X-Git-Url: https://git.archive.openwrt.org/?p=project%2Fprocd.git;a=blobdiff_plain;f=instance.c;h=7895a04cc7b03cb74ff3749112bd75c5327e83b1;hp=b2cdfa1dd57397d4e36d7886d8f5bafa5e710f23;hb=8fd40e8352bac9992ca4ebd50998d4e961f5404b;hpb=8158e9052afbf2ee47de645167b90b6e7fc7d86d

diff --git a/instance.c b/instance.c
index b2cdfa1..7895a04 100644
--- a/instance.c
+++ b/instance.c
@@ -31,6 +31,8 @@ enum {
 	INSTANCE_ATTR_DATA,
 	INSTANCE_ATTR_NETDEV,
 	INSTANCE_ATTR_FILE,
+	INSTANCE_ATTR_TRIGGER,
+	INSTANCE_ATTR_RESPAWN,
 	INSTANCE_ATTR_NICE,
 	__INSTANCE_ATTR_MAX
 };
@@ -41,6 +43,8 @@ static const struct blobmsg_policy instance_attr[__INSTANCE_ATTR_MAX] = {
 	[INSTANCE_ATTR_DATA] = { "data", BLOBMSG_TYPE_TABLE },
 	[INSTANCE_ATTR_NETDEV] = { "netdev", BLOBMSG_TYPE_ARRAY },
 	[INSTANCE_ATTR_FILE] = { "file", BLOBMSG_TYPE_ARRAY },
+	[INSTANCE_ATTR_TRIGGER] = { "triggers", BLOBMSG_TYPE_ARRAY },
+	[INSTANCE_ATTR_RESPAWN] = { "respawn", BLOBMSG_TYPE_ARRAY },
 	[INSTANCE_ATTR_NICE] = { "nice", BLOBMSG_TYPE_INT32 },
 };
 
@@ -61,7 +65,7 @@ instance_run(struct service_instance *in)
 	struct blob_attr *cur;
 	char **argv;
 	int argc = 1; /* NULL terminated */
-	int rem;
+	int rem, fd;
 
 	if (in->nice)
 		setpriority(PRIO_PROCESS, 0, in->nice);
@@ -79,6 +83,14 @@ instance_run(struct service_instance *in)
 		argv[argc++] = blobmsg_data(cur);
 
 	argv[argc] = NULL;
+	fd = open("/dev/null", O_RDWR);
+	if (fd > -1) {
+		dup2(fd, STDIN_FILENO);
+		dup2(fd, STDOUT_FILENO);
+		dup2(fd, STDERR_FILENO);
+		if (fd > STDERR_FILENO)
+			close(fd);
+	}
 	execvp(argv[0], argv);
 	exit(127);
 }
@@ -92,6 +104,8 @@ instance_start(struct service_instance *in)
 		return;
 
 	in->restart = false;
+	in->halt = !in->respawn;
+
 	if (!in->valid)
 		return;
 
@@ -100,12 +114,14 @@ instance_start(struct service_instance *in)
 		return;
 
 	if (!pid) {
+		uloop_done();
 		instance_run(in);
 		return;
 	}
 
 	DEBUG(1, "Started instance %s::%s\n", in->srv->name, in->name);
 	in->proc.pid = pid;
+	clock_gettime(CLOCK_MONOTONIC, &in->start);
 	uloop_process_add(&in->proc);
 }
 
@@ -115,29 +131,62 @@ instance_timeout(struct uloop_timeout *t)
 	struct service_instance *in;
 
 	in = container_of(t, struct service_instance, timeout);
-	kill(in->proc.pid, SIGKILL);
-	uloop_process_delete(&in->proc);
-	in->proc.cb(&in->proc, -1);
+
+	if (!in->halt && (in->restart || in->respawn))
+		instance_start(in);
 }
 
 static void
 instance_exit(struct uloop_process *p, int ret)
 {
 	struct service_instance *in;
+	struct timespec tp;
+	long runtime;
 
 	in = container_of(p, struct service_instance, proc);
-	LOG("Instance %s::%s exit with error code %d\n", in->srv->name, in->name, ret);
+
+	clock_gettime(CLOCK_MONOTONIC, &tp);
+	runtime = tp.tv_sec - in->start.tv_sec;
+
+	DEBUG(1, "Instance %s::%s exit with error code %d after %ld seconds\n", in->srv->name, in->name, ret, runtime);
 	uloop_timeout_cancel(&in->timeout);
-	if (in->restart)
+	if (in->halt) {
+		/* no action */
+	} else if (in->restart) {
 		instance_start(in);
+	} else if (in->respawn) {
+		if (runtime < in->respawn_threshold)
+			in->respawn_count++;
+		else
+			in->respawn_count = 0;
+		if (in->respawn_count > in->respawn_retry) {
+			LOG("Instance %s::%s s in a crash loop %d crashes, %ld seconds since last crash\n",
+								in->srv->name, in->name, in->respawn_count, runtime);
+			in->restart = in->respawn = 0;
+			in->halt = 1;
+		} else {
+			uloop_timeout_set(&in->timeout, in->respawn_timeout * 1000);
+		}
+	}
 }
 
 void
-instance_stop(struct service_instance *in, bool restart)
+instance_stop(struct service_instance *in)
 {
 	if (!in->proc.pending)
 		return;
+	in->halt = true;
+	in->restart = in->respawn = false;
+	kill(in->proc.pid, SIGTERM);
+}
 
+static void
+instance_restart(struct service_instance *in)
+{
+	if (!in->proc.pending)
+		return;
+	in->halt = false;
+	in->restart = true;
 	kill(in->proc.pid, SIGTERM);
 }
 
@@ -273,6 +322,30 @@ instance_config_parse(struct service_instance *in)
 
 	in->command = cur;
 
+	if (tb[INSTANCE_ATTR_RESPAWN]) {
+		int i = 0;
+		uint32_t vals[3] = { 3600, 5, 5};
+
+		blobmsg_for_each_attr(cur2, tb[INSTANCE_ATTR_RESPAWN], rem) {
+			if ((i >= 3) && (blobmsg_type(cur2) == BLOBMSG_TYPE_STRING))
+				continue;
+			vals[i] = atoi(blobmsg_get_string(cur2));
+			i++;
+		}
+		in->respawn = true;
+		in->respawn_count = 0;
+		in->respawn_threshold = vals[0];
+		in->respawn_timeout = vals[1];
+		in->respawn_retry = vals[2];
+	}
+	if (tb[INSTANCE_ATTR_TRIGGER]) {
+		in->trigger = malloc(blob_pad_len(tb[INSTANCE_ATTR_TRIGGER]));
+		if (!in->trigger)
+			return -1;
+		memcpy(in->trigger, tb[INSTANCE_ATTR_TRIGGER], blob_pad_len(tb[INSTANCE_ATTR_TRIGGER]));
+		trigger_add(in->trigger, in);
+	}
+
 	if ((cur = tb[INSTANCE_ATTR_NICE])) {
 		in->nice = (int8_t) blobmsg_get_u32(cur);
 		if (in->nice < -20 || in->nice > 20)
@@ -309,6 +382,7 @@ instance_config_move(struct service_instance *in, struct service_instance *in_sr
 	blobmsg_list_move(&in->env, &in_src->env);
 	blobmsg_list_move(&in->data, &in_src->data);
 	blobmsg_list_move(&in->netdev, &in_src->netdev);
+	in->trigger = in_src->trigger;
 	in->command = in_src->command;
 	in->name = in_src->name;
 	in->node.avl.key = in_src->node.avl.key;
@@ -322,13 +396,20 @@ bool
 instance_update(struct service_instance *in, struct service_instance *in_new)
 {
 	bool changed = instance_config_changed(in, in_new);
+	bool running = in->proc.pending;
 
-	if (!changed)
+	if (!changed && running)
 		return false;
 
-	in->restart = true;
-	instance_stop(in, true);
-	instance_config_move(in, in_new);
+	if (!running) {
+		if (changed)
+			instance_config_move(in, in_new);
+		instance_start(in);
+	} else {
+		instance_restart(in);
+		instance_config_move(in, in_new);
+		/* restart happens in the child callback handler */
+	}
 	return true;
 }
 
@@ -337,6 +418,8 @@ instance_free(struct service_instance *in)
 {
 	uloop_process_delete(&in->proc);
 	uloop_timeout_cancel(&in->timeout);
+	trigger_del(in);
+	free(in->trigger);
 	instance_config_cleanup(in);
 	free(in->config);
 	free(in);
@@ -359,14 +442,47 @@ instance_init(struct service_instance *in, struct service *s, struct blob_attr *
 	in->valid = instance_config_parse(in);
 }
 
-void instance_dump(struct blob_buf *b, struct service_instance *in)
+void instance_dump(struct blob_buf *b, struct service_instance *in, int verbose)
 {
 	void *i;
+	struct pid_info pi;
 
 	i = blobmsg_open_table(b, in->name);
 	blobmsg_add_u8(b, "running", in->proc.pending);
 	if (in->proc.pending)
 		blobmsg_add_u32(b, "pid", in->proc.pid);
 	blobmsg_add_blob(b, in->command);
+
+	if (!avl_is_empty(&in->env.avl)) {
+		struct blobmsg_list_node *var;
+		void *e = blobmsg_open_table(b, "env");
+		blobmsg_list_for_each(&in->env, var)
+			blobmsg_add_string(b, blobmsg_name(var->data), blobmsg_data(var->data));
+		blobmsg_close_table(b, e);
+	}
+
+	if (in->respawn) {
+		void *r = blobmsg_open_table(b, "respawn");
+		blobmsg_add_u32(b, "timeout", in->respawn_timeout);
+		blobmsg_add_u32(b, "threshold", in->respawn_threshold);
+		blobmsg_add_u32(b, "retry", in->respawn_retry);
+		blobmsg_close_table(b, r);
+	}
+
+	if (verbose && in->trigger)
+		blobmsg_add_blob(b, in->trigger);
+	if (!measure_process(in->proc.pid, &pi)) {
+		struct timespec tp;
+		long uptime;
+
+		clock_gettime(CLOCK_MONOTONIC, &tp);
+		uptime = tp.tv_sec - in->start.tv_sec;
+
+		blobmsg_add_u8(b, "ppid", pi.ppid);
+		blobmsg_add_u16(b, "uid", pi.uid);
+		blobmsg_add_u32(b, "fdcount", pi.fdcount);
+		blobmsg_add_u32(b, "vmsize", pi.vmsize);
+		blobmsg_add_u32(b, "uptime", uptime);
+	}
 	blobmsg_close_table(b, i);
 }