]> git.ipfire.org Git - thirdparty/systemd.git/blob - src/core/main.c
projects / thirdparty / systemd.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge pull request #4536 from poettering/seccomp-namespaces
[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 "format-util.h"
57 #include "fs-util.h"
58 #include "hostname-setup.h"
59 #include "ima-setup.h"
60 #include "killall.h"
61 #include "kmod-setup.h"
62 #include "load-fragment.h"
63 #include "log.h"
64 #include "loopback-setup.h"
65 #include "machine-id-setup.h"
66 #include "manager.h"
67 #include "missing.h"
68 #include "mount-setup.h"
69 #include "pager.h"
70 #include "parse-util.h"
71 #include "proc-cmdline.h"
72 #include "process-util.h"
73 #include "raw-clone.h"
74 #include "rlimit-util.h"
75 #ifdef HAVE_SECCOMP
76 #include "seccomp-util.h"
77 #endif
78 #include "selinux-setup.h"
79 #include "selinux-util.h"
80 #include "signal-util.h"
81 #include "smack-setup.h"
82 #include "special.h"
83 #include "stat-util.h"
84 #include "stdio-util.h"
85 #include "strv.h"
86 #include "switch-root.h"
87 #include "terminal-util.h"
88 #include "umask-util.h"
89 #include "user-util.h"
90 #include "virt.h"
91 #include "watchdog.h"
92 #include "emergency-action.h"
93
94 static enum {
95 ACTION_RUN,
96 ACTION_HELP,
97 ACTION_VERSION,
98 ACTION_TEST,
99 ACTION_DUMP_CONFIGURATION_ITEMS
100 } arg_action = ACTION_RUN;
101 static char *arg_default_unit = NULL;
102 static bool arg_system = false;
103 static bool arg_dump_core = true;
104 static int arg_crash_chvt = -1;
105 static bool arg_crash_shell = false;
106 static bool arg_crash_reboot = false;
107 static bool arg_confirm_spawn = false;
108 static ShowStatus arg_show_status = _SHOW_STATUS_UNSET;
109 static bool arg_switched_root = false;
110 static bool arg_no_pager = false;
111 static char ***arg_join_controllers = NULL;
112 static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
113 static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
114 static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
115 static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
116 static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
117 static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
118 static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
119 static usec_t arg_runtime_watchdog = 0;
120 static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
121 static char **arg_default_environment = NULL;
122 static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {};
123 static uint64_t arg_capability_bounding_set = CAP_ALL;
124 static nsec_t arg_timer_slack_nsec = NSEC_INFINITY;
125 static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
126 static Set* arg_syscall_archs = NULL;
127 static FILE* arg_serialization = NULL;
128 static bool arg_default_cpu_accounting = false;
129 static bool arg_default_io_accounting = false;
130 static bool arg_default_blockio_accounting = false;
131 static bool arg_default_memory_accounting = false;
132 static bool arg_default_tasks_accounting = true;
133 static uint64_t arg_default_tasks_max = UINT64_MAX;
134 static sd_id128_t arg_machine_id = {};
135 static EmergencyAction arg_cad_burst_action = EMERGENCY_ACTION_REBOOT_FORCE;
136
137 noreturn static void freeze_or_reboot(void) {
138
139 if (arg_crash_reboot) {
140 log_notice("Rebooting in 10s...");
141 (void) sleep(10);
142
143 log_notice("Rebooting now...");
144 (void) reboot(RB_AUTOBOOT);
145 log_emergency_errno(errno, "Failed to reboot: %m");
146 }
147
148 log_emergency("Freezing execution.");
149 freeze();
150 }
151
152 noreturn static void crash(int sig) {
153 struct sigaction sa;
154 pid_t pid;
155
156 if (getpid() != 1)
157 /* Pass this on immediately, if this is not PID 1 */
158 (void) raise(sig);
159 else if (!arg_dump_core)
160 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
161 else {
162 sa = (struct sigaction) {
163 .sa_handler = nop_signal_handler,
164 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
165 };
166
167 /* We want to wait for the core process, hence let's enable SIGCHLD */
168 (void) sigaction(SIGCHLD, &sa, NULL);
169
170 pid = raw_clone(SIGCHLD);
171 if (pid < 0)
172 log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
173 else if (pid == 0) {
174 /* Enable default signal handler for core dump */
175
176 sa = (struct sigaction) {
177 .sa_handler = SIG_DFL,
178 };
179 (void) sigaction(sig, &sa, NULL);
180
181 /* Don't limit the coredump size */
182 (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
183
184 /* Just to be sure... */
185 (void) chdir("/");
186
187 /* Raise the signal again */
188 pid = raw_getpid();
189 (void) kill(pid, sig); /* raise() would kill the parent */
190
191 assert_not_reached("We shouldn't be here...");
192 _exit(EXIT_FAILURE);
193 } else {
194 siginfo_t status;
195 int r;
196
197 /* Order things nicely. */
198 r = wait_for_terminate(pid, &status);
199 if (r < 0)
200 log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
201 else if (status.si_code != CLD_DUMPED)
202 log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
203 signal_to_string(sig),
204 pid, sigchld_code_to_string(status.si_code),
205 status.si_status,
206 strna(status.si_code == CLD_EXITED
207 ? exit_status_to_string(status.si_status, EXIT_STATUS_MINIMAL)
208 : signal_to_string(status.si_status)));
209 else
210 log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
211 }
212 }
213
214 if (arg_crash_chvt >= 0)
215 (void) chvt(arg_crash_chvt);
216
217 sa = (struct sigaction) {
218 .sa_handler = SIG_IGN,
219 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
220 };
221
222 /* Let the kernel reap children for us */
223 (void) sigaction(SIGCHLD, &sa, NULL);
224
225 if (arg_crash_shell) {
226 log_notice("Executing crash shell in 10s...");
227 (void) sleep(10);
228
229 pid = raw_clone(SIGCHLD);
230 if (pid < 0)
231 log_emergency_errno(errno, "Failed to fork off crash shell: %m");
232 else if (pid == 0) {
233 (void) setsid();
234 (void) make_console_stdio();
235 (void) execle("/bin/sh", "/bin/sh", NULL, environ);
236
237 log_emergency_errno(errno, "execle() failed: %m");
238 _exit(EXIT_FAILURE);
239 } else {
240 log_info("Spawned crash shell as PID "PID_FMT".", pid);
241 (void) wait_for_terminate(pid, NULL);
242 }
243 }
244
245 freeze_or_reboot();
246 }
247
248 static void install_crash_handler(void) {
249 static const struct sigaction sa = {
250 .sa_handler = crash,
251 .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
252 };
253 int r;
254
255 /* We ignore the return value here, since, we don't mind if we
256 * cannot set up a crash handler */
257 r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
258 if (r < 0)
259 log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
260 }
261
262 static int console_setup(void) {
263 _cleanup_close_ int tty_fd = -1;
264 int r;
265
266 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
267 if (tty_fd < 0)
268 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
269
270 /* We don't want to force text mode. plymouth may be showing
271 * pictures already from initrd. */
272 r = reset_terminal_fd(tty_fd, false);
273 if (r < 0)
274 return log_error_errno(r, "Failed to reset /dev/console: %m");
275
276 return 0;
277 }
278
279 static int parse_crash_chvt(const char *value) {
280 int b;
281
282 if (safe_atoi(value, &arg_crash_chvt) >= 0)
283 return 0;
284
285 b = parse_boolean(value);
286 if (b < 0)
287 return b;
288
289 if (b > 0)
290 arg_crash_chvt = 0; /* switch to where kmsg goes */
291 else
292 arg_crash_chvt = -1; /* turn off switching */
293
294 return 0;
295 }
296
297 static int set_machine_id(const char *m) {
298 sd_id128_t t;
299 assert(m);
300
301 if (sd_id128_from_string(m, &t) < 0)
302 return -EINVAL;
303
304 if (sd_id128_is_null(t))
305 return -EINVAL;
306
307 arg_machine_id = t;
308 return 0;
309 }
310
311 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
312
313 int r;
314
315 assert(key);
316
317 if (streq(key, "systemd.unit") && value) {
318
319 if (!in_initrd())
320 return free_and_strdup(&arg_default_unit, value);
321
322 } else if (streq(key, "rd.systemd.unit") && value) {
323
324 if (in_initrd())
325 return free_and_strdup(&arg_default_unit, value);
326
327 } else if (streq(key, "systemd.dump_core") && value) {
328
329 r = parse_boolean(value);
330 if (r < 0)
331 log_warning("Failed to parse dump core switch %s. Ignoring.", value);
332 else
333 arg_dump_core = r;
334
335 } else if (streq(key, "systemd.crash_chvt") && value) {
336
337 if (parse_crash_chvt(value) < 0)
338 log_warning("Failed to parse crash chvt switch %s. Ignoring.", value);
339
340 } else if (streq(key, "systemd.crash_shell") && value) {
341
342 r = parse_boolean(value);
343 if (r < 0)
344 log_warning("Failed to parse crash shell switch %s. Ignoring.", value);
345 else
346 arg_crash_shell = r;
347
348 } else if (streq(key, "systemd.crash_reboot") && value) {
349
350 r = parse_boolean(value);
351 if (r < 0)
352 log_warning("Failed to parse crash reboot switch %s. Ignoring.", value);
353 else
354 arg_crash_reboot = r;
355
356 } else if (streq(key, "systemd.confirm_spawn") && value) {
357
358 r = parse_boolean(value);
359 if (r < 0)
360 log_warning("Failed to parse confirm spawn switch %s. Ignoring.", value);
361 else
362 arg_confirm_spawn = r;
363
364 } else if (streq(key, "systemd.show_status") && value) {
365
366 r = parse_show_status(value, &arg_show_status);
367 if (r < 0)
368 log_warning("Failed to parse show status switch %s. Ignoring.", value);
369
370 } else if (streq(key, "systemd.default_standard_output") && value) {
371
372 r = exec_output_from_string(value);
373 if (r < 0)
374 log_warning("Failed to parse default standard output switch %s. Ignoring.", value);
375 else
376 arg_default_std_output = r;
377
378 } else if (streq(key, "systemd.default_standard_error") && value) {
379
380 r = exec_output_from_string(value);
381 if (r < 0)
382 log_warning("Failed to parse default standard error switch %s. Ignoring.", value);
383 else
384 arg_default_std_error = r;
385
386 } else if (streq(key, "systemd.setenv") && value) {
387
388 if (env_assignment_is_valid(value)) {
389 char **env;
390
391 env = strv_env_set(arg_default_environment, value);
392 if (env)
393 arg_default_environment = env;
394 else
395 log_warning_errno(ENOMEM, "Setting environment variable '%s' failed, ignoring: %m", value);
396 } else
397 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
398
399 } else if (streq(key, "systemd.machine_id") && value) {
400
401 r = set_machine_id(value);
402 if (r < 0)
403 log_warning("MachineID '%s' is not valid. Ignoring.", value);
404
405 } else if (streq(key, "quiet") && !value) {
406
407 if (arg_show_status == _SHOW_STATUS_UNSET)
408 arg_show_status = SHOW_STATUS_AUTO;
409
410 } else if (streq(key, "debug") && !value) {
411
412 /* Note that log_parse_environment() handles 'debug'
413 * too, and sets the log level to LOG_DEBUG. */
414
415 if (detect_container() > 0)
416 log_set_target(LOG_TARGET_CONSOLE);
417
418 } else if (!value) {
419 const char *target;
420
421 /* SysV compatibility */
422 target = runlevel_to_target(key);
423 if (target)
424 return free_and_strdup(&arg_default_unit, target);
425
426 } else if (streq(key, "systemd.default_timeout_start_sec") && value) {
427
428 r = parse_sec(value, &arg_default_timeout_start_usec);
429 if (r < 0)
430 log_warning_errno(r, "Failed to parse default start timeout: %s, ignoring.", value);
431
432 if (arg_default_timeout_start_usec <= 0)
433 arg_default_timeout_start_usec = USEC_INFINITY;
434 }
435
436 return 0;
437 }
438
439 #define DEFINE_SETTER(name, func, descr) \
440 static int name(const char *unit, \
441 const char *filename, \
442 unsigned line, \
443 const char *section, \
444 unsigned section_line, \
445 const char *lvalue, \
446 int ltype, \
447 const char *rvalue, \
448 void *data, \
449 void *userdata) { \
450 \
451 int r; \
452 \
453 assert(filename); \
454 assert(lvalue); \
455 assert(rvalue); \
456 \
457 r = func(rvalue); \
458 if (r < 0) \
459 log_syntax(unit, LOG_ERR, filename, line, r, \
460 "Invalid " descr "'%s': %m", \
461 rvalue); \
462 \
463 return 0; \
464 }
465
466 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level")
467 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target")
468 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" )
469 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location")
470
471 static int config_parse_cpu_affinity2(
472 const char *unit,
473 const char *filename,
474 unsigned line,
475 const char *section,
476 unsigned section_line,
477 const char *lvalue,
478 int ltype,
479 const char *rvalue,
480 void *data,
481 void *userdata) {
482
483 _cleanup_cpu_free_ cpu_set_t *c = NULL;
484 int ncpus;
485
486 ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue);
487 if (ncpus < 0)
488 return ncpus;
489
490 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
491 log_warning("Failed to set CPU affinity: %m");
492
493 return 0;
494 }
495
496 static int config_parse_show_status(
497 const char* unit,
498 const char *filename,
499 unsigned line,
500 const char *section,
501 unsigned section_line,
502 const char *lvalue,
503 int ltype,
504 const char *rvalue,
505 void *data,
506 void *userdata) {
507
508 int k;
509 ShowStatus *b = data;
510
511 assert(filename);
512 assert(lvalue);
513 assert(rvalue);
514 assert(data);
515
516 k = parse_show_status(rvalue, b);
517 if (k < 0) {
518 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
519 return 0;
520 }
521
522 return 0;
523 }
524
525 static int config_parse_crash_chvt(
526 const char* unit,
527 const char *filename,
528 unsigned line,
529 const char *section,
530 unsigned section_line,
531 const char *lvalue,
532 int ltype,
533 const char *rvalue,
534 void *data,
535 void *userdata) {
536
537 int r;
538
539 assert(filename);
540 assert(lvalue);
541 assert(rvalue);
542
543 r = parse_crash_chvt(rvalue);
544 if (r < 0) {
545 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
546 return 0;
547 }
548
549 return 0;
550 }
551
552 static int config_parse_join_controllers(const char *unit,
553 const char *filename,
554 unsigned line,
555 const char *section,
556 unsigned section_line,
557 const char *lvalue,
558 int ltype,
559 const char *rvalue,
560 void *data,
561 void *userdata) {
562
563 const char *whole_rvalue = rvalue;
564 unsigned n = 0;
565
566 assert(filename);
567 assert(lvalue);
568 assert(rvalue);
569
570 arg_join_controllers = strv_free_free(arg_join_controllers);
571
572 for (;;) {
573 _cleanup_free_ char *word = NULL;
574 char **l;
575 int r;
576
577 r = extract_first_word(&rvalue, &word, NULL, EXTRACT_QUOTES);
578 if (r < 0) {
579 log_syntax(unit, LOG_ERR, filename, line, r, "Invalid value for %s: %s", lvalue, whole_rvalue);
580 return r;
581 }
582 if (r == 0)
583 break;
584
585 l = strv_split(word, ",");
586 if (!l)
587 return log_oom();
588 strv_uniq(l);
589
590 if (strv_length(l) <= 1) {
591 strv_free(l);
592 continue;
593 }
594
595 if (!arg_join_controllers) {
596 arg_join_controllers = new(char**, 2);
597 if (!arg_join_controllers) {
598 strv_free(l);
599 return log_oom();
600 }
601
602 arg_join_controllers[0] = l;
603 arg_join_controllers[1] = NULL;
604
605 n = 1;
606 } else {
607 char ***a;
608 char ***t;
609
610 t = new0(char**, n+2);
611 if (!t) {
612 strv_free(l);
613 return log_oom();
614 }
615
616 n = 0;
617
618 for (a = arg_join_controllers; *a; a++) {
619
620 if (strv_overlap(*a, l)) {
621 if (strv_extend_strv(&l, *a, false) < 0) {
622 strv_free(l);
623 strv_free_free(t);
624 return log_oom();
625 }
626
627 } else {
628 char **c;
629
630 c = strv_copy(*a);
631 if (!c) {
632 strv_free(l);
633 strv_free_free(t);
634 return log_oom();
635 }
636
637 t[n++] = c;
638 }
639 }
640
641 t[n++] = strv_uniq(l);
642
643 strv_free_free(arg_join_controllers);
644 arg_join_controllers = t;
645 }
646 }
647 if (!isempty(rvalue))
648 log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring.");
649
650 return 0;
651 }
652
653 static int parse_config_file(void) {
654
655 const ConfigTableItem items[] = {
656 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
657 { "Manager", "LogTarget", config_parse_target, 0, NULL },
658 { "Manager", "LogColor", config_parse_color, 0, NULL },
659 { "Manager", "LogLocation", config_parse_location, 0, NULL },
660 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
661 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL },
662 { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL },
663 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
664 { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
665 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
666 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
667 { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
668 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
669 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
670 { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
671 #ifdef HAVE_SECCOMP
672 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
673 #endif
674 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
675 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
676 { "Manager", "DefaultStandardOutput", config_parse_output, 0, &arg_default_std_output },
677 { "Manager", "DefaultStandardError", config_parse_output, 0, &arg_default_std_error },
678 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
679 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
680 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
681 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */
682 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec, 0, &arg_default_start_limit_interval },
683 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
684 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
685 { "Manager", "DefaultLimitCPU", config_parse_limit, RLIMIT_CPU, arg_default_rlimit },
686 { "Manager", "DefaultLimitFSIZE", config_parse_limit, RLIMIT_FSIZE, arg_default_rlimit },
687 { "Manager", "DefaultLimitDATA", config_parse_limit, RLIMIT_DATA, arg_default_rlimit },
688 { "Manager", "DefaultLimitSTACK", config_parse_limit, RLIMIT_STACK, arg_default_rlimit },
689 { "Manager", "DefaultLimitCORE", config_parse_limit, RLIMIT_CORE, arg_default_rlimit },
690 { "Manager", "DefaultLimitRSS", config_parse_limit, RLIMIT_RSS, arg_default_rlimit },
691 { "Manager", "DefaultLimitNOFILE", config_parse_limit, RLIMIT_NOFILE, arg_default_rlimit },
692 { "Manager", "DefaultLimitAS", config_parse_limit, RLIMIT_AS, arg_default_rlimit },
693 { "Manager", "DefaultLimitNPROC", config_parse_limit, RLIMIT_NPROC, arg_default_rlimit },
694 { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, RLIMIT_MEMLOCK, arg_default_rlimit },
695 { "Manager", "DefaultLimitLOCKS", config_parse_limit, RLIMIT_LOCKS, arg_default_rlimit },
696 { "Manager", "DefaultLimitSIGPENDING", config_parse_limit, RLIMIT_SIGPENDING, arg_default_rlimit },
697 { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, RLIMIT_MSGQUEUE, arg_default_rlimit },
698 { "Manager", "DefaultLimitNICE", config_parse_limit, RLIMIT_NICE, arg_default_rlimit },
699 { "Manager", "DefaultLimitRTPRIO", config_parse_limit, RLIMIT_RTPRIO, arg_default_rlimit },
700 { "Manager", "DefaultLimitRTTIME", config_parse_limit, RLIMIT_RTTIME, arg_default_rlimit },
701 { "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_default_cpu_accounting },
702 { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting },
703 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
704 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
705 { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
706 { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max },
707 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action, 0, &arg_cad_burst_action },
708 {}
709 };
710
711 const char *fn, *conf_dirs_nulstr;
712
713 fn = arg_system ?
714 PKGSYSCONFDIR "/system.conf" :
715 PKGSYSCONFDIR "/user.conf";
716
717 conf_dirs_nulstr = arg_system ?
718 CONF_PATHS_NULSTR("systemd/system.conf.d") :
719 CONF_PATHS_NULSTR("systemd/user.conf.d");
720
721 config_parse_many_nulstr(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, false, NULL);
722
723 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
724 * like everywhere else. */
725 if (arg_default_timeout_start_usec <= 0)
726 arg_default_timeout_start_usec = USEC_INFINITY;
727 if (arg_default_timeout_stop_usec <= 0)
728 arg_default_timeout_stop_usec = USEC_INFINITY;
729
730 return 0;
731 }
732
733 static void manager_set_defaults(Manager *m) {
734
735 assert(m);
736
737 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
738 m->default_std_output = arg_default_std_output;
739 m->default_std_error = arg_default_std_error;
740 m->default_timeout_start_usec = arg_default_timeout_start_usec;
741 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
742 m->default_restart_usec = arg_default_restart_usec;
743 m->default_start_limit_interval = arg_default_start_limit_interval;
744 m->default_start_limit_burst = arg_default_start_limit_burst;
745 m->default_cpu_accounting = arg_default_cpu_accounting;
746 m->default_io_accounting = arg_default_io_accounting;
747 m->default_blockio_accounting = arg_default_blockio_accounting;
748 m->default_memory_accounting = arg_default_memory_accounting;
749 m->default_tasks_accounting = arg_default_tasks_accounting;
750 m->default_tasks_max = arg_default_tasks_max;
751
752 manager_set_default_rlimits(m, arg_default_rlimit);
753 manager_environment_add(m, NULL, arg_default_environment);
754 }
755
756 static int parse_argv(int argc, char *argv[]) {
757
758 enum {
759 ARG_LOG_LEVEL = 0x100,
760 ARG_LOG_TARGET,
761 ARG_LOG_COLOR,
762 ARG_LOG_LOCATION,
763 ARG_UNIT,
764 ARG_SYSTEM,
765 ARG_USER,
766 ARG_TEST,
767 ARG_NO_PAGER,
768 ARG_VERSION,
769 ARG_DUMP_CONFIGURATION_ITEMS,
770 ARG_DUMP_CORE,
771 ARG_CRASH_CHVT,
772 ARG_CRASH_SHELL,
773 ARG_CRASH_REBOOT,
774 ARG_CONFIRM_SPAWN,
775 ARG_SHOW_STATUS,
776 ARG_DESERIALIZE,
777 ARG_SWITCHED_ROOT,
778 ARG_DEFAULT_STD_OUTPUT,
779 ARG_DEFAULT_STD_ERROR,
780 ARG_MACHINE_ID
781 };
782
783 static const struct option options[] = {
784 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
785 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
786 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
787 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
788 { "unit", required_argument, NULL, ARG_UNIT },
789 { "system", no_argument, NULL, ARG_SYSTEM },
790 { "user", no_argument, NULL, ARG_USER },
791 { "test", no_argument, NULL, ARG_TEST },
792 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
793 { "help", no_argument, NULL, 'h' },
794 { "version", no_argument, NULL, ARG_VERSION },
795 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
796 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
797 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
798 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
799 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
800 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
801 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
802 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
803 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
804 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
805 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
806 { "machine-id", required_argument, NULL, ARG_MACHINE_ID },
807 {}
808 };
809
810 int c, r;
811
812 assert(argc >= 1);
813 assert(argv);
814
815 if (getpid() == 1)
816 opterr = 0;
817
818 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
819
820 switch (c) {
821
822 case ARG_LOG_LEVEL:
823 r = log_set_max_level_from_string(optarg);
824 if (r < 0) {
825 log_error("Failed to parse log level %s.", optarg);
826 return r;
827 }
828
829 break;
830
831 case ARG_LOG_TARGET:
832 r = log_set_target_from_string(optarg);
833 if (r < 0) {
834 log_error("Failed to parse log target %s.", optarg);
835 return r;
836 }
837
838 break;
839
840 case ARG_LOG_COLOR:
841
842 if (optarg) {
843 r = log_show_color_from_string(optarg);
844 if (r < 0) {
845 log_error("Failed to parse log color setting %s.", optarg);
846 return r;
847 }
848 } else
849 log_show_color(true);
850
851 break;
852
853 case ARG_LOG_LOCATION:
854 if (optarg) {
855 r = log_show_location_from_string(optarg);
856 if (r < 0) {
857 log_error("Failed to parse log location setting %s.", optarg);
858 return r;
859 }
860 } else
861 log_show_location(true);
862
863 break;
864
865 case ARG_DEFAULT_STD_OUTPUT:
866 r = exec_output_from_string(optarg);
867 if (r < 0) {
868 log_error("Failed to parse default standard output setting %s.", optarg);
869 return r;
870 } else
871 arg_default_std_output = r;
872 break;
873
874 case ARG_DEFAULT_STD_ERROR:
875 r = exec_output_from_string(optarg);
876 if (r < 0) {
877 log_error("Failed to parse default standard error output setting %s.", optarg);
878 return r;
879 } else
880 arg_default_std_error = r;
881 break;
882
883 case ARG_UNIT:
884
885 r = free_and_strdup(&arg_default_unit, optarg);
886 if (r < 0)
887 return log_error_errno(r, "Failed to set default unit %s: %m", optarg);
888
889 break;
890
891 case ARG_SYSTEM:
892 arg_system = true;
893 break;
894
895 case ARG_USER:
896 arg_system = false;
897 break;
898
899 case ARG_TEST:
900 arg_action = ACTION_TEST;
901 break;
902
903 case ARG_NO_PAGER:
904 arg_no_pager = true;
905 break;
906
907 case ARG_VERSION:
908 arg_action = ACTION_VERSION;
909 break;
910
911 case ARG_DUMP_CONFIGURATION_ITEMS:
912 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
913 break;
914
915 case ARG_DUMP_CORE:
916 if (!optarg)
917 arg_dump_core = true;
918 else {
919 r = parse_boolean(optarg);
920 if (r < 0)
921 return log_error_errno(r, "Failed to parse dump core boolean: %s", optarg);
922 arg_dump_core = r;
923 }
924 break;
925
926 case ARG_CRASH_CHVT:
927 r = parse_crash_chvt(optarg);
928 if (r < 0)
929 return log_error_errno(r, "Failed to parse crash virtual terminal index: %s", optarg);
930 break;
931
932 case ARG_CRASH_SHELL:
933 if (!optarg)
934 arg_crash_shell = true;
935 else {
936 r = parse_boolean(optarg);
937 if (r < 0)
938 return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg);
939 arg_crash_shell = r;
940 }
941 break;
942
943 case ARG_CRASH_REBOOT:
944 if (!optarg)
945 arg_crash_reboot = true;
946 else {
947 r = parse_boolean(optarg);
948 if (r < 0)
949 return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg);
950 arg_crash_reboot = r;
951 }
952 break;
953
954 case ARG_CONFIRM_SPAWN:
955 r = optarg ? parse_boolean(optarg) : 1;
956 if (r < 0) {
957 log_error("Failed to parse confirm spawn boolean %s.", optarg);
958 return r;
959 }
960 arg_confirm_spawn = r;
961 break;
962
963 case ARG_SHOW_STATUS:
964 if (optarg) {
965 r = parse_show_status(optarg, &arg_show_status);
966 if (r < 0) {
967 log_error("Failed to parse show status boolean %s.", optarg);
968 return r;
969 }
970 } else
971 arg_show_status = SHOW_STATUS_YES;
972 break;
973
974 case ARG_DESERIALIZE: {
975 int fd;
976 FILE *f;
977
978 r = safe_atoi(optarg, &fd);
979 if (r < 0 || fd < 0) {
980 log_error("Failed to parse deserialize option %s.", optarg);
981 return -EINVAL;
982 }
983
984 (void) fd_cloexec(fd, true);
985
986 f = fdopen(fd, "r");
987 if (!f)
988 return log_error_errno(errno, "Failed to open serialization fd: %m");
989
990 safe_fclose(arg_serialization);
991 arg_serialization = f;
992
993 break;
994 }
995
996 case ARG_SWITCHED_ROOT:
997 arg_switched_root = true;
998 break;
999
1000 case ARG_MACHINE_ID:
1001 r = set_machine_id(optarg);
1002 if (r < 0)
1003 return log_error_errno(r, "MachineID '%s' is not valid.", optarg);
1004 break;
1005
1006 case 'h':
1007 arg_action = ACTION_HELP;
1008 break;
1009
1010 case 'D':
1011 log_set_max_level(LOG_DEBUG);
1012 break;
1013
1014 case 'b':
1015 case 's':
1016 case 'z':
1017 /* Just to eat away the sysvinit kernel
1018 * cmdline args without getopt() error
1019 * messages that we'll parse in
1020 * parse_proc_cmdline_word() or ignore. */
1021
1022 case '?':
1023 if (getpid() != 1)
1024 return -EINVAL;
1025 else
1026 return 0;
1027
1028 default:
1029 assert_not_reached("Unhandled option code.");
1030 }
1031
1032 if (optind < argc && getpid() != 1) {
1033 /* Hmm, when we aren't run as init system
1034 * let's complain about excess arguments */
1035
1036 log_error("Excess arguments.");
1037 return -EINVAL;
1038 }
1039
1040 return 0;
1041 }
1042
1043 static int help(void) {
1044
1045 printf("%s [OPTIONS...]\n\n"
1046 "Starts up and maintains the system or user services.\n\n"
1047 " -h --help Show this help\n"
1048 " --test Determine startup sequence, dump it and exit\n"
1049 " --no-pager Do not pipe output into a pager\n"
1050 " --dump-configuration-items Dump understood unit configuration items\n"
1051 " --unit=UNIT Set default unit\n"
1052 " --system Run a system instance, even if PID != 1\n"
1053 " --user Run a user instance\n"
1054 " --dump-core[=BOOL] Dump core on crash\n"
1055 " --crash-vt=NR Change to specified VT on crash\n"
1056 " --crash-reboot[=BOOL] Reboot on crash\n"
1057 " --crash-shell[=BOOL] Run shell on crash\n"
1058 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1059 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1060 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1061 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1062 " --log-color[=BOOL] Highlight important log messages\n"
1063 " --log-location[=BOOL] Include code location in log messages\n"
1064 " --default-standard-output= Set default standard output for services\n"
1065 " --default-standard-error= Set default standard error output for services\n",
1066 program_invocation_short_name);
1067
1068 return 0;
1069 }
1070
1071 static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
1072 _cleanup_fdset_free_ FDSet *fds = NULL;
1073 _cleanup_fclose_ FILE *f = NULL;
1074 int r;
1075
1076 assert(m);
1077 assert(_f);
1078 assert(_fds);
1079
1080 r = manager_open_serialization(m, &f);
1081 if (r < 0)
1082 return log_error_errno(r, "Failed to create serialization file: %m");
1083
1084 /* Make sure nothing is really destructed when we shut down */
1085 m->n_reloading++;
1086 bus_manager_send_reloading(m, true);
1087
1088 fds = fdset_new();
1089 if (!fds)
1090 return log_oom();
1091
1092 r = manager_serialize(m, f, fds, switching_root);
1093 if (r < 0)
1094 return log_error_errno(r, "Failed to serialize state: %m");
1095
1096 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1097 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1098
1099 r = fd_cloexec(fileno(f), false);
1100 if (r < 0)
1101 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1102
1103 r = fdset_cloexec(fds, false);
1104 if (r < 0)
1105 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1106
1107 *_f = f;
1108 *_fds = fds;
1109
1110 f = NULL;
1111 fds = NULL;
1112
1113 return 0;
1114 }
1115
1116 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1117 struct rlimit nl;
1118 int r;
1119
1120 assert(saved_rlimit);
1121
1122 /* Save the original RLIMIT_NOFILE so that we can reset it
1123 * later when transitioning from the initrd to the main
1124 * systemd or suchlike. */
1125 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1126 return log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1127
1128 /* Make sure forked processes get the default kernel setting */
1129 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1130 struct rlimit *rl;
1131
1132 rl = newdup(struct rlimit, saved_rlimit, 1);
1133 if (!rl)
1134 return log_oom();
1135
1136 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1137 }
1138
1139 /* Bump up the resource limit for ourselves substantially */
1140 nl.rlim_cur = nl.rlim_max = 64*1024;
1141 r = setrlimit_closest(RLIMIT_NOFILE, &nl);
1142 if (r < 0)
1143 return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1144
1145 return 0;
1146 }
1147
1148 static void test_usr(void) {
1149
1150 /* Check that /usr is not a separate fs */
1151
1152 if (dir_is_empty("/usr") <= 0)
1153 return;
1154
1155 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1156 "Some things will probably break (sometimes even silently) in mysterious ways. "
1157 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1158 }
1159
1160 static int initialize_join_controllers(void) {
1161 /* By default, mount "cpu" + "cpuacct" together, and "net_cls"
1162 * + "net_prio". We'd like to add "cpuset" to the mix, but
1163 * "cpuset" doesn't really work for groups with no initialized
1164 * attributes. */
1165
1166 arg_join_controllers = new(char**, 3);
1167 if (!arg_join_controllers)
1168 return -ENOMEM;
1169
1170 arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL);
1171 if (!arg_join_controllers[0])
1172 goto oom;
1173
1174 arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL);
1175 if (!arg_join_controllers[1])
1176 goto oom;
1177
1178 arg_join_controllers[2] = NULL;
1179 return 0;
1180
1181 oom:
1182 arg_join_controllers = strv_free_free(arg_join_controllers);
1183 return -ENOMEM;
1184 }
1185
1186 static int enforce_syscall_archs(Set *archs) {
1187 #ifdef HAVE_SECCOMP
1188 scmp_filter_ctx *seccomp;
1189 Iterator i;
1190 void *id;
1191 int r;
1192
1193 if (!is_seccomp_available())
1194 return 0;
1195
1196 seccomp = seccomp_init(SCMP_ACT_ALLOW);
1197 if (!seccomp)
1198 return log_oom();
1199
1200 SET_FOREACH(id, arg_syscall_archs, i) {
1201 r = seccomp_arch_add(seccomp, PTR_TO_UINT32(id) - 1);
1202 if (r == -EEXIST)
1203 continue;
1204 if (r < 0) {
1205 log_error_errno(r, "Failed to add architecture to seccomp: %m");
1206 goto finish;
1207 }
1208 }
1209
1210 r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
1211 if (r < 0) {
1212 log_error_errno(r, "Failed to unset NO_NEW_PRIVS: %m");
1213 goto finish;
1214 }
1215
1216 r = seccomp_load(seccomp);
1217 if (r < 0)
1218 log_error_errno(r, "Failed to add install architecture seccomp: %m");
1219
1220 finish:
1221 seccomp_release(seccomp);
1222 return r;
1223 #else
1224 return 0;
1225 #endif
1226 }
1227
1228 static int status_welcome(void) {
1229 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1230 int r;
1231
1232 r = parse_env_file("/etc/os-release", NEWLINE,
1233 "PRETTY_NAME", &pretty_name,
1234 "ANSI_COLOR", &ansi_color,
1235 NULL);
1236 if (r == -ENOENT)
1237 r = parse_env_file("/usr/lib/os-release", NEWLINE,
1238 "PRETTY_NAME", &pretty_name,
1239 "ANSI_COLOR", &ansi_color,
1240 NULL);
1241
1242 if (r < 0 && r != -ENOENT)
1243 log_warning_errno(r, "Failed to read os-release file: %m");
1244
1245 if (log_get_show_color())
1246 return status_printf(NULL, false, false,
1247 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1248 isempty(ansi_color) ? "1" : ansi_color,
1249 isempty(pretty_name) ? "Linux" : pretty_name);
1250 else
1251 return status_printf(NULL, false, false,
1252 "\nWelcome to %s!\n",
1253 isempty(pretty_name) ? "Linux" : pretty_name);
1254 }
1255
1256 static int write_container_id(void) {
1257 const char *c;
1258 int r;
1259
1260 c = getenv("container");
1261 if (isempty(c))
1262 return 0;
1263
1264 RUN_WITH_UMASK(0022)
1265 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1266 if (r < 0)
1267 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
1268
1269 return 1;
1270 }
1271
1272 static int bump_unix_max_dgram_qlen(void) {
1273 _cleanup_free_ char *qlen = NULL;
1274 unsigned long v;
1275 int r;
1276
1277 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel
1278 * default of 16 is simply too low. We set the value really
1279 * really early during boot, so that it is actually applied to
1280 * all our sockets, including the $NOTIFY_SOCKET one. */
1281
1282 r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1283 if (r < 0)
1284 return log_warning_errno(r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1285
1286 r = safe_atolu(qlen, &v);
1287 if (r < 0)
1288 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length, ignoring: %m");
1289
1290 if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1291 return 0;
1292
1293 qlen = mfree(qlen);
1294 if (asprintf(&qlen, "%lu\n", DEFAULT_UNIX_MAX_DGRAM_QLEN) < 0)
1295 return log_oom();
1296
1297 r = write_string_file("/proc/sys/net/unix/max_dgram_qlen", qlen, 0);
1298 if (r < 0)
1299 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1300 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1301
1302 return 1;
1303 }
1304
1305 static int fixup_environment(void) {
1306 _cleanup_free_ char *term = NULL;
1307 int r;
1308
1309 /* We expect the environment to be set correctly
1310 * if run inside a container. */
1311 if (detect_container() > 0)
1312 return 0;
1313
1314 /* When started as PID1, the kernel uses /dev/console
1315 * for our stdios and uses TERM=linux whatever the
1316 * backend device used by the console. We try to make
1317 * a better guess here since some consoles might not
1318 * have support for color mode for example.
1319 *
1320 * However if TERM was configured through the kernel
1321 * command line then leave it alone. */
1322
1323 r = get_proc_cmdline_key("TERM=", &term);
1324 if (r < 0)
1325 return r;
1326
1327 if (r == 0) {
1328 term = strdup(default_term_for_tty("/dev/console"));
1329 if (!term)
1330 return -ENOMEM;
1331 }
1332
1333 if (setenv("TERM", term, 1) < 0)
1334 return -errno;
1335
1336 return 0;
1337 }
1338
1339 int main(int argc, char *argv[]) {
1340 Manager *m = NULL;
1341 int r, retval = EXIT_FAILURE;
1342 usec_t before_startup, after_startup;
1343 char timespan[FORMAT_TIMESPAN_MAX];
1344 FDSet *fds = NULL;
1345 bool reexecute = false;
1346 const char *shutdown_verb = NULL;
1347 dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL;
1348 dual_timestamp userspace_timestamp = DUAL_TIMESTAMP_NULL;
1349 dual_timestamp kernel_timestamp = DUAL_TIMESTAMP_NULL;
1350 dual_timestamp security_start_timestamp = DUAL_TIMESTAMP_NULL;
1351 dual_timestamp security_finish_timestamp = DUAL_TIMESTAMP_NULL;
1352 static char systemd[] = "systemd";
1353 bool skip_setup = false;
1354 unsigned j;
1355 bool loaded_policy = false;
1356 bool arm_reboot_watchdog = false;
1357 bool queue_default_job = false;
1358 bool empty_etc = false;
1359 char *switch_root_dir = NULL, *switch_root_init = NULL;
1360 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0);
1361 const char *error_message = NULL;
1362
1363 #ifdef HAVE_SYSV_COMPAT
1364 if (getpid() != 1 && strstr(program_invocation_short_name, "init")) {
1365 /* This is compatibility support for SysV, where
1366 * calling init as a user is identical to telinit. */
1367
1368 execv(SYSTEMCTL_BINARY_PATH, argv);
1369 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1370 return 1;
1371 }
1372 #endif
1373
1374 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
1375 dual_timestamp_get(&userspace_timestamp);
1376
1377 /* Determine if this is a reexecution or normal bootup. We do
1378 * the full command line parsing much later, so let's just
1379 * have a quick peek here. */
1380 if (strv_find(argv+1, "--deserialize"))
1381 skip_setup = true;
1382
1383 /* If we have switched root, do all the special setup
1384 * things */
1385 if (strv_find(argv+1, "--switched-root"))
1386 skip_setup = false;
1387
1388 /* If we get started via the /sbin/init symlink then we are
1389 called 'init'. After a subsequent reexecution we are then
1390 called 'systemd'. That is confusing, hence let's call us
1391 systemd right-away. */
1392 program_invocation_short_name = systemd;
1393 prctl(PR_SET_NAME, systemd);
1394
1395 saved_argv = argv;
1396 saved_argc = argc;
1397
1398 log_set_upgrade_syslog_to_journal(true);
1399
1400 /* Disable the umask logic */
1401 if (getpid() == 1)
1402 umask(0);
1403
1404 if (getpid() == 1 && detect_container() <= 0) {
1405
1406 /* Running outside of a container as PID 1 */
1407 arg_system = true;
1408 log_set_target(LOG_TARGET_KMSG);
1409 log_open();
1410
1411 if (in_initrd())
1412 initrd_timestamp = userspace_timestamp;
1413
1414 if (!skip_setup) {
1415 r = mount_setup_early();
1416 if (r < 0) {
1417 error_message = "Failed to early mount API filesystems";
1418 goto finish;
1419 }
1420 dual_timestamp_get(&security_start_timestamp);
1421 if (mac_selinux_setup(&loaded_policy) < 0) {
1422 error_message = "Failed to load SELinux policy";
1423 goto finish;
1424 } else if (mac_smack_setup(&loaded_policy) < 0) {
1425 error_message = "Failed to load SMACK policy";
1426 goto finish;
1427 } else if (ima_setup() < 0) {
1428 error_message = "Failed to load IMA policy";
1429 goto finish;
1430 }
1431 dual_timestamp_get(&security_finish_timestamp);
1432 }
1433
1434 if (mac_selinux_init() < 0) {
1435 error_message = "Failed to initialize SELinux policy";
1436 goto finish;
1437 }
1438
1439 if (!skip_setup) {
1440 if (clock_is_localtime(NULL) > 0) {
1441 int min;
1442
1443 /*
1444 * The very first call of settimeofday() also does a time warp in the kernel.
1445 *
1446 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on
1447 * external tools to take care of maintaining the RTC and do all adjustments.
1448 * This matches the behavior of Windows, which leaves the RTC alone if the
1449 * registry tells that the RTC runs in UTC.
1450 */
1451 r = clock_set_timezone(&min);
1452 if (r < 0)
1453 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1454 else
1455 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1456 } else if (!in_initrd()) {
1457 /*
1458 * Do a dummy very first call to seal the kernel's time warp magic.
1459 *
1460 * Do not call this from inside the initrd. The initrd might not
1461 * carry /etc/adjtime with LOCAL, but the real system could be set up
1462 * that way. In such case, we need to delay the time-warp or the sealing
1463 * until we reach the real system.
1464 *
1465 * Do no set the kernel's timezone. The concept of local time cannot
1466 * be supported reliably, the time will jump or be incorrect at every daylight
1467 * saving time change. All kernel local time concepts will be treated
1468 * as UTC that way.
1469 */
1470 (void) clock_reset_timewarp();
1471 }
1472
1473 r = clock_apply_epoch();
1474 if (r < 0)
1475 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1476 else if (r > 0)
1477 log_info("System time before build time, advancing clock.");
1478 }
1479
1480 /* Set the default for later on, but don't actually
1481 * open the logs like this for now. Note that if we
1482 * are transitioning from the initrd there might still
1483 * be journal fd open, and we shouldn't attempt
1484 * opening that before we parsed /proc/cmdline which
1485 * might redirect output elsewhere. */
1486 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
1487
1488 } else if (getpid() == 1) {
1489 /* Running inside a container, as PID 1 */
1490 arg_system = true;
1491 log_set_target(LOG_TARGET_CONSOLE);
1492 log_close_console(); /* force reopen of /dev/console */
1493 log_open();
1494
1495 /* For the later on, see above... */
1496 log_set_target(LOG_TARGET_JOURNAL);
1497
1498 /* clear the kernel timestamp,
1499 * because we are in a container */
1500 kernel_timestamp = DUAL_TIMESTAMP_NULL;
1501 } else {
1502 /* Running as user instance */
1503 arg_system = false;
1504 log_set_target(LOG_TARGET_AUTO);
1505 log_open();
1506
1507 /* clear the kernel timestamp,
1508 * because we are not PID 1 */
1509 kernel_timestamp = DUAL_TIMESTAMP_NULL;
1510 }
1511
1512 if (getpid() == 1) {
1513 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1514 * will process core dumps for system services by default. */
1515 if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0)
1516 log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m");
1517
1518 /* But at the same time, turn off the core_pattern logic by default, so that no coredumps are stored
1519 * until the systemd-coredump tool is enabled via sysctl. */
1520 if (!skip_setup)
1521 (void) write_string_file("/proc/sys/kernel/core_pattern", "|/bin/false", 0);
1522 }
1523
1524 if (arg_system) {
1525 if (fixup_environment() < 0) {
1526 error_message = "Failed to fix up PID1 environment";
1527 goto finish;
1528 }
1529
1530 /* Try to figure out if we can use colors with the console. No
1531 * need to do that for user instances since they never log
1532 * into the console. */
1533 log_show_color(colors_enabled());
1534 r = make_null_stdio();
1535 if (r < 0)
1536 log_warning_errno(r, "Failed to redirect standard streams to /dev/null: %m");
1537 }
1538
1539 r = initialize_join_controllers();
1540 if (r < 0) {
1541 error_message = "Failed to initialize cgroup controllers";
1542 goto finish;
1543 }
1544
1545 /* Mount /proc, /sys and friends, so that /proc/cmdline and
1546 * /proc/$PID/fd is available. */
1547 if (getpid() == 1) {
1548
1549 /* Load the kernel modules early, so that we kdbus.ko is loaded before kdbusfs shall be mounted */
1550 if (!skip_setup)
1551 kmod_setup();
1552
1553 r = mount_setup(loaded_policy);
1554 if (r < 0) {
1555 error_message = "Failed to mount API filesystems";
1556 goto finish;
1557 }
1558 }
1559
1560 /* Reset all signal handlers. */
1561 (void) reset_all_signal_handlers();
1562 (void) ignore_signals(SIGNALS_IGNORE, -1);
1563
1564 arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U);
1565
1566 if (parse_config_file() < 0) {
1567 error_message = "Failed to parse config file";
1568 goto finish;
1569 }
1570
1571 if (arg_system) {
1572 r = parse_proc_cmdline(parse_proc_cmdline_item, NULL, false);
1573 if (r < 0)
1574 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
1575 }
1576
1577 /* Note that this also parses bits from the kernel command
1578 * line, including "debug". */
1579 log_parse_environment();
1580
1581 if (parse_argv(argc, argv) < 0) {
1582 error_message = "Failed to parse commandline arguments";
1583 goto finish;
1584 }
1585
1586 /* Initialize default unit */
1587 if (!arg_default_unit) {
1588 arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET);
1589 if (!arg_default_unit) {
1590 r = log_oom();
1591 error_message = "Failed to set default unit";
1592 goto finish;
1593 }
1594 }
1595
1596 if (arg_action == ACTION_TEST &&
1597 geteuid() == 0) {
1598 log_error("Don't run test mode as root.");
1599 goto finish;
1600 }
1601
1602 if (!arg_system &&
1603 arg_action == ACTION_RUN &&
1604 sd_booted() <= 0) {
1605 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
1606 goto finish;
1607 }
1608
1609 if (arg_system &&
1610 arg_action == ACTION_RUN &&
1611 running_in_chroot() > 0) {
1612 log_error("Cannot be run in a chroot() environment.");
1613 goto finish;
1614 }
1615
1616 if (arg_action == ACTION_TEST || arg_action == ACTION_HELP) {
1617 pager_open(arg_no_pager, false);
1618 skip_setup = true;
1619 }
1620
1621 if (arg_action == ACTION_HELP) {
1622 retval = help();
1623 goto finish;
1624 } else if (arg_action == ACTION_VERSION) {
1625 retval = version();
1626 goto finish;
1627 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
1628 pager_open(arg_no_pager, false);
1629 unit_dump_config_items(stdout);
1630 retval = EXIT_SUCCESS;
1631 goto finish;
1632 }
1633
1634 if (!arg_system &&
1635 !getenv("XDG_RUNTIME_DIR")) {
1636 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
1637 goto finish;
1638 }
1639
1640 assert_se(arg_action == ACTION_RUN || arg_action == ACTION_TEST);
1641
1642 /* Close logging fds, in order not to confuse fdset below */
1643 log_close();
1644
1645 /* Remember open file descriptors for later deserialization */
1646 r = fdset_new_fill(&fds);
1647 if (r < 0) {
1648 log_emergency_errno(r, "Failed to allocate fd set: %m");
1649 error_message = "Failed to allocate fd set";
1650 goto finish;
1651 } else
1652 fdset_cloexec(fds, true);
1653
1654 if (arg_serialization)
1655 assert_se(fdset_remove(fds, fileno(arg_serialization)) >= 0);
1656
1657 if (arg_system)
1658 /* Become a session leader if we aren't one yet. */
1659 setsid();
1660
1661 /* Move out of the way, so that we won't block unmounts */
1662 assert_se(chdir("/") == 0);
1663
1664 /* Reset the console, but only if this is really init and we
1665 * are freshly booted */
1666 if (arg_system && arg_action == ACTION_RUN) {
1667
1668 /* If we are init, we connect stdin/stdout/stderr to
1669 * /dev/null and make sure we don't have a controlling
1670 * tty. */
1671 release_terminal();
1672
1673 if (getpid() == 1 && !skip_setup)
1674 console_setup();
1675 }
1676
1677 /* Open the logging devices, if possible and necessary */
1678 log_open();
1679
1680 if (arg_show_status == _SHOW_STATUS_UNSET)
1681 arg_show_status = SHOW_STATUS_YES;
1682
1683 /* Make sure we leave a core dump without panicing the
1684 * kernel. */
1685 if (getpid() == 1) {
1686 install_crash_handler();
1687
1688 r = mount_cgroup_controllers(arg_join_controllers);
1689 if (r < 0)
1690 goto finish;
1691 }
1692
1693 if (arg_system) {
1694 int v;
1695
1696 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1697 arg_action == ACTION_TEST ? "test " : "" );
1698
1699 v = detect_virtualization();
1700 if (v > 0)
1701 log_info("Detected virtualization %s.", virtualization_to_string(v));
1702
1703 write_container_id();
1704
1705 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1706
1707 if (in_initrd())
1708 log_info("Running in initial RAM disk.");
1709
1710 /* Let's check whether /etc is already populated. We
1711 * don't actually really check for that, but use
1712 * /etc/machine-id as flag file. This allows container
1713 * managers and installers to provision a couple of
1714 * files already. If the container manager wants to
1715 * provision the machine ID itself it should pass
1716 * $container_uuid to PID 1. */
1717
1718 empty_etc = access("/etc/machine-id", F_OK) < 0;
1719 if (empty_etc)
1720 log_info("Running with unpopulated /etc.");
1721 } else {
1722 _cleanup_free_ char *t;
1723
1724 t = uid_to_name(getuid());
1725 log_debug(PACKAGE_STRING " running in %suser mode for user "UID_FMT"/%s. (" SYSTEMD_FEATURES ")",
1726 arg_action == ACTION_TEST ? " test" : "", getuid(), t);
1727 }
1728
1729 if (arg_system && !skip_setup) {
1730 if (arg_show_status > 0)
1731 status_welcome();
1732
1733 hostname_setup();
1734 machine_id_setup(NULL, arg_machine_id, NULL);
1735 loopback_setup();
1736 bump_unix_max_dgram_qlen();
1737
1738 test_usr();
1739 }
1740
1741 if (arg_system && arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY)
1742 watchdog_set_timeout(&arg_runtime_watchdog);
1743
1744 if (arg_timer_slack_nsec != NSEC_INFINITY)
1745 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1746 log_error_errno(errno, "Failed to adjust timer slack: %m");
1747
1748 if (!cap_test_all(arg_capability_bounding_set)) {
1749 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
1750 if (r < 0) {
1751 log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
1752 error_message = "Failed to drop capability bounding set of usermode helpers";
1753 goto finish;
1754 }
1755 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
1756 if (r < 0) {
1757 log_emergency_errno(r, "Failed to drop capability bounding set: %m");
1758 error_message = "Failed to drop capability bounding set";
1759 goto finish;
1760 }
1761 }
1762
1763 if (arg_syscall_archs) {
1764 r = enforce_syscall_archs(arg_syscall_archs);
1765 if (r < 0) {
1766 error_message = "Failed to set syscall architectures";
1767 goto finish;
1768 }
1769 }
1770
1771 if (!arg_system)
1772 /* Become reaper of our children */
1773 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
1774 log_warning_errno(errno, "Failed to make us a subreaper: %m");
1775
1776 if (arg_system) {
1777 (void) bump_rlimit_nofile(&saved_rlimit_nofile);
1778
1779 if (empty_etc) {
1780 r = unit_file_preset_all(UNIT_FILE_SYSTEM, 0, NULL, UNIT_FILE_PRESET_ENABLE_ONLY, NULL, 0);
1781 if (r < 0)
1782 log_full_errno(r == -EEXIST ? LOG_NOTICE : LOG_WARNING, r, "Failed to populate /etc with preset unit settings, ignoring: %m");
1783 else
1784 log_info("Populated /etc with preset unit settings.");
1785 }
1786 }
1787
1788 r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER, arg_action == ACTION_TEST, &m);
1789 if (r < 0) {
1790 log_emergency_errno(r, "Failed to allocate manager object: %m");
1791 error_message = "Failed to allocate manager object";
1792 goto finish;
1793 }
1794
1795 m->confirm_spawn = arg_confirm_spawn;
1796 m->runtime_watchdog = arg_runtime_watchdog;
1797 m->shutdown_watchdog = arg_shutdown_watchdog;
1798 m->userspace_timestamp = userspace_timestamp;
1799 m->kernel_timestamp = kernel_timestamp;
1800 m->initrd_timestamp = initrd_timestamp;
1801 m->security_start_timestamp = security_start_timestamp;
1802 m->security_finish_timestamp = security_finish_timestamp;
1803 m->cad_burst_action = arg_cad_burst_action;
1804
1805 manager_set_defaults(m);
1806 manager_set_show_status(m, arg_show_status);
1807 manager_set_first_boot(m, empty_etc);
1808
1809 /* Remember whether we should queue the default job */
1810 queue_default_job = !arg_serialization || arg_switched_root;
1811
1812 before_startup = now(CLOCK_MONOTONIC);
1813
1814 r = manager_startup(m, arg_serialization, fds);
1815 if (r < 0)
1816 log_error_errno(r, "Failed to fully start up daemon: %m");
1817
1818 /* This will close all file descriptors that were opened, but
1819 * not claimed by any unit. */
1820 fds = fdset_free(fds);
1821
1822 arg_serialization = safe_fclose(arg_serialization);
1823
1824 if (queue_default_job) {
1825 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
1826 Unit *target = NULL;
1827 Job *default_unit_job;
1828
1829 log_debug("Activating default unit: %s", arg_default_unit);
1830
1831 r = manager_load_unit(m, arg_default_unit, NULL, &error, &target);
1832 if (r < 0)
1833 log_error("Failed to load default target: %s", bus_error_message(&error, r));
1834 else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND)
1835 log_error_errno(target->load_error, "Failed to load default target: %m");
1836 else if (target->load_state == UNIT_MASKED)
1837 log_error("Default target masked.");
1838
1839 if (!target || target->load_state != UNIT_LOADED) {
1840 log_info("Trying to load rescue target...");
1841
1842 r = manager_load_unit(m, SPECIAL_RESCUE_TARGET, NULL, &error, &target);
1843 if (r < 0) {
1844 log_emergency("Failed to load rescue target: %s", bus_error_message(&error, r));
1845 error_message = "Failed to load rescue target";
1846 goto finish;
1847 } else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND) {
1848 log_emergency_errno(target->load_error, "Failed to load rescue target: %m");
1849 error_message = "Failed to load rescue target";
1850 goto finish;
1851 } else if (target->load_state == UNIT_MASKED) {
1852 log_emergency("Rescue target masked.");
1853 error_message = "Rescue target masked";
1854 goto finish;
1855 }
1856 }
1857
1858 assert(target->load_state == UNIT_LOADED);
1859
1860 if (arg_action == ACTION_TEST) {
1861 printf("-> By units:\n");
1862 manager_dump_units(m, stdout, "\t");
1863 }
1864
1865 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, &error, &default_unit_job);
1866 if (r == -EPERM) {
1867 log_debug("Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1868
1869 sd_bus_error_free(&error);
1870
1871 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, &error, &default_unit_job);
1872 if (r < 0) {
1873 log_emergency("Failed to start default target: %s", bus_error_message(&error, r));
1874 error_message = "Failed to start default target";
1875 goto finish;
1876 }
1877 } else if (r < 0) {
1878 log_emergency("Failed to isolate default target: %s", bus_error_message(&error, r));
1879 error_message = "Failed to isolate default target";
1880 goto finish;
1881 }
1882
1883 m->default_unit_job_id = default_unit_job->id;
1884
1885 after_startup = now(CLOCK_MONOTONIC);
1886 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
1887 "Loaded units and determined initial transaction in %s.",
1888 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
1889
1890 if (arg_action == ACTION_TEST) {
1891 printf("-> By jobs:\n");
1892 manager_dump_jobs(m, stdout, "\t");
1893 retval = EXIT_SUCCESS;
1894 goto finish;
1895 }
1896 }
1897
1898 for (;;) {
1899 r = manager_loop(m);
1900 if (r < 0) {
1901 log_emergency_errno(r, "Failed to run main loop: %m");
1902 error_message = "Failed to run main loop";
1903 goto finish;
1904 }
1905
1906 switch (m->exit_code) {
1907
1908 case MANAGER_RELOAD:
1909 log_info("Reloading.");
1910
1911 r = parse_config_file();
1912 if (r < 0)
1913 log_error("Failed to parse config file.");
1914
1915 manager_set_defaults(m);
1916
1917 r = manager_reload(m);
1918 if (r < 0)
1919 log_error_errno(r, "Failed to reload: %m");
1920 break;
1921
1922 case MANAGER_REEXECUTE:
1923
1924 if (prepare_reexecute(m, &arg_serialization, &fds, false) < 0) {
1925 error_message = "Failed to prepare for reexecution";
1926 goto finish;
1927 }
1928
1929 reexecute = true;
1930 log_notice("Reexecuting.");
1931 goto finish;
1932
1933 case MANAGER_SWITCH_ROOT:
1934 /* Steal the switch root parameters */
1935 switch_root_dir = m->switch_root;
1936 switch_root_init = m->switch_root_init;
1937 m->switch_root = m->switch_root_init = NULL;
1938
1939 if (!switch_root_init)
1940 if (prepare_reexecute(m, &arg_serialization, &fds, true) < 0) {
1941 error_message = "Failed to prepare for reexecution";
1942 goto finish;
1943 }
1944
1945 reexecute = true;
1946 log_notice("Switching root.");
1947 goto finish;
1948
1949 case MANAGER_EXIT:
1950 retval = m->return_value;
1951
1952 if (MANAGER_IS_USER(m)) {
1953 log_debug("Exit.");
1954 goto finish;
1955 }
1956
1957 /* fallthrough */
1958 case MANAGER_REBOOT:
1959 case MANAGER_POWEROFF:
1960 case MANAGER_HALT:
1961 case MANAGER_KEXEC: {
1962 static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
1963 [MANAGER_EXIT] = "exit",
1964 [MANAGER_REBOOT] = "reboot",
1965 [MANAGER_POWEROFF] = "poweroff",
1966 [MANAGER_HALT] = "halt",
1967 [MANAGER_KEXEC] = "kexec"
1968 };
1969
1970 assert_se(shutdown_verb = table[m->exit_code]);
1971 arm_reboot_watchdog = m->exit_code == MANAGER_REBOOT;
1972
1973 log_notice("Shutting down.");
1974 goto finish;
1975 }
1976
1977 default:
1978 assert_not_reached("Unknown exit code.");
1979 }
1980 }
1981
1982 finish:
1983 pager_close();
1984
1985 if (m)
1986 arg_shutdown_watchdog = m->shutdown_watchdog;
1987
1988 m = manager_free(m);
1989
1990 for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++)
1991 arg_default_rlimit[j] = mfree(arg_default_rlimit[j]);
1992
1993 arg_default_unit = mfree(arg_default_unit);
1994 arg_join_controllers = strv_free_free(arg_join_controllers);
1995 arg_default_environment = strv_free(arg_default_environment);
1996 arg_syscall_archs = set_free(arg_syscall_archs);
1997
1998 mac_selinux_finish();
1999
2000 if (reexecute) {
2001 const char **args;
2002 unsigned i, args_size;
2003
2004 /* Close and disarm the watchdog, so that the new
2005 * instance can reinitialize it, but doesn't get
2006 * rebooted while we do that */
2007 watchdog_close(true);
2008
2009 /* Reset the RLIMIT_NOFILE to the kernel default, so
2010 * that the new systemd can pass the kernel default to
2011 * its child processes */
2012 if (saved_rlimit_nofile.rlim_cur > 0)
2013 (void) setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile);
2014
2015 if (switch_root_dir) {
2016 /* Kill all remaining processes from the
2017 * initrd, but don't wait for them, so that we
2018 * can handle the SIGCHLD for them after
2019 * deserializing. */
2020 broadcast_signal(SIGTERM, false, true);
2021
2022 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
2023 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
2024 if (r < 0)
2025 log_error_errno(r, "Failed to switch root, trying to continue: %m");
2026 }
2027
2028 args_size = MAX(6, argc+1);
2029 args = newa(const char*, args_size);
2030
2031 if (!switch_root_init) {
2032 char sfd[DECIMAL_STR_MAX(int) + 1];
2033
2034 /* First try to spawn ourselves with the right
2035 * path, and with full serialization. We do
2036 * this only if the user didn't specify an
2037 * explicit init to spawn. */
2038
2039 assert(arg_serialization);
2040 assert(fds);
2041
2042 xsprintf(sfd, "%i", fileno(arg_serialization));
2043
2044 i = 0;
2045 args[i++] = SYSTEMD_BINARY_PATH;
2046 if (switch_root_dir)
2047 args[i++] = "--switched-root";
2048 args[i++] = arg_system ? "--system" : "--user";
2049 args[i++] = "--deserialize";
2050 args[i++] = sfd;
2051 args[i++] = NULL;
2052
2053 assert(i <= args_size);
2054
2055 /*
2056 * We want valgrind to print its memory usage summary before reexecution.
2057 * Valgrind won't do this is on its own on exec(), but it will do it on exit().
2058 * Hence, to ensure we get a summary here, fork() off a child, let it exit() cleanly,
2059 * so that it prints the summary, and wait() for it in the parent, before proceeding into the exec().
2060 */
2061 valgrind_summary_hack();
2062
2063 (void) execv(args[0], (char* const*) args);
2064 }
2065
2066 /* Try the fallback, if there is any, without any
2067 * serialization. We pass the original argv[] and
2068 * envp[]. (Well, modulo the ordering changes due to
2069 * getopt() in argv[], and some cleanups in envp[],
2070 * but let's hope that doesn't matter.) */
2071
2072 arg_serialization = safe_fclose(arg_serialization);
2073 fds = fdset_free(fds);
2074
2075 /* Reopen the console */
2076 (void) make_console_stdio();
2077
2078 for (j = 1, i = 1; j < (unsigned) argc; j++)
2079 args[i++] = argv[j];
2080 args[i++] = NULL;
2081 assert(i <= args_size);
2082
2083 /* Reenable any blocked signals, especially important
2084 * if we switch from initial ramdisk to init=... */
2085 (void) reset_all_signal_handlers();
2086 (void) reset_signal_mask();
2087
2088 if (switch_root_init) {
2089 args[0] = switch_root_init;
2090 (void) execv(args[0], (char* const*) args);
2091 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
2092 }
2093
2094 args[0] = "/sbin/init";
2095 (void) execv(args[0], (char* const*) args);
2096
2097 if (errno == ENOENT) {
2098 log_warning("No /sbin/init, trying fallback");
2099
2100 args[0] = "/bin/sh";
2101 args[1] = NULL;
2102 (void) execv(args[0], (char* const*) args);
2103 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
2104 } else
2105 log_warning_errno(errno, "Failed to execute /sbin/init, giving up: %m");
2106 }
2107
2108 arg_serialization = safe_fclose(arg_serialization);
2109 fds = fdset_free(fds);
2110
2111 #ifdef HAVE_VALGRIND_VALGRIND_H
2112 /* If we are PID 1 and running under valgrind, then let's exit
2113 * here explicitly. valgrind will only generate nice output on
2114 * exit(), not on exec(), hence let's do the former not the
2115 * latter here. */
2116 if (getpid() == 1 && RUNNING_ON_VALGRIND)
2117 return 0;
2118 #endif
2119
2120 if (shutdown_verb) {
2121 char log_level[DECIMAL_STR_MAX(int) + 1];
2122 char exit_code[DECIMAL_STR_MAX(uint8_t) + 1];
2123 const char* command_line[11] = {
2124 SYSTEMD_SHUTDOWN_BINARY_PATH,
2125 shutdown_verb,
2126 "--log-level", log_level,
2127 "--log-target",
2128 };
2129 unsigned pos = 5;
2130 _cleanup_strv_free_ char **env_block = NULL;
2131
2132 assert(command_line[pos] == NULL);
2133 env_block = strv_copy(environ);
2134
2135 xsprintf(log_level, "%d", log_get_max_level());
2136
2137 switch (log_get_target()) {
2138
2139 case LOG_TARGET_KMSG:
2140 case LOG_TARGET_JOURNAL_OR_KMSG:
2141 case LOG_TARGET_SYSLOG_OR_KMSG:
2142 command_line[pos++] = "kmsg";
2143 break;
2144
2145 case LOG_TARGET_NULL:
2146 command_line[pos++] = "null";
2147 break;
2148
2149 case LOG_TARGET_CONSOLE:
2150 default:
2151 command_line[pos++] = "console";
2152 break;
2153 };
2154
2155 if (log_get_show_color())
2156 command_line[pos++] = "--log-color";
2157
2158 if (log_get_show_location())
2159 command_line[pos++] = "--log-location";
2160
2161 if (streq(shutdown_verb, "exit")) {
2162 command_line[pos++] = "--exit-code";
2163 command_line[pos++] = exit_code;
2164 xsprintf(exit_code, "%d", retval);
2165 }
2166
2167 assert(pos < ELEMENTSOF(command_line));
2168
2169 if (arm_reboot_watchdog && arg_shutdown_watchdog > 0 && arg_shutdown_watchdog != USEC_INFINITY) {
2170 char *e;
2171
2172 /* If we reboot let's set the shutdown
2173 * watchdog and tell the shutdown binary to
2174 * repeatedly ping it */
2175 r = watchdog_set_timeout(&arg_shutdown_watchdog);
2176 watchdog_close(r < 0);
2177
2178 /* Tell the binary how often to ping, ignore failure */
2179 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
2180 (void) strv_push(&env_block, e);
2181 } else
2182 watchdog_close(true);
2183
2184 /* Avoid the creation of new processes forked by the
2185 * kernel; at this point, we will not listen to the
2186 * signals anyway */
2187 if (detect_container() <= 0)
2188 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
2189
2190 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
2191 log_error_errno(errno, "Failed to execute shutdown binary, %s: %m",
2192 getpid() == 1 ? "freezing" : "quitting");
2193 }
2194
2195 if (getpid() == 1) {
2196 if (error_message)
2197 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2198 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
2199 "%s, freezing.", error_message);
2200 freeze_or_reboot();
2201 }
2202
2203 return retval;
2204 }
Unnamed repository; edit this file 'description' to name the repository.