]> 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
Fix DPI for MX Master 2s bluetooth mouse
[thirdparty/systemd.git] / src / core / main.c
CommitLineData
53e1b683 1/* SPDX-License-Identifier: LGPL-2.1+ */
a7334b09 2
60918275 3#include <errno.h>
3dfc9763 4#include <fcntl.h>
f170852a 5#include <getopt.h>
664f88a7 6#include <sys/mount.h>
3dfc9763 7#include <sys/prctl.h>
b9e74c39 8#include <sys/reboot.h>
3dfc9763 9#include <unistd.h>
349cc4a5 10#if HAVE_SECCOMP
b64a3d86
LP		 11#include <seccomp.h>
12#endif
349cc4a5 13#if HAVE_VALGRIND_VALGRIND_H
3dfc9763
LP		 14#include <valgrind/valgrind.h>
15#endif
54b434b1 16
718db961 17#include "sd-bus.h"
cf0fbc49 18#include "sd-daemon.h"
b2e7486c 19#include "sd-messages.h"
3dfc9763 20
b5efdb8a 21#include "alloc-util.h"
d9d93745 22#include "architecture.h"
3dfc9763
LP		 23#include "build.h"
24#include "bus-error.h"
25#include "bus-util.h"
430f0182 26#include "capability-util.h"
a88c5b8a 27#include "cgroup-util.h"
24efb112 28#include "clock-util.h"
3dfc9763 29#include "conf-parser.h"
618234a5 30#include "cpu-set-util.h"
3dfc9763 31#include "dbus-manager.h"
c18ecf03 32#include "dbus.h"
3dfc9763 33#include "def.h"
c18ecf03 34#include "efi-random.h"
eee8b7ab 35#include "emergency-action.h"
3dfc9763 36#include "env-util.h"
57b7a260 37#include "exit-status.h"
3ffd4af2 38#include "fd-util.h"
3dfc9763 39#include "fdset.h"
718db961 40#include "fileio.h"
f97b34a6 41#include "format-util.h"
f4f15635 42#include "fs-util.h"
3dfc9763
LP		 43#include "hostname-setup.h"
44#include "ima-setup.h"
45#include "killall.h"
46#include "kmod-setup.h"
eefc66aa 47#include "limits-util.h"
d7b8eec7 48#include "load-fragment.h"
3dfc9763 49#include "log.h"
b6e66135 50#include "loopback-setup.h"
b6e66135 51#include "machine-id-setup.h"
3dfc9763 52#include "manager.h"
3dfc9763 53#include "mount-setup.h"
d58ad743 54#include "os-util.h"
3dfc9763 55#include "pager.h"
6bedfcbb 56#include "parse-util.h"
7d5ceb64 57#include "path-util.h"
294bf0c3 58#include "pretty-print.h"
4e731273 59#include "proc-cmdline.h"
3dfc9763 60#include "process-util.h"
8869a0b4 61#include "raw-clone.h"
78f22b97 62#include "rlimit-util.h"
349cc4a5 63#if HAVE_SECCOMP
83f12b27
FS		 64#include "seccomp-util.h"
65#endif
b6e66135 66#include "selinux-setup.h"
3dfc9763
LP		 67#include "selinux-util.h"
68#include "signal-util.h"
ffbd2c4d 69#include "smack-setup.h"
3dfc9763 70#include "special.h"
8fcde012 71#include "stat-util.h"
15a5e950 72#include "stdio-util.h"
3dfc9763
LP		 73#include "strv.h"
74#include "switch-root.h"
a8b627aa 75#include "sysctl-util.h"
3dfc9763 76#include "terminal-util.h"
8612da97 77#include "umask-util.h"
b1d4f8e1 78#include "user-util.h"
9ce17593 79#include "util.h"
3dfc9763
LP		 80#include "virt.h"
81#include "watchdog.h"
b6e66135 82
7e11a95e
EV		 83#if HAS_FEATURE_ADDRESS_SANITIZER
84#include <sanitizer/lsan_interface.h>
85#endif
86
3a0f06c4
ZJS		 87#define DEFAULT_TASKS_MAX ((TasksMax) { 15U, 100U }) /* 15% */
88
f170852a
LP		 89static enum {
90 ACTION_RUN,
e965d56d 91 ACTION_HELP,
9ba0bc4e 92 ACTION_VERSION,
e537352b 93 ACTION_TEST,
bbc1acab
YW		 94 ACTION_DUMP_CONFIGURATION_ITEMS,
95 ACTION_DUMP_BUS_PROPERTIES,
fa0f4d8a 96} arg_action = ACTION_RUN;
fb39af4c
ZJS		 97
98/* Those variables are initalized to 0 automatically, so we avoid uninitialized memory access.
99 * Real defaults are assigned in reset_arguments() below. */
100static char *arg_default_unit;
101static bool arg_system;
102static bool arg_dump_core;
103static int arg_crash_chvt;
104static bool arg_crash_shell;
105static bool arg_crash_reboot;
106static char *arg_confirm_spawn;
107static ShowStatus arg_show_status;
36cf4507 108static StatusUnitFormat arg_status_unit_format;
fb39af4c
ZJS		 109static bool arg_switched_root;
110static PagerFlags arg_pager_flags;
111static bool arg_service_watchdogs;
112static ExecOutput arg_default_std_output;
113static ExecOutput arg_default_std_error;
114static usec_t arg_default_restart_usec;
115static usec_t arg_default_timeout_start_usec;
116static usec_t arg_default_timeout_stop_usec;
117static usec_t arg_default_timeout_abort_usec;
118static bool arg_default_timeout_abort_set;
119static usec_t arg_default_start_limit_interval;
120static unsigned arg_default_start_limit_burst;
121static usec_t arg_runtime_watchdog;
65224c1d 122static usec_t arg_reboot_watchdog;
acafd7d8 123static usec_t arg_kexec_watchdog;
fb39af4c
ZJS		 124static char *arg_early_core_pattern;
125static char *arg_watchdog_device;
126static char **arg_default_environment;
127static struct rlimit *arg_default_rlimit[_RLIMIT_MAX];
128static uint64_t arg_capability_bounding_set;
129static bool arg_no_new_privs;
130static nsec_t arg_timer_slack_nsec;
131static usec_t arg_default_timer_accuracy_usec;
132static Set* arg_syscall_archs;
133static FILE* arg_serialization;
134static int arg_default_cpu_accounting;
135static bool arg_default_io_accounting;
136static bool arg_default_ip_accounting;
137static bool arg_default_blockio_accounting;
138static bool arg_default_memory_accounting;
139static bool arg_default_tasks_accounting;
3a0f06c4 140static TasksMax arg_default_tasks_max;
fb39af4c
ZJS		 141static sd_id128_t arg_machine_id;
142static EmergencyAction arg_cad_burst_action;
143static OOMPolicy arg_default_oom_policy;
144static CPUSet arg_cpu_affinity;
b070c7c0 145static NUMAPolicy arg_numa_policy;
61fbbac1 146
0e06a031
LP		 147/* A copy of the original environment block */
148static char **saved_env = NULL;
149
a9fd4cd1
FB		 150static int parse_configuration(const struct rlimit *saved_rlimit_nofile,
151 const struct rlimit *saved_rlimit_memlock);
4fc935ca 152
bb259772
LP		 153_noreturn_ static void freeze_or_exit_or_reboot(void) {
154
c3b6a348
LP		 155 /* If we are running in a container, let's prefer exiting, after all we can propagate an exit code to
156 * the container manager, and thus inform it that something went wrong. */
bb259772
LP		 157 if (detect_container() > 0) {
158 log_emergency("Exiting PID 1...");
c3b6a348 159 _exit(EXIT_EXCEPTION);
bb259772 160 }
b9e74c39
LP		 161
162 if (arg_crash_reboot) {
163 log_notice("Rebooting in 10s...");
164 (void) sleep(10);
165
166 log_notice("Rebooting now...");
167 (void) reboot(RB_AUTOBOOT);
168 log_emergency_errno(errno, "Failed to reboot: %m");
169 }
170
171 log_emergency("Freezing execution.");
172 freeze();
173}
174
848e863a 175_noreturn_ static void crash(int sig) {
7d06dad9
MS		 176 struct sigaction sa;
177 pid_t pid;
97c4f35c 178
df0ff127 179 if (getpid_cached() != 1)
abb26902 180 /* Pass this on immediately, if this is not PID 1 */
92ca4cac 181 (void) raise(sig);
abb26902 182 else if (!arg_dump_core)
4104970e 183 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
97c4f35c 184 else {
7d06dad9 185 sa = (struct sigaction) {
189d5bac 186 .sa_handler = nop_signal_handler,
b92bea5d
ZJS		 187 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
188 };
97c4f35c 189
6f5e3f35 190 /* We want to wait for the core process, hence let's enable SIGCHLD */
92ca4cac 191 (void) sigaction(SIGCHLD, &sa, NULL);
6f5e3f35 192
8869a0b4 193 pid = raw_clone(SIGCHLD);
e62d8c39 194 if (pid < 0)
56f64d95 195 log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
97c4f35c 196 else if (pid == 0) {
97c4f35c 197 /* Enable default signal handler for core dump */
15a90032 198
92ca4cac
LP		 199 sa = (struct sigaction) {
200 .sa_handler = SIG_DFL,
201 };
202 (void) sigaction(sig, &sa, NULL);
97c4f35c 203
15a90032
LP		 204 /* Don't limit the coredump size */
205 (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
97c4f35c
LP		 206
207 /* Just to be sure... */
e62d9b81 208 (void) chdir("/");
97c4f35c
LP		 209
210 /* Raise the signal again */
ee05e779 211 pid = raw_getpid();
92ca4cac 212 (void) kill(pid, sig); /* raise() would kill the parent */
97c4f35c
LP		 213
214 assert_not_reached("We shouldn't be here...");
bb85a582 215 _exit(EXIT_EXCEPTION);
4fc935ca 216 } else {
8e12a6ae
LP		 217 siginfo_t status;
218 int r;
4fc935ca
LP		 219
220 /* Order things nicely. */
e62d8c39
ZJS		 221 r = wait_for_terminate(pid, &status);
222 if (r < 0)
da927ba9 223 log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
e1714f02
ZJS		 224 else if (status.si_code != CLD_DUMPED) {
225 const char *s = status.si_code == CLD_EXITED
e04ed6db 226 ? exit_status_to_string(status.si_status, EXIT_STATUS_LIBC)
e1714f02
ZJS		 227 : signal_to_string(status.si_status);
228
ee05e779
ZJS		 229 log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
230 signal_to_string(sig),
e1714f02
ZJS		 231 pid,
232 sigchld_code_to_string(status.si_code),
233 status.si_status, strna(s));
234 } else
235 log_emergency("Caught <%s>, dumped core as pid "PID_FMT".",
236 signal_to_string(sig), pid);
97c4f35c
LP		 237 }
238 }
239
b9e74c39 240 if (arg_crash_chvt >= 0)
92ca4cac 241 (void) chvt(arg_crash_chvt);
601f6a1e 242
7d06dad9
MS		 243 sa = (struct sigaction) {
244 .sa_handler = SIG_IGN,
245 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
246 };
247
248 /* Let the kernel reap children for us */
249 (void) sigaction(SIGCHLD, &sa, NULL);
8c43883a 250
7d06dad9 251 if (arg_crash_shell) {
b9e74c39 252 log_notice("Executing crash shell in 10s...");
92ca4cac 253 (void) sleep(10);
4fc935ca 254
8869a0b4 255 pid = raw_clone(SIGCHLD);
cd3bd60a 256 if (pid < 0)
56f64d95 257 log_emergency_errno(errno, "Failed to fork off crash shell: %m");
6f5e3f35 258 else if (pid == 0) {
b9e74c39 259 (void) setsid();
92ca4cac 260 (void) make_console_stdio();
595225af 261 (void) rlimit_nofile_safe();
92ca4cac 262 (void) execle("/bin/sh", "/bin/sh", NULL, environ);
6f5e3f35 263
ee05e779 264 log_emergency_errno(errno, "execle() failed: %m");
bb85a582 265 _exit(EXIT_EXCEPTION);
b9e74c39
LP		 266 } else {
267 log_info("Spawned crash shell as PID "PID_FMT".", pid);
4cf0b03b 268 (void) wait_for_terminate(pid, NULL);
b9e74c39 269 }
4fc935ca
LP		 270 }
271
bb259772 272 freeze_or_exit_or_reboot();
97c4f35c
LP		 273}
274
275static void install_crash_handler(void) {
297d563d 276 static const struct sigaction sa = {
b92bea5d 277 .sa_handler = crash,
297d563d 278 .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
b92bea5d 279 };
297d563d 280 int r;
97c4f35c 281
297d563d
LP		 282 /* We ignore the return value here, since, we don't mind if we
283 * cannot set up a crash handler */
284 r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
285 if (r < 0)
286 log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
97c4f35c 287}
f170852a 288
56d96fc0
LP		 289static int console_setup(void) {
290 _cleanup_close_ int tty_fd = -1;
291 int r;
80876c20 292
512947d4 293 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
23bbb0de
MS		 294 if (tty_fd < 0)
295 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
80876c20 296
56d96fc0
LP		 297 /* We don't want to force text mode. plymouth may be showing
298 * pictures already from initrd. */
512947d4 299 r = reset_terminal_fd(tty_fd, false);
23bbb0de
MS		 300 if (r < 0)
301 return log_error_errno(r, "Failed to reset /dev/console: %m");
843d2643 302
56d96fc0 303 return 0;
80876c20
LP		 304}
305
ee48dbd5 306static int set_machine_id(const char *m) {
e042eab7 307 sd_id128_t t;
8b26cdbd 308 assert(m);
ee48dbd5 309
e042eab7 310 if (sd_id128_from_string(m, &t) < 0)
ee48dbd5
NC		 311 return -EINVAL;
312
e042eab7 313 if (sd_id128_is_null(t))
ee48dbd5
NC		 314 return -EINVAL;
315
e042eab7 316 arg_machine_id = t;
ee48dbd5
NC		 317 return 0;
318}
319
96287a49 320static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
f170852a 321
059cb385 322 int r;
f170852a 323
059cb385 324 assert(key);
5192bd19 325
1d84ad94 326 if (STR_IN_SET(key, "systemd.unit", "rd.systemd.unit")) {
bf4df7c3 327
1d84ad94
LP		 328 if (proc_cmdline_value_missing(key, value))
329 return 0;
bf4df7c3 330
1d84ad94
LP		 331 if (!unit_name_is_valid(value, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE))
332 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key, value);
333 else if (in_initrd() == !!startswith(key, "rd.")) {
334 if (free_and_strdup(&arg_default_unit, value) < 0)
335 return log_oom();
336 }
f170852a 337
1d84ad94 338 } else if (proc_cmdline_key_streq(key, "systemd.dump_core")) {
4fc935ca 339
1d84ad94 340 r = value ? parse_boolean(value) : true;
fb472900 341 if (r < 0)
5e1ee764 342 log_warning_errno(r, "Failed to parse dump core switch %s, ignoring: %m", value);
4fc935ca 343 else
fa0f4d8a 344 arg_dump_core = r;
4fc935ca 345
c6885f5f
FB		 346 } else if (proc_cmdline_key_streq(key, "systemd.early_core_pattern")) {
347
348 if (proc_cmdline_value_missing(key, value))
349 return 0;
350
351 if (path_is_absolute(value))
352 (void) parse_path_argument_and_warn(value, false, &arg_early_core_pattern);
353 else
354 log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value);
355
1d84ad94 356 } else if (proc_cmdline_key_streq(key, "systemd.crash_chvt")) {
b9e74c39 357
1d84ad94
LP		 358 if (!value)
359 arg_crash_chvt = 0; /* turn on */
5e1ee764 360 else {
a07a7324 361 r = parse_crash_chvt(value, &arg_crash_chvt);
5e1ee764
YW		 362 if (r < 0)
363 log_warning_errno(r, "Failed to parse crash chvt switch %s, ignoring: %m", value);
364 }
b9e74c39 365
1d84ad94 366 } else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) {
4fc935ca 367
1d84ad94 368 r = value ? parse_boolean(value) : true;
fb472900 369 if (r < 0)
5e1ee764 370 log_warning_errno(r, "Failed to parse crash shell switch %s, ignoring: %m", value);
4fc935ca 371 else
fa0f4d8a 372 arg_crash_shell = r;
5e7ee61c 373
1d84ad94 374 } else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) {
5e7ee61c 375
1d84ad94 376 r = value ? parse_boolean(value) : true;
b9e74c39 377 if (r < 0)
5e1ee764 378 log_warning_errno(r, "Failed to parse crash reboot switch %s, ignoring: %m", value);
5e7ee61c 379 else
b9e74c39 380 arg_crash_reboot = r;
5e7ee61c 381
1d84ad94
LP		 382 } else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) {
383 char *s;
7d5ceb64 384
1d84ad94 385 r = parse_confirm_spawn(value, &s);
059cb385 386 if (r < 0)
5e1ee764
YW		 387 log_warning_errno(r, "Failed to parse confirm_spawn switch %s, ignoring: %m", value);
388 else
389 free_and_replace(arg_confirm_spawn, s);
601f6a1e 390
2a12e32e
JK		 391 } else if (proc_cmdline_key_streq(key, "systemd.service_watchdogs")) {
392
393 r = value ? parse_boolean(value) : true;
394 if (r < 0)
5e1ee764 395 log_warning_errno(r, "Failed to parse service watchdog switch %s, ignoring: %m", value);
2a12e32e
JK		 396 else
397 arg_service_watchdogs = r;
398
1d84ad94 399 } else if (proc_cmdline_key_streq(key, "systemd.show_status")) {
9e58ff9c 400
1d84ad94
LP		 401 if (value) {
402 r = parse_show_status(value, &arg_show_status);
403 if (r < 0)
5e1ee764 404 log_warning_errno(r, "Failed to parse show status switch %s, ignoring: %m", value);
1d84ad94
LP		 405 } else
406 arg_show_status = SHOW_STATUS_YES;
059cb385 407
36cf4507
ZJS		 408 } else if (proc_cmdline_key_streq(key, "systemd.status_unit_format")) {
409
410 if (proc_cmdline_value_missing(key, value))
411 return 0;
412
413 r = status_unit_format_from_string(value);
414 if (r < 0)
415 log_warning_errno(r, "Failed to parse %s=%s, ignoring: %m", key, value);
416 else
417 arg_status_unit_format = r;
418
1d84ad94
LP		 419 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) {
420
421 if (proc_cmdline_value_missing(key, value))
422 return 0;
0a494f1f 423
059cb385 424 r = exec_output_from_string(value);
fb472900 425 if (r < 0)
5e1ee764 426 log_warning_errno(r, "Failed to parse default standard output switch %s, ignoring: %m", value);
0a494f1f
LP		 427 else
428 arg_default_std_output = r;
0a494f1f 429
1d84ad94
LP		 430 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) {
431
432 if (proc_cmdline_value_missing(key, value))
433 return 0;
059cb385
LP		 434
435 r = exec_output_from_string(value);
fb472900 436 if (r < 0)
5e1ee764 437 log_warning_errno(r, "Failed to parse default standard error switch %s, ignoring: %m", value);
0a494f1f
LP		 438 else
439 arg_default_std_error = r;
9e7c5357 440
1d84ad94
LP		 441 } else if (streq(key, "systemd.setenv")) {
442
443 if (proc_cmdline_value_missing(key, value))
444 return 0;
059cb385
LP		 445
446 if (env_assignment_is_valid(value)) {
e21fea24
KS		 447 char **env;
448
059cb385 449 env = strv_env_set(arg_default_environment, value);
1d84ad94
LP		 450 if (!env)
451 return log_oom();
452
453 arg_default_environment = env;
e21fea24 454 } else
059cb385 455 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
9e58ff9c 456
1d84ad94
LP		 457 } else if (proc_cmdline_key_streq(key, "systemd.machine_id")) {
458
459 if (proc_cmdline_value_missing(key, value))
460 return 0;
461
462 r = set_machine_id(value);
463 if (r < 0)
5e1ee764 464 log_warning_errno(r, "MachineID '%s' is not valid, ignoring: %m", value);
ee48dbd5 465
1d84ad94
LP		 466 } else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) {
467
468 if (proc_cmdline_value_missing(key, value))
469 return 0;
470
471 r = parse_sec(value, &arg_default_timeout_start_usec);
472 if (r < 0)
5e1ee764 473 log_warning_errno(r, "Failed to parse default start timeout '%s', ignoring: %m", value);
1d84ad94
LP		 474
475 if (arg_default_timeout_start_usec <= 0)
476 arg_default_timeout_start_usec = USEC_INFINITY;
ee48dbd5 477
8a2c1fbf
EJ		 478 } else if (proc_cmdline_key_streq(key, "systemd.watchdog_device")) {
479
480 if (proc_cmdline_value_missing(key, value))
481 return 0;
482
c6885f5f 483 (void) parse_path_argument_and_warn(value, false, &arg_watchdog_device);
8a2c1fbf 484
059cb385 485 } else if (streq(key, "quiet") && !value) {
d7b15e0a 486
7a293242 487 if (arg_show_status == _SHOW_STATUS_INVALID)
d450b6f2 488 arg_show_status = SHOW_STATUS_AUTO;
059cb385
LP		 489
490 } else if (streq(key, "debug") && !value) {
d7b15e0a 491
1de1c9c3
LP		 492 /* Note that log_parse_environment() handles 'debug'
493 * too, and sets the log level to LOG_DEBUG. */
d7b15e0a 494
75f86906 495 if (detect_container() > 0)
b2103dcc 496 log_set_target(LOG_TARGET_CONSOLE);
059cb385 497
dcd61450 498 } else if (!value) {
e2c9a131 499 const char *target;
f170852a
LP		 500
501 /* SysV compatibility */
e2c9a131
EV		 502 target = runlevel_to_target(key);
503 if (target)
504 return free_and_strdup(&arg_default_unit, target);
f170852a
LP		 505 }
506
507 return 0;
508}
509
e8e581bf
ZJS		 510#define DEFINE_SETTER(name, func, descr) \
511 static int name(const char *unit, \
512 const char *filename, \
513 unsigned line, \
514 const char *section, \
71a61510 515 unsigned section_line, \
e8e581bf
ZJS		 516 const char *lvalue, \
517 int ltype, \
518 const char *rvalue, \
519 void *data, \
520 void *userdata) { \
521 \
522 int r; \
523 \
524 assert(filename); \
525 assert(lvalue); \
526 assert(rvalue); \
527 \
528 r = func(rvalue); \
529 if (r < 0) \
d1cefe0a
LP		 530 log_syntax(unit, LOG_ERR, filename, line, r, \
531 "Invalid " descr "'%s': %m", \
532 rvalue); \
e8e581bf
ZJS		 533 \
534 return 0; \
535 }
487393e9 536
a6ecbf83
FB		 537DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level");
538DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target");
539DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" );
540DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location");
487393e9 541
a61d6874
ZJS		 542static int config_parse_default_timeout_abort(
543 const char *unit,
544 const char *filename,
545 unsigned line,
546 const char *section,
547 unsigned section_line,
548 const char *lvalue,
549 int ltype,
550 const char *rvalue,
551 void *data,
552 void *userdata) {
553 int r;
554
555 r = config_parse_timeout_abort(unit, filename, line, section, section_line, lvalue, ltype, rvalue,
556 &arg_default_timeout_abort_usec, userdata);
557 if (r >= 0)
558 arg_default_timeout_abort_set = r;
559 return 0;
560}
487393e9 561
a61d6874 562static int parse_config_file(void) {
f975e971 563 const ConfigTableItem items[] = {
a61d6874
ZJS		 564 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
565 { "Manager", "LogTarget", config_parse_target, 0, NULL },
566 { "Manager", "LogColor", config_parse_color, 0, NULL },
567 { "Manager", "LogLocation", config_parse_location, 0, NULL },
568 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
569 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, &arg_crash_chvt },
570 { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, &arg_crash_chvt },
571 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
572 { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
573 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
574 { "Manager", "StatusUnitFormat", config_parse_status_unit_format, 0, &arg_status_unit_format },
575 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, &arg_cpu_affinity },
576 { "Manager", "NUMAPolicy", config_parse_numa_policy, 0, &arg_numa_policy.type },
577 { "Manager", "NUMAMask", config_parse_numa_mask, 0, &arg_numa_policy },
578 { "Manager", "JoinControllers", config_parse_warn_compat, DISABLED_CONFIGURATION, NULL },
579 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
580 { "Manager", "RebootWatchdogSec", config_parse_sec, 0, &arg_reboot_watchdog },
581 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_reboot_watchdog }, /* obsolete alias */
582 { "Manager", "KExecWatchdogSec", config_parse_sec, 0, &arg_kexec_watchdog },
583 { "Manager", "WatchdogDevice", config_parse_path, 0, &arg_watchdog_device },
584 { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
585 { "Manager", "NoNewPrivileges", config_parse_bool, 0, &arg_no_new_privs },
349cc4a5 586#if HAVE_SECCOMP
a61d6874 587 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
89fffa27 588#endif
a61d6874
ZJS		 589 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
590 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
591 { "Manager", "DefaultStandardOutput", config_parse_output_restricted, 0, &arg_default_std_output },
592 { "Manager", "DefaultStandardError", config_parse_output_restricted, 0, &arg_default_std_error },
593 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
594 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
595 { "Manager", "DefaultTimeoutAbortSec", config_parse_default_timeout_abort, 0, NULL },
596 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
597 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */
598 { "Manager", "DefaultStartLimitIntervalSec", config_parse_sec, 0, &arg_default_start_limit_interval },
599 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
600 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
601 { "Manager", "DefaultLimitCPU", config_parse_rlimit, RLIMIT_CPU, arg_default_rlimit },
602 { "Manager", "DefaultLimitFSIZE", config_parse_rlimit, RLIMIT_FSIZE, arg_default_rlimit },
603 { "Manager", "DefaultLimitDATA", config_parse_rlimit, RLIMIT_DATA, arg_default_rlimit },
604 { "Manager", "DefaultLimitSTACK", config_parse_rlimit, RLIMIT_STACK, arg_default_rlimit },
605 { "Manager", "DefaultLimitCORE", config_parse_rlimit, RLIMIT_CORE, arg_default_rlimit },
606 { "Manager", "DefaultLimitRSS", config_parse_rlimit, RLIMIT_RSS, arg_default_rlimit },
607 { "Manager", "DefaultLimitNOFILE", config_parse_rlimit, RLIMIT_NOFILE, arg_default_rlimit },
608 { "Manager", "DefaultLimitAS", config_parse_rlimit, RLIMIT_AS, arg_default_rlimit },
609 { "Manager", "DefaultLimitNPROC", config_parse_rlimit, RLIMIT_NPROC, arg_default_rlimit },
610 { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit, RLIMIT_MEMLOCK, arg_default_rlimit },
611 { "Manager", "DefaultLimitLOCKS", config_parse_rlimit, RLIMIT_LOCKS, arg_default_rlimit },
612 { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit, RLIMIT_SIGPENDING, arg_default_rlimit },
613 { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit, RLIMIT_MSGQUEUE, arg_default_rlimit },
614 { "Manager", "DefaultLimitNICE", config_parse_rlimit, RLIMIT_NICE, arg_default_rlimit },
615 { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit, RLIMIT_RTPRIO, arg_default_rlimit },
616 { "Manager", "DefaultLimitRTTIME", config_parse_rlimit, RLIMIT_RTTIME, arg_default_rlimit },
617 { "Manager", "DefaultCPUAccounting", config_parse_tristate, 0, &arg_default_cpu_accounting },
618 { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting },
619 { "Manager", "DefaultIPAccounting", config_parse_bool, 0, &arg_default_ip_accounting },
620 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
621 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
622 { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
623 { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max },
624 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action, 0, &arg_cad_burst_action },
625 { "Manager", "DefaultOOMPolicy", config_parse_oom_policy, 0, &arg_default_oom_policy },
d3b1c508 626 {}
487393e9
LP		 627 };
628
1b907b5c 629 const char *fn, *conf_dirs_nulstr;
487393e9 630
463d0d15 631 fn = arg_system ?
75eb6154
LP		 632 PKGSYSCONFDIR "/system.conf" :
633 PKGSYSCONFDIR "/user.conf";
634
463d0d15 635 conf_dirs_nulstr = arg_system ?
75eb6154
LP		 636 CONF_PATHS_NULSTR("systemd/system.conf.d") :
637 CONF_PATHS_NULSTR("systemd/user.conf.d");
638
bcde742e 639 (void) config_parse_many_nulstr(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, CONFIG_PARSE_WARN, NULL);
36c16a7c
LP		 640
641 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
642 * like everywhere else. */
643 if (arg_default_timeout_start_usec <= 0)
644 arg_default_timeout_start_usec = USEC_INFINITY;
645 if (arg_default_timeout_stop_usec <= 0)
646 arg_default_timeout_stop_usec = USEC_INFINITY;
487393e9 647
487393e9
LP		 648 return 0;
649}
650
85cb4151 651static void set_manager_defaults(Manager *m) {
06af2a04
TB		 652
653 assert(m);
654
5b65ae15
LP		 655 /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
656 * affect the manager itself, but are just what newly allocated units will have set if they haven't set
657 * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
658
06af2a04
TB		 659 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
660 m->default_std_output = arg_default_std_output;
661 m->default_std_error = arg_default_std_error;
662 m->default_timeout_start_usec = arg_default_timeout_start_usec;
663 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
dc653bf4
JK		 664 m->default_timeout_abort_usec = arg_default_timeout_abort_usec;
665 m->default_timeout_abort_set = arg_default_timeout_abort_set;
06af2a04
TB		 666 m->default_restart_usec = arg_default_restart_usec;
667 m->default_start_limit_interval = arg_default_start_limit_interval;
668 m->default_start_limit_burst = arg_default_start_limit_burst;
a88c5b8a
CD		 669
670 /* On 4.15+ with unified hierarchy, CPU accounting is essentially free as it doesn't require the CPU
671 * controller to be enabled, so the default is to enable it unless we got told otherwise. */
672 if (arg_default_cpu_accounting >= 0)
673 m->default_cpu_accounting = arg_default_cpu_accounting;
674 else
675 m->default_cpu_accounting = cpu_accounting_is_cheap();
676
13c31542 677 m->default_io_accounting = arg_default_io_accounting;
377bfd2d 678 m->default_ip_accounting = arg_default_ip_accounting;
06af2a04
TB		 679 m->default_blockio_accounting = arg_default_blockio_accounting;
680 m->default_memory_accounting = arg_default_memory_accounting;
03a7b521 681 m->default_tasks_accounting = arg_default_tasks_accounting;
0af20ea2 682 m->default_tasks_max = arg_default_tasks_max;
afcfaa69 683 m->default_oom_policy = arg_default_oom_policy;
06af2a04 684
79a224c4
LP		 685 (void) manager_set_default_rlimits(m, arg_default_rlimit);
686
687 (void) manager_default_environment(m);
688 (void) manager_transient_environment_add(m, arg_default_environment);
06af2a04
TB		 689}
690
7b46fc6a
LP		 691static void set_manager_settings(Manager *m) {
692
693 assert(m);
694
5b65ae15
LP		 695 /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
696 * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
697
7b46fc6a 698 m->confirm_spawn = arg_confirm_spawn;
2a12e32e 699 m->service_watchdogs = arg_service_watchdogs;
7b46fc6a 700 m->runtime_watchdog = arg_runtime_watchdog;
65224c1d 701 m->reboot_watchdog = arg_reboot_watchdog;
acafd7d8 702 m->kexec_watchdog = arg_kexec_watchdog;
7b46fc6a
LP		 703 m->cad_burst_action = arg_cad_burst_action;
704
705 manager_set_show_status(m, arg_show_status);
36cf4507 706 m->status_unit_format = arg_status_unit_format;
7b46fc6a
LP		 707}
708
f170852a 709static int parse_argv(int argc, char *argv[]) {
f170852a
LP		 710 enum {
711 ARG_LOG_LEVEL = 0x100,
712 ARG_LOG_TARGET,
bbe63281
LP		 713 ARG_LOG_COLOR,
714 ARG_LOG_LOCATION,
2f198e2f 715 ARG_UNIT,
edb9aaa8 716 ARG_SYSTEM,
af2d49f7 717 ARG_USER,
e537352b 718 ARG_TEST,
b87c2aa6 719 ARG_NO_PAGER,
9ba0bc4e 720 ARG_VERSION,
80876c20 721 ARG_DUMP_CONFIGURATION_ITEMS,
bbc1acab 722 ARG_DUMP_BUS_PROPERTIES,
9e58ff9c 723 ARG_DUMP_CORE,
b9e74c39 724 ARG_CRASH_CHVT,
9e58ff9c 725 ARG_CRASH_SHELL,
b9e74c39 726 ARG_CRASH_REBOOT,
a16e1123 727 ARG_CONFIRM_SPAWN,
9e58ff9c 728 ARG_SHOW_STATUS,
4288f619 729 ARG_DESERIALIZE,
2660882b 730 ARG_SWITCHED_ROOT,
0a494f1f 731 ARG_DEFAULT_STD_OUTPUT,
ee48dbd5 732 ARG_DEFAULT_STD_ERROR,
2a12e32e
JK		 733 ARG_MACHINE_ID,
734 ARG_SERVICE_WATCHDOGS,
f170852a
LP		 735 };
736
737 static const struct option options[] = {
a16e1123
LP		 738 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
739 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
bbe63281
LP		 740 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
741 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
2f198e2f 742 { "unit", required_argument, NULL, ARG_UNIT },
edb9aaa8 743 { "system", no_argument, NULL, ARG_SYSTEM },
af2d49f7 744 { "user", no_argument, NULL, ARG_USER },
a16e1123 745 { "test", no_argument, NULL, ARG_TEST },
b87c2aa6 746 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
a16e1123 747 { "help", no_argument, NULL, 'h' },
9ba0bc4e 748 { "version", no_argument, NULL, ARG_VERSION },
a16e1123 749 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
bbc1acab 750 { "dump-bus-properties", no_argument, NULL, ARG_DUMP_BUS_PROPERTIES },
a5d87bf0 751 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
b9e74c39 752 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
a5d87bf0 753 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
b9e74c39 754 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
a5d87bf0 755 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
6e98720f 756 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
a16e1123 757 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
2660882b 758 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
0a494f1f
LP		 759 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
760 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
ee48dbd5 761 { "machine-id", required_argument, NULL, ARG_MACHINE_ID },
2a12e32e 762 { "service-watchdogs", required_argument, NULL, ARG_SERVICE_WATCHDOGS },
fb472900 763 {}
f170852a
LP		 764 };
765
766 int c, r;
767
768 assert(argc >= 1);
769 assert(argv);
770
df0ff127 771 if (getpid_cached() == 1)
b770165a
LP		 772 opterr = 0;
773
099663ff 774 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
f170852a
LP		 775
776 switch (c) {
777
778 case ARG_LOG_LEVEL:
fb472900 779 r = log_set_max_level_from_string(optarg);
2b5107e1
ZJS		 780 if (r < 0)
781 return log_error_errno(r, "Failed to parse log level \"%s\": %m", optarg);
f170852a
LP		 782
783 break;
784
785 case ARG_LOG_TARGET:
fb472900 786 r = log_set_target_from_string(optarg);
2b5107e1
ZJS		 787 if (r < 0)
788 return log_error_errno(r, "Failed to parse log target \"%s\": %m", optarg);
f170852a
LP		 789
790 break;
791
bbe63281
LP		 792 case ARG_LOG_COLOR:
793
d0b170c8 794 if (optarg) {
fb472900 795 r = log_show_color_from_string(optarg);
2b5107e1
ZJS		 796 if (r < 0)
797 return log_error_errno(r, "Failed to parse log color setting \"%s\": %m",
798 optarg);
d0b170c8
LP		 799 } else
800 log_show_color(true);
bbe63281
LP		 801
802 break;
803
804 case ARG_LOG_LOCATION:
d0b170c8 805 if (optarg) {
fb472900 806 r = log_show_location_from_string(optarg);
2b5107e1
ZJS		 807 if (r < 0)
808 return log_error_errno(r, "Failed to parse log location setting \"%s\": %m",
809 optarg);
d0b170c8
LP		 810 } else
811 log_show_location(true);
bbe63281
LP		 812
813 break;
814
0a494f1f 815 case ARG_DEFAULT_STD_OUTPUT:
fb472900 816 r = exec_output_from_string(optarg);
2b5107e1
ZJS		 817 if (r < 0)
818 return log_error_errno(r, "Failed to parse default standard output setting \"%s\": %m",
819 optarg);
820 arg_default_std_output = r;
0a494f1f
LP		 821 break;
822
823 case ARG_DEFAULT_STD_ERROR:
fb472900 824 r = exec_output_from_string(optarg);
2b5107e1
ZJS		 825 if (r < 0)
826 return log_error_errno(r, "Failed to parse default standard error output setting \"%s\": %m",
827 optarg);
828 arg_default_std_error = r;
0a494f1f
LP		 829 break;
830
2f198e2f 831 case ARG_UNIT:
e6e242ad 832 r = free_and_strdup(&arg_default_unit, optarg);
23bbb0de 833 if (r < 0)
2b5107e1 834 return log_error_errno(r, "Failed to set default unit \"%s\": %m", optarg);
f170852a
LP		 835
836 break;
837
edb9aaa8 838 case ARG_SYSTEM:
463d0d15 839 arg_system = true;
edb9aaa8 840 break;
a5dab5ce 841
af2d49f7 842 case ARG_USER:
463d0d15 843 arg_system = false;
a5dab5ce 844 break;
a5dab5ce 845
e965d56d 846 case ARG_TEST:
fa0f4d8a 847 arg_action = ACTION_TEST;
b87c2aa6
ZJS		 848 break;
849
850 case ARG_NO_PAGER:
0221d68a 851 arg_pager_flags |= PAGER_DISABLE;
e965d56d
LP		 852 break;
853
9ba0bc4e
ZJS		 854 case ARG_VERSION:
855 arg_action = ACTION_VERSION;
856 break;
857
e537352b 858 case ARG_DUMP_CONFIGURATION_ITEMS:
fa0f4d8a 859 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
e537352b
LP		 860 break;
861
bbc1acab
YW		 862 case ARG_DUMP_BUS_PROPERTIES:
863 arg_action = ACTION_DUMP_BUS_PROPERTIES;
864 break;
865
9e58ff9c 866 case ARG_DUMP_CORE:
b9e74c39
LP		 867 if (!optarg)
868 arg_dump_core = true;
869 else {
870 r = parse_boolean(optarg);
871 if (r < 0)
2b5107e1
ZJS		 872 return log_error_errno(r, "Failed to parse dump core boolean: \"%s\": %m",
873 optarg);
b9e74c39 874 arg_dump_core = r;
a5d87bf0 875 }
b9e74c39
LP		 876 break;
877
878 case ARG_CRASH_CHVT:
a07a7324 879 r = parse_crash_chvt(optarg, &arg_crash_chvt);
b9e74c39 880 if (r < 0)
2b5107e1
ZJS		 881 return log_error_errno(r, "Failed to parse crash virtual terminal index: \"%s\": %m",
882 optarg);
9e58ff9c
LP		 883 break;
884
885 case ARG_CRASH_SHELL:
b9e74c39
LP		 886 if (!optarg)
887 arg_crash_shell = true;
888 else {
889 r = parse_boolean(optarg);
890 if (r < 0)
2b5107e1
ZJS		 891 return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
892 optarg);
b9e74c39
LP		 893 arg_crash_shell = r;
894 }
895 break;
896
897 case ARG_CRASH_REBOOT:
898 if (!optarg)
899 arg_crash_reboot = true;
900 else {
901 r = parse_boolean(optarg);
902 if (r < 0)
2b5107e1
ZJS		 903 return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
904 optarg);
b9e74c39 905 arg_crash_reboot = r;
a5d87bf0 906 }
9e58ff9c
LP		 907 break;
908
80876c20 909 case ARG_CONFIRM_SPAWN:
7d5ceb64
FB		 910 arg_confirm_spawn = mfree(arg_confirm_spawn);
911
912 r = parse_confirm_spawn(optarg, &arg_confirm_spawn);
913 if (r < 0)
2b5107e1
ZJS		 914 return log_error_errno(r, "Failed to parse confirm spawn option: \"%s\": %m",
915 optarg);
80876c20
LP		 916 break;
917
2a12e32e
JK		 918 case ARG_SERVICE_WATCHDOGS:
919 r = parse_boolean(optarg);
920 if (r < 0)
2b5107e1
ZJS		 921 return log_error_errno(r, "Failed to parse service watchdogs boolean: \"%s\": %m",
922 optarg);
2a12e32e
JK		 923 arg_service_watchdogs = r;
924 break;
925
9e58ff9c 926 case ARG_SHOW_STATUS:
d450b6f2
ZJS		 927 if (optarg) {
928 r = parse_show_status(optarg, &arg_show_status);
ac7ec288 929 if (r < 0)
2b5107e1
ZJS		 930 return log_error_errno(r, "Failed to parse show status boolean: \"%s\": %m",
931 optarg);
d450b6f2
ZJS		 932 } else
933 arg_show_status = SHOW_STATUS_YES;
6e98720f 934 break;
a5d87bf0 935
a16e1123
LP		 936 case ARG_DESERIALIZE: {
937 int fd;
938 FILE *f;
939
01e10de3 940 r = safe_atoi(optarg, &fd);
2b5107e1
ZJS		 941 if (r < 0)
942 log_error_errno(r, "Failed to parse deserialize option \"%s\": %m", optarg);
baaa35ad
ZJS		 943 if (fd < 0)
944 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
945 "Invalid deserialize fd: %d",
946 fd);
a16e1123 947
b9e74c39 948 (void) fd_cloexec(fd, true);
01e10de3
LP		 949
950 f = fdopen(fd, "r");
4a62c710 951 if (!f)
2b5107e1 952 return log_error_errno(errno, "Failed to open serialization fd %d: %m", fd);
a16e1123 953
74ca738f 954 safe_fclose(arg_serialization);
d3b1c508 955 arg_serialization = f;
a16e1123
LP		 956
957 break;
958 }
959
2660882b 960 case ARG_SWITCHED_ROOT:
bf4df7c3 961 arg_switched_root = true;
d03bc1b8
HH		 962 break;
963
ee48dbd5
NC		 964 case ARG_MACHINE_ID:
965 r = set_machine_id(optarg);
54500613 966 if (r < 0)
2b5107e1 967 return log_error_errno(r, "MachineID '%s' is not valid: %m", optarg);
ee48dbd5
NC		 968 break;
969
f170852a 970 case 'h':
fa0f4d8a 971 arg_action = ACTION_HELP;
f170852a
LP		 972 break;
973
1d2e23ab
LP		 974 case 'D':
975 log_set_max_level(LOG_DEBUG);
976 break;
977
099663ff
LP		 978 case 'b':
979 case 's':
980 case 'z':
981 /* Just to eat away the sysvinit kernel
982 * cmdline args without getopt() error
983 * messages that we'll parse in
984 * parse_proc_cmdline_word() or ignore. */
f170852a 985
099663ff 986 case '?':
df0ff127 987 if (getpid_cached() != 1)
099663ff 988 return -EINVAL;
601185b4
ZJS		 989 else
990 return 0;
099663ff 991
601185b4
ZJS		 992 default:
993 assert_not_reached("Unhandled option code.");
f170852a
LP		 994 }
995
df0ff127 996 if (optind < argc && getpid_cached() != 1) {
d821e6d6
LP		 997 /* Hmm, when we aren't run as init system
998 * let's complain about excess arguments */
999
baaa35ad
ZJS		 1000 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1001 "Excess arguments.");
d821e6d6
LP		 1002 }
1003
f170852a
LP		 1004 return 0;
1005}
1006
1007static int help(void) {
37ec0fdd
LP		 1008 _cleanup_free_ char *link = NULL;
1009 int r;
1010
1011 r = terminal_urlify_man("systemd", "1", &link);
1012 if (r < 0)
1013 return log_oom();
f170852a 1014
2e33c433 1015 printf("%s [OPTIONS...]\n\n"
af2d49f7 1016 "Starts up and maintains the system or user services.\n\n"
e537352b 1017 " -h --help Show this help\n"
cb4069d9 1018 " --version Show version\n"
cd69e88b
LP		 1019 " --test Determine initial transaction, dump it and exit\n"
1020 " --system In combination with --test: operate as system service manager\n"
1021 " --user In combination with --test: operate as per-user service manager\n"
b87c2aa6 1022 " --no-pager Do not pipe output into a pager\n"
80876c20 1023 " --dump-configuration-items Dump understood unit configuration items\n"
bbc1acab 1024 " --dump-bus-properties Dump exposed bus properties\n"
9e58ff9c 1025 " --unit=UNIT Set default unit\n"
b9e74c39
LP		 1026 " --dump-core[=BOOL] Dump core on crash\n"
1027 " --crash-vt=NR Change to specified VT on crash\n"
1028 " --crash-reboot[=BOOL] Reboot on crash\n"
1029 " --crash-shell[=BOOL] Run shell on crash\n"
1030 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1031 " --show-status[=BOOL] Show status updates on the console during bootup\n"
c1dc6153 1032 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
9e58ff9c 1033 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
b9e74c39
LP		 1034 " --log-color[=BOOL] Highlight important log messages\n"
1035 " --log-location[=BOOL] Include code location in log messages\n"
0a494f1f 1036 " --default-standard-output= Set default standard output for services\n"
37ec0fdd
LP		 1037 " --default-standard-error= Set default standard error output for services\n"
1038 "\nSee the %s for details.\n"
1039 , program_invocation_short_name
1040 , link
1041 );
f170852a
LP		 1042
1043 return 0;
1044}
1045
2cc856ac
LP		 1046static int prepare_reexecute(
1047 Manager *m,
1048 FILE **ret_f,
1049 FDSet **ret_fds,
1050 bool switching_root) {
1051
48b90859
LP		 1052 _cleanup_fdset_free_ FDSet *fds = NULL;
1053 _cleanup_fclose_ FILE *f = NULL;
a16e1123
LP		 1054 int r;
1055
1056 assert(m);
2cc856ac
LP		 1057 assert(ret_f);
1058 assert(ret_fds);
a16e1123 1059
6b78f9b4 1060 r = manager_open_serialization(m, &f);
48b90859
LP		 1061 if (r < 0)
1062 return log_error_errno(r, "Failed to create serialization file: %m");
a16e1123 1063
71445ae7 1064 /* Make sure nothing is really destructed when we shut down */
313cefa1 1065 m->n_reloading++;
718db961 1066 bus_manager_send_reloading(m, true);
71445ae7 1067
6b78f9b4 1068 fds = fdset_new();
48b90859
LP		 1069 if (!fds)
1070 return log_oom();
a16e1123 1071
b3680f49 1072 r = manager_serialize(m, f, fds, switching_root);
48b90859 1073 if (r < 0)
d68c645b 1074 return r;
a16e1123 1075
48b90859
LP		 1076 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1077 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
a16e1123 1078
6b78f9b4 1079 r = fd_cloexec(fileno(f), false);
48b90859
LP		 1080 if (r < 0)
1081 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
a16e1123 1082
6b78f9b4 1083 r = fdset_cloexec(fds, false);
48b90859
LP		 1084 if (r < 0)
1085 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
a16e1123 1086
2cc856ac
LP		 1087 *ret_f = TAKE_PTR(f);
1088 *ret_fds = TAKE_PTR(fds);
a16e1123 1089
48b90859 1090 return 0;
a16e1123
LP		 1091}
1092
a8b627aa
LP		 1093static void bump_file_max_and_nr_open(void) {
1094
1095 /* Let's bump fs.file-max and fs.nr_open to their respective maximums. On current kernels large numbers of file
1096 * descriptors are no longer a performance problem and their memory is properly tracked by memcg, thus counting
1097 * them and limiting them in another two layers of limits is unnecessary and just complicates things. This
1098 * function hence turns off 2 of the 4 levels of limits on file descriptors, and makes RLIMIT_NOLIMIT (soft +
1099 * hard) the only ones that really matter. */
1100
1101#if BUMP_PROC_SYS_FS_FILE_MAX || BUMP_PROC_SYS_FS_NR_OPEN
a8b627aa
LP		 1102 int r;
1103#endif
1104
1105#if BUMP_PROC_SYS_FS_FILE_MAX
6e2f7894
LP		 1106 /* The maximum the kernel allows for this since 5.2 is LONG_MAX, use that. (Previously thing where
1107 * different but the operation would fail silently.) */
56e8419a 1108 r = sysctl_writef("fs/file-max", "%li\n", LONG_MAX);
a8b627aa
LP		 1109 if (r < 0)
1110 log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.file-max, ignoring: %m");
1111#endif
1112
a8b627aa
LP		 1113#if BUMP_PROC_SYS_FS_NR_OPEN
1114 int v = INT_MAX;
1115
1116 /* Arg! The kernel enforces maximum and minimum values on the fs.nr_open, but we don't really know what they
1117 * are. The expression by which the maximum is determined is dependent on the architecture, and is something we
1118 * don't really want to copy to userspace, as it is dependent on implementation details of the kernel. Since
1119 * the kernel doesn't expose the maximum value to us, we can only try and hope. Hence, let's start with
1120 * INT_MAX, and then keep halving the value until we find one that works. Ugly? Yes, absolutely, but kernel
1121 * APIs are kernel APIs, so what do can we do... 🤯 */
1122
1123 for (;;) {
1124 int k;
1125
1126 v &= ~(__SIZEOF_POINTER__ - 1); /* Round down to next multiple of the pointer size */
1127 if (v < 1024) {
1128 log_warning("Can't bump fs.nr_open, value too small.");
1129 break;
1130 }
1131
1132 k = read_nr_open();
1133 if (k < 0) {
1134 log_error_errno(k, "Failed to read fs.nr_open: %m");
1135 break;
1136 }
1137 if (k >= v) { /* Already larger */
1138 log_debug("Skipping bump, value is already larger.");
1139 break;
1140 }
1141
56e8419a 1142 r = sysctl_writef("fs/nr_open", "%i\n", v);
a8b627aa
LP		 1143 if (r == -EINVAL) {
1144 log_debug("Couldn't write fs.nr_open as %i, halving it.", v);
1145 v /= 2;
1146 continue;
1147 }
1148 if (r < 0) {
1149 log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.nr_open, ignoring: %m");
1150 break;
1151 }
1152
1153 log_debug("Successfully bumped fs.nr_open to %i", v);
1154 break;
1155 }
1156#endif
1157}
1158
4096d6f5 1159static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
cda7faa9 1160 struct rlimit new_rlimit;
9264cc39 1161 int r, nr;
4096d6f5 1162
52d62075
LP		 1163 /* Get the underlying absolute limit the kernel enforces */
1164 nr = read_nr_open();
1165
cda7faa9
LP		 1166 /* Calculate the new limits to use for us. Never lower from what we inherited. */
1167 new_rlimit = (struct rlimit) {
1168 .rlim_cur = MAX((rlim_t) nr, saved_rlimit->rlim_cur),
1169 .rlim_max = MAX((rlim_t) nr, saved_rlimit->rlim_max),
1170 };
1171
1172 /* Shortcut if nothing changes. */
1173 if (saved_rlimit->rlim_max >= new_rlimit.rlim_max &&
1174 saved_rlimit->rlim_cur >= new_rlimit.rlim_cur) {
1175 log_debug("RLIMIT_NOFILE is already as high or higher than we need it, not bumping.");
1176 return 0;
1177 }
1178
52d62075
LP		 1179 /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for
1180 * both hard and soft. */
cda7faa9 1181 r = setrlimit_closest(RLIMIT_NOFILE, &new_rlimit);
23bbb0de 1182 if (r < 0)
3ce40911 1183 return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m");
4096d6f5
LP		 1184
1185 return 0;
1186}
1187
fb3ae275 1188static int bump_rlimit_memlock(struct rlimit *saved_rlimit) {
cda7faa9 1189 struct rlimit new_rlimit;
fb3ae275
LP		 1190 int r;
1191
a17c1712
LP		 1192 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK which should
1193 * normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's bump
1194 * the value high enough for our user. */
fb3ae275 1195
cda7faa9
LP		 1196 /* Using MAX() on resource limits only is safe if RLIM_INFINITY is > 0. POSIX declares that rlim_t
1197 * must be unsigned, hence this is a given, but let's make this clear here. */
1198 assert_cc(RLIM_INFINITY > 0);
1199
1200 new_rlimit = (struct rlimit) {
1201 .rlim_cur = MAX(HIGH_RLIMIT_MEMLOCK, saved_rlimit->rlim_cur),
1202 .rlim_max = MAX(HIGH_RLIMIT_MEMLOCK, saved_rlimit->rlim_max),
1203 };
1204
1205 if (saved_rlimit->rlim_max >= new_rlimit.rlim_cur &&
1206 saved_rlimit->rlim_cur >= new_rlimit.rlim_max) {
1207 log_debug("RLIMIT_MEMLOCK is already as high or higher than we need it, not bumping.");
1208 return 0;
1209 }
1210
1211 r = setrlimit_closest(RLIMIT_MEMLOCK, &new_rlimit);
fb3ae275
LP		 1212 if (r < 0)
1213 return log_warning_errno(r, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1214
1215 return 0;
1216}
1217
80758717 1218static void test_usr(void) {
80758717 1219
796ac4c1 1220 /* Check that /usr is either on the same file system as / or mounted already. */
80758717 1221
871c44a7
LP		 1222 if (dir_is_empty("/usr") <= 0)
1223 return;
1224
8b173b5e 1225 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
871c44a7
LP		 1226 "Some things will probably break (sometimes even silently) in mysterious ways. "
1227 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1228}
1229
d3b1c508 1230static int enforce_syscall_archs(Set *archs) {
349cc4a5 1231#if HAVE_SECCOMP
d3b1c508
LP		 1232 int r;
1233
83f12b27
FS		 1234 if (!is_seccomp_available())
1235 return 0;
1236
469830d1 1237 r = seccomp_restrict_archs(arg_syscall_archs);
d3b1c508 1238 if (r < 0)
469830d1 1239 return log_error_errno(r, "Failed to enforce system call architecture restrication: %m");
d3b1c508 1240#endif
469830d1 1241 return 0;
d3b1c508
LP		 1242}
1243
b6e2f329
LP		 1244static int status_welcome(void) {
1245 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1246 int r;
1247
7a293242 1248 if (IN_SET(arg_show_status, SHOW_STATUS_NO, SHOW_STATUS_AUTO))
fd8c85c6
LP		 1249 return 0;
1250
d58ad743
LP		 1251 r = parse_os_release(NULL,
1252 "PRETTY_NAME", &pretty_name,
1253 "ANSI_COLOR", &ansi_color,
1254 NULL);
1255 if (r < 0)
1256 log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r,
1257 "Failed to read os-release file, ignoring: %m");
b6e2f329 1258
dc9b5816 1259 if (log_get_show_color())
a885727a 1260 return status_printf(NULL, 0,
dc9b5816
ZJS		 1261 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1262 isempty(ansi_color) ? "1" : ansi_color,
1263 isempty(pretty_name) ? "Linux" : pretty_name);
1264 else
a885727a 1265 return status_printf(NULL, 0,
dc9b5816
ZJS		 1266 "\nWelcome to %s!\n",
1267 isempty(pretty_name) ? "Linux" : pretty_name);
b6e2f329
LP		 1268}
1269
fdd25311
LP		 1270static int write_container_id(void) {
1271 const char *c;
19854865 1272 int r;
fdd25311
LP		 1273
1274 c = getenv("container");
1275 if (isempty(c))
1276 return 0;
1277
8612da97
LP		 1278 RUN_WITH_UMASK(0022)
1279 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
19854865 1280 if (r < 0)
f1f849b0 1281 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
19854865
LP		 1282
1283 return 1;
1284}
1285
1286static int bump_unix_max_dgram_qlen(void) {
1287 _cleanup_free_ char *qlen = NULL;
1288 unsigned long v;
1289 int r;
1290
3130fca5
LP		 1291 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1292 * really really early during boot, so that it is actually applied to all our sockets, including the
1293 * $NOTIFY_SOCKET one. */
19854865
LP		 1294
1295 r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1296 if (r < 0)
875622c3 1297 return log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
19854865
LP		 1298
1299 r = safe_atolu(qlen, &v);
1300 if (r < 0)
3130fca5 1301 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen);
19854865
LP		 1302
1303 if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1304 return 0;
1305
57512c89 1306 r = write_string_filef("/proc/sys/net/unix/max_dgram_qlen", WRITE_STRING_FILE_DISABLE_BUFFER, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN);
19854865
LP		 1307 if (r < 0)
1308 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1309 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1310
1311 return 1;
fdd25311
LP		 1312}
1313
32391275
FB		 1314static int fixup_environment(void) {
1315 _cleanup_free_ char *term = NULL;
4dc63c4b 1316 const char *t;
32391275
FB		 1317 int r;
1318
43db615b
LP		 1319 /* Only fix up the environment when we are started as PID 1 */
1320 if (getpid_cached() != 1)
1321 return 0;
1322
1323 /* We expect the environment to be set correctly if run inside a container. */
84af7821
LP		 1324 if (detect_container() > 0)
1325 return 0;
1326
43db615b
LP		 1327 /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1328 * device used by the console. We try to make a better guess here since some consoles might not have support
1329 * for color mode for example.
32391275 1330 *
43db615b 1331 * However if TERM was configured through the kernel command line then leave it alone. */
1d84ad94 1332 r = proc_cmdline_get_key("TERM", 0, &term);
32391275
FB		 1333 if (r < 0)
1334 return r;
32391275 1335
4dc63c4b
LP		 1336 t = term ?: default_term_for_tty("/dev/console");
1337
1338 if (setenv("TERM", t, 1) < 0)
32391275
FB		 1339 return -errno;
1340
9d48671c
ZJS		 1341 /* The kernels sets HOME=/ for init. Let's undo this. */
1342 if (path_equal_ptr(getenv("HOME"), "/") &&
1343 unsetenv("HOME") < 0)
1344 log_warning_errno(errno, "Failed to unset $HOME: %m");
1345
32391275
FB		 1346 return 0;
1347}
1348
6808a0bc
LP		 1349static void redirect_telinit(int argc, char *argv[]) {
1350
1351 /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1352
1353#if HAVE_SYSV_COMPAT
1354 if (getpid_cached() == 1)
1355 return;
1356
1357 if (!strstr(program_invocation_short_name, "init"))
1358 return;
1359
1360 execv(SYSTEMCTL_BINARY_PATH, argv);
1361 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
a45d7127 1362 exit(EXIT_FAILURE);
6808a0bc
LP		 1363#endif
1364}
1365
4a36297c
LP		 1366static int become_shutdown(
1367 const char *shutdown_verb,
7eb35049 1368 int retval) {
4a36297c
LP		 1369
1370 char log_level[DECIMAL_STR_MAX(int) + 1],
e73c54b8
JK		 1371 exit_code[DECIMAL_STR_MAX(uint8_t) + 1],
1372 timeout[DECIMAL_STR_MAX(usec_t) + 1];
4a36297c 1373
e73c54b8 1374 const char* command_line[13] = {
4a36297c
LP		 1375 SYSTEMD_SHUTDOWN_BINARY_PATH,
1376 shutdown_verb,
e73c54b8 1377 "--timeout", timeout,
4a36297c
LP		 1378 "--log-level", log_level,
1379 "--log-target",
1380 };
1381
1382 _cleanup_strv_free_ char **env_block = NULL;
e73c54b8 1383 size_t pos = 7;
4a36297c 1384 int r;
acafd7d8 1385 usec_t watchdog_timer = 0;
4a36297c 1386
7eb35049 1387 assert(shutdown_verb);
234519ae 1388 assert(!command_line[pos]);
4a36297c
LP		 1389 env_block = strv_copy(environ);
1390
1391 xsprintf(log_level, "%d", log_get_max_level());
e73c54b8 1392 xsprintf(timeout, "%" PRI_USEC "us", arg_default_timeout_stop_usec);
4a36297c
LP		 1393
1394 switch (log_get_target()) {
1395
1396 case LOG_TARGET_KMSG:
1397 case LOG_TARGET_JOURNAL_OR_KMSG:
1398 case LOG_TARGET_SYSLOG_OR_KMSG:
1399 command_line[pos++] = "kmsg";
1400 break;
1401
1402 case LOG_TARGET_NULL:
1403 command_line[pos++] = "null";
1404 break;
1405
1406 case LOG_TARGET_CONSOLE:
1407 default:
1408 command_line[pos++] = "console";
1409 break;
1410 };
1411
1412 if (log_get_show_color())
1413 command_line[pos++] = "--log-color";
1414
1415 if (log_get_show_location())
1416 command_line[pos++] = "--log-location";
1417
1418 if (streq(shutdown_verb, "exit")) {
1419 command_line[pos++] = "--exit-code";
1420 command_line[pos++] = exit_code;
1421 xsprintf(exit_code, "%d", retval);
1422 }
1423
1424 assert(pos < ELEMENTSOF(command_line));
1425
acafd7d8 1426 if (streq(shutdown_verb, "reboot"))
65224c1d 1427 watchdog_timer = arg_reboot_watchdog;
acafd7d8
LB		 1428 else if (streq(shutdown_verb, "kexec"))
1429 watchdog_timer = arg_kexec_watchdog;
1430
1431 if (watchdog_timer > 0 && watchdog_timer != USEC_INFINITY) {
7eb35049 1432
4a36297c
LP		 1433 char *e;
1434
acafd7d8 1435 /* If we reboot or kexec let's set the shutdown
4a36297c
LP		 1436 * watchdog and tell the shutdown binary to
1437 * repeatedly ping it */
acafd7d8 1438 r = watchdog_set_timeout(&watchdog_timer);
4a36297c
LP		 1439 watchdog_close(r < 0);
1440
1441 /* Tell the binary how often to ping, ignore failure */
acafd7d8 1442 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, watchdog_timer) > 0)
8a2c1fbf
EJ		 1443 (void) strv_consume(&env_block, e);
1444
1445 if (arg_watchdog_device &&
1446 asprintf(&e, "WATCHDOG_DEVICE=%s", arg_watchdog_device) > 0)
1447 (void) strv_consume(&env_block, e);
4a36297c
LP		 1448 } else
1449 watchdog_close(true);
1450
1451 /* Avoid the creation of new processes forked by the
1452 * kernel; at this point, we will not listen to the
1453 * signals anyway */
1454 if (detect_container() <= 0)
1455 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
1456
1457 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
1458 return -errno;
1459}
1460
e839bafd
LP		 1461static void initialize_clock(void) {
1462 int r;
1463
1464 if (clock_is_localtime(NULL) > 0) {
1465 int min;
1466
1467 /*
1468 * The very first call of settimeofday() also does a time warp in the kernel.
1469 *
1470 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1471 * of maintaining the RTC and do all adjustments. This matches the behavior of Windows, which leaves
1472 * the RTC alone if the registry tells that the RTC runs in UTC.
1473 */
1474 r = clock_set_timezone(&min);
1475 if (r < 0)
1476 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1477 else
1478 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1479
1480 } else if (!in_initrd()) {
1481 /*
1482 * Do a dummy very first call to seal the kernel's time warp magic.
1483 *
1484 * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1485 * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1486 * until we reach the real system.
1487 *
1488 * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1489 * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1490 * be treated as UTC that way.
1491 */
1492 (void) clock_reset_timewarp();
1493 }
1494
1495 r = clock_apply_epoch();
1496 if (r < 0)
1497 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1498 else if (r > 0)
1499 log_info("System time before build time, advancing clock.");
1500}
1501
1e41242e 1502static void initialize_coredump(bool skip_setup) {
752bcb77 1503#if ENABLE_COREDUMP
1e41242e
LP		 1504 if (getpid_cached() != 1)
1505 return;
1506
1507 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1508 * will process core dumps for system services by default. */
1509 if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0)
1510 log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m");
1511
c6885f5f
FB		 1512 /* But at the same time, turn off the core_pattern logic by default, so that no
1513 * coredumps are stored until the systemd-coredump tool is enabled via
1514 * sysctl. However it can be changed via the kernel command line later so core
1515 * dumps can still be generated during early startup and in initramfs. */
1e41242e 1516 if (!skip_setup)
e557b1a6 1517 disable_coredumps();
752bcb77 1518#endif
1e41242e
LP		 1519}
1520
c6885f5f
FB		 1521static void initialize_core_pattern(bool skip_setup) {
1522 int r;
1523
1524 if (skip_setup || !arg_early_core_pattern)
1525 return;
1526
1527 if (getpid_cached() != 1)
1528 return;
1529
57512c89 1530 r = write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern, WRITE_STRING_FILE_DISABLE_BUFFER);
c6885f5f
FB		 1531 if (r < 0)
1532 log_warning_errno(r, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", arg_early_core_pattern);
1533}
1534
61fbbac1
ZJS		 1535static void update_cpu_affinity(bool skip_setup) {
1536 _cleanup_free_ char *mask = NULL;
1537
1538 if (skip_setup || !arg_cpu_affinity.set)
1539 return;
1540
1541 assert(arg_cpu_affinity.allocated > 0);
1542
1543 mask = cpu_set_to_string(&arg_cpu_affinity);
1544 log_debug("Setting CPU affinity to %s.", strnull(mask));
1545
1546 if (sched_setaffinity(0, arg_cpu_affinity.allocated, arg_cpu_affinity.set) < 0)
1547 log_warning_errno(errno, "Failed to set CPU affinity: %m");
1548}
1549
b070c7c0
MS		 1550static void update_numa_policy(bool skip_setup) {
1551 int r;
1552 _cleanup_free_ char *nodes = NULL;
1553 const char * policy = NULL;
1554
1555 if (skip_setup || !mpol_is_valid(numa_policy_get_type(&arg_numa_policy)))
1556 return;
1557
1558 if (DEBUG_LOGGING) {
1559 policy = mpol_to_string(numa_policy_get_type(&arg_numa_policy));
1560 nodes = cpu_set_to_range_string(&arg_numa_policy.nodes);
1561 log_debug("Setting NUMA policy to %s, with nodes %s.", strnull(policy), strnull(nodes));
1562 }
1563
1564 r = apply_numa_policy(&arg_numa_policy);
1565 if (r == -EOPNOTSUPP)
1566 log_debug_errno(r, "NUMA support not available, ignoring.");
1567 else if (r < 0)
1568 log_warning_errno(r, "Failed to set NUMA memory policy: %m");
1569}
1570
3c7878f9
LP		 1571static void do_reexecute(
1572 int argc,
1573 char *argv[],
1574 const struct rlimit *saved_rlimit_nofile,
1575 const struct rlimit *saved_rlimit_memlock,
1576 FDSet *fds,
1577 const char *switch_root_dir,
1578 const char *switch_root_init,
1579 const char **ret_error_message) {
1580
1581 unsigned i, j, args_size;
1582 const char **args;
1583 int r;
1584
1585 assert(saved_rlimit_nofile);
1586 assert(saved_rlimit_memlock);
1587 assert(ret_error_message);
1588
1589 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1590 * we do that */
1591 watchdog_close(true);
1592
ddfa8b0b
LP		 1593 /* Reset RLIMIT_NOFILE + RLIMIT_MEMLOCK back to the kernel defaults, so that the new systemd can pass
1594 * the kernel default to its child processes */
1595 if (saved_rlimit_nofile->rlim_cur != 0)
3c7878f9 1596 (void) setrlimit(RLIMIT_NOFILE, saved_rlimit_nofile);
ddfa8b0b 1597 if (saved_rlimit_memlock->rlim_cur != RLIM_INFINITY)
3c7878f9
LP		 1598 (void) setrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock);
1599
1600 if (switch_root_dir) {
1601 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1602 * SIGCHLD for them after deserializing. */
e73c54b8 1603 broadcast_signal(SIGTERM, false, true, arg_default_timeout_stop_usec);
3c7878f9
LP		 1604
1605 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1606 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1607 if (r < 0)
1608 log_error_errno(r, "Failed to switch root, trying to continue: %m");
1609 }
1610
1611 args_size = MAX(6, argc+1);
1612 args = newa(const char*, args_size);
1613
1614 if (!switch_root_init) {
1615 char sfd[DECIMAL_STR_MAX(int) + 1];
1616
1617 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1618 * the user didn't specify an explicit init to spawn. */
1619
1620 assert(arg_serialization);
1621 assert(fds);
1622
1623 xsprintf(sfd, "%i", fileno(arg_serialization));
1624
1625 i = 0;
1626 args[i++] = SYSTEMD_BINARY_PATH;
1627 if (switch_root_dir)
1628 args[i++] = "--switched-root";
1629 args[i++] = arg_system ? "--system" : "--user";
1630 args[i++] = "--deserialize";
1631 args[i++] = sfd;
1632 args[i++] = NULL;
1633
1634 assert(i <= args_size);
1635
1636 /*
1637 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do this is on
1638 * its own on exec(), but it will do it on exit(). Hence, to ensure we get a summary here, fork() off
1639 * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1640 * before proceeding into the exec().
1641 */
1642 valgrind_summary_hack();
1643
1644 (void) execv(args[0], (char* const*) args);
1645 log_debug_errno(errno, "Failed to execute our own binary, trying fallback: %m");
1646 }
1647
1648 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1649 * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1650 * doesn't matter.) */
1651
1652 arg_serialization = safe_fclose(arg_serialization);
1653 fds = fdset_free(fds);
1654
1655 /* Reopen the console */
1656 (void) make_console_stdio();
1657
1658 for (j = 1, i = 1; j < (unsigned) argc; j++)
1659 args[i++] = argv[j];
1660 args[i++] = NULL;
1661 assert(i <= args_size);
1662
5238e957 1663 /* Re-enable any blocked signals, especially important if we switch from initial ramdisk to init=... */
3c7878f9
LP		 1664 (void) reset_all_signal_handlers();
1665 (void) reset_signal_mask();
595225af 1666 (void) rlimit_nofile_safe();
3c7878f9
LP		 1667
1668 if (switch_root_init) {
1669 args[0] = switch_root_init;
a5cede8c 1670 (void) execve(args[0], (char* const*) args, saved_env);
3c7878f9
LP		 1671 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
1672 }
1673
1674 args[0] = "/sbin/init";
1675 (void) execv(args[0], (char* const*) args);
1676 r = -errno;
1677
1678 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
1679 ANSI_HIGHLIGHT_RED " !! " ANSI_NORMAL,
1680 "Failed to execute /sbin/init");
1681
1682 if (r == -ENOENT) {
1683 log_warning("No /sbin/init, trying fallback");
1684
1685 args[0] = "/bin/sh";
1686 args[1] = NULL;
a5cede8c 1687 (void) execve(args[0], (char* const*) args, saved_env);
3c7878f9
LP		 1688 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
1689 } else
1690 log_warning_errno(r, "Failed to execute /sbin/init, giving up: %m");
1691
1692 *ret_error_message = "Failed to execute fallback shell";
1693}
1694
7eb35049
LP		 1695static int invoke_main_loop(
1696 Manager *m,
a9fd4cd1
FB		 1697 const struct rlimit *saved_rlimit_nofile,
1698 const struct rlimit *saved_rlimit_memlock,
7eb35049
LP		 1699 bool *ret_reexecute,
1700 int *ret_retval, /* Return parameters relevant for shutting down */
1701 const char **ret_shutdown_verb, /* … */
1702 FDSet **ret_fds, /* Return parameters for reexecuting */
1703 char **ret_switch_root_dir, /* … */
1704 char **ret_switch_root_init, /* … */
1705 const char **ret_error_message) {
1706
1707 int r;
1708
1709 assert(m);
a9fd4cd1
FB		 1710 assert(saved_rlimit_nofile);
1711 assert(saved_rlimit_memlock);
7eb35049
LP		 1712 assert(ret_reexecute);
1713 assert(ret_retval);
1714 assert(ret_shutdown_verb);
1715 assert(ret_fds);
1716 assert(ret_switch_root_dir);
1717 assert(ret_switch_root_init);
1718 assert(ret_error_message);
1719
1720 for (;;) {
1721 r = manager_loop(m);
1722 if (r < 0) {
1723 *ret_error_message = "Failed to run main loop";
1724 return log_emergency_errno(r, "Failed to run main loop: %m");
1725 }
1726
3ca4d0b3 1727 switch ((ManagerObjective) r) {
7eb35049 1728
a6ecbf83 1729 case MANAGER_RELOAD: {
bda7d78b 1730 LogTarget saved_log_target;
a6ecbf83
FB		 1731 int saved_log_level;
1732
7eb35049
LP		 1733 log_info("Reloading.");
1734
3fe91079 1735 /* First, save any overridden log level/target, then parse the configuration file, which might
bda7d78b
FB		 1736 * change the log level to new settings. */
1737
a6ecbf83 1738 saved_log_level = m->log_level_overridden ? log_get_max_level() : -1;
bda7d78b 1739 saved_log_target = m->log_target_overridden ? log_get_target() : _LOG_TARGET_INVALID;
a6ecbf83 1740
a9fd4cd1 1741 (void) parse_configuration(saved_rlimit_nofile, saved_rlimit_memlock);
7eb35049
LP		 1742
1743 set_manager_defaults(m);
1744
61fbbac1 1745 update_cpu_affinity(false);
b070c7c0 1746 update_numa_policy(false);
61fbbac1 1747
a6ecbf83
FB		 1748 if (saved_log_level >= 0)
1749 manager_override_log_level(m, saved_log_level);
bda7d78b
FB		 1750 if (saved_log_target >= 0)
1751 manager_override_log_target(m, saved_log_target);
a6ecbf83 1752
7eb35049
LP		 1753 r = manager_reload(m);
1754 if (r < 0)
7a35fa24
LP		 1755 /* Reloading failed before the point of no return. Let's continue running as if nothing happened. */
1756 m->objective = MANAGER_OK;
7eb35049
LP		 1757
1758 break;
a6ecbf83 1759 }
7eb35049
LP		 1760
1761 case MANAGER_REEXECUTE:
1762
1763 r = prepare_reexecute(m, &arg_serialization, ret_fds, false);
1764 if (r < 0) {
1765 *ret_error_message = "Failed to prepare for reexecution";
1766 return r;
1767 }
1768
1769 log_notice("Reexecuting.");
1770
1771 *ret_reexecute = true;
1772 *ret_retval = EXIT_SUCCESS;
1773 *ret_shutdown_verb = NULL;
1774 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1775
1776 return 0;
1777
1778 case MANAGER_SWITCH_ROOT:
1779 if (!m->switch_root_init) {
1780 r = prepare_reexecute(m, &arg_serialization, ret_fds, true);
1781 if (r < 0) {
1782 *ret_error_message = "Failed to prepare for reexecution";
1783 return r;
1784 }
1785 } else
1786 *ret_fds = NULL;
1787
1788 log_notice("Switching root.");
1789
1790 *ret_reexecute = true;
1791 *ret_retval = EXIT_SUCCESS;
1792 *ret_shutdown_verb = NULL;
1793
1794 /* Steal the switch root parameters */
49052946
YW		 1795 *ret_switch_root_dir = TAKE_PTR(m->switch_root);
1796 *ret_switch_root_init = TAKE_PTR(m->switch_root_init);
7eb35049
LP		 1797
1798 return 0;
1799
1800 case MANAGER_EXIT:
1801
1802 if (MANAGER_IS_USER(m)) {
1803 log_debug("Exit.");
1804
1805 *ret_reexecute = false;
1806 *ret_retval = m->return_value;
1807 *ret_shutdown_verb = NULL;
1808 *ret_fds = NULL;
1809 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1810
1811 return 0;
1812 }
1813
1814 _fallthrough_;
1815 case MANAGER_REBOOT:
1816 case MANAGER_POWEROFF:
1817 case MANAGER_HALT:
1818 case MANAGER_KEXEC: {
af41e508
LP		 1819 static const char * const table[_MANAGER_OBJECTIVE_MAX] = {
1820 [MANAGER_EXIT] = "exit",
1821 [MANAGER_REBOOT] = "reboot",
7eb35049 1822 [MANAGER_POWEROFF] = "poweroff",
af41e508
LP		 1823 [MANAGER_HALT] = "halt",
1824 [MANAGER_KEXEC] = "kexec",
7eb35049
LP		 1825 };
1826
1827 log_notice("Shutting down.");
1828
1829 *ret_reexecute = false;
1830 *ret_retval = m->return_value;
af41e508 1831 assert_se(*ret_shutdown_verb = table[m->objective]);
7eb35049
LP		 1832 *ret_fds = NULL;
1833 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1834
1835 return 0;
1836 }
1837
1838 default:
af41e508 1839 assert_not_reached("Unknown or unexpected manager objective.");
7eb35049
LP		 1840 }
1841 }
1842}
1843
31aef7ff
LP		 1844static void log_execution_mode(bool *ret_first_boot) {
1845 assert(ret_first_boot);
1846
1847 if (arg_system) {
1848 int v;
1849
681bd2c5 1850 log_info("systemd " GIT_VERSION " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
31aef7ff
LP		 1851 arg_action == ACTION_TEST ? "test " : "" );
1852
1853 v = detect_virtualization();
1854 if (v > 0)
1855 log_info("Detected virtualization %s.", virtualization_to_string(v));
1856
1857 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1858
1859 if (in_initrd()) {
1860 *ret_first_boot = false;
1861 log_info("Running in initial RAM disk.");
1862 } else {
1863 /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
1864 * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
1865 * doesn't it's unpopulated. This allows container managers and installers to provision a
1866 * couple of files already. If the container manager wants to provision the machine ID itself
1867 * it should pass $container_uuid to PID 1. */
1868
1869 *ret_first_boot = access("/etc/machine-id", F_OK) < 0;
1870 if (*ret_first_boot)
1871 log_info("Running with unpopulated /etc.");
1872 }
1873 } else {
b9e90f3a
LP		 1874 if (DEBUG_LOGGING) {
1875 _cleanup_free_ char *t;
31aef7ff 1876
b9e90f3a 1877 t = uid_to_name(getuid());
681bd2c5 1878 log_debug("systemd " GIT_VERSION " running in %suser mode for user " UID_FMT "/%s. (" SYSTEMD_FEATURES ")",
b9e90f3a
LP		 1879 arg_action == ACTION_TEST ? " test" : "", getuid(), strna(t));
1880 }
31aef7ff
LP		 1881
1882 *ret_first_boot = false;
1883 }
1884}
1885
5afbaa36
LP		 1886static int initialize_runtime(
1887 bool skip_setup,
1888 struct rlimit *saved_rlimit_nofile,
1889 struct rlimit *saved_rlimit_memlock,
1890 const char **ret_error_message) {
5afbaa36
LP		 1891 int r;
1892
1893 assert(ret_error_message);
1894
1895 /* Sets up various runtime parameters. Many of these initializations are conditionalized:
1896 *
1897 * - Some only apply to --system instances
1898 * - Some only apply to --user instances
1899 * - Some only apply when we first start up, but not when we reexecute
1900 */
1901
2d776038
LP		 1902 if (arg_action != ACTION_RUN)
1903 return 0;
1904
61fbbac1 1905 update_cpu_affinity(skip_setup);
b070c7c0 1906 update_numa_policy(skip_setup);
61fbbac1 1907
3c3c6cb9 1908 if (arg_system) {
5238e957 1909 /* Make sure we leave a core dump without panicking the kernel. */
3c3c6cb9 1910 install_crash_handler();
5afbaa36 1911
3c3c6cb9 1912 if (!skip_setup) {
143fadf3 1913 r = mount_cgroup_controllers();
3c3c6cb9
LP		 1914 if (r < 0) {
1915 *ret_error_message = "Failed to mount cgroup hierarchies";
1916 return r;
1917 }
1918
1919 status_welcome();
1920 hostname_setup();
1921 machine_id_setup(NULL, arg_machine_id, NULL);
1922 loopback_setup();
1923 bump_unix_max_dgram_qlen();
a8b627aa 1924 bump_file_max_and_nr_open();
3c3c6cb9
LP		 1925 test_usr();
1926 write_container_id();
1927 }
8a2c1fbf 1928
3c3c6cb9
LP		 1929 if (arg_watchdog_device) {
1930 r = watchdog_set_device(arg_watchdog_device);
1931 if (r < 0)
1932 log_warning_errno(r, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device);
1933 }
1934
1f65fd49 1935 if (timestamp_is_set(arg_runtime_watchdog))
3c3c6cb9
LP		 1936 watchdog_set_timeout(&arg_runtime_watchdog);
1937 }
5afbaa36
LP		 1938
1939 if (arg_timer_slack_nsec != NSEC_INFINITY)
1940 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
3a671cd1 1941 log_warning_errno(errno, "Failed to adjust timer slack, ignoring: %m");
5afbaa36
LP		 1942
1943 if (arg_system && !cap_test_all(arg_capability_bounding_set)) {
1944 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
1945 if (r < 0) {
1946 *ret_error_message = "Failed to drop capability bounding set of usermode helpers";
1947 return log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
1948 }
1949
1950 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
1951 if (r < 0) {
1952 *ret_error_message = "Failed to drop capability bounding set";
1953 return log_emergency_errno(r, "Failed to drop capability bounding set: %m");
1954 }
1955 }
1956
39362f6f
JB		 1957 if (arg_system && arg_no_new_privs) {
1958 if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) {
1959 *ret_error_message = "Failed to disable new privileges";
1960 return log_emergency_errno(errno, "Failed to disable new privileges: %m");
1961 }
1962 }
1963
5afbaa36
LP		 1964 if (arg_syscall_archs) {
1965 r = enforce_syscall_archs(arg_syscall_archs);
1966 if (r < 0) {
1967 *ret_error_message = "Failed to set syscall architectures";
1968 return r;
1969 }
1970 }
1971
1972 if (!arg_system)
1973 /* Become reaper of our children */
1974 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
1975 log_warning_errno(errno, "Failed to make us a subreaper: %m");
1976
a17c1712
LP		 1977 /* Bump up RLIMIT_NOFILE for systemd itself */
1978 (void) bump_rlimit_nofile(saved_rlimit_nofile);
1979 (void) bump_rlimit_memlock(saved_rlimit_memlock);
5afbaa36
LP		 1980
1981 return 0;
1982}
1983
6acca5fc
LP		 1984static int do_queue_default_job(
1985 Manager *m,
1986 const char **ret_error_message) {
1987
1988 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
1989 Job *default_unit_job;
1990 Unit *target = NULL;
1991 int r;
1992
1993 log_debug("Activating default unit: %s", arg_default_unit);
1994
4109ede7
ZJS		 1995 r = manager_load_startable_unit_or_warn(m, arg_default_unit, NULL, &target);
1996 if (r < 0) {
1997 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET);
6acca5fc 1998
4109ede7 1999 r = manager_load_startable_unit_or_warn(m, SPECIAL_RESCUE_TARGET, NULL, &target);
6acca5fc 2000 if (r < 0) {
4109ede7
ZJS		 2001 *ret_error_message = r == -ERFKILL ? "Rescue target masked"
2002 : "Failed to load rescue target";
2003 return r;
6acca5fc
LP		 2004 }
2005 }
2006
2007 assert(target->load_state == UNIT_LOADED);
2008
50cbaba4 2009 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, NULL, &error, &default_unit_job);
6acca5fc
LP		 2010 if (r == -EPERM) {
2011 log_debug_errno(r, "Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
2012
2013 sd_bus_error_free(&error);
2014
50cbaba4 2015 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, NULL, &error, &default_unit_job);
6acca5fc
LP		 2016 if (r < 0) {
2017 *ret_error_message = "Failed to start default target";
2018 return log_emergency_errno(r, "Failed to start default target: %s", bus_error_message(&error, r));
2019 }
2020
2021 } else if (r < 0) {
2022 *ret_error_message = "Failed to isolate default target";
2023 return log_emergency_errno(r, "Failed to isolate default target: %s", bus_error_message(&error, r));
2024 }
2025
2026 m->default_unit_job_id = default_unit_job->id;
2027
2028 return 0;
2029}
2030
a9fd4cd1
FB		 2031static void save_rlimits(struct rlimit *saved_rlimit_nofile,
2032 struct rlimit *saved_rlimit_memlock) {
2033
2034 assert(saved_rlimit_nofile);
2035 assert(saved_rlimit_memlock);
2036
2037 if (getrlimit(RLIMIT_NOFILE, saved_rlimit_nofile) < 0)
2038 log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m");
2039
2040 if (getrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock) < 0)
2041 log_warning_errno(errno, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
2042}
2043
2044static void fallback_rlimit_nofile(const struct rlimit *saved_rlimit_nofile) {
2045 struct rlimit *rl;
2046
2047 if (arg_default_rlimit[RLIMIT_NOFILE])
2048 return;
2049
2050 /* Make sure forked processes get limits based on the original kernel setting */
2051
2052 rl = newdup(struct rlimit, saved_rlimit_nofile, 1);
2053 if (!rl) {
2054 log_oom();
2055 return;
2056 }
2057
2058 /* Bump the hard limit for system services to a substantially higher value. The default
2059 * hard limit current kernels set is pretty low (4K), mostly for historical
2060 * reasons. According to kernel developers, the fd handling in recent kernels has been
2061 * optimized substantially enough, so that we can bump the limit now, without paying too
2062 * high a price in memory or performance. Note however that we only bump the hard limit,
2063 * not the soft limit. That's because select() works the way it works, and chokes on fds
2064 * >= 1024. If we'd bump the soft limit globally, it might accidentally happen to
2065 * unexpecting programs that they get fds higher than what they can process using
2066 * select(). By only bumping the hard limit but leaving the low limit as it is we avoid
2067 * this pitfall: programs that are written by folks aware of the select() problem in mind
2068 * (and thus use poll()/epoll instead of select(), the way everybody should) can
2069 * explicitly opt into high fds by bumping their soft limit beyond 1024, to the hard limit
2070 * we pass. */
2071 if (arg_system) {
2072 int nr;
2073
2074 /* Get the underlying absolute limit the kernel enforces */
2075 nr = read_nr_open();
2076
2077 rl->rlim_max = MIN((rlim_t) nr, MAX(rl->rlim_max, (rlim_t) HIGH_RLIMIT_NOFILE));
2078 }
2079
2080 /* If for some reason we were invoked with a soft limit above 1024 (which should never
2081 * happen!, but who knows what we get passed in from pam_limit when invoked as --user
2082 * instance), then lower what we pass on to not confuse our children */
2083 rl->rlim_cur = MIN(rl->rlim_cur, (rlim_t) FD_SETSIZE);
2084
2085 arg_default_rlimit[RLIMIT_NOFILE] = rl;
2086}
2087
2088static void fallback_rlimit_memlock(const struct rlimit *saved_rlimit_memlock) {
2089 struct rlimit *rl;
2090
2091 /* Pass the original value down to invoked processes */
2092
2093 if (arg_default_rlimit[RLIMIT_MEMLOCK])
2094 return;
2095
2096 rl = newdup(struct rlimit, saved_rlimit_memlock, 1);
2097 if (!rl) {
2098 log_oom();
2099 return;
2100 }
2101
2102 arg_default_rlimit[RLIMIT_MEMLOCK] = rl;
2103}
2104
fb39af4c
ZJS		 2105static void reset_arguments(void) {
2106 /* Frees/resets arg_* variables, with a few exceptions commented below. */
970777b5
LP		 2107
2108 arg_default_unit = mfree(arg_default_unit);
fb39af4c
ZJS		 2109
2110 /* arg_system — ignore */
2111
2112 arg_dump_core = true;
2113 arg_crash_chvt = -1;
2114 arg_crash_shell = false;
2115 arg_crash_reboot = false;
970777b5 2116 arg_confirm_spawn = mfree(arg_confirm_spawn);
fb39af4c 2117 arg_show_status = _SHOW_STATUS_INVALID;
36cf4507 2118 arg_status_unit_format = STATUS_UNIT_FORMAT_DEFAULT;
fb39af4c
ZJS		 2119 arg_switched_root = false;
2120 arg_pager_flags = 0;
2121 arg_service_watchdogs = true;
2122 arg_default_std_output = EXEC_OUTPUT_JOURNAL;
2123 arg_default_std_error = EXEC_OUTPUT_INHERIT;
2124 arg_default_restart_usec = DEFAULT_RESTART_USEC;
2125 arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
2126 arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
2127 arg_default_timeout_abort_usec = DEFAULT_TIMEOUT_USEC;
2128 arg_default_timeout_abort_set = false;
2129 arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
2130 arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
2131 arg_runtime_watchdog = 0;
65224c1d 2132 arg_reboot_watchdog = 10 * USEC_PER_MINUTE;
acafd7d8 2133 arg_kexec_watchdog = 0;
fb39af4c
ZJS		 2134 arg_early_core_pattern = NULL;
2135 arg_watchdog_device = NULL;
2136
970777b5 2137 arg_default_environment = strv_free(arg_default_environment);
fb39af4c
ZJS		 2138 rlimit_free_all(arg_default_rlimit);
2139
2140 arg_capability_bounding_set = CAP_ALL;
2141 arg_no_new_privs = false;
2142 arg_timer_slack_nsec = NSEC_INFINITY;
2143 arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
2144
970777b5 2145 arg_syscall_archs = set_free(arg_syscall_archs);
61fbbac1 2146
fb39af4c
ZJS		 2147 /* arg_serialization — ignore */
2148
2149 arg_default_cpu_accounting = -1;
2150 arg_default_io_accounting = false;
2151 arg_default_ip_accounting = false;
2152 arg_default_blockio_accounting = false;
2153 arg_default_memory_accounting = MEMORY_ACCOUNTING_DEFAULT;
2154 arg_default_tasks_accounting = true;
3a0f06c4 2155 arg_default_tasks_max = DEFAULT_TASKS_MAX;
fb39af4c
ZJS		 2156 arg_machine_id = (sd_id128_t) {};
2157 arg_cad_burst_action = EMERGENCY_ACTION_REBOOT_FORCE;
2158 arg_default_oom_policy = OOM_STOP;
2159
61fbbac1 2160 cpu_set_reset(&arg_cpu_affinity);
b070c7c0 2161 numa_policy_reset(&arg_numa_policy);
970777b5
LP		 2162}
2163
a9fd4cd1
FB		 2164static int parse_configuration(const struct rlimit *saved_rlimit_nofile,
2165 const struct rlimit *saved_rlimit_memlock) {
97d1fb94
LP		 2166 int r;
2167
a9fd4cd1
FB		 2168 assert(saved_rlimit_nofile);
2169 assert(saved_rlimit_memlock);
2170
fb39af4c
ZJS		 2171 /* Assign configuration defaults */
2172 reset_arguments();
2173
97d1fb94 2174 r = parse_config_file();
470a5e6d
ZJS		 2175 if (r < 0)
2176 log_warning_errno(r, "Failed to parse config file, ignoring: %m");
97d1fb94
LP		 2177
2178 if (arg_system) {
2179 r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
2180 if (r < 0)
2181 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
2182 }
2183
a9fd4cd1
FB		 2184 /* Initialize some default rlimits for services if they haven't been configured */
2185 fallback_rlimit_nofile(saved_rlimit_nofile);
2186 fallback_rlimit_memlock(saved_rlimit_memlock);
2187
97d1fb94
LP		 2188 /* Note that this also parses bits from the kernel command line, including "debug". */
2189 log_parse_environment();
2190
470a5e6d
ZJS		 2191 return 0;
2192}
2193
a9fd4cd1
FB		 2194static int load_configuration(
2195 int argc,
2196 char **argv,
2197 const struct rlimit *saved_rlimit_nofile,
2198 const struct rlimit *saved_rlimit_memlock,
2199 const char **ret_error_message) {
470a5e6d
ZJS		 2200 int r;
2201
a9fd4cd1
FB		 2202 assert(saved_rlimit_nofile);
2203 assert(saved_rlimit_memlock);
470a5e6d
ZJS		 2204 assert(ret_error_message);
2205
a9fd4cd1 2206 (void) parse_configuration(saved_rlimit_nofile, saved_rlimit_memlock);
470a5e6d 2207
97d1fb94
LP		 2208 r = parse_argv(argc, argv);
2209 if (r < 0) {
2210 *ret_error_message = "Failed to parse commandline arguments";
2211 return r;
2212 }
2213
2214 /* Initialize default unit */
2215 if (!arg_default_unit) {
2216 arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET);
2217 if (!arg_default_unit) {
2218 *ret_error_message = "Failed to set default unit";
2219 return log_oom();
2220 }
2221 }
2222
db33214b 2223 /* Initialize the show status setting if it hasn't been set explicitly yet */
7a293242 2224 if (arg_show_status == _SHOW_STATUS_INVALID)
db33214b
LP		 2225 arg_show_status = SHOW_STATUS_YES;
2226
97d1fb94
LP		 2227 return 0;
2228}
2229
b0d7c989
LP		 2230static int safety_checks(void) {
2231
febf46a4 2232 if (getpid_cached() == 1 &&
baaa35ad
ZJS		 2233 arg_action != ACTION_RUN)
2234 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2235 "Unsupported execution mode while PID 1.");
febf46a4
LP		 2236
2237 if (getpid_cached() == 1 &&
baaa35ad
ZJS		 2238 !arg_system)
2239 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2240 "Can't run --user mode as PID 1.");
febf46a4
LP		 2241
2242 if (arg_action == ACTION_RUN &&
2243 arg_system &&
baaa35ad
ZJS		 2244 getpid_cached() != 1)
2245 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2246 "Can't run system mode unless PID 1.");
febf46a4 2247
b0d7c989 2248 if (arg_action == ACTION_TEST &&
baaa35ad
ZJS		 2249 geteuid() == 0)
2250 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2251 "Don't run test mode as root.");
b0d7c989
LP		 2252
2253 if (!arg_system &&
2254 arg_action == ACTION_RUN &&
baaa35ad
ZJS		 2255 sd_booted() <= 0)
2256 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
2257 "Trying to run as user instance, but the system has not been booted with systemd.");
b0d7c989
LP		 2258
2259 if (!arg_system &&
2260 arg_action == ACTION_RUN &&
baaa35ad
ZJS		 2261 !getenv("XDG_RUNTIME_DIR"))
2262 return log_error_errno(SYNTHETIC_ERRNO(EUNATCH),
2263 "Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
b0d7c989
LP		 2264
2265 if (arg_system &&
2266 arg_action == ACTION_RUN &&
baaa35ad
ZJS		 2267 running_in_chroot() > 0)
2268 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
2269 "Cannot be run in a chroot() environment.");
b0d7c989
LP		 2270
2271 return 0;
2272}
2273
74da609f
LP		 2274static int initialize_security(
2275 bool *loaded_policy,
2276 dual_timestamp *security_start_timestamp,
2277 dual_timestamp *security_finish_timestamp,
2278 const char **ret_error_message) {
2279
2280 int r;
2281
2282 assert(loaded_policy);
2283 assert(security_start_timestamp);
2284 assert(security_finish_timestamp);
2285 assert(ret_error_message);
2286
2287 dual_timestamp_get(security_start_timestamp);
2288
97149f40 2289 r = mac_selinux_setup(loaded_policy);
74da609f
LP		 2290 if (r < 0) {
2291 *ret_error_message = "Failed to load SELinux policy";
2292 return r;
2293 }
2294
2295 r = mac_smack_setup(loaded_policy);
2296 if (r < 0) {
2297 *ret_error_message = "Failed to load SMACK policy";
2298 return r;
2299 }
2300
2301 r = ima_setup();
2302 if (r < 0) {
2303 *ret_error_message = "Failed to load IMA policy";
2304 return r;
2305 }
2306
2307 dual_timestamp_get(security_finish_timestamp);
2308 return 0;
2309}
2310
263162da
LP		 2311static void test_summary(Manager *m) {
2312 assert(m);
2313
2314 printf("-> By units:\n");
2315 manager_dump_units(m, stdout, "\t");
2316
2317 printf("-> By jobs:\n");
2318 manager_dump_jobs(m, stdout, "\t");
2319}
2320
efeb853f
LP		 2321static int collect_fds(FDSet **ret_fds, const char **ret_error_message) {
2322 int r;
2323
2324 assert(ret_fds);
2325 assert(ret_error_message);
2326
2327 r = fdset_new_fill(ret_fds);
2328 if (r < 0) {
2329 *ret_error_message = "Failed to allocate fd set";
2330 return log_emergency_errno(r, "Failed to allocate fd set: %m");
2331 }
2332
2333 fdset_cloexec(*ret_fds, true);
2334
2335 if (arg_serialization)
2336 assert_se(fdset_remove(*ret_fds, fileno(arg_serialization)) >= 0);
2337
2338 return 0;
2339}
2340
2e51b31c
LP		 2341static void setup_console_terminal(bool skip_setup) {
2342
2343 if (!arg_system)
2344 return;
2345
2346 /* Become a session leader if we aren't one yet. */
2347 (void) setsid();
2348
2349 /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2350 * tty. */
2351 (void) release_terminal();
2352
2353 /* Reset the console, but only if this is really init and we are freshly booted */
2354 if (getpid_cached() == 1 && !skip_setup)
2355 (void) console_setup();
2356}
2357
aa40ff07
LP		 2358static bool early_skip_setup_check(int argc, char *argv[]) {
2359 bool found_deserialize = false;
2360 int i;
2361
2362 /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2363 * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2364 * anyway, even if in that case we also do deserialization. */
2365
2366 for (i = 1; i < argc; i++) {
aa40ff07
LP		 2367 if (streq(argv[i], "--switched-root"))
2368 return false; /* If we switched root, don't skip the setup. */
2369 else if (streq(argv[i], "--deserialize"))
2370 found_deserialize = true;
2371 }
2372
2373 return found_deserialize; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2374}
2375
0e06a031
LP		 2376static int save_env(void) {
2377 char **l;
2378
2379 l = strv_copy(environ);
2380 if (!l)
2381 return -ENOMEM;
2382
2383 strv_free_and_replace(saved_env, l);
2384 return 0;
2385}
2386
60918275 2387int main(int argc, char *argv[]) {
625e8690
LP		 2388
2389 dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL, userspace_timestamp = DUAL_TIMESTAMP_NULL, kernel_timestamp = DUAL_TIMESTAMP_NULL,
2390 security_start_timestamp = DUAL_TIMESTAMP_NULL, security_finish_timestamp = DUAL_TIMESTAMP_NULL;
ddfa8b0b
LP		 2391 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0),
2392 saved_rlimit_memlock = RLIMIT_MAKE_CONST(RLIM_INFINITY); /* The original rlimits we passed
2393 * in. Note we use different values
2394 * for the two that indicate whether
2395 * these fields are initialized! */
625e8690
LP		 2396 bool skip_setup, loaded_policy = false, queue_default_job = false, first_boot = false, reexecute = false;
2397 char *switch_root_dir = NULL, *switch_root_init = NULL;
9d76d730 2398 usec_t before_startup, after_startup;
625e8690 2399 static char systemd[] = "systemd";
9d76d730 2400 char timespan[FORMAT_TIMESPAN_MAX];
625e8690
LP		 2401 const char *shutdown_verb = NULL, *error_message = NULL;
2402 int r, retval = EXIT_FAILURE;
2403 Manager *m = NULL;
a16e1123 2404 FDSet *fds = NULL;
27b14a22 2405
d72a8f10 2406 /* SysV compatibility: redirect init → telinit */
6808a0bc 2407 redirect_telinit(argc, argv);
2cb1a60d 2408
d72a8f10 2409 /* Take timestamps early on */
c3a170f3
HH		 2410 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
2411 dual_timestamp_get(&userspace_timestamp);
2412
d72a8f10
LP		 2413 /* Figure out whether we need to do initialize the system, or if we already did that because we are
2414 * reexecuting */
aa40ff07 2415 skip_setup = early_skip_setup_check(argc, argv);
d03bc1b8 2416
d72a8f10
LP		 2417 /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2418 * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
f3b6a3ed 2419 program_invocation_short_name = systemd;
eee8b7ab 2420 (void) prctl(PR_SET_NAME, systemd);
5d6b1584 2421
d72a8f10 2422 /* Save the original command line */
36fea155 2423 save_argc_argv(argc, argv);
f3b6a3ed 2424
0e06a031
LP		 2425 /* Save the original environment as we might need to restore it if we're requested to execute another
2426 * system manager later. */
2427 r = save_env();
2428 if (r < 0) {
2429 error_message = "Failed to copy environment block";
2430 goto finish;
2431 }
a5cede8c 2432
6fdb8de4 2433 /* Make sure that if the user says "syslog" we actually log to the journal. */
c1dc6153 2434 log_set_upgrade_syslog_to_journal(true);
bbe63281 2435
df0ff127 2436 if (getpid_cached() == 1) {
b5752d23
LP		 2437 /* When we run as PID 1 force system mode */
2438 arg_system = true;
2439
48a601fe 2440 /* Disable the umask logic */
90dc8c2e
MG		 2441 umask(0);
2442
92890452
LP		 2443 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2444 * activated yet (even though the log socket for it exists). */
d075092f
LP		 2445 log_set_prohibit_ipc(true);
2446
48a601fe
LP		 2447 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2448 * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2449 * child process right before execve()'ing the actual binary, at a point in time where socket
2450 * activation stderr/stdout area already set up. */
2451 log_set_always_reopen_console(true);
48a601fe 2452
92890452 2453 if (detect_container() <= 0) {
4f8d551f 2454
92890452 2455 /* Running outside of a container as PID 1 */
92890452
LP		 2456 log_set_target(LOG_TARGET_KMSG);
2457 log_open();
a866073d 2458
92890452
LP		 2459 if (in_initrd())
2460 initrd_timestamp = userspace_timestamp;
c3ba6250 2461
92890452
LP		 2462 if (!skip_setup) {
2463 r = mount_setup_early();
2464 if (r < 0) {
2465 error_message = "Failed to mount early API filesystems";
2466 goto finish;
2467 }
2468
0a2eef1e
LP		 2469 /* Let's open the log backend a second time, in case the first time didn't
2470 * work. Quite possibly we have mounted /dev just now, so /dev/kmsg became
2471 * available, and it previously wasn't. */
2472 log_open();
2473
6123dfaa
ZJS		 2474 disable_printk_ratelimit();
2475
92890452
LP		 2476 r = initialize_security(
2477 &loaded_policy,
2478 &security_start_timestamp,
2479 &security_finish_timestamp,
2480 &error_message);
2481 if (r < 0)
2482 goto finish;
d723cd65 2483 }
eee8b7ab 2484
92890452
LP		 2485 if (mac_selinux_init() < 0) {
2486 error_message = "Failed to initialize SELinux policy";
96694e99 2487 goto finish;
92890452 2488 }
0b3325e7 2489
92890452
LP		 2490 if (!skip_setup)
2491 initialize_clock();
2492
2493 /* Set the default for later on, but don't actually open the logs like this for now. Note that
2494 * if we are transitioning from the initrd there might still be journal fd open, and we
2495 * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2496 * elsewhere. */
2497 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
2498
2499 } else {
2500 /* Running inside a container, as PID 1 */
92890452
LP		 2501 log_set_target(LOG_TARGET_CONSOLE);
2502 log_open();
2503
2504 /* For later on, see above... */
2505 log_set_target(LOG_TARGET_JOURNAL);
2506
2507 /* clear the kernel timestamp,
2508 * because we are in a container */
2509 kernel_timestamp = DUAL_TIMESTAMP_NULL;
cb6531be 2510 }
7948c4df 2511
92890452 2512 initialize_coredump(skip_setup);
a866073d 2513
92890452
LP		 2514 r = fixup_environment();
2515 if (r < 0) {
2516 log_emergency_errno(r, "Failed to fix up PID 1 environment: %m");
2517 error_message = "Failed to fix up PID1 environment";
2518 goto finish;
2519 }
a866073d 2520
c3a170f3 2521 } else {
a866073d 2522 /* Running as user instance */
463d0d15 2523 arg_system = false;
eeecf6e6 2524 log_set_target(LOG_TARGET_AUTO);
871e5809 2525 log_open();
c3a170f3
HH		 2526
2527 /* clear the kernel timestamp,
2528 * because we are not PID 1 */
6513d561 2529 kernel_timestamp = DUAL_TIMESTAMP_NULL;
bbe63281 2530 }
a5dab5ce 2531
6edefe0b 2532 if (arg_system) {
92890452
LP		 2533 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2534 * they never log into the console. */
3a18b604 2535 log_show_color(colors_enabled());
92890452 2536
c76cf844
AK		 2537 r = make_null_stdio();
2538 if (r < 0)
92890452 2539 log_warning_errno(r, "Failed to redirect standard streams to /dev/null, ignoring: %m");
3a18b604
FB		 2540 }
2541
f170852a
LP		 2542 /* Mount /proc, /sys and friends, so that /proc/cmdline and
2543 * /proc/$PID/fd is available. */
df0ff127 2544 if (getpid_cached() == 1) {
f84f9974 2545
a132bef0 2546 /* Load the kernel modules early. */
2e75e2a8
DM		 2547 if (!skip_setup)
2548 kmod_setup();
2e75e2a8 2549
0c85a4f3 2550 r = mount_setup(loaded_policy);
cb6531be
ZJS		 2551 if (r < 0) {
2552 error_message = "Failed to mount API filesystems";
8efe3c01 2553 goto finish;
cb6531be 2554 }
c18ecf03
LP		 2555
2556 /* The efivarfs is now mounted, let's read the random seed off it */
2557 (void) efi_take_random_seed();
0c85a4f3 2558 }
4ade7963 2559
a9fd4cd1
FB		 2560 /* Save the original RLIMIT_NOFILE/RLIMIT_MEMLOCK so that we can reset it later when
2561 * transitioning from the initrd to the main systemd or suchlike. */
2562 save_rlimits(&saved_rlimit_nofile, &saved_rlimit_memlock);
2563
4ade7963 2564 /* Reset all signal handlers. */
ce30c8dc
LP		 2565 (void) reset_all_signal_handlers();
2566 (void) ignore_signals(SIGNALS_IGNORE, -1);
078e4539 2567
a9fd4cd1 2568 r = load_configuration(argc, argv, &saved_rlimit_nofile, &saved_rlimit_memlock, &error_message);
97d1fb94 2569 if (r < 0)
f170852a 2570 goto finish;
10c961b9 2571
b0d7c989
LP		 2572 r = safety_checks();
2573 if (r < 0)
fe783b03 2574 goto finish;
fe783b03 2575
bbc1acab 2576 if (IN_SET(arg_action, ACTION_TEST, ACTION_HELP, ACTION_DUMP_CONFIGURATION_ITEMS, ACTION_DUMP_BUS_PROPERTIES))
0221d68a 2577 (void) pager_open(arg_pager_flags);
b0d7c989
LP		 2578
2579 if (arg_action != ACTION_RUN)
74e7579c 2580 skip_setup = true;
b87c2aa6 2581
fa0f4d8a 2582 if (arg_action == ACTION_HELP) {
37ec0fdd 2583 retval = help() < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
f170852a 2584 goto finish;
9ba0bc4e
ZJS		 2585 } else if (arg_action == ACTION_VERSION) {
2586 retval = version();
2587 goto finish;
fa0f4d8a 2588 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
e537352b 2589 unit_dump_config_items(stdout);
22f4096c 2590 retval = EXIT_SUCCESS;
e537352b 2591 goto finish;
bbc1acab
YW		 2592 } else if (arg_action == ACTION_DUMP_BUS_PROPERTIES) {
2593 dump_bus_properties(stdout);
2594 retval = EXIT_SUCCESS;
2595 goto finish;
f170852a
LP		 2596 }
2597
4c701096 2598 assert_se(IN_SET(arg_action, ACTION_RUN, ACTION_TEST));
f170852a 2599
5a2e0c62
LP		 2600 /* Move out of the way, so that we won't block unmounts */
2601 assert_se(chdir("/") == 0);
2602
dea374e8 2603 if (arg_action == ACTION_RUN) {
a70c72a0 2604
c6885f5f
FB		 2605 /* A core pattern might have been specified via the cmdline. */
2606 initialize_core_pattern(skip_setup);
2607
efeb853f 2608 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
a70c72a0
LP		 2609 log_close();
2610
2611 /* Remember open file descriptors for later deserialization */
efeb853f
LP		 2612 r = collect_fds(&fds, &error_message);
2613 if (r < 0)
dea374e8 2614 goto finish;
a16e1123 2615
2e51b31c
LP		 2616 /* Give up any control of the console, but make sure its initialized. */
2617 setup_console_terminal(skip_setup);
56d96fc0 2618
a70c72a0
LP		 2619 /* Open the logging devices, if possible and necessary */
2620 log_open();
56d96fc0 2621 }
4ade7963 2622
31aef7ff 2623 log_execution_mode(&first_boot);
a5dab5ce 2624
2d776038
LP		 2625 r = initialize_runtime(skip_setup,
2626 &saved_rlimit_nofile,
2627 &saved_rlimit_memlock,
2628 &error_message);
2629 if (r < 0)
2630 goto finish;
4096d6f5 2631
e0a3da1f
ZJS		 2632 r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER,
2633 arg_action == ACTION_TEST ? MANAGER_TEST_FULL : 0,
2634 &m);
e96d6be7 2635 if (r < 0) {
da927ba9 2636 log_emergency_errno(r, "Failed to allocate manager object: %m");
cb6531be 2637 error_message = "Failed to allocate manager object";
60918275
LP		 2638 goto finish;
2639 }
2640
9f9f0342
LP		 2641 m->timestamps[MANAGER_TIMESTAMP_KERNEL] = kernel_timestamp;
2642 m->timestamps[MANAGER_TIMESTAMP_INITRD] = initrd_timestamp;
2643 m->timestamps[MANAGER_TIMESTAMP_USERSPACE] = userspace_timestamp;
d4ee7bd8
YW		 2644 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START)] = security_start_timestamp;
2645 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH)] = security_finish_timestamp;
9e58ff9c 2646
85cb4151 2647 set_manager_defaults(m);
7b46fc6a 2648 set_manager_settings(m);
fd130612 2649 manager_set_first_boot(m, first_boot);
27d340c7 2650
bf4df7c3 2651 /* Remember whether we should queue the default job */
d3b1c508 2652 queue_default_job = !arg_serialization || arg_switched_root;
bf4df7c3 2653
9d76d730
LP		 2654 before_startup = now(CLOCK_MONOTONIC);
2655
d3b1c508 2656 r = manager_startup(m, arg_serialization, fds);
58f88d92 2657 if (r < 0) {
cefb3eda 2658 error_message = "Failed to start up manager";
58f88d92
ZJS		 2659 goto finish;
2660 }
a16e1123 2661
6acca5fc 2662 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2feceb5e 2663 fds = fdset_free(fds);
74ca738f 2664 arg_serialization = safe_fclose(arg_serialization);
bf4df7c3
LP		 2665
2666 if (queue_default_job) {
6acca5fc 2667 r = do_queue_default_job(m, &error_message);
718db961 2668 if (r < 0)
37d88da7 2669 goto finish;
6acca5fc 2670 }
ab17a050 2671
6acca5fc 2672 after_startup = now(CLOCK_MONOTONIC);
60918275 2673
6acca5fc
LP		 2674 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
2675 "Loaded units and determined initial transaction in %s.",
2676 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
07672f49 2677
6acca5fc 2678 if (arg_action == ACTION_TEST) {
263162da 2679 test_summary(m);
6acca5fc
LP		 2680 retval = EXIT_SUCCESS;
2681 goto finish;
e965d56d 2682 }
d46de8a1 2683
3046b6db 2684 (void) invoke_main_loop(m,
a9fd4cd1
FB		 2685 &saved_rlimit_nofile,
2686 &saved_rlimit_memlock,
3046b6db
LP		 2687 &reexecute,
2688 &retval,
2689 &shutdown_verb,
2690 &fds,
2691 &switch_root_dir,
2692 &switch_root_init,
2693 &error_message);
f170852a 2694
60918275 2695finish:
b87c2aa6
ZJS		 2696 pager_close();
2697
92890452 2698 if (m) {
65224c1d 2699 arg_reboot_watchdog = m->reboot_watchdog;
acafd7d8 2700 arg_kexec_watchdog = m->kexec_watchdog;
92890452
LP		 2701 m = manager_free(m);
2702 }
60918275 2703
fb39af4c 2704 reset_arguments();
cc56fafe 2705 mac_selinux_finish();
b2bb3dbe 2706
3c7878f9
LP		 2707 if (reexecute)
2708 do_reexecute(argc, argv,
2709 &saved_rlimit_nofile,
2710 &saved_rlimit_memlock,
2711 fds,
2712 switch_root_dir,
2713 switch_root_init,
2714 &error_message); /* This only returns if reexecution failed */
a16e1123 2715
74ca738f 2716 arg_serialization = safe_fclose(arg_serialization);
2feceb5e 2717 fds = fdset_free(fds);
a16e1123 2718
0e06a031
LP		 2719 saved_env = strv_free(saved_env);
2720
349cc4a5 2721#if HAVE_VALGRIND_VALGRIND_H
54b434b1
LP		 2722 /* If we are PID 1 and running under valgrind, then let's exit
2723 * here explicitly. valgrind will only generate nice output on
2724 * exit(), not on exec(), hence let's do the former not the
2725 * latter here. */
8a2c1fbf
EJ		 2726 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND) {
2727 /* Cleanup watchdog_device strings for valgrind. We need them
2728 * in become_shutdown() so normally we cannot free them yet. */
2729 watchdog_free_device();
2730 arg_watchdog_device = mfree(arg_watchdog_device);
27fe58b7 2731 return retval;
8a2c1fbf 2732 }
54b434b1
LP		 2733#endif
2734
7e11a95e
EV		 2735#if HAS_FEATURE_ADDRESS_SANITIZER
2736 __lsan_do_leak_check();
2737#endif
2738
b9080b03 2739 if (shutdown_verb) {
7eb35049 2740 r = become_shutdown(shutdown_verb, retval);
4a36297c 2741 log_error_errno(r, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
9b9881d7 2742 error_message = "Failed to execute shutdown binary";
b9080b03
FF		 2743 }
2744
8a2c1fbf
EJ		 2745 watchdog_free_device();
2746 arg_watchdog_device = mfree(arg_watchdog_device);
2747
df0ff127 2748 if (getpid_cached() == 1) {
cb6531be
ZJS		 2749 if (error_message)
2750 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
1fc464f6 2751 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
bb259772
LP		 2752 "%s.", error_message);
2753 freeze_or_exit_or_reboot();
cb6531be 2754 }
c3b3c274 2755
60918275
LP		 2756 return retval;
2757}
Unnamed repository; edit this file 'description' to name the repository.