]> git.ipfire.org Git - thirdparty/systemd.git/blob - src/core/main.c
projects / thirdparty / systemd.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
core: bump RLIMIT_NOFILE soft+hard limit for systemd itself in all cases
[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_early_core_pattern = NULL;
113 static char *arg_watchdog_device = NULL;
114 static char **arg_default_environment = NULL;
115 static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {};
116 static uint64_t arg_capability_bounding_set = CAP_ALL;
117 static bool arg_no_new_privs = false;
118 static nsec_t arg_timer_slack_nsec = NSEC_INFINITY;
119 static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
120 static Set* arg_syscall_archs = NULL;
121 static FILE* arg_serialization = NULL;
122 static bool arg_default_cpu_accounting = false;
123 static bool arg_default_io_accounting = false;
124 static bool arg_default_ip_accounting = false;
125 static bool arg_default_blockio_accounting = false;
126 static bool arg_default_memory_accounting = MEMORY_ACCOUNTING_DEFAULT;
127 static bool arg_default_tasks_accounting = true;
128 static uint64_t arg_default_tasks_max = UINT64_MAX;
129 static sd_id128_t arg_machine_id = {};
130 static EmergencyAction arg_cad_burst_action = EMERGENCY_ACTION_REBOOT_FORCE;
131
132 _noreturn_ static void freeze_or_reboot(void) {
133
134 if (arg_crash_reboot) {
135 log_notice("Rebooting in 10s...");
136 (void) sleep(10);
137
138 log_notice("Rebooting now...");
139 (void) reboot(RB_AUTOBOOT);
140 log_emergency_errno(errno, "Failed to reboot: %m");
141 }
142
143 log_emergency("Freezing execution.");
144 freeze();
145 }
146
147 _noreturn_ static void crash(int sig) {
148 struct sigaction sa;
149 pid_t pid;
150
151 if (getpid_cached() != 1)
152 /* Pass this on immediately, if this is not PID 1 */
153 (void) raise(sig);
154 else if (!arg_dump_core)
155 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
156 else {
157 sa = (struct sigaction) {
158 .sa_handler = nop_signal_handler,
159 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
160 };
161
162 /* We want to wait for the core process, hence let's enable SIGCHLD */
163 (void) sigaction(SIGCHLD, &sa, NULL);
164
165 pid = raw_clone(SIGCHLD);
166 if (pid < 0)
167 log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
168 else if (pid == 0) {
169 /* Enable default signal handler for core dump */
170
171 sa = (struct sigaction) {
172 .sa_handler = SIG_DFL,
173 };
174 (void) sigaction(sig, &sa, NULL);
175
176 /* Don't limit the coredump size */
177 (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
178
179 /* Just to be sure... */
180 (void) chdir("/");
181
182 /* Raise the signal again */
183 pid = raw_getpid();
184 (void) kill(pid, sig); /* raise() would kill the parent */
185
186 assert_not_reached("We shouldn't be here...");
187 _exit(EXIT_FAILURE);
188 } else {
189 siginfo_t status;
190 int r;
191
192 /* Order things nicely. */
193 r = wait_for_terminate(pid, &status);
194 if (r < 0)
195 log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
196 else if (status.si_code != CLD_DUMPED)
197 log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
198 signal_to_string(sig),
199 pid, sigchld_code_to_string(status.si_code),
200 status.si_status,
201 strna(status.si_code == CLD_EXITED
202 ? exit_status_to_string(status.si_status, EXIT_STATUS_MINIMAL)
203 : signal_to_string(status.si_status)));
204 else
205 log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
206 }
207 }
208
209 if (arg_crash_chvt >= 0)
210 (void) chvt(arg_crash_chvt);
211
212 sa = (struct sigaction) {
213 .sa_handler = SIG_IGN,
214 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
215 };
216
217 /* Let the kernel reap children for us */
218 (void) sigaction(SIGCHLD, &sa, NULL);
219
220 if (arg_crash_shell) {
221 log_notice("Executing crash shell in 10s...");
222 (void) sleep(10);
223
224 pid = raw_clone(SIGCHLD);
225 if (pid < 0)
226 log_emergency_errno(errno, "Failed to fork off crash shell: %m");
227 else if (pid == 0) {
228 (void) setsid();
229 (void) make_console_stdio();
230 (void) execle("/bin/sh", "/bin/sh", NULL, environ);
231
232 log_emergency_errno(errno, "execle() failed: %m");
233 _exit(EXIT_FAILURE);
234 } else {
235 log_info("Spawned crash shell as PID "PID_FMT".", pid);
236 (void) wait_for_terminate(pid, NULL);
237 }
238 }
239
240 freeze_or_reboot();
241 }
242
243 static void install_crash_handler(void) {
244 static const struct sigaction sa = {
245 .sa_handler = crash,
246 .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
247 };
248 int r;
249
250 /* We ignore the return value here, since, we don't mind if we
251 * cannot set up a crash handler */
252 r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
253 if (r < 0)
254 log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
255 }
256
257 static int console_setup(void) {
258 _cleanup_close_ int tty_fd = -1;
259 int r;
260
261 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
262 if (tty_fd < 0)
263 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
264
265 /* We don't want to force text mode. plymouth may be showing
266 * pictures already from initrd. */
267 r = reset_terminal_fd(tty_fd, false);
268 if (r < 0)
269 return log_error_errno(r, "Failed to reset /dev/console: %m");
270
271 return 0;
272 }
273
274 static int parse_crash_chvt(const char *value) {
275 int b;
276
277 if (safe_atoi(value, &arg_crash_chvt) >= 0)
278 return 0;
279
280 b = parse_boolean(value);
281 if (b < 0)
282 return b;
283
284 if (b > 0)
285 arg_crash_chvt = 0; /* switch to where kmsg goes */
286 else
287 arg_crash_chvt = -1; /* turn off switching */
288
289 return 0;
290 }
291
292 static int parse_confirm_spawn(const char *value, char **console) {
293 char *s;
294 int r;
295
296 r = value ? parse_boolean(value) : 1;
297 if (r == 0) {
298 *console = NULL;
299 return 0;
300 }
301
302 if (r > 0) /* on with default tty */
303 s = strdup("/dev/console");
304 else if (is_path(value)) /* on with fully qualified path */
305 s = strdup(value);
306 else /* on with only a tty file name, not a fully qualified path */
307 s = strjoin("/dev/", value);
308 if (!s)
309 return -ENOMEM;
310
311 *console = s;
312 return 0;
313 }
314
315 static int set_machine_id(const char *m) {
316 sd_id128_t t;
317 assert(m);
318
319 if (sd_id128_from_string(m, &t) < 0)
320 return -EINVAL;
321
322 if (sd_id128_is_null(t))
323 return -EINVAL;
324
325 arg_machine_id = t;
326 return 0;
327 }
328
329 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
330
331 int r;
332
333 assert(key);
334
335 if (STR_IN_SET(key, "systemd.unit", "rd.systemd.unit")) {
336
337 if (proc_cmdline_value_missing(key, value))
338 return 0;
339
340 if (!unit_name_is_valid(value, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE))
341 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key, value);
342 else if (in_initrd() == !!startswith(key, "rd.")) {
343 if (free_and_strdup(&arg_default_unit, value) < 0)
344 return log_oom();
345 }
346
347 } else if (proc_cmdline_key_streq(key, "systemd.dump_core")) {
348
349 r = value ? parse_boolean(value) : true;
350 if (r < 0)
351 log_warning("Failed to parse dump core switch %s. Ignoring.", value);
352 else
353 arg_dump_core = r;
354
355 } else if (proc_cmdline_key_streq(key, "systemd.early_core_pattern")) {
356
357 if (proc_cmdline_value_missing(key, value))
358 return 0;
359
360 if (path_is_absolute(value))
361 (void) parse_path_argument_and_warn(value, false, &arg_early_core_pattern);
362 else
363 log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value);
364
365 } else if (proc_cmdline_key_streq(key, "systemd.crash_chvt")) {
366
367 if (!value)
368 arg_crash_chvt = 0; /* turn on */
369 else if (parse_crash_chvt(value) < 0)
370 log_warning("Failed to parse crash chvt switch %s. Ignoring.", value);
371
372 } else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) {
373
374 r = value ? parse_boolean(value) : true;
375 if (r < 0)
376 log_warning("Failed to parse crash shell switch %s. Ignoring.", value);
377 else
378 arg_crash_shell = r;
379
380 } else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) {
381
382 r = value ? parse_boolean(value) : true;
383 if (r < 0)
384 log_warning("Failed to parse crash reboot switch %s. Ignoring.", value);
385 else
386 arg_crash_reboot = r;
387
388 } else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) {
389 char *s;
390
391 r = parse_confirm_spawn(value, &s);
392 if (r < 0)
393 log_warning_errno(r, "Failed to parse confirm_spawn switch %s. Ignoring.", value);
394 else {
395 free(arg_confirm_spawn);
396 arg_confirm_spawn = s;
397 }
398
399 } else if (proc_cmdline_key_streq(key, "systemd.service_watchdogs")) {
400
401 r = value ? parse_boolean(value) : true;
402 if (r < 0)
403 log_warning("Failed to parse service watchdog switch %s. Ignoring.", value);
404 else
405 arg_service_watchdogs = r;
406
407 } else if (proc_cmdline_key_streq(key, "systemd.show_status")) {
408
409 if (value) {
410 r = parse_show_status(value, &arg_show_status);
411 if (r < 0)
412 log_warning("Failed to parse show status switch %s. Ignoring.", value);
413 } else
414 arg_show_status = SHOW_STATUS_YES;
415
416 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) {
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 output switch %s. Ignoring.", value);
424 else
425 arg_default_std_output = r;
426
427 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) {
428
429 if (proc_cmdline_value_missing(key, value))
430 return 0;
431
432 r = exec_output_from_string(value);
433 if (r < 0)
434 log_warning("Failed to parse default standard error switch %s. Ignoring.", value);
435 else
436 arg_default_std_error = r;
437
438 } else if (streq(key, "systemd.setenv")) {
439
440 if (proc_cmdline_value_missing(key, value))
441 return 0;
442
443 if (env_assignment_is_valid(value)) {
444 char **env;
445
446 env = strv_env_set(arg_default_environment, value);
447 if (!env)
448 return log_oom();
449
450 arg_default_environment = env;
451 } else
452 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
453
454 } else if (proc_cmdline_key_streq(key, "systemd.machine_id")) {
455
456 if (proc_cmdline_value_missing(key, value))
457 return 0;
458
459 r = set_machine_id(value);
460 if (r < 0)
461 log_warning("MachineID '%s' is not valid. Ignoring.", value);
462
463 } else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) {
464
465 if (proc_cmdline_value_missing(key, value))
466 return 0;
467
468 r = parse_sec(value, &arg_default_timeout_start_usec);
469 if (r < 0)
470 log_warning_errno(r, "Failed to parse default start timeout: %s, ignoring.", value);
471
472 if (arg_default_timeout_start_usec <= 0)
473 arg_default_timeout_start_usec = USEC_INFINITY;
474
475 } else if (proc_cmdline_key_streq(key, "systemd.watchdog_device")) {
476
477 if (proc_cmdline_value_missing(key, value))
478 return 0;
479
480 (void) parse_path_argument_and_warn(value, false, &arg_watchdog_device);
481
482 } else if (streq(key, "quiet") && !value) {
483
484 if (arg_show_status == _SHOW_STATUS_INVALID)
485 arg_show_status = SHOW_STATUS_AUTO;
486
487 } else if (streq(key, "debug") && !value) {
488
489 /* Note that log_parse_environment() handles 'debug'
490 * too, and sets the log level to LOG_DEBUG. */
491
492 if (detect_container() > 0)
493 log_set_target(LOG_TARGET_CONSOLE);
494
495 } else if (!value) {
496 const char *target;
497
498 /* SysV compatibility */
499 target = runlevel_to_target(key);
500 if (target)
501 return free_and_strdup(&arg_default_unit, target);
502 }
503
504 return 0;
505 }
506
507 #define DEFINE_SETTER(name, func, descr) \
508 static int name(const char *unit, \
509 const char *filename, \
510 unsigned line, \
511 const char *section, \
512 unsigned section_line, \
513 const char *lvalue, \
514 int ltype, \
515 const char *rvalue, \
516 void *data, \
517 void *userdata) { \
518 \
519 int r; \
520 \
521 assert(filename); \
522 assert(lvalue); \
523 assert(rvalue); \
524 \
525 r = func(rvalue); \
526 if (r < 0) \
527 log_syntax(unit, LOG_ERR, filename, line, r, \
528 "Invalid " descr "'%s': %m", \
529 rvalue); \
530 \
531 return 0; \
532 }
533
534 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level");
535 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target");
536 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" );
537 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location");
538
539 static int config_parse_cpu_affinity2(
540 const char *unit,
541 const char *filename,
542 unsigned line,
543 const char *section,
544 unsigned section_line,
545 const char *lvalue,
546 int ltype,
547 const char *rvalue,
548 void *data,
549 void *userdata) {
550
551 _cleanup_cpu_free_ cpu_set_t *c = NULL;
552 int ncpus;
553
554 ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue);
555 if (ncpus < 0)
556 return ncpus;
557
558 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
559 log_warning_errno(errno, "Failed to set CPU affinity: %m");
560
561 return 0;
562 }
563
564 static int config_parse_show_status(
565 const char* unit,
566 const char *filename,
567 unsigned line,
568 const char *section,
569 unsigned section_line,
570 const char *lvalue,
571 int ltype,
572 const char *rvalue,
573 void *data,
574 void *userdata) {
575
576 int k;
577 ShowStatus *b = data;
578
579 assert(filename);
580 assert(lvalue);
581 assert(rvalue);
582 assert(data);
583
584 k = parse_show_status(rvalue, b);
585 if (k < 0) {
586 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
587 return 0;
588 }
589
590 return 0;
591 }
592
593 static int config_parse_output_restricted(
594 const char* unit,
595 const char *filename,
596 unsigned line,
597 const char *section,
598 unsigned section_line,
599 const char *lvalue,
600 int ltype,
601 const char *rvalue,
602 void *data,
603 void *userdata) {
604
605 ExecOutput t, *eo = data;
606
607 assert(filename);
608 assert(lvalue);
609 assert(rvalue);
610 assert(data);
611
612 t = exec_output_from_string(rvalue);
613 if (t < 0) {
614 log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse output type, ignoring: %s", rvalue);
615 return 0;
616 }
617
618 if (IN_SET(t, EXEC_OUTPUT_SOCKET, EXEC_OUTPUT_NAMED_FD, EXEC_OUTPUT_FILE, EXEC_OUTPUT_FILE_APPEND)) {
619 log_syntax(unit, LOG_ERR, filename, line, 0, "Standard output types socket, fd:, file:, append: are not supported as defaults, ignoring: %s", rvalue);
620 return 0;
621 }
622
623 *eo = t;
624 return 0;
625 }
626
627 static int config_parse_crash_chvt(
628 const char* unit,
629 const char *filename,
630 unsigned line,
631 const char *section,
632 unsigned section_line,
633 const char *lvalue,
634 int ltype,
635 const char *rvalue,
636 void *data,
637 void *userdata) {
638
639 int r;
640
641 assert(filename);
642 assert(lvalue);
643 assert(rvalue);
644
645 r = parse_crash_chvt(rvalue);
646 if (r < 0) {
647 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
648 return 0;
649 }
650
651 return 0;
652 }
653
654 static int parse_config_file(void) {
655
656 const ConfigTableItem items[] = {
657 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
658 { "Manager", "LogTarget", config_parse_target, 0, NULL },
659 { "Manager", "LogColor", config_parse_color, 0, NULL },
660 { "Manager", "LogLocation", config_parse_location, 0, NULL },
661 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
662 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL },
663 { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL },
664 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
665 { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
666 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
667 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
668 { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
669 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
670 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
671 { "Manager", "WatchdogDevice", config_parse_path, 0, &arg_watchdog_device },
672 { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
673 { "Manager", "NoNewPrivileges", config_parse_bool, 0, &arg_no_new_privs },
674 #if HAVE_SECCOMP
675 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
676 #endif
677 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
678 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
679 { "Manager", "DefaultStandardOutput", config_parse_output_restricted,0, &arg_default_std_output },
680 { "Manager", "DefaultStandardError", config_parse_output_restricted,0, &arg_default_std_error },
681 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
682 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
683 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
684 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */
685 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec, 0, &arg_default_start_limit_interval },
686 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
687 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
688 { "Manager", "DefaultLimitCPU", config_parse_rlimit, RLIMIT_CPU, arg_default_rlimit },
689 { "Manager", "DefaultLimitFSIZE", config_parse_rlimit, RLIMIT_FSIZE, arg_default_rlimit },
690 { "Manager", "DefaultLimitDATA", config_parse_rlimit, RLIMIT_DATA, arg_default_rlimit },
691 { "Manager", "DefaultLimitSTACK", config_parse_rlimit, RLIMIT_STACK, arg_default_rlimit },
692 { "Manager", "DefaultLimitCORE", config_parse_rlimit, RLIMIT_CORE, arg_default_rlimit },
693 { "Manager", "DefaultLimitRSS", config_parse_rlimit, RLIMIT_RSS, arg_default_rlimit },
694 { "Manager", "DefaultLimitNOFILE", config_parse_rlimit, RLIMIT_NOFILE, arg_default_rlimit },
695 { "Manager", "DefaultLimitAS", config_parse_rlimit, RLIMIT_AS, arg_default_rlimit },
696 { "Manager", "DefaultLimitNPROC", config_parse_rlimit, RLIMIT_NPROC, arg_default_rlimit },
697 { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit, RLIMIT_MEMLOCK, arg_default_rlimit },
698 { "Manager", "DefaultLimitLOCKS", config_parse_rlimit, RLIMIT_LOCKS, arg_default_rlimit },
699 { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit, RLIMIT_SIGPENDING, arg_default_rlimit },
700 { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit, RLIMIT_MSGQUEUE, arg_default_rlimit },
701 { "Manager", "DefaultLimitNICE", config_parse_rlimit, RLIMIT_NICE, arg_default_rlimit },
702 { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit, RLIMIT_RTPRIO, arg_default_rlimit },
703 { "Manager", "DefaultLimitRTTIME", config_parse_rlimit, RLIMIT_RTTIME, arg_default_rlimit },
704 { "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_default_cpu_accounting },
705 { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting },
706 { "Manager", "DefaultIPAccounting", config_parse_bool, 0, &arg_default_ip_accounting },
707 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
708 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
709 { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
710 { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max },
711 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action, 0, &arg_cad_burst_action },
712 {}
713 };
714
715 const char *fn, *conf_dirs_nulstr;
716
717 fn = arg_system ?
718 PKGSYSCONFDIR "/system.conf" :
719 PKGSYSCONFDIR "/user.conf";
720
721 conf_dirs_nulstr = arg_system ?
722 CONF_PATHS_NULSTR("systemd/system.conf.d") :
723 CONF_PATHS_NULSTR("systemd/user.conf.d");
724
725 (void) config_parse_many_nulstr(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, CONFIG_PARSE_WARN, NULL);
726
727 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
728 * like everywhere else. */
729 if (arg_default_timeout_start_usec <= 0)
730 arg_default_timeout_start_usec = USEC_INFINITY;
731 if (arg_default_timeout_stop_usec <= 0)
732 arg_default_timeout_stop_usec = USEC_INFINITY;
733
734 return 0;
735 }
736
737 static void set_manager_defaults(Manager *m) {
738
739 assert(m);
740
741 /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
742 * affect the manager itself, but are just what newly allocated units will have set if they haven't set
743 * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
744
745 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
746 m->default_std_output = arg_default_std_output;
747 m->default_std_error = arg_default_std_error;
748 m->default_timeout_start_usec = arg_default_timeout_start_usec;
749 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
750 m->default_restart_usec = arg_default_restart_usec;
751 m->default_start_limit_interval = arg_default_start_limit_interval;
752 m->default_start_limit_burst = arg_default_start_limit_burst;
753 m->default_cpu_accounting = arg_default_cpu_accounting;
754 m->default_io_accounting = arg_default_io_accounting;
755 m->default_ip_accounting = arg_default_ip_accounting;
756 m->default_blockio_accounting = arg_default_blockio_accounting;
757 m->default_memory_accounting = arg_default_memory_accounting;
758 m->default_tasks_accounting = arg_default_tasks_accounting;
759 m->default_tasks_max = arg_default_tasks_max;
760
761 manager_set_default_rlimits(m, arg_default_rlimit);
762 manager_environment_add(m, NULL, arg_default_environment);
763 }
764
765 static void set_manager_settings(Manager *m) {
766
767 assert(m);
768
769 /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
770 * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
771
772 m->confirm_spawn = arg_confirm_spawn;
773 m->service_watchdogs = arg_service_watchdogs;
774 m->runtime_watchdog = arg_runtime_watchdog;
775 m->shutdown_watchdog = arg_shutdown_watchdog;
776 m->cad_burst_action = arg_cad_burst_action;
777
778 manager_set_show_status(m, arg_show_status);
779 }
780
781 static int parse_argv(int argc, char *argv[]) {
782 enum {
783 ARG_LOG_LEVEL = 0x100,
784 ARG_LOG_TARGET,
785 ARG_LOG_COLOR,
786 ARG_LOG_LOCATION,
787 ARG_UNIT,
788 ARG_SYSTEM,
789 ARG_USER,
790 ARG_TEST,
791 ARG_NO_PAGER,
792 ARG_VERSION,
793 ARG_DUMP_CONFIGURATION_ITEMS,
794 ARG_DUMP_BUS_PROPERTIES,
795 ARG_DUMP_CORE,
796 ARG_CRASH_CHVT,
797 ARG_CRASH_SHELL,
798 ARG_CRASH_REBOOT,
799 ARG_CONFIRM_SPAWN,
800 ARG_SHOW_STATUS,
801 ARG_DESERIALIZE,
802 ARG_SWITCHED_ROOT,
803 ARG_DEFAULT_STD_OUTPUT,
804 ARG_DEFAULT_STD_ERROR,
805 ARG_MACHINE_ID,
806 ARG_SERVICE_WATCHDOGS,
807 };
808
809 static const struct option options[] = {
810 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
811 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
812 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
813 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
814 { "unit", required_argument, NULL, ARG_UNIT },
815 { "system", no_argument, NULL, ARG_SYSTEM },
816 { "user", no_argument, NULL, ARG_USER },
817 { "test", no_argument, NULL, ARG_TEST },
818 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
819 { "help", no_argument, NULL, 'h' },
820 { "version", no_argument, NULL, ARG_VERSION },
821 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
822 { "dump-bus-properties", no_argument, NULL, ARG_DUMP_BUS_PROPERTIES },
823 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
824 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
825 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
826 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
827 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
828 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
829 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
830 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
831 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
832 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
833 { "machine-id", required_argument, NULL, ARG_MACHINE_ID },
834 { "service-watchdogs", required_argument, NULL, ARG_SERVICE_WATCHDOGS },
835 {}
836 };
837
838 int c, r;
839
840 assert(argc >= 1);
841 assert(argv);
842
843 if (getpid_cached() == 1)
844 opterr = 0;
845
846 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
847
848 switch (c) {
849
850 case ARG_LOG_LEVEL:
851 r = log_set_max_level_from_string(optarg);
852 if (r < 0)
853 return log_error_errno(r, "Failed to parse log level \"%s\": %m", optarg);
854
855 break;
856
857 case ARG_LOG_TARGET:
858 r = log_set_target_from_string(optarg);
859 if (r < 0)
860 return log_error_errno(r, "Failed to parse log target \"%s\": %m", optarg);
861
862 break;
863
864 case ARG_LOG_COLOR:
865
866 if (optarg) {
867 r = log_show_color_from_string(optarg);
868 if (r < 0)
869 return log_error_errno(r, "Failed to parse log color setting \"%s\": %m",
870 optarg);
871 } else
872 log_show_color(true);
873
874 break;
875
876 case ARG_LOG_LOCATION:
877 if (optarg) {
878 r = log_show_location_from_string(optarg);
879 if (r < 0)
880 return log_error_errno(r, "Failed to parse log location setting \"%s\": %m",
881 optarg);
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 return log_error_errno(r, "Failed to parse default standard output setting \"%s\": %m",
891 optarg);
892 arg_default_std_output = r;
893 break;
894
895 case ARG_DEFAULT_STD_ERROR:
896 r = exec_output_from_string(optarg);
897 if (r < 0)
898 return log_error_errno(r, "Failed to parse default standard error output setting \"%s\": %m",
899 optarg);
900 arg_default_std_error = r;
901 break;
902
903 case ARG_UNIT:
904 r = free_and_strdup(&arg_default_unit, optarg);
905 if (r < 0)
906 return log_error_errno(r, "Failed to set default unit \"%s\": %m", optarg);
907
908 break;
909
910 case ARG_SYSTEM:
911 arg_system = true;
912 break;
913
914 case ARG_USER:
915 arg_system = false;
916 break;
917
918 case ARG_TEST:
919 arg_action = ACTION_TEST;
920 break;
921
922 case ARG_NO_PAGER:
923 arg_no_pager = true;
924 break;
925
926 case ARG_VERSION:
927 arg_action = ACTION_VERSION;
928 break;
929
930 case ARG_DUMP_CONFIGURATION_ITEMS:
931 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
932 break;
933
934 case ARG_DUMP_BUS_PROPERTIES:
935 arg_action = ACTION_DUMP_BUS_PROPERTIES;
936 break;
937
938 case ARG_DUMP_CORE:
939 if (!optarg)
940 arg_dump_core = true;
941 else {
942 r = parse_boolean(optarg);
943 if (r < 0)
944 return log_error_errno(r, "Failed to parse dump core boolean: \"%s\": %m",
945 optarg);
946 arg_dump_core = r;
947 }
948 break;
949
950 case ARG_CRASH_CHVT:
951 r = parse_crash_chvt(optarg);
952 if (r < 0)
953 return log_error_errno(r, "Failed to parse crash virtual terminal index: \"%s\": %m",
954 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\": %m",
964 optarg);
965 arg_crash_shell = r;
966 }
967 break;
968
969 case ARG_CRASH_REBOOT:
970 if (!optarg)
971 arg_crash_reboot = true;
972 else {
973 r = parse_boolean(optarg);
974 if (r < 0)
975 return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
976 optarg);
977 arg_crash_reboot = r;
978 }
979 break;
980
981 case ARG_CONFIRM_SPAWN:
982 arg_confirm_spawn = mfree(arg_confirm_spawn);
983
984 r = parse_confirm_spawn(optarg, &arg_confirm_spawn);
985 if (r < 0)
986 return log_error_errno(r, "Failed to parse confirm spawn option: \"%s\": %m",
987 optarg);
988 break;
989
990 case ARG_SERVICE_WATCHDOGS:
991 r = parse_boolean(optarg);
992 if (r < 0)
993 return log_error_errno(r, "Failed to parse service watchdogs boolean: \"%s\": %m",
994 optarg);
995 arg_service_watchdogs = r;
996 break;
997
998 case ARG_SHOW_STATUS:
999 if (optarg) {
1000 r = parse_show_status(optarg, &arg_show_status);
1001 if (r < 0)
1002 return log_error_errno(r, "Failed to parse show status boolean: \"%s\": %m",
1003 optarg);
1004 } else
1005 arg_show_status = SHOW_STATUS_YES;
1006 break;
1007
1008 case ARG_DESERIALIZE: {
1009 int fd;
1010 FILE *f;
1011
1012 r = safe_atoi(optarg, &fd);
1013 if (r < 0)
1014 log_error_errno(r, "Failed to parse deserialize option \"%s\": %m", optarg);
1015 if (fd < 0) {
1016 log_error("Invalid deserialize fd: %d", fd);
1017 return -EINVAL;
1018 }
1019
1020 (void) fd_cloexec(fd, true);
1021
1022 f = fdopen(fd, "r");
1023 if (!f)
1024 return log_error_errno(errno, "Failed to open serialization fd %d: %m", fd);
1025
1026 safe_fclose(arg_serialization);
1027 arg_serialization = f;
1028
1029 break;
1030 }
1031
1032 case ARG_SWITCHED_ROOT:
1033 arg_switched_root = true;
1034 break;
1035
1036 case ARG_MACHINE_ID:
1037 r = set_machine_id(optarg);
1038 if (r < 0)
1039 return log_error_errno(r, "MachineID '%s' is not valid: %m", optarg);
1040 break;
1041
1042 case 'h':
1043 arg_action = ACTION_HELP;
1044 break;
1045
1046 case 'D':
1047 log_set_max_level(LOG_DEBUG);
1048 break;
1049
1050 case 'b':
1051 case 's':
1052 case 'z':
1053 /* Just to eat away the sysvinit kernel
1054 * cmdline args without getopt() error
1055 * messages that we'll parse in
1056 * parse_proc_cmdline_word() or ignore. */
1057
1058 case '?':
1059 if (getpid_cached() != 1)
1060 return -EINVAL;
1061 else
1062 return 0;
1063
1064 default:
1065 assert_not_reached("Unhandled option code.");
1066 }
1067
1068 if (optind < argc && getpid_cached() != 1) {
1069 /* Hmm, when we aren't run as init system
1070 * let's complain about excess arguments */
1071
1072 log_error("Excess arguments.");
1073 return -EINVAL;
1074 }
1075
1076 return 0;
1077 }
1078
1079 static int help(void) {
1080 _cleanup_free_ char *link = NULL;
1081 int r;
1082
1083 r = terminal_urlify_man("systemd", "1", &link);
1084 if (r < 0)
1085 return log_oom();
1086
1087 printf("%s [OPTIONS...]\n\n"
1088 "Starts up and maintains the system or user services.\n\n"
1089 " -h --help Show this help\n"
1090 " --version Show version\n"
1091 " --test Determine startup sequence, dump it and exit\n"
1092 " --no-pager Do not pipe output into a pager\n"
1093 " --dump-configuration-items Dump understood unit configuration items\n"
1094 " --dump-bus-properties Dump exposed bus properties\n"
1095 " --unit=UNIT Set default unit\n"
1096 " --system Run a system instance, even if PID != 1\n"
1097 " --user Run a user instance\n"
1098 " --dump-core[=BOOL] Dump core on crash\n"
1099 " --crash-vt=NR Change to specified VT on crash\n"
1100 " --crash-reboot[=BOOL] Reboot on crash\n"
1101 " --crash-shell[=BOOL] Run shell on crash\n"
1102 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1103 " --show-status[=BOOL] Show status updates on the console during bootup\n"
1104 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1105 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1106 " --log-color[=BOOL] Highlight important log messages\n"
1107 " --log-location[=BOOL] Include code location in log messages\n"
1108 " --default-standard-output= Set default standard output for services\n"
1109 " --default-standard-error= Set default standard error output for services\n"
1110 "\nSee the %s for details.\n"
1111 , program_invocation_short_name
1112 , link
1113 );
1114
1115 return 0;
1116 }
1117
1118 static int prepare_reexecute(
1119 Manager *m,
1120 FILE **ret_f,
1121 FDSet **ret_fds,
1122 bool switching_root) {
1123
1124 _cleanup_fdset_free_ FDSet *fds = NULL;
1125 _cleanup_fclose_ FILE *f = NULL;
1126 int r;
1127
1128 assert(m);
1129 assert(ret_f);
1130 assert(ret_fds);
1131
1132 r = manager_open_serialization(m, &f);
1133 if (r < 0)
1134 return log_error_errno(r, "Failed to create serialization file: %m");
1135
1136 /* Make sure nothing is really destructed when we shut down */
1137 m->n_reloading++;
1138 bus_manager_send_reloading(m, true);
1139
1140 fds = fdset_new();
1141 if (!fds)
1142 return log_oom();
1143
1144 r = manager_serialize(m, f, fds, switching_root);
1145 if (r < 0)
1146 return log_error_errno(r, "Failed to serialize state: %m");
1147
1148 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1149 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1150
1151 r = fd_cloexec(fileno(f), false);
1152 if (r < 0)
1153 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1154
1155 r = fdset_cloexec(fds, false);
1156 if (r < 0)
1157 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1158
1159 *ret_f = TAKE_PTR(f);
1160 *ret_fds = TAKE_PTR(fds);
1161
1162 return 0;
1163 }
1164
1165 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1166 int r, nr;
1167
1168 assert(saved_rlimit);
1169
1170 /* Save the original RLIMIT_NOFILE so that we can reset it later when transitioning from the initrd to the main
1171 * systemd or suchlike. */
1172 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1173 return log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1174
1175 /* Get the underlying absolute limit the kernel enforces */
1176 nr = read_nr_open();
1177
1178 /* Make sure forked processes get limits based on the original kernel setting */
1179 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1180 struct rlimit *rl;
1181
1182 rl = newdup(struct rlimit, saved_rlimit, 1);
1183 if (!rl)
1184 return log_oom();
1185
1186 /* Bump the hard limit for system services to a substantially higher value. The default hard limit
1187 * current kernels set is pretty low (4K), mostly for historical reasons. According to kernel
1188 * developers, the fd handling in recent kernels has been optimized substantially enough, so that we
1189 * can bump the limit now, without paying too high a price in memory or performance. Note however that
1190 * we only bump the hard limit, not the soft limit. That's because select() works the way it works, and
1191 * chokes on fds >= 1024. If we'd bump the soft limit globally, it might accidentally happen to
1192 * unexpecting programs that they get fds higher than what they can process using select(). By only
1193 * bumping the hard limit but leaving the low limit as it is we avoid this pitfall: programs that are
1194 * written by folks aware of the select() problem in mind (and thus use poll()/epoll instead of
1195 * select(), the way everybody should) can explicitly opt into high fds by bumping their soft limit
1196 * beyond 1024, to the hard limit we pass. */
1197 if (arg_system)
1198 rl->rlim_max = MIN((rlim_t) nr, MAX(rl->rlim_max, (rlim_t) HIGH_RLIMIT_NOFILE));
1199
1200 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1201 }
1202
1203 /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for
1204 * both hard and soft. */
1205 r = setrlimit_closest(RLIMIT_NOFILE, &RLIMIT_MAKE_CONST(nr));
1206 if (r < 0)
1207 return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1208
1209 return 0;
1210 }
1211
1212 static int bump_rlimit_memlock(struct rlimit *saved_rlimit) {
1213 int r;
1214
1215 assert(saved_rlimit);
1216
1217 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK which should
1218 * normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's bump
1219 * the value high enough for our user. */
1220
1221 if (getrlimit(RLIMIT_MEMLOCK, saved_rlimit) < 0)
1222 return log_warning_errno(errno, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1223
1224 r = setrlimit_closest(RLIMIT_MEMLOCK, &RLIMIT_MAKE_CONST(1024ULL*1024ULL*64ULL));
1225 if (r < 0)
1226 return log_warning_errno(r, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1227
1228 return 0;
1229 }
1230
1231 static void test_usr(void) {
1232
1233 /* Check that /usr is not a separate fs */
1234
1235 if (dir_is_empty("/usr") <= 0)
1236 return;
1237
1238 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1239 "Some things will probably break (sometimes even silently) in mysterious ways. "
1240 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1241 }
1242
1243 static int enforce_syscall_archs(Set *archs) {
1244 #if HAVE_SECCOMP
1245 int r;
1246
1247 if (!is_seccomp_available())
1248 return 0;
1249
1250 r = seccomp_restrict_archs(arg_syscall_archs);
1251 if (r < 0)
1252 return log_error_errno(r, "Failed to enforce system call architecture restrication: %m");
1253 #endif
1254 return 0;
1255 }
1256
1257 static int status_welcome(void) {
1258 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1259 int r;
1260
1261 if (IN_SET(arg_show_status, SHOW_STATUS_NO, SHOW_STATUS_AUTO))
1262 return 0;
1263
1264 r = parse_os_release(NULL,
1265 "PRETTY_NAME", &pretty_name,
1266 "ANSI_COLOR", &ansi_color,
1267 NULL);
1268 if (r < 0)
1269 log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r,
1270 "Failed to read os-release file, ignoring: %m");
1271
1272 if (log_get_show_color())
1273 return status_printf(NULL, false, false,
1274 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1275 isempty(ansi_color) ? "1" : ansi_color,
1276 isempty(pretty_name) ? "Linux" : pretty_name);
1277 else
1278 return status_printf(NULL, false, false,
1279 "\nWelcome to %s!\n",
1280 isempty(pretty_name) ? "Linux" : pretty_name);
1281 }
1282
1283 static int write_container_id(void) {
1284 const char *c;
1285 int r;
1286
1287 c = getenv("container");
1288 if (isempty(c))
1289 return 0;
1290
1291 RUN_WITH_UMASK(0022)
1292 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1293 if (r < 0)
1294 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
1295
1296 return 1;
1297 }
1298
1299 static int bump_unix_max_dgram_qlen(void) {
1300 _cleanup_free_ char *qlen = NULL;
1301 unsigned long v;
1302 int r;
1303
1304 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1305 * really really early during boot, so that it is actually applied to all our sockets, including the
1306 * $NOTIFY_SOCKET one. */
1307
1308 r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1309 if (r < 0)
1310 return log_warning_errno(r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1311
1312 r = safe_atolu(qlen, &v);
1313 if (r < 0)
1314 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen);
1315
1316 if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1317 return 0;
1318
1319 r = write_string_filef("/proc/sys/net/unix/max_dgram_qlen", 0, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN);
1320 if (r < 0)
1321 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1322 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1323
1324 return 1;
1325 }
1326
1327 static int fixup_environment(void) {
1328 _cleanup_free_ char *term = NULL;
1329 const char *t;
1330 int r;
1331
1332 /* Only fix up the environment when we are started as PID 1 */
1333 if (getpid_cached() != 1)
1334 return 0;
1335
1336 /* We expect the environment to be set correctly if run inside a container. */
1337 if (detect_container() > 0)
1338 return 0;
1339
1340 /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1341 * device used by the console. We try to make a better guess here since some consoles might not have support
1342 * for color mode for example.
1343 *
1344 * However if TERM was configured through the kernel command line then leave it alone. */
1345 r = proc_cmdline_get_key("TERM", 0, &term);
1346 if (r < 0)
1347 return r;
1348
1349 t = term ?: default_term_for_tty("/dev/console");
1350
1351 if (setenv("TERM", t, 1) < 0)
1352 return -errno;
1353
1354 return 0;
1355 }
1356
1357 static void redirect_telinit(int argc, char *argv[]) {
1358
1359 /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1360
1361 #if HAVE_SYSV_COMPAT
1362 if (getpid_cached() == 1)
1363 return;
1364
1365 if (!strstr(program_invocation_short_name, "init"))
1366 return;
1367
1368 execv(SYSTEMCTL_BINARY_PATH, argv);
1369 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1370 exit(EXIT_FAILURE);
1371 #endif
1372 }
1373
1374 static int become_shutdown(
1375 const char *shutdown_verb,
1376 int retval) {
1377
1378 char log_level[DECIMAL_STR_MAX(int) + 1],
1379 exit_code[DECIMAL_STR_MAX(uint8_t) + 1],
1380 timeout[DECIMAL_STR_MAX(usec_t) + 1];
1381
1382 const char* command_line[13] = {
1383 SYSTEMD_SHUTDOWN_BINARY_PATH,
1384 shutdown_verb,
1385 "--timeout", timeout,
1386 "--log-level", log_level,
1387 "--log-target",
1388 };
1389
1390 _cleanup_strv_free_ char **env_block = NULL;
1391 size_t pos = 7;
1392 int r;
1393
1394 assert(shutdown_verb);
1395 assert(!command_line[pos]);
1396 env_block = strv_copy(environ);
1397
1398 xsprintf(log_level, "%d", log_get_max_level());
1399 xsprintf(timeout, "%" PRI_USEC "us", arg_default_timeout_stop_usec);
1400
1401 switch (log_get_target()) {
1402
1403 case LOG_TARGET_KMSG:
1404 case LOG_TARGET_JOURNAL_OR_KMSG:
1405 case LOG_TARGET_SYSLOG_OR_KMSG:
1406 command_line[pos++] = "kmsg";
1407 break;
1408
1409 case LOG_TARGET_NULL:
1410 command_line[pos++] = "null";
1411 break;
1412
1413 case LOG_TARGET_CONSOLE:
1414 default:
1415 command_line[pos++] = "console";
1416 break;
1417 };
1418
1419 if (log_get_show_color())
1420 command_line[pos++] = "--log-color";
1421
1422 if (log_get_show_location())
1423 command_line[pos++] = "--log-location";
1424
1425 if (streq(shutdown_verb, "exit")) {
1426 command_line[pos++] = "--exit-code";
1427 command_line[pos++] = exit_code;
1428 xsprintf(exit_code, "%d", retval);
1429 }
1430
1431 assert(pos < ELEMENTSOF(command_line));
1432
1433 if (streq(shutdown_verb, "reboot") &&
1434 arg_shutdown_watchdog > 0 &&
1435 arg_shutdown_watchdog != USEC_INFINITY) {
1436
1437 char *e;
1438
1439 /* If we reboot let's set the shutdown
1440 * watchdog and tell the shutdown binary to
1441 * repeatedly ping it */
1442 r = watchdog_set_timeout(&arg_shutdown_watchdog);
1443 watchdog_close(r < 0);
1444
1445 /* Tell the binary how often to ping, ignore failure */
1446 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
1447 (void) strv_consume(&env_block, e);
1448
1449 if (arg_watchdog_device &&
1450 asprintf(&e, "WATCHDOG_DEVICE=%s", arg_watchdog_device) > 0)
1451 (void) strv_consume(&env_block, e);
1452 } else
1453 watchdog_close(true);
1454
1455 /* Avoid the creation of new processes forked by the
1456 * kernel; at this point, we will not listen to the
1457 * signals anyway */
1458 if (detect_container() <= 0)
1459 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
1460
1461 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
1462 return -errno;
1463 }
1464
1465 static void initialize_clock(void) {
1466 int r;
1467
1468 if (clock_is_localtime(NULL) > 0) {
1469 int min;
1470
1471 /*
1472 * The very first call of settimeofday() also does a time warp in the kernel.
1473 *
1474 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1475 * of maintaining the RTC and do all adjustments. This matches the behavior of Windows, which leaves
1476 * the RTC alone if the registry tells that the RTC runs in UTC.
1477 */
1478 r = clock_set_timezone(&min);
1479 if (r < 0)
1480 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1481 else
1482 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1483
1484 } else if (!in_initrd()) {
1485 /*
1486 * Do a dummy very first call to seal the kernel's time warp magic.
1487 *
1488 * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1489 * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1490 * until we reach the real system.
1491 *
1492 * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1493 * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1494 * be treated as UTC that way.
1495 */
1496 (void) clock_reset_timewarp();
1497 }
1498
1499 r = clock_apply_epoch();
1500 if (r < 0)
1501 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1502 else if (r > 0)
1503 log_info("System time before build time, advancing clock.");
1504 }
1505
1506 static void initialize_coredump(bool skip_setup) {
1507 #if ENABLE_COREDUMP
1508 if (getpid_cached() != 1)
1509 return;
1510
1511 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1512 * will process core dumps for system services by default. */
1513 if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0)
1514 log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m");
1515
1516 /* But at the same time, turn off the core_pattern logic by default, so that no
1517 * coredumps are stored until the systemd-coredump tool is enabled via
1518 * sysctl. However it can be changed via the kernel command line later so core
1519 * dumps can still be generated during early startup and in initramfs. */
1520 if (!skip_setup)
1521 disable_coredumps();
1522 #endif
1523 }
1524
1525 static void initialize_core_pattern(bool skip_setup) {
1526 int r;
1527
1528 if (skip_setup || !arg_early_core_pattern)
1529 return;
1530
1531 if (getpid_cached() != 1)
1532 return;
1533
1534 r = write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern, 0);
1535 if (r < 0)
1536 log_warning_errno(r, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", arg_early_core_pattern);
1537 }
1538
1539 static void do_reexecute(
1540 int argc,
1541 char *argv[],
1542 const struct rlimit *saved_rlimit_nofile,
1543 const struct rlimit *saved_rlimit_memlock,
1544 FDSet *fds,
1545 const char *switch_root_dir,
1546 const char *switch_root_init,
1547 const char **ret_error_message) {
1548
1549 unsigned i, j, args_size;
1550 const char **args;
1551 int r;
1552
1553 assert(saved_rlimit_nofile);
1554 assert(saved_rlimit_memlock);
1555 assert(ret_error_message);
1556
1557 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1558 * we do that */
1559 watchdog_close(true);
1560
1561 /* Reset the RLIMIT_NOFILE to the kernel default, so that the new systemd can pass the kernel default to its
1562 * child processes */
1563
1564 if (saved_rlimit_nofile->rlim_cur > 0)
1565 (void) setrlimit(RLIMIT_NOFILE, saved_rlimit_nofile);
1566 if (saved_rlimit_memlock->rlim_cur != (rlim_t) -1)
1567 (void) setrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock);
1568
1569 if (switch_root_dir) {
1570 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1571 * SIGCHLD for them after deserializing. */
1572 broadcast_signal(SIGTERM, false, true, arg_default_timeout_stop_usec);
1573
1574 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1575 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1576 if (r < 0)
1577 log_error_errno(r, "Failed to switch root, trying to continue: %m");
1578 }
1579
1580 args_size = MAX(6, argc+1);
1581 args = newa(const char*, args_size);
1582
1583 if (!switch_root_init) {
1584 char sfd[DECIMAL_STR_MAX(int) + 1];
1585
1586 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1587 * the user didn't specify an explicit init to spawn. */
1588
1589 assert(arg_serialization);
1590 assert(fds);
1591
1592 xsprintf(sfd, "%i", fileno(arg_serialization));
1593
1594 i = 0;
1595 args[i++] = SYSTEMD_BINARY_PATH;
1596 if (switch_root_dir)
1597 args[i++] = "--switched-root";
1598 args[i++] = arg_system ? "--system" : "--user";
1599 args[i++] = "--deserialize";
1600 args[i++] = sfd;
1601 args[i++] = NULL;
1602
1603 assert(i <= args_size);
1604
1605 /*
1606 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do this is on
1607 * its own on exec(), but it will do it on exit(). Hence, to ensure we get a summary here, fork() off
1608 * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1609 * before proceeding into the exec().
1610 */
1611 valgrind_summary_hack();
1612
1613 (void) execv(args[0], (char* const*) args);
1614 log_debug_errno(errno, "Failed to execute our own binary, trying fallback: %m");
1615 }
1616
1617 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1618 * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1619 * doesn't matter.) */
1620
1621 arg_serialization = safe_fclose(arg_serialization);
1622 fds = fdset_free(fds);
1623
1624 /* Reopen the console */
1625 (void) make_console_stdio();
1626
1627 for (j = 1, i = 1; j < (unsigned) argc; j++)
1628 args[i++] = argv[j];
1629 args[i++] = NULL;
1630 assert(i <= args_size);
1631
1632 /* Reenable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1633 (void) reset_all_signal_handlers();
1634 (void) reset_signal_mask();
1635
1636 if (switch_root_init) {
1637 args[0] = switch_root_init;
1638 (void) execv(args[0], (char* const*) args);
1639 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
1640 }
1641
1642 args[0] = "/sbin/init";
1643 (void) execv(args[0], (char* const*) args);
1644 r = -errno;
1645
1646 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
1647 ANSI_HIGHLIGHT_RED " !! " ANSI_NORMAL,
1648 "Failed to execute /sbin/init");
1649
1650 if (r == -ENOENT) {
1651 log_warning("No /sbin/init, trying fallback");
1652
1653 args[0] = "/bin/sh";
1654 args[1] = NULL;
1655 (void) execv(args[0], (char* const*) args);
1656 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
1657 } else
1658 log_warning_errno(r, "Failed to execute /sbin/init, giving up: %m");
1659
1660 *ret_error_message = "Failed to execute fallback shell";
1661 }
1662
1663 static int invoke_main_loop(
1664 Manager *m,
1665 bool *ret_reexecute,
1666 int *ret_retval, /* Return parameters relevant for shutting down */
1667 const char **ret_shutdown_verb, /* … */
1668 FDSet **ret_fds, /* Return parameters for reexecuting */
1669 char **ret_switch_root_dir, /* … */
1670 char **ret_switch_root_init, /* … */
1671 const char **ret_error_message) {
1672
1673 int r;
1674
1675 assert(m);
1676 assert(ret_reexecute);
1677 assert(ret_retval);
1678 assert(ret_shutdown_verb);
1679 assert(ret_fds);
1680 assert(ret_switch_root_dir);
1681 assert(ret_switch_root_init);
1682 assert(ret_error_message);
1683
1684 for (;;) {
1685 r = manager_loop(m);
1686 if (r < 0) {
1687 *ret_error_message = "Failed to run main loop";
1688 return log_emergency_errno(r, "Failed to run main loop: %m");
1689 }
1690
1691 switch ((ManagerObjective) r) {
1692
1693 case MANAGER_RELOAD: {
1694 LogTarget saved_log_target;
1695 int saved_log_level;
1696
1697 log_info("Reloading.");
1698
1699 /* First, save any overridden log level/target, then parse the configuration file, which might
1700 * change the log level to new settings. */
1701
1702 saved_log_level = m->log_level_overridden ? log_get_max_level() : -1;
1703 saved_log_target = m->log_target_overridden ? log_get_target() : _LOG_TARGET_INVALID;
1704
1705 r = parse_config_file();
1706 if (r < 0)
1707 log_warning_errno(r, "Failed to parse config file, ignoring: %m");
1708
1709 set_manager_defaults(m);
1710
1711 if (saved_log_level >= 0)
1712 manager_override_log_level(m, saved_log_level);
1713 if (saved_log_target >= 0)
1714 manager_override_log_target(m, saved_log_target);
1715
1716 r = manager_reload(m);
1717 if (r < 0)
1718 /* Reloading failed before the point of no return. Let's continue running as if nothing happened. */
1719 m->objective = MANAGER_OK;
1720
1721 break;
1722 }
1723
1724 case MANAGER_REEXECUTE:
1725
1726 r = prepare_reexecute(m, &arg_serialization, ret_fds, false);
1727 if (r < 0) {
1728 *ret_error_message = "Failed to prepare for reexecution";
1729 return r;
1730 }
1731
1732 log_notice("Reexecuting.");
1733
1734 *ret_reexecute = true;
1735 *ret_retval = EXIT_SUCCESS;
1736 *ret_shutdown_verb = NULL;
1737 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1738
1739 return 0;
1740
1741 case MANAGER_SWITCH_ROOT:
1742 if (!m->switch_root_init) {
1743 r = prepare_reexecute(m, &arg_serialization, ret_fds, true);
1744 if (r < 0) {
1745 *ret_error_message = "Failed to prepare for reexecution";
1746 return r;
1747 }
1748 } else
1749 *ret_fds = NULL;
1750
1751 log_notice("Switching root.");
1752
1753 *ret_reexecute = true;
1754 *ret_retval = EXIT_SUCCESS;
1755 *ret_shutdown_verb = NULL;
1756
1757 /* Steal the switch root parameters */
1758 *ret_switch_root_dir = m->switch_root;
1759 *ret_switch_root_init = m->switch_root_init;
1760 m->switch_root = m->switch_root_init = NULL;
1761
1762 return 0;
1763
1764 case MANAGER_EXIT:
1765
1766 if (MANAGER_IS_USER(m)) {
1767 log_debug("Exit.");
1768
1769 *ret_reexecute = false;
1770 *ret_retval = m->return_value;
1771 *ret_shutdown_verb = NULL;
1772 *ret_fds = NULL;
1773 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1774
1775 return 0;
1776 }
1777
1778 _fallthrough_;
1779 case MANAGER_REBOOT:
1780 case MANAGER_POWEROFF:
1781 case MANAGER_HALT:
1782 case MANAGER_KEXEC: {
1783 static const char * const table[_MANAGER_OBJECTIVE_MAX] = {
1784 [MANAGER_EXIT] = "exit",
1785 [MANAGER_REBOOT] = "reboot",
1786 [MANAGER_POWEROFF] = "poweroff",
1787 [MANAGER_HALT] = "halt",
1788 [MANAGER_KEXEC] = "kexec",
1789 };
1790
1791 log_notice("Shutting down.");
1792
1793 *ret_reexecute = false;
1794 *ret_retval = m->return_value;
1795 assert_se(*ret_shutdown_verb = table[m->objective]);
1796 *ret_fds = NULL;
1797 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1798
1799 return 0;
1800 }
1801
1802 default:
1803 assert_not_reached("Unknown or unexpected manager objective.");
1804 }
1805 }
1806 }
1807
1808 static void log_execution_mode(bool *ret_first_boot) {
1809 assert(ret_first_boot);
1810
1811 if (arg_system) {
1812 int v;
1813
1814 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1815 arg_action == ACTION_TEST ? "test " : "" );
1816
1817 v = detect_virtualization();
1818 if (v > 0)
1819 log_info("Detected virtualization %s.", virtualization_to_string(v));
1820
1821 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1822
1823 if (in_initrd()) {
1824 *ret_first_boot = false;
1825 log_info("Running in initial RAM disk.");
1826 } else {
1827 /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
1828 * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
1829 * doesn't it's unpopulated. This allows container managers and installers to provision a
1830 * couple of files already. If the container manager wants to provision the machine ID itself
1831 * it should pass $container_uuid to PID 1. */
1832
1833 *ret_first_boot = access("/etc/machine-id", F_OK) < 0;
1834 if (*ret_first_boot)
1835 log_info("Running with unpopulated /etc.");
1836 }
1837 } else {
1838 if (DEBUG_LOGGING) {
1839 _cleanup_free_ char *t;
1840
1841 t = uid_to_name(getuid());
1842 log_debug(PACKAGE_STRING " running in %suser mode for user " UID_FMT "/%s. (" SYSTEMD_FEATURES ")",
1843 arg_action == ACTION_TEST ? " test" : "", getuid(), strna(t));
1844 }
1845
1846 *ret_first_boot = false;
1847 }
1848 }
1849
1850 static int initialize_runtime(
1851 bool skip_setup,
1852 struct rlimit *saved_rlimit_nofile,
1853 struct rlimit *saved_rlimit_memlock,
1854 const char **ret_error_message) {
1855
1856 int r;
1857
1858 assert(ret_error_message);
1859
1860 /* Sets up various runtime parameters. Many of these initializations are conditionalized:
1861 *
1862 * - Some only apply to --system instances
1863 * - Some only apply to --user instances
1864 * - Some only apply when we first start up, but not when we reexecute
1865 */
1866
1867 if (arg_action != ACTION_RUN)
1868 return 0;
1869
1870 if (arg_system) {
1871 /* Make sure we leave a core dump without panicing the kernel. */
1872 install_crash_handler();
1873
1874 if (!skip_setup) {
1875 r = mount_cgroup_controllers(arg_join_controllers);
1876 if (r < 0) {
1877 *ret_error_message = "Failed to mount cgroup hierarchies";
1878 return r;
1879 }
1880
1881 status_welcome();
1882 hostname_setup();
1883 machine_id_setup(NULL, arg_machine_id, NULL);
1884 loopback_setup();
1885 bump_unix_max_dgram_qlen();
1886 test_usr();
1887 write_container_id();
1888 }
1889
1890 if (arg_watchdog_device) {
1891 r = watchdog_set_device(arg_watchdog_device);
1892 if (r < 0)
1893 log_warning_errno(r, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device);
1894 }
1895
1896 if (arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY)
1897 watchdog_set_timeout(&arg_runtime_watchdog);
1898 }
1899
1900 if (arg_timer_slack_nsec != NSEC_INFINITY)
1901 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1902 log_warning_errno(errno, "Failed to adjust timer slack, ignoring: %m");
1903
1904 if (arg_system && !cap_test_all(arg_capability_bounding_set)) {
1905 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
1906 if (r < 0) {
1907 *ret_error_message = "Failed to drop capability bounding set of usermode helpers";
1908 return log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
1909 }
1910
1911 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
1912 if (r < 0) {
1913 *ret_error_message = "Failed to drop capability bounding set";
1914 return log_emergency_errno(r, "Failed to drop capability bounding set: %m");
1915 }
1916 }
1917
1918 if (arg_system && arg_no_new_privs) {
1919 if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) {
1920 *ret_error_message = "Failed to disable new privileges";
1921 return log_emergency_errno(errno, "Failed to disable new privileges: %m");
1922 }
1923 }
1924
1925 if (arg_syscall_archs) {
1926 r = enforce_syscall_archs(arg_syscall_archs);
1927 if (r < 0) {
1928 *ret_error_message = "Failed to set syscall architectures";
1929 return r;
1930 }
1931 }
1932
1933 if (!arg_system)
1934 /* Become reaper of our children */
1935 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
1936 log_warning_errno(errno, "Failed to make us a subreaper: %m");
1937
1938 /* Bump up RLIMIT_NOFILE for systemd itself */
1939 (void) bump_rlimit_nofile(saved_rlimit_nofile);
1940 (void) bump_rlimit_memlock(saved_rlimit_memlock);
1941
1942 return 0;
1943 }
1944
1945 static int do_queue_default_job(
1946 Manager *m,
1947 const char **ret_error_message) {
1948
1949 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
1950 Job *default_unit_job;
1951 Unit *target = NULL;
1952 int r;
1953
1954 log_debug("Activating default unit: %s", arg_default_unit);
1955
1956 r = manager_load_startable_unit_or_warn(m, arg_default_unit, NULL, &target);
1957 if (r < 0) {
1958 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET);
1959
1960 r = manager_load_startable_unit_or_warn(m, SPECIAL_RESCUE_TARGET, NULL, &target);
1961 if (r < 0) {
1962 *ret_error_message = r == -ERFKILL ? "Rescue target masked"
1963 : "Failed to load rescue target";
1964 return r;
1965 }
1966 }
1967
1968 assert(target->load_state == UNIT_LOADED);
1969
1970 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, &error, &default_unit_job);
1971 if (r == -EPERM) {
1972 log_debug_errno(r, "Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1973
1974 sd_bus_error_free(&error);
1975
1976 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, &error, &default_unit_job);
1977 if (r < 0) {
1978 *ret_error_message = "Failed to start default target";
1979 return log_emergency_errno(r, "Failed to start default target: %s", bus_error_message(&error, r));
1980 }
1981
1982 } else if (r < 0) {
1983 *ret_error_message = "Failed to isolate default target";
1984 return log_emergency_errno(r, "Failed to isolate default target: %s", bus_error_message(&error, r));
1985 }
1986
1987 m->default_unit_job_id = default_unit_job->id;
1988
1989 return 0;
1990 }
1991
1992 static void free_arguments(void) {
1993
1994 /* Frees all arg_* variables, with the exception of arg_serialization */
1995 rlimit_free_all(arg_default_rlimit);
1996
1997 arg_default_unit = mfree(arg_default_unit);
1998 arg_confirm_spawn = mfree(arg_confirm_spawn);
1999 arg_join_controllers = strv_free_free(arg_join_controllers);
2000 arg_default_environment = strv_free(arg_default_environment);
2001 arg_syscall_archs = set_free(arg_syscall_archs);
2002 }
2003
2004 static int load_configuration(int argc, char **argv, const char **ret_error_message) {
2005 int r;
2006
2007 assert(ret_error_message);
2008
2009 arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U);
2010
2011 r = parse_config_file();
2012 if (r < 0) {
2013 *ret_error_message = "Failed to parse config file";
2014 return r;
2015 }
2016
2017 if (arg_system) {
2018 r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
2019 if (r < 0)
2020 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
2021 }
2022
2023 /* Note that this also parses bits from the kernel command line, including "debug". */
2024 log_parse_environment();
2025
2026 r = parse_argv(argc, argv);
2027 if (r < 0) {
2028 *ret_error_message = "Failed to parse commandline arguments";
2029 return r;
2030 }
2031
2032 /* Initialize default unit */
2033 if (!arg_default_unit) {
2034 arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET);
2035 if (!arg_default_unit) {
2036 *ret_error_message = "Failed to set default unit";
2037 return log_oom();
2038 }
2039 }
2040
2041 /* Initialize the show status setting if it hasn't been set explicitly yet */
2042 if (arg_show_status == _SHOW_STATUS_INVALID)
2043 arg_show_status = SHOW_STATUS_YES;
2044
2045 return 0;
2046 }
2047
2048 static int safety_checks(void) {
2049
2050 if (getpid_cached() == 1 &&
2051 arg_action != ACTION_RUN) {
2052 log_error("Unsupported execution mode while PID 1.");
2053 return -EPERM;
2054 }
2055
2056 if (getpid_cached() == 1 &&
2057 !arg_system) {
2058 log_error("Can't run --user mode as PID 1.");
2059 return -EPERM;
2060 }
2061
2062 if (arg_action == ACTION_RUN &&
2063 arg_system &&
2064 getpid_cached() != 1) {
2065 log_error("Can't run system mode unless PID 1.");
2066 return -EPERM;
2067 }
2068
2069 if (arg_action == ACTION_TEST &&
2070 geteuid() == 0) {
2071 log_error("Don't run test mode as root.");
2072 return -EPERM;
2073 }
2074
2075 if (!arg_system &&
2076 arg_action == ACTION_RUN &&
2077 sd_booted() <= 0) {
2078 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
2079 return -EOPNOTSUPP;
2080 }
2081
2082 if (!arg_system &&
2083 arg_action == ACTION_RUN &&
2084 !getenv("XDG_RUNTIME_DIR")) {
2085 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
2086 return -EUNATCH;
2087 }
2088
2089 if (arg_system &&
2090 arg_action == ACTION_RUN &&
2091 running_in_chroot() > 0) {
2092 log_error("Cannot be run in a chroot() environment.");
2093 return -EOPNOTSUPP;
2094 }
2095
2096 return 0;
2097 }
2098
2099 static int initialize_security(
2100 bool *loaded_policy,
2101 dual_timestamp *security_start_timestamp,
2102 dual_timestamp *security_finish_timestamp,
2103 const char **ret_error_message) {
2104
2105 int r;
2106
2107 assert(loaded_policy);
2108 assert(security_start_timestamp);
2109 assert(security_finish_timestamp);
2110 assert(ret_error_message);
2111
2112 dual_timestamp_get(security_start_timestamp);
2113
2114 r = mac_selinux_setup(loaded_policy);
2115 if (r < 0) {
2116 *ret_error_message = "Failed to load SELinux policy";
2117 return r;
2118 }
2119
2120 r = mac_smack_setup(loaded_policy);
2121 if (r < 0) {
2122 *ret_error_message = "Failed to load SMACK policy";
2123 return r;
2124 }
2125
2126 r = ima_setup();
2127 if (r < 0) {
2128 *ret_error_message = "Failed to load IMA policy";
2129 return r;
2130 }
2131
2132 dual_timestamp_get(security_finish_timestamp);
2133 return 0;
2134 }
2135
2136 static void test_summary(Manager *m) {
2137 assert(m);
2138
2139 printf("-> By units:\n");
2140 manager_dump_units(m, stdout, "\t");
2141
2142 printf("-> By jobs:\n");
2143 manager_dump_jobs(m, stdout, "\t");
2144 }
2145
2146 static int collect_fds(FDSet **ret_fds, const char **ret_error_message) {
2147 int r;
2148
2149 assert(ret_fds);
2150 assert(ret_error_message);
2151
2152 r = fdset_new_fill(ret_fds);
2153 if (r < 0) {
2154 *ret_error_message = "Failed to allocate fd set";
2155 return log_emergency_errno(r, "Failed to allocate fd set: %m");
2156 }
2157
2158 fdset_cloexec(*ret_fds, true);
2159
2160 if (arg_serialization)
2161 assert_se(fdset_remove(*ret_fds, fileno(arg_serialization)) >= 0);
2162
2163 return 0;
2164 }
2165
2166 static void setup_console_terminal(bool skip_setup) {
2167
2168 if (!arg_system)
2169 return;
2170
2171 /* Become a session leader if we aren't one yet. */
2172 (void) setsid();
2173
2174 /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2175 * tty. */
2176 (void) release_terminal();
2177
2178 /* Reset the console, but only if this is really init and we are freshly booted */
2179 if (getpid_cached() == 1 && !skip_setup)
2180 (void) console_setup();
2181 }
2182
2183 static bool early_skip_setup_check(int argc, char *argv[]) {
2184 bool found_deserialize = false;
2185 int i;
2186
2187 /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2188 * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2189 * anyway, even if in that case we also do deserialization. */
2190
2191 for (i = 1; i < argc; i++) {
2192 if (streq(argv[i], "--switched-root"))
2193 return false; /* If we switched root, don't skip the setup. */
2194 else if (streq(argv[i], "--deserialize"))
2195 found_deserialize = true;
2196 }
2197
2198 return found_deserialize; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2199 }
2200
2201 int main(int argc, char *argv[]) {
2202
2203 dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL, userspace_timestamp = DUAL_TIMESTAMP_NULL, kernel_timestamp = DUAL_TIMESTAMP_NULL,
2204 security_start_timestamp = DUAL_TIMESTAMP_NULL, security_finish_timestamp = DUAL_TIMESTAMP_NULL;
2205 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0), saved_rlimit_memlock = RLIMIT_MAKE_CONST((rlim_t) -1);
2206 bool skip_setup, loaded_policy = false, queue_default_job = false, first_boot = false, reexecute = false;
2207 char *switch_root_dir = NULL, *switch_root_init = NULL;
2208 usec_t before_startup, after_startup;
2209 static char systemd[] = "systemd";
2210 char timespan[FORMAT_TIMESPAN_MAX];
2211 const char *shutdown_verb = NULL, *error_message = NULL;
2212 int r, retval = EXIT_FAILURE;
2213 Manager *m = NULL;
2214 FDSet *fds = NULL;
2215
2216 /* SysV compatibility: redirect init → telinit */
2217 redirect_telinit(argc, argv);
2218
2219 /* Take timestamps early on */
2220 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
2221 dual_timestamp_get(&userspace_timestamp);
2222
2223 /* Figure out whether we need to do initialize the system, or if we already did that because we are
2224 * reexecuting */
2225 skip_setup = early_skip_setup_check(argc, argv);
2226
2227 /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2228 * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
2229 program_invocation_short_name = systemd;
2230 (void) prctl(PR_SET_NAME, systemd);
2231
2232 /* Save the original command line */
2233 saved_argv = argv;
2234 saved_argc = argc;
2235
2236 /* Make sure that if the user says "syslog" we actually log to the journal. */
2237 log_set_upgrade_syslog_to_journal(true);
2238
2239 if (getpid_cached() == 1) {
2240 /* When we run as PID 1 force system mode */
2241 arg_system = true;
2242
2243 /* Disable the umask logic */
2244 umask(0);
2245
2246 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2247 * activated yet (even though the log socket for it exists). */
2248 log_set_prohibit_ipc(true);
2249
2250 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2251 * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2252 * child process right before execve()'ing the actual binary, at a point in time where socket
2253 * activation stderr/stdout area already set up. */
2254 log_set_always_reopen_console(true);
2255
2256 if (detect_container() <= 0) {
2257
2258 /* Running outside of a container as PID 1 */
2259 log_set_target(LOG_TARGET_KMSG);
2260 log_open();
2261
2262 if (in_initrd())
2263 initrd_timestamp = userspace_timestamp;
2264
2265 if (!skip_setup) {
2266 r = mount_setup_early();
2267 if (r < 0) {
2268 error_message = "Failed to mount early API filesystems";
2269 goto finish;
2270 }
2271
2272 r = initialize_security(
2273 &loaded_policy,
2274 &security_start_timestamp,
2275 &security_finish_timestamp,
2276 &error_message);
2277 if (r < 0)
2278 goto finish;
2279 }
2280
2281 if (mac_selinux_init() < 0) {
2282 error_message = "Failed to initialize SELinux policy";
2283 goto finish;
2284 }
2285
2286 if (!skip_setup)
2287 initialize_clock();
2288
2289 /* Set the default for later on, but don't actually open the logs like this for now. Note that
2290 * if we are transitioning from the initrd there might still be journal fd open, and we
2291 * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2292 * elsewhere. */
2293 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
2294
2295 } else {
2296 /* Running inside a container, as PID 1 */
2297 log_set_target(LOG_TARGET_CONSOLE);
2298 log_open();
2299
2300 /* For later on, see above... */
2301 log_set_target(LOG_TARGET_JOURNAL);
2302
2303 /* clear the kernel timestamp,
2304 * because we are in a container */
2305 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2306 }
2307
2308 initialize_coredump(skip_setup);
2309
2310 r = fixup_environment();
2311 if (r < 0) {
2312 log_emergency_errno(r, "Failed to fix up PID 1 environment: %m");
2313 error_message = "Failed to fix up PID1 environment";
2314 goto finish;
2315 }
2316
2317 } else {
2318 /* Running as user instance */
2319 arg_system = false;
2320 log_set_target(LOG_TARGET_AUTO);
2321 log_open();
2322
2323 /* clear the kernel timestamp,
2324 * because we are not PID 1 */
2325 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2326 }
2327
2328 if (arg_system) {
2329 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2330 * they never log into the console. */
2331 log_show_color(colors_enabled());
2332
2333 r = make_null_stdio();
2334 if (r < 0)
2335 log_warning_errno(r, "Failed to redirect standard streams to /dev/null, ignoring: %m");
2336 }
2337
2338 /* Mount /proc, /sys and friends, so that /proc/cmdline and
2339 * /proc/$PID/fd is available. */
2340 if (getpid_cached() == 1) {
2341
2342 /* Load the kernel modules early. */
2343 if (!skip_setup)
2344 kmod_setup();
2345
2346 r = mount_setup(loaded_policy);
2347 if (r < 0) {
2348 error_message = "Failed to mount API filesystems";
2349 goto finish;
2350 }
2351 }
2352
2353 /* Reset all signal handlers. */
2354 (void) reset_all_signal_handlers();
2355 (void) ignore_signals(SIGNALS_IGNORE, -1);
2356
2357 r = load_configuration(argc, argv, &error_message);
2358 if (r < 0)
2359 goto finish;
2360
2361 r = safety_checks();
2362 if (r < 0)
2363 goto finish;
2364
2365 if (IN_SET(arg_action, ACTION_TEST, ACTION_HELP, ACTION_DUMP_CONFIGURATION_ITEMS, ACTION_DUMP_BUS_PROPERTIES))
2366 (void) pager_open(arg_no_pager, false);
2367
2368 if (arg_action != ACTION_RUN)
2369 skip_setup = true;
2370
2371 if (arg_action == ACTION_HELP) {
2372 retval = help() < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
2373 goto finish;
2374 } else if (arg_action == ACTION_VERSION) {
2375 retval = version();
2376 goto finish;
2377 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
2378 unit_dump_config_items(stdout);
2379 retval = EXIT_SUCCESS;
2380 goto finish;
2381 } else if (arg_action == ACTION_DUMP_BUS_PROPERTIES) {
2382 dump_bus_properties(stdout);
2383 retval = EXIT_SUCCESS;
2384 goto finish;
2385 }
2386
2387 assert_se(IN_SET(arg_action, ACTION_RUN, ACTION_TEST));
2388
2389 /* Move out of the way, so that we won't block unmounts */
2390 assert_se(chdir("/") == 0);
2391
2392 if (arg_action == ACTION_RUN) {
2393
2394 /* A core pattern might have been specified via the cmdline. */
2395 initialize_core_pattern(skip_setup);
2396
2397 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2398 log_close();
2399
2400 /* Remember open file descriptors for later deserialization */
2401 r = collect_fds(&fds, &error_message);
2402 if (r < 0)
2403 goto finish;
2404
2405 /* Give up any control of the console, but make sure its initialized. */
2406 setup_console_terminal(skip_setup);
2407
2408 /* Open the logging devices, if possible and necessary */
2409 log_open();
2410 }
2411
2412 log_execution_mode(&first_boot);
2413
2414 r = initialize_runtime(skip_setup,
2415 &saved_rlimit_nofile,
2416 &saved_rlimit_memlock,
2417 &error_message);
2418 if (r < 0)
2419 goto finish;
2420
2421 r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER,
2422 arg_action == ACTION_TEST ? MANAGER_TEST_FULL : 0,
2423 &m);
2424 if (r < 0) {
2425 log_emergency_errno(r, "Failed to allocate manager object: %m");
2426 error_message = "Failed to allocate manager object";
2427 goto finish;
2428 }
2429
2430 m->timestamps[MANAGER_TIMESTAMP_KERNEL] = kernel_timestamp;
2431 m->timestamps[MANAGER_TIMESTAMP_INITRD] = initrd_timestamp;
2432 m->timestamps[MANAGER_TIMESTAMP_USERSPACE] = userspace_timestamp;
2433 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START)] = security_start_timestamp;
2434 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH)] = security_finish_timestamp;
2435
2436 set_manager_defaults(m);
2437 set_manager_settings(m);
2438 manager_set_first_boot(m, first_boot);
2439
2440 /* Remember whether we should queue the default job */
2441 queue_default_job = !arg_serialization || arg_switched_root;
2442
2443 before_startup = now(CLOCK_MONOTONIC);
2444
2445 r = manager_startup(m, arg_serialization, fds);
2446 if (r < 0) {
2447 error_message = "Failed to start up manager";
2448 goto finish;
2449 }
2450
2451 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2452 fds = fdset_free(fds);
2453 arg_serialization = safe_fclose(arg_serialization);
2454
2455 if (queue_default_job) {
2456 r = do_queue_default_job(m, &error_message);
2457 if (r < 0)
2458 goto finish;
2459 }
2460
2461 after_startup = now(CLOCK_MONOTONIC);
2462
2463 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
2464 "Loaded units and determined initial transaction in %s.",
2465 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
2466
2467 if (arg_action == ACTION_TEST) {
2468 test_summary(m);
2469 retval = EXIT_SUCCESS;
2470 goto finish;
2471 }
2472
2473 (void) invoke_main_loop(m,
2474 &reexecute,
2475 &retval,
2476 &shutdown_verb,
2477 &fds,
2478 &switch_root_dir,
2479 &switch_root_init,
2480 &error_message);
2481
2482 finish:
2483 pager_close();
2484
2485 if (m) {
2486 arg_shutdown_watchdog = m->shutdown_watchdog;
2487 m = manager_free(m);
2488 }
2489
2490 free_arguments();
2491 mac_selinux_finish();
2492
2493 if (reexecute)
2494 do_reexecute(argc, argv,
2495 &saved_rlimit_nofile,
2496 &saved_rlimit_memlock,
2497 fds,
2498 switch_root_dir,
2499 switch_root_init,
2500 &error_message); /* This only returns if reexecution failed */
2501
2502 arg_serialization = safe_fclose(arg_serialization);
2503 fds = fdset_free(fds);
2504
2505 #if HAVE_VALGRIND_VALGRIND_H
2506 /* If we are PID 1 and running under valgrind, then let's exit
2507 * here explicitly. valgrind will only generate nice output on
2508 * exit(), not on exec(), hence let's do the former not the
2509 * latter here. */
2510 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND) {
2511 /* Cleanup watchdog_device strings for valgrind. We need them
2512 * in become_shutdown() so normally we cannot free them yet. */
2513 watchdog_free_device();
2514 arg_watchdog_device = mfree(arg_watchdog_device);
2515 return retval;
2516 }
2517 #endif
2518
2519 if (shutdown_verb) {
2520 r = become_shutdown(shutdown_verb, retval);
2521 log_error_errno(r, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
2522 error_message = "Failed to execute shutdown binary";
2523 }
2524
2525 watchdog_free_device();
2526 arg_watchdog_device = mfree(arg_watchdog_device);
2527
2528 if (getpid_cached() == 1) {
2529 if (error_message)
2530 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2531 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
2532 "%s, freezing.", error_message);
2533 freeze_or_reboot();
2534 }
2535
2536 return retval;
2537 }
Unnamed repository; edit this file 'description' to name the repository.