Fix: failure to launch agent thread is not reported

[deliverable/lttng-tools.git] / src / bin / lttng-sessiond / main.c

diff --git a/src/bin/lttng-sessiond/main.c b/src/bin/lttng-sessiond/main.c
index 0df0b1f2e04bc4601149088740b6feec2b0bfdb2..cd2f07b60e8f0e6cce6556dab502a161cf351157 100644 (file)
--- a/src/bin/lttng-sessiond/main.c
+++ b/src/bin/lttng-sessiond/main.c
@@ -313,7 +313,7 @@ struct lttng_ht *agent_apps_ht_by_sock = NULL;
  * NR_LTTNG_SESSIOND_READY must match the number of calls to
  * sessiond_notify_ready().
  */
-#define NR_LTTNG_SESSIOND_READY                3
+#define NR_LTTNG_SESSIOND_READY                4
 int lttng_sessiond_ready = NR_LTTNG_SESSIOND_READY;
 
 int sessiond_check_thread_quit_pipe(int fd, uint32_t events)
@@ -1868,12 +1868,16 @@ static void *thread_dispatch_ust_registration(void *data)
 
        DBG("[thread] Dispatch UST command started");
 
-       while (!CMM_LOAD_SHARED(dispatch_thread_exit)) {
+       for (;;) {
                health_code_update();
 
                /* Atomically prepare the queue futex */
                futex_nto1_prepare(&ust_cmd_queue.futex);
 
+               if (CMM_LOAD_SHARED(dispatch_thread_exit)) {
+                       break;
+               }
+
                do {
                        struct ust_app *app = NULL;
                        ust_cmd = NULL;
@@ -2522,7 +2526,7 @@ static pid_t spawn_consumerd(struct consumer_data *consumer_data)
                        } else {
                                DBG("Could not find any valid consumerd executable");
                                ret = -EINVAL;
-                               break;
+                               goto error;
                        }
                        DBG("Using kernel consumer at: %s",  consumer_to_use);
                        ret = execl(consumer_to_use,
@@ -2619,7 +2623,7 @@ static pid_t spawn_consumerd(struct consumer_data *consumer_data)
                        break;
                }
                default:
-                       PERROR("unknown consumer type");
+                       ERR("unknown consumer type");
                        exit(EXIT_FAILURE);
                }
                if (errno != 0) {
@@ -2712,7 +2716,6 @@ static int init_kernel_tracer(void)
        kernel_tracer_fd = open(module_proc_lttng, O_RDWR);
        if (kernel_tracer_fd < 0) {
                DBG("Failed to open %s", module_proc_lttng);
-               ret = -1;
                goto error_open;
        }
 
@@ -4373,12 +4376,41 @@ static void *thread_manage_clients(void *data)
        }
 
        sessiond_notify_ready();
+
        ret = sem_post(&load_info->message_thread_ready);
        if (ret) {
                PERROR("sem_post message_thread_ready");
                goto error;
        }
 
+       /*
+        * Wait until all support threads are initialized before accepting
+        * commands.
+        */
+       while (uatomic_read(&lttng_sessiond_ready) != 0) {
+               fd_set read_fds;
+               struct timeval timeout;
+
+               FD_ZERO(&read_fds);
+               FD_SET(thread_quit_pipe[0], &read_fds);
+               memset(&timeout, 0, sizeof(timeout));
+               timeout.tv_usec = 1000;
+
+               /*
+                * If a support thread failed to launch, it may signal that
+                * we must exit and the sessiond would never be marked as
+                * "ready".
+                *
+                * The timeout is set to 1ms, which serves as a way to
+                * pace down this check.
+                */
+               ret = select(thread_quit_pipe[0] + 1, &read_fds, NULL, NULL,
+                               &timeout);
+               if (ret > 0 || (ret < 0 && errno != EINTR)) {
+                       goto exit;
+               }
+       }
+
        /* This testpoint is after we signal readiness to the parent. */
        if (testpoint(sessiond_thread_manage_clients)) {
                goto error;
@@ -5522,6 +5554,44 @@ error:
        return ret;
 }
 
+static int set_clock_plugin_env(void)
+{
+       int ret = 0;
+       const char *original_env_value;
+       char *full_path = NULL;
+       char *new_env_value = NULL;
+
+       original_env_value = getenv("LTTNG_UST_CLOCK_PLUGIN");
+       if (!original_env_value) {
+               goto end;
+       }
+
+       full_path = utils_expand_path(original_env_value);
+       if (!full_path) {
+               ERR("Failed to expand LTTNG_UST_CLOCK_PLUGIN path \"%s\"",
+                               original_env_value);
+               ret = -1;
+               goto end;
+       }
+        ret = asprintf(&new_env_value, "LTTNG_UST_CLOCK_PLUGIN=%s",
+                       full_path);
+       free(full_path);
+       if (ret < 0) {
+               PERROR("asprintf");
+               goto end;
+       }
+
+       DBG("Updating environment: %s", new_env_value);
+       ret = putenv(new_env_value);
+       if (ret) {
+               free(new_env_value);
+               PERROR("putenv of LTTNG_UST_CLOCK_PLUGIN");
+               goto end;
+       }
+end:
+       return ret;
+}
+
 /*
  * main
  */
@@ -5563,6 +5633,12 @@ int main(int argc, char **argv)
                goto exit_options;
        }
 
+       ret = set_clock_plugin_env();
+       if (ret) {
+               retval = -1;
+               goto exit_options;
+       }
+
        /* Daemonize */
        if (opt_daemon || opt_background) {
                int i;