1 /*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
4 This file is part of systemd.
6 Copyright 2010 Lennart Poettering
8 systemd is free software; you can redistribute it and/or modify it
9 under the terms of the GNU Lesser General Public License as published by
10 the Free Software Foundation; either version 2.1 of the License, or
11 (at your option) any later version.
13 systemd is distributed in the hope that it will be useful, but
14 WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 Lesser General Public License for more details.
18 You should have received a copy of the GNU Lesser General Public License
19 along with systemd; If not, see <http://www.gnu.org/licenses/>.
30 #include <sys/prctl.h>
31 #include <sys/mount.h>
33 #ifdef HAVE_VALGRIND_VALGRIND_H
34 #include <valgrind/valgrind.h>
40 #include "sd-daemon.h"
45 #include "conf-parser.h"
52 #include "architecture.h"
54 #include "switch-root.h"
55 #include "capability.h"
58 #include "clock-util.h"
60 #include "bus-error.h"
62 #include "selinux-util.h"
64 #include "dbus-manager.h"
65 #include "load-fragment.h"
67 #include "mount-setup.h"
68 #include "loopback-setup.h"
69 #include "hostname-setup.h"
70 #include "machine-id-setup.h"
71 #include "selinux-setup.h"
72 #include "ima-setup.h"
73 #include "smack-setup.h"
74 #include "kmod-setup.h"
81 ACTION_DUMP_CONFIGURATION_ITEMS,
83 } arg_action = ACTION_RUN;
84 static char *arg_default_unit = NULL;
85 static SystemdRunningAs arg_running_as = _SYSTEMD_RUNNING_AS_INVALID;
86 static bool arg_dump_core = true;
87 static bool arg_crash_shell = false;
88 static int arg_crash_chvt = -1;
89 static bool arg_confirm_spawn = false;
90 static ShowStatus arg_show_status = _SHOW_STATUS_UNSET;
91 static bool arg_switched_root = false;
92 static int arg_no_pager = -1;
93 static char ***arg_join_controllers = NULL;
94 static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
95 static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
96 static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
97 static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
98 static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
99 static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
100 static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
101 static usec_t arg_runtime_watchdog = 0;
102 static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
103 static char **arg_default_environment = NULL;
104 static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {};
105 static uint64_t arg_capability_bounding_set_drop = 0;
106 static nsec_t arg_timer_slack_nsec = NSEC_INFINITY;
107 static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
108 static Set* arg_syscall_archs = NULL;
109 static FILE* arg_serialization = NULL;
110 static bool arg_default_cpu_accounting = false;
111 static bool arg_default_blockio_accounting = false;
112 static bool arg_default_memory_accounting = false;
114 static void nop_handler(int sig) {}
116 static void pager_open_if_enabled(void) {
118 if (arg_no_pager <= 0)
124 noreturn static void crash(int sig) {
127 /* Pass this on immediately, if this is not PID 1 */
129 else if (!arg_dump_core)
130 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
132 struct sigaction sa = {
133 .sa_handler = nop_handler,
134 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
138 /* We want to wait for the core process, hence let's enable SIGCHLD */
139 sigaction(SIGCHLD, &sa, NULL);
141 pid = raw_clone(SIGCHLD, NULL);
143 log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
146 struct rlimit rl = {};
148 /* Enable default signal handler for core dump */
150 sa.sa_handler = SIG_DFL;
151 sigaction(sig, &sa, NULL);
153 /* Don't limit the core dump size */
154 rl.rlim_cur = RLIM_INFINITY;
155 rl.rlim_max = RLIM_INFINITY;
156 setrlimit(RLIMIT_CORE, &rl);
158 /* Just to be sure... */
161 /* Raise the signal again */
163 kill(pid, sig); /* raise() would kill the parent */
165 assert_not_reached("We shouldn't be here...");
171 /* Order things nicely. */
172 r = wait_for_terminate(pid, &status);
174 log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
175 else if (status.si_code != CLD_DUMPED)
176 log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
177 signal_to_string(sig),
178 pid, sigchld_code_to_string(status.si_code),
180 strna(status.si_code == CLD_EXITED
181 ? exit_status_to_string(status.si_status, EXIT_STATUS_FULL)
182 : signal_to_string(status.si_status)));
184 log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
189 chvt(arg_crash_chvt);
191 if (arg_crash_shell) {
192 struct sigaction sa = {
193 .sa_handler = SIG_IGN,
194 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
198 log_info("Executing crash shell in 10s...");
201 /* Let the kernel reap children for us */
202 assert_se(sigaction(SIGCHLD, &sa, NULL) == 0);
204 pid = raw_clone(SIGCHLD, NULL);
206 log_emergency_errno(errno, "Failed to fork off crash shell: %m");
208 make_console_stdio();
209 execle("/bin/sh", "/bin/sh", NULL, environ);
211 log_emergency_errno(errno, "execle() failed: %m");
214 log_info("Successfully spawned crash shell as PID "PID_FMT".", pid);
217 log_emergency("Freezing execution.");
221 static void install_crash_handler(void) {
222 static const struct sigaction sa = {
224 .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
228 /* We ignore the return value here, since, we don't mind if we
229 * cannot set up a crash handler */
230 r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
232 log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
235 static int console_setup(void) {
236 _cleanup_close_ int tty_fd = -1;
239 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
241 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
243 /* We don't want to force text mode. plymouth may be showing
244 * pictures already from initrd. */
245 r = reset_terminal_fd(tty_fd, false);
247 return log_error_errno(r, "Failed to reset /dev/console: %m");
252 static int set_default_unit(const char *u) {
261 free(arg_default_unit);
262 arg_default_unit = c;
267 static int parse_proc_cmdline_item(const char *key, const char *value) {
269 static const char * const rlmap[] = {
270 "emergency", SPECIAL_EMERGENCY_TARGET,
271 "-b", SPECIAL_EMERGENCY_TARGET,
272 "rescue", SPECIAL_RESCUE_TARGET,
273 "single", SPECIAL_RESCUE_TARGET,
274 "-s", SPECIAL_RESCUE_TARGET,
275 "s", SPECIAL_RESCUE_TARGET,
276 "S", SPECIAL_RESCUE_TARGET,
277 "1", SPECIAL_RESCUE_TARGET,
278 "2", SPECIAL_MULTI_USER_TARGET,
279 "3", SPECIAL_MULTI_USER_TARGET,
280 "4", SPECIAL_MULTI_USER_TARGET,
281 "5", SPECIAL_GRAPHICAL_TARGET,
287 if (streq(key, "systemd.unit") && value) {
290 return set_default_unit(value);
292 } else if (streq(key, "rd.systemd.unit") && value) {
295 return set_default_unit(value);
297 } else if (streq(key, "systemd.dump_core") && value) {
299 r = parse_boolean(value);
301 log_warning("Failed to parse dump core switch %s. Ignoring.", value);
305 } else if (streq(key, "systemd.crash_shell") && value) {
307 r = parse_boolean(value);
309 log_warning("Failed to parse crash shell switch %s. Ignoring.", value);
313 } else if (streq(key, "systemd.crash_chvt") && value) {
315 if (safe_atoi(value, &r) < 0)
316 log_warning("Failed to parse crash chvt switch %s. Ignoring.", value);
320 } else if (streq(key, "systemd.confirm_spawn") && value) {
322 r = parse_boolean(value);
324 log_warning("Failed to parse confirm spawn switch %s. Ignoring.", value);
326 arg_confirm_spawn = r;
328 } else if (streq(key, "systemd.show_status") && value) {
330 r = parse_show_status(value, &arg_show_status);
332 log_warning("Failed to parse show status switch %s. Ignoring.", value);
334 } else if (streq(key, "systemd.default_standard_output") && value) {
336 r = exec_output_from_string(value);
338 log_warning("Failed to parse default standard output switch %s. Ignoring.", value);
340 arg_default_std_output = r;
342 } else if (streq(key, "systemd.default_standard_error") && value) {
344 r = exec_output_from_string(value);
346 log_warning("Failed to parse default standard error switch %s. Ignoring.", value);
348 arg_default_std_error = r;
350 } else if (streq(key, "systemd.setenv") && value) {
352 if (env_assignment_is_valid(value)) {
355 env = strv_env_set(arg_default_environment, value);
357 arg_default_environment = env;
359 log_warning_errno(ENOMEM, "Setting environment variable '%s' failed, ignoring: %m", value);
361 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
363 } else if (streq(key, "quiet") && !value) {
365 if (arg_show_status == _SHOW_STATUS_UNSET)
366 arg_show_status = SHOW_STATUS_AUTO;
368 } else if (streq(key, "debug") && !value) {
370 /* Note that log_parse_environment() handles 'debug'
371 * too, and sets the log level to LOG_DEBUG. */
373 if (detect_container(NULL) > 0)
374 log_set_target(LOG_TARGET_CONSOLE);
376 } else if (!in_initrd() && !value) {
379 /* SysV compatibility */
380 for (i = 0; i < ELEMENTSOF(rlmap); i += 2)
381 if (streq(key, rlmap[i]))
382 return set_default_unit(rlmap[i+1]);
388 #define DEFINE_SETTER(name, func, descr) \
389 static int name(const char *unit, \
390 const char *filename, \
392 const char *section, \
393 unsigned section_line, \
394 const char *lvalue, \
396 const char *rvalue, \
408 log_syntax(unit, LOG_ERR, filename, line, -r, \
409 "Invalid " descr "'%s': %s", \
410 rvalue, strerror(-r)); \
415 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level")
416 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target")
417 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" )
418 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location")
420 static int config_parse_cpu_affinity2(
422 const char *filename,
425 unsigned section_line,
432 const char *word, *state;
441 FOREACH_WORD_QUOTED(word, l, rvalue, state) {
446 if (!(t = strndup(word, l)))
449 r = safe_atou(t, &cpu);
453 if (!(c = cpu_set_malloc(&ncpus)))
456 if (r < 0 || cpu >= ncpus) {
457 log_syntax(unit, LOG_ERR, filename, line, -r,
458 "Failed to parse CPU affinity '%s'", rvalue);
463 CPU_SET_S(cpu, CPU_ALLOC_SIZE(ncpus), c);
466 log_syntax(unit, LOG_ERR, filename, line, EINVAL,
467 "Trailing garbage, ignoring.");
470 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
471 log_unit_warning(unit, "Failed to set CPU affinity: %m");
479 static int config_parse_show_status(
481 const char *filename,
484 unsigned section_line,
492 ShowStatus *b = data;
499 k = parse_show_status(rvalue, b);
501 log_syntax(unit, LOG_ERR, filename, line, -k,
502 "Failed to parse show status setting, ignoring: %s", rvalue);
509 static void strv_free_free(char ***l) {
521 static void free_join_controllers(void) {
522 strv_free_free(arg_join_controllers);
523 arg_join_controllers = NULL;
526 static int config_parse_join_controllers(const char *unit,
527 const char *filename,
530 unsigned section_line,
538 const char *word, *state;
545 free_join_controllers();
547 FOREACH_WORD_QUOTED(word, length, rvalue, state) {
550 s = strndup(word, length);
554 l = strv_split(s, ",");
559 if (strv_length(l) <= 1) {
564 if (!arg_join_controllers) {
565 arg_join_controllers = new(char**, 2);
566 if (!arg_join_controllers) {
571 arg_join_controllers[0] = l;
572 arg_join_controllers[1] = NULL;
579 t = new0(char**, n+2);
587 for (a = arg_join_controllers; *a; a++) {
589 if (strv_overlap(*a, l)) {
590 if (strv_extend_strv(&l, *a) < 0) {
610 t[n++] = strv_uniq(l);
612 strv_free_free(arg_join_controllers);
613 arg_join_controllers = t;
617 log_syntax(unit, LOG_ERR, filename, line, EINVAL,
618 "Trailing garbage, ignoring.");
623 static int parse_config_file(void) {
625 const ConfigTableItem items[] = {
626 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
627 { "Manager", "LogTarget", config_parse_target, 0, NULL },
628 { "Manager", "LogColor", config_parse_color, 0, NULL },
629 { "Manager", "LogLocation", config_parse_location, 0, NULL },
630 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
631 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
632 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
633 { "Manager", "CrashChVT", config_parse_int, 0, &arg_crash_chvt },
634 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
635 { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
636 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
637 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
638 { "Manager", "CapabilityBoundingSet", config_parse_bounding_set, 0, &arg_capability_bounding_set_drop },
640 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
642 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
643 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
644 { "Manager", "DefaultStandardOutput", config_parse_output, 0, &arg_default_std_output },
645 { "Manager", "DefaultStandardError", config_parse_output, 0, &arg_default_std_error },
646 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
647 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
648 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
649 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval },
650 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
651 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
652 { "Manager", "DefaultLimitCPU", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CPU] },
653 { "Manager", "DefaultLimitFSIZE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_FSIZE] },
654 { "Manager", "DefaultLimitDATA", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_DATA] },
655 { "Manager", "DefaultLimitSTACK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_STACK] },
656 { "Manager", "DefaultLimitCORE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CORE] },
657 { "Manager", "DefaultLimitRSS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RSS] },
658 { "Manager", "DefaultLimitNOFILE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NOFILE] },
659 { "Manager", "DefaultLimitAS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_AS] },
660 { "Manager", "DefaultLimitNPROC", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NPROC] },
661 { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MEMLOCK] },
662 { "Manager", "DefaultLimitLOCKS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_LOCKS] },
663 { "Manager", "DefaultLimitSIGPENDING", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_SIGPENDING] },
664 { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MSGQUEUE] },
665 { "Manager", "DefaultLimitNICE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NICE] },
666 { "Manager", "DefaultLimitRTPRIO", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTPRIO] },
667 { "Manager", "DefaultLimitRTTIME", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTTIME] },
668 { "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_default_cpu_accounting },
669 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
670 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
674 const char *fn, *conf_dirs_nulstr;
676 fn = arg_running_as == SYSTEMD_SYSTEM ? PKGSYSCONFDIR "/system.conf" : PKGSYSCONFDIR "/user.conf";
677 conf_dirs_nulstr = arg_running_as == SYSTEMD_SYSTEM ? CONF_DIRS_NULSTR("systemd/system.conf") : CONF_DIRS_NULSTR("systemd/user.conf");
678 config_parse_many(fn, conf_dirs_nulstr, "Manager\0",
679 config_item_table_lookup, items, false, NULL);
684 static int parse_argv(int argc, char *argv[]) {
687 ARG_LOG_LEVEL = 0x100,
697 ARG_DUMP_CONFIGURATION_ITEMS,
704 ARG_DEFAULT_STD_OUTPUT,
705 ARG_DEFAULT_STD_ERROR
708 static const struct option options[] = {
709 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
710 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
711 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
712 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
713 { "unit", required_argument, NULL, ARG_UNIT },
714 { "system", no_argument, NULL, ARG_SYSTEM },
715 { "user", no_argument, NULL, ARG_USER },
716 { "test", no_argument, NULL, ARG_TEST },
717 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
718 { "help", no_argument, NULL, 'h' },
719 { "version", no_argument, NULL, ARG_VERSION },
720 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
721 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
722 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
723 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
724 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
725 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
726 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
727 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
728 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
740 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
745 r = log_set_max_level_from_string(optarg);
747 log_error("Failed to parse log level %s.", optarg);
754 r = log_set_target_from_string(optarg);
756 log_error("Failed to parse log target %s.", optarg);
765 r = log_show_color_from_string(optarg);
767 log_error("Failed to parse log color setting %s.", optarg);
771 log_show_color(true);
775 case ARG_LOG_LOCATION:
777 r = log_show_location_from_string(optarg);
779 log_error("Failed to parse log location setting %s.", optarg);
783 log_show_location(true);
787 case ARG_DEFAULT_STD_OUTPUT:
788 r = exec_output_from_string(optarg);
790 log_error("Failed to parse default standard output setting %s.", optarg);
793 arg_default_std_output = r;
796 case ARG_DEFAULT_STD_ERROR:
797 r = exec_output_from_string(optarg);
799 log_error("Failed to parse default standard error output setting %s.", optarg);
802 arg_default_std_error = r;
807 r = set_default_unit(optarg);
809 return log_error_errno(r, "Failed to set default unit %s: %m", optarg);
814 arg_running_as = SYSTEMD_SYSTEM;
818 arg_running_as = SYSTEMD_USER;
822 arg_action = ACTION_TEST;
823 if (arg_no_pager < 0)
832 arg_action = ACTION_VERSION;
835 case ARG_DUMP_CONFIGURATION_ITEMS:
836 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
840 r = optarg ? parse_boolean(optarg) : 1;
842 log_error("Failed to parse dump core boolean %s.", optarg);
848 case ARG_CRASH_SHELL:
849 r = optarg ? parse_boolean(optarg) : 1;
851 log_error("Failed to parse crash shell boolean %s.", optarg);
857 case ARG_CONFIRM_SPAWN:
858 r = optarg ? parse_boolean(optarg) : 1;
860 log_error("Failed to parse confirm spawn boolean %s.", optarg);
863 arg_confirm_spawn = r;
866 case ARG_SHOW_STATUS:
868 r = parse_show_status(optarg, &arg_show_status);
870 log_error("Failed to parse show status boolean %s.", optarg);
874 arg_show_status = SHOW_STATUS_YES;
877 case ARG_DESERIALIZE: {
881 r = safe_atoi(optarg, &fd);
882 if (r < 0 || fd < 0) {
883 log_error("Failed to parse deserialize option %s.", optarg);
884 return r < 0 ? r : -EINVAL;
887 fd_cloexec(fd, true);
891 return log_error_errno(errno, "Failed to open serialization fd: %m");
893 if (arg_serialization)
894 fclose(arg_serialization);
896 arg_serialization = f;
901 case ARG_SWITCHED_ROOT:
902 arg_switched_root = true;
906 arg_action = ACTION_HELP;
907 if (arg_no_pager < 0)
912 log_set_max_level(LOG_DEBUG);
918 /* Just to eat away the sysvinit kernel
919 * cmdline args without getopt() error
920 * messages that we'll parse in
921 * parse_proc_cmdline_word() or ignore. */
930 assert_not_reached("Unhandled option code.");
933 if (optind < argc && getpid() != 1) {
934 /* Hmm, when we aren't run as init system
935 * let's complain about excess arguments */
937 log_error("Excess arguments.");
944 static int help(void) {
946 printf("%s [OPTIONS...]\n\n"
947 "Starts up and maintains the system or user services.\n\n"
948 " -h --help Show this help\n"
949 " --test Determine startup sequence, dump it and exit\n"
950 " --no-pager Do not pipe output into a pager\n"
951 " --dump-configuration-items Dump understood unit configuration items\n"
952 " --unit=UNIT Set default unit\n"
953 " --system Run a system instance, even if PID != 1\n"
954 " --user Run a user instance\n"
955 " --dump-core[=0|1] Dump core on crash\n"
956 " --crash-shell[=0|1] Run shell on crash\n"
957 " --confirm-spawn[=0|1] Ask for confirmation when spawning processes\n"
958 " --show-status[=0|1] Show status updates on the console during bootup\n"
959 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
960 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
961 " --log-color[=0|1] Highlight important log messages\n"
962 " --log-location[=0|1] Include code location in log messages\n"
963 " --default-standard-output= Set default standard output for services\n"
964 " --default-standard-error= Set default standard error output for services\n",
965 program_invocation_short_name);
970 static int version(void) {
971 puts(PACKAGE_STRING);
972 puts(SYSTEMD_FEATURES);
977 static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
986 r = manager_open_serialization(m, &f);
988 log_error_errno(r, "Failed to create serialization file: %m");
992 /* Make sure nothing is really destructed when we shut down */
994 bus_manager_send_reloading(m, true);
999 log_error_errno(r, "Failed to allocate fd set: %m");
1003 r = manager_serialize(m, f, fds, switching_root);
1005 log_error_errno(r, "Failed to serialize state: %m");
1009 if (fseeko(f, 0, SEEK_SET) < 0) {
1010 log_error_errno(errno, "Failed to rewind serialization fd: %m");
1014 r = fd_cloexec(fileno(f), false);
1016 log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1020 r = fdset_cloexec(fds, false);
1022 log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1040 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1044 assert(saved_rlimit);
1046 /* Save the original RLIMIT_NOFILE so that we can reset it
1047 * later when transitioning from the initrd to the main
1048 * systemd or suchlike. */
1049 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1050 return log_error_errno(errno, "Reading RLIMIT_NOFILE failed: %m");
1052 /* Make sure forked processes get the default kernel setting */
1053 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1056 rl = newdup(struct rlimit, saved_rlimit, 1);
1060 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1063 /* Bump up the resource limit for ourselves substantially */
1064 nl.rlim_cur = nl.rlim_max = 64*1024;
1065 r = setrlimit_closest(RLIMIT_NOFILE, &nl);
1067 return log_error_errno(r, "Setting RLIMIT_NOFILE failed: %m");
1072 static void test_mtab(void) {
1074 static const char ok[] =
1075 "/proc/self/mounts\0"
1077 "../proc/self/mounts\0"
1080 _cleanup_free_ char *p = NULL;
1083 /* Check that /etc/mtab is a symlink to the right place or
1084 * non-existing. But certainly not a file, or a symlink to
1085 * some weird place... */
1087 r = readlink_malloc("/etc/mtab", &p);
1090 if (r >= 0 && nulstr_contains(ok, p))
1093 log_warning("/etc/mtab is not a symlink or not pointing to /proc/self/mounts. "
1094 "This is not supported anymore. "
1095 "Please make sure to replace this file by a symlink to avoid incorrect or misleading mount(8) output.");
1098 static void test_usr(void) {
1100 /* Check that /usr is not a separate fs */
1102 if (dir_is_empty("/usr") <= 0)
1105 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1106 "Some things will probably break (sometimes even silently) in mysterious ways. "
1107 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1110 static int initialize_join_controllers(void) {
1111 /* By default, mount "cpu" + "cpuacct" together, and "net_cls"
1112 * + "net_prio". We'd like to add "cpuset" to the mix, but
1113 * "cpuset" doesn't really work for groups with no initialized
1116 arg_join_controllers = new(char**, 3);
1117 if (!arg_join_controllers)
1120 arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL);
1121 arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL);
1122 arg_join_controllers[2] = NULL;
1124 if (!arg_join_controllers[0] || !arg_join_controllers[1]) {
1125 free_join_controllers();
1132 static int enforce_syscall_archs(Set *archs) {
1134 scmp_filter_ctx *seccomp;
1139 seccomp = seccomp_init(SCMP_ACT_ALLOW);
1143 SET_FOREACH(id, arg_syscall_archs, i) {
1144 r = seccomp_arch_add(seccomp, PTR_TO_UINT32(id) - 1);
1148 log_error_errno(r, "Failed to add architecture to seccomp: %m");
1153 r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
1155 log_error_errno(r, "Failed to unset NO_NEW_PRIVS: %m");
1159 r = seccomp_load(seccomp);
1161 log_error_errno(r, "Failed to add install architecture seccomp: %m");
1164 seccomp_release(seccomp);
1171 static int status_welcome(void) {
1172 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1175 r = parse_env_file("/etc/os-release", NEWLINE,
1176 "PRETTY_NAME", &pretty_name,
1177 "ANSI_COLOR", &ansi_color,
1180 r = parse_env_file("/usr/lib/os-release", NEWLINE,
1181 "PRETTY_NAME", &pretty_name,
1182 "ANSI_COLOR", &ansi_color,
1186 if (r < 0 && r != -ENOENT)
1187 log_warning_errno(r, "Failed to read os-release file: %m");
1189 return status_printf(NULL, false, false,
1190 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1191 isempty(ansi_color) ? "1" : ansi_color,
1192 isempty(pretty_name) ? "Linux" : pretty_name);
1195 static int write_container_id(void) {
1198 c = getenv("container");
1202 return write_string_file("/run/systemd/container", c);
1205 int main(int argc, char *argv[]) {
1207 int r, retval = EXIT_FAILURE;
1208 usec_t before_startup, after_startup;
1209 char timespan[FORMAT_TIMESPAN_MAX];
1211 bool reexecute = false;
1212 const char *shutdown_verb = NULL;
1213 dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL;
1214 dual_timestamp userspace_timestamp = DUAL_TIMESTAMP_NULL;
1215 dual_timestamp kernel_timestamp = DUAL_TIMESTAMP_NULL;
1216 dual_timestamp security_start_timestamp = DUAL_TIMESTAMP_NULL;
1217 dual_timestamp security_finish_timestamp = DUAL_TIMESTAMP_NULL;
1218 static char systemd[] = "systemd";
1219 bool skip_setup = false;
1221 bool loaded_policy = false;
1222 bool arm_reboot_watchdog = false;
1223 bool queue_default_job = false;
1224 bool empty_etc = false;
1225 char *switch_root_dir = NULL, *switch_root_init = NULL;
1226 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0);
1227 const char *error_message = NULL;
1229 #ifdef HAVE_SYSV_COMPAT
1230 if (getpid() != 1 && strstr(program_invocation_short_name, "init")) {
1231 /* This is compatibility support for SysV, where
1232 * calling init as a user is identical to telinit. */
1235 execv(SYSTEMCTL_BINARY_PATH, argv);
1236 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1241 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
1242 dual_timestamp_get(&userspace_timestamp);
1244 /* Determine if this is a reexecution or normal bootup. We do
1245 * the full command line parsing much later, so let's just
1246 * have a quick peek here. */
1247 if (strv_find(argv+1, "--deserialize"))
1250 /* If we have switched root, do all the special setup
1252 if (strv_find(argv+1, "--switched-root"))
1255 /* If we get started via the /sbin/init symlink then we are
1256 called 'init'. After a subsequent reexecution we are then
1257 called 'systemd'. That is confusing, hence let's call us
1258 systemd right-away. */
1259 program_invocation_short_name = systemd;
1260 prctl(PR_SET_NAME, systemd);
1265 log_show_color(isatty(STDERR_FILENO) > 0);
1266 log_set_upgrade_syslog_to_journal(true);
1268 /* Disable the umask logic */
1272 if (getpid() == 1 && detect_container(NULL) <= 0) {
1274 /* Running outside of a container as PID 1 */
1275 arg_running_as = SYSTEMD_SYSTEM;
1277 log_set_target(LOG_TARGET_KMSG);
1281 initrd_timestamp = userspace_timestamp;
1284 mount_setup_early();
1285 dual_timestamp_get(&security_start_timestamp);
1286 if (mac_selinux_setup(&loaded_policy) < 0) {
1287 error_message = "Failed to load SELinux policy";
1289 } else if (ima_setup() < 0) {
1290 error_message = "Failed to load IMA policy";
1292 } else if (mac_smack_setup(&loaded_policy) < 0) {
1293 error_message = "Failed to load SMACK policy";
1296 dual_timestamp_get(&security_finish_timestamp);
1299 if (mac_selinux_init(NULL) < 0) {
1300 error_message = "Failed to initialize SELinux policy";
1305 if (clock_is_localtime() > 0) {
1309 * The very first call of settimeofday() also does a time warp in the kernel.
1311 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on
1312 * external tools to take care of maintaining the RTC and do all adjustments.
1313 * This matches the behavior of Windows, which leaves the RTC alone if the
1314 * registry tells that the RTC runs in UTC.
1316 r = clock_set_timezone(&min);
1318 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1320 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1321 } else if (!in_initrd()) {
1323 * Do a dummy very first call to seal the kernel's time warp magic.
1325 * Do not call this this from inside the initrd. The initrd might not
1326 * carry /etc/adjtime with LOCAL, but the real system could be set up
1327 * that way. In such case, we need to delay the time-warp or the sealing
1328 * until we reach the real system.
1330 * Do no set the kernel's timezone. The concept of local time cannot
1331 * be supported reliably, the time will jump or be incorrect at every daylight
1332 * saving time change. All kernel local time concepts will be treated
1335 clock_reset_timewarp();
1339 /* Set the default for later on, but don't actually
1340 * open the logs like this for now. Note that if we
1341 * are transitioning from the initrd there might still
1342 * be journal fd open, and we shouldn't attempt
1343 * opening that before we parsed /proc/cmdline which
1344 * might redirect output elsewhere. */
1345 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
1347 } else if (getpid() == 1) {
1348 /* Running inside a container, as PID 1 */
1349 arg_running_as = SYSTEMD_SYSTEM;
1350 log_set_target(LOG_TARGET_CONSOLE);
1351 log_close_console(); /* force reopen of /dev/console */
1354 /* For the later on, see above... */
1355 log_set_target(LOG_TARGET_JOURNAL);
1357 /* clear the kernel timestamp,
1358 * because we are in a container */
1359 kernel_timestamp.monotonic = 0ULL;
1360 kernel_timestamp.realtime = 0ULL;
1363 /* Running as user instance */
1364 arg_running_as = SYSTEMD_USER;
1365 log_set_target(LOG_TARGET_AUTO);
1368 /* clear the kernel timestamp,
1369 * because we are not PID 1 */
1370 kernel_timestamp.monotonic = 0ULL;
1371 kernel_timestamp.realtime = 0ULL;
1374 /* Initialize default unit */
1375 r = set_default_unit(SPECIAL_DEFAULT_TARGET);
1377 log_emergency_errno(r, "Failed to set default unit %s: %m", SPECIAL_DEFAULT_TARGET);
1378 error_message = "Failed to set default unit";
1382 r = initialize_join_controllers();
1384 error_message = "Failed to initalize cgroup controllers";
1388 /* Mount /proc, /sys and friends, so that /proc/cmdline and
1389 * /proc/$PID/fd is available. */
1390 if (getpid() == 1) {
1392 /* Load the kernel modules early, so that we kdbus.ko is loaded before kdbusfs shall be mounted */
1396 r = mount_setup(loaded_policy);
1398 error_message = "Failed to mount API filesystems";
1403 /* Reset all signal handlers. */
1404 assert_se(reset_all_signal_handlers() == 0);
1406 ignore_signals(SIGNALS_IGNORE, -1);
1408 if (parse_config_file() < 0) {
1409 error_message = "Failed to parse config file";
1413 if (arg_running_as == SYSTEMD_SYSTEM) {
1414 r = parse_proc_cmdline(parse_proc_cmdline_item);
1416 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
1419 /* Note that this also parses bits from the kernel command
1420 * line, including "debug". */
1421 log_parse_environment();
1423 if (parse_argv(argc, argv) < 0) {
1424 error_message = "Failed to parse commandline arguments";
1428 if (arg_action == ACTION_TEST &&
1430 log_error("Don't run test mode as root.");
1434 if (arg_running_as == SYSTEMD_USER &&
1435 arg_action == ACTION_RUN &&
1437 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
1441 if (arg_running_as == SYSTEMD_SYSTEM &&
1442 arg_action == ACTION_RUN &&
1443 running_in_chroot() > 0) {
1444 log_error("Cannot be run in a chroot() environment.");
1448 if (arg_action == ACTION_TEST)
1451 pager_open_if_enabled();
1453 if (arg_action == ACTION_HELP) {
1456 } else if (arg_action == ACTION_VERSION) {
1459 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
1460 unit_dump_config_items(stdout);
1461 retval = EXIT_SUCCESS;
1463 } else if (arg_action == ACTION_DONE) {
1464 retval = EXIT_SUCCESS;
1468 if (arg_running_as == SYSTEMD_USER &&
1469 !getenv("XDG_RUNTIME_DIR")) {
1470 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
1474 assert_se(arg_action == ACTION_RUN || arg_action == ACTION_TEST);
1476 /* Close logging fds, in order not to confuse fdset below */
1479 /* Remember open file descriptors for later deserialization */
1480 r = fdset_new_fill(&fds);
1482 log_emergency_errno(r, "Failed to allocate fd set: %m");
1483 error_message = "Failed to allocate fd set";
1486 fdset_cloexec(fds, true);
1488 if (arg_serialization)
1489 assert_se(fdset_remove(fds, fileno(arg_serialization)) >= 0);
1491 if (arg_running_as == SYSTEMD_SYSTEM)
1492 /* Become a session leader if we aren't one yet. */
1495 /* Move out of the way, so that we won't block unmounts */
1496 assert_se(chdir("/") == 0);
1498 /* Reset the console, but only if this is really init and we
1499 * are freshly booted */
1500 if (arg_running_as == SYSTEMD_SYSTEM && arg_action == ACTION_RUN) {
1502 /* If we are init, we connect stdin/stdout/stderr to
1503 * /dev/null and make sure we don't have a controlling
1507 if (getpid() == 1 && !skip_setup)
1511 /* Open the logging devices, if possible and necessary */
1514 if (arg_show_status == _SHOW_STATUS_UNSET)
1515 arg_show_status = SHOW_STATUS_YES;
1517 /* Make sure we leave a core dump without panicing the
1519 if (getpid() == 1) {
1520 install_crash_handler();
1522 r = mount_cgroup_controllers(arg_join_controllers);
1527 if (arg_running_as == SYSTEMD_SYSTEM) {
1528 const char *virtualization = NULL;
1530 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1531 arg_action == ACTION_TEST ? "test " : "" );
1533 detect_virtualization(&virtualization);
1535 log_info("Detected virtualization %s.", virtualization);
1537 write_container_id();
1539 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1542 log_info("Running in initial RAM disk.");
1544 /* Let's check whether /etc is already populated. We
1545 * don't actually really check for that, but use
1546 * /etc/machine-id as flag file. This allows container
1547 * managers and installers to provision a couple of
1548 * files already. If the container manager wants to
1549 * provision the machine ID itself it should pass
1550 * $container_uuid to PID 1. */
1552 empty_etc = access("/etc/machine-id", F_OK) < 0;
1554 log_info("Running with unpopulated /etc.");
1556 _cleanup_free_ char *t;
1558 t = uid_to_name(getuid());
1559 log_debug(PACKAGE_STRING " running in %suser mode for user "UID_FMT"/%s. (" SYSTEMD_FEATURES ")",
1560 arg_action == ACTION_TEST ? " test" : "", getuid(), t);
1563 if (arg_running_as == SYSTEMD_SYSTEM && !skip_setup) {
1564 if (arg_show_status > 0)
1568 machine_id_setup(NULL);
1575 if (arg_running_as == SYSTEMD_SYSTEM && arg_runtime_watchdog > 0)
1576 watchdog_set_timeout(&arg_runtime_watchdog);
1578 if (arg_timer_slack_nsec != NSEC_INFINITY)
1579 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1580 log_error_errno(errno, "Failed to adjust timer slack: %m");
1582 if (arg_capability_bounding_set_drop) {
1583 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set_drop);
1585 log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
1586 error_message = "Failed to drop capability bounding set of usermode helpers";
1589 r = capability_bounding_set_drop(arg_capability_bounding_set_drop, true);
1591 log_emergency_errno(r, "Failed to drop capability bounding set: %m");
1592 error_message = "Failed to drop capability bounding set";
1597 if (arg_syscall_archs) {
1598 r = enforce_syscall_archs(arg_syscall_archs);
1600 error_message = "Failed to set syscall architectures";
1605 if (arg_running_as == SYSTEMD_USER) {
1606 /* Become reaper of our children */
1607 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0) {
1608 log_warning_errno(errno, "Failed to make us a subreaper: %m");
1609 if (errno == EINVAL)
1610 log_info("Perhaps the kernel version is too old (< 3.4?)");
1614 if (arg_running_as == SYSTEMD_SYSTEM) {
1615 bump_rlimit_nofile(&saved_rlimit_nofile);
1618 r = unit_file_preset_all(UNIT_FILE_SYSTEM, false, NULL, UNIT_FILE_PRESET_FULL, false, NULL, 0);
1620 log_warning_errno(r, "Failed to populate /etc with preset unit settings, ignoring: %m");
1622 log_info("Populated /etc with preset unit settings.");
1626 r = manager_new(arg_running_as, arg_action == ACTION_TEST, &m);
1628 log_emergency_errno(r, "Failed to allocate manager object: %m");
1629 error_message = "Failed to allocate manager object";
1633 m->confirm_spawn = arg_confirm_spawn;
1634 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
1635 m->default_std_output = arg_default_std_output;
1636 m->default_std_error = arg_default_std_error;
1637 m->default_restart_usec = arg_default_restart_usec;
1638 m->default_timeout_start_usec = arg_default_timeout_start_usec;
1639 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
1640 m->default_start_limit_interval = arg_default_start_limit_interval;
1641 m->default_start_limit_burst = arg_default_start_limit_burst;
1642 m->default_cpu_accounting = arg_default_cpu_accounting;
1643 m->default_blockio_accounting = arg_default_blockio_accounting;
1644 m->default_memory_accounting = arg_default_memory_accounting;
1645 m->runtime_watchdog = arg_runtime_watchdog;
1646 m->shutdown_watchdog = arg_shutdown_watchdog;
1648 m->userspace_timestamp = userspace_timestamp;
1649 m->kernel_timestamp = kernel_timestamp;
1650 m->initrd_timestamp = initrd_timestamp;
1651 m->security_start_timestamp = security_start_timestamp;
1652 m->security_finish_timestamp = security_finish_timestamp;
1654 manager_set_default_rlimits(m, arg_default_rlimit);
1655 manager_environment_add(m, NULL, arg_default_environment);
1656 manager_set_show_status(m, arg_show_status);
1657 manager_set_first_boot(m, empty_etc);
1659 /* Remember whether we should queue the default job */
1660 queue_default_job = !arg_serialization || arg_switched_root;
1662 before_startup = now(CLOCK_MONOTONIC);
1664 r = manager_startup(m, arg_serialization, fds);
1666 log_error_errno(r, "Failed to fully start up daemon: %m");
1668 /* This will close all file descriptors that were opened, but
1669 * not claimed by any unit. */
1673 if (arg_serialization) {
1674 fclose(arg_serialization);
1675 arg_serialization = NULL;
1678 if (queue_default_job) {
1679 _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL;
1680 Unit *target = NULL;
1681 Job *default_unit_job;
1683 log_debug("Activating default unit: %s", arg_default_unit);
1685 r = manager_load_unit(m, arg_default_unit, NULL, &error, &target);
1687 log_error("Failed to load default target: %s", bus_error_message(&error, r));
1688 else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND)
1689 log_error_errno(target->load_error, "Failed to load default target: %m");
1690 else if (target->load_state == UNIT_MASKED)
1691 log_error("Default target masked.");
1693 if (!target || target->load_state != UNIT_LOADED) {
1694 log_info("Trying to load rescue target...");
1696 r = manager_load_unit(m, SPECIAL_RESCUE_TARGET, NULL, &error, &target);
1698 log_emergency("Failed to load rescue target: %s", bus_error_message(&error, r));
1699 error_message = "Failed to load rescue target";
1701 } else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND) {
1702 log_emergency_errno(target->load_error, "Failed to load rescue target: %m");
1703 error_message = "Failed to load rescue target";
1705 } else if (target->load_state == UNIT_MASKED) {
1706 log_emergency("Rescue target masked.");
1707 error_message = "Rescue target masked";
1712 assert(target->load_state == UNIT_LOADED);
1714 if (arg_action == ACTION_TEST) {
1715 printf("-> By units:\n");
1716 manager_dump_units(m, stdout, "\t");
1719 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, false, &error, &default_unit_job);
1721 log_debug("Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1723 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, false, &error, &default_unit_job);
1725 log_emergency("Failed to start default target: %s", bus_error_message(&error, r));
1726 error_message = "Failed to start default target";
1730 log_emergency("Failed to isolate default target: %s", bus_error_message(&error, r));
1731 error_message = "Failed to isolate default target";
1735 m->default_unit_job_id = default_unit_job->id;
1737 after_startup = now(CLOCK_MONOTONIC);
1738 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
1739 "Loaded units and determined initial transaction in %s.",
1740 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
1742 if (arg_action == ACTION_TEST) {
1743 printf("-> By jobs:\n");
1744 manager_dump_jobs(m, stdout, "\t");
1745 retval = EXIT_SUCCESS;
1751 r = manager_loop(m);
1753 log_emergency_errno(r, "Failed to run main loop: %m");
1754 error_message = "Failed to run main loop";
1758 switch (m->exit_code) {
1761 retval = EXIT_SUCCESS;
1765 case MANAGER_RELOAD:
1766 log_info("Reloading.");
1767 r = manager_reload(m);
1769 log_error_errno(r, "Failed to reload: %m");
1772 case MANAGER_REEXECUTE:
1774 if (prepare_reexecute(m, &arg_serialization, &fds, false) < 0) {
1775 error_message = "Failed to prepare for reexection";
1780 log_notice("Reexecuting.");
1783 case MANAGER_SWITCH_ROOT:
1784 /* Steal the switch root parameters */
1785 switch_root_dir = m->switch_root;
1786 switch_root_init = m->switch_root_init;
1787 m->switch_root = m->switch_root_init = NULL;
1789 if (!switch_root_init)
1790 if (prepare_reexecute(m, &arg_serialization, &fds, true) < 0) {
1791 error_message = "Failed to prepare for reexection";
1796 log_notice("Switching root.");
1799 case MANAGER_REBOOT:
1800 case MANAGER_POWEROFF:
1802 case MANAGER_KEXEC: {
1803 static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
1804 [MANAGER_REBOOT] = "reboot",
1805 [MANAGER_POWEROFF] = "poweroff",
1806 [MANAGER_HALT] = "halt",
1807 [MANAGER_KEXEC] = "kexec"
1810 assert_se(shutdown_verb = table[m->exit_code]);
1811 arm_reboot_watchdog = m->exit_code == MANAGER_REBOOT;
1813 log_notice("Shutting down.");
1818 assert_not_reached("Unknown exit code.");
1826 arg_shutdown_watchdog = m->shutdown_watchdog;
1827 m = manager_free(m);
1829 for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++) {
1830 free(arg_default_rlimit[j]);
1831 arg_default_rlimit[j] = NULL;
1834 free(arg_default_unit);
1835 arg_default_unit = NULL;
1837 free_join_controllers();
1839 strv_free(arg_default_environment);
1840 arg_default_environment = NULL;
1842 set_free(arg_syscall_archs);
1843 arg_syscall_archs = NULL;
1845 mac_selinux_finish();
1849 unsigned i, args_size;
1851 /* Close and disarm the watchdog, so that the new
1852 * instance can reinitialize it, but doesn't get
1853 * rebooted while we do that */
1854 watchdog_close(true);
1856 /* Reset the RLIMIT_NOFILE to the kernel default, so
1857 * that the new systemd can pass the kernel default to
1858 * its child processes */
1859 if (saved_rlimit_nofile.rlim_cur > 0)
1860 setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile);
1862 if (switch_root_dir) {
1863 /* Kill all remaining processes from the
1864 * initrd, but don't wait for them, so that we
1865 * can handle the SIGCHLD for them after
1867 broadcast_signal(SIGTERM, false, true);
1869 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1870 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1872 log_error_errno(r, "Failed to switch root, trying to continue: %m");
1875 args_size = MAX(6, argc+1);
1876 args = newa(const char*, args_size);
1878 if (!switch_root_init) {
1879 char sfd[DECIMAL_STR_MAX(int) + 1];
1881 /* First try to spawn ourselves with the right
1882 * path, and with full serialization. We do
1883 * this only if the user didn't specify an
1884 * explicit init to spawn. */
1886 assert(arg_serialization);
1889 xsprintf(sfd, "%i", fileno(arg_serialization));
1892 args[i++] = SYSTEMD_BINARY_PATH;
1893 if (switch_root_dir)
1894 args[i++] = "--switched-root";
1895 args[i++] = arg_running_as == SYSTEMD_SYSTEM ? "--system" : "--user";
1896 args[i++] = "--deserialize";
1900 /* do not pass along the environment we inherit from the kernel or initrd */
1901 if (switch_root_dir)
1904 assert(i <= args_size);
1905 execv(args[0], (char* const*) args);
1908 /* Try the fallback, if there is any, without any
1909 * serialization. We pass the original argv[] and
1910 * envp[]. (Well, modulo the ordering changes due to
1911 * getopt() in argv[], and some cleanups in envp[],
1912 * but let's hope that doesn't matter.) */
1914 if (arg_serialization) {
1915 fclose(arg_serialization);
1916 arg_serialization = NULL;
1924 /* Reopen the console */
1925 make_console_stdio();
1927 for (j = 1, i = 1; j < (unsigned) argc; j++)
1928 args[i++] = argv[j];
1930 assert(i <= args_size);
1932 /* Reenable any blocked signals, especially important
1933 * if we switch from initial ramdisk to init=... */
1934 reset_all_signal_handlers();
1935 reset_signal_mask();
1937 if (switch_root_init) {
1938 args[0] = switch_root_init;
1939 execv(args[0], (char* const*) args);
1940 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
1943 args[0] = "/sbin/init";
1944 execv(args[0], (char* const*) args);
1946 if (errno == ENOENT) {
1947 log_warning("No /sbin/init, trying fallback");
1949 args[0] = "/bin/sh";
1951 execv(args[0], (char* const*) args);
1952 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
1954 log_warning_errno(errno, "Failed to execute /sbin/init, giving up: %m");
1957 if (arg_serialization) {
1958 fclose(arg_serialization);
1959 arg_serialization = NULL;
1967 #ifdef HAVE_VALGRIND_VALGRIND_H
1968 /* If we are PID 1 and running under valgrind, then let's exit
1969 * here explicitly. valgrind will only generate nice output on
1970 * exit(), not on exec(), hence let's do the former not the
1972 if (getpid() == 1 && RUNNING_ON_VALGRIND)
1976 if (shutdown_verb) {
1977 char log_level[DECIMAL_STR_MAX(int) + 1];
1978 const char* command_line[9] = {
1979 SYSTEMD_SHUTDOWN_BINARY_PATH,
1981 "--log-level", log_level,
1985 _cleanup_strv_free_ char **env_block = NULL;
1987 assert(command_line[pos] == NULL);
1988 env_block = strv_copy(environ);
1990 xsprintf(log_level, "%d", log_get_max_level());
1992 switch (log_get_target()) {
1993 case LOG_TARGET_KMSG:
1994 case LOG_TARGET_JOURNAL_OR_KMSG:
1995 case LOG_TARGET_SYSLOG_OR_KMSG:
1996 command_line[pos++] = "kmsg";
1999 case LOG_TARGET_CONSOLE:
2001 command_line[pos++] = "console";
2005 if (log_get_show_color())
2006 command_line[pos++] = "--log-color";
2008 if (log_get_show_location())
2009 command_line[pos++] = "--log-location";
2011 assert(pos < ELEMENTSOF(command_line));
2013 if (arm_reboot_watchdog && arg_shutdown_watchdog > 0) {
2016 /* If we reboot let's set the shutdown
2017 * watchdog and tell the shutdown binary to
2018 * repeatedly ping it */
2019 r = watchdog_set_timeout(&arg_shutdown_watchdog);
2020 watchdog_close(r < 0);
2022 /* Tell the binary how often to ping, ignore failure */
2023 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
2024 strv_push(&env_block, e);
2026 watchdog_close(true);
2028 /* Avoid the creation of new processes forked by the
2029 * kernel; at this point, we will not listen to the
2031 if (detect_container(NULL) <= 0)
2032 cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
2034 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
2035 log_error_errno(errno, "Failed to execute shutdown binary, %s: %m",
2036 getpid() == 1 ? "freezing" : "quitting");
2039 if (getpid() == 1) {
2041 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2042 ANSI_HIGHLIGHT_RED_ON "!!!!!!" ANSI_HIGHLIGHT_OFF,
2043 "%s, freezing.", error_message);