knet: Allow ping_timers to be auto-configured

knet ping_timers are auto-configured according to token value.

This patch also fixes some knet config bugs that resulted in defaults
not being applied when values were removed from corosync.conf.

Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>
Reviewed-by: Jan Friesse <jfriesse@redhat.com>

exec/totemconfig.c

@@ -699,6 +699,38 @@ static int find_local_node_in_nodelist(struct totem_config *totem_config)
 	return (local_node_pos);
 }
 
+/*
+ * This needs to be done last of all. It would be nice to do it when reading the
+ * interface params, but the totem params need to have them to be read first. We
+ * need both, so this is a way round that circular dependancy.
+ */
+static void calc_knet_ping_timers(struct totem_config *totem_config)
+{
+	char runtime_key_name[ICMAP_KEYNAME_MAXLEN];
+	int interface;
+
+	for (interface = 0; interface < INTERFACE_MAX; interface++) {
+
+		if (totem_config->interfaces[interface].configured) {
+			if (!totem_config->interfaces[interface].knet_ping_timeout) {
+				totem_config->interfaces[interface].knet_ping_timeout =
+					totem_config->token_timeout / totem_config->interfaces[interface].knet_pong_count;
+			}
+			snprintf(runtime_key_name, sizeof(runtime_key_name),
+				 "runtime.config.totem.interface.%d.knet_ping_timeout", interface);
+			icmap_set_uint32(runtime_key_name, totem_config->interfaces[interface].knet_ping_timeout);
+
+			if (!totem_config->interfaces[interface].knet_ping_interval) {
+				totem_config->interfaces[interface].knet_ping_interval =
+					totem_config->token_timeout / (totem_config->interfaces[interface].knet_pong_count * 2);
+			}
+			snprintf(runtime_key_name, sizeof(runtime_key_name),
+				 "runtime.config.totem.interface.%d.knet_ping_interval", interface);
+			icmap_set_uint32(runtime_key_name, totem_config->interfaces[interface].knet_ping_interval);
+		}
+	}
+}
+
 /*
  * Compute difference between two set of totem interface arrays. set1 and set2
  * are changed so for same ring, ip existing in both set1 and set2 are cleared
@@ -804,10 +836,9 @@ static void reconfigure_links(struct totem_config *totem_config)
 		/* In case this is a new link, fill in the defaults if there was no interface{} section for it */
 		if (!totem_config->interfaces[i].knet_link_priority)
 			totem_config->interfaces[i].knet_link_priority = 1;
-		if (!totem_config->interfaces[i].knet_ping_interval)
-			totem_config->interfaces[i].knet_ping_interval = KNET_PING_INTERVAL;
-		if (!totem_config->interfaces[i].knet_ping_timeout)
-			totem_config->interfaces[i].knet_ping_timeout = KNET_PING_TIMEOUT;
+
+		/* knet_ping_interval & knet_ping_timeout are set later once we know all the other params */
+
 		if (!totem_config->interfaces[i].knet_ping_precision)
 			totem_config->interfaces[i].knet_ping_precision = KNET_PING_PRECISION;
 		if (!totem_config->interfaces[i].knet_pong_count)
@@ -1103,7 +1134,16 @@ static int get_interface_params(struct totem_config *totem_config,
 
 	if (reload) {
 		for (i=0; i<INTERFACE_MAX; i++) {
+			/*
+			 * Set back to defaults things that might have been configured and
+			 * now have been taken out of corosync.conf. These won't be caught by the
+			 * code below which only looks at interface{} sections that actually exist.
+			 */
 			totem_config->interfaces[i].configured = 0;
+			totem_config->interfaces[i].knet_ping_timeout = 0;
+			totem_config->interfaces[i].knet_ping_interval = 0;
+			totem_config->interfaces[i].knet_ping_precision = KNET_PING_PRECISION;
+			totem_config->interfaces[i].knet_pong_count = KNET_PONG_COUNT;
 		}
 	}
 	if (icmap_get_string("totem.cluster_name", &cluster_name) != CS_OK) {
@@ -1228,12 +1268,12 @@ static int get_interface_params(struct totem_config *totem_config,
 			totem_config->interfaces[linknumber].knet_link_priority = u8;
 		}
 
-		totem_config->interfaces[linknumber].knet_ping_interval = KNET_PING_INTERVAL;
+		totem_config->interfaces[linknumber].knet_ping_interval = 0; /* real default applied later */
 		snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "totem.interface.%u.knet_ping_interval", linknumber);
 		if (icmap_get_uint32(tmp_key, &u32) == CS_OK) {
 			totem_config->interfaces[linknumber].knet_ping_interval = u32;
 		}
-		totem_config->interfaces[linknumber].knet_ping_timeout = KNET_PING_TIMEOUT;
+		totem_config->interfaces[linknumber].knet_ping_timeout = 0; /* real default applied later */
 		snprintf(tmp_key, ICMAP_KEYNAME_MAXLEN, "totem.interface.%u.knet_ping_timeout", linknumber);
 		if (icmap_get_uint32(tmp_key, &u32) == CS_OK) {
 			totem_config->interfaces[linknumber].knet_ping_timeout = u32;
@@ -1460,6 +1500,8 @@ extern int totem_config_read (
 	 */
 	totem_volatile_config_read(totem_config, NULL);
 
+	calc_knet_ping_timers(totem_config);
+
 	icmap_set_uint8("config.totemconfig_reload_in_progress", 0);
 
 	add_totem_config_notification(totem_config);
@@ -1855,6 +1897,9 @@ static void totem_reload_notify(
 		get_interface_params(totem_config, &error_string, &warnings, 1);
 		put_nodelist_members_to_config (totem_config, 1);
 		totem_volatile_config_read (totem_config, NULL);
+
+		calc_knet_ping_timers(totem_config);
+
 		log_printf(LOGSYS_LEVEL_DEBUG, "Configuration reloaded. Dumping actual totem config.");
 		debug_dump_totem_config(totem_config);
 		if (totem_volatile_config_validate(totem_config, &error_string) == -1) {

exec/totemknet.c

@@ -656,6 +656,8 @@ static void timer_function_netif_check_timeout (
 						     i);
 	}
 }
+
+/* NOTE: this relies on the fact that totem_reload_notify() is called first */
 static void totemknet_refresh_config(
 	int32_t event,
 	const char *key_name,
@@ -670,7 +672,6 @@ static void totemknet_refresh_config(
 	knet_node_id_t host_ids[KNET_MAX_HOST];
 	int i;
 	int err;
-	char path[ICMAP_KEYNAME_MAXLEN];
 	struct totemknet_instance *instance = (struct totemknet_instance *)user_data;
 
 	ENTER();
@@ -693,42 +694,6 @@ static void totemknet_refresh_config(
 		}
 	}
 
-	/* Get link parameters */
-	for (i = 0; i < INTERFACE_MAX; i++) {
-		if (!instance->totem_config->interfaces[i].configured) {
-			continue;
-		}
-
-		sprintf(path, "totem.interface.%d.knet_link_priority", i);
-		if (icmap_get_uint32(path, &value) == CS_OK) {
-			instance->totem_config->interfaces[i].knet_link_priority = value;
-			knet_log_printf (LOGSYS_LEVEL_DEBUG, "knet_link_priority on link %d now %d", i, value);
-		}
-
-		sprintf(path, "totem.interface.%d.knet_ping_interval", i);
-		if (icmap_get_uint32(path, &value) == CS_OK) {
-			instance->totem_config->interfaces[i].knet_ping_interval = value;
-			knet_log_printf (LOGSYS_LEVEL_DEBUG, "knet_ping_interval on link %d now %d", i, value);
-		}
-
-		sprintf(path, "totem.interface.%d.knet_ping_timeout", i);
-		if (icmap_get_uint32(path, &value) == CS_OK) {
-			instance->totem_config->interfaces[i].knet_ping_timeout = value;
-			knet_log_printf (LOGSYS_LEVEL_DEBUG, "knet_ping_timeout on link %d now %d", i, value);
-		}
-		sprintf(path, "totem.interface.%d.knet_ping_precision", i);
-		if (icmap_get_uint32(path, &value) == CS_OK) {
-			instance->totem_config->interfaces[i].knet_ping_precision = value;
-			knet_log_printf (LOGSYS_LEVEL_DEBUG, "knet_ping_precision on link %d now %d", i, value);
-		}
-
-		sprintf(path, "totem.interface.%d.knet_pong_count", i);
-		if (icmap_get_uint32(path, &value) == CS_OK) {
-			instance->totem_config->interfaces[i].knet_pong_count = value;
-			knet_log_printf (LOGSYS_LEVEL_DEBUG, "knet_pong_count on link %d now %d", i, value);
-		}
-	}
-
 	/* Configure link parameters for each node */
 	err = knet_host_get_host_list(instance->knet_handle, host_ids, &num_nodes);
 	if (err != 0) {

man/corosync.conf.5

@@ -89,12 +89,18 @@ mode. (see link_mode below)
 .TP
 knet_ping_interval
 This specifies the interval between knet link pings. 
-(default 1000 ms)
+knet_ping_interval and knet_ping_timeout
+are a pair, if one is specified the other should be too, otherwise one will be calculated from
+the token timeout and one will be taken from the config file.
+(default is token timeout / (knet_pong_count*2))
 
 .TP
 knet_ping_timeout
 If no ping is received within this time, the knet link is declared dead. 
-(default 2000 ms)
+knet_ping_interval and knet_ping_timeout
+are a pair, if one is specified the other should be too, otherwise one will be calculated from
+the token timeout and one will be taken from the config file.
+(default is token timeout / knet_pong_count)
 
 .TP
 knet_ping_precision