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 "terminal-util.h"
77 #include "umask-util.h"
78 #include "user-util.h"
88 ACTION_DUMP_CONFIGURATION_ITEMS
,
89 ACTION_DUMP_BUS_PROPERTIES
,
90 } arg_action
= ACTION_RUN
;
91 static char *arg_default_unit
= NULL
;
92 static bool arg_system
= false;
93 static bool arg_dump_core
= true;
94 static int arg_crash_chvt
= -1;
95 static bool arg_crash_shell
= false;
96 static bool arg_crash_reboot
= false;
97 static char *arg_confirm_spawn
= NULL
;
98 static ShowStatus arg_show_status
= _SHOW_STATUS_INVALID
;
99 static bool arg_switched_root
= false;
100 static bool arg_no_pager
= false;
101 static bool arg_service_watchdogs
= true;
102 static char ***arg_join_controllers
= NULL
;
103 static ExecOutput arg_default_std_output
= EXEC_OUTPUT_JOURNAL
;
104 static ExecOutput arg_default_std_error
= EXEC_OUTPUT_INHERIT
;
105 static usec_t arg_default_restart_usec
= DEFAULT_RESTART_USEC
;
106 static usec_t arg_default_timeout_start_usec
= DEFAULT_TIMEOUT_USEC
;
107 static usec_t arg_default_timeout_stop_usec
= DEFAULT_TIMEOUT_USEC
;
108 static usec_t arg_default_start_limit_interval
= DEFAULT_START_LIMIT_INTERVAL
;
109 static unsigned arg_default_start_limit_burst
= DEFAULT_START_LIMIT_BURST
;
110 static usec_t arg_runtime_watchdog
= 0;
111 static usec_t arg_shutdown_watchdog
= 10 * USEC_PER_MINUTE
;
112 static char *arg_early_core_pattern
= NULL
;
113 static char *arg_watchdog_device
= NULL
;
114 static char **arg_default_environment
= NULL
;
115 static struct rlimit
*arg_default_rlimit
[_RLIMIT_MAX
] = {};
116 static uint64_t arg_capability_bounding_set
= CAP_ALL
;
117 static bool arg_no_new_privs
= false;
118 static nsec_t arg_timer_slack_nsec
= NSEC_INFINITY
;
119 static usec_t arg_default_timer_accuracy_usec
= 1 * USEC_PER_MINUTE
;
120 static Set
* arg_syscall_archs
= NULL
;
121 static FILE* arg_serialization
= NULL
;
122 static bool arg_default_cpu_accounting
= false;
123 static bool arg_default_io_accounting
= false;
124 static bool arg_default_ip_accounting
= false;
125 static bool arg_default_blockio_accounting
= false;
126 static bool arg_default_memory_accounting
= MEMORY_ACCOUNTING_DEFAULT
;
127 static bool arg_default_tasks_accounting
= true;
128 static uint64_t arg_default_tasks_max
= UINT64_MAX
;
129 static sd_id128_t arg_machine_id
= {};
130 static EmergencyAction arg_cad_burst_action
= EMERGENCY_ACTION_REBOOT_FORCE
;
132 _noreturn_
static void freeze_or_reboot(void) {
134 if (arg_crash_reboot
) {
135 log_notice("Rebooting in 10s...");
138 log_notice("Rebooting now...");
139 (void) reboot(RB_AUTOBOOT
);
140 log_emergency_errno(errno
, "Failed to reboot: %m");
143 log_emergency("Freezing execution.");
147 _noreturn_
static void crash(int sig
) {
151 if (getpid_cached() != 1)
152 /* Pass this on immediately, if this is not PID 1 */
154 else if (!arg_dump_core
)
155 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig
));
157 sa
= (struct sigaction
) {
158 .sa_handler
= nop_signal_handler
,
159 .sa_flags
= SA_NOCLDSTOP
|SA_RESTART
,
162 /* We want to wait for the core process, hence let's enable SIGCHLD */
163 (void) sigaction(SIGCHLD
, &sa
, NULL
);
165 pid
= raw_clone(SIGCHLD
);
167 log_emergency_errno(errno
, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig
));
169 /* Enable default signal handler for core dump */
171 sa
= (struct sigaction
) {
172 .sa_handler
= SIG_DFL
,
174 (void) sigaction(sig
, &sa
, NULL
);
176 /* Don't limit the coredump size */
177 (void) setrlimit(RLIMIT_CORE
, &RLIMIT_MAKE_CONST(RLIM_INFINITY
));
179 /* Just to be sure... */
182 /* Raise the signal again */
184 (void) kill(pid
, sig
); /* raise() would kill the parent */
186 assert_not_reached("We shouldn't be here...");
192 /* Order things nicely. */
193 r
= wait_for_terminate(pid
, &status
);
195 log_emergency_errno(r
, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig
));
196 else if (status
.si_code
!= CLD_DUMPED
)
197 log_emergency("Caught <%s>, core dump failed (child "PID_FMT
", code=%s, status=%i/%s).",
198 signal_to_string(sig
),
199 pid
, sigchld_code_to_string(status
.si_code
),
201 strna(status
.si_code
== CLD_EXITED
202 ? exit_status_to_string(status
.si_status
, EXIT_STATUS_MINIMAL
)
203 : signal_to_string(status
.si_status
)));
205 log_emergency("Caught <%s>, dumped core as pid "PID_FMT
".", signal_to_string(sig
), pid
);
209 if (arg_crash_chvt
>= 0)
210 (void) chvt(arg_crash_chvt
);
212 sa
= (struct sigaction
) {
213 .sa_handler
= SIG_IGN
,
214 .sa_flags
= SA_NOCLDSTOP
|SA_NOCLDWAIT
|SA_RESTART
,
217 /* Let the kernel reap children for us */
218 (void) sigaction(SIGCHLD
, &sa
, NULL
);
220 if (arg_crash_shell
) {
221 log_notice("Executing crash shell in 10s...");
224 pid
= raw_clone(SIGCHLD
);
226 log_emergency_errno(errno
, "Failed to fork off crash shell: %m");
229 (void) make_console_stdio();
230 (void) execle("/bin/sh", "/bin/sh", NULL
, environ
);
232 log_emergency_errno(errno
, "execle() failed: %m");
235 log_info("Spawned crash shell as PID "PID_FMT
".", pid
);
236 (void) wait_for_terminate(pid
, NULL
);
243 static void install_crash_handler(void) {
244 static const struct sigaction sa
= {
246 .sa_flags
= SA_NODEFER
, /* So that we can raise the signal again from the signal handler */
250 /* We ignore the return value here, since, we don't mind if we
251 * cannot set up a crash handler */
252 r
= sigaction_many(&sa
, SIGNALS_CRASH_HANDLER
, -1);
254 log_debug_errno(r
, "I had trouble setting up the crash handler, ignoring: %m");
257 static int console_setup(void) {
258 _cleanup_close_
int tty_fd
= -1;
261 tty_fd
= open_terminal("/dev/console", O_WRONLY
|O_NOCTTY
|O_CLOEXEC
);
263 return log_error_errno(tty_fd
, "Failed to open /dev/console: %m");
265 /* We don't want to force text mode. plymouth may be showing
266 * pictures already from initrd. */
267 r
= reset_terminal_fd(tty_fd
, false);
269 return log_error_errno(r
, "Failed to reset /dev/console: %m");
274 static int parse_crash_chvt(const char *value
) {
277 if (safe_atoi(value
, &arg_crash_chvt
) >= 0)
280 b
= parse_boolean(value
);
285 arg_crash_chvt
= 0; /* switch to where kmsg goes */
287 arg_crash_chvt
= -1; /* turn off switching */
292 static int parse_confirm_spawn(const char *value
, char **console
) {
296 r
= value
? parse_boolean(value
) : 1;
302 if (r
> 0) /* on with default tty */
303 s
= strdup("/dev/console");
304 else if (is_path(value
)) /* on with fully qualified path */
306 else /* on with only a tty file name, not a fully qualified path */
307 s
= strjoin("/dev/", value
);
315 static int set_machine_id(const char *m
) {
319 if (sd_id128_from_string(m
, &t
) < 0)
322 if (sd_id128_is_null(t
))
329 static int parse_proc_cmdline_item(const char *key
, const char *value
, void *data
) {
335 if (STR_IN_SET(key
, "systemd.unit", "rd.systemd.unit")) {
337 if (proc_cmdline_value_missing(key
, value
))
340 if (!unit_name_is_valid(value
, UNIT_NAME_PLAIN
|UNIT_NAME_INSTANCE
))
341 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key
, value
);
342 else if (in_initrd() == !!startswith(key
, "rd.")) {
343 if (free_and_strdup(&arg_default_unit
, value
) < 0)
347 } else if (proc_cmdline_key_streq(key
, "systemd.dump_core")) {
349 r
= value
? parse_boolean(value
) : true;
351 log_warning("Failed to parse dump core switch %s. Ignoring.", value
);
355 } else if (proc_cmdline_key_streq(key
, "systemd.early_core_pattern")) {
357 if (proc_cmdline_value_missing(key
, value
))
360 if (path_is_absolute(value
))
361 (void) parse_path_argument_and_warn(value
, false, &arg_early_core_pattern
);
363 log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value
);
365 } else if (proc_cmdline_key_streq(key
, "systemd.crash_chvt")) {
368 arg_crash_chvt
= 0; /* turn on */
369 else if (parse_crash_chvt(value
) < 0)
370 log_warning("Failed to parse crash chvt switch %s. Ignoring.", value
);
372 } else if (proc_cmdline_key_streq(key
, "systemd.crash_shell")) {
374 r
= value
? parse_boolean(value
) : true;
376 log_warning("Failed to parse crash shell switch %s. Ignoring.", value
);
380 } else if (proc_cmdline_key_streq(key
, "systemd.crash_reboot")) {
382 r
= value
? parse_boolean(value
) : true;
384 log_warning("Failed to parse crash reboot switch %s. Ignoring.", value
);
386 arg_crash_reboot
= r
;
388 } else if (proc_cmdline_key_streq(key
, "systemd.confirm_spawn")) {
391 r
= parse_confirm_spawn(value
, &s
);
393 log_warning_errno(r
, "Failed to parse confirm_spawn switch %s. Ignoring.", value
);
395 free(arg_confirm_spawn
);
396 arg_confirm_spawn
= s
;
399 } else if (proc_cmdline_key_streq(key
, "systemd.service_watchdogs")) {
401 r
= value
? parse_boolean(value
) : true;
403 log_warning("Failed to parse service watchdog switch %s. Ignoring.", value
);
405 arg_service_watchdogs
= r
;
407 } else if (proc_cmdline_key_streq(key
, "systemd.show_status")) {
410 r
= parse_show_status(value
, &arg_show_status
);
412 log_warning("Failed to parse show status switch %s. Ignoring.", value
);
414 arg_show_status
= SHOW_STATUS_YES
;
416 } else if (proc_cmdline_key_streq(key
, "systemd.default_standard_output")) {
418 if (proc_cmdline_value_missing(key
, value
))
421 r
= exec_output_from_string(value
);
423 log_warning("Failed to parse default standard output switch %s. Ignoring.", value
);
425 arg_default_std_output
= r
;
427 } else if (proc_cmdline_key_streq(key
, "systemd.default_standard_error")) {
429 if (proc_cmdline_value_missing(key
, value
))
432 r
= exec_output_from_string(value
);
434 log_warning("Failed to parse default standard error switch %s. Ignoring.", value
);
436 arg_default_std_error
= r
;
438 } else if (streq(key
, "systemd.setenv")) {
440 if (proc_cmdline_value_missing(key
, value
))
443 if (env_assignment_is_valid(value
)) {
446 env
= strv_env_set(arg_default_environment
, value
);
450 arg_default_environment
= env
;
452 log_warning("Environment variable name '%s' is not valid. Ignoring.", value
);
454 } else if (proc_cmdline_key_streq(key
, "systemd.machine_id")) {
456 if (proc_cmdline_value_missing(key
, value
))
459 r
= set_machine_id(value
);
461 log_warning("MachineID '%s' is not valid. Ignoring.", value
);
463 } else if (proc_cmdline_key_streq(key
, "systemd.default_timeout_start_sec")) {
465 if (proc_cmdline_value_missing(key
, value
))
468 r
= parse_sec(value
, &arg_default_timeout_start_usec
);
470 log_warning_errno(r
, "Failed to parse default start timeout: %s, ignoring.", value
);
472 if (arg_default_timeout_start_usec
<= 0)
473 arg_default_timeout_start_usec
= USEC_INFINITY
;
475 } else if (proc_cmdline_key_streq(key
, "systemd.watchdog_device")) {
477 if (proc_cmdline_value_missing(key
, value
))
480 (void) parse_path_argument_and_warn(value
, false, &arg_watchdog_device
);
482 } else if (streq(key
, "quiet") && !value
) {
484 if (arg_show_status
== _SHOW_STATUS_INVALID
)
485 arg_show_status
= SHOW_STATUS_AUTO
;
487 } else if (streq(key
, "debug") && !value
) {
489 /* Note that log_parse_environment() handles 'debug'
490 * too, and sets the log level to LOG_DEBUG. */
492 if (detect_container() > 0)
493 log_set_target(LOG_TARGET_CONSOLE
);
498 /* SysV compatibility */
499 target
= runlevel_to_target(key
);
501 return free_and_strdup(&arg_default_unit
, target
);
507 #define DEFINE_SETTER(name, func, descr) \
508 static int name(const char *unit, \
509 const char *filename, \
511 const char *section, \
512 unsigned section_line, \
513 const char *lvalue, \
515 const char *rvalue, \
527 log_syntax(unit, LOG_ERR, filename, line, r, \
528 "Invalid " descr "'%s': %m", \
534 DEFINE_SETTER(config_parse_level2
, log_set_max_level_from_string
, "log level");
535 DEFINE_SETTER(config_parse_target
, log_set_target_from_string
, "target");
536 DEFINE_SETTER(config_parse_color
, log_show_color_from_string
, "color" );
537 DEFINE_SETTER(config_parse_location
, log_show_location_from_string
, "location");
539 static int config_parse_cpu_affinity2(
541 const char *filename
,
544 unsigned section_line
,
551 _cleanup_cpu_free_ cpu_set_t
*c
= NULL
;
554 ncpus
= parse_cpu_set_and_warn(rvalue
, &c
, unit
, filename
, line
, lvalue
);
558 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus
), c
) < 0)
559 log_warning_errno(errno
, "Failed to set CPU affinity: %m");
564 static int config_parse_show_status(
566 const char *filename
,
569 unsigned section_line
,
577 ShowStatus
*b
= data
;
584 k
= parse_show_status(rvalue
, b
);
586 log_syntax(unit
, LOG_ERR
, filename
, line
, k
, "Failed to parse show status setting, ignoring: %s", rvalue
);
593 static int config_parse_output_restricted(
595 const char *filename
,
598 unsigned section_line
,
605 ExecOutput t
, *eo
= data
;
612 t
= exec_output_from_string(rvalue
);
614 log_syntax(unit
, LOG_ERR
, filename
, line
, 0, "Failed to parse output type, ignoring: %s", rvalue
);
618 if (IN_SET(t
, EXEC_OUTPUT_SOCKET
, EXEC_OUTPUT_NAMED_FD
, EXEC_OUTPUT_FILE
, EXEC_OUTPUT_FILE_APPEND
)) {
619 log_syntax(unit
, LOG_ERR
, filename
, line
, 0, "Standard output types socket, fd:, file:, append: are not supported as defaults, ignoring: %s", rvalue
);
627 static int config_parse_crash_chvt(
629 const char *filename
,
632 unsigned section_line
,
645 r
= parse_crash_chvt(rvalue
);
647 log_syntax(unit
, LOG_ERR
, filename
, line
, r
, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue
);
654 static int parse_config_file(void) {
656 const ConfigTableItem items
[] = {
657 { "Manager", "LogLevel", config_parse_level2
, 0, NULL
},
658 { "Manager", "LogTarget", config_parse_target
, 0, NULL
},
659 { "Manager", "LogColor", config_parse_color
, 0, NULL
},
660 { "Manager", "LogLocation", config_parse_location
, 0, NULL
},
661 { "Manager", "DumpCore", config_parse_bool
, 0, &arg_dump_core
},
662 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt
, 0, NULL
},
663 { "Manager", "CrashChangeVT", config_parse_crash_chvt
, 0, NULL
},
664 { "Manager", "CrashShell", config_parse_bool
, 0, &arg_crash_shell
},
665 { "Manager", "CrashReboot", config_parse_bool
, 0, &arg_crash_reboot
},
666 { "Manager", "ShowStatus", config_parse_show_status
, 0, &arg_show_status
},
667 { "Manager", "CPUAffinity", config_parse_cpu_affinity2
, 0, NULL
},
668 { "Manager", "JoinControllers", config_parse_join_controllers
, 0, &arg_join_controllers
},
669 { "Manager", "RuntimeWatchdogSec", config_parse_sec
, 0, &arg_runtime_watchdog
},
670 { "Manager", "ShutdownWatchdogSec", config_parse_sec
, 0, &arg_shutdown_watchdog
},
671 { "Manager", "WatchdogDevice", config_parse_path
, 0, &arg_watchdog_device
},
672 { "Manager", "CapabilityBoundingSet", config_parse_capability_set
, 0, &arg_capability_bounding_set
},
673 { "Manager", "NoNewPrivileges", config_parse_bool
, 0, &arg_no_new_privs
},
675 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs
, 0, &arg_syscall_archs
},
677 { "Manager", "TimerSlackNSec", config_parse_nsec
, 0, &arg_timer_slack_nsec
},
678 { "Manager", "DefaultTimerAccuracySec", config_parse_sec
, 0, &arg_default_timer_accuracy_usec
},
679 { "Manager", "DefaultStandardOutput", config_parse_output_restricted
,0, &arg_default_std_output
},
680 { "Manager", "DefaultStandardError", config_parse_output_restricted
,0, &arg_default_std_error
},
681 { "Manager", "DefaultTimeoutStartSec", config_parse_sec
, 0, &arg_default_timeout_start_usec
},
682 { "Manager", "DefaultTimeoutStopSec", config_parse_sec
, 0, &arg_default_timeout_stop_usec
},
683 { "Manager", "DefaultRestartSec", config_parse_sec
, 0, &arg_default_restart_usec
},
684 { "Manager", "DefaultStartLimitInterval", config_parse_sec
, 0, &arg_default_start_limit_interval
}, /* obsolete alias */
685 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec
, 0, &arg_default_start_limit_interval
},
686 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned
, 0, &arg_default_start_limit_burst
},
687 { "Manager", "DefaultEnvironment", config_parse_environ
, 0, &arg_default_environment
},
688 { "Manager", "DefaultLimitCPU", config_parse_rlimit
, RLIMIT_CPU
, arg_default_rlimit
},
689 { "Manager", "DefaultLimitFSIZE", config_parse_rlimit
, RLIMIT_FSIZE
, arg_default_rlimit
},
690 { "Manager", "DefaultLimitDATA", config_parse_rlimit
, RLIMIT_DATA
, arg_default_rlimit
},
691 { "Manager", "DefaultLimitSTACK", config_parse_rlimit
, RLIMIT_STACK
, arg_default_rlimit
},
692 { "Manager", "DefaultLimitCORE", config_parse_rlimit
, RLIMIT_CORE
, arg_default_rlimit
},
693 { "Manager", "DefaultLimitRSS", config_parse_rlimit
, RLIMIT_RSS
, arg_default_rlimit
},
694 { "Manager", "DefaultLimitNOFILE", config_parse_rlimit
, RLIMIT_NOFILE
, arg_default_rlimit
},
695 { "Manager", "DefaultLimitAS", config_parse_rlimit
, RLIMIT_AS
, arg_default_rlimit
},
696 { "Manager", "DefaultLimitNPROC", config_parse_rlimit
, RLIMIT_NPROC
, arg_default_rlimit
},
697 { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit
, RLIMIT_MEMLOCK
, arg_default_rlimit
},
698 { "Manager", "DefaultLimitLOCKS", config_parse_rlimit
, RLIMIT_LOCKS
, arg_default_rlimit
},
699 { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit
, RLIMIT_SIGPENDING
, arg_default_rlimit
},
700 { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit
, RLIMIT_MSGQUEUE
, arg_default_rlimit
},
701 { "Manager", "DefaultLimitNICE", config_parse_rlimit
, RLIMIT_NICE
, arg_default_rlimit
},
702 { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit
, RLIMIT_RTPRIO
, arg_default_rlimit
},
703 { "Manager", "DefaultLimitRTTIME", config_parse_rlimit
, RLIMIT_RTTIME
, arg_default_rlimit
},
704 { "Manager", "DefaultCPUAccounting", config_parse_bool
, 0, &arg_default_cpu_accounting
},
705 { "Manager", "DefaultIOAccounting", config_parse_bool
, 0, &arg_default_io_accounting
},
706 { "Manager", "DefaultIPAccounting", config_parse_bool
, 0, &arg_default_ip_accounting
},
707 { "Manager", "DefaultBlockIOAccounting", config_parse_bool
, 0, &arg_default_blockio_accounting
},
708 { "Manager", "DefaultMemoryAccounting", config_parse_bool
, 0, &arg_default_memory_accounting
},
709 { "Manager", "DefaultTasksAccounting", config_parse_bool
, 0, &arg_default_tasks_accounting
},
710 { "Manager", "DefaultTasksMax", config_parse_tasks_max
, 0, &arg_default_tasks_max
},
711 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action
, 0, &arg_cad_burst_action
},
715 const char *fn
, *conf_dirs_nulstr
;
718 PKGSYSCONFDIR
"/system.conf" :
719 PKGSYSCONFDIR
"/user.conf";
721 conf_dirs_nulstr
= arg_system
?
722 CONF_PATHS_NULSTR("systemd/system.conf.d") :
723 CONF_PATHS_NULSTR("systemd/user.conf.d");
725 (void) config_parse_many_nulstr(fn
, conf_dirs_nulstr
, "Manager\0", config_item_table_lookup
, items
, CONFIG_PARSE_WARN
, NULL
);
727 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
728 * like everywhere else. */
729 if (arg_default_timeout_start_usec
<= 0)
730 arg_default_timeout_start_usec
= USEC_INFINITY
;
731 if (arg_default_timeout_stop_usec
<= 0)
732 arg_default_timeout_stop_usec
= USEC_INFINITY
;
737 static void set_manager_defaults(Manager
*m
) {
741 /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
742 * affect the manager itself, but are just what newly allocated units will have set if they haven't set
743 * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
745 m
->default_timer_accuracy_usec
= arg_default_timer_accuracy_usec
;
746 m
->default_std_output
= arg_default_std_output
;
747 m
->default_std_error
= arg_default_std_error
;
748 m
->default_timeout_start_usec
= arg_default_timeout_start_usec
;
749 m
->default_timeout_stop_usec
= arg_default_timeout_stop_usec
;
750 m
->default_restart_usec
= arg_default_restart_usec
;
751 m
->default_start_limit_interval
= arg_default_start_limit_interval
;
752 m
->default_start_limit_burst
= arg_default_start_limit_burst
;
753 m
->default_cpu_accounting
= arg_default_cpu_accounting
;
754 m
->default_io_accounting
= arg_default_io_accounting
;
755 m
->default_ip_accounting
= arg_default_ip_accounting
;
756 m
->default_blockio_accounting
= arg_default_blockio_accounting
;
757 m
->default_memory_accounting
= arg_default_memory_accounting
;
758 m
->default_tasks_accounting
= arg_default_tasks_accounting
;
759 m
->default_tasks_max
= arg_default_tasks_max
;
761 manager_set_default_rlimits(m
, arg_default_rlimit
);
762 manager_environment_add(m
, NULL
, arg_default_environment
);
765 static void set_manager_settings(Manager
*m
) {
769 /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
770 * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
772 m
->confirm_spawn
= arg_confirm_spawn
;
773 m
->service_watchdogs
= arg_service_watchdogs
;
774 m
->runtime_watchdog
= arg_runtime_watchdog
;
775 m
->shutdown_watchdog
= arg_shutdown_watchdog
;
776 m
->cad_burst_action
= arg_cad_burst_action
;
778 manager_set_show_status(m
, arg_show_status
);
781 static int parse_argv(int argc
, char *argv
[]) {
783 ARG_LOG_LEVEL
= 0x100,
793 ARG_DUMP_CONFIGURATION_ITEMS
,
794 ARG_DUMP_BUS_PROPERTIES
,
803 ARG_DEFAULT_STD_OUTPUT
,
804 ARG_DEFAULT_STD_ERROR
,
806 ARG_SERVICE_WATCHDOGS
,
809 static const struct option options
[] = {
810 { "log-level", required_argument
, NULL
, ARG_LOG_LEVEL
},
811 { "log-target", required_argument
, NULL
, ARG_LOG_TARGET
},
812 { "log-color", optional_argument
, NULL
, ARG_LOG_COLOR
},
813 { "log-location", optional_argument
, NULL
, ARG_LOG_LOCATION
},
814 { "unit", required_argument
, NULL
, ARG_UNIT
},
815 { "system", no_argument
, NULL
, ARG_SYSTEM
},
816 { "user", no_argument
, NULL
, ARG_USER
},
817 { "test", no_argument
, NULL
, ARG_TEST
},
818 { "no-pager", no_argument
, NULL
, ARG_NO_PAGER
},
819 { "help", no_argument
, NULL
, 'h' },
820 { "version", no_argument
, NULL
, ARG_VERSION
},
821 { "dump-configuration-items", no_argument
, NULL
, ARG_DUMP_CONFIGURATION_ITEMS
},
822 { "dump-bus-properties", no_argument
, NULL
, ARG_DUMP_BUS_PROPERTIES
},
823 { "dump-core", optional_argument
, NULL
, ARG_DUMP_CORE
},
824 { "crash-chvt", required_argument
, NULL
, ARG_CRASH_CHVT
},
825 { "crash-shell", optional_argument
, NULL
, ARG_CRASH_SHELL
},
826 { "crash-reboot", optional_argument
, NULL
, ARG_CRASH_REBOOT
},
827 { "confirm-spawn", optional_argument
, NULL
, ARG_CONFIRM_SPAWN
},
828 { "show-status", optional_argument
, NULL
, ARG_SHOW_STATUS
},
829 { "deserialize", required_argument
, NULL
, ARG_DESERIALIZE
},
830 { "switched-root", no_argument
, NULL
, ARG_SWITCHED_ROOT
},
831 { "default-standard-output", required_argument
, NULL
, ARG_DEFAULT_STD_OUTPUT
, },
832 { "default-standard-error", required_argument
, NULL
, ARG_DEFAULT_STD_ERROR
, },
833 { "machine-id", required_argument
, NULL
, ARG_MACHINE_ID
},
834 { "service-watchdogs", required_argument
, NULL
, ARG_SERVICE_WATCHDOGS
},
843 if (getpid_cached() == 1)
846 while ((c
= getopt_long(argc
, argv
, "hDbsz:", options
, NULL
)) >= 0)
851 r
= log_set_max_level_from_string(optarg
);
853 return log_error_errno(r
, "Failed to parse log level \"%s\": %m", optarg
);
858 r
= log_set_target_from_string(optarg
);
860 return log_error_errno(r
, "Failed to parse log target \"%s\": %m", optarg
);
867 r
= log_show_color_from_string(optarg
);
869 return log_error_errno(r
, "Failed to parse log color setting \"%s\": %m",
872 log_show_color(true);
876 case ARG_LOG_LOCATION
:
878 r
= log_show_location_from_string(optarg
);
880 return log_error_errno(r
, "Failed to parse log location setting \"%s\": %m",
883 log_show_location(true);
887 case ARG_DEFAULT_STD_OUTPUT
:
888 r
= exec_output_from_string(optarg
);
890 return log_error_errno(r
, "Failed to parse default standard output setting \"%s\": %m",
892 arg_default_std_output
= r
;
895 case ARG_DEFAULT_STD_ERROR
:
896 r
= exec_output_from_string(optarg
);
898 return log_error_errno(r
, "Failed to parse default standard error output setting \"%s\": %m",
900 arg_default_std_error
= r
;
904 r
= free_and_strdup(&arg_default_unit
, optarg
);
906 return log_error_errno(r
, "Failed to set default unit \"%s\": %m", optarg
);
919 arg_action
= ACTION_TEST
;
927 arg_action
= ACTION_VERSION
;
930 case ARG_DUMP_CONFIGURATION_ITEMS
:
931 arg_action
= ACTION_DUMP_CONFIGURATION_ITEMS
;
934 case ARG_DUMP_BUS_PROPERTIES
:
935 arg_action
= ACTION_DUMP_BUS_PROPERTIES
;
940 arg_dump_core
= true;
942 r
= parse_boolean(optarg
);
944 return log_error_errno(r
, "Failed to parse dump core boolean: \"%s\": %m",
951 r
= parse_crash_chvt(optarg
);
953 return log_error_errno(r
, "Failed to parse crash virtual terminal index: \"%s\": %m",
957 case ARG_CRASH_SHELL
:
959 arg_crash_shell
= true;
961 r
= parse_boolean(optarg
);
963 return log_error_errno(r
, "Failed to parse crash shell boolean: \"%s\": %m",
969 case ARG_CRASH_REBOOT
:
971 arg_crash_reboot
= true;
973 r
= parse_boolean(optarg
);
975 return log_error_errno(r
, "Failed to parse crash shell boolean: \"%s\": %m",
977 arg_crash_reboot
= r
;
981 case ARG_CONFIRM_SPAWN
:
982 arg_confirm_spawn
= mfree(arg_confirm_spawn
);
984 r
= parse_confirm_spawn(optarg
, &arg_confirm_spawn
);
986 return log_error_errno(r
, "Failed to parse confirm spawn option: \"%s\": %m",
990 case ARG_SERVICE_WATCHDOGS
:
991 r
= parse_boolean(optarg
);
993 return log_error_errno(r
, "Failed to parse service watchdogs boolean: \"%s\": %m",
995 arg_service_watchdogs
= r
;
998 case ARG_SHOW_STATUS
:
1000 r
= parse_show_status(optarg
, &arg_show_status
);
1002 return log_error_errno(r
, "Failed to parse show status boolean: \"%s\": %m",
1005 arg_show_status
= SHOW_STATUS_YES
;
1008 case ARG_DESERIALIZE
: {
1012 r
= safe_atoi(optarg
, &fd
);
1014 log_error_errno(r
, "Failed to parse deserialize option \"%s\": %m", optarg
);
1016 log_error("Invalid deserialize fd: %d", fd
);
1020 (void) fd_cloexec(fd
, true);
1022 f
= fdopen(fd
, "r");
1024 return log_error_errno(errno
, "Failed to open serialization fd %d: %m", fd
);
1026 safe_fclose(arg_serialization
);
1027 arg_serialization
= f
;
1032 case ARG_SWITCHED_ROOT
:
1033 arg_switched_root
= true;
1036 case ARG_MACHINE_ID
:
1037 r
= set_machine_id(optarg
);
1039 return log_error_errno(r
, "MachineID '%s' is not valid: %m", optarg
);
1043 arg_action
= ACTION_HELP
;
1047 log_set_max_level(LOG_DEBUG
);
1053 /* Just to eat away the sysvinit kernel
1054 * cmdline args without getopt() error
1055 * messages that we'll parse in
1056 * parse_proc_cmdline_word() or ignore. */
1059 if (getpid_cached() != 1)
1065 assert_not_reached("Unhandled option code.");
1068 if (optind
< argc
&& getpid_cached() != 1) {
1069 /* Hmm, when we aren't run as init system
1070 * let's complain about excess arguments */
1072 log_error("Excess arguments.");
1079 static int help(void) {
1080 _cleanup_free_
char *link
= NULL
;
1083 r
= terminal_urlify_man("systemd", "1", &link
);
1087 printf("%s [OPTIONS...]\n\n"
1088 "Starts up and maintains the system or user services.\n\n"
1089 " -h --help Show this help\n"
1090 " --version Show version\n"
1091 " --test Determine startup sequence, dump it and exit\n"
1092 " --no-pager Do not pipe output into a pager\n"
1093 " --dump-configuration-items Dump understood unit configuration items\n"
1094 " --dump-bus-properties Dump exposed bus properties\n"
1095 " --unit=UNIT Set default unit\n"
1096 " --system Run a system instance, even if PID != 1\n"
1097 " --user Run a user instance\n"
1098 " --dump-core[=BOOL] Dump core on crash\n"
1099 " --crash-vt=NR Change to specified VT on crash\n"
1100 " --crash-reboot[=BOOL] Reboot on crash\n"
1101 " --crash-shell[=BOOL] Run shell on crash\n"
1102 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1103 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1104 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1105 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1106 " --log-color[=BOOL] Highlight important log messages\n"
1107 " --log-location[=BOOL] Include code location in log messages\n"
1108 " --default-standard-output= Set default standard output for services\n"
1109 " --default-standard-error= Set default standard error output for services\n"
1110 "\nSee the %s for details.\n"
1111 , program_invocation_short_name
1118 static int prepare_reexecute(
1122 bool switching_root
) {
1124 _cleanup_fdset_free_ FDSet
*fds
= NULL
;
1125 _cleanup_fclose_
FILE *f
= NULL
;
1132 r
= manager_open_serialization(m
, &f
);
1134 return log_error_errno(r
, "Failed to create serialization file: %m");
1136 /* Make sure nothing is really destructed when we shut down */
1138 bus_manager_send_reloading(m
, true);
1144 r
= manager_serialize(m
, f
, fds
, switching_root
);
1146 return log_error_errno(r
, "Failed to serialize state: %m");
1148 if (fseeko(f
, 0, SEEK_SET
) == (off_t
) -1)
1149 return log_error_errno(errno
, "Failed to rewind serialization fd: %m");
1151 r
= fd_cloexec(fileno(f
), false);
1153 return log_error_errno(r
, "Failed to disable O_CLOEXEC for serialization: %m");
1155 r
= fdset_cloexec(fds
, false);
1157 return log_error_errno(r
, "Failed to disable O_CLOEXEC for serialization fds: %m");
1159 *ret_f
= TAKE_PTR(f
);
1160 *ret_fds
= TAKE_PTR(fds
);
1165 static int bump_rlimit_nofile(struct rlimit
*saved_rlimit
) {
1168 assert(saved_rlimit
);
1170 /* Save the original RLIMIT_NOFILE so that we can reset it later when transitioning from the initrd to the main
1171 * systemd or suchlike. */
1172 if (getrlimit(RLIMIT_NOFILE
, saved_rlimit
) < 0)
1173 return log_warning_errno(errno
, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1175 /* Get the underlying absolute limit the kernel enforces */
1176 nr
= read_nr_open();
1178 /* Make sure forked processes get limits based on the original kernel setting */
1179 if (!arg_default_rlimit
[RLIMIT_NOFILE
]) {
1182 rl
= newdup(struct rlimit
, saved_rlimit
, 1);
1186 /* Bump the hard limit for system services to a substantially higher value. The default hard limit
1187 * current kernels set is pretty low (4K), mostly for historical reasons. According to kernel
1188 * developers, the fd handling in recent kernels has been optimized substantially enough, so that we
1189 * can bump the limit now, without paying too high a price in memory or performance. Note however that
1190 * we only bump the hard limit, not the soft limit. That's because select() works the way it works, and
1191 * chokes on fds >= 1024. If we'd bump the soft limit globally, it might accidentally happen to
1192 * unexpecting programs that they get fds higher than what they can process using select(). By only
1193 * bumping the hard limit but leaving the low limit as it is we avoid this pitfall: programs that are
1194 * written by folks aware of the select() problem in mind (and thus use poll()/epoll instead of
1195 * select(), the way everybody should) can explicitly opt into high fds by bumping their soft limit
1196 * beyond 1024, to the hard limit we pass. */
1198 rl
->rlim_max
= MIN((rlim_t
) nr
, MAX(rl
->rlim_max
, (rlim_t
) HIGH_RLIMIT_NOFILE
));
1200 arg_default_rlimit
[RLIMIT_NOFILE
] = rl
;
1203 /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for
1204 * both hard and soft. */
1205 r
= setrlimit_closest(RLIMIT_NOFILE
, &RLIMIT_MAKE_CONST(nr
));
1207 return log_warning_errno(r
, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1212 static int bump_rlimit_memlock(struct rlimit
*saved_rlimit
) {
1215 assert(saved_rlimit
);
1216 assert(getuid() == 0);
1218 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even though we have CAP_IPC_LOCK which
1219 * should normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's
1220 * bump the value high enough for the root user. */
1222 if (getrlimit(RLIMIT_MEMLOCK
, saved_rlimit
) < 0)
1223 return log_warning_errno(errno
, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1225 r
= setrlimit_closest(RLIMIT_MEMLOCK
, &RLIMIT_MAKE_CONST(1024ULL*1024ULL*64ULL));
1227 return log_warning_errno(r
, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1232 static void test_usr(void) {
1234 /* Check that /usr is not a separate fs */
1236 if (dir_is_empty("/usr") <= 0)
1239 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1240 "Some things will probably break (sometimes even silently) in mysterious ways. "
1241 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1244 static int enforce_syscall_archs(Set
*archs
) {
1248 if (!is_seccomp_available())
1251 r
= seccomp_restrict_archs(arg_syscall_archs
);
1253 return log_error_errno(r
, "Failed to enforce system call architecture restrication: %m");
1258 static int status_welcome(void) {
1259 _cleanup_free_
char *pretty_name
= NULL
, *ansi_color
= NULL
;
1262 if (IN_SET(arg_show_status
, SHOW_STATUS_NO
, SHOW_STATUS_AUTO
))
1265 r
= parse_os_release(NULL
,
1266 "PRETTY_NAME", &pretty_name
,
1267 "ANSI_COLOR", &ansi_color
,
1270 log_full_errno(r
== -ENOENT
? LOG_DEBUG
: LOG_WARNING
, r
,
1271 "Failed to read os-release file, ignoring: %m");
1273 if (log_get_show_color())
1274 return status_printf(NULL
, false, false,
1275 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1276 isempty(ansi_color
) ? "1" : ansi_color
,
1277 isempty(pretty_name
) ? "Linux" : pretty_name
);
1279 return status_printf(NULL
, false, false,
1280 "\nWelcome to %s!\n",
1281 isempty(pretty_name
) ? "Linux" : pretty_name
);
1284 static int write_container_id(void) {
1288 c
= getenv("container");
1292 RUN_WITH_UMASK(0022)
1293 r
= write_string_file("/run/systemd/container", c
, WRITE_STRING_FILE_CREATE
);
1295 return log_warning_errno(r
, "Failed to write /run/systemd/container, ignoring: %m");
1300 static int bump_unix_max_dgram_qlen(void) {
1301 _cleanup_free_
char *qlen
= NULL
;
1305 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1306 * really really early during boot, so that it is actually applied to all our sockets, including the
1307 * $NOTIFY_SOCKET one. */
1309 r
= read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen
);
1311 return log_warning_errno(r
, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1313 r
= safe_atolu(qlen
, &v
);
1315 return log_warning_errno(r
, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen
);
1317 if (v
>= DEFAULT_UNIX_MAX_DGRAM_QLEN
)
1320 r
= write_string_filef("/proc/sys/net/unix/max_dgram_qlen", 0, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN
);
1322 return log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
,
1323 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1328 static int fixup_environment(void) {
1329 _cleanup_free_
char *term
= NULL
;
1333 /* Only fix up the environment when we are started as PID 1 */
1334 if (getpid_cached() != 1)
1337 /* We expect the environment to be set correctly if run inside a container. */
1338 if (detect_container() > 0)
1341 /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1342 * device used by the console. We try to make a better guess here since some consoles might not have support
1343 * for color mode for example.
1345 * However if TERM was configured through the kernel command line then leave it alone. */
1346 r
= proc_cmdline_get_key("TERM", 0, &term
);
1350 t
= term
?: default_term_for_tty("/dev/console");
1352 if (setenv("TERM", t
, 1) < 0)
1358 static void redirect_telinit(int argc
, char *argv
[]) {
1360 /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1362 #if HAVE_SYSV_COMPAT
1363 if (getpid_cached() == 1)
1366 if (!strstr(program_invocation_short_name
, "init"))
1369 execv(SYSTEMCTL_BINARY_PATH
, argv
);
1370 log_error_errno(errno
, "Failed to exec " SYSTEMCTL_BINARY_PATH
": %m");
1375 static int become_shutdown(
1376 const char *shutdown_verb
,
1379 char log_level
[DECIMAL_STR_MAX(int) + 1],
1380 exit_code
[DECIMAL_STR_MAX(uint8_t) + 1],
1381 timeout
[DECIMAL_STR_MAX(usec_t
) + 1];
1383 const char* command_line
[13] = {
1384 SYSTEMD_SHUTDOWN_BINARY_PATH
,
1386 "--timeout", timeout
,
1387 "--log-level", log_level
,
1391 _cleanup_strv_free_
char **env_block
= NULL
;
1395 assert(shutdown_verb
);
1396 assert(!command_line
[pos
]);
1397 env_block
= strv_copy(environ
);
1399 xsprintf(log_level
, "%d", log_get_max_level());
1400 xsprintf(timeout
, "%" PRI_USEC
"us", arg_default_timeout_stop_usec
);
1402 switch (log_get_target()) {
1404 case LOG_TARGET_KMSG
:
1405 case LOG_TARGET_JOURNAL_OR_KMSG
:
1406 case LOG_TARGET_SYSLOG_OR_KMSG
:
1407 command_line
[pos
++] = "kmsg";
1410 case LOG_TARGET_NULL
:
1411 command_line
[pos
++] = "null";
1414 case LOG_TARGET_CONSOLE
:
1416 command_line
[pos
++] = "console";
1420 if (log_get_show_color())
1421 command_line
[pos
++] = "--log-color";
1423 if (log_get_show_location())
1424 command_line
[pos
++] = "--log-location";
1426 if (streq(shutdown_verb
, "exit")) {
1427 command_line
[pos
++] = "--exit-code";
1428 command_line
[pos
++] = exit_code
;
1429 xsprintf(exit_code
, "%d", retval
);
1432 assert(pos
< ELEMENTSOF(command_line
));
1434 if (streq(shutdown_verb
, "reboot") &&
1435 arg_shutdown_watchdog
> 0 &&
1436 arg_shutdown_watchdog
!= USEC_INFINITY
) {
1440 /* If we reboot let's set the shutdown
1441 * watchdog and tell the shutdown binary to
1442 * repeatedly ping it */
1443 r
= watchdog_set_timeout(&arg_shutdown_watchdog
);
1444 watchdog_close(r
< 0);
1446 /* Tell the binary how often to ping, ignore failure */
1447 if (asprintf(&e
, "WATCHDOG_USEC="USEC_FMT
, arg_shutdown_watchdog
) > 0)
1448 (void) strv_consume(&env_block
, e
);
1450 if (arg_watchdog_device
&&
1451 asprintf(&e
, "WATCHDOG_DEVICE=%s", arg_watchdog_device
) > 0)
1452 (void) strv_consume(&env_block
, e
);
1454 watchdog_close(true);
1456 /* Avoid the creation of new processes forked by the
1457 * kernel; at this point, we will not listen to the
1459 if (detect_container() <= 0)
1460 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER
);
1462 execve(SYSTEMD_SHUTDOWN_BINARY_PATH
, (char **) command_line
, env_block
);
1466 static void initialize_clock(void) {
1469 if (clock_is_localtime(NULL
) > 0) {
1473 * The very first call of settimeofday() also does a time warp in the kernel.
1475 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1476 * of maintaining the RTC and do all adjustments. This matches the behavior of Windows, which leaves
1477 * the RTC alone if the registry tells that the RTC runs in UTC.
1479 r
= clock_set_timezone(&min
);
1481 log_error_errno(r
, "Failed to apply local time delta, ignoring: %m");
1483 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min
);
1485 } else if (!in_initrd()) {
1487 * Do a dummy very first call to seal the kernel's time warp magic.
1489 * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1490 * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1491 * until we reach the real system.
1493 * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1494 * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1495 * be treated as UTC that way.
1497 (void) clock_reset_timewarp();
1500 r
= clock_apply_epoch();
1502 log_error_errno(r
, "Current system time is before build time, but cannot correct: %m");
1504 log_info("System time before build time, advancing clock.");
1507 static void initialize_coredump(bool skip_setup
) {
1509 if (getpid_cached() != 1)
1512 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1513 * will process core dumps for system services by default. */
1514 if (setrlimit(RLIMIT_CORE
, &RLIMIT_MAKE_CONST(RLIM_INFINITY
)) < 0)
1515 log_warning_errno(errno
, "Failed to set RLIMIT_CORE: %m");
1517 /* But at the same time, turn off the core_pattern logic by default, so that no
1518 * coredumps are stored until the systemd-coredump tool is enabled via
1519 * sysctl. However it can be changed via the kernel command line later so core
1520 * dumps can still be generated during early startup and in initramfs. */
1522 disable_coredumps();
1526 static void initialize_core_pattern(bool skip_setup
) {
1529 if (skip_setup
|| !arg_early_core_pattern
)
1532 if (getpid_cached() != 1)
1535 r
= write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern
, 0);
1537 log_warning_errno(r
, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", arg_early_core_pattern
);
1540 static void do_reexecute(
1543 const struct rlimit
*saved_rlimit_nofile
,
1544 const struct rlimit
*saved_rlimit_memlock
,
1546 const char *switch_root_dir
,
1547 const char *switch_root_init
,
1548 const char **ret_error_message
) {
1550 unsigned i
, j
, args_size
;
1554 assert(saved_rlimit_nofile
);
1555 assert(saved_rlimit_memlock
);
1556 assert(ret_error_message
);
1558 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1560 watchdog_close(true);
1562 /* Reset the RLIMIT_NOFILE to the kernel default, so that the new systemd can pass the kernel default to its
1563 * child processes */
1565 if (saved_rlimit_nofile
->rlim_cur
> 0)
1566 (void) setrlimit(RLIMIT_NOFILE
, saved_rlimit_nofile
);
1567 if (saved_rlimit_memlock
->rlim_cur
!= (rlim_t
) -1)
1568 (void) setrlimit(RLIMIT_MEMLOCK
, saved_rlimit_memlock
);
1570 if (switch_root_dir
) {
1571 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1572 * SIGCHLD for them after deserializing. */
1573 broadcast_signal(SIGTERM
, false, true, arg_default_timeout_stop_usec
);
1575 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1576 r
= switch_root(switch_root_dir
, "/mnt", true, MS_MOVE
);
1578 log_error_errno(r
, "Failed to switch root, trying to continue: %m");
1581 args_size
= MAX(6, argc
+1);
1582 args
= newa(const char*, args_size
);
1584 if (!switch_root_init
) {
1585 char sfd
[DECIMAL_STR_MAX(int) + 1];
1587 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1588 * the user didn't specify an explicit init to spawn. */
1590 assert(arg_serialization
);
1593 xsprintf(sfd
, "%i", fileno(arg_serialization
));
1596 args
[i
++] = SYSTEMD_BINARY_PATH
;
1597 if (switch_root_dir
)
1598 args
[i
++] = "--switched-root";
1599 args
[i
++] = arg_system
? "--system" : "--user";
1600 args
[i
++] = "--deserialize";
1604 assert(i
<= args_size
);
1607 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do this is on
1608 * its own on exec(), but it will do it on exit(). Hence, to ensure we get a summary here, fork() off
1609 * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1610 * before proceeding into the exec().
1612 valgrind_summary_hack();
1614 (void) execv(args
[0], (char* const*) args
);
1615 log_debug_errno(errno
, "Failed to execute our own binary, trying fallback: %m");
1618 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1619 * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1620 * doesn't matter.) */
1622 arg_serialization
= safe_fclose(arg_serialization
);
1623 fds
= fdset_free(fds
);
1625 /* Reopen the console */
1626 (void) make_console_stdio();
1628 for (j
= 1, i
= 1; j
< (unsigned) argc
; j
++)
1629 args
[i
++] = argv
[j
];
1631 assert(i
<= args_size
);
1633 /* Reenable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1634 (void) reset_all_signal_handlers();
1635 (void) reset_signal_mask();
1637 if (switch_root_init
) {
1638 args
[0] = switch_root_init
;
1639 (void) execv(args
[0], (char* const*) args
);
1640 log_warning_errno(errno
, "Failed to execute configured init, trying fallback: %m");
1643 args
[0] = "/sbin/init";
1644 (void) execv(args
[0], (char* const*) args
);
1647 manager_status_printf(NULL
, STATUS_TYPE_EMERGENCY
,
1648 ANSI_HIGHLIGHT_RED
" !! " ANSI_NORMAL
,
1649 "Failed to execute /sbin/init");
1652 log_warning("No /sbin/init, trying fallback");
1654 args
[0] = "/bin/sh";
1656 (void) execv(args
[0], (char* const*) args
);
1657 log_error_errno(errno
, "Failed to execute /bin/sh, giving up: %m");
1659 log_warning_errno(r
, "Failed to execute /sbin/init, giving up: %m");
1661 *ret_error_message
= "Failed to execute fallback shell";
1664 static int invoke_main_loop(
1666 bool *ret_reexecute
,
1667 int *ret_retval
, /* Return parameters relevant for shutting down */
1668 const char **ret_shutdown_verb
, /* … */
1669 FDSet
**ret_fds
, /* Return parameters for reexecuting */
1670 char **ret_switch_root_dir
, /* … */
1671 char **ret_switch_root_init
, /* … */
1672 const char **ret_error_message
) {
1677 assert(ret_reexecute
);
1679 assert(ret_shutdown_verb
);
1681 assert(ret_switch_root_dir
);
1682 assert(ret_switch_root_init
);
1683 assert(ret_error_message
);
1686 r
= manager_loop(m
);
1688 *ret_error_message
= "Failed to run main loop";
1689 return log_emergency_errno(r
, "Failed to run main loop: %m");
1692 switch ((ManagerObjective
) r
) {
1694 case MANAGER_RELOAD
: {
1695 LogTarget saved_log_target
;
1696 int saved_log_level
;
1698 log_info("Reloading.");
1700 /* First, save any overridden log level/target, then parse the configuration file, which might
1701 * change the log level to new settings. */
1703 saved_log_level
= m
->log_level_overridden
? log_get_max_level() : -1;
1704 saved_log_target
= m
->log_target_overridden
? log_get_target() : _LOG_TARGET_INVALID
;
1706 r
= parse_config_file();
1708 log_warning_errno(r
, "Failed to parse config file, ignoring: %m");
1710 set_manager_defaults(m
);
1712 if (saved_log_level
>= 0)
1713 manager_override_log_level(m
, saved_log_level
);
1714 if (saved_log_target
>= 0)
1715 manager_override_log_target(m
, saved_log_target
);
1717 r
= manager_reload(m
);
1719 /* Reloading failed before the point of no return. Let's continue running as if nothing happened. */
1720 m
->objective
= MANAGER_OK
;
1725 case MANAGER_REEXECUTE
:
1727 r
= prepare_reexecute(m
, &arg_serialization
, ret_fds
, false);
1729 *ret_error_message
= "Failed to prepare for reexecution";
1733 log_notice("Reexecuting.");
1735 *ret_reexecute
= true;
1736 *ret_retval
= EXIT_SUCCESS
;
1737 *ret_shutdown_verb
= NULL
;
1738 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
1742 case MANAGER_SWITCH_ROOT
:
1743 if (!m
->switch_root_init
) {
1744 r
= prepare_reexecute(m
, &arg_serialization
, ret_fds
, true);
1746 *ret_error_message
= "Failed to prepare for reexecution";
1752 log_notice("Switching root.");
1754 *ret_reexecute
= true;
1755 *ret_retval
= EXIT_SUCCESS
;
1756 *ret_shutdown_verb
= NULL
;
1758 /* Steal the switch root parameters */
1759 *ret_switch_root_dir
= m
->switch_root
;
1760 *ret_switch_root_init
= m
->switch_root_init
;
1761 m
->switch_root
= m
->switch_root_init
= NULL
;
1767 if (MANAGER_IS_USER(m
)) {
1770 *ret_reexecute
= false;
1771 *ret_retval
= m
->return_value
;
1772 *ret_shutdown_verb
= NULL
;
1774 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
1780 case MANAGER_REBOOT
:
1781 case MANAGER_POWEROFF
:
1783 case MANAGER_KEXEC
: {
1784 static const char * const table
[_MANAGER_OBJECTIVE_MAX
] = {
1785 [MANAGER_EXIT
] = "exit",
1786 [MANAGER_REBOOT
] = "reboot",
1787 [MANAGER_POWEROFF
] = "poweroff",
1788 [MANAGER_HALT
] = "halt",
1789 [MANAGER_KEXEC
] = "kexec",
1792 log_notice("Shutting down.");
1794 *ret_reexecute
= false;
1795 *ret_retval
= m
->return_value
;
1796 assert_se(*ret_shutdown_verb
= table
[m
->objective
]);
1798 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
1804 assert_not_reached("Unknown or unexpected manager objective.");
1809 static void log_execution_mode(bool *ret_first_boot
) {
1810 assert(ret_first_boot
);
1815 log_info(PACKAGE_STRING
" running in %ssystem mode. (" SYSTEMD_FEATURES
")",
1816 arg_action
== ACTION_TEST
? "test " : "" );
1818 v
= detect_virtualization();
1820 log_info("Detected virtualization %s.", virtualization_to_string(v
));
1822 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1825 *ret_first_boot
= false;
1826 log_info("Running in initial RAM disk.");
1828 /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
1829 * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
1830 * doesn't it's unpopulated. This allows container managers and installers to provision a
1831 * couple of files already. If the container manager wants to provision the machine ID itself
1832 * it should pass $container_uuid to PID 1. */
1834 *ret_first_boot
= access("/etc/machine-id", F_OK
) < 0;
1835 if (*ret_first_boot
)
1836 log_info("Running with unpopulated /etc.");
1839 if (DEBUG_LOGGING
) {
1840 _cleanup_free_
char *t
;
1842 t
= uid_to_name(getuid());
1843 log_debug(PACKAGE_STRING
" running in %suser mode for user " UID_FMT
"/%s. (" SYSTEMD_FEATURES
")",
1844 arg_action
== ACTION_TEST
? " test" : "", getuid(), strna(t
));
1847 *ret_first_boot
= false;
1851 static int initialize_runtime(
1853 struct rlimit
*saved_rlimit_nofile
,
1854 struct rlimit
*saved_rlimit_memlock
,
1855 const char **ret_error_message
) {
1859 assert(ret_error_message
);
1861 /* Sets up various runtime parameters. Many of these initializations are conditionalized:
1863 * - Some only apply to --system instances
1864 * - Some only apply to --user instances
1865 * - Some only apply when we first start up, but not when we reexecute
1868 if (arg_action
!= ACTION_RUN
)
1872 /* Make sure we leave a core dump without panicing the kernel. */
1873 install_crash_handler();
1876 r
= mount_cgroup_controllers(arg_join_controllers
);
1878 *ret_error_message
= "Failed to mount cgroup hierarchies";
1884 machine_id_setup(NULL
, arg_machine_id
, NULL
);
1886 bump_unix_max_dgram_qlen();
1888 write_container_id();
1891 if (arg_watchdog_device
) {
1892 r
= watchdog_set_device(arg_watchdog_device
);
1894 log_warning_errno(r
, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device
);
1897 if (arg_runtime_watchdog
> 0 && arg_runtime_watchdog
!= USEC_INFINITY
)
1898 watchdog_set_timeout(&arg_runtime_watchdog
);
1901 if (arg_timer_slack_nsec
!= NSEC_INFINITY
)
1902 if (prctl(PR_SET_TIMERSLACK
, arg_timer_slack_nsec
) < 0)
1903 log_warning_errno(errno
, "Failed to adjust timer slack, ignoring: %m");
1905 if (arg_system
&& !cap_test_all(arg_capability_bounding_set
)) {
1906 r
= capability_bounding_set_drop_usermode(arg_capability_bounding_set
);
1908 *ret_error_message
= "Failed to drop capability bounding set of usermode helpers";
1909 return log_emergency_errno(r
, "Failed to drop capability bounding set of usermode helpers: %m");
1912 r
= capability_bounding_set_drop(arg_capability_bounding_set
, true);
1914 *ret_error_message
= "Failed to drop capability bounding set";
1915 return log_emergency_errno(r
, "Failed to drop capability bounding set: %m");
1919 if (arg_system
&& arg_no_new_privs
) {
1920 if (prctl(PR_SET_NO_NEW_PRIVS
, 1, 0, 0, 0) < 0) {
1921 *ret_error_message
= "Failed to disable new privileges";
1922 return log_emergency_errno(errno
, "Failed to disable new privileges: %m");
1926 if (arg_syscall_archs
) {
1927 r
= enforce_syscall_archs(arg_syscall_archs
);
1929 *ret_error_message
= "Failed to set syscall architectures";
1935 /* Become reaper of our children */
1936 if (prctl(PR_SET_CHILD_SUBREAPER
, 1) < 0)
1937 log_warning_errno(errno
, "Failed to make us a subreaper: %m");
1940 /* Bump up RLIMIT_NOFILE for systemd itself */
1941 (void) bump_rlimit_nofile(saved_rlimit_nofile
);
1942 (void) bump_rlimit_memlock(saved_rlimit_memlock
);
1948 static int do_queue_default_job(
1950 const char **ret_error_message
) {
1952 _cleanup_(sd_bus_error_free
) sd_bus_error error
= SD_BUS_ERROR_NULL
;
1953 Job
*default_unit_job
;
1954 Unit
*target
= NULL
;
1957 log_debug("Activating default unit: %s", arg_default_unit
);
1959 r
= manager_load_startable_unit_or_warn(m
, arg_default_unit
, NULL
, &target
);
1961 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET
);
1963 r
= manager_load_startable_unit_or_warn(m
, SPECIAL_RESCUE_TARGET
, NULL
, &target
);
1965 *ret_error_message
= r
== -ERFKILL
? "Rescue target masked"
1966 : "Failed to load rescue target";
1971 assert(target
->load_state
== UNIT_LOADED
);
1973 r
= manager_add_job(m
, JOB_START
, target
, JOB_ISOLATE
, &error
, &default_unit_job
);
1975 log_debug_errno(r
, "Default target could not be isolated, starting instead: %s", bus_error_message(&error
, r
));
1977 sd_bus_error_free(&error
);
1979 r
= manager_add_job(m
, JOB_START
, target
, JOB_REPLACE
, &error
, &default_unit_job
);
1981 *ret_error_message
= "Failed to start default target";
1982 return log_emergency_errno(r
, "Failed to start default target: %s", bus_error_message(&error
, r
));
1986 *ret_error_message
= "Failed to isolate default target";
1987 return log_emergency_errno(r
, "Failed to isolate default target: %s", bus_error_message(&error
, r
));
1990 m
->default_unit_job_id
= default_unit_job
->id
;
1995 static void free_arguments(void) {
1997 /* Frees all arg_* variables, with the exception of arg_serialization */
1998 rlimit_free_all(arg_default_rlimit
);
2000 arg_default_unit
= mfree(arg_default_unit
);
2001 arg_confirm_spawn
= mfree(arg_confirm_spawn
);
2002 arg_join_controllers
= strv_free_free(arg_join_controllers
);
2003 arg_default_environment
= strv_free(arg_default_environment
);
2004 arg_syscall_archs
= set_free(arg_syscall_archs
);
2007 static int load_configuration(int argc
, char **argv
, const char **ret_error_message
) {
2010 assert(ret_error_message
);
2012 arg_default_tasks_max
= system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE
, 100U);
2014 r
= parse_config_file();
2016 *ret_error_message
= "Failed to parse config file";
2021 r
= proc_cmdline_parse(parse_proc_cmdline_item
, NULL
, 0);
2023 log_warning_errno(r
, "Failed to parse kernel command line, ignoring: %m");
2026 /* Note that this also parses bits from the kernel command line, including "debug". */
2027 log_parse_environment();
2029 r
= parse_argv(argc
, argv
);
2031 *ret_error_message
= "Failed to parse commandline arguments";
2035 /* Initialize default unit */
2036 if (!arg_default_unit
) {
2037 arg_default_unit
= strdup(SPECIAL_DEFAULT_TARGET
);
2038 if (!arg_default_unit
) {
2039 *ret_error_message
= "Failed to set default unit";
2044 /* Initialize the show status setting if it hasn't been set explicitly yet */
2045 if (arg_show_status
== _SHOW_STATUS_INVALID
)
2046 arg_show_status
= SHOW_STATUS_YES
;
2051 static int safety_checks(void) {
2053 if (getpid_cached() == 1 &&
2054 arg_action
!= ACTION_RUN
) {
2055 log_error("Unsupported execution mode while PID 1.");
2059 if (getpid_cached() == 1 &&
2061 log_error("Can't run --user mode as PID 1.");
2065 if (arg_action
== ACTION_RUN
&&
2067 getpid_cached() != 1) {
2068 log_error("Can't run system mode unless PID 1.");
2072 if (arg_action
== ACTION_TEST
&&
2074 log_error("Don't run test mode as root.");
2079 arg_action
== ACTION_RUN
&&
2081 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
2086 arg_action
== ACTION_RUN
&&
2087 !getenv("XDG_RUNTIME_DIR")) {
2088 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
2093 arg_action
== ACTION_RUN
&&
2094 running_in_chroot() > 0) {
2095 log_error("Cannot be run in a chroot() environment.");
2102 static int initialize_security(
2103 bool *loaded_policy
,
2104 dual_timestamp
*security_start_timestamp
,
2105 dual_timestamp
*security_finish_timestamp
,
2106 const char **ret_error_message
) {
2110 assert(loaded_policy
);
2111 assert(security_start_timestamp
);
2112 assert(security_finish_timestamp
);
2113 assert(ret_error_message
);
2115 dual_timestamp_get(security_start_timestamp
);
2117 r
= mac_selinux_setup(loaded_policy
);
2119 *ret_error_message
= "Failed to load SELinux policy";
2123 r
= mac_smack_setup(loaded_policy
);
2125 *ret_error_message
= "Failed to load SMACK policy";
2131 *ret_error_message
= "Failed to load IMA policy";
2135 dual_timestamp_get(security_finish_timestamp
);
2139 static void test_summary(Manager
*m
) {
2142 printf("-> By units:\n");
2143 manager_dump_units(m
, stdout
, "\t");
2145 printf("-> By jobs:\n");
2146 manager_dump_jobs(m
, stdout
, "\t");
2149 static int collect_fds(FDSet
**ret_fds
, const char **ret_error_message
) {
2153 assert(ret_error_message
);
2155 r
= fdset_new_fill(ret_fds
);
2157 *ret_error_message
= "Failed to allocate fd set";
2158 return log_emergency_errno(r
, "Failed to allocate fd set: %m");
2161 fdset_cloexec(*ret_fds
, true);
2163 if (arg_serialization
)
2164 assert_se(fdset_remove(*ret_fds
, fileno(arg_serialization
)) >= 0);
2169 static void setup_console_terminal(bool skip_setup
) {
2174 /* Become a session leader if we aren't one yet. */
2177 /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2179 (void) release_terminal();
2181 /* Reset the console, but only if this is really init and we are freshly booted */
2182 if (getpid_cached() == 1 && !skip_setup
)
2183 (void) console_setup();
2186 static bool early_skip_setup_check(int argc
, char *argv
[]) {
2187 bool found_deserialize
= false;
2190 /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2191 * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2192 * anyway, even if in that case we also do deserialization. */
2194 for (i
= 1; i
< argc
; i
++) {
2195 if (streq(argv
[i
], "--switched-root"))
2196 return false; /* If we switched root, don't skip the setup. */
2197 else if (streq(argv
[i
], "--deserialize"))
2198 found_deserialize
= true;
2201 return found_deserialize
; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2204 int main(int argc
, char *argv
[]) {
2206 dual_timestamp initrd_timestamp
= DUAL_TIMESTAMP_NULL
, userspace_timestamp
= DUAL_TIMESTAMP_NULL
, kernel_timestamp
= DUAL_TIMESTAMP_NULL
,
2207 security_start_timestamp
= DUAL_TIMESTAMP_NULL
, security_finish_timestamp
= DUAL_TIMESTAMP_NULL
;
2208 struct rlimit saved_rlimit_nofile
= RLIMIT_MAKE_CONST(0), saved_rlimit_memlock
= RLIMIT_MAKE_CONST((rlim_t
) -1);
2209 bool skip_setup
, loaded_policy
= false, queue_default_job
= false, first_boot
= false, reexecute
= false;
2210 char *switch_root_dir
= NULL
, *switch_root_init
= NULL
;
2211 usec_t before_startup
, after_startup
;
2212 static char systemd
[] = "systemd";
2213 char timespan
[FORMAT_TIMESPAN_MAX
];
2214 const char *shutdown_verb
= NULL
, *error_message
= NULL
;
2215 int r
, retval
= EXIT_FAILURE
;
2219 /* SysV compatibility: redirect init → telinit */
2220 redirect_telinit(argc
, argv
);
2222 /* Take timestamps early on */
2223 dual_timestamp_from_monotonic(&kernel_timestamp
, 0);
2224 dual_timestamp_get(&userspace_timestamp
);
2226 /* Figure out whether we need to do initialize the system, or if we already did that because we are
2228 skip_setup
= early_skip_setup_check(argc
, argv
);
2230 /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2231 * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
2232 program_invocation_short_name
= systemd
;
2233 (void) prctl(PR_SET_NAME
, systemd
);
2235 /* Save the original command line */
2239 /* Make sure that if the user says "syslog" we actually log to the journal. */
2240 log_set_upgrade_syslog_to_journal(true);
2242 if (getpid_cached() == 1) {
2243 /* When we run as PID 1 force system mode */
2246 /* Disable the umask logic */
2249 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2250 * activated yet (even though the log socket for it exists). */
2251 log_set_prohibit_ipc(true);
2253 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2254 * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2255 * child process right before execve()'ing the actual binary, at a point in time where socket
2256 * activation stderr/stdout area already set up. */
2257 log_set_always_reopen_console(true);
2259 if (detect_container() <= 0) {
2261 /* Running outside of a container as PID 1 */
2262 log_set_target(LOG_TARGET_KMSG
);
2266 initrd_timestamp
= userspace_timestamp
;
2269 r
= mount_setup_early();
2271 error_message
= "Failed to mount early API filesystems";
2275 r
= initialize_security(
2277 &security_start_timestamp
,
2278 &security_finish_timestamp
,
2284 if (mac_selinux_init() < 0) {
2285 error_message
= "Failed to initialize SELinux policy";
2292 /* Set the default for later on, but don't actually open the logs like this for now. Note that
2293 * if we are transitioning from the initrd there might still be journal fd open, and we
2294 * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2296 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG
);
2299 /* Running inside a container, as PID 1 */
2300 log_set_target(LOG_TARGET_CONSOLE
);
2303 /* For later on, see above... */
2304 log_set_target(LOG_TARGET_JOURNAL
);
2306 /* clear the kernel timestamp,
2307 * because we are in a container */
2308 kernel_timestamp
= DUAL_TIMESTAMP_NULL
;
2311 initialize_coredump(skip_setup
);
2313 r
= fixup_environment();
2315 log_emergency_errno(r
, "Failed to fix up PID 1 environment: %m");
2316 error_message
= "Failed to fix up PID1 environment";
2321 /* Running as user instance */
2323 log_set_target(LOG_TARGET_AUTO
);
2326 /* clear the kernel timestamp,
2327 * because we are not PID 1 */
2328 kernel_timestamp
= DUAL_TIMESTAMP_NULL
;
2332 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2333 * they never log into the console. */
2334 log_show_color(colors_enabled());
2336 r
= make_null_stdio();
2338 log_warning_errno(r
, "Failed to redirect standard streams to /dev/null, ignoring: %m");
2341 /* Mount /proc, /sys and friends, so that /proc/cmdline and
2342 * /proc/$PID/fd is available. */
2343 if (getpid_cached() == 1) {
2345 /* Load the kernel modules early. */
2349 r
= mount_setup(loaded_policy
);
2351 error_message
= "Failed to mount API filesystems";
2356 /* Reset all signal handlers. */
2357 (void) reset_all_signal_handlers();
2358 (void) ignore_signals(SIGNALS_IGNORE
, -1);
2360 r
= load_configuration(argc
, argv
, &error_message
);
2364 r
= safety_checks();
2368 if (IN_SET(arg_action
, ACTION_TEST
, ACTION_HELP
, ACTION_DUMP_CONFIGURATION_ITEMS
, ACTION_DUMP_BUS_PROPERTIES
))
2369 (void) pager_open(arg_no_pager
, false);
2371 if (arg_action
!= ACTION_RUN
)
2374 if (arg_action
== ACTION_HELP
) {
2375 retval
= help() < 0 ? EXIT_FAILURE
: EXIT_SUCCESS
;
2377 } else if (arg_action
== ACTION_VERSION
) {
2380 } else if (arg_action
== ACTION_DUMP_CONFIGURATION_ITEMS
) {
2381 unit_dump_config_items(stdout
);
2382 retval
= EXIT_SUCCESS
;
2384 } else if (arg_action
== ACTION_DUMP_BUS_PROPERTIES
) {
2385 dump_bus_properties(stdout
);
2386 retval
= EXIT_SUCCESS
;
2390 assert_se(IN_SET(arg_action
, ACTION_RUN
, ACTION_TEST
));
2392 /* Move out of the way, so that we won't block unmounts */
2393 assert_se(chdir("/") == 0);
2395 if (arg_action
== ACTION_RUN
) {
2397 /* A core pattern might have been specified via the cmdline. */
2398 initialize_core_pattern(skip_setup
);
2400 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2403 /* Remember open file descriptors for later deserialization */
2404 r
= collect_fds(&fds
, &error_message
);
2408 /* Give up any control of the console, but make sure its initialized. */
2409 setup_console_terminal(skip_setup
);
2411 /* Open the logging devices, if possible and necessary */
2415 log_execution_mode(&first_boot
);
2417 r
= initialize_runtime(skip_setup
,
2418 &saved_rlimit_nofile
,
2419 &saved_rlimit_memlock
,
2424 r
= manager_new(arg_system
? UNIT_FILE_SYSTEM
: UNIT_FILE_USER
,
2425 arg_action
== ACTION_TEST
? MANAGER_TEST_FULL
: 0,
2428 log_emergency_errno(r
, "Failed to allocate manager object: %m");
2429 error_message
= "Failed to allocate manager object";
2433 m
->timestamps
[MANAGER_TIMESTAMP_KERNEL
] = kernel_timestamp
;
2434 m
->timestamps
[MANAGER_TIMESTAMP_INITRD
] = initrd_timestamp
;
2435 m
->timestamps
[MANAGER_TIMESTAMP_USERSPACE
] = userspace_timestamp
;
2436 m
->timestamps
[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START
)] = security_start_timestamp
;
2437 m
->timestamps
[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH
)] = security_finish_timestamp
;
2439 set_manager_defaults(m
);
2440 set_manager_settings(m
);
2441 manager_set_first_boot(m
, first_boot
);
2443 /* Remember whether we should queue the default job */
2444 queue_default_job
= !arg_serialization
|| arg_switched_root
;
2446 before_startup
= now(CLOCK_MONOTONIC
);
2448 r
= manager_startup(m
, arg_serialization
, fds
);
2450 error_message
= "Failed to start up manager";
2454 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2455 fds
= fdset_free(fds
);
2456 arg_serialization
= safe_fclose(arg_serialization
);
2458 if (queue_default_job
) {
2459 r
= do_queue_default_job(m
, &error_message
);
2464 after_startup
= now(CLOCK_MONOTONIC
);
2466 log_full(arg_action
== ACTION_TEST
? LOG_INFO
: LOG_DEBUG
,
2467 "Loaded units and determined initial transaction in %s.",
2468 format_timespan(timespan
, sizeof(timespan
), after_startup
- before_startup
, 100 * USEC_PER_MSEC
));
2470 if (arg_action
== ACTION_TEST
) {
2472 retval
= EXIT_SUCCESS
;
2476 (void) invoke_main_loop(m
,
2489 arg_shutdown_watchdog
= m
->shutdown_watchdog
;
2490 m
= manager_free(m
);
2494 mac_selinux_finish();
2497 do_reexecute(argc
, argv
,
2498 &saved_rlimit_nofile
,
2499 &saved_rlimit_memlock
,
2503 &error_message
); /* This only returns if reexecution failed */
2505 arg_serialization
= safe_fclose(arg_serialization
);
2506 fds
= fdset_free(fds
);
2508 #if HAVE_VALGRIND_VALGRIND_H
2509 /* If we are PID 1 and running under valgrind, then let's exit
2510 * here explicitly. valgrind will only generate nice output on
2511 * exit(), not on exec(), hence let's do the former not the
2513 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND
) {
2514 /* Cleanup watchdog_device strings for valgrind. We need them
2515 * in become_shutdown() so normally we cannot free them yet. */
2516 watchdog_free_device();
2517 arg_watchdog_device
= mfree(arg_watchdog_device
);
2522 if (shutdown_verb
) {
2523 r
= become_shutdown(shutdown_verb
, retval
);
2524 log_error_errno(r
, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
2525 error_message
= "Failed to execute shutdown binary";
2528 watchdog_free_device();
2529 arg_watchdog_device
= mfree(arg_watchdog_device
);
2531 if (getpid_cached() == 1) {
2533 manager_status_printf(NULL
, STATUS_TYPE_EMERGENCY
,
2534 ANSI_HIGHLIGHT_RED
"!!!!!!" ANSI_NORMAL
,
2535 "%s, freezing.", error_message
);