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