]> git.ipfire.org Git - thirdparty/systemd.git/blob - src/core/main.c
projects / thirdparty / systemd.git / blob
? search:


b33ea1b5b5286335ac04de7c6f2178ab1a74e303
[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 "limits-util.h"
51 #include "load-fragment.h"
52 #include "log.h"
53 #include "loopback-setup.h"
54 #include "machine-id-setup.h"
55 #include "manager.h"
56 #include "missing.h"
57 #include "mount-setup.h"
58 #include "os-util.h"
59 #include "pager.h"
60 #include "parse-util.h"
61 #include "path-util.h"
62 #include "pretty-print.h"
63 #include "proc-cmdline.h"
64 #include "process-util.h"
65 #include "raw-clone.h"
66 #include "rlimit-util.h"
67 #if HAVE_SECCOMP
68 #include "seccomp-util.h"
69 #endif
70 #include "selinux-setup.h"
71 #include "selinux-util.h"
72 #include "signal-util.h"
73 #include "smack-setup.h"
74 #include "special.h"
75 #include "stat-util.h"
76 #include "stdio-util.h"
77 #include "strv.h"
78 #include "switch-root.h"
79 #include "sysctl-util.h"
80 #include "terminal-util.h"
81 #include "umask-util.h"
82 #include "user-util.h"
83 #include "util.h"
84 #include "virt.h"
85 #include "watchdog.h"
86
87 #if HAS_FEATURE_ADDRESS_SANITIZER
88 #include <sanitizer/lsan_interface.h>
89 #endif
90
91 static enum {
92 ACTION_RUN,
93 ACTION_HELP,
94 ACTION_VERSION,
95 ACTION_TEST,
96 ACTION_DUMP_CONFIGURATION_ITEMS,
97 ACTION_DUMP_BUS_PROPERTIES,
98 } arg_action = ACTION_RUN;
99
100 /* Those variables are initalized to 0 automatically, so we avoid uninitialized memory access.
101 * Real defaults are assigned in reset_arguments() below. */
102 static char *arg_default_unit;
103 static bool arg_system;
104 static bool arg_dump_core;
105 static int arg_crash_chvt;
106 static bool arg_crash_shell;
107 static bool arg_crash_reboot;
108 static char *arg_confirm_spawn;
109 static ShowStatus arg_show_status;
110 static bool arg_switched_root;
111 static PagerFlags arg_pager_flags;
112 static bool arg_service_watchdogs;
113 static ExecOutput arg_default_std_output;
114 static ExecOutput arg_default_std_error;
115 static usec_t arg_default_restart_usec;
116 static usec_t arg_default_timeout_start_usec;
117 static usec_t arg_default_timeout_stop_usec;
118 static usec_t arg_default_timeout_abort_usec;
119 static bool arg_default_timeout_abort_set;
120 static usec_t arg_default_start_limit_interval;
121 static unsigned arg_default_start_limit_burst;
122 static usec_t arg_runtime_watchdog;
123 static usec_t arg_shutdown_watchdog;
124 static char *arg_early_core_pattern;
125 static char *arg_watchdog_device;
126 static char **arg_default_environment;
127 static struct rlimit *arg_default_rlimit[_RLIMIT_MAX];
128 static uint64_t arg_capability_bounding_set;
129 static bool arg_no_new_privs;
130 static nsec_t arg_timer_slack_nsec;
131 static usec_t arg_default_timer_accuracy_usec;
132 static Set* arg_syscall_archs;
133 static FILE* arg_serialization;
134 static int arg_default_cpu_accounting;
135 static bool arg_default_io_accounting;
136 static bool arg_default_ip_accounting;
137 static bool arg_default_blockio_accounting;
138 static bool arg_default_memory_accounting;
139 static bool arg_default_tasks_accounting;
140 static uint64_t arg_default_tasks_max;
141 static sd_id128_t arg_machine_id;
142 static EmergencyAction arg_cad_burst_action;
143 static OOMPolicy arg_default_oom_policy;
144 static CPUSet arg_cpu_affinity;
145
146 static int parse_configuration(void);
147
148 _noreturn_ static void freeze_or_exit_or_reboot(void) {
149
150 /* If we are running in a container, let's prefer exiting, after all we can propagate an exit code to
151 * the container manager, and thus inform it that something went wrong. */
152 if (detect_container() > 0) {
153 log_emergency("Exiting PID 1...");
154 _exit(EXIT_EXCEPTION);
155 }
156
157 if (arg_crash_reboot) {
158 log_notice("Rebooting in 10s...");
159 (void) sleep(10);
160
161 log_notice("Rebooting now...");
162 (void) reboot(RB_AUTOBOOT);
163 log_emergency_errno(errno, "Failed to reboot: %m");
164 }
165
166 log_emergency("Freezing execution.");
167 freeze();
168 }
169
170 _noreturn_ static void crash(int sig) {
171 struct sigaction sa;
172 pid_t pid;
173
174 if (getpid_cached() != 1)
175 /* Pass this on immediately, if this is not PID 1 */
176 (void) raise(sig);
177 else if (!arg_dump_core)
178 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
179 else {
180 sa = (struct sigaction) {
181 .sa_handler = nop_signal_handler,
182 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
183 };
184
185 /* We want to wait for the core process, hence let's enable SIGCHLD */
186 (void) sigaction(SIGCHLD, &sa, NULL);
187
188 pid = raw_clone(SIGCHLD);
189 if (pid < 0)
190 log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
191 else if (pid == 0) {
192 /* Enable default signal handler for core dump */
193
194 sa = (struct sigaction) {
195 .sa_handler = SIG_DFL,
196 };
197 (void) sigaction(sig, &sa, NULL);
198
199 /* Don't limit the coredump size */
200 (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
201
202 /* Just to be sure... */
203 (void) chdir("/");
204
205 /* Raise the signal again */
206 pid = raw_getpid();
207 (void) kill(pid, sig); /* raise() would kill the parent */
208
209 assert_not_reached("We shouldn't be here...");
210 _exit(EXIT_EXCEPTION);
211 } else {
212 siginfo_t status;
213 int r;
214
215 /* Order things nicely. */
216 r = wait_for_terminate(pid, &status);
217 if (r < 0)
218 log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
219 else if (status.si_code != CLD_DUMPED)
220 log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
221 signal_to_string(sig),
222 pid, sigchld_code_to_string(status.si_code),
223 status.si_status,
224 strna(status.si_code == CLD_EXITED
225 ? exit_status_to_string(status.si_status, EXIT_STATUS_MINIMAL)
226 : signal_to_string(status.si_status)));
227 else
228 log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
229 }
230 }
231
232 if (arg_crash_chvt >= 0)
233 (void) chvt(arg_crash_chvt);
234
235 sa = (struct sigaction) {
236 .sa_handler = SIG_IGN,
237 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
238 };
239
240 /* Let the kernel reap children for us */
241 (void) sigaction(SIGCHLD, &sa, NULL);
242
243 if (arg_crash_shell) {
244 log_notice("Executing crash shell in 10s...");
245 (void) sleep(10);
246
247 pid = raw_clone(SIGCHLD);
248 if (pid < 0)
249 log_emergency_errno(errno, "Failed to fork off crash shell: %m");
250 else if (pid == 0) {
251 (void) setsid();
252 (void) make_console_stdio();
253 (void) rlimit_nofile_safe();
254 (void) execle("/bin/sh", "/bin/sh", NULL, environ);
255
256 log_emergency_errno(errno, "execle() failed: %m");
257 _exit(EXIT_EXCEPTION);
258 } else {
259 log_info("Spawned crash shell as PID "PID_FMT".", pid);
260 (void) wait_for_terminate(pid, NULL);
261 }
262 }
263
264 freeze_or_exit_or_reboot();
265 }
266
267 static void install_crash_handler(void) {
268 static const struct sigaction sa = {
269 .sa_handler = crash,
270 .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
271 };
272 int r;
273
274 /* We ignore the return value here, since, we don't mind if we
275 * cannot set up a crash handler */
276 r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
277 if (r < 0)
278 log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
279 }
280
281 static int console_setup(void) {
282 _cleanup_close_ int tty_fd = -1;
283 int r;
284
285 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
286 if (tty_fd < 0)
287 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
288
289 /* We don't want to force text mode. plymouth may be showing
290 * pictures already from initrd. */
291 r = reset_terminal_fd(tty_fd, false);
292 if (r < 0)
293 return log_error_errno(r, "Failed to reset /dev/console: %m");
294
295 return 0;
296 }
297
298 static int parse_crash_chvt(const char *value) {
299 int b;
300
301 if (safe_atoi(value, &arg_crash_chvt) >= 0)
302 return 0;
303
304 b = parse_boolean(value);
305 if (b < 0)
306 return b;
307
308 if (b > 0)
309 arg_crash_chvt = 0; /* switch to where kmsg goes */
310 else
311 arg_crash_chvt = -1; /* turn off switching */
312
313 return 0;
314 }
315
316 static int parse_confirm_spawn(const char *value, char **console) {
317 char *s;
318 int r;
319
320 r = value ? parse_boolean(value) : 1;
321 if (r == 0) {
322 *console = NULL;
323 return 0;
324 }
325
326 if (r > 0) /* on with default tty */
327 s = strdup("/dev/console");
328 else if (is_path(value)) /* on with fully qualified path */
329 s = strdup(value);
330 else /* on with only a tty file name, not a fully qualified path */
331 s = strjoin("/dev/", value);
332 if (!s)
333 return -ENOMEM;
334
335 *console = s;
336 return 0;
337 }
338
339 static int set_machine_id(const char *m) {
340 sd_id128_t t;
341 assert(m);
342
343 if (sd_id128_from_string(m, &t) < 0)
344 return -EINVAL;
345
346 if (sd_id128_is_null(t))
347 return -EINVAL;
348
349 arg_machine_id = t;
350 return 0;
351 }
352
353 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
354
355 int r;
356
357 assert(key);
358
359 if (STR_IN_SET(key, "systemd.unit", "rd.systemd.unit")) {
360
361 if (proc_cmdline_value_missing(key, value))
362 return 0;
363
364 if (!unit_name_is_valid(value, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE))
365 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key, value);
366 else if (in_initrd() == !!startswith(key, "rd.")) {
367 if (free_and_strdup(&arg_default_unit, value) < 0)
368 return log_oom();
369 }
370
371 } else if (proc_cmdline_key_streq(key, "systemd.dump_core")) {
372
373 r = value ? parse_boolean(value) : true;
374 if (r < 0)
375 log_warning_errno(r, "Failed to parse dump core switch %s, ignoring: %m", value);
376 else
377 arg_dump_core = r;
378
379 } else if (proc_cmdline_key_streq(key, "systemd.early_core_pattern")) {
380
381 if (proc_cmdline_value_missing(key, value))
382 return 0;
383
384 if (path_is_absolute(value))
385 (void) parse_path_argument_and_warn(value, false, &arg_early_core_pattern);
386 else
387 log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value);
388
389 } else if (proc_cmdline_key_streq(key, "systemd.crash_chvt")) {
390
391 if (!value)
392 arg_crash_chvt = 0; /* turn on */
393 else {
394 r = parse_crash_chvt(value);
395 if (r < 0)
396 log_warning_errno(r, "Failed to parse crash chvt switch %s, ignoring: %m", value);
397 }
398
399 } else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) {
400
401 r = value ? parse_boolean(value) : true;
402 if (r < 0)
403 log_warning_errno(r, "Failed to parse crash shell switch %s, ignoring: %m", value);
404 else
405 arg_crash_shell = r;
406
407 } else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) {
408
409 r = value ? parse_boolean(value) : true;
410 if (r < 0)
411 log_warning_errno(r, "Failed to parse crash reboot switch %s, ignoring: %m", value);
412 else
413 arg_crash_reboot = r;
414
415 } else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) {
416 char *s;
417
418 r = parse_confirm_spawn(value, &s);
419 if (r < 0)
420 log_warning_errno(r, "Failed to parse confirm_spawn switch %s, ignoring: %m", value);
421 else
422 free_and_replace(arg_confirm_spawn, s);
423
424 } else if (proc_cmdline_key_streq(key, "systemd.service_watchdogs")) {
425
426 r = value ? parse_boolean(value) : true;
427 if (r < 0)
428 log_warning_errno(r, "Failed to parse service watchdog switch %s, ignoring: %m", value);
429 else
430 arg_service_watchdogs = r;
431
432 } else if (proc_cmdline_key_streq(key, "systemd.show_status")) {
433
434 if (value) {
435 r = parse_show_status(value, &arg_show_status);
436 if (r < 0)
437 log_warning_errno(r, "Failed to parse show status switch %s, ignoring: %m", value);
438 } else
439 arg_show_status = SHOW_STATUS_YES;
440
441 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) {
442
443 if (proc_cmdline_value_missing(key, value))
444 return 0;
445
446 r = exec_output_from_string(value);
447 if (r < 0)
448 log_warning_errno(r, "Failed to parse default standard output switch %s, ignoring: %m", value);
449 else
450 arg_default_std_output = r;
451
452 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) {
453
454 if (proc_cmdline_value_missing(key, value))
455 return 0;
456
457 r = exec_output_from_string(value);
458 if (r < 0)
459 log_warning_errno(r, "Failed to parse default standard error switch %s, ignoring: %m", value);
460 else
461 arg_default_std_error = r;
462
463 } else if (streq(key, "systemd.setenv")) {
464
465 if (proc_cmdline_value_missing(key, value))
466 return 0;
467
468 if (env_assignment_is_valid(value)) {
469 char **env;
470
471 env = strv_env_set(arg_default_environment, value);
472 if (!env)
473 return log_oom();
474
475 arg_default_environment = env;
476 } else
477 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
478
479 } else if (proc_cmdline_key_streq(key, "systemd.machine_id")) {
480
481 if (proc_cmdline_value_missing(key, value))
482 return 0;
483
484 r = set_machine_id(value);
485 if (r < 0)
486 log_warning_errno(r, "MachineID '%s' is not valid, ignoring: %m", value);
487
488 } else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) {
489
490 if (proc_cmdline_value_missing(key, value))
491 return 0;
492
493 r = parse_sec(value, &arg_default_timeout_start_usec);
494 if (r < 0)
495 log_warning_errno(r, "Failed to parse default start timeout '%s', ignoring: %m", value);
496
497 if (arg_default_timeout_start_usec <= 0)
498 arg_default_timeout_start_usec = USEC_INFINITY;
499
500 } else if (proc_cmdline_key_streq(key, "systemd.watchdog_device")) {
501
502 if (proc_cmdline_value_missing(key, value))
503 return 0;
504
505 (void) parse_path_argument_and_warn(value, false, &arg_watchdog_device);
506
507 } else if (streq(key, "quiet") && !value) {
508
509 if (arg_show_status == _SHOW_STATUS_INVALID)
510 arg_show_status = SHOW_STATUS_AUTO;
511
512 } else if (streq(key, "debug") && !value) {
513
514 /* Note that log_parse_environment() handles 'debug'
515 * too, and sets the log level to LOG_DEBUG. */
516
517 if (detect_container() > 0)
518 log_set_target(LOG_TARGET_CONSOLE);
519
520 } else if (!value) {
521 const char *target;
522
523 /* SysV compatibility */
524 target = runlevel_to_target(key);
525 if (target)
526 return free_and_strdup(&arg_default_unit, target);
527 }
528
529 return 0;
530 }
531
532 #define DEFINE_SETTER(name, func, descr) \
533 static int name(const char *unit, \
534 const char *filename, \
535 unsigned line, \
536 const char *section, \
537 unsigned section_line, \
538 const char *lvalue, \
539 int ltype, \
540 const char *rvalue, \
541 void *data, \
542 void *userdata) { \
543 \
544 int r; \
545 \
546 assert(filename); \
547 assert(lvalue); \
548 assert(rvalue); \
549 \
550 r = func(rvalue); \
551 if (r < 0) \
552 log_syntax(unit, LOG_ERR, filename, line, r, \
553 "Invalid " descr "'%s': %m", \
554 rvalue); \
555 \
556 return 0; \
557 }
558
559 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level");
560 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target");
561 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" );
562 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location");
563
564 static int config_parse_cpu_affinity2(
565 const char *unit,
566 const char *filename,
567 unsigned line,
568 const char *section,
569 unsigned section_line,
570 const char *lvalue,
571 int ltype,
572 const char *rvalue,
573 void *data,
574 void *userdata) {
575
576 CPUSet *affinity = data;
577
578 assert(affinity);
579
580 (void) parse_cpu_set_extend(rvalue, affinity, true, unit, filename, line, lvalue);
581
582 return 0;
583 }
584
585 static int config_parse_show_status(
586 const char* unit,
587 const char *filename,
588 unsigned line,
589 const char *section,
590 unsigned section_line,
591 const char *lvalue,
592 int ltype,
593 const char *rvalue,
594 void *data,
595 void *userdata) {
596
597 int k;
598 ShowStatus *b = data;
599
600 assert(filename);
601 assert(lvalue);
602 assert(rvalue);
603 assert(data);
604
605 k = parse_show_status(rvalue, b);
606 if (k < 0) {
607 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
608 return 0;
609 }
610
611 return 0;
612 }
613
614 static int config_parse_output_restricted(
615 const char* unit,
616 const char *filename,
617 unsigned line,
618 const char *section,
619 unsigned section_line,
620 const char *lvalue,
621 int ltype,
622 const char *rvalue,
623 void *data,
624 void *userdata) {
625
626 ExecOutput t, *eo = data;
627
628 assert(filename);
629 assert(lvalue);
630 assert(rvalue);
631 assert(data);
632
633 t = exec_output_from_string(rvalue);
634 if (t < 0) {
635 log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse output type, ignoring: %s", rvalue);
636 return 0;
637 }
638
639 if (IN_SET(t, EXEC_OUTPUT_SOCKET, EXEC_OUTPUT_NAMED_FD, EXEC_OUTPUT_FILE, EXEC_OUTPUT_FILE_APPEND)) {
640 log_syntax(unit, LOG_ERR, filename, line, 0, "Standard output types socket, fd:, file:, append: are not supported as defaults, ignoring: %s", rvalue);
641 return 0;
642 }
643
644 *eo = t;
645 return 0;
646 }
647
648 static int config_parse_crash_chvt(
649 const char* unit,
650 const char *filename,
651 unsigned line,
652 const char *section,
653 unsigned section_line,
654 const char *lvalue,
655 int ltype,
656 const char *rvalue,
657 void *data,
658 void *userdata) {
659
660 int r;
661
662 assert(filename);
663 assert(lvalue);
664 assert(rvalue);
665
666 r = parse_crash_chvt(rvalue);
667 if (r < 0) {
668 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
669 return 0;
670 }
671
672 return 0;
673 }
674
675 static int config_parse_timeout_abort(
676 const char* unit,
677 const char *filename,
678 unsigned line,
679 const char *section,
680 unsigned section_line,
681 const char *lvalue,
682 int ltype,
683 const char *rvalue,
684 void *data,
685 void *userdata) {
686
687 int r;
688
689 assert(filename);
690 assert(lvalue);
691 assert(rvalue);
692
693 rvalue += strspn(rvalue, WHITESPACE);
694 if (isempty(rvalue)) {
695 arg_default_timeout_abort_set = false;
696 return 0;
697 }
698
699 r = parse_sec(rvalue, &arg_default_timeout_abort_usec);
700 if (r < 0) {
701 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse DefaultTimeoutAbortSec= setting, ignoring: %s", rvalue);
702 return 0;
703 }
704
705 arg_default_timeout_abort_set = true;
706 return 0;
707 }
708
709 static int parse_config_file(void) {
710
711 const ConfigTableItem items[] = {
712 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
713 { "Manager", "LogTarget", config_parse_target, 0, NULL },
714 { "Manager", "LogColor", config_parse_color, 0, NULL },
715 { "Manager", "LogLocation", config_parse_location, 0, NULL },
716 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
717 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL },
718 { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL },
719 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
720 { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
721 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
722 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, &arg_cpu_affinity },
723 { "Manager", "JoinControllers", config_parse_warn_compat, DISABLED_CONFIGURATION, NULL },
724 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
725 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
726 { "Manager", "WatchdogDevice", config_parse_path, 0, &arg_watchdog_device },
727 { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
728 { "Manager", "NoNewPrivileges", config_parse_bool, 0, &arg_no_new_privs },
729 #if HAVE_SECCOMP
730 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
731 #endif
732 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
733 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
734 { "Manager", "DefaultStandardOutput", config_parse_output_restricted,0, &arg_default_std_output },
735 { "Manager", "DefaultStandardError", config_parse_output_restricted,0, &arg_default_std_error },
736 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
737 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
738 { "Manager", "DefaultTimeoutAbortSec", config_parse_timeout_abort, 0, NULL },
739 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
740 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */
741 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec, 0, &arg_default_start_limit_interval },
742 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
743 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
744 { "Manager", "DefaultLimitCPU", config_parse_rlimit, RLIMIT_CPU, arg_default_rlimit },
745 { "Manager", "DefaultLimitFSIZE", config_parse_rlimit, RLIMIT_FSIZE, arg_default_rlimit },
746 { "Manager", "DefaultLimitDATA", config_parse_rlimit, RLIMIT_DATA, arg_default_rlimit },
747 { "Manager", "DefaultLimitSTACK", config_parse_rlimit, RLIMIT_STACK, arg_default_rlimit },
748 { "Manager", "DefaultLimitCORE", config_parse_rlimit, RLIMIT_CORE, arg_default_rlimit },
749 { "Manager", "DefaultLimitRSS", config_parse_rlimit, RLIMIT_RSS, arg_default_rlimit },
750 { "Manager", "DefaultLimitNOFILE", config_parse_rlimit, RLIMIT_NOFILE, arg_default_rlimit },
751 { "Manager", "DefaultLimitAS", config_parse_rlimit, RLIMIT_AS, arg_default_rlimit },
752 { "Manager", "DefaultLimitNPROC", config_parse_rlimit, RLIMIT_NPROC, arg_default_rlimit },
753 { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit, RLIMIT_MEMLOCK, arg_default_rlimit },
754 { "Manager", "DefaultLimitLOCKS", config_parse_rlimit, RLIMIT_LOCKS, arg_default_rlimit },
755 { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit, RLIMIT_SIGPENDING, arg_default_rlimit },
756 { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit, RLIMIT_MSGQUEUE, arg_default_rlimit },
757 { "Manager", "DefaultLimitNICE", config_parse_rlimit, RLIMIT_NICE, arg_default_rlimit },
758 { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit, RLIMIT_RTPRIO, arg_default_rlimit },
759 { "Manager", "DefaultLimitRTTIME", config_parse_rlimit, RLIMIT_RTTIME, arg_default_rlimit },
760 { "Manager", "DefaultCPUAccounting", config_parse_tristate, 0, &arg_default_cpu_accounting },
761 { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting },
762 { "Manager", "DefaultIPAccounting", config_parse_bool, 0, &arg_default_ip_accounting },
763 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
764 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
765 { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
766 { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max },
767 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action, 0, &arg_cad_burst_action },
768 { "Manager", "DefaultOOMPolicy", config_parse_oom_policy, 0, &arg_default_oom_policy },
769 {}
770 };
771
772 const char *fn, *conf_dirs_nulstr;
773
774 fn = arg_system ?
775 PKGSYSCONFDIR "/system.conf" :
776 PKGSYSCONFDIR "/user.conf";
777
778 conf_dirs_nulstr = arg_system ?
779 CONF_PATHS_NULSTR("systemd/system.conf.d") :
780 CONF_PATHS_NULSTR("systemd/user.conf.d");
781
782 (void) config_parse_many_nulstr(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, CONFIG_PARSE_WARN, NULL);
783
784 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
785 * like everywhere else. */
786 if (arg_default_timeout_start_usec <= 0)
787 arg_default_timeout_start_usec = USEC_INFINITY;
788 if (arg_default_timeout_stop_usec <= 0)
789 arg_default_timeout_stop_usec = USEC_INFINITY;
790
791 return 0;
792 }
793
794 static void set_manager_defaults(Manager *m) {
795
796 assert(m);
797
798 /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
799 * affect the manager itself, but are just what newly allocated units will have set if they haven't set
800 * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
801
802 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
803 m->default_std_output = arg_default_std_output;
804 m->default_std_error = arg_default_std_error;
805 m->default_timeout_start_usec = arg_default_timeout_start_usec;
806 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
807 m->default_timeout_abort_usec = arg_default_timeout_abort_usec;
808 m->default_timeout_abort_set = arg_default_timeout_abort_set;
809 m->default_restart_usec = arg_default_restart_usec;
810 m->default_start_limit_interval = arg_default_start_limit_interval;
811 m->default_start_limit_burst = arg_default_start_limit_burst;
812
813 /* On 4.15+ with unified hierarchy, CPU accounting is essentially free as it doesn't require the CPU
814 * controller to be enabled, so the default is to enable it unless we got told otherwise. */
815 if (arg_default_cpu_accounting >= 0)
816 m->default_cpu_accounting = arg_default_cpu_accounting;
817 else
818 m->default_cpu_accounting = cpu_accounting_is_cheap();
819
820 m->default_io_accounting = arg_default_io_accounting;
821 m->default_ip_accounting = arg_default_ip_accounting;
822 m->default_blockio_accounting = arg_default_blockio_accounting;
823 m->default_memory_accounting = arg_default_memory_accounting;
824 m->default_tasks_accounting = arg_default_tasks_accounting;
825 m->default_tasks_max = arg_default_tasks_max;
826 m->default_oom_policy = arg_default_oom_policy;
827
828 (void) manager_set_default_rlimits(m, arg_default_rlimit);
829
830 (void) manager_default_environment(m);
831 (void) manager_transient_environment_add(m, arg_default_environment);
832 }
833
834 static void set_manager_settings(Manager *m) {
835
836 assert(m);
837
838 /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
839 * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
840
841 m->confirm_spawn = arg_confirm_spawn;
842 m->service_watchdogs = arg_service_watchdogs;
843 m->runtime_watchdog = arg_runtime_watchdog;
844 m->shutdown_watchdog = arg_shutdown_watchdog;
845 m->cad_burst_action = arg_cad_burst_action;
846
847 manager_set_show_status(m, arg_show_status);
848 }
849
850 static int parse_argv(int argc, char *argv[]) {
851 enum {
852 ARG_LOG_LEVEL = 0x100,
853 ARG_LOG_TARGET,
854 ARG_LOG_COLOR,
855 ARG_LOG_LOCATION,
856 ARG_UNIT,
857 ARG_SYSTEM,
858 ARG_USER,
859 ARG_TEST,
860 ARG_NO_PAGER,
861 ARG_VERSION,
862 ARG_DUMP_CONFIGURATION_ITEMS,
863 ARG_DUMP_BUS_PROPERTIES,
864 ARG_DUMP_CORE,
865 ARG_CRASH_CHVT,
866 ARG_CRASH_SHELL,
867 ARG_CRASH_REBOOT,
868 ARG_CONFIRM_SPAWN,
869 ARG_SHOW_STATUS,
870 ARG_DESERIALIZE,
871 ARG_SWITCHED_ROOT,
872 ARG_DEFAULT_STD_OUTPUT,
873 ARG_DEFAULT_STD_ERROR,
874 ARG_MACHINE_ID,
875 ARG_SERVICE_WATCHDOGS,
876 };
877
878 static const struct option options[] = {
879 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
880 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
881 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
882 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
883 { "unit", required_argument, NULL, ARG_UNIT },
884 { "system", no_argument, NULL, ARG_SYSTEM },
885 { "user", no_argument, NULL, ARG_USER },
886 { "test", no_argument, NULL, ARG_TEST },
887 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
888 { "help", no_argument, NULL, 'h' },
889 { "version", no_argument, NULL, ARG_VERSION },
890 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
891 { "dump-bus-properties", no_argument, NULL, ARG_DUMP_BUS_PROPERTIES },
892 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
893 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
894 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
895 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
896 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
897 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
898 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
899 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
900 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
901 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
902 { "machine-id", required_argument, NULL, ARG_MACHINE_ID },
903 { "service-watchdogs", required_argument, NULL, ARG_SERVICE_WATCHDOGS },
904 {}
905 };
906
907 int c, r;
908
909 assert(argc >= 1);
910 assert(argv);
911
912 if (getpid_cached() == 1)
913 opterr = 0;
914
915 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
916
917 switch (c) {
918
919 case ARG_LOG_LEVEL:
920 r = log_set_max_level_from_string(optarg);
921 if (r < 0)
922 return log_error_errno(r, "Failed to parse log level \"%s\": %m", optarg);
923
924 break;
925
926 case ARG_LOG_TARGET:
927 r = log_set_target_from_string(optarg);
928 if (r < 0)
929 return log_error_errno(r, "Failed to parse log target \"%s\": %m", optarg);
930
931 break;
932
933 case ARG_LOG_COLOR:
934
935 if (optarg) {
936 r = log_show_color_from_string(optarg);
937 if (r < 0)
938 return log_error_errno(r, "Failed to parse log color setting \"%s\": %m",
939 optarg);
940 } else
941 log_show_color(true);
942
943 break;
944
945 case ARG_LOG_LOCATION:
946 if (optarg) {
947 r = log_show_location_from_string(optarg);
948 if (r < 0)
949 return log_error_errno(r, "Failed to parse log location setting \"%s\": %m",
950 optarg);
951 } else
952 log_show_location(true);
953
954 break;
955
956 case ARG_DEFAULT_STD_OUTPUT:
957 r = exec_output_from_string(optarg);
958 if (r < 0)
959 return log_error_errno(r, "Failed to parse default standard output setting \"%s\": %m",
960 optarg);
961 arg_default_std_output = r;
962 break;
963
964 case ARG_DEFAULT_STD_ERROR:
965 r = exec_output_from_string(optarg);
966 if (r < 0)
967 return log_error_errno(r, "Failed to parse default standard error output setting \"%s\": %m",
968 optarg);
969 arg_default_std_error = r;
970 break;
971
972 case ARG_UNIT:
973 r = free_and_strdup(&arg_default_unit, optarg);
974 if (r < 0)
975 return log_error_errno(r, "Failed to set default unit \"%s\": %m", optarg);
976
977 break;
978
979 case ARG_SYSTEM:
980 arg_system = true;
981 break;
982
983 case ARG_USER:
984 arg_system = false;
985 break;
986
987 case ARG_TEST:
988 arg_action = ACTION_TEST;
989 break;
990
991 case ARG_NO_PAGER:
992 arg_pager_flags |= PAGER_DISABLE;
993 break;
994
995 case ARG_VERSION:
996 arg_action = ACTION_VERSION;
997 break;
998
999 case ARG_DUMP_CONFIGURATION_ITEMS:
1000 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
1001 break;
1002
1003 case ARG_DUMP_BUS_PROPERTIES:
1004 arg_action = ACTION_DUMP_BUS_PROPERTIES;
1005 break;
1006
1007 case ARG_DUMP_CORE:
1008 if (!optarg)
1009 arg_dump_core = true;
1010 else {
1011 r = parse_boolean(optarg);
1012 if (r < 0)
1013 return log_error_errno(r, "Failed to parse dump core boolean: \"%s\": %m",
1014 optarg);
1015 arg_dump_core = r;
1016 }
1017 break;
1018
1019 case ARG_CRASH_CHVT:
1020 r = parse_crash_chvt(optarg);
1021 if (r < 0)
1022 return log_error_errno(r, "Failed to parse crash virtual terminal index: \"%s\": %m",
1023 optarg);
1024 break;
1025
1026 case ARG_CRASH_SHELL:
1027 if (!optarg)
1028 arg_crash_shell = true;
1029 else {
1030 r = parse_boolean(optarg);
1031 if (r < 0)
1032 return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
1033 optarg);
1034 arg_crash_shell = r;
1035 }
1036 break;
1037
1038 case ARG_CRASH_REBOOT:
1039 if (!optarg)
1040 arg_crash_reboot = true;
1041 else {
1042 r = parse_boolean(optarg);
1043 if (r < 0)
1044 return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
1045 optarg);
1046 arg_crash_reboot = r;
1047 }
1048 break;
1049
1050 case ARG_CONFIRM_SPAWN:
1051 arg_confirm_spawn = mfree(arg_confirm_spawn);
1052
1053 r = parse_confirm_spawn(optarg, &arg_confirm_spawn);
1054 if (r < 0)
1055 return log_error_errno(r, "Failed to parse confirm spawn option: \"%s\": %m",
1056 optarg);
1057 break;
1058
1059 case ARG_SERVICE_WATCHDOGS:
1060 r = parse_boolean(optarg);
1061 if (r < 0)
1062 return log_error_errno(r, "Failed to parse service watchdogs boolean: \"%s\": %m",
1063 optarg);
1064 arg_service_watchdogs = r;
1065 break;
1066
1067 case ARG_SHOW_STATUS:
1068 if (optarg) {
1069 r = parse_show_status(optarg, &arg_show_status);
1070 if (r < 0)
1071 return log_error_errno(r, "Failed to parse show status boolean: \"%s\": %m",
1072 optarg);
1073 } else
1074 arg_show_status = SHOW_STATUS_YES;
1075 break;
1076
1077 case ARG_DESERIALIZE: {
1078 int fd;
1079 FILE *f;
1080
1081 r = safe_atoi(optarg, &fd);
1082 if (r < 0)
1083 log_error_errno(r, "Failed to parse deserialize option \"%s\": %m", optarg);
1084 if (fd < 0)
1085 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1086 "Invalid deserialize fd: %d",
1087 fd);
1088
1089 (void) fd_cloexec(fd, true);
1090
1091 f = fdopen(fd, "r");
1092 if (!f)
1093 return log_error_errno(errno, "Failed to open serialization fd %d: %m", fd);
1094
1095 safe_fclose(arg_serialization);
1096 arg_serialization = f;
1097
1098 break;
1099 }
1100
1101 case ARG_SWITCHED_ROOT:
1102 arg_switched_root = true;
1103 break;
1104
1105 case ARG_MACHINE_ID:
1106 r = set_machine_id(optarg);
1107 if (r < 0)
1108 return log_error_errno(r, "MachineID '%s' is not valid: %m", optarg);
1109 break;
1110
1111 case 'h':
1112 arg_action = ACTION_HELP;
1113 break;
1114
1115 case 'D':
1116 log_set_max_level(LOG_DEBUG);
1117 break;
1118
1119 case 'b':
1120 case 's':
1121 case 'z':
1122 /* Just to eat away the sysvinit kernel
1123 * cmdline args without getopt() error
1124 * messages that we'll parse in
1125 * parse_proc_cmdline_word() or ignore. */
1126
1127 case '?':
1128 if (getpid_cached() != 1)
1129 return -EINVAL;
1130 else
1131 return 0;
1132
1133 default:
1134 assert_not_reached("Unhandled option code.");
1135 }
1136
1137 if (optind < argc && getpid_cached() != 1) {
1138 /* Hmm, when we aren't run as init system
1139 * let's complain about excess arguments */
1140
1141 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1142 "Excess arguments.");
1143 }
1144
1145 return 0;
1146 }
1147
1148 static int help(void) {
1149 _cleanup_free_ char *link = NULL;
1150 int r;
1151
1152 r = terminal_urlify_man("systemd", "1", &link);
1153 if (r < 0)
1154 return log_oom();
1155
1156 printf("%s [OPTIONS...]\n\n"
1157 "Starts up and maintains the system or user services.\n\n"
1158 " -h --help Show this help\n"
1159 " --version Show version\n"
1160 " --test Determine startup sequence, dump it and exit\n"
1161 " --no-pager Do not pipe output into a pager\n"
1162 " --dump-configuration-items Dump understood unit configuration items\n"
1163 " --dump-bus-properties Dump exposed bus properties\n"
1164 " --unit=UNIT Set default unit\n"
1165 " --system Run a system instance, even if PID != 1\n"
1166 " --user Run a user instance\n"
1167 " --dump-core[=BOOL] Dump core on crash\n"
1168 " --crash-vt=NR Change to specified VT on crash\n"
1169 " --crash-reboot[=BOOL] Reboot on crash\n"
1170 " --crash-shell[=BOOL] Run shell on crash\n"
1171 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1172 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1173 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1174 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1175 " --log-color[=BOOL] Highlight important log messages\n"
1176 " --log-location[=BOOL] Include code location in log messages\n"
1177 " --default-standard-output= Set default standard output for services\n"
1178 " --default-standard-error= Set default standard error output for services\n"
1179 "\nSee the %s for details.\n"
1180 , program_invocation_short_name
1181 , link
1182 );
1183
1184 return 0;
1185 }
1186
1187 static int prepare_reexecute(
1188 Manager *m,
1189 FILE **ret_f,
1190 FDSet **ret_fds,
1191 bool switching_root) {
1192
1193 _cleanup_fdset_free_ FDSet *fds = NULL;
1194 _cleanup_fclose_ FILE *f = NULL;
1195 int r;
1196
1197 assert(m);
1198 assert(ret_f);
1199 assert(ret_fds);
1200
1201 r = manager_open_serialization(m, &f);
1202 if (r < 0)
1203 return log_error_errno(r, "Failed to create serialization file: %m");
1204
1205 /* Make sure nothing is really destructed when we shut down */
1206 m->n_reloading++;
1207 bus_manager_send_reloading(m, true);
1208
1209 fds = fdset_new();
1210 if (!fds)
1211 return log_oom();
1212
1213 r = manager_serialize(m, f, fds, switching_root);
1214 if (r < 0)
1215 return r;
1216
1217 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1218 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1219
1220 r = fd_cloexec(fileno(f), false);
1221 if (r < 0)
1222 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1223
1224 r = fdset_cloexec(fds, false);
1225 if (r < 0)
1226 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1227
1228 *ret_f = TAKE_PTR(f);
1229 *ret_fds = TAKE_PTR(fds);
1230
1231 return 0;
1232 }
1233
1234 static void bump_file_max_and_nr_open(void) {
1235
1236 /* Let's bump fs.file-max and fs.nr_open to their respective maximums. On current kernels large numbers of file
1237 * descriptors are no longer a performance problem and their memory is properly tracked by memcg, thus counting
1238 * them and limiting them in another two layers of limits is unnecessary and just complicates things. This
1239 * function hence turns off 2 of the 4 levels of limits on file descriptors, and makes RLIMIT_NOLIMIT (soft +
1240 * hard) the only ones that really matter. */
1241
1242 #if BUMP_PROC_SYS_FS_FILE_MAX || BUMP_PROC_SYS_FS_NR_OPEN
1243 _cleanup_free_ char *t = NULL;
1244 int r;
1245 #endif
1246
1247 #if BUMP_PROC_SYS_FS_FILE_MAX
1248 /* I so wanted to use STRINGIFY(ULONG_MAX) here, but alas we can't as glibc/gcc define that as
1249 * "(0x7fffffffffffffffL * 2UL + 1UL)". Seriously. 😢 */
1250 if (asprintf(&t, "%lu\n", ULONG_MAX) < 0) {
1251 log_oom();
1252 return;
1253 }
1254
1255 r = sysctl_write("fs/file-max", t);
1256 if (r < 0)
1257 log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.file-max, ignoring: %m");
1258 #endif
1259
1260 #if BUMP_PROC_SYS_FS_FILE_MAX && BUMP_PROC_SYS_FS_NR_OPEN
1261 t = mfree(t);
1262 #endif
1263
1264 #if BUMP_PROC_SYS_FS_NR_OPEN
1265 int v = INT_MAX;
1266
1267 /* Arg! The kernel enforces maximum and minimum values on the fs.nr_open, but we don't really know what they
1268 * are. The expression by which the maximum is determined is dependent on the architecture, and is something we
1269 * don't really want to copy to userspace, as it is dependent on implementation details of the kernel. Since
1270 * the kernel doesn't expose the maximum value to us, we can only try and hope. Hence, let's start with
1271 * INT_MAX, and then keep halving the value until we find one that works. Ugly? Yes, absolutely, but kernel
1272 * APIs are kernel APIs, so what do can we do... 🤯 */
1273
1274 for (;;) {
1275 int k;
1276
1277 v &= ~(__SIZEOF_POINTER__ - 1); /* Round down to next multiple of the pointer size */
1278 if (v < 1024) {
1279 log_warning("Can't bump fs.nr_open, value too small.");
1280 break;
1281 }
1282
1283 k = read_nr_open();
1284 if (k < 0) {
1285 log_error_errno(k, "Failed to read fs.nr_open: %m");
1286 break;
1287 }
1288 if (k >= v) { /* Already larger */
1289 log_debug("Skipping bump, value is already larger.");
1290 break;
1291 }
1292
1293 if (asprintf(&t, "%i\n", v) < 0) {
1294 log_oom();
1295 return;
1296 }
1297
1298 r = sysctl_write("fs/nr_open", t);
1299 t = mfree(t);
1300 if (r == -EINVAL) {
1301 log_debug("Couldn't write fs.nr_open as %i, halving it.", v);
1302 v /= 2;
1303 continue;
1304 }
1305 if (r < 0) {
1306 log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.nr_open, ignoring: %m");
1307 break;
1308 }
1309
1310 log_debug("Successfully bumped fs.nr_open to %i", v);
1311 break;
1312 }
1313 #endif
1314 }
1315
1316 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1317 struct rlimit new_rlimit;
1318 int r, nr;
1319
1320 assert(saved_rlimit);
1321
1322 /* Save the original RLIMIT_NOFILE so that we can reset it later when transitioning from the initrd to the main
1323 * systemd or suchlike. */
1324 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1325 return log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1326
1327 /* Get the underlying absolute limit the kernel enforces */
1328 nr = read_nr_open();
1329
1330 /* Make sure forked processes get limits based on the original kernel setting */
1331 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1332 struct rlimit *rl;
1333
1334 rl = newdup(struct rlimit, saved_rlimit, 1);
1335 if (!rl)
1336 return log_oom();
1337
1338 /* Bump the hard limit for system services to a substantially higher value. The default hard limit
1339 * current kernels set is pretty low (4K), mostly for historical reasons. According to kernel
1340 * developers, the fd handling in recent kernels has been optimized substantially enough, so that we
1341 * can bump the limit now, without paying too high a price in memory or performance. Note however that
1342 * we only bump the hard limit, not the soft limit. That's because select() works the way it works, and
1343 * chokes on fds >= 1024. If we'd bump the soft limit globally, it might accidentally happen to
1344 * unexpecting programs that they get fds higher than what they can process using select(). By only
1345 * bumping the hard limit but leaving the low limit as it is we avoid this pitfall: programs that are
1346 * written by folks aware of the select() problem in mind (and thus use poll()/epoll instead of
1347 * select(), the way everybody should) can explicitly opt into high fds by bumping their soft limit
1348 * beyond 1024, to the hard limit we pass. */
1349 if (arg_system)
1350 rl->rlim_max = MIN((rlim_t) nr, MAX(rl->rlim_max, (rlim_t) HIGH_RLIMIT_NOFILE));
1351
1352 /* If for some reason we were invoked with a soft limit above 1024 (which should never
1353 * happen!, but who knows what we get passed in from pam_limit when invoked as --user
1354 * instance), then lower what we pass on to not confuse our children */
1355 rl->rlim_cur = MIN(rl->rlim_cur, (rlim_t) FD_SETSIZE);
1356
1357 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1358 }
1359
1360 /* Calculate the new limits to use for us. Never lower from what we inherited. */
1361 new_rlimit = (struct rlimit) {
1362 .rlim_cur = MAX((rlim_t) nr, saved_rlimit->rlim_cur),
1363 .rlim_max = MAX((rlim_t) nr, saved_rlimit->rlim_max),
1364 };
1365
1366 /* Shortcut if nothing changes. */
1367 if (saved_rlimit->rlim_max >= new_rlimit.rlim_max &&
1368 saved_rlimit->rlim_cur >= new_rlimit.rlim_cur) {
1369 log_debug("RLIMIT_NOFILE is already as high or higher than we need it, not bumping.");
1370 return 0;
1371 }
1372
1373 /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for
1374 * both hard and soft. */
1375 r = setrlimit_closest(RLIMIT_NOFILE, &new_rlimit);
1376 if (r < 0)
1377 return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1378
1379 return 0;
1380 }
1381
1382 static int bump_rlimit_memlock(struct rlimit *saved_rlimit) {
1383 struct rlimit new_rlimit;
1384 int r;
1385
1386 assert(saved_rlimit);
1387
1388 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK which should
1389 * normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's bump
1390 * the value high enough for our user. */
1391
1392 if (getrlimit(RLIMIT_MEMLOCK, saved_rlimit) < 0)
1393 return log_warning_errno(errno, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1394
1395 /* Pass the original value down to invoked processes */
1396 if (!arg_default_rlimit[RLIMIT_MEMLOCK]) {
1397 struct rlimit *rl;
1398
1399 rl = newdup(struct rlimit, saved_rlimit, 1);
1400 if (!rl)
1401 return log_oom();
1402
1403 arg_default_rlimit[RLIMIT_MEMLOCK] = rl;
1404 }
1405
1406 /* Using MAX() on resource limits only is safe if RLIM_INFINITY is > 0. POSIX declares that rlim_t
1407 * must be unsigned, hence this is a given, but let's make this clear here. */
1408 assert_cc(RLIM_INFINITY > 0);
1409
1410 new_rlimit = (struct rlimit) {
1411 .rlim_cur = MAX(HIGH_RLIMIT_MEMLOCK, saved_rlimit->rlim_cur),
1412 .rlim_max = MAX(HIGH_RLIMIT_MEMLOCK, saved_rlimit->rlim_max),
1413 };
1414
1415 if (saved_rlimit->rlim_max >= new_rlimit.rlim_cur &&
1416 saved_rlimit->rlim_cur >= new_rlimit.rlim_max) {
1417 log_debug("RLIMIT_MEMLOCK is already as high or higher than we need it, not bumping.");
1418 return 0;
1419 }
1420
1421 r = setrlimit_closest(RLIMIT_MEMLOCK, &new_rlimit);
1422 if (r < 0)
1423 return log_warning_errno(r, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1424
1425 return 0;
1426 }
1427
1428 static void test_usr(void) {
1429
1430 /* Check that /usr is either on the same file system as / or mounted already. */
1431
1432 if (dir_is_empty("/usr") <= 0)
1433 return;
1434
1435 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1436 "Some things will probably break (sometimes even silently) in mysterious ways. "
1437 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1438 }
1439
1440 static int enforce_syscall_archs(Set *archs) {
1441 #if HAVE_SECCOMP
1442 int r;
1443
1444 if (!is_seccomp_available())
1445 return 0;
1446
1447 r = seccomp_restrict_archs(arg_syscall_archs);
1448 if (r < 0)
1449 return log_error_errno(r, "Failed to enforce system call architecture restrication: %m");
1450 #endif
1451 return 0;
1452 }
1453
1454 static int status_welcome(void) {
1455 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1456 int r;
1457
1458 if (IN_SET(arg_show_status, SHOW_STATUS_NO, SHOW_STATUS_AUTO))
1459 return 0;
1460
1461 r = parse_os_release(NULL,
1462 "PRETTY_NAME", &pretty_name,
1463 "ANSI_COLOR", &ansi_color,
1464 NULL);
1465 if (r < 0)
1466 log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r,
1467 "Failed to read os-release file, ignoring: %m");
1468
1469 if (log_get_show_color())
1470 return status_printf(NULL, 0,
1471 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1472 isempty(ansi_color) ? "1" : ansi_color,
1473 isempty(pretty_name) ? "Linux" : pretty_name);
1474 else
1475 return status_printf(NULL, 0,
1476 "\nWelcome to %s!\n",
1477 isempty(pretty_name) ? "Linux" : pretty_name);
1478 }
1479
1480 static int write_container_id(void) {
1481 const char *c;
1482 int r;
1483
1484 c = getenv("container");
1485 if (isempty(c))
1486 return 0;
1487
1488 RUN_WITH_UMASK(0022)
1489 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1490 if (r < 0)
1491 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
1492
1493 return 1;
1494 }
1495
1496 static int bump_unix_max_dgram_qlen(void) {
1497 _cleanup_free_ char *qlen = NULL;
1498 unsigned long v;
1499 int r;
1500
1501 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1502 * really really early during boot, so that it is actually applied to all our sockets, including the
1503 * $NOTIFY_SOCKET one. */
1504
1505 r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1506 if (r < 0)
1507 return log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1508
1509 r = safe_atolu(qlen, &v);
1510 if (r < 0)
1511 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen);
1512
1513 if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1514 return 0;
1515
1516 r = write_string_filef("/proc/sys/net/unix/max_dgram_qlen", WRITE_STRING_FILE_DISABLE_BUFFER, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN);
1517 if (r < 0)
1518 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1519 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1520
1521 return 1;
1522 }
1523
1524 static int fixup_environment(void) {
1525 _cleanup_free_ char *term = NULL;
1526 const char *t;
1527 int r;
1528
1529 /* Only fix up the environment when we are started as PID 1 */
1530 if (getpid_cached() != 1)
1531 return 0;
1532
1533 /* We expect the environment to be set correctly if run inside a container. */
1534 if (detect_container() > 0)
1535 return 0;
1536
1537 /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1538 * device used by the console. We try to make a better guess here since some consoles might not have support
1539 * for color mode for example.
1540 *
1541 * However if TERM was configured through the kernel command line then leave it alone. */
1542 r = proc_cmdline_get_key("TERM", 0, &term);
1543 if (r < 0)
1544 return r;
1545
1546 t = term ?: default_term_for_tty("/dev/console");
1547
1548 if (setenv("TERM", t, 1) < 0)
1549 return -errno;
1550
1551 /* The kernels sets HOME=/ for init. Let's undo this. */
1552 if (path_equal_ptr(getenv("HOME"), "/") &&
1553 unsetenv("HOME") < 0)
1554 log_warning_errno(errno, "Failed to unset $HOME: %m");
1555
1556 return 0;
1557 }
1558
1559 static void redirect_telinit(int argc, char *argv[]) {
1560
1561 /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1562
1563 #if HAVE_SYSV_COMPAT
1564 if (getpid_cached() == 1)
1565 return;
1566
1567 if (!strstr(program_invocation_short_name, "init"))
1568 return;
1569
1570 execv(SYSTEMCTL_BINARY_PATH, argv);
1571 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1572 exit(EXIT_FAILURE);
1573 #endif
1574 }
1575
1576 static int become_shutdown(
1577 const char *shutdown_verb,
1578 int retval) {
1579
1580 char log_level[DECIMAL_STR_MAX(int) + 1],
1581 exit_code[DECIMAL_STR_MAX(uint8_t) + 1],
1582 timeout[DECIMAL_STR_MAX(usec_t) + 1];
1583
1584 const char* command_line[13] = {
1585 SYSTEMD_SHUTDOWN_BINARY_PATH,
1586 shutdown_verb,
1587 "--timeout", timeout,
1588 "--log-level", log_level,
1589 "--log-target",
1590 };
1591
1592 _cleanup_strv_free_ char **env_block = NULL;
1593 size_t pos = 7;
1594 int r;
1595
1596 assert(shutdown_verb);
1597 assert(!command_line[pos]);
1598 env_block = strv_copy(environ);
1599
1600 xsprintf(log_level, "%d", log_get_max_level());
1601 xsprintf(timeout, "%" PRI_USEC "us", arg_default_timeout_stop_usec);
1602
1603 switch (log_get_target()) {
1604
1605 case LOG_TARGET_KMSG:
1606 case LOG_TARGET_JOURNAL_OR_KMSG:
1607 case LOG_TARGET_SYSLOG_OR_KMSG:
1608 command_line[pos++] = "kmsg";
1609 break;
1610
1611 case LOG_TARGET_NULL:
1612 command_line[pos++] = "null";
1613 break;
1614
1615 case LOG_TARGET_CONSOLE:
1616 default:
1617 command_line[pos++] = "console";
1618 break;
1619 };
1620
1621 if (log_get_show_color())
1622 command_line[pos++] = "--log-color";
1623
1624 if (log_get_show_location())
1625 command_line[pos++] = "--log-location";
1626
1627 if (streq(shutdown_verb, "exit")) {
1628 command_line[pos++] = "--exit-code";
1629 command_line[pos++] = exit_code;
1630 xsprintf(exit_code, "%d", retval);
1631 }
1632
1633 assert(pos < ELEMENTSOF(command_line));
1634
1635 if (streq(shutdown_verb, "reboot") &&
1636 arg_shutdown_watchdog > 0 &&
1637 arg_shutdown_watchdog != USEC_INFINITY) {
1638
1639 char *e;
1640
1641 /* If we reboot let's set the shutdown
1642 * watchdog and tell the shutdown binary to
1643 * repeatedly ping it */
1644 r = watchdog_set_timeout(&arg_shutdown_watchdog);
1645 watchdog_close(r < 0);
1646
1647 /* Tell the binary how often to ping, ignore failure */
1648 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
1649 (void) strv_consume(&env_block, e);
1650
1651 if (arg_watchdog_device &&
1652 asprintf(&e, "WATCHDOG_DEVICE=%s", arg_watchdog_device) > 0)
1653 (void) strv_consume(&env_block, e);
1654 } else
1655 watchdog_close(true);
1656
1657 /* Avoid the creation of new processes forked by the
1658 * kernel; at this point, we will not listen to the
1659 * signals anyway */
1660 if (detect_container() <= 0)
1661 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
1662
1663 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
1664 return -errno;
1665 }
1666
1667 static void initialize_clock(void) {
1668 int r;
1669
1670 if (clock_is_localtime(NULL) > 0) {
1671 int min;
1672
1673 /*
1674 * The very first call of settimeofday() also does a time warp in the kernel.
1675 *
1676 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1677 * of maintaining the RTC and do all adjustments. This matches the behavior of Windows, which leaves
1678 * the RTC alone if the registry tells that the RTC runs in UTC.
1679 */
1680 r = clock_set_timezone(&min);
1681 if (r < 0)
1682 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1683 else
1684 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1685
1686 } else if (!in_initrd()) {
1687 /*
1688 * Do a dummy very first call to seal the kernel's time warp magic.
1689 *
1690 * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1691 * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1692 * until we reach the real system.
1693 *
1694 * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1695 * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1696 * be treated as UTC that way.
1697 */
1698 (void) clock_reset_timewarp();
1699 }
1700
1701 r = clock_apply_epoch();
1702 if (r < 0)
1703 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1704 else if (r > 0)
1705 log_info("System time before build time, advancing clock.");
1706 }
1707
1708 static void initialize_coredump(bool skip_setup) {
1709 #if ENABLE_COREDUMP
1710 if (getpid_cached() != 1)
1711 return;
1712
1713 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1714 * will process core dumps for system services by default. */
1715 if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0)
1716 log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m");
1717
1718 /* But at the same time, turn off the core_pattern logic by default, so that no
1719 * coredumps are stored until the systemd-coredump tool is enabled via
1720 * sysctl. However it can be changed via the kernel command line later so core
1721 * dumps can still be generated during early startup and in initramfs. */
1722 if (!skip_setup)
1723 disable_coredumps();
1724 #endif
1725 }
1726
1727 static void initialize_core_pattern(bool skip_setup) {
1728 int r;
1729
1730 if (skip_setup || !arg_early_core_pattern)
1731 return;
1732
1733 if (getpid_cached() != 1)
1734 return;
1735
1736 r = write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern, WRITE_STRING_FILE_DISABLE_BUFFER);
1737 if (r < 0)
1738 log_warning_errno(r, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", arg_early_core_pattern);
1739 }
1740
1741 static void update_cpu_affinity(bool skip_setup) {
1742 _cleanup_free_ char *mask = NULL;
1743
1744 if (skip_setup || !arg_cpu_affinity.set)
1745 return;
1746
1747 assert(arg_cpu_affinity.allocated > 0);
1748
1749 mask = cpu_set_to_string(&arg_cpu_affinity);
1750 log_debug("Setting CPU affinity to %s.", strnull(mask));
1751
1752 if (sched_setaffinity(0, arg_cpu_affinity.allocated, arg_cpu_affinity.set) < 0)
1753 log_warning_errno(errno, "Failed to set CPU affinity: %m");
1754 }
1755
1756 static void do_reexecute(
1757 int argc,
1758 char *argv[],
1759 const struct rlimit *saved_rlimit_nofile,
1760 const struct rlimit *saved_rlimit_memlock,
1761 FDSet *fds,
1762 const char *switch_root_dir,
1763 const char *switch_root_init,
1764 const char **ret_error_message) {
1765
1766 unsigned i, j, args_size;
1767 const char **args;
1768 int r;
1769
1770 assert(saved_rlimit_nofile);
1771 assert(saved_rlimit_memlock);
1772 assert(ret_error_message);
1773
1774 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1775 * we do that */
1776 watchdog_close(true);
1777
1778 /* Reset RLIMIT_NOFILE + RLIMIT_MEMLOCK back to the kernel defaults, so that the new systemd can pass
1779 * the kernel default to its child processes */
1780 if (saved_rlimit_nofile->rlim_cur != 0)
1781 (void) setrlimit(RLIMIT_NOFILE, saved_rlimit_nofile);
1782 if (saved_rlimit_memlock->rlim_cur != RLIM_INFINITY)
1783 (void) setrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock);
1784
1785 if (switch_root_dir) {
1786 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1787 * SIGCHLD for them after deserializing. */
1788 broadcast_signal(SIGTERM, false, true, arg_default_timeout_stop_usec);
1789
1790 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1791 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1792 if (r < 0)
1793 log_error_errno(r, "Failed to switch root, trying to continue: %m");
1794 }
1795
1796 args_size = MAX(6, argc+1);
1797 args = newa(const char*, args_size);
1798
1799 if (!switch_root_init) {
1800 char sfd[DECIMAL_STR_MAX(int) + 1];
1801
1802 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1803 * the user didn't specify an explicit init to spawn. */
1804
1805 assert(arg_serialization);
1806 assert(fds);
1807
1808 xsprintf(sfd, "%i", fileno(arg_serialization));
1809
1810 i = 0;
1811 args[i++] = SYSTEMD_BINARY_PATH;
1812 if (switch_root_dir)
1813 args[i++] = "--switched-root";
1814 args[i++] = arg_system ? "--system" : "--user";
1815 args[i++] = "--deserialize";
1816 args[i++] = sfd;
1817 args[i++] = NULL;
1818
1819 assert(i <= args_size);
1820
1821 /*
1822 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do this is on
1823 * its own on exec(), but it will do it on exit(). Hence, to ensure we get a summary here, fork() off
1824 * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1825 * before proceeding into the exec().
1826 */
1827 valgrind_summary_hack();
1828
1829 (void) execv(args[0], (char* const*) args);
1830 log_debug_errno(errno, "Failed to execute our own binary, trying fallback: %m");
1831 }
1832
1833 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1834 * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1835 * doesn't matter.) */
1836
1837 arg_serialization = safe_fclose(arg_serialization);
1838 fds = fdset_free(fds);
1839
1840 /* Reopen the console */
1841 (void) make_console_stdio();
1842
1843 for (j = 1, i = 1; j < (unsigned) argc; j++)
1844 args[i++] = argv[j];
1845 args[i++] = NULL;
1846 assert(i <= args_size);
1847
1848 /* Re-enable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1849 (void) reset_all_signal_handlers();
1850 (void) reset_signal_mask();
1851 (void) rlimit_nofile_safe();
1852
1853 if (switch_root_init) {
1854 args[0] = switch_root_init;
1855 (void) execv(args[0], (char* const*) args);
1856 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
1857 }
1858
1859 args[0] = "/sbin/init";
1860 (void) execv(args[0], (char* const*) args);
1861 r = -errno;
1862
1863 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
1864 ANSI_HIGHLIGHT_RED " !! " ANSI_NORMAL,
1865 "Failed to execute /sbin/init");
1866
1867 if (r == -ENOENT) {
1868 log_warning("No /sbin/init, trying fallback");
1869
1870 args[0] = "/bin/sh";
1871 args[1] = NULL;
1872 (void) execv(args[0], (char* const*) args);
1873 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
1874 } else
1875 log_warning_errno(r, "Failed to execute /sbin/init, giving up: %m");
1876
1877 *ret_error_message = "Failed to execute fallback shell";
1878 }
1879
1880 static int invoke_main_loop(
1881 Manager *m,
1882 bool *ret_reexecute,
1883 int *ret_retval, /* Return parameters relevant for shutting down */
1884 const char **ret_shutdown_verb, /* … */
1885 FDSet **ret_fds, /* Return parameters for reexecuting */
1886 char **ret_switch_root_dir, /* … */
1887 char **ret_switch_root_init, /* … */
1888 const char **ret_error_message) {
1889
1890 int r;
1891
1892 assert(m);
1893 assert(ret_reexecute);
1894 assert(ret_retval);
1895 assert(ret_shutdown_verb);
1896 assert(ret_fds);
1897 assert(ret_switch_root_dir);
1898 assert(ret_switch_root_init);
1899 assert(ret_error_message);
1900
1901 for (;;) {
1902 r = manager_loop(m);
1903 if (r < 0) {
1904 *ret_error_message = "Failed to run main loop";
1905 return log_emergency_errno(r, "Failed to run main loop: %m");
1906 }
1907
1908 switch ((ManagerObjective) r) {
1909
1910 case MANAGER_RELOAD: {
1911 LogTarget saved_log_target;
1912 int saved_log_level;
1913
1914 log_info("Reloading.");
1915
1916 /* First, save any overridden log level/target, then parse the configuration file, which might
1917 * change the log level to new settings. */
1918
1919 saved_log_level = m->log_level_overridden ? log_get_max_level() : -1;
1920 saved_log_target = m->log_target_overridden ? log_get_target() : _LOG_TARGET_INVALID;
1921
1922 (void) parse_configuration();
1923
1924 set_manager_defaults(m);
1925
1926 update_cpu_affinity(false);
1927
1928 if (saved_log_level >= 0)
1929 manager_override_log_level(m, saved_log_level);
1930 if (saved_log_target >= 0)
1931 manager_override_log_target(m, saved_log_target);
1932
1933 r = manager_reload(m);
1934 if (r < 0)
1935 /* Reloading failed before the point of no return. Let's continue running as if nothing happened. */
1936 m->objective = MANAGER_OK;
1937
1938 break;
1939 }
1940
1941 case MANAGER_REEXECUTE:
1942
1943 r = prepare_reexecute(m, &arg_serialization, ret_fds, false);
1944 if (r < 0) {
1945 *ret_error_message = "Failed to prepare for reexecution";
1946 return r;
1947 }
1948
1949 log_notice("Reexecuting.");
1950
1951 *ret_reexecute = true;
1952 *ret_retval = EXIT_SUCCESS;
1953 *ret_shutdown_verb = NULL;
1954 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1955
1956 return 0;
1957
1958 case MANAGER_SWITCH_ROOT:
1959 if (!m->switch_root_init) {
1960 r = prepare_reexecute(m, &arg_serialization, ret_fds, true);
1961 if (r < 0) {
1962 *ret_error_message = "Failed to prepare for reexecution";
1963 return r;
1964 }
1965 } else
1966 *ret_fds = NULL;
1967
1968 log_notice("Switching root.");
1969
1970 *ret_reexecute = true;
1971 *ret_retval = EXIT_SUCCESS;
1972 *ret_shutdown_verb = NULL;
1973
1974 /* Steal the switch root parameters */
1975 *ret_switch_root_dir = TAKE_PTR(m->switch_root);
1976 *ret_switch_root_init = TAKE_PTR(m->switch_root_init);
1977
1978 return 0;
1979
1980 case MANAGER_EXIT:
1981
1982 if (MANAGER_IS_USER(m)) {
1983 log_debug("Exit.");
1984
1985 *ret_reexecute = false;
1986 *ret_retval = m->return_value;
1987 *ret_shutdown_verb = NULL;
1988 *ret_fds = NULL;
1989 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1990
1991 return 0;
1992 }
1993
1994 _fallthrough_;
1995 case MANAGER_REBOOT:
1996 case MANAGER_POWEROFF:
1997 case MANAGER_HALT:
1998 case MANAGER_KEXEC: {
1999 static const char * const table[_MANAGER_OBJECTIVE_MAX] = {
2000 [MANAGER_EXIT] = "exit",
2001 [MANAGER_REBOOT] = "reboot",
2002 [MANAGER_POWEROFF] = "poweroff",
2003 [MANAGER_HALT] = "halt",
2004 [MANAGER_KEXEC] = "kexec",
2005 };
2006
2007 log_notice("Shutting down.");
2008
2009 *ret_reexecute = false;
2010 *ret_retval = m->return_value;
2011 assert_se(*ret_shutdown_verb = table[m->objective]);
2012 *ret_fds = NULL;
2013 *ret_switch_root_dir = *ret_switch_root_init = NULL;
2014
2015 return 0;
2016 }
2017
2018 default:
2019 assert_not_reached("Unknown or unexpected manager objective.");
2020 }
2021 }
2022 }
2023
2024 static void log_execution_mode(bool *ret_first_boot) {
2025 assert(ret_first_boot);
2026
2027 if (arg_system) {
2028 int v;
2029
2030 log_info("systemd " GIT_VERSION " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
2031 arg_action == ACTION_TEST ? "test " : "" );
2032
2033 v = detect_virtualization();
2034 if (v > 0)
2035 log_info("Detected virtualization %s.", virtualization_to_string(v));
2036
2037 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
2038
2039 if (in_initrd()) {
2040 *ret_first_boot = false;
2041 log_info("Running in initial RAM disk.");
2042 } else {
2043 /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
2044 * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
2045 * doesn't it's unpopulated. This allows container managers and installers to provision a
2046 * couple of files already. If the container manager wants to provision the machine ID itself
2047 * it should pass $container_uuid to PID 1. */
2048
2049 *ret_first_boot = access("/etc/machine-id", F_OK) < 0;
2050 if (*ret_first_boot)
2051 log_info("Running with unpopulated /etc.");
2052 }
2053 } else {
2054 if (DEBUG_LOGGING) {
2055 _cleanup_free_ char *t;
2056
2057 t = uid_to_name(getuid());
2058 log_debug("systemd " GIT_VERSION " running in %suser mode for user " UID_FMT "/%s. (" SYSTEMD_FEATURES ")",
2059 arg_action == ACTION_TEST ? " test" : "", getuid(), strna(t));
2060 }
2061
2062 *ret_first_boot = false;
2063 }
2064 }
2065
2066 static int initialize_runtime(
2067 bool skip_setup,
2068 struct rlimit *saved_rlimit_nofile,
2069 struct rlimit *saved_rlimit_memlock,
2070 const char **ret_error_message) {
2071
2072 int r;
2073
2074 assert(ret_error_message);
2075
2076 /* Sets up various runtime parameters. Many of these initializations are conditionalized:
2077 *
2078 * - Some only apply to --system instances
2079 * - Some only apply to --user instances
2080 * - Some only apply when we first start up, but not when we reexecute
2081 */
2082
2083 if (arg_action != ACTION_RUN)
2084 return 0;
2085
2086 update_cpu_affinity(skip_setup);
2087
2088 if (arg_system) {
2089 /* Make sure we leave a core dump without panicking the kernel. */
2090 install_crash_handler();
2091
2092 if (!skip_setup) {
2093 r = mount_cgroup_controllers();
2094 if (r < 0) {
2095 *ret_error_message = "Failed to mount cgroup hierarchies";
2096 return r;
2097 }
2098
2099 status_welcome();
2100 hostname_setup();
2101 machine_id_setup(NULL, arg_machine_id, NULL);
2102 loopback_setup();
2103 bump_unix_max_dgram_qlen();
2104 bump_file_max_and_nr_open();
2105 test_usr();
2106 write_container_id();
2107 }
2108
2109 if (arg_watchdog_device) {
2110 r = watchdog_set_device(arg_watchdog_device);
2111 if (r < 0)
2112 log_warning_errno(r, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device);
2113 }
2114
2115 if (arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY)
2116 watchdog_set_timeout(&arg_runtime_watchdog);
2117 }
2118
2119 if (arg_timer_slack_nsec != NSEC_INFINITY)
2120 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
2121 log_warning_errno(errno, "Failed to adjust timer slack, ignoring: %m");
2122
2123 if (arg_system && !cap_test_all(arg_capability_bounding_set)) {
2124 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
2125 if (r < 0) {
2126 *ret_error_message = "Failed to drop capability bounding set of usermode helpers";
2127 return log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
2128 }
2129
2130 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
2131 if (r < 0) {
2132 *ret_error_message = "Failed to drop capability bounding set";
2133 return log_emergency_errno(r, "Failed to drop capability bounding set: %m");
2134 }
2135 }
2136
2137 if (arg_system && arg_no_new_privs) {
2138 if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) {
2139 *ret_error_message = "Failed to disable new privileges";
2140 return log_emergency_errno(errno, "Failed to disable new privileges: %m");
2141 }
2142 }
2143
2144 if (arg_syscall_archs) {
2145 r = enforce_syscall_archs(arg_syscall_archs);
2146 if (r < 0) {
2147 *ret_error_message = "Failed to set syscall architectures";
2148 return r;
2149 }
2150 }
2151
2152 if (!arg_system)
2153 /* Become reaper of our children */
2154 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
2155 log_warning_errno(errno, "Failed to make us a subreaper: %m");
2156
2157 /* Bump up RLIMIT_NOFILE for systemd itself */
2158 (void) bump_rlimit_nofile(saved_rlimit_nofile);
2159 (void) bump_rlimit_memlock(saved_rlimit_memlock);
2160
2161 return 0;
2162 }
2163
2164 static int do_queue_default_job(
2165 Manager *m,
2166 const char **ret_error_message) {
2167
2168 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
2169 Job *default_unit_job;
2170 Unit *target = NULL;
2171 int r;
2172
2173 log_debug("Activating default unit: %s", arg_default_unit);
2174
2175 r = manager_load_startable_unit_or_warn(m, arg_default_unit, NULL, &target);
2176 if (r < 0) {
2177 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET);
2178
2179 r = manager_load_startable_unit_or_warn(m, SPECIAL_RESCUE_TARGET, NULL, &target);
2180 if (r < 0) {
2181 *ret_error_message = r == -ERFKILL ? "Rescue target masked"
2182 : "Failed to load rescue target";
2183 return r;
2184 }
2185 }
2186
2187 assert(target->load_state == UNIT_LOADED);
2188
2189 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, NULL, &error, &default_unit_job);
2190 if (r == -EPERM) {
2191 log_debug_errno(r, "Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
2192
2193 sd_bus_error_free(&error);
2194
2195 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, NULL, &error, &default_unit_job);
2196 if (r < 0) {
2197 *ret_error_message = "Failed to start default target";
2198 return log_emergency_errno(r, "Failed to start default target: %s", bus_error_message(&error, r));
2199 }
2200
2201 } else if (r < 0) {
2202 *ret_error_message = "Failed to isolate default target";
2203 return log_emergency_errno(r, "Failed to isolate default target: %s", bus_error_message(&error, r));
2204 }
2205
2206 m->default_unit_job_id = default_unit_job->id;
2207
2208 return 0;
2209 }
2210
2211 static void reset_arguments(void) {
2212 /* Frees/resets arg_* variables, with a few exceptions commented below. */
2213
2214 arg_default_unit = mfree(arg_default_unit);
2215
2216 /* arg_system — ignore */
2217
2218 arg_dump_core = true;
2219 arg_crash_chvt = -1;
2220 arg_crash_shell = false;
2221 arg_crash_reboot = false;
2222 arg_confirm_spawn = mfree(arg_confirm_spawn);
2223 arg_show_status = _SHOW_STATUS_INVALID;
2224 arg_switched_root = false;
2225 arg_pager_flags = 0;
2226 arg_service_watchdogs = true;
2227 arg_default_std_output = EXEC_OUTPUT_JOURNAL;
2228 arg_default_std_error = EXEC_OUTPUT_INHERIT;
2229 arg_default_restart_usec = DEFAULT_RESTART_USEC;
2230 arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
2231 arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
2232 arg_default_timeout_abort_usec = DEFAULT_TIMEOUT_USEC;
2233 arg_default_timeout_abort_set = false;
2234 arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
2235 arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
2236 arg_runtime_watchdog = 0;
2237 arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
2238 arg_early_core_pattern = NULL;
2239 arg_watchdog_device = NULL;
2240
2241 arg_default_environment = strv_free(arg_default_environment);
2242 rlimit_free_all(arg_default_rlimit);
2243
2244 arg_capability_bounding_set = CAP_ALL;
2245 arg_no_new_privs = false;
2246 arg_timer_slack_nsec = NSEC_INFINITY;
2247 arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
2248
2249 arg_syscall_archs = set_free(arg_syscall_archs);
2250
2251 /* arg_serialization — ignore */
2252
2253 arg_default_cpu_accounting = -1;
2254 arg_default_io_accounting = false;
2255 arg_default_ip_accounting = false;
2256 arg_default_blockio_accounting = false;
2257 arg_default_memory_accounting = MEMORY_ACCOUNTING_DEFAULT;
2258 arg_default_tasks_accounting = true;
2259 arg_default_tasks_max = UINT64_MAX;
2260 arg_machine_id = (sd_id128_t) {};
2261 arg_cad_burst_action = EMERGENCY_ACTION_REBOOT_FORCE;
2262 arg_default_oom_policy = OOM_STOP;
2263
2264 cpu_set_reset(&arg_cpu_affinity);
2265 }
2266
2267 static int parse_configuration(void) {
2268 int r;
2269
2270 arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U);
2271
2272 /* Assign configuration defaults */
2273 reset_arguments();
2274
2275 r = parse_config_file();
2276 if (r < 0)
2277 log_warning_errno(r, "Failed to parse config file, ignoring: %m");
2278
2279 if (arg_system) {
2280 r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
2281 if (r < 0)
2282 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
2283 }
2284
2285 /* Note that this also parses bits from the kernel command line, including "debug". */
2286 log_parse_environment();
2287
2288 return 0;
2289 }
2290
2291 static int load_configuration(int argc, char **argv, const char **ret_error_message) {
2292 int r;
2293
2294 assert(ret_error_message);
2295
2296 (void) parse_configuration();
2297
2298 r = parse_argv(argc, argv);
2299 if (r < 0) {
2300 *ret_error_message = "Failed to parse commandline arguments";
2301 return r;
2302 }
2303
2304 /* Initialize default unit */
2305 if (!arg_default_unit) {
2306 arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET);
2307 if (!arg_default_unit) {
2308 *ret_error_message = "Failed to set default unit";
2309 return log_oom();
2310 }
2311 }
2312
2313 /* Initialize the show status setting if it hasn't been set explicitly yet */
2314 if (arg_show_status == _SHOW_STATUS_INVALID)
2315 arg_show_status = SHOW_STATUS_YES;
2316
2317 return 0;
2318 }
2319
2320 static int safety_checks(void) {
2321
2322 if (getpid_cached() == 1 &&
2323 arg_action != ACTION_RUN)
2324 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2325 "Unsupported execution mode while PID 1.");
2326
2327 if (getpid_cached() == 1 &&
2328 !arg_system)
2329 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2330 "Can't run --user mode as PID 1.");
2331
2332 if (arg_action == ACTION_RUN &&
2333 arg_system &&
2334 getpid_cached() != 1)
2335 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2336 "Can't run system mode unless PID 1.");
2337
2338 if (arg_action == ACTION_TEST &&
2339 geteuid() == 0)
2340 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2341 "Don't run test mode as root.");
2342
2343 if (!arg_system &&
2344 arg_action == ACTION_RUN &&
2345 sd_booted() <= 0)
2346 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
2347 "Trying to run as user instance, but the system has not been booted with systemd.");
2348
2349 if (!arg_system &&
2350 arg_action == ACTION_RUN &&
2351 !getenv("XDG_RUNTIME_DIR"))
2352 return log_error_errno(SYNTHETIC_ERRNO(EUNATCH),
2353 "Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
2354
2355 if (arg_system &&
2356 arg_action == ACTION_RUN &&
2357 running_in_chroot() > 0)
2358 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
2359 "Cannot be run in a chroot() environment.");
2360
2361 return 0;
2362 }
2363
2364 static int initialize_security(
2365 bool *loaded_policy,
2366 dual_timestamp *security_start_timestamp,
2367 dual_timestamp *security_finish_timestamp,
2368 const char **ret_error_message) {
2369
2370 int r;
2371
2372 assert(loaded_policy);
2373 assert(security_start_timestamp);
2374 assert(security_finish_timestamp);
2375 assert(ret_error_message);
2376
2377 dual_timestamp_get(security_start_timestamp);
2378
2379 r = mac_selinux_setup(loaded_policy);
2380 if (r < 0) {
2381 *ret_error_message = "Failed to load SELinux policy";
2382 return r;
2383 }
2384
2385 r = mac_smack_setup(loaded_policy);
2386 if (r < 0) {
2387 *ret_error_message = "Failed to load SMACK policy";
2388 return r;
2389 }
2390
2391 r = ima_setup();
2392 if (r < 0) {
2393 *ret_error_message = "Failed to load IMA policy";
2394 return r;
2395 }
2396
2397 dual_timestamp_get(security_finish_timestamp);
2398 return 0;
2399 }
2400
2401 static void test_summary(Manager *m) {
2402 assert(m);
2403
2404 printf("-> By units:\n");
2405 manager_dump_units(m, stdout, "\t");
2406
2407 printf("-> By jobs:\n");
2408 manager_dump_jobs(m, stdout, "\t");
2409 }
2410
2411 static int collect_fds(FDSet **ret_fds, const char **ret_error_message) {
2412 int r;
2413
2414 assert(ret_fds);
2415 assert(ret_error_message);
2416
2417 r = fdset_new_fill(ret_fds);
2418 if (r < 0) {
2419 *ret_error_message = "Failed to allocate fd set";
2420 return log_emergency_errno(r, "Failed to allocate fd set: %m");
2421 }
2422
2423 fdset_cloexec(*ret_fds, true);
2424
2425 if (arg_serialization)
2426 assert_se(fdset_remove(*ret_fds, fileno(arg_serialization)) >= 0);
2427
2428 return 0;
2429 }
2430
2431 static void setup_console_terminal(bool skip_setup) {
2432
2433 if (!arg_system)
2434 return;
2435
2436 /* Become a session leader if we aren't one yet. */
2437 (void) setsid();
2438
2439 /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2440 * tty. */
2441 (void) release_terminal();
2442
2443 /* Reset the console, but only if this is really init and we are freshly booted */
2444 if (getpid_cached() == 1 && !skip_setup)
2445 (void) console_setup();
2446 }
2447
2448 static bool early_skip_setup_check(int argc, char *argv[]) {
2449 bool found_deserialize = false;
2450 int i;
2451
2452 /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2453 * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2454 * anyway, even if in that case we also do deserialization. */
2455
2456 for (i = 1; i < argc; i++) {
2457 if (streq(argv[i], "--switched-root"))
2458 return false; /* If we switched root, don't skip the setup. */
2459 else if (streq(argv[i], "--deserialize"))
2460 found_deserialize = true;
2461 }
2462
2463 return found_deserialize; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2464 }
2465
2466 int main(int argc, char *argv[]) {
2467
2468 dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL, userspace_timestamp = DUAL_TIMESTAMP_NULL, kernel_timestamp = DUAL_TIMESTAMP_NULL,
2469 security_start_timestamp = DUAL_TIMESTAMP_NULL, security_finish_timestamp = DUAL_TIMESTAMP_NULL;
2470 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0),
2471 saved_rlimit_memlock = RLIMIT_MAKE_CONST(RLIM_INFINITY); /* The original rlimits we passed
2472 * in. Note we use different values
2473 * for the two that indicate whether
2474 * these fields are initialized! */
2475 bool skip_setup, loaded_policy = false, queue_default_job = false, first_boot = false, reexecute = false;
2476 char *switch_root_dir = NULL, *switch_root_init = NULL;
2477 usec_t before_startup, after_startup;
2478 static char systemd[] = "systemd";
2479 char timespan[FORMAT_TIMESPAN_MAX];
2480 const char *shutdown_verb = NULL, *error_message = NULL;
2481 int r, retval = EXIT_FAILURE;
2482 Manager *m = NULL;
2483 FDSet *fds = NULL;
2484
2485 /* SysV compatibility: redirect init → telinit */
2486 redirect_telinit(argc, argv);
2487
2488 /* Take timestamps early on */
2489 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
2490 dual_timestamp_get(&userspace_timestamp);
2491
2492 /* Figure out whether we need to do initialize the system, or if we already did that because we are
2493 * reexecuting */
2494 skip_setup = early_skip_setup_check(argc, argv);
2495
2496 /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2497 * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
2498 program_invocation_short_name = systemd;
2499 (void) prctl(PR_SET_NAME, systemd);
2500
2501 /* Save the original command line */
2502 save_argc_argv(argc, argv);
2503
2504 /* Make sure that if the user says "syslog" we actually log to the journal. */
2505 log_set_upgrade_syslog_to_journal(true);
2506
2507 if (getpid_cached() == 1) {
2508 /* When we run as PID 1 force system mode */
2509 arg_system = true;
2510
2511 /* Disable the umask logic */
2512 umask(0);
2513
2514 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2515 * activated yet (even though the log socket for it exists). */
2516 log_set_prohibit_ipc(true);
2517
2518 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2519 * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2520 * child process right before execve()'ing the actual binary, at a point in time where socket
2521 * activation stderr/stdout area already set up. */
2522 log_set_always_reopen_console(true);
2523
2524 if (detect_container() <= 0) {
2525
2526 /* Running outside of a container as PID 1 */
2527 log_set_target(LOG_TARGET_KMSG);
2528 log_open();
2529
2530 if (in_initrd())
2531 initrd_timestamp = userspace_timestamp;
2532
2533 if (!skip_setup) {
2534 r = mount_setup_early();
2535 if (r < 0) {
2536 error_message = "Failed to mount early API filesystems";
2537 goto finish;
2538 }
2539
2540 r = initialize_security(
2541 &loaded_policy,
2542 &security_start_timestamp,
2543 &security_finish_timestamp,
2544 &error_message);
2545 if (r < 0)
2546 goto finish;
2547 }
2548
2549 if (mac_selinux_init() < 0) {
2550 error_message = "Failed to initialize SELinux policy";
2551 goto finish;
2552 }
2553
2554 if (!skip_setup)
2555 initialize_clock();
2556
2557 /* Set the default for later on, but don't actually open the logs like this for now. Note that
2558 * if we are transitioning from the initrd there might still be journal fd open, and we
2559 * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2560 * elsewhere. */
2561 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
2562
2563 } else {
2564 /* Running inside a container, as PID 1 */
2565 log_set_target(LOG_TARGET_CONSOLE);
2566 log_open();
2567
2568 /* For later on, see above... */
2569 log_set_target(LOG_TARGET_JOURNAL);
2570
2571 /* clear the kernel timestamp,
2572 * because we are in a container */
2573 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2574 }
2575
2576 initialize_coredump(skip_setup);
2577
2578 r = fixup_environment();
2579 if (r < 0) {
2580 log_emergency_errno(r, "Failed to fix up PID 1 environment: %m");
2581 error_message = "Failed to fix up PID1 environment";
2582 goto finish;
2583 }
2584
2585 } else {
2586 /* Running as user instance */
2587 arg_system = false;
2588 log_set_target(LOG_TARGET_AUTO);
2589 log_open();
2590
2591 /* clear the kernel timestamp,
2592 * because we are not PID 1 */
2593 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2594 }
2595
2596 if (arg_system) {
2597 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2598 * they never log into the console. */
2599 log_show_color(colors_enabled());
2600
2601 r = make_null_stdio();
2602 if (r < 0)
2603 log_warning_errno(r, "Failed to redirect standard streams to /dev/null, ignoring: %m");
2604 }
2605
2606 /* Mount /proc, /sys and friends, so that /proc/cmdline and
2607 * /proc/$PID/fd is available. */
2608 if (getpid_cached() == 1) {
2609
2610 /* Load the kernel modules early. */
2611 if (!skip_setup)
2612 kmod_setup();
2613
2614 r = mount_setup(loaded_policy);
2615 if (r < 0) {
2616 error_message = "Failed to mount API filesystems";
2617 goto finish;
2618 }
2619 }
2620
2621 /* Reset all signal handlers. */
2622 (void) reset_all_signal_handlers();
2623 (void) ignore_signals(SIGNALS_IGNORE, -1);
2624
2625 r = load_configuration(argc, argv, &error_message);
2626 if (r < 0)
2627 goto finish;
2628
2629 r = safety_checks();
2630 if (r < 0)
2631 goto finish;
2632
2633 if (IN_SET(arg_action, ACTION_TEST, ACTION_HELP, ACTION_DUMP_CONFIGURATION_ITEMS, ACTION_DUMP_BUS_PROPERTIES))
2634 (void) pager_open(arg_pager_flags);
2635
2636 if (arg_action != ACTION_RUN)
2637 skip_setup = true;
2638
2639 if (arg_action == ACTION_HELP) {
2640 retval = help() < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
2641 goto finish;
2642 } else if (arg_action == ACTION_VERSION) {
2643 retval = version();
2644 goto finish;
2645 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
2646 unit_dump_config_items(stdout);
2647 retval = EXIT_SUCCESS;
2648 goto finish;
2649 } else if (arg_action == ACTION_DUMP_BUS_PROPERTIES) {
2650 dump_bus_properties(stdout);
2651 retval = EXIT_SUCCESS;
2652 goto finish;
2653 }
2654
2655 assert_se(IN_SET(arg_action, ACTION_RUN, ACTION_TEST));
2656
2657 /* Move out of the way, so that we won't block unmounts */
2658 assert_se(chdir("/") == 0);
2659
2660 if (arg_action == ACTION_RUN) {
2661
2662 /* A core pattern might have been specified via the cmdline. */
2663 initialize_core_pattern(skip_setup);
2664
2665 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2666 log_close();
2667
2668 /* Remember open file descriptors for later deserialization */
2669 r = collect_fds(&fds, &error_message);
2670 if (r < 0)
2671 goto finish;
2672
2673 /* Give up any control of the console, but make sure its initialized. */
2674 setup_console_terminal(skip_setup);
2675
2676 /* Open the logging devices, if possible and necessary */
2677 log_open();
2678 }
2679
2680 log_execution_mode(&first_boot);
2681
2682 r = initialize_runtime(skip_setup,
2683 &saved_rlimit_nofile,
2684 &saved_rlimit_memlock,
2685 &error_message);
2686 if (r < 0)
2687 goto finish;
2688
2689 r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER,
2690 arg_action == ACTION_TEST ? MANAGER_TEST_FULL : 0,
2691 &m);
2692 if (r < 0) {
2693 log_emergency_errno(r, "Failed to allocate manager object: %m");
2694 error_message = "Failed to allocate manager object";
2695 goto finish;
2696 }
2697
2698 m->timestamps[MANAGER_TIMESTAMP_KERNEL] = kernel_timestamp;
2699 m->timestamps[MANAGER_TIMESTAMP_INITRD] = initrd_timestamp;
2700 m->timestamps[MANAGER_TIMESTAMP_USERSPACE] = userspace_timestamp;
2701 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START)] = security_start_timestamp;
2702 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH)] = security_finish_timestamp;
2703
2704 set_manager_defaults(m);
2705 set_manager_settings(m);
2706 manager_set_first_boot(m, first_boot);
2707
2708 /* Remember whether we should queue the default job */
2709 queue_default_job = !arg_serialization || arg_switched_root;
2710
2711 before_startup = now(CLOCK_MONOTONIC);
2712
2713 r = manager_startup(m, arg_serialization, fds);
2714 if (r < 0) {
2715 error_message = "Failed to start up manager";
2716 goto finish;
2717 }
2718
2719 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2720 fds = fdset_free(fds);
2721 arg_serialization = safe_fclose(arg_serialization);
2722
2723 if (queue_default_job) {
2724 r = do_queue_default_job(m, &error_message);
2725 if (r < 0)
2726 goto finish;
2727 }
2728
2729 after_startup = now(CLOCK_MONOTONIC);
2730
2731 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
2732 "Loaded units and determined initial transaction in %s.",
2733 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
2734
2735 if (arg_action == ACTION_TEST) {
2736 test_summary(m);
2737 retval = EXIT_SUCCESS;
2738 goto finish;
2739 }
2740
2741 (void) invoke_main_loop(m,
2742 &reexecute,
2743 &retval,
2744 &shutdown_verb,
2745 &fds,
2746 &switch_root_dir,
2747 &switch_root_init,
2748 &error_message);
2749
2750 finish:
2751 pager_close();
2752
2753 if (m) {
2754 arg_shutdown_watchdog = m->shutdown_watchdog;
2755 m = manager_free(m);
2756 }
2757
2758 reset_arguments();
2759 mac_selinux_finish();
2760
2761 if (reexecute)
2762 do_reexecute(argc, argv,
2763 &saved_rlimit_nofile,
2764 &saved_rlimit_memlock,
2765 fds,
2766 switch_root_dir,
2767 switch_root_init,
2768 &error_message); /* This only returns if reexecution failed */
2769
2770 arg_serialization = safe_fclose(arg_serialization);
2771 fds = fdset_free(fds);
2772
2773 #if HAVE_VALGRIND_VALGRIND_H
2774 /* If we are PID 1 and running under valgrind, then let's exit
2775 * here explicitly. valgrind will only generate nice output on
2776 * exit(), not on exec(), hence let's do the former not the
2777 * latter here. */
2778 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND) {
2779 /* Cleanup watchdog_device strings for valgrind. We need them
2780 * in become_shutdown() so normally we cannot free them yet. */
2781 watchdog_free_device();
2782 arg_watchdog_device = mfree(arg_watchdog_device);
2783 return retval;
2784 }
2785 #endif
2786
2787 #if HAS_FEATURE_ADDRESS_SANITIZER
2788 __lsan_do_leak_check();
2789 #endif
2790
2791 if (shutdown_verb) {
2792 r = become_shutdown(shutdown_verb, retval);
2793 log_error_errno(r, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
2794 error_message = "Failed to execute shutdown binary";
2795 }
2796
2797 watchdog_free_device();
2798 arg_watchdog_device = mfree(arg_watchdog_device);
2799
2800 if (getpid_cached() == 1) {
2801 if (error_message)
2802 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2803 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
2804 "%s.", error_message);
2805 freeze_or_exit_or_reboot();
2806 }
2807
2808 return retval;
2809 }
Unnamed repository; edit this file 'description' to name the repository.