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


44b0a65f222be9d117867ef1bb082351eedb814b
[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 static char *arg_default_unit = NULL;
100 static bool arg_system = false;
101 static bool arg_dump_core = true;
102 static int arg_crash_chvt = -1;
103 static bool arg_crash_shell = false;
104 static bool arg_crash_reboot = false;
105 static char *arg_confirm_spawn = NULL;
106 static ShowStatus arg_show_status = _SHOW_STATUS_INVALID;
107 static bool arg_switched_root = false;
108 static PagerFlags arg_pager_flags = 0;
109 static bool arg_service_watchdogs = true;
110 static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
111 static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
112 static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
113 static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
114 static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
115 static usec_t arg_default_timeout_abort_usec = DEFAULT_TIMEOUT_USEC;
116 static bool arg_default_timeout_abort_set = false;
117 static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
118 static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
119 static usec_t arg_runtime_watchdog = 0;
120 static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
121 static char *arg_early_core_pattern = NULL;
122 static char *arg_watchdog_device = NULL;
123 static char **arg_default_environment = NULL;
124 static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {};
125 static uint64_t arg_capability_bounding_set = CAP_ALL;
126 static bool arg_no_new_privs = false;
127 static nsec_t arg_timer_slack_nsec = NSEC_INFINITY;
128 static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
129 static Set* arg_syscall_archs = NULL;
130 static FILE* arg_serialization = NULL;
131 static int arg_default_cpu_accounting = -1;
132 static bool arg_default_io_accounting = false;
133 static bool arg_default_ip_accounting = false;
134 static bool arg_default_blockio_accounting = false;
135 static bool arg_default_memory_accounting = MEMORY_ACCOUNTING_DEFAULT;
136 static bool arg_default_tasks_accounting = true;
137 static uint64_t arg_default_tasks_max = UINT64_MAX;
138 static sd_id128_t arg_machine_id = {};
139 static EmergencyAction arg_cad_burst_action = EMERGENCY_ACTION_REBOOT_FORCE;
140 static OOMPolicy arg_default_oom_policy = OOM_STOP;
141
142 static CPUSet arg_cpu_affinity = {};
143
144 static int parse_configuration(void);
145
146 _noreturn_ static void freeze_or_exit_or_reboot(void) {
147
148 /* If we are running in a container, let's prefer exiting, after all we can propagate an exit code to
149 * the container manager, and thus inform it that something went wrong. */
150 if (detect_container() > 0) {
151 log_emergency("Exiting PID 1...");
152 _exit(EXIT_EXCEPTION);
153 }
154
155 if (arg_crash_reboot) {
156 log_notice("Rebooting in 10s...");
157 (void) sleep(10);
158
159 log_notice("Rebooting now...");
160 (void) reboot(RB_AUTOBOOT);
161 log_emergency_errno(errno, "Failed to reboot: %m");
162 }
163
164 log_emergency("Freezing execution.");
165 freeze();
166 }
167
168 _noreturn_ static void crash(int sig) {
169 struct sigaction sa;
170 pid_t pid;
171
172 if (getpid_cached() != 1)
173 /* Pass this on immediately, if this is not PID 1 */
174 (void) raise(sig);
175 else if (!arg_dump_core)
176 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
177 else {
178 sa = (struct sigaction) {
179 .sa_handler = nop_signal_handler,
180 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
181 };
182
183 /* We want to wait for the core process, hence let's enable SIGCHLD */
184 (void) sigaction(SIGCHLD, &sa, NULL);
185
186 pid = raw_clone(SIGCHLD);
187 if (pid < 0)
188 log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
189 else if (pid == 0) {
190 /* Enable default signal handler for core dump */
191
192 sa = (struct sigaction) {
193 .sa_handler = SIG_DFL,
194 };
195 (void) sigaction(sig, &sa, NULL);
196
197 /* Don't limit the coredump size */
198 (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
199
200 /* Just to be sure... */
201 (void) chdir("/");
202
203 /* Raise the signal again */
204 pid = raw_getpid();
205 (void) kill(pid, sig); /* raise() would kill the parent */
206
207 assert_not_reached("We shouldn't be here...");
208 _exit(EXIT_EXCEPTION);
209 } else {
210 siginfo_t status;
211 int r;
212
213 /* Order things nicely. */
214 r = wait_for_terminate(pid, &status);
215 if (r < 0)
216 log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
217 else if (status.si_code != CLD_DUMPED)
218 log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
219 signal_to_string(sig),
220 pid, sigchld_code_to_string(status.si_code),
221 status.si_status,
222 strna(status.si_code == CLD_EXITED
223 ? exit_status_to_string(status.si_status, EXIT_STATUS_MINIMAL)
224 : signal_to_string(status.si_status)));
225 else
226 log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
227 }
228 }
229
230 if (arg_crash_chvt >= 0)
231 (void) chvt(arg_crash_chvt);
232
233 sa = (struct sigaction) {
234 .sa_handler = SIG_IGN,
235 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
236 };
237
238 /* Let the kernel reap children for us */
239 (void) sigaction(SIGCHLD, &sa, NULL);
240
241 if (arg_crash_shell) {
242 log_notice("Executing crash shell in 10s...");
243 (void) sleep(10);
244
245 pid = raw_clone(SIGCHLD);
246 if (pid < 0)
247 log_emergency_errno(errno, "Failed to fork off crash shell: %m");
248 else if (pid == 0) {
249 (void) setsid();
250 (void) make_console_stdio();
251 (void) rlimit_nofile_safe();
252 (void) execle("/bin/sh", "/bin/sh", NULL, environ);
253
254 log_emergency_errno(errno, "execle() failed: %m");
255 _exit(EXIT_EXCEPTION);
256 } else {
257 log_info("Spawned crash shell as PID "PID_FMT".", pid);
258 (void) wait_for_terminate(pid, NULL);
259 }
260 }
261
262 freeze_or_exit_or_reboot();
263 }
264
265 static void install_crash_handler(void) {
266 static const struct sigaction sa = {
267 .sa_handler = crash,
268 .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
269 };
270 int r;
271
272 /* We ignore the return value here, since, we don't mind if we
273 * cannot set up a crash handler */
274 r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
275 if (r < 0)
276 log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
277 }
278
279 static int console_setup(void) {
280 _cleanup_close_ int tty_fd = -1;
281 int r;
282
283 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
284 if (tty_fd < 0)
285 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
286
287 /* We don't want to force text mode. plymouth may be showing
288 * pictures already from initrd. */
289 r = reset_terminal_fd(tty_fd, false);
290 if (r < 0)
291 return log_error_errno(r, "Failed to reset /dev/console: %m");
292
293 return 0;
294 }
295
296 static int parse_crash_chvt(const char *value) {
297 int b;
298
299 if (safe_atoi(value, &arg_crash_chvt) >= 0)
300 return 0;
301
302 b = parse_boolean(value);
303 if (b < 0)
304 return b;
305
306 if (b > 0)
307 arg_crash_chvt = 0; /* switch to where kmsg goes */
308 else
309 arg_crash_chvt = -1; /* turn off switching */
310
311 return 0;
312 }
313
314 static int parse_confirm_spawn(const char *value, char **console) {
315 char *s;
316 int r;
317
318 r = value ? parse_boolean(value) : 1;
319 if (r == 0) {
320 *console = NULL;
321 return 0;
322 }
323
324 if (r > 0) /* on with default tty */
325 s = strdup("/dev/console");
326 else if (is_path(value)) /* on with fully qualified path */
327 s = strdup(value);
328 else /* on with only a tty file name, not a fully qualified path */
329 s = strjoin("/dev/", value);
330 if (!s)
331 return -ENOMEM;
332
333 *console = s;
334 return 0;
335 }
336
337 static int set_machine_id(const char *m) {
338 sd_id128_t t;
339 assert(m);
340
341 if (sd_id128_from_string(m, &t) < 0)
342 return -EINVAL;
343
344 if (sd_id128_is_null(t))
345 return -EINVAL;
346
347 arg_machine_id = t;
348 return 0;
349 }
350
351 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
352
353 int r;
354
355 assert(key);
356
357 if (STR_IN_SET(key, "systemd.unit", "rd.systemd.unit")) {
358
359 if (proc_cmdline_value_missing(key, value))
360 return 0;
361
362 if (!unit_name_is_valid(value, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE))
363 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key, value);
364 else if (in_initrd() == !!startswith(key, "rd.")) {
365 if (free_and_strdup(&arg_default_unit, value) < 0)
366 return log_oom();
367 }
368
369 } else if (proc_cmdline_key_streq(key, "systemd.dump_core")) {
370
371 r = value ? parse_boolean(value) : true;
372 if (r < 0)
373 log_warning_errno(r, "Failed to parse dump core switch %s, ignoring: %m", value);
374 else
375 arg_dump_core = r;
376
377 } else if (proc_cmdline_key_streq(key, "systemd.early_core_pattern")) {
378
379 if (proc_cmdline_value_missing(key, value))
380 return 0;
381
382 if (path_is_absolute(value))
383 (void) parse_path_argument_and_warn(value, false, &arg_early_core_pattern);
384 else
385 log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value);
386
387 } else if (proc_cmdline_key_streq(key, "systemd.crash_chvt")) {
388
389 if (!value)
390 arg_crash_chvt = 0; /* turn on */
391 else {
392 r = parse_crash_chvt(value);
393 if (r < 0)
394 log_warning_errno(r, "Failed to parse crash chvt switch %s, ignoring: %m", value);
395 }
396
397 } else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) {
398
399 r = value ? parse_boolean(value) : true;
400 if (r < 0)
401 log_warning_errno(r, "Failed to parse crash shell switch %s, ignoring: %m", value);
402 else
403 arg_crash_shell = r;
404
405 } else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) {
406
407 r = value ? parse_boolean(value) : true;
408 if (r < 0)
409 log_warning_errno(r, "Failed to parse crash reboot switch %s, ignoring: %m", value);
410 else
411 arg_crash_reboot = r;
412
413 } else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) {
414 char *s;
415
416 r = parse_confirm_spawn(value, &s);
417 if (r < 0)
418 log_warning_errno(r, "Failed to parse confirm_spawn switch %s, ignoring: %m", value);
419 else
420 free_and_replace(arg_confirm_spawn, s);
421
422 } else if (proc_cmdline_key_streq(key, "systemd.service_watchdogs")) {
423
424 r = value ? parse_boolean(value) : true;
425 if (r < 0)
426 log_warning_errno(r, "Failed to parse service watchdog switch %s, ignoring: %m", value);
427 else
428 arg_service_watchdogs = r;
429
430 } else if (proc_cmdline_key_streq(key, "systemd.show_status")) {
431
432 if (value) {
433 r = parse_show_status(value, &arg_show_status);
434 if (r < 0)
435 log_warning_errno(r, "Failed to parse show status switch %s, ignoring: %m", value);
436 } else
437 arg_show_status = SHOW_STATUS_YES;
438
439 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) {
440
441 if (proc_cmdline_value_missing(key, value))
442 return 0;
443
444 r = exec_output_from_string(value);
445 if (r < 0)
446 log_warning_errno(r, "Failed to parse default standard output switch %s, ignoring: %m", value);
447 else
448 arg_default_std_output = r;
449
450 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) {
451
452 if (proc_cmdline_value_missing(key, value))
453 return 0;
454
455 r = exec_output_from_string(value);
456 if (r < 0)
457 log_warning_errno(r, "Failed to parse default standard error switch %s, ignoring: %m", value);
458 else
459 arg_default_std_error = r;
460
461 } else if (streq(key, "systemd.setenv")) {
462
463 if (proc_cmdline_value_missing(key, value))
464 return 0;
465
466 if (env_assignment_is_valid(value)) {
467 char **env;
468
469 env = strv_env_set(arg_default_environment, value);
470 if (!env)
471 return log_oom();
472
473 arg_default_environment = env;
474 } else
475 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
476
477 } else if (proc_cmdline_key_streq(key, "systemd.machine_id")) {
478
479 if (proc_cmdline_value_missing(key, value))
480 return 0;
481
482 r = set_machine_id(value);
483 if (r < 0)
484 log_warning_errno(r, "MachineID '%s' is not valid, ignoring: %m", value);
485
486 } else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) {
487
488 if (proc_cmdline_value_missing(key, value))
489 return 0;
490
491 r = parse_sec(value, &arg_default_timeout_start_usec);
492 if (r < 0)
493 log_warning_errno(r, "Failed to parse default start timeout '%s', ignoring: %m", value);
494
495 if (arg_default_timeout_start_usec <= 0)
496 arg_default_timeout_start_usec = USEC_INFINITY;
497
498 } else if (proc_cmdline_key_streq(key, "systemd.watchdog_device")) {
499
500 if (proc_cmdline_value_missing(key, value))
501 return 0;
502
503 (void) parse_path_argument_and_warn(value, false, &arg_watchdog_device);
504
505 } else if (streq(key, "quiet") && !value) {
506
507 if (arg_show_status == _SHOW_STATUS_INVALID)
508 arg_show_status = SHOW_STATUS_AUTO;
509
510 } else if (streq(key, "debug") && !value) {
511
512 /* Note that log_parse_environment() handles 'debug'
513 * too, and sets the log level to LOG_DEBUG. */
514
515 if (detect_container() > 0)
516 log_set_target(LOG_TARGET_CONSOLE);
517
518 } else if (!value) {
519 const char *target;
520
521 /* SysV compatibility */
522 target = runlevel_to_target(key);
523 if (target)
524 return free_and_strdup(&arg_default_unit, target);
525 }
526
527 return 0;
528 }
529
530 #define DEFINE_SETTER(name, func, descr) \
531 static int name(const char *unit, \
532 const char *filename, \
533 unsigned line, \
534 const char *section, \
535 unsigned section_line, \
536 const char *lvalue, \
537 int ltype, \
538 const char *rvalue, \
539 void *data, \
540 void *userdata) { \
541 \
542 int r; \
543 \
544 assert(filename); \
545 assert(lvalue); \
546 assert(rvalue); \
547 \
548 r = func(rvalue); \
549 if (r < 0) \
550 log_syntax(unit, LOG_ERR, filename, line, r, \
551 "Invalid " descr "'%s': %m", \
552 rvalue); \
553 \
554 return 0; \
555 }
556
557 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level");
558 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target");
559 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" );
560 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location");
561
562 static int config_parse_cpu_affinity2(
563 const char *unit,
564 const char *filename,
565 unsigned line,
566 const char *section,
567 unsigned section_line,
568 const char *lvalue,
569 int ltype,
570 const char *rvalue,
571 void *data,
572 void *userdata) {
573
574 CPUSet *affinity = data;
575
576 assert(affinity);
577
578 (void) parse_cpu_set_extend(rvalue, affinity, true, unit, filename, line, lvalue);
579
580 return 0;
581 }
582
583 static int config_parse_show_status(
584 const char* unit,
585 const char *filename,
586 unsigned line,
587 const char *section,
588 unsigned section_line,
589 const char *lvalue,
590 int ltype,
591 const char *rvalue,
592 void *data,
593 void *userdata) {
594
595 int k;
596 ShowStatus *b = data;
597
598 assert(filename);
599 assert(lvalue);
600 assert(rvalue);
601 assert(data);
602
603 k = parse_show_status(rvalue, b);
604 if (k < 0) {
605 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
606 return 0;
607 }
608
609 return 0;
610 }
611
612 static int config_parse_output_restricted(
613 const char* unit,
614 const char *filename,
615 unsigned line,
616 const char *section,
617 unsigned section_line,
618 const char *lvalue,
619 int ltype,
620 const char *rvalue,
621 void *data,
622 void *userdata) {
623
624 ExecOutput t, *eo = data;
625
626 assert(filename);
627 assert(lvalue);
628 assert(rvalue);
629 assert(data);
630
631 t = exec_output_from_string(rvalue);
632 if (t < 0) {
633 log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse output type, ignoring: %s", rvalue);
634 return 0;
635 }
636
637 if (IN_SET(t, EXEC_OUTPUT_SOCKET, EXEC_OUTPUT_NAMED_FD, EXEC_OUTPUT_FILE, EXEC_OUTPUT_FILE_APPEND)) {
638 log_syntax(unit, LOG_ERR, filename, line, 0, "Standard output types socket, fd:, file:, append: are not supported as defaults, ignoring: %s", rvalue);
639 return 0;
640 }
641
642 *eo = t;
643 return 0;
644 }
645
646 static int config_parse_crash_chvt(
647 const char* unit,
648 const char *filename,
649 unsigned line,
650 const char *section,
651 unsigned section_line,
652 const char *lvalue,
653 int ltype,
654 const char *rvalue,
655 void *data,
656 void *userdata) {
657
658 int r;
659
660 assert(filename);
661 assert(lvalue);
662 assert(rvalue);
663
664 r = parse_crash_chvt(rvalue);
665 if (r < 0) {
666 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
667 return 0;
668 }
669
670 return 0;
671 }
672
673 static int config_parse_timeout_abort(
674 const char* unit,
675 const char *filename,
676 unsigned line,
677 const char *section,
678 unsigned section_line,
679 const char *lvalue,
680 int ltype,
681 const char *rvalue,
682 void *data,
683 void *userdata) {
684
685 int r;
686
687 assert(filename);
688 assert(lvalue);
689 assert(rvalue);
690
691 rvalue += strspn(rvalue, WHITESPACE);
692 if (isempty(rvalue)) {
693 arg_default_timeout_abort_set = false;
694 return 0;
695 }
696
697 r = parse_sec(rvalue, &arg_default_timeout_abort_usec);
698 if (r < 0) {
699 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse DefaultTimeoutAbortSec= setting, ignoring: %s", rvalue);
700 return 0;
701 }
702
703 arg_default_timeout_abort_set = true;
704 return 0;
705 }
706
707 static int parse_config_file(void) {
708
709 const ConfigTableItem items[] = {
710 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
711 { "Manager", "LogTarget", config_parse_target, 0, NULL },
712 { "Manager", "LogColor", config_parse_color, 0, NULL },
713 { "Manager", "LogLocation", config_parse_location, 0, NULL },
714 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
715 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL },
716 { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL },
717 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
718 { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
719 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
720 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, &arg_cpu_affinity },
721 { "Manager", "JoinControllers", config_parse_warn_compat, DISABLED_CONFIGURATION, NULL },
722 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
723 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
724 { "Manager", "WatchdogDevice", config_parse_path, 0, &arg_watchdog_device },
725 { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
726 { "Manager", "NoNewPrivileges", config_parse_bool, 0, &arg_no_new_privs },
727 #if HAVE_SECCOMP
728 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
729 #endif
730 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
731 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
732 { "Manager", "DefaultStandardOutput", config_parse_output_restricted,0, &arg_default_std_output },
733 { "Manager", "DefaultStandardError", config_parse_output_restricted,0, &arg_default_std_error },
734 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
735 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
736 { "Manager", "DefaultTimeoutAbortSec", config_parse_timeout_abort, 0, NULL },
737 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
738 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */
739 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec, 0, &arg_default_start_limit_interval },
740 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
741 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
742 { "Manager", "DefaultLimitCPU", config_parse_rlimit, RLIMIT_CPU, arg_default_rlimit },
743 { "Manager", "DefaultLimitFSIZE", config_parse_rlimit, RLIMIT_FSIZE, arg_default_rlimit },
744 { "Manager", "DefaultLimitDATA", config_parse_rlimit, RLIMIT_DATA, arg_default_rlimit },
745 { "Manager", "DefaultLimitSTACK", config_parse_rlimit, RLIMIT_STACK, arg_default_rlimit },
746 { "Manager", "DefaultLimitCORE", config_parse_rlimit, RLIMIT_CORE, arg_default_rlimit },
747 { "Manager", "DefaultLimitRSS", config_parse_rlimit, RLIMIT_RSS, arg_default_rlimit },
748 { "Manager", "DefaultLimitNOFILE", config_parse_rlimit, RLIMIT_NOFILE, arg_default_rlimit },
749 { "Manager", "DefaultLimitAS", config_parse_rlimit, RLIMIT_AS, arg_default_rlimit },
750 { "Manager", "DefaultLimitNPROC", config_parse_rlimit, RLIMIT_NPROC, arg_default_rlimit },
751 { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit, RLIMIT_MEMLOCK, arg_default_rlimit },
752 { "Manager", "DefaultLimitLOCKS", config_parse_rlimit, RLIMIT_LOCKS, arg_default_rlimit },
753 { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit, RLIMIT_SIGPENDING, arg_default_rlimit },
754 { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit, RLIMIT_MSGQUEUE, arg_default_rlimit },
755 { "Manager", "DefaultLimitNICE", config_parse_rlimit, RLIMIT_NICE, arg_default_rlimit },
756 { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit, RLIMIT_RTPRIO, arg_default_rlimit },
757 { "Manager", "DefaultLimitRTTIME", config_parse_rlimit, RLIMIT_RTTIME, arg_default_rlimit },
758 { "Manager", "DefaultCPUAccounting", config_parse_tristate, 0, &arg_default_cpu_accounting },
759 { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting },
760 { "Manager", "DefaultIPAccounting", config_parse_bool, 0, &arg_default_ip_accounting },
761 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
762 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
763 { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
764 { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max },
765 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action, 0, &arg_cad_burst_action },
766 { "Manager", "DefaultOOMPolicy", config_parse_oom_policy, 0, &arg_default_oom_policy },
767 {}
768 };
769
770 const char *fn, *conf_dirs_nulstr;
771
772 fn = arg_system ?
773 PKGSYSCONFDIR "/system.conf" :
774 PKGSYSCONFDIR "/user.conf";
775
776 conf_dirs_nulstr = arg_system ?
777 CONF_PATHS_NULSTR("systemd/system.conf.d") :
778 CONF_PATHS_NULSTR("systemd/user.conf.d");
779
780 (void) config_parse_many_nulstr(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, CONFIG_PARSE_WARN, NULL);
781
782 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
783 * like everywhere else. */
784 if (arg_default_timeout_start_usec <= 0)
785 arg_default_timeout_start_usec = USEC_INFINITY;
786 if (arg_default_timeout_stop_usec <= 0)
787 arg_default_timeout_stop_usec = USEC_INFINITY;
788
789 return 0;
790 }
791
792 static void set_manager_defaults(Manager *m) {
793
794 assert(m);
795
796 /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
797 * affect the manager itself, but are just what newly allocated units will have set if they haven't set
798 * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
799
800 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
801 m->default_std_output = arg_default_std_output;
802 m->default_std_error = arg_default_std_error;
803 m->default_timeout_start_usec = arg_default_timeout_start_usec;
804 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
805 m->default_timeout_abort_usec = arg_default_timeout_abort_usec;
806 m->default_timeout_abort_set = arg_default_timeout_abort_set;
807 m->default_restart_usec = arg_default_restart_usec;
808 m->default_start_limit_interval = arg_default_start_limit_interval;
809 m->default_start_limit_burst = arg_default_start_limit_burst;
810
811 /* On 4.15+ with unified hierarchy, CPU accounting is essentially free as it doesn't require the CPU
812 * controller to be enabled, so the default is to enable it unless we got told otherwise. */
813 if (arg_default_cpu_accounting >= 0)
814 m->default_cpu_accounting = arg_default_cpu_accounting;
815 else
816 m->default_cpu_accounting = cpu_accounting_is_cheap();
817
818 m->default_io_accounting = arg_default_io_accounting;
819 m->default_ip_accounting = arg_default_ip_accounting;
820 m->default_blockio_accounting = arg_default_blockio_accounting;
821 m->default_memory_accounting = arg_default_memory_accounting;
822 m->default_tasks_accounting = arg_default_tasks_accounting;
823 m->default_tasks_max = arg_default_tasks_max;
824 m->default_oom_policy = arg_default_oom_policy;
825
826 (void) manager_set_default_rlimits(m, arg_default_rlimit);
827
828 (void) manager_default_environment(m);
829 (void) manager_transient_environment_add(m, arg_default_environment);
830 }
831
832 static void set_manager_settings(Manager *m) {
833
834 assert(m);
835
836 /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
837 * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
838
839 m->confirm_spawn = arg_confirm_spawn;
840 m->service_watchdogs = arg_service_watchdogs;
841 m->runtime_watchdog = arg_runtime_watchdog;
842 m->shutdown_watchdog = arg_shutdown_watchdog;
843 m->cad_burst_action = arg_cad_burst_action;
844
845 manager_set_show_status(m, arg_show_status);
846 }
847
848 static int parse_argv(int argc, char *argv[]) {
849 enum {
850 ARG_LOG_LEVEL = 0x100,
851 ARG_LOG_TARGET,
852 ARG_LOG_COLOR,
853 ARG_LOG_LOCATION,
854 ARG_UNIT,
855 ARG_SYSTEM,
856 ARG_USER,
857 ARG_TEST,
858 ARG_NO_PAGER,
859 ARG_VERSION,
860 ARG_DUMP_CONFIGURATION_ITEMS,
861 ARG_DUMP_BUS_PROPERTIES,
862 ARG_DUMP_CORE,
863 ARG_CRASH_CHVT,
864 ARG_CRASH_SHELL,
865 ARG_CRASH_REBOOT,
866 ARG_CONFIRM_SPAWN,
867 ARG_SHOW_STATUS,
868 ARG_DESERIALIZE,
869 ARG_SWITCHED_ROOT,
870 ARG_DEFAULT_STD_OUTPUT,
871 ARG_DEFAULT_STD_ERROR,
872 ARG_MACHINE_ID,
873 ARG_SERVICE_WATCHDOGS,
874 };
875
876 static const struct option options[] = {
877 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
878 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
879 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
880 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
881 { "unit", required_argument, NULL, ARG_UNIT },
882 { "system", no_argument, NULL, ARG_SYSTEM },
883 { "user", no_argument, NULL, ARG_USER },
884 { "test", no_argument, NULL, ARG_TEST },
885 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
886 { "help", no_argument, NULL, 'h' },
887 { "version", no_argument, NULL, ARG_VERSION },
888 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
889 { "dump-bus-properties", no_argument, NULL, ARG_DUMP_BUS_PROPERTIES },
890 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
891 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
892 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
893 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
894 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
895 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
896 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
897 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
898 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
899 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
900 { "machine-id", required_argument, NULL, ARG_MACHINE_ID },
901 { "service-watchdogs", required_argument, NULL, ARG_SERVICE_WATCHDOGS },
902 {}
903 };
904
905 int c, r;
906
907 assert(argc >= 1);
908 assert(argv);
909
910 if (getpid_cached() == 1)
911 opterr = 0;
912
913 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
914
915 switch (c) {
916
917 case ARG_LOG_LEVEL:
918 r = log_set_max_level_from_string(optarg);
919 if (r < 0)
920 return log_error_errno(r, "Failed to parse log level \"%s\": %m", optarg);
921
922 break;
923
924 case ARG_LOG_TARGET:
925 r = log_set_target_from_string(optarg);
926 if (r < 0)
927 return log_error_errno(r, "Failed to parse log target \"%s\": %m", optarg);
928
929 break;
930
931 case ARG_LOG_COLOR:
932
933 if (optarg) {
934 r = log_show_color_from_string(optarg);
935 if (r < 0)
936 return log_error_errno(r, "Failed to parse log color setting \"%s\": %m",
937 optarg);
938 } else
939 log_show_color(true);
940
941 break;
942
943 case ARG_LOG_LOCATION:
944 if (optarg) {
945 r = log_show_location_from_string(optarg);
946 if (r < 0)
947 return log_error_errno(r, "Failed to parse log location setting \"%s\": %m",
948 optarg);
949 } else
950 log_show_location(true);
951
952 break;
953
954 case ARG_DEFAULT_STD_OUTPUT:
955 r = exec_output_from_string(optarg);
956 if (r < 0)
957 return log_error_errno(r, "Failed to parse default standard output setting \"%s\": %m",
958 optarg);
959 arg_default_std_output = r;
960 break;
961
962 case ARG_DEFAULT_STD_ERROR:
963 r = exec_output_from_string(optarg);
964 if (r < 0)
965 return log_error_errno(r, "Failed to parse default standard error output setting \"%s\": %m",
966 optarg);
967 arg_default_std_error = r;
968 break;
969
970 case ARG_UNIT:
971 r = free_and_strdup(&arg_default_unit, optarg);
972 if (r < 0)
973 return log_error_errno(r, "Failed to set default unit \"%s\": %m", optarg);
974
975 break;
976
977 case ARG_SYSTEM:
978 arg_system = true;
979 break;
980
981 case ARG_USER:
982 arg_system = false;
983 break;
984
985 case ARG_TEST:
986 arg_action = ACTION_TEST;
987 break;
988
989 case ARG_NO_PAGER:
990 arg_pager_flags |= PAGER_DISABLE;
991 break;
992
993 case ARG_VERSION:
994 arg_action = ACTION_VERSION;
995 break;
996
997 case ARG_DUMP_CONFIGURATION_ITEMS:
998 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
999 break;
1000
1001 case ARG_DUMP_BUS_PROPERTIES:
1002 arg_action = ACTION_DUMP_BUS_PROPERTIES;
1003 break;
1004
1005 case ARG_DUMP_CORE:
1006 if (!optarg)
1007 arg_dump_core = true;
1008 else {
1009 r = parse_boolean(optarg);
1010 if (r < 0)
1011 return log_error_errno(r, "Failed to parse dump core boolean: \"%s\": %m",
1012 optarg);
1013 arg_dump_core = r;
1014 }
1015 break;
1016
1017 case ARG_CRASH_CHVT:
1018 r = parse_crash_chvt(optarg);
1019 if (r < 0)
1020 return log_error_errno(r, "Failed to parse crash virtual terminal index: \"%s\": %m",
1021 optarg);
1022 break;
1023
1024 case ARG_CRASH_SHELL:
1025 if (!optarg)
1026 arg_crash_shell = true;
1027 else {
1028 r = parse_boolean(optarg);
1029 if (r < 0)
1030 return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
1031 optarg);
1032 arg_crash_shell = r;
1033 }
1034 break;
1035
1036 case ARG_CRASH_REBOOT:
1037 if (!optarg)
1038 arg_crash_reboot = true;
1039 else {
1040 r = parse_boolean(optarg);
1041 if (r < 0)
1042 return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
1043 optarg);
1044 arg_crash_reboot = r;
1045 }
1046 break;
1047
1048 case ARG_CONFIRM_SPAWN:
1049 arg_confirm_spawn = mfree(arg_confirm_spawn);
1050
1051 r = parse_confirm_spawn(optarg, &arg_confirm_spawn);
1052 if (r < 0)
1053 return log_error_errno(r, "Failed to parse confirm spawn option: \"%s\": %m",
1054 optarg);
1055 break;
1056
1057 case ARG_SERVICE_WATCHDOGS:
1058 r = parse_boolean(optarg);
1059 if (r < 0)
1060 return log_error_errno(r, "Failed to parse service watchdogs boolean: \"%s\": %m",
1061 optarg);
1062 arg_service_watchdogs = r;
1063 break;
1064
1065 case ARG_SHOW_STATUS:
1066 if (optarg) {
1067 r = parse_show_status(optarg, &arg_show_status);
1068 if (r < 0)
1069 return log_error_errno(r, "Failed to parse show status boolean: \"%s\": %m",
1070 optarg);
1071 } else
1072 arg_show_status = SHOW_STATUS_YES;
1073 break;
1074
1075 case ARG_DESERIALIZE: {
1076 int fd;
1077 FILE *f;
1078
1079 r = safe_atoi(optarg, &fd);
1080 if (r < 0)
1081 log_error_errno(r, "Failed to parse deserialize option \"%s\": %m", optarg);
1082 if (fd < 0)
1083 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1084 "Invalid deserialize fd: %d",
1085 fd);
1086
1087 (void) fd_cloexec(fd, true);
1088
1089 f = fdopen(fd, "r");
1090 if (!f)
1091 return log_error_errno(errno, "Failed to open serialization fd %d: %m", fd);
1092
1093 safe_fclose(arg_serialization);
1094 arg_serialization = f;
1095
1096 break;
1097 }
1098
1099 case ARG_SWITCHED_ROOT:
1100 arg_switched_root = true;
1101 break;
1102
1103 case ARG_MACHINE_ID:
1104 r = set_machine_id(optarg);
1105 if (r < 0)
1106 return log_error_errno(r, "MachineID '%s' is not valid: %m", optarg);
1107 break;
1108
1109 case 'h':
1110 arg_action = ACTION_HELP;
1111 break;
1112
1113 case 'D':
1114 log_set_max_level(LOG_DEBUG);
1115 break;
1116
1117 case 'b':
1118 case 's':
1119 case 'z':
1120 /* Just to eat away the sysvinit kernel
1121 * cmdline args without getopt() error
1122 * messages that we'll parse in
1123 * parse_proc_cmdline_word() or ignore. */
1124
1125 case '?':
1126 if (getpid_cached() != 1)
1127 return -EINVAL;
1128 else
1129 return 0;
1130
1131 default:
1132 assert_not_reached("Unhandled option code.");
1133 }
1134
1135 if (optind < argc && getpid_cached() != 1) {
1136 /* Hmm, when we aren't run as init system
1137 * let's complain about excess arguments */
1138
1139 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1140 "Excess arguments.");
1141 }
1142
1143 return 0;
1144 }
1145
1146 static int help(void) {
1147 _cleanup_free_ char *link = NULL;
1148 int r;
1149
1150 r = terminal_urlify_man("systemd", "1", &link);
1151 if (r < 0)
1152 return log_oom();
1153
1154 printf("%s [OPTIONS...]\n\n"
1155 "Starts up and maintains the system or user services.\n\n"
1156 " -h --help Show this help\n"
1157 " --version Show version\n"
1158 " --test Determine startup sequence, dump it and exit\n"
1159 " --no-pager Do not pipe output into a pager\n"
1160 " --dump-configuration-items Dump understood unit configuration items\n"
1161 " --dump-bus-properties Dump exposed bus properties\n"
1162 " --unit=UNIT Set default unit\n"
1163 " --system Run a system instance, even if PID != 1\n"
1164 " --user Run a user instance\n"
1165 " --dump-core[=BOOL] Dump core on crash\n"
1166 " --crash-vt=NR Change to specified VT on crash\n"
1167 " --crash-reboot[=BOOL] Reboot on crash\n"
1168 " --crash-shell[=BOOL] Run shell on crash\n"
1169 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1170 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1171 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1172 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1173 " --log-color[=BOOL] Highlight important log messages\n"
1174 " --log-location[=BOOL] Include code location in log messages\n"
1175 " --default-standard-output= Set default standard output for services\n"
1176 " --default-standard-error= Set default standard error output for services\n"
1177 "\nSee the %s for details.\n"
1178 , program_invocation_short_name
1179 , link
1180 );
1181
1182 return 0;
1183 }
1184
1185 static int prepare_reexecute(
1186 Manager *m,
1187 FILE **ret_f,
1188 FDSet **ret_fds,
1189 bool switching_root) {
1190
1191 _cleanup_fdset_free_ FDSet *fds = NULL;
1192 _cleanup_fclose_ FILE *f = NULL;
1193 int r;
1194
1195 assert(m);
1196 assert(ret_f);
1197 assert(ret_fds);
1198
1199 r = manager_open_serialization(m, &f);
1200 if (r < 0)
1201 return log_error_errno(r, "Failed to create serialization file: %m");
1202
1203 /* Make sure nothing is really destructed when we shut down */
1204 m->n_reloading++;
1205 bus_manager_send_reloading(m, true);
1206
1207 fds = fdset_new();
1208 if (!fds)
1209 return log_oom();
1210
1211 r = manager_serialize(m, f, fds, switching_root);
1212 if (r < 0)
1213 return r;
1214
1215 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1216 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1217
1218 r = fd_cloexec(fileno(f), false);
1219 if (r < 0)
1220 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1221
1222 r = fdset_cloexec(fds, false);
1223 if (r < 0)
1224 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1225
1226 *ret_f = TAKE_PTR(f);
1227 *ret_fds = TAKE_PTR(fds);
1228
1229 return 0;
1230 }
1231
1232 static void bump_file_max_and_nr_open(void) {
1233
1234 /* Let's bump fs.file-max and fs.nr_open to their respective maximums. On current kernels large numbers of file
1235 * descriptors are no longer a performance problem and their memory is properly tracked by memcg, thus counting
1236 * them and limiting them in another two layers of limits is unnecessary and just complicates things. This
1237 * function hence turns off 2 of the 4 levels of limits on file descriptors, and makes RLIMIT_NOLIMIT (soft +
1238 * hard) the only ones that really matter. */
1239
1240 #if BUMP_PROC_SYS_FS_FILE_MAX || BUMP_PROC_SYS_FS_NR_OPEN
1241 _cleanup_free_ char *t = NULL;
1242 int r;
1243 #endif
1244
1245 #if BUMP_PROC_SYS_FS_FILE_MAX
1246 /* I so wanted to use STRINGIFY(ULONG_MAX) here, but alas we can't as glibc/gcc define that as
1247 * "(0x7fffffffffffffffL * 2UL + 1UL)". Seriously. 😢 */
1248 if (asprintf(&t, "%lu\n", ULONG_MAX) < 0) {
1249 log_oom();
1250 return;
1251 }
1252
1253 r = sysctl_write("fs/file-max", t);
1254 if (r < 0)
1255 log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.file-max, ignoring: %m");
1256 #endif
1257
1258 #if BUMP_PROC_SYS_FS_FILE_MAX && BUMP_PROC_SYS_FS_NR_OPEN
1259 t = mfree(t);
1260 #endif
1261
1262 #if BUMP_PROC_SYS_FS_NR_OPEN
1263 int v = INT_MAX;
1264
1265 /* Arg! The kernel enforces maximum and minimum values on the fs.nr_open, but we don't really know what they
1266 * are. The expression by which the maximum is determined is dependent on the architecture, and is something we
1267 * don't really want to copy to userspace, as it is dependent on implementation details of the kernel. Since
1268 * the kernel doesn't expose the maximum value to us, we can only try and hope. Hence, let's start with
1269 * INT_MAX, and then keep halving the value until we find one that works. Ugly? Yes, absolutely, but kernel
1270 * APIs are kernel APIs, so what do can we do... 🤯 */
1271
1272 for (;;) {
1273 int k;
1274
1275 v &= ~(__SIZEOF_POINTER__ - 1); /* Round down to next multiple of the pointer size */
1276 if (v < 1024) {
1277 log_warning("Can't bump fs.nr_open, value too small.");
1278 break;
1279 }
1280
1281 k = read_nr_open();
1282 if (k < 0) {
1283 log_error_errno(k, "Failed to read fs.nr_open: %m");
1284 break;
1285 }
1286 if (k >= v) { /* Already larger */
1287 log_debug("Skipping bump, value is already larger.");
1288 break;
1289 }
1290
1291 if (asprintf(&t, "%i\n", v) < 0) {
1292 log_oom();
1293 return;
1294 }
1295
1296 r = sysctl_write("fs/nr_open", t);
1297 t = mfree(t);
1298 if (r == -EINVAL) {
1299 log_debug("Couldn't write fs.nr_open as %i, halving it.", v);
1300 v /= 2;
1301 continue;
1302 }
1303 if (r < 0) {
1304 log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.nr_open, ignoring: %m");
1305 break;
1306 }
1307
1308 log_debug("Successfully bumped fs.nr_open to %i", v);
1309 break;
1310 }
1311 #endif
1312 }
1313
1314 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1315 struct rlimit new_rlimit;
1316 int r, nr;
1317
1318 assert(saved_rlimit);
1319
1320 /* Save the original RLIMIT_NOFILE so that we can reset it later when transitioning from the initrd to the main
1321 * systemd or suchlike. */
1322 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1323 return log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1324
1325 /* Get the underlying absolute limit the kernel enforces */
1326 nr = read_nr_open();
1327
1328 /* Make sure forked processes get limits based on the original kernel setting */
1329 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1330 struct rlimit *rl;
1331
1332 rl = newdup(struct rlimit, saved_rlimit, 1);
1333 if (!rl)
1334 return log_oom();
1335
1336 /* Bump the hard limit for system services to a substantially higher value. The default hard limit
1337 * current kernels set is pretty low (4K), mostly for historical reasons. According to kernel
1338 * developers, the fd handling in recent kernels has been optimized substantially enough, so that we
1339 * can bump the limit now, without paying too high a price in memory or performance. Note however that
1340 * we only bump the hard limit, not the soft limit. That's because select() works the way it works, and
1341 * chokes on fds >= 1024. If we'd bump the soft limit globally, it might accidentally happen to
1342 * unexpecting programs that they get fds higher than what they can process using select(). By only
1343 * bumping the hard limit but leaving the low limit as it is we avoid this pitfall: programs that are
1344 * written by folks aware of the select() problem in mind (and thus use poll()/epoll instead of
1345 * select(), the way everybody should) can explicitly opt into high fds by bumping their soft limit
1346 * beyond 1024, to the hard limit we pass. */
1347 if (arg_system)
1348 rl->rlim_max = MIN((rlim_t) nr, MAX(rl->rlim_max, (rlim_t) HIGH_RLIMIT_NOFILE));
1349
1350 /* If for some reason we were invoked with a soft limit above 1024 (which should never
1351 * happen!, but who knows what we get passed in from pam_limit when invoked as --user
1352 * instance), then lower what we pass on to not confuse our children */
1353 rl->rlim_cur = MIN(rl->rlim_cur, (rlim_t) FD_SETSIZE);
1354
1355 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1356 }
1357
1358 /* Calculate the new limits to use for us. Never lower from what we inherited. */
1359 new_rlimit = (struct rlimit) {
1360 .rlim_cur = MAX((rlim_t) nr, saved_rlimit->rlim_cur),
1361 .rlim_max = MAX((rlim_t) nr, saved_rlimit->rlim_max),
1362 };
1363
1364 /* Shortcut if nothing changes. */
1365 if (saved_rlimit->rlim_max >= new_rlimit.rlim_max &&
1366 saved_rlimit->rlim_cur >= new_rlimit.rlim_cur) {
1367 log_debug("RLIMIT_NOFILE is already as high or higher than we need it, not bumping.");
1368 return 0;
1369 }
1370
1371 /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for
1372 * both hard and soft. */
1373 r = setrlimit_closest(RLIMIT_NOFILE, &new_rlimit);
1374 if (r < 0)
1375 return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1376
1377 return 0;
1378 }
1379
1380 static int bump_rlimit_memlock(struct rlimit *saved_rlimit) {
1381 struct rlimit new_rlimit;
1382 int r;
1383
1384 assert(saved_rlimit);
1385
1386 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK which should
1387 * normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's bump
1388 * the value high enough for our user. */
1389
1390 if (getrlimit(RLIMIT_MEMLOCK, saved_rlimit) < 0)
1391 return log_warning_errno(errno, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1392
1393 /* Pass the original value down to invoked processes */
1394 if (!arg_default_rlimit[RLIMIT_MEMLOCK]) {
1395 struct rlimit *rl;
1396
1397 rl = newdup(struct rlimit, saved_rlimit, 1);
1398 if (!rl)
1399 return log_oom();
1400
1401 arg_default_rlimit[RLIMIT_MEMLOCK] = rl;
1402 }
1403
1404 /* Using MAX() on resource limits only is safe if RLIM_INFINITY is > 0. POSIX declares that rlim_t
1405 * must be unsigned, hence this is a given, but let's make this clear here. */
1406 assert_cc(RLIM_INFINITY > 0);
1407
1408 new_rlimit = (struct rlimit) {
1409 .rlim_cur = MAX(HIGH_RLIMIT_MEMLOCK, saved_rlimit->rlim_cur),
1410 .rlim_max = MAX(HIGH_RLIMIT_MEMLOCK, saved_rlimit->rlim_max),
1411 };
1412
1413 if (saved_rlimit->rlim_max >= new_rlimit.rlim_cur &&
1414 saved_rlimit->rlim_cur >= new_rlimit.rlim_max) {
1415 log_debug("RLIMIT_MEMLOCK is already as high or higher than we need it, not bumping.");
1416 return 0;
1417 }
1418
1419 r = setrlimit_closest(RLIMIT_MEMLOCK, &new_rlimit);
1420 if (r < 0)
1421 return log_warning_errno(r, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1422
1423 return 0;
1424 }
1425
1426 static void test_usr(void) {
1427
1428 /* Check that /usr is either on the same file system as / or mounted already. */
1429
1430 if (dir_is_empty("/usr") <= 0)
1431 return;
1432
1433 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1434 "Some things will probably break (sometimes even silently) in mysterious ways. "
1435 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1436 }
1437
1438 static int enforce_syscall_archs(Set *archs) {
1439 #if HAVE_SECCOMP
1440 int r;
1441
1442 if (!is_seccomp_available())
1443 return 0;
1444
1445 r = seccomp_restrict_archs(arg_syscall_archs);
1446 if (r < 0)
1447 return log_error_errno(r, "Failed to enforce system call architecture restrication: %m");
1448 #endif
1449 return 0;
1450 }
1451
1452 static int status_welcome(void) {
1453 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1454 int r;
1455
1456 if (IN_SET(arg_show_status, SHOW_STATUS_NO, SHOW_STATUS_AUTO))
1457 return 0;
1458
1459 r = parse_os_release(NULL,
1460 "PRETTY_NAME", &pretty_name,
1461 "ANSI_COLOR", &ansi_color,
1462 NULL);
1463 if (r < 0)
1464 log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r,
1465 "Failed to read os-release file, ignoring: %m");
1466
1467 if (log_get_show_color())
1468 return status_printf(NULL, 0,
1469 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1470 isempty(ansi_color) ? "1" : ansi_color,
1471 isempty(pretty_name) ? "Linux" : pretty_name);
1472 else
1473 return status_printf(NULL, 0,
1474 "\nWelcome to %s!\n",
1475 isempty(pretty_name) ? "Linux" : pretty_name);
1476 }
1477
1478 static int write_container_id(void) {
1479 const char *c;
1480 int r;
1481
1482 c = getenv("container");
1483 if (isempty(c))
1484 return 0;
1485
1486 RUN_WITH_UMASK(0022)
1487 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1488 if (r < 0)
1489 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
1490
1491 return 1;
1492 }
1493
1494 static int bump_unix_max_dgram_qlen(void) {
1495 _cleanup_free_ char *qlen = NULL;
1496 unsigned long v;
1497 int r;
1498
1499 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1500 * really really early during boot, so that it is actually applied to all our sockets, including the
1501 * $NOTIFY_SOCKET one. */
1502
1503 r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1504 if (r < 0)
1505 return log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1506
1507 r = safe_atolu(qlen, &v);
1508 if (r < 0)
1509 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen);
1510
1511 if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1512 return 0;
1513
1514 r = write_string_filef("/proc/sys/net/unix/max_dgram_qlen", WRITE_STRING_FILE_DISABLE_BUFFER, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN);
1515 if (r < 0)
1516 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1517 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1518
1519 return 1;
1520 }
1521
1522 static int fixup_environment(void) {
1523 _cleanup_free_ char *term = NULL;
1524 const char *t;
1525 int r;
1526
1527 /* Only fix up the environment when we are started as PID 1 */
1528 if (getpid_cached() != 1)
1529 return 0;
1530
1531 /* We expect the environment to be set correctly if run inside a container. */
1532 if (detect_container() > 0)
1533 return 0;
1534
1535 /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1536 * device used by the console. We try to make a better guess here since some consoles might not have support
1537 * for color mode for example.
1538 *
1539 * However if TERM was configured through the kernel command line then leave it alone. */
1540 r = proc_cmdline_get_key("TERM", 0, &term);
1541 if (r < 0)
1542 return r;
1543
1544 t = term ?: default_term_for_tty("/dev/console");
1545
1546 if (setenv("TERM", t, 1) < 0)
1547 return -errno;
1548
1549 return 0;
1550 }
1551
1552 static void redirect_telinit(int argc, char *argv[]) {
1553
1554 /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1555
1556 #if HAVE_SYSV_COMPAT
1557 if (getpid_cached() == 1)
1558 return;
1559
1560 if (!strstr(program_invocation_short_name, "init"))
1561 return;
1562
1563 execv(SYSTEMCTL_BINARY_PATH, argv);
1564 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1565 exit(EXIT_FAILURE);
1566 #endif
1567 }
1568
1569 static int become_shutdown(
1570 const char *shutdown_verb,
1571 int retval) {
1572
1573 char log_level[DECIMAL_STR_MAX(int) + 1],
1574 exit_code[DECIMAL_STR_MAX(uint8_t) + 1],
1575 timeout[DECIMAL_STR_MAX(usec_t) + 1];
1576
1577 const char* command_line[13] = {
1578 SYSTEMD_SHUTDOWN_BINARY_PATH,
1579 shutdown_verb,
1580 "--timeout", timeout,
1581 "--log-level", log_level,
1582 "--log-target",
1583 };
1584
1585 _cleanup_strv_free_ char **env_block = NULL;
1586 size_t pos = 7;
1587 int r;
1588
1589 assert(shutdown_verb);
1590 assert(!command_line[pos]);
1591 env_block = strv_copy(environ);
1592
1593 xsprintf(log_level, "%d", log_get_max_level());
1594 xsprintf(timeout, "%" PRI_USEC "us", arg_default_timeout_stop_usec);
1595
1596 switch (log_get_target()) {
1597
1598 case LOG_TARGET_KMSG:
1599 case LOG_TARGET_JOURNAL_OR_KMSG:
1600 case LOG_TARGET_SYSLOG_OR_KMSG:
1601 command_line[pos++] = "kmsg";
1602 break;
1603
1604 case LOG_TARGET_NULL:
1605 command_line[pos++] = "null";
1606 break;
1607
1608 case LOG_TARGET_CONSOLE:
1609 default:
1610 command_line[pos++] = "console";
1611 break;
1612 };
1613
1614 if (log_get_show_color())
1615 command_line[pos++] = "--log-color";
1616
1617 if (log_get_show_location())
1618 command_line[pos++] = "--log-location";
1619
1620 if (streq(shutdown_verb, "exit")) {
1621 command_line[pos++] = "--exit-code";
1622 command_line[pos++] = exit_code;
1623 xsprintf(exit_code, "%d", retval);
1624 }
1625
1626 assert(pos < ELEMENTSOF(command_line));
1627
1628 if (streq(shutdown_verb, "reboot") &&
1629 arg_shutdown_watchdog > 0 &&
1630 arg_shutdown_watchdog != USEC_INFINITY) {
1631
1632 char *e;
1633
1634 /* If we reboot let's set the shutdown
1635 * watchdog and tell the shutdown binary to
1636 * repeatedly ping it */
1637 r = watchdog_set_timeout(&arg_shutdown_watchdog);
1638 watchdog_close(r < 0);
1639
1640 /* Tell the binary how often to ping, ignore failure */
1641 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
1642 (void) strv_consume(&env_block, e);
1643
1644 if (arg_watchdog_device &&
1645 asprintf(&e, "WATCHDOG_DEVICE=%s", arg_watchdog_device) > 0)
1646 (void) strv_consume(&env_block, e);
1647 } else
1648 watchdog_close(true);
1649
1650 /* Avoid the creation of new processes forked by the
1651 * kernel; at this point, we will not listen to the
1652 * signals anyway */
1653 if (detect_container() <= 0)
1654 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
1655
1656 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
1657 return -errno;
1658 }
1659
1660 static void initialize_clock(void) {
1661 int r;
1662
1663 if (clock_is_localtime(NULL) > 0) {
1664 int min;
1665
1666 /*
1667 * The very first call of settimeofday() also does a time warp in the kernel.
1668 *
1669 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1670 * of maintaining the RTC and do all adjustments. This matches the behavior of Windows, which leaves
1671 * the RTC alone if the registry tells that the RTC runs in UTC.
1672 */
1673 r = clock_set_timezone(&min);
1674 if (r < 0)
1675 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1676 else
1677 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1678
1679 } else if (!in_initrd()) {
1680 /*
1681 * Do a dummy very first call to seal the kernel's time warp magic.
1682 *
1683 * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1684 * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1685 * until we reach the real system.
1686 *
1687 * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1688 * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1689 * be treated as UTC that way.
1690 */
1691 (void) clock_reset_timewarp();
1692 }
1693
1694 r = clock_apply_epoch();
1695 if (r < 0)
1696 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1697 else if (r > 0)
1698 log_info("System time before build time, advancing clock.");
1699 }
1700
1701 static void initialize_coredump(bool skip_setup) {
1702 #if ENABLE_COREDUMP
1703 if (getpid_cached() != 1)
1704 return;
1705
1706 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1707 * will process core dumps for system services by default. */
1708 if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0)
1709 log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m");
1710
1711 /* But at the same time, turn off the core_pattern logic by default, so that no
1712 * coredumps are stored until the systemd-coredump tool is enabled via
1713 * sysctl. However it can be changed via the kernel command line later so core
1714 * dumps can still be generated during early startup and in initramfs. */
1715 if (!skip_setup)
1716 disable_coredumps();
1717 #endif
1718 }
1719
1720 static void initialize_core_pattern(bool skip_setup) {
1721 int r;
1722
1723 if (skip_setup || !arg_early_core_pattern)
1724 return;
1725
1726 if (getpid_cached() != 1)
1727 return;
1728
1729 r = write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern, WRITE_STRING_FILE_DISABLE_BUFFER);
1730 if (r < 0)
1731 log_warning_errno(r, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", arg_early_core_pattern);
1732 }
1733
1734 static void update_cpu_affinity(bool skip_setup) {
1735 _cleanup_free_ char *mask = NULL;
1736
1737 if (skip_setup || !arg_cpu_affinity.set)
1738 return;
1739
1740 assert(arg_cpu_affinity.allocated > 0);
1741
1742 mask = cpu_set_to_string(&arg_cpu_affinity);
1743 log_debug("Setting CPU affinity to %s.", strnull(mask));
1744
1745 if (sched_setaffinity(0, arg_cpu_affinity.allocated, arg_cpu_affinity.set) < 0)
1746 log_warning_errno(errno, "Failed to set CPU affinity: %m");
1747 }
1748
1749 static void do_reexecute(
1750 int argc,
1751 char *argv[],
1752 const struct rlimit *saved_rlimit_nofile,
1753 const struct rlimit *saved_rlimit_memlock,
1754 FDSet *fds,
1755 const char *switch_root_dir,
1756 const char *switch_root_init,
1757 const char **ret_error_message) {
1758
1759 unsigned i, j, args_size;
1760 const char **args;
1761 int r;
1762
1763 assert(saved_rlimit_nofile);
1764 assert(saved_rlimit_memlock);
1765 assert(ret_error_message);
1766
1767 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1768 * we do that */
1769 watchdog_close(true);
1770
1771 /* Reset RLIMIT_NOFILE + RLIMIT_MEMLOCK back to the kernel defaults, so that the new systemd can pass
1772 * the kernel default to its child processes */
1773 if (saved_rlimit_nofile->rlim_cur != 0)
1774 (void) setrlimit(RLIMIT_NOFILE, saved_rlimit_nofile);
1775 if (saved_rlimit_memlock->rlim_cur != RLIM_INFINITY)
1776 (void) setrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock);
1777
1778 if (switch_root_dir) {
1779 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1780 * SIGCHLD for them after deserializing. */
1781 broadcast_signal(SIGTERM, false, true, arg_default_timeout_stop_usec);
1782
1783 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1784 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1785 if (r < 0)
1786 log_error_errno(r, "Failed to switch root, trying to continue: %m");
1787 }
1788
1789 args_size = MAX(6, argc+1);
1790 args = newa(const char*, args_size);
1791
1792 if (!switch_root_init) {
1793 char sfd[DECIMAL_STR_MAX(int) + 1];
1794
1795 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1796 * the user didn't specify an explicit init to spawn. */
1797
1798 assert(arg_serialization);
1799 assert(fds);
1800
1801 xsprintf(sfd, "%i", fileno(arg_serialization));
1802
1803 i = 0;
1804 args[i++] = SYSTEMD_BINARY_PATH;
1805 if (switch_root_dir)
1806 args[i++] = "--switched-root";
1807 args[i++] = arg_system ? "--system" : "--user";
1808 args[i++] = "--deserialize";
1809 args[i++] = sfd;
1810 args[i++] = NULL;
1811
1812 assert(i <= args_size);
1813
1814 /*
1815 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do this is on
1816 * its own on exec(), but it will do it on exit(). Hence, to ensure we get a summary here, fork() off
1817 * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1818 * before proceeding into the exec().
1819 */
1820 valgrind_summary_hack();
1821
1822 (void) execv(args[0], (char* const*) args);
1823 log_debug_errno(errno, "Failed to execute our own binary, trying fallback: %m");
1824 }
1825
1826 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1827 * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1828 * doesn't matter.) */
1829
1830 arg_serialization = safe_fclose(arg_serialization);
1831 fds = fdset_free(fds);
1832
1833 /* Reopen the console */
1834 (void) make_console_stdio();
1835
1836 for (j = 1, i = 1; j < (unsigned) argc; j++)
1837 args[i++] = argv[j];
1838 args[i++] = NULL;
1839 assert(i <= args_size);
1840
1841 /* Re-enable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1842 (void) reset_all_signal_handlers();
1843 (void) reset_signal_mask();
1844 (void) rlimit_nofile_safe();
1845
1846 if (switch_root_init) {
1847 args[0] = switch_root_init;
1848 (void) execv(args[0], (char* const*) args);
1849 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
1850 }
1851
1852 args[0] = "/sbin/init";
1853 (void) execv(args[0], (char* const*) args);
1854 r = -errno;
1855
1856 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
1857 ANSI_HIGHLIGHT_RED " !! " ANSI_NORMAL,
1858 "Failed to execute /sbin/init");
1859
1860 if (r == -ENOENT) {
1861 log_warning("No /sbin/init, trying fallback");
1862
1863 args[0] = "/bin/sh";
1864 args[1] = NULL;
1865 (void) execv(args[0], (char* const*) args);
1866 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
1867 } else
1868 log_warning_errno(r, "Failed to execute /sbin/init, giving up: %m");
1869
1870 *ret_error_message = "Failed to execute fallback shell";
1871 }
1872
1873 static int invoke_main_loop(
1874 Manager *m,
1875 bool *ret_reexecute,
1876 int *ret_retval, /* Return parameters relevant for shutting down */
1877 const char **ret_shutdown_verb, /* … */
1878 FDSet **ret_fds, /* Return parameters for reexecuting */
1879 char **ret_switch_root_dir, /* … */
1880 char **ret_switch_root_init, /* … */
1881 const char **ret_error_message) {
1882
1883 int r;
1884
1885 assert(m);
1886 assert(ret_reexecute);
1887 assert(ret_retval);
1888 assert(ret_shutdown_verb);
1889 assert(ret_fds);
1890 assert(ret_switch_root_dir);
1891 assert(ret_switch_root_init);
1892 assert(ret_error_message);
1893
1894 for (;;) {
1895 r = manager_loop(m);
1896 if (r < 0) {
1897 *ret_error_message = "Failed to run main loop";
1898 return log_emergency_errno(r, "Failed to run main loop: %m");
1899 }
1900
1901 switch ((ManagerObjective) r) {
1902
1903 case MANAGER_RELOAD: {
1904 LogTarget saved_log_target;
1905 int saved_log_level;
1906
1907 log_info("Reloading.");
1908
1909 /* First, save any overridden log level/target, then parse the configuration file, which might
1910 * change the log level to new settings. */
1911
1912 saved_log_level = m->log_level_overridden ? log_get_max_level() : -1;
1913 saved_log_target = m->log_target_overridden ? log_get_target() : _LOG_TARGET_INVALID;
1914
1915 (void) parse_configuration();
1916
1917 set_manager_defaults(m);
1918
1919 update_cpu_affinity(false);
1920
1921 if (saved_log_level >= 0)
1922 manager_override_log_level(m, saved_log_level);
1923 if (saved_log_target >= 0)
1924 manager_override_log_target(m, saved_log_target);
1925
1926 r = manager_reload(m);
1927 if (r < 0)
1928 /* Reloading failed before the point of no return. Let's continue running as if nothing happened. */
1929 m->objective = MANAGER_OK;
1930
1931 break;
1932 }
1933
1934 case MANAGER_REEXECUTE:
1935
1936 r = prepare_reexecute(m, &arg_serialization, ret_fds, false);
1937 if (r < 0) {
1938 *ret_error_message = "Failed to prepare for reexecution";
1939 return r;
1940 }
1941
1942 log_notice("Reexecuting.");
1943
1944 *ret_reexecute = true;
1945 *ret_retval = EXIT_SUCCESS;
1946 *ret_shutdown_verb = NULL;
1947 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1948
1949 return 0;
1950
1951 case MANAGER_SWITCH_ROOT:
1952 if (!m->switch_root_init) {
1953 r = prepare_reexecute(m, &arg_serialization, ret_fds, true);
1954 if (r < 0) {
1955 *ret_error_message = "Failed to prepare for reexecution";
1956 return r;
1957 }
1958 } else
1959 *ret_fds = NULL;
1960
1961 log_notice("Switching root.");
1962
1963 *ret_reexecute = true;
1964 *ret_retval = EXIT_SUCCESS;
1965 *ret_shutdown_verb = NULL;
1966
1967 /* Steal the switch root parameters */
1968 *ret_switch_root_dir = TAKE_PTR(m->switch_root);
1969 *ret_switch_root_init = TAKE_PTR(m->switch_root_init);
1970
1971 return 0;
1972
1973 case MANAGER_EXIT:
1974
1975 if (MANAGER_IS_USER(m)) {
1976 log_debug("Exit.");
1977
1978 *ret_reexecute = false;
1979 *ret_retval = m->return_value;
1980 *ret_shutdown_verb = NULL;
1981 *ret_fds = NULL;
1982 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1983
1984 return 0;
1985 }
1986
1987 _fallthrough_;
1988 case MANAGER_REBOOT:
1989 case MANAGER_POWEROFF:
1990 case MANAGER_HALT:
1991 case MANAGER_KEXEC: {
1992 static const char * const table[_MANAGER_OBJECTIVE_MAX] = {
1993 [MANAGER_EXIT] = "exit",
1994 [MANAGER_REBOOT] = "reboot",
1995 [MANAGER_POWEROFF] = "poweroff",
1996 [MANAGER_HALT] = "halt",
1997 [MANAGER_KEXEC] = "kexec",
1998 };
1999
2000 log_notice("Shutting down.");
2001
2002 *ret_reexecute = false;
2003 *ret_retval = m->return_value;
2004 assert_se(*ret_shutdown_verb = table[m->objective]);
2005 *ret_fds = NULL;
2006 *ret_switch_root_dir = *ret_switch_root_init = NULL;
2007
2008 return 0;
2009 }
2010
2011 default:
2012 assert_not_reached("Unknown or unexpected manager objective.");
2013 }
2014 }
2015 }
2016
2017 static void log_execution_mode(bool *ret_first_boot) {
2018 assert(ret_first_boot);
2019
2020 if (arg_system) {
2021 int v;
2022
2023 log_info("systemd " GIT_VERSION " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
2024 arg_action == ACTION_TEST ? "test " : "" );
2025
2026 v = detect_virtualization();
2027 if (v > 0)
2028 log_info("Detected virtualization %s.", virtualization_to_string(v));
2029
2030 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
2031
2032 if (in_initrd()) {
2033 *ret_first_boot = false;
2034 log_info("Running in initial RAM disk.");
2035 } else {
2036 /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
2037 * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
2038 * doesn't it's unpopulated. This allows container managers and installers to provision a
2039 * couple of files already. If the container manager wants to provision the machine ID itself
2040 * it should pass $container_uuid to PID 1. */
2041
2042 *ret_first_boot = access("/etc/machine-id", F_OK) < 0;
2043 if (*ret_first_boot)
2044 log_info("Running with unpopulated /etc.");
2045 }
2046 } else {
2047 if (DEBUG_LOGGING) {
2048 _cleanup_free_ char *t;
2049
2050 t = uid_to_name(getuid());
2051 log_debug("systemd " GIT_VERSION " running in %suser mode for user " UID_FMT "/%s. (" SYSTEMD_FEATURES ")",
2052 arg_action == ACTION_TEST ? " test" : "", getuid(), strna(t));
2053 }
2054
2055 *ret_first_boot = false;
2056 }
2057 }
2058
2059 static int initialize_runtime(
2060 bool skip_setup,
2061 struct rlimit *saved_rlimit_nofile,
2062 struct rlimit *saved_rlimit_memlock,
2063 const char **ret_error_message) {
2064
2065 int r;
2066
2067 assert(ret_error_message);
2068
2069 /* Sets up various runtime parameters. Many of these initializations are conditionalized:
2070 *
2071 * - Some only apply to --system instances
2072 * - Some only apply to --user instances
2073 * - Some only apply when we first start up, but not when we reexecute
2074 */
2075
2076 if (arg_action != ACTION_RUN)
2077 return 0;
2078
2079 update_cpu_affinity(skip_setup);
2080
2081 if (arg_system) {
2082 /* Make sure we leave a core dump without panicking the kernel. */
2083 install_crash_handler();
2084
2085 if (!skip_setup) {
2086 r = mount_cgroup_controllers();
2087 if (r < 0) {
2088 *ret_error_message = "Failed to mount cgroup hierarchies";
2089 return r;
2090 }
2091
2092 status_welcome();
2093 hostname_setup();
2094 machine_id_setup(NULL, arg_machine_id, NULL);
2095 loopback_setup();
2096 bump_unix_max_dgram_qlen();
2097 bump_file_max_and_nr_open();
2098 test_usr();
2099 write_container_id();
2100 }
2101
2102 if (arg_watchdog_device) {
2103 r = watchdog_set_device(arg_watchdog_device);
2104 if (r < 0)
2105 log_warning_errno(r, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device);
2106 }
2107
2108 if (arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY)
2109 watchdog_set_timeout(&arg_runtime_watchdog);
2110 }
2111
2112 if (arg_timer_slack_nsec != NSEC_INFINITY)
2113 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
2114 log_warning_errno(errno, "Failed to adjust timer slack, ignoring: %m");
2115
2116 if (arg_system && !cap_test_all(arg_capability_bounding_set)) {
2117 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
2118 if (r < 0) {
2119 *ret_error_message = "Failed to drop capability bounding set of usermode helpers";
2120 return log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
2121 }
2122
2123 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
2124 if (r < 0) {
2125 *ret_error_message = "Failed to drop capability bounding set";
2126 return log_emergency_errno(r, "Failed to drop capability bounding set: %m");
2127 }
2128 }
2129
2130 if (arg_system && arg_no_new_privs) {
2131 if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) {
2132 *ret_error_message = "Failed to disable new privileges";
2133 return log_emergency_errno(errno, "Failed to disable new privileges: %m");
2134 }
2135 }
2136
2137 if (arg_syscall_archs) {
2138 r = enforce_syscall_archs(arg_syscall_archs);
2139 if (r < 0) {
2140 *ret_error_message = "Failed to set syscall architectures";
2141 return r;
2142 }
2143 }
2144
2145 if (!arg_system)
2146 /* Become reaper of our children */
2147 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
2148 log_warning_errno(errno, "Failed to make us a subreaper: %m");
2149
2150 /* Bump up RLIMIT_NOFILE for systemd itself */
2151 (void) bump_rlimit_nofile(saved_rlimit_nofile);
2152 (void) bump_rlimit_memlock(saved_rlimit_memlock);
2153
2154 return 0;
2155 }
2156
2157 static int do_queue_default_job(
2158 Manager *m,
2159 const char **ret_error_message) {
2160
2161 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
2162 Job *default_unit_job;
2163 Unit *target = NULL;
2164 int r;
2165
2166 log_debug("Activating default unit: %s", arg_default_unit);
2167
2168 r = manager_load_startable_unit_or_warn(m, arg_default_unit, NULL, &target);
2169 if (r < 0) {
2170 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET);
2171
2172 r = manager_load_startable_unit_or_warn(m, SPECIAL_RESCUE_TARGET, NULL, &target);
2173 if (r < 0) {
2174 *ret_error_message = r == -ERFKILL ? "Rescue target masked"
2175 : "Failed to load rescue target";
2176 return r;
2177 }
2178 }
2179
2180 assert(target->load_state == UNIT_LOADED);
2181
2182 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, NULL, &error, &default_unit_job);
2183 if (r == -EPERM) {
2184 log_debug_errno(r, "Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
2185
2186 sd_bus_error_free(&error);
2187
2188 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, NULL, &error, &default_unit_job);
2189 if (r < 0) {
2190 *ret_error_message = "Failed to start default target";
2191 return log_emergency_errno(r, "Failed to start default target: %s", bus_error_message(&error, r));
2192 }
2193
2194 } else if (r < 0) {
2195 *ret_error_message = "Failed to isolate default target";
2196 return log_emergency_errno(r, "Failed to isolate default target: %s", bus_error_message(&error, r));
2197 }
2198
2199 m->default_unit_job_id = default_unit_job->id;
2200
2201 return 0;
2202 }
2203
2204 static void free_arguments(void) {
2205
2206 /* Frees all arg_* variables, with the exception of arg_serialization */
2207 rlimit_free_all(arg_default_rlimit);
2208
2209 arg_default_unit = mfree(arg_default_unit);
2210 arg_confirm_spawn = mfree(arg_confirm_spawn);
2211 arg_default_environment = strv_free(arg_default_environment);
2212 arg_syscall_archs = set_free(arg_syscall_archs);
2213
2214 cpu_set_reset(&arg_cpu_affinity);
2215 }
2216
2217 static int parse_configuration(void) {
2218 int r;
2219
2220 arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U);
2221
2222 r = parse_config_file();
2223 if (r < 0)
2224 log_warning_errno(r, "Failed to parse config file, ignoring: %m");
2225
2226 if (arg_system) {
2227 r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
2228 if (r < 0)
2229 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
2230 }
2231
2232 /* Note that this also parses bits from the kernel command line, including "debug". */
2233 log_parse_environment();
2234
2235 return 0;
2236 }
2237
2238 static int load_configuration(int argc, char **argv, const char **ret_error_message) {
2239 int r;
2240
2241 assert(ret_error_message);
2242
2243 (void) parse_configuration();
2244
2245 r = parse_argv(argc, argv);
2246 if (r < 0) {
2247 *ret_error_message = "Failed to parse commandline arguments";
2248 return r;
2249 }
2250
2251 /* Initialize default unit */
2252 if (!arg_default_unit) {
2253 arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET);
2254 if (!arg_default_unit) {
2255 *ret_error_message = "Failed to set default unit";
2256 return log_oom();
2257 }
2258 }
2259
2260 /* Initialize the show status setting if it hasn't been set explicitly yet */
2261 if (arg_show_status == _SHOW_STATUS_INVALID)
2262 arg_show_status = SHOW_STATUS_YES;
2263
2264 return 0;
2265 }
2266
2267 static int safety_checks(void) {
2268
2269 if (getpid_cached() == 1 &&
2270 arg_action != ACTION_RUN)
2271 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2272 "Unsupported execution mode while PID 1.");
2273
2274 if (getpid_cached() == 1 &&
2275 !arg_system)
2276 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2277 "Can't run --user mode as PID 1.");
2278
2279 if (arg_action == ACTION_RUN &&
2280 arg_system &&
2281 getpid_cached() != 1)
2282 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2283 "Can't run system mode unless PID 1.");
2284
2285 if (arg_action == ACTION_TEST &&
2286 geteuid() == 0)
2287 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2288 "Don't run test mode as root.");
2289
2290 if (!arg_system &&
2291 arg_action == ACTION_RUN &&
2292 sd_booted() <= 0)
2293 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
2294 "Trying to run as user instance, but the system has not been booted with systemd.");
2295
2296 if (!arg_system &&
2297 arg_action == ACTION_RUN &&
2298 !getenv("XDG_RUNTIME_DIR"))
2299 return log_error_errno(SYNTHETIC_ERRNO(EUNATCH),
2300 "Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
2301
2302 if (arg_system &&
2303 arg_action == ACTION_RUN &&
2304 running_in_chroot() > 0)
2305 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
2306 "Cannot be run in a chroot() environment.");
2307
2308 return 0;
2309 }
2310
2311 static int initialize_security(
2312 bool *loaded_policy,
2313 dual_timestamp *security_start_timestamp,
2314 dual_timestamp *security_finish_timestamp,
2315 const char **ret_error_message) {
2316
2317 int r;
2318
2319 assert(loaded_policy);
2320 assert(security_start_timestamp);
2321 assert(security_finish_timestamp);
2322 assert(ret_error_message);
2323
2324 dual_timestamp_get(security_start_timestamp);
2325
2326 r = mac_selinux_setup(loaded_policy);
2327 if (r < 0) {
2328 *ret_error_message = "Failed to load SELinux policy";
2329 return r;
2330 }
2331
2332 r = mac_smack_setup(loaded_policy);
2333 if (r < 0) {
2334 *ret_error_message = "Failed to load SMACK policy";
2335 return r;
2336 }
2337
2338 r = ima_setup();
2339 if (r < 0) {
2340 *ret_error_message = "Failed to load IMA policy";
2341 return r;
2342 }
2343
2344 dual_timestamp_get(security_finish_timestamp);
2345 return 0;
2346 }
2347
2348 static void test_summary(Manager *m) {
2349 assert(m);
2350
2351 printf("-> By units:\n");
2352 manager_dump_units(m, stdout, "\t");
2353
2354 printf("-> By jobs:\n");
2355 manager_dump_jobs(m, stdout, "\t");
2356 }
2357
2358 static int collect_fds(FDSet **ret_fds, const char **ret_error_message) {
2359 int r;
2360
2361 assert(ret_fds);
2362 assert(ret_error_message);
2363
2364 r = fdset_new_fill(ret_fds);
2365 if (r < 0) {
2366 *ret_error_message = "Failed to allocate fd set";
2367 return log_emergency_errno(r, "Failed to allocate fd set: %m");
2368 }
2369
2370 fdset_cloexec(*ret_fds, true);
2371
2372 if (arg_serialization)
2373 assert_se(fdset_remove(*ret_fds, fileno(arg_serialization)) >= 0);
2374
2375 return 0;
2376 }
2377
2378 static void setup_console_terminal(bool skip_setup) {
2379
2380 if (!arg_system)
2381 return;
2382
2383 /* Become a session leader if we aren't one yet. */
2384 (void) setsid();
2385
2386 /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2387 * tty. */
2388 (void) release_terminal();
2389
2390 /* Reset the console, but only if this is really init and we are freshly booted */
2391 if (getpid_cached() == 1 && !skip_setup)
2392 (void) console_setup();
2393 }
2394
2395 static bool early_skip_setup_check(int argc, char *argv[]) {
2396 bool found_deserialize = false;
2397 int i;
2398
2399 /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2400 * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2401 * anyway, even if in that case we also do deserialization. */
2402
2403 for (i = 1; i < argc; i++) {
2404 if (streq(argv[i], "--switched-root"))
2405 return false; /* If we switched root, don't skip the setup. */
2406 else if (streq(argv[i], "--deserialize"))
2407 found_deserialize = true;
2408 }
2409
2410 return found_deserialize; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2411 }
2412
2413 int main(int argc, char *argv[]) {
2414
2415 dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL, userspace_timestamp = DUAL_TIMESTAMP_NULL, kernel_timestamp = DUAL_TIMESTAMP_NULL,
2416 security_start_timestamp = DUAL_TIMESTAMP_NULL, security_finish_timestamp = DUAL_TIMESTAMP_NULL;
2417 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0),
2418 saved_rlimit_memlock = RLIMIT_MAKE_CONST(RLIM_INFINITY); /* The original rlimits we passed
2419 * in. Note we use different values
2420 * for the two that indicate whether
2421 * these fields are initialized! */
2422 bool skip_setup, loaded_policy = false, queue_default_job = false, first_boot = false, reexecute = false;
2423 char *switch_root_dir = NULL, *switch_root_init = NULL;
2424 usec_t before_startup, after_startup;
2425 static char systemd[] = "systemd";
2426 char timespan[FORMAT_TIMESPAN_MAX];
2427 const char *shutdown_verb = NULL, *error_message = NULL;
2428 int r, retval = EXIT_FAILURE;
2429 Manager *m = NULL;
2430 FDSet *fds = NULL;
2431
2432 /* SysV compatibility: redirect init → telinit */
2433 redirect_telinit(argc, argv);
2434
2435 /* Take timestamps early on */
2436 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
2437 dual_timestamp_get(&userspace_timestamp);
2438
2439 /* Figure out whether we need to do initialize the system, or if we already did that because we are
2440 * reexecuting */
2441 skip_setup = early_skip_setup_check(argc, argv);
2442
2443 /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2444 * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
2445 program_invocation_short_name = systemd;
2446 (void) prctl(PR_SET_NAME, systemd);
2447
2448 /* Save the original command line */
2449 save_argc_argv(argc, argv);
2450
2451 /* Make sure that if the user says "syslog" we actually log to the journal. */
2452 log_set_upgrade_syslog_to_journal(true);
2453
2454 if (getpid_cached() == 1) {
2455 /* When we run as PID 1 force system mode */
2456 arg_system = true;
2457
2458 /* Disable the umask logic */
2459 umask(0);
2460
2461 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2462 * activated yet (even though the log socket for it exists). */
2463 log_set_prohibit_ipc(true);
2464
2465 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2466 * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2467 * child process right before execve()'ing the actual binary, at a point in time where socket
2468 * activation stderr/stdout area already set up. */
2469 log_set_always_reopen_console(true);
2470
2471 if (detect_container() <= 0) {
2472
2473 /* Running outside of a container as PID 1 */
2474 log_set_target(LOG_TARGET_KMSG);
2475 log_open();
2476
2477 if (in_initrd())
2478 initrd_timestamp = userspace_timestamp;
2479
2480 if (!skip_setup) {
2481 r = mount_setup_early();
2482 if (r < 0) {
2483 error_message = "Failed to mount early API filesystems";
2484 goto finish;
2485 }
2486
2487 r = initialize_security(
2488 &loaded_policy,
2489 &security_start_timestamp,
2490 &security_finish_timestamp,
2491 &error_message);
2492 if (r < 0)
2493 goto finish;
2494 }
2495
2496 if (mac_selinux_init() < 0) {
2497 error_message = "Failed to initialize SELinux policy";
2498 goto finish;
2499 }
2500
2501 if (!skip_setup)
2502 initialize_clock();
2503
2504 /* Set the default for later on, but don't actually open the logs like this for now. Note that
2505 * if we are transitioning from the initrd there might still be journal fd open, and we
2506 * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2507 * elsewhere. */
2508 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
2509
2510 } else {
2511 /* Running inside a container, as PID 1 */
2512 log_set_target(LOG_TARGET_CONSOLE);
2513 log_open();
2514
2515 /* For later on, see above... */
2516 log_set_target(LOG_TARGET_JOURNAL);
2517
2518 /* clear the kernel timestamp,
2519 * because we are in a container */
2520 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2521 }
2522
2523 initialize_coredump(skip_setup);
2524
2525 r = fixup_environment();
2526 if (r < 0) {
2527 log_emergency_errno(r, "Failed to fix up PID 1 environment: %m");
2528 error_message = "Failed to fix up PID1 environment";
2529 goto finish;
2530 }
2531
2532 } else {
2533 /* Running as user instance */
2534 arg_system = false;
2535 log_set_target(LOG_TARGET_AUTO);
2536 log_open();
2537
2538 /* clear the kernel timestamp,
2539 * because we are not PID 1 */
2540 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2541 }
2542
2543 if (arg_system) {
2544 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2545 * they never log into the console. */
2546 log_show_color(colors_enabled());
2547
2548 r = make_null_stdio();
2549 if (r < 0)
2550 log_warning_errno(r, "Failed to redirect standard streams to /dev/null, ignoring: %m");
2551 }
2552
2553 /* Mount /proc, /sys and friends, so that /proc/cmdline and
2554 * /proc/$PID/fd is available. */
2555 if (getpid_cached() == 1) {
2556
2557 /* Load the kernel modules early. */
2558 if (!skip_setup)
2559 kmod_setup();
2560
2561 r = mount_setup(loaded_policy);
2562 if (r < 0) {
2563 error_message = "Failed to mount API filesystems";
2564 goto finish;
2565 }
2566 }
2567
2568 /* Reset all signal handlers. */
2569 (void) reset_all_signal_handlers();
2570 (void) ignore_signals(SIGNALS_IGNORE, -1);
2571
2572 r = load_configuration(argc, argv, &error_message);
2573 if (r < 0)
2574 goto finish;
2575
2576 r = safety_checks();
2577 if (r < 0)
2578 goto finish;
2579
2580 if (IN_SET(arg_action, ACTION_TEST, ACTION_HELP, ACTION_DUMP_CONFIGURATION_ITEMS, ACTION_DUMP_BUS_PROPERTIES))
2581 (void) pager_open(arg_pager_flags);
2582
2583 if (arg_action != ACTION_RUN)
2584 skip_setup = true;
2585
2586 if (arg_action == ACTION_HELP) {
2587 retval = help() < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
2588 goto finish;
2589 } else if (arg_action == ACTION_VERSION) {
2590 retval = version();
2591 goto finish;
2592 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
2593 unit_dump_config_items(stdout);
2594 retval = EXIT_SUCCESS;
2595 goto finish;
2596 } else if (arg_action == ACTION_DUMP_BUS_PROPERTIES) {
2597 dump_bus_properties(stdout);
2598 retval = EXIT_SUCCESS;
2599 goto finish;
2600 }
2601
2602 assert_se(IN_SET(arg_action, ACTION_RUN, ACTION_TEST));
2603
2604 /* Move out of the way, so that we won't block unmounts */
2605 assert_se(chdir("/") == 0);
2606
2607 if (arg_action == ACTION_RUN) {
2608
2609 /* A core pattern might have been specified via the cmdline. */
2610 initialize_core_pattern(skip_setup);
2611
2612 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2613 log_close();
2614
2615 /* Remember open file descriptors for later deserialization */
2616 r = collect_fds(&fds, &error_message);
2617 if (r < 0)
2618 goto finish;
2619
2620 /* Give up any control of the console, but make sure its initialized. */
2621 setup_console_terminal(skip_setup);
2622
2623 /* Open the logging devices, if possible and necessary */
2624 log_open();
2625 }
2626
2627 log_execution_mode(&first_boot);
2628
2629 r = initialize_runtime(skip_setup,
2630 &saved_rlimit_nofile,
2631 &saved_rlimit_memlock,
2632 &error_message);
2633 if (r < 0)
2634 goto finish;
2635
2636 r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER,
2637 arg_action == ACTION_TEST ? MANAGER_TEST_FULL : 0,
2638 &m);
2639 if (r < 0) {
2640 log_emergency_errno(r, "Failed to allocate manager object: %m");
2641 error_message = "Failed to allocate manager object";
2642 goto finish;
2643 }
2644
2645 m->timestamps[MANAGER_TIMESTAMP_KERNEL] = kernel_timestamp;
2646 m->timestamps[MANAGER_TIMESTAMP_INITRD] = initrd_timestamp;
2647 m->timestamps[MANAGER_TIMESTAMP_USERSPACE] = userspace_timestamp;
2648 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START)] = security_start_timestamp;
2649 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH)] = security_finish_timestamp;
2650
2651 set_manager_defaults(m);
2652 set_manager_settings(m);
2653 manager_set_first_boot(m, first_boot);
2654
2655 /* Remember whether we should queue the default job */
2656 queue_default_job = !arg_serialization || arg_switched_root;
2657
2658 before_startup = now(CLOCK_MONOTONIC);
2659
2660 r = manager_startup(m, arg_serialization, fds);
2661 if (r < 0) {
2662 error_message = "Failed to start up manager";
2663 goto finish;
2664 }
2665
2666 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2667 fds = fdset_free(fds);
2668 arg_serialization = safe_fclose(arg_serialization);
2669
2670 if (queue_default_job) {
2671 r = do_queue_default_job(m, &error_message);
2672 if (r < 0)
2673 goto finish;
2674 }
2675
2676 after_startup = now(CLOCK_MONOTONIC);
2677
2678 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
2679 "Loaded units and determined initial transaction in %s.",
2680 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
2681
2682 if (arg_action == ACTION_TEST) {
2683 test_summary(m);
2684 retval = EXIT_SUCCESS;
2685 goto finish;
2686 }
2687
2688 (void) invoke_main_loop(m,
2689 &reexecute,
2690 &retval,
2691 &shutdown_verb,
2692 &fds,
2693 &switch_root_dir,
2694 &switch_root_init,
2695 &error_message);
2696
2697 finish:
2698 pager_close();
2699
2700 if (m) {
2701 arg_shutdown_watchdog = m->shutdown_watchdog;
2702 m = manager_free(m);
2703 }
2704
2705 free_arguments();
2706 mac_selinux_finish();
2707
2708 if (reexecute)
2709 do_reexecute(argc, argv,
2710 &saved_rlimit_nofile,
2711 &saved_rlimit_memlock,
2712 fds,
2713 switch_root_dir,
2714 switch_root_init,
2715 &error_message); /* This only returns if reexecution failed */
2716
2717 arg_serialization = safe_fclose(arg_serialization);
2718 fds = fdset_free(fds);
2719
2720 #if HAVE_VALGRIND_VALGRIND_H
2721 /* If we are PID 1 and running under valgrind, then let's exit
2722 * here explicitly. valgrind will only generate nice output on
2723 * exit(), not on exec(), hence let's do the former not the
2724 * latter here. */
2725 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND) {
2726 /* Cleanup watchdog_device strings for valgrind. We need them
2727 * in become_shutdown() so normally we cannot free them yet. */
2728 watchdog_free_device();
2729 arg_watchdog_device = mfree(arg_watchdog_device);
2730 return retval;
2731 }
2732 #endif
2733
2734 #if HAS_FEATURE_ADDRESS_SANITIZER
2735 __lsan_do_leak_check();
2736 #endif
2737
2738 if (shutdown_verb) {
2739 r = become_shutdown(shutdown_verb, retval);
2740 log_error_errno(r, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
2741 error_message = "Failed to execute shutdown binary";
2742 }
2743
2744 watchdog_free_device();
2745 arg_watchdog_device = mfree(arg_watchdog_device);
2746
2747 if (getpid_cached() == 1) {
2748 if (error_message)
2749 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2750 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
2751 "%s.", error_message);
2752 freeze_or_exit_or_reboot();
2753 }
2754
2755 return retval;
2756 }
Unnamed repository; edit this file 'description' to name the repository.