]> 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
core: change TasksMax= default for system services to 15%
[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 #include "selinux-setup.h"
76 #include "selinux-util.h"
77 #include "signal-util.h"
78 #include "smack-setup.h"
79 #include "special.h"
80 #include "stat-util.h"
81 #include "stdio-util.h"
82 #include "strv.h"
83 #include "switch-root.h"
84 #include "terminal-util.h"
85 #include "umask-util.h"
86 #include "user-util.h"
87 #include "virt.h"
88 #include "watchdog.h"
89
90 static enum {
91 ACTION_RUN,
92 ACTION_HELP,
93 ACTION_VERSION,
94 ACTION_TEST,
95 ACTION_DUMP_CONFIGURATION_ITEMS,
96 ACTION_DONE
97 } arg_action = ACTION_RUN;
98 static char *arg_default_unit = NULL;
99 static bool arg_system = false;
100 static bool arg_dump_core = true;
101 static int arg_crash_chvt = -1;
102 static bool arg_crash_shell = false;
103 static bool arg_crash_reboot = false;
104 static bool arg_confirm_spawn = false;
105 static ShowStatus arg_show_status = _SHOW_STATUS_UNSET;
106 static bool arg_switched_root = false;
107 static bool arg_no_pager = false;
108 static char ***arg_join_controllers = NULL;
109 static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
110 static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
111 static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
112 static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
113 static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
114 static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
115 static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
116 static usec_t arg_runtime_watchdog = 0;
117 static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
118 static char **arg_default_environment = NULL;
119 static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {};
120 static uint64_t arg_capability_bounding_set = CAP_ALL;
121 static nsec_t arg_timer_slack_nsec = NSEC_INFINITY;
122 static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
123 static Set* arg_syscall_archs = NULL;
124 static FILE* arg_serialization = NULL;
125 static bool arg_default_cpu_accounting = false;
126 static bool arg_default_io_accounting = false;
127 static bool arg_default_blockio_accounting = false;
128 static bool arg_default_memory_accounting = false;
129 static bool arg_default_tasks_accounting = true;
130 static uint64_t arg_default_tasks_max = UINT64_MAX;
131 static sd_id128_t arg_machine_id = {};
132
133 noreturn static void freeze_or_reboot(void) {
134
135 if (arg_crash_reboot) {
136 log_notice("Rebooting in 10s...");
137 (void) sleep(10);
138
139 log_notice("Rebooting now...");
140 (void) reboot(RB_AUTOBOOT);
141 log_emergency_errno(errno, "Failed to reboot: %m");
142 }
143
144 log_emergency("Freezing execution.");
145 freeze();
146 }
147
148 noreturn static void crash(int sig) {
149 struct sigaction sa;
150 pid_t pid;
151
152 if (getpid() != 1)
153 /* Pass this on immediately, if this is not PID 1 */
154 (void) raise(sig);
155 else if (!arg_dump_core)
156 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
157 else {
158 sa = (struct sigaction) {
159 .sa_handler = nop_signal_handler,
160 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
161 };
162
163 /* We want to wait for the core process, hence let's enable SIGCHLD */
164 (void) sigaction(SIGCHLD, &sa, NULL);
165
166 pid = raw_clone(SIGCHLD);
167 if (pid < 0)
168 log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
169 else if (pid == 0) {
170 /* Enable default signal handler for core dump */
171
172 sa = (struct sigaction) {
173 .sa_handler = SIG_DFL,
174 };
175 (void) sigaction(sig, &sa, NULL);
176
177 /* Don't limit the coredump size */
178 (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
179
180 /* Just to be sure... */
181 (void) chdir("/");
182
183 /* Raise the signal again */
184 pid = raw_getpid();
185 (void) kill(pid, sig); /* raise() would kill the parent */
186
187 assert_not_reached("We shouldn't be here...");
188 _exit(EXIT_FAILURE);
189 } else {
190 siginfo_t status;
191 int r;
192
193 /* Order things nicely. */
194 r = wait_for_terminate(pid, &status);
195 if (r < 0)
196 log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
197 else if (status.si_code != CLD_DUMPED)
198 log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
199 signal_to_string(sig),
200 pid, sigchld_code_to_string(status.si_code),
201 status.si_status,
202 strna(status.si_code == CLD_EXITED
203 ? exit_status_to_string(status.si_status, EXIT_STATUS_FULL)
204 : signal_to_string(status.si_status)));
205 else
206 log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
207 }
208 }
209
210 if (arg_crash_chvt >= 0)
211 (void) chvt(arg_crash_chvt);
212
213 sa = (struct sigaction) {
214 .sa_handler = SIG_IGN,
215 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
216 };
217
218 /* Let the kernel reap children for us */
219 (void) sigaction(SIGCHLD, &sa, NULL);
220
221 if (arg_crash_shell) {
222 log_notice("Executing crash shell in 10s...");
223 (void) sleep(10);
224
225 pid = raw_clone(SIGCHLD);
226 if (pid < 0)
227 log_emergency_errno(errno, "Failed to fork off crash shell: %m");
228 else if (pid == 0) {
229 (void) setsid();
230 (void) make_console_stdio();
231 (void) execle("/bin/sh", "/bin/sh", NULL, environ);
232
233 log_emergency_errno(errno, "execle() failed: %m");
234 _exit(EXIT_FAILURE);
235 } else {
236 log_info("Spawned crash shell as PID "PID_FMT".", pid);
237 (void) wait_for_terminate(pid, NULL);
238 }
239 }
240
241 freeze_or_reboot();
242 }
243
244 static void install_crash_handler(void) {
245 static const struct sigaction sa = {
246 .sa_handler = crash,
247 .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
248 };
249 int r;
250
251 /* We ignore the return value here, since, we don't mind if we
252 * cannot set up a crash handler */
253 r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
254 if (r < 0)
255 log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
256 }
257
258 static int console_setup(void) {
259 _cleanup_close_ int tty_fd = -1;
260 int r;
261
262 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
263 if (tty_fd < 0)
264 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
265
266 /* We don't want to force text mode. plymouth may be showing
267 * pictures already from initrd. */
268 r = reset_terminal_fd(tty_fd, false);
269 if (r < 0)
270 return log_error_errno(r, "Failed to reset /dev/console: %m");
271
272 return 0;
273 }
274
275 static int parse_crash_chvt(const char *value) {
276 int b;
277
278 if (safe_atoi(value, &arg_crash_chvt) >= 0)
279 return 0;
280
281 b = parse_boolean(value);
282 if (b < 0)
283 return b;
284
285 if (b > 0)
286 arg_crash_chvt = 0; /* switch to where kmsg goes */
287 else
288 arg_crash_chvt = -1; /* turn off switching */
289
290 return 0;
291 }
292
293 static int set_machine_id(const char *m) {
294 assert(m);
295
296 if (sd_id128_from_string(m, &arg_machine_id) < 0)
297 return -EINVAL;
298
299 if (sd_id128_is_null(arg_machine_id))
300 return -EINVAL;
301
302 return 0;
303 }
304
305 static int parse_proc_cmdline_item(const char *key, const char *value) {
306
307 int r;
308
309 assert(key);
310
311 if (streq(key, "systemd.unit") && value) {
312
313 if (!in_initrd())
314 return free_and_strdup(&arg_default_unit, value);
315
316 } else if (streq(key, "rd.systemd.unit") && value) {
317
318 if (in_initrd())
319 return free_and_strdup(&arg_default_unit, value);
320
321 } else if (streq(key, "systemd.dump_core") && value) {
322
323 r = parse_boolean(value);
324 if (r < 0)
325 log_warning("Failed to parse dump core switch %s. Ignoring.", value);
326 else
327 arg_dump_core = r;
328
329 } else if (streq(key, "systemd.crash_chvt") && value) {
330
331 if (parse_crash_chvt(value) < 0)
332 log_warning("Failed to parse crash chvt switch %s. Ignoring.", value);
333
334 } else if (streq(key, "systemd.crash_shell") && value) {
335
336 r = parse_boolean(value);
337 if (r < 0)
338 log_warning("Failed to parse crash shell switch %s. Ignoring.", value);
339 else
340 arg_crash_shell = r;
341
342 } else if (streq(key, "systemd.crash_reboot") && value) {
343
344 r = parse_boolean(value);
345 if (r < 0)
346 log_warning("Failed to parse crash reboot switch %s. Ignoring.", value);
347 else
348 arg_crash_reboot = r;
349
350 } else if (streq(key, "systemd.confirm_spawn") && value) {
351
352 r = parse_boolean(value);
353 if (r < 0)
354 log_warning("Failed to parse confirm spawn switch %s. Ignoring.", value);
355 else
356 arg_confirm_spawn = r;
357
358 } else if (streq(key, "systemd.show_status") && value) {
359
360 r = parse_show_status(value, &arg_show_status);
361 if (r < 0)
362 log_warning("Failed to parse show status switch %s. Ignoring.", value);
363
364 } else if (streq(key, "systemd.default_standard_output") && value) {
365
366 r = exec_output_from_string(value);
367 if (r < 0)
368 log_warning("Failed to parse default standard output switch %s. Ignoring.", value);
369 else
370 arg_default_std_output = r;
371
372 } else if (streq(key, "systemd.default_standard_error") && value) {
373
374 r = exec_output_from_string(value);
375 if (r < 0)
376 log_warning("Failed to parse default standard error switch %s. Ignoring.", value);
377 else
378 arg_default_std_error = r;
379
380 } else if (streq(key, "systemd.setenv") && value) {
381
382 if (env_assignment_is_valid(value)) {
383 char **env;
384
385 env = strv_env_set(arg_default_environment, value);
386 if (env)
387 arg_default_environment = env;
388 else
389 log_warning_errno(ENOMEM, "Setting environment variable '%s' failed, ignoring: %m", value);
390 } else
391 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
392
393 } else if (streq(key, "systemd.machine_id") && value) {
394
395 r = set_machine_id(value);
396 if (r < 0)
397 log_warning("MachineID '%s' is not valid. Ignoring.", value);
398
399 } else if (streq(key, "quiet") && !value) {
400
401 if (arg_show_status == _SHOW_STATUS_UNSET)
402 arg_show_status = SHOW_STATUS_AUTO;
403
404 } else if (streq(key, "debug") && !value) {
405
406 /* Note that log_parse_environment() handles 'debug'
407 * too, and sets the log level to LOG_DEBUG. */
408
409 if (detect_container() > 0)
410 log_set_target(LOG_TARGET_CONSOLE);
411
412 } else if (!value) {
413 const char *target;
414
415 /* SysV compatibility */
416 target = runlevel_to_target(key);
417 if (target)
418 return free_and_strdup(&arg_default_unit, target);
419
420 } else if (streq(key, "systemd.default_timeout_start_sec") && value) {
421
422 r = parse_sec(value, &arg_default_timeout_start_usec);
423 if (r < 0)
424 log_warning_errno(r, "Failed to parse default start timeout: %s, ignoring.", value);
425
426 if (arg_default_timeout_start_usec <= 0)
427 arg_default_timeout_start_usec = USEC_INFINITY;
428 }
429
430 return 0;
431 }
432
433 #define DEFINE_SETTER(name, func, descr) \
434 static int name(const char *unit, \
435 const char *filename, \
436 unsigned line, \
437 const char *section, \
438 unsigned section_line, \
439 const char *lvalue, \
440 int ltype, \
441 const char *rvalue, \
442 void *data, \
443 void *userdata) { \
444 \
445 int r; \
446 \
447 assert(filename); \
448 assert(lvalue); \
449 assert(rvalue); \
450 \
451 r = func(rvalue); \
452 if (r < 0) \
453 log_syntax(unit, LOG_ERR, filename, line, r, \
454 "Invalid " descr "'%s': %m", \
455 rvalue); \
456 \
457 return 0; \
458 }
459
460 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level")
461 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target")
462 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" )
463 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location")
464
465 static int config_parse_cpu_affinity2(
466 const char *unit,
467 const char *filename,
468 unsigned line,
469 const char *section,
470 unsigned section_line,
471 const char *lvalue,
472 int ltype,
473 const char *rvalue,
474 void *data,
475 void *userdata) {
476
477 _cleanup_cpu_free_ cpu_set_t *c = NULL;
478 int ncpus;
479
480 ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue);
481 if (ncpus < 0)
482 return ncpus;
483
484 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
485 log_warning("Failed to set CPU affinity: %m");
486
487 return 0;
488 }
489
490 static int config_parse_show_status(
491 const char* unit,
492 const char *filename,
493 unsigned line,
494 const char *section,
495 unsigned section_line,
496 const char *lvalue,
497 int ltype,
498 const char *rvalue,
499 void *data,
500 void *userdata) {
501
502 int k;
503 ShowStatus *b = data;
504
505 assert(filename);
506 assert(lvalue);
507 assert(rvalue);
508 assert(data);
509
510 k = parse_show_status(rvalue, b);
511 if (k < 0) {
512 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
513 return 0;
514 }
515
516 return 0;
517 }
518
519 static int config_parse_crash_chvt(
520 const char* unit,
521 const char *filename,
522 unsigned line,
523 const char *section,
524 unsigned section_line,
525 const char *lvalue,
526 int ltype,
527 const char *rvalue,
528 void *data,
529 void *userdata) {
530
531 int r;
532
533 assert(filename);
534 assert(lvalue);
535 assert(rvalue);
536
537 r = parse_crash_chvt(rvalue);
538 if (r < 0) {
539 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
540 return 0;
541 }
542
543 return 0;
544 }
545
546 static int config_parse_join_controllers(const char *unit,
547 const char *filename,
548 unsigned line,
549 const char *section,
550 unsigned section_line,
551 const char *lvalue,
552 int ltype,
553 const char *rvalue,
554 void *data,
555 void *userdata) {
556
557 const char *whole_rvalue = rvalue;
558 unsigned n = 0;
559
560 assert(filename);
561 assert(lvalue);
562 assert(rvalue);
563
564 arg_join_controllers = strv_free_free(arg_join_controllers);
565
566 for (;;) {
567 _cleanup_free_ char *word = NULL;
568 char **l;
569 int r;
570
571 r = extract_first_word(&rvalue, &word, WHITESPACE, EXTRACT_QUOTES);
572 if (r < 0) {
573 log_syntax(unit, LOG_ERR, filename, line, r, "Invalid value for %s: %s", lvalue, whole_rvalue);
574 return r;
575 }
576 if (r == 0)
577 break;
578
579 l = strv_split(word, ",");
580 if (!l)
581 return log_oom();
582 strv_uniq(l);
583
584 if (strv_length(l) <= 1) {
585 strv_free(l);
586 continue;
587 }
588
589 if (!arg_join_controllers) {
590 arg_join_controllers = new(char**, 2);
591 if (!arg_join_controllers) {
592 strv_free(l);
593 return log_oom();
594 }
595
596 arg_join_controllers[0] = l;
597 arg_join_controllers[1] = NULL;
598
599 n = 1;
600 } else {
601 char ***a;
602 char ***t;
603
604 t = new0(char**, n+2);
605 if (!t) {
606 strv_free(l);
607 return log_oom();
608 }
609
610 n = 0;
611
612 for (a = arg_join_controllers; *a; a++) {
613
614 if (strv_overlap(*a, l)) {
615 if (strv_extend_strv(&l, *a, false) < 0) {
616 strv_free(l);
617 strv_free_free(t);
618 return log_oom();
619 }
620
621 } else {
622 char **c;
623
624 c = strv_copy(*a);
625 if (!c) {
626 strv_free(l);
627 strv_free_free(t);
628 return log_oom();
629 }
630
631 t[n++] = c;
632 }
633 }
634
635 t[n++] = strv_uniq(l);
636
637 strv_free_free(arg_join_controllers);
638 arg_join_controllers = t;
639 }
640 }
641 if (!isempty(rvalue))
642 log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring.");
643
644 return 0;
645 }
646
647 static int parse_config_file(void) {
648
649 const ConfigTableItem items[] = {
650 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
651 { "Manager", "LogTarget", config_parse_target, 0, NULL },
652 { "Manager", "LogColor", config_parse_color, 0, NULL },
653 { "Manager", "LogLocation", config_parse_location, 0, NULL },
654 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
655 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL },
656 { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL },
657 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
658 { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
659 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
660 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
661 { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
662 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
663 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
664 { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
665 #ifdef HAVE_SECCOMP
666 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
667 #endif
668 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
669 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
670 { "Manager", "DefaultStandardOutput", config_parse_output, 0, &arg_default_std_output },
671 { "Manager", "DefaultStandardError", config_parse_output, 0, &arg_default_std_error },
672 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
673 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
674 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
675 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */
676 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec, 0, &arg_default_start_limit_interval },
677 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
678 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
679 { "Manager", "DefaultLimitCPU", config_parse_limit, RLIMIT_CPU, arg_default_rlimit },
680 { "Manager", "DefaultLimitFSIZE", config_parse_limit, RLIMIT_FSIZE, arg_default_rlimit },
681 { "Manager", "DefaultLimitDATA", config_parse_limit, RLIMIT_DATA, arg_default_rlimit },
682 { "Manager", "DefaultLimitSTACK", config_parse_limit, RLIMIT_STACK, arg_default_rlimit },
683 { "Manager", "DefaultLimitCORE", config_parse_limit, RLIMIT_CORE, arg_default_rlimit },
684 { "Manager", "DefaultLimitRSS", config_parse_limit, RLIMIT_RSS, arg_default_rlimit },
685 { "Manager", "DefaultLimitNOFILE", config_parse_limit, RLIMIT_NOFILE, arg_default_rlimit },
686 { "Manager", "DefaultLimitAS", config_parse_limit, RLIMIT_AS, arg_default_rlimit },
687 { "Manager", "DefaultLimitNPROC", config_parse_limit, RLIMIT_NPROC, arg_default_rlimit },
688 { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, RLIMIT_MEMLOCK, arg_default_rlimit },
689 { "Manager", "DefaultLimitLOCKS", config_parse_limit, RLIMIT_LOCKS, arg_default_rlimit },
690 { "Manager", "DefaultLimitSIGPENDING", config_parse_limit, RLIMIT_SIGPENDING, arg_default_rlimit },
691 { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, RLIMIT_MSGQUEUE, arg_default_rlimit },
692 { "Manager", "DefaultLimitNICE", config_parse_limit, RLIMIT_NICE, arg_default_rlimit },
693 { "Manager", "DefaultLimitRTPRIO", config_parse_limit, RLIMIT_RTPRIO, arg_default_rlimit },
694 { "Manager", "DefaultLimitRTTIME", config_parse_limit, RLIMIT_RTTIME, arg_default_rlimit },
695 { "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_default_cpu_accounting },
696 { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting },
697 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
698 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
699 { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
700 { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max },
701 {}
702 };
703
704 const char *fn, *conf_dirs_nulstr;
705
706 fn = arg_system ?
707 PKGSYSCONFDIR "/system.conf" :
708 PKGSYSCONFDIR "/user.conf";
709
710 conf_dirs_nulstr = arg_system ?
711 CONF_PATHS_NULSTR("systemd/system.conf.d") :
712 CONF_PATHS_NULSTR("systemd/user.conf.d");
713
714 config_parse_many(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, false, NULL);
715
716 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
717 * like everywhere else. */
718 if (arg_default_timeout_start_usec <= 0)
719 arg_default_timeout_start_usec = USEC_INFINITY;
720 if (arg_default_timeout_stop_usec <= 0)
721 arg_default_timeout_stop_usec = USEC_INFINITY;
722
723 return 0;
724 }
725
726 static void manager_set_defaults(Manager *m) {
727
728 assert(m);
729
730 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
731 m->default_std_output = arg_default_std_output;
732 m->default_std_error = arg_default_std_error;
733 m->default_timeout_start_usec = arg_default_timeout_start_usec;
734 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
735 m->default_restart_usec = arg_default_restart_usec;
736 m->default_start_limit_interval = arg_default_start_limit_interval;
737 m->default_start_limit_burst = arg_default_start_limit_burst;
738 m->default_cpu_accounting = arg_default_cpu_accounting;
739 m->default_io_accounting = arg_default_io_accounting;
740 m->default_blockio_accounting = arg_default_blockio_accounting;
741 m->default_memory_accounting = arg_default_memory_accounting;
742 m->default_tasks_accounting = arg_default_tasks_accounting;
743 m->default_tasks_max = arg_default_tasks_max;
744
745 manager_set_default_rlimits(m, arg_default_rlimit);
746 manager_environment_add(m, NULL, arg_default_environment);
747 }
748
749 static int parse_argv(int argc, char *argv[]) {
750
751 enum {
752 ARG_LOG_LEVEL = 0x100,
753 ARG_LOG_TARGET,
754 ARG_LOG_COLOR,
755 ARG_LOG_LOCATION,
756 ARG_UNIT,
757 ARG_SYSTEM,
758 ARG_USER,
759 ARG_TEST,
760 ARG_NO_PAGER,
761 ARG_VERSION,
762 ARG_DUMP_CONFIGURATION_ITEMS,
763 ARG_DUMP_CORE,
764 ARG_CRASH_CHVT,
765 ARG_CRASH_SHELL,
766 ARG_CRASH_REBOOT,
767 ARG_CONFIRM_SPAWN,
768 ARG_SHOW_STATUS,
769 ARG_DESERIALIZE,
770 ARG_SWITCHED_ROOT,
771 ARG_DEFAULT_STD_OUTPUT,
772 ARG_DEFAULT_STD_ERROR,
773 ARG_MACHINE_ID
774 };
775
776 static const struct option options[] = {
777 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
778 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
779 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
780 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
781 { "unit", required_argument, NULL, ARG_UNIT },
782 { "system", no_argument, NULL, ARG_SYSTEM },
783 { "user", no_argument, NULL, ARG_USER },
784 { "test", no_argument, NULL, ARG_TEST },
785 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
786 { "help", no_argument, NULL, 'h' },
787 { "version", no_argument, NULL, ARG_VERSION },
788 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
789 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
790 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
791 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
792 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
793 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
794 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
795 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
796 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
797 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
798 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
799 { "machine-id", required_argument, NULL, ARG_MACHINE_ID },
800 {}
801 };
802
803 int c, r;
804
805 assert(argc >= 1);
806 assert(argv);
807
808 if (getpid() == 1)
809 opterr = 0;
810
811 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
812
813 switch (c) {
814
815 case ARG_LOG_LEVEL:
816 r = log_set_max_level_from_string(optarg);
817 if (r < 0) {
818 log_error("Failed to parse log level %s.", optarg);
819 return r;
820 }
821
822 break;
823
824 case ARG_LOG_TARGET:
825 r = log_set_target_from_string(optarg);
826 if (r < 0) {
827 log_error("Failed to parse log target %s.", optarg);
828 return r;
829 }
830
831 break;
832
833 case ARG_LOG_COLOR:
834
835 if (optarg) {
836 r = log_show_color_from_string(optarg);
837 if (r < 0) {
838 log_error("Failed to parse log color setting %s.", optarg);
839 return r;
840 }
841 } else
842 log_show_color(true);
843
844 break;
845
846 case ARG_LOG_LOCATION:
847 if (optarg) {
848 r = log_show_location_from_string(optarg);
849 if (r < 0) {
850 log_error("Failed to parse log location setting %s.", optarg);
851 return r;
852 }
853 } else
854 log_show_location(true);
855
856 break;
857
858 case ARG_DEFAULT_STD_OUTPUT:
859 r = exec_output_from_string(optarg);
860 if (r < 0) {
861 log_error("Failed to parse default standard output setting %s.", optarg);
862 return r;
863 } else
864 arg_default_std_output = r;
865 break;
866
867 case ARG_DEFAULT_STD_ERROR:
868 r = exec_output_from_string(optarg);
869 if (r < 0) {
870 log_error("Failed to parse default standard error output setting %s.", optarg);
871 return r;
872 } else
873 arg_default_std_error = r;
874 break;
875
876 case ARG_UNIT:
877
878 r = free_and_strdup(&arg_default_unit, optarg);
879 if (r < 0)
880 return log_error_errno(r, "Failed to set default unit %s: %m", optarg);
881
882 break;
883
884 case ARG_SYSTEM:
885 arg_system = true;
886 break;
887
888 case ARG_USER:
889 arg_system = false;
890 break;
891
892 case ARG_TEST:
893 arg_action = ACTION_TEST;
894 break;
895
896 case ARG_NO_PAGER:
897 arg_no_pager = true;
898 break;
899
900 case ARG_VERSION:
901 arg_action = ACTION_VERSION;
902 break;
903
904 case ARG_DUMP_CONFIGURATION_ITEMS:
905 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
906 break;
907
908 case ARG_DUMP_CORE:
909 if (!optarg)
910 arg_dump_core = true;
911 else {
912 r = parse_boolean(optarg);
913 if (r < 0)
914 return log_error_errno(r, "Failed to parse dump core boolean: %s", optarg);
915 arg_dump_core = r;
916 }
917 break;
918
919 case ARG_CRASH_CHVT:
920 r = parse_crash_chvt(optarg);
921 if (r < 0)
922 return log_error_errno(r, "Failed to parse crash virtual terminal index: %s", optarg);
923 break;
924
925 case ARG_CRASH_SHELL:
926 if (!optarg)
927 arg_crash_shell = true;
928 else {
929 r = parse_boolean(optarg);
930 if (r < 0)
931 return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg);
932 arg_crash_shell = r;
933 }
934 break;
935
936 case ARG_CRASH_REBOOT:
937 if (!optarg)
938 arg_crash_reboot = true;
939 else {
940 r = parse_boolean(optarg);
941 if (r < 0)
942 return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg);
943 arg_crash_reboot = r;
944 }
945 break;
946
947 case ARG_CONFIRM_SPAWN:
948 r = optarg ? parse_boolean(optarg) : 1;
949 if (r < 0) {
950 log_error("Failed to parse confirm spawn boolean %s.", optarg);
951 return r;
952 }
953 arg_confirm_spawn = r;
954 break;
955
956 case ARG_SHOW_STATUS:
957 if (optarg) {
958 r = parse_show_status(optarg, &arg_show_status);
959 if (r < 0) {
960 log_error("Failed to parse show status boolean %s.", optarg);
961 return r;
962 }
963 } else
964 arg_show_status = SHOW_STATUS_YES;
965 break;
966
967 case ARG_DESERIALIZE: {
968 int fd;
969 FILE *f;
970
971 r = safe_atoi(optarg, &fd);
972 if (r < 0 || fd < 0) {
973 log_error("Failed to parse deserialize option %s.", optarg);
974 return -EINVAL;
975 }
976
977 (void) fd_cloexec(fd, true);
978
979 f = fdopen(fd, "r");
980 if (!f)
981 return log_error_errno(errno, "Failed to open serialization fd: %m");
982
983 safe_fclose(arg_serialization);
984 arg_serialization = f;
985
986 break;
987 }
988
989 case ARG_SWITCHED_ROOT:
990 arg_switched_root = true;
991 break;
992
993 case ARG_MACHINE_ID:
994 r = set_machine_id(optarg);
995 if (r < 0) {
996 log_error("MachineID '%s' is not valid.", optarg);
997 return r;
998 }
999 break;
1000
1001 case 'h':
1002 arg_action = ACTION_HELP;
1003 break;
1004
1005 case 'D':
1006 log_set_max_level(LOG_DEBUG);
1007 break;
1008
1009 case 'b':
1010 case 's':
1011 case 'z':
1012 /* Just to eat away the sysvinit kernel
1013 * cmdline args without getopt() error
1014 * messages that we'll parse in
1015 * parse_proc_cmdline_word() or ignore. */
1016
1017 case '?':
1018 if (getpid() != 1)
1019 return -EINVAL;
1020 else
1021 return 0;
1022
1023 default:
1024 assert_not_reached("Unhandled option code.");
1025 }
1026
1027 if (optind < argc && getpid() != 1) {
1028 /* Hmm, when we aren't run as init system
1029 * let's complain about excess arguments */
1030
1031 log_error("Excess arguments.");
1032 return -EINVAL;
1033 }
1034
1035 return 0;
1036 }
1037
1038 static int help(void) {
1039
1040 printf("%s [OPTIONS...]\n\n"
1041 "Starts up and maintains the system or user services.\n\n"
1042 " -h --help Show this help\n"
1043 " --test Determine startup sequence, dump it and exit\n"
1044 " --no-pager Do not pipe output into a pager\n"
1045 " --dump-configuration-items Dump understood unit configuration items\n"
1046 " --unit=UNIT Set default unit\n"
1047 " --system Run a system instance, even if PID != 1\n"
1048 " --user Run a user instance\n"
1049 " --dump-core[=BOOL] Dump core on crash\n"
1050 " --crash-vt=NR Change to specified VT on crash\n"
1051 " --crash-reboot[=BOOL] Reboot on crash\n"
1052 " --crash-shell[=BOOL] Run shell on crash\n"
1053 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1054 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1055 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1056 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1057 " --log-color[=BOOL] Highlight important log messages\n"
1058 " --log-location[=BOOL] Include code location in log messages\n"
1059 " --default-standard-output= Set default standard output for services\n"
1060 " --default-standard-error= Set default standard error output for services\n",
1061 program_invocation_short_name);
1062
1063 return 0;
1064 }
1065
1066 static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
1067 _cleanup_fdset_free_ FDSet *fds = NULL;
1068 _cleanup_fclose_ FILE *f = NULL;
1069 int r;
1070
1071 assert(m);
1072 assert(_f);
1073 assert(_fds);
1074
1075 r = manager_open_serialization(m, &f);
1076 if (r < 0)
1077 return log_error_errno(r, "Failed to create serialization file: %m");
1078
1079 /* Make sure nothing is really destructed when we shut down */
1080 m->n_reloading++;
1081 bus_manager_send_reloading(m, true);
1082
1083 fds = fdset_new();
1084 if (!fds)
1085 return log_oom();
1086
1087 r = manager_serialize(m, f, fds, switching_root);
1088 if (r < 0)
1089 return log_error_errno(r, "Failed to serialize state: %m");
1090
1091 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1092 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1093
1094 r = fd_cloexec(fileno(f), false);
1095 if (r < 0)
1096 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1097
1098 r = fdset_cloexec(fds, false);
1099 if (r < 0)
1100 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1101
1102 *_f = f;
1103 *_fds = fds;
1104
1105 f = NULL;
1106 fds = NULL;
1107
1108 return 0;
1109 }
1110
1111 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1112 struct rlimit nl;
1113 int r;
1114
1115 assert(saved_rlimit);
1116
1117 /* Save the original RLIMIT_NOFILE so that we can reset it
1118 * later when transitioning from the initrd to the main
1119 * systemd or suchlike. */
1120 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1121 return log_error_errno(errno, "Reading RLIMIT_NOFILE failed: %m");
1122
1123 /* Make sure forked processes get the default kernel setting */
1124 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1125 struct rlimit *rl;
1126
1127 rl = newdup(struct rlimit, saved_rlimit, 1);
1128 if (!rl)
1129 return log_oom();
1130
1131 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1132 }
1133
1134 /* Bump up the resource limit for ourselves substantially */
1135 nl.rlim_cur = nl.rlim_max = 64*1024;
1136 r = setrlimit_closest(RLIMIT_NOFILE, &nl);
1137 if (r < 0)
1138 return log_error_errno(r, "Setting RLIMIT_NOFILE failed: %m");
1139
1140 return 0;
1141 }
1142
1143 static void test_usr(void) {
1144
1145 /* Check that /usr is not a separate fs */
1146
1147 if (dir_is_empty("/usr") <= 0)
1148 return;
1149
1150 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1151 "Some things will probably break (sometimes even silently) in mysterious ways. "
1152 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1153 }
1154
1155 static int initialize_join_controllers(void) {
1156 /* By default, mount "cpu" + "cpuacct" together, and "net_cls"
1157 * + "net_prio". We'd like to add "cpuset" to the mix, but
1158 * "cpuset" doesn't really work for groups with no initialized
1159 * attributes. */
1160
1161 arg_join_controllers = new(char**, 3);
1162 if (!arg_join_controllers)
1163 return -ENOMEM;
1164
1165 arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL);
1166 if (!arg_join_controllers[0])
1167 goto oom;
1168
1169 arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL);
1170 if (!arg_join_controllers[1])
1171 goto oom;
1172
1173 arg_join_controllers[2] = NULL;
1174 return 0;
1175
1176 oom:
1177 arg_join_controllers = strv_free_free(arg_join_controllers);
1178 return -ENOMEM;
1179 }
1180
1181 static int enforce_syscall_archs(Set *archs) {
1182 #ifdef HAVE_SECCOMP
1183 scmp_filter_ctx *seccomp;
1184 Iterator i;
1185 void *id;
1186 int r;
1187
1188 seccomp = seccomp_init(SCMP_ACT_ALLOW);
1189 if (!seccomp)
1190 return log_oom();
1191
1192 SET_FOREACH(id, arg_syscall_archs, i) {
1193 r = seccomp_arch_add(seccomp, PTR_TO_UINT32(id) - 1);
1194 if (r == -EEXIST)
1195 continue;
1196 if (r < 0) {
1197 log_error_errno(r, "Failed to add architecture to seccomp: %m");
1198 goto finish;
1199 }
1200 }
1201
1202 r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
1203 if (r < 0) {
1204 log_error_errno(r, "Failed to unset NO_NEW_PRIVS: %m");
1205 goto finish;
1206 }
1207
1208 r = seccomp_load(seccomp);
1209 if (r < 0)
1210 log_error_errno(r, "Failed to add install architecture seccomp: %m");
1211
1212 finish:
1213 seccomp_release(seccomp);
1214 return r;
1215 #else
1216 return 0;
1217 #endif
1218 }
1219
1220 static int status_welcome(void) {
1221 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1222 int r;
1223
1224 r = parse_env_file("/etc/os-release", NEWLINE,
1225 "PRETTY_NAME", &pretty_name,
1226 "ANSI_COLOR", &ansi_color,
1227 NULL);
1228 if (r == -ENOENT)
1229 r = parse_env_file("/usr/lib/os-release", NEWLINE,
1230 "PRETTY_NAME", &pretty_name,
1231 "ANSI_COLOR", &ansi_color,
1232 NULL);
1233
1234 if (r < 0 && r != -ENOENT)
1235 log_warning_errno(r, "Failed to read os-release file: %m");
1236
1237 if (log_get_show_color())
1238 return status_printf(NULL, false, false,
1239 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1240 isempty(ansi_color) ? "1" : ansi_color,
1241 isempty(pretty_name) ? "Linux" : pretty_name);
1242 else
1243 return status_printf(NULL, false, false,
1244 "\nWelcome to %s!\n",
1245 isempty(pretty_name) ? "Linux" : pretty_name);
1246 }
1247
1248 static int write_container_id(void) {
1249 const char *c;
1250 int r;
1251
1252 c = getenv("container");
1253 if (isempty(c))
1254 return 0;
1255
1256 RUN_WITH_UMASK(0022)
1257 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1258 if (r < 0)
1259 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
1260
1261 return 1;
1262 }
1263
1264 static int bump_unix_max_dgram_qlen(void) {
1265 _cleanup_free_ char *qlen = NULL;
1266 unsigned long v;
1267 int r;
1268
1269 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel
1270 * default of 16 is simply too low. We set the value really
1271 * really early during boot, so that it is actually applied to
1272 * all our sockets, including the $NOTIFY_SOCKET one. */
1273
1274 r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1275 if (r < 0)
1276 return log_warning_errno(r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1277
1278 r = safe_atolu(qlen, &v);
1279 if (r < 0)
1280 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length, ignoring: %m");
1281
1282 if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1283 return 0;
1284
1285 qlen = mfree(qlen);
1286 if (asprintf(&qlen, "%lu\n", DEFAULT_UNIX_MAX_DGRAM_QLEN) < 0)
1287 return log_oom();
1288
1289 r = write_string_file("/proc/sys/net/unix/max_dgram_qlen", qlen, 0);
1290 if (r < 0)
1291 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1292 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1293
1294 return 1;
1295 }
1296
1297 static int fixup_environment(void) {
1298 _cleanup_free_ char *term = NULL;
1299 int r;
1300
1301 /* We expect the environment to be set correctly
1302 * if run inside a container. */
1303 if (detect_container() > 0)
1304 return 0;
1305
1306 /* When started as PID1, the kernel uses /dev/console
1307 * for our stdios and uses TERM=linux whatever the
1308 * backend device used by the console. We try to make
1309 * a better guess here since some consoles might not
1310 * have support for color mode for example.
1311 *
1312 * However if TERM was configured through the kernel
1313 * command line then leave it alone. */
1314
1315 r = get_proc_cmdline_key("TERM=", &term);
1316 if (r < 0)
1317 return r;
1318
1319 if (r == 0) {
1320 term = strdup(default_term_for_tty("/dev/console") + 5);
1321 if (!term)
1322 return -ENOMEM;
1323 }
1324
1325 if (setenv("TERM", term, 1) < 0)
1326 return -errno;
1327
1328 return 0;
1329 }
1330
1331 int main(int argc, char *argv[]) {
1332 Manager *m = NULL;
1333 int r, retval = EXIT_FAILURE;
1334 usec_t before_startup, after_startup;
1335 char timespan[FORMAT_TIMESPAN_MAX];
1336 FDSet *fds = NULL;
1337 bool reexecute = false;
1338 const char *shutdown_verb = NULL;
1339 dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL;
1340 dual_timestamp userspace_timestamp = DUAL_TIMESTAMP_NULL;
1341 dual_timestamp kernel_timestamp = DUAL_TIMESTAMP_NULL;
1342 dual_timestamp security_start_timestamp = DUAL_TIMESTAMP_NULL;
1343 dual_timestamp security_finish_timestamp = DUAL_TIMESTAMP_NULL;
1344 static char systemd[] = "systemd";
1345 bool skip_setup = false;
1346 unsigned j;
1347 bool loaded_policy = false;
1348 bool arm_reboot_watchdog = false;
1349 bool queue_default_job = false;
1350 bool empty_etc = false;
1351 char *switch_root_dir = NULL, *switch_root_init = NULL;
1352 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0);
1353 const char *error_message = NULL;
1354
1355 #ifdef HAVE_SYSV_COMPAT
1356 if (getpid() != 1 && strstr(program_invocation_short_name, "init")) {
1357 /* This is compatibility support for SysV, where
1358 * calling init as a user is identical to telinit. */
1359
1360 execv(SYSTEMCTL_BINARY_PATH, argv);
1361 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1362 return 1;
1363 }
1364 #endif
1365
1366 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
1367 dual_timestamp_get(&userspace_timestamp);
1368
1369 /* Determine if this is a reexecution or normal bootup. We do
1370 * the full command line parsing much later, so let's just
1371 * have a quick peek here. */
1372 if (strv_find(argv+1, "--deserialize"))
1373 skip_setup = true;
1374
1375 /* If we have switched root, do all the special setup
1376 * things */
1377 if (strv_find(argv+1, "--switched-root"))
1378 skip_setup = false;
1379
1380 /* If we get started via the /sbin/init symlink then we are
1381 called 'init'. After a subsequent reexecution we are then
1382 called 'systemd'. That is confusing, hence let's call us
1383 systemd right-away. */
1384 program_invocation_short_name = systemd;
1385 prctl(PR_SET_NAME, systemd);
1386
1387 saved_argv = argv;
1388 saved_argc = argc;
1389
1390 log_set_upgrade_syslog_to_journal(true);
1391
1392 /* Disable the umask logic */
1393 if (getpid() == 1)
1394 umask(0);
1395
1396 if (getpid() == 1 && detect_container() <= 0) {
1397
1398 /* Running outside of a container as PID 1 */
1399 arg_system = true;
1400 log_set_target(LOG_TARGET_KMSG);
1401 log_open();
1402
1403 if (in_initrd())
1404 initrd_timestamp = userspace_timestamp;
1405
1406 if (!skip_setup) {
1407 r = mount_setup_early();
1408 if (r < 0) {
1409 error_message = "Failed to early mount API filesystems";
1410 goto finish;
1411 }
1412 dual_timestamp_get(&security_start_timestamp);
1413 if (mac_selinux_setup(&loaded_policy) < 0) {
1414 error_message = "Failed to load SELinux policy";
1415 goto finish;
1416 } else if (ima_setup() < 0) {
1417 error_message = "Failed to load IMA policy";
1418 goto finish;
1419 } else if (mac_smack_setup(&loaded_policy) < 0) {
1420 error_message = "Failed to load SMACK policy";
1421 goto finish;
1422 }
1423 dual_timestamp_get(&security_finish_timestamp);
1424 }
1425
1426 if (mac_selinux_init() < 0) {
1427 error_message = "Failed to initialize SELinux policy";
1428 goto finish;
1429 }
1430
1431 if (!skip_setup) {
1432 if (clock_is_localtime(NULL) > 0) {
1433 int min;
1434
1435 /*
1436 * The very first call of settimeofday() also does a time warp in the kernel.
1437 *
1438 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on
1439 * external tools to take care of maintaining the RTC and do all adjustments.
1440 * This matches the behavior of Windows, which leaves the RTC alone if the
1441 * registry tells that the RTC runs in UTC.
1442 */
1443 r = clock_set_timezone(&min);
1444 if (r < 0)
1445 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1446 else
1447 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1448 } else if (!in_initrd()) {
1449 /*
1450 * Do a dummy very first call to seal the kernel's time warp magic.
1451 *
1452 * Do not call this from inside the initrd. The initrd might not
1453 * carry /etc/adjtime with LOCAL, but the real system could be set up
1454 * that way. In such case, we need to delay the time-warp or the sealing
1455 * until we reach the real system.
1456 *
1457 * Do no set the kernel's timezone. The concept of local time cannot
1458 * be supported reliably, the time will jump or be incorrect at every daylight
1459 * saving time change. All kernel local time concepts will be treated
1460 * as UTC that way.
1461 */
1462 (void) clock_reset_timewarp();
1463 }
1464
1465 r = clock_apply_epoch();
1466 if (r < 0)
1467 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1468 else if (r > 0)
1469 log_info("System time before build time, advancing clock.");
1470 }
1471
1472 /* Set the default for later on, but don't actually
1473 * open the logs like this for now. Note that if we
1474 * are transitioning from the initrd there might still
1475 * be journal fd open, and we shouldn't attempt
1476 * opening that before we parsed /proc/cmdline which
1477 * might redirect output elsewhere. */
1478 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
1479
1480 } else if (getpid() == 1) {
1481 /* Running inside a container, as PID 1 */
1482 arg_system = true;
1483 log_set_target(LOG_TARGET_CONSOLE);
1484 log_close_console(); /* force reopen of /dev/console */
1485 log_open();
1486
1487 /* For the later on, see above... */
1488 log_set_target(LOG_TARGET_JOURNAL);
1489
1490 /* clear the kernel timestamp,
1491 * because we are in a container */
1492 kernel_timestamp = DUAL_TIMESTAMP_NULL;
1493 } else {
1494 /* Running as user instance */
1495 arg_system = false;
1496 log_set_target(LOG_TARGET_AUTO);
1497 log_open();
1498
1499 /* clear the kernel timestamp,
1500 * because we are not PID 1 */
1501 kernel_timestamp = DUAL_TIMESTAMP_NULL;
1502 }
1503
1504 if (getpid() == 1) {
1505 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1506 * will process core dumps for system services by default. */
1507 (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
1508
1509 /* But at the same time, turn off the core_pattern logic by default, so that no coredumps are stored
1510 * until the systemd-coredump tool is enabled via sysctl. */
1511 if (!skip_setup)
1512 (void) write_string_file("/proc/sys/kernel/core_pattern", "|/bin/false", 0);
1513 }
1514
1515 if (arg_system) {
1516 if (fixup_environment() < 0) {
1517 error_message = "Failed to fix up PID1 environment";
1518 goto finish;
1519 }
1520
1521 /* Try to figure out if we can use colors with the console. No
1522 * need to do that for user instances since they never log
1523 * into the console. */
1524 log_show_color(colors_enabled());
1525 make_null_stdio();
1526 }
1527
1528 /* Initialize default unit */
1529 r = free_and_strdup(&arg_default_unit, SPECIAL_DEFAULT_TARGET);
1530 if (r < 0) {
1531 log_emergency_errno(r, "Failed to set default unit %s: %m", SPECIAL_DEFAULT_TARGET);
1532 error_message = "Failed to set default unit";
1533 goto finish;
1534 }
1535
1536 r = initialize_join_controllers();
1537 if (r < 0) {
1538 error_message = "Failed to initialize cgroup controllers";
1539 goto finish;
1540 }
1541
1542 /* Mount /proc, /sys and friends, so that /proc/cmdline and
1543 * /proc/$PID/fd is available. */
1544 if (getpid() == 1) {
1545
1546 /* Load the kernel modules early, so that we kdbus.ko is loaded before kdbusfs shall be mounted */
1547 if (!skip_setup)
1548 kmod_setup();
1549
1550 r = mount_setup(loaded_policy);
1551 if (r < 0) {
1552 error_message = "Failed to mount API filesystems";
1553 goto finish;
1554 }
1555 }
1556
1557 /* Reset all signal handlers. */
1558 (void) reset_all_signal_handlers();
1559 (void) ignore_signals(SIGNALS_IGNORE, -1);
1560
1561 arg_default_tasks_max = system_tasks_max_scale(15U, 100U); /* 15% the system PIDs equals 4915 by default. */
1562
1563 if (parse_config_file() < 0) {
1564 error_message = "Failed to parse config file";
1565 goto finish;
1566 }
1567
1568 if (arg_system) {
1569 r = parse_proc_cmdline(parse_proc_cmdline_item);
1570 if (r < 0)
1571 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
1572 }
1573
1574 /* Note that this also parses bits from the kernel command
1575 * line, including "debug". */
1576 log_parse_environment();
1577
1578 if (parse_argv(argc, argv) < 0) {
1579 error_message = "Failed to parse commandline arguments";
1580 goto finish;
1581 }
1582
1583 if (arg_action == ACTION_TEST &&
1584 geteuid() == 0) {
1585 log_error("Don't run test mode as root.");
1586 goto finish;
1587 }
1588
1589 if (!arg_system &&
1590 arg_action == ACTION_RUN &&
1591 sd_booted() <= 0) {
1592 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
1593 goto finish;
1594 }
1595
1596 if (arg_system &&
1597 arg_action == ACTION_RUN &&
1598 running_in_chroot() > 0) {
1599 log_error("Cannot be run in a chroot() environment.");
1600 goto finish;
1601 }
1602
1603 if (arg_action == ACTION_TEST)
1604 skip_setup = true;
1605
1606 if (arg_action == ACTION_TEST || arg_action == ACTION_HELP)
1607 pager_open(arg_no_pager, false);
1608
1609 if (arg_action == ACTION_HELP) {
1610 retval = help();
1611 goto finish;
1612 } else if (arg_action == ACTION_VERSION) {
1613 retval = version();
1614 goto finish;
1615 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
1616 unit_dump_config_items(stdout);
1617 retval = EXIT_SUCCESS;
1618 goto finish;
1619 } else if (arg_action == ACTION_DONE) {
1620 retval = EXIT_SUCCESS;
1621 goto finish;
1622 }
1623
1624 if (!arg_system &&
1625 !getenv("XDG_RUNTIME_DIR")) {
1626 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
1627 goto finish;
1628 }
1629
1630 assert_se(arg_action == ACTION_RUN || arg_action == ACTION_TEST);
1631
1632 /* Close logging fds, in order not to confuse fdset below */
1633 log_close();
1634
1635 /* Remember open file descriptors for later deserialization */
1636 r = fdset_new_fill(&fds);
1637 if (r < 0) {
1638 log_emergency_errno(r, "Failed to allocate fd set: %m");
1639 error_message = "Failed to allocate fd set";
1640 goto finish;
1641 } else
1642 fdset_cloexec(fds, true);
1643
1644 if (arg_serialization)
1645 assert_se(fdset_remove(fds, fileno(arg_serialization)) >= 0);
1646
1647 if (arg_system)
1648 /* Become a session leader if we aren't one yet. */
1649 setsid();
1650
1651 /* Move out of the way, so that we won't block unmounts */
1652 assert_se(chdir("/") == 0);
1653
1654 /* Reset the console, but only if this is really init and we
1655 * are freshly booted */
1656 if (arg_system && arg_action == ACTION_RUN) {
1657
1658 /* If we are init, we connect stdin/stdout/stderr to
1659 * /dev/null and make sure we don't have a controlling
1660 * tty. */
1661 release_terminal();
1662
1663 if (getpid() == 1 && !skip_setup)
1664 console_setup();
1665 }
1666
1667 /* Open the logging devices, if possible and necessary */
1668 log_open();
1669
1670 if (arg_show_status == _SHOW_STATUS_UNSET)
1671 arg_show_status = SHOW_STATUS_YES;
1672
1673 /* Make sure we leave a core dump without panicing the
1674 * kernel. */
1675 if (getpid() == 1) {
1676 install_crash_handler();
1677
1678 r = mount_cgroup_controllers(arg_join_controllers);
1679 if (r < 0)
1680 goto finish;
1681 }
1682
1683 if (arg_system) {
1684 int v;
1685
1686 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1687 arg_action == ACTION_TEST ? "test " : "" );
1688
1689 v = detect_virtualization();
1690 if (v > 0)
1691 log_info("Detected virtualization %s.", virtualization_to_string(v));
1692
1693 write_container_id();
1694
1695 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1696
1697 if (in_initrd())
1698 log_info("Running in initial RAM disk.");
1699
1700 /* Let's check whether /etc is already populated. We
1701 * don't actually really check for that, but use
1702 * /etc/machine-id as flag file. This allows container
1703 * managers and installers to provision a couple of
1704 * files already. If the container manager wants to
1705 * provision the machine ID itself it should pass
1706 * $container_uuid to PID 1. */
1707
1708 empty_etc = access("/etc/machine-id", F_OK) < 0;
1709 if (empty_etc)
1710 log_info("Running with unpopulated /etc.");
1711 } else {
1712 _cleanup_free_ char *t;
1713
1714 t = uid_to_name(getuid());
1715 log_debug(PACKAGE_STRING " running in %suser mode for user "UID_FMT"/%s. (" SYSTEMD_FEATURES ")",
1716 arg_action == ACTION_TEST ? " test" : "", getuid(), t);
1717 }
1718
1719 if (arg_system && !skip_setup) {
1720 if (arg_show_status > 0)
1721 status_welcome();
1722
1723 hostname_setup();
1724 machine_id_setup(NULL, arg_machine_id);
1725 loopback_setup();
1726 bump_unix_max_dgram_qlen();
1727
1728 test_usr();
1729 }
1730
1731 if (arg_system && arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY)
1732 watchdog_set_timeout(&arg_runtime_watchdog);
1733
1734 if (arg_timer_slack_nsec != NSEC_INFINITY)
1735 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1736 log_error_errno(errno, "Failed to adjust timer slack: %m");
1737
1738 if (!cap_test_all(arg_capability_bounding_set)) {
1739 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
1740 if (r < 0) {
1741 log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
1742 error_message = "Failed to drop capability bounding set of usermode helpers";
1743 goto finish;
1744 }
1745 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
1746 if (r < 0) {
1747 log_emergency_errno(r, "Failed to drop capability bounding set: %m");
1748 error_message = "Failed to drop capability bounding set";
1749 goto finish;
1750 }
1751 }
1752
1753 if (arg_syscall_archs) {
1754 r = enforce_syscall_archs(arg_syscall_archs);
1755 if (r < 0) {
1756 error_message = "Failed to set syscall architectures";
1757 goto finish;
1758 }
1759 }
1760
1761 if (!arg_system)
1762 /* Become reaper of our children */
1763 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
1764 log_warning_errno(errno, "Failed to make us a subreaper: %m");
1765
1766 if (arg_system) {
1767 bump_rlimit_nofile(&saved_rlimit_nofile);
1768
1769 if (empty_etc) {
1770 r = unit_file_preset_all(UNIT_FILE_SYSTEM, false, NULL, UNIT_FILE_PRESET_ENABLE_ONLY, false, NULL, 0);
1771 if (r < 0)
1772 log_full_errno(r == -EEXIST ? LOG_NOTICE : LOG_WARNING, r, "Failed to populate /etc with preset unit settings, ignoring: %m");
1773 else
1774 log_info("Populated /etc with preset unit settings.");
1775 }
1776 }
1777
1778 r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER, arg_action == ACTION_TEST, &m);
1779 if (r < 0) {
1780 log_emergency_errno(r, "Failed to allocate manager object: %m");
1781 error_message = "Failed to allocate manager object";
1782 goto finish;
1783 }
1784
1785 m->confirm_spawn = arg_confirm_spawn;
1786 m->runtime_watchdog = arg_runtime_watchdog;
1787 m->shutdown_watchdog = arg_shutdown_watchdog;
1788 m->userspace_timestamp = userspace_timestamp;
1789 m->kernel_timestamp = kernel_timestamp;
1790 m->initrd_timestamp = initrd_timestamp;
1791 m->security_start_timestamp = security_start_timestamp;
1792 m->security_finish_timestamp = security_finish_timestamp;
1793
1794 manager_set_defaults(m);
1795 manager_set_show_status(m, arg_show_status);
1796 manager_set_first_boot(m, empty_etc);
1797
1798 /* Remember whether we should queue the default job */
1799 queue_default_job = !arg_serialization || arg_switched_root;
1800
1801 before_startup = now(CLOCK_MONOTONIC);
1802
1803 r = manager_startup(m, arg_serialization, fds);
1804 if (r < 0)
1805 log_error_errno(r, "Failed to fully start up daemon: %m");
1806
1807 /* This will close all file descriptors that were opened, but
1808 * not claimed by any unit. */
1809 fds = fdset_free(fds);
1810
1811 arg_serialization = safe_fclose(arg_serialization);
1812
1813 if (queue_default_job) {
1814 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
1815 Unit *target = NULL;
1816 Job *default_unit_job;
1817
1818 log_debug("Activating default unit: %s", arg_default_unit);
1819
1820 r = manager_load_unit(m, arg_default_unit, NULL, &error, &target);
1821 if (r < 0)
1822 log_error("Failed to load default target: %s", bus_error_message(&error, r));
1823 else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND)
1824 log_error_errno(target->load_error, "Failed to load default target: %m");
1825 else if (target->load_state == UNIT_MASKED)
1826 log_error("Default target masked.");
1827
1828 if (!target || target->load_state != UNIT_LOADED) {
1829 log_info("Trying to load rescue target...");
1830
1831 r = manager_load_unit(m, SPECIAL_RESCUE_TARGET, NULL, &error, &target);
1832 if (r < 0) {
1833 log_emergency("Failed to load rescue target: %s", bus_error_message(&error, r));
1834 error_message = "Failed to load rescue target";
1835 goto finish;
1836 } else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND) {
1837 log_emergency_errno(target->load_error, "Failed to load rescue target: %m");
1838 error_message = "Failed to load rescue target";
1839 goto finish;
1840 } else if (target->load_state == UNIT_MASKED) {
1841 log_emergency("Rescue target masked.");
1842 error_message = "Rescue target masked";
1843 goto finish;
1844 }
1845 }
1846
1847 assert(target->load_state == UNIT_LOADED);
1848
1849 if (arg_action == ACTION_TEST) {
1850 printf("-> By units:\n");
1851 manager_dump_units(m, stdout, "\t");
1852 }
1853
1854 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, &error, &default_unit_job);
1855 if (r == -EPERM) {
1856 log_debug("Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1857
1858 sd_bus_error_free(&error);
1859
1860 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, &error, &default_unit_job);
1861 if (r < 0) {
1862 log_emergency("Failed to start default target: %s", bus_error_message(&error, r));
1863 error_message = "Failed to start default target";
1864 goto finish;
1865 }
1866 } else if (r < 0) {
1867 log_emergency("Failed to isolate default target: %s", bus_error_message(&error, r));
1868 error_message = "Failed to isolate default target";
1869 goto finish;
1870 }
1871
1872 m->default_unit_job_id = default_unit_job->id;
1873
1874 after_startup = now(CLOCK_MONOTONIC);
1875 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
1876 "Loaded units and determined initial transaction in %s.",
1877 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
1878
1879 if (arg_action == ACTION_TEST) {
1880 printf("-> By jobs:\n");
1881 manager_dump_jobs(m, stdout, "\t");
1882 retval = EXIT_SUCCESS;
1883 goto finish;
1884 }
1885 }
1886
1887 for (;;) {
1888 r = manager_loop(m);
1889 if (r < 0) {
1890 log_emergency_errno(r, "Failed to run main loop: %m");
1891 error_message = "Failed to run main loop";
1892 goto finish;
1893 }
1894
1895 switch (m->exit_code) {
1896
1897 case MANAGER_RELOAD:
1898 log_info("Reloading.");
1899
1900 r = parse_config_file();
1901 if (r < 0)
1902 log_error("Failed to parse config file.");
1903
1904 manager_set_defaults(m);
1905
1906 r = manager_reload(m);
1907 if (r < 0)
1908 log_error_errno(r, "Failed to reload: %m");
1909 break;
1910
1911 case MANAGER_REEXECUTE:
1912
1913 if (prepare_reexecute(m, &arg_serialization, &fds, false) < 0) {
1914 error_message = "Failed to prepare for reexecution";
1915 goto finish;
1916 }
1917
1918 reexecute = true;
1919 log_notice("Reexecuting.");
1920 goto finish;
1921
1922 case MANAGER_SWITCH_ROOT:
1923 /* Steal the switch root parameters */
1924 switch_root_dir = m->switch_root;
1925 switch_root_init = m->switch_root_init;
1926 m->switch_root = m->switch_root_init = NULL;
1927
1928 if (!switch_root_init)
1929 if (prepare_reexecute(m, &arg_serialization, &fds, true) < 0) {
1930 error_message = "Failed to prepare for reexecution";
1931 goto finish;
1932 }
1933
1934 reexecute = true;
1935 log_notice("Switching root.");
1936 goto finish;
1937
1938 case MANAGER_EXIT:
1939 retval = m->return_value;
1940
1941 if (MANAGER_IS_USER(m)) {
1942 log_debug("Exit.");
1943 goto finish;
1944 }
1945
1946 /* fallthrough */
1947 case MANAGER_REBOOT:
1948 case MANAGER_POWEROFF:
1949 case MANAGER_HALT:
1950 case MANAGER_KEXEC: {
1951 static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
1952 [MANAGER_EXIT] = "exit",
1953 [MANAGER_REBOOT] = "reboot",
1954 [MANAGER_POWEROFF] = "poweroff",
1955 [MANAGER_HALT] = "halt",
1956 [MANAGER_KEXEC] = "kexec"
1957 };
1958
1959 assert_se(shutdown_verb = table[m->exit_code]);
1960 arm_reboot_watchdog = m->exit_code == MANAGER_REBOOT;
1961
1962 log_notice("Shutting down.");
1963 goto finish;
1964 }
1965
1966 default:
1967 assert_not_reached("Unknown exit code.");
1968 }
1969 }
1970
1971 finish:
1972 pager_close();
1973
1974 if (m)
1975 arg_shutdown_watchdog = m->shutdown_watchdog;
1976
1977 m = manager_free(m);
1978
1979 for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++)
1980 arg_default_rlimit[j] = mfree(arg_default_rlimit[j]);
1981
1982 arg_default_unit = mfree(arg_default_unit);
1983 arg_join_controllers = strv_free_free(arg_join_controllers);
1984 arg_default_environment = strv_free(arg_default_environment);
1985 arg_syscall_archs = set_free(arg_syscall_archs);
1986
1987 mac_selinux_finish();
1988
1989 if (reexecute) {
1990 const char **args;
1991 unsigned i, args_size;
1992
1993 /* Close and disarm the watchdog, so that the new
1994 * instance can reinitialize it, but doesn't get
1995 * rebooted while we do that */
1996 watchdog_close(true);
1997
1998 /* Reset the RLIMIT_NOFILE to the kernel default, so
1999 * that the new systemd can pass the kernel default to
2000 * its child processes */
2001 if (saved_rlimit_nofile.rlim_cur > 0)
2002 (void) setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile);
2003
2004 if (switch_root_dir) {
2005 /* Kill all remaining processes from the
2006 * initrd, but don't wait for them, so that we
2007 * can handle the SIGCHLD for them after
2008 * deserializing. */
2009 broadcast_signal(SIGTERM, false, true);
2010
2011 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
2012 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
2013 if (r < 0)
2014 log_error_errno(r, "Failed to switch root, trying to continue: %m");
2015 }
2016
2017 /* Reopen the console */
2018 (void) make_console_stdio();
2019
2020 args_size = MAX(6, argc+1);
2021 args = newa(const char*, args_size);
2022
2023 if (!switch_root_init) {
2024 char sfd[DECIMAL_STR_MAX(int) + 1];
2025
2026 /* First try to spawn ourselves with the right
2027 * path, and with full serialization. We do
2028 * this only if the user didn't specify an
2029 * explicit init to spawn. */
2030
2031 assert(arg_serialization);
2032 assert(fds);
2033
2034 xsprintf(sfd, "%i", fileno(arg_serialization));
2035
2036 i = 0;
2037 args[i++] = SYSTEMD_BINARY_PATH;
2038 if (switch_root_dir)
2039 args[i++] = "--switched-root";
2040 args[i++] = arg_system ? "--system" : "--user";
2041 args[i++] = "--deserialize";
2042 args[i++] = sfd;
2043 args[i++] = NULL;
2044
2045 assert(i <= args_size);
2046
2047 /*
2048 * We want valgrind to print its memory usage summary before reexecution.
2049 * Valgrind won't do this is on its own on exec(), but it will do it on exit().
2050 * Hence, to ensure we get a summary here, fork() off a child, let it exit() cleanly,
2051 * so that it prints the summary, and wait() for it in the parent, before proceeding into the exec().
2052 */
2053 valgrind_summary_hack();
2054
2055 (void) execv(args[0], (char* const*) args);
2056 }
2057
2058 /* Try the fallback, if there is any, without any
2059 * serialization. We pass the original argv[] and
2060 * envp[]. (Well, modulo the ordering changes due to
2061 * getopt() in argv[], and some cleanups in envp[],
2062 * but let's hope that doesn't matter.) */
2063
2064 arg_serialization = safe_fclose(arg_serialization);
2065 fds = fdset_free(fds);
2066
2067 for (j = 1, i = 1; j < (unsigned) argc; j++)
2068 args[i++] = argv[j];
2069 args[i++] = NULL;
2070 assert(i <= args_size);
2071
2072 /* Reenable any blocked signals, especially important
2073 * if we switch from initial ramdisk to init=... */
2074 (void) reset_all_signal_handlers();
2075 (void) reset_signal_mask();
2076
2077 if (switch_root_init) {
2078 args[0] = switch_root_init;
2079 (void) execv(args[0], (char* const*) args);
2080 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
2081 }
2082
2083 args[0] = "/sbin/init";
2084 (void) execv(args[0], (char* const*) args);
2085
2086 if (errno == ENOENT) {
2087 log_warning("No /sbin/init, trying fallback");
2088
2089 args[0] = "/bin/sh";
2090 args[1] = NULL;
2091 (void) execv(args[0], (char* const*) args);
2092 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
2093 } else
2094 log_warning_errno(errno, "Failed to execute /sbin/init, giving up: %m");
2095 }
2096
2097 arg_serialization = safe_fclose(arg_serialization);
2098 fds = fdset_free(fds);
2099
2100 #ifdef HAVE_VALGRIND_VALGRIND_H
2101 /* If we are PID 1 and running under valgrind, then let's exit
2102 * here explicitly. valgrind will only generate nice output on
2103 * exit(), not on exec(), hence let's do the former not the
2104 * latter here. */
2105 if (getpid() == 1 && RUNNING_ON_VALGRIND)
2106 return 0;
2107 #endif
2108
2109 if (shutdown_verb) {
2110 char log_level[DECIMAL_STR_MAX(int) + 1];
2111 char exit_code[DECIMAL_STR_MAX(uint8_t) + 1];
2112 const char* command_line[11] = {
2113 SYSTEMD_SHUTDOWN_BINARY_PATH,
2114 shutdown_verb,
2115 "--log-level", log_level,
2116 "--log-target",
2117 };
2118 unsigned pos = 5;
2119 _cleanup_strv_free_ char **env_block = NULL;
2120
2121 assert(command_line[pos] == NULL);
2122 env_block = strv_copy(environ);
2123
2124 xsprintf(log_level, "%d", log_get_max_level());
2125
2126 switch (log_get_target()) {
2127
2128 case LOG_TARGET_KMSG:
2129 case LOG_TARGET_JOURNAL_OR_KMSG:
2130 case LOG_TARGET_SYSLOG_OR_KMSG:
2131 command_line[pos++] = "kmsg";
2132 break;
2133
2134 case LOG_TARGET_NULL:
2135 command_line[pos++] = "null";
2136 break;
2137
2138 case LOG_TARGET_CONSOLE:
2139 default:
2140 command_line[pos++] = "console";
2141 break;
2142 };
2143
2144 if (log_get_show_color())
2145 command_line[pos++] = "--log-color";
2146
2147 if (log_get_show_location())
2148 command_line[pos++] = "--log-location";
2149
2150 if (streq(shutdown_verb, "exit")) {
2151 command_line[pos++] = "--exit-code";
2152 command_line[pos++] = exit_code;
2153 xsprintf(exit_code, "%d", retval);
2154 }
2155
2156 assert(pos < ELEMENTSOF(command_line));
2157
2158 if (arm_reboot_watchdog && arg_shutdown_watchdog > 0 && arg_shutdown_watchdog != USEC_INFINITY) {
2159 char *e;
2160
2161 /* If we reboot let's set the shutdown
2162 * watchdog and tell the shutdown binary to
2163 * repeatedly ping it */
2164 r = watchdog_set_timeout(&arg_shutdown_watchdog);
2165 watchdog_close(r < 0);
2166
2167 /* Tell the binary how often to ping, ignore failure */
2168 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
2169 (void) strv_push(&env_block, e);
2170 } else
2171 watchdog_close(true);
2172
2173 /* Avoid the creation of new processes forked by the
2174 * kernel; at this point, we will not listen to the
2175 * signals anyway */
2176 if (detect_container() <= 0)
2177 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
2178
2179 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
2180 log_error_errno(errno, "Failed to execute shutdown binary, %s: %m",
2181 getpid() == 1 ? "freezing" : "quitting");
2182 }
2183
2184 if (getpid() == 1) {
2185 if (error_message)
2186 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2187 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
2188 "%s, freezing.", error_message);
2189 freeze_or_reboot();
2190 }
2191
2192 return retval;
2193 }
Unnamed repository; edit this file 'description' to name the repository.