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