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 if ((r = parse_boolean(word + 18)) < 0)
307 log_warning("Failed to parse dump core switch %s. Ignoring.", word + 18);
311 } else if (startswith(word, "systemd.crash_shell=")) {
314 if ((r = parse_boolean(word + 20)) < 0)
315 log_warning("Failed to parse crash shell switch %s. Ignoring.", word + 20);
319 } else if (startswith(word, "systemd.confirm_spawn=")) {
322 if ((r = parse_boolean(word + 22)) < 0)
323 log_warning("Failed to parse confirm spawn switch %s. Ignoring.", word + 22);
325 arg_confirm_spawn = r;
327 } else if (startswith(word, "systemd.crash_chvt=")) {
330 if (safe_atoi(word + 19, &k) < 0)
331 log_warning("Failed to parse crash chvt switch %s. Ignoring.", word + 19);
335 } else if (startswith(word, "systemd.show_status=")) {
338 r = parse_show_status(word + 20, &arg_show_status);
340 log_warning("Failed to parse show status switch %s. Ignoring.", word + 20);
341 } else if (startswith(word, "systemd.default_standard_output=")) {
344 if ((r = exec_output_from_string(word + 32)) < 0)
345 log_warning("Failed to parse default standard output switch %s. Ignoring.", word + 32);
347 arg_default_std_output = r;
348 } else if (startswith(word, "systemd.default_standard_error=")) {
351 if ((r = exec_output_from_string(word + 31)) < 0)
352 log_warning("Failed to parse default standard error switch %s. Ignoring.", word + 31);
354 arg_default_std_error = r;
355 } else if (startswith(word, "systemd.setenv=")) {
356 _cleanup_free_ char *cenv = NULL;
358 cenv = strdup(word + 15);
362 if (env_assignment_is_valid(cenv)) {
365 env = strv_env_set(arg_default_environment, cenv);
367 arg_default_environment = env;
369 log_warning("Setting environment variable '%s' failed, ignoring: %m", cenv);
371 log_warning("Environment variable name '%s' is not valid. Ignoring.", cenv);
373 } else if (startswith(word, "systemd.") ||
374 (in_initrd() && startswith(word, "rd.systemd."))) {
378 /* Ignore systemd.journald.xyz and friends */
380 if (startswith(c, "rd."))
382 if (startswith(c, "systemd."))
384 if (c[strcspn(c, ".=")] != '.') {
386 log_warning("Unknown kernel switch %s. Ignoring.", word);
388 log_info("Supported kernel switches:\n"
389 "systemd.unit=UNIT Default unit to start\n"
390 "rd.systemd.unit=UNIT Default unit to start when run in initrd\n"
391 "systemd.dump_core=0|1 Dump core on crash\n"
392 "systemd.crash_shell=0|1 Run shell on crash\n"
393 "systemd.crash_chvt=N Change to VT #N on crash\n"
394 "systemd.confirm_spawn=0|1 Confirm every process spawn\n"
395 "systemd.show_status=0|1|auto Show status updates on the console during bootup\n"
396 "systemd.log_target=console|kmsg|journal|journal-or-kmsg|syslog|syslog-or-kmsg|null\n"
398 "systemd.log_level=LEVEL Log level\n"
399 "systemd.log_color=0|1 Highlight important log messages\n"
400 "systemd.log_location=0|1 Include code location in log messages\n"
401 "systemd.default_standard_output=null|tty|syslog|syslog+console|kmsg|kmsg+console|journal|journal+console\n"
402 " Set default log output for services\n"
403 "systemd.default_standard_error=null|tty|syslog|syslog+console|kmsg|kmsg+console|journal|journal+console\n"
404 " Set default log error output for services\n"
405 "systemd.setenv=ASSIGNMENT Set an environment variable for all spawned processes\n");
408 } else if (streq(word, "quiet")) {
409 if (arg_show_status == SHOW_STATUS_UNSET)
410 arg_show_status = SHOW_STATUS_AUTO;
411 } else if (streq(word, "debug")) {
412 /* Log to kmsg, the journal socket will fill up before the
413 * journal is started and tools running during that time
414 * will block with every log message for for 60 seconds,
415 * before they give up. */
416 log_set_max_level(LOG_DEBUG);
417 log_set_target(detect_container(NULL) > 0 ? LOG_TARGET_CONSOLE : LOG_TARGET_KMSG);
418 } else if (!in_initrd()) {
421 /* SysV compatibility */
422 for (i = 0; i < ELEMENTSOF(rlmap); i += 2)
423 if (streq(word, rlmap[i]))
424 return set_default_unit(rlmap[i+1]);
430 #define DEFINE_SETTER(name, func, descr) \
431 static int name(const char *unit, \
432 const char *filename, \
434 const char *section, \
435 unsigned section_line, \
436 const char *lvalue, \
438 const char *rvalue, \
450 log_syntax(unit, LOG_ERR, filename, line, -r, \
451 "Invalid " descr "'%s': %s", \
452 rvalue, strerror(-r)); \
457 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level")
458 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target")
459 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" )
460 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location")
462 static int config_parse_cpu_affinity2(const char *unit,
463 const char *filename,
466 unsigned section_line,
483 FOREACH_WORD_QUOTED(w, l, rvalue, state) {
488 if (!(t = strndup(w, l)))
491 r = safe_atou(t, &cpu);
495 if (!(c = cpu_set_malloc(&ncpus)))
498 if (r < 0 || cpu >= ncpus) {
499 log_syntax(unit, LOG_ERR, filename, line, -r,
500 "Failed to parse CPU affinity '%s'", rvalue);
505 CPU_SET_S(cpu, CPU_ALLOC_SIZE(ncpus), c);
509 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
510 log_warning_unit(unit, "Failed to set CPU affinity: %m");
518 static void strv_free_free(char ***l) {
530 static void free_join_controllers(void) {
531 strv_free_free(arg_join_controllers);
532 arg_join_controllers = NULL;
535 static int config_parse_join_controllers(const char *unit,
536 const char *filename,
539 unsigned section_line,
554 free_join_controllers();
556 FOREACH_WORD_QUOTED(w, length, rvalue, state) {
559 s = strndup(w, length);
563 l = strv_split(s, ",");
568 if (strv_length(l) <= 1) {
573 if (!arg_join_controllers) {
574 arg_join_controllers = new(char**, 2);
575 if (!arg_join_controllers) {
580 arg_join_controllers[0] = l;
581 arg_join_controllers[1] = NULL;
588 t = new0(char**, n+2);
596 for (a = arg_join_controllers; *a; a++) {
598 if (strv_overlap(*a, l)) {
599 if (strv_extend_strv(&l, *a) < 0) {
619 t[n++] = strv_uniq(l);
621 strv_free_free(arg_join_controllers);
622 arg_join_controllers = t;
629 static int parse_config_file(void) {
631 const ConfigTableItem items[] = {
632 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
633 { "Manager", "LogTarget", config_parse_target, 0, NULL },
634 { "Manager", "LogColor", config_parse_color, 0, NULL },
635 { "Manager", "LogLocation", config_parse_location, 0, NULL },
636 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
637 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
638 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
639 { "Manager", "CrashChVT", config_parse_int, 0, &arg_crash_chvt },
640 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
641 { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
642 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
643 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
644 { "Manager", "CapabilityBoundingSet", config_parse_bounding_set, 0, &arg_capability_bounding_set_drop },
646 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
648 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
649 { "Manager", "DefaultStandardOutput", config_parse_output, 0, &arg_default_std_output },
650 { "Manager", "DefaultStandardError", config_parse_output, 0, &arg_default_std_error },
651 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
652 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
653 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
654 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval },
655 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
656 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
657 { "Manager", "DefaultLimitCPU", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CPU] },
658 { "Manager", "DefaultLimitFSIZE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_FSIZE] },
659 { "Manager", "DefaultLimitDATA", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_DATA] },
660 { "Manager", "DefaultLimitSTACK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_STACK] },
661 { "Manager", "DefaultLimitCORE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CORE] },
662 { "Manager", "DefaultLimitRSS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RSS] },
663 { "Manager", "DefaultLimitNOFILE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NOFILE] },
664 { "Manager", "DefaultLimitAS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_AS] },
665 { "Manager", "DefaultLimitNPROC", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NPROC] },
666 { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MEMLOCK] },
667 { "Manager", "DefaultLimitLOCKS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_LOCKS] },
668 { "Manager", "DefaultLimitSIGPENDING", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_SIGPENDING] },
669 { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MSGQUEUE] },
670 { "Manager", "DefaultLimitNICE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NICE] },
671 { "Manager", "DefaultLimitRTPRIO", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTPRIO] },
672 { "Manager", "DefaultLimitRTTIME", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTTIME] },
676 _cleanup_fclose_ FILE *f;
680 fn = arg_running_as == SYSTEMD_SYSTEM ? PKGSYSCONFDIR "/system.conf" : PKGSYSCONFDIR "/user.conf";
686 log_warning("Failed to open configuration file '%s': %m", fn);
690 r = config_parse(NULL, fn, f, "Manager\0", config_item_table_lookup, (void*) items, false, false, NULL);
692 log_warning("Failed to parse configuration file: %s", strerror(-r));
697 static int parse_argv(int argc, char *argv[]) {
700 ARG_LOG_LEVEL = 0x100,
709 ARG_DUMP_CONFIGURATION_ITEMS,
716 ARG_DEFAULT_STD_OUTPUT,
717 ARG_DEFAULT_STD_ERROR
720 static const struct option options[] = {
721 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
722 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
723 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
724 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
725 { "unit", required_argument, NULL, ARG_UNIT },
726 { "system", no_argument, NULL, ARG_SYSTEM },
727 { "user", no_argument, NULL, ARG_USER },
728 { "test", no_argument, NULL, ARG_TEST },
729 { "help", no_argument, NULL, 'h' },
730 { "version", no_argument, NULL, ARG_VERSION },
731 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
732 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
733 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
734 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
735 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
736 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
737 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
738 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
739 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
751 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
756 if ((r = log_set_max_level_from_string(optarg)) < 0) {
757 log_error("Failed to parse log level %s.", optarg);
765 if ((r = log_set_target_from_string(optarg)) < 0) {
766 log_error("Failed to parse log target %s.", optarg);
775 if ((r = log_show_color_from_string(optarg)) < 0) {
776 log_error("Failed to parse log color setting %s.", optarg);
780 log_show_color(true);
784 case ARG_LOG_LOCATION:
787 if ((r = log_show_location_from_string(optarg)) < 0) {
788 log_error("Failed to parse log location setting %s.", optarg);
792 log_show_location(true);
796 case ARG_DEFAULT_STD_OUTPUT:
798 if ((r = exec_output_from_string(optarg)) < 0) {
799 log_error("Failed to parse default standard output setting %s.", optarg);
802 arg_default_std_output = r;
805 case ARG_DEFAULT_STD_ERROR:
807 if ((r = exec_output_from_string(optarg)) < 0) {
808 log_error("Failed to parse default standard error output setting %s.", optarg);
811 arg_default_std_error = r;
816 if ((r = set_default_unit(optarg)) < 0) {
817 log_error("Failed to set default unit %s: %s", optarg, strerror(-r));
824 arg_running_as = SYSTEMD_SYSTEM;
828 arg_running_as = SYSTEMD_USER;
832 arg_action = ACTION_TEST;
836 arg_action = ACTION_VERSION;
839 case ARG_DUMP_CONFIGURATION_ITEMS:
840 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
844 r = optarg ? parse_boolean(optarg) : 1;
846 log_error("Failed to parse dump core boolean %s.", optarg);
852 case ARG_CRASH_SHELL:
853 r = optarg ? parse_boolean(optarg) : 1;
855 log_error("Failed to parse crash shell boolean %s.", optarg);
861 case ARG_CONFIRM_SPAWN:
862 r = optarg ? parse_boolean(optarg) : 1;
864 log_error("Failed to parse confirm spawn boolean %s.", optarg);
867 arg_confirm_spawn = r;
870 case ARG_SHOW_STATUS:
872 r = parse_show_status(optarg, &arg_show_status);
874 log_error("Failed to parse show status boolean %s.", optarg);
878 arg_show_status = SHOW_STATUS_YES;
881 case ARG_DESERIALIZE: {
885 r = safe_atoi(optarg, &fd);
886 if (r < 0 || fd < 0) {
887 log_error("Failed to parse deserialize option %s.", optarg);
888 return r < 0 ? r : -EINVAL;
891 fd_cloexec(fd, true);
895 log_error("Failed to open serialization fd: %m");
899 if (arg_serialization)
900 fclose(arg_serialization);
902 arg_serialization = f;
907 case ARG_SWITCHED_ROOT:
908 arg_switched_root = true;
912 arg_action = ACTION_HELP;
916 log_set_max_level(LOG_DEBUG);
922 /* Just to eat away the sysvinit kernel
923 * cmdline args without getopt() error
924 * messages that we'll parse in
925 * parse_proc_cmdline_word() or ignore. */
930 log_error("Unknown option code %c", c);
937 if (optind < argc && getpid() != 1) {
938 /* Hmm, when we aren't run as init system
939 * let's complain about excess arguments */
941 log_error("Excess arguments.");
945 if (detect_container(NULL) > 0) {
948 /* All /proc/cmdline arguments the kernel didn't
949 * understand it passed to us. We're not really
950 * interested in that usually since /proc/cmdline is
951 * more interesting and complete. With one exception:
952 * if we are run in a container /proc/cmdline is not
953 * relevant for the container, hence we rely on argv[]
956 for (a = argv; a < argv + argc; a++) {
957 r = parse_proc_cmdline_word(*a);
959 log_error("Failed on cmdline argument %s: %s", *a, strerror(-r));
968 static int help(void) {
970 printf("%s [OPTIONS...]\n\n"
971 "Starts up and maintains the system or user services.\n\n"
972 " -h --help Show this help\n"
973 " --test Determine startup sequence, dump it and exit\n"
974 " --dump-configuration-items Dump understood unit configuration items\n"
975 " --unit=UNIT Set default unit\n"
976 " --system Run a system instance, even if PID != 1\n"
977 " --user Run a user instance\n"
978 " --dump-core[=0|1] Dump core on crash\n"
979 " --crash-shell[=0|1] Run shell on crash\n"
980 " --confirm-spawn[=0|1] Ask for confirmation when spawning processes\n"
981 " --show-status[=0|1] Show status updates on the console during bootup\n"
982 " --log-target=TARGET Set log target (console, journal, syslog, kmsg, journal-or-kmsg, syslog-or-kmsg, null)\n"
983 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
984 " --log-color[=0|1] Highlight important log messages\n"
985 " --log-location[=0|1] Include code location in log messages\n"
986 " --default-standard-output= Set default standard output for services\n"
987 " --default-standard-error= Set default standard error output for services\n",
988 program_invocation_short_name);
993 static int version(void) {
994 puts(PACKAGE_STRING);
995 puts(SYSTEMD_FEATURES);
1000 static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
1009 r = manager_open_serialization(m, &f);
1011 log_error("Failed to create serialization file: %s", strerror(-r));
1015 /* Make sure nothing is really destructed when we shut down */
1017 bus_manager_send_reloading(m, true);
1022 log_error("Failed to allocate fd set: %s", strerror(-r));
1026 r = manager_serialize(m, f, fds, switching_root);
1028 log_error("Failed to serialize state: %s", strerror(-r));
1032 if (fseeko(f, 0, SEEK_SET) < 0) {
1033 log_error("Failed to rewind serialization fd: %m");
1037 r = fd_cloexec(fileno(f), false);
1039 log_error("Failed to disable O_CLOEXEC for serialization: %s", strerror(-r));
1043 r = fdset_cloexec(fds, false);
1045 log_error("Failed to disable O_CLOEXEC for serialization fds: %s", strerror(-r));
1063 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1067 assert(saved_rlimit);
1069 /* Save the original RLIMIT_NOFILE so that we can reset it
1070 * later when transitioning from the initrd to the main
1071 * systemd or suchlike. */
1072 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0) {
1073 log_error("Reading RLIMIT_NOFILE failed: %m");
1077 /* Make sure forked processes get the default kernel setting */
1078 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1081 rl = newdup(struct rlimit, saved_rlimit, 1);
1085 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1088 /* Bump up the resource limit for ourselves substantially */
1089 nl.rlim_cur = nl.rlim_max = 64*1024;
1090 r = setrlimit_closest(RLIMIT_NOFILE, &nl);
1092 log_error("Setting RLIMIT_NOFILE failed: %s", strerror(-r));
1099 static void test_mtab(void) {
1102 /* Check that /etc/mtab is a symlink */
1104 if (readlink_malloc("/etc/mtab", &p) >= 0) {
1107 b = streq(p, "/proc/self/mounts") || streq(p, "/proc/mounts");
1114 log_warning("/etc/mtab is not a symlink or not pointing to /proc/self/mounts. "
1115 "This is not supported anymore. "
1116 "Please make sure to replace this file by a symlink to avoid incorrect or misleading mount(8) output.");
1119 static void test_usr(void) {
1121 /* Check that /usr is not a separate fs */
1123 if (dir_is_empty("/usr") <= 0)
1126 log_warning("/usr appears to be on its own filesytem and is not already mounted. This is not a supported setup. "
1127 "Some things will probably break (sometimes even silently) in mysterious ways. "
1128 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1131 static void test_cgroups(void) {
1133 if (access("/proc/cgroups", F_OK) >= 0)
1136 log_warning("CONFIG_CGROUPS was not set when your kernel was compiled. "
1137 "Systems without control groups are not supported. "
1138 "We will now sleep for 10s, and then continue boot-up. "
1139 "Expect breakage and please do not file bugs. "
1140 "Instead fix your kernel and enable CONFIG_CGROUPS. "
1141 "Consult http://0pointer.de/blog/projects/cgroups-vs-cgroups.html for more information.");
1146 static int initialize_join_controllers(void) {
1147 /* By default, mount "cpu" + "cpuacct" together, and "net_cls"
1148 * + "net_prio". We'd like to add "cpuset" to the mix, but
1149 * "cpuset" does't really work for groups with no initialized
1152 arg_join_controllers = new(char**, 3);
1153 if (!arg_join_controllers)
1156 arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL);
1157 arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL);
1158 arg_join_controllers[2] = NULL;
1160 if (!arg_join_controllers[0] || !arg_join_controllers[1]) {
1161 free_join_controllers();
1168 static int enforce_syscall_archs(Set *archs) {
1170 scmp_filter_ctx *seccomp;
1175 seccomp = seccomp_init(SCMP_ACT_ALLOW);
1179 SET_FOREACH(id, arg_syscall_archs, i) {
1180 r = seccomp_arch_add(seccomp, PTR_TO_UINT32(id) - 1);
1184 log_error("Failed to add architecture to seccomp: %s", strerror(-r));
1189 r = seccomp_load(seccomp);
1191 log_error("Failed to add install architecture seccomp: %s", strerror(-r));
1194 seccomp_release(seccomp);
1201 int main(int argc, char *argv[]) {
1203 int r, retval = EXIT_FAILURE;
1204 usec_t before_startup, after_startup;
1205 char timespan[FORMAT_TIMESPAN_MAX];
1207 bool reexecute = false;
1208 const char *shutdown_verb = NULL;
1209 dual_timestamp initrd_timestamp = { 0ULL, 0ULL };
1210 dual_timestamp userspace_timestamp = { 0ULL, 0ULL };
1211 dual_timestamp kernel_timestamp = { 0ULL, 0ULL };
1212 dual_timestamp security_start_timestamp = { 0ULL, 0ULL };
1213 dual_timestamp security_finish_timestamp = { 0ULL, 0ULL };
1214 static char systemd[] = "systemd";
1215 bool skip_setup = false;
1217 bool loaded_policy = false;
1218 bool arm_reboot_watchdog = false;
1219 bool queue_default_job = false;
1220 char *switch_root_dir = NULL, *switch_root_init = NULL;
1221 static struct rlimit saved_rlimit_nofile = { 0, 0 };
1223 #ifdef HAVE_SYSV_COMPAT
1224 if (getpid() != 1 && strstr(program_invocation_short_name, "init")) {
1225 /* This is compatibility support for SysV, where
1226 * calling init as a user is identical to telinit. */
1229 execv(SYSTEMCTL_BINARY_PATH, argv);
1230 log_error("Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1235 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
1236 dual_timestamp_get(&userspace_timestamp);
1238 /* Determine if this is a reexecution or normal bootup. We do
1239 * the full command line parsing much later, so let's just
1240 * have a quick peek here. */
1241 if (strv_find(argv+1, "--deserialize"))
1244 /* If we have switched root, do all the special setup
1246 if (strv_find(argv+1, "--switched-root"))
1249 /* If we get started via the /sbin/init symlink then we are
1250 called 'init'. After a subsequent reexecution we are then
1251 called 'systemd'. That is confusing, hence let's call us
1252 systemd right-away. */
1253 program_invocation_short_name = systemd;
1254 prctl(PR_SET_NAME, systemd);
1259 log_show_color(isatty(STDERR_FILENO) > 0);
1261 /* Disable the umask logic */
1265 if (getpid() == 1 && detect_container(NULL) <= 0) {
1267 /* Running outside of a container as PID 1 */
1268 arg_running_as = SYSTEMD_SYSTEM;
1270 log_set_target(LOG_TARGET_KMSG);
1274 initrd_timestamp = userspace_timestamp;
1277 mount_setup_early();
1278 dual_timestamp_get(&security_start_timestamp);
1279 if (selinux_setup(&loaded_policy) < 0)
1281 if (ima_setup() < 0)
1283 if (smack_setup() < 0)
1285 dual_timestamp_get(&security_finish_timestamp);
1288 if (label_init(NULL) < 0)
1292 if (hwclock_is_localtime() > 0) {
1295 /* The first-time call to settimeofday() does a time warp in the kernel */
1296 r = hwclock_set_timezone(&min);
1298 log_error("Failed to apply local time delta, ignoring: %s", strerror(-r));
1300 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1301 } else if (!in_initrd()) {
1303 * Do dummy first-time call to seal the kernel's time warp magic
1305 * Do not call this this from inside the initrd. The initrd might not
1306 * carry /etc/adjtime with LOCAL, but the real system could be set up
1307 * that way. In such case, we need to delay the time-warp or the sealing
1308 * until we reach the real system.
1310 hwclock_reset_timezone();
1312 /* Tell the kernel our timezone */
1313 r = hwclock_set_timezone(NULL);
1315 log_error("Failed to set the kernel's timezone, ignoring: %s", strerror(-r));
1319 /* Set the default for later on, but don't actually
1320 * open the logs like this for now. Note that if we
1321 * are transitioning from the initrd there might still
1322 * be journal fd open, and we shouldn't attempt
1323 * opening that before we parsed /proc/cmdline which
1324 * might redirect output elsewhere. */
1325 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
1327 } else if (getpid() == 1) {
1328 /* Running inside a container, as PID 1 */
1329 arg_running_as = SYSTEMD_SYSTEM;
1330 log_set_target(LOG_TARGET_CONSOLE);
1331 log_close_console(); /* force reopen of /dev/console */
1334 /* For the later on, see above... */
1335 log_set_target(LOG_TARGET_JOURNAL);
1337 /* clear the kernel timestamp,
1338 * because we are in a container */
1339 kernel_timestamp.monotonic = 0ULL;
1340 kernel_timestamp.realtime = 0ULL;
1343 /* Running as user instance */
1344 arg_running_as = SYSTEMD_USER;
1345 log_set_target(LOG_TARGET_AUTO);
1348 /* clear the kernel timestamp,
1349 * because we are not PID 1 */
1350 kernel_timestamp.monotonic = 0ULL;
1351 kernel_timestamp.realtime = 0ULL;
1354 /* Initialize default unit */
1355 r = set_default_unit(SPECIAL_DEFAULT_TARGET);
1357 log_error("Failed to set default unit %s: %s", SPECIAL_DEFAULT_TARGET, strerror(-r));
1361 r = initialize_join_controllers();
1365 /* Mount /proc, /sys and friends, so that /proc/cmdline and
1366 * /proc/$PID/fd is available. */
1367 if (getpid() == 1) {
1368 r = mount_setup(loaded_policy);
1373 /* Reset all signal handlers. */
1374 assert_se(reset_all_signal_handlers() == 0);
1376 ignore_signals(SIGNALS_IGNORE, -1);
1378 if (parse_config_file() < 0)
1381 if (arg_running_as == SYSTEMD_SYSTEM)
1382 if (parse_proc_cmdline(parse_proc_cmdline_word) < 0)
1385 log_parse_environment();
1387 if (parse_argv(argc, argv) < 0)
1390 if (arg_action == ACTION_TEST &&
1392 log_error("Don't run test mode as root.");
1396 if (arg_running_as == SYSTEMD_USER &&
1397 arg_action == ACTION_RUN &&
1399 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
1403 if (arg_running_as == SYSTEMD_SYSTEM &&
1404 arg_action == ACTION_RUN &&
1405 running_in_chroot() > 0) {
1406 log_error("Cannot be run in a chroot() environment.");
1410 if (arg_action == ACTION_HELP) {
1413 } else if (arg_action == ACTION_VERSION) {
1416 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
1417 unit_dump_config_items(stdout);
1418 retval = EXIT_SUCCESS;
1420 } else if (arg_action == ACTION_DONE) {
1421 retval = EXIT_SUCCESS;
1425 if (arg_running_as == SYSTEMD_USER &&
1426 !getenv("XDG_RUNTIME_DIR")) {
1427 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
1431 assert_se(arg_action == ACTION_RUN || arg_action == ACTION_TEST);
1433 /* Close logging fds, in order not to confuse fdset below */
1436 /* Remember open file descriptors for later deserialization */
1437 r = fdset_new_fill(&fds);
1439 log_error("Failed to allocate fd set: %s", strerror(-r));
1442 fdset_cloexec(fds, true);
1444 if (arg_serialization)
1445 assert_se(fdset_remove(fds, fileno(arg_serialization)) >= 0);
1447 if (arg_running_as == SYSTEMD_SYSTEM)
1448 /* Become a session leader if we aren't one yet. */
1451 /* Move out of the way, so that we won't block unmounts */
1452 assert_se(chdir("/") == 0);
1454 /* Reset the console, but only if this is really init and we
1455 * are freshly booted */
1456 if (arg_running_as == SYSTEMD_SYSTEM && arg_action == ACTION_RUN)
1457 console_setup(getpid() == 1 && !skip_setup);
1459 /* Open the logging devices, if possible and necessary */
1462 /* Make sure we leave a core dump without panicing the
1464 if (getpid() == 1) {
1465 install_crash_handler();
1467 r = mount_cgroup_controllers(arg_join_controllers);
1472 if (arg_running_as == SYSTEMD_SYSTEM) {
1473 const char *virtualization = NULL;
1475 log_info(PACKAGE_STRING " running in system mode. (" SYSTEMD_FEATURES ")");
1477 detect_virtualization(&virtualization);
1479 log_info("Detected virtualization '%s'.", virtualization);
1482 log_info("Running in initial RAM disk.");
1485 _cleanup_free_ char *t = uid_to_name(getuid());
1486 log_debug(PACKAGE_STRING " running in user mode for user "PID_FMT"/%s. (" SYSTEMD_FEATURES ")",
1490 if (arg_running_as == SYSTEMD_SYSTEM && !skip_setup) {
1491 if (arg_show_status > 0 || plymouth_running())
1495 if (detect_container(NULL) <= 0)
1507 if (arg_running_as == SYSTEMD_SYSTEM && arg_runtime_watchdog > 0)
1508 watchdog_set_timeout(&arg_runtime_watchdog);
1510 if (arg_timer_slack_nsec != (nsec_t) -1)
1511 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1512 log_error("Failed to adjust timer slack: %m");
1514 if (arg_capability_bounding_set_drop) {
1515 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set_drop);
1517 log_error("Failed to drop capability bounding set of usermode helpers: %s", strerror(-r));
1520 r = capability_bounding_set_drop(arg_capability_bounding_set_drop, true);
1522 log_error("Failed to drop capability bounding set: %s", strerror(-r));
1527 if (arg_syscall_archs) {
1528 r = enforce_syscall_archs(arg_syscall_archs);
1533 if (arg_running_as == SYSTEMD_USER) {
1534 /* Become reaper of our children */
1535 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0) {
1536 log_warning("Failed to make us a subreaper: %m");
1537 if (errno == EINVAL)
1538 log_info("Perhaps the kernel version is too old (< 3.4?)");
1542 if (arg_running_as == SYSTEMD_SYSTEM)
1543 bump_rlimit_nofile(&saved_rlimit_nofile);
1545 r = manager_new(arg_running_as, &m);
1547 log_error("Failed to allocate manager object: %s", strerror(-r));
1551 m->confirm_spawn = arg_confirm_spawn;
1552 m->default_std_output = arg_default_std_output;
1553 m->default_std_error = arg_default_std_error;
1554 m->default_restart_usec = arg_default_restart_usec;
1555 m->default_timeout_start_usec = arg_default_timeout_start_usec;
1556 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
1557 m->default_start_limit_interval = arg_default_start_limit_interval;
1558 m->default_start_limit_burst = arg_default_start_limit_burst;
1559 m->runtime_watchdog = arg_runtime_watchdog;
1560 m->shutdown_watchdog = arg_shutdown_watchdog;
1561 m->userspace_timestamp = userspace_timestamp;
1562 m->kernel_timestamp = kernel_timestamp;
1563 m->initrd_timestamp = initrd_timestamp;
1564 m->security_start_timestamp = security_start_timestamp;
1565 m->security_finish_timestamp = security_finish_timestamp;
1567 manager_set_default_rlimits(m, arg_default_rlimit);
1569 if (arg_default_environment)
1570 manager_environment_add(m, NULL, arg_default_environment);
1572 if (arg_show_status == SHOW_STATUS_UNSET)
1573 arg_show_status = SHOW_STATUS_YES;
1574 manager_set_show_status(m, arg_show_status);
1576 /* Remember whether we should queue the default job */
1577 queue_default_job = !arg_serialization || arg_switched_root;
1579 before_startup = now(CLOCK_MONOTONIC);
1581 r = manager_startup(m, arg_serialization, fds);
1583 log_error("Failed to fully start up daemon: %s", strerror(-r));
1585 /* This will close all file descriptors that were opened, but
1586 * not claimed by any unit. */
1590 if (arg_serialization) {
1591 fclose(arg_serialization);
1592 arg_serialization = NULL;
1595 if (queue_default_job) {
1596 _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL;
1597 Unit *target = NULL;
1598 Job *default_unit_job;
1600 log_debug("Activating default unit: %s", arg_default_unit);
1602 r = manager_load_unit(m, arg_default_unit, NULL, &error, &target);
1604 log_error("Failed to load default target: %s", bus_error_message(&error, r));
1605 else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND)
1606 log_error("Failed to load default target: %s", strerror(-target->load_error));
1607 else if (target->load_state == UNIT_MASKED)
1608 log_error("Default target masked.");
1610 if (!target || target->load_state != UNIT_LOADED) {
1611 log_info("Trying to load rescue target...");
1613 r = manager_load_unit(m, SPECIAL_RESCUE_TARGET, NULL, &error, &target);
1615 log_error("Failed to load rescue target: %s", bus_error_message(&error, r));
1617 } else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND) {
1618 log_error("Failed to load rescue target: %s", strerror(-target->load_error));
1620 } else if (target->load_state == UNIT_MASKED) {
1621 log_error("Rescue target masked.");
1626 assert(target->load_state == UNIT_LOADED);
1628 if (arg_action == ACTION_TEST) {
1629 printf("-> By units:\n");
1630 manager_dump_units(m, stdout, "\t");
1633 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, false, &error, &default_unit_job);
1635 log_debug("Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1637 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, false, &error, &default_unit_job);
1639 log_error("Failed to start default target: %s", bus_error_message(&error, r));
1643 log_error("Failed to isolate default target: %s", bus_error_message(&error, r));
1647 m->default_unit_job_id = default_unit_job->id;
1649 after_startup = now(CLOCK_MONOTONIC);
1650 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
1651 "Loaded units and determined initial transaction in %s.",
1652 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 0));
1654 if (arg_action == ACTION_TEST) {
1655 printf("-> By jobs:\n");
1656 manager_dump_jobs(m, stdout, "\t");
1657 retval = EXIT_SUCCESS;
1663 r = manager_loop(m);
1665 log_error("Failed to run mainloop: %s", strerror(-r));
1669 switch (m->exit_code) {
1672 retval = EXIT_SUCCESS;
1676 case MANAGER_RELOAD:
1677 log_info("Reloading.");
1678 r = manager_reload(m);
1680 log_error("Failed to reload: %s", strerror(-r));
1683 case MANAGER_REEXECUTE:
1685 if (prepare_reexecute(m, &arg_serialization, &fds, false) < 0)
1689 log_notice("Reexecuting.");
1692 case MANAGER_SWITCH_ROOT:
1693 /* Steal the switch root parameters */
1694 switch_root_dir = m->switch_root;
1695 switch_root_init = m->switch_root_init;
1696 m->switch_root = m->switch_root_init = NULL;
1698 if (!switch_root_init)
1699 if (prepare_reexecute(m, &arg_serialization, &fds, true) < 0)
1703 log_notice("Switching root.");
1706 case MANAGER_REBOOT:
1707 case MANAGER_POWEROFF:
1709 case MANAGER_KEXEC: {
1710 static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
1711 [MANAGER_REBOOT] = "reboot",
1712 [MANAGER_POWEROFF] = "poweroff",
1713 [MANAGER_HALT] = "halt",
1714 [MANAGER_KEXEC] = "kexec"
1717 assert_se(shutdown_verb = table[m->exit_code]);
1718 arm_reboot_watchdog = m->exit_code == MANAGER_REBOOT;
1720 log_notice("Shutting down.");
1725 assert_not_reached("Unknown exit code.");
1735 for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++) {
1736 free(arg_default_rlimit[j]);
1737 arg_default_rlimit[j] = NULL;
1740 free(arg_default_unit);
1741 arg_default_unit = NULL;
1743 free_join_controllers();
1745 strv_free(arg_default_environment);
1746 arg_default_environment = NULL;
1748 set_free(arg_syscall_archs);
1749 arg_syscall_archs = NULL;
1755 unsigned i, args_size;
1757 /* Close and disarm the watchdog, so that the new
1758 * instance can reinitialize it, but doesn't get
1759 * rebooted while we do that */
1760 watchdog_close(true);
1762 /* Reset the RLIMIT_NOFILE to the kernel default, so
1763 * that the new systemd can pass the kernel default to
1764 * its child processes */
1765 if (saved_rlimit_nofile.rlim_cur > 0)
1766 setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile);
1768 if (switch_root_dir) {
1769 /* Kill all remaining processes from the
1770 * initrd, but don't wait for them, so that we
1771 * can handle the SIGCHLD for them after
1773 broadcast_signal(SIGTERM, false, true);
1775 /* And switch root */
1776 r = switch_root(switch_root_dir);
1778 log_error("Failed to switch root, ignoring: %s", strerror(-r));
1781 args_size = MAX(6, argc+1);
1782 args = newa(const char*, args_size);
1784 if (!switch_root_init) {
1787 /* First try to spawn ourselves with the right
1788 * path, and with full serialization. We do
1789 * this only if the user didn't specify an
1790 * explicit init to spawn. */
1792 assert(arg_serialization);
1795 snprintf(sfd, sizeof(sfd), "%i", fileno(arg_serialization));
1799 args[i++] = SYSTEMD_BINARY_PATH;
1800 if (switch_root_dir)
1801 args[i++] = "--switched-root";
1802 args[i++] = arg_running_as == SYSTEMD_SYSTEM ? "--system" : "--user";
1803 args[i++] = "--deserialize";
1807 /* do not pass along the environment we inherit from the kernel or initrd */
1808 if (switch_root_dir)
1811 assert(i <= args_size);
1812 execv(args[0], (char* const*) args);
1815 /* Try the fallback, if there is any, without any
1816 * serialization. We pass the original argv[] and
1817 * envp[]. (Well, modulo the ordering changes due to
1818 * getopt() in argv[], and some cleanups in envp[],
1819 * but let's hope that doesn't matter.) */
1821 if (arg_serialization) {
1822 fclose(arg_serialization);
1823 arg_serialization = NULL;
1831 /* Reopen the console */
1832 make_console_stdio();
1834 for (j = 1, i = 1; j < (unsigned) argc; j++)
1835 args[i++] = argv[j];
1837 assert(i <= args_size);
1839 if (switch_root_init) {
1840 args[0] = switch_root_init;
1841 execv(args[0], (char* const*) args);
1842 log_warning("Failed to execute configured init, trying fallback: %m");
1845 args[0] = "/sbin/init";
1846 execv(args[0], (char* const*) args);
1848 if (errno == ENOENT) {
1849 log_warning("No /sbin/init, trying fallback");
1851 args[0] = "/bin/sh";
1853 execv(args[0], (char* const*) args);
1854 log_error("Failed to execute /bin/sh, giving up: %m");
1856 log_warning("Failed to execute /sbin/init, giving up: %m");
1859 if (arg_serialization) {
1860 fclose(arg_serialization);
1861 arg_serialization = NULL;
1869 #ifdef HAVE_VALGRIND_VALGRIND_H
1870 /* If we are PID 1 and running under valgrind, then let's exit
1871 * here explicitly. valgrind will only generate nice output on
1872 * exit(), not on exec(), hence let's do the former not the
1874 if (getpid() == 1 && RUNNING_ON_VALGRIND)
1878 if (shutdown_verb) {
1879 const char * command_line[] = {
1880 SYSTEMD_SHUTDOWN_BINARY_PATH,
1884 _cleanup_strv_free_ char **env_block = NULL;
1885 env_block = strv_copy(environ);
1887 if (arm_reboot_watchdog && arg_shutdown_watchdog > 0) {
1890 /* If we reboot let's set the shutdown
1891 * watchdog and tell the shutdown binary to
1892 * repeatedly ping it */
1893 watchdog_set_timeout(&arg_shutdown_watchdog);
1894 watchdog_close(false);
1896 /* Tell the binary how often to ping, ignore failure */
1897 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
1898 strv_push(&env_block, e);
1900 watchdog_close(true);
1902 /* Avoid the creation of new processes forked by the
1903 * kernel; at this point, we will not listen to the
1905 if (detect_container(NULL) <= 0)
1906 cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
1908 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
1909 log_error("Failed to execute shutdown binary, freezing: %m");