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
;
99 static char *arg_default_unit
= NULL
;
100 static bool arg_system
= false;
101 static bool arg_dump_core
= true;
102 static int arg_crash_chvt
= -1;
103 static bool arg_crash_shell
= false;
104 static bool arg_crash_reboot
= false;
105 static char *arg_confirm_spawn
= NULL
;
106 static ShowStatus arg_show_status
= _SHOW_STATUS_INVALID
;
107 static bool arg_switched_root
= false;
108 static PagerFlags arg_pager_flags
= 0;
109 static bool arg_service_watchdogs
= true;
110 static ExecOutput arg_default_std_output
= EXEC_OUTPUT_JOURNAL
;
111 static ExecOutput arg_default_std_error
= EXEC_OUTPUT_INHERIT
;
112 static usec_t arg_default_restart_usec
= DEFAULT_RESTART_USEC
;
113 static usec_t arg_default_timeout_start_usec
= DEFAULT_TIMEOUT_USEC
;
114 static usec_t arg_default_timeout_stop_usec
= DEFAULT_TIMEOUT_USEC
;
115 static usec_t arg_default_timeout_abort_usec
= DEFAULT_TIMEOUT_USEC
;
116 static bool arg_default_timeout_abort_set
= false;
117 static usec_t arg_default_start_limit_interval
= DEFAULT_START_LIMIT_INTERVAL
;
118 static unsigned arg_default_start_limit_burst
= DEFAULT_START_LIMIT_BURST
;
119 static usec_t arg_runtime_watchdog
= 0;
120 static usec_t arg_shutdown_watchdog
= 10 * USEC_PER_MINUTE
;
121 static char *arg_early_core_pattern
= NULL
;
122 static char *arg_watchdog_device
= NULL
;
123 static char **arg_default_environment
= NULL
;
124 static struct rlimit
*arg_default_rlimit
[_RLIMIT_MAX
] = {};
125 static uint64_t arg_capability_bounding_set
= CAP_ALL
;
126 static bool arg_no_new_privs
= false;
127 static nsec_t arg_timer_slack_nsec
= NSEC_INFINITY
;
128 static usec_t arg_default_timer_accuracy_usec
= 1 * USEC_PER_MINUTE
;
129 static Set
* arg_syscall_archs
= NULL
;
130 static FILE* arg_serialization
= NULL
;
131 static int arg_default_cpu_accounting
= -1;
132 static bool arg_default_io_accounting
= false;
133 static bool arg_default_ip_accounting
= false;
134 static bool arg_default_blockio_accounting
= false;
135 static bool arg_default_memory_accounting
= MEMORY_ACCOUNTING_DEFAULT
;
136 static bool arg_default_tasks_accounting
= true;
137 static uint64_t arg_default_tasks_max
= UINT64_MAX
;
138 static sd_id128_t arg_machine_id
= {};
139 static EmergencyAction arg_cad_burst_action
= EMERGENCY_ACTION_REBOOT_FORCE
;
140 static OOMPolicy arg_default_oom_policy
= OOM_STOP
;
142 _noreturn_
static void freeze_or_exit_or_reboot(void) {
144 /* If we are running in a container, let's prefer exiting, after all we can propagate an exit code to
145 * the container manager, and thus inform it that something went wrong. */
146 if (detect_container() > 0) {
147 log_emergency("Exiting PID 1...");
148 _exit(EXIT_EXCEPTION
);
151 if (arg_crash_reboot
) {
152 log_notice("Rebooting in 10s...");
155 log_notice("Rebooting now...");
156 (void) reboot(RB_AUTOBOOT
);
157 log_emergency_errno(errno
, "Failed to reboot: %m");
160 log_emergency("Freezing execution.");
164 _noreturn_
static void crash(int sig
) {
168 if (getpid_cached() != 1)
169 /* Pass this on immediately, if this is not PID 1 */
171 else if (!arg_dump_core
)
172 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig
));
174 sa
= (struct sigaction
) {
175 .sa_handler
= nop_signal_handler
,
176 .sa_flags
= SA_NOCLDSTOP
|SA_RESTART
,
179 /* We want to wait for the core process, hence let's enable SIGCHLD */
180 (void) sigaction(SIGCHLD
, &sa
, NULL
);
182 pid
= raw_clone(SIGCHLD
);
184 log_emergency_errno(errno
, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig
));
186 /* Enable default signal handler for core dump */
188 sa
= (struct sigaction
) {
189 .sa_handler
= SIG_DFL
,
191 (void) sigaction(sig
, &sa
, NULL
);
193 /* Don't limit the coredump size */
194 (void) setrlimit(RLIMIT_CORE
, &RLIMIT_MAKE_CONST(RLIM_INFINITY
));
196 /* Just to be sure... */
199 /* Raise the signal again */
201 (void) kill(pid
, sig
); /* raise() would kill the parent */
203 assert_not_reached("We shouldn't be here...");
204 _exit(EXIT_EXCEPTION
);
209 /* Order things nicely. */
210 r
= wait_for_terminate(pid
, &status
);
212 log_emergency_errno(r
, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig
));
213 else if (status
.si_code
!= CLD_DUMPED
)
214 log_emergency("Caught <%s>, core dump failed (child "PID_FMT
", code=%s, status=%i/%s).",
215 signal_to_string(sig
),
216 pid
, sigchld_code_to_string(status
.si_code
),
218 strna(status
.si_code
== CLD_EXITED
219 ? exit_status_to_string(status
.si_status
, EXIT_STATUS_MINIMAL
)
220 : signal_to_string(status
.si_status
)));
222 log_emergency("Caught <%s>, dumped core as pid "PID_FMT
".", signal_to_string(sig
), pid
);
226 if (arg_crash_chvt
>= 0)
227 (void) chvt(arg_crash_chvt
);
229 sa
= (struct sigaction
) {
230 .sa_handler
= SIG_IGN
,
231 .sa_flags
= SA_NOCLDSTOP
|SA_NOCLDWAIT
|SA_RESTART
,
234 /* Let the kernel reap children for us */
235 (void) sigaction(SIGCHLD
, &sa
, NULL
);
237 if (arg_crash_shell
) {
238 log_notice("Executing crash shell in 10s...");
241 pid
= raw_clone(SIGCHLD
);
243 log_emergency_errno(errno
, "Failed to fork off crash shell: %m");
246 (void) make_console_stdio();
247 (void) rlimit_nofile_safe();
248 (void) execle("/bin/sh", "/bin/sh", NULL
, environ
);
250 log_emergency_errno(errno
, "execle() failed: %m");
251 _exit(EXIT_EXCEPTION
);
253 log_info("Spawned crash shell as PID "PID_FMT
".", pid
);
254 (void) wait_for_terminate(pid
, NULL
);
258 freeze_or_exit_or_reboot();
261 static void install_crash_handler(void) {
262 static const struct sigaction sa
= {
264 .sa_flags
= SA_NODEFER
, /* So that we can raise the signal again from the signal handler */
268 /* We ignore the return value here, since, we don't mind if we
269 * cannot set up a crash handler */
270 r
= sigaction_many(&sa
, SIGNALS_CRASH_HANDLER
, -1);
272 log_debug_errno(r
, "I had trouble setting up the crash handler, ignoring: %m");
275 static int console_setup(void) {
276 _cleanup_close_
int tty_fd
= -1;
279 tty_fd
= open_terminal("/dev/console", O_WRONLY
|O_NOCTTY
|O_CLOEXEC
);
281 return log_error_errno(tty_fd
, "Failed to open /dev/console: %m");
283 /* We don't want to force text mode. plymouth may be showing
284 * pictures already from initrd. */
285 r
= reset_terminal_fd(tty_fd
, false);
287 return log_error_errno(r
, "Failed to reset /dev/console: %m");
292 static int parse_crash_chvt(const char *value
) {
295 if (safe_atoi(value
, &arg_crash_chvt
) >= 0)
298 b
= parse_boolean(value
);
303 arg_crash_chvt
= 0; /* switch to where kmsg goes */
305 arg_crash_chvt
= -1; /* turn off switching */
310 static int parse_confirm_spawn(const char *value
, char **console
) {
314 r
= value
? parse_boolean(value
) : 1;
320 if (r
> 0) /* on with default tty */
321 s
= strdup("/dev/console");
322 else if (is_path(value
)) /* on with fully qualified path */
324 else /* on with only a tty file name, not a fully qualified path */
325 s
= strjoin("/dev/", value
);
333 static int set_machine_id(const char *m
) {
337 if (sd_id128_from_string(m
, &t
) < 0)
340 if (sd_id128_is_null(t
))
347 static int parse_proc_cmdline_item(const char *key
, const char *value
, void *data
) {
353 if (STR_IN_SET(key
, "systemd.unit", "rd.systemd.unit")) {
355 if (proc_cmdline_value_missing(key
, value
))
358 if (!unit_name_is_valid(value
, UNIT_NAME_PLAIN
|UNIT_NAME_INSTANCE
))
359 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key
, value
);
360 else if (in_initrd() == !!startswith(key
, "rd.")) {
361 if (free_and_strdup(&arg_default_unit
, value
) < 0)
365 } else if (proc_cmdline_key_streq(key
, "systemd.dump_core")) {
367 r
= value
? parse_boolean(value
) : true;
369 log_warning_errno(r
, "Failed to parse dump core switch %s, ignoring: %m", value
);
373 } else if (proc_cmdline_key_streq(key
, "systemd.early_core_pattern")) {
375 if (proc_cmdline_value_missing(key
, value
))
378 if (path_is_absolute(value
))
379 (void) parse_path_argument_and_warn(value
, false, &arg_early_core_pattern
);
381 log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value
);
383 } else if (proc_cmdline_key_streq(key
, "systemd.crash_chvt")) {
386 arg_crash_chvt
= 0; /* turn on */
388 r
= parse_crash_chvt(value
);
390 log_warning_errno(r
, "Failed to parse crash chvt switch %s, ignoring: %m", value
);
393 } else if (proc_cmdline_key_streq(key
, "systemd.crash_shell")) {
395 r
= value
? parse_boolean(value
) : true;
397 log_warning_errno(r
, "Failed to parse crash shell switch %s, ignoring: %m", value
);
401 } else if (proc_cmdline_key_streq(key
, "systemd.crash_reboot")) {
403 r
= value
? parse_boolean(value
) : true;
405 log_warning_errno(r
, "Failed to parse crash reboot switch %s, ignoring: %m", value
);
407 arg_crash_reboot
= r
;
409 } else if (proc_cmdline_key_streq(key
, "systemd.confirm_spawn")) {
412 r
= parse_confirm_spawn(value
, &s
);
414 log_warning_errno(r
, "Failed to parse confirm_spawn switch %s, ignoring: %m", value
);
416 free_and_replace(arg_confirm_spawn
, s
);
418 } else if (proc_cmdline_key_streq(key
, "systemd.service_watchdogs")) {
420 r
= value
? parse_boolean(value
) : true;
422 log_warning_errno(r
, "Failed to parse service watchdog switch %s, ignoring: %m", value
);
424 arg_service_watchdogs
= r
;
426 } else if (proc_cmdline_key_streq(key
, "systemd.show_status")) {
429 r
= parse_show_status(value
, &arg_show_status
);
431 log_warning_errno(r
, "Failed to parse show status switch %s, ignoring: %m", value
);
433 arg_show_status
= SHOW_STATUS_YES
;
435 } else if (proc_cmdline_key_streq(key
, "systemd.default_standard_output")) {
437 if (proc_cmdline_value_missing(key
, value
))
440 r
= exec_output_from_string(value
);
442 log_warning_errno(r
, "Failed to parse default standard output switch %s, ignoring: %m", value
);
444 arg_default_std_output
= r
;
446 } else if (proc_cmdline_key_streq(key
, "systemd.default_standard_error")) {
448 if (proc_cmdline_value_missing(key
, value
))
451 r
= exec_output_from_string(value
);
453 log_warning_errno(r
, "Failed to parse default standard error switch %s, ignoring: %m", value
);
455 arg_default_std_error
= r
;
457 } else if (streq(key
, "systemd.setenv")) {
459 if (proc_cmdline_value_missing(key
, value
))
462 if (env_assignment_is_valid(value
)) {
465 env
= strv_env_set(arg_default_environment
, value
);
469 arg_default_environment
= env
;
471 log_warning("Environment variable name '%s' is not valid. Ignoring.", value
);
473 } else if (proc_cmdline_key_streq(key
, "systemd.machine_id")) {
475 if (proc_cmdline_value_missing(key
, value
))
478 r
= set_machine_id(value
);
480 log_warning_errno(r
, "MachineID '%s' is not valid, ignoring: %m", value
);
482 } else if (proc_cmdline_key_streq(key
, "systemd.default_timeout_start_sec")) {
484 if (proc_cmdline_value_missing(key
, value
))
487 r
= parse_sec(value
, &arg_default_timeout_start_usec
);
489 log_warning_errno(r
, "Failed to parse default start timeout '%s', ignoring: %m", value
);
491 if (arg_default_timeout_start_usec
<= 0)
492 arg_default_timeout_start_usec
= USEC_INFINITY
;
494 } else if (proc_cmdline_key_streq(key
, "systemd.watchdog_device")) {
496 if (proc_cmdline_value_missing(key
, value
))
499 (void) parse_path_argument_and_warn(value
, false, &arg_watchdog_device
);
501 } else if (streq(key
, "quiet") && !value
) {
503 if (arg_show_status
== _SHOW_STATUS_INVALID
)
504 arg_show_status
= SHOW_STATUS_AUTO
;
506 } else if (streq(key
, "debug") && !value
) {
508 /* Note that log_parse_environment() handles 'debug'
509 * too, and sets the log level to LOG_DEBUG. */
511 if (detect_container() > 0)
512 log_set_target(LOG_TARGET_CONSOLE
);
517 /* SysV compatibility */
518 target
= runlevel_to_target(key
);
520 return free_and_strdup(&arg_default_unit
, target
);
526 #define DEFINE_SETTER(name, func, descr) \
527 static int name(const char *unit, \
528 const char *filename, \
530 const char *section, \
531 unsigned section_line, \
532 const char *lvalue, \
534 const char *rvalue, \
546 log_syntax(unit, LOG_ERR, filename, line, r, \
547 "Invalid " descr "'%s': %m", \
553 DEFINE_SETTER(config_parse_level2
, log_set_max_level_from_string
, "log level");
554 DEFINE_SETTER(config_parse_target
, log_set_target_from_string
, "target");
555 DEFINE_SETTER(config_parse_color
, log_show_color_from_string
, "color" );
556 DEFINE_SETTER(config_parse_location
, log_show_location_from_string
, "location");
558 static int config_parse_cpu_affinity2(
560 const char *filename
,
563 unsigned section_line
,
570 _cleanup_cpu_free_ cpu_set_t
*c
= NULL
;
573 ncpus
= parse_cpu_set_and_warn(rvalue
, &c
, unit
, filename
, line
, lvalue
);
577 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus
), c
) < 0)
578 log_warning_errno(errno
, "Failed to set CPU affinity: %m");
583 static int config_parse_show_status(
585 const char *filename
,
588 unsigned section_line
,
596 ShowStatus
*b
= data
;
603 k
= parse_show_status(rvalue
, b
);
605 log_syntax(unit
, LOG_ERR
, filename
, line
, k
, "Failed to parse show status setting, ignoring: %s", rvalue
);
612 static int config_parse_output_restricted(
614 const char *filename
,
617 unsigned section_line
,
624 ExecOutput t
, *eo
= data
;
631 t
= exec_output_from_string(rvalue
);
633 log_syntax(unit
, LOG_ERR
, filename
, line
, 0, "Failed to parse output type, ignoring: %s", rvalue
);
637 if (IN_SET(t
, EXEC_OUTPUT_SOCKET
, EXEC_OUTPUT_NAMED_FD
, EXEC_OUTPUT_FILE
, EXEC_OUTPUT_FILE_APPEND
)) {
638 log_syntax(unit
, LOG_ERR
, filename
, line
, 0, "Standard output types socket, fd:, file:, append: are not supported as defaults, ignoring: %s", rvalue
);
646 static int config_parse_crash_chvt(
648 const char *filename
,
651 unsigned section_line
,
664 r
= parse_crash_chvt(rvalue
);
666 log_syntax(unit
, LOG_ERR
, filename
, line
, r
, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue
);
673 static int config_parse_timeout_abort(
675 const char *filename
,
678 unsigned section_line
,
691 rvalue
+= strspn(rvalue
, WHITESPACE
);
692 if (isempty(rvalue
)) {
693 arg_default_timeout_abort_set
= false;
697 r
= parse_sec(rvalue
, &arg_default_timeout_abort_usec
);
699 log_syntax(unit
, LOG_ERR
, filename
, line
, r
, "Failed to parse DefaultTimeoutAbortSec= setting, ignoring: %s", rvalue
);
703 arg_default_timeout_abort_set
= true;
707 static int parse_config_file(void) {
709 const ConfigTableItem items
[] = {
710 { "Manager", "LogLevel", config_parse_level2
, 0, NULL
},
711 { "Manager", "LogTarget", config_parse_target
, 0, NULL
},
712 { "Manager", "LogColor", config_parse_color
, 0, NULL
},
713 { "Manager", "LogLocation", config_parse_location
, 0, NULL
},
714 { "Manager", "DumpCore", config_parse_bool
, 0, &arg_dump_core
},
715 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt
, 0, NULL
},
716 { "Manager", "CrashChangeVT", config_parse_crash_chvt
, 0, NULL
},
717 { "Manager", "CrashShell", config_parse_bool
, 0, &arg_crash_shell
},
718 { "Manager", "CrashReboot", config_parse_bool
, 0, &arg_crash_reboot
},
719 { "Manager", "ShowStatus", config_parse_show_status
, 0, &arg_show_status
},
720 { "Manager", "CPUAffinity", config_parse_cpu_affinity2
, 0, NULL
},
721 { "Manager", "JoinControllers", config_parse_warn_compat
, DISABLED_CONFIGURATION
, NULL
},
722 { "Manager", "RuntimeWatchdogSec", config_parse_sec
, 0, &arg_runtime_watchdog
},
723 { "Manager", "ShutdownWatchdogSec", config_parse_sec
, 0, &arg_shutdown_watchdog
},
724 { "Manager", "WatchdogDevice", config_parse_path
, 0, &arg_watchdog_device
},
725 { "Manager", "CapabilityBoundingSet", config_parse_capability_set
, 0, &arg_capability_bounding_set
},
726 { "Manager", "NoNewPrivileges", config_parse_bool
, 0, &arg_no_new_privs
},
728 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs
, 0, &arg_syscall_archs
},
730 { "Manager", "TimerSlackNSec", config_parse_nsec
, 0, &arg_timer_slack_nsec
},
731 { "Manager", "DefaultTimerAccuracySec", config_parse_sec
, 0, &arg_default_timer_accuracy_usec
},
732 { "Manager", "DefaultStandardOutput", config_parse_output_restricted
,0, &arg_default_std_output
},
733 { "Manager", "DefaultStandardError", config_parse_output_restricted
,0, &arg_default_std_error
},
734 { "Manager", "DefaultTimeoutStartSec", config_parse_sec
, 0, &arg_default_timeout_start_usec
},
735 { "Manager", "DefaultTimeoutStopSec", config_parse_sec
, 0, &arg_default_timeout_stop_usec
},
736 { "Manager", "DefaultTimeoutAbortSec", config_parse_timeout_abort
, 0, NULL
},
737 { "Manager", "DefaultRestartSec", config_parse_sec
, 0, &arg_default_restart_usec
},
738 { "Manager", "DefaultStartLimitInterval", config_parse_sec
, 0, &arg_default_start_limit_interval
}, /* obsolete alias */
739 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec
, 0, &arg_default_start_limit_interval
},
740 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned
, 0, &arg_default_start_limit_burst
},
741 { "Manager", "DefaultEnvironment", config_parse_environ
, 0, &arg_default_environment
},
742 { "Manager", "DefaultLimitCPU", config_parse_rlimit
, RLIMIT_CPU
, arg_default_rlimit
},
743 { "Manager", "DefaultLimitFSIZE", config_parse_rlimit
, RLIMIT_FSIZE
, arg_default_rlimit
},
744 { "Manager", "DefaultLimitDATA", config_parse_rlimit
, RLIMIT_DATA
, arg_default_rlimit
},
745 { "Manager", "DefaultLimitSTACK", config_parse_rlimit
, RLIMIT_STACK
, arg_default_rlimit
},
746 { "Manager", "DefaultLimitCORE", config_parse_rlimit
, RLIMIT_CORE
, arg_default_rlimit
},
747 { "Manager", "DefaultLimitRSS", config_parse_rlimit
, RLIMIT_RSS
, arg_default_rlimit
},
748 { "Manager", "DefaultLimitNOFILE", config_parse_rlimit
, RLIMIT_NOFILE
, arg_default_rlimit
},
749 { "Manager", "DefaultLimitAS", config_parse_rlimit
, RLIMIT_AS
, arg_default_rlimit
},
750 { "Manager", "DefaultLimitNPROC", config_parse_rlimit
, RLIMIT_NPROC
, arg_default_rlimit
},
751 { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit
, RLIMIT_MEMLOCK
, arg_default_rlimit
},
752 { "Manager", "DefaultLimitLOCKS", config_parse_rlimit
, RLIMIT_LOCKS
, arg_default_rlimit
},
753 { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit
, RLIMIT_SIGPENDING
, arg_default_rlimit
},
754 { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit
, RLIMIT_MSGQUEUE
, arg_default_rlimit
},
755 { "Manager", "DefaultLimitNICE", config_parse_rlimit
, RLIMIT_NICE
, arg_default_rlimit
},
756 { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit
, RLIMIT_RTPRIO
, arg_default_rlimit
},
757 { "Manager", "DefaultLimitRTTIME", config_parse_rlimit
, RLIMIT_RTTIME
, arg_default_rlimit
},
758 { "Manager", "DefaultCPUAccounting", config_parse_tristate
, 0, &arg_default_cpu_accounting
},
759 { "Manager", "DefaultIOAccounting", config_parse_bool
, 0, &arg_default_io_accounting
},
760 { "Manager", "DefaultIPAccounting", config_parse_bool
, 0, &arg_default_ip_accounting
},
761 { "Manager", "DefaultBlockIOAccounting", config_parse_bool
, 0, &arg_default_blockio_accounting
},
762 { "Manager", "DefaultMemoryAccounting", config_parse_bool
, 0, &arg_default_memory_accounting
},
763 { "Manager", "DefaultTasksAccounting", config_parse_bool
, 0, &arg_default_tasks_accounting
},
764 { "Manager", "DefaultTasksMax", config_parse_tasks_max
, 0, &arg_default_tasks_max
},
765 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action
, 0, &arg_cad_burst_action
},
766 { "Manager", "DefaultOOMPolicy", config_parse_oom_policy
, 0, &arg_default_oom_policy
},
770 const char *fn
, *conf_dirs_nulstr
;
773 PKGSYSCONFDIR
"/system.conf" :
774 PKGSYSCONFDIR
"/user.conf";
776 conf_dirs_nulstr
= arg_system
?
777 CONF_PATHS_NULSTR("systemd/system.conf.d") :
778 CONF_PATHS_NULSTR("systemd/user.conf.d");
780 (void) config_parse_many_nulstr(fn
, conf_dirs_nulstr
, "Manager\0", config_item_table_lookup
, items
, CONFIG_PARSE_WARN
, NULL
);
782 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
783 * like everywhere else. */
784 if (arg_default_timeout_start_usec
<= 0)
785 arg_default_timeout_start_usec
= USEC_INFINITY
;
786 if (arg_default_timeout_stop_usec
<= 0)
787 arg_default_timeout_stop_usec
= USEC_INFINITY
;
792 static void set_manager_defaults(Manager
*m
) {
796 /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
797 * affect the manager itself, but are just what newly allocated units will have set if they haven't set
798 * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
800 m
->default_timer_accuracy_usec
= arg_default_timer_accuracy_usec
;
801 m
->default_std_output
= arg_default_std_output
;
802 m
->default_std_error
= arg_default_std_error
;
803 m
->default_timeout_start_usec
= arg_default_timeout_start_usec
;
804 m
->default_timeout_stop_usec
= arg_default_timeout_stop_usec
;
805 m
->default_timeout_abort_usec
= arg_default_timeout_abort_usec
;
806 m
->default_timeout_abort_set
= arg_default_timeout_abort_set
;
807 m
->default_restart_usec
= arg_default_restart_usec
;
808 m
->default_start_limit_interval
= arg_default_start_limit_interval
;
809 m
->default_start_limit_burst
= arg_default_start_limit_burst
;
811 /* On 4.15+ with unified hierarchy, CPU accounting is essentially free as it doesn't require the CPU
812 * controller to be enabled, so the default is to enable it unless we got told otherwise. */
813 if (arg_default_cpu_accounting
>= 0)
814 m
->default_cpu_accounting
= arg_default_cpu_accounting
;
816 m
->default_cpu_accounting
= cpu_accounting_is_cheap();
818 m
->default_io_accounting
= arg_default_io_accounting
;
819 m
->default_ip_accounting
= arg_default_ip_accounting
;
820 m
->default_blockio_accounting
= arg_default_blockio_accounting
;
821 m
->default_memory_accounting
= arg_default_memory_accounting
;
822 m
->default_tasks_accounting
= arg_default_tasks_accounting
;
823 m
->default_tasks_max
= arg_default_tasks_max
;
824 m
->default_oom_policy
= arg_default_oom_policy
;
826 (void) manager_set_default_rlimits(m
, arg_default_rlimit
);
828 (void) manager_default_environment(m
);
829 (void) manager_transient_environment_add(m
, arg_default_environment
);
832 static void set_manager_settings(Manager
*m
) {
836 /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
837 * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
839 m
->confirm_spawn
= arg_confirm_spawn
;
840 m
->service_watchdogs
= arg_service_watchdogs
;
841 m
->runtime_watchdog
= arg_runtime_watchdog
;
842 m
->shutdown_watchdog
= arg_shutdown_watchdog
;
843 m
->cad_burst_action
= arg_cad_burst_action
;
845 manager_set_show_status(m
, arg_show_status
);
848 static int parse_argv(int argc
, char *argv
[]) {
850 ARG_LOG_LEVEL
= 0x100,
860 ARG_DUMP_CONFIGURATION_ITEMS
,
861 ARG_DUMP_BUS_PROPERTIES
,
870 ARG_DEFAULT_STD_OUTPUT
,
871 ARG_DEFAULT_STD_ERROR
,
873 ARG_SERVICE_WATCHDOGS
,
876 static const struct option options
[] = {
877 { "log-level", required_argument
, NULL
, ARG_LOG_LEVEL
},
878 { "log-target", required_argument
, NULL
, ARG_LOG_TARGET
},
879 { "log-color", optional_argument
, NULL
, ARG_LOG_COLOR
},
880 { "log-location", optional_argument
, NULL
, ARG_LOG_LOCATION
},
881 { "unit", required_argument
, NULL
, ARG_UNIT
},
882 { "system", no_argument
, NULL
, ARG_SYSTEM
},
883 { "user", no_argument
, NULL
, ARG_USER
},
884 { "test", no_argument
, NULL
, ARG_TEST
},
885 { "no-pager", no_argument
, NULL
, ARG_NO_PAGER
},
886 { "help", no_argument
, NULL
, 'h' },
887 { "version", no_argument
, NULL
, ARG_VERSION
},
888 { "dump-configuration-items", no_argument
, NULL
, ARG_DUMP_CONFIGURATION_ITEMS
},
889 { "dump-bus-properties", no_argument
, NULL
, ARG_DUMP_BUS_PROPERTIES
},
890 { "dump-core", optional_argument
, NULL
, ARG_DUMP_CORE
},
891 { "crash-chvt", required_argument
, NULL
, ARG_CRASH_CHVT
},
892 { "crash-shell", optional_argument
, NULL
, ARG_CRASH_SHELL
},
893 { "crash-reboot", optional_argument
, NULL
, ARG_CRASH_REBOOT
},
894 { "confirm-spawn", optional_argument
, NULL
, ARG_CONFIRM_SPAWN
},
895 { "show-status", optional_argument
, NULL
, ARG_SHOW_STATUS
},
896 { "deserialize", required_argument
, NULL
, ARG_DESERIALIZE
},
897 { "switched-root", no_argument
, NULL
, ARG_SWITCHED_ROOT
},
898 { "default-standard-output", required_argument
, NULL
, ARG_DEFAULT_STD_OUTPUT
, },
899 { "default-standard-error", required_argument
, NULL
, ARG_DEFAULT_STD_ERROR
, },
900 { "machine-id", required_argument
, NULL
, ARG_MACHINE_ID
},
901 { "service-watchdogs", required_argument
, NULL
, ARG_SERVICE_WATCHDOGS
},
910 if (getpid_cached() == 1)
913 while ((c
= getopt_long(argc
, argv
, "hDbsz:", options
, NULL
)) >= 0)
918 r
= log_set_max_level_from_string(optarg
);
920 return log_error_errno(r
, "Failed to parse log level \"%s\": %m", optarg
);
925 r
= log_set_target_from_string(optarg
);
927 return log_error_errno(r
, "Failed to parse log target \"%s\": %m", optarg
);
934 r
= log_show_color_from_string(optarg
);
936 return log_error_errno(r
, "Failed to parse log color setting \"%s\": %m",
939 log_show_color(true);
943 case ARG_LOG_LOCATION
:
945 r
= log_show_location_from_string(optarg
);
947 return log_error_errno(r
, "Failed to parse log location setting \"%s\": %m",
950 log_show_location(true);
954 case ARG_DEFAULT_STD_OUTPUT
:
955 r
= exec_output_from_string(optarg
);
957 return log_error_errno(r
, "Failed to parse default standard output setting \"%s\": %m",
959 arg_default_std_output
= r
;
962 case ARG_DEFAULT_STD_ERROR
:
963 r
= exec_output_from_string(optarg
);
965 return log_error_errno(r
, "Failed to parse default standard error output setting \"%s\": %m",
967 arg_default_std_error
= r
;
971 r
= free_and_strdup(&arg_default_unit
, optarg
);
973 return log_error_errno(r
, "Failed to set default unit \"%s\": %m", optarg
);
986 arg_action
= ACTION_TEST
;
990 arg_pager_flags
|= PAGER_DISABLE
;
994 arg_action
= ACTION_VERSION
;
997 case ARG_DUMP_CONFIGURATION_ITEMS
:
998 arg_action
= ACTION_DUMP_CONFIGURATION_ITEMS
;
1001 case ARG_DUMP_BUS_PROPERTIES
:
1002 arg_action
= ACTION_DUMP_BUS_PROPERTIES
;
1007 arg_dump_core
= true;
1009 r
= parse_boolean(optarg
);
1011 return log_error_errno(r
, "Failed to parse dump core boolean: \"%s\": %m",
1017 case ARG_CRASH_CHVT
:
1018 r
= parse_crash_chvt(optarg
);
1020 return log_error_errno(r
, "Failed to parse crash virtual terminal index: \"%s\": %m",
1024 case ARG_CRASH_SHELL
:
1026 arg_crash_shell
= true;
1028 r
= parse_boolean(optarg
);
1030 return log_error_errno(r
, "Failed to parse crash shell boolean: \"%s\": %m",
1032 arg_crash_shell
= r
;
1036 case ARG_CRASH_REBOOT
:
1038 arg_crash_reboot
= true;
1040 r
= parse_boolean(optarg
);
1042 return log_error_errno(r
, "Failed to parse crash shell boolean: \"%s\": %m",
1044 arg_crash_reboot
= r
;
1048 case ARG_CONFIRM_SPAWN
:
1049 arg_confirm_spawn
= mfree(arg_confirm_spawn
);
1051 r
= parse_confirm_spawn(optarg
, &arg_confirm_spawn
);
1053 return log_error_errno(r
, "Failed to parse confirm spawn option: \"%s\": %m",
1057 case ARG_SERVICE_WATCHDOGS
:
1058 r
= parse_boolean(optarg
);
1060 return log_error_errno(r
, "Failed to parse service watchdogs boolean: \"%s\": %m",
1062 arg_service_watchdogs
= r
;
1065 case ARG_SHOW_STATUS
:
1067 r
= parse_show_status(optarg
, &arg_show_status
);
1069 return log_error_errno(r
, "Failed to parse show status boolean: \"%s\": %m",
1072 arg_show_status
= SHOW_STATUS_YES
;
1075 case ARG_DESERIALIZE
: {
1079 r
= safe_atoi(optarg
, &fd
);
1081 log_error_errno(r
, "Failed to parse deserialize option \"%s\": %m", optarg
);
1083 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
),
1084 "Invalid deserialize fd: %d",
1087 (void) fd_cloexec(fd
, true);
1089 f
= fdopen(fd
, "r");
1091 return log_error_errno(errno
, "Failed to open serialization fd %d: %m", fd
);
1093 safe_fclose(arg_serialization
);
1094 arg_serialization
= f
;
1099 case ARG_SWITCHED_ROOT
:
1100 arg_switched_root
= true;
1103 case ARG_MACHINE_ID
:
1104 r
= set_machine_id(optarg
);
1106 return log_error_errno(r
, "MachineID '%s' is not valid: %m", optarg
);
1110 arg_action
= ACTION_HELP
;
1114 log_set_max_level(LOG_DEBUG
);
1120 /* Just to eat away the sysvinit kernel
1121 * cmdline args without getopt() error
1122 * messages that we'll parse in
1123 * parse_proc_cmdline_word() or ignore. */
1126 if (getpid_cached() != 1)
1132 assert_not_reached("Unhandled option code.");
1135 if (optind
< argc
&& getpid_cached() != 1) {
1136 /* Hmm, when we aren't run as init system
1137 * let's complain about excess arguments */
1139 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
),
1140 "Excess arguments.");
1146 static int help(void) {
1147 _cleanup_free_
char *link
= NULL
;
1150 r
= terminal_urlify_man("systemd", "1", &link
);
1154 printf("%s [OPTIONS...]\n\n"
1155 "Starts up and maintains the system or user services.\n\n"
1156 " -h --help Show this help\n"
1157 " --version Show version\n"
1158 " --test Determine startup sequence, dump it and exit\n"
1159 " --no-pager Do not pipe output into a pager\n"
1160 " --dump-configuration-items Dump understood unit configuration items\n"
1161 " --dump-bus-properties Dump exposed bus properties\n"
1162 " --unit=UNIT Set default unit\n"
1163 " --system Run a system instance, even if PID != 1\n"
1164 " --user Run a user instance\n"
1165 " --dump-core[=BOOL] Dump core on crash\n"
1166 " --crash-vt=NR Change to specified VT on crash\n"
1167 " --crash-reboot[=BOOL] Reboot on crash\n"
1168 " --crash-shell[=BOOL] Run shell on crash\n"
1169 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1170 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1171 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1172 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1173 " --log-color[=BOOL] Highlight important log messages\n"
1174 " --log-location[=BOOL] Include code location in log messages\n"
1175 " --default-standard-output= Set default standard output for services\n"
1176 " --default-standard-error= Set default standard error output for services\n"
1177 "\nSee the %s for details.\n"
1178 , program_invocation_short_name
1185 static int prepare_reexecute(
1189 bool switching_root
) {
1191 _cleanup_fdset_free_ FDSet
*fds
= NULL
;
1192 _cleanup_fclose_
FILE *f
= NULL
;
1199 r
= manager_open_serialization(m
, &f
);
1201 return log_error_errno(r
, "Failed to create serialization file: %m");
1203 /* Make sure nothing is really destructed when we shut down */
1205 bus_manager_send_reloading(m
, true);
1211 r
= manager_serialize(m
, f
, fds
, switching_root
);
1215 if (fseeko(f
, 0, SEEK_SET
) == (off_t
) -1)
1216 return log_error_errno(errno
, "Failed to rewind serialization fd: %m");
1218 r
= fd_cloexec(fileno(f
), false);
1220 return log_error_errno(r
, "Failed to disable O_CLOEXEC for serialization: %m");
1222 r
= fdset_cloexec(fds
, false);
1224 return log_error_errno(r
, "Failed to disable O_CLOEXEC for serialization fds: %m");
1226 *ret_f
= TAKE_PTR(f
);
1227 *ret_fds
= TAKE_PTR(fds
);
1232 static void bump_file_max_and_nr_open(void) {
1234 /* Let's bump fs.file-max and fs.nr_open to their respective maximums. On current kernels large numbers of file
1235 * descriptors are no longer a performance problem and their memory is properly tracked by memcg, thus counting
1236 * them and limiting them in another two layers of limits is unnecessary and just complicates things. This
1237 * function hence turns off 2 of the 4 levels of limits on file descriptors, and makes RLIMIT_NOLIMIT (soft +
1238 * hard) the only ones that really matter. */
1240 #if BUMP_PROC_SYS_FS_FILE_MAX || BUMP_PROC_SYS_FS_NR_OPEN
1241 _cleanup_free_
char *t
= NULL
;
1245 #if BUMP_PROC_SYS_FS_FILE_MAX
1246 /* I so wanted to use STRINGIFY(ULONG_MAX) here, but alas we can't as glibc/gcc define that as
1247 * "(0x7fffffffffffffffL * 2UL + 1UL)". Seriously. 😢 */
1248 if (asprintf(&t
, "%lu\n", ULONG_MAX
) < 0) {
1253 r
= sysctl_write("fs/file-max", t
);
1255 log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
, "Failed to bump fs.file-max, ignoring: %m");
1258 #if BUMP_PROC_SYS_FS_FILE_MAX && BUMP_PROC_SYS_FS_NR_OPEN
1262 #if BUMP_PROC_SYS_FS_NR_OPEN
1265 /* Arg! The kernel enforces maximum and minimum values on the fs.nr_open, but we don't really know what they
1266 * are. The expression by which the maximum is determined is dependent on the architecture, and is something we
1267 * don't really want to copy to userspace, as it is dependent on implementation details of the kernel. Since
1268 * the kernel doesn't expose the maximum value to us, we can only try and hope. Hence, let's start with
1269 * INT_MAX, and then keep halving the value until we find one that works. Ugly? Yes, absolutely, but kernel
1270 * APIs are kernel APIs, so what do can we do... 🤯 */
1275 v
&= ~(__SIZEOF_POINTER__
- 1); /* Round down to next multiple of the pointer size */
1277 log_warning("Can't bump fs.nr_open, value too small.");
1283 log_error_errno(k
, "Failed to read fs.nr_open: %m");
1286 if (k
>= v
) { /* Already larger */
1287 log_debug("Skipping bump, value is already larger.");
1291 if (asprintf(&t
, "%i\n", v
) < 0) {
1296 r
= sysctl_write("fs/nr_open", t
);
1299 log_debug("Couldn't write fs.nr_open as %i, halving it.", v
);
1304 log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
, "Failed to bump fs.nr_open, ignoring: %m");
1308 log_debug("Successfully bumped fs.nr_open to %i", v
);
1314 static int bump_rlimit_nofile(struct rlimit
*saved_rlimit
) {
1315 struct rlimit new_rlimit
;
1318 assert(saved_rlimit
);
1320 /* Save the original RLIMIT_NOFILE so that we can reset it later when transitioning from the initrd to the main
1321 * systemd or suchlike. */
1322 if (getrlimit(RLIMIT_NOFILE
, saved_rlimit
) < 0)
1323 return log_warning_errno(errno
, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1325 /* Get the underlying absolute limit the kernel enforces */
1326 nr
= read_nr_open();
1328 /* Make sure forked processes get limits based on the original kernel setting */
1329 if (!arg_default_rlimit
[RLIMIT_NOFILE
]) {
1332 rl
= newdup(struct rlimit
, saved_rlimit
, 1);
1336 /* Bump the hard limit for system services to a substantially higher value. The default hard limit
1337 * current kernels set is pretty low (4K), mostly for historical reasons. According to kernel
1338 * developers, the fd handling in recent kernels has been optimized substantially enough, so that we
1339 * can bump the limit now, without paying too high a price in memory or performance. Note however that
1340 * we only bump the hard limit, not the soft limit. That's because select() works the way it works, and
1341 * chokes on fds >= 1024. If we'd bump the soft limit globally, it might accidentally happen to
1342 * unexpecting programs that they get fds higher than what they can process using select(). By only
1343 * bumping the hard limit but leaving the low limit as it is we avoid this pitfall: programs that are
1344 * written by folks aware of the select() problem in mind (and thus use poll()/epoll instead of
1345 * select(), the way everybody should) can explicitly opt into high fds by bumping their soft limit
1346 * beyond 1024, to the hard limit we pass. */
1348 rl
->rlim_max
= MIN((rlim_t
) nr
, MAX(rl
->rlim_max
, (rlim_t
) HIGH_RLIMIT_NOFILE
));
1350 /* If for some reason we were invoked with a soft limit above 1024 (which should never
1351 * happen!, but who knows what we get passed in from pam_limit when invoked as --user
1352 * instance), then lower what we pass on to not confuse our children */
1353 rl
->rlim_cur
= MIN(rl
->rlim_cur
, (rlim_t
) FD_SETSIZE
);
1355 arg_default_rlimit
[RLIMIT_NOFILE
] = rl
;
1358 /* Calculate the new limits to use for us. Never lower from what we inherited. */
1359 new_rlimit
= (struct rlimit
) {
1360 .rlim_cur
= MAX((rlim_t
) nr
, saved_rlimit
->rlim_cur
),
1361 .rlim_max
= MAX((rlim_t
) nr
, saved_rlimit
->rlim_max
),
1364 /* Shortcut if nothing changes. */
1365 if (saved_rlimit
->rlim_max
>= new_rlimit
.rlim_max
&&
1366 saved_rlimit
->rlim_cur
>= new_rlimit
.rlim_cur
) {
1367 log_debug("RLIMIT_NOFILE is already as high or higher than we need it, not bumping.");
1371 /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for
1372 * both hard and soft. */
1373 r
= setrlimit_closest(RLIMIT_NOFILE
, &new_rlimit
);
1375 return log_warning_errno(r
, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1380 static int bump_rlimit_memlock(struct rlimit
*saved_rlimit
) {
1381 struct rlimit new_rlimit
;
1384 assert(saved_rlimit
);
1386 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK which should
1387 * normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's bump
1388 * the value high enough for our user. */
1390 if (getrlimit(RLIMIT_MEMLOCK
, saved_rlimit
) < 0)
1391 return log_warning_errno(errno
, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1393 /* Pass the original value down to invoked processes */
1394 if (!arg_default_rlimit
[RLIMIT_MEMLOCK
]) {
1397 rl
= newdup(struct rlimit
, saved_rlimit
, 1);
1401 arg_default_rlimit
[RLIMIT_MEMLOCK
] = rl
;
1404 /* Using MAX() on resource limits only is safe if RLIM_INFINITY is > 0. POSIX declares that rlim_t
1405 * must be unsigned, hence this is a given, but let's make this clear here. */
1406 assert_cc(RLIM_INFINITY
> 0);
1408 new_rlimit
= (struct rlimit
) {
1409 .rlim_cur
= MAX(HIGH_RLIMIT_MEMLOCK
, saved_rlimit
->rlim_cur
),
1410 .rlim_max
= MAX(HIGH_RLIMIT_MEMLOCK
, saved_rlimit
->rlim_max
),
1413 if (saved_rlimit
->rlim_max
>= new_rlimit
.rlim_cur
&&
1414 saved_rlimit
->rlim_cur
>= new_rlimit
.rlim_max
) {
1415 log_debug("RLIMIT_MEMLOCK is already as high or higher than we need it, not bumping.");
1419 r
= setrlimit_closest(RLIMIT_MEMLOCK
, &new_rlimit
);
1421 return log_warning_errno(r
, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1426 static void test_usr(void) {
1428 /* Check that /usr is either on the same file system as / or mounted already. */
1430 if (dir_is_empty("/usr") <= 0)
1433 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1434 "Some things will probably break (sometimes even silently) in mysterious ways. "
1435 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1438 static int enforce_syscall_archs(Set
*archs
) {
1442 if (!is_seccomp_available())
1445 r
= seccomp_restrict_archs(arg_syscall_archs
);
1447 return log_error_errno(r
, "Failed to enforce system call architecture restrication: %m");
1452 static int status_welcome(void) {
1453 _cleanup_free_
char *pretty_name
= NULL
, *ansi_color
= NULL
;
1456 if (IN_SET(arg_show_status
, SHOW_STATUS_NO
, SHOW_STATUS_AUTO
))
1459 r
= parse_os_release(NULL
,
1460 "PRETTY_NAME", &pretty_name
,
1461 "ANSI_COLOR", &ansi_color
,
1464 log_full_errno(r
== -ENOENT
? LOG_DEBUG
: LOG_WARNING
, r
,
1465 "Failed to read os-release file, ignoring: %m");
1467 if (log_get_show_color())
1468 return status_printf(NULL
, 0,
1469 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1470 isempty(ansi_color
) ? "1" : ansi_color
,
1471 isempty(pretty_name
) ? "Linux" : pretty_name
);
1473 return status_printf(NULL
, 0,
1474 "\nWelcome to %s!\n",
1475 isempty(pretty_name
) ? "Linux" : pretty_name
);
1478 static int write_container_id(void) {
1482 c
= getenv("container");
1486 RUN_WITH_UMASK(0022)
1487 r
= write_string_file("/run/systemd/container", c
, WRITE_STRING_FILE_CREATE
);
1489 return log_warning_errno(r
, "Failed to write /run/systemd/container, ignoring: %m");
1494 static int bump_unix_max_dgram_qlen(void) {
1495 _cleanup_free_
char *qlen
= NULL
;
1499 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1500 * really really early during boot, so that it is actually applied to all our sockets, including the
1501 * $NOTIFY_SOCKET one. */
1503 r
= read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen
);
1505 return log_full_errno(r
== -ENOENT
? LOG_DEBUG
: LOG_WARNING
, r
, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1507 r
= safe_atolu(qlen
, &v
);
1509 return log_warning_errno(r
, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen
);
1511 if (v
>= DEFAULT_UNIX_MAX_DGRAM_QLEN
)
1514 r
= write_string_filef("/proc/sys/net/unix/max_dgram_qlen", WRITE_STRING_FILE_DISABLE_BUFFER
, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN
);
1516 return log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
,
1517 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1522 static int fixup_environment(void) {
1523 _cleanup_free_
char *term
= NULL
;
1527 /* Only fix up the environment when we are started as PID 1 */
1528 if (getpid_cached() != 1)
1531 /* We expect the environment to be set correctly if run inside a container. */
1532 if (detect_container() > 0)
1535 /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1536 * device used by the console. We try to make a better guess here since some consoles might not have support
1537 * for color mode for example.
1539 * However if TERM was configured through the kernel command line then leave it alone. */
1540 r
= proc_cmdline_get_key("TERM", 0, &term
);
1544 t
= term
?: default_term_for_tty("/dev/console");
1546 if (setenv("TERM", t
, 1) < 0)
1549 /* The kernels sets HOME=/ for init. Let's undo this. */
1550 if (path_equal_ptr(getenv("HOME"), "/") &&
1551 unsetenv("HOME") < 0)
1552 log_warning_errno(errno
, "Failed to unset $HOME: %m");
1557 static void redirect_telinit(int argc
, char *argv
[]) {
1559 /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1561 #if HAVE_SYSV_COMPAT
1562 if (getpid_cached() == 1)
1565 if (!strstr(program_invocation_short_name
, "init"))
1568 execv(SYSTEMCTL_BINARY_PATH
, argv
);
1569 log_error_errno(errno
, "Failed to exec " SYSTEMCTL_BINARY_PATH
": %m");
1574 static int become_shutdown(
1575 const char *shutdown_verb
,
1578 char log_level
[DECIMAL_STR_MAX(int) + 1],
1579 exit_code
[DECIMAL_STR_MAX(uint8_t) + 1],
1580 timeout
[DECIMAL_STR_MAX(usec_t
) + 1];
1582 const char* command_line
[13] = {
1583 SYSTEMD_SHUTDOWN_BINARY_PATH
,
1585 "--timeout", timeout
,
1586 "--log-level", log_level
,
1590 _cleanup_strv_free_
char **env_block
= NULL
;
1594 assert(shutdown_verb
);
1595 assert(!command_line
[pos
]);
1596 env_block
= strv_copy(environ
);
1598 xsprintf(log_level
, "%d", log_get_max_level());
1599 xsprintf(timeout
, "%" PRI_USEC
"us", arg_default_timeout_stop_usec
);
1601 switch (log_get_target()) {
1603 case LOG_TARGET_KMSG
:
1604 case LOG_TARGET_JOURNAL_OR_KMSG
:
1605 case LOG_TARGET_SYSLOG_OR_KMSG
:
1606 command_line
[pos
++] = "kmsg";
1609 case LOG_TARGET_NULL
:
1610 command_line
[pos
++] = "null";
1613 case LOG_TARGET_CONSOLE
:
1615 command_line
[pos
++] = "console";
1619 if (log_get_show_color())
1620 command_line
[pos
++] = "--log-color";
1622 if (log_get_show_location())
1623 command_line
[pos
++] = "--log-location";
1625 if (streq(shutdown_verb
, "exit")) {
1626 command_line
[pos
++] = "--exit-code";
1627 command_line
[pos
++] = exit_code
;
1628 xsprintf(exit_code
, "%d", retval
);
1631 assert(pos
< ELEMENTSOF(command_line
));
1633 if (streq(shutdown_verb
, "reboot") &&
1634 arg_shutdown_watchdog
> 0 &&
1635 arg_shutdown_watchdog
!= USEC_INFINITY
) {
1639 /* If we reboot let's set the shutdown
1640 * watchdog and tell the shutdown binary to
1641 * repeatedly ping it */
1642 r
= watchdog_set_timeout(&arg_shutdown_watchdog
);
1643 watchdog_close(r
< 0);
1645 /* Tell the binary how often to ping, ignore failure */
1646 if (asprintf(&e
, "WATCHDOG_USEC="USEC_FMT
, arg_shutdown_watchdog
) > 0)
1647 (void) strv_consume(&env_block
, e
);
1649 if (arg_watchdog_device
&&
1650 asprintf(&e
, "WATCHDOG_DEVICE=%s", arg_watchdog_device
) > 0)
1651 (void) strv_consume(&env_block
, e
);
1653 watchdog_close(true);
1655 /* Avoid the creation of new processes forked by the
1656 * kernel; at this point, we will not listen to the
1658 if (detect_container() <= 0)
1659 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER
);
1661 execve(SYSTEMD_SHUTDOWN_BINARY_PATH
, (char **) command_line
, env_block
);
1665 static void initialize_clock(void) {
1668 if (clock_is_localtime(NULL
) > 0) {
1672 * The very first call of settimeofday() also does a time warp in the kernel.
1674 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1675 * of maintaining the RTC and do all adjustments. This matches the behavior of Windows, which leaves
1676 * the RTC alone if the registry tells that the RTC runs in UTC.
1678 r
= clock_set_timezone(&min
);
1680 log_error_errno(r
, "Failed to apply local time delta, ignoring: %m");
1682 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min
);
1684 } else if (!in_initrd()) {
1686 * Do a dummy very first call to seal the kernel's time warp magic.
1688 * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1689 * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1690 * until we reach the real system.
1692 * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1693 * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1694 * be treated as UTC that way.
1696 (void) clock_reset_timewarp();
1699 r
= clock_apply_epoch();
1701 log_error_errno(r
, "Current system time is before build time, but cannot correct: %m");
1703 log_info("System time before build time, advancing clock.");
1706 static void initialize_coredump(bool skip_setup
) {
1708 if (getpid_cached() != 1)
1711 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1712 * will process core dumps for system services by default. */
1713 if (setrlimit(RLIMIT_CORE
, &RLIMIT_MAKE_CONST(RLIM_INFINITY
)) < 0)
1714 log_warning_errno(errno
, "Failed to set RLIMIT_CORE: %m");
1716 /* But at the same time, turn off the core_pattern logic by default, so that no
1717 * coredumps are stored until the systemd-coredump tool is enabled via
1718 * sysctl. However it can be changed via the kernel command line later so core
1719 * dumps can still be generated during early startup and in initramfs. */
1721 disable_coredumps();
1725 static void initialize_core_pattern(bool skip_setup
) {
1728 if (skip_setup
|| !arg_early_core_pattern
)
1731 if (getpid_cached() != 1)
1734 r
= write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern
, WRITE_STRING_FILE_DISABLE_BUFFER
);
1736 log_warning_errno(r
, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", arg_early_core_pattern
);
1739 static void do_reexecute(
1742 const struct rlimit
*saved_rlimit_nofile
,
1743 const struct rlimit
*saved_rlimit_memlock
,
1745 const char *switch_root_dir
,
1746 const char *switch_root_init
,
1747 const char **ret_error_message
) {
1749 unsigned i
, j
, args_size
;
1753 assert(saved_rlimit_nofile
);
1754 assert(saved_rlimit_memlock
);
1755 assert(ret_error_message
);
1757 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1759 watchdog_close(true);
1761 /* Reset RLIMIT_NOFILE + RLIMIT_MEMLOCK back to the kernel defaults, so that the new systemd can pass
1762 * the kernel default to its child processes */
1763 if (saved_rlimit_nofile
->rlim_cur
!= 0)
1764 (void) setrlimit(RLIMIT_NOFILE
, saved_rlimit_nofile
);
1765 if (saved_rlimit_memlock
->rlim_cur
!= RLIM_INFINITY
)
1766 (void) setrlimit(RLIMIT_MEMLOCK
, saved_rlimit_memlock
);
1768 if (switch_root_dir
) {
1769 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1770 * SIGCHLD for them after deserializing. */
1771 broadcast_signal(SIGTERM
, false, true, arg_default_timeout_stop_usec
);
1773 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1774 r
= switch_root(switch_root_dir
, "/mnt", true, MS_MOVE
);
1776 log_error_errno(r
, "Failed to switch root, trying to continue: %m");
1779 args_size
= MAX(6, argc
+1);
1780 args
= newa(const char*, args_size
);
1782 if (!switch_root_init
) {
1783 char sfd
[DECIMAL_STR_MAX(int) + 1];
1785 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1786 * the user didn't specify an explicit init to spawn. */
1788 assert(arg_serialization
);
1791 xsprintf(sfd
, "%i", fileno(arg_serialization
));
1794 args
[i
++] = SYSTEMD_BINARY_PATH
;
1795 if (switch_root_dir
)
1796 args
[i
++] = "--switched-root";
1797 args
[i
++] = arg_system
? "--system" : "--user";
1798 args
[i
++] = "--deserialize";
1802 assert(i
<= args_size
);
1805 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do this is on
1806 * its own on exec(), but it will do it on exit(). Hence, to ensure we get a summary here, fork() off
1807 * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1808 * before proceeding into the exec().
1810 valgrind_summary_hack();
1812 (void) execv(args
[0], (char* const*) args
);
1813 log_debug_errno(errno
, "Failed to execute our own binary, trying fallback: %m");
1816 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1817 * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1818 * doesn't matter.) */
1820 arg_serialization
= safe_fclose(arg_serialization
);
1821 fds
= fdset_free(fds
);
1823 /* Reopen the console */
1824 (void) make_console_stdio();
1826 for (j
= 1, i
= 1; j
< (unsigned) argc
; j
++)
1827 args
[i
++] = argv
[j
];
1829 assert(i
<= args_size
);
1831 /* Re-enable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1832 (void) reset_all_signal_handlers();
1833 (void) reset_signal_mask();
1834 (void) rlimit_nofile_safe();
1836 if (switch_root_init
) {
1837 args
[0] = switch_root_init
;
1838 (void) execv(args
[0], (char* const*) args
);
1839 log_warning_errno(errno
, "Failed to execute configured init, trying fallback: %m");
1842 args
[0] = "/sbin/init";
1843 (void) execv(args
[0], (char* const*) args
);
1846 manager_status_printf(NULL
, STATUS_TYPE_EMERGENCY
,
1847 ANSI_HIGHLIGHT_RED
" !! " ANSI_NORMAL
,
1848 "Failed to execute /sbin/init");
1851 log_warning("No /sbin/init, trying fallback");
1853 args
[0] = "/bin/sh";
1855 (void) execv(args
[0], (char* const*) args
);
1856 log_error_errno(errno
, "Failed to execute /bin/sh, giving up: %m");
1858 log_warning_errno(r
, "Failed to execute /sbin/init, giving up: %m");
1860 *ret_error_message
= "Failed to execute fallback shell";
1863 static int invoke_main_loop(
1865 bool *ret_reexecute
,
1866 int *ret_retval
, /* Return parameters relevant for shutting down */
1867 const char **ret_shutdown_verb
, /* … */
1868 FDSet
**ret_fds
, /* Return parameters for reexecuting */
1869 char **ret_switch_root_dir
, /* … */
1870 char **ret_switch_root_init
, /* … */
1871 const char **ret_error_message
) {
1876 assert(ret_reexecute
);
1878 assert(ret_shutdown_verb
);
1880 assert(ret_switch_root_dir
);
1881 assert(ret_switch_root_init
);
1882 assert(ret_error_message
);
1885 r
= manager_loop(m
);
1887 *ret_error_message
= "Failed to run main loop";
1888 return log_emergency_errno(r
, "Failed to run main loop: %m");
1891 switch ((ManagerObjective
) r
) {
1893 case MANAGER_RELOAD
: {
1894 LogTarget saved_log_target
;
1895 int saved_log_level
;
1897 log_info("Reloading.");
1899 /* First, save any overridden log level/target, then parse the configuration file, which might
1900 * change the log level to new settings. */
1902 saved_log_level
= m
->log_level_overridden
? log_get_max_level() : -1;
1903 saved_log_target
= m
->log_target_overridden
? log_get_target() : _LOG_TARGET_INVALID
;
1905 r
= parse_config_file();
1907 log_warning_errno(r
, "Failed to parse config file, ignoring: %m");
1909 set_manager_defaults(m
);
1911 if (saved_log_level
>= 0)
1912 manager_override_log_level(m
, saved_log_level
);
1913 if (saved_log_target
>= 0)
1914 manager_override_log_target(m
, saved_log_target
);
1916 r
= manager_reload(m
);
1918 /* Reloading failed before the point of no return. Let's continue running as if nothing happened. */
1919 m
->objective
= MANAGER_OK
;
1924 case MANAGER_REEXECUTE
:
1926 r
= prepare_reexecute(m
, &arg_serialization
, ret_fds
, false);
1928 *ret_error_message
= "Failed to prepare for reexecution";
1932 log_notice("Reexecuting.");
1934 *ret_reexecute
= true;
1935 *ret_retval
= EXIT_SUCCESS
;
1936 *ret_shutdown_verb
= NULL
;
1937 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
1941 case MANAGER_SWITCH_ROOT
:
1942 if (!m
->switch_root_init
) {
1943 r
= prepare_reexecute(m
, &arg_serialization
, ret_fds
, true);
1945 *ret_error_message
= "Failed to prepare for reexecution";
1951 log_notice("Switching root.");
1953 *ret_reexecute
= true;
1954 *ret_retval
= EXIT_SUCCESS
;
1955 *ret_shutdown_verb
= NULL
;
1957 /* Steal the switch root parameters */
1958 *ret_switch_root_dir
= TAKE_PTR(m
->switch_root
);
1959 *ret_switch_root_init
= TAKE_PTR(m
->switch_root_init
);
1965 if (MANAGER_IS_USER(m
)) {
1968 *ret_reexecute
= false;
1969 *ret_retval
= m
->return_value
;
1970 *ret_shutdown_verb
= NULL
;
1972 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
1978 case MANAGER_REBOOT
:
1979 case MANAGER_POWEROFF
:
1981 case MANAGER_KEXEC
: {
1982 static const char * const table
[_MANAGER_OBJECTIVE_MAX
] = {
1983 [MANAGER_EXIT
] = "exit",
1984 [MANAGER_REBOOT
] = "reboot",
1985 [MANAGER_POWEROFF
] = "poweroff",
1986 [MANAGER_HALT
] = "halt",
1987 [MANAGER_KEXEC
] = "kexec",
1990 log_notice("Shutting down.");
1992 *ret_reexecute
= false;
1993 *ret_retval
= m
->return_value
;
1994 assert_se(*ret_shutdown_verb
= table
[m
->objective
]);
1996 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
2002 assert_not_reached("Unknown or unexpected manager objective.");
2007 static void log_execution_mode(bool *ret_first_boot
) {
2008 assert(ret_first_boot
);
2013 log_info("systemd " GIT_VERSION
" running in %ssystem mode. (" SYSTEMD_FEATURES
")",
2014 arg_action
== ACTION_TEST
? "test " : "" );
2016 v
= detect_virtualization();
2018 log_info("Detected virtualization %s.", virtualization_to_string(v
));
2020 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
2023 *ret_first_boot
= false;
2024 log_info("Running in initial RAM disk.");
2026 /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
2027 * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
2028 * doesn't it's unpopulated. This allows container managers and installers to provision a
2029 * couple of files already. If the container manager wants to provision the machine ID itself
2030 * it should pass $container_uuid to PID 1. */
2032 *ret_first_boot
= access("/etc/machine-id", F_OK
) < 0;
2033 if (*ret_first_boot
)
2034 log_info("Running with unpopulated /etc.");
2037 if (DEBUG_LOGGING
) {
2038 _cleanup_free_
char *t
;
2040 t
= uid_to_name(getuid());
2041 log_debug("systemd " GIT_VERSION
" running in %suser mode for user " UID_FMT
"/%s. (" SYSTEMD_FEATURES
")",
2042 arg_action
== ACTION_TEST
? " test" : "", getuid(), strna(t
));
2045 *ret_first_boot
= false;
2049 static int initialize_runtime(
2051 struct rlimit
*saved_rlimit_nofile
,
2052 struct rlimit
*saved_rlimit_memlock
,
2053 const char **ret_error_message
) {
2057 assert(ret_error_message
);
2059 /* Sets up various runtime parameters. Many of these initializations are conditionalized:
2061 * - Some only apply to --system instances
2062 * - Some only apply to --user instances
2063 * - Some only apply when we first start up, but not when we reexecute
2066 if (arg_action
!= ACTION_RUN
)
2070 /* Make sure we leave a core dump without panicking the kernel. */
2071 install_crash_handler();
2074 r
= mount_cgroup_controllers();
2076 *ret_error_message
= "Failed to mount cgroup hierarchies";
2082 machine_id_setup(NULL
, arg_machine_id
, NULL
);
2084 bump_unix_max_dgram_qlen();
2085 bump_file_max_and_nr_open();
2087 write_container_id();
2090 if (arg_watchdog_device
) {
2091 r
= watchdog_set_device(arg_watchdog_device
);
2093 log_warning_errno(r
, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device
);
2096 if (arg_runtime_watchdog
> 0 && arg_runtime_watchdog
!= USEC_INFINITY
)
2097 watchdog_set_timeout(&arg_runtime_watchdog
);
2100 if (arg_timer_slack_nsec
!= NSEC_INFINITY
)
2101 if (prctl(PR_SET_TIMERSLACK
, arg_timer_slack_nsec
) < 0)
2102 log_warning_errno(errno
, "Failed to adjust timer slack, ignoring: %m");
2104 if (arg_system
&& !cap_test_all(arg_capability_bounding_set
)) {
2105 r
= capability_bounding_set_drop_usermode(arg_capability_bounding_set
);
2107 *ret_error_message
= "Failed to drop capability bounding set of usermode helpers";
2108 return log_emergency_errno(r
, "Failed to drop capability bounding set of usermode helpers: %m");
2111 r
= capability_bounding_set_drop(arg_capability_bounding_set
, true);
2113 *ret_error_message
= "Failed to drop capability bounding set";
2114 return log_emergency_errno(r
, "Failed to drop capability bounding set: %m");
2118 if (arg_system
&& arg_no_new_privs
) {
2119 if (prctl(PR_SET_NO_NEW_PRIVS
, 1, 0, 0, 0) < 0) {
2120 *ret_error_message
= "Failed to disable new privileges";
2121 return log_emergency_errno(errno
, "Failed to disable new privileges: %m");
2125 if (arg_syscall_archs
) {
2126 r
= enforce_syscall_archs(arg_syscall_archs
);
2128 *ret_error_message
= "Failed to set syscall architectures";
2134 /* Become reaper of our children */
2135 if (prctl(PR_SET_CHILD_SUBREAPER
, 1) < 0)
2136 log_warning_errno(errno
, "Failed to make us a subreaper: %m");
2138 /* Bump up RLIMIT_NOFILE for systemd itself */
2139 (void) bump_rlimit_nofile(saved_rlimit_nofile
);
2140 (void) bump_rlimit_memlock(saved_rlimit_memlock
);
2145 static int do_queue_default_job(
2147 const char **ret_error_message
) {
2149 _cleanup_(sd_bus_error_free
) sd_bus_error error
= SD_BUS_ERROR_NULL
;
2150 Job
*default_unit_job
;
2151 Unit
*target
= NULL
;
2154 log_debug("Activating default unit: %s", arg_default_unit
);
2156 r
= manager_load_startable_unit_or_warn(m
, arg_default_unit
, NULL
, &target
);
2158 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET
);
2160 r
= manager_load_startable_unit_or_warn(m
, SPECIAL_RESCUE_TARGET
, NULL
, &target
);
2162 *ret_error_message
= r
== -ERFKILL
? "Rescue target masked"
2163 : "Failed to load rescue target";
2168 assert(target
->load_state
== UNIT_LOADED
);
2170 r
= manager_add_job(m
, JOB_START
, target
, JOB_ISOLATE
, NULL
, &error
, &default_unit_job
);
2172 log_debug_errno(r
, "Default target could not be isolated, starting instead: %s", bus_error_message(&error
, r
));
2174 sd_bus_error_free(&error
);
2176 r
= manager_add_job(m
, JOB_START
, target
, JOB_REPLACE
, NULL
, &error
, &default_unit_job
);
2178 *ret_error_message
= "Failed to start default target";
2179 return log_emergency_errno(r
, "Failed to start default target: %s", bus_error_message(&error
, r
));
2183 *ret_error_message
= "Failed to isolate default target";
2184 return log_emergency_errno(r
, "Failed to isolate default target: %s", bus_error_message(&error
, r
));
2187 m
->default_unit_job_id
= default_unit_job
->id
;
2192 static void free_arguments(void) {
2194 /* Frees all arg_* variables, with the exception of arg_serialization */
2195 rlimit_free_all(arg_default_rlimit
);
2197 arg_default_unit
= mfree(arg_default_unit
);
2198 arg_confirm_spawn
= mfree(arg_confirm_spawn
);
2199 arg_default_environment
= strv_free(arg_default_environment
);
2200 arg_syscall_archs
= set_free(arg_syscall_archs
);
2203 static int load_configuration(int argc
, char **argv
, const char **ret_error_message
) {
2206 assert(ret_error_message
);
2208 arg_default_tasks_max
= system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE
, 100U);
2210 r
= parse_config_file();
2212 *ret_error_message
= "Failed to parse config file";
2217 r
= proc_cmdline_parse(parse_proc_cmdline_item
, NULL
, 0);
2219 log_warning_errno(r
, "Failed to parse kernel command line, ignoring: %m");
2222 /* Note that this also parses bits from the kernel command line, including "debug". */
2223 log_parse_environment();
2225 r
= parse_argv(argc
, argv
);
2227 *ret_error_message
= "Failed to parse commandline arguments";
2231 /* Initialize default unit */
2232 if (!arg_default_unit
) {
2233 arg_default_unit
= strdup(SPECIAL_DEFAULT_TARGET
);
2234 if (!arg_default_unit
) {
2235 *ret_error_message
= "Failed to set default unit";
2240 /* Initialize the show status setting if it hasn't been set explicitly yet */
2241 if (arg_show_status
== _SHOW_STATUS_INVALID
)
2242 arg_show_status
= SHOW_STATUS_YES
;
2247 static int safety_checks(void) {
2249 if (getpid_cached() == 1 &&
2250 arg_action
!= ACTION_RUN
)
2251 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
2252 "Unsupported execution mode while PID 1.");
2254 if (getpid_cached() == 1 &&
2256 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
2257 "Can't run --user mode as PID 1.");
2259 if (arg_action
== ACTION_RUN
&&
2261 getpid_cached() != 1)
2262 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
2263 "Can't run system mode unless PID 1.");
2265 if (arg_action
== ACTION_TEST
&&
2267 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
2268 "Don't run test mode as root.");
2271 arg_action
== ACTION_RUN
&&
2273 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP
),
2274 "Trying to run as user instance, but the system has not been booted with systemd.");
2277 arg_action
== ACTION_RUN
&&
2278 !getenv("XDG_RUNTIME_DIR"))
2279 return log_error_errno(SYNTHETIC_ERRNO(EUNATCH
),
2280 "Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
2283 arg_action
== ACTION_RUN
&&
2284 running_in_chroot() > 0)
2285 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP
),
2286 "Cannot be run in a chroot() environment.");
2291 static int initialize_security(
2292 bool *loaded_policy
,
2293 dual_timestamp
*security_start_timestamp
,
2294 dual_timestamp
*security_finish_timestamp
,
2295 const char **ret_error_message
) {
2299 assert(loaded_policy
);
2300 assert(security_start_timestamp
);
2301 assert(security_finish_timestamp
);
2302 assert(ret_error_message
);
2304 dual_timestamp_get(security_start_timestamp
);
2306 r
= mac_selinux_setup(loaded_policy
);
2308 *ret_error_message
= "Failed to load SELinux policy";
2312 r
= mac_smack_setup(loaded_policy
);
2314 *ret_error_message
= "Failed to load SMACK policy";
2320 *ret_error_message
= "Failed to load IMA policy";
2324 dual_timestamp_get(security_finish_timestamp
);
2328 static void test_summary(Manager
*m
) {
2331 printf("-> By units:\n");
2332 manager_dump_units(m
, stdout
, "\t");
2334 printf("-> By jobs:\n");
2335 manager_dump_jobs(m
, stdout
, "\t");
2338 static int collect_fds(FDSet
**ret_fds
, const char **ret_error_message
) {
2342 assert(ret_error_message
);
2344 r
= fdset_new_fill(ret_fds
);
2346 *ret_error_message
= "Failed to allocate fd set";
2347 return log_emergency_errno(r
, "Failed to allocate fd set: %m");
2350 fdset_cloexec(*ret_fds
, true);
2352 if (arg_serialization
)
2353 assert_se(fdset_remove(*ret_fds
, fileno(arg_serialization
)) >= 0);
2358 static void setup_console_terminal(bool skip_setup
) {
2363 /* Become a session leader if we aren't one yet. */
2366 /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2368 (void) release_terminal();
2370 /* Reset the console, but only if this is really init and we are freshly booted */
2371 if (getpid_cached() == 1 && !skip_setup
)
2372 (void) console_setup();
2375 static bool early_skip_setup_check(int argc
, char *argv
[]) {
2376 bool found_deserialize
= false;
2379 /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2380 * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2381 * anyway, even if in that case we also do deserialization. */
2383 for (i
= 1; i
< argc
; i
++) {
2384 if (streq(argv
[i
], "--switched-root"))
2385 return false; /* If we switched root, don't skip the setup. */
2386 else if (streq(argv
[i
], "--deserialize"))
2387 found_deserialize
= true;
2390 return found_deserialize
; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2393 int main(int argc
, char *argv
[]) {
2395 dual_timestamp initrd_timestamp
= DUAL_TIMESTAMP_NULL
, userspace_timestamp
= DUAL_TIMESTAMP_NULL
, kernel_timestamp
= DUAL_TIMESTAMP_NULL
,
2396 security_start_timestamp
= DUAL_TIMESTAMP_NULL
, security_finish_timestamp
= DUAL_TIMESTAMP_NULL
;
2397 struct rlimit saved_rlimit_nofile
= RLIMIT_MAKE_CONST(0),
2398 saved_rlimit_memlock
= RLIMIT_MAKE_CONST(RLIM_INFINITY
); /* The original rlimits we passed
2399 * in. Note we use different values
2400 * for the two that indicate whether
2401 * these fields are initialized! */
2402 bool skip_setup
, loaded_policy
= false, queue_default_job
= false, first_boot
= false, reexecute
= false;
2403 char *switch_root_dir
= NULL
, *switch_root_init
= NULL
;
2404 usec_t before_startup
, after_startup
;
2405 static char systemd
[] = "systemd";
2406 char timespan
[FORMAT_TIMESPAN_MAX
];
2407 const char *shutdown_verb
= NULL
, *error_message
= NULL
;
2408 int r
, retval
= EXIT_FAILURE
;
2412 /* SysV compatibility: redirect init → telinit */
2413 redirect_telinit(argc
, argv
);
2415 /* Take timestamps early on */
2416 dual_timestamp_from_monotonic(&kernel_timestamp
, 0);
2417 dual_timestamp_get(&userspace_timestamp
);
2419 /* Figure out whether we need to do initialize the system, or if we already did that because we are
2421 skip_setup
= early_skip_setup_check(argc
, argv
);
2423 /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2424 * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
2425 program_invocation_short_name
= systemd
;
2426 (void) prctl(PR_SET_NAME
, systemd
);
2428 /* Save the original command line */
2429 save_argc_argv(argc
, argv
);
2431 /* Make sure that if the user says "syslog" we actually log to the journal. */
2432 log_set_upgrade_syslog_to_journal(true);
2434 if (getpid_cached() == 1) {
2435 /* When we run as PID 1 force system mode */
2438 /* Disable the umask logic */
2441 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2442 * activated yet (even though the log socket for it exists). */
2443 log_set_prohibit_ipc(true);
2445 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2446 * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2447 * child process right before execve()'ing the actual binary, at a point in time where socket
2448 * activation stderr/stdout area already set up. */
2449 log_set_always_reopen_console(true);
2451 if (detect_container() <= 0) {
2453 /* Running outside of a container as PID 1 */
2454 log_set_target(LOG_TARGET_KMSG
);
2458 initrd_timestamp
= userspace_timestamp
;
2461 r
= mount_setup_early();
2463 error_message
= "Failed to mount early API filesystems";
2467 r
= initialize_security(
2469 &security_start_timestamp
,
2470 &security_finish_timestamp
,
2476 if (mac_selinux_init() < 0) {
2477 error_message
= "Failed to initialize SELinux policy";
2484 /* Set the default for later on, but don't actually open the logs like this for now. Note that
2485 * if we are transitioning from the initrd there might still be journal fd open, and we
2486 * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2488 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG
);
2491 /* Running inside a container, as PID 1 */
2492 log_set_target(LOG_TARGET_CONSOLE
);
2495 /* For later on, see above... */
2496 log_set_target(LOG_TARGET_JOURNAL
);
2498 /* clear the kernel timestamp,
2499 * because we are in a container */
2500 kernel_timestamp
= DUAL_TIMESTAMP_NULL
;
2503 initialize_coredump(skip_setup
);
2505 r
= fixup_environment();
2507 log_emergency_errno(r
, "Failed to fix up PID 1 environment: %m");
2508 error_message
= "Failed to fix up PID1 environment";
2513 /* Running as user instance */
2515 log_set_target(LOG_TARGET_AUTO
);
2518 /* clear the kernel timestamp,
2519 * because we are not PID 1 */
2520 kernel_timestamp
= DUAL_TIMESTAMP_NULL
;
2524 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2525 * they never log into the console. */
2526 log_show_color(colors_enabled());
2528 r
= make_null_stdio();
2530 log_warning_errno(r
, "Failed to redirect standard streams to /dev/null, ignoring: %m");
2533 /* Mount /proc, /sys and friends, so that /proc/cmdline and
2534 * /proc/$PID/fd is available. */
2535 if (getpid_cached() == 1) {
2537 /* Load the kernel modules early. */
2541 r
= mount_setup(loaded_policy
);
2543 error_message
= "Failed to mount API filesystems";
2548 /* Reset all signal handlers. */
2549 (void) reset_all_signal_handlers();
2550 (void) ignore_signals(SIGNALS_IGNORE
, -1);
2552 r
= load_configuration(argc
, argv
, &error_message
);
2556 r
= safety_checks();
2560 if (IN_SET(arg_action
, ACTION_TEST
, ACTION_HELP
, ACTION_DUMP_CONFIGURATION_ITEMS
, ACTION_DUMP_BUS_PROPERTIES
))
2561 (void) pager_open(arg_pager_flags
);
2563 if (arg_action
!= ACTION_RUN
)
2566 if (arg_action
== ACTION_HELP
) {
2567 retval
= help() < 0 ? EXIT_FAILURE
: EXIT_SUCCESS
;
2569 } else if (arg_action
== ACTION_VERSION
) {
2572 } else if (arg_action
== ACTION_DUMP_CONFIGURATION_ITEMS
) {
2573 unit_dump_config_items(stdout
);
2574 retval
= EXIT_SUCCESS
;
2576 } else if (arg_action
== ACTION_DUMP_BUS_PROPERTIES
) {
2577 dump_bus_properties(stdout
);
2578 retval
= EXIT_SUCCESS
;
2582 assert_se(IN_SET(arg_action
, ACTION_RUN
, ACTION_TEST
));
2584 /* Move out of the way, so that we won't block unmounts */
2585 assert_se(chdir("/") == 0);
2587 if (arg_action
== ACTION_RUN
) {
2589 /* A core pattern might have been specified via the cmdline. */
2590 initialize_core_pattern(skip_setup
);
2592 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2595 /* Remember open file descriptors for later deserialization */
2596 r
= collect_fds(&fds
, &error_message
);
2600 /* Give up any control of the console, but make sure its initialized. */
2601 setup_console_terminal(skip_setup
);
2603 /* Open the logging devices, if possible and necessary */
2607 log_execution_mode(&first_boot
);
2609 r
= initialize_runtime(skip_setup
,
2610 &saved_rlimit_nofile
,
2611 &saved_rlimit_memlock
,
2616 r
= manager_new(arg_system
? UNIT_FILE_SYSTEM
: UNIT_FILE_USER
,
2617 arg_action
== ACTION_TEST
? MANAGER_TEST_FULL
: 0,
2620 log_emergency_errno(r
, "Failed to allocate manager object: %m");
2621 error_message
= "Failed to allocate manager object";
2625 m
->timestamps
[MANAGER_TIMESTAMP_KERNEL
] = kernel_timestamp
;
2626 m
->timestamps
[MANAGER_TIMESTAMP_INITRD
] = initrd_timestamp
;
2627 m
->timestamps
[MANAGER_TIMESTAMP_USERSPACE
] = userspace_timestamp
;
2628 m
->timestamps
[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START
)] = security_start_timestamp
;
2629 m
->timestamps
[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH
)] = security_finish_timestamp
;
2631 set_manager_defaults(m
);
2632 set_manager_settings(m
);
2633 manager_set_first_boot(m
, first_boot
);
2635 /* Remember whether we should queue the default job */
2636 queue_default_job
= !arg_serialization
|| arg_switched_root
;
2638 before_startup
= now(CLOCK_MONOTONIC
);
2640 r
= manager_startup(m
, arg_serialization
, fds
);
2642 error_message
= "Failed to start up manager";
2646 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2647 fds
= fdset_free(fds
);
2648 arg_serialization
= safe_fclose(arg_serialization
);
2650 if (queue_default_job
) {
2651 r
= do_queue_default_job(m
, &error_message
);
2656 after_startup
= now(CLOCK_MONOTONIC
);
2658 log_full(arg_action
== ACTION_TEST
? LOG_INFO
: LOG_DEBUG
,
2659 "Loaded units and determined initial transaction in %s.",
2660 format_timespan(timespan
, sizeof(timespan
), after_startup
- before_startup
, 100 * USEC_PER_MSEC
));
2662 if (arg_action
== ACTION_TEST
) {
2664 retval
= EXIT_SUCCESS
;
2668 (void) invoke_main_loop(m
,
2681 arg_shutdown_watchdog
= m
->shutdown_watchdog
;
2682 m
= manager_free(m
);
2686 mac_selinux_finish();
2689 do_reexecute(argc
, argv
,
2690 &saved_rlimit_nofile
,
2691 &saved_rlimit_memlock
,
2695 &error_message
); /* This only returns if reexecution failed */
2697 arg_serialization
= safe_fclose(arg_serialization
);
2698 fds
= fdset_free(fds
);
2700 #if HAVE_VALGRIND_VALGRIND_H
2701 /* If we are PID 1 and running under valgrind, then let's exit
2702 * here explicitly. valgrind will only generate nice output on
2703 * exit(), not on exec(), hence let's do the former not the
2705 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND
) {
2706 /* Cleanup watchdog_device strings for valgrind. We need them
2707 * in become_shutdown() so normally we cannot free them yet. */
2708 watchdog_free_device();
2709 arg_watchdog_device
= mfree(arg_watchdog_device
);
2714 #if HAS_FEATURE_ADDRESS_SANITIZER
2715 __lsan_do_leak_check();
2718 if (shutdown_verb
) {
2719 r
= become_shutdown(shutdown_verb
, retval
);
2720 log_error_errno(r
, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
2721 error_message
= "Failed to execute shutdown binary";
2724 watchdog_free_device();
2725 arg_watchdog_device
= mfree(arg_watchdog_device
);
2727 if (getpid_cached() == 1) {
2729 manager_status_printf(NULL
, STATUS_TYPE_EMERGENCY
,
2730 ANSI_HIGHLIGHT_RED
"!!!!!!" ANSI_NORMAL
,
2731 "%s.", error_message
);
2732 freeze_or_exit_or_reboot();