]> git.ipfire.org Git - thirdparty/systemd.git/blob - src/core/main.c
projects / thirdparty / systemd.git / blob
? search:


1783b9c7af127922108de0b2bdd98273425bccd7
[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 "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 "stat-util.h"
80 #include "stdio-util.h"
81 #include "strv.h"
82 #include "switch-root.h"
83 #include "terminal-util.h"
84 #include "user-util.h"
85 #include "virt.h"
86 #include "watchdog.h"
87
88 static enum {
89 ACTION_RUN,
90 ACTION_HELP,
91 ACTION_VERSION,
92 ACTION_TEST,
93 ACTION_DUMP_CONFIGURATION_ITEMS,
94 ACTION_DONE
95 } arg_action = ACTION_RUN;
96 static char *arg_default_unit = NULL;
97 static ManagerRunningAs arg_running_as = _MANAGER_RUNNING_AS_INVALID;
98 static bool arg_dump_core = true;
99 static int arg_crash_chvt = -1;
100 static bool arg_crash_shell = false;
101 static bool arg_crash_reboot = false;
102 static bool arg_confirm_spawn = false;
103 static ShowStatus arg_show_status = _SHOW_STATUS_UNSET;
104 static bool arg_switched_root = false;
105 static bool arg_no_pager = false;
106 static char ***arg_join_controllers = NULL;
107 static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
108 static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
109 static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
110 static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
111 static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
112 static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
113 static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
114 static usec_t arg_runtime_watchdog = 0;
115 static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
116 static char **arg_default_environment = NULL;
117 static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {};
118 static uint64_t arg_capability_bounding_set = CAP_ALL;
119 static nsec_t arg_timer_slack_nsec = NSEC_INFINITY;
120 static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
121 static Set* arg_syscall_archs = NULL;
122 static FILE* arg_serialization = NULL;
123 static bool arg_default_cpu_accounting = false;
124 static bool arg_default_blockio_accounting = false;
125 static bool arg_default_memory_accounting = false;
126 static bool arg_default_tasks_accounting = true;
127 static uint64_t arg_default_tasks_max = UINT64_C(512);
128 static sd_id128_t arg_machine_id = {};
129
130 noreturn static void freeze_or_reboot(void) {
131
132 if (arg_crash_reboot) {
133 log_notice("Rebooting in 10s...");
134 (void) sleep(10);
135
136 log_notice("Rebooting now...");
137 (void) reboot(RB_AUTOBOOT);
138 log_emergency_errno(errno, "Failed to reboot: %m");
139 }
140
141 log_emergency("Freezing execution.");
142 freeze();
143 }
144
145 noreturn static void crash(int sig) {
146 struct sigaction sa;
147 pid_t pid;
148
149 if (getpid() != 1)
150 /* Pass this on immediately, if this is not PID 1 */
151 (void) raise(sig);
152 else if (!arg_dump_core)
153 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
154 else {
155 sa = (struct sigaction) {
156 .sa_handler = nop_signal_handler,
157 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
158 };
159
160 /* We want to wait for the core process, hence let's enable SIGCHLD */
161 (void) sigaction(SIGCHLD, &sa, NULL);
162
163 pid = raw_clone(SIGCHLD, NULL);
164 if (pid < 0)
165 log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
166 else if (pid == 0) {
167 /* Enable default signal handler for core dump */
168
169 sa = (struct sigaction) {
170 .sa_handler = SIG_DFL,
171 };
172 (void) sigaction(sig, &sa, NULL);
173
174 /* Don't limit the coredump size */
175 (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
176
177 /* Just to be sure... */
178 (void) chdir("/");
179
180 /* Raise the signal again */
181 pid = raw_getpid();
182 (void) kill(pid, sig); /* raise() would kill the parent */
183
184 assert_not_reached("We shouldn't be here...");
185 _exit(EXIT_FAILURE);
186 } else {
187 siginfo_t status;
188 int r;
189
190 /* Order things nicely. */
191 r = wait_for_terminate(pid, &status);
192 if (r < 0)
193 log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
194 else if (status.si_code != CLD_DUMPED)
195 log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
196 signal_to_string(sig),
197 pid, sigchld_code_to_string(status.si_code),
198 status.si_status,
199 strna(status.si_code == CLD_EXITED
200 ? exit_status_to_string(status.si_status, EXIT_STATUS_FULL)
201 : signal_to_string(status.si_status)));
202 else
203 log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
204 }
205 }
206
207 if (arg_crash_chvt >= 0)
208 (void) chvt(arg_crash_chvt);
209
210 sa = (struct sigaction) {
211 .sa_handler = SIG_IGN,
212 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
213 };
214
215 /* Let the kernel reap children for us */
216 (void) sigaction(SIGCHLD, &sa, NULL);
217
218 if (arg_crash_shell) {
219 log_notice("Executing crash shell in 10s...");
220 (void) sleep(10);
221
222 pid = raw_clone(SIGCHLD, NULL);
223 if (pid < 0)
224 log_emergency_errno(errno, "Failed to fork off crash shell: %m");
225 else if (pid == 0) {
226 (void) setsid();
227 (void) make_console_stdio();
228 (void) execle("/bin/sh", "/bin/sh", NULL, environ);
229
230 log_emergency_errno(errno, "execle() failed: %m");
231 _exit(EXIT_FAILURE);
232 } else {
233 log_info("Spawned crash shell as PID "PID_FMT".", pid);
234 (void) wait_for_terminate(pid, NULL);
235 }
236 }
237
238 freeze_or_reboot();
239 }
240
241 static void install_crash_handler(void) {
242 static const struct sigaction sa = {
243 .sa_handler = crash,
244 .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
245 };
246 int r;
247
248 /* We ignore the return value here, since, we don't mind if we
249 * cannot set up a crash handler */
250 r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
251 if (r < 0)
252 log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
253 }
254
255 static int console_setup(void) {
256 _cleanup_close_ int tty_fd = -1;
257 int r;
258
259 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
260 if (tty_fd < 0)
261 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
262
263 /* We don't want to force text mode. plymouth may be showing
264 * pictures already from initrd. */
265 r = reset_terminal_fd(tty_fd, false);
266 if (r < 0)
267 return log_error_errno(r, "Failed to reset /dev/console: %m");
268
269 return 0;
270 }
271
272 static int parse_crash_chvt(const char *value) {
273 int b;
274
275 if (safe_atoi(value, &arg_crash_chvt) >= 0)
276 return 0;
277
278 b = parse_boolean(value);
279 if (b < 0)
280 return b;
281
282 if (b > 0)
283 arg_crash_chvt = 0; /* switch to where kmsg goes */
284 else
285 arg_crash_chvt = -1; /* turn off switching */
286
287 return 0;
288 }
289
290 static int set_machine_id(const char *m) {
291
292 if (sd_id128_from_string(m, &arg_machine_id) < 0)
293 return -EINVAL;
294
295 if (sd_id128_is_null(arg_machine_id))
296 return -EINVAL;
297
298 return 0;
299 }
300
301 static int parse_proc_cmdline_item(const char *key, const char *value) {
302
303 int r;
304
305 assert(key);
306
307 if (streq(key, "systemd.unit") && value) {
308
309 if (!in_initrd())
310 return free_and_strdup(&arg_default_unit, value);
311
312 } else if (streq(key, "rd.systemd.unit") && value) {
313
314 if (in_initrd())
315 return free_and_strdup(&arg_default_unit, value);
316
317 } else if (streq(key, "systemd.dump_core") && value) {
318
319 r = parse_boolean(value);
320 if (r < 0)
321 log_warning("Failed to parse dump core switch %s. Ignoring.", value);
322 else
323 arg_dump_core = r;
324
325 } else if (streq(key, "systemd.crash_chvt") && value) {
326
327 if (parse_crash_chvt(value) < 0)
328 log_warning("Failed to parse crash chvt switch %s. Ignoring.", value);
329
330 } else if (streq(key, "systemd.crash_shell") && value) {
331
332 r = parse_boolean(value);
333 if (r < 0)
334 log_warning("Failed to parse crash shell switch %s. Ignoring.", value);
335 else
336 arg_crash_shell = r;
337
338 } else if (streq(key, "systemd.crash_reboot") && value) {
339
340 r = parse_boolean(value);
341 if (r < 0)
342 log_warning("Failed to parse crash reboot switch %s. Ignoring.", value);
343 else
344 arg_crash_reboot = r;
345
346 } else if (streq(key, "systemd.confirm_spawn") && value) {
347
348 r = parse_boolean(value);
349 if (r < 0)
350 log_warning("Failed to parse confirm spawn switch %s. Ignoring.", value);
351 else
352 arg_confirm_spawn = r;
353
354 } else if (streq(key, "systemd.show_status") && value) {
355
356 r = parse_show_status(value, &arg_show_status);
357 if (r < 0)
358 log_warning("Failed to parse show status switch %s. Ignoring.", value);
359
360 } else if (streq(key, "systemd.default_standard_output") && value) {
361
362 r = exec_output_from_string(value);
363 if (r < 0)
364 log_warning("Failed to parse default standard output switch %s. Ignoring.", value);
365 else
366 arg_default_std_output = r;
367
368 } else if (streq(key, "systemd.default_standard_error") && value) {
369
370 r = exec_output_from_string(value);
371 if (r < 0)
372 log_warning("Failed to parse default standard error switch %s. Ignoring.", value);
373 else
374 arg_default_std_error = r;
375
376 } else if (streq(key, "systemd.setenv") && value) {
377
378 if (env_assignment_is_valid(value)) {
379 char **env;
380
381 env = strv_env_set(arg_default_environment, value);
382 if (env)
383 arg_default_environment = env;
384 else
385 log_warning_errno(ENOMEM, "Setting environment variable '%s' failed, ignoring: %m", value);
386 } else
387 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
388
389 } else if (streq(key, "systemd.machine_id") && value) {
390
391 r = set_machine_id(value);
392 if (r < 0)
393 log_warning("MachineID '%s' is not valid. Ignoring.", value);
394
395 } else if (streq(key, "quiet") && !value) {
396
397 if (arg_show_status == _SHOW_STATUS_UNSET)
398 arg_show_status = SHOW_STATUS_AUTO;
399
400 } else if (streq(key, "debug") && !value) {
401
402 /* Note that log_parse_environment() handles 'debug'
403 * too, and sets the log level to LOG_DEBUG. */
404
405 if (detect_container() > 0)
406 log_set_target(LOG_TARGET_CONSOLE);
407
408 } else if (!in_initrd() && !value) {
409 const char *target;
410
411 /* SysV compatibility */
412 target = runlevel_to_target(key);
413 if (target)
414 return free_and_strdup(&arg_default_unit, target);
415 }
416
417 return 0;
418 }
419
420 #define DEFINE_SETTER(name, func, descr) \
421 static int name(const char *unit, \
422 const char *filename, \
423 unsigned line, \
424 const char *section, \
425 unsigned section_line, \
426 const char *lvalue, \
427 int ltype, \
428 const char *rvalue, \
429 void *data, \
430 void *userdata) { \
431 \
432 int r; \
433 \
434 assert(filename); \
435 assert(lvalue); \
436 assert(rvalue); \
437 \
438 r = func(rvalue); \
439 if (r < 0) \
440 log_syntax(unit, LOG_ERR, filename, line, r, \
441 "Invalid " descr "'%s': %m", \
442 rvalue); \
443 \
444 return 0; \
445 }
446
447 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level")
448 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target")
449 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" )
450 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location")
451
452 static int config_parse_cpu_affinity2(
453 const char *unit,
454 const char *filename,
455 unsigned line,
456 const char *section,
457 unsigned section_line,
458 const char *lvalue,
459 int ltype,
460 const char *rvalue,
461 void *data,
462 void *userdata) {
463
464 _cleanup_cpu_free_ cpu_set_t *c = NULL;
465 int ncpus;
466
467 ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue);
468 if (ncpus < 0)
469 return ncpus;
470
471 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
472 log_warning("Failed to set CPU affinity: %m");
473
474 return 0;
475 }
476
477 static int config_parse_show_status(
478 const char* unit,
479 const char *filename,
480 unsigned line,
481 const char *section,
482 unsigned section_line,
483 const char *lvalue,
484 int ltype,
485 const char *rvalue,
486 void *data,
487 void *userdata) {
488
489 int k;
490 ShowStatus *b = data;
491
492 assert(filename);
493 assert(lvalue);
494 assert(rvalue);
495 assert(data);
496
497 k = parse_show_status(rvalue, b);
498 if (k < 0) {
499 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
500 return 0;
501 }
502
503 return 0;
504 }
505
506 static int config_parse_crash_chvt(
507 const char* unit,
508 const char *filename,
509 unsigned line,
510 const char *section,
511 unsigned section_line,
512 const char *lvalue,
513 int ltype,
514 const char *rvalue,
515 void *data,
516 void *userdata) {
517
518 int r;
519
520 assert(filename);
521 assert(lvalue);
522 assert(rvalue);
523
524 r = parse_crash_chvt(rvalue);
525 if (r < 0) {
526 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
527 return 0;
528 }
529
530 return 0;
531 }
532
533 static int config_parse_join_controllers(const char *unit,
534 const char *filename,
535 unsigned line,
536 const char *section,
537 unsigned section_line,
538 const char *lvalue,
539 int ltype,
540 const char *rvalue,
541 void *data,
542 void *userdata) {
543
544 const char *whole_rvalue = rvalue;
545 unsigned n = 0;
546
547 assert(filename);
548 assert(lvalue);
549 assert(rvalue);
550
551 arg_join_controllers = strv_free_free(arg_join_controllers);
552
553 for (;;) {
554 _cleanup_free_ char *word = NULL;
555 char **l;
556 int r;
557
558 r = extract_first_word(&rvalue, &word, WHITESPACE, EXTRACT_QUOTES);
559 if (r < 0) {
560 log_syntax(unit, LOG_ERR, filename, line, r, "Invalid value for %s: %s", lvalue, whole_rvalue);
561 return r;
562 }
563 if (r == 0)
564 break;
565
566 l = strv_split(word, ",");
567 if (!l)
568 return log_oom();
569 strv_uniq(l);
570
571 if (strv_length(l) <= 1) {
572 strv_free(l);
573 continue;
574 }
575
576 if (!arg_join_controllers) {
577 arg_join_controllers = new(char**, 2);
578 if (!arg_join_controllers) {
579 strv_free(l);
580 return log_oom();
581 }
582
583 arg_join_controllers[0] = l;
584 arg_join_controllers[1] = NULL;
585
586 n = 1;
587 } else {
588 char ***a;
589 char ***t;
590
591 t = new0(char**, n+2);
592 if (!t) {
593 strv_free(l);
594 return log_oom();
595 }
596
597 n = 0;
598
599 for (a = arg_join_controllers; *a; a++) {
600
601 if (strv_overlap(*a, l)) {
602 if (strv_extend_strv(&l, *a, false) < 0) {
603 strv_free(l);
604 strv_free_free(t);
605 return log_oom();
606 }
607
608 } else {
609 char **c;
610
611 c = strv_copy(*a);
612 if (!c) {
613 strv_free(l);
614 strv_free_free(t);
615 return log_oom();
616 }
617
618 t[n++] = c;
619 }
620 }
621
622 t[n++] = strv_uniq(l);
623
624 strv_free_free(arg_join_controllers);
625 arg_join_controllers = t;
626 }
627 }
628 if (!isempty(rvalue))
629 log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring.");
630
631 return 0;
632 }
633
634 static int parse_config_file(void) {
635
636 const ConfigTableItem items[] = {
637 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
638 { "Manager", "LogTarget", config_parse_target, 0, NULL },
639 { "Manager", "LogColor", config_parse_color, 0, NULL },
640 { "Manager", "LogLocation", config_parse_location, 0, NULL },
641 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
642 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL },
643 { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL },
644 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
645 { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
646 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
647 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
648 { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
649 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
650 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
651 { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
652 #ifdef HAVE_SECCOMP
653 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
654 #endif
655 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
656 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
657 { "Manager", "DefaultStandardOutput", config_parse_output, 0, &arg_default_std_output },
658 { "Manager", "DefaultStandardError", config_parse_output, 0, &arg_default_std_error },
659 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
660 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
661 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
662 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval },
663 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
664 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
665 { "Manager", "DefaultLimitCPU", config_parse_limit, RLIMIT_CPU, arg_default_rlimit },
666 { "Manager", "DefaultLimitFSIZE", config_parse_limit, RLIMIT_FSIZE, arg_default_rlimit },
667 { "Manager", "DefaultLimitDATA", config_parse_limit, RLIMIT_DATA, arg_default_rlimit },
668 { "Manager", "DefaultLimitSTACK", config_parse_limit, RLIMIT_STACK, arg_default_rlimit },
669 { "Manager", "DefaultLimitCORE", config_parse_limit, RLIMIT_CORE, arg_default_rlimit },
670 { "Manager", "DefaultLimitRSS", config_parse_limit, RLIMIT_RSS, arg_default_rlimit },
671 { "Manager", "DefaultLimitNOFILE", config_parse_limit, RLIMIT_NOFILE, arg_default_rlimit },
672 { "Manager", "DefaultLimitAS", config_parse_limit, RLIMIT_AS, arg_default_rlimit },
673 { "Manager", "DefaultLimitNPROC", config_parse_limit, RLIMIT_NPROC, arg_default_rlimit },
674 { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, RLIMIT_MEMLOCK, arg_default_rlimit },
675 { "Manager", "DefaultLimitLOCKS", config_parse_limit, RLIMIT_LOCKS, arg_default_rlimit },
676 { "Manager", "DefaultLimitSIGPENDING", config_parse_limit, RLIMIT_SIGPENDING, arg_default_rlimit },
677 { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, RLIMIT_MSGQUEUE, arg_default_rlimit },
678 { "Manager", "DefaultLimitNICE", config_parse_limit, RLIMIT_NICE, arg_default_rlimit },
679 { "Manager", "DefaultLimitRTPRIO", config_parse_limit, RLIMIT_RTPRIO, arg_default_rlimit },
680 { "Manager", "DefaultLimitRTTIME", config_parse_limit, RLIMIT_RTTIME, arg_default_rlimit },
681 { "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_default_cpu_accounting },
682 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
683 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
684 { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
685 { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max },
686 {}
687 };
688
689 const char *fn, *conf_dirs_nulstr;
690
691 fn = arg_running_as == MANAGER_SYSTEM ?
692 PKGSYSCONFDIR "/system.conf" :
693 PKGSYSCONFDIR "/user.conf";
694
695 conf_dirs_nulstr = arg_running_as == MANAGER_SYSTEM ?
696 CONF_PATHS_NULSTR("systemd/system.conf.d") :
697 CONF_PATHS_NULSTR("systemd/user.conf.d");
698
699 config_parse_many(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, false, NULL);
700
701 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
702 * like everywhere else. */
703 if (arg_default_timeout_start_usec <= 0)
704 arg_default_timeout_start_usec = USEC_INFINITY;
705 if (arg_default_timeout_stop_usec <= 0)
706 arg_default_timeout_stop_usec = USEC_INFINITY;
707
708 return 0;
709 }
710
711 static void manager_set_defaults(Manager *m) {
712
713 assert(m);
714
715 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
716 m->default_std_output = arg_default_std_output;
717 m->default_std_error = arg_default_std_error;
718 m->default_timeout_start_usec = arg_default_timeout_start_usec;
719 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
720 m->default_restart_usec = arg_default_restart_usec;
721 m->default_start_limit_interval = arg_default_start_limit_interval;
722 m->default_start_limit_burst = arg_default_start_limit_burst;
723 m->default_cpu_accounting = arg_default_cpu_accounting;
724 m->default_blockio_accounting = arg_default_blockio_accounting;
725 m->default_memory_accounting = arg_default_memory_accounting;
726 m->default_tasks_accounting = arg_default_tasks_accounting;
727 m->default_tasks_max = arg_default_tasks_max;
728
729 manager_set_default_rlimits(m, arg_default_rlimit);
730 manager_environment_add(m, NULL, arg_default_environment);
731 }
732
733 static int parse_argv(int argc, char *argv[]) {
734
735 enum {
736 ARG_LOG_LEVEL = 0x100,
737 ARG_LOG_TARGET,
738 ARG_LOG_COLOR,
739 ARG_LOG_LOCATION,
740 ARG_UNIT,
741 ARG_SYSTEM,
742 ARG_USER,
743 ARG_TEST,
744 ARG_NO_PAGER,
745 ARG_VERSION,
746 ARG_DUMP_CONFIGURATION_ITEMS,
747 ARG_DUMP_CORE,
748 ARG_CRASH_CHVT,
749 ARG_CRASH_SHELL,
750 ARG_CRASH_REBOOT,
751 ARG_CONFIRM_SPAWN,
752 ARG_SHOW_STATUS,
753 ARG_DESERIALIZE,
754 ARG_SWITCHED_ROOT,
755 ARG_DEFAULT_STD_OUTPUT,
756 ARG_DEFAULT_STD_ERROR,
757 ARG_MACHINE_ID
758 };
759
760 static const struct option options[] = {
761 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
762 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
763 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
764 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
765 { "unit", required_argument, NULL, ARG_UNIT },
766 { "system", no_argument, NULL, ARG_SYSTEM },
767 { "user", no_argument, NULL, ARG_USER },
768 { "test", no_argument, NULL, ARG_TEST },
769 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
770 { "help", no_argument, NULL, 'h' },
771 { "version", no_argument, NULL, ARG_VERSION },
772 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
773 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
774 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
775 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
776 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
777 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
778 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
779 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
780 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
781 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
782 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
783 { "machine-id", required_argument, NULL, ARG_MACHINE_ID },
784 {}
785 };
786
787 int c, r;
788
789 assert(argc >= 1);
790 assert(argv);
791
792 if (getpid() == 1)
793 opterr = 0;
794
795 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
796
797 switch (c) {
798
799 case ARG_LOG_LEVEL:
800 r = log_set_max_level_from_string(optarg);
801 if (r < 0) {
802 log_error("Failed to parse log level %s.", optarg);
803 return r;
804 }
805
806 break;
807
808 case ARG_LOG_TARGET:
809 r = log_set_target_from_string(optarg);
810 if (r < 0) {
811 log_error("Failed to parse log target %s.", optarg);
812 return r;
813 }
814
815 break;
816
817 case ARG_LOG_COLOR:
818
819 if (optarg) {
820 r = log_show_color_from_string(optarg);
821 if (r < 0) {
822 log_error("Failed to parse log color setting %s.", optarg);
823 return r;
824 }
825 } else
826 log_show_color(true);
827
828 break;
829
830 case ARG_LOG_LOCATION:
831 if (optarg) {
832 r = log_show_location_from_string(optarg);
833 if (r < 0) {
834 log_error("Failed to parse log location setting %s.", optarg);
835 return r;
836 }
837 } else
838 log_show_location(true);
839
840 break;
841
842 case ARG_DEFAULT_STD_OUTPUT:
843 r = exec_output_from_string(optarg);
844 if (r < 0) {
845 log_error("Failed to parse default standard output setting %s.", optarg);
846 return r;
847 } else
848 arg_default_std_output = r;
849 break;
850
851 case ARG_DEFAULT_STD_ERROR:
852 r = exec_output_from_string(optarg);
853 if (r < 0) {
854 log_error("Failed to parse default standard error output setting %s.", optarg);
855 return r;
856 } else
857 arg_default_std_error = r;
858 break;
859
860 case ARG_UNIT:
861
862 r = free_and_strdup(&arg_default_unit, optarg);
863 if (r < 0)
864 return log_error_errno(r, "Failed to set default unit %s: %m", optarg);
865
866 break;
867
868 case ARG_SYSTEM:
869 arg_running_as = MANAGER_SYSTEM;
870 break;
871
872 case ARG_USER:
873 arg_running_as = MANAGER_USER;
874 break;
875
876 case ARG_TEST:
877 arg_action = ACTION_TEST;
878 break;
879
880 case ARG_NO_PAGER:
881 arg_no_pager = true;
882 break;
883
884 case ARG_VERSION:
885 arg_action = ACTION_VERSION;
886 break;
887
888 case ARG_DUMP_CONFIGURATION_ITEMS:
889 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
890 break;
891
892 case ARG_DUMP_CORE:
893 if (!optarg)
894 arg_dump_core = true;
895 else {
896 r = parse_boolean(optarg);
897 if (r < 0)
898 return log_error_errno(r, "Failed to parse dump core boolean: %s", optarg);
899 arg_dump_core = r;
900 }
901 break;
902
903 case ARG_CRASH_CHVT:
904 r = parse_crash_chvt(optarg);
905 if (r < 0)
906 return log_error_errno(r, "Failed to parse crash virtual terminal index: %s", optarg);
907 break;
908
909 case ARG_CRASH_SHELL:
910 if (!optarg)
911 arg_crash_shell = true;
912 else {
913 r = parse_boolean(optarg);
914 if (r < 0)
915 return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg);
916 arg_crash_shell = r;
917 }
918 break;
919
920 case ARG_CRASH_REBOOT:
921 if (!optarg)
922 arg_crash_reboot = true;
923 else {
924 r = parse_boolean(optarg);
925 if (r < 0)
926 return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg);
927 arg_crash_reboot = r;
928 }
929 break;
930
931 case ARG_CONFIRM_SPAWN:
932 r = optarg ? parse_boolean(optarg) : 1;
933 if (r < 0) {
934 log_error("Failed to parse confirm spawn boolean %s.", optarg);
935 return r;
936 }
937 arg_confirm_spawn = r;
938 break;
939
940 case ARG_SHOW_STATUS:
941 if (optarg) {
942 r = parse_show_status(optarg, &arg_show_status);
943 if (r < 0) {
944 log_error("Failed to parse show status boolean %s.", optarg);
945 return r;
946 }
947 } else
948 arg_show_status = SHOW_STATUS_YES;
949 break;
950
951 case ARG_DESERIALIZE: {
952 int fd;
953 FILE *f;
954
955 r = safe_atoi(optarg, &fd);
956 if (r < 0 || fd < 0) {
957 log_error("Failed to parse deserialize option %s.", optarg);
958 return -EINVAL;
959 }
960
961 (void) fd_cloexec(fd, true);
962
963 f = fdopen(fd, "r");
964 if (!f)
965 return log_error_errno(errno, "Failed to open serialization fd: %m");
966
967 safe_fclose(arg_serialization);
968 arg_serialization = f;
969
970 break;
971 }
972
973 case ARG_SWITCHED_ROOT:
974 arg_switched_root = true;
975 break;
976
977 case ARG_MACHINE_ID:
978 r = set_machine_id(optarg);
979 if (r < 0) {
980 log_error("MachineID '%s' is not valid.", optarg);
981 return r;
982 }
983 break;
984
985 case 'h':
986 arg_action = ACTION_HELP;
987 break;
988
989 case 'D':
990 log_set_max_level(LOG_DEBUG);
991 break;
992
993 case 'b':
994 case 's':
995 case 'z':
996 /* Just to eat away the sysvinit kernel
997 * cmdline args without getopt() error
998 * messages that we'll parse in
999 * parse_proc_cmdline_word() or ignore. */
1000
1001 case '?':
1002 if (getpid() != 1)
1003 return -EINVAL;
1004 else
1005 return 0;
1006
1007 default:
1008 assert_not_reached("Unhandled option code.");
1009 }
1010
1011 if (optind < argc && getpid() != 1) {
1012 /* Hmm, when we aren't run as init system
1013 * let's complain about excess arguments */
1014
1015 log_error("Excess arguments.");
1016 return -EINVAL;
1017 }
1018
1019 return 0;
1020 }
1021
1022 static int help(void) {
1023
1024 printf("%s [OPTIONS...]\n\n"
1025 "Starts up and maintains the system or user services.\n\n"
1026 " -h --help Show this help\n"
1027 " --test Determine startup sequence, dump it and exit\n"
1028 " --no-pager Do not pipe output into a pager\n"
1029 " --dump-configuration-items Dump understood unit configuration items\n"
1030 " --unit=UNIT Set default unit\n"
1031 " --system Run a system instance, even if PID != 1\n"
1032 " --user Run a user instance\n"
1033 " --dump-core[=BOOL] Dump core on crash\n"
1034 " --crash-vt=NR Change to specified VT on crash\n"
1035 " --crash-reboot[=BOOL] Reboot on crash\n"
1036 " --crash-shell[=BOOL] Run shell on crash\n"
1037 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1038 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1039 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1040 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1041 " --log-color[=BOOL] Highlight important log messages\n"
1042 " --log-location[=BOOL] Include code location in log messages\n"
1043 " --default-standard-output= Set default standard output for services\n"
1044 " --default-standard-error= Set default standard error output for services\n",
1045 program_invocation_short_name);
1046
1047 return 0;
1048 }
1049
1050 static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
1051 _cleanup_fdset_free_ FDSet *fds = NULL;
1052 _cleanup_fclose_ FILE *f = NULL;
1053 int r;
1054
1055 assert(m);
1056 assert(_f);
1057 assert(_fds);
1058
1059 r = manager_open_serialization(m, &f);
1060 if (r < 0)
1061 return log_error_errno(r, "Failed to create serialization file: %m");
1062
1063 /* Make sure nothing is really destructed when we shut down */
1064 m->n_reloading++;
1065 bus_manager_send_reloading(m, true);
1066
1067 fds = fdset_new();
1068 if (!fds)
1069 return log_oom();
1070
1071 r = manager_serialize(m, f, fds, switching_root);
1072 if (r < 0)
1073 return log_error_errno(r, "Failed to serialize state: %m");
1074
1075 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1076 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1077
1078 r = fd_cloexec(fileno(f), false);
1079 if (r < 0)
1080 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1081
1082 r = fdset_cloexec(fds, false);
1083 if (r < 0)
1084 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1085
1086 *_f = f;
1087 *_fds = fds;
1088
1089 f = NULL;
1090 fds = NULL;
1091
1092 return 0;
1093 }
1094
1095 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1096 struct rlimit nl;
1097 int r;
1098
1099 assert(saved_rlimit);
1100
1101 /* Save the original RLIMIT_NOFILE so that we can reset it
1102 * later when transitioning from the initrd to the main
1103 * systemd or suchlike. */
1104 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1105 return log_error_errno(errno, "Reading RLIMIT_NOFILE failed: %m");
1106
1107 /* Make sure forked processes get the default kernel setting */
1108 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1109 struct rlimit *rl;
1110
1111 rl = newdup(struct rlimit, saved_rlimit, 1);
1112 if (!rl)
1113 return log_oom();
1114
1115 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1116 }
1117
1118 /* Bump up the resource limit for ourselves substantially */
1119 nl.rlim_cur = nl.rlim_max = 64*1024;
1120 r = setrlimit_closest(RLIMIT_NOFILE, &nl);
1121 if (r < 0)
1122 return log_error_errno(r, "Setting RLIMIT_NOFILE failed: %m");
1123
1124 return 0;
1125 }
1126
1127 static void test_usr(void) {
1128
1129 /* Check that /usr is not a separate fs */
1130
1131 if (dir_is_empty("/usr") <= 0)
1132 return;
1133
1134 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1135 "Some things will probably break (sometimes even silently) in mysterious ways. "
1136 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1137 }
1138
1139 static int initialize_join_controllers(void) {
1140 /* By default, mount "cpu" + "cpuacct" together, and "net_cls"
1141 * + "net_prio". We'd like to add "cpuset" to the mix, but
1142 * "cpuset" doesn't really work for groups with no initialized
1143 * attributes. */
1144
1145 arg_join_controllers = new(char**, 3);
1146 if (!arg_join_controllers)
1147 return -ENOMEM;
1148
1149 arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL);
1150 if (!arg_join_controllers[0])
1151 goto oom;
1152
1153 arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL);
1154 if (!arg_join_controllers[1])
1155 goto oom;
1156
1157 arg_join_controllers[2] = NULL;
1158 return 0;
1159
1160 oom:
1161 arg_join_controllers = strv_free_free(arg_join_controllers);
1162 return -ENOMEM;
1163 }
1164
1165 static int enforce_syscall_archs(Set *archs) {
1166 #ifdef HAVE_SECCOMP
1167 scmp_filter_ctx *seccomp;
1168 Iterator i;
1169 void *id;
1170 int r;
1171
1172 seccomp = seccomp_init(SCMP_ACT_ALLOW);
1173 if (!seccomp)
1174 return log_oom();
1175
1176 SET_FOREACH(id, arg_syscall_archs, i) {
1177 r = seccomp_arch_add(seccomp, PTR_TO_UINT32(id) - 1);
1178 if (r == -EEXIST)
1179 continue;
1180 if (r < 0) {
1181 log_error_errno(r, "Failed to add architecture to seccomp: %m");
1182 goto finish;
1183 }
1184 }
1185
1186 r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
1187 if (r < 0) {
1188 log_error_errno(r, "Failed to unset NO_NEW_PRIVS: %m");
1189 goto finish;
1190 }
1191
1192 r = seccomp_load(seccomp);
1193 if (r < 0)
1194 log_error_errno(r, "Failed to add install architecture seccomp: %m");
1195
1196 finish:
1197 seccomp_release(seccomp);
1198 return r;
1199 #else
1200 return 0;
1201 #endif
1202 }
1203
1204 static int status_welcome(void) {
1205 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1206 int r;
1207
1208 r = parse_env_file("/etc/os-release", NEWLINE,
1209 "PRETTY_NAME", &pretty_name,
1210 "ANSI_COLOR", &ansi_color,
1211 NULL);
1212 if (r == -ENOENT)
1213 r = parse_env_file("/usr/lib/os-release", NEWLINE,
1214 "PRETTY_NAME", &pretty_name,
1215 "ANSI_COLOR", &ansi_color,
1216 NULL);
1217
1218 if (r < 0 && r != -ENOENT)
1219 log_warning_errno(r, "Failed to read os-release file: %m");
1220
1221 return status_printf(NULL, false, false,
1222 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1223 isempty(ansi_color) ? "1" : ansi_color,
1224 isempty(pretty_name) ? "Linux" : pretty_name);
1225 }
1226
1227 static int write_container_id(void) {
1228 const char *c;
1229 int r;
1230
1231 c = getenv("container");
1232 if (isempty(c))
1233 return 0;
1234
1235 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1236 if (r < 0)
1237 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
1238
1239 return 1;
1240 }
1241
1242 static int bump_unix_max_dgram_qlen(void) {
1243 _cleanup_free_ char *qlen = NULL;
1244 unsigned long v;
1245 int r;
1246
1247 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel
1248 * default of 16 is simply too low. We set the value really
1249 * really early during boot, so that it is actually applied to
1250 * all our sockets, including the $NOTIFY_SOCKET one. */
1251
1252 r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1253 if (r < 0)
1254 return log_warning_errno(r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1255
1256 r = safe_atolu(qlen, &v);
1257 if (r < 0)
1258 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length, ignoring: %m");
1259
1260 if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1261 return 0;
1262
1263 qlen = mfree(qlen);
1264 if (asprintf(&qlen, "%lu\n", DEFAULT_UNIX_MAX_DGRAM_QLEN) < 0)
1265 return log_oom();
1266
1267 r = write_string_file("/proc/sys/net/unix/max_dgram_qlen", qlen, 0);
1268 if (r < 0)
1269 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1270 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1271
1272 return 1;
1273 }
1274
1275 int main(int argc, char *argv[]) {
1276 Manager *m = NULL;
1277 int r, retval = EXIT_FAILURE;
1278 usec_t before_startup, after_startup;
1279 char timespan[FORMAT_TIMESPAN_MAX];
1280 FDSet *fds = NULL;
1281 bool reexecute = false;
1282 const char *shutdown_verb = NULL;
1283 dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL;
1284 dual_timestamp userspace_timestamp = DUAL_TIMESTAMP_NULL;
1285 dual_timestamp kernel_timestamp = DUAL_TIMESTAMP_NULL;
1286 dual_timestamp security_start_timestamp = DUAL_TIMESTAMP_NULL;
1287 dual_timestamp security_finish_timestamp = DUAL_TIMESTAMP_NULL;
1288 static char systemd[] = "systemd";
1289 bool skip_setup = false;
1290 unsigned j;
1291 bool loaded_policy = false;
1292 bool arm_reboot_watchdog = false;
1293 bool queue_default_job = false;
1294 bool empty_etc = false;
1295 char *switch_root_dir = NULL, *switch_root_init = NULL;
1296 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0);
1297 const char *error_message = NULL;
1298
1299 #ifdef HAVE_SYSV_COMPAT
1300 if (getpid() != 1 && strstr(program_invocation_short_name, "init")) {
1301 /* This is compatibility support for SysV, where
1302 * calling init as a user is identical to telinit. */
1303
1304 execv(SYSTEMCTL_BINARY_PATH, argv);
1305 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1306 return 1;
1307 }
1308 #endif
1309
1310 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
1311 dual_timestamp_get(&userspace_timestamp);
1312
1313 /* Determine if this is a reexecution or normal bootup. We do
1314 * the full command line parsing much later, so let's just
1315 * have a quick peek here. */
1316 if (strv_find(argv+1, "--deserialize"))
1317 skip_setup = true;
1318
1319 /* If we have switched root, do all the special setup
1320 * things */
1321 if (strv_find(argv+1, "--switched-root"))
1322 skip_setup = false;
1323
1324 /* If we get started via the /sbin/init symlink then we are
1325 called 'init'. After a subsequent reexecution we are then
1326 called 'systemd'. That is confusing, hence let's call us
1327 systemd right-away. */
1328 program_invocation_short_name = systemd;
1329 prctl(PR_SET_NAME, systemd);
1330
1331 saved_argv = argv;
1332 saved_argc = argc;
1333
1334 log_show_color(isatty(STDERR_FILENO) > 0);
1335 log_set_upgrade_syslog_to_journal(true);
1336
1337 /* Disable the umask logic */
1338 if (getpid() == 1)
1339 umask(0);
1340
1341 if (getpid() == 1 && detect_container() <= 0) {
1342
1343 /* Running outside of a container as PID 1 */
1344 arg_running_as = MANAGER_SYSTEM;
1345 make_null_stdio();
1346 log_set_target(LOG_TARGET_KMSG);
1347 log_open();
1348
1349 if (in_initrd())
1350 initrd_timestamp = userspace_timestamp;
1351
1352 if (!skip_setup) {
1353 r = mount_setup_early();
1354 if (r < 0) {
1355 error_message = "Failed to early mount API filesystems";
1356 goto finish;
1357 }
1358 dual_timestamp_get(&security_start_timestamp);
1359 if (mac_selinux_setup(&loaded_policy) < 0) {
1360 error_message = "Failed to load SELinux policy";
1361 goto finish;
1362 } else if (ima_setup() < 0) {
1363 error_message = "Failed to load IMA policy";
1364 goto finish;
1365 } else if (mac_smack_setup(&loaded_policy) < 0) {
1366 error_message = "Failed to load SMACK policy";
1367 goto finish;
1368 }
1369 dual_timestamp_get(&security_finish_timestamp);
1370 }
1371
1372 if (mac_selinux_init() < 0) {
1373 error_message = "Failed to initialize SELinux policy";
1374 goto finish;
1375 }
1376
1377 if (!skip_setup) {
1378 if (clock_is_localtime(NULL) > 0) {
1379 int min;
1380
1381 /*
1382 * The very first call of settimeofday() also does a time warp in the kernel.
1383 *
1384 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on
1385 * external tools to take care of maintaining the RTC and do all adjustments.
1386 * This matches the behavior of Windows, which leaves the RTC alone if the
1387 * registry tells that the RTC runs in UTC.
1388 */
1389 r = clock_set_timezone(&min);
1390 if (r < 0)
1391 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1392 else
1393 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1394 } else if (!in_initrd()) {
1395 /*
1396 * Do a dummy very first call to seal the kernel's time warp magic.
1397 *
1398 * Do not call this this from inside the initrd. The initrd might not
1399 * carry /etc/adjtime with LOCAL, but the real system could be set up
1400 * that way. In such case, we need to delay the time-warp or the sealing
1401 * until we reach the real system.
1402 *
1403 * Do no set the kernel's timezone. The concept of local time cannot
1404 * be supported reliably, the time will jump or be incorrect at every daylight
1405 * saving time change. All kernel local time concepts will be treated
1406 * as UTC that way.
1407 */
1408 (void) clock_reset_timewarp();
1409 }
1410
1411 r = clock_apply_epoch();
1412 if (r < 0)
1413 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1414 else if (r > 0)
1415 log_info("System time before build time, advancing clock.");
1416 }
1417
1418 /* Set the default for later on, but don't actually
1419 * open the logs like this for now. Note that if we
1420 * are transitioning from the initrd there might still
1421 * be journal fd open, and we shouldn't attempt
1422 * opening that before we parsed /proc/cmdline which
1423 * might redirect output elsewhere. */
1424 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
1425
1426 } else if (getpid() == 1) {
1427 /* Running inside a container, as PID 1 */
1428 arg_running_as = MANAGER_SYSTEM;
1429 log_set_target(LOG_TARGET_CONSOLE);
1430 log_close_console(); /* force reopen of /dev/console */
1431 log_open();
1432
1433 /* For the later on, see above... */
1434 log_set_target(LOG_TARGET_JOURNAL);
1435
1436 /* clear the kernel timestamp,
1437 * because we are in a container */
1438 kernel_timestamp = DUAL_TIMESTAMP_NULL;
1439 } else {
1440 /* Running as user instance */
1441 arg_running_as = MANAGER_USER;
1442 log_set_target(LOG_TARGET_AUTO);
1443 log_open();
1444
1445 /* clear the kernel timestamp,
1446 * because we are not PID 1 */
1447 kernel_timestamp = DUAL_TIMESTAMP_NULL;
1448 }
1449
1450 if (getpid() == 1) {
1451 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1452 * will process core dumps for system services by default. */
1453 (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
1454
1455 /* But at the same time, turn off the core_pattern logic by default, so that no coredumps are stored
1456 * until the systemd-coredump tool is enabled via sysctl. */
1457 if (!skip_setup)
1458 (void) write_string_file("/proc/sys/kernel/core_pattern", "|/bin/false", 0);
1459 }
1460
1461 /* Initialize default unit */
1462 r = free_and_strdup(&arg_default_unit, SPECIAL_DEFAULT_TARGET);
1463 if (r < 0) {
1464 log_emergency_errno(r, "Failed to set default unit %s: %m", SPECIAL_DEFAULT_TARGET);
1465 error_message = "Failed to set default unit";
1466 goto finish;
1467 }
1468
1469 r = initialize_join_controllers();
1470 if (r < 0) {
1471 error_message = "Failed to initialize cgroup controllers";
1472 goto finish;
1473 }
1474
1475 /* Mount /proc, /sys and friends, so that /proc/cmdline and
1476 * /proc/$PID/fd is available. */
1477 if (getpid() == 1) {
1478
1479 /* Load the kernel modules early, so that we kdbus.ko is loaded before kdbusfs shall be mounted */
1480 if (!skip_setup)
1481 kmod_setup();
1482
1483 r = mount_setup(loaded_policy);
1484 if (r < 0) {
1485 error_message = "Failed to mount API filesystems";
1486 goto finish;
1487 }
1488 }
1489
1490 /* Reset all signal handlers. */
1491 (void) reset_all_signal_handlers();
1492 (void) ignore_signals(SIGNALS_IGNORE, -1);
1493
1494 if (parse_config_file() < 0) {
1495 error_message = "Failed to parse config file";
1496 goto finish;
1497 }
1498
1499 if (arg_running_as == MANAGER_SYSTEM) {
1500 r = parse_proc_cmdline(parse_proc_cmdline_item);
1501 if (r < 0)
1502 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
1503 }
1504
1505 /* Note that this also parses bits from the kernel command
1506 * line, including "debug". */
1507 log_parse_environment();
1508
1509 if (parse_argv(argc, argv) < 0) {
1510 error_message = "Failed to parse commandline arguments";
1511 goto finish;
1512 }
1513
1514 if (arg_action == ACTION_TEST &&
1515 geteuid() == 0) {
1516 log_error("Don't run test mode as root.");
1517 goto finish;
1518 }
1519
1520 if (arg_running_as == MANAGER_USER &&
1521 arg_action == ACTION_RUN &&
1522 sd_booted() <= 0) {
1523 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
1524 goto finish;
1525 }
1526
1527 if (arg_running_as == MANAGER_SYSTEM &&
1528 arg_action == ACTION_RUN &&
1529 running_in_chroot() > 0) {
1530 log_error("Cannot be run in a chroot() environment.");
1531 goto finish;
1532 }
1533
1534 if (arg_action == ACTION_TEST)
1535 skip_setup = true;
1536
1537 if (arg_action == ACTION_TEST || arg_action == ACTION_HELP)
1538 pager_open(arg_no_pager, false);
1539
1540 if (arg_action == ACTION_HELP) {
1541 retval = help();
1542 goto finish;
1543 } else if (arg_action == ACTION_VERSION) {
1544 retval = version();
1545 goto finish;
1546 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
1547 unit_dump_config_items(stdout);
1548 retval = EXIT_SUCCESS;
1549 goto finish;
1550 } else if (arg_action == ACTION_DONE) {
1551 retval = EXIT_SUCCESS;
1552 goto finish;
1553 }
1554
1555 if (arg_running_as == MANAGER_USER &&
1556 !getenv("XDG_RUNTIME_DIR")) {
1557 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
1558 goto finish;
1559 }
1560
1561 assert_se(arg_action == ACTION_RUN || arg_action == ACTION_TEST);
1562
1563 /* Close logging fds, in order not to confuse fdset below */
1564 log_close();
1565
1566 /* Remember open file descriptors for later deserialization */
1567 r = fdset_new_fill(&fds);
1568 if (r < 0) {
1569 log_emergency_errno(r, "Failed to allocate fd set: %m");
1570 error_message = "Failed to allocate fd set";
1571 goto finish;
1572 } else
1573 fdset_cloexec(fds, true);
1574
1575 if (arg_serialization)
1576 assert_se(fdset_remove(fds, fileno(arg_serialization)) >= 0);
1577
1578 if (arg_running_as == MANAGER_SYSTEM)
1579 /* Become a session leader if we aren't one yet. */
1580 setsid();
1581
1582 /* Move out of the way, so that we won't block unmounts */
1583 assert_se(chdir("/") == 0);
1584
1585 /* Reset the console, but only if this is really init and we
1586 * are freshly booted */
1587 if (arg_running_as == MANAGER_SYSTEM && arg_action == ACTION_RUN) {
1588
1589 /* If we are init, we connect stdin/stdout/stderr to
1590 * /dev/null and make sure we don't have a controlling
1591 * tty. */
1592 release_terminal();
1593
1594 if (getpid() == 1 && !skip_setup)
1595 console_setup();
1596 }
1597
1598 /* Open the logging devices, if possible and necessary */
1599 log_open();
1600
1601 if (arg_show_status == _SHOW_STATUS_UNSET)
1602 arg_show_status = SHOW_STATUS_YES;
1603
1604 /* Make sure we leave a core dump without panicing the
1605 * kernel. */
1606 if (getpid() == 1) {
1607 install_crash_handler();
1608
1609 r = mount_cgroup_controllers(arg_join_controllers);
1610 if (r < 0)
1611 goto finish;
1612 }
1613
1614 if (arg_running_as == MANAGER_SYSTEM) {
1615 int v;
1616
1617 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1618 arg_action == ACTION_TEST ? "test " : "" );
1619
1620 v = detect_virtualization();
1621 if (v > 0)
1622 log_info("Detected virtualization %s.", virtualization_to_string(v));
1623
1624 write_container_id();
1625
1626 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1627
1628 if (in_initrd())
1629 log_info("Running in initial RAM disk.");
1630
1631 /* Let's check whether /etc is already populated. We
1632 * don't actually really check for that, but use
1633 * /etc/machine-id as flag file. This allows container
1634 * managers and installers to provision a couple of
1635 * files already. If the container manager wants to
1636 * provision the machine ID itself it should pass
1637 * $container_uuid to PID 1. */
1638
1639 empty_etc = access("/etc/machine-id", F_OK) < 0;
1640 if (empty_etc)
1641 log_info("Running with unpopulated /etc.");
1642 } else {
1643 _cleanup_free_ char *t;
1644
1645 t = uid_to_name(getuid());
1646 log_debug(PACKAGE_STRING " running in %suser mode for user "UID_FMT"/%s. (" SYSTEMD_FEATURES ")",
1647 arg_action == ACTION_TEST ? " test" : "", getuid(), t);
1648 }
1649
1650 if (arg_running_as == MANAGER_SYSTEM && !skip_setup) {
1651 if (arg_show_status > 0)
1652 status_welcome();
1653
1654 hostname_setup();
1655 machine_id_setup(NULL, arg_machine_id);
1656 loopback_setup();
1657 bump_unix_max_dgram_qlen();
1658
1659 test_usr();
1660 }
1661
1662 if (arg_running_as == MANAGER_SYSTEM && arg_runtime_watchdog > 0)
1663 watchdog_set_timeout(&arg_runtime_watchdog);
1664
1665 if (arg_timer_slack_nsec != NSEC_INFINITY)
1666 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1667 log_error_errno(errno, "Failed to adjust timer slack: %m");
1668
1669 if (!cap_test_all(arg_capability_bounding_set)) {
1670 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
1671 if (r < 0) {
1672 log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
1673 error_message = "Failed to drop capability bounding set of usermode helpers";
1674 goto finish;
1675 }
1676 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
1677 if (r < 0) {
1678 log_emergency_errno(r, "Failed to drop capability bounding set: %m");
1679 error_message = "Failed to drop capability bounding set";
1680 goto finish;
1681 }
1682 }
1683
1684 if (arg_syscall_archs) {
1685 r = enforce_syscall_archs(arg_syscall_archs);
1686 if (r < 0) {
1687 error_message = "Failed to set syscall architectures";
1688 goto finish;
1689 }
1690 }
1691
1692 if (arg_running_as == MANAGER_USER)
1693 /* Become reaper of our children */
1694 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
1695 log_warning_errno(errno, "Failed to make us a subreaper: %m");
1696
1697 if (arg_running_as == MANAGER_SYSTEM) {
1698 bump_rlimit_nofile(&saved_rlimit_nofile);
1699
1700 if (empty_etc) {
1701 r = unit_file_preset_all(UNIT_FILE_SYSTEM, false, NULL, UNIT_FILE_PRESET_ENABLE_ONLY, false, NULL, 0);
1702 if (r < 0)
1703 log_full_errno(r == -EEXIST ? LOG_NOTICE : LOG_WARNING, r, "Failed to populate /etc with preset unit settings, ignoring: %m");
1704 else
1705 log_info("Populated /etc with preset unit settings.");
1706 }
1707 }
1708
1709 r = manager_new(arg_running_as, arg_action == ACTION_TEST, &m);
1710 if (r < 0) {
1711 log_emergency_errno(r, "Failed to allocate manager object: %m");
1712 error_message = "Failed to allocate manager object";
1713 goto finish;
1714 }
1715
1716 m->confirm_spawn = arg_confirm_spawn;
1717 m->runtime_watchdog = arg_runtime_watchdog;
1718 m->shutdown_watchdog = arg_shutdown_watchdog;
1719 m->userspace_timestamp = userspace_timestamp;
1720 m->kernel_timestamp = kernel_timestamp;
1721 m->initrd_timestamp = initrd_timestamp;
1722 m->security_start_timestamp = security_start_timestamp;
1723 m->security_finish_timestamp = security_finish_timestamp;
1724
1725 manager_set_defaults(m);
1726 manager_set_show_status(m, arg_show_status);
1727 manager_set_first_boot(m, empty_etc);
1728
1729 /* Remember whether we should queue the default job */
1730 queue_default_job = !arg_serialization || arg_switched_root;
1731
1732 before_startup = now(CLOCK_MONOTONIC);
1733
1734 r = manager_startup(m, arg_serialization, fds);
1735 if (r < 0)
1736 log_error_errno(r, "Failed to fully start up daemon: %m");
1737
1738 /* This will close all file descriptors that were opened, but
1739 * not claimed by any unit. */
1740 fds = fdset_free(fds);
1741
1742 arg_serialization = safe_fclose(arg_serialization);
1743
1744 if (queue_default_job) {
1745 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
1746 Unit *target = NULL;
1747 Job *default_unit_job;
1748
1749 log_debug("Activating default unit: %s", arg_default_unit);
1750
1751 r = manager_load_unit(m, arg_default_unit, NULL, &error, &target);
1752 if (r < 0)
1753 log_error("Failed to load default target: %s", bus_error_message(&error, r));
1754 else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND)
1755 log_error_errno(target->load_error, "Failed to load default target: %m");
1756 else if (target->load_state == UNIT_MASKED)
1757 log_error("Default target masked.");
1758
1759 if (!target || target->load_state != UNIT_LOADED) {
1760 log_info("Trying to load rescue target...");
1761
1762 r = manager_load_unit(m, SPECIAL_RESCUE_TARGET, NULL, &error, &target);
1763 if (r < 0) {
1764 log_emergency("Failed to load rescue target: %s", bus_error_message(&error, r));
1765 error_message = "Failed to load rescue target";
1766 goto finish;
1767 } else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND) {
1768 log_emergency_errno(target->load_error, "Failed to load rescue target: %m");
1769 error_message = "Failed to load rescue target";
1770 goto finish;
1771 } else if (target->load_state == UNIT_MASKED) {
1772 log_emergency("Rescue target masked.");
1773 error_message = "Rescue target masked";
1774 goto finish;
1775 }
1776 }
1777
1778 assert(target->load_state == UNIT_LOADED);
1779
1780 if (arg_action == ACTION_TEST) {
1781 printf("-> By units:\n");
1782 manager_dump_units(m, stdout, "\t");
1783 }
1784
1785 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, &error, &default_unit_job);
1786 if (r == -EPERM) {
1787 log_debug("Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1788
1789 sd_bus_error_free(&error);
1790
1791 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, &error, &default_unit_job);
1792 if (r < 0) {
1793 log_emergency("Failed to start default target: %s", bus_error_message(&error, r));
1794 error_message = "Failed to start default target";
1795 goto finish;
1796 }
1797 } else if (r < 0) {
1798 log_emergency("Failed to isolate default target: %s", bus_error_message(&error, r));
1799 error_message = "Failed to isolate default target";
1800 goto finish;
1801 }
1802
1803 m->default_unit_job_id = default_unit_job->id;
1804
1805 after_startup = now(CLOCK_MONOTONIC);
1806 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
1807 "Loaded units and determined initial transaction in %s.",
1808 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
1809
1810 if (arg_action == ACTION_TEST) {
1811 printf("-> By jobs:\n");
1812 manager_dump_jobs(m, stdout, "\t");
1813 retval = EXIT_SUCCESS;
1814 goto finish;
1815 }
1816 }
1817
1818 for (;;) {
1819 r = manager_loop(m);
1820 if (r < 0) {
1821 log_emergency_errno(r, "Failed to run main loop: %m");
1822 error_message = "Failed to run main loop";
1823 goto finish;
1824 }
1825
1826 switch (m->exit_code) {
1827
1828 case MANAGER_RELOAD:
1829 log_info("Reloading.");
1830
1831 r = parse_config_file();
1832 if (r < 0)
1833 log_error("Failed to parse config file.");
1834
1835 manager_set_defaults(m);
1836
1837 r = manager_reload(m);
1838 if (r < 0)
1839 log_error_errno(r, "Failed to reload: %m");
1840 break;
1841
1842 case MANAGER_REEXECUTE:
1843
1844 if (prepare_reexecute(m, &arg_serialization, &fds, false) < 0) {
1845 error_message = "Failed to prepare for reexecution";
1846 goto finish;
1847 }
1848
1849 reexecute = true;
1850 log_notice("Reexecuting.");
1851 goto finish;
1852
1853 case MANAGER_SWITCH_ROOT:
1854 /* Steal the switch root parameters */
1855 switch_root_dir = m->switch_root;
1856 switch_root_init = m->switch_root_init;
1857 m->switch_root = m->switch_root_init = NULL;
1858
1859 if (!switch_root_init)
1860 if (prepare_reexecute(m, &arg_serialization, &fds, true) < 0) {
1861 error_message = "Failed to prepare for reexecution";
1862 goto finish;
1863 }
1864
1865 reexecute = true;
1866 log_notice("Switching root.");
1867 goto finish;
1868
1869 case MANAGER_EXIT:
1870 retval = m->return_value;
1871
1872 if (m->running_as == MANAGER_USER) {
1873 log_debug("Exit.");
1874 goto finish;
1875 }
1876
1877 /* fallthrough */
1878 case MANAGER_REBOOT:
1879 case MANAGER_POWEROFF:
1880 case MANAGER_HALT:
1881 case MANAGER_KEXEC: {
1882 static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
1883 [MANAGER_EXIT] = "exit",
1884 [MANAGER_REBOOT] = "reboot",
1885 [MANAGER_POWEROFF] = "poweroff",
1886 [MANAGER_HALT] = "halt",
1887 [MANAGER_KEXEC] = "kexec"
1888 };
1889
1890 assert_se(shutdown_verb = table[m->exit_code]);
1891 arm_reboot_watchdog = m->exit_code == MANAGER_REBOOT;
1892
1893 log_notice("Shutting down.");
1894 goto finish;
1895 }
1896
1897 default:
1898 assert_not_reached("Unknown exit code.");
1899 }
1900 }
1901
1902 finish:
1903 pager_close();
1904
1905 if (m)
1906 arg_shutdown_watchdog = m->shutdown_watchdog;
1907
1908 m = manager_free(m);
1909
1910 for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++)
1911 arg_default_rlimit[j] = mfree(arg_default_rlimit[j]);
1912
1913 arg_default_unit = mfree(arg_default_unit);
1914 arg_join_controllers = strv_free_free(arg_join_controllers);
1915 arg_default_environment = strv_free(arg_default_environment);
1916 arg_syscall_archs = set_free(arg_syscall_archs);
1917
1918 mac_selinux_finish();
1919
1920 if (reexecute) {
1921 const char **args;
1922 unsigned i, args_size;
1923
1924 /* Close and disarm the watchdog, so that the new
1925 * instance can reinitialize it, but doesn't get
1926 * rebooted while we do that */
1927 watchdog_close(true);
1928
1929 /* Reset the RLIMIT_NOFILE to the kernel default, so
1930 * that the new systemd can pass the kernel default to
1931 * its child processes */
1932 if (saved_rlimit_nofile.rlim_cur > 0)
1933 (void) setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile);
1934
1935 if (switch_root_dir) {
1936 /* Kill all remaining processes from the
1937 * initrd, but don't wait for them, so that we
1938 * can handle the SIGCHLD for them after
1939 * deserializing. */
1940 broadcast_signal(SIGTERM, false, true);
1941
1942 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1943 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1944 if (r < 0)
1945 log_error_errno(r, "Failed to switch root, trying to continue: %m");
1946 }
1947
1948 args_size = MAX(6, argc+1);
1949 args = newa(const char*, args_size);
1950
1951 if (!switch_root_init) {
1952 char sfd[DECIMAL_STR_MAX(int) + 1];
1953
1954 /* First try to spawn ourselves with the right
1955 * path, and with full serialization. We do
1956 * this only if the user didn't specify an
1957 * explicit init to spawn. */
1958
1959 assert(arg_serialization);
1960 assert(fds);
1961
1962 xsprintf(sfd, "%i", fileno(arg_serialization));
1963
1964 i = 0;
1965 args[i++] = SYSTEMD_BINARY_PATH;
1966 if (switch_root_dir)
1967 args[i++] = "--switched-root";
1968 args[i++] = arg_running_as == MANAGER_SYSTEM ? "--system" : "--user";
1969 args[i++] = "--deserialize";
1970 args[i++] = sfd;
1971 args[i++] = NULL;
1972
1973 /* do not pass along the environment we inherit from the kernel or initrd */
1974 if (switch_root_dir)
1975 (void) clearenv();
1976
1977 assert(i <= args_size);
1978
1979 /*
1980 * We want valgrind to print its memory usage summary before reexecution.
1981 * Valgrind won't do this is on its own on exec(), but it will do it on exit().
1982 * Hence, to ensure we get a summary here, fork() off a child, let it exit() cleanly,
1983 * so that it prints the summary, and wait() for it in the parent, before proceeding into the exec().
1984 */
1985 valgrind_summary_hack();
1986
1987 (void) execv(args[0], (char* const*) args);
1988 }
1989
1990 /* Try the fallback, if there is any, without any
1991 * serialization. We pass the original argv[] and
1992 * envp[]. (Well, modulo the ordering changes due to
1993 * getopt() in argv[], and some cleanups in envp[],
1994 * but let's hope that doesn't matter.) */
1995
1996 arg_serialization = safe_fclose(arg_serialization);
1997 fds = fdset_free(fds);
1998
1999 /* Reopen the console */
2000 (void) make_console_stdio();
2001
2002 for (j = 1, i = 1; j < (unsigned) argc; j++)
2003 args[i++] = argv[j];
2004 args[i++] = NULL;
2005 assert(i <= args_size);
2006
2007 /* Reenable any blocked signals, especially important
2008 * if we switch from initial ramdisk to init=... */
2009 (void) reset_all_signal_handlers();
2010 (void) reset_signal_mask();
2011
2012 if (switch_root_init) {
2013 args[0] = switch_root_init;
2014 (void) execv(args[0], (char* const*) args);
2015 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
2016 }
2017
2018 args[0] = "/sbin/init";
2019 (void) execv(args[0], (char* const*) args);
2020
2021 if (errno == ENOENT) {
2022 log_warning("No /sbin/init, trying fallback");
2023
2024 args[0] = "/bin/sh";
2025 args[1] = NULL;
2026 (void) execv(args[0], (char* const*) args);
2027 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
2028 } else
2029 log_warning_errno(errno, "Failed to execute /sbin/init, giving up: %m");
2030 }
2031
2032 arg_serialization = safe_fclose(arg_serialization);
2033 fds = fdset_free(fds);
2034
2035 #ifdef HAVE_VALGRIND_VALGRIND_H
2036 /* If we are PID 1 and running under valgrind, then let's exit
2037 * here explicitly. valgrind will only generate nice output on
2038 * exit(), not on exec(), hence let's do the former not the
2039 * latter here. */
2040 if (getpid() == 1 && RUNNING_ON_VALGRIND)
2041 return 0;
2042 #endif
2043
2044 if (shutdown_verb) {
2045 char log_level[DECIMAL_STR_MAX(int) + 1];
2046 char exit_code[DECIMAL_STR_MAX(uint8_t) + 1];
2047 const char* command_line[11] = {
2048 SYSTEMD_SHUTDOWN_BINARY_PATH,
2049 shutdown_verb,
2050 "--log-level", log_level,
2051 "--log-target",
2052 };
2053 unsigned pos = 5;
2054 _cleanup_strv_free_ char **env_block = NULL;
2055
2056 assert(command_line[pos] == NULL);
2057 env_block = strv_copy(environ);
2058
2059 xsprintf(log_level, "%d", log_get_max_level());
2060
2061 switch (log_get_target()) {
2062
2063 case LOG_TARGET_KMSG:
2064 case LOG_TARGET_JOURNAL_OR_KMSG:
2065 case LOG_TARGET_SYSLOG_OR_KMSG:
2066 command_line[pos++] = "kmsg";
2067 break;
2068
2069 case LOG_TARGET_NULL:
2070 command_line[pos++] = "null";
2071 break;
2072
2073 case LOG_TARGET_CONSOLE:
2074 default:
2075 command_line[pos++] = "console";
2076 break;
2077 };
2078
2079 if (log_get_show_color())
2080 command_line[pos++] = "--log-color";
2081
2082 if (log_get_show_location())
2083 command_line[pos++] = "--log-location";
2084
2085 if (streq(shutdown_verb, "exit")) {
2086 command_line[pos++] = "--exit-code";
2087 command_line[pos++] = exit_code;
2088 xsprintf(exit_code, "%d", retval);
2089 }
2090
2091 assert(pos < ELEMENTSOF(command_line));
2092
2093 if (arm_reboot_watchdog && arg_shutdown_watchdog > 0) {
2094 char *e;
2095
2096 /* If we reboot let's set the shutdown
2097 * watchdog and tell the shutdown binary to
2098 * repeatedly ping it */
2099 r = watchdog_set_timeout(&arg_shutdown_watchdog);
2100 watchdog_close(r < 0);
2101
2102 /* Tell the binary how often to ping, ignore failure */
2103 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
2104 (void) strv_push(&env_block, e);
2105 } else
2106 watchdog_close(true);
2107
2108 /* Avoid the creation of new processes forked by the
2109 * kernel; at this point, we will not listen to the
2110 * signals anyway */
2111 if (detect_container() <= 0)
2112 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
2113
2114 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
2115 log_error_errno(errno, "Failed to execute shutdown binary, %s: %m",
2116 getpid() == 1 ? "freezing" : "quitting");
2117 }
2118
2119 if (getpid() == 1) {
2120 if (error_message)
2121 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2122 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
2123 "%s, freezing.", error_message);
2124 freeze_or_reboot();
2125 }
2126
2127 return retval;
2128 }
Unnamed repository; edit this file 'description' to name the repository.