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 "single", SPECIAL_RESCUE_TARGET,
273 "-s", SPECIAL_RESCUE_TARGET,
274 "s", SPECIAL_RESCUE_TARGET,
275 "S", SPECIAL_RESCUE_TARGET,
276 "1", SPECIAL_RESCUE_TARGET,
277 "2", SPECIAL_RUNLEVEL2_TARGET,
278 "3", SPECIAL_RUNLEVEL3_TARGET,
279 "4", SPECIAL_RUNLEVEL4_TARGET,
280 "5", SPECIAL_RUNLEVEL5_TARGET,
286 if (streq(key, "systemd.unit") && value) {
289 return set_default_unit(value);
291 } else if (streq(key, "rd.systemd.unit") && value) {
294 return set_default_unit(value);
296 } else if (streq(key, "systemd.dump_core") && value) {
298 r = parse_boolean(value);
300 log_warning("Failed to parse dump core switch %s. Ignoring.", value);
304 } else if (streq(key, "systemd.crash_shell") && value) {
306 r = parse_boolean(value);
308 log_warning("Failed to parse crash shell switch %s. Ignoring.", value);
312 } else if (streq(key, "systemd.crash_chvt") && value) {
314 if (safe_atoi(value, &r) < 0)
315 log_warning("Failed to parse crash chvt switch %s. Ignoring.", value);
319 } else if (streq(key, "systemd.confirm_spawn") && value) {
321 r = parse_boolean(value);
323 log_warning("Failed to parse confirm spawn switch %s. Ignoring.", value);
325 arg_confirm_spawn = r;
327 } else if (streq(key, "systemd.show_status") && value) {
329 r = parse_show_status(value, &arg_show_status);
331 log_warning("Failed to parse show status switch %s. Ignoring.", value);
333 } else if (streq(key, "systemd.default_standard_output") && value) {
335 r = exec_output_from_string(value);
337 log_warning("Failed to parse default standard output switch %s. Ignoring.", value);
339 arg_default_std_output = r;
341 } else if (streq(key, "systemd.default_standard_error") && value) {
343 r = exec_output_from_string(value);
345 log_warning("Failed to parse default standard error switch %s. Ignoring.", value);
347 arg_default_std_error = r;
349 } else if (streq(key, "systemd.setenv") && value) {
351 if (env_assignment_is_valid(value)) {
354 env = strv_env_set(arg_default_environment, value);
356 arg_default_environment = env;
358 log_warning("Setting environment variable '%s' failed, ignoring: %s", value, strerror(ENOMEM));
360 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
362 } else if (streq(key, "quiet") && !value) {
364 log_set_max_level(LOG_NOTICE);
366 if (arg_show_status == _SHOW_STATUS_UNSET)
367 arg_show_status = SHOW_STATUS_AUTO;
369 } else if (streq(key, "debug") && !value) {
371 /* Note that log_parse_environment() handles 'debug'
372 * too, and sets the log level to LOG_DEBUG. */
374 if (detect_container(NULL) > 0)
375 log_set_target(LOG_TARGET_CONSOLE);
377 } else if (!in_initrd() && !value) {
380 /* SysV compatibility */
381 for (i = 0; i < ELEMENTSOF(rlmap); i += 2)
382 if (streq(key, rlmap[i]))
383 return set_default_unit(rlmap[i+1]);
389 #define DEFINE_SETTER(name, func, descr) \
390 static int name(const char *unit, \
391 const char *filename, \
393 const char *section, \
394 unsigned section_line, \
395 const char *lvalue, \
397 const char *rvalue, \
409 log_syntax(unit, LOG_ERR, filename, line, -r, \
410 "Invalid " descr "'%s': %s", \
411 rvalue, strerror(-r)); \
416 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level")
417 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target")
418 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" )
419 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location")
421 static int config_parse_cpu_affinity2(
423 const char *filename,
426 unsigned section_line,
433 const char *word, *state;
442 FOREACH_WORD_QUOTED(word, l, rvalue, state) {
447 if (!(t = strndup(word, l)))
450 r = safe_atou(t, &cpu);
454 if (!(c = cpu_set_malloc(&ncpus)))
457 if (r < 0 || cpu >= ncpus) {
458 log_syntax(unit, LOG_ERR, filename, line, -r,
459 "Failed to parse CPU affinity '%s'", rvalue);
464 CPU_SET_S(cpu, CPU_ALLOC_SIZE(ncpus), c);
467 log_syntax(unit, LOG_ERR, filename, line, EINVAL,
468 "Trailing garbage, ignoring.");
471 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
472 log_warning_unit(unit, "Failed to set CPU affinity: %m");
480 static int config_parse_show_status(
482 const char *filename,
485 unsigned section_line,
493 ShowStatus *b = data;
500 k = parse_show_status(rvalue, b);
502 log_syntax(unit, LOG_ERR, filename, line, -k,
503 "Failed to parse show status setting, ignoring: %s", rvalue);
510 static void strv_free_free(char ***l) {
522 static void free_join_controllers(void) {
523 strv_free_free(arg_join_controllers);
524 arg_join_controllers = NULL;
527 static int config_parse_join_controllers(const char *unit,
528 const char *filename,
531 unsigned section_line,
539 const char *word, *state;
546 free_join_controllers();
548 FOREACH_WORD_QUOTED(word, length, rvalue, state) {
551 s = strndup(word, length);
555 l = strv_split(s, ",");
560 if (strv_length(l) <= 1) {
565 if (!arg_join_controllers) {
566 arg_join_controllers = new(char**, 2);
567 if (!arg_join_controllers) {
572 arg_join_controllers[0] = l;
573 arg_join_controllers[1] = NULL;
580 t = new0(char**, n+2);
588 for (a = arg_join_controllers; *a; a++) {
590 if (strv_overlap(*a, l)) {
591 if (strv_extend_strv(&l, *a) < 0) {
611 t[n++] = strv_uniq(l);
613 strv_free_free(arg_join_controllers);
614 arg_join_controllers = t;
618 log_syntax(unit, LOG_ERR, filename, line, EINVAL,
619 "Trailing garbage, ignoring.");
624 static int parse_config_file(void) {
626 const ConfigTableItem items[] = {
627 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
628 { "Manager", "LogTarget", config_parse_target, 0, NULL },
629 { "Manager", "LogColor", config_parse_color, 0, NULL },
630 { "Manager", "LogLocation", config_parse_location, 0, NULL },
631 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
632 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
633 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
634 { "Manager", "CrashChVT", config_parse_int, 0, &arg_crash_chvt },
635 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
636 { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
637 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
638 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
639 { "Manager", "CapabilityBoundingSet", config_parse_bounding_set, 0, &arg_capability_bounding_set_drop },
641 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
643 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
644 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
645 { "Manager", "DefaultStandardOutput", config_parse_output, 0, &arg_default_std_output },
646 { "Manager", "DefaultStandardError", config_parse_output, 0, &arg_default_std_error },
647 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
648 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
649 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
650 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval },
651 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
652 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
653 { "Manager", "DefaultLimitCPU", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CPU] },
654 { "Manager", "DefaultLimitFSIZE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_FSIZE] },
655 { "Manager", "DefaultLimitDATA", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_DATA] },
656 { "Manager", "DefaultLimitSTACK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_STACK] },
657 { "Manager", "DefaultLimitCORE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CORE] },
658 { "Manager", "DefaultLimitRSS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RSS] },
659 { "Manager", "DefaultLimitNOFILE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NOFILE] },
660 { "Manager", "DefaultLimitAS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_AS] },
661 { "Manager", "DefaultLimitNPROC", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NPROC] },
662 { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MEMLOCK] },
663 { "Manager", "DefaultLimitLOCKS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_LOCKS] },
664 { "Manager", "DefaultLimitSIGPENDING", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_SIGPENDING] },
665 { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MSGQUEUE] },
666 { "Manager", "DefaultLimitNICE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NICE] },
667 { "Manager", "DefaultLimitRTPRIO", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTPRIO] },
668 { "Manager", "DefaultLimitRTTIME", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTTIME] },
669 { "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_default_cpu_accounting },
670 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
671 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
677 fn = arg_running_as == SYSTEMD_SYSTEM ? PKGSYSCONFDIR "/system.conf" : PKGSYSCONFDIR "/user.conf";
678 config_parse(NULL, fn, NULL,
680 config_item_table_lookup, items,
681 false, false, true, NULL);
686 static int parse_argv(int argc, char *argv[]) {
689 ARG_LOG_LEVEL = 0x100,
699 ARG_DUMP_CONFIGURATION_ITEMS,
706 ARG_DEFAULT_STD_OUTPUT,
707 ARG_DEFAULT_STD_ERROR
710 static const struct option options[] = {
711 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
712 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
713 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
714 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
715 { "unit", required_argument, NULL, ARG_UNIT },
716 { "system", no_argument, NULL, ARG_SYSTEM },
717 { "user", no_argument, NULL, ARG_USER },
718 { "test", no_argument, NULL, ARG_TEST },
719 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
720 { "help", no_argument, NULL, 'h' },
721 { "version", no_argument, NULL, ARG_VERSION },
722 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
723 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
724 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
725 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
726 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
727 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
728 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
729 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
730 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
742 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
747 r = log_set_max_level_from_string(optarg);
749 log_error("Failed to parse log level %s.", optarg);
756 r = log_set_target_from_string(optarg);
758 log_error("Failed to parse log target %s.", optarg);
767 r = log_show_color_from_string(optarg);
769 log_error("Failed to parse log color setting %s.", optarg);
773 log_show_color(true);
777 case ARG_LOG_LOCATION:
779 r = log_show_location_from_string(optarg);
781 log_error("Failed to parse log location setting %s.", optarg);
785 log_show_location(true);
789 case ARG_DEFAULT_STD_OUTPUT:
790 r = exec_output_from_string(optarg);
792 log_error("Failed to parse default standard output setting %s.", optarg);
795 arg_default_std_output = r;
798 case ARG_DEFAULT_STD_ERROR:
799 r = exec_output_from_string(optarg);
801 log_error("Failed to parse default standard error output setting %s.", optarg);
804 arg_default_std_error = r;
809 r = set_default_unit(optarg);
811 log_error("Failed to set default unit %s: %s", optarg, strerror(-r));
818 arg_running_as = SYSTEMD_SYSTEM;
822 arg_running_as = SYSTEMD_USER;
826 arg_action = ACTION_TEST;
827 if (arg_no_pager < 0)
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;
913 if (arg_no_pager < 0)
918 log_set_max_level(LOG_DEBUG);
924 /* Just to eat away the sysvinit kernel
925 * cmdline args without getopt() error
926 * messages that we'll parse in
927 * parse_proc_cmdline_word() or ignore. */
936 assert_not_reached("Unhandled option code.");
939 if (optind < argc && getpid() != 1) {
940 /* Hmm, when we aren't run as init system
941 * let's complain about excess arguments */
943 log_error("Excess arguments.");
950 static int help(void) {
952 printf("%s [OPTIONS...]\n\n"
953 "Starts up and maintains the system or user services.\n\n"
954 " -h --help Show this help\n"
955 " --test Determine startup sequence, dump it and exit\n"
956 " --no-pager Do not pipe output into a pager\n"
957 " --dump-configuration-items Dump understood unit configuration items\n"
958 " --unit=UNIT Set default unit\n"
959 " --system Run a system instance, even if PID != 1\n"
960 " --user Run a user instance\n"
961 " --dump-core[=0|1] Dump core on crash\n"
962 " --crash-shell[=0|1] Run shell on crash\n"
963 " --confirm-spawn[=0|1] Ask for confirmation when spawning processes\n"
964 " --show-status[=0|1] Show status updates on the console during bootup\n"
965 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
966 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
967 " --log-color[=0|1] Highlight important log messages\n"
968 " --log-location[=0|1] Include code location in log messages\n"
969 " --default-standard-output= Set default standard output for services\n"
970 " --default-standard-error= Set default standard error output for services\n",
971 program_invocation_short_name);
976 static int version(void) {
977 puts(PACKAGE_STRING);
978 puts(SYSTEMD_FEATURES);
983 static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
992 r = manager_open_serialization(m, &f);
994 log_error("Failed to create serialization file: %s", strerror(-r));
998 /* Make sure nothing is really destructed when we shut down */
1000 bus_manager_send_reloading(m, true);
1005 log_error("Failed to allocate fd set: %s", strerror(-r));
1009 r = manager_serialize(m, f, fds, switching_root);
1011 log_error("Failed to serialize state: %s", strerror(-r));
1015 if (fseeko(f, 0, SEEK_SET) < 0) {
1016 log_error("Failed to rewind serialization fd: %m");
1020 r = fd_cloexec(fileno(f), false);
1022 log_error("Failed to disable O_CLOEXEC for serialization: %s", strerror(-r));
1026 r = fdset_cloexec(fds, false);
1028 log_error("Failed to disable O_CLOEXEC for serialization fds: %s", strerror(-r));
1046 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1050 assert(saved_rlimit);
1052 /* Save the original RLIMIT_NOFILE so that we can reset it
1053 * later when transitioning from the initrd to the main
1054 * systemd or suchlike. */
1055 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0) {
1056 log_error("Reading RLIMIT_NOFILE failed: %m");
1060 /* Make sure forked processes get the default kernel setting */
1061 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1064 rl = newdup(struct rlimit, saved_rlimit, 1);
1068 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1071 /* Bump up the resource limit for ourselves substantially */
1072 nl.rlim_cur = nl.rlim_max = 64*1024;
1073 r = setrlimit_closest(RLIMIT_NOFILE, &nl);
1075 log_error("Setting RLIMIT_NOFILE failed: %s", strerror(-r));
1082 static void test_mtab(void) {
1084 static const char ok[] =
1085 "/proc/self/mounts\0"
1087 "../proc/self/mounts\0"
1090 _cleanup_free_ char *p = NULL;
1093 /* Check that /etc/mtab is a symlink to the right place or
1094 * non-existing. But certainly not a file, or a symlink to
1095 * some weird place... */
1097 r = readlink_malloc("/etc/mtab", &p);
1100 if (r >= 0 && nulstr_contains(ok, p))
1103 log_warning("/etc/mtab is not a symlink or not pointing to /proc/self/mounts. "
1104 "This is not supported anymore. "
1105 "Please make sure to replace this file by a symlink to avoid incorrect or misleading mount(8) output.");
1108 static void test_usr(void) {
1110 /* Check that /usr is not a separate fs */
1112 if (dir_is_empty("/usr") <= 0)
1115 log_warning("/usr appears to be on its own filesytem and is not already mounted. This is not a supported setup. "
1116 "Some things will probably break (sometimes even silently) in mysterious ways. "
1117 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1120 static int initialize_join_controllers(void) {
1121 /* By default, mount "cpu" + "cpuacct" together, and "net_cls"
1122 * + "net_prio". We'd like to add "cpuset" to the mix, but
1123 * "cpuset" does't really work for groups with no initialized
1126 arg_join_controllers = new(char**, 3);
1127 if (!arg_join_controllers)
1130 arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL);
1131 arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL);
1132 arg_join_controllers[2] = NULL;
1134 if (!arg_join_controllers[0] || !arg_join_controllers[1]) {
1135 free_join_controllers();
1142 static int enforce_syscall_archs(Set *archs) {
1144 scmp_filter_ctx *seccomp;
1149 seccomp = seccomp_init(SCMP_ACT_ALLOW);
1153 SET_FOREACH(id, arg_syscall_archs, i) {
1154 r = seccomp_arch_add(seccomp, PTR_TO_UINT32(id) - 1);
1158 log_error("Failed to add architecture to seccomp: %s", strerror(-r));
1163 r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
1165 log_error("Failed to unset NO_NEW_PRIVS: %s", strerror(-r));
1169 r = seccomp_load(seccomp);
1171 log_error("Failed to add install architecture seccomp: %s", strerror(-r));
1174 seccomp_release(seccomp);
1181 static int status_welcome(void) {
1182 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1185 r = parse_env_file("/etc/os-release", NEWLINE,
1186 "PRETTY_NAME", &pretty_name,
1187 "ANSI_COLOR", &ansi_color,
1190 r = parse_env_file("/usr/lib/os-release", NEWLINE,
1191 "PRETTY_NAME", &pretty_name,
1192 "ANSI_COLOR", &ansi_color,
1196 if (r < 0 && r != -ENOENT)
1197 log_warning("Failed to read os-release file: %s", strerror(-r));
1199 return status_printf(NULL, false, false,
1200 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1201 isempty(ansi_color) ? "1" : ansi_color,
1202 isempty(pretty_name) ? "Linux" : pretty_name);
1205 static int write_container_id(void) {
1208 c = getenv("container");
1212 return write_string_file("/run/systemd/container", c);
1215 int main(int argc, char *argv[]) {
1217 int r, retval = EXIT_FAILURE;
1218 usec_t before_startup, after_startup;
1219 char timespan[FORMAT_TIMESPAN_MAX];
1221 bool reexecute = false;
1222 const char *shutdown_verb = NULL;
1223 dual_timestamp initrd_timestamp = { 0ULL, 0ULL };
1224 dual_timestamp userspace_timestamp = { 0ULL, 0ULL };
1225 dual_timestamp kernel_timestamp = { 0ULL, 0ULL };
1226 dual_timestamp security_start_timestamp = { 0ULL, 0ULL };
1227 dual_timestamp security_finish_timestamp = { 0ULL, 0ULL };
1228 static char systemd[] = "systemd";
1229 bool skip_setup = false;
1231 bool loaded_policy = false;
1232 bool arm_reboot_watchdog = false;
1233 bool queue_default_job = false;
1234 bool empty_etc = false;
1235 char *switch_root_dir = NULL, *switch_root_init = NULL;
1236 static struct rlimit saved_rlimit_nofile = { 0, 0 };
1238 #ifdef HAVE_SYSV_COMPAT
1239 if (getpid() != 1 && strstr(program_invocation_short_name, "init")) {
1240 /* This is compatibility support for SysV, where
1241 * calling init as a user is identical to telinit. */
1244 execv(SYSTEMCTL_BINARY_PATH, argv);
1245 log_error("Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1250 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
1251 dual_timestamp_get(&userspace_timestamp);
1253 /* Determine if this is a reexecution or normal bootup. We do
1254 * the full command line parsing much later, so let's just
1255 * have a quick peek here. */
1256 if (strv_find(argv+1, "--deserialize"))
1259 /* If we have switched root, do all the special setup
1261 if (strv_find(argv+1, "--switched-root"))
1264 /* If we get started via the /sbin/init symlink then we are
1265 called 'init'. After a subsequent reexecution we are then
1266 called 'systemd'. That is confusing, hence let's call us
1267 systemd right-away. */
1268 program_invocation_short_name = systemd;
1269 prctl(PR_SET_NAME, systemd);
1274 log_show_color(isatty(STDERR_FILENO) > 0);
1275 log_set_upgrade_syslog_to_journal(true);
1277 /* Disable the umask logic */
1281 if (getpid() == 1 && detect_container(NULL) <= 0) {
1283 /* Running outside of a container as PID 1 */
1284 arg_running_as = SYSTEMD_SYSTEM;
1286 log_set_target(LOG_TARGET_KMSG);
1290 initrd_timestamp = userspace_timestamp;
1293 mount_setup_early();
1294 dual_timestamp_get(&security_start_timestamp);
1295 if (selinux_setup(&loaded_policy) < 0)
1297 if (ima_setup() < 0)
1299 if (smack_setup(&loaded_policy) < 0)
1301 dual_timestamp_get(&security_finish_timestamp);
1304 if (label_init(NULL) < 0)
1308 if (clock_is_localtime() > 0) {
1312 * The very first call of settimeofday() also does a time warp in the kernel.
1314 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on
1315 * external tools to take care of maintaining the RTC and do all adjustments.
1316 * This matches the behavior of Windows, which leaves the RTC alone if the
1317 * registry tells that the RTC runs in UTC.
1319 r = clock_set_timezone(&min);
1321 log_error("Failed to apply local time delta, ignoring: %s", strerror(-r));
1323 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1324 } else if (!in_initrd()) {
1326 * Do a dummy very first call to seal the kernel's time warp magic.
1328 * Do not call this this from inside the initrd. The initrd might not
1329 * carry /etc/adjtime with LOCAL, but the real system could be set up
1330 * that way. In such case, we need to delay the time-warp or the sealing
1331 * until we reach the real system.
1333 * Do no set the kernel's timezone. The concept of local time cannot
1334 * be supported reliably, the time will jump or be incorrect at every daylight
1335 * saving time change. All kernel local time concepts will be treated
1338 clock_reset_timewarp();
1342 /* Set the default for later on, but don't actually
1343 * open the logs like this for now. Note that if we
1344 * are transitioning from the initrd there might still
1345 * be journal fd open, and we shouldn't attempt
1346 * opening that before we parsed /proc/cmdline which
1347 * might redirect output elsewhere. */
1348 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
1350 } else if (getpid() == 1) {
1351 /* Running inside a container, as PID 1 */
1352 arg_running_as = SYSTEMD_SYSTEM;
1353 log_set_target(LOG_TARGET_CONSOLE);
1354 log_close_console(); /* force reopen of /dev/console */
1357 /* For the later on, see above... */
1358 log_set_target(LOG_TARGET_JOURNAL);
1360 /* clear the kernel timestamp,
1361 * because we are in a container */
1362 kernel_timestamp.monotonic = 0ULL;
1363 kernel_timestamp.realtime = 0ULL;
1366 /* Running as user instance */
1367 arg_running_as = SYSTEMD_USER;
1368 log_set_target(LOG_TARGET_AUTO);
1371 /* clear the kernel timestamp,
1372 * because we are not PID 1 */
1373 kernel_timestamp.monotonic = 0ULL;
1374 kernel_timestamp.realtime = 0ULL;
1377 /* Initialize default unit */
1378 r = set_default_unit(SPECIAL_DEFAULT_TARGET);
1380 log_error("Failed to set default unit %s: %s", SPECIAL_DEFAULT_TARGET, strerror(-r));
1384 r = initialize_join_controllers();
1388 /* Mount /proc, /sys and friends, so that /proc/cmdline and
1389 * /proc/$PID/fd is available. */
1390 if (getpid() == 1) {
1391 r = mount_setup(loaded_policy);
1396 /* Reset all signal handlers. */
1397 assert_se(reset_all_signal_handlers() == 0);
1399 ignore_signals(SIGNALS_IGNORE, -1);
1401 if (parse_config_file() < 0)
1404 if (arg_running_as == SYSTEMD_SYSTEM)
1405 if (parse_proc_cmdline(parse_proc_cmdline_item) < 0)
1408 /* Note that this also parses bits from the kernel command
1409 * line, including "debug". */
1410 log_parse_environment();
1412 if (parse_argv(argc, argv) < 0)
1415 if (arg_action == ACTION_TEST &&
1417 log_error("Don't run test mode as root.");
1421 if (arg_running_as == SYSTEMD_USER &&
1422 arg_action == ACTION_RUN &&
1424 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
1428 if (arg_running_as == SYSTEMD_SYSTEM &&
1429 arg_action == ACTION_RUN &&
1430 running_in_chroot() > 0) {
1431 log_error("Cannot be run in a chroot() environment.");
1435 if (arg_action == ACTION_TEST)
1438 pager_open_if_enabled();
1440 if (arg_action == ACTION_HELP) {
1443 } else if (arg_action == ACTION_VERSION) {
1446 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
1447 unit_dump_config_items(stdout);
1448 retval = EXIT_SUCCESS;
1450 } else if (arg_action == ACTION_DONE) {
1451 retval = EXIT_SUCCESS;
1455 if (arg_running_as == SYSTEMD_USER &&
1456 !getenv("XDG_RUNTIME_DIR")) {
1457 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
1461 assert_se(arg_action == ACTION_RUN || arg_action == ACTION_TEST);
1463 /* Close logging fds, in order not to confuse fdset below */
1466 /* Remember open file descriptors for later deserialization */
1467 r = fdset_new_fill(&fds);
1469 log_error("Failed to allocate fd set: %s", strerror(-r));
1472 fdset_cloexec(fds, true);
1474 if (arg_serialization)
1475 assert_se(fdset_remove(fds, fileno(arg_serialization)) >= 0);
1477 if (arg_running_as == SYSTEMD_SYSTEM)
1478 /* Become a session leader if we aren't one yet. */
1481 /* Move out of the way, so that we won't block unmounts */
1482 assert_se(chdir("/") == 0);
1484 /* Reset the console, but only if this is really init and we
1485 * are freshly booted */
1486 if (arg_running_as == SYSTEMD_SYSTEM && arg_action == ACTION_RUN) {
1488 /* If we are init, we connect stdin/stdout/stderr to
1489 * /dev/null and make sure we don't have a controlling
1493 if (getpid() == 1 && !skip_setup)
1497 /* Open the logging devices, if possible and necessary */
1500 if (arg_show_status == _SHOW_STATUS_UNSET)
1501 arg_show_status = SHOW_STATUS_YES;
1503 /* Make sure we leave a core dump without panicing the
1505 if (getpid() == 1) {
1506 install_crash_handler();
1508 r = mount_cgroup_controllers(arg_join_controllers);
1513 if (arg_running_as == SYSTEMD_SYSTEM) {
1514 const char *virtualization = NULL;
1516 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1517 arg_action == ACTION_TEST ? "test " : "" );
1519 detect_virtualization(&virtualization);
1521 log_info("Detected virtualization '%s'.", virtualization);
1523 write_container_id();
1525 log_info("Detected architecture '%s'.", architecture_to_string(uname_architecture()));
1528 log_info("Running in initial RAM disk.");
1530 /* Let's check whether /etc is already populated. We
1531 * don't actually really check for that, but use
1532 * /etc/machine-id as flag file. This allows container
1533 * managers and installers to provision a couple of
1534 * files already. If the container manager wants to
1535 * provision the machine ID itself it should pass
1536 * $container_uuid to PID 1.*/
1538 empty_etc = access("/etc/machine-id", F_OK) < 0;
1540 log_info("Running with unpopulated /etc.");
1542 _cleanup_free_ char *t;
1544 t = uid_to_name(getuid());
1545 log_debug(PACKAGE_STRING " running in %suser mode for user "UID_FMT"/%s. (" SYSTEMD_FEATURES ")",
1546 arg_action == ACTION_TEST ? " test" : "", getuid(), t);
1549 if (arg_running_as == SYSTEMD_SYSTEM && !skip_setup) {
1550 if (arg_show_status > 0 || plymouth_running())
1557 machine_id_setup(NULL);
1564 if (arg_running_as == SYSTEMD_SYSTEM && arg_runtime_watchdog > 0)
1565 watchdog_set_timeout(&arg_runtime_watchdog);
1567 if (arg_timer_slack_nsec != NSEC_INFINITY)
1568 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1569 log_error("Failed to adjust timer slack: %m");
1571 if (arg_capability_bounding_set_drop) {
1572 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set_drop);
1574 log_error("Failed to drop capability bounding set of usermode helpers: %s", strerror(-r));
1577 r = capability_bounding_set_drop(arg_capability_bounding_set_drop, true);
1579 log_error("Failed to drop capability bounding set: %s", strerror(-r));
1584 if (arg_syscall_archs) {
1585 r = enforce_syscall_archs(arg_syscall_archs);
1590 if (arg_running_as == SYSTEMD_USER) {
1591 /* Become reaper of our children */
1592 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0) {
1593 log_warning("Failed to make us a subreaper: %m");
1594 if (errno == EINVAL)
1595 log_info("Perhaps the kernel version is too old (< 3.4?)");
1599 if (arg_running_as == SYSTEMD_SYSTEM) {
1600 bump_rlimit_nofile(&saved_rlimit_nofile);
1603 r = unit_file_preset_all(UNIT_FILE_SYSTEM, false, NULL, UNIT_FILE_PRESET_FULL, false, NULL, 0);
1605 log_warning("Failed to populate /etc with preset unit settings, ignoring: %s", strerror(-r));
1607 log_info("Populated /etc with preset unit settings.");
1611 r = manager_new(arg_running_as, arg_action == ACTION_TEST, &m);
1613 log_error("Failed to allocate manager object: %s", strerror(-r));
1617 m->confirm_spawn = arg_confirm_spawn;
1618 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
1619 m->default_std_output = arg_default_std_output;
1620 m->default_std_error = arg_default_std_error;
1621 m->default_restart_usec = arg_default_restart_usec;
1622 m->default_timeout_start_usec = arg_default_timeout_start_usec;
1623 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
1624 m->default_start_limit_interval = arg_default_start_limit_interval;
1625 m->default_start_limit_burst = arg_default_start_limit_burst;
1626 m->default_cpu_accounting = arg_default_cpu_accounting;
1627 m->default_blockio_accounting = arg_default_blockio_accounting;
1628 m->default_memory_accounting = arg_default_memory_accounting;
1629 m->runtime_watchdog = arg_runtime_watchdog;
1630 m->shutdown_watchdog = arg_shutdown_watchdog;
1631 m->userspace_timestamp = userspace_timestamp;
1632 m->kernel_timestamp = kernel_timestamp;
1633 m->initrd_timestamp = initrd_timestamp;
1634 m->security_start_timestamp = security_start_timestamp;
1635 m->security_finish_timestamp = security_finish_timestamp;
1637 manager_set_default_rlimits(m, arg_default_rlimit);
1638 manager_environment_add(m, NULL, arg_default_environment);
1639 manager_set_show_status(m, arg_show_status);
1640 manager_set_first_boot(m, empty_etc);
1642 /* Remember whether we should queue the default job */
1643 queue_default_job = !arg_serialization || arg_switched_root;
1645 before_startup = now(CLOCK_MONOTONIC);
1647 r = manager_startup(m, arg_serialization, fds);
1649 log_error("Failed to fully start up daemon: %s", strerror(-r));
1651 /* This will close all file descriptors that were opened, but
1652 * not claimed by any unit. */
1656 if (arg_serialization) {
1657 fclose(arg_serialization);
1658 arg_serialization = NULL;
1661 if (queue_default_job) {
1662 _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL;
1663 Unit *target = NULL;
1664 Job *default_unit_job;
1666 log_debug("Activating default unit: %s", arg_default_unit);
1668 r = manager_load_unit(m, arg_default_unit, NULL, &error, &target);
1670 log_error("Failed to load default target: %s", bus_error_message(&error, r));
1671 else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND)
1672 log_error("Failed to load default target: %s", strerror(-target->load_error));
1673 else if (target->load_state == UNIT_MASKED)
1674 log_error("Default target masked.");
1676 if (!target || target->load_state != UNIT_LOADED) {
1677 log_info("Trying to load rescue target...");
1679 r = manager_load_unit(m, SPECIAL_RESCUE_TARGET, NULL, &error, &target);
1681 log_error("Failed to load rescue target: %s", bus_error_message(&error, r));
1683 } else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND) {
1684 log_error("Failed to load rescue target: %s", strerror(-target->load_error));
1686 } else if (target->load_state == UNIT_MASKED) {
1687 log_error("Rescue target masked.");
1692 assert(target->load_state == UNIT_LOADED);
1694 if (arg_action == ACTION_TEST) {
1695 printf("-> By units:\n");
1696 manager_dump_units(m, stdout, "\t");
1699 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, false, &error, &default_unit_job);
1701 log_debug("Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1703 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, false, &error, &default_unit_job);
1705 log_error("Failed to start default target: %s", bus_error_message(&error, r));
1709 log_error("Failed to isolate default target: %s", bus_error_message(&error, r));
1713 m->default_unit_job_id = default_unit_job->id;
1715 after_startup = now(CLOCK_MONOTONIC);
1716 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
1717 "Loaded units and determined initial transaction in %s.",
1718 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 0));
1720 if (arg_action == ACTION_TEST) {
1721 printf("-> By jobs:\n");
1722 manager_dump_jobs(m, stdout, "\t");
1723 retval = EXIT_SUCCESS;
1729 r = manager_loop(m);
1731 log_error("Failed to run mainloop: %s", strerror(-r));
1735 switch (m->exit_code) {
1738 retval = EXIT_SUCCESS;
1742 case MANAGER_RELOAD:
1743 log_info("Reloading.");
1744 r = manager_reload(m);
1746 log_error("Failed to reload: %s", strerror(-r));
1749 case MANAGER_REEXECUTE:
1751 if (prepare_reexecute(m, &arg_serialization, &fds, false) < 0)
1755 log_notice("Reexecuting.");
1758 case MANAGER_SWITCH_ROOT:
1759 /* Steal the switch root parameters */
1760 switch_root_dir = m->switch_root;
1761 switch_root_init = m->switch_root_init;
1762 m->switch_root = m->switch_root_init = NULL;
1764 if (!switch_root_init)
1765 if (prepare_reexecute(m, &arg_serialization, &fds, true) < 0)
1769 log_notice("Switching root.");
1772 case MANAGER_REBOOT:
1773 case MANAGER_POWEROFF:
1775 case MANAGER_KEXEC: {
1776 static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
1777 [MANAGER_REBOOT] = "reboot",
1778 [MANAGER_POWEROFF] = "poweroff",
1779 [MANAGER_HALT] = "halt",
1780 [MANAGER_KEXEC] = "kexec"
1783 assert_se(shutdown_verb = table[m->exit_code]);
1784 arm_reboot_watchdog = m->exit_code == MANAGER_REBOOT;
1786 log_notice("Shutting down.");
1791 assert_not_reached("Unknown exit code.");
1803 for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++) {
1804 free(arg_default_rlimit[j]);
1805 arg_default_rlimit[j] = NULL;
1808 free(arg_default_unit);
1809 arg_default_unit = NULL;
1811 free_join_controllers();
1813 strv_free(arg_default_environment);
1814 arg_default_environment = NULL;
1816 set_free(arg_syscall_archs);
1817 arg_syscall_archs = NULL;
1823 unsigned i, args_size;
1826 /* Close and disarm the watchdog, so that the new
1827 * instance can reinitialize it, but doesn't get
1828 * rebooted while we do that */
1829 watchdog_close(true);
1831 /* Reset the RLIMIT_NOFILE to the kernel default, so
1832 * that the new systemd can pass the kernel default to
1833 * its child processes */
1834 if (saved_rlimit_nofile.rlim_cur > 0)
1835 setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile);
1837 if (switch_root_dir) {
1838 /* Kill all remaining processes from the
1839 * initrd, but don't wait for them, so that we
1840 * can handle the SIGCHLD for them after
1842 broadcast_signal(SIGTERM, false, true);
1844 /* And switch root */
1845 r = switch_root(switch_root_dir);
1847 log_error("Failed to switch root, ignoring: %s", strerror(-r));
1850 args_size = MAX(6, argc+1);
1851 args = newa(const char*, args_size);
1853 if (!switch_root_init) {
1856 /* First try to spawn ourselves with the right
1857 * path, and with full serialization. We do
1858 * this only if the user didn't specify an
1859 * explicit init to spawn. */
1861 assert(arg_serialization);
1864 snprintf(sfd, sizeof(sfd), "%i", fileno(arg_serialization));
1868 args[i++] = SYSTEMD_BINARY_PATH;
1869 if (switch_root_dir)
1870 args[i++] = "--switched-root";
1871 args[i++] = arg_running_as == SYSTEMD_SYSTEM ? "--system" : "--user";
1872 args[i++] = "--deserialize";
1876 /* do not pass along the environment we inherit from the kernel or initrd */
1877 if (switch_root_dir)
1880 assert(i <= args_size);
1881 execv(args[0], (char* const*) args);
1884 /* Try the fallback, if there is any, without any
1885 * serialization. We pass the original argv[] and
1886 * envp[]. (Well, modulo the ordering changes due to
1887 * getopt() in argv[], and some cleanups in envp[],
1888 * but let's hope that doesn't matter.) */
1890 if (arg_serialization) {
1891 fclose(arg_serialization);
1892 arg_serialization = NULL;
1900 /* Reopen the console */
1901 make_console_stdio();
1903 for (j = 1, i = 1; j < (unsigned) argc; j++)
1904 args[i++] = argv[j];
1906 assert(i <= args_size);
1908 /* reenable any blocked signals, especially important
1909 * if we switch from initial ramdisk to init=... */
1910 reset_all_signal_handlers();
1912 assert_se(sigemptyset(&ss) == 0);
1913 assert_se(sigprocmask(SIG_SETMASK, &ss, NULL) == 0);
1915 if (switch_root_init) {
1916 args[0] = switch_root_init;
1917 execv(args[0], (char* const*) args);
1918 log_warning("Failed to execute configured init, trying fallback: %m");
1921 args[0] = "/sbin/init";
1922 execv(args[0], (char* const*) args);
1924 if (errno == ENOENT) {
1925 log_warning("No /sbin/init, trying fallback");
1927 args[0] = "/bin/sh";
1929 execv(args[0], (char* const*) args);
1930 log_error("Failed to execute /bin/sh, giving up: %m");
1932 log_warning("Failed to execute /sbin/init, giving up: %m");
1935 if (arg_serialization) {
1936 fclose(arg_serialization);
1937 arg_serialization = NULL;
1945 #ifdef HAVE_VALGRIND_VALGRIND_H
1946 /* If we are PID 1 and running under valgrind, then let's exit
1947 * here explicitly. valgrind will only generate nice output on
1948 * exit(), not on exec(), hence let's do the former not the
1950 if (getpid() == 1 && RUNNING_ON_VALGRIND)
1954 if (shutdown_verb) {
1955 char log_level[DECIMAL_STR_MAX(int) + 1];
1956 const char* command_line[9] = {
1957 SYSTEMD_SHUTDOWN_BINARY_PATH,
1959 "--log-level", log_level,
1963 _cleanup_strv_free_ char **env_block = NULL;
1965 assert(command_line[pos] == NULL);
1966 env_block = strv_copy(environ);
1968 snprintf(log_level, sizeof(log_level), "%d", log_get_max_level());
1970 switch (log_get_target()) {
1971 case LOG_TARGET_KMSG:
1972 case LOG_TARGET_JOURNAL_OR_KMSG:
1973 case LOG_TARGET_SYSLOG_OR_KMSG:
1974 command_line[pos++] = "kmsg";
1977 case LOG_TARGET_CONSOLE:
1979 command_line[pos++] = "console";
1983 if (log_get_show_color())
1984 command_line[pos++] = "--log-color";
1986 if (log_get_show_location())
1987 command_line[pos++] = "--log-location";
1989 assert(pos < ELEMENTSOF(command_line));
1991 if (arm_reboot_watchdog && arg_shutdown_watchdog > 0) {
1994 /* If we reboot let's set the shutdown
1995 * watchdog and tell the shutdown binary to
1996 * repeatedly ping it */
1997 watchdog_set_timeout(&arg_shutdown_watchdog);
1998 watchdog_close(false);
2000 /* Tell the binary how often to ping, ignore failure */
2001 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
2002 strv_push(&env_block, e);
2004 watchdog_close(true);
2006 /* Avoid the creation of new processes forked by the
2007 * kernel; at this point, we will not listen to the
2009 if (detect_container(NULL) <= 0)
2010 cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
2012 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
2013 log_error("Failed to execute shutdown binary, %s: %m",
2014 getpid() == 1 ? "freezing" : "quitting");