]> git.ipfire.org Git - thirdparty/systemd.git/blob - src/core/main.c
projects / thirdparty / systemd.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge pull request #9005 from fsateler/circular-deps
[thirdparty/systemd.git] / src / core / main.c
1 /* SPDX-License-Identifier: LGPL-2.1+ */
2 /***
3 This file is part of systemd.
4
5 Copyright 2010 Lennart Poettering
6 ***/
7
8 #include <errno.h>
9 #include <fcntl.h>
10 #include <getopt.h>
11 #include <signal.h>
12 #include <stdio.h>
13 #include <string.h>
14 #include <sys/mount.h>
15 #include <sys/prctl.h>
16 #include <sys/reboot.h>
17 #include <sys/stat.h>
18 #include <unistd.h>
19 #if HAVE_SECCOMP
20 #include <seccomp.h>
21 #endif
22 #if HAVE_VALGRIND_VALGRIND_H
23 #include <valgrind/valgrind.h>
24 #endif
25
26 #include "sd-bus.h"
27 #include "sd-daemon.h"
28 #include "sd-messages.h"
29
30 #include "alloc-util.h"
31 #include "architecture.h"
32 #include "build.h"
33 #include "bus-error.h"
34 #include "bus-util.h"
35 #include "capability-util.h"
36 #include "clock-util.h"
37 #include "conf-parser.h"
38 #include "cpu-set-util.h"
39 #include "dbus-manager.h"
40 #include "def.h"
41 #include "emergency-action.h"
42 #include "env-util.h"
43 #include "exit-status.h"
44 #include "fd-util.h"
45 #include "fdset.h"
46 #include "fileio.h"
47 #include "format-util.h"
48 #include "fs-util.h"
49 #include "hostname-setup.h"
50 #include "ima-setup.h"
51 #include "killall.h"
52 #include "kmod-setup.h"
53 #include "load-fragment.h"
54 #include "log.h"
55 #include "loopback-setup.h"
56 #include "machine-id-setup.h"
57 #include "manager.h"
58 #include "missing.h"
59 #include "mount-setup.h"
60 #include "pager.h"
61 #include "parse-util.h"
62 #include "path-util.h"
63 #include "proc-cmdline.h"
64 #include "process-util.h"
65 #include "raw-clone.h"
66 #include "rlimit-util.h"
67 #if HAVE_SECCOMP
68 #include "seccomp-util.h"
69 #endif
70 #include "selinux-setup.h"
71 #include "selinux-util.h"
72 #include "signal-util.h"
73 #include "smack-setup.h"
74 #include "special.h"
75 #include "stat-util.h"
76 #include "stdio-util.h"
77 #include "strv.h"
78 #include "switch-root.h"
79 #include "terminal-util.h"
80 #include "umask-util.h"
81 #include "user-util.h"
82 #include "util.h"
83 #include "virt.h"
84 #include "watchdog.h"
85
86 static enum {
87 ACTION_RUN,
88 ACTION_HELP,
89 ACTION_VERSION,
90 ACTION_TEST,
91 ACTION_DUMP_CONFIGURATION_ITEMS
92 } arg_action = ACTION_RUN;
93 static char *arg_default_unit = NULL;
94 static bool arg_system = false;
95 static bool arg_dump_core = true;
96 static int arg_crash_chvt = -1;
97 static bool arg_crash_shell = false;
98 static bool arg_crash_reboot = false;
99 static char *arg_confirm_spawn = NULL;
100 static ShowStatus arg_show_status = _SHOW_STATUS_UNSET;
101 static bool arg_switched_root = false;
102 static bool arg_no_pager = false;
103 static bool arg_service_watchdogs = true;
104 static char ***arg_join_controllers = NULL;
105 static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
106 static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
107 static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
108 static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
109 static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
110 static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
111 static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
112 static usec_t arg_runtime_watchdog = 0;
113 static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
114 static char *arg_watchdog_device = NULL;
115 static char **arg_default_environment = NULL;
116 static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {};
117 static uint64_t arg_capability_bounding_set = CAP_ALL;
118 static bool arg_no_new_privs = false;
119 static nsec_t arg_timer_slack_nsec = NSEC_INFINITY;
120 static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
121 static Set* arg_syscall_archs = NULL;
122 static FILE* arg_serialization = NULL;
123 static bool arg_default_cpu_accounting = false;
124 static bool arg_default_io_accounting = false;
125 static bool arg_default_ip_accounting = false;
126 static bool arg_default_blockio_accounting = false;
127 static bool arg_default_memory_accounting = MEMORY_ACCOUNTING_DEFAULT;
128 static bool arg_default_tasks_accounting = true;
129 static uint64_t arg_default_tasks_max = UINT64_MAX;
130 static sd_id128_t arg_machine_id = {};
131 static EmergencyAction arg_cad_burst_action = EMERGENCY_ACTION_REBOOT_FORCE;
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_cached() != 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_MINIMAL)
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 parse_confirm_spawn(const char *value, char **console) {
294 char *s;
295 int r;
296
297 r = value ? parse_boolean(value) : 1;
298 if (r == 0) {
299 *console = NULL;
300 return 0;
301 }
302
303 if (r > 0) /* on with default tty */
304 s = strdup("/dev/console");
305 else if (is_path(value)) /* on with fully qualified path */
306 s = strdup(value);
307 else /* on with only a tty file name, not a fully qualified path */
308 s = strjoin("/dev/", value);
309 if (!s)
310 return -ENOMEM;
311 *console = s;
312 return 0;
313 }
314
315 static int set_machine_id(const char *m) {
316 sd_id128_t t;
317 assert(m);
318
319 if (sd_id128_from_string(m, &t) < 0)
320 return -EINVAL;
321
322 if (sd_id128_is_null(t))
323 return -EINVAL;
324
325 arg_machine_id = t;
326 return 0;
327 }
328
329 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
330
331 int r;
332
333 assert(key);
334
335 if (STR_IN_SET(key, "systemd.unit", "rd.systemd.unit")) {
336
337 if (proc_cmdline_value_missing(key, value))
338 return 0;
339
340 if (!unit_name_is_valid(value, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE))
341 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key, value);
342 else if (in_initrd() == !!startswith(key, "rd.")) {
343 if (free_and_strdup(&arg_default_unit, value) < 0)
344 return log_oom();
345 }
346
347 } else if (proc_cmdline_key_streq(key, "systemd.dump_core")) {
348
349 r = value ? parse_boolean(value) : true;
350 if (r < 0)
351 log_warning("Failed to parse dump core switch %s. Ignoring.", value);
352 else
353 arg_dump_core = r;
354
355 } else if (proc_cmdline_key_streq(key, "systemd.crash_chvt")) {
356
357 if (!value)
358 arg_crash_chvt = 0; /* turn on */
359 else if (parse_crash_chvt(value) < 0)
360 log_warning("Failed to parse crash chvt switch %s. Ignoring.", value);
361
362 } else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) {
363
364 r = value ? parse_boolean(value) : true;
365 if (r < 0)
366 log_warning("Failed to parse crash shell switch %s. Ignoring.", value);
367 else
368 arg_crash_shell = r;
369
370 } else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) {
371
372 r = value ? parse_boolean(value) : true;
373 if (r < 0)
374 log_warning("Failed to parse crash reboot switch %s. Ignoring.", value);
375 else
376 arg_crash_reboot = r;
377
378 } else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) {
379 char *s;
380
381 r = parse_confirm_spawn(value, &s);
382 if (r < 0)
383 log_warning_errno(r, "Failed to parse confirm_spawn switch %s. Ignoring.", value);
384 else {
385 free(arg_confirm_spawn);
386 arg_confirm_spawn = s;
387 }
388
389 } else if (proc_cmdline_key_streq(key, "systemd.service_watchdogs")) {
390
391 r = value ? parse_boolean(value) : true;
392 if (r < 0)
393 log_warning("Failed to parse service watchdog switch %s. Ignoring.", value);
394 else
395 arg_service_watchdogs = r;
396
397 } else if (proc_cmdline_key_streq(key, "systemd.show_status")) {
398
399 if (value) {
400 r = parse_show_status(value, &arg_show_status);
401 if (r < 0)
402 log_warning("Failed to parse show status switch %s. Ignoring.", value);
403 } else
404 arg_show_status = SHOW_STATUS_YES;
405
406 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) {
407
408 if (proc_cmdline_value_missing(key, value))
409 return 0;
410
411 r = exec_output_from_string(value);
412 if (r < 0)
413 log_warning("Failed to parse default standard output switch %s. Ignoring.", value);
414 else
415 arg_default_std_output = r;
416
417 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) {
418
419 if (proc_cmdline_value_missing(key, value))
420 return 0;
421
422 r = exec_output_from_string(value);
423 if (r < 0)
424 log_warning("Failed to parse default standard error switch %s. Ignoring.", value);
425 else
426 arg_default_std_error = r;
427
428 } else if (streq(key, "systemd.setenv")) {
429
430 if (proc_cmdline_value_missing(key, value))
431 return 0;
432
433 if (env_assignment_is_valid(value)) {
434 char **env;
435
436 env = strv_env_set(arg_default_environment, value);
437 if (!env)
438 return log_oom();
439
440 arg_default_environment = env;
441 } else
442 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
443
444 } else if (proc_cmdline_key_streq(key, "systemd.machine_id")) {
445
446 if (proc_cmdline_value_missing(key, value))
447 return 0;
448
449 r = set_machine_id(value);
450 if (r < 0)
451 log_warning("MachineID '%s' is not valid. Ignoring.", value);
452
453 } else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) {
454
455 if (proc_cmdline_value_missing(key, value))
456 return 0;
457
458 r = parse_sec(value, &arg_default_timeout_start_usec);
459 if (r < 0)
460 log_warning_errno(r, "Failed to parse default start timeout: %s, ignoring.", value);
461
462 if (arg_default_timeout_start_usec <= 0)
463 arg_default_timeout_start_usec = USEC_INFINITY;
464
465 } else if (proc_cmdline_key_streq(key, "systemd.watchdog_device")) {
466
467 if (proc_cmdline_value_missing(key, value))
468 return 0;
469
470 parse_path_argument_and_warn(value, false, &arg_watchdog_device);
471
472 } else if (streq(key, "quiet") && !value) {
473
474 if (arg_show_status == _SHOW_STATUS_UNSET)
475 arg_show_status = SHOW_STATUS_AUTO;
476
477 } else if (streq(key, "debug") && !value) {
478
479 /* Note that log_parse_environment() handles 'debug'
480 * too, and sets the log level to LOG_DEBUG. */
481
482 if (detect_container() > 0)
483 log_set_target(LOG_TARGET_CONSOLE);
484
485 } else if (!value) {
486 const char *target;
487
488 /* SysV compatibility */
489 target = runlevel_to_target(key);
490 if (target)
491 return free_and_strdup(&arg_default_unit, target);
492 }
493
494 return 0;
495 }
496
497 #define DEFINE_SETTER(name, func, descr) \
498 static int name(const char *unit, \
499 const char *filename, \
500 unsigned line, \
501 const char *section, \
502 unsigned section_line, \
503 const char *lvalue, \
504 int ltype, \
505 const char *rvalue, \
506 void *data, \
507 void *userdata) { \
508 \
509 int r; \
510 \
511 assert(filename); \
512 assert(lvalue); \
513 assert(rvalue); \
514 \
515 r = func(rvalue); \
516 if (r < 0) \
517 log_syntax(unit, LOG_ERR, filename, line, r, \
518 "Invalid " descr "'%s': %m", \
519 rvalue); \
520 \
521 return 0; \
522 }
523
524 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level")
525 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target")
526 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" )
527 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location")
528
529 static int config_parse_cpu_affinity2(
530 const char *unit,
531 const char *filename,
532 unsigned line,
533 const char *section,
534 unsigned section_line,
535 const char *lvalue,
536 int ltype,
537 const char *rvalue,
538 void *data,
539 void *userdata) {
540
541 _cleanup_cpu_free_ cpu_set_t *c = NULL;
542 int ncpus;
543
544 ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue);
545 if (ncpus < 0)
546 return ncpus;
547
548 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
549 log_warning_errno(errno, "Failed to set CPU affinity: %m");
550
551 return 0;
552 }
553
554 static int config_parse_show_status(
555 const char* unit,
556 const char *filename,
557 unsigned line,
558 const char *section,
559 unsigned section_line,
560 const char *lvalue,
561 int ltype,
562 const char *rvalue,
563 void *data,
564 void *userdata) {
565
566 int k;
567 ShowStatus *b = data;
568
569 assert(filename);
570 assert(lvalue);
571 assert(rvalue);
572 assert(data);
573
574 k = parse_show_status(rvalue, b);
575 if (k < 0) {
576 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
577 return 0;
578 }
579
580 return 0;
581 }
582
583 static int config_parse_output_restricted(
584 const char* unit,
585 const char *filename,
586 unsigned line,
587 const char *section,
588 unsigned section_line,
589 const char *lvalue,
590 int ltype,
591 const char *rvalue,
592 void *data,
593 void *userdata) {
594
595 ExecOutput t, *eo = data;
596
597 assert(filename);
598 assert(lvalue);
599 assert(rvalue);
600 assert(data);
601
602 t = exec_output_from_string(rvalue);
603 if (t < 0) {
604 log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse output type, ignoring: %s", rvalue);
605 return 0;
606 }
607
608 if (IN_SET(t, EXEC_OUTPUT_SOCKET, EXEC_OUTPUT_NAMED_FD, EXEC_OUTPUT_FILE)) {
609 log_syntax(unit, LOG_ERR, filename, line, 0, "Standard output types socket, fd:, file: are not supported as defaults, ignoring: %s", rvalue);
610 return 0;
611 }
612
613 *eo = t;
614 return 0;
615 }
616
617 static int config_parse_crash_chvt(
618 const char* unit,
619 const char *filename,
620 unsigned line,
621 const char *section,
622 unsigned section_line,
623 const char *lvalue,
624 int ltype,
625 const char *rvalue,
626 void *data,
627 void *userdata) {
628
629 int r;
630
631 assert(filename);
632 assert(lvalue);
633 assert(rvalue);
634
635 r = parse_crash_chvt(rvalue);
636 if (r < 0) {
637 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
638 return 0;
639 }
640
641 return 0;
642 }
643
644 static int parse_config_file(void) {
645
646 const ConfigTableItem items[] = {
647 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
648 { "Manager", "LogTarget", config_parse_target, 0, NULL },
649 { "Manager", "LogColor", config_parse_color, 0, NULL },
650 { "Manager", "LogLocation", config_parse_location, 0, NULL },
651 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
652 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL },
653 { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL },
654 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
655 { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
656 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
657 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
658 { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
659 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
660 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
661 { "Manager", "WatchdogDevice", config_parse_path, 0, &arg_watchdog_device },
662 { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
663 { "Manager", "NoNewPrivileges", config_parse_bool, 0, &arg_no_new_privs },
664 #if HAVE_SECCOMP
665 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
666 #endif
667 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
668 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
669 { "Manager", "DefaultStandardOutput", config_parse_output_restricted,0, &arg_default_std_output },
670 { "Manager", "DefaultStandardError", config_parse_output_restricted,0, &arg_default_std_error },
671 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
672 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
673 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
674 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */
675 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec, 0, &arg_default_start_limit_interval },
676 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
677 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
678 { "Manager", "DefaultLimitCPU", config_parse_limit, RLIMIT_CPU, arg_default_rlimit },
679 { "Manager", "DefaultLimitFSIZE", config_parse_limit, RLIMIT_FSIZE, arg_default_rlimit },
680 { "Manager", "DefaultLimitDATA", config_parse_limit, RLIMIT_DATA, arg_default_rlimit },
681 { "Manager", "DefaultLimitSTACK", config_parse_limit, RLIMIT_STACK, arg_default_rlimit },
682 { "Manager", "DefaultLimitCORE", config_parse_limit, RLIMIT_CORE, arg_default_rlimit },
683 { "Manager", "DefaultLimitRSS", config_parse_limit, RLIMIT_RSS, arg_default_rlimit },
684 { "Manager", "DefaultLimitNOFILE", config_parse_limit, RLIMIT_NOFILE, arg_default_rlimit },
685 { "Manager", "DefaultLimitAS", config_parse_limit, RLIMIT_AS, arg_default_rlimit },
686 { "Manager", "DefaultLimitNPROC", config_parse_limit, RLIMIT_NPROC, arg_default_rlimit },
687 { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, RLIMIT_MEMLOCK, arg_default_rlimit },
688 { "Manager", "DefaultLimitLOCKS", config_parse_limit, RLIMIT_LOCKS, arg_default_rlimit },
689 { "Manager", "DefaultLimitSIGPENDING", config_parse_limit, RLIMIT_SIGPENDING, arg_default_rlimit },
690 { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, RLIMIT_MSGQUEUE, arg_default_rlimit },
691 { "Manager", "DefaultLimitNICE", config_parse_limit, RLIMIT_NICE, arg_default_rlimit },
692 { "Manager", "DefaultLimitRTPRIO", config_parse_limit, RLIMIT_RTPRIO, arg_default_rlimit },
693 { "Manager", "DefaultLimitRTTIME", config_parse_limit, RLIMIT_RTTIME, arg_default_rlimit },
694 { "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_default_cpu_accounting },
695 { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting },
696 { "Manager", "DefaultIPAccounting", config_parse_bool, 0, &arg_default_ip_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 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action, 0, &arg_cad_burst_action },
702 {}
703 };
704
705 const char *fn, *conf_dirs_nulstr;
706
707 fn = arg_system ?
708 PKGSYSCONFDIR "/system.conf" :
709 PKGSYSCONFDIR "/user.conf";
710
711 conf_dirs_nulstr = arg_system ?
712 CONF_PATHS_NULSTR("systemd/system.conf.d") :
713 CONF_PATHS_NULSTR("systemd/user.conf.d");
714
715 (void) config_parse_many_nulstr(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, CONFIG_PARSE_WARN, NULL);
716
717 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
718 * like everywhere else. */
719 if (arg_default_timeout_start_usec <= 0)
720 arg_default_timeout_start_usec = USEC_INFINITY;
721 if (arg_default_timeout_stop_usec <= 0)
722 arg_default_timeout_stop_usec = USEC_INFINITY;
723
724 return 0;
725 }
726
727 static void set_manager_defaults(Manager *m) {
728
729 assert(m);
730
731 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
732 m->default_std_output = arg_default_std_output;
733 m->default_std_error = arg_default_std_error;
734 m->default_timeout_start_usec = arg_default_timeout_start_usec;
735 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
736 m->default_restart_usec = arg_default_restart_usec;
737 m->default_start_limit_interval = arg_default_start_limit_interval;
738 m->default_start_limit_burst = arg_default_start_limit_burst;
739 m->default_cpu_accounting = arg_default_cpu_accounting;
740 m->default_io_accounting = arg_default_io_accounting;
741 m->default_ip_accounting = arg_default_ip_accounting;
742 m->default_blockio_accounting = arg_default_blockio_accounting;
743 m->default_memory_accounting = arg_default_memory_accounting;
744 m->default_tasks_accounting = arg_default_tasks_accounting;
745 m->default_tasks_max = arg_default_tasks_max;
746
747 manager_set_default_rlimits(m, arg_default_rlimit);
748 manager_environment_add(m, NULL, arg_default_environment);
749 }
750
751 static void set_manager_settings(Manager *m) {
752
753 assert(m);
754
755 m->confirm_spawn = arg_confirm_spawn;
756 m->service_watchdogs = arg_service_watchdogs;
757 m->runtime_watchdog = arg_runtime_watchdog;
758 m->shutdown_watchdog = arg_shutdown_watchdog;
759 m->cad_burst_action = arg_cad_burst_action;
760
761 manager_set_show_status(m, arg_show_status);
762 }
763
764 static int parse_argv(int argc, char *argv[]) {
765 enum {
766 ARG_LOG_LEVEL = 0x100,
767 ARG_LOG_TARGET,
768 ARG_LOG_COLOR,
769 ARG_LOG_LOCATION,
770 ARG_UNIT,
771 ARG_SYSTEM,
772 ARG_USER,
773 ARG_TEST,
774 ARG_NO_PAGER,
775 ARG_VERSION,
776 ARG_DUMP_CONFIGURATION_ITEMS,
777 ARG_DUMP_CORE,
778 ARG_CRASH_CHVT,
779 ARG_CRASH_SHELL,
780 ARG_CRASH_REBOOT,
781 ARG_CONFIRM_SPAWN,
782 ARG_SHOW_STATUS,
783 ARG_DESERIALIZE,
784 ARG_SWITCHED_ROOT,
785 ARG_DEFAULT_STD_OUTPUT,
786 ARG_DEFAULT_STD_ERROR,
787 ARG_MACHINE_ID,
788 ARG_SERVICE_WATCHDOGS,
789 };
790
791 static const struct option options[] = {
792 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
793 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
794 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
795 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
796 { "unit", required_argument, NULL, ARG_UNIT },
797 { "system", no_argument, NULL, ARG_SYSTEM },
798 { "user", no_argument, NULL, ARG_USER },
799 { "test", no_argument, NULL, ARG_TEST },
800 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
801 { "help", no_argument, NULL, 'h' },
802 { "version", no_argument, NULL, ARG_VERSION },
803 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
804 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
805 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
806 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
807 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
808 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
809 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
810 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
811 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
812 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
813 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
814 { "machine-id", required_argument, NULL, ARG_MACHINE_ID },
815 { "service-watchdogs", required_argument, NULL, ARG_SERVICE_WATCHDOGS },
816 {}
817 };
818
819 int c, r;
820
821 assert(argc >= 1);
822 assert(argv);
823
824 if (getpid_cached() == 1)
825 opterr = 0;
826
827 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
828
829 switch (c) {
830
831 case ARG_LOG_LEVEL:
832 r = log_set_max_level_from_string(optarg);
833 if (r < 0) {
834 log_error("Failed to parse log level %s.", optarg);
835 return r;
836 }
837
838 break;
839
840 case ARG_LOG_TARGET:
841 r = log_set_target_from_string(optarg);
842 if (r < 0) {
843 log_error("Failed to parse log target %s.", optarg);
844 return r;
845 }
846
847 break;
848
849 case ARG_LOG_COLOR:
850
851 if (optarg) {
852 r = log_show_color_from_string(optarg);
853 if (r < 0) {
854 log_error("Failed to parse log color setting %s.", optarg);
855 return r;
856 }
857 } else
858 log_show_color(true);
859
860 break;
861
862 case ARG_LOG_LOCATION:
863 if (optarg) {
864 r = log_show_location_from_string(optarg);
865 if (r < 0) {
866 log_error("Failed to parse log location setting %s.", optarg);
867 return r;
868 }
869 } else
870 log_show_location(true);
871
872 break;
873
874 case ARG_DEFAULT_STD_OUTPUT:
875 r = exec_output_from_string(optarg);
876 if (r < 0) {
877 log_error("Failed to parse default standard output setting %s.", optarg);
878 return r;
879 } else
880 arg_default_std_output = r;
881 break;
882
883 case ARG_DEFAULT_STD_ERROR:
884 r = exec_output_from_string(optarg);
885 if (r < 0) {
886 log_error("Failed to parse default standard error output setting %s.", optarg);
887 return r;
888 } else
889 arg_default_std_error = r;
890 break;
891
892 case ARG_UNIT:
893 r = free_and_strdup(&arg_default_unit, optarg);
894 if (r < 0)
895 return log_error_errno(r, "Failed to set default unit %s: %m", optarg);
896
897 break;
898
899 case ARG_SYSTEM:
900 arg_system = true;
901 break;
902
903 case ARG_USER:
904 arg_system = false;
905 break;
906
907 case ARG_TEST:
908 arg_action = ACTION_TEST;
909 break;
910
911 case ARG_NO_PAGER:
912 arg_no_pager = true;
913 break;
914
915 case ARG_VERSION:
916 arg_action = ACTION_VERSION;
917 break;
918
919 case ARG_DUMP_CONFIGURATION_ITEMS:
920 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
921 break;
922
923 case ARG_DUMP_CORE:
924 if (!optarg)
925 arg_dump_core = true;
926 else {
927 r = parse_boolean(optarg);
928 if (r < 0)
929 return log_error_errno(r, "Failed to parse dump core boolean: %s", optarg);
930 arg_dump_core = r;
931 }
932 break;
933
934 case ARG_CRASH_CHVT:
935 r = parse_crash_chvt(optarg);
936 if (r < 0)
937 return log_error_errno(r, "Failed to parse crash virtual terminal index: %s", optarg);
938 break;
939
940 case ARG_CRASH_SHELL:
941 if (!optarg)
942 arg_crash_shell = true;
943 else {
944 r = parse_boolean(optarg);
945 if (r < 0)
946 return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg);
947 arg_crash_shell = r;
948 }
949 break;
950
951 case ARG_CRASH_REBOOT:
952 if (!optarg)
953 arg_crash_reboot = true;
954 else {
955 r = parse_boolean(optarg);
956 if (r < 0)
957 return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg);
958 arg_crash_reboot = r;
959 }
960 break;
961
962 case ARG_CONFIRM_SPAWN:
963 arg_confirm_spawn = mfree(arg_confirm_spawn);
964
965 r = parse_confirm_spawn(optarg, &arg_confirm_spawn);
966 if (r < 0)
967 return log_error_errno(r, "Failed to parse confirm spawn option: %m");
968 break;
969
970 case ARG_SERVICE_WATCHDOGS:
971 r = parse_boolean(optarg);
972 if (r < 0)
973 return log_error_errno(r, "Failed to parse service watchdogs boolean: %s", optarg);
974 arg_service_watchdogs = r;
975 break;
976
977 case ARG_SHOW_STATUS:
978 if (optarg) {
979 r = parse_show_status(optarg, &arg_show_status);
980 if (r < 0) {
981 log_error("Failed to parse show status boolean %s.", optarg);
982 return r;
983 }
984 } else
985 arg_show_status = SHOW_STATUS_YES;
986 break;
987
988 case ARG_DESERIALIZE: {
989 int fd;
990 FILE *f;
991
992 r = safe_atoi(optarg, &fd);
993 if (r < 0 || fd < 0) {
994 log_error("Failed to parse deserialize option %s.", optarg);
995 return -EINVAL;
996 }
997
998 (void) fd_cloexec(fd, true);
999
1000 f = fdopen(fd, "r");
1001 if (!f)
1002 return log_error_errno(errno, "Failed to open serialization fd: %m");
1003
1004 safe_fclose(arg_serialization);
1005 arg_serialization = f;
1006
1007 break;
1008 }
1009
1010 case ARG_SWITCHED_ROOT:
1011 arg_switched_root = true;
1012 break;
1013
1014 case ARG_MACHINE_ID:
1015 r = set_machine_id(optarg);
1016 if (r < 0)
1017 return log_error_errno(r, "MachineID '%s' is not valid.", optarg);
1018 break;
1019
1020 case 'h':
1021 arg_action = ACTION_HELP;
1022 break;
1023
1024 case 'D':
1025 log_set_max_level(LOG_DEBUG);
1026 break;
1027
1028 case 'b':
1029 case 's':
1030 case 'z':
1031 /* Just to eat away the sysvinit kernel
1032 * cmdline args without getopt() error
1033 * messages that we'll parse in
1034 * parse_proc_cmdline_word() or ignore. */
1035
1036 case '?':
1037 if (getpid_cached() != 1)
1038 return -EINVAL;
1039 else
1040 return 0;
1041
1042 default:
1043 assert_not_reached("Unhandled option code.");
1044 }
1045
1046 if (optind < argc && getpid_cached() != 1) {
1047 /* Hmm, when we aren't run as init system
1048 * let's complain about excess arguments */
1049
1050 log_error("Excess arguments.");
1051 return -EINVAL;
1052 }
1053
1054 return 0;
1055 }
1056
1057 static int help(void) {
1058
1059 printf("%s [OPTIONS...]\n\n"
1060 "Starts up and maintains the system or user services.\n\n"
1061 " -h --help Show this help\n"
1062 " --version Show version\n"
1063 " --test Determine startup sequence, dump it and exit\n"
1064 " --no-pager Do not pipe output into a pager\n"
1065 " --dump-configuration-items Dump understood unit configuration items\n"
1066 " --unit=UNIT Set default unit\n"
1067 " --system Run a system instance, even if PID != 1\n"
1068 " --user Run a user instance\n"
1069 " --dump-core[=BOOL] Dump core on crash\n"
1070 " --crash-vt=NR Change to specified VT on crash\n"
1071 " --crash-reboot[=BOOL] Reboot on crash\n"
1072 " --crash-shell[=BOOL] Run shell on crash\n"
1073 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1074 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1075 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1076 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1077 " --log-color[=BOOL] Highlight important log messages\n"
1078 " --log-location[=BOOL] Include code location in log messages\n"
1079 " --default-standard-output= Set default standard output for services\n"
1080 " --default-standard-error= Set default standard error output for services\n",
1081 program_invocation_short_name);
1082
1083 return 0;
1084 }
1085
1086 static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
1087 _cleanup_fdset_free_ FDSet *fds = NULL;
1088 _cleanup_fclose_ FILE *f = NULL;
1089 int r;
1090
1091 assert(m);
1092 assert(_f);
1093 assert(_fds);
1094
1095 r = manager_open_serialization(m, &f);
1096 if (r < 0)
1097 return log_error_errno(r, "Failed to create serialization file: %m");
1098
1099 /* Make sure nothing is really destructed when we shut down */
1100 m->n_reloading++;
1101 bus_manager_send_reloading(m, true);
1102
1103 fds = fdset_new();
1104 if (!fds)
1105 return log_oom();
1106
1107 r = manager_serialize(m, f, fds, switching_root);
1108 if (r < 0)
1109 return log_error_errno(r, "Failed to serialize state: %m");
1110
1111 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1112 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1113
1114 r = fd_cloexec(fileno(f), false);
1115 if (r < 0)
1116 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1117
1118 r = fdset_cloexec(fds, false);
1119 if (r < 0)
1120 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1121
1122 *_f = TAKE_PTR(f);
1123 *_fds = TAKE_PTR(fds);
1124
1125 return 0;
1126 }
1127
1128 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1129 struct rlimit nl;
1130 int r;
1131 int min_max;
1132 _cleanup_free_ char *nr_open = NULL;
1133
1134 assert(saved_rlimit);
1135
1136 /* Save the original RLIMIT_NOFILE so that we can reset it
1137 * later when transitioning from the initrd to the main
1138 * systemd or suchlike. */
1139 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1140 return log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1141
1142 /* Make sure forked processes get the default kernel setting */
1143 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1144 struct rlimit *rl;
1145
1146 rl = newdup(struct rlimit, saved_rlimit, 1);
1147 if (!rl)
1148 return log_oom();
1149
1150 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1151 }
1152
1153 /* Get current RLIMIT_NOFILE maximum compiled into the kernel. */
1154 r = read_one_line_file("/proc/sys/fs/nr_open", &nr_open);
1155 if (r >= 0)
1156 r = safe_atoi(nr_open, &min_max);
1157 /* If we fail, fallback to the hard-coded kernel limit of 1024 * 1024. */
1158 if (r < 0)
1159 min_max = 1024 * 1024;
1160
1161 /* Bump up the resource limit for ourselves substantially */
1162 nl.rlim_cur = nl.rlim_max = min_max;
1163 r = setrlimit_closest(RLIMIT_NOFILE, &nl);
1164 if (r < 0)
1165 return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1166
1167 return 0;
1168 }
1169
1170 static int bump_rlimit_memlock(struct rlimit *saved_rlimit) {
1171 int r;
1172
1173 assert(saved_rlimit);
1174 assert(getuid() == 0);
1175
1176 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even though we have CAP_IPC_LOCK which
1177 * should normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's
1178 * bump the value high enough for the root user. */
1179
1180 if (getrlimit(RLIMIT_MEMLOCK, saved_rlimit) < 0)
1181 return log_warning_errno(errno, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1182
1183 r = setrlimit_closest(RLIMIT_MEMLOCK, &RLIMIT_MAKE_CONST(1024ULL*1024ULL*16ULL));
1184 if (r < 0)
1185 return log_warning_errno(r, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1186
1187 return 0;
1188 }
1189
1190 static void test_usr(void) {
1191
1192 /* Check that /usr is not a separate fs */
1193
1194 if (dir_is_empty("/usr") <= 0)
1195 return;
1196
1197 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1198 "Some things will probably break (sometimes even silently) in mysterious ways. "
1199 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1200 }
1201
1202 static int enforce_syscall_archs(Set *archs) {
1203 #if HAVE_SECCOMP
1204 int r;
1205
1206 if (!is_seccomp_available())
1207 return 0;
1208
1209 r = seccomp_restrict_archs(arg_syscall_archs);
1210 if (r < 0)
1211 return log_error_errno(r, "Failed to enforce system call architecture restrication: %m");
1212 #endif
1213 return 0;
1214 }
1215
1216 static int status_welcome(void) {
1217 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1218 const char *fn;
1219 int r;
1220
1221 if (arg_show_status <= 0)
1222 return 0;
1223
1224 FOREACH_STRING(fn, "/etc/os-release", "/usr/lib/os-release") {
1225 r = parse_env_file(fn, NEWLINE,
1226 "PRETTY_NAME", &pretty_name,
1227 "ANSI_COLOR", &ansi_color,
1228 NULL);
1229
1230 if (r != -ENOENT)
1231 break;
1232 }
1233 if (r < 0 && r != -ENOENT)
1234 log_warning_errno(r, "Failed to read os-release file, ignoring: %m");
1235
1236 if (log_get_show_color())
1237 return status_printf(NULL, false, false,
1238 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1239 isempty(ansi_color) ? "1" : ansi_color,
1240 isempty(pretty_name) ? "Linux" : pretty_name);
1241 else
1242 return status_printf(NULL, false, false,
1243 "\nWelcome to %s!\n",
1244 isempty(pretty_name) ? "Linux" : pretty_name);
1245 }
1246
1247 static int write_container_id(void) {
1248 const char *c;
1249 int r;
1250
1251 c = getenv("container");
1252 if (isempty(c))
1253 return 0;
1254
1255 RUN_WITH_UMASK(0022)
1256 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1257 if (r < 0)
1258 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
1259
1260 return 1;
1261 }
1262
1263 static int bump_unix_max_dgram_qlen(void) {
1264 _cleanup_free_ char *qlen = NULL;
1265 unsigned long v;
1266 int r;
1267
1268 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel
1269 * default of 16 is simply too low. We set the value really
1270 * really early during boot, so that it is actually applied to
1271 * all our sockets, including the $NOTIFY_SOCKET one. */
1272
1273 r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1274 if (r < 0)
1275 return log_warning_errno(r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1276
1277 r = safe_atolu(qlen, &v);
1278 if (r < 0)
1279 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length, ignoring: %m");
1280
1281 if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1282 return 0;
1283
1284 qlen = mfree(qlen);
1285 if (asprintf(&qlen, "%lu\n", DEFAULT_UNIX_MAX_DGRAM_QLEN) < 0)
1286 return log_oom();
1287
1288 r = write_string_file("/proc/sys/net/unix/max_dgram_qlen", qlen, 0);
1289 if (r < 0)
1290 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1291 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1292
1293 return 1;
1294 }
1295
1296 static int fixup_environment(void) {
1297 _cleanup_free_ char *term = NULL;
1298 const char *t;
1299 int r;
1300
1301 /* Only fix up the environment when we are started as PID 1 */
1302 if (getpid_cached() != 1)
1303 return 0;
1304
1305 /* We expect the environment to be set correctly if run inside a container. */
1306 if (detect_container() > 0)
1307 return 0;
1308
1309 /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1310 * device used by the console. We try to make a better guess here since some consoles might not have support
1311 * for color mode for example.
1312 *
1313 * However if TERM was configured through the kernel command line then leave it alone. */
1314 r = proc_cmdline_get_key("TERM", 0, &term);
1315 if (r < 0)
1316 return r;
1317
1318 t = term ?: default_term_for_tty("/dev/console");
1319
1320 if (setenv("TERM", t, 1) < 0)
1321 return -errno;
1322
1323 return 0;
1324 }
1325
1326 static void redirect_telinit(int argc, char *argv[]) {
1327
1328 /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1329
1330 #if HAVE_SYSV_COMPAT
1331 if (getpid_cached() == 1)
1332 return;
1333
1334 if (!strstr(program_invocation_short_name, "init"))
1335 return;
1336
1337 execv(SYSTEMCTL_BINARY_PATH, argv);
1338 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1339 exit(EXIT_FAILURE);
1340 #endif
1341 }
1342
1343 static int become_shutdown(
1344 const char *shutdown_verb,
1345 int retval) {
1346
1347 char log_level[DECIMAL_STR_MAX(int) + 1],
1348 exit_code[DECIMAL_STR_MAX(uint8_t) + 1],
1349 timeout[DECIMAL_STR_MAX(usec_t) + 1];
1350
1351 const char* command_line[13] = {
1352 SYSTEMD_SHUTDOWN_BINARY_PATH,
1353 shutdown_verb,
1354 "--timeout", timeout,
1355 "--log-level", log_level,
1356 "--log-target",
1357 };
1358
1359 _cleanup_strv_free_ char **env_block = NULL;
1360 size_t pos = 7;
1361 int r;
1362
1363 assert(shutdown_verb);
1364 assert(!command_line[pos]);
1365 env_block = strv_copy(environ);
1366
1367 xsprintf(log_level, "%d", log_get_max_level());
1368 xsprintf(timeout, "%" PRI_USEC "us", arg_default_timeout_stop_usec);
1369
1370 switch (log_get_target()) {
1371
1372 case LOG_TARGET_KMSG:
1373 case LOG_TARGET_JOURNAL_OR_KMSG:
1374 case LOG_TARGET_SYSLOG_OR_KMSG:
1375 command_line[pos++] = "kmsg";
1376 break;
1377
1378 case LOG_TARGET_NULL:
1379 command_line[pos++] = "null";
1380 break;
1381
1382 case LOG_TARGET_CONSOLE:
1383 default:
1384 command_line[pos++] = "console";
1385 break;
1386 };
1387
1388 if (log_get_show_color())
1389 command_line[pos++] = "--log-color";
1390
1391 if (log_get_show_location())
1392 command_line[pos++] = "--log-location";
1393
1394 if (streq(shutdown_verb, "exit")) {
1395 command_line[pos++] = "--exit-code";
1396 command_line[pos++] = exit_code;
1397 xsprintf(exit_code, "%d", retval);
1398 }
1399
1400 assert(pos < ELEMENTSOF(command_line));
1401
1402 if (streq(shutdown_verb, "reboot") &&
1403 arg_shutdown_watchdog > 0 &&
1404 arg_shutdown_watchdog != USEC_INFINITY) {
1405
1406 char *e;
1407
1408 /* If we reboot let's set the shutdown
1409 * watchdog and tell the shutdown binary to
1410 * repeatedly ping it */
1411 r = watchdog_set_timeout(&arg_shutdown_watchdog);
1412 watchdog_close(r < 0);
1413
1414 /* Tell the binary how often to ping, ignore failure */
1415 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
1416 (void) strv_consume(&env_block, e);
1417
1418 if (arg_watchdog_device &&
1419 asprintf(&e, "WATCHDOG_DEVICE=%s", arg_watchdog_device) > 0)
1420 (void) strv_consume(&env_block, e);
1421 } else
1422 watchdog_close(true);
1423
1424 /* Avoid the creation of new processes forked by the
1425 * kernel; at this point, we will not listen to the
1426 * signals anyway */
1427 if (detect_container() <= 0)
1428 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
1429
1430 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
1431 return -errno;
1432 }
1433
1434 static void initialize_clock(void) {
1435 int r;
1436
1437 if (clock_is_localtime(NULL) > 0) {
1438 int min;
1439
1440 /*
1441 * The very first call of settimeofday() also does a time warp in the kernel.
1442 *
1443 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1444 * of maintaining the RTC and do all adjustments. This matches the behavior of Windows, which leaves
1445 * the RTC alone if the registry tells that the RTC runs in UTC.
1446 */
1447 r = clock_set_timezone(&min);
1448 if (r < 0)
1449 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1450 else
1451 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1452
1453 } else if (!in_initrd()) {
1454 /*
1455 * Do a dummy very first call to seal the kernel's time warp magic.
1456 *
1457 * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1458 * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1459 * until we reach the real system.
1460 *
1461 * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1462 * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1463 * be treated as UTC that way.
1464 */
1465 (void) clock_reset_timewarp();
1466 }
1467
1468 r = clock_apply_epoch();
1469 if (r < 0)
1470 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1471 else if (r > 0)
1472 log_info("System time before build time, advancing clock.");
1473 }
1474
1475 static void initialize_coredump(bool skip_setup) {
1476
1477 if (getpid_cached() != 1)
1478 return;
1479
1480 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1481 * will process core dumps for system services by default. */
1482 if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0)
1483 log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m");
1484
1485 /* But at the same time, turn off the core_pattern logic by default, so that no coredumps are stored
1486 * until the systemd-coredump tool is enabled via sysctl. */
1487 if (!skip_setup)
1488 disable_coredumps();
1489 }
1490
1491 static void do_reexecute(
1492 int argc,
1493 char *argv[],
1494 const struct rlimit *saved_rlimit_nofile,
1495 const struct rlimit *saved_rlimit_memlock,
1496 FDSet *fds,
1497 const char *switch_root_dir,
1498 const char *switch_root_init,
1499 const char **ret_error_message) {
1500
1501 unsigned i, j, args_size;
1502 const char **args;
1503 int r;
1504
1505 assert(saved_rlimit_nofile);
1506 assert(saved_rlimit_memlock);
1507 assert(ret_error_message);
1508
1509 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1510 * we do that */
1511 watchdog_close(true);
1512
1513 /* Reset the RLIMIT_NOFILE to the kernel default, so that the new systemd can pass the kernel default to its
1514 * child processes */
1515
1516 if (saved_rlimit_nofile->rlim_cur > 0)
1517 (void) setrlimit(RLIMIT_NOFILE, saved_rlimit_nofile);
1518 if (saved_rlimit_memlock->rlim_cur != (rlim_t) -1)
1519 (void) setrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock);
1520
1521 if (switch_root_dir) {
1522 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1523 * SIGCHLD for them after deserializing. */
1524 broadcast_signal(SIGTERM, false, true, arg_default_timeout_stop_usec);
1525
1526 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1527 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1528 if (r < 0)
1529 log_error_errno(r, "Failed to switch root, trying to continue: %m");
1530 }
1531
1532 args_size = MAX(6, argc+1);
1533 args = newa(const char*, args_size);
1534
1535 if (!switch_root_init) {
1536 char sfd[DECIMAL_STR_MAX(int) + 1];
1537
1538 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1539 * the user didn't specify an explicit init to spawn. */
1540
1541 assert(arg_serialization);
1542 assert(fds);
1543
1544 xsprintf(sfd, "%i", fileno(arg_serialization));
1545
1546 i = 0;
1547 args[i++] = SYSTEMD_BINARY_PATH;
1548 if (switch_root_dir)
1549 args[i++] = "--switched-root";
1550 args[i++] = arg_system ? "--system" : "--user";
1551 args[i++] = "--deserialize";
1552 args[i++] = sfd;
1553 args[i++] = NULL;
1554
1555 assert(i <= args_size);
1556
1557 /*
1558 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do this is on
1559 * its own on exec(), but it will do it on exit(). Hence, to ensure we get a summary here, fork() off
1560 * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1561 * before proceeding into the exec().
1562 */
1563 valgrind_summary_hack();
1564
1565 (void) execv(args[0], (char* const*) args);
1566 log_debug_errno(errno, "Failed to execute our own binary, trying fallback: %m");
1567 }
1568
1569 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1570 * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1571 * doesn't matter.) */
1572
1573 arg_serialization = safe_fclose(arg_serialization);
1574 fds = fdset_free(fds);
1575
1576 /* Reopen the console */
1577 (void) make_console_stdio();
1578
1579 for (j = 1, i = 1; j < (unsigned) argc; j++)
1580 args[i++] = argv[j];
1581 args[i++] = NULL;
1582 assert(i <= args_size);
1583
1584 /* Reenable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1585 (void) reset_all_signal_handlers();
1586 (void) reset_signal_mask();
1587
1588 if (switch_root_init) {
1589 args[0] = switch_root_init;
1590 (void) execv(args[0], (char* const*) args);
1591 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
1592 }
1593
1594 args[0] = "/sbin/init";
1595 (void) execv(args[0], (char* const*) args);
1596 r = -errno;
1597
1598 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
1599 ANSI_HIGHLIGHT_RED " !! " ANSI_NORMAL,
1600 "Failed to execute /sbin/init");
1601
1602 if (r == -ENOENT) {
1603 log_warning("No /sbin/init, trying fallback");
1604
1605 args[0] = "/bin/sh";
1606 args[1] = NULL;
1607 (void) execv(args[0], (char* const*) args);
1608 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
1609 } else
1610 log_warning_errno(r, "Failed to execute /sbin/init, giving up: %m");
1611
1612 *ret_error_message = "Failed to execute fallback shell";
1613 }
1614
1615 static int invoke_main_loop(
1616 Manager *m,
1617 bool *ret_reexecute,
1618 int *ret_retval, /* Return parameters relevant for shutting down */
1619 const char **ret_shutdown_verb, /* … */
1620 FDSet **ret_fds, /* Return parameters for reexecuting */
1621 char **ret_switch_root_dir, /* … */
1622 char **ret_switch_root_init, /* … */
1623 const char **ret_error_message) {
1624
1625 int r;
1626
1627 assert(m);
1628 assert(ret_reexecute);
1629 assert(ret_retval);
1630 assert(ret_shutdown_verb);
1631 assert(ret_fds);
1632 assert(ret_switch_root_dir);
1633 assert(ret_switch_root_init);
1634 assert(ret_error_message);
1635
1636 for (;;) {
1637 r = manager_loop(m);
1638 if (r < 0) {
1639 *ret_error_message = "Failed to run main loop";
1640 return log_emergency_errno(r, "Failed to run main loop: %m");
1641 }
1642
1643 switch (m->exit_code) {
1644
1645 case MANAGER_RELOAD:
1646 log_info("Reloading.");
1647
1648 r = parse_config_file();
1649 if (r < 0)
1650 log_warning_errno(r, "Failed to parse config file, ignoring: %m");
1651
1652 set_manager_defaults(m);
1653
1654 r = manager_reload(m);
1655 if (r < 0)
1656 log_warning_errno(r, "Failed to reload, ignoring: %m");
1657
1658 break;
1659
1660 case MANAGER_REEXECUTE:
1661
1662 r = prepare_reexecute(m, &arg_serialization, ret_fds, false);
1663 if (r < 0) {
1664 *ret_error_message = "Failed to prepare for reexecution";
1665 return r;
1666 }
1667
1668 log_notice("Reexecuting.");
1669
1670 *ret_reexecute = true;
1671 *ret_retval = EXIT_SUCCESS;
1672 *ret_shutdown_verb = NULL;
1673 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1674
1675 return 0;
1676
1677 case MANAGER_SWITCH_ROOT:
1678 if (!m->switch_root_init) {
1679 r = prepare_reexecute(m, &arg_serialization, ret_fds, true);
1680 if (r < 0) {
1681 *ret_error_message = "Failed to prepare for reexecution";
1682 return r;
1683 }
1684 } else
1685 *ret_fds = NULL;
1686
1687 log_notice("Switching root.");
1688
1689 *ret_reexecute = true;
1690 *ret_retval = EXIT_SUCCESS;
1691 *ret_shutdown_verb = NULL;
1692
1693 /* Steal the switch root parameters */
1694 *ret_switch_root_dir = m->switch_root;
1695 *ret_switch_root_init = m->switch_root_init;
1696 m->switch_root = m->switch_root_init = NULL;
1697
1698 return 0;
1699
1700 case MANAGER_EXIT:
1701
1702 if (MANAGER_IS_USER(m)) {
1703 log_debug("Exit.");
1704
1705 *ret_reexecute = false;
1706 *ret_retval = m->return_value;
1707 *ret_shutdown_verb = NULL;
1708 *ret_fds = NULL;
1709 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1710
1711 return 0;
1712 }
1713
1714 _fallthrough_;
1715 case MANAGER_REBOOT:
1716 case MANAGER_POWEROFF:
1717 case MANAGER_HALT:
1718 case MANAGER_KEXEC: {
1719 static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
1720 [MANAGER_EXIT] = "exit",
1721 [MANAGER_REBOOT] = "reboot",
1722 [MANAGER_POWEROFF] = "poweroff",
1723 [MANAGER_HALT] = "halt",
1724 [MANAGER_KEXEC] = "kexec"
1725 };
1726
1727 log_notice("Shutting down.");
1728
1729 *ret_reexecute = false;
1730 *ret_retval = m->return_value;
1731 assert_se(*ret_shutdown_verb = table[m->exit_code]);
1732 *ret_fds = NULL;
1733 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1734
1735 return 0;
1736 }
1737
1738 default:
1739 assert_not_reached("Unknown exit code.");
1740 }
1741 }
1742 }
1743
1744 static void log_execution_mode(bool *ret_first_boot) {
1745 assert(ret_first_boot);
1746
1747 if (arg_system) {
1748 int v;
1749
1750 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1751 arg_action == ACTION_TEST ? "test " : "" );
1752
1753 v = detect_virtualization();
1754 if (v > 0)
1755 log_info("Detected virtualization %s.", virtualization_to_string(v));
1756
1757 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1758
1759 if (in_initrd()) {
1760 *ret_first_boot = false;
1761 log_info("Running in initial RAM disk.");
1762 } else {
1763 /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
1764 * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
1765 * doesn't it's unpopulated. This allows container managers and installers to provision a
1766 * couple of files already. If the container manager wants to provision the machine ID itself
1767 * it should pass $container_uuid to PID 1. */
1768
1769 *ret_first_boot = access("/etc/machine-id", F_OK) < 0;
1770 if (*ret_first_boot)
1771 log_info("Running with unpopulated /etc.");
1772 }
1773 } else {
1774 if (DEBUG_LOGGING) {
1775 _cleanup_free_ char *t;
1776
1777 t = uid_to_name(getuid());
1778 log_debug(PACKAGE_STRING " running in %suser mode for user " UID_FMT "/%s. (" SYSTEMD_FEATURES ")",
1779 arg_action == ACTION_TEST ? " test" : "", getuid(), strna(t));
1780 }
1781
1782 *ret_first_boot = false;
1783 }
1784 }
1785
1786 static int initialize_runtime(
1787 bool skip_setup,
1788 struct rlimit *saved_rlimit_nofile,
1789 struct rlimit *saved_rlimit_memlock,
1790 const char **ret_error_message) {
1791
1792 int r;
1793
1794 assert(ret_error_message);
1795
1796 /* Sets up various runtime parameters. Many of these initializations are conditionalized:
1797 *
1798 * - Some only apply to --system instances
1799 * - Some only apply to --user instances
1800 * - Some only apply when we first start up, but not when we reexecute
1801 */
1802
1803 if (arg_action != ACTION_RUN)
1804 return 0;
1805
1806 if (arg_system) {
1807 /* Make sure we leave a core dump without panicing the kernel. */
1808 install_crash_handler();
1809
1810 if (!skip_setup) {
1811 r = mount_cgroup_controllers(arg_join_controllers);
1812 if (r < 0) {
1813 *ret_error_message = "Failed to mount cgroup hierarchies";
1814 return r;
1815 }
1816
1817 status_welcome();
1818 hostname_setup();
1819 machine_id_setup(NULL, arg_machine_id, NULL);
1820 loopback_setup();
1821 bump_unix_max_dgram_qlen();
1822 test_usr();
1823 write_container_id();
1824 }
1825
1826 if (arg_watchdog_device) {
1827 r = watchdog_set_device(arg_watchdog_device);
1828 if (r < 0)
1829 log_warning_errno(r, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device);
1830 }
1831
1832 if (arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY)
1833 watchdog_set_timeout(&arg_runtime_watchdog);
1834 }
1835
1836 if (arg_timer_slack_nsec != NSEC_INFINITY)
1837 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1838 log_warning_errno(errno, "Failed to adjust timer slack, ignoring: %m");
1839
1840 if (arg_system && !cap_test_all(arg_capability_bounding_set)) {
1841 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
1842 if (r < 0) {
1843 *ret_error_message = "Failed to drop capability bounding set of usermode helpers";
1844 return log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
1845 }
1846
1847 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
1848 if (r < 0) {
1849 *ret_error_message = "Failed to drop capability bounding set";
1850 return log_emergency_errno(r, "Failed to drop capability bounding set: %m");
1851 }
1852 }
1853
1854 if (arg_system && arg_no_new_privs) {
1855 if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) {
1856 *ret_error_message = "Failed to disable new privileges";
1857 return log_emergency_errno(errno, "Failed to disable new privileges: %m");
1858 }
1859 }
1860
1861 if (arg_syscall_archs) {
1862 r = enforce_syscall_archs(arg_syscall_archs);
1863 if (r < 0) {
1864 *ret_error_message = "Failed to set syscall architectures";
1865 return r;
1866 }
1867 }
1868
1869 if (!arg_system)
1870 /* Become reaper of our children */
1871 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
1872 log_warning_errno(errno, "Failed to make us a subreaper: %m");
1873
1874 if (arg_system) {
1875 /* Bump up RLIMIT_NOFILE for systemd itself */
1876 (void) bump_rlimit_nofile(saved_rlimit_nofile);
1877 (void) bump_rlimit_memlock(saved_rlimit_memlock);
1878 }
1879
1880 return 0;
1881 }
1882
1883 static int do_queue_default_job(
1884 Manager *m,
1885 const char **ret_error_message) {
1886
1887 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
1888 Job *default_unit_job;
1889 Unit *target = NULL;
1890 int r;
1891
1892 log_debug("Activating default unit: %s", arg_default_unit);
1893
1894 r = manager_load_startable_unit_or_warn(m, arg_default_unit, NULL, &target);
1895 if (r < 0) {
1896 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET);
1897
1898 r = manager_load_startable_unit_or_warn(m, SPECIAL_RESCUE_TARGET, NULL, &target);
1899 if (r < 0) {
1900 *ret_error_message = r == -ERFKILL ? "Rescue target masked"
1901 : "Failed to load rescue target";
1902 return r;
1903 }
1904 }
1905
1906 assert(target->load_state == UNIT_LOADED);
1907
1908 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, &error, &default_unit_job);
1909 if (r == -EPERM) {
1910 log_debug_errno(r, "Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1911
1912 sd_bus_error_free(&error);
1913
1914 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, &error, &default_unit_job);
1915 if (r < 0) {
1916 *ret_error_message = "Failed to start default target";
1917 return log_emergency_errno(r, "Failed to start default target: %s", bus_error_message(&error, r));
1918 }
1919
1920 } else if (r < 0) {
1921 *ret_error_message = "Failed to isolate default target";
1922 return log_emergency_errno(r, "Failed to isolate default target: %s", bus_error_message(&error, r));
1923 }
1924
1925 m->default_unit_job_id = default_unit_job->id;
1926
1927 return 0;
1928 }
1929
1930 static void free_arguments(void) {
1931 size_t j;
1932
1933 /* Frees all arg_* variables, with the exception of arg_serialization */
1934
1935 for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++)
1936 arg_default_rlimit[j] = mfree(arg_default_rlimit[j]);
1937
1938 arg_default_unit = mfree(arg_default_unit);
1939 arg_confirm_spawn = mfree(arg_confirm_spawn);
1940 arg_join_controllers = strv_free_free(arg_join_controllers);
1941 arg_default_environment = strv_free(arg_default_environment);
1942 arg_syscall_archs = set_free(arg_syscall_archs);
1943 }
1944
1945 static int load_configuration(int argc, char **argv, const char **ret_error_message) {
1946 int r;
1947
1948 assert(ret_error_message);
1949
1950 arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U);
1951
1952 r = parse_config_file();
1953 if (r < 0) {
1954 *ret_error_message = "Failed to parse config file";
1955 return r;
1956 }
1957
1958 if (arg_system) {
1959 r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
1960 if (r < 0)
1961 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
1962 }
1963
1964 /* Note that this also parses bits from the kernel command line, including "debug". */
1965 log_parse_environment();
1966
1967 r = parse_argv(argc, argv);
1968 if (r < 0) {
1969 *ret_error_message = "Failed to parse commandline arguments";
1970 return r;
1971 }
1972
1973 /* Initialize default unit */
1974 if (!arg_default_unit) {
1975 arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET);
1976 if (!arg_default_unit) {
1977 *ret_error_message = "Failed to set default unit";
1978 return log_oom();
1979 }
1980 }
1981
1982 /* Initialize the show status setting if it hasn't been set explicitly yet */
1983 if (arg_show_status == _SHOW_STATUS_UNSET)
1984 arg_show_status = SHOW_STATUS_YES;
1985
1986 return 0;
1987 }
1988
1989 static int safety_checks(void) {
1990
1991 if (getpid_cached() == 1 &&
1992 arg_action != ACTION_RUN) {
1993 log_error("Unsupported execution mode while PID 1.");
1994 return -EPERM;
1995 }
1996
1997 if (getpid_cached() == 1 &&
1998 !arg_system) {
1999 log_error("Can't run --user mode as PID 1.");
2000 return -EPERM;
2001 }
2002
2003 if (arg_action == ACTION_RUN &&
2004 arg_system &&
2005 getpid_cached() != 1) {
2006 log_error("Can't run system mode unless PID 1.");
2007 return -EPERM;
2008 }
2009
2010 if (arg_action == ACTION_TEST &&
2011 geteuid() == 0) {
2012 log_error("Don't run test mode as root.");
2013 return -EPERM;
2014 }
2015
2016 if (!arg_system &&
2017 arg_action == ACTION_RUN &&
2018 sd_booted() <= 0) {
2019 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
2020 return -EOPNOTSUPP;
2021 }
2022
2023 if (!arg_system &&
2024 arg_action == ACTION_RUN &&
2025 !getenv("XDG_RUNTIME_DIR")) {
2026 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
2027 return -EUNATCH;
2028 }
2029
2030 if (arg_system &&
2031 arg_action == ACTION_RUN &&
2032 running_in_chroot() > 0) {
2033 log_error("Cannot be run in a chroot() environment.");
2034 return -EOPNOTSUPP;
2035 }
2036
2037 return 0;
2038 }
2039
2040 static int initialize_security(
2041 bool *loaded_policy,
2042 dual_timestamp *security_start_timestamp,
2043 dual_timestamp *security_finish_timestamp,
2044 const char **ret_error_message) {
2045
2046 int r;
2047
2048 assert(loaded_policy);
2049 assert(security_start_timestamp);
2050 assert(security_finish_timestamp);
2051 assert(ret_error_message);
2052
2053 dual_timestamp_get(security_start_timestamp);
2054
2055 r = mac_selinux_setup(loaded_policy);
2056 if (r < 0) {
2057 *ret_error_message = "Failed to load SELinux policy";
2058 return r;
2059 }
2060
2061 r = mac_smack_setup(loaded_policy);
2062 if (r < 0) {
2063 *ret_error_message = "Failed to load SMACK policy";
2064 return r;
2065 }
2066
2067 r = ima_setup();
2068 if (r < 0) {
2069 *ret_error_message = "Failed to load IMA policy";
2070 return r;
2071 }
2072
2073 dual_timestamp_get(security_finish_timestamp);
2074 return 0;
2075 }
2076
2077 static void test_summary(Manager *m) {
2078 assert(m);
2079
2080 printf("-> By units:\n");
2081 manager_dump_units(m, stdout, "\t");
2082
2083 printf("-> By jobs:\n");
2084 manager_dump_jobs(m, stdout, "\t");
2085 }
2086
2087 static int collect_fds(FDSet **ret_fds, const char **ret_error_message) {
2088 int r;
2089
2090 assert(ret_fds);
2091 assert(ret_error_message);
2092
2093 r = fdset_new_fill(ret_fds);
2094 if (r < 0) {
2095 *ret_error_message = "Failed to allocate fd set";
2096 return log_emergency_errno(r, "Failed to allocate fd set: %m");
2097 }
2098
2099 fdset_cloexec(*ret_fds, true);
2100
2101 if (arg_serialization)
2102 assert_se(fdset_remove(*ret_fds, fileno(arg_serialization)) >= 0);
2103
2104 return 0;
2105 }
2106
2107 static void setup_console_terminal(bool skip_setup) {
2108
2109 if (!arg_system)
2110 return;
2111
2112 /* Become a session leader if we aren't one yet. */
2113 (void) setsid();
2114
2115 /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2116 * tty. */
2117 (void) release_terminal();
2118
2119 /* Reset the console, but only if this is really init and we are freshly booted */
2120 if (getpid_cached() == 1 && !skip_setup)
2121 (void) console_setup();
2122 }
2123
2124 static bool early_skip_setup_check(int argc, char *argv[]) {
2125 bool found_deserialize = false;
2126 int i;
2127
2128 /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2129 * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2130 * anyway, even if in that case we also do deserialization. */
2131
2132 for (i = 1; i < argc; i++) {
2133
2134 if (streq(argv[i], "--switched-root"))
2135 return false; /* If we switched root, don't skip the setup. */
2136 else if (streq(argv[i], "--deserialize"))
2137 found_deserialize = true;
2138 }
2139
2140 return found_deserialize; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2141 }
2142
2143 int main(int argc, char *argv[]) {
2144
2145 dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL, userspace_timestamp = DUAL_TIMESTAMP_NULL, kernel_timestamp = DUAL_TIMESTAMP_NULL,
2146 security_start_timestamp = DUAL_TIMESTAMP_NULL, security_finish_timestamp = DUAL_TIMESTAMP_NULL;
2147 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0), saved_rlimit_memlock = RLIMIT_MAKE_CONST((rlim_t) -1);
2148 bool skip_setup, loaded_policy = false, queue_default_job = false, first_boot = false, reexecute = false;
2149 char *switch_root_dir = NULL, *switch_root_init = NULL;
2150 usec_t before_startup, after_startup;
2151 static char systemd[] = "systemd";
2152 char timespan[FORMAT_TIMESPAN_MAX];
2153 const char *shutdown_verb = NULL, *error_message = NULL;
2154 int r, retval = EXIT_FAILURE;
2155 Manager *m = NULL;
2156 FDSet *fds = NULL;
2157
2158 /* SysV compatibility: redirect init → telinit */
2159 redirect_telinit(argc, argv);
2160
2161 /* Take timestamps early on */
2162 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
2163 dual_timestamp_get(&userspace_timestamp);
2164
2165 /* Figure out whether we need to do initialize the system, or if we already did that because we are
2166 * reexecuting */
2167 skip_setup = early_skip_setup_check(argc, argv);
2168
2169 /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2170 * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
2171 program_invocation_short_name = systemd;
2172 (void) prctl(PR_SET_NAME, systemd);
2173
2174 /* Save the original command line */
2175 saved_argv = argv;
2176 saved_argc = argc;
2177
2178 /* Make sure that if the user says "syslog" we actually log to the journal. */
2179 log_set_upgrade_syslog_to_journal(true);
2180
2181 if (getpid_cached() == 1) {
2182 /* Disable the umask logic */
2183 umask(0);
2184
2185 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be activated
2186 * yet (even though the log socket for it exists). */
2187 log_set_prohibit_ipc(true);
2188
2189 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2190 * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2191 * child process right before execve()'ing the actual binary, at a point in time where socket
2192 * activation stderr/stdout area already set up. */
2193 log_set_always_reopen_console(true);
2194 }
2195
2196 if (getpid_cached() == 1 && detect_container() <= 0) {
2197
2198 /* Running outside of a container as PID 1 */
2199 arg_system = true;
2200 log_set_target(LOG_TARGET_KMSG);
2201 log_open();
2202
2203 if (in_initrd())
2204 initrd_timestamp = userspace_timestamp;
2205
2206 if (!skip_setup) {
2207 r = mount_setup_early();
2208 if (r < 0) {
2209 error_message = "Failed to mount early API filesystems";
2210 goto finish;
2211 }
2212
2213 r = initialize_security(
2214 &loaded_policy,
2215 &security_start_timestamp,
2216 &security_finish_timestamp,
2217 &error_message);
2218 if (r < 0)
2219 goto finish;
2220 }
2221
2222 if (mac_selinux_init() < 0) {
2223 error_message = "Failed to initialize SELinux policy";
2224 goto finish;
2225 }
2226
2227 if (!skip_setup)
2228 initialize_clock();
2229
2230 /* Set the default for later on, but don't actually
2231 * open the logs like this for now. Note that if we
2232 * are transitioning from the initrd there might still
2233 * be journal fd open, and we shouldn't attempt
2234 * opening that before we parsed /proc/cmdline which
2235 * might redirect output elsewhere. */
2236 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
2237
2238 } else if (getpid_cached() == 1) {
2239 /* Running inside a container, as PID 1 */
2240 arg_system = true;
2241 log_set_target(LOG_TARGET_CONSOLE);
2242 log_open();
2243
2244 /* For later on, see above... */
2245 log_set_target(LOG_TARGET_JOURNAL);
2246
2247 /* clear the kernel timestamp,
2248 * because we are in a container */
2249 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2250 } else {
2251 /* Running as user instance */
2252 arg_system = false;
2253 log_set_target(LOG_TARGET_AUTO);
2254 log_open();
2255
2256 /* clear the kernel timestamp,
2257 * because we are not PID 1 */
2258 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2259 }
2260
2261 initialize_coredump(skip_setup);
2262
2263 r = fixup_environment();
2264 if (r < 0) {
2265 log_emergency_errno(r, "Failed to fix up PID 1 environment: %m");
2266 error_message = "Failed to fix up PID1 environment";
2267 goto finish;
2268 }
2269
2270 if (arg_system) {
2271
2272 /* Try to figure out if we can use colors with the console. No
2273 * need to do that for user instances since they never log
2274 * into the console. */
2275 log_show_color(colors_enabled());
2276 r = make_null_stdio();
2277 if (r < 0)
2278 log_warning_errno(r, "Failed to redirect standard streams to /dev/null: %m");
2279 }
2280
2281 /* Mount /proc, /sys and friends, so that /proc/cmdline and
2282 * /proc/$PID/fd is available. */
2283 if (getpid_cached() == 1) {
2284
2285 /* Load the kernel modules early. */
2286 if (!skip_setup)
2287 kmod_setup();
2288
2289 r = mount_setup(loaded_policy);
2290 if (r < 0) {
2291 error_message = "Failed to mount API filesystems";
2292 goto finish;
2293 }
2294 }
2295
2296 /* Reset all signal handlers. */
2297 (void) reset_all_signal_handlers();
2298 (void) ignore_signals(SIGNALS_IGNORE, -1);
2299
2300 r = load_configuration(argc, argv, &error_message);
2301 if (r < 0)
2302 goto finish;
2303
2304 r = safety_checks();
2305 if (r < 0)
2306 goto finish;
2307
2308 if (IN_SET(arg_action, ACTION_TEST, ACTION_HELP, ACTION_DUMP_CONFIGURATION_ITEMS))
2309 (void) pager_open(arg_no_pager, false);
2310
2311 if (arg_action != ACTION_RUN)
2312 skip_setup = true;
2313
2314 if (arg_action == ACTION_HELP) {
2315 retval = help();
2316 goto finish;
2317 } else if (arg_action == ACTION_VERSION) {
2318 retval = version();
2319 goto finish;
2320 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
2321 unit_dump_config_items(stdout);
2322 retval = EXIT_SUCCESS;
2323 goto finish;
2324 }
2325
2326 assert_se(IN_SET(arg_action, ACTION_RUN, ACTION_TEST));
2327
2328 /* Move out of the way, so that we won't block unmounts */
2329 assert_se(chdir("/") == 0);
2330
2331 if (arg_action == ACTION_RUN) {
2332
2333 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2334 log_close();
2335
2336 /* Remember open file descriptors for later deserialization */
2337 r = collect_fds(&fds, &error_message);
2338 if (r < 0)
2339 goto finish;
2340
2341 /* Give up any control of the console, but make sure its initialized. */
2342 setup_console_terminal(skip_setup);
2343
2344 /* Open the logging devices, if possible and necessary */
2345 log_open();
2346 }
2347
2348 log_execution_mode(&first_boot);
2349
2350 r = initialize_runtime(skip_setup,
2351 &saved_rlimit_nofile,
2352 &saved_rlimit_memlock,
2353 &error_message);
2354 if (r < 0)
2355 goto finish;
2356
2357 r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER,
2358 arg_action == ACTION_TEST ? MANAGER_TEST_FULL : 0,
2359 &m);
2360 if (r < 0) {
2361 log_emergency_errno(r, "Failed to allocate manager object: %m");
2362 error_message = "Failed to allocate manager object";
2363 goto finish;
2364 }
2365
2366 m->timestamps[MANAGER_TIMESTAMP_KERNEL] = kernel_timestamp;
2367 m->timestamps[MANAGER_TIMESTAMP_INITRD] = initrd_timestamp;
2368 m->timestamps[MANAGER_TIMESTAMP_USERSPACE] = userspace_timestamp;
2369 m->timestamps[MANAGER_TIMESTAMP_SECURITY_START] = security_start_timestamp;
2370 m->timestamps[MANAGER_TIMESTAMP_SECURITY_FINISH] = security_finish_timestamp;
2371
2372 set_manager_defaults(m);
2373 set_manager_settings(m);
2374 manager_set_first_boot(m, first_boot);
2375
2376 /* Remember whether we should queue the default job */
2377 queue_default_job = !arg_serialization || arg_switched_root;
2378
2379 before_startup = now(CLOCK_MONOTONIC);
2380
2381 r = manager_startup(m, arg_serialization, fds);
2382 if (r < 0) {
2383 log_error_errno(r, "Failed to fully start up daemon: %m");
2384 error_message = "Failed to start up manager";
2385 goto finish;
2386 }
2387
2388 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2389 fds = fdset_free(fds);
2390 arg_serialization = safe_fclose(arg_serialization);
2391
2392 if (queue_default_job) {
2393 r = do_queue_default_job(m, &error_message);
2394 if (r < 0)
2395 goto finish;
2396 }
2397
2398 after_startup = now(CLOCK_MONOTONIC);
2399
2400 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
2401 "Loaded units and determined initial transaction in %s.",
2402 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
2403
2404 if (arg_action == ACTION_TEST) {
2405 test_summary(m);
2406 retval = EXIT_SUCCESS;
2407 goto finish;
2408 }
2409
2410 (void) invoke_main_loop(m,
2411 &reexecute,
2412 &retval,
2413 &shutdown_verb,
2414 &fds,
2415 &switch_root_dir,
2416 &switch_root_init,
2417 &error_message);
2418
2419 finish:
2420 pager_close();
2421
2422 if (m)
2423 arg_shutdown_watchdog = m->shutdown_watchdog;
2424
2425 m = manager_free(m);
2426
2427 free_arguments();
2428 mac_selinux_finish();
2429
2430 if (reexecute)
2431 do_reexecute(argc, argv,
2432 &saved_rlimit_nofile,
2433 &saved_rlimit_memlock,
2434 fds,
2435 switch_root_dir,
2436 switch_root_init,
2437 &error_message); /* This only returns if reexecution failed */
2438
2439 arg_serialization = safe_fclose(arg_serialization);
2440 fds = fdset_free(fds);
2441
2442 #if HAVE_VALGRIND_VALGRIND_H
2443 /* If we are PID 1 and running under valgrind, then let's exit
2444 * here explicitly. valgrind will only generate nice output on
2445 * exit(), not on exec(), hence let's do the former not the
2446 * latter here. */
2447 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND) {
2448 /* Cleanup watchdog_device strings for valgrind. We need them
2449 * in become_shutdown() so normally we cannot free them yet. */
2450 watchdog_free_device();
2451 arg_watchdog_device = mfree(arg_watchdog_device);
2452 return retval;
2453 }
2454 #endif
2455
2456 if (shutdown_verb) {
2457 r = become_shutdown(shutdown_verb, retval);
2458
2459 log_error_errno(r, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
2460 error_message = "Failed to execute shutdown binary";
2461 }
2462
2463 watchdog_free_device();
2464 arg_watchdog_device = mfree(arg_watchdog_device);
2465
2466 if (getpid_cached() == 1) {
2467 if (error_message)
2468 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2469 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
2470 "%s, freezing.", error_message);
2471 freeze_or_reboot();
2472 }
2473
2474 return retval;
2475 }
Unnamed repository; edit this file 'description' to name the repository.