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/>.
26 #include <sys/types.h>
32 #include <sys/prctl.h>
33 #include <sys/mount.h>
35 #ifdef HAVE_VALGRIND_VALGRIND_H
36 #include <valgrind/valgrind.h>
39 #include "sd-daemon.h"
40 #include "sd-messages.h"
44 #include "load-fragment.h"
47 #include "conf-parser.h"
55 #include "path-util.h"
56 #include "switch-root.h"
57 #include "capability.h"
62 #include "dbus-manager.h"
63 #include "bus-error.h"
66 #include "mount-setup.h"
67 #include "loopback-setup.h"
68 #include "hostname-setup.h"
69 #include "machine-id-setup.h"
70 #include "selinux-setup.h"
71 #include "ima-setup.h"
72 #include "smack-setup.h"
74 #include "kmod-setup.h"
82 ACTION_DUMP_CONFIGURATION_ITEMS,
84 } arg_action = ACTION_RUN;
85 static char *arg_default_unit = NULL;
86 static SystemdRunningAs arg_running_as = _SYSTEMD_RUNNING_AS_INVALID;
87 static bool arg_dump_core = true;
88 static bool arg_crash_shell = false;
89 static int arg_crash_chvt = -1;
90 static bool arg_confirm_spawn = false;
91 static ShowStatus arg_show_status = _SHOW_STATUS_UNSET;
92 static bool arg_switched_root = false;
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_NLIMITS] = {};
105 static uint64_t arg_capability_bounding_set_drop = 0;
106 static nsec_t arg_timer_slack_nsec = (nsec_t) -1;
107 static Set* arg_syscall_archs = NULL;
108 static FILE* arg_serialization = NULL;
110 static void nop_handler(int sig) {}
112 noreturn static void crash(int sig) {
115 /* Pass this on immediately, if this is not PID 1 */
117 else if (!arg_dump_core)
118 log_error("Caught <%s>, not dumping core.", signal_to_string(sig));
120 struct sigaction sa = {
121 .sa_handler = nop_handler,
122 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
126 /* We want to wait for the core process, hence let's enable SIGCHLD */
127 sigaction(SIGCHLD, &sa, NULL);
131 log_error("Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
134 struct rlimit rl = {};
136 /* Enable default signal handler for core dump */
138 sa.sa_handler = SIG_DFL;
139 sigaction(sig, &sa, NULL);
141 /* Don't limit the core dump size */
142 rl.rlim_cur = RLIM_INFINITY;
143 rl.rlim_max = RLIM_INFINITY;
144 setrlimit(RLIMIT_CORE, &rl);
146 /* Just to be sure... */
149 /* Raise the signal again */
152 assert_not_reached("We shouldn't be here...");
159 /* Order things nicely. */
160 r = wait_for_terminate(pid, &status);
162 log_error("Caught <%s>, waitpid() failed: %s", signal_to_string(sig), strerror(-r));
163 else if (status.si_code != CLD_DUMPED)
164 log_error("Caught <%s>, core dump failed.", signal_to_string(sig));
166 log_error("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
171 chvt(arg_crash_chvt);
173 if (arg_crash_shell) {
174 struct sigaction sa = {
175 .sa_handler = SIG_IGN,
176 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
180 log_info("Executing crash shell in 10s...");
183 /* Let the kernel reap children for us */
184 assert_se(sigaction(SIGCHLD, &sa, NULL) == 0);
188 log_error("Failed to fork off crash shell: %m");
190 make_console_stdio();
191 execl("/bin/sh", "/bin/sh", NULL);
193 log_error("execl() failed: %m");
197 log_info("Successfully spawned crash shell as pid "PID_FMT".", pid);
200 log_info("Freezing execution.");
204 static void install_crash_handler(void) {
205 struct sigaction sa = {
207 .sa_flags = SA_NODEFER,
210 sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
213 static int console_setup(bool do_reset) {
216 /* If we are init, we connect stdin/stdout/stderr to /dev/null
217 * and make sure we don't have a controlling tty. */
224 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
226 log_error("Failed to open /dev/console: %s", strerror(-tty_fd));
230 /* We don't want to force text mode.
231 * plymouth may be showing pictures already from initrd. */
232 r = reset_terminal_fd(tty_fd, false);
234 log_error("Failed to reset /dev/console: %s", strerror(-r));
236 close_nointr_nofail(tty_fd);
240 static int set_default_unit(const char *u) {
249 free(arg_default_unit);
250 arg_default_unit = c;
255 static int parse_proc_cmdline_word(const char *word) {
257 static const char * const rlmap[] = {
258 "emergency", SPECIAL_EMERGENCY_TARGET,
259 "-b", SPECIAL_EMERGENCY_TARGET,
260 "single", SPECIAL_RESCUE_TARGET,
261 "-s", SPECIAL_RESCUE_TARGET,
262 "s", SPECIAL_RESCUE_TARGET,
263 "S", SPECIAL_RESCUE_TARGET,
264 "1", SPECIAL_RESCUE_TARGET,
265 "2", SPECIAL_RUNLEVEL2_TARGET,
266 "3", SPECIAL_RUNLEVEL3_TARGET,
267 "4", SPECIAL_RUNLEVEL4_TARGET,
268 "5", SPECIAL_RUNLEVEL5_TARGET,
273 if (startswith(word, "systemd.unit=")) {
276 return set_default_unit(word + 13);
278 } else if (startswith(word, "rd.systemd.unit=")) {
281 return set_default_unit(word + 16);
283 } else if (startswith(word, "systemd.log_target=")) {
285 if (log_set_target_from_string(word + 19) < 0)
286 log_warning("Failed to parse log target %s. Ignoring.", word + 19);
288 } else if (startswith(word, "systemd.log_level=")) {
290 if (log_set_max_level_from_string(word + 18) < 0)
291 log_warning("Failed to parse log level %s. Ignoring.", word + 18);
293 } else if (startswith(word, "systemd.log_color=")) {
295 if (log_show_color_from_string(word + 18) < 0)
296 log_warning("Failed to parse log color setting %s. Ignoring.", word + 18);
298 } else if (startswith(word, "systemd.log_location=")) {
300 if (log_show_location_from_string(word + 21) < 0)
301 log_warning("Failed to parse log location setting %s. Ignoring.", word + 21);
303 } else if (startswith(word, "systemd.dump_core=")) {
306 r = parse_boolean(word + 18);
308 log_warning("Failed to parse dump core switch %s. Ignoring.", word + 18);
312 } else if (startswith(word, "systemd.crash_shell=")) {
315 r = parse_boolean(word + 20);
317 log_warning("Failed to parse crash shell switch %s. Ignoring.", word + 20);
321 } else if (startswith(word, "systemd.confirm_spawn=")) {
324 r = parse_boolean(word + 22);
326 log_warning("Failed to parse confirm spawn switch %s. Ignoring.", word + 22);
328 arg_confirm_spawn = r;
330 } else if (startswith(word, "systemd.crash_chvt=")) {
333 if (safe_atoi(word + 19, &k) < 0)
334 log_warning("Failed to parse crash chvt switch %s. Ignoring.", word + 19);
338 } else if (startswith(word, "systemd.show_status=")) {
341 r = parse_show_status(word + 20, &arg_show_status);
343 log_warning("Failed to parse show status switch %s. Ignoring.", word + 20);
344 } else if (startswith(word, "systemd.default_standard_output=")) {
347 r = exec_output_from_string(word + 32);
349 log_warning("Failed to parse default standard output switch %s. Ignoring.", word + 32);
351 arg_default_std_output = r;
352 } else if (startswith(word, "systemd.default_standard_error=")) {
355 r = exec_output_from_string(word + 31);
357 log_warning("Failed to parse default standard error switch %s. Ignoring.", word + 31);
359 arg_default_std_error = r;
360 } else if (startswith(word, "systemd.setenv=")) {
361 const char *cenv = word + 15;
363 if (env_assignment_is_valid(cenv)) {
366 env = strv_env_set(arg_default_environment, cenv);
368 arg_default_environment = env;
370 log_warning("Setting environment variable '%s' failed, ignoring: %s",
371 cenv, strerror(ENOMEM));
373 log_warning("Environment variable name '%s' is not valid. Ignoring.", cenv);
375 } else if (startswith(word, "systemd.") ||
376 (in_initrd() && startswith(word, "rd.systemd."))) {
380 /* Ignore systemd.journald.xyz and friends */
382 if (startswith(c, "rd."))
384 if (startswith(c, "systemd."))
386 if (c[strcspn(c, ".=")] != '.') {
388 log_warning("Unknown kernel switch %s. Ignoring.", word);
390 log_info("Supported kernel switches:\n"
391 "systemd.unit=UNIT Default unit to start\n"
392 "rd.systemd.unit=UNIT Default unit to start when run in initrd\n"
393 "systemd.dump_core=0|1 Dump core on crash\n"
394 "systemd.crash_shell=0|1 Run shell on crash\n"
395 "systemd.crash_chvt=N Change to VT #N on crash\n"
396 "systemd.confirm_spawn=0|1 Confirm every process spawn\n"
397 "systemd.show_status=0|1|auto Show status updates on the console during bootup\n"
398 "systemd.log_target=console|kmsg|journal|journal-or-kmsg|syslog|syslog-or-kmsg|null\n"
400 "systemd.log_level=LEVEL Log level\n"
401 "systemd.log_color=0|1 Highlight important log messages\n"
402 "systemd.log_location=0|1 Include code location in log messages\n"
403 "systemd.default_standard_output=null|tty|syslog|syslog+console|kmsg|kmsg+console|journal|journal+console\n"
404 " Set default log output for services\n"
405 "systemd.default_standard_error=null|tty|syslog|syslog+console|kmsg|kmsg+console|journal|journal+console\n"
406 " Set default log error output for services\n"
407 "systemd.setenv=ASSIGNMENT Set an environment variable for all spawned processes\n");
410 } else if (streq(word, "quiet")) {
411 if (arg_show_status == _SHOW_STATUS_UNSET)
412 arg_show_status = SHOW_STATUS_AUTO;
413 } else if (streq(word, "debug")) {
414 /* Log to kmsg, the journal socket will fill up before the
415 * journal is started and tools running during that time
416 * will block with every log message for for 60 seconds,
417 * before they give up. */
418 log_set_max_level(LOG_DEBUG);
419 log_set_target(detect_container(NULL) > 0 ? LOG_TARGET_CONSOLE : LOG_TARGET_KMSG);
420 } else if (!in_initrd()) {
423 /* SysV compatibility */
424 for (i = 0; i < ELEMENTSOF(rlmap); i += 2)
425 if (streq(word, rlmap[i]))
426 return set_default_unit(rlmap[i+1]);
432 #define DEFINE_SETTER(name, func, descr) \
433 static int name(const char *unit, \
434 const char *filename, \
436 const char *section, \
437 unsigned section_line, \
438 const char *lvalue, \
440 const char *rvalue, \
452 log_syntax(unit, LOG_ERR, filename, line, -r, \
453 "Invalid " descr "'%s': %s", \
454 rvalue, strerror(-r)); \
459 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level")
460 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target")
461 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" )
462 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location")
464 static int config_parse_cpu_affinity2(const char *unit,
465 const char *filename,
468 unsigned section_line,
485 FOREACH_WORD_QUOTED(w, l, rvalue, state) {
490 if (!(t = strndup(w, l)))
493 r = safe_atou(t, &cpu);
497 if (!(c = cpu_set_malloc(&ncpus)))
500 if (r < 0 || cpu >= ncpus) {
501 log_syntax(unit, LOG_ERR, filename, line, -r,
502 "Failed to parse CPU affinity '%s'", rvalue);
507 CPU_SET_S(cpu, CPU_ALLOC_SIZE(ncpus), c);
511 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
512 log_warning_unit(unit, "Failed to set CPU affinity: %m");
520 static void strv_free_free(char ***l) {
532 static void free_join_controllers(void) {
533 strv_free_free(arg_join_controllers);
534 arg_join_controllers = NULL;
537 static int config_parse_join_controllers(const char *unit,
538 const char *filename,
541 unsigned section_line,
556 free_join_controllers();
558 FOREACH_WORD_QUOTED(w, length, rvalue, state) {
561 s = strndup(w, length);
565 l = strv_split(s, ",");
570 if (strv_length(l) <= 1) {
575 if (!arg_join_controllers) {
576 arg_join_controllers = new(char**, 2);
577 if (!arg_join_controllers) {
582 arg_join_controllers[0] = l;
583 arg_join_controllers[1] = NULL;
590 t = new0(char**, n+2);
598 for (a = arg_join_controllers; *a; a++) {
600 if (strv_overlap(*a, l)) {
601 if (strv_extend_strv(&l, *a) < 0) {
621 t[n++] = strv_uniq(l);
623 strv_free_free(arg_join_controllers);
624 arg_join_controllers = t;
631 static int parse_config_file(void) {
633 const ConfigTableItem items[] = {
634 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
635 { "Manager", "LogTarget", config_parse_target, 0, NULL },
636 { "Manager", "LogColor", config_parse_color, 0, NULL },
637 { "Manager", "LogLocation", config_parse_location, 0, NULL },
638 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
639 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
640 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
641 { "Manager", "CrashChVT", config_parse_int, 0, &arg_crash_chvt },
642 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
643 { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
644 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
645 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
646 { "Manager", "CapabilityBoundingSet", config_parse_bounding_set, 0, &arg_capability_bounding_set_drop },
648 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
650 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
651 { "Manager", "DefaultStandardOutput", config_parse_output, 0, &arg_default_std_output },
652 { "Manager", "DefaultStandardError", config_parse_output, 0, &arg_default_std_error },
653 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
654 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
655 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
656 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval },
657 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
658 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
659 { "Manager", "DefaultLimitCPU", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CPU] },
660 { "Manager", "DefaultLimitFSIZE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_FSIZE] },
661 { "Manager", "DefaultLimitDATA", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_DATA] },
662 { "Manager", "DefaultLimitSTACK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_STACK] },
663 { "Manager", "DefaultLimitCORE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CORE] },
664 { "Manager", "DefaultLimitRSS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RSS] },
665 { "Manager", "DefaultLimitNOFILE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NOFILE] },
666 { "Manager", "DefaultLimitAS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_AS] },
667 { "Manager", "DefaultLimitNPROC", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NPROC] },
668 { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MEMLOCK] },
669 { "Manager", "DefaultLimitLOCKS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_LOCKS] },
670 { "Manager", "DefaultLimitSIGPENDING", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_SIGPENDING] },
671 { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MSGQUEUE] },
672 { "Manager", "DefaultLimitNICE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NICE] },
673 { "Manager", "DefaultLimitRTPRIO", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTPRIO] },
674 { "Manager", "DefaultLimitRTTIME", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTTIME] },
678 _cleanup_fclose_ FILE *f;
682 fn = arg_running_as == SYSTEMD_SYSTEM ? PKGSYSCONFDIR "/system.conf" : PKGSYSCONFDIR "/user.conf";
688 log_warning("Failed to open configuration file '%s': %m", fn);
692 r = config_parse(NULL, fn, f, "Manager\0", config_item_table_lookup, (void*) items, false, false, NULL);
694 log_warning("Failed to parse configuration file: %s", strerror(-r));
699 static int parse_argv(int argc, char *argv[]) {
702 ARG_LOG_LEVEL = 0x100,
711 ARG_DUMP_CONFIGURATION_ITEMS,
718 ARG_DEFAULT_STD_OUTPUT,
719 ARG_DEFAULT_STD_ERROR
722 static const struct option options[] = {
723 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
724 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
725 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
726 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
727 { "unit", required_argument, NULL, ARG_UNIT },
728 { "system", no_argument, NULL, ARG_SYSTEM },
729 { "user", no_argument, NULL, ARG_USER },
730 { "test", no_argument, NULL, ARG_TEST },
731 { "help", no_argument, NULL, 'h' },
732 { "version", no_argument, NULL, ARG_VERSION },
733 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
734 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
735 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
736 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
737 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
738 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
739 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
740 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
741 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
753 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
758 r = log_set_max_level_from_string(optarg);
760 log_error("Failed to parse log level %s.", optarg);
767 r = log_set_target_from_string(optarg);
769 log_error("Failed to parse log target %s.", optarg);
778 r = log_show_color_from_string(optarg);
780 log_error("Failed to parse log color setting %s.", optarg);
784 log_show_color(true);
788 case ARG_LOG_LOCATION:
790 r = log_show_location_from_string(optarg);
792 log_error("Failed to parse log location setting %s.", optarg);
796 log_show_location(true);
800 case ARG_DEFAULT_STD_OUTPUT:
801 r = exec_output_from_string(optarg);
803 log_error("Failed to parse default standard output setting %s.", optarg);
806 arg_default_std_output = r;
809 case ARG_DEFAULT_STD_ERROR:
810 r = exec_output_from_string(optarg);
812 log_error("Failed to parse default standard error output setting %s.", optarg);
815 arg_default_std_error = r;
820 r = set_default_unit(optarg);
822 log_error("Failed to set default unit %s: %s", optarg, strerror(-r));
829 arg_running_as = SYSTEMD_SYSTEM;
833 arg_running_as = SYSTEMD_USER;
837 arg_action = ACTION_TEST;
841 arg_action = ACTION_VERSION;
844 case ARG_DUMP_CONFIGURATION_ITEMS:
845 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
849 r = optarg ? parse_boolean(optarg) : 1;
851 log_error("Failed to parse dump core boolean %s.", optarg);
857 case ARG_CRASH_SHELL:
858 r = optarg ? parse_boolean(optarg) : 1;
860 log_error("Failed to parse crash shell boolean %s.", optarg);
866 case ARG_CONFIRM_SPAWN:
867 r = optarg ? parse_boolean(optarg) : 1;
869 log_error("Failed to parse confirm spawn boolean %s.", optarg);
872 arg_confirm_spawn = r;
875 case ARG_SHOW_STATUS:
877 r = parse_show_status(optarg, &arg_show_status);
879 log_error("Failed to parse show status boolean %s.", optarg);
883 arg_show_status = SHOW_STATUS_YES;
886 case ARG_DESERIALIZE: {
890 r = safe_atoi(optarg, &fd);
891 if (r < 0 || fd < 0) {
892 log_error("Failed to parse deserialize option %s.", optarg);
893 return r < 0 ? r : -EINVAL;
896 fd_cloexec(fd, true);
900 log_error("Failed to open serialization fd: %m");
904 if (arg_serialization)
905 fclose(arg_serialization);
907 arg_serialization = f;
912 case ARG_SWITCHED_ROOT:
913 arg_switched_root = true;
917 arg_action = ACTION_HELP;
921 log_set_max_level(LOG_DEBUG);
927 /* Just to eat away the sysvinit kernel
928 * cmdline args without getopt() error
929 * messages that we'll parse in
930 * parse_proc_cmdline_word() or ignore. */
935 log_error("Unknown option code %c", c);
942 if (optind < argc && getpid() != 1) {
943 /* Hmm, when we aren't run as init system
944 * let's complain about excess arguments */
946 log_error("Excess arguments.");
950 if (detect_container(NULL) > 0) {
953 /* All /proc/cmdline arguments the kernel didn't
954 * understand it passed to us. We're not really
955 * interested in that usually since /proc/cmdline is
956 * more interesting and complete. With one exception:
957 * if we are run in a container /proc/cmdline is not
958 * relevant for the container, hence we rely on argv[]
961 for (a = argv; a < argv + argc; a++) {
962 r = parse_proc_cmdline_word(*a);
964 log_error("Failed on cmdline argument %s: %s", *a, strerror(-r));
973 static int help(void) {
975 printf("%s [OPTIONS...]\n\n"
976 "Starts up and maintains the system or user services.\n\n"
977 " -h --help Show this help\n"
978 " --test Determine startup sequence, dump it and exit\n"
979 " --dump-configuration-items Dump understood unit configuration items\n"
980 " --unit=UNIT Set default unit\n"
981 " --system Run a system instance, even if PID != 1\n"
982 " --user Run a user instance\n"
983 " --dump-core[=0|1] Dump core on crash\n"
984 " --crash-shell[=0|1] Run shell on crash\n"
985 " --confirm-spawn[=0|1] Ask for confirmation when spawning processes\n"
986 " --show-status[=0|1] Show status updates on the console during bootup\n"
987 " --log-target=TARGET Set log target (console, journal, syslog, kmsg, journal-or-kmsg, syslog-or-kmsg, null)\n"
988 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
989 " --log-color[=0|1] Highlight important log messages\n"
990 " --log-location[=0|1] Include code location in log messages\n"
991 " --default-standard-output= Set default standard output for services\n"
992 " --default-standard-error= Set default standard error output for services\n",
993 program_invocation_short_name);
998 static int version(void) {
999 puts(PACKAGE_STRING);
1000 puts(SYSTEMD_FEATURES);
1005 static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
1014 r = manager_open_serialization(m, &f);
1016 log_error("Failed to create serialization file: %s", strerror(-r));
1020 /* Make sure nothing is really destructed when we shut down */
1022 bus_manager_send_reloading(m, true);
1027 log_error("Failed to allocate fd set: %s", strerror(-r));
1031 r = manager_serialize(m, f, fds, switching_root);
1033 log_error("Failed to serialize state: %s", strerror(-r));
1037 if (fseeko(f, 0, SEEK_SET) < 0) {
1038 log_error("Failed to rewind serialization fd: %m");
1042 r = fd_cloexec(fileno(f), false);
1044 log_error("Failed to disable O_CLOEXEC for serialization: %s", strerror(-r));
1048 r = fdset_cloexec(fds, false);
1050 log_error("Failed to disable O_CLOEXEC for serialization fds: %s", strerror(-r));
1068 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1072 assert(saved_rlimit);
1074 /* Save the original RLIMIT_NOFILE so that we can reset it
1075 * later when transitioning from the initrd to the main
1076 * systemd or suchlike. */
1077 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0) {
1078 log_error("Reading RLIMIT_NOFILE failed: %m");
1082 /* Make sure forked processes get the default kernel setting */
1083 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1086 rl = newdup(struct rlimit, saved_rlimit, 1);
1090 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1093 /* Bump up the resource limit for ourselves substantially */
1094 nl.rlim_cur = nl.rlim_max = 64*1024;
1095 r = setrlimit_closest(RLIMIT_NOFILE, &nl);
1097 log_error("Setting RLIMIT_NOFILE failed: %s", strerror(-r));
1104 static void test_mtab(void) {
1107 /* Check that /etc/mtab is a symlink */
1109 if (readlink_malloc("/etc/mtab", &p) >= 0) {
1112 b = streq(p, "/proc/self/mounts") || streq(p, "/proc/mounts");
1119 log_warning("/etc/mtab is not a symlink or not pointing to /proc/self/mounts. "
1120 "This is not supported anymore. "
1121 "Please make sure to replace this file by a symlink to avoid incorrect or misleading mount(8) output.");
1124 static void test_usr(void) {
1126 /* Check that /usr is not a separate fs */
1128 if (dir_is_empty("/usr") <= 0)
1131 log_warning("/usr appears to be on its own filesytem and is not already mounted. This is not a supported setup. "
1132 "Some things will probably break (sometimes even silently) in mysterious ways. "
1133 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1136 static void test_cgroups(void) {
1138 if (access("/proc/cgroups", F_OK) >= 0)
1141 log_warning("CONFIG_CGROUPS was not set when your kernel was compiled. "
1142 "Systems without control groups are not supported. "
1143 "We will now sleep for 10s, and then continue boot-up. "
1144 "Expect breakage and please do not file bugs. "
1145 "Instead fix your kernel and enable CONFIG_CGROUPS. "
1146 "Consult http://0pointer.de/blog/projects/cgroups-vs-cgroups.html for more information.");
1151 static int initialize_join_controllers(void) {
1152 /* By default, mount "cpu" + "cpuacct" together, and "net_cls"
1153 * + "net_prio". We'd like to add "cpuset" to the mix, but
1154 * "cpuset" does't really work for groups with no initialized
1157 arg_join_controllers = new(char**, 3);
1158 if (!arg_join_controllers)
1161 arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL);
1162 arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL);
1163 arg_join_controllers[2] = NULL;
1165 if (!arg_join_controllers[0] || !arg_join_controllers[1]) {
1166 free_join_controllers();
1173 static int enforce_syscall_archs(Set *archs) {
1175 scmp_filter_ctx *seccomp;
1180 seccomp = seccomp_init(SCMP_ACT_ALLOW);
1184 SET_FOREACH(id, arg_syscall_archs, i) {
1185 r = seccomp_arch_add(seccomp, PTR_TO_UINT32(id) - 1);
1189 log_error("Failed to add architecture to seccomp: %s", strerror(-r));
1194 r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
1196 log_error("Failed to unset NO_NEW_PRIVS: %s", strerror(-r));
1200 r = seccomp_load(seccomp);
1202 log_error("Failed to add install architecture seccomp: %s", strerror(-r));
1205 seccomp_release(seccomp);
1212 static int status_welcome(void) {
1213 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1216 r = parse_env_file("/etc/os-release", NEWLINE,
1217 "PRETTY_NAME", &pretty_name,
1218 "ANSI_COLOR", &ansi_color,
1221 if (r < 0 && r != -ENOENT)
1222 log_warning("Failed to read /etc/os-release: %s", strerror(-r));
1224 return status_printf(NULL, false, false,
1225 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1226 isempty(ansi_color) ? "1" : ansi_color,
1227 isempty(pretty_name) ? "Linux" : pretty_name);
1230 int main(int argc, char *argv[]) {
1232 int r, retval = EXIT_FAILURE;
1233 usec_t before_startup, after_startup;
1234 char timespan[FORMAT_TIMESPAN_MAX];
1236 bool reexecute = false;
1237 const char *shutdown_verb = NULL;
1238 dual_timestamp initrd_timestamp = { 0ULL, 0ULL };
1239 dual_timestamp userspace_timestamp = { 0ULL, 0ULL };
1240 dual_timestamp kernel_timestamp = { 0ULL, 0ULL };
1241 dual_timestamp security_start_timestamp = { 0ULL, 0ULL };
1242 dual_timestamp security_finish_timestamp = { 0ULL, 0ULL };
1243 static char systemd[] = "systemd";
1244 bool skip_setup = false;
1246 bool loaded_policy = false;
1247 bool arm_reboot_watchdog = false;
1248 bool queue_default_job = false;
1249 char *switch_root_dir = NULL, *switch_root_init = NULL;
1250 static struct rlimit saved_rlimit_nofile = { 0, 0 };
1252 #ifdef HAVE_SYSV_COMPAT
1253 if (getpid() != 1 && strstr(program_invocation_short_name, "init")) {
1254 /* This is compatibility support for SysV, where
1255 * calling init as a user is identical to telinit. */
1258 execv(SYSTEMCTL_BINARY_PATH, argv);
1259 log_error("Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1264 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
1265 dual_timestamp_get(&userspace_timestamp);
1267 /* Determine if this is a reexecution or normal bootup. We do
1268 * the full command line parsing much later, so let's just
1269 * have a quick peek here. */
1270 if (strv_find(argv+1, "--deserialize"))
1273 /* If we have switched root, do all the special setup
1275 if (strv_find(argv+1, "--switched-root"))
1278 /* If we get started via the /sbin/init symlink then we are
1279 called 'init'. After a subsequent reexecution we are then
1280 called 'systemd'. That is confusing, hence let's call us
1281 systemd right-away. */
1282 program_invocation_short_name = systemd;
1283 prctl(PR_SET_NAME, systemd);
1288 log_show_color(isatty(STDERR_FILENO) > 0);
1290 /* Disable the umask logic */
1294 if (getpid() == 1 && detect_container(NULL) <= 0) {
1296 /* Running outside of a container as PID 1 */
1297 arg_running_as = SYSTEMD_SYSTEM;
1299 log_set_target(LOG_TARGET_KMSG);
1303 initrd_timestamp = userspace_timestamp;
1306 mount_setup_early();
1307 dual_timestamp_get(&security_start_timestamp);
1308 if (selinux_setup(&loaded_policy) < 0)
1310 if (ima_setup() < 0)
1312 if (smack_setup() < 0)
1314 dual_timestamp_get(&security_finish_timestamp);
1317 if (label_init(NULL) < 0)
1321 if (hwclock_is_localtime() > 0) {
1324 /* The first-time call to settimeofday() does a time warp in the kernel */
1325 r = hwclock_set_timezone(&min);
1327 log_error("Failed to apply local time delta, ignoring: %s", strerror(-r));
1329 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1330 } else if (!in_initrd()) {
1332 * Do dummy first-time call to seal the kernel's time warp magic
1334 * Do not call this this from inside the initrd. The initrd might not
1335 * carry /etc/adjtime with LOCAL, but the real system could be set up
1336 * that way. In such case, we need to delay the time-warp or the sealing
1337 * until we reach the real system.
1339 hwclock_reset_timezone();
1341 /* Tell the kernel our timezone */
1342 r = hwclock_set_timezone(NULL);
1344 log_error("Failed to set the kernel's timezone, ignoring: %s", strerror(-r));
1348 /* Set the default for later on, but don't actually
1349 * open the logs like this for now. Note that if we
1350 * are transitioning from the initrd there might still
1351 * be journal fd open, and we shouldn't attempt
1352 * opening that before we parsed /proc/cmdline which
1353 * might redirect output elsewhere. */
1354 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
1356 } else if (getpid() == 1) {
1357 /* Running inside a container, as PID 1 */
1358 arg_running_as = SYSTEMD_SYSTEM;
1359 log_set_target(LOG_TARGET_CONSOLE);
1360 log_close_console(); /* force reopen of /dev/console */
1363 /* For the later on, see above... */
1364 log_set_target(LOG_TARGET_JOURNAL);
1366 /* clear the kernel timestamp,
1367 * because we are in a container */
1368 kernel_timestamp.monotonic = 0ULL;
1369 kernel_timestamp.realtime = 0ULL;
1372 /* Running as user instance */
1373 arg_running_as = SYSTEMD_USER;
1374 log_set_target(LOG_TARGET_AUTO);
1377 /* clear the kernel timestamp,
1378 * because we are not PID 1 */
1379 kernel_timestamp.monotonic = 0ULL;
1380 kernel_timestamp.realtime = 0ULL;
1383 /* Initialize default unit */
1384 r = set_default_unit(SPECIAL_DEFAULT_TARGET);
1386 log_error("Failed to set default unit %s: %s", SPECIAL_DEFAULT_TARGET, strerror(-r));
1390 r = initialize_join_controllers();
1394 /* Mount /proc, /sys and friends, so that /proc/cmdline and
1395 * /proc/$PID/fd is available. */
1396 if (getpid() == 1) {
1397 r = mount_setup(loaded_policy);
1402 /* Reset all signal handlers. */
1403 assert_se(reset_all_signal_handlers() == 0);
1405 ignore_signals(SIGNALS_IGNORE, -1);
1407 if (parse_config_file() < 0)
1410 if (arg_running_as == SYSTEMD_SYSTEM)
1411 if (parse_proc_cmdline(parse_proc_cmdline_word) < 0)
1414 log_parse_environment();
1416 if (parse_argv(argc, argv) < 0)
1419 if (arg_action == ACTION_TEST &&
1421 log_error("Don't run test mode as root.");
1425 if (arg_running_as == SYSTEMD_USER &&
1426 arg_action == ACTION_RUN &&
1428 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
1432 if (arg_running_as == SYSTEMD_SYSTEM &&
1433 arg_action == ACTION_RUN &&
1434 running_in_chroot() > 0) {
1435 log_error("Cannot be run in a chroot() environment.");
1439 if (arg_action == ACTION_HELP) {
1442 } else if (arg_action == ACTION_VERSION) {
1445 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
1446 unit_dump_config_items(stdout);
1447 retval = EXIT_SUCCESS;
1449 } else if (arg_action == ACTION_DONE) {
1450 retval = EXIT_SUCCESS;
1454 if (arg_running_as == SYSTEMD_USER &&
1455 !getenv("XDG_RUNTIME_DIR")) {
1456 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
1460 assert_se(arg_action == ACTION_RUN || arg_action == ACTION_TEST);
1462 /* Close logging fds, in order not to confuse fdset below */
1465 /* Remember open file descriptors for later deserialization */
1466 r = fdset_new_fill(&fds);
1468 log_error("Failed to allocate fd set: %s", strerror(-r));
1471 fdset_cloexec(fds, true);
1473 if (arg_serialization)
1474 assert_se(fdset_remove(fds, fileno(arg_serialization)) >= 0);
1476 if (arg_running_as == SYSTEMD_SYSTEM)
1477 /* Become a session leader if we aren't one yet. */
1480 /* Move out of the way, so that we won't block unmounts */
1481 assert_se(chdir("/") == 0);
1483 /* Reset the console, but only if this is really init and we
1484 * are freshly booted */
1485 if (arg_running_as == SYSTEMD_SYSTEM && arg_action == ACTION_RUN)
1486 console_setup(getpid() == 1 && !skip_setup);
1488 /* Open the logging devices, if possible and necessary */
1491 if (arg_show_status == _SHOW_STATUS_UNSET)
1492 arg_show_status = SHOW_STATUS_YES;
1494 /* Make sure we leave a core dump without panicing the
1496 if (getpid() == 1) {
1497 install_crash_handler();
1499 r = mount_cgroup_controllers(arg_join_controllers);
1504 if (arg_running_as == SYSTEMD_SYSTEM) {
1505 const char *virtualization = NULL;
1507 log_info(PACKAGE_STRING " running in system mode. (" SYSTEMD_FEATURES ")");
1509 detect_virtualization(&virtualization);
1511 log_info("Detected virtualization '%s'.", virtualization);
1514 log_info("Running in initial RAM disk.");
1517 _cleanup_free_ char *t = uid_to_name(getuid());
1518 log_debug(PACKAGE_STRING " running in user mode for user "PID_FMT"/%s. (" SYSTEMD_FEATURES ")",
1522 if (arg_running_as == SYSTEMD_SYSTEM && !skip_setup) {
1523 if (arg_show_status > 0 || plymouth_running())
1527 if (detect_container(NULL) <= 0)
1539 if (arg_running_as == SYSTEMD_SYSTEM && arg_runtime_watchdog > 0)
1540 watchdog_set_timeout(&arg_runtime_watchdog);
1542 if (arg_timer_slack_nsec != (nsec_t) -1)
1543 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1544 log_error("Failed to adjust timer slack: %m");
1546 if (arg_capability_bounding_set_drop) {
1547 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set_drop);
1549 log_error("Failed to drop capability bounding set of usermode helpers: %s", strerror(-r));
1552 r = capability_bounding_set_drop(arg_capability_bounding_set_drop, true);
1554 log_error("Failed to drop capability bounding set: %s", strerror(-r));
1559 if (arg_syscall_archs) {
1560 r = enforce_syscall_archs(arg_syscall_archs);
1565 if (arg_running_as == SYSTEMD_USER) {
1566 /* Become reaper of our children */
1567 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0) {
1568 log_warning("Failed to make us a subreaper: %m");
1569 if (errno == EINVAL)
1570 log_info("Perhaps the kernel version is too old (< 3.4?)");
1574 if (arg_running_as == SYSTEMD_SYSTEM)
1575 bump_rlimit_nofile(&saved_rlimit_nofile);
1577 r = manager_new(arg_running_as, &m);
1579 log_error("Failed to allocate manager object: %s", strerror(-r));
1583 m->confirm_spawn = arg_confirm_spawn;
1584 m->default_std_output = arg_default_std_output;
1585 m->default_std_error = arg_default_std_error;
1586 m->default_restart_usec = arg_default_restart_usec;
1587 m->default_timeout_start_usec = arg_default_timeout_start_usec;
1588 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
1589 m->default_start_limit_interval = arg_default_start_limit_interval;
1590 m->default_start_limit_burst = arg_default_start_limit_burst;
1591 m->runtime_watchdog = arg_runtime_watchdog;
1592 m->shutdown_watchdog = arg_shutdown_watchdog;
1593 m->userspace_timestamp = userspace_timestamp;
1594 m->kernel_timestamp = kernel_timestamp;
1595 m->initrd_timestamp = initrd_timestamp;
1596 m->security_start_timestamp = security_start_timestamp;
1597 m->security_finish_timestamp = security_finish_timestamp;
1599 manager_set_default_rlimits(m, arg_default_rlimit);
1600 manager_environment_add(m, NULL, arg_default_environment);
1601 manager_set_show_status(m, arg_show_status);
1603 /* Remember whether we should queue the default job */
1604 queue_default_job = !arg_serialization || arg_switched_root;
1606 before_startup = now(CLOCK_MONOTONIC);
1608 r = manager_startup(m, arg_serialization, fds);
1610 log_error("Failed to fully start up daemon: %s", strerror(-r));
1612 /* This will close all file descriptors that were opened, but
1613 * not claimed by any unit. */
1617 if (arg_serialization) {
1618 fclose(arg_serialization);
1619 arg_serialization = NULL;
1622 if (queue_default_job) {
1623 _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL;
1624 Unit *target = NULL;
1625 Job *default_unit_job;
1627 log_debug("Activating default unit: %s", arg_default_unit);
1629 r = manager_load_unit(m, arg_default_unit, NULL, &error, &target);
1631 log_error("Failed to load default target: %s", bus_error_message(&error, r));
1632 else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND)
1633 log_error("Failed to load default target: %s", strerror(-target->load_error));
1634 else if (target->load_state == UNIT_MASKED)
1635 log_error("Default target masked.");
1637 if (!target || target->load_state != UNIT_LOADED) {
1638 log_info("Trying to load rescue target...");
1640 r = manager_load_unit(m, SPECIAL_RESCUE_TARGET, NULL, &error, &target);
1642 log_error("Failed to load rescue target: %s", bus_error_message(&error, r));
1644 } else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND) {
1645 log_error("Failed to load rescue target: %s", strerror(-target->load_error));
1647 } else if (target->load_state == UNIT_MASKED) {
1648 log_error("Rescue target masked.");
1653 assert(target->load_state == UNIT_LOADED);
1655 if (arg_action == ACTION_TEST) {
1656 printf("-> By units:\n");
1657 manager_dump_units(m, stdout, "\t");
1660 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, false, &error, &default_unit_job);
1662 log_debug("Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1664 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, false, &error, &default_unit_job);
1666 log_error("Failed to start default target: %s", bus_error_message(&error, r));
1670 log_error("Failed to isolate default target: %s", bus_error_message(&error, r));
1674 m->default_unit_job_id = default_unit_job->id;
1676 after_startup = now(CLOCK_MONOTONIC);
1677 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
1678 "Loaded units and determined initial transaction in %s.",
1679 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 0));
1681 if (arg_action == ACTION_TEST) {
1682 printf("-> By jobs:\n");
1683 manager_dump_jobs(m, stdout, "\t");
1684 retval = EXIT_SUCCESS;
1690 r = manager_loop(m);
1692 log_error("Failed to run mainloop: %s", strerror(-r));
1696 switch (m->exit_code) {
1699 retval = EXIT_SUCCESS;
1703 case MANAGER_RELOAD:
1704 log_info("Reloading.");
1705 r = manager_reload(m);
1707 log_error("Failed to reload: %s", strerror(-r));
1710 case MANAGER_REEXECUTE:
1712 if (prepare_reexecute(m, &arg_serialization, &fds, false) < 0)
1716 log_notice("Reexecuting.");
1719 case MANAGER_SWITCH_ROOT:
1720 /* Steal the switch root parameters */
1721 switch_root_dir = m->switch_root;
1722 switch_root_init = m->switch_root_init;
1723 m->switch_root = m->switch_root_init = NULL;
1725 if (!switch_root_init)
1726 if (prepare_reexecute(m, &arg_serialization, &fds, true) < 0)
1730 log_notice("Switching root.");
1733 case MANAGER_REBOOT:
1734 case MANAGER_POWEROFF:
1736 case MANAGER_KEXEC: {
1737 static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
1738 [MANAGER_REBOOT] = "reboot",
1739 [MANAGER_POWEROFF] = "poweroff",
1740 [MANAGER_HALT] = "halt",
1741 [MANAGER_KEXEC] = "kexec"
1744 assert_se(shutdown_verb = table[m->exit_code]);
1745 arm_reboot_watchdog = m->exit_code == MANAGER_REBOOT;
1747 log_notice("Shutting down.");
1752 assert_not_reached("Unknown exit code.");
1762 for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++) {
1763 free(arg_default_rlimit[j]);
1764 arg_default_rlimit[j] = NULL;
1767 free(arg_default_unit);
1768 arg_default_unit = NULL;
1770 free_join_controllers();
1772 strv_free(arg_default_environment);
1773 arg_default_environment = NULL;
1775 set_free(arg_syscall_archs);
1776 arg_syscall_archs = NULL;
1782 unsigned i, args_size;
1784 /* Close and disarm the watchdog, so that the new
1785 * instance can reinitialize it, but doesn't get
1786 * rebooted while we do that */
1787 watchdog_close(true);
1789 /* Reset the RLIMIT_NOFILE to the kernel default, so
1790 * that the new systemd can pass the kernel default to
1791 * its child processes */
1792 if (saved_rlimit_nofile.rlim_cur > 0)
1793 setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile);
1795 if (switch_root_dir) {
1796 /* Kill all remaining processes from the
1797 * initrd, but don't wait for them, so that we
1798 * can handle the SIGCHLD for them after
1800 broadcast_signal(SIGTERM, false, true);
1802 /* And switch root */
1803 r = switch_root(switch_root_dir);
1805 log_error("Failed to switch root, ignoring: %s", strerror(-r));
1808 args_size = MAX(6, argc+1);
1809 args = newa(const char*, args_size);
1811 if (!switch_root_init) {
1814 /* First try to spawn ourselves with the right
1815 * path, and with full serialization. We do
1816 * this only if the user didn't specify an
1817 * explicit init to spawn. */
1819 assert(arg_serialization);
1822 snprintf(sfd, sizeof(sfd), "%i", fileno(arg_serialization));
1826 args[i++] = SYSTEMD_BINARY_PATH;
1827 if (switch_root_dir)
1828 args[i++] = "--switched-root";
1829 args[i++] = arg_running_as == SYSTEMD_SYSTEM ? "--system" : "--user";
1830 args[i++] = "--deserialize";
1834 /* do not pass along the environment we inherit from the kernel or initrd */
1835 if (switch_root_dir)
1838 assert(i <= args_size);
1839 execv(args[0], (char* const*) args);
1842 /* Try the fallback, if there is any, without any
1843 * serialization. We pass the original argv[] and
1844 * envp[]. (Well, modulo the ordering changes due to
1845 * getopt() in argv[], and some cleanups in envp[],
1846 * but let's hope that doesn't matter.) */
1848 if (arg_serialization) {
1849 fclose(arg_serialization);
1850 arg_serialization = NULL;
1858 /* Reopen the console */
1859 make_console_stdio();
1861 for (j = 1, i = 1; j < (unsigned) argc; j++)
1862 args[i++] = argv[j];
1864 assert(i <= args_size);
1866 if (switch_root_init) {
1867 args[0] = switch_root_init;
1868 execv(args[0], (char* const*) args);
1869 log_warning("Failed to execute configured init, trying fallback: %m");
1872 args[0] = "/sbin/init";
1873 execv(args[0], (char* const*) args);
1875 if (errno == ENOENT) {
1876 log_warning("No /sbin/init, trying fallback");
1878 args[0] = "/bin/sh";
1880 execv(args[0], (char* const*) args);
1881 log_error("Failed to execute /bin/sh, giving up: %m");
1883 log_warning("Failed to execute /sbin/init, giving up: %m");
1886 if (arg_serialization) {
1887 fclose(arg_serialization);
1888 arg_serialization = NULL;
1896 #ifdef HAVE_VALGRIND_VALGRIND_H
1897 /* If we are PID 1 and running under valgrind, then let's exit
1898 * here explicitly. valgrind will only generate nice output on
1899 * exit(), not on exec(), hence let's do the former not the
1901 if (getpid() == 1 && RUNNING_ON_VALGRIND)
1905 if (shutdown_verb) {
1906 char log_level[DECIMAL_STR_MAX(int) + 1];
1907 const char* command_line[9] = {
1908 SYSTEMD_SHUTDOWN_BINARY_PATH,
1910 "--log-level", log_level,
1914 _cleanup_strv_free_ char **env_block = NULL;
1916 assert(command_line[pos] == NULL);
1917 env_block = strv_copy(environ);
1919 snprintf(log_level, sizeof(log_level), "%d", log_get_max_level());
1921 switch (log_get_target()) {
1922 case LOG_TARGET_KMSG:
1923 case LOG_TARGET_JOURNAL_OR_KMSG:
1924 case LOG_TARGET_SYSLOG_OR_KMSG:
1925 command_line[pos++] = "kmsg";
1928 case LOG_TARGET_CONSOLE:
1930 command_line[pos++] = "console";
1934 if (log_get_show_color())
1935 command_line[pos++] = "--log-color";
1937 if (log_get_show_location())
1938 command_line[pos++] = "--log-location";
1940 assert(pos + 1 < ELEMENTSOF(command_line));
1942 if (arm_reboot_watchdog && arg_shutdown_watchdog > 0) {
1945 /* If we reboot let's set the shutdown
1946 * watchdog and tell the shutdown binary to
1947 * repeatedly ping it */
1948 watchdog_set_timeout(&arg_shutdown_watchdog);
1949 watchdog_close(false);
1951 /* Tell the binary how often to ping, ignore failure */
1952 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
1953 strv_push(&env_block, e);
1955 watchdog_close(true);
1957 /* Avoid the creation of new processes forked by the
1958 * kernel; at this point, we will not listen to the
1960 if (detect_container(NULL) <= 0)
1961 cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
1963 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
1964 log_error("Failed to execute shutdown binary, %s: %m",
1965 getpid() == 1 ? "freezing" : "quitting");