X-Git-Url: http://git.archive.openwrt.org/?p=project%2Fprocd.git;a=blobdiff_plain;f=instance.c;h=7895a04cc7b03cb74ff3749112bd75c5327e83b1;hp=d61bb332e2c7047b99f5d3d069576df63e224557;hb=eba428f6672068d819d6296db3f635e6ac5a8be7;hpb=85a5ce27f5f02e4ddab6916d69ba7f7375d289d4

diff --git a/instance.c b/instance.c
index d61bb33..7895a04 100644
--- a/instance.c
+++ b/instance.c
@@ -32,6 +32,7 @@ enum {
 	INSTANCE_ATTR_NETDEV,
 	INSTANCE_ATTR_FILE,
 	INSTANCE_ATTR_TRIGGER,
+	INSTANCE_ATTR_RESPAWN,
 	INSTANCE_ATTR_NICE,
 	__INSTANCE_ATTR_MAX
 };
@@ -43,6 +44,7 @@ static const struct blobmsg_policy instance_attr[__INSTANCE_ATTR_MAX] = {
 	[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 },
 };
 
@@ -102,6 +104,8 @@ instance_start(struct service_instance *in)
 		return;
 
 	in->restart = false;
+	in->halt = !in->respawn;
+
 	if (!in->valid)
 		return;
 
@@ -117,6 +121,7 @@ instance_start(struct service_instance *in)
 
 	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);
 }
 
@@ -126,29 +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);
-	DEBUG(1, "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);
 }
 
@@ -283,11 +321,31 @@ instance_config_parse(struct service_instance *in)
 		return false;
 
 	in->command = cur;
-	in->trigger = tb[INSTANCE_ATTR_TRIGGER];
 
-	if (in->trigger) {
+	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)
@@ -348,9 +406,9 @@ instance_update(struct service_instance *in, struct service_instance *in_new)
 			instance_config_move(in, in_new);
 		instance_start(in);
 	} else {
-		in->restart = true;
-		instance_stop(in, true);
+		instance_restart(in);
 		instance_config_move(in, in_new);
+		/* restart happens in the child callback handler */
 	}
 	return true;
 }
@@ -361,6 +419,7 @@ 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);
@@ -386,13 +445,44 @@ instance_init(struct service_instance *in, struct service *s, struct blob_attr *
 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);
 }