]> 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 #3984 from poettering/refcnt
[thirdparty/systemd.git] / src / core / main.c
1 /***
2 This file is part of systemd.
3
4 Copyright 2010 Lennart Poettering
5
6 systemd is free software; you can redistribute it and/or modify it
7 under the terms of the GNU Lesser General Public License as published by
8 the Free Software Foundation; either version 2.1 of the License, or
9 (at your option) any later version.
10
11 systemd is distributed in the hope that it will be useful, but
12 WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
15
16 You should have received a copy of the GNU Lesser General Public License
17 along with systemd; If not, see <http://www.gnu.org/licenses/>.
18 ***/
19
20 #include <errno.h>
21 #include <fcntl.h>
22 #include <getopt.h>
23 #include <signal.h>
24 #include <stdio.h>
25 #include <string.h>
26 #include <sys/mount.h>
27 #include <sys/prctl.h>
28 #include <sys/reboot.h>
29 #include <sys/stat.h>
30 #include <unistd.h>
31 #ifdef HAVE_SECCOMP
32 #include <seccomp.h>
33 #endif
34 #ifdef HAVE_VALGRIND_VALGRIND_H
35 #include <valgrind/valgrind.h>
36 #endif
37
38 #include "sd-bus.h"
39 #include "sd-daemon.h"
40
41 #include "alloc-util.h"
42 #include "architecture.h"
43 #include "build.h"
44 #include "bus-error.h"
45 #include "bus-util.h"
46 #include "capability-util.h"
47 #include "clock-util.h"
48 #include "conf-parser.h"
49 #include "cpu-set-util.h"
50 #include "dbus-manager.h"
51 #include "def.h"
52 #include "env-util.h"
53 #include "fd-util.h"
54 #include "fdset.h"
55 #include "fileio.h"
56 #include "formats-util.h"
57 #include "fs-util.h"
58 #include "hostname-setup.h"
59 #include "ima-setup.h"
60 #include "killall.h"
61 #include "kmod-setup.h"
62 #include "load-fragment.h"
63 #include "log.h"
64 #include "loopback-setup.h"
65 #include "machine-id-setup.h"
66 #include "manager.h"
67 #include "missing.h"
68 #include "mount-setup.h"
69 #include "pager.h"
70 #include "parse-util.h"
71 #include "proc-cmdline.h"
72 #include "process-util.h"
73 #include "raw-clone.h"
74 #include "rlimit-util.h"
75 #ifdef HAVE_SECCOMP
76 #include "seccomp-util.h"
77 #endif
78 #include "selinux-setup.h"
79 #include "selinux-util.h"
80 #include "signal-util.h"
81 #include "smack-setup.h"
82 #include "special.h"
83 #include "stat-util.h"
84 #include "stdio-util.h"
85 #include "strv.h"
86 #include "switch-root.h"
87 #include "terminal-util.h"
88 #include "umask-util.h"
89 #include "user-util.h"
90 #include "virt.h"
91 #include "watchdog.h"
92
93 static enum {
94 ACTION_RUN,
95 ACTION_HELP,
96 ACTION_VERSION,
97 ACTION_TEST,
98 ACTION_DUMP_CONFIGURATION_ITEMS
99 } arg_action = ACTION_RUN;
100 static char *arg_default_unit = NULL;
101 static bool arg_system = false;
102 static bool arg_dump_core = true;
103 static int arg_crash_chvt = -1;
104 static bool arg_crash_shell = false;
105 static bool arg_crash_reboot = false;
106 static bool arg_confirm_spawn = false;
107 static ShowStatus arg_show_status = _SHOW_STATUS_UNSET;
108 static bool arg_switched_root = false;
109 static bool arg_no_pager = false;
110 static char ***arg_join_controllers = NULL;
111 static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
112 static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
113 static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
114 static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
115 static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
116 static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
117 static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
118 static usec_t arg_runtime_watchdog = 0;
119 static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
120 static char **arg_default_environment = NULL;
121 static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {};
122 static uint64_t arg_capability_bounding_set = CAP_ALL;
123 static nsec_t arg_timer_slack_nsec = NSEC_INFINITY;
124 static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
125 static Set* arg_syscall_archs = NULL;
126 static FILE* arg_serialization = NULL;
127 static bool arg_default_cpu_accounting = false;
128 static bool arg_default_io_accounting = false;
129 static bool arg_default_blockio_accounting = false;
130 static bool arg_default_memory_accounting = false;
131 static bool arg_default_tasks_accounting = true;
132 static uint64_t arg_default_tasks_max = UINT64_MAX;
133 static sd_id128_t arg_machine_id = {};
134
135 noreturn static void freeze_or_reboot(void) {
136
137 if (arg_crash_reboot) {
138 log_notice("Rebooting in 10s...");
139 (void) sleep(10);
140
141 log_notice("Rebooting now...");
142 (void) reboot(RB_AUTOBOOT);
143 log_emergency_errno(errno, "Failed to reboot: %m");
144 }
145
146 log_emergency("Freezing execution.");
147 freeze();
148 }
149
150 noreturn static void crash(int sig) {
151 struct sigaction sa;
152 pid_t pid;
153
154 if (getpid() != 1)
155 /* Pass this on immediately, if this is not PID 1 */
156 (void) raise(sig);
157 else if (!arg_dump_core)
158 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
159 else {
160 sa = (struct sigaction) {
161 .sa_handler = nop_signal_handler,
162 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
163 };
164
165 /* We want to wait for the core process, hence let's enable SIGCHLD */
166 (void) sigaction(SIGCHLD, &sa, NULL);
167
168 pid = raw_clone(SIGCHLD);
169 if (pid < 0)
170 log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
171 else if (pid == 0) {
172 /* Enable default signal handler for core dump */
173
174 sa = (struct sigaction) {
175 .sa_handler = SIG_DFL,
176 };
177 (void) sigaction(sig, &sa, NULL);
178
179 /* Don't limit the coredump size */
180 (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
181
182 /* Just to be sure... */
183 (void) chdir("/");
184
185 /* Raise the signal again */
186 pid = raw_getpid();
187 (void) kill(pid, sig); /* raise() would kill the parent */
188
189 assert_not_reached("We shouldn't be here...");
190 _exit(EXIT_FAILURE);
191 } else {
192 siginfo_t status;
193 int r;
194
195 /* Order things nicely. */
196 r = wait_for_terminate(pid, &status);
197 if (r < 0)
198 log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
199 else if (status.si_code != CLD_DUMPED)
200 log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
201 signal_to_string(sig),
202 pid, sigchld_code_to_string(status.si_code),
203 status.si_status,
204 strna(status.si_code == CLD_EXITED
205 ? exit_status_to_string(status.si_status, EXIT_STATUS_FULL)
206 : signal_to_string(status.si_status)));
207 else
208 log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
209 }
210 }
211
212 if (arg_crash_chvt >= 0)
213 (void) chvt(arg_crash_chvt);
214
215 sa = (struct sigaction) {
216 .sa_handler = SIG_IGN,
217 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
218 };
219
220 /* Let the kernel reap children for us */
221 (void) sigaction(SIGCHLD, &sa, NULL);
222
223 if (arg_crash_shell) {
224 log_notice("Executing crash shell in 10s...");
225 (void) sleep(10);
226
227 pid = raw_clone(SIGCHLD);
228 if (pid < 0)
229 log_emergency_errno(errno, "Failed to fork off crash shell: %m");
230 else if (pid == 0) {
231 (void) setsid();
232 (void) make_console_stdio();
233 (void) execle("/bin/sh", "/bin/sh", NULL, environ);
234
235 log_emergency_errno(errno, "execle() failed: %m");
236 _exit(EXIT_FAILURE);
237 } else {
238 log_info("Spawned crash shell as PID "PID_FMT".", pid);
239 (void) wait_for_terminate(pid, NULL);
240 }
241 }
242
243 freeze_or_reboot();
244 }
245
246 static void install_crash_handler(void) {
247 static const struct sigaction sa = {
248 .sa_handler = crash,
249 .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
250 };
251 int r;
252
253 /* We ignore the return value here, since, we don't mind if we
254 * cannot set up a crash handler */
255 r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
256 if (r < 0)
257 log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
258 }
259
260 static int console_setup(void) {
261 _cleanup_close_ int tty_fd = -1;
262 int r;
263
264 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
265 if (tty_fd < 0)
266 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
267
268 /* We don't want to force text mode. plymouth may be showing
269 * pictures already from initrd. */
270 r = reset_terminal_fd(tty_fd, false);
271 if (r < 0)
272 return log_error_errno(r, "Failed to reset /dev/console: %m");
273
274 return 0;
275 }
276
277 static int parse_crash_chvt(const char *value) {
278 int b;
279
280 if (safe_atoi(value, &arg_crash_chvt) >= 0)
281 return 0;
282
283 b = parse_boolean(value);
284 if (b < 0)
285 return b;
286
287 if (b > 0)
288 arg_crash_chvt = 0; /* switch to where kmsg goes */
289 else
290 arg_crash_chvt = -1; /* turn off switching */
291
292 return 0;
293 }
294
295 static int set_machine_id(const char *m) {
296 sd_id128_t t;
297 assert(m);
298
299 if (sd_id128_from_string(m, &t) < 0)
300 return -EINVAL;
301
302 if (sd_id128_is_null(t))
303 return -EINVAL;
304
305 arg_machine_id = t;
306 return 0;
307 }
308
309 static int parse_proc_cmdline_item(const char *key, const char *value) {
310
311 int r;
312
313 assert(key);
314
315 if (streq(key, "systemd.unit") && value) {
316
317 if (!in_initrd())
318 return free_and_strdup(&arg_default_unit, value);
319
320 } else if (streq(key, "rd.systemd.unit") && value) {
321
322 if (in_initrd())
323 return free_and_strdup(&arg_default_unit, value);
324
325 } else if (streq(key, "systemd.dump_core") && value) {
326
327 r = parse_boolean(value);
328 if (r < 0)
329 log_warning("Failed to parse dump core switch %s. Ignoring.", value);
330 else
331 arg_dump_core = r;
332
333 } else if (streq(key, "systemd.crash_chvt") && value) {
334
335 if (parse_crash_chvt(value) < 0)
336 log_warning("Failed to parse crash chvt switch %s. Ignoring.", value);
337
338 } else if (streq(key, "systemd.crash_shell") && value) {
339
340 r = parse_boolean(value);
341 if (r < 0)
342 log_warning("Failed to parse crash shell switch %s. Ignoring.", value);
343 else
344 arg_crash_shell = r;
345
346 } else if (streq(key, "systemd.crash_reboot") && value) {
347
348 r = parse_boolean(value);
349 if (r < 0)
350 log_warning("Failed to parse crash reboot switch %s. Ignoring.", value);
351 else
352 arg_crash_reboot = r;
353
354 } else if (streq(key, "systemd.confirm_spawn") && value) {
355
356 r = parse_boolean(value);
357 if (r < 0)
358 log_warning("Failed to parse confirm spawn switch %s. Ignoring.", value);
359 else
360 arg_confirm_spawn = r;
361
362 } else if (streq(key, "systemd.show_status") && value) {
363
364 r = parse_show_status(value, &arg_show_status);
365 if (r < 0)
366 log_warning("Failed to parse show status switch %s. Ignoring.", value);
367
368 } else if (streq(key, "systemd.default_standard_output") && value) {
369
370 r = exec_output_from_string(value);
371 if (r < 0)
372 log_warning("Failed to parse default standard output switch %s. Ignoring.", value);
373 else
374 arg_default_std_output = r;
375
376 } else if (streq(key, "systemd.default_standard_error") && value) {
377
378 r = exec_output_from_string(value);
379 if (r < 0)
380 log_warning("Failed to parse default standard error switch %s. Ignoring.", value);
381 else
382 arg_default_std_error = r;
383
384 } else if (streq(key, "systemd.setenv") && value) {
385
386 if (env_assignment_is_valid(value)) {
387 char **env;
388
389 env = strv_env_set(arg_default_environment, value);
390 if (env)
391 arg_default_environment = env;
392 else
393 log_warning_errno(ENOMEM, "Setting environment variable '%s' failed, ignoring: %m", value);
394 } else
395 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
396
397 } else if (streq(key, "systemd.machine_id") && value) {
398
399 r = set_machine_id(value);
400 if (r < 0)
401 log_warning("MachineID '%s' is not valid. Ignoring.", value);
402
403 } else if (streq(key, "quiet") && !value) {
404
405 if (arg_show_status == _SHOW_STATUS_UNSET)
406 arg_show_status = SHOW_STATUS_AUTO;
407
408 } else if (streq(key, "debug") && !value) {
409
410 /* Note that log_parse_environment() handles 'debug'
411 * too, and sets the log level to LOG_DEBUG. */
412
413 if (detect_container() > 0)
414 log_set_target(LOG_TARGET_CONSOLE);
415
416 } else if (!value) {
417 const char *target;
418
419 /* SysV compatibility */
420 target = runlevel_to_target(key);
421 if (target)
422 return free_and_strdup(&arg_default_unit, target);
423
424 } else if (streq(key, "systemd.default_timeout_start_sec") && value) {
425
426 r = parse_sec(value, &arg_default_timeout_start_usec);
427 if (r < 0)
428 log_warning_errno(r, "Failed to parse default start timeout: %s, ignoring.", value);
429
430 if (arg_default_timeout_start_usec <= 0)
431 arg_default_timeout_start_usec = USEC_INFINITY;
432 }
433
434 return 0;
435 }
436
437 #define DEFINE_SETTER(name, func, descr) \
438 static int name(const char *unit, \
439 const char *filename, \
440 unsigned line, \
441 const char *section, \
442 unsigned section_line, \
443 const char *lvalue, \
444 int ltype, \
445 const char *rvalue, \
446 void *data, \
447 void *userdata) { \
448 \
449 int r; \
450 \
451 assert(filename); \
452 assert(lvalue); \
453 assert(rvalue); \
454 \
455 r = func(rvalue); \
456 if (r < 0) \
457 log_syntax(unit, LOG_ERR, filename, line, r, \
458 "Invalid " descr "'%s': %m", \
459 rvalue); \
460 \
461 return 0; \
462 }
463
464 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level")
465 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target")
466 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" )
467 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location")
468
469 static int config_parse_cpu_affinity2(
470 const char *unit,
471 const char *filename,
472 unsigned line,
473 const char *section,
474 unsigned section_line,
475 const char *lvalue,
476 int ltype,
477 const char *rvalue,
478 void *data,
479 void *userdata) {
480
481 _cleanup_cpu_free_ cpu_set_t *c = NULL;
482 int ncpus;
483
484 ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue);
485 if (ncpus < 0)
486 return ncpus;
487
488 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
489 log_warning("Failed to set CPU affinity: %m");
490
491 return 0;
492 }
493
494 static int config_parse_show_status(
495 const char* unit,
496 const char *filename,
497 unsigned line,
498 const char *section,
499 unsigned section_line,
500 const char *lvalue,
501 int ltype,
502 const char *rvalue,
503 void *data,
504 void *userdata) {
505
506 int k;
507 ShowStatus *b = data;
508
509 assert(filename);
510 assert(lvalue);
511 assert(rvalue);
512 assert(data);
513
514 k = parse_show_status(rvalue, b);
515 if (k < 0) {
516 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
517 return 0;
518 }
519
520 return 0;
521 }
522
523 static int config_parse_crash_chvt(
524 const char* unit,
525 const char *filename,
526 unsigned line,
527 const char *section,
528 unsigned section_line,
529 const char *lvalue,
530 int ltype,
531 const char *rvalue,
532 void *data,
533 void *userdata) {
534
535 int r;
536
537 assert(filename);
538 assert(lvalue);
539 assert(rvalue);
540
541 r = parse_crash_chvt(rvalue);
542 if (r < 0) {
543 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
544 return 0;
545 }
546
547 return 0;
548 }
549
550 static int config_parse_join_controllers(const char *unit,
551 const char *filename,
552 unsigned line,
553 const char *section,
554 unsigned section_line,
555 const char *lvalue,
556 int ltype,
557 const char *rvalue,
558 void *data,
559 void *userdata) {
560
561 const char *whole_rvalue = rvalue;
562 unsigned n = 0;
563
564 assert(filename);
565 assert(lvalue);
566 assert(rvalue);
567
568 arg_join_controllers = strv_free_free(arg_join_controllers);
569
570 for (;;) {
571 _cleanup_free_ char *word = NULL;
572 char **l;
573 int r;
574
575 r = extract_first_word(&rvalue, &word, WHITESPACE, EXTRACT_QUOTES);
576 if (r < 0) {
577 log_syntax(unit, LOG_ERR, filename, line, r, "Invalid value for %s: %s", lvalue, whole_rvalue);
578 return r;
579 }
580 if (r == 0)
581 break;
582
583 l = strv_split(word, ",");
584 if (!l)
585 return log_oom();
586 strv_uniq(l);
587
588 if (strv_length(l) <= 1) {
589 strv_free(l);
590 continue;
591 }
592
593 if (!arg_join_controllers) {
594 arg_join_controllers = new(char**, 2);
595 if (!arg_join_controllers) {
596 strv_free(l);
597 return log_oom();
598 }
599
600 arg_join_controllers[0] = l;
601 arg_join_controllers[1] = NULL;
602
603 n = 1;
604 } else {
605 char ***a;
606 char ***t;
607
608 t = new0(char**, n+2);
609 if (!t) {
610 strv_free(l);
611 return log_oom();
612 }
613
614 n = 0;
615
616 for (a = arg_join_controllers; *a; a++) {
617
618 if (strv_overlap(*a, l)) {
619 if (strv_extend_strv(&l, *a, false) < 0) {
620 strv_free(l);
621 strv_free_free(t);
622 return log_oom();
623 }
624
625 } else {
626 char **c;
627
628 c = strv_copy(*a);
629 if (!c) {
630 strv_free(l);
631 strv_free_free(t);
632 return log_oom();
633 }
634
635 t[n++] = c;
636 }
637 }
638
639 t[n++] = strv_uniq(l);
640
641 strv_free_free(arg_join_controllers);
642 arg_join_controllers = t;
643 }
644 }
645 if (!isempty(rvalue))
646 log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring.");
647
648 return 0;
649 }
650
651 static int parse_config_file(void) {
652
653 const ConfigTableItem items[] = {
654 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
655 { "Manager", "LogTarget", config_parse_target, 0, NULL },
656 { "Manager", "LogColor", config_parse_color, 0, NULL },
657 { "Manager", "LogLocation", config_parse_location, 0, NULL },
658 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
659 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL },
660 { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL },
661 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
662 { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
663 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
664 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
665 { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
666 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
667 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
668 { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
669 #ifdef HAVE_SECCOMP
670 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
671 #endif
672 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
673 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
674 { "Manager", "DefaultStandardOutput", config_parse_output, 0, &arg_default_std_output },
675 { "Manager", "DefaultStandardError", config_parse_output, 0, &arg_default_std_error },
676 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
677 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
678 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
679 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */
680 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec, 0, &arg_default_start_limit_interval },
681 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
682 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
683 { "Manager", "DefaultLimitCPU", config_parse_limit, RLIMIT_CPU, arg_default_rlimit },
684 { "Manager", "DefaultLimitFSIZE", config_parse_limit, RLIMIT_FSIZE, arg_default_rlimit },
685 { "Manager", "DefaultLimitDATA", config_parse_limit, RLIMIT_DATA, arg_default_rlimit },
686 { "Manager", "DefaultLimitSTACK", config_parse_limit, RLIMIT_STACK, arg_default_rlimit },
687 { "Manager", "DefaultLimitCORE", config_parse_limit, RLIMIT_CORE, arg_default_rlimit },
688 { "Manager", "DefaultLimitRSS", config_parse_limit, RLIMIT_RSS, arg_default_rlimit },
689 { "Manager", "DefaultLimitNOFILE", config_parse_limit, RLIMIT_NOFILE, arg_default_rlimit },
690 { "Manager", "DefaultLimitAS", config_parse_limit, RLIMIT_AS, arg_default_rlimit },
691 { "Manager", "DefaultLimitNPROC", config_parse_limit, RLIMIT_NPROC, arg_default_rlimit },
692 { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, RLIMIT_MEMLOCK, arg_default_rlimit },
693 { "Manager", "DefaultLimitLOCKS", config_parse_limit, RLIMIT_LOCKS, arg_default_rlimit },
694 { "Manager", "DefaultLimitSIGPENDING", config_parse_limit, RLIMIT_SIGPENDING, arg_default_rlimit },
695 { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, RLIMIT_MSGQUEUE, arg_default_rlimit },
696 { "Manager", "DefaultLimitNICE", config_parse_limit, RLIMIT_NICE, arg_default_rlimit },
697 { "Manager", "DefaultLimitRTPRIO", config_parse_limit, RLIMIT_RTPRIO, arg_default_rlimit },
698 { "Manager", "DefaultLimitRTTIME", config_parse_limit, RLIMIT_RTTIME, arg_default_rlimit },
699 { "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_default_cpu_accounting },
700 { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting },
701 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
702 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
703 { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
704 { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max },
705 {}
706 };
707
708 const char *fn, *conf_dirs_nulstr;
709
710 fn = arg_system ?
711 PKGSYSCONFDIR "/system.conf" :
712 PKGSYSCONFDIR "/user.conf";
713
714 conf_dirs_nulstr = arg_system ?
715 CONF_PATHS_NULSTR("systemd/system.conf.d") :
716 CONF_PATHS_NULSTR("systemd/user.conf.d");
717
718 config_parse_many(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, false, NULL);
719
720 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
721 * like everywhere else. */
722 if (arg_default_timeout_start_usec <= 0)
723 arg_default_timeout_start_usec = USEC_INFINITY;
724 if (arg_default_timeout_stop_usec <= 0)
725 arg_default_timeout_stop_usec = USEC_INFINITY;
726
727 return 0;
728 }
729
730 static void manager_set_defaults(Manager *m) {
731
732 assert(m);
733
734 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
735 m->default_std_output = arg_default_std_output;
736 m->default_std_error = arg_default_std_error;
737 m->default_timeout_start_usec = arg_default_timeout_start_usec;
738 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
739 m->default_restart_usec = arg_default_restart_usec;
740 m->default_start_limit_interval = arg_default_start_limit_interval;
741 m->default_start_limit_burst = arg_default_start_limit_burst;
742 m->default_cpu_accounting = arg_default_cpu_accounting;
743 m->default_io_accounting = arg_default_io_accounting;
744 m->default_blockio_accounting = arg_default_blockio_accounting;
745 m->default_memory_accounting = arg_default_memory_accounting;
746 m->default_tasks_accounting = arg_default_tasks_accounting;
747 m->default_tasks_max = arg_default_tasks_max;
748
749 manager_set_default_rlimits(m, arg_default_rlimit);
750 manager_environment_add(m, NULL, arg_default_environment);
751 }
752
753 static int parse_argv(int argc, char *argv[]) {
754
755 enum {
756 ARG_LOG_LEVEL = 0x100,
757 ARG_LOG_TARGET,
758 ARG_LOG_COLOR,
759 ARG_LOG_LOCATION,
760 ARG_UNIT,
761 ARG_SYSTEM,
762 ARG_USER,
763 ARG_TEST,
764 ARG_NO_PAGER,
765 ARG_VERSION,
766 ARG_DUMP_CONFIGURATION_ITEMS,
767 ARG_DUMP_CORE,
768 ARG_CRASH_CHVT,
769 ARG_CRASH_SHELL,
770 ARG_CRASH_REBOOT,
771 ARG_CONFIRM_SPAWN,
772 ARG_SHOW_STATUS,
773 ARG_DESERIALIZE,
774 ARG_SWITCHED_ROOT,
775 ARG_DEFAULT_STD_OUTPUT,
776 ARG_DEFAULT_STD_ERROR,
777 ARG_MACHINE_ID
778 };
779
780 static const struct option options[] = {
781 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
782 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
783 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
784 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
785 { "unit", required_argument, NULL, ARG_UNIT },
786 { "system", no_argument, NULL, ARG_SYSTEM },
787 { "user", no_argument, NULL, ARG_USER },
788 { "test", no_argument, NULL, ARG_TEST },
789 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
790 { "help", no_argument, NULL, 'h' },
791 { "version", no_argument, NULL, ARG_VERSION },
792 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
793 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
794 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
795 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
796 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
797 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
798 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
799 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
800 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
801 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
802 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
803 { "machine-id", required_argument, NULL, ARG_MACHINE_ID },
804 {}
805 };
806
807 int c, r;
808
809 assert(argc >= 1);
810 assert(argv);
811
812 if (getpid() == 1)
813 opterr = 0;
814
815 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
816
817 switch (c) {
818
819 case ARG_LOG_LEVEL:
820 r = log_set_max_level_from_string(optarg);
821 if (r < 0) {
822 log_error("Failed to parse log level %s.", optarg);
823 return r;
824 }
825
826 break;
827
828 case ARG_LOG_TARGET:
829 r = log_set_target_from_string(optarg);
830 if (r < 0) {
831 log_error("Failed to parse log target %s.", optarg);
832 return r;
833 }
834
835 break;
836
837 case ARG_LOG_COLOR:
838
839 if (optarg) {
840 r = log_show_color_from_string(optarg);
841 if (r < 0) {
842 log_error("Failed to parse log color setting %s.", optarg);
843 return r;
844 }
845 } else
846 log_show_color(true);
847
848 break;
849
850 case ARG_LOG_LOCATION:
851 if (optarg) {
852 r = log_show_location_from_string(optarg);
853 if (r < 0) {
854 log_error("Failed to parse log location setting %s.", optarg);
855 return r;
856 }
857 } else
858 log_show_location(true);
859
860 break;
861
862 case ARG_DEFAULT_STD_OUTPUT:
863 r = exec_output_from_string(optarg);
864 if (r < 0) {
865 log_error("Failed to parse default standard output setting %s.", optarg);
866 return r;
867 } else
868 arg_default_std_output = r;
869 break;
870
871 case ARG_DEFAULT_STD_ERROR:
872 r = exec_output_from_string(optarg);
873 if (r < 0) {
874 log_error("Failed to parse default standard error output setting %s.", optarg);
875 return r;
876 } else
877 arg_default_std_error = r;
878 break;
879
880 case ARG_UNIT:
881
882 r = free_and_strdup(&arg_default_unit, optarg);
883 if (r < 0)
884 return log_error_errno(r, "Failed to set default unit %s: %m", optarg);
885
886 break;
887
888 case ARG_SYSTEM:
889 arg_system = true;
890 break;
891
892 case ARG_USER:
893 arg_system = false;
894 break;
895
896 case ARG_TEST:
897 arg_action = ACTION_TEST;
898 break;
899
900 case ARG_NO_PAGER:
901 arg_no_pager = true;
902 break;
903
904 case ARG_VERSION:
905 arg_action = ACTION_VERSION;
906 break;
907
908 case ARG_DUMP_CONFIGURATION_ITEMS:
909 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
910 break;
911
912 case ARG_DUMP_CORE:
913 if (!optarg)
914 arg_dump_core = true;
915 else {
916 r = parse_boolean(optarg);
917 if (r < 0)
918 return log_error_errno(r, "Failed to parse dump core boolean: %s", optarg);
919 arg_dump_core = r;
920 }
921 break;
922
923 case ARG_CRASH_CHVT:
924 r = parse_crash_chvt(optarg);
925 if (r < 0)
926 return log_error_errno(r, "Failed to parse crash virtual terminal index: %s", optarg);
927 break;
928
929 case ARG_CRASH_SHELL:
930 if (!optarg)
931 arg_crash_shell = true;
932 else {
933 r = parse_boolean(optarg);
934 if (r < 0)
935 return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg);
936 arg_crash_shell = r;
937 }
938 break;
939
940 case ARG_CRASH_REBOOT:
941 if (!optarg)
942 arg_crash_reboot = true;
943 else {
944 r = parse_boolean(optarg);
945 if (r < 0)
946 return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg);
947 arg_crash_reboot = r;
948 }
949 break;
950
951 case ARG_CONFIRM_SPAWN:
952 r = optarg ? parse_boolean(optarg) : 1;
953 if (r < 0) {
954 log_error("Failed to parse confirm spawn boolean %s.", optarg);
955 return r;
956 }
957 arg_confirm_spawn = r;
958 break;
959
960 case ARG_SHOW_STATUS:
961 if (optarg) {
962 r = parse_show_status(optarg, &arg_show_status);
963 if (r < 0) {
964 log_error("Failed to parse show status boolean %s.", optarg);
965 return r;
966 }
967 } else
968 arg_show_status = SHOW_STATUS_YES;
969 break;
970
971 case ARG_DESERIALIZE: {
972 int fd;
973 FILE *f;
974
975 r = safe_atoi(optarg, &fd);
976 if (r < 0 || fd < 0) {
977 log_error("Failed to parse deserialize option %s.", optarg);
978 return -EINVAL;
979 }
980
981 (void) fd_cloexec(fd, true);
982
983 f = fdopen(fd, "r");
984 if (!f)
985 return log_error_errno(errno, "Failed to open serialization fd: %m");
986
987 safe_fclose(arg_serialization);
988 arg_serialization = f;
989
990 break;
991 }
992
993 case ARG_SWITCHED_ROOT:
994 arg_switched_root = true;
995 break;
996
997 case ARG_MACHINE_ID:
998 r = set_machine_id(optarg);
999 if (r < 0) {
1000 log_error("MachineID '%s' is not valid.", optarg);
1001 return r;
1002 }
1003 break;
1004
1005 case 'h':
1006 arg_action = ACTION_HELP;
1007 break;
1008
1009 case 'D':
1010 log_set_max_level(LOG_DEBUG);
1011 break;
1012
1013 case 'b':
1014 case 's':
1015 case 'z':
1016 /* Just to eat away the sysvinit kernel
1017 * cmdline args without getopt() error
1018 * messages that we'll parse in
1019 * parse_proc_cmdline_word() or ignore. */
1020
1021 case '?':
1022 if (getpid() != 1)
1023 return -EINVAL;
1024 else
1025 return 0;
1026
1027 default:
1028 assert_not_reached("Unhandled option code.");
1029 }
1030
1031 if (optind < argc && getpid() != 1) {
1032 /* Hmm, when we aren't run as init system
1033 * let's complain about excess arguments */
1034
1035 log_error("Excess arguments.");
1036 return -EINVAL;
1037 }
1038
1039 return 0;
1040 }
1041
1042 static int help(void) {
1043
1044 printf("%s [OPTIONS...]\n\n"
1045 "Starts up and maintains the system or user services.\n\n"
1046 " -h --help Show this help\n"
1047 " --test Determine startup sequence, dump it and exit\n"
1048 " --no-pager Do not pipe output into a pager\n"
1049 " --dump-configuration-items Dump understood unit configuration items\n"
1050 " --unit=UNIT Set default unit\n"
1051 " --system Run a system instance, even if PID != 1\n"
1052 " --user Run a user instance\n"
1053 " --dump-core[=BOOL] Dump core on crash\n"
1054 " --crash-vt=NR Change to specified VT on crash\n"
1055 " --crash-reboot[=BOOL] Reboot on crash\n"
1056 " --crash-shell[=BOOL] Run shell on crash\n"
1057 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1058 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1059 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1060 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1061 " --log-color[=BOOL] Highlight important log messages\n"
1062 " --log-location[=BOOL] Include code location in log messages\n"
1063 " --default-standard-output= Set default standard output for services\n"
1064 " --default-standard-error= Set default standard error output for services\n",
1065 program_invocation_short_name);
1066
1067 return 0;
1068 }
1069
1070 static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
1071 _cleanup_fdset_free_ FDSet *fds = NULL;
1072 _cleanup_fclose_ FILE *f = NULL;
1073 int r;
1074
1075 assert(m);
1076 assert(_f);
1077 assert(_fds);
1078
1079 r = manager_open_serialization(m, &f);
1080 if (r < 0)
1081 return log_error_errno(r, "Failed to create serialization file: %m");
1082
1083 /* Make sure nothing is really destructed when we shut down */
1084 m->n_reloading++;
1085 bus_manager_send_reloading(m, true);
1086
1087 fds = fdset_new();
1088 if (!fds)
1089 return log_oom();
1090
1091 r = manager_serialize(m, f, fds, switching_root);
1092 if (r < 0)
1093 return log_error_errno(r, "Failed to serialize state: %m");
1094
1095 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1096 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1097
1098 r = fd_cloexec(fileno(f), false);
1099 if (r < 0)
1100 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1101
1102 r = fdset_cloexec(fds, false);
1103 if (r < 0)
1104 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1105
1106 *_f = f;
1107 *_fds = fds;
1108
1109 f = NULL;
1110 fds = NULL;
1111
1112 return 0;
1113 }
1114
1115 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1116 struct rlimit nl;
1117 int r;
1118
1119 assert(saved_rlimit);
1120
1121 /* Save the original RLIMIT_NOFILE so that we can reset it
1122 * later when transitioning from the initrd to the main
1123 * systemd or suchlike. */
1124 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1125 return log_error_errno(errno, "Reading RLIMIT_NOFILE failed: %m");
1126
1127 /* Make sure forked processes get the default kernel setting */
1128 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1129 struct rlimit *rl;
1130
1131 rl = newdup(struct rlimit, saved_rlimit, 1);
1132 if (!rl)
1133 return log_oom();
1134
1135 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1136 }
1137
1138 /* Bump up the resource limit for ourselves substantially */
1139 nl.rlim_cur = nl.rlim_max = 64*1024;
1140 r = setrlimit_closest(RLIMIT_NOFILE, &nl);
1141 if (r < 0)
1142 return log_error_errno(r, "Setting RLIMIT_NOFILE failed: %m");
1143
1144 return 0;
1145 }
1146
1147 static void test_usr(void) {
1148
1149 /* Check that /usr is not a separate fs */
1150
1151 if (dir_is_empty("/usr") <= 0)
1152 return;
1153
1154 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1155 "Some things will probably break (sometimes even silently) in mysterious ways. "
1156 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1157 }
1158
1159 static int initialize_join_controllers(void) {
1160 /* By default, mount "cpu" + "cpuacct" together, and "net_cls"
1161 * + "net_prio". We'd like to add "cpuset" to the mix, but
1162 * "cpuset" doesn't really work for groups with no initialized
1163 * attributes. */
1164
1165 arg_join_controllers = new(char**, 3);
1166 if (!arg_join_controllers)
1167 return -ENOMEM;
1168
1169 arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL);
1170 if (!arg_join_controllers[0])
1171 goto oom;
1172
1173 arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL);
1174 if (!arg_join_controllers[1])
1175 goto oom;
1176
1177 arg_join_controllers[2] = NULL;
1178 return 0;
1179
1180 oom:
1181 arg_join_controllers = strv_free_free(arg_join_controllers);
1182 return -ENOMEM;
1183 }
1184
1185 static int enforce_syscall_archs(Set *archs) {
1186 #ifdef HAVE_SECCOMP
1187 scmp_filter_ctx *seccomp;
1188 Iterator i;
1189 void *id;
1190 int r;
1191
1192 if (!is_seccomp_available())
1193 return 0;
1194
1195 seccomp = seccomp_init(SCMP_ACT_ALLOW);
1196 if (!seccomp)
1197 return log_oom();
1198
1199 SET_FOREACH(id, arg_syscall_archs, i) {
1200 r = seccomp_arch_add(seccomp, PTR_TO_UINT32(id) - 1);
1201 if (r == -EEXIST)
1202 continue;
1203 if (r < 0) {
1204 log_error_errno(r, "Failed to add architecture to seccomp: %m");
1205 goto finish;
1206 }
1207 }
1208
1209 r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
1210 if (r < 0) {
1211 log_error_errno(r, "Failed to unset NO_NEW_PRIVS: %m");
1212 goto finish;
1213 }
1214
1215 r = seccomp_load(seccomp);
1216 if (r < 0)
1217 log_error_errno(r, "Failed to add install architecture seccomp: %m");
1218
1219 finish:
1220 seccomp_release(seccomp);
1221 return r;
1222 #else
1223 return 0;
1224 #endif
1225 }
1226
1227 static int status_welcome(void) {
1228 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1229 int r;
1230
1231 r = parse_env_file("/etc/os-release", NEWLINE,
1232 "PRETTY_NAME", &pretty_name,
1233 "ANSI_COLOR", &ansi_color,
1234 NULL);
1235 if (r == -ENOENT)
1236 r = parse_env_file("/usr/lib/os-release", NEWLINE,
1237 "PRETTY_NAME", &pretty_name,
1238 "ANSI_COLOR", &ansi_color,
1239 NULL);
1240
1241 if (r < 0 && r != -ENOENT)
1242 log_warning_errno(r, "Failed to read os-release file: %m");
1243
1244 if (log_get_show_color())
1245 return status_printf(NULL, false, false,
1246 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1247 isempty(ansi_color) ? "1" : ansi_color,
1248 isempty(pretty_name) ? "Linux" : pretty_name);
1249 else
1250 return status_printf(NULL, false, false,
1251 "\nWelcome to %s!\n",
1252 isempty(pretty_name) ? "Linux" : pretty_name);
1253 }
1254
1255 static int write_container_id(void) {
1256 const char *c;
1257 int r;
1258
1259 c = getenv("container");
1260 if (isempty(c))
1261 return 0;
1262
1263 RUN_WITH_UMASK(0022)
1264 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1265 if (r < 0)
1266 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
1267
1268 return 1;
1269 }
1270
1271 static int bump_unix_max_dgram_qlen(void) {
1272 _cleanup_free_ char *qlen = NULL;
1273 unsigned long v;
1274 int r;
1275
1276 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel
1277 * default of 16 is simply too low. We set the value really
1278 * really early during boot, so that it is actually applied to
1279 * all our sockets, including the $NOTIFY_SOCKET one. */
1280
1281 r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1282 if (r < 0)
1283 return log_warning_errno(r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1284
1285 r = safe_atolu(qlen, &v);
1286 if (r < 0)
1287 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length, ignoring: %m");
1288
1289 if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1290 return 0;
1291
1292 qlen = mfree(qlen);
1293 if (asprintf(&qlen, "%lu\n", DEFAULT_UNIX_MAX_DGRAM_QLEN) < 0)
1294 return log_oom();
1295
1296 r = write_string_file("/proc/sys/net/unix/max_dgram_qlen", qlen, 0);
1297 if (r < 0)
1298 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1299 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1300
1301 return 1;
1302 }
1303
1304 static int fixup_environment(void) {
1305 _cleanup_free_ char *term = NULL;
1306 int r;
1307
1308 /* We expect the environment to be set correctly
1309 * if run inside a container. */
1310 if (detect_container() > 0)
1311 return 0;
1312
1313 /* When started as PID1, the kernel uses /dev/console
1314 * for our stdios and uses TERM=linux whatever the
1315 * backend device used by the console. We try to make
1316 * a better guess here since some consoles might not
1317 * have support for color mode for example.
1318 *
1319 * However if TERM was configured through the kernel
1320 * command line then leave it alone. */
1321
1322 r = get_proc_cmdline_key("TERM=", &term);
1323 if (r < 0)
1324 return r;
1325
1326 if (r == 0) {
1327 term = strdup(default_term_for_tty("/dev/console"));
1328 if (!term)
1329 return -ENOMEM;
1330 }
1331
1332 if (setenv("TERM", term, 1) < 0)
1333 return -errno;
1334
1335 return 0;
1336 }
1337
1338 int main(int argc, char *argv[]) {
1339 Manager *m = NULL;
1340 int r, retval = EXIT_FAILURE;
1341 usec_t before_startup, after_startup;
1342 char timespan[FORMAT_TIMESPAN_MAX];
1343 FDSet *fds = NULL;
1344 bool reexecute = false;
1345 const char *shutdown_verb = NULL;
1346 dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL;
1347 dual_timestamp userspace_timestamp = DUAL_TIMESTAMP_NULL;
1348 dual_timestamp kernel_timestamp = DUAL_TIMESTAMP_NULL;
1349 dual_timestamp security_start_timestamp = DUAL_TIMESTAMP_NULL;
1350 dual_timestamp security_finish_timestamp = DUAL_TIMESTAMP_NULL;
1351 static char systemd[] = "systemd";
1352 bool skip_setup = false;
1353 unsigned j;
1354 bool loaded_policy = false;
1355 bool arm_reboot_watchdog = false;
1356 bool queue_default_job = false;
1357 bool empty_etc = false;
1358 char *switch_root_dir = NULL, *switch_root_init = NULL;
1359 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0);
1360 const char *error_message = NULL;
1361
1362 #ifdef HAVE_SYSV_COMPAT
1363 if (getpid() != 1 && strstr(program_invocation_short_name, "init")) {
1364 /* This is compatibility support for SysV, where
1365 * calling init as a user is identical to telinit. */
1366
1367 execv(SYSTEMCTL_BINARY_PATH, argv);
1368 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1369 return 1;
1370 }
1371 #endif
1372
1373 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
1374 dual_timestamp_get(&userspace_timestamp);
1375
1376 /* Determine if this is a reexecution or normal bootup. We do
1377 * the full command line parsing much later, so let's just
1378 * have a quick peek here. */
1379 if (strv_find(argv+1, "--deserialize"))
1380 skip_setup = true;
1381
1382 /* If we have switched root, do all the special setup
1383 * things */
1384 if (strv_find(argv+1, "--switched-root"))
1385 skip_setup = false;
1386
1387 /* If we get started via the /sbin/init symlink then we are
1388 called 'init'. After a subsequent reexecution we are then
1389 called 'systemd'. That is confusing, hence let's call us
1390 systemd right-away. */
1391 program_invocation_short_name = systemd;
1392 prctl(PR_SET_NAME, systemd);
1393
1394 saved_argv = argv;
1395 saved_argc = argc;
1396
1397 log_set_upgrade_syslog_to_journal(true);
1398
1399 /* Disable the umask logic */
1400 if (getpid() == 1)
1401 umask(0);
1402
1403 if (getpid() == 1 && detect_container() <= 0) {
1404
1405 /* Running outside of a container as PID 1 */
1406 arg_system = true;
1407 log_set_target(LOG_TARGET_KMSG);
1408 log_open();
1409
1410 if (in_initrd())
1411 initrd_timestamp = userspace_timestamp;
1412
1413 if (!skip_setup) {
1414 r = mount_setup_early();
1415 if (r < 0) {
1416 error_message = "Failed to early mount API filesystems";
1417 goto finish;
1418 }
1419 dual_timestamp_get(&security_start_timestamp);
1420 if (mac_selinux_setup(&loaded_policy) < 0) {
1421 error_message = "Failed to load SELinux policy";
1422 goto finish;
1423 } else if (mac_smack_setup(&loaded_policy) < 0) {
1424 error_message = "Failed to load SMACK policy";
1425 goto finish;
1426 } else if (ima_setup() < 0) {
1427 error_message = "Failed to load IMA policy";
1428 goto finish;
1429 }
1430 dual_timestamp_get(&security_finish_timestamp);
1431 }
1432
1433 if (mac_selinux_init() < 0) {
1434 error_message = "Failed to initialize SELinux policy";
1435 goto finish;
1436 }
1437
1438 if (!skip_setup) {
1439 if (clock_is_localtime(NULL) > 0) {
1440 int min;
1441
1442 /*
1443 * The very first call of settimeofday() also does a time warp in the kernel.
1444 *
1445 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on
1446 * external tools to take care of maintaining the RTC and do all adjustments.
1447 * This matches the behavior of Windows, which leaves the RTC alone if the
1448 * registry tells that the RTC runs in UTC.
1449 */
1450 r = clock_set_timezone(&min);
1451 if (r < 0)
1452 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1453 else
1454 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1455 } else if (!in_initrd()) {
1456 /*
1457 * Do a dummy very first call to seal the kernel's time warp magic.
1458 *
1459 * Do not call this from inside the initrd. The initrd might not
1460 * carry /etc/adjtime with LOCAL, but the real system could be set up
1461 * that way. In such case, we need to delay the time-warp or the sealing
1462 * until we reach the real system.
1463 *
1464 * Do no set the kernel's timezone. The concept of local time cannot
1465 * be supported reliably, the time will jump or be incorrect at every daylight
1466 * saving time change. All kernel local time concepts will be treated
1467 * as UTC that way.
1468 */
1469 (void) clock_reset_timewarp();
1470 }
1471
1472 r = clock_apply_epoch();
1473 if (r < 0)
1474 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1475 else if (r > 0)
1476 log_info("System time before build time, advancing clock.");
1477 }
1478
1479 /* Set the default for later on, but don't actually
1480 * open the logs like this for now. Note that if we
1481 * are transitioning from the initrd there might still
1482 * be journal fd open, and we shouldn't attempt
1483 * opening that before we parsed /proc/cmdline which
1484 * might redirect output elsewhere. */
1485 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
1486
1487 } else if (getpid() == 1) {
1488 /* Running inside a container, as PID 1 */
1489 arg_system = true;
1490 log_set_target(LOG_TARGET_CONSOLE);
1491 log_close_console(); /* force reopen of /dev/console */
1492 log_open();
1493
1494 /* For the later on, see above... */
1495 log_set_target(LOG_TARGET_JOURNAL);
1496
1497 /* clear the kernel timestamp,
1498 * because we are in a container */
1499 kernel_timestamp = DUAL_TIMESTAMP_NULL;
1500 } else {
1501 /* Running as user instance */
1502 arg_system = false;
1503 log_set_target(LOG_TARGET_AUTO);
1504 log_open();
1505
1506 /* clear the kernel timestamp,
1507 * because we are not PID 1 */
1508 kernel_timestamp = DUAL_TIMESTAMP_NULL;
1509 }
1510
1511 if (getpid() == 1) {
1512 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1513 * will process core dumps for system services by default. */
1514 if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0)
1515 log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m");
1516
1517 /* But at the same time, turn off the core_pattern logic by default, so that no coredumps are stored
1518 * until the systemd-coredump tool is enabled via sysctl. */
1519 if (!skip_setup)
1520 (void) write_string_file("/proc/sys/kernel/core_pattern", "|/bin/false", 0);
1521 }
1522
1523 if (arg_system) {
1524 if (fixup_environment() < 0) {
1525 error_message = "Failed to fix up PID1 environment";
1526 goto finish;
1527 }
1528
1529 /* Try to figure out if we can use colors with the console. No
1530 * need to do that for user instances since they never log
1531 * into the console. */
1532 log_show_color(colors_enabled());
1533 make_null_stdio();
1534 }
1535
1536 /* Initialize default unit */
1537 r = free_and_strdup(&arg_default_unit, SPECIAL_DEFAULT_TARGET);
1538 if (r < 0) {
1539 log_emergency_errno(r, "Failed to set default unit %s: %m", SPECIAL_DEFAULT_TARGET);
1540 error_message = "Failed to set default unit";
1541 goto finish;
1542 }
1543
1544 r = initialize_join_controllers();
1545 if (r < 0) {
1546 error_message = "Failed to initialize cgroup controllers";
1547 goto finish;
1548 }
1549
1550 /* Mount /proc, /sys and friends, so that /proc/cmdline and
1551 * /proc/$PID/fd is available. */
1552 if (getpid() == 1) {
1553
1554 /* Load the kernel modules early, so that we kdbus.ko is loaded before kdbusfs shall be mounted */
1555 if (!skip_setup)
1556 kmod_setup();
1557
1558 r = mount_setup(loaded_policy);
1559 if (r < 0) {
1560 error_message = "Failed to mount API filesystems";
1561 goto finish;
1562 }
1563 }
1564
1565 /* Reset all signal handlers. */
1566 (void) reset_all_signal_handlers();
1567 (void) ignore_signals(SIGNALS_IGNORE, -1);
1568
1569 arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U);
1570
1571 if (parse_config_file() < 0) {
1572 error_message = "Failed to parse config file";
1573 goto finish;
1574 }
1575
1576 if (arg_system) {
1577 r = parse_proc_cmdline(parse_proc_cmdline_item);
1578 if (r < 0)
1579 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
1580 }
1581
1582 /* Note that this also parses bits from the kernel command
1583 * line, including "debug". */
1584 log_parse_environment();
1585
1586 if (parse_argv(argc, argv) < 0) {
1587 error_message = "Failed to parse commandline arguments";
1588 goto finish;
1589 }
1590
1591 if (arg_action == ACTION_TEST &&
1592 geteuid() == 0) {
1593 log_error("Don't run test mode as root.");
1594 goto finish;
1595 }
1596
1597 if (!arg_system &&
1598 arg_action == ACTION_RUN &&
1599 sd_booted() <= 0) {
1600 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
1601 goto finish;
1602 }
1603
1604 if (arg_system &&
1605 arg_action == ACTION_RUN &&
1606 running_in_chroot() > 0) {
1607 log_error("Cannot be run in a chroot() environment.");
1608 goto finish;
1609 }
1610
1611 if (arg_action == ACTION_TEST)
1612 skip_setup = true;
1613
1614 if (arg_action == ACTION_TEST || arg_action == ACTION_HELP)
1615 pager_open(arg_no_pager, false);
1616
1617 if (arg_action == ACTION_HELP) {
1618 retval = help();
1619 goto finish;
1620 } else if (arg_action == ACTION_VERSION) {
1621 retval = version();
1622 goto finish;
1623 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
1624 pager_open(arg_no_pager, false);
1625 unit_dump_config_items(stdout);
1626 retval = EXIT_SUCCESS;
1627 goto finish;
1628 }
1629
1630 if (!arg_system &&
1631 !getenv("XDG_RUNTIME_DIR")) {
1632 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
1633 goto finish;
1634 }
1635
1636 assert_se(arg_action == ACTION_RUN || arg_action == ACTION_TEST);
1637
1638 /* Close logging fds, in order not to confuse fdset below */
1639 log_close();
1640
1641 /* Remember open file descriptors for later deserialization */
1642 r = fdset_new_fill(&fds);
1643 if (r < 0) {
1644 log_emergency_errno(r, "Failed to allocate fd set: %m");
1645 error_message = "Failed to allocate fd set";
1646 goto finish;
1647 } else
1648 fdset_cloexec(fds, true);
1649
1650 if (arg_serialization)
1651 assert_se(fdset_remove(fds, fileno(arg_serialization)) >= 0);
1652
1653 if (arg_system)
1654 /* Become a session leader if we aren't one yet. */
1655 setsid();
1656
1657 /* Move out of the way, so that we won't block unmounts */
1658 assert_se(chdir("/") == 0);
1659
1660 /* Reset the console, but only if this is really init and we
1661 * are freshly booted */
1662 if (arg_system && arg_action == ACTION_RUN) {
1663
1664 /* If we are init, we connect stdin/stdout/stderr to
1665 * /dev/null and make sure we don't have a controlling
1666 * tty. */
1667 release_terminal();
1668
1669 if (getpid() == 1 && !skip_setup)
1670 console_setup();
1671 }
1672
1673 /* Open the logging devices, if possible and necessary */
1674 log_open();
1675
1676 if (arg_show_status == _SHOW_STATUS_UNSET)
1677 arg_show_status = SHOW_STATUS_YES;
1678
1679 /* Make sure we leave a core dump without panicing the
1680 * kernel. */
1681 if (getpid() == 1) {
1682 install_crash_handler();
1683
1684 r = mount_cgroup_controllers(arg_join_controllers);
1685 if (r < 0)
1686 goto finish;
1687 }
1688
1689 if (arg_system) {
1690 int v;
1691
1692 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1693 arg_action == ACTION_TEST ? "test " : "" );
1694
1695 v = detect_virtualization();
1696 if (v > 0)
1697 log_info("Detected virtualization %s.", virtualization_to_string(v));
1698
1699 write_container_id();
1700
1701 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1702
1703 if (in_initrd())
1704 log_info("Running in initial RAM disk.");
1705
1706 /* Let's check whether /etc is already populated. We
1707 * don't actually really check for that, but use
1708 * /etc/machine-id as flag file. This allows container
1709 * managers and installers to provision a couple of
1710 * files already. If the container manager wants to
1711 * provision the machine ID itself it should pass
1712 * $container_uuid to PID 1. */
1713
1714 empty_etc = access("/etc/machine-id", F_OK) < 0;
1715 if (empty_etc)
1716 log_info("Running with unpopulated /etc.");
1717 } else {
1718 _cleanup_free_ char *t;
1719
1720 t = uid_to_name(getuid());
1721 log_debug(PACKAGE_STRING " running in %suser mode for user "UID_FMT"/%s. (" SYSTEMD_FEATURES ")",
1722 arg_action == ACTION_TEST ? " test" : "", getuid(), t);
1723 }
1724
1725 if (arg_system && !skip_setup) {
1726 if (arg_show_status > 0)
1727 status_welcome();
1728
1729 hostname_setup();
1730 machine_id_setup(NULL, arg_machine_id, NULL);
1731 loopback_setup();
1732 bump_unix_max_dgram_qlen();
1733
1734 test_usr();
1735 }
1736
1737 if (arg_system && arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY)
1738 watchdog_set_timeout(&arg_runtime_watchdog);
1739
1740 if (arg_timer_slack_nsec != NSEC_INFINITY)
1741 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1742 log_error_errno(errno, "Failed to adjust timer slack: %m");
1743
1744 if (!cap_test_all(arg_capability_bounding_set)) {
1745 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
1746 if (r < 0) {
1747 log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
1748 error_message = "Failed to drop capability bounding set of usermode helpers";
1749 goto finish;
1750 }
1751 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
1752 if (r < 0) {
1753 log_emergency_errno(r, "Failed to drop capability bounding set: %m");
1754 error_message = "Failed to drop capability bounding set";
1755 goto finish;
1756 }
1757 }
1758
1759 if (arg_syscall_archs) {
1760 r = enforce_syscall_archs(arg_syscall_archs);
1761 if (r < 0) {
1762 error_message = "Failed to set syscall architectures";
1763 goto finish;
1764 }
1765 }
1766
1767 if (!arg_system)
1768 /* Become reaper of our children */
1769 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
1770 log_warning_errno(errno, "Failed to make us a subreaper: %m");
1771
1772 if (arg_system) {
1773 bump_rlimit_nofile(&saved_rlimit_nofile);
1774
1775 if (empty_etc) {
1776 r = unit_file_preset_all(UNIT_FILE_SYSTEM, false, NULL, UNIT_FILE_PRESET_ENABLE_ONLY, false, NULL, 0);
1777 if (r < 0)
1778 log_full_errno(r == -EEXIST ? LOG_NOTICE : LOG_WARNING, r, "Failed to populate /etc with preset unit settings, ignoring: %m");
1779 else
1780 log_info("Populated /etc with preset unit settings.");
1781 }
1782 }
1783
1784 r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER, arg_action == ACTION_TEST, &m);
1785 if (r < 0) {
1786 log_emergency_errno(r, "Failed to allocate manager object: %m");
1787 error_message = "Failed to allocate manager object";
1788 goto finish;
1789 }
1790
1791 m->confirm_spawn = arg_confirm_spawn;
1792 m->runtime_watchdog = arg_runtime_watchdog;
1793 m->shutdown_watchdog = arg_shutdown_watchdog;
1794 m->userspace_timestamp = userspace_timestamp;
1795 m->kernel_timestamp = kernel_timestamp;
1796 m->initrd_timestamp = initrd_timestamp;
1797 m->security_start_timestamp = security_start_timestamp;
1798 m->security_finish_timestamp = security_finish_timestamp;
1799
1800 manager_set_defaults(m);
1801 manager_set_show_status(m, arg_show_status);
1802 manager_set_first_boot(m, empty_etc);
1803
1804 /* Remember whether we should queue the default job */
1805 queue_default_job = !arg_serialization || arg_switched_root;
1806
1807 before_startup = now(CLOCK_MONOTONIC);
1808
1809 r = manager_startup(m, arg_serialization, fds);
1810 if (r < 0)
1811 log_error_errno(r, "Failed to fully start up daemon: %m");
1812
1813 /* This will close all file descriptors that were opened, but
1814 * not claimed by any unit. */
1815 fds = fdset_free(fds);
1816
1817 arg_serialization = safe_fclose(arg_serialization);
1818
1819 if (queue_default_job) {
1820 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
1821 Unit *target = NULL;
1822 Job *default_unit_job;
1823
1824 log_debug("Activating default unit: %s", arg_default_unit);
1825
1826 r = manager_load_unit(m, arg_default_unit, NULL, &error, &target);
1827 if (r < 0)
1828 log_error("Failed to load default target: %s", bus_error_message(&error, r));
1829 else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND)
1830 log_error_errno(target->load_error, "Failed to load default target: %m");
1831 else if (target->load_state == UNIT_MASKED)
1832 log_error("Default target masked.");
1833
1834 if (!target || target->load_state != UNIT_LOADED) {
1835 log_info("Trying to load rescue target...");
1836
1837 r = manager_load_unit(m, SPECIAL_RESCUE_TARGET, NULL, &error, &target);
1838 if (r < 0) {
1839 log_emergency("Failed to load rescue target: %s", bus_error_message(&error, r));
1840 error_message = "Failed to load rescue target";
1841 goto finish;
1842 } else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND) {
1843 log_emergency_errno(target->load_error, "Failed to load rescue target: %m");
1844 error_message = "Failed to load rescue target";
1845 goto finish;
1846 } else if (target->load_state == UNIT_MASKED) {
1847 log_emergency("Rescue target masked.");
1848 error_message = "Rescue target masked";
1849 goto finish;
1850 }
1851 }
1852
1853 assert(target->load_state == UNIT_LOADED);
1854
1855 if (arg_action == ACTION_TEST) {
1856 printf("-> By units:\n");
1857 manager_dump_units(m, stdout, "\t");
1858 }
1859
1860 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, &error, &default_unit_job);
1861 if (r == -EPERM) {
1862 log_debug("Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1863
1864 sd_bus_error_free(&error);
1865
1866 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, &error, &default_unit_job);
1867 if (r < 0) {
1868 log_emergency("Failed to start default target: %s", bus_error_message(&error, r));
1869 error_message = "Failed to start default target";
1870 goto finish;
1871 }
1872 } else if (r < 0) {
1873 log_emergency("Failed to isolate default target: %s", bus_error_message(&error, r));
1874 error_message = "Failed to isolate default target";
1875 goto finish;
1876 }
1877
1878 m->default_unit_job_id = default_unit_job->id;
1879
1880 after_startup = now(CLOCK_MONOTONIC);
1881 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
1882 "Loaded units and determined initial transaction in %s.",
1883 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
1884
1885 if (arg_action == ACTION_TEST) {
1886 printf("-> By jobs:\n");
1887 manager_dump_jobs(m, stdout, "\t");
1888 retval = EXIT_SUCCESS;
1889 goto finish;
1890 }
1891 }
1892
1893 for (;;) {
1894 r = manager_loop(m);
1895 if (r < 0) {
1896 log_emergency_errno(r, "Failed to run main loop: %m");
1897 error_message = "Failed to run main loop";
1898 goto finish;
1899 }
1900
1901 switch (m->exit_code) {
1902
1903 case MANAGER_RELOAD:
1904 log_info("Reloading.");
1905
1906 r = parse_config_file();
1907 if (r < 0)
1908 log_error("Failed to parse config file.");
1909
1910 manager_set_defaults(m);
1911
1912 r = manager_reload(m);
1913 if (r < 0)
1914 log_error_errno(r, "Failed to reload: %m");
1915 break;
1916
1917 case MANAGER_REEXECUTE:
1918
1919 if (prepare_reexecute(m, &arg_serialization, &fds, false) < 0) {
1920 error_message = "Failed to prepare for reexecution";
1921 goto finish;
1922 }
1923
1924 reexecute = true;
1925 log_notice("Reexecuting.");
1926 goto finish;
1927
1928 case MANAGER_SWITCH_ROOT:
1929 /* Steal the switch root parameters */
1930 switch_root_dir = m->switch_root;
1931 switch_root_init = m->switch_root_init;
1932 m->switch_root = m->switch_root_init = NULL;
1933
1934 if (!switch_root_init)
1935 if (prepare_reexecute(m, &arg_serialization, &fds, true) < 0) {
1936 error_message = "Failed to prepare for reexecution";
1937 goto finish;
1938 }
1939
1940 reexecute = true;
1941 log_notice("Switching root.");
1942 goto finish;
1943
1944 case MANAGER_EXIT:
1945 retval = m->return_value;
1946
1947 if (MANAGER_IS_USER(m)) {
1948 log_debug("Exit.");
1949 goto finish;
1950 }
1951
1952 /* fallthrough */
1953 case MANAGER_REBOOT:
1954 case MANAGER_POWEROFF:
1955 case MANAGER_HALT:
1956 case MANAGER_KEXEC: {
1957 static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
1958 [MANAGER_EXIT] = "exit",
1959 [MANAGER_REBOOT] = "reboot",
1960 [MANAGER_POWEROFF] = "poweroff",
1961 [MANAGER_HALT] = "halt",
1962 [MANAGER_KEXEC] = "kexec"
1963 };
1964
1965 assert_se(shutdown_verb = table[m->exit_code]);
1966 arm_reboot_watchdog = m->exit_code == MANAGER_REBOOT;
1967
1968 log_notice("Shutting down.");
1969 goto finish;
1970 }
1971
1972 default:
1973 assert_not_reached("Unknown exit code.");
1974 }
1975 }
1976
1977 finish:
1978 pager_close();
1979
1980 if (m)
1981 arg_shutdown_watchdog = m->shutdown_watchdog;
1982
1983 m = manager_free(m);
1984
1985 for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++)
1986 arg_default_rlimit[j] = mfree(arg_default_rlimit[j]);
1987
1988 arg_default_unit = mfree(arg_default_unit);
1989 arg_join_controllers = strv_free_free(arg_join_controllers);
1990 arg_default_environment = strv_free(arg_default_environment);
1991 arg_syscall_archs = set_free(arg_syscall_archs);
1992
1993 mac_selinux_finish();
1994
1995 if (reexecute) {
1996 const char **args;
1997 unsigned i, args_size;
1998
1999 /* Close and disarm the watchdog, so that the new
2000 * instance can reinitialize it, but doesn't get
2001 * rebooted while we do that */
2002 watchdog_close(true);
2003
2004 /* Reset the RLIMIT_NOFILE to the kernel default, so
2005 * that the new systemd can pass the kernel default to
2006 * its child processes */
2007 if (saved_rlimit_nofile.rlim_cur > 0)
2008 (void) setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile);
2009
2010 if (switch_root_dir) {
2011 /* Kill all remaining processes from the
2012 * initrd, but don't wait for them, so that we
2013 * can handle the SIGCHLD for them after
2014 * deserializing. */
2015 broadcast_signal(SIGTERM, false, true);
2016
2017 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
2018 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
2019 if (r < 0)
2020 log_error_errno(r, "Failed to switch root, trying to continue: %m");
2021 }
2022
2023 args_size = MAX(6, argc+1);
2024 args = newa(const char*, args_size);
2025
2026 if (!switch_root_init) {
2027 char sfd[DECIMAL_STR_MAX(int) + 1];
2028
2029 /* First try to spawn ourselves with the right
2030 * path, and with full serialization. We do
2031 * this only if the user didn't specify an
2032 * explicit init to spawn. */
2033
2034 assert(arg_serialization);
2035 assert(fds);
2036
2037 xsprintf(sfd, "%i", fileno(arg_serialization));
2038
2039 i = 0;
2040 args[i++] = SYSTEMD_BINARY_PATH;
2041 if (switch_root_dir)
2042 args[i++] = "--switched-root";
2043 args[i++] = arg_system ? "--system" : "--user";
2044 args[i++] = "--deserialize";
2045 args[i++] = sfd;
2046 args[i++] = NULL;
2047
2048 assert(i <= args_size);
2049
2050 /*
2051 * We want valgrind to print its memory usage summary before reexecution.
2052 * Valgrind won't do this is on its own on exec(), but it will do it on exit().
2053 * Hence, to ensure we get a summary here, fork() off a child, let it exit() cleanly,
2054 * so that it prints the summary, and wait() for it in the parent, before proceeding into the exec().
2055 */
2056 valgrind_summary_hack();
2057
2058 (void) execv(args[0], (char* const*) args);
2059 }
2060
2061 /* Try the fallback, if there is any, without any
2062 * serialization. We pass the original argv[] and
2063 * envp[]. (Well, modulo the ordering changes due to
2064 * getopt() in argv[], and some cleanups in envp[],
2065 * but let's hope that doesn't matter.) */
2066
2067 arg_serialization = safe_fclose(arg_serialization);
2068 fds = fdset_free(fds);
2069
2070 /* Reopen the console */
2071 (void) make_console_stdio();
2072
2073 for (j = 1, i = 1; j < (unsigned) argc; j++)
2074 args[i++] = argv[j];
2075 args[i++] = NULL;
2076 assert(i <= args_size);
2077
2078 /* Reenable any blocked signals, especially important
2079 * if we switch from initial ramdisk to init=... */
2080 (void) reset_all_signal_handlers();
2081 (void) reset_signal_mask();
2082
2083 if (switch_root_init) {
2084 args[0] = switch_root_init;
2085 (void) execv(args[0], (char* const*) args);
2086 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
2087 }
2088
2089 args[0] = "/sbin/init";
2090 (void) execv(args[0], (char* const*) args);
2091
2092 if (errno == ENOENT) {
2093 log_warning("No /sbin/init, trying fallback");
2094
2095 args[0] = "/bin/sh";
2096 args[1] = NULL;
2097 (void) execv(args[0], (char* const*) args);
2098 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
2099 } else
2100 log_warning_errno(errno, "Failed to execute /sbin/init, giving up: %m");
2101 }
2102
2103 arg_serialization = safe_fclose(arg_serialization);
2104 fds = fdset_free(fds);
2105
2106 #ifdef HAVE_VALGRIND_VALGRIND_H
2107 /* If we are PID 1 and running under valgrind, then let's exit
2108 * here explicitly. valgrind will only generate nice output on
2109 * exit(), not on exec(), hence let's do the former not the
2110 * latter here. */
2111 if (getpid() == 1 && RUNNING_ON_VALGRIND)
2112 return 0;
2113 #endif
2114
2115 if (shutdown_verb) {
2116 char log_level[DECIMAL_STR_MAX(int) + 1];
2117 char exit_code[DECIMAL_STR_MAX(uint8_t) + 1];
2118 const char* command_line[11] = {
2119 SYSTEMD_SHUTDOWN_BINARY_PATH,
2120 shutdown_verb,
2121 "--log-level", log_level,
2122 "--log-target",
2123 };
2124 unsigned pos = 5;
2125 _cleanup_strv_free_ char **env_block = NULL;
2126
2127 assert(command_line[pos] == NULL);
2128 env_block = strv_copy(environ);
2129
2130 xsprintf(log_level, "%d", log_get_max_level());
2131
2132 switch (log_get_target()) {
2133
2134 case LOG_TARGET_KMSG:
2135 case LOG_TARGET_JOURNAL_OR_KMSG:
2136 case LOG_TARGET_SYSLOG_OR_KMSG:
2137 command_line[pos++] = "kmsg";
2138 break;
2139
2140 case LOG_TARGET_NULL:
2141 command_line[pos++] = "null";
2142 break;
2143
2144 case LOG_TARGET_CONSOLE:
2145 default:
2146 command_line[pos++] = "console";
2147 break;
2148 };
2149
2150 if (log_get_show_color())
2151 command_line[pos++] = "--log-color";
2152
2153 if (log_get_show_location())
2154 command_line[pos++] = "--log-location";
2155
2156 if (streq(shutdown_verb, "exit")) {
2157 command_line[pos++] = "--exit-code";
2158 command_line[pos++] = exit_code;
2159 xsprintf(exit_code, "%d", retval);
2160 }
2161
2162 assert(pos < ELEMENTSOF(command_line));
2163
2164 if (arm_reboot_watchdog && arg_shutdown_watchdog > 0 && arg_shutdown_watchdog != USEC_INFINITY) {
2165 char *e;
2166
2167 /* If we reboot let's set the shutdown
2168 * watchdog and tell the shutdown binary to
2169 * repeatedly ping it */
2170 r = watchdog_set_timeout(&arg_shutdown_watchdog);
2171 watchdog_close(r < 0);
2172
2173 /* Tell the binary how often to ping, ignore failure */
2174 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
2175 (void) strv_push(&env_block, e);
2176 } else
2177 watchdog_close(true);
2178
2179 /* Avoid the creation of new processes forked by the
2180 * kernel; at this point, we will not listen to the
2181 * signals anyway */
2182 if (detect_container() <= 0)
2183 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
2184
2185 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
2186 log_error_errno(errno, "Failed to execute shutdown binary, %s: %m",
2187 getpid() == 1 ? "freezing" : "quitting");
2188 }
2189
2190 if (getpid() == 1) {
2191 if (error_message)
2192 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2193 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
2194 "%s, freezing.", error_message);
2195 freeze_or_reboot();
2196 }
2197
2198 return retval;
2199 }
Unnamed repository; edit this file 'description' to name the repository.