1 /* SPDX-License-Identifier: LGPL-2.1+ */
10 #include <sys/prctl.h>
11 #include <sys/reboot.h>
17 #if HAVE_VALGRIND_VALGRIND_H
18 #include <valgrind/valgrind.h>
22 #include "sd-daemon.h"
23 #include "sd-messages.h"
25 #include "alloc-util.h"
26 #include "architecture.h"
28 #include "bus-error.h"
30 #include "capability-util.h"
31 #include "clock-util.h"
32 #include "conf-parser.h"
33 #include "cpu-set-util.h"
35 #include "dbus-manager.h"
37 #include "emergency-action.h"
39 #include "exit-status.h"
43 #include "format-util.h"
45 #include "hostname-setup.h"
46 #include "ima-setup.h"
48 #include "kmod-setup.h"
49 #include "load-fragment.h"
51 #include "loopback-setup.h"
52 #include "machine-id-setup.h"
55 #include "mount-setup.h"
58 #include "parse-util.h"
59 #include "path-util.h"
60 #include "proc-cmdline.h"
61 #include "process-util.h"
62 #include "raw-clone.h"
63 #include "rlimit-util.h"
65 #include "seccomp-util.h"
67 #include "selinux-setup.h"
68 #include "selinux-util.h"
69 #include "signal-util.h"
70 #include "smack-setup.h"
72 #include "stat-util.h"
73 #include "stdio-util.h"
75 #include "switch-root.h"
76 #include "sysctl-util.h"
77 #include "terminal-util.h"
78 #include "umask-util.h"
79 #include "user-util.h"
89 ACTION_DUMP_CONFIGURATION_ITEMS
,
90 ACTION_DUMP_BUS_PROPERTIES
,
91 } arg_action
= ACTION_RUN
;
92 static char *arg_default_unit
= NULL
;
93 static bool arg_system
= false;
94 static bool arg_dump_core
= true;
95 static int arg_crash_chvt
= -1;
96 static bool arg_crash_shell
= false;
97 static bool arg_crash_reboot
= false;
98 static char *arg_confirm_spawn
= NULL
;
99 static ShowStatus arg_show_status
= _SHOW_STATUS_INVALID
;
100 static bool arg_switched_root
= false;
101 static bool arg_no_pager
= false;
102 static bool arg_service_watchdogs
= true;
103 static char ***arg_join_controllers
= NULL
;
104 static ExecOutput arg_default_std_output
= EXEC_OUTPUT_JOURNAL
;
105 static ExecOutput arg_default_std_error
= EXEC_OUTPUT_INHERIT
;
106 static usec_t arg_default_restart_usec
= DEFAULT_RESTART_USEC
;
107 static usec_t arg_default_timeout_start_usec
= DEFAULT_TIMEOUT_USEC
;
108 static usec_t arg_default_timeout_stop_usec
= DEFAULT_TIMEOUT_USEC
;
109 static usec_t arg_default_start_limit_interval
= DEFAULT_START_LIMIT_INTERVAL
;
110 static unsigned arg_default_start_limit_burst
= DEFAULT_START_LIMIT_BURST
;
111 static usec_t arg_runtime_watchdog
= 0;
112 static usec_t arg_shutdown_watchdog
= 10 * USEC_PER_MINUTE
;
113 static char *arg_early_core_pattern
= NULL
;
114 static char *arg_watchdog_device
= NULL
;
115 static char **arg_default_environment
= NULL
;
116 static struct rlimit
*arg_default_rlimit
[_RLIMIT_MAX
] = {};
117 static uint64_t arg_capability_bounding_set
= CAP_ALL
;
118 static bool arg_no_new_privs
= false;
119 static nsec_t arg_timer_slack_nsec
= NSEC_INFINITY
;
120 static usec_t arg_default_timer_accuracy_usec
= 1 * USEC_PER_MINUTE
;
121 static Set
* arg_syscall_archs
= NULL
;
122 static FILE* arg_serialization
= NULL
;
123 static bool arg_default_cpu_accounting
= false;
124 static bool arg_default_io_accounting
= false;
125 static bool arg_default_ip_accounting
= false;
126 static bool arg_default_blockio_accounting
= false;
127 static bool arg_default_memory_accounting
= MEMORY_ACCOUNTING_DEFAULT
;
128 static bool arg_default_tasks_accounting
= true;
129 static uint64_t arg_default_tasks_max
= UINT64_MAX
;
130 static sd_id128_t arg_machine_id
= {};
131 static EmergencyAction arg_cad_burst_action
= EMERGENCY_ACTION_REBOOT_FORCE
;
133 _noreturn_
static void freeze_or_reboot(void) {
135 if (arg_crash_reboot
) {
136 log_notice("Rebooting in 10s...");
139 log_notice("Rebooting now...");
140 (void) reboot(RB_AUTOBOOT
);
141 log_emergency_errno(errno
, "Failed to reboot: %m");
144 log_emergency("Freezing execution.");
148 _noreturn_
static void crash(int sig
) {
152 if (getpid_cached() != 1)
153 /* Pass this on immediately, if this is not PID 1 */
155 else if (!arg_dump_core
)
156 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig
));
158 sa
= (struct sigaction
) {
159 .sa_handler
= nop_signal_handler
,
160 .sa_flags
= SA_NOCLDSTOP
|SA_RESTART
,
163 /* We want to wait for the core process, hence let's enable SIGCHLD */
164 (void) sigaction(SIGCHLD
, &sa
, NULL
);
166 pid
= raw_clone(SIGCHLD
);
168 log_emergency_errno(errno
, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig
));
170 /* Enable default signal handler for core dump */
172 sa
= (struct sigaction
) {
173 .sa_handler
= SIG_DFL
,
175 (void) sigaction(sig
, &sa
, NULL
);
177 /* Don't limit the coredump size */
178 (void) setrlimit(RLIMIT_CORE
, &RLIMIT_MAKE_CONST(RLIM_INFINITY
));
180 /* Just to be sure... */
183 /* Raise the signal again */
185 (void) kill(pid
, sig
); /* raise() would kill the parent */
187 assert_not_reached("We shouldn't be here...");
193 /* Order things nicely. */
194 r
= wait_for_terminate(pid
, &status
);
196 log_emergency_errno(r
, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig
));
197 else if (status
.si_code
!= CLD_DUMPED
)
198 log_emergency("Caught <%s>, core dump failed (child "PID_FMT
", code=%s, status=%i/%s).",
199 signal_to_string(sig
),
200 pid
, sigchld_code_to_string(status
.si_code
),
202 strna(status
.si_code
== CLD_EXITED
203 ? exit_status_to_string(status
.si_status
, EXIT_STATUS_MINIMAL
)
204 : signal_to_string(status
.si_status
)));
206 log_emergency("Caught <%s>, dumped core as pid "PID_FMT
".", signal_to_string(sig
), pid
);
210 if (arg_crash_chvt
>= 0)
211 (void) chvt(arg_crash_chvt
);
213 sa
= (struct sigaction
) {
214 .sa_handler
= SIG_IGN
,
215 .sa_flags
= SA_NOCLDSTOP
|SA_NOCLDWAIT
|SA_RESTART
,
218 /* Let the kernel reap children for us */
219 (void) sigaction(SIGCHLD
, &sa
, NULL
);
221 if (arg_crash_shell
) {
222 log_notice("Executing crash shell in 10s...");
225 pid
= raw_clone(SIGCHLD
);
227 log_emergency_errno(errno
, "Failed to fork off crash shell: %m");
230 (void) make_console_stdio();
231 (void) execle("/bin/sh", "/bin/sh", NULL
, environ
);
233 log_emergency_errno(errno
, "execle() failed: %m");
236 log_info("Spawned crash shell as PID "PID_FMT
".", pid
);
237 (void) wait_for_terminate(pid
, NULL
);
244 static void install_crash_handler(void) {
245 static const struct sigaction sa
= {
247 .sa_flags
= SA_NODEFER
, /* So that we can raise the signal again from the signal handler */
251 /* We ignore the return value here, since, we don't mind if we
252 * cannot set up a crash handler */
253 r
= sigaction_many(&sa
, SIGNALS_CRASH_HANDLER
, -1);
255 log_debug_errno(r
, "I had trouble setting up the crash handler, ignoring: %m");
258 static int console_setup(void) {
259 _cleanup_close_
int tty_fd
= -1;
262 tty_fd
= open_terminal("/dev/console", O_WRONLY
|O_NOCTTY
|O_CLOEXEC
);
264 return log_error_errno(tty_fd
, "Failed to open /dev/console: %m");
266 /* We don't want to force text mode. plymouth may be showing
267 * pictures already from initrd. */
268 r
= reset_terminal_fd(tty_fd
, false);
270 return log_error_errno(r
, "Failed to reset /dev/console: %m");
275 static int parse_crash_chvt(const char *value
) {
278 if (safe_atoi(value
, &arg_crash_chvt
) >= 0)
281 b
= parse_boolean(value
);
286 arg_crash_chvt
= 0; /* switch to where kmsg goes */
288 arg_crash_chvt
= -1; /* turn off switching */
293 static int parse_confirm_spawn(const char *value
, char **console
) {
297 r
= value
? parse_boolean(value
) : 1;
303 if (r
> 0) /* on with default tty */
304 s
= strdup("/dev/console");
305 else if (is_path(value
)) /* on with fully qualified path */
307 else /* on with only a tty file name, not a fully qualified path */
308 s
= strjoin("/dev/", value
);
316 static int set_machine_id(const char *m
) {
320 if (sd_id128_from_string(m
, &t
) < 0)
323 if (sd_id128_is_null(t
))
330 static int parse_proc_cmdline_item(const char *key
, const char *value
, void *data
) {
336 if (STR_IN_SET(key
, "systemd.unit", "rd.systemd.unit")) {
338 if (proc_cmdline_value_missing(key
, value
))
341 if (!unit_name_is_valid(value
, UNIT_NAME_PLAIN
|UNIT_NAME_INSTANCE
))
342 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key
, value
);
343 else if (in_initrd() == !!startswith(key
, "rd.")) {
344 if (free_and_strdup(&arg_default_unit
, value
) < 0)
348 } else if (proc_cmdline_key_streq(key
, "systemd.dump_core")) {
350 r
= value
? parse_boolean(value
) : true;
352 log_warning("Failed to parse dump core switch %s. Ignoring.", value
);
356 } else if (proc_cmdline_key_streq(key
, "systemd.early_core_pattern")) {
358 if (proc_cmdline_value_missing(key
, value
))
361 if (path_is_absolute(value
))
362 (void) parse_path_argument_and_warn(value
, false, &arg_early_core_pattern
);
364 log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value
);
366 } else if (proc_cmdline_key_streq(key
, "systemd.crash_chvt")) {
369 arg_crash_chvt
= 0; /* turn on */
370 else if (parse_crash_chvt(value
) < 0)
371 log_warning("Failed to parse crash chvt switch %s. Ignoring.", value
);
373 } else if (proc_cmdline_key_streq(key
, "systemd.crash_shell")) {
375 r
= value
? parse_boolean(value
) : true;
377 log_warning("Failed to parse crash shell switch %s. Ignoring.", value
);
381 } else if (proc_cmdline_key_streq(key
, "systemd.crash_reboot")) {
383 r
= value
? parse_boolean(value
) : true;
385 log_warning("Failed to parse crash reboot switch %s. Ignoring.", value
);
387 arg_crash_reboot
= r
;
389 } else if (proc_cmdline_key_streq(key
, "systemd.confirm_spawn")) {
392 r
= parse_confirm_spawn(value
, &s
);
394 log_warning_errno(r
, "Failed to parse confirm_spawn switch %s. Ignoring.", value
);
396 free(arg_confirm_spawn
);
397 arg_confirm_spawn
= s
;
400 } else if (proc_cmdline_key_streq(key
, "systemd.service_watchdogs")) {
402 r
= value
? parse_boolean(value
) : true;
404 log_warning("Failed to parse service watchdog switch %s. Ignoring.", value
);
406 arg_service_watchdogs
= r
;
408 } else if (proc_cmdline_key_streq(key
, "systemd.show_status")) {
411 r
= parse_show_status(value
, &arg_show_status
);
413 log_warning("Failed to parse show status switch %s. Ignoring.", value
);
415 arg_show_status
= SHOW_STATUS_YES
;
417 } else if (proc_cmdline_key_streq(key
, "systemd.default_standard_output")) {
419 if (proc_cmdline_value_missing(key
, value
))
422 r
= exec_output_from_string(value
);
424 log_warning("Failed to parse default standard output switch %s. Ignoring.", value
);
426 arg_default_std_output
= r
;
428 } else if (proc_cmdline_key_streq(key
, "systemd.default_standard_error")) {
430 if (proc_cmdline_value_missing(key
, value
))
433 r
= exec_output_from_string(value
);
435 log_warning("Failed to parse default standard error switch %s. Ignoring.", value
);
437 arg_default_std_error
= r
;
439 } else if (streq(key
, "systemd.setenv")) {
441 if (proc_cmdline_value_missing(key
, value
))
444 if (env_assignment_is_valid(value
)) {
447 env
= strv_env_set(arg_default_environment
, value
);
451 arg_default_environment
= env
;
453 log_warning("Environment variable name '%s' is not valid. Ignoring.", value
);
455 } else if (proc_cmdline_key_streq(key
, "systemd.machine_id")) {
457 if (proc_cmdline_value_missing(key
, value
))
460 r
= set_machine_id(value
);
462 log_warning("MachineID '%s' is not valid. Ignoring.", value
);
464 } else if (proc_cmdline_key_streq(key
, "systemd.default_timeout_start_sec")) {
466 if (proc_cmdline_value_missing(key
, value
))
469 r
= parse_sec(value
, &arg_default_timeout_start_usec
);
471 log_warning_errno(r
, "Failed to parse default start timeout: %s, ignoring.", value
);
473 if (arg_default_timeout_start_usec
<= 0)
474 arg_default_timeout_start_usec
= USEC_INFINITY
;
476 } else if (proc_cmdline_key_streq(key
, "systemd.watchdog_device")) {
478 if (proc_cmdline_value_missing(key
, value
))
481 (void) parse_path_argument_and_warn(value
, false, &arg_watchdog_device
);
483 } else if (streq(key
, "quiet") && !value
) {
485 if (arg_show_status
== _SHOW_STATUS_INVALID
)
486 arg_show_status
= SHOW_STATUS_AUTO
;
488 } else if (streq(key
, "debug") && !value
) {
490 /* Note that log_parse_environment() handles 'debug'
491 * too, and sets the log level to LOG_DEBUG. */
493 if (detect_container() > 0)
494 log_set_target(LOG_TARGET_CONSOLE
);
499 /* SysV compatibility */
500 target
= runlevel_to_target(key
);
502 return free_and_strdup(&arg_default_unit
, target
);
508 #define DEFINE_SETTER(name, func, descr) \
509 static int name(const char *unit, \
510 const char *filename, \
512 const char *section, \
513 unsigned section_line, \
514 const char *lvalue, \
516 const char *rvalue, \
528 log_syntax(unit, LOG_ERR, filename, line, r, \
529 "Invalid " descr "'%s': %m", \
535 DEFINE_SETTER(config_parse_level2
, log_set_max_level_from_string
, "log level");
536 DEFINE_SETTER(config_parse_target
, log_set_target_from_string
, "target");
537 DEFINE_SETTER(config_parse_color
, log_show_color_from_string
, "color" );
538 DEFINE_SETTER(config_parse_location
, log_show_location_from_string
, "location");
540 static int config_parse_cpu_affinity2(
542 const char *filename
,
545 unsigned section_line
,
552 _cleanup_cpu_free_ cpu_set_t
*c
= NULL
;
555 ncpus
= parse_cpu_set_and_warn(rvalue
, &c
, unit
, filename
, line
, lvalue
);
559 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus
), c
) < 0)
560 log_warning_errno(errno
, "Failed to set CPU affinity: %m");
565 static int config_parse_show_status(
567 const char *filename
,
570 unsigned section_line
,
578 ShowStatus
*b
= data
;
585 k
= parse_show_status(rvalue
, b
);
587 log_syntax(unit
, LOG_ERR
, filename
, line
, k
, "Failed to parse show status setting, ignoring: %s", rvalue
);
594 static int config_parse_output_restricted(
596 const char *filename
,
599 unsigned section_line
,
606 ExecOutput t
, *eo
= data
;
613 t
= exec_output_from_string(rvalue
);
615 log_syntax(unit
, LOG_ERR
, filename
, line
, 0, "Failed to parse output type, ignoring: %s", rvalue
);
619 if (IN_SET(t
, EXEC_OUTPUT_SOCKET
, EXEC_OUTPUT_NAMED_FD
, EXEC_OUTPUT_FILE
, EXEC_OUTPUT_FILE_APPEND
)) {
620 log_syntax(unit
, LOG_ERR
, filename
, line
, 0, "Standard output types socket, fd:, file:, append: are not supported as defaults, ignoring: %s", rvalue
);
628 static int config_parse_crash_chvt(
630 const char *filename
,
633 unsigned section_line
,
646 r
= parse_crash_chvt(rvalue
);
648 log_syntax(unit
, LOG_ERR
, filename
, line
, r
, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue
);
655 static int parse_config_file(void) {
657 const ConfigTableItem items
[] = {
658 { "Manager", "LogLevel", config_parse_level2
, 0, NULL
},
659 { "Manager", "LogTarget", config_parse_target
, 0, NULL
},
660 { "Manager", "LogColor", config_parse_color
, 0, NULL
},
661 { "Manager", "LogLocation", config_parse_location
, 0, NULL
},
662 { "Manager", "DumpCore", config_parse_bool
, 0, &arg_dump_core
},
663 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt
, 0, NULL
},
664 { "Manager", "CrashChangeVT", config_parse_crash_chvt
, 0, NULL
},
665 { "Manager", "CrashShell", config_parse_bool
, 0, &arg_crash_shell
},
666 { "Manager", "CrashReboot", config_parse_bool
, 0, &arg_crash_reboot
},
667 { "Manager", "ShowStatus", config_parse_show_status
, 0, &arg_show_status
},
668 { "Manager", "CPUAffinity", config_parse_cpu_affinity2
, 0, NULL
},
669 { "Manager", "JoinControllers", config_parse_join_controllers
, 0, &arg_join_controllers
},
670 { "Manager", "RuntimeWatchdogSec", config_parse_sec
, 0, &arg_runtime_watchdog
},
671 { "Manager", "ShutdownWatchdogSec", config_parse_sec
, 0, &arg_shutdown_watchdog
},
672 { "Manager", "WatchdogDevice", config_parse_path
, 0, &arg_watchdog_device
},
673 { "Manager", "CapabilityBoundingSet", config_parse_capability_set
, 0, &arg_capability_bounding_set
},
674 { "Manager", "NoNewPrivileges", config_parse_bool
, 0, &arg_no_new_privs
},
676 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs
, 0, &arg_syscall_archs
},
678 { "Manager", "TimerSlackNSec", config_parse_nsec
, 0, &arg_timer_slack_nsec
},
679 { "Manager", "DefaultTimerAccuracySec", config_parse_sec
, 0, &arg_default_timer_accuracy_usec
},
680 { "Manager", "DefaultStandardOutput", config_parse_output_restricted
,0, &arg_default_std_output
},
681 { "Manager", "DefaultStandardError", config_parse_output_restricted
,0, &arg_default_std_error
},
682 { "Manager", "DefaultTimeoutStartSec", config_parse_sec
, 0, &arg_default_timeout_start_usec
},
683 { "Manager", "DefaultTimeoutStopSec", config_parse_sec
, 0, &arg_default_timeout_stop_usec
},
684 { "Manager", "DefaultRestartSec", config_parse_sec
, 0, &arg_default_restart_usec
},
685 { "Manager", "DefaultStartLimitInterval", config_parse_sec
, 0, &arg_default_start_limit_interval
}, /* obsolete alias */
686 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec
, 0, &arg_default_start_limit_interval
},
687 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned
, 0, &arg_default_start_limit_burst
},
688 { "Manager", "DefaultEnvironment", config_parse_environ
, 0, &arg_default_environment
},
689 { "Manager", "DefaultLimitCPU", config_parse_rlimit
, RLIMIT_CPU
, arg_default_rlimit
},
690 { "Manager", "DefaultLimitFSIZE", config_parse_rlimit
, RLIMIT_FSIZE
, arg_default_rlimit
},
691 { "Manager", "DefaultLimitDATA", config_parse_rlimit
, RLIMIT_DATA
, arg_default_rlimit
},
692 { "Manager", "DefaultLimitSTACK", config_parse_rlimit
, RLIMIT_STACK
, arg_default_rlimit
},
693 { "Manager", "DefaultLimitCORE", config_parse_rlimit
, RLIMIT_CORE
, arg_default_rlimit
},
694 { "Manager", "DefaultLimitRSS", config_parse_rlimit
, RLIMIT_RSS
, arg_default_rlimit
},
695 { "Manager", "DefaultLimitNOFILE", config_parse_rlimit
, RLIMIT_NOFILE
, arg_default_rlimit
},
696 { "Manager", "DefaultLimitAS", config_parse_rlimit
, RLIMIT_AS
, arg_default_rlimit
},
697 { "Manager", "DefaultLimitNPROC", config_parse_rlimit
, RLIMIT_NPROC
, arg_default_rlimit
},
698 { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit
, RLIMIT_MEMLOCK
, arg_default_rlimit
},
699 { "Manager", "DefaultLimitLOCKS", config_parse_rlimit
, RLIMIT_LOCKS
, arg_default_rlimit
},
700 { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit
, RLIMIT_SIGPENDING
, arg_default_rlimit
},
701 { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit
, RLIMIT_MSGQUEUE
, arg_default_rlimit
},
702 { "Manager", "DefaultLimitNICE", config_parse_rlimit
, RLIMIT_NICE
, arg_default_rlimit
},
703 { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit
, RLIMIT_RTPRIO
, arg_default_rlimit
},
704 { "Manager", "DefaultLimitRTTIME", config_parse_rlimit
, RLIMIT_RTTIME
, arg_default_rlimit
},
705 { "Manager", "DefaultCPUAccounting", config_parse_bool
, 0, &arg_default_cpu_accounting
},
706 { "Manager", "DefaultIOAccounting", config_parse_bool
, 0, &arg_default_io_accounting
},
707 { "Manager", "DefaultIPAccounting", config_parse_bool
, 0, &arg_default_ip_accounting
},
708 { "Manager", "DefaultBlockIOAccounting", config_parse_bool
, 0, &arg_default_blockio_accounting
},
709 { "Manager", "DefaultMemoryAccounting", config_parse_bool
, 0, &arg_default_memory_accounting
},
710 { "Manager", "DefaultTasksAccounting", config_parse_bool
, 0, &arg_default_tasks_accounting
},
711 { "Manager", "DefaultTasksMax", config_parse_tasks_max
, 0, &arg_default_tasks_max
},
712 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action
, 0, &arg_cad_burst_action
},
716 const char *fn
, *conf_dirs_nulstr
;
719 PKGSYSCONFDIR
"/system.conf" :
720 PKGSYSCONFDIR
"/user.conf";
722 conf_dirs_nulstr
= arg_system
?
723 CONF_PATHS_NULSTR("systemd/system.conf.d") :
724 CONF_PATHS_NULSTR("systemd/user.conf.d");
726 (void) config_parse_many_nulstr(fn
, conf_dirs_nulstr
, "Manager\0", config_item_table_lookup
, items
, CONFIG_PARSE_WARN
, NULL
);
728 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
729 * like everywhere else. */
730 if (arg_default_timeout_start_usec
<= 0)
731 arg_default_timeout_start_usec
= USEC_INFINITY
;
732 if (arg_default_timeout_stop_usec
<= 0)
733 arg_default_timeout_stop_usec
= USEC_INFINITY
;
738 static void set_manager_defaults(Manager
*m
) {
742 /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
743 * affect the manager itself, but are just what newly allocated units will have set if they haven't set
744 * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
746 m
->default_timer_accuracy_usec
= arg_default_timer_accuracy_usec
;
747 m
->default_std_output
= arg_default_std_output
;
748 m
->default_std_error
= arg_default_std_error
;
749 m
->default_timeout_start_usec
= arg_default_timeout_start_usec
;
750 m
->default_timeout_stop_usec
= arg_default_timeout_stop_usec
;
751 m
->default_restart_usec
= arg_default_restart_usec
;
752 m
->default_start_limit_interval
= arg_default_start_limit_interval
;
753 m
->default_start_limit_burst
= arg_default_start_limit_burst
;
754 m
->default_cpu_accounting
= arg_default_cpu_accounting
;
755 m
->default_io_accounting
= arg_default_io_accounting
;
756 m
->default_ip_accounting
= arg_default_ip_accounting
;
757 m
->default_blockio_accounting
= arg_default_blockio_accounting
;
758 m
->default_memory_accounting
= arg_default_memory_accounting
;
759 m
->default_tasks_accounting
= arg_default_tasks_accounting
;
760 m
->default_tasks_max
= arg_default_tasks_max
;
762 manager_set_default_rlimits(m
, arg_default_rlimit
);
763 manager_environment_add(m
, NULL
, arg_default_environment
);
766 static void set_manager_settings(Manager
*m
) {
770 /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
771 * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
773 m
->confirm_spawn
= arg_confirm_spawn
;
774 m
->service_watchdogs
= arg_service_watchdogs
;
775 m
->runtime_watchdog
= arg_runtime_watchdog
;
776 m
->shutdown_watchdog
= arg_shutdown_watchdog
;
777 m
->cad_burst_action
= arg_cad_burst_action
;
779 manager_set_show_status(m
, arg_show_status
);
782 static int parse_argv(int argc
, char *argv
[]) {
784 ARG_LOG_LEVEL
= 0x100,
794 ARG_DUMP_CONFIGURATION_ITEMS
,
795 ARG_DUMP_BUS_PROPERTIES
,
804 ARG_DEFAULT_STD_OUTPUT
,
805 ARG_DEFAULT_STD_ERROR
,
807 ARG_SERVICE_WATCHDOGS
,
810 static const struct option options
[] = {
811 { "log-level", required_argument
, NULL
, ARG_LOG_LEVEL
},
812 { "log-target", required_argument
, NULL
, ARG_LOG_TARGET
},
813 { "log-color", optional_argument
, NULL
, ARG_LOG_COLOR
},
814 { "log-location", optional_argument
, NULL
, ARG_LOG_LOCATION
},
815 { "unit", required_argument
, NULL
, ARG_UNIT
},
816 { "system", no_argument
, NULL
, ARG_SYSTEM
},
817 { "user", no_argument
, NULL
, ARG_USER
},
818 { "test", no_argument
, NULL
, ARG_TEST
},
819 { "no-pager", no_argument
, NULL
, ARG_NO_PAGER
},
820 { "help", no_argument
, NULL
, 'h' },
821 { "version", no_argument
, NULL
, ARG_VERSION
},
822 { "dump-configuration-items", no_argument
, NULL
, ARG_DUMP_CONFIGURATION_ITEMS
},
823 { "dump-bus-properties", no_argument
, NULL
, ARG_DUMP_BUS_PROPERTIES
},
824 { "dump-core", optional_argument
, NULL
, ARG_DUMP_CORE
},
825 { "crash-chvt", required_argument
, NULL
, ARG_CRASH_CHVT
},
826 { "crash-shell", optional_argument
, NULL
, ARG_CRASH_SHELL
},
827 { "crash-reboot", optional_argument
, NULL
, ARG_CRASH_REBOOT
},
828 { "confirm-spawn", optional_argument
, NULL
, ARG_CONFIRM_SPAWN
},
829 { "show-status", optional_argument
, NULL
, ARG_SHOW_STATUS
},
830 { "deserialize", required_argument
, NULL
, ARG_DESERIALIZE
},
831 { "switched-root", no_argument
, NULL
, ARG_SWITCHED_ROOT
},
832 { "default-standard-output", required_argument
, NULL
, ARG_DEFAULT_STD_OUTPUT
, },
833 { "default-standard-error", required_argument
, NULL
, ARG_DEFAULT_STD_ERROR
, },
834 { "machine-id", required_argument
, NULL
, ARG_MACHINE_ID
},
835 { "service-watchdogs", required_argument
, NULL
, ARG_SERVICE_WATCHDOGS
},
844 if (getpid_cached() == 1)
847 while ((c
= getopt_long(argc
, argv
, "hDbsz:", options
, NULL
)) >= 0)
852 r
= log_set_max_level_from_string(optarg
);
854 return log_error_errno(r
, "Failed to parse log level \"%s\": %m", optarg
);
859 r
= log_set_target_from_string(optarg
);
861 return log_error_errno(r
, "Failed to parse log target \"%s\": %m", optarg
);
868 r
= log_show_color_from_string(optarg
);
870 return log_error_errno(r
, "Failed to parse log color setting \"%s\": %m",
873 log_show_color(true);
877 case ARG_LOG_LOCATION
:
879 r
= log_show_location_from_string(optarg
);
881 return log_error_errno(r
, "Failed to parse log location setting \"%s\": %m",
884 log_show_location(true);
888 case ARG_DEFAULT_STD_OUTPUT
:
889 r
= exec_output_from_string(optarg
);
891 return log_error_errno(r
, "Failed to parse default standard output setting \"%s\": %m",
893 arg_default_std_output
= r
;
896 case ARG_DEFAULT_STD_ERROR
:
897 r
= exec_output_from_string(optarg
);
899 return log_error_errno(r
, "Failed to parse default standard error output setting \"%s\": %m",
901 arg_default_std_error
= r
;
905 r
= free_and_strdup(&arg_default_unit
, optarg
);
907 return log_error_errno(r
, "Failed to set default unit \"%s\": %m", optarg
);
920 arg_action
= ACTION_TEST
;
928 arg_action
= ACTION_VERSION
;
931 case ARG_DUMP_CONFIGURATION_ITEMS
:
932 arg_action
= ACTION_DUMP_CONFIGURATION_ITEMS
;
935 case ARG_DUMP_BUS_PROPERTIES
:
936 arg_action
= ACTION_DUMP_BUS_PROPERTIES
;
941 arg_dump_core
= true;
943 r
= parse_boolean(optarg
);
945 return log_error_errno(r
, "Failed to parse dump core boolean: \"%s\": %m",
952 r
= parse_crash_chvt(optarg
);
954 return log_error_errno(r
, "Failed to parse crash virtual terminal index: \"%s\": %m",
958 case ARG_CRASH_SHELL
:
960 arg_crash_shell
= true;
962 r
= parse_boolean(optarg
);
964 return log_error_errno(r
, "Failed to parse crash shell boolean: \"%s\": %m",
970 case ARG_CRASH_REBOOT
:
972 arg_crash_reboot
= true;
974 r
= parse_boolean(optarg
);
976 return log_error_errno(r
, "Failed to parse crash shell boolean: \"%s\": %m",
978 arg_crash_reboot
= r
;
982 case ARG_CONFIRM_SPAWN
:
983 arg_confirm_spawn
= mfree(arg_confirm_spawn
);
985 r
= parse_confirm_spawn(optarg
, &arg_confirm_spawn
);
987 return log_error_errno(r
, "Failed to parse confirm spawn option: \"%s\": %m",
991 case ARG_SERVICE_WATCHDOGS
:
992 r
= parse_boolean(optarg
);
994 return log_error_errno(r
, "Failed to parse service watchdogs boolean: \"%s\": %m",
996 arg_service_watchdogs
= r
;
999 case ARG_SHOW_STATUS
:
1001 r
= parse_show_status(optarg
, &arg_show_status
);
1003 return log_error_errno(r
, "Failed to parse show status boolean: \"%s\": %m",
1006 arg_show_status
= SHOW_STATUS_YES
;
1009 case ARG_DESERIALIZE
: {
1013 r
= safe_atoi(optarg
, &fd
);
1015 log_error_errno(r
, "Failed to parse deserialize option \"%s\": %m", optarg
);
1017 log_error("Invalid deserialize fd: %d", fd
);
1021 (void) fd_cloexec(fd
, true);
1023 f
= fdopen(fd
, "r");
1025 return log_error_errno(errno
, "Failed to open serialization fd %d: %m", fd
);
1027 safe_fclose(arg_serialization
);
1028 arg_serialization
= f
;
1033 case ARG_SWITCHED_ROOT
:
1034 arg_switched_root
= true;
1037 case ARG_MACHINE_ID
:
1038 r
= set_machine_id(optarg
);
1040 return log_error_errno(r
, "MachineID '%s' is not valid: %m", optarg
);
1044 arg_action
= ACTION_HELP
;
1048 log_set_max_level(LOG_DEBUG
);
1054 /* Just to eat away the sysvinit kernel
1055 * cmdline args without getopt() error
1056 * messages that we'll parse in
1057 * parse_proc_cmdline_word() or ignore. */
1060 if (getpid_cached() != 1)
1066 assert_not_reached("Unhandled option code.");
1069 if (optind
< argc
&& getpid_cached() != 1) {
1070 /* Hmm, when we aren't run as init system
1071 * let's complain about excess arguments */
1073 log_error("Excess arguments.");
1080 static int help(void) {
1081 _cleanup_free_
char *link
= NULL
;
1084 r
= terminal_urlify_man("systemd", "1", &link
);
1088 printf("%s [OPTIONS...]\n\n"
1089 "Starts up and maintains the system or user services.\n\n"
1090 " -h --help Show this help\n"
1091 " --version Show version\n"
1092 " --test Determine startup sequence, dump it and exit\n"
1093 " --no-pager Do not pipe output into a pager\n"
1094 " --dump-configuration-items Dump understood unit configuration items\n"
1095 " --dump-bus-properties Dump exposed bus properties\n"
1096 " --unit=UNIT Set default unit\n"
1097 " --system Run a system instance, even if PID != 1\n"
1098 " --user Run a user instance\n"
1099 " --dump-core[=BOOL] Dump core on crash\n"
1100 " --crash-vt=NR Change to specified VT on crash\n"
1101 " --crash-reboot[=BOOL] Reboot on crash\n"
1102 " --crash-shell[=BOOL] Run shell on crash\n"
1103 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1104 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1105 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1106 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1107 " --log-color[=BOOL] Highlight important log messages\n"
1108 " --log-location[=BOOL] Include code location in log messages\n"
1109 " --default-standard-output= Set default standard output for services\n"
1110 " --default-standard-error= Set default standard error output for services\n"
1111 "\nSee the %s for details.\n"
1112 , program_invocation_short_name
1119 static int prepare_reexecute(
1123 bool switching_root
) {
1125 _cleanup_fdset_free_ FDSet
*fds
= NULL
;
1126 _cleanup_fclose_
FILE *f
= NULL
;
1133 r
= manager_open_serialization(m
, &f
);
1135 return log_error_errno(r
, "Failed to create serialization file: %m");
1137 /* Make sure nothing is really destructed when we shut down */
1139 bus_manager_send_reloading(m
, true);
1145 r
= manager_serialize(m
, f
, fds
, switching_root
);
1147 return log_error_errno(r
, "Failed to serialize state: %m");
1149 if (fseeko(f
, 0, SEEK_SET
) == (off_t
) -1)
1150 return log_error_errno(errno
, "Failed to rewind serialization fd: %m");
1152 r
= fd_cloexec(fileno(f
), false);
1154 return log_error_errno(r
, "Failed to disable O_CLOEXEC for serialization: %m");
1156 r
= fdset_cloexec(fds
, false);
1158 return log_error_errno(r
, "Failed to disable O_CLOEXEC for serialization fds: %m");
1160 *ret_f
= TAKE_PTR(f
);
1161 *ret_fds
= TAKE_PTR(fds
);
1166 static void bump_file_max_and_nr_open(void) {
1168 /* Let's bump fs.file-max and fs.nr_open to their respective maximums. On current kernels large numbers of file
1169 * descriptors are no longer a performance problem and their memory is properly tracked by memcg, thus counting
1170 * them and limiting them in another two layers of limits is unnecessary and just complicates things. This
1171 * function hence turns off 2 of the 4 levels of limits on file descriptors, and makes RLIMIT_NOLIMIT (soft +
1172 * hard) the only ones that really matter. */
1174 #if BUMP_PROC_SYS_FS_FILE_MAX || BUMP_PROC_SYS_FS_NR_OPEN
1175 _cleanup_free_
char *t
= NULL
;
1179 #if BUMP_PROC_SYS_FS_FILE_MAX
1180 /* I so wanted to use STRINGIFY(ULONG_MAX) here, but alas we can't as glibc/gcc define that as
1181 * "(0x7fffffffffffffffL * 2UL + 1UL)". Seriously. 😢 */
1182 if (asprintf(&t
, "%lu\n", ULONG_MAX
) < 0) {
1187 r
= sysctl_write("fs/file-max", t
);
1189 log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
, "Failed to bump fs.file-max, ignoring: %m");
1192 #if BUMP_PROC_SYS_FS_FILE_MAX && BUMP_PROC_SYS_FS_NR_OPEN
1196 #if BUMP_PROC_SYS_FS_NR_OPEN
1199 /* Arg! The kernel enforces maximum and minimum values on the fs.nr_open, but we don't really know what they
1200 * are. The expression by which the maximum is determined is dependent on the architecture, and is something we
1201 * don't really want to copy to userspace, as it is dependent on implementation details of the kernel. Since
1202 * the kernel doesn't expose the maximum value to us, we can only try and hope. Hence, let's start with
1203 * INT_MAX, and then keep halving the value until we find one that works. Ugly? Yes, absolutely, but kernel
1204 * APIs are kernel APIs, so what do can we do... 🤯 */
1209 v
&= ~(__SIZEOF_POINTER__
- 1); /* Round down to next multiple of the pointer size */
1211 log_warning("Can't bump fs.nr_open, value too small.");
1217 log_error_errno(k
, "Failed to read fs.nr_open: %m");
1220 if (k
>= v
) { /* Already larger */
1221 log_debug("Skipping bump, value is already larger.");
1225 if (asprintf(&t
, "%i\n", v
) < 0) {
1230 r
= sysctl_write("fs/nr_open", t
);
1233 log_debug("Couldn't write fs.nr_open as %i, halving it.", v
);
1238 log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
, "Failed to bump fs.nr_open, ignoring: %m");
1242 log_debug("Successfully bumped fs.nr_open to %i", v
);
1248 static int bump_rlimit_nofile(struct rlimit
*saved_rlimit
) {
1251 assert(saved_rlimit
);
1253 /* Save the original RLIMIT_NOFILE so that we can reset it later when transitioning from the initrd to the main
1254 * systemd or suchlike. */
1255 if (getrlimit(RLIMIT_NOFILE
, saved_rlimit
) < 0)
1256 return log_warning_errno(errno
, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1258 /* Get the underlying absolute limit the kernel enforces */
1259 nr
= read_nr_open();
1261 /* Make sure forked processes get limits based on the original kernel setting */
1262 if (!arg_default_rlimit
[RLIMIT_NOFILE
]) {
1265 rl
= newdup(struct rlimit
, saved_rlimit
, 1);
1269 /* Bump the hard limit for system services to a substantially higher value. The default hard limit
1270 * current kernels set is pretty low (4K), mostly for historical reasons. According to kernel
1271 * developers, the fd handling in recent kernels has been optimized substantially enough, so that we
1272 * can bump the limit now, without paying too high a price in memory or performance. Note however that
1273 * we only bump the hard limit, not the soft limit. That's because select() works the way it works, and
1274 * chokes on fds >= 1024. If we'd bump the soft limit globally, it might accidentally happen to
1275 * unexpecting programs that they get fds higher than what they can process using select(). By only
1276 * bumping the hard limit but leaving the low limit as it is we avoid this pitfall: programs that are
1277 * written by folks aware of the select() problem in mind (and thus use poll()/epoll instead of
1278 * select(), the way everybody should) can explicitly opt into high fds by bumping their soft limit
1279 * beyond 1024, to the hard limit we pass. */
1281 rl
->rlim_max
= MIN((rlim_t
) nr
, MAX(rl
->rlim_max
, (rlim_t
) HIGH_RLIMIT_NOFILE
));
1283 arg_default_rlimit
[RLIMIT_NOFILE
] = rl
;
1286 /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for
1287 * both hard and soft. */
1288 r
= setrlimit_closest(RLIMIT_NOFILE
, &RLIMIT_MAKE_CONST(nr
));
1290 return log_warning_errno(r
, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1295 static int bump_rlimit_memlock(struct rlimit
*saved_rlimit
) {
1298 assert(saved_rlimit
);
1300 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK which should
1301 * normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's bump
1302 * the value high enough for our user. */
1304 if (getrlimit(RLIMIT_MEMLOCK
, saved_rlimit
) < 0)
1305 return log_warning_errno(errno
, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1307 r
= setrlimit_closest(RLIMIT_MEMLOCK
, &RLIMIT_MAKE_CONST(HIGH_RLIMIT_MEMLOCK
));
1309 return log_warning_errno(r
, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1314 static void test_usr(void) {
1316 /* Check that /usr is not a separate fs */
1318 if (dir_is_empty("/usr") <= 0)
1321 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1322 "Some things will probably break (sometimes even silently) in mysterious ways. "
1323 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1326 static int enforce_syscall_archs(Set
*archs
) {
1330 if (!is_seccomp_available())
1333 r
= seccomp_restrict_archs(arg_syscall_archs
);
1335 return log_error_errno(r
, "Failed to enforce system call architecture restrication: %m");
1340 static int status_welcome(void) {
1341 _cleanup_free_
char *pretty_name
= NULL
, *ansi_color
= NULL
;
1344 if (IN_SET(arg_show_status
, SHOW_STATUS_NO
, SHOW_STATUS_AUTO
))
1347 r
= parse_os_release(NULL
,
1348 "PRETTY_NAME", &pretty_name
,
1349 "ANSI_COLOR", &ansi_color
,
1352 log_full_errno(r
== -ENOENT
? LOG_DEBUG
: LOG_WARNING
, r
,
1353 "Failed to read os-release file, ignoring: %m");
1355 if (log_get_show_color())
1356 return status_printf(NULL
, false, false,
1357 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1358 isempty(ansi_color
) ? "1" : ansi_color
,
1359 isempty(pretty_name
) ? "Linux" : pretty_name
);
1361 return status_printf(NULL
, false, false,
1362 "\nWelcome to %s!\n",
1363 isempty(pretty_name
) ? "Linux" : pretty_name
);
1366 static int write_container_id(void) {
1370 c
= getenv("container");
1374 RUN_WITH_UMASK(0022)
1375 r
= write_string_file("/run/systemd/container", c
, WRITE_STRING_FILE_CREATE
);
1377 return log_warning_errno(r
, "Failed to write /run/systemd/container, ignoring: %m");
1382 static int bump_unix_max_dgram_qlen(void) {
1383 _cleanup_free_
char *qlen
= NULL
;
1387 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1388 * really really early during boot, so that it is actually applied to all our sockets, including the
1389 * $NOTIFY_SOCKET one. */
1391 r
= read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen
);
1393 return log_warning_errno(r
, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1395 r
= safe_atolu(qlen
, &v
);
1397 return log_warning_errno(r
, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen
);
1399 if (v
>= DEFAULT_UNIX_MAX_DGRAM_QLEN
)
1402 r
= write_string_filef("/proc/sys/net/unix/max_dgram_qlen", 0, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN
);
1404 return log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
,
1405 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1410 static int fixup_environment(void) {
1411 _cleanup_free_
char *term
= NULL
;
1415 /* Only fix up the environment when we are started as PID 1 */
1416 if (getpid_cached() != 1)
1419 /* We expect the environment to be set correctly if run inside a container. */
1420 if (detect_container() > 0)
1423 /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1424 * device used by the console. We try to make a better guess here since some consoles might not have support
1425 * for color mode for example.
1427 * However if TERM was configured through the kernel command line then leave it alone. */
1428 r
= proc_cmdline_get_key("TERM", 0, &term
);
1432 t
= term
?: default_term_for_tty("/dev/console");
1434 if (setenv("TERM", t
, 1) < 0)
1440 static void redirect_telinit(int argc
, char *argv
[]) {
1442 /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1444 #if HAVE_SYSV_COMPAT
1445 if (getpid_cached() == 1)
1448 if (!strstr(program_invocation_short_name
, "init"))
1451 execv(SYSTEMCTL_BINARY_PATH
, argv
);
1452 log_error_errno(errno
, "Failed to exec " SYSTEMCTL_BINARY_PATH
": %m");
1457 static int become_shutdown(
1458 const char *shutdown_verb
,
1461 char log_level
[DECIMAL_STR_MAX(int) + 1],
1462 exit_code
[DECIMAL_STR_MAX(uint8_t) + 1],
1463 timeout
[DECIMAL_STR_MAX(usec_t
) + 1];
1465 const char* command_line
[13] = {
1466 SYSTEMD_SHUTDOWN_BINARY_PATH
,
1468 "--timeout", timeout
,
1469 "--log-level", log_level
,
1473 _cleanup_strv_free_
char **env_block
= NULL
;
1477 assert(shutdown_verb
);
1478 assert(!command_line
[pos
]);
1479 env_block
= strv_copy(environ
);
1481 xsprintf(log_level
, "%d", log_get_max_level());
1482 xsprintf(timeout
, "%" PRI_USEC
"us", arg_default_timeout_stop_usec
);
1484 switch (log_get_target()) {
1486 case LOG_TARGET_KMSG
:
1487 case LOG_TARGET_JOURNAL_OR_KMSG
:
1488 case LOG_TARGET_SYSLOG_OR_KMSG
:
1489 command_line
[pos
++] = "kmsg";
1492 case LOG_TARGET_NULL
:
1493 command_line
[pos
++] = "null";
1496 case LOG_TARGET_CONSOLE
:
1498 command_line
[pos
++] = "console";
1502 if (log_get_show_color())
1503 command_line
[pos
++] = "--log-color";
1505 if (log_get_show_location())
1506 command_line
[pos
++] = "--log-location";
1508 if (streq(shutdown_verb
, "exit")) {
1509 command_line
[pos
++] = "--exit-code";
1510 command_line
[pos
++] = exit_code
;
1511 xsprintf(exit_code
, "%d", retval
);
1514 assert(pos
< ELEMENTSOF(command_line
));
1516 if (streq(shutdown_verb
, "reboot") &&
1517 arg_shutdown_watchdog
> 0 &&
1518 arg_shutdown_watchdog
!= USEC_INFINITY
) {
1522 /* If we reboot let's set the shutdown
1523 * watchdog and tell the shutdown binary to
1524 * repeatedly ping it */
1525 r
= watchdog_set_timeout(&arg_shutdown_watchdog
);
1526 watchdog_close(r
< 0);
1528 /* Tell the binary how often to ping, ignore failure */
1529 if (asprintf(&e
, "WATCHDOG_USEC="USEC_FMT
, arg_shutdown_watchdog
) > 0)
1530 (void) strv_consume(&env_block
, e
);
1532 if (arg_watchdog_device
&&
1533 asprintf(&e
, "WATCHDOG_DEVICE=%s", arg_watchdog_device
) > 0)
1534 (void) strv_consume(&env_block
, e
);
1536 watchdog_close(true);
1538 /* Avoid the creation of new processes forked by the
1539 * kernel; at this point, we will not listen to the
1541 if (detect_container() <= 0)
1542 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER
);
1544 execve(SYSTEMD_SHUTDOWN_BINARY_PATH
, (char **) command_line
, env_block
);
1548 static void initialize_clock(void) {
1551 if (clock_is_localtime(NULL
) > 0) {
1555 * The very first call of settimeofday() also does a time warp in the kernel.
1557 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1558 * of maintaining the RTC and do all adjustments. This matches the behavior of Windows, which leaves
1559 * the RTC alone if the registry tells that the RTC runs in UTC.
1561 r
= clock_set_timezone(&min
);
1563 log_error_errno(r
, "Failed to apply local time delta, ignoring: %m");
1565 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min
);
1567 } else if (!in_initrd()) {
1569 * Do a dummy very first call to seal the kernel's time warp magic.
1571 * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1572 * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1573 * until we reach the real system.
1575 * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1576 * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1577 * be treated as UTC that way.
1579 (void) clock_reset_timewarp();
1582 r
= clock_apply_epoch();
1584 log_error_errno(r
, "Current system time is before build time, but cannot correct: %m");
1586 log_info("System time before build time, advancing clock.");
1589 static void initialize_coredump(bool skip_setup
) {
1591 if (getpid_cached() != 1)
1594 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1595 * will process core dumps for system services by default. */
1596 if (setrlimit(RLIMIT_CORE
, &RLIMIT_MAKE_CONST(RLIM_INFINITY
)) < 0)
1597 log_warning_errno(errno
, "Failed to set RLIMIT_CORE: %m");
1599 /* But at the same time, turn off the core_pattern logic by default, so that no
1600 * coredumps are stored until the systemd-coredump tool is enabled via
1601 * sysctl. However it can be changed via the kernel command line later so core
1602 * dumps can still be generated during early startup and in initramfs. */
1604 disable_coredumps();
1608 static void initialize_core_pattern(bool skip_setup
) {
1611 if (skip_setup
|| !arg_early_core_pattern
)
1614 if (getpid_cached() != 1)
1617 r
= write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern
, 0);
1619 log_warning_errno(r
, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", arg_early_core_pattern
);
1622 static void do_reexecute(
1625 const struct rlimit
*saved_rlimit_nofile
,
1626 const struct rlimit
*saved_rlimit_memlock
,
1628 const char *switch_root_dir
,
1629 const char *switch_root_init
,
1630 const char **ret_error_message
) {
1632 unsigned i
, j
, args_size
;
1636 assert(saved_rlimit_nofile
);
1637 assert(saved_rlimit_memlock
);
1638 assert(ret_error_message
);
1640 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1642 watchdog_close(true);
1644 /* Reset the RLIMIT_NOFILE to the kernel default, so that the new systemd can pass the kernel default to its
1645 * child processes */
1647 if (saved_rlimit_nofile
->rlim_cur
> 0)
1648 (void) setrlimit(RLIMIT_NOFILE
, saved_rlimit_nofile
);
1649 if (saved_rlimit_memlock
->rlim_cur
!= (rlim_t
) -1)
1650 (void) setrlimit(RLIMIT_MEMLOCK
, saved_rlimit_memlock
);
1652 if (switch_root_dir
) {
1653 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1654 * SIGCHLD for them after deserializing. */
1655 broadcast_signal(SIGTERM
, false, true, arg_default_timeout_stop_usec
);
1657 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1658 r
= switch_root(switch_root_dir
, "/mnt", true, MS_MOVE
);
1660 log_error_errno(r
, "Failed to switch root, trying to continue: %m");
1663 args_size
= MAX(6, argc
+1);
1664 args
= newa(const char*, args_size
);
1666 if (!switch_root_init
) {
1667 char sfd
[DECIMAL_STR_MAX(int) + 1];
1669 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1670 * the user didn't specify an explicit init to spawn. */
1672 assert(arg_serialization
);
1675 xsprintf(sfd
, "%i", fileno(arg_serialization
));
1678 args
[i
++] = SYSTEMD_BINARY_PATH
;
1679 if (switch_root_dir
)
1680 args
[i
++] = "--switched-root";
1681 args
[i
++] = arg_system
? "--system" : "--user";
1682 args
[i
++] = "--deserialize";
1686 assert(i
<= args_size
);
1689 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do this is on
1690 * its own on exec(), but it will do it on exit(). Hence, to ensure we get a summary here, fork() off
1691 * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1692 * before proceeding into the exec().
1694 valgrind_summary_hack();
1696 (void) execv(args
[0], (char* const*) args
);
1697 log_debug_errno(errno
, "Failed to execute our own binary, trying fallback: %m");
1700 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1701 * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1702 * doesn't matter.) */
1704 arg_serialization
= safe_fclose(arg_serialization
);
1705 fds
= fdset_free(fds
);
1707 /* Reopen the console */
1708 (void) make_console_stdio();
1710 for (j
= 1, i
= 1; j
< (unsigned) argc
; j
++)
1711 args
[i
++] = argv
[j
];
1713 assert(i
<= args_size
);
1715 /* Reenable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1716 (void) reset_all_signal_handlers();
1717 (void) reset_signal_mask();
1719 if (switch_root_init
) {
1720 args
[0] = switch_root_init
;
1721 (void) execv(args
[0], (char* const*) args
);
1722 log_warning_errno(errno
, "Failed to execute configured init, trying fallback: %m");
1725 args
[0] = "/sbin/init";
1726 (void) execv(args
[0], (char* const*) args
);
1729 manager_status_printf(NULL
, STATUS_TYPE_EMERGENCY
,
1730 ANSI_HIGHLIGHT_RED
" !! " ANSI_NORMAL
,
1731 "Failed to execute /sbin/init");
1734 log_warning("No /sbin/init, trying fallback");
1736 args
[0] = "/bin/sh";
1738 (void) execv(args
[0], (char* const*) args
);
1739 log_error_errno(errno
, "Failed to execute /bin/sh, giving up: %m");
1741 log_warning_errno(r
, "Failed to execute /sbin/init, giving up: %m");
1743 *ret_error_message
= "Failed to execute fallback shell";
1746 static int invoke_main_loop(
1748 bool *ret_reexecute
,
1749 int *ret_retval
, /* Return parameters relevant for shutting down */
1750 const char **ret_shutdown_verb
, /* … */
1751 FDSet
**ret_fds
, /* Return parameters for reexecuting */
1752 char **ret_switch_root_dir
, /* … */
1753 char **ret_switch_root_init
, /* … */
1754 const char **ret_error_message
) {
1759 assert(ret_reexecute
);
1761 assert(ret_shutdown_verb
);
1763 assert(ret_switch_root_dir
);
1764 assert(ret_switch_root_init
);
1765 assert(ret_error_message
);
1768 r
= manager_loop(m
);
1770 *ret_error_message
= "Failed to run main loop";
1771 return log_emergency_errno(r
, "Failed to run main loop: %m");
1774 switch ((ManagerObjective
) r
) {
1776 case MANAGER_RELOAD
: {
1777 LogTarget saved_log_target
;
1778 int saved_log_level
;
1780 log_info("Reloading.");
1782 /* First, save any overridden log level/target, then parse the configuration file, which might
1783 * change the log level to new settings. */
1785 saved_log_level
= m
->log_level_overridden
? log_get_max_level() : -1;
1786 saved_log_target
= m
->log_target_overridden
? log_get_target() : _LOG_TARGET_INVALID
;
1788 r
= parse_config_file();
1790 log_warning_errno(r
, "Failed to parse config file, ignoring: %m");
1792 set_manager_defaults(m
);
1794 if (saved_log_level
>= 0)
1795 manager_override_log_level(m
, saved_log_level
);
1796 if (saved_log_target
>= 0)
1797 manager_override_log_target(m
, saved_log_target
);
1799 r
= manager_reload(m
);
1801 /* Reloading failed before the point of no return. Let's continue running as if nothing happened. */
1802 m
->objective
= MANAGER_OK
;
1807 case MANAGER_REEXECUTE
:
1809 r
= prepare_reexecute(m
, &arg_serialization
, ret_fds
, false);
1811 *ret_error_message
= "Failed to prepare for reexecution";
1815 log_notice("Reexecuting.");
1817 *ret_reexecute
= true;
1818 *ret_retval
= EXIT_SUCCESS
;
1819 *ret_shutdown_verb
= NULL
;
1820 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
1824 case MANAGER_SWITCH_ROOT
:
1825 if (!m
->switch_root_init
) {
1826 r
= prepare_reexecute(m
, &arg_serialization
, ret_fds
, true);
1828 *ret_error_message
= "Failed to prepare for reexecution";
1834 log_notice("Switching root.");
1836 *ret_reexecute
= true;
1837 *ret_retval
= EXIT_SUCCESS
;
1838 *ret_shutdown_verb
= NULL
;
1840 /* Steal the switch root parameters */
1841 *ret_switch_root_dir
= m
->switch_root
;
1842 *ret_switch_root_init
= m
->switch_root_init
;
1843 m
->switch_root
= m
->switch_root_init
= NULL
;
1849 if (MANAGER_IS_USER(m
)) {
1852 *ret_reexecute
= false;
1853 *ret_retval
= m
->return_value
;
1854 *ret_shutdown_verb
= NULL
;
1856 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
1862 case MANAGER_REBOOT
:
1863 case MANAGER_POWEROFF
:
1865 case MANAGER_KEXEC
: {
1866 static const char * const table
[_MANAGER_OBJECTIVE_MAX
] = {
1867 [MANAGER_EXIT
] = "exit",
1868 [MANAGER_REBOOT
] = "reboot",
1869 [MANAGER_POWEROFF
] = "poweroff",
1870 [MANAGER_HALT
] = "halt",
1871 [MANAGER_KEXEC
] = "kexec",
1874 log_notice("Shutting down.");
1876 *ret_reexecute
= false;
1877 *ret_retval
= m
->return_value
;
1878 assert_se(*ret_shutdown_verb
= table
[m
->objective
]);
1880 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
1886 assert_not_reached("Unknown or unexpected manager objective.");
1891 static void log_execution_mode(bool *ret_first_boot
) {
1892 assert(ret_first_boot
);
1897 log_info(PACKAGE_STRING
" running in %ssystem mode. (" SYSTEMD_FEATURES
")",
1898 arg_action
== ACTION_TEST
? "test " : "" );
1900 v
= detect_virtualization();
1902 log_info("Detected virtualization %s.", virtualization_to_string(v
));
1904 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1907 *ret_first_boot
= false;
1908 log_info("Running in initial RAM disk.");
1910 /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
1911 * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
1912 * doesn't it's unpopulated. This allows container managers and installers to provision a
1913 * couple of files already. If the container manager wants to provision the machine ID itself
1914 * it should pass $container_uuid to PID 1. */
1916 *ret_first_boot
= access("/etc/machine-id", F_OK
) < 0;
1917 if (*ret_first_boot
)
1918 log_info("Running with unpopulated /etc.");
1921 if (DEBUG_LOGGING
) {
1922 _cleanup_free_
char *t
;
1924 t
= uid_to_name(getuid());
1925 log_debug(PACKAGE_STRING
" running in %suser mode for user " UID_FMT
"/%s. (" SYSTEMD_FEATURES
")",
1926 arg_action
== ACTION_TEST
? " test" : "", getuid(), strna(t
));
1929 *ret_first_boot
= false;
1933 static int initialize_runtime(
1935 struct rlimit
*saved_rlimit_nofile
,
1936 struct rlimit
*saved_rlimit_memlock
,
1937 const char **ret_error_message
) {
1941 assert(ret_error_message
);
1943 /* Sets up various runtime parameters. Many of these initializations are conditionalized:
1945 * - Some only apply to --system instances
1946 * - Some only apply to --user instances
1947 * - Some only apply when we first start up, but not when we reexecute
1950 if (arg_action
!= ACTION_RUN
)
1954 /* Make sure we leave a core dump without panicing the kernel. */
1955 install_crash_handler();
1958 r
= mount_cgroup_controllers(arg_join_controllers
);
1960 *ret_error_message
= "Failed to mount cgroup hierarchies";
1966 machine_id_setup(NULL
, arg_machine_id
, NULL
);
1968 bump_unix_max_dgram_qlen();
1969 bump_file_max_and_nr_open();
1971 write_container_id();
1974 if (arg_watchdog_device
) {
1975 r
= watchdog_set_device(arg_watchdog_device
);
1977 log_warning_errno(r
, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device
);
1980 if (arg_runtime_watchdog
> 0 && arg_runtime_watchdog
!= USEC_INFINITY
)
1981 watchdog_set_timeout(&arg_runtime_watchdog
);
1984 if (arg_timer_slack_nsec
!= NSEC_INFINITY
)
1985 if (prctl(PR_SET_TIMERSLACK
, arg_timer_slack_nsec
) < 0)
1986 log_warning_errno(errno
, "Failed to adjust timer slack, ignoring: %m");
1988 if (arg_system
&& !cap_test_all(arg_capability_bounding_set
)) {
1989 r
= capability_bounding_set_drop_usermode(arg_capability_bounding_set
);
1991 *ret_error_message
= "Failed to drop capability bounding set of usermode helpers";
1992 return log_emergency_errno(r
, "Failed to drop capability bounding set of usermode helpers: %m");
1995 r
= capability_bounding_set_drop(arg_capability_bounding_set
, true);
1997 *ret_error_message
= "Failed to drop capability bounding set";
1998 return log_emergency_errno(r
, "Failed to drop capability bounding set: %m");
2002 if (arg_system
&& arg_no_new_privs
) {
2003 if (prctl(PR_SET_NO_NEW_PRIVS
, 1, 0, 0, 0) < 0) {
2004 *ret_error_message
= "Failed to disable new privileges";
2005 return log_emergency_errno(errno
, "Failed to disable new privileges: %m");
2009 if (arg_syscall_archs
) {
2010 r
= enforce_syscall_archs(arg_syscall_archs
);
2012 *ret_error_message
= "Failed to set syscall architectures";
2018 /* Become reaper of our children */
2019 if (prctl(PR_SET_CHILD_SUBREAPER
, 1) < 0)
2020 log_warning_errno(errno
, "Failed to make us a subreaper: %m");
2022 /* Bump up RLIMIT_NOFILE for systemd itself */
2023 (void) bump_rlimit_nofile(saved_rlimit_nofile
);
2024 (void) bump_rlimit_memlock(saved_rlimit_memlock
);
2029 static int do_queue_default_job(
2031 const char **ret_error_message
) {
2033 _cleanup_(sd_bus_error_free
) sd_bus_error error
= SD_BUS_ERROR_NULL
;
2034 Job
*default_unit_job
;
2035 Unit
*target
= NULL
;
2038 log_debug("Activating default unit: %s", arg_default_unit
);
2040 r
= manager_load_startable_unit_or_warn(m
, arg_default_unit
, NULL
, &target
);
2042 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET
);
2044 r
= manager_load_startable_unit_or_warn(m
, SPECIAL_RESCUE_TARGET
, NULL
, &target
);
2046 *ret_error_message
= r
== -ERFKILL
? "Rescue target masked"
2047 : "Failed to load rescue target";
2052 assert(target
->load_state
== UNIT_LOADED
);
2054 r
= manager_add_job(m
, JOB_START
, target
, JOB_ISOLATE
, &error
, &default_unit_job
);
2056 log_debug_errno(r
, "Default target could not be isolated, starting instead: %s", bus_error_message(&error
, r
));
2058 sd_bus_error_free(&error
);
2060 r
= manager_add_job(m
, JOB_START
, target
, JOB_REPLACE
, &error
, &default_unit_job
);
2062 *ret_error_message
= "Failed to start default target";
2063 return log_emergency_errno(r
, "Failed to start default target: %s", bus_error_message(&error
, r
));
2067 *ret_error_message
= "Failed to isolate default target";
2068 return log_emergency_errno(r
, "Failed to isolate default target: %s", bus_error_message(&error
, r
));
2071 m
->default_unit_job_id
= default_unit_job
->id
;
2076 static void free_arguments(void) {
2078 /* Frees all arg_* variables, with the exception of arg_serialization */
2079 rlimit_free_all(arg_default_rlimit
);
2081 arg_default_unit
= mfree(arg_default_unit
);
2082 arg_confirm_spawn
= mfree(arg_confirm_spawn
);
2083 arg_join_controllers
= strv_free_free(arg_join_controllers
);
2084 arg_default_environment
= strv_free(arg_default_environment
);
2085 arg_syscall_archs
= set_free(arg_syscall_archs
);
2088 static int load_configuration(int argc
, char **argv
, const char **ret_error_message
) {
2091 assert(ret_error_message
);
2093 arg_default_tasks_max
= system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE
, 100U);
2095 r
= parse_config_file();
2097 *ret_error_message
= "Failed to parse config file";
2102 r
= proc_cmdline_parse(parse_proc_cmdline_item
, NULL
, 0);
2104 log_warning_errno(r
, "Failed to parse kernel command line, ignoring: %m");
2107 /* Note that this also parses bits from the kernel command line, including "debug". */
2108 log_parse_environment();
2110 r
= parse_argv(argc
, argv
);
2112 *ret_error_message
= "Failed to parse commandline arguments";
2116 /* Initialize default unit */
2117 if (!arg_default_unit
) {
2118 arg_default_unit
= strdup(SPECIAL_DEFAULT_TARGET
);
2119 if (!arg_default_unit
) {
2120 *ret_error_message
= "Failed to set default unit";
2125 /* Initialize the show status setting if it hasn't been set explicitly yet */
2126 if (arg_show_status
== _SHOW_STATUS_INVALID
)
2127 arg_show_status
= SHOW_STATUS_YES
;
2132 static int safety_checks(void) {
2134 if (getpid_cached() == 1 &&
2135 arg_action
!= ACTION_RUN
) {
2136 log_error("Unsupported execution mode while PID 1.");
2140 if (getpid_cached() == 1 &&
2142 log_error("Can't run --user mode as PID 1.");
2146 if (arg_action
== ACTION_RUN
&&
2148 getpid_cached() != 1) {
2149 log_error("Can't run system mode unless PID 1.");
2153 if (arg_action
== ACTION_TEST
&&
2155 log_error("Don't run test mode as root.");
2160 arg_action
== ACTION_RUN
&&
2162 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
2167 arg_action
== ACTION_RUN
&&
2168 !getenv("XDG_RUNTIME_DIR")) {
2169 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
2174 arg_action
== ACTION_RUN
&&
2175 running_in_chroot() > 0) {
2176 log_error("Cannot be run in a chroot() environment.");
2183 static int initialize_security(
2184 bool *loaded_policy
,
2185 dual_timestamp
*security_start_timestamp
,
2186 dual_timestamp
*security_finish_timestamp
,
2187 const char **ret_error_message
) {
2191 assert(loaded_policy
);
2192 assert(security_start_timestamp
);
2193 assert(security_finish_timestamp
);
2194 assert(ret_error_message
);
2196 dual_timestamp_get(security_start_timestamp
);
2198 r
= mac_selinux_setup(loaded_policy
);
2200 *ret_error_message
= "Failed to load SELinux policy";
2204 r
= mac_smack_setup(loaded_policy
);
2206 *ret_error_message
= "Failed to load SMACK policy";
2212 *ret_error_message
= "Failed to load IMA policy";
2216 dual_timestamp_get(security_finish_timestamp
);
2220 static void test_summary(Manager
*m
) {
2223 printf("-> By units:\n");
2224 manager_dump_units(m
, stdout
, "\t");
2226 printf("-> By jobs:\n");
2227 manager_dump_jobs(m
, stdout
, "\t");
2230 static int collect_fds(FDSet
**ret_fds
, const char **ret_error_message
) {
2234 assert(ret_error_message
);
2236 r
= fdset_new_fill(ret_fds
);
2238 *ret_error_message
= "Failed to allocate fd set";
2239 return log_emergency_errno(r
, "Failed to allocate fd set: %m");
2242 fdset_cloexec(*ret_fds
, true);
2244 if (arg_serialization
)
2245 assert_se(fdset_remove(*ret_fds
, fileno(arg_serialization
)) >= 0);
2250 static void setup_console_terminal(bool skip_setup
) {
2255 /* Become a session leader if we aren't one yet. */
2258 /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2260 (void) release_terminal();
2262 /* Reset the console, but only if this is really init and we are freshly booted */
2263 if (getpid_cached() == 1 && !skip_setup
)
2264 (void) console_setup();
2267 static bool early_skip_setup_check(int argc
, char *argv
[]) {
2268 bool found_deserialize
= false;
2271 /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2272 * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2273 * anyway, even if in that case we also do deserialization. */
2275 for (i
= 1; i
< argc
; i
++) {
2276 if (streq(argv
[i
], "--switched-root"))
2277 return false; /* If we switched root, don't skip the setup. */
2278 else if (streq(argv
[i
], "--deserialize"))
2279 found_deserialize
= true;
2282 return found_deserialize
; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2285 int main(int argc
, char *argv
[]) {
2287 dual_timestamp initrd_timestamp
= DUAL_TIMESTAMP_NULL
, userspace_timestamp
= DUAL_TIMESTAMP_NULL
, kernel_timestamp
= DUAL_TIMESTAMP_NULL
,
2288 security_start_timestamp
= DUAL_TIMESTAMP_NULL
, security_finish_timestamp
= DUAL_TIMESTAMP_NULL
;
2289 struct rlimit saved_rlimit_nofile
= RLIMIT_MAKE_CONST(0), saved_rlimit_memlock
= RLIMIT_MAKE_CONST((rlim_t
) -1);
2290 bool skip_setup
, loaded_policy
= false, queue_default_job
= false, first_boot
= false, reexecute
= false;
2291 char *switch_root_dir
= NULL
, *switch_root_init
= NULL
;
2292 usec_t before_startup
, after_startup
;
2293 static char systemd
[] = "systemd";
2294 char timespan
[FORMAT_TIMESPAN_MAX
];
2295 const char *shutdown_verb
= NULL
, *error_message
= NULL
;
2296 int r
, retval
= EXIT_FAILURE
;
2300 /* SysV compatibility: redirect init → telinit */
2301 redirect_telinit(argc
, argv
);
2303 /* Take timestamps early on */
2304 dual_timestamp_from_monotonic(&kernel_timestamp
, 0);
2305 dual_timestamp_get(&userspace_timestamp
);
2307 /* Figure out whether we need to do initialize the system, or if we already did that because we are
2309 skip_setup
= early_skip_setup_check(argc
, argv
);
2311 /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2312 * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
2313 program_invocation_short_name
= systemd
;
2314 (void) prctl(PR_SET_NAME
, systemd
);
2316 /* Save the original command line */
2320 /* Make sure that if the user says "syslog" we actually log to the journal. */
2321 log_set_upgrade_syslog_to_journal(true);
2323 if (getpid_cached() == 1) {
2324 /* When we run as PID 1 force system mode */
2327 /* Disable the umask logic */
2330 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2331 * activated yet (even though the log socket for it exists). */
2332 log_set_prohibit_ipc(true);
2334 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2335 * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2336 * child process right before execve()'ing the actual binary, at a point in time where socket
2337 * activation stderr/stdout area already set up. */
2338 log_set_always_reopen_console(true);
2340 if (detect_container() <= 0) {
2342 /* Running outside of a container as PID 1 */
2343 log_set_target(LOG_TARGET_KMSG
);
2347 initrd_timestamp
= userspace_timestamp
;
2350 r
= mount_setup_early();
2352 error_message
= "Failed to mount early API filesystems";
2356 r
= initialize_security(
2358 &security_start_timestamp
,
2359 &security_finish_timestamp
,
2365 if (mac_selinux_init() < 0) {
2366 error_message
= "Failed to initialize SELinux policy";
2373 /* Set the default for later on, but don't actually open the logs like this for now. Note that
2374 * if we are transitioning from the initrd there might still be journal fd open, and we
2375 * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2377 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG
);
2380 /* Running inside a container, as PID 1 */
2381 log_set_target(LOG_TARGET_CONSOLE
);
2384 /* For later on, see above... */
2385 log_set_target(LOG_TARGET_JOURNAL
);
2387 /* clear the kernel timestamp,
2388 * because we are in a container */
2389 kernel_timestamp
= DUAL_TIMESTAMP_NULL
;
2392 initialize_coredump(skip_setup
);
2394 r
= fixup_environment();
2396 log_emergency_errno(r
, "Failed to fix up PID 1 environment: %m");
2397 error_message
= "Failed to fix up PID1 environment";
2402 /* Running as user instance */
2404 log_set_target(LOG_TARGET_AUTO
);
2407 /* clear the kernel timestamp,
2408 * because we are not PID 1 */
2409 kernel_timestamp
= DUAL_TIMESTAMP_NULL
;
2413 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2414 * they never log into the console. */
2415 log_show_color(colors_enabled());
2417 r
= make_null_stdio();
2419 log_warning_errno(r
, "Failed to redirect standard streams to /dev/null, ignoring: %m");
2422 /* Mount /proc, /sys and friends, so that /proc/cmdline and
2423 * /proc/$PID/fd is available. */
2424 if (getpid_cached() == 1) {
2426 /* Load the kernel modules early. */
2430 r
= mount_setup(loaded_policy
);
2432 error_message
= "Failed to mount API filesystems";
2437 /* Reset all signal handlers. */
2438 (void) reset_all_signal_handlers();
2439 (void) ignore_signals(SIGNALS_IGNORE
, -1);
2441 r
= load_configuration(argc
, argv
, &error_message
);
2445 r
= safety_checks();
2449 if (IN_SET(arg_action
, ACTION_TEST
, ACTION_HELP
, ACTION_DUMP_CONFIGURATION_ITEMS
, ACTION_DUMP_BUS_PROPERTIES
))
2450 (void) pager_open(arg_no_pager
, false);
2452 if (arg_action
!= ACTION_RUN
)
2455 if (arg_action
== ACTION_HELP
) {
2456 retval
= help() < 0 ? EXIT_FAILURE
: EXIT_SUCCESS
;
2458 } else if (arg_action
== ACTION_VERSION
) {
2461 } else if (arg_action
== ACTION_DUMP_CONFIGURATION_ITEMS
) {
2462 unit_dump_config_items(stdout
);
2463 retval
= EXIT_SUCCESS
;
2465 } else if (arg_action
== ACTION_DUMP_BUS_PROPERTIES
) {
2466 dump_bus_properties(stdout
);
2467 retval
= EXIT_SUCCESS
;
2471 assert_se(IN_SET(arg_action
, ACTION_RUN
, ACTION_TEST
));
2473 /* Move out of the way, so that we won't block unmounts */
2474 assert_se(chdir("/") == 0);
2476 if (arg_action
== ACTION_RUN
) {
2478 /* A core pattern might have been specified via the cmdline. */
2479 initialize_core_pattern(skip_setup
);
2481 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2484 /* Remember open file descriptors for later deserialization */
2485 r
= collect_fds(&fds
, &error_message
);
2489 /* Give up any control of the console, but make sure its initialized. */
2490 setup_console_terminal(skip_setup
);
2492 /* Open the logging devices, if possible and necessary */
2496 log_execution_mode(&first_boot
);
2498 r
= initialize_runtime(skip_setup
,
2499 &saved_rlimit_nofile
,
2500 &saved_rlimit_memlock
,
2505 r
= manager_new(arg_system
? UNIT_FILE_SYSTEM
: UNIT_FILE_USER
,
2506 arg_action
== ACTION_TEST
? MANAGER_TEST_FULL
: 0,
2509 log_emergency_errno(r
, "Failed to allocate manager object: %m");
2510 error_message
= "Failed to allocate manager object";
2514 m
->timestamps
[MANAGER_TIMESTAMP_KERNEL
] = kernel_timestamp
;
2515 m
->timestamps
[MANAGER_TIMESTAMP_INITRD
] = initrd_timestamp
;
2516 m
->timestamps
[MANAGER_TIMESTAMP_USERSPACE
] = userspace_timestamp
;
2517 m
->timestamps
[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START
)] = security_start_timestamp
;
2518 m
->timestamps
[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH
)] = security_finish_timestamp
;
2520 set_manager_defaults(m
);
2521 set_manager_settings(m
);
2522 manager_set_first_boot(m
, first_boot
);
2524 /* Remember whether we should queue the default job */
2525 queue_default_job
= !arg_serialization
|| arg_switched_root
;
2527 before_startup
= now(CLOCK_MONOTONIC
);
2529 r
= manager_startup(m
, arg_serialization
, fds
);
2531 error_message
= "Failed to start up manager";
2535 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2536 fds
= fdset_free(fds
);
2537 arg_serialization
= safe_fclose(arg_serialization
);
2539 if (queue_default_job
) {
2540 r
= do_queue_default_job(m
, &error_message
);
2545 after_startup
= now(CLOCK_MONOTONIC
);
2547 log_full(arg_action
== ACTION_TEST
? LOG_INFO
: LOG_DEBUG
,
2548 "Loaded units and determined initial transaction in %s.",
2549 format_timespan(timespan
, sizeof(timespan
), after_startup
- before_startup
, 100 * USEC_PER_MSEC
));
2551 if (arg_action
== ACTION_TEST
) {
2553 retval
= EXIT_SUCCESS
;
2557 (void) invoke_main_loop(m
,
2570 arg_shutdown_watchdog
= m
->shutdown_watchdog
;
2571 m
= manager_free(m
);
2575 mac_selinux_finish();
2578 do_reexecute(argc
, argv
,
2579 &saved_rlimit_nofile
,
2580 &saved_rlimit_memlock
,
2584 &error_message
); /* This only returns if reexecution failed */
2586 arg_serialization
= safe_fclose(arg_serialization
);
2587 fds
= fdset_free(fds
);
2589 #if HAVE_VALGRIND_VALGRIND_H
2590 /* If we are PID 1 and running under valgrind, then let's exit
2591 * here explicitly. valgrind will only generate nice output on
2592 * exit(), not on exec(), hence let's do the former not the
2594 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND
) {
2595 /* Cleanup watchdog_device strings for valgrind. We need them
2596 * in become_shutdown() so normally we cannot free them yet. */
2597 watchdog_free_device();
2598 arg_watchdog_device
= mfree(arg_watchdog_device
);
2603 if (shutdown_verb
) {
2604 r
= become_shutdown(shutdown_verb
, retval
);
2605 log_error_errno(r
, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
2606 error_message
= "Failed to execute shutdown binary";
2609 watchdog_free_device();
2610 arg_watchdog_device
= mfree(arg_watchdog_device
);
2612 if (getpid_cached() == 1) {
2614 manager_status_printf(NULL
, STATUS_TYPE_EMERGENCY
,
2615 ANSI_HIGHLIGHT_RED
"!!!!!!" ANSI_NORMAL
,
2616 "%s, freezing.", error_message
);