]> 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
tree-wide: make parse_proc_cmdline() strip "rd." prefix automatically
[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 "env-util.h"
53 #include "fd-util.h"
54 #include "fdset.h"
55 #include "fileio.h"
56 #include "formats-util.h"
57 #include "fs-util.h"
58 #include "hostname-setup.h"
59 #include "ima-setup.h"
60 #include "killall.h"
61 #include "kmod-setup.h"
62 #include "load-fragment.h"
63 #include "log.h"
64 #include "loopback-setup.h"
65 #include "machine-id-setup.h"
66 #include "manager.h"
67 #include "missing.h"
68 #include "mount-setup.h"
69 #include "pager.h"
70 #include "parse-util.h"
71 #include "proc-cmdline.h"
72 #include "process-util.h"
73 #include "raw-clone.h"
74 #include "rlimit-util.h"
75 #ifdef HAVE_SECCOMP
76 #include "seccomp-util.h"
77 #endif
78 #include "selinux-setup.h"
79 #include "selinux-util.h"
80 #include "signal-util.h"
81 #include "smack-setup.h"
82 #include "special.h"
83 #include "stat-util.h"
84 #include "stdio-util.h"
85 #include "strv.h"
86 #include "switch-root.h"
87 #include "terminal-util.h"
88 #include "umask-util.h"
89 #include "user-util.h"
90 #include "virt.h"
91 #include "watchdog.h"
92
93 static enum {
94 ACTION_RUN,
95 ACTION_HELP,
96 ACTION_VERSION,
97 ACTION_TEST,
98 ACTION_DUMP_CONFIGURATION_ITEMS
99 } arg_action = ACTION_RUN;
100 static char *arg_default_unit = NULL;
101 static bool arg_system = false;
102 static bool arg_dump_core = true;
103 static int arg_crash_chvt = -1;
104 static bool arg_crash_shell = false;
105 static bool arg_crash_reboot = false;
106 static bool arg_confirm_spawn = false;
107 static ShowStatus arg_show_status = _SHOW_STATUS_UNSET;
108 static bool arg_switched_root = false;
109 static bool arg_no_pager = false;
110 static char ***arg_join_controllers = NULL;
111 static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
112 static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
113 static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
114 static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
115 static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
116 static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
117 static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
118 static usec_t arg_runtime_watchdog = 0;
119 static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
120 static char **arg_default_environment = NULL;
121 static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {};
122 static uint64_t arg_capability_bounding_set = CAP_ALL;
123 static nsec_t arg_timer_slack_nsec = NSEC_INFINITY;
124 static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
125 static Set* arg_syscall_archs = NULL;
126 static FILE* arg_serialization = NULL;
127 static bool arg_default_cpu_accounting = false;
128 static bool arg_default_io_accounting = false;
129 static bool arg_default_blockio_accounting = false;
130 static bool arg_default_memory_accounting = false;
131 static bool arg_default_tasks_accounting = true;
132 static uint64_t arg_default_tasks_max = UINT64_MAX;
133 static sd_id128_t arg_machine_id = {};
134 static CADBurstAction arg_cad_burst_action = CAD_BURST_ACTION_REBOOT;
135
136 noreturn static void freeze_or_reboot(void) {
137
138 if (arg_crash_reboot) {
139 log_notice("Rebooting in 10s...");
140 (void) sleep(10);
141
142 log_notice("Rebooting now...");
143 (void) reboot(RB_AUTOBOOT);
144 log_emergency_errno(errno, "Failed to reboot: %m");
145 }
146
147 log_emergency("Freezing execution.");
148 freeze();
149 }
150
151 noreturn static void crash(int sig) {
152 struct sigaction sa;
153 pid_t pid;
154
155 if (getpid() != 1)
156 /* Pass this on immediately, if this is not PID 1 */
157 (void) raise(sig);
158 else if (!arg_dump_core)
159 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
160 else {
161 sa = (struct sigaction) {
162 .sa_handler = nop_signal_handler,
163 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
164 };
165
166 /* We want to wait for the core process, hence let's enable SIGCHLD */
167 (void) sigaction(SIGCHLD, &sa, NULL);
168
169 pid = raw_clone(SIGCHLD);
170 if (pid < 0)
171 log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
172 else if (pid == 0) {
173 /* Enable default signal handler for core dump */
174
175 sa = (struct sigaction) {
176 .sa_handler = SIG_DFL,
177 };
178 (void) sigaction(sig, &sa, NULL);
179
180 /* Don't limit the coredump size */
181 (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
182
183 /* Just to be sure... */
184 (void) chdir("/");
185
186 /* Raise the signal again */
187 pid = raw_getpid();
188 (void) kill(pid, sig); /* raise() would kill the parent */
189
190 assert_not_reached("We shouldn't be here...");
191 _exit(EXIT_FAILURE);
192 } else {
193 siginfo_t status;
194 int r;
195
196 /* Order things nicely. */
197 r = wait_for_terminate(pid, &status);
198 if (r < 0)
199 log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
200 else if (status.si_code != CLD_DUMPED)
201 log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
202 signal_to_string(sig),
203 pid, sigchld_code_to_string(status.si_code),
204 status.si_status,
205 strna(status.si_code == CLD_EXITED
206 ? exit_status_to_string(status.si_status, EXIT_STATUS_MINIMAL)
207 : signal_to_string(status.si_status)));
208 else
209 log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
210 }
211 }
212
213 if (arg_crash_chvt >= 0)
214 (void) chvt(arg_crash_chvt);
215
216 sa = (struct sigaction) {
217 .sa_handler = SIG_IGN,
218 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
219 };
220
221 /* Let the kernel reap children for us */
222 (void) sigaction(SIGCHLD, &sa, NULL);
223
224 if (arg_crash_shell) {
225 log_notice("Executing crash shell in 10s...");
226 (void) sleep(10);
227
228 pid = raw_clone(SIGCHLD);
229 if (pid < 0)
230 log_emergency_errno(errno, "Failed to fork off crash shell: %m");
231 else if (pid == 0) {
232 (void) setsid();
233 (void) make_console_stdio();
234 (void) execle("/bin/sh", "/bin/sh", NULL, environ);
235
236 log_emergency_errno(errno, "execle() failed: %m");
237 _exit(EXIT_FAILURE);
238 } else {
239 log_info("Spawned crash shell as PID "PID_FMT".", pid);
240 (void) wait_for_terminate(pid, NULL);
241 }
242 }
243
244 freeze_or_reboot();
245 }
246
247 static void install_crash_handler(void) {
248 static const struct sigaction sa = {
249 .sa_handler = crash,
250 .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
251 };
252 int r;
253
254 /* We ignore the return value here, since, we don't mind if we
255 * cannot set up a crash handler */
256 r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
257 if (r < 0)
258 log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
259 }
260
261 static int console_setup(void) {
262 _cleanup_close_ int tty_fd = -1;
263 int r;
264
265 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
266 if (tty_fd < 0)
267 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
268
269 /* We don't want to force text mode. plymouth may be showing
270 * pictures already from initrd. */
271 r = reset_terminal_fd(tty_fd, false);
272 if (r < 0)
273 return log_error_errno(r, "Failed to reset /dev/console: %m");
274
275 return 0;
276 }
277
278 static int parse_crash_chvt(const char *value) {
279 int b;
280
281 if (safe_atoi(value, &arg_crash_chvt) >= 0)
282 return 0;
283
284 b = parse_boolean(value);
285 if (b < 0)
286 return b;
287
288 if (b > 0)
289 arg_crash_chvt = 0; /* switch to where kmsg goes */
290 else
291 arg_crash_chvt = -1; /* turn off switching */
292
293 return 0;
294 }
295
296 static int set_machine_id(const char *m) {
297 sd_id128_t t;
298 assert(m);
299
300 if (sd_id128_from_string(m, &t) < 0)
301 return -EINVAL;
302
303 if (sd_id128_is_null(t))
304 return -EINVAL;
305
306 arg_machine_id = t;
307 return 0;
308 }
309
310 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
311
312 int r;
313
314 assert(key);
315
316 if (streq(key, "systemd.unit") && value) {
317
318 if (!in_initrd())
319 return free_and_strdup(&arg_default_unit, value);
320
321 } else if (streq(key, "rd.systemd.unit") && value) {
322
323 if (in_initrd())
324 return free_and_strdup(&arg_default_unit, value);
325
326 } else if (streq(key, "systemd.dump_core") && value) {
327
328 r = parse_boolean(value);
329 if (r < 0)
330 log_warning("Failed to parse dump core switch %s. Ignoring.", value);
331 else
332 arg_dump_core = r;
333
334 } else if (streq(key, "systemd.crash_chvt") && value) {
335
336 if (parse_crash_chvt(value) < 0)
337 log_warning("Failed to parse crash chvt switch %s. Ignoring.", value);
338
339 } else if (streq(key, "systemd.crash_shell") && value) {
340
341 r = parse_boolean(value);
342 if (r < 0)
343 log_warning("Failed to parse crash shell switch %s. Ignoring.", value);
344 else
345 arg_crash_shell = r;
346
347 } else if (streq(key, "systemd.crash_reboot") && value) {
348
349 r = parse_boolean(value);
350 if (r < 0)
351 log_warning("Failed to parse crash reboot switch %s. Ignoring.", value);
352 else
353 arg_crash_reboot = r;
354
355 } else if (streq(key, "systemd.confirm_spawn") && value) {
356
357 r = parse_boolean(value);
358 if (r < 0)
359 log_warning("Failed to parse confirm spawn switch %s. Ignoring.", value);
360 else
361 arg_confirm_spawn = r;
362
363 } else if (streq(key, "systemd.show_status") && value) {
364
365 r = parse_show_status(value, &arg_show_status);
366 if (r < 0)
367 log_warning("Failed to parse show status switch %s. Ignoring.", value);
368
369 } else if (streq(key, "systemd.default_standard_output") && value) {
370
371 r = exec_output_from_string(value);
372 if (r < 0)
373 log_warning("Failed to parse default standard output switch %s. Ignoring.", value);
374 else
375 arg_default_std_output = r;
376
377 } else if (streq(key, "systemd.default_standard_error") && value) {
378
379 r = exec_output_from_string(value);
380 if (r < 0)
381 log_warning("Failed to parse default standard error switch %s. Ignoring.", value);
382 else
383 arg_default_std_error = r;
384
385 } else if (streq(key, "systemd.setenv") && value) {
386
387 if (env_assignment_is_valid(value)) {
388 char **env;
389
390 env = strv_env_set(arg_default_environment, value);
391 if (env)
392 arg_default_environment = env;
393 else
394 log_warning_errno(ENOMEM, "Setting environment variable '%s' failed, ignoring: %m", value);
395 } else
396 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
397
398 } else if (streq(key, "systemd.machine_id") && value) {
399
400 r = set_machine_id(value);
401 if (r < 0)
402 log_warning("MachineID '%s' is not valid. Ignoring.", value);
403
404 } else if (streq(key, "quiet") && !value) {
405
406 if (arg_show_status == _SHOW_STATUS_UNSET)
407 arg_show_status = SHOW_STATUS_AUTO;
408
409 } else if (streq(key, "debug") && !value) {
410
411 /* Note that log_parse_environment() handles 'debug'
412 * too, and sets the log level to LOG_DEBUG. */
413
414 if (detect_container() > 0)
415 log_set_target(LOG_TARGET_CONSOLE);
416
417 } else if (!value) {
418 const char *target;
419
420 /* SysV compatibility */
421 target = runlevel_to_target(key);
422 if (target)
423 return free_and_strdup(&arg_default_unit, target);
424
425 } else if (streq(key, "systemd.default_timeout_start_sec") && value) {
426
427 r = parse_sec(value, &arg_default_timeout_start_usec);
428 if (r < 0)
429 log_warning_errno(r, "Failed to parse default start timeout: %s, ignoring.", value);
430
431 if (arg_default_timeout_start_usec <= 0)
432 arg_default_timeout_start_usec = USEC_INFINITY;
433 }
434
435 return 0;
436 }
437
438 #define DEFINE_SETTER(name, func, descr) \
439 static int name(const char *unit, \
440 const char *filename, \
441 unsigned line, \
442 const char *section, \
443 unsigned section_line, \
444 const char *lvalue, \
445 int ltype, \
446 const char *rvalue, \
447 void *data, \
448 void *userdata) { \
449 \
450 int r; \
451 \
452 assert(filename); \
453 assert(lvalue); \
454 assert(rvalue); \
455 \
456 r = func(rvalue); \
457 if (r < 0) \
458 log_syntax(unit, LOG_ERR, filename, line, r, \
459 "Invalid " descr "'%s': %m", \
460 rvalue); \
461 \
462 return 0; \
463 }
464
465 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level")
466 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target")
467 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" )
468 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location")
469
470 static int config_parse_cpu_affinity2(
471 const char *unit,
472 const char *filename,
473 unsigned line,
474 const char *section,
475 unsigned section_line,
476 const char *lvalue,
477 int ltype,
478 const char *rvalue,
479 void *data,
480 void *userdata) {
481
482 _cleanup_cpu_free_ cpu_set_t *c = NULL;
483 int ncpus;
484
485 ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue);
486 if (ncpus < 0)
487 return ncpus;
488
489 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
490 log_warning("Failed to set CPU affinity: %m");
491
492 return 0;
493 }
494
495 static int config_parse_show_status(
496 const char* unit,
497 const char *filename,
498 unsigned line,
499 const char *section,
500 unsigned section_line,
501 const char *lvalue,
502 int ltype,
503 const char *rvalue,
504 void *data,
505 void *userdata) {
506
507 int k;
508 ShowStatus *b = data;
509
510 assert(filename);
511 assert(lvalue);
512 assert(rvalue);
513 assert(data);
514
515 k = parse_show_status(rvalue, b);
516 if (k < 0) {
517 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
518 return 0;
519 }
520
521 return 0;
522 }
523
524 static int config_parse_crash_chvt(
525 const char* unit,
526 const char *filename,
527 unsigned line,
528 const char *section,
529 unsigned section_line,
530 const char *lvalue,
531 int ltype,
532 const char *rvalue,
533 void *data,
534 void *userdata) {
535
536 int r;
537
538 assert(filename);
539 assert(lvalue);
540 assert(rvalue);
541
542 r = parse_crash_chvt(rvalue);
543 if (r < 0) {
544 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
545 return 0;
546 }
547
548 return 0;
549 }
550
551 static int config_parse_join_controllers(const char *unit,
552 const char *filename,
553 unsigned line,
554 const char *section,
555 unsigned section_line,
556 const char *lvalue,
557 int ltype,
558 const char *rvalue,
559 void *data,
560 void *userdata) {
561
562 const char *whole_rvalue = rvalue;
563 unsigned n = 0;
564
565 assert(filename);
566 assert(lvalue);
567 assert(rvalue);
568
569 arg_join_controllers = strv_free_free(arg_join_controllers);
570
571 for (;;) {
572 _cleanup_free_ char *word = NULL;
573 char **l;
574 int r;
575
576 r = extract_first_word(&rvalue, &word, WHITESPACE, EXTRACT_QUOTES);
577 if (r < 0) {
578 log_syntax(unit, LOG_ERR, filename, line, r, "Invalid value for %s: %s", lvalue, whole_rvalue);
579 return r;
580 }
581 if (r == 0)
582 break;
583
584 l = strv_split(word, ",");
585 if (!l)
586 return log_oom();
587 strv_uniq(l);
588
589 if (strv_length(l) <= 1) {
590 strv_free(l);
591 continue;
592 }
593
594 if (!arg_join_controllers) {
595 arg_join_controllers = new(char**, 2);
596 if (!arg_join_controllers) {
597 strv_free(l);
598 return log_oom();
599 }
600
601 arg_join_controllers[0] = l;
602 arg_join_controllers[1] = NULL;
603
604 n = 1;
605 } else {
606 char ***a;
607 char ***t;
608
609 t = new0(char**, n+2);
610 if (!t) {
611 strv_free(l);
612 return log_oom();
613 }
614
615 n = 0;
616
617 for (a = arg_join_controllers; *a; a++) {
618
619 if (strv_overlap(*a, l)) {
620 if (strv_extend_strv(&l, *a, false) < 0) {
621 strv_free(l);
622 strv_free_free(t);
623 return log_oom();
624 }
625
626 } else {
627 char **c;
628
629 c = strv_copy(*a);
630 if (!c) {
631 strv_free(l);
632 strv_free_free(t);
633 return log_oom();
634 }
635
636 t[n++] = c;
637 }
638 }
639
640 t[n++] = strv_uniq(l);
641
642 strv_free_free(arg_join_controllers);
643 arg_join_controllers = t;
644 }
645 }
646 if (!isempty(rvalue))
647 log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring.");
648
649 return 0;
650 }
651
652 static DEFINE_CONFIG_PARSE_ENUM(config_parse_cad_burst_action, cad_burst_action, CADBurstAction, "Failed to parse service restart specifier");
653
654 static int parse_config_file(void) {
655
656 const ConfigTableItem items[] = {
657 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
658 { "Manager", "LogTarget", config_parse_target, 0, NULL },
659 { "Manager", "LogColor", config_parse_color, 0, NULL },
660 { "Manager", "LogLocation", config_parse_location, 0, NULL },
661 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
662 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL },
663 { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL },
664 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
665 { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
666 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
667 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
668 { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
669 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
670 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
671 { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
672 #ifdef HAVE_SECCOMP
673 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
674 #endif
675 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
676 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
677 { "Manager", "DefaultStandardOutput", config_parse_output, 0, &arg_default_std_output },
678 { "Manager", "DefaultStandardError", config_parse_output, 0, &arg_default_std_error },
679 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
680 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
681 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
682 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */
683 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec, 0, &arg_default_start_limit_interval },
684 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
685 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
686 { "Manager", "DefaultLimitCPU", config_parse_limit, RLIMIT_CPU, arg_default_rlimit },
687 { "Manager", "DefaultLimitFSIZE", config_parse_limit, RLIMIT_FSIZE, arg_default_rlimit },
688 { "Manager", "DefaultLimitDATA", config_parse_limit, RLIMIT_DATA, arg_default_rlimit },
689 { "Manager", "DefaultLimitSTACK", config_parse_limit, RLIMIT_STACK, arg_default_rlimit },
690 { "Manager", "DefaultLimitCORE", config_parse_limit, RLIMIT_CORE, arg_default_rlimit },
691 { "Manager", "DefaultLimitRSS", config_parse_limit, RLIMIT_RSS, arg_default_rlimit },
692 { "Manager", "DefaultLimitNOFILE", config_parse_limit, RLIMIT_NOFILE, arg_default_rlimit },
693 { "Manager", "DefaultLimitAS", config_parse_limit, RLIMIT_AS, arg_default_rlimit },
694 { "Manager", "DefaultLimitNPROC", config_parse_limit, RLIMIT_NPROC, arg_default_rlimit },
695 { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, RLIMIT_MEMLOCK, arg_default_rlimit },
696 { "Manager", "DefaultLimitLOCKS", config_parse_limit, RLIMIT_LOCKS, arg_default_rlimit },
697 { "Manager", "DefaultLimitSIGPENDING", config_parse_limit, RLIMIT_SIGPENDING, arg_default_rlimit },
698 { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, RLIMIT_MSGQUEUE, arg_default_rlimit },
699 { "Manager", "DefaultLimitNICE", config_parse_limit, RLIMIT_NICE, arg_default_rlimit },
700 { "Manager", "DefaultLimitRTPRIO", config_parse_limit, RLIMIT_RTPRIO, arg_default_rlimit },
701 { "Manager", "DefaultLimitRTTIME", config_parse_limit, RLIMIT_RTTIME, arg_default_rlimit },
702 { "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_default_cpu_accounting },
703 { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting },
704 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
705 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
706 { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
707 { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max },
708 { "Manager", "CtrlAltDelBurstAction", config_parse_cad_burst_action, 0, &arg_cad_burst_action},
709 {}
710 };
711
712 const char *fn, *conf_dirs_nulstr;
713
714 fn = arg_system ?
715 PKGSYSCONFDIR "/system.conf" :
716 PKGSYSCONFDIR "/user.conf";
717
718 conf_dirs_nulstr = arg_system ?
719 CONF_PATHS_NULSTR("systemd/system.conf.d") :
720 CONF_PATHS_NULSTR("systemd/user.conf.d");
721
722 config_parse_many_nulstr(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, false, NULL);
723
724 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
725 * like everywhere else. */
726 if (arg_default_timeout_start_usec <= 0)
727 arg_default_timeout_start_usec = USEC_INFINITY;
728 if (arg_default_timeout_stop_usec <= 0)
729 arg_default_timeout_stop_usec = USEC_INFINITY;
730
731 return 0;
732 }
733
734 static void manager_set_defaults(Manager *m) {
735
736 assert(m);
737
738 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
739 m->default_std_output = arg_default_std_output;
740 m->default_std_error = arg_default_std_error;
741 m->default_timeout_start_usec = arg_default_timeout_start_usec;
742 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
743 m->default_restart_usec = arg_default_restart_usec;
744 m->default_start_limit_interval = arg_default_start_limit_interval;
745 m->default_start_limit_burst = arg_default_start_limit_burst;
746 m->default_cpu_accounting = arg_default_cpu_accounting;
747 m->default_io_accounting = arg_default_io_accounting;
748 m->default_blockio_accounting = arg_default_blockio_accounting;
749 m->default_memory_accounting = arg_default_memory_accounting;
750 m->default_tasks_accounting = arg_default_tasks_accounting;
751 m->default_tasks_max = arg_default_tasks_max;
752
753 manager_set_default_rlimits(m, arg_default_rlimit);
754 manager_environment_add(m, NULL, arg_default_environment);
755 }
756
757 static int parse_argv(int argc, char *argv[]) {
758
759 enum {
760 ARG_LOG_LEVEL = 0x100,
761 ARG_LOG_TARGET,
762 ARG_LOG_COLOR,
763 ARG_LOG_LOCATION,
764 ARG_UNIT,
765 ARG_SYSTEM,
766 ARG_USER,
767 ARG_TEST,
768 ARG_NO_PAGER,
769 ARG_VERSION,
770 ARG_DUMP_CONFIGURATION_ITEMS,
771 ARG_DUMP_CORE,
772 ARG_CRASH_CHVT,
773 ARG_CRASH_SHELL,
774 ARG_CRASH_REBOOT,
775 ARG_CONFIRM_SPAWN,
776 ARG_SHOW_STATUS,
777 ARG_DESERIALIZE,
778 ARG_SWITCHED_ROOT,
779 ARG_DEFAULT_STD_OUTPUT,
780 ARG_DEFAULT_STD_ERROR,
781 ARG_MACHINE_ID
782 };
783
784 static const struct option options[] = {
785 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
786 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
787 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
788 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
789 { "unit", required_argument, NULL, ARG_UNIT },
790 { "system", no_argument, NULL, ARG_SYSTEM },
791 { "user", no_argument, NULL, ARG_USER },
792 { "test", no_argument, NULL, ARG_TEST },
793 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
794 { "help", no_argument, NULL, 'h' },
795 { "version", no_argument, NULL, ARG_VERSION },
796 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
797 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
798 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
799 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
800 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
801 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
802 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
803 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
804 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
805 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
806 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
807 { "machine-id", required_argument, NULL, ARG_MACHINE_ID },
808 {}
809 };
810
811 int c, r;
812
813 assert(argc >= 1);
814 assert(argv);
815
816 if (getpid() == 1)
817 opterr = 0;
818
819 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
820
821 switch (c) {
822
823 case ARG_LOG_LEVEL:
824 r = log_set_max_level_from_string(optarg);
825 if (r < 0) {
826 log_error("Failed to parse log level %s.", optarg);
827 return r;
828 }
829
830 break;
831
832 case ARG_LOG_TARGET:
833 r = log_set_target_from_string(optarg);
834 if (r < 0) {
835 log_error("Failed to parse log target %s.", optarg);
836 return r;
837 }
838
839 break;
840
841 case ARG_LOG_COLOR:
842
843 if (optarg) {
844 r = log_show_color_from_string(optarg);
845 if (r < 0) {
846 log_error("Failed to parse log color setting %s.", optarg);
847 return r;
848 }
849 } else
850 log_show_color(true);
851
852 break;
853
854 case ARG_LOG_LOCATION:
855 if (optarg) {
856 r = log_show_location_from_string(optarg);
857 if (r < 0) {
858 log_error("Failed to parse log location setting %s.", optarg);
859 return r;
860 }
861 } else
862 log_show_location(true);
863
864 break;
865
866 case ARG_DEFAULT_STD_OUTPUT:
867 r = exec_output_from_string(optarg);
868 if (r < 0) {
869 log_error("Failed to parse default standard output setting %s.", optarg);
870 return r;
871 } else
872 arg_default_std_output = r;
873 break;
874
875 case ARG_DEFAULT_STD_ERROR:
876 r = exec_output_from_string(optarg);
877 if (r < 0) {
878 log_error("Failed to parse default standard error output setting %s.", optarg);
879 return r;
880 } else
881 arg_default_std_error = r;
882 break;
883
884 case ARG_UNIT:
885
886 r = free_and_strdup(&arg_default_unit, optarg);
887 if (r < 0)
888 return log_error_errno(r, "Failed to set default unit %s: %m", optarg);
889
890 break;
891
892 case ARG_SYSTEM:
893 arg_system = true;
894 break;
895
896 case ARG_USER:
897 arg_system = false;
898 break;
899
900 case ARG_TEST:
901 arg_action = ACTION_TEST;
902 break;
903
904 case ARG_NO_PAGER:
905 arg_no_pager = true;
906 break;
907
908 case ARG_VERSION:
909 arg_action = ACTION_VERSION;
910 break;
911
912 case ARG_DUMP_CONFIGURATION_ITEMS:
913 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
914 break;
915
916 case ARG_DUMP_CORE:
917 if (!optarg)
918 arg_dump_core = true;
919 else {
920 r = parse_boolean(optarg);
921 if (r < 0)
922 return log_error_errno(r, "Failed to parse dump core boolean: %s", optarg);
923 arg_dump_core = r;
924 }
925 break;
926
927 case ARG_CRASH_CHVT:
928 r = parse_crash_chvt(optarg);
929 if (r < 0)
930 return log_error_errno(r, "Failed to parse crash virtual terminal index: %s", optarg);
931 break;
932
933 case ARG_CRASH_SHELL:
934 if (!optarg)
935 arg_crash_shell = true;
936 else {
937 r = parse_boolean(optarg);
938 if (r < 0)
939 return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg);
940 arg_crash_shell = r;
941 }
942 break;
943
944 case ARG_CRASH_REBOOT:
945 if (!optarg)
946 arg_crash_reboot = true;
947 else {
948 r = parse_boolean(optarg);
949 if (r < 0)
950 return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg);
951 arg_crash_reboot = r;
952 }
953 break;
954
955 case ARG_CONFIRM_SPAWN:
956 r = optarg ? parse_boolean(optarg) : 1;
957 if (r < 0) {
958 log_error("Failed to parse confirm spawn boolean %s.", optarg);
959 return r;
960 }
961 arg_confirm_spawn = r;
962 break;
963
964 case ARG_SHOW_STATUS:
965 if (optarg) {
966 r = parse_show_status(optarg, &arg_show_status);
967 if (r < 0) {
968 log_error("Failed to parse show status boolean %s.", optarg);
969 return r;
970 }
971 } else
972 arg_show_status = SHOW_STATUS_YES;
973 break;
974
975 case ARG_DESERIALIZE: {
976 int fd;
977 FILE *f;
978
979 r = safe_atoi(optarg, &fd);
980 if (r < 0 || fd < 0) {
981 log_error("Failed to parse deserialize option %s.", optarg);
982 return -EINVAL;
983 }
984
985 (void) fd_cloexec(fd, true);
986
987 f = fdopen(fd, "r");
988 if (!f)
989 return log_error_errno(errno, "Failed to open serialization fd: %m");
990
991 safe_fclose(arg_serialization);
992 arg_serialization = f;
993
994 break;
995 }
996
997 case ARG_SWITCHED_ROOT:
998 arg_switched_root = true;
999 break;
1000
1001 case ARG_MACHINE_ID:
1002 r = set_machine_id(optarg);
1003 if (r < 0)
1004 return log_error_errno(r, "MachineID '%s' is not valid.", optarg);
1005 break;
1006
1007 case 'h':
1008 arg_action = ACTION_HELP;
1009 break;
1010
1011 case 'D':
1012 log_set_max_level(LOG_DEBUG);
1013 break;
1014
1015 case 'b':
1016 case 's':
1017 case 'z':
1018 /* Just to eat away the sysvinit kernel
1019 * cmdline args without getopt() error
1020 * messages that we'll parse in
1021 * parse_proc_cmdline_word() or ignore. */
1022
1023 case '?':
1024 if (getpid() != 1)
1025 return -EINVAL;
1026 else
1027 return 0;
1028
1029 default:
1030 assert_not_reached("Unhandled option code.");
1031 }
1032
1033 if (optind < argc && getpid() != 1) {
1034 /* Hmm, when we aren't run as init system
1035 * let's complain about excess arguments */
1036
1037 log_error("Excess arguments.");
1038 return -EINVAL;
1039 }
1040
1041 return 0;
1042 }
1043
1044 static int help(void) {
1045
1046 printf("%s [OPTIONS...]\n\n"
1047 "Starts up and maintains the system or user services.\n\n"
1048 " -h --help Show this help\n"
1049 " --test Determine startup sequence, dump it and exit\n"
1050 " --no-pager Do not pipe output into a pager\n"
1051 " --dump-configuration-items Dump understood unit configuration items\n"
1052 " --unit=UNIT Set default unit\n"
1053 " --system Run a system instance, even if PID != 1\n"
1054 " --user Run a user instance\n"
1055 " --dump-core[=BOOL] Dump core on crash\n"
1056 " --crash-vt=NR Change to specified VT on crash\n"
1057 " --crash-reboot[=BOOL] Reboot on crash\n"
1058 " --crash-shell[=BOOL] Run shell on crash\n"
1059 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1060 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1061 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1062 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1063 " --log-color[=BOOL] Highlight important log messages\n"
1064 " --log-location[=BOOL] Include code location in log messages\n"
1065 " --default-standard-output= Set default standard output for services\n"
1066 " --default-standard-error= Set default standard error output for services\n",
1067 program_invocation_short_name);
1068
1069 return 0;
1070 }
1071
1072 static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
1073 _cleanup_fdset_free_ FDSet *fds = NULL;
1074 _cleanup_fclose_ FILE *f = NULL;
1075 int r;
1076
1077 assert(m);
1078 assert(_f);
1079 assert(_fds);
1080
1081 r = manager_open_serialization(m, &f);
1082 if (r < 0)
1083 return log_error_errno(r, "Failed to create serialization file: %m");
1084
1085 /* Make sure nothing is really destructed when we shut down */
1086 m->n_reloading++;
1087 bus_manager_send_reloading(m, true);
1088
1089 fds = fdset_new();
1090 if (!fds)
1091 return log_oom();
1092
1093 r = manager_serialize(m, f, fds, switching_root);
1094 if (r < 0)
1095 return log_error_errno(r, "Failed to serialize state: %m");
1096
1097 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1098 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1099
1100 r = fd_cloexec(fileno(f), false);
1101 if (r < 0)
1102 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1103
1104 r = fdset_cloexec(fds, false);
1105 if (r < 0)
1106 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1107
1108 *_f = f;
1109 *_fds = fds;
1110
1111 f = NULL;
1112 fds = NULL;
1113
1114 return 0;
1115 }
1116
1117 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1118 struct rlimit nl;
1119 int r;
1120
1121 assert(saved_rlimit);
1122
1123 /* Save the original RLIMIT_NOFILE so that we can reset it
1124 * later when transitioning from the initrd to the main
1125 * systemd or suchlike. */
1126 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1127 return log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1128
1129 /* Make sure forked processes get the default kernel setting */
1130 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1131 struct rlimit *rl;
1132
1133 rl = newdup(struct rlimit, saved_rlimit, 1);
1134 if (!rl)
1135 return log_oom();
1136
1137 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1138 }
1139
1140 /* Bump up the resource limit for ourselves substantially */
1141 nl.rlim_cur = nl.rlim_max = 64*1024;
1142 r = setrlimit_closest(RLIMIT_NOFILE, &nl);
1143 if (r < 0)
1144 return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1145
1146 return 0;
1147 }
1148
1149 static void test_usr(void) {
1150
1151 /* Check that /usr is not a separate fs */
1152
1153 if (dir_is_empty("/usr") <= 0)
1154 return;
1155
1156 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1157 "Some things will probably break (sometimes even silently) in mysterious ways. "
1158 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1159 }
1160
1161 static int initialize_join_controllers(void) {
1162 /* By default, mount "cpu" + "cpuacct" together, and "net_cls"
1163 * + "net_prio". We'd like to add "cpuset" to the mix, but
1164 * "cpuset" doesn't really work for groups with no initialized
1165 * attributes. */
1166
1167 arg_join_controllers = new(char**, 3);
1168 if (!arg_join_controllers)
1169 return -ENOMEM;
1170
1171 arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL);
1172 if (!arg_join_controllers[0])
1173 goto oom;
1174
1175 arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL);
1176 if (!arg_join_controllers[1])
1177 goto oom;
1178
1179 arg_join_controllers[2] = NULL;
1180 return 0;
1181
1182 oom:
1183 arg_join_controllers = strv_free_free(arg_join_controllers);
1184 return -ENOMEM;
1185 }
1186
1187 static int enforce_syscall_archs(Set *archs) {
1188 #ifdef HAVE_SECCOMP
1189 scmp_filter_ctx *seccomp;
1190 Iterator i;
1191 void *id;
1192 int r;
1193
1194 if (!is_seccomp_available())
1195 return 0;
1196
1197 seccomp = seccomp_init(SCMP_ACT_ALLOW);
1198 if (!seccomp)
1199 return log_oom();
1200
1201 SET_FOREACH(id, arg_syscall_archs, i) {
1202 r = seccomp_arch_add(seccomp, PTR_TO_UINT32(id) - 1);
1203 if (r == -EEXIST)
1204 continue;
1205 if (r < 0) {
1206 log_error_errno(r, "Failed to add architecture to seccomp: %m");
1207 goto finish;
1208 }
1209 }
1210
1211 r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
1212 if (r < 0) {
1213 log_error_errno(r, "Failed to unset NO_NEW_PRIVS: %m");
1214 goto finish;
1215 }
1216
1217 r = seccomp_load(seccomp);
1218 if (r < 0)
1219 log_error_errno(r, "Failed to add install architecture seccomp: %m");
1220
1221 finish:
1222 seccomp_release(seccomp);
1223 return r;
1224 #else
1225 return 0;
1226 #endif
1227 }
1228
1229 static int status_welcome(void) {
1230 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1231 int r;
1232
1233 r = parse_env_file("/etc/os-release", NEWLINE,
1234 "PRETTY_NAME", &pretty_name,
1235 "ANSI_COLOR", &ansi_color,
1236 NULL);
1237 if (r == -ENOENT)
1238 r = parse_env_file("/usr/lib/os-release", NEWLINE,
1239 "PRETTY_NAME", &pretty_name,
1240 "ANSI_COLOR", &ansi_color,
1241 NULL);
1242
1243 if (r < 0 && r != -ENOENT)
1244 log_warning_errno(r, "Failed to read os-release file: %m");
1245
1246 if (log_get_show_color())
1247 return status_printf(NULL, false, false,
1248 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1249 isempty(ansi_color) ? "1" : ansi_color,
1250 isempty(pretty_name) ? "Linux" : pretty_name);
1251 else
1252 return status_printf(NULL, false, false,
1253 "\nWelcome to %s!\n",
1254 isempty(pretty_name) ? "Linux" : pretty_name);
1255 }
1256
1257 static int write_container_id(void) {
1258 const char *c;
1259 int r;
1260
1261 c = getenv("container");
1262 if (isempty(c))
1263 return 0;
1264
1265 RUN_WITH_UMASK(0022)
1266 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1267 if (r < 0)
1268 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
1269
1270 return 1;
1271 }
1272
1273 static int bump_unix_max_dgram_qlen(void) {
1274 _cleanup_free_ char *qlen = NULL;
1275 unsigned long v;
1276 int r;
1277
1278 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel
1279 * default of 16 is simply too low. We set the value really
1280 * really early during boot, so that it is actually applied to
1281 * all our sockets, including the $NOTIFY_SOCKET one. */
1282
1283 r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1284 if (r < 0)
1285 return log_warning_errno(r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1286
1287 r = safe_atolu(qlen, &v);
1288 if (r < 0)
1289 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length, ignoring: %m");
1290
1291 if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1292 return 0;
1293
1294 qlen = mfree(qlen);
1295 if (asprintf(&qlen, "%lu\n", DEFAULT_UNIX_MAX_DGRAM_QLEN) < 0)
1296 return log_oom();
1297
1298 r = write_string_file("/proc/sys/net/unix/max_dgram_qlen", qlen, 0);
1299 if (r < 0)
1300 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1301 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1302
1303 return 1;
1304 }
1305
1306 static int fixup_environment(void) {
1307 _cleanup_free_ char *term = NULL;
1308 int r;
1309
1310 /* We expect the environment to be set correctly
1311 * if run inside a container. */
1312 if (detect_container() > 0)
1313 return 0;
1314
1315 /* When started as PID1, the kernel uses /dev/console
1316 * for our stdios and uses TERM=linux whatever the
1317 * backend device used by the console. We try to make
1318 * a better guess here since some consoles might not
1319 * have support for color mode for example.
1320 *
1321 * However if TERM was configured through the kernel
1322 * command line then leave it alone. */
1323
1324 r = get_proc_cmdline_key("TERM=", &term);
1325 if (r < 0)
1326 return r;
1327
1328 if (r == 0) {
1329 term = strdup(default_term_for_tty("/dev/console"));
1330 if (!term)
1331 return -ENOMEM;
1332 }
1333
1334 if (setenv("TERM", term, 1) < 0)
1335 return -errno;
1336
1337 return 0;
1338 }
1339
1340 int main(int argc, char *argv[]) {
1341 Manager *m = NULL;
1342 int r, retval = EXIT_FAILURE;
1343 usec_t before_startup, after_startup;
1344 char timespan[FORMAT_TIMESPAN_MAX];
1345 FDSet *fds = NULL;
1346 bool reexecute = false;
1347 const char *shutdown_verb = NULL;
1348 dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL;
1349 dual_timestamp userspace_timestamp = DUAL_TIMESTAMP_NULL;
1350 dual_timestamp kernel_timestamp = DUAL_TIMESTAMP_NULL;
1351 dual_timestamp security_start_timestamp = DUAL_TIMESTAMP_NULL;
1352 dual_timestamp security_finish_timestamp = DUAL_TIMESTAMP_NULL;
1353 static char systemd[] = "systemd";
1354 bool skip_setup = false;
1355 unsigned j;
1356 bool loaded_policy = false;
1357 bool arm_reboot_watchdog = false;
1358 bool queue_default_job = false;
1359 bool empty_etc = false;
1360 char *switch_root_dir = NULL, *switch_root_init = NULL;
1361 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0);
1362 const char *error_message = NULL;
1363
1364 #ifdef HAVE_SYSV_COMPAT
1365 if (getpid() != 1 && strstr(program_invocation_short_name, "init")) {
1366 /* This is compatibility support for SysV, where
1367 * calling init as a user is identical to telinit. */
1368
1369 execv(SYSTEMCTL_BINARY_PATH, argv);
1370 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1371 return 1;
1372 }
1373 #endif
1374
1375 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
1376 dual_timestamp_get(&userspace_timestamp);
1377
1378 /* Determine if this is a reexecution or normal bootup. We do
1379 * the full command line parsing much later, so let's just
1380 * have a quick peek here. */
1381 if (strv_find(argv+1, "--deserialize"))
1382 skip_setup = true;
1383
1384 /* If we have switched root, do all the special setup
1385 * things */
1386 if (strv_find(argv+1, "--switched-root"))
1387 skip_setup = false;
1388
1389 /* If we get started via the /sbin/init symlink then we are
1390 called 'init'. After a subsequent reexecution we are then
1391 called 'systemd'. That is confusing, hence let's call us
1392 systemd right-away. */
1393 program_invocation_short_name = systemd;
1394 prctl(PR_SET_NAME, systemd);
1395
1396 saved_argv = argv;
1397 saved_argc = argc;
1398
1399 log_set_upgrade_syslog_to_journal(true);
1400
1401 /* Disable the umask logic */
1402 if (getpid() == 1)
1403 umask(0);
1404
1405 if (getpid() == 1 && detect_container() <= 0) {
1406
1407 /* Running outside of a container as PID 1 */
1408 arg_system = true;
1409 log_set_target(LOG_TARGET_KMSG);
1410 log_open();
1411
1412 if (in_initrd())
1413 initrd_timestamp = userspace_timestamp;
1414
1415 if (!skip_setup) {
1416 r = mount_setup_early();
1417 if (r < 0) {
1418 error_message = "Failed to early mount API filesystems";
1419 goto finish;
1420 }
1421 dual_timestamp_get(&security_start_timestamp);
1422 if (mac_selinux_setup(&loaded_policy) < 0) {
1423 error_message = "Failed to load SELinux policy";
1424 goto finish;
1425 } else if (mac_smack_setup(&loaded_policy) < 0) {
1426 error_message = "Failed to load SMACK policy";
1427 goto finish;
1428 } else if (ima_setup() < 0) {
1429 error_message = "Failed to load IMA policy";
1430 goto finish;
1431 }
1432 dual_timestamp_get(&security_finish_timestamp);
1433 }
1434
1435 if (mac_selinux_init() < 0) {
1436 error_message = "Failed to initialize SELinux policy";
1437 goto finish;
1438 }
1439
1440 if (!skip_setup) {
1441 if (clock_is_localtime(NULL) > 0) {
1442 int min;
1443
1444 /*
1445 * The very first call of settimeofday() also does a time warp in the kernel.
1446 *
1447 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on
1448 * external tools to take care of maintaining the RTC and do all adjustments.
1449 * This matches the behavior of Windows, which leaves the RTC alone if the
1450 * registry tells that the RTC runs in UTC.
1451 */
1452 r = clock_set_timezone(&min);
1453 if (r < 0)
1454 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1455 else
1456 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1457 } else if (!in_initrd()) {
1458 /*
1459 * Do a dummy very first call to seal the kernel's time warp magic.
1460 *
1461 * Do not call this from inside the initrd. The initrd might not
1462 * carry /etc/adjtime with LOCAL, but the real system could be set up
1463 * that way. In such case, we need to delay the time-warp or the sealing
1464 * until we reach the real system.
1465 *
1466 * Do no set the kernel's timezone. The concept of local time cannot
1467 * be supported reliably, the time will jump or be incorrect at every daylight
1468 * saving time change. All kernel local time concepts will be treated
1469 * as UTC that way.
1470 */
1471 (void) clock_reset_timewarp();
1472 }
1473
1474 r = clock_apply_epoch();
1475 if (r < 0)
1476 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1477 else if (r > 0)
1478 log_info("System time before build time, advancing clock.");
1479 }
1480
1481 /* Set the default for later on, but don't actually
1482 * open the logs like this for now. Note that if we
1483 * are transitioning from the initrd there might still
1484 * be journal fd open, and we shouldn't attempt
1485 * opening that before we parsed /proc/cmdline which
1486 * might redirect output elsewhere. */
1487 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
1488
1489 } else if (getpid() == 1) {
1490 /* Running inside a container, as PID 1 */
1491 arg_system = true;
1492 log_set_target(LOG_TARGET_CONSOLE);
1493 log_close_console(); /* force reopen of /dev/console */
1494 log_open();
1495
1496 /* For the later on, see above... */
1497 log_set_target(LOG_TARGET_JOURNAL);
1498
1499 /* clear the kernel timestamp,
1500 * because we are in a container */
1501 kernel_timestamp = DUAL_TIMESTAMP_NULL;
1502 } else {
1503 /* Running as user instance */
1504 arg_system = false;
1505 log_set_target(LOG_TARGET_AUTO);
1506 log_open();
1507
1508 /* clear the kernel timestamp,
1509 * because we are not PID 1 */
1510 kernel_timestamp = DUAL_TIMESTAMP_NULL;
1511 }
1512
1513 if (getpid() == 1) {
1514 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1515 * will process core dumps for system services by default. */
1516 if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0)
1517 log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m");
1518
1519 /* But at the same time, turn off the core_pattern logic by default, so that no coredumps are stored
1520 * until the systemd-coredump tool is enabled via sysctl. */
1521 if (!skip_setup)
1522 (void) write_string_file("/proc/sys/kernel/core_pattern", "|/bin/false", 0);
1523 }
1524
1525 if (arg_system) {
1526 if (fixup_environment() < 0) {
1527 error_message = "Failed to fix up PID1 environment";
1528 goto finish;
1529 }
1530
1531 /* Try to figure out if we can use colors with the console. No
1532 * need to do that for user instances since they never log
1533 * into the console. */
1534 log_show_color(colors_enabled());
1535 r = make_null_stdio();
1536 if (r < 0)
1537 log_warning_errno(r, "Failed to redirect standard streams to /dev/null: %m");
1538 }
1539
1540 r = initialize_join_controllers();
1541 if (r < 0) {
1542 error_message = "Failed to initialize cgroup controllers";
1543 goto finish;
1544 }
1545
1546 /* Mount /proc, /sys and friends, so that /proc/cmdline and
1547 * /proc/$PID/fd is available. */
1548 if (getpid() == 1) {
1549
1550 /* Load the kernel modules early, so that we kdbus.ko is loaded before kdbusfs shall be mounted */
1551 if (!skip_setup)
1552 kmod_setup();
1553
1554 r = mount_setup(loaded_policy);
1555 if (r < 0) {
1556 error_message = "Failed to mount API filesystems";
1557 goto finish;
1558 }
1559 }
1560
1561 /* Reset all signal handlers. */
1562 (void) reset_all_signal_handlers();
1563 (void) ignore_signals(SIGNALS_IGNORE, -1);
1564
1565 arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U);
1566
1567 if (parse_config_file() < 0) {
1568 error_message = "Failed to parse config file";
1569 goto finish;
1570 }
1571
1572 if (arg_system) {
1573 r = parse_proc_cmdline(parse_proc_cmdline_item, NULL, false);
1574 if (r < 0)
1575 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
1576 }
1577
1578 /* Note that this also parses bits from the kernel command
1579 * line, including "debug". */
1580 log_parse_environment();
1581
1582 if (parse_argv(argc, argv) < 0) {
1583 error_message = "Failed to parse commandline arguments";
1584 goto finish;
1585 }
1586
1587 /* Initialize default unit */
1588 if (!arg_default_unit) {
1589 arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET);
1590 if (!arg_default_unit) {
1591 r = log_oom();
1592 error_message = "Failed to set default unit";
1593 goto finish;
1594 }
1595 }
1596
1597 if (arg_action == ACTION_TEST &&
1598 geteuid() == 0) {
1599 log_error("Don't run test mode as root.");
1600 goto finish;
1601 }
1602
1603 if (!arg_system &&
1604 arg_action == ACTION_RUN &&
1605 sd_booted() <= 0) {
1606 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
1607 goto finish;
1608 }
1609
1610 if (arg_system &&
1611 arg_action == ACTION_RUN &&
1612 running_in_chroot() > 0) {
1613 log_error("Cannot be run in a chroot() environment.");
1614 goto finish;
1615 }
1616
1617 if (arg_action == ACTION_TEST || arg_action == ACTION_HELP) {
1618 pager_open(arg_no_pager, false);
1619 skip_setup = true;
1620 }
1621
1622 if (arg_action == ACTION_HELP) {
1623 retval = help();
1624 goto finish;
1625 } else if (arg_action == ACTION_VERSION) {
1626 retval = version();
1627 goto finish;
1628 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
1629 pager_open(arg_no_pager, false);
1630 unit_dump_config_items(stdout);
1631 retval = EXIT_SUCCESS;
1632 goto finish;
1633 }
1634
1635 if (!arg_system &&
1636 !getenv("XDG_RUNTIME_DIR")) {
1637 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
1638 goto finish;
1639 }
1640
1641 assert_se(arg_action == ACTION_RUN || arg_action == ACTION_TEST);
1642
1643 /* Close logging fds, in order not to confuse fdset below */
1644 log_close();
1645
1646 /* Remember open file descriptors for later deserialization */
1647 r = fdset_new_fill(&fds);
1648 if (r < 0) {
1649 log_emergency_errno(r, "Failed to allocate fd set: %m");
1650 error_message = "Failed to allocate fd set";
1651 goto finish;
1652 } else
1653 fdset_cloexec(fds, true);
1654
1655 if (arg_serialization)
1656 assert_se(fdset_remove(fds, fileno(arg_serialization)) >= 0);
1657
1658 if (arg_system)
1659 /* Become a session leader if we aren't one yet. */
1660 setsid();
1661
1662 /* Move out of the way, so that we won't block unmounts */
1663 assert_se(chdir("/") == 0);
1664
1665 /* Reset the console, but only if this is really init and we
1666 * are freshly booted */
1667 if (arg_system && arg_action == ACTION_RUN) {
1668
1669 /* If we are init, we connect stdin/stdout/stderr to
1670 * /dev/null and make sure we don't have a controlling
1671 * tty. */
1672 release_terminal();
1673
1674 if (getpid() == 1 && !skip_setup)
1675 console_setup();
1676 }
1677
1678 /* Open the logging devices, if possible and necessary */
1679 log_open();
1680
1681 if (arg_show_status == _SHOW_STATUS_UNSET)
1682 arg_show_status = SHOW_STATUS_YES;
1683
1684 /* Make sure we leave a core dump without panicing the
1685 * kernel. */
1686 if (getpid() == 1) {
1687 install_crash_handler();
1688
1689 r = mount_cgroup_controllers(arg_join_controllers);
1690 if (r < 0)
1691 goto finish;
1692 }
1693
1694 if (arg_system) {
1695 int v;
1696
1697 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1698 arg_action == ACTION_TEST ? "test " : "" );
1699
1700 v = detect_virtualization();
1701 if (v > 0)
1702 log_info("Detected virtualization %s.", virtualization_to_string(v));
1703
1704 write_container_id();
1705
1706 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1707
1708 if (in_initrd())
1709 log_info("Running in initial RAM disk.");
1710
1711 /* Let's check whether /etc is already populated. We
1712 * don't actually really check for that, but use
1713 * /etc/machine-id as flag file. This allows container
1714 * managers and installers to provision a couple of
1715 * files already. If the container manager wants to
1716 * provision the machine ID itself it should pass
1717 * $container_uuid to PID 1. */
1718
1719 empty_etc = access("/etc/machine-id", F_OK) < 0;
1720 if (empty_etc)
1721 log_info("Running with unpopulated /etc.");
1722 } else {
1723 _cleanup_free_ char *t;
1724
1725 t = uid_to_name(getuid());
1726 log_debug(PACKAGE_STRING " running in %suser mode for user "UID_FMT"/%s. (" SYSTEMD_FEATURES ")",
1727 arg_action == ACTION_TEST ? " test" : "", getuid(), t);
1728 }
1729
1730 if (arg_system && !skip_setup) {
1731 if (arg_show_status > 0)
1732 status_welcome();
1733
1734 hostname_setup();
1735 machine_id_setup(NULL, arg_machine_id, NULL);
1736 loopback_setup();
1737 bump_unix_max_dgram_qlen();
1738
1739 test_usr();
1740 }
1741
1742 if (arg_system && arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY)
1743 watchdog_set_timeout(&arg_runtime_watchdog);
1744
1745 if (arg_timer_slack_nsec != NSEC_INFINITY)
1746 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1747 log_error_errno(errno, "Failed to adjust timer slack: %m");
1748
1749 if (!cap_test_all(arg_capability_bounding_set)) {
1750 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
1751 if (r < 0) {
1752 log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
1753 error_message = "Failed to drop capability bounding set of usermode helpers";
1754 goto finish;
1755 }
1756 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
1757 if (r < 0) {
1758 log_emergency_errno(r, "Failed to drop capability bounding set: %m");
1759 error_message = "Failed to drop capability bounding set";
1760 goto finish;
1761 }
1762 }
1763
1764 if (arg_syscall_archs) {
1765 r = enforce_syscall_archs(arg_syscall_archs);
1766 if (r < 0) {
1767 error_message = "Failed to set syscall architectures";
1768 goto finish;
1769 }
1770 }
1771
1772 if (!arg_system)
1773 /* Become reaper of our children */
1774 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
1775 log_warning_errno(errno, "Failed to make us a subreaper: %m");
1776
1777 if (arg_system) {
1778 (void) bump_rlimit_nofile(&saved_rlimit_nofile);
1779
1780 if (empty_etc) {
1781 r = unit_file_preset_all(UNIT_FILE_SYSTEM, false, NULL, UNIT_FILE_PRESET_ENABLE_ONLY, false, NULL, 0);
1782 if (r < 0)
1783 log_full_errno(r == -EEXIST ? LOG_NOTICE : LOG_WARNING, r, "Failed to populate /etc with preset unit settings, ignoring: %m");
1784 else
1785 log_info("Populated /etc with preset unit settings.");
1786 }
1787 }
1788
1789 r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER, arg_action == ACTION_TEST, &m);
1790 if (r < 0) {
1791 log_emergency_errno(r, "Failed to allocate manager object: %m");
1792 error_message = "Failed to allocate manager object";
1793 goto finish;
1794 }
1795
1796 m->confirm_spawn = arg_confirm_spawn;
1797 m->runtime_watchdog = arg_runtime_watchdog;
1798 m->shutdown_watchdog = arg_shutdown_watchdog;
1799 m->userspace_timestamp = userspace_timestamp;
1800 m->kernel_timestamp = kernel_timestamp;
1801 m->initrd_timestamp = initrd_timestamp;
1802 m->security_start_timestamp = security_start_timestamp;
1803 m->security_finish_timestamp = security_finish_timestamp;
1804 m->cad_burst_action = arg_cad_burst_action;
1805
1806 manager_set_defaults(m);
1807 manager_set_show_status(m, arg_show_status);
1808 manager_set_first_boot(m, empty_etc);
1809
1810 /* Remember whether we should queue the default job */
1811 queue_default_job = !arg_serialization || arg_switched_root;
1812
1813 before_startup = now(CLOCK_MONOTONIC);
1814
1815 r = manager_startup(m, arg_serialization, fds);
1816 if (r < 0)
1817 log_error_errno(r, "Failed to fully start up daemon: %m");
1818
1819 /* This will close all file descriptors that were opened, but
1820 * not claimed by any unit. */
1821 fds = fdset_free(fds);
1822
1823 arg_serialization = safe_fclose(arg_serialization);
1824
1825 if (queue_default_job) {
1826 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
1827 Unit *target = NULL;
1828 Job *default_unit_job;
1829
1830 log_debug("Activating default unit: %s", arg_default_unit);
1831
1832 r = manager_load_unit(m, arg_default_unit, NULL, &error, &target);
1833 if (r < 0)
1834 log_error("Failed to load default target: %s", bus_error_message(&error, r));
1835 else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND)
1836 log_error_errno(target->load_error, "Failed to load default target: %m");
1837 else if (target->load_state == UNIT_MASKED)
1838 log_error("Default target masked.");
1839
1840 if (!target || target->load_state != UNIT_LOADED) {
1841 log_info("Trying to load rescue target...");
1842
1843 r = manager_load_unit(m, SPECIAL_RESCUE_TARGET, NULL, &error, &target);
1844 if (r < 0) {
1845 log_emergency("Failed to load rescue target: %s", bus_error_message(&error, r));
1846 error_message = "Failed to load rescue target";
1847 goto finish;
1848 } else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND) {
1849 log_emergency_errno(target->load_error, "Failed to load rescue target: %m");
1850 error_message = "Failed to load rescue target";
1851 goto finish;
1852 } else if (target->load_state == UNIT_MASKED) {
1853 log_emergency("Rescue target masked.");
1854 error_message = "Rescue target masked";
1855 goto finish;
1856 }
1857 }
1858
1859 assert(target->load_state == UNIT_LOADED);
1860
1861 if (arg_action == ACTION_TEST) {
1862 printf("-> By units:\n");
1863 manager_dump_units(m, stdout, "\t");
1864 }
1865
1866 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, &error, &default_unit_job);
1867 if (r == -EPERM) {
1868 log_debug("Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1869
1870 sd_bus_error_free(&error);
1871
1872 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, &error, &default_unit_job);
1873 if (r < 0) {
1874 log_emergency("Failed to start default target: %s", bus_error_message(&error, r));
1875 error_message = "Failed to start default target";
1876 goto finish;
1877 }
1878 } else if (r < 0) {
1879 log_emergency("Failed to isolate default target: %s", bus_error_message(&error, r));
1880 error_message = "Failed to isolate default target";
1881 goto finish;
1882 }
1883
1884 m->default_unit_job_id = default_unit_job->id;
1885
1886 after_startup = now(CLOCK_MONOTONIC);
1887 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
1888 "Loaded units and determined initial transaction in %s.",
1889 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
1890
1891 if (arg_action == ACTION_TEST) {
1892 printf("-> By jobs:\n");
1893 manager_dump_jobs(m, stdout, "\t");
1894 retval = EXIT_SUCCESS;
1895 goto finish;
1896 }
1897 }
1898
1899 for (;;) {
1900 r = manager_loop(m);
1901 if (r < 0) {
1902 log_emergency_errno(r, "Failed to run main loop: %m");
1903 error_message = "Failed to run main loop";
1904 goto finish;
1905 }
1906
1907 switch (m->exit_code) {
1908
1909 case MANAGER_RELOAD:
1910 log_info("Reloading.");
1911
1912 r = parse_config_file();
1913 if (r < 0)
1914 log_error("Failed to parse config file.");
1915
1916 manager_set_defaults(m);
1917
1918 r = manager_reload(m);
1919 if (r < 0)
1920 log_error_errno(r, "Failed to reload: %m");
1921 break;
1922
1923 case MANAGER_REEXECUTE:
1924
1925 if (prepare_reexecute(m, &arg_serialization, &fds, false) < 0) {
1926 error_message = "Failed to prepare for reexecution";
1927 goto finish;
1928 }
1929
1930 reexecute = true;
1931 log_notice("Reexecuting.");
1932 goto finish;
1933
1934 case MANAGER_SWITCH_ROOT:
1935 /* Steal the switch root parameters */
1936 switch_root_dir = m->switch_root;
1937 switch_root_init = m->switch_root_init;
1938 m->switch_root = m->switch_root_init = NULL;
1939
1940 if (!switch_root_init)
1941 if (prepare_reexecute(m, &arg_serialization, &fds, true) < 0) {
1942 error_message = "Failed to prepare for reexecution";
1943 goto finish;
1944 }
1945
1946 reexecute = true;
1947 log_notice("Switching root.");
1948 goto finish;
1949
1950 case MANAGER_EXIT:
1951 retval = m->return_value;
1952
1953 if (MANAGER_IS_USER(m)) {
1954 log_debug("Exit.");
1955 goto finish;
1956 }
1957
1958 /* fallthrough */
1959 case MANAGER_REBOOT:
1960 case MANAGER_POWEROFF:
1961 case MANAGER_HALT:
1962 case MANAGER_KEXEC: {
1963 static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
1964 [MANAGER_EXIT] = "exit",
1965 [MANAGER_REBOOT] = "reboot",
1966 [MANAGER_POWEROFF] = "poweroff",
1967 [MANAGER_HALT] = "halt",
1968 [MANAGER_KEXEC] = "kexec"
1969 };
1970
1971 assert_se(shutdown_verb = table[m->exit_code]);
1972 arm_reboot_watchdog = m->exit_code == MANAGER_REBOOT;
1973
1974 log_notice("Shutting down.");
1975 goto finish;
1976 }
1977
1978 default:
1979 assert_not_reached("Unknown exit code.");
1980 }
1981 }
1982
1983 finish:
1984 pager_close();
1985
1986 if (m)
1987 arg_shutdown_watchdog = m->shutdown_watchdog;
1988
1989 m = manager_free(m);
1990
1991 for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++)
1992 arg_default_rlimit[j] = mfree(arg_default_rlimit[j]);
1993
1994 arg_default_unit = mfree(arg_default_unit);
1995 arg_join_controllers = strv_free_free(arg_join_controllers);
1996 arg_default_environment = strv_free(arg_default_environment);
1997 arg_syscall_archs = set_free(arg_syscall_archs);
1998
1999 mac_selinux_finish();
2000
2001 if (reexecute) {
2002 const char **args;
2003 unsigned i, args_size;
2004
2005 /* Close and disarm the watchdog, so that the new
2006 * instance can reinitialize it, but doesn't get
2007 * rebooted while we do that */
2008 watchdog_close(true);
2009
2010 /* Reset the RLIMIT_NOFILE to the kernel default, so
2011 * that the new systemd can pass the kernel default to
2012 * its child processes */
2013 if (saved_rlimit_nofile.rlim_cur > 0)
2014 (void) setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile);
2015
2016 if (switch_root_dir) {
2017 /* Kill all remaining processes from the
2018 * initrd, but don't wait for them, so that we
2019 * can handle the SIGCHLD for them after
2020 * deserializing. */
2021 broadcast_signal(SIGTERM, false, true);
2022
2023 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
2024 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
2025 if (r < 0)
2026 log_error_errno(r, "Failed to switch root, trying to continue: %m");
2027 }
2028
2029 args_size = MAX(6, argc+1);
2030 args = newa(const char*, args_size);
2031
2032 if (!switch_root_init) {
2033 char sfd[DECIMAL_STR_MAX(int) + 1];
2034
2035 /* First try to spawn ourselves with the right
2036 * path, and with full serialization. We do
2037 * this only if the user didn't specify an
2038 * explicit init to spawn. */
2039
2040 assert(arg_serialization);
2041 assert(fds);
2042
2043 xsprintf(sfd, "%i", fileno(arg_serialization));
2044
2045 i = 0;
2046 args[i++] = SYSTEMD_BINARY_PATH;
2047 if (switch_root_dir)
2048 args[i++] = "--switched-root";
2049 args[i++] = arg_system ? "--system" : "--user";
2050 args[i++] = "--deserialize";
2051 args[i++] = sfd;
2052 args[i++] = NULL;
2053
2054 assert(i <= args_size);
2055
2056 /*
2057 * We want valgrind to print its memory usage summary before reexecution.
2058 * Valgrind won't do this is on its own on exec(), but it will do it on exit().
2059 * Hence, to ensure we get a summary here, fork() off a child, let it exit() cleanly,
2060 * so that it prints the summary, and wait() for it in the parent, before proceeding into the exec().
2061 */
2062 valgrind_summary_hack();
2063
2064 (void) execv(args[0], (char* const*) args);
2065 }
2066
2067 /* Try the fallback, if there is any, without any
2068 * serialization. We pass the original argv[] and
2069 * envp[]. (Well, modulo the ordering changes due to
2070 * getopt() in argv[], and some cleanups in envp[],
2071 * but let's hope that doesn't matter.) */
2072
2073 arg_serialization = safe_fclose(arg_serialization);
2074 fds = fdset_free(fds);
2075
2076 /* Reopen the console */
2077 (void) make_console_stdio();
2078
2079 for (j = 1, i = 1; j < (unsigned) argc; j++)
2080 args[i++] = argv[j];
2081 args[i++] = NULL;
2082 assert(i <= args_size);
2083
2084 /* Reenable any blocked signals, especially important
2085 * if we switch from initial ramdisk to init=... */
2086 (void) reset_all_signal_handlers();
2087 (void) reset_signal_mask();
2088
2089 if (switch_root_init) {
2090 args[0] = switch_root_init;
2091 (void) execv(args[0], (char* const*) args);
2092 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
2093 }
2094
2095 args[0] = "/sbin/init";
2096 (void) execv(args[0], (char* const*) args);
2097
2098 if (errno == ENOENT) {
2099 log_warning("No /sbin/init, trying fallback");
2100
2101 args[0] = "/bin/sh";
2102 args[1] = NULL;
2103 (void) execv(args[0], (char* const*) args);
2104 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
2105 } else
2106 log_warning_errno(errno, "Failed to execute /sbin/init, giving up: %m");
2107 }
2108
2109 arg_serialization = safe_fclose(arg_serialization);
2110 fds = fdset_free(fds);
2111
2112 #ifdef HAVE_VALGRIND_VALGRIND_H
2113 /* If we are PID 1 and running under valgrind, then let's exit
2114 * here explicitly. valgrind will only generate nice output on
2115 * exit(), not on exec(), hence let's do the former not the
2116 * latter here. */
2117 if (getpid() == 1 && RUNNING_ON_VALGRIND)
2118 return 0;
2119 #endif
2120
2121 if (shutdown_verb) {
2122 char log_level[DECIMAL_STR_MAX(int) + 1];
2123 char exit_code[DECIMAL_STR_MAX(uint8_t) + 1];
2124 const char* command_line[11] = {
2125 SYSTEMD_SHUTDOWN_BINARY_PATH,
2126 shutdown_verb,
2127 "--log-level", log_level,
2128 "--log-target",
2129 };
2130 unsigned pos = 5;
2131 _cleanup_strv_free_ char **env_block = NULL;
2132
2133 assert(command_line[pos] == NULL);
2134 env_block = strv_copy(environ);
2135
2136 xsprintf(log_level, "%d", log_get_max_level());
2137
2138 switch (log_get_target()) {
2139
2140 case LOG_TARGET_KMSG:
2141 case LOG_TARGET_JOURNAL_OR_KMSG:
2142 case LOG_TARGET_SYSLOG_OR_KMSG:
2143 command_line[pos++] = "kmsg";
2144 break;
2145
2146 case LOG_TARGET_NULL:
2147 command_line[pos++] = "null";
2148 break;
2149
2150 case LOG_TARGET_CONSOLE:
2151 default:
2152 command_line[pos++] = "console";
2153 break;
2154 };
2155
2156 if (log_get_show_color())
2157 command_line[pos++] = "--log-color";
2158
2159 if (log_get_show_location())
2160 command_line[pos++] = "--log-location";
2161
2162 if (streq(shutdown_verb, "exit")) {
2163 command_line[pos++] = "--exit-code";
2164 command_line[pos++] = exit_code;
2165 xsprintf(exit_code, "%d", retval);
2166 }
2167
2168 assert(pos < ELEMENTSOF(command_line));
2169
2170 if (arm_reboot_watchdog && arg_shutdown_watchdog > 0 && arg_shutdown_watchdog != USEC_INFINITY) {
2171 char *e;
2172
2173 /* If we reboot let's set the shutdown
2174 * watchdog and tell the shutdown binary to
2175 * repeatedly ping it */
2176 r = watchdog_set_timeout(&arg_shutdown_watchdog);
2177 watchdog_close(r < 0);
2178
2179 /* Tell the binary how often to ping, ignore failure */
2180 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
2181 (void) strv_push(&env_block, e);
2182 } else
2183 watchdog_close(true);
2184
2185 /* Avoid the creation of new processes forked by the
2186 * kernel; at this point, we will not listen to the
2187 * signals anyway */
2188 if (detect_container() <= 0)
2189 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
2190
2191 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
2192 log_error_errno(errno, "Failed to execute shutdown binary, %s: %m",
2193 getpid() == 1 ? "freezing" : "quitting");
2194 }
2195
2196 if (getpid() == 1) {
2197 if (error_message)
2198 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2199 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
2200 "%s, freezing.", error_message);
2201 freeze_or_reboot();
2202 }
2203
2204 return retval;
2205 }
Unnamed repository; edit this file 'description' to name the repository.