
 	bgp_path_info_extra_free(&path->extra);
 	bgp_path_info_mpath_free(&path->mpath);
+	bgp_addpath_free_info_data(&path->tx_addpath,
+				 path->net ? &path->net->tx_addpath : NULL);
 
 	peer_unlock(path->peer); /* bgp_path_info peer reference */
 
@@ -1472,7 +1475,7 @@ int subgroup_announce_check(struct bgp_node *rn, struct bgp_path_info *pi,
 	 * addpath
 	 * feature that requires us to advertise it */
 	if (!CHECK_FLAG(pi->flags, BGP_PATH_SELECTED)) {
-		if (!bgp_addpath_tx_path(peer, afi, safi, pi)) {
+		if (!bgp_addpath_tx_path(peer->addpath_type[afi][safi], pi)) {
 			return 0;
 		}
 	}
@@ -2078,6 +2081,8 @@ void bgp_best_selection(struct bgp *bgp, struct bgp_node *rn,
 	bgp_path_info_mpath_aggregate_update(new_select, old_select);
 	bgp_mp_list_clear(&mp_list);
 
+	bgp_addpath_update_ids(bgp, rn, afi, safi);
+
 	result->old = old_select;
 	result->new = new_select;
 
@@ -2127,7 +2132,7 @@ int subgroup_process_announce_selected(struct update_subgroup *subgrp,
 			bgp_adj_out_set_subgroup(rn, subgrp, &attr, selected);
 		else
 			bgp_adj_out_unset_subgroup(rn, subgrp, 1,
-						   selected->addpath_tx_id);
+						   addpath_tx_id);
 	}
 
 	/* If selected is NULL we must withdraw the path using addpath_tx_id */
@@ -2303,7 +2308,7 @@ static void bgp_process_main_one(struct bgp *bgp, struct bgp_node *rn,
 	if (old_select && old_select == new_select
 	    && !CHECK_FLAG(rn->flags, BGP_NODE_USER_CLEAR)
 	    && !CHECK_FLAG(old_select->flags, BGP_PATH_ATTR_CHANGED)
-	    && !bgp->addpath_tx_used[afi][safi]) {
+	    && !bgp_addpath_is_addpath_used(&bgp->tx_addpath, afi, safi)) {
 		if (bgp_zebra_has_route_changed(rn, old_select)) {
 #if ENABLE_BGP_VNC
 			vnc_import_bgp_add_route(bgp, p, old_select);
@@ -2776,7 +2781,6 @@ struct bgp_path_info *info_make(int type, int sub_type, unsigned short instance,
 	new->attr = attr;
 	new->uptime = bgp_clock();
 	new->net = rn;
-	new->addpath_tx_id = ++peer->bgp->addpath_tx_id;
 	return new;
 }
 
@@ -7486,6 +7490,18 @@ static void route_vty_out_advertised_to(struct vty *vty, struct peer *peer,
 	}
 }
 
+static void route_vty_out_tx_ids(struct vty *vty,
+				 struct bgp_addpath_info_data *d)
+{
+	int i;
+
+	for (i = 0; i < BGP_ADDPATH_MAX; i++) {
+		vty_out(vty, "TX-%s %u%s", bgp_addpath_names(i)->human_name,
+			d->addpath_tx_id[i],
+			i < BGP_ADDPATH_MAX - 1 ? " " : "\n");
+	}
+}
+
 void route_vty_out_detail(struct vty *vty, struct bgp *bgp, struct prefix *p,
 			  struct bgp_path_info *path, afi_t afi, safi_t safi,
 			  json_object *json_paths)
@@ -7517,6 +7533,7 @@ void route_vty_out_detail(struct vty *vty, struct bgp *bgp, struct prefix *p,
 	unsigned int first_as;
 	bool nexthop_self =
 		CHECK_FLAG(path->flags, BGP_PATH_ANNC_NH_SELF) ? true : false;
+	int i;
 
 	if (json_paths) {
 		json_path = json_object_new_object();
@@ -8228,29 +8245,53 @@ void route_vty_out_detail(struct vty *vty, struct bgp *bgp, struct prefix *p,
 		}
 
 		/* Line 8 display Addpath IDs */
-		if (path->addpath_rx_id || path->addpath_tx_id) {
+		if (path->addpath_rx_id
+		    || bgp_addpath_info_has_ids(&path->tx_addpath)) {
 			if (json_paths) {
 				json_object_int_add(json_path, "addpathRxId",
 						    path->addpath_rx_id);
-				json_object_int_add(json_path, "addpathTxId",
-						    path->addpath_tx_id);
+
+				/* Keep backwards compatibility with the old API
+				 * by putting TX All's ID in the old field
+				 */
+				json_object_int_add(
+					json_path, "addpathTxId",
+					path->tx_addpath.addpath_tx_id
+						[BGP_ADDPATH_ALL]);
+
+				/* ... but create a specific field for each
+				 * strategy
+				 */
+				for (i = 0; i < BGP_ADDPATH_MAX; i++) {
+					json_object_int_add(
+						json_path,
+						bgp_addpath_names(i)
+							->id_json_name,
+						path->tx_addpath
+							.addpath_tx_id[i]);
+				}
 			} else {
-				vty_out(vty, "      AddPath ID: RX %u, TX %u\n",
-					path->addpath_rx_id,
-					path->addpath_tx_id);
+				vty_out(vty, "      AddPath ID: RX %u, ",
+					path->addpath_rx_id);
+
+				route_vty_out_tx_ids(vty, &path->tx_addpath);
 			}
 		}
 
 		/* If we used addpath to TX a non-bestpath we need to display
-		 * "Advertised to" on a path-by-path basis */
-		if (bgp->addpath_tx_used[afi][safi]) {
+		 * "Advertised to" on a path-by-path basis
+		 */
+		if (bgp_addpath_is_addpath_used(&bgp->tx_addpath, afi, safi)) {
 			first = 1;
 
 			for (ALL_LIST_ELEMENTS(bgp->peer, node, nnode, peer)) {
 				addpath_capable =
 					bgp_addpath_encode_tx(peer, afi, safi);
 				has_adj = bgp_adj_out_lookup(
-					peer, path->net, path->addpath_tx_id);
+					peer, path->net,
+					bgp_addpath_id_for_peer(
+						peer, afi, safi,
+						&path->tx_addpath));
 
 				if ((addpath_capable && has_adj)
 				    || (!addpath_capable && has_adj
@@ -8958,7 +8999,7 @@ void route_vty_out_detail_header(struct vty *vty, struct bgp *bgp,
 	 * show what peers we advertised the bestpath to.  If we are using
 	 * addpath
 	 * though then we must display Advertised to on a path-by-path basis. */
-	if (!bgp->addpath_tx_used[afi][safi]) {
+	if (!bgp_addpath_is_addpath_used(&bgp->tx_addpath, afi, safi)) {
 		for (ALL_LIST_ELEMENTS(bgp->peer, node, nnode, peer)) {
 			if (bgp_adj_out_lookup(peer, rn, 0)) {
 				if (json && !json_adv_to)
diff --git a/bgpd/bgp_route.h b/bgpd/bgp_route.h
index f0edc8d49a..466d52ed64 100644
--- a/bgpd/bgp_route.h
+++ b/bgpd/bgp_route.h
@@ -24,6 +24,7 @@
 #include "queue.h"
 #include "nexthop.h"
 #include "bgp_table.h"
+#include "bgp_addpath_types.h"
 
 struct bgp_nexthop_cache;
 struct bgp_route_evpn;
@@ -220,7 +221,7 @@ struct bgp_path_info {
 
 	/* Addpath identifiers */
 	uint32_t addpath_rx_id;
-	uint32_t addpath_tx_id;
+	struct bgp_addpath_info_data tx_addpath;
 };
 
 /* Structure used in BGP path selection */
diff --git a/bgpd/bgp_table.c b/bgpd/bgp_table.c
index 231f326cef..728eeaa3a9 100644
--- a/bgpd/bgp_table.c
+++ b/bgpd/bgp_table.c
@@ -29,6 +29,7 @@
 
 #include "bgpd/bgpd.h"
 #include "bgpd/bgp_table.h"
+#include "bgp_addpath.h"
 
 void bgp_table_lock(struct bgp_table *rt)
 {
@@ -76,7 +77,16 @@ static void bgp_node_destroy(route_table_delegate_t *delegate,
 			     struct route_table *table, struct route_node *node)
 {
 	struct bgp_node *bgp_node;
+	struct bgp_table *rt;
 	bgp_node = bgp_node_from_rnode(node);
+	rt = table->info;
+
+	if (rt->bgp) {
+		bgp_addpath_free_node_data(&rt->bgp->tx_addpath,
+					 &bgp_node->tx_addpath,
+					 rt->afi, rt->safi);
+	}
+
 	XFREE(MTYPE_BGP_NODE, bgp_node);
 }
 
diff --git a/bgpd/bgp_table.h b/bgpd/bgp_table.h
index 60a26b76cc..c267b4fe8a 100644
--- a/bgpd/bgp_table.h
+++ b/bgpd/bgp_table.h
@@ -25,6 +25,7 @@
 #include "table.h"
 #include "queue.h"
 #include "linklist.h"
+#include "bgpd.h"
 
 struct bgp_table {
 	/* table belongs to this instance */
@@ -67,6 +68,8 @@ struct bgp_node {
 #define BGP_NODE_USER_CLEAR             (1 << 1)
 #define BGP_NODE_LABEL_CHANGED          (1 << 2)
 #define BGP_NODE_REGISTERED_FOR_LABEL   (1 << 3)
+
+	struct bgp_addpath_node_data tx_addpath;
 };
 
 /*
diff --git a/bgpd/bgp_updgrp.c b/bgpd/bgp_updgrp.c
index 393586dbec..2e2ad88314 100644
--- a/bgpd/bgp_updgrp.c
+++ b/bgpd/bgp_updgrp.c
@@ -152,6 +152,7 @@ static void conf_copy(struct peer *dst, struct peer *src, afi_t afi,
 	dst->af_cap[afi][safi] = src->af_cap[afi][safi];
 	dst->afc_nego[afi][safi] = src->afc_nego[afi][safi];
 	dst->orf_plist[afi][safi] = src->orf_plist[afi][safi];
+	dst->addpath_type[afi][safi] = src->addpath_type[afi][safi];
 	dst->local_as = src->local_as;
 	dst->change_local_as = src->change_local_as;
 	dst->shared_network = src->shared_network;
@@ -322,6 +323,7 @@ static unsigned int updgrp_hash_key_make(void *p)
 	key = jhash_1word(peer->sort, key); /* EBGP or IBGP */
 	key = jhash_1word((peer->flags & PEER_UPDGRP_FLAGS), key);
 	key = jhash_1word((flags & PEER_UPDGRP_AF_FLAGS), key);
+	key = jhash_1word((uint32_t)peer->addpath_type[afi][safi], key);
 	key = jhash_1word((peer->cap & PEER_UPDGRP_CAP_FLAGS), key);
 	key = jhash_1word((peer->af_cap[afi][safi] & PEER_UPDGRP_AF_CAP_FLAGS),
 			  key);
@@ -437,6 +439,9 @@ static bool updgrp_hash_cmp(const void *p1, const void *p2)
 	if ((flags1 & PEER_UPDGRP_AF_FLAGS) != (flags2 & PEER_UPDGRP_AF_FLAGS))
 		return false;
 
+	if (pe1->addpath_type[afi][safi] != pe2->addpath_type[afi][safi])
+		return 0;
+
 	if ((pe1->cap & PEER_UPDGRP_CAP_FLAGS)
 	    != (pe2->cap & PEER_UPDGRP_CAP_FLAGS))
 		return false;
@@ -1900,22 +1905,3 @@ int bgp_addpath_encode_tx(struct peer *peer, afi_t afi, safi_t safi)
 		&& CHECK_FLAG(peer->af_cap[afi][safi],
 			      PEER_CAP_ADDPATH_AF_RX_RCV));
 }
-
-/*
- * Return true if this is a path we should advertise due to a
- * configured addpath-tx knob
- */
-int bgp_addpath_tx_path(struct peer *peer, afi_t afi, safi_t safi,
-			struct bgp_path_info *pi)
-{
-	if (CHECK_FLAG(peer->af_flags[afi][safi],
-		       PEER_FLAG_ADDPATH_TX_ALL_PATHS))
-		return 1;
-
-	if (CHECK_FLAG(peer->af_flags[afi][safi],
-		       PEER_FLAG_ADDPATH_TX_BESTPATH_PER_AS)
-	    && CHECK_FLAG(pi->flags, BGP_PATH_DMED_SELECTED))
-		return 1;
-
-	return 0;
-}
diff --git a/bgpd/bgp_updgrp.h b/bgpd/bgp_updgrp.h
index b1d852d384..6b3bf9d1f7 100644
--- a/bgpd/bgp_updgrp.h
+++ b/bgpd/bgp_updgrp.h
@@ -64,8 +64,7 @@
 	 | PEER_FLAG_REMOVE_PRIVATE_AS_ALL                                     \
 	 | PEER_FLAG_REMOVE_PRIVATE_AS_REPLACE                                 \
 	 | PEER_FLAG_REMOVE_PRIVATE_AS_ALL_REPLACE                             \
-	 | PEER_FLAG_ADDPATH_TX_ALL_PATHS                                      \
-	 | PEER_FLAG_ADDPATH_TX_BESTPATH_PER_AS | PEER_FLAG_AS_OVERRIDE)
+	 | PEER_FLAG_AS_OVERRIDE)
 
 #define PEER_UPDGRP_CAP_FLAGS (PEER_CAP_AS4_RCV)
 
@@ -469,8 +468,6 @@ extern int update_group_clear_update_dbg(struct update_group *updgrp,
 
 extern void update_bgp_group_free(struct bgp *bgp);
 extern int bgp_addpath_encode_tx(struct peer *peer, afi_t afi, safi_t safi);
-extern int bgp_addpath_tx_path(struct peer *peer, afi_t afi, safi_t safi,
-			       struct bgp_path_info *pi);
 
 /*
  * Inline functions
diff --git a/bgpd/bgp_updgrp_adv.c b/bgpd/bgp_updgrp_adv.c
index 81b782c2b0..7196bbbf12 100644
--- a/bgpd/bgp_updgrp_adv.c
+++ b/bgpd/bgp_updgrp_adv.c
@@ -49,6 +49,7 @@
 #include "bgpd/bgp_mplsvpn.h"
 #include "bgpd/bgp_updgrp.h"
 #include "bgpd/bgp_advertise.h"
+#include "bgpd/bgp_addpath.h"
 
 
 /********************
@@ -97,6 +98,40 @@ static void adj_free(struct bgp_adj_out *adj)
 	XFREE(MTYPE_BGP_ADJ_OUT, adj);
 }
 
+static void subgrp_withdraw_stale_addpath(struct updwalk_context *ctx,
+					  struct update_subgroup *subgrp)
+{
+	struct bgp_adj_out *adj, *adj_next;
+	uint32_t id;
+	struct bgp_path_info *pi;
+	afi_t afi = SUBGRP_AFI(subgrp);
+	safi_t safi = SUBGRP_SAFI(subgrp);
+	struct peer *peer = SUBGRP_PEER(subgrp);
+
+	/* Look through all of the paths we have advertised for this rn and send
+	 * a withdraw for the ones that are no longer present */
+	for (adj = ctx->rn->adj_out; adj; adj = adj_next) {
+		adj_next = adj->next;
+
+		if (adj->subgroup == subgrp) {
+			for (pi = ctx->rn->info; pi; pi = pi->next) {
+				id = bgp_addpath_id_for_peer(peer, afi, safi,
+					&pi->tx_addpath);
+
+				if (id == adj->addpath_tx_id) {
+					break;
+				}
+			}
+
+			if (!pi) {
+				subgroup_process_announce_selected(
+					subgrp, NULL, ctx->rn,
+					adj->addpath_tx_id);
+			}
+		}
+	}
+}
+
 static int group_announce_route_walkcb(struct update_group *updgrp, void *arg)
 {
 	struct updwalk_context *ctx = arg;
@@ -131,31 +166,7 @@ static int group_announce_route_walkcb(struct update_group *updgrp, void *arg)
 		if (!subgrp->t_coalesce) {
 			/* An update-group that uses addpath */
 			if (addpath_capable) {
-				/* Look through all of the paths we have
-				 * advertised for this rn and
-				 * send a withdraw for the ones that are no
-				 * longer present */
-				for (adj = ctx->rn->adj_out; adj;
-				     adj = adj_next) {
-					adj_next = adj->next;
-
-					if (adj->subgroup == subgrp) {
-						for (pi = ctx->rn->info; pi;
-						     pi = pi->next) {
-							if (pi->addpath_tx_id
-							    == adj->addpath_tx_id) {
-								break;
-							}
-						}
-
-						if (!pi) {
-							subgroup_process_announce_selected(
-								subgrp, NULL,
-								ctx->rn,
-								adj->addpath_tx_id);
-						}
-					}
-				}
+				subgrp_withdraw_stale_addpath(ctx, subgrp);
 
 				for (pi = ctx->rn->info; pi; pi = pi->next) {
 					/* Skip the bestpath for now */
@@ -164,7 +175,9 @@ static int group_announce_route_walkcb(struct update_group *updgrp, void *arg)
 
 					subgroup_process_announce_selected(
 						subgrp, pi, ctx->rn,
-						pi->addpath_tx_id);
+						bgp_addpath_id_for_peer(
+							peer, afi, safi,
+							&pi->tx_addpath));
 				}
 
 				/* Process the bestpath last so the "show [ip]
@@ -174,7 +187,9 @@ static int group_announce_route_walkcb(struct update_group *updgrp, void *arg)
 				if (ctx->pi)
 					subgroup_process_announce_selected(
 						subgrp, ctx->pi, ctx->rn,
-						ctx->pi->addpath_tx_id);
+						bgp_addpath_id_for_peer(
+							peer, afi, safi,
+							&ctx->pi->tx_addpath));
 			}
 
 			/* An update-group that does not use addpath */
@@ -182,7 +197,9 @@ static int group_announce_route_walkcb(struct update_group *updgrp, void *arg)
 				if (ctx->pi) {
 					subgroup_process_announce_selected(
 						subgrp, ctx->pi, ctx->rn,
-						ctx->pi->addpath_tx_id);
+						bgp_addpath_id_for_peer(
+							peer, afi, safi,
+							&ctx->pi->tx_addpath));
 				} else {
 					/* Find the addpath_tx_id of the path we
 					 * had advertised and
@@ -433,15 +450,27 @@ void bgp_adj_out_set_subgroup(struct bgp_node *rn,
 {
 	struct bgp_adj_out *adj = NULL;
 	struct bgp_advertise *adv;
+	struct peer *peer;
+	afi_t afi;
+	safi_t safi;
+
+	peer = SUBGRP_PEER(subgrp);
+	afi = SUBGRP_AFI(subgrp);
+	safi = SUBGRP_SAFI(subgrp);
 
 	if (DISABLE_BGP_ANNOUNCE)
 		return;
 
 	/* Look for adjacency information. */
-	adj = adj_lookup(rn, subgrp, path->addpath_tx_id);
+	adj = adj_lookup(
+		rn, subgrp,
+		bgp_addpath_id_for_peer(peer, afi, safi, &path->tx_addpath));
 
 	if (!adj) {
-		adj = bgp_adj_out_alloc(subgrp, rn, path->addpath_tx_id);
+		adj = bgp_adj_out_alloc(
+			subgrp, rn,
+			bgp_addpath_id_for_peer(peer, afi, safi,
+					      &path->tx_addpath));
 		if (!adj)
 			return;
 	}
@@ -597,7 +626,9 @@ void subgroup_announce_table(struct update_subgroup *subgrp,
 
 			if (CHECK_FLAG(ri->flags, BGP_PATH_SELECTED)
 			    || (addpath_capable
-				&& bgp_addpath_tx_path(peer, afi, safi, ri))) {
+				&& bgp_addpath_tx_path(
+					   peer->addpath_type[afi][safi],
+					   ri))) {
 				if (subgroup_announce_check(rn, ri, subgrp,
 							    &rn->p, &attr))
 					bgp_adj_out_set_subgroup(rn, subgrp,
@@ -605,7 +636,9 @@ void subgroup_announce_table(struct update_subgroup *subgrp,
 				else
 					bgp_adj_out_unset_subgroup(
 						rn, subgrp, 1,
-						ri->addpath_tx_id);
+						bgp_addpath_id_for_peer(
+							peer, afi, safi,
+							&ri->tx_addpath));
 			}
 
 	/*
diff --git a/bgpd/bgp_updgrp_packet.c b/bgpd/bgp_updgrp_packet.c
index c4fcea948f..ca6bb5ab82 100644
--- a/bgpd/bgp_updgrp_packet.c
+++ b/bgpd/bgp_updgrp_packet.c
@@ -56,6 +56,7 @@
 #include "bgpd/bgp_nht.h"
 #include "bgpd/bgp_mplsvpn.h"
 #include "bgpd/bgp_label.h"
+#include "bgpd/bgp_addpath.h"
 
 /********************
  * PRIVATE FUNCTIONS
diff --git a/bgpd/bgp_vty.c b/bgpd/bgp_vty.c
index c57cd38151..4dd148bb41 100644
--- a/bgpd/bgp_vty.c
+++ b/bgpd/bgp_vty.c
@@ -61,6 +61,7 @@
 #include "bgpd/bgp_bfd.h"
 #include "bgpd/bgp_io.h"
 #include "bgpd/bgp_evpn.h"
+#include "bgpd/bgp_addpath.h"
 
 static struct peer_group *listen_range_exists(struct bgp *bgp,
 					      struct prefix *range, int exact);
@@ -1888,9 +1889,8 @@ DEFUN (no_bgp_deterministic_med,
 
 		for (ALL_LIST_ELEMENTS(bgp->peer, node, nnode, peer)) {
 			FOREACH_AFI_SAFI (afi, safi)
-				if (CHECK_FLAG(
-					    peer->af_flags[afi][safi],
-					    PEER_FLAG_ADDPATH_TX_BESTPATH_PER_AS)) {
+				if (bgp_addpath_dmed_required(
+					peer->addpath_type[afi][safi])) {
 					bestpath_per_as_used = 1;
 					break;
 				}
@@ -6234,9 +6234,9 @@ DEFUN (neighbor_addpath_tx_all_paths,
 	if (!peer)
 		return CMD_WARNING_CONFIG_FAILED;
 
-	return peer_af_flag_set_vty(vty, argv[idx_peer]->arg, bgp_node_afi(vty),
-				    bgp_node_safi(vty),
-				    PEER_FLAG_ADDPATH_TX_ALL_PATHS);
+	bgp_addpath_set_peer_type(peer, bgp_node_afi(vty), bgp_node_safi(vty),
+				 BGP_ADDPATH_ALL);
+	return CMD_SUCCESS;
 }
 
 ALIAS_HIDDEN(neighbor_addpath_tx_all_paths,
@@ -6254,9 +6254,23 @@ DEFUN (no_neighbor_addpath_tx_all_paths,
        "Use addpath to advertise all paths to a neighbor\n")
 {
 	int idx_peer = 2;
-	return peer_af_flag_unset_vty(vty, argv[idx_peer]->arg,
-				      bgp_node_afi(vty), bgp_node_safi(vty),
-				      PEER_FLAG_ADDPATH_TX_ALL_PATHS);
+	struct peer *peer;
+
+	peer = peer_and_group_lookup_vty(vty, argv[idx_peer]->arg);
+	if (!peer)
+		return CMD_WARNING_CONFIG_FAILED;
+
+	if (peer->addpath_type[bgp_node_afi(vty)][bgp_node_safi(vty)]
+	    != BGP_ADDPATH_ALL) {
+		vty_out(vty,
+			"%% Peer not currently configured to transmit all paths.");
+		return CMD_WARNING_CONFIG_FAILED;
+	}
+
+	bgp_addpath_set_peer_type(peer, bgp_node_afi(vty), bgp_node_safi(vty),
+				 BGP_ADDPATH_NONE);
+
+	return CMD_SUCCESS;
 }
 
 ALIAS_HIDDEN(no_neighbor_addpath_tx_all_paths,
@@ -6279,9 +6293,10 @@ DEFUN (neighbor_addpath_tx_bestpath_per_as,
 	if (!peer)
 		return CMD_WARNING_CONFIG_FAILED;
 
-	return peer_af_flag_set_vty(vty, argv[idx_peer]->arg, bgp_node_afi(vty),
-				    bgp_node_safi(vty),
-				    PEER_FLAG_ADDPATH_TX_BESTPATH_PER_AS);
+	bgp_addpath_set_peer_type(peer, bgp_node_afi(vty), bgp_node_safi(vty),
+				 BGP_ADDPATH_BEST_PER_AS);
+
+	return CMD_SUCCESS;
 }
 
 ALIAS_HIDDEN(neighbor_addpath_tx_bestpath_per_as,
@@ -6299,9 +6314,23 @@ DEFUN (no_neighbor_addpath_tx_bestpath_per_as,
        "Use addpath to advertise the bestpath per each neighboring AS\n")
 {
 	int idx_peer = 2;
-	return peer_af_flag_unset_vty(vty, argv[idx_peer]->arg,
-				      bgp_node_afi(vty), bgp_node_safi(vty),
-				      PEER_FLAG_ADDPATH_TX_BESTPATH_PER_AS);
+	struct peer *peer;
+
+	peer = peer_and_group_lookup_vty(vty, argv[idx_peer]->arg);
+	if (!peer)
+		return CMD_WARNING_CONFIG_FAILED;
+
+	if (peer->addpath_type[bgp_node_afi(vty)][bgp_node_safi(vty)]
+	    != BGP_ADDPATH_BEST_PER_AS) {
+		vty_out(vty,
+			"%% Peer not currently configured to transmit all best path per as.");
+		return CMD_WARNING_CONFIG_FAILED;
+	}
+
+	bgp_addpath_set_peer_type(peer, bgp_node_afi(vty), bgp_node_safi(vty),
+				 BGP_ADDPATH_NONE);
+
+	return CMD_SUCCESS;
 }
 
 ALIAS_HIDDEN(no_neighbor_addpath_tx_bestpath_per_as,
@@ -8656,15 +8685,11 @@ static void bgp_show_peer_afi(struct vty *vty, struct peer *p, afi_t afi,
 				json_addr,
 				"privateAsNumsRemovedInUpdatesToNbr");
 
-		if (CHECK_FLAG(p->af_flags[afi][safi],
-			       PEER_FLAG_ADDPATH_TX_ALL_PATHS))
-			json_object_boolean_true_add(json_addr,
-						     "addpathTxAllPaths");
-
-		if (CHECK_FLAG(p->af_flags[afi][safi],
-			       PEER_FLAG_ADDPATH_TX_BESTPATH_PER_AS))
-			json_object_boolean_true_add(json_addr,
-						     "addpathTxBestpathPerAS");
+		if (p->addpath_type[afi][safi] != BGP_ADDPATH_NONE)
+			json_object_boolean_true_add(
+				json_addr,
+				bgp_addpath_names(p->addpath_type[afi][safi])
+					->type_json_name);
 
 		if (CHECK_FLAG(p->af_flags[afi][safi], PEER_FLAG_AS_OVERRIDE))
 			json_object_string_add(json_addr,
@@ -8930,14 +8955,10 @@ static void bgp_show_peer_afi(struct vty *vty, struct peer *p, afi_t afi,
 			vty_out(vty,
 				"  Private AS numbers removed in updates to this neighbor\n");
 
-		if (CHECK_FLAG(p->af_flags[afi][safi],
-			       PEER_FLAG_ADDPATH_TX_ALL_PATHS))
-			vty_out(vty, "  Advertise all paths via addpath\n");
-
-		if (CHECK_FLAG(p->af_flags[afi][safi],
-			       PEER_FLAG_ADDPATH_TX_BESTPATH_PER_AS))
-			vty_out(vty,
-				"  Advertise bestpath per AS via addpath\n");
+		if (p->addpath_type[afi][safi] != BGP_ADDPATH_NONE)
+			vty_out(vty, "  %s\n",
+				bgp_addpath_names(p->addpath_type[afi][safi])
+					->human_description);
 
 		if (CHECK_FLAG(p->af_flags[afi][safi], PEER_FLAG_AS_OVERRIDE))
 			vty_out(vty,
diff --git a/bgpd/bgpd.c b/bgpd/bgpd.c
index 0c26fa6720..32bbb78ef3 100644
--- a/bgpd/bgpd.c
+++ b/bgpd/bgpd.c
@@ -85,6 +85,7 @@
 #include "bgpd/bgp_flowspec.h"
 #include "bgpd/bgp_labelpool.h"
 #include "bgpd/bgp_pbr.h"
+#include "bgpd/bgp_addpath.h"
 
 DEFINE_MTYPE_STATIC(BGPD, PEER_TX_SHUTDOWN_MSG, "Peer shutdown message (TX)");
 DEFINE_QOBJ_TYPE(bgp_master)
@@ -883,6 +884,31 @@ static bool peergroup_filter_check(struct peer *peer, afi_t afi, safi_t safi,
 	}
 }
 
+/* Return true if the addpath type is set for peer and different from
+ * peer-group.
+ */
+static int peergroup_af_addpath_check(struct peer *peer, afi_t afi, safi_t safi)
+{
+	enum bgp_addpath_strat type, g_type;
+
+	type = peer->addpath_type[afi][safi];
+
+	if (type != BGP_ADDPATH_NONE) {
+		if (peer_group_active(peer)) {
+			g_type = peer->group->conf->addpath_type[afi][safi];
+
+			if (type != g_type)
+				return 1;
+			else
+				return 0;
+		}
+
+		return 1;
+	}
+
+	return 0;
+}
+
 /* Check peer's AS number and determines if this peer is IBGP or EBGP */
 static inline bgp_peer_sort_t peer_calc_sort(struct peer *peer)
 {
@@ -960,6 +986,9 @@ bgp_peer_sort_t peer_sort(struct peer *peer)
 
 static void peer_free(struct peer *peer)
 {
+	afi_t afi;
+	safi_t safi;
+
 	assert(peer->status == Deleted);
 
 	QOBJ_UNREG(peer);
@@ -1032,6 +1061,13 @@ static void peer_free(struct peer *peer)
 
 	bfd_info_free(&(peer->bfd_info));
 
+	for (afi = AFI_IP; afi < AFI_MAX; afi++) {
+		for (safi = SAFI_UNICAST; safi < SAFI_MAX; safi++) {
+			bgp_addpath_set_peer_type(peer, afi, safi,
+						 BGP_ADDPATH_NONE);
+		}
+	}
+
 	bgp_unlock(peer->bgp);
 
 	memset(peer, 0, sizeof(struct peer));
@@ -1118,6 +1154,7 @@ struct peer *peer_new(struct bgp *bgp)
 			SET_FLAG(peer->af_flags_invert[afi][safi],
 				 PEER_FLAG_SEND_LARGE_COMMUNITY);
 		}
+		peer->addpath_type[afi][safi] = BGP_ADDPATH_NONE;
 	}
 
 	/* set nexthop-unchanged for l2vpn evpn by default */
@@ -1210,6 +1247,8 @@ void peer_xfer_config(struct peer *peer_dst, struct peer *peer_src)
 		peer_dst->allowas_in[afi][safi] =
 			peer_src->allowas_in[afi][safi];
 		peer_dst->weight[afi][safi] = peer_src->weight[afi][safi];
+		peer_dst->addpath_type[afi][safi] =
+			peer_src->addpath_type[afi][safi];
 	}
 
 	for (afidx = BGP_AF_START; afidx < BGP_AF_MAX; afidx++) {
@@ -1808,6 +1847,11 @@ static void peer_group2peer_config_copy_af(struct peer_group *group,
 				      MTYPE_BGP_FILTER_NAME);
 		PEER_ATTR_INHERIT(peer, group, filter[afi][safi].usmap.map);
 	}
+
+	if (peer->addpath_type[afi][safi] == BGP_ADDPATH_NONE) {
+		peer->addpath_type[afi][safi] = conf->addpath_type[afi][safi];
+		bgp_addpath_type_changed(conf->bgp);
+	}
 }
 
 static int peer_activate_af(struct peer *peer, afi_t afi, safi_t safi)
@@ -2836,7 +2880,7 @@ static struct bgp *bgp_create(as_t *as, const char *name,
 #if DFLT_BGP_DETERMINISTIC_MED
 	bgp_flag_set(bgp, BGP_FLAG_DETERMINISTIC_MED);
 #endif
-	bgp->addpath_tx_id = BGP_ADDPATH_TX_ID_FOR_DEFAULT_ORIGINATE;
+	bgp_addpath_init_bgp_data(&bgp->tx_addpath);
 
 	bgp->as = *as;
 
@@ -3627,15 +3671,7 @@ int peer_active_nego(struct peer *peer)
 	return 0;
 }
 
-/* peer_flag_change_type. */
-enum peer_change_type {
-	peer_change_none,
-	peer_change_reset,
-	peer_change_reset_in,
-	peer_change_reset_out,
-};
-
-static void peer_change_action(struct peer *peer, afi_t afi, safi_t safi,
+void peer_change_action(struct peer *peer, afi_t afi, safi_t safi,
 			       enum peer_change_type type)
 {
 	if (CHECK_FLAG(peer->sflags, PEER_STATUS_GROUP))
@@ -3731,8 +3767,6 @@ static const struct peer_flag_action peer_af_flag_action_list[] = {
 	{PEER_FLAG_REMOVE_PRIVATE_AS_REPLACE, 1, peer_change_reset_out},
 	{PEER_FLAG_AS_OVERRIDE, 1, peer_change_reset_out},
 	{PEER_FLAG_REMOVE_PRIVATE_AS_ALL_REPLACE, 1, peer_change_reset_out},
-	{PEER_FLAG_ADDPATH_TX_ALL_PATHS, 1, peer_change_reset},
-	{PEER_FLAG_ADDPATH_TX_BESTPATH_PER_AS, 1, peer_change_reset},
 	{PEER_FLAG_WEIGHT, 0, peer_change_reset_in},
 	{0, 0, 0}};
 
@@ -3957,9 +3991,7 @@ static int peer_af_flag_modify(struct peer *peer, afi_t afi, safi_t safi,
 {
 	int found;
 	int size;
-	int addpath_tx_used;
 	bool invert, member_invert;
-	struct bgp *bgp;
 	struct peer *member;
 	struct listnode *node, *nnode;
 	struct peer_flag_action action;
@@ -4122,45 +4154,6 @@ static int peer_af_flag_modify(struct peer *peer, afi_t afi, safi_t safi,
 		}
 	}
 
-	/* Track if addpath TX is in use */
-	if (flag & (PEER_FLAG_ADDPATH_TX_ALL_PATHS
-		    | PEER_FLAG_ADDPATH_TX_BESTPATH_PER_AS)) {
-		bgp = peer->bgp;
-		addpath_tx_used = 0;
-
-		if (set) {
-			addpath_tx_used = 1;
-
-			if (flag & PEER_FLAG_ADDPATH_TX_BESTPATH_PER_AS) {
-				if (!bgp_flag_check(
-					    bgp, BGP_FLAG_DETERMINISTIC_MED)) {
-					zlog_info(
-						"%s: enabling bgp deterministic-med, this is required"
-						" for addpath-tx-bestpath-per-AS",
-						peer->host);
-					bgp_flag_set(
-						bgp,
-						BGP_FLAG_DETERMINISTIC_MED);
-					bgp_recalculate_all_bestpaths(bgp);
-				}
-			}
-		} else {
-			for (ALL_LIST_ELEMENTS(bgp->peer, node, nnode,
-					       member)) {
-				if (CHECK_FLAG(member->af_flags[afi][safi],
-					       PEER_FLAG_ADDPATH_TX_ALL_PATHS)
-				    || CHECK_FLAG(
-					       member->af_flags[afi][safi],
-					       PEER_FLAG_ADDPATH_TX_BESTPATH_PER_AS)) {
-					addpath_tx_used = 1;
-					break;
-				}
-			}
-		}
-
-		bgp->addpath_tx_used[afi][safi] = addpath_tx_used;
-	}
-
 	return 0;
 }
 
@@ -7065,15 +7058,21 @@ static void bgp_config_write_peer_af(struct vty *vty, struct bgp *bgp,
 	}
 
 	/* addpath TX knobs */
-	if (peergroup_af_flag_check(peer, afi, safi,
-				    PEER_FLAG_ADDPATH_TX_ALL_PATHS)) {
-		vty_out(vty, "  neighbor %s addpath-tx-all-paths\n", addr);
-	}
-
-	if (peergroup_af_flag_check(peer, afi, safi,
-				    PEER_FLAG_ADDPATH_TX_BESTPATH_PER_AS)) {
-		vty_out(vty, "  neighbor %s addpath-tx-bestpath-per-AS\n",
-			addr);
+	if (peergroup_af_addpath_check(peer, afi, safi)) {
+		switch (peer->addpath_type[afi][safi]) {
+		case BGP_ADDPATH_ALL:
+			vty_out(vty, "  neighbor %s addpath-tx-all-paths\n",
+				addr);
+			break;
+		case BGP_ADDPATH_BEST_PER_AS:
+			vty_out(vty,
+				"  neighbor %s addpath-tx-bestpath-per-AS\n",
+				addr);
+			break;
+		case BGP_ADDPATH_MAX:
+		case BGP_ADDPATH_NONE:
+			break;
+		}
 	}
 
 	/* ORF capability.  */
diff --git a/bgpd/bgpd.h b/bgpd/bgpd.h
index e14b0f39e0..5ee54e3c70 100644
--- a/bgpd/bgpd.h
+++ b/bgpd/bgpd.h
@@ -39,6 +39,7 @@
 #include "bitfield.h"
 #include "vxlan.h"
 #include "bgp_labelpool.h"
+#include "bgp_addpath_types.h"
 
 #define BGP_MAX_HOSTNAME 64	/* Linux max, is larger than most other sys */
 #define BGP_PEER_MAX_HASH_SIZE 16384
@@ -462,8 +463,7 @@ struct bgp {
 	/* Auto-shutdown new peers */
 	bool autoshutdown;
 
-	uint32_t addpath_tx_id;
-	int addpath_tx_used[AFI_MAX][SAFI_MAX];
+	struct bgp_addpath_bgp_data tx_addpath;
 
 #if ENABLE_BGP_VNC
 	struct rfapi_cfg *rfapi_cfg;
@@ -938,12 +938,12 @@ struct peer {
 #define PEER_FLAG_REMOVE_PRIVATE_AS_REPLACE (1 << 19) /* remove-private-as replace-as */
 #define PEER_FLAG_AS_OVERRIDE               (1 << 20) /* as-override */
 #define PEER_FLAG_REMOVE_PRIVATE_AS_ALL_REPLACE (1 << 21) /* remove-private-as all replace-as */
-#define PEER_FLAG_ADDPATH_TX_ALL_PATHS      (1 << 22) /* addpath-tx-all-paths */
-#define PEER_FLAG_ADDPATH_TX_BESTPATH_PER_AS (1 << 23) /* addpath-tx-bestpath-per-AS */
 #define PEER_FLAG_WEIGHT                    (1 << 24) /* weight */
 #define PEER_FLAG_ALLOWAS_IN_ORIGIN         (1 << 25) /* allowas-in origin */
 #define PEER_FLAG_SEND_LARGE_COMMUNITY      (1 << 26) /* Send large Communities */
 
+	enum bgp_addpath_strat addpath_type[AFI_MAX][SAFI_MAX];
+
 	/* MD5 password */
 	char *password;
 
@@ -1466,6 +1466,14 @@ typedef enum {
 	BGP_POLICY_DISTRIBUTE_LIST,
 } bgp_policy_type_e;
 
+/* peer_flag_change_type. */
+enum peer_change_type {
+	peer_change_none,
+	peer_change_reset,
+	peer_change_reset_in,
+	peer_change_reset_out,
+};
+
 extern struct bgp_master *bm;
 extern unsigned int multipath_num;
 
@@ -1597,6 +1605,8 @@ extern int peer_af_flag_unset(struct peer *, afi_t, safi_t, uint32_t);
 extern int peer_af_flag_check(struct peer *, afi_t, safi_t, uint32_t);
 extern void peer_af_flag_inherit(struct peer *peer, afi_t afi, safi_t safi,
 				 uint32_t flag);
+extern void peer_change_action(struct peer *peer, afi_t afi, safi_t safi,
+			       enum peer_change_type type);
 
 extern int peer_ebgp_multihop_set(struct peer *, int);
 extern int peer_ebgp_multihop_unset(struct peer *);
diff --git a/bgpd/subdir.am b/bgpd/subdir.am
index 83f55939ce..a13f058f74 100644
--- a/bgpd/subdir.am
+++ b/bgpd/subdir.am
@@ -46,6 +46,7 @@ man8 += $(MANBUILD)/bgpd.8
 endif
 
 bgpd_libbgp_a_SOURCES = \
+	bgpd/bgp_addpath.c \
 	bgpd/bgp_advertise.c \
 	bgpd/bgp_aspath.c \
 	bgpd/bgp_attr.c \
@@ -115,6 +116,8 @@ bgpd_libbgp_a_SOURCES += \
 endif
 
 noinst_HEADERS += \
+	bgpd/bgp_addpath.h \
+	bgpd/bgp_addpath_types.h \
 	bgpd/bgp_advertise.h \
 	bgpd/bgp_aspath.h \
 	bgpd/bgp_attr.h \
diff --git a/lib/id_alloc.c b/lib/id_alloc.c
new file mode 100644
index 0000000000..b4d37dbdcf
--- /dev/null
+++ b/lib/id_alloc.c
@@ -0,0 +1,406 @@
+/*
+ * FRR ID Number Allocator
+ * Copyright (C) 2018  Amazon.com, Inc. or its affiliates
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the Free
+ * Software Foundation; either version 2 of the License, or (at your option)
+ * any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; see the file COPYING; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "id_alloc.h"
+
+#include "log.h"
+#include "lib_errors.h"
+#include "memory.h"
+
+#include <inttypes.h>
+
+DEFINE_MTYPE_STATIC(LIB, IDALLOC_ALLOCATOR, "ID Number Allocator")
+DEFINE_MTYPE_STATIC(LIB, IDALLOC_ALLOCATOR_NAME, "ID Number Allocator Name")
+DEFINE_MTYPE_STATIC(LIB, IDALLOC_DIRECTORY, "ID Number Allocator Directory")
+DEFINE_MTYPE_STATIC(LIB, IDALLOC_SUBDIRECTORY,
+		    "ID Number Allocator Subdirectory")
+DEFINE_MTYPE_STATIC(LIB, IDALLOC_PAGE, "ID Number Allocator Page")
+DEFINE_MTYPE_STATIC(LIB, IDALLOC_POOL, "ID Number temporary holding pool entry")
+
+#if UINT_MAX >= UINT32_MAX
+#define FFS32(x) ffs(x)
+#else
+/* ints less than 32 bits? Yikes. */
+#define FFS32(x) ffsl(x)
+#endif
+
+#define DIR_MASK    ((1<<IDALLOC_DIR_BITS)-1)
+#define SUBDIR_MASK ((1<<IDALLOC_SUBDIR_BITS)-1)
+#define PAGE_MASK   ((1<<IDALLOC_PAGE_BITS)-1)
+#define WORD_MASK   ((1<<IDALLOC_WORD_BITS)-1)
+#define OFFSET_MASK ((1<<IDALLOC_OFFSET_BITS)-1)
+
+#define DIR_SHIFT    (IDALLOC_OFFSET_BITS + IDALLOC_WORD_BITS + \
+		      IDALLOC_PAGE_BITS + IDALLOC_SUBDIR_BITS)
+#define SUBDIR_SHIFT (IDALLOC_OFFSET_BITS + IDALLOC_WORD_BITS + \
+		      IDALLOC_PAGE_BITS)
+#define PAGE_SHIFT   (IDALLOC_OFFSET_BITS + IDALLOC_WORD_BITS)
+#define WORD_SHIFT   (IDALLOC_OFFSET_BITS)
+#define OFFSET_SHIFT (0)
+
+#define ID_DIR(id)    ((id >> DIR_SHIFT)    & DIR_MASK)
+#define ID_SUBDIR(id) ((id >> SUBDIR_SHIFT) & SUBDIR_MASK)
+#define ID_PAGE(id)   ((id >> PAGE_SHIFT)   & PAGE_MASK)
+#define ID_WORD(id)   ((id >> WORD_SHIFT)   & WORD_MASK)
+#define ID_OFFSET(id) ((id >> OFFSET_SHIFT) & OFFSET_MASK)
+
+/*
+ * Find the page that an ID number belongs to in an allocator.
+ * Optionally create the page if it doesn't exist.
+ */
+static struct id_alloc_page *find_or_create_page(struct id_alloc *alloc,
+						 uint32_t id, int create)
+{
+	struct id_alloc_dir *dir = NULL;
+	struct id_alloc_subdir *subdir = NULL;
+	struct id_alloc_page *page = NULL;
+
+	dir = alloc->sublevels[ID_DIR(id)];
+	if (dir == NULL) {
+		if (create) {
+			dir = XCALLOC(MTYPE_IDALLOC_DIRECTORY, sizeof(*dir));
+			alloc->sublevels[ID_DIR(id)] = dir;
+		} else {
+			return NULL;
+		}
+	}
+
+	subdir = dir->sublevels[ID_SUBDIR(id)];
+	if (subdir == NULL) {
+		if (create) {
+			subdir = XCALLOC(MTYPE_IDALLOC_SUBDIRECTORY,
+					 sizeof(*subdir));
+			dir->sublevels[ID_SUBDIR(id)] = subdir;
+		} else {
+			return NULL;
+		}
+	}
+
+	page = subdir->sublevels[ID_PAGE(id)];
+	if (page == NULL && create) {
+		page = XCALLOC(MTYPE_IDALLOC_PAGE, sizeof(*page));
+		page->base_value = id;
+		subdir->sublevels[ID_PAGE(id)] = page;
+
+		alloc->capacity += 1 << PAGE_SHIFT;
+		page->next_has_free = alloc->has_free;
+		alloc->has_free = page;
+	} else if (page != NULL && create) {
+		flog_err(
+			EC_LIB_ID_CONSISTENCY,
+			"ID Allocator %s attempt to re-create page at %" PRIu32,
+			alloc->name, id);
+	}
+
+	return page;
+}
+
+/*
+ * Return an ID number back to the allocator.
+ * While this ID can be re-assigned through idalloc_allocate, the underlying
+ * memory will not be freed. If this is the first free ID in the page, the page
+ * will be added to the allocator's list of pages with free IDs.
+ */
+void idalloc_free(struct id_alloc *alloc, uint32_t id)
+{
+	struct id_alloc_page *page = NULL;
+
+	int word, offset;
+	uint32_t old_word, old_word_mask;
+
+	page = find_or_create_page(alloc, id, 0);
+	if (!page) {
+		flog_err(EC_LIB_ID_CONSISTENCY,
+			"ID Allocator %s cannot free #%" PRIu32
+			". ID Block does not exist.",
+			alloc->name, id);
+		return;
+	}
+
+	word = ID_WORD(id);
+	offset = ID_OFFSET(id);
+
+	if ((page->allocated_mask[word] & (1 << offset)) == 0) {
+		flog_err(EC_LIB_ID_CONSISTENCY,
+			"ID Allocator %s cannot free #%" PRIu32
+			". ID was not allocated at the time of free.",
+			alloc->name, id);
+		return;
+	}
+
+	old_word = page->allocated_mask[word];
+	page->allocated_mask[word] &= ~(((uint32_t)1) << offset);
+	alloc->allocated -= 1;
+
+	if (old_word == UINT32_MAX) {
+		/* first bit in this block of 32 to be freed.*/
+
+		old_word_mask = page->full_word_mask;
+		page->full_word_mask &= ~(((uint32_t)1) << word);
+
+		if (old_word_mask == UINT32_MAX) {
+			/* first bit in page freed, add this to the allocator's
+			 * list of pages with free space
+			 */
+			page->next_has_free = alloc->has_free;
+			alloc->has_free = page;
+		}
+	}
+}
+
+/*
+ * Add a allocation page to the end of the allocator's current range.
+ * Returns null if the allocator has had all possible pages allocated already.
+ */
+static struct id_alloc_page *create_next_page(struct id_alloc *alloc)
+{
+	if (alloc->capacity == 0 && alloc->sublevels[0])
+		return NULL; /* All IDs allocated and the capacity looped. */
+
+	return find_or_create_page(alloc, alloc->capacity, 1);
+}
+
+/*
+ * Marks an ID within an allocator page as in use.
+ * If the ID was the last free ID in the page, the page is removed from the
+ * allocator's list of free IDs. In the typical allocation case, this page is
+ * the first page in the list, and removing the page is fast. If instead an ID
+ * is being reserved by number, this may end up scanning the whole single linked
+ * list of pages in order to remove it.
+ */
+static void reserve_bit(struct id_alloc *alloc, struct id_alloc_page *page,
+			int word, int offset)
+{
+	struct id_alloc_page *itr;
+
+	page->allocated_mask[word] |= ((uint32_t)1) << offset;
+	alloc->allocated += 1;
+
+	if (page->allocated_mask[word] == UINT32_MAX) {
+		page->full_word_mask |= ((uint32_t)1) << word;
+		if (page->full_word_mask == UINT32_MAX) {
+			if (alloc->has_free == page) {
+				/* allocate always pulls from alloc->has_free */
+				alloc->has_free = page->next_has_free;
+			} else {
+				/* reserve could pull from any page with free
+				 * bits
+				 */
+				itr = alloc->has_free;
+				while (itr) {
+					if (itr->next_has_free == page) {
+						itr->next_has_free =
+							page->next_has_free;
+						return;
+					}
+
+					itr = itr->next_has_free;
+				}
+			}
+		}
+	}
+}
+
+/*
+ * Reserve an ID number from the allocator. Returns IDALLOC_INVALID (0) if the
+ * allocator has no more IDs available.
+ */
+uint32_t idalloc_allocate(struct id_alloc *alloc)
+{
+	struct id_alloc_page *page;
+	int word, offset;
+	uint32_t return_value;
+
+	if (alloc->has_free == NULL)
+		create_next_page(alloc);
+
+	if (alloc->has_free == NULL) {
+		flog_err(EC_LIB_ID_EXHAUST,
+			"ID Allocator %s has run out of IDs.", alloc->name);
+		return IDALLOC_INVALID;
+	}
+
+	page = alloc->has_free;
+	word = FFS32(~(page->full_word_mask)) - 1;
+
+	if (word < 0 || word >= 32) {
+		flog_err(EC_LIB_ID_CONSISTENCY,
+			"ID Allocator %s internal error. Page starting at %d is inconsistent.",
+			alloc->name, page->base_value);
+		return IDALLOC_INVALID;
+	}
+
+	offset = FFS32(~(page->allocated_mask[word])) - 1;
+	if (offset < 0 || offset >= 32) {
+		flog_err(EC_LIB_ID_CONSISTENCY,
+			"ID Allocator %s internal error. Page starting at %d is inconsistent on word %d",
+			alloc->name, page->base_value, word);
+		return IDALLOC_INVALID;
+	}
+	return_value = page->base_value + word * 32 + offset;
+
+	reserve_bit(alloc, page, word, offset);
+
+	return return_value;
+}
+
+/*
+ * Tries to allocate a specific ID from the allocator. Returns IDALLOC_INVALID
+ * when the ID being "reserved" has allready been assigned/reserved. This should
+ * only be done with low numbered IDs, as the allocator needs to reserve bit-map
+ * pages in order
+ */
+uint32_t idalloc_reserve(struct id_alloc *alloc, uint32_t id)
+{
+	struct id_alloc_page *page;
+	int word, offset;
+
+	while (alloc->capacity <= id)
+		create_next_page(alloc);
+
+	word = ID_WORD(id);
+	offset = ID_OFFSET(id);
+	page = find_or_create_page(alloc, id, 0);
+	/* page can't be null because the loop above ensured it was created. */
+
+	if (page->allocated_mask[word] & (((uint32_t)1) << offset)) {
+		flog_err(EC_LIB_ID_CONSISTENCY,
+			"ID Allocator %s could not reserve %" PRIu32
+			" because it is already allocated.",
+			alloc->name, id);
+		return IDALLOC_INVALID;
+	}
+
+	reserve_bit(alloc, page, word, offset);
+	return id;
+}
+
+/*
+ * Set up an empty ID allocator, with IDALLOC_INVALID pre-reserved.
+ */
+struct id_alloc *idalloc_new(const char *name)
+{
+	struct id_alloc *ret;
+
+	ret = XCALLOC(MTYPE_IDALLOC_ALLOCATOR, sizeof(*ret));
+	ret->name = XSTRDUP(MTYPE_IDALLOC_ALLOCATOR_NAME, name);
+
+	idalloc_reserve(ret, IDALLOC_INVALID);
+
+	return ret;
+}
+
+/*
+ * Free a subdir, and all pages below it.
+ */
+static void idalloc_destroy_subdir(struct id_alloc_subdir *subdir)
+{
+	int i;
+
+	for (i = 0; i < IDALLOC_PAGE_COUNT; i++) {
+		if (subdir->sublevels[i])
+			XFREE(MTYPE_IDALLOC_PAGE, subdir->sublevels[i]);
+		else
+			break;
+	}
+	XFREE(MTYPE_IDALLOC_SUBDIRECTORY, subdir);
+}
+
+/*
+ * Free a dir, and all subdirs/pages below it.
+ */
+static void idalloc_destroy_dir(struct id_alloc_dir *dir)
+{
+	int i;
+
+	for (i = 0; i < IDALLOC_SUBDIR_COUNT; i++) {
+		if (dir->sublevels[i])
+			idalloc_destroy_subdir(dir->sublevels[i]);
+		else
+			break;
+	}
+	XFREE(MTYPE_IDALLOC_DIRECTORY, dir);
+}
+
+/*
+ * Free all memory associated with an ID allocator.
+ */
+void idalloc_destroy(struct id_alloc *alloc)
+{
+	int i;
+
+	for (i = 0; i < IDALLOC_DIR_COUNT; i++) {
+		if (alloc->sublevels[i])
+			idalloc_destroy_dir(alloc->sublevels[i]);
+		else
+			break;
+	}
+
+	XFREE(MTYPE_IDALLOC_ALLOCATOR_NAME, alloc->name);
+	XFREE(MTYPE_IDALLOC_ALLOCATOR, alloc);
+}
+
+/*
+ * Give an ID number to temporary holding pool.
+ */
+void idalloc_free_to_pool(struct id_alloc_pool **pool_ptr, uint32_t id)
+{
+	struct id_alloc_pool *new_pool;
+
+	new_pool = XMALLOC(MTYPE_IDALLOC_POOL, sizeof(*new_pool));
+	new_pool->id = id;
+	new_pool->next = *pool_ptr;
+	*pool_ptr = new_pool;
+}
+
+/*
+ * Free all ID numbers held in a holding pool back to the main allocator.
+ */
+void idalloc_drain_pool(struct id_alloc *alloc, struct id_alloc_pool **pool_ptr)
+{
+	struct id_alloc_pool *current, *next;
+
+	while (*pool_ptr) {
+		current = *pool_ptr;
+		next = current->next;
+		idalloc_free(alloc, current->id);
+		XFREE(MTYPE_IDALLOC_POOL, current);
+		*pool_ptr = next;
+	}
+}
+
+/*
+ * Allocate an ID from either a holding pool, or the main allocator. IDs will
+ * only be pulled form the main allocator when the pool is empty.
+ */
+uint32_t idalloc_allocate_prefer_pool(struct id_alloc *alloc,
+				      struct id_alloc_pool **pool_ptr)
+{
+	uint32_t ret;
+	struct id_alloc_pool *pool_head = *pool_ptr;
+
+	if (pool_head) {
+		ret = pool_head->id;
+		*pool_ptr = pool_head->next;
+		XFREE(MTYPE_IDALLOC_POOL, pool_head);
+		return ret;
+	} else {
+		return idalloc_allocate(alloc);
+	}
+}
diff --git a/lib/id_alloc.h b/lib/id_alloc.h
new file mode 100644
index 0000000000..efe355658a
--- /dev/null
+++ b/lib/id_alloc.h
@@ -0,0 +1,90 @@
+/*
+ * FRR ID Number Allocator
+ * Copyright (C) 2018  Amazon.com, Inc. or its affiliates
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the Free
+ * Software Foundation; either version 2 of the License, or (at your option)
+ * any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; see the file COPYING; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef _ZEBRA_ID_ALLOC_H
+#define _ZEBRA_ID_ALLOC_H
+
+#include <strings.h>
+#include <limits.h>
+#include <stdint.h>
+
+#define IDALLOC_INVALID 0
+
+#define IDALLOC_DIR_BITS     8
+#define IDALLOC_SUBDIR_BITS  7
+#define IDALLOC_PAGE_BITS    7
+#define IDALLOC_WORD_BITS    5
+#define IDALLOC_OFFSET_BITS  5
+
+#define IDALLOC_DIR_COUNT (1 << IDALLOC_DIR_BITS)
+#define IDALLOC_SUBDIR_COUNT (1 << IDALLOC_SUBDIR_BITS)
+#define IDALLOC_PAGE_COUNT (1 << IDALLOC_PAGE_BITS)
+#define IDALLOC_WORD_COUNT (1 << IDALLOC_WORD_BITS)
+
+struct id_alloc_page {
+	/* Bitmask of allocations. 1s indicates the ID is already allocated. */
+	uint32_t allocated_mask[IDALLOC_WORD_COUNT];
+
+	/* Bitmask for free space in allocated_mask. 1s indicate whole 32 bit
+	 * section is full.
+	 */
+	uint32_t full_word_mask;
+
+	/* The ID that bit 0 in allocated_mask corresponds to. */
+	uint32_t base_value;
+
+	struct id_alloc_page
+		*next_has_free; /* Next page with at least one bit open */
+};
+
+struct id_alloc_subdir {
+	struct id_alloc_page *sublevels[IDALLOC_PAGE_COUNT];
+};
+
+struct id_alloc_dir {
+	struct id_alloc_subdir *sublevels[IDALLOC_SUBDIR_COUNT];
+};
+
+struct id_alloc {
+	struct id_alloc_dir *sublevels[IDALLOC_DIR_COUNT];
+
+	struct id_alloc_page *has_free;
+
+	char *name;
+
+	uint32_t allocated, capacity;
+};
+
+struct id_alloc_pool {
+	struct id_alloc_pool *next;
+	uint32_t id;
+};
+
+void idalloc_free(struct id_alloc *alloc, uint32_t id);
+void idalloc_free_to_pool(struct id_alloc_pool **pool_ptr, uint32_t id);
+void idalloc_drain_pool(struct id_alloc *alloc,
+			struct id_alloc_pool **pool_ptr);
+uint32_t idalloc_allocate(struct id_alloc *alloc);
+uint32_t idalloc_allocate_prefer_pool(struct id_alloc *alloc,
+				      struct id_alloc_pool **pool_ptr);
+uint32_t idalloc_reserve(struct id_alloc *alloc, uint32_t id);
+struct id_alloc *idalloc_new(const char *name);
+void idalloc_destroy(struct id_alloc *alloc);
+
+#endif
diff --git a/lib/lib_errors.h b/lib/lib_errors.h
index 3831116dd3..38c75f913e 100644
--- a/lib/lib_errors.h
+++ b/lib/lib_errors.h
@@ -72,6 +72,8 @@ enum lib_log_refs {
 	EC_LIB_SYSREPO_INIT,
 	EC_LIB_SYSREPO_DATA_CONVERT,
 	EC_LIB_LIBSYSREPO,
+	EC_LIB_ID_CONSISTENCY,
+	EC_LIB_ID_EXHAUST,
 };
 
 extern void lib_error_init(void);
diff --git a/lib/subdir.am b/lib/subdir.am
index 356796fbb1..c144c2c2e1 100644
--- a/lib/subdir.am
+++ b/lib/subdir.am
@@ -29,6 +29,7 @@ lib_libfrr_la_SOURCES = \
 	lib/graph.c \
 	lib/hash.c \
 	lib/hook.c \
+	lib/id_alloc.c \
 	lib/if.c \
 	lib/if_rmap.c \
 	lib/imsg-buffer.c \
@@ -147,6 +148,7 @@ pkginclude_HEADERS += \
 	lib/graph.h \
 	lib/hash.h \
 	lib/hook.h \
+	lib/id_alloc.h \
 	lib/if.h \
 	lib/if_rmap.h \
 	lib/imsg.h \
diff --git a/tests/.gitignore b/tests/.gitignore
index 49a4b6e470..a6202786be 100644
--- a/tests/.gitignore
+++ b/tests/.gitignore
@@ -20,23 +20,24 @@
 /lib/cli/test_commands_defun.c
 /lib/test_buffer
 /lib/test_checksum
+/lib/test_graph
 /lib/test_heavy
 /lib/test_heavy_thread
 /lib/test_heavy_wq
+/lib/test_idalloc
 /lib/test_memory
 /lib/test_nexthop_iter
 /lib/test_privs
 /lib/test_ringbuf
-/lib/test_srcdest_table
 /lib/test_segv
 /lib/test_sig
+/lib/test_srcdest_table
 /lib/test_stream
 /lib/test_table
 /lib/test_timer_correctness
 /lib/test_timer_performance
 /lib/test_ttable
-/lib/test_zmq
 /lib/test_zlog
-/lib/test_graph
+/lib/test_zmq
 /ospf6d/test_lsdb
 /ospf6d/test_lsdb_clippy.c
diff --git a/tests/bgpd/test_bgp_table.c b/tests/bgpd/test_bgp_table.c
index 171b1aa22f..73243dcacf 100644
--- a/tests/bgpd/test_bgp_table.c
+++ b/tests/bgpd/test_bgp_table.c
@@ -28,6 +28,8 @@
 #include "bgpd/bgp_table.h"
 #include "linklist.h"
 
+/* Satisfy link requirements from including bgpd.h */
+struct zebra_privs_t bgpd_privs = {0};
 /*
  * test_node_t
  *
diff --git a/tests/bgpd/test_peer_attr.c b/tests/bgpd/test_peer_attr.c
index 2f464d7217..2fbc686e1e 100644
--- a/tests/bgpd/test_peer_attr.c
+++ b/tests/bgpd/test_peer_attr.c
@@ -254,6 +254,8 @@ TEST_STR_ATTR_HANDLER_DECL(password, password, "FRR-Peer", "FRR-Group");
 TEST_ATTR_HANDLER_DECL(local_as, change_local_as, 1, 2);
 TEST_ATTR_HANDLER_DECL(timers_1, keepalive, 10, 20);
 TEST_ATTR_HANDLER_DECL(timers_2, holdtime, 30, 60);
+TEST_ATTR_HANDLER_DECL(addpath_types, addpath_type[pa->afi][pa->safi],
+		       BGP_ADDPATH_ALL, BGP_ADDPATH_BEST_PER_AS);
 TEST_SU_ATTR_HANDLER_DECL(update_source_su, update_source, "255.255.255.1",
 			  "255.255.255.2");
 TEST_STR_ATTR_HANDLER_DECL(update_source_if, update_if, "IF-PEER", "IF-GROUP");
@@ -414,12 +416,11 @@ static struct test_peer_attr test_peer_attrs[] = {
 
 	/* Address Family Attributes */
 	{
-		.cmd = "addpath-tx-all-paths",
-		.u.flag = PEER_FLAG_ADDPATH_TX_ALL_PATHS,
-	},
-	{
-		.cmd = "addpath-tx-bestpath-per-AS",
-		.u.flag = PEER_FLAG_ADDPATH_TX_BESTPATH_PER_AS,
+		.cmd = "addpath",
+		.peer_cmd = "addpath-tx-all-paths",
+		.group_cmd = "addpath-tx-bestpath-per-AS",
+		.type = PEER_AT_AF_CUSTOM,
+		.handlers[0] = TEST_HANDLER(addpath_types),
 	},
 	{
 		.cmd = "allowas-in",
diff --git a/tests/bgpd/test_peer_attr.py b/tests/bgpd/test_peer_attr.py
index 7a4dd2b04a..bafd131b78 100644
--- a/tests/bgpd/test_peer_attr.py
+++ b/tests/bgpd/test_peer_attr.py
@@ -26,14 +26,10 @@ TestFlag.okfail('peer\\timers')
 TestFlag.okfail('peer\\timers connect')
 TestFlag.okfail('peer\\update-source')
 TestFlag.okfail('peer\\update-source')
-TestFlag.okfail('peer\\ipv4-unicast\\addpath-tx-all-paths')
-TestFlag.okfail('peer\\ipv4-multicast\\addpath-tx-all-paths')
-TestFlag.okfail('peer\\ipv6-unicast\\addpath-tx-all-paths')
-TestFlag.okfail('peer\\ipv6-multicast\\addpath-tx-all-paths')
-TestFlag.okfail('peer\\ipv4-unicast\\addpath-tx-bestpath-per-AS')
-TestFlag.okfail('peer\\ipv4-multicast\\addpath-tx-bestpath-per-AS')
-TestFlag.okfail('peer\\ipv6-unicast\\addpath-tx-bestpath-per-AS')
-TestFlag.okfail('peer\\ipv6-multicast\\addpath-tx-bestpath-per-AS')
+TestFlag.okfail('peer\\ipv4-unicast\\addpath')
+TestFlag.okfail('peer\\ipv4-multicast\\addpath')
+TestFlag.okfail('peer\\ipv6-unicast\\addpath')
+TestFlag.okfail('peer\\ipv6-multicast\\addpath')
 TestFlag.okfail('peer\\ipv4-unicast\\allowas-in')
 TestFlag.okfail('peer\\ipv4-multicast\\allowas-in')
 TestFlag.okfail('peer\\ipv6-unicast\\allowas-in')
diff --git a/tests/lib/test_idalloc.c b/tests/lib/test_idalloc.c
new file mode 100644
index 0000000000..3053c1c074
--- /dev/null
+++ b/tests/lib/test_idalloc.c
@@ -0,0 +1,193 @@
+#include "id_alloc.h"
+
+#include <inttypes.h>
+#include <string.h>
+#include <assert.h>
+#include <stdio.h>
+
+#define IDS_PER_PAGE (1<<(IDALLOC_OFFSET_BITS + IDALLOC_WORD_BITS))
+char allocated_markers[IDS_PER_PAGE*3];
+
+int main(int argc, char **argv)
+{
+	int i, val;
+	uint32_t pg;
+	struct id_alloc *a;
+
+	/* 1. Rattle test, shake it a little and make sure it doesn't make any
+	 * noise :)
+	 */
+	a = idalloc_new("Rattle test");
+	for (i = 0; i < 1000000; i++)
+		assert(idalloc_allocate(a) != 0);
+
+	idalloc_destroy(a);
+
+	/* 2. Reserve a few low IDs, make sure they are skipped by normal
+	 * allocation.
+	 */
+	a = idalloc_new("Low Reservations");
+	assert(idalloc_reserve(a, 1) == 1);
+	assert(idalloc_reserve(a, 3) == 3);
+	assert(idalloc_reserve(a, 5) == 5);
+	for (i = 0; i < 100; i++) {
+		val = idalloc_allocate(a);
+		assert(val != 1 && val != 3 && val != 5);
+	}
+	idalloc_destroy(a);
+
+	/* 3. Single page testing. Check that IDs are kept unique, and all IDs
+	 * in the existing page are allocated before a new page is added.
+	 */
+	memset(allocated_markers, 0, sizeof(allocated_markers));
+	allocated_markers[IDALLOC_INVALID] = 1;
+
+	a = idalloc_new("Single Page");
+
+	/* reserve the rest of the first page */
+	for (i = 0; i < IDS_PER_PAGE - 1; i++) {
+		val = idalloc_allocate(a);
+		assert(val < IDS_PER_PAGE);
+		assert(allocated_markers[val] == 0);
+		assert(a->capacity == IDS_PER_PAGE);
+		allocated_markers[val] = 1;
+	}
+	/* Check that the count is right */
+	assert(a->allocated == IDS_PER_PAGE);
+
+	/* Free some IDs out of the middle. */
+	idalloc_free(a, 300);
+	allocated_markers[300] = 0;
+	idalloc_free(a, 400);
+	allocated_markers[400] = 0;
+	idalloc_free(a, 500);
+	allocated_markers[500] = 0;
+
+	assert(a->allocated == IDS_PER_PAGE-3);
+
+	/* Allocate the three IDs back and make sure they are pulled from the
+	 * set just freed
+	 */
+	for (i = 0; i < 3; i++) {
+		val = idalloc_allocate(a);
+		assert(val < IDS_PER_PAGE);
+		assert(allocated_markers[val] == 0);
+		assert(a->capacity == IDS_PER_PAGE);
+		allocated_markers[val] = 1;
+	}
+	idalloc_destroy(a);
+
+	/* 4. Multi-page testing. */
+	memset(allocated_markers, 0, sizeof(allocated_markers));
+	allocated_markers[IDALLOC_INVALID] = 1;
+
+	a = idalloc_new("Multi-page");
+
+	/* reserve the rest of the first page and all of the second and third */
+	for (i = 0; i < 3 * IDS_PER_PAGE - 1; i++) {
+		val = idalloc_allocate(a);
+		assert(val < 3*IDS_PER_PAGE);
+		assert(allocated_markers[val] == 0);
+		allocated_markers[val] = 1;
+	}
+	assert(a->capacity == 3*IDS_PER_PAGE);
+	assert(a->allocated == 3*IDS_PER_PAGE);
+
+	/* Free two IDs from each page. */
+	for (i = 0; i < 3; i++) {
+		idalloc_free(a, 7 + i*IDS_PER_PAGE);
+		allocated_markers[7 + i*IDS_PER_PAGE] = 0;
+
+		idalloc_free(a, 4 + i*IDS_PER_PAGE);
+		allocated_markers[4 + i*IDS_PER_PAGE] = 0;
+	}
+
+	assert(a->allocated == 3*IDS_PER_PAGE - 6);
+
+	/* Allocate the six IDs back and make sure they are pulled from the set
+	 * just freed.
+	 */
+	for (i = 0; i < 6; i++) {
+		val = idalloc_allocate(a);
+		assert(val < 3*IDS_PER_PAGE);
+		assert(allocated_markers[val] == 0);
+		assert(a->capacity == 3*IDS_PER_PAGE);
+		allocated_markers[val] = 1;
+	}
+
+	assert(a->capacity == 3*IDS_PER_PAGE);
+	assert(a->allocated == 3*IDS_PER_PAGE);
+
+	/* Walk each allocated ID. Free it, then re-allocate it back. */
+	for (i = 1; i < 3 * IDS_PER_PAGE - 1; i++) {
+		idalloc_free(a, i);
+		val = idalloc_allocate(a);
+		assert(val == i);
+		assert(a->capacity == 3*IDS_PER_PAGE);
+		assert(a->allocated == 3*IDS_PER_PAGE);
+	}
+	idalloc_destroy(a);
+
+	/* 5. Weird Reservations
+	 * idalloc_reserve exists primarily to black out low numbered IDs that
+	 * are reserved for special cases. However, we will test it for more
+	 * complex use cases to avoid unpleasant surprises.
+	 */
+
+	memset(allocated_markers, 0, sizeof(allocated_markers));
+	allocated_markers[IDALLOC_INVALID] = 1;
+
+	a = idalloc_new("Weird Reservations");
+
+	/* Start with 3 pages fully allocated. */
+	for (i = 0; i < 3 * IDS_PER_PAGE - 1; i++) {
+		val = idalloc_allocate(a);
+		assert(val < 3*IDS_PER_PAGE);
+		assert(allocated_markers[val] == 0);
+		allocated_markers[val] = 1;
+	}
+	assert(a->capacity == 3*IDS_PER_PAGE);
+	assert(a->allocated == 3*IDS_PER_PAGE);
+
+	/* Free a bit out of each of the three pages. Then reserve one of the
+	 * three freed IDs. Finally, allocate the other two freed IDs. Do this
+	 * each of three ways. (Reserve out of the first, seconds then third
+	 * page.)
+	 * The intent here is to exercise the rare cases on reserve_bit's
+	 * linked-list removal in the case that it is not removing the first
+	 * page with a free bit in its list of pages with free bits.
+	 */
+
+	for (pg = 0; pg < 3; pg++) {
+		/* free a bit out of each of the three pages */
+		for (i = 0; i < 3; i++) {
+			idalloc_free(a, i*IDS_PER_PAGE + 17);
+			allocated_markers[i*IDS_PER_PAGE + 17] = 0;
+		}
+
+		assert(a->capacity == 3*IDS_PER_PAGE);
+		assert(a->allocated == 3*IDS_PER_PAGE-3);
+
+		/* Reserve one of the freed IDs */
+		assert(idalloc_reserve(a, pg*IDS_PER_PAGE + 17) ==
+		       pg*IDS_PER_PAGE + 17);
+		allocated_markers[pg*IDS_PER_PAGE + 17] = 1;
+
+		assert(a->capacity == 3*IDS_PER_PAGE);
+		assert(a->allocated == 3*IDS_PER_PAGE-2);
+
+		/* Allocate the other two back */
+		for (i = 0; i < 2; i++) {
+			val = idalloc_allocate(a);
+			assert(val < 3*IDS_PER_PAGE);
+			assert(allocated_markers[val] == 0);
+			allocated_markers[val] = 1;
+		}
+		assert(a->capacity == 3*IDS_PER_PAGE);
+		assert(a->allocated == 3*IDS_PER_PAGE);
+	}
+	idalloc_destroy(a);
+
+	puts("ID Allocator test successful.\n");
+	return 0;
+}
diff --git a/tests/lib/test_idalloc.py b/tests/lib/test_idalloc.py
new file mode 100644
index 0000000000..22de082be4
--- /dev/null
+++ b/tests/lib/test_idalloc.py
@@ -0,0 +1,6 @@
+import frrtest
+
+class TestIDAlloc(frrtest.TestMultiOut):
+    program = './test_idalloc'
+
+TestIDAlloc.onesimple('ID Allocator test successful.')
diff --git a/tests/subdir.am b/tests/subdir.am
index 7a693ac0f7..6b52c90bc0 100644
--- a/tests/subdir.am
+++ b/tests/subdir.am
@@ -51,6 +51,7 @@ check_PROGRAMS = \
 	tests/lib/test_heavy_thread \
 	tests/lib/test_heavy_wq \
 	tests/lib/test_heavy \
+	tests/lib/test_idalloc \
 	tests/lib/test_memory \
 	tests/lib/test_nexthop_iter \
 	tests/lib/test_privs \
@@ -198,6 +199,9 @@ tests_lib_test_heavy_wq_CFLAGS = $(TESTS_CFLAGS)
 tests_lib_test_heavy_wq_CPPFLAGS = $(TESTS_CPPFLAGS)
 tests_lib_test_heavy_wq_LDADD = $(ALL_TESTS_LDADD) -lm
 tests_lib_test_heavy_wq_SOURCES = tests/lib/test_heavy_wq.c tests/helpers/c/main.c
+tests_lib_test_idalloc_CFLAGS = $(TESTS_CFLAGS)
+tests_lib_test_idalloc_LDADD = $(ALL_TESTS_LDADD)
+tests_lib_test_idalloc_SOURCES = tests/lib/test_idalloc.c
 tests_lib_test_memory_CFLAGS = $(TESTS_CFLAGS)
 tests_lib_test_memory_CPPFLAGS = $(TESTS_CPPFLAGS)
 tests_lib_test_memory_LDADD = $(ALL_TESTS_LDADD)