X-Git-Url: http://git.lttng.org./?a=blobdiff_plain;f=src%2Fbin%2Flttng-sessiond%2Fmain.c;h=9b3828bd08feec5654c6a73cc1fdc3e01abfa672;hb=ec4c7666d3907424f62e874b78e44ce4b9456f41;hp=de4736643010143c6af8c7a554f1682ddaebc651;hpb=8c6c56c2c6577c540786b341e5334f676c12d43a;p=lttng-tools.git diff --git a/src/bin/lttng-sessiond/main.c b/src/bin/lttng-sessiond/main.c index de4736643..9b3828bd0 100644 --- a/src/bin/lttng-sessiond/main.c +++ b/src/bin/lttng-sessiond/main.c @@ -20,6 +20,7 @@ #include #include #include +#include #include #include #include @@ -44,6 +45,7 @@ #include #include #include +#include #include "lttng-sessiond.h" #include "buffer-registry.h" @@ -63,6 +65,7 @@ #include "health-sessiond.h" #include "testpoint.h" #include "ust-thread.h" +#include "jul-thread.h" #define CONSUMERD_FILE "lttng-consumerd" @@ -71,11 +74,15 @@ static const char *tracing_group_name = DEFAULT_TRACING_GROUP; static const char *opt_pidfile; static int opt_sig_parent; static int opt_verbose_consumer; -static int opt_daemon; +static int opt_daemon, opt_background; static int opt_no_kernel; -static int is_root; /* Set to 1 if the daemon is running as root */ static pid_t ppid; /* Parent PID for --sig-parent option */ +static pid_t child_ppid; /* Internal parent PID use with daemonize. */ static char *rundir; +static int lockfile_fd = -1; + +/* Set to 1 when a SIGUSR1 signal is received. */ +static int recv_child_signal; /* * Consumer daemon specific control data. Every value not initialized here is @@ -156,6 +163,7 @@ static pthread_t kernel_thread; static pthread_t dispatch_thread; static pthread_t health_thread; static pthread_t ht_cleanup_thread; +static pthread_t jul_reg_thread; /* * UST registration command queue. This queue is tied with a futex and uses a N @@ -233,6 +241,44 @@ long page_size; /* Application health monitoring */ struct health_app *health_sessiond; +/* JUL TCP port for registration. Used by the JUL thread. */ +unsigned int jul_tcp_port = DEFAULT_JUL_TCP_PORT; + +/* Am I root or not. */ +int is_root; /* Set to 1 if the daemon is running as root */ + +/* + * Whether sessiond is ready for commands/health check requests. + * NR_LTTNG_SESSIOND_READY must match the number of calls to + * lttng_sessiond_notify_ready(). + */ +#define NR_LTTNG_SESSIOND_READY 2 +int lttng_sessiond_ready = NR_LTTNG_SESSIOND_READY; + +/* Notify parents that we are ready for cmd and health check */ +static +void lttng_sessiond_notify_ready(void) +{ + if (uatomic_sub_return(<tng_sessiond_ready, 1) == 0) { + /* + * Notify parent pid that we are ready to accept command + * for client side. This ppid is the one from the + * external process that spawned us. + */ + if (opt_sig_parent) { + kill(ppid, SIGUSR1); + } + + /* + * Notify the parent of the fork() process that we are + * ready. + */ + if (opt_daemon || opt_background) { + kill(child_ppid, SIGUSR1); + } + } +} + static void setup_consumerd_path(void) { @@ -413,6 +459,27 @@ static void close_consumer_sockets(void) } } +/* + * Generate the full lock file path using the rundir. + * + * Return the snprintf() return value thus a negative value is an error. + */ +static int generate_lock_file_path(char *path, size_t len) +{ + int ret; + + assert(path); + assert(rundir); + + /* Build lockfile path from rundir. */ + ret = snprintf(path, len, "%s/" DEFAULT_LTTNG_SESSIOND_LOCKFILE, rundir); + if (ret < 0) { + PERROR("snprintf lockfile path"); + } + + return ret; +} + /* * Cleanup the daemon */ @@ -450,6 +517,11 @@ static void cleanup(void) DBG("Removing %s", path); (void) unlink(path); + snprintf(path, PATH_MAX, "%s/%s", rundir, + DEFAULT_LTTNG_SESSIOND_JULPORT_FILE); + DBG("Removing %s", path); + (void) unlink(path); + /* kconsumerd */ snprintf(path, PATH_MAX, DEFAULT_KCONSUMERD_ERR_SOCK_PATH, @@ -489,8 +561,6 @@ static void cleanup(void) DBG("Removing directory %s", path); (void) rmdir(path); - free(rundir); - DBG("Cleaning up all sessions"); /* Destroy session list mutex */ @@ -522,6 +592,35 @@ static void cleanup(void) close_consumer_sockets(); + + /* + * Cleanup lock file by deleting it and finaly closing it which will + * release the file system lock. + */ + if (lockfile_fd >= 0) { + char lockfile_path[PATH_MAX]; + + ret = generate_lock_file_path(lockfile_path, sizeof(lockfile_path)); + if (ret > 0) { + ret = remove(lockfile_path); + if (ret < 0) { + PERROR("remove lock file"); + } + ret = close(lockfile_fd); + if (ret < 0) { + PERROR("close lock file"); + } + } + } + + /* + * We do NOT rmdir rundir because there are other processes + * using it, for instance lttng-relayd, which can start in + * parallel with this teardown. + */ + + free(rundir); + /* */ DBG("%c[%d;%dm*** assert failed :-) *** ==> %c[%dm%c[%d;%dm" "Matthew, BEET driven development works!%c[%dm", @@ -780,13 +879,13 @@ static void *thread_manage_kernel(void *data) */ lttng_poll_init(&events); - if (testpoint(thread_manage_kernel)) { + if (testpoint(sessiond_thread_manage_kernel)) { goto error_testpoint; } health_code_update(); - if (testpoint(thread_manage_kernel_before_loop)) { + if (testpoint(sessiond_thread_manage_kernel_before_loop)) { goto error_testpoint; } @@ -815,12 +914,14 @@ static void *thread_manage_kernel(void *data) update_poll_flag = 0; } - DBG("Thread kernel polling on %d fds", LTTNG_POLL_GETNB(&events)); + DBG("Thread kernel polling"); /* Poll infinite value of time */ restart: health_poll_entry(); ret = lttng_poll_wait(&events, -1); + DBG("Thread kernel return from poll on %d fds", + LTTNG_POLL_GETNB(&events)); health_poll_exit(); if (ret < 0) { /* @@ -846,6 +947,11 @@ static void *thread_manage_kernel(void *data) health_code_update(); + if (!revents) { + /* No activity for this FD (poll implementation). */ + continue; + } + /* Thread quit pipe has been closed. Killing thread. */ ret = sessiond_check_thread_quit_pipe(pollfd, revents); if (ret) { @@ -855,9 +961,8 @@ static void *thread_manage_kernel(void *data) /* Check for data on kernel pipe */ if (pollfd == kernel_poll_pipe[0] && (revents & LPOLLIN)) { - do { - ret = read(kernel_poll_pipe[0], &tmp, 1); - } while (ret < 0 && errno == EINTR); + (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. @@ -930,7 +1035,7 @@ static void signal_consumer_condition(struct consumer_data *data, int state) */ static void *thread_manage_consumer(void *data) { - int sock = -1, i, ret, pollfd, err = -1; + int sock = -1, i, ret, pollfd, err = -1, should_quit = 0; uint32_t revents, nb_fd; enum lttcomm_return_code code; struct lttng_poll_event events; @@ -967,7 +1072,7 @@ static void *thread_manage_consumer(void *data) restart: health_poll_entry(); - if (testpoint(thread_manage_consumer)) { + if (testpoint(sessiond_thread_manage_consumer)) { goto error; } @@ -992,6 +1097,11 @@ restart: health_code_update(); + if (!revents) { + /* No activity for this FD (poll implementation). */ + continue; + } + /* Thread quit pipe has been closed. Killing thread. */ ret = sessiond_check_thread_quit_pipe(pollfd, revents); if (ret) { @@ -1031,7 +1141,6 @@ restart: } health_code_update(); - if (code == LTTCOMM_CONSUMERD_COMMAND_SOCK_READY) { /* Connect both socket, command and metadata. */ consumer_data->cmd_sock = @@ -1089,6 +1198,15 @@ restart: /* Infinite blocking call, waiting for transmission */ restart_poll: while (1) { + health_code_update(); + + /* Exit the thread because the thread quit pipe has been triggered. */ + if (should_quit) { + /* Not a health error. */ + err = 0; + goto exit; + } + health_poll_entry(); ret = lttng_poll_wait(&events, -1); health_poll_exit(); @@ -1111,13 +1229,18 @@ restart_poll: health_code_update(); - /* Thread quit pipe has been closed. Killing thread. */ - ret = sessiond_check_thread_quit_pipe(pollfd, revents); - if (ret) { - err = 0; - goto exit; + if (!revents) { + /* No activity for this FD (poll implementation). */ + continue; } + /* + * Thread quit pipe has been triggered, flag that we should stop + * but continue the current loop to handle potential data from + * consumer. + */ + should_quit = sessiond_check_thread_quit_pipe(pollfd, revents); + if (pollfd == sock) { /* Event on the consumerd socket */ if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { @@ -1145,11 +1268,8 @@ restart_poll: ERR("Handling metadata request"); goto error; } - break; - } else { - ERR("Unknown pollfd"); - goto error; } + /* No need for an else branch all FDs are tested prior. */ } health_code_update(); } @@ -1188,13 +1308,13 @@ error: } consumer_data->cmd_sock = -1; } - if (*consumer_data->metadata_sock.fd_ptr >= 0) { + if (consumer_data->metadata_sock.fd_ptr && + *consumer_data->metadata_sock.fd_ptr >= 0) { ret = close(*consumer_data->metadata_sock.fd_ptr); if (ret) { PERROR("close"); } } - if (sock >= 0) { ret = close(sock); if (ret) { @@ -1208,9 +1328,10 @@ error: pthread_mutex_unlock(&consumer_data->lock); /* Cleanup metadata socket mutex. */ - pthread_mutex_destroy(consumer_data->metadata_sock.lock); - free(consumer_data->metadata_sock.lock); - + if (consumer_data->metadata_sock.lock) { + pthread_mutex_destroy(consumer_data->metadata_sock.lock); + free(consumer_data->metadata_sock.lock); + } lttng_poll_clean(&events); error_poll: if (err) { @@ -1229,6 +1350,7 @@ error_poll: static void *thread_manage_apps(void *data) { int i, ret, pollfd, err = -1; + ssize_t size_ret; uint32_t revents, nb_fd; struct lttng_poll_event events; @@ -1239,7 +1361,7 @@ static void *thread_manage_apps(void *data) health_register(health_sessiond, HEALTH_SESSIOND_TYPE_APP_MANAGE); - if (testpoint(thread_manage_apps)) { + if (testpoint(sessiond_thread_manage_apps)) { goto error_testpoint; } @@ -1255,19 +1377,21 @@ static void *thread_manage_apps(void *data) goto error; } - if (testpoint(thread_manage_apps_before_loop)) { + if (testpoint(sessiond_thread_manage_apps_before_loop)) { goto error; } health_code_update(); while (1) { - DBG("Apps thread polling on %d fds", LTTNG_POLL_GETNB(&events)); + DBG("Apps thread polling"); /* Inifinite blocking call, waiting for transmission */ restart: health_poll_entry(); ret = lttng_poll_wait(&events, -1); + DBG("Apps thread return from poll on %d fds", + LTTNG_POLL_GETNB(&events)); health_poll_exit(); if (ret < 0) { /* @@ -1288,6 +1412,11 @@ static void *thread_manage_apps(void *data) health_code_update(); + if (!revents) { + /* No activity for this FD (poll implementation). */ + continue; + } + /* Thread quit pipe has been closed. Killing thread. */ ret = sessiond_check_thread_quit_pipe(pollfd, revents); if (ret) { @@ -1304,10 +1433,8 @@ static void *thread_manage_apps(void *data) int sock; /* Empty pipe */ - do { - ret = read(apps_cmd_pipe[0], &sock, sizeof(sock)); - } while (ret < 0 && errno == EINTR); - if (ret < 0 || ret < sizeof(sock)) { + size_ret = lttng_read(apps_cmd_pipe[0], &sock, sizeof(sock)); + if (size_ret < sizeof(sock)) { PERROR("read apps cmd pipe"); goto error; } @@ -1325,23 +1452,7 @@ static void *thread_manage_apps(void *data) goto error; } - /* - * Set socket timeout for both receiving and ending. - * app_socket_timeout is in seconds, whereas - * lttcomm_setsockopt_rcv_timeout and - * 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); - DBG("Apps with sock %d added to poll set", sock); - - health_code_update(); - - break; } } else { /* @@ -1357,7 +1468,6 @@ static void *thread_manage_apps(void *data) /* Socket closed on remote end. */ ust_app_unregister(pollfd); - break; } } @@ -1402,7 +1512,7 @@ error_testpoint: */ static int send_socket_to_thread(int fd, int sock) { - int ret; + ssize_t ret; /* * It's possible that the FD is set as invalid with -1 concurrently just @@ -1413,10 +1523,8 @@ static int send_socket_to_thread(int fd, int sock) goto error; } - do { - ret = write(fd, &sock, sizeof(sock)); - } while (ret < 0 && errno == EINTR); - if (ret < 0 || ret != sizeof(sock)) { + ret = lttng_write(fd, &sock, sizeof(sock)); + if (ret < sizeof(sock)) { PERROR("write apps pipe %d", fd); if (ret < 0) { ret = -errno; @@ -1427,7 +1535,7 @@ static int send_socket_to_thread(int fd, int sock) /* All good. Don't send back the write positive ret value. */ ret = 0; error: - return ret; + return (int) ret; } /* @@ -1487,6 +1595,11 @@ static void sanitize_wait_queue(struct ust_reg_wait_queue *wait_queue) uint32_t revents = LTTNG_POLL_GETEV(&events, i); int pollfd = LTTNG_POLL_GETFD(&events, i); + if (!revents) { + /* No activity for this FD (poll implementation). */ + continue; + } + cds_list_for_each_entry_safe(wait_node, tmp_wait_node, &wait_queue->head, head) { if (pollfd == wait_node->app->sock && @@ -1531,6 +1644,10 @@ static void *thread_dispatch_ust_registration(void *data) health_register(health_sessiond, HEALTH_SESSIOND_TYPE_APP_REG_DISPATCH); + if (testpoint(sessiond_thread_app_reg_dispatch)) { + goto error_testpoint; + } + health_code_update(); CDS_INIT_LIST_HEAD(&wait_queue.head); @@ -1580,7 +1697,7 @@ static void *thread_dispatch_ust_registration(void *data) if (ret < 0) { PERROR("close ust sock dispatch %d", ust_cmd->sock); } - lttng_fd_put(1, LTTNG_FD_APPS); + lttng_fd_put(LTTNG_FD_APPS, 1); free(ust_cmd); goto error; } @@ -1594,7 +1711,7 @@ static void *thread_dispatch_ust_registration(void *data) if (ret < 0) { PERROR("close ust sock dispatch %d", ust_cmd->sock); } - lttng_fd_put(1, LTTNG_FD_APPS); + lttng_fd_put(LTTNG_FD_APPS, 1); free(wait_node); free(ust_cmd); continue; @@ -1642,7 +1759,7 @@ static void *thread_dispatch_ust_registration(void *data) if (ret < 0) { PERROR("close ust sock dispatch %d", ust_cmd->sock); } - lttng_fd_put(1, LTTNG_FD_APPS); + lttng_fd_put(LTTNG_FD_APPS, 1); } free(ust_cmd); } @@ -1735,6 +1852,7 @@ error: free(wait_node); } +error_testpoint: DBG("Dispatch thread dying"); if (err) { health_error(); @@ -1762,7 +1880,7 @@ static void *thread_registration_apps(void *data) health_register(health_sessiond, HEALTH_SESSIOND_TYPE_APP_REG); - if (testpoint(thread_registration_apps)) { + if (testpoint(sessiond_thread_registration_apps)) { goto error_testpoint; } @@ -1821,6 +1939,11 @@ static void *thread_registration_apps(void *data) revents = LTTNG_POLL_GETEV(&events, i); pollfd = LTTNG_POLL_GETFD(&events, i); + if (!revents) { + /* No activity for this FD (poll implementation). */ + continue; + } + /* Thread quit pipe has been closed. Killing thread. */ ret = sessiond_check_thread_quit_pipe(pollfd, revents); if (ret) { @@ -1839,6 +1962,18 @@ static void *thread_registration_apps(void *data) goto error; } + /* + * Set socket timeout for both receiving and ending. + * app_socket_timeout is in seconds, whereas + * lttcomm_setsockopt_rcv_timeout and + * 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); + /* * Set the CLOEXEC flag. Return code is useless because * either way, the show must go on. @@ -1911,11 +2046,6 @@ static void *thread_registration_apps(void *data) exit: error: - if (err) { - health_error(); - ERR("Health error occurred in %s", __func__); - } - /* Notify that the registration thread is gone */ notify_ust_apps(0); @@ -1940,6 +2070,10 @@ error_listen: error_create_poll: error_testpoint: DBG("UST Registration thread cleanup complete"); + if (err) { + health_error(); + ERR("Health error occurred in %s", __func__); + } health_unregister(health_sessiond); return NULL; @@ -2036,19 +2170,23 @@ static int spawn_consumer_thread(struct consumer_data *consumer_data) if (ret != 0) { errno = ret; if (ret == ETIMEDOUT) { + int pth_ret; + /* * Call has timed out so we kill the kconsumerd_thread and return * an error. */ ERR("Condition timed out. The consumer thread was never ready." " Killing it"); - ret = pthread_cancel(consumer_data->thread); - if (ret < 0) { + pth_ret = pthread_cancel(consumer_data->thread); + if (pth_ret < 0) { PERROR("pthread_cancel consumer thread"); } } else { PERROR("pthread_cond_wait failed consumer thread"); } + /* Caller is expecting a negative value on failure. */ + ret = -1; goto error; } @@ -2134,10 +2272,11 @@ static pid_t spawn_consumerd(struct consumer_data *consumer_data) consumer_to_use = consumerd32_bin; } else { DBG("Could not find any valid consumerd executable"); + ret = -EINVAL; break; } DBG("Using kernel consumer at: %s", consumer_to_use); - execl(consumer_to_use, + ret = execl(consumer_to_use, "lttng-consumerd", verbosity, "-k", "--consumerd-cmd-sock", consumer_data->cmd_unix_sock_path, "--consumerd-err-sock", consumer_data->err_unix_sock_path, @@ -2185,9 +2324,6 @@ static pid_t spawn_consumerd(struct consumer_data *consumer_data) if (consumerd64_libdir[0] != '\0') { free(tmpnew); } - if (ret) { - goto error; - } break; } case LTTNG_CONSUMER32_UST: @@ -2231,9 +2367,6 @@ static pid_t spawn_consumerd(struct consumer_data *consumer_data) if (consumerd32_libdir[0] != '\0') { free(tmpnew); } - if (ret) { - goto error; - } break; } default: @@ -2241,8 +2374,9 @@ static pid_t spawn_consumerd(struct consumer_data *consumer_data) exit(EXIT_FAILURE); } if (errno != 0) { - PERROR("kernel start consumer exec"); + PERROR("Consumer execl()"); } + /* Reaching this point, we got a failure on our execl(). */ exit(EXIT_FAILURE); } else if (pid > 0) { ret = pid; @@ -2410,6 +2544,7 @@ static int copy_session_consumer(int domain, struct ltt_session *session) consumer = session->kernel_session->consumer; dir_name = DEFAULT_KERNEL_TRACE_DIR; break; + case LTTNG_DOMAIN_JUL: case LTTNG_DOMAIN_UST: DBG3("Copying tracing session consumer output in UST session"); if (session->ust_session->consumer) { @@ -2453,6 +2588,7 @@ static int create_ust_session(struct ltt_session *session, assert(session->consumer); switch (domain->type) { + case LTTNG_DOMAIN_JUL: case LTTNG_DOMAIN_UST: break; default: @@ -2744,6 +2880,7 @@ static int process_client_msg(struct command_ctx *cmd_ctx, int sock, } break; + case LTTNG_DOMAIN_JUL: case LTTNG_DOMAIN_UST: { if (!ust_app_supported()) { @@ -2797,6 +2934,7 @@ static int process_client_msg(struct command_ctx *cmd_ctx, int sock, } /* 32-bit */ + pthread_mutex_lock(&ustconsumer32_data.pid_mutex); if (consumerd32_bin[0] != '\0' && ustconsumer32_data.pid == 0 && cmd_ctx->lsm->cmd_type != LTTNG_REGISTER_CONSUMER) { @@ -2835,6 +2973,7 @@ 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_JUL: case LTTNG_DOMAIN_UST: if (uatomic_read(&ust_consumerd_state) != CONSUMER_STARTED) { ret = LTTNG_ERR_NO_USTCONSUMERD; @@ -2916,9 +3055,74 @@ skip_domain: } case LTTNG_ENABLE_EVENT: { + struct lttng_event_exclusion *exclusion = NULL; + struct lttng_filter_bytecode *bytecode = NULL; + + /* Handle exclusion events and receive it from the client. */ + if (cmd_ctx->lsm->u.enable.exclusion_count > 0) { + size_t count = cmd_ctx->lsm->u.enable.exclusion_count; + + exclusion = zmalloc(sizeof(struct lttng_event_exclusion) + + (count * LTTNG_SYMBOL_NAME_LEN)); + if (!exclusion) { + ret = LTTNG_ERR_EXCLUSION_NOMEM; + goto error; + } + + DBG("Receiving var len exclusion event list from client ..."); + exclusion->count = count; + ret = lttcomm_recv_unix_sock(sock, exclusion->names, + count * LTTNG_SYMBOL_NAME_LEN); + if (ret <= 0) { + DBG("Nothing recv() from client var len data... continuing"); + *sock_error = 1; + free(exclusion); + ret = LTTNG_ERR_EXCLUSION_INVAL; + goto error; + } + } + + /* Handle filter and get bytecode from client. */ + if (cmd_ctx->lsm->u.enable.bytecode_len > 0) { + size_t bytecode_len = cmd_ctx->lsm->u.enable.bytecode_len; + + if (bytecode_len > LTTNG_FILTER_MAX_LEN) { + ret = LTTNG_ERR_FILTER_INVAL; + free(exclusion); + goto error; + } + + bytecode = zmalloc(bytecode_len); + if (!bytecode) { + free(exclusion); + ret = LTTNG_ERR_FILTER_NOMEM; + goto error; + } + + /* Receive var. len. data */ + DBG("Receiving var len filter's bytecode from client ..."); + ret = lttcomm_recv_unix_sock(sock, bytecode, bytecode_len); + if (ret <= 0) { + DBG("Nothing recv() from client car len data... continuing"); + *sock_error = 1; + free(bytecode); + free(exclusion); + ret = LTTNG_ERR_FILTER_INVAL; + goto error; + } + + if ((bytecode->len + sizeof(*bytecode)) != bytecode_len) { + free(bytecode); + free(exclusion); + ret = LTTNG_ERR_FILTER_INVAL; + goto error; + } + } + ret = cmd_enable_event(cmd_ctx->session, &cmd_ctx->lsm->domain, cmd_ctx->lsm->u.enable.channel_name, - &cmd_ctx->lsm->u.enable.event, NULL, kernel_poll_pipe[1]); + &cmd_ctx->lsm->u.enable.event, bytecode, exclusion, + kernel_poll_pipe[1]); break; } case LTTNG_ENABLE_ALL_EVENT: @@ -2935,7 +3139,9 @@ skip_domain: struct lttng_event *events; ssize_t nb_events; + session_lock_list(); nb_events = cmd_list_tracepoints(cmd_ctx->lsm->domain.type, &events); + session_unlock_list(); if (nb_events < 0) { /* Return value is a negative lttng_error_code. */ ret = -nb_events; @@ -2966,8 +3172,10 @@ skip_domain: struct lttng_event_field *fields; ssize_t nb_fields; + session_lock_list(); nb_fields = cmd_list_tracepoint_fields(cmd_ctx->lsm->domain.type, &fields); + session_unlock_list(); if (nb_fields < 0) { /* Return value is a negative lttng_error_code. */ ret = -nb_fields; @@ -3141,7 +3349,7 @@ skip_domain: case LTTNG_LIST_CHANNELS: { int nb_chan; - struct lttng_channel *channels; + struct lttng_channel *channels = NULL; nb_chan = cmd_list_channels(cmd_ctx->lsm->domain.type, cmd_ctx->session, &channels); @@ -3242,46 +3450,6 @@ skip_domain: cmd_ctx->lsm->u.reg.path, cdata); break; } - case LTTNG_ENABLE_EVENT_WITH_FILTER: - { - struct lttng_filter_bytecode *bytecode; - - if (cmd_ctx->lsm->u.enable.bytecode_len > LTTNG_FILTER_MAX_LEN) { - ret = LTTNG_ERR_FILTER_INVAL; - goto error; - } - if (cmd_ctx->lsm->u.enable.bytecode_len == 0) { - ret = LTTNG_ERR_FILTER_INVAL; - goto error; - } - bytecode = zmalloc(cmd_ctx->lsm->u.enable.bytecode_len); - if (!bytecode) { - ret = LTTNG_ERR_FILTER_NOMEM; - goto error; - } - /* Receive var. len. data */ - DBG("Receiving var len data from client ..."); - ret = lttcomm_recv_unix_sock(sock, bytecode, - cmd_ctx->lsm->u.enable.bytecode_len); - if (ret <= 0) { - DBG("Nothing recv() from client var len data... continuing"); - *sock_error = 1; - ret = LTTNG_ERR_FILTER_INVAL; - goto error; - } - - if (bytecode->len + sizeof(*bytecode) - != cmd_ctx->lsm->u.enable.bytecode_len) { - free(bytecode); - ret = LTTNG_ERR_FILTER_INVAL; - goto error; - } - - ret = cmd_enable_event(cmd_ctx->session, &cmd_ctx->lsm->domain, - cmd_ctx->lsm->u.enable.channel_name, - &cmd_ctx->lsm->u.enable.event, bytecode, kernel_poll_pipe[1]); - break; - } case LTTNG_DATA_PENDING: { ret = cmd_data_pending(cmd_ctx->session); @@ -3524,6 +3692,8 @@ static void *thread_manage_health(void *data) goto error; } + lttng_sessiond_notify_ready(); + while (1) { DBG("Health check ready"); @@ -3547,6 +3717,11 @@ restart: revents = LTTNG_POLL_GETEV(&events, i); pollfd = LTTNG_POLL_GETFD(&events, i); + if (!revents) { + /* No activity for this FD (poll implementation). */ + continue; + } + /* Thread quit pipe has been closed. Killing thread. */ ret = sessiond_check_thread_quit_pipe(pollfd, revents); if (ret) { @@ -3588,7 +3763,7 @@ restart: rcu_thread_online(); - reply.ret_code = 0; + memset(&reply, 0, sizeof(reply)); for (i = 0; i < NR_HEALTH_SESSIOND_TYPES; i++) { /* * health_check_state returns 0 if health is @@ -3652,10 +3827,6 @@ static void *thread_manage_clients(void *data) health_register(health_sessiond, HEALTH_SESSIOND_TYPE_CMD); - if (testpoint(thread_manage_clients)) { - goto error_testpoint; - } - health_code_update(); ret = lttcomm_listen_unix_sock(client_sock); @@ -3678,14 +3849,14 @@ static void *thread_manage_clients(void *data) goto error; } - /* - * Notify parent pid that we are ready to accept command for client side. - */ - if (opt_sig_parent) { - kill(ppid, SIGUSR1); + lttng_sessiond_notify_ready(); + + /* This testpoint is after we signal readiness to the parent. */ + if (testpoint(sessiond_thread_manage_clients)) { + goto error; } - if (testpoint(thread_manage_clients_before_loop)) { + if (testpoint(sessiond_thread_manage_clients_before_loop)) { goto error; } @@ -3718,6 +3889,11 @@ static void *thread_manage_clients(void *data) health_code_update(); + if (!revents) { + /* No activity for this FD (poll implementation). */ + continue; + } + /* Thread quit pipe has been closed. Killing thread. */ ret = sessiond_check_thread_quit_pipe(pollfd, revents); if (ret) { @@ -3860,7 +4036,6 @@ error: error_listen: error_create_poll: -error_testpoint: unlink(client_unix_sock_path); if (client_sock >= 0) { ret = close(client_sock); @@ -3903,14 +4078,16 @@ static void usage(void) fprintf(stderr, " --consumerd64-path PATH Specify path for the 64-bit UST consumer daemon binary\n"); fprintf(stderr, " --consumerd64-libdir PATH Specify path for the 64-bit UST consumer daemon libraries\n"); fprintf(stderr, " -d, --daemonize Start as a daemon.\n"); + fprintf(stderr, " -b, --background Start as a daemon, keeping console open.\n"); fprintf(stderr, " -g, --group NAME Specify the tracing group name. (default: tracing)\n"); fprintf(stderr, " -V, --version Show version number.\n"); - fprintf(stderr, " -S, --sig-parent Send SIGCHLD to parent pid to notify readiness.\n"); + fprintf(stderr, " -S, --sig-parent Send SIGUSR1 to parent pid to notify readiness.\n"); fprintf(stderr, " -q, --quiet No output at all.\n"); fprintf(stderr, " -v, --verbose Verbose mode. Activate DBG() macro.\n"); fprintf(stderr, " -p, --pidfile FILE Write a pid to FILE name overriding the default value.\n"); fprintf(stderr, " --verbose-consumer Verbose mode for consumer. Activate DBG() macro.\n"); fprintf(stderr, " --no-kernel Disable kernel tracer\n"); + fprintf(stderr, " --jul-tcp-port JUL application registration TCP port\n"); } /* @@ -3943,12 +4120,14 @@ static int parse_args(int argc, char **argv) { "verbose-consumer", 0, 0, 'Z' }, { "no-kernel", 0, 0, 'N' }, { "pidfile", 1, 0, 'p' }, + { "jul-tcp-port", 1, 0, 'J' }, + { "background", 0, 0, 'b' }, { NULL, 0, 0, 0 } }; while (1) { int option_index = 0; - c = getopt_long(argc, argv, "dhqvVSN" "a:c:g:s:C:E:D:F:Z:u:t:p:", + c = getopt_long(argc, argv, "dhqvVSN" "a:c:g:s:C:E:D:F:Z:u:t:p:J:b", long_options, &option_index); if (c == -1) { break; @@ -3970,6 +4149,9 @@ static int parse_args(int argc, char **argv) case 'd': opt_daemon = 1; break; + case 'b': + opt_background = 1; + break; case 'g': tracing_group_name = optarg; break; @@ -4028,6 +4210,24 @@ static int parse_args(int argc, char **argv) case 'p': opt_pidfile = optarg; break; + case 'J': /* JUL TCP port. */ + { + unsigned long v; + + errno = 0; + v = strtoul(optarg, NULL, 0); + if (errno != 0 || !isdigit(optarg[0])) { + ERR("Wrong value in --jul-tcp-port parameter: %s", optarg); + return -1; + } + if (v == 0 || v >= 65535) { + ERR("Port overflow in --jul-tcp-port parameter: %s", optarg); + return -1; + } + jul_tcp_port = (uint32_t) v; + DBG3("JUL TCP port set to non default: %u", jul_tcp_port); + break; + } default: /* Unknown option or other error. * Error is printed by getopt, just return */ @@ -4306,6 +4506,9 @@ static void sighandler(int sig) DBG("SIGTERM caught"); stop_threads(); break; + case SIGUSR1: + CMM_STORE_SHARED(recv_child_signal, 1); + break; default: break; } @@ -4344,7 +4547,12 @@ static int set_signal_handler(void) return ret; } - DBG("Signal handler set for SIGTERM, SIGPIPE and SIGINT"); + if ((ret = sigaction(SIGUSR1, &sa, NULL)) < 0) { + PERROR("sigaction"); + return ret; + } + + DBG("Signal handler set for SIGTERM, SIGUSR1, SIGPIPE and SIGINT"); return ret; } @@ -4400,6 +4608,52 @@ error: return; } +/* + * Create lockfile using the rundir and return its fd. + */ +static int create_lockfile(void) +{ + int ret; + char lockfile_path[PATH_MAX]; + + ret = generate_lock_file_path(lockfile_path, sizeof(lockfile_path)); + if (ret < 0) { + goto error; + } + + ret = utils_create_lock_file(lockfile_path); +error: + return ret; +} + +/* + * Write JUL TCP port using the rundir. + */ +static void write_julport(void) +{ + int ret; + char path[PATH_MAX]; + + assert(rundir); + + ret = snprintf(path, sizeof(path), "%s/" + DEFAULT_LTTNG_SESSIOND_JULPORT_FILE, rundir); + if (ret < 0) { + PERROR("snprintf julport path"); + goto error; + } + + /* + * Create TCP JUL port file in rundir. Return value is of no importance. + * The execution will continue even though we are not able to write the + * file. + */ + (void) utils_create_pid_file(jul_tcp_port, path); + +error: + return; +} + /* * main */ @@ -4413,6 +4667,10 @@ int main(int argc, char **argv) rcu_register_thread(); + if ((ret = set_signal_handler()) < 0) { + goto error; + } + setup_consumerd_path(); page_size = sysconf(_SC_PAGESIZE); @@ -4429,23 +4687,19 @@ int main(int argc, char **argv) } /* Daemonize */ - if (opt_daemon) { + if (opt_daemon || opt_background) { int i; - /* - * fork - * child: setsid, close FD 0, 1, 2, chdir / - * parent: exit (if fork is successful) - */ - ret = daemon(0, 0); + ret = lttng_daemonize(&child_ppid, &recv_child_signal, + !opt_background); if (ret < 0) { - PERROR("daemon"); goto error; } + /* - * We are in the child. Make sure all other file - * descriptors are closed, in case we are called with - * more opened file descriptors than the standard ones. + * We are in the child. Make sure all other file descriptors are + * closed, in case we are called with more opened file descriptors than + * the standard ones. */ for (i = 3; i < sysconf(_SC_OPEN_MAX); i++) { (void) close(i); @@ -4462,6 +4716,10 @@ int main(int argc, char **argv) if (is_root) { rundir = strdup(DEFAULT_LTTNG_RUNDIR); + if (!rundir) { + ret = -ENOMEM; + goto error; + } /* Create global run dir with root access */ ret = create_lttng_rundir(rundir); @@ -4548,6 +4806,11 @@ int main(int argc, char **argv) } } + lockfile_fd = create_lockfile(); + if (lockfile_fd < 0) { + goto error; + } + /* Set consumer initial state */ kernel_consumerd_state = CONSUMER_STOPPED; ust_consumerd_state = CONSUMER_STOPPED; @@ -4597,6 +4860,12 @@ int main(int argc, char **argv) */ ust_app_ht_alloc(); + /* Initialize JUL domain subsystem. */ + if ((ret = jul_init()) < 0) { + /* ENOMEM at this point. */ + goto error; + } + /* After this point, we can safely call cleanup() with "goto exit" */ /* @@ -4632,10 +4901,6 @@ int main(int argc, char **argv) goto exit; } - if ((ret = set_signal_handler()) < 0) { - goto exit; - } - /* Setup the needed unix socket */ if ((ret = init_daemon_socket()) < 0) { goto exit; @@ -4700,6 +4965,7 @@ int main(int argc, char **argv) } write_pidfile(); + write_julport(); /* Initialize communication library */ lttcomm_init(); @@ -4716,7 +4982,7 @@ int main(int argc, char **argv) goto exit_health_sessiond_cleanup; } - /* Create thread to manage the client socket */ + /* Create thread to clean up RCU hash tables */ ret = pthread_create(&ht_cleanup_thread, NULL, thread_ht_cleanup, (void *) NULL); if (ret != 0) { @@ -4724,7 +4990,7 @@ int main(int argc, char **argv) goto exit_ht_cleanup; } - /* Create thread to manage the client socket */ + /* Create health-check thread */ ret = pthread_create(&health_thread, NULL, thread_manage_health, (void *) NULL); if (ret != 0) { @@ -4768,10 +5034,18 @@ int main(int argc, char **argv) ret = pthread_create(&apps_notify_thread, NULL, ust_thread_manage_notify, (void *) NULL); if (ret != 0) { - PERROR("pthread_create apps"); + PERROR("pthread_create notify"); goto exit_apps_notify; } + /* Create JUL registration thread. */ + ret = pthread_create(&jul_reg_thread, NULL, + jul_thread_manage_registration, (void *) NULL); + if (ret != 0) { + PERROR("pthread_create JUL"); + goto exit_jul_reg; + } + /* Don't start this thread if kernel tracing is not requested nor root */ if (is_root && !opt_no_kernel) { /* Create kernel thread to manage kernel event */ @@ -4790,6 +5064,13 @@ int main(int argc, char **argv) } exit_kernel: + ret = pthread_join(jul_reg_thread, &status); + if (ret != 0) { + PERROR("pthread_join JUL"); + goto error; /* join error, exit without cleanup */ + } + +exit_jul_reg: ret = pthread_join(apps_notify_thread, &status); if (ret != 0) { PERROR("pthread_join apps notify");