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