2 This file is part of systemd.
4 Copyright 2010 Lennart Poettering
6 systemd is free software; you can redistribute it and/or modify it
7 under the terms of the GNU Lesser General Public License as published by
8 the Free Software Foundation; either version 2.1 of the License, or
9 (at your option) any later version.
11 systemd is distributed in the hope that it will be useful, but
12 WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
16 You should have received a copy of the GNU Lesser General Public License
17 along with systemd; If not, see <http://www.gnu.org/licenses/>.
26 #include <sys/mount.h>
27 #include <sys/prctl.h>
28 #include <sys/reboot.h>
34 #ifdef HAVE_VALGRIND_VALGRIND_H
35 #include <valgrind/valgrind.h>
39 #include "sd-daemon.h"
41 #include "alloc-util.h"
42 #include "architecture.h"
44 #include "bus-error.h"
46 #include "capability-util.h"
47 #include "clock-util.h"
48 #include "conf-parser.h"
49 #include "cpu-set-util.h"
50 #include "dbus-manager.h"
52 #include "emergency-action.h"
57 #include "format-util.h"
59 #include "hostname-setup.h"
60 #include "ima-setup.h"
62 #include "kmod-setup.h"
63 #include "load-fragment.h"
65 #include "loopback-setup.h"
66 #include "machine-id-setup.h"
69 #include "mount-setup.h"
71 #include "parse-util.h"
72 #include "path-util.h"
73 #include "proc-cmdline.h"
74 #include "process-util.h"
75 #include "raw-clone.h"
76 #include "rlimit-util.h"
78 #include "seccomp-util.h"
80 #include "selinux-setup.h"
81 #include "selinux-util.h"
82 #include "signal-util.h"
83 #include "smack-setup.h"
85 #include "stat-util.h"
86 #include "stdio-util.h"
88 #include "switch-root.h"
89 #include "terminal-util.h"
90 #include "umask-util.h"
91 #include "user-util.h"
100 ACTION_DUMP_CONFIGURATION_ITEMS
101 } arg_action
= ACTION_RUN
;
102 static char *arg_default_unit
= NULL
;
103 static bool arg_system
= false;
104 static bool arg_dump_core
= true;
105 static int arg_crash_chvt
= -1;
106 static bool arg_crash_shell
= false;
107 static bool arg_crash_reboot
= false;
108 static char *arg_confirm_spawn
= NULL
;
109 static ShowStatus arg_show_status
= _SHOW_STATUS_UNSET
;
110 static bool arg_switched_root
= false;
111 static bool arg_no_pager
= false;
112 static char ***arg_join_controllers
= NULL
;
113 static ExecOutput arg_default_std_output
= EXEC_OUTPUT_JOURNAL
;
114 static ExecOutput arg_default_std_error
= EXEC_OUTPUT_INHERIT
;
115 static usec_t arg_default_restart_usec
= DEFAULT_RESTART_USEC
;
116 static usec_t arg_default_timeout_start_usec
= DEFAULT_TIMEOUT_USEC
;
117 static usec_t arg_default_timeout_stop_usec
= DEFAULT_TIMEOUT_USEC
;
118 static usec_t arg_default_start_limit_interval
= DEFAULT_START_LIMIT_INTERVAL
;
119 static unsigned arg_default_start_limit_burst
= DEFAULT_START_LIMIT_BURST
;
120 static usec_t arg_runtime_watchdog
= 0;
121 static usec_t arg_shutdown_watchdog
= 10 * USEC_PER_MINUTE
;
122 static char **arg_default_environment
= NULL
;
123 static struct rlimit
*arg_default_rlimit
[_RLIMIT_MAX
] = {};
124 static uint64_t arg_capability_bounding_set
= CAP_ALL
;
125 static nsec_t arg_timer_slack_nsec
= NSEC_INFINITY
;
126 static usec_t arg_default_timer_accuracy_usec
= 1 * USEC_PER_MINUTE
;
127 static Set
* arg_syscall_archs
= NULL
;
128 static FILE* arg_serialization
= NULL
;
129 static bool arg_default_cpu_accounting
= false;
130 static bool arg_default_io_accounting
= false;
131 static bool arg_default_blockio_accounting
= false;
132 static bool arg_default_memory_accounting
= false;
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_reboot(void) {
140 if (arg_crash_reboot
) {
141 log_notice("Rebooting in 10s...");
144 log_notice("Rebooting now...");
145 (void) reboot(RB_AUTOBOOT
);
146 log_emergency_errno(errno
, "Failed to reboot: %m");
149 log_emergency("Freezing execution.");
153 noreturn
static void crash(int sig
) {
158 /* Pass this on immediately, if this is not PID 1 */
160 else if (!arg_dump_core
)
161 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig
));
163 sa
= (struct sigaction
) {
164 .sa_handler
= nop_signal_handler
,
165 .sa_flags
= SA_NOCLDSTOP
|SA_RESTART
,
168 /* We want to wait for the core process, hence let's enable SIGCHLD */
169 (void) sigaction(SIGCHLD
, &sa
, NULL
);
171 pid
= raw_clone(SIGCHLD
);
173 log_emergency_errno(errno
, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig
));
175 /* Enable default signal handler for core dump */
177 sa
= (struct sigaction
) {
178 .sa_handler
= SIG_DFL
,
180 (void) sigaction(sig
, &sa
, NULL
);
182 /* Don't limit the coredump size */
183 (void) setrlimit(RLIMIT_CORE
, &RLIMIT_MAKE_CONST(RLIM_INFINITY
));
185 /* Just to be sure... */
188 /* Raise the signal again */
190 (void) kill(pid
, sig
); /* raise() would kill the parent */
192 assert_not_reached("We shouldn't be here...");
198 /* Order things nicely. */
199 r
= wait_for_terminate(pid
, &status
);
201 log_emergency_errno(r
, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig
));
202 else if (status
.si_code
!= CLD_DUMPED
)
203 log_emergency("Caught <%s>, core dump failed (child "PID_FMT
", code=%s, status=%i/%s).",
204 signal_to_string(sig
),
205 pid
, sigchld_code_to_string(status
.si_code
),
207 strna(status
.si_code
== CLD_EXITED
208 ? exit_status_to_string(status
.si_status
, EXIT_STATUS_MINIMAL
)
209 : signal_to_string(status
.si_status
)));
211 log_emergency("Caught <%s>, dumped core as pid "PID_FMT
".", signal_to_string(sig
), pid
);
215 if (arg_crash_chvt
>= 0)
216 (void) chvt(arg_crash_chvt
);
218 sa
= (struct sigaction
) {
219 .sa_handler
= SIG_IGN
,
220 .sa_flags
= SA_NOCLDSTOP
|SA_NOCLDWAIT
|SA_RESTART
,
223 /* Let the kernel reap children for us */
224 (void) sigaction(SIGCHLD
, &sa
, NULL
);
226 if (arg_crash_shell
) {
227 log_notice("Executing crash shell in 10s...");
230 pid
= raw_clone(SIGCHLD
);
232 log_emergency_errno(errno
, "Failed to fork off crash shell: %m");
235 (void) make_console_stdio();
236 (void) execle("/bin/sh", "/bin/sh", NULL
, environ
);
238 log_emergency_errno(errno
, "execle() failed: %m");
241 log_info("Spawned crash shell as PID "PID_FMT
".", pid
);
242 (void) wait_for_terminate(pid
, NULL
);
249 static void install_crash_handler(void) {
250 static const struct sigaction sa
= {
252 .sa_flags
= SA_NODEFER
, /* So that we can raise the signal again from the signal handler */
256 /* We ignore the return value here, since, we don't mind if we
257 * cannot set up a crash handler */
258 r
= sigaction_many(&sa
, SIGNALS_CRASH_HANDLER
, -1);
260 log_debug_errno(r
, "I had trouble setting up the crash handler, ignoring: %m");
263 static int console_setup(void) {
264 _cleanup_close_
int tty_fd
= -1;
267 tty_fd
= open_terminal("/dev/console", O_WRONLY
|O_NOCTTY
|O_CLOEXEC
);
269 return log_error_errno(tty_fd
, "Failed to open /dev/console: %m");
271 /* We don't want to force text mode. plymouth may be showing
272 * pictures already from initrd. */
273 r
= reset_terminal_fd(tty_fd
, false);
275 return log_error_errno(r
, "Failed to reset /dev/console: %m");
280 static int parse_crash_chvt(const char *value
) {
283 if (safe_atoi(value
, &arg_crash_chvt
) >= 0)
286 b
= parse_boolean(value
);
291 arg_crash_chvt
= 0; /* switch to where kmsg goes */
293 arg_crash_chvt
= -1; /* turn off switching */
298 static int parse_confirm_spawn(const char *value
, char **console
) {
302 r
= value
? parse_boolean(value
) : 1;
308 if (r
> 0) /* on with default tty */
309 s
= strdup("/dev/console");
310 else if (is_path(value
)) /* on with fully qualified path */
312 else /* on with only a tty file name, not a fully qualified path */
313 s
= strjoin("/dev/", value
);
320 static int set_machine_id(const char *m
) {
324 if (sd_id128_from_string(m
, &t
) < 0)
327 if (sd_id128_is_null(t
))
334 static int parse_proc_cmdline_item(const char *key
, const char *value
, void *data
) {
340 if (STR_IN_SET(key
, "systemd.unit", "rd.systemd.unit")) {
342 if (proc_cmdline_value_missing(key
, value
))
345 if (!unit_name_is_valid(value
, UNIT_NAME_PLAIN
|UNIT_NAME_INSTANCE
))
346 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key
, value
);
347 else if (in_initrd() == !!startswith(key
, "rd.")) {
348 if (free_and_strdup(&arg_default_unit
, value
) < 0)
352 } else if (proc_cmdline_key_streq(key
, "systemd.dump_core")) {
354 r
= value
? parse_boolean(value
) : true;
356 log_warning("Failed to parse dump core switch %s. Ignoring.", value
);
360 } else if (proc_cmdline_key_streq(key
, "systemd.crash_chvt")) {
363 arg_crash_chvt
= 0; /* turn on */
364 else if (parse_crash_chvt(value
) < 0)
365 log_warning("Failed to parse crash chvt switch %s. Ignoring.", value
);
367 } else if (proc_cmdline_key_streq(key
, "systemd.crash_shell")) {
369 r
= value
? parse_boolean(value
) : true;
371 log_warning("Failed to parse crash shell switch %s. Ignoring.", value
);
375 } else if (proc_cmdline_key_streq(key
, "systemd.crash_reboot")) {
377 r
= value
? parse_boolean(value
) : true;
379 log_warning("Failed to parse crash reboot switch %s. Ignoring.", value
);
381 arg_crash_reboot
= r
;
383 } else if (proc_cmdline_key_streq(key
, "systemd.confirm_spawn")) {
386 r
= parse_confirm_spawn(value
, &s
);
388 log_warning_errno(r
, "Failed to parse confirm_spawn switch %s. Ignoring.", value
);
390 free(arg_confirm_spawn
);
391 arg_confirm_spawn
= s
;
394 } else if (proc_cmdline_key_streq(key
, "systemd.show_status")) {
397 r
= parse_show_status(value
, &arg_show_status
);
399 log_warning("Failed to parse show status switch %s. Ignoring.", value
);
401 arg_show_status
= SHOW_STATUS_YES
;
403 } else if (proc_cmdline_key_streq(key
, "systemd.default_standard_output")) {
405 if (proc_cmdline_value_missing(key
, value
))
408 r
= exec_output_from_string(value
);
410 log_warning("Failed to parse default standard output switch %s. Ignoring.", value
);
412 arg_default_std_output
= r
;
414 } else if (proc_cmdline_key_streq(key
, "systemd.default_standard_error")) {
416 if (proc_cmdline_value_missing(key
, value
))
419 r
= exec_output_from_string(value
);
421 log_warning("Failed to parse default standard error switch %s. Ignoring.", value
);
423 arg_default_std_error
= r
;
425 } else if (streq(key
, "systemd.setenv")) {
427 if (proc_cmdline_value_missing(key
, value
))
430 if (env_assignment_is_valid(value
)) {
433 env
= strv_env_set(arg_default_environment
, value
);
437 arg_default_environment
= env
;
439 log_warning("Environment variable name '%s' is not valid. Ignoring.", value
);
441 } else if (proc_cmdline_key_streq(key
, "systemd.machine_id")) {
443 if (proc_cmdline_value_missing(key
, value
))
446 r
= set_machine_id(value
);
448 log_warning("MachineID '%s' is not valid. Ignoring.", value
);
450 } else if (proc_cmdline_key_streq(key
, "systemd.default_timeout_start_sec")) {
452 if (proc_cmdline_value_missing(key
, value
))
455 r
= parse_sec(value
, &arg_default_timeout_start_usec
);
457 log_warning_errno(r
, "Failed to parse default start timeout: %s, ignoring.", value
);
459 if (arg_default_timeout_start_usec
<= 0)
460 arg_default_timeout_start_usec
= USEC_INFINITY
;
462 } else if (streq(key
, "quiet") && !value
) {
464 if (arg_show_status
== _SHOW_STATUS_UNSET
)
465 arg_show_status
= SHOW_STATUS_AUTO
;
467 } else if (streq(key
, "debug") && !value
) {
469 /* Note that log_parse_environment() handles 'debug'
470 * too, and sets the log level to LOG_DEBUG. */
472 if (detect_container() > 0)
473 log_set_target(LOG_TARGET_CONSOLE
);
478 /* SysV compatibility */
479 target
= runlevel_to_target(key
);
481 return free_and_strdup(&arg_default_unit
, target
);
487 #define DEFINE_SETTER(name, func, descr) \
488 static int name(const char *unit, \
489 const char *filename, \
491 const char *section, \
492 unsigned section_line, \
493 const char *lvalue, \
495 const char *rvalue, \
507 log_syntax(unit, LOG_ERR, filename, line, r, \
508 "Invalid " descr "'%s': %m", \
514 DEFINE_SETTER(config_parse_level2
, log_set_max_level_from_string
, "log level")
515 DEFINE_SETTER(config_parse_target
, log_set_target_from_string
, "target")
516 DEFINE_SETTER(config_parse_color
, log_show_color_from_string
, "color" )
517 DEFINE_SETTER(config_parse_location
, log_show_location_from_string
, "location")
519 static int config_parse_cpu_affinity2(
521 const char *filename
,
524 unsigned section_line
,
531 _cleanup_cpu_free_ cpu_set_t
*c
= NULL
;
534 ncpus
= parse_cpu_set_and_warn(rvalue
, &c
, unit
, filename
, line
, lvalue
);
538 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus
), c
) < 0)
539 log_warning("Failed to set CPU affinity: %m");
544 static int config_parse_show_status(
546 const char *filename
,
549 unsigned section_line
,
557 ShowStatus
*b
= data
;
564 k
= parse_show_status(rvalue
, b
);
566 log_syntax(unit
, LOG_ERR
, filename
, line
, k
, "Failed to parse show status setting, ignoring: %s", rvalue
);
573 static int config_parse_crash_chvt(
575 const char *filename
,
578 unsigned section_line
,
591 r
= parse_crash_chvt(rvalue
);
593 log_syntax(unit
, LOG_ERR
, filename
, line
, r
, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue
);
600 static int config_parse_join_controllers(const char *unit
,
601 const char *filename
,
604 unsigned section_line
,
611 const char *whole_rvalue
= rvalue
;
618 arg_join_controllers
= strv_free_free(arg_join_controllers
);
621 _cleanup_free_
char *word
= NULL
;
625 r
= extract_first_word(&rvalue
, &word
, NULL
, EXTRACT_QUOTES
);
627 log_syntax(unit
, LOG_ERR
, filename
, line
, r
, "Invalid value for %s: %s", lvalue
, whole_rvalue
);
633 l
= strv_split(word
, ",");
638 if (strv_length(l
) <= 1) {
643 if (!arg_join_controllers
) {
644 arg_join_controllers
= new(char**, 2);
645 if (!arg_join_controllers
) {
650 arg_join_controllers
[0] = l
;
651 arg_join_controllers
[1] = NULL
;
658 t
= new0(char**, n
+2);
666 for (a
= arg_join_controllers
; *a
; a
++) {
668 if (strv_overlap(*a
, l
)) {
669 if (strv_extend_strv(&l
, *a
, false) < 0) {
689 t
[n
++] = strv_uniq(l
);
691 strv_free_free(arg_join_controllers
);
692 arg_join_controllers
= t
;
695 if (!isempty(rvalue
))
696 log_syntax(unit
, LOG_ERR
, filename
, line
, 0, "Trailing garbage, ignoring.");
701 static int parse_config_file(void) {
703 const ConfigTableItem items
[] = {
704 { "Manager", "LogLevel", config_parse_level2
, 0, NULL
},
705 { "Manager", "LogTarget", config_parse_target
, 0, NULL
},
706 { "Manager", "LogColor", config_parse_color
, 0, NULL
},
707 { "Manager", "LogLocation", config_parse_location
, 0, NULL
},
708 { "Manager", "DumpCore", config_parse_bool
, 0, &arg_dump_core
},
709 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt
, 0, NULL
},
710 { "Manager", "CrashChangeVT", config_parse_crash_chvt
, 0, NULL
},
711 { "Manager", "CrashShell", config_parse_bool
, 0, &arg_crash_shell
},
712 { "Manager", "CrashReboot", config_parse_bool
, 0, &arg_crash_reboot
},
713 { "Manager", "ShowStatus", config_parse_show_status
, 0, &arg_show_status
},
714 { "Manager", "CPUAffinity", config_parse_cpu_affinity2
, 0, NULL
},
715 { "Manager", "JoinControllers", config_parse_join_controllers
, 0, &arg_join_controllers
},
716 { "Manager", "RuntimeWatchdogSec", config_parse_sec
, 0, &arg_runtime_watchdog
},
717 { "Manager", "ShutdownWatchdogSec", config_parse_sec
, 0, &arg_shutdown_watchdog
},
718 { "Manager", "CapabilityBoundingSet", config_parse_capability_set
, 0, &arg_capability_bounding_set
},
720 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs
, 0, &arg_syscall_archs
},
722 { "Manager", "TimerSlackNSec", config_parse_nsec
, 0, &arg_timer_slack_nsec
},
723 { "Manager", "DefaultTimerAccuracySec", config_parse_sec
, 0, &arg_default_timer_accuracy_usec
},
724 { "Manager", "DefaultStandardOutput", config_parse_output
, 0, &arg_default_std_output
},
725 { "Manager", "DefaultStandardError", config_parse_output
, 0, &arg_default_std_error
},
726 { "Manager", "DefaultTimeoutStartSec", config_parse_sec
, 0, &arg_default_timeout_start_usec
},
727 { "Manager", "DefaultTimeoutStopSec", config_parse_sec
, 0, &arg_default_timeout_stop_usec
},
728 { "Manager", "DefaultRestartSec", config_parse_sec
, 0, &arg_default_restart_usec
},
729 { "Manager", "DefaultStartLimitInterval", config_parse_sec
, 0, &arg_default_start_limit_interval
}, /* obsolete alias */
730 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec
, 0, &arg_default_start_limit_interval
},
731 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned
, 0, &arg_default_start_limit_burst
},
732 { "Manager", "DefaultEnvironment", config_parse_environ
, 0, &arg_default_environment
},
733 { "Manager", "DefaultLimitCPU", config_parse_limit
, RLIMIT_CPU
, arg_default_rlimit
},
734 { "Manager", "DefaultLimitFSIZE", config_parse_limit
, RLIMIT_FSIZE
, arg_default_rlimit
},
735 { "Manager", "DefaultLimitDATA", config_parse_limit
, RLIMIT_DATA
, arg_default_rlimit
},
736 { "Manager", "DefaultLimitSTACK", config_parse_limit
, RLIMIT_STACK
, arg_default_rlimit
},
737 { "Manager", "DefaultLimitCORE", config_parse_limit
, RLIMIT_CORE
, arg_default_rlimit
},
738 { "Manager", "DefaultLimitRSS", config_parse_limit
, RLIMIT_RSS
, arg_default_rlimit
},
739 { "Manager", "DefaultLimitNOFILE", config_parse_limit
, RLIMIT_NOFILE
, arg_default_rlimit
},
740 { "Manager", "DefaultLimitAS", config_parse_limit
, RLIMIT_AS
, arg_default_rlimit
},
741 { "Manager", "DefaultLimitNPROC", config_parse_limit
, RLIMIT_NPROC
, arg_default_rlimit
},
742 { "Manager", "DefaultLimitMEMLOCK", config_parse_limit
, RLIMIT_MEMLOCK
, arg_default_rlimit
},
743 { "Manager", "DefaultLimitLOCKS", config_parse_limit
, RLIMIT_LOCKS
, arg_default_rlimit
},
744 { "Manager", "DefaultLimitSIGPENDING", config_parse_limit
, RLIMIT_SIGPENDING
, arg_default_rlimit
},
745 { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit
, RLIMIT_MSGQUEUE
, arg_default_rlimit
},
746 { "Manager", "DefaultLimitNICE", config_parse_limit
, RLIMIT_NICE
, arg_default_rlimit
},
747 { "Manager", "DefaultLimitRTPRIO", config_parse_limit
, RLIMIT_RTPRIO
, arg_default_rlimit
},
748 { "Manager", "DefaultLimitRTTIME", config_parse_limit
, RLIMIT_RTTIME
, arg_default_rlimit
},
749 { "Manager", "DefaultCPUAccounting", config_parse_bool
, 0, &arg_default_cpu_accounting
},
750 { "Manager", "DefaultIOAccounting", config_parse_bool
, 0, &arg_default_io_accounting
},
751 { "Manager", "DefaultBlockIOAccounting", config_parse_bool
, 0, &arg_default_blockio_accounting
},
752 { "Manager", "DefaultMemoryAccounting", config_parse_bool
, 0, &arg_default_memory_accounting
},
753 { "Manager", "DefaultTasksAccounting", config_parse_bool
, 0, &arg_default_tasks_accounting
},
754 { "Manager", "DefaultTasksMax", config_parse_tasks_max
, 0, &arg_default_tasks_max
},
755 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action
, 0, &arg_cad_burst_action
},
759 const char *fn
, *conf_dirs_nulstr
;
762 PKGSYSCONFDIR
"/system.conf" :
763 PKGSYSCONFDIR
"/user.conf";
765 conf_dirs_nulstr
= arg_system
?
766 CONF_PATHS_NULSTR("systemd/system.conf.d") :
767 CONF_PATHS_NULSTR("systemd/user.conf.d");
769 config_parse_many_nulstr(fn
, conf_dirs_nulstr
, "Manager\0", config_item_table_lookup
, items
, false, NULL
);
771 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
772 * like everywhere else. */
773 if (arg_default_timeout_start_usec
<= 0)
774 arg_default_timeout_start_usec
= USEC_INFINITY
;
775 if (arg_default_timeout_stop_usec
<= 0)
776 arg_default_timeout_stop_usec
= USEC_INFINITY
;
781 static void manager_set_defaults(Manager
*m
) {
785 m
->default_timer_accuracy_usec
= arg_default_timer_accuracy_usec
;
786 m
->default_std_output
= arg_default_std_output
;
787 m
->default_std_error
= arg_default_std_error
;
788 m
->default_timeout_start_usec
= arg_default_timeout_start_usec
;
789 m
->default_timeout_stop_usec
= arg_default_timeout_stop_usec
;
790 m
->default_restart_usec
= arg_default_restart_usec
;
791 m
->default_start_limit_interval
= arg_default_start_limit_interval
;
792 m
->default_start_limit_burst
= arg_default_start_limit_burst
;
793 m
->default_cpu_accounting
= arg_default_cpu_accounting
;
794 m
->default_io_accounting
= arg_default_io_accounting
;
795 m
->default_blockio_accounting
= arg_default_blockio_accounting
;
796 m
->default_memory_accounting
= arg_default_memory_accounting
;
797 m
->default_tasks_accounting
= arg_default_tasks_accounting
;
798 m
->default_tasks_max
= arg_default_tasks_max
;
800 manager_set_default_rlimits(m
, arg_default_rlimit
);
801 manager_environment_add(m
, NULL
, arg_default_environment
);
804 static int parse_argv(int argc
, char *argv
[]) {
807 ARG_LOG_LEVEL
= 0x100,
817 ARG_DUMP_CONFIGURATION_ITEMS
,
826 ARG_DEFAULT_STD_OUTPUT
,
827 ARG_DEFAULT_STD_ERROR
,
831 static const struct option options
[] = {
832 { "log-level", required_argument
, NULL
, ARG_LOG_LEVEL
},
833 { "log-target", required_argument
, NULL
, ARG_LOG_TARGET
},
834 { "log-color", optional_argument
, NULL
, ARG_LOG_COLOR
},
835 { "log-location", optional_argument
, NULL
, ARG_LOG_LOCATION
},
836 { "unit", required_argument
, NULL
, ARG_UNIT
},
837 { "system", no_argument
, NULL
, ARG_SYSTEM
},
838 { "user", no_argument
, NULL
, ARG_USER
},
839 { "test", no_argument
, NULL
, ARG_TEST
},
840 { "no-pager", no_argument
, NULL
, ARG_NO_PAGER
},
841 { "help", no_argument
, NULL
, 'h' },
842 { "version", no_argument
, NULL
, ARG_VERSION
},
843 { "dump-configuration-items", no_argument
, NULL
, ARG_DUMP_CONFIGURATION_ITEMS
},
844 { "dump-core", optional_argument
, NULL
, ARG_DUMP_CORE
},
845 { "crash-chvt", required_argument
, NULL
, ARG_CRASH_CHVT
},
846 { "crash-shell", optional_argument
, NULL
, ARG_CRASH_SHELL
},
847 { "crash-reboot", optional_argument
, NULL
, ARG_CRASH_REBOOT
},
848 { "confirm-spawn", optional_argument
, NULL
, ARG_CONFIRM_SPAWN
},
849 { "show-status", optional_argument
, NULL
, ARG_SHOW_STATUS
},
850 { "deserialize", required_argument
, NULL
, ARG_DESERIALIZE
},
851 { "switched-root", no_argument
, NULL
, ARG_SWITCHED_ROOT
},
852 { "default-standard-output", required_argument
, NULL
, ARG_DEFAULT_STD_OUTPUT
, },
853 { "default-standard-error", required_argument
, NULL
, ARG_DEFAULT_STD_ERROR
, },
854 { "machine-id", required_argument
, NULL
, ARG_MACHINE_ID
},
866 while ((c
= getopt_long(argc
, argv
, "hDbsz:", options
, NULL
)) >= 0)
871 r
= log_set_max_level_from_string(optarg
);
873 log_error("Failed to parse log level %s.", optarg
);
880 r
= log_set_target_from_string(optarg
);
882 log_error("Failed to parse log target %s.", optarg
);
891 r
= log_show_color_from_string(optarg
);
893 log_error("Failed to parse log color setting %s.", optarg
);
897 log_show_color(true);
901 case ARG_LOG_LOCATION
:
903 r
= log_show_location_from_string(optarg
);
905 log_error("Failed to parse log location setting %s.", optarg
);
909 log_show_location(true);
913 case ARG_DEFAULT_STD_OUTPUT
:
914 r
= exec_output_from_string(optarg
);
916 log_error("Failed to parse default standard output setting %s.", optarg
);
919 arg_default_std_output
= r
;
922 case ARG_DEFAULT_STD_ERROR
:
923 r
= exec_output_from_string(optarg
);
925 log_error("Failed to parse default standard error output setting %s.", optarg
);
928 arg_default_std_error
= r
;
932 r
= free_and_strdup(&arg_default_unit
, optarg
);
934 return log_error_errno(r
, "Failed to set default unit %s: %m", optarg
);
947 arg_action
= ACTION_TEST
;
955 arg_action
= ACTION_VERSION
;
958 case ARG_DUMP_CONFIGURATION_ITEMS
:
959 arg_action
= ACTION_DUMP_CONFIGURATION_ITEMS
;
964 arg_dump_core
= true;
966 r
= parse_boolean(optarg
);
968 return log_error_errno(r
, "Failed to parse dump core boolean: %s", optarg
);
974 r
= parse_crash_chvt(optarg
);
976 return log_error_errno(r
, "Failed to parse crash virtual terminal index: %s", optarg
);
979 case ARG_CRASH_SHELL
:
981 arg_crash_shell
= true;
983 r
= parse_boolean(optarg
);
985 return log_error_errno(r
, "Failed to parse crash shell boolean: %s", optarg
);
990 case ARG_CRASH_REBOOT
:
992 arg_crash_reboot
= true;
994 r
= parse_boolean(optarg
);
996 return log_error_errno(r
, "Failed to parse crash shell boolean: %s", optarg
);
997 arg_crash_reboot
= r
;
1001 case ARG_CONFIRM_SPAWN
:
1002 arg_confirm_spawn
= mfree(arg_confirm_spawn
);
1004 r
= parse_confirm_spawn(optarg
, &arg_confirm_spawn
);
1006 return log_error_errno(r
, "Failed to parse confirm spawn option: %m");
1009 case ARG_SHOW_STATUS
:
1011 r
= parse_show_status(optarg
, &arg_show_status
);
1013 log_error("Failed to parse show status boolean %s.", optarg
);
1017 arg_show_status
= SHOW_STATUS_YES
;
1020 case ARG_DESERIALIZE
: {
1024 r
= safe_atoi(optarg
, &fd
);
1025 if (r
< 0 || fd
< 0) {
1026 log_error("Failed to parse deserialize option %s.", optarg
);
1030 (void) fd_cloexec(fd
, true);
1032 f
= fdopen(fd
, "r");
1034 return log_error_errno(errno
, "Failed to open serialization fd: %m");
1036 safe_fclose(arg_serialization
);
1037 arg_serialization
= f
;
1042 case ARG_SWITCHED_ROOT
:
1043 arg_switched_root
= true;
1046 case ARG_MACHINE_ID
:
1047 r
= set_machine_id(optarg
);
1049 return log_error_errno(r
, "MachineID '%s' is not valid.", optarg
);
1053 arg_action
= ACTION_HELP
;
1057 log_set_max_level(LOG_DEBUG
);
1063 /* Just to eat away the sysvinit kernel
1064 * cmdline args without getopt() error
1065 * messages that we'll parse in
1066 * parse_proc_cmdline_word() or ignore. */
1075 assert_not_reached("Unhandled option code.");
1078 if (optind
< argc
&& getpid() != 1) {
1079 /* Hmm, when we aren't run as init system
1080 * let's complain about excess arguments */
1082 log_error("Excess arguments.");
1089 static int help(void) {
1091 printf("%s [OPTIONS...]\n\n"
1092 "Starts up and maintains the system or user services.\n\n"
1093 " -h --help Show this help\n"
1094 " --test Determine startup sequence, dump it and exit\n"
1095 " --no-pager Do not pipe output into a pager\n"
1096 " --dump-configuration-items Dump understood unit configuration items\n"
1097 " --unit=UNIT Set default unit\n"
1098 " --system Run a system instance, even if PID != 1\n"
1099 " --user Run a user instance\n"
1100 " --dump-core[=BOOL] Dump core on crash\n"
1101 " --crash-vt=NR Change to specified VT on crash\n"
1102 " --crash-reboot[=BOOL] Reboot on crash\n"
1103 " --crash-shell[=BOOL] Run shell on crash\n"
1104 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1105 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1106 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1107 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1108 " --log-color[=BOOL] Highlight important log messages\n"
1109 " --log-location[=BOOL] Include code location in log messages\n"
1110 " --default-standard-output= Set default standard output for services\n"
1111 " --default-standard-error= Set default standard error output for services\n",
1112 program_invocation_short_name
);
1117 static int prepare_reexecute(Manager
*m
, FILE **_f
, FDSet
**_fds
, bool switching_root
) {
1118 _cleanup_fdset_free_ FDSet
*fds
= NULL
;
1119 _cleanup_fclose_
FILE *f
= NULL
;
1126 r
= manager_open_serialization(m
, &f
);
1128 return log_error_errno(r
, "Failed to create serialization file: %m");
1130 /* Make sure nothing is really destructed when we shut down */
1132 bus_manager_send_reloading(m
, true);
1138 r
= manager_serialize(m
, f
, fds
, switching_root
);
1140 return log_error_errno(r
, "Failed to serialize state: %m");
1142 if (fseeko(f
, 0, SEEK_SET
) == (off_t
) -1)
1143 return log_error_errno(errno
, "Failed to rewind serialization fd: %m");
1145 r
= fd_cloexec(fileno(f
), false);
1147 return log_error_errno(r
, "Failed to disable O_CLOEXEC for serialization: %m");
1149 r
= fdset_cloexec(fds
, false);
1151 return log_error_errno(r
, "Failed to disable O_CLOEXEC for serialization fds: %m");
1162 static int bump_rlimit_nofile(struct rlimit
*saved_rlimit
) {
1166 assert(saved_rlimit
);
1168 /* Save the original RLIMIT_NOFILE so that we can reset it
1169 * later when transitioning from the initrd to the main
1170 * systemd or suchlike. */
1171 if (getrlimit(RLIMIT_NOFILE
, saved_rlimit
) < 0)
1172 return log_warning_errno(errno
, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1174 /* Make sure forked processes get the default kernel setting */
1175 if (!arg_default_rlimit
[RLIMIT_NOFILE
]) {
1178 rl
= newdup(struct rlimit
, saved_rlimit
, 1);
1182 arg_default_rlimit
[RLIMIT_NOFILE
] = rl
;
1185 /* Bump up the resource limit for ourselves substantially */
1186 nl
.rlim_cur
= nl
.rlim_max
= 64*1024;
1187 r
= setrlimit_closest(RLIMIT_NOFILE
, &nl
);
1189 return log_warning_errno(r
, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1194 static void test_usr(void) {
1196 /* Check that /usr is not a separate fs */
1198 if (dir_is_empty("/usr") <= 0)
1201 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1202 "Some things will probably break (sometimes even silently) in mysterious ways. "
1203 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1206 static int initialize_join_controllers(void) {
1207 /* By default, mount "cpu" + "cpuacct" together, and "net_cls"
1208 * + "net_prio". We'd like to add "cpuset" to the mix, but
1209 * "cpuset" doesn't really work for groups with no initialized
1212 arg_join_controllers
= new(char**, 3);
1213 if (!arg_join_controllers
)
1216 arg_join_controllers
[0] = strv_new("cpu", "cpuacct", NULL
);
1217 if (!arg_join_controllers
[0])
1220 arg_join_controllers
[1] = strv_new("net_cls", "net_prio", NULL
);
1221 if (!arg_join_controllers
[1])
1224 arg_join_controllers
[2] = NULL
;
1228 arg_join_controllers
= strv_free_free(arg_join_controllers
);
1232 static int enforce_syscall_archs(Set
*archs
) {
1236 if (!is_seccomp_available())
1239 r
= seccomp_restrict_archs(arg_syscall_archs
);
1241 return log_error_errno(r
, "Failed to enforce system call architecture restrication: %m");
1246 static int status_welcome(void) {
1247 _cleanup_free_
char *pretty_name
= NULL
, *ansi_color
= NULL
;
1250 r
= parse_env_file("/etc/os-release", NEWLINE
,
1251 "PRETTY_NAME", &pretty_name
,
1252 "ANSI_COLOR", &ansi_color
,
1255 r
= parse_env_file("/usr/lib/os-release", NEWLINE
,
1256 "PRETTY_NAME", &pretty_name
,
1257 "ANSI_COLOR", &ansi_color
,
1260 if (r
< 0 && r
!= -ENOENT
)
1261 log_warning_errno(r
, "Failed to read os-release file: %m");
1263 if (log_get_show_color())
1264 return status_printf(NULL
, false, false,
1265 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1266 isempty(ansi_color
) ? "1" : ansi_color
,
1267 isempty(pretty_name
) ? "Linux" : pretty_name
);
1269 return status_printf(NULL
, false, false,
1270 "\nWelcome to %s!\n",
1271 isempty(pretty_name
) ? "Linux" : pretty_name
);
1274 static int write_container_id(void) {
1278 c
= getenv("container");
1282 RUN_WITH_UMASK(0022)
1283 r
= write_string_file("/run/systemd/container", c
, WRITE_STRING_FILE_CREATE
);
1285 return log_warning_errno(r
, "Failed to write /run/systemd/container, ignoring: %m");
1290 static int bump_unix_max_dgram_qlen(void) {
1291 _cleanup_free_
char *qlen
= NULL
;
1295 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel
1296 * default of 16 is simply too low. We set the value really
1297 * really early during boot, so that it is actually applied to
1298 * all our sockets, including the $NOTIFY_SOCKET one. */
1300 r
= read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen
);
1302 return log_warning_errno(r
, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1304 r
= safe_atolu(qlen
, &v
);
1306 return log_warning_errno(r
, "Failed to parse AF_UNIX datagram queue length, ignoring: %m");
1308 if (v
>= DEFAULT_UNIX_MAX_DGRAM_QLEN
)
1312 if (asprintf(&qlen
, "%lu\n", DEFAULT_UNIX_MAX_DGRAM_QLEN
) < 0)
1315 r
= write_string_file("/proc/sys/net/unix/max_dgram_qlen", qlen
, 0);
1317 return log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
,
1318 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1323 static int fixup_environment(void) {
1324 _cleanup_free_
char *term
= NULL
;
1327 /* We expect the environment to be set correctly
1328 * if run inside a container. */
1329 if (detect_container() > 0)
1332 /* When started as PID1, the kernel uses /dev/console
1333 * for our stdios and uses TERM=linux whatever the
1334 * backend device used by the console. We try to make
1335 * a better guess here since some consoles might not
1336 * have support for color mode for example.
1338 * However if TERM was configured through the kernel
1339 * command line then leave it alone. */
1341 r
= proc_cmdline_get_key("TERM", 0, &term
);
1345 term
= strdup(default_term_for_tty("/dev/console"));
1350 if (setenv("TERM", term
, 1) < 0)
1356 int main(int argc
, char *argv
[]) {
1358 int r
, retval
= EXIT_FAILURE
;
1359 usec_t before_startup
, after_startup
;
1360 char timespan
[FORMAT_TIMESPAN_MAX
];
1362 bool reexecute
= false;
1363 const char *shutdown_verb
= NULL
;
1364 dual_timestamp initrd_timestamp
= DUAL_TIMESTAMP_NULL
;
1365 dual_timestamp userspace_timestamp
= DUAL_TIMESTAMP_NULL
;
1366 dual_timestamp kernel_timestamp
= DUAL_TIMESTAMP_NULL
;
1367 dual_timestamp security_start_timestamp
= DUAL_TIMESTAMP_NULL
;
1368 dual_timestamp security_finish_timestamp
= DUAL_TIMESTAMP_NULL
;
1369 static char systemd
[] = "systemd";
1370 bool skip_setup
= false;
1372 bool loaded_policy
= false;
1373 bool arm_reboot_watchdog
= false;
1374 bool queue_default_job
= false;
1375 bool empty_etc
= false;
1376 char *switch_root_dir
= NULL
, *switch_root_init
= NULL
;
1377 struct rlimit saved_rlimit_nofile
= RLIMIT_MAKE_CONST(0);
1378 const char *error_message
= NULL
;
1380 #ifdef HAVE_SYSV_COMPAT
1381 if (getpid() != 1 && strstr(program_invocation_short_name
, "init")) {
1382 /* This is compatibility support for SysV, where
1383 * calling init as a user is identical to telinit. */
1385 execv(SYSTEMCTL_BINARY_PATH
, argv
);
1386 log_error_errno(errno
, "Failed to exec " SYSTEMCTL_BINARY_PATH
": %m");
1391 dual_timestamp_from_monotonic(&kernel_timestamp
, 0);
1392 dual_timestamp_get(&userspace_timestamp
);
1394 /* Determine if this is a reexecution or normal bootup. We do
1395 * the full command line parsing much later, so let's just
1396 * have a quick peek here. */
1397 if (strv_find(argv
+1, "--deserialize"))
1400 /* If we have switched root, do all the special setup
1402 if (strv_find(argv
+1, "--switched-root"))
1405 /* If we get started via the /sbin/init symlink then we are
1406 called 'init'. After a subsequent reexecution we are then
1407 called 'systemd'. That is confusing, hence let's call us
1408 systemd right-away. */
1409 program_invocation_short_name
= systemd
;
1410 (void) prctl(PR_SET_NAME
, systemd
);
1415 log_set_upgrade_syslog_to_journal(true);
1417 /* Disable the umask logic */
1421 if (getpid() == 1 && detect_container() <= 0) {
1423 /* Running outside of a container as PID 1 */
1425 log_set_target(LOG_TARGET_KMSG
);
1429 initrd_timestamp
= userspace_timestamp
;
1432 r
= mount_setup_early();
1434 error_message
= "Failed to mount early API filesystems";
1438 dual_timestamp_get(&security_start_timestamp
);
1439 if (mac_selinux_setup(&loaded_policy
) < 0) {
1440 error_message
= "Failed to load SELinux policy";
1442 } else if (mac_smack_setup(&loaded_policy
) < 0) {
1443 error_message
= "Failed to load SMACK policy";
1445 } else if (ima_setup() < 0) {
1446 error_message
= "Failed to load IMA policy";
1449 dual_timestamp_get(&security_finish_timestamp
);
1452 if (mac_selinux_init() < 0) {
1453 error_message
= "Failed to initialize SELinux policy";
1458 if (clock_is_localtime(NULL
) > 0) {
1462 * The very first call of settimeofday() also does a time warp in the kernel.
1464 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on
1465 * external tools to take care of maintaining the RTC and do all adjustments.
1466 * This matches the behavior of Windows, which leaves the RTC alone if the
1467 * registry tells that the RTC runs in UTC.
1469 r
= clock_set_timezone(&min
);
1471 log_error_errno(r
, "Failed to apply local time delta, ignoring: %m");
1473 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min
);
1474 } else if (!in_initrd()) {
1476 * Do a dummy very first call to seal the kernel's time warp magic.
1478 * Do not call this from inside the initrd. The initrd might not
1479 * carry /etc/adjtime with LOCAL, but the real system could be set up
1480 * that way. In such case, we need to delay the time-warp or the sealing
1481 * until we reach the real system.
1483 * Do no set the kernel's timezone. The concept of local time cannot
1484 * be supported reliably, the time will jump or be incorrect at every daylight
1485 * saving time change. All kernel local time concepts will be treated
1488 (void) clock_reset_timewarp();
1491 r
= clock_apply_epoch();
1493 log_error_errno(r
, "Current system time is before build time, but cannot correct: %m");
1495 log_info("System time before build time, advancing clock.");
1498 /* Set the default for later on, but don't actually
1499 * open the logs like this for now. Note that if we
1500 * are transitioning from the initrd there might still
1501 * be journal fd open, and we shouldn't attempt
1502 * opening that before we parsed /proc/cmdline which
1503 * might redirect output elsewhere. */
1504 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG
);
1506 } else if (getpid() == 1) {
1507 /* Running inside a container, as PID 1 */
1509 log_set_target(LOG_TARGET_CONSOLE
);
1510 log_close_console(); /* force reopen of /dev/console */
1513 /* For later on, see above... */
1514 log_set_target(LOG_TARGET_JOURNAL
);
1516 /* clear the kernel timestamp,
1517 * because we are in a container */
1518 kernel_timestamp
= DUAL_TIMESTAMP_NULL
;
1520 /* Running as user instance */
1522 log_set_target(LOG_TARGET_AUTO
);
1525 /* clear the kernel timestamp,
1526 * because we are not PID 1 */
1527 kernel_timestamp
= DUAL_TIMESTAMP_NULL
;
1530 if (getpid() == 1) {
1531 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1532 * will process core dumps for system services by default. */
1533 if (setrlimit(RLIMIT_CORE
, &RLIMIT_MAKE_CONST(RLIM_INFINITY
)) < 0)
1534 log_warning_errno(errno
, "Failed to set RLIMIT_CORE: %m");
1536 /* But at the same time, turn off the core_pattern logic by default, so that no coredumps are stored
1537 * until the systemd-coredump tool is enabled via sysctl. */
1539 (void) write_string_file("/proc/sys/kernel/core_pattern", "|/bin/false", 0);
1543 if (fixup_environment() < 0) {
1544 error_message
= "Failed to fix up PID1 environment";
1548 /* Try to figure out if we can use colors with the console. No
1549 * need to do that for user instances since they never log
1550 * into the console. */
1551 log_show_color(colors_enabled());
1552 r
= make_null_stdio();
1554 log_warning_errno(r
, "Failed to redirect standard streams to /dev/null: %m");
1557 r
= initialize_join_controllers();
1559 error_message
= "Failed to initialize cgroup controllers";
1563 /* Mount /proc, /sys and friends, so that /proc/cmdline and
1564 * /proc/$PID/fd is available. */
1565 if (getpid() == 1) {
1567 /* Load the kernel modules early, so that we kdbus.ko is loaded before kdbusfs shall be mounted */
1571 r
= mount_setup(loaded_policy
);
1573 error_message
= "Failed to mount API filesystems";
1578 /* Reset all signal handlers. */
1579 (void) reset_all_signal_handlers();
1580 (void) ignore_signals(SIGNALS_IGNORE
, -1);
1582 arg_default_tasks_max
= system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE
, 100U);
1584 if (parse_config_file() < 0) {
1585 error_message
= "Failed to parse config file";
1590 r
= proc_cmdline_parse(parse_proc_cmdline_item
, NULL
, 0);
1592 log_warning_errno(r
, "Failed to parse kernel command line, ignoring: %m");
1595 /* Note that this also parses bits from the kernel command
1596 * line, including "debug". */
1597 log_parse_environment();
1599 if (parse_argv(argc
, argv
) < 0) {
1600 error_message
= "Failed to parse commandline arguments";
1604 /* Initialize default unit */
1605 if (!arg_default_unit
) {
1606 arg_default_unit
= strdup(SPECIAL_DEFAULT_TARGET
);
1607 if (!arg_default_unit
) {
1609 error_message
= "Failed to set default unit";
1614 if (arg_action
== ACTION_TEST
&&
1616 log_error("Don't run test mode as root.");
1621 arg_action
== ACTION_RUN
&&
1623 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
1628 arg_action
== ACTION_RUN
&&
1629 running_in_chroot() > 0) {
1630 log_error("Cannot be run in a chroot() environment.");
1634 if (arg_action
== ACTION_TEST
|| arg_action
== ACTION_HELP
) {
1635 pager_open(arg_no_pager
, false);
1639 if (arg_action
== ACTION_HELP
) {
1642 } else if (arg_action
== ACTION_VERSION
) {
1645 } else if (arg_action
== ACTION_DUMP_CONFIGURATION_ITEMS
) {
1646 pager_open(arg_no_pager
, false);
1647 unit_dump_config_items(stdout
);
1648 retval
= EXIT_SUCCESS
;
1653 !getenv("XDG_RUNTIME_DIR")) {
1654 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
1658 assert_se(arg_action
== ACTION_RUN
|| arg_action
== ACTION_TEST
);
1660 /* Close logging fds, in order not to confuse fdset below */
1663 /* Remember open file descriptors for later deserialization */
1664 r
= fdset_new_fill(&fds
);
1666 log_emergency_errno(r
, "Failed to allocate fd set: %m");
1667 error_message
= "Failed to allocate fd set";
1670 fdset_cloexec(fds
, true);
1672 if (arg_serialization
)
1673 assert_se(fdset_remove(fds
, fileno(arg_serialization
)) >= 0);
1676 /* Become a session leader if we aren't one yet. */
1679 /* Move out of the way, so that we won't block unmounts */
1680 assert_se(chdir("/") == 0);
1682 /* Reset the console, but only if this is really init and we
1683 * are freshly booted */
1684 if (arg_system
&& arg_action
== ACTION_RUN
) {
1686 /* If we are init, we connect stdin/stdout/stderr to
1687 * /dev/null and make sure we don't have a controlling
1691 if (getpid() == 1 && !skip_setup
)
1695 /* Open the logging devices, if possible and necessary */
1698 if (arg_show_status
== _SHOW_STATUS_UNSET
)
1699 arg_show_status
= SHOW_STATUS_YES
;
1701 /* Make sure we leave a core dump without panicing the
1703 if (getpid() == 1) {
1704 install_crash_handler();
1706 r
= mount_cgroup_controllers(arg_join_controllers
);
1714 log_info(PACKAGE_STRING
" running in %ssystem mode. (" SYSTEMD_FEATURES
")",
1715 arg_action
== ACTION_TEST
? "test " : "" );
1717 v
= detect_virtualization();
1719 log_info("Detected virtualization %s.", virtualization_to_string(v
));
1721 write_container_id();
1723 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1726 log_info("Running in initial RAM disk.");
1728 /* Let's check whether /etc is already populated. We
1729 * don't actually really check for that, but use
1730 * /etc/machine-id as flag file. This allows container
1731 * managers and installers to provision a couple of
1732 * files already. If the container manager wants to
1733 * provision the machine ID itself it should pass
1734 * $container_uuid to PID 1. */
1736 empty_etc
= access("/etc/machine-id", F_OK
) < 0;
1738 log_info("Running with unpopulated /etc.");
1740 _cleanup_free_
char *t
;
1742 t
= uid_to_name(getuid());
1743 log_debug(PACKAGE_STRING
" running in %suser mode for user "UID_FMT
"/%s. (" SYSTEMD_FEATURES
")",
1744 arg_action
== ACTION_TEST
? " test" : "", getuid(), t
);
1747 if (arg_system
&& !skip_setup
) {
1748 if (arg_show_status
> 0)
1752 machine_id_setup(NULL
, arg_machine_id
, NULL
);
1754 bump_unix_max_dgram_qlen();
1759 if (arg_system
&& arg_runtime_watchdog
> 0 && arg_runtime_watchdog
!= USEC_INFINITY
)
1760 watchdog_set_timeout(&arg_runtime_watchdog
);
1762 if (arg_timer_slack_nsec
!= NSEC_INFINITY
)
1763 if (prctl(PR_SET_TIMERSLACK
, arg_timer_slack_nsec
) < 0)
1764 log_error_errno(errno
, "Failed to adjust timer slack: %m");
1766 if (!cap_test_all(arg_capability_bounding_set
)) {
1767 r
= capability_bounding_set_drop_usermode(arg_capability_bounding_set
);
1769 log_emergency_errno(r
, "Failed to drop capability bounding set of usermode helpers: %m");
1770 error_message
= "Failed to drop capability bounding set of usermode helpers";
1773 r
= capability_bounding_set_drop(arg_capability_bounding_set
, true);
1775 log_emergency_errno(r
, "Failed to drop capability bounding set: %m");
1776 error_message
= "Failed to drop capability bounding set";
1781 if (arg_syscall_archs
) {
1782 r
= enforce_syscall_archs(arg_syscall_archs
);
1784 error_message
= "Failed to set syscall architectures";
1790 /* Become reaper of our children */
1791 if (prctl(PR_SET_CHILD_SUBREAPER
, 1) < 0)
1792 log_warning_errno(errno
, "Failed to make us a subreaper: %m");
1795 (void) bump_rlimit_nofile(&saved_rlimit_nofile
);
1798 r
= unit_file_preset_all(UNIT_FILE_SYSTEM
, 0, NULL
, UNIT_FILE_PRESET_ENABLE_ONLY
, NULL
, 0);
1800 log_full_errno(r
== -EEXIST
? LOG_NOTICE
: LOG_WARNING
, r
, "Failed to populate /etc with preset unit settings, ignoring: %m");
1802 log_info("Populated /etc with preset unit settings.");
1806 r
= manager_new(arg_system
? UNIT_FILE_SYSTEM
: UNIT_FILE_USER
, arg_action
== ACTION_TEST
, &m
);
1808 log_emergency_errno(r
, "Failed to allocate manager object: %m");
1809 error_message
= "Failed to allocate manager object";
1813 m
->confirm_spawn
= arg_confirm_spawn
;
1814 m
->runtime_watchdog
= arg_runtime_watchdog
;
1815 m
->shutdown_watchdog
= arg_shutdown_watchdog
;
1816 m
->userspace_timestamp
= userspace_timestamp
;
1817 m
->kernel_timestamp
= kernel_timestamp
;
1818 m
->initrd_timestamp
= initrd_timestamp
;
1819 m
->security_start_timestamp
= security_start_timestamp
;
1820 m
->security_finish_timestamp
= security_finish_timestamp
;
1821 m
->cad_burst_action
= arg_cad_burst_action
;
1823 manager_set_defaults(m
);
1824 manager_set_show_status(m
, arg_show_status
);
1825 manager_set_first_boot(m
, empty_etc
);
1827 /* Remember whether we should queue the default job */
1828 queue_default_job
= !arg_serialization
|| arg_switched_root
;
1830 before_startup
= now(CLOCK_MONOTONIC
);
1832 r
= manager_startup(m
, arg_serialization
, fds
);
1834 log_error_errno(r
, "Failed to fully start up daemon: %m");
1836 /* This will close all file descriptors that were opened, but
1837 * not claimed by any unit. */
1838 fds
= fdset_free(fds
);
1840 arg_serialization
= safe_fclose(arg_serialization
);
1842 if (queue_default_job
) {
1843 _cleanup_(sd_bus_error_free
) sd_bus_error error
= SD_BUS_ERROR_NULL
;
1844 Unit
*target
= NULL
;
1845 Job
*default_unit_job
;
1847 log_debug("Activating default unit: %s", arg_default_unit
);
1849 r
= manager_load_unit(m
, arg_default_unit
, NULL
, &error
, &target
);
1851 log_error("Failed to load default target: %s", bus_error_message(&error
, r
));
1852 else if (target
->load_state
== UNIT_ERROR
|| target
->load_state
== UNIT_NOT_FOUND
)
1853 log_error_errno(target
->load_error
, "Failed to load default target: %m");
1854 else if (target
->load_state
== UNIT_MASKED
)
1855 log_error("Default target masked.");
1857 if (!target
|| target
->load_state
!= UNIT_LOADED
) {
1858 log_info("Trying to load rescue target...");
1860 r
= manager_load_unit(m
, SPECIAL_RESCUE_TARGET
, NULL
, &error
, &target
);
1862 log_emergency("Failed to load rescue target: %s", bus_error_message(&error
, r
));
1863 error_message
= "Failed to load rescue target";
1865 } else if (target
->load_state
== UNIT_ERROR
|| target
->load_state
== UNIT_NOT_FOUND
) {
1866 log_emergency_errno(target
->load_error
, "Failed to load rescue target: %m");
1867 error_message
= "Failed to load rescue target";
1869 } else if (target
->load_state
== UNIT_MASKED
) {
1870 log_emergency("Rescue target masked.");
1871 error_message
= "Rescue target masked";
1876 assert(target
->load_state
== UNIT_LOADED
);
1878 if (arg_action
== ACTION_TEST
) {
1879 printf("-> By units:\n");
1880 manager_dump_units(m
, stdout
, "\t");
1883 r
= manager_add_job(m
, JOB_START
, target
, JOB_ISOLATE
, &error
, &default_unit_job
);
1885 log_debug("Default target could not be isolated, starting instead: %s", bus_error_message(&error
, r
));
1887 sd_bus_error_free(&error
);
1889 r
= manager_add_job(m
, JOB_START
, target
, JOB_REPLACE
, &error
, &default_unit_job
);
1891 log_emergency("Failed to start default target: %s", bus_error_message(&error
, r
));
1892 error_message
= "Failed to start default target";
1896 log_emergency("Failed to isolate default target: %s", bus_error_message(&error
, r
));
1897 error_message
= "Failed to isolate default target";
1901 m
->default_unit_job_id
= default_unit_job
->id
;
1903 after_startup
= now(CLOCK_MONOTONIC
);
1904 log_full(arg_action
== ACTION_TEST
? LOG_INFO
: LOG_DEBUG
,
1905 "Loaded units and determined initial transaction in %s.",
1906 format_timespan(timespan
, sizeof(timespan
), after_startup
- before_startup
, 100 * USEC_PER_MSEC
));
1908 if (arg_action
== ACTION_TEST
) {
1909 printf("-> By jobs:\n");
1910 manager_dump_jobs(m
, stdout
, "\t");
1911 retval
= EXIT_SUCCESS
;
1917 r
= manager_loop(m
);
1919 log_emergency_errno(r
, "Failed to run main loop: %m");
1920 error_message
= "Failed to run main loop";
1924 switch (m
->exit_code
) {
1926 case MANAGER_RELOAD
:
1927 log_info("Reloading.");
1929 r
= parse_config_file();
1931 log_error("Failed to parse config file.");
1933 manager_set_defaults(m
);
1935 r
= manager_reload(m
);
1937 log_error_errno(r
, "Failed to reload: %m");
1940 case MANAGER_REEXECUTE
:
1942 if (prepare_reexecute(m
, &arg_serialization
, &fds
, false) < 0) {
1943 error_message
= "Failed to prepare for reexecution";
1948 log_notice("Reexecuting.");
1951 case MANAGER_SWITCH_ROOT
:
1952 /* Steal the switch root parameters */
1953 switch_root_dir
= m
->switch_root
;
1954 switch_root_init
= m
->switch_root_init
;
1955 m
->switch_root
= m
->switch_root_init
= NULL
;
1957 if (!switch_root_init
)
1958 if (prepare_reexecute(m
, &arg_serialization
, &fds
, true) < 0) {
1959 error_message
= "Failed to prepare for reexecution";
1964 log_notice("Switching root.");
1968 retval
= m
->return_value
;
1970 if (MANAGER_IS_USER(m
)) {
1976 case MANAGER_REBOOT
:
1977 case MANAGER_POWEROFF
:
1979 case MANAGER_KEXEC
: {
1980 static const char * const table
[_MANAGER_EXIT_CODE_MAX
] = {
1981 [MANAGER_EXIT
] = "exit",
1982 [MANAGER_REBOOT
] = "reboot",
1983 [MANAGER_POWEROFF
] = "poweroff",
1984 [MANAGER_HALT
] = "halt",
1985 [MANAGER_KEXEC
] = "kexec"
1988 assert_se(shutdown_verb
= table
[m
->exit_code
]);
1989 arm_reboot_watchdog
= m
->exit_code
== MANAGER_REBOOT
;
1991 log_notice("Shutting down.");
1996 assert_not_reached("Unknown exit code.");
2004 arg_shutdown_watchdog
= m
->shutdown_watchdog
;
2006 m
= manager_free(m
);
2008 for (j
= 0; j
< ELEMENTSOF(arg_default_rlimit
); j
++)
2009 arg_default_rlimit
[j
] = mfree(arg_default_rlimit
[j
]);
2011 arg_default_unit
= mfree(arg_default_unit
);
2012 arg_confirm_spawn
= mfree(arg_confirm_spawn
);
2013 arg_join_controllers
= strv_free_free(arg_join_controllers
);
2014 arg_default_environment
= strv_free(arg_default_environment
);
2015 arg_syscall_archs
= set_free(arg_syscall_archs
);
2017 mac_selinux_finish();
2021 unsigned i
, args_size
;
2023 /* Close and disarm the watchdog, so that the new
2024 * instance can reinitialize it, but doesn't get
2025 * rebooted while we do that */
2026 watchdog_close(true);
2028 /* Reset the RLIMIT_NOFILE to the kernel default, so
2029 * that the new systemd can pass the kernel default to
2030 * its child processes */
2031 if (saved_rlimit_nofile
.rlim_cur
> 0)
2032 (void) setrlimit(RLIMIT_NOFILE
, &saved_rlimit_nofile
);
2034 if (switch_root_dir
) {
2035 /* Kill all remaining processes from the
2036 * initrd, but don't wait for them, so that we
2037 * can handle the SIGCHLD for them after
2039 broadcast_signal(SIGTERM
, false, true);
2041 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
2042 r
= switch_root(switch_root_dir
, "/mnt", true, MS_MOVE
);
2044 log_error_errno(r
, "Failed to switch root, trying to continue: %m");
2047 args_size
= MAX(6, argc
+1);
2048 args
= newa(const char*, args_size
);
2050 if (!switch_root_init
) {
2051 char sfd
[DECIMAL_STR_MAX(int) + 1];
2053 /* First try to spawn ourselves with the right
2054 * path, and with full serialization. We do
2055 * this only if the user didn't specify an
2056 * explicit init to spawn. */
2058 assert(arg_serialization
);
2061 xsprintf(sfd
, "%i", fileno(arg_serialization
));
2064 args
[i
++] = SYSTEMD_BINARY_PATH
;
2065 if (switch_root_dir
)
2066 args
[i
++] = "--switched-root";
2067 args
[i
++] = arg_system
? "--system" : "--user";
2068 args
[i
++] = "--deserialize";
2072 assert(i
<= args_size
);
2075 * We want valgrind to print its memory usage summary before reexecution.
2076 * Valgrind won't do this is on its own on exec(), but it will do it on exit().
2077 * Hence, to ensure we get a summary here, fork() off a child, let it exit() cleanly,
2078 * so that it prints the summary, and wait() for it in the parent, before proceeding into the exec().
2080 valgrind_summary_hack();
2082 (void) execv(args
[0], (char* const*) args
);
2085 /* Try the fallback, if there is any, without any
2086 * serialization. We pass the original argv[] and
2087 * envp[]. (Well, modulo the ordering changes due to
2088 * getopt() in argv[], and some cleanups in envp[],
2089 * but let's hope that doesn't matter.) */
2091 arg_serialization
= safe_fclose(arg_serialization
);
2092 fds
= fdset_free(fds
);
2094 /* Reopen the console */
2095 (void) make_console_stdio();
2097 for (j
= 1, i
= 1; j
< (unsigned) argc
; j
++)
2098 args
[i
++] = argv
[j
];
2100 assert(i
<= args_size
);
2102 /* Reenable any blocked signals, especially important
2103 * if we switch from initial ramdisk to init=... */
2104 (void) reset_all_signal_handlers();
2105 (void) reset_signal_mask();
2107 if (switch_root_init
) {
2108 args
[0] = switch_root_init
;
2109 (void) execv(args
[0], (char* const*) args
);
2110 log_warning_errno(errno
, "Failed to execute configured init, trying fallback: %m");
2113 args
[0] = "/sbin/init";
2114 (void) execv(args
[0], (char* const*) args
);
2116 if (errno
== ENOENT
) {
2117 log_warning("No /sbin/init, trying fallback");
2119 args
[0] = "/bin/sh";
2121 (void) execv(args
[0], (char* const*) args
);
2122 log_error_errno(errno
, "Failed to execute /bin/sh, giving up: %m");
2124 log_warning_errno(errno
, "Failed to execute /sbin/init, giving up: %m");
2127 arg_serialization
= safe_fclose(arg_serialization
);
2128 fds
= fdset_free(fds
);
2130 #ifdef HAVE_VALGRIND_VALGRIND_H
2131 /* If we are PID 1 and running under valgrind, then let's exit
2132 * here explicitly. valgrind will only generate nice output on
2133 * exit(), not on exec(), hence let's do the former not the
2135 if (getpid() == 1 && RUNNING_ON_VALGRIND
)
2139 if (shutdown_verb
) {
2140 char log_level
[DECIMAL_STR_MAX(int) + 1];
2141 char exit_code
[DECIMAL_STR_MAX(uint8_t) + 1];
2142 const char* command_line
[11] = {
2143 SYSTEMD_SHUTDOWN_BINARY_PATH
,
2145 "--log-level", log_level
,
2149 _cleanup_strv_free_
char **env_block
= NULL
;
2151 assert(command_line
[pos
] == NULL
);
2152 env_block
= strv_copy(environ
);
2154 xsprintf(log_level
, "%d", log_get_max_level());
2156 switch (log_get_target()) {
2158 case LOG_TARGET_KMSG
:
2159 case LOG_TARGET_JOURNAL_OR_KMSG
:
2160 case LOG_TARGET_SYSLOG_OR_KMSG
:
2161 command_line
[pos
++] = "kmsg";
2164 case LOG_TARGET_NULL
:
2165 command_line
[pos
++] = "null";
2168 case LOG_TARGET_CONSOLE
:
2170 command_line
[pos
++] = "console";
2174 if (log_get_show_color())
2175 command_line
[pos
++] = "--log-color";
2177 if (log_get_show_location())
2178 command_line
[pos
++] = "--log-location";
2180 if (streq(shutdown_verb
, "exit")) {
2181 command_line
[pos
++] = "--exit-code";
2182 command_line
[pos
++] = exit_code
;
2183 xsprintf(exit_code
, "%d", retval
);
2186 assert(pos
< ELEMENTSOF(command_line
));
2188 if (arm_reboot_watchdog
&& arg_shutdown_watchdog
> 0 && arg_shutdown_watchdog
!= USEC_INFINITY
) {
2191 /* If we reboot let's set the shutdown
2192 * watchdog and tell the shutdown binary to
2193 * repeatedly ping it */
2194 r
= watchdog_set_timeout(&arg_shutdown_watchdog
);
2195 watchdog_close(r
< 0);
2197 /* Tell the binary how often to ping, ignore failure */
2198 if (asprintf(&e
, "WATCHDOG_USEC="USEC_FMT
, arg_shutdown_watchdog
) > 0)
2199 (void) strv_push(&env_block
, e
);
2201 watchdog_close(true);
2203 /* Avoid the creation of new processes forked by the
2204 * kernel; at this point, we will not listen to the
2206 if (detect_container() <= 0)
2207 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER
);
2209 execve(SYSTEMD_SHUTDOWN_BINARY_PATH
, (char **) command_line
, env_block
);
2210 log_error_errno(errno
, "Failed to execute shutdown binary, %s: %m",
2211 getpid() == 1 ? "freezing" : "quitting");
2214 if (getpid() == 1) {
2216 manager_status_printf(NULL
, STATUS_TYPE_EMERGENCY
,
2217 ANSI_HIGHLIGHT_RED
"!!!!!!" ANSI_NORMAL
,
2218 "%s, freezing.", error_message
);