1 /* SPDX-License-Identifier: LGPL-2.1+ */
10 #include <sys/prctl.h>
11 #include <sys/reboot.h>
17 #if HAVE_VALGRIND_VALGRIND_H
18 #include <valgrind/valgrind.h>
22 #include "sd-daemon.h"
23 #include "sd-messages.h"
25 #include "alloc-util.h"
26 #include "architecture.h"
28 #include "bus-error.h"
30 #include "capability-util.h"
31 #include "cgroup-util.h"
32 #include "clock-util.h"
33 #include "conf-parser.h"
34 #include "cpu-set-util.h"
36 #include "dbus-manager.h"
38 #include "emergency-action.h"
40 #include "exit-status.h"
44 #include "format-util.h"
46 #include "hostname-setup.h"
47 #include "ima-setup.h"
49 #include "kmod-setup.h"
50 #include "limits-util.h"
51 #include "load-fragment.h"
53 #include "loopback-setup.h"
54 #include "machine-id-setup.h"
57 #include "mount-setup.h"
60 #include "parse-util.h"
61 #include "path-util.h"
62 #include "pretty-print.h"
63 #include "proc-cmdline.h"
64 #include "process-util.h"
65 #include "raw-clone.h"
66 #include "rlimit-util.h"
68 #include "seccomp-util.h"
70 #include "selinux-setup.h"
71 #include "selinux-util.h"
72 #include "signal-util.h"
73 #include "smack-setup.h"
75 #include "stat-util.h"
76 #include "stdio-util.h"
78 #include "switch-root.h"
79 #include "sysctl-util.h"
80 #include "terminal-util.h"
81 #include "umask-util.h"
82 #include "user-util.h"
87 #if HAS_FEATURE_ADDRESS_SANITIZER
88 #include <sanitizer/lsan_interface.h>
96 ACTION_DUMP_CONFIGURATION_ITEMS
,
97 ACTION_DUMP_BUS_PROPERTIES
,
98 } arg_action
= ACTION_RUN
;
100 /* Those variables are initalized to 0 automatically, so we avoid uninitialized memory access.
101 * Real defaults are assigned in reset_arguments() below. */
102 static char *arg_default_unit
;
103 static bool arg_system
;
104 static bool arg_dump_core
;
105 static int arg_crash_chvt
;
106 static bool arg_crash_shell
;
107 static bool arg_crash_reboot
;
108 static char *arg_confirm_spawn
;
109 static ShowStatus arg_show_status
;
110 static bool arg_switched_root
;
111 static PagerFlags arg_pager_flags
;
112 static bool arg_service_watchdogs
;
113 static ExecOutput arg_default_std_output
;
114 static ExecOutput arg_default_std_error
;
115 static usec_t arg_default_restart_usec
;
116 static usec_t arg_default_timeout_start_usec
;
117 static usec_t arg_default_timeout_stop_usec
;
118 static usec_t arg_default_timeout_abort_usec
;
119 static bool arg_default_timeout_abort_set
;
120 static usec_t arg_default_start_limit_interval
;
121 static unsigned arg_default_start_limit_burst
;
122 static usec_t arg_runtime_watchdog
;
123 static usec_t arg_shutdown_watchdog
;
124 static char *arg_early_core_pattern
;
125 static char *arg_watchdog_device
;
126 static char **arg_default_environment
;
127 static struct rlimit
*arg_default_rlimit
[_RLIMIT_MAX
];
128 static uint64_t arg_capability_bounding_set
;
129 static bool arg_no_new_privs
;
130 static nsec_t arg_timer_slack_nsec
;
131 static usec_t arg_default_timer_accuracy_usec
;
132 static Set
* arg_syscall_archs
;
133 static FILE* arg_serialization
;
134 static int arg_default_cpu_accounting
;
135 static bool arg_default_io_accounting
;
136 static bool arg_default_ip_accounting
;
137 static bool arg_default_blockio_accounting
;
138 static bool arg_default_memory_accounting
;
139 static bool arg_default_tasks_accounting
;
140 static uint64_t arg_default_tasks_max
;
141 static sd_id128_t arg_machine_id
;
142 static EmergencyAction arg_cad_burst_action
;
143 static OOMPolicy arg_default_oom_policy
;
144 static CPUSet arg_cpu_affinity
;
145 static NUMAPolicy arg_numa_policy
;
147 static int parse_configuration(void);
149 _noreturn_
static void freeze_or_exit_or_reboot(void) {
151 /* If we are running in a container, let's prefer exiting, after all we can propagate an exit code to
152 * the container manager, and thus inform it that something went wrong. */
153 if (detect_container() > 0) {
154 log_emergency("Exiting PID 1...");
155 _exit(EXIT_EXCEPTION
);
158 if (arg_crash_reboot
) {
159 log_notice("Rebooting in 10s...");
162 log_notice("Rebooting now...");
163 (void) reboot(RB_AUTOBOOT
);
164 log_emergency_errno(errno
, "Failed to reboot: %m");
167 log_emergency("Freezing execution.");
171 _noreturn_
static void crash(int sig
) {
175 if (getpid_cached() != 1)
176 /* Pass this on immediately, if this is not PID 1 */
178 else if (!arg_dump_core
)
179 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig
));
181 sa
= (struct sigaction
) {
182 .sa_handler
= nop_signal_handler
,
183 .sa_flags
= SA_NOCLDSTOP
|SA_RESTART
,
186 /* We want to wait for the core process, hence let's enable SIGCHLD */
187 (void) sigaction(SIGCHLD
, &sa
, NULL
);
189 pid
= raw_clone(SIGCHLD
);
191 log_emergency_errno(errno
, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig
));
193 /* Enable default signal handler for core dump */
195 sa
= (struct sigaction
) {
196 .sa_handler
= SIG_DFL
,
198 (void) sigaction(sig
, &sa
, NULL
);
200 /* Don't limit the coredump size */
201 (void) setrlimit(RLIMIT_CORE
, &RLIMIT_MAKE_CONST(RLIM_INFINITY
));
203 /* Just to be sure... */
206 /* Raise the signal again */
208 (void) kill(pid
, sig
); /* raise() would kill the parent */
210 assert_not_reached("We shouldn't be here...");
211 _exit(EXIT_EXCEPTION
);
216 /* Order things nicely. */
217 r
= wait_for_terminate(pid
, &status
);
219 log_emergency_errno(r
, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig
));
220 else if (status
.si_code
!= CLD_DUMPED
)
221 log_emergency("Caught <%s>, core dump failed (child "PID_FMT
", code=%s, status=%i/%s).",
222 signal_to_string(sig
),
223 pid
, sigchld_code_to_string(status
.si_code
),
225 strna(status
.si_code
== CLD_EXITED
226 ? exit_status_to_string(status
.si_status
, EXIT_STATUS_MINIMAL
)
227 : signal_to_string(status
.si_status
)));
229 log_emergency("Caught <%s>, dumped core as pid "PID_FMT
".", signal_to_string(sig
), pid
);
233 if (arg_crash_chvt
>= 0)
234 (void) chvt(arg_crash_chvt
);
236 sa
= (struct sigaction
) {
237 .sa_handler
= SIG_IGN
,
238 .sa_flags
= SA_NOCLDSTOP
|SA_NOCLDWAIT
|SA_RESTART
,
241 /* Let the kernel reap children for us */
242 (void) sigaction(SIGCHLD
, &sa
, NULL
);
244 if (arg_crash_shell
) {
245 log_notice("Executing crash shell in 10s...");
248 pid
= raw_clone(SIGCHLD
);
250 log_emergency_errno(errno
, "Failed to fork off crash shell: %m");
253 (void) make_console_stdio();
254 (void) rlimit_nofile_safe();
255 (void) execle("/bin/sh", "/bin/sh", NULL
, environ
);
257 log_emergency_errno(errno
, "execle() failed: %m");
258 _exit(EXIT_EXCEPTION
);
260 log_info("Spawned crash shell as PID "PID_FMT
".", pid
);
261 (void) wait_for_terminate(pid
, NULL
);
265 freeze_or_exit_or_reboot();
268 static void install_crash_handler(void) {
269 static const struct sigaction sa
= {
271 .sa_flags
= SA_NODEFER
, /* So that we can raise the signal again from the signal handler */
275 /* We ignore the return value here, since, we don't mind if we
276 * cannot set up a crash handler */
277 r
= sigaction_many(&sa
, SIGNALS_CRASH_HANDLER
, -1);
279 log_debug_errno(r
, "I had trouble setting up the crash handler, ignoring: %m");
282 static int console_setup(void) {
283 _cleanup_close_
int tty_fd
= -1;
286 tty_fd
= open_terminal("/dev/console", O_WRONLY
|O_NOCTTY
|O_CLOEXEC
);
288 return log_error_errno(tty_fd
, "Failed to open /dev/console: %m");
290 /* We don't want to force text mode. plymouth may be showing
291 * pictures already from initrd. */
292 r
= reset_terminal_fd(tty_fd
, false);
294 return log_error_errno(r
, "Failed to reset /dev/console: %m");
299 static int parse_crash_chvt(const char *value
) {
302 if (safe_atoi(value
, &arg_crash_chvt
) >= 0)
305 b
= parse_boolean(value
);
310 arg_crash_chvt
= 0; /* switch to where kmsg goes */
312 arg_crash_chvt
= -1; /* turn off switching */
317 static int parse_confirm_spawn(const char *value
, char **console
) {
321 r
= value
? parse_boolean(value
) : 1;
327 if (r
> 0) /* on with default tty */
328 s
= strdup("/dev/console");
329 else if (is_path(value
)) /* on with fully qualified path */
331 else /* on with only a tty file name, not a fully qualified path */
332 s
= path_join("/dev", value
);
340 static int set_machine_id(const char *m
) {
344 if (sd_id128_from_string(m
, &t
) < 0)
347 if (sd_id128_is_null(t
))
354 static int parse_proc_cmdline_item(const char *key
, const char *value
, void *data
) {
360 if (STR_IN_SET(key
, "systemd.unit", "rd.systemd.unit")) {
362 if (proc_cmdline_value_missing(key
, value
))
365 if (!unit_name_is_valid(value
, UNIT_NAME_PLAIN
|UNIT_NAME_INSTANCE
))
366 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key
, value
);
367 else if (in_initrd() == !!startswith(key
, "rd.")) {
368 if (free_and_strdup(&arg_default_unit
, value
) < 0)
372 } else if (proc_cmdline_key_streq(key
, "systemd.dump_core")) {
374 r
= value
? parse_boolean(value
) : true;
376 log_warning_errno(r
, "Failed to parse dump core switch %s, ignoring: %m", value
);
380 } else if (proc_cmdline_key_streq(key
, "systemd.early_core_pattern")) {
382 if (proc_cmdline_value_missing(key
, value
))
385 if (path_is_absolute(value
))
386 (void) parse_path_argument_and_warn(value
, false, &arg_early_core_pattern
);
388 log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value
);
390 } else if (proc_cmdline_key_streq(key
, "systemd.crash_chvt")) {
393 arg_crash_chvt
= 0; /* turn on */
395 r
= parse_crash_chvt(value
);
397 log_warning_errno(r
, "Failed to parse crash chvt switch %s, ignoring: %m", value
);
400 } else if (proc_cmdline_key_streq(key
, "systemd.crash_shell")) {
402 r
= value
? parse_boolean(value
) : true;
404 log_warning_errno(r
, "Failed to parse crash shell switch %s, ignoring: %m", value
);
408 } else if (proc_cmdline_key_streq(key
, "systemd.crash_reboot")) {
410 r
= value
? parse_boolean(value
) : true;
412 log_warning_errno(r
, "Failed to parse crash reboot switch %s, ignoring: %m", value
);
414 arg_crash_reboot
= r
;
416 } else if (proc_cmdline_key_streq(key
, "systemd.confirm_spawn")) {
419 r
= parse_confirm_spawn(value
, &s
);
421 log_warning_errno(r
, "Failed to parse confirm_spawn switch %s, ignoring: %m", value
);
423 free_and_replace(arg_confirm_spawn
, s
);
425 } else if (proc_cmdline_key_streq(key
, "systemd.service_watchdogs")) {
427 r
= value
? parse_boolean(value
) : true;
429 log_warning_errno(r
, "Failed to parse service watchdog switch %s, ignoring: %m", value
);
431 arg_service_watchdogs
= r
;
433 } else if (proc_cmdline_key_streq(key
, "systemd.show_status")) {
436 r
= parse_show_status(value
, &arg_show_status
);
438 log_warning_errno(r
, "Failed to parse show status switch %s, ignoring: %m", value
);
440 arg_show_status
= SHOW_STATUS_YES
;
442 } else if (proc_cmdline_key_streq(key
, "systemd.default_standard_output")) {
444 if (proc_cmdline_value_missing(key
, value
))
447 r
= exec_output_from_string(value
);
449 log_warning_errno(r
, "Failed to parse default standard output switch %s, ignoring: %m", value
);
451 arg_default_std_output
= r
;
453 } else if (proc_cmdline_key_streq(key
, "systemd.default_standard_error")) {
455 if (proc_cmdline_value_missing(key
, value
))
458 r
= exec_output_from_string(value
);
460 log_warning_errno(r
, "Failed to parse default standard error switch %s, ignoring: %m", value
);
462 arg_default_std_error
= r
;
464 } else if (streq(key
, "systemd.setenv")) {
466 if (proc_cmdline_value_missing(key
, value
))
469 if (env_assignment_is_valid(value
)) {
472 env
= strv_env_set(arg_default_environment
, value
);
476 arg_default_environment
= env
;
478 log_warning("Environment variable name '%s' is not valid. Ignoring.", value
);
480 } else if (proc_cmdline_key_streq(key
, "systemd.machine_id")) {
482 if (proc_cmdline_value_missing(key
, value
))
485 r
= set_machine_id(value
);
487 log_warning_errno(r
, "MachineID '%s' is not valid, ignoring: %m", value
);
489 } else if (proc_cmdline_key_streq(key
, "systemd.default_timeout_start_sec")) {
491 if (proc_cmdline_value_missing(key
, value
))
494 r
= parse_sec(value
, &arg_default_timeout_start_usec
);
496 log_warning_errno(r
, "Failed to parse default start timeout '%s', ignoring: %m", value
);
498 if (arg_default_timeout_start_usec
<= 0)
499 arg_default_timeout_start_usec
= USEC_INFINITY
;
501 } else if (proc_cmdline_key_streq(key
, "systemd.watchdog_device")) {
503 if (proc_cmdline_value_missing(key
, value
))
506 (void) parse_path_argument_and_warn(value
, false, &arg_watchdog_device
);
508 } else if (streq(key
, "quiet") && !value
) {
510 if (arg_show_status
== _SHOW_STATUS_INVALID
)
511 arg_show_status
= SHOW_STATUS_AUTO
;
513 } else if (streq(key
, "debug") && !value
) {
515 /* Note that log_parse_environment() handles 'debug'
516 * too, and sets the log level to LOG_DEBUG. */
518 if (detect_container() > 0)
519 log_set_target(LOG_TARGET_CONSOLE
);
524 /* SysV compatibility */
525 target
= runlevel_to_target(key
);
527 return free_and_strdup(&arg_default_unit
, target
);
533 #define DEFINE_SETTER(name, func, descr) \
534 static int name(const char *unit, \
535 const char *filename, \
537 const char *section, \
538 unsigned section_line, \
539 const char *lvalue, \
541 const char *rvalue, \
553 log_syntax(unit, LOG_ERR, filename, line, r, \
554 "Invalid " descr "'%s': %m", \
560 DEFINE_SETTER(config_parse_level2
, log_set_max_level_from_string
, "log level");
561 DEFINE_SETTER(config_parse_target
, log_set_target_from_string
, "target");
562 DEFINE_SETTER(config_parse_color
, log_show_color_from_string
, "color" );
563 DEFINE_SETTER(config_parse_location
, log_show_location_from_string
, "location");
565 static int config_parse_cpu_affinity2(
567 const char *filename
,
570 unsigned section_line
,
577 CPUSet
*affinity
= data
;
581 (void) parse_cpu_set_extend(rvalue
, affinity
, true, unit
, filename
, line
, lvalue
);
586 static int config_parse_show_status(
588 const char *filename
,
591 unsigned section_line
,
599 ShowStatus
*b
= data
;
606 k
= parse_show_status(rvalue
, b
);
608 log_syntax(unit
, LOG_ERR
, filename
, line
, k
, "Failed to parse show status setting, ignoring: %s", rvalue
);
615 static int config_parse_output_restricted(
617 const char *filename
,
620 unsigned section_line
,
627 ExecOutput t
, *eo
= data
;
634 t
= exec_output_from_string(rvalue
);
636 log_syntax(unit
, LOG_ERR
, filename
, line
, 0, "Failed to parse output type, ignoring: %s", rvalue
);
640 if (IN_SET(t
, EXEC_OUTPUT_SOCKET
, EXEC_OUTPUT_NAMED_FD
, EXEC_OUTPUT_FILE
, EXEC_OUTPUT_FILE_APPEND
)) {
641 log_syntax(unit
, LOG_ERR
, filename
, line
, 0, "Standard output types socket, fd:, file:, append: are not supported as defaults, ignoring: %s", rvalue
);
649 static int config_parse_crash_chvt(
651 const char *filename
,
654 unsigned section_line
,
667 r
= parse_crash_chvt(rvalue
);
669 log_syntax(unit
, LOG_ERR
, filename
, line
, r
, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue
);
676 static int config_parse_timeout_abort(
678 const char *filename
,
681 unsigned section_line
,
694 rvalue
+= strspn(rvalue
, WHITESPACE
);
695 if (isempty(rvalue
)) {
696 arg_default_timeout_abort_set
= false;
700 r
= parse_sec(rvalue
, &arg_default_timeout_abort_usec
);
702 log_syntax(unit
, LOG_ERR
, filename
, line
, r
, "Failed to parse DefaultTimeoutAbortSec= setting, ignoring: %s", rvalue
);
706 arg_default_timeout_abort_set
= true;
710 static int parse_config_file(void) {
712 const ConfigTableItem items
[] = {
713 { "Manager", "LogLevel", config_parse_level2
, 0, NULL
},
714 { "Manager", "LogTarget", config_parse_target
, 0, NULL
},
715 { "Manager", "LogColor", config_parse_color
, 0, NULL
},
716 { "Manager", "LogLocation", config_parse_location
, 0, NULL
},
717 { "Manager", "DumpCore", config_parse_bool
, 0, &arg_dump_core
},
718 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt
, 0, NULL
},
719 { "Manager", "CrashChangeVT", config_parse_crash_chvt
, 0, NULL
},
720 { "Manager", "CrashShell", config_parse_bool
, 0, &arg_crash_shell
},
721 { "Manager", "CrashReboot", config_parse_bool
, 0, &arg_crash_reboot
},
722 { "Manager", "ShowStatus", config_parse_show_status
, 0, &arg_show_status
},
723 { "Manager", "CPUAffinity", config_parse_cpu_affinity2
, 0, &arg_cpu_affinity
},
724 { "Manager", "NUMAPolicy", config_parse_numa_policy
, 0, &arg_numa_policy
.type
},
725 { "Manager", "NUMAMask", config_parse_numa_mask
, 0, &arg_numa_policy
},
726 { "Manager", "JoinControllers", config_parse_warn_compat
, DISABLED_CONFIGURATION
, NULL
},
727 { "Manager", "RuntimeWatchdogSec", config_parse_sec
, 0, &arg_runtime_watchdog
},
728 { "Manager", "ShutdownWatchdogSec", config_parse_sec
, 0, &arg_shutdown_watchdog
},
729 { "Manager", "WatchdogDevice", config_parse_path
, 0, &arg_watchdog_device
},
730 { "Manager", "CapabilityBoundingSet", config_parse_capability_set
, 0, &arg_capability_bounding_set
},
731 { "Manager", "NoNewPrivileges", config_parse_bool
, 0, &arg_no_new_privs
},
733 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs
, 0, &arg_syscall_archs
},
735 { "Manager", "TimerSlackNSec", config_parse_nsec
, 0, &arg_timer_slack_nsec
},
736 { "Manager", "DefaultTimerAccuracySec", config_parse_sec
, 0, &arg_default_timer_accuracy_usec
},
737 { "Manager", "DefaultStandardOutput", config_parse_output_restricted
,0, &arg_default_std_output
},
738 { "Manager", "DefaultStandardError", config_parse_output_restricted
,0, &arg_default_std_error
},
739 { "Manager", "DefaultTimeoutStartSec", config_parse_sec
, 0, &arg_default_timeout_start_usec
},
740 { "Manager", "DefaultTimeoutStopSec", config_parse_sec
, 0, &arg_default_timeout_stop_usec
},
741 { "Manager", "DefaultTimeoutAbortSec", config_parse_timeout_abort
, 0, NULL
},
742 { "Manager", "DefaultRestartSec", config_parse_sec
, 0, &arg_default_restart_usec
},
743 { "Manager", "DefaultStartLimitInterval", config_parse_sec
, 0, &arg_default_start_limit_interval
}, /* obsolete alias */
744 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec
, 0, &arg_default_start_limit_interval
},
745 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned
, 0, &arg_default_start_limit_burst
},
746 { "Manager", "DefaultEnvironment", config_parse_environ
, 0, &arg_default_environment
},
747 { "Manager", "DefaultLimitCPU", config_parse_rlimit
, RLIMIT_CPU
, arg_default_rlimit
},
748 { "Manager", "DefaultLimitFSIZE", config_parse_rlimit
, RLIMIT_FSIZE
, arg_default_rlimit
},
749 { "Manager", "DefaultLimitDATA", config_parse_rlimit
, RLIMIT_DATA
, arg_default_rlimit
},
750 { "Manager", "DefaultLimitSTACK", config_parse_rlimit
, RLIMIT_STACK
, arg_default_rlimit
},
751 { "Manager", "DefaultLimitCORE", config_parse_rlimit
, RLIMIT_CORE
, arg_default_rlimit
},
752 { "Manager", "DefaultLimitRSS", config_parse_rlimit
, RLIMIT_RSS
, arg_default_rlimit
},
753 { "Manager", "DefaultLimitNOFILE", config_parse_rlimit
, RLIMIT_NOFILE
, arg_default_rlimit
},
754 { "Manager", "DefaultLimitAS", config_parse_rlimit
, RLIMIT_AS
, arg_default_rlimit
},
755 { "Manager", "DefaultLimitNPROC", config_parse_rlimit
, RLIMIT_NPROC
, arg_default_rlimit
},
756 { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit
, RLIMIT_MEMLOCK
, arg_default_rlimit
},
757 { "Manager", "DefaultLimitLOCKS", config_parse_rlimit
, RLIMIT_LOCKS
, arg_default_rlimit
},
758 { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit
, RLIMIT_SIGPENDING
, arg_default_rlimit
},
759 { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit
, RLIMIT_MSGQUEUE
, arg_default_rlimit
},
760 { "Manager", "DefaultLimitNICE", config_parse_rlimit
, RLIMIT_NICE
, arg_default_rlimit
},
761 { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit
, RLIMIT_RTPRIO
, arg_default_rlimit
},
762 { "Manager", "DefaultLimitRTTIME", config_parse_rlimit
, RLIMIT_RTTIME
, arg_default_rlimit
},
763 { "Manager", "DefaultCPUAccounting", config_parse_tristate
, 0, &arg_default_cpu_accounting
},
764 { "Manager", "DefaultIOAccounting", config_parse_bool
, 0, &arg_default_io_accounting
},
765 { "Manager", "DefaultIPAccounting", config_parse_bool
, 0, &arg_default_ip_accounting
},
766 { "Manager", "DefaultBlockIOAccounting", config_parse_bool
, 0, &arg_default_blockio_accounting
},
767 { "Manager", "DefaultMemoryAccounting", config_parse_bool
, 0, &arg_default_memory_accounting
},
768 { "Manager", "DefaultTasksAccounting", config_parse_bool
, 0, &arg_default_tasks_accounting
},
769 { "Manager", "DefaultTasksMax", config_parse_tasks_max
, 0, &arg_default_tasks_max
},
770 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action
, 0, &arg_cad_burst_action
},
771 { "Manager", "DefaultOOMPolicy", config_parse_oom_policy
, 0, &arg_default_oom_policy
},
775 const char *fn
, *conf_dirs_nulstr
;
778 PKGSYSCONFDIR
"/system.conf" :
779 PKGSYSCONFDIR
"/user.conf";
781 conf_dirs_nulstr
= arg_system
?
782 CONF_PATHS_NULSTR("systemd/system.conf.d") :
783 CONF_PATHS_NULSTR("systemd/user.conf.d");
785 (void) config_parse_many_nulstr(fn
, conf_dirs_nulstr
, "Manager\0", config_item_table_lookup
, items
, CONFIG_PARSE_WARN
, NULL
);
787 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
788 * like everywhere else. */
789 if (arg_default_timeout_start_usec
<= 0)
790 arg_default_timeout_start_usec
= USEC_INFINITY
;
791 if (arg_default_timeout_stop_usec
<= 0)
792 arg_default_timeout_stop_usec
= USEC_INFINITY
;
797 static void set_manager_defaults(Manager
*m
) {
801 /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
802 * affect the manager itself, but are just what newly allocated units will have set if they haven't set
803 * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
805 m
->default_timer_accuracy_usec
= arg_default_timer_accuracy_usec
;
806 m
->default_std_output
= arg_default_std_output
;
807 m
->default_std_error
= arg_default_std_error
;
808 m
->default_timeout_start_usec
= arg_default_timeout_start_usec
;
809 m
->default_timeout_stop_usec
= arg_default_timeout_stop_usec
;
810 m
->default_timeout_abort_usec
= arg_default_timeout_abort_usec
;
811 m
->default_timeout_abort_set
= arg_default_timeout_abort_set
;
812 m
->default_restart_usec
= arg_default_restart_usec
;
813 m
->default_start_limit_interval
= arg_default_start_limit_interval
;
814 m
->default_start_limit_burst
= arg_default_start_limit_burst
;
816 /* On 4.15+ with unified hierarchy, CPU accounting is essentially free as it doesn't require the CPU
817 * controller to be enabled, so the default is to enable it unless we got told otherwise. */
818 if (arg_default_cpu_accounting
>= 0)
819 m
->default_cpu_accounting
= arg_default_cpu_accounting
;
821 m
->default_cpu_accounting
= cpu_accounting_is_cheap();
823 m
->default_io_accounting
= arg_default_io_accounting
;
824 m
->default_ip_accounting
= arg_default_ip_accounting
;
825 m
->default_blockio_accounting
= arg_default_blockio_accounting
;
826 m
->default_memory_accounting
= arg_default_memory_accounting
;
827 m
->default_tasks_accounting
= arg_default_tasks_accounting
;
828 m
->default_tasks_max
= arg_default_tasks_max
;
829 m
->default_oom_policy
= arg_default_oom_policy
;
831 (void) manager_set_default_rlimits(m
, arg_default_rlimit
);
833 (void) manager_default_environment(m
);
834 (void) manager_transient_environment_add(m
, arg_default_environment
);
837 static void set_manager_settings(Manager
*m
) {
841 /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
842 * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
844 m
->confirm_spawn
= arg_confirm_spawn
;
845 m
->service_watchdogs
= arg_service_watchdogs
;
846 m
->runtime_watchdog
= arg_runtime_watchdog
;
847 m
->shutdown_watchdog
= arg_shutdown_watchdog
;
848 m
->cad_burst_action
= arg_cad_burst_action
;
850 manager_set_show_status(m
, arg_show_status
);
853 static int parse_argv(int argc
, char *argv
[]) {
855 ARG_LOG_LEVEL
= 0x100,
865 ARG_DUMP_CONFIGURATION_ITEMS
,
866 ARG_DUMP_BUS_PROPERTIES
,
875 ARG_DEFAULT_STD_OUTPUT
,
876 ARG_DEFAULT_STD_ERROR
,
878 ARG_SERVICE_WATCHDOGS
,
881 static const struct option options
[] = {
882 { "log-level", required_argument
, NULL
, ARG_LOG_LEVEL
},
883 { "log-target", required_argument
, NULL
, ARG_LOG_TARGET
},
884 { "log-color", optional_argument
, NULL
, ARG_LOG_COLOR
},
885 { "log-location", optional_argument
, NULL
, ARG_LOG_LOCATION
},
886 { "unit", required_argument
, NULL
, ARG_UNIT
},
887 { "system", no_argument
, NULL
, ARG_SYSTEM
},
888 { "user", no_argument
, NULL
, ARG_USER
},
889 { "test", no_argument
, NULL
, ARG_TEST
},
890 { "no-pager", no_argument
, NULL
, ARG_NO_PAGER
},
891 { "help", no_argument
, NULL
, 'h' },
892 { "version", no_argument
, NULL
, ARG_VERSION
},
893 { "dump-configuration-items", no_argument
, NULL
, ARG_DUMP_CONFIGURATION_ITEMS
},
894 { "dump-bus-properties", no_argument
, NULL
, ARG_DUMP_BUS_PROPERTIES
},
895 { "dump-core", optional_argument
, NULL
, ARG_DUMP_CORE
},
896 { "crash-chvt", required_argument
, NULL
, ARG_CRASH_CHVT
},
897 { "crash-shell", optional_argument
, NULL
, ARG_CRASH_SHELL
},
898 { "crash-reboot", optional_argument
, NULL
, ARG_CRASH_REBOOT
},
899 { "confirm-spawn", optional_argument
, NULL
, ARG_CONFIRM_SPAWN
},
900 { "show-status", optional_argument
, NULL
, ARG_SHOW_STATUS
},
901 { "deserialize", required_argument
, NULL
, ARG_DESERIALIZE
},
902 { "switched-root", no_argument
, NULL
, ARG_SWITCHED_ROOT
},
903 { "default-standard-output", required_argument
, NULL
, ARG_DEFAULT_STD_OUTPUT
, },
904 { "default-standard-error", required_argument
, NULL
, ARG_DEFAULT_STD_ERROR
, },
905 { "machine-id", required_argument
, NULL
, ARG_MACHINE_ID
},
906 { "service-watchdogs", required_argument
, NULL
, ARG_SERVICE_WATCHDOGS
},
915 if (getpid_cached() == 1)
918 while ((c
= getopt_long(argc
, argv
, "hDbsz:", options
, NULL
)) >= 0)
923 r
= log_set_max_level_from_string(optarg
);
925 return log_error_errno(r
, "Failed to parse log level \"%s\": %m", optarg
);
930 r
= log_set_target_from_string(optarg
);
932 return log_error_errno(r
, "Failed to parse log target \"%s\": %m", optarg
);
939 r
= log_show_color_from_string(optarg
);
941 return log_error_errno(r
, "Failed to parse log color setting \"%s\": %m",
944 log_show_color(true);
948 case ARG_LOG_LOCATION
:
950 r
= log_show_location_from_string(optarg
);
952 return log_error_errno(r
, "Failed to parse log location setting \"%s\": %m",
955 log_show_location(true);
959 case ARG_DEFAULT_STD_OUTPUT
:
960 r
= exec_output_from_string(optarg
);
962 return log_error_errno(r
, "Failed to parse default standard output setting \"%s\": %m",
964 arg_default_std_output
= r
;
967 case ARG_DEFAULT_STD_ERROR
:
968 r
= exec_output_from_string(optarg
);
970 return log_error_errno(r
, "Failed to parse default standard error output setting \"%s\": %m",
972 arg_default_std_error
= r
;
976 r
= free_and_strdup(&arg_default_unit
, optarg
);
978 return log_error_errno(r
, "Failed to set default unit \"%s\": %m", optarg
);
991 arg_action
= ACTION_TEST
;
995 arg_pager_flags
|= PAGER_DISABLE
;
999 arg_action
= ACTION_VERSION
;
1002 case ARG_DUMP_CONFIGURATION_ITEMS
:
1003 arg_action
= ACTION_DUMP_CONFIGURATION_ITEMS
;
1006 case ARG_DUMP_BUS_PROPERTIES
:
1007 arg_action
= ACTION_DUMP_BUS_PROPERTIES
;
1012 arg_dump_core
= true;
1014 r
= parse_boolean(optarg
);
1016 return log_error_errno(r
, "Failed to parse dump core boolean: \"%s\": %m",
1022 case ARG_CRASH_CHVT
:
1023 r
= parse_crash_chvt(optarg
);
1025 return log_error_errno(r
, "Failed to parse crash virtual terminal index: \"%s\": %m",
1029 case ARG_CRASH_SHELL
:
1031 arg_crash_shell
= true;
1033 r
= parse_boolean(optarg
);
1035 return log_error_errno(r
, "Failed to parse crash shell boolean: \"%s\": %m",
1037 arg_crash_shell
= r
;
1041 case ARG_CRASH_REBOOT
:
1043 arg_crash_reboot
= true;
1045 r
= parse_boolean(optarg
);
1047 return log_error_errno(r
, "Failed to parse crash shell boolean: \"%s\": %m",
1049 arg_crash_reboot
= r
;
1053 case ARG_CONFIRM_SPAWN
:
1054 arg_confirm_spawn
= mfree(arg_confirm_spawn
);
1056 r
= parse_confirm_spawn(optarg
, &arg_confirm_spawn
);
1058 return log_error_errno(r
, "Failed to parse confirm spawn option: \"%s\": %m",
1062 case ARG_SERVICE_WATCHDOGS
:
1063 r
= parse_boolean(optarg
);
1065 return log_error_errno(r
, "Failed to parse service watchdogs boolean: \"%s\": %m",
1067 arg_service_watchdogs
= r
;
1070 case ARG_SHOW_STATUS
:
1072 r
= parse_show_status(optarg
, &arg_show_status
);
1074 return log_error_errno(r
, "Failed to parse show status boolean: \"%s\": %m",
1077 arg_show_status
= SHOW_STATUS_YES
;
1080 case ARG_DESERIALIZE
: {
1084 r
= safe_atoi(optarg
, &fd
);
1086 log_error_errno(r
, "Failed to parse deserialize option \"%s\": %m", optarg
);
1088 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
),
1089 "Invalid deserialize fd: %d",
1092 (void) fd_cloexec(fd
, true);
1094 f
= fdopen(fd
, "r");
1096 return log_error_errno(errno
, "Failed to open serialization fd %d: %m", fd
);
1098 safe_fclose(arg_serialization
);
1099 arg_serialization
= f
;
1104 case ARG_SWITCHED_ROOT
:
1105 arg_switched_root
= true;
1108 case ARG_MACHINE_ID
:
1109 r
= set_machine_id(optarg
);
1111 return log_error_errno(r
, "MachineID '%s' is not valid: %m", optarg
);
1115 arg_action
= ACTION_HELP
;
1119 log_set_max_level(LOG_DEBUG
);
1125 /* Just to eat away the sysvinit kernel
1126 * cmdline args without getopt() error
1127 * messages that we'll parse in
1128 * parse_proc_cmdline_word() or ignore. */
1131 if (getpid_cached() != 1)
1137 assert_not_reached("Unhandled option code.");
1140 if (optind
< argc
&& getpid_cached() != 1) {
1141 /* Hmm, when we aren't run as init system
1142 * let's complain about excess arguments */
1144 return log_error_errno(SYNTHETIC_ERRNO(EINVAL
),
1145 "Excess arguments.");
1151 static int help(void) {
1152 _cleanup_free_
char *link
= NULL
;
1155 r
= terminal_urlify_man("systemd", "1", &link
);
1159 printf("%s [OPTIONS...]\n\n"
1160 "Starts up and maintains the system or user services.\n\n"
1161 " -h --help Show this help\n"
1162 " --version Show version\n"
1163 " --test Determine initial transaction, dump it and exit\n"
1164 " --system In combination with --test: operate as system service manager\n"
1165 " --user In combination with --test: operate as per-user service manager\n"
1166 " --no-pager Do not pipe output into a pager\n"
1167 " --dump-configuration-items Dump understood unit configuration items\n"
1168 " --dump-bus-properties Dump exposed bus properties\n"
1169 " --unit=UNIT Set default unit\n"
1170 " --dump-core[=BOOL] Dump core on crash\n"
1171 " --crash-vt=NR Change to specified VT on crash\n"
1172 " --crash-reboot[=BOOL] Reboot on crash\n"
1173 " --crash-shell[=BOOL] Run shell on crash\n"
1174 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1175 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1176 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1177 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1178 " --log-color[=BOOL] Highlight important log messages\n"
1179 " --log-location[=BOOL] Include code location in log messages\n"
1180 " --default-standard-output= Set default standard output for services\n"
1181 " --default-standard-error= Set default standard error output for services\n"
1182 "\nSee the %s for details.\n"
1183 , program_invocation_short_name
1190 static int prepare_reexecute(
1194 bool switching_root
) {
1196 _cleanup_fdset_free_ FDSet
*fds
= NULL
;
1197 _cleanup_fclose_
FILE *f
= NULL
;
1204 r
= manager_open_serialization(m
, &f
);
1206 return log_error_errno(r
, "Failed to create serialization file: %m");
1208 /* Make sure nothing is really destructed when we shut down */
1210 bus_manager_send_reloading(m
, true);
1216 r
= manager_serialize(m
, f
, fds
, switching_root
);
1220 if (fseeko(f
, 0, SEEK_SET
) == (off_t
) -1)
1221 return log_error_errno(errno
, "Failed to rewind serialization fd: %m");
1223 r
= fd_cloexec(fileno(f
), false);
1225 return log_error_errno(r
, "Failed to disable O_CLOEXEC for serialization: %m");
1227 r
= fdset_cloexec(fds
, false);
1229 return log_error_errno(r
, "Failed to disable O_CLOEXEC for serialization fds: %m");
1231 *ret_f
= TAKE_PTR(f
);
1232 *ret_fds
= TAKE_PTR(fds
);
1237 static void bump_file_max_and_nr_open(void) {
1239 /* Let's bump fs.file-max and fs.nr_open to their respective maximums. On current kernels large numbers of file
1240 * descriptors are no longer a performance problem and their memory is properly tracked by memcg, thus counting
1241 * them and limiting them in another two layers of limits is unnecessary and just complicates things. This
1242 * function hence turns off 2 of the 4 levels of limits on file descriptors, and makes RLIMIT_NOLIMIT (soft +
1243 * hard) the only ones that really matter. */
1245 #if BUMP_PROC_SYS_FS_FILE_MAX || BUMP_PROC_SYS_FS_NR_OPEN
1246 _cleanup_free_
char *t
= NULL
;
1250 #if BUMP_PROC_SYS_FS_FILE_MAX
1251 /* The maximum the kernel allows for this since 5.2 is LONG_MAX, use that. (Previously thing where
1252 * different but the operation would fail silently.) */
1253 if (asprintf(&t
, "%li\n", LONG_MAX
) < 0) {
1258 r
= sysctl_write("fs/file-max", t
);
1260 log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
, "Failed to bump fs.file-max, ignoring: %m");
1263 #if BUMP_PROC_SYS_FS_FILE_MAX && BUMP_PROC_SYS_FS_NR_OPEN
1267 #if BUMP_PROC_SYS_FS_NR_OPEN
1270 /* Arg! The kernel enforces maximum and minimum values on the fs.nr_open, but we don't really know what they
1271 * are. The expression by which the maximum is determined is dependent on the architecture, and is something we
1272 * don't really want to copy to userspace, as it is dependent on implementation details of the kernel. Since
1273 * the kernel doesn't expose the maximum value to us, we can only try and hope. Hence, let's start with
1274 * INT_MAX, and then keep halving the value until we find one that works. Ugly? Yes, absolutely, but kernel
1275 * APIs are kernel APIs, so what do can we do... 🤯 */
1280 v
&= ~(__SIZEOF_POINTER__
- 1); /* Round down to next multiple of the pointer size */
1282 log_warning("Can't bump fs.nr_open, value too small.");
1288 log_error_errno(k
, "Failed to read fs.nr_open: %m");
1291 if (k
>= v
) { /* Already larger */
1292 log_debug("Skipping bump, value is already larger.");
1296 if (asprintf(&t
, "%i\n", v
) < 0) {
1301 r
= sysctl_write("fs/nr_open", t
);
1304 log_debug("Couldn't write fs.nr_open as %i, halving it.", v
);
1309 log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
, "Failed to bump fs.nr_open, ignoring: %m");
1313 log_debug("Successfully bumped fs.nr_open to %i", v
);
1319 static int bump_rlimit_nofile(struct rlimit
*saved_rlimit
) {
1320 struct rlimit new_rlimit
;
1323 assert(saved_rlimit
);
1325 /* Save the original RLIMIT_NOFILE so that we can reset it later when transitioning from the initrd to the main
1326 * systemd or suchlike. */
1327 if (getrlimit(RLIMIT_NOFILE
, saved_rlimit
) < 0)
1328 return log_warning_errno(errno
, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1330 /* Get the underlying absolute limit the kernel enforces */
1331 nr
= read_nr_open();
1333 /* Make sure forked processes get limits based on the original kernel setting */
1334 if (!arg_default_rlimit
[RLIMIT_NOFILE
]) {
1337 rl
= newdup(struct rlimit
, saved_rlimit
, 1);
1341 /* Bump the hard limit for system services to a substantially higher value. The default hard limit
1342 * current kernels set is pretty low (4K), mostly for historical reasons. According to kernel
1343 * developers, the fd handling in recent kernels has been optimized substantially enough, so that we
1344 * can bump the limit now, without paying too high a price in memory or performance. Note however that
1345 * we only bump the hard limit, not the soft limit. That's because select() works the way it works, and
1346 * chokes on fds >= 1024. If we'd bump the soft limit globally, it might accidentally happen to
1347 * unexpecting programs that they get fds higher than what they can process using select(). By only
1348 * bumping the hard limit but leaving the low limit as it is we avoid this pitfall: programs that are
1349 * written by folks aware of the select() problem in mind (and thus use poll()/epoll instead of
1350 * select(), the way everybody should) can explicitly opt into high fds by bumping their soft limit
1351 * beyond 1024, to the hard limit we pass. */
1353 rl
->rlim_max
= MIN((rlim_t
) nr
, MAX(rl
->rlim_max
, (rlim_t
) HIGH_RLIMIT_NOFILE
));
1355 /* If for some reason we were invoked with a soft limit above 1024 (which should never
1356 * happen!, but who knows what we get passed in from pam_limit when invoked as --user
1357 * instance), then lower what we pass on to not confuse our children */
1358 rl
->rlim_cur
= MIN(rl
->rlim_cur
, (rlim_t
) FD_SETSIZE
);
1360 arg_default_rlimit
[RLIMIT_NOFILE
] = rl
;
1363 /* Calculate the new limits to use for us. Never lower from what we inherited. */
1364 new_rlimit
= (struct rlimit
) {
1365 .rlim_cur
= MAX((rlim_t
) nr
, saved_rlimit
->rlim_cur
),
1366 .rlim_max
= MAX((rlim_t
) nr
, saved_rlimit
->rlim_max
),
1369 /* Shortcut if nothing changes. */
1370 if (saved_rlimit
->rlim_max
>= new_rlimit
.rlim_max
&&
1371 saved_rlimit
->rlim_cur
>= new_rlimit
.rlim_cur
) {
1372 log_debug("RLIMIT_NOFILE is already as high or higher than we need it, not bumping.");
1376 /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for
1377 * both hard and soft. */
1378 r
= setrlimit_closest(RLIMIT_NOFILE
, &new_rlimit
);
1380 return log_warning_errno(r
, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1385 static int bump_rlimit_memlock(struct rlimit
*saved_rlimit
) {
1386 struct rlimit new_rlimit
;
1389 assert(saved_rlimit
);
1391 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK which should
1392 * normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's bump
1393 * the value high enough for our user. */
1395 if (getrlimit(RLIMIT_MEMLOCK
, saved_rlimit
) < 0)
1396 return log_warning_errno(errno
, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1398 /* Pass the original value down to invoked processes */
1399 if (!arg_default_rlimit
[RLIMIT_MEMLOCK
]) {
1402 rl
= newdup(struct rlimit
, saved_rlimit
, 1);
1406 arg_default_rlimit
[RLIMIT_MEMLOCK
] = rl
;
1409 /* Using MAX() on resource limits only is safe if RLIM_INFINITY is > 0. POSIX declares that rlim_t
1410 * must be unsigned, hence this is a given, but let's make this clear here. */
1411 assert_cc(RLIM_INFINITY
> 0);
1413 new_rlimit
= (struct rlimit
) {
1414 .rlim_cur
= MAX(HIGH_RLIMIT_MEMLOCK
, saved_rlimit
->rlim_cur
),
1415 .rlim_max
= MAX(HIGH_RLIMIT_MEMLOCK
, saved_rlimit
->rlim_max
),
1418 if (saved_rlimit
->rlim_max
>= new_rlimit
.rlim_cur
&&
1419 saved_rlimit
->rlim_cur
>= new_rlimit
.rlim_max
) {
1420 log_debug("RLIMIT_MEMLOCK is already as high or higher than we need it, not bumping.");
1424 r
= setrlimit_closest(RLIMIT_MEMLOCK
, &new_rlimit
);
1426 return log_warning_errno(r
, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1431 static void test_usr(void) {
1433 /* Check that /usr is either on the same file system as / or mounted already. */
1435 if (dir_is_empty("/usr") <= 0)
1438 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1439 "Some things will probably break (sometimes even silently) in mysterious ways. "
1440 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1443 static int enforce_syscall_archs(Set
*archs
) {
1447 if (!is_seccomp_available())
1450 r
= seccomp_restrict_archs(arg_syscall_archs
);
1452 return log_error_errno(r
, "Failed to enforce system call architecture restrication: %m");
1457 static int status_welcome(void) {
1458 _cleanup_free_
char *pretty_name
= NULL
, *ansi_color
= NULL
;
1461 if (IN_SET(arg_show_status
, SHOW_STATUS_NO
, SHOW_STATUS_AUTO
))
1464 r
= parse_os_release(NULL
,
1465 "PRETTY_NAME", &pretty_name
,
1466 "ANSI_COLOR", &ansi_color
,
1469 log_full_errno(r
== -ENOENT
? LOG_DEBUG
: LOG_WARNING
, r
,
1470 "Failed to read os-release file, ignoring: %m");
1472 if (log_get_show_color())
1473 return status_printf(NULL
, 0,
1474 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1475 isempty(ansi_color
) ? "1" : ansi_color
,
1476 isempty(pretty_name
) ? "Linux" : pretty_name
);
1478 return status_printf(NULL
, 0,
1479 "\nWelcome to %s!\n",
1480 isempty(pretty_name
) ? "Linux" : pretty_name
);
1483 static int write_container_id(void) {
1487 c
= getenv("container");
1491 RUN_WITH_UMASK(0022)
1492 r
= write_string_file("/run/systemd/container", c
, WRITE_STRING_FILE_CREATE
);
1494 return log_warning_errno(r
, "Failed to write /run/systemd/container, ignoring: %m");
1499 static int bump_unix_max_dgram_qlen(void) {
1500 _cleanup_free_
char *qlen
= NULL
;
1504 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1505 * really really early during boot, so that it is actually applied to all our sockets, including the
1506 * $NOTIFY_SOCKET one. */
1508 r
= read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen
);
1510 return log_full_errno(r
== -ENOENT
? LOG_DEBUG
: LOG_WARNING
, r
, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1512 r
= safe_atolu(qlen
, &v
);
1514 return log_warning_errno(r
, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen
);
1516 if (v
>= DEFAULT_UNIX_MAX_DGRAM_QLEN
)
1519 r
= write_string_filef("/proc/sys/net/unix/max_dgram_qlen", WRITE_STRING_FILE_DISABLE_BUFFER
, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN
);
1521 return log_full_errno(IN_SET(r
, -EROFS
, -EPERM
, -EACCES
) ? LOG_DEBUG
: LOG_WARNING
, r
,
1522 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1527 static int fixup_environment(void) {
1528 _cleanup_free_
char *term
= NULL
;
1532 /* Only fix up the environment when we are started as PID 1 */
1533 if (getpid_cached() != 1)
1536 /* We expect the environment to be set correctly if run inside a container. */
1537 if (detect_container() > 0)
1540 /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1541 * device used by the console. We try to make a better guess here since some consoles might not have support
1542 * for color mode for example.
1544 * However if TERM was configured through the kernel command line then leave it alone. */
1545 r
= proc_cmdline_get_key("TERM", 0, &term
);
1549 t
= term
?: default_term_for_tty("/dev/console");
1551 if (setenv("TERM", t
, 1) < 0)
1554 /* The kernels sets HOME=/ for init. Let's undo this. */
1555 if (path_equal_ptr(getenv("HOME"), "/") &&
1556 unsetenv("HOME") < 0)
1557 log_warning_errno(errno
, "Failed to unset $HOME: %m");
1562 static void redirect_telinit(int argc
, char *argv
[]) {
1564 /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1566 #if HAVE_SYSV_COMPAT
1567 if (getpid_cached() == 1)
1570 if (!strstr(program_invocation_short_name
, "init"))
1573 execv(SYSTEMCTL_BINARY_PATH
, argv
);
1574 log_error_errno(errno
, "Failed to exec " SYSTEMCTL_BINARY_PATH
": %m");
1579 static int become_shutdown(
1580 const char *shutdown_verb
,
1583 char log_level
[DECIMAL_STR_MAX(int) + 1],
1584 exit_code
[DECIMAL_STR_MAX(uint8_t) + 1],
1585 timeout
[DECIMAL_STR_MAX(usec_t
) + 1];
1587 const char* command_line
[13] = {
1588 SYSTEMD_SHUTDOWN_BINARY_PATH
,
1590 "--timeout", timeout
,
1591 "--log-level", log_level
,
1595 _cleanup_strv_free_
char **env_block
= NULL
;
1599 assert(shutdown_verb
);
1600 assert(!command_line
[pos
]);
1601 env_block
= strv_copy(environ
);
1603 xsprintf(log_level
, "%d", log_get_max_level());
1604 xsprintf(timeout
, "%" PRI_USEC
"us", arg_default_timeout_stop_usec
);
1606 switch (log_get_target()) {
1608 case LOG_TARGET_KMSG
:
1609 case LOG_TARGET_JOURNAL_OR_KMSG
:
1610 case LOG_TARGET_SYSLOG_OR_KMSG
:
1611 command_line
[pos
++] = "kmsg";
1614 case LOG_TARGET_NULL
:
1615 command_line
[pos
++] = "null";
1618 case LOG_TARGET_CONSOLE
:
1620 command_line
[pos
++] = "console";
1624 if (log_get_show_color())
1625 command_line
[pos
++] = "--log-color";
1627 if (log_get_show_location())
1628 command_line
[pos
++] = "--log-location";
1630 if (streq(shutdown_verb
, "exit")) {
1631 command_line
[pos
++] = "--exit-code";
1632 command_line
[pos
++] = exit_code
;
1633 xsprintf(exit_code
, "%d", retval
);
1636 assert(pos
< ELEMENTSOF(command_line
));
1638 if (streq(shutdown_verb
, "reboot") &&
1639 arg_shutdown_watchdog
> 0 &&
1640 arg_shutdown_watchdog
!= USEC_INFINITY
) {
1644 /* If we reboot let's set the shutdown
1645 * watchdog and tell the shutdown binary to
1646 * repeatedly ping it */
1647 r
= watchdog_set_timeout(&arg_shutdown_watchdog
);
1648 watchdog_close(r
< 0);
1650 /* Tell the binary how often to ping, ignore failure */
1651 if (asprintf(&e
, "WATCHDOG_USEC="USEC_FMT
, arg_shutdown_watchdog
) > 0)
1652 (void) strv_consume(&env_block
, e
);
1654 if (arg_watchdog_device
&&
1655 asprintf(&e
, "WATCHDOG_DEVICE=%s", arg_watchdog_device
) > 0)
1656 (void) strv_consume(&env_block
, e
);
1658 watchdog_close(true);
1660 /* Avoid the creation of new processes forked by the
1661 * kernel; at this point, we will not listen to the
1663 if (detect_container() <= 0)
1664 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER
);
1666 execve(SYSTEMD_SHUTDOWN_BINARY_PATH
, (char **) command_line
, env_block
);
1670 static void initialize_clock(void) {
1673 if (clock_is_localtime(NULL
) > 0) {
1677 * The very first call of settimeofday() also does a time warp in the kernel.
1679 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1680 * of maintaining the RTC and do all adjustments. This matches the behavior of Windows, which leaves
1681 * the RTC alone if the registry tells that the RTC runs in UTC.
1683 r
= clock_set_timezone(&min
);
1685 log_error_errno(r
, "Failed to apply local time delta, ignoring: %m");
1687 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min
);
1689 } else if (!in_initrd()) {
1691 * Do a dummy very first call to seal the kernel's time warp magic.
1693 * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1694 * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1695 * until we reach the real system.
1697 * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1698 * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1699 * be treated as UTC that way.
1701 (void) clock_reset_timewarp();
1704 r
= clock_apply_epoch();
1706 log_error_errno(r
, "Current system time is before build time, but cannot correct: %m");
1708 log_info("System time before build time, advancing clock.");
1711 static void initialize_coredump(bool skip_setup
) {
1713 if (getpid_cached() != 1)
1716 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1717 * will process core dumps for system services by default. */
1718 if (setrlimit(RLIMIT_CORE
, &RLIMIT_MAKE_CONST(RLIM_INFINITY
)) < 0)
1719 log_warning_errno(errno
, "Failed to set RLIMIT_CORE: %m");
1721 /* But at the same time, turn off the core_pattern logic by default, so that no
1722 * coredumps are stored until the systemd-coredump tool is enabled via
1723 * sysctl. However it can be changed via the kernel command line later so core
1724 * dumps can still be generated during early startup and in initramfs. */
1726 disable_coredumps();
1730 static void initialize_core_pattern(bool skip_setup
) {
1733 if (skip_setup
|| !arg_early_core_pattern
)
1736 if (getpid_cached() != 1)
1739 r
= write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern
, WRITE_STRING_FILE_DISABLE_BUFFER
);
1741 log_warning_errno(r
, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", arg_early_core_pattern
);
1744 static void update_cpu_affinity(bool skip_setup
) {
1745 _cleanup_free_
char *mask
= NULL
;
1747 if (skip_setup
|| !arg_cpu_affinity
.set
)
1750 assert(arg_cpu_affinity
.allocated
> 0);
1752 mask
= cpu_set_to_string(&arg_cpu_affinity
);
1753 log_debug("Setting CPU affinity to %s.", strnull(mask
));
1755 if (sched_setaffinity(0, arg_cpu_affinity
.allocated
, arg_cpu_affinity
.set
) < 0)
1756 log_warning_errno(errno
, "Failed to set CPU affinity: %m");
1759 static void update_numa_policy(bool skip_setup
) {
1761 _cleanup_free_
char *nodes
= NULL
;
1762 const char * policy
= NULL
;
1764 if (skip_setup
|| !mpol_is_valid(numa_policy_get_type(&arg_numa_policy
)))
1767 if (DEBUG_LOGGING
) {
1768 policy
= mpol_to_string(numa_policy_get_type(&arg_numa_policy
));
1769 nodes
= cpu_set_to_range_string(&arg_numa_policy
.nodes
);
1770 log_debug("Setting NUMA policy to %s, with nodes %s.", strnull(policy
), strnull(nodes
));
1773 r
= apply_numa_policy(&arg_numa_policy
);
1774 if (r
== -EOPNOTSUPP
)
1775 log_debug_errno(r
, "NUMA support not available, ignoring.");
1777 log_warning_errno(r
, "Failed to set NUMA memory policy: %m");
1780 static void do_reexecute(
1783 const struct rlimit
*saved_rlimit_nofile
,
1784 const struct rlimit
*saved_rlimit_memlock
,
1786 const char *switch_root_dir
,
1787 const char *switch_root_init
,
1788 const char **ret_error_message
) {
1790 unsigned i
, j
, args_size
;
1794 assert(saved_rlimit_nofile
);
1795 assert(saved_rlimit_memlock
);
1796 assert(ret_error_message
);
1798 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1800 watchdog_close(true);
1802 /* Reset RLIMIT_NOFILE + RLIMIT_MEMLOCK back to the kernel defaults, so that the new systemd can pass
1803 * the kernel default to its child processes */
1804 if (saved_rlimit_nofile
->rlim_cur
!= 0)
1805 (void) setrlimit(RLIMIT_NOFILE
, saved_rlimit_nofile
);
1806 if (saved_rlimit_memlock
->rlim_cur
!= RLIM_INFINITY
)
1807 (void) setrlimit(RLIMIT_MEMLOCK
, saved_rlimit_memlock
);
1809 if (switch_root_dir
) {
1810 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1811 * SIGCHLD for them after deserializing. */
1812 broadcast_signal(SIGTERM
, false, true, arg_default_timeout_stop_usec
);
1814 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1815 r
= switch_root(switch_root_dir
, "/mnt", true, MS_MOVE
);
1817 log_error_errno(r
, "Failed to switch root, trying to continue: %m");
1820 args_size
= MAX(6, argc
+1);
1821 args
= newa(const char*, args_size
);
1823 if (!switch_root_init
) {
1824 char sfd
[DECIMAL_STR_MAX(int) + 1];
1826 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1827 * the user didn't specify an explicit init to spawn. */
1829 assert(arg_serialization
);
1832 xsprintf(sfd
, "%i", fileno(arg_serialization
));
1835 args
[i
++] = SYSTEMD_BINARY_PATH
;
1836 if (switch_root_dir
)
1837 args
[i
++] = "--switched-root";
1838 args
[i
++] = arg_system
? "--system" : "--user";
1839 args
[i
++] = "--deserialize";
1843 assert(i
<= args_size
);
1846 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do this is on
1847 * its own on exec(), but it will do it on exit(). Hence, to ensure we get a summary here, fork() off
1848 * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1849 * before proceeding into the exec().
1851 valgrind_summary_hack();
1853 (void) execv(args
[0], (char* const*) args
);
1854 log_debug_errno(errno
, "Failed to execute our own binary, trying fallback: %m");
1857 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1858 * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1859 * doesn't matter.) */
1861 arg_serialization
= safe_fclose(arg_serialization
);
1862 fds
= fdset_free(fds
);
1864 /* Reopen the console */
1865 (void) make_console_stdio();
1867 for (j
= 1, i
= 1; j
< (unsigned) argc
; j
++)
1868 args
[i
++] = argv
[j
];
1870 assert(i
<= args_size
);
1872 /* Re-enable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1873 (void) reset_all_signal_handlers();
1874 (void) reset_signal_mask();
1875 (void) rlimit_nofile_safe();
1877 if (switch_root_init
) {
1878 args
[0] = switch_root_init
;
1879 (void) execv(args
[0], (char* const*) args
);
1880 log_warning_errno(errno
, "Failed to execute configured init, trying fallback: %m");
1883 args
[0] = "/sbin/init";
1884 (void) execv(args
[0], (char* const*) args
);
1887 manager_status_printf(NULL
, STATUS_TYPE_EMERGENCY
,
1888 ANSI_HIGHLIGHT_RED
" !! " ANSI_NORMAL
,
1889 "Failed to execute /sbin/init");
1892 log_warning("No /sbin/init, trying fallback");
1894 args
[0] = "/bin/sh";
1896 (void) execv(args
[0], (char* const*) args
);
1897 log_error_errno(errno
, "Failed to execute /bin/sh, giving up: %m");
1899 log_warning_errno(r
, "Failed to execute /sbin/init, giving up: %m");
1901 *ret_error_message
= "Failed to execute fallback shell";
1904 static int invoke_main_loop(
1906 bool *ret_reexecute
,
1907 int *ret_retval
, /* Return parameters relevant for shutting down */
1908 const char **ret_shutdown_verb
, /* … */
1909 FDSet
**ret_fds
, /* Return parameters for reexecuting */
1910 char **ret_switch_root_dir
, /* … */
1911 char **ret_switch_root_init
, /* … */
1912 const char **ret_error_message
) {
1917 assert(ret_reexecute
);
1919 assert(ret_shutdown_verb
);
1921 assert(ret_switch_root_dir
);
1922 assert(ret_switch_root_init
);
1923 assert(ret_error_message
);
1926 r
= manager_loop(m
);
1928 *ret_error_message
= "Failed to run main loop";
1929 return log_emergency_errno(r
, "Failed to run main loop: %m");
1932 switch ((ManagerObjective
) r
) {
1934 case MANAGER_RELOAD
: {
1935 LogTarget saved_log_target
;
1936 int saved_log_level
;
1938 log_info("Reloading.");
1940 /* First, save any overridden log level/target, then parse the configuration file, which might
1941 * change the log level to new settings. */
1943 saved_log_level
= m
->log_level_overridden
? log_get_max_level() : -1;
1944 saved_log_target
= m
->log_target_overridden
? log_get_target() : _LOG_TARGET_INVALID
;
1946 (void) parse_configuration();
1948 set_manager_defaults(m
);
1950 update_cpu_affinity(false);
1951 update_numa_policy(false);
1953 if (saved_log_level
>= 0)
1954 manager_override_log_level(m
, saved_log_level
);
1955 if (saved_log_target
>= 0)
1956 manager_override_log_target(m
, saved_log_target
);
1958 r
= manager_reload(m
);
1960 /* Reloading failed before the point of no return. Let's continue running as if nothing happened. */
1961 m
->objective
= MANAGER_OK
;
1966 case MANAGER_REEXECUTE
:
1968 r
= prepare_reexecute(m
, &arg_serialization
, ret_fds
, false);
1970 *ret_error_message
= "Failed to prepare for reexecution";
1974 log_notice("Reexecuting.");
1976 *ret_reexecute
= true;
1977 *ret_retval
= EXIT_SUCCESS
;
1978 *ret_shutdown_verb
= NULL
;
1979 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
1983 case MANAGER_SWITCH_ROOT
:
1984 if (!m
->switch_root_init
) {
1985 r
= prepare_reexecute(m
, &arg_serialization
, ret_fds
, true);
1987 *ret_error_message
= "Failed to prepare for reexecution";
1993 log_notice("Switching root.");
1995 *ret_reexecute
= true;
1996 *ret_retval
= EXIT_SUCCESS
;
1997 *ret_shutdown_verb
= NULL
;
1999 /* Steal the switch root parameters */
2000 *ret_switch_root_dir
= TAKE_PTR(m
->switch_root
);
2001 *ret_switch_root_init
= TAKE_PTR(m
->switch_root_init
);
2007 if (MANAGER_IS_USER(m
)) {
2010 *ret_reexecute
= false;
2011 *ret_retval
= m
->return_value
;
2012 *ret_shutdown_verb
= NULL
;
2014 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
2020 case MANAGER_REBOOT
:
2021 case MANAGER_POWEROFF
:
2023 case MANAGER_KEXEC
: {
2024 static const char * const table
[_MANAGER_OBJECTIVE_MAX
] = {
2025 [MANAGER_EXIT
] = "exit",
2026 [MANAGER_REBOOT
] = "reboot",
2027 [MANAGER_POWEROFF
] = "poweroff",
2028 [MANAGER_HALT
] = "halt",
2029 [MANAGER_KEXEC
] = "kexec",
2032 log_notice("Shutting down.");
2034 *ret_reexecute
= false;
2035 *ret_retval
= m
->return_value
;
2036 assert_se(*ret_shutdown_verb
= table
[m
->objective
]);
2038 *ret_switch_root_dir
= *ret_switch_root_init
= NULL
;
2044 assert_not_reached("Unknown or unexpected manager objective.");
2049 static void log_execution_mode(bool *ret_first_boot
) {
2050 assert(ret_first_boot
);
2055 log_info("systemd " GIT_VERSION
" running in %ssystem mode. (" SYSTEMD_FEATURES
")",
2056 arg_action
== ACTION_TEST
? "test " : "" );
2058 v
= detect_virtualization();
2060 log_info("Detected virtualization %s.", virtualization_to_string(v
));
2062 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
2065 *ret_first_boot
= false;
2066 log_info("Running in initial RAM disk.");
2068 /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
2069 * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
2070 * doesn't it's unpopulated. This allows container managers and installers to provision a
2071 * couple of files already. If the container manager wants to provision the machine ID itself
2072 * it should pass $container_uuid to PID 1. */
2074 *ret_first_boot
= access("/etc/machine-id", F_OK
) < 0;
2075 if (*ret_first_boot
)
2076 log_info("Running with unpopulated /etc.");
2079 if (DEBUG_LOGGING
) {
2080 _cleanup_free_
char *t
;
2082 t
= uid_to_name(getuid());
2083 log_debug("systemd " GIT_VERSION
" running in %suser mode for user " UID_FMT
"/%s. (" SYSTEMD_FEATURES
")",
2084 arg_action
== ACTION_TEST
? " test" : "", getuid(), strna(t
));
2087 *ret_first_boot
= false;
2091 static int initialize_runtime(
2093 struct rlimit
*saved_rlimit_nofile
,
2094 struct rlimit
*saved_rlimit_memlock
,
2095 const char **ret_error_message
) {
2099 assert(ret_error_message
);
2101 /* Sets up various runtime parameters. Many of these initializations are conditionalized:
2103 * - Some only apply to --system instances
2104 * - Some only apply to --user instances
2105 * - Some only apply when we first start up, but not when we reexecute
2108 if (arg_action
!= ACTION_RUN
)
2111 update_cpu_affinity(skip_setup
);
2112 update_numa_policy(skip_setup
);
2115 /* Make sure we leave a core dump without panicking the kernel. */
2116 install_crash_handler();
2119 r
= mount_cgroup_controllers();
2121 *ret_error_message
= "Failed to mount cgroup hierarchies";
2127 machine_id_setup(NULL
, arg_machine_id
, NULL
);
2129 bump_unix_max_dgram_qlen();
2130 bump_file_max_and_nr_open();
2132 write_container_id();
2135 if (arg_watchdog_device
) {
2136 r
= watchdog_set_device(arg_watchdog_device
);
2138 log_warning_errno(r
, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device
);
2141 if (arg_runtime_watchdog
> 0 && arg_runtime_watchdog
!= USEC_INFINITY
)
2142 watchdog_set_timeout(&arg_runtime_watchdog
);
2145 if (arg_timer_slack_nsec
!= NSEC_INFINITY
)
2146 if (prctl(PR_SET_TIMERSLACK
, arg_timer_slack_nsec
) < 0)
2147 log_warning_errno(errno
, "Failed to adjust timer slack, ignoring: %m");
2149 if (arg_system
&& !cap_test_all(arg_capability_bounding_set
)) {
2150 r
= capability_bounding_set_drop_usermode(arg_capability_bounding_set
);
2152 *ret_error_message
= "Failed to drop capability bounding set of usermode helpers";
2153 return log_emergency_errno(r
, "Failed to drop capability bounding set of usermode helpers: %m");
2156 r
= capability_bounding_set_drop(arg_capability_bounding_set
, true);
2158 *ret_error_message
= "Failed to drop capability bounding set";
2159 return log_emergency_errno(r
, "Failed to drop capability bounding set: %m");
2163 if (arg_system
&& arg_no_new_privs
) {
2164 if (prctl(PR_SET_NO_NEW_PRIVS
, 1, 0, 0, 0) < 0) {
2165 *ret_error_message
= "Failed to disable new privileges";
2166 return log_emergency_errno(errno
, "Failed to disable new privileges: %m");
2170 if (arg_syscall_archs
) {
2171 r
= enforce_syscall_archs(arg_syscall_archs
);
2173 *ret_error_message
= "Failed to set syscall architectures";
2179 /* Become reaper of our children */
2180 if (prctl(PR_SET_CHILD_SUBREAPER
, 1) < 0)
2181 log_warning_errno(errno
, "Failed to make us a subreaper: %m");
2183 /* Bump up RLIMIT_NOFILE for systemd itself */
2184 (void) bump_rlimit_nofile(saved_rlimit_nofile
);
2185 (void) bump_rlimit_memlock(saved_rlimit_memlock
);
2190 static int do_queue_default_job(
2192 const char **ret_error_message
) {
2194 _cleanup_(sd_bus_error_free
) sd_bus_error error
= SD_BUS_ERROR_NULL
;
2195 Job
*default_unit_job
;
2196 Unit
*target
= NULL
;
2199 log_debug("Activating default unit: %s", arg_default_unit
);
2201 r
= manager_load_startable_unit_or_warn(m
, arg_default_unit
, NULL
, &target
);
2203 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET
);
2205 r
= manager_load_startable_unit_or_warn(m
, SPECIAL_RESCUE_TARGET
, NULL
, &target
);
2207 *ret_error_message
= r
== -ERFKILL
? "Rescue target masked"
2208 : "Failed to load rescue target";
2213 assert(target
->load_state
== UNIT_LOADED
);
2215 r
= manager_add_job(m
, JOB_START
, target
, JOB_ISOLATE
, NULL
, &error
, &default_unit_job
);
2217 log_debug_errno(r
, "Default target could not be isolated, starting instead: %s", bus_error_message(&error
, r
));
2219 sd_bus_error_free(&error
);
2221 r
= manager_add_job(m
, JOB_START
, target
, JOB_REPLACE
, NULL
, &error
, &default_unit_job
);
2223 *ret_error_message
= "Failed to start default target";
2224 return log_emergency_errno(r
, "Failed to start default target: %s", bus_error_message(&error
, r
));
2228 *ret_error_message
= "Failed to isolate default target";
2229 return log_emergency_errno(r
, "Failed to isolate default target: %s", bus_error_message(&error
, r
));
2232 m
->default_unit_job_id
= default_unit_job
->id
;
2237 static void reset_arguments(void) {
2238 /* Frees/resets arg_* variables, with a few exceptions commented below. */
2240 arg_default_unit
= mfree(arg_default_unit
);
2242 /* arg_system — ignore */
2244 arg_dump_core
= true;
2245 arg_crash_chvt
= -1;
2246 arg_crash_shell
= false;
2247 arg_crash_reboot
= false;
2248 arg_confirm_spawn
= mfree(arg_confirm_spawn
);
2249 arg_show_status
= _SHOW_STATUS_INVALID
;
2250 arg_switched_root
= false;
2251 arg_pager_flags
= 0;
2252 arg_service_watchdogs
= true;
2253 arg_default_std_output
= EXEC_OUTPUT_JOURNAL
;
2254 arg_default_std_error
= EXEC_OUTPUT_INHERIT
;
2255 arg_default_restart_usec
= DEFAULT_RESTART_USEC
;
2256 arg_default_timeout_start_usec
= DEFAULT_TIMEOUT_USEC
;
2257 arg_default_timeout_stop_usec
= DEFAULT_TIMEOUT_USEC
;
2258 arg_default_timeout_abort_usec
= DEFAULT_TIMEOUT_USEC
;
2259 arg_default_timeout_abort_set
= false;
2260 arg_default_start_limit_interval
= DEFAULT_START_LIMIT_INTERVAL
;
2261 arg_default_start_limit_burst
= DEFAULT_START_LIMIT_BURST
;
2262 arg_runtime_watchdog
= 0;
2263 arg_shutdown_watchdog
= 10 * USEC_PER_MINUTE
;
2264 arg_early_core_pattern
= NULL
;
2265 arg_watchdog_device
= NULL
;
2267 arg_default_environment
= strv_free(arg_default_environment
);
2268 rlimit_free_all(arg_default_rlimit
);
2270 arg_capability_bounding_set
= CAP_ALL
;
2271 arg_no_new_privs
= false;
2272 arg_timer_slack_nsec
= NSEC_INFINITY
;
2273 arg_default_timer_accuracy_usec
= 1 * USEC_PER_MINUTE
;
2275 arg_syscall_archs
= set_free(arg_syscall_archs
);
2277 /* arg_serialization — ignore */
2279 arg_default_cpu_accounting
= -1;
2280 arg_default_io_accounting
= false;
2281 arg_default_ip_accounting
= false;
2282 arg_default_blockio_accounting
= false;
2283 arg_default_memory_accounting
= MEMORY_ACCOUNTING_DEFAULT
;
2284 arg_default_tasks_accounting
= true;
2285 arg_default_tasks_max
= UINT64_MAX
;
2286 arg_machine_id
= (sd_id128_t
) {};
2287 arg_cad_burst_action
= EMERGENCY_ACTION_REBOOT_FORCE
;
2288 arg_default_oom_policy
= OOM_STOP
;
2290 cpu_set_reset(&arg_cpu_affinity
);
2291 numa_policy_reset(&arg_numa_policy
);
2294 static int parse_configuration(void) {
2297 arg_default_tasks_max
= system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE
, 100U);
2299 /* Assign configuration defaults */
2302 r
= parse_config_file();
2304 log_warning_errno(r
, "Failed to parse config file, ignoring: %m");
2307 r
= proc_cmdline_parse(parse_proc_cmdline_item
, NULL
, 0);
2309 log_warning_errno(r
, "Failed to parse kernel command line, ignoring: %m");
2312 /* Note that this also parses bits from the kernel command line, including "debug". */
2313 log_parse_environment();
2318 static int load_configuration(int argc
, char **argv
, const char **ret_error_message
) {
2321 assert(ret_error_message
);
2323 (void) parse_configuration();
2325 r
= parse_argv(argc
, argv
);
2327 *ret_error_message
= "Failed to parse commandline arguments";
2331 /* Initialize default unit */
2332 if (!arg_default_unit
) {
2333 arg_default_unit
= strdup(SPECIAL_DEFAULT_TARGET
);
2334 if (!arg_default_unit
) {
2335 *ret_error_message
= "Failed to set default unit";
2340 /* Initialize the show status setting if it hasn't been set explicitly yet */
2341 if (arg_show_status
== _SHOW_STATUS_INVALID
)
2342 arg_show_status
= SHOW_STATUS_YES
;
2347 static int safety_checks(void) {
2349 if (getpid_cached() == 1 &&
2350 arg_action
!= ACTION_RUN
)
2351 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
2352 "Unsupported execution mode while PID 1.");
2354 if (getpid_cached() == 1 &&
2356 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
2357 "Can't run --user mode as PID 1.");
2359 if (arg_action
== ACTION_RUN
&&
2361 getpid_cached() != 1)
2362 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
2363 "Can't run system mode unless PID 1.");
2365 if (arg_action
== ACTION_TEST
&&
2367 return log_error_errno(SYNTHETIC_ERRNO(EPERM
),
2368 "Don't run test mode as root.");
2371 arg_action
== ACTION_RUN
&&
2373 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP
),
2374 "Trying to run as user instance, but the system has not been booted with systemd.");
2377 arg_action
== ACTION_RUN
&&
2378 !getenv("XDG_RUNTIME_DIR"))
2379 return log_error_errno(SYNTHETIC_ERRNO(EUNATCH
),
2380 "Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
2383 arg_action
== ACTION_RUN
&&
2384 running_in_chroot() > 0)
2385 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP
),
2386 "Cannot be run in a chroot() environment.");
2391 static int initialize_security(
2392 bool *loaded_policy
,
2393 dual_timestamp
*security_start_timestamp
,
2394 dual_timestamp
*security_finish_timestamp
,
2395 const char **ret_error_message
) {
2399 assert(loaded_policy
);
2400 assert(security_start_timestamp
);
2401 assert(security_finish_timestamp
);
2402 assert(ret_error_message
);
2404 dual_timestamp_get(security_start_timestamp
);
2406 r
= mac_selinux_setup(loaded_policy
);
2408 *ret_error_message
= "Failed to load SELinux policy";
2412 r
= mac_smack_setup(loaded_policy
);
2414 *ret_error_message
= "Failed to load SMACK policy";
2420 *ret_error_message
= "Failed to load IMA policy";
2424 dual_timestamp_get(security_finish_timestamp
);
2428 static void test_summary(Manager
*m
) {
2431 printf("-> By units:\n");
2432 manager_dump_units(m
, stdout
, "\t");
2434 printf("-> By jobs:\n");
2435 manager_dump_jobs(m
, stdout
, "\t");
2438 static int collect_fds(FDSet
**ret_fds
, const char **ret_error_message
) {
2442 assert(ret_error_message
);
2444 r
= fdset_new_fill(ret_fds
);
2446 *ret_error_message
= "Failed to allocate fd set";
2447 return log_emergency_errno(r
, "Failed to allocate fd set: %m");
2450 fdset_cloexec(*ret_fds
, true);
2452 if (arg_serialization
)
2453 assert_se(fdset_remove(*ret_fds
, fileno(arg_serialization
)) >= 0);
2458 static void setup_console_terminal(bool skip_setup
) {
2463 /* Become a session leader if we aren't one yet. */
2466 /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2468 (void) release_terminal();
2470 /* Reset the console, but only if this is really init and we are freshly booted */
2471 if (getpid_cached() == 1 && !skip_setup
)
2472 (void) console_setup();
2475 static bool early_skip_setup_check(int argc
, char *argv
[]) {
2476 bool found_deserialize
= false;
2479 /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2480 * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2481 * anyway, even if in that case we also do deserialization. */
2483 for (i
= 1; i
< argc
; i
++) {
2484 if (streq(argv
[i
], "--switched-root"))
2485 return false; /* If we switched root, don't skip the setup. */
2486 else if (streq(argv
[i
], "--deserialize"))
2487 found_deserialize
= true;
2490 return found_deserialize
; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2493 int main(int argc
, char *argv
[]) {
2495 dual_timestamp initrd_timestamp
= DUAL_TIMESTAMP_NULL
, userspace_timestamp
= DUAL_TIMESTAMP_NULL
, kernel_timestamp
= DUAL_TIMESTAMP_NULL
,
2496 security_start_timestamp
= DUAL_TIMESTAMP_NULL
, security_finish_timestamp
= DUAL_TIMESTAMP_NULL
;
2497 struct rlimit saved_rlimit_nofile
= RLIMIT_MAKE_CONST(0),
2498 saved_rlimit_memlock
= RLIMIT_MAKE_CONST(RLIM_INFINITY
); /* The original rlimits we passed
2499 * in. Note we use different values
2500 * for the two that indicate whether
2501 * these fields are initialized! */
2502 bool skip_setup
, loaded_policy
= false, queue_default_job
= false, first_boot
= false, reexecute
= false;
2503 char *switch_root_dir
= NULL
, *switch_root_init
= NULL
;
2504 usec_t before_startup
, after_startup
;
2505 static char systemd
[] = "systemd";
2506 char timespan
[FORMAT_TIMESPAN_MAX
];
2507 const char *shutdown_verb
= NULL
, *error_message
= NULL
;
2508 int r
, retval
= EXIT_FAILURE
;
2512 /* SysV compatibility: redirect init → telinit */
2513 redirect_telinit(argc
, argv
);
2515 /* Take timestamps early on */
2516 dual_timestamp_from_monotonic(&kernel_timestamp
, 0);
2517 dual_timestamp_get(&userspace_timestamp
);
2519 /* Figure out whether we need to do initialize the system, or if we already did that because we are
2521 skip_setup
= early_skip_setup_check(argc
, argv
);
2523 /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2524 * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
2525 program_invocation_short_name
= systemd
;
2526 (void) prctl(PR_SET_NAME
, systemd
);
2528 /* Save the original command line */
2529 save_argc_argv(argc
, argv
);
2531 /* Make sure that if the user says "syslog" we actually log to the journal. */
2532 log_set_upgrade_syslog_to_journal(true);
2534 if (getpid_cached() == 1) {
2535 /* When we run as PID 1 force system mode */
2538 /* Disable the umask logic */
2541 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2542 * activated yet (even though the log socket for it exists). */
2543 log_set_prohibit_ipc(true);
2545 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2546 * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2547 * child process right before execve()'ing the actual binary, at a point in time where socket
2548 * activation stderr/stdout area already set up. */
2549 log_set_always_reopen_console(true);
2551 if (detect_container() <= 0) {
2553 /* Running outside of a container as PID 1 */
2554 log_set_target(LOG_TARGET_KMSG
);
2558 initrd_timestamp
= userspace_timestamp
;
2561 r
= mount_setup_early();
2563 error_message
= "Failed to mount early API filesystems";
2567 r
= initialize_security(
2569 &security_start_timestamp
,
2570 &security_finish_timestamp
,
2576 if (mac_selinux_init() < 0) {
2577 error_message
= "Failed to initialize SELinux policy";
2584 /* Set the default for later on, but don't actually open the logs like this for now. Note that
2585 * if we are transitioning from the initrd there might still be journal fd open, and we
2586 * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2588 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG
);
2591 /* Running inside a container, as PID 1 */
2592 log_set_target(LOG_TARGET_CONSOLE
);
2595 /* For later on, see above... */
2596 log_set_target(LOG_TARGET_JOURNAL
);
2598 /* clear the kernel timestamp,
2599 * because we are in a container */
2600 kernel_timestamp
= DUAL_TIMESTAMP_NULL
;
2603 initialize_coredump(skip_setup
);
2605 r
= fixup_environment();
2607 log_emergency_errno(r
, "Failed to fix up PID 1 environment: %m");
2608 error_message
= "Failed to fix up PID1 environment";
2613 /* Running as user instance */
2615 log_set_target(LOG_TARGET_AUTO
);
2618 /* clear the kernel timestamp,
2619 * because we are not PID 1 */
2620 kernel_timestamp
= DUAL_TIMESTAMP_NULL
;
2624 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2625 * they never log into the console. */
2626 log_show_color(colors_enabled());
2628 r
= make_null_stdio();
2630 log_warning_errno(r
, "Failed to redirect standard streams to /dev/null, ignoring: %m");
2633 /* Mount /proc, /sys and friends, so that /proc/cmdline and
2634 * /proc/$PID/fd is available. */
2635 if (getpid_cached() == 1) {
2637 /* Load the kernel modules early. */
2641 r
= mount_setup(loaded_policy
);
2643 error_message
= "Failed to mount API filesystems";
2648 /* Reset all signal handlers. */
2649 (void) reset_all_signal_handlers();
2650 (void) ignore_signals(SIGNALS_IGNORE
, -1);
2652 r
= load_configuration(argc
, argv
, &error_message
);
2656 r
= safety_checks();
2660 if (IN_SET(arg_action
, ACTION_TEST
, ACTION_HELP
, ACTION_DUMP_CONFIGURATION_ITEMS
, ACTION_DUMP_BUS_PROPERTIES
))
2661 (void) pager_open(arg_pager_flags
);
2663 if (arg_action
!= ACTION_RUN
)
2666 if (arg_action
== ACTION_HELP
) {
2667 retval
= help() < 0 ? EXIT_FAILURE
: EXIT_SUCCESS
;
2669 } else if (arg_action
== ACTION_VERSION
) {
2672 } else if (arg_action
== ACTION_DUMP_CONFIGURATION_ITEMS
) {
2673 unit_dump_config_items(stdout
);
2674 retval
= EXIT_SUCCESS
;
2676 } else if (arg_action
== ACTION_DUMP_BUS_PROPERTIES
) {
2677 dump_bus_properties(stdout
);
2678 retval
= EXIT_SUCCESS
;
2682 assert_se(IN_SET(arg_action
, ACTION_RUN
, ACTION_TEST
));
2684 /* Move out of the way, so that we won't block unmounts */
2685 assert_se(chdir("/") == 0);
2687 if (arg_action
== ACTION_RUN
) {
2689 /* A core pattern might have been specified via the cmdline. */
2690 initialize_core_pattern(skip_setup
);
2692 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2695 /* Remember open file descriptors for later deserialization */
2696 r
= collect_fds(&fds
, &error_message
);
2700 /* Give up any control of the console, but make sure its initialized. */
2701 setup_console_terminal(skip_setup
);
2703 /* Open the logging devices, if possible and necessary */
2707 log_execution_mode(&first_boot
);
2709 r
= initialize_runtime(skip_setup
,
2710 &saved_rlimit_nofile
,
2711 &saved_rlimit_memlock
,
2716 r
= manager_new(arg_system
? UNIT_FILE_SYSTEM
: UNIT_FILE_USER
,
2717 arg_action
== ACTION_TEST
? MANAGER_TEST_FULL
: 0,
2720 log_emergency_errno(r
, "Failed to allocate manager object: %m");
2721 error_message
= "Failed to allocate manager object";
2725 m
->timestamps
[MANAGER_TIMESTAMP_KERNEL
] = kernel_timestamp
;
2726 m
->timestamps
[MANAGER_TIMESTAMP_INITRD
] = initrd_timestamp
;
2727 m
->timestamps
[MANAGER_TIMESTAMP_USERSPACE
] = userspace_timestamp
;
2728 m
->timestamps
[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START
)] = security_start_timestamp
;
2729 m
->timestamps
[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH
)] = security_finish_timestamp
;
2731 set_manager_defaults(m
);
2732 set_manager_settings(m
);
2733 manager_set_first_boot(m
, first_boot
);
2735 /* Remember whether we should queue the default job */
2736 queue_default_job
= !arg_serialization
|| arg_switched_root
;
2738 before_startup
= now(CLOCK_MONOTONIC
);
2740 r
= manager_startup(m
, arg_serialization
, fds
);
2742 error_message
= "Failed to start up manager";
2746 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2747 fds
= fdset_free(fds
);
2748 arg_serialization
= safe_fclose(arg_serialization
);
2750 if (queue_default_job
) {
2751 r
= do_queue_default_job(m
, &error_message
);
2756 after_startup
= now(CLOCK_MONOTONIC
);
2758 log_full(arg_action
== ACTION_TEST
? LOG_INFO
: LOG_DEBUG
,
2759 "Loaded units and determined initial transaction in %s.",
2760 format_timespan(timespan
, sizeof(timespan
), after_startup
- before_startup
, 100 * USEC_PER_MSEC
));
2762 if (arg_action
== ACTION_TEST
) {
2764 retval
= EXIT_SUCCESS
;
2768 (void) invoke_main_loop(m
,
2781 arg_shutdown_watchdog
= m
->shutdown_watchdog
;
2782 m
= manager_free(m
);
2786 mac_selinux_finish();
2789 do_reexecute(argc
, argv
,
2790 &saved_rlimit_nofile
,
2791 &saved_rlimit_memlock
,
2795 &error_message
); /* This only returns if reexecution failed */
2797 arg_serialization
= safe_fclose(arg_serialization
);
2798 fds
= fdset_free(fds
);
2800 #if HAVE_VALGRIND_VALGRIND_H
2801 /* If we are PID 1 and running under valgrind, then let's exit
2802 * here explicitly. valgrind will only generate nice output on
2803 * exit(), not on exec(), hence let's do the former not the
2805 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND
) {
2806 /* Cleanup watchdog_device strings for valgrind. We need them
2807 * in become_shutdown() so normally we cannot free them yet. */
2808 watchdog_free_device();
2809 arg_watchdog_device
= mfree(arg_watchdog_device
);
2814 #if HAS_FEATURE_ADDRESS_SANITIZER
2815 __lsan_do_leak_check();
2818 if (shutdown_verb
) {
2819 r
= become_shutdown(shutdown_verb
, retval
);
2820 log_error_errno(r
, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
2821 error_message
= "Failed to execute shutdown binary";
2824 watchdog_free_device();
2825 arg_watchdog_device
= mfree(arg_watchdog_device
);
2827 if (getpid_cached() == 1) {
2829 manager_status_printf(NULL
, STATUS_TYPE_EMERGENCY
,
2830 ANSI_HIGHLIGHT_RED
"!!!!!!" ANSI_NORMAL
,
2831 "%s.", error_message
);
2832 freeze_or_exit_or_reboot();