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 "limits-util.h"
51 #include "load-fragment.h"
53 #include "loopback-setup.h"
54 #include "machine-id-setup.h"
57 #include "mount-setup.h"
60 #include "parse-util.h"
61 #include "path-util.h"
62 #include "pretty-print.h"
63 #include "proc-cmdline.h"
64 #include "process-util.h"
65 #include "raw-clone.h"
66 #include "rlimit-util.h"
68 #include "seccomp-util.h"
70 #include "selinux-setup.h"
71 #include "selinux-util.h"
72 #include "signal-util.h"
73 #include "smack-setup.h"
75 #include "stat-util.h"
76 #include "stdio-util.h"
78 #include "switch-root.h"
79 #include "sysctl-util.h"
80 #include "terminal-util.h"
81 #include "umask-util.h"
82 #include "user-util.h"
87 #if HAS_FEATURE_ADDRESS_SANITIZER
88 #include <sanitizer/lsan_interface.h>
96 ACTION_DUMP_CONFIGURATION_ITEMS
,
97 ACTION_DUMP_BUS_PROPERTIES
,
98 } arg_action
= ACTION_RUN
;
100 /* Those variables are initalized to 0 automatically, so we avoid uninitialized memory access.
101 * Real defaults are assigned in reset_arguments() below. */
102 static char *arg_default_unit
;
103 static bool arg_system
;
104 static bool arg_dump_core
;
105 static int arg_crash_chvt
;
106 static bool arg_crash_shell
;
107 static bool arg_crash_reboot
;
108 static char *arg_confirm_spawn
;
109 static ShowStatus arg_show_status
;
110 static bool arg_switched_root
;
111 static PagerFlags arg_pager_flags
;
112 static bool arg_service_watchdogs
;
113 static ExecOutput arg_default_std_output
;
114 static ExecOutput arg_default_std_error
;
115 static usec_t arg_default_restart_usec
;
116 static usec_t arg_default_timeout_start_usec
;
117 static usec_t arg_default_timeout_stop_usec
;
118 static usec_t arg_default_timeout_abort_usec
;
119 static bool arg_default_timeout_abort_set
;
120 static usec_t arg_default_start_limit_interval
;
121 static unsigned arg_default_start_limit_burst
;
122 static usec_t arg_runtime_watchdog
;
123 static usec_t arg_shutdown_watchdog
;
124 static char *arg_early_core_pattern
;
125 static char *arg_watchdog_device
;
126 static char **arg_default_environment
;
127 static struct rlimit
*arg_default_rlimit
[_RLIMIT_MAX
];
128 static uint64_t arg_capability_bounding_set
;
129 static bool arg_no_new_privs
;
130 static nsec_t arg_timer_slack_nsec
;
131 static usec_t arg_default_timer_accuracy_usec
;
132 static Set
* arg_syscall_archs
;
133 static FILE* arg_serialization
;
134 static int arg_default_cpu_accounting
;
135 static bool arg_default_io_accounting
;
136 static bool arg_default_ip_accounting
;
137 static bool arg_default_blockio_accounting
;
138 static bool arg_default_memory_accounting
;
139 static bool arg_default_tasks_accounting
;
140 static uint64_t arg_default_tasks_max
;
141 static sd_id128_t arg_machine_id
;
142 static EmergencyAction arg_cad_burst_action
;
143 static OOMPolicy arg_default_oom_policy
;
144 static CPUSet arg_cpu_affinity
;
146 static int parse_configuration(void);
148 _noreturn_
static void freeze_or_exit_or_reboot(void) {
150 /* If we are running in a container, let's prefer exiting, after all we can propagate an exit code to
151 * the container manager, and thus inform it that something went wrong. */
152 if (detect_container() > 0) {
153 log_emergency("Exiting PID 1...");
154 _exit(EXIT_EXCEPTION
);
157 if (arg_crash_reboot
) {
158 log_notice("Rebooting in 10s...");
161 log_notice("Rebooting now...");
162 (void) reboot(RB_AUTOBOOT
);
163 log_emergency_errno(errno
, "Failed to reboot: %m");
166 log_emergency("Freezing execution.");
170 _noreturn_
static void crash(int sig
) {
174 if (getpid_cached() != 1)
175 /* Pass this on immediately, if this is not PID 1 */
177 else if (!arg_dump_core
)
178 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig
));
180 sa
= (struct sigaction
) {
181 .sa_handler
= nop_signal_handler
,
182 .sa_flags
= SA_NOCLDSTOP
|SA_RESTART
,
185 /* We want to wait for the core process, hence let's enable SIGCHLD */
186 (void) sigaction(SIGCHLD
, &sa
, NULL
);
188 pid
= raw_clone(SIGCHLD
);
190 log_emergency_errno(errno
, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig
));
192 /* Enable default signal handler for core dump */
194 sa
= (struct sigaction
) {
195 .sa_handler
= SIG_DFL
,
197 (void) sigaction(sig
, &sa
, NULL
);
199 /* Don't limit the coredump size */
200 (void) setrlimit(RLIMIT_CORE
, &RLIMIT_MAKE_CONST(RLIM_INFINITY
));
202 /* Just to be sure... */
205 /* Raise the signal again */
207 (void) kill(pid
, sig
); /* raise() would kill the parent */
209 assert_not_reached("We shouldn't be here...");
210 _exit(EXIT_EXCEPTION
);
215 /* Order things nicely. */
216 r
= wait_for_terminate(pid
, &status
);
218 log_emergency_errno(r
, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig
));
219 else if (status
.si_code
!= CLD_DUMPED
)
220 log_emergency("Caught <%s>, core dump failed (child "PID_FMT
", code=%s, status=%i/%s).",
221 signal_to_string(sig
),
222 pid
, sigchld_code_to_string(status
.si_code
),
224 strna(status
.si_code
== CLD_EXITED
225 ? exit_status_to_string(status
.si_status
, EXIT_STATUS_MINIMAL
)
226 : signal_to_string(status
.si_status
)));
228 log_emergency("Caught <%s>, dumped core as pid "PID_FMT
".", signal_to_string(sig
), pid
);
232 if (arg_crash_chvt
>= 0)
233 (void) chvt(arg_crash_chvt
);
235 sa
= (struct sigaction
) {
236 .sa_handler
= SIG_IGN
,
237 .sa_flags
= SA_NOCLDSTOP
|SA_NOCLDWAIT
|SA_RESTART
,
240 /* Let the kernel reap children for us */
241 (void) sigaction(SIGCHLD
, &sa
, NULL
);
243 if (arg_crash_shell
) {
244 log_notice("Executing crash shell in 10s...");
247 pid
= raw_clone(SIGCHLD
);
249 log_emergency_errno(errno
, "Failed to fork off crash shell: %m");
252 (void) make_console_stdio();
253 (void) rlimit_nofile_safe();
254 (void) execle("/bin/sh", "/bin/sh", NULL
, environ
);
256 log_emergency_errno(errno
, "execle() failed: %m");
257 _exit(EXIT_EXCEPTION
);
259 log_info("Spawned crash shell as PID "PID_FMT
".", pid
);
260 (void) wait_for_terminate(pid
, NULL
);
264 freeze_or_exit_or_reboot();
267 static void install_crash_handler(void) {
268 static const struct sigaction sa
= {
270 .sa_flags
= SA_NODEFER
, /* So that we can raise the signal again from the signal handler */
274 /* We ignore the return value here, since, we don't mind if we
275 * cannot set up a crash handler */
276 r
= sigaction_many(&sa
, SIGNALS_CRASH_HANDLER
, -1);
278 log_debug_errno(r
, "I had trouble setting up the crash handler, ignoring: %m");
281 static int console_setup(void) {
282 _cleanup_close_
int tty_fd
= -1;
285 tty_fd
= open_terminal("/dev/console", O_WRONLY
|O_NOCTTY
|O_CLOEXEC
);
287 return log_error_errno(tty_fd
, "Failed to open /dev/console: %m");
289 /* We don't want to force text mode. plymouth may be showing
290 * pictures already from initrd. */
291 r
= reset_terminal_fd(tty_fd
, false);
293 return log_error_errno(r
, "Failed to reset /dev/console: %m");
298 static int parse_crash_chvt(const char *value
) {
301 if (safe_atoi(value
, &arg_crash_chvt
) >= 0)
304 b
= parse_boolean(value
);
309 arg_crash_chvt
= 0; /* switch to where kmsg goes */
311 arg_crash_chvt
= -1; /* turn off switching */
316 static int parse_confirm_spawn(const char *value
, char **console
) {
320 r
= value
? parse_boolean(value
) : 1;
326 if (r
> 0) /* on with default tty */
327 s
= strdup("/dev/console");
328 else if (is_path(value
)) /* on with fully qualified path */
330 else /* on with only a tty file name, not a fully qualified path */
331 s
= strjoin("/dev/", value
);
339 static int set_machine_id(const char *m
) {
343 if (sd_id128_from_string(m
, &t
) < 0)
346 if (sd_id128_is_null(t
))
353 static int parse_proc_cmdline_item(const char *key
, const char *value
, void *data
) {
359 if (STR_IN_SET(key
, "systemd.unit", "rd.systemd.unit")) {
361 if (proc_cmdline_value_missing(key
, value
))
364 if (!unit_name_is_valid(value
, UNIT_NAME_PLAIN
|UNIT_NAME_INSTANCE
))
365 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key
, value
);
366 else if (in_initrd() == !!startswith(key
, "rd.")) {
367 if (free_and_strdup(&arg_default_unit
, value
) < 0)
371 } else if (proc_cmdline_key_streq(key
, "systemd.dump_core")) {
373 r
= value
? parse_boolean(value
) : true;
375 log_warning_errno(r
, "Failed to parse dump core switch %s, ignoring: %m", value
);
379 } else if (proc_cmdline_key_streq(key
, "systemd.early_core_pattern")) {
381 if (proc_cmdline_value_missing(key
, value
))
384 if (path_is_absolute(value
))
385 (void) parse_path_argument_and_warn(value
, false, &arg_early_core_pattern
);
387 log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value
);
389 } else if (proc_cmdline_key_streq(key
, "systemd.crash_chvt")) {
392 arg_crash_chvt
= 0; /* turn on */
394 r
= parse_crash_chvt(value
);
396 log_warning_errno(r
, "Failed to parse crash chvt switch %s, ignoring: %m", value
);
399 } else if (proc_cmdline_key_streq(key
, "systemd.crash_shell")) {
401 r
= value
? parse_boolean(value
) : true;
403 log_warning_errno(r
, "Failed to parse crash shell switch %s, ignoring: %m", value
);
407 } else if (proc_cmdline_key_streq(key
, "systemd.crash_reboot")) {
409 r
= value
? parse_boolean(value
) : true;
411 log_warning_errno(r
, "Failed to parse crash reboot switch %s, ignoring: %m", value
);
413 arg_crash_reboot
= r
;
415 } else if (proc_cmdline_key_streq(key
, "systemd.confirm_spawn")) {
418 r
= parse_confirm_spawn(value
, &s
);
420 log_warning_errno(r
, "Failed to parse confirm_spawn switch %s, ignoring: %m", value
);
422 free_and_replace(arg_confirm_spawn
, s
);
424 } else if (proc_cmdline_key_streq(key
, "systemd.service_watchdogs")) {
426 r
= value
? parse_boolean(value
) : true;
428 log_warning_errno(r
, "Failed to parse service watchdog switch %s, ignoring: %m", value
);
430 arg_service_watchdogs
= r
;
432 } else if (proc_cmdline_key_streq(key
, "systemd.show_status")) {
435 r
= parse_show_status(value
, &arg_show_status
);
437 log_warning_errno(r
, "Failed to parse show status switch %s, ignoring: %m", value
);
439 arg_show_status
= SHOW_STATUS_YES
;
441 } else if (proc_cmdline_key_streq(key
, "systemd.default_standard_output")) {
443 if (proc_cmdline_value_missing(key
, value
))
446 r
= exec_output_from_string(value
);
448 log_warning_errno(r
, "Failed to parse default standard output switch %s, ignoring: %m", value
);
450 arg_default_std_output
= r
;
452 } else if (proc_cmdline_key_streq(key
, "systemd.default_standard_error")) {
454 if (proc_cmdline_value_missing(key
, value
))
457 r
= exec_output_from_string(value
);
459 log_warning_errno(r
, "Failed to parse default standard error switch %s, ignoring: %m", value
);
461 arg_default_std_error
= r
;
463 } else if (streq(key
, "systemd.setenv")) {
465 if (proc_cmdline_value_missing(key
, value
))
468 if (env_assignment_is_valid(value
)) {
471 env
= strv_env_set(arg_default_environment
, value
);
475 arg_default_environment
= env
;
477 log_warning("Environment variable name '%s' is not valid. Ignoring.", value
);
479 } else if (proc_cmdline_key_streq(key
, "systemd.machine_id")) {
481 if (proc_cmdline_value_missing(key
, value
))
484 r
= set_machine_id(value
);
486 log_warning_errno(r
, "MachineID '%s' is not valid, ignoring: %m", value
);
488 } else if (proc_cmdline_key_streq(key
, "systemd.default_timeout_start_sec")) {
490 if (proc_cmdline_value_missing(key
, value
))
493 r
= parse_sec(value
, &arg_default_timeout_start_usec
);
495 log_warning_errno(r
, "Failed to parse default start timeout '%s', ignoring: %m", value
);
497 if (arg_default_timeout_start_usec
<= 0)
498 arg_default_timeout_start_usec
= USEC_INFINITY
;
500 } else if (proc_cmdline_key_streq(key
, "systemd.watchdog_device")) {
502 if (proc_cmdline_value_missing(key
, value
))
505 (void) parse_path_argument_and_warn(value
, false, &arg_watchdog_device
);
507 } else if (streq(key
, "quiet") && !value
) {
509 if (arg_show_status
== _SHOW_STATUS_INVALID
)
510 arg_show_status
= SHOW_STATUS_AUTO
;
512 } else if (streq(key
, "debug") && !value
) {
514 /* Note that log_parse_environment() handles 'debug'
515 * too, and sets the log level to LOG_DEBUG. */
517 if (detect_container() > 0)
518 log_set_target(LOG_TARGET_CONSOLE
);
523 /* SysV compatibility */
524 target
= runlevel_to_target(key
);
526 return free_and_strdup(&arg_default_unit
, target
);
532 #define DEFINE_SETTER(name, func, descr) \
533 static int name(const char *unit, \
534 const char *filename, \
536 const char *section, \
537 unsigned section_line, \
538 const char *lvalue, \
540 const char *rvalue, \
552 log_syntax(unit, LOG_ERR, filename, line, r, \
553 "Invalid " descr "'%s': %m", \
559 DEFINE_SETTER(config_parse_level2
, log_set_max_level_from_string
, "log level");
560 DEFINE_SETTER(config_parse_target
, log_set_target_from_string
, "target");
561 DEFINE_SETTER(config_parse_color
, log_show_color_from_string
, "color" );
562 DEFINE_SETTER(config_parse_location
, log_show_location_from_string
, "location");
564 static int config_parse_cpu_affinity2(
566 const char *filename
,
569 unsigned section_line
,
576 CPUSet
*affinity
= data
;
580 (void) parse_cpu_set_extend(rvalue
, affinity
, true, unit
, filename
, line
, lvalue
);
585 static int config_parse_show_status(
587 const char *filename
,
590 unsigned section_line
,
598 ShowStatus
*b
= data
;
605 k
= parse_show_status(rvalue
, b
);
607 log_syntax(unit
, LOG_ERR
, filename
, line
, k
, "Failed to parse show status setting, ignoring: %s", rvalue
);
614 static int config_parse_output_restricted(
616 const char *filename
,
619 unsigned section_line
,
626 ExecOutput t
, *eo
= data
;
633 t
= exec_output_from_string(rvalue
);
635 log_syntax(unit
, LOG_ERR
, filename
, line
, 0, "Failed to parse output type, ignoring: %s", rvalue
);
639 if (IN_SET(t
, EXEC_OUTPUT_SOCKET
, EXEC_OUTPUT_NAMED_FD
, EXEC_OUTPUT_FILE
, EXEC_OUTPUT_FILE_APPEND
)) {
640 log_syntax(unit
, LOG_ERR
, filename
, line
, 0, "Standard output types socket, fd:, file:, append: are not supported as defaults, ignoring: %s", rvalue
);
648 static int config_parse_crash_chvt(
650 const char *filename
,
653 unsigned section_line
,
666 r
= parse_crash_chvt(rvalue
);
668 log_syntax(unit
, LOG_ERR
, filename
, line
, r
, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue
);
675 static int config_parse_timeout_abort(
677 const char *filename
,
680 unsigned section_line
,
693 rvalue
+= strspn(rvalue
, WHITESPACE
);
694 if (isempty(rvalue
)) {
695 arg_default_timeout_abort_set
= false;
699 r
= parse_sec(rvalue
, &arg_default_timeout_abort_usec
);
701 log_syntax(unit
, LOG_ERR
, filename
, line
, r
, "Failed to parse DefaultTimeoutAbortSec= setting, ignoring: %s", rvalue
);
705 arg_default_timeout_abort_set
= true;
709 static int parse_config_file(void) {
711 const ConfigTableItem items
[] = {
712 { "Manager", "LogLevel", config_parse_level2
, 0, NULL
},
713 { "Manager", "LogTarget", config_parse_target
, 0, NULL
},
714 { "Manager", "LogColor", config_parse_color
, 0, NULL
},
715 { "Manager", "LogLocation", config_parse_location
, 0, NULL
},
716 { "Manager", "DumpCore", config_parse_bool
, 0, &arg_dump_core
},
717 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt
, 0, NULL
},
718 { "Manager", "CrashChangeVT", config_parse_crash_chvt
, 0, NULL
},
719 { "Manager", "CrashShell", config_parse_bool
, 0, &arg_crash_shell
},
720 { "Manager", "CrashReboot", config_parse_bool
, 0, &arg_crash_reboot
},
721 { "Manager", "ShowStatus", config_parse_show_status
, 0, &arg_show_status
},
722 { "Manager", "CPUAffinity", config_parse_cpu_affinity2
, 0, &arg_cpu_affinity
},
723 { "Manager", "JoinControllers", config_parse_warn_compat
, DISABLED_CONFIGURATION
, NULL
},
724 { "Manager", "RuntimeWatchdogSec", config_parse_sec
, 0, &arg_runtime_watchdog
},
725 { "Manager", "ShutdownWatchdogSec", config_parse_sec
, 0, &arg_shutdown_watchdog
},
726 { "Manager", "WatchdogDevice", config_parse_path
, 0, &arg_watchdog_device
},
727 { "Manager", "CapabilityBoundingSet", config_parse_capability_set
, 0, &arg_capability_bounding_set
},
728 { "Manager", "NoNewPrivileges", config_parse_bool
, 0, &arg_no_new_privs
},
730 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs
, 0, &arg_syscall_archs
},
732 { "Manager", "TimerSlackNSec", config_parse_nsec
, 0, &arg_timer_slack_nsec
},
733 { "Manager", "DefaultTimerAccuracySec", config_parse_sec
, 0, &arg_default_timer_accuracy_usec
},
734 { "Manager", "DefaultStandardOutput", config_parse_output_restricted
,0, &arg_default_std_output
},
735 { "Manager", "DefaultStandardError", config_parse_output_restricted
,0, &arg_default_std_error
},
736 { "Manager", "DefaultTimeoutStartSec", config_parse_sec
, 0, &arg_default_timeout_start_usec
},
737 { "Manager", "DefaultTimeoutStopSec", config_parse_sec
, 0, &arg_default_timeout_stop_usec
},
738 { "Manager", "DefaultTimeoutAbortSec", config_parse_timeout_abort
, 0, NULL
},
739 { "Manager", "DefaultRestartSec", config_parse_sec
, 0, &arg_default_restart_usec
},
740 { "Manager", "DefaultStartLimitInterval", config_parse_sec
, 0, &arg_default_start_limit_interval
}, /* obsolete alias */
741 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec
, 0, &arg_default_start_limit_interval
},
742 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned
, 0, &arg_default_start_limit_burst
},
743 { "Manager", "DefaultEnvironment", config_parse_environ
, 0, &arg_default_environment
},
744 { "Manager", "DefaultLimitCPU", config_parse_rlimit
, RLIMIT_CPU
, arg_default_rlimit
},
745 { "Manager", "DefaultLimitFSIZE", config_parse_rlimit
, RLIMIT_FSIZE
, arg_default_rlimit
},
746 { "Manager", "DefaultLimitDATA", config_parse_rlimit
, RLIMIT_DATA
, arg_default_rlimit
},
747 { "Manager", "DefaultLimitSTACK", config_parse_rlimit
, RLIMIT_STACK
, arg_default_rlimit
},
748 { "Manager", "DefaultLimitCORE", config_parse_rlimit
, RLIMIT_CORE
, arg_default_rlimit
},
749 { "Manager", "DefaultLimitRSS", config_parse_rlimit
, RLIMIT_RSS
, arg_default_rlimit
},
750 { "Manager", "DefaultLimitNOFILE", config_parse_rlimit
, RLIMIT_NOFILE
, arg_default_rlimit
},
751 { "Manager", "DefaultLimitAS", config_parse_rlimit
, RLIMIT_AS
, arg_default_rlimit
},
752 { "Manager", "DefaultLimitNPROC", config_parse_rlimit
, RLIMIT_NPROC
, arg_default_rlimit
},
753 { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit
, RLIMIT_MEMLOCK
, arg_default_rlimit
},
754 { "Manager", "DefaultLimitLOCKS", config_parse_rlimit
, RLIMIT_LOCKS
, arg_default_rlimit
},
755 { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit
, RLIMIT_SIGPENDING
, arg_default_rlimit
},
756 { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit
, RLIMIT_MSGQUEUE
, arg_default_rlimit
},
757 { "Manager", "DefaultLimitNICE", config_parse_rlimit
, RLIMIT_NICE
, arg_default_rlimit
},
758 { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit
, RLIMIT_RTPRIO
, arg_default_rlimit
},
759 { "Manager", "DefaultLimitRTTIME", config_parse_rlimit
, RLIMIT_RTTIME
, arg_default_rlimit
},
760 { "Manager", "DefaultCPUAccounting", config_parse_tristate
, 0, &arg_default_cpu_accounting
},
761 { "Manager", "DefaultIOAccounting", config_parse_bool
, 0, &arg_default_io_accounting
},
762 { "Manager", "DefaultIPAccounting", config_parse_bool
, 0, &arg_default_ip_accounting
},
763 { "Manager", "DefaultBlockIOAccounting", config_parse_bool
, 0, &arg_default_blockio_accounting
},
764 { "Manager", "DefaultMemoryAccounting", config_parse_bool
, 0, &arg_default_memory_accounting
},
765 { "Manager", "DefaultTasksAccounting", config_parse_bool
, 0, &arg_default_tasks_accounting
},
766 { "Manager", "DefaultTasksMax", config_parse_tasks_max
, 0, &arg_default_tasks_max
},
767 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action
, 0, &arg_cad_burst_action
},
768 { "Manager", "DefaultOOMPolicy", config_parse_oom_policy
, 0, &arg_default_oom_policy
},
772 const char *fn
, *conf_dirs_nulstr
;
775 PKGSYSCONFDIR
"/system.conf" :
776 PKGSYSCONFDIR
"/user.conf";
778 conf_dirs_nulstr
= arg_system
?
779 CONF_PATHS_NULSTR("systemd/system.conf.d") :
780 CONF_PATHS_NULSTR("systemd/user.conf.d");
782 (void) config_parse_many_nulstr(fn
, conf_dirs_nulstr
, "Manager\0", config_item_table_lookup
, items
, CONFIG_PARSE_WARN
, NULL
);
784 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
785 * like everywhere else. */
786 if (arg_default_timeout_start_usec
<= 0)
787 arg_default_timeout_start_usec
= USEC_INFINITY
;
788 if (arg_default_timeout_stop_usec
<= 0)
789 arg_default_timeout_stop_usec
= USEC_INFINITY
;
794 static void set_manager_defaults(Manager
*m
) {
798 /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
799 * affect the manager itself, but are just what newly allocated units will have set if they haven't set
800 * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
802 m
->default_timer_accuracy_usec
= arg_default_timer_accuracy_usec
;
803 m
->default_std_output
= arg_default_std_output
;
804 m
->default_std_error
= arg_default_std_error
;
805 m
->default_timeout_start_usec
= arg_default_timeout_start_usec
;
806 m
->default_timeout_stop_usec
= arg_default_timeout_stop_usec
;
807 m
->default_timeout_abort_usec
= arg_default_timeout_abort_usec
;
808 m
->default_timeout_abort_set
= arg_default_timeout_abort_set
;
809 m
->default_restart_usec
= arg_default_restart_usec
;
810 m
->default_start_limit_interval
= arg_default_start_limit_interval
;
811 m
->default_start_limit_burst
= arg_default_start_limit_burst
;
813 /* On 4.15+ with unified hierarchy, CPU accounting is essentially free as it doesn't require the CPU
814 * controller to be enabled, so the default is to enable it unless we got told otherwise. */
815 if (arg_default_cpu_accounting
>= 0)
816 m
->default_cpu_accounting
= arg_default_cpu_accounting
;
818 m
->default_cpu_accounting
= cpu_accounting_is_cheap();
820 m
->default_io_accounting
= arg_default_io_accounting
;
821 m
->default_ip_accounting
= arg_default_ip_accounting
;
822 m
->default_blockio_accounting
= arg_default_blockio_accounting
;
823 m
->default_memory_accounting
= arg_default_memory_accounting
;
824 m
->default_tasks_accounting
= arg_default_tasks_accounting
;
825 m
->default_tasks_max
= arg_default_tasks_max
;
826 m
->default_oom_policy
= arg_default_oom_policy
;
828 (void) manager_set_default_rlimits(m
, arg_default_rlimit
);
830 (void) manager_default_environment(m
);
831 (void) manager_transient_environment_add(m
, arg_default_environment
);
834 static void set_manager_settings(Manager
*m
) {
838 /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
839 * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
841 m
->confirm_spawn
= arg_confirm_spawn
;
842 m
->service_watchdogs
= arg_service_watchdogs
;
843 m
->runtime_watchdog
= arg_runtime_watchdog
;
844 m
->shutdown_watchdog
= arg_shutdown_watchdog
;
845 m
->cad_burst_action
= arg_cad_burst_action
;
847 manager_set_show_status(m
, arg_show_status
);
850 static int parse_argv(int argc
, char *argv
[]) {
852 ARG_LOG_LEVEL
= 0x100,
862 ARG_DUMP_CONFIGURATION_ITEMS
,
863 ARG_DUMP_BUS_PROPERTIES
,
872 ARG_DEFAULT_STD_OUTPUT
,
873 ARG_DEFAULT_STD_ERROR
,
875 ARG_SERVICE_WATCHDOGS
,
878 static const struct option options
[] = {
879 { "log-level", required_argument
, NULL
, ARG_LOG_LEVEL
},
880 { "log-target", required_argument
, NULL
, ARG_LOG_TARGET
},
881 { "log-color", optional_argument
, NULL
, ARG_LOG_COLOR
},
882 { "log-location", optional_argument
, NULL
, ARG_LOG_LOCATION
},
883 { "unit", required_argument
, NULL
, ARG_UNIT
},
884 { "system", no_argument
, NULL
, ARG_SYSTEM
},
885 { "user", no_argument
, NULL
, ARG_USER
},
886 { "test", no_argument
, NULL
, ARG_TEST
},
887 { "no-pager", no_argument
, NULL
, ARG_NO_PAGER
},
888 { "help", no_argument
, NULL
, 'h' },
889 { "version", no_argument
, NULL
, ARG_VERSION
},
890 { "dump-configuration-items", no_argument
, NULL
, ARG_DUMP_CONFIGURATION_ITEMS
},
891 { "dump-bus-properties", no_argument
, NULL
, ARG_DUMP_BUS_PROPERTIES
},
892 { "dump-core", optional_argument
, NULL
, ARG_DUMP_CORE
},
893 { "crash-chvt", required_argument
, NULL
, ARG_CRASH_CHVT
},
894 { "crash-shell", optional_argument
, NULL
, ARG_CRASH_SHELL
},
895 { "crash-reboot", optional_argument
, NULL
, ARG_CRASH_REBOOT
},
896 { "confirm-spawn", optional_argument
, NULL
, ARG_CONFIRM_SPAWN
},
897 { "show-status", optional_argument
, NULL
, ARG_SHOW_STATUS
},
898 { "deserialize", required_argument
, NULL
, ARG_DESERIALIZE
},
899 { "switched-root", no_argument
, NULL
, ARG_SWITCHED_ROOT
},
900 { "default-standard-output", required_argument
, NULL
, ARG_DEFAULT_STD_OUTPUT
, },
901 { "default-standard-error", required_argument
, NULL
, ARG_DEFAULT_STD_ERROR
, },
902 { "machine-id", required_argument
, NULL
, ARG_MACHINE_ID
},
903 { "service-watchdogs", required_argument
, NULL
, ARG_SERVICE_WATCHDOGS
},
912 if (getpid_cached() == 1)
915 while ((c
= getopt_long(argc
, argv
, "hDbsz:", options
, NULL
)) >= 0)
920 r
= log_set_max_level_from_string(optarg
);
922 return log_error_errno(r
, "Failed to parse log level \"%s\": %m", optarg
);
927 r
= log_set_target_from_string(optarg
);
929 return log_error_errno(r
, "Failed to parse log target \"%s\": %m", optarg
);
936 r
= log_show_color_from_string(optarg
);
938 return log_error_errno(r
, "Failed to parse log color setting \"%s\": %m",
941 log_show_color(true);
945 case ARG_LOG_LOCATION
:
947 r
= log_show_location_from_string(optarg
);
949 return log_error_errno(r
, "Failed to parse log location setting \"%s\": %m",
952 log_show_location(true);
956 case ARG_DEFAULT_STD_OUTPUT
:
957 r
= exec_output_from_string(optarg
);
959 return log_error_errno(r
, "Failed to parse default standard output setting \"%s\": %m",
961 arg_default_std_output
= r
;
964 case ARG_DEFAULT_STD_ERROR
:
965 r
= exec_output_from_string(optarg
);
967 return log_error_errno(r
, "Failed to parse default standard error output setting \"%s\": %m",
969 arg_default_std_error
= r
;
973 r
= free_and_strdup(&arg_default_unit
, optarg
);
975 return log_error_errno(r
, "Failed to set default unit \"%s\": %m", optarg
);
988 arg_action
= ACTION_TEST
;
992 arg_pager_flags
|= PAGER_DISABLE
;
996 arg_action
= ACTION_VERSION
;
999 case ARG_DUMP_CONFIGURATION_ITEMS
:
1000 arg_action
= ACTION_DUMP_CONFIGURATION_ITEMS
;
1003 case ARG_DUMP_BUS_PROPERTIES
:
1004 arg_action
= ACTION_DUMP_BUS_PROPERTIES
;
1009 arg_dump_core
= true;
1011 r
= parse_boolean(optarg
);
1013 return log_error_errno(r
, "Failed to parse dump core boolean: \"%s\": %m",
1019 case ARG_CRASH_CHVT
:
1020 r
= parse_crash_chvt(optarg
);
1022 return log_error_errno(r
, "Failed to parse crash virtual terminal index: \"%s\": %m",
1026 case ARG_CRASH_SHELL
:
1028 arg_crash_shell
= true;
1030 r
= parse_boolean(optarg
);
1032 return log_error_errno(r
, "Failed to parse crash shell boolean: \"%s\": %m",
1034 arg_crash_shell
= r
;
1038 case ARG_CRASH_REBOOT
:
1040 arg_crash_reboot
= true;
1042 r
= parse_boolean(optarg
);
1044 return log_error_errno(r
, "Failed to parse crash shell boolean: \"%s\": %m",
1046 arg_crash_reboot
= r
;
1050 case ARG_CONFIRM_SPAWN
:
1051 arg_confirm_spawn
= mfree(arg_confirm_spawn
);
1053 r
= parse_confirm_spawn(optarg
, &arg_confirm_spawn
);
1055 return log_error_errno(r
, "Failed to parse confirm spawn option: \"%s\": %m",
1059 case ARG_SERVICE_WATCHDOGS
:
1060 r
= parse_boolean(optarg
);
1062 return log_error_errno(r
, "Failed to parse service watchdogs boolean: \"%s\": %m",
1064 arg_service_watchdogs
= r
;
1067 case ARG_SHOW_STATUS
:
1069 r
= parse_show_status(optarg
, &arg_show_status
);
1071 return log_error_errno(r
, "Failed to parse show status boolean: \"%s\": %m",
1074 arg_show_status
= SHOW_STATUS_YES
;
1077 case ARG_DESERIALIZE
: {
1081 r
= safe_atoi(optarg
, &fd
);
1083 log_error_errno(r
, "Failed to parse deserialize option \"%s\": %m", optarg
);
1085 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
),
1086 "Invalid deserialize fd: %d",
1089 (void) fd_cloexec(fd
, true);
1091 f
= fdopen(fd
, "r");
1093 return log_error_errno(errno
, "Failed to open serialization fd %d: %m", fd
);
1095 safe_fclose(arg_serialization
);
1096 arg_serialization
= f
;
1101 case ARG_SWITCHED_ROOT
:
1102 arg_switched_root
= true;
1105 case ARG_MACHINE_ID
:
1106 r
= set_machine_id(optarg
);
1108 return log_error_errno(r
, "MachineID '%s' is not valid: %m", optarg
);
1112 arg_action
= ACTION_HELP
;
1116 log_set_max_level(LOG_DEBUG
);
1122 /* Just to eat away the sysvinit kernel
1123 * cmdline args without getopt() error
1124 * messages that we'll parse in
1125 * parse_proc_cmdline_word() or ignore. */
1128 if (getpid_cached() != 1)
1134 assert_not_reached("Unhandled option code.");
1137 if (optind
< argc
&& getpid_cached() != 1) {
1138 /* Hmm, when we aren't run as init system
1139 * let's complain about excess arguments */
1141 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
),
1142 "Excess arguments.");
1148 static int help(void) {
1149 _cleanup_free_
char *link
= NULL
;
1152 r
= terminal_urlify_man("systemd", "1", &link
);
1156 printf("%s [OPTIONS...]\n\n"
1157 "Starts up and maintains the system or user services.\n\n"
1158 " -h --help Show this help\n"
1159 " --version Show version\n"
1160 " --test Determine startup sequence, dump it and exit\n"
1161 " --no-pager Do not pipe output into a pager\n"
1162 " --dump-configuration-items Dump understood unit configuration items\n"
1163 " --dump-bus-properties Dump exposed bus properties\n"
1164 " --unit=UNIT Set default unit\n"
1165 " --system Run a system instance, even if PID != 1\n"
1166 " --user Run a user instance\n"
1167 " --dump-core[=BOOL] Dump core on crash\n"
1168 " --crash-vt=NR Change to specified VT on crash\n"
1169 " --crash-reboot[=BOOL] Reboot on crash\n"
1170 " --crash-shell[=BOOL] Run shell on crash\n"
1171 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1172 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1173 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1174 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1175 " --log-color[=BOOL] Highlight important log messages\n"
1176 " --log-location[=BOOL] Include code location in log messages\n"
1177 " --default-standard-output= Set default standard output for services\n"
1178 " --default-standard-error= Set default standard error output for services\n"
1179 "\nSee the %s for details.\n"
1180 , program_invocation_short_name
1187 static int prepare_reexecute(
1191 bool switching_root
) {
1193 _cleanup_fdset_free_ FDSet
*fds
= NULL
;
1194 _cleanup_fclose_
FILE *f
= NULL
;
1201 r
= manager_open_serialization(m
, &f
);
1203 return log_error_errno(r
, "Failed to create serialization file: %m");
1205 /* Make sure nothing is really destructed when we shut down */
1207 bus_manager_send_reloading(m
, true);
1213 r
= manager_serialize(m
, f
, fds
, switching_root
);
1217 if (fseeko(f
, 0, SEEK_SET
) == (off_t
) -1)
1218 return log_error_errno(errno
, "Failed to rewind serialization fd: %m");
1220 r
= fd_cloexec(fileno(f
), false);
1222 return log_error_errno(r
, "Failed to disable O_CLOEXEC for serialization: %m");
1224 r
= fdset_cloexec(fds
, false);
1226 return log_error_errno(r
, "Failed to disable O_CLOEXEC for serialization fds: %m");
1228 *ret_f
= TAKE_PTR(f
);
1229 *ret_fds
= TAKE_PTR(fds
);
1234 static void bump_file_max_and_nr_open(void) {
1236 /* Let's bump fs.file-max and fs.nr_open to their respective maximums. On current kernels large numbers of file
1237 * descriptors are no longer a performance problem and their memory is properly tracked by memcg, thus counting
1238 * them and limiting them in another two layers of limits is unnecessary and just complicates things. This
1239 * function hence turns off 2 of the 4 levels of limits on file descriptors, and makes RLIMIT_NOLIMIT (soft +
1240 * hard) the only ones that really matter. */
1242 #if BUMP_PROC_SYS_FS_FILE_MAX || BUMP_PROC_SYS_FS_NR_OPEN
1243 _cleanup_free_
char *t
= NULL
;
1247 #if BUMP_PROC_SYS_FS_FILE_MAX
1248 /* I so wanted to use STRINGIFY(ULONG_MAX) here, but alas we can't as glibc/gcc define that as
1249 * "(0x7fffffffffffffffL * 2UL + 1UL)". Seriously. 😢 */
1250 if (asprintf(&t
, "%lu\n", ULONG_MAX
) < 0) {
1255 r
= sysctl_write("fs/file-max", t
);
1257 log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
, "Failed to bump fs.file-max, ignoring: %m");
1260 #if BUMP_PROC_SYS_FS_FILE_MAX && BUMP_PROC_SYS_FS_NR_OPEN
1264 #if BUMP_PROC_SYS_FS_NR_OPEN
1267 /* Arg! The kernel enforces maximum and minimum values on the fs.nr_open, but we don't really know what they
1268 * are. The expression by which the maximum is determined is dependent on the architecture, and is something we
1269 * don't really want to copy to userspace, as it is dependent on implementation details of the kernel. Since
1270 * the kernel doesn't expose the maximum value to us, we can only try and hope. Hence, let's start with
1271 * INT_MAX, and then keep halving the value until we find one that works. Ugly? Yes, absolutely, but kernel
1272 * APIs are kernel APIs, so what do can we do... 🤯 */
1277 v
&= ~(__SIZEOF_POINTER__
- 1); /* Round down to next multiple of the pointer size */
1279 log_warning("Can't bump fs.nr_open, value too small.");
1285 log_error_errno(k
, "Failed to read fs.nr_open: %m");
1288 if (k
>= v
) { /* Already larger */
1289 log_debug("Skipping bump, value is already larger.");
1293 if (asprintf(&t
, "%i\n", v
) < 0) {
1298 r
= sysctl_write("fs/nr_open", t
);
1301 log_debug("Couldn't write fs.nr_open as %i, halving it.", v
);
1306 log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
, "Failed to bump fs.nr_open, ignoring: %m");
1310 log_debug("Successfully bumped fs.nr_open to %i", v
);
1316 static int bump_rlimit_nofile(struct rlimit
*saved_rlimit
) {
1317 struct rlimit new_rlimit
;
1320 assert(saved_rlimit
);
1322 /* Save the original RLIMIT_NOFILE so that we can reset it later when transitioning from the initrd to the main
1323 * systemd or suchlike. */
1324 if (getrlimit(RLIMIT_NOFILE
, saved_rlimit
) < 0)
1325 return log_warning_errno(errno
, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1327 /* Get the underlying absolute limit the kernel enforces */
1328 nr
= read_nr_open();
1330 /* Make sure forked processes get limits based on the original kernel setting */
1331 if (!arg_default_rlimit
[RLIMIT_NOFILE
]) {
1334 rl
= newdup(struct rlimit
, saved_rlimit
, 1);
1338 /* Bump the hard limit for system services to a substantially higher value. The default hard limit
1339 * current kernels set is pretty low (4K), mostly for historical reasons. According to kernel
1340 * developers, the fd handling in recent kernels has been optimized substantially enough, so that we
1341 * can bump the limit now, without paying too high a price in memory or performance. Note however that
1342 * we only bump the hard limit, not the soft limit. That's because select() works the way it works, and
1343 * chokes on fds >= 1024. If we'd bump the soft limit globally, it might accidentally happen to
1344 * unexpecting programs that they get fds higher than what they can process using select(). By only
1345 * bumping the hard limit but leaving the low limit as it is we avoid this pitfall: programs that are
1346 * written by folks aware of the select() problem in mind (and thus use poll()/epoll instead of
1347 * select(), the way everybody should) can explicitly opt into high fds by bumping their soft limit
1348 * beyond 1024, to the hard limit we pass. */
1350 rl
->rlim_max
= MIN((rlim_t
) nr
, MAX(rl
->rlim_max
, (rlim_t
) HIGH_RLIMIT_NOFILE
));
1352 /* If for some reason we were invoked with a soft limit above 1024 (which should never
1353 * happen!, but who knows what we get passed in from pam_limit when invoked as --user
1354 * instance), then lower what we pass on to not confuse our children */
1355 rl
->rlim_cur
= MIN(rl
->rlim_cur
, (rlim_t
) FD_SETSIZE
);
1357 arg_default_rlimit
[RLIMIT_NOFILE
] = rl
;
1360 /* Calculate the new limits to use for us. Never lower from what we inherited. */
1361 new_rlimit
= (struct rlimit
) {
1362 .rlim_cur
= MAX((rlim_t
) nr
, saved_rlimit
->rlim_cur
),
1363 .rlim_max
= MAX((rlim_t
) nr
, saved_rlimit
->rlim_max
),
1366 /* Shortcut if nothing changes. */
1367 if (saved_rlimit
->rlim_max
>= new_rlimit
.rlim_max
&&
1368 saved_rlimit
->rlim_cur
>= new_rlimit
.rlim_cur
) {
1369 log_debug("RLIMIT_NOFILE is already as high or higher than we need it, not bumping.");
1373 /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for
1374 * both hard and soft. */
1375 r
= setrlimit_closest(RLIMIT_NOFILE
, &new_rlimit
);
1377 return log_warning_errno(r
, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1382 static int bump_rlimit_memlock(struct rlimit
*saved_rlimit
) {
1383 struct rlimit new_rlimit
;
1386 assert(saved_rlimit
);
1388 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK which should
1389 * normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's bump
1390 * the value high enough for our user. */
1392 if (getrlimit(RLIMIT_MEMLOCK
, saved_rlimit
) < 0)
1393 return log_warning_errno(errno
, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1395 /* Pass the original value down to invoked processes */
1396 if (!arg_default_rlimit
[RLIMIT_MEMLOCK
]) {
1399 rl
= newdup(struct rlimit
, saved_rlimit
, 1);
1403 arg_default_rlimit
[RLIMIT_MEMLOCK
] = rl
;
1406 /* Using MAX() on resource limits only is safe if RLIM_INFINITY is > 0. POSIX declares that rlim_t
1407 * must be unsigned, hence this is a given, but let's make this clear here. */
1408 assert_cc(RLIM_INFINITY
> 0);
1410 new_rlimit
= (struct rlimit
) {
1411 .rlim_cur
= MAX(HIGH_RLIMIT_MEMLOCK
, saved_rlimit
->rlim_cur
),
1412 .rlim_max
= MAX(HIGH_RLIMIT_MEMLOCK
, saved_rlimit
->rlim_max
),
1415 if (saved_rlimit
->rlim_max
>= new_rlimit
.rlim_cur
&&
1416 saved_rlimit
->rlim_cur
>= new_rlimit
.rlim_max
) {
1417 log_debug("RLIMIT_MEMLOCK is already as high or higher than we need it, not bumping.");
1421 r
= setrlimit_closest(RLIMIT_MEMLOCK
, &new_rlimit
);
1423 return log_warning_errno(r
, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1428 static void test_usr(void) {
1430 /* Check that /usr is either on the same file system as / or mounted already. */
1432 if (dir_is_empty("/usr") <= 0)
1435 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1436 "Some things will probably break (sometimes even silently) in mysterious ways. "
1437 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1440 static int enforce_syscall_archs(Set
*archs
) {
1444 if (!is_seccomp_available())
1447 r
= seccomp_restrict_archs(arg_syscall_archs
);
1449 return log_error_errno(r
, "Failed to enforce system call architecture restrication: %m");
1454 static int status_welcome(void) {
1455 _cleanup_free_
char *pretty_name
= NULL
, *ansi_color
= NULL
;
1458 if (IN_SET(arg_show_status
, SHOW_STATUS_NO
, SHOW_STATUS_AUTO
))
1461 r
= parse_os_release(NULL
,
1462 "PRETTY_NAME", &pretty_name
,
1463 "ANSI_COLOR", &ansi_color
,
1466 log_full_errno(r
== -ENOENT
? LOG_DEBUG
: LOG_WARNING
, r
,
1467 "Failed to read os-release file, ignoring: %m");
1469 if (log_get_show_color())
1470 return status_printf(NULL
, 0,
1471 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1472 isempty(ansi_color
) ? "1" : ansi_color
,
1473 isempty(pretty_name
) ? "Linux" : pretty_name
);
1475 return status_printf(NULL
, 0,
1476 "\nWelcome to %s!\n",
1477 isempty(pretty_name
) ? "Linux" : pretty_name
);
1480 static int write_container_id(void) {
1484 c
= getenv("container");
1488 RUN_WITH_UMASK(0022)
1489 r
= write_string_file("/run/systemd/container", c
, WRITE_STRING_FILE_CREATE
);
1491 return log_warning_errno(r
, "Failed to write /run/systemd/container, ignoring: %m");
1496 static int bump_unix_max_dgram_qlen(void) {
1497 _cleanup_free_
char *qlen
= NULL
;
1501 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1502 * really really early during boot, so that it is actually applied to all our sockets, including the
1503 * $NOTIFY_SOCKET one. */
1505 r
= read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen
);
1507 return log_full_errno(r
== -ENOENT
? LOG_DEBUG
: LOG_WARNING
, r
, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1509 r
= safe_atolu(qlen
, &v
);
1511 return log_warning_errno(r
, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen
);
1513 if (v
>= DEFAULT_UNIX_MAX_DGRAM_QLEN
)
1516 r
= write_string_filef("/proc/sys/net/unix/max_dgram_qlen", WRITE_STRING_FILE_DISABLE_BUFFER
, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN
);
1518 return log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
,
1519 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1524 static int fixup_environment(void) {
1525 _cleanup_free_
char *term
= NULL
;
1529 /* Only fix up the environment when we are started as PID 1 */
1530 if (getpid_cached() != 1)
1533 /* We expect the environment to be set correctly if run inside a container. */
1534 if (detect_container() > 0)
1537 /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1538 * device used by the console. We try to make a better guess here since some consoles might not have support
1539 * for color mode for example.
1541 * However if TERM was configured through the kernel command line then leave it alone. */
1542 r
= proc_cmdline_get_key("TERM", 0, &term
);
1546 t
= term
?: default_term_for_tty("/dev/console");
1548 if (setenv("TERM", t
, 1) < 0)
1554 static void redirect_telinit(int argc
, char *argv
[]) {
1556 /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1558 #if HAVE_SYSV_COMPAT
1559 if (getpid_cached() == 1)
1562 if (!strstr(program_invocation_short_name
, "init"))
1565 execv(SYSTEMCTL_BINARY_PATH
, argv
);
1566 log_error_errno(errno
, "Failed to exec " SYSTEMCTL_BINARY_PATH
": %m");
1571 static int become_shutdown(
1572 const char *shutdown_verb
,
1575 char log_level
[DECIMAL_STR_MAX(int) + 1],
1576 exit_code
[DECIMAL_STR_MAX(uint8_t) + 1],
1577 timeout
[DECIMAL_STR_MAX(usec_t
) + 1];
1579 const char* command_line
[13] = {
1580 SYSTEMD_SHUTDOWN_BINARY_PATH
,
1582 "--timeout", timeout
,
1583 "--log-level", log_level
,
1587 _cleanup_strv_free_
char **env_block
= NULL
;
1591 assert(shutdown_verb
);
1592 assert(!command_line
[pos
]);
1593 env_block
= strv_copy(environ
);
1595 xsprintf(log_level
, "%d", log_get_max_level());
1596 xsprintf(timeout
, "%" PRI_USEC
"us", arg_default_timeout_stop_usec
);
1598 switch (log_get_target()) {
1600 case LOG_TARGET_KMSG
:
1601 case LOG_TARGET_JOURNAL_OR_KMSG
:
1602 case LOG_TARGET_SYSLOG_OR_KMSG
:
1603 command_line
[pos
++] = "kmsg";
1606 case LOG_TARGET_NULL
:
1607 command_line
[pos
++] = "null";
1610 case LOG_TARGET_CONSOLE
:
1612 command_line
[pos
++] = "console";
1616 if (log_get_show_color())
1617 command_line
[pos
++] = "--log-color";
1619 if (log_get_show_location())
1620 command_line
[pos
++] = "--log-location";
1622 if (streq(shutdown_verb
, "exit")) {
1623 command_line
[pos
++] = "--exit-code";
1624 command_line
[pos
++] = exit_code
;
1625 xsprintf(exit_code
, "%d", retval
);
1628 assert(pos
< ELEMENTSOF(command_line
));
1630 if (streq(shutdown_verb
, "reboot") &&
1631 arg_shutdown_watchdog
> 0 &&
1632 arg_shutdown_watchdog
!= USEC_INFINITY
) {
1636 /* If we reboot let's set the shutdown
1637 * watchdog and tell the shutdown binary to
1638 * repeatedly ping it */
1639 r
= watchdog_set_timeout(&arg_shutdown_watchdog
);
1640 watchdog_close(r
< 0);
1642 /* Tell the binary how often to ping, ignore failure */
1643 if (asprintf(&e
, "WATCHDOG_USEC="USEC_FMT
, arg_shutdown_watchdog
) > 0)
1644 (void) strv_consume(&env_block
, e
);
1646 if (arg_watchdog_device
&&
1647 asprintf(&e
, "WATCHDOG_DEVICE=%s", arg_watchdog_device
) > 0)
1648 (void) strv_consume(&env_block
, e
);
1650 watchdog_close(true);
1652 /* Avoid the creation of new processes forked by the
1653 * kernel; at this point, we will not listen to the
1655 if (detect_container() <= 0)
1656 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER
);
1658 execve(SYSTEMD_SHUTDOWN_BINARY_PATH
, (char **) command_line
, env_block
);
1662 static void initialize_clock(void) {
1665 if (clock_is_localtime(NULL
) > 0) {
1669 * The very first call of settimeofday() also does a time warp in the kernel.
1671 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1672 * of maintaining the RTC and do all adjustments. This matches the behavior of Windows, which leaves
1673 * the RTC alone if the registry tells that the RTC runs in UTC.
1675 r
= clock_set_timezone(&min
);
1677 log_error_errno(r
, "Failed to apply local time delta, ignoring: %m");
1679 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min
);
1681 } else if (!in_initrd()) {
1683 * Do a dummy very first call to seal the kernel's time warp magic.
1685 * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1686 * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1687 * until we reach the real system.
1689 * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1690 * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1691 * be treated as UTC that way.
1693 (void) clock_reset_timewarp();
1696 r
= clock_apply_epoch();
1698 log_error_errno(r
, "Current system time is before build time, but cannot correct: %m");
1700 log_info("System time before build time, advancing clock.");
1703 static void initialize_coredump(bool skip_setup
) {
1705 if (getpid_cached() != 1)
1708 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1709 * will process core dumps for system services by default. */
1710 if (setrlimit(RLIMIT_CORE
, &RLIMIT_MAKE_CONST(RLIM_INFINITY
)) < 0)
1711 log_warning_errno(errno
, "Failed to set RLIMIT_CORE: %m");
1713 /* But at the same time, turn off the core_pattern logic by default, so that no
1714 * coredumps are stored until the systemd-coredump tool is enabled via
1715 * sysctl. However it can be changed via the kernel command line later so core
1716 * dumps can still be generated during early startup and in initramfs. */
1718 disable_coredumps();
1722 static void initialize_core_pattern(bool skip_setup
) {
1725 if (skip_setup
|| !arg_early_core_pattern
)
1728 if (getpid_cached() != 1)
1731 r
= write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern
, WRITE_STRING_FILE_DISABLE_BUFFER
);
1733 log_warning_errno(r
, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", arg_early_core_pattern
);
1736 static void update_cpu_affinity(bool skip_setup
) {
1737 _cleanup_free_
char *mask
= NULL
;
1739 if (skip_setup
|| !arg_cpu_affinity
.set
)
1742 assert(arg_cpu_affinity
.allocated
> 0);
1744 mask
= cpu_set_to_string(&arg_cpu_affinity
);
1745 log_debug("Setting CPU affinity to %s.", strnull(mask
));
1747 if (sched_setaffinity(0, arg_cpu_affinity
.allocated
, arg_cpu_affinity
.set
) < 0)
1748 log_warning_errno(errno
, "Failed to set CPU affinity: %m");
1751 static void do_reexecute(
1754 const struct rlimit
*saved_rlimit_nofile
,
1755 const struct rlimit
*saved_rlimit_memlock
,
1757 const char *switch_root_dir
,
1758 const char *switch_root_init
,
1759 const char **ret_error_message
) {
1761 unsigned i
, j
, args_size
;
1765 assert(saved_rlimit_nofile
);
1766 assert(saved_rlimit_memlock
);
1767 assert(ret_error_message
);
1769 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1771 watchdog_close(true);
1773 /* Reset RLIMIT_NOFILE + RLIMIT_MEMLOCK back to the kernel defaults, so that the new systemd can pass
1774 * the kernel default to its child processes */
1775 if (saved_rlimit_nofile
->rlim_cur
!= 0)
1776 (void) setrlimit(RLIMIT_NOFILE
, saved_rlimit_nofile
);
1777 if (saved_rlimit_memlock
->rlim_cur
!= RLIM_INFINITY
)
1778 (void) setrlimit(RLIMIT_MEMLOCK
, saved_rlimit_memlock
);
1780 if (switch_root_dir
) {
1781 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1782 * SIGCHLD for them after deserializing. */
1783 broadcast_signal(SIGTERM
, false, true, arg_default_timeout_stop_usec
);
1785 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1786 r
= switch_root(switch_root_dir
, "/mnt", true, MS_MOVE
);
1788 log_error_errno(r
, "Failed to switch root, trying to continue: %m");
1791 args_size
= MAX(6, argc
+1);
1792 args
= newa(const char*, args_size
);
1794 if (!switch_root_init
) {
1795 char sfd
[DECIMAL_STR_MAX(int) + 1];
1797 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1798 * the user didn't specify an explicit init to spawn. */
1800 assert(arg_serialization
);
1803 xsprintf(sfd
, "%i", fileno(arg_serialization
));
1806 args
[i
++] = SYSTEMD_BINARY_PATH
;
1807 if (switch_root_dir
)
1808 args
[i
++] = "--switched-root";
1809 args
[i
++] = arg_system
? "--system" : "--user";
1810 args
[i
++] = "--deserialize";
1814 assert(i
<= args_size
);
1817 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do this is on
1818 * its own on exec(), but it will do it on exit(). Hence, to ensure we get a summary here, fork() off
1819 * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1820 * before proceeding into the exec().
1822 valgrind_summary_hack();
1824 (void) execv(args
[0], (char* const*) args
);
1825 log_debug_errno(errno
, "Failed to execute our own binary, trying fallback: %m");
1828 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1829 * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1830 * doesn't matter.) */
1832 arg_serialization
= safe_fclose(arg_serialization
);
1833 fds
= fdset_free(fds
);
1835 /* Reopen the console */
1836 (void) make_console_stdio();
1838 for (j
= 1, i
= 1; j
< (unsigned) argc
; j
++)
1839 args
[i
++] = argv
[j
];
1841 assert(i
<= args_size
);
1843 /* Re-enable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1844 (void) reset_all_signal_handlers();
1845 (void) reset_signal_mask();
1846 (void) rlimit_nofile_safe();
1848 if (switch_root_init
) {
1849 args
[0] = switch_root_init
;
1850 (void) execv(args
[0], (char* const*) args
);
1851 log_warning_errno(errno
, "Failed to execute configured init, trying fallback: %m");
1854 args
[0] = "/sbin/init";
1855 (void) execv(args
[0], (char* const*) args
);
1858 manager_status_printf(NULL
, STATUS_TYPE_EMERGENCY
,
1859 ANSI_HIGHLIGHT_RED
" !! " ANSI_NORMAL
,
1860 "Failed to execute /sbin/init");
1863 log_warning("No /sbin/init, trying fallback");
1865 args
[0] = "/bin/sh";
1867 (void) execv(args
[0], (char* const*) args
);
1868 log_error_errno(errno
, "Failed to execute /bin/sh, giving up: %m");
1870 log_warning_errno(r
, "Failed to execute /sbin/init, giving up: %m");
1872 *ret_error_message
= "Failed to execute fallback shell";
1875 static int invoke_main_loop(
1877 bool *ret_reexecute
,
1878 int *ret_retval
, /* Return parameters relevant for shutting down */
1879 const char **ret_shutdown_verb
, /* … */
1880 FDSet
**ret_fds
, /* Return parameters for reexecuting */
1881 char **ret_switch_root_dir
, /* … */
1882 char **ret_switch_root_init
, /* … */
1883 const char **ret_error_message
) {
1888 assert(ret_reexecute
);
1890 assert(ret_shutdown_verb
);
1892 assert(ret_switch_root_dir
);
1893 assert(ret_switch_root_init
);
1894 assert(ret_error_message
);
1897 r
= manager_loop(m
);
1899 *ret_error_message
= "Failed to run main loop";
1900 return log_emergency_errno(r
, "Failed to run main loop: %m");
1903 switch ((ManagerObjective
) r
) {
1905 case MANAGER_RELOAD
: {
1906 LogTarget saved_log_target
;
1907 int saved_log_level
;
1909 log_info("Reloading.");
1911 /* First, save any overridden log level/target, then parse the configuration file, which might
1912 * change the log level to new settings. */
1914 saved_log_level
= m
->log_level_overridden
? log_get_max_level() : -1;
1915 saved_log_target
= m
->log_target_overridden
? log_get_target() : _LOG_TARGET_INVALID
;
1917 (void) parse_configuration();
1919 set_manager_defaults(m
);
1921 update_cpu_affinity(false);
1923 if (saved_log_level
>= 0)
1924 manager_override_log_level(m
, saved_log_level
);
1925 if (saved_log_target
>= 0)
1926 manager_override_log_target(m
, saved_log_target
);
1928 r
= manager_reload(m
);
1930 /* Reloading failed before the point of no return. Let's continue running as if nothing happened. */
1931 m
->objective
= MANAGER_OK
;
1936 case MANAGER_REEXECUTE
:
1938 r
= prepare_reexecute(m
, &arg_serialization
, ret_fds
, false);
1940 *ret_error_message
= "Failed to prepare for reexecution";
1944 log_notice("Reexecuting.");
1946 *ret_reexecute
= true;
1947 *ret_retval
= EXIT_SUCCESS
;
1948 *ret_shutdown_verb
= NULL
;
1949 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
1953 case MANAGER_SWITCH_ROOT
:
1954 if (!m
->switch_root_init
) {
1955 r
= prepare_reexecute(m
, &arg_serialization
, ret_fds
, true);
1957 *ret_error_message
= "Failed to prepare for reexecution";
1963 log_notice("Switching root.");
1965 *ret_reexecute
= true;
1966 *ret_retval
= EXIT_SUCCESS
;
1967 *ret_shutdown_verb
= NULL
;
1969 /* Steal the switch root parameters */
1970 *ret_switch_root_dir
= TAKE_PTR(m
->switch_root
);
1971 *ret_switch_root_init
= TAKE_PTR(m
->switch_root_init
);
1977 if (MANAGER_IS_USER(m
)) {
1980 *ret_reexecute
= false;
1981 *ret_retval
= m
->return_value
;
1982 *ret_shutdown_verb
= NULL
;
1984 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
1990 case MANAGER_REBOOT
:
1991 case MANAGER_POWEROFF
:
1993 case MANAGER_KEXEC
: {
1994 static const char * const table
[_MANAGER_OBJECTIVE_MAX
] = {
1995 [MANAGER_EXIT
] = "exit",
1996 [MANAGER_REBOOT
] = "reboot",
1997 [MANAGER_POWEROFF
] = "poweroff",
1998 [MANAGER_HALT
] = "halt",
1999 [MANAGER_KEXEC
] = "kexec",
2002 log_notice("Shutting down.");
2004 *ret_reexecute
= false;
2005 *ret_retval
= m
->return_value
;
2006 assert_se(*ret_shutdown_verb
= table
[m
->objective
]);
2008 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
2014 assert_not_reached("Unknown or unexpected manager objective.");
2019 static void log_execution_mode(bool *ret_first_boot
) {
2020 assert(ret_first_boot
);
2025 log_info("systemd " GIT_VERSION
" running in %ssystem mode. (" SYSTEMD_FEATURES
")",
2026 arg_action
== ACTION_TEST
? "test " : "" );
2028 v
= detect_virtualization();
2030 log_info("Detected virtualization %s.", virtualization_to_string(v
));
2032 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
2035 *ret_first_boot
= false;
2036 log_info("Running in initial RAM disk.");
2038 /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
2039 * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
2040 * doesn't it's unpopulated. This allows container managers and installers to provision a
2041 * couple of files already. If the container manager wants to provision the machine ID itself
2042 * it should pass $container_uuid to PID 1. */
2044 *ret_first_boot
= access("/etc/machine-id", F_OK
) < 0;
2045 if (*ret_first_boot
)
2046 log_info("Running with unpopulated /etc.");
2049 if (DEBUG_LOGGING
) {
2050 _cleanup_free_
char *t
;
2052 t
= uid_to_name(getuid());
2053 log_debug("systemd " GIT_VERSION
" running in %suser mode for user " UID_FMT
"/%s. (" SYSTEMD_FEATURES
")",
2054 arg_action
== ACTION_TEST
? " test" : "", getuid(), strna(t
));
2057 *ret_first_boot
= false;
2061 static int initialize_runtime(
2063 struct rlimit
*saved_rlimit_nofile
,
2064 struct rlimit
*saved_rlimit_memlock
,
2065 const char **ret_error_message
) {
2069 assert(ret_error_message
);
2071 /* Sets up various runtime parameters. Many of these initializations are conditionalized:
2073 * - Some only apply to --system instances
2074 * - Some only apply to --user instances
2075 * - Some only apply when we first start up, but not when we reexecute
2078 if (arg_action
!= ACTION_RUN
)
2081 update_cpu_affinity(skip_setup
);
2084 /* Make sure we leave a core dump without panicking the kernel. */
2085 install_crash_handler();
2088 r
= mount_cgroup_controllers();
2090 *ret_error_message
= "Failed to mount cgroup hierarchies";
2096 machine_id_setup(NULL
, arg_machine_id
, NULL
);
2098 bump_unix_max_dgram_qlen();
2099 bump_file_max_and_nr_open();
2101 write_container_id();
2104 if (arg_watchdog_device
) {
2105 r
= watchdog_set_device(arg_watchdog_device
);
2107 log_warning_errno(r
, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device
);
2110 if (arg_runtime_watchdog
> 0 && arg_runtime_watchdog
!= USEC_INFINITY
)
2111 watchdog_set_timeout(&arg_runtime_watchdog
);
2114 if (arg_timer_slack_nsec
!= NSEC_INFINITY
)
2115 if (prctl(PR_SET_TIMERSLACK
, arg_timer_slack_nsec
) < 0)
2116 log_warning_errno(errno
, "Failed to adjust timer slack, ignoring: %m");
2118 if (arg_system
&& !cap_test_all(arg_capability_bounding_set
)) {
2119 r
= capability_bounding_set_drop_usermode(arg_capability_bounding_set
);
2121 *ret_error_message
= "Failed to drop capability bounding set of usermode helpers";
2122 return log_emergency_errno(r
, "Failed to drop capability bounding set of usermode helpers: %m");
2125 r
= capability_bounding_set_drop(arg_capability_bounding_set
, true);
2127 *ret_error_message
= "Failed to drop capability bounding set";
2128 return log_emergency_errno(r
, "Failed to drop capability bounding set: %m");
2132 if (arg_system
&& arg_no_new_privs
) {
2133 if (prctl(PR_SET_NO_NEW_PRIVS
, 1, 0, 0, 0) < 0) {
2134 *ret_error_message
= "Failed to disable new privileges";
2135 return log_emergency_errno(errno
, "Failed to disable new privileges: %m");
2139 if (arg_syscall_archs
) {
2140 r
= enforce_syscall_archs(arg_syscall_archs
);
2142 *ret_error_message
= "Failed to set syscall architectures";
2148 /* Become reaper of our children */
2149 if (prctl(PR_SET_CHILD_SUBREAPER
, 1) < 0)
2150 log_warning_errno(errno
, "Failed to make us a subreaper: %m");
2152 /* Bump up RLIMIT_NOFILE for systemd itself */
2153 (void) bump_rlimit_nofile(saved_rlimit_nofile
);
2154 (void) bump_rlimit_memlock(saved_rlimit_memlock
);
2159 static int do_queue_default_job(
2161 const char **ret_error_message
) {
2163 _cleanup_(sd_bus_error_free
) sd_bus_error error
= SD_BUS_ERROR_NULL
;
2164 Job
*default_unit_job
;
2165 Unit
*target
= NULL
;
2168 log_debug("Activating default unit: %s", arg_default_unit
);
2170 r
= manager_load_startable_unit_or_warn(m
, arg_default_unit
, NULL
, &target
);
2172 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET
);
2174 r
= manager_load_startable_unit_or_warn(m
, SPECIAL_RESCUE_TARGET
, NULL
, &target
);
2176 *ret_error_message
= r
== -ERFKILL
? "Rescue target masked"
2177 : "Failed to load rescue target";
2182 assert(target
->load_state
== UNIT_LOADED
);
2184 r
= manager_add_job(m
, JOB_START
, target
, JOB_ISOLATE
, NULL
, &error
, &default_unit_job
);
2186 log_debug_errno(r
, "Default target could not be isolated, starting instead: %s", bus_error_message(&error
, r
));
2188 sd_bus_error_free(&error
);
2190 r
= manager_add_job(m
, JOB_START
, target
, JOB_REPLACE
, NULL
, &error
, &default_unit_job
);
2192 *ret_error_message
= "Failed to start default target";
2193 return log_emergency_errno(r
, "Failed to start default target: %s", bus_error_message(&error
, r
));
2197 *ret_error_message
= "Failed to isolate default target";
2198 return log_emergency_errno(r
, "Failed to isolate default target: %s", bus_error_message(&error
, r
));
2201 m
->default_unit_job_id
= default_unit_job
->id
;
2206 static void reset_arguments(void) {
2207 /* Frees/resets arg_* variables, with a few exceptions commented below. */
2209 arg_default_unit
= mfree(arg_default_unit
);
2211 /* arg_system — ignore */
2213 arg_dump_core
= true;
2214 arg_crash_chvt
= -1;
2215 arg_crash_shell
= false;
2216 arg_crash_reboot
= false;
2217 arg_confirm_spawn
= mfree(arg_confirm_spawn
);
2218 arg_show_status
= _SHOW_STATUS_INVALID
;
2219 arg_switched_root
= false;
2220 arg_pager_flags
= 0;
2221 arg_service_watchdogs
= true;
2222 arg_default_std_output
= EXEC_OUTPUT_JOURNAL
;
2223 arg_default_std_error
= EXEC_OUTPUT_INHERIT
;
2224 arg_default_restart_usec
= DEFAULT_RESTART_USEC
;
2225 arg_default_timeout_start_usec
= DEFAULT_TIMEOUT_USEC
;
2226 arg_default_timeout_stop_usec
= DEFAULT_TIMEOUT_USEC
;
2227 arg_default_timeout_abort_usec
= DEFAULT_TIMEOUT_USEC
;
2228 arg_default_timeout_abort_set
= false;
2229 arg_default_start_limit_interval
= DEFAULT_START_LIMIT_INTERVAL
;
2230 arg_default_start_limit_burst
= DEFAULT_START_LIMIT_BURST
;
2231 arg_runtime_watchdog
= 0;
2232 arg_shutdown_watchdog
= 10 * USEC_PER_MINUTE
;
2233 arg_early_core_pattern
= NULL
;
2234 arg_watchdog_device
= NULL
;
2236 arg_default_environment
= strv_free(arg_default_environment
);
2237 rlimit_free_all(arg_default_rlimit
);
2239 arg_capability_bounding_set
= CAP_ALL
;
2240 arg_no_new_privs
= false;
2241 arg_timer_slack_nsec
= NSEC_INFINITY
;
2242 arg_default_timer_accuracy_usec
= 1 * USEC_PER_MINUTE
;
2244 arg_syscall_archs
= set_free(arg_syscall_archs
);
2246 /* arg_serialization — ignore */
2248 arg_default_cpu_accounting
= -1;
2249 arg_default_io_accounting
= false;
2250 arg_default_ip_accounting
= false;
2251 arg_default_blockio_accounting
= false;
2252 arg_default_memory_accounting
= MEMORY_ACCOUNTING_DEFAULT
;
2253 arg_default_tasks_accounting
= true;
2254 arg_default_tasks_max
= UINT64_MAX
;
2255 arg_machine_id
= (sd_id128_t
) {};
2256 arg_cad_burst_action
= EMERGENCY_ACTION_REBOOT_FORCE
;
2257 arg_default_oom_policy
= OOM_STOP
;
2259 cpu_set_reset(&arg_cpu_affinity
);
2262 static int parse_configuration(void) {
2265 arg_default_tasks_max
= system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE
, 100U);
2267 /* Assign configuration defaults */
2270 r
= parse_config_file();
2272 log_warning_errno(r
, "Failed to parse config file, ignoring: %m");
2275 r
= proc_cmdline_parse(parse_proc_cmdline_item
, NULL
, 0);
2277 log_warning_errno(r
, "Failed to parse kernel command line, ignoring: %m");
2280 /* Note that this also parses bits from the kernel command line, including "debug". */
2281 log_parse_environment();
2286 static int load_configuration(int argc
, char **argv
, const char **ret_error_message
) {
2289 assert(ret_error_message
);
2291 (void) parse_configuration();
2293 r
= parse_argv(argc
, argv
);
2295 *ret_error_message
= "Failed to parse commandline arguments";
2299 /* Initialize default unit */
2300 if (!arg_default_unit
) {
2301 arg_default_unit
= strdup(SPECIAL_DEFAULT_TARGET
);
2302 if (!arg_default_unit
) {
2303 *ret_error_message
= "Failed to set default unit";
2308 /* Initialize the show status setting if it hasn't been set explicitly yet */
2309 if (arg_show_status
== _SHOW_STATUS_INVALID
)
2310 arg_show_status
= SHOW_STATUS_YES
;
2315 static int safety_checks(void) {
2317 if (getpid_cached() == 1 &&
2318 arg_action
!= ACTION_RUN
)
2319 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
2320 "Unsupported execution mode while PID 1.");
2322 if (getpid_cached() == 1 &&
2324 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
2325 "Can't run --user mode as PID 1.");
2327 if (arg_action
== ACTION_RUN
&&
2329 getpid_cached() != 1)
2330 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
2331 "Can't run system mode unless PID 1.");
2333 if (arg_action
== ACTION_TEST
&&
2335 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
2336 "Don't run test mode as root.");
2339 arg_action
== ACTION_RUN
&&
2341 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP
),
2342 "Trying to run as user instance, but the system has not been booted with systemd.");
2345 arg_action
== ACTION_RUN
&&
2346 !getenv("XDG_RUNTIME_DIR"))
2347 return log_error_errno(SYNTHETIC_ERRNO(EUNATCH
),
2348 "Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
2351 arg_action
== ACTION_RUN
&&
2352 running_in_chroot() > 0)
2353 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP
),
2354 "Cannot be run in a chroot() environment.");
2359 static int initialize_security(
2360 bool *loaded_policy
,
2361 dual_timestamp
*security_start_timestamp
,
2362 dual_timestamp
*security_finish_timestamp
,
2363 const char **ret_error_message
) {
2367 assert(loaded_policy
);
2368 assert(security_start_timestamp
);
2369 assert(security_finish_timestamp
);
2370 assert(ret_error_message
);
2372 dual_timestamp_get(security_start_timestamp
);
2374 r
= mac_selinux_setup(loaded_policy
);
2376 *ret_error_message
= "Failed to load SELinux policy";
2380 r
= mac_smack_setup(loaded_policy
);
2382 *ret_error_message
= "Failed to load SMACK policy";
2388 *ret_error_message
= "Failed to load IMA policy";
2392 dual_timestamp_get(security_finish_timestamp
);
2396 static void test_summary(Manager
*m
) {
2399 printf("-> By units:\n");
2400 manager_dump_units(m
, stdout
, "\t");
2402 printf("-> By jobs:\n");
2403 manager_dump_jobs(m
, stdout
, "\t");
2406 static int collect_fds(FDSet
**ret_fds
, const char **ret_error_message
) {
2410 assert(ret_error_message
);
2412 r
= fdset_new_fill(ret_fds
);
2414 *ret_error_message
= "Failed to allocate fd set";
2415 return log_emergency_errno(r
, "Failed to allocate fd set: %m");
2418 fdset_cloexec(*ret_fds
, true);
2420 if (arg_serialization
)
2421 assert_se(fdset_remove(*ret_fds
, fileno(arg_serialization
)) >= 0);
2426 static void setup_console_terminal(bool skip_setup
) {
2431 /* Become a session leader if we aren't one yet. */
2434 /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2436 (void) release_terminal();
2438 /* Reset the console, but only if this is really init and we are freshly booted */
2439 if (getpid_cached() == 1 && !skip_setup
)
2440 (void) console_setup();
2443 static bool early_skip_setup_check(int argc
, char *argv
[]) {
2444 bool found_deserialize
= false;
2447 /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2448 * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2449 * anyway, even if in that case we also do deserialization. */
2451 for (i
= 1; i
< argc
; i
++) {
2452 if (streq(argv
[i
], "--switched-root"))
2453 return false; /* If we switched root, don't skip the setup. */
2454 else if (streq(argv
[i
], "--deserialize"))
2455 found_deserialize
= true;
2458 return found_deserialize
; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2461 int main(int argc
, char *argv
[]) {
2463 dual_timestamp initrd_timestamp
= DUAL_TIMESTAMP_NULL
, userspace_timestamp
= DUAL_TIMESTAMP_NULL
, kernel_timestamp
= DUAL_TIMESTAMP_NULL
,
2464 security_start_timestamp
= DUAL_TIMESTAMP_NULL
, security_finish_timestamp
= DUAL_TIMESTAMP_NULL
;
2465 struct rlimit saved_rlimit_nofile
= RLIMIT_MAKE_CONST(0),
2466 saved_rlimit_memlock
= RLIMIT_MAKE_CONST(RLIM_INFINITY
); /* The original rlimits we passed
2467 * in. Note we use different values
2468 * for the two that indicate whether
2469 * these fields are initialized! */
2470 bool skip_setup
, loaded_policy
= false, queue_default_job
= false, first_boot
= false, reexecute
= false;
2471 char *switch_root_dir
= NULL
, *switch_root_init
= NULL
;
2472 usec_t before_startup
, after_startup
;
2473 static char systemd
[] = "systemd";
2474 char timespan
[FORMAT_TIMESPAN_MAX
];
2475 const char *shutdown_verb
= NULL
, *error_message
= NULL
;
2476 int r
, retval
= EXIT_FAILURE
;
2480 /* SysV compatibility: redirect init → telinit */
2481 redirect_telinit(argc
, argv
);
2483 /* Take timestamps early on */
2484 dual_timestamp_from_monotonic(&kernel_timestamp
, 0);
2485 dual_timestamp_get(&userspace_timestamp
);
2487 /* Figure out whether we need to do initialize the system, or if we already did that because we are
2489 skip_setup
= early_skip_setup_check(argc
, argv
);
2491 /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2492 * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
2493 program_invocation_short_name
= systemd
;
2494 (void) prctl(PR_SET_NAME
, systemd
);
2496 /* Save the original command line */
2497 save_argc_argv(argc
, argv
);
2499 /* Make sure that if the user says "syslog" we actually log to the journal. */
2500 log_set_upgrade_syslog_to_journal(true);
2502 if (getpid_cached() == 1) {
2503 /* When we run as PID 1 force system mode */
2506 /* Disable the umask logic */
2509 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2510 * activated yet (even though the log socket for it exists). */
2511 log_set_prohibit_ipc(true);
2513 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2514 * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2515 * child process right before execve()'ing the actual binary, at a point in time where socket
2516 * activation stderr/stdout area already set up. */
2517 log_set_always_reopen_console(true);
2519 if (detect_container() <= 0) {
2521 /* Running outside of a container as PID 1 */
2522 log_set_target(LOG_TARGET_KMSG
);
2526 initrd_timestamp
= userspace_timestamp
;
2529 r
= mount_setup_early();
2531 error_message
= "Failed to mount early API filesystems";
2535 r
= initialize_security(
2537 &security_start_timestamp
,
2538 &security_finish_timestamp
,
2544 if (mac_selinux_init() < 0) {
2545 error_message
= "Failed to initialize SELinux policy";
2552 /* Set the default for later on, but don't actually open the logs like this for now. Note that
2553 * if we are transitioning from the initrd there might still be journal fd open, and we
2554 * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2556 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG
);
2559 /* Running inside a container, as PID 1 */
2560 log_set_target(LOG_TARGET_CONSOLE
);
2563 /* For later on, see above... */
2564 log_set_target(LOG_TARGET_JOURNAL
);
2566 /* clear the kernel timestamp,
2567 * because we are in a container */
2568 kernel_timestamp
= DUAL_TIMESTAMP_NULL
;
2571 initialize_coredump(skip_setup
);
2573 r
= fixup_environment();
2575 log_emergency_errno(r
, "Failed to fix up PID 1 environment: %m");
2576 error_message
= "Failed to fix up PID1 environment";
2581 /* Running as user instance */
2583 log_set_target(LOG_TARGET_AUTO
);
2586 /* clear the kernel timestamp,
2587 * because we are not PID 1 */
2588 kernel_timestamp
= DUAL_TIMESTAMP_NULL
;
2592 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2593 * they never log into the console. */
2594 log_show_color(colors_enabled());
2596 r
= make_null_stdio();
2598 log_warning_errno(r
, "Failed to redirect standard streams to /dev/null, ignoring: %m");
2601 /* Mount /proc, /sys and friends, so that /proc/cmdline and
2602 * /proc/$PID/fd is available. */
2603 if (getpid_cached() == 1) {
2605 /* Load the kernel modules early. */
2609 r
= mount_setup(loaded_policy
);
2611 error_message
= "Failed to mount API filesystems";
2616 /* Reset all signal handlers. */
2617 (void) reset_all_signal_handlers();
2618 (void) ignore_signals(SIGNALS_IGNORE
, -1);
2620 r
= load_configuration(argc
, argv
, &error_message
);
2624 r
= safety_checks();
2628 if (IN_SET(arg_action
, ACTION_TEST
, ACTION_HELP
, ACTION_DUMP_CONFIGURATION_ITEMS
, ACTION_DUMP_BUS_PROPERTIES
))
2629 (void) pager_open(arg_pager_flags
);
2631 if (arg_action
!= ACTION_RUN
)
2634 if (arg_action
== ACTION_HELP
) {
2635 retval
= help() < 0 ? EXIT_FAILURE
: EXIT_SUCCESS
;
2637 } else if (arg_action
== ACTION_VERSION
) {
2640 } else if (arg_action
== ACTION_DUMP_CONFIGURATION_ITEMS
) {
2641 unit_dump_config_items(stdout
);
2642 retval
= EXIT_SUCCESS
;
2644 } else if (arg_action
== ACTION_DUMP_BUS_PROPERTIES
) {
2645 dump_bus_properties(stdout
);
2646 retval
= EXIT_SUCCESS
;
2650 assert_se(IN_SET(arg_action
, ACTION_RUN
, ACTION_TEST
));
2652 /* Move out of the way, so that we won't block unmounts */
2653 assert_se(chdir("/") == 0);
2655 if (arg_action
== ACTION_RUN
) {
2657 /* A core pattern might have been specified via the cmdline. */
2658 initialize_core_pattern(skip_setup
);
2660 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2663 /* Remember open file descriptors for later deserialization */
2664 r
= collect_fds(&fds
, &error_message
);
2668 /* Give up any control of the console, but make sure its initialized. */
2669 setup_console_terminal(skip_setup
);
2671 /* Open the logging devices, if possible and necessary */
2675 log_execution_mode(&first_boot
);
2677 r
= initialize_runtime(skip_setup
,
2678 &saved_rlimit_nofile
,
2679 &saved_rlimit_memlock
,
2684 r
= manager_new(arg_system
? UNIT_FILE_SYSTEM
: UNIT_FILE_USER
,
2685 arg_action
== ACTION_TEST
? MANAGER_TEST_FULL
: 0,
2688 log_emergency_errno(r
, "Failed to allocate manager object: %m");
2689 error_message
= "Failed to allocate manager object";
2693 m
->timestamps
[MANAGER_TIMESTAMP_KERNEL
] = kernel_timestamp
;
2694 m
->timestamps
[MANAGER_TIMESTAMP_INITRD
] = initrd_timestamp
;
2695 m
->timestamps
[MANAGER_TIMESTAMP_USERSPACE
] = userspace_timestamp
;
2696 m
->timestamps
[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START
)] = security_start_timestamp
;
2697 m
->timestamps
[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH
)] = security_finish_timestamp
;
2699 set_manager_defaults(m
);
2700 set_manager_settings(m
);
2701 manager_set_first_boot(m
, first_boot
);
2703 /* Remember whether we should queue the default job */
2704 queue_default_job
= !arg_serialization
|| arg_switched_root
;
2706 before_startup
= now(CLOCK_MONOTONIC
);
2708 r
= manager_startup(m
, arg_serialization
, fds
);
2710 error_message
= "Failed to start up manager";
2714 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2715 fds
= fdset_free(fds
);
2716 arg_serialization
= safe_fclose(arg_serialization
);
2718 if (queue_default_job
) {
2719 r
= do_queue_default_job(m
, &error_message
);
2724 after_startup
= now(CLOCK_MONOTONIC
);
2726 log_full(arg_action
== ACTION_TEST
? LOG_INFO
: LOG_DEBUG
,
2727 "Loaded units and determined initial transaction in %s.",
2728 format_timespan(timespan
, sizeof(timespan
), after_startup
- before_startup
, 100 * USEC_PER_MSEC
));
2730 if (arg_action
== ACTION_TEST
) {
2732 retval
= EXIT_SUCCESS
;
2736 (void) invoke_main_loop(m
,
2749 arg_shutdown_watchdog
= m
->shutdown_watchdog
;
2750 m
= manager_free(m
);
2754 mac_selinux_finish();
2757 do_reexecute(argc
, argv
,
2758 &saved_rlimit_nofile
,
2759 &saved_rlimit_memlock
,
2763 &error_message
); /* This only returns if reexecution failed */
2765 arg_serialization
= safe_fclose(arg_serialization
);
2766 fds
= fdset_free(fds
);
2768 #if HAVE_VALGRIND_VALGRIND_H
2769 /* If we are PID 1 and running under valgrind, then let's exit
2770 * here explicitly. valgrind will only generate nice output on
2771 * exit(), not on exec(), hence let's do the former not the
2773 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND
) {
2774 /* Cleanup watchdog_device strings for valgrind. We need them
2775 * in become_shutdown() so normally we cannot free them yet. */
2776 watchdog_free_device();
2777 arg_watchdog_device
= mfree(arg_watchdog_device
);
2782 #if HAS_FEATURE_ADDRESS_SANITIZER
2783 __lsan_do_leak_check();
2786 if (shutdown_verb
) {
2787 r
= become_shutdown(shutdown_verb
, retval
);
2788 log_error_errno(r
, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
2789 error_message
= "Failed to execute shutdown binary";
2792 watchdog_free_device();
2793 arg_watchdog_device
= mfree(arg_watchdog_device
);
2795 if (getpid_cached() == 1) {
2797 manager_status_printf(NULL
, STATUS_TYPE_EMERGENCY
,
2798 ANSI_HIGHLIGHT_RED
"!!!!!!" ANSI_NORMAL
,
2799 "%s.", error_message
);
2800 freeze_or_exit_or_reboot();