1 /* SPDX-License-Identifier: LGPL-2.1+ */
10 #include <sys/prctl.h>
11 #include <sys/reboot.h>
17 #if HAVE_VALGRIND_VALGRIND_H
18 #include <valgrind/valgrind.h>
22 #include "sd-daemon.h"
23 #include "sd-messages.h"
25 #include "alloc-util.h"
26 #include "architecture.h"
28 #include "bus-error.h"
30 #include "capability-util.h"
31 #include "cgroup-util.h"
32 #include "clock-util.h"
33 #include "conf-parser.h"
34 #include "cpu-set-util.h"
36 #include "dbus-manager.h"
38 #include "emergency-action.h"
40 #include "exit-status.h"
44 #include "format-util.h"
46 #include "hostname-setup.h"
47 #include "ima-setup.h"
49 #include "kmod-setup.h"
50 #include "limits-util.h"
51 #include "load-fragment.h"
53 #include "loopback-setup.h"
54 #include "machine-id-setup.h"
57 #include "mount-setup.h"
60 #include "parse-util.h"
61 #include "path-util.h"
62 #include "pretty-print.h"
63 #include "proc-cmdline.h"
64 #include "process-util.h"
65 #include "raw-clone.h"
66 #include "rlimit-util.h"
68 #include "seccomp-util.h"
70 #include "selinux-setup.h"
71 #include "selinux-util.h"
72 #include "signal-util.h"
73 #include "smack-setup.h"
75 #include "stat-util.h"
76 #include "stdio-util.h"
78 #include "switch-root.h"
79 #include "sysctl-util.h"
80 #include "terminal-util.h"
81 #include "umask-util.h"
82 #include "user-util.h"
87 #if HAS_FEATURE_ADDRESS_SANITIZER
88 #include <sanitizer/lsan_interface.h>
96 ACTION_DUMP_CONFIGURATION_ITEMS
,
97 ACTION_DUMP_BUS_PROPERTIES
,
98 } arg_action
= ACTION_RUN
;
99 static char *arg_default_unit
= NULL
;
100 static bool arg_system
= false;
101 static bool arg_dump_core
= true;
102 static int arg_crash_chvt
= -1;
103 static bool arg_crash_shell
= false;
104 static bool arg_crash_reboot
= false;
105 static char *arg_confirm_spawn
= NULL
;
106 static ShowStatus arg_show_status
= _SHOW_STATUS_INVALID
;
107 static bool arg_switched_root
= false;
108 static PagerFlags arg_pager_flags
= 0;
109 static bool arg_service_watchdogs
= true;
110 static ExecOutput arg_default_std_output
= EXEC_OUTPUT_JOURNAL
;
111 static ExecOutput arg_default_std_error
= EXEC_OUTPUT_INHERIT
;
112 static usec_t arg_default_restart_usec
= DEFAULT_RESTART_USEC
;
113 static usec_t arg_default_timeout_start_usec
= DEFAULT_TIMEOUT_USEC
;
114 static usec_t arg_default_timeout_stop_usec
= DEFAULT_TIMEOUT_USEC
;
115 static usec_t arg_default_start_limit_interval
= DEFAULT_START_LIMIT_INTERVAL
;
116 static unsigned arg_default_start_limit_burst
= DEFAULT_START_LIMIT_BURST
;
117 static usec_t arg_runtime_watchdog
= 0;
118 static usec_t arg_shutdown_watchdog
= 10 * USEC_PER_MINUTE
;
119 static char *arg_early_core_pattern
= NULL
;
120 static char *arg_watchdog_device
= NULL
;
121 static char **arg_default_environment
= NULL
;
122 static struct rlimit
*arg_default_rlimit
[_RLIMIT_MAX
] = {};
123 static uint64_t arg_capability_bounding_set
= CAP_ALL
;
124 static bool arg_no_new_privs
= false;
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 int arg_default_cpu_accounting
= -1;
130 static bool arg_default_io_accounting
= false;
131 static bool arg_default_ip_accounting
= false;
132 static bool arg_default_blockio_accounting
= false;
133 static bool arg_default_memory_accounting
= MEMORY_ACCOUNTING_DEFAULT
;
134 static bool arg_default_tasks_accounting
= true;
135 static uint64_t arg_default_tasks_max
= UINT64_MAX
;
136 static sd_id128_t arg_machine_id
= {};
137 static EmergencyAction arg_cad_burst_action
= EMERGENCY_ACTION_REBOOT_FORCE
;
139 _noreturn_
static void freeze_or_exit_or_reboot(void) {
141 /* If we are running in a container, let's prefer exiting, after all we can propagate an exit code to
142 * the container manager, and thus inform it that something went wrong. */
143 if (detect_container() > 0) {
144 log_emergency("Exiting PID 1...");
145 _exit(EXIT_EXCEPTION
);
148 if (arg_crash_reboot
) {
149 log_notice("Rebooting in 10s...");
152 log_notice("Rebooting now...");
153 (void) reboot(RB_AUTOBOOT
);
154 log_emergency_errno(errno
, "Failed to reboot: %m");
157 log_emergency("Freezing execution.");
161 _noreturn_
static void crash(int sig
) {
165 if (getpid_cached() != 1)
166 /* Pass this on immediately, if this is not PID 1 */
168 else if (!arg_dump_core
)
169 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig
));
171 sa
= (struct sigaction
) {
172 .sa_handler
= nop_signal_handler
,
173 .sa_flags
= SA_NOCLDSTOP
|SA_RESTART
,
176 /* We want to wait for the core process, hence let's enable SIGCHLD */
177 (void) sigaction(SIGCHLD
, &sa
, NULL
);
179 pid
= raw_clone(SIGCHLD
);
181 log_emergency_errno(errno
, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig
));
183 /* Enable default signal handler for core dump */
185 sa
= (struct sigaction
) {
186 .sa_handler
= SIG_DFL
,
188 (void) sigaction(sig
, &sa
, NULL
);
190 /* Don't limit the coredump size */
191 (void) setrlimit(RLIMIT_CORE
, &RLIMIT_MAKE_CONST(RLIM_INFINITY
));
193 /* Just to be sure... */
196 /* Raise the signal again */
198 (void) kill(pid
, sig
); /* raise() would kill the parent */
200 assert_not_reached("We shouldn't be here...");
201 _exit(EXIT_EXCEPTION
);
206 /* Order things nicely. */
207 r
= wait_for_terminate(pid
, &status
);
209 log_emergency_errno(r
, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig
));
210 else if (status
.si_code
!= CLD_DUMPED
)
211 log_emergency("Caught <%s>, core dump failed (child "PID_FMT
", code=%s, status=%i/%s).",
212 signal_to_string(sig
),
213 pid
, sigchld_code_to_string(status
.si_code
),
215 strna(status
.si_code
== CLD_EXITED
216 ? exit_status_to_string(status
.si_status
, EXIT_STATUS_MINIMAL
)
217 : signal_to_string(status
.si_status
)));
219 log_emergency("Caught <%s>, dumped core as pid "PID_FMT
".", signal_to_string(sig
), pid
);
223 if (arg_crash_chvt
>= 0)
224 (void) chvt(arg_crash_chvt
);
226 sa
= (struct sigaction
) {
227 .sa_handler
= SIG_IGN
,
228 .sa_flags
= SA_NOCLDSTOP
|SA_NOCLDWAIT
|SA_RESTART
,
231 /* Let the kernel reap children for us */
232 (void) sigaction(SIGCHLD
, &sa
, NULL
);
234 if (arg_crash_shell
) {
235 log_notice("Executing crash shell in 10s...");
238 pid
= raw_clone(SIGCHLD
);
240 log_emergency_errno(errno
, "Failed to fork off crash shell: %m");
243 (void) make_console_stdio();
244 (void) rlimit_nofile_safe();
245 (void) execle("/bin/sh", "/bin/sh", NULL
, environ
);
247 log_emergency_errno(errno
, "execle() failed: %m");
248 _exit(EXIT_EXCEPTION
);
250 log_info("Spawned crash shell as PID "PID_FMT
".", pid
);
251 (void) wait_for_terminate(pid
, NULL
);
255 freeze_or_exit_or_reboot();
258 static void install_crash_handler(void) {
259 static const struct sigaction sa
= {
261 .sa_flags
= SA_NODEFER
, /* So that we can raise the signal again from the signal handler */
265 /* We ignore the return value here, since, we don't mind if we
266 * cannot set up a crash handler */
267 r
= sigaction_many(&sa
, SIGNALS_CRASH_HANDLER
, -1);
269 log_debug_errno(r
, "I had trouble setting up the crash handler, ignoring: %m");
272 static int console_setup(void) {
273 _cleanup_close_
int tty_fd
= -1;
276 tty_fd
= open_terminal("/dev/console", O_WRONLY
|O_NOCTTY
|O_CLOEXEC
);
278 return log_error_errno(tty_fd
, "Failed to open /dev/console: %m");
280 /* We don't want to force text mode. plymouth may be showing
281 * pictures already from initrd. */
282 r
= reset_terminal_fd(tty_fd
, false);
284 return log_error_errno(r
, "Failed to reset /dev/console: %m");
289 static int parse_crash_chvt(const char *value
) {
292 if (safe_atoi(value
, &arg_crash_chvt
) >= 0)
295 b
= parse_boolean(value
);
300 arg_crash_chvt
= 0; /* switch to where kmsg goes */
302 arg_crash_chvt
= -1; /* turn off switching */
307 static int parse_confirm_spawn(const char *value
, char **console
) {
311 r
= value
? parse_boolean(value
) : 1;
317 if (r
> 0) /* on with default tty */
318 s
= strdup("/dev/console");
319 else if (is_path(value
)) /* on with fully qualified path */
321 else /* on with only a tty file name, not a fully qualified path */
322 s
= strjoin("/dev/", value
);
330 static int set_machine_id(const char *m
) {
334 if (sd_id128_from_string(m
, &t
) < 0)
337 if (sd_id128_is_null(t
))
344 static int parse_proc_cmdline_item(const char *key
, const char *value
, void *data
) {
350 if (STR_IN_SET(key
, "systemd.unit", "rd.systemd.unit")) {
352 if (proc_cmdline_value_missing(key
, value
))
355 if (!unit_name_is_valid(value
, UNIT_NAME_PLAIN
|UNIT_NAME_INSTANCE
))
356 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key
, value
);
357 else if (in_initrd() == !!startswith(key
, "rd.")) {
358 if (free_and_strdup(&arg_default_unit
, value
) < 0)
362 } else if (proc_cmdline_key_streq(key
, "systemd.dump_core")) {
364 r
= value
? parse_boolean(value
) : true;
366 log_warning_errno(r
, "Failed to parse dump core switch %s, ignoring: %m", value
);
370 } else if (proc_cmdline_key_streq(key
, "systemd.early_core_pattern")) {
372 if (proc_cmdline_value_missing(key
, value
))
375 if (path_is_absolute(value
))
376 (void) parse_path_argument_and_warn(value
, false, &arg_early_core_pattern
);
378 log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value
);
380 } else if (proc_cmdline_key_streq(key
, "systemd.crash_chvt")) {
383 arg_crash_chvt
= 0; /* turn on */
385 r
= parse_crash_chvt(value
);
387 log_warning_errno(r
, "Failed to parse crash chvt switch %s, ignoring: %m", value
);
390 } else if (proc_cmdline_key_streq(key
, "systemd.crash_shell")) {
392 r
= value
? parse_boolean(value
) : true;
394 log_warning_errno(r
, "Failed to parse crash shell switch %s, ignoring: %m", value
);
398 } else if (proc_cmdline_key_streq(key
, "systemd.crash_reboot")) {
400 r
= value
? parse_boolean(value
) : true;
402 log_warning_errno(r
, "Failed to parse crash reboot switch %s, ignoring: %m", value
);
404 arg_crash_reboot
= r
;
406 } else if (proc_cmdline_key_streq(key
, "systemd.confirm_spawn")) {
409 r
= parse_confirm_spawn(value
, &s
);
411 log_warning_errno(r
, "Failed to parse confirm_spawn switch %s, ignoring: %m", value
);
413 free_and_replace(arg_confirm_spawn
, s
);
415 } else if (proc_cmdline_key_streq(key
, "systemd.service_watchdogs")) {
417 r
= value
? parse_boolean(value
) : true;
419 log_warning_errno(r
, "Failed to parse service watchdog switch %s, ignoring: %m", value
);
421 arg_service_watchdogs
= r
;
423 } else if (proc_cmdline_key_streq(key
, "systemd.show_status")) {
426 r
= parse_show_status(value
, &arg_show_status
);
428 log_warning_errno(r
, "Failed to parse show status switch %s, ignoring: %m", value
);
430 arg_show_status
= SHOW_STATUS_YES
;
432 } else if (proc_cmdline_key_streq(key
, "systemd.default_standard_output")) {
434 if (proc_cmdline_value_missing(key
, value
))
437 r
= exec_output_from_string(value
);
439 log_warning_errno(r
, "Failed to parse default standard output switch %s, ignoring: %m", value
);
441 arg_default_std_output
= r
;
443 } else if (proc_cmdline_key_streq(key
, "systemd.default_standard_error")) {
445 if (proc_cmdline_value_missing(key
, value
))
448 r
= exec_output_from_string(value
);
450 log_warning_errno(r
, "Failed to parse default standard error switch %s, ignoring: %m", value
);
452 arg_default_std_error
= r
;
454 } else if (streq(key
, "systemd.setenv")) {
456 if (proc_cmdline_value_missing(key
, value
))
459 if (env_assignment_is_valid(value
)) {
462 env
= strv_env_set(arg_default_environment
, value
);
466 arg_default_environment
= env
;
468 log_warning("Environment variable name '%s' is not valid. Ignoring.", value
);
470 } else if (proc_cmdline_key_streq(key
, "systemd.machine_id")) {
472 if (proc_cmdline_value_missing(key
, value
))
475 r
= set_machine_id(value
);
477 log_warning_errno(r
, "MachineID '%s' is not valid, ignoring: %m", value
);
479 } else if (proc_cmdline_key_streq(key
, "systemd.default_timeout_start_sec")) {
481 if (proc_cmdline_value_missing(key
, value
))
484 r
= parse_sec(value
, &arg_default_timeout_start_usec
);
486 log_warning_errno(r
, "Failed to parse default start timeout '%s', ignoring: %m", value
);
488 if (arg_default_timeout_start_usec
<= 0)
489 arg_default_timeout_start_usec
= USEC_INFINITY
;
491 } else if (proc_cmdline_key_streq(key
, "systemd.watchdog_device")) {
493 if (proc_cmdline_value_missing(key
, value
))
496 (void) parse_path_argument_and_warn(value
, false, &arg_watchdog_device
);
498 } else if (streq(key
, "quiet") && !value
) {
500 if (arg_show_status
== _SHOW_STATUS_INVALID
)
501 arg_show_status
= SHOW_STATUS_AUTO
;
503 } else if (streq(key
, "debug") && !value
) {
505 /* Note that log_parse_environment() handles 'debug'
506 * too, and sets the log level to LOG_DEBUG. */
508 if (detect_container() > 0)
509 log_set_target(LOG_TARGET_CONSOLE
);
514 /* SysV compatibility */
515 target
= runlevel_to_target(key
);
517 return free_and_strdup(&arg_default_unit
, target
);
523 #define DEFINE_SETTER(name, func, descr) \
524 static int name(const char *unit, \
525 const char *filename, \
527 const char *section, \
528 unsigned section_line, \
529 const char *lvalue, \
531 const char *rvalue, \
543 log_syntax(unit, LOG_ERR, filename, line, r, \
544 "Invalid " descr "'%s': %m", \
550 DEFINE_SETTER(config_parse_level2
, log_set_max_level_from_string
, "log level");
551 DEFINE_SETTER(config_parse_target
, log_set_target_from_string
, "target");
552 DEFINE_SETTER(config_parse_color
, log_show_color_from_string
, "color" );
553 DEFINE_SETTER(config_parse_location
, log_show_location_from_string
, "location");
555 static int config_parse_cpu_affinity2(
557 const char *filename
,
560 unsigned section_line
,
567 _cleanup_cpu_free_ cpu_set_t
*c
= NULL
;
570 ncpus
= parse_cpu_set_and_warn(rvalue
, &c
, unit
, filename
, line
, lvalue
);
574 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus
), c
) < 0)
575 log_warning_errno(errno
, "Failed to set CPU affinity: %m");
580 static int config_parse_show_status(
582 const char *filename
,
585 unsigned section_line
,
593 ShowStatus
*b
= data
;
600 k
= parse_show_status(rvalue
, b
);
602 log_syntax(unit
, LOG_ERR
, filename
, line
, k
, "Failed to parse show status setting, ignoring: %s", rvalue
);
609 static int config_parse_output_restricted(
611 const char *filename
,
614 unsigned section_line
,
621 ExecOutput t
, *eo
= data
;
628 t
= exec_output_from_string(rvalue
);
630 log_syntax(unit
, LOG_ERR
, filename
, line
, 0, "Failed to parse output type, ignoring: %s", rvalue
);
634 if (IN_SET(t
, EXEC_OUTPUT_SOCKET
, EXEC_OUTPUT_NAMED_FD
, EXEC_OUTPUT_FILE
, EXEC_OUTPUT_FILE_APPEND
)) {
635 log_syntax(unit
, LOG_ERR
, filename
, line
, 0, "Standard output types socket, fd:, file:, append: are not supported as defaults, ignoring: %s", rvalue
);
643 static int config_parse_crash_chvt(
645 const char *filename
,
648 unsigned section_line
,
661 r
= parse_crash_chvt(rvalue
);
663 log_syntax(unit
, LOG_ERR
, filename
, line
, r
, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue
);
670 static int parse_config_file(void) {
672 const ConfigTableItem items
[] = {
673 { "Manager", "LogLevel", config_parse_level2
, 0, NULL
},
674 { "Manager", "LogTarget", config_parse_target
, 0, NULL
},
675 { "Manager", "LogColor", config_parse_color
, 0, NULL
},
676 { "Manager", "LogLocation", config_parse_location
, 0, NULL
},
677 { "Manager", "DumpCore", config_parse_bool
, 0, &arg_dump_core
},
678 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt
, 0, NULL
},
679 { "Manager", "CrashChangeVT", config_parse_crash_chvt
, 0, NULL
},
680 { "Manager", "CrashShell", config_parse_bool
, 0, &arg_crash_shell
},
681 { "Manager", "CrashReboot", config_parse_bool
, 0, &arg_crash_reboot
},
682 { "Manager", "ShowStatus", config_parse_show_status
, 0, &arg_show_status
},
683 { "Manager", "CPUAffinity", config_parse_cpu_affinity2
, 0, NULL
},
684 { "Manager", "JoinControllers", config_parse_warn_compat
, DISABLED_CONFIGURATION
, NULL
},
685 { "Manager", "RuntimeWatchdogSec", config_parse_sec
, 0, &arg_runtime_watchdog
},
686 { "Manager", "ShutdownWatchdogSec", config_parse_sec
, 0, &arg_shutdown_watchdog
},
687 { "Manager", "WatchdogDevice", config_parse_path
, 0, &arg_watchdog_device
},
688 { "Manager", "CapabilityBoundingSet", config_parse_capability_set
, 0, &arg_capability_bounding_set
},
689 { "Manager", "NoNewPrivileges", config_parse_bool
, 0, &arg_no_new_privs
},
691 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs
, 0, &arg_syscall_archs
},
693 { "Manager", "TimerSlackNSec", config_parse_nsec
, 0, &arg_timer_slack_nsec
},
694 { "Manager", "DefaultTimerAccuracySec", config_parse_sec
, 0, &arg_default_timer_accuracy_usec
},
695 { "Manager", "DefaultStandardOutput", config_parse_output_restricted
,0, &arg_default_std_output
},
696 { "Manager", "DefaultStandardError", config_parse_output_restricted
,0, &arg_default_std_error
},
697 { "Manager", "DefaultTimeoutStartSec", config_parse_sec
, 0, &arg_default_timeout_start_usec
},
698 { "Manager", "DefaultTimeoutStopSec", config_parse_sec
, 0, &arg_default_timeout_stop_usec
},
699 { "Manager", "DefaultRestartSec", config_parse_sec
, 0, &arg_default_restart_usec
},
700 { "Manager", "DefaultStartLimitInterval", config_parse_sec
, 0, &arg_default_start_limit_interval
}, /* obsolete alias */
701 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec
, 0, &arg_default_start_limit_interval
},
702 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned
, 0, &arg_default_start_limit_burst
},
703 { "Manager", "DefaultEnvironment", config_parse_environ
, 0, &arg_default_environment
},
704 { "Manager", "DefaultLimitCPU", config_parse_rlimit
, RLIMIT_CPU
, arg_default_rlimit
},
705 { "Manager", "DefaultLimitFSIZE", config_parse_rlimit
, RLIMIT_FSIZE
, arg_default_rlimit
},
706 { "Manager", "DefaultLimitDATA", config_parse_rlimit
, RLIMIT_DATA
, arg_default_rlimit
},
707 { "Manager", "DefaultLimitSTACK", config_parse_rlimit
, RLIMIT_STACK
, arg_default_rlimit
},
708 { "Manager", "DefaultLimitCORE", config_parse_rlimit
, RLIMIT_CORE
, arg_default_rlimit
},
709 { "Manager", "DefaultLimitRSS", config_parse_rlimit
, RLIMIT_RSS
, arg_default_rlimit
},
710 { "Manager", "DefaultLimitNOFILE", config_parse_rlimit
, RLIMIT_NOFILE
, arg_default_rlimit
},
711 { "Manager", "DefaultLimitAS", config_parse_rlimit
, RLIMIT_AS
, arg_default_rlimit
},
712 { "Manager", "DefaultLimitNPROC", config_parse_rlimit
, RLIMIT_NPROC
, arg_default_rlimit
},
713 { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit
, RLIMIT_MEMLOCK
, arg_default_rlimit
},
714 { "Manager", "DefaultLimitLOCKS", config_parse_rlimit
, RLIMIT_LOCKS
, arg_default_rlimit
},
715 { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit
, RLIMIT_SIGPENDING
, arg_default_rlimit
},
716 { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit
, RLIMIT_MSGQUEUE
, arg_default_rlimit
},
717 { "Manager", "DefaultLimitNICE", config_parse_rlimit
, RLIMIT_NICE
, arg_default_rlimit
},
718 { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit
, RLIMIT_RTPRIO
, arg_default_rlimit
},
719 { "Manager", "DefaultLimitRTTIME", config_parse_rlimit
, RLIMIT_RTTIME
, arg_default_rlimit
},
720 { "Manager", "DefaultCPUAccounting", config_parse_tristate
, 0, &arg_default_cpu_accounting
},
721 { "Manager", "DefaultIOAccounting", config_parse_bool
, 0, &arg_default_io_accounting
},
722 { "Manager", "DefaultIPAccounting", config_parse_bool
, 0, &arg_default_ip_accounting
},
723 { "Manager", "DefaultBlockIOAccounting", config_parse_bool
, 0, &arg_default_blockio_accounting
},
724 { "Manager", "DefaultMemoryAccounting", config_parse_bool
, 0, &arg_default_memory_accounting
},
725 { "Manager", "DefaultTasksAccounting", config_parse_bool
, 0, &arg_default_tasks_accounting
},
726 { "Manager", "DefaultTasksMax", config_parse_tasks_max
, 0, &arg_default_tasks_max
},
727 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action
, 0, &arg_cad_burst_action
},
731 const char *fn
, *conf_dirs_nulstr
;
734 PKGSYSCONFDIR
"/system.conf" :
735 PKGSYSCONFDIR
"/user.conf";
737 conf_dirs_nulstr
= arg_system
?
738 CONF_PATHS_NULSTR("systemd/system.conf.d") :
739 CONF_PATHS_NULSTR("systemd/user.conf.d");
741 (void) config_parse_many_nulstr(fn
, conf_dirs_nulstr
, "Manager\0", config_item_table_lookup
, items
, CONFIG_PARSE_WARN
, NULL
);
743 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
744 * like everywhere else. */
745 if (arg_default_timeout_start_usec
<= 0)
746 arg_default_timeout_start_usec
= USEC_INFINITY
;
747 if (arg_default_timeout_stop_usec
<= 0)
748 arg_default_timeout_stop_usec
= USEC_INFINITY
;
753 static void set_manager_defaults(Manager
*m
) {
757 /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
758 * affect the manager itself, but are just what newly allocated units will have set if they haven't set
759 * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
761 m
->default_timer_accuracy_usec
= arg_default_timer_accuracy_usec
;
762 m
->default_std_output
= arg_default_std_output
;
763 m
->default_std_error
= arg_default_std_error
;
764 m
->default_timeout_start_usec
= arg_default_timeout_start_usec
;
765 m
->default_timeout_stop_usec
= arg_default_timeout_stop_usec
;
766 m
->default_restart_usec
= arg_default_restart_usec
;
767 m
->default_start_limit_interval
= arg_default_start_limit_interval
;
768 m
->default_start_limit_burst
= arg_default_start_limit_burst
;
770 /* On 4.15+ with unified hierarchy, CPU accounting is essentially free as it doesn't require the CPU
771 * controller to be enabled, so the default is to enable it unless we got told otherwise. */
772 if (arg_default_cpu_accounting
>= 0)
773 m
->default_cpu_accounting
= arg_default_cpu_accounting
;
775 m
->default_cpu_accounting
= cpu_accounting_is_cheap();
777 m
->default_io_accounting
= arg_default_io_accounting
;
778 m
->default_ip_accounting
= arg_default_ip_accounting
;
779 m
->default_blockio_accounting
= arg_default_blockio_accounting
;
780 m
->default_memory_accounting
= arg_default_memory_accounting
;
781 m
->default_tasks_accounting
= arg_default_tasks_accounting
;
782 m
->default_tasks_max
= arg_default_tasks_max
;
784 (void) manager_set_default_rlimits(m
, arg_default_rlimit
);
786 (void) manager_default_environment(m
);
787 (void) manager_transient_environment_add(m
, arg_default_environment
);
790 static void set_manager_settings(Manager
*m
) {
794 /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
795 * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
797 m
->confirm_spawn
= arg_confirm_spawn
;
798 m
->service_watchdogs
= arg_service_watchdogs
;
799 m
->runtime_watchdog
= arg_runtime_watchdog
;
800 m
->shutdown_watchdog
= arg_shutdown_watchdog
;
801 m
->cad_burst_action
= arg_cad_burst_action
;
803 manager_set_show_status(m
, arg_show_status
);
806 static int parse_argv(int argc
, char *argv
[]) {
808 ARG_LOG_LEVEL
= 0x100,
818 ARG_DUMP_CONFIGURATION_ITEMS
,
819 ARG_DUMP_BUS_PROPERTIES
,
828 ARG_DEFAULT_STD_OUTPUT
,
829 ARG_DEFAULT_STD_ERROR
,
831 ARG_SERVICE_WATCHDOGS
,
834 static const struct option options
[] = {
835 { "log-level", required_argument
, NULL
, ARG_LOG_LEVEL
},
836 { "log-target", required_argument
, NULL
, ARG_LOG_TARGET
},
837 { "log-color", optional_argument
, NULL
, ARG_LOG_COLOR
},
838 { "log-location", optional_argument
, NULL
, ARG_LOG_LOCATION
},
839 { "unit", required_argument
, NULL
, ARG_UNIT
},
840 { "system", no_argument
, NULL
, ARG_SYSTEM
},
841 { "user", no_argument
, NULL
, ARG_USER
},
842 { "test", no_argument
, NULL
, ARG_TEST
},
843 { "no-pager", no_argument
, NULL
, ARG_NO_PAGER
},
844 { "help", no_argument
, NULL
, 'h' },
845 { "version", no_argument
, NULL
, ARG_VERSION
},
846 { "dump-configuration-items", no_argument
, NULL
, ARG_DUMP_CONFIGURATION_ITEMS
},
847 { "dump-bus-properties", no_argument
, NULL
, ARG_DUMP_BUS_PROPERTIES
},
848 { "dump-core", optional_argument
, NULL
, ARG_DUMP_CORE
},
849 { "crash-chvt", required_argument
, NULL
, ARG_CRASH_CHVT
},
850 { "crash-shell", optional_argument
, NULL
, ARG_CRASH_SHELL
},
851 { "crash-reboot", optional_argument
, NULL
, ARG_CRASH_REBOOT
},
852 { "confirm-spawn", optional_argument
, NULL
, ARG_CONFIRM_SPAWN
},
853 { "show-status", optional_argument
, NULL
, ARG_SHOW_STATUS
},
854 { "deserialize", required_argument
, NULL
, ARG_DESERIALIZE
},
855 { "switched-root", no_argument
, NULL
, ARG_SWITCHED_ROOT
},
856 { "default-standard-output", required_argument
, NULL
, ARG_DEFAULT_STD_OUTPUT
, },
857 { "default-standard-error", required_argument
, NULL
, ARG_DEFAULT_STD_ERROR
, },
858 { "machine-id", required_argument
, NULL
, ARG_MACHINE_ID
},
859 { "service-watchdogs", required_argument
, NULL
, ARG_SERVICE_WATCHDOGS
},
868 if (getpid_cached() == 1)
871 while ((c
= getopt_long(argc
, argv
, "hDbsz:", options
, NULL
)) >= 0)
876 r
= log_set_max_level_from_string(optarg
);
878 return log_error_errno(r
, "Failed to parse log level \"%s\": %m", optarg
);
883 r
= log_set_target_from_string(optarg
);
885 return log_error_errno(r
, "Failed to parse log target \"%s\": %m", optarg
);
892 r
= log_show_color_from_string(optarg
);
894 return log_error_errno(r
, "Failed to parse log color setting \"%s\": %m",
897 log_show_color(true);
901 case ARG_LOG_LOCATION
:
903 r
= log_show_location_from_string(optarg
);
905 return log_error_errno(r
, "Failed to parse log location setting \"%s\": %m",
908 log_show_location(true);
912 case ARG_DEFAULT_STD_OUTPUT
:
913 r
= exec_output_from_string(optarg
);
915 return log_error_errno(r
, "Failed to parse default standard output setting \"%s\": %m",
917 arg_default_std_output
= r
;
920 case ARG_DEFAULT_STD_ERROR
:
921 r
= exec_output_from_string(optarg
);
923 return log_error_errno(r
, "Failed to parse default standard error output setting \"%s\": %m",
925 arg_default_std_error
= r
;
929 r
= free_and_strdup(&arg_default_unit
, optarg
);
931 return log_error_errno(r
, "Failed to set default unit \"%s\": %m", optarg
);
944 arg_action
= ACTION_TEST
;
948 arg_pager_flags
|= PAGER_DISABLE
;
952 arg_action
= ACTION_VERSION
;
955 case ARG_DUMP_CONFIGURATION_ITEMS
:
956 arg_action
= ACTION_DUMP_CONFIGURATION_ITEMS
;
959 case ARG_DUMP_BUS_PROPERTIES
:
960 arg_action
= ACTION_DUMP_BUS_PROPERTIES
;
965 arg_dump_core
= true;
967 r
= parse_boolean(optarg
);
969 return log_error_errno(r
, "Failed to parse dump core boolean: \"%s\": %m",
976 r
= parse_crash_chvt(optarg
);
978 return log_error_errno(r
, "Failed to parse crash virtual terminal index: \"%s\": %m",
982 case ARG_CRASH_SHELL
:
984 arg_crash_shell
= true;
986 r
= parse_boolean(optarg
);
988 return log_error_errno(r
, "Failed to parse crash shell boolean: \"%s\": %m",
994 case ARG_CRASH_REBOOT
:
996 arg_crash_reboot
= true;
998 r
= parse_boolean(optarg
);
1000 return log_error_errno(r
, "Failed to parse crash shell boolean: \"%s\": %m",
1002 arg_crash_reboot
= r
;
1006 case ARG_CONFIRM_SPAWN
:
1007 arg_confirm_spawn
= mfree(arg_confirm_spawn
);
1009 r
= parse_confirm_spawn(optarg
, &arg_confirm_spawn
);
1011 return log_error_errno(r
, "Failed to parse confirm spawn option: \"%s\": %m",
1015 case ARG_SERVICE_WATCHDOGS
:
1016 r
= parse_boolean(optarg
);
1018 return log_error_errno(r
, "Failed to parse service watchdogs boolean: \"%s\": %m",
1020 arg_service_watchdogs
= r
;
1023 case ARG_SHOW_STATUS
:
1025 r
= parse_show_status(optarg
, &arg_show_status
);
1027 return log_error_errno(r
, "Failed to parse show status boolean: \"%s\": %m",
1030 arg_show_status
= SHOW_STATUS_YES
;
1033 case ARG_DESERIALIZE
: {
1037 r
= safe_atoi(optarg
, &fd
);
1039 log_error_errno(r
, "Failed to parse deserialize option \"%s\": %m", optarg
);
1041 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
),
1042 "Invalid deserialize fd: %d",
1045 (void) fd_cloexec(fd
, true);
1047 f
= fdopen(fd
, "r");
1049 return log_error_errno(errno
, "Failed to open serialization fd %d: %m", fd
);
1051 safe_fclose(arg_serialization
);
1052 arg_serialization
= f
;
1057 case ARG_SWITCHED_ROOT
:
1058 arg_switched_root
= true;
1061 case ARG_MACHINE_ID
:
1062 r
= set_machine_id(optarg
);
1064 return log_error_errno(r
, "MachineID '%s' is not valid: %m", optarg
);
1068 arg_action
= ACTION_HELP
;
1072 log_set_max_level(LOG_DEBUG
);
1078 /* Just to eat away the sysvinit kernel
1079 * cmdline args without getopt() error
1080 * messages that we'll parse in
1081 * parse_proc_cmdline_word() or ignore. */
1084 if (getpid_cached() != 1)
1090 assert_not_reached("Unhandled option code.");
1093 if (optind
< argc
&& getpid_cached() != 1) {
1094 /* Hmm, when we aren't run as init system
1095 * let's complain about excess arguments */
1097 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
),
1098 "Excess arguments.");
1104 static int help(void) {
1105 _cleanup_free_
char *link
= NULL
;
1108 r
= terminal_urlify_man("systemd", "1", &link
);
1112 printf("%s [OPTIONS...]\n\n"
1113 "Starts up and maintains the system or user services.\n\n"
1114 " -h --help Show this help\n"
1115 " --version Show version\n"
1116 " --test Determine startup sequence, dump it and exit\n"
1117 " --no-pager Do not pipe output into a pager\n"
1118 " --dump-configuration-items Dump understood unit configuration items\n"
1119 " --dump-bus-properties Dump exposed bus properties\n"
1120 " --unit=UNIT Set default unit\n"
1121 " --system Run a system instance, even if PID != 1\n"
1122 " --user Run a user instance\n"
1123 " --dump-core[=BOOL] Dump core on crash\n"
1124 " --crash-vt=NR Change to specified VT on crash\n"
1125 " --crash-reboot[=BOOL] Reboot on crash\n"
1126 " --crash-shell[=BOOL] Run shell on crash\n"
1127 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1128 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1129 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1130 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1131 " --log-color[=BOOL] Highlight important log messages\n"
1132 " --log-location[=BOOL] Include code location in log messages\n"
1133 " --default-standard-output= Set default standard output for services\n"
1134 " --default-standard-error= Set default standard error output for services\n"
1135 "\nSee the %s for details.\n"
1136 , program_invocation_short_name
1143 static int prepare_reexecute(
1147 bool switching_root
) {
1149 _cleanup_fdset_free_ FDSet
*fds
= NULL
;
1150 _cleanup_fclose_
FILE *f
= NULL
;
1157 r
= manager_open_serialization(m
, &f
);
1159 return log_error_errno(r
, "Failed to create serialization file: %m");
1161 /* Make sure nothing is really destructed when we shut down */
1163 bus_manager_send_reloading(m
, true);
1169 r
= manager_serialize(m
, f
, fds
, switching_root
);
1173 if (fseeko(f
, 0, SEEK_SET
) == (off_t
) -1)
1174 return log_error_errno(errno
, "Failed to rewind serialization fd: %m");
1176 r
= fd_cloexec(fileno(f
), false);
1178 return log_error_errno(r
, "Failed to disable O_CLOEXEC for serialization: %m");
1180 r
= fdset_cloexec(fds
, false);
1182 return log_error_errno(r
, "Failed to disable O_CLOEXEC for serialization fds: %m");
1184 *ret_f
= TAKE_PTR(f
);
1185 *ret_fds
= TAKE_PTR(fds
);
1190 static void bump_file_max_and_nr_open(void) {
1192 /* Let's bump fs.file-max and fs.nr_open to their respective maximums. On current kernels large numbers of file
1193 * descriptors are no longer a performance problem and their memory is properly tracked by memcg, thus counting
1194 * them and limiting them in another two layers of limits is unnecessary and just complicates things. This
1195 * function hence turns off 2 of the 4 levels of limits on file descriptors, and makes RLIMIT_NOLIMIT (soft +
1196 * hard) the only ones that really matter. */
1198 #if BUMP_PROC_SYS_FS_FILE_MAX || BUMP_PROC_SYS_FS_NR_OPEN
1199 _cleanup_free_
char *t
= NULL
;
1203 #if BUMP_PROC_SYS_FS_FILE_MAX
1204 /* I so wanted to use STRINGIFY(ULONG_MAX) here, but alas we can't as glibc/gcc define that as
1205 * "(0x7fffffffffffffffL * 2UL + 1UL)". Seriously. 😢 */
1206 if (asprintf(&t
, "%lu\n", ULONG_MAX
) < 0) {
1211 r
= sysctl_write("fs/file-max", t
);
1213 log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
, "Failed to bump fs.file-max, ignoring: %m");
1216 #if BUMP_PROC_SYS_FS_FILE_MAX && BUMP_PROC_SYS_FS_NR_OPEN
1220 #if BUMP_PROC_SYS_FS_NR_OPEN
1223 /* Arg! The kernel enforces maximum and minimum values on the fs.nr_open, but we don't really know what they
1224 * are. The expression by which the maximum is determined is dependent on the architecture, and is something we
1225 * don't really want to copy to userspace, as it is dependent on implementation details of the kernel. Since
1226 * the kernel doesn't expose the maximum value to us, we can only try and hope. Hence, let's start with
1227 * INT_MAX, and then keep halving the value until we find one that works. Ugly? Yes, absolutely, but kernel
1228 * APIs are kernel APIs, so what do can we do... 🤯 */
1233 v
&= ~(__SIZEOF_POINTER__
- 1); /* Round down to next multiple of the pointer size */
1235 log_warning("Can't bump fs.nr_open, value too small.");
1241 log_error_errno(k
, "Failed to read fs.nr_open: %m");
1244 if (k
>= v
) { /* Already larger */
1245 log_debug("Skipping bump, value is already larger.");
1249 if (asprintf(&t
, "%i\n", v
) < 0) {
1254 r
= sysctl_write("fs/nr_open", t
);
1257 log_debug("Couldn't write fs.nr_open as %i, halving it.", v
);
1262 log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
, "Failed to bump fs.nr_open, ignoring: %m");
1266 log_debug("Successfully bumped fs.nr_open to %i", v
);
1272 static int bump_rlimit_nofile(struct rlimit
*saved_rlimit
) {
1273 struct rlimit new_rlimit
;
1276 assert(saved_rlimit
);
1278 /* Save the original RLIMIT_NOFILE so that we can reset it later when transitioning from the initrd to the main
1279 * systemd or suchlike. */
1280 if (getrlimit(RLIMIT_NOFILE
, saved_rlimit
) < 0)
1281 return log_warning_errno(errno
, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1283 /* Get the underlying absolute limit the kernel enforces */
1284 nr
= read_nr_open();
1286 /* Make sure forked processes get limits based on the original kernel setting */
1287 if (!arg_default_rlimit
[RLIMIT_NOFILE
]) {
1290 rl
= newdup(struct rlimit
, saved_rlimit
, 1);
1294 /* Bump the hard limit for system services to a substantially higher value. The default hard limit
1295 * current kernels set is pretty low (4K), mostly for historical reasons. According to kernel
1296 * developers, the fd handling in recent kernels has been optimized substantially enough, so that we
1297 * can bump the limit now, without paying too high a price in memory or performance. Note however that
1298 * we only bump the hard limit, not the soft limit. That's because select() works the way it works, and
1299 * chokes on fds >= 1024. If we'd bump the soft limit globally, it might accidentally happen to
1300 * unexpecting programs that they get fds higher than what they can process using select(). By only
1301 * bumping the hard limit but leaving the low limit as it is we avoid this pitfall: programs that are
1302 * written by folks aware of the select() problem in mind (and thus use poll()/epoll instead of
1303 * select(), the way everybody should) can explicitly opt into high fds by bumping their soft limit
1304 * beyond 1024, to the hard limit we pass. */
1306 rl
->rlim_max
= MIN((rlim_t
) nr
, MAX(rl
->rlim_max
, (rlim_t
) HIGH_RLIMIT_NOFILE
));
1308 /* If for some reason we were invoked with a soft limit above 1024 (which should never
1309 * happen!, but who knows what we get passed in from pam_limit when invoked as --user
1310 * instance), then lower what we pass on to not confuse our children */
1311 rl
->rlim_cur
= MIN(rl
->rlim_cur
, (rlim_t
) FD_SETSIZE
);
1313 arg_default_rlimit
[RLIMIT_NOFILE
] = rl
;
1316 /* Calculate the new limits to use for us. Never lower from what we inherited. */
1317 new_rlimit
= (struct rlimit
) {
1318 .rlim_cur
= MAX((rlim_t
) nr
, saved_rlimit
->rlim_cur
),
1319 .rlim_max
= MAX((rlim_t
) nr
, saved_rlimit
->rlim_max
),
1322 /* Shortcut if nothing changes. */
1323 if (saved_rlimit
->rlim_max
>= new_rlimit
.rlim_max
&&
1324 saved_rlimit
->rlim_cur
>= new_rlimit
.rlim_cur
) {
1325 log_debug("RLIMIT_NOFILE is already as high or higher than we need it, not bumping.");
1329 /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for
1330 * both hard and soft. */
1331 r
= setrlimit_closest(RLIMIT_NOFILE
, &new_rlimit
);
1333 return log_warning_errno(r
, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1338 static int bump_rlimit_memlock(struct rlimit
*saved_rlimit
) {
1339 struct rlimit new_rlimit
;
1342 assert(saved_rlimit
);
1344 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK which should
1345 * normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's bump
1346 * the value high enough for our user. */
1348 if (getrlimit(RLIMIT_MEMLOCK
, saved_rlimit
) < 0)
1349 return log_warning_errno(errno
, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1351 /* Pass the original value down to invoked processes */
1352 if (!arg_default_rlimit
[RLIMIT_MEMLOCK
]) {
1355 rl
= newdup(struct rlimit
, saved_rlimit
, 1);
1359 arg_default_rlimit
[RLIMIT_MEMLOCK
] = rl
;
1362 /* Using MAX() on resource limits only is safe if RLIM_INFINITY is > 0. POSIX declares that rlim_t
1363 * must be unsigned, hence this is a given, but let's make this clear here. */
1364 assert_cc(RLIM_INFINITY
> 0);
1366 new_rlimit
= (struct rlimit
) {
1367 .rlim_cur
= MAX(HIGH_RLIMIT_MEMLOCK
, saved_rlimit
->rlim_cur
),
1368 .rlim_max
= MAX(HIGH_RLIMIT_MEMLOCK
, saved_rlimit
->rlim_max
),
1371 if (saved_rlimit
->rlim_max
>= new_rlimit
.rlim_cur
&&
1372 saved_rlimit
->rlim_cur
>= new_rlimit
.rlim_max
) {
1373 log_debug("RLIMIT_MEMLOCK is already as high or higher than we need it, not bumping.");
1377 r
= setrlimit_closest(RLIMIT_MEMLOCK
, &new_rlimit
);
1379 return log_warning_errno(r
, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1384 static void test_usr(void) {
1386 /* Check that /usr is either on the same file system as / or mounted already. */
1388 if (dir_is_empty("/usr") <= 0)
1391 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1392 "Some things will probably break (sometimes even silently) in mysterious ways. "
1393 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1396 static int enforce_syscall_archs(Set
*archs
) {
1400 if (!is_seccomp_available())
1403 r
= seccomp_restrict_archs(arg_syscall_archs
);
1405 return log_error_errno(r
, "Failed to enforce system call architecture restrication: %m");
1410 static int status_welcome(void) {
1411 _cleanup_free_
char *pretty_name
= NULL
, *ansi_color
= NULL
;
1414 if (IN_SET(arg_show_status
, SHOW_STATUS_NO
, SHOW_STATUS_AUTO
))
1417 r
= parse_os_release(NULL
,
1418 "PRETTY_NAME", &pretty_name
,
1419 "ANSI_COLOR", &ansi_color
,
1422 log_full_errno(r
== -ENOENT
? LOG_DEBUG
: LOG_WARNING
, r
,
1423 "Failed to read os-release file, ignoring: %m");
1425 if (log_get_show_color())
1426 return status_printf(NULL
, 0,
1427 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1428 isempty(ansi_color
) ? "1" : ansi_color
,
1429 isempty(pretty_name
) ? "Linux" : pretty_name
);
1431 return status_printf(NULL
, 0,
1432 "\nWelcome to %s!\n",
1433 isempty(pretty_name
) ? "Linux" : pretty_name
);
1436 static int write_container_id(void) {
1440 c
= getenv("container");
1444 RUN_WITH_UMASK(0022)
1445 r
= write_string_file("/run/systemd/container", c
, WRITE_STRING_FILE_CREATE
);
1447 return log_warning_errno(r
, "Failed to write /run/systemd/container, ignoring: %m");
1452 static int bump_unix_max_dgram_qlen(void) {
1453 _cleanup_free_
char *qlen
= NULL
;
1457 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1458 * really really early during boot, so that it is actually applied to all our sockets, including the
1459 * $NOTIFY_SOCKET one. */
1461 r
= read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen
);
1463 return log_full_errno(r
== -ENOENT
? LOG_DEBUG
: LOG_WARNING
, r
, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1465 r
= safe_atolu(qlen
, &v
);
1467 return log_warning_errno(r
, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen
);
1469 if (v
>= DEFAULT_UNIX_MAX_DGRAM_QLEN
)
1472 r
= write_string_filef("/proc/sys/net/unix/max_dgram_qlen", WRITE_STRING_FILE_DISABLE_BUFFER
, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN
);
1474 return log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
,
1475 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1480 static int fixup_environment(void) {
1481 _cleanup_free_
char *term
= NULL
;
1485 /* Only fix up the environment when we are started as PID 1 */
1486 if (getpid_cached() != 1)
1489 /* We expect the environment to be set correctly if run inside a container. */
1490 if (detect_container() > 0)
1493 /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1494 * device used by the console. We try to make a better guess here since some consoles might not have support
1495 * for color mode for example.
1497 * However if TERM was configured through the kernel command line then leave it alone. */
1498 r
= proc_cmdline_get_key("TERM", 0, &term
);
1502 t
= term
?: default_term_for_tty("/dev/console");
1504 if (setenv("TERM", t
, 1) < 0)
1510 static void redirect_telinit(int argc
, char *argv
[]) {
1512 /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1514 #if HAVE_SYSV_COMPAT
1515 if (getpid_cached() == 1)
1518 if (!strstr(program_invocation_short_name
, "init"))
1521 execv(SYSTEMCTL_BINARY_PATH
, argv
);
1522 log_error_errno(errno
, "Failed to exec " SYSTEMCTL_BINARY_PATH
": %m");
1527 static int become_shutdown(
1528 const char *shutdown_verb
,
1531 char log_level
[DECIMAL_STR_MAX(int) + 1],
1532 exit_code
[DECIMAL_STR_MAX(uint8_t) + 1],
1533 timeout
[DECIMAL_STR_MAX(usec_t
) + 1];
1535 const char* command_line
[13] = {
1536 SYSTEMD_SHUTDOWN_BINARY_PATH
,
1538 "--timeout", timeout
,
1539 "--log-level", log_level
,
1543 _cleanup_strv_free_
char **env_block
= NULL
;
1547 assert(shutdown_verb
);
1548 assert(!command_line
[pos
]);
1549 env_block
= strv_copy(environ
);
1551 xsprintf(log_level
, "%d", log_get_max_level());
1552 xsprintf(timeout
, "%" PRI_USEC
"us", arg_default_timeout_stop_usec
);
1554 switch (log_get_target()) {
1556 case LOG_TARGET_KMSG
:
1557 case LOG_TARGET_JOURNAL_OR_KMSG
:
1558 case LOG_TARGET_SYSLOG_OR_KMSG
:
1559 command_line
[pos
++] = "kmsg";
1562 case LOG_TARGET_NULL
:
1563 command_line
[pos
++] = "null";
1566 case LOG_TARGET_CONSOLE
:
1568 command_line
[pos
++] = "console";
1572 if (log_get_show_color())
1573 command_line
[pos
++] = "--log-color";
1575 if (log_get_show_location())
1576 command_line
[pos
++] = "--log-location";
1578 if (streq(shutdown_verb
, "exit")) {
1579 command_line
[pos
++] = "--exit-code";
1580 command_line
[pos
++] = exit_code
;
1581 xsprintf(exit_code
, "%d", retval
);
1584 assert(pos
< ELEMENTSOF(command_line
));
1586 if (streq(shutdown_verb
, "reboot") &&
1587 arg_shutdown_watchdog
> 0 &&
1588 arg_shutdown_watchdog
!= USEC_INFINITY
) {
1592 /* If we reboot let's set the shutdown
1593 * watchdog and tell the shutdown binary to
1594 * repeatedly ping it */
1595 r
= watchdog_set_timeout(&arg_shutdown_watchdog
);
1596 watchdog_close(r
< 0);
1598 /* Tell the binary how often to ping, ignore failure */
1599 if (asprintf(&e
, "WATCHDOG_USEC="USEC_FMT
, arg_shutdown_watchdog
) > 0)
1600 (void) strv_consume(&env_block
, e
);
1602 if (arg_watchdog_device
&&
1603 asprintf(&e
, "WATCHDOG_DEVICE=%s", arg_watchdog_device
) > 0)
1604 (void) strv_consume(&env_block
, e
);
1606 watchdog_close(true);
1608 /* Avoid the creation of new processes forked by the
1609 * kernel; at this point, we will not listen to the
1611 if (detect_container() <= 0)
1612 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER
);
1614 execve(SYSTEMD_SHUTDOWN_BINARY_PATH
, (char **) command_line
, env_block
);
1618 static void initialize_clock(void) {
1621 if (clock_is_localtime(NULL
) > 0) {
1625 * The very first call of settimeofday() also does a time warp in the kernel.
1627 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1628 * of maintaining the RTC and do all adjustments. This matches the behavior of Windows, which leaves
1629 * the RTC alone if the registry tells that the RTC runs in UTC.
1631 r
= clock_set_timezone(&min
);
1633 log_error_errno(r
, "Failed to apply local time delta, ignoring: %m");
1635 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min
);
1637 } else if (!in_initrd()) {
1639 * Do a dummy very first call to seal the kernel's time warp magic.
1641 * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1642 * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1643 * until we reach the real system.
1645 * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1646 * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1647 * be treated as UTC that way.
1649 (void) clock_reset_timewarp();
1652 r
= clock_apply_epoch();
1654 log_error_errno(r
, "Current system time is before build time, but cannot correct: %m");
1656 log_info("System time before build time, advancing clock.");
1659 static void initialize_coredump(bool skip_setup
) {
1661 if (getpid_cached() != 1)
1664 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1665 * will process core dumps for system services by default. */
1666 if (setrlimit(RLIMIT_CORE
, &RLIMIT_MAKE_CONST(RLIM_INFINITY
)) < 0)
1667 log_warning_errno(errno
, "Failed to set RLIMIT_CORE: %m");
1669 /* But at the same time, turn off the core_pattern logic by default, so that no
1670 * coredumps are stored until the systemd-coredump tool is enabled via
1671 * sysctl. However it can be changed via the kernel command line later so core
1672 * dumps can still be generated during early startup and in initramfs. */
1674 disable_coredumps();
1678 static void initialize_core_pattern(bool skip_setup
) {
1681 if (skip_setup
|| !arg_early_core_pattern
)
1684 if (getpid_cached() != 1)
1687 r
= write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern
, WRITE_STRING_FILE_DISABLE_BUFFER
);
1689 log_warning_errno(r
, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", arg_early_core_pattern
);
1692 static void do_reexecute(
1695 const struct rlimit
*saved_rlimit_nofile
,
1696 const struct rlimit
*saved_rlimit_memlock
,
1698 const char *switch_root_dir
,
1699 const char *switch_root_init
,
1700 const char **ret_error_message
) {
1702 unsigned i
, j
, args_size
;
1706 assert(saved_rlimit_nofile
);
1707 assert(saved_rlimit_memlock
);
1708 assert(ret_error_message
);
1710 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1712 watchdog_close(true);
1714 /* Reset RLIMIT_NOFILE + RLIMIT_MEMLOCK back to the kernel defaults, so that the new systemd can pass
1715 * the kernel default to its child processes */
1716 if (saved_rlimit_nofile
->rlim_cur
!= 0)
1717 (void) setrlimit(RLIMIT_NOFILE
, saved_rlimit_nofile
);
1718 if (saved_rlimit_memlock
->rlim_cur
!= RLIM_INFINITY
)
1719 (void) setrlimit(RLIMIT_MEMLOCK
, saved_rlimit_memlock
);
1721 if (switch_root_dir
) {
1722 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1723 * SIGCHLD for them after deserializing. */
1724 broadcast_signal(SIGTERM
, false, true, arg_default_timeout_stop_usec
);
1726 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1727 r
= switch_root(switch_root_dir
, "/mnt", true, MS_MOVE
);
1729 log_error_errno(r
, "Failed to switch root, trying to continue: %m");
1732 args_size
= MAX(6, argc
+1);
1733 args
= newa(const char*, args_size
);
1735 if (!switch_root_init
) {
1736 char sfd
[DECIMAL_STR_MAX(int) + 1];
1738 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1739 * the user didn't specify an explicit init to spawn. */
1741 assert(arg_serialization
);
1744 xsprintf(sfd
, "%i", fileno(arg_serialization
));
1747 args
[i
++] = SYSTEMD_BINARY_PATH
;
1748 if (switch_root_dir
)
1749 args
[i
++] = "--switched-root";
1750 args
[i
++] = arg_system
? "--system" : "--user";
1751 args
[i
++] = "--deserialize";
1755 assert(i
<= args_size
);
1758 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do this is on
1759 * its own on exec(), but it will do it on exit(). Hence, to ensure we get a summary here, fork() off
1760 * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1761 * before proceeding into the exec().
1763 valgrind_summary_hack();
1765 (void) execv(args
[0], (char* const*) args
);
1766 log_debug_errno(errno
, "Failed to execute our own binary, trying fallback: %m");
1769 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1770 * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1771 * doesn't matter.) */
1773 arg_serialization
= safe_fclose(arg_serialization
);
1774 fds
= fdset_free(fds
);
1776 /* Reopen the console */
1777 (void) make_console_stdio();
1779 for (j
= 1, i
= 1; j
< (unsigned) argc
; j
++)
1780 args
[i
++] = argv
[j
];
1782 assert(i
<= args_size
);
1784 /* Reenable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1785 (void) reset_all_signal_handlers();
1786 (void) reset_signal_mask();
1787 (void) rlimit_nofile_safe();
1789 if (switch_root_init
) {
1790 args
[0] = switch_root_init
;
1791 (void) execv(args
[0], (char* const*) args
);
1792 log_warning_errno(errno
, "Failed to execute configured init, trying fallback: %m");
1795 args
[0] = "/sbin/init";
1796 (void) execv(args
[0], (char* const*) args
);
1799 manager_status_printf(NULL
, STATUS_TYPE_EMERGENCY
,
1800 ANSI_HIGHLIGHT_RED
" !! " ANSI_NORMAL
,
1801 "Failed to execute /sbin/init");
1804 log_warning("No /sbin/init, trying fallback");
1806 args
[0] = "/bin/sh";
1808 (void) execv(args
[0], (char* const*) args
);
1809 log_error_errno(errno
, "Failed to execute /bin/sh, giving up: %m");
1811 log_warning_errno(r
, "Failed to execute /sbin/init, giving up: %m");
1813 *ret_error_message
= "Failed to execute fallback shell";
1816 static int invoke_main_loop(
1818 bool *ret_reexecute
,
1819 int *ret_retval
, /* Return parameters relevant for shutting down */
1820 const char **ret_shutdown_verb
, /* … */
1821 FDSet
**ret_fds
, /* Return parameters for reexecuting */
1822 char **ret_switch_root_dir
, /* … */
1823 char **ret_switch_root_init
, /* … */
1824 const char **ret_error_message
) {
1829 assert(ret_reexecute
);
1831 assert(ret_shutdown_verb
);
1833 assert(ret_switch_root_dir
);
1834 assert(ret_switch_root_init
);
1835 assert(ret_error_message
);
1838 r
= manager_loop(m
);
1840 *ret_error_message
= "Failed to run main loop";
1841 return log_emergency_errno(r
, "Failed to run main loop: %m");
1844 switch ((ManagerObjective
) r
) {
1846 case MANAGER_RELOAD
: {
1847 LogTarget saved_log_target
;
1848 int saved_log_level
;
1850 log_info("Reloading.");
1852 /* First, save any overridden log level/target, then parse the configuration file, which might
1853 * change the log level to new settings. */
1855 saved_log_level
= m
->log_level_overridden
? log_get_max_level() : -1;
1856 saved_log_target
= m
->log_target_overridden
? log_get_target() : _LOG_TARGET_INVALID
;
1858 r
= parse_config_file();
1860 log_warning_errno(r
, "Failed to parse config file, ignoring: %m");
1862 set_manager_defaults(m
);
1864 if (saved_log_level
>= 0)
1865 manager_override_log_level(m
, saved_log_level
);
1866 if (saved_log_target
>= 0)
1867 manager_override_log_target(m
, saved_log_target
);
1869 r
= manager_reload(m
);
1871 /* Reloading failed before the point of no return. Let's continue running as if nothing happened. */
1872 m
->objective
= MANAGER_OK
;
1877 case MANAGER_REEXECUTE
:
1879 r
= prepare_reexecute(m
, &arg_serialization
, ret_fds
, false);
1881 *ret_error_message
= "Failed to prepare for reexecution";
1885 log_notice("Reexecuting.");
1887 *ret_reexecute
= true;
1888 *ret_retval
= EXIT_SUCCESS
;
1889 *ret_shutdown_verb
= NULL
;
1890 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
1894 case MANAGER_SWITCH_ROOT
:
1895 if (!m
->switch_root_init
) {
1896 r
= prepare_reexecute(m
, &arg_serialization
, ret_fds
, true);
1898 *ret_error_message
= "Failed to prepare for reexecution";
1904 log_notice("Switching root.");
1906 *ret_reexecute
= true;
1907 *ret_retval
= EXIT_SUCCESS
;
1908 *ret_shutdown_verb
= NULL
;
1910 /* Steal the switch root parameters */
1911 *ret_switch_root_dir
= TAKE_PTR(m
->switch_root
);
1912 *ret_switch_root_init
= TAKE_PTR(m
->switch_root_init
);
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
, NULL
, &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
, NULL
, &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 */
2382 save_argc_argv(argc
, argv
);
2384 /* Make sure that if the user says "syslog" we actually log to the journal. */
2385 log_set_upgrade_syslog_to_journal(true);
2387 if (getpid_cached() == 1) {
2388 /* When we run as PID 1 force system mode */
2391 /* Disable the umask logic */
2394 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2395 * activated yet (even though the log socket for it exists). */
2396 log_set_prohibit_ipc(true);
2398 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2399 * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2400 * child process right before execve()'ing the actual binary, at a point in time where socket
2401 * activation stderr/stdout area already set up. */
2402 log_set_always_reopen_console(true);
2404 if (detect_container() <= 0) {
2406 /* Running outside of a container as PID 1 */
2407 log_set_target(LOG_TARGET_KMSG
);
2411 initrd_timestamp
= userspace_timestamp
;
2414 r
= mount_setup_early();
2416 error_message
= "Failed to mount early API filesystems";
2420 r
= initialize_security(
2422 &security_start_timestamp
,
2423 &security_finish_timestamp
,
2429 if (mac_selinux_init() < 0) {
2430 error_message
= "Failed to initialize SELinux policy";
2437 /* Set the default for later on, but don't actually open the logs like this for now. Note that
2438 * if we are transitioning from the initrd there might still be journal fd open, and we
2439 * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2441 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG
);
2444 /* Running inside a container, as PID 1 */
2445 log_set_target(LOG_TARGET_CONSOLE
);
2448 /* For later on, see above... */
2449 log_set_target(LOG_TARGET_JOURNAL
);
2451 /* clear the kernel timestamp,
2452 * because we are in a container */
2453 kernel_timestamp
= DUAL_TIMESTAMP_NULL
;
2456 initialize_coredump(skip_setup
);
2458 r
= fixup_environment();
2460 log_emergency_errno(r
, "Failed to fix up PID 1 environment: %m");
2461 error_message
= "Failed to fix up PID1 environment";
2466 /* Running as user instance */
2468 log_set_target(LOG_TARGET_AUTO
);
2471 /* clear the kernel timestamp,
2472 * because we are not PID 1 */
2473 kernel_timestamp
= DUAL_TIMESTAMP_NULL
;
2477 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2478 * they never log into the console. */
2479 log_show_color(colors_enabled());
2481 r
= make_null_stdio();
2483 log_warning_errno(r
, "Failed to redirect standard streams to /dev/null, ignoring: %m");
2486 /* Mount /proc, /sys and friends, so that /proc/cmdline and
2487 * /proc/$PID/fd is available. */
2488 if (getpid_cached() == 1) {
2490 /* Load the kernel modules early. */
2494 r
= mount_setup(loaded_policy
);
2496 error_message
= "Failed to mount API filesystems";
2501 /* Reset all signal handlers. */
2502 (void) reset_all_signal_handlers();
2503 (void) ignore_signals(SIGNALS_IGNORE
, -1);
2505 r
= load_configuration(argc
, argv
, &error_message
);
2509 r
= safety_checks();
2513 if (IN_SET(arg_action
, ACTION_TEST
, ACTION_HELP
, ACTION_DUMP_CONFIGURATION_ITEMS
, ACTION_DUMP_BUS_PROPERTIES
))
2514 (void) pager_open(arg_pager_flags
);
2516 if (arg_action
!= ACTION_RUN
)
2519 if (arg_action
== ACTION_HELP
) {
2520 retval
= help() < 0 ? EXIT_FAILURE
: EXIT_SUCCESS
;
2522 } else if (arg_action
== ACTION_VERSION
) {
2525 } else if (arg_action
== ACTION_DUMP_CONFIGURATION_ITEMS
) {
2526 unit_dump_config_items(stdout
);
2527 retval
= EXIT_SUCCESS
;
2529 } else if (arg_action
== ACTION_DUMP_BUS_PROPERTIES
) {
2530 dump_bus_properties(stdout
);
2531 retval
= EXIT_SUCCESS
;
2535 assert_se(IN_SET(arg_action
, ACTION_RUN
, ACTION_TEST
));
2537 /* Move out of the way, so that we won't block unmounts */
2538 assert_se(chdir("/") == 0);
2540 if (arg_action
== ACTION_RUN
) {
2542 /* A core pattern might have been specified via the cmdline. */
2543 initialize_core_pattern(skip_setup
);
2545 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2548 /* Remember open file descriptors for later deserialization */
2549 r
= collect_fds(&fds
, &error_message
);
2553 /* Give up any control of the console, but make sure its initialized. */
2554 setup_console_terminal(skip_setup
);
2556 /* Open the logging devices, if possible and necessary */
2560 log_execution_mode(&first_boot
);
2562 r
= initialize_runtime(skip_setup
,
2563 &saved_rlimit_nofile
,
2564 &saved_rlimit_memlock
,
2569 r
= manager_new(arg_system
? UNIT_FILE_SYSTEM
: UNIT_FILE_USER
,
2570 arg_action
== ACTION_TEST
? MANAGER_TEST_FULL
: 0,
2573 log_emergency_errno(r
, "Failed to allocate manager object: %m");
2574 error_message
= "Failed to allocate manager object";
2578 m
->timestamps
[MANAGER_TIMESTAMP_KERNEL
] = kernel_timestamp
;
2579 m
->timestamps
[MANAGER_TIMESTAMP_INITRD
] = initrd_timestamp
;
2580 m
->timestamps
[MANAGER_TIMESTAMP_USERSPACE
] = userspace_timestamp
;
2581 m
->timestamps
[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START
)] = security_start_timestamp
;
2582 m
->timestamps
[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH
)] = security_finish_timestamp
;
2584 set_manager_defaults(m
);
2585 set_manager_settings(m
);
2586 manager_set_first_boot(m
, first_boot
);
2588 /* Remember whether we should queue the default job */
2589 queue_default_job
= !arg_serialization
|| arg_switched_root
;
2591 before_startup
= now(CLOCK_MONOTONIC
);
2593 r
= manager_startup(m
, arg_serialization
, fds
);
2595 error_message
= "Failed to start up manager";
2599 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2600 fds
= fdset_free(fds
);
2601 arg_serialization
= safe_fclose(arg_serialization
);
2603 if (queue_default_job
) {
2604 r
= do_queue_default_job(m
, &error_message
);
2609 after_startup
= now(CLOCK_MONOTONIC
);
2611 log_full(arg_action
== ACTION_TEST
? LOG_INFO
: LOG_DEBUG
,
2612 "Loaded units and determined initial transaction in %s.",
2613 format_timespan(timespan
, sizeof(timespan
), after_startup
- before_startup
, 100 * USEC_PER_MSEC
));
2615 if (arg_action
== ACTION_TEST
) {
2617 retval
= EXIT_SUCCESS
;
2621 (void) invoke_main_loop(m
,
2634 arg_shutdown_watchdog
= m
->shutdown_watchdog
;
2635 m
= manager_free(m
);
2639 mac_selinux_finish();
2642 do_reexecute(argc
, argv
,
2643 &saved_rlimit_nofile
,
2644 &saved_rlimit_memlock
,
2648 &error_message
); /* This only returns if reexecution failed */
2650 arg_serialization
= safe_fclose(arg_serialization
);
2651 fds
= fdset_free(fds
);
2653 #if HAVE_VALGRIND_VALGRIND_H
2654 /* If we are PID 1 and running under valgrind, then let's exit
2655 * here explicitly. valgrind will only generate nice output on
2656 * exit(), not on exec(), hence let's do the former not the
2658 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND
) {
2659 /* Cleanup watchdog_device strings for valgrind. We need them
2660 * in become_shutdown() so normally we cannot free them yet. */
2661 watchdog_free_device();
2662 arg_watchdog_device
= mfree(arg_watchdog_device
);
2667 #if HAS_FEATURE_ADDRESS_SANITIZER
2668 __lsan_do_leak_check();
2671 if (shutdown_verb
) {
2672 r
= become_shutdown(shutdown_verb
, retval
);
2673 log_error_errno(r
, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
2674 error_message
= "Failed to execute shutdown binary";
2677 watchdog_free_device();
2678 arg_watchdog_device
= mfree(arg_watchdog_device
);
2680 if (getpid_cached() == 1) {
2682 manager_status_printf(NULL
, STATUS_TYPE_EMERGENCY
,
2683 ANSI_HIGHLIGHT_RED
"!!!!!!" ANSI_NORMAL
,
2684 "%s.", error_message
);
2685 freeze_or_exit_or_reboot();