X-Git-Url: http://git.lttng.org/?a=blobdiff_plain;f=src%2Fbin%2Flttng-sessiond%2Fmain.c;h=476b1b35fcbd65fcf6daeeb829309ad7613900e6;hb=25c629438a161e3ae35fa50baf63339b031bd749;hp=2582e886d1e1a5f5352a6f91b8bd82b2ea914c7b;hpb=77cdfc81210ea4cfca740335af03be732e652e6d;p=lttng-tools.git diff --git a/src/bin/lttng-sessiond/main.c b/src/bin/lttng-sessiond/main.c index 2582e886d..476b1b35f 100644 --- a/src/bin/lttng-sessiond/main.c +++ b/src/bin/lttng-sessiond/main.c @@ -582,6 +582,34 @@ static int generate_lock_file_path(char *path, size_t len) return ret; } +/* + * Wait on consumer process termination. + * + * Need to be called with the consumer data lock held or from a context + * ensuring no concurrent access to data (e.g: cleanup). + */ +static void wait_consumer(struct consumer_data *consumer_data) +{ + pid_t ret; + int status; + + if (consumer_data->pid <= 0) { + return; + } + + DBG("Waiting for complete teardown of consumerd (PID: %d)", + consumer_data->pid); + ret = waitpid(consumer_data->pid, &status, 0); + if (ret == -1) { + PERROR("consumerd waitpid pid: %d", consumer_data->pid) + } + if (!WIFEXITED(status)) { + ERR("consumerd termination with error: %d", + WEXITSTATUS(ret)); + } + consumer_data->pid = 0; +} + /* * Cleanup the session daemon's data structures. */ @@ -676,6 +704,10 @@ static void sessiond_cleanup(void) } } + wait_consumer(&kconsumer_data); + wait_consumer(&ustconsumer64_data); + wait_consumer(&ustconsumer32_data); + DBG("Cleaning up all agent apps"); agent_app_ht_clean(); @@ -765,6 +797,8 @@ static void sessiond_cleanup_options(void) free(kmod_probes_list); free(kmod_extra_probes_list); + run_as_destroy_worker(); + /* */ DBG("%c[%d;%dm*** assert failed :-) *** ==> %c[%dm%c[%d;%dm" "Matthew, BEET driven development works!%c[%dm", @@ -1129,31 +1163,33 @@ static void *thread_manage_kernel(void *data) } /* Check for data on kernel pipe */ - if (pollfd == kernel_poll_pipe[0] && (revents & LPOLLIN)) { - (void) lttng_read(kernel_poll_pipe[0], - &tmp, 1); - /* - * Ret value is useless here, if this pipe gets any actions an - * update is required anyway. - */ - update_poll_flag = 1; - continue; - } else { - /* - * New CPU detected by the kernel. Adding kernel stream to - * kernel session and updating the kernel consumer - */ - if (revents & LPOLLIN) { + if (revents & LPOLLIN) { + if (pollfd == kernel_poll_pipe[0]) { + (void) lttng_read(kernel_poll_pipe[0], + &tmp, 1); + /* + * Ret value is useless here, if this pipe gets any actions an + * update is required anyway. + */ + update_poll_flag = 1; + continue; + } else { + /* + * New CPU detected by the kernel. Adding kernel stream to + * kernel session and updating the kernel consumer + */ ret = update_kernel_stream(&kconsumer_data, pollfd); if (ret < 0) { continue; } break; - /* - * TODO: We might want to handle the LPOLLERR | LPOLLHUP - * and unregister kernel stream at this point. - */ } + } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { + update_poll_flag = 1; + continue; + } else { + ERR("Unexpected poll events %u for sock %d", revents, pollfd); + goto error; } } } @@ -1212,6 +1248,9 @@ static void *thread_manage_consumer(void *data) DBG("[thread] Manage consumer started"); + rcu_register_thread(); + rcu_thread_online(); + health_register(health_sessiond, HEALTH_SESSIOND_TYPE_CONSUMER); health_code_update(); @@ -1280,9 +1319,14 @@ restart: /* Event on the registration socket */ if (pollfd == consumer_data->err_sock) { - if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { + if (revents & LPOLLIN) { + continue; + } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { ERR("consumer err socket poll error"); goto error; + } else { + ERR("Unexpected poll events %u for sock %d", revents, pollfd); + goto error; } } } @@ -1412,7 +1456,8 @@ restart_poll: if (pollfd == sock) { /* Event on the consumerd socket */ - if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { + if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP) + && !(revents & LPOLLIN)) { ERR("consumer err socket second poll error"); goto error; } @@ -1430,6 +1475,11 @@ restart_poll: goto exit; } else if (pollfd == consumer_data->metadata_fd) { + if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP) + && !(revents & LPOLLIN)) { + ERR("consumer err metadata socket second poll error"); + goto error; + } /* UST metadata requests */ ret = ust_consumer_metadata_request( &consumer_data->metadata_sock); @@ -1493,7 +1543,6 @@ error: unlink(consumer_data->err_unix_sock_path); unlink(consumer_data->cmd_unix_sock_path); - consumer_data->pid = 0; pthread_mutex_unlock(&consumer_data->lock); /* Cleanup metadata socket mutex. */ @@ -1510,6 +1559,9 @@ error_poll: health_unregister(health_sessiond); DBG("consumer thread cleanup completed"); + rcu_thread_offline(); + rcu_unregister_thread(); + return NULL; } @@ -1595,10 +1647,7 @@ static void *thread_manage_apps(void *data) /* Inspect the apps cmd pipe */ if (pollfd == apps_cmd_pipe[0]) { - if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { - ERR("Apps command pipe error"); - goto error; - } else if (revents & LPOLLIN) { + if (revents & LPOLLIN) { int sock; /* Empty pipe */ @@ -1611,9 +1660,8 @@ static void *thread_manage_apps(void *data) health_code_update(); /* - * We only monitor the error events of the socket. This - * thread does not handle any incoming data from UST - * (POLLIN). + * Since this is a command socket (write then read), + * we only monitor the error events of the socket. */ ret = lttng_poll_add(&events, sock, LPOLLERR | LPOLLHUP | LPOLLRDHUP); @@ -1622,6 +1670,12 @@ static void *thread_manage_apps(void *data) } DBG("Apps with sock %d added to poll set", sock); + } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { + ERR("Apps command pipe error"); + goto error; + } else { + ERR("Unknown poll events %u for sock %d", revents, pollfd); + goto error; } } else { /* @@ -1637,6 +1691,9 @@ static void *thread_manage_apps(void *data) /* Socket closed on remote end. */ ust_app_unregister(pollfd); + } else { + ERR("Unexpected poll events %u for sock %d", revents, pollfd); + goto error; } } @@ -1778,6 +1835,9 @@ static void sanitize_wait_queue(struct ust_reg_wait_queue *wait_queue) ust_app_destroy(wait_node->app); free(wait_node); break; + } else { + ERR("Unexpected poll events %u for sock %d", revents, pollfd); + goto error; } } } @@ -1979,7 +2039,7 @@ static void *thread_dispatch_ust_registration(void *data) * Don't care about return value. Let the manage apps threads * handle app unregistration upon socket close. */ - (void) ust_app_register_done(app->sock); + (void) ust_app_register_done(app); /* * Even if the application socket has been closed, send the app @@ -2021,6 +2081,22 @@ error: free(wait_node); } + /* Empty command queue. */ + for (;;) { + /* Dequeue command for registration */ + node = cds_wfcq_dequeue_blocking(&ust_cmd_queue.head, &ust_cmd_queue.tail); + if (node == NULL) { + break; + } + ust_cmd = caa_container_of(node, struct ust_command, node); + ret = close(ust_cmd->sock); + if (ret < 0) { + PERROR("close ust sock exit dispatch %d", ust_cmd->sock); + } + lttng_fd_put(LTTNG_FD_APPS, 1); + free(ust_cmd); + } + error_testpoint: DBG("Dispatch thread dying"); if (err) { @@ -2122,10 +2198,7 @@ static void *thread_registration_apps(void *data) /* Event on the registration socket */ if (pollfd == apps_sock) { - if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { - ERR("Register apps socket poll error"); - goto error; - } else if (revents & LPOLLIN) { + if (revents & LPOLLIN) { sock = lttcomm_accept_unix_sock(apps_sock); if (sock < 0) { goto error; @@ -2138,10 +2211,12 @@ static void *thread_registration_apps(void *data) * lttcomm_setsockopt_snd_timeout expect msec as * parameter. */ - (void) lttcomm_setsockopt_rcv_timeout(sock, - app_socket_timeout * 1000); - (void) lttcomm_setsockopt_snd_timeout(sock, - app_socket_timeout * 1000); + if (app_socket_timeout >= 0) { + (void) lttcomm_setsockopt_rcv_timeout(sock, + app_socket_timeout * 1000); + (void) lttcomm_setsockopt_snd_timeout(sock, + app_socket_timeout * 1000); + } /* * Set the CLOEXEC flag. Return code is useless because @@ -2153,6 +2228,10 @@ static void *thread_registration_apps(void *data) ust_cmd = zmalloc(sizeof(struct ust_command)); if (ust_cmd == NULL) { PERROR("ust command zmalloc"); + ret = close(sock); + if (ret) { + PERROR("close"); + } goto error; } @@ -2208,6 +2287,12 @@ static void *thread_registration_apps(void *data) * barrier with the exchange in cds_wfcq_enqueue. */ futex_nto1_wake(&ust_cmd_queue.futex); + } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { + ERR("Register apps socket poll error"); + goto error; + } else { + ERR("Unexpected poll events %u for sock %d", revents, pollfd); + goto error; } } } @@ -2709,7 +2794,7 @@ static int copy_session_consumer(int domain, struct ltt_session *session) * domain. */ if (session->kernel_session->consumer) { - consumer_destroy_output(session->kernel_session->consumer); + consumer_output_put(session->kernel_session->consumer); } session->kernel_session->consumer = consumer_copy_output(session->consumer); @@ -2723,7 +2808,7 @@ static int copy_session_consumer(int domain, struct ltt_session *session) case LTTNG_DOMAIN_UST: DBG3("Copying tracing session consumer output in UST session"); if (session->ust_session->consumer) { - consumer_destroy_output(session->ust_session->consumer); + consumer_output_put(session->ust_session->consumer); } session->ust_session->consumer = consumer_copy_output(session->consumer); @@ -2843,7 +2928,7 @@ static int create_kernel_session(struct ltt_session *session) session->kernel_session->consumer->dst.trace_path, S_IRWXU | S_IRWXG, session->uid, session->gid); if (ret < 0) { - if (ret != -EEXIST) { + if (errno != EEXIST) { ERR("Trace directory creation error"); goto error; } @@ -2905,6 +2990,8 @@ static int process_client_msg(struct command_ctx *cmd_ctx, int sock, DBG("Processing client command %d", cmd_ctx->lsm->cmd_type); + assert(!rcu_read_ongoing()); + *sock_error = 0; switch (cmd_ctx->lsm->cmd_type) { @@ -2964,6 +3051,7 @@ static int process_client_msg(struct command_ctx *cmd_ctx, int sock, case LTTNG_LIST_EVENTS: case LTTNG_LIST_SYSCALLS: case LTTNG_LIST_TRACKER_PIDS: + case LTTNG_DATA_PENDING: break; default: /* Setup lttng message with no payload */ @@ -3200,6 +3288,8 @@ skip_domain: if (cmd_ctx->lsm->cmd_type == LTTNG_START_TRACE || cmd_ctx->lsm->cmd_type == LTTNG_STOP_TRACE) { switch (cmd_ctx->lsm->domain.type) { + case LTTNG_DOMAIN_NONE: + break; case LTTNG_DOMAIN_JUL: case LTTNG_DOMAIN_LOG4J: case LTTNG_DOMAIN_PYTHON: @@ -3215,6 +3305,9 @@ skip_domain: goto error; } break; + default: + ret = LTTNG_ERR_UNKNOWN_DOMAIN; + goto error; } } @@ -3263,8 +3356,34 @@ skip_domain: } case LTTNG_DISABLE_EVENT: { + + /* + * FIXME: handle filter; for now we just receive the filter's + * bytecode along with the filter expression which are sent by + * liblttng-ctl and discard them. + * + * This fixes an issue where the client may block while sending + * the filter payload and encounter an error because the session + * daemon closes the socket without ever handling this data. + */ + size_t count = cmd_ctx->lsm->u.disable.expression_len + + cmd_ctx->lsm->u.disable.bytecode_len; + + if (count) { + char data[LTTNG_FILTER_MAX_LEN]; + + DBG("Discarding disable event command payload of size %zu", count); + while (count) { + ret = lttcomm_recv_unix_sock(sock, data, + count > sizeof(data) ? sizeof(data) : count); + if (ret < 0) { + goto error; + } + + count -= (size_t) ret; + } + } /* FIXME: passing packed structure to non-packed pointer */ - /* TODO: handle filter */ ret = cmd_disable_event(cmd_ctx->session, cmd_ctx->lsm->domain.type, cmd_ctx->lsm->u.disable.channel_name, &cmd_ctx->lsm->u.disable.event); @@ -3938,6 +4057,7 @@ setup_error: session_unlock_list(); } init_setup_error: + assert(!rcu_read_ongoing()); return ret; } @@ -4053,9 +4173,14 @@ restart: /* Event on the registration socket */ if (pollfd == sock) { - if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { + if (revents & LPOLLIN) { + continue; + } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { ERR("Health socket poll error"); goto error; + } else { + ERR("Unexpected poll events %u for sock %d", revents, pollfd); + goto error; } } } @@ -4126,7 +4251,7 @@ error: } lttng_poll_clean(&events); - + stop_threads(); rcu_unregister_thread(); return NULL; } @@ -4230,9 +4355,14 @@ static void *thread_manage_clients(void *data) /* Event on the registration socket */ if (pollfd == client_sock) { - if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { + if (revents & LPOLLIN) { + continue; + } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { ERR("Client socket poll error"); goto error; + } else { + ERR("Unexpected poll events %u for sock %d", revents, pollfd); + goto error; } } } @@ -4458,18 +4588,11 @@ static int set_option(int opt, const char *arg, const char *optname) { int ret = 0; - if (arg && arg[0] == '\0') { - /* - * This only happens if the value is read from daemon config - * file. This means the option requires an argument and the - * configuration file contains a line such as: - * my_option = - */ - ret = -EINVAL; - goto end; - } - if (string_match(optname, "client-sock") || opt == 'c') { + if (!arg || *arg == '\0') { + ret = -EINVAL; + goto end; + } if (lttng_is_setuid_setgid()) { WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", "-c, --client-sock"); @@ -4477,6 +4600,10 @@ static int set_option(int opt, const char *arg, const char *optname) snprintf(client_unix_sock_path, PATH_MAX, "%s", arg); } } else if (string_match(optname, "apps-sock") || opt == 'a') { + if (!arg || *arg == '\0') { + ret = -EINVAL; + goto end; + } if (lttng_is_setuid_setgid()) { WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", "-a, --apps-sock"); @@ -4488,6 +4615,10 @@ static int set_option(int opt, const char *arg, const char *optname) } else if (string_match(optname, "background") || opt == 'b') { opt_background = 1; } else if (string_match(optname, "group") || opt == 'g') { + if (!arg || *arg == '\0') { + ret = -EINVAL; + goto end; + } if (lttng_is_setuid_setgid()) { WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", "-g, --group"); @@ -4516,6 +4647,10 @@ static int set_option(int opt, const char *arg, const char *optname) } else if (string_match(optname, "sig-parent") || opt == 'S') { opt_sig_parent = 1; } else if (string_match(optname, "kconsumerd-err-sock")) { + if (!arg || *arg == '\0') { + ret = -EINVAL; + goto end; + } if (lttng_is_setuid_setgid()) { WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", "--kconsumerd-err-sock"); @@ -4523,6 +4658,10 @@ static int set_option(int opt, const char *arg, const char *optname) snprintf(kconsumer_data.err_unix_sock_path, PATH_MAX, "%s", arg); } } else if (string_match(optname, "kconsumerd-cmd-sock")) { + if (!arg || *arg == '\0') { + ret = -EINVAL; + goto end; + } if (lttng_is_setuid_setgid()) { WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", "--kconsumerd-cmd-sock"); @@ -4530,6 +4669,10 @@ static int set_option(int opt, const char *arg, const char *optname) snprintf(kconsumer_data.cmd_unix_sock_path, PATH_MAX, "%s", arg); } } else if (string_match(optname, "ustconsumerd64-err-sock")) { + if (!arg || *arg == '\0') { + ret = -EINVAL; + goto end; + } if (lttng_is_setuid_setgid()) { WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", "--ustconsumerd64-err-sock"); @@ -4537,6 +4680,10 @@ static int set_option(int opt, const char *arg, const char *optname) snprintf(ustconsumer64_data.err_unix_sock_path, PATH_MAX, "%s", arg); } } else if (string_match(optname, "ustconsumerd64-cmd-sock")) { + if (!arg || *arg == '\0') { + ret = -EINVAL; + goto end; + } if (lttng_is_setuid_setgid()) { WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", "--ustconsumerd64-cmd-sock"); @@ -4544,6 +4691,10 @@ static int set_option(int opt, const char *arg, const char *optname) snprintf(ustconsumer64_data.cmd_unix_sock_path, PATH_MAX, "%s", arg); } } else if (string_match(optname, "ustconsumerd32-err-sock")) { + if (!arg || *arg == '\0') { + ret = -EINVAL; + goto end; + } if (lttng_is_setuid_setgid()) { WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", "--ustconsumerd32-err-sock"); @@ -4551,6 +4702,10 @@ static int set_option(int opt, const char *arg, const char *optname) snprintf(ustconsumer32_data.err_unix_sock_path, PATH_MAX, "%s", arg); } } else if (string_match(optname, "ustconsumerd32-cmd-sock")) { + if (!arg || *arg == '\0') { + ret = -EINVAL; + goto end; + } if (lttng_is_setuid_setgid()) { WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", "--ustconsumerd32-cmd-sock"); @@ -4580,6 +4735,10 @@ static int set_option(int opt, const char *arg, const char *optname) opt_verbose_consumer += 1; } } else if (string_match(optname, "consumerd32-path")) { + if (!arg || *arg == '\0') { + ret = -EINVAL; + goto end; + } if (lttng_is_setuid_setgid()) { WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", "--consumerd32-path"); @@ -4595,6 +4754,10 @@ static int set_option(int opt, const char *arg, const char *optname) consumerd32_bin_override = 1; } } else if (string_match(optname, "consumerd32-libdir")) { + if (!arg || *arg == '\0') { + ret = -EINVAL; + goto end; + } if (lttng_is_setuid_setgid()) { WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", "--consumerd32-libdir"); @@ -4610,6 +4773,10 @@ static int set_option(int opt, const char *arg, const char *optname) consumerd32_libdir_override = 1; } } else if (string_match(optname, "consumerd64-path")) { + if (!arg || *arg == '\0') { + ret = -EINVAL; + goto end; + } if (lttng_is_setuid_setgid()) { WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", "--consumerd64-path"); @@ -4625,6 +4792,10 @@ static int set_option(int opt, const char *arg, const char *optname) consumerd64_bin_override = 1; } } else if (string_match(optname, "consumerd64-libdir")) { + if (!arg || *arg == '\0') { + ret = -EINVAL; + goto end; + } if (lttng_is_setuid_setgid()) { WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", "--consumerd64-libdir"); @@ -4640,6 +4811,10 @@ static int set_option(int opt, const char *arg, const char *optname) consumerd64_libdir_override = 1; } } else if (string_match(optname, "pidfile") || opt == 'p') { + if (!arg || *arg == '\0') { + ret = -EINVAL; + goto end; + } if (lttng_is_setuid_setgid()) { WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", "-p, --pidfile"); @@ -4652,6 +4827,10 @@ static int set_option(int opt, const char *arg, const char *optname) } } } else if (string_match(optname, "agent-tcp-port")) { + if (!arg || *arg == '\0') { + ret = -EINVAL; + goto end; + } if (lttng_is_setuid_setgid()) { WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", "--agent-tcp-port"); @@ -4676,6 +4855,10 @@ static int set_option(int opt, const char *arg, const char *optname) DBG3("Agent TCP port set to non default: %u", agent_tcp_port); } } else if (string_match(optname, "load") || opt == 'l') { + if (!arg || *arg == '\0') { + ret = -EINVAL; + goto end; + } if (lttng_is_setuid_setgid()) { WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", "-l, --load"); @@ -4688,6 +4871,10 @@ static int set_option(int opt, const char *arg, const char *optname) } } } else if (string_match(optname, "kmod-probes")) { + if (!arg || *arg == '\0') { + ret = -EINVAL; + goto end; + } if (lttng_is_setuid_setgid()) { WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", "--kmod-probes"); @@ -4700,6 +4887,10 @@ static int set_option(int opt, const char *arg, const char *optname) } } } else if (string_match(optname, "extra-kmod-probes")) { + if (!arg || *arg == '\0') { + ret = -EINVAL; + goto end; + } if (lttng_is_setuid_setgid()) { WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", "--extra-kmod-probes"); @@ -5134,9 +5325,6 @@ error: static void sighandler(int sig) { switch (sig) { - case SIGPIPE: - DBG("SIGPIPE caught"); - return; case SIGINT: DBG("SIGINT caught"); stop_threads(); @@ -5168,9 +5356,10 @@ static int set_signal_handler(void) return ret; } - sa.sa_handler = sighandler; sa.sa_mask = sigset; sa.sa_flags = 0; + + sa.sa_handler = sighandler; if ((ret = sigaction(SIGTERM, &sa, NULL)) < 0) { PERROR("sigaction"); return ret; @@ -5181,12 +5370,13 @@ static int set_signal_handler(void) return ret; } - if ((ret = sigaction(SIGPIPE, &sa, NULL)) < 0) { + if ((ret = sigaction(SIGUSR1, &sa, NULL)) < 0) { PERROR("sigaction"); return ret; } - if ((ret = sigaction(SIGUSR1, &sa, NULL)) < 0) { + sa.sa_handler = SIG_IGN; + if ((ret = sigaction(SIGPIPE, &sa, NULL)) < 0) { PERROR("sigaction"); return ret; } @@ -5226,7 +5416,10 @@ static int write_pidfile(void) assert(rundir); if (opt_pidfile) { - strncpy(pidfile_path, opt_pidfile, sizeof(pidfile_path)); + if (lttng_strncpy(pidfile_path, opt_pidfile, sizeof(pidfile_path))) { + ret = -1; + goto error; + } } else { /* Build pidfile path from rundir and opt_pidfile. */ ret = snprintf(pidfile_path, sizeof(pidfile_path), "%s/" @@ -5354,6 +5547,10 @@ int main(int argc, char **argv) } } + if (run_as_create_worker(argv[0]) < 0) { + goto exit_create_run_as_worker_cleanup; + } + /* * Starting from here, we can create threads. This needs to be after * lttng_daemonize due to RCU. @@ -5907,6 +6104,10 @@ exit_apps: } exit_reg_apps: + /* + * Join dispatch thread after joining reg_apps_thread to ensure + * we don't leak applications in the queue. + */ ret = pthread_join(dispatch_thread, &status); if (ret) { errno = ret; @@ -5941,6 +6142,13 @@ exit_init_data: rcu_thread_offline(); rcu_unregister_thread(); + /* + * Ensure all prior call_rcu are done. call_rcu callbacks may push + * hash tables to the ht_cleanup thread. Therefore, we ensure that + * the queue is empty before shutting down the clean-up thread. + */ + rcu_barrier(); + ret = notify_thread_pipe(ht_cleanup_quit_pipe[1]); if (ret < 0) { ERR("write error on ht_cleanup quit pipe"); @@ -5967,11 +6175,13 @@ exit_ht_cleanup_quit_pipe: health_app_destroy(health_sessiond); exit_health_sessiond_cleanup: +exit_create_run_as_worker_cleanup: exit_options: sessiond_cleanup_options(); exit_set_signal_handler: + if (!retval) { exit(EXIT_SUCCESS); } else {