Add support for multiple instances of the same command

This patch adds support for starting and supervising multiple
services (run, task, or service) using an extra #ID number.

service #1 [2345] /sbin/httpd -f -h /http -p 80   -- Web server
service #2 [2345] /sbin/httpd -f -h /http -p 8080 -- Old web server

Also included in this patch is a fix for endless respawn of faulty
services.  For instance, a buggy daemon that crashes repeatedly will now
be stopped after 10 respawns.

Signed-off-by: Joachim Nilsson <troglobit@gmail.com>

README.md

@@ -206,6 +206,16 @@ the `/path/to/cmd` should be executed with.  Simply prefix the path with
 
     run [2345] @joe:users /usr/bin/logger "Hello world"
 
+For multiple instances of the same command, e.g. a DHCP client or
+multiple web servers, add `#ID` somewhere between the `run`, `task`,
+`service` keyword and the command, like this:
+
+    service #1 [2345] /sbin/httpd -f -h /http -p 80   -- Web server
+    service #2 [2345] /sbin/httpd -f -h /http -p 8080 -- Old web server
+
+Without the `#ID` to the service the latter will overwrite the former
+and only the old web server would be started and supervised.
+
 
 /etc/finit.d
 ------------

conf.c

@@ -214,7 +214,7 @@ static void parse_dynamic(char *line, time_t mtime)
 {
 	char *x;
 
-	/* Skip comments. */
+	/* Skip comments, i.e. lines beginning with # */
 	if (MATCH_CMD(line, "#", x))
 		return;
 

finit.conf

@@ -1,4 +1,8 @@
-# Sample Debian finit.conf
+# Sample finit.conf to demonstatre the syntax
+#
+# Comments are lines that start with #, comments are only allowed
+# on separate lines.  The # sign is used inside service/run/task
+# stanzas to denote instance ID, see below.
 user admin
 host testbed
 
@@ -41,3 +45,9 @@ inetd time/udp                wait [2345] internal          -- UNIX rdate servic
 inetd time/tcp              nowait [2345] internal          -- UNIX rdate service
 inetd ssh@eth0:222/tcp      nowait [2345] /usr/sbin/sshd -i -- SSH service
 inetd ssh/tcp               nowait [2345] /usr/sbin/sshd -i -- SSH service
+
+# For multiple instances of the same service, add #ID somewhere between
+# the service/run/task keyword and the command.
+service #1 [2345] /sbin/httpd -f -h /http -p 80   -- Web server
+service #2 [2345] /sbin/httpd -f -h /http -p 8080 -- Old web server
+

helpers.c

@@ -267,72 +267,10 @@ char *pid_get_name(pid_t pid, char *name, size_t len)
 	if (ret)
 		return NULL;
 
-	return name;
+	if (name)
-}
+		return name;
 
-/**
+	return pname;
- * procname_kill - Send a signal to a process group by name.
- * @name: Name of process to send signal to.
- * @signo: Signal to send.
- *
- * Send a signal to a running process (group). This function searches
- * for a given process @name and sends a signal, @signo, to each
- * process ID matching that @name.
- *
- * Returns:
- * Number of signals sent, i.e. number of processes who have received the signal.
- */
-int procname_kill(char *name, int signo)
-{
-	int result = 0;
-	char path[32], line[64];
-	FILE *fp;
-	DIR *dir;
-	struct dirent *entry;
-
-	if (!name) {
-		errno = EINVAL;
-		return 0;
-	}
-
-	dir = opendir("/proc");
-	if (!dir)
-		return 0;
-
-	while ((entry = readdir(dir)) != NULL) {
-		/* Skip non-process entries in /proc */
-		if (!isdigit(*entry->d_name))
-			continue;
-
-		snprintf(path, sizeof(path), "/proc/%s/status", entry->d_name);
-		/* Skip non-readable files (protected?) */
-		if ((fp = fopen(path, "r")) == NULL)
-			continue;
-
-		if (fgets(line, sizeof (line), fp)) {
-			char *pname = line + 6; /* Skip first part of line --> "Name:\t" */
-
-			if (strncmp(pname, name, strlen(pname) - 1) == 0) {
-				int error = errno;
-				int pid = atonum(entry->d_name);
-
-				if (-1 == pid) {
-					FLOG_ERROR("Failed converting %s to a PID: %s", entry->d_name, strerror(error));
-				} else {
-					if (kill(pid, signo))
-						FLOG_ERROR("Failed signalling(%d) %s: %s!", signo, name, strerror(error));
-					else
-						result++; /* Track number of processes we deliver the signal to. */
-				}
-			}
-		}
-
-		fclose(fp);
-	}
-
-	closedir(dir);
-
-	return result;
 }
 
 

helpers.h

@@ -103,7 +103,6 @@ int     pid_alive       (pid_t pid);
 char   *pid_get_name    (pid_t pid, char *name, size_t len);
 
 void    procname_set    (char *name, char *args[]);
-int     procname_kill   (char *name, int signo);
 
 void    print           (int action, const char *fmt, ...);
 void    print_desc      (char *action, char *desc);

plugin.c

@@ -87,9 +87,13 @@ int plugin_unregister(plugin_t *plugin)
 	TAILQ_REMOVE(&plugins, plugin, link);
 
 	if (plugin->svc.cb) {
-		svc_t *svc = svc_find(plugin->name);
+		svc_t *svc;
+
+		/* Unregister plugin callback for all matching services */
+		for (svc = svc_iterator(1); svc; svc = svc_iterator(0)) {
+			if (strcmp(svc->cmd, plugin->name))
+				continue;
 
-		if (svc) {
 			svc->cb           = NULL;
 			svc->dynamic      = 0;
 			svc->dynamic_stop = 0;

svc.c

@@ -36,6 +36,8 @@
 #include "svc.h"
 #include "inetd.h"
 
+#define RESPAWN_MAX    10	        /* Prevent endless respawn of faulty services. */
+
 /* The registered services are kept in shared memory for easy read-only access
  * by 3rd party APIs.  Mostly because of the ability to, from the outside, query
  * liveness state of all daemons that should run. */
@@ -58,12 +60,13 @@ static void __connect_shm(void)
 }
 
 /**
- * svc_new - Create a new service object
+ * svc_new - Create a new service
+ * @id: Instance id
  *
  * Returns:
- * A pointer to an empty &svc_t object, or %NULL if out of empty slots.
+ * A pointer to a new &svc_t object, or %NULL if out of empty slots.
  */
-svc_t *svc_new(void)
+svc_t *svc_new(int id)
 {
 	int i;
 
@@ -73,6 +76,8 @@ svc_t *svc_new(void)
 
 		if (svc->type == SVC_TYPE_FREE) {
 			memset(svc, 0, sizeof(*svc));
+			svc->id = id;
+
 			return svc;
 		}
 	}
@@ -127,12 +132,12 @@ svc_t *svc_iterator(int first)
  * Returns:
  * A pointer to an &svc_t object, or %NULL if not found.
  */
-svc_t *svc_find(char *path)
+svc_t *svc_find(char *path, int id)
 {
 	svc_t *svc;
 
 	for (svc = svc_iterator(1); svc; svc = svc_iterator(0)) {
-		if (!strncmp(path, svc->cmd, strlen(svc->cmd))) {
+		if (svc->id == id && !strncmp(path, svc->cmd, strlen(svc->cmd))) {
 			_d("Found a matching svc for %s", path);
 			return svc;
 		}
@@ -252,12 +257,12 @@ static void svc_dance(svc_t *svc)
 	svc_cmd_t cmd = svc_enabled(svc, 0, NULL);
 
 	if (svc->pid) {
-		if (cmd == SVC_STOP)
+		if (SVC_STOP == cmd)
 			svc_stop(svc);
-		else if (cmd == SVC_RELOAD)
+		else if (SVC_RELOAD == cmd)
 			svc_reload(svc);
 	} else {
-		if (cmd == SVC_START)
+		if (SVC_START == cmd || SVC_RELOAD == cmd)
 			svc_start(svc);
 	}
 }
@@ -336,18 +341,27 @@ void svc_runlevel(int newlevel)
  * The @line can optionally start with a username, denoted by an @
  * character. Like this:
  *
- *   "service @username [!0-6,S] /path/to/daemon arg -- Description text"
+ *     service @username [!0-6,S] /path/to/daemon arg       -- Description
- *   "task @username [!0-6,S] /path/to/task arg -- Description text"
+ *     task @username [!0-6,S] /path/to/task arg            -- Description
- *   "run  @username [!0-6,S] /path/to/cmd arg -- Description text"
+ *     run  @username [!0-6,S] /path/to/cmd arg             -- Description
- *   "inetd tcp/ssh nowait [2345] @root:root /usr/sbin/sshd -i -- Description"
+ *     inetd tcp/ssh nowait [2345] @root:root /sbin/sshd -i -- Description
  *
- * If the username is left out the command is started as root.
+ * If the username is left out the command is started as root.  The []
- * Inetd services, launched on demand (SVC_TYPE_INETD)
+ * brackets denote the allowed runlevels, if left out the default for a
- * The [] brackets are there to denote the allowed runlevels. Allowed
+ * service is set to [2-5].  Allowed runlevels mimic that of SysV init
- * runlevels mimic that of SysV init with the addition of the 'S'
+ * with the addition of the 'S' runlevel, which is only run once at
- * runlevel, which is only run once at startup. It can be seen as the
+ * startup.  It can be seen as the system bootstrap.  If a task or run
- * system bootstrap. If a task or run command is listed in more than the
+ * command is listed in more than the [S] runlevel they will be called
- * [S] runlevel they will be called when changing runlevel.
+ * when changing runlevel.
+ *
+ * For multiple instances of the same command, e.g. multiple DHCP
+ * clients, the user may enter an optional ID, using the #ID syntax.
+ *
+ *     service #1 /sbin/udhcpc -i eth1
+ *     service #2 /sbin/udhcpc -i eth2
+ *
+ * Without the #ID syntax Finit will overwrite the first service line
+ * with the contents of the second.  The #ID must be [1,MAXINT].
  *
  * Returns:
  * POSIX OK(0) on success, or non-zero errno exit status on failure.
@@ -355,6 +369,7 @@ void svc_runlevel(int newlevel)
 int svc_register(int type, char *line, time_t mtime, char *username)
 {
 	int i = 0;
+	int id = 1;		/* Default to ID:1 */
 #ifndef INETD_DISABLED
 	int forking = 0;
 #endif
@@ -392,6 +407,8 @@ int svc_register(int type, char *line, time_t mtime, char *username)
 			username = &cmd[1];
 		else if (cmd[0] == '[')	/* [runlevels] */
 			runlevels = &cmd[0];
+		else if (cmd[0] == '#')	/* #ID */
+			id = atoi(&cmd[1]);
 		else
 			break;
 
@@ -433,9 +450,10 @@ int svc_register(int type, char *line, time_t mtime, char *username)
 		return inetd_allow(&svc->inetd, iface);
 #endif
 
-	svc = svc_find(cmd);
+	svc = svc_find(cmd, id);
 	if (!svc) {
-		svc = svc_new();
+		_d("Creating new svc for %s id #%d", cmd, id);
+		svc = svc_new(id);
 		if (!svc) {
 			_e("Out of memory, cannot register service %s", cmd);
 			return errno = ENOMEM;
@@ -635,12 +653,14 @@ void svc_monitor(pid_t lost)
 			break;
 		}
 
-		/* Cleanup any lingering or semi-restarting tasks before respawning */
-		_d("Sending SIGTERM to service group %s", basename(svc->cmd));
-		procname_kill(basename(svc->cmd), SIGTERM);
-
 		/* Restarting lost service. */
 		if (svc_enabled(svc, 0, NULL)) {
+			if (svc->restart_counter > RESPAWN_MAX) {
+				_e("Not restarting %s id %d, respawn MAX (%d) reached!",
+				   svc->cmd, svc->id, RESPAWN_MAX);
+				break;
+			}
+
 			svc->restart_counter++;
 			svc_start(svc);
 		}
@@ -807,16 +827,6 @@ int svc_start(svc_t *svc)
 	return 0;
 }
 
-int svc_start_by_name(char *name)
-{
-	svc_t *svc = svc_find(name);
-
-	if (svc && svc_enabled(svc, 0, NULL))
-		return svc_start(svc);
-
-	return 1;
-}
-
 int svc_stop(svc_t *svc)
 {
 	int res = 1;
@@ -838,10 +848,13 @@ int svc_stop(svc_t *svc)
 
 	if (runlevel != 1)
 		print_desc("Stopping ", svc->desc);
-	_d("Sending SIGTERM to pid:%d name:'%s'", svc->pid, pid_get_name(svc->pid, NULL, 0));
+
+	_d("Sending SIGTERM to pid:%d name:%s", svc->pid, pid_get_name(svc->pid, NULL, 0));
 	res = kill(svc->pid, SIGTERM);
+
 	if (runlevel != 1)
 		print_result(res);
+
 	svc->pid = 0;
 	svc->restart_counter = 0;
 
@@ -890,8 +903,11 @@ void svc_reload_dynamic(void)
 {
 	svc_t *svc;
 
+	_d("Checking if any dynamically loaded services need to be reloaded ...");
 	for (svc = svc_iterator(1); svc; svc = svc_iterator(0)) {
 		if (svc->mtime && svc->dirty > 0) {
+			_d("Service %s id %d modified, must stop-start for changes to take effect!", svc->cmd, svc->id);
+			svc_stop(svc); /* This will reset the restart_counter as well! */
 			svc_dance(svc);
 			svc->dirty = 0;
 		}

svc.h

@@ -61,6 +61,10 @@ typedef enum {
  * of issuing an initctl call. E.g.
  *   initctl <stop|start|restart> service */
 typedef struct svc {
+	/* Instance specifics */
+	int            id;
+
+	/* Service details */
 	pid_t	       pid;
 	svc_type_t     type;
 	time_t	       mtime;	       /* Modification time for .conf from /etc/finit.d/ */
@@ -111,9 +115,9 @@ static inline svc_t *finit_svc_connect(void)
 	return (svc_t *)ptr;
 }
 
-svc_t	 *svc_new	    (void);
+svc_t	 *svc_new	    (int id);
 int	  svc_del	    (svc_t *svc);
-svc_t	 *svc_find	    (char *path);
+svc_t	 *svc_find	    (char *path, int id);
 svc_t    *svc_find_by_pid   (pid_t pid);
 svc_t    *svc_find_inetd    (char *path, char *service, char *proto, char *port);
 svc_t	 *svc_iterator	    (int first);
@@ -124,7 +128,6 @@ int	  svc_register	    (int type, char *line, time_t mtime, char *username);
 int	  svc_id_by_name    (char *name);
 svc_cmd_t svc_enabled	    (svc_t *svc, int event, void *arg);
 int	  svc_start	    (svc_t *svc);
-int	  svc_start_by_name (char *name);
 int	  svc_stop	    (svc_t *svc);
 int	  svc_reload	    (svc_t *svc);
 void      svc_reload_dynamic(void);