Merge pull request #11417 from donaldsharp/nhg_timer

Nhg timer

doc/user/zebra.rst

@@ -273,6 +273,12 @@ Nexthop tracking doesn't resolve nexthops via the default route by default.
 Allowing this might be useful when e.g. you want to allow BGP to peer across
 the default route.
 
+.. clicmd:: zebra nexthop-group keep (1-3600)
+
+   Set the time that zebra will keep a created and installed nexthop group
+   before removing it from the system if the nexthop group is no longer
+   being used.  The default time is 180 seconds.
+
 .. clicmd:: ip nht resolve-via-default
 
    Allow IPv4 nexthop tracking to resolve via the default route. This parameter

zebra/interface.c

@@ -184,13 +184,6 @@ static int if_zebra_new_hook(struct interface *ifp)
 static void if_nhg_dependents_check_valid(struct nhg_hash_entry *nhe)
 {
 	zebra_nhg_check_valid(nhe);
-	if (!CHECK_FLAG(nhe->flags, NEXTHOP_GROUP_VALID)) {
-		/* If we're in shutdown, this interface event needs to clean
-		 * up installed NHGs, so don't clear that flag directly.
-		 */
-		if (!zrouter.in_shutdown)
-			UNSET_FLAG(nhe->flags, NEXTHOP_GROUP_INSTALLED);
-	}
 }
 
 static void if_down_nhg_dependents(const struct interface *ifp)

zebra/zebra_nhg.c

@@ -1055,6 +1055,12 @@ static void zebra_nhg_set_invalid(struct nhg_hash_entry *nhe)
 
 	UNSET_FLAG(nhe->flags, NEXTHOP_GROUP_VALID);
 
+	/* If we're in shutdown, this interface event needs to clean
+	 * up installed NHGs, so don't clear that flag directly.
+	 */
+	if (!zrouter.in_shutdown)
+		UNSET_FLAG(nhe->flags, NEXTHOP_GROUP_INSTALLED);
+
 	/* Update validity of nexthops depending on it */
 	frr_each(nhg_connected_tree, &nhe->nhg_dependents, rb_node_dep)
 		zebra_nhg_check_valid(rb_node_dep->nhe);
@@ -1619,6 +1625,17 @@ void zebra_nhg_hash_free(void *p)
 	zebra_nhg_free((struct nhg_hash_entry *)p);
 }
 
+static void zebra_nhg_timer(struct thread *thread)
+{
+	struct nhg_hash_entry *nhe = THREAD_ARG(thread);
+
+	if (IS_ZEBRA_DEBUG_NHG_DETAIL)
+		zlog_debug("Nexthop Timer for nhe: %pNG", nhe);
+
+	if (nhe->refcnt == 1)
+		zebra_nhg_decrement_ref(nhe);
+}
+
 void zebra_nhg_decrement_ref(struct nhg_hash_entry *nhe)
 {
 	if (IS_ZEBRA_DEBUG_NHG_DETAIL)
@@ -1627,6 +1644,15 @@ void zebra_nhg_decrement_ref(struct nhg_hash_entry *nhe)
 
 	nhe->refcnt--;
 
+	if (!zrouter.in_shutdown && nhe->refcnt <= 0 &&
+	    CHECK_FLAG(nhe->flags, NEXTHOP_GROUP_INSTALLED) &&
+	    !CHECK_FLAG(nhe->flags, NEXTHOP_GROUP_KEEP_AROUND)) {
+		nhe->refcnt = 1;
+		SET_FLAG(nhe->flags, NEXTHOP_GROUP_KEEP_AROUND);
+		thread_add_timer(zrouter.master, zebra_nhg_timer, nhe,
+				 zrouter.nhg_keep, &nhe->timer);
+	}
+
 	if (!zebra_nhg_depends_is_empty(nhe))
 		nhg_connected_tree_decrement_ref(&nhe->nhg_depends);
 
@@ -1642,6 +1668,12 @@ void zebra_nhg_increment_ref(struct nhg_hash_entry *nhe)
 
 	nhe->refcnt++;
 
+	if (thread_is_scheduled(nhe->timer)) {
+		THREAD_OFF(nhe->timer);
+		nhe->refcnt--;
+		UNSET_FLAG(nhe->flags, NEXTHOP_GROUP_KEEP_AROUND);
+	}
+
 	if (!zebra_nhg_depends_is_empty(nhe))
 		nhg_connected_tree_increment_ref(&nhe->nhg_depends);
 }
@@ -3290,9 +3322,6 @@ struct nhg_hash_entry *zebra_nhg_proto_add(uint32_t id, int type,
 
 		rib_handle_nhg_replace(old, new);
 
-		/* if this != 1 at this point, we have a bug */
-		assert(old->refcnt == 1);
-
 		/* We have to decrement its singletons
 		 * because some might not exist in NEW.
 		 */
@@ -3304,6 +3333,7 @@ struct nhg_hash_entry *zebra_nhg_proto_add(uint32_t id, int type,
 
 		/* Dont call the dec API, we dont want to uninstall the ID */
 		old->refcnt = 0;
+		THREAD_OFF(old->timer);
 		zebra_nhg_free(old);
 		old = NULL;
 	}

zebra/zebra_nhg.h

@@ -79,16 +79,34 @@ struct nhg_hash_entry {
 
 	uint32_t flags;
 
-	/* Dependency tree for other entries.
+	/* Dependency trees for other entries.
 	 * For instance a group with two
 	 * nexthops will have two dependencies
 	 * pointing to those nhg_hash_entries.
 	 *
 	 * Using a rb tree here to make lookups
 	 * faster with ID's.
+	 *
+	 * nhg_depends the RB tree of entries that this
+	 * group contains.
+	 *
+	 * nhg_dependents the RB tree of entries that
+	 * this group is being used by
+	 *
+	 * NHG id 3 with nexthops id 1/2
+	 * nhg(3)->nhg_depends has 1 and 2 in the tree
+	 * nhg(3)->nhg_dependents is empty
+	 *
+	 * nhg(1)->nhg_depends is empty
+	 * nhg(1)->nhg_dependents is 3 in the tree
+	 *
+	 * nhg(2)->nhg_depends is empty
+	 * nhg(3)->nhg_dependents is 3 in the tree
 	 */
 	struct nhg_connected_tree_head nhg_depends, nhg_dependents;
 
+	struct thread *timer;
+
 /*
  * Is this nexthop group valid, ie all nexthops are fully resolved.
  * What is fully resolved?  It's a nexthop that is either self contained
@@ -129,6 +147,15 @@ struct nhg_hash_entry {
  */
 #define NEXTHOP_GROUP_PROTO_RELEASED (1 << 5)
 
+/*
+ * When deleting a NHG notice that it is still installed
+ * and if it is, slightly delay the actual removal to
+ * the future.  So that upper level protocols might
+ * be able to take advantage of some NHG's that
+ * are there
+ */
+#define NEXTHOP_GROUP_KEEP_AROUND (1 << 6)
+
 /*
  * Track FPM installation status..
  */

zebra/zebra_router.c

@@ -278,6 +278,8 @@ void zebra_router_init(bool asic_offload, bool notify_on_ack)
 
 	zrouter.packets_to_process = ZEBRA_ZAPI_PACKETS_TO_PROCESS;
 
+	zrouter.nhg_keep = ZEBRA_DEFAULT_NHG_KEEP_TIMER;
+
 	zebra_vxlan_init();
 	zebra_mlag_init();
 

zebra/zebra_router.h

@@ -219,6 +219,9 @@ struct zebra_router {
 	bool notify_on_ack;
 
 	bool supports_nhgs;
+
+#define ZEBRA_DEFAULT_NHG_KEEP_TIMER 180
+	uint32_t nhg_keep;
 };
 
 #define GRACEFUL_RESTART_TIME 60

zebra/zebra_vty.c

@@ -1433,14 +1433,22 @@ static void show_nexthop_group_out(struct vty *vty, struct nhg_hash_entry *nhe)
 	struct nhg_connected *rb_node_dep = NULL;
 	struct nexthop_group *backup_nhg;
 	char up_str[MONOTIME_STRLEN];
+	char time_left[MONOTIME_STRLEN];
 
 	uptime2str(nhe->uptime, up_str, sizeof(up_str));
 
 	vty_out(vty, "ID: %u (%s)\n", nhe->id, zebra_route_string(nhe->type));
-	vty_out(vty, "     RefCnt: %u\n", nhe->refcnt);
+	vty_out(vty, "     RefCnt: %u", nhe->refcnt);
+	if (thread_is_scheduled(nhe->timer))
+		vty_out(vty, " Time to Deletion: %s",
+			thread_timer_to_hhmmss(time_left, sizeof(time_left),
+					       nhe->timer));
+	vty_out(vty, "\n");
+
 	vty_out(vty, "     Uptime: %s\n", up_str);
 	vty_out(vty, "     VRF: %s\n", vrf_id_to_name(nhe->vrf_id));
 
+
 	if (CHECK_FLAG(nhe->flags, NEXTHOP_GROUP_VALID)) {
 		vty_out(vty, "     Valid");
 		if (CHECK_FLAG(nhe->flags, NEXTHOP_GROUP_INSTALLED))
@@ -3842,11 +3850,31 @@ DEFUN (no_ip_zebra_import_table,
 	return (zebra_import_table(AFI_IP, VRF_DEFAULT, table_id, 0, NULL, 0));
 }
 
+DEFPY (zebra_nexthop_group_keep,
+       zebra_nexthop_group_keep_cmd,
+       "[no] zebra nexthop-group keep (1-3600)",
+       NO_STR
+       ZEBRA_STR
+       "Nexthop-Group\n"
+       "How long to keep\n"
+       "Time in seconds from 1-3600\n")
+{
+	if (no)
+		zrouter.nhg_keep = ZEBRA_DEFAULT_NHG_KEEP_TIMER;
+	else
+		zrouter.nhg_keep = keep;
+
+	return CMD_SUCCESS;
+}
+
 static int config_write_protocol(struct vty *vty)
 {
 	if (allow_delete)
 		vty_out(vty, "allow-external-route-update\n");
 
+	if (zrouter.nhg_keep != ZEBRA_DEFAULT_NHG_KEEP_TIMER)
+		vty_out(vty, "zebra nexthop-group keep %u\n", zrouter.nhg_keep);
+
 	if (zrouter.ribq->spec.hold != ZEBRA_RIB_PROCESS_HOLD_TIME)
 		vty_out(vty, "zebra work-queue %u\n", zrouter.ribq->spec.hold);
 
@@ -4425,6 +4453,7 @@ void zebra_vty_init(void)
 	install_element(CONFIG_NODE, &ip_multicast_mode_cmd);
 	install_element(CONFIG_NODE, &no_ip_multicast_mode_cmd);
 
+	install_element(CONFIG_NODE, &zebra_nexthop_group_keep_cmd);
 	install_element(CONFIG_NODE, &ip_zebra_import_table_distance_cmd);
 	install_element(CONFIG_NODE, &no_ip_zebra_import_table_cmd);
 	install_element(CONFIG_NODE, &zebra_workqueue_timer_cmd);