]> git.ipfire.org Git - thirdparty/systemd.git/blame - src/core/main.c
projects / thirdparty / systemd.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
pid1: also process memory pressure events (and SIGRTMIN+18)
[thirdparty/systemd.git] / src / core / main.c
CommitLineData
db9ecf05 1/* SPDX-License-Identifier: LGPL-2.1-or-later */
a7334b09 2
60918275 3#include <errno.h>
3dfc9763 4#include <fcntl.h>
f170852a 5#include <getopt.h>
d4a402e4 6#include <linux/oom.h>
664f88a7 7#include <sys/mount.h>
3dfc9763 8#include <sys/prctl.h>
40efaaed 9#include <sys/utsname.h>
3dfc9763 10#include <unistd.h>
349cc4a5 11#if HAVE_SECCOMP
b64a3d86
LP		 12#include <seccomp.h>
13#endif
349cc4a5 14#if HAVE_VALGRIND_VALGRIND_H
50b35193 15# include <valgrind/valgrind.h>
3dfc9763 16#endif
54b434b1 17
718db961 18#include "sd-bus.h"
cf0fbc49 19#include "sd-daemon.h"
b2e7486c 20#include "sd-messages.h"
3dfc9763 21
b5efdb8a 22#include "alloc-util.h"
2ffadd3c 23#include "apparmor-setup.h"
d9d93745 24#include "architecture.h"
ee617a4e 25#include "argv-util.h"
b1994387
ILG		 26#if HAVE_LIBBPF
27#include "bpf-lsm.h"
28#endif
3dfc9763
LP		 29#include "build.h"
30#include "bus-error.h"
31#include "bus-util.h"
430f0182 32#include "capability-util.h"
a88c5b8a 33#include "cgroup-util.h"
24efb112 34#include "clock-util.h"
3dfc9763 35#include "conf-parser.h"
618234a5 36#include "cpu-set-util.h"
898c9a6f 37#include "crash-handler.h"
3dfc9763 38#include "dbus-manager.h"
c18ecf03 39#include "dbus.h"
28db6fbf 40#include "constants.h"
32429805 41#include "dev-setup.h"
c18ecf03 42#include "efi-random.h"
209b2592 43#include "efivars.h"
eee8b7ab 44#include "emergency-action.h"
3dfc9763 45#include "env-util.h"
57b7a260 46#include "exit-status.h"
3ffd4af2 47#include "fd-util.h"
3dfc9763 48#include "fdset.h"
718db961 49#include "fileio.h"
f97b34a6 50#include "format-util.h"
f4f15635 51#include "fs-util.h"
d247f232 52#include "hexdecoct.h"
3dfc9763
LP		 53#include "hostname-setup.h"
54#include "ima-setup.h"
4b9a4b01 55#include "import-creds.h"
baa6a42d 56#include "initrd-util.h"
3dfc9763
LP		 57#include "killall.h"
58#include "kmod-setup.h"
eefc66aa 59#include "limits-util.h"
d7b8eec7 60#include "load-fragment.h"
3dfc9763 61#include "log.h"
b6e66135 62#include "loopback-setup.h"
b6e66135 63#include "machine-id-setup.h"
898c9a6f 64#include "main.h"
3dfc9763 65#include "manager.h"
2a341bb9 66#include "manager-dump.h"
a01ba4b2 67#include "manager-serialize.h"
35cd0ba5 68#include "mkdir-label.h"
3dfc9763 69#include "mount-setup.h"
d58ad743 70#include "os-util.h"
3dfc9763 71#include "pager.h"
614b022c 72#include "parse-argument.h"
6bedfcbb 73#include "parse-util.h"
7d5ceb64 74#include "path-util.h"
294bf0c3 75#include "pretty-print.h"
4e731273 76#include "proc-cmdline.h"
3dfc9763 77#include "process-util.h"
d247f232 78#include "random-util.h"
78f22b97 79#include "rlimit-util.h"
349cc4a5 80#if HAVE_SECCOMP
83f12b27
FS		 81#include "seccomp-util.h"
82#endif
b6e66135 83#include "selinux-setup.h"
3dfc9763
LP		 84#include "selinux-util.h"
85#include "signal-util.h"
ffbd2c4d 86#include "smack-setup.h"
3dfc9763 87#include "special.h"
8fcde012 88#include "stat-util.h"
15a5e950 89#include "stdio-util.h"
3dfc9763
LP		 90#include "strv.h"
91#include "switch-root.h"
a8b627aa 92#include "sysctl-util.h"
3dfc9763 93#include "terminal-util.h"
b10abe4b 94#include "time-util.h"
8612da97 95#include "umask-util.h"
b1d4f8e1 96#include "user-util.h"
3dfc9763
LP		 97#include "virt.h"
98#include "watchdog.h"
b6e66135 99
7e11a95e
EV		 100#if HAS_FEATURE_ADDRESS_SANITIZER
101#include <sanitizer/lsan_interface.h>
102#endif
103
3a0f06c4
ZJS		 104#define DEFAULT_TASKS_MAX ((TasksMax) { 15U, 100U }) /* 15% */
105
f170852a
LP		 106static enum {
107 ACTION_RUN,
e965d56d 108 ACTION_HELP,
9ba0bc4e 109 ACTION_VERSION,
e537352b 110 ACTION_TEST,
bbc1acab
YW		 111 ACTION_DUMP_CONFIGURATION_ITEMS,
112 ACTION_DUMP_BUS_PROPERTIES,
5c08257b 113 ACTION_BUS_INTROSPECT,
fa0f4d8a 114} arg_action = ACTION_RUN;
fb39af4c 115
5c08257b
ZJS		 116static const char *arg_bus_introspect = NULL;
117
45250e66
LP		 118/* Those variables are initialized to 0 automatically, so we avoid uninitialized memory access. Real
119 * defaults are assigned in reset_arguments() below. */
fb39af4c
ZJS		 120static char *arg_default_unit;
121static bool arg_system;
898c9a6f
LP		 122bool arg_dump_core;
123int arg_crash_chvt;
124bool arg_crash_shell;
125bool arg_crash_reboot;
fb39af4c
ZJS		 126static char *arg_confirm_spawn;
127static ShowStatus arg_show_status;
36cf4507 128static StatusUnitFormat arg_status_unit_format;
fb39af4c
ZJS		 129static bool arg_switched_root;
130static PagerFlags arg_pager_flags;
131static bool arg_service_watchdogs;
132static ExecOutput arg_default_std_output;
133static ExecOutput arg_default_std_error;
134static usec_t arg_default_restart_usec;
135static usec_t arg_default_timeout_start_usec;
136static usec_t arg_default_timeout_stop_usec;
137static usec_t arg_default_timeout_abort_usec;
a0fe19f9 138static usec_t arg_default_device_timeout_usec;
fb39af4c
ZJS		 139static bool arg_default_timeout_abort_set;
140static usec_t arg_default_start_limit_interval;
141static unsigned arg_default_start_limit_burst;
142static usec_t arg_runtime_watchdog;
65224c1d 143static usec_t arg_reboot_watchdog;
acafd7d8 144static usec_t arg_kexec_watchdog;
5717062e 145static usec_t arg_pretimeout_watchdog;
fb39af4c 146static char *arg_early_core_pattern;
aff3a9e1 147static char *arg_watchdog_pretimeout_governor;
fb39af4c
ZJS		 148static char *arg_watchdog_device;
149static char **arg_default_environment;
d55ed7de 150static char **arg_manager_environment;
fb39af4c
ZJS		 151static struct rlimit *arg_default_rlimit[_RLIMIT_MAX];
152static uint64_t arg_capability_bounding_set;
153static bool arg_no_new_privs;
154static nsec_t arg_timer_slack_nsec;
155static usec_t arg_default_timer_accuracy_usec;
156static Set* arg_syscall_archs;
157static FILE* arg_serialization;
158static int arg_default_cpu_accounting;
159static bool arg_default_io_accounting;
160static bool arg_default_ip_accounting;
161static bool arg_default_blockio_accounting;
162static bool arg_default_memory_accounting;
163static bool arg_default_tasks_accounting;
3a0f06c4 164static TasksMax arg_default_tasks_max;
fb39af4c
ZJS		 165static sd_id128_t arg_machine_id;
166static EmergencyAction arg_cad_burst_action;
167static OOMPolicy arg_default_oom_policy;
168static CPUSet arg_cpu_affinity;
b070c7c0 169static NUMAPolicy arg_numa_policy;
3753325b 170static usec_t arg_clock_usec;
d247f232
LP		 171static void *arg_random_seed;
172static size_t arg_random_seed_size;
d4a402e4
LP		 173static int arg_default_oom_score_adjust;
174static bool arg_default_oom_score_adjust_set;
aa5ae971 175static char *arg_default_smack_process_label;
856bfaeb
LB		 176static usec_t arg_reload_limit_interval_sec;
177static unsigned arg_reload_limit_burst;
61fbbac1 178
0e06a031
LP		 179/* A copy of the original environment block */
180static char **saved_env = NULL;
181
a9fd4cd1
FB		 182static int parse_configuration(const struct rlimit *saved_rlimit_nofile,
183 const struct rlimit *saved_rlimit_memlock);
4fc935ca 184
f70e6fb4
ZJS		 185static int manager_find_user_config_paths(char ***ret_files, char ***ret_dirs) {
186 _cleanup_free_ char *base = NULL;
187 _cleanup_strv_free_ char **files = NULL, **dirs = NULL;
188 int r;
189
190 r = xdg_user_config_dir(&base, "/systemd");
191 if (r < 0)
192 return r;
193
194 r = strv_extendf(&files, "%s/user.conf", base);
195 if (r < 0)
196 return r;
197
198 r = strv_extend(&files, PKGSYSCONFDIR "/user.conf");
199 if (r < 0)
200 return r;
201
202 r = strv_consume(&dirs, TAKE_PTR(base));
203 if (r < 0)
204 return r;
205
206 r = strv_extend_strv(&dirs, CONF_PATHS_STRV("systemd"), false);
207 if (r < 0)
208 return r;
209
210 *ret_files = TAKE_PTR(files);
211 *ret_dirs = TAKE_PTR(dirs);
212 return 0;
213}
214
56d96fc0 215static int console_setup(void) {
254d1313 216 _cleanup_close_ int tty_fd = -EBADF;
56d96fc0 217 int r;
80876c20 218
512947d4 219 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
23bbb0de
MS		 220 if (tty_fd < 0)
221 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
80876c20 222
56d96fc0
LP		 223 /* We don't want to force text mode. plymouth may be showing
224 * pictures already from initrd. */
512947d4 225 r = reset_terminal_fd(tty_fd, false);
23bbb0de
MS		 226 if (r < 0)
227 return log_error_errno(r, "Failed to reset /dev/console: %m");
843d2643 228
56d96fc0 229 return 0;
80876c20
LP		 230}
231
ee48dbd5 232static int set_machine_id(const char *m) {
e042eab7 233 sd_id128_t t;
8b26cdbd 234 assert(m);
ee48dbd5 235
e042eab7 236 if (sd_id128_from_string(m, &t) < 0)
ee48dbd5
NC		 237 return -EINVAL;
238
e042eab7 239 if (sd_id128_is_null(t))
ee48dbd5
NC		 240 return -EINVAL;
241
e042eab7 242 arg_machine_id = t;
ee48dbd5
NC		 243 return 0;
244}
245
96287a49 246static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
059cb385 247 int r;
f170852a 248
059cb385 249 assert(key);
5192bd19 250
1d84ad94 251 if (STR_IN_SET(key, "systemd.unit", "rd.systemd.unit")) {
bf4df7c3 252
1d84ad94
LP		 253 if (proc_cmdline_value_missing(key, value))
254 return 0;
bf4df7c3 255
1d84ad94
LP		 256 if (!unit_name_is_valid(value, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE))
257 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key, value);
cd57038a
ZJS		 258 else if (in_initrd() == !!startswith(key, "rd."))
259 return free_and_strdup_warn(&arg_default_unit, value);
f170852a 260
1d84ad94 261 } else if (proc_cmdline_key_streq(key, "systemd.dump_core")) {
4fc935ca 262
1d84ad94 263 r = value ? parse_boolean(value) : true;
fb472900 264 if (r < 0)
5e1ee764 265 log_warning_errno(r, "Failed to parse dump core switch %s, ignoring: %m", value);
4fc935ca 266 else
fa0f4d8a 267 arg_dump_core = r;
4fc935ca 268
c6885f5f
FB		 269 } else if (proc_cmdline_key_streq(key, "systemd.early_core_pattern")) {
270
271 if (proc_cmdline_value_missing(key, value))
272 return 0;
273
274 if (path_is_absolute(value))
614b022c 275 (void) parse_path_argument(value, false, &arg_early_core_pattern);
c6885f5f
FB		 276 else
277 log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value);
278
1d84ad94 279 } else if (proc_cmdline_key_streq(key, "systemd.crash_chvt")) {
b9e74c39 280
1d84ad94
LP		 281 if (!value)
282 arg_crash_chvt = 0; /* turn on */
5e1ee764 283 else {
a07a7324 284 r = parse_crash_chvt(value, &arg_crash_chvt);
5e1ee764
YW		 285 if (r < 0)
286 log_warning_errno(r, "Failed to parse crash chvt switch %s, ignoring: %m", value);
287 }
b9e74c39 288
1d84ad94 289 } else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) {
4fc935ca 290
1d84ad94 291 r = value ? parse_boolean(value) : true;
fb472900 292 if (r < 0)
5e1ee764 293 log_warning_errno(r, "Failed to parse crash shell switch %s, ignoring: %m", value);
4fc935ca 294 else
fa0f4d8a 295 arg_crash_shell = r;
5e7ee61c 296
1d84ad94 297 } else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) {
5e7ee61c 298
1d84ad94 299 r = value ? parse_boolean(value) : true;
b9e74c39 300 if (r < 0)
5e1ee764 301 log_warning_errno(r, "Failed to parse crash reboot switch %s, ignoring: %m", value);
5e7ee61c 302 else
b9e74c39 303 arg_crash_reboot = r;
5e7ee61c 304
1d84ad94
LP		 305 } else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) {
306 char *s;
7d5ceb64 307
1d84ad94 308 r = parse_confirm_spawn(value, &s);
059cb385 309 if (r < 0)
5e1ee764
YW		 310 log_warning_errno(r, "Failed to parse confirm_spawn switch %s, ignoring: %m", value);
311 else
312 free_and_replace(arg_confirm_spawn, s);
601f6a1e 313
2a12e32e
JK		 314 } else if (proc_cmdline_key_streq(key, "systemd.service_watchdogs")) {
315
316 r = value ? parse_boolean(value) : true;
317 if (r < 0)
5e1ee764 318 log_warning_errno(r, "Failed to parse service watchdog switch %s, ignoring: %m", value);
2a12e32e
JK		 319 else
320 arg_service_watchdogs = r;
321
1d84ad94 322 } else if (proc_cmdline_key_streq(key, "systemd.show_status")) {
9e58ff9c 323
1d84ad94
LP		 324 if (value) {
325 r = parse_show_status(value, &arg_show_status);
326 if (r < 0)
5e1ee764 327 log_warning_errno(r, "Failed to parse show status switch %s, ignoring: %m", value);
1d84ad94
LP		 328 } else
329 arg_show_status = SHOW_STATUS_YES;
059cb385 330
36cf4507
ZJS		 331 } else if (proc_cmdline_key_streq(key, "systemd.status_unit_format")) {
332
333 if (proc_cmdline_value_missing(key, value))
334 return 0;
335
336 r = status_unit_format_from_string(value);
337 if (r < 0)
338 log_warning_errno(r, "Failed to parse %s=%s, ignoring: %m", key, value);
339 else
340 arg_status_unit_format = r;
341
1d84ad94
LP		 342 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) {
343
344 if (proc_cmdline_value_missing(key, value))
345 return 0;
0a494f1f 346
059cb385 347 r = exec_output_from_string(value);
fb472900 348 if (r < 0)
5e1ee764 349 log_warning_errno(r, "Failed to parse default standard output switch %s, ignoring: %m", value);
0a494f1f
LP		 350 else
351 arg_default_std_output = r;
0a494f1f 352
1d84ad94
LP		 353 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) {
354
355 if (proc_cmdline_value_missing(key, value))
356 return 0;
059cb385
LP		 357
358 r = exec_output_from_string(value);
fb472900 359 if (r < 0)
5e1ee764 360 log_warning_errno(r, "Failed to parse default standard error switch %s, ignoring: %m", value);
0a494f1f
LP		 361 else
362 arg_default_std_error = r;
9e7c5357 363
1d84ad94
LP		 364 } else if (streq(key, "systemd.setenv")) {
365
366 if (proc_cmdline_value_missing(key, value))
367 return 0;
059cb385 368
b70935ac
ZJS		 369 if (!env_assignment_is_valid(value))
370 log_warning("Environment variable assignment '%s' is not valid. Ignoring.", value);
371 else {
372 r = strv_env_replace_strdup(&arg_default_environment, value);
373 if (r < 0)
1d84ad94 374 return log_oom();
b70935ac 375 }
9e58ff9c 376
1d84ad94
LP		 377 } else if (proc_cmdline_key_streq(key, "systemd.machine_id")) {
378
379 if (proc_cmdline_value_missing(key, value))
380 return 0;
381
382 r = set_machine_id(value);
383 if (r < 0)
5e1ee764 384 log_warning_errno(r, "MachineID '%s' is not valid, ignoring: %m", value);
ee48dbd5 385
1d84ad94
LP		 386 } else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) {
387
388 if (proc_cmdline_value_missing(key, value))
389 return 0;
390
391 r = parse_sec(value, &arg_default_timeout_start_usec);
392 if (r < 0)
5e1ee764 393 log_warning_errno(r, "Failed to parse default start timeout '%s', ignoring: %m", value);
1d84ad94
LP		 394
395 if (arg_default_timeout_start_usec <= 0)
396 arg_default_timeout_start_usec = USEC_INFINITY;
ee48dbd5 397
68d58f38
LP		 398 } else if (proc_cmdline_key_streq(key, "systemd.cpu_affinity")) {
399
400 if (proc_cmdline_value_missing(key, value))
401 return 0;
402
403 r = parse_cpu_set(value, &arg_cpu_affinity);
404 if (r < 0)
162392b7 405 log_warning_errno(r, "Failed to parse CPU affinity mask '%s', ignoring: %m", value);
68d58f38 406
8a2c1fbf
EJ		 407 } else if (proc_cmdline_key_streq(key, "systemd.watchdog_device")) {
408
409 if (proc_cmdline_value_missing(key, value))
410 return 0;
411
614b022c 412 (void) parse_path_argument(value, false, &arg_watchdog_device);
8a2c1fbf 413
b3aa73e4
FB		 414 } else if (proc_cmdline_key_streq(key, "systemd.watchdog_sec")) {
415
416 if (proc_cmdline_value_missing(key, value))
417 return 0;
418
8a85c5b6
FB		 419 if (streq(value, "default"))
420 arg_runtime_watchdog = USEC_INFINITY;
c91c95e6
LP		 421 else if (streq(value, "off"))
422 arg_runtime_watchdog = 0;
8a85c5b6
FB		 423 else {
424 r = parse_sec(value, &arg_runtime_watchdog);
425 if (r < 0) {
426 log_warning_errno(r, "Failed to parse systemd.watchdog_sec= argument '%s', ignoring: %m", value);
427 return 0;
428 }
429 }
430
431 arg_kexec_watchdog = arg_reboot_watchdog = arg_runtime_watchdog;
b3aa73e4 432
5717062e
CK		 433 } else if (proc_cmdline_key_streq(key, "systemd.watchdog_pre_sec")) {
434
435 if (proc_cmdline_value_missing(key, value))
436 return 0;
437
438 if (streq(value, "default"))
439 arg_pretimeout_watchdog = USEC_INFINITY;
440 else if (streq(value, "off"))
441 arg_pretimeout_watchdog = 0;
442 else {
443 r = parse_sec(value, &arg_pretimeout_watchdog);
444 if (r < 0) {
445 log_warning_errno(r, "Failed to parse systemd.watchdog_pre_sec= argument '%s', ignoring: %m", value);
446 return 0;
447 }
448 }
449
aff3a9e1
LB		 450 } else if (proc_cmdline_key_streq(key, "systemd.watchdog_pretimeout_governor")) {
451
452 if (proc_cmdline_value_missing(key, value) || isempty(value)) {
453 arg_watchdog_pretimeout_governor = mfree(arg_watchdog_pretimeout_governor);
454 return 0;
455 }
456
457 if (!string_is_safe(value)) {
458 log_warning("Watchdog pretimeout governor '%s' is not valid, ignoring.", value);
459 return 0;
460 }
461
462 return free_and_strdup_warn(&arg_watchdog_pretimeout_governor, value);
463
3753325b
LP		 464 } else if (proc_cmdline_key_streq(key, "systemd.clock_usec")) {
465
466 if (proc_cmdline_value_missing(key, value))
467 return 0;
468
469 r = safe_atou64(value, &arg_clock_usec);
470 if (r < 0)
471 log_warning_errno(r, "Failed to parse systemd.clock_usec= argument, ignoring: %s", value);
472
d247f232
LP		 473 } else if (proc_cmdline_key_streq(key, "systemd.random_seed")) {
474 void *p;
475 size_t sz;
476
477 if (proc_cmdline_value_missing(key, value))
478 return 0;
479
f5fbe71d 480 r = unbase64mem(value, SIZE_MAX, &p, &sz);
d247f232
LP		 481 if (r < 0)
482 log_warning_errno(r, "Failed to parse systemd.random_seed= argument, ignoring: %s", value);
483
484 free(arg_random_seed);
485 arg_random_seed = sz > 0 ? p : mfree(p);
486 arg_random_seed_size = sz;
487
856bfaeb
LB		 488 } else if (proc_cmdline_key_streq(key, "systemd.reload_limit_interval_sec")) {
489
490 if (proc_cmdline_value_missing(key, value))
491 return 0;
492
493 r = parse_sec(value, &arg_reload_limit_interval_sec);
494 if (r < 0) {
495 log_warning_errno(r, "Failed to parse systemd.reload_limit_interval_sec= argument '%s', ignoring: %m", value);
496 return 0;
497 }
498
499 } else if (proc_cmdline_key_streq(key, "systemd.reload_limit_burst")) {
500
501 if (proc_cmdline_value_missing(key, value))
502 return 0;
503
504 r = safe_atou(value, &arg_reload_limit_burst);
505 if (r < 0) {
506 log_warning_errno(r, "Failed to parse systemd.reload_limit_burst= argument '%s', ignoring: %m", value);
507 return 0;
508 }
509
059cb385 510 } else if (streq(key, "quiet") && !value) {
d7b15e0a 511
7a293242 512 if (arg_show_status == _SHOW_STATUS_INVALID)
0d066dd1 513 arg_show_status = SHOW_STATUS_ERROR;
059cb385
LP		 514
515 } else if (streq(key, "debug") && !value) {
d7b15e0a 516
1de1c9c3
LP		 517 /* Note that log_parse_environment() handles 'debug'
518 * too, and sets the log level to LOG_DEBUG. */
d7b15e0a 519
75f86906 520 if (detect_container() > 0)
b2103dcc 521 log_set_target(LOG_TARGET_CONSOLE);
059cb385 522
dcd61450 523 } else if (!value) {
e2c9a131 524 const char *target;
f170852a 525
ceae6295 526 /* Compatible with SysV, but supported independently even if SysV compatibility is disabled. */
e2c9a131
EV		 527 target = runlevel_to_target(key);
528 if (target)
cd57038a 529 return free_and_strdup_warn(&arg_default_unit, target);
f170852a
LP		 530 }
531
532 return 0;
533}
534
e8e581bf
ZJS		 535#define DEFINE_SETTER(name, func, descr) \
536 static int name(const char *unit, \
537 const char *filename, \
538 unsigned line, \
539 const char *section, \
71a61510 540 unsigned section_line, \
e8e581bf
ZJS		 541 const char *lvalue, \
542 int ltype, \
543 const char *rvalue, \
544 void *data, \
545 void *userdata) { \
546 \
547 int r; \
548 \
549 assert(filename); \
550 assert(lvalue); \
551 assert(rvalue); \
552 \
553 r = func(rvalue); \
554 if (r < 0) \
d1cefe0a
LP		 555 log_syntax(unit, LOG_ERR, filename, line, r, \
556 "Invalid " descr "'%s': %m", \
557 rvalue); \
e8e581bf
ZJS		 558 \
559 return 0; \
560 }
487393e9 561
a6ecbf83
FB		 562DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level");
563DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target");
c5673ed0 564DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color");
a6ecbf83 565DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location");
c5673ed0 566DEFINE_SETTER(config_parse_time, log_show_time_from_string, "time");
487393e9 567
a61d6874
ZJS		 568static int config_parse_default_timeout_abort(
569 const char *unit,
570 const char *filename,
571 unsigned line,
572 const char *section,
573 unsigned section_line,
574 const char *lvalue,
575 int ltype,
576 const char *rvalue,
577 void *data,
578 void *userdata) {
579 int r;
580
581 r = config_parse_timeout_abort(unit, filename, line, section, section_line, lvalue, ltype, rvalue,
582 &arg_default_timeout_abort_usec, userdata);
583 if (r >= 0)
584 arg_default_timeout_abort_set = r;
585 return 0;
586}
487393e9 587
d4a402e4
LP		 588static int config_parse_oom_score_adjust(
589 const char *unit,
590 const char *filename,
591 unsigned line,
592 const char *section,
593 unsigned section_line,
594 const char *lvalue,
595 int ltype,
596 const char *rvalue,
597 void *data,
598 void *userdata) {
599
600 int oa, r;
601
602 if (isempty(rvalue)) {
603 arg_default_oom_score_adjust_set = false;
604 return 0;
605 }
606
607 r = parse_oom_score_adjust(rvalue, &oa);
608 if (r < 0) {
609 log_syntax(unit, LOG_WARNING, filename, line, r, "Failed to parse the OOM score adjust value '%s', ignoring: %m", rvalue);
610 return 0;
611 }
612
613 arg_default_oom_score_adjust = oa;
614 arg_default_oom_score_adjust_set = true;
615
616 return 0;
617}
618
a61d6874 619static int parse_config_file(void) {
f975e971 620 const ConfigTableItem items[] = {
3f87eaa5
YW		 621 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
622 { "Manager", "LogTarget", config_parse_target, 0, NULL },
623 { "Manager", "LogColor", config_parse_color, 0, NULL },
624 { "Manager", "LogLocation", config_parse_location, 0, NULL },
625 { "Manager", "LogTime", config_parse_time, 0, NULL },
626 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
627 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, &arg_crash_chvt },
628 { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, &arg_crash_chvt },
629 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
630 { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
631 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
632 { "Manager", "StatusUnitFormat", config_parse_status_unit_format, 0, &arg_status_unit_format },
633 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, &arg_cpu_affinity },
634 { "Manager", "NUMAPolicy", config_parse_numa_policy, 0, &arg_numa_policy.type },
635 { "Manager", "NUMAMask", config_parse_numa_mask, 0, &arg_numa_policy },
636 { "Manager", "JoinControllers", config_parse_warn_compat, DISABLED_CONFIGURATION, NULL },
637 { "Manager", "RuntimeWatchdogSec", config_parse_watchdog_sec, 0, &arg_runtime_watchdog },
638 { "Manager", "RuntimeWatchdogPreSec", config_parse_watchdog_sec, 0, &arg_pretimeout_watchdog },
639 { "Manager", "RebootWatchdogSec", config_parse_watchdog_sec, 0, &arg_reboot_watchdog },
640 { "Manager", "ShutdownWatchdogSec", config_parse_watchdog_sec, 0, &arg_reboot_watchdog }, /* obsolete alias */
641 { "Manager", "KExecWatchdogSec", config_parse_watchdog_sec, 0, &arg_kexec_watchdog },
642 { "Manager", "WatchdogDevice", config_parse_path, 0, &arg_watchdog_device },
643 { "Manager", "RuntimeWatchdogPreGovernor", config_parse_string, CONFIG_PARSE_STRING_SAFE, &arg_watchdog_pretimeout_governor },
644 { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
645 { "Manager", "NoNewPrivileges", config_parse_bool, 0, &arg_no_new_privs },
349cc4a5 646#if HAVE_SECCOMP
3f87eaa5 647 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
6aa2c555
LP		 648#else
649 { "Manager", "SystemCallArchitectures", config_parse_warn_compat, DISABLED_CONFIGURATION, NULL },
650
89fffa27 651#endif
3f87eaa5
YW		 652 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
653 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
654 { "Manager", "DefaultStandardOutput", config_parse_output_restricted, 0, &arg_default_std_output },
655 { "Manager", "DefaultStandardError", config_parse_output_restricted, 0, &arg_default_std_error },
656 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
657 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
658 { "Manager", "DefaultTimeoutAbortSec", config_parse_default_timeout_abort, 0, NULL },
a0fe19f9 659 { "Manager", "DefaultDeviceTimeoutSec", config_parse_sec, 0, &arg_default_device_timeout_usec },
3f87eaa5
YW		 660 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
661 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */
662 { "Manager", "DefaultStartLimitIntervalSec", config_parse_sec, 0, &arg_default_start_limit_interval },
663 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
664 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
665 { "Manager", "ManagerEnvironment", config_parse_environ, 0, &arg_manager_environment },
666 { "Manager", "DefaultLimitCPU", config_parse_rlimit, RLIMIT_CPU, arg_default_rlimit },
667 { "Manager", "DefaultLimitFSIZE", config_parse_rlimit, RLIMIT_FSIZE, arg_default_rlimit },
668 { "Manager", "DefaultLimitDATA", config_parse_rlimit, RLIMIT_DATA, arg_default_rlimit },
669 { "Manager", "DefaultLimitSTACK", config_parse_rlimit, RLIMIT_STACK, arg_default_rlimit },
670 { "Manager", "DefaultLimitCORE", config_parse_rlimit, RLIMIT_CORE, arg_default_rlimit },
671 { "Manager", "DefaultLimitRSS", config_parse_rlimit, RLIMIT_RSS, arg_default_rlimit },
672 { "Manager", "DefaultLimitNOFILE", config_parse_rlimit, RLIMIT_NOFILE, arg_default_rlimit },
673 { "Manager", "DefaultLimitAS", config_parse_rlimit, RLIMIT_AS, arg_default_rlimit },
674 { "Manager", "DefaultLimitNPROC", config_parse_rlimit, RLIMIT_NPROC, arg_default_rlimit },
675 { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit, RLIMIT_MEMLOCK, arg_default_rlimit },
676 { "Manager", "DefaultLimitLOCKS", config_parse_rlimit, RLIMIT_LOCKS, arg_default_rlimit },
677 { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit, RLIMIT_SIGPENDING, arg_default_rlimit },
678 { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit, RLIMIT_MSGQUEUE, arg_default_rlimit },
679 { "Manager", "DefaultLimitNICE", config_parse_rlimit, RLIMIT_NICE, arg_default_rlimit },
680 { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit, RLIMIT_RTPRIO, arg_default_rlimit },
681 { "Manager", "DefaultLimitRTTIME", config_parse_rlimit, RLIMIT_RTTIME, arg_default_rlimit },
682 { "Manager", "DefaultCPUAccounting", config_parse_tristate, 0, &arg_default_cpu_accounting },
683 { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting },
684 { "Manager", "DefaultIPAccounting", config_parse_bool, 0, &arg_default_ip_accounting },
685 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
686 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
687 { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
688 { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max },
724f061d 689 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action, arg_system, &arg_cad_burst_action },
3f87eaa5
YW		 690 { "Manager", "DefaultOOMPolicy", config_parse_oom_policy, 0, &arg_default_oom_policy },
691 { "Manager", "DefaultOOMScoreAdjust", config_parse_oom_score_adjust, 0, NULL },
856bfaeb
LB		 692 { "Manager", "ReloadLimitIntervalSec", config_parse_sec, 0, &arg_reload_limit_interval_sec },
693 { "Manager", "ReloadLimitBurst", config_parse_unsigned, 0, &arg_reload_limit_burst },
aa5ae971
ŁS		 694#if ENABLE_SMACK
695 { "Manager", "DefaultSmackProcessLabel", config_parse_string, 0, &arg_default_smack_process_label },
696#else
697 { "Manager", "DefaultSmackProcessLabel", config_parse_warn_compat, DISABLED_CONFIGURATION, NULL },
698#endif
d3b1c508 699 {}
487393e9
LP		 700 };
701
e94a009c
YW		 702 _cleanup_strv_free_ char **files = NULL, **dirs = NULL;
703 const char *suffix;
f70e6fb4 704 int r;
75eb6154 705
e94a009c 706 if (arg_system)
f70e6fb4 707 suffix = "system.conf.d";
e94a009c
YW		 708 else {
709 r = manager_find_user_config_paths(&files, &dirs);
f70e6fb4
ZJS		 710 if (r < 0)
711 return log_error_errno(r, "Failed to determine config file paths: %m");
e94a009c 712
f70e6fb4
ZJS		 713 suffix = "user.conf.d";
714 }
75eb6154 715
f70e6fb4 716 (void) config_parse_many(
e94a009c
YW		 717 (const char* const*) (files ?: STRV_MAKE(PKGSYSCONFDIR "/system.conf")),
718 (const char* const*) (dirs ?: CONF_PATHS_STRV("systemd")),
719 suffix,
4f9ff96a
LP		 720 "Manager\0",
721 config_item_table_lookup, items,
722 CONFIG_PARSE_WARN,
723 NULL,
ead3a3fc 724 NULL,
4f9ff96a 725 NULL);
36c16a7c 726
f70e6fb4
ZJS		 727 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we use
728 * USEC_INFINITY like everywhere else. */
36c16a7c
LP		 729 if (arg_default_timeout_start_usec <= 0)
730 arg_default_timeout_start_usec = USEC_INFINITY;
731 if (arg_default_timeout_stop_usec <= 0)
732 arg_default_timeout_stop_usec = USEC_INFINITY;
487393e9 733
487393e9
LP		 734 return 0;
735}
736
85cb4151 737static void set_manager_defaults(Manager *m) {
06af2a04
TB		 738
739 assert(m);
740
5b65ae15
LP		 741 /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
742 * affect the manager itself, but are just what newly allocated units will have set if they haven't set
743 * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
744
06af2a04
TB		 745 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
746 m->default_std_output = arg_default_std_output;
747 m->default_std_error = arg_default_std_error;
748 m->default_timeout_start_usec = arg_default_timeout_start_usec;
749 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
dc653bf4
JK		 750 m->default_timeout_abort_usec = arg_default_timeout_abort_usec;
751 m->default_timeout_abort_set = arg_default_timeout_abort_set;
a0fe19f9 752 m->default_device_timeout_usec = arg_default_device_timeout_usec;
06af2a04
TB		 753 m->default_restart_usec = arg_default_restart_usec;
754 m->default_start_limit_interval = arg_default_start_limit_interval;
755 m->default_start_limit_burst = arg_default_start_limit_burst;
a88c5b8a
CD		 756
757 /* On 4.15+ with unified hierarchy, CPU accounting is essentially free as it doesn't require the CPU
758 * controller to be enabled, so the default is to enable it unless we got told otherwise. */
759 if (arg_default_cpu_accounting >= 0)
760 m->default_cpu_accounting = arg_default_cpu_accounting;
761 else
762 m->default_cpu_accounting = cpu_accounting_is_cheap();
763
13c31542 764 m->default_io_accounting = arg_default_io_accounting;
377bfd2d 765 m->default_ip_accounting = arg_default_ip_accounting;
06af2a04
TB		 766 m->default_blockio_accounting = arg_default_blockio_accounting;
767 m->default_memory_accounting = arg_default_memory_accounting;
03a7b521 768 m->default_tasks_accounting = arg_default_tasks_accounting;
0af20ea2 769 m->default_tasks_max = arg_default_tasks_max;
afcfaa69 770 m->default_oom_policy = arg_default_oom_policy;
d4a402e4
LP		 771 m->default_oom_score_adjust_set = arg_default_oom_score_adjust_set;
772 m->default_oom_score_adjust = arg_default_oom_score_adjust;
06af2a04 773
aa5ae971
ŁS		 774 (void) manager_set_default_smack_process_label(m, arg_default_smack_process_label);
775
79a224c4
LP		 776 (void) manager_set_default_rlimits(m, arg_default_rlimit);
777
778 (void) manager_default_environment(m);
779 (void) manager_transient_environment_add(m, arg_default_environment);
06af2a04
TB		 780}
781
7b46fc6a 782static void set_manager_settings(Manager *m) {
aff3a9e1 783 int r;
7b46fc6a
LP		 784
785 assert(m);
786
986935cf
FB		 787 /* Propagates the various manager settings into the manager object, i.e. properties that
788 * effect the manager itself (as opposed to just being inherited into newly allocated
789 * units, see set_manager_defaults() above). */
5b65ae15 790
7b46fc6a 791 m->confirm_spawn = arg_confirm_spawn;
2a12e32e 792 m->service_watchdogs = arg_service_watchdogs;
7b46fc6a 793 m->cad_burst_action = arg_cad_burst_action;
856bfaeb
LB		 794 /* Note that we don't do structured initialization here, otherwise it will reset the rate limit
795 * counter on every daemon-reload. */
796 m->reload_ratelimit.interval = arg_reload_limit_interval_sec;
797 m->reload_ratelimit.burst = arg_reload_limit_burst;
7b46fc6a 798
986935cf
FB		 799 manager_set_watchdog(m, WATCHDOG_RUNTIME, arg_runtime_watchdog);
800 manager_set_watchdog(m, WATCHDOG_REBOOT, arg_reboot_watchdog);
801 manager_set_watchdog(m, WATCHDOG_KEXEC, arg_kexec_watchdog);
5717062e 802 manager_set_watchdog(m, WATCHDOG_PRETIMEOUT, arg_pretimeout_watchdog);
aff3a9e1
LB		 803 r = manager_set_watchdog_pretimeout_governor(m, arg_watchdog_pretimeout_governor);
804 if (r < 0)
805 log_warning_errno(r, "Failed to set watchdog pretimeout governor to '%s', ignoring: %m", arg_watchdog_pretimeout_governor);
986935cf 806
7365a296 807 manager_set_show_status(m, arg_show_status, "commandline");
36cf4507 808 m->status_unit_format = arg_status_unit_format;
7b46fc6a
LP		 809}
810
f170852a 811static int parse_argv(int argc, char *argv[]) {
f170852a
LP		 812 enum {
813 ARG_LOG_LEVEL = 0x100,
814 ARG_LOG_TARGET,
bbe63281
LP		 815 ARG_LOG_COLOR,
816 ARG_LOG_LOCATION,
c5673ed0 817 ARG_LOG_TIME,
2f198e2f 818 ARG_UNIT,
edb9aaa8 819 ARG_SYSTEM,
af2d49f7 820 ARG_USER,
e537352b 821 ARG_TEST,
b87c2aa6 822 ARG_NO_PAGER,
9ba0bc4e 823 ARG_VERSION,
80876c20 824 ARG_DUMP_CONFIGURATION_ITEMS,
bbc1acab 825 ARG_DUMP_BUS_PROPERTIES,
5c08257b 826 ARG_BUS_INTROSPECT,
9e58ff9c 827 ARG_DUMP_CORE,
b9e74c39 828 ARG_CRASH_CHVT,
9e58ff9c 829 ARG_CRASH_SHELL,
b9e74c39 830 ARG_CRASH_REBOOT,
a16e1123 831 ARG_CONFIRM_SPAWN,
9e58ff9c 832 ARG_SHOW_STATUS,
4288f619 833 ARG_DESERIALIZE,
2660882b 834 ARG_SWITCHED_ROOT,
0a494f1f 835 ARG_DEFAULT_STD_OUTPUT,
ee48dbd5 836 ARG_DEFAULT_STD_ERROR,
2a12e32e
JK		 837 ARG_MACHINE_ID,
838 ARG_SERVICE_WATCHDOGS,
f170852a
LP		 839 };
840
841 static const struct option options[] = {
a16e1123
LP		 842 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
843 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
bbe63281
LP		 844 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
845 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
c5673ed0 846 { "log-time", optional_argument, NULL, ARG_LOG_TIME },
2f198e2f 847 { "unit", required_argument, NULL, ARG_UNIT },
edb9aaa8 848 { "system", no_argument, NULL, ARG_SYSTEM },
af2d49f7 849 { "user", no_argument, NULL, ARG_USER },
a16e1123 850 { "test", no_argument, NULL, ARG_TEST },
b87c2aa6 851 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
a16e1123 852 { "help", no_argument, NULL, 'h' },
9ba0bc4e 853 { "version", no_argument, NULL, ARG_VERSION },
a16e1123 854 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
bbc1acab 855 { "dump-bus-properties", no_argument, NULL, ARG_DUMP_BUS_PROPERTIES },
5c08257b 856 { "bus-introspect", required_argument, NULL, ARG_BUS_INTROSPECT },
a5d87bf0 857 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
b9e74c39 858 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
a5d87bf0 859 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
b9e74c39 860 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
a5d87bf0 861 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
6e98720f 862 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
a16e1123 863 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
2660882b 864 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
0a494f1f
LP		 865 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
866 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
ee48dbd5 867 { "machine-id", required_argument, NULL, ARG_MACHINE_ID },
2a12e32e 868 { "service-watchdogs", required_argument, NULL, ARG_SERVICE_WATCHDOGS },
fb472900 869 {}
f170852a
LP		 870 };
871
872 int c, r;
9a9ca408 873 bool user_arg_seen = false;
f170852a
LP		 874
875 assert(argc >= 1);
876 assert(argv);
877
df0ff127 878 if (getpid_cached() == 1)
b770165a
LP		 879 opterr = 0;
880
099663ff 881 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
f170852a
LP		 882
883 switch (c) {
884
885 case ARG_LOG_LEVEL:
fb472900 886 r = log_set_max_level_from_string(optarg);
2b5107e1
ZJS		 887 if (r < 0)
888 return log_error_errno(r, "Failed to parse log level \"%s\": %m", optarg);
f170852a
LP		 889
890 break;
891
892 case ARG_LOG_TARGET:
fb472900 893 r = log_set_target_from_string(optarg);
2b5107e1
ZJS		 894 if (r < 0)
895 return log_error_errno(r, "Failed to parse log target \"%s\": %m", optarg);
f170852a
LP		 896
897 break;
898
bbe63281
LP		 899 case ARG_LOG_COLOR:
900
d0b170c8 901 if (optarg) {
fb472900 902 r = log_show_color_from_string(optarg);
2b5107e1
ZJS		 903 if (r < 0)
904 return log_error_errno(r, "Failed to parse log color setting \"%s\": %m",
905 optarg);
d0b170c8
LP		 906 } else
907 log_show_color(true);
bbe63281
LP		 908
909 break;
910
911 case ARG_LOG_LOCATION:
d0b170c8 912 if (optarg) {
fb472900 913 r = log_show_location_from_string(optarg);
2b5107e1
ZJS		 914 if (r < 0)
915 return log_error_errno(r, "Failed to parse log location setting \"%s\": %m",
916 optarg);
d0b170c8
LP		 917 } else
918 log_show_location(true);
bbe63281
LP		 919
920 break;
921
c5673ed0
DS		 922 case ARG_LOG_TIME:
923
924 if (optarg) {
925 r = log_show_time_from_string(optarg);
926 if (r < 0)
927 return log_error_errno(r, "Failed to parse log time setting \"%s\": %m",
928 optarg);
929 } else
930 log_show_time(true);
931
932 break;
933
0a494f1f 934 case ARG_DEFAULT_STD_OUTPUT:
fb472900 935 r = exec_output_from_string(optarg);
2b5107e1
ZJS		 936 if (r < 0)
937 return log_error_errno(r, "Failed to parse default standard output setting \"%s\": %m",
938 optarg);
939 arg_default_std_output = r;
0a494f1f
LP		 940 break;
941
942 case ARG_DEFAULT_STD_ERROR:
fb472900 943 r = exec_output_from_string(optarg);
2b5107e1
ZJS		 944 if (r < 0)
945 return log_error_errno(r, "Failed to parse default standard error output setting \"%s\": %m",
946 optarg);
947 arg_default_std_error = r;
0a494f1f
LP		 948 break;
949
2f198e2f 950 case ARG_UNIT:
e6e242ad 951 r = free_and_strdup(&arg_default_unit, optarg);
23bbb0de 952 if (r < 0)
2b5107e1 953 return log_error_errno(r, "Failed to set default unit \"%s\": %m", optarg);
f170852a
LP		 954
955 break;
956
edb9aaa8 957 case ARG_SYSTEM:
463d0d15 958 arg_system = true;
edb9aaa8 959 break;
a5dab5ce 960
af2d49f7 961 case ARG_USER:
463d0d15 962 arg_system = false;
9a9ca408 963 user_arg_seen = true;
a5dab5ce 964 break;
a5dab5ce 965
e965d56d 966 case ARG_TEST:
fa0f4d8a 967 arg_action = ACTION_TEST;
b87c2aa6
ZJS		 968 break;
969
970 case ARG_NO_PAGER:
0221d68a 971 arg_pager_flags |= PAGER_DISABLE;
e965d56d
LP		 972 break;
973
9ba0bc4e
ZJS		 974 case ARG_VERSION:
975 arg_action = ACTION_VERSION;
976 break;
977
e537352b 978 case ARG_DUMP_CONFIGURATION_ITEMS:
fa0f4d8a 979 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
e537352b
LP		 980 break;
981
bbc1acab
YW		 982 case ARG_DUMP_BUS_PROPERTIES:
983 arg_action = ACTION_DUMP_BUS_PROPERTIES;
984 break;
985
5c08257b
ZJS		 986 case ARG_BUS_INTROSPECT:
987 arg_bus_introspect = optarg;
988 arg_action = ACTION_BUS_INTROSPECT;
989 break;
990
9e58ff9c 991 case ARG_DUMP_CORE:
599c7c54
ZJS		 992 r = parse_boolean_argument("--dump-core", optarg, &arg_dump_core);
993 if (r < 0)
994 return r;
b9e74c39
LP		 995 break;
996
997 case ARG_CRASH_CHVT:
a07a7324 998 r = parse_crash_chvt(optarg, &arg_crash_chvt);
b9e74c39 999 if (r < 0)
2b5107e1
ZJS		 1000 return log_error_errno(r, "Failed to parse crash virtual terminal index: \"%s\": %m",
1001 optarg);
9e58ff9c
LP		 1002 break;
1003
1004 case ARG_CRASH_SHELL:
599c7c54
ZJS		 1005 r = parse_boolean_argument("--crash-shell", optarg, &arg_crash_shell);
1006 if (r < 0)
1007 return r;
b9e74c39
LP		 1008 break;
1009
1010 case ARG_CRASH_REBOOT:
599c7c54
ZJS		 1011 r = parse_boolean_argument("--crash-reboot", optarg, &arg_crash_reboot);
1012 if (r < 0)
1013 return r;
9e58ff9c
LP		 1014 break;
1015
80876c20 1016 case ARG_CONFIRM_SPAWN:
7d5ceb64
FB		 1017 arg_confirm_spawn = mfree(arg_confirm_spawn);
1018
1019 r = parse_confirm_spawn(optarg, &arg_confirm_spawn);
1020 if (r < 0)
2b5107e1
ZJS		 1021 return log_error_errno(r, "Failed to parse confirm spawn option: \"%s\": %m",
1022 optarg);
80876c20
LP		 1023 break;
1024
2a12e32e 1025 case ARG_SERVICE_WATCHDOGS:
599c7c54 1026 r = parse_boolean_argument("--service-watchdogs=", optarg, &arg_service_watchdogs);
2a12e32e 1027 if (r < 0)
599c7c54 1028 return r;
2a12e32e
JK		 1029 break;
1030
9e58ff9c 1031 case ARG_SHOW_STATUS:
d450b6f2
ZJS		 1032 if (optarg) {
1033 r = parse_show_status(optarg, &arg_show_status);
ac7ec288 1034 if (r < 0)
2b5107e1
ZJS		 1035 return log_error_errno(r, "Failed to parse show status boolean: \"%s\": %m",
1036 optarg);
d450b6f2
ZJS		 1037 } else
1038 arg_show_status = SHOW_STATUS_YES;
6e98720f 1039 break;
a5d87bf0 1040
a16e1123
LP		 1041 case ARG_DESERIALIZE: {
1042 int fd;
1043 FILE *f;
1044
01e10de3 1045 r = safe_atoi(optarg, &fd);
2b5107e1
ZJS		 1046 if (r < 0)
1047 log_error_errno(r, "Failed to parse deserialize option \"%s\": %m", optarg);
baaa35ad
ZJS		 1048 if (fd < 0)
1049 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1050 "Invalid deserialize fd: %d",
1051 fd);
a16e1123 1052
b9e74c39 1053 (void) fd_cloexec(fd, true);
01e10de3
LP		 1054
1055 f = fdopen(fd, "r");
4a62c710 1056 if (!f)
2b5107e1 1057 return log_error_errno(errno, "Failed to open serialization fd %d: %m", fd);
a16e1123 1058
74ca738f 1059 safe_fclose(arg_serialization);
d3b1c508 1060 arg_serialization = f;
a16e1123
LP		 1061
1062 break;
1063 }
1064
2660882b 1065 case ARG_SWITCHED_ROOT:
bf4df7c3 1066 arg_switched_root = true;
d03bc1b8
HH		 1067 break;
1068
ee48dbd5
NC		 1069 case ARG_MACHINE_ID:
1070 r = set_machine_id(optarg);
54500613 1071 if (r < 0)
2b5107e1 1072 return log_error_errno(r, "MachineID '%s' is not valid: %m", optarg);
ee48dbd5
NC		 1073 break;
1074
f170852a 1075 case 'h':
fa0f4d8a 1076 arg_action = ACTION_HELP;
f170852a
LP		 1077 break;
1078
1d2e23ab
LP		 1079 case 'D':
1080 log_set_max_level(LOG_DEBUG);
1081 break;
1082
099663ff
LP		 1083 case 'b':
1084 case 's':
1085 case 'z':
cd57038a
ZJS		 1086 /* Just to eat away the sysvinit kernel cmdline args that we'll parse in
1087 * parse_proc_cmdline_item() or ignore, without any getopt() error messages.
1088 */
099663ff 1089 case '?':
df0ff127 1090 if (getpid_cached() != 1)
099663ff 1091 return -EINVAL;
601185b4
ZJS		 1092 else
1093 return 0;
099663ff 1094
601185b4 1095 default:
04499a70 1096 assert_not_reached();
f170852a
LP		 1097 }
1098
d7a0f1f4 1099 if (optind < argc && getpid_cached() != 1)
9a9ca408
ZJS		 1100 /* Hmm, when we aren't run as init system let's complain about excess arguments */
1101 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Excess arguments.");
1102
1103 if (arg_action == ACTION_RUN && !arg_system && !user_arg_seen)
baaa35ad 1104 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
9a9ca408 1105 "Explicit --user argument required to run as user manager.");
d821e6d6 1106
f170852a
LP		 1107 return 0;
1108}
1109
1110static int help(void) {
37ec0fdd
LP		 1111 _cleanup_free_ char *link = NULL;
1112 int r;
1113
1114 r = terminal_urlify_man("systemd", "1", &link);
1115 if (r < 0)
1116 return log_oom();
f170852a 1117
2e33c433 1118 printf("%s [OPTIONS...]\n\n"
7ae47326
ZJS		 1119 "%sStarts and monitors system and user services.%s\n\n"
1120 "This program takes no positional arguments.\n\n"
1121 "%sOptions%s:\n"
e537352b 1122 " -h --help Show this help\n"
cb4069d9 1123 " --version Show version\n"
cd69e88b 1124 " --test Determine initial transaction, dump it and exit\n"
03b3b55e
ZJS		 1125 " --system Combined with --test: operate in system mode\n"
1126 " --user Combined with --test: operate in user mode\n"
80876c20 1127 " --dump-configuration-items Dump understood unit configuration items\n"
bbc1acab 1128 " --dump-bus-properties Dump exposed bus properties\n"
5c08257b 1129 " --bus-introspect=PATH Write XML introspection data\n"
9e58ff9c 1130 " --unit=UNIT Set default unit\n"
b9e74c39
LP		 1131 " --dump-core[=BOOL] Dump core on crash\n"
1132 " --crash-vt=NR Change to specified VT on crash\n"
1133 " --crash-reboot[=BOOL] Reboot on crash\n"
1134 " --crash-shell[=BOOL] Run shell on crash\n"
1135 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
03b3b55e
ZJS		 1136 " --show-status[=BOOL] Show status updates on the console during boot\n"
1137 " --log-target=TARGET Set log target (console, journal, kmsg,\n"
1138 " journal-or-kmsg, null)\n"
1139 " --log-level=LEVEL Set log level (debug, info, notice, warning,\n"
1140 " err, crit, alert, emerg)\n"
b9e74c39
LP		 1141 " --log-color[=BOOL] Highlight important log messages\n"
1142 " --log-location[=BOOL] Include code location in log messages\n"
c5673ed0 1143 " --log-time[=BOOL] Prefix log messages with current time\n"
0a494f1f 1144 " --default-standard-output= Set default standard output for services\n"
37ec0fdd 1145 " --default-standard-error= Set default standard error output for services\n"
03b3b55e 1146 " --no-pager Do not pipe output into a pager\n"
bc556335
DDM		 1147 "\nSee the %s for details.\n",
1148 program_invocation_short_name,
1149 ansi_highlight(),
1150 ansi_normal(),
1151 ansi_underline(),
1152 ansi_normal(),
1153 link);
f170852a
LP		 1154
1155 return 0;
1156}
1157
2cc856ac
LP		 1158static int prepare_reexecute(
1159 Manager *m,
1160 FILE **ret_f,
1161 FDSet **ret_fds,
1162 bool switching_root) {
1163
48b90859
LP		 1164 _cleanup_fdset_free_ FDSet *fds = NULL;
1165 _cleanup_fclose_ FILE *f = NULL;
a16e1123
LP		 1166 int r;
1167
1168 assert(m);
2cc856ac
LP		 1169 assert(ret_f);
1170 assert(ret_fds);
a16e1123 1171
6b78f9b4 1172 r = manager_open_serialization(m, &f);
48b90859
LP		 1173 if (r < 0)
1174 return log_error_errno(r, "Failed to create serialization file: %m");
a16e1123 1175
71445ae7 1176 /* Make sure nothing is really destructed when we shut down */
313cefa1 1177 m->n_reloading++;
718db961 1178 bus_manager_send_reloading(m, true);
71445ae7 1179
6b78f9b4 1180 fds = fdset_new();
48b90859
LP		 1181 if (!fds)
1182 return log_oom();
a16e1123 1183
b3680f49 1184 r = manager_serialize(m, f, fds, switching_root);
48b90859 1185 if (r < 0)
d68c645b 1186 return r;
a16e1123 1187
48b90859
LP		 1188 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1189 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
a16e1123 1190
6b78f9b4 1191 r = fd_cloexec(fileno(f), false);
48b90859
LP		 1192 if (r < 0)
1193 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
a16e1123 1194
6b78f9b4 1195 r = fdset_cloexec(fds, false);
48b90859
LP		 1196 if (r < 0)
1197 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
a16e1123 1198
2cc856ac
LP		 1199 *ret_f = TAKE_PTR(f);
1200 *ret_fds = TAKE_PTR(fds);
a16e1123 1201
48b90859 1202 return 0;
a16e1123
LP		 1203}
1204
a8b627aa
LP		 1205static void bump_file_max_and_nr_open(void) {
1206
ff3a7019
ZJS		 1207 /* Let's bump fs.file-max and fs.nr_open to their respective maximums. On current kernels large
1208 * numbers of file descriptors are no longer a performance problem and their memory is properly
1209 * tracked by memcg, thus counting them and limiting them in another two layers of limits is
1210 * unnecessary and just complicates things. This function hence turns off 2 of the 4 levels of limits
1211 * on file descriptors, and makes RLIMIT_NOLIMIT (soft + hard) the only ones that really matter. */
a8b627aa
LP		 1212
1213#if BUMP_PROC_SYS_FS_FILE_MAX || BUMP_PROC_SYS_FS_NR_OPEN
a8b627aa
LP		 1214 int r;
1215#endif
1216
1217#if BUMP_PROC_SYS_FS_FILE_MAX
409607c1
ZJS		 1218 /* The maximum the kernel allows for this since 5.2 is LONG_MAX, use that. (Previously things were
1219 * different, but the operation would fail silently.) */
a5fac1df 1220 r = sysctl_write("fs/file-max", LONG_MAX_STR);
a8b627aa 1221 if (r < 0)
b47e0fac
ZJS		 1222 log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING,
1223 r, "Failed to bump fs.file-max, ignoring: %m");
a8b627aa
LP		 1224#endif
1225
a8b627aa
LP		 1226#if BUMP_PROC_SYS_FS_NR_OPEN
1227 int v = INT_MAX;
1228
ff3a7019
ZJS		 1229 /* Argh! The kernel enforces maximum and minimum values on the fs.nr_open, but we don't really know
1230 * what they are. The expression by which the maximum is determined is dependent on the architecture,
1231 * and is something we don't really want to copy to userspace, as it is dependent on implementation
1232 * details of the kernel. Since the kernel doesn't expose the maximum value to us, we can only try
1233 * and hope. Hence, let's start with INT_MAX, and then keep halving the value until we find one that
1234 * works. Ugly? Yes, absolutely, but kernel APIs are kernel APIs, so what do can we do... 🤯 */
a8b627aa
LP		 1235
1236 for (;;) {
1237 int k;
1238
1239 v &= ~(__SIZEOF_POINTER__ - 1); /* Round down to next multiple of the pointer size */
1240 if (v < 1024) {
1241 log_warning("Can't bump fs.nr_open, value too small.");
1242 break;
1243 }
1244
1245 k = read_nr_open();
1246 if (k < 0) {
1247 log_error_errno(k, "Failed to read fs.nr_open: %m");
1248 break;
1249 }
1250 if (k >= v) { /* Already larger */
1251 log_debug("Skipping bump, value is already larger.");
1252 break;
1253 }
1254
b47e0fac 1255 r = sysctl_writef("fs/nr_open", "%i", v);
a8b627aa
LP		 1256 if (r == -EINVAL) {
1257 log_debug("Couldn't write fs.nr_open as %i, halving it.", v);
1258 v /= 2;
1259 continue;
1260 }
1261 if (r < 0) {
1262 log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.nr_open, ignoring: %m");
1263 break;
1264 }
1265
1266 log_debug("Successfully bumped fs.nr_open to %i", v);
1267 break;
1268 }
1269#endif
1270}
1271
eadb4f19 1272static int bump_rlimit_nofile(const struct rlimit *saved_rlimit) {
cda7faa9 1273 struct rlimit new_rlimit;
9264cc39 1274 int r, nr;
4096d6f5 1275
52d62075
LP		 1276 /* Get the underlying absolute limit the kernel enforces */
1277 nr = read_nr_open();
1278
cda7faa9
LP		 1279 /* Calculate the new limits to use for us. Never lower from what we inherited. */
1280 new_rlimit = (struct rlimit) {
1281 .rlim_cur = MAX((rlim_t) nr, saved_rlimit->rlim_cur),
1282 .rlim_max = MAX((rlim_t) nr, saved_rlimit->rlim_max),
1283 };
1284
1285 /* Shortcut if nothing changes. */
1286 if (saved_rlimit->rlim_max >= new_rlimit.rlim_max &&
1287 saved_rlimit->rlim_cur >= new_rlimit.rlim_cur) {
1288 log_debug("RLIMIT_NOFILE is already as high or higher than we need it, not bumping.");
1289 return 0;
1290 }
1291
52d62075
LP		 1292 /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for
1293 * both hard and soft. */
cda7faa9 1294 r = setrlimit_closest(RLIMIT_NOFILE, &new_rlimit);
23bbb0de 1295 if (r < 0)
3ce40911 1296 return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m");
4096d6f5
LP		 1297
1298 return 0;
1299}
1300
eadb4f19 1301static int bump_rlimit_memlock(const struct rlimit *saved_rlimit) {
cda7faa9 1302 struct rlimit new_rlimit;
04d1ee0f 1303 uint64_t mm;
fb3ae275
LP		 1304 int r;
1305
ff3a7019
ZJS		 1306 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK
1307 * which should normally disable such checks. We need them to implement IPAddressAllow= and
1308 * IPAddressDeny=, hence let's bump the value high enough for our user. */
fb3ae275 1309
cda7faa9
LP		 1310 /* Using MAX() on resource limits only is safe if RLIM_INFINITY is > 0. POSIX declares that rlim_t
1311 * must be unsigned, hence this is a given, but let's make this clear here. */
1312 assert_cc(RLIM_INFINITY > 0);
1313
ff3a7019
ZJS		 1314 mm = physical_memory_scale(1, 8); /* Let's scale how much we allow to be locked by the amount of
1315 * physical RAM. We allow an eighth to be locked by us, just to
1316 * pick a value. */
04d1ee0f 1317
cda7faa9 1318 new_rlimit = (struct rlimit) {
04d1ee0f
LP		 1319 .rlim_cur = MAX3(HIGH_RLIMIT_MEMLOCK, saved_rlimit->rlim_cur, mm),
1320 .rlim_max = MAX3(HIGH_RLIMIT_MEMLOCK, saved_rlimit->rlim_max, mm),
cda7faa9
LP		 1321 };
1322
1323 if (saved_rlimit->rlim_max >= new_rlimit.rlim_cur &&
1324 saved_rlimit->rlim_cur >= new_rlimit.rlim_max) {
1325 log_debug("RLIMIT_MEMLOCK is already as high or higher than we need it, not bumping.");
1326 return 0;
1327 }
1328
1329 r = setrlimit_closest(RLIMIT_MEMLOCK, &new_rlimit);
fb3ae275
LP		 1330 if (r < 0)
1331 return log_warning_errno(r, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1332
1333 return 0;
1334}
1335
80758717 1336static void test_usr(void) {
80758717 1337
796ac4c1 1338 /* Check that /usr is either on the same file system as / or mounted already. */
80758717 1339
db55bbf2 1340 if (dir_is_empty("/usr", /* ignore_hidden_or_backup= */ false) <= 0)
871c44a7
LP		 1341 return;
1342
8b173b5e 1343 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
871c44a7 1344 "Some things will probably break (sometimes even silently) in mysterious ways. "
a25d9395 1345 "Consult https://www.freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
871c44a7
LP		 1346}
1347
d3b1c508 1348static int enforce_syscall_archs(Set *archs) {
349cc4a5 1349#if HAVE_SECCOMP
d3b1c508
LP		 1350 int r;
1351
83f12b27
FS		 1352 if (!is_seccomp_available())
1353 return 0;
1354
469830d1 1355 r = seccomp_restrict_archs(arg_syscall_archs);
d3b1c508 1356 if (r < 0)
469830d1 1357 return log_error_errno(r, "Failed to enforce system call architecture restrication: %m");
d3b1c508 1358#endif
469830d1 1359 return 0;
d3b1c508
LP		 1360}
1361
4bd03515
ZJS		 1362static int os_release_status(void) {
1363 _cleanup_free_ char *pretty_name = NULL, *name = NULL, *version = NULL,
1364 *ansi_color = NULL, *support_end = NULL;
b6e2f329
LP		 1365 int r;
1366
d58ad743
LP		 1367 r = parse_os_release(NULL,
1368 "PRETTY_NAME", &pretty_name,
4bd03515
ZJS		 1369 "NAME", &name,
1370 "VERSION", &version,
1371 "ANSI_COLOR", &ansi_color,
1372 "SUPPORT_END", &support_end);
d58ad743 1373 if (r < 0)
4bd03515
ZJS		 1374 return log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r,
1375 "Failed to read os-release file, ignoring: %m");
b6e2f329 1376
02b7005e 1377 const char *label = os_release_pretty_name(pretty_name, name);
4bd03515
ZJS		 1378
1379 if (show_status_on(arg_show_status)) {
1380 if (log_get_show_color())
1381 status_printf(NULL, 0,
1382 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1383 empty_to_null(ansi_color) ?: "1",
1384 label);
1385 else
1386 status_printf(NULL, 0,
1387 "\nWelcome to %s!\n",
1388 label);
1389 }
1390
6bfe9b3b 1391 if (support_end && os_release_support_ended(support_end, /* quiet */ false, NULL) > 0)
4bd03515
ZJS		 1392 /* pretty_name may include the version already, so we'll print the version only if we
1393 * have it and we're not using pretty_name. */
1394 status_printf(ANSI_HIGHLIGHT_RED " !! " ANSI_NORMAL, 0,
1395 "This OS version (%s%s%s) is past its end-of-support date (%s)",
1396 label,
1397 (pretty_name || !version) ? "" : " version ",
1398 (pretty_name || !version) ? "" : version,
1399 support_end);
1400
1401 return 0;
b6e2f329
LP		 1402}
1403
fdd25311
LP		 1404static int write_container_id(void) {
1405 const char *c;
7756528e 1406 int r = 0; /* avoid false maybe-uninitialized warning */
fdd25311
LP		 1407
1408 c = getenv("container");
1409 if (isempty(c))
1410 return 0;
1411
2053593f 1412 WITH_UMASK(0022)
8612da97 1413 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
19854865 1414 if (r < 0)
f1f849b0 1415 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
19854865
LP		 1416
1417 return 1;
1418}
1419
1420static int bump_unix_max_dgram_qlen(void) {
1421 _cleanup_free_ char *qlen = NULL;
1422 unsigned long v;
1423 int r;
1424
ff3a7019
ZJS		 1425 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set
1426 * the value really really early during boot, so that it is actually applied to all our sockets,
1427 * including the $NOTIFY_SOCKET one. */
19854865
LP		 1428
1429 r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1430 if (r < 0)
ff3a7019
ZJS		 1431 return log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r,
1432 "Failed to read AF_UNIX datagram queue length, ignoring: %m");
19854865
LP		 1433
1434 r = safe_atolu(qlen, &v);
1435 if (r < 0)
3130fca5 1436 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen);
19854865
LP		 1437
1438 if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1439 return 0;
1440
27c8ca43 1441 r = sysctl_write("net/unix/max_dgram_qlen", STRINGIFY(DEFAULT_UNIX_MAX_DGRAM_QLEN));
19854865
LP		 1442 if (r < 0)
1443 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1444 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1445
1446 return 1;
fdd25311
LP		 1447}
1448
32391275
FB		 1449static int fixup_environment(void) {
1450 _cleanup_free_ char *term = NULL;
4dc63c4b 1451 const char *t;
32391275
FB		 1452 int r;
1453
43db615b
LP		 1454 /* Only fix up the environment when we are started as PID 1 */
1455 if (getpid_cached() != 1)
1456 return 0;
1457
1458 /* We expect the environment to be set correctly if run inside a container. */
84af7821
LP		 1459 if (detect_container() > 0)
1460 return 0;
1461
ff3a7019
ZJS		 1462 /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the
1463 * backend device used by the console. We try to make a better guess here since some consoles might
1464 * not have support for color mode for example.
32391275 1465 *
43db615b 1466 * However if TERM was configured through the kernel command line then leave it alone. */
1d84ad94 1467 r = proc_cmdline_get_key("TERM", 0, &term);
32391275
FB		 1468 if (r < 0)
1469 return r;
32391275 1470
4dc63c4b
LP		 1471 t = term ?: default_term_for_tty("/dev/console");
1472
1473 if (setenv("TERM", t, 1) < 0)
32391275
FB		 1474 return -errno;
1475
9d48671c 1476 /* The kernels sets HOME=/ for init. Let's undo this. */
44ee03d1
ZJS		 1477 if (path_equal_ptr(getenv("HOME"), "/"))
1478 assert_se(unsetenv("HOME") == 0);
9d48671c 1479
32391275
FB		 1480 return 0;
1481}
1482
6808a0bc
LP		 1483static void redirect_telinit(int argc, char *argv[]) {
1484
1485 /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1486
1487#if HAVE_SYSV_COMPAT
1488 if (getpid_cached() == 1)
1489 return;
1490
2306d177 1491 if (!invoked_as(argv, "init"))
6808a0bc
LP		 1492 return;
1493
1494 execv(SYSTEMCTL_BINARY_PATH, argv);
1495 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
a45d7127 1496 exit(EXIT_FAILURE);
6808a0bc
LP		 1497#endif
1498}
1499
b22d392d
VC		 1500static int become_shutdown(int objective, int retval) {
1501
1502 static const char* const table[_MANAGER_OBJECTIVE_MAX] = {
1503 [MANAGER_EXIT] = "exit",
1504 [MANAGER_REBOOT] = "reboot",
1505 [MANAGER_POWEROFF] = "poweroff",
1506 [MANAGER_HALT] = "halt",
1507 [MANAGER_KEXEC] = "kexec",
1508 };
4a36297c
LP		 1509
1510 char log_level[DECIMAL_STR_MAX(int) + 1],
e73c54b8
JK		 1511 exit_code[DECIMAL_STR_MAX(uint8_t) + 1],
1512 timeout[DECIMAL_STR_MAX(usec_t) + 1];
4a36297c 1513
e73c54b8 1514 const char* command_line[13] = {
4a36297c 1515 SYSTEMD_SHUTDOWN_BINARY_PATH,
b22d392d 1516 table[objective],
e73c54b8 1517 "--timeout", timeout,
4a36297c
LP		 1518 "--log-level", log_level,
1519 "--log-target",
1520 };
1521
1522 _cleanup_strv_free_ char **env_block = NULL;
f16890f8 1523 usec_t watchdog_timer = 0;
e73c54b8 1524 size_t pos = 7;
4a36297c
LP		 1525 int r;
1526
b22d392d
VC		 1527 assert(objective >= 0 && objective < _MANAGER_OBJECTIVE_MAX);
1528 assert(table[objective]);
234519ae 1529 assert(!command_line[pos]);
4a36297c
LP		 1530 env_block = strv_copy(environ);
1531
1532 xsprintf(log_level, "%d", log_get_max_level());
e73c54b8 1533 xsprintf(timeout, "%" PRI_USEC "us", arg_default_timeout_stop_usec);
4a36297c
LP		 1534
1535 switch (log_get_target()) {
1536
1537 case LOG_TARGET_KMSG:
1538 case LOG_TARGET_JOURNAL_OR_KMSG:
1539 case LOG_TARGET_SYSLOG_OR_KMSG:
1540 command_line[pos++] = "kmsg";
1541 break;
1542
1543 case LOG_TARGET_NULL:
1544 command_line[pos++] = "null";
1545 break;
1546
1547 case LOG_TARGET_CONSOLE:
1548 default:
1549 command_line[pos++] = "console";
1550 break;
1551 };
1552
1553 if (log_get_show_color())
1554 command_line[pos++] = "--log-color";
1555
1556 if (log_get_show_location())
1557 command_line[pos++] = "--log-location";
1558
c5673ed0
DS		 1559 if (log_get_show_time())
1560 command_line[pos++] = "--log-time";
1561
b22d392d 1562 if (objective == MANAGER_EXIT) {
4a36297c
LP		 1563 command_line[pos++] = "--exit-code";
1564 command_line[pos++] = exit_code;
1565 xsprintf(exit_code, "%d", retval);
1566 }
1567
1568 assert(pos < ELEMENTSOF(command_line));
1569
b22d392d 1570 if (objective == MANAGER_REBOOT)
65224c1d 1571 watchdog_timer = arg_reboot_watchdog;
b22d392d 1572 else if (objective == MANAGER_KEXEC)
acafd7d8
LB		 1573 watchdog_timer = arg_kexec_watchdog;
1574
f16890f8 1575 /* If we reboot or kexec let's set the shutdown watchdog and tell the
5717062e
CK		 1576 * shutdown binary to repeatedly ping it.
1577 * Disable the pretimeout watchdog, as we do not support it from the shutdown binary. */
1578 (void) watchdog_setup_pretimeout(0);
aff3a9e1 1579 (void) watchdog_setup_pretimeout_governor(NULL);
f16890f8
FB		 1580 r = watchdog_setup(watchdog_timer);
1581 watchdog_close(r < 0);
4a36297c 1582
f16890f8
FB		 1583 /* Tell the binary how often to ping, ignore failure */
1584 (void) strv_extendf(&env_block, "WATCHDOG_USEC="USEC_FMT, watchdog_timer);
8a2c1fbf 1585
f16890f8
FB		 1586 if (arg_watchdog_device)
1587 (void) strv_extendf(&env_block, "WATCHDOG_DEVICE=%s", arg_watchdog_device);
4a36297c 1588
3178d23d
FB		 1589 /* Avoid the creation of new processes forked by the kernel; at this
1590 * point, we will not listen to the signals anyway */
4a36297c
LP		 1591 if (detect_container() <= 0)
1592 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
1593
1594 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
1595 return -errno;
1596}
1597
e839bafd
LP		 1598static void initialize_clock(void) {
1599 int r;
1600
3753325b
LP		 1601 /* This is called very early on, before we parse the kernel command line or otherwise figure out why
1602 * we are running, but only once. */
1603
e839bafd
LP		 1604 if (clock_is_localtime(NULL) > 0) {
1605 int min;
1606
ff3a7019 1607 /* The very first call of settimeofday() also does a time warp in the kernel.
e839bafd 1608 *
ff3a7019
ZJS		 1609 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to
1610 * take care of maintaining the RTC and do all adjustments. This matches the behavior of
1611 * Windows, which leaves the RTC alone if the registry tells that the RTC runs in UTC.
e839bafd
LP		 1612 */
1613 r = clock_set_timezone(&min);
1614 if (r < 0)
1615 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1616 else
1617 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1618
d46b79bb 1619 } else if (!in_initrd())
e839bafd
LP		 1620 /*
1621 * Do a dummy very first call to seal the kernel's time warp magic.
1622 *
ff3a7019
ZJS		 1623 * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with
1624 * LOCAL, but the real system could be set up that way. In such case, we need to delay the
1625 * time-warp or the sealing until we reach the real system.
e839bafd 1626 *
ff3a7019
ZJS		 1627 * Do no set the kernel's timezone. The concept of local time cannot be supported reliably,
1628 * the time will jump or be incorrect at every daylight saving time change. All kernel local
1629 * time concepts will be treated as UTC that way.
e839bafd
LP		 1630 */
1631 (void) clock_reset_timewarp();
e839bafd 1632
b10abe4b
EI		 1633 ClockChangeDirection change_dir;
1634 r = clock_apply_epoch(&change_dir);
1635 if (r > 0 && change_dir == CLOCK_CHANGE_FORWARD)
e839bafd 1636 log_info("System time before build time, advancing clock.");
b10abe4b
EI		 1637 else if (r > 0 && change_dir == CLOCK_CHANGE_BACKWARD)
1638 log_info("System time is further ahead than %s after build time, resetting clock to build time.",
1639 FORMAT_TIMESPAN(CLOCK_VALID_RANGE_USEC_MAX, USEC_PER_DAY));
1640 else if (r < 0 && change_dir == CLOCK_CHANGE_FORWARD)
1641 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1642 else if (r < 0 && change_dir == CLOCK_CHANGE_BACKWARD)
1643 log_error_errno(r, "Current system time is further ahead %s after build time, but cannot correct: %m",
1644 FORMAT_TIMESPAN(CLOCK_VALID_RANGE_USEC_MAX, USEC_PER_DAY));
e839bafd
LP		 1645}
1646
3753325b 1647static void apply_clock_update(void) {
3753325b
LP		 1648 /* This is called later than initialize_clock(), i.e. after we parsed configuration files/kernel
1649 * command line and such. */
1650
1651 if (arg_clock_usec == 0)
1652 return;
1653
45250e66
LP		 1654 if (getpid_cached() != 1)
1655 return;
1656
52bb308c 1657 if (clock_settime(CLOCK_REALTIME, TIMESPEC_STORE(arg_clock_usec)) < 0)
3753325b 1658 log_error_errno(errno, "Failed to set system clock to time specified on kernel command line: %m");
04f5c018 1659 else
3753325b 1660 log_info("Set system clock to %s, as specified on the kernel command line.",
04f5c018 1661 FORMAT_TIMESTAMP(arg_clock_usec));
3753325b
LP		 1662}
1663
d247f232 1664static void cmdline_take_random_seed(void) {
d247f232
LP		 1665 size_t suggested;
1666 int r;
1667
1668 if (arg_random_seed_size == 0)
1669 return;
1670
1671 if (getpid_cached() != 1)
1672 return;
1673
1674 assert(arg_random_seed);
1675 suggested = random_pool_size();
1676
1677 if (arg_random_seed_size < suggested)
1678 log_warning("Random seed specified on kernel command line has size %zu, but %zu bytes required to fill entropy pool.",
1679 arg_random_seed_size, suggested);
1680
61bd7d1e 1681 r = random_write_entropy(-1, arg_random_seed, arg_random_seed_size, true);
d247f232
LP		 1682 if (r < 0) {
1683 log_warning_errno(r, "Failed to credit entropy specified on kernel command line, ignoring: %m");
1684 return;
1685 }
1686
1687 log_notice("Successfully credited entropy passed on kernel command line.\n"
ff3a7019
ZJS		 1688 "Note that the seed provided this way is accessible to unprivileged programs. "
1689 "This functionality should not be used outside of testing environments.");
d247f232
LP		 1690}
1691
1e41242e 1692static void initialize_coredump(bool skip_setup) {
752bcb77 1693#if ENABLE_COREDUMP
1e41242e
LP		 1694 if (getpid_cached() != 1)
1695 return;
1696
ff3a7019
ZJS		 1697 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour
1698 * the limit) will process core dumps for system services by default. */
1e41242e
LP		 1699 if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0)
1700 log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m");
1701
ff3a7019
ZJS		 1702 /* But at the same time, turn off the core_pattern logic by default, so that no coredumps are stored
1703 * until the systemd-coredump tool is enabled via sysctl. However it can be changed via the kernel
32e27670 1704 * command line later so core dumps can still be generated during early startup and in initrd. */
1e41242e 1705 if (!skip_setup)
e557b1a6 1706 disable_coredumps();
752bcb77 1707#endif
1e41242e
LP		 1708}
1709
c6885f5f
FB		 1710static void initialize_core_pattern(bool skip_setup) {
1711 int r;
1712
1713 if (skip_setup || !arg_early_core_pattern)
1714 return;
1715
1716 if (getpid_cached() != 1)
1717 return;
1718
57512c89 1719 r = write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern, WRITE_STRING_FILE_DISABLE_BUFFER);
c6885f5f 1720 if (r < 0)
ff3a7019
ZJS		 1721 log_warning_errno(r, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m",
1722 arg_early_core_pattern);
c6885f5f
FB		 1723}
1724
61fbbac1
ZJS		 1725static void update_cpu_affinity(bool skip_setup) {
1726 _cleanup_free_ char *mask = NULL;
1727
1728 if (skip_setup || !arg_cpu_affinity.set)
1729 return;
1730
1731 assert(arg_cpu_affinity.allocated > 0);
1732
667030bf
ZJS		 1733 mask = cpu_set_to_range_string(&arg_cpu_affinity);
1734 log_debug("Setting CPU affinity to {%s}.", strnull(mask));
61fbbac1
ZJS		 1735
1736 if (sched_setaffinity(0, arg_cpu_affinity.allocated, arg_cpu_affinity.set) < 0)
6b1fa539 1737 log_warning_errno(errno, "Failed to set CPU affinity, ignoring: %m");
61fbbac1
ZJS		 1738}
1739
b070c7c0
MS		 1740static void update_numa_policy(bool skip_setup) {
1741 int r;
1742 _cleanup_free_ char *nodes = NULL;
1743 const char * policy = NULL;
1744
1745 if (skip_setup || !mpol_is_valid(numa_policy_get_type(&arg_numa_policy)))
1746 return;
1747
1748 if (DEBUG_LOGGING) {
1749 policy = mpol_to_string(numa_policy_get_type(&arg_numa_policy));
1750 nodes = cpu_set_to_range_string(&arg_numa_policy.nodes);
667030bf 1751 log_debug("Setting NUMA policy to %s, with nodes {%s}.", strnull(policy), strnull(nodes));
b070c7c0
MS		 1752 }
1753
1754 r = apply_numa_policy(&arg_numa_policy);
1755 if (r == -EOPNOTSUPP)
1756 log_debug_errno(r, "NUMA support not available, ignoring.");
1757 else if (r < 0)
6b1fa539 1758 log_warning_errno(r, "Failed to set NUMA memory policy, ignoring: %m");
b070c7c0
MS		 1759}
1760
19fd72df
LP		 1761static void filter_args(
1762 const char* dst[],
1763 size_t *dst_index,
1764 char **src,
1765 int argc) {
1766
846f1da4 1767 assert(dst);
19fd72df 1768 assert(dst_index);
846f1da4
ZJS		 1769
1770 /* Copy some filtered arguments into the dst array from src. */
1771 for (int i = 1; i < argc; i++) {
1772 if (STR_IN_SET(src[i],
1773 "--switched-root",
1774 "--system",
1775 "--user"))
1776 continue;
1777
1778 if (startswith(src[i], "--deserialize="))
1779 continue;
1780 if (streq(src[i], "--deserialize")) {
1781 i++; /* Skip the argument too */
1782 continue;
1783 }
1784
1785 /* Skip target unit designators. We already acted upon this information and have queued
1786 * appropriate jobs. We don't want to redo all this after reexecution. */
1787 if (startswith(src[i], "--unit="))
1788 continue;
1789 if (streq(src[i], "--unit")) {
1790 i++; /* Skip the argument too */
1791 continue;
1792 }
1793
846f1da4 1794 /* Seems we have a good old option. Let's pass it over to the new instance. */
19fd72df 1795 dst[(*dst_index)++] = src[i];
846f1da4
ZJS		 1796 }
1797}
1798
1e3eee8c
ZJS		 1799static int do_reexecute(
1800 ManagerObjective objective,
3c7878f9 1801 int argc,
846f1da4 1802 char* argv[],
3c7878f9
LP		 1803 const struct rlimit *saved_rlimit_nofile,
1804 const struct rlimit *saved_rlimit_memlock,
1805 FDSet *fds,
1806 const char *switch_root_dir,
1807 const char *switch_root_init,
1808 const char **ret_error_message) {
1809
19fd72df 1810 size_t i, args_size;
3c7878f9
LP		 1811 const char **args;
1812 int r;
1813
1e3eee8c 1814 assert(IN_SET(objective, MANAGER_REEXECUTE, MANAGER_SWITCH_ROOT));
19fd72df 1815 assert(argc >= 0);
3c7878f9
LP		 1816 assert(saved_rlimit_nofile);
1817 assert(saved_rlimit_memlock);
1818 assert(ret_error_message);
1819
ff3a7019
ZJS		 1820 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get
1821 * rebooted while we do that */
3c7878f9
LP		 1822 watchdog_close(true);
1823
ddfa8b0b
LP		 1824 /* Reset RLIMIT_NOFILE + RLIMIT_MEMLOCK back to the kernel defaults, so that the new systemd can pass
1825 * the kernel default to its child processes */
1826 if (saved_rlimit_nofile->rlim_cur != 0)
3c7878f9 1827 (void) setrlimit(RLIMIT_NOFILE, saved_rlimit_nofile);
ddfa8b0b 1828 if (saved_rlimit_memlock->rlim_cur != RLIM_INFINITY)
3c7878f9
LP		 1829 (void) setrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock);
1830
1831 if (switch_root_dir) {
ff3a7019
ZJS		 1832 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can
1833 * handle the SIGCHLD for them after deserializing. */
e73c54b8 1834 broadcast_signal(SIGTERM, false, true, arg_default_timeout_stop_usec);
3c7878f9
LP		 1835
1836 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1837 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1838 if (r < 0)
1839 log_error_errno(r, "Failed to switch root, trying to continue: %m");
1840 }
1841
846f1da4 1842 args_size = argc + 6;
3c7878f9
LP		 1843 args = newa(const char*, args_size);
1844
1845 if (!switch_root_init) {
846f1da4 1846 char sfd[DECIMAL_STR_MAX(int)];
3c7878f9 1847
ff3a7019
ZJS		 1848 /* First try to spawn ourselves with the right path, and with full serialization. We do this
1849 * only if the user didn't specify an explicit init to spawn. */
3c7878f9
LP		 1850
1851 assert(arg_serialization);
1852 assert(fds);
1853
1854 xsprintf(sfd, "%i", fileno(arg_serialization));
1855
846f1da4
ZJS		 1856 i = 1; /* Leave args[0] empty for now. */
1857 filter_args(args, &i, argv, argc);
1858
3c7878f9
LP		 1859 if (switch_root_dir)
1860 args[i++] = "--switched-root";
1861 args[i++] = arg_system ? "--system" : "--user";
1862 args[i++] = "--deserialize";
1863 args[i++] = sfd;
1864 args[i++] = NULL;
1865
1866 assert(i <= args_size);
1867
1868 /*
50b35193
ZJS		 1869 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do
1870 * this is on its own on exec(), but it will do it on exit(). Hence, to ensure we get a
ff3a7019
ZJS		 1871 * summary here, fork() off a child, let it exit() cleanly, so that it prints the summary,
1872 * and wait() for it in the parent, before proceeding into the exec().
3c7878f9
LP		 1873 */
1874 valgrind_summary_hack();
1875
846f1da4 1876 args[0] = SYSTEMD_BINARY_PATH;
3c7878f9 1877 (void) execv(args[0], (char* const*) args);
1e3eee8c
ZJS		 1878
1879 if (objective == MANAGER_REEXECUTE) {
1880 *ret_error_message = "Failed to execute our own binary";
1881 return log_error_errno(errno, "Failed to execute our own binary %s: %m", args[0]);
1882 }
1883
846f1da4 1884 log_debug_errno(errno, "Failed to execute our own binary %s, trying fallback: %m", args[0]);
3c7878f9
LP		 1885 }
1886
ff3a7019
ZJS		 1887 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and
1888 * envp[]. (Well, modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[],
1889 * but let's hope that doesn't matter.) */
3c7878f9
LP		 1890
1891 arg_serialization = safe_fclose(arg_serialization);
1892 fds = fdset_free(fds);
1893
1894 /* Reopen the console */
1895 (void) make_console_stdio();
1896
846f1da4
ZJS		 1897 i = 1; /* Leave args[0] empty for now. */
1898 for (int j = 1; j <= argc; j++)
3c7878f9 1899 args[i++] = argv[j];
3c7878f9
LP		 1900 assert(i <= args_size);
1901
55c041b4 1902 /* Re-enable any blocked signals, especially important if we switch from initrd to init=... */
3c7878f9
LP		 1903 (void) reset_all_signal_handlers();
1904 (void) reset_signal_mask();
595225af 1905 (void) rlimit_nofile_safe();
3c7878f9
LP		 1906
1907 if (switch_root_init) {
1908 args[0] = switch_root_init;
a5cede8c 1909 (void) execve(args[0], (char* const*) args, saved_env);
846f1da4 1910 log_warning_errno(errno, "Failed to execute configured init %s, trying fallback: %m", args[0]);
3c7878f9
LP		 1911 }
1912
1913 args[0] = "/sbin/init";
1914 (void) execv(args[0], (char* const*) args);
1915 r = -errno;
1916
1917 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
1918 ANSI_HIGHLIGHT_RED " !! " ANSI_NORMAL,
1919 "Failed to execute /sbin/init");
1920
1e3eee8c 1921 *ret_error_message = "Failed to execute fallback shell";
3c7878f9
LP		 1922 if (r == -ENOENT) {
1923 log_warning("No /sbin/init, trying fallback");
1924
1925 args[0] = "/bin/sh";
1926 args[1] = NULL;
a5cede8c 1927 (void) execve(args[0], (char* const*) args, saved_env);
1e3eee8c 1928 return log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
3c7878f9 1929 } else
6b1fa539 1930 return log_error_errno(r, "Failed to execute /sbin/init, giving up: %m");
3c7878f9
LP		 1931}
1932
7eb35049
LP		 1933static int invoke_main_loop(
1934 Manager *m,
a9fd4cd1
FB		 1935 const struct rlimit *saved_rlimit_nofile,
1936 const struct rlimit *saved_rlimit_memlock,
7eb35049 1937 int *ret_retval, /* Return parameters relevant for shutting down */
7eb35049
LP		 1938 FDSet **ret_fds, /* Return parameters for reexecuting */
1939 char **ret_switch_root_dir, /* … */
1940 char **ret_switch_root_init, /* … */
1941 const char **ret_error_message) {
1942
1943 int r;
1944
1945 assert(m);
a9fd4cd1
FB		 1946 assert(saved_rlimit_nofile);
1947 assert(saved_rlimit_memlock);
7eb35049 1948 assert(ret_retval);
7eb35049
LP		 1949 assert(ret_fds);
1950 assert(ret_switch_root_dir);
1951 assert(ret_switch_root_init);
1952 assert(ret_error_message);
1953
1954 for (;;) {
5409c6fc
ZJS		 1955 int objective = manager_loop(m);
1956 if (objective < 0) {
7eb35049 1957 *ret_error_message = "Failed to run main loop";
5409c6fc 1958 return log_emergency_errno(objective, "Failed to run main loop: %m");
7eb35049
LP		 1959 }
1960
5409c6fc 1961 switch (objective) {
7eb35049 1962
a6ecbf83 1963 case MANAGER_RELOAD: {
bda7d78b 1964 LogTarget saved_log_target;
a6ecbf83
FB		 1965 int saved_log_level;
1966
dd0ab174
LP		 1967 manager_send_reloading(m);
1968
7eb35049
LP		 1969 log_info("Reloading.");
1970
ff3a7019
ZJS		 1971 /* First, save any overridden log level/target, then parse the configuration file,
1972 * which might change the log level to new settings. */
bda7d78b 1973
a6ecbf83 1974 saved_log_level = m->log_level_overridden ? log_get_max_level() : -1;
bda7d78b 1975 saved_log_target = m->log_target_overridden ? log_get_target() : _LOG_TARGET_INVALID;
a6ecbf83 1976
a9fd4cd1 1977 (void) parse_configuration(saved_rlimit_nofile, saved_rlimit_memlock);
7eb35049
LP		 1978
1979 set_manager_defaults(m);
986935cf 1980 set_manager_settings(m);
7eb35049 1981
61fbbac1 1982 update_cpu_affinity(false);
b070c7c0 1983 update_numa_policy(false);
61fbbac1 1984
a6ecbf83
FB		 1985 if (saved_log_level >= 0)
1986 manager_override_log_level(m, saved_log_level);
bda7d78b
FB		 1987 if (saved_log_target >= 0)
1988 manager_override_log_target(m, saved_log_target);
a6ecbf83 1989
5409c6fc 1990 if (manager_reload(m) < 0)
ff3a7019
ZJS		 1991 /* Reloading failed before the point of no return.
1992 * Let's continue running as if nothing happened. */
7a35fa24 1993 m->objective = MANAGER_OK;
7eb35049 1994
5409c6fc 1995 continue;
a6ecbf83 1996 }
7eb35049
LP		 1997
1998 case MANAGER_REEXECUTE:
dd0ab174
LP		 1999
2000 manager_send_reloading(m); /* From the perspective of the manager calling us this is
2001 * pretty much the same as a reload */
2002
7eb35049
LP		 2003 r = prepare_reexecute(m, &arg_serialization, ret_fds, false);
2004 if (r < 0) {
2005 *ret_error_message = "Failed to prepare for reexecution";
2006 return r;
2007 }
2008
2009 log_notice("Reexecuting.");
2010
7eb35049 2011 *ret_retval = EXIT_SUCCESS;
7eb35049
LP		 2012 *ret_switch_root_dir = *ret_switch_root_init = NULL;
2013
5409c6fc 2014 return objective;
7eb35049
LP		 2015
2016 case MANAGER_SWITCH_ROOT:
dd0ab174
LP		 2017
2018 manager_send_reloading(m); /* From the perspective of the manager calling us this is
2019 * pretty much the same as a reload */
2020
d35fe8c0
FB		 2021 manager_set_switching_root(m, true);
2022
7eb35049
LP		 2023 if (!m->switch_root_init) {
2024 r = prepare_reexecute(m, &arg_serialization, ret_fds, true);
2025 if (r < 0) {
2026 *ret_error_message = "Failed to prepare for reexecution";
2027 return r;
2028 }
2029 } else
2030 *ret_fds = NULL;
2031
2032 log_notice("Switching root.");
2033
7eb35049 2034 *ret_retval = EXIT_SUCCESS;
7eb35049
LP		 2035
2036 /* Steal the switch root parameters */
49052946
YW		 2037 *ret_switch_root_dir = TAKE_PTR(m->switch_root);
2038 *ret_switch_root_init = TAKE_PTR(m->switch_root_init);
7eb35049 2039
5409c6fc 2040 return objective;
7eb35049
LP		 2041
2042 case MANAGER_EXIT:
7eb35049
LP		 2043 if (MANAGER_IS_USER(m)) {
2044 log_debug("Exit.");
2045
7eb35049 2046 *ret_retval = m->return_value;
7eb35049
LP		 2047 *ret_fds = NULL;
2048 *ret_switch_root_dir = *ret_switch_root_init = NULL;
2049
5409c6fc 2050 return objective;
7eb35049
LP		 2051 }
2052
2053 _fallthrough_;
2054 case MANAGER_REBOOT:
2055 case MANAGER_POWEROFF:
2056 case MANAGER_HALT:
2057 case MANAGER_KEXEC: {
7eb35049
LP		 2058 log_notice("Shutting down.");
2059
7eb35049 2060 *ret_retval = m->return_value;
7eb35049
LP		 2061 *ret_fds = NULL;
2062 *ret_switch_root_dir = *ret_switch_root_init = NULL;
2063
5409c6fc 2064 return objective;
7eb35049
LP		 2065 }
2066
2067 default:
04499a70 2068 assert_not_reached();
7eb35049
LP		 2069 }
2070 }
2071}
2072
31aef7ff 2073static void log_execution_mode(bool *ret_first_boot) {
7cd43e34
ZJS		 2074 bool first_boot = false;
2075
31aef7ff
LP		 2076 assert(ret_first_boot);
2077
2078 if (arg_system) {
40efaaed 2079 struct utsname uts;
31aef7ff
LP		 2080 int v;
2081
e7b18106 2082 log_info("systemd " GIT_VERSION " running in %ssystem mode (%s)",
91b79ba8
ZJS		 2083 arg_action == ACTION_TEST ? "test " : "",
2084 systemd_features);
31aef7ff
LP		 2085
2086 v = detect_virtualization();
2087 if (v > 0)
2088 log_info("Detected virtualization %s.", virtualization_to_string(v));
2089
2090 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
2091
7cd43e34 2092 if (in_initrd())
55c041b4 2093 log_info("Running in initrd.");
7cd43e34 2094 else {
583cef3b
HS		 2095 int r;
2096 _cleanup_free_ char *id_text = NULL;
2097
7cd43e34
ZJS		 2098 /* Let's check whether we are in first boot. First, check if an override was
2099 * specified on the kernel commandline. If yes, we honour that. */
2100
2101 r = proc_cmdline_get_bool("systemd.condition-first-boot", &first_boot);
2102 if (r < 0)
2103 log_debug_errno(r, "Failed to parse systemd.condition-first-boot= kernel commandline argument, ignoring: %m");
2104
2105 if (r > 0)
2106 log_full(first_boot ? LOG_INFO : LOG_DEBUG,
2107 "Kernel commandline argument says we are %s first boot.",
2108 first_boot ? "in" : "not in");
2109 else {
2110 /* Second, perform autodetection. We use /etc/machine-id as flag file for
2111 * this: If it is missing or contains the value "uninitialized", this is the
2112 * first boot. In other cases, it is not. This allows container managers and
2113 * installers to provision a couple of files in /etc but still permit the
2114 * first-boot initialization to occur. If the container manager wants to
2115 * provision the machine ID it should pass $container_uuid to PID 1. */
2116
2117 r = read_one_line_file("/etc/machine-id", &id_text);
2118 if (r < 0 || streq(id_text, "uninitialized")) {
2119 if (r < 0 && r != -ENOENT)
2120 log_warning_errno(r, "Unexpected error while reading /etc/machine-id, ignoring: %m");
2121
2122 first_boot = true;
2123 log_info("Detected first boot.");
2124 } else
2125 log_debug("Detected initialized system, this is not the first boot.");
583cef3b 2126 }
31aef7ff 2127 }
40efaaed 2128
5180394b 2129 assert_se(uname(&uts) >= 0);
40efaaed
LP		 2130
2131 if (strverscmp_improved(uts.release, KERNEL_BASELINE_VERSION) < 0)
2132 log_warning("Warning! Reported kernel version %s is older than systemd's required baseline kernel version %s. "
2133 "Your mileage may vary.", uts.release, KERNEL_BASELINE_VERSION);
2134 else
2135 log_debug("Kernel version %s, our baseline is %s", uts.release, KERNEL_BASELINE_VERSION);
31aef7ff 2136 } else {
b9e90f3a 2137 if (DEBUG_LOGGING) {
c2b2df60 2138 _cleanup_free_ char *t = NULL;
31aef7ff 2139
b9e90f3a 2140 t = uid_to_name(getuid());
91b79ba8
ZJS		 2141 log_debug("systemd " GIT_VERSION " running in %suser mode for user " UID_FMT "/%s. (%s)",
2142 arg_action == ACTION_TEST ? " test" : "",
2143 getuid(), strna(t), systemd_features);
b9e90f3a 2144 }
31aef7ff 2145 }
7cd43e34
ZJS		 2146
2147 *ret_first_boot = first_boot;
31aef7ff
LP		 2148}
2149
5afbaa36
LP		 2150static int initialize_runtime(
2151 bool skip_setup,
3023f2fe 2152 bool first_boot,
5afbaa36
LP		 2153 struct rlimit *saved_rlimit_nofile,
2154 struct rlimit *saved_rlimit_memlock,
2155 const char **ret_error_message) {
5afbaa36
LP		 2156 int r;
2157
2158 assert(ret_error_message);
2159
2160 /* Sets up various runtime parameters. Many of these initializations are conditionalized:
2161 *
2162 * - Some only apply to --system instances
2163 * - Some only apply to --user instances
2164 * - Some only apply when we first start up, but not when we reexecute
2165 */
2166
2d776038
LP		 2167 if (arg_action != ACTION_RUN)
2168 return 0;
2169
61fbbac1 2170 update_cpu_affinity(skip_setup);
b070c7c0 2171 update_numa_policy(skip_setup);
61fbbac1 2172
3c3c6cb9 2173 if (arg_system) {
5238e957 2174 /* Make sure we leave a core dump without panicking the kernel. */
3c3c6cb9 2175 install_crash_handler();
5afbaa36 2176
3c3c6cb9 2177 if (!skip_setup) {
143fadf3 2178 r = mount_cgroup_controllers();
3c3c6cb9
LP		 2179 if (r < 0) {
2180 *ret_error_message = "Failed to mount cgroup hierarchies";
2181 return r;
2182 }
2183
4bd03515 2184 (void) os_release_status();
b6fad306 2185 (void) hostname_setup(true);
3023f2fe 2186 /* Force transient machine-id on first boot. */
7cd43e34 2187 machine_id_setup(NULL, /* force_transient= */ first_boot, arg_machine_id, NULL);
df883de9 2188 (void) loopback_setup();
3c3c6cb9 2189 bump_unix_max_dgram_qlen();
a8b627aa 2190 bump_file_max_and_nr_open();
3c3c6cb9
LP		 2191 test_usr();
2192 write_container_id();
2193 }
8a2c1fbf 2194
82d7a151
YW		 2195 r = watchdog_set_device(arg_watchdog_device);
2196 if (r < 0)
2197 log_warning_errno(r, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device);
32429805
LP		 2198 } else {
2199 _cleanup_free_ char *p = NULL;
2200
2201 /* Create the runtime directory and place the inaccessible device nodes there, if we run in
2202 * user mode. In system mode mount_setup() already did that. */
2203
2204 r = xdg_user_runtime_dir(&p, "/systemd");
2205 if (r < 0) {
2206 *ret_error_message = "$XDG_RUNTIME_DIR is not set";
2207 return log_emergency_errno(r, "Failed to determine $XDG_RUNTIME_DIR path: %m");
2208 }
2209
e813a74a 2210 (void) mkdir_p_label(p, 0755);
32429805 2211 (void) make_inaccessible_nodes(p, UID_INVALID, GID_INVALID);
3c3c6cb9 2212 }
5afbaa36
LP		 2213
2214 if (arg_timer_slack_nsec != NSEC_INFINITY)
2215 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
3a671cd1 2216 log_warning_errno(errno, "Failed to adjust timer slack, ignoring: %m");
5afbaa36
LP		 2217
2218 if (arg_system && !cap_test_all(arg_capability_bounding_set)) {
2219 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
2220 if (r < 0) {
2221 *ret_error_message = "Failed to drop capability bounding set of usermode helpers";
2222 return log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
2223 }
2224
2225 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
2226 if (r < 0) {
2227 *ret_error_message = "Failed to drop capability bounding set";
2228 return log_emergency_errno(r, "Failed to drop capability bounding set: %m");
2229 }
2230 }
2231
39362f6f
JB		 2232 if (arg_system && arg_no_new_privs) {
2233 if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) {
2234 *ret_error_message = "Failed to disable new privileges";
2235 return log_emergency_errno(errno, "Failed to disable new privileges: %m");
2236 }
2237 }
2238
5afbaa36
LP		 2239 if (arg_syscall_archs) {
2240 r = enforce_syscall_archs(arg_syscall_archs);
2241 if (r < 0) {
2242 *ret_error_message = "Failed to set syscall architectures";
2243 return r;
2244 }
2245 }
2246
2247 if (!arg_system)
2248 /* Become reaper of our children */
2249 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
6b1fa539 2250 log_warning_errno(errno, "Failed to make us a subreaper, ignoring: %m");
5afbaa36 2251
a17c1712
LP		 2252 /* Bump up RLIMIT_NOFILE for systemd itself */
2253 (void) bump_rlimit_nofile(saved_rlimit_nofile);
2254 (void) bump_rlimit_memlock(saved_rlimit_memlock);
5afbaa36 2255
4b9a4b01
LP		 2256 /* Pull credentials from various sources into a common credential directory */
2257 if (arg_system && !skip_setup)
2258 (void) import_credentials();
2259
5afbaa36
LP		 2260 return 0;
2261}
2262
6acca5fc
LP		 2263static int do_queue_default_job(
2264 Manager *m,
2265 const char **ret_error_message) {
2266
2267 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
f1d075dc
ZJS		 2268 const char *unit;
2269 Job *job;
2270 Unit *target;
6acca5fc
LP		 2271 int r;
2272
8755dbad 2273 if (arg_default_unit)
f1d075dc 2274 unit = arg_default_unit;
8755dbad 2275 else if (in_initrd())
f1d075dc 2276 unit = SPECIAL_INITRD_TARGET;
8755dbad 2277 else
f1d075dc 2278 unit = SPECIAL_DEFAULT_TARGET;
8755dbad 2279
f1d075dc 2280 log_debug("Activating default unit: %s", unit);
8755dbad 2281
f1d075dc 2282 r = manager_load_startable_unit_or_warn(m, unit, NULL, &target);
8755dbad
ZJS		 2283 if (r < 0 && in_initrd() && !arg_default_unit) {
2284 /* Fall back to default.target, which we used to always use by default. Only do this if no
2285 * explicit configuration was given. */
2286
2287 log_info("Falling back to " SPECIAL_DEFAULT_TARGET ".");
6acca5fc 2288
8755dbad
ZJS		 2289 r = manager_load_startable_unit_or_warn(m, SPECIAL_DEFAULT_TARGET, NULL, &target);
2290 }
4109ede7 2291 if (r < 0) {
8755dbad 2292 log_info("Falling back to " SPECIAL_RESCUE_TARGET ".");
6acca5fc 2293
4109ede7 2294 r = manager_load_startable_unit_or_warn(m, SPECIAL_RESCUE_TARGET, NULL, &target);
6acca5fc 2295 if (r < 0) {
8755dbad
ZJS		 2296 *ret_error_message = r == -ERFKILL ? SPECIAL_RESCUE_TARGET " masked"
2297 : "Failed to load " SPECIAL_RESCUE_TARGET;
4109ede7 2298 return r;
6acca5fc
LP		 2299 }
2300 }
2301
2302 assert(target->load_state == UNIT_LOADED);
2303
f1d075dc 2304 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, NULL, &error, &job);
6acca5fc
LP		 2305 if (r == -EPERM) {
2306 log_debug_errno(r, "Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
2307
2308 sd_bus_error_free(&error);
2309
f1d075dc 2310 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, NULL, &error, &job);
6acca5fc
LP		 2311 if (r < 0) {
2312 *ret_error_message = "Failed to start default target";
2313 return log_emergency_errno(r, "Failed to start default target: %s", bus_error_message(&error, r));
2314 }
2315
2316 } else if (r < 0) {
2317 *ret_error_message = "Failed to isolate default target";
2318 return log_emergency_errno(r, "Failed to isolate default target: %s", bus_error_message(&error, r));
c86c31d9
ZJS		 2319 } else
2320 log_info("Queued %s job for default target %s.",
2321 job_type_to_string(job->type),
04d232d8 2322 unit_status_string(job->unit, NULL));
6acca5fc 2323
f1d075dc 2324 m->default_unit_job_id = job->id;
6acca5fc
LP		 2325
2326 return 0;
2327}
2328
a9fd4cd1
FB		 2329static void save_rlimits(struct rlimit *saved_rlimit_nofile,
2330 struct rlimit *saved_rlimit_memlock) {
2331
2332 assert(saved_rlimit_nofile);
2333 assert(saved_rlimit_memlock);
2334
2335 if (getrlimit(RLIMIT_NOFILE, saved_rlimit_nofile) < 0)
2336 log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m");
2337
2338 if (getrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock) < 0)
2339 log_warning_errno(errno, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
2340}
2341
2342static void fallback_rlimit_nofile(const struct rlimit *saved_rlimit_nofile) {
2343 struct rlimit *rl;
2344
2345 if (arg_default_rlimit[RLIMIT_NOFILE])
2346 return;
2347
2348 /* Make sure forked processes get limits based on the original kernel setting */
2349
2350 rl = newdup(struct rlimit, saved_rlimit_nofile, 1);
2351 if (!rl) {
2352 log_oom();
2353 return;
2354 }
2355
2356 /* Bump the hard limit for system services to a substantially higher value. The default
2357 * hard limit current kernels set is pretty low (4K), mostly for historical
2358 * reasons. According to kernel developers, the fd handling in recent kernels has been
2359 * optimized substantially enough, so that we can bump the limit now, without paying too
2360 * high a price in memory or performance. Note however that we only bump the hard limit,
2361 * not the soft limit. That's because select() works the way it works, and chokes on fds
2362 * >= 1024. If we'd bump the soft limit globally, it might accidentally happen to
2363 * unexpecting programs that they get fds higher than what they can process using
2364 * select(). By only bumping the hard limit but leaving the low limit as it is we avoid
2365 * this pitfall: programs that are written by folks aware of the select() problem in mind
2366 * (and thus use poll()/epoll instead of select(), the way everybody should) can
2367 * explicitly opt into high fds by bumping their soft limit beyond 1024, to the hard limit
2368 * we pass. */
2369 if (arg_system) {
2370 int nr;
2371
2372 /* Get the underlying absolute limit the kernel enforces */
2373 nr = read_nr_open();
2374
2375 rl->rlim_max = MIN((rlim_t) nr, MAX(rl->rlim_max, (rlim_t) HIGH_RLIMIT_NOFILE));
2376 }
2377
2378 /* If for some reason we were invoked with a soft limit above 1024 (which should never
2379 * happen!, but who knows what we get passed in from pam_limit when invoked as --user
2380 * instance), then lower what we pass on to not confuse our children */
2381 rl->rlim_cur = MIN(rl->rlim_cur, (rlim_t) FD_SETSIZE);
2382
2383 arg_default_rlimit[RLIMIT_NOFILE] = rl;
2384}
2385
2386static void fallback_rlimit_memlock(const struct rlimit *saved_rlimit_memlock) {
2387 struct rlimit *rl;
2388
2389 /* Pass the original value down to invoked processes */
2390
2391 if (arg_default_rlimit[RLIMIT_MEMLOCK])
2392 return;
2393
2394 rl = newdup(struct rlimit, saved_rlimit_memlock, 1);
2395 if (!rl) {
2396 log_oom();
2397 return;
2398 }
2399
852b6250
LP		 2400 if (arg_system) {
2401 /* Raise the default limit to 8M also on old kernels and in containers (8M is the kernel
2402 * default for this since kernel 5.16) */
2403 rl->rlim_max = MAX(rl->rlim_max, (rlim_t) DEFAULT_RLIMIT_MEMLOCK);
2404 rl->rlim_cur = MAX(rl->rlim_cur, (rlim_t) DEFAULT_RLIMIT_MEMLOCK);
2405 }
2406
a9fd4cd1
FB		 2407 arg_default_rlimit[RLIMIT_MEMLOCK] = rl;
2408}
2409
d55ed7de 2410static void setenv_manager_environment(void) {
d55ed7de
ZJS		 2411 int r;
2412
2413 STRV_FOREACH(p, arg_manager_environment) {
2414 log_debug("Setting '%s' in our own environment.", *p);
2415
2416 r = putenv_dup(*p, true);
2417 if (r < 0)
2418 log_warning_errno(errno, "Failed to setenv \"%s\", ignoring: %m", *p);
2419 }
2420}
2421
fb39af4c
ZJS		 2422static void reset_arguments(void) {
2423 /* Frees/resets arg_* variables, with a few exceptions commented below. */
970777b5
LP		 2424
2425 arg_default_unit = mfree(arg_default_unit);
fb39af4c
ZJS		 2426
2427 /* arg_system — ignore */
2428
2429 arg_dump_core = true;
2430 arg_crash_chvt = -1;
2431 arg_crash_shell = false;
2432 arg_crash_reboot = false;
970777b5 2433 arg_confirm_spawn = mfree(arg_confirm_spawn);
fb39af4c 2434 arg_show_status = _SHOW_STATUS_INVALID;
36cf4507 2435 arg_status_unit_format = STATUS_UNIT_FORMAT_DEFAULT;
fb39af4c
ZJS		 2436 arg_switched_root = false;
2437 arg_pager_flags = 0;
2438 arg_service_watchdogs = true;
2439 arg_default_std_output = EXEC_OUTPUT_JOURNAL;
2440 arg_default_std_error = EXEC_OUTPUT_INHERIT;
2441 arg_default_restart_usec = DEFAULT_RESTART_USEC;
e4c7b5f5
ZJS		 2442 arg_default_timeout_start_usec = manager_default_timeout(arg_system);
2443 arg_default_timeout_stop_usec = manager_default_timeout(arg_system);
2444 arg_default_timeout_abort_usec = manager_default_timeout(arg_system);
fb39af4c 2445 arg_default_timeout_abort_set = false;
e4c7b5f5 2446 arg_default_device_timeout_usec = manager_default_timeout(arg_system);
fb39af4c
ZJS		 2447 arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
2448 arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
2449 arg_runtime_watchdog = 0;
65224c1d 2450 arg_reboot_watchdog = 10 * USEC_PER_MINUTE;
acafd7d8 2451 arg_kexec_watchdog = 0;
5717062e 2452 arg_pretimeout_watchdog = 0;
919ea64f
ŁS		 2453 arg_early_core_pattern = mfree(arg_early_core_pattern);
2454 arg_watchdog_device = mfree(arg_watchdog_device);
aff3a9e1 2455 arg_watchdog_pretimeout_governor = mfree(arg_watchdog_pretimeout_governor);
fb39af4c 2456
970777b5 2457 arg_default_environment = strv_free(arg_default_environment);
d55ed7de 2458 arg_manager_environment = strv_free(arg_manager_environment);
fb39af4c
ZJS		 2459 rlimit_free_all(arg_default_rlimit);
2460
3fd5190b 2461 arg_capability_bounding_set = CAP_MASK_UNSET;
fb39af4c
ZJS		 2462 arg_no_new_privs = false;
2463 arg_timer_slack_nsec = NSEC_INFINITY;
2464 arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
2465
970777b5 2466 arg_syscall_archs = set_free(arg_syscall_archs);
61fbbac1 2467
fb39af4c
ZJS		 2468 /* arg_serialization — ignore */
2469
2470 arg_default_cpu_accounting = -1;
2471 arg_default_io_accounting = false;
2472 arg_default_ip_accounting = false;
2473 arg_default_blockio_accounting = false;
2474 arg_default_memory_accounting = MEMORY_ACCOUNTING_DEFAULT;
2475 arg_default_tasks_accounting = true;
3a0f06c4 2476 arg_default_tasks_max = DEFAULT_TASKS_MAX;
fb39af4c
ZJS		 2477 arg_machine_id = (sd_id128_t) {};
2478 arg_cad_burst_action = EMERGENCY_ACTION_REBOOT_FORCE;
2479 arg_default_oom_policy = OOM_STOP;
2480
61fbbac1 2481 cpu_set_reset(&arg_cpu_affinity);
b070c7c0 2482 numa_policy_reset(&arg_numa_policy);
d247f232
LP		 2483
2484 arg_random_seed = mfree(arg_random_seed);
2485 arg_random_seed_size = 0;
33d943d1 2486 arg_clock_usec = 0;
d4a402e4
LP		 2487
2488 arg_default_oom_score_adjust_set = false;
aa5ae971 2489 arg_default_smack_process_label = mfree(arg_default_smack_process_label);
856bfaeb
LB		 2490
2491 arg_reload_limit_interval_sec = 0;
2492 arg_reload_limit_burst = 0;
d4a402e4
LP		 2493}
2494
2495static void determine_default_oom_score_adjust(void) {
2496 int r, a, b;
2497
2498 /* Run our services at slightly higher OOM score than ourselves. But let's be conservative here, and
2499 * do this only if we don't run as root (i.e. only if we are run in user mode, for an unprivileged
2500 * user). */
2501
2502 if (arg_default_oom_score_adjust_set)
2503 return;
2504
2505 if (getuid() == 0)
2506 return;
2507
2508 r = get_oom_score_adjust(&a);
2509 if (r < 0)
2510 return (void) log_warning_errno(r, "Failed to determine current OOM score adjustment value, ignoring: %m");
2511
2512 assert_cc(100 <= OOM_SCORE_ADJ_MAX);
2513 b = a >= OOM_SCORE_ADJ_MAX - 100 ? OOM_SCORE_ADJ_MAX : a + 100;
2514
2515 if (a == b)
2516 return;
2517
2518 arg_default_oom_score_adjust = b;
2519 arg_default_oom_score_adjust_set = true;
970777b5
LP		 2520}
2521
a9fd4cd1
FB		 2522static int parse_configuration(const struct rlimit *saved_rlimit_nofile,
2523 const struct rlimit *saved_rlimit_memlock) {
97d1fb94
LP		 2524 int r;
2525
a9fd4cd1
FB		 2526 assert(saved_rlimit_nofile);
2527 assert(saved_rlimit_memlock);
2528
fb39af4c
ZJS		 2529 /* Assign configuration defaults */
2530 reset_arguments();
2531
97d1fb94 2532 r = parse_config_file();
470a5e6d
ZJS		 2533 if (r < 0)
2534 log_warning_errno(r, "Failed to parse config file, ignoring: %m");
97d1fb94
LP		 2535
2536 if (arg_system) {
2537 r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
2538 if (r < 0)
2539 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
2540 }
2541
a9fd4cd1
FB		 2542 /* Initialize some default rlimits for services if they haven't been configured */
2543 fallback_rlimit_nofile(saved_rlimit_nofile);
2544 fallback_rlimit_memlock(saved_rlimit_memlock);
2545
97d1fb94
LP		 2546 /* Note that this also parses bits from the kernel command line, including "debug". */
2547 log_parse_environment();
2548
db33214b 2549 /* Initialize the show status setting if it hasn't been set explicitly yet */
7a293242 2550 if (arg_show_status == _SHOW_STATUS_INVALID)
db33214b
LP		 2551 arg_show_status = SHOW_STATUS_YES;
2552
d4a402e4
LP		 2553 /* Slightly raise the OOM score for our services if we are running for unprivileged users. */
2554 determine_default_oom_score_adjust();
2555
d55ed7de
ZJS		 2556 /* Push variables into the manager environment block */
2557 setenv_manager_environment();
2558
a4303b40
DDM		 2559 /* Parse log environment variables again to take into account any new environment variables. */
2560 log_parse_environment();
2561
97d1fb94
LP		 2562 return 0;
2563}
2564
b0d7c989
LP		 2565static int safety_checks(void) {
2566
febf46a4 2567 if (getpid_cached() == 1 &&
baaa35ad
ZJS		 2568 arg_action != ACTION_RUN)
2569 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2570 "Unsupported execution mode while PID 1.");
febf46a4
LP		 2571
2572 if (getpid_cached() == 1 &&
baaa35ad
ZJS		 2573 !arg_system)
2574 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2575 "Can't run --user mode as PID 1.");
febf46a4
LP		 2576
2577 if (arg_action == ACTION_RUN &&
2578 arg_system &&
baaa35ad
ZJS		 2579 getpid_cached() != 1)
2580 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2581 "Can't run system mode unless PID 1.");
febf46a4 2582
b0d7c989 2583 if (arg_action == ACTION_TEST &&
baaa35ad
ZJS		 2584 geteuid() == 0)
2585 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2586 "Don't run test mode as root.");
b0d7c989
LP		 2587
2588 if (!arg_system &&
2589 arg_action == ACTION_RUN &&
baaa35ad
ZJS		 2590 sd_booted() <= 0)
2591 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
2592 "Trying to run as user instance, but the system has not been booted with systemd.");
b0d7c989
LP		 2593
2594 if (!arg_system &&
2595 arg_action == ACTION_RUN &&
baaa35ad
ZJS		 2596 !getenv("XDG_RUNTIME_DIR"))
2597 return log_error_errno(SYNTHETIC_ERRNO(EUNATCH),
2598 "Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
b0d7c989
LP		 2599
2600 if (arg_system &&
2601 arg_action == ACTION_RUN &&
baaa35ad
ZJS		 2602 running_in_chroot() > 0)
2603 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
2604 "Cannot be run in a chroot() environment.");
b0d7c989
LP		 2605
2606 return 0;
2607}
2608
74da609f
LP		 2609static int initialize_security(
2610 bool *loaded_policy,
2611 dual_timestamp *security_start_timestamp,
2612 dual_timestamp *security_finish_timestamp,
2613 const char **ret_error_message) {
2614
2615 int r;
2616
2617 assert(loaded_policy);
2618 assert(security_start_timestamp);
2619 assert(security_finish_timestamp);
2620 assert(ret_error_message);
2621
2622 dual_timestamp_get(security_start_timestamp);
2623
97149f40 2624 r = mac_selinux_setup(loaded_policy);
74da609f
LP		 2625 if (r < 0) {
2626 *ret_error_message = "Failed to load SELinux policy";
2627 return r;
2628 }
2629
2630 r = mac_smack_setup(loaded_policy);
2631 if (r < 0) {
2632 *ret_error_message = "Failed to load SMACK policy";
2633 return r;
2634 }
2635
2ffadd3c
Y		 2636 r = mac_apparmor_setup();
2637 if (r < 0) {
2638 *ret_error_message = "Failed to load AppArmor policy";
2639 return r;
2640 }
2641
74da609f
LP		 2642 r = ima_setup();
2643 if (r < 0) {
2644 *ret_error_message = "Failed to load IMA policy";
2645 return r;
2646 }
2647
2648 dual_timestamp_get(security_finish_timestamp);
2649 return 0;
2650}
2651
efeb853f
LP		 2652static int collect_fds(FDSet **ret_fds, const char **ret_error_message) {
2653 int r;
2654
2655 assert(ret_fds);
2656 assert(ret_error_message);
2657
2658 r = fdset_new_fill(ret_fds);
2659 if (r < 0) {
2660 *ret_error_message = "Failed to allocate fd set";
2661 return log_emergency_errno(r, "Failed to allocate fd set: %m");
2662 }
2663
2664 fdset_cloexec(*ret_fds, true);
2665
2666 if (arg_serialization)
2667 assert_se(fdset_remove(*ret_fds, fileno(arg_serialization)) >= 0);
2668
2669 return 0;
2670}
2671
2e51b31c
LP		 2672static void setup_console_terminal(bool skip_setup) {
2673
2674 if (!arg_system)
2675 return;
2676
2677 /* Become a session leader if we aren't one yet. */
2678 (void) setsid();
2679
ff3a7019
ZJS		 2680 /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a
2681 * controlling tty. */
2e51b31c
LP		 2682 (void) release_terminal();
2683
2684 /* Reset the console, but only if this is really init and we are freshly booted */
2685 if (getpid_cached() == 1 && !skip_setup)
2686 (void) console_setup();
2687}
2688
aa40ff07
LP		 2689static bool early_skip_setup_check(int argc, char *argv[]) {
2690 bool found_deserialize = false;
aa40ff07 2691
ff3a7019
ZJS		 2692 /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much
2693 * later, so let's just have a quick peek here. Note that if we have switched root, do all the
2694 * special setup things anyway, even if in that case we also do deserialization. */
aa40ff07 2695
431733b8 2696 for (int i = 1; i < argc; i++)
aa40ff07
LP		 2697 if (streq(argv[i], "--switched-root"))
2698 return false; /* If we switched root, don't skip the setup. */
2699 else if (streq(argv[i], "--deserialize"))
2700 found_deserialize = true;
aa40ff07
LP		 2701
2702 return found_deserialize; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2703}
2704
0e06a031
LP		 2705static int save_env(void) {
2706 char **l;
2707
2708 l = strv_copy(environ);
2709 if (!l)
2710 return -ENOMEM;
2711
2712 strv_free_and_replace(saved_env, l);
2713 return 0;
2714}
2715
60918275 2716int main(int argc, char *argv[]) {
5409c6fc
ZJS		 2717 dual_timestamp
2718 initrd_timestamp = DUAL_TIMESTAMP_NULL,
2719 userspace_timestamp = DUAL_TIMESTAMP_NULL,
2720 kernel_timestamp = DUAL_TIMESTAMP_NULL,
2721 security_start_timestamp = DUAL_TIMESTAMP_NULL,
2722 security_finish_timestamp = DUAL_TIMESTAMP_NULL;
ddfa8b0b
LP		 2723 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0),
2724 saved_rlimit_memlock = RLIMIT_MAKE_CONST(RLIM_INFINITY); /* The original rlimits we passed
2725 * in. Note we use different values
2726 * for the two that indicate whether
2727 * these fields are initialized! */
5409c6fc 2728 bool skip_setup, loaded_policy = false, queue_default_job = false, first_boot = false;
625e8690 2729 char *switch_root_dir = NULL, *switch_root_init = NULL;
9d76d730 2730 usec_t before_startup, after_startup;
625e8690 2731 static char systemd[] = "systemd";
b22d392d 2732 const char *error_message = NULL;
625e8690
LP		 2733 int r, retval = EXIT_FAILURE;
2734 Manager *m = NULL;
a16e1123 2735 FDSet *fds = NULL;
27b14a22 2736
61b9769b 2737 assert_se(argc > 0 && !isempty(argv[0]));
cf3095ac 2738
d72a8f10 2739 /* SysV compatibility: redirect init → telinit */
6808a0bc 2740 redirect_telinit(argc, argv);
2cb1a60d 2741
d72a8f10 2742 /* Take timestamps early on */
c3a170f3
HH		 2743 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
2744 dual_timestamp_get(&userspace_timestamp);
2745
d72a8f10 2746 /* Figure out whether we need to do initialize the system, or if we already did that because we are
ff3a7019 2747 * reexecuting. */
aa40ff07 2748 skip_setup = early_skip_setup_check(argc, argv);
d03bc1b8 2749
ff3a7019
ZJS		 2750 /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent
2751 * reexecution we are then called 'systemd'. That is confusing, hence let's call us systemd
2752 * right-away. */
f3b6a3ed 2753 program_invocation_short_name = systemd;
eee8b7ab 2754 (void) prctl(PR_SET_NAME, systemd);
5d6b1584 2755
d72a8f10 2756 /* Save the original command line */
36fea155 2757 save_argc_argv(argc, argv);
f3b6a3ed 2758
0e06a031
LP		 2759 /* Save the original environment as we might need to restore it if we're requested to execute another
2760 * system manager later. */
2761 r = save_env();
2762 if (r < 0) {
2763 error_message = "Failed to copy environment block";
2764 goto finish;
2765 }
a5cede8c 2766
6fdb8de4 2767 /* Make sure that if the user says "syslog" we actually log to the journal. */
c1dc6153 2768 log_set_upgrade_syslog_to_journal(true);
bbe63281 2769
df0ff127 2770 if (getpid_cached() == 1) {
b5752d23
LP		 2771 /* When we run as PID 1 force system mode */
2772 arg_system = true;
2773
48a601fe 2774 /* Disable the umask logic */
90dc8c2e
MG		 2775 umask(0);
2776
ff3a7019
ZJS		 2777 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might
2778 * not be activated yet (even though the log socket for it exists). */
d075092f
LP		 2779 log_set_prohibit_ipc(true);
2780
ff3a7019
ZJS		 2781 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This
2782 * is important so that we never end up logging to any foreign stderr, for example if we have
2783 * to log in a child process right before execve()'ing the actual binary, at a point in time
2784 * where socket activation stderr/stdout area already set up. */
48a601fe 2785 log_set_always_reopen_console(true);
48a601fe 2786
92890452 2787 if (detect_container() <= 0) {
4f8d551f 2788
92890452 2789 /* Running outside of a container as PID 1 */
1e344c1d 2790 log_set_target_and_open(LOG_TARGET_KMSG);
a866073d 2791
92890452
LP		 2792 if (in_initrd())
2793 initrd_timestamp = userspace_timestamp;
c3ba6250 2794
92890452
LP		 2795 if (!skip_setup) {
2796 r = mount_setup_early();
2797 if (r < 0) {
2798 error_message = "Failed to mount early API filesystems";
2799 goto finish;
2800 }
2801
0a2eef1e
LP		 2802 /* Let's open the log backend a second time, in case the first time didn't
2803 * work. Quite possibly we have mounted /dev just now, so /dev/kmsg became
2804 * available, and it previously wasn't. */
2805 log_open();
2806
6123dfaa
ZJS		 2807 disable_printk_ratelimit();
2808
92890452
LP		 2809 r = initialize_security(
2810 &loaded_policy,
2811 &security_start_timestamp,
2812 &security_finish_timestamp,
2813 &error_message);
2814 if (r < 0)
2815 goto finish;
d723cd65 2816 }
eee8b7ab 2817
92890452 2818 if (mac_selinux_init() < 0) {
a9ba0e32 2819 error_message = "Failed to initialize SELinux support";
96694e99 2820 goto finish;
92890452 2821 }
0b3325e7 2822
92890452
LP		 2823 if (!skip_setup)
2824 initialize_clock();
2825
ff3a7019
ZJS		 2826 /* Set the default for later on, but don't actually open the logs like this for
2827 * now. Note that if we are transitioning from the initrd there might still be
2828 * journal fd open, and we shouldn't attempt opening that before we parsed
2829 * /proc/cmdline which might redirect output elsewhere. */
92890452
LP		 2830 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
2831
2832 } else {
2833 /* Running inside a container, as PID 1 */
1e344c1d 2834 log_set_target_and_open(LOG_TARGET_CONSOLE);
92890452
LP		 2835
2836 /* For later on, see above... */
2837 log_set_target(LOG_TARGET_JOURNAL);
2838
45250e66 2839 /* clear the kernel timestamp, because we are in a container */
92890452 2840 kernel_timestamp = DUAL_TIMESTAMP_NULL;
cb6531be 2841 }
7948c4df 2842
92890452 2843 initialize_coredump(skip_setup);
a866073d 2844
92890452
LP		 2845 r = fixup_environment();
2846 if (r < 0) {
2847 log_emergency_errno(r, "Failed to fix up PID 1 environment: %m");
2848 error_message = "Failed to fix up PID1 environment";
2849 goto finish;
2850 }
a866073d 2851
ff3a7019
ZJS		 2852 /* Try to figure out if we can use colors with the console. No need to do that for user
2853 * instances since they never log into the console. */
3a18b604 2854 log_show_color(colors_enabled());
92890452 2855
c76cf844
AK		 2856 r = make_null_stdio();
2857 if (r < 0)
92890452 2858 log_warning_errno(r, "Failed to redirect standard streams to /dev/null, ignoring: %m");
f84f9974 2859
a132bef0 2860 /* Load the kernel modules early. */
2e75e2a8 2861 if (!skip_setup)
e921a00d 2862 (void) kmod_setup();
2e75e2a8 2863
3196e423 2864 /* Mount /proc, /sys and friends, so that /proc/cmdline and /proc/$PID/fd is available. */
f74349d8 2865 r = mount_setup(loaded_policy, skip_setup);
cb6531be
ZJS		 2866 if (r < 0) {
2867 error_message = "Failed to mount API filesystems";
8efe3c01 2868 goto finish;
cb6531be 2869 }
c18ecf03 2870
0be72218
JD		 2871 /* The efivarfs is now mounted, let's lock down the system token. */
2872 lock_down_efi_variables();
209b2592
FB		 2873
2874 /* Cache command-line options passed from EFI variables */
2875 if (!skip_setup)
2876 (void) cache_efi_options_variable();
3196e423
LP		 2877 } else {
2878 /* Running as user instance */
2879 arg_system = false;
2a646b1d 2880 log_set_always_reopen_console(true);
1e344c1d 2881 log_set_target_and_open(LOG_TARGET_AUTO);
3196e423
LP		 2882
2883 /* clear the kernel timestamp, because we are not PID 1 */
2884 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2885
963b6b90
ŁS		 2886 /* Clear ambient capabilities, so services do not inherit them implicitly. Dropping them does
2887 * not affect the permitted and effective sets which are important for the manager itself to
2888 * operate. */
2889 capability_ambient_set_apply(0, /* also_inherit= */ false);
2890
3196e423
LP		 2891 if (mac_selinux_init() < 0) {
2892 error_message = "Failed to initialize SELinux support";
2893 goto finish;
2894 }
0c85a4f3 2895 }
4ade7963 2896
a9fd4cd1
FB		 2897 /* Save the original RLIMIT_NOFILE/RLIMIT_MEMLOCK so that we can reset it later when
2898 * transitioning from the initrd to the main systemd or suchlike. */
2899 save_rlimits(&saved_rlimit_nofile, &saved_rlimit_memlock);
2900
4ade7963 2901 /* Reset all signal handlers. */
ce30c8dc 2902 (void) reset_all_signal_handlers();
9c274488 2903 (void) ignore_signals(SIGNALS_IGNORE);
078e4539 2904
ffe5c01e
FB		 2905 (void) parse_configuration(&saved_rlimit_nofile, &saved_rlimit_memlock);
2906
2907 r = parse_argv(argc, argv);
2908 if (r < 0) {
2909 error_message = "Failed to parse commandline arguments";
f170852a 2910 goto finish;
ffe5c01e 2911 }
10c961b9 2912
b0d7c989
LP		 2913 r = safety_checks();
2914 if (r < 0)
fe783b03 2915 goto finish;
fe783b03 2916
5c08257b 2917 if (IN_SET(arg_action, ACTION_TEST, ACTION_HELP, ACTION_DUMP_CONFIGURATION_ITEMS, ACTION_DUMP_BUS_PROPERTIES, ACTION_BUS_INTROSPECT))
384c2c32 2918 pager_open(arg_pager_flags);
b0d7c989
LP		 2919
2920 if (arg_action != ACTION_RUN)
74e7579c 2921 skip_setup = true;
b87c2aa6 2922
fa0f4d8a 2923 if (arg_action == ACTION_HELP) {
37ec0fdd 2924 retval = help() < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
f170852a 2925 goto finish;
9ba0bc4e
ZJS		 2926 } else if (arg_action == ACTION_VERSION) {
2927 retval = version();
2928 goto finish;
fa0f4d8a 2929 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
e537352b 2930 unit_dump_config_items(stdout);
22f4096c 2931 retval = EXIT_SUCCESS;
e537352b 2932 goto finish;
bbc1acab
YW		 2933 } else if (arg_action == ACTION_DUMP_BUS_PROPERTIES) {
2934 dump_bus_properties(stdout);
2935 retval = EXIT_SUCCESS;
2936 goto finish;
5c08257b
ZJS		 2937 } else if (arg_action == ACTION_BUS_INTROSPECT) {
2938 r = bus_manager_introspect_implementations(stdout, arg_bus_introspect);
2939 retval = r >= 0 ? EXIT_SUCCESS : EXIT_FAILURE;
2940 goto finish;
f170852a
LP		 2941 }
2942
4c701096 2943 assert_se(IN_SET(arg_action, ACTION_RUN, ACTION_TEST));
f170852a 2944
5a2e0c62
LP		 2945 /* Move out of the way, so that we won't block unmounts */
2946 assert_se(chdir("/") == 0);
2947
dea374e8 2948 if (arg_action == ACTION_RUN) {
d247f232
LP		 2949 if (!skip_setup) {
2950 /* Apply the systemd.clock_usec= kernel command line switch */
45250e66 2951 apply_clock_update();
a70c72a0 2952
d247f232
LP		 2953 /* Apply random seed from kernel command line */
2954 cmdline_take_random_seed();
2955 }
2956
c6885f5f
FB		 2957 /* A core pattern might have been specified via the cmdline. */
2958 initialize_core_pattern(skip_setup);
2959
efeb853f 2960 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
a70c72a0
LP		 2961 log_close();
2962
2963 /* Remember open file descriptors for later deserialization */
efeb853f
LP		 2964 r = collect_fds(&fds, &error_message);
2965 if (r < 0)
dea374e8 2966 goto finish;
a16e1123 2967
2e51b31c
LP		 2968 /* Give up any control of the console, but make sure its initialized. */
2969 setup_console_terminal(skip_setup);
56d96fc0 2970
a70c72a0
LP		 2971 /* Open the logging devices, if possible and necessary */
2972 log_open();
56d96fc0 2973 }
4ade7963 2974
31aef7ff 2975 log_execution_mode(&first_boot);
a5dab5ce 2976
2d776038 2977 r = initialize_runtime(skip_setup,
3023f2fe 2978 first_boot,
2d776038
LP		 2979 &saved_rlimit_nofile,
2980 &saved_rlimit_memlock,
2981 &error_message);
2982 if (r < 0)
2983 goto finish;
4096d6f5 2984
b380b643 2985 r = manager_new(arg_system ? LOOKUP_SCOPE_SYSTEM : LOOKUP_SCOPE_USER,
e0a3da1f
ZJS		 2986 arg_action == ACTION_TEST ? MANAGER_TEST_FULL : 0,
2987 &m);
e96d6be7 2988 if (r < 0) {
da927ba9 2989 log_emergency_errno(r, "Failed to allocate manager object: %m");
cb6531be 2990 error_message = "Failed to allocate manager object";
60918275
LP		 2991 goto finish;
2992 }
2993
9f9f0342
LP		 2994 m->timestamps[MANAGER_TIMESTAMP_KERNEL] = kernel_timestamp;
2995 m->timestamps[MANAGER_TIMESTAMP_INITRD] = initrd_timestamp;
2996 m->timestamps[MANAGER_TIMESTAMP_USERSPACE] = userspace_timestamp;
d4ee7bd8
YW		 2997 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START)] = security_start_timestamp;
2998 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH)] = security_finish_timestamp;
9e58ff9c 2999
85cb4151 3000 set_manager_defaults(m);
7b46fc6a 3001 set_manager_settings(m);
fd130612 3002 manager_set_first_boot(m, first_boot);
d35fe8c0 3003 manager_set_switching_root(m, arg_switched_root);
27d340c7 3004
bf4df7c3 3005 /* Remember whether we should queue the default job */
d3b1c508 3006 queue_default_job = !arg_serialization || arg_switched_root;
bf4df7c3 3007
9d76d730
LP		 3008 before_startup = now(CLOCK_MONOTONIC);
3009
2a7cf953 3010 r = manager_startup(m, arg_serialization, fds, /* root= */ NULL);
58f88d92 3011 if (r < 0) {
cefb3eda 3012 error_message = "Failed to start up manager";
58f88d92
ZJS		 3013 goto finish;
3014 }
a16e1123 3015
6acca5fc 3016 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2feceb5e 3017 fds = fdset_free(fds);
74ca738f 3018 arg_serialization = safe_fclose(arg_serialization);
bf4df7c3
LP		 3019
3020 if (queue_default_job) {
6acca5fc 3021 r = do_queue_default_job(m, &error_message);
718db961 3022 if (r < 0)
37d88da7 3023 goto finish;
6acca5fc 3024 }
ab17a050 3025
6acca5fc 3026 after_startup = now(CLOCK_MONOTONIC);
60918275 3027
6acca5fc
LP		 3028 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
3029 "Loaded units and determined initial transaction in %s.",
5291f26d 3030 FORMAT_TIMESPAN(after_startup - before_startup, 100 * USEC_PER_MSEC));
07672f49 3031
6acca5fc 3032 if (arg_action == ACTION_TEST) {
2a341bb9 3033 manager_test_summary(m);
6acca5fc
LP		 3034 retval = EXIT_SUCCESS;
3035 goto finish;
e965d56d 3036 }
d46de8a1 3037
5409c6fc
ZJS		 3038 r = invoke_main_loop(m,
3039 &saved_rlimit_nofile,
3040 &saved_rlimit_memlock,
3041 &retval,
5409c6fc
ZJS		 3042 &fds,
3043 &switch_root_dir,
3044 &switch_root_init,
3045 &error_message);
3046 assert(r < 0 || IN_SET(r, MANAGER_EXIT, /* MANAGER_OK is not expected here. */
3047 MANAGER_RELOAD,
3048 MANAGER_REEXECUTE,
3049 MANAGER_REBOOT,
3050 MANAGER_POWEROFF,
3051 MANAGER_HALT,
3052 MANAGER_KEXEC,
3053 MANAGER_SWITCH_ROOT));
f170852a 3054
60918275 3055finish:
b87c2aa6
ZJS		 3056 pager_close();
3057
92890452 3058 if (m) {
986935cf
FB		 3059 arg_reboot_watchdog = manager_get_watchdog(m, WATCHDOG_REBOOT);
3060 arg_kexec_watchdog = manager_get_watchdog(m, WATCHDOG_KEXEC);
92890452
LP		 3061 m = manager_free(m);
3062 }
60918275 3063
cc56fafe 3064 mac_selinux_finish();
b2bb3dbe 3065
5409c6fc 3066 if (IN_SET(r, MANAGER_REEXECUTE, MANAGER_SWITCH_ROOT))
1e3eee8c
ZJS		 3067 r = do_reexecute(r,
3068 argc, argv,
3069 &saved_rlimit_nofile,
3070 &saved_rlimit_memlock,
3071 fds,
3072 switch_root_dir,
3073 switch_root_init,
3074 &error_message); /* This only returns if reexecution failed */
a16e1123 3075
74ca738f 3076 arg_serialization = safe_fclose(arg_serialization);
2feceb5e 3077 fds = fdset_free(fds);
a16e1123 3078
0e06a031
LP		 3079 saved_env = strv_free(saved_env);
3080
349cc4a5 3081#if HAVE_VALGRIND_VALGRIND_H
54b434b1
LP		 3082 /* If we are PID 1 and running under valgrind, then let's exit
3083 * here explicitly. valgrind will only generate nice output on
3084 * exit(), not on exec(), hence let's do the former not the
3085 * latter here. */
8a2c1fbf
EJ		 3086 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND) {
3087 /* Cleanup watchdog_device strings for valgrind. We need them
3088 * in become_shutdown() so normally we cannot free them yet. */
3089 watchdog_free_device();
7d9eea2b 3090 reset_arguments();
27fe58b7 3091 return retval;
8a2c1fbf 3092 }
54b434b1
LP		 3093#endif
3094
7e11a95e
EV		 3095#if HAS_FEATURE_ADDRESS_SANITIZER
3096 __lsan_do_leak_check();
3097#endif
3098
5409c6fc
ZJS		 3099 /* Try to invoke the shutdown binary unless we already failed.
3100 * If we failed above, we want to freeze after finishing cleanup. */
b22d392d
VC		 3101 if (arg_system && IN_SET(r, MANAGER_EXIT,
3102 MANAGER_REBOOT,
3103 MANAGER_POWEROFF,
3104 MANAGER_HALT,
3105 MANAGER_KEXEC)) {
3106
3107 r = become_shutdown(r, retval);
4a36297c 3108 log_error_errno(r, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
9b9881d7 3109 error_message = "Failed to execute shutdown binary";
b9080b03
FF		 3110 }
3111
8a2c1fbf
EJ		 3112 watchdog_free_device();
3113 arg_watchdog_device = mfree(arg_watchdog_device);
3114
df0ff127 3115 if (getpid_cached() == 1) {
cb6531be
ZJS		 3116 if (error_message)
3117 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
1fc464f6 3118 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
bb259772
LP		 3119 "%s.", error_message);
3120 freeze_or_exit_or_reboot();
cb6531be 3121 }
c3b3c274 3122
7d9eea2b 3123 reset_arguments();
60918275
LP		 3124 return retval;
3125}
Unnamed repository; edit this file 'description' to name the repository.