]> git.ipfire.org Git - thirdparty/systemd.git/blob - src/core/main.c
projects / thirdparty / systemd.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge pull request #23927 from AndreKalb/feature/ipv4-link-local-start-address
[thirdparty/systemd.git] / src / core / main.c
1 /* SPDX-License-Identifier: LGPL-2.1-or-later */
2
3 #include <errno.h>
4 #include <fcntl.h>
5 #include <getopt.h>
6 #include <linux/oom.h>
7 #include <sys/mount.h>
8 #include <sys/prctl.h>
9 #include <sys/utsname.h>
10 #include <unistd.h>
11 #if HAVE_SECCOMP
12 #include <seccomp.h>
13 #endif
14 #if HAVE_VALGRIND_VALGRIND_H
15 #include <valgrind/valgrind.h>
16 #endif
17
18 #include "sd-bus.h"
19 #include "sd-daemon.h"
20 #include "sd-messages.h"
21
22 #include "alloc-util.h"
23 #include "apparmor-setup.h"
24 #include "architecture.h"
25 #if HAVE_LIBBPF
26 #include "bpf-lsm.h"
27 #endif
28 #include "build.h"
29 #include "bus-error.h"
30 #include "bus-util.h"
31 #include "capability-util.h"
32 #include "cgroup-util.h"
33 #include "clock-util.h"
34 #include "conf-parser.h"
35 #include "cpu-set-util.h"
36 #include "crash-handler.h"
37 #include "dbus-manager.h"
38 #include "dbus.h"
39 #include "def.h"
40 #include "dev-setup.h"
41 #include "efi-random.h"
42 #include "efivars.h"
43 #include "emergency-action.h"
44 #include "env-util.h"
45 #include "exit-status.h"
46 #include "fd-util.h"
47 #include "fdset.h"
48 #include "fileio.h"
49 #include "format-util.h"
50 #include "fs-util.h"
51 #include "hexdecoct.h"
52 #include "hostname-setup.h"
53 #include "ima-setup.h"
54 #include "import-creds.h"
55 #include "killall.h"
56 #include "kmod-setup.h"
57 #include "limits-util.h"
58 #include "load-fragment.h"
59 #include "log.h"
60 #include "loopback-setup.h"
61 #include "machine-id-setup.h"
62 #include "main.h"
63 #include "manager.h"
64 #include "manager-dump.h"
65 #include "manager-serialize.h"
66 #include "mkdir-label.h"
67 #include "mount-setup.h"
68 #include "os-util.h"
69 #include "pager.h"
70 #include "parse-argument.h"
71 #include "parse-util.h"
72 #include "path-util.h"
73 #include "pretty-print.h"
74 #include "proc-cmdline.h"
75 #include "process-util.h"
76 #include "random-util.h"
77 #include "rlimit-util.h"
78 #if HAVE_SECCOMP
79 #include "seccomp-util.h"
80 #endif
81 #include "selinux-setup.h"
82 #include "selinux-util.h"
83 #include "signal-util.h"
84 #include "smack-setup.h"
85 #include "special.h"
86 #include "stat-util.h"
87 #include "stdio-util.h"
88 #include "strv.h"
89 #include "switch-root.h"
90 #include "sysctl-util.h"
91 #include "terminal-util.h"
92 #include "time-util.h"
93 #include "umask-util.h"
94 #include "user-util.h"
95 #include "util.h"
96 #include "virt.h"
97 #include "watchdog.h"
98
99 #if HAS_FEATURE_ADDRESS_SANITIZER
100 #include <sanitizer/lsan_interface.h>
101 #endif
102
103 #define DEFAULT_TASKS_MAX ((TasksMax) { 15U, 100U }) /* 15% */
104
105 static enum {
106 ACTION_RUN,
107 ACTION_HELP,
108 ACTION_VERSION,
109 ACTION_TEST,
110 ACTION_DUMP_CONFIGURATION_ITEMS,
111 ACTION_DUMP_BUS_PROPERTIES,
112 ACTION_BUS_INTROSPECT,
113 } arg_action = ACTION_RUN;
114
115 static const char *arg_bus_introspect = NULL;
116
117 /* Those variables are initialized to 0 automatically, so we avoid uninitialized memory access. Real
118 * defaults are assigned in reset_arguments() below. */
119 static char *arg_default_unit;
120 static bool arg_system;
121 bool arg_dump_core;
122 int arg_crash_chvt;
123 bool arg_crash_shell;
124 bool arg_crash_reboot;
125 static char *arg_confirm_spawn;
126 static ShowStatus arg_show_status;
127 static StatusUnitFormat arg_status_unit_format;
128 static bool arg_switched_root;
129 static PagerFlags arg_pager_flags;
130 static bool arg_service_watchdogs;
131 static ExecOutput arg_default_std_output;
132 static ExecOutput arg_default_std_error;
133 static usec_t arg_default_restart_usec;
134 static usec_t arg_default_timeout_start_usec;
135 static usec_t arg_default_timeout_stop_usec;
136 static usec_t arg_default_timeout_abort_usec;
137 static bool arg_default_timeout_abort_set;
138 static usec_t arg_default_start_limit_interval;
139 static unsigned arg_default_start_limit_burst;
140 static usec_t arg_runtime_watchdog;
141 static usec_t arg_reboot_watchdog;
142 static usec_t arg_kexec_watchdog;
143 static usec_t arg_pretimeout_watchdog;
144 static char *arg_early_core_pattern;
145 static char *arg_watchdog_pretimeout_governor;
146 static char *arg_watchdog_device;
147 static char **arg_default_environment;
148 static char **arg_manager_environment;
149 static struct rlimit *arg_default_rlimit[_RLIMIT_MAX];
150 static uint64_t arg_capability_bounding_set;
151 static bool arg_no_new_privs;
152 static nsec_t arg_timer_slack_nsec;
153 static usec_t arg_default_timer_accuracy_usec;
154 static Set* arg_syscall_archs;
155 static FILE* arg_serialization;
156 static int arg_default_cpu_accounting;
157 static bool arg_default_io_accounting;
158 static bool arg_default_ip_accounting;
159 static bool arg_default_blockio_accounting;
160 static bool arg_default_memory_accounting;
161 static bool arg_default_tasks_accounting;
162 static TasksMax arg_default_tasks_max;
163 static sd_id128_t arg_machine_id;
164 static EmergencyAction arg_cad_burst_action;
165 static OOMPolicy arg_default_oom_policy;
166 static CPUSet arg_cpu_affinity;
167 static NUMAPolicy arg_numa_policy;
168 static usec_t arg_clock_usec;
169 static void *arg_random_seed;
170 static size_t arg_random_seed_size;
171 static int arg_default_oom_score_adjust;
172 static bool arg_default_oom_score_adjust_set;
173 static char *arg_default_smack_process_label;
174
175 /* A copy of the original environment block */
176 static char **saved_env = NULL;
177
178 static int parse_configuration(const struct rlimit *saved_rlimit_nofile,
179 const struct rlimit *saved_rlimit_memlock);
180
181 static int manager_find_user_config_paths(char ***ret_files, char ***ret_dirs) {
182 _cleanup_free_ char *base = NULL;
183 _cleanup_strv_free_ char **files = NULL, **dirs = NULL;
184 int r;
185
186 r = xdg_user_config_dir(&base, "/systemd");
187 if (r < 0)
188 return r;
189
190 r = strv_extendf(&files, "%s/user.conf", base);
191 if (r < 0)
192 return r;
193
194 r = strv_extend(&files, PKGSYSCONFDIR "/user.conf");
195 if (r < 0)
196 return r;
197
198 r = strv_consume(&dirs, TAKE_PTR(base));
199 if (r < 0)
200 return r;
201
202 r = strv_extend_strv(&dirs, CONF_PATHS_STRV("systemd"), false);
203 if (r < 0)
204 return r;
205
206 *ret_files = TAKE_PTR(files);
207 *ret_dirs = TAKE_PTR(dirs);
208 return 0;
209 }
210
211 static int console_setup(void) {
212 _cleanup_close_ int tty_fd = -1;
213 int r;
214
215 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
216 if (tty_fd < 0)
217 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
218
219 /* We don't want to force text mode. plymouth may be showing
220 * pictures already from initrd. */
221 r = reset_terminal_fd(tty_fd, false);
222 if (r < 0)
223 return log_error_errno(r, "Failed to reset /dev/console: %m");
224
225 return 0;
226 }
227
228 static int set_machine_id(const char *m) {
229 sd_id128_t t;
230 assert(m);
231
232 if (sd_id128_from_string(m, &t) < 0)
233 return -EINVAL;
234
235 if (sd_id128_is_null(t))
236 return -EINVAL;
237
238 arg_machine_id = t;
239 return 0;
240 }
241
242 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
243 int r;
244
245 assert(key);
246
247 if (STR_IN_SET(key, "systemd.unit", "rd.systemd.unit")) {
248
249 if (proc_cmdline_value_missing(key, value))
250 return 0;
251
252 if (!unit_name_is_valid(value, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE))
253 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key, value);
254 else if (in_initrd() == !!startswith(key, "rd."))
255 return free_and_strdup_warn(&arg_default_unit, value);
256
257 } else if (proc_cmdline_key_streq(key, "systemd.dump_core")) {
258
259 r = value ? parse_boolean(value) : true;
260 if (r < 0)
261 log_warning_errno(r, "Failed to parse dump core switch %s, ignoring: %m", value);
262 else
263 arg_dump_core = r;
264
265 } else if (proc_cmdline_key_streq(key, "systemd.early_core_pattern")) {
266
267 if (proc_cmdline_value_missing(key, value))
268 return 0;
269
270 if (path_is_absolute(value))
271 (void) parse_path_argument(value, false, &arg_early_core_pattern);
272 else
273 log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value);
274
275 } else if (proc_cmdline_key_streq(key, "systemd.crash_chvt")) {
276
277 if (!value)
278 arg_crash_chvt = 0; /* turn on */
279 else {
280 r = parse_crash_chvt(value, &arg_crash_chvt);
281 if (r < 0)
282 log_warning_errno(r, "Failed to parse crash chvt switch %s, ignoring: %m", value);
283 }
284
285 } else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) {
286
287 r = value ? parse_boolean(value) : true;
288 if (r < 0)
289 log_warning_errno(r, "Failed to parse crash shell switch %s, ignoring: %m", value);
290 else
291 arg_crash_shell = r;
292
293 } else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) {
294
295 r = value ? parse_boolean(value) : true;
296 if (r < 0)
297 log_warning_errno(r, "Failed to parse crash reboot switch %s, ignoring: %m", value);
298 else
299 arg_crash_reboot = r;
300
301 } else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) {
302 char *s;
303
304 r = parse_confirm_spawn(value, &s);
305 if (r < 0)
306 log_warning_errno(r, "Failed to parse confirm_spawn switch %s, ignoring: %m", value);
307 else
308 free_and_replace(arg_confirm_spawn, s);
309
310 } else if (proc_cmdline_key_streq(key, "systemd.service_watchdogs")) {
311
312 r = value ? parse_boolean(value) : true;
313 if (r < 0)
314 log_warning_errno(r, "Failed to parse service watchdog switch %s, ignoring: %m", value);
315 else
316 arg_service_watchdogs = r;
317
318 } else if (proc_cmdline_key_streq(key, "systemd.show_status")) {
319
320 if (value) {
321 r = parse_show_status(value, &arg_show_status);
322 if (r < 0)
323 log_warning_errno(r, "Failed to parse show status switch %s, ignoring: %m", value);
324 } else
325 arg_show_status = SHOW_STATUS_YES;
326
327 } else if (proc_cmdline_key_streq(key, "systemd.status_unit_format")) {
328
329 if (proc_cmdline_value_missing(key, value))
330 return 0;
331
332 r = status_unit_format_from_string(value);
333 if (r < 0)
334 log_warning_errno(r, "Failed to parse %s=%s, ignoring: %m", key, value);
335 else
336 arg_status_unit_format = r;
337
338 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) {
339
340 if (proc_cmdline_value_missing(key, value))
341 return 0;
342
343 r = exec_output_from_string(value);
344 if (r < 0)
345 log_warning_errno(r, "Failed to parse default standard output switch %s, ignoring: %m", value);
346 else
347 arg_default_std_output = r;
348
349 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) {
350
351 if (proc_cmdline_value_missing(key, value))
352 return 0;
353
354 r = exec_output_from_string(value);
355 if (r < 0)
356 log_warning_errno(r, "Failed to parse default standard error switch %s, ignoring: %m", value);
357 else
358 arg_default_std_error = r;
359
360 } else if (streq(key, "systemd.setenv")) {
361
362 if (proc_cmdline_value_missing(key, value))
363 return 0;
364
365 if (!env_assignment_is_valid(value))
366 log_warning("Environment variable assignment '%s' is not valid. Ignoring.", value);
367 else {
368 r = strv_env_replace_strdup(&arg_default_environment, value);
369 if (r < 0)
370 return log_oom();
371 }
372
373 } else if (proc_cmdline_key_streq(key, "systemd.machine_id")) {
374
375 if (proc_cmdline_value_missing(key, value))
376 return 0;
377
378 r = set_machine_id(value);
379 if (r < 0)
380 log_warning_errno(r, "MachineID '%s' is not valid, ignoring: %m", value);
381
382 } else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) {
383
384 if (proc_cmdline_value_missing(key, value))
385 return 0;
386
387 r = parse_sec(value, &arg_default_timeout_start_usec);
388 if (r < 0)
389 log_warning_errno(r, "Failed to parse default start timeout '%s', ignoring: %m", value);
390
391 if (arg_default_timeout_start_usec <= 0)
392 arg_default_timeout_start_usec = USEC_INFINITY;
393
394 } else if (proc_cmdline_key_streq(key, "systemd.cpu_affinity")) {
395
396 if (proc_cmdline_value_missing(key, value))
397 return 0;
398
399 r = parse_cpu_set(value, &arg_cpu_affinity);
400 if (r < 0)
401 log_warning_errno(r, "Failed to parse CPU affinity mask '%s', ignoring: %m", value);
402
403 } else if (proc_cmdline_key_streq(key, "systemd.watchdog_device")) {
404
405 if (proc_cmdline_value_missing(key, value))
406 return 0;
407
408 (void) parse_path_argument(value, false, &arg_watchdog_device);
409
410 } else if (proc_cmdline_key_streq(key, "systemd.watchdog_sec")) {
411
412 if (proc_cmdline_value_missing(key, value))
413 return 0;
414
415 if (streq(value, "default"))
416 arg_runtime_watchdog = USEC_INFINITY;
417 else if (streq(value, "off"))
418 arg_runtime_watchdog = 0;
419 else {
420 r = parse_sec(value, &arg_runtime_watchdog);
421 if (r < 0) {
422 log_warning_errno(r, "Failed to parse systemd.watchdog_sec= argument '%s', ignoring: %m", value);
423 return 0;
424 }
425 }
426
427 arg_kexec_watchdog = arg_reboot_watchdog = arg_runtime_watchdog;
428
429 } else if (proc_cmdline_key_streq(key, "systemd.watchdog_pre_sec")) {
430
431 if (proc_cmdline_value_missing(key, value))
432 return 0;
433
434 if (streq(value, "default"))
435 arg_pretimeout_watchdog = USEC_INFINITY;
436 else if (streq(value, "off"))
437 arg_pretimeout_watchdog = 0;
438 else {
439 r = parse_sec(value, &arg_pretimeout_watchdog);
440 if (r < 0) {
441 log_warning_errno(r, "Failed to parse systemd.watchdog_pre_sec= argument '%s', ignoring: %m", value);
442 return 0;
443 }
444 }
445
446 } else if (proc_cmdline_key_streq(key, "systemd.watchdog_pretimeout_governor")) {
447
448 if (proc_cmdline_value_missing(key, value) || isempty(value)) {
449 arg_watchdog_pretimeout_governor = mfree(arg_watchdog_pretimeout_governor);
450 return 0;
451 }
452
453 if (!string_is_safe(value)) {
454 log_warning("Watchdog pretimeout governor '%s' is not valid, ignoring.", value);
455 return 0;
456 }
457
458 return free_and_strdup_warn(&arg_watchdog_pretimeout_governor, value);
459
460 } else if (proc_cmdline_key_streq(key, "systemd.clock_usec")) {
461
462 if (proc_cmdline_value_missing(key, value))
463 return 0;
464
465 r = safe_atou64(value, &arg_clock_usec);
466 if (r < 0)
467 log_warning_errno(r, "Failed to parse systemd.clock_usec= argument, ignoring: %s", value);
468
469 } else if (proc_cmdline_key_streq(key, "systemd.random_seed")) {
470 void *p;
471 size_t sz;
472
473 if (proc_cmdline_value_missing(key, value))
474 return 0;
475
476 r = unbase64mem(value, SIZE_MAX, &p, &sz);
477 if (r < 0)
478 log_warning_errno(r, "Failed to parse systemd.random_seed= argument, ignoring: %s", value);
479
480 free(arg_random_seed);
481 arg_random_seed = sz > 0 ? p : mfree(p);
482 arg_random_seed_size = sz;
483
484 } else if (streq(key, "quiet") && !value) {
485
486 if (arg_show_status == _SHOW_STATUS_INVALID)
487 arg_show_status = SHOW_STATUS_ERROR;
488
489 } else if (streq(key, "debug") && !value) {
490
491 /* Note that log_parse_environment() handles 'debug'
492 * too, and sets the log level to LOG_DEBUG. */
493
494 if (detect_container() > 0)
495 log_set_target(LOG_TARGET_CONSOLE);
496
497 } else if (!value) {
498 const char *target;
499
500 /* Compatible with SysV, but supported independently even if SysV compatibility is disabled. */
501 target = runlevel_to_target(key);
502 if (target)
503 return free_and_strdup_warn(&arg_default_unit, target);
504 }
505
506 return 0;
507 }
508
509 #define DEFINE_SETTER(name, func, descr) \
510 static int name(const char *unit, \
511 const char *filename, \
512 unsigned line, \
513 const char *section, \
514 unsigned section_line, \
515 const char *lvalue, \
516 int ltype, \
517 const char *rvalue, \
518 void *data, \
519 void *userdata) { \
520 \
521 int r; \
522 \
523 assert(filename); \
524 assert(lvalue); \
525 assert(rvalue); \
526 \
527 r = func(rvalue); \
528 if (r < 0) \
529 log_syntax(unit, LOG_ERR, filename, line, r, \
530 "Invalid " descr "'%s': %m", \
531 rvalue); \
532 \
533 return 0; \
534 }
535
536 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level");
537 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target");
538 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color");
539 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location");
540 DEFINE_SETTER(config_parse_time, log_show_time_from_string, "time");
541
542 static int config_parse_default_timeout_abort(
543 const char *unit,
544 const char *filename,
545 unsigned line,
546 const char *section,
547 unsigned section_line,
548 const char *lvalue,
549 int ltype,
550 const char *rvalue,
551 void *data,
552 void *userdata) {
553 int r;
554
555 r = config_parse_timeout_abort(unit, filename, line, section, section_line, lvalue, ltype, rvalue,
556 &arg_default_timeout_abort_usec, userdata);
557 if (r >= 0)
558 arg_default_timeout_abort_set = r;
559 return 0;
560 }
561
562 static int config_parse_oom_score_adjust(
563 const char *unit,
564 const char *filename,
565 unsigned line,
566 const char *section,
567 unsigned section_line,
568 const char *lvalue,
569 int ltype,
570 const char *rvalue,
571 void *data,
572 void *userdata) {
573
574 int oa, r;
575
576 if (isempty(rvalue)) {
577 arg_default_oom_score_adjust_set = false;
578 return 0;
579 }
580
581 r = parse_oom_score_adjust(rvalue, &oa);
582 if (r < 0) {
583 log_syntax(unit, LOG_WARNING, filename, line, r, "Failed to parse the OOM score adjust value '%s', ignoring: %m", rvalue);
584 return 0;
585 }
586
587 arg_default_oom_score_adjust = oa;
588 arg_default_oom_score_adjust_set = true;
589
590 return 0;
591 }
592
593 static int parse_config_file(void) {
594 const ConfigTableItem items[] = {
595 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
596 { "Manager", "LogTarget", config_parse_target, 0, NULL },
597 { "Manager", "LogColor", config_parse_color, 0, NULL },
598 { "Manager", "LogLocation", config_parse_location, 0, NULL },
599 { "Manager", "LogTime", config_parse_time, 0, NULL },
600 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
601 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, &arg_crash_chvt },
602 { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, &arg_crash_chvt },
603 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
604 { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
605 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
606 { "Manager", "StatusUnitFormat", config_parse_status_unit_format, 0, &arg_status_unit_format },
607 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, &arg_cpu_affinity },
608 { "Manager", "NUMAPolicy", config_parse_numa_policy, 0, &arg_numa_policy.type },
609 { "Manager", "NUMAMask", config_parse_numa_mask, 0, &arg_numa_policy },
610 { "Manager", "JoinControllers", config_parse_warn_compat, DISABLED_CONFIGURATION, NULL },
611 { "Manager", "RuntimeWatchdogSec", config_parse_watchdog_sec, 0, &arg_runtime_watchdog },
612 { "Manager", "RuntimeWatchdogPreSec", config_parse_watchdog_sec, 0, &arg_pretimeout_watchdog },
613 { "Manager", "RebootWatchdogSec", config_parse_watchdog_sec, 0, &arg_reboot_watchdog },
614 { "Manager", "ShutdownWatchdogSec", config_parse_watchdog_sec, 0, &arg_reboot_watchdog }, /* obsolete alias */
615 { "Manager", "KExecWatchdogSec", config_parse_watchdog_sec, 0, &arg_kexec_watchdog },
616 { "Manager", "WatchdogDevice", config_parse_path, 0, &arg_watchdog_device },
617 { "Manager", "RuntimeWatchdogPreGovernor", config_parse_string, CONFIG_PARSE_STRING_SAFE, &arg_watchdog_pretimeout_governor },
618 { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
619 { "Manager", "NoNewPrivileges", config_parse_bool, 0, &arg_no_new_privs },
620 #if HAVE_SECCOMP
621 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
622 #endif
623 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
624 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
625 { "Manager", "DefaultStandardOutput", config_parse_output_restricted, 0, &arg_default_std_output },
626 { "Manager", "DefaultStandardError", config_parse_output_restricted, 0, &arg_default_std_error },
627 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
628 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
629 { "Manager", "DefaultTimeoutAbortSec", config_parse_default_timeout_abort, 0, NULL },
630 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
631 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */
632 { "Manager", "DefaultStartLimitIntervalSec", config_parse_sec, 0, &arg_default_start_limit_interval },
633 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
634 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
635 { "Manager", "ManagerEnvironment", config_parse_environ, 0, &arg_manager_environment },
636 { "Manager", "DefaultLimitCPU", config_parse_rlimit, RLIMIT_CPU, arg_default_rlimit },
637 { "Manager", "DefaultLimitFSIZE", config_parse_rlimit, RLIMIT_FSIZE, arg_default_rlimit },
638 { "Manager", "DefaultLimitDATA", config_parse_rlimit, RLIMIT_DATA, arg_default_rlimit },
639 { "Manager", "DefaultLimitSTACK", config_parse_rlimit, RLIMIT_STACK, arg_default_rlimit },
640 { "Manager", "DefaultLimitCORE", config_parse_rlimit, RLIMIT_CORE, arg_default_rlimit },
641 { "Manager", "DefaultLimitRSS", config_parse_rlimit, RLIMIT_RSS, arg_default_rlimit },
642 { "Manager", "DefaultLimitNOFILE", config_parse_rlimit, RLIMIT_NOFILE, arg_default_rlimit },
643 { "Manager", "DefaultLimitAS", config_parse_rlimit, RLIMIT_AS, arg_default_rlimit },
644 { "Manager", "DefaultLimitNPROC", config_parse_rlimit, RLIMIT_NPROC, arg_default_rlimit },
645 { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit, RLIMIT_MEMLOCK, arg_default_rlimit },
646 { "Manager", "DefaultLimitLOCKS", config_parse_rlimit, RLIMIT_LOCKS, arg_default_rlimit },
647 { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit, RLIMIT_SIGPENDING, arg_default_rlimit },
648 { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit, RLIMIT_MSGQUEUE, arg_default_rlimit },
649 { "Manager", "DefaultLimitNICE", config_parse_rlimit, RLIMIT_NICE, arg_default_rlimit },
650 { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit, RLIMIT_RTPRIO, arg_default_rlimit },
651 { "Manager", "DefaultLimitRTTIME", config_parse_rlimit, RLIMIT_RTTIME, arg_default_rlimit },
652 { "Manager", "DefaultCPUAccounting", config_parse_tristate, 0, &arg_default_cpu_accounting },
653 { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting },
654 { "Manager", "DefaultIPAccounting", config_parse_bool, 0, &arg_default_ip_accounting },
655 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
656 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
657 { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
658 { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max },
659 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action, 0, &arg_cad_burst_action },
660 { "Manager", "DefaultOOMPolicy", config_parse_oom_policy, 0, &arg_default_oom_policy },
661 { "Manager", "DefaultOOMScoreAdjust", config_parse_oom_score_adjust, 0, NULL },
662 #if ENABLE_SMACK
663 { "Manager", "DefaultSmackProcessLabel", config_parse_string, 0, &arg_default_smack_process_label },
664 #else
665 { "Manager", "DefaultSmackProcessLabel", config_parse_warn_compat, DISABLED_CONFIGURATION, NULL },
666 #endif
667 {}
668 };
669
670 _cleanup_strv_free_ char **files = NULL, **dirs = NULL;
671 const char *suffix;
672 int r;
673
674 if (arg_system)
675 suffix = "system.conf.d";
676 else {
677 r = manager_find_user_config_paths(&files, &dirs);
678 if (r < 0)
679 return log_error_errno(r, "Failed to determine config file paths: %m");
680
681 suffix = "user.conf.d";
682 }
683
684 (void) config_parse_many(
685 (const char* const*) (files ?: STRV_MAKE(PKGSYSCONFDIR "/system.conf")),
686 (const char* const*) (dirs ?: CONF_PATHS_STRV("systemd")),
687 suffix,
688 "Manager\0",
689 config_item_table_lookup, items,
690 CONFIG_PARSE_WARN,
691 NULL,
692 NULL);
693
694 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we use
695 * USEC_INFINITY like everywhere else. */
696 if (arg_default_timeout_start_usec <= 0)
697 arg_default_timeout_start_usec = USEC_INFINITY;
698 if (arg_default_timeout_stop_usec <= 0)
699 arg_default_timeout_stop_usec = USEC_INFINITY;
700
701 return 0;
702 }
703
704 static void set_manager_defaults(Manager *m) {
705
706 assert(m);
707
708 /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
709 * affect the manager itself, but are just what newly allocated units will have set if they haven't set
710 * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
711
712 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
713 m->default_std_output = arg_default_std_output;
714 m->default_std_error = arg_default_std_error;
715 m->default_timeout_start_usec = arg_default_timeout_start_usec;
716 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
717 m->default_timeout_abort_usec = arg_default_timeout_abort_usec;
718 m->default_timeout_abort_set = arg_default_timeout_abort_set;
719 m->default_restart_usec = arg_default_restart_usec;
720 m->default_start_limit_interval = arg_default_start_limit_interval;
721 m->default_start_limit_burst = arg_default_start_limit_burst;
722
723 /* On 4.15+ with unified hierarchy, CPU accounting is essentially free as it doesn't require the CPU
724 * controller to be enabled, so the default is to enable it unless we got told otherwise. */
725 if (arg_default_cpu_accounting >= 0)
726 m->default_cpu_accounting = arg_default_cpu_accounting;
727 else
728 m->default_cpu_accounting = cpu_accounting_is_cheap();
729
730 m->default_io_accounting = arg_default_io_accounting;
731 m->default_ip_accounting = arg_default_ip_accounting;
732 m->default_blockio_accounting = arg_default_blockio_accounting;
733 m->default_memory_accounting = arg_default_memory_accounting;
734 m->default_tasks_accounting = arg_default_tasks_accounting;
735 m->default_tasks_max = arg_default_tasks_max;
736 m->default_oom_policy = arg_default_oom_policy;
737 m->default_oom_score_adjust_set = arg_default_oom_score_adjust_set;
738 m->default_oom_score_adjust = arg_default_oom_score_adjust;
739
740 (void) manager_set_default_smack_process_label(m, arg_default_smack_process_label);
741
742 (void) manager_set_default_rlimits(m, arg_default_rlimit);
743
744 (void) manager_default_environment(m);
745 (void) manager_transient_environment_add(m, arg_default_environment);
746 }
747
748 static void set_manager_settings(Manager *m) {
749 int r;
750
751 assert(m);
752
753 /* Propagates the various manager settings into the manager object, i.e. properties that
754 * effect the manager itself (as opposed to just being inherited into newly allocated
755 * units, see set_manager_defaults() above). */
756
757 m->confirm_spawn = arg_confirm_spawn;
758 m->service_watchdogs = arg_service_watchdogs;
759 m->cad_burst_action = arg_cad_burst_action;
760
761 manager_set_watchdog(m, WATCHDOG_RUNTIME, arg_runtime_watchdog);
762 manager_set_watchdog(m, WATCHDOG_REBOOT, arg_reboot_watchdog);
763 manager_set_watchdog(m, WATCHDOG_KEXEC, arg_kexec_watchdog);
764 manager_set_watchdog(m, WATCHDOG_PRETIMEOUT, arg_pretimeout_watchdog);
765 r = manager_set_watchdog_pretimeout_governor(m, arg_watchdog_pretimeout_governor);
766 if (r < 0)
767 log_warning_errno(r, "Failed to set watchdog pretimeout governor to '%s', ignoring: %m", arg_watchdog_pretimeout_governor);
768
769 manager_set_show_status(m, arg_show_status, "commandline");
770 m->status_unit_format = arg_status_unit_format;
771 }
772
773 static int parse_argv(int argc, char *argv[]) {
774 enum {
775 ARG_LOG_LEVEL = 0x100,
776 ARG_LOG_TARGET,
777 ARG_LOG_COLOR,
778 ARG_LOG_LOCATION,
779 ARG_LOG_TIME,
780 ARG_UNIT,
781 ARG_SYSTEM,
782 ARG_USER,
783 ARG_TEST,
784 ARG_NO_PAGER,
785 ARG_VERSION,
786 ARG_DUMP_CONFIGURATION_ITEMS,
787 ARG_DUMP_BUS_PROPERTIES,
788 ARG_BUS_INTROSPECT,
789 ARG_DUMP_CORE,
790 ARG_CRASH_CHVT,
791 ARG_CRASH_SHELL,
792 ARG_CRASH_REBOOT,
793 ARG_CONFIRM_SPAWN,
794 ARG_SHOW_STATUS,
795 ARG_DESERIALIZE,
796 ARG_SWITCHED_ROOT,
797 ARG_DEFAULT_STD_OUTPUT,
798 ARG_DEFAULT_STD_ERROR,
799 ARG_MACHINE_ID,
800 ARG_SERVICE_WATCHDOGS,
801 };
802
803 static const struct option options[] = {
804 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
805 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
806 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
807 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
808 { "log-time", optional_argument, NULL, ARG_LOG_TIME },
809 { "unit", required_argument, NULL, ARG_UNIT },
810 { "system", no_argument, NULL, ARG_SYSTEM },
811 { "user", no_argument, NULL, ARG_USER },
812 { "test", no_argument, NULL, ARG_TEST },
813 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
814 { "help", no_argument, NULL, 'h' },
815 { "version", no_argument, NULL, ARG_VERSION },
816 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
817 { "dump-bus-properties", no_argument, NULL, ARG_DUMP_BUS_PROPERTIES },
818 { "bus-introspect", required_argument, NULL, ARG_BUS_INTROSPECT },
819 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
820 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
821 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
822 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
823 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
824 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
825 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
826 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
827 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
828 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
829 { "machine-id", required_argument, NULL, ARG_MACHINE_ID },
830 { "service-watchdogs", required_argument, NULL, ARG_SERVICE_WATCHDOGS },
831 {}
832 };
833
834 int c, r;
835 bool user_arg_seen = false;
836
837 assert(argc >= 1);
838 assert(argv);
839
840 if (getpid_cached() == 1)
841 opterr = 0;
842
843 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
844
845 switch (c) {
846
847 case ARG_LOG_LEVEL:
848 r = log_set_max_level_from_string(optarg);
849 if (r < 0)
850 return log_error_errno(r, "Failed to parse log level \"%s\": %m", optarg);
851
852 break;
853
854 case ARG_LOG_TARGET:
855 r = log_set_target_from_string(optarg);
856 if (r < 0)
857 return log_error_errno(r, "Failed to parse log target \"%s\": %m", optarg);
858
859 break;
860
861 case ARG_LOG_COLOR:
862
863 if (optarg) {
864 r = log_show_color_from_string(optarg);
865 if (r < 0)
866 return log_error_errno(r, "Failed to parse log color setting \"%s\": %m",
867 optarg);
868 } else
869 log_show_color(true);
870
871 break;
872
873 case ARG_LOG_LOCATION:
874 if (optarg) {
875 r = log_show_location_from_string(optarg);
876 if (r < 0)
877 return log_error_errno(r, "Failed to parse log location setting \"%s\": %m",
878 optarg);
879 } else
880 log_show_location(true);
881
882 break;
883
884 case ARG_LOG_TIME:
885
886 if (optarg) {
887 r = log_show_time_from_string(optarg);
888 if (r < 0)
889 return log_error_errno(r, "Failed to parse log time setting \"%s\": %m",
890 optarg);
891 } else
892 log_show_time(true);
893
894 break;
895
896 case ARG_DEFAULT_STD_OUTPUT:
897 r = exec_output_from_string(optarg);
898 if (r < 0)
899 return log_error_errno(r, "Failed to parse default standard output setting \"%s\": %m",
900 optarg);
901 arg_default_std_output = r;
902 break;
903
904 case ARG_DEFAULT_STD_ERROR:
905 r = exec_output_from_string(optarg);
906 if (r < 0)
907 return log_error_errno(r, "Failed to parse default standard error output setting \"%s\": %m",
908 optarg);
909 arg_default_std_error = r;
910 break;
911
912 case ARG_UNIT:
913 r = free_and_strdup(&arg_default_unit, optarg);
914 if (r < 0)
915 return log_error_errno(r, "Failed to set default unit \"%s\": %m", optarg);
916
917 break;
918
919 case ARG_SYSTEM:
920 arg_system = true;
921 break;
922
923 case ARG_USER:
924 arg_system = false;
925 user_arg_seen = true;
926 break;
927
928 case ARG_TEST:
929 arg_action = ACTION_TEST;
930 break;
931
932 case ARG_NO_PAGER:
933 arg_pager_flags |= PAGER_DISABLE;
934 break;
935
936 case ARG_VERSION:
937 arg_action = ACTION_VERSION;
938 break;
939
940 case ARG_DUMP_CONFIGURATION_ITEMS:
941 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
942 break;
943
944 case ARG_DUMP_BUS_PROPERTIES:
945 arg_action = ACTION_DUMP_BUS_PROPERTIES;
946 break;
947
948 case ARG_BUS_INTROSPECT:
949 arg_bus_introspect = optarg;
950 arg_action = ACTION_BUS_INTROSPECT;
951 break;
952
953 case ARG_DUMP_CORE:
954 r = parse_boolean_argument("--dump-core", optarg, &arg_dump_core);
955 if (r < 0)
956 return r;
957 break;
958
959 case ARG_CRASH_CHVT:
960 r = parse_crash_chvt(optarg, &arg_crash_chvt);
961 if (r < 0)
962 return log_error_errno(r, "Failed to parse crash virtual terminal index: \"%s\": %m",
963 optarg);
964 break;
965
966 case ARG_CRASH_SHELL:
967 r = parse_boolean_argument("--crash-shell", optarg, &arg_crash_shell);
968 if (r < 0)
969 return r;
970 break;
971
972 case ARG_CRASH_REBOOT:
973 r = parse_boolean_argument("--crash-reboot", optarg, &arg_crash_reboot);
974 if (r < 0)
975 return r;
976 break;
977
978 case ARG_CONFIRM_SPAWN:
979 arg_confirm_spawn = mfree(arg_confirm_spawn);
980
981 r = parse_confirm_spawn(optarg, &arg_confirm_spawn);
982 if (r < 0)
983 return log_error_errno(r, "Failed to parse confirm spawn option: \"%s\": %m",
984 optarg);
985 break;
986
987 case ARG_SERVICE_WATCHDOGS:
988 r = parse_boolean_argument("--service-watchdogs=", optarg, &arg_service_watchdogs);
989 if (r < 0)
990 return r;
991 break;
992
993 case ARG_SHOW_STATUS:
994 if (optarg) {
995 r = parse_show_status(optarg, &arg_show_status);
996 if (r < 0)
997 return log_error_errno(r, "Failed to parse show status boolean: \"%s\": %m",
998 optarg);
999 } else
1000 arg_show_status = SHOW_STATUS_YES;
1001 break;
1002
1003 case ARG_DESERIALIZE: {
1004 int fd;
1005 FILE *f;
1006
1007 r = safe_atoi(optarg, &fd);
1008 if (r < 0)
1009 log_error_errno(r, "Failed to parse deserialize option \"%s\": %m", optarg);
1010 if (fd < 0)
1011 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1012 "Invalid deserialize fd: %d",
1013 fd);
1014
1015 (void) fd_cloexec(fd, true);
1016
1017 f = fdopen(fd, "r");
1018 if (!f)
1019 return log_error_errno(errno, "Failed to open serialization fd %d: %m", fd);
1020
1021 safe_fclose(arg_serialization);
1022 arg_serialization = f;
1023
1024 break;
1025 }
1026
1027 case ARG_SWITCHED_ROOT:
1028 arg_switched_root = true;
1029 break;
1030
1031 case ARG_MACHINE_ID:
1032 r = set_machine_id(optarg);
1033 if (r < 0)
1034 return log_error_errno(r, "MachineID '%s' is not valid: %m", optarg);
1035 break;
1036
1037 case 'h':
1038 arg_action = ACTION_HELP;
1039 break;
1040
1041 case 'D':
1042 log_set_max_level(LOG_DEBUG);
1043 break;
1044
1045 case 'b':
1046 case 's':
1047 case 'z':
1048 /* Just to eat away the sysvinit kernel cmdline args that we'll parse in
1049 * parse_proc_cmdline_item() or ignore, without any getopt() error messages.
1050 */
1051 case '?':
1052 if (getpid_cached() != 1)
1053 return -EINVAL;
1054 else
1055 return 0;
1056
1057 default:
1058 assert_not_reached();
1059 }
1060
1061 if (optind < argc && getpid_cached() != 1)
1062 /* Hmm, when we aren't run as init system let's complain about excess arguments */
1063 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Excess arguments.");
1064
1065 if (arg_action == ACTION_RUN && !arg_system && !user_arg_seen)
1066 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1067 "Explicit --user argument required to run as user manager.");
1068
1069 return 0;
1070 }
1071
1072 static int help(void) {
1073 _cleanup_free_ char *link = NULL;
1074 int r;
1075
1076 r = terminal_urlify_man("systemd", "1", &link);
1077 if (r < 0)
1078 return log_oom();
1079
1080 printf("%s [OPTIONS...]\n\n"
1081 "%sStarts and monitors system and user services.%s\n\n"
1082 "This program takes no positional arguments.\n\n"
1083 "%sOptions%s:\n"
1084 " -h --help Show this help\n"
1085 " --version Show version\n"
1086 " --test Determine initial transaction, dump it and exit\n"
1087 " --system Combined with --test: operate in system mode\n"
1088 " --user Combined with --test: operate in user mode\n"
1089 " --dump-configuration-items Dump understood unit configuration items\n"
1090 " --dump-bus-properties Dump exposed bus properties\n"
1091 " --bus-introspect=PATH Write XML introspection data\n"
1092 " --unit=UNIT Set default unit\n"
1093 " --dump-core[=BOOL] Dump core on crash\n"
1094 " --crash-vt=NR Change to specified VT on crash\n"
1095 " --crash-reboot[=BOOL] Reboot on crash\n"
1096 " --crash-shell[=BOOL] Run shell on crash\n"
1097 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1098 " --show-status[=BOOL] Show status updates on the console during boot\n"
1099 " --log-target=TARGET Set log target (console, journal, kmsg,\n"
1100 " journal-or-kmsg, null)\n"
1101 " --log-level=LEVEL Set log level (debug, info, notice, warning,\n"
1102 " err, crit, alert, emerg)\n"
1103 " --log-color[=BOOL] Highlight important log messages\n"
1104 " --log-location[=BOOL] Include code location in log messages\n"
1105 " --log-time[=BOOL] Prefix log messages with current time\n"
1106 " --default-standard-output= Set default standard output for services\n"
1107 " --default-standard-error= Set default standard error output for services\n"
1108 " --no-pager Do not pipe output into a pager\n"
1109 "\nSee the %s for details.\n",
1110 program_invocation_short_name,
1111 ansi_highlight(),
1112 ansi_normal(),
1113 ansi_underline(),
1114 ansi_normal(),
1115 link);
1116
1117 return 0;
1118 }
1119
1120 static int prepare_reexecute(
1121 Manager *m,
1122 FILE **ret_f,
1123 FDSet **ret_fds,
1124 bool switching_root) {
1125
1126 _cleanup_fdset_free_ FDSet *fds = NULL;
1127 _cleanup_fclose_ FILE *f = NULL;
1128 int r;
1129
1130 assert(m);
1131 assert(ret_f);
1132 assert(ret_fds);
1133
1134 r = manager_open_serialization(m, &f);
1135 if (r < 0)
1136 return log_error_errno(r, "Failed to create serialization file: %m");
1137
1138 /* Make sure nothing is really destructed when we shut down */
1139 m->n_reloading++;
1140 bus_manager_send_reloading(m, true);
1141
1142 fds = fdset_new();
1143 if (!fds)
1144 return log_oom();
1145
1146 r = manager_serialize(m, f, fds, switching_root);
1147 if (r < 0)
1148 return r;
1149
1150 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1151 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1152
1153 r = fd_cloexec(fileno(f), false);
1154 if (r < 0)
1155 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1156
1157 r = fdset_cloexec(fds, false);
1158 if (r < 0)
1159 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1160
1161 *ret_f = TAKE_PTR(f);
1162 *ret_fds = TAKE_PTR(fds);
1163
1164 return 0;
1165 }
1166
1167 static void bump_file_max_and_nr_open(void) {
1168
1169 /* Let's bump fs.file-max and fs.nr_open to their respective maximums. On current kernels large
1170 * numbers of file descriptors are no longer a performance problem and their memory is properly
1171 * tracked by memcg, thus counting them and limiting them in another two layers of limits is
1172 * unnecessary and just complicates things. This function hence turns off 2 of the 4 levels of limits
1173 * on file descriptors, and makes RLIMIT_NOLIMIT (soft + hard) the only ones that really matter. */
1174
1175 #if BUMP_PROC_SYS_FS_FILE_MAX || BUMP_PROC_SYS_FS_NR_OPEN
1176 int r;
1177 #endif
1178
1179 #if BUMP_PROC_SYS_FS_FILE_MAX
1180 /* The maximum the kernel allows for this since 5.2 is LONG_MAX, use that. (Previously things were
1181 * different, but the operation would fail silently.) */
1182 r = sysctl_writef("fs/file-max", "%li\n", LONG_MAX);
1183 if (r < 0)
1184 log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.file-max, ignoring: %m");
1185 #endif
1186
1187 #if BUMP_PROC_SYS_FS_NR_OPEN
1188 int v = INT_MAX;
1189
1190 /* Argh! The kernel enforces maximum and minimum values on the fs.nr_open, but we don't really know
1191 * what they are. The expression by which the maximum is determined is dependent on the architecture,
1192 * and is something we don't really want to copy to userspace, as it is dependent on implementation
1193 * details of the kernel. Since the kernel doesn't expose the maximum value to us, we can only try
1194 * and hope. Hence, let's start with INT_MAX, and then keep halving the value until we find one that
1195 * works. Ugly? Yes, absolutely, but kernel APIs are kernel APIs, so what do can we do... 🤯 */
1196
1197 for (;;) {
1198 int k;
1199
1200 v &= ~(__SIZEOF_POINTER__ - 1); /* Round down to next multiple of the pointer size */
1201 if (v < 1024) {
1202 log_warning("Can't bump fs.nr_open, value too small.");
1203 break;
1204 }
1205
1206 k = read_nr_open();
1207 if (k < 0) {
1208 log_error_errno(k, "Failed to read fs.nr_open: %m");
1209 break;
1210 }
1211 if (k >= v) { /* Already larger */
1212 log_debug("Skipping bump, value is already larger.");
1213 break;
1214 }
1215
1216 r = sysctl_writef("fs/nr_open", "%i\n", v);
1217 if (r == -EINVAL) {
1218 log_debug("Couldn't write fs.nr_open as %i, halving it.", v);
1219 v /= 2;
1220 continue;
1221 }
1222 if (r < 0) {
1223 log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.nr_open, ignoring: %m");
1224 break;
1225 }
1226
1227 log_debug("Successfully bumped fs.nr_open to %i", v);
1228 break;
1229 }
1230 #endif
1231 }
1232
1233 static int bump_rlimit_nofile(const struct rlimit *saved_rlimit) {
1234 struct rlimit new_rlimit;
1235 int r, nr;
1236
1237 /* Get the underlying absolute limit the kernel enforces */
1238 nr = read_nr_open();
1239
1240 /* Calculate the new limits to use for us. Never lower from what we inherited. */
1241 new_rlimit = (struct rlimit) {
1242 .rlim_cur = MAX((rlim_t) nr, saved_rlimit->rlim_cur),
1243 .rlim_max = MAX((rlim_t) nr, saved_rlimit->rlim_max),
1244 };
1245
1246 /* Shortcut if nothing changes. */
1247 if (saved_rlimit->rlim_max >= new_rlimit.rlim_max &&
1248 saved_rlimit->rlim_cur >= new_rlimit.rlim_cur) {
1249 log_debug("RLIMIT_NOFILE is already as high or higher than we need it, not bumping.");
1250 return 0;
1251 }
1252
1253 /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for
1254 * both hard and soft. */
1255 r = setrlimit_closest(RLIMIT_NOFILE, &new_rlimit);
1256 if (r < 0)
1257 return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1258
1259 return 0;
1260 }
1261
1262 static int bump_rlimit_memlock(const struct rlimit *saved_rlimit) {
1263 struct rlimit new_rlimit;
1264 uint64_t mm;
1265 int r;
1266
1267 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK
1268 * which should normally disable such checks. We need them to implement IPAddressAllow= and
1269 * IPAddressDeny=, hence let's bump the value high enough for our user. */
1270
1271 /* Using MAX() on resource limits only is safe if RLIM_INFINITY is > 0. POSIX declares that rlim_t
1272 * must be unsigned, hence this is a given, but let's make this clear here. */
1273 assert_cc(RLIM_INFINITY > 0);
1274
1275 mm = physical_memory_scale(1, 8); /* Let's scale how much we allow to be locked by the amount of
1276 * physical RAM. We allow an eighth to be locked by us, just to
1277 * pick a value. */
1278
1279 new_rlimit = (struct rlimit) {
1280 .rlim_cur = MAX3(HIGH_RLIMIT_MEMLOCK, saved_rlimit->rlim_cur, mm),
1281 .rlim_max = MAX3(HIGH_RLIMIT_MEMLOCK, saved_rlimit->rlim_max, mm),
1282 };
1283
1284 if (saved_rlimit->rlim_max >= new_rlimit.rlim_cur &&
1285 saved_rlimit->rlim_cur >= new_rlimit.rlim_max) {
1286 log_debug("RLIMIT_MEMLOCK is already as high or higher than we need it, not bumping.");
1287 return 0;
1288 }
1289
1290 r = setrlimit_closest(RLIMIT_MEMLOCK, &new_rlimit);
1291 if (r < 0)
1292 return log_warning_errno(r, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1293
1294 return 0;
1295 }
1296
1297 static void test_usr(void) {
1298
1299 /* Check that /usr is either on the same file system as / or mounted already. */
1300
1301 if (dir_is_empty("/usr", /* ignore_hidden_or_backup= */ false) <= 0)
1302 return;
1303
1304 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1305 "Some things will probably break (sometimes even silently) in mysterious ways. "
1306 "Consult https://www.freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1307 }
1308
1309 static int enforce_syscall_archs(Set *archs) {
1310 #if HAVE_SECCOMP
1311 int r;
1312
1313 if (!is_seccomp_available())
1314 return 0;
1315
1316 r = seccomp_restrict_archs(arg_syscall_archs);
1317 if (r < 0)
1318 return log_error_errno(r, "Failed to enforce system call architecture restrication: %m");
1319 #endif
1320 return 0;
1321 }
1322
1323 static int os_release_status(void) {
1324 _cleanup_free_ char *pretty_name = NULL, *name = NULL, *version = NULL,
1325 *ansi_color = NULL, *support_end = NULL;
1326 int r;
1327
1328 r = parse_os_release(NULL,
1329 "PRETTY_NAME", &pretty_name,
1330 "NAME", &name,
1331 "VERSION", &version,
1332 "ANSI_COLOR", &ansi_color,
1333 "SUPPORT_END", &support_end);
1334 if (r < 0)
1335 return log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r,
1336 "Failed to read os-release file, ignoring: %m");
1337
1338 const char *label = empty_to_null(pretty_name) ?: empty_to_null(name) ?: "Linux";
1339
1340 if (show_status_on(arg_show_status)) {
1341 if (log_get_show_color())
1342 status_printf(NULL, 0,
1343 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1344 empty_to_null(ansi_color) ?: "1",
1345 label);
1346 else
1347 status_printf(NULL, 0,
1348 "\nWelcome to %s!\n",
1349 label);
1350 }
1351
1352 if (support_end && os_release_support_ended(support_end, false) > 0)
1353 /* pretty_name may include the version already, so we'll print the version only if we
1354 * have it and we're not using pretty_name. */
1355 status_printf(ANSI_HIGHLIGHT_RED " !! " ANSI_NORMAL, 0,
1356 "This OS version (%s%s%s) is past its end-of-support date (%s)",
1357 label,
1358 (pretty_name || !version) ? "" : " version ",
1359 (pretty_name || !version) ? "" : version,
1360 support_end);
1361
1362 return 0;
1363 }
1364
1365 static int write_container_id(void) {
1366 const char *c;
1367 int r = 0; /* avoid false maybe-uninitialized warning */
1368
1369 c = getenv("container");
1370 if (isempty(c))
1371 return 0;
1372
1373 RUN_WITH_UMASK(0022)
1374 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1375 if (r < 0)
1376 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
1377
1378 return 1;
1379 }
1380
1381 static int bump_unix_max_dgram_qlen(void) {
1382 _cleanup_free_ char *qlen = NULL;
1383 unsigned long v;
1384 int r;
1385
1386 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set
1387 * the value really really early during boot, so that it is actually applied to all our sockets,
1388 * including the $NOTIFY_SOCKET one. */
1389
1390 r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1391 if (r < 0)
1392 return log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r,
1393 "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1394
1395 r = safe_atolu(qlen, &v);
1396 if (r < 0)
1397 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen);
1398
1399 if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1400 return 0;
1401
1402 r = write_string_filef("/proc/sys/net/unix/max_dgram_qlen", WRITE_STRING_FILE_DISABLE_BUFFER,
1403 "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN);
1404 if (r < 0)
1405 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1406 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1407
1408 return 1;
1409 }
1410
1411 static int fixup_environment(void) {
1412 _cleanup_free_ char *term = NULL;
1413 const char *t;
1414 int r;
1415
1416 /* Only fix up the environment when we are started as PID 1 */
1417 if (getpid_cached() != 1)
1418 return 0;
1419
1420 /* We expect the environment to be set correctly if run inside a container. */
1421 if (detect_container() > 0)
1422 return 0;
1423
1424 /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the
1425 * backend device used by the console. We try to make a better guess here since some consoles might
1426 * not have support for color mode for example.
1427 *
1428 * However if TERM was configured through the kernel command line then leave it alone. */
1429 r = proc_cmdline_get_key("TERM", 0, &term);
1430 if (r < 0)
1431 return r;
1432
1433 t = term ?: default_term_for_tty("/dev/console");
1434
1435 if (setenv("TERM", t, 1) < 0)
1436 return -errno;
1437
1438 /* The kernels sets HOME=/ for init. Let's undo this. */
1439 if (path_equal_ptr(getenv("HOME"), "/"))
1440 assert_se(unsetenv("HOME") == 0);
1441
1442 return 0;
1443 }
1444
1445 static void redirect_telinit(int argc, char *argv[]) {
1446
1447 /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1448
1449 #if HAVE_SYSV_COMPAT
1450 if (getpid_cached() == 1)
1451 return;
1452
1453 if (!invoked_as(argv, "init"))
1454 return;
1455
1456 execv(SYSTEMCTL_BINARY_PATH, argv);
1457 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1458 exit(EXIT_FAILURE);
1459 #endif
1460 }
1461
1462 static int become_shutdown(
1463 const char *shutdown_verb,
1464 int retval) {
1465
1466 char log_level[DECIMAL_STR_MAX(int) + 1],
1467 exit_code[DECIMAL_STR_MAX(uint8_t) + 1],
1468 timeout[DECIMAL_STR_MAX(usec_t) + 1];
1469
1470 const char* command_line[13] = {
1471 SYSTEMD_SHUTDOWN_BINARY_PATH,
1472 shutdown_verb,
1473 "--timeout", timeout,
1474 "--log-level", log_level,
1475 "--log-target",
1476 };
1477
1478 _cleanup_strv_free_ char **env_block = NULL;
1479 usec_t watchdog_timer = 0;
1480 size_t pos = 7;
1481 int r;
1482
1483 assert(shutdown_verb);
1484 assert(!command_line[pos]);
1485 env_block = strv_copy(environ);
1486
1487 xsprintf(log_level, "%d", log_get_max_level());
1488 xsprintf(timeout, "%" PRI_USEC "us", arg_default_timeout_stop_usec);
1489
1490 switch (log_get_target()) {
1491
1492 case LOG_TARGET_KMSG:
1493 case LOG_TARGET_JOURNAL_OR_KMSG:
1494 case LOG_TARGET_SYSLOG_OR_KMSG:
1495 command_line[pos++] = "kmsg";
1496 break;
1497
1498 case LOG_TARGET_NULL:
1499 command_line[pos++] = "null";
1500 break;
1501
1502 case LOG_TARGET_CONSOLE:
1503 default:
1504 command_line[pos++] = "console";
1505 break;
1506 };
1507
1508 if (log_get_show_color())
1509 command_line[pos++] = "--log-color";
1510
1511 if (log_get_show_location())
1512 command_line[pos++] = "--log-location";
1513
1514 if (log_get_show_time())
1515 command_line[pos++] = "--log-time";
1516
1517 if (streq(shutdown_verb, "exit")) {
1518 command_line[pos++] = "--exit-code";
1519 command_line[pos++] = exit_code;
1520 xsprintf(exit_code, "%d", retval);
1521 }
1522
1523 assert(pos < ELEMENTSOF(command_line));
1524
1525 if (streq(shutdown_verb, "reboot"))
1526 watchdog_timer = arg_reboot_watchdog;
1527 else if (streq(shutdown_verb, "kexec"))
1528 watchdog_timer = arg_kexec_watchdog;
1529
1530 /* If we reboot or kexec let's set the shutdown watchdog and tell the
1531 * shutdown binary to repeatedly ping it.
1532 * Disable the pretimeout watchdog, as we do not support it from the shutdown binary. */
1533 (void) watchdog_setup_pretimeout(0);
1534 (void) watchdog_setup_pretimeout_governor(NULL);
1535 r = watchdog_setup(watchdog_timer);
1536 watchdog_close(r < 0);
1537
1538 /* Tell the binary how often to ping, ignore failure */
1539 (void) strv_extendf(&env_block, "WATCHDOG_USEC="USEC_FMT, watchdog_timer);
1540
1541 if (arg_watchdog_device)
1542 (void) strv_extendf(&env_block, "WATCHDOG_DEVICE=%s", arg_watchdog_device);
1543
1544 /* Avoid the creation of new processes forked by the kernel; at this
1545 * point, we will not listen to the signals anyway */
1546 if (detect_container() <= 0)
1547 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
1548
1549 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
1550 return -errno;
1551 }
1552
1553 static void initialize_clock(void) {
1554 int r;
1555
1556 /* This is called very early on, before we parse the kernel command line or otherwise figure out why
1557 * we are running, but only once. */
1558
1559 if (clock_is_localtime(NULL) > 0) {
1560 int min;
1561
1562 /* The very first call of settimeofday() also does a time warp in the kernel.
1563 *
1564 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to
1565 * take care of maintaining the RTC and do all adjustments. This matches the behavior of
1566 * Windows, which leaves the RTC alone if the registry tells that the RTC runs in UTC.
1567 */
1568 r = clock_set_timezone(&min);
1569 if (r < 0)
1570 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1571 else
1572 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1573
1574 } else if (!in_initrd())
1575 /*
1576 * Do a dummy very first call to seal the kernel's time warp magic.
1577 *
1578 * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with
1579 * LOCAL, but the real system could be set up that way. In such case, we need to delay the
1580 * time-warp or the sealing until we reach the real system.
1581 *
1582 * Do no set the kernel's timezone. The concept of local time cannot be supported reliably,
1583 * the time will jump or be incorrect at every daylight saving time change. All kernel local
1584 * time concepts will be treated as UTC that way.
1585 */
1586 (void) clock_reset_timewarp();
1587
1588 ClockChangeDirection change_dir;
1589 r = clock_apply_epoch(&change_dir);
1590 if (r > 0 && change_dir == CLOCK_CHANGE_FORWARD)
1591 log_info("System time before build time, advancing clock.");
1592 else if (r > 0 && change_dir == CLOCK_CHANGE_BACKWARD)
1593 log_info("System time is further ahead than %s after build time, resetting clock to build time.",
1594 FORMAT_TIMESPAN(CLOCK_VALID_RANGE_USEC_MAX, USEC_PER_DAY));
1595 else if (r < 0 && change_dir == CLOCK_CHANGE_FORWARD)
1596 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1597 else if (r < 0 && change_dir == CLOCK_CHANGE_BACKWARD)
1598 log_error_errno(r, "Current system time is further ahead %s after build time, but cannot correct: %m",
1599 FORMAT_TIMESPAN(CLOCK_VALID_RANGE_USEC_MAX, USEC_PER_DAY));
1600 }
1601
1602 static void apply_clock_update(void) {
1603 /* This is called later than initialize_clock(), i.e. after we parsed configuration files/kernel
1604 * command line and such. */
1605
1606 if (arg_clock_usec == 0)
1607 return;
1608
1609 if (getpid_cached() != 1)
1610 return;
1611
1612 if (clock_settime(CLOCK_REALTIME, TIMESPEC_STORE(arg_clock_usec)) < 0)
1613 log_error_errno(errno, "Failed to set system clock to time specified on kernel command line: %m");
1614 else
1615 log_info("Set system clock to %s, as specified on the kernel command line.",
1616 FORMAT_TIMESTAMP(arg_clock_usec));
1617 }
1618
1619 static void cmdline_take_random_seed(void) {
1620 size_t suggested;
1621 int r;
1622
1623 if (arg_random_seed_size == 0)
1624 return;
1625
1626 if (getpid_cached() != 1)
1627 return;
1628
1629 assert(arg_random_seed);
1630 suggested = random_pool_size();
1631
1632 if (arg_random_seed_size < suggested)
1633 log_warning("Random seed specified on kernel command line has size %zu, but %zu bytes required to fill entropy pool.",
1634 arg_random_seed_size, suggested);
1635
1636 r = random_write_entropy(-1, arg_random_seed, arg_random_seed_size, true);
1637 if (r < 0) {
1638 log_warning_errno(r, "Failed to credit entropy specified on kernel command line, ignoring: %m");
1639 return;
1640 }
1641
1642 log_notice("Successfully credited entropy passed on kernel command line.\n"
1643 "Note that the seed provided this way is accessible to unprivileged programs. "
1644 "This functionality should not be used outside of testing environments.");
1645 }
1646
1647 static void initialize_coredump(bool skip_setup) {
1648 #if ENABLE_COREDUMP
1649 if (getpid_cached() != 1)
1650 return;
1651
1652 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour
1653 * the limit) will process core dumps for system services by default. */
1654 if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0)
1655 log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m");
1656
1657 /* But at the same time, turn off the core_pattern logic by default, so that no coredumps are stored
1658 * until the systemd-coredump tool is enabled via sysctl. However it can be changed via the kernel
1659 * command line later so core dumps can still be generated during early startup and in initramfs. */
1660 if (!skip_setup)
1661 disable_coredumps();
1662 #endif
1663 }
1664
1665 static void initialize_core_pattern(bool skip_setup) {
1666 int r;
1667
1668 if (skip_setup || !arg_early_core_pattern)
1669 return;
1670
1671 if (getpid_cached() != 1)
1672 return;
1673
1674 r = write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern, WRITE_STRING_FILE_DISABLE_BUFFER);
1675 if (r < 0)
1676 log_warning_errno(r, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m",
1677 arg_early_core_pattern);
1678 }
1679
1680 static void update_cpu_affinity(bool skip_setup) {
1681 _cleanup_free_ char *mask = NULL;
1682
1683 if (skip_setup || !arg_cpu_affinity.set)
1684 return;
1685
1686 assert(arg_cpu_affinity.allocated > 0);
1687
1688 mask = cpu_set_to_range_string(&arg_cpu_affinity);
1689 log_debug("Setting CPU affinity to {%s}.", strnull(mask));
1690
1691 if (sched_setaffinity(0, arg_cpu_affinity.allocated, arg_cpu_affinity.set) < 0)
1692 log_warning_errno(errno, "Failed to set CPU affinity, ignoring: %m");
1693 }
1694
1695 static void update_numa_policy(bool skip_setup) {
1696 int r;
1697 _cleanup_free_ char *nodes = NULL;
1698 const char * policy = NULL;
1699
1700 if (skip_setup || !mpol_is_valid(numa_policy_get_type(&arg_numa_policy)))
1701 return;
1702
1703 if (DEBUG_LOGGING) {
1704 policy = mpol_to_string(numa_policy_get_type(&arg_numa_policy));
1705 nodes = cpu_set_to_range_string(&arg_numa_policy.nodes);
1706 log_debug("Setting NUMA policy to %s, with nodes {%s}.", strnull(policy), strnull(nodes));
1707 }
1708
1709 r = apply_numa_policy(&arg_numa_policy);
1710 if (r == -EOPNOTSUPP)
1711 log_debug_errno(r, "NUMA support not available, ignoring.");
1712 else if (r < 0)
1713 log_warning_errno(r, "Failed to set NUMA memory policy, ignoring: %m");
1714 }
1715
1716 static void filter_args(
1717 const char* dst[],
1718 size_t *dst_index,
1719 char **src,
1720 int argc) {
1721
1722 assert(dst);
1723 assert(dst_index);
1724
1725 /* Copy some filtered arguments into the dst array from src. */
1726 for (int i = 1; i < argc; i++) {
1727 if (STR_IN_SET(src[i],
1728 "--switched-root",
1729 "--system",
1730 "--user"))
1731 continue;
1732
1733 if (startswith(src[i], "--deserialize="))
1734 continue;
1735 if (streq(src[i], "--deserialize")) {
1736 i++; /* Skip the argument too */
1737 continue;
1738 }
1739
1740 /* Skip target unit designators. We already acted upon this information and have queued
1741 * appropriate jobs. We don't want to redo all this after reexecution. */
1742 if (startswith(src[i], "--unit="))
1743 continue;
1744 if (streq(src[i], "--unit")) {
1745 i++; /* Skip the argument too */
1746 continue;
1747 }
1748
1749 /* Seems we have a good old option. Let's pass it over to the new instance. */
1750 dst[(*dst_index)++] = src[i];
1751 }
1752 }
1753
1754 static int do_reexecute(
1755 ManagerObjective objective,
1756 int argc,
1757 char* argv[],
1758 const struct rlimit *saved_rlimit_nofile,
1759 const struct rlimit *saved_rlimit_memlock,
1760 FDSet *fds,
1761 const char *switch_root_dir,
1762 const char *switch_root_init,
1763 const char **ret_error_message) {
1764
1765 size_t i, args_size;
1766 const char **args;
1767 int r;
1768
1769 assert(IN_SET(objective, MANAGER_REEXECUTE, MANAGER_SWITCH_ROOT));
1770 assert(argc >= 0);
1771 assert(saved_rlimit_nofile);
1772 assert(saved_rlimit_memlock);
1773 assert(ret_error_message);
1774
1775 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get
1776 * rebooted while we do that */
1777 watchdog_close(true);
1778
1779 /* Reset RLIMIT_NOFILE + RLIMIT_MEMLOCK back to the kernel defaults, so that the new systemd can pass
1780 * the kernel default to its child processes */
1781 if (saved_rlimit_nofile->rlim_cur != 0)
1782 (void) setrlimit(RLIMIT_NOFILE, saved_rlimit_nofile);
1783 if (saved_rlimit_memlock->rlim_cur != RLIM_INFINITY)
1784 (void) setrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock);
1785
1786 if (switch_root_dir) {
1787 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can
1788 * handle the SIGCHLD for them after deserializing. */
1789 broadcast_signal(SIGTERM, false, true, arg_default_timeout_stop_usec);
1790
1791 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1792 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1793 if (r < 0)
1794 log_error_errno(r, "Failed to switch root, trying to continue: %m");
1795 }
1796
1797 args_size = argc + 6;
1798 args = newa(const char*, args_size);
1799
1800 if (!switch_root_init) {
1801 char sfd[DECIMAL_STR_MAX(int)];
1802
1803 /* First try to spawn ourselves with the right path, and with full serialization. We do this
1804 * only if the user didn't specify an explicit init to spawn. */
1805
1806 assert(arg_serialization);
1807 assert(fds);
1808
1809 xsprintf(sfd, "%i", fileno(arg_serialization));
1810
1811 i = 1; /* Leave args[0] empty for now. */
1812 filter_args(args, &i, argv, argc);
1813
1814 if (switch_root_dir)
1815 args[i++] = "--switched-root";
1816 args[i++] = arg_system ? "--system" : "--user";
1817 args[i++] = "--deserialize";
1818 args[i++] = sfd;
1819 args[i++] = NULL;
1820
1821 assert(i <= args_size);
1822
1823 /*
1824 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do
1825 * this is on its own on exec(), but it will do it on exit(). Hence, to ensure we get a
1826 * summary here, fork() off a child, let it exit() cleanly, so that it prints the summary,
1827 * and wait() for it in the parent, before proceeding into the exec().
1828 */
1829 valgrind_summary_hack();
1830
1831 args[0] = SYSTEMD_BINARY_PATH;
1832 (void) execv(args[0], (char* const*) args);
1833
1834 if (objective == MANAGER_REEXECUTE) {
1835 *ret_error_message = "Failed to execute our own binary";
1836 return log_error_errno(errno, "Failed to execute our own binary %s: %m", args[0]);
1837 }
1838
1839 log_debug_errno(errno, "Failed to execute our own binary %s, trying fallback: %m", args[0]);
1840 }
1841
1842 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and
1843 * envp[]. (Well, modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[],
1844 * but let's hope that doesn't matter.) */
1845
1846 arg_serialization = safe_fclose(arg_serialization);
1847 fds = fdset_free(fds);
1848
1849 /* Reopen the console */
1850 (void) make_console_stdio();
1851
1852 i = 1; /* Leave args[0] empty for now. */
1853 for (int j = 1; j <= argc; j++)
1854 args[i++] = argv[j];
1855 assert(i <= args_size);
1856
1857 /* Re-enable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1858 (void) reset_all_signal_handlers();
1859 (void) reset_signal_mask();
1860 (void) rlimit_nofile_safe();
1861
1862 if (switch_root_init) {
1863 args[0] = switch_root_init;
1864 (void) execve(args[0], (char* const*) args, saved_env);
1865 log_warning_errno(errno, "Failed to execute configured init %s, trying fallback: %m", args[0]);
1866 }
1867
1868 args[0] = "/sbin/init";
1869 (void) execv(args[0], (char* const*) args);
1870 r = -errno;
1871
1872 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
1873 ANSI_HIGHLIGHT_RED " !! " ANSI_NORMAL,
1874 "Failed to execute /sbin/init");
1875
1876 *ret_error_message = "Failed to execute fallback shell";
1877 if (r == -ENOENT) {
1878 log_warning("No /sbin/init, trying fallback");
1879
1880 args[0] = "/bin/sh";
1881 args[1] = NULL;
1882 (void) execve(args[0], (char* const*) args, saved_env);
1883 return log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
1884 } else
1885 return log_error_errno(r, "Failed to execute /sbin/init, giving up: %m");
1886 }
1887
1888 static int invoke_main_loop(
1889 Manager *m,
1890 const struct rlimit *saved_rlimit_nofile,
1891 const struct rlimit *saved_rlimit_memlock,
1892 int *ret_retval, /* Return parameters relevant for shutting down */
1893 const char **ret_shutdown_verb, /* … */
1894 FDSet **ret_fds, /* Return parameters for reexecuting */
1895 char **ret_switch_root_dir, /* … */
1896 char **ret_switch_root_init, /* … */
1897 const char **ret_error_message) {
1898
1899 int r;
1900
1901 assert(m);
1902 assert(saved_rlimit_nofile);
1903 assert(saved_rlimit_memlock);
1904 assert(ret_retval);
1905 assert(ret_shutdown_verb);
1906 assert(ret_fds);
1907 assert(ret_switch_root_dir);
1908 assert(ret_switch_root_init);
1909 assert(ret_error_message);
1910
1911 for (;;) {
1912 int objective = manager_loop(m);
1913 if (objective < 0) {
1914 *ret_error_message = "Failed to run main loop";
1915 return log_emergency_errno(objective, "Failed to run main loop: %m");
1916 }
1917
1918 switch (objective) {
1919
1920 case MANAGER_RELOAD: {
1921 LogTarget saved_log_target;
1922 int saved_log_level;
1923
1924 log_info("Reloading.");
1925
1926 /* First, save any overridden log level/target, then parse the configuration file,
1927 * which might change the log level to new settings. */
1928
1929 saved_log_level = m->log_level_overridden ? log_get_max_level() : -1;
1930 saved_log_target = m->log_target_overridden ? log_get_target() : _LOG_TARGET_INVALID;
1931
1932 (void) parse_configuration(saved_rlimit_nofile, saved_rlimit_memlock);
1933
1934 set_manager_defaults(m);
1935 set_manager_settings(m);
1936
1937 update_cpu_affinity(false);
1938 update_numa_policy(false);
1939
1940 if (saved_log_level >= 0)
1941 manager_override_log_level(m, saved_log_level);
1942 if (saved_log_target >= 0)
1943 manager_override_log_target(m, saved_log_target);
1944
1945 if (manager_reload(m) < 0)
1946 /* Reloading failed before the point of no return.
1947 * Let's continue running as if nothing happened. */
1948 m->objective = MANAGER_OK;
1949
1950 continue;
1951 }
1952
1953 case MANAGER_REEXECUTE:
1954 r = prepare_reexecute(m, &arg_serialization, ret_fds, false);
1955 if (r < 0) {
1956 *ret_error_message = "Failed to prepare for reexecution";
1957 return r;
1958 }
1959
1960 log_notice("Reexecuting.");
1961
1962 *ret_retval = EXIT_SUCCESS;
1963 *ret_shutdown_verb = NULL;
1964 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1965
1966 return objective;
1967
1968 case MANAGER_SWITCH_ROOT:
1969 if (!m->switch_root_init) {
1970 r = prepare_reexecute(m, &arg_serialization, ret_fds, true);
1971 if (r < 0) {
1972 *ret_error_message = "Failed to prepare for reexecution";
1973 return r;
1974 }
1975 } else
1976 *ret_fds = NULL;
1977
1978 log_notice("Switching root.");
1979
1980 *ret_retval = EXIT_SUCCESS;
1981 *ret_shutdown_verb = NULL;
1982
1983 /* Steal the switch root parameters */
1984 *ret_switch_root_dir = TAKE_PTR(m->switch_root);
1985 *ret_switch_root_init = TAKE_PTR(m->switch_root_init);
1986
1987 return objective;
1988
1989 case MANAGER_EXIT:
1990 if (MANAGER_IS_USER(m)) {
1991 log_debug("Exit.");
1992
1993 *ret_retval = m->return_value;
1994 *ret_shutdown_verb = NULL;
1995 *ret_fds = NULL;
1996 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1997
1998 return objective;
1999 }
2000
2001 _fallthrough_;
2002 case MANAGER_REBOOT:
2003 case MANAGER_POWEROFF:
2004 case MANAGER_HALT:
2005 case MANAGER_KEXEC: {
2006 static const char* const table[_MANAGER_OBJECTIVE_MAX] = {
2007 [MANAGER_EXIT] = "exit",
2008 [MANAGER_REBOOT] = "reboot",
2009 [MANAGER_POWEROFF] = "poweroff",
2010 [MANAGER_HALT] = "halt",
2011 [MANAGER_KEXEC] = "kexec",
2012 };
2013
2014 log_notice("Shutting down.");
2015
2016 *ret_retval = m->return_value;
2017 assert_se(*ret_shutdown_verb = table[m->objective]);
2018 *ret_fds = NULL;
2019 *ret_switch_root_dir = *ret_switch_root_init = NULL;
2020
2021 return objective;
2022 }
2023
2024 default:
2025 assert_not_reached();
2026 }
2027 }
2028 }
2029
2030 static void log_execution_mode(bool *ret_first_boot) {
2031 assert(ret_first_boot);
2032
2033 if (arg_system) {
2034 struct utsname uts;
2035 int v;
2036
2037 log_info("systemd " GIT_VERSION " running in %ssystem mode (%s)",
2038 arg_action == ACTION_TEST ? "test " : "",
2039 systemd_features);
2040
2041 v = detect_virtualization();
2042 if (v > 0)
2043 log_info("Detected virtualization %s.", virtualization_to_string(v));
2044
2045 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
2046
2047 if (in_initrd()) {
2048 *ret_first_boot = false;
2049 log_info("Running in initial RAM disk.");
2050 } else {
2051 int r;
2052 _cleanup_free_ char *id_text = NULL;
2053
2054 /* Let's check whether we are in first boot. We use /etc/machine-id as flag file
2055 * for this: If it is missing or contains the value "uninitialized", this is the
2056 * first boot. In any other case, it is not. This allows container managers and
2057 * installers to provision a couple of files already. If the container manager
2058 * wants to provision the machine ID itself it should pass $container_uuid to PID 1. */
2059
2060 r = read_one_line_file("/etc/machine-id", &id_text);
2061 if (r < 0 || streq(id_text, "uninitialized")) {
2062 if (r < 0 && r != -ENOENT)
2063 log_warning_errno(r, "Unexpected error while reading /etc/machine-id, ignoring: %m");
2064
2065 *ret_first_boot = true;
2066 log_info("Detected first boot.");
2067 } else {
2068 *ret_first_boot = false;
2069 log_debug("Detected initialized system, this is not the first boot.");
2070 }
2071 }
2072
2073 assert_se(uname(&uts) >= 0);
2074
2075 if (strverscmp_improved(uts.release, KERNEL_BASELINE_VERSION) < 0)
2076 log_warning("Warning! Reported kernel version %s is older than systemd's required baseline kernel version %s. "
2077 "Your mileage may vary.", uts.release, KERNEL_BASELINE_VERSION);
2078 else
2079 log_debug("Kernel version %s, our baseline is %s", uts.release, KERNEL_BASELINE_VERSION);
2080 } else {
2081 if (DEBUG_LOGGING) {
2082 _cleanup_free_ char *t = NULL;
2083
2084 t = uid_to_name(getuid());
2085 log_debug("systemd " GIT_VERSION " running in %suser mode for user " UID_FMT "/%s. (%s)",
2086 arg_action == ACTION_TEST ? " test" : "",
2087 getuid(), strna(t), systemd_features);
2088 }
2089
2090 *ret_first_boot = false;
2091 }
2092 }
2093
2094 static int initialize_runtime(
2095 bool skip_setup,
2096 bool first_boot,
2097 struct rlimit *saved_rlimit_nofile,
2098 struct rlimit *saved_rlimit_memlock,
2099 const char **ret_error_message) {
2100 int r;
2101
2102 assert(ret_error_message);
2103
2104 /* Sets up various runtime parameters. Many of these initializations are conditionalized:
2105 *
2106 * - Some only apply to --system instances
2107 * - Some only apply to --user instances
2108 * - Some only apply when we first start up, but not when we reexecute
2109 */
2110
2111 if (arg_action != ACTION_RUN)
2112 return 0;
2113
2114 update_cpu_affinity(skip_setup);
2115 update_numa_policy(skip_setup);
2116
2117 if (arg_system) {
2118 /* Make sure we leave a core dump without panicking the kernel. */
2119 install_crash_handler();
2120
2121 if (!skip_setup) {
2122 r = mount_cgroup_controllers();
2123 if (r < 0) {
2124 *ret_error_message = "Failed to mount cgroup hierarchies";
2125 return r;
2126 }
2127
2128 (void) os_release_status();
2129 (void) hostname_setup(true);
2130 /* Force transient machine-id on first boot. */
2131 machine_id_setup(NULL, first_boot, arg_machine_id, NULL);
2132 (void) loopback_setup();
2133 bump_unix_max_dgram_qlen();
2134 bump_file_max_and_nr_open();
2135 test_usr();
2136 write_container_id();
2137 }
2138
2139 r = watchdog_set_device(arg_watchdog_device);
2140 if (r < 0)
2141 log_warning_errno(r, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device);
2142 } else {
2143 _cleanup_free_ char *p = NULL;
2144
2145 /* Create the runtime directory and place the inaccessible device nodes there, if we run in
2146 * user mode. In system mode mount_setup() already did that. */
2147
2148 r = xdg_user_runtime_dir(&p, "/systemd");
2149 if (r < 0) {
2150 *ret_error_message = "$XDG_RUNTIME_DIR is not set";
2151 return log_emergency_errno(r, "Failed to determine $XDG_RUNTIME_DIR path: %m");
2152 }
2153
2154 (void) mkdir_p_label(p, 0755);
2155 (void) make_inaccessible_nodes(p, UID_INVALID, GID_INVALID);
2156 }
2157
2158 if (arg_timer_slack_nsec != NSEC_INFINITY)
2159 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
2160 log_warning_errno(errno, "Failed to adjust timer slack, ignoring: %m");
2161
2162 if (arg_system && !cap_test_all(arg_capability_bounding_set)) {
2163 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
2164 if (r < 0) {
2165 *ret_error_message = "Failed to drop capability bounding set of usermode helpers";
2166 return log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
2167 }
2168
2169 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
2170 if (r < 0) {
2171 *ret_error_message = "Failed to drop capability bounding set";
2172 return log_emergency_errno(r, "Failed to drop capability bounding set: %m");
2173 }
2174 }
2175
2176 if (arg_system && arg_no_new_privs) {
2177 if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) {
2178 *ret_error_message = "Failed to disable new privileges";
2179 return log_emergency_errno(errno, "Failed to disable new privileges: %m");
2180 }
2181 }
2182
2183 if (arg_syscall_archs) {
2184 r = enforce_syscall_archs(arg_syscall_archs);
2185 if (r < 0) {
2186 *ret_error_message = "Failed to set syscall architectures";
2187 return r;
2188 }
2189 }
2190
2191 if (!arg_system)
2192 /* Become reaper of our children */
2193 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
2194 log_warning_errno(errno, "Failed to make us a subreaper, ignoring: %m");
2195
2196 /* Bump up RLIMIT_NOFILE for systemd itself */
2197 (void) bump_rlimit_nofile(saved_rlimit_nofile);
2198 (void) bump_rlimit_memlock(saved_rlimit_memlock);
2199
2200 /* Pull credentials from various sources into a common credential directory */
2201 if (arg_system && !skip_setup)
2202 (void) import_credentials();
2203
2204 return 0;
2205 }
2206
2207 static int do_queue_default_job(
2208 Manager *m,
2209 const char **ret_error_message) {
2210
2211 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
2212 const char *unit;
2213 Job *job;
2214 Unit *target;
2215 int r;
2216
2217 if (arg_default_unit)
2218 unit = arg_default_unit;
2219 else if (in_initrd())
2220 unit = SPECIAL_INITRD_TARGET;
2221 else
2222 unit = SPECIAL_DEFAULT_TARGET;
2223
2224 log_debug("Activating default unit: %s", unit);
2225
2226 r = manager_load_startable_unit_or_warn(m, unit, NULL, &target);
2227 if (r < 0 && in_initrd() && !arg_default_unit) {
2228 /* Fall back to default.target, which we used to always use by default. Only do this if no
2229 * explicit configuration was given. */
2230
2231 log_info("Falling back to " SPECIAL_DEFAULT_TARGET ".");
2232
2233 r = manager_load_startable_unit_or_warn(m, SPECIAL_DEFAULT_TARGET, NULL, &target);
2234 }
2235 if (r < 0) {
2236 log_info("Falling back to " SPECIAL_RESCUE_TARGET ".");
2237
2238 r = manager_load_startable_unit_or_warn(m, SPECIAL_RESCUE_TARGET, NULL, &target);
2239 if (r < 0) {
2240 *ret_error_message = r == -ERFKILL ? SPECIAL_RESCUE_TARGET " masked"
2241 : "Failed to load " SPECIAL_RESCUE_TARGET;
2242 return r;
2243 }
2244 }
2245
2246 assert(target->load_state == UNIT_LOADED);
2247
2248 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, NULL, &error, &job);
2249 if (r == -EPERM) {
2250 log_debug_errno(r, "Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
2251
2252 sd_bus_error_free(&error);
2253
2254 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, NULL, &error, &job);
2255 if (r < 0) {
2256 *ret_error_message = "Failed to start default target";
2257 return log_emergency_errno(r, "Failed to start default target: %s", bus_error_message(&error, r));
2258 }
2259
2260 } else if (r < 0) {
2261 *ret_error_message = "Failed to isolate default target";
2262 return log_emergency_errno(r, "Failed to isolate default target: %s", bus_error_message(&error, r));
2263 } else
2264 log_info("Queued %s job for default target %s.",
2265 job_type_to_string(job->type),
2266 unit_status_string(job->unit, NULL));
2267
2268 m->default_unit_job_id = job->id;
2269
2270 return 0;
2271 }
2272
2273 static void save_rlimits(struct rlimit *saved_rlimit_nofile,
2274 struct rlimit *saved_rlimit_memlock) {
2275
2276 assert(saved_rlimit_nofile);
2277 assert(saved_rlimit_memlock);
2278
2279 if (getrlimit(RLIMIT_NOFILE, saved_rlimit_nofile) < 0)
2280 log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m");
2281
2282 if (getrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock) < 0)
2283 log_warning_errno(errno, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
2284 }
2285
2286 static void fallback_rlimit_nofile(const struct rlimit *saved_rlimit_nofile) {
2287 struct rlimit *rl;
2288
2289 if (arg_default_rlimit[RLIMIT_NOFILE])
2290 return;
2291
2292 /* Make sure forked processes get limits based on the original kernel setting */
2293
2294 rl = newdup(struct rlimit, saved_rlimit_nofile, 1);
2295 if (!rl) {
2296 log_oom();
2297 return;
2298 }
2299
2300 /* Bump the hard limit for system services to a substantially higher value. The default
2301 * hard limit current kernels set is pretty low (4K), mostly for historical
2302 * reasons. According to kernel developers, the fd handling in recent kernels has been
2303 * optimized substantially enough, so that we can bump the limit now, without paying too
2304 * high a price in memory or performance. Note however that we only bump the hard limit,
2305 * not the soft limit. That's because select() works the way it works, and chokes on fds
2306 * >= 1024. If we'd bump the soft limit globally, it might accidentally happen to
2307 * unexpecting programs that they get fds higher than what they can process using
2308 * select(). By only bumping the hard limit but leaving the low limit as it is we avoid
2309 * this pitfall: programs that are written by folks aware of the select() problem in mind
2310 * (and thus use poll()/epoll instead of select(), the way everybody should) can
2311 * explicitly opt into high fds by bumping their soft limit beyond 1024, to the hard limit
2312 * we pass. */
2313 if (arg_system) {
2314 int nr;
2315
2316 /* Get the underlying absolute limit the kernel enforces */
2317 nr = read_nr_open();
2318
2319 rl->rlim_max = MIN((rlim_t) nr, MAX(rl->rlim_max, (rlim_t) HIGH_RLIMIT_NOFILE));
2320 }
2321
2322 /* If for some reason we were invoked with a soft limit above 1024 (which should never
2323 * happen!, but who knows what we get passed in from pam_limit when invoked as --user
2324 * instance), then lower what we pass on to not confuse our children */
2325 rl->rlim_cur = MIN(rl->rlim_cur, (rlim_t) FD_SETSIZE);
2326
2327 arg_default_rlimit[RLIMIT_NOFILE] = rl;
2328 }
2329
2330 static void fallback_rlimit_memlock(const struct rlimit *saved_rlimit_memlock) {
2331 struct rlimit *rl;
2332
2333 /* Pass the original value down to invoked processes */
2334
2335 if (arg_default_rlimit[RLIMIT_MEMLOCK])
2336 return;
2337
2338 rl = newdup(struct rlimit, saved_rlimit_memlock, 1);
2339 if (!rl) {
2340 log_oom();
2341 return;
2342 }
2343
2344 if (arg_system) {
2345 /* Raise the default limit to 8M also on old kernels and in containers (8M is the kernel
2346 * default for this since kernel 5.16) */
2347 rl->rlim_max = MAX(rl->rlim_max, (rlim_t) DEFAULT_RLIMIT_MEMLOCK);
2348 rl->rlim_cur = MAX(rl->rlim_cur, (rlim_t) DEFAULT_RLIMIT_MEMLOCK);
2349 }
2350
2351 arg_default_rlimit[RLIMIT_MEMLOCK] = rl;
2352 }
2353
2354 static void setenv_manager_environment(void) {
2355 int r;
2356
2357 STRV_FOREACH(p, arg_manager_environment) {
2358 log_debug("Setting '%s' in our own environment.", *p);
2359
2360 r = putenv_dup(*p, true);
2361 if (r < 0)
2362 log_warning_errno(errno, "Failed to setenv \"%s\", ignoring: %m", *p);
2363 }
2364 }
2365
2366 static void reset_arguments(void) {
2367 /* Frees/resets arg_* variables, with a few exceptions commented below. */
2368
2369 arg_default_unit = mfree(arg_default_unit);
2370
2371 /* arg_system — ignore */
2372
2373 arg_dump_core = true;
2374 arg_crash_chvt = -1;
2375 arg_crash_shell = false;
2376 arg_crash_reboot = false;
2377 arg_confirm_spawn = mfree(arg_confirm_spawn);
2378 arg_show_status = _SHOW_STATUS_INVALID;
2379 arg_status_unit_format = STATUS_UNIT_FORMAT_DEFAULT;
2380 arg_switched_root = false;
2381 arg_pager_flags = 0;
2382 arg_service_watchdogs = true;
2383 arg_default_std_output = EXEC_OUTPUT_JOURNAL;
2384 arg_default_std_error = EXEC_OUTPUT_INHERIT;
2385 arg_default_restart_usec = DEFAULT_RESTART_USEC;
2386 arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
2387 arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
2388 arg_default_timeout_abort_usec = DEFAULT_TIMEOUT_USEC;
2389 arg_default_timeout_abort_set = false;
2390 arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
2391 arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
2392 arg_runtime_watchdog = 0;
2393 arg_reboot_watchdog = 10 * USEC_PER_MINUTE;
2394 arg_kexec_watchdog = 0;
2395 arg_pretimeout_watchdog = 0;
2396 arg_early_core_pattern = mfree(arg_early_core_pattern);
2397 arg_watchdog_device = mfree(arg_watchdog_device);
2398 arg_watchdog_pretimeout_governor = mfree(arg_watchdog_pretimeout_governor);
2399
2400 arg_default_environment = strv_free(arg_default_environment);
2401 arg_manager_environment = strv_free(arg_manager_environment);
2402 rlimit_free_all(arg_default_rlimit);
2403
2404 arg_capability_bounding_set = CAP_ALL;
2405 arg_no_new_privs = false;
2406 arg_timer_slack_nsec = NSEC_INFINITY;
2407 arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
2408
2409 arg_syscall_archs = set_free(arg_syscall_archs);
2410
2411 /* arg_serialization — ignore */
2412
2413 arg_default_cpu_accounting = -1;
2414 arg_default_io_accounting = false;
2415 arg_default_ip_accounting = false;
2416 arg_default_blockio_accounting = false;
2417 arg_default_memory_accounting = MEMORY_ACCOUNTING_DEFAULT;
2418 arg_default_tasks_accounting = true;
2419 arg_default_tasks_max = DEFAULT_TASKS_MAX;
2420 arg_machine_id = (sd_id128_t) {};
2421 arg_cad_burst_action = EMERGENCY_ACTION_REBOOT_FORCE;
2422 arg_default_oom_policy = OOM_STOP;
2423
2424 cpu_set_reset(&arg_cpu_affinity);
2425 numa_policy_reset(&arg_numa_policy);
2426
2427 arg_random_seed = mfree(arg_random_seed);
2428 arg_random_seed_size = 0;
2429 arg_clock_usec = 0;
2430
2431 arg_default_oom_score_adjust_set = false;
2432 arg_default_smack_process_label = mfree(arg_default_smack_process_label);
2433 }
2434
2435 static void determine_default_oom_score_adjust(void) {
2436 int r, a, b;
2437
2438 /* Run our services at slightly higher OOM score than ourselves. But let's be conservative here, and
2439 * do this only if we don't run as root (i.e. only if we are run in user mode, for an unprivileged
2440 * user). */
2441
2442 if (arg_default_oom_score_adjust_set)
2443 return;
2444
2445 if (getuid() == 0)
2446 return;
2447
2448 r = get_oom_score_adjust(&a);
2449 if (r < 0)
2450 return (void) log_warning_errno(r, "Failed to determine current OOM score adjustment value, ignoring: %m");
2451
2452 assert_cc(100 <= OOM_SCORE_ADJ_MAX);
2453 b = a >= OOM_SCORE_ADJ_MAX - 100 ? OOM_SCORE_ADJ_MAX : a + 100;
2454
2455 if (a == b)
2456 return;
2457
2458 arg_default_oom_score_adjust = b;
2459 arg_default_oom_score_adjust_set = true;
2460 }
2461
2462 static int parse_configuration(const struct rlimit *saved_rlimit_nofile,
2463 const struct rlimit *saved_rlimit_memlock) {
2464 int r;
2465
2466 assert(saved_rlimit_nofile);
2467 assert(saved_rlimit_memlock);
2468
2469 /* Assign configuration defaults */
2470 reset_arguments();
2471
2472 r = parse_config_file();
2473 if (r < 0)
2474 log_warning_errno(r, "Failed to parse config file, ignoring: %m");
2475
2476 if (arg_system) {
2477 r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
2478 if (r < 0)
2479 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
2480 }
2481
2482 /* Initialize some default rlimits for services if they haven't been configured */
2483 fallback_rlimit_nofile(saved_rlimit_nofile);
2484 fallback_rlimit_memlock(saved_rlimit_memlock);
2485
2486 /* Note that this also parses bits from the kernel command line, including "debug". */
2487 log_parse_environment();
2488
2489 /* Initialize the show status setting if it hasn't been set explicitly yet */
2490 if (arg_show_status == _SHOW_STATUS_INVALID)
2491 arg_show_status = SHOW_STATUS_YES;
2492
2493 /* Slightly raise the OOM score for our services if we are running for unprivileged users. */
2494 determine_default_oom_score_adjust();
2495
2496 /* Push variables into the manager environment block */
2497 setenv_manager_environment();
2498
2499 /* Parse log environment variables again to take into account any new environment variables. */
2500 log_parse_environment();
2501
2502 return 0;
2503 }
2504
2505 static int safety_checks(void) {
2506
2507 if (getpid_cached() == 1 &&
2508 arg_action != ACTION_RUN)
2509 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2510 "Unsupported execution mode while PID 1.");
2511
2512 if (getpid_cached() == 1 &&
2513 !arg_system)
2514 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2515 "Can't run --user mode as PID 1.");
2516
2517 if (arg_action == ACTION_RUN &&
2518 arg_system &&
2519 getpid_cached() != 1)
2520 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2521 "Can't run system mode unless PID 1.");
2522
2523 if (arg_action == ACTION_TEST &&
2524 geteuid() == 0)
2525 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2526 "Don't run test mode as root.");
2527
2528 if (!arg_system &&
2529 arg_action == ACTION_RUN &&
2530 sd_booted() <= 0)
2531 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
2532 "Trying to run as user instance, but the system has not been booted with systemd.");
2533
2534 if (!arg_system &&
2535 arg_action == ACTION_RUN &&
2536 !getenv("XDG_RUNTIME_DIR"))
2537 return log_error_errno(SYNTHETIC_ERRNO(EUNATCH),
2538 "Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
2539
2540 if (arg_system &&
2541 arg_action == ACTION_RUN &&
2542 running_in_chroot() > 0)
2543 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
2544 "Cannot be run in a chroot() environment.");
2545
2546 return 0;
2547 }
2548
2549 static int initialize_security(
2550 bool *loaded_policy,
2551 dual_timestamp *security_start_timestamp,
2552 dual_timestamp *security_finish_timestamp,
2553 const char **ret_error_message) {
2554
2555 int r;
2556
2557 assert(loaded_policy);
2558 assert(security_start_timestamp);
2559 assert(security_finish_timestamp);
2560 assert(ret_error_message);
2561
2562 dual_timestamp_get(security_start_timestamp);
2563
2564 r = mac_selinux_setup(loaded_policy);
2565 if (r < 0) {
2566 *ret_error_message = "Failed to load SELinux policy";
2567 return r;
2568 }
2569
2570 r = mac_smack_setup(loaded_policy);
2571 if (r < 0) {
2572 *ret_error_message = "Failed to load SMACK policy";
2573 return r;
2574 }
2575
2576 r = mac_apparmor_setup();
2577 if (r < 0) {
2578 *ret_error_message = "Failed to load AppArmor policy";
2579 return r;
2580 }
2581
2582 r = ima_setup();
2583 if (r < 0) {
2584 *ret_error_message = "Failed to load IMA policy";
2585 return r;
2586 }
2587
2588 dual_timestamp_get(security_finish_timestamp);
2589 return 0;
2590 }
2591
2592 static int collect_fds(FDSet **ret_fds, const char **ret_error_message) {
2593 int r;
2594
2595 assert(ret_fds);
2596 assert(ret_error_message);
2597
2598 r = fdset_new_fill(ret_fds);
2599 if (r < 0) {
2600 *ret_error_message = "Failed to allocate fd set";
2601 return log_emergency_errno(r, "Failed to allocate fd set: %m");
2602 }
2603
2604 fdset_cloexec(*ret_fds, true);
2605
2606 if (arg_serialization)
2607 assert_se(fdset_remove(*ret_fds, fileno(arg_serialization)) >= 0);
2608
2609 return 0;
2610 }
2611
2612 static void setup_console_terminal(bool skip_setup) {
2613
2614 if (!arg_system)
2615 return;
2616
2617 /* Become a session leader if we aren't one yet. */
2618 (void) setsid();
2619
2620 /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a
2621 * controlling tty. */
2622 (void) release_terminal();
2623
2624 /* Reset the console, but only if this is really init and we are freshly booted */
2625 if (getpid_cached() == 1 && !skip_setup)
2626 (void) console_setup();
2627 }
2628
2629 static bool early_skip_setup_check(int argc, char *argv[]) {
2630 bool found_deserialize = false;
2631
2632 /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much
2633 * later, so let's just have a quick peek here. Note that if we have switched root, do all the
2634 * special setup things anyway, even if in that case we also do deserialization. */
2635
2636 for (int i = 1; i < argc; i++)
2637 if (streq(argv[i], "--switched-root"))
2638 return false; /* If we switched root, don't skip the setup. */
2639 else if (streq(argv[i], "--deserialize"))
2640 found_deserialize = true;
2641
2642 return found_deserialize; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2643 }
2644
2645 static int save_env(void) {
2646 char **l;
2647
2648 l = strv_copy(environ);
2649 if (!l)
2650 return -ENOMEM;
2651
2652 strv_free_and_replace(saved_env, l);
2653 return 0;
2654 }
2655
2656 int main(int argc, char *argv[]) {
2657 dual_timestamp
2658 initrd_timestamp = DUAL_TIMESTAMP_NULL,
2659 userspace_timestamp = DUAL_TIMESTAMP_NULL,
2660 kernel_timestamp = DUAL_TIMESTAMP_NULL,
2661 security_start_timestamp = DUAL_TIMESTAMP_NULL,
2662 security_finish_timestamp = DUAL_TIMESTAMP_NULL;
2663 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0),
2664 saved_rlimit_memlock = RLIMIT_MAKE_CONST(RLIM_INFINITY); /* The original rlimits we passed
2665 * in. Note we use different values
2666 * for the two that indicate whether
2667 * these fields are initialized! */
2668 bool skip_setup, loaded_policy = false, queue_default_job = false, first_boot = false;
2669 char *switch_root_dir = NULL, *switch_root_init = NULL;
2670 usec_t before_startup, after_startup;
2671 static char systemd[] = "systemd";
2672 const char *shutdown_verb = NULL, *error_message = NULL;
2673 int r, retval = EXIT_FAILURE;
2674 Manager *m = NULL;
2675 FDSet *fds = NULL;
2676
2677 assert_se(argc > 0 && !isempty(argv[0]));
2678
2679 /* SysV compatibility: redirect init → telinit */
2680 redirect_telinit(argc, argv);
2681
2682 /* Take timestamps early on */
2683 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
2684 dual_timestamp_get(&userspace_timestamp);
2685
2686 /* Figure out whether we need to do initialize the system, or if we already did that because we are
2687 * reexecuting. */
2688 skip_setup = early_skip_setup_check(argc, argv);
2689
2690 /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent
2691 * reexecution we are then called 'systemd'. That is confusing, hence let's call us systemd
2692 * right-away. */
2693 program_invocation_short_name = systemd;
2694 (void) prctl(PR_SET_NAME, systemd);
2695
2696 /* Save the original command line */
2697 save_argc_argv(argc, argv);
2698
2699 /* Save the original environment as we might need to restore it if we're requested to execute another
2700 * system manager later. */
2701 r = save_env();
2702 if (r < 0) {
2703 error_message = "Failed to copy environment block";
2704 goto finish;
2705 }
2706
2707 /* Make sure that if the user says "syslog" we actually log to the journal. */
2708 log_set_upgrade_syslog_to_journal(true);
2709
2710 if (getpid_cached() == 1) {
2711 /* When we run as PID 1 force system mode */
2712 arg_system = true;
2713
2714 /* Disable the umask logic */
2715 umask(0);
2716
2717 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might
2718 * not be activated yet (even though the log socket for it exists). */
2719 log_set_prohibit_ipc(true);
2720
2721 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This
2722 * is important so that we never end up logging to any foreign stderr, for example if we have
2723 * to log in a child process right before execve()'ing the actual binary, at a point in time
2724 * where socket activation stderr/stdout area already set up. */
2725 log_set_always_reopen_console(true);
2726
2727 if (detect_container() <= 0) {
2728
2729 /* Running outside of a container as PID 1 */
2730 log_set_target(LOG_TARGET_KMSG);
2731 log_open();
2732
2733 if (in_initrd())
2734 initrd_timestamp = userspace_timestamp;
2735
2736 if (!skip_setup) {
2737 r = mount_setup_early();
2738 if (r < 0) {
2739 error_message = "Failed to mount early API filesystems";
2740 goto finish;
2741 }
2742
2743 /* Let's open the log backend a second time, in case the first time didn't
2744 * work. Quite possibly we have mounted /dev just now, so /dev/kmsg became
2745 * available, and it previously wasn't. */
2746 log_open();
2747
2748 disable_printk_ratelimit();
2749
2750 r = initialize_security(
2751 &loaded_policy,
2752 &security_start_timestamp,
2753 &security_finish_timestamp,
2754 &error_message);
2755 if (r < 0)
2756 goto finish;
2757 }
2758
2759 if (mac_selinux_init() < 0) {
2760 error_message = "Failed to initialize SELinux support";
2761 goto finish;
2762 }
2763
2764 if (!skip_setup)
2765 initialize_clock();
2766
2767 /* Set the default for later on, but don't actually open the logs like this for
2768 * now. Note that if we are transitioning from the initrd there might still be
2769 * journal fd open, and we shouldn't attempt opening that before we parsed
2770 * /proc/cmdline which might redirect output elsewhere. */
2771 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
2772
2773 } else {
2774 /* Running inside a container, as PID 1 */
2775 log_set_target(LOG_TARGET_CONSOLE);
2776 log_open();
2777
2778 /* For later on, see above... */
2779 log_set_target(LOG_TARGET_JOURNAL);
2780
2781 /* clear the kernel timestamp, because we are in a container */
2782 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2783 }
2784
2785 initialize_coredump(skip_setup);
2786
2787 r = fixup_environment();
2788 if (r < 0) {
2789 log_emergency_errno(r, "Failed to fix up PID 1 environment: %m");
2790 error_message = "Failed to fix up PID1 environment";
2791 goto finish;
2792 }
2793
2794 /* Try to figure out if we can use colors with the console. No need to do that for user
2795 * instances since they never log into the console. */
2796 log_show_color(colors_enabled());
2797
2798 r = make_null_stdio();
2799 if (r < 0)
2800 log_warning_errno(r, "Failed to redirect standard streams to /dev/null, ignoring: %m");
2801
2802 /* Load the kernel modules early. */
2803 if (!skip_setup)
2804 (void) kmod_setup();
2805
2806 /* Mount /proc, /sys and friends, so that /proc/cmdline and /proc/$PID/fd is available. */
2807 r = mount_setup(loaded_policy, skip_setup);
2808 if (r < 0) {
2809 error_message = "Failed to mount API filesystems";
2810 goto finish;
2811 }
2812
2813 /* The efivarfs is now mounted, let's read the random seed off it */
2814 (void) efi_take_random_seed();
2815
2816 /* Cache command-line options passed from EFI variables */
2817 if (!skip_setup)
2818 (void) cache_efi_options_variable();
2819 } else {
2820 /* Running as user instance */
2821 arg_system = false;
2822 log_set_target(LOG_TARGET_AUTO);
2823 log_open();
2824
2825 /* clear the kernel timestamp, because we are not PID 1 */
2826 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2827
2828 if (mac_selinux_init() < 0) {
2829 error_message = "Failed to initialize SELinux support";
2830 goto finish;
2831 }
2832 }
2833
2834 /* Save the original RLIMIT_NOFILE/RLIMIT_MEMLOCK so that we can reset it later when
2835 * transitioning from the initrd to the main systemd or suchlike. */
2836 save_rlimits(&saved_rlimit_nofile, &saved_rlimit_memlock);
2837
2838 /* Reset all signal handlers. */
2839 (void) reset_all_signal_handlers();
2840 (void) ignore_signals(SIGNALS_IGNORE);
2841
2842 (void) parse_configuration(&saved_rlimit_nofile, &saved_rlimit_memlock);
2843
2844 r = parse_argv(argc, argv);
2845 if (r < 0) {
2846 error_message = "Failed to parse commandline arguments";
2847 goto finish;
2848 }
2849
2850 r = safety_checks();
2851 if (r < 0)
2852 goto finish;
2853
2854 if (IN_SET(arg_action, ACTION_TEST, ACTION_HELP, ACTION_DUMP_CONFIGURATION_ITEMS, ACTION_DUMP_BUS_PROPERTIES, ACTION_BUS_INTROSPECT))
2855 pager_open(arg_pager_flags);
2856
2857 if (arg_action != ACTION_RUN)
2858 skip_setup = true;
2859
2860 if (arg_action == ACTION_HELP) {
2861 retval = help() < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
2862 goto finish;
2863 } else if (arg_action == ACTION_VERSION) {
2864 retval = version();
2865 goto finish;
2866 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
2867 unit_dump_config_items(stdout);
2868 retval = EXIT_SUCCESS;
2869 goto finish;
2870 } else if (arg_action == ACTION_DUMP_BUS_PROPERTIES) {
2871 dump_bus_properties(stdout);
2872 retval = EXIT_SUCCESS;
2873 goto finish;
2874 } else if (arg_action == ACTION_BUS_INTROSPECT) {
2875 r = bus_manager_introspect_implementations(stdout, arg_bus_introspect);
2876 retval = r >= 0 ? EXIT_SUCCESS : EXIT_FAILURE;
2877 goto finish;
2878 }
2879
2880 assert_se(IN_SET(arg_action, ACTION_RUN, ACTION_TEST));
2881
2882 /* Move out of the way, so that we won't block unmounts */
2883 assert_se(chdir("/") == 0);
2884
2885 if (arg_action == ACTION_RUN) {
2886 if (!skip_setup) {
2887 /* Apply the systemd.clock_usec= kernel command line switch */
2888 apply_clock_update();
2889
2890 /* Apply random seed from kernel command line */
2891 cmdline_take_random_seed();
2892 }
2893
2894 /* A core pattern might have been specified via the cmdline. */
2895 initialize_core_pattern(skip_setup);
2896
2897 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2898 log_close();
2899
2900 /* Remember open file descriptors for later deserialization */
2901 r = collect_fds(&fds, &error_message);
2902 if (r < 0)
2903 goto finish;
2904
2905 /* Give up any control of the console, but make sure its initialized. */
2906 setup_console_terminal(skip_setup);
2907
2908 /* Open the logging devices, if possible and necessary */
2909 log_open();
2910 }
2911
2912 log_execution_mode(&first_boot);
2913
2914 r = initialize_runtime(skip_setup,
2915 first_boot,
2916 &saved_rlimit_nofile,
2917 &saved_rlimit_memlock,
2918 &error_message);
2919 if (r < 0)
2920 goto finish;
2921
2922 r = manager_new(arg_system ? LOOKUP_SCOPE_SYSTEM : LOOKUP_SCOPE_USER,
2923 arg_action == ACTION_TEST ? MANAGER_TEST_FULL : 0,
2924 &m);
2925 if (r < 0) {
2926 log_emergency_errno(r, "Failed to allocate manager object: %m");
2927 error_message = "Failed to allocate manager object";
2928 goto finish;
2929 }
2930
2931 m->timestamps[MANAGER_TIMESTAMP_KERNEL] = kernel_timestamp;
2932 m->timestamps[MANAGER_TIMESTAMP_INITRD] = initrd_timestamp;
2933 m->timestamps[MANAGER_TIMESTAMP_USERSPACE] = userspace_timestamp;
2934 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START)] = security_start_timestamp;
2935 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH)] = security_finish_timestamp;
2936
2937 set_manager_defaults(m);
2938 set_manager_settings(m);
2939 manager_set_first_boot(m, first_boot);
2940
2941 /* Remember whether we should queue the default job */
2942 queue_default_job = !arg_serialization || arg_switched_root;
2943
2944 before_startup = now(CLOCK_MONOTONIC);
2945
2946 r = manager_startup(m, arg_serialization, fds, /* root= */ NULL);
2947 if (r < 0) {
2948 error_message = "Failed to start up manager";
2949 goto finish;
2950 }
2951
2952 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2953 fds = fdset_free(fds);
2954 arg_serialization = safe_fclose(arg_serialization);
2955
2956 if (queue_default_job) {
2957 r = do_queue_default_job(m, &error_message);
2958 if (r < 0)
2959 goto finish;
2960 }
2961
2962 after_startup = now(CLOCK_MONOTONIC);
2963
2964 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
2965 "Loaded units and determined initial transaction in %s.",
2966 FORMAT_TIMESPAN(after_startup - before_startup, 100 * USEC_PER_MSEC));
2967
2968 if (arg_action == ACTION_TEST) {
2969 manager_test_summary(m);
2970 retval = EXIT_SUCCESS;
2971 goto finish;
2972 }
2973
2974 r = invoke_main_loop(m,
2975 &saved_rlimit_nofile,
2976 &saved_rlimit_memlock,
2977 &retval,
2978 &shutdown_verb,
2979 &fds,
2980 &switch_root_dir,
2981 &switch_root_init,
2982 &error_message);
2983 assert(r < 0 || IN_SET(r, MANAGER_EXIT, /* MANAGER_OK is not expected here. */
2984 MANAGER_RELOAD,
2985 MANAGER_REEXECUTE,
2986 MANAGER_REBOOT,
2987 MANAGER_POWEROFF,
2988 MANAGER_HALT,
2989 MANAGER_KEXEC,
2990 MANAGER_SWITCH_ROOT));
2991
2992 finish:
2993 pager_close();
2994
2995 if (m) {
2996 arg_reboot_watchdog = manager_get_watchdog(m, WATCHDOG_REBOOT);
2997 arg_kexec_watchdog = manager_get_watchdog(m, WATCHDOG_KEXEC);
2998 m = manager_free(m);
2999 }
3000
3001 mac_selinux_finish();
3002
3003 if (IN_SET(r, MANAGER_REEXECUTE, MANAGER_SWITCH_ROOT))
3004 r = do_reexecute(r,
3005 argc, argv,
3006 &saved_rlimit_nofile,
3007 &saved_rlimit_memlock,
3008 fds,
3009 switch_root_dir,
3010 switch_root_init,
3011 &error_message); /* This only returns if reexecution failed */
3012
3013 arg_serialization = safe_fclose(arg_serialization);
3014 fds = fdset_free(fds);
3015
3016 saved_env = strv_free(saved_env);
3017
3018 #if HAVE_VALGRIND_VALGRIND_H
3019 /* If we are PID 1 and running under valgrind, then let's exit
3020 * here explicitly. valgrind will only generate nice output on
3021 * exit(), not on exec(), hence let's do the former not the
3022 * latter here. */
3023 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND) {
3024 /* Cleanup watchdog_device strings for valgrind. We need them
3025 * in become_shutdown() so normally we cannot free them yet. */
3026 watchdog_free_device();
3027 reset_arguments();
3028 return retval;
3029 }
3030 #endif
3031
3032 #if HAS_FEATURE_ADDRESS_SANITIZER
3033 __lsan_do_leak_check();
3034 #endif
3035
3036 /* Try to invoke the shutdown binary unless we already failed.
3037 * If we failed above, we want to freeze after finishing cleanup. */
3038 if (r >= 0 && shutdown_verb) {
3039 r = become_shutdown(shutdown_verb, retval);
3040 log_error_errno(r, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
3041 error_message = "Failed to execute shutdown binary";
3042 }
3043
3044 watchdog_free_device();
3045 arg_watchdog_device = mfree(arg_watchdog_device);
3046
3047 if (getpid_cached() == 1) {
3048 if (error_message)
3049 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
3050 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
3051 "%s.", error_message);
3052 freeze_or_exit_or_reboot();
3053 }
3054
3055 reset_arguments();
3056 return retval;
3057 }
Unnamed repository; edit this file 'description' to name the repository.