]> git.ipfire.org Git - thirdparty/systemd.git/blob - src/core/main.c
util-lib: move a number of fs operations into fs-util.[ch]
[thirdparty/systemd.git] / src / core / main.c
1 /*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
2
3 /***
4 This file is part of systemd.
5
6 Copyright 2010 Lennart Poettering
7
8 systemd is free software; you can redistribute it and/or modify it
9 under the terms of the GNU Lesser General Public License as published by
10 the Free Software Foundation; either version 2.1 of the License, or
11 (at your option) any later version.
12
13 systemd is distributed in the hope that it will be useful, but
14 WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 Lesser General Public License for more details.
17
18 You should have received a copy of the GNU Lesser General Public License
19 along with systemd; If not, see <http://www.gnu.org/licenses/>.
20 ***/
21
22 #include <errno.h>
23 #include <fcntl.h>
24 #include <getopt.h>
25 #include <signal.h>
26 #include <stdio.h>
27 #include <string.h>
28 #include <sys/mount.h>
29 #include <sys/prctl.h>
30 #include <sys/reboot.h>
31 #include <sys/stat.h>
32 #include <unistd.h>
33 #ifdef HAVE_SECCOMP
34 #include <seccomp.h>
35 #endif
36 #ifdef HAVE_VALGRIND_VALGRIND_H
37 #include <valgrind/valgrind.h>
38 #endif
39
40 #include "sd-daemon.h"
41 #include "sd-bus.h"
42
43 #include "architecture.h"
44 #include "build.h"
45 #include "bus-error.h"
46 #include "bus-util.h"
47 #include "capability.h"
48 #include "clock-util.h"
49 #include "conf-parser.h"
50 #include "cpu-set-util.h"
51 #include "dbus-manager.h"
52 #include "def.h"
53 #include "env-util.h"
54 #include "fd-util.h"
55 #include "fdset.h"
56 #include "fileio.h"
57 #include "formats-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 "process-util.h"
73 #include "rlimit-util.h"
74 #include "selinux-setup.h"
75 #include "selinux-util.h"
76 #include "signal-util.h"
77 #include "smack-setup.h"
78 #include "special.h"
79 #include "strv.h"
80 #include "switch-root.h"
81 #include "terminal-util.h"
82 #include "user-util.h"
83 #include "virt.h"
84 #include "watchdog.h"
85
86 static enum {
87 ACTION_RUN,
88 ACTION_HELP,
89 ACTION_VERSION,
90 ACTION_TEST,
91 ACTION_DUMP_CONFIGURATION_ITEMS,
92 ACTION_DONE
93 } arg_action = ACTION_RUN;
94 static char *arg_default_unit = NULL;
95 static ManagerRunningAs arg_running_as = _MANAGER_RUNNING_AS_INVALID;
96 static bool arg_dump_core = true;
97 static int arg_crash_chvt = -1;
98 static bool arg_crash_shell = false;
99 static bool arg_crash_reboot = false;
100 static bool arg_confirm_spawn = false;
101 static ShowStatus arg_show_status = _SHOW_STATUS_UNSET;
102 static bool arg_switched_root = false;
103 static int arg_no_pager = -1;
104 static char ***arg_join_controllers = NULL;
105 static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
106 static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
107 static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
108 static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
109 static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
110 static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
111 static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
112 static usec_t arg_runtime_watchdog = 0;
113 static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
114 static char **arg_default_environment = NULL;
115 static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {};
116 static uint64_t arg_capability_bounding_set_drop = 0;
117 static nsec_t arg_timer_slack_nsec = NSEC_INFINITY;
118 static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
119 static Set* arg_syscall_archs = NULL;
120 static FILE* arg_serialization = NULL;
121 static bool arg_default_cpu_accounting = false;
122 static bool arg_default_blockio_accounting = false;
123 static bool arg_default_memory_accounting = false;
124 static bool arg_default_tasks_accounting = false;
125
126 static void pager_open_if_enabled(void) {
127
128 if (arg_no_pager <= 0)
129 return;
130
131 pager_open(false);
132 }
133
134 noreturn static void freeze_or_reboot(void) {
135
136 if (arg_crash_reboot) {
137 log_notice("Rebooting in 10s...");
138 (void) sleep(10);
139
140 log_notice("Rebooting now...");
141 (void) reboot(RB_AUTOBOOT);
142 log_emergency_errno(errno, "Failed to reboot: %m");
143 }
144
145 log_emergency("Freezing execution.");
146 freeze();
147 }
148
149 noreturn static void crash(int sig) {
150 struct sigaction sa;
151 pid_t pid;
152
153 if (getpid() != 1)
154 /* Pass this on immediately, if this is not PID 1 */
155 (void) raise(sig);
156 else if (!arg_dump_core)
157 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
158 else {
159 sa = (struct sigaction) {
160 .sa_handler = nop_signal_handler,
161 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
162 };
163
164 /* We want to wait for the core process, hence let's enable SIGCHLD */
165 (void) sigaction(SIGCHLD, &sa, NULL);
166
167 pid = raw_clone(SIGCHLD, NULL);
168 if (pid < 0)
169 log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
170 else if (pid == 0) {
171 struct rlimit rl = {
172 .rlim_cur = RLIM_INFINITY,
173 .rlim_max = RLIM_INFINITY,
174 };
175
176 /* Enable default signal handler for core dump */
177 sa = (struct sigaction) {
178 .sa_handler = SIG_DFL,
179 };
180 (void) sigaction(sig, &sa, NULL);
181
182 /* Don't limit the core dump size */
183 (void) setrlimit(RLIMIT_CORE, &rl);
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_FULL)
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, NULL);
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_proc_cmdline_item(const char *key, const char *value) {
299
300 static const char * const rlmap[] = {
301 "emergency", SPECIAL_EMERGENCY_TARGET,
302 "-b", SPECIAL_EMERGENCY_TARGET,
303 "rescue", SPECIAL_RESCUE_TARGET,
304 "single", SPECIAL_RESCUE_TARGET,
305 "-s", SPECIAL_RESCUE_TARGET,
306 "s", SPECIAL_RESCUE_TARGET,
307 "S", SPECIAL_RESCUE_TARGET,
308 "1", SPECIAL_RESCUE_TARGET,
309 "2", SPECIAL_MULTI_USER_TARGET,
310 "3", SPECIAL_MULTI_USER_TARGET,
311 "4", SPECIAL_MULTI_USER_TARGET,
312 "5", SPECIAL_GRAPHICAL_TARGET,
313 };
314 int r;
315
316 assert(key);
317
318 if (streq(key, "systemd.unit") && value) {
319
320 if (!in_initrd())
321 return free_and_strdup(&arg_default_unit, value);
322
323 } else if (streq(key, "rd.systemd.unit") && value) {
324
325 if (in_initrd())
326 return free_and_strdup(&arg_default_unit, value);
327
328 } else if (streq(key, "systemd.dump_core") && value) {
329
330 r = parse_boolean(value);
331 if (r < 0)
332 log_warning("Failed to parse dump core switch %s. Ignoring.", value);
333 else
334 arg_dump_core = r;
335
336 } else if (streq(key, "systemd.crash_chvt") && value) {
337
338 if (parse_crash_chvt(value) < 0)
339 log_warning("Failed to parse crash chvt switch %s. Ignoring.", value);
340
341 } else if (streq(key, "systemd.crash_shell") && value) {
342
343 r = parse_boolean(value);
344 if (r < 0)
345 log_warning("Failed to parse crash shell switch %s. Ignoring.", value);
346 else
347 arg_crash_shell = r;
348
349 } else if (streq(key, "systemd.crash_reboot") && value) {
350
351 r = parse_boolean(value);
352 if (r < 0)
353 log_warning("Failed to parse crash reboot switch %s. Ignoring.", value);
354 else
355 arg_crash_reboot = r;
356
357 } else if (streq(key, "systemd.confirm_spawn") && value) {
358
359 r = parse_boolean(value);
360 if (r < 0)
361 log_warning("Failed to parse confirm spawn switch %s. Ignoring.", value);
362 else
363 arg_confirm_spawn = r;
364
365 } else if (streq(key, "systemd.show_status") && value) {
366
367 r = parse_show_status(value, &arg_show_status);
368 if (r < 0)
369 log_warning("Failed to parse show status switch %s. Ignoring.", value);
370
371 } else if (streq(key, "systemd.default_standard_output") && value) {
372
373 r = exec_output_from_string(value);
374 if (r < 0)
375 log_warning("Failed to parse default standard output switch %s. Ignoring.", value);
376 else
377 arg_default_std_output = r;
378
379 } else if (streq(key, "systemd.default_standard_error") && value) {
380
381 r = exec_output_from_string(value);
382 if (r < 0)
383 log_warning("Failed to parse default standard error switch %s. Ignoring.", value);
384 else
385 arg_default_std_error = r;
386
387 } else if (streq(key, "systemd.setenv") && value) {
388
389 if (env_assignment_is_valid(value)) {
390 char **env;
391
392 env = strv_env_set(arg_default_environment, value);
393 if (env)
394 arg_default_environment = env;
395 else
396 log_warning_errno(ENOMEM, "Setting environment variable '%s' failed, ignoring: %m", value);
397 } else
398 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
399
400 } else if (streq(key, "quiet") && !value) {
401
402 if (arg_show_status == _SHOW_STATUS_UNSET)
403 arg_show_status = SHOW_STATUS_AUTO;
404
405 } else if (streq(key, "debug") && !value) {
406
407 /* Note that log_parse_environment() handles 'debug'
408 * too, and sets the log level to LOG_DEBUG. */
409
410 if (detect_container() > 0)
411 log_set_target(LOG_TARGET_CONSOLE);
412
413 } else if (!in_initrd() && !value) {
414 unsigned i;
415
416 /* SysV compatibility */
417 for (i = 0; i < ELEMENTSOF(rlmap); i += 2)
418 if (streq(key, rlmap[i]))
419 return free_and_strdup(&arg_default_unit, rlmap[i+1]);
420 }
421
422 return 0;
423 }
424
425 #define DEFINE_SETTER(name, func, descr) \
426 static int name(const char *unit, \
427 const char *filename, \
428 unsigned line, \
429 const char *section, \
430 unsigned section_line, \
431 const char *lvalue, \
432 int ltype, \
433 const char *rvalue, \
434 void *data, \
435 void *userdata) { \
436 \
437 int r; \
438 \
439 assert(filename); \
440 assert(lvalue); \
441 assert(rvalue); \
442 \
443 r = func(rvalue); \
444 if (r < 0) \
445 log_syntax(unit, LOG_ERR, filename, line, r, \
446 "Invalid " descr "'%s': %m", \
447 rvalue); \
448 \
449 return 0; \
450 }
451
452 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level")
453 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target")
454 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" )
455 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location")
456
457 static int config_parse_cpu_affinity2(
458 const char *unit,
459 const char *filename,
460 unsigned line,
461 const char *section,
462 unsigned section_line,
463 const char *lvalue,
464 int ltype,
465 const char *rvalue,
466 void *data,
467 void *userdata) {
468
469 _cleanup_cpu_free_ cpu_set_t *c = NULL;
470 int ncpus;
471
472 ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue);
473 if (ncpus < 0)
474 return ncpus;
475
476 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
477 log_warning("Failed to set CPU affinity: %m");
478
479 return 0;
480 }
481
482 static int config_parse_show_status(
483 const char* unit,
484 const char *filename,
485 unsigned line,
486 const char *section,
487 unsigned section_line,
488 const char *lvalue,
489 int ltype,
490 const char *rvalue,
491 void *data,
492 void *userdata) {
493
494 int k;
495 ShowStatus *b = data;
496
497 assert(filename);
498 assert(lvalue);
499 assert(rvalue);
500 assert(data);
501
502 k = parse_show_status(rvalue, b);
503 if (k < 0) {
504 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
505 return 0;
506 }
507
508 return 0;
509 }
510
511 static int config_parse_crash_chvt(
512 const char* unit,
513 const char *filename,
514 unsigned line,
515 const char *section,
516 unsigned section_line,
517 const char *lvalue,
518 int ltype,
519 const char *rvalue,
520 void *data,
521 void *userdata) {
522
523 int r;
524
525 assert(filename);
526 assert(lvalue);
527 assert(rvalue);
528
529 r = parse_crash_chvt(rvalue);
530 if (r < 0) {
531 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
532 return 0;
533 }
534
535 return 0;
536 }
537
538 static int config_parse_join_controllers(const char *unit,
539 const char *filename,
540 unsigned line,
541 const char *section,
542 unsigned section_line,
543 const char *lvalue,
544 int ltype,
545 const char *rvalue,
546 void *data,
547 void *userdata) {
548
549 const char *whole_rvalue = rvalue;
550 unsigned n = 0;
551
552 assert(filename);
553 assert(lvalue);
554 assert(rvalue);
555
556 arg_join_controllers = strv_free_free(arg_join_controllers);
557
558 for (;;) {
559 _cleanup_free_ char *word = NULL;
560 char **l;
561 int r;
562
563 r = extract_first_word(&rvalue, &word, WHITESPACE, EXTRACT_QUOTES);
564 if (r < 0) {
565 log_syntax(unit, LOG_ERR, filename, line, r, "Invalid value for %s: %s", lvalue, whole_rvalue);
566 return r;
567 }
568 if (r == 0)
569 break;
570
571 l = strv_split(word, ",");
572 if (!l)
573 return log_oom();
574 strv_uniq(l);
575
576 if (strv_length(l) <= 1) {
577 strv_free(l);
578 continue;
579 }
580
581 if (!arg_join_controllers) {
582 arg_join_controllers = new(char**, 2);
583 if (!arg_join_controllers) {
584 strv_free(l);
585 return log_oom();
586 }
587
588 arg_join_controllers[0] = l;
589 arg_join_controllers[1] = NULL;
590
591 n = 1;
592 } else {
593 char ***a;
594 char ***t;
595
596 t = new0(char**, n+2);
597 if (!t) {
598 strv_free(l);
599 return log_oom();
600 }
601
602 n = 0;
603
604 for (a = arg_join_controllers; *a; a++) {
605
606 if (strv_overlap(*a, l)) {
607 if (strv_extend_strv(&l, *a, false) < 0) {
608 strv_free(l);
609 strv_free_free(t);
610 return log_oom();
611 }
612
613 } else {
614 char **c;
615
616 c = strv_copy(*a);
617 if (!c) {
618 strv_free(l);
619 strv_free_free(t);
620 return log_oom();
621 }
622
623 t[n++] = c;
624 }
625 }
626
627 t[n++] = strv_uniq(l);
628
629 strv_free_free(arg_join_controllers);
630 arg_join_controllers = t;
631 }
632 }
633 if (!isempty(rvalue))
634 log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring.");
635
636 return 0;
637 }
638
639 static int parse_config_file(void) {
640
641 const ConfigTableItem items[] = {
642 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
643 { "Manager", "LogTarget", config_parse_target, 0, NULL },
644 { "Manager", "LogColor", config_parse_color, 0, NULL },
645 { "Manager", "LogLocation", config_parse_location, 0, NULL },
646 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
647 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL },
648 { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL },
649 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
650 { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
651 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
652 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
653 { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
654 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
655 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
656 { "Manager", "CapabilityBoundingSet", config_parse_bounding_set, 0, &arg_capability_bounding_set_drop },
657 #ifdef HAVE_SECCOMP
658 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
659 #endif
660 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
661 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
662 { "Manager", "DefaultStandardOutput", config_parse_output, 0, &arg_default_std_output },
663 { "Manager", "DefaultStandardError", config_parse_output, 0, &arg_default_std_error },
664 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
665 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
666 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
667 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval },
668 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
669 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
670 { "Manager", "DefaultLimitCPU", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CPU] },
671 { "Manager", "DefaultLimitFSIZE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_FSIZE] },
672 { "Manager", "DefaultLimitDATA", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_DATA] },
673 { "Manager", "DefaultLimitSTACK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_STACK] },
674 { "Manager", "DefaultLimitCORE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CORE] },
675 { "Manager", "DefaultLimitRSS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RSS] },
676 { "Manager", "DefaultLimitNOFILE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NOFILE] },
677 { "Manager", "DefaultLimitAS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_AS] },
678 { "Manager", "DefaultLimitNPROC", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NPROC] },
679 { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MEMLOCK] },
680 { "Manager", "DefaultLimitLOCKS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_LOCKS] },
681 { "Manager", "DefaultLimitSIGPENDING", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_SIGPENDING] },
682 { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MSGQUEUE] },
683 { "Manager", "DefaultLimitNICE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NICE] },
684 { "Manager", "DefaultLimitRTPRIO", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTPRIO] },
685 { "Manager", "DefaultLimitRTTIME", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTTIME] },
686 { "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_default_cpu_accounting },
687 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
688 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
689 { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
690 {}
691 };
692
693 const char *fn, *conf_dirs_nulstr;
694
695 fn = arg_running_as == MANAGER_SYSTEM ? PKGSYSCONFDIR "/system.conf" : PKGSYSCONFDIR "/user.conf";
696 conf_dirs_nulstr = arg_running_as == MANAGER_SYSTEM ? CONF_DIRS_NULSTR("systemd/system.conf") : CONF_DIRS_NULSTR("systemd/user.conf");
697 config_parse_many(fn, conf_dirs_nulstr, "Manager\0",
698 config_item_table_lookup, items, false, NULL);
699
700 return 0;
701 }
702
703 static void manager_set_defaults(Manager *m) {
704
705 assert(m);
706
707 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
708 m->default_std_output = arg_default_std_output;
709 m->default_std_error = arg_default_std_error;
710 m->default_timeout_start_usec = arg_default_timeout_start_usec;
711 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
712 m->default_restart_usec = arg_default_restart_usec;
713 m->default_start_limit_interval = arg_default_start_limit_interval;
714 m->default_start_limit_burst = arg_default_start_limit_burst;
715 m->default_cpu_accounting = arg_default_cpu_accounting;
716 m->default_blockio_accounting = arg_default_blockio_accounting;
717 m->default_memory_accounting = arg_default_memory_accounting;
718 m->default_tasks_accounting = arg_default_tasks_accounting;
719
720 manager_set_default_rlimits(m, arg_default_rlimit);
721 manager_environment_add(m, NULL, arg_default_environment);
722 }
723
724 static int parse_argv(int argc, char *argv[]) {
725
726 enum {
727 ARG_LOG_LEVEL = 0x100,
728 ARG_LOG_TARGET,
729 ARG_LOG_COLOR,
730 ARG_LOG_LOCATION,
731 ARG_UNIT,
732 ARG_SYSTEM,
733 ARG_USER,
734 ARG_TEST,
735 ARG_NO_PAGER,
736 ARG_VERSION,
737 ARG_DUMP_CONFIGURATION_ITEMS,
738 ARG_DUMP_CORE,
739 ARG_CRASH_CHVT,
740 ARG_CRASH_SHELL,
741 ARG_CRASH_REBOOT,
742 ARG_CONFIRM_SPAWN,
743 ARG_SHOW_STATUS,
744 ARG_DESERIALIZE,
745 ARG_SWITCHED_ROOT,
746 ARG_DEFAULT_STD_OUTPUT,
747 ARG_DEFAULT_STD_ERROR
748 };
749
750 static const struct option options[] = {
751 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
752 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
753 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
754 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
755 { "unit", required_argument, NULL, ARG_UNIT },
756 { "system", no_argument, NULL, ARG_SYSTEM },
757 { "user", no_argument, NULL, ARG_USER },
758 { "test", no_argument, NULL, ARG_TEST },
759 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
760 { "help", no_argument, NULL, 'h' },
761 { "version", no_argument, NULL, ARG_VERSION },
762 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
763 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
764 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
765 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
766 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
767 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
768 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
769 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
770 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
771 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
772 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
773 {}
774 };
775
776 int c, r;
777
778 assert(argc >= 1);
779 assert(argv);
780
781 if (getpid() == 1)
782 opterr = 0;
783
784 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
785
786 switch (c) {
787
788 case ARG_LOG_LEVEL:
789 r = log_set_max_level_from_string(optarg);
790 if (r < 0) {
791 log_error("Failed to parse log level %s.", optarg);
792 return r;
793 }
794
795 break;
796
797 case ARG_LOG_TARGET:
798 r = log_set_target_from_string(optarg);
799 if (r < 0) {
800 log_error("Failed to parse log target %s.", optarg);
801 return r;
802 }
803
804 break;
805
806 case ARG_LOG_COLOR:
807
808 if (optarg) {
809 r = log_show_color_from_string(optarg);
810 if (r < 0) {
811 log_error("Failed to parse log color setting %s.", optarg);
812 return r;
813 }
814 } else
815 log_show_color(true);
816
817 break;
818
819 case ARG_LOG_LOCATION:
820 if (optarg) {
821 r = log_show_location_from_string(optarg);
822 if (r < 0) {
823 log_error("Failed to parse log location setting %s.", optarg);
824 return r;
825 }
826 } else
827 log_show_location(true);
828
829 break;
830
831 case ARG_DEFAULT_STD_OUTPUT:
832 r = exec_output_from_string(optarg);
833 if (r < 0) {
834 log_error("Failed to parse default standard output setting %s.", optarg);
835 return r;
836 } else
837 arg_default_std_output = r;
838 break;
839
840 case ARG_DEFAULT_STD_ERROR:
841 r = exec_output_from_string(optarg);
842 if (r < 0) {
843 log_error("Failed to parse default standard error output setting %s.", optarg);
844 return r;
845 } else
846 arg_default_std_error = r;
847 break;
848
849 case ARG_UNIT:
850
851 r = free_and_strdup(&arg_default_unit, optarg);
852 if (r < 0)
853 return log_error_errno(r, "Failed to set default unit %s: %m", optarg);
854
855 break;
856
857 case ARG_SYSTEM:
858 arg_running_as = MANAGER_SYSTEM;
859 break;
860
861 case ARG_USER:
862 arg_running_as = MANAGER_USER;
863 break;
864
865 case ARG_TEST:
866 arg_action = ACTION_TEST;
867 if (arg_no_pager < 0)
868 arg_no_pager = true;
869 break;
870
871 case ARG_NO_PAGER:
872 arg_no_pager = true;
873 break;
874
875 case ARG_VERSION:
876 arg_action = ACTION_VERSION;
877 break;
878
879 case ARG_DUMP_CONFIGURATION_ITEMS:
880 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
881 break;
882
883 case ARG_DUMP_CORE:
884 if (!optarg)
885 arg_dump_core = true;
886 else {
887 r = parse_boolean(optarg);
888 if (r < 0)
889 return log_error_errno(r, "Failed to parse dump core boolean: %s", optarg);
890 arg_dump_core = r;
891 }
892 break;
893
894 case ARG_CRASH_CHVT:
895 r = parse_crash_chvt(optarg);
896 if (r < 0)
897 return log_error_errno(r, "Failed to parse crash virtual terminal index: %s", optarg);
898 break;
899
900 case ARG_CRASH_SHELL:
901 if (!optarg)
902 arg_crash_shell = true;
903 else {
904 r = parse_boolean(optarg);
905 if (r < 0)
906 return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg);
907 arg_crash_shell = r;
908 }
909 break;
910
911 case ARG_CRASH_REBOOT:
912 if (!optarg)
913 arg_crash_reboot = true;
914 else {
915 r = parse_boolean(optarg);
916 if (r < 0)
917 return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg);
918 arg_crash_reboot = r;
919 }
920 break;
921
922 case ARG_CONFIRM_SPAWN:
923 r = optarg ? parse_boolean(optarg) : 1;
924 if (r < 0) {
925 log_error("Failed to parse confirm spawn boolean %s.", optarg);
926 return r;
927 }
928 arg_confirm_spawn = r;
929 break;
930
931 case ARG_SHOW_STATUS:
932 if (optarg) {
933 r = parse_show_status(optarg, &arg_show_status);
934 if (r < 0) {
935 log_error("Failed to parse show status boolean %s.", optarg);
936 return r;
937 }
938 } else
939 arg_show_status = SHOW_STATUS_YES;
940 break;
941
942 case ARG_DESERIALIZE: {
943 int fd;
944 FILE *f;
945
946 r = safe_atoi(optarg, &fd);
947 if (r < 0 || fd < 0) {
948 log_error("Failed to parse deserialize option %s.", optarg);
949 return -EINVAL;
950 }
951
952 (void) fd_cloexec(fd, true);
953
954 f = fdopen(fd, "r");
955 if (!f)
956 return log_error_errno(errno, "Failed to open serialization fd: %m");
957
958 safe_fclose(arg_serialization);
959 arg_serialization = f;
960
961 break;
962 }
963
964 case ARG_SWITCHED_ROOT:
965 arg_switched_root = true;
966 break;
967
968 case 'h':
969 arg_action = ACTION_HELP;
970 if (arg_no_pager < 0)
971 arg_no_pager = true;
972 break;
973
974 case 'D':
975 log_set_max_level(LOG_DEBUG);
976 break;
977
978 case 'b':
979 case 's':
980 case 'z':
981 /* Just to eat away the sysvinit kernel
982 * cmdline args without getopt() error
983 * messages that we'll parse in
984 * parse_proc_cmdline_word() or ignore. */
985
986 case '?':
987 if (getpid() != 1)
988 return -EINVAL;
989 else
990 return 0;
991
992 default:
993 assert_not_reached("Unhandled option code.");
994 }
995
996 if (optind < argc && getpid() != 1) {
997 /* Hmm, when we aren't run as init system
998 * let's complain about excess arguments */
999
1000 log_error("Excess arguments.");
1001 return -EINVAL;
1002 }
1003
1004 return 0;
1005 }
1006
1007 static int help(void) {
1008
1009 printf("%s [OPTIONS...]\n\n"
1010 "Starts up and maintains the system or user services.\n\n"
1011 " -h --help Show this help\n"
1012 " --test Determine startup sequence, dump it and exit\n"
1013 " --no-pager Do not pipe output into a pager\n"
1014 " --dump-configuration-items Dump understood unit configuration items\n"
1015 " --unit=UNIT Set default unit\n"
1016 " --system Run a system instance, even if PID != 1\n"
1017 " --user Run a user instance\n"
1018 " --dump-core[=BOOL] Dump core on crash\n"
1019 " --crash-vt=NR Change to specified VT on crash\n"
1020 " --crash-reboot[=BOOL] Reboot on crash\n"
1021 " --crash-shell[=BOOL] Run shell on crash\n"
1022 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1023 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1024 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1025 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1026 " --log-color[=BOOL] Highlight important log messages\n"
1027 " --log-location[=BOOL] Include code location in log messages\n"
1028 " --default-standard-output= Set default standard output for services\n"
1029 " --default-standard-error= Set default standard error output for services\n",
1030 program_invocation_short_name);
1031
1032 return 0;
1033 }
1034
1035 static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
1036 _cleanup_fdset_free_ FDSet *fds = NULL;
1037 _cleanup_fclose_ FILE *f = NULL;
1038 int r;
1039
1040 assert(m);
1041 assert(_f);
1042 assert(_fds);
1043
1044 r = manager_open_serialization(m, &f);
1045 if (r < 0)
1046 return log_error_errno(r, "Failed to create serialization file: %m");
1047
1048 /* Make sure nothing is really destructed when we shut down */
1049 m->n_reloading ++;
1050 bus_manager_send_reloading(m, true);
1051
1052 fds = fdset_new();
1053 if (!fds)
1054 return log_oom();
1055
1056 r = manager_serialize(m, f, fds, switching_root);
1057 if (r < 0)
1058 return log_error_errno(r, "Failed to serialize state: %m");
1059
1060 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1061 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1062
1063 r = fd_cloexec(fileno(f), false);
1064 if (r < 0)
1065 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1066
1067 r = fdset_cloexec(fds, false);
1068 if (r < 0)
1069 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1070
1071 *_f = f;
1072 *_fds = fds;
1073
1074 f = NULL;
1075 fds = NULL;
1076
1077 return 0;
1078 }
1079
1080 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1081 struct rlimit nl;
1082 int r;
1083
1084 assert(saved_rlimit);
1085
1086 /* Save the original RLIMIT_NOFILE so that we can reset it
1087 * later when transitioning from the initrd to the main
1088 * systemd or suchlike. */
1089 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1090 return log_error_errno(errno, "Reading RLIMIT_NOFILE failed: %m");
1091
1092 /* Make sure forked processes get the default kernel setting */
1093 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1094 struct rlimit *rl;
1095
1096 rl = newdup(struct rlimit, saved_rlimit, 1);
1097 if (!rl)
1098 return log_oom();
1099
1100 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1101 }
1102
1103 /* Bump up the resource limit for ourselves substantially */
1104 nl.rlim_cur = nl.rlim_max = 64*1024;
1105 r = setrlimit_closest(RLIMIT_NOFILE, &nl);
1106 if (r < 0)
1107 return log_error_errno(r, "Setting RLIMIT_NOFILE failed: %m");
1108
1109 return 0;
1110 }
1111
1112 static void test_mtab(void) {
1113
1114 static const char ok[] =
1115 "/proc/self/mounts\0"
1116 "/proc/mounts\0"
1117 "../proc/self/mounts\0"
1118 "../proc/mounts\0";
1119
1120 _cleanup_free_ char *p = NULL;
1121 int r;
1122
1123 /* Check that /etc/mtab is a symlink to the right place or
1124 * non-existing. But certainly not a file, or a symlink to
1125 * some weird place... */
1126
1127 r = readlink_malloc("/etc/mtab", &p);
1128 if (r == -ENOENT)
1129 return;
1130 if (r >= 0 && nulstr_contains(ok, p))
1131 return;
1132
1133 log_error("/etc/mtab is not a symlink or not pointing to /proc/self/mounts. "
1134 "This is not supported anymore. "
1135 "Please replace /etc/mtab with a symlink to /proc/self/mounts.");
1136 freeze_or_reboot();
1137 }
1138
1139 static void test_usr(void) {
1140
1141 /* Check that /usr is not a separate fs */
1142
1143 if (dir_is_empty("/usr") <= 0)
1144 return;
1145
1146 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1147 "Some things will probably break (sometimes even silently) in mysterious ways. "
1148 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1149 }
1150
1151 static int initialize_join_controllers(void) {
1152 /* By default, mount "cpu" + "cpuacct" together, and "net_cls"
1153 * + "net_prio". We'd like to add "cpuset" to the mix, but
1154 * "cpuset" doesn't really work for groups with no initialized
1155 * attributes. */
1156
1157 arg_join_controllers = new(char**, 3);
1158 if (!arg_join_controllers)
1159 return -ENOMEM;
1160
1161 arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL);
1162 if (!arg_join_controllers[0])
1163 goto oom;
1164
1165 arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL);
1166 if (!arg_join_controllers[1])
1167 goto oom;
1168
1169 arg_join_controllers[2] = NULL;
1170 return 0;
1171
1172 oom:
1173 arg_join_controllers = strv_free_free(arg_join_controllers);
1174 return -ENOMEM;
1175 }
1176
1177 static int enforce_syscall_archs(Set *archs) {
1178 #ifdef HAVE_SECCOMP
1179 scmp_filter_ctx *seccomp;
1180 Iterator i;
1181 void *id;
1182 int r;
1183
1184 seccomp = seccomp_init(SCMP_ACT_ALLOW);
1185 if (!seccomp)
1186 return log_oom();
1187
1188 SET_FOREACH(id, arg_syscall_archs, i) {
1189 r = seccomp_arch_add(seccomp, PTR_TO_UINT32(id) - 1);
1190 if (r == -EEXIST)
1191 continue;
1192 if (r < 0) {
1193 log_error_errno(r, "Failed to add architecture to seccomp: %m");
1194 goto finish;
1195 }
1196 }
1197
1198 r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
1199 if (r < 0) {
1200 log_error_errno(r, "Failed to unset NO_NEW_PRIVS: %m");
1201 goto finish;
1202 }
1203
1204 r = seccomp_load(seccomp);
1205 if (r < 0)
1206 log_error_errno(r, "Failed to add install architecture seccomp: %m");
1207
1208 finish:
1209 seccomp_release(seccomp);
1210 return r;
1211 #else
1212 return 0;
1213 #endif
1214 }
1215
1216 static int status_welcome(void) {
1217 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1218 int r;
1219
1220 r = parse_env_file("/etc/os-release", NEWLINE,
1221 "PRETTY_NAME", &pretty_name,
1222 "ANSI_COLOR", &ansi_color,
1223 NULL);
1224 if (r == -ENOENT)
1225 r = parse_env_file("/usr/lib/os-release", NEWLINE,
1226 "PRETTY_NAME", &pretty_name,
1227 "ANSI_COLOR", &ansi_color,
1228 NULL);
1229
1230 if (r < 0 && r != -ENOENT)
1231 log_warning_errno(r, "Failed to read os-release file: %m");
1232
1233 return status_printf(NULL, false, false,
1234 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1235 isempty(ansi_color) ? "1" : ansi_color,
1236 isempty(pretty_name) ? "Linux" : pretty_name);
1237 }
1238
1239 static int write_container_id(void) {
1240 const char *c;
1241
1242 c = getenv("container");
1243 if (isempty(c))
1244 return 0;
1245
1246 return write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1247 }
1248
1249 int main(int argc, char *argv[]) {
1250 Manager *m = NULL;
1251 int r, retval = EXIT_FAILURE;
1252 usec_t before_startup, after_startup;
1253 char timespan[FORMAT_TIMESPAN_MAX];
1254 FDSet *fds = NULL;
1255 bool reexecute = false;
1256 const char *shutdown_verb = NULL;
1257 dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL;
1258 dual_timestamp userspace_timestamp = DUAL_TIMESTAMP_NULL;
1259 dual_timestamp kernel_timestamp = DUAL_TIMESTAMP_NULL;
1260 dual_timestamp security_start_timestamp = DUAL_TIMESTAMP_NULL;
1261 dual_timestamp security_finish_timestamp = DUAL_TIMESTAMP_NULL;
1262 static char systemd[] = "systemd";
1263 bool skip_setup = false;
1264 unsigned j;
1265 bool loaded_policy = false;
1266 bool arm_reboot_watchdog = false;
1267 bool queue_default_job = false;
1268 bool empty_etc = false;
1269 char *switch_root_dir = NULL, *switch_root_init = NULL;
1270 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0);
1271 const char *error_message = NULL;
1272
1273 #ifdef HAVE_SYSV_COMPAT
1274 if (getpid() != 1 && strstr(program_invocation_short_name, "init")) {
1275 /* This is compatibility support for SysV, where
1276 * calling init as a user is identical to telinit. */
1277
1278 errno = -ENOENT;
1279 execv(SYSTEMCTL_BINARY_PATH, argv);
1280 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1281 return 1;
1282 }
1283 #endif
1284
1285 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
1286 dual_timestamp_get(&userspace_timestamp);
1287
1288 /* Determine if this is a reexecution or normal bootup. We do
1289 * the full command line parsing much later, so let's just
1290 * have a quick peek here. */
1291 if (strv_find(argv+1, "--deserialize"))
1292 skip_setup = true;
1293
1294 /* If we have switched root, do all the special setup
1295 * things */
1296 if (strv_find(argv+1, "--switched-root"))
1297 skip_setup = false;
1298
1299 /* If we get started via the /sbin/init symlink then we are
1300 called 'init'. After a subsequent reexecution we are then
1301 called 'systemd'. That is confusing, hence let's call us
1302 systemd right-away. */
1303 program_invocation_short_name = systemd;
1304 prctl(PR_SET_NAME, systemd);
1305
1306 saved_argv = argv;
1307 saved_argc = argc;
1308
1309 log_show_color(isatty(STDERR_FILENO) > 0);
1310 log_set_upgrade_syslog_to_journal(true);
1311
1312 /* Disable the umask logic */
1313 if (getpid() == 1)
1314 umask(0);
1315
1316 if (getpid() == 1 && detect_container() <= 0) {
1317
1318 /* Running outside of a container as PID 1 */
1319 arg_running_as = MANAGER_SYSTEM;
1320 make_null_stdio();
1321 log_set_target(LOG_TARGET_KMSG);
1322 log_open();
1323
1324 if (in_initrd())
1325 initrd_timestamp = userspace_timestamp;
1326
1327 if (!skip_setup) {
1328 mount_setup_early();
1329 dual_timestamp_get(&security_start_timestamp);
1330 if (mac_selinux_setup(&loaded_policy) < 0) {
1331 error_message = "Failed to load SELinux policy";
1332 goto finish;
1333 } else if (ima_setup() < 0) {
1334 error_message = "Failed to load IMA policy";
1335 goto finish;
1336 } else if (mac_smack_setup(&loaded_policy) < 0) {
1337 error_message = "Failed to load SMACK policy";
1338 goto finish;
1339 }
1340 dual_timestamp_get(&security_finish_timestamp);
1341 }
1342
1343 if (mac_selinux_init(NULL) < 0) {
1344 error_message = "Failed to initialize SELinux policy";
1345 goto finish;
1346 }
1347
1348 if (!skip_setup) {
1349 if (clock_is_localtime() > 0) {
1350 int min;
1351
1352 /*
1353 * The very first call of settimeofday() also does a time warp in the kernel.
1354 *
1355 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on
1356 * external tools to take care of maintaining the RTC and do all adjustments.
1357 * This matches the behavior of Windows, which leaves the RTC alone if the
1358 * registry tells that the RTC runs in UTC.
1359 */
1360 r = clock_set_timezone(&min);
1361 if (r < 0)
1362 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1363 else
1364 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1365 } else if (!in_initrd()) {
1366 /*
1367 * Do a dummy very first call to seal the kernel's time warp magic.
1368 *
1369 * Do not call this this from inside the initrd. The initrd might not
1370 * carry /etc/adjtime with LOCAL, but the real system could be set up
1371 * that way. In such case, we need to delay the time-warp or the sealing
1372 * until we reach the real system.
1373 *
1374 * Do no set the kernel's timezone. The concept of local time cannot
1375 * be supported reliably, the time will jump or be incorrect at every daylight
1376 * saving time change. All kernel local time concepts will be treated
1377 * as UTC that way.
1378 */
1379 clock_reset_timewarp();
1380 }
1381 }
1382
1383 /* Set the default for later on, but don't actually
1384 * open the logs like this for now. Note that if we
1385 * are transitioning from the initrd there might still
1386 * be journal fd open, and we shouldn't attempt
1387 * opening that before we parsed /proc/cmdline which
1388 * might redirect output elsewhere. */
1389 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
1390
1391 } else if (getpid() == 1) {
1392 /* Running inside a container, as PID 1 */
1393 arg_running_as = MANAGER_SYSTEM;
1394 log_set_target(LOG_TARGET_CONSOLE);
1395 log_close_console(); /* force reopen of /dev/console */
1396 log_open();
1397
1398 /* For the later on, see above... */
1399 log_set_target(LOG_TARGET_JOURNAL);
1400
1401 /* clear the kernel timestamp,
1402 * because we are in a container */
1403 kernel_timestamp.monotonic = 0ULL;
1404 kernel_timestamp.realtime = 0ULL;
1405
1406 } else {
1407 /* Running as user instance */
1408 arg_running_as = MANAGER_USER;
1409 log_set_target(LOG_TARGET_AUTO);
1410 log_open();
1411
1412 /* clear the kernel timestamp,
1413 * because we are not PID 1 */
1414 kernel_timestamp = DUAL_TIMESTAMP_NULL;
1415 }
1416
1417 /* Initialize default unit */
1418 r = free_and_strdup(&arg_default_unit, SPECIAL_DEFAULT_TARGET);
1419 if (r < 0) {
1420 log_emergency_errno(r, "Failed to set default unit %s: %m", SPECIAL_DEFAULT_TARGET);
1421 error_message = "Failed to set default unit";
1422 goto finish;
1423 }
1424
1425 r = initialize_join_controllers();
1426 if (r < 0) {
1427 error_message = "Failed to initialize cgroup controllers";
1428 goto finish;
1429 }
1430
1431 /* Mount /proc, /sys and friends, so that /proc/cmdline and
1432 * /proc/$PID/fd is available. */
1433 if (getpid() == 1) {
1434
1435 /* Load the kernel modules early, so that we kdbus.ko is loaded before kdbusfs shall be mounted */
1436 if (!skip_setup)
1437 kmod_setup();
1438
1439 r = mount_setup(loaded_policy);
1440 if (r < 0) {
1441 error_message = "Failed to mount API filesystems";
1442 goto finish;
1443 }
1444 }
1445
1446 /* Reset all signal handlers. */
1447 (void) reset_all_signal_handlers();
1448 (void) ignore_signals(SIGNALS_IGNORE, -1);
1449
1450 if (parse_config_file() < 0) {
1451 error_message = "Failed to parse config file";
1452 goto finish;
1453 }
1454
1455 if (arg_running_as == MANAGER_SYSTEM) {
1456 r = parse_proc_cmdline(parse_proc_cmdline_item);
1457 if (r < 0)
1458 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
1459 }
1460
1461 /* Note that this also parses bits from the kernel command
1462 * line, including "debug". */
1463 log_parse_environment();
1464
1465 if (parse_argv(argc, argv) < 0) {
1466 error_message = "Failed to parse commandline arguments";
1467 goto finish;
1468 }
1469
1470 if (arg_action == ACTION_TEST &&
1471 geteuid() == 0) {
1472 log_error("Don't run test mode as root.");
1473 goto finish;
1474 }
1475
1476 if (arg_running_as == MANAGER_USER &&
1477 arg_action == ACTION_RUN &&
1478 sd_booted() <= 0) {
1479 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
1480 goto finish;
1481 }
1482
1483 if (arg_running_as == MANAGER_SYSTEM &&
1484 arg_action == ACTION_RUN &&
1485 running_in_chroot() > 0) {
1486 log_error("Cannot be run in a chroot() environment.");
1487 goto finish;
1488 }
1489
1490 if (arg_action == ACTION_TEST)
1491 skip_setup = true;
1492
1493 pager_open_if_enabled();
1494
1495 if (arg_action == ACTION_HELP) {
1496 retval = help();
1497 goto finish;
1498 } else if (arg_action == ACTION_VERSION) {
1499 retval = version();
1500 goto finish;
1501 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
1502 unit_dump_config_items(stdout);
1503 retval = EXIT_SUCCESS;
1504 goto finish;
1505 } else if (arg_action == ACTION_DONE) {
1506 retval = EXIT_SUCCESS;
1507 goto finish;
1508 }
1509
1510 if (arg_running_as == MANAGER_USER &&
1511 !getenv("XDG_RUNTIME_DIR")) {
1512 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
1513 goto finish;
1514 }
1515
1516 assert_se(arg_action == ACTION_RUN || arg_action == ACTION_TEST);
1517
1518 /* Close logging fds, in order not to confuse fdset below */
1519 log_close();
1520
1521 /* Remember open file descriptors for later deserialization */
1522 r = fdset_new_fill(&fds);
1523 if (r < 0) {
1524 log_emergency_errno(r, "Failed to allocate fd set: %m");
1525 error_message = "Failed to allocate fd set";
1526 goto finish;
1527 } else
1528 fdset_cloexec(fds, true);
1529
1530 if (arg_serialization)
1531 assert_se(fdset_remove(fds, fileno(arg_serialization)) >= 0);
1532
1533 if (arg_running_as == MANAGER_SYSTEM)
1534 /* Become a session leader if we aren't one yet. */
1535 setsid();
1536
1537 /* Move out of the way, so that we won't block unmounts */
1538 assert_se(chdir("/") == 0);
1539
1540 /* Reset the console, but only if this is really init and we
1541 * are freshly booted */
1542 if (arg_running_as == MANAGER_SYSTEM && arg_action == ACTION_RUN) {
1543
1544 /* If we are init, we connect stdin/stdout/stderr to
1545 * /dev/null and make sure we don't have a controlling
1546 * tty. */
1547 release_terminal();
1548
1549 if (getpid() == 1 && !skip_setup)
1550 console_setup();
1551 }
1552
1553 /* Open the logging devices, if possible and necessary */
1554 log_open();
1555
1556 if (arg_show_status == _SHOW_STATUS_UNSET)
1557 arg_show_status = SHOW_STATUS_YES;
1558
1559 /* Make sure we leave a core dump without panicing the
1560 * kernel. */
1561 if (getpid() == 1) {
1562 install_crash_handler();
1563
1564 r = mount_cgroup_controllers(arg_join_controllers);
1565 if (r < 0)
1566 goto finish;
1567 }
1568
1569 if (arg_running_as == MANAGER_SYSTEM) {
1570 int v;
1571
1572 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1573 arg_action == ACTION_TEST ? "test " : "" );
1574
1575 v = detect_virtualization();
1576 if (v > 0)
1577 log_info("Detected virtualization %s.", virtualization_to_string(v));
1578
1579 write_container_id();
1580
1581 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1582
1583 if (in_initrd())
1584 log_info("Running in initial RAM disk.");
1585
1586 /* Let's check whether /etc is already populated. We
1587 * don't actually really check for that, but use
1588 * /etc/machine-id as flag file. This allows container
1589 * managers and installers to provision a couple of
1590 * files already. If the container manager wants to
1591 * provision the machine ID itself it should pass
1592 * $container_uuid to PID 1. */
1593
1594 empty_etc = access("/etc/machine-id", F_OK) < 0;
1595 if (empty_etc)
1596 log_info("Running with unpopulated /etc.");
1597 } else {
1598 _cleanup_free_ char *t;
1599
1600 t = uid_to_name(getuid());
1601 log_debug(PACKAGE_STRING " running in %suser mode for user "UID_FMT"/%s. (" SYSTEMD_FEATURES ")",
1602 arg_action == ACTION_TEST ? " test" : "", getuid(), t);
1603 }
1604
1605 if (arg_running_as == MANAGER_SYSTEM && !skip_setup) {
1606 if (arg_show_status > 0)
1607 status_welcome();
1608
1609 hostname_setup();
1610 machine_id_setup(NULL);
1611 loopback_setup();
1612
1613 test_mtab();
1614 test_usr();
1615 }
1616
1617 if (arg_running_as == MANAGER_SYSTEM && arg_runtime_watchdog > 0)
1618 watchdog_set_timeout(&arg_runtime_watchdog);
1619
1620 if (arg_timer_slack_nsec != NSEC_INFINITY)
1621 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1622 log_error_errno(errno, "Failed to adjust timer slack: %m");
1623
1624 if (arg_capability_bounding_set_drop) {
1625 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set_drop);
1626 if (r < 0) {
1627 log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
1628 error_message = "Failed to drop capability bounding set of usermode helpers";
1629 goto finish;
1630 }
1631 r = capability_bounding_set_drop(arg_capability_bounding_set_drop, true);
1632 if (r < 0) {
1633 log_emergency_errno(r, "Failed to drop capability bounding set: %m");
1634 error_message = "Failed to drop capability bounding set";
1635 goto finish;
1636 }
1637 }
1638
1639 if (arg_syscall_archs) {
1640 r = enforce_syscall_archs(arg_syscall_archs);
1641 if (r < 0) {
1642 error_message = "Failed to set syscall architectures";
1643 goto finish;
1644 }
1645 }
1646
1647 if (arg_running_as == MANAGER_USER)
1648 /* Become reaper of our children */
1649 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
1650 log_warning_errno(errno, "Failed to make us a subreaper: %m");
1651
1652 if (arg_running_as == MANAGER_SYSTEM) {
1653 bump_rlimit_nofile(&saved_rlimit_nofile);
1654
1655 if (empty_etc) {
1656 r = unit_file_preset_all(UNIT_FILE_SYSTEM, false, NULL, UNIT_FILE_PRESET_ENABLE_ONLY, false, NULL, 0);
1657 if (r < 0)
1658 log_warning_errno(r, "Failed to populate /etc with preset unit settings, ignoring: %m");
1659 else
1660 log_info("Populated /etc with preset unit settings.");
1661 }
1662 }
1663
1664 r = manager_new(arg_running_as, arg_action == ACTION_TEST, &m);
1665 if (r < 0) {
1666 log_emergency_errno(r, "Failed to allocate manager object: %m");
1667 error_message = "Failed to allocate manager object";
1668 goto finish;
1669 }
1670
1671 m->confirm_spawn = arg_confirm_spawn;
1672 m->runtime_watchdog = arg_runtime_watchdog;
1673 m->shutdown_watchdog = arg_shutdown_watchdog;
1674 m->userspace_timestamp = userspace_timestamp;
1675 m->kernel_timestamp = kernel_timestamp;
1676 m->initrd_timestamp = initrd_timestamp;
1677 m->security_start_timestamp = security_start_timestamp;
1678 m->security_finish_timestamp = security_finish_timestamp;
1679
1680 manager_set_defaults(m);
1681 manager_set_show_status(m, arg_show_status);
1682 manager_set_first_boot(m, empty_etc);
1683
1684 /* Remember whether we should queue the default job */
1685 queue_default_job = !arg_serialization || arg_switched_root;
1686
1687 before_startup = now(CLOCK_MONOTONIC);
1688
1689 r = manager_startup(m, arg_serialization, fds);
1690 if (r < 0)
1691 log_error_errno(r, "Failed to fully start up daemon: %m");
1692
1693 /* This will close all file descriptors that were opened, but
1694 * not claimed by any unit. */
1695 fds = fdset_free(fds);
1696
1697 arg_serialization = safe_fclose(arg_serialization);
1698
1699 if (queue_default_job) {
1700 _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL;
1701 Unit *target = NULL;
1702 Job *default_unit_job;
1703
1704 log_debug("Activating default unit: %s", arg_default_unit);
1705
1706 r = manager_load_unit(m, arg_default_unit, NULL, &error, &target);
1707 if (r < 0)
1708 log_error("Failed to load default target: %s", bus_error_message(&error, r));
1709 else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND)
1710 log_error_errno(target->load_error, "Failed to load default target: %m");
1711 else if (target->load_state == UNIT_MASKED)
1712 log_error("Default target masked.");
1713
1714 if (!target || target->load_state != UNIT_LOADED) {
1715 log_info("Trying to load rescue target...");
1716
1717 r = manager_load_unit(m, SPECIAL_RESCUE_TARGET, NULL, &error, &target);
1718 if (r < 0) {
1719 log_emergency("Failed to load rescue target: %s", bus_error_message(&error, r));
1720 error_message = "Failed to load rescue target";
1721 goto finish;
1722 } else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND) {
1723 log_emergency_errno(target->load_error, "Failed to load rescue target: %m");
1724 error_message = "Failed to load rescue target";
1725 goto finish;
1726 } else if (target->load_state == UNIT_MASKED) {
1727 log_emergency("Rescue target masked.");
1728 error_message = "Rescue target masked";
1729 goto finish;
1730 }
1731 }
1732
1733 assert(target->load_state == UNIT_LOADED);
1734
1735 if (arg_action == ACTION_TEST) {
1736 printf("-> By units:\n");
1737 manager_dump_units(m, stdout, "\t");
1738 }
1739
1740 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, false, &error, &default_unit_job);
1741 if (r == -EPERM) {
1742 log_debug("Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1743
1744 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, false, &error, &default_unit_job);
1745 if (r < 0) {
1746 log_emergency("Failed to start default target: %s", bus_error_message(&error, r));
1747 error_message = "Failed to start default target";
1748 goto finish;
1749 }
1750 } else if (r < 0) {
1751 log_emergency("Failed to isolate default target: %s", bus_error_message(&error, r));
1752 error_message = "Failed to isolate default target";
1753 goto finish;
1754 }
1755
1756 m->default_unit_job_id = default_unit_job->id;
1757
1758 after_startup = now(CLOCK_MONOTONIC);
1759 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
1760 "Loaded units and determined initial transaction in %s.",
1761 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
1762
1763 if (arg_action == ACTION_TEST) {
1764 printf("-> By jobs:\n");
1765 manager_dump_jobs(m, stdout, "\t");
1766 retval = EXIT_SUCCESS;
1767 goto finish;
1768 }
1769 }
1770
1771 for (;;) {
1772 r = manager_loop(m);
1773 if (r < 0) {
1774 log_emergency_errno(r, "Failed to run main loop: %m");
1775 error_message = "Failed to run main loop";
1776 goto finish;
1777 }
1778
1779 switch (m->exit_code) {
1780
1781 case MANAGER_RELOAD:
1782 log_info("Reloading.");
1783
1784 r = parse_config_file();
1785 if (r < 0)
1786 log_error("Failed to parse config file.");
1787
1788 manager_set_defaults(m);
1789
1790 r = manager_reload(m);
1791 if (r < 0)
1792 log_error_errno(r, "Failed to reload: %m");
1793 break;
1794
1795 case MANAGER_REEXECUTE:
1796
1797 if (prepare_reexecute(m, &arg_serialization, &fds, false) < 0) {
1798 error_message = "Failed to prepare for reexecution";
1799 goto finish;
1800 }
1801
1802 reexecute = true;
1803 log_notice("Reexecuting.");
1804 goto finish;
1805
1806 case MANAGER_SWITCH_ROOT:
1807 /* Steal the switch root parameters */
1808 switch_root_dir = m->switch_root;
1809 switch_root_init = m->switch_root_init;
1810 m->switch_root = m->switch_root_init = NULL;
1811
1812 if (!switch_root_init)
1813 if (prepare_reexecute(m, &arg_serialization, &fds, true) < 0) {
1814 error_message = "Failed to prepare for reexecution";
1815 goto finish;
1816 }
1817
1818 reexecute = true;
1819 log_notice("Switching root.");
1820 goto finish;
1821
1822 case MANAGER_EXIT:
1823 retval = m->return_value;
1824
1825 if (m->running_as == MANAGER_USER) {
1826 log_debug("Exit.");
1827 goto finish;
1828 }
1829
1830 /* fallthrough */
1831 case MANAGER_REBOOT:
1832 case MANAGER_POWEROFF:
1833 case MANAGER_HALT:
1834 case MANAGER_KEXEC: {
1835 static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
1836 [MANAGER_EXIT] = "exit",
1837 [MANAGER_REBOOT] = "reboot",
1838 [MANAGER_POWEROFF] = "poweroff",
1839 [MANAGER_HALT] = "halt",
1840 [MANAGER_KEXEC] = "kexec"
1841 };
1842
1843 assert_se(shutdown_verb = table[m->exit_code]);
1844 arm_reboot_watchdog = m->exit_code == MANAGER_REBOOT;
1845
1846 log_notice("Shutting down.");
1847 goto finish;
1848 }
1849
1850 default:
1851 assert_not_reached("Unknown exit code.");
1852 }
1853 }
1854
1855 finish:
1856 pager_close();
1857
1858 if (m)
1859 arg_shutdown_watchdog = m->shutdown_watchdog;
1860
1861 m = manager_free(m);
1862
1863 for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++)
1864 arg_default_rlimit[j] = mfree(arg_default_rlimit[j]);
1865
1866 arg_default_unit = mfree(arg_default_unit);
1867 arg_join_controllers = strv_free_free(arg_join_controllers);
1868 arg_default_environment = strv_free(arg_default_environment);
1869 arg_syscall_archs = set_free(arg_syscall_archs);
1870
1871 mac_selinux_finish();
1872
1873 if (reexecute) {
1874 const char **args;
1875 unsigned i, args_size;
1876
1877 /* Close and disarm the watchdog, so that the new
1878 * instance can reinitialize it, but doesn't get
1879 * rebooted while we do that */
1880 watchdog_close(true);
1881
1882 /* Reset the RLIMIT_NOFILE to the kernel default, so
1883 * that the new systemd can pass the kernel default to
1884 * its child processes */
1885 if (saved_rlimit_nofile.rlim_cur > 0)
1886 (void) setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile);
1887
1888 if (switch_root_dir) {
1889 /* Kill all remaining processes from the
1890 * initrd, but don't wait for them, so that we
1891 * can handle the SIGCHLD for them after
1892 * deserializing. */
1893 broadcast_signal(SIGTERM, false, true);
1894
1895 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1896 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1897 if (r < 0)
1898 log_error_errno(r, "Failed to switch root, trying to continue: %m");
1899 }
1900
1901 args_size = MAX(6, argc+1);
1902 args = newa(const char*, args_size);
1903
1904 if (!switch_root_init) {
1905 char sfd[DECIMAL_STR_MAX(int) + 1];
1906
1907 /* First try to spawn ourselves with the right
1908 * path, and with full serialization. We do
1909 * this only if the user didn't specify an
1910 * explicit init to spawn. */
1911
1912 assert(arg_serialization);
1913 assert(fds);
1914
1915 xsprintf(sfd, "%i", fileno(arg_serialization));
1916
1917 i = 0;
1918 args[i++] = SYSTEMD_BINARY_PATH;
1919 if (switch_root_dir)
1920 args[i++] = "--switched-root";
1921 args[i++] = arg_running_as == MANAGER_SYSTEM ? "--system" : "--user";
1922 args[i++] = "--deserialize";
1923 args[i++] = sfd;
1924 args[i++] = NULL;
1925
1926 /* do not pass along the environment we inherit from the kernel or initrd */
1927 if (switch_root_dir)
1928 (void) clearenv();
1929
1930 assert(i <= args_size);
1931 (void) execv(args[0], (char* const*) args);
1932 }
1933
1934 /* Try the fallback, if there is any, without any
1935 * serialization. We pass the original argv[] and
1936 * envp[]. (Well, modulo the ordering changes due to
1937 * getopt() in argv[], and some cleanups in envp[],
1938 * but let's hope that doesn't matter.) */
1939
1940 arg_serialization = safe_fclose(arg_serialization);
1941 fds = fdset_free(fds);
1942
1943 /* Reopen the console */
1944 (void) make_console_stdio();
1945
1946 for (j = 1, i = 1; j < (unsigned) argc; j++)
1947 args[i++] = argv[j];
1948 args[i++] = NULL;
1949 assert(i <= args_size);
1950
1951 /* Reenable any blocked signals, especially important
1952 * if we switch from initial ramdisk to init=... */
1953 (void) reset_all_signal_handlers();
1954 (void) reset_signal_mask();
1955
1956 if (switch_root_init) {
1957 args[0] = switch_root_init;
1958 (void) execv(args[0], (char* const*) args);
1959 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
1960 }
1961
1962 args[0] = "/sbin/init";
1963 (void) execv(args[0], (char* const*) args);
1964
1965 if (errno == ENOENT) {
1966 log_warning("No /sbin/init, trying fallback");
1967
1968 args[0] = "/bin/sh";
1969 args[1] = NULL;
1970 (void) execv(args[0], (char* const*) args);
1971 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
1972 } else
1973 log_warning_errno(errno, "Failed to execute /sbin/init, giving up: %m");
1974 }
1975
1976 arg_serialization = safe_fclose(arg_serialization);
1977 fds = fdset_free(fds);
1978
1979 #ifdef HAVE_VALGRIND_VALGRIND_H
1980 /* If we are PID 1 and running under valgrind, then let's exit
1981 * here explicitly. valgrind will only generate nice output on
1982 * exit(), not on exec(), hence let's do the former not the
1983 * latter here. */
1984 if (getpid() == 1 && RUNNING_ON_VALGRIND)
1985 return 0;
1986 #endif
1987
1988 if (shutdown_verb) {
1989 char log_level[DECIMAL_STR_MAX(int) + 1];
1990 char exit_code[DECIMAL_STR_MAX(uint8_t) + 1];
1991 const char* command_line[11] = {
1992 SYSTEMD_SHUTDOWN_BINARY_PATH,
1993 shutdown_verb,
1994 "--log-level", log_level,
1995 "--log-target",
1996 };
1997 unsigned pos = 5;
1998 _cleanup_strv_free_ char **env_block = NULL;
1999
2000 assert(command_line[pos] == NULL);
2001 env_block = strv_copy(environ);
2002
2003 xsprintf(log_level, "%d", log_get_max_level());
2004
2005 switch (log_get_target()) {
2006
2007 case LOG_TARGET_KMSG:
2008 case LOG_TARGET_JOURNAL_OR_KMSG:
2009 case LOG_TARGET_SYSLOG_OR_KMSG:
2010 command_line[pos++] = "kmsg";
2011 break;
2012
2013 case LOG_TARGET_NULL:
2014 command_line[pos++] = "null";
2015 break;
2016
2017 case LOG_TARGET_CONSOLE:
2018 default:
2019 command_line[pos++] = "console";
2020 break;
2021 };
2022
2023 if (log_get_show_color())
2024 command_line[pos++] = "--log-color";
2025
2026 if (log_get_show_location())
2027 command_line[pos++] = "--log-location";
2028
2029 if (streq(shutdown_verb, "exit")) {
2030 command_line[pos++] = "--exit-code";
2031 command_line[pos++] = exit_code;
2032 xsprintf(exit_code, "%d", retval);
2033 }
2034
2035 assert(pos < ELEMENTSOF(command_line));
2036
2037 if (arm_reboot_watchdog && arg_shutdown_watchdog > 0) {
2038 char *e;
2039
2040 /* If we reboot let's set the shutdown
2041 * watchdog and tell the shutdown binary to
2042 * repeatedly ping it */
2043 r = watchdog_set_timeout(&arg_shutdown_watchdog);
2044 watchdog_close(r < 0);
2045
2046 /* Tell the binary how often to ping, ignore failure */
2047 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
2048 (void) strv_push(&env_block, e);
2049 } else
2050 watchdog_close(true);
2051
2052 /* Avoid the creation of new processes forked by the
2053 * kernel; at this point, we will not listen to the
2054 * signals anyway */
2055 if (detect_container() <= 0)
2056 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
2057
2058 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
2059 log_error_errno(errno, "Failed to execute shutdown binary, %s: %m",
2060 getpid() == 1 ? "freezing" : "quitting");
2061 }
2062
2063 if (getpid() == 1) {
2064 if (error_message)
2065 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2066 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
2067 "%s, freezing.", error_message);
2068 freeze_or_reboot();
2069 }
2070
2071 return retval;
2072 }