]> git.ipfire.org Git - thirdparty/systemd.git/blob - src/core/main.c
projects / thirdparty / systemd.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge pull request #11460 from yuwata/fix-11458
[thirdparty/systemd.git] / src / core / main.c
1 /* SPDX-License-Identifier: LGPL-2.1+ */
2
3 #include <errno.h>
4 #include <fcntl.h>
5 #include <getopt.h>
6 #include <signal.h>
7 #include <stdio.h>
8 #include <string.h>
9 #include <sys/mount.h>
10 #include <sys/prctl.h>
11 #include <sys/reboot.h>
12 #include <sys/stat.h>
13 #include <unistd.h>
14 #if HAVE_SECCOMP
15 #include <seccomp.h>
16 #endif
17 #if HAVE_VALGRIND_VALGRIND_H
18 #include <valgrind/valgrind.h>
19 #endif
20
21 #include "sd-bus.h"
22 #include "sd-daemon.h"
23 #include "sd-messages.h"
24
25 #include "alloc-util.h"
26 #include "architecture.h"
27 #include "build.h"
28 #include "bus-error.h"
29 #include "bus-util.h"
30 #include "capability-util.h"
31 #include "cgroup-util.h"
32 #include "clock-util.h"
33 #include "conf-parser.h"
34 #include "cpu-set-util.h"
35 #include "dbus.h"
36 #include "dbus-manager.h"
37 #include "def.h"
38 #include "emergency-action.h"
39 #include "env-util.h"
40 #include "exit-status.h"
41 #include "fd-util.h"
42 #include "fdset.h"
43 #include "fileio.h"
44 #include "format-util.h"
45 #include "fs-util.h"
46 #include "hostname-setup.h"
47 #include "ima-setup.h"
48 #include "killall.h"
49 #include "kmod-setup.h"
50 #include "load-fragment.h"
51 #include "log.h"
52 #include "loopback-setup.h"
53 #include "machine-id-setup.h"
54 #include "manager.h"
55 #include "missing.h"
56 #include "mount-setup.h"
57 #include "os-util.h"
58 #include "pager.h"
59 #include "parse-util.h"
60 #include "path-util.h"
61 #include "pretty-print.h"
62 #include "proc-cmdline.h"
63 #include "process-util.h"
64 #include "raw-clone.h"
65 #include "rlimit-util.h"
66 #if HAVE_SECCOMP
67 #include "seccomp-util.h"
68 #endif
69 #include "selinux-setup.h"
70 #include "selinux-util.h"
71 #include "signal-util.h"
72 #include "smack-setup.h"
73 #include "special.h"
74 #include "stat-util.h"
75 #include "stdio-util.h"
76 #include "strv.h"
77 #include "switch-root.h"
78 #include "sysctl-util.h"
79 #include "terminal-util.h"
80 #include "umask-util.h"
81 #include "user-util.h"
82 #include "util.h"
83 #include "virt.h"
84 #include "watchdog.h"
85
86 #if HAS_FEATURE_ADDRESS_SANITIZER
87 #include <sanitizer/lsan_interface.h>
88 #endif
89
90 static enum {
91 ACTION_RUN,
92 ACTION_HELP,
93 ACTION_VERSION,
94 ACTION_TEST,
95 ACTION_DUMP_CONFIGURATION_ITEMS,
96 ACTION_DUMP_BUS_PROPERTIES,
97 } arg_action = ACTION_RUN;
98 static char *arg_default_unit = NULL;
99 static bool arg_system = false;
100 static bool arg_dump_core = true;
101 static int arg_crash_chvt = -1;
102 static bool arg_crash_shell = false;
103 static bool arg_crash_reboot = false;
104 static char *arg_confirm_spawn = NULL;
105 static ShowStatus arg_show_status = _SHOW_STATUS_INVALID;
106 static bool arg_switched_root = false;
107 static PagerFlags arg_pager_flags = 0;
108 static bool arg_service_watchdogs = true;
109 static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
110 static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
111 static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
112 static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
113 static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
114 static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
115 static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
116 static usec_t arg_runtime_watchdog = 0;
117 static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
118 static char *arg_early_core_pattern = NULL;
119 static char *arg_watchdog_device = NULL;
120 static char **arg_default_environment = NULL;
121 static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {};
122 static uint64_t arg_capability_bounding_set = CAP_ALL;
123 static bool arg_no_new_privs = false;
124 static nsec_t arg_timer_slack_nsec = NSEC_INFINITY;
125 static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
126 static Set* arg_syscall_archs = NULL;
127 static FILE* arg_serialization = NULL;
128 static int arg_default_cpu_accounting = -1;
129 static bool arg_default_io_accounting = false;
130 static bool arg_default_ip_accounting = false;
131 static bool arg_default_blockio_accounting = false;
132 static bool arg_default_memory_accounting = MEMORY_ACCOUNTING_DEFAULT;
133 static bool arg_default_tasks_accounting = true;
134 static uint64_t arg_default_tasks_max = UINT64_MAX;
135 static sd_id128_t arg_machine_id = {};
136 static EmergencyAction arg_cad_burst_action = EMERGENCY_ACTION_REBOOT_FORCE;
137
138 _noreturn_ static void freeze_or_exit_or_reboot(void) {
139
140 /* If we are running in a contianer, let's prefer exiting, after all we can propagate an exit code to the
141 * container manager, and thus inform it that something went wrong. */
142 if (detect_container() > 0) {
143 log_emergency("Exiting PID 1...");
144 exit(EXIT_EXCEPTION);
145 }
146
147 if (arg_crash_reboot) {
148 log_notice("Rebooting in 10s...");
149 (void) sleep(10);
150
151 log_notice("Rebooting now...");
152 (void) reboot(RB_AUTOBOOT);
153 log_emergency_errno(errno, "Failed to reboot: %m");
154 }
155
156 log_emergency("Freezing execution.");
157 freeze();
158 }
159
160 _noreturn_ static void crash(int sig) {
161 struct sigaction sa;
162 pid_t pid;
163
164 if (getpid_cached() != 1)
165 /* Pass this on immediately, if this is not PID 1 */
166 (void) raise(sig);
167 else if (!arg_dump_core)
168 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
169 else {
170 sa = (struct sigaction) {
171 .sa_handler = nop_signal_handler,
172 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
173 };
174
175 /* We want to wait for the core process, hence let's enable SIGCHLD */
176 (void) sigaction(SIGCHLD, &sa, NULL);
177
178 pid = raw_clone(SIGCHLD);
179 if (pid < 0)
180 log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
181 else if (pid == 0) {
182 /* Enable default signal handler for core dump */
183
184 sa = (struct sigaction) {
185 .sa_handler = SIG_DFL,
186 };
187 (void) sigaction(sig, &sa, NULL);
188
189 /* Don't limit the coredump size */
190 (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
191
192 /* Just to be sure... */
193 (void) chdir("/");
194
195 /* Raise the signal again */
196 pid = raw_getpid();
197 (void) kill(pid, sig); /* raise() would kill the parent */
198
199 assert_not_reached("We shouldn't be here...");
200 _exit(EXIT_EXCEPTION);
201 } else {
202 siginfo_t status;
203 int r;
204
205 /* Order things nicely. */
206 r = wait_for_terminate(pid, &status);
207 if (r < 0)
208 log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
209 else if (status.si_code != CLD_DUMPED)
210 log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
211 signal_to_string(sig),
212 pid, sigchld_code_to_string(status.si_code),
213 status.si_status,
214 strna(status.si_code == CLD_EXITED
215 ? exit_status_to_string(status.si_status, EXIT_STATUS_MINIMAL)
216 : signal_to_string(status.si_status)));
217 else
218 log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
219 }
220 }
221
222 if (arg_crash_chvt >= 0)
223 (void) chvt(arg_crash_chvt);
224
225 sa = (struct sigaction) {
226 .sa_handler = SIG_IGN,
227 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
228 };
229
230 /* Let the kernel reap children for us */
231 (void) sigaction(SIGCHLD, &sa, NULL);
232
233 if (arg_crash_shell) {
234 log_notice("Executing crash shell in 10s...");
235 (void) sleep(10);
236
237 pid = raw_clone(SIGCHLD);
238 if (pid < 0)
239 log_emergency_errno(errno, "Failed to fork off crash shell: %m");
240 else if (pid == 0) {
241 (void) setsid();
242 (void) make_console_stdio();
243 (void) rlimit_nofile_safe();
244 (void) execle("/bin/sh", "/bin/sh", NULL, environ);
245
246 log_emergency_errno(errno, "execle() failed: %m");
247 _exit(EXIT_EXCEPTION);
248 } else {
249 log_info("Spawned crash shell as PID "PID_FMT".", pid);
250 (void) wait_for_terminate(pid, NULL);
251 }
252 }
253
254 freeze_or_exit_or_reboot();
255 }
256
257 static void install_crash_handler(void) {
258 static const struct sigaction sa = {
259 .sa_handler = crash,
260 .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
261 };
262 int r;
263
264 /* We ignore the return value here, since, we don't mind if we
265 * cannot set up a crash handler */
266 r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
267 if (r < 0)
268 log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
269 }
270
271 static int console_setup(void) {
272 _cleanup_close_ int tty_fd = -1;
273 int r;
274
275 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
276 if (tty_fd < 0)
277 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
278
279 /* We don't want to force text mode. plymouth may be showing
280 * pictures already from initrd. */
281 r = reset_terminal_fd(tty_fd, false);
282 if (r < 0)
283 return log_error_errno(r, "Failed to reset /dev/console: %m");
284
285 return 0;
286 }
287
288 static int parse_crash_chvt(const char *value) {
289 int b;
290
291 if (safe_atoi(value, &arg_crash_chvt) >= 0)
292 return 0;
293
294 b = parse_boolean(value);
295 if (b < 0)
296 return b;
297
298 if (b > 0)
299 arg_crash_chvt = 0; /* switch to where kmsg goes */
300 else
301 arg_crash_chvt = -1; /* turn off switching */
302
303 return 0;
304 }
305
306 static int parse_confirm_spawn(const char *value, char **console) {
307 char *s;
308 int r;
309
310 r = value ? parse_boolean(value) : 1;
311 if (r == 0) {
312 *console = NULL;
313 return 0;
314 }
315
316 if (r > 0) /* on with default tty */
317 s = strdup("/dev/console");
318 else if (is_path(value)) /* on with fully qualified path */
319 s = strdup(value);
320 else /* on with only a tty file name, not a fully qualified path */
321 s = strjoin("/dev/", value);
322 if (!s)
323 return -ENOMEM;
324
325 *console = s;
326 return 0;
327 }
328
329 static int set_machine_id(const char *m) {
330 sd_id128_t t;
331 assert(m);
332
333 if (sd_id128_from_string(m, &t) < 0)
334 return -EINVAL;
335
336 if (sd_id128_is_null(t))
337 return -EINVAL;
338
339 arg_machine_id = t;
340 return 0;
341 }
342
343 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
344
345 int r;
346
347 assert(key);
348
349 if (STR_IN_SET(key, "systemd.unit", "rd.systemd.unit")) {
350
351 if (proc_cmdline_value_missing(key, value))
352 return 0;
353
354 if (!unit_name_is_valid(value, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE))
355 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key, value);
356 else if (in_initrd() == !!startswith(key, "rd.")) {
357 if (free_and_strdup(&arg_default_unit, value) < 0)
358 return log_oom();
359 }
360
361 } else if (proc_cmdline_key_streq(key, "systemd.dump_core")) {
362
363 r = value ? parse_boolean(value) : true;
364 if (r < 0)
365 log_warning_errno(r, "Failed to parse dump core switch %s, ignoring: %m", value);
366 else
367 arg_dump_core = r;
368
369 } else if (proc_cmdline_key_streq(key, "systemd.early_core_pattern")) {
370
371 if (proc_cmdline_value_missing(key, value))
372 return 0;
373
374 if (path_is_absolute(value))
375 (void) parse_path_argument_and_warn(value, false, &arg_early_core_pattern);
376 else
377 log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value);
378
379 } else if (proc_cmdline_key_streq(key, "systemd.crash_chvt")) {
380
381 if (!value)
382 arg_crash_chvt = 0; /* turn on */
383 else {
384 r = parse_crash_chvt(value);
385 if (r < 0)
386 log_warning_errno(r, "Failed to parse crash chvt switch %s, ignoring: %m", value);
387 }
388
389 } else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) {
390
391 r = value ? parse_boolean(value) : true;
392 if (r < 0)
393 log_warning_errno(r, "Failed to parse crash shell switch %s, ignoring: %m", value);
394 else
395 arg_crash_shell = r;
396
397 } else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) {
398
399 r = value ? parse_boolean(value) : true;
400 if (r < 0)
401 log_warning_errno(r, "Failed to parse crash reboot switch %s, ignoring: %m", value);
402 else
403 arg_crash_reboot = r;
404
405 } else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) {
406 char *s;
407
408 r = parse_confirm_spawn(value, &s);
409 if (r < 0)
410 log_warning_errno(r, "Failed to parse confirm_spawn switch %s, ignoring: %m", value);
411 else
412 free_and_replace(arg_confirm_spawn, s);
413
414 } else if (proc_cmdline_key_streq(key, "systemd.service_watchdogs")) {
415
416 r = value ? parse_boolean(value) : true;
417 if (r < 0)
418 log_warning_errno(r, "Failed to parse service watchdog switch %s, ignoring: %m", value);
419 else
420 arg_service_watchdogs = r;
421
422 } else if (proc_cmdline_key_streq(key, "systemd.show_status")) {
423
424 if (value) {
425 r = parse_show_status(value, &arg_show_status);
426 if (r < 0)
427 log_warning_errno(r, "Failed to parse show status switch %s, ignoring: %m", value);
428 } else
429 arg_show_status = SHOW_STATUS_YES;
430
431 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) {
432
433 if (proc_cmdline_value_missing(key, value))
434 return 0;
435
436 r = exec_output_from_string(value);
437 if (r < 0)
438 log_warning_errno(r, "Failed to parse default standard output switch %s, ignoring: %m", value);
439 else
440 arg_default_std_output = r;
441
442 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) {
443
444 if (proc_cmdline_value_missing(key, value))
445 return 0;
446
447 r = exec_output_from_string(value);
448 if (r < 0)
449 log_warning_errno(r, "Failed to parse default standard error switch %s, ignoring: %m", value);
450 else
451 arg_default_std_error = r;
452
453 } else if (streq(key, "systemd.setenv")) {
454
455 if (proc_cmdline_value_missing(key, value))
456 return 0;
457
458 if (env_assignment_is_valid(value)) {
459 char **env;
460
461 env = strv_env_set(arg_default_environment, value);
462 if (!env)
463 return log_oom();
464
465 arg_default_environment = env;
466 } else
467 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
468
469 } else if (proc_cmdline_key_streq(key, "systemd.machine_id")) {
470
471 if (proc_cmdline_value_missing(key, value))
472 return 0;
473
474 r = set_machine_id(value);
475 if (r < 0)
476 log_warning_errno(r, "MachineID '%s' is not valid, ignoring: %m", value);
477
478 } else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) {
479
480 if (proc_cmdline_value_missing(key, value))
481 return 0;
482
483 r = parse_sec(value, &arg_default_timeout_start_usec);
484 if (r < 0)
485 log_warning_errno(r, "Failed to parse default start timeout '%s', ignoring: %m", value);
486
487 if (arg_default_timeout_start_usec <= 0)
488 arg_default_timeout_start_usec = USEC_INFINITY;
489
490 } else if (proc_cmdline_key_streq(key, "systemd.watchdog_device")) {
491
492 if (proc_cmdline_value_missing(key, value))
493 return 0;
494
495 (void) parse_path_argument_and_warn(value, false, &arg_watchdog_device);
496
497 } else if (streq(key, "quiet") && !value) {
498
499 if (arg_show_status == _SHOW_STATUS_INVALID)
500 arg_show_status = SHOW_STATUS_AUTO;
501
502 } else if (streq(key, "debug") && !value) {
503
504 /* Note that log_parse_environment() handles 'debug'
505 * too, and sets the log level to LOG_DEBUG. */
506
507 if (detect_container() > 0)
508 log_set_target(LOG_TARGET_CONSOLE);
509
510 } else if (!value) {
511 const char *target;
512
513 /* SysV compatibility */
514 target = runlevel_to_target(key);
515 if (target)
516 return free_and_strdup(&arg_default_unit, target);
517 }
518
519 return 0;
520 }
521
522 #define DEFINE_SETTER(name, func, descr) \
523 static int name(const char *unit, \
524 const char *filename, \
525 unsigned line, \
526 const char *section, \
527 unsigned section_line, \
528 const char *lvalue, \
529 int ltype, \
530 const char *rvalue, \
531 void *data, \
532 void *userdata) { \
533 \
534 int r; \
535 \
536 assert(filename); \
537 assert(lvalue); \
538 assert(rvalue); \
539 \
540 r = func(rvalue); \
541 if (r < 0) \
542 log_syntax(unit, LOG_ERR, filename, line, r, \
543 "Invalid " descr "'%s': %m", \
544 rvalue); \
545 \
546 return 0; \
547 }
548
549 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level");
550 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target");
551 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" );
552 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location");
553
554 static int config_parse_cpu_affinity2(
555 const char *unit,
556 const char *filename,
557 unsigned line,
558 const char *section,
559 unsigned section_line,
560 const char *lvalue,
561 int ltype,
562 const char *rvalue,
563 void *data,
564 void *userdata) {
565
566 _cleanup_cpu_free_ cpu_set_t *c = NULL;
567 int ncpus;
568
569 ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue);
570 if (ncpus < 0)
571 return ncpus;
572
573 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
574 log_warning_errno(errno, "Failed to set CPU affinity: %m");
575
576 return 0;
577 }
578
579 static int config_parse_show_status(
580 const char* unit,
581 const char *filename,
582 unsigned line,
583 const char *section,
584 unsigned section_line,
585 const char *lvalue,
586 int ltype,
587 const char *rvalue,
588 void *data,
589 void *userdata) {
590
591 int k;
592 ShowStatus *b = data;
593
594 assert(filename);
595 assert(lvalue);
596 assert(rvalue);
597 assert(data);
598
599 k = parse_show_status(rvalue, b);
600 if (k < 0) {
601 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
602 return 0;
603 }
604
605 return 0;
606 }
607
608 static int config_parse_output_restricted(
609 const char* unit,
610 const char *filename,
611 unsigned line,
612 const char *section,
613 unsigned section_line,
614 const char *lvalue,
615 int ltype,
616 const char *rvalue,
617 void *data,
618 void *userdata) {
619
620 ExecOutput t, *eo = data;
621
622 assert(filename);
623 assert(lvalue);
624 assert(rvalue);
625 assert(data);
626
627 t = exec_output_from_string(rvalue);
628 if (t < 0) {
629 log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse output type, ignoring: %s", rvalue);
630 return 0;
631 }
632
633 if (IN_SET(t, EXEC_OUTPUT_SOCKET, EXEC_OUTPUT_NAMED_FD, EXEC_OUTPUT_FILE, EXEC_OUTPUT_FILE_APPEND)) {
634 log_syntax(unit, LOG_ERR, filename, line, 0, "Standard output types socket, fd:, file:, append: are not supported as defaults, ignoring: %s", rvalue);
635 return 0;
636 }
637
638 *eo = t;
639 return 0;
640 }
641
642 static int config_parse_crash_chvt(
643 const char* unit,
644 const char *filename,
645 unsigned line,
646 const char *section,
647 unsigned section_line,
648 const char *lvalue,
649 int ltype,
650 const char *rvalue,
651 void *data,
652 void *userdata) {
653
654 int r;
655
656 assert(filename);
657 assert(lvalue);
658 assert(rvalue);
659
660 r = parse_crash_chvt(rvalue);
661 if (r < 0) {
662 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
663 return 0;
664 }
665
666 return 0;
667 }
668
669 static int parse_config_file(void) {
670
671 const ConfigTableItem items[] = {
672 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
673 { "Manager", "LogTarget", config_parse_target, 0, NULL },
674 { "Manager", "LogColor", config_parse_color, 0, NULL },
675 { "Manager", "LogLocation", config_parse_location, 0, NULL },
676 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
677 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL },
678 { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL },
679 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
680 { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
681 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
682 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
683 { "Manager", "JoinControllers", config_parse_warn_compat, DISABLED_CONFIGURATION, NULL },
684 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
685 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
686 { "Manager", "WatchdogDevice", config_parse_path, 0, &arg_watchdog_device },
687 { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
688 { "Manager", "NoNewPrivileges", config_parse_bool, 0, &arg_no_new_privs },
689 #if HAVE_SECCOMP
690 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
691 #endif
692 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
693 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
694 { "Manager", "DefaultStandardOutput", config_parse_output_restricted,0, &arg_default_std_output },
695 { "Manager", "DefaultStandardError", config_parse_output_restricted,0, &arg_default_std_error },
696 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
697 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
698 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
699 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */
700 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec, 0, &arg_default_start_limit_interval },
701 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
702 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
703 { "Manager", "DefaultLimitCPU", config_parse_rlimit, RLIMIT_CPU, arg_default_rlimit },
704 { "Manager", "DefaultLimitFSIZE", config_parse_rlimit, RLIMIT_FSIZE, arg_default_rlimit },
705 { "Manager", "DefaultLimitDATA", config_parse_rlimit, RLIMIT_DATA, arg_default_rlimit },
706 { "Manager", "DefaultLimitSTACK", config_parse_rlimit, RLIMIT_STACK, arg_default_rlimit },
707 { "Manager", "DefaultLimitCORE", config_parse_rlimit, RLIMIT_CORE, arg_default_rlimit },
708 { "Manager", "DefaultLimitRSS", config_parse_rlimit, RLIMIT_RSS, arg_default_rlimit },
709 { "Manager", "DefaultLimitNOFILE", config_parse_rlimit, RLIMIT_NOFILE, arg_default_rlimit },
710 { "Manager", "DefaultLimitAS", config_parse_rlimit, RLIMIT_AS, arg_default_rlimit },
711 { "Manager", "DefaultLimitNPROC", config_parse_rlimit, RLIMIT_NPROC, arg_default_rlimit },
712 { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit, RLIMIT_MEMLOCK, arg_default_rlimit },
713 { "Manager", "DefaultLimitLOCKS", config_parse_rlimit, RLIMIT_LOCKS, arg_default_rlimit },
714 { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit, RLIMIT_SIGPENDING, arg_default_rlimit },
715 { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit, RLIMIT_MSGQUEUE, arg_default_rlimit },
716 { "Manager", "DefaultLimitNICE", config_parse_rlimit, RLIMIT_NICE, arg_default_rlimit },
717 { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit, RLIMIT_RTPRIO, arg_default_rlimit },
718 { "Manager", "DefaultLimitRTTIME", config_parse_rlimit, RLIMIT_RTTIME, arg_default_rlimit },
719 { "Manager", "DefaultCPUAccounting", config_parse_tristate, 0, &arg_default_cpu_accounting },
720 { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting },
721 { "Manager", "DefaultIPAccounting", config_parse_bool, 0, &arg_default_ip_accounting },
722 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
723 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
724 { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
725 { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max },
726 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action, 0, &arg_cad_burst_action },
727 {}
728 };
729
730 const char *fn, *conf_dirs_nulstr;
731
732 fn = arg_system ?
733 PKGSYSCONFDIR "/system.conf" :
734 PKGSYSCONFDIR "/user.conf";
735
736 conf_dirs_nulstr = arg_system ?
737 CONF_PATHS_NULSTR("systemd/system.conf.d") :
738 CONF_PATHS_NULSTR("systemd/user.conf.d");
739
740 (void) config_parse_many_nulstr(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, CONFIG_PARSE_WARN, NULL);
741
742 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
743 * like everywhere else. */
744 if (arg_default_timeout_start_usec <= 0)
745 arg_default_timeout_start_usec = USEC_INFINITY;
746 if (arg_default_timeout_stop_usec <= 0)
747 arg_default_timeout_stop_usec = USEC_INFINITY;
748
749 return 0;
750 }
751
752 static void set_manager_defaults(Manager *m) {
753
754 assert(m);
755
756 /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
757 * affect the manager itself, but are just what newly allocated units will have set if they haven't set
758 * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
759
760 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
761 m->default_std_output = arg_default_std_output;
762 m->default_std_error = arg_default_std_error;
763 m->default_timeout_start_usec = arg_default_timeout_start_usec;
764 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
765 m->default_restart_usec = arg_default_restart_usec;
766 m->default_start_limit_interval = arg_default_start_limit_interval;
767 m->default_start_limit_burst = arg_default_start_limit_burst;
768
769 /* On 4.15+ with unified hierarchy, CPU accounting is essentially free as it doesn't require the CPU
770 * controller to be enabled, so the default is to enable it unless we got told otherwise. */
771 if (arg_default_cpu_accounting >= 0)
772 m->default_cpu_accounting = arg_default_cpu_accounting;
773 else
774 m->default_cpu_accounting = cpu_accounting_is_cheap();
775
776 m->default_io_accounting = arg_default_io_accounting;
777 m->default_ip_accounting = arg_default_ip_accounting;
778 m->default_blockio_accounting = arg_default_blockio_accounting;
779 m->default_memory_accounting = arg_default_memory_accounting;
780 m->default_tasks_accounting = arg_default_tasks_accounting;
781 m->default_tasks_max = arg_default_tasks_max;
782
783 (void) manager_set_default_rlimits(m, arg_default_rlimit);
784
785 (void) manager_default_environment(m);
786 (void) manager_transient_environment_add(m, arg_default_environment);
787 }
788
789 static void set_manager_settings(Manager *m) {
790
791 assert(m);
792
793 /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
794 * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
795
796 m->confirm_spawn = arg_confirm_spawn;
797 m->service_watchdogs = arg_service_watchdogs;
798 m->runtime_watchdog = arg_runtime_watchdog;
799 m->shutdown_watchdog = arg_shutdown_watchdog;
800 m->cad_burst_action = arg_cad_burst_action;
801
802 manager_set_show_status(m, arg_show_status);
803 }
804
805 static int parse_argv(int argc, char *argv[]) {
806 enum {
807 ARG_LOG_LEVEL = 0x100,
808 ARG_LOG_TARGET,
809 ARG_LOG_COLOR,
810 ARG_LOG_LOCATION,
811 ARG_UNIT,
812 ARG_SYSTEM,
813 ARG_USER,
814 ARG_TEST,
815 ARG_NO_PAGER,
816 ARG_VERSION,
817 ARG_DUMP_CONFIGURATION_ITEMS,
818 ARG_DUMP_BUS_PROPERTIES,
819 ARG_DUMP_CORE,
820 ARG_CRASH_CHVT,
821 ARG_CRASH_SHELL,
822 ARG_CRASH_REBOOT,
823 ARG_CONFIRM_SPAWN,
824 ARG_SHOW_STATUS,
825 ARG_DESERIALIZE,
826 ARG_SWITCHED_ROOT,
827 ARG_DEFAULT_STD_OUTPUT,
828 ARG_DEFAULT_STD_ERROR,
829 ARG_MACHINE_ID,
830 ARG_SERVICE_WATCHDOGS,
831 };
832
833 static const struct option options[] = {
834 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
835 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
836 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
837 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
838 { "unit", required_argument, NULL, ARG_UNIT },
839 { "system", no_argument, NULL, ARG_SYSTEM },
840 { "user", no_argument, NULL, ARG_USER },
841 { "test", no_argument, NULL, ARG_TEST },
842 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
843 { "help", no_argument, NULL, 'h' },
844 { "version", no_argument, NULL, ARG_VERSION },
845 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
846 { "dump-bus-properties", no_argument, NULL, ARG_DUMP_BUS_PROPERTIES },
847 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
848 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
849 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
850 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
851 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
852 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
853 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
854 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
855 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
856 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
857 { "machine-id", required_argument, NULL, ARG_MACHINE_ID },
858 { "service-watchdogs", required_argument, NULL, ARG_SERVICE_WATCHDOGS },
859 {}
860 };
861
862 int c, r;
863
864 assert(argc >= 1);
865 assert(argv);
866
867 if (getpid_cached() == 1)
868 opterr = 0;
869
870 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
871
872 switch (c) {
873
874 case ARG_LOG_LEVEL:
875 r = log_set_max_level_from_string(optarg);
876 if (r < 0)
877 return log_error_errno(r, "Failed to parse log level \"%s\": %m", optarg);
878
879 break;
880
881 case ARG_LOG_TARGET:
882 r = log_set_target_from_string(optarg);
883 if (r < 0)
884 return log_error_errno(r, "Failed to parse log target \"%s\": %m", optarg);
885
886 break;
887
888 case ARG_LOG_COLOR:
889
890 if (optarg) {
891 r = log_show_color_from_string(optarg);
892 if (r < 0)
893 return log_error_errno(r, "Failed to parse log color setting \"%s\": %m",
894 optarg);
895 } else
896 log_show_color(true);
897
898 break;
899
900 case ARG_LOG_LOCATION:
901 if (optarg) {
902 r = log_show_location_from_string(optarg);
903 if (r < 0)
904 return log_error_errno(r, "Failed to parse log location setting \"%s\": %m",
905 optarg);
906 } else
907 log_show_location(true);
908
909 break;
910
911 case ARG_DEFAULT_STD_OUTPUT:
912 r = exec_output_from_string(optarg);
913 if (r < 0)
914 return log_error_errno(r, "Failed to parse default standard output setting \"%s\": %m",
915 optarg);
916 arg_default_std_output = r;
917 break;
918
919 case ARG_DEFAULT_STD_ERROR:
920 r = exec_output_from_string(optarg);
921 if (r < 0)
922 return log_error_errno(r, "Failed to parse default standard error output setting \"%s\": %m",
923 optarg);
924 arg_default_std_error = r;
925 break;
926
927 case ARG_UNIT:
928 r = free_and_strdup(&arg_default_unit, optarg);
929 if (r < 0)
930 return log_error_errno(r, "Failed to set default unit \"%s\": %m", optarg);
931
932 break;
933
934 case ARG_SYSTEM:
935 arg_system = true;
936 break;
937
938 case ARG_USER:
939 arg_system = false;
940 break;
941
942 case ARG_TEST:
943 arg_action = ACTION_TEST;
944 break;
945
946 case ARG_NO_PAGER:
947 arg_pager_flags |= PAGER_DISABLE;
948 break;
949
950 case ARG_VERSION:
951 arg_action = ACTION_VERSION;
952 break;
953
954 case ARG_DUMP_CONFIGURATION_ITEMS:
955 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
956 break;
957
958 case ARG_DUMP_BUS_PROPERTIES:
959 arg_action = ACTION_DUMP_BUS_PROPERTIES;
960 break;
961
962 case ARG_DUMP_CORE:
963 if (!optarg)
964 arg_dump_core = true;
965 else {
966 r = parse_boolean(optarg);
967 if (r < 0)
968 return log_error_errno(r, "Failed to parse dump core boolean: \"%s\": %m",
969 optarg);
970 arg_dump_core = r;
971 }
972 break;
973
974 case ARG_CRASH_CHVT:
975 r = parse_crash_chvt(optarg);
976 if (r < 0)
977 return log_error_errno(r, "Failed to parse crash virtual terminal index: \"%s\": %m",
978 optarg);
979 break;
980
981 case ARG_CRASH_SHELL:
982 if (!optarg)
983 arg_crash_shell = true;
984 else {
985 r = parse_boolean(optarg);
986 if (r < 0)
987 return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
988 optarg);
989 arg_crash_shell = r;
990 }
991 break;
992
993 case ARG_CRASH_REBOOT:
994 if (!optarg)
995 arg_crash_reboot = true;
996 else {
997 r = parse_boolean(optarg);
998 if (r < 0)
999 return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
1000 optarg);
1001 arg_crash_reboot = r;
1002 }
1003 break;
1004
1005 case ARG_CONFIRM_SPAWN:
1006 arg_confirm_spawn = mfree(arg_confirm_spawn);
1007
1008 r = parse_confirm_spawn(optarg, &arg_confirm_spawn);
1009 if (r < 0)
1010 return log_error_errno(r, "Failed to parse confirm spawn option: \"%s\": %m",
1011 optarg);
1012 break;
1013
1014 case ARG_SERVICE_WATCHDOGS:
1015 r = parse_boolean(optarg);
1016 if (r < 0)
1017 return log_error_errno(r, "Failed to parse service watchdogs boolean: \"%s\": %m",
1018 optarg);
1019 arg_service_watchdogs = r;
1020 break;
1021
1022 case ARG_SHOW_STATUS:
1023 if (optarg) {
1024 r = parse_show_status(optarg, &arg_show_status);
1025 if (r < 0)
1026 return log_error_errno(r, "Failed to parse show status boolean: \"%s\": %m",
1027 optarg);
1028 } else
1029 arg_show_status = SHOW_STATUS_YES;
1030 break;
1031
1032 case ARG_DESERIALIZE: {
1033 int fd;
1034 FILE *f;
1035
1036 r = safe_atoi(optarg, &fd);
1037 if (r < 0)
1038 log_error_errno(r, "Failed to parse deserialize option \"%s\": %m", optarg);
1039 if (fd < 0)
1040 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1041 "Invalid deserialize fd: %d",
1042 fd);
1043
1044 (void) fd_cloexec(fd, true);
1045
1046 f = fdopen(fd, "r");
1047 if (!f)
1048 return log_error_errno(errno, "Failed to open serialization fd %d: %m", fd);
1049
1050 safe_fclose(arg_serialization);
1051 arg_serialization = f;
1052
1053 break;
1054 }
1055
1056 case ARG_SWITCHED_ROOT:
1057 arg_switched_root = true;
1058 break;
1059
1060 case ARG_MACHINE_ID:
1061 r = set_machine_id(optarg);
1062 if (r < 0)
1063 return log_error_errno(r, "MachineID '%s' is not valid: %m", optarg);
1064 break;
1065
1066 case 'h':
1067 arg_action = ACTION_HELP;
1068 break;
1069
1070 case 'D':
1071 log_set_max_level(LOG_DEBUG);
1072 break;
1073
1074 case 'b':
1075 case 's':
1076 case 'z':
1077 /* Just to eat away the sysvinit kernel
1078 * cmdline args without getopt() error
1079 * messages that we'll parse in
1080 * parse_proc_cmdline_word() or ignore. */
1081
1082 case '?':
1083 if (getpid_cached() != 1)
1084 return -EINVAL;
1085 else
1086 return 0;
1087
1088 default:
1089 assert_not_reached("Unhandled option code.");
1090 }
1091
1092 if (optind < argc && getpid_cached() != 1) {
1093 /* Hmm, when we aren't run as init system
1094 * let's complain about excess arguments */
1095
1096 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1097 "Excess arguments.");
1098 }
1099
1100 return 0;
1101 }
1102
1103 static int help(void) {
1104 _cleanup_free_ char *link = NULL;
1105 int r;
1106
1107 r = terminal_urlify_man("systemd", "1", &link);
1108 if (r < 0)
1109 return log_oom();
1110
1111 printf("%s [OPTIONS...]\n\n"
1112 "Starts up and maintains the system or user services.\n\n"
1113 " -h --help Show this help\n"
1114 " --version Show version\n"
1115 " --test Determine startup sequence, dump it and exit\n"
1116 " --no-pager Do not pipe output into a pager\n"
1117 " --dump-configuration-items Dump understood unit configuration items\n"
1118 " --dump-bus-properties Dump exposed bus properties\n"
1119 " --unit=UNIT Set default unit\n"
1120 " --system Run a system instance, even if PID != 1\n"
1121 " --user Run a user instance\n"
1122 " --dump-core[=BOOL] Dump core on crash\n"
1123 " --crash-vt=NR Change to specified VT on crash\n"
1124 " --crash-reboot[=BOOL] Reboot on crash\n"
1125 " --crash-shell[=BOOL] Run shell on crash\n"
1126 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1127 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1128 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1129 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1130 " --log-color[=BOOL] Highlight important log messages\n"
1131 " --log-location[=BOOL] Include code location in log messages\n"
1132 " --default-standard-output= Set default standard output for services\n"
1133 " --default-standard-error= Set default standard error output for services\n"
1134 "\nSee the %s for details.\n"
1135 , program_invocation_short_name
1136 , link
1137 );
1138
1139 return 0;
1140 }
1141
1142 static int prepare_reexecute(
1143 Manager *m,
1144 FILE **ret_f,
1145 FDSet **ret_fds,
1146 bool switching_root) {
1147
1148 _cleanup_fdset_free_ FDSet *fds = NULL;
1149 _cleanup_fclose_ FILE *f = NULL;
1150 int r;
1151
1152 assert(m);
1153 assert(ret_f);
1154 assert(ret_fds);
1155
1156 r = manager_open_serialization(m, &f);
1157 if (r < 0)
1158 return log_error_errno(r, "Failed to create serialization file: %m");
1159
1160 /* Make sure nothing is really destructed when we shut down */
1161 m->n_reloading++;
1162 bus_manager_send_reloading(m, true);
1163
1164 fds = fdset_new();
1165 if (!fds)
1166 return log_oom();
1167
1168 r = manager_serialize(m, f, fds, switching_root);
1169 if (r < 0)
1170 return r;
1171
1172 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1173 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1174
1175 r = fd_cloexec(fileno(f), false);
1176 if (r < 0)
1177 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1178
1179 r = fdset_cloexec(fds, false);
1180 if (r < 0)
1181 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1182
1183 *ret_f = TAKE_PTR(f);
1184 *ret_fds = TAKE_PTR(fds);
1185
1186 return 0;
1187 }
1188
1189 static void bump_file_max_and_nr_open(void) {
1190
1191 /* Let's bump fs.file-max and fs.nr_open to their respective maximums. On current kernels large numbers of file
1192 * descriptors are no longer a performance problem and their memory is properly tracked by memcg, thus counting
1193 * them and limiting them in another two layers of limits is unnecessary and just complicates things. This
1194 * function hence turns off 2 of the 4 levels of limits on file descriptors, and makes RLIMIT_NOLIMIT (soft +
1195 * hard) the only ones that really matter. */
1196
1197 #if BUMP_PROC_SYS_FS_FILE_MAX || BUMP_PROC_SYS_FS_NR_OPEN
1198 _cleanup_free_ char *t = NULL;
1199 int r;
1200 #endif
1201
1202 #if BUMP_PROC_SYS_FS_FILE_MAX
1203 /* I so wanted to use STRINGIFY(ULONG_MAX) here, but alas we can't as glibc/gcc define that as
1204 * "(0x7fffffffffffffffL * 2UL + 1UL)". Seriously. 😢 */
1205 if (asprintf(&t, "%lu\n", ULONG_MAX) < 0) {
1206 log_oom();
1207 return;
1208 }
1209
1210 r = sysctl_write("fs/file-max", t);
1211 if (r < 0)
1212 log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.file-max, ignoring: %m");
1213 #endif
1214
1215 #if BUMP_PROC_SYS_FS_FILE_MAX && BUMP_PROC_SYS_FS_NR_OPEN
1216 t = mfree(t);
1217 #endif
1218
1219 #if BUMP_PROC_SYS_FS_NR_OPEN
1220 int v = INT_MAX;
1221
1222 /* Arg! The kernel enforces maximum and minimum values on the fs.nr_open, but we don't really know what they
1223 * are. The expression by which the maximum is determined is dependent on the architecture, and is something we
1224 * don't really want to copy to userspace, as it is dependent on implementation details of the kernel. Since
1225 * the kernel doesn't expose the maximum value to us, we can only try and hope. Hence, let's start with
1226 * INT_MAX, and then keep halving the value until we find one that works. Ugly? Yes, absolutely, but kernel
1227 * APIs are kernel APIs, so what do can we do... 🤯 */
1228
1229 for (;;) {
1230 int k;
1231
1232 v &= ~(__SIZEOF_POINTER__ - 1); /* Round down to next multiple of the pointer size */
1233 if (v < 1024) {
1234 log_warning("Can't bump fs.nr_open, value too small.");
1235 break;
1236 }
1237
1238 k = read_nr_open();
1239 if (k < 0) {
1240 log_error_errno(k, "Failed to read fs.nr_open: %m");
1241 break;
1242 }
1243 if (k >= v) { /* Already larger */
1244 log_debug("Skipping bump, value is already larger.");
1245 break;
1246 }
1247
1248 if (asprintf(&t, "%i\n", v) < 0) {
1249 log_oom();
1250 return;
1251 }
1252
1253 r = sysctl_write("fs/nr_open", t);
1254 t = mfree(t);
1255 if (r == -EINVAL) {
1256 log_debug("Couldn't write fs.nr_open as %i, halving it.", v);
1257 v /= 2;
1258 continue;
1259 }
1260 if (r < 0) {
1261 log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.nr_open, ignoring: %m");
1262 break;
1263 }
1264
1265 log_debug("Successfully bumped fs.nr_open to %i", v);
1266 break;
1267 }
1268 #endif
1269 }
1270
1271 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1272 struct rlimit new_rlimit;
1273 int r, nr;
1274
1275 assert(saved_rlimit);
1276
1277 /* Save the original RLIMIT_NOFILE so that we can reset it later when transitioning from the initrd to the main
1278 * systemd or suchlike. */
1279 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1280 return log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1281
1282 /* Get the underlying absolute limit the kernel enforces */
1283 nr = read_nr_open();
1284
1285 /* Make sure forked processes get limits based on the original kernel setting */
1286 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1287 struct rlimit *rl;
1288
1289 rl = newdup(struct rlimit, saved_rlimit, 1);
1290 if (!rl)
1291 return log_oom();
1292
1293 /* Bump the hard limit for system services to a substantially higher value. The default hard limit
1294 * current kernels set is pretty low (4K), mostly for historical reasons. According to kernel
1295 * developers, the fd handling in recent kernels has been optimized substantially enough, so that we
1296 * can bump the limit now, without paying too high a price in memory or performance. Note however that
1297 * we only bump the hard limit, not the soft limit. That's because select() works the way it works, and
1298 * chokes on fds >= 1024. If we'd bump the soft limit globally, it might accidentally happen to
1299 * unexpecting programs that they get fds higher than what they can process using select(). By only
1300 * bumping the hard limit but leaving the low limit as it is we avoid this pitfall: programs that are
1301 * written by folks aware of the select() problem in mind (and thus use poll()/epoll instead of
1302 * select(), the way everybody should) can explicitly opt into high fds by bumping their soft limit
1303 * beyond 1024, to the hard limit we pass. */
1304 if (arg_system)
1305 rl->rlim_max = MIN((rlim_t) nr, MAX(rl->rlim_max, (rlim_t) HIGH_RLIMIT_NOFILE));
1306
1307 /* If for some reason we were invoked with a soft limit above 1024 (which should never
1308 * happen!, but who knows what we get passed in from pam_limit when invoked as --user
1309 * instance), then lower what we pass on to not confuse our children */
1310 rl->rlim_cur = MIN(rl->rlim_cur, (rlim_t) FD_SETSIZE);
1311
1312 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1313 }
1314
1315 /* Calculate the new limits to use for us. Never lower from what we inherited. */
1316 new_rlimit = (struct rlimit) {
1317 .rlim_cur = MAX((rlim_t) nr, saved_rlimit->rlim_cur),
1318 .rlim_max = MAX((rlim_t) nr, saved_rlimit->rlim_max),
1319 };
1320
1321 /* Shortcut if nothing changes. */
1322 if (saved_rlimit->rlim_max >= new_rlimit.rlim_max &&
1323 saved_rlimit->rlim_cur >= new_rlimit.rlim_cur) {
1324 log_debug("RLIMIT_NOFILE is already as high or higher than we need it, not bumping.");
1325 return 0;
1326 }
1327
1328 /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for
1329 * both hard and soft. */
1330 r = setrlimit_closest(RLIMIT_NOFILE, &new_rlimit);
1331 if (r < 0)
1332 return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1333
1334 return 0;
1335 }
1336
1337 static int bump_rlimit_memlock(struct rlimit *saved_rlimit) {
1338 struct rlimit new_rlimit;
1339 int r;
1340
1341 assert(saved_rlimit);
1342
1343 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK which should
1344 * normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's bump
1345 * the value high enough for our user. */
1346
1347 if (getrlimit(RLIMIT_MEMLOCK, saved_rlimit) < 0)
1348 return log_warning_errno(errno, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1349
1350 /* Pass the original value down to invoked processes */
1351 if (!arg_default_rlimit[RLIMIT_MEMLOCK]) {
1352 struct rlimit *rl;
1353
1354 rl = newdup(struct rlimit, saved_rlimit, 1);
1355 if (!rl)
1356 return log_oom();
1357
1358 arg_default_rlimit[RLIMIT_MEMLOCK] = rl;
1359 }
1360
1361 /* Using MAX() on resource limits only is safe if RLIM_INFINITY is > 0. POSIX declares that rlim_t
1362 * must be unsigned, hence this is a given, but let's make this clear here. */
1363 assert_cc(RLIM_INFINITY > 0);
1364
1365 new_rlimit = (struct rlimit) {
1366 .rlim_cur = MAX(HIGH_RLIMIT_MEMLOCK, saved_rlimit->rlim_cur),
1367 .rlim_max = MAX(HIGH_RLIMIT_MEMLOCK, saved_rlimit->rlim_max),
1368 };
1369
1370 if (saved_rlimit->rlim_max >= new_rlimit.rlim_cur &&
1371 saved_rlimit->rlim_cur >= new_rlimit.rlim_max) {
1372 log_debug("RLIMIT_MEMLOCK is already as high or higher than we need it, not bumping.");
1373 return 0;
1374 }
1375
1376 r = setrlimit_closest(RLIMIT_MEMLOCK, &new_rlimit);
1377 if (r < 0)
1378 return log_warning_errno(r, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1379
1380 return 0;
1381 }
1382
1383 static void test_usr(void) {
1384
1385 /* Check that /usr is not a separate fs */
1386
1387 if (dir_is_empty("/usr") <= 0)
1388 return;
1389
1390 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1391 "Some things will probably break (sometimes even silently) in mysterious ways. "
1392 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1393 }
1394
1395 static int enforce_syscall_archs(Set *archs) {
1396 #if HAVE_SECCOMP
1397 int r;
1398
1399 if (!is_seccomp_available())
1400 return 0;
1401
1402 r = seccomp_restrict_archs(arg_syscall_archs);
1403 if (r < 0)
1404 return log_error_errno(r, "Failed to enforce system call architecture restrication: %m");
1405 #endif
1406 return 0;
1407 }
1408
1409 static int status_welcome(void) {
1410 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1411 int r;
1412
1413 if (IN_SET(arg_show_status, SHOW_STATUS_NO, SHOW_STATUS_AUTO))
1414 return 0;
1415
1416 r = parse_os_release(NULL,
1417 "PRETTY_NAME", &pretty_name,
1418 "ANSI_COLOR", &ansi_color,
1419 NULL);
1420 if (r < 0)
1421 log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r,
1422 "Failed to read os-release file, ignoring: %m");
1423
1424 if (log_get_show_color())
1425 return status_printf(NULL, 0,
1426 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1427 isempty(ansi_color) ? "1" : ansi_color,
1428 isempty(pretty_name) ? "Linux" : pretty_name);
1429 else
1430 return status_printf(NULL, 0,
1431 "\nWelcome to %s!\n",
1432 isempty(pretty_name) ? "Linux" : pretty_name);
1433 }
1434
1435 static int write_container_id(void) {
1436 const char *c;
1437 int r;
1438
1439 c = getenv("container");
1440 if (isempty(c))
1441 return 0;
1442
1443 RUN_WITH_UMASK(0022)
1444 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1445 if (r < 0)
1446 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
1447
1448 return 1;
1449 }
1450
1451 static int bump_unix_max_dgram_qlen(void) {
1452 _cleanup_free_ char *qlen = NULL;
1453 unsigned long v;
1454 int r;
1455
1456 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1457 * really really early during boot, so that it is actually applied to all our sockets, including the
1458 * $NOTIFY_SOCKET one. */
1459
1460 r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1461 if (r < 0)
1462 return log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1463
1464 r = safe_atolu(qlen, &v);
1465 if (r < 0)
1466 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen);
1467
1468 if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1469 return 0;
1470
1471 r = write_string_filef("/proc/sys/net/unix/max_dgram_qlen", WRITE_STRING_FILE_DISABLE_BUFFER, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN);
1472 if (r < 0)
1473 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1474 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1475
1476 return 1;
1477 }
1478
1479 static int fixup_environment(void) {
1480 _cleanup_free_ char *term = NULL;
1481 const char *t;
1482 int r;
1483
1484 /* Only fix up the environment when we are started as PID 1 */
1485 if (getpid_cached() != 1)
1486 return 0;
1487
1488 /* We expect the environment to be set correctly if run inside a container. */
1489 if (detect_container() > 0)
1490 return 0;
1491
1492 /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1493 * device used by the console. We try to make a better guess here since some consoles might not have support
1494 * for color mode for example.
1495 *
1496 * However if TERM was configured through the kernel command line then leave it alone. */
1497 r = proc_cmdline_get_key("TERM", 0, &term);
1498 if (r < 0)
1499 return r;
1500
1501 t = term ?: default_term_for_tty("/dev/console");
1502
1503 if (setenv("TERM", t, 1) < 0)
1504 return -errno;
1505
1506 return 0;
1507 }
1508
1509 static void redirect_telinit(int argc, char *argv[]) {
1510
1511 /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1512
1513 #if HAVE_SYSV_COMPAT
1514 if (getpid_cached() == 1)
1515 return;
1516
1517 if (!strstr(program_invocation_short_name, "init"))
1518 return;
1519
1520 execv(SYSTEMCTL_BINARY_PATH, argv);
1521 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1522 exit(EXIT_FAILURE);
1523 #endif
1524 }
1525
1526 static int become_shutdown(
1527 const char *shutdown_verb,
1528 int retval) {
1529
1530 char log_level[DECIMAL_STR_MAX(int) + 1],
1531 exit_code[DECIMAL_STR_MAX(uint8_t) + 1],
1532 timeout[DECIMAL_STR_MAX(usec_t) + 1];
1533
1534 const char* command_line[13] = {
1535 SYSTEMD_SHUTDOWN_BINARY_PATH,
1536 shutdown_verb,
1537 "--timeout", timeout,
1538 "--log-level", log_level,
1539 "--log-target",
1540 };
1541
1542 _cleanup_strv_free_ char **env_block = NULL;
1543 size_t pos = 7;
1544 int r;
1545
1546 assert(shutdown_verb);
1547 assert(!command_line[pos]);
1548 env_block = strv_copy(environ);
1549
1550 xsprintf(log_level, "%d", log_get_max_level());
1551 xsprintf(timeout, "%" PRI_USEC "us", arg_default_timeout_stop_usec);
1552
1553 switch (log_get_target()) {
1554
1555 case LOG_TARGET_KMSG:
1556 case LOG_TARGET_JOURNAL_OR_KMSG:
1557 case LOG_TARGET_SYSLOG_OR_KMSG:
1558 command_line[pos++] = "kmsg";
1559 break;
1560
1561 case LOG_TARGET_NULL:
1562 command_line[pos++] = "null";
1563 break;
1564
1565 case LOG_TARGET_CONSOLE:
1566 default:
1567 command_line[pos++] = "console";
1568 break;
1569 };
1570
1571 if (log_get_show_color())
1572 command_line[pos++] = "--log-color";
1573
1574 if (log_get_show_location())
1575 command_line[pos++] = "--log-location";
1576
1577 if (streq(shutdown_verb, "exit")) {
1578 command_line[pos++] = "--exit-code";
1579 command_line[pos++] = exit_code;
1580 xsprintf(exit_code, "%d", retval);
1581 }
1582
1583 assert(pos < ELEMENTSOF(command_line));
1584
1585 if (streq(shutdown_verb, "reboot") &&
1586 arg_shutdown_watchdog > 0 &&
1587 arg_shutdown_watchdog != USEC_INFINITY) {
1588
1589 char *e;
1590
1591 /* If we reboot let's set the shutdown
1592 * watchdog and tell the shutdown binary to
1593 * repeatedly ping it */
1594 r = watchdog_set_timeout(&arg_shutdown_watchdog);
1595 watchdog_close(r < 0);
1596
1597 /* Tell the binary how often to ping, ignore failure */
1598 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
1599 (void) strv_consume(&env_block, e);
1600
1601 if (arg_watchdog_device &&
1602 asprintf(&e, "WATCHDOG_DEVICE=%s", arg_watchdog_device) > 0)
1603 (void) strv_consume(&env_block, e);
1604 } else
1605 watchdog_close(true);
1606
1607 /* Avoid the creation of new processes forked by the
1608 * kernel; at this point, we will not listen to the
1609 * signals anyway */
1610 if (detect_container() <= 0)
1611 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
1612
1613 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
1614 return -errno;
1615 }
1616
1617 static void initialize_clock(void) {
1618 int r;
1619
1620 if (clock_is_localtime(NULL) > 0) {
1621 int min;
1622
1623 /*
1624 * The very first call of settimeofday() also does a time warp in the kernel.
1625 *
1626 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1627 * of maintaining the RTC and do all adjustments. This matches the behavior of Windows, which leaves
1628 * the RTC alone if the registry tells that the RTC runs in UTC.
1629 */
1630 r = clock_set_timezone(&min);
1631 if (r < 0)
1632 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1633 else
1634 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1635
1636 } else if (!in_initrd()) {
1637 /*
1638 * Do a dummy very first call to seal the kernel's time warp magic.
1639 *
1640 * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1641 * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1642 * until we reach the real system.
1643 *
1644 * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1645 * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1646 * be treated as UTC that way.
1647 */
1648 (void) clock_reset_timewarp();
1649 }
1650
1651 r = clock_apply_epoch();
1652 if (r < 0)
1653 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1654 else if (r > 0)
1655 log_info("System time before build time, advancing clock.");
1656 }
1657
1658 static void initialize_coredump(bool skip_setup) {
1659 #if ENABLE_COREDUMP
1660 if (getpid_cached() != 1)
1661 return;
1662
1663 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1664 * will process core dumps for system services by default. */
1665 if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0)
1666 log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m");
1667
1668 /* But at the same time, turn off the core_pattern logic by default, so that no
1669 * coredumps are stored until the systemd-coredump tool is enabled via
1670 * sysctl. However it can be changed via the kernel command line later so core
1671 * dumps can still be generated during early startup and in initramfs. */
1672 if (!skip_setup)
1673 disable_coredumps();
1674 #endif
1675 }
1676
1677 static void initialize_core_pattern(bool skip_setup) {
1678 int r;
1679
1680 if (skip_setup || !arg_early_core_pattern)
1681 return;
1682
1683 if (getpid_cached() != 1)
1684 return;
1685
1686 r = write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern, WRITE_STRING_FILE_DISABLE_BUFFER);
1687 if (r < 0)
1688 log_warning_errno(r, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", arg_early_core_pattern);
1689 }
1690
1691 static void do_reexecute(
1692 int argc,
1693 char *argv[],
1694 const struct rlimit *saved_rlimit_nofile,
1695 const struct rlimit *saved_rlimit_memlock,
1696 FDSet *fds,
1697 const char *switch_root_dir,
1698 const char *switch_root_init,
1699 const char **ret_error_message) {
1700
1701 unsigned i, j, args_size;
1702 const char **args;
1703 int r;
1704
1705 assert(saved_rlimit_nofile);
1706 assert(saved_rlimit_memlock);
1707 assert(ret_error_message);
1708
1709 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1710 * we do that */
1711 watchdog_close(true);
1712
1713 /* Reset RLIMIT_NOFILE + RLIMIT_MEMLOCK back to the kernel defaults, so that the new systemd can pass
1714 * the kernel default to its child processes */
1715 if (saved_rlimit_nofile->rlim_cur != 0)
1716 (void) setrlimit(RLIMIT_NOFILE, saved_rlimit_nofile);
1717 if (saved_rlimit_memlock->rlim_cur != RLIM_INFINITY)
1718 (void) setrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock);
1719
1720 if (switch_root_dir) {
1721 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1722 * SIGCHLD for them after deserializing. */
1723 broadcast_signal(SIGTERM, false, true, arg_default_timeout_stop_usec);
1724
1725 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1726 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1727 if (r < 0)
1728 log_error_errno(r, "Failed to switch root, trying to continue: %m");
1729 }
1730
1731 args_size = MAX(6, argc+1);
1732 args = newa(const char*, args_size);
1733
1734 if (!switch_root_init) {
1735 char sfd[DECIMAL_STR_MAX(int) + 1];
1736
1737 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1738 * the user didn't specify an explicit init to spawn. */
1739
1740 assert(arg_serialization);
1741 assert(fds);
1742
1743 xsprintf(sfd, "%i", fileno(arg_serialization));
1744
1745 i = 0;
1746 args[i++] = SYSTEMD_BINARY_PATH;
1747 if (switch_root_dir)
1748 args[i++] = "--switched-root";
1749 args[i++] = arg_system ? "--system" : "--user";
1750 args[i++] = "--deserialize";
1751 args[i++] = sfd;
1752 args[i++] = NULL;
1753
1754 assert(i <= args_size);
1755
1756 /*
1757 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do this is on
1758 * its own on exec(), but it will do it on exit(). Hence, to ensure we get a summary here, fork() off
1759 * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1760 * before proceeding into the exec().
1761 */
1762 valgrind_summary_hack();
1763
1764 (void) execv(args[0], (char* const*) args);
1765 log_debug_errno(errno, "Failed to execute our own binary, trying fallback: %m");
1766 }
1767
1768 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1769 * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1770 * doesn't matter.) */
1771
1772 arg_serialization = safe_fclose(arg_serialization);
1773 fds = fdset_free(fds);
1774
1775 /* Reopen the console */
1776 (void) make_console_stdio();
1777
1778 for (j = 1, i = 1; j < (unsigned) argc; j++)
1779 args[i++] = argv[j];
1780 args[i++] = NULL;
1781 assert(i <= args_size);
1782
1783 /* Reenable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1784 (void) reset_all_signal_handlers();
1785 (void) reset_signal_mask();
1786 (void) rlimit_nofile_safe();
1787
1788 if (switch_root_init) {
1789 args[0] = switch_root_init;
1790 (void) execv(args[0], (char* const*) args);
1791 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
1792 }
1793
1794 args[0] = "/sbin/init";
1795 (void) execv(args[0], (char* const*) args);
1796 r = -errno;
1797
1798 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
1799 ANSI_HIGHLIGHT_RED " !! " ANSI_NORMAL,
1800 "Failed to execute /sbin/init");
1801
1802 if (r == -ENOENT) {
1803 log_warning("No /sbin/init, trying fallback");
1804
1805 args[0] = "/bin/sh";
1806 args[1] = NULL;
1807 (void) execv(args[0], (char* const*) args);
1808 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
1809 } else
1810 log_warning_errno(r, "Failed to execute /sbin/init, giving up: %m");
1811
1812 *ret_error_message = "Failed to execute fallback shell";
1813 }
1814
1815 static int invoke_main_loop(
1816 Manager *m,
1817 bool *ret_reexecute,
1818 int *ret_retval, /* Return parameters relevant for shutting down */
1819 const char **ret_shutdown_verb, /* … */
1820 FDSet **ret_fds, /* Return parameters for reexecuting */
1821 char **ret_switch_root_dir, /* … */
1822 char **ret_switch_root_init, /* … */
1823 const char **ret_error_message) {
1824
1825 int r;
1826
1827 assert(m);
1828 assert(ret_reexecute);
1829 assert(ret_retval);
1830 assert(ret_shutdown_verb);
1831 assert(ret_fds);
1832 assert(ret_switch_root_dir);
1833 assert(ret_switch_root_init);
1834 assert(ret_error_message);
1835
1836 for (;;) {
1837 r = manager_loop(m);
1838 if (r < 0) {
1839 *ret_error_message = "Failed to run main loop";
1840 return log_emergency_errno(r, "Failed to run main loop: %m");
1841 }
1842
1843 switch ((ManagerObjective) r) {
1844
1845 case MANAGER_RELOAD: {
1846 LogTarget saved_log_target;
1847 int saved_log_level;
1848
1849 log_info("Reloading.");
1850
1851 /* First, save any overridden log level/target, then parse the configuration file, which might
1852 * change the log level to new settings. */
1853
1854 saved_log_level = m->log_level_overridden ? log_get_max_level() : -1;
1855 saved_log_target = m->log_target_overridden ? log_get_target() : _LOG_TARGET_INVALID;
1856
1857 r = parse_config_file();
1858 if (r < 0)
1859 log_warning_errno(r, "Failed to parse config file, ignoring: %m");
1860
1861 set_manager_defaults(m);
1862
1863 if (saved_log_level >= 0)
1864 manager_override_log_level(m, saved_log_level);
1865 if (saved_log_target >= 0)
1866 manager_override_log_target(m, saved_log_target);
1867
1868 r = manager_reload(m);
1869 if (r < 0)
1870 /* Reloading failed before the point of no return. Let's continue running as if nothing happened. */
1871 m->objective = MANAGER_OK;
1872
1873 break;
1874 }
1875
1876 case MANAGER_REEXECUTE:
1877
1878 r = prepare_reexecute(m, &arg_serialization, ret_fds, false);
1879 if (r < 0) {
1880 *ret_error_message = "Failed to prepare for reexecution";
1881 return r;
1882 }
1883
1884 log_notice("Reexecuting.");
1885
1886 *ret_reexecute = true;
1887 *ret_retval = EXIT_SUCCESS;
1888 *ret_shutdown_verb = NULL;
1889 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1890
1891 return 0;
1892
1893 case MANAGER_SWITCH_ROOT:
1894 if (!m->switch_root_init) {
1895 r = prepare_reexecute(m, &arg_serialization, ret_fds, true);
1896 if (r < 0) {
1897 *ret_error_message = "Failed to prepare for reexecution";
1898 return r;
1899 }
1900 } else
1901 *ret_fds = NULL;
1902
1903 log_notice("Switching root.");
1904
1905 *ret_reexecute = true;
1906 *ret_retval = EXIT_SUCCESS;
1907 *ret_shutdown_verb = NULL;
1908
1909 /* Steal the switch root parameters */
1910 *ret_switch_root_dir = m->switch_root;
1911 *ret_switch_root_init = m->switch_root_init;
1912 m->switch_root = m->switch_root_init = NULL;
1913
1914 return 0;
1915
1916 case MANAGER_EXIT:
1917
1918 if (MANAGER_IS_USER(m)) {
1919 log_debug("Exit.");
1920
1921 *ret_reexecute = false;
1922 *ret_retval = m->return_value;
1923 *ret_shutdown_verb = NULL;
1924 *ret_fds = NULL;
1925 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1926
1927 return 0;
1928 }
1929
1930 _fallthrough_;
1931 case MANAGER_REBOOT:
1932 case MANAGER_POWEROFF:
1933 case MANAGER_HALT:
1934 case MANAGER_KEXEC: {
1935 static const char * const table[_MANAGER_OBJECTIVE_MAX] = {
1936 [MANAGER_EXIT] = "exit",
1937 [MANAGER_REBOOT] = "reboot",
1938 [MANAGER_POWEROFF] = "poweroff",
1939 [MANAGER_HALT] = "halt",
1940 [MANAGER_KEXEC] = "kexec",
1941 };
1942
1943 log_notice("Shutting down.");
1944
1945 *ret_reexecute = false;
1946 *ret_retval = m->return_value;
1947 assert_se(*ret_shutdown_verb = table[m->objective]);
1948 *ret_fds = NULL;
1949 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1950
1951 return 0;
1952 }
1953
1954 default:
1955 assert_not_reached("Unknown or unexpected manager objective.");
1956 }
1957 }
1958 }
1959
1960 static void log_execution_mode(bool *ret_first_boot) {
1961 assert(ret_first_boot);
1962
1963 if (arg_system) {
1964 int v;
1965
1966 log_info("systemd " GIT_VERSION " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1967 arg_action == ACTION_TEST ? "test " : "" );
1968
1969 v = detect_virtualization();
1970 if (v > 0)
1971 log_info("Detected virtualization %s.", virtualization_to_string(v));
1972
1973 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1974
1975 if (in_initrd()) {
1976 *ret_first_boot = false;
1977 log_info("Running in initial RAM disk.");
1978 } else {
1979 /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
1980 * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
1981 * doesn't it's unpopulated. This allows container managers and installers to provision a
1982 * couple of files already. If the container manager wants to provision the machine ID itself
1983 * it should pass $container_uuid to PID 1. */
1984
1985 *ret_first_boot = access("/etc/machine-id", F_OK) < 0;
1986 if (*ret_first_boot)
1987 log_info("Running with unpopulated /etc.");
1988 }
1989 } else {
1990 if (DEBUG_LOGGING) {
1991 _cleanup_free_ char *t;
1992
1993 t = uid_to_name(getuid());
1994 log_debug("systemd " GIT_VERSION " running in %suser mode for user " UID_FMT "/%s. (" SYSTEMD_FEATURES ")",
1995 arg_action == ACTION_TEST ? " test" : "", getuid(), strna(t));
1996 }
1997
1998 *ret_first_boot = false;
1999 }
2000 }
2001
2002 static int initialize_runtime(
2003 bool skip_setup,
2004 struct rlimit *saved_rlimit_nofile,
2005 struct rlimit *saved_rlimit_memlock,
2006 const char **ret_error_message) {
2007
2008 int r;
2009
2010 assert(ret_error_message);
2011
2012 /* Sets up various runtime parameters. Many of these initializations are conditionalized:
2013 *
2014 * - Some only apply to --system instances
2015 * - Some only apply to --user instances
2016 * - Some only apply when we first start up, but not when we reexecute
2017 */
2018
2019 if (arg_action != ACTION_RUN)
2020 return 0;
2021
2022 if (arg_system) {
2023 /* Make sure we leave a core dump without panicing the kernel. */
2024 install_crash_handler();
2025
2026 if (!skip_setup) {
2027 r = mount_cgroup_controllers();
2028 if (r < 0) {
2029 *ret_error_message = "Failed to mount cgroup hierarchies";
2030 return r;
2031 }
2032
2033 status_welcome();
2034 hostname_setup();
2035 machine_id_setup(NULL, arg_machine_id, NULL);
2036 loopback_setup();
2037 bump_unix_max_dgram_qlen();
2038 bump_file_max_and_nr_open();
2039 test_usr();
2040 write_container_id();
2041 }
2042
2043 if (arg_watchdog_device) {
2044 r = watchdog_set_device(arg_watchdog_device);
2045 if (r < 0)
2046 log_warning_errno(r, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device);
2047 }
2048
2049 if (arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY)
2050 watchdog_set_timeout(&arg_runtime_watchdog);
2051 }
2052
2053 if (arg_timer_slack_nsec != NSEC_INFINITY)
2054 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
2055 log_warning_errno(errno, "Failed to adjust timer slack, ignoring: %m");
2056
2057 if (arg_system && !cap_test_all(arg_capability_bounding_set)) {
2058 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
2059 if (r < 0) {
2060 *ret_error_message = "Failed to drop capability bounding set of usermode helpers";
2061 return log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
2062 }
2063
2064 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
2065 if (r < 0) {
2066 *ret_error_message = "Failed to drop capability bounding set";
2067 return log_emergency_errno(r, "Failed to drop capability bounding set: %m");
2068 }
2069 }
2070
2071 if (arg_system && arg_no_new_privs) {
2072 if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) {
2073 *ret_error_message = "Failed to disable new privileges";
2074 return log_emergency_errno(errno, "Failed to disable new privileges: %m");
2075 }
2076 }
2077
2078 if (arg_syscall_archs) {
2079 r = enforce_syscall_archs(arg_syscall_archs);
2080 if (r < 0) {
2081 *ret_error_message = "Failed to set syscall architectures";
2082 return r;
2083 }
2084 }
2085
2086 if (!arg_system)
2087 /* Become reaper of our children */
2088 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
2089 log_warning_errno(errno, "Failed to make us a subreaper: %m");
2090
2091 /* Bump up RLIMIT_NOFILE for systemd itself */
2092 (void) bump_rlimit_nofile(saved_rlimit_nofile);
2093 (void) bump_rlimit_memlock(saved_rlimit_memlock);
2094
2095 return 0;
2096 }
2097
2098 static int do_queue_default_job(
2099 Manager *m,
2100 const char **ret_error_message) {
2101
2102 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
2103 Job *default_unit_job;
2104 Unit *target = NULL;
2105 int r;
2106
2107 log_debug("Activating default unit: %s", arg_default_unit);
2108
2109 r = manager_load_startable_unit_or_warn(m, arg_default_unit, NULL, &target);
2110 if (r < 0) {
2111 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET);
2112
2113 r = manager_load_startable_unit_or_warn(m, SPECIAL_RESCUE_TARGET, NULL, &target);
2114 if (r < 0) {
2115 *ret_error_message = r == -ERFKILL ? "Rescue target masked"
2116 : "Failed to load rescue target";
2117 return r;
2118 }
2119 }
2120
2121 assert(target->load_state == UNIT_LOADED);
2122
2123 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, &error, &default_unit_job);
2124 if (r == -EPERM) {
2125 log_debug_errno(r, "Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
2126
2127 sd_bus_error_free(&error);
2128
2129 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, &error, &default_unit_job);
2130 if (r < 0) {
2131 *ret_error_message = "Failed to start default target";
2132 return log_emergency_errno(r, "Failed to start default target: %s", bus_error_message(&error, r));
2133 }
2134
2135 } else if (r < 0) {
2136 *ret_error_message = "Failed to isolate default target";
2137 return log_emergency_errno(r, "Failed to isolate default target: %s", bus_error_message(&error, r));
2138 }
2139
2140 m->default_unit_job_id = default_unit_job->id;
2141
2142 return 0;
2143 }
2144
2145 static void free_arguments(void) {
2146
2147 /* Frees all arg_* variables, with the exception of arg_serialization */
2148 rlimit_free_all(arg_default_rlimit);
2149
2150 arg_default_unit = mfree(arg_default_unit);
2151 arg_confirm_spawn = mfree(arg_confirm_spawn);
2152 arg_default_environment = strv_free(arg_default_environment);
2153 arg_syscall_archs = set_free(arg_syscall_archs);
2154 }
2155
2156 static int load_configuration(int argc, char **argv, const char **ret_error_message) {
2157 int r;
2158
2159 assert(ret_error_message);
2160
2161 arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U);
2162
2163 r = parse_config_file();
2164 if (r < 0) {
2165 *ret_error_message = "Failed to parse config file";
2166 return r;
2167 }
2168
2169 if (arg_system) {
2170 r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
2171 if (r < 0)
2172 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
2173 }
2174
2175 /* Note that this also parses bits from the kernel command line, including "debug". */
2176 log_parse_environment();
2177
2178 r = parse_argv(argc, argv);
2179 if (r < 0) {
2180 *ret_error_message = "Failed to parse commandline arguments";
2181 return r;
2182 }
2183
2184 /* Initialize default unit */
2185 if (!arg_default_unit) {
2186 arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET);
2187 if (!arg_default_unit) {
2188 *ret_error_message = "Failed to set default unit";
2189 return log_oom();
2190 }
2191 }
2192
2193 /* Initialize the show status setting if it hasn't been set explicitly yet */
2194 if (arg_show_status == _SHOW_STATUS_INVALID)
2195 arg_show_status = SHOW_STATUS_YES;
2196
2197 return 0;
2198 }
2199
2200 static int safety_checks(void) {
2201
2202 if (getpid_cached() == 1 &&
2203 arg_action != ACTION_RUN)
2204 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2205 "Unsupported execution mode while PID 1.");
2206
2207 if (getpid_cached() == 1 &&
2208 !arg_system)
2209 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2210 "Can't run --user mode as PID 1.");
2211
2212 if (arg_action == ACTION_RUN &&
2213 arg_system &&
2214 getpid_cached() != 1)
2215 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2216 "Can't run system mode unless PID 1.");
2217
2218 if (arg_action == ACTION_TEST &&
2219 geteuid() == 0)
2220 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2221 "Don't run test mode as root.");
2222
2223 if (!arg_system &&
2224 arg_action == ACTION_RUN &&
2225 sd_booted() <= 0)
2226 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
2227 "Trying to run as user instance, but the system has not been booted with systemd.");
2228
2229 if (!arg_system &&
2230 arg_action == ACTION_RUN &&
2231 !getenv("XDG_RUNTIME_DIR"))
2232 return log_error_errno(SYNTHETIC_ERRNO(EUNATCH),
2233 "Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
2234
2235 if (arg_system &&
2236 arg_action == ACTION_RUN &&
2237 running_in_chroot() > 0)
2238 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
2239 "Cannot be run in a chroot() environment.");
2240
2241 return 0;
2242 }
2243
2244 static int initialize_security(
2245 bool *loaded_policy,
2246 dual_timestamp *security_start_timestamp,
2247 dual_timestamp *security_finish_timestamp,
2248 const char **ret_error_message) {
2249
2250 int r;
2251
2252 assert(loaded_policy);
2253 assert(security_start_timestamp);
2254 assert(security_finish_timestamp);
2255 assert(ret_error_message);
2256
2257 dual_timestamp_get(security_start_timestamp);
2258
2259 r = mac_selinux_setup(loaded_policy);
2260 if (r < 0) {
2261 *ret_error_message = "Failed to load SELinux policy";
2262 return r;
2263 }
2264
2265 r = mac_smack_setup(loaded_policy);
2266 if (r < 0) {
2267 *ret_error_message = "Failed to load SMACK policy";
2268 return r;
2269 }
2270
2271 r = ima_setup();
2272 if (r < 0) {
2273 *ret_error_message = "Failed to load IMA policy";
2274 return r;
2275 }
2276
2277 dual_timestamp_get(security_finish_timestamp);
2278 return 0;
2279 }
2280
2281 static void test_summary(Manager *m) {
2282 assert(m);
2283
2284 printf("-> By units:\n");
2285 manager_dump_units(m, stdout, "\t");
2286
2287 printf("-> By jobs:\n");
2288 manager_dump_jobs(m, stdout, "\t");
2289 }
2290
2291 static int collect_fds(FDSet **ret_fds, const char **ret_error_message) {
2292 int r;
2293
2294 assert(ret_fds);
2295 assert(ret_error_message);
2296
2297 r = fdset_new_fill(ret_fds);
2298 if (r < 0) {
2299 *ret_error_message = "Failed to allocate fd set";
2300 return log_emergency_errno(r, "Failed to allocate fd set: %m");
2301 }
2302
2303 fdset_cloexec(*ret_fds, true);
2304
2305 if (arg_serialization)
2306 assert_se(fdset_remove(*ret_fds, fileno(arg_serialization)) >= 0);
2307
2308 return 0;
2309 }
2310
2311 static void setup_console_terminal(bool skip_setup) {
2312
2313 if (!arg_system)
2314 return;
2315
2316 /* Become a session leader if we aren't one yet. */
2317 (void) setsid();
2318
2319 /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2320 * tty. */
2321 (void) release_terminal();
2322
2323 /* Reset the console, but only if this is really init and we are freshly booted */
2324 if (getpid_cached() == 1 && !skip_setup)
2325 (void) console_setup();
2326 }
2327
2328 static bool early_skip_setup_check(int argc, char *argv[]) {
2329 bool found_deserialize = false;
2330 int i;
2331
2332 /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2333 * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2334 * anyway, even if in that case we also do deserialization. */
2335
2336 for (i = 1; i < argc; i++) {
2337 if (streq(argv[i], "--switched-root"))
2338 return false; /* If we switched root, don't skip the setup. */
2339 else if (streq(argv[i], "--deserialize"))
2340 found_deserialize = true;
2341 }
2342
2343 return found_deserialize; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2344 }
2345
2346 int main(int argc, char *argv[]) {
2347
2348 dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL, userspace_timestamp = DUAL_TIMESTAMP_NULL, kernel_timestamp = DUAL_TIMESTAMP_NULL,
2349 security_start_timestamp = DUAL_TIMESTAMP_NULL, security_finish_timestamp = DUAL_TIMESTAMP_NULL;
2350 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0),
2351 saved_rlimit_memlock = RLIMIT_MAKE_CONST(RLIM_INFINITY); /* The original rlimits we passed
2352 * in. Note we use different values
2353 * for the two that indicate whether
2354 * these fields are initialized! */
2355 bool skip_setup, loaded_policy = false, queue_default_job = false, first_boot = false, reexecute = false;
2356 char *switch_root_dir = NULL, *switch_root_init = NULL;
2357 usec_t before_startup, after_startup;
2358 static char systemd[] = "systemd";
2359 char timespan[FORMAT_TIMESPAN_MAX];
2360 const char *shutdown_verb = NULL, *error_message = NULL;
2361 int r, retval = EXIT_FAILURE;
2362 Manager *m = NULL;
2363 FDSet *fds = NULL;
2364
2365 /* SysV compatibility: redirect init → telinit */
2366 redirect_telinit(argc, argv);
2367
2368 /* Take timestamps early on */
2369 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
2370 dual_timestamp_get(&userspace_timestamp);
2371
2372 /* Figure out whether we need to do initialize the system, or if we already did that because we are
2373 * reexecuting */
2374 skip_setup = early_skip_setup_check(argc, argv);
2375
2376 /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2377 * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
2378 program_invocation_short_name = systemd;
2379 (void) prctl(PR_SET_NAME, systemd);
2380
2381 /* Save the original command line */
2382 saved_argv = argv;
2383 saved_argc = argc;
2384
2385 /* Make sure that if the user says "syslog" we actually log to the journal. */
2386 log_set_upgrade_syslog_to_journal(true);
2387
2388 if (getpid_cached() == 1) {
2389 /* When we run as PID 1 force system mode */
2390 arg_system = true;
2391
2392 /* Disable the umask logic */
2393 umask(0);
2394
2395 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2396 * activated yet (even though the log socket for it exists). */
2397 log_set_prohibit_ipc(true);
2398
2399 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2400 * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2401 * child process right before execve()'ing the actual binary, at a point in time where socket
2402 * activation stderr/stdout area already set up. */
2403 log_set_always_reopen_console(true);
2404
2405 if (detect_container() <= 0) {
2406
2407 /* Running outside of a container as PID 1 */
2408 log_set_target(LOG_TARGET_KMSG);
2409 log_open();
2410
2411 if (in_initrd())
2412 initrd_timestamp = userspace_timestamp;
2413
2414 if (!skip_setup) {
2415 r = mount_setup_early();
2416 if (r < 0) {
2417 error_message = "Failed to mount early API filesystems";
2418 goto finish;
2419 }
2420
2421 r = initialize_security(
2422 &loaded_policy,
2423 &security_start_timestamp,
2424 &security_finish_timestamp,
2425 &error_message);
2426 if (r < 0)
2427 goto finish;
2428 }
2429
2430 if (mac_selinux_init() < 0) {
2431 error_message = "Failed to initialize SELinux policy";
2432 goto finish;
2433 }
2434
2435 if (!skip_setup)
2436 initialize_clock();
2437
2438 /* Set the default for later on, but don't actually open the logs like this for now. Note that
2439 * if we are transitioning from the initrd there might still be journal fd open, and we
2440 * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2441 * elsewhere. */
2442 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
2443
2444 } else {
2445 /* Running inside a container, as PID 1 */
2446 log_set_target(LOG_TARGET_CONSOLE);
2447 log_open();
2448
2449 /* For later on, see above... */
2450 log_set_target(LOG_TARGET_JOURNAL);
2451
2452 /* clear the kernel timestamp,
2453 * because we are in a container */
2454 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2455 }
2456
2457 initialize_coredump(skip_setup);
2458
2459 r = fixup_environment();
2460 if (r < 0) {
2461 log_emergency_errno(r, "Failed to fix up PID 1 environment: %m");
2462 error_message = "Failed to fix up PID1 environment";
2463 goto finish;
2464 }
2465
2466 } else {
2467 /* Running as user instance */
2468 arg_system = false;
2469 log_set_target(LOG_TARGET_AUTO);
2470 log_open();
2471
2472 /* clear the kernel timestamp,
2473 * because we are not PID 1 */
2474 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2475 }
2476
2477 if (arg_system) {
2478 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2479 * they never log into the console. */
2480 log_show_color(colors_enabled());
2481
2482 r = make_null_stdio();
2483 if (r < 0)
2484 log_warning_errno(r, "Failed to redirect standard streams to /dev/null, ignoring: %m");
2485 }
2486
2487 /* Mount /proc, /sys and friends, so that /proc/cmdline and
2488 * /proc/$PID/fd is available. */
2489 if (getpid_cached() == 1) {
2490
2491 /* Load the kernel modules early. */
2492 if (!skip_setup)
2493 kmod_setup();
2494
2495 r = mount_setup(loaded_policy);
2496 if (r < 0) {
2497 error_message = "Failed to mount API filesystems";
2498 goto finish;
2499 }
2500 }
2501
2502 /* Reset all signal handlers. */
2503 (void) reset_all_signal_handlers();
2504 (void) ignore_signals(SIGNALS_IGNORE, -1);
2505
2506 r = load_configuration(argc, argv, &error_message);
2507 if (r < 0)
2508 goto finish;
2509
2510 r = safety_checks();
2511 if (r < 0)
2512 goto finish;
2513
2514 if (IN_SET(arg_action, ACTION_TEST, ACTION_HELP, ACTION_DUMP_CONFIGURATION_ITEMS, ACTION_DUMP_BUS_PROPERTIES))
2515 (void) pager_open(arg_pager_flags);
2516
2517 if (arg_action != ACTION_RUN)
2518 skip_setup = true;
2519
2520 if (arg_action == ACTION_HELP) {
2521 retval = help() < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
2522 goto finish;
2523 } else if (arg_action == ACTION_VERSION) {
2524 retval = version();
2525 goto finish;
2526 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
2527 unit_dump_config_items(stdout);
2528 retval = EXIT_SUCCESS;
2529 goto finish;
2530 } else if (arg_action == ACTION_DUMP_BUS_PROPERTIES) {
2531 dump_bus_properties(stdout);
2532 retval = EXIT_SUCCESS;
2533 goto finish;
2534 }
2535
2536 assert_se(IN_SET(arg_action, ACTION_RUN, ACTION_TEST));
2537
2538 /* Move out of the way, so that we won't block unmounts */
2539 assert_se(chdir("/") == 0);
2540
2541 if (arg_action == ACTION_RUN) {
2542
2543 /* A core pattern might have been specified via the cmdline. */
2544 initialize_core_pattern(skip_setup);
2545
2546 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2547 log_close();
2548
2549 /* Remember open file descriptors for later deserialization */
2550 r = collect_fds(&fds, &error_message);
2551 if (r < 0)
2552 goto finish;
2553
2554 /* Give up any control of the console, but make sure its initialized. */
2555 setup_console_terminal(skip_setup);
2556
2557 /* Open the logging devices, if possible and necessary */
2558 log_open();
2559 }
2560
2561 log_execution_mode(&first_boot);
2562
2563 r = initialize_runtime(skip_setup,
2564 &saved_rlimit_nofile,
2565 &saved_rlimit_memlock,
2566 &error_message);
2567 if (r < 0)
2568 goto finish;
2569
2570 r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER,
2571 arg_action == ACTION_TEST ? MANAGER_TEST_FULL : 0,
2572 &m);
2573 if (r < 0) {
2574 log_emergency_errno(r, "Failed to allocate manager object: %m");
2575 error_message = "Failed to allocate manager object";
2576 goto finish;
2577 }
2578
2579 m->timestamps[MANAGER_TIMESTAMP_KERNEL] = kernel_timestamp;
2580 m->timestamps[MANAGER_TIMESTAMP_INITRD] = initrd_timestamp;
2581 m->timestamps[MANAGER_TIMESTAMP_USERSPACE] = userspace_timestamp;
2582 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START)] = security_start_timestamp;
2583 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH)] = security_finish_timestamp;
2584
2585 set_manager_defaults(m);
2586 set_manager_settings(m);
2587 manager_set_first_boot(m, first_boot);
2588
2589 /* Remember whether we should queue the default job */
2590 queue_default_job = !arg_serialization || arg_switched_root;
2591
2592 before_startup = now(CLOCK_MONOTONIC);
2593
2594 r = manager_startup(m, arg_serialization, fds);
2595 if (r < 0) {
2596 error_message = "Failed to start up manager";
2597 goto finish;
2598 }
2599
2600 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2601 fds = fdset_free(fds);
2602 arg_serialization = safe_fclose(arg_serialization);
2603
2604 if (queue_default_job) {
2605 r = do_queue_default_job(m, &error_message);
2606 if (r < 0)
2607 goto finish;
2608 }
2609
2610 after_startup = now(CLOCK_MONOTONIC);
2611
2612 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
2613 "Loaded units and determined initial transaction in %s.",
2614 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
2615
2616 if (arg_action == ACTION_TEST) {
2617 test_summary(m);
2618 retval = EXIT_SUCCESS;
2619 goto finish;
2620 }
2621
2622 (void) invoke_main_loop(m,
2623 &reexecute,
2624 &retval,
2625 &shutdown_verb,
2626 &fds,
2627 &switch_root_dir,
2628 &switch_root_init,
2629 &error_message);
2630
2631 finish:
2632 pager_close();
2633
2634 if (m) {
2635 arg_shutdown_watchdog = m->shutdown_watchdog;
2636 m = manager_free(m);
2637 }
2638
2639 free_arguments();
2640 mac_selinux_finish();
2641
2642 if (reexecute)
2643 do_reexecute(argc, argv,
2644 &saved_rlimit_nofile,
2645 &saved_rlimit_memlock,
2646 fds,
2647 switch_root_dir,
2648 switch_root_init,
2649 &error_message); /* This only returns if reexecution failed */
2650
2651 arg_serialization = safe_fclose(arg_serialization);
2652 fds = fdset_free(fds);
2653
2654 #if HAVE_VALGRIND_VALGRIND_H
2655 /* If we are PID 1 and running under valgrind, then let's exit
2656 * here explicitly. valgrind will only generate nice output on
2657 * exit(), not on exec(), hence let's do the former not the
2658 * latter here. */
2659 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND) {
2660 /* Cleanup watchdog_device strings for valgrind. We need them
2661 * in become_shutdown() so normally we cannot free them yet. */
2662 watchdog_free_device();
2663 arg_watchdog_device = mfree(arg_watchdog_device);
2664 return retval;
2665 }
2666 #endif
2667
2668 #if HAS_FEATURE_ADDRESS_SANITIZER
2669 __lsan_do_leak_check();
2670 #endif
2671
2672 if (shutdown_verb) {
2673 r = become_shutdown(shutdown_verb, retval);
2674 log_error_errno(r, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
2675 error_message = "Failed to execute shutdown binary";
2676 }
2677
2678 watchdog_free_device();
2679 arg_watchdog_device = mfree(arg_watchdog_device);
2680
2681 if (getpid_cached() == 1) {
2682 if (error_message)
2683 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2684 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
2685 "%s.", error_message);
2686 freeze_or_exit_or_reboot();
2687 }
2688
2689 return retval;
2690 }
Unnamed repository; edit this file 'description' to name the repository.