]> git.ipfire.org Git - thirdparty/systemd.git/blob - src/core/main.c
projects / thirdparty / systemd.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge pull request #10428 from keszybz/failure-actions
[thirdparty/systemd.git] / src / core / main.c
1 /* SPDX-License-Identifier: LGPL-2.1+ */
2
3 #include <errno.h>
4 #include <fcntl.h>
5 #include <getopt.h>
6 #include <signal.h>
7 #include <stdio.h>
8 #include <string.h>
9 #include <sys/mount.h>
10 #include <sys/prctl.h>
11 #include <sys/reboot.h>
12 #include <sys/stat.h>
13 #include <unistd.h>
14 #if HAVE_SECCOMP
15 #include <seccomp.h>
16 #endif
17 #if HAVE_VALGRIND_VALGRIND_H
18 #include <valgrind/valgrind.h>
19 #endif
20
21 #include "sd-bus.h"
22 #include "sd-daemon.h"
23 #include "sd-messages.h"
24
25 #include "alloc-util.h"
26 #include "architecture.h"
27 #include "build.h"
28 #include "bus-error.h"
29 #include "bus-util.h"
30 #include "capability-util.h"
31 #include "clock-util.h"
32 #include "conf-parser.h"
33 #include "cpu-set-util.h"
34 #include "dbus.h"
35 #include "dbus-manager.h"
36 #include "def.h"
37 #include "emergency-action.h"
38 #include "env-util.h"
39 #include "exit-status.h"
40 #include "fd-util.h"
41 #include "fdset.h"
42 #include "fileio.h"
43 #include "format-util.h"
44 #include "fs-util.h"
45 #include "hostname-setup.h"
46 #include "ima-setup.h"
47 #include "killall.h"
48 #include "kmod-setup.h"
49 #include "load-fragment.h"
50 #include "log.h"
51 #include "loopback-setup.h"
52 #include "machine-id-setup.h"
53 #include "manager.h"
54 #include "missing.h"
55 #include "mount-setup.h"
56 #include "os-util.h"
57 #include "pager.h"
58 #include "parse-util.h"
59 #include "path-util.h"
60 #include "proc-cmdline.h"
61 #include "process-util.h"
62 #include "raw-clone.h"
63 #include "rlimit-util.h"
64 #if HAVE_SECCOMP
65 #include "seccomp-util.h"
66 #endif
67 #include "selinux-setup.h"
68 #include "selinux-util.h"
69 #include "signal-util.h"
70 #include "smack-setup.h"
71 #include "special.h"
72 #include "stat-util.h"
73 #include "stdio-util.h"
74 #include "strv.h"
75 #include "switch-root.h"
76 #include "sysctl-util.h"
77 #include "terminal-util.h"
78 #include "umask-util.h"
79 #include "user-util.h"
80 #include "util.h"
81 #include "virt.h"
82 #include "watchdog.h"
83
84 static enum {
85 ACTION_RUN,
86 ACTION_HELP,
87 ACTION_VERSION,
88 ACTION_TEST,
89 ACTION_DUMP_CONFIGURATION_ITEMS,
90 ACTION_DUMP_BUS_PROPERTIES,
91 } arg_action = ACTION_RUN;
92 static char *arg_default_unit = NULL;
93 static bool arg_system = false;
94 static bool arg_dump_core = true;
95 static int arg_crash_chvt = -1;
96 static bool arg_crash_shell = false;
97 static bool arg_crash_reboot = false;
98 static char *arg_confirm_spawn = NULL;
99 static ShowStatus arg_show_status = _SHOW_STATUS_INVALID;
100 static bool arg_switched_root = false;
101 static bool arg_no_pager = false;
102 static bool arg_service_watchdogs = true;
103 static char ***arg_join_controllers = NULL;
104 static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
105 static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
106 static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
107 static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
108 static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
109 static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
110 static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
111 static usec_t arg_runtime_watchdog = 0;
112 static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
113 static char *arg_early_core_pattern = NULL;
114 static char *arg_watchdog_device = NULL;
115 static char **arg_default_environment = NULL;
116 static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {};
117 static uint64_t arg_capability_bounding_set = CAP_ALL;
118 static bool arg_no_new_privs = false;
119 static nsec_t arg_timer_slack_nsec = NSEC_INFINITY;
120 static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
121 static Set* arg_syscall_archs = NULL;
122 static FILE* arg_serialization = NULL;
123 static bool arg_default_cpu_accounting = false;
124 static bool arg_default_io_accounting = false;
125 static bool arg_default_ip_accounting = false;
126 static bool arg_default_blockio_accounting = false;
127 static bool arg_default_memory_accounting = MEMORY_ACCOUNTING_DEFAULT;
128 static bool arg_default_tasks_accounting = true;
129 static uint64_t arg_default_tasks_max = UINT64_MAX;
130 static sd_id128_t arg_machine_id = {};
131 static EmergencyAction arg_cad_burst_action = EMERGENCY_ACTION_REBOOT_FORCE;
132
133 _noreturn_ static void freeze_or_reboot(void) {
134
135 if (arg_crash_reboot) {
136 log_notice("Rebooting in 10s...");
137 (void) sleep(10);
138
139 log_notice("Rebooting now...");
140 (void) reboot(RB_AUTOBOOT);
141 log_emergency_errno(errno, "Failed to reboot: %m");
142 }
143
144 log_emergency("Freezing execution.");
145 freeze();
146 }
147
148 _noreturn_ static void crash(int sig) {
149 struct sigaction sa;
150 pid_t pid;
151
152 if (getpid_cached() != 1)
153 /* Pass this on immediately, if this is not PID 1 */
154 (void) raise(sig);
155 else if (!arg_dump_core)
156 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
157 else {
158 sa = (struct sigaction) {
159 .sa_handler = nop_signal_handler,
160 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
161 };
162
163 /* We want to wait for the core process, hence let's enable SIGCHLD */
164 (void) sigaction(SIGCHLD, &sa, NULL);
165
166 pid = raw_clone(SIGCHLD);
167 if (pid < 0)
168 log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
169 else if (pid == 0) {
170 /* Enable default signal handler for core dump */
171
172 sa = (struct sigaction) {
173 .sa_handler = SIG_DFL,
174 };
175 (void) sigaction(sig, &sa, NULL);
176
177 /* Don't limit the coredump size */
178 (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
179
180 /* Just to be sure... */
181 (void) chdir("/");
182
183 /* Raise the signal again */
184 pid = raw_getpid();
185 (void) kill(pid, sig); /* raise() would kill the parent */
186
187 assert_not_reached("We shouldn't be here...");
188 _exit(EXIT_FAILURE);
189 } else {
190 siginfo_t status;
191 int r;
192
193 /* Order things nicely. */
194 r = wait_for_terminate(pid, &status);
195 if (r < 0)
196 log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
197 else if (status.si_code != CLD_DUMPED)
198 log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
199 signal_to_string(sig),
200 pid, sigchld_code_to_string(status.si_code),
201 status.si_status,
202 strna(status.si_code == CLD_EXITED
203 ? exit_status_to_string(status.si_status, EXIT_STATUS_MINIMAL)
204 : signal_to_string(status.si_status)));
205 else
206 log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
207 }
208 }
209
210 if (arg_crash_chvt >= 0)
211 (void) chvt(arg_crash_chvt);
212
213 sa = (struct sigaction) {
214 .sa_handler = SIG_IGN,
215 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
216 };
217
218 /* Let the kernel reap children for us */
219 (void) sigaction(SIGCHLD, &sa, NULL);
220
221 if (arg_crash_shell) {
222 log_notice("Executing crash shell in 10s...");
223 (void) sleep(10);
224
225 pid = raw_clone(SIGCHLD);
226 if (pid < 0)
227 log_emergency_errno(errno, "Failed to fork off crash shell: %m");
228 else if (pid == 0) {
229 (void) setsid();
230 (void) make_console_stdio();
231 (void) execle("/bin/sh", "/bin/sh", NULL, environ);
232
233 log_emergency_errno(errno, "execle() failed: %m");
234 _exit(EXIT_FAILURE);
235 } else {
236 log_info("Spawned crash shell as PID "PID_FMT".", pid);
237 (void) wait_for_terminate(pid, NULL);
238 }
239 }
240
241 freeze_or_reboot();
242 }
243
244 static void install_crash_handler(void) {
245 static const struct sigaction sa = {
246 .sa_handler = crash,
247 .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
248 };
249 int r;
250
251 /* We ignore the return value here, since, we don't mind if we
252 * cannot set up a crash handler */
253 r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
254 if (r < 0)
255 log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
256 }
257
258 static int console_setup(void) {
259 _cleanup_close_ int tty_fd = -1;
260 int r;
261
262 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
263 if (tty_fd < 0)
264 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
265
266 /* We don't want to force text mode. plymouth may be showing
267 * pictures already from initrd. */
268 r = reset_terminal_fd(tty_fd, false);
269 if (r < 0)
270 return log_error_errno(r, "Failed to reset /dev/console: %m");
271
272 return 0;
273 }
274
275 static int parse_crash_chvt(const char *value) {
276 int b;
277
278 if (safe_atoi(value, &arg_crash_chvt) >= 0)
279 return 0;
280
281 b = parse_boolean(value);
282 if (b < 0)
283 return b;
284
285 if (b > 0)
286 arg_crash_chvt = 0; /* switch to where kmsg goes */
287 else
288 arg_crash_chvt = -1; /* turn off switching */
289
290 return 0;
291 }
292
293 static int parse_confirm_spawn(const char *value, char **console) {
294 char *s;
295 int r;
296
297 r = value ? parse_boolean(value) : 1;
298 if (r == 0) {
299 *console = NULL;
300 return 0;
301 }
302
303 if (r > 0) /* on with default tty */
304 s = strdup("/dev/console");
305 else if (is_path(value)) /* on with fully qualified path */
306 s = strdup(value);
307 else /* on with only a tty file name, not a fully qualified path */
308 s = strjoin("/dev/", value);
309 if (!s)
310 return -ENOMEM;
311
312 *console = s;
313 return 0;
314 }
315
316 static int set_machine_id(const char *m) {
317 sd_id128_t t;
318 assert(m);
319
320 if (sd_id128_from_string(m, &t) < 0)
321 return -EINVAL;
322
323 if (sd_id128_is_null(t))
324 return -EINVAL;
325
326 arg_machine_id = t;
327 return 0;
328 }
329
330 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
331
332 int r;
333
334 assert(key);
335
336 if (STR_IN_SET(key, "systemd.unit", "rd.systemd.unit")) {
337
338 if (proc_cmdline_value_missing(key, value))
339 return 0;
340
341 if (!unit_name_is_valid(value, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE))
342 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key, value);
343 else if (in_initrd() == !!startswith(key, "rd.")) {
344 if (free_and_strdup(&arg_default_unit, value) < 0)
345 return log_oom();
346 }
347
348 } else if (proc_cmdline_key_streq(key, "systemd.dump_core")) {
349
350 r = value ? parse_boolean(value) : true;
351 if (r < 0)
352 log_warning("Failed to parse dump core switch %s. Ignoring.", value);
353 else
354 arg_dump_core = r;
355
356 } else if (proc_cmdline_key_streq(key, "systemd.early_core_pattern")) {
357
358 if (proc_cmdline_value_missing(key, value))
359 return 0;
360
361 if (path_is_absolute(value))
362 (void) parse_path_argument_and_warn(value, false, &arg_early_core_pattern);
363 else
364 log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value);
365
366 } else if (proc_cmdline_key_streq(key, "systemd.crash_chvt")) {
367
368 if (!value)
369 arg_crash_chvt = 0; /* turn on */
370 else if (parse_crash_chvt(value) < 0)
371 log_warning("Failed to parse crash chvt switch %s. Ignoring.", value);
372
373 } else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) {
374
375 r = value ? parse_boolean(value) : true;
376 if (r < 0)
377 log_warning("Failed to parse crash shell switch %s. Ignoring.", value);
378 else
379 arg_crash_shell = r;
380
381 } else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) {
382
383 r = value ? parse_boolean(value) : true;
384 if (r < 0)
385 log_warning("Failed to parse crash reboot switch %s. Ignoring.", value);
386 else
387 arg_crash_reboot = r;
388
389 } else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) {
390 char *s;
391
392 r = parse_confirm_spawn(value, &s);
393 if (r < 0)
394 log_warning_errno(r, "Failed to parse confirm_spawn switch %s. Ignoring.", value);
395 else {
396 free(arg_confirm_spawn);
397 arg_confirm_spawn = s;
398 }
399
400 } else if (proc_cmdline_key_streq(key, "systemd.service_watchdogs")) {
401
402 r = value ? parse_boolean(value) : true;
403 if (r < 0)
404 log_warning("Failed to parse service watchdog switch %s. Ignoring.", value);
405 else
406 arg_service_watchdogs = r;
407
408 } else if (proc_cmdline_key_streq(key, "systemd.show_status")) {
409
410 if (value) {
411 r = parse_show_status(value, &arg_show_status);
412 if (r < 0)
413 log_warning("Failed to parse show status switch %s. Ignoring.", value);
414 } else
415 arg_show_status = SHOW_STATUS_YES;
416
417 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) {
418
419 if (proc_cmdline_value_missing(key, value))
420 return 0;
421
422 r = exec_output_from_string(value);
423 if (r < 0)
424 log_warning("Failed to parse default standard output switch %s. Ignoring.", value);
425 else
426 arg_default_std_output = r;
427
428 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) {
429
430 if (proc_cmdline_value_missing(key, value))
431 return 0;
432
433 r = exec_output_from_string(value);
434 if (r < 0)
435 log_warning("Failed to parse default standard error switch %s. Ignoring.", value);
436 else
437 arg_default_std_error = r;
438
439 } else if (streq(key, "systemd.setenv")) {
440
441 if (proc_cmdline_value_missing(key, value))
442 return 0;
443
444 if (env_assignment_is_valid(value)) {
445 char **env;
446
447 env = strv_env_set(arg_default_environment, value);
448 if (!env)
449 return log_oom();
450
451 arg_default_environment = env;
452 } else
453 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
454
455 } else if (proc_cmdline_key_streq(key, "systemd.machine_id")) {
456
457 if (proc_cmdline_value_missing(key, value))
458 return 0;
459
460 r = set_machine_id(value);
461 if (r < 0)
462 log_warning("MachineID '%s' is not valid. Ignoring.", value);
463
464 } else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) {
465
466 if (proc_cmdline_value_missing(key, value))
467 return 0;
468
469 r = parse_sec(value, &arg_default_timeout_start_usec);
470 if (r < 0)
471 log_warning_errno(r, "Failed to parse default start timeout: %s, ignoring.", value);
472
473 if (arg_default_timeout_start_usec <= 0)
474 arg_default_timeout_start_usec = USEC_INFINITY;
475
476 } else if (proc_cmdline_key_streq(key, "systemd.watchdog_device")) {
477
478 if (proc_cmdline_value_missing(key, value))
479 return 0;
480
481 (void) parse_path_argument_and_warn(value, false, &arg_watchdog_device);
482
483 } else if (streq(key, "quiet") && !value) {
484
485 if (arg_show_status == _SHOW_STATUS_INVALID)
486 arg_show_status = SHOW_STATUS_AUTO;
487
488 } else if (streq(key, "debug") && !value) {
489
490 /* Note that log_parse_environment() handles 'debug'
491 * too, and sets the log level to LOG_DEBUG. */
492
493 if (detect_container() > 0)
494 log_set_target(LOG_TARGET_CONSOLE);
495
496 } else if (!value) {
497 const char *target;
498
499 /* SysV compatibility */
500 target = runlevel_to_target(key);
501 if (target)
502 return free_and_strdup(&arg_default_unit, target);
503 }
504
505 return 0;
506 }
507
508 #define DEFINE_SETTER(name, func, descr) \
509 static int name(const char *unit, \
510 const char *filename, \
511 unsigned line, \
512 const char *section, \
513 unsigned section_line, \
514 const char *lvalue, \
515 int ltype, \
516 const char *rvalue, \
517 void *data, \
518 void *userdata) { \
519 \
520 int r; \
521 \
522 assert(filename); \
523 assert(lvalue); \
524 assert(rvalue); \
525 \
526 r = func(rvalue); \
527 if (r < 0) \
528 log_syntax(unit, LOG_ERR, filename, line, r, \
529 "Invalid " descr "'%s': %m", \
530 rvalue); \
531 \
532 return 0; \
533 }
534
535 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level");
536 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target");
537 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" );
538 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location");
539
540 static int config_parse_cpu_affinity2(
541 const char *unit,
542 const char *filename,
543 unsigned line,
544 const char *section,
545 unsigned section_line,
546 const char *lvalue,
547 int ltype,
548 const char *rvalue,
549 void *data,
550 void *userdata) {
551
552 _cleanup_cpu_free_ cpu_set_t *c = NULL;
553 int ncpus;
554
555 ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue);
556 if (ncpus < 0)
557 return ncpus;
558
559 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
560 log_warning_errno(errno, "Failed to set CPU affinity: %m");
561
562 return 0;
563 }
564
565 static int config_parse_show_status(
566 const char* unit,
567 const char *filename,
568 unsigned line,
569 const char *section,
570 unsigned section_line,
571 const char *lvalue,
572 int ltype,
573 const char *rvalue,
574 void *data,
575 void *userdata) {
576
577 int k;
578 ShowStatus *b = data;
579
580 assert(filename);
581 assert(lvalue);
582 assert(rvalue);
583 assert(data);
584
585 k = parse_show_status(rvalue, b);
586 if (k < 0) {
587 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
588 return 0;
589 }
590
591 return 0;
592 }
593
594 static int config_parse_output_restricted(
595 const char* unit,
596 const char *filename,
597 unsigned line,
598 const char *section,
599 unsigned section_line,
600 const char *lvalue,
601 int ltype,
602 const char *rvalue,
603 void *data,
604 void *userdata) {
605
606 ExecOutput t, *eo = data;
607
608 assert(filename);
609 assert(lvalue);
610 assert(rvalue);
611 assert(data);
612
613 t = exec_output_from_string(rvalue);
614 if (t < 0) {
615 log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse output type, ignoring: %s", rvalue);
616 return 0;
617 }
618
619 if (IN_SET(t, EXEC_OUTPUT_SOCKET, EXEC_OUTPUT_NAMED_FD, EXEC_OUTPUT_FILE, EXEC_OUTPUT_FILE_APPEND)) {
620 log_syntax(unit, LOG_ERR, filename, line, 0, "Standard output types socket, fd:, file:, append: are not supported as defaults, ignoring: %s", rvalue);
621 return 0;
622 }
623
624 *eo = t;
625 return 0;
626 }
627
628 static int config_parse_crash_chvt(
629 const char* unit,
630 const char *filename,
631 unsigned line,
632 const char *section,
633 unsigned section_line,
634 const char *lvalue,
635 int ltype,
636 const char *rvalue,
637 void *data,
638 void *userdata) {
639
640 int r;
641
642 assert(filename);
643 assert(lvalue);
644 assert(rvalue);
645
646 r = parse_crash_chvt(rvalue);
647 if (r < 0) {
648 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
649 return 0;
650 }
651
652 return 0;
653 }
654
655 static int parse_config_file(void) {
656
657 const ConfigTableItem items[] = {
658 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
659 { "Manager", "LogTarget", config_parse_target, 0, NULL },
660 { "Manager", "LogColor", config_parse_color, 0, NULL },
661 { "Manager", "LogLocation", config_parse_location, 0, NULL },
662 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
663 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL },
664 { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL },
665 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
666 { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
667 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
668 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
669 { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
670 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
671 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
672 { "Manager", "WatchdogDevice", config_parse_path, 0, &arg_watchdog_device },
673 { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
674 { "Manager", "NoNewPrivileges", config_parse_bool, 0, &arg_no_new_privs },
675 #if HAVE_SECCOMP
676 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
677 #endif
678 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
679 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
680 { "Manager", "DefaultStandardOutput", config_parse_output_restricted,0, &arg_default_std_output },
681 { "Manager", "DefaultStandardError", config_parse_output_restricted,0, &arg_default_std_error },
682 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
683 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
684 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
685 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */
686 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec, 0, &arg_default_start_limit_interval },
687 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
688 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
689 { "Manager", "DefaultLimitCPU", config_parse_rlimit, RLIMIT_CPU, arg_default_rlimit },
690 { "Manager", "DefaultLimitFSIZE", config_parse_rlimit, RLIMIT_FSIZE, arg_default_rlimit },
691 { "Manager", "DefaultLimitDATA", config_parse_rlimit, RLIMIT_DATA, arg_default_rlimit },
692 { "Manager", "DefaultLimitSTACK", config_parse_rlimit, RLIMIT_STACK, arg_default_rlimit },
693 { "Manager", "DefaultLimitCORE", config_parse_rlimit, RLIMIT_CORE, arg_default_rlimit },
694 { "Manager", "DefaultLimitRSS", config_parse_rlimit, RLIMIT_RSS, arg_default_rlimit },
695 { "Manager", "DefaultLimitNOFILE", config_parse_rlimit, RLIMIT_NOFILE, arg_default_rlimit },
696 { "Manager", "DefaultLimitAS", config_parse_rlimit, RLIMIT_AS, arg_default_rlimit },
697 { "Manager", "DefaultLimitNPROC", config_parse_rlimit, RLIMIT_NPROC, arg_default_rlimit },
698 { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit, RLIMIT_MEMLOCK, arg_default_rlimit },
699 { "Manager", "DefaultLimitLOCKS", config_parse_rlimit, RLIMIT_LOCKS, arg_default_rlimit },
700 { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit, RLIMIT_SIGPENDING, arg_default_rlimit },
701 { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit, RLIMIT_MSGQUEUE, arg_default_rlimit },
702 { "Manager", "DefaultLimitNICE", config_parse_rlimit, RLIMIT_NICE, arg_default_rlimit },
703 { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit, RLIMIT_RTPRIO, arg_default_rlimit },
704 { "Manager", "DefaultLimitRTTIME", config_parse_rlimit, RLIMIT_RTTIME, arg_default_rlimit },
705 { "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_default_cpu_accounting },
706 { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting },
707 { "Manager", "DefaultIPAccounting", config_parse_bool, 0, &arg_default_ip_accounting },
708 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
709 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
710 { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
711 { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max },
712 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action, 0, &arg_cad_burst_action },
713 {}
714 };
715
716 const char *fn, *conf_dirs_nulstr;
717
718 fn = arg_system ?
719 PKGSYSCONFDIR "/system.conf" :
720 PKGSYSCONFDIR "/user.conf";
721
722 conf_dirs_nulstr = arg_system ?
723 CONF_PATHS_NULSTR("systemd/system.conf.d") :
724 CONF_PATHS_NULSTR("systemd/user.conf.d");
725
726 (void) config_parse_many_nulstr(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, CONFIG_PARSE_WARN, NULL);
727
728 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
729 * like everywhere else. */
730 if (arg_default_timeout_start_usec <= 0)
731 arg_default_timeout_start_usec = USEC_INFINITY;
732 if (arg_default_timeout_stop_usec <= 0)
733 arg_default_timeout_stop_usec = USEC_INFINITY;
734
735 return 0;
736 }
737
738 static void set_manager_defaults(Manager *m) {
739
740 assert(m);
741
742 /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
743 * affect the manager itself, but are just what newly allocated units will have set if they haven't set
744 * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
745
746 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
747 m->default_std_output = arg_default_std_output;
748 m->default_std_error = arg_default_std_error;
749 m->default_timeout_start_usec = arg_default_timeout_start_usec;
750 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
751 m->default_restart_usec = arg_default_restart_usec;
752 m->default_start_limit_interval = arg_default_start_limit_interval;
753 m->default_start_limit_burst = arg_default_start_limit_burst;
754 m->default_cpu_accounting = arg_default_cpu_accounting;
755 m->default_io_accounting = arg_default_io_accounting;
756 m->default_ip_accounting = arg_default_ip_accounting;
757 m->default_blockio_accounting = arg_default_blockio_accounting;
758 m->default_memory_accounting = arg_default_memory_accounting;
759 m->default_tasks_accounting = arg_default_tasks_accounting;
760 m->default_tasks_max = arg_default_tasks_max;
761
762 manager_set_default_rlimits(m, arg_default_rlimit);
763 manager_environment_add(m, NULL, arg_default_environment);
764 }
765
766 static void set_manager_settings(Manager *m) {
767
768 assert(m);
769
770 /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
771 * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
772
773 m->confirm_spawn = arg_confirm_spawn;
774 m->service_watchdogs = arg_service_watchdogs;
775 m->runtime_watchdog = arg_runtime_watchdog;
776 m->shutdown_watchdog = arg_shutdown_watchdog;
777 m->cad_burst_action = arg_cad_burst_action;
778
779 manager_set_show_status(m, arg_show_status);
780 }
781
782 static int parse_argv(int argc, char *argv[]) {
783 enum {
784 ARG_LOG_LEVEL = 0x100,
785 ARG_LOG_TARGET,
786 ARG_LOG_COLOR,
787 ARG_LOG_LOCATION,
788 ARG_UNIT,
789 ARG_SYSTEM,
790 ARG_USER,
791 ARG_TEST,
792 ARG_NO_PAGER,
793 ARG_VERSION,
794 ARG_DUMP_CONFIGURATION_ITEMS,
795 ARG_DUMP_BUS_PROPERTIES,
796 ARG_DUMP_CORE,
797 ARG_CRASH_CHVT,
798 ARG_CRASH_SHELL,
799 ARG_CRASH_REBOOT,
800 ARG_CONFIRM_SPAWN,
801 ARG_SHOW_STATUS,
802 ARG_DESERIALIZE,
803 ARG_SWITCHED_ROOT,
804 ARG_DEFAULT_STD_OUTPUT,
805 ARG_DEFAULT_STD_ERROR,
806 ARG_MACHINE_ID,
807 ARG_SERVICE_WATCHDOGS,
808 };
809
810 static const struct option options[] = {
811 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
812 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
813 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
814 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
815 { "unit", required_argument, NULL, ARG_UNIT },
816 { "system", no_argument, NULL, ARG_SYSTEM },
817 { "user", no_argument, NULL, ARG_USER },
818 { "test", no_argument, NULL, ARG_TEST },
819 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
820 { "help", no_argument, NULL, 'h' },
821 { "version", no_argument, NULL, ARG_VERSION },
822 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
823 { "dump-bus-properties", no_argument, NULL, ARG_DUMP_BUS_PROPERTIES },
824 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
825 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
826 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
827 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
828 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
829 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
830 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
831 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
832 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
833 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
834 { "machine-id", required_argument, NULL, ARG_MACHINE_ID },
835 { "service-watchdogs", required_argument, NULL, ARG_SERVICE_WATCHDOGS },
836 {}
837 };
838
839 int c, r;
840
841 assert(argc >= 1);
842 assert(argv);
843
844 if (getpid_cached() == 1)
845 opterr = 0;
846
847 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
848
849 switch (c) {
850
851 case ARG_LOG_LEVEL:
852 r = log_set_max_level_from_string(optarg);
853 if (r < 0)
854 return log_error_errno(r, "Failed to parse log level \"%s\": %m", optarg);
855
856 break;
857
858 case ARG_LOG_TARGET:
859 r = log_set_target_from_string(optarg);
860 if (r < 0)
861 return log_error_errno(r, "Failed to parse log target \"%s\": %m", optarg);
862
863 break;
864
865 case ARG_LOG_COLOR:
866
867 if (optarg) {
868 r = log_show_color_from_string(optarg);
869 if (r < 0)
870 return log_error_errno(r, "Failed to parse log color setting \"%s\": %m",
871 optarg);
872 } else
873 log_show_color(true);
874
875 break;
876
877 case ARG_LOG_LOCATION:
878 if (optarg) {
879 r = log_show_location_from_string(optarg);
880 if (r < 0)
881 return log_error_errno(r, "Failed to parse log location setting \"%s\": %m",
882 optarg);
883 } else
884 log_show_location(true);
885
886 break;
887
888 case ARG_DEFAULT_STD_OUTPUT:
889 r = exec_output_from_string(optarg);
890 if (r < 0)
891 return log_error_errno(r, "Failed to parse default standard output setting \"%s\": %m",
892 optarg);
893 arg_default_std_output = r;
894 break;
895
896 case ARG_DEFAULT_STD_ERROR:
897 r = exec_output_from_string(optarg);
898 if (r < 0)
899 return log_error_errno(r, "Failed to parse default standard error output setting \"%s\": %m",
900 optarg);
901 arg_default_std_error = r;
902 break;
903
904 case ARG_UNIT:
905 r = free_and_strdup(&arg_default_unit, optarg);
906 if (r < 0)
907 return log_error_errno(r, "Failed to set default unit \"%s\": %m", optarg);
908
909 break;
910
911 case ARG_SYSTEM:
912 arg_system = true;
913 break;
914
915 case ARG_USER:
916 arg_system = false;
917 break;
918
919 case ARG_TEST:
920 arg_action = ACTION_TEST;
921 break;
922
923 case ARG_NO_PAGER:
924 arg_no_pager = true;
925 break;
926
927 case ARG_VERSION:
928 arg_action = ACTION_VERSION;
929 break;
930
931 case ARG_DUMP_CONFIGURATION_ITEMS:
932 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
933 break;
934
935 case ARG_DUMP_BUS_PROPERTIES:
936 arg_action = ACTION_DUMP_BUS_PROPERTIES;
937 break;
938
939 case ARG_DUMP_CORE:
940 if (!optarg)
941 arg_dump_core = true;
942 else {
943 r = parse_boolean(optarg);
944 if (r < 0)
945 return log_error_errno(r, "Failed to parse dump core boolean: \"%s\": %m",
946 optarg);
947 arg_dump_core = r;
948 }
949 break;
950
951 case ARG_CRASH_CHVT:
952 r = parse_crash_chvt(optarg);
953 if (r < 0)
954 return log_error_errno(r, "Failed to parse crash virtual terminal index: \"%s\": %m",
955 optarg);
956 break;
957
958 case ARG_CRASH_SHELL:
959 if (!optarg)
960 arg_crash_shell = true;
961 else {
962 r = parse_boolean(optarg);
963 if (r < 0)
964 return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
965 optarg);
966 arg_crash_shell = r;
967 }
968 break;
969
970 case ARG_CRASH_REBOOT:
971 if (!optarg)
972 arg_crash_reboot = true;
973 else {
974 r = parse_boolean(optarg);
975 if (r < 0)
976 return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
977 optarg);
978 arg_crash_reboot = r;
979 }
980 break;
981
982 case ARG_CONFIRM_SPAWN:
983 arg_confirm_spawn = mfree(arg_confirm_spawn);
984
985 r = parse_confirm_spawn(optarg, &arg_confirm_spawn);
986 if (r < 0)
987 return log_error_errno(r, "Failed to parse confirm spawn option: \"%s\": %m",
988 optarg);
989 break;
990
991 case ARG_SERVICE_WATCHDOGS:
992 r = parse_boolean(optarg);
993 if (r < 0)
994 return log_error_errno(r, "Failed to parse service watchdogs boolean: \"%s\": %m",
995 optarg);
996 arg_service_watchdogs = r;
997 break;
998
999 case ARG_SHOW_STATUS:
1000 if (optarg) {
1001 r = parse_show_status(optarg, &arg_show_status);
1002 if (r < 0)
1003 return log_error_errno(r, "Failed to parse show status boolean: \"%s\": %m",
1004 optarg);
1005 } else
1006 arg_show_status = SHOW_STATUS_YES;
1007 break;
1008
1009 case ARG_DESERIALIZE: {
1010 int fd;
1011 FILE *f;
1012
1013 r = safe_atoi(optarg, &fd);
1014 if (r < 0)
1015 log_error_errno(r, "Failed to parse deserialize option \"%s\": %m", optarg);
1016 if (fd < 0) {
1017 log_error("Invalid deserialize fd: %d", fd);
1018 return -EINVAL;
1019 }
1020
1021 (void) fd_cloexec(fd, true);
1022
1023 f = fdopen(fd, "r");
1024 if (!f)
1025 return log_error_errno(errno, "Failed to open serialization fd %d: %m", fd);
1026
1027 safe_fclose(arg_serialization);
1028 arg_serialization = f;
1029
1030 break;
1031 }
1032
1033 case ARG_SWITCHED_ROOT:
1034 arg_switched_root = true;
1035 break;
1036
1037 case ARG_MACHINE_ID:
1038 r = set_machine_id(optarg);
1039 if (r < 0)
1040 return log_error_errno(r, "MachineID '%s' is not valid: %m", optarg);
1041 break;
1042
1043 case 'h':
1044 arg_action = ACTION_HELP;
1045 break;
1046
1047 case 'D':
1048 log_set_max_level(LOG_DEBUG);
1049 break;
1050
1051 case 'b':
1052 case 's':
1053 case 'z':
1054 /* Just to eat away the sysvinit kernel
1055 * cmdline args without getopt() error
1056 * messages that we'll parse in
1057 * parse_proc_cmdline_word() or ignore. */
1058
1059 case '?':
1060 if (getpid_cached() != 1)
1061 return -EINVAL;
1062 else
1063 return 0;
1064
1065 default:
1066 assert_not_reached("Unhandled option code.");
1067 }
1068
1069 if (optind < argc && getpid_cached() != 1) {
1070 /* Hmm, when we aren't run as init system
1071 * let's complain about excess arguments */
1072
1073 log_error("Excess arguments.");
1074 return -EINVAL;
1075 }
1076
1077 return 0;
1078 }
1079
1080 static int help(void) {
1081 _cleanup_free_ char *link = NULL;
1082 int r;
1083
1084 r = terminal_urlify_man("systemd", "1", &link);
1085 if (r < 0)
1086 return log_oom();
1087
1088 printf("%s [OPTIONS...]\n\n"
1089 "Starts up and maintains the system or user services.\n\n"
1090 " -h --help Show this help\n"
1091 " --version Show version\n"
1092 " --test Determine startup sequence, dump it and exit\n"
1093 " --no-pager Do not pipe output into a pager\n"
1094 " --dump-configuration-items Dump understood unit configuration items\n"
1095 " --dump-bus-properties Dump exposed bus properties\n"
1096 " --unit=UNIT Set default unit\n"
1097 " --system Run a system instance, even if PID != 1\n"
1098 " --user Run a user instance\n"
1099 " --dump-core[=BOOL] Dump core on crash\n"
1100 " --crash-vt=NR Change to specified VT on crash\n"
1101 " --crash-reboot[=BOOL] Reboot on crash\n"
1102 " --crash-shell[=BOOL] Run shell on crash\n"
1103 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1104 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1105 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1106 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1107 " --log-color[=BOOL] Highlight important log messages\n"
1108 " --log-location[=BOOL] Include code location in log messages\n"
1109 " --default-standard-output= Set default standard output for services\n"
1110 " --default-standard-error= Set default standard error output for services\n"
1111 "\nSee the %s for details.\n"
1112 , program_invocation_short_name
1113 , link
1114 );
1115
1116 return 0;
1117 }
1118
1119 static int prepare_reexecute(
1120 Manager *m,
1121 FILE **ret_f,
1122 FDSet **ret_fds,
1123 bool switching_root) {
1124
1125 _cleanup_fdset_free_ FDSet *fds = NULL;
1126 _cleanup_fclose_ FILE *f = NULL;
1127 int r;
1128
1129 assert(m);
1130 assert(ret_f);
1131 assert(ret_fds);
1132
1133 r = manager_open_serialization(m, &f);
1134 if (r < 0)
1135 return log_error_errno(r, "Failed to create serialization file: %m");
1136
1137 /* Make sure nothing is really destructed when we shut down */
1138 m->n_reloading++;
1139 bus_manager_send_reloading(m, true);
1140
1141 fds = fdset_new();
1142 if (!fds)
1143 return log_oom();
1144
1145 r = manager_serialize(m, f, fds, switching_root);
1146 if (r < 0)
1147 return log_error_errno(r, "Failed to serialize state: %m");
1148
1149 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1150 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1151
1152 r = fd_cloexec(fileno(f), false);
1153 if (r < 0)
1154 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1155
1156 r = fdset_cloexec(fds, false);
1157 if (r < 0)
1158 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1159
1160 *ret_f = TAKE_PTR(f);
1161 *ret_fds = TAKE_PTR(fds);
1162
1163 return 0;
1164 }
1165
1166 static void bump_file_max_and_nr_open(void) {
1167
1168 /* Let's bump fs.file-max and fs.nr_open to their respective maximums. On current kernels large numbers of file
1169 * descriptors are no longer a performance problem and their memory is properly tracked by memcg, thus counting
1170 * them and limiting them in another two layers of limits is unnecessary and just complicates things. This
1171 * function hence turns off 2 of the 4 levels of limits on file descriptors, and makes RLIMIT_NOLIMIT (soft +
1172 * hard) the only ones that really matter. */
1173
1174 #if BUMP_PROC_SYS_FS_FILE_MAX || BUMP_PROC_SYS_FS_NR_OPEN
1175 _cleanup_free_ char *t = NULL;
1176 int r;
1177 #endif
1178
1179 #if BUMP_PROC_SYS_FS_FILE_MAX
1180 /* I so wanted to use STRINGIFY(ULONG_MAX) here, but alas we can't as glibc/gcc define that as
1181 * "(0x7fffffffffffffffL * 2UL + 1UL)". Seriously. 😢 */
1182 if (asprintf(&t, "%lu\n", ULONG_MAX) < 0) {
1183 log_oom();
1184 return;
1185 }
1186
1187 r = sysctl_write("fs/file-max", t);
1188 if (r < 0)
1189 log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.file-max, ignoring: %m");
1190 #endif
1191
1192 #if BUMP_PROC_SYS_FS_FILE_MAX && BUMP_PROC_SYS_FS_NR_OPEN
1193 t = mfree(t);
1194 #endif
1195
1196 #if BUMP_PROC_SYS_FS_NR_OPEN
1197 int v = INT_MAX;
1198
1199 /* Arg! The kernel enforces maximum and minimum values on the fs.nr_open, but we don't really know what they
1200 * are. The expression by which the maximum is determined is dependent on the architecture, and is something we
1201 * don't really want to copy to userspace, as it is dependent on implementation details of the kernel. Since
1202 * the kernel doesn't expose the maximum value to us, we can only try and hope. Hence, let's start with
1203 * INT_MAX, and then keep halving the value until we find one that works. Ugly? Yes, absolutely, but kernel
1204 * APIs are kernel APIs, so what do can we do... 🤯 */
1205
1206 for (;;) {
1207 int k;
1208
1209 v &= ~(__SIZEOF_POINTER__ - 1); /* Round down to next multiple of the pointer size */
1210 if (v < 1024) {
1211 log_warning("Can't bump fs.nr_open, value too small.");
1212 break;
1213 }
1214
1215 k = read_nr_open();
1216 if (k < 0) {
1217 log_error_errno(k, "Failed to read fs.nr_open: %m");
1218 break;
1219 }
1220 if (k >= v) { /* Already larger */
1221 log_debug("Skipping bump, value is already larger.");
1222 break;
1223 }
1224
1225 if (asprintf(&t, "%i\n", v) < 0) {
1226 log_oom();
1227 return;
1228 }
1229
1230 r = sysctl_write("fs/nr_open", t);
1231 t = mfree(t);
1232 if (r == -EINVAL) {
1233 log_debug("Couldn't write fs.nr_open as %i, halving it.", v);
1234 v /= 2;
1235 continue;
1236 }
1237 if (r < 0) {
1238 log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.nr_open, ignoring: %m");
1239 break;
1240 }
1241
1242 log_debug("Successfully bumped fs.nr_open to %i", v);
1243 break;
1244 }
1245 #endif
1246 }
1247
1248 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1249 int r, nr;
1250
1251 assert(saved_rlimit);
1252
1253 /* Save the original RLIMIT_NOFILE so that we can reset it later when transitioning from the initrd to the main
1254 * systemd or suchlike. */
1255 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1256 return log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1257
1258 /* Get the underlying absolute limit the kernel enforces */
1259 nr = read_nr_open();
1260
1261 /* Make sure forked processes get limits based on the original kernel setting */
1262 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1263 struct rlimit *rl;
1264
1265 rl = newdup(struct rlimit, saved_rlimit, 1);
1266 if (!rl)
1267 return log_oom();
1268
1269 /* Bump the hard limit for system services to a substantially higher value. The default hard limit
1270 * current kernels set is pretty low (4K), mostly for historical reasons. According to kernel
1271 * developers, the fd handling in recent kernels has been optimized substantially enough, so that we
1272 * can bump the limit now, without paying too high a price in memory or performance. Note however that
1273 * we only bump the hard limit, not the soft limit. That's because select() works the way it works, and
1274 * chokes on fds >= 1024. If we'd bump the soft limit globally, it might accidentally happen to
1275 * unexpecting programs that they get fds higher than what they can process using select(). By only
1276 * bumping the hard limit but leaving the low limit as it is we avoid this pitfall: programs that are
1277 * written by folks aware of the select() problem in mind (and thus use poll()/epoll instead of
1278 * select(), the way everybody should) can explicitly opt into high fds by bumping their soft limit
1279 * beyond 1024, to the hard limit we pass. */
1280 if (arg_system)
1281 rl->rlim_max = MIN((rlim_t) nr, MAX(rl->rlim_max, (rlim_t) HIGH_RLIMIT_NOFILE));
1282
1283 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1284 }
1285
1286 /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for
1287 * both hard and soft. */
1288 r = setrlimit_closest(RLIMIT_NOFILE, &RLIMIT_MAKE_CONST(nr));
1289 if (r < 0)
1290 return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1291
1292 return 0;
1293 }
1294
1295 static int bump_rlimit_memlock(struct rlimit *saved_rlimit) {
1296 int r;
1297
1298 assert(saved_rlimit);
1299
1300 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK which should
1301 * normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's bump
1302 * the value high enough for our user. */
1303
1304 if (getrlimit(RLIMIT_MEMLOCK, saved_rlimit) < 0)
1305 return log_warning_errno(errno, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1306
1307 r = setrlimit_closest(RLIMIT_MEMLOCK, &RLIMIT_MAKE_CONST(HIGH_RLIMIT_MEMLOCK));
1308 if (r < 0)
1309 return log_warning_errno(r, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1310
1311 return 0;
1312 }
1313
1314 static void test_usr(void) {
1315
1316 /* Check that /usr is not a separate fs */
1317
1318 if (dir_is_empty("/usr") <= 0)
1319 return;
1320
1321 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1322 "Some things will probably break (sometimes even silently) in mysterious ways. "
1323 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1324 }
1325
1326 static int enforce_syscall_archs(Set *archs) {
1327 #if HAVE_SECCOMP
1328 int r;
1329
1330 if (!is_seccomp_available())
1331 return 0;
1332
1333 r = seccomp_restrict_archs(arg_syscall_archs);
1334 if (r < 0)
1335 return log_error_errno(r, "Failed to enforce system call architecture restrication: %m");
1336 #endif
1337 return 0;
1338 }
1339
1340 static int status_welcome(void) {
1341 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1342 int r;
1343
1344 if (IN_SET(arg_show_status, SHOW_STATUS_NO, SHOW_STATUS_AUTO))
1345 return 0;
1346
1347 r = parse_os_release(NULL,
1348 "PRETTY_NAME", &pretty_name,
1349 "ANSI_COLOR", &ansi_color,
1350 NULL);
1351 if (r < 0)
1352 log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r,
1353 "Failed to read os-release file, ignoring: %m");
1354
1355 if (log_get_show_color())
1356 return status_printf(NULL, false, false,
1357 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1358 isempty(ansi_color) ? "1" : ansi_color,
1359 isempty(pretty_name) ? "Linux" : pretty_name);
1360 else
1361 return status_printf(NULL, false, false,
1362 "\nWelcome to %s!\n",
1363 isempty(pretty_name) ? "Linux" : pretty_name);
1364 }
1365
1366 static int write_container_id(void) {
1367 const char *c;
1368 int r;
1369
1370 c = getenv("container");
1371 if (isempty(c))
1372 return 0;
1373
1374 RUN_WITH_UMASK(0022)
1375 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1376 if (r < 0)
1377 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
1378
1379 return 1;
1380 }
1381
1382 static int bump_unix_max_dgram_qlen(void) {
1383 _cleanup_free_ char *qlen = NULL;
1384 unsigned long v;
1385 int r;
1386
1387 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1388 * really really early during boot, so that it is actually applied to all our sockets, including the
1389 * $NOTIFY_SOCKET one. */
1390
1391 r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1392 if (r < 0)
1393 return log_warning_errno(r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1394
1395 r = safe_atolu(qlen, &v);
1396 if (r < 0)
1397 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen);
1398
1399 if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1400 return 0;
1401
1402 r = write_string_filef("/proc/sys/net/unix/max_dgram_qlen", 0, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN);
1403 if (r < 0)
1404 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1405 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1406
1407 return 1;
1408 }
1409
1410 static int fixup_environment(void) {
1411 _cleanup_free_ char *term = NULL;
1412 const char *t;
1413 int r;
1414
1415 /* Only fix up the environment when we are started as PID 1 */
1416 if (getpid_cached() != 1)
1417 return 0;
1418
1419 /* We expect the environment to be set correctly if run inside a container. */
1420 if (detect_container() > 0)
1421 return 0;
1422
1423 /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1424 * device used by the console. We try to make a better guess here since some consoles might not have support
1425 * for color mode for example.
1426 *
1427 * However if TERM was configured through the kernel command line then leave it alone. */
1428 r = proc_cmdline_get_key("TERM", 0, &term);
1429 if (r < 0)
1430 return r;
1431
1432 t = term ?: default_term_for_tty("/dev/console");
1433
1434 if (setenv("TERM", t, 1) < 0)
1435 return -errno;
1436
1437 return 0;
1438 }
1439
1440 static void redirect_telinit(int argc, char *argv[]) {
1441
1442 /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1443
1444 #if HAVE_SYSV_COMPAT
1445 if (getpid_cached() == 1)
1446 return;
1447
1448 if (!strstr(program_invocation_short_name, "init"))
1449 return;
1450
1451 execv(SYSTEMCTL_BINARY_PATH, argv);
1452 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1453 exit(EXIT_FAILURE);
1454 #endif
1455 }
1456
1457 static int become_shutdown(
1458 const char *shutdown_verb,
1459 int retval) {
1460
1461 char log_level[DECIMAL_STR_MAX(int) + 1],
1462 exit_code[DECIMAL_STR_MAX(uint8_t) + 1],
1463 timeout[DECIMAL_STR_MAX(usec_t) + 1];
1464
1465 const char* command_line[13] = {
1466 SYSTEMD_SHUTDOWN_BINARY_PATH,
1467 shutdown_verb,
1468 "--timeout", timeout,
1469 "--log-level", log_level,
1470 "--log-target",
1471 };
1472
1473 _cleanup_strv_free_ char **env_block = NULL;
1474 size_t pos = 7;
1475 int r;
1476
1477 assert(shutdown_verb);
1478 assert(!command_line[pos]);
1479 env_block = strv_copy(environ);
1480
1481 xsprintf(log_level, "%d", log_get_max_level());
1482 xsprintf(timeout, "%" PRI_USEC "us", arg_default_timeout_stop_usec);
1483
1484 switch (log_get_target()) {
1485
1486 case LOG_TARGET_KMSG:
1487 case LOG_TARGET_JOURNAL_OR_KMSG:
1488 case LOG_TARGET_SYSLOG_OR_KMSG:
1489 command_line[pos++] = "kmsg";
1490 break;
1491
1492 case LOG_TARGET_NULL:
1493 command_line[pos++] = "null";
1494 break;
1495
1496 case LOG_TARGET_CONSOLE:
1497 default:
1498 command_line[pos++] = "console";
1499 break;
1500 };
1501
1502 if (log_get_show_color())
1503 command_line[pos++] = "--log-color";
1504
1505 if (log_get_show_location())
1506 command_line[pos++] = "--log-location";
1507
1508 if (streq(shutdown_verb, "exit")) {
1509 command_line[pos++] = "--exit-code";
1510 command_line[pos++] = exit_code;
1511 xsprintf(exit_code, "%d", retval);
1512 }
1513
1514 assert(pos < ELEMENTSOF(command_line));
1515
1516 if (streq(shutdown_verb, "reboot") &&
1517 arg_shutdown_watchdog > 0 &&
1518 arg_shutdown_watchdog != USEC_INFINITY) {
1519
1520 char *e;
1521
1522 /* If we reboot let's set the shutdown
1523 * watchdog and tell the shutdown binary to
1524 * repeatedly ping it */
1525 r = watchdog_set_timeout(&arg_shutdown_watchdog);
1526 watchdog_close(r < 0);
1527
1528 /* Tell the binary how often to ping, ignore failure */
1529 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
1530 (void) strv_consume(&env_block, e);
1531
1532 if (arg_watchdog_device &&
1533 asprintf(&e, "WATCHDOG_DEVICE=%s", arg_watchdog_device) > 0)
1534 (void) strv_consume(&env_block, e);
1535 } else
1536 watchdog_close(true);
1537
1538 /* Avoid the creation of new processes forked by the
1539 * kernel; at this point, we will not listen to the
1540 * signals anyway */
1541 if (detect_container() <= 0)
1542 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
1543
1544 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
1545 return -errno;
1546 }
1547
1548 static void initialize_clock(void) {
1549 int r;
1550
1551 if (clock_is_localtime(NULL) > 0) {
1552 int min;
1553
1554 /*
1555 * The very first call of settimeofday() also does a time warp in the kernel.
1556 *
1557 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1558 * of maintaining the RTC and do all adjustments. This matches the behavior of Windows, which leaves
1559 * the RTC alone if the registry tells that the RTC runs in UTC.
1560 */
1561 r = clock_set_timezone(&min);
1562 if (r < 0)
1563 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1564 else
1565 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1566
1567 } else if (!in_initrd()) {
1568 /*
1569 * Do a dummy very first call to seal the kernel's time warp magic.
1570 *
1571 * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1572 * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1573 * until we reach the real system.
1574 *
1575 * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1576 * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1577 * be treated as UTC that way.
1578 */
1579 (void) clock_reset_timewarp();
1580 }
1581
1582 r = clock_apply_epoch();
1583 if (r < 0)
1584 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1585 else if (r > 0)
1586 log_info("System time before build time, advancing clock.");
1587 }
1588
1589 static void initialize_coredump(bool skip_setup) {
1590 #if ENABLE_COREDUMP
1591 if (getpid_cached() != 1)
1592 return;
1593
1594 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1595 * will process core dumps for system services by default. */
1596 if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0)
1597 log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m");
1598
1599 /* But at the same time, turn off the core_pattern logic by default, so that no
1600 * coredumps are stored until the systemd-coredump tool is enabled via
1601 * sysctl. However it can be changed via the kernel command line later so core
1602 * dumps can still be generated during early startup and in initramfs. */
1603 if (!skip_setup)
1604 disable_coredumps();
1605 #endif
1606 }
1607
1608 static void initialize_core_pattern(bool skip_setup) {
1609 int r;
1610
1611 if (skip_setup || !arg_early_core_pattern)
1612 return;
1613
1614 if (getpid_cached() != 1)
1615 return;
1616
1617 r = write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern, 0);
1618 if (r < 0)
1619 log_warning_errno(r, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", arg_early_core_pattern);
1620 }
1621
1622 static void do_reexecute(
1623 int argc,
1624 char *argv[],
1625 const struct rlimit *saved_rlimit_nofile,
1626 const struct rlimit *saved_rlimit_memlock,
1627 FDSet *fds,
1628 const char *switch_root_dir,
1629 const char *switch_root_init,
1630 const char **ret_error_message) {
1631
1632 unsigned i, j, args_size;
1633 const char **args;
1634 int r;
1635
1636 assert(saved_rlimit_nofile);
1637 assert(saved_rlimit_memlock);
1638 assert(ret_error_message);
1639
1640 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1641 * we do that */
1642 watchdog_close(true);
1643
1644 /* Reset the RLIMIT_NOFILE to the kernel default, so that the new systemd can pass the kernel default to its
1645 * child processes */
1646
1647 if (saved_rlimit_nofile->rlim_cur > 0)
1648 (void) setrlimit(RLIMIT_NOFILE, saved_rlimit_nofile);
1649 if (saved_rlimit_memlock->rlim_cur != (rlim_t) -1)
1650 (void) setrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock);
1651
1652 if (switch_root_dir) {
1653 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1654 * SIGCHLD for them after deserializing. */
1655 broadcast_signal(SIGTERM, false, true, arg_default_timeout_stop_usec);
1656
1657 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1658 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1659 if (r < 0)
1660 log_error_errno(r, "Failed to switch root, trying to continue: %m");
1661 }
1662
1663 args_size = MAX(6, argc+1);
1664 args = newa(const char*, args_size);
1665
1666 if (!switch_root_init) {
1667 char sfd[DECIMAL_STR_MAX(int) + 1];
1668
1669 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1670 * the user didn't specify an explicit init to spawn. */
1671
1672 assert(arg_serialization);
1673 assert(fds);
1674
1675 xsprintf(sfd, "%i", fileno(arg_serialization));
1676
1677 i = 0;
1678 args[i++] = SYSTEMD_BINARY_PATH;
1679 if (switch_root_dir)
1680 args[i++] = "--switched-root";
1681 args[i++] = arg_system ? "--system" : "--user";
1682 args[i++] = "--deserialize";
1683 args[i++] = sfd;
1684 args[i++] = NULL;
1685
1686 assert(i <= args_size);
1687
1688 /*
1689 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do this is on
1690 * its own on exec(), but it will do it on exit(). Hence, to ensure we get a summary here, fork() off
1691 * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1692 * before proceeding into the exec().
1693 */
1694 valgrind_summary_hack();
1695
1696 (void) execv(args[0], (char* const*) args);
1697 log_debug_errno(errno, "Failed to execute our own binary, trying fallback: %m");
1698 }
1699
1700 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1701 * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1702 * doesn't matter.) */
1703
1704 arg_serialization = safe_fclose(arg_serialization);
1705 fds = fdset_free(fds);
1706
1707 /* Reopen the console */
1708 (void) make_console_stdio();
1709
1710 for (j = 1, i = 1; j < (unsigned) argc; j++)
1711 args[i++] = argv[j];
1712 args[i++] = NULL;
1713 assert(i <= args_size);
1714
1715 /* Reenable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1716 (void) reset_all_signal_handlers();
1717 (void) reset_signal_mask();
1718
1719 if (switch_root_init) {
1720 args[0] = switch_root_init;
1721 (void) execv(args[0], (char* const*) args);
1722 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
1723 }
1724
1725 args[0] = "/sbin/init";
1726 (void) execv(args[0], (char* const*) args);
1727 r = -errno;
1728
1729 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
1730 ANSI_HIGHLIGHT_RED " !! " ANSI_NORMAL,
1731 "Failed to execute /sbin/init");
1732
1733 if (r == -ENOENT) {
1734 log_warning("No /sbin/init, trying fallback");
1735
1736 args[0] = "/bin/sh";
1737 args[1] = NULL;
1738 (void) execv(args[0], (char* const*) args);
1739 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
1740 } else
1741 log_warning_errno(r, "Failed to execute /sbin/init, giving up: %m");
1742
1743 *ret_error_message = "Failed to execute fallback shell";
1744 }
1745
1746 static int invoke_main_loop(
1747 Manager *m,
1748 bool *ret_reexecute,
1749 int *ret_retval, /* Return parameters relevant for shutting down */
1750 const char **ret_shutdown_verb, /* … */
1751 FDSet **ret_fds, /* Return parameters for reexecuting */
1752 char **ret_switch_root_dir, /* … */
1753 char **ret_switch_root_init, /* … */
1754 const char **ret_error_message) {
1755
1756 int r;
1757
1758 assert(m);
1759 assert(ret_reexecute);
1760 assert(ret_retval);
1761 assert(ret_shutdown_verb);
1762 assert(ret_fds);
1763 assert(ret_switch_root_dir);
1764 assert(ret_switch_root_init);
1765 assert(ret_error_message);
1766
1767 for (;;) {
1768 r = manager_loop(m);
1769 if (r < 0) {
1770 *ret_error_message = "Failed to run main loop";
1771 return log_emergency_errno(r, "Failed to run main loop: %m");
1772 }
1773
1774 switch ((ManagerObjective) r) {
1775
1776 case MANAGER_RELOAD: {
1777 LogTarget saved_log_target;
1778 int saved_log_level;
1779
1780 log_info("Reloading.");
1781
1782 /* First, save any overridden log level/target, then parse the configuration file, which might
1783 * change the log level to new settings. */
1784
1785 saved_log_level = m->log_level_overridden ? log_get_max_level() : -1;
1786 saved_log_target = m->log_target_overridden ? log_get_target() : _LOG_TARGET_INVALID;
1787
1788 r = parse_config_file();
1789 if (r < 0)
1790 log_warning_errno(r, "Failed to parse config file, ignoring: %m");
1791
1792 set_manager_defaults(m);
1793
1794 if (saved_log_level >= 0)
1795 manager_override_log_level(m, saved_log_level);
1796 if (saved_log_target >= 0)
1797 manager_override_log_target(m, saved_log_target);
1798
1799 r = manager_reload(m);
1800 if (r < 0)
1801 /* Reloading failed before the point of no return. Let's continue running as if nothing happened. */
1802 m->objective = MANAGER_OK;
1803
1804 break;
1805 }
1806
1807 case MANAGER_REEXECUTE:
1808
1809 r = prepare_reexecute(m, &arg_serialization, ret_fds, false);
1810 if (r < 0) {
1811 *ret_error_message = "Failed to prepare for reexecution";
1812 return r;
1813 }
1814
1815 log_notice("Reexecuting.");
1816
1817 *ret_reexecute = true;
1818 *ret_retval = EXIT_SUCCESS;
1819 *ret_shutdown_verb = NULL;
1820 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1821
1822 return 0;
1823
1824 case MANAGER_SWITCH_ROOT:
1825 if (!m->switch_root_init) {
1826 r = prepare_reexecute(m, &arg_serialization, ret_fds, true);
1827 if (r < 0) {
1828 *ret_error_message = "Failed to prepare for reexecution";
1829 return r;
1830 }
1831 } else
1832 *ret_fds = NULL;
1833
1834 log_notice("Switching root.");
1835
1836 *ret_reexecute = true;
1837 *ret_retval = EXIT_SUCCESS;
1838 *ret_shutdown_verb = NULL;
1839
1840 /* Steal the switch root parameters */
1841 *ret_switch_root_dir = m->switch_root;
1842 *ret_switch_root_init = m->switch_root_init;
1843 m->switch_root = m->switch_root_init = NULL;
1844
1845 return 0;
1846
1847 case MANAGER_EXIT:
1848
1849 if (MANAGER_IS_USER(m)) {
1850 log_debug("Exit.");
1851
1852 *ret_reexecute = false;
1853 *ret_retval = m->return_value;
1854 *ret_shutdown_verb = NULL;
1855 *ret_fds = NULL;
1856 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1857
1858 return 0;
1859 }
1860
1861 _fallthrough_;
1862 case MANAGER_REBOOT:
1863 case MANAGER_POWEROFF:
1864 case MANAGER_HALT:
1865 case MANAGER_KEXEC: {
1866 static const char * const table[_MANAGER_OBJECTIVE_MAX] = {
1867 [MANAGER_EXIT] = "exit",
1868 [MANAGER_REBOOT] = "reboot",
1869 [MANAGER_POWEROFF] = "poweroff",
1870 [MANAGER_HALT] = "halt",
1871 [MANAGER_KEXEC] = "kexec",
1872 };
1873
1874 log_notice("Shutting down.");
1875
1876 *ret_reexecute = false;
1877 *ret_retval = m->return_value;
1878 assert_se(*ret_shutdown_verb = table[m->objective]);
1879 *ret_fds = NULL;
1880 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1881
1882 return 0;
1883 }
1884
1885 default:
1886 assert_not_reached("Unknown or unexpected manager objective.");
1887 }
1888 }
1889 }
1890
1891 static void log_execution_mode(bool *ret_first_boot) {
1892 assert(ret_first_boot);
1893
1894 if (arg_system) {
1895 int v;
1896
1897 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1898 arg_action == ACTION_TEST ? "test " : "" );
1899
1900 v = detect_virtualization();
1901 if (v > 0)
1902 log_info("Detected virtualization %s.", virtualization_to_string(v));
1903
1904 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1905
1906 if (in_initrd()) {
1907 *ret_first_boot = false;
1908 log_info("Running in initial RAM disk.");
1909 } else {
1910 /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
1911 * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
1912 * doesn't it's unpopulated. This allows container managers and installers to provision a
1913 * couple of files already. If the container manager wants to provision the machine ID itself
1914 * it should pass $container_uuid to PID 1. */
1915
1916 *ret_first_boot = access("/etc/machine-id", F_OK) < 0;
1917 if (*ret_first_boot)
1918 log_info("Running with unpopulated /etc.");
1919 }
1920 } else {
1921 if (DEBUG_LOGGING) {
1922 _cleanup_free_ char *t;
1923
1924 t = uid_to_name(getuid());
1925 log_debug(PACKAGE_STRING " running in %suser mode for user " UID_FMT "/%s. (" SYSTEMD_FEATURES ")",
1926 arg_action == ACTION_TEST ? " test" : "", getuid(), strna(t));
1927 }
1928
1929 *ret_first_boot = false;
1930 }
1931 }
1932
1933 static int initialize_runtime(
1934 bool skip_setup,
1935 struct rlimit *saved_rlimit_nofile,
1936 struct rlimit *saved_rlimit_memlock,
1937 const char **ret_error_message) {
1938
1939 int r;
1940
1941 assert(ret_error_message);
1942
1943 /* Sets up various runtime parameters. Many of these initializations are conditionalized:
1944 *
1945 * - Some only apply to --system instances
1946 * - Some only apply to --user instances
1947 * - Some only apply when we first start up, but not when we reexecute
1948 */
1949
1950 if (arg_action != ACTION_RUN)
1951 return 0;
1952
1953 if (arg_system) {
1954 /* Make sure we leave a core dump without panicing the kernel. */
1955 install_crash_handler();
1956
1957 if (!skip_setup) {
1958 r = mount_cgroup_controllers(arg_join_controllers);
1959 if (r < 0) {
1960 *ret_error_message = "Failed to mount cgroup hierarchies";
1961 return r;
1962 }
1963
1964 status_welcome();
1965 hostname_setup();
1966 machine_id_setup(NULL, arg_machine_id, NULL);
1967 loopback_setup();
1968 bump_unix_max_dgram_qlen();
1969 bump_file_max_and_nr_open();
1970 test_usr();
1971 write_container_id();
1972 }
1973
1974 if (arg_watchdog_device) {
1975 r = watchdog_set_device(arg_watchdog_device);
1976 if (r < 0)
1977 log_warning_errno(r, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device);
1978 }
1979
1980 if (arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY)
1981 watchdog_set_timeout(&arg_runtime_watchdog);
1982 }
1983
1984 if (arg_timer_slack_nsec != NSEC_INFINITY)
1985 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1986 log_warning_errno(errno, "Failed to adjust timer slack, ignoring: %m");
1987
1988 if (arg_system && !cap_test_all(arg_capability_bounding_set)) {
1989 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
1990 if (r < 0) {
1991 *ret_error_message = "Failed to drop capability bounding set of usermode helpers";
1992 return log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
1993 }
1994
1995 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
1996 if (r < 0) {
1997 *ret_error_message = "Failed to drop capability bounding set";
1998 return log_emergency_errno(r, "Failed to drop capability bounding set: %m");
1999 }
2000 }
2001
2002 if (arg_system && arg_no_new_privs) {
2003 if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) {
2004 *ret_error_message = "Failed to disable new privileges";
2005 return log_emergency_errno(errno, "Failed to disable new privileges: %m");
2006 }
2007 }
2008
2009 if (arg_syscall_archs) {
2010 r = enforce_syscall_archs(arg_syscall_archs);
2011 if (r < 0) {
2012 *ret_error_message = "Failed to set syscall architectures";
2013 return r;
2014 }
2015 }
2016
2017 if (!arg_system)
2018 /* Become reaper of our children */
2019 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
2020 log_warning_errno(errno, "Failed to make us a subreaper: %m");
2021
2022 /* Bump up RLIMIT_NOFILE for systemd itself */
2023 (void) bump_rlimit_nofile(saved_rlimit_nofile);
2024 (void) bump_rlimit_memlock(saved_rlimit_memlock);
2025
2026 return 0;
2027 }
2028
2029 static int do_queue_default_job(
2030 Manager *m,
2031 const char **ret_error_message) {
2032
2033 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
2034 Job *default_unit_job;
2035 Unit *target = NULL;
2036 int r;
2037
2038 log_debug("Activating default unit: %s", arg_default_unit);
2039
2040 r = manager_load_startable_unit_or_warn(m, arg_default_unit, NULL, &target);
2041 if (r < 0) {
2042 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET);
2043
2044 r = manager_load_startable_unit_or_warn(m, SPECIAL_RESCUE_TARGET, NULL, &target);
2045 if (r < 0) {
2046 *ret_error_message = r == -ERFKILL ? "Rescue target masked"
2047 : "Failed to load rescue target";
2048 return r;
2049 }
2050 }
2051
2052 assert(target->load_state == UNIT_LOADED);
2053
2054 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, &error, &default_unit_job);
2055 if (r == -EPERM) {
2056 log_debug_errno(r, "Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
2057
2058 sd_bus_error_free(&error);
2059
2060 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, &error, &default_unit_job);
2061 if (r < 0) {
2062 *ret_error_message = "Failed to start default target";
2063 return log_emergency_errno(r, "Failed to start default target: %s", bus_error_message(&error, r));
2064 }
2065
2066 } else if (r < 0) {
2067 *ret_error_message = "Failed to isolate default target";
2068 return log_emergency_errno(r, "Failed to isolate default target: %s", bus_error_message(&error, r));
2069 }
2070
2071 m->default_unit_job_id = default_unit_job->id;
2072
2073 return 0;
2074 }
2075
2076 static void free_arguments(void) {
2077
2078 /* Frees all arg_* variables, with the exception of arg_serialization */
2079 rlimit_free_all(arg_default_rlimit);
2080
2081 arg_default_unit = mfree(arg_default_unit);
2082 arg_confirm_spawn = mfree(arg_confirm_spawn);
2083 arg_join_controllers = strv_free_free(arg_join_controllers);
2084 arg_default_environment = strv_free(arg_default_environment);
2085 arg_syscall_archs = set_free(arg_syscall_archs);
2086 }
2087
2088 static int load_configuration(int argc, char **argv, const char **ret_error_message) {
2089 int r;
2090
2091 assert(ret_error_message);
2092
2093 arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U);
2094
2095 r = parse_config_file();
2096 if (r < 0) {
2097 *ret_error_message = "Failed to parse config file";
2098 return r;
2099 }
2100
2101 if (arg_system) {
2102 r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
2103 if (r < 0)
2104 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
2105 }
2106
2107 /* Note that this also parses bits from the kernel command line, including "debug". */
2108 log_parse_environment();
2109
2110 r = parse_argv(argc, argv);
2111 if (r < 0) {
2112 *ret_error_message = "Failed to parse commandline arguments";
2113 return r;
2114 }
2115
2116 /* Initialize default unit */
2117 if (!arg_default_unit) {
2118 arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET);
2119 if (!arg_default_unit) {
2120 *ret_error_message = "Failed to set default unit";
2121 return log_oom();
2122 }
2123 }
2124
2125 /* Initialize the show status setting if it hasn't been set explicitly yet */
2126 if (arg_show_status == _SHOW_STATUS_INVALID)
2127 arg_show_status = SHOW_STATUS_YES;
2128
2129 return 0;
2130 }
2131
2132 static int safety_checks(void) {
2133
2134 if (getpid_cached() == 1 &&
2135 arg_action != ACTION_RUN) {
2136 log_error("Unsupported execution mode while PID 1.");
2137 return -EPERM;
2138 }
2139
2140 if (getpid_cached() == 1 &&
2141 !arg_system) {
2142 log_error("Can't run --user mode as PID 1.");
2143 return -EPERM;
2144 }
2145
2146 if (arg_action == ACTION_RUN &&
2147 arg_system &&
2148 getpid_cached() != 1) {
2149 log_error("Can't run system mode unless PID 1.");
2150 return -EPERM;
2151 }
2152
2153 if (arg_action == ACTION_TEST &&
2154 geteuid() == 0) {
2155 log_error("Don't run test mode as root.");
2156 return -EPERM;
2157 }
2158
2159 if (!arg_system &&
2160 arg_action == ACTION_RUN &&
2161 sd_booted() <= 0) {
2162 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
2163 return -EOPNOTSUPP;
2164 }
2165
2166 if (!arg_system &&
2167 arg_action == ACTION_RUN &&
2168 !getenv("XDG_RUNTIME_DIR")) {
2169 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
2170 return -EUNATCH;
2171 }
2172
2173 if (arg_system &&
2174 arg_action == ACTION_RUN &&
2175 running_in_chroot() > 0) {
2176 log_error("Cannot be run in a chroot() environment.");
2177 return -EOPNOTSUPP;
2178 }
2179
2180 return 0;
2181 }
2182
2183 static int initialize_security(
2184 bool *loaded_policy,
2185 dual_timestamp *security_start_timestamp,
2186 dual_timestamp *security_finish_timestamp,
2187 const char **ret_error_message) {
2188
2189 int r;
2190
2191 assert(loaded_policy);
2192 assert(security_start_timestamp);
2193 assert(security_finish_timestamp);
2194 assert(ret_error_message);
2195
2196 dual_timestamp_get(security_start_timestamp);
2197
2198 r = mac_selinux_setup(loaded_policy);
2199 if (r < 0) {
2200 *ret_error_message = "Failed to load SELinux policy";
2201 return r;
2202 }
2203
2204 r = mac_smack_setup(loaded_policy);
2205 if (r < 0) {
2206 *ret_error_message = "Failed to load SMACK policy";
2207 return r;
2208 }
2209
2210 r = ima_setup();
2211 if (r < 0) {
2212 *ret_error_message = "Failed to load IMA policy";
2213 return r;
2214 }
2215
2216 dual_timestamp_get(security_finish_timestamp);
2217 return 0;
2218 }
2219
2220 static void test_summary(Manager *m) {
2221 assert(m);
2222
2223 printf("-> By units:\n");
2224 manager_dump_units(m, stdout, "\t");
2225
2226 printf("-> By jobs:\n");
2227 manager_dump_jobs(m, stdout, "\t");
2228 }
2229
2230 static int collect_fds(FDSet **ret_fds, const char **ret_error_message) {
2231 int r;
2232
2233 assert(ret_fds);
2234 assert(ret_error_message);
2235
2236 r = fdset_new_fill(ret_fds);
2237 if (r < 0) {
2238 *ret_error_message = "Failed to allocate fd set";
2239 return log_emergency_errno(r, "Failed to allocate fd set: %m");
2240 }
2241
2242 fdset_cloexec(*ret_fds, true);
2243
2244 if (arg_serialization)
2245 assert_se(fdset_remove(*ret_fds, fileno(arg_serialization)) >= 0);
2246
2247 return 0;
2248 }
2249
2250 static void setup_console_terminal(bool skip_setup) {
2251
2252 if (!arg_system)
2253 return;
2254
2255 /* Become a session leader if we aren't one yet. */
2256 (void) setsid();
2257
2258 /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2259 * tty. */
2260 (void) release_terminal();
2261
2262 /* Reset the console, but only if this is really init and we are freshly booted */
2263 if (getpid_cached() == 1 && !skip_setup)
2264 (void) console_setup();
2265 }
2266
2267 static bool early_skip_setup_check(int argc, char *argv[]) {
2268 bool found_deserialize = false;
2269 int i;
2270
2271 /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2272 * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2273 * anyway, even if in that case we also do deserialization. */
2274
2275 for (i = 1; i < argc; i++) {
2276 if (streq(argv[i], "--switched-root"))
2277 return false; /* If we switched root, don't skip the setup. */
2278 else if (streq(argv[i], "--deserialize"))
2279 found_deserialize = true;
2280 }
2281
2282 return found_deserialize; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2283 }
2284
2285 int main(int argc, char *argv[]) {
2286
2287 dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL, userspace_timestamp = DUAL_TIMESTAMP_NULL, kernel_timestamp = DUAL_TIMESTAMP_NULL,
2288 security_start_timestamp = DUAL_TIMESTAMP_NULL, security_finish_timestamp = DUAL_TIMESTAMP_NULL;
2289 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0), saved_rlimit_memlock = RLIMIT_MAKE_CONST((rlim_t) -1);
2290 bool skip_setup, loaded_policy = false, queue_default_job = false, first_boot = false, reexecute = false;
2291 char *switch_root_dir = NULL, *switch_root_init = NULL;
2292 usec_t before_startup, after_startup;
2293 static char systemd[] = "systemd";
2294 char timespan[FORMAT_TIMESPAN_MAX];
2295 const char *shutdown_verb = NULL, *error_message = NULL;
2296 int r, retval = EXIT_FAILURE;
2297 Manager *m = NULL;
2298 FDSet *fds = NULL;
2299
2300 /* SysV compatibility: redirect init → telinit */
2301 redirect_telinit(argc, argv);
2302
2303 /* Take timestamps early on */
2304 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
2305 dual_timestamp_get(&userspace_timestamp);
2306
2307 /* Figure out whether we need to do initialize the system, or if we already did that because we are
2308 * reexecuting */
2309 skip_setup = early_skip_setup_check(argc, argv);
2310
2311 /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2312 * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
2313 program_invocation_short_name = systemd;
2314 (void) prctl(PR_SET_NAME, systemd);
2315
2316 /* Save the original command line */
2317 saved_argv = argv;
2318 saved_argc = argc;
2319
2320 /* Make sure that if the user says "syslog" we actually log to the journal. */
2321 log_set_upgrade_syslog_to_journal(true);
2322
2323 if (getpid_cached() == 1) {
2324 /* When we run as PID 1 force system mode */
2325 arg_system = true;
2326
2327 /* Disable the umask logic */
2328 umask(0);
2329
2330 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2331 * activated yet (even though the log socket for it exists). */
2332 log_set_prohibit_ipc(true);
2333
2334 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2335 * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2336 * child process right before execve()'ing the actual binary, at a point in time where socket
2337 * activation stderr/stdout area already set up. */
2338 log_set_always_reopen_console(true);
2339
2340 if (detect_container() <= 0) {
2341
2342 /* Running outside of a container as PID 1 */
2343 log_set_target(LOG_TARGET_KMSG);
2344 log_open();
2345
2346 if (in_initrd())
2347 initrd_timestamp = userspace_timestamp;
2348
2349 if (!skip_setup) {
2350 r = mount_setup_early();
2351 if (r < 0) {
2352 error_message = "Failed to mount early API filesystems";
2353 goto finish;
2354 }
2355
2356 r = initialize_security(
2357 &loaded_policy,
2358 &security_start_timestamp,
2359 &security_finish_timestamp,
2360 &error_message);
2361 if (r < 0)
2362 goto finish;
2363 }
2364
2365 if (mac_selinux_init() < 0) {
2366 error_message = "Failed to initialize SELinux policy";
2367 goto finish;
2368 }
2369
2370 if (!skip_setup)
2371 initialize_clock();
2372
2373 /* Set the default for later on, but don't actually open the logs like this for now. Note that
2374 * if we are transitioning from the initrd there might still be journal fd open, and we
2375 * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2376 * elsewhere. */
2377 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
2378
2379 } else {
2380 /* Running inside a container, as PID 1 */
2381 log_set_target(LOG_TARGET_CONSOLE);
2382 log_open();
2383
2384 /* For later on, see above... */
2385 log_set_target(LOG_TARGET_JOURNAL);
2386
2387 /* clear the kernel timestamp,
2388 * because we are in a container */
2389 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2390 }
2391
2392 initialize_coredump(skip_setup);
2393
2394 r = fixup_environment();
2395 if (r < 0) {
2396 log_emergency_errno(r, "Failed to fix up PID 1 environment: %m");
2397 error_message = "Failed to fix up PID1 environment";
2398 goto finish;
2399 }
2400
2401 } else {
2402 /* Running as user instance */
2403 arg_system = false;
2404 log_set_target(LOG_TARGET_AUTO);
2405 log_open();
2406
2407 /* clear the kernel timestamp,
2408 * because we are not PID 1 */
2409 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2410 }
2411
2412 if (arg_system) {
2413 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2414 * they never log into the console. */
2415 log_show_color(colors_enabled());
2416
2417 r = make_null_stdio();
2418 if (r < 0)
2419 log_warning_errno(r, "Failed to redirect standard streams to /dev/null, ignoring: %m");
2420 }
2421
2422 /* Mount /proc, /sys and friends, so that /proc/cmdline and
2423 * /proc/$PID/fd is available. */
2424 if (getpid_cached() == 1) {
2425
2426 /* Load the kernel modules early. */
2427 if (!skip_setup)
2428 kmod_setup();
2429
2430 r = mount_setup(loaded_policy);
2431 if (r < 0) {
2432 error_message = "Failed to mount API filesystems";
2433 goto finish;
2434 }
2435 }
2436
2437 /* Reset all signal handlers. */
2438 (void) reset_all_signal_handlers();
2439 (void) ignore_signals(SIGNALS_IGNORE, -1);
2440
2441 r = load_configuration(argc, argv, &error_message);
2442 if (r < 0)
2443 goto finish;
2444
2445 r = safety_checks();
2446 if (r < 0)
2447 goto finish;
2448
2449 if (IN_SET(arg_action, ACTION_TEST, ACTION_HELP, ACTION_DUMP_CONFIGURATION_ITEMS, ACTION_DUMP_BUS_PROPERTIES))
2450 (void) pager_open(arg_no_pager, false);
2451
2452 if (arg_action != ACTION_RUN)
2453 skip_setup = true;
2454
2455 if (arg_action == ACTION_HELP) {
2456 retval = help() < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
2457 goto finish;
2458 } else if (arg_action == ACTION_VERSION) {
2459 retval = version();
2460 goto finish;
2461 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
2462 unit_dump_config_items(stdout);
2463 retval = EXIT_SUCCESS;
2464 goto finish;
2465 } else if (arg_action == ACTION_DUMP_BUS_PROPERTIES) {
2466 dump_bus_properties(stdout);
2467 retval = EXIT_SUCCESS;
2468 goto finish;
2469 }
2470
2471 assert_se(IN_SET(arg_action, ACTION_RUN, ACTION_TEST));
2472
2473 /* Move out of the way, so that we won't block unmounts */
2474 assert_se(chdir("/") == 0);
2475
2476 if (arg_action == ACTION_RUN) {
2477
2478 /* A core pattern might have been specified via the cmdline. */
2479 initialize_core_pattern(skip_setup);
2480
2481 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2482 log_close();
2483
2484 /* Remember open file descriptors for later deserialization */
2485 r = collect_fds(&fds, &error_message);
2486 if (r < 0)
2487 goto finish;
2488
2489 /* Give up any control of the console, but make sure its initialized. */
2490 setup_console_terminal(skip_setup);
2491
2492 /* Open the logging devices, if possible and necessary */
2493 log_open();
2494 }
2495
2496 log_execution_mode(&first_boot);
2497
2498 r = initialize_runtime(skip_setup,
2499 &saved_rlimit_nofile,
2500 &saved_rlimit_memlock,
2501 &error_message);
2502 if (r < 0)
2503 goto finish;
2504
2505 r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER,
2506 arg_action == ACTION_TEST ? MANAGER_TEST_FULL : 0,
2507 &m);
2508 if (r < 0) {
2509 log_emergency_errno(r, "Failed to allocate manager object: %m");
2510 error_message = "Failed to allocate manager object";
2511 goto finish;
2512 }
2513
2514 m->timestamps[MANAGER_TIMESTAMP_KERNEL] = kernel_timestamp;
2515 m->timestamps[MANAGER_TIMESTAMP_INITRD] = initrd_timestamp;
2516 m->timestamps[MANAGER_TIMESTAMP_USERSPACE] = userspace_timestamp;
2517 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START)] = security_start_timestamp;
2518 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH)] = security_finish_timestamp;
2519
2520 set_manager_defaults(m);
2521 set_manager_settings(m);
2522 manager_set_first_boot(m, first_boot);
2523
2524 /* Remember whether we should queue the default job */
2525 queue_default_job = !arg_serialization || arg_switched_root;
2526
2527 before_startup = now(CLOCK_MONOTONIC);
2528
2529 r = manager_startup(m, arg_serialization, fds);
2530 if (r < 0) {
2531 error_message = "Failed to start up manager";
2532 goto finish;
2533 }
2534
2535 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2536 fds = fdset_free(fds);
2537 arg_serialization = safe_fclose(arg_serialization);
2538
2539 if (queue_default_job) {
2540 r = do_queue_default_job(m, &error_message);
2541 if (r < 0)
2542 goto finish;
2543 }
2544
2545 after_startup = now(CLOCK_MONOTONIC);
2546
2547 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
2548 "Loaded units and determined initial transaction in %s.",
2549 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
2550
2551 if (arg_action == ACTION_TEST) {
2552 test_summary(m);
2553 retval = EXIT_SUCCESS;
2554 goto finish;
2555 }
2556
2557 (void) invoke_main_loop(m,
2558 &reexecute,
2559 &retval,
2560 &shutdown_verb,
2561 &fds,
2562 &switch_root_dir,
2563 &switch_root_init,
2564 &error_message);
2565
2566 finish:
2567 pager_close();
2568
2569 if (m) {
2570 arg_shutdown_watchdog = m->shutdown_watchdog;
2571 m = manager_free(m);
2572 }
2573
2574 free_arguments();
2575 mac_selinux_finish();
2576
2577 if (reexecute)
2578 do_reexecute(argc, argv,
2579 &saved_rlimit_nofile,
2580 &saved_rlimit_memlock,
2581 fds,
2582 switch_root_dir,
2583 switch_root_init,
2584 &error_message); /* This only returns if reexecution failed */
2585
2586 arg_serialization = safe_fclose(arg_serialization);
2587 fds = fdset_free(fds);
2588
2589 #if HAVE_VALGRIND_VALGRIND_H
2590 /* If we are PID 1 and running under valgrind, then let's exit
2591 * here explicitly. valgrind will only generate nice output on
2592 * exit(), not on exec(), hence let's do the former not the
2593 * latter here. */
2594 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND) {
2595 /* Cleanup watchdog_device strings for valgrind. We need them
2596 * in become_shutdown() so normally we cannot free them yet. */
2597 watchdog_free_device();
2598 arg_watchdog_device = mfree(arg_watchdog_device);
2599 return retval;
2600 }
2601 #endif
2602
2603 if (shutdown_verb) {
2604 r = become_shutdown(shutdown_verb, retval);
2605 log_error_errno(r, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
2606 error_message = "Failed to execute shutdown binary";
2607 }
2608
2609 watchdog_free_device();
2610 arg_watchdog_device = mfree(arg_watchdog_device);
2611
2612 if (getpid_cached() == 1) {
2613 if (error_message)
2614 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2615 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
2616 "%s, freezing.", error_message);
2617 freeze_or_reboot();
2618 }
2619
2620 return retval;
2621 }
Unnamed repository; edit this file 'description' to name the repository.