watchfrr, lib: cleanup & delay detaching

This cleans up watchfrr to be more "normal" like the other daemons in
terms of what it does in main(), i.e. using the full frr_*() call set.

Also, this changes the startup behaviour on watchfrr to stay attached on
the daemon's parent process until startup is really complete.  This
should allow removing the "watchfrr.started" hack at some point.

Signed-off-by: David Lamparter <equinox@diac24.net>

2 files changed, 69 insertions, 100 deletions

diff --git a/watchfrr/watchfrr.c b/watchfrr/watchfrr.c
index ec922490e..e32bf3359 100644
--- a/watchfrr/watchfrr.c
+++ b/watchfrr/watchfrr.c
@@ -55,9 +55,11 @@
 
 #define PING_TOKEN	"PING"
 
+DEFINE_MGROUP(WATCHFRR, "watchfrr")
+DEFINE_MTYPE_STATIC(WATCHFRR, WATCHFRR_DAEMON, "watchfrr daemon entry")
+
 /* Needs to be global, referenced somewhere inside libfrr. */
 struct thread_master *master;
-static char pidfile_default[256];
 
 static bool watch_only = false;
 
@@ -230,7 +232,7 @@ Otherwise, the interval is doubled (but capped at the -M value).\n\n",
 		name of the daemon should be substituted.\n\
     --dry	Do not start or restart anything, just log.\n\
 -p, --pid-file	Set process identifier file name\n\
-		(default is %s).\n\
+		(default is %s/watchfrr.pid).\n\
 -b, --blank-string\n\
 		When the supplied argument string is found in any of the\n\
 		various shell command arguments (-r, -s, or -k), replace\n\
@@ -240,7 +242,7 @@ Otherwise, the interval is doubled (but capped at the -M value).\n\n",
 -h, --help	Display this help and exit\n",
 		frr_vtydir, DEFAULT_LOGLEVEL, LOG_EMERG, LOG_DEBUG, LOG_DEBUG,
 		DEFAULT_MIN_RESTART, DEFAULT_MAX_RESTART, DEFAULT_PERIOD,
-		DEFAULT_TIMEOUT, DEFAULT_RESTART_TIMEOUT, pidfile_default);
+		DEFAULT_TIMEOUT, DEFAULT_RESTART_TIMEOUT, frr_vtydir);
 }
 
 static pid_t run_background(char *shell_cmd)
@@ -609,12 +611,13 @@ static void daemon_send_ready(void)
 	if (!sent && gs.numdown == 0) {
 		FILE *fp;
 
+		zlog_notice("all daemons up, doing startup-complete notify");
+		frr_detach();
+
 		fp = fopen(DAEMON_VTY_DIR "/watchfrr.started", "w");
 		if (fp)
 			fclose(fp);
 #if defined HAVE_SYSTEMD
-		zlog_notice(
-			"Watchfrr: Notifying Systemd we are up and running");
 		systemd_send_started(master, 0);
 #endif
 		sent = 1;
@@ -961,6 +964,53 @@ static char *translate_blanks(const char *cmd, const char *blankstr)
 	return res;
 }
 
+static void watchfrr_init(int argc, char **argv)
+{
+	const char *special = "zebra";
+	int i;
+	struct daemon *dmn, **add = &gs.daemons;
+	char alldaemons[512] = "", *p = alldaemons;
+
+	for (i = optind; i < argc; i++) {
+		dmn = XCALLOC(MTYPE_WATCHFRR_DAEMON, sizeof(*dmn));
+
+		dmn->name = dmn->restart.name = argv[i];
+		dmn->state = DAEMON_INIT;
+		gs.numdaemons++;
+		gs.numdown++;
+		dmn->fd = -1;
+		dmn->t_wakeup = NULL;
+		thread_add_timer_msec(master, wakeup_init, dmn,
+				      100 + (random() % 900),
+				      &dmn->t_wakeup);
+		dmn->restart.interval = gs.min_restart_interval;
+		*add = dmn;
+		add = &dmn->next;
+
+		if (!strcmp(dmn->name, special))
+			gs.special = dmn;
+	}
+
+	if (!gs.daemons) {
+		fprintf(stderr,
+			"Must specify one or more daemons to monitor.\n\n");
+		frr_help_exit(1);
+	}
+	if (!watch_only && !gs.special) {
+		fprintf(stderr, "\"%s\" daemon must be in daemon lists\n\n",
+			special);
+		frr_help_exit(1);
+	}
+
+	for (dmn = gs.daemons; dmn; dmn = dmn->next) {
+		snprintf(p, alldaemons + sizeof(alldaemons) - p, "%s%s",
+			 (p == alldaemons) ? "" : " ", dmn->name);
+		p += strlen(p);
+	}
+	zlog_notice("%s %s watching [%s]%s", progname, FRR_VERSION, alldaemons,
+		    watch_only ? ", monitor mode" : "");
+}
+
 struct zebra_privs_t watchfrr_privs = {
 #ifdef VTY_GROUP
 	.vty_group = VTY_GROUP,
@@ -984,7 +1034,8 @@ static struct quagga_signal_t watchfrr_signals[] = {
 
 FRR_DAEMON_INFO(watchfrr, WATCHFRR,
 		.flags = FRR_NO_PRIVSEP | FRR_NO_TCPVTY | FRR_LIMITED_CLI
-			 | FRR_NO_CFG_PID_DRY | FRR_NO_ZCLIENT,
+			 | FRR_NO_CFG_PID_DRY | FRR_NO_ZCLIENT
+			 | FRR_DETACH_LATER,
 
 		.printhelp = printhelp,
 		.copyright = "Copyright 2004 Andrew J. Schorr",
@@ -999,13 +1050,8 @@ FRR_DAEMON_INFO(watchfrr, WATCHFRR,
 int main(int argc, char **argv)
 {
 	int opt;
-	const char *pidfile = pidfile_default;
-	const char *special = "zebra";
 	const char *blankstr = NULL;
 
-	snprintf(pidfile_default, sizeof(pidfile_default), "%s/watchfrr.pid",
-		 frr_vtydir);
-
 	frr_preinit(&watchfrr_di, argc, argv);
 	progname = watchfrr_di.progname;
 
@@ -1087,7 +1133,7 @@ int main(int argc, char **argv)
 			gs.period = 1000 * period;
 		} break;
 		case 'p':
-			pidfile = optarg;
+			watchfrr_di.pid_file = optarg;
 			break;
 		case 'r':
 			if (!valid_command(optarg)) {
@@ -1167,99 +1213,18 @@ int main(int argc, char **argv)
 
 	master = frr_init();
 	watchfrr_error_init();
+	watchfrr_init(argc, argv);
+	watchfrr_vty_init();
+
+	frr_config_fork();
 
 	zlog_set_level(ZLOG_DEST_MONITOR, ZLOG_DISABLED);
-	if (watchfrr_di.daemon_mode) {
+	if (watchfrr_di.daemon_mode)
 		zlog_set_level(ZLOG_DEST_SYSLOG, MIN(gs.loglevel, LOG_DEBUG));
-		if (daemon(0, 0) < 0) {
-			fprintf(stderr, "Watchfrr daemon failed: %s",
-				strerror(errno));
-			exit(1);
-		}
-	} else
+	else
 		zlog_set_level(ZLOG_DEST_STDOUT, MIN(gs.loglevel, LOG_DEBUG));
 
-	watchfrr_vty_init();
-
-	frr_vty_serv();
-
-	{
-		int i;
-		struct daemon *tail = NULL;
-
-		for (i = optind; i < argc; i++) {
-			struct daemon *dmn;
-
-			if (!(dmn = (struct daemon *)calloc(1, sizeof(*dmn)))) {
-				fprintf(stderr, "calloc(1,%u) failed: %s\n",
-					(unsigned int)sizeof(*dmn),
-					safe_strerror(errno));
-				return 1;
-			}
-			dmn->name = dmn->restart.name = argv[i];
-			dmn->state = DAEMON_INIT;
-			gs.numdaemons++;
-			gs.numdown++;
-			dmn->fd = -1;
-			dmn->t_wakeup = NULL;
-			thread_add_timer_msec(master, wakeup_init, dmn,
-					      100 + (random() % 900),
-					      &dmn->t_wakeup);
-			dmn->restart.interval = gs.min_restart_interval;
-			if (tail)
-				tail->next = dmn;
-			else
-				gs.daemons = dmn;
-			tail = dmn;
-
-			if (!strcmp(dmn->name, special))
-				gs.special = dmn;
-		}
-	}
-	if (!gs.daemons) {
-		flog_err(EC_WATCHFRR_UNEXPECTED_DAEMONS,
-			  "Must specify one or more daemons to monitor.");
-		frr_help_exit(1);
-	}
-	if (!watch_only && !gs.special) {
-		flog_err(EC_WATCHFRR_UNEXPECTED_DAEMONS,
-			  "\"%s\" daemon must be in daemon lists", special);
-		frr_help_exit(1);
-	}
-
-	/* Make sure we're not already running. */
-	pid_output(pidfile);
-
-	/* Announce which daemons are being monitored. */
-	{
-		struct daemon *dmn;
-		size_t len = 0;
-
-		for (dmn = gs.daemons; dmn; dmn = dmn->next)
-			len += strlen(dmn->name) + 1;
-
-		{
-			char buf[len + 1];
-			char *p = buf;
-
-			for (dmn = gs.daemons; dmn; dmn = dmn->next) {
-				if (p != buf)
-					*p++ = ' ';
-				strcpy(p, dmn->name);
-				p += strlen(p);
-			}
-			zlog_notice("%s %s watching [%s]%s", progname,
-				    FRR_VERSION, buf,
-				    watch_only ? ", monitor mode" : "");
-		}
-	}
-
-	{
-		struct thread thread;
-
-		while (thread_fetch(master, &thread))
-			thread_call(&thread);
-	}
+	frr_run(master);
 
 	systemd_send_stopping();
 	/* Not reached. */
diff --git a/watchfrr/watchfrr.h b/watchfrr/watchfrr.h
index 1a1c19056..ee16846a1 100644
--- a/watchfrr/watchfrr.h
+++ b/watchfrr/watchfrr.h
@@ -21,6 +21,10 @@
 #ifndef FRR_WATCHFRR_H
 #define FRR_WATCHFRR_H
 
+#include "lib/memory.h"
+
+DECLARE_MGROUP(WATCHFRR)
+
 extern void watchfrr_vty_init(void);
 
 extern pid_t integrated_write_pid;