]> 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 #10811 from keszybz/define-main-through-macro
[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 "cgroup-util.h"
32 #include "clock-util.h"
33 #include "conf-parser.h"
34 #include "cpu-set-util.h"
35 #include "dbus.h"
36 #include "dbus-manager.h"
37 #include "def.h"
38 #include "emergency-action.h"
39 #include "env-util.h"
40 #include "exit-status.h"
41 #include "fd-util.h"
42 #include "fdset.h"
43 #include "fileio.h"
44 #include "format-util.h"
45 #include "fs-util.h"
46 #include "hostname-setup.h"
47 #include "ima-setup.h"
48 #include "killall.h"
49 #include "kmod-setup.h"
50 #include "load-fragment.h"
51 #include "log.h"
52 #include "loopback-setup.h"
53 #include "machine-id-setup.h"
54 #include "manager.h"
55 #include "missing.h"
56 #include "mount-setup.h"
57 #include "os-util.h"
58 #include "pager.h"
59 #include "parse-util.h"
60 #include "path-util.h"
61 #include "proc-cmdline.h"
62 #include "process-util.h"
63 #include "raw-clone.h"
64 #include "rlimit-util.h"
65 #if HAVE_SECCOMP
66 #include "seccomp-util.h"
67 #endif
68 #include "selinux-setup.h"
69 #include "selinux-util.h"
70 #include "signal-util.h"
71 #include "smack-setup.h"
72 #include "special.h"
73 #include "stat-util.h"
74 #include "stdio-util.h"
75 #include "strv.h"
76 #include "switch-root.h"
77 #include "sysctl-util.h"
78 #include "terminal-util.h"
79 #include "umask-util.h"
80 #include "user-util.h"
81 #include "util.h"
82 #include "virt.h"
83 #include "watchdog.h"
84
85 static enum {
86 ACTION_RUN,
87 ACTION_HELP,
88 ACTION_VERSION,
89 ACTION_TEST,
90 ACTION_DUMP_CONFIGURATION_ITEMS,
91 ACTION_DUMP_BUS_PROPERTIES,
92 } arg_action = ACTION_RUN;
93 static char *arg_default_unit = NULL;
94 static bool arg_system = false;
95 static bool arg_dump_core = true;
96 static int arg_crash_chvt = -1;
97 static bool arg_crash_shell = false;
98 static bool arg_crash_reboot = false;
99 static char *arg_confirm_spawn = NULL;
100 static ShowStatus arg_show_status = _SHOW_STATUS_INVALID;
101 static bool arg_switched_root = false;
102 static PagerFlags arg_pager_flags = 0;
103 static bool arg_service_watchdogs = true;
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 int arg_default_cpu_accounting = -1;
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_errno(r, "Failed to parse dump core switch %s, ignoring: %m", 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 {
371 r = parse_crash_chvt(value);
372 if (r < 0)
373 log_warning_errno(r, "Failed to parse crash chvt switch %s, ignoring: %m", value);
374 }
375
376 } else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) {
377
378 r = value ? parse_boolean(value) : true;
379 if (r < 0)
380 log_warning_errno(r, "Failed to parse crash shell switch %s, ignoring: %m", value);
381 else
382 arg_crash_shell = r;
383
384 } else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) {
385
386 r = value ? parse_boolean(value) : true;
387 if (r < 0)
388 log_warning_errno(r, "Failed to parse crash reboot switch %s, ignoring: %m", value);
389 else
390 arg_crash_reboot = r;
391
392 } else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) {
393 char *s;
394
395 r = parse_confirm_spawn(value, &s);
396 if (r < 0)
397 log_warning_errno(r, "Failed to parse confirm_spawn switch %s, ignoring: %m", value);
398 else
399 free_and_replace(arg_confirm_spawn, s);
400
401 } else if (proc_cmdline_key_streq(key, "systemd.service_watchdogs")) {
402
403 r = value ? parse_boolean(value) : true;
404 if (r < 0)
405 log_warning_errno(r, "Failed to parse service watchdog switch %s, ignoring: %m", value);
406 else
407 arg_service_watchdogs = r;
408
409 } else if (proc_cmdline_key_streq(key, "systemd.show_status")) {
410
411 if (value) {
412 r = parse_show_status(value, &arg_show_status);
413 if (r < 0)
414 log_warning_errno(r, "Failed to parse show status switch %s, ignoring: %m", value);
415 } else
416 arg_show_status = SHOW_STATUS_YES;
417
418 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) {
419
420 if (proc_cmdline_value_missing(key, value))
421 return 0;
422
423 r = exec_output_from_string(value);
424 if (r < 0)
425 log_warning_errno(r, "Failed to parse default standard output switch %s, ignoring: %m", value);
426 else
427 arg_default_std_output = r;
428
429 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) {
430
431 if (proc_cmdline_value_missing(key, value))
432 return 0;
433
434 r = exec_output_from_string(value);
435 if (r < 0)
436 log_warning_errno(r, "Failed to parse default standard error switch %s, ignoring: %m", value);
437 else
438 arg_default_std_error = r;
439
440 } else if (streq(key, "systemd.setenv")) {
441
442 if (proc_cmdline_value_missing(key, value))
443 return 0;
444
445 if (env_assignment_is_valid(value)) {
446 char **env;
447
448 env = strv_env_set(arg_default_environment, value);
449 if (!env)
450 return log_oom();
451
452 arg_default_environment = env;
453 } else
454 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
455
456 } else if (proc_cmdline_key_streq(key, "systemd.machine_id")) {
457
458 if (proc_cmdline_value_missing(key, value))
459 return 0;
460
461 r = set_machine_id(value);
462 if (r < 0)
463 log_warning_errno(r, "MachineID '%s' is not valid, ignoring: %m", value);
464
465 } else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) {
466
467 if (proc_cmdline_value_missing(key, value))
468 return 0;
469
470 r = parse_sec(value, &arg_default_timeout_start_usec);
471 if (r < 0)
472 log_warning_errno(r, "Failed to parse default start timeout '%s', ignoring: %m", value);
473
474 if (arg_default_timeout_start_usec <= 0)
475 arg_default_timeout_start_usec = USEC_INFINITY;
476
477 } else if (proc_cmdline_key_streq(key, "systemd.watchdog_device")) {
478
479 if (proc_cmdline_value_missing(key, value))
480 return 0;
481
482 (void) parse_path_argument_and_warn(value, false, &arg_watchdog_device);
483
484 } else if (streq(key, "quiet") && !value) {
485
486 if (arg_show_status == _SHOW_STATUS_INVALID)
487 arg_show_status = SHOW_STATUS_AUTO;
488
489 } else if (streq(key, "debug") && !value) {
490
491 /* Note that log_parse_environment() handles 'debug'
492 * too, and sets the log level to LOG_DEBUG. */
493
494 if (detect_container() > 0)
495 log_set_target(LOG_TARGET_CONSOLE);
496
497 } else if (!value) {
498 const char *target;
499
500 /* SysV compatibility */
501 target = runlevel_to_target(key);
502 if (target)
503 return free_and_strdup(&arg_default_unit, target);
504 }
505
506 return 0;
507 }
508
509 #define DEFINE_SETTER(name, func, descr) \
510 static int name(const char *unit, \
511 const char *filename, \
512 unsigned line, \
513 const char *section, \
514 unsigned section_line, \
515 const char *lvalue, \
516 int ltype, \
517 const char *rvalue, \
518 void *data, \
519 void *userdata) { \
520 \
521 int r; \
522 \
523 assert(filename); \
524 assert(lvalue); \
525 assert(rvalue); \
526 \
527 r = func(rvalue); \
528 if (r < 0) \
529 log_syntax(unit, LOG_ERR, filename, line, r, \
530 "Invalid " descr "'%s': %m", \
531 rvalue); \
532 \
533 return 0; \
534 }
535
536 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level");
537 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target");
538 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" );
539 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location");
540
541 static int config_parse_cpu_affinity2(
542 const char *unit,
543 const char *filename,
544 unsigned line,
545 const char *section,
546 unsigned section_line,
547 const char *lvalue,
548 int ltype,
549 const char *rvalue,
550 void *data,
551 void *userdata) {
552
553 _cleanup_cpu_free_ cpu_set_t *c = NULL;
554 int ncpus;
555
556 ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue);
557 if (ncpus < 0)
558 return ncpus;
559
560 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
561 log_warning_errno(errno, "Failed to set CPU affinity: %m");
562
563 return 0;
564 }
565
566 static int config_parse_show_status(
567 const char* unit,
568 const char *filename,
569 unsigned line,
570 const char *section,
571 unsigned section_line,
572 const char *lvalue,
573 int ltype,
574 const char *rvalue,
575 void *data,
576 void *userdata) {
577
578 int k;
579 ShowStatus *b = data;
580
581 assert(filename);
582 assert(lvalue);
583 assert(rvalue);
584 assert(data);
585
586 k = parse_show_status(rvalue, b);
587 if (k < 0) {
588 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
589 return 0;
590 }
591
592 return 0;
593 }
594
595 static int config_parse_output_restricted(
596 const char* unit,
597 const char *filename,
598 unsigned line,
599 const char *section,
600 unsigned section_line,
601 const char *lvalue,
602 int ltype,
603 const char *rvalue,
604 void *data,
605 void *userdata) {
606
607 ExecOutput t, *eo = data;
608
609 assert(filename);
610 assert(lvalue);
611 assert(rvalue);
612 assert(data);
613
614 t = exec_output_from_string(rvalue);
615 if (t < 0) {
616 log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse output type, ignoring: %s", rvalue);
617 return 0;
618 }
619
620 if (IN_SET(t, EXEC_OUTPUT_SOCKET, EXEC_OUTPUT_NAMED_FD, EXEC_OUTPUT_FILE, EXEC_OUTPUT_FILE_APPEND)) {
621 log_syntax(unit, LOG_ERR, filename, line, 0, "Standard output types socket, fd:, file:, append: are not supported as defaults, ignoring: %s", rvalue);
622 return 0;
623 }
624
625 *eo = t;
626 return 0;
627 }
628
629 static int config_parse_crash_chvt(
630 const char* unit,
631 const char *filename,
632 unsigned line,
633 const char *section,
634 unsigned section_line,
635 const char *lvalue,
636 int ltype,
637 const char *rvalue,
638 void *data,
639 void *userdata) {
640
641 int r;
642
643 assert(filename);
644 assert(lvalue);
645 assert(rvalue);
646
647 r = parse_crash_chvt(rvalue);
648 if (r < 0) {
649 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
650 return 0;
651 }
652
653 return 0;
654 }
655
656 static int parse_config_file(void) {
657
658 const ConfigTableItem items[] = {
659 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
660 { "Manager", "LogTarget", config_parse_target, 0, NULL },
661 { "Manager", "LogColor", config_parse_color, 0, NULL },
662 { "Manager", "LogLocation", config_parse_location, 0, NULL },
663 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
664 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL },
665 { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL },
666 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
667 { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
668 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
669 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
670 { "Manager", "JoinControllers", config_parse_warn_compat, DISABLED_CONFIGURATION, NULL },
671 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
672 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
673 { "Manager", "WatchdogDevice", config_parse_path, 0, &arg_watchdog_device },
674 { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
675 { "Manager", "NoNewPrivileges", config_parse_bool, 0, &arg_no_new_privs },
676 #if HAVE_SECCOMP
677 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
678 #endif
679 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
680 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
681 { "Manager", "DefaultStandardOutput", config_parse_output_restricted,0, &arg_default_std_output },
682 { "Manager", "DefaultStandardError", config_parse_output_restricted,0, &arg_default_std_error },
683 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
684 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
685 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
686 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */
687 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec, 0, &arg_default_start_limit_interval },
688 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
689 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
690 { "Manager", "DefaultLimitCPU", config_parse_rlimit, RLIMIT_CPU, arg_default_rlimit },
691 { "Manager", "DefaultLimitFSIZE", config_parse_rlimit, RLIMIT_FSIZE, arg_default_rlimit },
692 { "Manager", "DefaultLimitDATA", config_parse_rlimit, RLIMIT_DATA, arg_default_rlimit },
693 { "Manager", "DefaultLimitSTACK", config_parse_rlimit, RLIMIT_STACK, arg_default_rlimit },
694 { "Manager", "DefaultLimitCORE", config_parse_rlimit, RLIMIT_CORE, arg_default_rlimit },
695 { "Manager", "DefaultLimitRSS", config_parse_rlimit, RLIMIT_RSS, arg_default_rlimit },
696 { "Manager", "DefaultLimitNOFILE", config_parse_rlimit, RLIMIT_NOFILE, arg_default_rlimit },
697 { "Manager", "DefaultLimitAS", config_parse_rlimit, RLIMIT_AS, arg_default_rlimit },
698 { "Manager", "DefaultLimitNPROC", config_parse_rlimit, RLIMIT_NPROC, arg_default_rlimit },
699 { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit, RLIMIT_MEMLOCK, arg_default_rlimit },
700 { "Manager", "DefaultLimitLOCKS", config_parse_rlimit, RLIMIT_LOCKS, arg_default_rlimit },
701 { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit, RLIMIT_SIGPENDING, arg_default_rlimit },
702 { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit, RLIMIT_MSGQUEUE, arg_default_rlimit },
703 { "Manager", "DefaultLimitNICE", config_parse_rlimit, RLIMIT_NICE, arg_default_rlimit },
704 { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit, RLIMIT_RTPRIO, arg_default_rlimit },
705 { "Manager", "DefaultLimitRTTIME", config_parse_rlimit, RLIMIT_RTTIME, arg_default_rlimit },
706 { "Manager", "DefaultCPUAccounting", config_parse_tristate, 0, &arg_default_cpu_accounting },
707 { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting },
708 { "Manager", "DefaultIPAccounting", config_parse_bool, 0, &arg_default_ip_accounting },
709 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
710 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
711 { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
712 { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max },
713 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action, 0, &arg_cad_burst_action },
714 {}
715 };
716
717 const char *fn, *conf_dirs_nulstr;
718
719 fn = arg_system ?
720 PKGSYSCONFDIR "/system.conf" :
721 PKGSYSCONFDIR "/user.conf";
722
723 conf_dirs_nulstr = arg_system ?
724 CONF_PATHS_NULSTR("systemd/system.conf.d") :
725 CONF_PATHS_NULSTR("systemd/user.conf.d");
726
727 (void) config_parse_many_nulstr(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, CONFIG_PARSE_WARN, NULL);
728
729 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
730 * like everywhere else. */
731 if (arg_default_timeout_start_usec <= 0)
732 arg_default_timeout_start_usec = USEC_INFINITY;
733 if (arg_default_timeout_stop_usec <= 0)
734 arg_default_timeout_stop_usec = USEC_INFINITY;
735
736 return 0;
737 }
738
739 static void set_manager_defaults(Manager *m) {
740
741 assert(m);
742
743 /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
744 * affect the manager itself, but are just what newly allocated units will have set if they haven't set
745 * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
746
747 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
748 m->default_std_output = arg_default_std_output;
749 m->default_std_error = arg_default_std_error;
750 m->default_timeout_start_usec = arg_default_timeout_start_usec;
751 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
752 m->default_restart_usec = arg_default_restart_usec;
753 m->default_start_limit_interval = arg_default_start_limit_interval;
754 m->default_start_limit_burst = arg_default_start_limit_burst;
755
756 /* On 4.15+ with unified hierarchy, CPU accounting is essentially free as it doesn't require the CPU
757 * controller to be enabled, so the default is to enable it unless we got told otherwise. */
758 if (arg_default_cpu_accounting >= 0)
759 m->default_cpu_accounting = arg_default_cpu_accounting;
760 else
761 m->default_cpu_accounting = cpu_accounting_is_cheap();
762
763 m->default_io_accounting = arg_default_io_accounting;
764 m->default_ip_accounting = arg_default_ip_accounting;
765 m->default_blockio_accounting = arg_default_blockio_accounting;
766 m->default_memory_accounting = arg_default_memory_accounting;
767 m->default_tasks_accounting = arg_default_tasks_accounting;
768 m->default_tasks_max = arg_default_tasks_max;
769
770 manager_set_default_rlimits(m, arg_default_rlimit);
771 manager_transient_environment_add(m, arg_default_environment);
772 }
773
774 static void set_manager_settings(Manager *m) {
775
776 assert(m);
777
778 /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
779 * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
780
781 m->confirm_spawn = arg_confirm_spawn;
782 m->service_watchdogs = arg_service_watchdogs;
783 m->runtime_watchdog = arg_runtime_watchdog;
784 m->shutdown_watchdog = arg_shutdown_watchdog;
785 m->cad_burst_action = arg_cad_burst_action;
786
787 manager_set_show_status(m, arg_show_status);
788 }
789
790 static int parse_argv(int argc, char *argv[]) {
791 enum {
792 ARG_LOG_LEVEL = 0x100,
793 ARG_LOG_TARGET,
794 ARG_LOG_COLOR,
795 ARG_LOG_LOCATION,
796 ARG_UNIT,
797 ARG_SYSTEM,
798 ARG_USER,
799 ARG_TEST,
800 ARG_NO_PAGER,
801 ARG_VERSION,
802 ARG_DUMP_CONFIGURATION_ITEMS,
803 ARG_DUMP_BUS_PROPERTIES,
804 ARG_DUMP_CORE,
805 ARG_CRASH_CHVT,
806 ARG_CRASH_SHELL,
807 ARG_CRASH_REBOOT,
808 ARG_CONFIRM_SPAWN,
809 ARG_SHOW_STATUS,
810 ARG_DESERIALIZE,
811 ARG_SWITCHED_ROOT,
812 ARG_DEFAULT_STD_OUTPUT,
813 ARG_DEFAULT_STD_ERROR,
814 ARG_MACHINE_ID,
815 ARG_SERVICE_WATCHDOGS,
816 };
817
818 static const struct option options[] = {
819 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
820 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
821 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
822 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
823 { "unit", required_argument, NULL, ARG_UNIT },
824 { "system", no_argument, NULL, ARG_SYSTEM },
825 { "user", no_argument, NULL, ARG_USER },
826 { "test", no_argument, NULL, ARG_TEST },
827 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
828 { "help", no_argument, NULL, 'h' },
829 { "version", no_argument, NULL, ARG_VERSION },
830 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
831 { "dump-bus-properties", no_argument, NULL, ARG_DUMP_BUS_PROPERTIES },
832 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
833 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
834 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
835 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
836 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
837 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
838 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
839 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
840 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
841 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
842 { "machine-id", required_argument, NULL, ARG_MACHINE_ID },
843 { "service-watchdogs", required_argument, NULL, ARG_SERVICE_WATCHDOGS },
844 {}
845 };
846
847 int c, r;
848
849 assert(argc >= 1);
850 assert(argv);
851
852 if (getpid_cached() == 1)
853 opterr = 0;
854
855 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
856
857 switch (c) {
858
859 case ARG_LOG_LEVEL:
860 r = log_set_max_level_from_string(optarg);
861 if (r < 0)
862 return log_error_errno(r, "Failed to parse log level \"%s\": %m", optarg);
863
864 break;
865
866 case ARG_LOG_TARGET:
867 r = log_set_target_from_string(optarg);
868 if (r < 0)
869 return log_error_errno(r, "Failed to parse log target \"%s\": %m", optarg);
870
871 break;
872
873 case ARG_LOG_COLOR:
874
875 if (optarg) {
876 r = log_show_color_from_string(optarg);
877 if (r < 0)
878 return log_error_errno(r, "Failed to parse log color setting \"%s\": %m",
879 optarg);
880 } else
881 log_show_color(true);
882
883 break;
884
885 case ARG_LOG_LOCATION:
886 if (optarg) {
887 r = log_show_location_from_string(optarg);
888 if (r < 0)
889 return log_error_errno(r, "Failed to parse log location setting \"%s\": %m",
890 optarg);
891 } else
892 log_show_location(true);
893
894 break;
895
896 case ARG_DEFAULT_STD_OUTPUT:
897 r = exec_output_from_string(optarg);
898 if (r < 0)
899 return log_error_errno(r, "Failed to parse default standard output setting \"%s\": %m",
900 optarg);
901 arg_default_std_output = r;
902 break;
903
904 case ARG_DEFAULT_STD_ERROR:
905 r = exec_output_from_string(optarg);
906 if (r < 0)
907 return log_error_errno(r, "Failed to parse default standard error output setting \"%s\": %m",
908 optarg);
909 arg_default_std_error = r;
910 break;
911
912 case ARG_UNIT:
913 r = free_and_strdup(&arg_default_unit, optarg);
914 if (r < 0)
915 return log_error_errno(r, "Failed to set default unit \"%s\": %m", optarg);
916
917 break;
918
919 case ARG_SYSTEM:
920 arg_system = true;
921 break;
922
923 case ARG_USER:
924 arg_system = false;
925 break;
926
927 case ARG_TEST:
928 arg_action = ACTION_TEST;
929 break;
930
931 case ARG_NO_PAGER:
932 arg_pager_flags |= PAGER_DISABLE;
933 break;
934
935 case ARG_VERSION:
936 arg_action = ACTION_VERSION;
937 break;
938
939 case ARG_DUMP_CONFIGURATION_ITEMS:
940 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
941 break;
942
943 case ARG_DUMP_BUS_PROPERTIES:
944 arg_action = ACTION_DUMP_BUS_PROPERTIES;
945 break;
946
947 case ARG_DUMP_CORE:
948 if (!optarg)
949 arg_dump_core = true;
950 else {
951 r = parse_boolean(optarg);
952 if (r < 0)
953 return log_error_errno(r, "Failed to parse dump core boolean: \"%s\": %m",
954 optarg);
955 arg_dump_core = r;
956 }
957 break;
958
959 case ARG_CRASH_CHVT:
960 r = parse_crash_chvt(optarg);
961 if (r < 0)
962 return log_error_errno(r, "Failed to parse crash virtual terminal index: \"%s\": %m",
963 optarg);
964 break;
965
966 case ARG_CRASH_SHELL:
967 if (!optarg)
968 arg_crash_shell = true;
969 else {
970 r = parse_boolean(optarg);
971 if (r < 0)
972 return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
973 optarg);
974 arg_crash_shell = r;
975 }
976 break;
977
978 case ARG_CRASH_REBOOT:
979 if (!optarg)
980 arg_crash_reboot = true;
981 else {
982 r = parse_boolean(optarg);
983 if (r < 0)
984 return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
985 optarg);
986 arg_crash_reboot = r;
987 }
988 break;
989
990 case ARG_CONFIRM_SPAWN:
991 arg_confirm_spawn = mfree(arg_confirm_spawn);
992
993 r = parse_confirm_spawn(optarg, &arg_confirm_spawn);
994 if (r < 0)
995 return log_error_errno(r, "Failed to parse confirm spawn option: \"%s\": %m",
996 optarg);
997 break;
998
999 case ARG_SERVICE_WATCHDOGS:
1000 r = parse_boolean(optarg);
1001 if (r < 0)
1002 return log_error_errno(r, "Failed to parse service watchdogs boolean: \"%s\": %m",
1003 optarg);
1004 arg_service_watchdogs = r;
1005 break;
1006
1007 case ARG_SHOW_STATUS:
1008 if (optarg) {
1009 r = parse_show_status(optarg, &arg_show_status);
1010 if (r < 0)
1011 return log_error_errno(r, "Failed to parse show status boolean: \"%s\": %m",
1012 optarg);
1013 } else
1014 arg_show_status = SHOW_STATUS_YES;
1015 break;
1016
1017 case ARG_DESERIALIZE: {
1018 int fd;
1019 FILE *f;
1020
1021 r = safe_atoi(optarg, &fd);
1022 if (r < 0)
1023 log_error_errno(r, "Failed to parse deserialize option \"%s\": %m", optarg);
1024 if (fd < 0) {
1025 log_error("Invalid deserialize fd: %d", fd);
1026 return -EINVAL;
1027 }
1028
1029 (void) fd_cloexec(fd, true);
1030
1031 f = fdopen(fd, "r");
1032 if (!f)
1033 return log_error_errno(errno, "Failed to open serialization fd %d: %m", fd);
1034
1035 safe_fclose(arg_serialization);
1036 arg_serialization = f;
1037
1038 break;
1039 }
1040
1041 case ARG_SWITCHED_ROOT:
1042 arg_switched_root = true;
1043 break;
1044
1045 case ARG_MACHINE_ID:
1046 r = set_machine_id(optarg);
1047 if (r < 0)
1048 return log_error_errno(r, "MachineID '%s' is not valid: %m", optarg);
1049 break;
1050
1051 case 'h':
1052 arg_action = ACTION_HELP;
1053 break;
1054
1055 case 'D':
1056 log_set_max_level(LOG_DEBUG);
1057 break;
1058
1059 case 'b':
1060 case 's':
1061 case 'z':
1062 /* Just to eat away the sysvinit kernel
1063 * cmdline args without getopt() error
1064 * messages that we'll parse in
1065 * parse_proc_cmdline_word() or ignore. */
1066
1067 case '?':
1068 if (getpid_cached() != 1)
1069 return -EINVAL;
1070 else
1071 return 0;
1072
1073 default:
1074 assert_not_reached("Unhandled option code.");
1075 }
1076
1077 if (optind < argc && getpid_cached() != 1) {
1078 /* Hmm, when we aren't run as init system
1079 * let's complain about excess arguments */
1080
1081 log_error("Excess arguments.");
1082 return -EINVAL;
1083 }
1084
1085 return 0;
1086 }
1087
1088 static int help(void) {
1089 _cleanup_free_ char *link = NULL;
1090 int r;
1091
1092 r = terminal_urlify_man("systemd", "1", &link);
1093 if (r < 0)
1094 return log_oom();
1095
1096 printf("%s [OPTIONS...]\n\n"
1097 "Starts up and maintains the system or user services.\n\n"
1098 " -h --help Show this help\n"
1099 " --version Show version\n"
1100 " --test Determine startup sequence, dump it and exit\n"
1101 " --no-pager Do not pipe output into a pager\n"
1102 " --dump-configuration-items Dump understood unit configuration items\n"
1103 " --dump-bus-properties Dump exposed bus properties\n"
1104 " --unit=UNIT Set default unit\n"
1105 " --system Run a system instance, even if PID != 1\n"
1106 " --user Run a user instance\n"
1107 " --dump-core[=BOOL] Dump core on crash\n"
1108 " --crash-vt=NR Change to specified VT on crash\n"
1109 " --crash-reboot[=BOOL] Reboot on crash\n"
1110 " --crash-shell[=BOOL] Run shell on crash\n"
1111 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1112 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1113 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1114 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1115 " --log-color[=BOOL] Highlight important log messages\n"
1116 " --log-location[=BOOL] Include code location in log messages\n"
1117 " --default-standard-output= Set default standard output for services\n"
1118 " --default-standard-error= Set default standard error output for services\n"
1119 "\nSee the %s for details.\n"
1120 , program_invocation_short_name
1121 , link
1122 );
1123
1124 return 0;
1125 }
1126
1127 static int prepare_reexecute(
1128 Manager *m,
1129 FILE **ret_f,
1130 FDSet **ret_fds,
1131 bool switching_root) {
1132
1133 _cleanup_fdset_free_ FDSet *fds = NULL;
1134 _cleanup_fclose_ FILE *f = NULL;
1135 int r;
1136
1137 assert(m);
1138 assert(ret_f);
1139 assert(ret_fds);
1140
1141 r = manager_open_serialization(m, &f);
1142 if (r < 0)
1143 return log_error_errno(r, "Failed to create serialization file: %m");
1144
1145 /* Make sure nothing is really destructed when we shut down */
1146 m->n_reloading++;
1147 bus_manager_send_reloading(m, true);
1148
1149 fds = fdset_new();
1150 if (!fds)
1151 return log_oom();
1152
1153 r = manager_serialize(m, f, fds, switching_root);
1154 if (r < 0)
1155 return r;
1156
1157 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1158 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1159
1160 r = fd_cloexec(fileno(f), false);
1161 if (r < 0)
1162 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1163
1164 r = fdset_cloexec(fds, false);
1165 if (r < 0)
1166 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1167
1168 *ret_f = TAKE_PTR(f);
1169 *ret_fds = TAKE_PTR(fds);
1170
1171 return 0;
1172 }
1173
1174 static void bump_file_max_and_nr_open(void) {
1175
1176 /* Let's bump fs.file-max and fs.nr_open to their respective maximums. On current kernels large numbers of file
1177 * descriptors are no longer a performance problem and their memory is properly tracked by memcg, thus counting
1178 * them and limiting them in another two layers of limits is unnecessary and just complicates things. This
1179 * function hence turns off 2 of the 4 levels of limits on file descriptors, and makes RLIMIT_NOLIMIT (soft +
1180 * hard) the only ones that really matter. */
1181
1182 #if BUMP_PROC_SYS_FS_FILE_MAX || BUMP_PROC_SYS_FS_NR_OPEN
1183 _cleanup_free_ char *t = NULL;
1184 int r;
1185 #endif
1186
1187 #if BUMP_PROC_SYS_FS_FILE_MAX
1188 /* I so wanted to use STRINGIFY(ULONG_MAX) here, but alas we can't as glibc/gcc define that as
1189 * "(0x7fffffffffffffffL * 2UL + 1UL)". Seriously. 😢 */
1190 if (asprintf(&t, "%lu\n", ULONG_MAX) < 0) {
1191 log_oom();
1192 return;
1193 }
1194
1195 r = sysctl_write("fs/file-max", t);
1196 if (r < 0)
1197 log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.file-max, ignoring: %m");
1198 #endif
1199
1200 #if BUMP_PROC_SYS_FS_FILE_MAX && BUMP_PROC_SYS_FS_NR_OPEN
1201 t = mfree(t);
1202 #endif
1203
1204 #if BUMP_PROC_SYS_FS_NR_OPEN
1205 int v = INT_MAX;
1206
1207 /* Arg! The kernel enforces maximum and minimum values on the fs.nr_open, but we don't really know what they
1208 * are. The expression by which the maximum is determined is dependent on the architecture, and is something we
1209 * don't really want to copy to userspace, as it is dependent on implementation details of the kernel. Since
1210 * the kernel doesn't expose the maximum value to us, we can only try and hope. Hence, let's start with
1211 * INT_MAX, and then keep halving the value until we find one that works. Ugly? Yes, absolutely, but kernel
1212 * APIs are kernel APIs, so what do can we do... 🤯 */
1213
1214 for (;;) {
1215 int k;
1216
1217 v &= ~(__SIZEOF_POINTER__ - 1); /* Round down to next multiple of the pointer size */
1218 if (v < 1024) {
1219 log_warning("Can't bump fs.nr_open, value too small.");
1220 break;
1221 }
1222
1223 k = read_nr_open();
1224 if (k < 0) {
1225 log_error_errno(k, "Failed to read fs.nr_open: %m");
1226 break;
1227 }
1228 if (k >= v) { /* Already larger */
1229 log_debug("Skipping bump, value is already larger.");
1230 break;
1231 }
1232
1233 if (asprintf(&t, "%i\n", v) < 0) {
1234 log_oom();
1235 return;
1236 }
1237
1238 r = sysctl_write("fs/nr_open", t);
1239 t = mfree(t);
1240 if (r == -EINVAL) {
1241 log_debug("Couldn't write fs.nr_open as %i, halving it.", v);
1242 v /= 2;
1243 continue;
1244 }
1245 if (r < 0) {
1246 log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.nr_open, ignoring: %m");
1247 break;
1248 }
1249
1250 log_debug("Successfully bumped fs.nr_open to %i", v);
1251 break;
1252 }
1253 #endif
1254 }
1255
1256 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1257 int r, nr;
1258
1259 assert(saved_rlimit);
1260
1261 /* Save the original RLIMIT_NOFILE so that we can reset it later when transitioning from the initrd to the main
1262 * systemd or suchlike. */
1263 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1264 return log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1265
1266 /* Get the underlying absolute limit the kernel enforces */
1267 nr = read_nr_open();
1268
1269 /* Make sure forked processes get limits based on the original kernel setting */
1270 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1271 struct rlimit *rl;
1272
1273 rl = newdup(struct rlimit, saved_rlimit, 1);
1274 if (!rl)
1275 return log_oom();
1276
1277 /* Bump the hard limit for system services to a substantially higher value. The default hard limit
1278 * current kernels set is pretty low (4K), mostly for historical reasons. According to kernel
1279 * developers, the fd handling in recent kernels has been optimized substantially enough, so that we
1280 * can bump the limit now, without paying too high a price in memory or performance. Note however that
1281 * we only bump the hard limit, not the soft limit. That's because select() works the way it works, and
1282 * chokes on fds >= 1024. If we'd bump the soft limit globally, it might accidentally happen to
1283 * unexpecting programs that they get fds higher than what they can process using select(). By only
1284 * bumping the hard limit but leaving the low limit as it is we avoid this pitfall: programs that are
1285 * written by folks aware of the select() problem in mind (and thus use poll()/epoll instead of
1286 * select(), the way everybody should) can explicitly opt into high fds by bumping their soft limit
1287 * beyond 1024, to the hard limit we pass. */
1288 if (arg_system)
1289 rl->rlim_max = MIN((rlim_t) nr, MAX(rl->rlim_max, (rlim_t) HIGH_RLIMIT_NOFILE));
1290
1291 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1292 }
1293
1294 /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for
1295 * both hard and soft. */
1296 r = setrlimit_closest(RLIMIT_NOFILE, &RLIMIT_MAKE_CONST(nr));
1297 if (r < 0)
1298 return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1299
1300 return 0;
1301 }
1302
1303 static int bump_rlimit_memlock(struct rlimit *saved_rlimit) {
1304 int r;
1305
1306 assert(saved_rlimit);
1307
1308 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK which should
1309 * normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's bump
1310 * the value high enough for our user. */
1311
1312 if (getrlimit(RLIMIT_MEMLOCK, saved_rlimit) < 0)
1313 return log_warning_errno(errno, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1314
1315 r = setrlimit_closest(RLIMIT_MEMLOCK, &RLIMIT_MAKE_CONST(HIGH_RLIMIT_MEMLOCK));
1316 if (r < 0)
1317 return log_warning_errno(r, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1318
1319 return 0;
1320 }
1321
1322 static void test_usr(void) {
1323
1324 /* Check that /usr is not a separate fs */
1325
1326 if (dir_is_empty("/usr") <= 0)
1327 return;
1328
1329 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1330 "Some things will probably break (sometimes even silently) in mysterious ways. "
1331 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1332 }
1333
1334 static int enforce_syscall_archs(Set *archs) {
1335 #if HAVE_SECCOMP
1336 int r;
1337
1338 if (!is_seccomp_available())
1339 return 0;
1340
1341 r = seccomp_restrict_archs(arg_syscall_archs);
1342 if (r < 0)
1343 return log_error_errno(r, "Failed to enforce system call architecture restrication: %m");
1344 #endif
1345 return 0;
1346 }
1347
1348 static int status_welcome(void) {
1349 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1350 int r;
1351
1352 if (IN_SET(arg_show_status, SHOW_STATUS_NO, SHOW_STATUS_AUTO))
1353 return 0;
1354
1355 r = parse_os_release(NULL,
1356 "PRETTY_NAME", &pretty_name,
1357 "ANSI_COLOR", &ansi_color,
1358 NULL);
1359 if (r < 0)
1360 log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r,
1361 "Failed to read os-release file, ignoring: %m");
1362
1363 if (log_get_show_color())
1364 return status_printf(NULL, false, false,
1365 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1366 isempty(ansi_color) ? "1" : ansi_color,
1367 isempty(pretty_name) ? "Linux" : pretty_name);
1368 else
1369 return status_printf(NULL, false, false,
1370 "\nWelcome to %s!\n",
1371 isempty(pretty_name) ? "Linux" : pretty_name);
1372 }
1373
1374 static int write_container_id(void) {
1375 const char *c;
1376 int r;
1377
1378 c = getenv("container");
1379 if (isempty(c))
1380 return 0;
1381
1382 RUN_WITH_UMASK(0022)
1383 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1384 if (r < 0)
1385 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
1386
1387 return 1;
1388 }
1389
1390 static int bump_unix_max_dgram_qlen(void) {
1391 _cleanup_free_ char *qlen = NULL;
1392 unsigned long v;
1393 int r;
1394
1395 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1396 * really really early during boot, so that it is actually applied to all our sockets, including the
1397 * $NOTIFY_SOCKET one. */
1398
1399 r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1400 if (r < 0)
1401 return log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1402
1403 r = safe_atolu(qlen, &v);
1404 if (r < 0)
1405 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen);
1406
1407 if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1408 return 0;
1409
1410 r = write_string_filef("/proc/sys/net/unix/max_dgram_qlen", WRITE_STRING_FILE_DISABLE_BUFFER, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN);
1411 if (r < 0)
1412 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1413 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1414
1415 return 1;
1416 }
1417
1418 static int fixup_environment(void) {
1419 _cleanup_free_ char *term = NULL;
1420 const char *t;
1421 int r;
1422
1423 /* Only fix up the environment when we are started as PID 1 */
1424 if (getpid_cached() != 1)
1425 return 0;
1426
1427 /* We expect the environment to be set correctly if run inside a container. */
1428 if (detect_container() > 0)
1429 return 0;
1430
1431 /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1432 * device used by the console. We try to make a better guess here since some consoles might not have support
1433 * for color mode for example.
1434 *
1435 * However if TERM was configured through the kernel command line then leave it alone. */
1436 r = proc_cmdline_get_key("TERM", 0, &term);
1437 if (r < 0)
1438 return r;
1439
1440 t = term ?: default_term_for_tty("/dev/console");
1441
1442 if (setenv("TERM", t, 1) < 0)
1443 return -errno;
1444
1445 return 0;
1446 }
1447
1448 static void redirect_telinit(int argc, char *argv[]) {
1449
1450 /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1451
1452 #if HAVE_SYSV_COMPAT
1453 if (getpid_cached() == 1)
1454 return;
1455
1456 if (!strstr(program_invocation_short_name, "init"))
1457 return;
1458
1459 execv(SYSTEMCTL_BINARY_PATH, argv);
1460 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1461 exit(EXIT_FAILURE);
1462 #endif
1463 }
1464
1465 static int become_shutdown(
1466 const char *shutdown_verb,
1467 int retval) {
1468
1469 char log_level[DECIMAL_STR_MAX(int) + 1],
1470 exit_code[DECIMAL_STR_MAX(uint8_t) + 1],
1471 timeout[DECIMAL_STR_MAX(usec_t) + 1];
1472
1473 const char* command_line[13] = {
1474 SYSTEMD_SHUTDOWN_BINARY_PATH,
1475 shutdown_verb,
1476 "--timeout", timeout,
1477 "--log-level", log_level,
1478 "--log-target",
1479 };
1480
1481 _cleanup_strv_free_ char **env_block = NULL;
1482 size_t pos = 7;
1483 int r;
1484
1485 assert(shutdown_verb);
1486 assert(!command_line[pos]);
1487 env_block = strv_copy(environ);
1488
1489 xsprintf(log_level, "%d", log_get_max_level());
1490 xsprintf(timeout, "%" PRI_USEC "us", arg_default_timeout_stop_usec);
1491
1492 switch (log_get_target()) {
1493
1494 case LOG_TARGET_KMSG:
1495 case LOG_TARGET_JOURNAL_OR_KMSG:
1496 case LOG_TARGET_SYSLOG_OR_KMSG:
1497 command_line[pos++] = "kmsg";
1498 break;
1499
1500 case LOG_TARGET_NULL:
1501 command_line[pos++] = "null";
1502 break;
1503
1504 case LOG_TARGET_CONSOLE:
1505 default:
1506 command_line[pos++] = "console";
1507 break;
1508 };
1509
1510 if (log_get_show_color())
1511 command_line[pos++] = "--log-color";
1512
1513 if (log_get_show_location())
1514 command_line[pos++] = "--log-location";
1515
1516 if (streq(shutdown_verb, "exit")) {
1517 command_line[pos++] = "--exit-code";
1518 command_line[pos++] = exit_code;
1519 xsprintf(exit_code, "%d", retval);
1520 }
1521
1522 assert(pos < ELEMENTSOF(command_line));
1523
1524 if (streq(shutdown_verb, "reboot") &&
1525 arg_shutdown_watchdog > 0 &&
1526 arg_shutdown_watchdog != USEC_INFINITY) {
1527
1528 char *e;
1529
1530 /* If we reboot let's set the shutdown
1531 * watchdog and tell the shutdown binary to
1532 * repeatedly ping it */
1533 r = watchdog_set_timeout(&arg_shutdown_watchdog);
1534 watchdog_close(r < 0);
1535
1536 /* Tell the binary how often to ping, ignore failure */
1537 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
1538 (void) strv_consume(&env_block, e);
1539
1540 if (arg_watchdog_device &&
1541 asprintf(&e, "WATCHDOG_DEVICE=%s", arg_watchdog_device) > 0)
1542 (void) strv_consume(&env_block, e);
1543 } else
1544 watchdog_close(true);
1545
1546 /* Avoid the creation of new processes forked by the
1547 * kernel; at this point, we will not listen to the
1548 * signals anyway */
1549 if (detect_container() <= 0)
1550 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
1551
1552 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
1553 return -errno;
1554 }
1555
1556 static void initialize_clock(void) {
1557 int r;
1558
1559 if (clock_is_localtime(NULL) > 0) {
1560 int min;
1561
1562 /*
1563 * The very first call of settimeofday() also does a time warp in the kernel.
1564 *
1565 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1566 * of maintaining the RTC and do all adjustments. This matches the behavior of Windows, which leaves
1567 * the RTC alone if the registry tells that the RTC runs in UTC.
1568 */
1569 r = clock_set_timezone(&min);
1570 if (r < 0)
1571 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1572 else
1573 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1574
1575 } else if (!in_initrd()) {
1576 /*
1577 * Do a dummy very first call to seal the kernel's time warp magic.
1578 *
1579 * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1580 * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1581 * until we reach the real system.
1582 *
1583 * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1584 * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1585 * be treated as UTC that way.
1586 */
1587 (void) clock_reset_timewarp();
1588 }
1589
1590 r = clock_apply_epoch();
1591 if (r < 0)
1592 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1593 else if (r > 0)
1594 log_info("System time before build time, advancing clock.");
1595 }
1596
1597 static void initialize_coredump(bool skip_setup) {
1598 #if ENABLE_COREDUMP
1599 if (getpid_cached() != 1)
1600 return;
1601
1602 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1603 * will process core dumps for system services by default. */
1604 if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0)
1605 log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m");
1606
1607 /* But at the same time, turn off the core_pattern logic by default, so that no
1608 * coredumps are stored until the systemd-coredump tool is enabled via
1609 * sysctl. However it can be changed via the kernel command line later so core
1610 * dumps can still be generated during early startup and in initramfs. */
1611 if (!skip_setup)
1612 disable_coredumps();
1613 #endif
1614 }
1615
1616 static void initialize_core_pattern(bool skip_setup) {
1617 int r;
1618
1619 if (skip_setup || !arg_early_core_pattern)
1620 return;
1621
1622 if (getpid_cached() != 1)
1623 return;
1624
1625 r = write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern, WRITE_STRING_FILE_DISABLE_BUFFER);
1626 if (r < 0)
1627 log_warning_errno(r, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", arg_early_core_pattern);
1628 }
1629
1630 static void do_reexecute(
1631 int argc,
1632 char *argv[],
1633 const struct rlimit *saved_rlimit_nofile,
1634 const struct rlimit *saved_rlimit_memlock,
1635 FDSet *fds,
1636 const char *switch_root_dir,
1637 const char *switch_root_init,
1638 const char **ret_error_message) {
1639
1640 unsigned i, j, args_size;
1641 const char **args;
1642 int r;
1643
1644 assert(saved_rlimit_nofile);
1645 assert(saved_rlimit_memlock);
1646 assert(ret_error_message);
1647
1648 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1649 * we do that */
1650 watchdog_close(true);
1651
1652 /* Reset the RLIMIT_NOFILE to the kernel default, so that the new systemd can pass the kernel default to its
1653 * child processes */
1654
1655 if (saved_rlimit_nofile->rlim_cur > 0)
1656 (void) setrlimit(RLIMIT_NOFILE, saved_rlimit_nofile);
1657 if (saved_rlimit_memlock->rlim_cur != (rlim_t) -1)
1658 (void) setrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock);
1659
1660 if (switch_root_dir) {
1661 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1662 * SIGCHLD for them after deserializing. */
1663 broadcast_signal(SIGTERM, false, true, arg_default_timeout_stop_usec);
1664
1665 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1666 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1667 if (r < 0)
1668 log_error_errno(r, "Failed to switch root, trying to continue: %m");
1669 }
1670
1671 args_size = MAX(6, argc+1);
1672 args = newa(const char*, args_size);
1673
1674 if (!switch_root_init) {
1675 char sfd[DECIMAL_STR_MAX(int) + 1];
1676
1677 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1678 * the user didn't specify an explicit init to spawn. */
1679
1680 assert(arg_serialization);
1681 assert(fds);
1682
1683 xsprintf(sfd, "%i", fileno(arg_serialization));
1684
1685 i = 0;
1686 args[i++] = SYSTEMD_BINARY_PATH;
1687 if (switch_root_dir)
1688 args[i++] = "--switched-root";
1689 args[i++] = arg_system ? "--system" : "--user";
1690 args[i++] = "--deserialize";
1691 args[i++] = sfd;
1692 args[i++] = NULL;
1693
1694 assert(i <= args_size);
1695
1696 /*
1697 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do this is on
1698 * its own on exec(), but it will do it on exit(). Hence, to ensure we get a summary here, fork() off
1699 * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1700 * before proceeding into the exec().
1701 */
1702 valgrind_summary_hack();
1703
1704 (void) execv(args[0], (char* const*) args);
1705 log_debug_errno(errno, "Failed to execute our own binary, trying fallback: %m");
1706 }
1707
1708 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1709 * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1710 * doesn't matter.) */
1711
1712 arg_serialization = safe_fclose(arg_serialization);
1713 fds = fdset_free(fds);
1714
1715 /* Reopen the console */
1716 (void) make_console_stdio();
1717
1718 for (j = 1, i = 1; j < (unsigned) argc; j++)
1719 args[i++] = argv[j];
1720 args[i++] = NULL;
1721 assert(i <= args_size);
1722
1723 /* Reenable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1724 (void) reset_all_signal_handlers();
1725 (void) reset_signal_mask();
1726
1727 if (switch_root_init) {
1728 args[0] = switch_root_init;
1729 (void) execv(args[0], (char* const*) args);
1730 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
1731 }
1732
1733 args[0] = "/sbin/init";
1734 (void) execv(args[0], (char* const*) args);
1735 r = -errno;
1736
1737 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
1738 ANSI_HIGHLIGHT_RED " !! " ANSI_NORMAL,
1739 "Failed to execute /sbin/init");
1740
1741 if (r == -ENOENT) {
1742 log_warning("No /sbin/init, trying fallback");
1743
1744 args[0] = "/bin/sh";
1745 args[1] = NULL;
1746 (void) execv(args[0], (char* const*) args);
1747 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
1748 } else
1749 log_warning_errno(r, "Failed to execute /sbin/init, giving up: %m");
1750
1751 *ret_error_message = "Failed to execute fallback shell";
1752 }
1753
1754 static int invoke_main_loop(
1755 Manager *m,
1756 bool *ret_reexecute,
1757 int *ret_retval, /* Return parameters relevant for shutting down */
1758 const char **ret_shutdown_verb, /* … */
1759 FDSet **ret_fds, /* Return parameters for reexecuting */
1760 char **ret_switch_root_dir, /* … */
1761 char **ret_switch_root_init, /* … */
1762 const char **ret_error_message) {
1763
1764 int r;
1765
1766 assert(m);
1767 assert(ret_reexecute);
1768 assert(ret_retval);
1769 assert(ret_shutdown_verb);
1770 assert(ret_fds);
1771 assert(ret_switch_root_dir);
1772 assert(ret_switch_root_init);
1773 assert(ret_error_message);
1774
1775 for (;;) {
1776 r = manager_loop(m);
1777 if (r < 0) {
1778 *ret_error_message = "Failed to run main loop";
1779 return log_emergency_errno(r, "Failed to run main loop: %m");
1780 }
1781
1782 switch ((ManagerObjective) r) {
1783
1784 case MANAGER_RELOAD: {
1785 LogTarget saved_log_target;
1786 int saved_log_level;
1787
1788 log_info("Reloading.");
1789
1790 /* First, save any overridden log level/target, then parse the configuration file, which might
1791 * change the log level to new settings. */
1792
1793 saved_log_level = m->log_level_overridden ? log_get_max_level() : -1;
1794 saved_log_target = m->log_target_overridden ? log_get_target() : _LOG_TARGET_INVALID;
1795
1796 r = parse_config_file();
1797 if (r < 0)
1798 log_warning_errno(r, "Failed to parse config file, ignoring: %m");
1799
1800 set_manager_defaults(m);
1801
1802 if (saved_log_level >= 0)
1803 manager_override_log_level(m, saved_log_level);
1804 if (saved_log_target >= 0)
1805 manager_override_log_target(m, saved_log_target);
1806
1807 r = manager_reload(m);
1808 if (r < 0)
1809 /* Reloading failed before the point of no return. Let's continue running as if nothing happened. */
1810 m->objective = MANAGER_OK;
1811
1812 break;
1813 }
1814
1815 case MANAGER_REEXECUTE:
1816
1817 r = prepare_reexecute(m, &arg_serialization, ret_fds, false);
1818 if (r < 0) {
1819 *ret_error_message = "Failed to prepare for reexecution";
1820 return r;
1821 }
1822
1823 log_notice("Reexecuting.");
1824
1825 *ret_reexecute = true;
1826 *ret_retval = EXIT_SUCCESS;
1827 *ret_shutdown_verb = NULL;
1828 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1829
1830 return 0;
1831
1832 case MANAGER_SWITCH_ROOT:
1833 if (!m->switch_root_init) {
1834 r = prepare_reexecute(m, &arg_serialization, ret_fds, true);
1835 if (r < 0) {
1836 *ret_error_message = "Failed to prepare for reexecution";
1837 return r;
1838 }
1839 } else
1840 *ret_fds = NULL;
1841
1842 log_notice("Switching root.");
1843
1844 *ret_reexecute = true;
1845 *ret_retval = EXIT_SUCCESS;
1846 *ret_shutdown_verb = NULL;
1847
1848 /* Steal the switch root parameters */
1849 *ret_switch_root_dir = m->switch_root;
1850 *ret_switch_root_init = m->switch_root_init;
1851 m->switch_root = m->switch_root_init = NULL;
1852
1853 return 0;
1854
1855 case MANAGER_EXIT:
1856
1857 if (MANAGER_IS_USER(m)) {
1858 log_debug("Exit.");
1859
1860 *ret_reexecute = false;
1861 *ret_retval = m->return_value;
1862 *ret_shutdown_verb = NULL;
1863 *ret_fds = NULL;
1864 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1865
1866 return 0;
1867 }
1868
1869 _fallthrough_;
1870 case MANAGER_REBOOT:
1871 case MANAGER_POWEROFF:
1872 case MANAGER_HALT:
1873 case MANAGER_KEXEC: {
1874 static const char * const table[_MANAGER_OBJECTIVE_MAX] = {
1875 [MANAGER_EXIT] = "exit",
1876 [MANAGER_REBOOT] = "reboot",
1877 [MANAGER_POWEROFF] = "poweroff",
1878 [MANAGER_HALT] = "halt",
1879 [MANAGER_KEXEC] = "kexec",
1880 };
1881
1882 log_notice("Shutting down.");
1883
1884 *ret_reexecute = false;
1885 *ret_retval = m->return_value;
1886 assert_se(*ret_shutdown_verb = table[m->objective]);
1887 *ret_fds = NULL;
1888 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1889
1890 return 0;
1891 }
1892
1893 default:
1894 assert_not_reached("Unknown or unexpected manager objective.");
1895 }
1896 }
1897 }
1898
1899 static void log_execution_mode(bool *ret_first_boot) {
1900 assert(ret_first_boot);
1901
1902 if (arg_system) {
1903 int v;
1904
1905 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1906 arg_action == ACTION_TEST ? "test " : "" );
1907
1908 v = detect_virtualization();
1909 if (v > 0)
1910 log_info("Detected virtualization %s.", virtualization_to_string(v));
1911
1912 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1913
1914 if (in_initrd()) {
1915 *ret_first_boot = false;
1916 log_info("Running in initial RAM disk.");
1917 } else {
1918 /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
1919 * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
1920 * doesn't it's unpopulated. This allows container managers and installers to provision a
1921 * couple of files already. If the container manager wants to provision the machine ID itself
1922 * it should pass $container_uuid to PID 1. */
1923
1924 *ret_first_boot = access("/etc/machine-id", F_OK) < 0;
1925 if (*ret_first_boot)
1926 log_info("Running with unpopulated /etc.");
1927 }
1928 } else {
1929 if (DEBUG_LOGGING) {
1930 _cleanup_free_ char *t;
1931
1932 t = uid_to_name(getuid());
1933 log_debug(PACKAGE_STRING " running in %suser mode for user " UID_FMT "/%s. (" SYSTEMD_FEATURES ")",
1934 arg_action == ACTION_TEST ? " test" : "", getuid(), strna(t));
1935 }
1936
1937 *ret_first_boot = false;
1938 }
1939 }
1940
1941 static int initialize_runtime(
1942 bool skip_setup,
1943 struct rlimit *saved_rlimit_nofile,
1944 struct rlimit *saved_rlimit_memlock,
1945 const char **ret_error_message) {
1946
1947 int r;
1948
1949 assert(ret_error_message);
1950
1951 /* Sets up various runtime parameters. Many of these initializations are conditionalized:
1952 *
1953 * - Some only apply to --system instances
1954 * - Some only apply to --user instances
1955 * - Some only apply when we first start up, but not when we reexecute
1956 */
1957
1958 if (arg_action != ACTION_RUN)
1959 return 0;
1960
1961 if (arg_system) {
1962 /* Make sure we leave a core dump without panicing the kernel. */
1963 install_crash_handler();
1964
1965 if (!skip_setup) {
1966 r = mount_cgroup_controllers();
1967 if (r < 0) {
1968 *ret_error_message = "Failed to mount cgroup hierarchies";
1969 return r;
1970 }
1971
1972 status_welcome();
1973 hostname_setup();
1974 machine_id_setup(NULL, arg_machine_id, NULL);
1975 loopback_setup();
1976 bump_unix_max_dgram_qlen();
1977 bump_file_max_and_nr_open();
1978 test_usr();
1979 write_container_id();
1980 }
1981
1982 if (arg_watchdog_device) {
1983 r = watchdog_set_device(arg_watchdog_device);
1984 if (r < 0)
1985 log_warning_errno(r, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device);
1986 }
1987
1988 if (arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY)
1989 watchdog_set_timeout(&arg_runtime_watchdog);
1990 }
1991
1992 if (arg_timer_slack_nsec != NSEC_INFINITY)
1993 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1994 log_warning_errno(errno, "Failed to adjust timer slack, ignoring: %m");
1995
1996 if (arg_system && !cap_test_all(arg_capability_bounding_set)) {
1997 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
1998 if (r < 0) {
1999 *ret_error_message = "Failed to drop capability bounding set of usermode helpers";
2000 return log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
2001 }
2002
2003 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
2004 if (r < 0) {
2005 *ret_error_message = "Failed to drop capability bounding set";
2006 return log_emergency_errno(r, "Failed to drop capability bounding set: %m");
2007 }
2008 }
2009
2010 if (arg_system && arg_no_new_privs) {
2011 if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) {
2012 *ret_error_message = "Failed to disable new privileges";
2013 return log_emergency_errno(errno, "Failed to disable new privileges: %m");
2014 }
2015 }
2016
2017 if (arg_syscall_archs) {
2018 r = enforce_syscall_archs(arg_syscall_archs);
2019 if (r < 0) {
2020 *ret_error_message = "Failed to set syscall architectures";
2021 return r;
2022 }
2023 }
2024
2025 if (!arg_system)
2026 /* Become reaper of our children */
2027 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
2028 log_warning_errno(errno, "Failed to make us a subreaper: %m");
2029
2030 /* Bump up RLIMIT_NOFILE for systemd itself */
2031 (void) bump_rlimit_nofile(saved_rlimit_nofile);
2032 (void) bump_rlimit_memlock(saved_rlimit_memlock);
2033
2034 return 0;
2035 }
2036
2037 static int do_queue_default_job(
2038 Manager *m,
2039 const char **ret_error_message) {
2040
2041 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
2042 Job *default_unit_job;
2043 Unit *target = NULL;
2044 int r;
2045
2046 log_debug("Activating default unit: %s", arg_default_unit);
2047
2048 r = manager_load_startable_unit_or_warn(m, arg_default_unit, NULL, &target);
2049 if (r < 0) {
2050 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET);
2051
2052 r = manager_load_startable_unit_or_warn(m, SPECIAL_RESCUE_TARGET, NULL, &target);
2053 if (r < 0) {
2054 *ret_error_message = r == -ERFKILL ? "Rescue target masked"
2055 : "Failed to load rescue target";
2056 return r;
2057 }
2058 }
2059
2060 assert(target->load_state == UNIT_LOADED);
2061
2062 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, &error, &default_unit_job);
2063 if (r == -EPERM) {
2064 log_debug_errno(r, "Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
2065
2066 sd_bus_error_free(&error);
2067
2068 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, &error, &default_unit_job);
2069 if (r < 0) {
2070 *ret_error_message = "Failed to start default target";
2071 return log_emergency_errno(r, "Failed to start default target: %s", bus_error_message(&error, r));
2072 }
2073
2074 } else if (r < 0) {
2075 *ret_error_message = "Failed to isolate default target";
2076 return log_emergency_errno(r, "Failed to isolate default target: %s", bus_error_message(&error, r));
2077 }
2078
2079 m->default_unit_job_id = default_unit_job->id;
2080
2081 return 0;
2082 }
2083
2084 static void free_arguments(void) {
2085
2086 /* Frees all arg_* variables, with the exception of arg_serialization */
2087 rlimit_free_all(arg_default_rlimit);
2088
2089 arg_default_unit = mfree(arg_default_unit);
2090 arg_confirm_spawn = mfree(arg_confirm_spawn);
2091 arg_default_environment = strv_free(arg_default_environment);
2092 arg_syscall_archs = set_free(arg_syscall_archs);
2093 }
2094
2095 static int load_configuration(int argc, char **argv, const char **ret_error_message) {
2096 int r;
2097
2098 assert(ret_error_message);
2099
2100 arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U);
2101
2102 r = parse_config_file();
2103 if (r < 0) {
2104 *ret_error_message = "Failed to parse config file";
2105 return r;
2106 }
2107
2108 if (arg_system) {
2109 r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
2110 if (r < 0)
2111 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
2112 }
2113
2114 /* Note that this also parses bits from the kernel command line, including "debug". */
2115 log_parse_environment();
2116
2117 r = parse_argv(argc, argv);
2118 if (r < 0) {
2119 *ret_error_message = "Failed to parse commandline arguments";
2120 return r;
2121 }
2122
2123 /* Initialize default unit */
2124 if (!arg_default_unit) {
2125 arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET);
2126 if (!arg_default_unit) {
2127 *ret_error_message = "Failed to set default unit";
2128 return log_oom();
2129 }
2130 }
2131
2132 /* Initialize the show status setting if it hasn't been set explicitly yet */
2133 if (arg_show_status == _SHOW_STATUS_INVALID)
2134 arg_show_status = SHOW_STATUS_YES;
2135
2136 return 0;
2137 }
2138
2139 static int safety_checks(void) {
2140
2141 if (getpid_cached() == 1 &&
2142 arg_action != ACTION_RUN) {
2143 log_error("Unsupported execution mode while PID 1.");
2144 return -EPERM;
2145 }
2146
2147 if (getpid_cached() == 1 &&
2148 !arg_system) {
2149 log_error("Can't run --user mode as PID 1.");
2150 return -EPERM;
2151 }
2152
2153 if (arg_action == ACTION_RUN &&
2154 arg_system &&
2155 getpid_cached() != 1) {
2156 log_error("Can't run system mode unless PID 1.");
2157 return -EPERM;
2158 }
2159
2160 if (arg_action == ACTION_TEST &&
2161 geteuid() == 0) {
2162 log_error("Don't run test mode as root.");
2163 return -EPERM;
2164 }
2165
2166 if (!arg_system &&
2167 arg_action == ACTION_RUN &&
2168 sd_booted() <= 0) {
2169 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
2170 return -EOPNOTSUPP;
2171 }
2172
2173 if (!arg_system &&
2174 arg_action == ACTION_RUN &&
2175 !getenv("XDG_RUNTIME_DIR")) {
2176 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
2177 return -EUNATCH;
2178 }
2179
2180 if (arg_system &&
2181 arg_action == ACTION_RUN &&
2182 running_in_chroot() > 0) {
2183 log_error("Cannot be run in a chroot() environment.");
2184 return -EOPNOTSUPP;
2185 }
2186
2187 return 0;
2188 }
2189
2190 static int initialize_security(
2191 bool *loaded_policy,
2192 dual_timestamp *security_start_timestamp,
2193 dual_timestamp *security_finish_timestamp,
2194 const char **ret_error_message) {
2195
2196 int r;
2197
2198 assert(loaded_policy);
2199 assert(security_start_timestamp);
2200 assert(security_finish_timestamp);
2201 assert(ret_error_message);
2202
2203 dual_timestamp_get(security_start_timestamp);
2204
2205 r = mac_selinux_setup(loaded_policy);
2206 if (r < 0) {
2207 *ret_error_message = "Failed to load SELinux policy";
2208 return r;
2209 }
2210
2211 r = mac_smack_setup(loaded_policy);
2212 if (r < 0) {
2213 *ret_error_message = "Failed to load SMACK policy";
2214 return r;
2215 }
2216
2217 r = ima_setup();
2218 if (r < 0) {
2219 *ret_error_message = "Failed to load IMA policy";
2220 return r;
2221 }
2222
2223 dual_timestamp_get(security_finish_timestamp);
2224 return 0;
2225 }
2226
2227 static void test_summary(Manager *m) {
2228 assert(m);
2229
2230 printf("-> By units:\n");
2231 manager_dump_units(m, stdout, "\t");
2232
2233 printf("-> By jobs:\n");
2234 manager_dump_jobs(m, stdout, "\t");
2235 }
2236
2237 static int collect_fds(FDSet **ret_fds, const char **ret_error_message) {
2238 int r;
2239
2240 assert(ret_fds);
2241 assert(ret_error_message);
2242
2243 r = fdset_new_fill(ret_fds);
2244 if (r < 0) {
2245 *ret_error_message = "Failed to allocate fd set";
2246 return log_emergency_errno(r, "Failed to allocate fd set: %m");
2247 }
2248
2249 fdset_cloexec(*ret_fds, true);
2250
2251 if (arg_serialization)
2252 assert_se(fdset_remove(*ret_fds, fileno(arg_serialization)) >= 0);
2253
2254 return 0;
2255 }
2256
2257 static void setup_console_terminal(bool skip_setup) {
2258
2259 if (!arg_system)
2260 return;
2261
2262 /* Become a session leader if we aren't one yet. */
2263 (void) setsid();
2264
2265 /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2266 * tty. */
2267 (void) release_terminal();
2268
2269 /* Reset the console, but only if this is really init and we are freshly booted */
2270 if (getpid_cached() == 1 && !skip_setup)
2271 (void) console_setup();
2272 }
2273
2274 static bool early_skip_setup_check(int argc, char *argv[]) {
2275 bool found_deserialize = false;
2276 int i;
2277
2278 /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2279 * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2280 * anyway, even if in that case we also do deserialization. */
2281
2282 for (i = 1; i < argc; i++) {
2283 if (streq(argv[i], "--switched-root"))
2284 return false; /* If we switched root, don't skip the setup. */
2285 else if (streq(argv[i], "--deserialize"))
2286 found_deserialize = true;
2287 }
2288
2289 return found_deserialize; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2290 }
2291
2292 int main(int argc, char *argv[]) {
2293
2294 dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL, userspace_timestamp = DUAL_TIMESTAMP_NULL, kernel_timestamp = DUAL_TIMESTAMP_NULL,
2295 security_start_timestamp = DUAL_TIMESTAMP_NULL, security_finish_timestamp = DUAL_TIMESTAMP_NULL;
2296 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0), saved_rlimit_memlock = RLIMIT_MAKE_CONST((rlim_t) -1);
2297 bool skip_setup, loaded_policy = false, queue_default_job = false, first_boot = false, reexecute = false;
2298 char *switch_root_dir = NULL, *switch_root_init = NULL;
2299 usec_t before_startup, after_startup;
2300 static char systemd[] = "systemd";
2301 char timespan[FORMAT_TIMESPAN_MAX];
2302 const char *shutdown_verb = NULL, *error_message = NULL;
2303 int r, retval = EXIT_FAILURE;
2304 Manager *m = NULL;
2305 FDSet *fds = NULL;
2306
2307 /* SysV compatibility: redirect init → telinit */
2308 redirect_telinit(argc, argv);
2309
2310 /* Take timestamps early on */
2311 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
2312 dual_timestamp_get(&userspace_timestamp);
2313
2314 /* Figure out whether we need to do initialize the system, or if we already did that because we are
2315 * reexecuting */
2316 skip_setup = early_skip_setup_check(argc, argv);
2317
2318 /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2319 * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
2320 program_invocation_short_name = systemd;
2321 (void) prctl(PR_SET_NAME, systemd);
2322
2323 /* Save the original command line */
2324 saved_argv = argv;
2325 saved_argc = argc;
2326
2327 /* Make sure that if the user says "syslog" we actually log to the journal. */
2328 log_set_upgrade_syslog_to_journal(true);
2329
2330 if (getpid_cached() == 1) {
2331 /* When we run as PID 1 force system mode */
2332 arg_system = true;
2333
2334 /* Disable the umask logic */
2335 umask(0);
2336
2337 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2338 * activated yet (even though the log socket for it exists). */
2339 log_set_prohibit_ipc(true);
2340
2341 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2342 * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2343 * child process right before execve()'ing the actual binary, at a point in time where socket
2344 * activation stderr/stdout area already set up. */
2345 log_set_always_reopen_console(true);
2346
2347 if (detect_container() <= 0) {
2348
2349 /* Running outside of a container as PID 1 */
2350 log_set_target(LOG_TARGET_KMSG);
2351 log_open();
2352
2353 if (in_initrd())
2354 initrd_timestamp = userspace_timestamp;
2355
2356 if (!skip_setup) {
2357 r = mount_setup_early();
2358 if (r < 0) {
2359 error_message = "Failed to mount early API filesystems";
2360 goto finish;
2361 }
2362
2363 r = initialize_security(
2364 &loaded_policy,
2365 &security_start_timestamp,
2366 &security_finish_timestamp,
2367 &error_message);
2368 if (r < 0)
2369 goto finish;
2370 }
2371
2372 if (mac_selinux_init() < 0) {
2373 error_message = "Failed to initialize SELinux policy";
2374 goto finish;
2375 }
2376
2377 if (!skip_setup)
2378 initialize_clock();
2379
2380 /* Set the default for later on, but don't actually open the logs like this for now. Note that
2381 * if we are transitioning from the initrd there might still be journal fd open, and we
2382 * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2383 * elsewhere. */
2384 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
2385
2386 } else {
2387 /* Running inside a container, as PID 1 */
2388 log_set_target(LOG_TARGET_CONSOLE);
2389 log_open();
2390
2391 /* For later on, see above... */
2392 log_set_target(LOG_TARGET_JOURNAL);
2393
2394 /* clear the kernel timestamp,
2395 * because we are in a container */
2396 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2397 }
2398
2399 initialize_coredump(skip_setup);
2400
2401 r = fixup_environment();
2402 if (r < 0) {
2403 log_emergency_errno(r, "Failed to fix up PID 1 environment: %m");
2404 error_message = "Failed to fix up PID1 environment";
2405 goto finish;
2406 }
2407
2408 } else {
2409 /* Running as user instance */
2410 arg_system = false;
2411 log_set_target(LOG_TARGET_AUTO);
2412 log_open();
2413
2414 /* clear the kernel timestamp,
2415 * because we are not PID 1 */
2416 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2417 }
2418
2419 if (arg_system) {
2420 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2421 * they never log into the console. */
2422 log_show_color(colors_enabled());
2423
2424 r = make_null_stdio();
2425 if (r < 0)
2426 log_warning_errno(r, "Failed to redirect standard streams to /dev/null, ignoring: %m");
2427 }
2428
2429 /* Mount /proc, /sys and friends, so that /proc/cmdline and
2430 * /proc/$PID/fd is available. */
2431 if (getpid_cached() == 1) {
2432
2433 /* Load the kernel modules early. */
2434 if (!skip_setup)
2435 kmod_setup();
2436
2437 r = mount_setup(loaded_policy);
2438 if (r < 0) {
2439 error_message = "Failed to mount API filesystems";
2440 goto finish;
2441 }
2442 }
2443
2444 /* Reset all signal handlers. */
2445 (void) reset_all_signal_handlers();
2446 (void) ignore_signals(SIGNALS_IGNORE, -1);
2447
2448 r = load_configuration(argc, argv, &error_message);
2449 if (r < 0)
2450 goto finish;
2451
2452 r = safety_checks();
2453 if (r < 0)
2454 goto finish;
2455
2456 if (IN_SET(arg_action, ACTION_TEST, ACTION_HELP, ACTION_DUMP_CONFIGURATION_ITEMS, ACTION_DUMP_BUS_PROPERTIES))
2457 (void) pager_open(arg_pager_flags);
2458
2459 if (arg_action != ACTION_RUN)
2460 skip_setup = true;
2461
2462 if (arg_action == ACTION_HELP) {
2463 retval = help() < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
2464 goto finish;
2465 } else if (arg_action == ACTION_VERSION) {
2466 retval = version();
2467 goto finish;
2468 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
2469 unit_dump_config_items(stdout);
2470 retval = EXIT_SUCCESS;
2471 goto finish;
2472 } else if (arg_action == ACTION_DUMP_BUS_PROPERTIES) {
2473 dump_bus_properties(stdout);
2474 retval = EXIT_SUCCESS;
2475 goto finish;
2476 }
2477
2478 assert_se(IN_SET(arg_action, ACTION_RUN, ACTION_TEST));
2479
2480 /* Move out of the way, so that we won't block unmounts */
2481 assert_se(chdir("/") == 0);
2482
2483 if (arg_action == ACTION_RUN) {
2484
2485 /* A core pattern might have been specified via the cmdline. */
2486 initialize_core_pattern(skip_setup);
2487
2488 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2489 log_close();
2490
2491 /* Remember open file descriptors for later deserialization */
2492 r = collect_fds(&fds, &error_message);
2493 if (r < 0)
2494 goto finish;
2495
2496 /* Give up any control of the console, but make sure its initialized. */
2497 setup_console_terminal(skip_setup);
2498
2499 /* Open the logging devices, if possible and necessary */
2500 log_open();
2501 }
2502
2503 log_execution_mode(&first_boot);
2504
2505 r = initialize_runtime(skip_setup,
2506 &saved_rlimit_nofile,
2507 &saved_rlimit_memlock,
2508 &error_message);
2509 if (r < 0)
2510 goto finish;
2511
2512 r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER,
2513 arg_action == ACTION_TEST ? MANAGER_TEST_FULL : 0,
2514 &m);
2515 if (r < 0) {
2516 log_emergency_errno(r, "Failed to allocate manager object: %m");
2517 error_message = "Failed to allocate manager object";
2518 goto finish;
2519 }
2520
2521 m->timestamps[MANAGER_TIMESTAMP_KERNEL] = kernel_timestamp;
2522 m->timestamps[MANAGER_TIMESTAMP_INITRD] = initrd_timestamp;
2523 m->timestamps[MANAGER_TIMESTAMP_USERSPACE] = userspace_timestamp;
2524 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START)] = security_start_timestamp;
2525 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH)] = security_finish_timestamp;
2526
2527 set_manager_defaults(m);
2528 set_manager_settings(m);
2529 manager_set_first_boot(m, first_boot);
2530
2531 /* Remember whether we should queue the default job */
2532 queue_default_job = !arg_serialization || arg_switched_root;
2533
2534 before_startup = now(CLOCK_MONOTONIC);
2535
2536 r = manager_startup(m, arg_serialization, fds);
2537 if (r < 0) {
2538 error_message = "Failed to start up manager";
2539 goto finish;
2540 }
2541
2542 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2543 fds = fdset_free(fds);
2544 arg_serialization = safe_fclose(arg_serialization);
2545
2546 if (queue_default_job) {
2547 r = do_queue_default_job(m, &error_message);
2548 if (r < 0)
2549 goto finish;
2550 }
2551
2552 after_startup = now(CLOCK_MONOTONIC);
2553
2554 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
2555 "Loaded units and determined initial transaction in %s.",
2556 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
2557
2558 if (arg_action == ACTION_TEST) {
2559 test_summary(m);
2560 retval = EXIT_SUCCESS;
2561 goto finish;
2562 }
2563
2564 (void) invoke_main_loop(m,
2565 &reexecute,
2566 &retval,
2567 &shutdown_verb,
2568 &fds,
2569 &switch_root_dir,
2570 &switch_root_init,
2571 &error_message);
2572
2573 finish:
2574 pager_close();
2575
2576 if (m) {
2577 arg_shutdown_watchdog = m->shutdown_watchdog;
2578 m = manager_free(m);
2579 }
2580
2581 free_arguments();
2582 mac_selinux_finish();
2583
2584 if (reexecute)
2585 do_reexecute(argc, argv,
2586 &saved_rlimit_nofile,
2587 &saved_rlimit_memlock,
2588 fds,
2589 switch_root_dir,
2590 switch_root_init,
2591 &error_message); /* This only returns if reexecution failed */
2592
2593 arg_serialization = safe_fclose(arg_serialization);
2594 fds = fdset_free(fds);
2595
2596 #if HAVE_VALGRIND_VALGRIND_H
2597 /* If we are PID 1 and running under valgrind, then let's exit
2598 * here explicitly. valgrind will only generate nice output on
2599 * exit(), not on exec(), hence let's do the former not the
2600 * latter here. */
2601 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND) {
2602 /* Cleanup watchdog_device strings for valgrind. We need them
2603 * in become_shutdown() so normally we cannot free them yet. */
2604 watchdog_free_device();
2605 arg_watchdog_device = mfree(arg_watchdog_device);
2606 return retval;
2607 }
2608 #endif
2609
2610 if (shutdown_verb) {
2611 r = become_shutdown(shutdown_verb, retval);
2612 log_error_errno(r, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
2613 error_message = "Failed to execute shutdown binary";
2614 }
2615
2616 watchdog_free_device();
2617 arg_watchdog_device = mfree(arg_watchdog_device);
2618
2619 if (getpid_cached() == 1) {
2620 if (error_message)
2621 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2622 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
2623 "%s, freezing.", error_message);
2624 freeze_or_reboot();
2625 }
2626
2627 return retval;
2628 }
Unnamed repository; edit this file 'description' to name the repository.