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>
42 #include "sd-daemon.h"
43 #include "sd-messages.h"
47 #include "load-fragment.h"
50 #include "conf-parser.h"
58 #include "architecture.h"
60 #include "path-util.h"
61 #include "switch-root.h"
62 #include "capability.h"
65 #include "clock-util.h"
67 #include "dbus-manager.h"
68 #include "bus-error.h"
71 #include "mount-setup.h"
72 #include "loopback-setup.h"
73 #include "hostname-setup.h"
74 #include "machine-id-setup.h"
75 #include "selinux-setup.h"
76 #include "ima-setup.h"
77 #include "smack-setup.h"
79 #include "kmod-setup.h"
87 ACTION_DUMP_CONFIGURATION_ITEMS,
89 } arg_action = ACTION_RUN;
90 static char *arg_default_unit = NULL;
91 static SystemdRunningAs arg_running_as = _SYSTEMD_RUNNING_AS_INVALID;
92 static bool arg_dump_core = true;
93 static bool arg_crash_shell = false;
94 static int arg_crash_chvt = -1;
95 static bool arg_confirm_spawn = false;
96 static ShowStatus arg_show_status = _SHOW_STATUS_UNSET;
97 static bool arg_switched_root = false;
98 static int arg_no_pager = -1;
99 static char ***arg_join_controllers = NULL;
100 static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
101 static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
102 static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
103 static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
104 static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
105 static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
106 static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
107 static usec_t arg_runtime_watchdog = 0;
108 static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
109 static char **arg_default_environment = NULL;
110 static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {};
111 static uint64_t arg_capability_bounding_set_drop = 0;
112 static nsec_t arg_timer_slack_nsec = NSEC_INFINITY;
113 static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
114 static Set* arg_syscall_archs = NULL;
115 static FILE* arg_serialization = NULL;
116 static bool arg_default_cpu_accounting = false;
117 static bool arg_default_blockio_accounting = false;
118 static bool arg_default_memory_accounting = false;
120 static void nop_handler(int sig) {}
122 static void pager_open_if_enabled(void) {
124 if (arg_no_pager <= 0)
130 noreturn static void crash(int sig) {
133 /* Pass this on immediately, if this is not PID 1 */
135 else if (!arg_dump_core)
136 log_error("Caught <%s>, not dumping core.", signal_to_string(sig));
138 struct sigaction sa = {
139 .sa_handler = nop_handler,
140 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
144 /* We want to wait for the core process, hence let's enable SIGCHLD */
145 sigaction(SIGCHLD, &sa, NULL);
149 log_error("Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
152 struct rlimit rl = {};
154 /* Enable default signal handler for core dump */
156 sa.sa_handler = SIG_DFL;
157 sigaction(sig, &sa, NULL);
159 /* Don't limit the core dump size */
160 rl.rlim_cur = RLIM_INFINITY;
161 rl.rlim_max = RLIM_INFINITY;
162 setrlimit(RLIMIT_CORE, &rl);
164 /* Just to be sure... */
167 /* Raise the signal again */
170 assert_not_reached("We shouldn't be here...");
177 /* Order things nicely. */
178 r = wait_for_terminate(pid, &status);
180 log_error("Caught <%s>, waitpid() failed: %s", signal_to_string(sig), strerror(-r));
181 else if (status.si_code != CLD_DUMPED)
182 log_error("Caught <%s>, core dump failed.", signal_to_string(sig));
184 log_error("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);
206 log_error("Failed to fork off crash shell: %m");
208 make_console_stdio();
209 execl("/bin/sh", "/bin/sh", NULL);
211 log_error("execl() failed: %m");
215 log_info("Successfully spawned crash shell as pid "PID_FMT".", pid);
218 log_info("Freezing execution.");
222 static void install_crash_handler(void) {
223 struct sigaction sa = {
225 .sa_flags = SA_NODEFER,
228 sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
231 static int console_setup(void) {
232 _cleanup_close_ int tty_fd = -1;
235 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
237 log_error("Failed to open /dev/console: %s", strerror(-tty_fd));
241 /* We don't want to force text mode. plymouth may be showing
242 * pictures already from initrd. */
243 r = reset_terminal_fd(tty_fd, false);
245 log_error("Failed to reset /dev/console: %s", strerror(-r));
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_RUNLEVEL2_TARGET,
279 "3", SPECIAL_RUNLEVEL3_TARGET,
280 "4", SPECIAL_RUNLEVEL4_TARGET,
281 "5", SPECIAL_RUNLEVEL5_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("Setting environment variable '%s' failed, ignoring: %s", value, strerror(ENOMEM));
361 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
363 } else if (streq(key, "quiet") && !value) {
365 log_set_max_level(LOG_NOTICE);
367 if (arg_show_status == _SHOW_STATUS_UNSET)
368 arg_show_status = SHOW_STATUS_AUTO;
370 } else if (streq(key, "debug") && !value) {
372 /* Note that log_parse_environment() handles 'debug'
373 * too, and sets the log level to LOG_DEBUG. */
375 if (detect_container(NULL) > 0)
376 log_set_target(LOG_TARGET_CONSOLE);
378 } else if (!in_initrd() && !value) {
381 /* SysV compatibility */
382 for (i = 0; i < ELEMENTSOF(rlmap); i += 2)
383 if (streq(key, rlmap[i]))
384 return set_default_unit(rlmap[i+1]);
390 #define DEFINE_SETTER(name, func, descr) \
391 static int name(const char *unit, \
392 const char *filename, \
394 const char *section, \
395 unsigned section_line, \
396 const char *lvalue, \
398 const char *rvalue, \
410 log_syntax(unit, LOG_ERR, filename, line, -r, \
411 "Invalid " descr "'%s': %s", \
412 rvalue, strerror(-r)); \
417 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level")
418 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target")
419 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" )
420 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location")
422 static int config_parse_cpu_affinity2(
424 const char *filename,
427 unsigned section_line,
434 const char *word, *state;
443 FOREACH_WORD_QUOTED(word, l, rvalue, state) {
448 if (!(t = strndup(word, l)))
451 r = safe_atou(t, &cpu);
455 if (!(c = cpu_set_malloc(&ncpus)))
458 if (r < 0 || cpu >= ncpus) {
459 log_syntax(unit, LOG_ERR, filename, line, -r,
460 "Failed to parse CPU affinity '%s'", rvalue);
465 CPU_SET_S(cpu, CPU_ALLOC_SIZE(ncpus), c);
468 log_syntax(unit, LOG_ERR, filename, line, EINVAL,
469 "Trailing garbage, ignoring.");
472 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
473 log_warning_unit(unit, "Failed to set CPU affinity: %m");
481 static int config_parse_show_status(
483 const char *filename,
486 unsigned section_line,
494 ShowStatus *b = data;
501 k = parse_show_status(rvalue, b);
503 log_syntax(unit, LOG_ERR, filename, line, -k,
504 "Failed to parse show status setting, ignoring: %s", rvalue);
511 static void strv_free_free(char ***l) {
523 static void free_join_controllers(void) {
524 strv_free_free(arg_join_controllers);
525 arg_join_controllers = NULL;
528 static int config_parse_join_controllers(const char *unit,
529 const char *filename,
532 unsigned section_line,
540 const char *word, *state;
547 free_join_controllers();
549 FOREACH_WORD_QUOTED(word, length, rvalue, state) {
552 s = strndup(word, length);
556 l = strv_split(s, ",");
561 if (strv_length(l) <= 1) {
566 if (!arg_join_controllers) {
567 arg_join_controllers = new(char**, 2);
568 if (!arg_join_controllers) {
573 arg_join_controllers[0] = l;
574 arg_join_controllers[1] = NULL;
581 t = new0(char**, n+2);
589 for (a = arg_join_controllers; *a; a++) {
591 if (strv_overlap(*a, l)) {
592 if (strv_extend_strv(&l, *a) < 0) {
612 t[n++] = strv_uniq(l);
614 strv_free_free(arg_join_controllers);
615 arg_join_controllers = t;
619 log_syntax(unit, LOG_ERR, filename, line, EINVAL,
620 "Trailing garbage, ignoring.");
625 static int parse_config_file(void) {
627 const ConfigTableItem items[] = {
628 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
629 { "Manager", "LogTarget", config_parse_target, 0, NULL },
630 { "Manager", "LogColor", config_parse_color, 0, NULL },
631 { "Manager", "LogLocation", config_parse_location, 0, NULL },
632 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
633 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
634 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
635 { "Manager", "CrashChVT", config_parse_int, 0, &arg_crash_chvt },
636 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
637 { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
638 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
639 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
640 { "Manager", "CapabilityBoundingSet", config_parse_bounding_set, 0, &arg_capability_bounding_set_drop },
642 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
644 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
645 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
646 { "Manager", "DefaultStandardOutput", config_parse_output, 0, &arg_default_std_output },
647 { "Manager", "DefaultStandardError", config_parse_output, 0, &arg_default_std_error },
648 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
649 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
650 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
651 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval },
652 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
653 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
654 { "Manager", "DefaultLimitCPU", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CPU] },
655 { "Manager", "DefaultLimitFSIZE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_FSIZE] },
656 { "Manager", "DefaultLimitDATA", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_DATA] },
657 { "Manager", "DefaultLimitSTACK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_STACK] },
658 { "Manager", "DefaultLimitCORE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CORE] },
659 { "Manager", "DefaultLimitRSS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RSS] },
660 { "Manager", "DefaultLimitNOFILE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NOFILE] },
661 { "Manager", "DefaultLimitAS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_AS] },
662 { "Manager", "DefaultLimitNPROC", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NPROC] },
663 { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MEMLOCK] },
664 { "Manager", "DefaultLimitLOCKS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_LOCKS] },
665 { "Manager", "DefaultLimitSIGPENDING", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_SIGPENDING] },
666 { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MSGQUEUE] },
667 { "Manager", "DefaultLimitNICE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NICE] },
668 { "Manager", "DefaultLimitRTPRIO", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTPRIO] },
669 { "Manager", "DefaultLimitRTTIME", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTTIME] },
670 { "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_default_cpu_accounting },
671 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
672 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
678 fn = arg_running_as == SYSTEMD_SYSTEM ? PKGSYSCONFDIR "/system.conf" : PKGSYSCONFDIR "/user.conf";
679 config_parse(NULL, fn, NULL,
681 config_item_table_lookup, items,
682 false, false, true, NULL);
687 static int parse_argv(int argc, char *argv[]) {
690 ARG_LOG_LEVEL = 0x100,
700 ARG_DUMP_CONFIGURATION_ITEMS,
707 ARG_DEFAULT_STD_OUTPUT,
708 ARG_DEFAULT_STD_ERROR
711 static const struct option options[] = {
712 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
713 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
714 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
715 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
716 { "unit", required_argument, NULL, ARG_UNIT },
717 { "system", no_argument, NULL, ARG_SYSTEM },
718 { "user", no_argument, NULL, ARG_USER },
719 { "test", no_argument, NULL, ARG_TEST },
720 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
721 { "help", no_argument, NULL, 'h' },
722 { "version", no_argument, NULL, ARG_VERSION },
723 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
724 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
725 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
726 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
727 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
728 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
729 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
730 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
731 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
743 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
748 r = log_set_max_level_from_string(optarg);
750 log_error("Failed to parse log level %s.", optarg);
757 r = log_set_target_from_string(optarg);
759 log_error("Failed to parse log target %s.", optarg);
768 r = log_show_color_from_string(optarg);
770 log_error("Failed to parse log color setting %s.", optarg);
774 log_show_color(true);
778 case ARG_LOG_LOCATION:
780 r = log_show_location_from_string(optarg);
782 log_error("Failed to parse log location setting %s.", optarg);
786 log_show_location(true);
790 case ARG_DEFAULT_STD_OUTPUT:
791 r = exec_output_from_string(optarg);
793 log_error("Failed to parse default standard output setting %s.", optarg);
796 arg_default_std_output = r;
799 case ARG_DEFAULT_STD_ERROR:
800 r = exec_output_from_string(optarg);
802 log_error("Failed to parse default standard error output setting %s.", optarg);
805 arg_default_std_error = r;
810 r = set_default_unit(optarg);
812 log_error("Failed to set default unit %s: %s", optarg, strerror(-r));
819 arg_running_as = SYSTEMD_SYSTEM;
823 arg_running_as = SYSTEMD_USER;
827 arg_action = ACTION_TEST;
828 if (arg_no_pager < 0)
837 arg_action = ACTION_VERSION;
840 case ARG_DUMP_CONFIGURATION_ITEMS:
841 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
845 r = optarg ? parse_boolean(optarg) : 1;
847 log_error("Failed to parse dump core boolean %s.", optarg);
853 case ARG_CRASH_SHELL:
854 r = optarg ? parse_boolean(optarg) : 1;
856 log_error("Failed to parse crash shell boolean %s.", optarg);
862 case ARG_CONFIRM_SPAWN:
863 r = optarg ? parse_boolean(optarg) : 1;
865 log_error("Failed to parse confirm spawn boolean %s.", optarg);
868 arg_confirm_spawn = r;
871 case ARG_SHOW_STATUS:
873 r = parse_show_status(optarg, &arg_show_status);
875 log_error("Failed to parse show status boolean %s.", optarg);
879 arg_show_status = SHOW_STATUS_YES;
882 case ARG_DESERIALIZE: {
886 r = safe_atoi(optarg, &fd);
887 if (r < 0 || fd < 0) {
888 log_error("Failed to parse deserialize option %s.", optarg);
889 return r < 0 ? r : -EINVAL;
892 fd_cloexec(fd, true);
896 log_error("Failed to open serialization fd: %m");
900 if (arg_serialization)
901 fclose(arg_serialization);
903 arg_serialization = f;
908 case ARG_SWITCHED_ROOT:
909 arg_switched_root = true;
913 arg_action = ACTION_HELP;
914 if (arg_no_pager < 0)
919 log_set_max_level(LOG_DEBUG);
925 /* Just to eat away the sysvinit kernel
926 * cmdline args without getopt() error
927 * messages that we'll parse in
928 * parse_proc_cmdline_word() or ignore. */
937 assert_not_reached("Unhandled option code.");
940 if (optind < argc && getpid() != 1) {
941 /* Hmm, when we aren't run as init system
942 * let's complain about excess arguments */
944 log_error("Excess arguments.");
951 static int help(void) {
953 printf("%s [OPTIONS...]\n\n"
954 "Starts up and maintains the system or user services.\n\n"
955 " -h --help Show this help\n"
956 " --test Determine startup sequence, dump it and exit\n"
957 " --no-pager Do not pipe output into a pager\n"
958 " --dump-configuration-items Dump understood unit configuration items\n"
959 " --unit=UNIT Set default unit\n"
960 " --system Run a system instance, even if PID != 1\n"
961 " --user Run a user instance\n"
962 " --dump-core[=0|1] Dump core on crash\n"
963 " --crash-shell[=0|1] Run shell on crash\n"
964 " --confirm-spawn[=0|1] Ask for confirmation when spawning processes\n"
965 " --show-status[=0|1] Show status updates on the console during bootup\n"
966 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
967 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
968 " --log-color[=0|1] Highlight important log messages\n"
969 " --log-location[=0|1] Include code location in log messages\n"
970 " --default-standard-output= Set default standard output for services\n"
971 " --default-standard-error= Set default standard error output for services\n",
972 program_invocation_short_name);
977 static int version(void) {
978 puts(PACKAGE_STRING);
979 puts(SYSTEMD_FEATURES);
984 static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
993 r = manager_open_serialization(m, &f);
995 log_error("Failed to create serialization file: %s", strerror(-r));
999 /* Make sure nothing is really destructed when we shut down */
1001 bus_manager_send_reloading(m, true);
1006 log_error("Failed to allocate fd set: %s", strerror(-r));
1010 r = manager_serialize(m, f, fds, switching_root);
1012 log_error("Failed to serialize state: %s", strerror(-r));
1016 if (fseeko(f, 0, SEEK_SET) < 0) {
1017 log_error("Failed to rewind serialization fd: %m");
1021 r = fd_cloexec(fileno(f), false);
1023 log_error("Failed to disable O_CLOEXEC for serialization: %s", strerror(-r));
1027 r = fdset_cloexec(fds, false);
1029 log_error("Failed to disable O_CLOEXEC for serialization fds: %s", strerror(-r));
1047 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1051 assert(saved_rlimit);
1053 /* Save the original RLIMIT_NOFILE so that we can reset it
1054 * later when transitioning from the initrd to the main
1055 * systemd or suchlike. */
1056 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0) {
1057 log_error("Reading RLIMIT_NOFILE failed: %m");
1061 /* Make sure forked processes get the default kernel setting */
1062 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1065 rl = newdup(struct rlimit, saved_rlimit, 1);
1069 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1072 /* Bump up the resource limit for ourselves substantially */
1073 nl.rlim_cur = nl.rlim_max = 64*1024;
1074 r = setrlimit_closest(RLIMIT_NOFILE, &nl);
1076 log_error("Setting RLIMIT_NOFILE failed: %s", strerror(-r));
1083 static void test_mtab(void) {
1085 static const char ok[] =
1086 "/proc/self/mounts\0"
1088 "../proc/self/mounts\0"
1091 _cleanup_free_ char *p = NULL;
1094 /* Check that /etc/mtab is a symlink to the right place or
1095 * non-existing. But certainly not a file, or a symlink to
1096 * some weird place... */
1098 r = readlink_malloc("/etc/mtab", &p);
1101 if (r >= 0 && nulstr_contains(ok, p))
1104 log_warning("/etc/mtab is not a symlink or not pointing to /proc/self/mounts. "
1105 "This is not supported anymore. "
1106 "Please make sure to replace this file by a symlink to avoid incorrect or misleading mount(8) output.");
1109 static void test_usr(void) {
1111 /* Check that /usr is not a separate fs */
1113 if (dir_is_empty("/usr") <= 0)
1116 log_warning("/usr appears to be on its own filesytem and is not already mounted. This is not a supported setup. "
1117 "Some things will probably break (sometimes even silently) in mysterious ways. "
1118 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1121 static int initialize_join_controllers(void) {
1122 /* By default, mount "cpu" + "cpuacct" together, and "net_cls"
1123 * + "net_prio". We'd like to add "cpuset" to the mix, but
1124 * "cpuset" does't really work for groups with no initialized
1127 arg_join_controllers = new(char**, 3);
1128 if (!arg_join_controllers)
1131 arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL);
1132 arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL);
1133 arg_join_controllers[2] = NULL;
1135 if (!arg_join_controllers[0] || !arg_join_controllers[1]) {
1136 free_join_controllers();
1143 static int enforce_syscall_archs(Set *archs) {
1145 scmp_filter_ctx *seccomp;
1150 seccomp = seccomp_init(SCMP_ACT_ALLOW);
1154 SET_FOREACH(id, arg_syscall_archs, i) {
1155 r = seccomp_arch_add(seccomp, PTR_TO_UINT32(id) - 1);
1159 log_error("Failed to add architecture to seccomp: %s", strerror(-r));
1164 r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
1166 log_error("Failed to unset NO_NEW_PRIVS: %s", strerror(-r));
1170 r = seccomp_load(seccomp);
1172 log_error("Failed to add install architecture seccomp: %s", strerror(-r));
1175 seccomp_release(seccomp);
1182 static int status_welcome(void) {
1183 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1186 r = parse_env_file("/etc/os-release", NEWLINE,
1187 "PRETTY_NAME", &pretty_name,
1188 "ANSI_COLOR", &ansi_color,
1191 r = parse_env_file("/usr/lib/os-release", NEWLINE,
1192 "PRETTY_NAME", &pretty_name,
1193 "ANSI_COLOR", &ansi_color,
1197 if (r < 0 && r != -ENOENT)
1198 log_warning("Failed to read os-release file: %s", strerror(-r));
1200 return status_printf(NULL, false, false,
1201 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1202 isempty(ansi_color) ? "1" : ansi_color,
1203 isempty(pretty_name) ? "Linux" : pretty_name);
1206 static int write_container_id(void) {
1209 c = getenv("container");
1213 return write_string_file("/run/systemd/container", c);
1216 int main(int argc, char *argv[]) {
1218 int r, retval = EXIT_FAILURE;
1219 usec_t before_startup, after_startup;
1220 char timespan[FORMAT_TIMESPAN_MAX];
1222 bool reexecute = false;
1223 const char *shutdown_verb = NULL;
1224 dual_timestamp initrd_timestamp = { 0ULL, 0ULL };
1225 dual_timestamp userspace_timestamp = { 0ULL, 0ULL };
1226 dual_timestamp kernel_timestamp = { 0ULL, 0ULL };
1227 dual_timestamp security_start_timestamp = { 0ULL, 0ULL };
1228 dual_timestamp security_finish_timestamp = { 0ULL, 0ULL };
1229 static char systemd[] = "systemd";
1230 bool skip_setup = false;
1232 bool loaded_policy = false;
1233 bool arm_reboot_watchdog = false;
1234 bool queue_default_job = false;
1235 bool empty_etc = false;
1236 char *switch_root_dir = NULL, *switch_root_init = NULL;
1237 static struct rlimit saved_rlimit_nofile = { 0, 0 };
1239 #ifdef HAVE_SYSV_COMPAT
1240 if (getpid() != 1 && strstr(program_invocation_short_name, "init")) {
1241 /* This is compatibility support for SysV, where
1242 * calling init as a user is identical to telinit. */
1245 execv(SYSTEMCTL_BINARY_PATH, argv);
1246 log_error("Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1251 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
1252 dual_timestamp_get(&userspace_timestamp);
1254 /* Determine if this is a reexecution or normal bootup. We do
1255 * the full command line parsing much later, so let's just
1256 * have a quick peek here. */
1257 if (strv_find(argv+1, "--deserialize"))
1260 /* If we have switched root, do all the special setup
1262 if (strv_find(argv+1, "--switched-root"))
1265 /* If we get started via the /sbin/init symlink then we are
1266 called 'init'. After a subsequent reexecution we are then
1267 called 'systemd'. That is confusing, hence let's call us
1268 systemd right-away. */
1269 program_invocation_short_name = systemd;
1270 prctl(PR_SET_NAME, systemd);
1275 log_show_color(isatty(STDERR_FILENO) > 0);
1276 log_set_upgrade_syslog_to_journal(true);
1278 /* Disable the umask logic */
1282 if (getpid() == 1 && detect_container(NULL) <= 0) {
1284 /* Running outside of a container as PID 1 */
1285 arg_running_as = SYSTEMD_SYSTEM;
1287 log_set_target(LOG_TARGET_KMSG);
1291 initrd_timestamp = userspace_timestamp;
1294 mount_setup_early();
1295 dual_timestamp_get(&security_start_timestamp);
1296 if (mac_selinux_setup(&loaded_policy) < 0)
1298 if (ima_setup() < 0)
1300 if (mac_smack_setup(&loaded_policy) < 0)
1302 dual_timestamp_get(&security_finish_timestamp);
1305 if (mac_selinux_init(NULL) < 0)
1309 if (clock_is_localtime() > 0) {
1313 * The very first call of settimeofday() also does a time warp in the kernel.
1315 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on
1316 * external tools to take care of maintaining the RTC and do all adjustments.
1317 * This matches the behavior of Windows, which leaves the RTC alone if the
1318 * registry tells that the RTC runs in UTC.
1320 r = clock_set_timezone(&min);
1322 log_error("Failed to apply local time delta, ignoring: %s", strerror(-r));
1324 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1325 } else if (!in_initrd()) {
1327 * Do a dummy very first call to seal the kernel's time warp magic.
1329 * Do not call this this from inside the initrd. The initrd might not
1330 * carry /etc/adjtime with LOCAL, but the real system could be set up
1331 * that way. In such case, we need to delay the time-warp or the sealing
1332 * until we reach the real system.
1334 * Do no set the kernel's timezone. The concept of local time cannot
1335 * be supported reliably, the time will jump or be incorrect at every daylight
1336 * saving time change. All kernel local time concepts will be treated
1339 clock_reset_timewarp();
1343 /* Set the default for later on, but don't actually
1344 * open the logs like this for now. Note that if we
1345 * are transitioning from the initrd there might still
1346 * be journal fd open, and we shouldn't attempt
1347 * opening that before we parsed /proc/cmdline which
1348 * might redirect output elsewhere. */
1349 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
1351 } else if (getpid() == 1) {
1352 /* Running inside a container, as PID 1 */
1353 arg_running_as = SYSTEMD_SYSTEM;
1354 log_set_target(LOG_TARGET_CONSOLE);
1355 log_close_console(); /* force reopen of /dev/console */
1358 /* For the later on, see above... */
1359 log_set_target(LOG_TARGET_JOURNAL);
1361 /* clear the kernel timestamp,
1362 * because we are in a container */
1363 kernel_timestamp.monotonic = 0ULL;
1364 kernel_timestamp.realtime = 0ULL;
1367 /* Running as user instance */
1368 arg_running_as = SYSTEMD_USER;
1369 log_set_target(LOG_TARGET_AUTO);
1372 /* clear the kernel timestamp,
1373 * because we are not PID 1 */
1374 kernel_timestamp.monotonic = 0ULL;
1375 kernel_timestamp.realtime = 0ULL;
1378 /* Initialize default unit */
1379 r = set_default_unit(SPECIAL_DEFAULT_TARGET);
1381 log_error("Failed to set default unit %s: %s", SPECIAL_DEFAULT_TARGET, strerror(-r));
1385 r = initialize_join_controllers();
1389 /* Mount /proc, /sys and friends, so that /proc/cmdline and
1390 * /proc/$PID/fd is available. */
1391 if (getpid() == 1) {
1392 r = mount_setup(loaded_policy);
1397 /* Reset all signal handlers. */
1398 assert_se(reset_all_signal_handlers() == 0);
1400 ignore_signals(SIGNALS_IGNORE, -1);
1402 if (parse_config_file() < 0)
1405 if (arg_running_as == SYSTEMD_SYSTEM)
1406 if (parse_proc_cmdline(parse_proc_cmdline_item) < 0)
1409 /* Note that this also parses bits from the kernel command
1410 * line, including "debug". */
1411 log_parse_environment();
1413 if (parse_argv(argc, argv) < 0)
1416 if (arg_action == ACTION_TEST &&
1418 log_error("Don't run test mode as root.");
1422 if (arg_running_as == SYSTEMD_USER &&
1423 arg_action == ACTION_RUN &&
1425 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
1429 if (arg_running_as == SYSTEMD_SYSTEM &&
1430 arg_action == ACTION_RUN &&
1431 running_in_chroot() > 0) {
1432 log_error("Cannot be run in a chroot() environment.");
1436 if (arg_action == ACTION_TEST)
1439 pager_open_if_enabled();
1441 if (arg_action == ACTION_HELP) {
1444 } else if (arg_action == ACTION_VERSION) {
1447 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
1448 unit_dump_config_items(stdout);
1449 retval = EXIT_SUCCESS;
1451 } else if (arg_action == ACTION_DONE) {
1452 retval = EXIT_SUCCESS;
1456 if (arg_running_as == SYSTEMD_USER &&
1457 !getenv("XDG_RUNTIME_DIR")) {
1458 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
1462 assert_se(arg_action == ACTION_RUN || arg_action == ACTION_TEST);
1464 /* Close logging fds, in order not to confuse fdset below */
1467 /* Remember open file descriptors for later deserialization */
1468 r = fdset_new_fill(&fds);
1470 log_error("Failed to allocate fd set: %s", strerror(-r));
1473 fdset_cloexec(fds, true);
1475 if (arg_serialization)
1476 assert_se(fdset_remove(fds, fileno(arg_serialization)) >= 0);
1478 if (arg_running_as == SYSTEMD_SYSTEM)
1479 /* Become a session leader if we aren't one yet. */
1482 /* Move out of the way, so that we won't block unmounts */
1483 assert_se(chdir("/") == 0);
1485 /* Reset the console, but only if this is really init and we
1486 * are freshly booted */
1487 if (arg_running_as == SYSTEMD_SYSTEM && arg_action == ACTION_RUN) {
1489 /* If we are init, we connect stdin/stdout/stderr to
1490 * /dev/null and make sure we don't have a controlling
1494 if (getpid() == 1 && !skip_setup)
1498 /* Open the logging devices, if possible and necessary */
1501 if (arg_show_status == _SHOW_STATUS_UNSET)
1502 arg_show_status = SHOW_STATUS_YES;
1504 /* Make sure we leave a core dump without panicing the
1506 if (getpid() == 1) {
1507 install_crash_handler();
1509 r = mount_cgroup_controllers(arg_join_controllers);
1514 if (arg_running_as == SYSTEMD_SYSTEM) {
1515 const char *virtualization = NULL;
1517 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1518 arg_action == ACTION_TEST ? "test " : "" );
1520 detect_virtualization(&virtualization);
1522 log_info("Detected virtualization '%s'.", virtualization);
1524 write_container_id();
1526 log_info("Detected architecture '%s'.", architecture_to_string(uname_architecture()));
1529 log_info("Running in initial RAM disk.");
1531 /* Let's check whether /etc is already populated. We
1532 * don't actually really check for that, but use
1533 * /etc/machine-id as flag file. This allows container
1534 * managers and installers to provision a couple of
1535 * files already. If the container manager wants to
1536 * provision the machine ID itself it should pass
1537 * $container_uuid to PID 1.*/
1539 empty_etc = access("/etc/machine-id", F_OK) < 0;
1541 log_info("Running with unpopulated /etc.");
1543 _cleanup_free_ char *t;
1545 t = uid_to_name(getuid());
1546 log_debug(PACKAGE_STRING " running in %suser mode for user "UID_FMT"/%s. (" SYSTEMD_FEATURES ")",
1547 arg_action == ACTION_TEST ? " test" : "", getuid(), t);
1550 if (arg_running_as == SYSTEMD_SYSTEM && !skip_setup) {
1551 if (arg_show_status > 0 || plymouth_running())
1558 machine_id_setup(NULL);
1565 if (arg_running_as == SYSTEMD_SYSTEM && arg_runtime_watchdog > 0)
1566 watchdog_set_timeout(&arg_runtime_watchdog);
1568 if (arg_timer_slack_nsec != NSEC_INFINITY)
1569 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1570 log_error("Failed to adjust timer slack: %m");
1572 if (arg_capability_bounding_set_drop) {
1573 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set_drop);
1575 log_error("Failed to drop capability bounding set of usermode helpers: %s", strerror(-r));
1578 r = capability_bounding_set_drop(arg_capability_bounding_set_drop, true);
1580 log_error("Failed to drop capability bounding set: %s", strerror(-r));
1585 if (arg_syscall_archs) {
1586 r = enforce_syscall_archs(arg_syscall_archs);
1591 if (arg_running_as == SYSTEMD_USER) {
1592 /* Become reaper of our children */
1593 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0) {
1594 log_warning("Failed to make us a subreaper: %m");
1595 if (errno == EINVAL)
1596 log_info("Perhaps the kernel version is too old (< 3.4?)");
1600 if (arg_running_as == SYSTEMD_SYSTEM) {
1601 bump_rlimit_nofile(&saved_rlimit_nofile);
1604 r = unit_file_preset_all(UNIT_FILE_SYSTEM, false, NULL, UNIT_FILE_PRESET_FULL, false, NULL, 0);
1606 log_warning("Failed to populate /etc with preset unit settings, ignoring: %s", strerror(-r));
1608 log_info("Populated /etc with preset unit settings.");
1612 r = manager_new(arg_running_as, arg_action == ACTION_TEST, &m);
1614 log_error("Failed to allocate manager object: %s", strerror(-r));
1618 m->confirm_spawn = arg_confirm_spawn;
1619 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
1620 m->default_std_output = arg_default_std_output;
1621 m->default_std_error = arg_default_std_error;
1622 m->default_restart_usec = arg_default_restart_usec;
1623 m->default_timeout_start_usec = arg_default_timeout_start_usec;
1624 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
1625 m->default_start_limit_interval = arg_default_start_limit_interval;
1626 m->default_start_limit_burst = arg_default_start_limit_burst;
1627 m->default_cpu_accounting = arg_default_cpu_accounting;
1628 m->default_blockio_accounting = arg_default_blockio_accounting;
1629 m->default_memory_accounting = arg_default_memory_accounting;
1630 m->runtime_watchdog = arg_runtime_watchdog;
1631 m->shutdown_watchdog = arg_shutdown_watchdog;
1633 m->userspace_timestamp = userspace_timestamp;
1634 m->kernel_timestamp = kernel_timestamp;
1635 m->initrd_timestamp = initrd_timestamp;
1636 m->security_start_timestamp = security_start_timestamp;
1637 m->security_finish_timestamp = security_finish_timestamp;
1639 manager_set_default_rlimits(m, arg_default_rlimit);
1640 manager_environment_add(m, NULL, arg_default_environment);
1641 manager_set_show_status(m, arg_show_status);
1642 manager_set_first_boot(m, empty_etc);
1644 /* Remember whether we should queue the default job */
1645 queue_default_job = !arg_serialization || arg_switched_root;
1647 before_startup = now(CLOCK_MONOTONIC);
1649 r = manager_startup(m, arg_serialization, fds);
1651 log_error("Failed to fully start up daemon: %s", strerror(-r));
1653 /* This will close all file descriptors that were opened, but
1654 * not claimed by any unit. */
1658 if (arg_serialization) {
1659 fclose(arg_serialization);
1660 arg_serialization = NULL;
1663 if (queue_default_job) {
1664 _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL;
1665 Unit *target = NULL;
1666 Job *default_unit_job;
1668 log_debug("Activating default unit: %s", arg_default_unit);
1670 r = manager_load_unit(m, arg_default_unit, NULL, &error, &target);
1672 log_error("Failed to load default target: %s", bus_error_message(&error, r));
1673 else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND)
1674 log_error("Failed to load default target: %s", strerror(-target->load_error));
1675 else if (target->load_state == UNIT_MASKED)
1676 log_error("Default target masked.");
1678 if (!target || target->load_state != UNIT_LOADED) {
1679 log_info("Trying to load rescue target...");
1681 r = manager_load_unit(m, SPECIAL_RESCUE_TARGET, NULL, &error, &target);
1683 log_error("Failed to load rescue target: %s", bus_error_message(&error, r));
1685 } else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND) {
1686 log_error("Failed to load rescue target: %s", strerror(-target->load_error));
1688 } else if (target->load_state == UNIT_MASKED) {
1689 log_error("Rescue target masked.");
1694 assert(target->load_state == UNIT_LOADED);
1696 if (arg_action == ACTION_TEST) {
1697 printf("-> By units:\n");
1698 manager_dump_units(m, stdout, "\t");
1701 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, false, &error, &default_unit_job);
1703 log_debug("Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1705 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, false, &error, &default_unit_job);
1707 log_error("Failed to start default target: %s", bus_error_message(&error, r));
1711 log_error("Failed to isolate default target: %s", bus_error_message(&error, r));
1715 m->default_unit_job_id = default_unit_job->id;
1717 after_startup = now(CLOCK_MONOTONIC);
1718 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
1719 "Loaded units and determined initial transaction in %s.",
1720 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
1722 if (arg_action == ACTION_TEST) {
1723 printf("-> By jobs:\n");
1724 manager_dump_jobs(m, stdout, "\t");
1725 retval = EXIT_SUCCESS;
1731 r = manager_loop(m);
1733 log_error("Failed to run mainloop: %s", strerror(-r));
1737 switch (m->exit_code) {
1740 retval = EXIT_SUCCESS;
1744 case MANAGER_RELOAD:
1745 log_info("Reloading.");
1746 r = manager_reload(m);
1748 log_error("Failed to reload: %s", strerror(-r));
1751 case MANAGER_REEXECUTE:
1753 if (prepare_reexecute(m, &arg_serialization, &fds, false) < 0)
1757 log_notice("Reexecuting.");
1760 case MANAGER_SWITCH_ROOT:
1761 /* Steal the switch root parameters */
1762 switch_root_dir = m->switch_root;
1763 switch_root_init = m->switch_root_init;
1764 m->switch_root = m->switch_root_init = NULL;
1766 if (!switch_root_init)
1767 if (prepare_reexecute(m, &arg_serialization, &fds, true) < 0)
1771 log_notice("Switching root.");
1774 case MANAGER_REBOOT:
1775 case MANAGER_POWEROFF:
1777 case MANAGER_KEXEC: {
1778 static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
1779 [MANAGER_REBOOT] = "reboot",
1780 [MANAGER_POWEROFF] = "poweroff",
1781 [MANAGER_HALT] = "halt",
1782 [MANAGER_KEXEC] = "kexec"
1785 assert_se(shutdown_verb = table[m->exit_code]);
1786 arm_reboot_watchdog = m->exit_code == MANAGER_REBOOT;
1788 log_notice("Shutting down.");
1793 assert_not_reached("Unknown exit code.");
1805 for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++) {
1806 free(arg_default_rlimit[j]);
1807 arg_default_rlimit[j] = NULL;
1810 free(arg_default_unit);
1811 arg_default_unit = NULL;
1813 free_join_controllers();
1815 strv_free(arg_default_environment);
1816 arg_default_environment = NULL;
1818 set_free(arg_syscall_archs);
1819 arg_syscall_archs = NULL;
1821 mac_selinux_finish();
1825 unsigned i, args_size;
1827 /* Close and disarm the watchdog, so that the new
1828 * instance can reinitialize it, but doesn't get
1829 * rebooted while we do that */
1830 watchdog_close(true);
1832 /* Reset the RLIMIT_NOFILE to the kernel default, so
1833 * that the new systemd can pass the kernel default to
1834 * its child processes */
1835 if (saved_rlimit_nofile.rlim_cur > 0)
1836 setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile);
1838 if (switch_root_dir) {
1839 /* Kill all remaining processes from the
1840 * initrd, but don't wait for them, so that we
1841 * can handle the SIGCHLD for them after
1843 broadcast_signal(SIGTERM, false, true);
1845 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1846 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1848 log_error("Failed to switch root, trying to continue: %s", strerror(-r));
1851 args_size = MAX(6, argc+1);
1852 args = newa(const char*, args_size);
1854 if (!switch_root_init) {
1857 /* First try to spawn ourselves with the right
1858 * path, and with full serialization. We do
1859 * this only if the user didn't specify an
1860 * explicit init to spawn. */
1862 assert(arg_serialization);
1865 snprintf(sfd, sizeof(sfd), "%i", fileno(arg_serialization));
1869 args[i++] = SYSTEMD_BINARY_PATH;
1870 if (switch_root_dir)
1871 args[i++] = "--switched-root";
1872 args[i++] = arg_running_as == SYSTEMD_SYSTEM ? "--system" : "--user";
1873 args[i++] = "--deserialize";
1877 /* do not pass along the environment we inherit from the kernel or initrd */
1878 if (switch_root_dir)
1881 assert(i <= args_size);
1882 execv(args[0], (char* const*) args);
1885 /* Try the fallback, if there is any, without any
1886 * serialization. We pass the original argv[] and
1887 * envp[]. (Well, modulo the ordering changes due to
1888 * getopt() in argv[], and some cleanups in envp[],
1889 * but let's hope that doesn't matter.) */
1891 if (arg_serialization) {
1892 fclose(arg_serialization);
1893 arg_serialization = NULL;
1901 /* Reopen the console */
1902 make_console_stdio();
1904 for (j = 1, i = 1; j < (unsigned) argc; j++)
1905 args[i++] = argv[j];
1907 assert(i <= args_size);
1909 /* Reenable any blocked signals, especially important
1910 * if we switch from initial ramdisk to init=... */
1911 reset_all_signal_handlers();
1912 reset_signal_mask();
1914 if (switch_root_init) {
1915 args[0] = switch_root_init;
1916 execv(args[0], (char* const*) args);
1917 log_warning("Failed to execute configured init, trying fallback: %m");
1920 args[0] = "/sbin/init";
1921 execv(args[0], (char* const*) args);
1923 if (errno == ENOENT) {
1924 log_warning("No /sbin/init, trying fallback");
1926 args[0] = "/bin/sh";
1928 execv(args[0], (char* const*) args);
1929 log_error("Failed to execute /bin/sh, giving up: %m");
1931 log_warning("Failed to execute /sbin/init, giving up: %m");
1934 if (arg_serialization) {
1935 fclose(arg_serialization);
1936 arg_serialization = NULL;
1944 #ifdef HAVE_VALGRIND_VALGRIND_H
1945 /* If we are PID 1 and running under valgrind, then let's exit
1946 * here explicitly. valgrind will only generate nice output on
1947 * exit(), not on exec(), hence let's do the former not the
1949 if (getpid() == 1 && RUNNING_ON_VALGRIND)
1953 if (shutdown_verb) {
1954 char log_level[DECIMAL_STR_MAX(int) + 1];
1955 const char* command_line[9] = {
1956 SYSTEMD_SHUTDOWN_BINARY_PATH,
1958 "--log-level", log_level,
1962 _cleanup_strv_free_ char **env_block = NULL;
1964 assert(command_line[pos] == NULL);
1965 env_block = strv_copy(environ);
1967 snprintf(log_level, sizeof(log_level), "%d", log_get_max_level());
1969 switch (log_get_target()) {
1970 case LOG_TARGET_KMSG:
1971 case LOG_TARGET_JOURNAL_OR_KMSG:
1972 case LOG_TARGET_SYSLOG_OR_KMSG:
1973 command_line[pos++] = "kmsg";
1976 case LOG_TARGET_CONSOLE:
1978 command_line[pos++] = "console";
1982 if (log_get_show_color())
1983 command_line[pos++] = "--log-color";
1985 if (log_get_show_location())
1986 command_line[pos++] = "--log-location";
1988 assert(pos < ELEMENTSOF(command_line));
1990 if (arm_reboot_watchdog && arg_shutdown_watchdog > 0) {
1993 /* If we reboot let's set the shutdown
1994 * watchdog and tell the shutdown binary to
1995 * repeatedly ping it */
1996 watchdog_set_timeout(&arg_shutdown_watchdog);
1997 watchdog_close(false);
1999 /* Tell the binary how often to ping, ignore failure */
2000 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
2001 strv_push(&env_block, e);
2003 watchdog_close(true);
2005 /* Avoid the creation of new processes forked by the
2006 * kernel; at this point, we will not listen to the
2008 if (detect_container(NULL) <= 0)
2009 cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
2011 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
2012 log_error("Failed to execute shutdown binary, %s: %m",
2013 getpid() == 1 ? "freezing" : "quitting");