trivial code cleanup

[project/procd.git] / instance.c

diff --git a/instance.c b/instance.c
index a1459b7..d9530e5 100644 (file)
--- 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 +131,58 @@ 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 < RESPAWN_ERROR)
+                       in->respawn_count++;
+               else
+                       in->respawn_count = 0;
+               if (in->respawn_count > 5)
+                       DEBUG(1, "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);
+               uloop_timeout_set(&in->timeout, 5000);
+       }
 }
 
 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);
 }
 
@@ -272,6 +317,10 @@ instance_config_parse(struct service_instance *in)
                return false;
 
        in->command = cur;
+       in->trigger = tb[INSTANCE_ATTR_TRIGGER];
+
+       if (in->trigger)
+               trigger_add(in->trigger, in);
 
        if ((cur = tb[INSTANCE_ATTR_NICE])) {
                in->nice = (int8_t) blobmsg_get_u32(cur);
@@ -309,6 +358,7 @@ 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;
@@ -332,9 +382,9 @@ 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;
 }
@@ -344,6 +394,7 @@ instance_free(struct service_instance *in)
 {
        uloop_process_delete(&in->proc);
        uloop_timeout_cancel(&in->timeout);
+       trigger_del(in);
        instance_config_cleanup(in);
        free(in->config);
        free(in);
@@ -358,6 +409,8 @@ instance_init(struct service_instance *in, struct service *s, struct blob_attr *
        in->config = config;
        in->timeout.cb = instance_timeout;
        in->proc.cb = instance_exit;
+       in->respawn = true;
+       in->respawn_count = 0;
 
        blobmsg_list_init(&in->netdev, struct instance_netdev, node, instance_netdev_cmp);
        blobmsg_list_init(&in->file, struct instance_file, node, instance_file_cmp);
@@ -366,7 +419,7 @@ 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;
 
@@ -375,5 +428,7 @@ void instance_dump(struct blob_buf *b, struct service_instance *in)
        if (in->proc.pending)
                blobmsg_add_u32(b, "pid", in->proc.pid);
        blobmsg_add_blob(b, in->command);
+       if (verbose && in->trigger)
+               blobmsg_add_blob(b, in->trigger);
        blobmsg_close_table(b, i);
 }