X-Git-Url: http://git.lttng.org./?a=blobdiff_plain;f=src%2Fbin%2Flttng-sessiond%2Fmain.c;h=bbdf141b2409825ee1b900453b2914ec03d7c9d3;hb=06b55dbbd4ce78bc7a3aad8a811ecb23c9f02862;hp=e119c8bf0305f6d19084dcfe04f9f12abbe12c9e;hpb=0c89d7959f4af7a2b39e925b2e8dc3a8954ccfa5;p=lttng-tools.git diff --git a/src/bin/lttng-sessiond/main.c b/src/bin/lttng-sessiond/main.c index e119c8bf0..bbdf141b2 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" @@ -63,23 +65,25 @@ #include "health-sessiond.h" #include "testpoint.h" #include "ust-thread.h" +#include "jul-thread.h" #define CONSUMERD_FILE "lttng-consumerd" -/* Const values */ -const char default_tracing_group[] = DEFAULT_TRACING_GROUP; - const char *progname; -const char *opt_tracing_group; +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; +/* 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 * set to 0 by the static definition. @@ -159,6 +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 @@ -236,6 +241,41 @@ 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; + +/* + * 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) { @@ -325,25 +365,6 @@ int sessiond_check_thread_quit_pipe(int fd, uint32_t events) return 0; } -/* - * Return group ID of the tracing group or -1 if not found. - */ -static gid_t allowed_group(void) -{ - struct group *grp; - - if (opt_tracing_group) { - grp = getgrnam(opt_tracing_group); - } else { - grp = getgrnam(default_tracing_group); - } - if (!grp) { - return -1; - } else { - return grp->gr_gid; - } -} - /* * Init thread quit pipe. * @@ -441,8 +462,8 @@ static void close_consumer_sockets(void) static void cleanup(void) { int ret; - char *cmd = NULL; struct ltt_session *sess, *stmp; + char path[PATH_MAX]; DBG("Cleaning up"); @@ -463,18 +484,65 @@ static void cleanup(void) } } - DBG("Removing %s directory", rundir); - ret = asprintf(&cmd, "rm -rf %s", rundir); - if (ret < 0) { - ERR("asprintf failed. Something is really wrong!"); - } + DBG("Removing sessiond and consumerd content of directory %s", rundir); + + /* sessiond */ + snprintf(path, PATH_MAX, + "%s/%s", + rundir, DEFAULT_LTTNG_SESSIOND_PIDFILE); + 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, + rundir); + DBG("Removing %s", path); + (void) unlink(path); + + snprintf(path, PATH_MAX, + DEFAULT_KCONSUMERD_PATH, + rundir); + DBG("Removing directory %s", path); + (void) rmdir(path); + + /* ust consumerd 32 */ + snprintf(path, PATH_MAX, + DEFAULT_USTCONSUMERD32_ERR_SOCK_PATH, + rundir); + DBG("Removing %s", path); + (void) unlink(path); + + snprintf(path, PATH_MAX, + DEFAULT_USTCONSUMERD32_PATH, + rundir); + DBG("Removing directory %s", path); + (void) rmdir(path); + + /* ust consumerd 64 */ + snprintf(path, PATH_MAX, + DEFAULT_USTCONSUMERD64_ERR_SOCK_PATH, + rundir); + DBG("Removing %s", path); + (void) unlink(path); + + snprintf(path, PATH_MAX, + DEFAULT_USTCONSUMERD64_PATH, + rundir); + DBG("Removing directory %s", path); + (void) rmdir(path); + + /* + * We do NOT rmdir rundir because there are other processes + * using it, for instance lttng-relayd, which can start in + * parallel with this teardown. + */ - /* Remove lttng run directory */ - ret = system(cmd); - if (ret < 0) { - ERR("Unable to clean %s", rundir); - } - free(cmd); free(rundir); DBG("Cleaning up all sessions"); @@ -758,7 +826,7 @@ static void *thread_manage_kernel(void *data) DBG("[thread] Thread manage kernel started"); - health_register(health_sessiond, HEALTH_TYPE_KERNEL); + health_register(health_sessiond, HEALTH_SESSIOND_TYPE_KERNEL); /* * This first step of the while is to clean this structure which could free @@ -766,13 +834,13 @@ 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; } @@ -841,9 +909,8 @@ 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. @@ -924,7 +991,7 @@ static void *thread_manage_consumer(void *data) DBG("[thread] Manage consumer started"); - health_register(health_sessiond, HEALTH_TYPE_CONSUMER); + health_register(health_sessiond, HEALTH_SESSIOND_TYPE_CONSUMER); health_code_update(); @@ -953,7 +1020,7 @@ static void *thread_manage_consumer(void *data) restart: health_poll_entry(); - if (testpoint(thread_manage_consumer)) { + if (testpoint(sessiond_thread_manage_consumer)) { goto error; } @@ -1215,6 +1282,7 @@ 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; @@ -1223,9 +1291,9 @@ static void *thread_manage_apps(void *data) rcu_register_thread(); rcu_thread_online(); - health_register(health_sessiond, HEALTH_TYPE_APP_MANAGE); + health_register(health_sessiond, HEALTH_SESSIOND_TYPE_APP_MANAGE); - if (testpoint(thread_manage_apps)) { + if (testpoint(sessiond_thread_manage_apps)) { goto error_testpoint; } @@ -1241,7 +1309,7 @@ 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; } @@ -1290,10 +1358,8 @@ 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; } @@ -1311,23 +1377,7 @@ 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 { /* @@ -1343,7 +1393,6 @@ static void *thread_manage_apps(void *data) /* Socket closed on remote end. */ ust_app_unregister(pollfd); - break; } } @@ -1388,7 +1437,7 @@ 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 @@ -1399,10 +1448,8 @@ 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; @@ -1413,7 +1460,7 @@ 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; } /* @@ -1515,7 +1562,11 @@ static void *thread_dispatch_ust_registration(void *data) .count = 0, }; - health_register(health_sessiond, HEALTH_TYPE_APP_REG_DISPATCH); + health_register(health_sessiond, HEALTH_SESSIOND_TYPE_APP_REG_DISPATCH); + + if (testpoint(sessiond_thread_app_reg_dispatch)) { + goto error_testpoint; + } health_code_update(); @@ -1566,7 +1617,7 @@ 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; } @@ -1580,7 +1631,7 @@ 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; @@ -1628,7 +1679,7 @@ 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); } @@ -1721,6 +1772,7 @@ error: free(wait_node); } +error_testpoint: DBG("Dispatch thread dying"); if (err) { health_error(); @@ -1746,9 +1798,9 @@ static void *thread_registration_apps(void *data) DBG("[thread] Manage application registration started"); - health_register(health_sessiond, HEALTH_TYPE_APP_REG); + health_register(health_sessiond, HEALTH_SESSIOND_TYPE_APP_REG); - if (testpoint(thread_registration_apps)) { + if (testpoint(sessiond_thread_registration_apps)) { goto error_testpoint; } @@ -1825,6 +1877,18 @@ 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. @@ -1897,11 +1961,6 @@ 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); @@ -1926,6 +1985,10 @@ 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; @@ -2127,6 +2190,7 @@ static pid_t spawn_consumerd(struct consumer_data *consumer_data) "lttng-consumerd", verbosity, "-k", "--consumerd-cmd-sock", consumer_data->cmd_unix_sock_path, "--consumerd-err-sock", consumer_data->err_unix_sock_path, + "--group", tracing_group_name, NULL); break; case LTTNG_CONSUMER64_UST: @@ -2165,6 +2229,7 @@ static pid_t spawn_consumerd(struct consumer_data *consumer_data) ret = execl(consumerd64_bin, "lttng-consumerd", verbosity, "-u", "--consumerd-cmd-sock", consumer_data->cmd_unix_sock_path, "--consumerd-err-sock", consumer_data->err_unix_sock_path, + "--group", tracing_group_name, NULL); if (consumerd64_libdir[0] != '\0') { free(tmpnew); @@ -2210,6 +2275,7 @@ static pid_t spawn_consumerd(struct consumer_data *consumer_data) ret = execl(consumerd32_bin, "lttng-consumerd", verbosity, "-u", "--consumerd-cmd-sock", consumer_data->cmd_unix_sock_path, "--consumerd-err-sock", consumer_data->err_unix_sock_path, + "--group", tracing_group_name, NULL); if (consumerd32_libdir[0] != '\0') { free(tmpnew); @@ -2295,21 +2361,6 @@ error: return ret; } -/* - * Compute health status of each consumer. If one of them is zero (bad - * state), we return 0. - */ -static int check_consumer_health(void) -{ - int ret; - - ret = health_check_state(health_sessiond, HEALTH_TYPE_CONSUMER); - - DBG3("Health consumer check %d", ret); - - return ret; -} - /* * Setup necessary data for kernel tracer action. */ @@ -2408,6 +2459,7 @@ 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) { @@ -2451,6 +2503,7 @@ 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: @@ -2742,6 +2795,7 @@ 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()) { @@ -2833,6 +2887,7 @@ 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; @@ -2914,9 +2969,74 @@ 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: @@ -3240,46 +3360,6 @@ 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); @@ -3475,6 +3555,27 @@ static void *thread_manage_health(void *data) goto error; } + if (is_root) { + /* lttng health client socket path permissions */ + ret = chown(health_unix_sock_path, 0, + utils_get_group_id(tracing_group_name)); + if (ret < 0) { + ERR("Unable to set group on %s", health_unix_sock_path); + PERROR("chown"); + ret = -1; + goto error; + } + + ret = chmod(health_unix_sock_path, + S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP); + if (ret < 0) { + ERR("Unable to set permissions on %s", health_unix_sock_path); + PERROR("chmod"); + ret = -1; + goto error; + } + } + /* * Set the CLOEXEC flag. Return code is useless because either way, the * show must go on. @@ -3501,6 +3602,8 @@ static void *thread_manage_health(void *data) goto error; } + lttng_sessiond_notify_ready(); + while (1) { DBG("Health check ready"); @@ -3565,58 +3668,18 @@ restart: rcu_thread_online(); - switch (msg.component) { - case LTTNG_HEALTH_CMD: - reply.ret_code = health_check_state(health_sessiond, HEALTH_TYPE_CMD); - break; - case LTTNG_HEALTH_APP_MANAGE: - reply.ret_code = health_check_state(health_sessiond, HEALTH_TYPE_APP_MANAGE); - break; - case LTTNG_HEALTH_APP_REG: - reply.ret_code = health_check_state(health_sessiond, HEALTH_TYPE_APP_REG); - break; - case LTTNG_HEALTH_KERNEL: - reply.ret_code = health_check_state(health_sessiond, HEALTH_TYPE_KERNEL); - break; - case LTTNG_HEALTH_CONSUMER: - reply.ret_code = check_consumer_health(); - break; - case LTTNG_HEALTH_HT_CLEANUP: - reply.ret_code = health_check_state(health_sessiond, HEALTH_TYPE_HT_CLEANUP); - break; - case LTTNG_HEALTH_APP_MANAGE_NOTIFY: - reply.ret_code = health_check_state(health_sessiond, HEALTH_TYPE_APP_MANAGE_NOTIFY); - break; - case LTTNG_HEALTH_APP_REG_DISPATCH: - reply.ret_code = health_check_state(health_sessiond, HEALTH_TYPE_APP_REG_DISPATCH); - break; - case LTTNG_HEALTH_ALL: - reply.ret_code = - health_check_state(health_sessiond, HEALTH_TYPE_APP_MANAGE) && - health_check_state(health_sessiond, HEALTH_TYPE_APP_REG) && - health_check_state(health_sessiond, HEALTH_TYPE_CMD) && - health_check_state(health_sessiond, HEALTH_TYPE_KERNEL) && - check_consumer_health() && - health_check_state(health_sessiond, HEALTH_TYPE_HT_CLEANUP) && - health_check_state(health_sessiond, HEALTH_TYPE_APP_MANAGE_NOTIFY) && - health_check_state(health_sessiond, HEALTH_TYPE_APP_REG_DISPATCH); - break; - default: - reply.ret_code = LTTNG_ERR_UND; - break; - } - - /* - * Flip ret value since 0 is a success and 1 indicates a bad health for - * the client where in the sessiond it is the opposite. Again, this is - * just to make things easier for us poor developer which enjoy a lot - * lazyness. - */ - if (reply.ret_code == 0 || reply.ret_code == 1) { - reply.ret_code = !reply.ret_code; + reply.ret_code = 0; + for (i = 0; i < NR_HEALTH_SESSIOND_TYPES; i++) { + /* + * health_check_state returns 0 if health is + * bad. + */ + if (!health_check_state(health_sessiond, i)) { + reply.ret_code |= 1ULL << i; + } } - DBG2("Health check return value %d", reply.ret_code); + DBG2("Health check return value %" PRIx64, reply.ret_code); ret = send_unix_sock(new_sock, (void *) &reply, sizeof(reply)); if (ret < 0) { @@ -3667,11 +3730,7 @@ static void *thread_manage_clients(void *data) rcu_register_thread(); - health_register(health_sessiond, HEALTH_TYPE_CMD); - - if (testpoint(thread_manage_clients)) { - goto error_testpoint; - } + health_register(health_sessiond, HEALTH_SESSIOND_TYPE_CMD); health_code_update(); @@ -3695,14 +3754,14 @@ 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; } @@ -3877,7 +3936,6 @@ error: error_listen: error_create_poll: -error_testpoint: unlink(client_unix_sock_path); if (client_sock >= 0) { ret = close(client_sock); @@ -3920,14 +3978,16 @@ 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"); } /* @@ -3960,12 +4020,14 @@ 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; @@ -3987,8 +4049,11 @@ static int parse_args(int argc, char **argv) case 'd': opt_daemon = 1; break; + case 'b': + opt_background = 1; + break; case 'g': - opt_tracing_group = optarg; + tracing_group_name = optarg; break; case 'h': usage(); @@ -4045,6 +4110,24 @@ 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 */ @@ -4149,14 +4232,7 @@ static int set_permissions(char *rundir) int ret; gid_t gid; - ret = allowed_group(); - if (ret < 0) { - WARN("No tracing group detected"); - /* Setting gid to 0 if no tracing group is found */ - gid = 0; - } else { - gid = ret; - } + gid = utils_get_group_id(tracing_group_name); /* Set lttng run dir */ ret = chown(rundir, 0, gid); @@ -4165,8 +4241,12 @@ static int set_permissions(char *rundir) PERROR("chown"); } - /* Ensure all applications and tracing group can search the run dir */ - ret = chmod(rundir, S_IRWXU | S_IXGRP | S_IXOTH); + /* + * Ensure all applications and tracing group can search the run + * dir. Allow everyone to read the directory, since it does not + * buy us anything to hide its content. + */ + ret = chmod(rundir, S_IRWXU | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH); if (ret < 0) { ERR("Unable to set permissions on %s", rundir); PERROR("chmod"); @@ -4180,21 +4260,21 @@ static int set_permissions(char *rundir) } /* kconsumer error socket path */ - ret = chown(kconsumer_data.err_unix_sock_path, 0, gid); + ret = chown(kconsumer_data.err_unix_sock_path, 0, 0); if (ret < 0) { ERR("Unable to set group on %s", kconsumer_data.err_unix_sock_path); PERROR("chown"); } /* 64-bit ustconsumer error socket path */ - ret = chown(ustconsumer64_data.err_unix_sock_path, 0, gid); + ret = chown(ustconsumer64_data.err_unix_sock_path, 0, 0); if (ret < 0) { ERR("Unable to set group on %s", ustconsumer64_data.err_unix_sock_path); PERROR("chown"); } /* 32-bit ustconsumer compat32 error socket path */ - ret = chown(ustconsumer32_data.err_unix_sock_path, 0, gid); + ret = chown(ustconsumer32_data.err_unix_sock_path, 0, 0); if (ret < 0) { ERR("Unable to set group on %s", ustconsumer32_data.err_unix_sock_path); PERROR("chown"); @@ -4238,7 +4318,7 @@ static int set_consumer_sockets(struct consumer_data *consumer_data, int ret; char path[PATH_MAX]; - switch (consumer_data->type) { + switch (consumer_data->type) { case LTTNG_CONSUMER_KERNEL: snprintf(path, PATH_MAX, DEFAULT_KCONSUMERD_PATH, rundir); break; @@ -4256,7 +4336,7 @@ static int set_consumer_sockets(struct consumer_data *consumer_data, DBG2("Creating consumer directory: %s", path); - ret = mkdir(path, S_IRWXU); + ret = mkdir(path, S_IRWXU | S_IRGRP | S_IXGRP); if (ret < 0) { if (errno != EEXIST) { PERROR("mkdir"); @@ -4265,6 +4345,14 @@ static int set_consumer_sockets(struct consumer_data *consumer_data, } ret = -1; } + if (is_root) { + ret = chown(path, 0, utils_get_group_id(tracing_group_name)); + if (ret < 0) { + ERR("Unable to set group on %s", path); + PERROR("chown"); + goto error; + } + } /* Create the kconsumerd error unix socket */ consumer_data->err_sock = @@ -4318,6 +4406,9 @@ static void sighandler(int sig) DBG("SIGTERM caught"); stop_threads(); break; + case SIGUSR1: + CMM_STORE_SHARED(recv_child_signal, 1); + break; default: break; } @@ -4356,7 +4447,12 @@ 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; } @@ -4412,6 +4508,34 @@ error: return; } +/* + * 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 */ @@ -4425,6 +4549,10 @@ 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); @@ -4441,23 +4569,19 @@ 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); @@ -4609,6 +4733,12 @@ 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" */ /* @@ -4644,10 +4774,6 @@ 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; @@ -4712,6 +4838,7 @@ int main(int argc, char **argv) } write_pidfile(); + write_julport(); /* Initialize communication library */ lttcomm_init(); @@ -4722,13 +4849,13 @@ int main(int argc, char **argv) * Initialize the health check subsystem. This call should set the * appropriate time values. */ - health_sessiond = health_app_create(HEALTH_NUM_TYPE); + health_sessiond = health_app_create(NR_HEALTH_SESSIOND_TYPES); if (!health_sessiond) { PERROR("health_app_create error"); 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) { @@ -4736,7 +4863,7 @@ 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) { @@ -4784,6 +4911,14 @@ int main(int argc, char **argv) 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 apps"); + 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 */ @@ -4802,6 +4937,13 @@ 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");