]> 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
core: set fs.file-max sysctl to LONG_MAX rather than ULONG_MAX
[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>
97c4f35c 6#include <signal.h>
3dfc9763
LP		 7#include <stdio.h>
8#include <string.h>
664f88a7 9#include <sys/mount.h>
3dfc9763 10#include <sys/prctl.h>
b9e74c39 11#include <sys/reboot.h>
3dfc9763
LP		 12#include <sys/stat.h>
13#include <unistd.h>
349cc4a5 14#if HAVE_SECCOMP
b64a3d86
LP		 15#include <seccomp.h>
16#endif
349cc4a5 17#if HAVE_VALGRIND_VALGRIND_H
3dfc9763
LP		 18#include <valgrind/valgrind.h>
19#endif
54b434b1 20
718db961 21#include "sd-bus.h"
cf0fbc49 22#include "sd-daemon.h"
b2e7486c 23#include "sd-messages.h"
3dfc9763 24
b5efdb8a 25#include "alloc-util.h"
d9d93745 26#include "architecture.h"
3dfc9763
LP		 27#include "build.h"
28#include "bus-error.h"
29#include "bus-util.h"
430f0182 30#include "capability-util.h"
a88c5b8a 31#include "cgroup-util.h"
24efb112 32#include "clock-util.h"
3dfc9763 33#include "conf-parser.h"
618234a5 34#include "cpu-set-util.h"
bbc1acab 35#include "dbus.h"
3dfc9763
LP		 36#include "dbus-manager.h"
37#include "def.h"
eee8b7ab 38#include "emergency-action.h"
3dfc9763 39#include "env-util.h"
57b7a260 40#include "exit-status.h"
3ffd4af2 41#include "fd-util.h"
3dfc9763 42#include "fdset.h"
718db961 43#include "fileio.h"
f97b34a6 44#include "format-util.h"
f4f15635 45#include "fs-util.h"
3dfc9763
LP		 46#include "hostname-setup.h"
47#include "ima-setup.h"
48#include "killall.h"
49#include "kmod-setup.h"
eefc66aa 50#include "limits-util.h"
d7b8eec7 51#include "load-fragment.h"
3dfc9763 52#include "log.h"
b6e66135 53#include "loopback-setup.h"
b6e66135 54#include "machine-id-setup.h"
3dfc9763
LP		 55#include "manager.h"
56#include "missing.h"
57#include "mount-setup.h"
d58ad743 58#include "os-util.h"
3dfc9763 59#include "pager.h"
6bedfcbb 60#include "parse-util.h"
7d5ceb64 61#include "path-util.h"
294bf0c3 62#include "pretty-print.h"
4e731273 63#include "proc-cmdline.h"
3dfc9763 64#include "process-util.h"
8869a0b4 65#include "raw-clone.h"
78f22b97 66#include "rlimit-util.h"
349cc4a5 67#if HAVE_SECCOMP
83f12b27
FS		 68#include "seccomp-util.h"
69#endif
b6e66135 70#include "selinux-setup.h"
3dfc9763
LP		 71#include "selinux-util.h"
72#include "signal-util.h"
ffbd2c4d 73#include "smack-setup.h"
3dfc9763 74#include "special.h"
8fcde012 75#include "stat-util.h"
15a5e950 76#include "stdio-util.h"
3dfc9763
LP		 77#include "strv.h"
78#include "switch-root.h"
a8b627aa 79#include "sysctl-util.h"
3dfc9763 80#include "terminal-util.h"
8612da97 81#include "umask-util.h"
b1d4f8e1 82#include "user-util.h"
9ce17593 83#include "util.h"
3dfc9763
LP		 84#include "virt.h"
85#include "watchdog.h"
b6e66135 86
7e11a95e
EV		 87#if HAS_FEATURE_ADDRESS_SANITIZER
88#include <sanitizer/lsan_interface.h>
89#endif
90
f170852a
LP		 91static enum {
92 ACTION_RUN,
e965d56d 93 ACTION_HELP,
9ba0bc4e 94 ACTION_VERSION,
e537352b 95 ACTION_TEST,
bbc1acab
YW		 96 ACTION_DUMP_CONFIGURATION_ITEMS,
97 ACTION_DUMP_BUS_PROPERTIES,
fa0f4d8a 98} arg_action = ACTION_RUN;
fb39af4c
ZJS		 99
100/* Those variables are initalized to 0 automatically, so we avoid uninitialized memory access.
101 * Real defaults are assigned in reset_arguments() below. */
102static char *arg_default_unit;
103static bool arg_system;
104static bool arg_dump_core;
105static int arg_crash_chvt;
106static bool arg_crash_shell;
107static bool arg_crash_reboot;
108static char *arg_confirm_spawn;
109static ShowStatus arg_show_status;
110static bool arg_switched_root;
111static PagerFlags arg_pager_flags;
112static bool arg_service_watchdogs;
113static ExecOutput arg_default_std_output;
114static ExecOutput arg_default_std_error;
115static usec_t arg_default_restart_usec;
116static usec_t arg_default_timeout_start_usec;
117static usec_t arg_default_timeout_stop_usec;
118static usec_t arg_default_timeout_abort_usec;
119static bool arg_default_timeout_abort_set;
120static usec_t arg_default_start_limit_interval;
121static unsigned arg_default_start_limit_burst;
122static usec_t arg_runtime_watchdog;
123static usec_t arg_shutdown_watchdog;
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;
140static uint64_t arg_default_tasks_max;
141static sd_id128_t arg_machine_id;
142static EmergencyAction arg_cad_burst_action;
143static OOMPolicy arg_default_oom_policy;
144static CPUSet arg_cpu_affinity;
61fbbac1 145
470a5e6d 146static int parse_configuration(void);
4fc935ca 147
bb259772
LP		 148_noreturn_ static void freeze_or_exit_or_reboot(void) {
149
c3b6a348
LP		 150 /* If we are running in a container, let's prefer exiting, after all we can propagate an exit code to
151 * the container manager, and thus inform it that something went wrong. */
bb259772
LP		 152 if (detect_container() > 0) {
153 log_emergency("Exiting PID 1...");
c3b6a348 154 _exit(EXIT_EXCEPTION);
bb259772 155 }
b9e74c39
LP		 156
157 if (arg_crash_reboot) {
158 log_notice("Rebooting in 10s...");
159 (void) sleep(10);
160
161 log_notice("Rebooting now...");
162 (void) reboot(RB_AUTOBOOT);
163 log_emergency_errno(errno, "Failed to reboot: %m");
164 }
165
166 log_emergency("Freezing execution.");
167 freeze();
168}
169
848e863a 170_noreturn_ static void crash(int sig) {
7d06dad9
MS		 171 struct sigaction sa;
172 pid_t pid;
97c4f35c 173
df0ff127 174 if (getpid_cached() != 1)
abb26902 175 /* Pass this on immediately, if this is not PID 1 */
92ca4cac 176 (void) raise(sig);
abb26902 177 else if (!arg_dump_core)
4104970e 178 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
97c4f35c 179 else {
7d06dad9 180 sa = (struct sigaction) {
189d5bac 181 .sa_handler = nop_signal_handler,
b92bea5d
ZJS		 182 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
183 };
97c4f35c 184
6f5e3f35 185 /* We want to wait for the core process, hence let's enable SIGCHLD */
92ca4cac 186 (void) sigaction(SIGCHLD, &sa, NULL);
6f5e3f35 187
8869a0b4 188 pid = raw_clone(SIGCHLD);
e62d8c39 189 if (pid < 0)
56f64d95 190 log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
97c4f35c 191 else if (pid == 0) {
97c4f35c 192 /* Enable default signal handler for core dump */
15a90032 193
92ca4cac
LP		 194 sa = (struct sigaction) {
195 .sa_handler = SIG_DFL,
196 };
197 (void) sigaction(sig, &sa, NULL);
97c4f35c 198
15a90032
LP		 199 /* Don't limit the coredump size */
200 (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
97c4f35c
LP		 201
202 /* Just to be sure... */
e62d9b81 203 (void) chdir("/");
97c4f35c
LP		 204
205 /* Raise the signal again */
ee05e779 206 pid = raw_getpid();
92ca4cac 207 (void) kill(pid, sig); /* raise() would kill the parent */
97c4f35c
LP		 208
209 assert_not_reached("We shouldn't be here...");
bb85a582 210 _exit(EXIT_EXCEPTION);
4fc935ca 211 } else {
8e12a6ae
LP		 212 siginfo_t status;
213 int r;
4fc935ca
LP		 214
215 /* Order things nicely. */
e62d8c39
ZJS		 216 r = wait_for_terminate(pid, &status);
217 if (r < 0)
da927ba9 218 log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
8e12a6ae 219 else if (status.si_code != CLD_DUMPED)
ee05e779
ZJS		 220 log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
221 signal_to_string(sig),
222 pid, sigchld_code_to_string(status.si_code),
223 status.si_status,
224 strna(status.si_code == CLD_EXITED
38107f5a 225 ? exit_status_to_string(status.si_status, EXIT_STATUS_MINIMAL)
ee05e779 226 : signal_to_string(status.si_status)));
4fc935ca 227 else
4104970e 228 log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
97c4f35c
LP		 229 }
230 }
231
b9e74c39 232 if (arg_crash_chvt >= 0)
92ca4cac 233 (void) chvt(arg_crash_chvt);
601f6a1e 234
7d06dad9
MS		 235 sa = (struct sigaction) {
236 .sa_handler = SIG_IGN,
237 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
238 };
239
240 /* Let the kernel reap children for us */
241 (void) sigaction(SIGCHLD, &sa, NULL);
8c43883a 242
7d06dad9 243 if (arg_crash_shell) {
b9e74c39 244 log_notice("Executing crash shell in 10s...");
92ca4cac 245 (void) sleep(10);
4fc935ca 246
8869a0b4 247 pid = raw_clone(SIGCHLD);
cd3bd60a 248 if (pid < 0)
56f64d95 249 log_emergency_errno(errno, "Failed to fork off crash shell: %m");
6f5e3f35 250 else if (pid == 0) {
b9e74c39 251 (void) setsid();
92ca4cac 252 (void) make_console_stdio();
595225af 253 (void) rlimit_nofile_safe();
92ca4cac 254 (void) execle("/bin/sh", "/bin/sh", NULL, environ);
6f5e3f35 255
ee05e779 256 log_emergency_errno(errno, "execle() failed: %m");
bb85a582 257 _exit(EXIT_EXCEPTION);
b9e74c39
LP		 258 } else {
259 log_info("Spawned crash shell as PID "PID_FMT".", pid);
4cf0b03b 260 (void) wait_for_terminate(pid, NULL);
b9e74c39 261 }
4fc935ca
LP		 262 }
263
bb259772 264 freeze_or_exit_or_reboot();
97c4f35c
LP		 265}
266
267static void install_crash_handler(void) {
297d563d 268 static const struct sigaction sa = {
b92bea5d 269 .sa_handler = crash,
297d563d 270 .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
b92bea5d 271 };
297d563d 272 int r;
97c4f35c 273
297d563d
LP		 274 /* We ignore the return value here, since, we don't mind if we
275 * cannot set up a crash handler */
276 r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
277 if (r < 0)
278 log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
97c4f35c 279}
f170852a 280
56d96fc0
LP		 281static int console_setup(void) {
282 _cleanup_close_ int tty_fd = -1;
283 int r;
80876c20 284
512947d4 285 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
23bbb0de
MS		 286 if (tty_fd < 0)
287 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
80876c20 288
56d96fc0
LP		 289 /* We don't want to force text mode. plymouth may be showing
290 * pictures already from initrd. */
512947d4 291 r = reset_terminal_fd(tty_fd, false);
23bbb0de
MS		 292 if (r < 0)
293 return log_error_errno(r, "Failed to reset /dev/console: %m");
843d2643 294
56d96fc0 295 return 0;
80876c20
LP		 296}
297
b9e74c39
LP		 298static int parse_crash_chvt(const char *value) {
299 int b;
300
301 if (safe_atoi(value, &arg_crash_chvt) >= 0)
302 return 0;
303
304 b = parse_boolean(value);
305 if (b < 0)
306 return b;
307
308 if (b > 0)
309 arg_crash_chvt = 0; /* switch to where kmsg goes */
310 else
311 arg_crash_chvt = -1; /* turn off switching */
312
313 return 0;
314}
315
7d5ceb64
FB		 316static int parse_confirm_spawn(const char *value, char **console) {
317 char *s;
318 int r;
319
320 r = value ? parse_boolean(value) : 1;
321 if (r == 0) {
322 *console = NULL;
323 return 0;
324 }
325
326 if (r > 0) /* on with default tty */
327 s = strdup("/dev/console");
328 else if (is_path(value)) /* on with fully qualified path */
329 s = strdup(value);
330 else /* on with only a tty file name, not a fully qualified path */
331 s = strjoin("/dev/", value);
332 if (!s)
333 return -ENOMEM;
5b65ae15 334
7d5ceb64
FB		 335 *console = s;
336 return 0;
337}
338
ee48dbd5 339static int set_machine_id(const char *m) {
e042eab7 340 sd_id128_t t;
8b26cdbd 341 assert(m);
ee48dbd5 342
e042eab7 343 if (sd_id128_from_string(m, &t) < 0)
ee48dbd5
NC		 344 return -EINVAL;
345
e042eab7 346 if (sd_id128_is_null(t))
ee48dbd5
NC		 347 return -EINVAL;
348
e042eab7 349 arg_machine_id = t;
ee48dbd5
NC		 350 return 0;
351}
352
96287a49 353static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
f170852a 354
059cb385 355 int r;
f170852a 356
059cb385 357 assert(key);
5192bd19 358
1d84ad94 359 if (STR_IN_SET(key, "systemd.unit", "rd.systemd.unit")) {
bf4df7c3 360
1d84ad94
LP		 361 if (proc_cmdline_value_missing(key, value))
362 return 0;
bf4df7c3 363
1d84ad94
LP		 364 if (!unit_name_is_valid(value, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE))
365 log_warning("Unit name specified on %s= is not valid, ignoring: %s", key, value);
366 else if (in_initrd() == !!startswith(key, "rd.")) {
367 if (free_and_strdup(&arg_default_unit, value) < 0)
368 return log_oom();
369 }
f170852a 370
1d84ad94 371 } else if (proc_cmdline_key_streq(key, "systemd.dump_core")) {
4fc935ca 372
1d84ad94 373 r = value ? parse_boolean(value) : true;
fb472900 374 if (r < 0)
5e1ee764 375 log_warning_errno(r, "Failed to parse dump core switch %s, ignoring: %m", value);
4fc935ca 376 else
fa0f4d8a 377 arg_dump_core = r;
4fc935ca 378
c6885f5f
FB		 379 } else if (proc_cmdline_key_streq(key, "systemd.early_core_pattern")) {
380
381 if (proc_cmdline_value_missing(key, value))
382 return 0;
383
384 if (path_is_absolute(value))
385 (void) parse_path_argument_and_warn(value, false, &arg_early_core_pattern);
386 else
387 log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value);
388
1d84ad94 389 } else if (proc_cmdline_key_streq(key, "systemd.crash_chvt")) {
b9e74c39 390
1d84ad94
LP		 391 if (!value)
392 arg_crash_chvt = 0; /* turn on */
5e1ee764
YW		 393 else {
394 r = parse_crash_chvt(value);
395 if (r < 0)
396 log_warning_errno(r, "Failed to parse crash chvt switch %s, ignoring: %m", value);
397 }
b9e74c39 398
1d84ad94 399 } else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) {
4fc935ca 400
1d84ad94 401 r = value ? parse_boolean(value) : true;
fb472900 402 if (r < 0)
5e1ee764 403 log_warning_errno(r, "Failed to parse crash shell switch %s, ignoring: %m", value);
4fc935ca 404 else
fa0f4d8a 405 arg_crash_shell = r;
5e7ee61c 406
1d84ad94 407 } else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) {
5e7ee61c 408
1d84ad94 409 r = value ? parse_boolean(value) : true;
b9e74c39 410 if (r < 0)
5e1ee764 411 log_warning_errno(r, "Failed to parse crash reboot switch %s, ignoring: %m", value);
5e7ee61c 412 else
b9e74c39 413 arg_crash_reboot = r;
5e7ee61c 414
1d84ad94
LP		 415 } else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) {
416 char *s;
7d5ceb64 417
1d84ad94 418 r = parse_confirm_spawn(value, &s);
059cb385 419 if (r < 0)
5e1ee764
YW		 420 log_warning_errno(r, "Failed to parse confirm_spawn switch %s, ignoring: %m", value);
421 else
422 free_and_replace(arg_confirm_spawn, s);
601f6a1e 423
2a12e32e
JK		 424 } else if (proc_cmdline_key_streq(key, "systemd.service_watchdogs")) {
425
426 r = value ? parse_boolean(value) : true;
427 if (r < 0)
5e1ee764 428 log_warning_errno(r, "Failed to parse service watchdog switch %s, ignoring: %m", value);
2a12e32e
JK		 429 else
430 arg_service_watchdogs = r;
431
1d84ad94 432 } else if (proc_cmdline_key_streq(key, "systemd.show_status")) {
9e58ff9c 433
1d84ad94
LP		 434 if (value) {
435 r = parse_show_status(value, &arg_show_status);
436 if (r < 0)
5e1ee764 437 log_warning_errno(r, "Failed to parse show status switch %s, ignoring: %m", value);
1d84ad94
LP		 438 } else
439 arg_show_status = SHOW_STATUS_YES;
059cb385 440
1d84ad94
LP		 441 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) {
442
443 if (proc_cmdline_value_missing(key, value))
444 return 0;
0a494f1f 445
059cb385 446 r = exec_output_from_string(value);
fb472900 447 if (r < 0)
5e1ee764 448 log_warning_errno(r, "Failed to parse default standard output switch %s, ignoring: %m", value);
0a494f1f
LP		 449 else
450 arg_default_std_output = r;
0a494f1f 451
1d84ad94
LP		 452 } else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) {
453
454 if (proc_cmdline_value_missing(key, value))
455 return 0;
059cb385
LP		 456
457 r = exec_output_from_string(value);
fb472900 458 if (r < 0)
5e1ee764 459 log_warning_errno(r, "Failed to parse default standard error switch %s, ignoring: %m", value);
0a494f1f
LP		 460 else
461 arg_default_std_error = r;
9e7c5357 462
1d84ad94
LP		 463 } else if (streq(key, "systemd.setenv")) {
464
465 if (proc_cmdline_value_missing(key, value))
466 return 0;
059cb385
LP		 467
468 if (env_assignment_is_valid(value)) {
e21fea24
KS		 469 char **env;
470
059cb385 471 env = strv_env_set(arg_default_environment, value);
1d84ad94
LP		 472 if (!env)
473 return log_oom();
474
475 arg_default_environment = env;
e21fea24 476 } else
059cb385 477 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
9e58ff9c 478
1d84ad94
LP		 479 } else if (proc_cmdline_key_streq(key, "systemd.machine_id")) {
480
481 if (proc_cmdline_value_missing(key, value))
482 return 0;
483
484 r = set_machine_id(value);
485 if (r < 0)
5e1ee764 486 log_warning_errno(r, "MachineID '%s' is not valid, ignoring: %m", value);
ee48dbd5 487
1d84ad94
LP		 488 } else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) {
489
490 if (proc_cmdline_value_missing(key, value))
491 return 0;
492
493 r = parse_sec(value, &arg_default_timeout_start_usec);
494 if (r < 0)
5e1ee764 495 log_warning_errno(r, "Failed to parse default start timeout '%s', ignoring: %m", value);
1d84ad94
LP		 496
497 if (arg_default_timeout_start_usec <= 0)
498 arg_default_timeout_start_usec = USEC_INFINITY;
ee48dbd5 499
8a2c1fbf
EJ		 500 } else if (proc_cmdline_key_streq(key, "systemd.watchdog_device")) {
501
502 if (proc_cmdline_value_missing(key, value))
503 return 0;
504
c6885f5f 505 (void) parse_path_argument_and_warn(value, false, &arg_watchdog_device);
8a2c1fbf 506
059cb385 507 } else if (streq(key, "quiet") && !value) {
d7b15e0a 508
7a293242 509 if (arg_show_status == _SHOW_STATUS_INVALID)
d450b6f2 510 arg_show_status = SHOW_STATUS_AUTO;
059cb385
LP		 511
512 } else if (streq(key, "debug") && !value) {
d7b15e0a 513
1de1c9c3
LP		 514 /* Note that log_parse_environment() handles 'debug'
515 * too, and sets the log level to LOG_DEBUG. */
d7b15e0a 516
75f86906 517 if (detect_container() > 0)
b2103dcc 518 log_set_target(LOG_TARGET_CONSOLE);
059cb385 519
dcd61450 520 } else if (!value) {
e2c9a131 521 const char *target;
f170852a
LP		 522
523 /* SysV compatibility */
e2c9a131
EV		 524 target = runlevel_to_target(key);
525 if (target)
526 return free_and_strdup(&arg_default_unit, target);
f170852a
LP		 527 }
528
529 return 0;
530}
531
e8e581bf
ZJS		 532#define DEFINE_SETTER(name, func, descr) \
533 static int name(const char *unit, \
534 const char *filename, \
535 unsigned line, \
536 const char *section, \
71a61510 537 unsigned section_line, \
e8e581bf
ZJS		 538 const char *lvalue, \
539 int ltype, \
540 const char *rvalue, \
541 void *data, \
542 void *userdata) { \
543 \
544 int r; \
545 \
546 assert(filename); \
547 assert(lvalue); \
548 assert(rvalue); \
549 \
550 r = func(rvalue); \
551 if (r < 0) \
d1cefe0a
LP		 552 log_syntax(unit, LOG_ERR, filename, line, r, \
553 "Invalid " descr "'%s': %m", \
554 rvalue); \
e8e581bf
ZJS		 555 \
556 return 0; \
557 }
487393e9 558
a6ecbf83
FB		 559DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level");
560DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target");
561DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" );
562DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location");
487393e9 563
4d7213b2
LP		 564static int config_parse_cpu_affinity2(
565 const char *unit,
566 const char *filename,
567 unsigned line,
568 const char *section,
569 unsigned section_line,
570 const char *lvalue,
571 int ltype,
572 const char *rvalue,
573 void *data,
574 void *userdata) {
487393e9 575
61fbbac1 576 CPUSet *affinity = data;
487393e9 577
61fbbac1 578 assert(affinity);
487393e9 579
61fbbac1 580 (void) parse_cpu_set_extend(rvalue, affinity, true, unit, filename, line, lvalue);
487393e9 581
487393e9
LP		 582 return 0;
583}
584
4d7213b2
LP		 585static int config_parse_show_status(
586 const char* unit,
587 const char *filename,
588 unsigned line,
589 const char *section,
590 unsigned section_line,
591 const char *lvalue,
592 int ltype,
593 const char *rvalue,
594 void *data,
595 void *userdata) {
596
597 int k;
598 ShowStatus *b = data;
599
600 assert(filename);
601 assert(lvalue);
602 assert(rvalue);
603 assert(data);
604
605 k = parse_show_status(rvalue, b);
606 if (k < 0) {
12ca818f 607 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
4d7213b2
LP		 608 return 0;
609 }
610
611 return 0;
612}
613
5db98187
LP		 614static int config_parse_output_restricted(
615 const char* unit,
616 const char *filename,
617 unsigned line,
618 const char *section,
619 unsigned section_line,
620 const char *lvalue,
621 int ltype,
622 const char *rvalue,
623 void *data,
624 void *userdata) {
625
626 ExecOutput t, *eo = data;
627
628 assert(filename);
629 assert(lvalue);
630 assert(rvalue);
631 assert(data);
632
633 t = exec_output_from_string(rvalue);
634 if (t < 0) {
635 log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse output type, ignoring: %s", rvalue);
636 return 0;
637 }
638
566b7d23
ZD		 639 if (IN_SET(t, EXEC_OUTPUT_SOCKET, EXEC_OUTPUT_NAMED_FD, EXEC_OUTPUT_FILE, EXEC_OUTPUT_FILE_APPEND)) {
640 log_syntax(unit, LOG_ERR, filename, line, 0, "Standard output types socket, fd:, file:, append: are not supported as defaults, ignoring: %s", rvalue);
5db98187
LP		 641 return 0;
642 }
643
644 *eo = t;
645 return 0;
646}
647
b9e74c39
LP		 648static int config_parse_crash_chvt(
649 const char* unit,
650 const char *filename,
651 unsigned line,
652 const char *section,
653 unsigned section_line,
654 const char *lvalue,
655 int ltype,
656 const char *rvalue,
657 void *data,
658 void *userdata) {
659
660 int r;
661
662 assert(filename);
663 assert(lvalue);
664 assert(rvalue);
b9e74c39
LP		 665
666 r = parse_crash_chvt(rvalue);
667 if (r < 0) {
668 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
669 return 0;
670 }
671
672 return 0;
673}
674
dc653bf4
JK		 675static int config_parse_timeout_abort(
676 const char* unit,
677 const char *filename,
678 unsigned line,
679 const char *section,
680 unsigned section_line,
681 const char *lvalue,
682 int ltype,
683 const char *rvalue,
684 void *data,
685 void *userdata) {
686
687 int r;
688
689 assert(filename);
690 assert(lvalue);
691 assert(rvalue);
692
693 rvalue += strspn(rvalue, WHITESPACE);
694 if (isempty(rvalue)) {
695 arg_default_timeout_abort_set = false;
696 return 0;
697 }
698
699 r = parse_sec(rvalue, &arg_default_timeout_abort_usec);
700 if (r < 0) {
701 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse DefaultTimeoutAbortSec= setting, ignoring: %s", rvalue);
702 return 0;
703 }
704
705 arg_default_timeout_abort_set = true;
706 return 0;
707}
708
487393e9
LP		 709static int parse_config_file(void) {
710
f975e971 711 const ConfigTableItem items[] = {
d3b1c508
LP		 712 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
713 { "Manager", "LogTarget", config_parse_target, 0, NULL },
714 { "Manager", "LogColor", config_parse_color, 0, NULL },
715 { "Manager", "LogLocation", config_parse_location, 0, NULL },
716 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
b9e74c39
LP		 717 { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL },
718 { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL },
d3b1c508 719 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
b9e74c39 720 { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot },
d3b1c508 721 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
61fbbac1 722 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, &arg_cpu_affinity },
143fadf3 723 { "Manager", "JoinControllers", config_parse_warn_compat, DISABLED_CONFIGURATION, NULL },
d3b1c508
LP		 724 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
725 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
8a2c1fbf 726 { "Manager", "WatchdogDevice", config_parse_path, 0, &arg_watchdog_device },
a103496c 727 { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set },
39362f6f 728 { "Manager", "NoNewPrivileges", config_parse_bool, 0, &arg_no_new_privs },
349cc4a5 729#if HAVE_SECCOMP
d3b1c508 730 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
89fffa27 731#endif
d3b1c508 732 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
bd8f585b 733 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
5db98187
LP		 734 { "Manager", "DefaultStandardOutput", config_parse_output_restricted,0, &arg_default_std_output },
735 { "Manager", "DefaultStandardError", config_parse_output_restricted,0, &arg_default_std_error },
d3b1c508
LP		 736 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
737 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
dc653bf4 738 { "Manager", "DefaultTimeoutAbortSec", config_parse_timeout_abort, 0, NULL },
d3b1c508 739 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
f0367da7
LP		 740 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */
741 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec, 0, &arg_default_start_limit_interval },
d3b1c508
LP		 742 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
743 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
4f424df7
LP		 744 { "Manager", "DefaultLimitCPU", config_parse_rlimit, RLIMIT_CPU, arg_default_rlimit },
745 { "Manager", "DefaultLimitFSIZE", config_parse_rlimit, RLIMIT_FSIZE, arg_default_rlimit },
746 { "Manager", "DefaultLimitDATA", config_parse_rlimit, RLIMIT_DATA, arg_default_rlimit },
747 { "Manager", "DefaultLimitSTACK", config_parse_rlimit, RLIMIT_STACK, arg_default_rlimit },
748 { "Manager", "DefaultLimitCORE", config_parse_rlimit, RLIMIT_CORE, arg_default_rlimit },
749 { "Manager", "DefaultLimitRSS", config_parse_rlimit, RLIMIT_RSS, arg_default_rlimit },
750 { "Manager", "DefaultLimitNOFILE", config_parse_rlimit, RLIMIT_NOFILE, arg_default_rlimit },
751 { "Manager", "DefaultLimitAS", config_parse_rlimit, RLIMIT_AS, arg_default_rlimit },
752 { "Manager", "DefaultLimitNPROC", config_parse_rlimit, RLIMIT_NPROC, arg_default_rlimit },
753 { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit, RLIMIT_MEMLOCK, arg_default_rlimit },
754 { "Manager", "DefaultLimitLOCKS", config_parse_rlimit, RLIMIT_LOCKS, arg_default_rlimit },
755 { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit, RLIMIT_SIGPENDING, arg_default_rlimit },
756 { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit, RLIMIT_MSGQUEUE, arg_default_rlimit },
757 { "Manager", "DefaultLimitNICE", config_parse_rlimit, RLIMIT_NICE, arg_default_rlimit },
758 { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit, RLIMIT_RTPRIO, arg_default_rlimit },
759 { "Manager", "DefaultLimitRTTIME", config_parse_rlimit, RLIMIT_RTTIME, arg_default_rlimit },
a88c5b8a 760 { "Manager", "DefaultCPUAccounting", config_parse_tristate, 0, &arg_default_cpu_accounting },
13c31542 761 { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting },
377bfd2d 762 { "Manager", "DefaultIPAccounting", config_parse_bool, 0, &arg_default_ip_accounting },
085afe36
LP		 763 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
764 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
03a7b521 765 { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting },
0af20ea2 766 { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max },
ae8c7939 767 { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action, 0, &arg_cad_burst_action },
afcfaa69 768 { "Manager", "DefaultOOMPolicy", config_parse_oom_policy, 0, &arg_default_oom_policy },
d3b1c508 769 {}
487393e9
LP		 770 };
771
1b907b5c 772 const char *fn, *conf_dirs_nulstr;
487393e9 773
463d0d15 774 fn = arg_system ?
75eb6154
LP		 775 PKGSYSCONFDIR "/system.conf" :
776 PKGSYSCONFDIR "/user.conf";
777
463d0d15 778 conf_dirs_nulstr = arg_system ?
75eb6154
LP		 779 CONF_PATHS_NULSTR("systemd/system.conf.d") :
780 CONF_PATHS_NULSTR("systemd/user.conf.d");
781
bcde742e 782 (void) config_parse_many_nulstr(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, CONFIG_PARSE_WARN, NULL);
36c16a7c
LP		 783
784 /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
785 * like everywhere else. */
786 if (arg_default_timeout_start_usec <= 0)
787 arg_default_timeout_start_usec = USEC_INFINITY;
788 if (arg_default_timeout_stop_usec <= 0)
789 arg_default_timeout_stop_usec = USEC_INFINITY;
487393e9 790
487393e9
LP		 791 return 0;
792}
793
85cb4151 794static void set_manager_defaults(Manager *m) {
06af2a04
TB		 795
796 assert(m);
797
5b65ae15
LP		 798 /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
799 * affect the manager itself, but are just what newly allocated units will have set if they haven't set
800 * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
801
06af2a04
TB		 802 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
803 m->default_std_output = arg_default_std_output;
804 m->default_std_error = arg_default_std_error;
805 m->default_timeout_start_usec = arg_default_timeout_start_usec;
806 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
dc653bf4
JK		 807 m->default_timeout_abort_usec = arg_default_timeout_abort_usec;
808 m->default_timeout_abort_set = arg_default_timeout_abort_set;
06af2a04
TB		 809 m->default_restart_usec = arg_default_restart_usec;
810 m->default_start_limit_interval = arg_default_start_limit_interval;
811 m->default_start_limit_burst = arg_default_start_limit_burst;
a88c5b8a
CD		 812
813 /* On 4.15+ with unified hierarchy, CPU accounting is essentially free as it doesn't require the CPU
814 * controller to be enabled, so the default is to enable it unless we got told otherwise. */
815 if (arg_default_cpu_accounting >= 0)
816 m->default_cpu_accounting = arg_default_cpu_accounting;
817 else
818 m->default_cpu_accounting = cpu_accounting_is_cheap();
819
13c31542 820 m->default_io_accounting = arg_default_io_accounting;
377bfd2d 821 m->default_ip_accounting = arg_default_ip_accounting;
06af2a04
TB		 822 m->default_blockio_accounting = arg_default_blockio_accounting;
823 m->default_memory_accounting = arg_default_memory_accounting;
03a7b521 824 m->default_tasks_accounting = arg_default_tasks_accounting;
0af20ea2 825 m->default_tasks_max = arg_default_tasks_max;
afcfaa69 826 m->default_oom_policy = arg_default_oom_policy;
06af2a04 827
79a224c4
LP		 828 (void) manager_set_default_rlimits(m, arg_default_rlimit);
829
830 (void) manager_default_environment(m);
831 (void) manager_transient_environment_add(m, arg_default_environment);
06af2a04
TB		 832}
833
7b46fc6a
LP		 834static void set_manager_settings(Manager *m) {
835
836 assert(m);
837
5b65ae15
LP		 838 /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
839 * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
840
7b46fc6a 841 m->confirm_spawn = arg_confirm_spawn;
2a12e32e 842 m->service_watchdogs = arg_service_watchdogs;
7b46fc6a
LP		 843 m->runtime_watchdog = arg_runtime_watchdog;
844 m->shutdown_watchdog = arg_shutdown_watchdog;
845 m->cad_burst_action = arg_cad_burst_action;
846
847 manager_set_show_status(m, arg_show_status);
848}
849
f170852a 850static int parse_argv(int argc, char *argv[]) {
f170852a
LP		 851 enum {
852 ARG_LOG_LEVEL = 0x100,
853 ARG_LOG_TARGET,
bbe63281
LP		 854 ARG_LOG_COLOR,
855 ARG_LOG_LOCATION,
2f198e2f 856 ARG_UNIT,
edb9aaa8 857 ARG_SYSTEM,
af2d49f7 858 ARG_USER,
e537352b 859 ARG_TEST,
b87c2aa6 860 ARG_NO_PAGER,
9ba0bc4e 861 ARG_VERSION,
80876c20 862 ARG_DUMP_CONFIGURATION_ITEMS,
bbc1acab 863 ARG_DUMP_BUS_PROPERTIES,
9e58ff9c 864 ARG_DUMP_CORE,
b9e74c39 865 ARG_CRASH_CHVT,
9e58ff9c 866 ARG_CRASH_SHELL,
b9e74c39 867 ARG_CRASH_REBOOT,
a16e1123 868 ARG_CONFIRM_SPAWN,
9e58ff9c 869 ARG_SHOW_STATUS,
4288f619 870 ARG_DESERIALIZE,
2660882b 871 ARG_SWITCHED_ROOT,
0a494f1f 872 ARG_DEFAULT_STD_OUTPUT,
ee48dbd5 873 ARG_DEFAULT_STD_ERROR,
2a12e32e
JK		 874 ARG_MACHINE_ID,
875 ARG_SERVICE_WATCHDOGS,
f170852a
LP		 876 };
877
878 static const struct option options[] = {
a16e1123
LP		 879 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
880 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
bbe63281
LP		 881 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
882 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
2f198e2f 883 { "unit", required_argument, NULL, ARG_UNIT },
edb9aaa8 884 { "system", no_argument, NULL, ARG_SYSTEM },
af2d49f7 885 { "user", no_argument, NULL, ARG_USER },
a16e1123 886 { "test", no_argument, NULL, ARG_TEST },
b87c2aa6 887 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
a16e1123 888 { "help", no_argument, NULL, 'h' },
9ba0bc4e 889 { "version", no_argument, NULL, ARG_VERSION },
a16e1123 890 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
bbc1acab 891 { "dump-bus-properties", no_argument, NULL, ARG_DUMP_BUS_PROPERTIES },
a5d87bf0 892 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
b9e74c39 893 { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT },
a5d87bf0 894 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
b9e74c39 895 { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT },
a5d87bf0 896 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
6e98720f 897 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
a16e1123 898 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
2660882b 899 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
0a494f1f
LP		 900 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
901 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
ee48dbd5 902 { "machine-id", required_argument, NULL, ARG_MACHINE_ID },
2a12e32e 903 { "service-watchdogs", required_argument, NULL, ARG_SERVICE_WATCHDOGS },
fb472900 904 {}
f170852a
LP		 905 };
906
907 int c, r;
908
909 assert(argc >= 1);
910 assert(argv);
911
df0ff127 912 if (getpid_cached() == 1)
b770165a
LP		 913 opterr = 0;
914
099663ff 915 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
f170852a
LP		 916
917 switch (c) {
918
919 case ARG_LOG_LEVEL:
fb472900 920 r = log_set_max_level_from_string(optarg);
2b5107e1
ZJS		 921 if (r < 0)
922 return log_error_errno(r, "Failed to parse log level \"%s\": %m", optarg);
f170852a
LP		 923
924 break;
925
926 case ARG_LOG_TARGET:
fb472900 927 r = log_set_target_from_string(optarg);
2b5107e1
ZJS		 928 if (r < 0)
929 return log_error_errno(r, "Failed to parse log target \"%s\": %m", optarg);
f170852a
LP		 930
931 break;
932
bbe63281
LP		 933 case ARG_LOG_COLOR:
934
d0b170c8 935 if (optarg) {
fb472900 936 r = log_show_color_from_string(optarg);
2b5107e1
ZJS		 937 if (r < 0)
938 return log_error_errno(r, "Failed to parse log color setting \"%s\": %m",
939 optarg);
d0b170c8
LP		 940 } else
941 log_show_color(true);
bbe63281
LP		 942
943 break;
944
945 case ARG_LOG_LOCATION:
d0b170c8 946 if (optarg) {
fb472900 947 r = log_show_location_from_string(optarg);
2b5107e1
ZJS		 948 if (r < 0)
949 return log_error_errno(r, "Failed to parse log location setting \"%s\": %m",
950 optarg);
d0b170c8
LP		 951 } else
952 log_show_location(true);
bbe63281
LP		 953
954 break;
955
0a494f1f 956 case ARG_DEFAULT_STD_OUTPUT:
fb472900 957 r = exec_output_from_string(optarg);
2b5107e1
ZJS		 958 if (r < 0)
959 return log_error_errno(r, "Failed to parse default standard output setting \"%s\": %m",
960 optarg);
961 arg_default_std_output = r;
0a494f1f
LP		 962 break;
963
964 case ARG_DEFAULT_STD_ERROR:
fb472900 965 r = exec_output_from_string(optarg);
2b5107e1
ZJS		 966 if (r < 0)
967 return log_error_errno(r, "Failed to parse default standard error output setting \"%s\": %m",
968 optarg);
969 arg_default_std_error = r;
0a494f1f
LP		 970 break;
971
2f198e2f 972 case ARG_UNIT:
e6e242ad 973 r = free_and_strdup(&arg_default_unit, optarg);
23bbb0de 974 if (r < 0)
2b5107e1 975 return log_error_errno(r, "Failed to set default unit \"%s\": %m", optarg);
f170852a
LP		 976
977 break;
978
edb9aaa8 979 case ARG_SYSTEM:
463d0d15 980 arg_system = true;
edb9aaa8 981 break;
a5dab5ce 982
af2d49f7 983 case ARG_USER:
463d0d15 984 arg_system = false;
a5dab5ce 985 break;
a5dab5ce 986
e965d56d 987 case ARG_TEST:
fa0f4d8a 988 arg_action = ACTION_TEST;
b87c2aa6
ZJS		 989 break;
990
991 case ARG_NO_PAGER:
0221d68a 992 arg_pager_flags |= PAGER_DISABLE;
e965d56d
LP		 993 break;
994
9ba0bc4e
ZJS		 995 case ARG_VERSION:
996 arg_action = ACTION_VERSION;
997 break;
998
e537352b 999 case ARG_DUMP_CONFIGURATION_ITEMS:
fa0f4d8a 1000 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
e537352b
LP		 1001 break;
1002
bbc1acab
YW		 1003 case ARG_DUMP_BUS_PROPERTIES:
1004 arg_action = ACTION_DUMP_BUS_PROPERTIES;
1005 break;
1006
9e58ff9c 1007 case ARG_DUMP_CORE:
b9e74c39
LP		 1008 if (!optarg)
1009 arg_dump_core = true;
1010 else {
1011 r = parse_boolean(optarg);
1012 if (r < 0)
2b5107e1
ZJS		 1013 return log_error_errno(r, "Failed to parse dump core boolean: \"%s\": %m",
1014 optarg);
b9e74c39 1015 arg_dump_core = r;
a5d87bf0 1016 }
b9e74c39
LP		 1017 break;
1018
1019 case ARG_CRASH_CHVT:
1020 r = parse_crash_chvt(optarg);
1021 if (r < 0)
2b5107e1
ZJS		 1022 return log_error_errno(r, "Failed to parse crash virtual terminal index: \"%s\": %m",
1023 optarg);
9e58ff9c
LP		 1024 break;
1025
1026 case ARG_CRASH_SHELL:
b9e74c39
LP		 1027 if (!optarg)
1028 arg_crash_shell = true;
1029 else {
1030 r = parse_boolean(optarg);
1031 if (r < 0)
2b5107e1
ZJS		 1032 return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
1033 optarg);
b9e74c39
LP		 1034 arg_crash_shell = r;
1035 }
1036 break;
1037
1038 case ARG_CRASH_REBOOT:
1039 if (!optarg)
1040 arg_crash_reboot = true;
1041 else {
1042 r = parse_boolean(optarg);
1043 if (r < 0)
2b5107e1
ZJS		 1044 return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
1045 optarg);
b9e74c39 1046 arg_crash_reboot = r;
a5d87bf0 1047 }
9e58ff9c
LP		 1048 break;
1049
80876c20 1050 case ARG_CONFIRM_SPAWN:
7d5ceb64
FB		 1051 arg_confirm_spawn = mfree(arg_confirm_spawn);
1052
1053 r = parse_confirm_spawn(optarg, &arg_confirm_spawn);
1054 if (r < 0)
2b5107e1
ZJS		 1055 return log_error_errno(r, "Failed to parse confirm spawn option: \"%s\": %m",
1056 optarg);
80876c20
LP		 1057 break;
1058
2a12e32e
JK		 1059 case ARG_SERVICE_WATCHDOGS:
1060 r = parse_boolean(optarg);
1061 if (r < 0)
2b5107e1
ZJS		 1062 return log_error_errno(r, "Failed to parse service watchdogs boolean: \"%s\": %m",
1063 optarg);
2a12e32e
JK		 1064 arg_service_watchdogs = r;
1065 break;
1066
9e58ff9c 1067 case ARG_SHOW_STATUS:
d450b6f2
ZJS		 1068 if (optarg) {
1069 r = parse_show_status(optarg, &arg_show_status);
ac7ec288 1070 if (r < 0)
2b5107e1
ZJS		 1071 return log_error_errno(r, "Failed to parse show status boolean: \"%s\": %m",
1072 optarg);
d450b6f2
ZJS		 1073 } else
1074 arg_show_status = SHOW_STATUS_YES;
6e98720f 1075 break;
a5d87bf0 1076
a16e1123
LP		 1077 case ARG_DESERIALIZE: {
1078 int fd;
1079 FILE *f;
1080
01e10de3 1081 r = safe_atoi(optarg, &fd);
2b5107e1
ZJS		 1082 if (r < 0)
1083 log_error_errno(r, "Failed to parse deserialize option \"%s\": %m", optarg);
baaa35ad
ZJS		 1084 if (fd < 0)
1085 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1086 "Invalid deserialize fd: %d",
1087 fd);
a16e1123 1088
b9e74c39 1089 (void) fd_cloexec(fd, true);
01e10de3
LP		 1090
1091 f = fdopen(fd, "r");
4a62c710 1092 if (!f)
2b5107e1 1093 return log_error_errno(errno, "Failed to open serialization fd %d: %m", fd);
a16e1123 1094
74ca738f 1095 safe_fclose(arg_serialization);
d3b1c508 1096 arg_serialization = f;
a16e1123
LP		 1097
1098 break;
1099 }
1100
2660882b 1101 case ARG_SWITCHED_ROOT:
bf4df7c3 1102 arg_switched_root = true;
d03bc1b8
HH		 1103 break;
1104
ee48dbd5
NC		 1105 case ARG_MACHINE_ID:
1106 r = set_machine_id(optarg);
54500613 1107 if (r < 0)
2b5107e1 1108 return log_error_errno(r, "MachineID '%s' is not valid: %m", optarg);
ee48dbd5
NC		 1109 break;
1110
f170852a 1111 case 'h':
fa0f4d8a 1112 arg_action = ACTION_HELP;
f170852a
LP		 1113 break;
1114
1d2e23ab
LP		 1115 case 'D':
1116 log_set_max_level(LOG_DEBUG);
1117 break;
1118
099663ff
LP		 1119 case 'b':
1120 case 's':
1121 case 'z':
1122 /* Just to eat away the sysvinit kernel
1123 * cmdline args without getopt() error
1124 * messages that we'll parse in
1125 * parse_proc_cmdline_word() or ignore. */
f170852a 1126
099663ff 1127 case '?':
df0ff127 1128 if (getpid_cached() != 1)
099663ff 1129 return -EINVAL;
601185b4
ZJS		 1130 else
1131 return 0;
099663ff 1132
601185b4
ZJS		 1133 default:
1134 assert_not_reached("Unhandled option code.");
f170852a
LP		 1135 }
1136
df0ff127 1137 if (optind < argc && getpid_cached() != 1) {
d821e6d6
LP		 1138 /* Hmm, when we aren't run as init system
1139 * let's complain about excess arguments */
1140
baaa35ad
ZJS		 1141 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1142 "Excess arguments.");
d821e6d6
LP		 1143 }
1144
f170852a
LP		 1145 return 0;
1146}
1147
1148static int help(void) {
37ec0fdd
LP		 1149 _cleanup_free_ char *link = NULL;
1150 int r;
1151
1152 r = terminal_urlify_man("systemd", "1", &link);
1153 if (r < 0)
1154 return log_oom();
f170852a 1155
2e33c433 1156 printf("%s [OPTIONS...]\n\n"
af2d49f7 1157 "Starts up and maintains the system or user services.\n\n"
e537352b 1158 " -h --help Show this help\n"
cb4069d9 1159 " --version Show version\n"
e537352b 1160 " --test Determine startup sequence, dump it and exit\n"
b87c2aa6 1161 " --no-pager Do not pipe output into a pager\n"
80876c20 1162 " --dump-configuration-items Dump understood unit configuration items\n"
bbc1acab 1163 " --dump-bus-properties Dump exposed bus properties\n"
9e58ff9c 1164 " --unit=UNIT Set default unit\n"
edb9aaa8 1165 " --system Run a system instance, even if PID != 1\n"
af2d49f7 1166 " --user Run a user instance\n"
b9e74c39
LP		 1167 " --dump-core[=BOOL] Dump core on crash\n"
1168 " --crash-vt=NR Change to specified VT on crash\n"
1169 " --crash-reboot[=BOOL] Reboot on crash\n"
1170 " --crash-shell[=BOOL] Run shell on crash\n"
1171 " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n"
1172 " --show-status[=BOOL] Show status updates on the console during bootup\n"
c1dc6153 1173 " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
9e58ff9c 1174 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
b9e74c39
LP		 1175 " --log-color[=BOOL] Highlight important log messages\n"
1176 " --log-location[=BOOL] Include code location in log messages\n"
0a494f1f 1177 " --default-standard-output= Set default standard output for services\n"
37ec0fdd
LP		 1178 " --default-standard-error= Set default standard error output for services\n"
1179 "\nSee the %s for details.\n"
1180 , program_invocation_short_name
1181 , link
1182 );
f170852a
LP		 1183
1184 return 0;
1185}
1186
2cc856ac
LP		 1187static int prepare_reexecute(
1188 Manager *m,
1189 FILE **ret_f,
1190 FDSet **ret_fds,
1191 bool switching_root) {
1192
48b90859
LP		 1193 _cleanup_fdset_free_ FDSet *fds = NULL;
1194 _cleanup_fclose_ FILE *f = NULL;
a16e1123
LP		 1195 int r;
1196
1197 assert(m);
2cc856ac
LP		 1198 assert(ret_f);
1199 assert(ret_fds);
a16e1123 1200
6b78f9b4 1201 r = manager_open_serialization(m, &f);
48b90859
LP		 1202 if (r < 0)
1203 return log_error_errno(r, "Failed to create serialization file: %m");
a16e1123 1204
71445ae7 1205 /* Make sure nothing is really destructed when we shut down */
313cefa1 1206 m->n_reloading++;
718db961 1207 bus_manager_send_reloading(m, true);
71445ae7 1208
6b78f9b4 1209 fds = fdset_new();
48b90859
LP		 1210 if (!fds)
1211 return log_oom();
a16e1123 1212
b3680f49 1213 r = manager_serialize(m, f, fds, switching_root);
48b90859 1214 if (r < 0)
d68c645b 1215 return r;
a16e1123 1216
48b90859
LP		 1217 if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1218 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
a16e1123 1219
6b78f9b4 1220 r = fd_cloexec(fileno(f), false);
48b90859
LP		 1221 if (r < 0)
1222 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
a16e1123 1223
6b78f9b4 1224 r = fdset_cloexec(fds, false);
48b90859
LP		 1225 if (r < 0)
1226 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
a16e1123 1227
2cc856ac
LP		 1228 *ret_f = TAKE_PTR(f);
1229 *ret_fds = TAKE_PTR(fds);
a16e1123 1230
48b90859 1231 return 0;
a16e1123
LP		 1232}
1233
a8b627aa
LP		 1234static void bump_file_max_and_nr_open(void) {
1235
1236 /* Let's bump fs.file-max and fs.nr_open to their respective maximums. On current kernels large numbers of file
1237 * descriptors are no longer a performance problem and their memory is properly tracked by memcg, thus counting
1238 * them and limiting them in another two layers of limits is unnecessary and just complicates things. This
1239 * function hence turns off 2 of the 4 levels of limits on file descriptors, and makes RLIMIT_NOLIMIT (soft +
1240 * hard) the only ones that really matter. */
1241
1242#if BUMP_PROC_SYS_FS_FILE_MAX || BUMP_PROC_SYS_FS_NR_OPEN
1243 _cleanup_free_ char *t = NULL;
1244 int r;
1245#endif
1246
1247#if BUMP_PROC_SYS_FS_FILE_MAX
6e2f7894
LP		 1248 /* The maximum the kernel allows for this since 5.2 is LONG_MAX, use that. (Previously thing where
1249 * different but the operation would fail silently.) */
1250 if (asprintf(&t, "%li\n", LONG_MAX) < 0) {
a8b627aa
LP		 1251 log_oom();
1252 return;
1253 }
1254
1255 r = sysctl_write("fs/file-max", t);
1256 if (r < 0)
1257 log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.file-max, ignoring: %m");
1258#endif
1259
1260#if BUMP_PROC_SYS_FS_FILE_MAX && BUMP_PROC_SYS_FS_NR_OPEN
1261 t = mfree(t);
1262#endif
1263
1264#if BUMP_PROC_SYS_FS_NR_OPEN
1265 int v = INT_MAX;
1266
1267 /* Arg! The kernel enforces maximum and minimum values on the fs.nr_open, but we don't really know what they
1268 * are. The expression by which the maximum is determined is dependent on the architecture, and is something we
1269 * don't really want to copy to userspace, as it is dependent on implementation details of the kernel. Since
1270 * the kernel doesn't expose the maximum value to us, we can only try and hope. Hence, let's start with
1271 * INT_MAX, and then keep halving the value until we find one that works. Ugly? Yes, absolutely, but kernel
1272 * APIs are kernel APIs, so what do can we do... 🤯 */
1273
1274 for (;;) {
1275 int k;
1276
1277 v &= ~(__SIZEOF_POINTER__ - 1); /* Round down to next multiple of the pointer size */
1278 if (v < 1024) {
1279 log_warning("Can't bump fs.nr_open, value too small.");
1280 break;
1281 }
1282
1283 k = read_nr_open();
1284 if (k < 0) {
1285 log_error_errno(k, "Failed to read fs.nr_open: %m");
1286 break;
1287 }
1288 if (k >= v) { /* Already larger */
1289 log_debug("Skipping bump, value is already larger.");
1290 break;
1291 }
1292
1293 if (asprintf(&t, "%i\n", v) < 0) {
1294 log_oom();
1295 return;
1296 }
1297
1298 r = sysctl_write("fs/nr_open", t);
1299 t = mfree(t);
1300 if (r == -EINVAL) {
1301 log_debug("Couldn't write fs.nr_open as %i, halving it.", v);
1302 v /= 2;
1303 continue;
1304 }
1305 if (r < 0) {
1306 log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.nr_open, ignoring: %m");
1307 break;
1308 }
1309
1310 log_debug("Successfully bumped fs.nr_open to %i", v);
1311 break;
1312 }
1313#endif
1314}
1315
4096d6f5 1316static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
cda7faa9 1317 struct rlimit new_rlimit;
9264cc39 1318 int r, nr;
4096d6f5
LP		 1319
1320 assert(saved_rlimit);
1321
52d62075 1322 /* Save the original RLIMIT_NOFILE so that we can reset it later when transitioning from the initrd to the main
4096d6f5 1323 * systemd or suchlike. */
4a62c710 1324 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
3ce40911 1325 return log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m");
4096d6f5 1326
52d62075
LP		 1327 /* Get the underlying absolute limit the kernel enforces */
1328 nr = read_nr_open();
1329
1330 /* Make sure forked processes get limits based on the original kernel setting */
4096d6f5
LP		 1331 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1332 struct rlimit *rl;
1333
1334 rl = newdup(struct rlimit, saved_rlimit, 1);
1335 if (!rl)
1336 return log_oom();
1337
52d62075
LP		 1338 /* Bump the hard limit for system services to a substantially higher value. The default hard limit
1339 * current kernels set is pretty low (4K), mostly for historical reasons. According to kernel
1340 * developers, the fd handling in recent kernels has been optimized substantially enough, so that we
1341 * can bump the limit now, without paying too high a price in memory or performance. Note however that
1342 * we only bump the hard limit, not the soft limit. That's because select() works the way it works, and
1343 * chokes on fds >= 1024. If we'd bump the soft limit globally, it might accidentally happen to
1344 * unexpecting programs that they get fds higher than what they can process using select(). By only
1345 * bumping the hard limit but leaving the low limit as it is we avoid this pitfall: programs that are
1346 * written by folks aware of the select() problem in mind (and thus use poll()/epoll instead of
1347 * select(), the way everybody should) can explicitly opt into high fds by bumping their soft limit
1348 * beyond 1024, to the hard limit we pass. */
1349 if (arg_system)
1350 rl->rlim_max = MIN((rlim_t) nr, MAX(rl->rlim_max, (rlim_t) HIGH_RLIMIT_NOFILE));
1351
99a2fd3b
LP		 1352 /* If for some reason we were invoked with a soft limit above 1024 (which should never
1353 * happen!, but who knows what we get passed in from pam_limit when invoked as --user
1354 * instance), then lower what we pass on to not confuse our children */
1355 rl->rlim_cur = MIN(rl->rlim_cur, (rlim_t) FD_SETSIZE);
1356
4096d6f5
LP		 1357 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1358 }
1359
cda7faa9
LP		 1360 /* Calculate the new limits to use for us. Never lower from what we inherited. */
1361 new_rlimit = (struct rlimit) {
1362 .rlim_cur = MAX((rlim_t) nr, saved_rlimit->rlim_cur),
1363 .rlim_max = MAX((rlim_t) nr, saved_rlimit->rlim_max),
1364 };
1365
1366 /* Shortcut if nothing changes. */
1367 if (saved_rlimit->rlim_max >= new_rlimit.rlim_max &&
1368 saved_rlimit->rlim_cur >= new_rlimit.rlim_cur) {
1369 log_debug("RLIMIT_NOFILE is already as high or higher than we need it, not bumping.");
1370 return 0;
1371 }
1372
52d62075
LP		 1373 /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for
1374 * both hard and soft. */
cda7faa9 1375 r = setrlimit_closest(RLIMIT_NOFILE, &new_rlimit);
23bbb0de 1376 if (r < 0)
3ce40911 1377 return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m");
4096d6f5
LP		 1378
1379 return 0;
1380}
1381
fb3ae275 1382static int bump_rlimit_memlock(struct rlimit *saved_rlimit) {
cda7faa9 1383 struct rlimit new_rlimit;
fb3ae275
LP		 1384 int r;
1385
1386 assert(saved_rlimit);
fb3ae275 1387
a17c1712
LP		 1388 /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK which should
1389 * normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's bump
1390 * the value high enough for our user. */
fb3ae275
LP		 1391
1392 if (getrlimit(RLIMIT_MEMLOCK, saved_rlimit) < 0)
1393 return log_warning_errno(errno, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1394
c0d76959
LP		 1395 /* Pass the original value down to invoked processes */
1396 if (!arg_default_rlimit[RLIMIT_MEMLOCK]) {
1397 struct rlimit *rl;
1398
1399 rl = newdup(struct rlimit, saved_rlimit, 1);
1400 if (!rl)
1401 return log_oom();
1402
1403 arg_default_rlimit[RLIMIT_MEMLOCK] = rl;
1404 }
1405
cda7faa9
LP		 1406 /* Using MAX() on resource limits only is safe if RLIM_INFINITY is > 0. POSIX declares that rlim_t
1407 * must be unsigned, hence this is a given, but let's make this clear here. */
1408 assert_cc(RLIM_INFINITY > 0);
1409
1410 new_rlimit = (struct rlimit) {
1411 .rlim_cur = MAX(HIGH_RLIMIT_MEMLOCK, saved_rlimit->rlim_cur),
1412 .rlim_max = MAX(HIGH_RLIMIT_MEMLOCK, saved_rlimit->rlim_max),
1413 };
1414
1415 if (saved_rlimit->rlim_max >= new_rlimit.rlim_cur &&
1416 saved_rlimit->rlim_cur >= new_rlimit.rlim_max) {
1417 log_debug("RLIMIT_MEMLOCK is already as high or higher than we need it, not bumping.");
1418 return 0;
1419 }
1420
1421 r = setrlimit_closest(RLIMIT_MEMLOCK, &new_rlimit);
fb3ae275
LP		 1422 if (r < 0)
1423 return log_warning_errno(r, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1424
1425 return 0;
1426}
1427
80758717 1428static void test_usr(void) {
80758717 1429
796ac4c1 1430 /* Check that /usr is either on the same file system as / or mounted already. */
80758717 1431
871c44a7
LP		 1432 if (dir_is_empty("/usr") <= 0)
1433 return;
1434
8b173b5e 1435 log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
871c44a7
LP		 1436 "Some things will probably break (sometimes even silently) in mysterious ways. "
1437 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1438}
1439
d3b1c508 1440static int enforce_syscall_archs(Set *archs) {
349cc4a5 1441#if HAVE_SECCOMP
d3b1c508
LP		 1442 int r;
1443
83f12b27
FS		 1444 if (!is_seccomp_available())
1445 return 0;
1446
469830d1 1447 r = seccomp_restrict_archs(arg_syscall_archs);
d3b1c508 1448 if (r < 0)
469830d1 1449 return log_error_errno(r, "Failed to enforce system call architecture restrication: %m");
d3b1c508 1450#endif
469830d1 1451 return 0;
d3b1c508
LP		 1452}
1453
b6e2f329
LP		 1454static int status_welcome(void) {
1455 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1456 int r;
1457
7a293242 1458 if (IN_SET(arg_show_status, SHOW_STATUS_NO, SHOW_STATUS_AUTO))
fd8c85c6
LP		 1459 return 0;
1460
d58ad743
LP		 1461 r = parse_os_release(NULL,
1462 "PRETTY_NAME", &pretty_name,
1463 "ANSI_COLOR", &ansi_color,
1464 NULL);
1465 if (r < 0)
1466 log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r,
1467 "Failed to read os-release file, ignoring: %m");
b6e2f329 1468
dc9b5816 1469 if (log_get_show_color())
a885727a 1470 return status_printf(NULL, 0,
dc9b5816
ZJS		 1471 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1472 isempty(ansi_color) ? "1" : ansi_color,
1473 isempty(pretty_name) ? "Linux" : pretty_name);
1474 else
a885727a 1475 return status_printf(NULL, 0,
dc9b5816
ZJS		 1476 "\nWelcome to %s!\n",
1477 isempty(pretty_name) ? "Linux" : pretty_name);
b6e2f329
LP		 1478}
1479
fdd25311
LP		 1480static int write_container_id(void) {
1481 const char *c;
19854865 1482 int r;
fdd25311
LP		 1483
1484 c = getenv("container");
1485 if (isempty(c))
1486 return 0;
1487
8612da97
LP		 1488 RUN_WITH_UMASK(0022)
1489 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
19854865 1490 if (r < 0)
f1f849b0 1491 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
19854865
LP		 1492
1493 return 1;
1494}
1495
1496static int bump_unix_max_dgram_qlen(void) {
1497 _cleanup_free_ char *qlen = NULL;
1498 unsigned long v;
1499 int r;
1500
3130fca5
LP		 1501 /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1502 * really really early during boot, so that it is actually applied to all our sockets, including the
1503 * $NOTIFY_SOCKET one. */
19854865
LP		 1504
1505 r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1506 if (r < 0)
875622c3 1507 return log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
19854865
LP		 1508
1509 r = safe_atolu(qlen, &v);
1510 if (r < 0)
3130fca5 1511 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen);
19854865
LP		 1512
1513 if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1514 return 0;
1515
57512c89 1516 r = write_string_filef("/proc/sys/net/unix/max_dgram_qlen", WRITE_STRING_FILE_DISABLE_BUFFER, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN);
19854865
LP		 1517 if (r < 0)
1518 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1519 "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1520
1521 return 1;
fdd25311
LP		 1522}
1523
32391275
FB		 1524static int fixup_environment(void) {
1525 _cleanup_free_ char *term = NULL;
4dc63c4b 1526 const char *t;
32391275
FB		 1527 int r;
1528
43db615b
LP		 1529 /* Only fix up the environment when we are started as PID 1 */
1530 if (getpid_cached() != 1)
1531 return 0;
1532
1533 /* We expect the environment to be set correctly if run inside a container. */
84af7821
LP		 1534 if (detect_container() > 0)
1535 return 0;
1536
43db615b
LP		 1537 /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1538 * device used by the console. We try to make a better guess here since some consoles might not have support
1539 * for color mode for example.
32391275 1540 *
43db615b 1541 * However if TERM was configured through the kernel command line then leave it alone. */
1d84ad94 1542 r = proc_cmdline_get_key("TERM", 0, &term);
32391275
FB		 1543 if (r < 0)
1544 return r;
32391275 1545
4dc63c4b
LP		 1546 t = term ?: default_term_for_tty("/dev/console");
1547
1548 if (setenv("TERM", t, 1) < 0)
32391275
FB		 1549 return -errno;
1550
9d48671c
ZJS		 1551 /* The kernels sets HOME=/ for init. Let's undo this. */
1552 if (path_equal_ptr(getenv("HOME"), "/") &&
1553 unsetenv("HOME") < 0)
1554 log_warning_errno(errno, "Failed to unset $HOME: %m");
1555
32391275
FB		 1556 return 0;
1557}
1558
6808a0bc
LP		 1559static void redirect_telinit(int argc, char *argv[]) {
1560
1561 /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1562
1563#if HAVE_SYSV_COMPAT
1564 if (getpid_cached() == 1)
1565 return;
1566
1567 if (!strstr(program_invocation_short_name, "init"))
1568 return;
1569
1570 execv(SYSTEMCTL_BINARY_PATH, argv);
1571 log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
a45d7127 1572 exit(EXIT_FAILURE);
6808a0bc
LP		 1573#endif
1574}
1575
4a36297c
LP		 1576static int become_shutdown(
1577 const char *shutdown_verb,
7eb35049 1578 int retval) {
4a36297c
LP		 1579
1580 char log_level[DECIMAL_STR_MAX(int) + 1],
e73c54b8
JK		 1581 exit_code[DECIMAL_STR_MAX(uint8_t) + 1],
1582 timeout[DECIMAL_STR_MAX(usec_t) + 1];
4a36297c 1583
e73c54b8 1584 const char* command_line[13] = {
4a36297c
LP		 1585 SYSTEMD_SHUTDOWN_BINARY_PATH,
1586 shutdown_verb,
e73c54b8 1587 "--timeout", timeout,
4a36297c
LP		 1588 "--log-level", log_level,
1589 "--log-target",
1590 };
1591
1592 _cleanup_strv_free_ char **env_block = NULL;
e73c54b8 1593 size_t pos = 7;
4a36297c
LP		 1594 int r;
1595
7eb35049 1596 assert(shutdown_verb);
234519ae 1597 assert(!command_line[pos]);
4a36297c
LP		 1598 env_block = strv_copy(environ);
1599
1600 xsprintf(log_level, "%d", log_get_max_level());
e73c54b8 1601 xsprintf(timeout, "%" PRI_USEC "us", arg_default_timeout_stop_usec);
4a36297c
LP		 1602
1603 switch (log_get_target()) {
1604
1605 case LOG_TARGET_KMSG:
1606 case LOG_TARGET_JOURNAL_OR_KMSG:
1607 case LOG_TARGET_SYSLOG_OR_KMSG:
1608 command_line[pos++] = "kmsg";
1609 break;
1610
1611 case LOG_TARGET_NULL:
1612 command_line[pos++] = "null";
1613 break;
1614
1615 case LOG_TARGET_CONSOLE:
1616 default:
1617 command_line[pos++] = "console";
1618 break;
1619 };
1620
1621 if (log_get_show_color())
1622 command_line[pos++] = "--log-color";
1623
1624 if (log_get_show_location())
1625 command_line[pos++] = "--log-location";
1626
1627 if (streq(shutdown_verb, "exit")) {
1628 command_line[pos++] = "--exit-code";
1629 command_line[pos++] = exit_code;
1630 xsprintf(exit_code, "%d", retval);
1631 }
1632
1633 assert(pos < ELEMENTSOF(command_line));
1634
7eb35049
LP		 1635 if (streq(shutdown_verb, "reboot") &&
1636 arg_shutdown_watchdog > 0 &&
1637 arg_shutdown_watchdog != USEC_INFINITY) {
1638
4a36297c
LP		 1639 char *e;
1640
1641 /* If we reboot let's set the shutdown
1642 * watchdog and tell the shutdown binary to
1643 * repeatedly ping it */
1644 r = watchdog_set_timeout(&arg_shutdown_watchdog);
1645 watchdog_close(r < 0);
1646
1647 /* Tell the binary how often to ping, ignore failure */
1648 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
8a2c1fbf
EJ		 1649 (void) strv_consume(&env_block, e);
1650
1651 if (arg_watchdog_device &&
1652 asprintf(&e, "WATCHDOG_DEVICE=%s", arg_watchdog_device) > 0)
1653 (void) strv_consume(&env_block, e);
4a36297c
LP		 1654 } else
1655 watchdog_close(true);
1656
1657 /* Avoid the creation of new processes forked by the
1658 * kernel; at this point, we will not listen to the
1659 * signals anyway */
1660 if (detect_container() <= 0)
1661 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
1662
1663 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
1664 return -errno;
1665}
1666
e839bafd
LP		 1667static void initialize_clock(void) {
1668 int r;
1669
1670 if (clock_is_localtime(NULL) > 0) {
1671 int min;
1672
1673 /*
1674 * The very first call of settimeofday() also does a time warp in the kernel.
1675 *
1676 * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1677 * of maintaining the RTC and do all adjustments. This matches the behavior of Windows, which leaves
1678 * the RTC alone if the registry tells that the RTC runs in UTC.
1679 */
1680 r = clock_set_timezone(&min);
1681 if (r < 0)
1682 log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1683 else
1684 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1685
1686 } else if (!in_initrd()) {
1687 /*
1688 * Do a dummy very first call to seal the kernel's time warp magic.
1689 *
1690 * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1691 * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1692 * until we reach the real system.
1693 *
1694 * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1695 * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1696 * be treated as UTC that way.
1697 */
1698 (void) clock_reset_timewarp();
1699 }
1700
1701 r = clock_apply_epoch();
1702 if (r < 0)
1703 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1704 else if (r > 0)
1705 log_info("System time before build time, advancing clock.");
1706}
1707
1e41242e 1708static void initialize_coredump(bool skip_setup) {
752bcb77 1709#if ENABLE_COREDUMP
1e41242e
LP		 1710 if (getpid_cached() != 1)
1711 return;
1712
1713 /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1714 * will process core dumps for system services by default. */
1715 if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0)
1716 log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m");
1717
c6885f5f
FB		 1718 /* But at the same time, turn off the core_pattern logic by default, so that no
1719 * coredumps are stored until the systemd-coredump tool is enabled via
1720 * sysctl. However it can be changed via the kernel command line later so core
1721 * dumps can still be generated during early startup and in initramfs. */
1e41242e 1722 if (!skip_setup)
e557b1a6 1723 disable_coredumps();
752bcb77 1724#endif
1e41242e
LP		 1725}
1726
c6885f5f
FB		 1727static void initialize_core_pattern(bool skip_setup) {
1728 int r;
1729
1730 if (skip_setup || !arg_early_core_pattern)
1731 return;
1732
1733 if (getpid_cached() != 1)
1734 return;
1735
57512c89 1736 r = write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern, WRITE_STRING_FILE_DISABLE_BUFFER);
c6885f5f
FB		 1737 if (r < 0)
1738 log_warning_errno(r, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", arg_early_core_pattern);
1739}
1740
61fbbac1
ZJS		 1741static void update_cpu_affinity(bool skip_setup) {
1742 _cleanup_free_ char *mask = NULL;
1743
1744 if (skip_setup || !arg_cpu_affinity.set)
1745 return;
1746
1747 assert(arg_cpu_affinity.allocated > 0);
1748
1749 mask = cpu_set_to_string(&arg_cpu_affinity);
1750 log_debug("Setting CPU affinity to %s.", strnull(mask));
1751
1752 if (sched_setaffinity(0, arg_cpu_affinity.allocated, arg_cpu_affinity.set) < 0)
1753 log_warning_errno(errno, "Failed to set CPU affinity: %m");
1754}
1755
3c7878f9
LP		 1756static void do_reexecute(
1757 int argc,
1758 char *argv[],
1759 const struct rlimit *saved_rlimit_nofile,
1760 const struct rlimit *saved_rlimit_memlock,
1761 FDSet *fds,
1762 const char *switch_root_dir,
1763 const char *switch_root_init,
1764 const char **ret_error_message) {
1765
1766 unsigned i, j, args_size;
1767 const char **args;
1768 int r;
1769
1770 assert(saved_rlimit_nofile);
1771 assert(saved_rlimit_memlock);
1772 assert(ret_error_message);
1773
1774 /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1775 * we do that */
1776 watchdog_close(true);
1777
ddfa8b0b
LP		 1778 /* Reset RLIMIT_NOFILE + RLIMIT_MEMLOCK back to the kernel defaults, so that the new systemd can pass
1779 * the kernel default to its child processes */
1780 if (saved_rlimit_nofile->rlim_cur != 0)
3c7878f9 1781 (void) setrlimit(RLIMIT_NOFILE, saved_rlimit_nofile);
ddfa8b0b 1782 if (saved_rlimit_memlock->rlim_cur != RLIM_INFINITY)
3c7878f9
LP		 1783 (void) setrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock);
1784
1785 if (switch_root_dir) {
1786 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1787 * SIGCHLD for them after deserializing. */
e73c54b8 1788 broadcast_signal(SIGTERM, false, true, arg_default_timeout_stop_usec);
3c7878f9
LP		 1789
1790 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1791 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1792 if (r < 0)
1793 log_error_errno(r, "Failed to switch root, trying to continue: %m");
1794 }
1795
1796 args_size = MAX(6, argc+1);
1797 args = newa(const char*, args_size);
1798
1799 if (!switch_root_init) {
1800 char sfd[DECIMAL_STR_MAX(int) + 1];
1801
1802 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1803 * the user didn't specify an explicit init to spawn. */
1804
1805 assert(arg_serialization);
1806 assert(fds);
1807
1808 xsprintf(sfd, "%i", fileno(arg_serialization));
1809
1810 i = 0;
1811 args[i++] = SYSTEMD_BINARY_PATH;
1812 if (switch_root_dir)
1813 args[i++] = "--switched-root";
1814 args[i++] = arg_system ? "--system" : "--user";
1815 args[i++] = "--deserialize";
1816 args[i++] = sfd;
1817 args[i++] = NULL;
1818
1819 assert(i <= args_size);
1820
1821 /*
1822 * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do this is on
1823 * its own on exec(), but it will do it on exit(). Hence, to ensure we get a summary here, fork() off
1824 * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1825 * before proceeding into the exec().
1826 */
1827 valgrind_summary_hack();
1828
1829 (void) execv(args[0], (char* const*) args);
1830 log_debug_errno(errno, "Failed to execute our own binary, trying fallback: %m");
1831 }
1832
1833 /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1834 * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1835 * doesn't matter.) */
1836
1837 arg_serialization = safe_fclose(arg_serialization);
1838 fds = fdset_free(fds);
1839
1840 /* Reopen the console */
1841 (void) make_console_stdio();
1842
1843 for (j = 1, i = 1; j < (unsigned) argc; j++)
1844 args[i++] = argv[j];
1845 args[i++] = NULL;
1846 assert(i <= args_size);
1847
5238e957 1848 /* Re-enable any blocked signals, especially important if we switch from initial ramdisk to init=... */
3c7878f9
LP		 1849 (void) reset_all_signal_handlers();
1850 (void) reset_signal_mask();
595225af 1851 (void) rlimit_nofile_safe();
3c7878f9
LP		 1852
1853 if (switch_root_init) {
1854 args[0] = switch_root_init;
1855 (void) execv(args[0], (char* const*) args);
1856 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
1857 }
1858
1859 args[0] = "/sbin/init";
1860 (void) execv(args[0], (char* const*) args);
1861 r = -errno;
1862
1863 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
1864 ANSI_HIGHLIGHT_RED " !! " ANSI_NORMAL,
1865 "Failed to execute /sbin/init");
1866
1867 if (r == -ENOENT) {
1868 log_warning("No /sbin/init, trying fallback");
1869
1870 args[0] = "/bin/sh";
1871 args[1] = NULL;
1872 (void) execv(args[0], (char* const*) args);
1873 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
1874 } else
1875 log_warning_errno(r, "Failed to execute /sbin/init, giving up: %m");
1876
1877 *ret_error_message = "Failed to execute fallback shell";
1878}
1879
7eb35049
LP		 1880static int invoke_main_loop(
1881 Manager *m,
1882 bool *ret_reexecute,
1883 int *ret_retval, /* Return parameters relevant for shutting down */
1884 const char **ret_shutdown_verb, /* … */
1885 FDSet **ret_fds, /* Return parameters for reexecuting */
1886 char **ret_switch_root_dir, /* … */
1887 char **ret_switch_root_init, /* … */
1888 const char **ret_error_message) {
1889
1890 int r;
1891
1892 assert(m);
1893 assert(ret_reexecute);
1894 assert(ret_retval);
1895 assert(ret_shutdown_verb);
1896 assert(ret_fds);
1897 assert(ret_switch_root_dir);
1898 assert(ret_switch_root_init);
1899 assert(ret_error_message);
1900
1901 for (;;) {
1902 r = manager_loop(m);
1903 if (r < 0) {
1904 *ret_error_message = "Failed to run main loop";
1905 return log_emergency_errno(r, "Failed to run main loop: %m");
1906 }
1907
3ca4d0b3 1908 switch ((ManagerObjective) r) {
7eb35049 1909
a6ecbf83 1910 case MANAGER_RELOAD: {
bda7d78b 1911 LogTarget saved_log_target;
a6ecbf83
FB		 1912 int saved_log_level;
1913
7eb35049
LP		 1914 log_info("Reloading.");
1915
3fe91079 1916 /* First, save any overridden log level/target, then parse the configuration file, which might
bda7d78b
FB		 1917 * change the log level to new settings. */
1918
a6ecbf83 1919 saved_log_level = m->log_level_overridden ? log_get_max_level() : -1;
bda7d78b 1920 saved_log_target = m->log_target_overridden ? log_get_target() : _LOG_TARGET_INVALID;
a6ecbf83 1921
470a5e6d 1922 (void) parse_configuration();
7eb35049
LP		 1923
1924 set_manager_defaults(m);
1925
61fbbac1
ZJS		 1926 update_cpu_affinity(false);
1927
a6ecbf83
FB		 1928 if (saved_log_level >= 0)
1929 manager_override_log_level(m, saved_log_level);
bda7d78b
FB		 1930 if (saved_log_target >= 0)
1931 manager_override_log_target(m, saved_log_target);
a6ecbf83 1932
7eb35049
LP		 1933 r = manager_reload(m);
1934 if (r < 0)
7a35fa24
LP		 1935 /* Reloading failed before the point of no return. Let's continue running as if nothing happened. */
1936 m->objective = MANAGER_OK;
7eb35049
LP		 1937
1938 break;
a6ecbf83 1939 }
7eb35049
LP		 1940
1941 case MANAGER_REEXECUTE:
1942
1943 r = prepare_reexecute(m, &arg_serialization, ret_fds, false);
1944 if (r < 0) {
1945 *ret_error_message = "Failed to prepare for reexecution";
1946 return r;
1947 }
1948
1949 log_notice("Reexecuting.");
1950
1951 *ret_reexecute = true;
1952 *ret_retval = EXIT_SUCCESS;
1953 *ret_shutdown_verb = NULL;
1954 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1955
1956 return 0;
1957
1958 case MANAGER_SWITCH_ROOT:
1959 if (!m->switch_root_init) {
1960 r = prepare_reexecute(m, &arg_serialization, ret_fds, true);
1961 if (r < 0) {
1962 *ret_error_message = "Failed to prepare for reexecution";
1963 return r;
1964 }
1965 } else
1966 *ret_fds = NULL;
1967
1968 log_notice("Switching root.");
1969
1970 *ret_reexecute = true;
1971 *ret_retval = EXIT_SUCCESS;
1972 *ret_shutdown_verb = NULL;
1973
1974 /* Steal the switch root parameters */
49052946
YW		 1975 *ret_switch_root_dir = TAKE_PTR(m->switch_root);
1976 *ret_switch_root_init = TAKE_PTR(m->switch_root_init);
7eb35049
LP		 1977
1978 return 0;
1979
1980 case MANAGER_EXIT:
1981
1982 if (MANAGER_IS_USER(m)) {
1983 log_debug("Exit.");
1984
1985 *ret_reexecute = false;
1986 *ret_retval = m->return_value;
1987 *ret_shutdown_verb = NULL;
1988 *ret_fds = NULL;
1989 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1990
1991 return 0;
1992 }
1993
1994 _fallthrough_;
1995 case MANAGER_REBOOT:
1996 case MANAGER_POWEROFF:
1997 case MANAGER_HALT:
1998 case MANAGER_KEXEC: {
af41e508
LP		 1999 static const char * const table[_MANAGER_OBJECTIVE_MAX] = {
2000 [MANAGER_EXIT] = "exit",
2001 [MANAGER_REBOOT] = "reboot",
7eb35049 2002 [MANAGER_POWEROFF] = "poweroff",
af41e508
LP		 2003 [MANAGER_HALT] = "halt",
2004 [MANAGER_KEXEC] = "kexec",
7eb35049
LP		 2005 };
2006
2007 log_notice("Shutting down.");
2008
2009 *ret_reexecute = false;
2010 *ret_retval = m->return_value;
af41e508 2011 assert_se(*ret_shutdown_verb = table[m->objective]);
7eb35049
LP		 2012 *ret_fds = NULL;
2013 *ret_switch_root_dir = *ret_switch_root_init = NULL;
2014
2015 return 0;
2016 }
2017
2018 default:
af41e508 2019 assert_not_reached("Unknown or unexpected manager objective.");
7eb35049
LP		 2020 }
2021 }
2022}
2023
31aef7ff
LP		 2024static void log_execution_mode(bool *ret_first_boot) {
2025 assert(ret_first_boot);
2026
2027 if (arg_system) {
2028 int v;
2029
681bd2c5 2030 log_info("systemd " GIT_VERSION " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
31aef7ff
LP		 2031 arg_action == ACTION_TEST ? "test " : "" );
2032
2033 v = detect_virtualization();
2034 if (v > 0)
2035 log_info("Detected virtualization %s.", virtualization_to_string(v));
2036
2037 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
2038
2039 if (in_initrd()) {
2040 *ret_first_boot = false;
2041 log_info("Running in initial RAM disk.");
2042 } else {
2043 /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
2044 * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
2045 * doesn't it's unpopulated. This allows container managers and installers to provision a
2046 * couple of files already. If the container manager wants to provision the machine ID itself
2047 * it should pass $container_uuid to PID 1. */
2048
2049 *ret_first_boot = access("/etc/machine-id", F_OK) < 0;
2050 if (*ret_first_boot)
2051 log_info("Running with unpopulated /etc.");
2052 }
2053 } else {
b9e90f3a
LP		 2054 if (DEBUG_LOGGING) {
2055 _cleanup_free_ char *t;
31aef7ff 2056
b9e90f3a 2057 t = uid_to_name(getuid());
681bd2c5 2058 log_debug("systemd " GIT_VERSION " running in %suser mode for user " UID_FMT "/%s. (" SYSTEMD_FEATURES ")",
b9e90f3a
LP		 2059 arg_action == ACTION_TEST ? " test" : "", getuid(), strna(t));
2060 }
31aef7ff
LP		 2061
2062 *ret_first_boot = false;
2063 }
2064}
2065
5afbaa36
LP		 2066static int initialize_runtime(
2067 bool skip_setup,
2068 struct rlimit *saved_rlimit_nofile,
2069 struct rlimit *saved_rlimit_memlock,
2070 const char **ret_error_message) {
2071
2072 int r;
2073
2074 assert(ret_error_message);
2075
2076 /* Sets up various runtime parameters. Many of these initializations are conditionalized:
2077 *
2078 * - Some only apply to --system instances
2079 * - Some only apply to --user instances
2080 * - Some only apply when we first start up, but not when we reexecute
2081 */
2082
2d776038
LP		 2083 if (arg_action != ACTION_RUN)
2084 return 0;
2085
61fbbac1
ZJS		 2086 update_cpu_affinity(skip_setup);
2087
3c3c6cb9 2088 if (arg_system) {
5238e957 2089 /* Make sure we leave a core dump without panicking the kernel. */
3c3c6cb9 2090 install_crash_handler();
5afbaa36 2091
3c3c6cb9 2092 if (!skip_setup) {
143fadf3 2093 r = mount_cgroup_controllers();
3c3c6cb9
LP		 2094 if (r < 0) {
2095 *ret_error_message = "Failed to mount cgroup hierarchies";
2096 return r;
2097 }
2098
2099 status_welcome();
2100 hostname_setup();
2101 machine_id_setup(NULL, arg_machine_id, NULL);
2102 loopback_setup();
2103 bump_unix_max_dgram_qlen();
a8b627aa 2104 bump_file_max_and_nr_open();
3c3c6cb9
LP		 2105 test_usr();
2106 write_container_id();
2107 }
8a2c1fbf 2108
3c3c6cb9
LP		 2109 if (arg_watchdog_device) {
2110 r = watchdog_set_device(arg_watchdog_device);
2111 if (r < 0)
2112 log_warning_errno(r, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device);
2113 }
2114
2115 if (arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY)
2116 watchdog_set_timeout(&arg_runtime_watchdog);
2117 }
5afbaa36
LP		 2118
2119 if (arg_timer_slack_nsec != NSEC_INFINITY)
2120 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
3a671cd1 2121 log_warning_errno(errno, "Failed to adjust timer slack, ignoring: %m");
5afbaa36
LP		 2122
2123 if (arg_system && !cap_test_all(arg_capability_bounding_set)) {
2124 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
2125 if (r < 0) {
2126 *ret_error_message = "Failed to drop capability bounding set of usermode helpers";
2127 return log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
2128 }
2129
2130 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
2131 if (r < 0) {
2132 *ret_error_message = "Failed to drop capability bounding set";
2133 return log_emergency_errno(r, "Failed to drop capability bounding set: %m");
2134 }
2135 }
2136
39362f6f
JB		 2137 if (arg_system && arg_no_new_privs) {
2138 if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) {
2139 *ret_error_message = "Failed to disable new privileges";
2140 return log_emergency_errno(errno, "Failed to disable new privileges: %m");
2141 }
2142 }
2143
5afbaa36
LP		 2144 if (arg_syscall_archs) {
2145 r = enforce_syscall_archs(arg_syscall_archs);
2146 if (r < 0) {
2147 *ret_error_message = "Failed to set syscall architectures";
2148 return r;
2149 }
2150 }
2151
2152 if (!arg_system)
2153 /* Become reaper of our children */
2154 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
2155 log_warning_errno(errno, "Failed to make us a subreaper: %m");
2156
a17c1712
LP		 2157 /* Bump up RLIMIT_NOFILE for systemd itself */
2158 (void) bump_rlimit_nofile(saved_rlimit_nofile);
2159 (void) bump_rlimit_memlock(saved_rlimit_memlock);
5afbaa36
LP		 2160
2161 return 0;
2162}
2163
6acca5fc
LP		 2164static int do_queue_default_job(
2165 Manager *m,
2166 const char **ret_error_message) {
2167
2168 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
2169 Job *default_unit_job;
2170 Unit *target = NULL;
2171 int r;
2172
2173 log_debug("Activating default unit: %s", arg_default_unit);
2174
4109ede7
ZJS		 2175 r = manager_load_startable_unit_or_warn(m, arg_default_unit, NULL, &target);
2176 if (r < 0) {
2177 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET);
6acca5fc 2178
4109ede7 2179 r = manager_load_startable_unit_or_warn(m, SPECIAL_RESCUE_TARGET, NULL, &target);
6acca5fc 2180 if (r < 0) {
4109ede7
ZJS		 2181 *ret_error_message = r == -ERFKILL ? "Rescue target masked"
2182 : "Failed to load rescue target";
2183 return r;
6acca5fc
LP		 2184 }
2185 }
2186
2187 assert(target->load_state == UNIT_LOADED);
2188
50cbaba4 2189 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, NULL, &error, &default_unit_job);
6acca5fc
LP		 2190 if (r == -EPERM) {
2191 log_debug_errno(r, "Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
2192
2193 sd_bus_error_free(&error);
2194
50cbaba4 2195 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, NULL, &error, &default_unit_job);
6acca5fc
LP		 2196 if (r < 0) {
2197 *ret_error_message = "Failed to start default target";
2198 return log_emergency_errno(r, "Failed to start default target: %s", bus_error_message(&error, r));
2199 }
2200
2201 } else if (r < 0) {
2202 *ret_error_message = "Failed to isolate default target";
2203 return log_emergency_errno(r, "Failed to isolate default target: %s", bus_error_message(&error, r));
2204 }
2205
2206 m->default_unit_job_id = default_unit_job->id;
2207
2208 return 0;
2209}
2210
fb39af4c
ZJS		 2211static void reset_arguments(void) {
2212 /* Frees/resets arg_* variables, with a few exceptions commented below. */
970777b5
LP		 2213
2214 arg_default_unit = mfree(arg_default_unit);
fb39af4c
ZJS		 2215
2216 /* arg_system — ignore */
2217
2218 arg_dump_core = true;
2219 arg_crash_chvt = -1;
2220 arg_crash_shell = false;
2221 arg_crash_reboot = false;
970777b5 2222 arg_confirm_spawn = mfree(arg_confirm_spawn);
fb39af4c
ZJS		 2223 arg_show_status = _SHOW_STATUS_INVALID;
2224 arg_switched_root = false;
2225 arg_pager_flags = 0;
2226 arg_service_watchdogs = true;
2227 arg_default_std_output = EXEC_OUTPUT_JOURNAL;
2228 arg_default_std_error = EXEC_OUTPUT_INHERIT;
2229 arg_default_restart_usec = DEFAULT_RESTART_USEC;
2230 arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
2231 arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
2232 arg_default_timeout_abort_usec = DEFAULT_TIMEOUT_USEC;
2233 arg_default_timeout_abort_set = false;
2234 arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
2235 arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
2236 arg_runtime_watchdog = 0;
2237 arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
2238 arg_early_core_pattern = NULL;
2239 arg_watchdog_device = NULL;
2240
970777b5 2241 arg_default_environment = strv_free(arg_default_environment);
fb39af4c
ZJS		 2242 rlimit_free_all(arg_default_rlimit);
2243
2244 arg_capability_bounding_set = CAP_ALL;
2245 arg_no_new_privs = false;
2246 arg_timer_slack_nsec = NSEC_INFINITY;
2247 arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
2248
970777b5 2249 arg_syscall_archs = set_free(arg_syscall_archs);
61fbbac1 2250
fb39af4c
ZJS		 2251 /* arg_serialization — ignore */
2252
2253 arg_default_cpu_accounting = -1;
2254 arg_default_io_accounting = false;
2255 arg_default_ip_accounting = false;
2256 arg_default_blockio_accounting = false;
2257 arg_default_memory_accounting = MEMORY_ACCOUNTING_DEFAULT;
2258 arg_default_tasks_accounting = true;
2259 arg_default_tasks_max = UINT64_MAX;
2260 arg_machine_id = (sd_id128_t) {};
2261 arg_cad_burst_action = EMERGENCY_ACTION_REBOOT_FORCE;
2262 arg_default_oom_policy = OOM_STOP;
2263
61fbbac1 2264 cpu_set_reset(&arg_cpu_affinity);
970777b5
LP		 2265}
2266
470a5e6d 2267static int parse_configuration(void) {
97d1fb94
LP		 2268 int r;
2269
97d1fb94
LP		 2270 arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U);
2271
fb39af4c
ZJS		 2272 /* Assign configuration defaults */
2273 reset_arguments();
2274
97d1fb94 2275 r = parse_config_file();
470a5e6d
ZJS		 2276 if (r < 0)
2277 log_warning_errno(r, "Failed to parse config file, ignoring: %m");
97d1fb94
LP		 2278
2279 if (arg_system) {
2280 r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
2281 if (r < 0)
2282 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
2283 }
2284
2285 /* Note that this also parses bits from the kernel command line, including "debug". */
2286 log_parse_environment();
2287
470a5e6d
ZJS		 2288 return 0;
2289}
2290
2291static int load_configuration(int argc, char **argv, const char **ret_error_message) {
2292 int r;
2293
2294 assert(ret_error_message);
2295
2296 (void) parse_configuration();
2297
97d1fb94
LP		 2298 r = parse_argv(argc, argv);
2299 if (r < 0) {
2300 *ret_error_message = "Failed to parse commandline arguments";
2301 return r;
2302 }
2303
2304 /* Initialize default unit */
2305 if (!arg_default_unit) {
2306 arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET);
2307 if (!arg_default_unit) {
2308 *ret_error_message = "Failed to set default unit";
2309 return log_oom();
2310 }
2311 }
2312
db33214b 2313 /* Initialize the show status setting if it hasn't been set explicitly yet */
7a293242 2314 if (arg_show_status == _SHOW_STATUS_INVALID)
db33214b
LP		 2315 arg_show_status = SHOW_STATUS_YES;
2316
97d1fb94
LP		 2317 return 0;
2318}
2319
b0d7c989
LP		 2320static int safety_checks(void) {
2321
febf46a4 2322 if (getpid_cached() == 1 &&
baaa35ad
ZJS		 2323 arg_action != ACTION_RUN)
2324 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2325 "Unsupported execution mode while PID 1.");
febf46a4
LP		 2326
2327 if (getpid_cached() == 1 &&
baaa35ad
ZJS		 2328 !arg_system)
2329 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2330 "Can't run --user mode as PID 1.");
febf46a4
LP		 2331
2332 if (arg_action == ACTION_RUN &&
2333 arg_system &&
baaa35ad
ZJS		 2334 getpid_cached() != 1)
2335 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2336 "Can't run system mode unless PID 1.");
febf46a4 2337
b0d7c989 2338 if (arg_action == ACTION_TEST &&
baaa35ad
ZJS		 2339 geteuid() == 0)
2340 return log_error_errno(SYNTHETIC_ERRNO(EPERM),
2341 "Don't run test mode as root.");
b0d7c989
LP		 2342
2343 if (!arg_system &&
2344 arg_action == ACTION_RUN &&
baaa35ad
ZJS		 2345 sd_booted() <= 0)
2346 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
2347 "Trying to run as user instance, but the system has not been booted with systemd.");
b0d7c989
LP		 2348
2349 if (!arg_system &&
2350 arg_action == ACTION_RUN &&
baaa35ad
ZJS		 2351 !getenv("XDG_RUNTIME_DIR"))
2352 return log_error_errno(SYNTHETIC_ERRNO(EUNATCH),
2353 "Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
b0d7c989
LP		 2354
2355 if (arg_system &&
2356 arg_action == ACTION_RUN &&
baaa35ad
ZJS		 2357 running_in_chroot() > 0)
2358 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
2359 "Cannot be run in a chroot() environment.");
b0d7c989
LP		 2360
2361 return 0;
2362}
2363
74da609f
LP		 2364static int initialize_security(
2365 bool *loaded_policy,
2366 dual_timestamp *security_start_timestamp,
2367 dual_timestamp *security_finish_timestamp,
2368 const char **ret_error_message) {
2369
2370 int r;
2371
2372 assert(loaded_policy);
2373 assert(security_start_timestamp);
2374 assert(security_finish_timestamp);
2375 assert(ret_error_message);
2376
2377 dual_timestamp_get(security_start_timestamp);
2378
97149f40 2379 r = mac_selinux_setup(loaded_policy);
74da609f
LP		 2380 if (r < 0) {
2381 *ret_error_message = "Failed to load SELinux policy";
2382 return r;
2383 }
2384
2385 r = mac_smack_setup(loaded_policy);
2386 if (r < 0) {
2387 *ret_error_message = "Failed to load SMACK policy";
2388 return r;
2389 }
2390
2391 r = ima_setup();
2392 if (r < 0) {
2393 *ret_error_message = "Failed to load IMA policy";
2394 return r;
2395 }
2396
2397 dual_timestamp_get(security_finish_timestamp);
2398 return 0;
2399}
2400
263162da
LP		 2401static void test_summary(Manager *m) {
2402 assert(m);
2403
2404 printf("-> By units:\n");
2405 manager_dump_units(m, stdout, "\t");
2406
2407 printf("-> By jobs:\n");
2408 manager_dump_jobs(m, stdout, "\t");
2409}
2410
efeb853f
LP		 2411static int collect_fds(FDSet **ret_fds, const char **ret_error_message) {
2412 int r;
2413
2414 assert(ret_fds);
2415 assert(ret_error_message);
2416
2417 r = fdset_new_fill(ret_fds);
2418 if (r < 0) {
2419 *ret_error_message = "Failed to allocate fd set";
2420 return log_emergency_errno(r, "Failed to allocate fd set: %m");
2421 }
2422
2423 fdset_cloexec(*ret_fds, true);
2424
2425 if (arg_serialization)
2426 assert_se(fdset_remove(*ret_fds, fileno(arg_serialization)) >= 0);
2427
2428 return 0;
2429}
2430
2e51b31c
LP		 2431static void setup_console_terminal(bool skip_setup) {
2432
2433 if (!arg_system)
2434 return;
2435
2436 /* Become a session leader if we aren't one yet. */
2437 (void) setsid();
2438
2439 /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2440 * tty. */
2441 (void) release_terminal();
2442
2443 /* Reset the console, but only if this is really init and we are freshly booted */
2444 if (getpid_cached() == 1 && !skip_setup)
2445 (void) console_setup();
2446}
2447
aa40ff07
LP		 2448static bool early_skip_setup_check(int argc, char *argv[]) {
2449 bool found_deserialize = false;
2450 int i;
2451
2452 /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2453 * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2454 * anyway, even if in that case we also do deserialization. */
2455
2456 for (i = 1; i < argc; i++) {
aa40ff07
LP		 2457 if (streq(argv[i], "--switched-root"))
2458 return false; /* If we switched root, don't skip the setup. */
2459 else if (streq(argv[i], "--deserialize"))
2460 found_deserialize = true;
2461 }
2462
2463 return found_deserialize; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2464}
2465
60918275 2466int main(int argc, char *argv[]) {
625e8690
LP		 2467
2468 dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL, userspace_timestamp = DUAL_TIMESTAMP_NULL, kernel_timestamp = DUAL_TIMESTAMP_NULL,
2469 security_start_timestamp = DUAL_TIMESTAMP_NULL, security_finish_timestamp = DUAL_TIMESTAMP_NULL;
ddfa8b0b
LP		 2470 struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0),
2471 saved_rlimit_memlock = RLIMIT_MAKE_CONST(RLIM_INFINITY); /* The original rlimits we passed
2472 * in. Note we use different values
2473 * for the two that indicate whether
2474 * these fields are initialized! */
625e8690
LP		 2475 bool skip_setup, loaded_policy = false, queue_default_job = false, first_boot = false, reexecute = false;
2476 char *switch_root_dir = NULL, *switch_root_init = NULL;
9d76d730 2477 usec_t before_startup, after_startup;
625e8690 2478 static char systemd[] = "systemd";
9d76d730 2479 char timespan[FORMAT_TIMESPAN_MAX];
625e8690
LP		 2480 const char *shutdown_verb = NULL, *error_message = NULL;
2481 int r, retval = EXIT_FAILURE;
2482 Manager *m = NULL;
a16e1123 2483 FDSet *fds = NULL;
27b14a22 2484
d72a8f10 2485 /* SysV compatibility: redirect init → telinit */
6808a0bc 2486 redirect_telinit(argc, argv);
2cb1a60d 2487
d72a8f10 2488 /* Take timestamps early on */
c3a170f3
HH		 2489 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
2490 dual_timestamp_get(&userspace_timestamp);
2491
d72a8f10
LP		 2492 /* Figure out whether we need to do initialize the system, or if we already did that because we are
2493 * reexecuting */
aa40ff07 2494 skip_setup = early_skip_setup_check(argc, argv);
d03bc1b8 2495
d72a8f10
LP		 2496 /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2497 * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
f3b6a3ed 2498 program_invocation_short_name = systemd;
eee8b7ab 2499 (void) prctl(PR_SET_NAME, systemd);
5d6b1584 2500
d72a8f10 2501 /* Save the original command line */
36fea155 2502 save_argc_argv(argc, argv);
f3b6a3ed 2503
6fdb8de4 2504 /* Make sure that if the user says "syslog" we actually log to the journal. */
c1dc6153 2505 log_set_upgrade_syslog_to_journal(true);
bbe63281 2506
df0ff127 2507 if (getpid_cached() == 1) {
b5752d23
LP		 2508 /* When we run as PID 1 force system mode */
2509 arg_system = true;
2510
48a601fe 2511 /* Disable the umask logic */
90dc8c2e
MG		 2512 umask(0);
2513
92890452
LP		 2514 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2515 * activated yet (even though the log socket for it exists). */
d075092f
LP		 2516 log_set_prohibit_ipc(true);
2517
48a601fe
LP		 2518 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2519 * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2520 * child process right before execve()'ing the actual binary, at a point in time where socket
2521 * activation stderr/stdout area already set up. */
2522 log_set_always_reopen_console(true);
48a601fe 2523
92890452 2524 if (detect_container() <= 0) {
4f8d551f 2525
92890452 2526 /* Running outside of a container as PID 1 */
92890452
LP		 2527 log_set_target(LOG_TARGET_KMSG);
2528 log_open();
a866073d 2529
92890452
LP		 2530 if (in_initrd())
2531 initrd_timestamp = userspace_timestamp;
c3ba6250 2532
92890452
LP		 2533 if (!skip_setup) {
2534 r = mount_setup_early();
2535 if (r < 0) {
2536 error_message = "Failed to mount early API filesystems";
2537 goto finish;
2538 }
2539
2540 r = initialize_security(
2541 &loaded_policy,
2542 &security_start_timestamp,
2543 &security_finish_timestamp,
2544 &error_message);
2545 if (r < 0)
2546 goto finish;
d723cd65 2547 }
eee8b7ab 2548
92890452
LP		 2549 if (mac_selinux_init() < 0) {
2550 error_message = "Failed to initialize SELinux policy";
96694e99 2551 goto finish;
92890452 2552 }
0b3325e7 2553
92890452
LP		 2554 if (!skip_setup)
2555 initialize_clock();
2556
2557 /* Set the default for later on, but don't actually open the logs like this for now. Note that
2558 * if we are transitioning from the initrd there might still be journal fd open, and we
2559 * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2560 * elsewhere. */
2561 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
2562
2563 } else {
2564 /* Running inside a container, as PID 1 */
92890452
LP		 2565 log_set_target(LOG_TARGET_CONSOLE);
2566 log_open();
2567
2568 /* For later on, see above... */
2569 log_set_target(LOG_TARGET_JOURNAL);
2570
2571 /* clear the kernel timestamp,
2572 * because we are in a container */
2573 kernel_timestamp = DUAL_TIMESTAMP_NULL;
cb6531be 2574 }
7948c4df 2575
92890452 2576 initialize_coredump(skip_setup);
a866073d 2577
92890452
LP		 2578 r = fixup_environment();
2579 if (r < 0) {
2580 log_emergency_errno(r, "Failed to fix up PID 1 environment: %m");
2581 error_message = "Failed to fix up PID1 environment";
2582 goto finish;
2583 }
a866073d 2584
c3a170f3 2585 } else {
a866073d 2586 /* Running as user instance */
463d0d15 2587 arg_system = false;
eeecf6e6 2588 log_set_target(LOG_TARGET_AUTO);
871e5809 2589 log_open();
c3a170f3
HH		 2590
2591 /* clear the kernel timestamp,
2592 * because we are not PID 1 */
6513d561 2593 kernel_timestamp = DUAL_TIMESTAMP_NULL;
bbe63281 2594 }
a5dab5ce 2595
6edefe0b 2596 if (arg_system) {
92890452
LP		 2597 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2598 * they never log into the console. */
3a18b604 2599 log_show_color(colors_enabled());
92890452 2600
c76cf844
AK		 2601 r = make_null_stdio();
2602 if (r < 0)
92890452 2603 log_warning_errno(r, "Failed to redirect standard streams to /dev/null, ignoring: %m");
3a18b604
FB		 2604 }
2605
f170852a
LP		 2606 /* Mount /proc, /sys and friends, so that /proc/cmdline and
2607 * /proc/$PID/fd is available. */
df0ff127 2608 if (getpid_cached() == 1) {
f84f9974 2609
a132bef0 2610 /* Load the kernel modules early. */
2e75e2a8
DM		 2611 if (!skip_setup)
2612 kmod_setup();
2e75e2a8 2613
0c85a4f3 2614 r = mount_setup(loaded_policy);
cb6531be
ZJS		 2615 if (r < 0) {
2616 error_message = "Failed to mount API filesystems";
8efe3c01 2617 goto finish;
cb6531be 2618 }
0c85a4f3 2619 }
4ade7963
LP		 2620
2621 /* Reset all signal handlers. */
ce30c8dc
LP		 2622 (void) reset_all_signal_handlers();
2623 (void) ignore_signals(SIGNALS_IGNORE, -1);
078e4539 2624
97d1fb94
LP		 2625 r = load_configuration(argc, argv, &error_message);
2626 if (r < 0)
f170852a 2627 goto finish;
10c961b9 2628
b0d7c989
LP		 2629 r = safety_checks();
2630 if (r < 0)
fe783b03 2631 goto finish;
fe783b03 2632
bbc1acab 2633 if (IN_SET(arg_action, ACTION_TEST, ACTION_HELP, ACTION_DUMP_CONFIGURATION_ITEMS, ACTION_DUMP_BUS_PROPERTIES))
0221d68a 2634 (void) pager_open(arg_pager_flags);
b0d7c989
LP		 2635
2636 if (arg_action != ACTION_RUN)
74e7579c 2637 skip_setup = true;
b87c2aa6 2638
fa0f4d8a 2639 if (arg_action == ACTION_HELP) {
37ec0fdd 2640 retval = help() < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
f170852a 2641 goto finish;
9ba0bc4e
ZJS		 2642 } else if (arg_action == ACTION_VERSION) {
2643 retval = version();
2644 goto finish;
fa0f4d8a 2645 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
e537352b 2646 unit_dump_config_items(stdout);
22f4096c 2647 retval = EXIT_SUCCESS;
e537352b 2648 goto finish;
bbc1acab
YW		 2649 } else if (arg_action == ACTION_DUMP_BUS_PROPERTIES) {
2650 dump_bus_properties(stdout);
2651 retval = EXIT_SUCCESS;
2652 goto finish;
f170852a
LP		 2653 }
2654
4c701096 2655 assert_se(IN_SET(arg_action, ACTION_RUN, ACTION_TEST));
f170852a 2656
5a2e0c62
LP		 2657 /* Move out of the way, so that we won't block unmounts */
2658 assert_se(chdir("/") == 0);
2659
dea374e8 2660 if (arg_action == ACTION_RUN) {
a70c72a0 2661
c6885f5f
FB		 2662 /* A core pattern might have been specified via the cmdline. */
2663 initialize_core_pattern(skip_setup);
2664
efeb853f 2665 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
a70c72a0
LP		 2666 log_close();
2667
2668 /* Remember open file descriptors for later deserialization */
efeb853f
LP		 2669 r = collect_fds(&fds, &error_message);
2670 if (r < 0)
dea374e8 2671 goto finish;
a16e1123 2672
2e51b31c
LP		 2673 /* Give up any control of the console, but make sure its initialized. */
2674 setup_console_terminal(skip_setup);
56d96fc0 2675
a70c72a0
LP		 2676 /* Open the logging devices, if possible and necessary */
2677 log_open();
56d96fc0 2678 }
4ade7963 2679
31aef7ff 2680 log_execution_mode(&first_boot);
a5dab5ce 2681
2d776038
LP		 2682 r = initialize_runtime(skip_setup,
2683 &saved_rlimit_nofile,
2684 &saved_rlimit_memlock,
2685 &error_message);
2686 if (r < 0)
2687 goto finish;
4096d6f5 2688
e0a3da1f
ZJS		 2689 r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER,
2690 arg_action == ACTION_TEST ? MANAGER_TEST_FULL : 0,
2691 &m);
e96d6be7 2692 if (r < 0) {
da927ba9 2693 log_emergency_errno(r, "Failed to allocate manager object: %m");
cb6531be 2694 error_message = "Failed to allocate manager object";
60918275
LP		 2695 goto finish;
2696 }
2697
9f9f0342
LP		 2698 m->timestamps[MANAGER_TIMESTAMP_KERNEL] = kernel_timestamp;
2699 m->timestamps[MANAGER_TIMESTAMP_INITRD] = initrd_timestamp;
2700 m->timestamps[MANAGER_TIMESTAMP_USERSPACE] = userspace_timestamp;
d4ee7bd8
YW		 2701 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START)] = security_start_timestamp;
2702 m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH)] = security_finish_timestamp;
9e58ff9c 2703
85cb4151 2704 set_manager_defaults(m);
7b46fc6a 2705 set_manager_settings(m);
fd130612 2706 manager_set_first_boot(m, first_boot);
27d340c7 2707
bf4df7c3 2708 /* Remember whether we should queue the default job */
d3b1c508 2709 queue_default_job = !arg_serialization || arg_switched_root;
bf4df7c3 2710
9d76d730
LP		 2711 before_startup = now(CLOCK_MONOTONIC);
2712
d3b1c508 2713 r = manager_startup(m, arg_serialization, fds);
58f88d92 2714 if (r < 0) {
cefb3eda 2715 error_message = "Failed to start up manager";
58f88d92
ZJS		 2716 goto finish;
2717 }
a16e1123 2718
6acca5fc 2719 /* This will close all file descriptors that were opened, but not claimed by any unit. */
2feceb5e 2720 fds = fdset_free(fds);
74ca738f 2721 arg_serialization = safe_fclose(arg_serialization);
bf4df7c3
LP		 2722
2723 if (queue_default_job) {
6acca5fc 2724 r = do_queue_default_job(m, &error_message);
718db961 2725 if (r < 0)
37d88da7 2726 goto finish;
6acca5fc 2727 }
ab17a050 2728
6acca5fc 2729 after_startup = now(CLOCK_MONOTONIC);
60918275 2730
6acca5fc
LP		 2731 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
2732 "Loaded units and determined initial transaction in %s.",
2733 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
07672f49 2734
6acca5fc 2735 if (arg_action == ACTION_TEST) {
263162da 2736 test_summary(m);
6acca5fc
LP		 2737 retval = EXIT_SUCCESS;
2738 goto finish;
e965d56d 2739 }
d46de8a1 2740
3046b6db
LP		 2741 (void) invoke_main_loop(m,
2742 &reexecute,
2743 &retval,
2744 &shutdown_verb,
2745 &fds,
2746 &switch_root_dir,
2747 &switch_root_init,
2748 &error_message);
f170852a 2749
60918275 2750finish:
b87c2aa6
ZJS		 2751 pager_close();
2752
92890452 2753 if (m) {
81f5fc2d 2754 arg_shutdown_watchdog = m->shutdown_watchdog;
92890452
LP		 2755 m = manager_free(m);
2756 }
60918275 2757
fb39af4c 2758 reset_arguments();
cc56fafe 2759 mac_selinux_finish();
b2bb3dbe 2760
3c7878f9
LP		 2761 if (reexecute)
2762 do_reexecute(argc, argv,
2763 &saved_rlimit_nofile,
2764 &saved_rlimit_memlock,
2765 fds,
2766 switch_root_dir,
2767 switch_root_init,
2768 &error_message); /* This only returns if reexecution failed */
a16e1123 2769
74ca738f 2770 arg_serialization = safe_fclose(arg_serialization);
2feceb5e 2771 fds = fdset_free(fds);
a16e1123 2772
349cc4a5 2773#if HAVE_VALGRIND_VALGRIND_H
54b434b1
LP		 2774 /* If we are PID 1 and running under valgrind, then let's exit
2775 * here explicitly. valgrind will only generate nice output on
2776 * exit(), not on exec(), hence let's do the former not the
2777 * latter here. */
8a2c1fbf
EJ		 2778 if (getpid_cached() == 1 && RUNNING_ON_VALGRIND) {
2779 /* Cleanup watchdog_device strings for valgrind. We need them
2780 * in become_shutdown() so normally we cannot free them yet. */
2781 watchdog_free_device();
2782 arg_watchdog_device = mfree(arg_watchdog_device);
27fe58b7 2783 return retval;
8a2c1fbf 2784 }
54b434b1
LP		 2785#endif
2786
7e11a95e
EV		 2787#if HAS_FEATURE_ADDRESS_SANITIZER
2788 __lsan_do_leak_check();
2789#endif
2790
b9080b03 2791 if (shutdown_verb) {
7eb35049 2792 r = become_shutdown(shutdown_verb, retval);
4a36297c 2793 log_error_errno(r, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
9b9881d7 2794 error_message = "Failed to execute shutdown binary";
b9080b03
FF		 2795 }
2796
8a2c1fbf
EJ		 2797 watchdog_free_device();
2798 arg_watchdog_device = mfree(arg_watchdog_device);
2799
df0ff127 2800 if (getpid_cached() == 1) {
cb6531be
ZJS		 2801 if (error_message)
2802 manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
1fc464f6 2803 ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
bb259772
LP		 2804 "%s.", error_message);
2805 freeze_or_exit_or_reboot();
cb6531be 2806 }
c3b3c274 2807
60918275
LP		 2808 return retval;
2809}
Unnamed repository; edit this file 'description' to name the repository.