]> 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 #9201 from poettering/integration-test-unfuck
[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.h"
40 #include "dbus-manager.h"
41 #include "def.h"
42 #include "emergency-action.h"
43 #include "env-util.h"
44 #include "exit-status.h"
45 #include "fd-util.h"
46 #include "fdset.h"
47 #include "fileio.h"
48 #include "format-util.h"
49 #include "fs-util.h"
50 #include "hostname-setup.h"
51 #include "ima-setup.h"
52 #include "killall.h"
53 #include "kmod-setup.h"
54 #include "load-fragment.h"
55 #include "log.h"
56 #include "loopback-setup.h"
57 #include "machine-id-setup.h"
58 #include "manager.h"
59 #include "missing.h"
60 #include "mount-setup.h"
61 #include "os-util.h"
62 #include "pager.h"
63 #include "parse-util.h"
64 #include "path-util.h"
65 #include "proc-cmdline.h"
66 #include "process-util.h"
67 #include "raw-clone.h"
68 #include "rlimit-util.h"
69 #if HAVE_SECCOMP
70 #include "seccomp-util.h"
71 #endif
72 #include "selinux-setup.h"
73 #include "selinux-util.h"
74 #include "signal-util.h"
75 #include "smack-setup.h"
76 #include "special.h"
77 #include "stat-util.h"
78 #include "stdio-util.h"
79 #include "strv.h"
80 #include "switch-root.h"
81 #include "terminal-util.h"
82 #include "umask-util.h"
83 #include "user-util.h"
84 #include "util.h"
85 #include "virt.h"
86 #include "watchdog.h"
87
88 static enum {
89 ACTION_RUN,
90 ACTION_HELP,
91 ACTION_VERSION,
92 ACTION_TEST,
93 ACTION_DUMP_CONFIGURATION_ITEMS,
94 ACTION_DUMP_BUS_PROPERTIES,
95 } arg_action = ACTION_RUN;
96 static char *arg_default_unit = NULL;
97 static bool arg_system = false;
98 static bool arg_dump_core = true;
99 static int arg_crash_chvt = -1;
100 static bool arg_crash_shell = false;
101 static bool arg_crash_reboot = false;
102 static char *arg_confirm_spawn = NULL;
103 static ShowStatus arg_show_status = _SHOW_STATUS_UNSET;
104 static bool arg_switched_root = false;
105 static bool arg_no_pager = false;
106 static bool arg_service_watchdogs = true;
107 static char ***arg_join_controllers = NULL;
108 static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
109 static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
110 static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
111 static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
112 static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
113 static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
114 static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
115 static usec_t arg_runtime_watchdog = 0;
116 static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
117 static char *arg_watchdog_device = NULL;
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 bool arg_no_new_privs = false;
122 static nsec_t arg_timer_slack_nsec = NSEC_INFINITY;
123 static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
124 static Set* arg_syscall_archs = NULL;
125 static FILE* arg_serialization = NULL;
126 static bool arg_default_cpu_accounting = false;
127 static bool arg_default_io_accounting = false;
128 static bool arg_default_ip_accounting = false;
129 static bool arg_default_blockio_accounting = false;
130 static bool arg_default_memory_accounting = MEMORY_ACCOUNTING_DEFAULT;
131 static bool arg_default_tasks_accounting = true;
132 static uint64_t arg_default_tasks_max = UINT64_MAX;
133 static sd_id128_t arg_machine_id = {};
134 static EmergencyAction arg_cad_burst_action = EMERGENCY_ACTION_REBOOT_FORCE;
135
136 _noreturn_ static void freeze_or_reboot(void) {
137
138 if (arg_crash_reboot) {
139 log_notice("Rebooting in 10s...");
140 (void) sleep(10);
141
142 log_notice("Rebooting now...");
143 (void) reboot(RB_AUTOBOOT);
144 log_emergency_errno(errno, "Failed to reboot: %m");
145 }
146
147 log_emergency("Freezing execution.");
148 freeze();
149 }
150
151 _noreturn_ static void crash(int sig) {
152 struct sigaction sa;
153 pid_t pid;
154
155 if (getpid_cached() != 1)
156 /* Pass this on immediately, if this is not PID 1 */
157 (void) raise(sig);
158 else if (!arg_dump_core)
159 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
160 else {
161 sa = (struct sigaction) {
162 .sa_handler = nop_signal_handler,
163 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
164 };
165
166 /* We want to wait for the core process, hence let's enable SIGCHLD */
167 (void) sigaction(SIGCHLD, &sa, NULL);
168
169 pid = raw_clone(SIGCHLD);
170 if (pid < 0)
171 log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
172 else if (pid == 0) {
173 /* Enable default signal handler for core dump */
174
175 sa = (struct sigaction) {
176 .sa_handler = SIG_DFL,
177 };
178 (void) sigaction(sig, &sa, NULL);
179
180 /* Don't limit the coredump size */
181 (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
182
183 /* Just to be sure... */
184 (void) chdir("/");
185
186 /* Raise the signal again */
187 pid = raw_getpid();
188 (void) kill(pid, sig); /* raise() would kill the parent */
189
190 assert_not_reached("We shouldn't be here...");
191 _exit(EXIT_FAILURE);
192 } else {
193 siginfo_t status;
194 int r;
195
196 /* Order things nicely. */
197 r = wait_for_terminate(pid, &status);
198 if (r < 0)
199 log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
200 else if (status.si_code != CLD_DUMPED)
201 log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
202 signal_to_string(sig),
203 pid, sigchld_code_to_string(status.si_code),
204 status.si_status,
205 strna(status.si_code == CLD_EXITED
206 ? exit_status_to_string(status.si_status, EXIT_STATUS_MINIMAL)
207 : signal_to_string(status.si_status)));
208 else
209 log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
210 }
211 }
212
213 if (arg_crash_chvt >= 0)
214 (void) chvt(arg_crash_chvt);
215
216 sa = (struct sigaction) {
217 .sa_handler = SIG_IGN,
218 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
219 };
220
221 /* Let the kernel reap children for us */
222 (void) sigaction(SIGCHLD, &sa, NULL);
223
224 if (arg_crash_shell) {
225 log_notice("Executing crash shell in 10s...");
226 (void) sleep(10);
227
228 pid = raw_clone(SIGCHLD);
229 if (pid < 0)
230 log_emergency_errno(errno, "Failed to fork off crash shell: %m");
231 else if (pid == 0) {
232 (void) setsid();
233 (void) make_console_stdio();
234 (void) execle("/bin/sh", "/bin/sh", NULL, environ);
235
236 log_emergency_errno(errno, "execle() failed: %m");
237 _exit(EXIT_FAILURE);
238 } else {
239 log_info("Spawned crash shell as PID "PID_FMT".", pid);
240 (void) wait_for_terminate(pid, NULL);
241 }
242 }
243
244 freeze_or_reboot();
245 }
246
247 static void install_crash_handler(void) {
248 static const struct sigaction sa = {
249 .sa_handler = crash,
250 .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
251 };
252 int r;
253
254 /* We ignore the return value here, since, we don't mind if we
255 * cannot set up a crash handler */
256 r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
257 if (r < 0)
258 log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
259 }
260
261 static int console_setup(void) {
262 _cleanup_close_ int tty_fd = -1;
263 int r;
264
265 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
266 if (tty_fd < 0)
267 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
268
269 /* We don't want to force text mode. plymouth may be showing
270 * pictures already from initrd. */
271 r = reset_terminal_fd(tty_fd, false);
272 if (r < 0)
273 return log_error_errno(r, "Failed to reset /dev/console: %m");
274
275 return 0;
276 }
277
278 static int parse_crash_chvt(const char *value) {
279 int b;
280
281 if (safe_atoi(value, &arg_crash_chvt) >= 0)
282 return 0;
283
284 b = parse_boolean(value);
285 if (b < 0)
286 return b;
287
288 if (b > 0)
289 arg_crash_chvt = 0; /* switch to where kmsg goes */
290 else
291 arg_crash_chvt = -1; /* turn off switching */
292
293 return 0;
294 }
295
296 static int parse_confirm_spawn(const char *value, char **console) {
297 char *s;
298 int r;
299
300 r = value ? parse_boolean(value) : 1;
301 if (r == 0) {
302 *console = NULL;
303 return 0;
304 }
305
306 if (r > 0) /* on with default tty */
307 s = strdup("/dev/console");
308 else if (is_path(value)) /* on with fully qualified path */
309 s = strdup(value);
310 else /* on with only a tty file name, not a fully qualified path */
311 s = strjoin("/dev/", value);
312 if (!s)
313 return -ENOMEM;
314
315 *console = s;
316 return 0;
317 }
318
319 static int set_machine_id(const char *m) {
320 sd_id128_t t;
321 assert(m);
322
323 if (sd_id128_from_string(m, &t) < 0)
324 return -EINVAL;
325
326 if (sd_id128_is_null(t))
327 return -EINVAL;
328
329 arg_machine_id = t;
330 return 0;
331 }
332
333 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
334
335 int r;
336
337 assert(key);
338
339 if (STR_IN_SET(key, "systemd.unit", "rd.systemd.unit")) {
340
341 if (proc_cmdline_value_missing(key, value))
342 return 0;
343
344 if (!unit_name_is_valid(value, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE))
345 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key, value);
346 else if (in_initrd() == !!startswith(key, "rd.")) {
347 if (free_and_strdup(&arg_default_unit, value) < 0)
348 return log_oom();
349 }
350
351 } else if (proc_cmdline_key_streq(key, "systemd.dump_core")) {
352
353 r = value ? parse_boolean(value) : true;
354 if (r < 0)
355 log_warning("Failed to parse dump core switch %s. Ignoring.", value);
356 else
357 arg_dump_core = r;
358
359 } else if (proc_cmdline_key_streq(key, "systemd.crash_chvt")) {
360
361 if (!value)
362 arg_crash_chvt = 0; /* turn on */
363 else if (parse_crash_chvt(value) < 0)
364 log_warning("Failed to parse crash chvt switch %s. Ignoring.", value);
365
366 } else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) {
367
368 r = value ? parse_boolean(value) : true;
369 if (r < 0)
370 log_warning("Failed to parse crash shell switch %s. Ignoring.", value);
371 else
372 arg_crash_shell = r;
373
374 } else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) {
375
376 r = value ? parse_boolean(value) : true;
377 if (r < 0)
378 log_warning("Failed to parse crash reboot switch %s. Ignoring.", value);
379 else
380 arg_crash_reboot = r;
381
382 } else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) {
383 char *s;
384
385 r = parse_confirm_spawn(value, &s);
386 if (r < 0)
387 log_warning_errno(r, "Failed to parse confirm_spawn switch %s. Ignoring.", value);
388 else {
389 free(arg_confirm_spawn);
390 arg_confirm_spawn = s;
391 }
392
393 } else if (proc_cmdline_key_streq(key, "systemd.service_watchdogs")) {
394
395 r = value ? parse_boolean(value) : true;
396 if (r < 0)
397 log_warning("Failed to parse service watchdog switch %s. Ignoring.", value);
398 else
399 arg_service_watchdogs = r;
400
401 } else if (proc_cmdline_key_streq(key, "systemd.show_status")) {
402
403 if (value) {
404 r = parse_show_status(value, &arg_show_status);
405 if (r < 0)
406 log_warning("Failed to parse show status switch %s. Ignoring.", value);
407 } else
408 arg_show_status = SHOW_STATUS_YES;
409
410 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) {
411
412 if (proc_cmdline_value_missing(key, value))
413 return 0;
414
415 r = exec_output_from_string(value);
416 if (r < 0)
417 log_warning("Failed to parse default standard output switch %s. Ignoring.", value);
418 else
419 arg_default_std_output = r;
420
421 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) {
422
423 if (proc_cmdline_value_missing(key, value))
424 return 0;
425
426 r = exec_output_from_string(value);
427 if (r < 0)
428 log_warning("Failed to parse default standard error switch %s. Ignoring.", value);
429 else
430 arg_default_std_error = r;
431
432 } else if (streq(key, "systemd.setenv")) {
433
434 if (proc_cmdline_value_missing(key, value))
435 return 0;
436
437 if (env_assignment_is_valid(value)) {
438 char **env;
439
440 env = strv_env_set(arg_default_environment, value);
441 if (!env)
442 return log_oom();
443
444 arg_default_environment = env;
445 } else
446 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
447
448 } else if (proc_cmdline_key_streq(key, "systemd.machine_id")) {
449
450 if (proc_cmdline_value_missing(key, value))
451 return 0;
452
453 r = set_machine_id(value);
454 if (r < 0)
455 log_warning("MachineID '%s' is not valid. Ignoring.", value);
456
457 } else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) {
458
459 if (proc_cmdline_value_missing(key, value))
460 return 0;
461
462 r = parse_sec(value, &arg_default_timeout_start_usec);
463 if (r < 0)
464 log_warning_errno(r, "Failed to parse default start timeout: %s, ignoring.", value);
465
466 if (arg_default_timeout_start_usec <= 0)
467 arg_default_timeout_start_usec = USEC_INFINITY;
468
469 } else if (proc_cmdline_key_streq(key, "systemd.watchdog_device")) {
470
471 if (proc_cmdline_value_missing(key, value))
472 return 0;
473
474 parse_path_argument_and_warn(value, false, &arg_watchdog_device);
475
476 } else if (streq(key, "quiet") && !value) {
477
478 if (arg_show_status == _SHOW_STATUS_UNSET)
479 arg_show_status = SHOW_STATUS_AUTO;
480
481 } else if (streq(key, "debug") && !value) {
482
483 /* Note that log_parse_environment() handles 'debug'
484 * too, and sets the log level to LOG_DEBUG. */
485
486 if (detect_container() > 0)
487 log_set_target(LOG_TARGET_CONSOLE);
488
489 } else if (!value) {
490 const char *target;
491
492 /* SysV compatibility */
493 target = runlevel_to_target(key);
494 if (target)
495 return free_and_strdup(&arg_default_unit, target);
496 }
497
498 return 0;
499 }
500
501 #define DEFINE_SETTER(name, func, descr) \
502 static int name(const char *unit, \
503 const char *filename, \
504 unsigned line, \
505 const char *section, \
506 unsigned section_line, \
507 const char *lvalue, \
508 int ltype, \
509 const char *rvalue, \
510 void *data, \
511 void *userdata) { \
512 \
513 int r; \
514 \
515 assert(filename); \
516 assert(lvalue); \
517 assert(rvalue); \
518 \
519 r = func(rvalue); \
520 if (r < 0) \
521 log_syntax(unit, LOG_ERR, filename, line, r, \
522 "Invalid " descr "'%s': %m", \
523 rvalue); \
524 \
525 return 0; \
526 }
527
528 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level")
529 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target")
530 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" )
531 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location")
532
533 static int config_parse_cpu_affinity2(
534 const char *unit,
535 const char *filename,
536 unsigned line,
537 const char *section,
538 unsigned section_line,
539 const char *lvalue,
540 int ltype,
541 const char *rvalue,
542 void *data,
543 void *userdata) {
544
545 _cleanup_cpu_free_ cpu_set_t *c = NULL;
546 int ncpus;
547
548 ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue);
549 if (ncpus < 0)
550 return ncpus;
551
552 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
553 log_warning_errno(errno, "Failed to set CPU affinity: %m");
554
555 return 0;
556 }
557
558 static int config_parse_show_status(
559 const char* unit,
560 const char *filename,
561 unsigned line,
562 const char *section,
563 unsigned section_line,
564 const char *lvalue,
565 int ltype,
566 const char *rvalue,
567 void *data,
568 void *userdata) {
569
570 int k;
571 ShowStatus *b = data;
572
573 assert(filename);
574 assert(lvalue);
575 assert(rvalue);
576 assert(data);
577
578 k = parse_show_status(rvalue, b);
579 if (k < 0) {
580 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
581 return 0;
582 }
583
584 return 0;
585 }
586
587 static int config_parse_output_restricted(
588 const char* unit,
589 const char *filename,
590 unsigned line,
591 const char *section,
592 unsigned section_line,
593 const char *lvalue,
594 int ltype,
595 const char *rvalue,
596 void *data,
597 void *userdata) {
598
599 ExecOutput t, *eo = data;
600
601 assert(filename);
602 assert(lvalue);
603 assert(rvalue);
604 assert(data);
605
606 t = exec_output_from_string(rvalue);
607 if (t < 0) {
608 log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse output type, ignoring: %s", rvalue);
609 return 0;
610 }
611
612 if (IN_SET(t, EXEC_OUTPUT_SOCKET, EXEC_OUTPUT_NAMED_FD, EXEC_OUTPUT_FILE)) {
613 log_syntax(unit, LOG_ERR, filename, line, 0, "Standard output types socket, fd:, file: are not supported as defaults, ignoring: %s", rvalue);
614 return 0;
615 }
616
617 *eo = t;
618 return 0;
619 }
620
621 static int config_parse_crash_chvt(
622 const char* unit,
623 const char *filename,
624 unsigned line,
625 const char *section,
626 unsigned section_line,
627 const char *lvalue,
628 int ltype,
629 const char *rvalue,
630 void *data,
631 void *userdata) {
632
633 int r;
634
635 assert(filename);
636 assert(lvalue);
637 assert(rvalue);
638
639 r = parse_crash_chvt(rvalue);
640 if (r < 0) {
641 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
642 return 0;
643 }
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", "WatchdogDevice", config_parse_path, 0, &arg_watchdog_device },
666 { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
667 { "Manager", "NoNewPrivileges", config_parse_bool, 0, &arg_no_new_privs },
668 #if HAVE_SECCOMP
669 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
670 #endif
671 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
672 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
673 { "Manager", "DefaultStandardOutput", config_parse_output_restricted,0, &arg_default_std_output },
674 { "Manager", "DefaultStandardError", config_parse_output_restricted,0, &arg_default_std_error },
675 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
676 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
677 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
678 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */
679 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec, 0, &arg_default_start_limit_interval },
680 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
681 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
682 { "Manager", "DefaultLimitCPU", config_parse_rlimit, RLIMIT_CPU, arg_default_rlimit },
683 { "Manager", "DefaultLimitFSIZE", config_parse_rlimit, RLIMIT_FSIZE, arg_default_rlimit },
684 { "Manager", "DefaultLimitDATA", config_parse_rlimit, RLIMIT_DATA, arg_default_rlimit },
685 { "Manager", "DefaultLimitSTACK", config_parse_rlimit, RLIMIT_STACK, arg_default_rlimit },
686 { "Manager", "DefaultLimitCORE", config_parse_rlimit, RLIMIT_CORE, arg_default_rlimit },
687 { "Manager", "DefaultLimitRSS", config_parse_rlimit, RLIMIT_RSS, arg_default_rlimit },
688 { "Manager", "DefaultLimitNOFILE", config_parse_rlimit, RLIMIT_NOFILE, arg_default_rlimit },
689 { "Manager", "DefaultLimitAS", config_parse_rlimit, RLIMIT_AS, arg_default_rlimit },
690 { "Manager", "DefaultLimitNPROC", config_parse_rlimit, RLIMIT_NPROC, arg_default_rlimit },
691 { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit, RLIMIT_MEMLOCK, arg_default_rlimit },
692 { "Manager", "DefaultLimitLOCKS", config_parse_rlimit, RLIMIT_LOCKS, arg_default_rlimit },
693 { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit, RLIMIT_SIGPENDING, arg_default_rlimit },
694 { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit, RLIMIT_MSGQUEUE, arg_default_rlimit },
695 { "Manager", "DefaultLimitNICE", config_parse_rlimit, RLIMIT_NICE, arg_default_rlimit },
696 { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit, RLIMIT_RTPRIO, arg_default_rlimit },
697 { "Manager", "DefaultLimitRTTIME", config_parse_rlimit, RLIMIT_RTTIME, arg_default_rlimit },
698 { "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_default_cpu_accounting },
699 { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting },
700 { "Manager", "DefaultIPAccounting", config_parse_bool, 0, &arg_default_ip_accounting },
701 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
702 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
703 { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
704 { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max },
705 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action, 0, &arg_cad_burst_action },
706 {}
707 };
708
709 const char *fn, *conf_dirs_nulstr;
710
711 fn = arg_system ?
712 PKGSYSCONFDIR "/system.conf" :
713 PKGSYSCONFDIR "/user.conf";
714
715 conf_dirs_nulstr = arg_system ?
716 CONF_PATHS_NULSTR("systemd/system.conf.d") :
717 CONF_PATHS_NULSTR("systemd/user.conf.d");
718
719 (void) config_parse_many_nulstr(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, CONFIG_PARSE_WARN, NULL);
720
721 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
722 * like everywhere else. */
723 if (arg_default_timeout_start_usec <= 0)
724 arg_default_timeout_start_usec = USEC_INFINITY;
725 if (arg_default_timeout_stop_usec <= 0)
726 arg_default_timeout_stop_usec = USEC_INFINITY;
727
728 return 0;
729 }
730
731 static void set_manager_defaults(Manager *m) {
732
733 assert(m);
734
735 /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
736 * affect the manager itself, but are just what newly allocated units will have set if they haven't set
737 * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
738
739 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
740 m->default_std_output = arg_default_std_output;
741 m->default_std_error = arg_default_std_error;
742 m->default_timeout_start_usec = arg_default_timeout_start_usec;
743 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
744 m->default_restart_usec = arg_default_restart_usec;
745 m->default_start_limit_interval = arg_default_start_limit_interval;
746 m->default_start_limit_burst = arg_default_start_limit_burst;
747 m->default_cpu_accounting = arg_default_cpu_accounting;
748 m->default_io_accounting = arg_default_io_accounting;
749 m->default_ip_accounting = arg_default_ip_accounting;
750 m->default_blockio_accounting = arg_default_blockio_accounting;
751 m->default_memory_accounting = arg_default_memory_accounting;
752 m->default_tasks_accounting = arg_default_tasks_accounting;
753 m->default_tasks_max = arg_default_tasks_max;
754
755 manager_set_default_rlimits(m, arg_default_rlimit);
756 manager_environment_add(m, NULL, arg_default_environment);
757 }
758
759 static void set_manager_settings(Manager *m) {
760
761 assert(m);
762
763 /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
764 * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
765
766 m->confirm_spawn = arg_confirm_spawn;
767 m->service_watchdogs = arg_service_watchdogs;
768 m->runtime_watchdog = arg_runtime_watchdog;
769 m->shutdown_watchdog = arg_shutdown_watchdog;
770 m->cad_burst_action = arg_cad_burst_action;
771
772 manager_set_show_status(m, arg_show_status);
773 }
774
775 static int parse_argv(int argc, char *argv[]) {
776 enum {
777 ARG_LOG_LEVEL = 0x100,
778 ARG_LOG_TARGET,
779 ARG_LOG_COLOR,
780 ARG_LOG_LOCATION,
781 ARG_UNIT,
782 ARG_SYSTEM,
783 ARG_USER,
784 ARG_TEST,
785 ARG_NO_PAGER,
786 ARG_VERSION,
787 ARG_DUMP_CONFIGURATION_ITEMS,
788 ARG_DUMP_BUS_PROPERTIES,
789 ARG_DUMP_CORE,
790 ARG_CRASH_CHVT,
791 ARG_CRASH_SHELL,
792 ARG_CRASH_REBOOT,
793 ARG_CONFIRM_SPAWN,
794 ARG_SHOW_STATUS,
795 ARG_DESERIALIZE,
796 ARG_SWITCHED_ROOT,
797 ARG_DEFAULT_STD_OUTPUT,
798 ARG_DEFAULT_STD_ERROR,
799 ARG_MACHINE_ID,
800 ARG_SERVICE_WATCHDOGS,
801 };
802
803 static const struct option options[] = {
804 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
805 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
806 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
807 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
808 { "unit", required_argument, NULL, ARG_UNIT },
809 { "system", no_argument, NULL, ARG_SYSTEM },
810 { "user", no_argument, NULL, ARG_USER },
811 { "test", no_argument, NULL, ARG_TEST },
812 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
813 { "help", no_argument, NULL, 'h' },
814 { "version", no_argument, NULL, ARG_VERSION },
815 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
816 { "dump-bus-properties", no_argument, NULL, ARG_DUMP_BUS_PROPERTIES },
817 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
818 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
819 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
820 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
821 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
822 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
823 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
824 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
825 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
826 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
827 { "machine-id", required_argument, NULL, ARG_MACHINE_ID },
828 { "service-watchdogs", required_argument, NULL, ARG_SERVICE_WATCHDOGS },
829 {}
830 };
831
832 int c, r;
833
834 assert(argc >= 1);
835 assert(argv);
836
837 if (getpid_cached() == 1)
838 opterr = 0;
839
840 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
841
842 switch (c) {
843
844 case ARG_LOG_LEVEL:
845 r = log_set_max_level_from_string(optarg);
846 if (r < 0) {
847 log_error("Failed to parse log level %s.", optarg);
848 return r;
849 }
850
851 break;
852
853 case ARG_LOG_TARGET:
854 r = log_set_target_from_string(optarg);
855 if (r < 0) {
856 log_error("Failed to parse log target %s.", optarg);
857 return r;
858 }
859
860 break;
861
862 case ARG_LOG_COLOR:
863
864 if (optarg) {
865 r = log_show_color_from_string(optarg);
866 if (r < 0) {
867 log_error("Failed to parse log color setting %s.", optarg);
868 return r;
869 }
870 } else
871 log_show_color(true);
872
873 break;
874
875 case ARG_LOG_LOCATION:
876 if (optarg) {
877 r = log_show_location_from_string(optarg);
878 if (r < 0) {
879 log_error("Failed to parse log location setting %s.", optarg);
880 return r;
881 }
882 } else
883 log_show_location(true);
884
885 break;
886
887 case ARG_DEFAULT_STD_OUTPUT:
888 r = exec_output_from_string(optarg);
889 if (r < 0) {
890 log_error("Failed to parse default standard output setting %s.", optarg);
891 return r;
892 } else
893 arg_default_std_output = r;
894 break;
895
896 case ARG_DEFAULT_STD_ERROR:
897 r = exec_output_from_string(optarg);
898 if (r < 0) {
899 log_error("Failed to parse default standard error output setting %s.", optarg);
900 return r;
901 } else
902 arg_default_std_error = r;
903 break;
904
905 case ARG_UNIT:
906 r = free_and_strdup(&arg_default_unit, optarg);
907 if (r < 0)
908 return log_error_errno(r, "Failed to set default unit %s: %m", optarg);
909
910 break;
911
912 case ARG_SYSTEM:
913 arg_system = true;
914 break;
915
916 case ARG_USER:
917 arg_system = false;
918 break;
919
920 case ARG_TEST:
921 arg_action = ACTION_TEST;
922 break;
923
924 case ARG_NO_PAGER:
925 arg_no_pager = true;
926 break;
927
928 case ARG_VERSION:
929 arg_action = ACTION_VERSION;
930 break;
931
932 case ARG_DUMP_CONFIGURATION_ITEMS:
933 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
934 break;
935
936 case ARG_DUMP_BUS_PROPERTIES:
937 arg_action = ACTION_DUMP_BUS_PROPERTIES;
938 break;
939
940 case ARG_DUMP_CORE:
941 if (!optarg)
942 arg_dump_core = true;
943 else {
944 r = parse_boolean(optarg);
945 if (r < 0)
946 return log_error_errno(r, "Failed to parse dump core boolean: %s", optarg);
947 arg_dump_core = r;
948 }
949 break;
950
951 case ARG_CRASH_CHVT:
952 r = parse_crash_chvt(optarg);
953 if (r < 0)
954 return log_error_errno(r, "Failed to parse crash virtual terminal index: %s", optarg);
955 break;
956
957 case ARG_CRASH_SHELL:
958 if (!optarg)
959 arg_crash_shell = true;
960 else {
961 r = parse_boolean(optarg);
962 if (r < 0)
963 return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg);
964 arg_crash_shell = r;
965 }
966 break;
967
968 case ARG_CRASH_REBOOT:
969 if (!optarg)
970 arg_crash_reboot = true;
971 else {
972 r = parse_boolean(optarg);
973 if (r < 0)
974 return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg);
975 arg_crash_reboot = r;
976 }
977 break;
978
979 case ARG_CONFIRM_SPAWN:
980 arg_confirm_spawn = mfree(arg_confirm_spawn);
981
982 r = parse_confirm_spawn(optarg, &arg_confirm_spawn);
983 if (r < 0)
984 return log_error_errno(r, "Failed to parse confirm spawn option: %m");
985 break;
986
987 case ARG_SERVICE_WATCHDOGS:
988 r = parse_boolean(optarg);
989 if (r < 0)
990 return log_error_errno(r, "Failed to parse service watchdogs boolean: %s", optarg);
991 arg_service_watchdogs = r;
992 break;
993
994 case ARG_SHOW_STATUS:
995 if (optarg) {
996 r = parse_show_status(optarg, &arg_show_status);
997 if (r < 0) {
998 log_error("Failed to parse show status boolean %s.", optarg);
999 return r;
1000 }
1001 } else
1002 arg_show_status = SHOW_STATUS_YES;
1003 break;
1004
1005 case ARG_DESERIALIZE: {
1006 int fd;
1007 FILE *f;
1008
1009 r = safe_atoi(optarg, &fd);
1010 if (r < 0 || fd < 0) {
1011 log_error("Failed to parse deserialize option %s.", optarg);
1012 return -EINVAL;
1013 }
1014
1015 (void) fd_cloexec(fd, true);
1016
1017 f = fdopen(fd, "r");
1018 if (!f)
1019 return log_error_errno(errno, "Failed to open serialization fd: %m");
1020
1021 safe_fclose(arg_serialization);
1022 arg_serialization = f;
1023
1024 break;
1025 }
1026
1027 case ARG_SWITCHED_ROOT:
1028 arg_switched_root = true;
1029 break;
1030
1031 case ARG_MACHINE_ID:
1032 r = set_machine_id(optarg);
1033 if (r < 0)
1034 return log_error_errno(r, "MachineID '%s' is not valid.", optarg);
1035 break;
1036
1037 case 'h':
1038 arg_action = ACTION_HELP;
1039 break;
1040
1041 case 'D':
1042 log_set_max_level(LOG_DEBUG);
1043 break;
1044
1045 case 'b':
1046 case 's':
1047 case 'z':
1048 /* Just to eat away the sysvinit kernel
1049 * cmdline args without getopt() error
1050 * messages that we'll parse in
1051 * parse_proc_cmdline_word() or ignore. */
1052
1053 case '?':
1054 if (getpid_cached() != 1)
1055 return -EINVAL;
1056 else
1057 return 0;
1058
1059 default:
1060 assert_not_reached("Unhandled option code.");
1061 }
1062
1063 if (optind < argc && getpid_cached() != 1) {
1064 /* Hmm, when we aren't run as init system
1065 * let's complain about excess arguments */
1066
1067 log_error("Excess arguments.");
1068 return -EINVAL;
1069 }
1070
1071 return 0;
1072 }
1073
1074 static int help(void) {
1075
1076 printf("%s [OPTIONS...]\n\n"
1077 "Starts up and maintains the system or user services.\n\n"
1078 " -h --help Show this help\n"
1079 " --version Show version\n"
1080 " --test Determine startup sequence, dump it and exit\n"
1081 " --no-pager Do not pipe output into a pager\n"
1082 " --dump-configuration-items Dump understood unit configuration items\n"
1083 " --dump-bus-properties Dump exposed bus properties\n"
1084 " --unit=UNIT Set default unit\n"
1085 " --system Run a system instance, even if PID != 1\n"
1086 " --user Run a user instance\n"
1087 " --dump-core[=BOOL] Dump core on crash\n"
1088 " --crash-vt=NR Change to specified VT on crash\n"
1089 " --crash-reboot[=BOOL] Reboot on crash\n"
1090 " --crash-shell[=BOOL] Run shell on crash\n"
1091 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1092 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1093 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1094 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1095 " --log-color[=BOOL] Highlight important log messages\n"
1096 " --log-location[=BOOL] Include code location in log messages\n"
1097 " --default-standard-output= Set default standard output for services\n"
1098 " --default-standard-error= Set default standard error output for services\n",
1099 program_invocation_short_name);
1100
1101 return 0;
1102 }
1103
1104 static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
1105 _cleanup_fdset_free_ FDSet *fds = NULL;
1106 _cleanup_fclose_ FILE *f = NULL;
1107 int r;
1108
1109 assert(m);
1110 assert(_f);
1111 assert(_fds);
1112
1113 r = manager_open_serialization(m, &f);
1114 if (r < 0)
1115 return log_error_errno(r, "Failed to create serialization file: %m");
1116
1117 /* Make sure nothing is really destructed when we shut down */
1118 m->n_reloading++;
1119 bus_manager_send_reloading(m, true);
1120
1121 fds = fdset_new();
1122 if (!fds)
1123 return log_oom();
1124
1125 r = manager_serialize(m, f, fds, switching_root);
1126 if (r < 0)
1127 return log_error_errno(r, "Failed to serialize state: %m");
1128
1129 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1130 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1131
1132 r = fd_cloexec(fileno(f), false);
1133 if (r < 0)
1134 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1135
1136 r = fdset_cloexec(fds, false);
1137 if (r < 0)
1138 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1139
1140 *_f = TAKE_PTR(f);
1141 *_fds = TAKE_PTR(fds);
1142
1143 return 0;
1144 }
1145
1146 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1147 int r, nr;
1148
1149 assert(saved_rlimit);
1150
1151 /* Save the original RLIMIT_NOFILE so that we can reset it
1152 * later when transitioning from the initrd to the main
1153 * systemd or suchlike. */
1154 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1155 return log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1156
1157 /* Make sure forked processes get the default kernel setting */
1158 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1159 struct rlimit *rl;
1160
1161 rl = newdup(struct rlimit, saved_rlimit, 1);
1162 if (!rl)
1163 return log_oom();
1164
1165 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1166 }
1167
1168 /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows */
1169 nr = read_nr_open();
1170 r = setrlimit_closest(RLIMIT_NOFILE, &RLIMIT_MAKE_CONST(nr));
1171 if (r < 0)
1172 return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1173
1174 return 0;
1175 }
1176
1177 static int bump_rlimit_memlock(struct rlimit *saved_rlimit) {
1178 int r;
1179
1180 assert(saved_rlimit);
1181 assert(getuid() == 0);
1182
1183 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even though we have CAP_IPC_LOCK which
1184 * should normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's
1185 * bump the value high enough for the root user. */
1186
1187 if (getrlimit(RLIMIT_MEMLOCK, saved_rlimit) < 0)
1188 return log_warning_errno(errno, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1189
1190 r = setrlimit_closest(RLIMIT_MEMLOCK, &RLIMIT_MAKE_CONST(1024ULL*1024ULL*16ULL));
1191 if (r < 0)
1192 return log_warning_errno(r, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1193
1194 return 0;
1195 }
1196
1197 static void test_usr(void) {
1198
1199 /* Check that /usr is not a separate fs */
1200
1201 if (dir_is_empty("/usr") <= 0)
1202 return;
1203
1204 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1205 "Some things will probably break (sometimes even silently) in mysterious ways. "
1206 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1207 }
1208
1209 static int enforce_syscall_archs(Set *archs) {
1210 #if HAVE_SECCOMP
1211 int r;
1212
1213 if (!is_seccomp_available())
1214 return 0;
1215
1216 r = seccomp_restrict_archs(arg_syscall_archs);
1217 if (r < 0)
1218 return log_error_errno(r, "Failed to enforce system call architecture restrication: %m");
1219 #endif
1220 return 0;
1221 }
1222
1223 static int status_welcome(void) {
1224 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1225 int r;
1226
1227 if (arg_show_status <= 0)
1228 return 0;
1229
1230 r = parse_os_release(NULL,
1231 "PRETTY_NAME", &pretty_name,
1232 "ANSI_COLOR", &ansi_color,
1233 NULL);
1234 if (r < 0)
1235 log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r,
1236 "Failed to read os-release file, ignoring: %m");
1237
1238 if (log_get_show_color())
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 else
1244 return status_printf(NULL, false, false,
1245 "\nWelcome to %s!\n",
1246 isempty(pretty_name) ? "Linux" : pretty_name);
1247 }
1248
1249 static int write_container_id(void) {
1250 const char *c;
1251 int r;
1252
1253 c = getenv("container");
1254 if (isempty(c))
1255 return 0;
1256
1257 RUN_WITH_UMASK(0022)
1258 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1259 if (r < 0)
1260 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
1261
1262 return 1;
1263 }
1264
1265 static int bump_unix_max_dgram_qlen(void) {
1266 _cleanup_free_ char *qlen = NULL;
1267 unsigned long v;
1268 int r;
1269
1270 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1271 * really really early during boot, so that it is actually applied to all our sockets, including the
1272 * $NOTIFY_SOCKET one. */
1273
1274 r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1275 if (r < 0)
1276 return log_warning_errno(r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1277
1278 r = safe_atolu(qlen, &v);
1279 if (r < 0)
1280 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen);
1281
1282 if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1283 return 0;
1284
1285 r = write_string_filef("/proc/sys/net/unix/max_dgram_qlen", 0, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN);
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 static int fixup_environment(void) {
1294 _cleanup_free_ char *term = NULL;
1295 const char *t;
1296 int r;
1297
1298 /* Only fix up the environment when we are started as PID 1 */
1299 if (getpid_cached() != 1)
1300 return 0;
1301
1302 /* We expect the environment to be set correctly if run inside a container. */
1303 if (detect_container() > 0)
1304 return 0;
1305
1306 /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1307 * device used by the console. We try to make a better guess here since some consoles might not have support
1308 * for color mode for example.
1309 *
1310 * However if TERM was configured through the kernel command line then leave it alone. */
1311 r = proc_cmdline_get_key("TERM", 0, &term);
1312 if (r < 0)
1313 return r;
1314
1315 t = term ?: default_term_for_tty("/dev/console");
1316
1317 if (setenv("TERM", t, 1) < 0)
1318 return -errno;
1319
1320 return 0;
1321 }
1322
1323 static void redirect_telinit(int argc, char *argv[]) {
1324
1325 /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1326
1327 #if HAVE_SYSV_COMPAT
1328 if (getpid_cached() == 1)
1329 return;
1330
1331 if (!strstr(program_invocation_short_name, "init"))
1332 return;
1333
1334 execv(SYSTEMCTL_BINARY_PATH, argv);
1335 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1336 exit(EXIT_FAILURE);
1337 #endif
1338 }
1339
1340 static int become_shutdown(
1341 const char *shutdown_verb,
1342 int retval) {
1343
1344 char log_level[DECIMAL_STR_MAX(int) + 1],
1345 exit_code[DECIMAL_STR_MAX(uint8_t) + 1],
1346 timeout[DECIMAL_STR_MAX(usec_t) + 1];
1347
1348 const char* command_line[13] = {
1349 SYSTEMD_SHUTDOWN_BINARY_PATH,
1350 shutdown_verb,
1351 "--timeout", timeout,
1352 "--log-level", log_level,
1353 "--log-target",
1354 };
1355
1356 _cleanup_strv_free_ char **env_block = NULL;
1357 size_t pos = 7;
1358 int r;
1359
1360 assert(shutdown_verb);
1361 assert(!command_line[pos]);
1362 env_block = strv_copy(environ);
1363
1364 xsprintf(log_level, "%d", log_get_max_level());
1365 xsprintf(timeout, "%" PRI_USEC "us", arg_default_timeout_stop_usec);
1366
1367 switch (log_get_target()) {
1368
1369 case LOG_TARGET_KMSG:
1370 case LOG_TARGET_JOURNAL_OR_KMSG:
1371 case LOG_TARGET_SYSLOG_OR_KMSG:
1372 command_line[pos++] = "kmsg";
1373 break;
1374
1375 case LOG_TARGET_NULL:
1376 command_line[pos++] = "null";
1377 break;
1378
1379 case LOG_TARGET_CONSOLE:
1380 default:
1381 command_line[pos++] = "console";
1382 break;
1383 };
1384
1385 if (log_get_show_color())
1386 command_line[pos++] = "--log-color";
1387
1388 if (log_get_show_location())
1389 command_line[pos++] = "--log-location";
1390
1391 if (streq(shutdown_verb, "exit")) {
1392 command_line[pos++] = "--exit-code";
1393 command_line[pos++] = exit_code;
1394 xsprintf(exit_code, "%d", retval);
1395 }
1396
1397 assert(pos < ELEMENTSOF(command_line));
1398
1399 if (streq(shutdown_verb, "reboot") &&
1400 arg_shutdown_watchdog > 0 &&
1401 arg_shutdown_watchdog != USEC_INFINITY) {
1402
1403 char *e;
1404
1405 /* If we reboot let's set the shutdown
1406 * watchdog and tell the shutdown binary to
1407 * repeatedly ping it */
1408 r = watchdog_set_timeout(&arg_shutdown_watchdog);
1409 watchdog_close(r < 0);
1410
1411 /* Tell the binary how often to ping, ignore failure */
1412 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
1413 (void) strv_consume(&env_block, e);
1414
1415 if (arg_watchdog_device &&
1416 asprintf(&e, "WATCHDOG_DEVICE=%s", arg_watchdog_device) > 0)
1417 (void) strv_consume(&env_block, e);
1418 } else
1419 watchdog_close(true);
1420
1421 /* Avoid the creation of new processes forked by the
1422 * kernel; at this point, we will not listen to the
1423 * signals anyway */
1424 if (detect_container() <= 0)
1425 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
1426
1427 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
1428 return -errno;
1429 }
1430
1431 static void initialize_clock(void) {
1432 int r;
1433
1434 if (clock_is_localtime(NULL) > 0) {
1435 int min;
1436
1437 /*
1438 * The very first call of settimeofday() also does a time warp in the kernel.
1439 *
1440 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1441 * of maintaining the RTC and do all adjustments. This matches the behavior of Windows, which leaves
1442 * the RTC alone if the registry tells that the RTC runs in UTC.
1443 */
1444 r = clock_set_timezone(&min);
1445 if (r < 0)
1446 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1447 else
1448 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1449
1450 } else if (!in_initrd()) {
1451 /*
1452 * Do a dummy very first call to seal the kernel's time warp magic.
1453 *
1454 * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1455 * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1456 * until we reach the real system.
1457 *
1458 * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1459 * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1460 * be treated as UTC that way.
1461 */
1462 (void) clock_reset_timewarp();
1463 }
1464
1465 r = clock_apply_epoch();
1466 if (r < 0)
1467 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1468 else if (r > 0)
1469 log_info("System time before build time, advancing clock.");
1470 }
1471
1472 static void initialize_coredump(bool skip_setup) {
1473 #if ENABLE_COREDUMP
1474 if (getpid_cached() != 1)
1475 return;
1476
1477 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1478 * will process core dumps for system services by default. */
1479 if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0)
1480 log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m");
1481
1482 /* But at the same time, turn off the core_pattern logic by default, so that no coredumps are stored
1483 * until the systemd-coredump tool is enabled via sysctl. */
1484 if (!skip_setup)
1485 disable_coredumps();
1486 #endif
1487 }
1488
1489 static void do_reexecute(
1490 int argc,
1491 char *argv[],
1492 const struct rlimit *saved_rlimit_nofile,
1493 const struct rlimit *saved_rlimit_memlock,
1494 FDSet *fds,
1495 const char *switch_root_dir,
1496 const char *switch_root_init,
1497 const char **ret_error_message) {
1498
1499 unsigned i, j, args_size;
1500 const char **args;
1501 int r;
1502
1503 assert(saved_rlimit_nofile);
1504 assert(saved_rlimit_memlock);
1505 assert(ret_error_message);
1506
1507 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1508 * we do that */
1509 watchdog_close(true);
1510
1511 /* Reset the RLIMIT_NOFILE to the kernel default, so that the new systemd can pass the kernel default to its
1512 * child processes */
1513
1514 if (saved_rlimit_nofile->rlim_cur > 0)
1515 (void) setrlimit(RLIMIT_NOFILE, saved_rlimit_nofile);
1516 if (saved_rlimit_memlock->rlim_cur != (rlim_t) -1)
1517 (void) setrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock);
1518
1519 if (switch_root_dir) {
1520 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1521 * SIGCHLD for them after deserializing. */
1522 broadcast_signal(SIGTERM, false, true, arg_default_timeout_stop_usec);
1523
1524 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1525 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1526 if (r < 0)
1527 log_error_errno(r, "Failed to switch root, trying to continue: %m");
1528 }
1529
1530 args_size = MAX(6, argc+1);
1531 args = newa(const char*, args_size);
1532
1533 if (!switch_root_init) {
1534 char sfd[DECIMAL_STR_MAX(int) + 1];
1535
1536 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1537 * the user didn't specify an explicit init to spawn. */
1538
1539 assert(arg_serialization);
1540 assert(fds);
1541
1542 xsprintf(sfd, "%i", fileno(arg_serialization));
1543
1544 i = 0;
1545 args[i++] = SYSTEMD_BINARY_PATH;
1546 if (switch_root_dir)
1547 args[i++] = "--switched-root";
1548 args[i++] = arg_system ? "--system" : "--user";
1549 args[i++] = "--deserialize";
1550 args[i++] = sfd;
1551 args[i++] = NULL;
1552
1553 assert(i <= args_size);
1554
1555 /*
1556 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do this is on
1557 * its own on exec(), but it will do it on exit(). Hence, to ensure we get a summary here, fork() off
1558 * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1559 * before proceeding into the exec().
1560 */
1561 valgrind_summary_hack();
1562
1563 (void) execv(args[0], (char* const*) args);
1564 log_debug_errno(errno, "Failed to execute our own binary, trying fallback: %m");
1565 }
1566
1567 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1568 * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1569 * doesn't matter.) */
1570
1571 arg_serialization = safe_fclose(arg_serialization);
1572 fds = fdset_free(fds);
1573
1574 /* Reopen the console */
1575 (void) make_console_stdio();
1576
1577 for (j = 1, i = 1; j < (unsigned) argc; j++)
1578 args[i++] = argv[j];
1579 args[i++] = NULL;
1580 assert(i <= args_size);
1581
1582 /* Reenable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1583 (void) reset_all_signal_handlers();
1584 (void) reset_signal_mask();
1585
1586 if (switch_root_init) {
1587 args[0] = switch_root_init;
1588 (void) execv(args[0], (char* const*) args);
1589 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
1590 }
1591
1592 args[0] = "/sbin/init";
1593 (void) execv(args[0], (char* const*) args);
1594 r = -errno;
1595
1596 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
1597 ANSI_HIGHLIGHT_RED " !! " ANSI_NORMAL,
1598 "Failed to execute /sbin/init");
1599
1600 if (r == -ENOENT) {
1601 log_warning("No /sbin/init, trying fallback");
1602
1603 args[0] = "/bin/sh";
1604 args[1] = NULL;
1605 (void) execv(args[0], (char* const*) args);
1606 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
1607 } else
1608 log_warning_errno(r, "Failed to execute /sbin/init, giving up: %m");
1609
1610 *ret_error_message = "Failed to execute fallback shell";
1611 }
1612
1613 static int invoke_main_loop(
1614 Manager *m,
1615 bool *ret_reexecute,
1616 int *ret_retval, /* Return parameters relevant for shutting down */
1617 const char **ret_shutdown_verb, /* … */
1618 FDSet **ret_fds, /* Return parameters for reexecuting */
1619 char **ret_switch_root_dir, /* … */
1620 char **ret_switch_root_init, /* … */
1621 const char **ret_error_message) {
1622
1623 int r;
1624
1625 assert(m);
1626 assert(ret_reexecute);
1627 assert(ret_retval);
1628 assert(ret_shutdown_verb);
1629 assert(ret_fds);
1630 assert(ret_switch_root_dir);
1631 assert(ret_switch_root_init);
1632 assert(ret_error_message);
1633
1634 for (;;) {
1635 r = manager_loop(m);
1636 if (r < 0) {
1637 *ret_error_message = "Failed to run main loop";
1638 return log_emergency_errno(r, "Failed to run main loop: %m");
1639 }
1640
1641 switch (m->exit_code) {
1642
1643 case MANAGER_RELOAD:
1644 log_info("Reloading.");
1645
1646 r = parse_config_file();
1647 if (r < 0)
1648 log_warning_errno(r, "Failed to parse config file, ignoring: %m");
1649
1650 set_manager_defaults(m);
1651
1652 r = manager_reload(m);
1653 if (r < 0)
1654 log_warning_errno(r, "Failed to reload, ignoring: %m");
1655
1656 break;
1657
1658 case MANAGER_REEXECUTE:
1659
1660 r = prepare_reexecute(m, &arg_serialization, ret_fds, false);
1661 if (r < 0) {
1662 *ret_error_message = "Failed to prepare for reexecution";
1663 return r;
1664 }
1665
1666 log_notice("Reexecuting.");
1667
1668 *ret_reexecute = true;
1669 *ret_retval = EXIT_SUCCESS;
1670 *ret_shutdown_verb = NULL;
1671 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1672
1673 return 0;
1674
1675 case MANAGER_SWITCH_ROOT:
1676 if (!m->switch_root_init) {
1677 r = prepare_reexecute(m, &arg_serialization, ret_fds, true);
1678 if (r < 0) {
1679 *ret_error_message = "Failed to prepare for reexecution";
1680 return r;
1681 }
1682 } else
1683 *ret_fds = NULL;
1684
1685 log_notice("Switching root.");
1686
1687 *ret_reexecute = true;
1688 *ret_retval = EXIT_SUCCESS;
1689 *ret_shutdown_verb = NULL;
1690
1691 /* Steal the switch root parameters */
1692 *ret_switch_root_dir = m->switch_root;
1693 *ret_switch_root_init = m->switch_root_init;
1694 m->switch_root = m->switch_root_init = NULL;
1695
1696 return 0;
1697
1698 case MANAGER_EXIT:
1699
1700 if (MANAGER_IS_USER(m)) {
1701 log_debug("Exit.");
1702
1703 *ret_reexecute = false;
1704 *ret_retval = m->return_value;
1705 *ret_shutdown_verb = NULL;
1706 *ret_fds = NULL;
1707 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1708
1709 return 0;
1710 }
1711
1712 _fallthrough_;
1713 case MANAGER_REBOOT:
1714 case MANAGER_POWEROFF:
1715 case MANAGER_HALT:
1716 case MANAGER_KEXEC: {
1717 static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
1718 [MANAGER_EXIT] = "exit",
1719 [MANAGER_REBOOT] = "reboot",
1720 [MANAGER_POWEROFF] = "poweroff",
1721 [MANAGER_HALT] = "halt",
1722 [MANAGER_KEXEC] = "kexec"
1723 };
1724
1725 log_notice("Shutting down.");
1726
1727 *ret_reexecute = false;
1728 *ret_retval = m->return_value;
1729 assert_se(*ret_shutdown_verb = table[m->exit_code]);
1730 *ret_fds = NULL;
1731 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1732
1733 return 0;
1734 }
1735
1736 default:
1737 assert_not_reached("Unknown exit code.");
1738 }
1739 }
1740 }
1741
1742 static void log_execution_mode(bool *ret_first_boot) {
1743 assert(ret_first_boot);
1744
1745 if (arg_system) {
1746 int v;
1747
1748 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1749 arg_action == ACTION_TEST ? "test " : "" );
1750
1751 v = detect_virtualization();
1752 if (v > 0)
1753 log_info("Detected virtualization %s.", virtualization_to_string(v));
1754
1755 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1756
1757 if (in_initrd()) {
1758 *ret_first_boot = false;
1759 log_info("Running in initial RAM disk.");
1760 } else {
1761 /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
1762 * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
1763 * doesn't it's unpopulated. This allows container managers and installers to provision a
1764 * couple of files already. If the container manager wants to provision the machine ID itself
1765 * it should pass $container_uuid to PID 1. */
1766
1767 *ret_first_boot = access("/etc/machine-id", F_OK) < 0;
1768 if (*ret_first_boot)
1769 log_info("Running with unpopulated /etc.");
1770 }
1771 } else {
1772 if (DEBUG_LOGGING) {
1773 _cleanup_free_ char *t;
1774
1775 t = uid_to_name(getuid());
1776 log_debug(PACKAGE_STRING " running in %suser mode for user " UID_FMT "/%s. (" SYSTEMD_FEATURES ")",
1777 arg_action == ACTION_TEST ? " test" : "", getuid(), strna(t));
1778 }
1779
1780 *ret_first_boot = false;
1781 }
1782 }
1783
1784 static int initialize_runtime(
1785 bool skip_setup,
1786 struct rlimit *saved_rlimit_nofile,
1787 struct rlimit *saved_rlimit_memlock,
1788 const char **ret_error_message) {
1789
1790 int r;
1791
1792 assert(ret_error_message);
1793
1794 /* Sets up various runtime parameters. Many of these initializations are conditionalized:
1795 *
1796 * - Some only apply to --system instances
1797 * - Some only apply to --user instances
1798 * - Some only apply when we first start up, but not when we reexecute
1799 */
1800
1801 if (arg_action != ACTION_RUN)
1802 return 0;
1803
1804 if (arg_system) {
1805 /* Make sure we leave a core dump without panicing the kernel. */
1806 install_crash_handler();
1807
1808 if (!skip_setup) {
1809 r = mount_cgroup_controllers(arg_join_controllers);
1810 if (r < 0) {
1811 *ret_error_message = "Failed to mount cgroup hierarchies";
1812 return r;
1813 }
1814
1815 status_welcome();
1816 hostname_setup();
1817 machine_id_setup(NULL, arg_machine_id, NULL);
1818 loopback_setup();
1819 bump_unix_max_dgram_qlen();
1820 test_usr();
1821 write_container_id();
1822 }
1823
1824 if (arg_watchdog_device) {
1825 r = watchdog_set_device(arg_watchdog_device);
1826 if (r < 0)
1827 log_warning_errno(r, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device);
1828 }
1829
1830 if (arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY)
1831 watchdog_set_timeout(&arg_runtime_watchdog);
1832 }
1833
1834 if (arg_timer_slack_nsec != NSEC_INFINITY)
1835 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1836 log_warning_errno(errno, "Failed to adjust timer slack, ignoring: %m");
1837
1838 if (arg_system && !cap_test_all(arg_capability_bounding_set)) {
1839 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
1840 if (r < 0) {
1841 *ret_error_message = "Failed to drop capability bounding set of usermode helpers";
1842 return log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
1843 }
1844
1845 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
1846 if (r < 0) {
1847 *ret_error_message = "Failed to drop capability bounding set";
1848 return log_emergency_errno(r, "Failed to drop capability bounding set: %m");
1849 }
1850 }
1851
1852 if (arg_system && arg_no_new_privs) {
1853 if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) {
1854 *ret_error_message = "Failed to disable new privileges";
1855 return log_emergency_errno(errno, "Failed to disable new privileges: %m");
1856 }
1857 }
1858
1859 if (arg_syscall_archs) {
1860 r = enforce_syscall_archs(arg_syscall_archs);
1861 if (r < 0) {
1862 *ret_error_message = "Failed to set syscall architectures";
1863 return r;
1864 }
1865 }
1866
1867 if (!arg_system)
1868 /* Become reaper of our children */
1869 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
1870 log_warning_errno(errno, "Failed to make us a subreaper: %m");
1871
1872 if (arg_system) {
1873 /* Bump up RLIMIT_NOFILE for systemd itself */
1874 (void) bump_rlimit_nofile(saved_rlimit_nofile);
1875 (void) bump_rlimit_memlock(saved_rlimit_memlock);
1876 }
1877
1878 return 0;
1879 }
1880
1881 static int do_queue_default_job(
1882 Manager *m,
1883 const char **ret_error_message) {
1884
1885 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
1886 Job *default_unit_job;
1887 Unit *target = NULL;
1888 int r;
1889
1890 log_debug("Activating default unit: %s", arg_default_unit);
1891
1892 r = manager_load_startable_unit_or_warn(m, arg_default_unit, NULL, &target);
1893 if (r < 0) {
1894 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET);
1895
1896 r = manager_load_startable_unit_or_warn(m, SPECIAL_RESCUE_TARGET, NULL, &target);
1897 if (r < 0) {
1898 *ret_error_message = r == -ERFKILL ? "Rescue target masked"
1899 : "Failed to load rescue target";
1900 return r;
1901 }
1902 }
1903
1904 assert(target->load_state == UNIT_LOADED);
1905
1906 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, &error, &default_unit_job);
1907 if (r == -EPERM) {
1908 log_debug_errno(r, "Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1909
1910 sd_bus_error_free(&error);
1911
1912 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, &error, &default_unit_job);
1913 if (r < 0) {
1914 *ret_error_message = "Failed to start default target";
1915 return log_emergency_errno(r, "Failed to start default target: %s", bus_error_message(&error, r));
1916 }
1917
1918 } else if (r < 0) {
1919 *ret_error_message = "Failed to isolate default target";
1920 return log_emergency_errno(r, "Failed to isolate default target: %s", bus_error_message(&error, r));
1921 }
1922
1923 m->default_unit_job_id = default_unit_job->id;
1924
1925 return 0;
1926 }
1927
1928 static void free_arguments(void) {
1929
1930 /* Frees all arg_* variables, with the exception of arg_serialization */
1931 rlimit_free_all(arg_default_rlimit);
1932
1933 arg_default_unit = mfree(arg_default_unit);
1934 arg_confirm_spawn = mfree(arg_confirm_spawn);
1935 arg_join_controllers = strv_free_free(arg_join_controllers);
1936 arg_default_environment = strv_free(arg_default_environment);
1937 arg_syscall_archs = set_free(arg_syscall_archs);
1938 }
1939
1940 static int load_configuration(int argc, char **argv, const char **ret_error_message) {
1941 int r;
1942
1943 assert(ret_error_message);
1944
1945 arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U);
1946
1947 r = parse_config_file();
1948 if (r < 0) {
1949 *ret_error_message = "Failed to parse config file";
1950 return r;
1951 }
1952
1953 if (arg_system) {
1954 r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
1955 if (r < 0)
1956 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
1957 }
1958
1959 /* Note that this also parses bits from the kernel command line, including "debug". */
1960 log_parse_environment();
1961
1962 r = parse_argv(argc, argv);
1963 if (r < 0) {
1964 *ret_error_message = "Failed to parse commandline arguments";
1965 return r;
1966 }
1967
1968 /* Initialize default unit */
1969 if (!arg_default_unit) {
1970 arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET);
1971 if (!arg_default_unit) {
1972 *ret_error_message = "Failed to set default unit";
1973 return log_oom();
1974 }
1975 }
1976
1977 /* Initialize the show status setting if it hasn't been set explicitly yet */
1978 if (arg_show_status == _SHOW_STATUS_UNSET)
1979 arg_show_status = SHOW_STATUS_YES;
1980
1981 return 0;
1982 }
1983
1984 static int safety_checks(void) {
1985
1986 if (getpid_cached() == 1 &&
1987 arg_action != ACTION_RUN) {
1988 log_error("Unsupported execution mode while PID 1.");
1989 return -EPERM;
1990 }
1991
1992 if (getpid_cached() == 1 &&
1993 !arg_system) {
1994 log_error("Can't run --user mode as PID 1.");
1995 return -EPERM;
1996 }
1997
1998 if (arg_action == ACTION_RUN &&
1999 arg_system &&
2000 getpid_cached() != 1) {
2001 log_error("Can't run system mode unless PID 1.");
2002 return -EPERM;
2003 }
2004
2005 if (arg_action == ACTION_TEST &&
2006 geteuid() == 0) {
2007 log_error("Don't run test mode as root.");
2008 return -EPERM;
2009 }
2010
2011 if (!arg_system &&
2012 arg_action == ACTION_RUN &&
2013 sd_booted() <= 0) {
2014 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
2015 return -EOPNOTSUPP;
2016 }
2017
2018 if (!arg_system &&
2019 arg_action == ACTION_RUN &&
2020 !getenv("XDG_RUNTIME_DIR")) {
2021 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
2022 return -EUNATCH;
2023 }
2024
2025 if (arg_system &&
2026 arg_action == ACTION_RUN &&
2027 running_in_chroot() > 0) {
2028 log_error("Cannot be run in a chroot() environment.");
2029 return -EOPNOTSUPP;
2030 }
2031
2032 return 0;
2033 }
2034
2035 static int initialize_security(
2036 bool *loaded_policy,
2037 dual_timestamp *security_start_timestamp,
2038 dual_timestamp *security_finish_timestamp,
2039 const char **ret_error_message) {
2040
2041 int r;
2042
2043 assert(loaded_policy);
2044 assert(security_start_timestamp);
2045 assert(security_finish_timestamp);
2046 assert(ret_error_message);
2047
2048 dual_timestamp_get(security_start_timestamp);
2049
2050 r = mac_selinux_setup(loaded_policy);
2051 if (r < 0) {
2052 *ret_error_message = "Failed to load SELinux policy";
2053 return r;
2054 }
2055
2056 r = mac_smack_setup(loaded_policy);
2057 if (r < 0) {
2058 *ret_error_message = "Failed to load SMACK policy";
2059 return r;
2060 }
2061
2062 r = ima_setup();
2063 if (r < 0) {
2064 *ret_error_message = "Failed to load IMA policy";
2065 return r;
2066 }
2067
2068 dual_timestamp_get(security_finish_timestamp);
2069 return 0;
2070 }
2071
2072 static void test_summary(Manager *m) {
2073 assert(m);
2074
2075 printf("-> By units:\n");
2076 manager_dump_units(m, stdout, "\t");
2077
2078 printf("-> By jobs:\n");
2079 manager_dump_jobs(m, stdout, "\t");
2080 }
2081
2082 static int collect_fds(FDSet **ret_fds, const char **ret_error_message) {
2083 int r;
2084
2085 assert(ret_fds);
2086 assert(ret_error_message);
2087
2088 r = fdset_new_fill(ret_fds);
2089 if (r < 0) {
2090 *ret_error_message = "Failed to allocate fd set";
2091 return log_emergency_errno(r, "Failed to allocate fd set: %m");
2092 }
2093
2094 fdset_cloexec(*ret_fds, true);
2095
2096 if (arg_serialization)
2097 assert_se(fdset_remove(*ret_fds, fileno(arg_serialization)) >= 0);
2098
2099 return 0;
2100 }
2101
2102 static void setup_console_terminal(bool skip_setup) {
2103
2104 if (!arg_system)
2105 return;
2106
2107 /* Become a session leader if we aren't one yet. */
2108 (void) setsid();
2109
2110 /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2111 * tty. */
2112 (void) release_terminal();
2113
2114 /* Reset the console, but only if this is really init and we are freshly booted */
2115 if (getpid_cached() == 1 && !skip_setup)
2116 (void) console_setup();
2117 }
2118
2119 static bool early_skip_setup_check(int argc, char *argv[]) {
2120 bool found_deserialize = false;
2121 int i;
2122
2123 /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2124 * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2125 * anyway, even if in that case we also do deserialization. */
2126
2127 for (i = 1; i < argc; i++) {
2128 if (streq(argv[i], "--switched-root"))
2129 return false; /* If we switched root, don't skip the setup. */
2130 else if (streq(argv[i], "--deserialize"))
2131 found_deserialize = true;
2132 }
2133
2134 return found_deserialize; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2135 }
2136
2137 int main(int argc, char *argv[]) {
2138
2139 dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL, userspace_timestamp = DUAL_TIMESTAMP_NULL, kernel_timestamp = DUAL_TIMESTAMP_NULL,
2140 security_start_timestamp = DUAL_TIMESTAMP_NULL, security_finish_timestamp = DUAL_TIMESTAMP_NULL;
2141 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0), saved_rlimit_memlock = RLIMIT_MAKE_CONST((rlim_t) -1);
2142 bool skip_setup, loaded_policy = false, queue_default_job = false, first_boot = false, reexecute = false;
2143 char *switch_root_dir = NULL, *switch_root_init = NULL;
2144 usec_t before_startup, after_startup;
2145 static char systemd[] = "systemd";
2146 char timespan[FORMAT_TIMESPAN_MAX];
2147 const char *shutdown_verb = NULL, *error_message = NULL;
2148 int r, retval = EXIT_FAILURE;
2149 Manager *m = NULL;
2150 FDSet *fds = NULL;
2151
2152 /* SysV compatibility: redirect init → telinit */
2153 redirect_telinit(argc, argv);
2154
2155 /* Take timestamps early on */
2156 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
2157 dual_timestamp_get(&userspace_timestamp);
2158
2159 /* Figure out whether we need to do initialize the system, or if we already did that because we are
2160 * reexecuting */
2161 skip_setup = early_skip_setup_check(argc, argv);
2162
2163 /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2164 * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
2165 program_invocation_short_name = systemd;
2166 (void) prctl(PR_SET_NAME, systemd);
2167
2168 /* Save the original command line */
2169 saved_argv = argv;
2170 saved_argc = argc;
2171
2172 /* Make sure that if the user says "syslog" we actually log to the journal. */
2173 log_set_upgrade_syslog_to_journal(true);
2174
2175 if (getpid_cached() == 1) {
2176 /* Disable the umask logic */
2177 umask(0);
2178
2179 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2180 * activated yet (even though the log socket for it exists). */
2181 log_set_prohibit_ipc(true);
2182
2183 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2184 * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2185 * child process right before execve()'ing the actual binary, at a point in time where socket
2186 * activation stderr/stdout area already set up. */
2187 log_set_always_reopen_console(true);
2188
2189 if (detect_container() <= 0) {
2190
2191 /* Running outside of a container as PID 1 */
2192 arg_system = true;
2193 log_set_target(LOG_TARGET_KMSG);
2194 log_open();
2195
2196 if (in_initrd())
2197 initrd_timestamp = userspace_timestamp;
2198
2199 if (!skip_setup) {
2200 r = mount_setup_early();
2201 if (r < 0) {
2202 error_message = "Failed to mount early API filesystems";
2203 goto finish;
2204 }
2205
2206 r = initialize_security(
2207 &loaded_policy,
2208 &security_start_timestamp,
2209 &security_finish_timestamp,
2210 &error_message);
2211 if (r < 0)
2212 goto finish;
2213 }
2214
2215 if (mac_selinux_init() < 0) {
2216 error_message = "Failed to initialize SELinux policy";
2217 goto finish;
2218 }
2219
2220 if (!skip_setup)
2221 initialize_clock();
2222
2223 /* Set the default for later on, but don't actually open the logs like this for now. Note that
2224 * if we are transitioning from the initrd there might still be journal fd open, and we
2225 * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2226 * elsewhere. */
2227 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
2228
2229 } else {
2230 /* Running inside a container, as PID 1 */
2231 arg_system = true;
2232 log_set_target(LOG_TARGET_CONSOLE);
2233 log_open();
2234
2235 /* For later on, see above... */
2236 log_set_target(LOG_TARGET_JOURNAL);
2237
2238 /* clear the kernel timestamp,
2239 * because we are in a container */
2240 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2241 }
2242
2243 initialize_coredump(skip_setup);
2244
2245 r = fixup_environment();
2246 if (r < 0) {
2247 log_emergency_errno(r, "Failed to fix up PID 1 environment: %m");
2248 error_message = "Failed to fix up PID1 environment";
2249 goto finish;
2250 }
2251
2252 } else {
2253 /* Running as user instance */
2254 arg_system = false;
2255 log_set_target(LOG_TARGET_AUTO);
2256 log_open();
2257
2258 /* clear the kernel timestamp,
2259 * because we are not PID 1 */
2260 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2261 }
2262
2263 if (arg_system) {
2264 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2265 * they never log into the console. */
2266 log_show_color(colors_enabled());
2267
2268 r = make_null_stdio();
2269 if (r < 0)
2270 log_warning_errno(r, "Failed to redirect standard streams to /dev/null, ignoring: %m");
2271 }
2272
2273 /* Mount /proc, /sys and friends, so that /proc/cmdline and
2274 * /proc/$PID/fd is available. */
2275 if (getpid_cached() == 1) {
2276
2277 /* Load the kernel modules early. */
2278 if (!skip_setup)
2279 kmod_setup();
2280
2281 r = mount_setup(loaded_policy);
2282 if (r < 0) {
2283 error_message = "Failed to mount API filesystems";
2284 goto finish;
2285 }
2286 }
2287
2288 /* Reset all signal handlers. */
2289 (void) reset_all_signal_handlers();
2290 (void) ignore_signals(SIGNALS_IGNORE, -1);
2291
2292 r = load_configuration(argc, argv, &error_message);
2293 if (r < 0)
2294 goto finish;
2295
2296 r = safety_checks();
2297 if (r < 0)
2298 goto finish;
2299
2300 if (IN_SET(arg_action, ACTION_TEST, ACTION_HELP, ACTION_DUMP_CONFIGURATION_ITEMS, ACTION_DUMP_BUS_PROPERTIES))
2301 (void) pager_open(arg_no_pager, false);
2302
2303 if (arg_action != ACTION_RUN)
2304 skip_setup = true;
2305
2306 if (arg_action == ACTION_HELP) {
2307 retval = help();
2308 goto finish;
2309 } else if (arg_action == ACTION_VERSION) {
2310 retval = version();
2311 goto finish;
2312 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
2313 unit_dump_config_items(stdout);
2314 retval = EXIT_SUCCESS;
2315 goto finish;
2316 } else if (arg_action == ACTION_DUMP_BUS_PROPERTIES) {
2317 dump_bus_properties(stdout);
2318 retval = EXIT_SUCCESS;
2319 goto finish;
2320 }
2321
2322 assert_se(IN_SET(arg_action, ACTION_RUN, ACTION_TEST));
2323
2324 /* Move out of the way, so that we won't block unmounts */
2325 assert_se(chdir("/") == 0);
2326
2327 if (arg_action == ACTION_RUN) {
2328
2329 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2330 log_close();
2331
2332 /* Remember open file descriptors for later deserialization */
2333 r = collect_fds(&fds, &error_message);
2334 if (r < 0)
2335 goto finish;
2336
2337 /* Give up any control of the console, but make sure its initialized. */
2338 setup_console_terminal(skip_setup);
2339
2340 /* Open the logging devices, if possible and necessary */
2341 log_open();
2342 }
2343
2344 log_execution_mode(&first_boot);
2345
2346 r = initialize_runtime(skip_setup,
2347 &saved_rlimit_nofile,
2348 &saved_rlimit_memlock,
2349 &error_message);
2350 if (r < 0)
2351 goto finish;
2352
2353 r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER,
2354 arg_action == ACTION_TEST ? MANAGER_TEST_FULL : 0,
2355 &m);
2356 if (r < 0) {
2357 log_emergency_errno(r, "Failed to allocate manager object: %m");
2358 error_message = "Failed to allocate manager object";
2359 goto finish;
2360 }
2361
2362 m->timestamps[MANAGER_TIMESTAMP_KERNEL] = kernel_timestamp;
2363 m->timestamps[MANAGER_TIMESTAMP_INITRD] = initrd_timestamp;
2364 m->timestamps[MANAGER_TIMESTAMP_USERSPACE] = userspace_timestamp;
2365 m->timestamps[MANAGER_TIMESTAMP_SECURITY_START] = security_start_timestamp;
2366 m->timestamps[MANAGER_TIMESTAMP_SECURITY_FINISH] = security_finish_timestamp;
2367
2368 set_manager_defaults(m);
2369 set_manager_settings(m);
2370 manager_set_first_boot(m, first_boot);
2371
2372 /* Remember whether we should queue the default job */
2373 queue_default_job = !arg_serialization || arg_switched_root;
2374
2375 before_startup = now(CLOCK_MONOTONIC);
2376
2377 r = manager_startup(m, arg_serialization, fds);
2378 if (r < 0) {
2379 log_error_errno(r, "Failed to fully start up daemon: %m");
2380 error_message = "Failed to start up manager";
2381 goto finish;
2382 }
2383
2384 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2385 fds = fdset_free(fds);
2386 arg_serialization = safe_fclose(arg_serialization);
2387
2388 if (queue_default_job) {
2389 r = do_queue_default_job(m, &error_message);
2390 if (r < 0)
2391 goto finish;
2392 }
2393
2394 after_startup = now(CLOCK_MONOTONIC);
2395
2396 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
2397 "Loaded units and determined initial transaction in %s.",
2398 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
2399
2400 if (arg_action == ACTION_TEST) {
2401 test_summary(m);
2402 retval = EXIT_SUCCESS;
2403 goto finish;
2404 }
2405
2406 (void) invoke_main_loop(m,
2407 &reexecute,
2408 &retval,
2409 &shutdown_verb,
2410 &fds,
2411 &switch_root_dir,
2412 &switch_root_init,
2413 &error_message);
2414
2415 finish:
2416 pager_close();
2417
2418 if (m) {
2419 arg_shutdown_watchdog = m->shutdown_watchdog;
2420 m = manager_free(m);
2421 }
2422
2423 free_arguments();
2424 mac_selinux_finish();
2425
2426 if (reexecute)
2427 do_reexecute(argc, argv,
2428 &saved_rlimit_nofile,
2429 &saved_rlimit_memlock,
2430 fds,
2431 switch_root_dir,
2432 switch_root_init,
2433 &error_message); /* This only returns if reexecution failed */
2434
2435 arg_serialization = safe_fclose(arg_serialization);
2436 fds = fdset_free(fds);
2437
2438 #if HAVE_VALGRIND_VALGRIND_H
2439 /* If we are PID 1 and running under valgrind, then let's exit
2440 * here explicitly. valgrind will only generate nice output on
2441 * exit(), not on exec(), hence let's do the former not the
2442 * latter here. */
2443 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND) {
2444 /* Cleanup watchdog_device strings for valgrind. We need them
2445 * in become_shutdown() so normally we cannot free them yet. */
2446 watchdog_free_device();
2447 arg_watchdog_device = mfree(arg_watchdog_device);
2448 return retval;
2449 }
2450 #endif
2451
2452 if (shutdown_verb) {
2453 r = become_shutdown(shutdown_verb, retval);
2454 log_error_errno(r, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
2455 error_message = "Failed to execute shutdown binary";
2456 }
2457
2458 watchdog_free_device();
2459 arg_watchdog_device = mfree(arg_watchdog_device);
2460
2461 if (getpid_cached() == 1) {
2462 if (error_message)
2463 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2464 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
2465 "%s, freezing.", error_message);
2466 freeze_or_reboot();
2467 }
2468
2469 return retval;
2470 }
Unnamed repository; edit this file 'description' to name the repository.