]> git.ipfire.org Git - thirdparty/systemd.git/blob - src/core/main.c
projects / thirdparty / systemd.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge pull request #6746 from yuwata/parse-empty-string
[thirdparty/systemd.git] / src / core / main.c
1 /***
2 This file is part of systemd.
3
4 Copyright 2010 Lennart Poettering
5
6 systemd is free software; you can redistribute it and/or modify it
7 under the terms of the GNU Lesser General Public License as published by
8 the Free Software Foundation; either version 2.1 of the License, or
9 (at your option) any later version.
10
11 systemd is distributed in the hope that it will be useful, but
12 WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
15
16 You should have received a copy of the GNU Lesser General Public License
17 along with systemd; If not, see <http://www.gnu.org/licenses/>.
18 ***/
19
20 #include <errno.h>
21 #include <fcntl.h>
22 #include <getopt.h>
23 #include <signal.h>
24 #include <stdio.h>
25 #include <string.h>
26 #include <sys/mount.h>
27 #include <sys/prctl.h>
28 #include <sys/reboot.h>
29 #include <sys/stat.h>
30 #include <unistd.h>
31 #ifdef HAVE_SECCOMP
32 #include <seccomp.h>
33 #endif
34 #ifdef HAVE_VALGRIND_VALGRIND_H
35 #include <valgrind/valgrind.h>
36 #endif
37
38 #include "sd-bus.h"
39 #include "sd-daemon.h"
40
41 #include "alloc-util.h"
42 #include "architecture.h"
43 #include "build.h"
44 #include "bus-error.h"
45 #include "bus-util.h"
46 #include "capability-util.h"
47 #include "clock-util.h"
48 #include "conf-parser.h"
49 #include "cpu-set-util.h"
50 #include "dbus-manager.h"
51 #include "def.h"
52 #include "emergency-action.h"
53 #include "env-util.h"
54 #include "fd-util.h"
55 #include "fdset.h"
56 #include "fileio.h"
57 #include "format-util.h"
58 #include "fs-util.h"
59 #include "hostname-setup.h"
60 #include "ima-setup.h"
61 #include "killall.h"
62 #include "kmod-setup.h"
63 #include "load-fragment.h"
64 #include "log.h"
65 #include "loopback-setup.h"
66 #include "machine-id-setup.h"
67 #include "manager.h"
68 #include "missing.h"
69 #include "mount-setup.h"
70 #include "pager.h"
71 #include "parse-util.h"
72 #include "path-util.h"
73 #include "proc-cmdline.h"
74 #include "process-util.h"
75 #include "raw-clone.h"
76 #include "rlimit-util.h"
77 #ifdef HAVE_SECCOMP
78 #include "seccomp-util.h"
79 #endif
80 #include "selinux-setup.h"
81 #include "selinux-util.h"
82 #include "signal-util.h"
83 #include "smack-setup.h"
84 #include "special.h"
85 #include "stat-util.h"
86 #include "stdio-util.h"
87 #include "strv.h"
88 #include "switch-root.h"
89 #include "terminal-util.h"
90 #include "umask-util.h"
91 #include "user-util.h"
92 #include "virt.h"
93 #include "watchdog.h"
94
95 static enum {
96 ACTION_RUN,
97 ACTION_HELP,
98 ACTION_VERSION,
99 ACTION_TEST,
100 ACTION_DUMP_CONFIGURATION_ITEMS
101 } arg_action = ACTION_RUN;
102 static char *arg_default_unit = NULL;
103 static bool arg_system = false;
104 static bool arg_dump_core = true;
105 static int arg_crash_chvt = -1;
106 static bool arg_crash_shell = false;
107 static bool arg_crash_reboot = false;
108 static char *arg_confirm_spawn = NULL;
109 static ShowStatus arg_show_status = _SHOW_STATUS_UNSET;
110 static bool arg_switched_root = false;
111 static bool arg_no_pager = false;
112 static char ***arg_join_controllers = NULL;
113 static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
114 static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
115 static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
116 static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
117 static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
118 static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
119 static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
120 static usec_t arg_runtime_watchdog = 0;
121 static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
122 static char **arg_default_environment = NULL;
123 static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {};
124 static uint64_t arg_capability_bounding_set = CAP_ALL;
125 static nsec_t arg_timer_slack_nsec = NSEC_INFINITY;
126 static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
127 static Set* arg_syscall_archs = NULL;
128 static FILE* arg_serialization = NULL;
129 static bool arg_default_cpu_accounting = false;
130 static bool arg_default_io_accounting = false;
131 static bool arg_default_blockio_accounting = false;
132 static bool arg_default_memory_accounting = false;
133 static bool arg_default_tasks_accounting = true;
134 static uint64_t arg_default_tasks_max = UINT64_MAX;
135 static sd_id128_t arg_machine_id = {};
136 static EmergencyAction arg_cad_burst_action = EMERGENCY_ACTION_REBOOT_FORCE;
137
138 noreturn static void freeze_or_reboot(void) {
139
140 if (arg_crash_reboot) {
141 log_notice("Rebooting in 10s...");
142 (void) sleep(10);
143
144 log_notice("Rebooting now...");
145 (void) reboot(RB_AUTOBOOT);
146 log_emergency_errno(errno, "Failed to reboot: %m");
147 }
148
149 log_emergency("Freezing execution.");
150 freeze();
151 }
152
153 noreturn static void crash(int sig) {
154 struct sigaction sa;
155 pid_t pid;
156
157 if (getpid_cached() != 1)
158 /* Pass this on immediately, if this is not PID 1 */
159 (void) raise(sig);
160 else if (!arg_dump_core)
161 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
162 else {
163 sa = (struct sigaction) {
164 .sa_handler = nop_signal_handler,
165 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
166 };
167
168 /* We want to wait for the core process, hence let's enable SIGCHLD */
169 (void) sigaction(SIGCHLD, &sa, NULL);
170
171 pid = raw_clone(SIGCHLD);
172 if (pid < 0)
173 log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
174 else if (pid == 0) {
175 /* Enable default signal handler for core dump */
176
177 sa = (struct sigaction) {
178 .sa_handler = SIG_DFL,
179 };
180 (void) sigaction(sig, &sa, NULL);
181
182 /* Don't limit the coredump size */
183 (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
184
185 /* Just to be sure... */
186 (void) chdir("/");
187
188 /* Raise the signal again */
189 pid = raw_getpid();
190 (void) kill(pid, sig); /* raise() would kill the parent */
191
192 assert_not_reached("We shouldn't be here...");
193 _exit(EXIT_FAILURE);
194 } else {
195 siginfo_t status;
196 int r;
197
198 /* Order things nicely. */
199 r = wait_for_terminate(pid, &status);
200 if (r < 0)
201 log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
202 else if (status.si_code != CLD_DUMPED)
203 log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
204 signal_to_string(sig),
205 pid, sigchld_code_to_string(status.si_code),
206 status.si_status,
207 strna(status.si_code == CLD_EXITED
208 ? exit_status_to_string(status.si_status, EXIT_STATUS_MINIMAL)
209 : signal_to_string(status.si_status)));
210 else
211 log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
212 }
213 }
214
215 if (arg_crash_chvt >= 0)
216 (void) chvt(arg_crash_chvt);
217
218 sa = (struct sigaction) {
219 .sa_handler = SIG_IGN,
220 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
221 };
222
223 /* Let the kernel reap children for us */
224 (void) sigaction(SIGCHLD, &sa, NULL);
225
226 if (arg_crash_shell) {
227 log_notice("Executing crash shell in 10s...");
228 (void) sleep(10);
229
230 pid = raw_clone(SIGCHLD);
231 if (pid < 0)
232 log_emergency_errno(errno, "Failed to fork off crash shell: %m");
233 else if (pid == 0) {
234 (void) setsid();
235 (void) make_console_stdio();
236 (void) execle("/bin/sh", "/bin/sh", NULL, environ);
237
238 log_emergency_errno(errno, "execle() failed: %m");
239 _exit(EXIT_FAILURE);
240 } else {
241 log_info("Spawned crash shell as PID "PID_FMT".", pid);
242 (void) wait_for_terminate(pid, NULL);
243 }
244 }
245
246 freeze_or_reboot();
247 }
248
249 static void install_crash_handler(void) {
250 static const struct sigaction sa = {
251 .sa_handler = crash,
252 .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
253 };
254 int r;
255
256 /* We ignore the return value here, since, we don't mind if we
257 * cannot set up a crash handler */
258 r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
259 if (r < 0)
260 log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
261 }
262
263 static int console_setup(void) {
264 _cleanup_close_ int tty_fd = -1;
265 int r;
266
267 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
268 if (tty_fd < 0)
269 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
270
271 /* We don't want to force text mode. plymouth may be showing
272 * pictures already from initrd. */
273 r = reset_terminal_fd(tty_fd, false);
274 if (r < 0)
275 return log_error_errno(r, "Failed to reset /dev/console: %m");
276
277 return 0;
278 }
279
280 static int parse_crash_chvt(const char *value) {
281 int b;
282
283 if (safe_atoi(value, &arg_crash_chvt) >= 0)
284 return 0;
285
286 b = parse_boolean(value);
287 if (b < 0)
288 return b;
289
290 if (b > 0)
291 arg_crash_chvt = 0; /* switch to where kmsg goes */
292 else
293 arg_crash_chvt = -1; /* turn off switching */
294
295 return 0;
296 }
297
298 static int parse_confirm_spawn(const char *value, char **console) {
299 char *s;
300 int r;
301
302 r = value ? parse_boolean(value) : 1;
303 if (r == 0) {
304 *console = NULL;
305 return 0;
306 }
307
308 if (r > 0) /* on with default tty */
309 s = strdup("/dev/console");
310 else if (is_path(value)) /* on with fully qualified path */
311 s = strdup(value);
312 else /* on with only a tty file name, not a fully qualified path */
313 s = strjoin("/dev/", value);
314 if (!s)
315 return -ENOMEM;
316 *console = s;
317 return 0;
318 }
319
320 static int set_machine_id(const char *m) {
321 sd_id128_t t;
322 assert(m);
323
324 if (sd_id128_from_string(m, &t) < 0)
325 return -EINVAL;
326
327 if (sd_id128_is_null(t))
328 return -EINVAL;
329
330 arg_machine_id = t;
331 return 0;
332 }
333
334 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
335
336 int r;
337
338 assert(key);
339
340 if (STR_IN_SET(key, "systemd.unit", "rd.systemd.unit")) {
341
342 if (proc_cmdline_value_missing(key, value))
343 return 0;
344
345 if (!unit_name_is_valid(value, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE))
346 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key, value);
347 else if (in_initrd() == !!startswith(key, "rd.")) {
348 if (free_and_strdup(&arg_default_unit, value) < 0)
349 return log_oom();
350 }
351
352 } else if (proc_cmdline_key_streq(key, "systemd.dump_core")) {
353
354 r = value ? parse_boolean(value) : true;
355 if (r < 0)
356 log_warning("Failed to parse dump core switch %s. Ignoring.", value);
357 else
358 arg_dump_core = r;
359
360 } else if (proc_cmdline_key_streq(key, "systemd.crash_chvt")) {
361
362 if (!value)
363 arg_crash_chvt = 0; /* turn on */
364 else if (parse_crash_chvt(value) < 0)
365 log_warning("Failed to parse crash chvt switch %s. Ignoring.", value);
366
367 } else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) {
368
369 r = value ? parse_boolean(value) : true;
370 if (r < 0)
371 log_warning("Failed to parse crash shell switch %s. Ignoring.", value);
372 else
373 arg_crash_shell = r;
374
375 } else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) {
376
377 r = value ? parse_boolean(value) : true;
378 if (r < 0)
379 log_warning("Failed to parse crash reboot switch %s. Ignoring.", value);
380 else
381 arg_crash_reboot = r;
382
383 } else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) {
384 char *s;
385
386 r = parse_confirm_spawn(value, &s);
387 if (r < 0)
388 log_warning_errno(r, "Failed to parse confirm_spawn switch %s. Ignoring.", value);
389 else {
390 free(arg_confirm_spawn);
391 arg_confirm_spawn = s;
392 }
393
394 } else if (proc_cmdline_key_streq(key, "systemd.show_status")) {
395
396 if (value) {
397 r = parse_show_status(value, &arg_show_status);
398 if (r < 0)
399 log_warning("Failed to parse show status switch %s. Ignoring.", value);
400 } else
401 arg_show_status = SHOW_STATUS_YES;
402
403 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) {
404
405 if (proc_cmdline_value_missing(key, value))
406 return 0;
407
408 r = exec_output_from_string(value);
409 if (r < 0)
410 log_warning("Failed to parse default standard output switch %s. Ignoring.", value);
411 else
412 arg_default_std_output = r;
413
414 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) {
415
416 if (proc_cmdline_value_missing(key, value))
417 return 0;
418
419 r = exec_output_from_string(value);
420 if (r < 0)
421 log_warning("Failed to parse default standard error switch %s. Ignoring.", value);
422 else
423 arg_default_std_error = r;
424
425 } else if (streq(key, "systemd.setenv")) {
426
427 if (proc_cmdline_value_missing(key, value))
428 return 0;
429
430 if (env_assignment_is_valid(value)) {
431 char **env;
432
433 env = strv_env_set(arg_default_environment, value);
434 if (!env)
435 return log_oom();
436
437 arg_default_environment = env;
438 } else
439 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
440
441 } else if (proc_cmdline_key_streq(key, "systemd.machine_id")) {
442
443 if (proc_cmdline_value_missing(key, value))
444 return 0;
445
446 r = set_machine_id(value);
447 if (r < 0)
448 log_warning("MachineID '%s' is not valid. Ignoring.", value);
449
450 } else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) {
451
452 if (proc_cmdline_value_missing(key, value))
453 return 0;
454
455 r = parse_sec(value, &arg_default_timeout_start_usec);
456 if (r < 0)
457 log_warning_errno(r, "Failed to parse default start timeout: %s, ignoring.", value);
458
459 if (arg_default_timeout_start_usec <= 0)
460 arg_default_timeout_start_usec = USEC_INFINITY;
461
462 } else if (streq(key, "quiet") && !value) {
463
464 if (arg_show_status == _SHOW_STATUS_UNSET)
465 arg_show_status = SHOW_STATUS_AUTO;
466
467 } else if (streq(key, "debug") && !value) {
468
469 /* Note that log_parse_environment() handles 'debug'
470 * too, and sets the log level to LOG_DEBUG. */
471
472 if (detect_container() > 0)
473 log_set_target(LOG_TARGET_CONSOLE);
474
475 } else if (!value) {
476 const char *target;
477
478 /* SysV compatibility */
479 target = runlevel_to_target(key);
480 if (target)
481 return free_and_strdup(&arg_default_unit, target);
482 }
483
484 return 0;
485 }
486
487 #define DEFINE_SETTER(name, func, descr) \
488 static int name(const char *unit, \
489 const char *filename, \
490 unsigned line, \
491 const char *section, \
492 unsigned section_line, \
493 const char *lvalue, \
494 int ltype, \
495 const char *rvalue, \
496 void *data, \
497 void *userdata) { \
498 \
499 int r; \
500 \
501 assert(filename); \
502 assert(lvalue); \
503 assert(rvalue); \
504 \
505 r = func(rvalue); \
506 if (r < 0) \
507 log_syntax(unit, LOG_ERR, filename, line, r, \
508 "Invalid " descr "'%s': %m", \
509 rvalue); \
510 \
511 return 0; \
512 }
513
514 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level")
515 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target")
516 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" )
517 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location")
518
519 static int config_parse_cpu_affinity2(
520 const char *unit,
521 const char *filename,
522 unsigned line,
523 const char *section,
524 unsigned section_line,
525 const char *lvalue,
526 int ltype,
527 const char *rvalue,
528 void *data,
529 void *userdata) {
530
531 _cleanup_cpu_free_ cpu_set_t *c = NULL;
532 int ncpus;
533
534 ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue);
535 if (ncpus < 0)
536 return ncpus;
537
538 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
539 log_warning_errno(errno, "Failed to set CPU affinity: %m");
540
541 return 0;
542 }
543
544 static int config_parse_show_status(
545 const char* unit,
546 const char *filename,
547 unsigned line,
548 const char *section,
549 unsigned section_line,
550 const char *lvalue,
551 int ltype,
552 const char *rvalue,
553 void *data,
554 void *userdata) {
555
556 int k;
557 ShowStatus *b = data;
558
559 assert(filename);
560 assert(lvalue);
561 assert(rvalue);
562 assert(data);
563
564 k = parse_show_status(rvalue, b);
565 if (k < 0) {
566 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
567 return 0;
568 }
569
570 return 0;
571 }
572
573 static int config_parse_crash_chvt(
574 const char* unit,
575 const char *filename,
576 unsigned line,
577 const char *section,
578 unsigned section_line,
579 const char *lvalue,
580 int ltype,
581 const char *rvalue,
582 void *data,
583 void *userdata) {
584
585 int r;
586
587 assert(filename);
588 assert(lvalue);
589 assert(rvalue);
590
591 r = parse_crash_chvt(rvalue);
592 if (r < 0) {
593 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
594 return 0;
595 }
596
597 return 0;
598 }
599
600 static int config_parse_join_controllers(const char *unit,
601 const char *filename,
602 unsigned line,
603 const char *section,
604 unsigned section_line,
605 const char *lvalue,
606 int ltype,
607 const char *rvalue,
608 void *data,
609 void *userdata) {
610
611 const char *whole_rvalue = rvalue;
612 unsigned n = 0;
613
614 assert(filename);
615 assert(lvalue);
616 assert(rvalue);
617
618 arg_join_controllers = strv_free_free(arg_join_controllers);
619
620 for (;;) {
621 _cleanup_free_ char *word = NULL;
622 char **l;
623 int r;
624
625 r = extract_first_word(&rvalue, &word, NULL, EXTRACT_QUOTES);
626 if (r < 0) {
627 log_syntax(unit, LOG_ERR, filename, line, r, "Invalid value for %s: %s", lvalue, whole_rvalue);
628 return r;
629 }
630 if (r == 0)
631 break;
632
633 l = strv_split(word, ",");
634 if (!l)
635 return log_oom();
636 strv_uniq(l);
637
638 if (strv_length(l) <= 1) {
639 strv_free(l);
640 continue;
641 }
642
643 if (!arg_join_controllers) {
644 arg_join_controllers = new(char**, 2);
645 if (!arg_join_controllers) {
646 strv_free(l);
647 return log_oom();
648 }
649
650 arg_join_controllers[0] = l;
651 arg_join_controllers[1] = NULL;
652
653 n = 1;
654 } else {
655 char ***a;
656 char ***t;
657
658 t = new0(char**, n+2);
659 if (!t) {
660 strv_free(l);
661 return log_oom();
662 }
663
664 n = 0;
665
666 for (a = arg_join_controllers; *a; a++) {
667
668 if (strv_overlap(*a, l)) {
669 if (strv_extend_strv(&l, *a, false) < 0) {
670 strv_free(l);
671 strv_free_free(t);
672 return log_oom();
673 }
674
675 } else {
676 char **c;
677
678 c = strv_copy(*a);
679 if (!c) {
680 strv_free(l);
681 strv_free_free(t);
682 return log_oom();
683 }
684
685 t[n++] = c;
686 }
687 }
688
689 t[n++] = strv_uniq(l);
690
691 strv_free_free(arg_join_controllers);
692 arg_join_controllers = t;
693 }
694 }
695 if (!isempty(rvalue))
696 log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring.");
697
698 return 0;
699 }
700
701 static int parse_config_file(void) {
702
703 const ConfigTableItem items[] = {
704 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
705 { "Manager", "LogTarget", config_parse_target, 0, NULL },
706 { "Manager", "LogColor", config_parse_color, 0, NULL },
707 { "Manager", "LogLocation", config_parse_location, 0, NULL },
708 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
709 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL },
710 { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL },
711 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
712 { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
713 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
714 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
715 { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
716 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
717 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
718 { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
719 #ifdef HAVE_SECCOMP
720 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
721 #endif
722 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
723 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
724 { "Manager", "DefaultStandardOutput", config_parse_output, 0, &arg_default_std_output },
725 { "Manager", "DefaultStandardError", config_parse_output, 0, &arg_default_std_error },
726 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
727 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
728 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
729 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */
730 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec, 0, &arg_default_start_limit_interval },
731 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
732 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
733 { "Manager", "DefaultLimitCPU", config_parse_limit, RLIMIT_CPU, arg_default_rlimit },
734 { "Manager", "DefaultLimitFSIZE", config_parse_limit, RLIMIT_FSIZE, arg_default_rlimit },
735 { "Manager", "DefaultLimitDATA", config_parse_limit, RLIMIT_DATA, arg_default_rlimit },
736 { "Manager", "DefaultLimitSTACK", config_parse_limit, RLIMIT_STACK, arg_default_rlimit },
737 { "Manager", "DefaultLimitCORE", config_parse_limit, RLIMIT_CORE, arg_default_rlimit },
738 { "Manager", "DefaultLimitRSS", config_parse_limit, RLIMIT_RSS, arg_default_rlimit },
739 { "Manager", "DefaultLimitNOFILE", config_parse_limit, RLIMIT_NOFILE, arg_default_rlimit },
740 { "Manager", "DefaultLimitAS", config_parse_limit, RLIMIT_AS, arg_default_rlimit },
741 { "Manager", "DefaultLimitNPROC", config_parse_limit, RLIMIT_NPROC, arg_default_rlimit },
742 { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, RLIMIT_MEMLOCK, arg_default_rlimit },
743 { "Manager", "DefaultLimitLOCKS", config_parse_limit, RLIMIT_LOCKS, arg_default_rlimit },
744 { "Manager", "DefaultLimitSIGPENDING", config_parse_limit, RLIMIT_SIGPENDING, arg_default_rlimit },
745 { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, RLIMIT_MSGQUEUE, arg_default_rlimit },
746 { "Manager", "DefaultLimitNICE", config_parse_limit, RLIMIT_NICE, arg_default_rlimit },
747 { "Manager", "DefaultLimitRTPRIO", config_parse_limit, RLIMIT_RTPRIO, arg_default_rlimit },
748 { "Manager", "DefaultLimitRTTIME", config_parse_limit, RLIMIT_RTTIME, arg_default_rlimit },
749 { "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_default_cpu_accounting },
750 { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting },
751 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
752 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
753 { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
754 { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max },
755 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action, 0, &arg_cad_burst_action },
756 {}
757 };
758
759 const char *fn, *conf_dirs_nulstr;
760
761 fn = arg_system ?
762 PKGSYSCONFDIR "/system.conf" :
763 PKGSYSCONFDIR "/user.conf";
764
765 conf_dirs_nulstr = arg_system ?
766 CONF_PATHS_NULSTR("systemd/system.conf.d") :
767 CONF_PATHS_NULSTR("systemd/user.conf.d");
768
769 config_parse_many_nulstr(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, false, NULL);
770
771 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
772 * like everywhere else. */
773 if (arg_default_timeout_start_usec <= 0)
774 arg_default_timeout_start_usec = USEC_INFINITY;
775 if (arg_default_timeout_stop_usec <= 0)
776 arg_default_timeout_stop_usec = USEC_INFINITY;
777
778 return 0;
779 }
780
781 static void manager_set_defaults(Manager *m) {
782
783 assert(m);
784
785 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
786 m->default_std_output = arg_default_std_output;
787 m->default_std_error = arg_default_std_error;
788 m->default_timeout_start_usec = arg_default_timeout_start_usec;
789 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
790 m->default_restart_usec = arg_default_restart_usec;
791 m->default_start_limit_interval = arg_default_start_limit_interval;
792 m->default_start_limit_burst = arg_default_start_limit_burst;
793 m->default_cpu_accounting = arg_default_cpu_accounting;
794 m->default_io_accounting = arg_default_io_accounting;
795 m->default_blockio_accounting = arg_default_blockio_accounting;
796 m->default_memory_accounting = arg_default_memory_accounting;
797 m->default_tasks_accounting = arg_default_tasks_accounting;
798 m->default_tasks_max = arg_default_tasks_max;
799
800 manager_set_default_rlimits(m, arg_default_rlimit);
801 manager_environment_add(m, NULL, arg_default_environment);
802 }
803
804 static int parse_argv(int argc, char *argv[]) {
805
806 enum {
807 ARG_LOG_LEVEL = 0x100,
808 ARG_LOG_TARGET,
809 ARG_LOG_COLOR,
810 ARG_LOG_LOCATION,
811 ARG_UNIT,
812 ARG_SYSTEM,
813 ARG_USER,
814 ARG_TEST,
815 ARG_NO_PAGER,
816 ARG_VERSION,
817 ARG_DUMP_CONFIGURATION_ITEMS,
818 ARG_DUMP_CORE,
819 ARG_CRASH_CHVT,
820 ARG_CRASH_SHELL,
821 ARG_CRASH_REBOOT,
822 ARG_CONFIRM_SPAWN,
823 ARG_SHOW_STATUS,
824 ARG_DESERIALIZE,
825 ARG_SWITCHED_ROOT,
826 ARG_DEFAULT_STD_OUTPUT,
827 ARG_DEFAULT_STD_ERROR,
828 ARG_MACHINE_ID
829 };
830
831 static const struct option options[] = {
832 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
833 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
834 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
835 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
836 { "unit", required_argument, NULL, ARG_UNIT },
837 { "system", no_argument, NULL, ARG_SYSTEM },
838 { "user", no_argument, NULL, ARG_USER },
839 { "test", no_argument, NULL, ARG_TEST },
840 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
841 { "help", no_argument, NULL, 'h' },
842 { "version", no_argument, NULL, ARG_VERSION },
843 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
844 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
845 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
846 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
847 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
848 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
849 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
850 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
851 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
852 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
853 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
854 { "machine-id", required_argument, NULL, ARG_MACHINE_ID },
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 log_error("Failed to parse log level %s.", optarg);
874 return r;
875 }
876
877 break;
878
879 case ARG_LOG_TARGET:
880 r = log_set_target_from_string(optarg);
881 if (r < 0) {
882 log_error("Failed to parse log target %s.", optarg);
883 return r;
884 }
885
886 break;
887
888 case ARG_LOG_COLOR:
889
890 if (optarg) {
891 r = log_show_color_from_string(optarg);
892 if (r < 0) {
893 log_error("Failed to parse log color setting %s.", optarg);
894 return r;
895 }
896 } else
897 log_show_color(true);
898
899 break;
900
901 case ARG_LOG_LOCATION:
902 if (optarg) {
903 r = log_show_location_from_string(optarg);
904 if (r < 0) {
905 log_error("Failed to parse log location setting %s.", optarg);
906 return r;
907 }
908 } else
909 log_show_location(true);
910
911 break;
912
913 case ARG_DEFAULT_STD_OUTPUT:
914 r = exec_output_from_string(optarg);
915 if (r < 0) {
916 log_error("Failed to parse default standard output setting %s.", optarg);
917 return r;
918 } else
919 arg_default_std_output = r;
920 break;
921
922 case ARG_DEFAULT_STD_ERROR:
923 r = exec_output_from_string(optarg);
924 if (r < 0) {
925 log_error("Failed to parse default standard error output setting %s.", optarg);
926 return r;
927 } else
928 arg_default_std_error = r;
929 break;
930
931 case ARG_UNIT:
932 r = free_and_strdup(&arg_default_unit, optarg);
933 if (r < 0)
934 return log_error_errno(r, "Failed to set default unit %s: %m", optarg);
935
936 break;
937
938 case ARG_SYSTEM:
939 arg_system = true;
940 break;
941
942 case ARG_USER:
943 arg_system = false;
944 break;
945
946 case ARG_TEST:
947 arg_action = ACTION_TEST;
948 break;
949
950 case ARG_NO_PAGER:
951 arg_no_pager = true;
952 break;
953
954 case ARG_VERSION:
955 arg_action = ACTION_VERSION;
956 break;
957
958 case ARG_DUMP_CONFIGURATION_ITEMS:
959 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
960 break;
961
962 case ARG_DUMP_CORE:
963 if (!optarg)
964 arg_dump_core = true;
965 else {
966 r = parse_boolean(optarg);
967 if (r < 0)
968 return log_error_errno(r, "Failed to parse dump core boolean: %s", optarg);
969 arg_dump_core = r;
970 }
971 break;
972
973 case ARG_CRASH_CHVT:
974 r = parse_crash_chvt(optarg);
975 if (r < 0)
976 return log_error_errno(r, "Failed to parse crash virtual terminal index: %s", optarg);
977 break;
978
979 case ARG_CRASH_SHELL:
980 if (!optarg)
981 arg_crash_shell = true;
982 else {
983 r = parse_boolean(optarg);
984 if (r < 0)
985 return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg);
986 arg_crash_shell = r;
987 }
988 break;
989
990 case ARG_CRASH_REBOOT:
991 if (!optarg)
992 arg_crash_reboot = true;
993 else {
994 r = parse_boolean(optarg);
995 if (r < 0)
996 return log_error_errno(r, "Failed to parse crash shell boolean: %s", 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: %m");
1007 break;
1008
1009 case ARG_SHOW_STATUS:
1010 if (optarg) {
1011 r = parse_show_status(optarg, &arg_show_status);
1012 if (r < 0) {
1013 log_error("Failed to parse show status boolean %s.", optarg);
1014 return r;
1015 }
1016 } else
1017 arg_show_status = SHOW_STATUS_YES;
1018 break;
1019
1020 case ARG_DESERIALIZE: {
1021 int fd;
1022 FILE *f;
1023
1024 r = safe_atoi(optarg, &fd);
1025 if (r < 0 || fd < 0) {
1026 log_error("Failed to parse deserialize option %s.", optarg);
1027 return -EINVAL;
1028 }
1029
1030 (void) fd_cloexec(fd, true);
1031
1032 f = fdopen(fd, "r");
1033 if (!f)
1034 return log_error_errno(errno, "Failed to open serialization fd: %m");
1035
1036 safe_fclose(arg_serialization);
1037 arg_serialization = f;
1038
1039 break;
1040 }
1041
1042 case ARG_SWITCHED_ROOT:
1043 arg_switched_root = true;
1044 break;
1045
1046 case ARG_MACHINE_ID:
1047 r = set_machine_id(optarg);
1048 if (r < 0)
1049 return log_error_errno(r, "MachineID '%s' is not valid.", optarg);
1050 break;
1051
1052 case 'h':
1053 arg_action = ACTION_HELP;
1054 break;
1055
1056 case 'D':
1057 log_set_max_level(LOG_DEBUG);
1058 break;
1059
1060 case 'b':
1061 case 's':
1062 case 'z':
1063 /* Just to eat away the sysvinit kernel
1064 * cmdline args without getopt() error
1065 * messages that we'll parse in
1066 * parse_proc_cmdline_word() or ignore. */
1067
1068 case '?':
1069 if (getpid_cached() != 1)
1070 return -EINVAL;
1071 else
1072 return 0;
1073
1074 default:
1075 assert_not_reached("Unhandled option code.");
1076 }
1077
1078 if (optind < argc && getpid_cached() != 1) {
1079 /* Hmm, when we aren't run as init system
1080 * let's complain about excess arguments */
1081
1082 log_error("Excess arguments.");
1083 return -EINVAL;
1084 }
1085
1086 return 0;
1087 }
1088
1089 static int help(void) {
1090
1091 printf("%s [OPTIONS...]\n\n"
1092 "Starts up and maintains the system or user services.\n\n"
1093 " -h --help Show this help\n"
1094 " --version Show version\n"
1095 " --test Determine startup sequence, dump it and exit\n"
1096 " --no-pager Do not pipe output into a pager\n"
1097 " --dump-configuration-items Dump understood unit configuration items\n"
1098 " --unit=UNIT Set default unit\n"
1099 " --system Run a system instance, even if PID != 1\n"
1100 " --user Run a user instance\n"
1101 " --dump-core[=BOOL] Dump core on crash\n"
1102 " --crash-vt=NR Change to specified VT on crash\n"
1103 " --crash-reboot[=BOOL] Reboot on crash\n"
1104 " --crash-shell[=BOOL] Run shell on crash\n"
1105 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1106 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1107 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1108 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1109 " --log-color[=BOOL] Highlight important log messages\n"
1110 " --log-location[=BOOL] Include code location in log messages\n"
1111 " --default-standard-output= Set default standard output for services\n"
1112 " --default-standard-error= Set default standard error output for services\n",
1113 program_invocation_short_name);
1114
1115 return 0;
1116 }
1117
1118 static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
1119 _cleanup_fdset_free_ FDSet *fds = NULL;
1120 _cleanup_fclose_ FILE *f = NULL;
1121 int r;
1122
1123 assert(m);
1124 assert(_f);
1125 assert(_fds);
1126
1127 r = manager_open_serialization(m, &f);
1128 if (r < 0)
1129 return log_error_errno(r, "Failed to create serialization file: %m");
1130
1131 /* Make sure nothing is really destructed when we shut down */
1132 m->n_reloading++;
1133 bus_manager_send_reloading(m, true);
1134
1135 fds = fdset_new();
1136 if (!fds)
1137 return log_oom();
1138
1139 r = manager_serialize(m, f, fds, switching_root);
1140 if (r < 0)
1141 return log_error_errno(r, "Failed to serialize state: %m");
1142
1143 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1144 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1145
1146 r = fd_cloexec(fileno(f), false);
1147 if (r < 0)
1148 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1149
1150 r = fdset_cloexec(fds, false);
1151 if (r < 0)
1152 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1153
1154 *_f = f;
1155 *_fds = fds;
1156
1157 f = NULL;
1158 fds = NULL;
1159
1160 return 0;
1161 }
1162
1163 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1164 struct rlimit nl;
1165 int r;
1166 int min_max;
1167 _cleanup_free_ char *nr_open = NULL;
1168
1169 assert(saved_rlimit);
1170
1171 /* Save the original RLIMIT_NOFILE so that we can reset it
1172 * later when transitioning from the initrd to the main
1173 * systemd or suchlike. */
1174 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1175 return log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1176
1177 /* Make sure forked processes get the default kernel setting */
1178 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1179 struct rlimit *rl;
1180
1181 rl = newdup(struct rlimit, saved_rlimit, 1);
1182 if (!rl)
1183 return log_oom();
1184
1185 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1186 }
1187
1188 /* Get current RLIMIT_NOFILE maximum compiled into the kernel. */
1189 r = read_one_line_file("/proc/sys/fs/nr_open", &nr_open);
1190 if (r == 0)
1191 r = safe_atoi(nr_open, &min_max);
1192 /* If we fail, fallback to the hard-coded kernel limit of 1024 * 1024. */
1193 if (r < 0)
1194 min_max = 1024 * 1024;
1195
1196 /* Bump up the resource limit for ourselves substantially */
1197 nl.rlim_cur = nl.rlim_max = min_max;
1198 r = setrlimit_closest(RLIMIT_NOFILE, &nl);
1199 if (r < 0)
1200 return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1201
1202 return 0;
1203 }
1204
1205 static void test_usr(void) {
1206
1207 /* Check that /usr is not a separate fs */
1208
1209 if (dir_is_empty("/usr") <= 0)
1210 return;
1211
1212 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1213 "Some things will probably break (sometimes even silently) in mysterious ways. "
1214 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1215 }
1216
1217 static int initialize_join_controllers(void) {
1218 /* By default, mount "cpu" + "cpuacct" together, and "net_cls"
1219 * + "net_prio". We'd like to add "cpuset" to the mix, but
1220 * "cpuset" doesn't really work for groups with no initialized
1221 * attributes. */
1222
1223 arg_join_controllers = new(char**, 3);
1224 if (!arg_join_controllers)
1225 return -ENOMEM;
1226
1227 arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL);
1228 if (!arg_join_controllers[0])
1229 goto oom;
1230
1231 arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL);
1232 if (!arg_join_controllers[1])
1233 goto oom;
1234
1235 arg_join_controllers[2] = NULL;
1236 return 0;
1237
1238 oom:
1239 arg_join_controllers = strv_free_free(arg_join_controllers);
1240 return -ENOMEM;
1241 }
1242
1243 static int enforce_syscall_archs(Set *archs) {
1244 #ifdef HAVE_SECCOMP
1245 int r;
1246
1247 if (!is_seccomp_available())
1248 return 0;
1249
1250 r = seccomp_restrict_archs(arg_syscall_archs);
1251 if (r < 0)
1252 return log_error_errno(r, "Failed to enforce system call architecture restrication: %m");
1253 #endif
1254 return 0;
1255 }
1256
1257 static int status_welcome(void) {
1258 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1259 int r;
1260
1261 r = parse_env_file("/etc/os-release", NEWLINE,
1262 "PRETTY_NAME", &pretty_name,
1263 "ANSI_COLOR", &ansi_color,
1264 NULL);
1265 if (r == -ENOENT)
1266 r = parse_env_file("/usr/lib/os-release", NEWLINE,
1267 "PRETTY_NAME", &pretty_name,
1268 "ANSI_COLOR", &ansi_color,
1269 NULL);
1270
1271 if (r < 0 && r != -ENOENT)
1272 log_warning_errno(r, "Failed to read os-release file: %m");
1273
1274 if (log_get_show_color())
1275 return status_printf(NULL, false, false,
1276 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1277 isempty(ansi_color) ? "1" : ansi_color,
1278 isempty(pretty_name) ? "Linux" : pretty_name);
1279 else
1280 return status_printf(NULL, false, false,
1281 "\nWelcome to %s!\n",
1282 isempty(pretty_name) ? "Linux" : pretty_name);
1283 }
1284
1285 static int write_container_id(void) {
1286 const char *c;
1287 int r;
1288
1289 c = getenv("container");
1290 if (isempty(c))
1291 return 0;
1292
1293 RUN_WITH_UMASK(0022)
1294 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1295 if (r < 0)
1296 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
1297
1298 return 1;
1299 }
1300
1301 static int bump_unix_max_dgram_qlen(void) {
1302 _cleanup_free_ char *qlen = NULL;
1303 unsigned long v;
1304 int r;
1305
1306 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel
1307 * default of 16 is simply too low. We set the value really
1308 * really early during boot, so that it is actually applied to
1309 * all our sockets, including the $NOTIFY_SOCKET one. */
1310
1311 r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1312 if (r < 0)
1313 return log_warning_errno(r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1314
1315 r = safe_atolu(qlen, &v);
1316 if (r < 0)
1317 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length, ignoring: %m");
1318
1319 if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1320 return 0;
1321
1322 qlen = mfree(qlen);
1323 if (asprintf(&qlen, "%lu\n", DEFAULT_UNIX_MAX_DGRAM_QLEN) < 0)
1324 return log_oom();
1325
1326 r = write_string_file("/proc/sys/net/unix/max_dgram_qlen", qlen, 0);
1327 if (r < 0)
1328 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1329 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1330
1331 return 1;
1332 }
1333
1334 static int fixup_environment(void) {
1335 _cleanup_free_ char *term = NULL;
1336 int r;
1337
1338 /* We expect the environment to be set correctly
1339 * if run inside a container. */
1340 if (detect_container() > 0)
1341 return 0;
1342
1343 /* When started as PID1, the kernel uses /dev/console
1344 * for our stdios and uses TERM=linux whatever the
1345 * backend device used by the console. We try to make
1346 * a better guess here since some consoles might not
1347 * have support for color mode for example.
1348 *
1349 * However if TERM was configured through the kernel
1350 * command line then leave it alone. */
1351
1352 r = proc_cmdline_get_key("TERM", 0, &term);
1353 if (r < 0)
1354 return r;
1355 if (r == 0) {
1356 term = strdup(default_term_for_tty("/dev/console"));
1357 if (!term)
1358 return -ENOMEM;
1359 }
1360
1361 if (setenv("TERM", term, 1) < 0)
1362 return -errno;
1363
1364 return 0;
1365 }
1366
1367 int main(int argc, char *argv[]) {
1368 Manager *m = NULL;
1369 int r, retval = EXIT_FAILURE;
1370 usec_t before_startup, after_startup;
1371 char timespan[FORMAT_TIMESPAN_MAX];
1372 FDSet *fds = NULL;
1373 bool reexecute = false;
1374 const char *shutdown_verb = NULL;
1375 dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL;
1376 dual_timestamp userspace_timestamp = DUAL_TIMESTAMP_NULL;
1377 dual_timestamp kernel_timestamp = DUAL_TIMESTAMP_NULL;
1378 dual_timestamp security_start_timestamp = DUAL_TIMESTAMP_NULL;
1379 dual_timestamp security_finish_timestamp = DUAL_TIMESTAMP_NULL;
1380 static char systemd[] = "systemd";
1381 bool skip_setup = false;
1382 unsigned j;
1383 bool loaded_policy = false;
1384 bool arm_reboot_watchdog = false;
1385 bool queue_default_job = false;
1386 bool empty_etc = false;
1387 char *switch_root_dir = NULL, *switch_root_init = NULL;
1388 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0);
1389 const char *error_message = NULL;
1390
1391 #ifdef HAVE_SYSV_COMPAT
1392 if (getpid_cached() != 1 && strstr(program_invocation_short_name, "init")) {
1393 /* This is compatibility support for SysV, where
1394 * calling init as a user is identical to telinit. */
1395
1396 execv(SYSTEMCTL_BINARY_PATH, argv);
1397 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1398 return 1;
1399 }
1400 #endif
1401
1402 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
1403 dual_timestamp_get(&userspace_timestamp);
1404
1405 /* Determine if this is a reexecution or normal bootup. We do
1406 * the full command line parsing much later, so let's just
1407 * have a quick peek here. */
1408 if (strv_find(argv+1, "--deserialize"))
1409 skip_setup = true;
1410
1411 /* If we have switched root, do all the special setup
1412 * things */
1413 if (strv_find(argv+1, "--switched-root"))
1414 skip_setup = false;
1415
1416 /* If we get started via the /sbin/init symlink then we are
1417 called 'init'. After a subsequent reexecution we are then
1418 called 'systemd'. That is confusing, hence let's call us
1419 systemd right-away. */
1420 program_invocation_short_name = systemd;
1421 (void) prctl(PR_SET_NAME, systemd);
1422
1423 saved_argv = argv;
1424 saved_argc = argc;
1425
1426 log_set_upgrade_syslog_to_journal(true);
1427
1428 if (getpid_cached() == 1) {
1429 /* Disable the umask logic */
1430 umask(0);
1431
1432 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
1433 * important so that we never end up logging to any foreign stderr, for example if we have to log in a
1434 * child process right before execve()'ing the actual binary, at a point in time where socket
1435 * activation stderr/stdout area already set up. */
1436 log_set_always_reopen_console(true);
1437 }
1438
1439 if (getpid_cached() == 1 && detect_container() <= 0) {
1440
1441 /* Running outside of a container as PID 1 */
1442 arg_system = true;
1443 log_set_target(LOG_TARGET_KMSG);
1444 log_open();
1445
1446 if (in_initrd())
1447 initrd_timestamp = userspace_timestamp;
1448
1449 if (!skip_setup) {
1450 r = mount_setup_early();
1451 if (r < 0) {
1452 error_message = "Failed to mount early API filesystems";
1453 goto finish;
1454 }
1455
1456 dual_timestamp_get(&security_start_timestamp);
1457 if (mac_selinux_setup(&loaded_policy) < 0) {
1458 error_message = "Failed to load SELinux policy";
1459 goto finish;
1460 } else if (mac_smack_setup(&loaded_policy) < 0) {
1461 error_message = "Failed to load SMACK policy";
1462 goto finish;
1463 } else if (ima_setup() < 0) {
1464 error_message = "Failed to load IMA policy";
1465 goto finish;
1466 }
1467 dual_timestamp_get(&security_finish_timestamp);
1468 }
1469
1470 if (mac_selinux_init() < 0) {
1471 error_message = "Failed to initialize SELinux policy";
1472 goto finish;
1473 }
1474
1475 if (!skip_setup) {
1476 if (clock_is_localtime(NULL) > 0) {
1477 int min;
1478
1479 /*
1480 * The very first call of settimeofday() also does a time warp in the kernel.
1481 *
1482 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on
1483 * external tools to take care of maintaining the RTC and do all adjustments.
1484 * This matches the behavior of Windows, which leaves the RTC alone if the
1485 * registry tells that the RTC runs in UTC.
1486 */
1487 r = clock_set_timezone(&min);
1488 if (r < 0)
1489 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1490 else
1491 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1492 } else if (!in_initrd()) {
1493 /*
1494 * Do a dummy very first call to seal the kernel's time warp magic.
1495 *
1496 * Do not call this from inside the initrd. The initrd might not
1497 * carry /etc/adjtime with LOCAL, but the real system could be set up
1498 * that way. In such case, we need to delay the time-warp or the sealing
1499 * until we reach the real system.
1500 *
1501 * Do no set the kernel's timezone. The concept of local time cannot
1502 * be supported reliably, the time will jump or be incorrect at every daylight
1503 * saving time change. All kernel local time concepts will be treated
1504 * as UTC that way.
1505 */
1506 (void) clock_reset_timewarp();
1507 }
1508
1509 r = clock_apply_epoch();
1510 if (r < 0)
1511 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1512 else if (r > 0)
1513 log_info("System time before build time, advancing clock.");
1514 }
1515
1516 /* Set the default for later on, but don't actually
1517 * open the logs like this for now. Note that if we
1518 * are transitioning from the initrd there might still
1519 * be journal fd open, and we shouldn't attempt
1520 * opening that before we parsed /proc/cmdline which
1521 * might redirect output elsewhere. */
1522 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
1523
1524 } else if (getpid_cached() == 1) {
1525 /* Running inside a container, as PID 1 */
1526 arg_system = true;
1527 log_set_target(LOG_TARGET_CONSOLE);
1528 log_close_console(); /* force reopen of /dev/console */
1529 log_open();
1530
1531 /* For later on, see above... */
1532 log_set_target(LOG_TARGET_JOURNAL);
1533
1534 /* clear the kernel timestamp,
1535 * because we are in a container */
1536 kernel_timestamp = DUAL_TIMESTAMP_NULL;
1537 } else {
1538 /* Running as user instance */
1539 arg_system = false;
1540 log_set_target(LOG_TARGET_AUTO);
1541 log_open();
1542
1543 /* clear the kernel timestamp,
1544 * because we are not PID 1 */
1545 kernel_timestamp = DUAL_TIMESTAMP_NULL;
1546 }
1547
1548 if (getpid_cached() == 1) {
1549 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1550 * will process core dumps for system services by default. */
1551 if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0)
1552 log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m");
1553
1554 /* But at the same time, turn off the core_pattern logic by default, so that no coredumps are stored
1555 * until the systemd-coredump tool is enabled via sysctl. */
1556 if (!skip_setup)
1557 (void) write_string_file("/proc/sys/kernel/core_pattern", "|/bin/false", 0);
1558 }
1559
1560 if (arg_system) {
1561 if (fixup_environment() < 0) {
1562 error_message = "Failed to fix up PID1 environment";
1563 goto finish;
1564 }
1565
1566 /* Try to figure out if we can use colors with the console. No
1567 * need to do that for user instances since they never log
1568 * into the console. */
1569 log_show_color(colors_enabled());
1570 r = make_null_stdio();
1571 if (r < 0)
1572 log_warning_errno(r, "Failed to redirect standard streams to /dev/null: %m");
1573 }
1574
1575 r = initialize_join_controllers();
1576 if (r < 0) {
1577 error_message = "Failed to initialize cgroup controllers";
1578 goto finish;
1579 }
1580
1581 /* Mount /proc, /sys and friends, so that /proc/cmdline and
1582 * /proc/$PID/fd is available. */
1583 if (getpid_cached() == 1) {
1584
1585 /* Load the kernel modules early. */
1586 if (!skip_setup)
1587 kmod_setup();
1588
1589 r = mount_setup(loaded_policy);
1590 if (r < 0) {
1591 error_message = "Failed to mount API filesystems";
1592 goto finish;
1593 }
1594 }
1595
1596 /* Reset all signal handlers. */
1597 (void) reset_all_signal_handlers();
1598 (void) ignore_signals(SIGNALS_IGNORE, -1);
1599
1600 arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U);
1601
1602 if (parse_config_file() < 0) {
1603 error_message = "Failed to parse config file";
1604 goto finish;
1605 }
1606
1607 if (arg_system) {
1608 r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
1609 if (r < 0)
1610 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
1611 }
1612
1613 /* Note that this also parses bits from the kernel command
1614 * line, including "debug". */
1615 log_parse_environment();
1616
1617 if (parse_argv(argc, argv) < 0) {
1618 error_message = "Failed to parse commandline arguments";
1619 goto finish;
1620 }
1621
1622 /* Initialize default unit */
1623 if (!arg_default_unit) {
1624 arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET);
1625 if (!arg_default_unit) {
1626 r = log_oom();
1627 error_message = "Failed to set default unit";
1628 goto finish;
1629 }
1630 }
1631
1632 if (arg_action == ACTION_TEST &&
1633 geteuid() == 0) {
1634 log_error("Don't run test mode as root.");
1635 goto finish;
1636 }
1637
1638 if (!arg_system &&
1639 arg_action == ACTION_RUN &&
1640 sd_booted() <= 0) {
1641 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
1642 goto finish;
1643 }
1644
1645 if (arg_system &&
1646 arg_action == ACTION_RUN &&
1647 running_in_chroot() > 0) {
1648 log_error("Cannot be run in a chroot() environment.");
1649 goto finish;
1650 }
1651
1652 if (arg_action == ACTION_TEST || arg_action == ACTION_HELP) {
1653 pager_open(arg_no_pager, false);
1654 skip_setup = true;
1655 }
1656
1657 if (arg_action == ACTION_HELP) {
1658 retval = help();
1659 goto finish;
1660 } else if (arg_action == ACTION_VERSION) {
1661 retval = version();
1662 goto finish;
1663 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
1664 pager_open(arg_no_pager, false);
1665 unit_dump_config_items(stdout);
1666 retval = EXIT_SUCCESS;
1667 goto finish;
1668 }
1669
1670 if (!arg_system &&
1671 !getenv("XDG_RUNTIME_DIR")) {
1672 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
1673 goto finish;
1674 }
1675
1676 assert_se(arg_action == ACTION_RUN || arg_action == ACTION_TEST);
1677
1678 /* Close logging fds, in order not to confuse fdset below */
1679 log_close();
1680
1681 /* Remember open file descriptors for later deserialization */
1682 if (arg_action == ACTION_RUN) {
1683 r = fdset_new_fill(&fds);
1684 if (r < 0) {
1685 log_emergency_errno(r, "Failed to allocate fd set: %m");
1686 error_message = "Failed to allocate fd set";
1687 goto finish;
1688 } else
1689 fdset_cloexec(fds, true);
1690
1691 if (arg_serialization)
1692 assert_se(fdset_remove(fds, fileno(arg_serialization)) >= 0);
1693
1694 if (arg_system)
1695 /* Become a session leader if we aren't one yet. */
1696 setsid();
1697 }
1698
1699 /* Move out of the way, so that we won't block unmounts */
1700 assert_se(chdir("/") == 0);
1701
1702 /* Reset the console, but only if this is really init and we
1703 * are freshly booted */
1704 if (arg_system && arg_action == ACTION_RUN) {
1705
1706 /* If we are init, we connect stdin/stdout/stderr to
1707 * /dev/null and make sure we don't have a controlling
1708 * tty. */
1709 release_terminal();
1710
1711 if (getpid_cached() == 1 && !skip_setup)
1712 console_setup();
1713 }
1714
1715 /* Open the logging devices, if possible and necessary */
1716 log_open();
1717
1718 if (arg_show_status == _SHOW_STATUS_UNSET)
1719 arg_show_status = SHOW_STATUS_YES;
1720
1721 /* Make sure we leave a core dump without panicing the
1722 * kernel. */
1723 if (getpid_cached() == 1) {
1724 install_crash_handler();
1725
1726 r = mount_cgroup_controllers(arg_join_controllers);
1727 if (r < 0)
1728 goto finish;
1729 }
1730
1731 if (arg_system) {
1732 int v;
1733
1734 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1735 arg_action == ACTION_TEST ? "test " : "" );
1736
1737 v = detect_virtualization();
1738 if (v > 0)
1739 log_info("Detected virtualization %s.", virtualization_to_string(v));
1740
1741 write_container_id();
1742
1743 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1744
1745 if (in_initrd())
1746 log_info("Running in initial RAM disk.");
1747
1748 /* Let's check whether /etc is already populated. We
1749 * don't actually really check for that, but use
1750 * /etc/machine-id as flag file. This allows container
1751 * managers and installers to provision a couple of
1752 * files already. If the container manager wants to
1753 * provision the machine ID itself it should pass
1754 * $container_uuid to PID 1. */
1755
1756 empty_etc = access("/etc/machine-id", F_OK) < 0;
1757 if (empty_etc)
1758 log_info("Running with unpopulated /etc.");
1759 } else {
1760 _cleanup_free_ char *t;
1761
1762 t = uid_to_name(getuid());
1763 log_debug(PACKAGE_STRING " running in %suser mode for user "UID_FMT"/%s. (" SYSTEMD_FEATURES ")",
1764 arg_action == ACTION_TEST ? " test" : "", getuid(), t);
1765 }
1766
1767 if (arg_action == ACTION_RUN) {
1768 if (arg_system && !skip_setup) {
1769 if (arg_show_status > 0)
1770 status_welcome();
1771
1772 hostname_setup();
1773 machine_id_setup(NULL, arg_machine_id, NULL);
1774 loopback_setup();
1775 bump_unix_max_dgram_qlen();
1776
1777 test_usr();
1778 }
1779
1780 if (arg_system && arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY)
1781 watchdog_set_timeout(&arg_runtime_watchdog);
1782
1783 if (arg_timer_slack_nsec != NSEC_INFINITY)
1784 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1785 log_error_errno(errno, "Failed to adjust timer slack: %m");
1786
1787 if (arg_system && !cap_test_all(arg_capability_bounding_set)) {
1788 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
1789 if (r < 0) {
1790 log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
1791 error_message = "Failed to drop capability bounding set of usermode helpers";
1792 goto finish;
1793 }
1794 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
1795 if (r < 0) {
1796 log_emergency_errno(r, "Failed to drop capability bounding set: %m");
1797 error_message = "Failed to drop capability bounding set";
1798 goto finish;
1799 }
1800 }
1801
1802 if (arg_syscall_archs) {
1803 r = enforce_syscall_archs(arg_syscall_archs);
1804 if (r < 0) {
1805 error_message = "Failed to set syscall architectures";
1806 goto finish;
1807 }
1808 }
1809
1810 if (!arg_system)
1811 /* Become reaper of our children */
1812 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
1813 log_warning_errno(errno, "Failed to make us a subreaper: %m");
1814
1815 if (arg_system)
1816 /* Bump up RLIMIT_NOFILE for systemd itself */
1817 (void) bump_rlimit_nofile(&saved_rlimit_nofile);
1818 }
1819
1820 r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER, arg_action == ACTION_TEST, &m);
1821 if (r < 0) {
1822 log_emergency_errno(r, "Failed to allocate manager object: %m");
1823 error_message = "Failed to allocate manager object";
1824 goto finish;
1825 }
1826
1827 m->confirm_spawn = arg_confirm_spawn;
1828 m->runtime_watchdog = arg_runtime_watchdog;
1829 m->shutdown_watchdog = arg_shutdown_watchdog;
1830 m->userspace_timestamp = userspace_timestamp;
1831 m->kernel_timestamp = kernel_timestamp;
1832 m->initrd_timestamp = initrd_timestamp;
1833 m->security_start_timestamp = security_start_timestamp;
1834 m->security_finish_timestamp = security_finish_timestamp;
1835 m->cad_burst_action = arg_cad_burst_action;
1836
1837 manager_set_defaults(m);
1838 manager_set_show_status(m, arg_show_status);
1839 manager_set_first_boot(m, empty_etc);
1840
1841 /* Remember whether we should queue the default job */
1842 queue_default_job = !arg_serialization || arg_switched_root;
1843
1844 before_startup = now(CLOCK_MONOTONIC);
1845
1846 r = manager_startup(m, arg_serialization, fds);
1847 if (r < 0) {
1848 log_error_errno(r, "Failed to fully start up daemon: %m");
1849 goto finish;
1850 }
1851
1852 /* This will close all file descriptors that were opened, but
1853 * not claimed by any unit. */
1854 fds = fdset_free(fds);
1855
1856 arg_serialization = safe_fclose(arg_serialization);
1857
1858 if (queue_default_job) {
1859 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
1860 Unit *target = NULL;
1861 Job *default_unit_job;
1862
1863 log_debug("Activating default unit: %s", arg_default_unit);
1864
1865 r = manager_load_unit(m, arg_default_unit, NULL, &error, &target);
1866 if (r < 0)
1867 log_error("Failed to load default target: %s", bus_error_message(&error, r));
1868 else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND)
1869 log_error_errno(target->load_error, "Failed to load default target: %m");
1870 else if (target->load_state == UNIT_MASKED)
1871 log_error("Default target masked.");
1872
1873 if (!target || target->load_state != UNIT_LOADED) {
1874 log_info("Trying to load rescue target...");
1875
1876 r = manager_load_unit(m, SPECIAL_RESCUE_TARGET, NULL, &error, &target);
1877 if (r < 0) {
1878 log_emergency("Failed to load rescue target: %s", bus_error_message(&error, r));
1879 error_message = "Failed to load rescue target";
1880 goto finish;
1881 } else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND) {
1882 log_emergency_errno(target->load_error, "Failed to load rescue target: %m");
1883 error_message = "Failed to load rescue target";
1884 goto finish;
1885 } else if (target->load_state == UNIT_MASKED) {
1886 log_emergency("Rescue target masked.");
1887 error_message = "Rescue target masked";
1888 goto finish;
1889 }
1890 }
1891
1892 assert(target->load_state == UNIT_LOADED);
1893
1894 if (arg_action == ACTION_TEST) {
1895 printf("-> By units:\n");
1896 manager_dump_units(m, stdout, "\t");
1897 }
1898
1899 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, &error, &default_unit_job);
1900 if (r == -EPERM) {
1901 log_debug("Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1902
1903 sd_bus_error_free(&error);
1904
1905 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, &error, &default_unit_job);
1906 if (r < 0) {
1907 log_emergency("Failed to start default target: %s", bus_error_message(&error, r));
1908 error_message = "Failed to start default target";
1909 goto finish;
1910 }
1911 } else if (r < 0) {
1912 log_emergency("Failed to isolate default target: %s", bus_error_message(&error, r));
1913 error_message = "Failed to isolate default target";
1914 goto finish;
1915 }
1916
1917 m->default_unit_job_id = default_unit_job->id;
1918
1919 after_startup = now(CLOCK_MONOTONIC);
1920 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
1921 "Loaded units and determined initial transaction in %s.",
1922 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
1923
1924 if (arg_action == ACTION_TEST) {
1925 printf("-> By jobs:\n");
1926 manager_dump_jobs(m, stdout, "\t");
1927 retval = EXIT_SUCCESS;
1928 goto finish;
1929 }
1930 }
1931
1932 for (;;) {
1933 r = manager_loop(m);
1934 if (r < 0) {
1935 log_emergency_errno(r, "Failed to run main loop: %m");
1936 error_message = "Failed to run main loop";
1937 goto finish;
1938 }
1939
1940 switch (m->exit_code) {
1941
1942 case MANAGER_RELOAD:
1943 log_info("Reloading.");
1944
1945 r = parse_config_file();
1946 if (r < 0)
1947 log_error("Failed to parse config file.");
1948
1949 manager_set_defaults(m);
1950
1951 r = manager_reload(m);
1952 if (r < 0)
1953 log_error_errno(r, "Failed to reload: %m");
1954 break;
1955
1956 case MANAGER_REEXECUTE:
1957
1958 if (prepare_reexecute(m, &arg_serialization, &fds, false) < 0) {
1959 error_message = "Failed to prepare for reexecution";
1960 goto finish;
1961 }
1962
1963 reexecute = true;
1964 log_notice("Reexecuting.");
1965 goto finish;
1966
1967 case MANAGER_SWITCH_ROOT:
1968 /* Steal the switch root parameters */
1969 switch_root_dir = m->switch_root;
1970 switch_root_init = m->switch_root_init;
1971 m->switch_root = m->switch_root_init = NULL;
1972
1973 if (!switch_root_init)
1974 if (prepare_reexecute(m, &arg_serialization, &fds, true) < 0) {
1975 error_message = "Failed to prepare for reexecution";
1976 goto finish;
1977 }
1978
1979 reexecute = true;
1980 log_notice("Switching root.");
1981 goto finish;
1982
1983 case MANAGER_EXIT:
1984 retval = m->return_value;
1985
1986 if (MANAGER_IS_USER(m)) {
1987 log_debug("Exit.");
1988 goto finish;
1989 }
1990
1991 /* fallthrough */
1992 case MANAGER_REBOOT:
1993 case MANAGER_POWEROFF:
1994 case MANAGER_HALT:
1995 case MANAGER_KEXEC: {
1996 static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
1997 [MANAGER_EXIT] = "exit",
1998 [MANAGER_REBOOT] = "reboot",
1999 [MANAGER_POWEROFF] = "poweroff",
2000 [MANAGER_HALT] = "halt",
2001 [MANAGER_KEXEC] = "kexec"
2002 };
2003
2004 assert_se(shutdown_verb = table[m->exit_code]);
2005 arm_reboot_watchdog = m->exit_code == MANAGER_REBOOT;
2006
2007 log_notice("Shutting down.");
2008 goto finish;
2009 }
2010
2011 default:
2012 assert_not_reached("Unknown exit code.");
2013 }
2014 }
2015
2016 finish:
2017 pager_close();
2018
2019 if (m)
2020 arg_shutdown_watchdog = m->shutdown_watchdog;
2021
2022 m = manager_free(m);
2023
2024 for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++)
2025 arg_default_rlimit[j] = mfree(arg_default_rlimit[j]);
2026
2027 arg_default_unit = mfree(arg_default_unit);
2028 arg_confirm_spawn = mfree(arg_confirm_spawn);
2029 arg_join_controllers = strv_free_free(arg_join_controllers);
2030 arg_default_environment = strv_free(arg_default_environment);
2031 arg_syscall_archs = set_free(arg_syscall_archs);
2032
2033 mac_selinux_finish();
2034
2035 if (reexecute) {
2036 const char **args;
2037 unsigned i, args_size;
2038
2039 /* Close and disarm the watchdog, so that the new
2040 * instance can reinitialize it, but doesn't get
2041 * rebooted while we do that */
2042 watchdog_close(true);
2043
2044 /* Reset the RLIMIT_NOFILE to the kernel default, so
2045 * that the new systemd can pass the kernel default to
2046 * its child processes */
2047 if (saved_rlimit_nofile.rlim_cur > 0)
2048 (void) setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile);
2049
2050 if (switch_root_dir) {
2051 /* Kill all remaining processes from the
2052 * initrd, but don't wait for them, so that we
2053 * can handle the SIGCHLD for them after
2054 * deserializing. */
2055 broadcast_signal(SIGTERM, false, true);
2056
2057 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
2058 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
2059 if (r < 0)
2060 log_error_errno(r, "Failed to switch root, trying to continue: %m");
2061 }
2062
2063 args_size = MAX(6, argc+1);
2064 args = newa(const char*, args_size);
2065
2066 if (!switch_root_init) {
2067 char sfd[DECIMAL_STR_MAX(int) + 1];
2068
2069 /* First try to spawn ourselves with the right
2070 * path, and with full serialization. We do
2071 * this only if the user didn't specify an
2072 * explicit init to spawn. */
2073
2074 assert(arg_serialization);
2075 assert(fds);
2076
2077 xsprintf(sfd, "%i", fileno(arg_serialization));
2078
2079 i = 0;
2080 args[i++] = SYSTEMD_BINARY_PATH;
2081 if (switch_root_dir)
2082 args[i++] = "--switched-root";
2083 args[i++] = arg_system ? "--system" : "--user";
2084 args[i++] = "--deserialize";
2085 args[i++] = sfd;
2086 args[i++] = NULL;
2087
2088 assert(i <= args_size);
2089
2090 /*
2091 * We want valgrind to print its memory usage summary before reexecution.
2092 * Valgrind won't do this is on its own on exec(), but it will do it on exit().
2093 * Hence, to ensure we get a summary here, fork() off a child, let it exit() cleanly,
2094 * so that it prints the summary, and wait() for it in the parent, before proceeding into the exec().
2095 */
2096 valgrind_summary_hack();
2097
2098 (void) execv(args[0], (char* const*) args);
2099 }
2100
2101 /* Try the fallback, if there is any, without any
2102 * serialization. We pass the original argv[] and
2103 * envp[]. (Well, modulo the ordering changes due to
2104 * getopt() in argv[], and some cleanups in envp[],
2105 * but let's hope that doesn't matter.) */
2106
2107 arg_serialization = safe_fclose(arg_serialization);
2108 fds = fdset_free(fds);
2109
2110 /* Reopen the console */
2111 (void) make_console_stdio();
2112
2113 for (j = 1, i = 1; j < (unsigned) argc; j++)
2114 args[i++] = argv[j];
2115 args[i++] = NULL;
2116 assert(i <= args_size);
2117
2118 /* Reenable any blocked signals, especially important
2119 * if we switch from initial ramdisk to init=... */
2120 (void) reset_all_signal_handlers();
2121 (void) reset_signal_mask();
2122
2123 if (switch_root_init) {
2124 args[0] = switch_root_init;
2125 (void) execv(args[0], (char* const*) args);
2126 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
2127 }
2128
2129 args[0] = "/sbin/init";
2130 (void) execv(args[0], (char* const*) args);
2131
2132 if (errno == ENOENT) {
2133 log_warning("No /sbin/init, trying fallback");
2134
2135 args[0] = "/bin/sh";
2136 args[1] = NULL;
2137 (void) execv(args[0], (char* const*) args);
2138 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
2139 } else
2140 log_warning_errno(errno, "Failed to execute /sbin/init, giving up: %m");
2141 }
2142
2143 arg_serialization = safe_fclose(arg_serialization);
2144 fds = fdset_free(fds);
2145
2146 #ifdef HAVE_VALGRIND_VALGRIND_H
2147 /* If we are PID 1 and running under valgrind, then let's exit
2148 * here explicitly. valgrind will only generate nice output on
2149 * exit(), not on exec(), hence let's do the former not the
2150 * latter here. */
2151 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND)
2152 return 0;
2153 #endif
2154
2155 if (shutdown_verb) {
2156 char log_level[DECIMAL_STR_MAX(int) + 1];
2157 char exit_code[DECIMAL_STR_MAX(uint8_t) + 1];
2158 const char* command_line[11] = {
2159 SYSTEMD_SHUTDOWN_BINARY_PATH,
2160 shutdown_verb,
2161 "--log-level", log_level,
2162 "--log-target",
2163 };
2164 unsigned pos = 5;
2165 _cleanup_strv_free_ char **env_block = NULL;
2166
2167 assert(command_line[pos] == NULL);
2168 env_block = strv_copy(environ);
2169
2170 xsprintf(log_level, "%d", log_get_max_level());
2171
2172 switch (log_get_target()) {
2173
2174 case LOG_TARGET_KMSG:
2175 case LOG_TARGET_JOURNAL_OR_KMSG:
2176 case LOG_TARGET_SYSLOG_OR_KMSG:
2177 command_line[pos++] = "kmsg";
2178 break;
2179
2180 case LOG_TARGET_NULL:
2181 command_line[pos++] = "null";
2182 break;
2183
2184 case LOG_TARGET_CONSOLE:
2185 default:
2186 command_line[pos++] = "console";
2187 break;
2188 };
2189
2190 if (log_get_show_color())
2191 command_line[pos++] = "--log-color";
2192
2193 if (log_get_show_location())
2194 command_line[pos++] = "--log-location";
2195
2196 if (streq(shutdown_verb, "exit")) {
2197 command_line[pos++] = "--exit-code";
2198 command_line[pos++] = exit_code;
2199 xsprintf(exit_code, "%d", retval);
2200 }
2201
2202 assert(pos < ELEMENTSOF(command_line));
2203
2204 if (arm_reboot_watchdog && arg_shutdown_watchdog > 0 && arg_shutdown_watchdog != USEC_INFINITY) {
2205 char *e;
2206
2207 /* If we reboot let's set the shutdown
2208 * watchdog and tell the shutdown binary to
2209 * repeatedly ping it */
2210 r = watchdog_set_timeout(&arg_shutdown_watchdog);
2211 watchdog_close(r < 0);
2212
2213 /* Tell the binary how often to ping, ignore failure */
2214 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
2215 (void) strv_push(&env_block, e);
2216 } else
2217 watchdog_close(true);
2218
2219 /* Avoid the creation of new processes forked by the
2220 * kernel; at this point, we will not listen to the
2221 * signals anyway */
2222 if (detect_container() <= 0)
2223 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
2224
2225 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
2226 log_error_errno(errno, "Failed to execute shutdown binary, %s: %m",
2227 getpid_cached() == 1 ? "freezing" : "quitting");
2228 }
2229
2230 if (getpid_cached() == 1) {
2231 if (error_message)
2232 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2233 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
2234 "%s, freezing.", error_message);
2235 freeze_or_reboot();
2236 }
2237
2238 return retval;
2239 }
Unnamed repository; edit this file 'description' to name the repository.