]> git.ipfire.org Git - thirdparty/systemd.git/blob - src/core/main.c
tty-ask-password: Split out password sending
[thirdparty/systemd.git] / src / core / main.c
1 /*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
2
3 /***
4 This file is part of systemd.
5
6 Copyright 2010 Lennart Poettering
7
8 systemd is free software; you can redistribute it and/or modify it
9 under the terms of the GNU Lesser General Public License as published by
10 the Free Software Foundation; either version 2.1 of the License, or
11 (at your option) any later version.
12
13 systemd is distributed in the hope that it will be useful, but
14 WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 Lesser General Public License for more details.
17
18 You should have received a copy of the GNU Lesser General Public License
19 along with systemd; If not, see <http://www.gnu.org/licenses/>.
20 ***/
21
22 #include <errno.h>
23 #include <fcntl.h>
24 #include <getopt.h>
25 #include <signal.h>
26 #include <stdio.h>
27 #include <string.h>
28 #include <sys/mount.h>
29 #include <sys/prctl.h>
30 #include <sys/reboot.h>
31 #include <sys/stat.h>
32 #include <unistd.h>
33 #ifdef HAVE_SECCOMP
34 #include <seccomp.h>
35 #endif
36 #ifdef HAVE_VALGRIND_VALGRIND_H
37 #include <valgrind/valgrind.h>
38 #endif
39
40 #include "sd-bus.h"
41 #include "sd-daemon.h"
42
43 #include "alloc-util.h"
44 #include "architecture.h"
45 #include "build.h"
46 #include "bus-error.h"
47 #include "bus-util.h"
48 #include "capability-util.h"
49 #include "clock-util.h"
50 #include "conf-parser.h"
51 #include "cpu-set-util.h"
52 #include "dbus-manager.h"
53 #include "def.h"
54 #include "env-util.h"
55 #include "fd-util.h"
56 #include "fdset.h"
57 #include "fileio.h"
58 #include "formats-util.h"
59 #include "fs-util.h"
60 #include "hostname-setup.h"
61 #include "ima-setup.h"
62 #include "killall.h"
63 #include "kmod-setup.h"
64 #include "load-fragment.h"
65 #include "log.h"
66 #include "loopback-setup.h"
67 #include "machine-id-setup.h"
68 #include "manager.h"
69 #include "missing.h"
70 #include "mount-setup.h"
71 #include "pager.h"
72 #include "parse-util.h"
73 #include "proc-cmdline.h"
74 #include "process-util.h"
75 #include "rlimit-util.h"
76 #include "selinux-setup.h"
77 #include "selinux-util.h"
78 #include "signal-util.h"
79 #include "smack-setup.h"
80 #include "special.h"
81 #include "stat-util.h"
82 #include "stdio-util.h"
83 #include "strv.h"
84 #include "switch-root.h"
85 #include "terminal-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 ManagerRunningAs arg_running_as = _MANAGER_RUNNING_AS_INVALID;
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 int arg_no_pager = -1;
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_blockio_accounting = false;
127 static bool arg_default_memory_accounting = false;
128 static bool arg_default_tasks_accounting = true;
129 static uint64_t arg_default_tasks_max = UINT64_C(512);
130 static sd_id128_t arg_machine_id = {};
131
132 static void pager_open_if_enabled(void) {
133
134 if (arg_no_pager <= 0)
135 return;
136
137 pager_open(false);
138 }
139
140 noreturn static void freeze_or_reboot(void) {
141
142 if (arg_crash_reboot) {
143 log_notice("Rebooting in 10s...");
144 (void) sleep(10);
145
146 log_notice("Rebooting now...");
147 (void) reboot(RB_AUTOBOOT);
148 log_emergency_errno(errno, "Failed to reboot: %m");
149 }
150
151 log_emergency("Freezing execution.");
152 freeze();
153 }
154
155 noreturn static void crash(int sig) {
156 struct sigaction sa;
157 pid_t pid;
158
159 if (getpid() != 1)
160 /* Pass this on immediately, if this is not PID 1 */
161 (void) raise(sig);
162 else if (!arg_dump_core)
163 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
164 else {
165 sa = (struct sigaction) {
166 .sa_handler = nop_signal_handler,
167 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
168 };
169
170 /* We want to wait for the core process, hence let's enable SIGCHLD */
171 (void) sigaction(SIGCHLD, &sa, NULL);
172
173 pid = raw_clone(SIGCHLD, NULL);
174 if (pid < 0)
175 log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
176 else if (pid == 0) {
177 struct rlimit rl = {
178 .rlim_cur = RLIM_INFINITY,
179 .rlim_max = RLIM_INFINITY,
180 };
181
182 /* Enable default signal handler for core dump */
183 sa = (struct sigaction) {
184 .sa_handler = SIG_DFL,
185 };
186 (void) sigaction(sig, &sa, NULL);
187
188 /* Don't limit the core dump size */
189 (void) setrlimit(RLIMIT_CORE, &rl);
190
191 /* Just to be sure... */
192 (void) chdir("/");
193
194 /* Raise the signal again */
195 pid = raw_getpid();
196 (void) kill(pid, sig); /* raise() would kill the parent */
197
198 assert_not_reached("We shouldn't be here...");
199 _exit(EXIT_FAILURE);
200 } else {
201 siginfo_t status;
202 int r;
203
204 /* Order things nicely. */
205 r = wait_for_terminate(pid, &status);
206 if (r < 0)
207 log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
208 else if (status.si_code != CLD_DUMPED)
209 log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
210 signal_to_string(sig),
211 pid, sigchld_code_to_string(status.si_code),
212 status.si_status,
213 strna(status.si_code == CLD_EXITED
214 ? exit_status_to_string(status.si_status, EXIT_STATUS_FULL)
215 : signal_to_string(status.si_status)));
216 else
217 log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
218 }
219 }
220
221 if (arg_crash_chvt >= 0)
222 (void) chvt(arg_crash_chvt);
223
224 sa = (struct sigaction) {
225 .sa_handler = SIG_IGN,
226 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
227 };
228
229 /* Let the kernel reap children for us */
230 (void) sigaction(SIGCHLD, &sa, NULL);
231
232 if (arg_crash_shell) {
233 log_notice("Executing crash shell in 10s...");
234 (void) sleep(10);
235
236 pid = raw_clone(SIGCHLD, NULL);
237 if (pid < 0)
238 log_emergency_errno(errno, "Failed to fork off crash shell: %m");
239 else if (pid == 0) {
240 (void) setsid();
241 (void) make_console_stdio();
242 (void) execle("/bin/sh", "/bin/sh", NULL, environ);
243
244 log_emergency_errno(errno, "execle() failed: %m");
245 _exit(EXIT_FAILURE);
246 } else {
247 log_info("Spawned crash shell as PID "PID_FMT".", pid);
248 (void) wait_for_terminate(pid, NULL);
249 }
250 }
251
252 freeze_or_reboot();
253 }
254
255 static void install_crash_handler(void) {
256 static const struct sigaction sa = {
257 .sa_handler = crash,
258 .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
259 };
260 int r;
261
262 /* We ignore the return value here, since, we don't mind if we
263 * cannot set up a crash handler */
264 r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
265 if (r < 0)
266 log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
267 }
268
269 static int console_setup(void) {
270 _cleanup_close_ int tty_fd = -1;
271 int r;
272
273 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
274 if (tty_fd < 0)
275 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
276
277 /* We don't want to force text mode. plymouth may be showing
278 * pictures already from initrd. */
279 r = reset_terminal_fd(tty_fd, false);
280 if (r < 0)
281 return log_error_errno(r, "Failed to reset /dev/console: %m");
282
283 return 0;
284 }
285
286 static int parse_crash_chvt(const char *value) {
287 int b;
288
289 if (safe_atoi(value, &arg_crash_chvt) >= 0)
290 return 0;
291
292 b = parse_boolean(value);
293 if (b < 0)
294 return b;
295
296 if (b > 0)
297 arg_crash_chvt = 0; /* switch to where kmsg goes */
298 else
299 arg_crash_chvt = -1; /* turn off switching */
300
301 return 0;
302 }
303
304 static int set_machine_id(const char *m) {
305
306 if (sd_id128_from_string(m, &arg_machine_id) < 0)
307 return -EINVAL;
308
309 if (sd_id128_is_null(arg_machine_id))
310 return -EINVAL;
311
312 return 0;
313 }
314
315 static int parse_proc_cmdline_item(const char *key, const char *value) {
316
317 int r;
318
319 assert(key);
320
321 if (streq(key, "systemd.unit") && value) {
322
323 if (!in_initrd())
324 return free_and_strdup(&arg_default_unit, value);
325
326 } else if (streq(key, "rd.systemd.unit") && value) {
327
328 if (in_initrd())
329 return free_and_strdup(&arg_default_unit, value);
330
331 } else if (streq(key, "systemd.dump_core") && value) {
332
333 r = parse_boolean(value);
334 if (r < 0)
335 log_warning("Failed to parse dump core switch %s. Ignoring.", value);
336 else
337 arg_dump_core = r;
338
339 } else if (streq(key, "systemd.crash_chvt") && value) {
340
341 if (parse_crash_chvt(value) < 0)
342 log_warning("Failed to parse crash chvt switch %s. Ignoring.", value);
343
344 } else if (streq(key, "systemd.crash_shell") && value) {
345
346 r = parse_boolean(value);
347 if (r < 0)
348 log_warning("Failed to parse crash shell switch %s. Ignoring.", value);
349 else
350 arg_crash_shell = r;
351
352 } else if (streq(key, "systemd.crash_reboot") && value) {
353
354 r = parse_boolean(value);
355 if (r < 0)
356 log_warning("Failed to parse crash reboot switch %s. Ignoring.", value);
357 else
358 arg_crash_reboot = r;
359
360 } else if (streq(key, "systemd.confirm_spawn") && value) {
361
362 r = parse_boolean(value);
363 if (r < 0)
364 log_warning("Failed to parse confirm spawn switch %s. Ignoring.", value);
365 else
366 arg_confirm_spawn = r;
367
368 } else if (streq(key, "systemd.show_status") && value) {
369
370 r = parse_show_status(value, &arg_show_status);
371 if (r < 0)
372 log_warning("Failed to parse show status switch %s. Ignoring.", value);
373
374 } else if (streq(key, "systemd.default_standard_output") && value) {
375
376 r = exec_output_from_string(value);
377 if (r < 0)
378 log_warning("Failed to parse default standard output switch %s. Ignoring.", value);
379 else
380 arg_default_std_output = r;
381
382 } else if (streq(key, "systemd.default_standard_error") && value) {
383
384 r = exec_output_from_string(value);
385 if (r < 0)
386 log_warning("Failed to parse default standard error switch %s. Ignoring.", value);
387 else
388 arg_default_std_error = r;
389
390 } else if (streq(key, "systemd.setenv") && value) {
391
392 if (env_assignment_is_valid(value)) {
393 char **env;
394
395 env = strv_env_set(arg_default_environment, value);
396 if (env)
397 arg_default_environment = env;
398 else
399 log_warning_errno(ENOMEM, "Setting environment variable '%s' failed, ignoring: %m", value);
400 } else
401 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
402
403 } else if (streq(key, "systemd.machine_id") && value) {
404
405 r = set_machine_id(value);
406 if (r < 0)
407 log_warning("MachineID '%s' is not valid. Ignoring.", value);
408
409 } else if (streq(key, "quiet") && !value) {
410
411 if (arg_show_status == _SHOW_STATUS_UNSET)
412 arg_show_status = SHOW_STATUS_AUTO;
413
414 } else if (streq(key, "debug") && !value) {
415
416 /* Note that log_parse_environment() handles 'debug'
417 * too, and sets the log level to LOG_DEBUG. */
418
419 if (detect_container() > 0)
420 log_set_target(LOG_TARGET_CONSOLE);
421
422 } else if (!in_initrd() && !value) {
423 const char *target;
424
425 /* SysV compatibility */
426 target = runlevel_to_target(key);
427 if (target)
428 return free_and_strdup(&arg_default_unit, target);
429 }
430
431 return 0;
432 }
433
434 #define DEFINE_SETTER(name, func, descr) \
435 static int name(const char *unit, \
436 const char *filename, \
437 unsigned line, \
438 const char *section, \
439 unsigned section_line, \
440 const char *lvalue, \
441 int ltype, \
442 const char *rvalue, \
443 void *data, \
444 void *userdata) { \
445 \
446 int r; \
447 \
448 assert(filename); \
449 assert(lvalue); \
450 assert(rvalue); \
451 \
452 r = func(rvalue); \
453 if (r < 0) \
454 log_syntax(unit, LOG_ERR, filename, line, r, \
455 "Invalid " descr "'%s': %m", \
456 rvalue); \
457 \
458 return 0; \
459 }
460
461 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level")
462 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target")
463 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" )
464 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location")
465
466 static int config_parse_cpu_affinity2(
467 const char *unit,
468 const char *filename,
469 unsigned line,
470 const char *section,
471 unsigned section_line,
472 const char *lvalue,
473 int ltype,
474 const char *rvalue,
475 void *data,
476 void *userdata) {
477
478 _cleanup_cpu_free_ cpu_set_t *c = NULL;
479 int ncpus;
480
481 ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue);
482 if (ncpus < 0)
483 return ncpus;
484
485 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
486 log_warning("Failed to set CPU affinity: %m");
487
488 return 0;
489 }
490
491 static int config_parse_show_status(
492 const char* unit,
493 const char *filename,
494 unsigned line,
495 const char *section,
496 unsigned section_line,
497 const char *lvalue,
498 int ltype,
499 const char *rvalue,
500 void *data,
501 void *userdata) {
502
503 int k;
504 ShowStatus *b = data;
505
506 assert(filename);
507 assert(lvalue);
508 assert(rvalue);
509 assert(data);
510
511 k = parse_show_status(rvalue, b);
512 if (k < 0) {
513 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
514 return 0;
515 }
516
517 return 0;
518 }
519
520 static int config_parse_crash_chvt(
521 const char* unit,
522 const char *filename,
523 unsigned line,
524 const char *section,
525 unsigned section_line,
526 const char *lvalue,
527 int ltype,
528 const char *rvalue,
529 void *data,
530 void *userdata) {
531
532 int r;
533
534 assert(filename);
535 assert(lvalue);
536 assert(rvalue);
537
538 r = parse_crash_chvt(rvalue);
539 if (r < 0) {
540 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
541 return 0;
542 }
543
544 return 0;
545 }
546
547 static int config_parse_join_controllers(const char *unit,
548 const char *filename,
549 unsigned line,
550 const char *section,
551 unsigned section_line,
552 const char *lvalue,
553 int ltype,
554 const char *rvalue,
555 void *data,
556 void *userdata) {
557
558 const char *whole_rvalue = rvalue;
559 unsigned n = 0;
560
561 assert(filename);
562 assert(lvalue);
563 assert(rvalue);
564
565 arg_join_controllers = strv_free_free(arg_join_controllers);
566
567 for (;;) {
568 _cleanup_free_ char *word = NULL;
569 char **l;
570 int r;
571
572 r = extract_first_word(&rvalue, &word, WHITESPACE, EXTRACT_QUOTES);
573 if (r < 0) {
574 log_syntax(unit, LOG_ERR, filename, line, r, "Invalid value for %s: %s", lvalue, whole_rvalue);
575 return r;
576 }
577 if (r == 0)
578 break;
579
580 l = strv_split(word, ",");
581 if (!l)
582 return log_oom();
583 strv_uniq(l);
584
585 if (strv_length(l) <= 1) {
586 strv_free(l);
587 continue;
588 }
589
590 if (!arg_join_controllers) {
591 arg_join_controllers = new(char**, 2);
592 if (!arg_join_controllers) {
593 strv_free(l);
594 return log_oom();
595 }
596
597 arg_join_controllers[0] = l;
598 arg_join_controllers[1] = NULL;
599
600 n = 1;
601 } else {
602 char ***a;
603 char ***t;
604
605 t = new0(char**, n+2);
606 if (!t) {
607 strv_free(l);
608 return log_oom();
609 }
610
611 n = 0;
612
613 for (a = arg_join_controllers; *a; a++) {
614
615 if (strv_overlap(*a, l)) {
616 if (strv_extend_strv(&l, *a, false) < 0) {
617 strv_free(l);
618 strv_free_free(t);
619 return log_oom();
620 }
621
622 } else {
623 char **c;
624
625 c = strv_copy(*a);
626 if (!c) {
627 strv_free(l);
628 strv_free_free(t);
629 return log_oom();
630 }
631
632 t[n++] = c;
633 }
634 }
635
636 t[n++] = strv_uniq(l);
637
638 strv_free_free(arg_join_controllers);
639 arg_join_controllers = t;
640 }
641 }
642 if (!isempty(rvalue))
643 log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring.");
644
645 return 0;
646 }
647
648 static int parse_config_file(void) {
649
650 const ConfigTableItem items[] = {
651 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
652 { "Manager", "LogTarget", config_parse_target, 0, NULL },
653 { "Manager", "LogColor", config_parse_color, 0, NULL },
654 { "Manager", "LogLocation", config_parse_location, 0, NULL },
655 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
656 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL },
657 { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL },
658 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
659 { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
660 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
661 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
662 { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
663 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
664 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
665 { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
666 #ifdef HAVE_SECCOMP
667 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
668 #endif
669 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
670 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
671 { "Manager", "DefaultStandardOutput", config_parse_output, 0, &arg_default_std_output },
672 { "Manager", "DefaultStandardError", config_parse_output, 0, &arg_default_std_error },
673 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
674 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
675 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
676 { "Manager", "DefaultStartLimitInterval", 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", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
697 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
698 { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
699 { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max },
700 {}
701 };
702
703 const char *fn, *conf_dirs_nulstr;
704
705 fn = arg_running_as == MANAGER_SYSTEM ?
706 PKGSYSCONFDIR "/system.conf" :
707 PKGSYSCONFDIR "/user.conf";
708
709 conf_dirs_nulstr = arg_running_as == MANAGER_SYSTEM ?
710 CONF_PATHS_NULSTR("systemd/system.conf.d") :
711 CONF_PATHS_NULSTR("systemd/user.conf.d");
712
713 config_parse_many(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, false, NULL);
714
715 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
716 * like everywhere else. */
717 if (arg_default_timeout_start_usec <= 0)
718 arg_default_timeout_start_usec = USEC_INFINITY;
719 if (arg_default_timeout_stop_usec <= 0)
720 arg_default_timeout_stop_usec = USEC_INFINITY;
721
722 return 0;
723 }
724
725 static void manager_set_defaults(Manager *m) {
726
727 assert(m);
728
729 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
730 m->default_std_output = arg_default_std_output;
731 m->default_std_error = arg_default_std_error;
732 m->default_timeout_start_usec = arg_default_timeout_start_usec;
733 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
734 m->default_restart_usec = arg_default_restart_usec;
735 m->default_start_limit_interval = arg_default_start_limit_interval;
736 m->default_start_limit_burst = arg_default_start_limit_burst;
737 m->default_cpu_accounting = arg_default_cpu_accounting;
738 m->default_blockio_accounting = arg_default_blockio_accounting;
739 m->default_memory_accounting = arg_default_memory_accounting;
740 m->default_tasks_accounting = arg_default_tasks_accounting;
741 m->default_tasks_max = arg_default_tasks_max;
742
743 manager_set_default_rlimits(m, arg_default_rlimit);
744 manager_environment_add(m, NULL, arg_default_environment);
745 }
746
747 static int parse_argv(int argc, char *argv[]) {
748
749 enum {
750 ARG_LOG_LEVEL = 0x100,
751 ARG_LOG_TARGET,
752 ARG_LOG_COLOR,
753 ARG_LOG_LOCATION,
754 ARG_UNIT,
755 ARG_SYSTEM,
756 ARG_USER,
757 ARG_TEST,
758 ARG_NO_PAGER,
759 ARG_VERSION,
760 ARG_DUMP_CONFIGURATION_ITEMS,
761 ARG_DUMP_CORE,
762 ARG_CRASH_CHVT,
763 ARG_CRASH_SHELL,
764 ARG_CRASH_REBOOT,
765 ARG_CONFIRM_SPAWN,
766 ARG_SHOW_STATUS,
767 ARG_DESERIALIZE,
768 ARG_SWITCHED_ROOT,
769 ARG_DEFAULT_STD_OUTPUT,
770 ARG_DEFAULT_STD_ERROR,
771 ARG_MACHINE_ID
772 };
773
774 static const struct option options[] = {
775 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
776 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
777 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
778 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
779 { "unit", required_argument, NULL, ARG_UNIT },
780 { "system", no_argument, NULL, ARG_SYSTEM },
781 { "user", no_argument, NULL, ARG_USER },
782 { "test", no_argument, NULL, ARG_TEST },
783 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
784 { "help", no_argument, NULL, 'h' },
785 { "version", no_argument, NULL, ARG_VERSION },
786 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
787 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
788 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
789 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
790 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
791 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
792 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
793 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
794 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
795 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
796 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
797 { "machine-id", required_argument, NULL, ARG_MACHINE_ID },
798 {}
799 };
800
801 int c, r;
802
803 assert(argc >= 1);
804 assert(argv);
805
806 if (getpid() == 1)
807 opterr = 0;
808
809 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
810
811 switch (c) {
812
813 case ARG_LOG_LEVEL:
814 r = log_set_max_level_from_string(optarg);
815 if (r < 0) {
816 log_error("Failed to parse log level %s.", optarg);
817 return r;
818 }
819
820 break;
821
822 case ARG_LOG_TARGET:
823 r = log_set_target_from_string(optarg);
824 if (r < 0) {
825 log_error("Failed to parse log target %s.", optarg);
826 return r;
827 }
828
829 break;
830
831 case ARG_LOG_COLOR:
832
833 if (optarg) {
834 r = log_show_color_from_string(optarg);
835 if (r < 0) {
836 log_error("Failed to parse log color setting %s.", optarg);
837 return r;
838 }
839 } else
840 log_show_color(true);
841
842 break;
843
844 case ARG_LOG_LOCATION:
845 if (optarg) {
846 r = log_show_location_from_string(optarg);
847 if (r < 0) {
848 log_error("Failed to parse log location setting %s.", optarg);
849 return r;
850 }
851 } else
852 log_show_location(true);
853
854 break;
855
856 case ARG_DEFAULT_STD_OUTPUT:
857 r = exec_output_from_string(optarg);
858 if (r < 0) {
859 log_error("Failed to parse default standard output setting %s.", optarg);
860 return r;
861 } else
862 arg_default_std_output = r;
863 break;
864
865 case ARG_DEFAULT_STD_ERROR:
866 r = exec_output_from_string(optarg);
867 if (r < 0) {
868 log_error("Failed to parse default standard error output setting %s.", optarg);
869 return r;
870 } else
871 arg_default_std_error = r;
872 break;
873
874 case ARG_UNIT:
875
876 r = free_and_strdup(&arg_default_unit, optarg);
877 if (r < 0)
878 return log_error_errno(r, "Failed to set default unit %s: %m", optarg);
879
880 break;
881
882 case ARG_SYSTEM:
883 arg_running_as = MANAGER_SYSTEM;
884 break;
885
886 case ARG_USER:
887 arg_running_as = MANAGER_USER;
888 break;
889
890 case ARG_TEST:
891 arg_action = ACTION_TEST;
892 if (arg_no_pager < 0)
893 arg_no_pager = true;
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 if (arg_no_pager < 0)
1004 arg_no_pager = true;
1005 break;
1006
1007 case 'D':
1008 log_set_max_level(LOG_DEBUG);
1009 break;
1010
1011 case 'b':
1012 case 's':
1013 case 'z':
1014 /* Just to eat away the sysvinit kernel
1015 * cmdline args without getopt() error
1016 * messages that we'll parse in
1017 * parse_proc_cmdline_word() or ignore. */
1018
1019 case '?':
1020 if (getpid() != 1)
1021 return -EINVAL;
1022 else
1023 return 0;
1024
1025 default:
1026 assert_not_reached("Unhandled option code.");
1027 }
1028
1029 if (optind < argc && getpid() != 1) {
1030 /* Hmm, when we aren't run as init system
1031 * let's complain about excess arguments */
1032
1033 log_error("Excess arguments.");
1034 return -EINVAL;
1035 }
1036
1037 return 0;
1038 }
1039
1040 static int help(void) {
1041
1042 printf("%s [OPTIONS...]\n\n"
1043 "Starts up and maintains the system or user services.\n\n"
1044 " -h --help Show this help\n"
1045 " --test Determine startup sequence, dump it and exit\n"
1046 " --no-pager Do not pipe output into a pager\n"
1047 " --dump-configuration-items Dump understood unit configuration items\n"
1048 " --unit=UNIT Set default unit\n"
1049 " --system Run a system instance, even if PID != 1\n"
1050 " --user Run a user instance\n"
1051 " --dump-core[=BOOL] Dump core on crash\n"
1052 " --crash-vt=NR Change to specified VT on crash\n"
1053 " --crash-reboot[=BOOL] Reboot on crash\n"
1054 " --crash-shell[=BOOL] Run shell on crash\n"
1055 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1056 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1057 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1058 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1059 " --log-color[=BOOL] Highlight important log messages\n"
1060 " --log-location[=BOOL] Include code location in log messages\n"
1061 " --default-standard-output= Set default standard output for services\n"
1062 " --default-standard-error= Set default standard error output for services\n",
1063 program_invocation_short_name);
1064
1065 return 0;
1066 }
1067
1068 static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
1069 _cleanup_fdset_free_ FDSet *fds = NULL;
1070 _cleanup_fclose_ FILE *f = NULL;
1071 int r;
1072
1073 assert(m);
1074 assert(_f);
1075 assert(_fds);
1076
1077 r = manager_open_serialization(m, &f);
1078 if (r < 0)
1079 return log_error_errno(r, "Failed to create serialization file: %m");
1080
1081 /* Make sure nothing is really destructed when we shut down */
1082 m->n_reloading ++;
1083 bus_manager_send_reloading(m, true);
1084
1085 fds = fdset_new();
1086 if (!fds)
1087 return log_oom();
1088
1089 r = manager_serialize(m, f, fds, switching_root);
1090 if (r < 0)
1091 return log_error_errno(r, "Failed to serialize state: %m");
1092
1093 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1094 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1095
1096 r = fd_cloexec(fileno(f), false);
1097 if (r < 0)
1098 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1099
1100 r = fdset_cloexec(fds, false);
1101 if (r < 0)
1102 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1103
1104 *_f = f;
1105 *_fds = fds;
1106
1107 f = NULL;
1108 fds = NULL;
1109
1110 return 0;
1111 }
1112
1113 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1114 struct rlimit nl;
1115 int r;
1116
1117 assert(saved_rlimit);
1118
1119 /* Save the original RLIMIT_NOFILE so that we can reset it
1120 * later when transitioning from the initrd to the main
1121 * systemd or suchlike. */
1122 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1123 return log_error_errno(errno, "Reading RLIMIT_NOFILE failed: %m");
1124
1125 /* Make sure forked processes get the default kernel setting */
1126 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1127 struct rlimit *rl;
1128
1129 rl = newdup(struct rlimit, saved_rlimit, 1);
1130 if (!rl)
1131 return log_oom();
1132
1133 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1134 }
1135
1136 /* Bump up the resource limit for ourselves substantially */
1137 nl.rlim_cur = nl.rlim_max = 64*1024;
1138 r = setrlimit_closest(RLIMIT_NOFILE, &nl);
1139 if (r < 0)
1140 return log_error_errno(r, "Setting RLIMIT_NOFILE failed: %m");
1141
1142 return 0;
1143 }
1144
1145 static void test_usr(void) {
1146
1147 /* Check that /usr is not a separate fs */
1148
1149 if (dir_is_empty("/usr") <= 0)
1150 return;
1151
1152 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1153 "Some things will probably break (sometimes even silently) in mysterious ways. "
1154 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1155 }
1156
1157 static int initialize_join_controllers(void) {
1158 /* By default, mount "cpu" + "cpuacct" together, and "net_cls"
1159 * + "net_prio". We'd like to add "cpuset" to the mix, but
1160 * "cpuset" doesn't really work for groups with no initialized
1161 * attributes. */
1162
1163 arg_join_controllers = new(char**, 3);
1164 if (!arg_join_controllers)
1165 return -ENOMEM;
1166
1167 arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL);
1168 if (!arg_join_controllers[0])
1169 goto oom;
1170
1171 arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL);
1172 if (!arg_join_controllers[1])
1173 goto oom;
1174
1175 arg_join_controllers[2] = NULL;
1176 return 0;
1177
1178 oom:
1179 arg_join_controllers = strv_free_free(arg_join_controllers);
1180 return -ENOMEM;
1181 }
1182
1183 static int enforce_syscall_archs(Set *archs) {
1184 #ifdef HAVE_SECCOMP
1185 scmp_filter_ctx *seccomp;
1186 Iterator i;
1187 void *id;
1188 int r;
1189
1190 seccomp = seccomp_init(SCMP_ACT_ALLOW);
1191 if (!seccomp)
1192 return log_oom();
1193
1194 SET_FOREACH(id, arg_syscall_archs, i) {
1195 r = seccomp_arch_add(seccomp, PTR_TO_UINT32(id) - 1);
1196 if (r == -EEXIST)
1197 continue;
1198 if (r < 0) {
1199 log_error_errno(r, "Failed to add architecture to seccomp: %m");
1200 goto finish;
1201 }
1202 }
1203
1204 r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
1205 if (r < 0) {
1206 log_error_errno(r, "Failed to unset NO_NEW_PRIVS: %m");
1207 goto finish;
1208 }
1209
1210 r = seccomp_load(seccomp);
1211 if (r < 0)
1212 log_error_errno(r, "Failed to add install architecture seccomp: %m");
1213
1214 finish:
1215 seccomp_release(seccomp);
1216 return r;
1217 #else
1218 return 0;
1219 #endif
1220 }
1221
1222 static int status_welcome(void) {
1223 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1224 int r;
1225
1226 r = parse_env_file("/etc/os-release", NEWLINE,
1227 "PRETTY_NAME", &pretty_name,
1228 "ANSI_COLOR", &ansi_color,
1229 NULL);
1230 if (r == -ENOENT)
1231 r = parse_env_file("/usr/lib/os-release", NEWLINE,
1232 "PRETTY_NAME", &pretty_name,
1233 "ANSI_COLOR", &ansi_color,
1234 NULL);
1235
1236 if (r < 0 && r != -ENOENT)
1237 log_warning_errno(r, "Failed to read os-release file: %m");
1238
1239 return status_printf(NULL, false, false,
1240 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1241 isempty(ansi_color) ? "1" : ansi_color,
1242 isempty(pretty_name) ? "Linux" : pretty_name);
1243 }
1244
1245 static int write_container_id(void) {
1246 const char *c;
1247 int r;
1248
1249 c = getenv("container");
1250 if (isempty(c))
1251 return 0;
1252
1253 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1254 if (r < 0)
1255 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
1256
1257 return 1;
1258 }
1259
1260 static int bump_unix_max_dgram_qlen(void) {
1261 _cleanup_free_ char *qlen = NULL;
1262 unsigned long v;
1263 int r;
1264
1265 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel
1266 * default of 16 is simply too low. We set the value really
1267 * really early during boot, so that it is actually applied to
1268 * all our sockets, including the $NOTIFY_SOCKET one. */
1269
1270 r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1271 if (r < 0)
1272 return log_warning_errno(r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1273
1274 r = safe_atolu(qlen, &v);
1275 if (r < 0)
1276 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length, ignoring: %m");
1277
1278 if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1279 return 0;
1280
1281 qlen = mfree(qlen);
1282 if (asprintf(&qlen, "%lu\n", DEFAULT_UNIX_MAX_DGRAM_QLEN) < 0)
1283 return log_oom();
1284
1285 r = write_string_file("/proc/sys/net/unix/max_dgram_qlen", qlen, 0);
1286 if (r < 0)
1287 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1288 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1289
1290 return 1;
1291 }
1292
1293 int main(int argc, char *argv[]) {
1294 Manager *m = NULL;
1295 int r, retval = EXIT_FAILURE;
1296 usec_t before_startup, after_startup;
1297 char timespan[FORMAT_TIMESPAN_MAX];
1298 FDSet *fds = NULL;
1299 bool reexecute = false;
1300 const char *shutdown_verb = NULL;
1301 dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL;
1302 dual_timestamp userspace_timestamp = DUAL_TIMESTAMP_NULL;
1303 dual_timestamp kernel_timestamp = DUAL_TIMESTAMP_NULL;
1304 dual_timestamp security_start_timestamp = DUAL_TIMESTAMP_NULL;
1305 dual_timestamp security_finish_timestamp = DUAL_TIMESTAMP_NULL;
1306 static char systemd[] = "systemd";
1307 bool skip_setup = false;
1308 unsigned j;
1309 bool loaded_policy = false;
1310 bool arm_reboot_watchdog = false;
1311 bool queue_default_job = false;
1312 bool empty_etc = false;
1313 char *switch_root_dir = NULL, *switch_root_init = NULL;
1314 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0);
1315 const char *error_message = NULL;
1316
1317 #ifdef HAVE_SYSV_COMPAT
1318 if (getpid() != 1 && strstr(program_invocation_short_name, "init")) {
1319 /* This is compatibility support for SysV, where
1320 * calling init as a user is identical to telinit. */
1321
1322 errno = -ENOENT;
1323 execv(SYSTEMCTL_BINARY_PATH, argv);
1324 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1325 return 1;
1326 }
1327 #endif
1328
1329 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
1330 dual_timestamp_get(&userspace_timestamp);
1331
1332 /* Determine if this is a reexecution or normal bootup. We do
1333 * the full command line parsing much later, so let's just
1334 * have a quick peek here. */
1335 if (strv_find(argv+1, "--deserialize"))
1336 skip_setup = true;
1337
1338 /* If we have switched root, do all the special setup
1339 * things */
1340 if (strv_find(argv+1, "--switched-root"))
1341 skip_setup = false;
1342
1343 /* If we get started via the /sbin/init symlink then we are
1344 called 'init'. After a subsequent reexecution we are then
1345 called 'systemd'. That is confusing, hence let's call us
1346 systemd right-away. */
1347 program_invocation_short_name = systemd;
1348 prctl(PR_SET_NAME, systemd);
1349
1350 saved_argv = argv;
1351 saved_argc = argc;
1352
1353 log_show_color(isatty(STDERR_FILENO) > 0);
1354 log_set_upgrade_syslog_to_journal(true);
1355
1356 /* Disable the umask logic */
1357 if (getpid() == 1)
1358 umask(0);
1359
1360 if (getpid() == 1 && detect_container() <= 0) {
1361
1362 /* Running outside of a container as PID 1 */
1363 arg_running_as = MANAGER_SYSTEM;
1364 make_null_stdio();
1365 log_set_target(LOG_TARGET_KMSG);
1366 log_open();
1367
1368 if (in_initrd())
1369 initrd_timestamp = userspace_timestamp;
1370
1371 if (!skip_setup) {
1372 r = mount_setup_early();
1373 if (r < 0) {
1374 error_message = "Failed to early mount API filesystems";
1375 goto finish;
1376 }
1377 dual_timestamp_get(&security_start_timestamp);
1378 if (mac_selinux_setup(&loaded_policy) < 0) {
1379 error_message = "Failed to load SELinux policy";
1380 goto finish;
1381 } else if (ima_setup() < 0) {
1382 error_message = "Failed to load IMA policy";
1383 goto finish;
1384 } else if (mac_smack_setup(&loaded_policy) < 0) {
1385 error_message = "Failed to load SMACK policy";
1386 goto finish;
1387 }
1388 dual_timestamp_get(&security_finish_timestamp);
1389 }
1390
1391 if (mac_selinux_init(NULL) < 0) {
1392 error_message = "Failed to initialize SELinux policy";
1393 goto finish;
1394 }
1395
1396 if (!skip_setup) {
1397 if (clock_is_localtime() > 0) {
1398 int min;
1399
1400 /*
1401 * The very first call of settimeofday() also does a time warp in the kernel.
1402 *
1403 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on
1404 * external tools to take care of maintaining the RTC and do all adjustments.
1405 * This matches the behavior of Windows, which leaves the RTC alone if the
1406 * registry tells that the RTC runs in UTC.
1407 */
1408 r = clock_set_timezone(&min);
1409 if (r < 0)
1410 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1411 else
1412 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1413 } else if (!in_initrd()) {
1414 /*
1415 * Do a dummy very first call to seal the kernel's time warp magic.
1416 *
1417 * Do not call this this from inside the initrd. The initrd might not
1418 * carry /etc/adjtime with LOCAL, but the real system could be set up
1419 * that way. In such case, we need to delay the time-warp or the sealing
1420 * until we reach the real system.
1421 *
1422 * Do no set the kernel's timezone. The concept of local time cannot
1423 * be supported reliably, the time will jump or be incorrect at every daylight
1424 * saving time change. All kernel local time concepts will be treated
1425 * as UTC that way.
1426 */
1427 clock_reset_timewarp();
1428 }
1429 }
1430
1431 /* Set the default for later on, but don't actually
1432 * open the logs like this for now. Note that if we
1433 * are transitioning from the initrd there might still
1434 * be journal fd open, and we shouldn't attempt
1435 * opening that before we parsed /proc/cmdline which
1436 * might redirect output elsewhere. */
1437 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
1438
1439 } else if (getpid() == 1) {
1440 /* Running inside a container, as PID 1 */
1441 arg_running_as = MANAGER_SYSTEM;
1442 log_set_target(LOG_TARGET_CONSOLE);
1443 log_close_console(); /* force reopen of /dev/console */
1444 log_open();
1445
1446 /* For the later on, see above... */
1447 log_set_target(LOG_TARGET_JOURNAL);
1448
1449 /* clear the kernel timestamp,
1450 * because we are in a container */
1451 kernel_timestamp.monotonic = 0ULL;
1452 kernel_timestamp.realtime = 0ULL;
1453
1454 } else {
1455 /* Running as user instance */
1456 arg_running_as = MANAGER_USER;
1457 log_set_target(LOG_TARGET_AUTO);
1458 log_open();
1459
1460 /* clear the kernel timestamp,
1461 * because we are not PID 1 */
1462 kernel_timestamp = DUAL_TIMESTAMP_NULL;
1463 }
1464
1465 /* Initialize default unit */
1466 r = free_and_strdup(&arg_default_unit, SPECIAL_DEFAULT_TARGET);
1467 if (r < 0) {
1468 log_emergency_errno(r, "Failed to set default unit %s: %m", SPECIAL_DEFAULT_TARGET);
1469 error_message = "Failed to set default unit";
1470 goto finish;
1471 }
1472
1473 r = initialize_join_controllers();
1474 if (r < 0) {
1475 error_message = "Failed to initialize cgroup controllers";
1476 goto finish;
1477 }
1478
1479 /* Mount /proc, /sys and friends, so that /proc/cmdline and
1480 * /proc/$PID/fd is available. */
1481 if (getpid() == 1) {
1482
1483 /* Load the kernel modules early, so that we kdbus.ko is loaded before kdbusfs shall be mounted */
1484 if (!skip_setup)
1485 kmod_setup();
1486
1487 r = mount_setup(loaded_policy);
1488 if (r < 0) {
1489 error_message = "Failed to mount API filesystems";
1490 goto finish;
1491 }
1492 }
1493
1494 /* Reset all signal handlers. */
1495 (void) reset_all_signal_handlers();
1496 (void) ignore_signals(SIGNALS_IGNORE, -1);
1497
1498 if (parse_config_file() < 0) {
1499 error_message = "Failed to parse config file";
1500 goto finish;
1501 }
1502
1503 if (arg_running_as == MANAGER_SYSTEM) {
1504 r = parse_proc_cmdline(parse_proc_cmdline_item);
1505 if (r < 0)
1506 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
1507 }
1508
1509 /* Note that this also parses bits from the kernel command
1510 * line, including "debug". */
1511 log_parse_environment();
1512
1513 if (parse_argv(argc, argv) < 0) {
1514 error_message = "Failed to parse commandline arguments";
1515 goto finish;
1516 }
1517
1518 if (arg_action == ACTION_TEST &&
1519 geteuid() == 0) {
1520 log_error("Don't run test mode as root.");
1521 goto finish;
1522 }
1523
1524 if (arg_running_as == MANAGER_USER &&
1525 arg_action == ACTION_RUN &&
1526 sd_booted() <= 0) {
1527 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
1528 goto finish;
1529 }
1530
1531 if (arg_running_as == MANAGER_SYSTEM &&
1532 arg_action == ACTION_RUN &&
1533 running_in_chroot() > 0) {
1534 log_error("Cannot be run in a chroot() environment.");
1535 goto finish;
1536 }
1537
1538 if (arg_action == ACTION_TEST)
1539 skip_setup = true;
1540
1541 pager_open_if_enabled();
1542
1543 if (arg_action == ACTION_HELP) {
1544 retval = help();
1545 goto finish;
1546 } else if (arg_action == ACTION_VERSION) {
1547 retval = version();
1548 goto finish;
1549 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
1550 unit_dump_config_items(stdout);
1551 retval = EXIT_SUCCESS;
1552 goto finish;
1553 } else if (arg_action == ACTION_DONE) {
1554 retval = EXIT_SUCCESS;
1555 goto finish;
1556 }
1557
1558 if (arg_running_as == MANAGER_USER &&
1559 !getenv("XDG_RUNTIME_DIR")) {
1560 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
1561 goto finish;
1562 }
1563
1564 assert_se(arg_action == ACTION_RUN || arg_action == ACTION_TEST);
1565
1566 /* Close logging fds, in order not to confuse fdset below */
1567 log_close();
1568
1569 /* Remember open file descriptors for later deserialization */
1570 r = fdset_new_fill(&fds);
1571 if (r < 0) {
1572 log_emergency_errno(r, "Failed to allocate fd set: %m");
1573 error_message = "Failed to allocate fd set";
1574 goto finish;
1575 } else
1576 fdset_cloexec(fds, true);
1577
1578 if (arg_serialization)
1579 assert_se(fdset_remove(fds, fileno(arg_serialization)) >= 0);
1580
1581 if (arg_running_as == MANAGER_SYSTEM)
1582 /* Become a session leader if we aren't one yet. */
1583 setsid();
1584
1585 /* Move out of the way, so that we won't block unmounts */
1586 assert_se(chdir("/") == 0);
1587
1588 /* Reset the console, but only if this is really init and we
1589 * are freshly booted */
1590 if (arg_running_as == MANAGER_SYSTEM && arg_action == ACTION_RUN) {
1591
1592 /* If we are init, we connect stdin/stdout/stderr to
1593 * /dev/null and make sure we don't have a controlling
1594 * tty. */
1595 release_terminal();
1596
1597 if (getpid() == 1 && !skip_setup)
1598 console_setup();
1599 }
1600
1601 /* Open the logging devices, if possible and necessary */
1602 log_open();
1603
1604 if (arg_show_status == _SHOW_STATUS_UNSET)
1605 arg_show_status = SHOW_STATUS_YES;
1606
1607 /* Make sure we leave a core dump without panicing the
1608 * kernel. */
1609 if (getpid() == 1) {
1610 install_crash_handler();
1611
1612 r = mount_cgroup_controllers(arg_join_controllers);
1613 if (r < 0)
1614 goto finish;
1615 }
1616
1617 if (arg_running_as == MANAGER_SYSTEM) {
1618 int v;
1619
1620 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1621 arg_action == ACTION_TEST ? "test " : "" );
1622
1623 v = detect_virtualization();
1624 if (v > 0)
1625 log_info("Detected virtualization %s.", virtualization_to_string(v));
1626
1627 write_container_id();
1628
1629 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1630
1631 if (in_initrd())
1632 log_info("Running in initial RAM disk.");
1633
1634 /* Let's check whether /etc is already populated. We
1635 * don't actually really check for that, but use
1636 * /etc/machine-id as flag file. This allows container
1637 * managers and installers to provision a couple of
1638 * files already. If the container manager wants to
1639 * provision the machine ID itself it should pass
1640 * $container_uuid to PID 1. */
1641
1642 empty_etc = access("/etc/machine-id", F_OK) < 0;
1643 if (empty_etc)
1644 log_info("Running with unpopulated /etc.");
1645 } else {
1646 _cleanup_free_ char *t;
1647
1648 t = uid_to_name(getuid());
1649 log_debug(PACKAGE_STRING " running in %suser mode for user "UID_FMT"/%s. (" SYSTEMD_FEATURES ")",
1650 arg_action == ACTION_TEST ? " test" : "", getuid(), t);
1651 }
1652
1653 if (arg_running_as == MANAGER_SYSTEM && !skip_setup) {
1654 if (arg_show_status > 0)
1655 status_welcome();
1656
1657 hostname_setup();
1658 machine_id_setup(NULL, arg_machine_id);
1659 loopback_setup();
1660 bump_unix_max_dgram_qlen();
1661
1662 test_usr();
1663 }
1664
1665 if (arg_running_as == MANAGER_SYSTEM && arg_runtime_watchdog > 0)
1666 watchdog_set_timeout(&arg_runtime_watchdog);
1667
1668 if (arg_timer_slack_nsec != NSEC_INFINITY)
1669 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1670 log_error_errno(errno, "Failed to adjust timer slack: %m");
1671
1672 if (!cap_test_all(arg_capability_bounding_set)) {
1673 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
1674 if (r < 0) {
1675 log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
1676 error_message = "Failed to drop capability bounding set of usermode helpers";
1677 goto finish;
1678 }
1679 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
1680 if (r < 0) {
1681 log_emergency_errno(r, "Failed to drop capability bounding set: %m");
1682 error_message = "Failed to drop capability bounding set";
1683 goto finish;
1684 }
1685 }
1686
1687 if (arg_syscall_archs) {
1688 r = enforce_syscall_archs(arg_syscall_archs);
1689 if (r < 0) {
1690 error_message = "Failed to set syscall architectures";
1691 goto finish;
1692 }
1693 }
1694
1695 if (arg_running_as == MANAGER_USER)
1696 /* Become reaper of our children */
1697 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
1698 log_warning_errno(errno, "Failed to make us a subreaper: %m");
1699
1700 if (arg_running_as == MANAGER_SYSTEM) {
1701 bump_rlimit_nofile(&saved_rlimit_nofile);
1702
1703 if (empty_etc) {
1704 r = unit_file_preset_all(UNIT_FILE_SYSTEM, false, NULL, UNIT_FILE_PRESET_ENABLE_ONLY, false, NULL, 0);
1705 if (r < 0)
1706 log_full_errno(r == -EEXIST ? LOG_NOTICE : LOG_WARNING, r, "Failed to populate /etc with preset unit settings, ignoring: %m");
1707 else
1708 log_info("Populated /etc with preset unit settings.");
1709 }
1710 }
1711
1712 r = manager_new(arg_running_as, arg_action == ACTION_TEST, &m);
1713 if (r < 0) {
1714 log_emergency_errno(r, "Failed to allocate manager object: %m");
1715 error_message = "Failed to allocate manager object";
1716 goto finish;
1717 }
1718
1719 m->confirm_spawn = arg_confirm_spawn;
1720 m->runtime_watchdog = arg_runtime_watchdog;
1721 m->shutdown_watchdog = arg_shutdown_watchdog;
1722 m->userspace_timestamp = userspace_timestamp;
1723 m->kernel_timestamp = kernel_timestamp;
1724 m->initrd_timestamp = initrd_timestamp;
1725 m->security_start_timestamp = security_start_timestamp;
1726 m->security_finish_timestamp = security_finish_timestamp;
1727
1728 manager_set_defaults(m);
1729 manager_set_show_status(m, arg_show_status);
1730 manager_set_first_boot(m, empty_etc);
1731
1732 /* Remember whether we should queue the default job */
1733 queue_default_job = !arg_serialization || arg_switched_root;
1734
1735 before_startup = now(CLOCK_MONOTONIC);
1736
1737 r = manager_startup(m, arg_serialization, fds);
1738 if (r < 0)
1739 log_error_errno(r, "Failed to fully start up daemon: %m");
1740
1741 /* This will close all file descriptors that were opened, but
1742 * not claimed by any unit. */
1743 fds = fdset_free(fds);
1744
1745 arg_serialization = safe_fclose(arg_serialization);
1746
1747 if (queue_default_job) {
1748 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
1749 Unit *target = NULL;
1750 Job *default_unit_job;
1751
1752 log_debug("Activating default unit: %s", arg_default_unit);
1753
1754 r = manager_load_unit(m, arg_default_unit, NULL, &error, &target);
1755 if (r < 0)
1756 log_error("Failed to load default target: %s", bus_error_message(&error, r));
1757 else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND)
1758 log_error_errno(target->load_error, "Failed to load default target: %m");
1759 else if (target->load_state == UNIT_MASKED)
1760 log_error("Default target masked.");
1761
1762 if (!target || target->load_state != UNIT_LOADED) {
1763 log_info("Trying to load rescue target...");
1764
1765 r = manager_load_unit(m, SPECIAL_RESCUE_TARGET, NULL, &error, &target);
1766 if (r < 0) {
1767 log_emergency("Failed to load rescue target: %s", bus_error_message(&error, r));
1768 error_message = "Failed to load rescue target";
1769 goto finish;
1770 } else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND) {
1771 log_emergency_errno(target->load_error, "Failed to load rescue target: %m");
1772 error_message = "Failed to load rescue target";
1773 goto finish;
1774 } else if (target->load_state == UNIT_MASKED) {
1775 log_emergency("Rescue target masked.");
1776 error_message = "Rescue target masked";
1777 goto finish;
1778 }
1779 }
1780
1781 assert(target->load_state == UNIT_LOADED);
1782
1783 if (arg_action == ACTION_TEST) {
1784 printf("-> By units:\n");
1785 manager_dump_units(m, stdout, "\t");
1786 }
1787
1788 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, &error, &default_unit_job);
1789 if (r == -EPERM) {
1790 log_debug("Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1791
1792 sd_bus_error_free(&error);
1793
1794 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, &error, &default_unit_job);
1795 if (r < 0) {
1796 log_emergency("Failed to start default target: %s", bus_error_message(&error, r));
1797 error_message = "Failed to start default target";
1798 goto finish;
1799 }
1800 } else if (r < 0) {
1801 log_emergency("Failed to isolate default target: %s", bus_error_message(&error, r));
1802 error_message = "Failed to isolate default target";
1803 goto finish;
1804 }
1805
1806 m->default_unit_job_id = default_unit_job->id;
1807
1808 after_startup = now(CLOCK_MONOTONIC);
1809 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
1810 "Loaded units and determined initial transaction in %s.",
1811 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
1812
1813 if (arg_action == ACTION_TEST) {
1814 printf("-> By jobs:\n");
1815 manager_dump_jobs(m, stdout, "\t");
1816 retval = EXIT_SUCCESS;
1817 goto finish;
1818 }
1819 }
1820
1821 for (;;) {
1822 r = manager_loop(m);
1823 if (r < 0) {
1824 log_emergency_errno(r, "Failed to run main loop: %m");
1825 error_message = "Failed to run main loop";
1826 goto finish;
1827 }
1828
1829 switch (m->exit_code) {
1830
1831 case MANAGER_RELOAD:
1832 log_info("Reloading.");
1833
1834 r = parse_config_file();
1835 if (r < 0)
1836 log_error("Failed to parse config file.");
1837
1838 manager_set_defaults(m);
1839
1840 r = manager_reload(m);
1841 if (r < 0)
1842 log_error_errno(r, "Failed to reload: %m");
1843 break;
1844
1845 case MANAGER_REEXECUTE:
1846
1847 if (prepare_reexecute(m, &arg_serialization, &fds, false) < 0) {
1848 error_message = "Failed to prepare for reexecution";
1849 goto finish;
1850 }
1851
1852 reexecute = true;
1853 log_notice("Reexecuting.");
1854 goto finish;
1855
1856 case MANAGER_SWITCH_ROOT:
1857 /* Steal the switch root parameters */
1858 switch_root_dir = m->switch_root;
1859 switch_root_init = m->switch_root_init;
1860 m->switch_root = m->switch_root_init = NULL;
1861
1862 if (!switch_root_init)
1863 if (prepare_reexecute(m, &arg_serialization, &fds, true) < 0) {
1864 error_message = "Failed to prepare for reexecution";
1865 goto finish;
1866 }
1867
1868 reexecute = true;
1869 log_notice("Switching root.");
1870 goto finish;
1871
1872 case MANAGER_EXIT:
1873 retval = m->return_value;
1874
1875 if (m->running_as == MANAGER_USER) {
1876 log_debug("Exit.");
1877 goto finish;
1878 }
1879
1880 /* fallthrough */
1881 case MANAGER_REBOOT:
1882 case MANAGER_POWEROFF:
1883 case MANAGER_HALT:
1884 case MANAGER_KEXEC: {
1885 static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
1886 [MANAGER_EXIT] = "exit",
1887 [MANAGER_REBOOT] = "reboot",
1888 [MANAGER_POWEROFF] = "poweroff",
1889 [MANAGER_HALT] = "halt",
1890 [MANAGER_KEXEC] = "kexec"
1891 };
1892
1893 assert_se(shutdown_verb = table[m->exit_code]);
1894 arm_reboot_watchdog = m->exit_code == MANAGER_REBOOT;
1895
1896 log_notice("Shutting down.");
1897 goto finish;
1898 }
1899
1900 default:
1901 assert_not_reached("Unknown exit code.");
1902 }
1903 }
1904
1905 finish:
1906 pager_close();
1907
1908 if (m)
1909 arg_shutdown_watchdog = m->shutdown_watchdog;
1910
1911 m = manager_free(m);
1912
1913 for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++)
1914 arg_default_rlimit[j] = mfree(arg_default_rlimit[j]);
1915
1916 arg_default_unit = mfree(arg_default_unit);
1917 arg_join_controllers = strv_free_free(arg_join_controllers);
1918 arg_default_environment = strv_free(arg_default_environment);
1919 arg_syscall_archs = set_free(arg_syscall_archs);
1920
1921 mac_selinux_finish();
1922
1923 if (reexecute) {
1924 const char **args;
1925 unsigned i, args_size;
1926
1927 /* Close and disarm the watchdog, so that the new
1928 * instance can reinitialize it, but doesn't get
1929 * rebooted while we do that */
1930 watchdog_close(true);
1931
1932 /* Reset the RLIMIT_NOFILE to the kernel default, so
1933 * that the new systemd can pass the kernel default to
1934 * its child processes */
1935 if (saved_rlimit_nofile.rlim_cur > 0)
1936 (void) setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile);
1937
1938 if (switch_root_dir) {
1939 /* Kill all remaining processes from the
1940 * initrd, but don't wait for them, so that we
1941 * can handle the SIGCHLD for them after
1942 * deserializing. */
1943 broadcast_signal(SIGTERM, false, true);
1944
1945 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1946 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1947 if (r < 0)
1948 log_error_errno(r, "Failed to switch root, trying to continue: %m");
1949 }
1950
1951 args_size = MAX(6, argc+1);
1952 args = newa(const char*, args_size);
1953
1954 if (!switch_root_init) {
1955 char sfd[DECIMAL_STR_MAX(int) + 1];
1956
1957 /* First try to spawn ourselves with the right
1958 * path, and with full serialization. We do
1959 * this only if the user didn't specify an
1960 * explicit init to spawn. */
1961
1962 assert(arg_serialization);
1963 assert(fds);
1964
1965 xsprintf(sfd, "%i", fileno(arg_serialization));
1966
1967 i = 0;
1968 args[i++] = SYSTEMD_BINARY_PATH;
1969 if (switch_root_dir)
1970 args[i++] = "--switched-root";
1971 args[i++] = arg_running_as == MANAGER_SYSTEM ? "--system" : "--user";
1972 args[i++] = "--deserialize";
1973 args[i++] = sfd;
1974 args[i++] = NULL;
1975
1976 /* do not pass along the environment we inherit from the kernel or initrd */
1977 if (switch_root_dir)
1978 (void) clearenv();
1979
1980 assert(i <= args_size);
1981
1982 /*
1983 * We want valgrind to print its memory usage summary before reexecution.
1984 * Valgrind won't do this is on its own on exec(), but it will do it on exit().
1985 * Hence, to ensure we get a summary here, fork() off a child, let it exit() cleanly,
1986 * so that it prints the summary, and wait() for it in the parent, before proceeding into the exec().
1987 */
1988 valgrind_summary_hack();
1989
1990 (void) execv(args[0], (char* const*) args);
1991 }
1992
1993 /* Try the fallback, if there is any, without any
1994 * serialization. We pass the original argv[] and
1995 * envp[]. (Well, modulo the ordering changes due to
1996 * getopt() in argv[], and some cleanups in envp[],
1997 * but let's hope that doesn't matter.) */
1998
1999 arg_serialization = safe_fclose(arg_serialization);
2000 fds = fdset_free(fds);
2001
2002 /* Reopen the console */
2003 (void) make_console_stdio();
2004
2005 for (j = 1, i = 1; j < (unsigned) argc; j++)
2006 args[i++] = argv[j];
2007 args[i++] = NULL;
2008 assert(i <= args_size);
2009
2010 /* Reenable any blocked signals, especially important
2011 * if we switch from initial ramdisk to init=... */
2012 (void) reset_all_signal_handlers();
2013 (void) reset_signal_mask();
2014
2015 if (switch_root_init) {
2016 args[0] = switch_root_init;
2017 (void) execv(args[0], (char* const*) args);
2018 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
2019 }
2020
2021 args[0] = "/sbin/init";
2022 (void) execv(args[0], (char* const*) args);
2023
2024 if (errno == ENOENT) {
2025 log_warning("No /sbin/init, trying fallback");
2026
2027 args[0] = "/bin/sh";
2028 args[1] = NULL;
2029 (void) execv(args[0], (char* const*) args);
2030 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
2031 } else
2032 log_warning_errno(errno, "Failed to execute /sbin/init, giving up: %m");
2033 }
2034
2035 arg_serialization = safe_fclose(arg_serialization);
2036 fds = fdset_free(fds);
2037
2038 #ifdef HAVE_VALGRIND_VALGRIND_H
2039 /* If we are PID 1 and running under valgrind, then let's exit
2040 * here explicitly. valgrind will only generate nice output on
2041 * exit(), not on exec(), hence let's do the former not the
2042 * latter here. */
2043 if (getpid() == 1 && RUNNING_ON_VALGRIND)
2044 return 0;
2045 #endif
2046
2047 if (shutdown_verb) {
2048 char log_level[DECIMAL_STR_MAX(int) + 1];
2049 char exit_code[DECIMAL_STR_MAX(uint8_t) + 1];
2050 const char* command_line[11] = {
2051 SYSTEMD_SHUTDOWN_BINARY_PATH,
2052 shutdown_verb,
2053 "--log-level", log_level,
2054 "--log-target",
2055 };
2056 unsigned pos = 5;
2057 _cleanup_strv_free_ char **env_block = NULL;
2058
2059 assert(command_line[pos] == NULL);
2060 env_block = strv_copy(environ);
2061
2062 xsprintf(log_level, "%d", log_get_max_level());
2063
2064 switch (log_get_target()) {
2065
2066 case LOG_TARGET_KMSG:
2067 case LOG_TARGET_JOURNAL_OR_KMSG:
2068 case LOG_TARGET_SYSLOG_OR_KMSG:
2069 command_line[pos++] = "kmsg";
2070 break;
2071
2072 case LOG_TARGET_NULL:
2073 command_line[pos++] = "null";
2074 break;
2075
2076 case LOG_TARGET_CONSOLE:
2077 default:
2078 command_line[pos++] = "console";
2079 break;
2080 };
2081
2082 if (log_get_show_color())
2083 command_line[pos++] = "--log-color";
2084
2085 if (log_get_show_location())
2086 command_line[pos++] = "--log-location";
2087
2088 if (streq(shutdown_verb, "exit")) {
2089 command_line[pos++] = "--exit-code";
2090 command_line[pos++] = exit_code;
2091 xsprintf(exit_code, "%d", retval);
2092 }
2093
2094 assert(pos < ELEMENTSOF(command_line));
2095
2096 if (arm_reboot_watchdog && arg_shutdown_watchdog > 0) {
2097 char *e;
2098
2099 /* If we reboot let's set the shutdown
2100 * watchdog and tell the shutdown binary to
2101 * repeatedly ping it */
2102 r = watchdog_set_timeout(&arg_shutdown_watchdog);
2103 watchdog_close(r < 0);
2104
2105 /* Tell the binary how often to ping, ignore failure */
2106 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
2107 (void) strv_push(&env_block, e);
2108 } else
2109 watchdog_close(true);
2110
2111 /* Avoid the creation of new processes forked by the
2112 * kernel; at this point, we will not listen to the
2113 * signals anyway */
2114 if (detect_container() <= 0)
2115 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
2116
2117 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
2118 log_error_errno(errno, "Failed to execute shutdown binary, %s: %m",
2119 getpid() == 1 ? "freezing" : "quitting");
2120 }
2121
2122 if (getpid() == 1) {
2123 if (error_message)
2124 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2125 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
2126 "%s, freezing.", error_message);
2127 freeze_or_reboot();
2128 }
2129
2130 return retval;
2131 }