1 /* SPDX-License-Identifier: LGPL-2.1+ */
10 #include <sys/prctl.h>
11 #include <sys/reboot.h>
17 #if HAVE_VALGRIND_VALGRIND_H
18 #include <valgrind/valgrind.h>
22 #include "sd-daemon.h"
23 #include "sd-messages.h"
25 #include "alloc-util.h"
26 #include "architecture.h"
28 #include "bus-error.h"
30 #include "capability-util.h"
31 #include "cgroup-util.h"
32 #include "clock-util.h"
33 #include "conf-parser.h"
34 #include "cpu-set-util.h"
36 #include "dbus-manager.h"
38 #include "emergency-action.h"
40 #include "exit-status.h"
44 #include "format-util.h"
46 #include "hostname-setup.h"
47 #include "ima-setup.h"
49 #include "kmod-setup.h"
50 #include "load-fragment.h"
52 #include "loopback-setup.h"
53 #include "machine-id-setup.h"
56 #include "mount-setup.h"
59 #include "parse-util.h"
60 #include "path-util.h"
61 #include "pretty-print.h"
62 #include "proc-cmdline.h"
63 #include "process-util.h"
64 #include "raw-clone.h"
65 #include "rlimit-util.h"
67 #include "seccomp-util.h"
69 #include "selinux-setup.h"
70 #include "selinux-util.h"
71 #include "signal-util.h"
72 #include "smack-setup.h"
74 #include "stat-util.h"
75 #include "stdio-util.h"
77 #include "switch-root.h"
78 #include "sysctl-util.h"
79 #include "terminal-util.h"
80 #include "umask-util.h"
81 #include "user-util.h"
86 #if HAS_FEATURE_ADDRESS_SANITIZER
87 #include <sanitizer/lsan_interface.h>
95 ACTION_DUMP_CONFIGURATION_ITEMS
,
96 ACTION_DUMP_BUS_PROPERTIES
,
97 } arg_action
= ACTION_RUN
;
98 static char *arg_default_unit
= NULL
;
99 static bool arg_system
= false;
100 static bool arg_dump_core
= true;
101 static int arg_crash_chvt
= -1;
102 static bool arg_crash_shell
= false;
103 static bool arg_crash_reboot
= false;
104 static char *arg_confirm_spawn
= NULL
;
105 static ShowStatus arg_show_status
= _SHOW_STATUS_INVALID
;
106 static bool arg_switched_root
= false;
107 static PagerFlags arg_pager_flags
= 0;
108 static bool arg_service_watchdogs
= true;
109 static ExecOutput arg_default_std_output
= EXEC_OUTPUT_JOURNAL
;
110 static ExecOutput arg_default_std_error
= EXEC_OUTPUT_INHERIT
;
111 static usec_t arg_default_restart_usec
= DEFAULT_RESTART_USEC
;
112 static usec_t arg_default_timeout_start_usec
= DEFAULT_TIMEOUT_USEC
;
113 static usec_t arg_default_timeout_stop_usec
= DEFAULT_TIMEOUT_USEC
;
114 static usec_t arg_default_start_limit_interval
= DEFAULT_START_LIMIT_INTERVAL
;
115 static unsigned arg_default_start_limit_burst
= DEFAULT_START_LIMIT_BURST
;
116 static usec_t arg_runtime_watchdog
= 0;
117 static usec_t arg_shutdown_watchdog
= 10 * USEC_PER_MINUTE
;
118 static char *arg_early_core_pattern
= NULL
;
119 static char *arg_watchdog_device
= NULL
;
120 static char **arg_default_environment
= NULL
;
121 static struct rlimit
*arg_default_rlimit
[_RLIMIT_MAX
] = {};
122 static uint64_t arg_capability_bounding_set
= CAP_ALL
;
123 static bool arg_no_new_privs
= false;
124 static nsec_t arg_timer_slack_nsec
= NSEC_INFINITY
;
125 static usec_t arg_default_timer_accuracy_usec
= 1 * USEC_PER_MINUTE
;
126 static Set
* arg_syscall_archs
= NULL
;
127 static FILE* arg_serialization
= NULL
;
128 static int arg_default_cpu_accounting
= -1;
129 static bool arg_default_io_accounting
= false;
130 static bool arg_default_ip_accounting
= false;
131 static bool arg_default_blockio_accounting
= false;
132 static bool arg_default_memory_accounting
= MEMORY_ACCOUNTING_DEFAULT
;
133 static bool arg_default_tasks_accounting
= true;
134 static uint64_t arg_default_tasks_max
= UINT64_MAX
;
135 static sd_id128_t arg_machine_id
= {};
136 static EmergencyAction arg_cad_burst_action
= EMERGENCY_ACTION_REBOOT_FORCE
;
138 _noreturn_
static void freeze_or_exit_or_reboot(void) {
140 /* If we are running in a contianer, let's prefer exiting, after all we can propagate an exit code to the
141 * container manager, and thus inform it that something went wrong. */
142 if (detect_container() > 0) {
143 log_emergency("Exiting PID 1...");
144 exit(EXIT_EXCEPTION
);
147 if (arg_crash_reboot
) {
148 log_notice("Rebooting in 10s...");
151 log_notice("Rebooting now...");
152 (void) reboot(RB_AUTOBOOT
);
153 log_emergency_errno(errno
, "Failed to reboot: %m");
156 log_emergency("Freezing execution.");
160 _noreturn_
static void crash(int sig
) {
164 if (getpid_cached() != 1)
165 /* Pass this on immediately, if this is not PID 1 */
167 else if (!arg_dump_core
)
168 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig
));
170 sa
= (struct sigaction
) {
171 .sa_handler
= nop_signal_handler
,
172 .sa_flags
= SA_NOCLDSTOP
|SA_RESTART
,
175 /* We want to wait for the core process, hence let's enable SIGCHLD */
176 (void) sigaction(SIGCHLD
, &sa
, NULL
);
178 pid
= raw_clone(SIGCHLD
);
180 log_emergency_errno(errno
, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig
));
182 /* Enable default signal handler for core dump */
184 sa
= (struct sigaction
) {
185 .sa_handler
= SIG_DFL
,
187 (void) sigaction(sig
, &sa
, NULL
);
189 /* Don't limit the coredump size */
190 (void) setrlimit(RLIMIT_CORE
, &RLIMIT_MAKE_CONST(RLIM_INFINITY
));
192 /* Just to be sure... */
195 /* Raise the signal again */
197 (void) kill(pid
, sig
); /* raise() would kill the parent */
199 assert_not_reached("We shouldn't be here...");
200 _exit(EXIT_EXCEPTION
);
205 /* Order things nicely. */
206 r
= wait_for_terminate(pid
, &status
);
208 log_emergency_errno(r
, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig
));
209 else if (status
.si_code
!= CLD_DUMPED
)
210 log_emergency("Caught <%s>, core dump failed (child "PID_FMT
", code=%s, status=%i/%s).",
211 signal_to_string(sig
),
212 pid
, sigchld_code_to_string(status
.si_code
),
214 strna(status
.si_code
== CLD_EXITED
215 ? exit_status_to_string(status
.si_status
, EXIT_STATUS_MINIMAL
)
216 : signal_to_string(status
.si_status
)));
218 log_emergency("Caught <%s>, dumped core as pid "PID_FMT
".", signal_to_string(sig
), pid
);
222 if (arg_crash_chvt
>= 0)
223 (void) chvt(arg_crash_chvt
);
225 sa
= (struct sigaction
) {
226 .sa_handler
= SIG_IGN
,
227 .sa_flags
= SA_NOCLDSTOP
|SA_NOCLDWAIT
|SA_RESTART
,
230 /* Let the kernel reap children for us */
231 (void) sigaction(SIGCHLD
, &sa
, NULL
);
233 if (arg_crash_shell
) {
234 log_notice("Executing crash shell in 10s...");
237 pid
= raw_clone(SIGCHLD
);
239 log_emergency_errno(errno
, "Failed to fork off crash shell: %m");
242 (void) make_console_stdio();
243 (void) rlimit_nofile_safe();
244 (void) execle("/bin/sh", "/bin/sh", NULL
, environ
);
246 log_emergency_errno(errno
, "execle() failed: %m");
247 _exit(EXIT_EXCEPTION
);
249 log_info("Spawned crash shell as PID "PID_FMT
".", pid
);
250 (void) wait_for_terminate(pid
, NULL
);
254 freeze_or_exit_or_reboot();
257 static void install_crash_handler(void) {
258 static const struct sigaction sa
= {
260 .sa_flags
= SA_NODEFER
, /* So that we can raise the signal again from the signal handler */
264 /* We ignore the return value here, since, we don't mind if we
265 * cannot set up a crash handler */
266 r
= sigaction_many(&sa
, SIGNALS_CRASH_HANDLER
, -1);
268 log_debug_errno(r
, "I had trouble setting up the crash handler, ignoring: %m");
271 static int console_setup(void) {
272 _cleanup_close_
int tty_fd
= -1;
275 tty_fd
= open_terminal("/dev/console", O_WRONLY
|O_NOCTTY
|O_CLOEXEC
);
277 return log_error_errno(tty_fd
, "Failed to open /dev/console: %m");
279 /* We don't want to force text mode. plymouth may be showing
280 * pictures already from initrd. */
281 r
= reset_terminal_fd(tty_fd
, false);
283 return log_error_errno(r
, "Failed to reset /dev/console: %m");
288 static int parse_crash_chvt(const char *value
) {
291 if (safe_atoi(value
, &arg_crash_chvt
) >= 0)
294 b
= parse_boolean(value
);
299 arg_crash_chvt
= 0; /* switch to where kmsg goes */
301 arg_crash_chvt
= -1; /* turn off switching */
306 static int parse_confirm_spawn(const char *value
, char **console
) {
310 r
= value
? parse_boolean(value
) : 1;
316 if (r
> 0) /* on with default tty */
317 s
= strdup("/dev/console");
318 else if (is_path(value
)) /* on with fully qualified path */
320 else /* on with only a tty file name, not a fully qualified path */
321 s
= strjoin("/dev/", value
);
329 static int set_machine_id(const char *m
) {
333 if (sd_id128_from_string(m
, &t
) < 0)
336 if (sd_id128_is_null(t
))
343 static int parse_proc_cmdline_item(const char *key
, const char *value
, void *data
) {
349 if (STR_IN_SET(key
, "systemd.unit", "rd.systemd.unit")) {
351 if (proc_cmdline_value_missing(key
, value
))
354 if (!unit_name_is_valid(value
, UNIT_NAME_PLAIN
|UNIT_NAME_INSTANCE
))
355 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key
, value
);
356 else if (in_initrd() == !!startswith(key
, "rd.")) {
357 if (free_and_strdup(&arg_default_unit
, value
) < 0)
361 } else if (proc_cmdline_key_streq(key
, "systemd.dump_core")) {
363 r
= value
? parse_boolean(value
) : true;
365 log_warning_errno(r
, "Failed to parse dump core switch %s, ignoring: %m", value
);
369 } else if (proc_cmdline_key_streq(key
, "systemd.early_core_pattern")) {
371 if (proc_cmdline_value_missing(key
, value
))
374 if (path_is_absolute(value
))
375 (void) parse_path_argument_and_warn(value
, false, &arg_early_core_pattern
);
377 log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value
);
379 } else if (proc_cmdline_key_streq(key
, "systemd.crash_chvt")) {
382 arg_crash_chvt
= 0; /* turn on */
384 r
= parse_crash_chvt(value
);
386 log_warning_errno(r
, "Failed to parse crash chvt switch %s, ignoring: %m", value
);
389 } else if (proc_cmdline_key_streq(key
, "systemd.crash_shell")) {
391 r
= value
? parse_boolean(value
) : true;
393 log_warning_errno(r
, "Failed to parse crash shell switch %s, ignoring: %m", value
);
397 } else if (proc_cmdline_key_streq(key
, "systemd.crash_reboot")) {
399 r
= value
? parse_boolean(value
) : true;
401 log_warning_errno(r
, "Failed to parse crash reboot switch %s, ignoring: %m", value
);
403 arg_crash_reboot
= r
;
405 } else if (proc_cmdline_key_streq(key
, "systemd.confirm_spawn")) {
408 r
= parse_confirm_spawn(value
, &s
);
410 log_warning_errno(r
, "Failed to parse confirm_spawn switch %s, ignoring: %m", value
);
412 free_and_replace(arg_confirm_spawn
, s
);
414 } else if (proc_cmdline_key_streq(key
, "systemd.service_watchdogs")) {
416 r
= value
? parse_boolean(value
) : true;
418 log_warning_errno(r
, "Failed to parse service watchdog switch %s, ignoring: %m", value
);
420 arg_service_watchdogs
= r
;
422 } else if (proc_cmdline_key_streq(key
, "systemd.show_status")) {
425 r
= parse_show_status(value
, &arg_show_status
);
427 log_warning_errno(r
, "Failed to parse show status switch %s, ignoring: %m", value
);
429 arg_show_status
= SHOW_STATUS_YES
;
431 } else if (proc_cmdline_key_streq(key
, "systemd.default_standard_output")) {
433 if (proc_cmdline_value_missing(key
, value
))
436 r
= exec_output_from_string(value
);
438 log_warning_errno(r
, "Failed to parse default standard output switch %s, ignoring: %m", value
);
440 arg_default_std_output
= r
;
442 } else if (proc_cmdline_key_streq(key
, "systemd.default_standard_error")) {
444 if (proc_cmdline_value_missing(key
, value
))
447 r
= exec_output_from_string(value
);
449 log_warning_errno(r
, "Failed to parse default standard error switch %s, ignoring: %m", value
);
451 arg_default_std_error
= r
;
453 } else if (streq(key
, "systemd.setenv")) {
455 if (proc_cmdline_value_missing(key
, value
))
458 if (env_assignment_is_valid(value
)) {
461 env
= strv_env_set(arg_default_environment
, value
);
465 arg_default_environment
= env
;
467 log_warning("Environment variable name '%s' is not valid. Ignoring.", value
);
469 } else if (proc_cmdline_key_streq(key
, "systemd.machine_id")) {
471 if (proc_cmdline_value_missing(key
, value
))
474 r
= set_machine_id(value
);
476 log_warning_errno(r
, "MachineID '%s' is not valid, ignoring: %m", value
);
478 } else if (proc_cmdline_key_streq(key
, "systemd.default_timeout_start_sec")) {
480 if (proc_cmdline_value_missing(key
, value
))
483 r
= parse_sec(value
, &arg_default_timeout_start_usec
);
485 log_warning_errno(r
, "Failed to parse default start timeout '%s', ignoring: %m", value
);
487 if (arg_default_timeout_start_usec
<= 0)
488 arg_default_timeout_start_usec
= USEC_INFINITY
;
490 } else if (proc_cmdline_key_streq(key
, "systemd.watchdog_device")) {
492 if (proc_cmdline_value_missing(key
, value
))
495 (void) parse_path_argument_and_warn(value
, false, &arg_watchdog_device
);
497 } else if (streq(key
, "quiet") && !value
) {
499 if (arg_show_status
== _SHOW_STATUS_INVALID
)
500 arg_show_status
= SHOW_STATUS_AUTO
;
502 } else if (streq(key
, "debug") && !value
) {
504 /* Note that log_parse_environment() handles 'debug'
505 * too, and sets the log level to LOG_DEBUG. */
507 if (detect_container() > 0)
508 log_set_target(LOG_TARGET_CONSOLE
);
513 /* SysV compatibility */
514 target
= runlevel_to_target(key
);
516 return free_and_strdup(&arg_default_unit
, target
);
522 #define DEFINE_SETTER(name, func, descr) \
523 static int name(const char *unit, \
524 const char *filename, \
526 const char *section, \
527 unsigned section_line, \
528 const char *lvalue, \
530 const char *rvalue, \
542 log_syntax(unit, LOG_ERR, filename, line, r, \
543 "Invalid " descr "'%s': %m", \
549 DEFINE_SETTER(config_parse_level2
, log_set_max_level_from_string
, "log level");
550 DEFINE_SETTER(config_parse_target
, log_set_target_from_string
, "target");
551 DEFINE_SETTER(config_parse_color
, log_show_color_from_string
, "color" );
552 DEFINE_SETTER(config_parse_location
, log_show_location_from_string
, "location");
554 static int config_parse_cpu_affinity2(
556 const char *filename
,
559 unsigned section_line
,
566 _cleanup_cpu_free_ cpu_set_t
*c
= NULL
;
569 ncpus
= parse_cpu_set_and_warn(rvalue
, &c
, unit
, filename
, line
, lvalue
);
573 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus
), c
) < 0)
574 log_warning_errno(errno
, "Failed to set CPU affinity: %m");
579 static int config_parse_show_status(
581 const char *filename
,
584 unsigned section_line
,
592 ShowStatus
*b
= data
;
599 k
= parse_show_status(rvalue
, b
);
601 log_syntax(unit
, LOG_ERR
, filename
, line
, k
, "Failed to parse show status setting, ignoring: %s", rvalue
);
608 static int config_parse_output_restricted(
610 const char *filename
,
613 unsigned section_line
,
620 ExecOutput t
, *eo
= data
;
627 t
= exec_output_from_string(rvalue
);
629 log_syntax(unit
, LOG_ERR
, filename
, line
, 0, "Failed to parse output type, ignoring: %s", rvalue
);
633 if (IN_SET(t
, EXEC_OUTPUT_SOCKET
, EXEC_OUTPUT_NAMED_FD
, EXEC_OUTPUT_FILE
, EXEC_OUTPUT_FILE_APPEND
)) {
634 log_syntax(unit
, LOG_ERR
, filename
, line
, 0, "Standard output types socket, fd:, file:, append: are not supported as defaults, ignoring: %s", rvalue
);
642 static int config_parse_crash_chvt(
644 const char *filename
,
647 unsigned section_line
,
660 r
= parse_crash_chvt(rvalue
);
662 log_syntax(unit
, LOG_ERR
, filename
, line
, r
, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue
);
669 static int parse_config_file(void) {
671 const ConfigTableItem items
[] = {
672 { "Manager", "LogLevel", config_parse_level2
, 0, NULL
},
673 { "Manager", "LogTarget", config_parse_target
, 0, NULL
},
674 { "Manager", "LogColor", config_parse_color
, 0, NULL
},
675 { "Manager", "LogLocation", config_parse_location
, 0, NULL
},
676 { "Manager", "DumpCore", config_parse_bool
, 0, &arg_dump_core
},
677 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt
, 0, NULL
},
678 { "Manager", "CrashChangeVT", config_parse_crash_chvt
, 0, NULL
},
679 { "Manager", "CrashShell", config_parse_bool
, 0, &arg_crash_shell
},
680 { "Manager", "CrashReboot", config_parse_bool
, 0, &arg_crash_reboot
},
681 { "Manager", "ShowStatus", config_parse_show_status
, 0, &arg_show_status
},
682 { "Manager", "CPUAffinity", config_parse_cpu_affinity2
, 0, NULL
},
683 { "Manager", "JoinControllers", config_parse_warn_compat
, DISABLED_CONFIGURATION
, NULL
},
684 { "Manager", "RuntimeWatchdogSec", config_parse_sec
, 0, &arg_runtime_watchdog
},
685 { "Manager", "ShutdownWatchdogSec", config_parse_sec
, 0, &arg_shutdown_watchdog
},
686 { "Manager", "WatchdogDevice", config_parse_path
, 0, &arg_watchdog_device
},
687 { "Manager", "CapabilityBoundingSet", config_parse_capability_set
, 0, &arg_capability_bounding_set
},
688 { "Manager", "NoNewPrivileges", config_parse_bool
, 0, &arg_no_new_privs
},
690 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs
, 0, &arg_syscall_archs
},
692 { "Manager", "TimerSlackNSec", config_parse_nsec
, 0, &arg_timer_slack_nsec
},
693 { "Manager", "DefaultTimerAccuracySec", config_parse_sec
, 0, &arg_default_timer_accuracy_usec
},
694 { "Manager", "DefaultStandardOutput", config_parse_output_restricted
,0, &arg_default_std_output
},
695 { "Manager", "DefaultStandardError", config_parse_output_restricted
,0, &arg_default_std_error
},
696 { "Manager", "DefaultTimeoutStartSec", config_parse_sec
, 0, &arg_default_timeout_start_usec
},
697 { "Manager", "DefaultTimeoutStopSec", config_parse_sec
, 0, &arg_default_timeout_stop_usec
},
698 { "Manager", "DefaultRestartSec", config_parse_sec
, 0, &arg_default_restart_usec
},
699 { "Manager", "DefaultStartLimitInterval", config_parse_sec
, 0, &arg_default_start_limit_interval
}, /* obsolete alias */
700 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec
, 0, &arg_default_start_limit_interval
},
701 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned
, 0, &arg_default_start_limit_burst
},
702 { "Manager", "DefaultEnvironment", config_parse_environ
, 0, &arg_default_environment
},
703 { "Manager", "DefaultLimitCPU", config_parse_rlimit
, RLIMIT_CPU
, arg_default_rlimit
},
704 { "Manager", "DefaultLimitFSIZE", config_parse_rlimit
, RLIMIT_FSIZE
, arg_default_rlimit
},
705 { "Manager", "DefaultLimitDATA", config_parse_rlimit
, RLIMIT_DATA
, arg_default_rlimit
},
706 { "Manager", "DefaultLimitSTACK", config_parse_rlimit
, RLIMIT_STACK
, arg_default_rlimit
},
707 { "Manager", "DefaultLimitCORE", config_parse_rlimit
, RLIMIT_CORE
, arg_default_rlimit
},
708 { "Manager", "DefaultLimitRSS", config_parse_rlimit
, RLIMIT_RSS
, arg_default_rlimit
},
709 { "Manager", "DefaultLimitNOFILE", config_parse_rlimit
, RLIMIT_NOFILE
, arg_default_rlimit
},
710 { "Manager", "DefaultLimitAS", config_parse_rlimit
, RLIMIT_AS
, arg_default_rlimit
},
711 { "Manager", "DefaultLimitNPROC", config_parse_rlimit
, RLIMIT_NPROC
, arg_default_rlimit
},
712 { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit
, RLIMIT_MEMLOCK
, arg_default_rlimit
},
713 { "Manager", "DefaultLimitLOCKS", config_parse_rlimit
, RLIMIT_LOCKS
, arg_default_rlimit
},
714 { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit
, RLIMIT_SIGPENDING
, arg_default_rlimit
},
715 { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit
, RLIMIT_MSGQUEUE
, arg_default_rlimit
},
716 { "Manager", "DefaultLimitNICE", config_parse_rlimit
, RLIMIT_NICE
, arg_default_rlimit
},
717 { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit
, RLIMIT_RTPRIO
, arg_default_rlimit
},
718 { "Manager", "DefaultLimitRTTIME", config_parse_rlimit
, RLIMIT_RTTIME
, arg_default_rlimit
},
719 { "Manager", "DefaultCPUAccounting", config_parse_tristate
, 0, &arg_default_cpu_accounting
},
720 { "Manager", "DefaultIOAccounting", config_parse_bool
, 0, &arg_default_io_accounting
},
721 { "Manager", "DefaultIPAccounting", config_parse_bool
, 0, &arg_default_ip_accounting
},
722 { "Manager", "DefaultBlockIOAccounting", config_parse_bool
, 0, &arg_default_blockio_accounting
},
723 { "Manager", "DefaultMemoryAccounting", config_parse_bool
, 0, &arg_default_memory_accounting
},
724 { "Manager", "DefaultTasksAccounting", config_parse_bool
, 0, &arg_default_tasks_accounting
},
725 { "Manager", "DefaultTasksMax", config_parse_tasks_max
, 0, &arg_default_tasks_max
},
726 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action
, 0, &arg_cad_burst_action
},
730 const char *fn
, *conf_dirs_nulstr
;
733 PKGSYSCONFDIR
"/system.conf" :
734 PKGSYSCONFDIR
"/user.conf";
736 conf_dirs_nulstr
= arg_system
?
737 CONF_PATHS_NULSTR("systemd/system.conf.d") :
738 CONF_PATHS_NULSTR("systemd/user.conf.d");
740 (void) config_parse_many_nulstr(fn
, conf_dirs_nulstr
, "Manager\0", config_item_table_lookup
, items
, CONFIG_PARSE_WARN
, NULL
);
742 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
743 * like everywhere else. */
744 if (arg_default_timeout_start_usec
<= 0)
745 arg_default_timeout_start_usec
= USEC_INFINITY
;
746 if (arg_default_timeout_stop_usec
<= 0)
747 arg_default_timeout_stop_usec
= USEC_INFINITY
;
752 static void set_manager_defaults(Manager
*m
) {
756 /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
757 * affect the manager itself, but are just what newly allocated units will have set if they haven't set
758 * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
760 m
->default_timer_accuracy_usec
= arg_default_timer_accuracy_usec
;
761 m
->default_std_output
= arg_default_std_output
;
762 m
->default_std_error
= arg_default_std_error
;
763 m
->default_timeout_start_usec
= arg_default_timeout_start_usec
;
764 m
->default_timeout_stop_usec
= arg_default_timeout_stop_usec
;
765 m
->default_restart_usec
= arg_default_restart_usec
;
766 m
->default_start_limit_interval
= arg_default_start_limit_interval
;
767 m
->default_start_limit_burst
= arg_default_start_limit_burst
;
769 /* On 4.15+ with unified hierarchy, CPU accounting is essentially free as it doesn't require the CPU
770 * controller to be enabled, so the default is to enable it unless we got told otherwise. */
771 if (arg_default_cpu_accounting
>= 0)
772 m
->default_cpu_accounting
= arg_default_cpu_accounting
;
774 m
->default_cpu_accounting
= cpu_accounting_is_cheap();
776 m
->default_io_accounting
= arg_default_io_accounting
;
777 m
->default_ip_accounting
= arg_default_ip_accounting
;
778 m
->default_blockio_accounting
= arg_default_blockio_accounting
;
779 m
->default_memory_accounting
= arg_default_memory_accounting
;
780 m
->default_tasks_accounting
= arg_default_tasks_accounting
;
781 m
->default_tasks_max
= arg_default_tasks_max
;
783 (void) manager_set_default_rlimits(m
, arg_default_rlimit
);
785 (void) manager_default_environment(m
);
786 (void) manager_transient_environment_add(m
, arg_default_environment
);
789 static void set_manager_settings(Manager
*m
) {
793 /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
794 * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
796 m
->confirm_spawn
= arg_confirm_spawn
;
797 m
->service_watchdogs
= arg_service_watchdogs
;
798 m
->runtime_watchdog
= arg_runtime_watchdog
;
799 m
->shutdown_watchdog
= arg_shutdown_watchdog
;
800 m
->cad_burst_action
= arg_cad_burst_action
;
802 manager_set_show_status(m
, arg_show_status
);
805 static int parse_argv(int argc
, char *argv
[]) {
807 ARG_LOG_LEVEL
= 0x100,
817 ARG_DUMP_CONFIGURATION_ITEMS
,
818 ARG_DUMP_BUS_PROPERTIES
,
827 ARG_DEFAULT_STD_OUTPUT
,
828 ARG_DEFAULT_STD_ERROR
,
830 ARG_SERVICE_WATCHDOGS
,
833 static const struct option options
[] = {
834 { "log-level", required_argument
, NULL
, ARG_LOG_LEVEL
},
835 { "log-target", required_argument
, NULL
, ARG_LOG_TARGET
},
836 { "log-color", optional_argument
, NULL
, ARG_LOG_COLOR
},
837 { "log-location", optional_argument
, NULL
, ARG_LOG_LOCATION
},
838 { "unit", required_argument
, NULL
, ARG_UNIT
},
839 { "system", no_argument
, NULL
, ARG_SYSTEM
},
840 { "user", no_argument
, NULL
, ARG_USER
},
841 { "test", no_argument
, NULL
, ARG_TEST
},
842 { "no-pager", no_argument
, NULL
, ARG_NO_PAGER
},
843 { "help", no_argument
, NULL
, 'h' },
844 { "version", no_argument
, NULL
, ARG_VERSION
},
845 { "dump-configuration-items", no_argument
, NULL
, ARG_DUMP_CONFIGURATION_ITEMS
},
846 { "dump-bus-properties", no_argument
, NULL
, ARG_DUMP_BUS_PROPERTIES
},
847 { "dump-core", optional_argument
, NULL
, ARG_DUMP_CORE
},
848 { "crash-chvt", required_argument
, NULL
, ARG_CRASH_CHVT
},
849 { "crash-shell", optional_argument
, NULL
, ARG_CRASH_SHELL
},
850 { "crash-reboot", optional_argument
, NULL
, ARG_CRASH_REBOOT
},
851 { "confirm-spawn", optional_argument
, NULL
, ARG_CONFIRM_SPAWN
},
852 { "show-status", optional_argument
, NULL
, ARG_SHOW_STATUS
},
853 { "deserialize", required_argument
, NULL
, ARG_DESERIALIZE
},
854 { "switched-root", no_argument
, NULL
, ARG_SWITCHED_ROOT
},
855 { "default-standard-output", required_argument
, NULL
, ARG_DEFAULT_STD_OUTPUT
, },
856 { "default-standard-error", required_argument
, NULL
, ARG_DEFAULT_STD_ERROR
, },
857 { "machine-id", required_argument
, NULL
, ARG_MACHINE_ID
},
858 { "service-watchdogs", required_argument
, NULL
, ARG_SERVICE_WATCHDOGS
},
867 if (getpid_cached() == 1)
870 while ((c
= getopt_long(argc
, argv
, "hDbsz:", options
, NULL
)) >= 0)
875 r
= log_set_max_level_from_string(optarg
);
877 return log_error_errno(r
, "Failed to parse log level \"%s\": %m", optarg
);
882 r
= log_set_target_from_string(optarg
);
884 return log_error_errno(r
, "Failed to parse log target \"%s\": %m", optarg
);
891 r
= log_show_color_from_string(optarg
);
893 return log_error_errno(r
, "Failed to parse log color setting \"%s\": %m",
896 log_show_color(true);
900 case ARG_LOG_LOCATION
:
902 r
= log_show_location_from_string(optarg
);
904 return log_error_errno(r
, "Failed to parse log location setting \"%s\": %m",
907 log_show_location(true);
911 case ARG_DEFAULT_STD_OUTPUT
:
912 r
= exec_output_from_string(optarg
);
914 return log_error_errno(r
, "Failed to parse default standard output setting \"%s\": %m",
916 arg_default_std_output
= r
;
919 case ARG_DEFAULT_STD_ERROR
:
920 r
= exec_output_from_string(optarg
);
922 return log_error_errno(r
, "Failed to parse default standard error output setting \"%s\": %m",
924 arg_default_std_error
= r
;
928 r
= free_and_strdup(&arg_default_unit
, optarg
);
930 return log_error_errno(r
, "Failed to set default unit \"%s\": %m", optarg
);
943 arg_action
= ACTION_TEST
;
947 arg_pager_flags
|= PAGER_DISABLE
;
951 arg_action
= ACTION_VERSION
;
954 case ARG_DUMP_CONFIGURATION_ITEMS
:
955 arg_action
= ACTION_DUMP_CONFIGURATION_ITEMS
;
958 case ARG_DUMP_BUS_PROPERTIES
:
959 arg_action
= ACTION_DUMP_BUS_PROPERTIES
;
964 arg_dump_core
= true;
966 r
= parse_boolean(optarg
);
968 return log_error_errno(r
, "Failed to parse dump core boolean: \"%s\": %m",
975 r
= parse_crash_chvt(optarg
);
977 return log_error_errno(r
, "Failed to parse crash virtual terminal index: \"%s\": %m",
981 case ARG_CRASH_SHELL
:
983 arg_crash_shell
= true;
985 r
= parse_boolean(optarg
);
987 return log_error_errno(r
, "Failed to parse crash shell boolean: \"%s\": %m",
993 case ARG_CRASH_REBOOT
:
995 arg_crash_reboot
= true;
997 r
= parse_boolean(optarg
);
999 return log_error_errno(r
, "Failed to parse crash shell boolean: \"%s\": %m",
1001 arg_crash_reboot
= r
;
1005 case ARG_CONFIRM_SPAWN
:
1006 arg_confirm_spawn
= mfree(arg_confirm_spawn
);
1008 r
= parse_confirm_spawn(optarg
, &arg_confirm_spawn
);
1010 return log_error_errno(r
, "Failed to parse confirm spawn option: \"%s\": %m",
1014 case ARG_SERVICE_WATCHDOGS
:
1015 r
= parse_boolean(optarg
);
1017 return log_error_errno(r
, "Failed to parse service watchdogs boolean: \"%s\": %m",
1019 arg_service_watchdogs
= r
;
1022 case ARG_SHOW_STATUS
:
1024 r
= parse_show_status(optarg
, &arg_show_status
);
1026 return log_error_errno(r
, "Failed to parse show status boolean: \"%s\": %m",
1029 arg_show_status
= SHOW_STATUS_YES
;
1032 case ARG_DESERIALIZE
: {
1036 r
= safe_atoi(optarg
, &fd
);
1038 log_error_errno(r
, "Failed to parse deserialize option \"%s\": %m", optarg
);
1040 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
),
1041 "Invalid deserialize fd: %d",
1044 (void) fd_cloexec(fd
, true);
1046 f
= fdopen(fd
, "r");
1048 return log_error_errno(errno
, "Failed to open serialization fd %d: %m", fd
);
1050 safe_fclose(arg_serialization
);
1051 arg_serialization
= f
;
1056 case ARG_SWITCHED_ROOT
:
1057 arg_switched_root
= true;
1060 case ARG_MACHINE_ID
:
1061 r
= set_machine_id(optarg
);
1063 return log_error_errno(r
, "MachineID '%s' is not valid: %m", optarg
);
1067 arg_action
= ACTION_HELP
;
1071 log_set_max_level(LOG_DEBUG
);
1077 /* Just to eat away the sysvinit kernel
1078 * cmdline args without getopt() error
1079 * messages that we'll parse in
1080 * parse_proc_cmdline_word() or ignore. */
1083 if (getpid_cached() != 1)
1089 assert_not_reached("Unhandled option code.");
1092 if (optind
< argc
&& getpid_cached() != 1) {
1093 /* Hmm, when we aren't run as init system
1094 * let's complain about excess arguments */
1096 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
),
1097 "Excess arguments.");
1103 static int help(void) {
1104 _cleanup_free_
char *link
= NULL
;
1107 r
= terminal_urlify_man("systemd", "1", &link
);
1111 printf("%s [OPTIONS...]\n\n"
1112 "Starts up and maintains the system or user services.\n\n"
1113 " -h --help Show this help\n"
1114 " --version Show version\n"
1115 " --test Determine startup sequence, dump it and exit\n"
1116 " --no-pager Do not pipe output into a pager\n"
1117 " --dump-configuration-items Dump understood unit configuration items\n"
1118 " --dump-bus-properties Dump exposed bus properties\n"
1119 " --unit=UNIT Set default unit\n"
1120 " --system Run a system instance, even if PID != 1\n"
1121 " --user Run a user instance\n"
1122 " --dump-core[=BOOL] Dump core on crash\n"
1123 " --crash-vt=NR Change to specified VT on crash\n"
1124 " --crash-reboot[=BOOL] Reboot on crash\n"
1125 " --crash-shell[=BOOL] Run shell on crash\n"
1126 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1127 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1128 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1129 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1130 " --log-color[=BOOL] Highlight important log messages\n"
1131 " --log-location[=BOOL] Include code location in log messages\n"
1132 " --default-standard-output= Set default standard output for services\n"
1133 " --default-standard-error= Set default standard error output for services\n"
1134 "\nSee the %s for details.\n"
1135 , program_invocation_short_name
1142 static int prepare_reexecute(
1146 bool switching_root
) {
1148 _cleanup_fdset_free_ FDSet
*fds
= NULL
;
1149 _cleanup_fclose_
FILE *f
= NULL
;
1156 r
= manager_open_serialization(m
, &f
);
1158 return log_error_errno(r
, "Failed to create serialization file: %m");
1160 /* Make sure nothing is really destructed when we shut down */
1162 bus_manager_send_reloading(m
, true);
1168 r
= manager_serialize(m
, f
, fds
, switching_root
);
1172 if (fseeko(f
, 0, SEEK_SET
) == (off_t
) -1)
1173 return log_error_errno(errno
, "Failed to rewind serialization fd: %m");
1175 r
= fd_cloexec(fileno(f
), false);
1177 return log_error_errno(r
, "Failed to disable O_CLOEXEC for serialization: %m");
1179 r
= fdset_cloexec(fds
, false);
1181 return log_error_errno(r
, "Failed to disable O_CLOEXEC for serialization fds: %m");
1183 *ret_f
= TAKE_PTR(f
);
1184 *ret_fds
= TAKE_PTR(fds
);
1189 static void bump_file_max_and_nr_open(void) {
1191 /* Let's bump fs.file-max and fs.nr_open to their respective maximums. On current kernels large numbers of file
1192 * descriptors are no longer a performance problem and their memory is properly tracked by memcg, thus counting
1193 * them and limiting them in another two layers of limits is unnecessary and just complicates things. This
1194 * function hence turns off 2 of the 4 levels of limits on file descriptors, and makes RLIMIT_NOLIMIT (soft +
1195 * hard) the only ones that really matter. */
1197 #if BUMP_PROC_SYS_FS_FILE_MAX || BUMP_PROC_SYS_FS_NR_OPEN
1198 _cleanup_free_
char *t
= NULL
;
1202 #if BUMP_PROC_SYS_FS_FILE_MAX
1203 /* I so wanted to use STRINGIFY(ULONG_MAX) here, but alas we can't as glibc/gcc define that as
1204 * "(0x7fffffffffffffffL * 2UL + 1UL)". Seriously. 😢 */
1205 if (asprintf(&t
, "%lu\n", ULONG_MAX
) < 0) {
1210 r
= sysctl_write("fs/file-max", t
);
1212 log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
, "Failed to bump fs.file-max, ignoring: %m");
1215 #if BUMP_PROC_SYS_FS_FILE_MAX && BUMP_PROC_SYS_FS_NR_OPEN
1219 #if BUMP_PROC_SYS_FS_NR_OPEN
1222 /* Arg! The kernel enforces maximum and minimum values on the fs.nr_open, but we don't really know what they
1223 * are. The expression by which the maximum is determined is dependent on the architecture, and is something we
1224 * don't really want to copy to userspace, as it is dependent on implementation details of the kernel. Since
1225 * the kernel doesn't expose the maximum value to us, we can only try and hope. Hence, let's start with
1226 * INT_MAX, and then keep halving the value until we find one that works. Ugly? Yes, absolutely, but kernel
1227 * APIs are kernel APIs, so what do can we do... 🤯 */
1232 v
&= ~(__SIZEOF_POINTER__
- 1); /* Round down to next multiple of the pointer size */
1234 log_warning("Can't bump fs.nr_open, value too small.");
1240 log_error_errno(k
, "Failed to read fs.nr_open: %m");
1243 if (k
>= v
) { /* Already larger */
1244 log_debug("Skipping bump, value is already larger.");
1248 if (asprintf(&t
, "%i\n", v
) < 0) {
1253 r
= sysctl_write("fs/nr_open", t
);
1256 log_debug("Couldn't write fs.nr_open as %i, halving it.", v
);
1261 log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
, "Failed to bump fs.nr_open, ignoring: %m");
1265 log_debug("Successfully bumped fs.nr_open to %i", v
);
1271 static int bump_rlimit_nofile(struct rlimit
*saved_rlimit
) {
1272 struct rlimit new_rlimit
;
1275 assert(saved_rlimit
);
1277 /* Save the original RLIMIT_NOFILE so that we can reset it later when transitioning from the initrd to the main
1278 * systemd or suchlike. */
1279 if (getrlimit(RLIMIT_NOFILE
, saved_rlimit
) < 0)
1280 return log_warning_errno(errno
, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1282 /* Get the underlying absolute limit the kernel enforces */
1283 nr
= read_nr_open();
1285 /* Make sure forked processes get limits based on the original kernel setting */
1286 if (!arg_default_rlimit
[RLIMIT_NOFILE
]) {
1289 rl
= newdup(struct rlimit
, saved_rlimit
, 1);
1293 /* Bump the hard limit for system services to a substantially higher value. The default hard limit
1294 * current kernels set is pretty low (4K), mostly for historical reasons. According to kernel
1295 * developers, the fd handling in recent kernels has been optimized substantially enough, so that we
1296 * can bump the limit now, without paying too high a price in memory or performance. Note however that
1297 * we only bump the hard limit, not the soft limit. That's because select() works the way it works, and
1298 * chokes on fds >= 1024. If we'd bump the soft limit globally, it might accidentally happen to
1299 * unexpecting programs that they get fds higher than what they can process using select(). By only
1300 * bumping the hard limit but leaving the low limit as it is we avoid this pitfall: programs that are
1301 * written by folks aware of the select() problem in mind (and thus use poll()/epoll instead of
1302 * select(), the way everybody should) can explicitly opt into high fds by bumping their soft limit
1303 * beyond 1024, to the hard limit we pass. */
1305 rl
->rlim_max
= MIN((rlim_t
) nr
, MAX(rl
->rlim_max
, (rlim_t
) HIGH_RLIMIT_NOFILE
));
1307 /* If for some reason we were invoked with a soft limit above 1024 (which should never
1308 * happen!, but who knows what we get passed in from pam_limit when invoked as --user
1309 * instance), then lower what we pass on to not confuse our children */
1310 rl
->rlim_cur
= MIN(rl
->rlim_cur
, (rlim_t
) FD_SETSIZE
);
1312 arg_default_rlimit
[RLIMIT_NOFILE
] = rl
;
1315 /* Calculate the new limits to use for us. Never lower from what we inherited. */
1316 new_rlimit
= (struct rlimit
) {
1317 .rlim_cur
= MAX((rlim_t
) nr
, saved_rlimit
->rlim_cur
),
1318 .rlim_max
= MAX((rlim_t
) nr
, saved_rlimit
->rlim_max
),
1321 /* Shortcut if nothing changes. */
1322 if (saved_rlimit
->rlim_max
>= new_rlimit
.rlim_max
&&
1323 saved_rlimit
->rlim_cur
>= new_rlimit
.rlim_cur
) {
1324 log_debug("RLIMIT_NOFILE is already as high or higher than we need it, not bumping.");
1328 /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for
1329 * both hard and soft. */
1330 r
= setrlimit_closest(RLIMIT_NOFILE
, &new_rlimit
);
1332 return log_warning_errno(r
, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1337 static int bump_rlimit_memlock(struct rlimit
*saved_rlimit
) {
1338 struct rlimit new_rlimit
;
1341 assert(saved_rlimit
);
1343 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK which should
1344 * normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's bump
1345 * the value high enough for our user. */
1347 if (getrlimit(RLIMIT_MEMLOCK
, saved_rlimit
) < 0)
1348 return log_warning_errno(errno
, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1350 /* Pass the original value down to invoked processes */
1351 if (!arg_default_rlimit
[RLIMIT_MEMLOCK
]) {
1354 rl
= newdup(struct rlimit
, saved_rlimit
, 1);
1358 arg_default_rlimit
[RLIMIT_MEMLOCK
] = rl
;
1361 /* Using MAX() on resource limits only is safe if RLIM_INFINITY is > 0. POSIX declares that rlim_t
1362 * must be unsigned, hence this is a given, but let's make this clear here. */
1363 assert_cc(RLIM_INFINITY
> 0);
1365 new_rlimit
= (struct rlimit
) {
1366 .rlim_cur
= MAX(HIGH_RLIMIT_MEMLOCK
, saved_rlimit
->rlim_cur
),
1367 .rlim_max
= MAX(HIGH_RLIMIT_MEMLOCK
, saved_rlimit
->rlim_max
),
1370 if (saved_rlimit
->rlim_max
>= new_rlimit
.rlim_cur
&&
1371 saved_rlimit
->rlim_cur
>= new_rlimit
.rlim_max
) {
1372 log_debug("RLIMIT_MEMLOCK is already as high or higher than we need it, not bumping.");
1376 r
= setrlimit_closest(RLIMIT_MEMLOCK
, &new_rlimit
);
1378 return log_warning_errno(r
, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1383 static void test_usr(void) {
1385 /* Check that /usr is either on the same file system as / or mounted already. */
1387 if (dir_is_empty("/usr") <= 0)
1390 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1391 "Some things will probably break (sometimes even silently) in mysterious ways. "
1392 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1395 static int enforce_syscall_archs(Set
*archs
) {
1399 if (!is_seccomp_available())
1402 r
= seccomp_restrict_archs(arg_syscall_archs
);
1404 return log_error_errno(r
, "Failed to enforce system call architecture restrication: %m");
1409 static int status_welcome(void) {
1410 _cleanup_free_
char *pretty_name
= NULL
, *ansi_color
= NULL
;
1413 if (IN_SET(arg_show_status
, SHOW_STATUS_NO
, SHOW_STATUS_AUTO
))
1416 r
= parse_os_release(NULL
,
1417 "PRETTY_NAME", &pretty_name
,
1418 "ANSI_COLOR", &ansi_color
,
1421 log_full_errno(r
== -ENOENT
? LOG_DEBUG
: LOG_WARNING
, r
,
1422 "Failed to read os-release file, ignoring: %m");
1424 if (log_get_show_color())
1425 return status_printf(NULL
, 0,
1426 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1427 isempty(ansi_color
) ? "1" : ansi_color
,
1428 isempty(pretty_name
) ? "Linux" : pretty_name
);
1430 return status_printf(NULL
, 0,
1431 "\nWelcome to %s!\n",
1432 isempty(pretty_name
) ? "Linux" : pretty_name
);
1435 static int write_container_id(void) {
1439 c
= getenv("container");
1443 RUN_WITH_UMASK(0022)
1444 r
= write_string_file("/run/systemd/container", c
, WRITE_STRING_FILE_CREATE
);
1446 return log_warning_errno(r
, "Failed to write /run/systemd/container, ignoring: %m");
1451 static int bump_unix_max_dgram_qlen(void) {
1452 _cleanup_free_
char *qlen
= NULL
;
1456 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1457 * really really early during boot, so that it is actually applied to all our sockets, including the
1458 * $NOTIFY_SOCKET one. */
1460 r
= read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen
);
1462 return log_full_errno(r
== -ENOENT
? LOG_DEBUG
: LOG_WARNING
, r
, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1464 r
= safe_atolu(qlen
, &v
);
1466 return log_warning_errno(r
, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen
);
1468 if (v
>= DEFAULT_UNIX_MAX_DGRAM_QLEN
)
1471 r
= write_string_filef("/proc/sys/net/unix/max_dgram_qlen", WRITE_STRING_FILE_DISABLE_BUFFER
, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN
);
1473 return log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
,
1474 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1479 static int fixup_environment(void) {
1480 _cleanup_free_
char *term
= NULL
;
1484 /* Only fix up the environment when we are started as PID 1 */
1485 if (getpid_cached() != 1)
1488 /* We expect the environment to be set correctly if run inside a container. */
1489 if (detect_container() > 0)
1492 /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1493 * device used by the console. We try to make a better guess here since some consoles might not have support
1494 * for color mode for example.
1496 * However if TERM was configured through the kernel command line then leave it alone. */
1497 r
= proc_cmdline_get_key("TERM", 0, &term
);
1501 t
= term
?: default_term_for_tty("/dev/console");
1503 if (setenv("TERM", t
, 1) < 0)
1509 static void redirect_telinit(int argc
, char *argv
[]) {
1511 /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1513 #if HAVE_SYSV_COMPAT
1514 if (getpid_cached() == 1)
1517 if (!strstr(program_invocation_short_name
, "init"))
1520 execv(SYSTEMCTL_BINARY_PATH
, argv
);
1521 log_error_errno(errno
, "Failed to exec " SYSTEMCTL_BINARY_PATH
": %m");
1526 static int become_shutdown(
1527 const char *shutdown_verb
,
1530 char log_level
[DECIMAL_STR_MAX(int) + 1],
1531 exit_code
[DECIMAL_STR_MAX(uint8_t) + 1],
1532 timeout
[DECIMAL_STR_MAX(usec_t
) + 1];
1534 const char* command_line
[13] = {
1535 SYSTEMD_SHUTDOWN_BINARY_PATH
,
1537 "--timeout", timeout
,
1538 "--log-level", log_level
,
1542 _cleanup_strv_free_
char **env_block
= NULL
;
1546 assert(shutdown_verb
);
1547 assert(!command_line
[pos
]);
1548 env_block
= strv_copy(environ
);
1550 xsprintf(log_level
, "%d", log_get_max_level());
1551 xsprintf(timeout
, "%" PRI_USEC
"us", arg_default_timeout_stop_usec
);
1553 switch (log_get_target()) {
1555 case LOG_TARGET_KMSG
:
1556 case LOG_TARGET_JOURNAL_OR_KMSG
:
1557 case LOG_TARGET_SYSLOG_OR_KMSG
:
1558 command_line
[pos
++] = "kmsg";
1561 case LOG_TARGET_NULL
:
1562 command_line
[pos
++] = "null";
1565 case LOG_TARGET_CONSOLE
:
1567 command_line
[pos
++] = "console";
1571 if (log_get_show_color())
1572 command_line
[pos
++] = "--log-color";
1574 if (log_get_show_location())
1575 command_line
[pos
++] = "--log-location";
1577 if (streq(shutdown_verb
, "exit")) {
1578 command_line
[pos
++] = "--exit-code";
1579 command_line
[pos
++] = exit_code
;
1580 xsprintf(exit_code
, "%d", retval
);
1583 assert(pos
< ELEMENTSOF(command_line
));
1585 if (streq(shutdown_verb
, "reboot") &&
1586 arg_shutdown_watchdog
> 0 &&
1587 arg_shutdown_watchdog
!= USEC_INFINITY
) {
1591 /* If we reboot let's set the shutdown
1592 * watchdog and tell the shutdown binary to
1593 * repeatedly ping it */
1594 r
= watchdog_set_timeout(&arg_shutdown_watchdog
);
1595 watchdog_close(r
< 0);
1597 /* Tell the binary how often to ping, ignore failure */
1598 if (asprintf(&e
, "WATCHDOG_USEC="USEC_FMT
, arg_shutdown_watchdog
) > 0)
1599 (void) strv_consume(&env_block
, e
);
1601 if (arg_watchdog_device
&&
1602 asprintf(&e
, "WATCHDOG_DEVICE=%s", arg_watchdog_device
) > 0)
1603 (void) strv_consume(&env_block
, e
);
1605 watchdog_close(true);
1607 /* Avoid the creation of new processes forked by the
1608 * kernel; at this point, we will not listen to the
1610 if (detect_container() <= 0)
1611 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER
);
1613 execve(SYSTEMD_SHUTDOWN_BINARY_PATH
, (char **) command_line
, env_block
);
1617 static void initialize_clock(void) {
1620 if (clock_is_localtime(NULL
) > 0) {
1624 * The very first call of settimeofday() also does a time warp in the kernel.
1626 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1627 * of maintaining the RTC and do all adjustments. This matches the behavior of Windows, which leaves
1628 * the RTC alone if the registry tells that the RTC runs in UTC.
1630 r
= clock_set_timezone(&min
);
1632 log_error_errno(r
, "Failed to apply local time delta, ignoring: %m");
1634 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min
);
1636 } else if (!in_initrd()) {
1638 * Do a dummy very first call to seal the kernel's time warp magic.
1640 * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1641 * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1642 * until we reach the real system.
1644 * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1645 * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1646 * be treated as UTC that way.
1648 (void) clock_reset_timewarp();
1651 r
= clock_apply_epoch();
1653 log_error_errno(r
, "Current system time is before build time, but cannot correct: %m");
1655 log_info("System time before build time, advancing clock.");
1658 static void initialize_coredump(bool skip_setup
) {
1660 if (getpid_cached() != 1)
1663 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1664 * will process core dumps for system services by default. */
1665 if (setrlimit(RLIMIT_CORE
, &RLIMIT_MAKE_CONST(RLIM_INFINITY
)) < 0)
1666 log_warning_errno(errno
, "Failed to set RLIMIT_CORE: %m");
1668 /* But at the same time, turn off the core_pattern logic by default, so that no
1669 * coredumps are stored until the systemd-coredump tool is enabled via
1670 * sysctl. However it can be changed via the kernel command line later so core
1671 * dumps can still be generated during early startup and in initramfs. */
1673 disable_coredumps();
1677 static void initialize_core_pattern(bool skip_setup
) {
1680 if (skip_setup
|| !arg_early_core_pattern
)
1683 if (getpid_cached() != 1)
1686 r
= write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern
, WRITE_STRING_FILE_DISABLE_BUFFER
);
1688 log_warning_errno(r
, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", arg_early_core_pattern
);
1691 static void do_reexecute(
1694 const struct rlimit
*saved_rlimit_nofile
,
1695 const struct rlimit
*saved_rlimit_memlock
,
1697 const char *switch_root_dir
,
1698 const char *switch_root_init
,
1699 const char **ret_error_message
) {
1701 unsigned i
, j
, args_size
;
1705 assert(saved_rlimit_nofile
);
1706 assert(saved_rlimit_memlock
);
1707 assert(ret_error_message
);
1709 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1711 watchdog_close(true);
1713 /* Reset RLIMIT_NOFILE + RLIMIT_MEMLOCK back to the kernel defaults, so that the new systemd can pass
1714 * the kernel default to its child processes */
1715 if (saved_rlimit_nofile
->rlim_cur
!= 0)
1716 (void) setrlimit(RLIMIT_NOFILE
, saved_rlimit_nofile
);
1717 if (saved_rlimit_memlock
->rlim_cur
!= RLIM_INFINITY
)
1718 (void) setrlimit(RLIMIT_MEMLOCK
, saved_rlimit_memlock
);
1720 if (switch_root_dir
) {
1721 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1722 * SIGCHLD for them after deserializing. */
1723 broadcast_signal(SIGTERM
, false, true, arg_default_timeout_stop_usec
);
1725 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1726 r
= switch_root(switch_root_dir
, "/mnt", true, MS_MOVE
);
1728 log_error_errno(r
, "Failed to switch root, trying to continue: %m");
1731 args_size
= MAX(6, argc
+1);
1732 args
= newa(const char*, args_size
);
1734 if (!switch_root_init
) {
1735 char sfd
[DECIMAL_STR_MAX(int) + 1];
1737 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1738 * the user didn't specify an explicit init to spawn. */
1740 assert(arg_serialization
);
1743 xsprintf(sfd
, "%i", fileno(arg_serialization
));
1746 args
[i
++] = SYSTEMD_BINARY_PATH
;
1747 if (switch_root_dir
)
1748 args
[i
++] = "--switched-root";
1749 args
[i
++] = arg_system
? "--system" : "--user";
1750 args
[i
++] = "--deserialize";
1754 assert(i
<= args_size
);
1757 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do this is on
1758 * its own on exec(), but it will do it on exit(). Hence, to ensure we get a summary here, fork() off
1759 * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1760 * before proceeding into the exec().
1762 valgrind_summary_hack();
1764 (void) execv(args
[0], (char* const*) args
);
1765 log_debug_errno(errno
, "Failed to execute our own binary, trying fallback: %m");
1768 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1769 * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1770 * doesn't matter.) */
1772 arg_serialization
= safe_fclose(arg_serialization
);
1773 fds
= fdset_free(fds
);
1775 /* Reopen the console */
1776 (void) make_console_stdio();
1778 for (j
= 1, i
= 1; j
< (unsigned) argc
; j
++)
1779 args
[i
++] = argv
[j
];
1781 assert(i
<= args_size
);
1783 /* Reenable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1784 (void) reset_all_signal_handlers();
1785 (void) reset_signal_mask();
1786 (void) rlimit_nofile_safe();
1788 if (switch_root_init
) {
1789 args
[0] = switch_root_init
;
1790 (void) execv(args
[0], (char* const*) args
);
1791 log_warning_errno(errno
, "Failed to execute configured init, trying fallback: %m");
1794 args
[0] = "/sbin/init";
1795 (void) execv(args
[0], (char* const*) args
);
1798 manager_status_printf(NULL
, STATUS_TYPE_EMERGENCY
,
1799 ANSI_HIGHLIGHT_RED
" !! " ANSI_NORMAL
,
1800 "Failed to execute /sbin/init");
1803 log_warning("No /sbin/init, trying fallback");
1805 args
[0] = "/bin/sh";
1807 (void) execv(args
[0], (char* const*) args
);
1808 log_error_errno(errno
, "Failed to execute /bin/sh, giving up: %m");
1810 log_warning_errno(r
, "Failed to execute /sbin/init, giving up: %m");
1812 *ret_error_message
= "Failed to execute fallback shell";
1815 static int invoke_main_loop(
1817 bool *ret_reexecute
,
1818 int *ret_retval
, /* Return parameters relevant for shutting down */
1819 const char **ret_shutdown_verb
, /* … */
1820 FDSet
**ret_fds
, /* Return parameters for reexecuting */
1821 char **ret_switch_root_dir
, /* … */
1822 char **ret_switch_root_init
, /* … */
1823 const char **ret_error_message
) {
1828 assert(ret_reexecute
);
1830 assert(ret_shutdown_verb
);
1832 assert(ret_switch_root_dir
);
1833 assert(ret_switch_root_init
);
1834 assert(ret_error_message
);
1837 r
= manager_loop(m
);
1839 *ret_error_message
= "Failed to run main loop";
1840 return log_emergency_errno(r
, "Failed to run main loop: %m");
1843 switch ((ManagerObjective
) r
) {
1845 case MANAGER_RELOAD
: {
1846 LogTarget saved_log_target
;
1847 int saved_log_level
;
1849 log_info("Reloading.");
1851 /* First, save any overridden log level/target, then parse the configuration file, which might
1852 * change the log level to new settings. */
1854 saved_log_level
= m
->log_level_overridden
? log_get_max_level() : -1;
1855 saved_log_target
= m
->log_target_overridden
? log_get_target() : _LOG_TARGET_INVALID
;
1857 r
= parse_config_file();
1859 log_warning_errno(r
, "Failed to parse config file, ignoring: %m");
1861 set_manager_defaults(m
);
1863 if (saved_log_level
>= 0)
1864 manager_override_log_level(m
, saved_log_level
);
1865 if (saved_log_target
>= 0)
1866 manager_override_log_target(m
, saved_log_target
);
1868 r
= manager_reload(m
);
1870 /* Reloading failed before the point of no return. Let's continue running as if nothing happened. */
1871 m
->objective
= MANAGER_OK
;
1876 case MANAGER_REEXECUTE
:
1878 r
= prepare_reexecute(m
, &arg_serialization
, ret_fds
, false);
1880 *ret_error_message
= "Failed to prepare for reexecution";
1884 log_notice("Reexecuting.");
1886 *ret_reexecute
= true;
1887 *ret_retval
= EXIT_SUCCESS
;
1888 *ret_shutdown_verb
= NULL
;
1889 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
1893 case MANAGER_SWITCH_ROOT
:
1894 if (!m
->switch_root_init
) {
1895 r
= prepare_reexecute(m
, &arg_serialization
, ret_fds
, true);
1897 *ret_error_message
= "Failed to prepare for reexecution";
1903 log_notice("Switching root.");
1905 *ret_reexecute
= true;
1906 *ret_retval
= EXIT_SUCCESS
;
1907 *ret_shutdown_verb
= NULL
;
1909 /* Steal the switch root parameters */
1910 *ret_switch_root_dir
= m
->switch_root
;
1911 *ret_switch_root_init
= m
->switch_root_init
;
1912 m
->switch_root
= m
->switch_root_init
= NULL
;
1918 if (MANAGER_IS_USER(m
)) {
1921 *ret_reexecute
= false;
1922 *ret_retval
= m
->return_value
;
1923 *ret_shutdown_verb
= NULL
;
1925 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
1931 case MANAGER_REBOOT
:
1932 case MANAGER_POWEROFF
:
1934 case MANAGER_KEXEC
: {
1935 static const char * const table
[_MANAGER_OBJECTIVE_MAX
] = {
1936 [MANAGER_EXIT
] = "exit",
1937 [MANAGER_REBOOT
] = "reboot",
1938 [MANAGER_POWEROFF
] = "poweroff",
1939 [MANAGER_HALT
] = "halt",
1940 [MANAGER_KEXEC
] = "kexec",
1943 log_notice("Shutting down.");
1945 *ret_reexecute
= false;
1946 *ret_retval
= m
->return_value
;
1947 assert_se(*ret_shutdown_verb
= table
[m
->objective
]);
1949 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
1955 assert_not_reached("Unknown or unexpected manager objective.");
1960 static void log_execution_mode(bool *ret_first_boot
) {
1961 assert(ret_first_boot
);
1966 log_info("systemd " GIT_VERSION
" running in %ssystem mode. (" SYSTEMD_FEATURES
")",
1967 arg_action
== ACTION_TEST
? "test " : "" );
1969 v
= detect_virtualization();
1971 log_info("Detected virtualization %s.", virtualization_to_string(v
));
1973 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1976 *ret_first_boot
= false;
1977 log_info("Running in initial RAM disk.");
1979 /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
1980 * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
1981 * doesn't it's unpopulated. This allows container managers and installers to provision a
1982 * couple of files already. If the container manager wants to provision the machine ID itself
1983 * it should pass $container_uuid to PID 1. */
1985 *ret_first_boot
= access("/etc/machine-id", F_OK
) < 0;
1986 if (*ret_first_boot
)
1987 log_info("Running with unpopulated /etc.");
1990 if (DEBUG_LOGGING
) {
1991 _cleanup_free_
char *t
;
1993 t
= uid_to_name(getuid());
1994 log_debug("systemd " GIT_VERSION
" running in %suser mode for user " UID_FMT
"/%s. (" SYSTEMD_FEATURES
")",
1995 arg_action
== ACTION_TEST
? " test" : "", getuid(), strna(t
));
1998 *ret_first_boot
= false;
2002 static int initialize_runtime(
2004 struct rlimit
*saved_rlimit_nofile
,
2005 struct rlimit
*saved_rlimit_memlock
,
2006 const char **ret_error_message
) {
2010 assert(ret_error_message
);
2012 /* Sets up various runtime parameters. Many of these initializations are conditionalized:
2014 * - Some only apply to --system instances
2015 * - Some only apply to --user instances
2016 * - Some only apply when we first start up, but not when we reexecute
2019 if (arg_action
!= ACTION_RUN
)
2023 /* Make sure we leave a core dump without panicing the kernel. */
2024 install_crash_handler();
2027 r
= mount_cgroup_controllers();
2029 *ret_error_message
= "Failed to mount cgroup hierarchies";
2035 machine_id_setup(NULL
, arg_machine_id
, NULL
);
2037 bump_unix_max_dgram_qlen();
2038 bump_file_max_and_nr_open();
2040 write_container_id();
2043 if (arg_watchdog_device
) {
2044 r
= watchdog_set_device(arg_watchdog_device
);
2046 log_warning_errno(r
, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device
);
2049 if (arg_runtime_watchdog
> 0 && arg_runtime_watchdog
!= USEC_INFINITY
)
2050 watchdog_set_timeout(&arg_runtime_watchdog
);
2053 if (arg_timer_slack_nsec
!= NSEC_INFINITY
)
2054 if (prctl(PR_SET_TIMERSLACK
, arg_timer_slack_nsec
) < 0)
2055 log_warning_errno(errno
, "Failed to adjust timer slack, ignoring: %m");
2057 if (arg_system
&& !cap_test_all(arg_capability_bounding_set
)) {
2058 r
= capability_bounding_set_drop_usermode(arg_capability_bounding_set
);
2060 *ret_error_message
= "Failed to drop capability bounding set of usermode helpers";
2061 return log_emergency_errno(r
, "Failed to drop capability bounding set of usermode helpers: %m");
2064 r
= capability_bounding_set_drop(arg_capability_bounding_set
, true);
2066 *ret_error_message
= "Failed to drop capability bounding set";
2067 return log_emergency_errno(r
, "Failed to drop capability bounding set: %m");
2071 if (arg_system
&& arg_no_new_privs
) {
2072 if (prctl(PR_SET_NO_NEW_PRIVS
, 1, 0, 0, 0) < 0) {
2073 *ret_error_message
= "Failed to disable new privileges";
2074 return log_emergency_errno(errno
, "Failed to disable new privileges: %m");
2078 if (arg_syscall_archs
) {
2079 r
= enforce_syscall_archs(arg_syscall_archs
);
2081 *ret_error_message
= "Failed to set syscall architectures";
2087 /* Become reaper of our children */
2088 if (prctl(PR_SET_CHILD_SUBREAPER
, 1) < 0)
2089 log_warning_errno(errno
, "Failed to make us a subreaper: %m");
2091 /* Bump up RLIMIT_NOFILE for systemd itself */
2092 (void) bump_rlimit_nofile(saved_rlimit_nofile
);
2093 (void) bump_rlimit_memlock(saved_rlimit_memlock
);
2098 static int do_queue_default_job(
2100 const char **ret_error_message
) {
2102 _cleanup_(sd_bus_error_free
) sd_bus_error error
= SD_BUS_ERROR_NULL
;
2103 Job
*default_unit_job
;
2104 Unit
*target
= NULL
;
2107 log_debug("Activating default unit: %s", arg_default_unit
);
2109 r
= manager_load_startable_unit_or_warn(m
, arg_default_unit
, NULL
, &target
);
2111 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET
);
2113 r
= manager_load_startable_unit_or_warn(m
, SPECIAL_RESCUE_TARGET
, NULL
, &target
);
2115 *ret_error_message
= r
== -ERFKILL
? "Rescue target masked"
2116 : "Failed to load rescue target";
2121 assert(target
->load_state
== UNIT_LOADED
);
2123 r
= manager_add_job(m
, JOB_START
, target
, JOB_ISOLATE
, &error
, &default_unit_job
);
2125 log_debug_errno(r
, "Default target could not be isolated, starting instead: %s", bus_error_message(&error
, r
));
2127 sd_bus_error_free(&error
);
2129 r
= manager_add_job(m
, JOB_START
, target
, JOB_REPLACE
, &error
, &default_unit_job
);
2131 *ret_error_message
= "Failed to start default target";
2132 return log_emergency_errno(r
, "Failed to start default target: %s", bus_error_message(&error
, r
));
2136 *ret_error_message
= "Failed to isolate default target";
2137 return log_emergency_errno(r
, "Failed to isolate default target: %s", bus_error_message(&error
, r
));
2140 m
->default_unit_job_id
= default_unit_job
->id
;
2145 static void free_arguments(void) {
2147 /* Frees all arg_* variables, with the exception of arg_serialization */
2148 rlimit_free_all(arg_default_rlimit
);
2150 arg_default_unit
= mfree(arg_default_unit
);
2151 arg_confirm_spawn
= mfree(arg_confirm_spawn
);
2152 arg_default_environment
= strv_free(arg_default_environment
);
2153 arg_syscall_archs
= set_free(arg_syscall_archs
);
2156 static int load_configuration(int argc
, char **argv
, const char **ret_error_message
) {
2159 assert(ret_error_message
);
2161 arg_default_tasks_max
= system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE
, 100U);
2163 r
= parse_config_file();
2165 *ret_error_message
= "Failed to parse config file";
2170 r
= proc_cmdline_parse(parse_proc_cmdline_item
, NULL
, 0);
2172 log_warning_errno(r
, "Failed to parse kernel command line, ignoring: %m");
2175 /* Note that this also parses bits from the kernel command line, including "debug". */
2176 log_parse_environment();
2178 r
= parse_argv(argc
, argv
);
2180 *ret_error_message
= "Failed to parse commandline arguments";
2184 /* Initialize default unit */
2185 if (!arg_default_unit
) {
2186 arg_default_unit
= strdup(SPECIAL_DEFAULT_TARGET
);
2187 if (!arg_default_unit
) {
2188 *ret_error_message
= "Failed to set default unit";
2193 /* Initialize the show status setting if it hasn't been set explicitly yet */
2194 if (arg_show_status
== _SHOW_STATUS_INVALID
)
2195 arg_show_status
= SHOW_STATUS_YES
;
2200 static int safety_checks(void) {
2202 if (getpid_cached() == 1 &&
2203 arg_action
!= ACTION_RUN
)
2204 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
2205 "Unsupported execution mode while PID 1.");
2207 if (getpid_cached() == 1 &&
2209 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
2210 "Can't run --user mode as PID 1.");
2212 if (arg_action
== ACTION_RUN
&&
2214 getpid_cached() != 1)
2215 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
2216 "Can't run system mode unless PID 1.");
2218 if (arg_action
== ACTION_TEST
&&
2220 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
2221 "Don't run test mode as root.");
2224 arg_action
== ACTION_RUN
&&
2226 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP
),
2227 "Trying to run as user instance, but the system has not been booted with systemd.");
2230 arg_action
== ACTION_RUN
&&
2231 !getenv("XDG_RUNTIME_DIR"))
2232 return log_error_errno(SYNTHETIC_ERRNO(EUNATCH
),
2233 "Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
2236 arg_action
== ACTION_RUN
&&
2237 running_in_chroot() > 0)
2238 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP
),
2239 "Cannot be run in a chroot() environment.");
2244 static int initialize_security(
2245 bool *loaded_policy
,
2246 dual_timestamp
*security_start_timestamp
,
2247 dual_timestamp
*security_finish_timestamp
,
2248 const char **ret_error_message
) {
2252 assert(loaded_policy
);
2253 assert(security_start_timestamp
);
2254 assert(security_finish_timestamp
);
2255 assert(ret_error_message
);
2257 dual_timestamp_get(security_start_timestamp
);
2259 r
= mac_selinux_setup(loaded_policy
);
2261 *ret_error_message
= "Failed to load SELinux policy";
2265 r
= mac_smack_setup(loaded_policy
);
2267 *ret_error_message
= "Failed to load SMACK policy";
2273 *ret_error_message
= "Failed to load IMA policy";
2277 dual_timestamp_get(security_finish_timestamp
);
2281 static void test_summary(Manager
*m
) {
2284 printf("-> By units:\n");
2285 manager_dump_units(m
, stdout
, "\t");
2287 printf("-> By jobs:\n");
2288 manager_dump_jobs(m
, stdout
, "\t");
2291 static int collect_fds(FDSet
**ret_fds
, const char **ret_error_message
) {
2295 assert(ret_error_message
);
2297 r
= fdset_new_fill(ret_fds
);
2299 *ret_error_message
= "Failed to allocate fd set";
2300 return log_emergency_errno(r
, "Failed to allocate fd set: %m");
2303 fdset_cloexec(*ret_fds
, true);
2305 if (arg_serialization
)
2306 assert_se(fdset_remove(*ret_fds
, fileno(arg_serialization
)) >= 0);
2311 static void setup_console_terminal(bool skip_setup
) {
2316 /* Become a session leader if we aren't one yet. */
2319 /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2321 (void) release_terminal();
2323 /* Reset the console, but only if this is really init and we are freshly booted */
2324 if (getpid_cached() == 1 && !skip_setup
)
2325 (void) console_setup();
2328 static bool early_skip_setup_check(int argc
, char *argv
[]) {
2329 bool found_deserialize
= false;
2332 /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2333 * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2334 * anyway, even if in that case we also do deserialization. */
2336 for (i
= 1; i
< argc
; i
++) {
2337 if (streq(argv
[i
], "--switched-root"))
2338 return false; /* If we switched root, don't skip the setup. */
2339 else if (streq(argv
[i
], "--deserialize"))
2340 found_deserialize
= true;
2343 return found_deserialize
; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2346 int main(int argc
, char *argv
[]) {
2348 dual_timestamp initrd_timestamp
= DUAL_TIMESTAMP_NULL
, userspace_timestamp
= DUAL_TIMESTAMP_NULL
, kernel_timestamp
= DUAL_TIMESTAMP_NULL
,
2349 security_start_timestamp
= DUAL_TIMESTAMP_NULL
, security_finish_timestamp
= DUAL_TIMESTAMP_NULL
;
2350 struct rlimit saved_rlimit_nofile
= RLIMIT_MAKE_CONST(0),
2351 saved_rlimit_memlock
= RLIMIT_MAKE_CONST(RLIM_INFINITY
); /* The original rlimits we passed
2352 * in. Note we use different values
2353 * for the two that indicate whether
2354 * these fields are initialized! */
2355 bool skip_setup
, loaded_policy
= false, queue_default_job
= false, first_boot
= false, reexecute
= false;
2356 char *switch_root_dir
= NULL
, *switch_root_init
= NULL
;
2357 usec_t before_startup
, after_startup
;
2358 static char systemd
[] = "systemd";
2359 char timespan
[FORMAT_TIMESPAN_MAX
];
2360 const char *shutdown_verb
= NULL
, *error_message
= NULL
;
2361 int r
, retval
= EXIT_FAILURE
;
2365 /* SysV compatibility: redirect init → telinit */
2366 redirect_telinit(argc
, argv
);
2368 /* Take timestamps early on */
2369 dual_timestamp_from_monotonic(&kernel_timestamp
, 0);
2370 dual_timestamp_get(&userspace_timestamp
);
2372 /* Figure out whether we need to do initialize the system, or if we already did that because we are
2374 skip_setup
= early_skip_setup_check(argc
, argv
);
2376 /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2377 * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
2378 program_invocation_short_name
= systemd
;
2379 (void) prctl(PR_SET_NAME
, systemd
);
2381 /* Save the original command line */
2385 /* Make sure that if the user says "syslog" we actually log to the journal. */
2386 log_set_upgrade_syslog_to_journal(true);
2388 if (getpid_cached() == 1) {
2389 /* When we run as PID 1 force system mode */
2392 /* Disable the umask logic */
2395 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2396 * activated yet (even though the log socket for it exists). */
2397 log_set_prohibit_ipc(true);
2399 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2400 * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2401 * child process right before execve()'ing the actual binary, at a point in time where socket
2402 * activation stderr/stdout area already set up. */
2403 log_set_always_reopen_console(true);
2405 if (detect_container() <= 0) {
2407 /* Running outside of a container as PID 1 */
2408 log_set_target(LOG_TARGET_KMSG
);
2412 initrd_timestamp
= userspace_timestamp
;
2415 r
= mount_setup_early();
2417 error_message
= "Failed to mount early API filesystems";
2421 r
= initialize_security(
2423 &security_start_timestamp
,
2424 &security_finish_timestamp
,
2430 if (mac_selinux_init() < 0) {
2431 error_message
= "Failed to initialize SELinux policy";
2438 /* Set the default for later on, but don't actually open the logs like this for now. Note that
2439 * if we are transitioning from the initrd there might still be journal fd open, and we
2440 * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2442 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG
);
2445 /* Running inside a container, as PID 1 */
2446 log_set_target(LOG_TARGET_CONSOLE
);
2449 /* For later on, see above... */
2450 log_set_target(LOG_TARGET_JOURNAL
);
2452 /* clear the kernel timestamp,
2453 * because we are in a container */
2454 kernel_timestamp
= DUAL_TIMESTAMP_NULL
;
2457 initialize_coredump(skip_setup
);
2459 r
= fixup_environment();
2461 log_emergency_errno(r
, "Failed to fix up PID 1 environment: %m");
2462 error_message
= "Failed to fix up PID1 environment";
2467 /* Running as user instance */
2469 log_set_target(LOG_TARGET_AUTO
);
2472 /* clear the kernel timestamp,
2473 * because we are not PID 1 */
2474 kernel_timestamp
= DUAL_TIMESTAMP_NULL
;
2478 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2479 * they never log into the console. */
2480 log_show_color(colors_enabled());
2482 r
= make_null_stdio();
2484 log_warning_errno(r
, "Failed to redirect standard streams to /dev/null, ignoring: %m");
2487 /* Mount /proc, /sys and friends, so that /proc/cmdline and
2488 * /proc/$PID/fd is available. */
2489 if (getpid_cached() == 1) {
2491 /* Load the kernel modules early. */
2495 r
= mount_setup(loaded_policy
);
2497 error_message
= "Failed to mount API filesystems";
2502 /* Reset all signal handlers. */
2503 (void) reset_all_signal_handlers();
2504 (void) ignore_signals(SIGNALS_IGNORE
, -1);
2506 r
= load_configuration(argc
, argv
, &error_message
);
2510 r
= safety_checks();
2514 if (IN_SET(arg_action
, ACTION_TEST
, ACTION_HELP
, ACTION_DUMP_CONFIGURATION_ITEMS
, ACTION_DUMP_BUS_PROPERTIES
))
2515 (void) pager_open(arg_pager_flags
);
2517 if (arg_action
!= ACTION_RUN
)
2520 if (arg_action
== ACTION_HELP
) {
2521 retval
= help() < 0 ? EXIT_FAILURE
: EXIT_SUCCESS
;
2523 } else if (arg_action
== ACTION_VERSION
) {
2526 } else if (arg_action
== ACTION_DUMP_CONFIGURATION_ITEMS
) {
2527 unit_dump_config_items(stdout
);
2528 retval
= EXIT_SUCCESS
;
2530 } else if (arg_action
== ACTION_DUMP_BUS_PROPERTIES
) {
2531 dump_bus_properties(stdout
);
2532 retval
= EXIT_SUCCESS
;
2536 assert_se(IN_SET(arg_action
, ACTION_RUN
, ACTION_TEST
));
2538 /* Move out of the way, so that we won't block unmounts */
2539 assert_se(chdir("/") == 0);
2541 if (arg_action
== ACTION_RUN
) {
2543 /* A core pattern might have been specified via the cmdline. */
2544 initialize_core_pattern(skip_setup
);
2546 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2549 /* Remember open file descriptors for later deserialization */
2550 r
= collect_fds(&fds
, &error_message
);
2554 /* Give up any control of the console, but make sure its initialized. */
2555 setup_console_terminal(skip_setup
);
2557 /* Open the logging devices, if possible and necessary */
2561 log_execution_mode(&first_boot
);
2563 r
= initialize_runtime(skip_setup
,
2564 &saved_rlimit_nofile
,
2565 &saved_rlimit_memlock
,
2570 r
= manager_new(arg_system
? UNIT_FILE_SYSTEM
: UNIT_FILE_USER
,
2571 arg_action
== ACTION_TEST
? MANAGER_TEST_FULL
: 0,
2574 log_emergency_errno(r
, "Failed to allocate manager object: %m");
2575 error_message
= "Failed to allocate manager object";
2579 m
->timestamps
[MANAGER_TIMESTAMP_KERNEL
] = kernel_timestamp
;
2580 m
->timestamps
[MANAGER_TIMESTAMP_INITRD
] = initrd_timestamp
;
2581 m
->timestamps
[MANAGER_TIMESTAMP_USERSPACE
] = userspace_timestamp
;
2582 m
->timestamps
[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START
)] = security_start_timestamp
;
2583 m
->timestamps
[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH
)] = security_finish_timestamp
;
2585 set_manager_defaults(m
);
2586 set_manager_settings(m
);
2587 manager_set_first_boot(m
, first_boot
);
2589 /* Remember whether we should queue the default job */
2590 queue_default_job
= !arg_serialization
|| arg_switched_root
;
2592 before_startup
= now(CLOCK_MONOTONIC
);
2594 r
= manager_startup(m
, arg_serialization
, fds
);
2596 error_message
= "Failed to start up manager";
2600 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2601 fds
= fdset_free(fds
);
2602 arg_serialization
= safe_fclose(arg_serialization
);
2604 if (queue_default_job
) {
2605 r
= do_queue_default_job(m
, &error_message
);
2610 after_startup
= now(CLOCK_MONOTONIC
);
2612 log_full(arg_action
== ACTION_TEST
? LOG_INFO
: LOG_DEBUG
,
2613 "Loaded units and determined initial transaction in %s.",
2614 format_timespan(timespan
, sizeof(timespan
), after_startup
- before_startup
, 100 * USEC_PER_MSEC
));
2616 if (arg_action
== ACTION_TEST
) {
2618 retval
= EXIT_SUCCESS
;
2622 (void) invoke_main_loop(m
,
2635 arg_shutdown_watchdog
= m
->shutdown_watchdog
;
2636 m
= manager_free(m
);
2640 mac_selinux_finish();
2643 do_reexecute(argc
, argv
,
2644 &saved_rlimit_nofile
,
2645 &saved_rlimit_memlock
,
2649 &error_message
); /* This only returns if reexecution failed */
2651 arg_serialization
= safe_fclose(arg_serialization
);
2652 fds
= fdset_free(fds
);
2654 #if HAVE_VALGRIND_VALGRIND_H
2655 /* If we are PID 1 and running under valgrind, then let's exit
2656 * here explicitly. valgrind will only generate nice output on
2657 * exit(), not on exec(), hence let's do the former not the
2659 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND
) {
2660 /* Cleanup watchdog_device strings for valgrind. We need them
2661 * in become_shutdown() so normally we cannot free them yet. */
2662 watchdog_free_device();
2663 arg_watchdog_device
= mfree(arg_watchdog_device
);
2668 #if HAS_FEATURE_ADDRESS_SANITIZER
2669 __lsan_do_leak_check();
2672 if (shutdown_verb
) {
2673 r
= become_shutdown(shutdown_verb
, retval
);
2674 log_error_errno(r
, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
2675 error_message
= "Failed to execute shutdown binary";
2678 watchdog_free_device();
2679 arg_watchdog_device
= mfree(arg_watchdog_device
);
2681 if (getpid_cached() == 1) {
2683 manager_status_printf(NULL
, STATUS_TYPE_EMERGENCY
,
2684 ANSI_HIGHLIGHT_RED
"!!!!!!" ANSI_NORMAL
,
2685 "%s.", error_message
);
2686 freeze_or_exit_or_reboot();