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 "cgroup-util.h"
32 #include "clock-util.h"
33 #include "conf-parser.h"
34 #include "cpu-set-util.h"
36 #include "dbus-manager.h"
38 #include "emergency-action.h"
40 #include "exit-status.h"
44 #include "format-util.h"
46 #include "hostname-setup.h"
47 #include "ima-setup.h"
49 #include "kmod-setup.h"
50 #include "load-fragment.h"
52 #include "loopback-setup.h"
53 #include "machine-id-setup.h"
56 #include "mount-setup.h"
59 #include "parse-util.h"
60 #include "path-util.h"
61 #include "pretty-print.h"
62 #include "proc-cmdline.h"
63 #include "process-util.h"
64 #include "raw-clone.h"
65 #include "rlimit-util.h"
67 #include "seccomp-util.h"
69 #include "selinux-setup.h"
70 #include "selinux-util.h"
71 #include "signal-util.h"
72 #include "smack-setup.h"
74 #include "stat-util.h"
75 #include "stdio-util.h"
77 #include "switch-root.h"
78 #include "sysctl-util.h"
79 #include "terminal-util.h"
80 #include "umask-util.h"
81 #include "user-util.h"
91 ACTION_DUMP_CONFIGURATION_ITEMS
,
92 ACTION_DUMP_BUS_PROPERTIES
,
93 } arg_action
= ACTION_RUN
;
94 static char *arg_default_unit
= NULL
;
95 static bool arg_system
= false;
96 static bool arg_dump_core
= true;
97 static int arg_crash_chvt
= -1;
98 static bool arg_crash_shell
= false;
99 static bool arg_crash_reboot
= false;
100 static char *arg_confirm_spawn
= NULL
;
101 static ShowStatus arg_show_status
= _SHOW_STATUS_INVALID
;
102 static bool arg_switched_root
= false;
103 static PagerFlags arg_pager_flags
= 0;
104 static bool arg_service_watchdogs
= true;
105 static ExecOutput arg_default_std_output
= EXEC_OUTPUT_JOURNAL
;
106 static ExecOutput arg_default_std_error
= EXEC_OUTPUT_INHERIT
;
107 static usec_t arg_default_restart_usec
= DEFAULT_RESTART_USEC
;
108 static usec_t arg_default_timeout_start_usec
= DEFAULT_TIMEOUT_USEC
;
109 static usec_t arg_default_timeout_stop_usec
= DEFAULT_TIMEOUT_USEC
;
110 static usec_t arg_default_start_limit_interval
= DEFAULT_START_LIMIT_INTERVAL
;
111 static unsigned arg_default_start_limit_burst
= DEFAULT_START_LIMIT_BURST
;
112 static usec_t arg_runtime_watchdog
= 0;
113 static usec_t arg_shutdown_watchdog
= 10 * USEC_PER_MINUTE
;
114 static char *arg_early_core_pattern
= NULL
;
115 static char *arg_watchdog_device
= NULL
;
116 static char **arg_default_environment
= NULL
;
117 static struct rlimit
*arg_default_rlimit
[_RLIMIT_MAX
] = {};
118 static uint64_t arg_capability_bounding_set
= CAP_ALL
;
119 static bool arg_no_new_privs
= false;
120 static nsec_t arg_timer_slack_nsec
= NSEC_INFINITY
;
121 static usec_t arg_default_timer_accuracy_usec
= 1 * USEC_PER_MINUTE
;
122 static Set
* arg_syscall_archs
= NULL
;
123 static FILE* arg_serialization
= NULL
;
124 static int arg_default_cpu_accounting
= -1;
125 static bool arg_default_io_accounting
= false;
126 static bool arg_default_ip_accounting
= false;
127 static bool arg_default_blockio_accounting
= false;
128 static bool arg_default_memory_accounting
= MEMORY_ACCOUNTING_DEFAULT
;
129 static bool arg_default_tasks_accounting
= true;
130 static uint64_t arg_default_tasks_max
= UINT64_MAX
;
131 static sd_id128_t arg_machine_id
= {};
132 static EmergencyAction arg_cad_burst_action
= EMERGENCY_ACTION_REBOOT_FORCE
;
134 _noreturn_
static void freeze_or_exit_or_reboot(void) {
136 /* If we are running in a contianer, let's prefer exiting, after all we can propagate an exit code to the
137 * container manager, and thus inform it that something went wrong. */
138 if (detect_container() > 0) {
139 log_emergency("Exiting PID 1...");
140 exit(EXIT_EXCEPTION
);
143 if (arg_crash_reboot
) {
144 log_notice("Rebooting in 10s...");
147 log_notice("Rebooting now...");
148 (void) reboot(RB_AUTOBOOT
);
149 log_emergency_errno(errno
, "Failed to reboot: %m");
152 log_emergency("Freezing execution.");
156 _noreturn_
static void crash(int sig
) {
160 if (getpid_cached() != 1)
161 /* Pass this on immediately, if this is not PID 1 */
163 else if (!arg_dump_core
)
164 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig
));
166 sa
= (struct sigaction
) {
167 .sa_handler
= nop_signal_handler
,
168 .sa_flags
= SA_NOCLDSTOP
|SA_RESTART
,
171 /* We want to wait for the core process, hence let's enable SIGCHLD */
172 (void) sigaction(SIGCHLD
, &sa
, NULL
);
174 pid
= raw_clone(SIGCHLD
);
176 log_emergency_errno(errno
, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig
));
178 /* Enable default signal handler for core dump */
180 sa
= (struct sigaction
) {
181 .sa_handler
= SIG_DFL
,
183 (void) sigaction(sig
, &sa
, NULL
);
185 /* Don't limit the coredump size */
186 (void) setrlimit(RLIMIT_CORE
, &RLIMIT_MAKE_CONST(RLIM_INFINITY
));
188 /* Just to be sure... */
191 /* Raise the signal again */
193 (void) kill(pid
, sig
); /* raise() would kill the parent */
195 assert_not_reached("We shouldn't be here...");
196 _exit(EXIT_EXCEPTION
);
201 /* Order things nicely. */
202 r
= wait_for_terminate(pid
, &status
);
204 log_emergency_errno(r
, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig
));
205 else if (status
.si_code
!= CLD_DUMPED
)
206 log_emergency("Caught <%s>, core dump failed (child "PID_FMT
", code=%s, status=%i/%s).",
207 signal_to_string(sig
),
208 pid
, sigchld_code_to_string(status
.si_code
),
210 strna(status
.si_code
== CLD_EXITED
211 ? exit_status_to_string(status
.si_status
, EXIT_STATUS_MINIMAL
)
212 : signal_to_string(status
.si_status
)));
214 log_emergency("Caught <%s>, dumped core as pid "PID_FMT
".", signal_to_string(sig
), pid
);
218 if (arg_crash_chvt
>= 0)
219 (void) chvt(arg_crash_chvt
);
221 sa
= (struct sigaction
) {
222 .sa_handler
= SIG_IGN
,
223 .sa_flags
= SA_NOCLDSTOP
|SA_NOCLDWAIT
|SA_RESTART
,
226 /* Let the kernel reap children for us */
227 (void) sigaction(SIGCHLD
, &sa
, NULL
);
229 if (arg_crash_shell
) {
230 log_notice("Executing crash shell in 10s...");
233 pid
= raw_clone(SIGCHLD
);
235 log_emergency_errno(errno
, "Failed to fork off crash shell: %m");
238 (void) make_console_stdio();
239 (void) execle("/bin/sh", "/bin/sh", NULL
, environ
);
241 log_emergency_errno(errno
, "execle() failed: %m");
242 _exit(EXIT_EXCEPTION
);
244 log_info("Spawned crash shell as PID "PID_FMT
".", pid
);
245 (void) wait_for_terminate(pid
, NULL
);
249 freeze_or_exit_or_reboot();
252 static void install_crash_handler(void) {
253 static const struct sigaction sa
= {
255 .sa_flags
= SA_NODEFER
, /* So that we can raise the signal again from the signal handler */
259 /* We ignore the return value here, since, we don't mind if we
260 * cannot set up a crash handler */
261 r
= sigaction_many(&sa
, SIGNALS_CRASH_HANDLER
, -1);
263 log_debug_errno(r
, "I had trouble setting up the crash handler, ignoring: %m");
266 static int console_setup(void) {
267 _cleanup_close_
int tty_fd
= -1;
270 tty_fd
= open_terminal("/dev/console", O_WRONLY
|O_NOCTTY
|O_CLOEXEC
);
272 return log_error_errno(tty_fd
, "Failed to open /dev/console: %m");
274 /* We don't want to force text mode. plymouth may be showing
275 * pictures already from initrd. */
276 r
= reset_terminal_fd(tty_fd
, false);
278 return log_error_errno(r
, "Failed to reset /dev/console: %m");
283 static int parse_crash_chvt(const char *value
) {
286 if (safe_atoi(value
, &arg_crash_chvt
) >= 0)
289 b
= parse_boolean(value
);
294 arg_crash_chvt
= 0; /* switch to where kmsg goes */
296 arg_crash_chvt
= -1; /* turn off switching */
301 static int parse_confirm_spawn(const char *value
, char **console
) {
305 r
= value
? parse_boolean(value
) : 1;
311 if (r
> 0) /* on with default tty */
312 s
= strdup("/dev/console");
313 else if (is_path(value
)) /* on with fully qualified path */
315 else /* on with only a tty file name, not a fully qualified path */
316 s
= strjoin("/dev/", value
);
324 static int set_machine_id(const char *m
) {
328 if (sd_id128_from_string(m
, &t
) < 0)
331 if (sd_id128_is_null(t
))
338 static int parse_proc_cmdline_item(const char *key
, const char *value
, void *data
) {
344 if (STR_IN_SET(key
, "systemd.unit", "rd.systemd.unit")) {
346 if (proc_cmdline_value_missing(key
, value
))
349 if (!unit_name_is_valid(value
, UNIT_NAME_PLAIN
|UNIT_NAME_INSTANCE
))
350 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key
, value
);
351 else if (in_initrd() == !!startswith(key
, "rd.")) {
352 if (free_and_strdup(&arg_default_unit
, value
) < 0)
356 } else if (proc_cmdline_key_streq(key
, "systemd.dump_core")) {
358 r
= value
? parse_boolean(value
) : true;
360 log_warning_errno(r
, "Failed to parse dump core switch %s, ignoring: %m", value
);
364 } else if (proc_cmdline_key_streq(key
, "systemd.early_core_pattern")) {
366 if (proc_cmdline_value_missing(key
, value
))
369 if (path_is_absolute(value
))
370 (void) parse_path_argument_and_warn(value
, false, &arg_early_core_pattern
);
372 log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value
);
374 } else if (proc_cmdline_key_streq(key
, "systemd.crash_chvt")) {
377 arg_crash_chvt
= 0; /* turn on */
379 r
= parse_crash_chvt(value
);
381 log_warning_errno(r
, "Failed to parse crash chvt switch %s, ignoring: %m", value
);
384 } else if (proc_cmdline_key_streq(key
, "systemd.crash_shell")) {
386 r
= value
? parse_boolean(value
) : true;
388 log_warning_errno(r
, "Failed to parse crash shell switch %s, ignoring: %m", value
);
392 } else if (proc_cmdline_key_streq(key
, "systemd.crash_reboot")) {
394 r
= value
? parse_boolean(value
) : true;
396 log_warning_errno(r
, "Failed to parse crash reboot switch %s, ignoring: %m", value
);
398 arg_crash_reboot
= r
;
400 } else if (proc_cmdline_key_streq(key
, "systemd.confirm_spawn")) {
403 r
= parse_confirm_spawn(value
, &s
);
405 log_warning_errno(r
, "Failed to parse confirm_spawn switch %s, ignoring: %m", value
);
407 free_and_replace(arg_confirm_spawn
, s
);
409 } else if (proc_cmdline_key_streq(key
, "systemd.service_watchdogs")) {
411 r
= value
? parse_boolean(value
) : true;
413 log_warning_errno(r
, "Failed to parse service watchdog switch %s, ignoring: %m", value
);
415 arg_service_watchdogs
= r
;
417 } else if (proc_cmdline_key_streq(key
, "systemd.show_status")) {
420 r
= parse_show_status(value
, &arg_show_status
);
422 log_warning_errno(r
, "Failed to parse show status switch %s, ignoring: %m", value
);
424 arg_show_status
= SHOW_STATUS_YES
;
426 } else if (proc_cmdline_key_streq(key
, "systemd.default_standard_output")) {
428 if (proc_cmdline_value_missing(key
, value
))
431 r
= exec_output_from_string(value
);
433 log_warning_errno(r
, "Failed to parse default standard output switch %s, ignoring: %m", value
);
435 arg_default_std_output
= r
;
437 } else if (proc_cmdline_key_streq(key
, "systemd.default_standard_error")) {
439 if (proc_cmdline_value_missing(key
, value
))
442 r
= exec_output_from_string(value
);
444 log_warning_errno(r
, "Failed to parse default standard error switch %s, ignoring: %m", value
);
446 arg_default_std_error
= r
;
448 } else if (streq(key
, "systemd.setenv")) {
450 if (proc_cmdline_value_missing(key
, value
))
453 if (env_assignment_is_valid(value
)) {
456 env
= strv_env_set(arg_default_environment
, value
);
460 arg_default_environment
= env
;
462 log_warning("Environment variable name '%s' is not valid. Ignoring.", value
);
464 } else if (proc_cmdline_key_streq(key
, "systemd.machine_id")) {
466 if (proc_cmdline_value_missing(key
, value
))
469 r
= set_machine_id(value
);
471 log_warning_errno(r
, "MachineID '%s' is not valid, ignoring: %m", value
);
473 } else if (proc_cmdline_key_streq(key
, "systemd.default_timeout_start_sec")) {
475 if (proc_cmdline_value_missing(key
, value
))
478 r
= parse_sec(value
, &arg_default_timeout_start_usec
);
480 log_warning_errno(r
, "Failed to parse default start timeout '%s', ignoring: %m", value
);
482 if (arg_default_timeout_start_usec
<= 0)
483 arg_default_timeout_start_usec
= USEC_INFINITY
;
485 } else if (proc_cmdline_key_streq(key
, "systemd.watchdog_device")) {
487 if (proc_cmdline_value_missing(key
, value
))
490 (void) parse_path_argument_and_warn(value
, false, &arg_watchdog_device
);
492 } else if (streq(key
, "quiet") && !value
) {
494 if (arg_show_status
== _SHOW_STATUS_INVALID
)
495 arg_show_status
= SHOW_STATUS_AUTO
;
497 } else if (streq(key
, "debug") && !value
) {
499 /* Note that log_parse_environment() handles 'debug'
500 * too, and sets the log level to LOG_DEBUG. */
502 if (detect_container() > 0)
503 log_set_target(LOG_TARGET_CONSOLE
);
508 /* SysV compatibility */
509 target
= runlevel_to_target(key
);
511 return free_and_strdup(&arg_default_unit
, target
);
517 #define DEFINE_SETTER(name, func, descr) \
518 static int name(const char *unit, \
519 const char *filename, \
521 const char *section, \
522 unsigned section_line, \
523 const char *lvalue, \
525 const char *rvalue, \
537 log_syntax(unit, LOG_ERR, filename, line, r, \
538 "Invalid " descr "'%s': %m", \
544 DEFINE_SETTER(config_parse_level2
, log_set_max_level_from_string
, "log level");
545 DEFINE_SETTER(config_parse_target
, log_set_target_from_string
, "target");
546 DEFINE_SETTER(config_parse_color
, log_show_color_from_string
, "color" );
547 DEFINE_SETTER(config_parse_location
, log_show_location_from_string
, "location");
549 static int config_parse_cpu_affinity2(
551 const char *filename
,
554 unsigned section_line
,
561 _cleanup_cpu_free_ cpu_set_t
*c
= NULL
;
564 ncpus
= parse_cpu_set_and_warn(rvalue
, &c
, unit
, filename
, line
, lvalue
);
568 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus
), c
) < 0)
569 log_warning_errno(errno
, "Failed to set CPU affinity: %m");
574 static int config_parse_show_status(
576 const char *filename
,
579 unsigned section_line
,
587 ShowStatus
*b
= data
;
594 k
= parse_show_status(rvalue
, b
);
596 log_syntax(unit
, LOG_ERR
, filename
, line
, k
, "Failed to parse show status setting, ignoring: %s", rvalue
);
603 static int config_parse_output_restricted(
605 const char *filename
,
608 unsigned section_line
,
615 ExecOutput t
, *eo
= data
;
622 t
= exec_output_from_string(rvalue
);
624 log_syntax(unit
, LOG_ERR
, filename
, line
, 0, "Failed to parse output type, ignoring: %s", rvalue
);
628 if (IN_SET(t
, EXEC_OUTPUT_SOCKET
, EXEC_OUTPUT_NAMED_FD
, EXEC_OUTPUT_FILE
, EXEC_OUTPUT_FILE_APPEND
)) {
629 log_syntax(unit
, LOG_ERR
, filename
, line
, 0, "Standard output types socket, fd:, file:, append: are not supported as defaults, ignoring: %s", rvalue
);
637 static int config_parse_crash_chvt(
639 const char *filename
,
642 unsigned section_line
,
655 r
= parse_crash_chvt(rvalue
);
657 log_syntax(unit
, LOG_ERR
, filename
, line
, r
, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue
);
664 static int parse_config_file(void) {
666 const ConfigTableItem items
[] = {
667 { "Manager", "LogLevel", config_parse_level2
, 0, NULL
},
668 { "Manager", "LogTarget", config_parse_target
, 0, NULL
},
669 { "Manager", "LogColor", config_parse_color
, 0, NULL
},
670 { "Manager", "LogLocation", config_parse_location
, 0, NULL
},
671 { "Manager", "DumpCore", config_parse_bool
, 0, &arg_dump_core
},
672 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt
, 0, NULL
},
673 { "Manager", "CrashChangeVT", config_parse_crash_chvt
, 0, NULL
},
674 { "Manager", "CrashShell", config_parse_bool
, 0, &arg_crash_shell
},
675 { "Manager", "CrashReboot", config_parse_bool
, 0, &arg_crash_reboot
},
676 { "Manager", "ShowStatus", config_parse_show_status
, 0, &arg_show_status
},
677 { "Manager", "CPUAffinity", config_parse_cpu_affinity2
, 0, NULL
},
678 { "Manager", "JoinControllers", config_parse_warn_compat
, DISABLED_CONFIGURATION
, NULL
},
679 { "Manager", "RuntimeWatchdogSec", config_parse_sec
, 0, &arg_runtime_watchdog
},
680 { "Manager", "ShutdownWatchdogSec", config_parse_sec
, 0, &arg_shutdown_watchdog
},
681 { "Manager", "WatchdogDevice", config_parse_path
, 0, &arg_watchdog_device
},
682 { "Manager", "CapabilityBoundingSet", config_parse_capability_set
, 0, &arg_capability_bounding_set
},
683 { "Manager", "NoNewPrivileges", config_parse_bool
, 0, &arg_no_new_privs
},
685 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs
, 0, &arg_syscall_archs
},
687 { "Manager", "TimerSlackNSec", config_parse_nsec
, 0, &arg_timer_slack_nsec
},
688 { "Manager", "DefaultTimerAccuracySec", config_parse_sec
, 0, &arg_default_timer_accuracy_usec
},
689 { "Manager", "DefaultStandardOutput", config_parse_output_restricted
,0, &arg_default_std_output
},
690 { "Manager", "DefaultStandardError", config_parse_output_restricted
,0, &arg_default_std_error
},
691 { "Manager", "DefaultTimeoutStartSec", config_parse_sec
, 0, &arg_default_timeout_start_usec
},
692 { "Manager", "DefaultTimeoutStopSec", config_parse_sec
, 0, &arg_default_timeout_stop_usec
},
693 { "Manager", "DefaultRestartSec", config_parse_sec
, 0, &arg_default_restart_usec
},
694 { "Manager", "DefaultStartLimitInterval", config_parse_sec
, 0, &arg_default_start_limit_interval
}, /* obsolete alias */
695 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec
, 0, &arg_default_start_limit_interval
},
696 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned
, 0, &arg_default_start_limit_burst
},
697 { "Manager", "DefaultEnvironment", config_parse_environ
, 0, &arg_default_environment
},
698 { "Manager", "DefaultLimitCPU", config_parse_rlimit
, RLIMIT_CPU
, arg_default_rlimit
},
699 { "Manager", "DefaultLimitFSIZE", config_parse_rlimit
, RLIMIT_FSIZE
, arg_default_rlimit
},
700 { "Manager", "DefaultLimitDATA", config_parse_rlimit
, RLIMIT_DATA
, arg_default_rlimit
},
701 { "Manager", "DefaultLimitSTACK", config_parse_rlimit
, RLIMIT_STACK
, arg_default_rlimit
},
702 { "Manager", "DefaultLimitCORE", config_parse_rlimit
, RLIMIT_CORE
, arg_default_rlimit
},
703 { "Manager", "DefaultLimitRSS", config_parse_rlimit
, RLIMIT_RSS
, arg_default_rlimit
},
704 { "Manager", "DefaultLimitNOFILE", config_parse_rlimit
, RLIMIT_NOFILE
, arg_default_rlimit
},
705 { "Manager", "DefaultLimitAS", config_parse_rlimit
, RLIMIT_AS
, arg_default_rlimit
},
706 { "Manager", "DefaultLimitNPROC", config_parse_rlimit
, RLIMIT_NPROC
, arg_default_rlimit
},
707 { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit
, RLIMIT_MEMLOCK
, arg_default_rlimit
},
708 { "Manager", "DefaultLimitLOCKS", config_parse_rlimit
, RLIMIT_LOCKS
, arg_default_rlimit
},
709 { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit
, RLIMIT_SIGPENDING
, arg_default_rlimit
},
710 { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit
, RLIMIT_MSGQUEUE
, arg_default_rlimit
},
711 { "Manager", "DefaultLimitNICE", config_parse_rlimit
, RLIMIT_NICE
, arg_default_rlimit
},
712 { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit
, RLIMIT_RTPRIO
, arg_default_rlimit
},
713 { "Manager", "DefaultLimitRTTIME", config_parse_rlimit
, RLIMIT_RTTIME
, arg_default_rlimit
},
714 { "Manager", "DefaultCPUAccounting", config_parse_tristate
, 0, &arg_default_cpu_accounting
},
715 { "Manager", "DefaultIOAccounting", config_parse_bool
, 0, &arg_default_io_accounting
},
716 { "Manager", "DefaultIPAccounting", config_parse_bool
, 0, &arg_default_ip_accounting
},
717 { "Manager", "DefaultBlockIOAccounting", config_parse_bool
, 0, &arg_default_blockio_accounting
},
718 { "Manager", "DefaultMemoryAccounting", config_parse_bool
, 0, &arg_default_memory_accounting
},
719 { "Manager", "DefaultTasksAccounting", config_parse_bool
, 0, &arg_default_tasks_accounting
},
720 { "Manager", "DefaultTasksMax", config_parse_tasks_max
, 0, &arg_default_tasks_max
},
721 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action
, 0, &arg_cad_burst_action
},
725 const char *fn
, *conf_dirs_nulstr
;
728 PKGSYSCONFDIR
"/system.conf" :
729 PKGSYSCONFDIR
"/user.conf";
731 conf_dirs_nulstr
= arg_system
?
732 CONF_PATHS_NULSTR("systemd/system.conf.d") :
733 CONF_PATHS_NULSTR("systemd/user.conf.d");
735 (void) config_parse_many_nulstr(fn
, conf_dirs_nulstr
, "Manager\0", config_item_table_lookup
, items
, CONFIG_PARSE_WARN
, NULL
);
737 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
738 * like everywhere else. */
739 if (arg_default_timeout_start_usec
<= 0)
740 arg_default_timeout_start_usec
= USEC_INFINITY
;
741 if (arg_default_timeout_stop_usec
<= 0)
742 arg_default_timeout_stop_usec
= USEC_INFINITY
;
747 static void set_manager_defaults(Manager
*m
) {
751 /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
752 * affect the manager itself, but are just what newly allocated units will have set if they haven't set
753 * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
755 m
->default_timer_accuracy_usec
= arg_default_timer_accuracy_usec
;
756 m
->default_std_output
= arg_default_std_output
;
757 m
->default_std_error
= arg_default_std_error
;
758 m
->default_timeout_start_usec
= arg_default_timeout_start_usec
;
759 m
->default_timeout_stop_usec
= arg_default_timeout_stop_usec
;
760 m
->default_restart_usec
= arg_default_restart_usec
;
761 m
->default_start_limit_interval
= arg_default_start_limit_interval
;
762 m
->default_start_limit_burst
= arg_default_start_limit_burst
;
764 /* On 4.15+ with unified hierarchy, CPU accounting is essentially free as it doesn't require the CPU
765 * controller to be enabled, so the default is to enable it unless we got told otherwise. */
766 if (arg_default_cpu_accounting
>= 0)
767 m
->default_cpu_accounting
= arg_default_cpu_accounting
;
769 m
->default_cpu_accounting
= cpu_accounting_is_cheap();
771 m
->default_io_accounting
= arg_default_io_accounting
;
772 m
->default_ip_accounting
= arg_default_ip_accounting
;
773 m
->default_blockio_accounting
= arg_default_blockio_accounting
;
774 m
->default_memory_accounting
= arg_default_memory_accounting
;
775 m
->default_tasks_accounting
= arg_default_tasks_accounting
;
776 m
->default_tasks_max
= arg_default_tasks_max
;
778 (void) manager_set_default_rlimits(m
, arg_default_rlimit
);
780 (void) manager_default_environment(m
);
781 (void) manager_transient_environment_add(m
, arg_default_environment
);
784 static void set_manager_settings(Manager
*m
) {
788 /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
789 * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
791 m
->confirm_spawn
= arg_confirm_spawn
;
792 m
->service_watchdogs
= arg_service_watchdogs
;
793 m
->runtime_watchdog
= arg_runtime_watchdog
;
794 m
->shutdown_watchdog
= arg_shutdown_watchdog
;
795 m
->cad_burst_action
= arg_cad_burst_action
;
797 manager_set_show_status(m
, arg_show_status
);
800 static int parse_argv(int argc
, char *argv
[]) {
802 ARG_LOG_LEVEL
= 0x100,
812 ARG_DUMP_CONFIGURATION_ITEMS
,
813 ARG_DUMP_BUS_PROPERTIES
,
822 ARG_DEFAULT_STD_OUTPUT
,
823 ARG_DEFAULT_STD_ERROR
,
825 ARG_SERVICE_WATCHDOGS
,
828 static const struct option options
[] = {
829 { "log-level", required_argument
, NULL
, ARG_LOG_LEVEL
},
830 { "log-target", required_argument
, NULL
, ARG_LOG_TARGET
},
831 { "log-color", optional_argument
, NULL
, ARG_LOG_COLOR
},
832 { "log-location", optional_argument
, NULL
, ARG_LOG_LOCATION
},
833 { "unit", required_argument
, NULL
, ARG_UNIT
},
834 { "system", no_argument
, NULL
, ARG_SYSTEM
},
835 { "user", no_argument
, NULL
, ARG_USER
},
836 { "test", no_argument
, NULL
, ARG_TEST
},
837 { "no-pager", no_argument
, NULL
, ARG_NO_PAGER
},
838 { "help", no_argument
, NULL
, 'h' },
839 { "version", no_argument
, NULL
, ARG_VERSION
},
840 { "dump-configuration-items", no_argument
, NULL
, ARG_DUMP_CONFIGURATION_ITEMS
},
841 { "dump-bus-properties", no_argument
, NULL
, ARG_DUMP_BUS_PROPERTIES
},
842 { "dump-core", optional_argument
, NULL
, ARG_DUMP_CORE
},
843 { "crash-chvt", required_argument
, NULL
, ARG_CRASH_CHVT
},
844 { "crash-shell", optional_argument
, NULL
, ARG_CRASH_SHELL
},
845 { "crash-reboot", optional_argument
, NULL
, ARG_CRASH_REBOOT
},
846 { "confirm-spawn", optional_argument
, NULL
, ARG_CONFIRM_SPAWN
},
847 { "show-status", optional_argument
, NULL
, ARG_SHOW_STATUS
},
848 { "deserialize", required_argument
, NULL
, ARG_DESERIALIZE
},
849 { "switched-root", no_argument
, NULL
, ARG_SWITCHED_ROOT
},
850 { "default-standard-output", required_argument
, NULL
, ARG_DEFAULT_STD_OUTPUT
, },
851 { "default-standard-error", required_argument
, NULL
, ARG_DEFAULT_STD_ERROR
, },
852 { "machine-id", required_argument
, NULL
, ARG_MACHINE_ID
},
853 { "service-watchdogs", required_argument
, NULL
, ARG_SERVICE_WATCHDOGS
},
862 if (getpid_cached() == 1)
865 while ((c
= getopt_long(argc
, argv
, "hDbsz:", options
, NULL
)) >= 0)
870 r
= log_set_max_level_from_string(optarg
);
872 return log_error_errno(r
, "Failed to parse log level \"%s\": %m", optarg
);
877 r
= log_set_target_from_string(optarg
);
879 return log_error_errno(r
, "Failed to parse log target \"%s\": %m", optarg
);
886 r
= log_show_color_from_string(optarg
);
888 return log_error_errno(r
, "Failed to parse log color setting \"%s\": %m",
891 log_show_color(true);
895 case ARG_LOG_LOCATION
:
897 r
= log_show_location_from_string(optarg
);
899 return log_error_errno(r
, "Failed to parse log location setting \"%s\": %m",
902 log_show_location(true);
906 case ARG_DEFAULT_STD_OUTPUT
:
907 r
= exec_output_from_string(optarg
);
909 return log_error_errno(r
, "Failed to parse default standard output setting \"%s\": %m",
911 arg_default_std_output
= r
;
914 case ARG_DEFAULT_STD_ERROR
:
915 r
= exec_output_from_string(optarg
);
917 return log_error_errno(r
, "Failed to parse default standard error output setting \"%s\": %m",
919 arg_default_std_error
= r
;
923 r
= free_and_strdup(&arg_default_unit
, optarg
);
925 return log_error_errno(r
, "Failed to set default unit \"%s\": %m", optarg
);
938 arg_action
= ACTION_TEST
;
942 arg_pager_flags
|= PAGER_DISABLE
;
946 arg_action
= ACTION_VERSION
;
949 case ARG_DUMP_CONFIGURATION_ITEMS
:
950 arg_action
= ACTION_DUMP_CONFIGURATION_ITEMS
;
953 case ARG_DUMP_BUS_PROPERTIES
:
954 arg_action
= ACTION_DUMP_BUS_PROPERTIES
;
959 arg_dump_core
= true;
961 r
= parse_boolean(optarg
);
963 return log_error_errno(r
, "Failed to parse dump core boolean: \"%s\": %m",
970 r
= parse_crash_chvt(optarg
);
972 return log_error_errno(r
, "Failed to parse crash virtual terminal index: \"%s\": %m",
976 case ARG_CRASH_SHELL
:
978 arg_crash_shell
= true;
980 r
= parse_boolean(optarg
);
982 return log_error_errno(r
, "Failed to parse crash shell boolean: \"%s\": %m",
988 case ARG_CRASH_REBOOT
:
990 arg_crash_reboot
= true;
992 r
= parse_boolean(optarg
);
994 return log_error_errno(r
, "Failed to parse crash shell boolean: \"%s\": %m",
996 arg_crash_reboot
= r
;
1000 case ARG_CONFIRM_SPAWN
:
1001 arg_confirm_spawn
= mfree(arg_confirm_spawn
);
1003 r
= parse_confirm_spawn(optarg
, &arg_confirm_spawn
);
1005 return log_error_errno(r
, "Failed to parse confirm spawn option: \"%s\": %m",
1009 case ARG_SERVICE_WATCHDOGS
:
1010 r
= parse_boolean(optarg
);
1012 return log_error_errno(r
, "Failed to parse service watchdogs boolean: \"%s\": %m",
1014 arg_service_watchdogs
= r
;
1017 case ARG_SHOW_STATUS
:
1019 r
= parse_show_status(optarg
, &arg_show_status
);
1021 return log_error_errno(r
, "Failed to parse show status boolean: \"%s\": %m",
1024 arg_show_status
= SHOW_STATUS_YES
;
1027 case ARG_DESERIALIZE
: {
1031 r
= safe_atoi(optarg
, &fd
);
1033 log_error_errno(r
, "Failed to parse deserialize option \"%s\": %m", optarg
);
1035 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
),
1036 "Invalid deserialize fd: %d",
1039 (void) fd_cloexec(fd
, true);
1041 f
= fdopen(fd
, "r");
1043 return log_error_errno(errno
, "Failed to open serialization fd %d: %m", fd
);
1045 safe_fclose(arg_serialization
);
1046 arg_serialization
= f
;
1051 case ARG_SWITCHED_ROOT
:
1052 arg_switched_root
= true;
1055 case ARG_MACHINE_ID
:
1056 r
= set_machine_id(optarg
);
1058 return log_error_errno(r
, "MachineID '%s' is not valid: %m", optarg
);
1062 arg_action
= ACTION_HELP
;
1066 log_set_max_level(LOG_DEBUG
);
1072 /* Just to eat away the sysvinit kernel
1073 * cmdline args without getopt() error
1074 * messages that we'll parse in
1075 * parse_proc_cmdline_word() or ignore. */
1078 if (getpid_cached() != 1)
1084 assert_not_reached("Unhandled option code.");
1087 if (optind
< argc
&& getpid_cached() != 1) {
1088 /* Hmm, when we aren't run as init system
1089 * let's complain about excess arguments */
1091 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
),
1092 "Excess arguments.");
1098 static int help(void) {
1099 _cleanup_free_
char *link
= NULL
;
1102 r
= terminal_urlify_man("systemd", "1", &link
);
1106 printf("%s [OPTIONS...]\n\n"
1107 "Starts up and maintains the system or user services.\n\n"
1108 " -h --help Show this help\n"
1109 " --version Show version\n"
1110 " --test Determine startup sequence, dump it and exit\n"
1111 " --no-pager Do not pipe output into a pager\n"
1112 " --dump-configuration-items Dump understood unit configuration items\n"
1113 " --dump-bus-properties Dump exposed bus properties\n"
1114 " --unit=UNIT Set default unit\n"
1115 " --system Run a system instance, even if PID != 1\n"
1116 " --user Run a user instance\n"
1117 " --dump-core[=BOOL] Dump core on crash\n"
1118 " --crash-vt=NR Change to specified VT on crash\n"
1119 " --crash-reboot[=BOOL] Reboot on crash\n"
1120 " --crash-shell[=BOOL] Run shell on crash\n"
1121 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1122 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1123 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1124 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1125 " --log-color[=BOOL] Highlight important log messages\n"
1126 " --log-location[=BOOL] Include code location in log messages\n"
1127 " --default-standard-output= Set default standard output for services\n"
1128 " --default-standard-error= Set default standard error output for services\n"
1129 "\nSee the %s for details.\n"
1130 , program_invocation_short_name
1137 static int prepare_reexecute(
1141 bool switching_root
) {
1143 _cleanup_fdset_free_ FDSet
*fds
= NULL
;
1144 _cleanup_fclose_
FILE *f
= NULL
;
1151 r
= manager_open_serialization(m
, &f
);
1153 return log_error_errno(r
, "Failed to create serialization file: %m");
1155 /* Make sure nothing is really destructed when we shut down */
1157 bus_manager_send_reloading(m
, true);
1163 r
= manager_serialize(m
, f
, fds
, switching_root
);
1167 if (fseeko(f
, 0, SEEK_SET
) == (off_t
) -1)
1168 return log_error_errno(errno
, "Failed to rewind serialization fd: %m");
1170 r
= fd_cloexec(fileno(f
), false);
1172 return log_error_errno(r
, "Failed to disable O_CLOEXEC for serialization: %m");
1174 r
= fdset_cloexec(fds
, false);
1176 return log_error_errno(r
, "Failed to disable O_CLOEXEC for serialization fds: %m");
1178 *ret_f
= TAKE_PTR(f
);
1179 *ret_fds
= TAKE_PTR(fds
);
1184 static void bump_file_max_and_nr_open(void) {
1186 /* Let's bump fs.file-max and fs.nr_open to their respective maximums. On current kernels large numbers of file
1187 * descriptors are no longer a performance problem and their memory is properly tracked by memcg, thus counting
1188 * them and limiting them in another two layers of limits is unnecessary and just complicates things. This
1189 * function hence turns off 2 of the 4 levels of limits on file descriptors, and makes RLIMIT_NOLIMIT (soft +
1190 * hard) the only ones that really matter. */
1192 #if BUMP_PROC_SYS_FS_FILE_MAX || BUMP_PROC_SYS_FS_NR_OPEN
1193 _cleanup_free_
char *t
= NULL
;
1197 #if BUMP_PROC_SYS_FS_FILE_MAX
1198 /* I so wanted to use STRINGIFY(ULONG_MAX) here, but alas we can't as glibc/gcc define that as
1199 * "(0x7fffffffffffffffL * 2UL + 1UL)". Seriously. 😢 */
1200 if (asprintf(&t
, "%lu\n", ULONG_MAX
) < 0) {
1205 r
= sysctl_write("fs/file-max", t
);
1207 log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
, "Failed to bump fs.file-max, ignoring: %m");
1210 #if BUMP_PROC_SYS_FS_FILE_MAX && BUMP_PROC_SYS_FS_NR_OPEN
1214 #if BUMP_PROC_SYS_FS_NR_OPEN
1217 /* Arg! The kernel enforces maximum and minimum values on the fs.nr_open, but we don't really know what they
1218 * are. The expression by which the maximum is determined is dependent on the architecture, and is something we
1219 * don't really want to copy to userspace, as it is dependent on implementation details of the kernel. Since
1220 * the kernel doesn't expose the maximum value to us, we can only try and hope. Hence, let's start with
1221 * INT_MAX, and then keep halving the value until we find one that works. Ugly? Yes, absolutely, but kernel
1222 * APIs are kernel APIs, so what do can we do... 🤯 */
1227 v
&= ~(__SIZEOF_POINTER__
- 1); /* Round down to next multiple of the pointer size */
1229 log_warning("Can't bump fs.nr_open, value too small.");
1235 log_error_errno(k
, "Failed to read fs.nr_open: %m");
1238 if (k
>= v
) { /* Already larger */
1239 log_debug("Skipping bump, value is already larger.");
1243 if (asprintf(&t
, "%i\n", v
) < 0) {
1248 r
= sysctl_write("fs/nr_open", t
);
1251 log_debug("Couldn't write fs.nr_open as %i, halving it.", v
);
1256 log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
, "Failed to bump fs.nr_open, ignoring: %m");
1260 log_debug("Successfully bumped fs.nr_open to %i", v
);
1266 static int bump_rlimit_nofile(struct rlimit
*saved_rlimit
) {
1269 assert(saved_rlimit
);
1271 /* Save the original RLIMIT_NOFILE so that we can reset it later when transitioning from the initrd to the main
1272 * systemd or suchlike. */
1273 if (getrlimit(RLIMIT_NOFILE
, saved_rlimit
) < 0)
1274 return log_warning_errno(errno
, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1276 /* Get the underlying absolute limit the kernel enforces */
1277 nr
= read_nr_open();
1279 /* Make sure forked processes get limits based on the original kernel setting */
1280 if (!arg_default_rlimit
[RLIMIT_NOFILE
]) {
1283 rl
= newdup(struct rlimit
, saved_rlimit
, 1);
1287 /* Bump the hard limit for system services to a substantially higher value. The default hard limit
1288 * current kernels set is pretty low (4K), mostly for historical reasons. According to kernel
1289 * developers, the fd handling in recent kernels has been optimized substantially enough, so that we
1290 * can bump the limit now, without paying too high a price in memory or performance. Note however that
1291 * we only bump the hard limit, not the soft limit. That's because select() works the way it works, and
1292 * chokes on fds >= 1024. If we'd bump the soft limit globally, it might accidentally happen to
1293 * unexpecting programs that they get fds higher than what they can process using select(). By only
1294 * bumping the hard limit but leaving the low limit as it is we avoid this pitfall: programs that are
1295 * written by folks aware of the select() problem in mind (and thus use poll()/epoll instead of
1296 * select(), the way everybody should) can explicitly opt into high fds by bumping their soft limit
1297 * beyond 1024, to the hard limit we pass. */
1299 rl
->rlim_max
= MIN((rlim_t
) nr
, MAX(rl
->rlim_max
, (rlim_t
) HIGH_RLIMIT_NOFILE
));
1301 arg_default_rlimit
[RLIMIT_NOFILE
] = rl
;
1304 /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for
1305 * both hard and soft. */
1306 r
= setrlimit_closest(RLIMIT_NOFILE
, &RLIMIT_MAKE_CONST(nr
));
1308 return log_warning_errno(r
, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1313 static int bump_rlimit_memlock(struct rlimit
*saved_rlimit
) {
1316 assert(saved_rlimit
);
1318 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK which should
1319 * normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's bump
1320 * the value high enough for our user. */
1322 if (getrlimit(RLIMIT_MEMLOCK
, saved_rlimit
) < 0)
1323 return log_warning_errno(errno
, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1325 r
= setrlimit_closest(RLIMIT_MEMLOCK
, &RLIMIT_MAKE_CONST(HIGH_RLIMIT_MEMLOCK
));
1327 return log_warning_errno(r
, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1332 static void test_usr(void) {
1334 /* Check that /usr is not a separate fs */
1336 if (dir_is_empty("/usr") <= 0)
1339 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1340 "Some things will probably break (sometimes even silently) in mysterious ways. "
1341 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1344 static int enforce_syscall_archs(Set
*archs
) {
1348 if (!is_seccomp_available())
1351 r
= seccomp_restrict_archs(arg_syscall_archs
);
1353 return log_error_errno(r
, "Failed to enforce system call architecture restrication: %m");
1358 static int status_welcome(void) {
1359 _cleanup_free_
char *pretty_name
= NULL
, *ansi_color
= NULL
;
1362 if (IN_SET(arg_show_status
, SHOW_STATUS_NO
, SHOW_STATUS_AUTO
))
1365 r
= parse_os_release(NULL
,
1366 "PRETTY_NAME", &pretty_name
,
1367 "ANSI_COLOR", &ansi_color
,
1370 log_full_errno(r
== -ENOENT
? LOG_DEBUG
: LOG_WARNING
, r
,
1371 "Failed to read os-release file, ignoring: %m");
1373 if (log_get_show_color())
1374 return status_printf(NULL
, false, false,
1375 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1376 isempty(ansi_color
) ? "1" : ansi_color
,
1377 isempty(pretty_name
) ? "Linux" : pretty_name
);
1379 return status_printf(NULL
, false, false,
1380 "\nWelcome to %s!\n",
1381 isempty(pretty_name
) ? "Linux" : pretty_name
);
1384 static int write_container_id(void) {
1388 c
= getenv("container");
1392 RUN_WITH_UMASK(0022)
1393 r
= write_string_file("/run/systemd/container", c
, WRITE_STRING_FILE_CREATE
);
1395 return log_warning_errno(r
, "Failed to write /run/systemd/container, ignoring: %m");
1400 static int bump_unix_max_dgram_qlen(void) {
1401 _cleanup_free_
char *qlen
= NULL
;
1405 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1406 * really really early during boot, so that it is actually applied to all our sockets, including the
1407 * $NOTIFY_SOCKET one. */
1409 r
= read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen
);
1411 return log_full_errno(r
== -ENOENT
? LOG_DEBUG
: LOG_WARNING
, r
, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1413 r
= safe_atolu(qlen
, &v
);
1415 return log_warning_errno(r
, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen
);
1417 if (v
>= DEFAULT_UNIX_MAX_DGRAM_QLEN
)
1420 r
= write_string_filef("/proc/sys/net/unix/max_dgram_qlen", WRITE_STRING_FILE_DISABLE_BUFFER
, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN
);
1422 return log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
,
1423 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1428 static int fixup_environment(void) {
1429 _cleanup_free_
char *term
= NULL
;
1433 /* Only fix up the environment when we are started as PID 1 */
1434 if (getpid_cached() != 1)
1437 /* We expect the environment to be set correctly if run inside a container. */
1438 if (detect_container() > 0)
1441 /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1442 * device used by the console. We try to make a better guess here since some consoles might not have support
1443 * for color mode for example.
1445 * However if TERM was configured through the kernel command line then leave it alone. */
1446 r
= proc_cmdline_get_key("TERM", 0, &term
);
1450 t
= term
?: default_term_for_tty("/dev/console");
1452 if (setenv("TERM", t
, 1) < 0)
1458 static void redirect_telinit(int argc
, char *argv
[]) {
1460 /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1462 #if HAVE_SYSV_COMPAT
1463 if (getpid_cached() == 1)
1466 if (!strstr(program_invocation_short_name
, "init"))
1469 execv(SYSTEMCTL_BINARY_PATH
, argv
);
1470 log_error_errno(errno
, "Failed to exec " SYSTEMCTL_BINARY_PATH
": %m");
1475 static int become_shutdown(
1476 const char *shutdown_verb
,
1479 char log_level
[DECIMAL_STR_MAX(int) + 1],
1480 exit_code
[DECIMAL_STR_MAX(uint8_t) + 1],
1481 timeout
[DECIMAL_STR_MAX(usec_t
) + 1];
1483 const char* command_line
[13] = {
1484 SYSTEMD_SHUTDOWN_BINARY_PATH
,
1486 "--timeout", timeout
,
1487 "--log-level", log_level
,
1491 _cleanup_strv_free_
char **env_block
= NULL
;
1495 assert(shutdown_verb
);
1496 assert(!command_line
[pos
]);
1497 env_block
= strv_copy(environ
);
1499 xsprintf(log_level
, "%d", log_get_max_level());
1500 xsprintf(timeout
, "%" PRI_USEC
"us", arg_default_timeout_stop_usec
);
1502 switch (log_get_target()) {
1504 case LOG_TARGET_KMSG
:
1505 case LOG_TARGET_JOURNAL_OR_KMSG
:
1506 case LOG_TARGET_SYSLOG_OR_KMSG
:
1507 command_line
[pos
++] = "kmsg";
1510 case LOG_TARGET_NULL
:
1511 command_line
[pos
++] = "null";
1514 case LOG_TARGET_CONSOLE
:
1516 command_line
[pos
++] = "console";
1520 if (log_get_show_color())
1521 command_line
[pos
++] = "--log-color";
1523 if (log_get_show_location())
1524 command_line
[pos
++] = "--log-location";
1526 if (streq(shutdown_verb
, "exit")) {
1527 command_line
[pos
++] = "--exit-code";
1528 command_line
[pos
++] = exit_code
;
1529 xsprintf(exit_code
, "%d", retval
);
1532 assert(pos
< ELEMENTSOF(command_line
));
1534 if (streq(shutdown_verb
, "reboot") &&
1535 arg_shutdown_watchdog
> 0 &&
1536 arg_shutdown_watchdog
!= USEC_INFINITY
) {
1540 /* If we reboot let's set the shutdown
1541 * watchdog and tell the shutdown binary to
1542 * repeatedly ping it */
1543 r
= watchdog_set_timeout(&arg_shutdown_watchdog
);
1544 watchdog_close(r
< 0);
1546 /* Tell the binary how often to ping, ignore failure */
1547 if (asprintf(&e
, "WATCHDOG_USEC="USEC_FMT
, arg_shutdown_watchdog
) > 0)
1548 (void) strv_consume(&env_block
, e
);
1550 if (arg_watchdog_device
&&
1551 asprintf(&e
, "WATCHDOG_DEVICE=%s", arg_watchdog_device
) > 0)
1552 (void) strv_consume(&env_block
, e
);
1554 watchdog_close(true);
1556 /* Avoid the creation of new processes forked by the
1557 * kernel; at this point, we will not listen to the
1559 if (detect_container() <= 0)
1560 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER
);
1562 execve(SYSTEMD_SHUTDOWN_BINARY_PATH
, (char **) command_line
, env_block
);
1566 static void initialize_clock(void) {
1569 if (clock_is_localtime(NULL
) > 0) {
1573 * The very first call of settimeofday() also does a time warp in the kernel.
1575 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1576 * of maintaining the RTC and do all adjustments. This matches the behavior of Windows, which leaves
1577 * the RTC alone if the registry tells that the RTC runs in UTC.
1579 r
= clock_set_timezone(&min
);
1581 log_error_errno(r
, "Failed to apply local time delta, ignoring: %m");
1583 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min
);
1585 } else if (!in_initrd()) {
1587 * Do a dummy very first call to seal the kernel's time warp magic.
1589 * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1590 * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1591 * until we reach the real system.
1593 * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1594 * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1595 * be treated as UTC that way.
1597 (void) clock_reset_timewarp();
1600 r
= clock_apply_epoch();
1602 log_error_errno(r
, "Current system time is before build time, but cannot correct: %m");
1604 log_info("System time before build time, advancing clock.");
1607 static void initialize_coredump(bool skip_setup
) {
1609 if (getpid_cached() != 1)
1612 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1613 * will process core dumps for system services by default. */
1614 if (setrlimit(RLIMIT_CORE
, &RLIMIT_MAKE_CONST(RLIM_INFINITY
)) < 0)
1615 log_warning_errno(errno
, "Failed to set RLIMIT_CORE: %m");
1617 /* But at the same time, turn off the core_pattern logic by default, so that no
1618 * coredumps are stored until the systemd-coredump tool is enabled via
1619 * sysctl. However it can be changed via the kernel command line later so core
1620 * dumps can still be generated during early startup and in initramfs. */
1622 disable_coredumps();
1626 static void initialize_core_pattern(bool skip_setup
) {
1629 if (skip_setup
|| !arg_early_core_pattern
)
1632 if (getpid_cached() != 1)
1635 r
= write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern
, WRITE_STRING_FILE_DISABLE_BUFFER
);
1637 log_warning_errno(r
, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", arg_early_core_pattern
);
1640 static void do_reexecute(
1643 const struct rlimit
*saved_rlimit_nofile
,
1644 const struct rlimit
*saved_rlimit_memlock
,
1646 const char *switch_root_dir
,
1647 const char *switch_root_init
,
1648 const char **ret_error_message
) {
1650 unsigned i
, j
, args_size
;
1654 assert(saved_rlimit_nofile
);
1655 assert(saved_rlimit_memlock
);
1656 assert(ret_error_message
);
1658 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1660 watchdog_close(true);
1662 /* Reset the RLIMIT_NOFILE to the kernel default, so that the new systemd can pass the kernel default to its
1663 * child processes */
1665 if (saved_rlimit_nofile
->rlim_cur
> 0)
1666 (void) setrlimit(RLIMIT_NOFILE
, saved_rlimit_nofile
);
1667 if (saved_rlimit_memlock
->rlim_cur
!= (rlim_t
) -1)
1668 (void) setrlimit(RLIMIT_MEMLOCK
, saved_rlimit_memlock
);
1670 if (switch_root_dir
) {
1671 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1672 * SIGCHLD for them after deserializing. */
1673 broadcast_signal(SIGTERM
, false, true, arg_default_timeout_stop_usec
);
1675 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1676 r
= switch_root(switch_root_dir
, "/mnt", true, MS_MOVE
);
1678 log_error_errno(r
, "Failed to switch root, trying to continue: %m");
1681 args_size
= MAX(6, argc
+1);
1682 args
= newa(const char*, args_size
);
1684 if (!switch_root_init
) {
1685 char sfd
[DECIMAL_STR_MAX(int) + 1];
1687 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1688 * the user didn't specify an explicit init to spawn. */
1690 assert(arg_serialization
);
1693 xsprintf(sfd
, "%i", fileno(arg_serialization
));
1696 args
[i
++] = SYSTEMD_BINARY_PATH
;
1697 if (switch_root_dir
)
1698 args
[i
++] = "--switched-root";
1699 args
[i
++] = arg_system
? "--system" : "--user";
1700 args
[i
++] = "--deserialize";
1704 assert(i
<= args_size
);
1707 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do this is on
1708 * its own on exec(), but it will do it on exit(). Hence, to ensure we get a summary here, fork() off
1709 * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1710 * before proceeding into the exec().
1712 valgrind_summary_hack();
1714 (void) execv(args
[0], (char* const*) args
);
1715 log_debug_errno(errno
, "Failed to execute our own binary, trying fallback: %m");
1718 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1719 * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1720 * doesn't matter.) */
1722 arg_serialization
= safe_fclose(arg_serialization
);
1723 fds
= fdset_free(fds
);
1725 /* Reopen the console */
1726 (void) make_console_stdio();
1728 for (j
= 1, i
= 1; j
< (unsigned) argc
; j
++)
1729 args
[i
++] = argv
[j
];
1731 assert(i
<= args_size
);
1733 /* Reenable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1734 (void) reset_all_signal_handlers();
1735 (void) reset_signal_mask();
1737 if (switch_root_init
) {
1738 args
[0] = switch_root_init
;
1739 (void) execv(args
[0], (char* const*) args
);
1740 log_warning_errno(errno
, "Failed to execute configured init, trying fallback: %m");
1743 args
[0] = "/sbin/init";
1744 (void) execv(args
[0], (char* const*) args
);
1747 manager_status_printf(NULL
, STATUS_TYPE_EMERGENCY
,
1748 ANSI_HIGHLIGHT_RED
" !! " ANSI_NORMAL
,
1749 "Failed to execute /sbin/init");
1752 log_warning("No /sbin/init, trying fallback");
1754 args
[0] = "/bin/sh";
1756 (void) execv(args
[0], (char* const*) args
);
1757 log_error_errno(errno
, "Failed to execute /bin/sh, giving up: %m");
1759 log_warning_errno(r
, "Failed to execute /sbin/init, giving up: %m");
1761 *ret_error_message
= "Failed to execute fallback shell";
1764 static int invoke_main_loop(
1766 bool *ret_reexecute
,
1767 int *ret_retval
, /* Return parameters relevant for shutting down */
1768 const char **ret_shutdown_verb
, /* … */
1769 FDSet
**ret_fds
, /* Return parameters for reexecuting */
1770 char **ret_switch_root_dir
, /* … */
1771 char **ret_switch_root_init
, /* … */
1772 const char **ret_error_message
) {
1777 assert(ret_reexecute
);
1779 assert(ret_shutdown_verb
);
1781 assert(ret_switch_root_dir
);
1782 assert(ret_switch_root_init
);
1783 assert(ret_error_message
);
1786 r
= manager_loop(m
);
1788 *ret_error_message
= "Failed to run main loop";
1789 return log_emergency_errno(r
, "Failed to run main loop: %m");
1792 switch ((ManagerObjective
) r
) {
1794 case MANAGER_RELOAD
: {
1795 LogTarget saved_log_target
;
1796 int saved_log_level
;
1798 log_info("Reloading.");
1800 /* First, save any overridden log level/target, then parse the configuration file, which might
1801 * change the log level to new settings. */
1803 saved_log_level
= m
->log_level_overridden
? log_get_max_level() : -1;
1804 saved_log_target
= m
->log_target_overridden
? log_get_target() : _LOG_TARGET_INVALID
;
1806 r
= parse_config_file();
1808 log_warning_errno(r
, "Failed to parse config file, ignoring: %m");
1810 set_manager_defaults(m
);
1812 if (saved_log_level
>= 0)
1813 manager_override_log_level(m
, saved_log_level
);
1814 if (saved_log_target
>= 0)
1815 manager_override_log_target(m
, saved_log_target
);
1817 r
= manager_reload(m
);
1819 /* Reloading failed before the point of no return. Let's continue running as if nothing happened. */
1820 m
->objective
= MANAGER_OK
;
1825 case MANAGER_REEXECUTE
:
1827 r
= prepare_reexecute(m
, &arg_serialization
, ret_fds
, false);
1829 *ret_error_message
= "Failed to prepare for reexecution";
1833 log_notice("Reexecuting.");
1835 *ret_reexecute
= true;
1836 *ret_retval
= EXIT_SUCCESS
;
1837 *ret_shutdown_verb
= NULL
;
1838 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
1842 case MANAGER_SWITCH_ROOT
:
1843 if (!m
->switch_root_init
) {
1844 r
= prepare_reexecute(m
, &arg_serialization
, ret_fds
, true);
1846 *ret_error_message
= "Failed to prepare for reexecution";
1852 log_notice("Switching root.");
1854 *ret_reexecute
= true;
1855 *ret_retval
= EXIT_SUCCESS
;
1856 *ret_shutdown_verb
= NULL
;
1858 /* Steal the switch root parameters */
1859 *ret_switch_root_dir
= m
->switch_root
;
1860 *ret_switch_root_init
= m
->switch_root_init
;
1861 m
->switch_root
= m
->switch_root_init
= NULL
;
1867 if (MANAGER_IS_USER(m
)) {
1870 *ret_reexecute
= false;
1871 *ret_retval
= m
->return_value
;
1872 *ret_shutdown_verb
= NULL
;
1874 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
1880 case MANAGER_REBOOT
:
1881 case MANAGER_POWEROFF
:
1883 case MANAGER_KEXEC
: {
1884 static const char * const table
[_MANAGER_OBJECTIVE_MAX
] = {
1885 [MANAGER_EXIT
] = "exit",
1886 [MANAGER_REBOOT
] = "reboot",
1887 [MANAGER_POWEROFF
] = "poweroff",
1888 [MANAGER_HALT
] = "halt",
1889 [MANAGER_KEXEC
] = "kexec",
1892 log_notice("Shutting down.");
1894 *ret_reexecute
= false;
1895 *ret_retval
= m
->return_value
;
1896 assert_se(*ret_shutdown_verb
= table
[m
->objective
]);
1898 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
1904 assert_not_reached("Unknown or unexpected manager objective.");
1909 static void log_execution_mode(bool *ret_first_boot
) {
1910 assert(ret_first_boot
);
1915 log_info(PACKAGE_STRING
" running in %ssystem mode. (" SYSTEMD_FEATURES
")",
1916 arg_action
== ACTION_TEST
? "test " : "" );
1918 v
= detect_virtualization();
1920 log_info("Detected virtualization %s.", virtualization_to_string(v
));
1922 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1925 *ret_first_boot
= false;
1926 log_info("Running in initial RAM disk.");
1928 /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
1929 * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
1930 * doesn't it's unpopulated. This allows container managers and installers to provision a
1931 * couple of files already. If the container manager wants to provision the machine ID itself
1932 * it should pass $container_uuid to PID 1. */
1934 *ret_first_boot
= access("/etc/machine-id", F_OK
) < 0;
1935 if (*ret_first_boot
)
1936 log_info("Running with unpopulated /etc.");
1939 if (DEBUG_LOGGING
) {
1940 _cleanup_free_
char *t
;
1942 t
= uid_to_name(getuid());
1943 log_debug(PACKAGE_STRING
" running in %suser mode for user " UID_FMT
"/%s. (" SYSTEMD_FEATURES
")",
1944 arg_action
== ACTION_TEST
? " test" : "", getuid(), strna(t
));
1947 *ret_first_boot
= false;
1951 static int initialize_runtime(
1953 struct rlimit
*saved_rlimit_nofile
,
1954 struct rlimit
*saved_rlimit_memlock
,
1955 const char **ret_error_message
) {
1959 assert(ret_error_message
);
1961 /* Sets up various runtime parameters. Many of these initializations are conditionalized:
1963 * - Some only apply to --system instances
1964 * - Some only apply to --user instances
1965 * - Some only apply when we first start up, but not when we reexecute
1968 if (arg_action
!= ACTION_RUN
)
1972 /* Make sure we leave a core dump without panicing the kernel. */
1973 install_crash_handler();
1976 r
= mount_cgroup_controllers();
1978 *ret_error_message
= "Failed to mount cgroup hierarchies";
1984 machine_id_setup(NULL
, arg_machine_id
, NULL
);
1986 bump_unix_max_dgram_qlen();
1987 bump_file_max_and_nr_open();
1989 write_container_id();
1992 if (arg_watchdog_device
) {
1993 r
= watchdog_set_device(arg_watchdog_device
);
1995 log_warning_errno(r
, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device
);
1998 if (arg_runtime_watchdog
> 0 && arg_runtime_watchdog
!= USEC_INFINITY
)
1999 watchdog_set_timeout(&arg_runtime_watchdog
);
2002 if (arg_timer_slack_nsec
!= NSEC_INFINITY
)
2003 if (prctl(PR_SET_TIMERSLACK
, arg_timer_slack_nsec
) < 0)
2004 log_warning_errno(errno
, "Failed to adjust timer slack, ignoring: %m");
2006 if (arg_system
&& !cap_test_all(arg_capability_bounding_set
)) {
2007 r
= capability_bounding_set_drop_usermode(arg_capability_bounding_set
);
2009 *ret_error_message
= "Failed to drop capability bounding set of usermode helpers";
2010 return log_emergency_errno(r
, "Failed to drop capability bounding set of usermode helpers: %m");
2013 r
= capability_bounding_set_drop(arg_capability_bounding_set
, true);
2015 *ret_error_message
= "Failed to drop capability bounding set";
2016 return log_emergency_errno(r
, "Failed to drop capability bounding set: %m");
2020 if (arg_system
&& arg_no_new_privs
) {
2021 if (prctl(PR_SET_NO_NEW_PRIVS
, 1, 0, 0, 0) < 0) {
2022 *ret_error_message
= "Failed to disable new privileges";
2023 return log_emergency_errno(errno
, "Failed to disable new privileges: %m");
2027 if (arg_syscall_archs
) {
2028 r
= enforce_syscall_archs(arg_syscall_archs
);
2030 *ret_error_message
= "Failed to set syscall architectures";
2036 /* Become reaper of our children */
2037 if (prctl(PR_SET_CHILD_SUBREAPER
, 1) < 0)
2038 log_warning_errno(errno
, "Failed to make us a subreaper: %m");
2040 /* Bump up RLIMIT_NOFILE for systemd itself */
2041 (void) bump_rlimit_nofile(saved_rlimit_nofile
);
2042 (void) bump_rlimit_memlock(saved_rlimit_memlock
);
2047 static int do_queue_default_job(
2049 const char **ret_error_message
) {
2051 _cleanup_(sd_bus_error_free
) sd_bus_error error
= SD_BUS_ERROR_NULL
;
2052 Job
*default_unit_job
;
2053 Unit
*target
= NULL
;
2056 log_debug("Activating default unit: %s", arg_default_unit
);
2058 r
= manager_load_startable_unit_or_warn(m
, arg_default_unit
, NULL
, &target
);
2060 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET
);
2062 r
= manager_load_startable_unit_or_warn(m
, SPECIAL_RESCUE_TARGET
, NULL
, &target
);
2064 *ret_error_message
= r
== -ERFKILL
? "Rescue target masked"
2065 : "Failed to load rescue target";
2070 assert(target
->load_state
== UNIT_LOADED
);
2072 r
= manager_add_job(m
, JOB_START
, target
, JOB_ISOLATE
, &error
, &default_unit_job
);
2074 log_debug_errno(r
, "Default target could not be isolated, starting instead: %s", bus_error_message(&error
, r
));
2076 sd_bus_error_free(&error
);
2078 r
= manager_add_job(m
, JOB_START
, target
, JOB_REPLACE
, &error
, &default_unit_job
);
2080 *ret_error_message
= "Failed to start default target";
2081 return log_emergency_errno(r
, "Failed to start default target: %s", bus_error_message(&error
, r
));
2085 *ret_error_message
= "Failed to isolate default target";
2086 return log_emergency_errno(r
, "Failed to isolate default target: %s", bus_error_message(&error
, r
));
2089 m
->default_unit_job_id
= default_unit_job
->id
;
2094 static void free_arguments(void) {
2096 /* Frees all arg_* variables, with the exception of arg_serialization */
2097 rlimit_free_all(arg_default_rlimit
);
2099 arg_default_unit
= mfree(arg_default_unit
);
2100 arg_confirm_spawn
= mfree(arg_confirm_spawn
);
2101 arg_default_environment
= strv_free(arg_default_environment
);
2102 arg_syscall_archs
= set_free(arg_syscall_archs
);
2105 static int load_configuration(int argc
, char **argv
, const char **ret_error_message
) {
2108 assert(ret_error_message
);
2110 arg_default_tasks_max
= system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE
, 100U);
2112 r
= parse_config_file();
2114 *ret_error_message
= "Failed to parse config file";
2119 r
= proc_cmdline_parse(parse_proc_cmdline_item
, NULL
, 0);
2121 log_warning_errno(r
, "Failed to parse kernel command line, ignoring: %m");
2124 /* Note that this also parses bits from the kernel command line, including "debug". */
2125 log_parse_environment();
2127 r
= parse_argv(argc
, argv
);
2129 *ret_error_message
= "Failed to parse commandline arguments";
2133 /* Initialize default unit */
2134 if (!arg_default_unit
) {
2135 arg_default_unit
= strdup(SPECIAL_DEFAULT_TARGET
);
2136 if (!arg_default_unit
) {
2137 *ret_error_message
= "Failed to set default unit";
2142 /* Initialize the show status setting if it hasn't been set explicitly yet */
2143 if (arg_show_status
== _SHOW_STATUS_INVALID
)
2144 arg_show_status
= SHOW_STATUS_YES
;
2149 static int safety_checks(void) {
2151 if (getpid_cached() == 1 &&
2152 arg_action
!= ACTION_RUN
)
2153 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
2154 "Unsupported execution mode while PID 1.");
2156 if (getpid_cached() == 1 &&
2158 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
2159 "Can't run --user mode as PID 1.");
2161 if (arg_action
== ACTION_RUN
&&
2163 getpid_cached() != 1)
2164 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
2165 "Can't run system mode unless PID 1.");
2167 if (arg_action
== ACTION_TEST
&&
2169 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
2170 "Don't run test mode as root.");
2173 arg_action
== ACTION_RUN
&&
2175 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP
),
2176 "Trying to run as user instance, but the system has not been booted with systemd.");
2179 arg_action
== ACTION_RUN
&&
2180 !getenv("XDG_RUNTIME_DIR"))
2181 return log_error_errno(SYNTHETIC_ERRNO(EUNATCH
),
2182 "Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
2185 arg_action
== ACTION_RUN
&&
2186 running_in_chroot() > 0)
2187 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP
),
2188 "Cannot be run in a chroot() environment.");
2193 static int initialize_security(
2194 bool *loaded_policy
,
2195 dual_timestamp
*security_start_timestamp
,
2196 dual_timestamp
*security_finish_timestamp
,
2197 const char **ret_error_message
) {
2201 assert(loaded_policy
);
2202 assert(security_start_timestamp
);
2203 assert(security_finish_timestamp
);
2204 assert(ret_error_message
);
2206 dual_timestamp_get(security_start_timestamp
);
2208 r
= mac_selinux_setup(loaded_policy
);
2210 *ret_error_message
= "Failed to load SELinux policy";
2214 r
= mac_smack_setup(loaded_policy
);
2216 *ret_error_message
= "Failed to load SMACK policy";
2222 *ret_error_message
= "Failed to load IMA policy";
2226 dual_timestamp_get(security_finish_timestamp
);
2230 static void test_summary(Manager
*m
) {
2233 printf("-> By units:\n");
2234 manager_dump_units(m
, stdout
, "\t");
2236 printf("-> By jobs:\n");
2237 manager_dump_jobs(m
, stdout
, "\t");
2240 static int collect_fds(FDSet
**ret_fds
, const char **ret_error_message
) {
2244 assert(ret_error_message
);
2246 r
= fdset_new_fill(ret_fds
);
2248 *ret_error_message
= "Failed to allocate fd set";
2249 return log_emergency_errno(r
, "Failed to allocate fd set: %m");
2252 fdset_cloexec(*ret_fds
, true);
2254 if (arg_serialization
)
2255 assert_se(fdset_remove(*ret_fds
, fileno(arg_serialization
)) >= 0);
2260 static void setup_console_terminal(bool skip_setup
) {
2265 /* Become a session leader if we aren't one yet. */
2268 /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2270 (void) release_terminal();
2272 /* Reset the console, but only if this is really init and we are freshly booted */
2273 if (getpid_cached() == 1 && !skip_setup
)
2274 (void) console_setup();
2277 static bool early_skip_setup_check(int argc
, char *argv
[]) {
2278 bool found_deserialize
= false;
2281 /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2282 * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2283 * anyway, even if in that case we also do deserialization. */
2285 for (i
= 1; i
< argc
; i
++) {
2286 if (streq(argv
[i
], "--switched-root"))
2287 return false; /* If we switched root, don't skip the setup. */
2288 else if (streq(argv
[i
], "--deserialize"))
2289 found_deserialize
= true;
2292 return found_deserialize
; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2295 int main(int argc
, char *argv
[]) {
2297 dual_timestamp initrd_timestamp
= DUAL_TIMESTAMP_NULL
, userspace_timestamp
= DUAL_TIMESTAMP_NULL
, kernel_timestamp
= DUAL_TIMESTAMP_NULL
,
2298 security_start_timestamp
= DUAL_TIMESTAMP_NULL
, security_finish_timestamp
= DUAL_TIMESTAMP_NULL
;
2299 struct rlimit saved_rlimit_nofile
= RLIMIT_MAKE_CONST(0), saved_rlimit_memlock
= RLIMIT_MAKE_CONST((rlim_t
) -1);
2300 bool skip_setup
, loaded_policy
= false, queue_default_job
= false, first_boot
= false, reexecute
= false;
2301 char *switch_root_dir
= NULL
, *switch_root_init
= NULL
;
2302 usec_t before_startup
, after_startup
;
2303 static char systemd
[] = "systemd";
2304 char timespan
[FORMAT_TIMESPAN_MAX
];
2305 const char *shutdown_verb
= NULL
, *error_message
= NULL
;
2306 int r
, retval
= EXIT_FAILURE
;
2310 /* SysV compatibility: redirect init → telinit */
2311 redirect_telinit(argc
, argv
);
2313 /* Take timestamps early on */
2314 dual_timestamp_from_monotonic(&kernel_timestamp
, 0);
2315 dual_timestamp_get(&userspace_timestamp
);
2317 /* Figure out whether we need to do initialize the system, or if we already did that because we are
2319 skip_setup
= early_skip_setup_check(argc
, argv
);
2321 /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2322 * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
2323 program_invocation_short_name
= systemd
;
2324 (void) prctl(PR_SET_NAME
, systemd
);
2326 /* Save the original command line */
2330 /* Make sure that if the user says "syslog" we actually log to the journal. */
2331 log_set_upgrade_syslog_to_journal(true);
2333 if (getpid_cached() == 1) {
2334 /* When we run as PID 1 force system mode */
2337 /* Disable the umask logic */
2340 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2341 * activated yet (even though the log socket for it exists). */
2342 log_set_prohibit_ipc(true);
2344 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2345 * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2346 * child process right before execve()'ing the actual binary, at a point in time where socket
2347 * activation stderr/stdout area already set up. */
2348 log_set_always_reopen_console(true);
2350 if (detect_container() <= 0) {
2352 /* Running outside of a container as PID 1 */
2353 log_set_target(LOG_TARGET_KMSG
);
2357 initrd_timestamp
= userspace_timestamp
;
2360 r
= mount_setup_early();
2362 error_message
= "Failed to mount early API filesystems";
2366 r
= initialize_security(
2368 &security_start_timestamp
,
2369 &security_finish_timestamp
,
2375 if (mac_selinux_init() < 0) {
2376 error_message
= "Failed to initialize SELinux policy";
2383 /* Set the default for later on, but don't actually open the logs like this for now. Note that
2384 * if we are transitioning from the initrd there might still be journal fd open, and we
2385 * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2387 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG
);
2390 /* Running inside a container, as PID 1 */
2391 log_set_target(LOG_TARGET_CONSOLE
);
2394 /* For later on, see above... */
2395 log_set_target(LOG_TARGET_JOURNAL
);
2397 /* clear the kernel timestamp,
2398 * because we are in a container */
2399 kernel_timestamp
= DUAL_TIMESTAMP_NULL
;
2402 initialize_coredump(skip_setup
);
2404 r
= fixup_environment();
2406 log_emergency_errno(r
, "Failed to fix up PID 1 environment: %m");
2407 error_message
= "Failed to fix up PID1 environment";
2412 /* Running as user instance */
2414 log_set_target(LOG_TARGET_AUTO
);
2417 /* clear the kernel timestamp,
2418 * because we are not PID 1 */
2419 kernel_timestamp
= DUAL_TIMESTAMP_NULL
;
2423 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2424 * they never log into the console. */
2425 log_show_color(colors_enabled());
2427 r
= make_null_stdio();
2429 log_warning_errno(r
, "Failed to redirect standard streams to /dev/null, ignoring: %m");
2432 /* Mount /proc, /sys and friends, so that /proc/cmdline and
2433 * /proc/$PID/fd is available. */
2434 if (getpid_cached() == 1) {
2436 /* Load the kernel modules early. */
2440 r
= mount_setup(loaded_policy
);
2442 error_message
= "Failed to mount API filesystems";
2447 /* Reset all signal handlers. */
2448 (void) reset_all_signal_handlers();
2449 (void) ignore_signals(SIGNALS_IGNORE
, -1);
2451 r
= load_configuration(argc
, argv
, &error_message
);
2455 r
= safety_checks();
2459 if (IN_SET(arg_action
, ACTION_TEST
, ACTION_HELP
, ACTION_DUMP_CONFIGURATION_ITEMS
, ACTION_DUMP_BUS_PROPERTIES
))
2460 (void) pager_open(arg_pager_flags
);
2462 if (arg_action
!= ACTION_RUN
)
2465 if (arg_action
== ACTION_HELP
) {
2466 retval
= help() < 0 ? EXIT_FAILURE
: EXIT_SUCCESS
;
2468 } else if (arg_action
== ACTION_VERSION
) {
2471 } else if (arg_action
== ACTION_DUMP_CONFIGURATION_ITEMS
) {
2472 unit_dump_config_items(stdout
);
2473 retval
= EXIT_SUCCESS
;
2475 } else if (arg_action
== ACTION_DUMP_BUS_PROPERTIES
) {
2476 dump_bus_properties(stdout
);
2477 retval
= EXIT_SUCCESS
;
2481 assert_se(IN_SET(arg_action
, ACTION_RUN
, ACTION_TEST
));
2483 /* Move out of the way, so that we won't block unmounts */
2484 assert_se(chdir("/") == 0);
2486 if (arg_action
== ACTION_RUN
) {
2488 /* A core pattern might have been specified via the cmdline. */
2489 initialize_core_pattern(skip_setup
);
2491 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2494 /* Remember open file descriptors for later deserialization */
2495 r
= collect_fds(&fds
, &error_message
);
2499 /* Give up any control of the console, but make sure its initialized. */
2500 setup_console_terminal(skip_setup
);
2502 /* Open the logging devices, if possible and necessary */
2506 log_execution_mode(&first_boot
);
2508 r
= initialize_runtime(skip_setup
,
2509 &saved_rlimit_nofile
,
2510 &saved_rlimit_memlock
,
2515 r
= manager_new(arg_system
? UNIT_FILE_SYSTEM
: UNIT_FILE_USER
,
2516 arg_action
== ACTION_TEST
? MANAGER_TEST_FULL
: 0,
2519 log_emergency_errno(r
, "Failed to allocate manager object: %m");
2520 error_message
= "Failed to allocate manager object";
2524 m
->timestamps
[MANAGER_TIMESTAMP_KERNEL
] = kernel_timestamp
;
2525 m
->timestamps
[MANAGER_TIMESTAMP_INITRD
] = initrd_timestamp
;
2526 m
->timestamps
[MANAGER_TIMESTAMP_USERSPACE
] = userspace_timestamp
;
2527 m
->timestamps
[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START
)] = security_start_timestamp
;
2528 m
->timestamps
[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH
)] = security_finish_timestamp
;
2530 set_manager_defaults(m
);
2531 set_manager_settings(m
);
2532 manager_set_first_boot(m
, first_boot
);
2534 /* Remember whether we should queue the default job */
2535 queue_default_job
= !arg_serialization
|| arg_switched_root
;
2537 before_startup
= now(CLOCK_MONOTONIC
);
2539 r
= manager_startup(m
, arg_serialization
, fds
);
2541 error_message
= "Failed to start up manager";
2545 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2546 fds
= fdset_free(fds
);
2547 arg_serialization
= safe_fclose(arg_serialization
);
2549 if (queue_default_job
) {
2550 r
= do_queue_default_job(m
, &error_message
);
2555 after_startup
= now(CLOCK_MONOTONIC
);
2557 log_full(arg_action
== ACTION_TEST
? LOG_INFO
: LOG_DEBUG
,
2558 "Loaded units and determined initial transaction in %s.",
2559 format_timespan(timespan
, sizeof(timespan
), after_startup
- before_startup
, 100 * USEC_PER_MSEC
));
2561 if (arg_action
== ACTION_TEST
) {
2563 retval
= EXIT_SUCCESS
;
2567 (void) invoke_main_loop(m
,
2580 arg_shutdown_watchdog
= m
->shutdown_watchdog
;
2581 m
= manager_free(m
);
2585 mac_selinux_finish();
2588 do_reexecute(argc
, argv
,
2589 &saved_rlimit_nofile
,
2590 &saved_rlimit_memlock
,
2594 &error_message
); /* This only returns if reexecution failed */
2596 arg_serialization
= safe_fclose(arg_serialization
);
2597 fds
= fdset_free(fds
);
2599 #if HAVE_VALGRIND_VALGRIND_H
2600 /* If we are PID 1 and running under valgrind, then let's exit
2601 * here explicitly. valgrind will only generate nice output on
2602 * exit(), not on exec(), hence let's do the former not the
2604 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND
) {
2605 /* Cleanup watchdog_device strings for valgrind. We need them
2606 * in become_shutdown() so normally we cannot free them yet. */
2607 watchdog_free_device();
2608 arg_watchdog_device
= mfree(arg_watchdog_device
);
2613 if (shutdown_verb
) {
2614 r
= become_shutdown(shutdown_verb
, retval
);
2615 log_error_errno(r
, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
2616 error_message
= "Failed to execute shutdown binary";
2619 watchdog_free_device();
2620 arg_watchdog_device
= mfree(arg_watchdog_device
);
2622 if (getpid_cached() == 1) {
2624 manager_status_printf(NULL
, STATUS_TYPE_EMERGENCY
,
2625 ANSI_HIGHLIGHT_RED
"!!!!!!" ANSI_NORMAL
,
2626 "%s.", error_message
);
2627 freeze_or_exit_or_reboot();