X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fcore%2Fservice.c;h=694a2658e9744beeb7ecff0031daa8a72d8c91b9;hp=4eb3d9e6688d1ff4990dc30a08d424f8001b2135;hb=bf500566323bbc2240d1fdd1165a8c908faf4098;hpb=09812eb764b440651f3ff4cb5d37bd343f800560 diff --git a/src/core/service.c b/src/core/service.c index 4eb3d9e66..694a2658e 100644 --- a/src/core/service.c +++ b/src/core/service.c @@ -24,6 +24,8 @@ #include #include #include +#include +#include #include "manager.h" #include "unit.h" @@ -143,13 +145,7 @@ static void service_init(Unit *u) { s->socket_fd = -1; s->guess_main_pid = true; - exec_context_init(&s->exec_context); - kill_context_init(&s->kill_context); - cgroup_context_init(&s->cgroup_context); - - RATELIMIT_INIT(s->start_limit, - u->manager->default_start_limit_interval, - u->manager->default_start_limit_burst); + RATELIMIT_INIT(s->start_limit, u->manager->default_start_limit_interval, u->manager->default_start_limit_burst); s->control_command_id = _SERVICE_EXEC_COMMAND_INVALID; } @@ -210,8 +206,8 @@ static int service_set_main_pid(Service *s, pid_t pid) { if (get_parent_of_pid(pid, &ppid) >= 0 && ppid != getpid()) { log_warning_unit(UNIT(s)->id, - "%s: Supervising process %lu which is not our child. We'll most likely not notice when it exits.", - UNIT(s)->id, (unsigned long) pid); + "%s: Supervising process "PID_FMT" which is not our child. We'll most likely not notice when it exits.", + UNIT(s)->id, pid); s->main_pid_alien = true; } else @@ -226,8 +222,7 @@ static void service_close_socket_fd(Service *s) { if (s->socket_fd < 0) return; - close_nointr_nofail(s->socket_fd); - s->socket_fd = -1; + s->socket_fd = safe_close(s->socket_fd); } static void service_connection_unref(Service *s) { @@ -264,7 +259,12 @@ static void service_start_watchdog(Service *s) { r = sd_event_source_set_enabled(s->watchdog_event_source, SD_EVENT_ONESHOT); } else { - r = sd_event_add_monotonic(UNIT(s)->manager->event, s->watchdog_timestamp.monotonic + s->watchdog_usec, 0, service_dispatch_watchdog, s, &s->watchdog_event_source); + r = sd_event_add_time( + UNIT(s)->manager->event, + &s->watchdog_event_source, + CLOCK_MONOTONIC, + s->watchdog_timestamp.monotonic + s->watchdog_usec, 0, + service_dispatch_watchdog, s); if (r < 0) { log_warning_unit(UNIT(s)->id, "%s failed to add watchdog timer: %s", UNIT(s)->id, strerror(-r)); return; @@ -302,8 +302,9 @@ static void service_done(Unit *u) { free(s->status_text); s->status_text = NULL; - cgroup_context_done(&s->cgroup_context); - exec_context_done(&s->exec_context); + free(s->reboot_arg); + s->reboot_arg = NULL; + s->exec_runtime = exec_runtime_unref(s->exec_runtime); exec_command_free_array(s->exec_command, _SERVICE_EXEC_COMMAND_MAX); s->control_command = NULL; @@ -354,14 +355,19 @@ static int service_arm_timer(Service *s, usec_t usec) { return sd_event_source_set_enabled(s->timer_event_source, SD_EVENT_ONESHOT); } - return sd_event_add_monotonic(UNIT(s)->manager->event, now(CLOCK_MONOTONIC) + usec, 0, service_dispatch_timer, s, &s->timer_event_source); + return sd_event_add_time( + UNIT(s)->manager->event, + &s->timer_event_source, + CLOCK_MONOTONIC, + now(CLOCK_MONOTONIC) + usec, 0, + service_dispatch_timer, s); } #ifdef HAVE_SYSV_COMPAT static char *sysv_translate_name(const char *name) { char *r; - r = new(char, strlen(name) + sizeof(".service")); + r = new(char, strlen(name) + strlen(".service") + 1); if (!r) return NULL; @@ -663,13 +669,8 @@ static int service_load_sysv_path(Service *s, const char *path) { state = NORMAL; - if (sscanf(t+10, "%15s %i %*i", - runlevels, - &start_priority) != 2) { - - log_warning_unit(u->id, - "[%s:%u] Failed to parse chkconfig line. Ignoring.", - path, line); + if (sscanf(t+10, "%15s %i %*i", runlevels, &start_priority) != 2) { + log_warning_unit(u->id, "[%s:%u] Failed to parse chkconfig line. Ignoring.", path, line); continue; } @@ -677,19 +678,17 @@ static int service_load_sysv_path(Service *s, const char *path) { * symlink farms is preferred over the * data from the LSB header. */ if (start_priority < 0 || start_priority > 99) - log_warning_unit(u->id, - "[%s:%u] Start priority out of range. Ignoring.", - path, line); + log_warning_unit(u->id, "[%s:%u] Start priority out of range. Ignoring.", path, line); else - s->sysv_start_priority = start_priority; + log_debug_unit(u->id, "[%s:%u] Ignoring start priority set in the chkconfig file.", path, line); char_array_0(runlevels); k = delete_chars(runlevels, WHITESPACE "-"); - if (k[0]) { char *d; - if (!(d = strdup(k))) { + d = strdup(k); + if (!d) { r = -ENOMEM; goto finish; } @@ -991,9 +990,9 @@ static int service_load_sysv_path(Service *s, const char *path) { u->description = d; } - /* The priority that has been set in /etc/rcN.d/ hierarchies - * takes precedence over what is stored as default in the LSB - * header */ + /* Initialize the start priority from what has been set in the + * /etc/rcN.d/ hierarchies if we load the unit file as SysV + * init script. */ if (s->sysv_start_priority_from_rcnd >= 0) s->sysv_start_priority = s->sysv_start_priority_from_rcnd; @@ -1092,37 +1091,31 @@ static int service_verify(Service *s) { return 0; if (!s->exec_command[SERVICE_EXEC_START]) { - log_error_unit(UNIT(s)->id, - "%s lacks ExecStart setting. Refusing.", UNIT(s)->id); + log_error_unit(UNIT(s)->id, "%s lacks ExecStart setting. Refusing.", UNIT(s)->id); return -EINVAL; } if (s->type != SERVICE_ONESHOT && s->exec_command[SERVICE_EXEC_START]->command_next) { - log_error_unit(UNIT(s)->id, - "%s has more than one ExecStart setting, which is only allowed for Type=oneshot services. Refusing.", UNIT(s)->id); + log_error_unit(UNIT(s)->id, "%s has more than one ExecStart setting, which is only allowed for Type=oneshot services. Refusing.", UNIT(s)->id); return -EINVAL; } if (s->type == SERVICE_ONESHOT && s->restart != SERVICE_RESTART_NO) { - log_error_unit(UNIT(s)->id, - "%s has Restart setting other than no, which isn't allowed for Type=oneshot services. Refusing.", UNIT(s)->id); + log_error_unit(UNIT(s)->id, "%s has Restart setting other than no, which isn't allowed for Type=oneshot services. Refusing.", UNIT(s)->id); return -EINVAL; } if (s->type == SERVICE_DBUS && !s->bus_name) { - log_error_unit(UNIT(s)->id, - "%s is of type D-Bus but no D-Bus service name has been specified. Refusing.", UNIT(s)->id); + log_error_unit(UNIT(s)->id, "%s is of type D-Bus but no D-Bus service name has been specified. Refusing.", UNIT(s)->id); return -EINVAL; } if (s->bus_name && s->type != SERVICE_DBUS) - log_warning_unit(UNIT(s)->id, - "%s has a D-Bus service name specified, but is not of type dbus. Ignoring.", UNIT(s)->id); + log_warning_unit(UNIT(s)->id, "%s has a D-Bus service name specified, but is not of type dbus. Ignoring.", UNIT(s)->id); - if (s->exec_context.pam_name && s->kill_context.kill_mode != KILL_CONTROL_GROUP) { - log_error_unit(UNIT(s)->id, - "%s has PAM enabled. Kill mode must be set to 'control-group'. Refusing.", UNIT(s)->id); + if (s->exec_context.pam_name && !(s->kill_context.kill_mode == KILL_CONTROL_GROUP || s->kill_context.kill_mode == KILL_MIXED)) { + log_error_unit(UNIT(s)->id, "%s has PAM enabled. Kill mode must be set to 'control-group' or 'mixed'. Refusing.", UNIT(s)->id); return -EINVAL; } @@ -1138,28 +1131,10 @@ static int service_add_default_dependencies(Service *s) { * majority of services. */ /* First, pull in base system */ - if (UNIT(s)->manager->running_as == SYSTEMD_SYSTEM) { - r = unit_add_two_dependencies_by_name(UNIT(s), UNIT_AFTER, UNIT_REQUIRES, - SPECIAL_BASIC_TARGET, NULL, true); - if (r < 0) - return r; - - } else if (UNIT(s)->manager->running_as == SYSTEMD_USER) { - r = unit_add_two_dependencies_by_name(UNIT(s), UNIT_AFTER, UNIT_REQUIRES, - SPECIAL_SOCKETS_TARGET, NULL, true); - if (r < 0) - return r; - - r = unit_add_two_dependencies_by_name(UNIT(s), UNIT_AFTER, UNIT_REQUIRES, - SPECIAL_TIMERS_TARGET, NULL, true); - if (r < 0) - return r; - - r = unit_add_two_dependencies_by_name(UNIT(s), UNIT_AFTER, UNIT_REQUIRES, - SPECIAL_PATHS_TARGET, NULL, true); - if (r < 0) - return r; - } + r = unit_add_two_dependencies_by_name(UNIT(s), UNIT_AFTER, UNIT_REQUIRES, + SPECIAL_BASIC_TARGET, NULL, true); + if (r < 0) + return r; /* Second, activate normal shutdown */ r = unit_add_two_dependencies_by_name(UNIT(s), UNIT_BEFORE, UNIT_CONFLICTS, @@ -1227,11 +1202,15 @@ static int service_load(Unit *u) { service_fix_output(s); + r = unit_patch_contexts(u); + if (r < 0) + return r; + r = unit_add_exec_dependencies(u, &s->exec_context); if (r < 0) return r; - r = unit_add_default_slice(u); + r = unit_add_default_slice(u, &s->cgroup_context); if (r < 0) return r; @@ -1241,33 +1220,23 @@ static int service_load(Unit *u) { return r; #endif - if (s->bus_name) { - r = unit_watch_bus_name(u, s->bus_name); - if (r < 0) - return r; - } - if (s->type == SERVICE_NOTIFY && s->notify_access == NOTIFY_NONE) s->notify_access = NOTIFY_MAIN; if (s->watchdog_usec > 0 && s->notify_access == NOTIFY_NONE) s->notify_access = NOTIFY_MAIN; - if (s->type == SERVICE_DBUS || s->bus_name) { - r = unit_add_two_dependencies_by_name(u, UNIT_AFTER, UNIT_REQUIRES, SPECIAL_DBUS_SOCKET, NULL, true); + if (s->bus_name) { + r = unit_watch_bus_name(u, s->bus_name); if (r < 0) return r; } - if (UNIT(s)->default_dependencies) { + if (u->default_dependencies) { r = service_add_default_dependencies(s); if (r < 0) return r; } - - r = unit_exec_context_defaults(u, &s->exec_context); - if (r < 0) - return r; } return service_verify(s); @@ -1309,15 +1278,15 @@ static void service_dump(Unit *u, FILE *f, const char *prefix) { if (s->control_pid > 0) fprintf(f, - "%sControl PID: %lu\n", - prefix, (unsigned long) s->control_pid); + "%sControl PID: "PID_FMT"\n", + prefix, s->control_pid); if (s->main_pid > 0) fprintf(f, - "%sMain PID: %lu\n" + "%sMain PID: "PID_FMT"\n" "%sMain PID Known: %s\n" "%sMain PID Alien: %s\n", - prefix, (unsigned long) s->main_pid, + prefix, s->main_pid, prefix, yes_no(s->main_pid_known), prefix, yes_no(s->main_pid_alien)); @@ -1398,11 +1367,10 @@ static int service_load_pid_file(Service *s, bool may_warn) { return r; } - if (kill(pid, 0) < 0 && errno != EPERM) { + if (!pid_is_alive(pid)) { if (may_warn) - log_info_unit(UNIT(s)->id, - "PID %lu read from file %s does not exist.", - (unsigned long) pid, s->pid_file); + log_info_unit(UNIT(s)->id, "PID "PID_FMT" read from file %s does not exist or is a zombie.", pid, s->pid_file); + return -ESRCH; } @@ -1411,13 +1379,13 @@ static int service_load_pid_file(Service *s, bool may_warn) { return 0; log_debug_unit(UNIT(s)->id, - "Main PID changing: %lu -> %lu", - (unsigned long) s->main_pid, (unsigned long) pid); + "Main PID changing: "PID_FMT" -> "PID_FMT, + s->main_pid, pid); service_unwatch_main_pid(s); s->main_pid_known = false; } else log_debug_unit(UNIT(s)->id, - "Main PID loaded: %lu", (unsigned long) pid); + "Main PID loaded: "PID_FMT, pid); r = service_set_main_pid(s, pid); if (r < 0) @@ -1427,8 +1395,8 @@ static int service_load_pid_file(Service *s, bool may_warn) { if (r < 0) { /* FIXME: we need to do something here */ log_warning_unit(UNIT(s)->id, - "Failed to watch PID %lu from service %s", - (unsigned long) pid, UNIT(s)->id); + "Failed to watch PID "PID_FMT" from service %s", + pid, UNIT(s)->id); return r; } @@ -1456,7 +1424,7 @@ static int service_search_main_pid(Service *s) { return -ENOENT; log_debug_unit(UNIT(s)->id, - "Main PID guessed: %lu", (unsigned long) pid); + "Main PID guessed: "PID_FMT, pid); r = service_set_main_pid(s, pid); if (r < 0) return r; @@ -1465,11 +1433,9 @@ static int service_search_main_pid(Service *s) { if (r < 0) /* FIXME: we need to do something here */ log_warning_unit(UNIT(s)->id, - "Failed to watch PID %lu from service %s", - (unsigned long) pid, UNIT(s)->id); - return r; - - return 0; + "Failed to watch PID "PID_FMT" from service %s", + pid, UNIT(s)->id); + return r; } static void service_set_state(Service *s, ServiceState state) { @@ -1515,6 +1481,9 @@ static void service_set_state(Service *s, ServiceState state) { s->control_command_id = _SERVICE_EXEC_COMMAND_INVALID; } + if (IN_SET(state, SERVICE_DEAD, SERVICE_FAILED, SERVICE_AUTO_RESTART)) + unit_unwatch_all_pids(UNIT(s)); + if (!IN_SET(state, SERVICE_START_PRE, SERVICE_START, SERVICE_START_POST, SERVICE_RUNNING, SERVICE_RELOAD, @@ -1592,7 +1561,7 @@ static int service_coldplug(Unit *u) { return r; } - if (pid_valid(s->main_pid) && + if (pid_is_unwaited(s->main_pid) && ((s->deserialized_state == SERVICE_START && IN_SET(s->type, SERVICE_FORKING, SERVICE_DBUS, SERVICE_ONESHOT, SERVICE_NOTIFY)) || IN_SET(s->deserialized_state, SERVICE_START, SERVICE_START_POST, @@ -1605,7 +1574,7 @@ static int service_coldplug(Unit *u) { return r; } - if (pid_valid(s->control_pid) && + if (pid_is_unwaited(s->control_pid) && IN_SET(s->deserialized_state, SERVICE_START_PRE, SERVICE_START, SERVICE_START_POST, SERVICE_RELOAD, @@ -1617,6 +1586,9 @@ static int service_coldplug(Unit *u) { return r; } + if (!IN_SET(s->deserialized_state, SERVICE_DEAD, SERVICE_FAILED, SERVICE_AUTO_RESTART)) + unit_watch_all_pids(UNIT(s)); + if (IN_SET(s->deserialized_state, SERVICE_START_POST, SERVICE_RUNNING, SERVICE_RELOAD)) service_start_watchdog(s); @@ -1763,13 +1735,13 @@ static int service_spawn( } if (s->main_pid > 0) - if (asprintf(our_env + n_env++, "MAINPID=%lu", (unsigned long) s->main_pid) < 0) { + if (asprintf(our_env + n_env++, "MAINPID="PID_FMT, s->main_pid) < 0) { r = -ENOMEM; goto fail; } if (UNIT(s)->manager->running_as != SYSTEMD_SYSTEM) - if (asprintf(our_env + n_env++, "MANAGERPID=%lu", (unsigned long) getpid()) < 0) { + if (asprintf(our_env + n_env++, "MANAGERPID="PID_FMT, getpid()) < 0) { r = -ENOMEM; goto fail; } @@ -1797,6 +1769,7 @@ static int service_spawn( UNIT(s)->manager->confirm_spawn, UNIT(s)->manager->cgroup_supported, path, + manager_get_runtime_prefix(UNIT(s)->manager), UNIT(s)->id, s->watchdog_usec, s->type == SERVICE_IDLE ? UNIT(s)->manager->idle_pipe : NULL, @@ -1834,7 +1807,7 @@ static int main_pid_good(Service *s) { /* If it's an alien child let's check if it is still * alive ... */ if (s->main_pid_alien && s->main_pid > 0) - return kill(s->main_pid, 0) >= 0 || errno != ESRCH; + return pid_is_alive(s->main_pid); /* .. otherwise assume we'll get a SIGCHLD for it, * which we really should wait for to collect exit @@ -1867,6 +1840,8 @@ static int cgroup_good(Service *s) { return !r; } +static int service_execute_action(Service *s, FailureAction action, const char *reason, bool log_action_none); + static void service_enter_dead(Service *s, ServiceResult f, bool allow_restart) { int r; assert(s); @@ -1876,6 +1851,9 @@ static void service_enter_dead(Service *s, ServiceResult f, bool allow_restart) service_set_state(s, s->result != SERVICE_SUCCESS ? SERVICE_FAILED : SERVICE_DEAD); + if (s->result != SERVICE_SUCCESS) + service_execute_action(s, s->failure_action, "failed", false); + if (allow_restart && !s->forbid_restart && (s->restart == SERVICE_RESTART_ALWAYS || @@ -1898,10 +1876,13 @@ static void service_enter_dead(Service *s, ServiceResult f, bool allow_restart) s->forbid_restart = false; - /* we want fresh tmpdirs in case service is started again immediately */ + /* We want fresh tmpdirs in case service is started again immediately */ exec_runtime_destroy(s->exec_runtime); s->exec_runtime = exec_runtime_unref(s->exec_runtime); + /* Also, remove the runtime directory in */ + exec_context_destroy_runtime_directory(&s->exec_context, manager_get_runtime_prefix(UNIT(s)->manager)); + /* Try to delete the pid file. At this point it will be * out-of-date, and some software might be confused by it, so * let's remove it. */ @@ -1925,6 +1906,7 @@ static void service_enter_stop_post(Service *s, ServiceResult f) { s->result = f; service_unwatch_control_pid(s); + unit_watch_all_pids(UNIT(s)); s->control_command = s->exec_command[SERVICE_EXEC_STOP_POST]; if (s->control_command) { @@ -1943,10 +1925,9 @@ static void service_enter_stop_post(Service *s, ServiceResult f) { if (r < 0) goto fail; - service_set_state(s, SERVICE_STOP_POST); } else - service_enter_dead(s, SERVICE_SUCCESS, true); + service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_SUCCESS); return; @@ -1965,6 +1946,8 @@ static void service_enter_signal(Service *s, ServiceState state, ServiceResult f if (f != SERVICE_SUCCESS) s->result = f; + unit_watch_all_pids(UNIT(s)); + r = unit_kill_context( UNIT(s), &s->kill_context, @@ -1972,6 +1955,7 @@ static void service_enter_signal(Service *s, ServiceState state, ServiceResult f s->main_pid, s->control_pid, s->main_pid_alien); + if (r < 0) goto fail; @@ -1983,8 +1967,12 @@ static void service_enter_signal(Service *s, ServiceState state, ServiceResult f } service_set_state(s, state); - } else if (state == SERVICE_STOP_SIGTERM || state == SERVICE_STOP_SIGKILL) + } else if (state == SERVICE_STOP_SIGTERM) + service_enter_signal(s, SERVICE_STOP_SIGKILL, SERVICE_SUCCESS); + else if (state == SERVICE_STOP_SIGKILL) service_enter_stop_post(s, SERVICE_SUCCESS); + else if (state == SERVICE_FINAL_SIGTERM) + service_enter_signal(s, SERVICE_FINAL_SIGKILL, SERVICE_SUCCESS); else service_enter_dead(s, SERVICE_SUCCESS, true); @@ -2009,6 +1997,7 @@ static void service_enter_stop(Service *s, ServiceResult f) { s->result = f; service_unwatch_control_pid(s); + unit_watch_all_pids(UNIT(s)); s->control_command = s->exec_command[SERVICE_EXEC_STOP]; if (s->control_command) { @@ -2387,26 +2376,27 @@ fail: service_enter_stop(s, SERVICE_FAILURE_RESOURCES); } -static int service_start_limit_test(Service *s) { +static int service_execute_action(Service *s, FailureAction action, const char *reason, bool log_action_none) { assert(s); - if (ratelimit_test(&s->start_limit)) - return 0; + if (action == SERVICE_FAILURE_ACTION_REBOOT || + action == SERVICE_FAILURE_ACTION_REBOOT_FORCE) + update_reboot_param_file(s->reboot_arg); - switch (s->start_limit_action) { + switch (action) { - case SERVICE_START_LIMIT_NONE: - log_warning_unit(UNIT(s)->id, - "%s start request repeated too quickly, refusing to start.", - UNIT(s)->id); + case SERVICE_FAILURE_ACTION_NONE: + if (log_action_none) + log_warning_unit(UNIT(s)->id, + "%s %s, refusing to start.", UNIT(s)->id, reason); break; - case SERVICE_START_LIMIT_REBOOT: { + case SERVICE_FAILURE_ACTION_REBOOT: { _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; int r; log_warning_unit(UNIT(s)->id, - "%s start request repeated too quickly, rebooting.", UNIT(s)->id); + "%s %s, rebooting.", UNIT(s)->id, reason); r = manager_add_job_by_name(UNIT(s)->manager, JOB_START, SPECIAL_REBOOT_TARGET, JOB_REPLACE, @@ -2418,28 +2408,44 @@ static int service_start_limit_test(Service *s) { break; } - case SERVICE_START_LIMIT_REBOOT_FORCE: + case SERVICE_FAILURE_ACTION_REBOOT_FORCE: log_warning_unit(UNIT(s)->id, - "%s start request repeated too quickly, forcibly rebooting.", UNIT(s)->id); + "%s %s, forcibly rebooting.", UNIT(s)->id, reason); UNIT(s)->manager->exit_code = MANAGER_REBOOT; break; - case SERVICE_START_LIMIT_REBOOT_IMMEDIATE: + case SERVICE_FAILURE_ACTION_REBOOT_IMMEDIATE: log_warning_unit(UNIT(s)->id, - "%s start request repeated too quickly, rebooting immediately.", UNIT(s)->id); + "%s %s, rebooting immediately.", UNIT(s)->id, reason); sync(); + if (s->reboot_arg) { + log_info("Rebooting with argument '%s'.", s->reboot_arg); + syscall(SYS_reboot, LINUX_REBOOT_MAGIC1, LINUX_REBOOT_MAGIC2, + LINUX_REBOOT_CMD_RESTART2, s->reboot_arg); + } + + log_info("Rebooting."); reboot(RB_AUTOBOOT); break; default: log_error_unit(UNIT(s)->id, - "start limit action=%i", s->start_limit_action); - assert_not_reached("Unknown StartLimitAction."); + "failure action=%i", action); + assert_not_reached("Unknown FailureAction."); } return -ECANCELED; } +static int service_start_limit_test(Service *s) { + assert(s); + + if (ratelimit_test(&s->start_limit)) + return 0; + + return service_execute_action(s, s->start_limit_action, "start request repeated too quickly", true); +} + static int service_start(Unit *u) { Service *s = SERVICE(u); int r; @@ -2562,11 +2568,11 @@ static int service_serialize(Unit *u, FILE *f, FDSet *fds) { unit_serialize_item(u, f, "reload-result", service_result_to_string(s->reload_result)); if (s->control_pid > 0) - unit_serialize_item_format(u, f, "control-pid", "%lu", - (unsigned long) s->control_pid); + unit_serialize_item_format(u, f, "control-pid", PID_FMT, + s->control_pid); if (s->main_pid_known && s->main_pid > 0) - unit_serialize_item_format(u, f, "main-pid", "%lu", (unsigned long) s->main_pid); + unit_serialize_item_format(u, f, "main-pid", PID_FMT, s->main_pid); unit_serialize_item(u, f, "main-pid-known", yes_no(s->main_pid_known)); @@ -2590,8 +2596,8 @@ static int service_serialize(Unit *u, FILE *f, FDSet *fds) { } if (s->main_exec_status.pid > 0) { - unit_serialize_item_format(u, f, "main-exec-status-pid", "%lu", - (unsigned long) s->main_exec_status.pid); + unit_serialize_item_format(u, f, "main-exec-status-pid", PID_FMT, + s->main_exec_status.pid); dual_timestamp_serialize(f, "main-exec-status-start", &s->main_exec_status.start_timestamp); dual_timestamp_serialize(f, "main-exec-status-exit", @@ -2699,8 +2705,7 @@ static int service_deserialize_item(Unit *u, const char *key, const char *value, log_debug_unit(u->id, "Failed to parse socket-fd value %s", value); else { - if (s->socket_fd >= 0) - close_nointr_nofail(s->socket_fd); + safe_close(s->socket_fd); s->socket_fd = fdset_remove(fds, fd); } } else if (streq(key, "main-exec-status-pid")) { @@ -2891,6 +2896,62 @@ fail: return 0; } +static void service_notify_cgroup_empty_event(Unit *u) { + Service *s = SERVICE(u); + + assert(u); + + log_debug_unit(u->id, "%s: cgroup is empty", u->id); + + switch (s->state) { + + /* Waiting for SIGCHLD is usually more interesting, + * because it includes return codes/signals. Which is + * why we ignore the cgroup events for most cases, + * except when we don't know pid which to expect the + * SIGCHLD for. */ + + case SERVICE_START: + case SERVICE_START_POST: + /* If we were hoping for the daemon to write its PID file, + * we can give up now. */ + if (s->pid_file_pathspec) { + log_warning_unit(u->id, + "%s never wrote its PID file. Failing.", UNIT(s)->id); + service_unwatch_pid_file(s); + if (s->state == SERVICE_START) + service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_FAILURE_RESOURCES); + else + service_enter_stop(s, SERVICE_FAILURE_RESOURCES); + } + break; + + case SERVICE_RUNNING: + /* service_enter_running() will figure out what to do */ + service_enter_running(s, SERVICE_SUCCESS); + break; + + case SERVICE_STOP_SIGTERM: + case SERVICE_STOP_SIGKILL: + + if (main_pid_good(s) <= 0 && !control_pid_good(s)) + service_enter_stop_post(s, SERVICE_SUCCESS); + + break; + + case SERVICE_STOP_POST: + case SERVICE_FINAL_SIGTERM: + case SERVICE_FINAL_SIGKILL: + if (main_pid_good(s) <= 0 && !control_pid_good(s)) + service_enter_dead(s, SERVICE_SUCCESS, true); + + break; + + default: + ; + } +} + static void service_sigchld_event(Unit *u, pid_t pid, int code, int status) { Service *s = SERVICE(u); ServiceResult f; @@ -3168,6 +3229,18 @@ static void service_sigchld_event(Unit *u, pid_t pid, int code, int status) { /* Notify clients about changed exit status */ unit_add_to_dbus_queue(u); + + /* We got one SIGCHLD for the service, let's watch all + * processes that are now running of the service, and watch + * that. Among the PIDs we then watch will be children + * reassigned to us, which hopefully allows us to identify + * when all children are gone */ + unit_tidy_watch_pids(u, s->main_pid, s->control_pid); + unit_watch_all_pids(u); + + /* If the PID set is empty now, then let's finish this off */ + if (set_isempty(u->pids)) + service_notify_cgroup_empty_event(u); } static int service_dispatch_timer(sd_event_source *source, usec_t usec, void *userdata) { @@ -3181,19 +3254,21 @@ static int service_dispatch_timer(sd_event_source *source, usec_t usec, void *us case SERVICE_START_PRE: case SERVICE_START: log_warning_unit(UNIT(s)->id, - "%s operation timed out. Terminating.", UNIT(s)->id); + "%s %s operation timed out. Terminating.", + UNIT(s)->id, + s->state == SERVICE_START ? "start" : "start-pre"); service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_FAILURE_TIMEOUT); break; case SERVICE_START_POST: log_warning_unit(UNIT(s)->id, - "%s operation timed out. Stopping.", UNIT(s)->id); + "%s start-post operation timed out. Stopping.", UNIT(s)->id); service_enter_stop(s, SERVICE_FAILURE_TIMEOUT); break; case SERVICE_RELOAD: log_warning_unit(UNIT(s)->id, - "%s operation timed out. Stopping.", UNIT(s)->id); + "%s reload operation timed out. Stopping.", UNIT(s)->id); s->reload_result = SERVICE_FAILURE_TIMEOUT; service_enter_running(s, SERVICE_SUCCESS); break; @@ -3207,11 +3282,11 @@ static int service_dispatch_timer(sd_event_source *source, usec_t usec, void *us case SERVICE_STOP_SIGTERM: if (s->kill_context.send_sigkill) { log_warning_unit(UNIT(s)->id, - "%s stopping timed out. Killing.", UNIT(s)->id); + "%s stop-sigterm timed out. Killing.", UNIT(s)->id); service_enter_signal(s, SERVICE_STOP_SIGKILL, SERVICE_FAILURE_TIMEOUT); } else { log_warning_unit(UNIT(s)->id, - "%s stopping timed out. Skipping SIGKILL.", UNIT(s)->id); + "%s stop-sigterm timed out. Skipping SIGKILL.", UNIT(s)->id); service_enter_stop_post(s, SERVICE_FAILURE_TIMEOUT); } @@ -3229,18 +3304,18 @@ static int service_dispatch_timer(sd_event_source *source, usec_t usec, void *us case SERVICE_STOP_POST: log_warning_unit(UNIT(s)->id, - "%s stopping timed out (2). Terminating.", UNIT(s)->id); + "%s stop-post timed out. Terminating.", UNIT(s)->id); service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_FAILURE_TIMEOUT); break; case SERVICE_FINAL_SIGTERM: if (s->kill_context.send_sigkill) { log_warning_unit(UNIT(s)->id, - "%s stopping timed out (2). Killing.", UNIT(s)->id); + "%s stop-final-sigterm timed out. Killing.", UNIT(s)->id); service_enter_signal(s, SERVICE_FINAL_SIGKILL, SERVICE_FAILURE_TIMEOUT); } else { log_warning_unit(UNIT(s)->id, - "%s stopping timed out (2). Skipping SIGKILL. Entering failed mode.", + "%s stop-final-sigterm timed out. Skipping SIGKILL. Entering failed mode.", UNIT(s)->id); service_enter_dead(s, SERVICE_FAILURE_TIMEOUT, false); } @@ -3249,13 +3324,16 @@ static int service_dispatch_timer(sd_event_source *source, usec_t usec, void *us case SERVICE_FINAL_SIGKILL: log_warning_unit(UNIT(s)->id, - "%s still around after SIGKILL (2). Entering failed mode.", UNIT(s)->id); + "%s still around after final SIGKILL. Entering failed mode.", UNIT(s)->id); service_enter_dead(s, SERVICE_FAILURE_TIMEOUT, true); break; case SERVICE_AUTO_RESTART: log_info_unit(UNIT(s)->id, - "%s holdoff time over, scheduling restart.", UNIT(s)->id); + s->restart_usec > 0 ? + "%s holdoff time over, scheduling restart." : + "%s has no holdoff time, scheduling restart.", + UNIT(s)->id); service_enter_restart(s); break; @@ -3278,62 +3356,6 @@ static int service_dispatch_watchdog(sd_event_source *source, usec_t usec, void return 0; } -static void service_notify_cgroup_empty_event(Unit *u) { - Service *s = SERVICE(u); - - assert(u); - - log_debug_unit(u->id, "%s: cgroup is empty", u->id); - - switch (s->state) { - - /* Waiting for SIGCHLD is usually more interesting, - * because it includes return codes/signals. Which is - * why we ignore the cgroup events for most cases, - * except when we don't know pid which to expect the - * SIGCHLD for. */ - - case SERVICE_START: - case SERVICE_START_POST: - /* If we were hoping for the daemon to write its PID file, - * we can give up now. */ - if (s->pid_file_pathspec) { - log_warning_unit(u->id, - "%s never wrote its PID file. Failing.", UNIT(s)->id); - service_unwatch_pid_file(s); - if (s->state == SERVICE_START) - service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_FAILURE_RESOURCES); - else - service_enter_stop(s, SERVICE_FAILURE_RESOURCES); - } - break; - - case SERVICE_RUNNING: - /* service_enter_running() will figure out what to do */ - service_enter_running(s, SERVICE_SUCCESS); - break; - - case SERVICE_STOP_SIGTERM: - case SERVICE_STOP_SIGKILL: - - if (main_pid_good(s) <= 0 && !control_pid_good(s)) - service_enter_stop_post(s, SERVICE_SUCCESS); - - break; - - case SERVICE_STOP_POST: - case SERVICE_FINAL_SIGTERM: - case SERVICE_FINAL_SIGKILL: - if (main_pid_good(s) <= 0 && !control_pid_good(s)) - service_enter_dead(s, SERVICE_SUCCESS, true); - - break; - - default: - ; - } -} - static void service_notify_message(Unit *u, pid_t pid, char **tags) { Service *s = SERVICE(u); const char *e; @@ -3341,22 +3363,25 @@ static void service_notify_message(Unit *u, pid_t pid, char **tags) { assert(u); + log_debug_unit(u->id, "%s: Got notification message from PID "PID_FMT" (%s...)", + u->id, pid, tags && *tags ? tags[0] : "(empty)"); + if (s->notify_access == NOTIFY_NONE) { log_warning_unit(u->id, - "%s: Got notification message from PID %lu, but reception is disabled.", - u->id, (unsigned long) pid); + "%s: Got notification message from PID "PID_FMT", but reception is disabled.", + u->id, pid); return; } if (s->notify_access == NOTIFY_MAIN && pid != s->main_pid) { - log_warning_unit(u->id, - "%s: Got notification message from PID %lu, but reception only permitted for PID %lu", - u->id, (unsigned long) pid, (unsigned long) s->main_pid); + + if (s->main_pid != 0) + log_warning_unit(u->id, "%s: Got notification message from PID "PID_FMT", but reception only permitted for main PID "PID_FMT, u->id, pid, s->main_pid); + else + log_debug_unit(u->id, "%s: Got notification message from PID "PID_FMT", but reception only permitted for main PID which is currently not known", u->id, pid); return; } - log_debug_unit(u->id, "%s: Got message", u->id); - /* Interpret MAINPID= */ if ((e = strv_find_prefix(tags, "MAINPID=")) && (s->state == SERVICE_START || @@ -3411,7 +3436,7 @@ static void service_notify_message(Unit *u, pid_t pid, char **tags) { free(t); } - /* Interpet WATCHDOG= */ + /* Interpret WATCHDOG= */ if (strv_find(tags, "WATCHDOG=1")) { log_debug_unit(u->id, "%s: got WATCHDOG=1", u->id); service_reset_watchdog(s); @@ -3422,6 +3447,20 @@ static void service_notify_message(Unit *u, pid_t pid, char **tags) { unit_add_to_dbus_queue(u); } +static int service_get_timeout(Unit *u, uint64_t *timeout) { + Service *s = SERVICE(u); + int r; + + if (!s->timer_event_source) + return 0; + + r = sd_event_source_get_time(s->timer_event_source, timeout); + if (r < 0) + return r; + + return 1; +} + #ifdef HAVE_SYSV_COMPAT static int service_enumerate(Manager *m) { @@ -3656,6 +3695,8 @@ static void service_bus_name_owner_change( } int service_set_socket_fd(Service *s, int fd, Socket *sock) { + _cleanup_free_ char *peer = NULL; + int r; assert(s); assert(fd >= 0); @@ -3673,6 +3714,23 @@ int service_set_socket_fd(Service *s, int fd, Socket *sock) { if (s->state != SERVICE_DEAD) return -EAGAIN; + if (getpeername_pretty(fd, &peer) >= 0) { + + if (UNIT(s)->description) { + _cleanup_free_ char *a; + + a = strjoin(UNIT(s)->description, " (", peer, ")", NULL); + if (!a) + return -ENOMEM; + + r = unit_set_description(UNIT(s), a); + } else + r = unit_set_description(UNIT(s), peer); + + if (r < 0) + return r; + } + s->socket_fd = fd; unit_ref_set(&s->accept_socket, UNIT(sock)); @@ -3774,13 +3832,13 @@ static const char* const service_result_table[_SERVICE_RESULT_MAX] = { DEFINE_STRING_TABLE_LOOKUP(service_result, ServiceResult); -static const char* const start_limit_action_table[_SERVICE_START_LIMIT_MAX] = { - [SERVICE_START_LIMIT_NONE] = "none", - [SERVICE_START_LIMIT_REBOOT] = "reboot", - [SERVICE_START_LIMIT_REBOOT_FORCE] = "reboot-force", - [SERVICE_START_LIMIT_REBOOT_IMMEDIATE] = "reboot-immediate" +static const char* const failure_action_table[_SERVICE_FAILURE_ACTION_MAX] = { + [SERVICE_FAILURE_ACTION_NONE] = "none", + [SERVICE_FAILURE_ACTION_REBOOT] = "reboot", + [SERVICE_FAILURE_ACTION_REBOOT_FORCE] = "reboot-force", + [SERVICE_FAILURE_ACTION_REBOOT_IMMEDIATE] = "reboot-immediate" }; -DEFINE_STRING_TABLE_LOOKUP(start_limit_action, StartLimitAction); +DEFINE_STRING_TABLE_LOOKUP(failure_action, FailureAction); const UnitVTable service_vtable = { .object_size = sizeof(Service), @@ -3834,6 +3892,8 @@ const UnitVTable service_vtable = { .bus_set_property = bus_service_set_property, .bus_commit_properties = bus_service_commit_properties, + .get_timeout = service_get_timeout, + #ifdef HAVE_SYSV_COMPAT .enumerate = service_enumerate, #endif