]> git.ipfire.org Git - thirdparty/systemd.git/blob - src/core/main.c
Merge pull request #9406 from yuwata/rfe-9228
[thirdparty/systemd.git] / src / core / main.c
1 /* SPDX-License-Identifier: LGPL-2.1+ */
2
3 #include <errno.h>
4 #include <fcntl.h>
5 #include <getopt.h>
6 #include <signal.h>
7 #include <stdio.h>
8 #include <string.h>
9 #include <sys/mount.h>
10 #include <sys/prctl.h>
11 #include <sys/reboot.h>
12 #include <sys/stat.h>
13 #include <unistd.h>
14 #if HAVE_SECCOMP
15 #include <seccomp.h>
16 #endif
17 #if HAVE_VALGRIND_VALGRIND_H
18 #include <valgrind/valgrind.h>
19 #endif
20
21 #include "sd-bus.h"
22 #include "sd-daemon.h"
23 #include "sd-messages.h"
24
25 #include "alloc-util.h"
26 #include "architecture.h"
27 #include "build.h"
28 #include "bus-error.h"
29 #include "bus-util.h"
30 #include "capability-util.h"
31 #include "clock-util.h"
32 #include "conf-parser.h"
33 #include "cpu-set-util.h"
34 #include "dbus.h"
35 #include "dbus-manager.h"
36 #include "def.h"
37 #include "emergency-action.h"
38 #include "env-util.h"
39 #include "exit-status.h"
40 #include "fd-util.h"
41 #include "fdset.h"
42 #include "fileio.h"
43 #include "format-util.h"
44 #include "fs-util.h"
45 #include "hostname-setup.h"
46 #include "ima-setup.h"
47 #include "killall.h"
48 #include "kmod-setup.h"
49 #include "load-fragment.h"
50 #include "log.h"
51 #include "loopback-setup.h"
52 #include "machine-id-setup.h"
53 #include "manager.h"
54 #include "missing.h"
55 #include "mount-setup.h"
56 #include "os-util.h"
57 #include "pager.h"
58 #include "parse-util.h"
59 #include "path-util.h"
60 #include "proc-cmdline.h"
61 #include "process-util.h"
62 #include "raw-clone.h"
63 #include "rlimit-util.h"
64 #if HAVE_SECCOMP
65 #include "seccomp-util.h"
66 #endif
67 #include "selinux-setup.h"
68 #include "selinux-util.h"
69 #include "signal-util.h"
70 #include "smack-setup.h"
71 #include "special.h"
72 #include "stat-util.h"
73 #include "stdio-util.h"
74 #include "strv.h"
75 #include "switch-root.h"
76 #include "terminal-util.h"
77 #include "umask-util.h"
78 #include "user-util.h"
79 #include "util.h"
80 #include "virt.h"
81 #include "watchdog.h"
82
83 static enum {
84 ACTION_RUN,
85 ACTION_HELP,
86 ACTION_VERSION,
87 ACTION_TEST,
88 ACTION_DUMP_CONFIGURATION_ITEMS,
89 ACTION_DUMP_BUS_PROPERTIES,
90 } arg_action = ACTION_RUN;
91 static char *arg_default_unit = NULL;
92 static bool arg_system = false;
93 static bool arg_dump_core = true;
94 static int arg_crash_chvt = -1;
95 static bool arg_crash_shell = false;
96 static bool arg_crash_reboot = false;
97 static char *arg_confirm_spawn = NULL;
98 static ShowStatus arg_show_status = _SHOW_STATUS_INVALID;
99 static bool arg_switched_root = false;
100 static bool arg_no_pager = false;
101 static bool arg_service_watchdogs = true;
102 static char ***arg_join_controllers = NULL;
103 static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
104 static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
105 static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
106 static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
107 static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
108 static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
109 static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
110 static usec_t arg_runtime_watchdog = 0;
111 static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
112 static char *arg_watchdog_device = NULL;
113 static char **arg_default_environment = NULL;
114 static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {};
115 static uint64_t arg_capability_bounding_set = CAP_ALL;
116 static bool arg_no_new_privs = false;
117 static nsec_t arg_timer_slack_nsec = NSEC_INFINITY;
118 static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
119 static Set* arg_syscall_archs = NULL;
120 static FILE* arg_serialization = NULL;
121 static bool arg_default_cpu_accounting = false;
122 static bool arg_default_io_accounting = false;
123 static bool arg_default_ip_accounting = false;
124 static bool arg_default_blockio_accounting = false;
125 static bool arg_default_memory_accounting = MEMORY_ACCOUNTING_DEFAULT;
126 static bool arg_default_tasks_accounting = true;
127 static uint64_t arg_default_tasks_max = UINT64_MAX;
128 static sd_id128_t arg_machine_id = {};
129 static EmergencyAction arg_cad_burst_action = EMERGENCY_ACTION_REBOOT_FORCE;
130
131 _noreturn_ static void freeze_or_reboot(void) {
132
133 if (arg_crash_reboot) {
134 log_notice("Rebooting in 10s...");
135 (void) sleep(10);
136
137 log_notice("Rebooting now...");
138 (void) reboot(RB_AUTOBOOT);
139 log_emergency_errno(errno, "Failed to reboot: %m");
140 }
141
142 log_emergency("Freezing execution.");
143 freeze();
144 }
145
146 _noreturn_ static void crash(int sig) {
147 struct sigaction sa;
148 pid_t pid;
149
150 if (getpid_cached() != 1)
151 /* Pass this on immediately, if this is not PID 1 */
152 (void) raise(sig);
153 else if (!arg_dump_core)
154 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
155 else {
156 sa = (struct sigaction) {
157 .sa_handler = nop_signal_handler,
158 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
159 };
160
161 /* We want to wait for the core process, hence let's enable SIGCHLD */
162 (void) sigaction(SIGCHLD, &sa, NULL);
163
164 pid = raw_clone(SIGCHLD);
165 if (pid < 0)
166 log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
167 else if (pid == 0) {
168 /* Enable default signal handler for core dump */
169
170 sa = (struct sigaction) {
171 .sa_handler = SIG_DFL,
172 };
173 (void) sigaction(sig, &sa, NULL);
174
175 /* Don't limit the coredump size */
176 (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
177
178 /* Just to be sure... */
179 (void) chdir("/");
180
181 /* Raise the signal again */
182 pid = raw_getpid();
183 (void) kill(pid, sig); /* raise() would kill the parent */
184
185 assert_not_reached("We shouldn't be here...");
186 _exit(EXIT_FAILURE);
187 } else {
188 siginfo_t status;
189 int r;
190
191 /* Order things nicely. */
192 r = wait_for_terminate(pid, &status);
193 if (r < 0)
194 log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
195 else if (status.si_code != CLD_DUMPED)
196 log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
197 signal_to_string(sig),
198 pid, sigchld_code_to_string(status.si_code),
199 status.si_status,
200 strna(status.si_code == CLD_EXITED
201 ? exit_status_to_string(status.si_status, EXIT_STATUS_MINIMAL)
202 : signal_to_string(status.si_status)));
203 else
204 log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
205 }
206 }
207
208 if (arg_crash_chvt >= 0)
209 (void) chvt(arg_crash_chvt);
210
211 sa = (struct sigaction) {
212 .sa_handler = SIG_IGN,
213 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
214 };
215
216 /* Let the kernel reap children for us */
217 (void) sigaction(SIGCHLD, &sa, NULL);
218
219 if (arg_crash_shell) {
220 log_notice("Executing crash shell in 10s...");
221 (void) sleep(10);
222
223 pid = raw_clone(SIGCHLD);
224 if (pid < 0)
225 log_emergency_errno(errno, "Failed to fork off crash shell: %m");
226 else if (pid == 0) {
227 (void) setsid();
228 (void) make_console_stdio();
229 (void) execle("/bin/sh", "/bin/sh", NULL, environ);
230
231 log_emergency_errno(errno, "execle() failed: %m");
232 _exit(EXIT_FAILURE);
233 } else {
234 log_info("Spawned crash shell as PID "PID_FMT".", pid);
235 (void) wait_for_terminate(pid, NULL);
236 }
237 }
238
239 freeze_or_reboot();
240 }
241
242 static void install_crash_handler(void) {
243 static const struct sigaction sa = {
244 .sa_handler = crash,
245 .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
246 };
247 int r;
248
249 /* We ignore the return value here, since, we don't mind if we
250 * cannot set up a crash handler */
251 r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
252 if (r < 0)
253 log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
254 }
255
256 static int console_setup(void) {
257 _cleanup_close_ int tty_fd = -1;
258 int r;
259
260 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
261 if (tty_fd < 0)
262 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
263
264 /* We don't want to force text mode. plymouth may be showing
265 * pictures already from initrd. */
266 r = reset_terminal_fd(tty_fd, false);
267 if (r < 0)
268 return log_error_errno(r, "Failed to reset /dev/console: %m");
269
270 return 0;
271 }
272
273 static int parse_crash_chvt(const char *value) {
274 int b;
275
276 if (safe_atoi(value, &arg_crash_chvt) >= 0)
277 return 0;
278
279 b = parse_boolean(value);
280 if (b < 0)
281 return b;
282
283 if (b > 0)
284 arg_crash_chvt = 0; /* switch to where kmsg goes */
285 else
286 arg_crash_chvt = -1; /* turn off switching */
287
288 return 0;
289 }
290
291 static int parse_confirm_spawn(const char *value, char **console) {
292 char *s;
293 int r;
294
295 r = value ? parse_boolean(value) : 1;
296 if (r == 0) {
297 *console = NULL;
298 return 0;
299 }
300
301 if (r > 0) /* on with default tty */
302 s = strdup("/dev/console");
303 else if (is_path(value)) /* on with fully qualified path */
304 s = strdup(value);
305 else /* on with only a tty file name, not a fully qualified path */
306 s = strjoin("/dev/", value);
307 if (!s)
308 return -ENOMEM;
309
310 *console = s;
311 return 0;
312 }
313
314 static int set_machine_id(const char *m) {
315 sd_id128_t t;
316 assert(m);
317
318 if (sd_id128_from_string(m, &t) < 0)
319 return -EINVAL;
320
321 if (sd_id128_is_null(t))
322 return -EINVAL;
323
324 arg_machine_id = t;
325 return 0;
326 }
327
328 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
329
330 int r;
331
332 assert(key);
333
334 if (STR_IN_SET(key, "systemd.unit", "rd.systemd.unit")) {
335
336 if (proc_cmdline_value_missing(key, value))
337 return 0;
338
339 if (!unit_name_is_valid(value, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE))
340 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key, value);
341 else if (in_initrd() == !!startswith(key, "rd.")) {
342 if (free_and_strdup(&arg_default_unit, value) < 0)
343 return log_oom();
344 }
345
346 } else if (proc_cmdline_key_streq(key, "systemd.dump_core")) {
347
348 r = value ? parse_boolean(value) : true;
349 if (r < 0)
350 log_warning("Failed to parse dump core switch %s. Ignoring.", value);
351 else
352 arg_dump_core = r;
353
354 } else if (proc_cmdline_key_streq(key, "systemd.crash_chvt")) {
355
356 if (!value)
357 arg_crash_chvt = 0; /* turn on */
358 else if (parse_crash_chvt(value) < 0)
359 log_warning("Failed to parse crash chvt switch %s. Ignoring.", value);
360
361 } else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) {
362
363 r = value ? parse_boolean(value) : true;
364 if (r < 0)
365 log_warning("Failed to parse crash shell switch %s. Ignoring.", value);
366 else
367 arg_crash_shell = r;
368
369 } else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) {
370
371 r = value ? parse_boolean(value) : true;
372 if (r < 0)
373 log_warning("Failed to parse crash reboot switch %s. Ignoring.", value);
374 else
375 arg_crash_reboot = r;
376
377 } else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) {
378 char *s;
379
380 r = parse_confirm_spawn(value, &s);
381 if (r < 0)
382 log_warning_errno(r, "Failed to parse confirm_spawn switch %s. Ignoring.", value);
383 else {
384 free(arg_confirm_spawn);
385 arg_confirm_spawn = s;
386 }
387
388 } else if (proc_cmdline_key_streq(key, "systemd.service_watchdogs")) {
389
390 r = value ? parse_boolean(value) : true;
391 if (r < 0)
392 log_warning("Failed to parse service watchdog switch %s. Ignoring.", value);
393 else
394 arg_service_watchdogs = r;
395
396 } else if (proc_cmdline_key_streq(key, "systemd.show_status")) {
397
398 if (value) {
399 r = parse_show_status(value, &arg_show_status);
400 if (r < 0)
401 log_warning("Failed to parse show status switch %s. Ignoring.", value);
402 } else
403 arg_show_status = SHOW_STATUS_YES;
404
405 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) {
406
407 if (proc_cmdline_value_missing(key, value))
408 return 0;
409
410 r = exec_output_from_string(value);
411 if (r < 0)
412 log_warning("Failed to parse default standard output switch %s. Ignoring.", value);
413 else
414 arg_default_std_output = r;
415
416 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) {
417
418 if (proc_cmdline_value_missing(key, value))
419 return 0;
420
421 r = exec_output_from_string(value);
422 if (r < 0)
423 log_warning("Failed to parse default standard error switch %s. Ignoring.", value);
424 else
425 arg_default_std_error = r;
426
427 } else if (streq(key, "systemd.setenv")) {
428
429 if (proc_cmdline_value_missing(key, value))
430 return 0;
431
432 if (env_assignment_is_valid(value)) {
433 char **env;
434
435 env = strv_env_set(arg_default_environment, value);
436 if (!env)
437 return log_oom();
438
439 arg_default_environment = env;
440 } else
441 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
442
443 } else if (proc_cmdline_key_streq(key, "systemd.machine_id")) {
444
445 if (proc_cmdline_value_missing(key, value))
446 return 0;
447
448 r = set_machine_id(value);
449 if (r < 0)
450 log_warning("MachineID '%s' is not valid. Ignoring.", value);
451
452 } else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) {
453
454 if (proc_cmdline_value_missing(key, value))
455 return 0;
456
457 r = parse_sec(value, &arg_default_timeout_start_usec);
458 if (r < 0)
459 log_warning_errno(r, "Failed to parse default start timeout: %s, ignoring.", value);
460
461 if (arg_default_timeout_start_usec <= 0)
462 arg_default_timeout_start_usec = USEC_INFINITY;
463
464 } else if (proc_cmdline_key_streq(key, "systemd.watchdog_device")) {
465
466 if (proc_cmdline_value_missing(key, value))
467 return 0;
468
469 parse_path_argument_and_warn(value, false, &arg_watchdog_device);
470
471 } else if (streq(key, "quiet") && !value) {
472
473 if (arg_show_status == _SHOW_STATUS_INVALID)
474 arg_show_status = SHOW_STATUS_AUTO;
475
476 } else if (streq(key, "debug") && !value) {
477
478 /* Note that log_parse_environment() handles 'debug'
479 * too, and sets the log level to LOG_DEBUG. */
480
481 if (detect_container() > 0)
482 log_set_target(LOG_TARGET_CONSOLE);
483
484 } else if (!value) {
485 const char *target;
486
487 /* SysV compatibility */
488 target = runlevel_to_target(key);
489 if (target)
490 return free_and_strdup(&arg_default_unit, target);
491 }
492
493 return 0;
494 }
495
496 #define DEFINE_SETTER(name, func, descr) \
497 static int name(const char *unit, \
498 const char *filename, \
499 unsigned line, \
500 const char *section, \
501 unsigned section_line, \
502 const char *lvalue, \
503 int ltype, \
504 const char *rvalue, \
505 void *data, \
506 void *userdata) { \
507 \
508 int r; \
509 \
510 assert(filename); \
511 assert(lvalue); \
512 assert(rvalue); \
513 \
514 r = func(rvalue); \
515 if (r < 0) \
516 log_syntax(unit, LOG_ERR, filename, line, r, \
517 "Invalid " descr "'%s': %m", \
518 rvalue); \
519 \
520 return 0; \
521 }
522
523 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level");
524 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target");
525 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" );
526 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location");
527
528 static int config_parse_cpu_affinity2(
529 const char *unit,
530 const char *filename,
531 unsigned line,
532 const char *section,
533 unsigned section_line,
534 const char *lvalue,
535 int ltype,
536 const char *rvalue,
537 void *data,
538 void *userdata) {
539
540 _cleanup_cpu_free_ cpu_set_t *c = NULL;
541 int ncpus;
542
543 ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue);
544 if (ncpus < 0)
545 return ncpus;
546
547 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
548 log_warning_errno(errno, "Failed to set CPU affinity: %m");
549
550 return 0;
551 }
552
553 static int config_parse_show_status(
554 const char* unit,
555 const char *filename,
556 unsigned line,
557 const char *section,
558 unsigned section_line,
559 const char *lvalue,
560 int ltype,
561 const char *rvalue,
562 void *data,
563 void *userdata) {
564
565 int k;
566 ShowStatus *b = data;
567
568 assert(filename);
569 assert(lvalue);
570 assert(rvalue);
571 assert(data);
572
573 k = parse_show_status(rvalue, b);
574 if (k < 0) {
575 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
576 return 0;
577 }
578
579 return 0;
580 }
581
582 static int config_parse_output_restricted(
583 const char* unit,
584 const char *filename,
585 unsigned line,
586 const char *section,
587 unsigned section_line,
588 const char *lvalue,
589 int ltype,
590 const char *rvalue,
591 void *data,
592 void *userdata) {
593
594 ExecOutput t, *eo = data;
595
596 assert(filename);
597 assert(lvalue);
598 assert(rvalue);
599 assert(data);
600
601 t = exec_output_from_string(rvalue);
602 if (t < 0) {
603 log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse output type, ignoring: %s", rvalue);
604 return 0;
605 }
606
607 if (IN_SET(t, EXEC_OUTPUT_SOCKET, EXEC_OUTPUT_NAMED_FD, EXEC_OUTPUT_FILE, EXEC_OUTPUT_FILE_APPEND)) {
608 log_syntax(unit, LOG_ERR, filename, line, 0, "Standard output types socket, fd:, file:, append: are not supported as defaults, ignoring: %s", rvalue);
609 return 0;
610 }
611
612 *eo = t;
613 return 0;
614 }
615
616 static int config_parse_crash_chvt(
617 const char* unit,
618 const char *filename,
619 unsigned line,
620 const char *section,
621 unsigned section_line,
622 const char *lvalue,
623 int ltype,
624 const char *rvalue,
625 void *data,
626 void *userdata) {
627
628 int r;
629
630 assert(filename);
631 assert(lvalue);
632 assert(rvalue);
633
634 r = parse_crash_chvt(rvalue);
635 if (r < 0) {
636 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
637 return 0;
638 }
639
640 return 0;
641 }
642
643 static int parse_config_file(void) {
644
645 const ConfigTableItem items[] = {
646 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
647 { "Manager", "LogTarget", config_parse_target, 0, NULL },
648 { "Manager", "LogColor", config_parse_color, 0, NULL },
649 { "Manager", "LogLocation", config_parse_location, 0, NULL },
650 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
651 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL },
652 { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL },
653 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
654 { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
655 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
656 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
657 { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
658 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
659 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
660 { "Manager", "WatchdogDevice", config_parse_path, 0, &arg_watchdog_device },
661 { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
662 { "Manager", "NoNewPrivileges", config_parse_bool, 0, &arg_no_new_privs },
663 #if HAVE_SECCOMP
664 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
665 #endif
666 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
667 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
668 { "Manager", "DefaultStandardOutput", config_parse_output_restricted,0, &arg_default_std_output },
669 { "Manager", "DefaultStandardError", config_parse_output_restricted,0, &arg_default_std_error },
670 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
671 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
672 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
673 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */
674 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec, 0, &arg_default_start_limit_interval },
675 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
676 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
677 { "Manager", "DefaultLimitCPU", config_parse_rlimit, RLIMIT_CPU, arg_default_rlimit },
678 { "Manager", "DefaultLimitFSIZE", config_parse_rlimit, RLIMIT_FSIZE, arg_default_rlimit },
679 { "Manager", "DefaultLimitDATA", config_parse_rlimit, RLIMIT_DATA, arg_default_rlimit },
680 { "Manager", "DefaultLimitSTACK", config_parse_rlimit, RLIMIT_STACK, arg_default_rlimit },
681 { "Manager", "DefaultLimitCORE", config_parse_rlimit, RLIMIT_CORE, arg_default_rlimit },
682 { "Manager", "DefaultLimitRSS", config_parse_rlimit, RLIMIT_RSS, arg_default_rlimit },
683 { "Manager", "DefaultLimitNOFILE", config_parse_rlimit, RLIMIT_NOFILE, arg_default_rlimit },
684 { "Manager", "DefaultLimitAS", config_parse_rlimit, RLIMIT_AS, arg_default_rlimit },
685 { "Manager", "DefaultLimitNPROC", config_parse_rlimit, RLIMIT_NPROC, arg_default_rlimit },
686 { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit, RLIMIT_MEMLOCK, arg_default_rlimit },
687 { "Manager", "DefaultLimitLOCKS", config_parse_rlimit, RLIMIT_LOCKS, arg_default_rlimit },
688 { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit, RLIMIT_SIGPENDING, arg_default_rlimit },
689 { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit, RLIMIT_MSGQUEUE, arg_default_rlimit },
690 { "Manager", "DefaultLimitNICE", config_parse_rlimit, RLIMIT_NICE, arg_default_rlimit },
691 { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit, RLIMIT_RTPRIO, arg_default_rlimit },
692 { "Manager", "DefaultLimitRTTIME", config_parse_rlimit, RLIMIT_RTTIME, arg_default_rlimit },
693 { "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_default_cpu_accounting },
694 { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting },
695 { "Manager", "DefaultIPAccounting", config_parse_bool, 0, &arg_default_ip_accounting },
696 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
697 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
698 { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
699 { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max },
700 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action, 0, &arg_cad_burst_action },
701 {}
702 };
703
704 const char *fn, *conf_dirs_nulstr;
705
706 fn = arg_system ?
707 PKGSYSCONFDIR "/system.conf" :
708 PKGSYSCONFDIR "/user.conf";
709
710 conf_dirs_nulstr = arg_system ?
711 CONF_PATHS_NULSTR("systemd/system.conf.d") :
712 CONF_PATHS_NULSTR("systemd/user.conf.d");
713
714 (void) config_parse_many_nulstr(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, CONFIG_PARSE_WARN, NULL);
715
716 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
717 * like everywhere else. */
718 if (arg_default_timeout_start_usec <= 0)
719 arg_default_timeout_start_usec = USEC_INFINITY;
720 if (arg_default_timeout_stop_usec <= 0)
721 arg_default_timeout_stop_usec = USEC_INFINITY;
722
723 return 0;
724 }
725
726 static void set_manager_defaults(Manager *m) {
727
728 assert(m);
729
730 /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
731 * affect the manager itself, but are just what newly allocated units will have set if they haven't set
732 * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
733
734 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
735 m->default_std_output = arg_default_std_output;
736 m->default_std_error = arg_default_std_error;
737 m->default_timeout_start_usec = arg_default_timeout_start_usec;
738 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
739 m->default_restart_usec = arg_default_restart_usec;
740 m->default_start_limit_interval = arg_default_start_limit_interval;
741 m->default_start_limit_burst = arg_default_start_limit_burst;
742 m->default_cpu_accounting = arg_default_cpu_accounting;
743 m->default_io_accounting = arg_default_io_accounting;
744 m->default_ip_accounting = arg_default_ip_accounting;
745 m->default_blockio_accounting = arg_default_blockio_accounting;
746 m->default_memory_accounting = arg_default_memory_accounting;
747 m->default_tasks_accounting = arg_default_tasks_accounting;
748 m->default_tasks_max = arg_default_tasks_max;
749
750 manager_set_default_rlimits(m, arg_default_rlimit);
751 manager_environment_add(m, NULL, arg_default_environment);
752 }
753
754 static void set_manager_settings(Manager *m) {
755
756 assert(m);
757
758 /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
759 * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
760
761 m->confirm_spawn = arg_confirm_spawn;
762 m->service_watchdogs = arg_service_watchdogs;
763 m->runtime_watchdog = arg_runtime_watchdog;
764 m->shutdown_watchdog = arg_shutdown_watchdog;
765 m->cad_burst_action = arg_cad_burst_action;
766
767 manager_set_show_status(m, arg_show_status);
768 }
769
770 static int parse_argv(int argc, char *argv[]) {
771 enum {
772 ARG_LOG_LEVEL = 0x100,
773 ARG_LOG_TARGET,
774 ARG_LOG_COLOR,
775 ARG_LOG_LOCATION,
776 ARG_UNIT,
777 ARG_SYSTEM,
778 ARG_USER,
779 ARG_TEST,
780 ARG_NO_PAGER,
781 ARG_VERSION,
782 ARG_DUMP_CONFIGURATION_ITEMS,
783 ARG_DUMP_BUS_PROPERTIES,
784 ARG_DUMP_CORE,
785 ARG_CRASH_CHVT,
786 ARG_CRASH_SHELL,
787 ARG_CRASH_REBOOT,
788 ARG_CONFIRM_SPAWN,
789 ARG_SHOW_STATUS,
790 ARG_DESERIALIZE,
791 ARG_SWITCHED_ROOT,
792 ARG_DEFAULT_STD_OUTPUT,
793 ARG_DEFAULT_STD_ERROR,
794 ARG_MACHINE_ID,
795 ARG_SERVICE_WATCHDOGS,
796 };
797
798 static const struct option options[] = {
799 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
800 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
801 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
802 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
803 { "unit", required_argument, NULL, ARG_UNIT },
804 { "system", no_argument, NULL, ARG_SYSTEM },
805 { "user", no_argument, NULL, ARG_USER },
806 { "test", no_argument, NULL, ARG_TEST },
807 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
808 { "help", no_argument, NULL, 'h' },
809 { "version", no_argument, NULL, ARG_VERSION },
810 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
811 { "dump-bus-properties", no_argument, NULL, ARG_DUMP_BUS_PROPERTIES },
812 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
813 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
814 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
815 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
816 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
817 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
818 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
819 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
820 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
821 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
822 { "machine-id", required_argument, NULL, ARG_MACHINE_ID },
823 { "service-watchdogs", required_argument, NULL, ARG_SERVICE_WATCHDOGS },
824 {}
825 };
826
827 int c, r;
828
829 assert(argc >= 1);
830 assert(argv);
831
832 if (getpid_cached() == 1)
833 opterr = 0;
834
835 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
836
837 switch (c) {
838
839 case ARG_LOG_LEVEL:
840 r = log_set_max_level_from_string(optarg);
841 if (r < 0)
842 return log_error_errno(r, "Failed to parse log level \"%s\": %m", optarg);
843
844 break;
845
846 case ARG_LOG_TARGET:
847 r = log_set_target_from_string(optarg);
848 if (r < 0)
849 return log_error_errno(r, "Failed to parse log target \"%s\": %m", optarg);
850
851 break;
852
853 case ARG_LOG_COLOR:
854
855 if (optarg) {
856 r = log_show_color_from_string(optarg);
857 if (r < 0)
858 return log_error_errno(r, "Failed to parse log color setting \"%s\": %m",
859 optarg);
860 } else
861 log_show_color(true);
862
863 break;
864
865 case ARG_LOG_LOCATION:
866 if (optarg) {
867 r = log_show_location_from_string(optarg);
868 if (r < 0)
869 return log_error_errno(r, "Failed to parse log location setting \"%s\": %m",
870 optarg);
871 } else
872 log_show_location(true);
873
874 break;
875
876 case ARG_DEFAULT_STD_OUTPUT:
877 r = exec_output_from_string(optarg);
878 if (r < 0)
879 return log_error_errno(r, "Failed to parse default standard output setting \"%s\": %m",
880 optarg);
881 arg_default_std_output = r;
882 break;
883
884 case ARG_DEFAULT_STD_ERROR:
885 r = exec_output_from_string(optarg);
886 if (r < 0)
887 return log_error_errno(r, "Failed to parse default standard error output setting \"%s\": %m",
888 optarg);
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_BUS_PROPERTIES:
924 arg_action = ACTION_DUMP_BUS_PROPERTIES;
925 break;
926
927 case ARG_DUMP_CORE:
928 if (!optarg)
929 arg_dump_core = true;
930 else {
931 r = parse_boolean(optarg);
932 if (r < 0)
933 return log_error_errno(r, "Failed to parse dump core boolean: \"%s\": %m",
934 optarg);
935 arg_dump_core = r;
936 }
937 break;
938
939 case ARG_CRASH_CHVT:
940 r = parse_crash_chvt(optarg);
941 if (r < 0)
942 return log_error_errno(r, "Failed to parse crash virtual terminal index: \"%s\": %m",
943 optarg);
944 break;
945
946 case ARG_CRASH_SHELL:
947 if (!optarg)
948 arg_crash_shell = true;
949 else {
950 r = parse_boolean(optarg);
951 if (r < 0)
952 return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
953 optarg);
954 arg_crash_shell = r;
955 }
956 break;
957
958 case ARG_CRASH_REBOOT:
959 if (!optarg)
960 arg_crash_reboot = true;
961 else {
962 r = parse_boolean(optarg);
963 if (r < 0)
964 return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
965 optarg);
966 arg_crash_reboot = r;
967 }
968 break;
969
970 case ARG_CONFIRM_SPAWN:
971 arg_confirm_spawn = mfree(arg_confirm_spawn);
972
973 r = parse_confirm_spawn(optarg, &arg_confirm_spawn);
974 if (r < 0)
975 return log_error_errno(r, "Failed to parse confirm spawn option: \"%s\": %m",
976 optarg);
977 break;
978
979 case ARG_SERVICE_WATCHDOGS:
980 r = parse_boolean(optarg);
981 if (r < 0)
982 return log_error_errno(r, "Failed to parse service watchdogs boolean: \"%s\": %m",
983 optarg);
984 arg_service_watchdogs = r;
985 break;
986
987 case ARG_SHOW_STATUS:
988 if (optarg) {
989 r = parse_show_status(optarg, &arg_show_status);
990 if (r < 0)
991 return log_error_errno(r, "Failed to parse show status boolean: \"%s\": %m",
992 optarg);
993 } else
994 arg_show_status = SHOW_STATUS_YES;
995 break;
996
997 case ARG_DESERIALIZE: {
998 int fd;
999 FILE *f;
1000
1001 r = safe_atoi(optarg, &fd);
1002 if (r < 0)
1003 log_error_errno(r, "Failed to parse deserialize option \"%s\": %m", optarg);
1004 if (fd < 0) {
1005 log_error("Invalid deserialize fd: %d", fd);
1006 return -EINVAL;
1007 }
1008
1009 (void) fd_cloexec(fd, true);
1010
1011 f = fdopen(fd, "r");
1012 if (!f)
1013 return log_error_errno(errno, "Failed to open serialization fd %d: %m", fd);
1014
1015 safe_fclose(arg_serialization);
1016 arg_serialization = f;
1017
1018 break;
1019 }
1020
1021 case ARG_SWITCHED_ROOT:
1022 arg_switched_root = true;
1023 break;
1024
1025 case ARG_MACHINE_ID:
1026 r = set_machine_id(optarg);
1027 if (r < 0)
1028 return log_error_errno(r, "MachineID '%s' is not valid: %m", optarg);
1029 break;
1030
1031 case 'h':
1032 arg_action = ACTION_HELP;
1033 break;
1034
1035 case 'D':
1036 log_set_max_level(LOG_DEBUG);
1037 break;
1038
1039 case 'b':
1040 case 's':
1041 case 'z':
1042 /* Just to eat away the sysvinit kernel
1043 * cmdline args without getopt() error
1044 * messages that we'll parse in
1045 * parse_proc_cmdline_word() or ignore. */
1046
1047 case '?':
1048 if (getpid_cached() != 1)
1049 return -EINVAL;
1050 else
1051 return 0;
1052
1053 default:
1054 assert_not_reached("Unhandled option code.");
1055 }
1056
1057 if (optind < argc && getpid_cached() != 1) {
1058 /* Hmm, when we aren't run as init system
1059 * let's complain about excess arguments */
1060
1061 log_error("Excess arguments.");
1062 return -EINVAL;
1063 }
1064
1065 return 0;
1066 }
1067
1068 static int help(void) {
1069 _cleanup_free_ char *link = NULL;
1070 int r;
1071
1072 r = terminal_urlify_man("systemd", "1", &link);
1073 if (r < 0)
1074 return log_oom();
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 "\nSee the %s for details.\n"
1100 , program_invocation_short_name
1101 , link
1102 );
1103
1104 return 0;
1105 }
1106
1107 static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
1108 _cleanup_fdset_free_ FDSet *fds = NULL;
1109 _cleanup_fclose_ FILE *f = NULL;
1110 int r;
1111
1112 assert(m);
1113 assert(_f);
1114 assert(_fds);
1115
1116 r = manager_open_serialization(m, &f);
1117 if (r < 0)
1118 return log_error_errno(r, "Failed to create serialization file: %m");
1119
1120 /* Make sure nothing is really destructed when we shut down */
1121 m->n_reloading++;
1122 bus_manager_send_reloading(m, true);
1123
1124 fds = fdset_new();
1125 if (!fds)
1126 return log_oom();
1127
1128 r = manager_serialize(m, f, fds, switching_root);
1129 if (r < 0)
1130 return log_error_errno(r, "Failed to serialize state: %m");
1131
1132 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1133 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1134
1135 r = fd_cloexec(fileno(f), false);
1136 if (r < 0)
1137 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1138
1139 r = fdset_cloexec(fds, false);
1140 if (r < 0)
1141 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1142
1143 *_f = TAKE_PTR(f);
1144 *_fds = TAKE_PTR(fds);
1145
1146 return 0;
1147 }
1148
1149 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1150 int r, nr;
1151
1152 assert(saved_rlimit);
1153
1154 /* Save the original RLIMIT_NOFILE so that we can reset it
1155 * later when transitioning from the initrd to the main
1156 * systemd or suchlike. */
1157 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1158 return log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1159
1160 /* Make sure forked processes get the default kernel setting */
1161 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1162 struct rlimit *rl;
1163
1164 rl = newdup(struct rlimit, saved_rlimit, 1);
1165 if (!rl)
1166 return log_oom();
1167
1168 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1169 }
1170
1171 /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows */
1172 nr = read_nr_open();
1173 r = setrlimit_closest(RLIMIT_NOFILE, &RLIMIT_MAKE_CONST(nr));
1174 if (r < 0)
1175 return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1176
1177 return 0;
1178 }
1179
1180 static int bump_rlimit_memlock(struct rlimit *saved_rlimit) {
1181 int r;
1182
1183 assert(saved_rlimit);
1184 assert(getuid() == 0);
1185
1186 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even though we have CAP_IPC_LOCK which
1187 * should normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's
1188 * bump the value high enough for the root user. */
1189
1190 if (getrlimit(RLIMIT_MEMLOCK, saved_rlimit) < 0)
1191 return log_warning_errno(errno, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1192
1193 r = setrlimit_closest(RLIMIT_MEMLOCK, &RLIMIT_MAKE_CONST(1024ULL*1024ULL*16ULL));
1194 if (r < 0)
1195 return log_warning_errno(r, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1196
1197 return 0;
1198 }
1199
1200 static void test_usr(void) {
1201
1202 /* Check that /usr is not a separate fs */
1203
1204 if (dir_is_empty("/usr") <= 0)
1205 return;
1206
1207 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1208 "Some things will probably break (sometimes even silently) in mysterious ways. "
1209 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1210 }
1211
1212 static int enforce_syscall_archs(Set *archs) {
1213 #if HAVE_SECCOMP
1214 int r;
1215
1216 if (!is_seccomp_available())
1217 return 0;
1218
1219 r = seccomp_restrict_archs(arg_syscall_archs);
1220 if (r < 0)
1221 return log_error_errno(r, "Failed to enforce system call architecture restrication: %m");
1222 #endif
1223 return 0;
1224 }
1225
1226 static int status_welcome(void) {
1227 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1228 int r;
1229
1230 if (IN_SET(arg_show_status, SHOW_STATUS_NO, SHOW_STATUS_AUTO))
1231 return 0;
1232
1233 r = parse_os_release(NULL,
1234 "PRETTY_NAME", &pretty_name,
1235 "ANSI_COLOR", &ansi_color,
1236 NULL);
1237 if (r < 0)
1238 log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r,
1239 "Failed to read os-release file, ignoring: %m");
1240
1241 if (log_get_show_color())
1242 return status_printf(NULL, false, false,
1243 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1244 isempty(ansi_color) ? "1" : ansi_color,
1245 isempty(pretty_name) ? "Linux" : pretty_name);
1246 else
1247 return status_printf(NULL, false, false,
1248 "\nWelcome to %s!\n",
1249 isempty(pretty_name) ? "Linux" : pretty_name);
1250 }
1251
1252 static int write_container_id(void) {
1253 const char *c;
1254 int r;
1255
1256 c = getenv("container");
1257 if (isempty(c))
1258 return 0;
1259
1260 RUN_WITH_UMASK(0022)
1261 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1262 if (r < 0)
1263 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
1264
1265 return 1;
1266 }
1267
1268 static int bump_unix_max_dgram_qlen(void) {
1269 _cleanup_free_ char *qlen = NULL;
1270 unsigned long v;
1271 int r;
1272
1273 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1274 * really really early during boot, so that it is actually applied to all our sockets, including the
1275 * $NOTIFY_SOCKET one. */
1276
1277 r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1278 if (r < 0)
1279 return log_warning_errno(r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1280
1281 r = safe_atolu(qlen, &v);
1282 if (r < 0)
1283 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen);
1284
1285 if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1286 return 0;
1287
1288 r = write_string_filef("/proc/sys/net/unix/max_dgram_qlen", 0, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN);
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 #if ENABLE_COREDUMP
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 #endif
1490 }
1491
1492 static void do_reexecute(
1493 int argc,
1494 char *argv[],
1495 const struct rlimit *saved_rlimit_nofile,
1496 const struct rlimit *saved_rlimit_memlock,
1497 FDSet *fds,
1498 const char *switch_root_dir,
1499 const char *switch_root_init,
1500 const char **ret_error_message) {
1501
1502 unsigned i, j, args_size;
1503 const char **args;
1504 int r;
1505
1506 assert(saved_rlimit_nofile);
1507 assert(saved_rlimit_memlock);
1508 assert(ret_error_message);
1509
1510 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1511 * we do that */
1512 watchdog_close(true);
1513
1514 /* Reset the RLIMIT_NOFILE to the kernel default, so that the new systemd can pass the kernel default to its
1515 * child processes */
1516
1517 if (saved_rlimit_nofile->rlim_cur > 0)
1518 (void) setrlimit(RLIMIT_NOFILE, saved_rlimit_nofile);
1519 if (saved_rlimit_memlock->rlim_cur != (rlim_t) -1)
1520 (void) setrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock);
1521
1522 if (switch_root_dir) {
1523 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1524 * SIGCHLD for them after deserializing. */
1525 broadcast_signal(SIGTERM, false, true, arg_default_timeout_stop_usec);
1526
1527 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1528 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1529 if (r < 0)
1530 log_error_errno(r, "Failed to switch root, trying to continue: %m");
1531 }
1532
1533 args_size = MAX(6, argc+1);
1534 args = newa(const char*, args_size);
1535
1536 if (!switch_root_init) {
1537 char sfd[DECIMAL_STR_MAX(int) + 1];
1538
1539 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1540 * the user didn't specify an explicit init to spawn. */
1541
1542 assert(arg_serialization);
1543 assert(fds);
1544
1545 xsprintf(sfd, "%i", fileno(arg_serialization));
1546
1547 i = 0;
1548 args[i++] = SYSTEMD_BINARY_PATH;
1549 if (switch_root_dir)
1550 args[i++] = "--switched-root";
1551 args[i++] = arg_system ? "--system" : "--user";
1552 args[i++] = "--deserialize";
1553 args[i++] = sfd;
1554 args[i++] = NULL;
1555
1556 assert(i <= args_size);
1557
1558 /*
1559 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do this is on
1560 * its own on exec(), but it will do it on exit(). Hence, to ensure we get a summary here, fork() off
1561 * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1562 * before proceeding into the exec().
1563 */
1564 valgrind_summary_hack();
1565
1566 (void) execv(args[0], (char* const*) args);
1567 log_debug_errno(errno, "Failed to execute our own binary, trying fallback: %m");
1568 }
1569
1570 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1571 * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1572 * doesn't matter.) */
1573
1574 arg_serialization = safe_fclose(arg_serialization);
1575 fds = fdset_free(fds);
1576
1577 /* Reopen the console */
1578 (void) make_console_stdio();
1579
1580 for (j = 1, i = 1; j < (unsigned) argc; j++)
1581 args[i++] = argv[j];
1582 args[i++] = NULL;
1583 assert(i <= args_size);
1584
1585 /* Reenable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1586 (void) reset_all_signal_handlers();
1587 (void) reset_signal_mask();
1588
1589 if (switch_root_init) {
1590 args[0] = switch_root_init;
1591 (void) execv(args[0], (char* const*) args);
1592 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
1593 }
1594
1595 args[0] = "/sbin/init";
1596 (void) execv(args[0], (char* const*) args);
1597 r = -errno;
1598
1599 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
1600 ANSI_HIGHLIGHT_RED " !! " ANSI_NORMAL,
1601 "Failed to execute /sbin/init");
1602
1603 if (r == -ENOENT) {
1604 log_warning("No /sbin/init, trying fallback");
1605
1606 args[0] = "/bin/sh";
1607 args[1] = NULL;
1608 (void) execv(args[0], (char* const*) args);
1609 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
1610 } else
1611 log_warning_errno(r, "Failed to execute /sbin/init, giving up: %m");
1612
1613 *ret_error_message = "Failed to execute fallback shell";
1614 }
1615
1616 static int invoke_main_loop(
1617 Manager *m,
1618 bool *ret_reexecute,
1619 int *ret_retval, /* Return parameters relevant for shutting down */
1620 const char **ret_shutdown_verb, /* … */
1621 FDSet **ret_fds, /* Return parameters for reexecuting */
1622 char **ret_switch_root_dir, /* … */
1623 char **ret_switch_root_init, /* … */
1624 const char **ret_error_message) {
1625
1626 int r;
1627
1628 assert(m);
1629 assert(ret_reexecute);
1630 assert(ret_retval);
1631 assert(ret_shutdown_verb);
1632 assert(ret_fds);
1633 assert(ret_switch_root_dir);
1634 assert(ret_switch_root_init);
1635 assert(ret_error_message);
1636
1637 for (;;) {
1638 r = manager_loop(m);
1639 if (r < 0) {
1640 *ret_error_message = "Failed to run main loop";
1641 return log_emergency_errno(r, "Failed to run main loop: %m");
1642 }
1643
1644 switch (m->exit_code) {
1645
1646 case MANAGER_RELOAD: {
1647 LogTarget saved_log_target;
1648 int saved_log_level;
1649
1650 log_info("Reloading.");
1651
1652 /* First, save any overridden log level/target, then parse the configuration file, which might
1653 * change the log level to new settings. */
1654
1655 saved_log_level = m->log_level_overridden ? log_get_max_level() : -1;
1656 saved_log_target = m->log_target_overridden ? log_get_target() : _LOG_TARGET_INVALID;
1657
1658 r = parse_config_file();
1659 if (r < 0)
1660 log_warning_errno(r, "Failed to parse config file, ignoring: %m");
1661
1662 set_manager_defaults(m);
1663
1664 if (saved_log_level >= 0)
1665 manager_override_log_level(m, saved_log_level);
1666 if (saved_log_target >= 0)
1667 manager_override_log_target(m, saved_log_target);
1668
1669 r = manager_reload(m);
1670 if (r < 0)
1671 log_warning_errno(r, "Failed to reload, ignoring: %m");
1672
1673 break;
1674 }
1675
1676 case MANAGER_REEXECUTE:
1677
1678 r = prepare_reexecute(m, &arg_serialization, ret_fds, false);
1679 if (r < 0) {
1680 *ret_error_message = "Failed to prepare for reexecution";
1681 return r;
1682 }
1683
1684 log_notice("Reexecuting.");
1685
1686 *ret_reexecute = true;
1687 *ret_retval = EXIT_SUCCESS;
1688 *ret_shutdown_verb = NULL;
1689 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1690
1691 return 0;
1692
1693 case MANAGER_SWITCH_ROOT:
1694 if (!m->switch_root_init) {
1695 r = prepare_reexecute(m, &arg_serialization, ret_fds, true);
1696 if (r < 0) {
1697 *ret_error_message = "Failed to prepare for reexecution";
1698 return r;
1699 }
1700 } else
1701 *ret_fds = NULL;
1702
1703 log_notice("Switching root.");
1704
1705 *ret_reexecute = true;
1706 *ret_retval = EXIT_SUCCESS;
1707 *ret_shutdown_verb = NULL;
1708
1709 /* Steal the switch root parameters */
1710 *ret_switch_root_dir = m->switch_root;
1711 *ret_switch_root_init = m->switch_root_init;
1712 m->switch_root = m->switch_root_init = NULL;
1713
1714 return 0;
1715
1716 case MANAGER_EXIT:
1717
1718 if (MANAGER_IS_USER(m)) {
1719 log_debug("Exit.");
1720
1721 *ret_reexecute = false;
1722 *ret_retval = m->return_value;
1723 *ret_shutdown_verb = NULL;
1724 *ret_fds = NULL;
1725 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1726
1727 return 0;
1728 }
1729
1730 _fallthrough_;
1731 case MANAGER_REBOOT:
1732 case MANAGER_POWEROFF:
1733 case MANAGER_HALT:
1734 case MANAGER_KEXEC: {
1735 static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
1736 [MANAGER_EXIT] = "exit",
1737 [MANAGER_REBOOT] = "reboot",
1738 [MANAGER_POWEROFF] = "poweroff",
1739 [MANAGER_HALT] = "halt",
1740 [MANAGER_KEXEC] = "kexec"
1741 };
1742
1743 log_notice("Shutting down.");
1744
1745 *ret_reexecute = false;
1746 *ret_retval = m->return_value;
1747 assert_se(*ret_shutdown_verb = table[m->exit_code]);
1748 *ret_fds = NULL;
1749 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1750
1751 return 0;
1752 }
1753
1754 default:
1755 assert_not_reached("Unknown exit code.");
1756 }
1757 }
1758 }
1759
1760 static void log_execution_mode(bool *ret_first_boot) {
1761 assert(ret_first_boot);
1762
1763 if (arg_system) {
1764 int v;
1765
1766 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1767 arg_action == ACTION_TEST ? "test " : "" );
1768
1769 v = detect_virtualization();
1770 if (v > 0)
1771 log_info("Detected virtualization %s.", virtualization_to_string(v));
1772
1773 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1774
1775 if (in_initrd()) {
1776 *ret_first_boot = false;
1777 log_info("Running in initial RAM disk.");
1778 } else {
1779 /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
1780 * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
1781 * doesn't it's unpopulated. This allows container managers and installers to provision a
1782 * couple of files already. If the container manager wants to provision the machine ID itself
1783 * it should pass $container_uuid to PID 1. */
1784
1785 *ret_first_boot = access("/etc/machine-id", F_OK) < 0;
1786 if (*ret_first_boot)
1787 log_info("Running with unpopulated /etc.");
1788 }
1789 } else {
1790 if (DEBUG_LOGGING) {
1791 _cleanup_free_ char *t;
1792
1793 t = uid_to_name(getuid());
1794 log_debug(PACKAGE_STRING " running in %suser mode for user " UID_FMT "/%s. (" SYSTEMD_FEATURES ")",
1795 arg_action == ACTION_TEST ? " test" : "", getuid(), strna(t));
1796 }
1797
1798 *ret_first_boot = false;
1799 }
1800 }
1801
1802 static int initialize_runtime(
1803 bool skip_setup,
1804 struct rlimit *saved_rlimit_nofile,
1805 struct rlimit *saved_rlimit_memlock,
1806 const char **ret_error_message) {
1807
1808 int r;
1809
1810 assert(ret_error_message);
1811
1812 /* Sets up various runtime parameters. Many of these initializations are conditionalized:
1813 *
1814 * - Some only apply to --system instances
1815 * - Some only apply to --user instances
1816 * - Some only apply when we first start up, but not when we reexecute
1817 */
1818
1819 if (arg_action != ACTION_RUN)
1820 return 0;
1821
1822 if (arg_system) {
1823 /* Make sure we leave a core dump without panicing the kernel. */
1824 install_crash_handler();
1825
1826 if (!skip_setup) {
1827 r = mount_cgroup_controllers(arg_join_controllers);
1828 if (r < 0) {
1829 *ret_error_message = "Failed to mount cgroup hierarchies";
1830 return r;
1831 }
1832
1833 status_welcome();
1834 hostname_setup();
1835 machine_id_setup(NULL, arg_machine_id, NULL);
1836 loopback_setup();
1837 bump_unix_max_dgram_qlen();
1838 test_usr();
1839 write_container_id();
1840 }
1841
1842 if (arg_watchdog_device) {
1843 r = watchdog_set_device(arg_watchdog_device);
1844 if (r < 0)
1845 log_warning_errno(r, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device);
1846 }
1847
1848 if (arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY)
1849 watchdog_set_timeout(&arg_runtime_watchdog);
1850 }
1851
1852 if (arg_timer_slack_nsec != NSEC_INFINITY)
1853 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1854 log_warning_errno(errno, "Failed to adjust timer slack, ignoring: %m");
1855
1856 if (arg_system && !cap_test_all(arg_capability_bounding_set)) {
1857 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
1858 if (r < 0) {
1859 *ret_error_message = "Failed to drop capability bounding set of usermode helpers";
1860 return log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
1861 }
1862
1863 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
1864 if (r < 0) {
1865 *ret_error_message = "Failed to drop capability bounding set";
1866 return log_emergency_errno(r, "Failed to drop capability bounding set: %m");
1867 }
1868 }
1869
1870 if (arg_system && arg_no_new_privs) {
1871 if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) {
1872 *ret_error_message = "Failed to disable new privileges";
1873 return log_emergency_errno(errno, "Failed to disable new privileges: %m");
1874 }
1875 }
1876
1877 if (arg_syscall_archs) {
1878 r = enforce_syscall_archs(arg_syscall_archs);
1879 if (r < 0) {
1880 *ret_error_message = "Failed to set syscall architectures";
1881 return r;
1882 }
1883 }
1884
1885 if (!arg_system)
1886 /* Become reaper of our children */
1887 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
1888 log_warning_errno(errno, "Failed to make us a subreaper: %m");
1889
1890 if (arg_system) {
1891 /* Bump up RLIMIT_NOFILE for systemd itself */
1892 (void) bump_rlimit_nofile(saved_rlimit_nofile);
1893 (void) bump_rlimit_memlock(saved_rlimit_memlock);
1894 }
1895
1896 return 0;
1897 }
1898
1899 static int do_queue_default_job(
1900 Manager *m,
1901 const char **ret_error_message) {
1902
1903 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
1904 Job *default_unit_job;
1905 Unit *target = NULL;
1906 int r;
1907
1908 log_debug("Activating default unit: %s", arg_default_unit);
1909
1910 r = manager_load_startable_unit_or_warn(m, arg_default_unit, NULL, &target);
1911 if (r < 0) {
1912 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET);
1913
1914 r = manager_load_startable_unit_or_warn(m, SPECIAL_RESCUE_TARGET, NULL, &target);
1915 if (r < 0) {
1916 *ret_error_message = r == -ERFKILL ? "Rescue target masked"
1917 : "Failed to load rescue target";
1918 return r;
1919 }
1920 }
1921
1922 assert(target->load_state == UNIT_LOADED);
1923
1924 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, &error, &default_unit_job);
1925 if (r == -EPERM) {
1926 log_debug_errno(r, "Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1927
1928 sd_bus_error_free(&error);
1929
1930 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, &error, &default_unit_job);
1931 if (r < 0) {
1932 *ret_error_message = "Failed to start default target";
1933 return log_emergency_errno(r, "Failed to start default target: %s", bus_error_message(&error, r));
1934 }
1935
1936 } else if (r < 0) {
1937 *ret_error_message = "Failed to isolate default target";
1938 return log_emergency_errno(r, "Failed to isolate default target: %s", bus_error_message(&error, r));
1939 }
1940
1941 m->default_unit_job_id = default_unit_job->id;
1942
1943 return 0;
1944 }
1945
1946 static void free_arguments(void) {
1947
1948 /* Frees all arg_* variables, with the exception of arg_serialization */
1949 rlimit_free_all(arg_default_rlimit);
1950
1951 arg_default_unit = mfree(arg_default_unit);
1952 arg_confirm_spawn = mfree(arg_confirm_spawn);
1953 arg_join_controllers = strv_free_free(arg_join_controllers);
1954 arg_default_environment = strv_free(arg_default_environment);
1955 arg_syscall_archs = set_free(arg_syscall_archs);
1956 }
1957
1958 static int load_configuration(int argc, char **argv, const char **ret_error_message) {
1959 int r;
1960
1961 assert(ret_error_message);
1962
1963 arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U);
1964
1965 r = parse_config_file();
1966 if (r < 0) {
1967 *ret_error_message = "Failed to parse config file";
1968 return r;
1969 }
1970
1971 if (arg_system) {
1972 r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
1973 if (r < 0)
1974 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
1975 }
1976
1977 /* Note that this also parses bits from the kernel command line, including "debug". */
1978 log_parse_environment();
1979
1980 r = parse_argv(argc, argv);
1981 if (r < 0) {
1982 *ret_error_message = "Failed to parse commandline arguments";
1983 return r;
1984 }
1985
1986 /* Initialize default unit */
1987 if (!arg_default_unit) {
1988 arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET);
1989 if (!arg_default_unit) {
1990 *ret_error_message = "Failed to set default unit";
1991 return log_oom();
1992 }
1993 }
1994
1995 /* Initialize the show status setting if it hasn't been set explicitly yet */
1996 if (arg_show_status == _SHOW_STATUS_INVALID)
1997 arg_show_status = SHOW_STATUS_YES;
1998
1999 return 0;
2000 }
2001
2002 static int safety_checks(void) {
2003
2004 if (getpid_cached() == 1 &&
2005 arg_action != ACTION_RUN) {
2006 log_error("Unsupported execution mode while PID 1.");
2007 return -EPERM;
2008 }
2009
2010 if (getpid_cached() == 1 &&
2011 !arg_system) {
2012 log_error("Can't run --user mode as PID 1.");
2013 return -EPERM;
2014 }
2015
2016 if (arg_action == ACTION_RUN &&
2017 arg_system &&
2018 getpid_cached() != 1) {
2019 log_error("Can't run system mode unless PID 1.");
2020 return -EPERM;
2021 }
2022
2023 if (arg_action == ACTION_TEST &&
2024 geteuid() == 0) {
2025 log_error("Don't run test mode as root.");
2026 return -EPERM;
2027 }
2028
2029 if (!arg_system &&
2030 arg_action == ACTION_RUN &&
2031 sd_booted() <= 0) {
2032 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
2033 return -EOPNOTSUPP;
2034 }
2035
2036 if (!arg_system &&
2037 arg_action == ACTION_RUN &&
2038 !getenv("XDG_RUNTIME_DIR")) {
2039 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
2040 return -EUNATCH;
2041 }
2042
2043 if (arg_system &&
2044 arg_action == ACTION_RUN &&
2045 running_in_chroot() > 0) {
2046 log_error("Cannot be run in a chroot() environment.");
2047 return -EOPNOTSUPP;
2048 }
2049
2050 return 0;
2051 }
2052
2053 static int initialize_security(
2054 bool *loaded_policy,
2055 dual_timestamp *security_start_timestamp,
2056 dual_timestamp *security_finish_timestamp,
2057 const char **ret_error_message) {
2058
2059 int r;
2060
2061 assert(loaded_policy);
2062 assert(security_start_timestamp);
2063 assert(security_finish_timestamp);
2064 assert(ret_error_message);
2065
2066 dual_timestamp_get(security_start_timestamp);
2067
2068 r = mac_selinux_setup(loaded_policy);
2069 if (r < 0) {
2070 *ret_error_message = "Failed to load SELinux policy";
2071 return r;
2072 }
2073
2074 r = mac_smack_setup(loaded_policy);
2075 if (r < 0) {
2076 *ret_error_message = "Failed to load SMACK policy";
2077 return r;
2078 }
2079
2080 r = ima_setup();
2081 if (r < 0) {
2082 *ret_error_message = "Failed to load IMA policy";
2083 return r;
2084 }
2085
2086 dual_timestamp_get(security_finish_timestamp);
2087 return 0;
2088 }
2089
2090 static void test_summary(Manager *m) {
2091 assert(m);
2092
2093 printf("-> By units:\n");
2094 manager_dump_units(m, stdout, "\t");
2095
2096 printf("-> By jobs:\n");
2097 manager_dump_jobs(m, stdout, "\t");
2098 }
2099
2100 static int collect_fds(FDSet **ret_fds, const char **ret_error_message) {
2101 int r;
2102
2103 assert(ret_fds);
2104 assert(ret_error_message);
2105
2106 r = fdset_new_fill(ret_fds);
2107 if (r < 0) {
2108 *ret_error_message = "Failed to allocate fd set";
2109 return log_emergency_errno(r, "Failed to allocate fd set: %m");
2110 }
2111
2112 fdset_cloexec(*ret_fds, true);
2113
2114 if (arg_serialization)
2115 assert_se(fdset_remove(*ret_fds, fileno(arg_serialization)) >= 0);
2116
2117 return 0;
2118 }
2119
2120 static void setup_console_terminal(bool skip_setup) {
2121
2122 if (!arg_system)
2123 return;
2124
2125 /* Become a session leader if we aren't one yet. */
2126 (void) setsid();
2127
2128 /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2129 * tty. */
2130 (void) release_terminal();
2131
2132 /* Reset the console, but only if this is really init and we are freshly booted */
2133 if (getpid_cached() == 1 && !skip_setup)
2134 (void) console_setup();
2135 }
2136
2137 static bool early_skip_setup_check(int argc, char *argv[]) {
2138 bool found_deserialize = false;
2139 int i;
2140
2141 /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2142 * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2143 * anyway, even if in that case we also do deserialization. */
2144
2145 for (i = 1; i < argc; i++) {
2146 if (streq(argv[i], "--switched-root"))
2147 return false; /* If we switched root, don't skip the setup. */
2148 else if (streq(argv[i], "--deserialize"))
2149 found_deserialize = true;
2150 }
2151
2152 return found_deserialize; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2153 }
2154
2155 int main(int argc, char *argv[]) {
2156
2157 dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL, userspace_timestamp = DUAL_TIMESTAMP_NULL, kernel_timestamp = DUAL_TIMESTAMP_NULL,
2158 security_start_timestamp = DUAL_TIMESTAMP_NULL, security_finish_timestamp = DUAL_TIMESTAMP_NULL;
2159 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0), saved_rlimit_memlock = RLIMIT_MAKE_CONST((rlim_t) -1);
2160 bool skip_setup, loaded_policy = false, queue_default_job = false, first_boot = false, reexecute = false;
2161 char *switch_root_dir = NULL, *switch_root_init = NULL;
2162 usec_t before_startup, after_startup;
2163 static char systemd[] = "systemd";
2164 char timespan[FORMAT_TIMESPAN_MAX];
2165 const char *shutdown_verb = NULL, *error_message = NULL;
2166 int r, retval = EXIT_FAILURE;
2167 Manager *m = NULL;
2168 FDSet *fds = NULL;
2169
2170 /* SysV compatibility: redirect init → telinit */
2171 redirect_telinit(argc, argv);
2172
2173 /* Take timestamps early on */
2174 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
2175 dual_timestamp_get(&userspace_timestamp);
2176
2177 /* Figure out whether we need to do initialize the system, or if we already did that because we are
2178 * reexecuting */
2179 skip_setup = early_skip_setup_check(argc, argv);
2180
2181 /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2182 * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
2183 program_invocation_short_name = systemd;
2184 (void) prctl(PR_SET_NAME, systemd);
2185
2186 /* Save the original command line */
2187 saved_argv = argv;
2188 saved_argc = argc;
2189
2190 /* Make sure that if the user says "syslog" we actually log to the journal. */
2191 log_set_upgrade_syslog_to_journal(true);
2192
2193 if (getpid_cached() == 1) {
2194 /* When we run as PID 1 force system mode */
2195 arg_system = true;
2196
2197 /* Disable the umask logic */
2198 umask(0);
2199
2200 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2201 * activated yet (even though the log socket for it exists). */
2202 log_set_prohibit_ipc(true);
2203
2204 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2205 * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2206 * child process right before execve()'ing the actual binary, at a point in time where socket
2207 * activation stderr/stdout area already set up. */
2208 log_set_always_reopen_console(true);
2209
2210 if (detect_container() <= 0) {
2211
2212 /* Running outside of a container as PID 1 */
2213 log_set_target(LOG_TARGET_KMSG);
2214 log_open();
2215
2216 if (in_initrd())
2217 initrd_timestamp = userspace_timestamp;
2218
2219 if (!skip_setup) {
2220 r = mount_setup_early();
2221 if (r < 0) {
2222 error_message = "Failed to mount early API filesystems";
2223 goto finish;
2224 }
2225
2226 r = initialize_security(
2227 &loaded_policy,
2228 &security_start_timestamp,
2229 &security_finish_timestamp,
2230 &error_message);
2231 if (r < 0)
2232 goto finish;
2233 }
2234
2235 if (mac_selinux_init() < 0) {
2236 error_message = "Failed to initialize SELinux policy";
2237 goto finish;
2238 }
2239
2240 if (!skip_setup)
2241 initialize_clock();
2242
2243 /* Set the default for later on, but don't actually open the logs like this for now. Note that
2244 * if we are transitioning from the initrd there might still be journal fd open, and we
2245 * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2246 * elsewhere. */
2247 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
2248
2249 } else {
2250 /* Running inside a container, as PID 1 */
2251 log_set_target(LOG_TARGET_CONSOLE);
2252 log_open();
2253
2254 /* For later on, see above... */
2255 log_set_target(LOG_TARGET_JOURNAL);
2256
2257 /* clear the kernel timestamp,
2258 * because we are in a container */
2259 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2260 }
2261
2262 initialize_coredump(skip_setup);
2263
2264 r = fixup_environment();
2265 if (r < 0) {
2266 log_emergency_errno(r, "Failed to fix up PID 1 environment: %m");
2267 error_message = "Failed to fix up PID1 environment";
2268 goto finish;
2269 }
2270
2271 } else {
2272 /* Running as user instance */
2273 arg_system = false;
2274 log_set_target(LOG_TARGET_AUTO);
2275 log_open();
2276
2277 /* clear the kernel timestamp,
2278 * because we are not PID 1 */
2279 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2280 }
2281
2282 if (arg_system) {
2283 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2284 * they never log into the console. */
2285 log_show_color(colors_enabled());
2286
2287 r = make_null_stdio();
2288 if (r < 0)
2289 log_warning_errno(r, "Failed to redirect standard streams to /dev/null, ignoring: %m");
2290 }
2291
2292 /* Mount /proc, /sys and friends, so that /proc/cmdline and
2293 * /proc/$PID/fd is available. */
2294 if (getpid_cached() == 1) {
2295
2296 /* Load the kernel modules early. */
2297 if (!skip_setup)
2298 kmod_setup();
2299
2300 r = mount_setup(loaded_policy);
2301 if (r < 0) {
2302 error_message = "Failed to mount API filesystems";
2303 goto finish;
2304 }
2305 }
2306
2307 /* Reset all signal handlers. */
2308 (void) reset_all_signal_handlers();
2309 (void) ignore_signals(SIGNALS_IGNORE, -1);
2310
2311 r = load_configuration(argc, argv, &error_message);
2312 if (r < 0)
2313 goto finish;
2314
2315 r = safety_checks();
2316 if (r < 0)
2317 goto finish;
2318
2319 if (IN_SET(arg_action, ACTION_TEST, ACTION_HELP, ACTION_DUMP_CONFIGURATION_ITEMS, ACTION_DUMP_BUS_PROPERTIES))
2320 (void) pager_open(arg_no_pager, false);
2321
2322 if (arg_action != ACTION_RUN)
2323 skip_setup = true;
2324
2325 if (arg_action == ACTION_HELP) {
2326 retval = help() < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
2327 goto finish;
2328 } else if (arg_action == ACTION_VERSION) {
2329 retval = version();
2330 goto finish;
2331 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
2332 unit_dump_config_items(stdout);
2333 retval = EXIT_SUCCESS;
2334 goto finish;
2335 } else if (arg_action == ACTION_DUMP_BUS_PROPERTIES) {
2336 dump_bus_properties(stdout);
2337 retval = EXIT_SUCCESS;
2338 goto finish;
2339 }
2340
2341 assert_se(IN_SET(arg_action, ACTION_RUN, ACTION_TEST));
2342
2343 /* Move out of the way, so that we won't block unmounts */
2344 assert_se(chdir("/") == 0);
2345
2346 if (arg_action == ACTION_RUN) {
2347
2348 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2349 log_close();
2350
2351 /* Remember open file descriptors for later deserialization */
2352 r = collect_fds(&fds, &error_message);
2353 if (r < 0)
2354 goto finish;
2355
2356 /* Give up any control of the console, but make sure its initialized. */
2357 setup_console_terminal(skip_setup);
2358
2359 /* Open the logging devices, if possible and necessary */
2360 log_open();
2361 }
2362
2363 log_execution_mode(&first_boot);
2364
2365 r = initialize_runtime(skip_setup,
2366 &saved_rlimit_nofile,
2367 &saved_rlimit_memlock,
2368 &error_message);
2369 if (r < 0)
2370 goto finish;
2371
2372 r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER,
2373 arg_action == ACTION_TEST ? MANAGER_TEST_FULL : 0,
2374 &m);
2375 if (r < 0) {
2376 log_emergency_errno(r, "Failed to allocate manager object: %m");
2377 error_message = "Failed to allocate manager object";
2378 goto finish;
2379 }
2380
2381 m->timestamps[MANAGER_TIMESTAMP_KERNEL] = kernel_timestamp;
2382 m->timestamps[MANAGER_TIMESTAMP_INITRD] = initrd_timestamp;
2383 m->timestamps[MANAGER_TIMESTAMP_USERSPACE] = userspace_timestamp;
2384 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START)] = security_start_timestamp;
2385 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH)] = security_finish_timestamp;
2386
2387 set_manager_defaults(m);
2388 set_manager_settings(m);
2389 manager_set_first_boot(m, first_boot);
2390
2391 /* Remember whether we should queue the default job */
2392 queue_default_job = !arg_serialization || arg_switched_root;
2393
2394 before_startup = now(CLOCK_MONOTONIC);
2395
2396 r = manager_startup(m, arg_serialization, fds);
2397 if (r < 0) {
2398 log_error_errno(r, "Failed to fully start up daemon: %m");
2399 error_message = "Failed to start up manager";
2400 goto finish;
2401 }
2402
2403 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2404 fds = fdset_free(fds);
2405 arg_serialization = safe_fclose(arg_serialization);
2406
2407 if (queue_default_job) {
2408 r = do_queue_default_job(m, &error_message);
2409 if (r < 0)
2410 goto finish;
2411 }
2412
2413 after_startup = now(CLOCK_MONOTONIC);
2414
2415 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
2416 "Loaded units and determined initial transaction in %s.",
2417 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
2418
2419 if (arg_action == ACTION_TEST) {
2420 test_summary(m);
2421 retval = EXIT_SUCCESS;
2422 goto finish;
2423 }
2424
2425 (void) invoke_main_loop(m,
2426 &reexecute,
2427 &retval,
2428 &shutdown_verb,
2429 &fds,
2430 &switch_root_dir,
2431 &switch_root_init,
2432 &error_message);
2433
2434 finish:
2435 pager_close();
2436
2437 if (m) {
2438 arg_shutdown_watchdog = m->shutdown_watchdog;
2439 m = manager_free(m);
2440 }
2441
2442 free_arguments();
2443 mac_selinux_finish();
2444
2445 if (reexecute)
2446 do_reexecute(argc, argv,
2447 &saved_rlimit_nofile,
2448 &saved_rlimit_memlock,
2449 fds,
2450 switch_root_dir,
2451 switch_root_init,
2452 &error_message); /* This only returns if reexecution failed */
2453
2454 arg_serialization = safe_fclose(arg_serialization);
2455 fds = fdset_free(fds);
2456
2457 #if HAVE_VALGRIND_VALGRIND_H
2458 /* If we are PID 1 and running under valgrind, then let's exit
2459 * here explicitly. valgrind will only generate nice output on
2460 * exit(), not on exec(), hence let's do the former not the
2461 * latter here. */
2462 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND) {
2463 /* Cleanup watchdog_device strings for valgrind. We need them
2464 * in become_shutdown() so normally we cannot free them yet. */
2465 watchdog_free_device();
2466 arg_watchdog_device = mfree(arg_watchdog_device);
2467 return retval;
2468 }
2469 #endif
2470
2471 if (shutdown_verb) {
2472 r = become_shutdown(shutdown_verb, retval);
2473 log_error_errno(r, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
2474 error_message = "Failed to execute shutdown binary";
2475 }
2476
2477 watchdog_free_device();
2478 arg_watchdog_device = mfree(arg_watchdog_device);
2479
2480 if (getpid_cached() == 1) {
2481 if (error_message)
2482 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2483 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
2484 "%s, freezing.", error_message);
2485 freeze_or_reboot();
2486 }
2487
2488 return retval;
2489 }