]>
Commit | Line | Data |
---|---|---|
1 | /*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ | |
2 | ||
3 | /*** | |
4 | This file is part of systemd. | |
5 | ||
6 | Copyright 2010 Lennart Poettering | |
7 | ||
8 | systemd is free software; you can redistribute it and/or modify it | |
9 | under the terms of the GNU Lesser General Public License as published by | |
10 | the Free Software Foundation; either version 2.1 of the License, or | |
11 | (at your option) any later version. | |
12 | ||
13 | systemd is distributed in the hope that it will be useful, but | |
14 | WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
16 | Lesser General Public License for more details. | |
17 | ||
18 | You should have received a copy of the GNU Lesser General Public License | |
19 | along with systemd; If not, see <http://www.gnu.org/licenses/>. | |
20 | ***/ | |
21 | ||
22 | #include <errno.h> | |
23 | #include <fcntl.h> | |
24 | #include <getopt.h> | |
25 | #include <signal.h> | |
26 | #include <stdio.h> | |
27 | #include <string.h> | |
28 | #include <sys/mount.h> | |
29 | #include <sys/prctl.h> | |
30 | #include <sys/reboot.h> | |
31 | #include <sys/stat.h> | |
32 | #include <unistd.h> | |
33 | #ifdef HAVE_SECCOMP | |
34 | #include <seccomp.h> | |
35 | #endif | |
36 | #ifdef HAVE_VALGRIND_VALGRIND_H | |
37 | #include <valgrind/valgrind.h> | |
38 | #endif | |
39 | ||
40 | #include "sd-daemon.h" | |
41 | #include "sd-bus.h" | |
42 | ||
43 | #include "architecture.h" | |
44 | #include "build.h" | |
45 | #include "bus-error.h" | |
46 | #include "bus-util.h" | |
47 | #include "capability.h" | |
48 | #include "clock-util.h" | |
49 | #include "conf-parser.h" | |
50 | #include "cpu-set-util.h" | |
51 | #include "dbus-manager.h" | |
52 | #include "def.h" | |
53 | #include "env-util.h" | |
54 | #include "fd-util.h" | |
55 | #include "fdset.h" | |
56 | #include "fileio.h" | |
57 | #include "formats-util.h" | |
58 | #include "hostname-setup.h" | |
59 | #include "ima-setup.h" | |
60 | #include "killall.h" | |
61 | #include "kmod-setup.h" | |
62 | #include "load-fragment.h" | |
63 | #include "log.h" | |
64 | #include "loopback-setup.h" | |
65 | #include "machine-id-setup.h" | |
66 | #include "manager.h" | |
67 | #include "missing.h" | |
68 | #include "mount-setup.h" | |
69 | #include "pager.h" | |
70 | #include "process-util.h" | |
71 | #include "selinux-setup.h" | |
72 | #include "selinux-util.h" | |
73 | #include "signal-util.h" | |
74 | #include "smack-setup.h" | |
75 | #include "special.h" | |
76 | #include "strv.h" | |
77 | #include "switch-root.h" | |
78 | #include "terminal-util.h" | |
79 | #include "virt.h" | |
80 | #include "watchdog.h" | |
81 | ||
82 | static enum { | |
83 | ACTION_RUN, | |
84 | ACTION_HELP, | |
85 | ACTION_VERSION, | |
86 | ACTION_TEST, | |
87 | ACTION_DUMP_CONFIGURATION_ITEMS, | |
88 | ACTION_DONE | |
89 | } arg_action = ACTION_RUN; | |
90 | static char *arg_default_unit = NULL; | |
91 | static ManagerRunningAs arg_running_as = _MANAGER_RUNNING_AS_INVALID; | |
92 | static bool arg_dump_core = true; | |
93 | static int arg_crash_chvt = -1; | |
94 | static bool arg_crash_shell = false; | |
95 | static bool arg_crash_reboot = false; | |
96 | static bool arg_confirm_spawn = false; | |
97 | static ShowStatus arg_show_status = _SHOW_STATUS_UNSET; | |
98 | static bool arg_switched_root = false; | |
99 | static int arg_no_pager = -1; | |
100 | static char ***arg_join_controllers = NULL; | |
101 | static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL; | |
102 | static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT; | |
103 | static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC; | |
104 | static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC; | |
105 | static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC; | |
106 | static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL; | |
107 | static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST; | |
108 | static usec_t arg_runtime_watchdog = 0; | |
109 | static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE; | |
110 | static char **arg_default_environment = NULL; | |
111 | static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {}; | |
112 | static uint64_t arg_capability_bounding_set_drop = 0; | |
113 | static nsec_t arg_timer_slack_nsec = NSEC_INFINITY; | |
114 | static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE; | |
115 | static Set* arg_syscall_archs = NULL; | |
116 | static FILE* arg_serialization = NULL; | |
117 | static bool arg_default_cpu_accounting = false; | |
118 | static bool arg_default_blockio_accounting = false; | |
119 | static bool arg_default_memory_accounting = false; | |
120 | static bool arg_default_tasks_accounting = false; | |
121 | ||
122 | static void pager_open_if_enabled(void) { | |
123 | ||
124 | if (arg_no_pager <= 0) | |
125 | return; | |
126 | ||
127 | pager_open(false); | |
128 | } | |
129 | ||
130 | noreturn static void freeze_or_reboot(void) { | |
131 | ||
132 | if (arg_crash_reboot) { | |
133 | log_notice("Rebooting in 10s..."); | |
134 | (void) sleep(10); | |
135 | ||
136 | log_notice("Rebooting now..."); | |
137 | (void) reboot(RB_AUTOBOOT); | |
138 | log_emergency_errno(errno, "Failed to reboot: %m"); | |
139 | } | |
140 | ||
141 | log_emergency("Freezing execution."); | |
142 | freeze(); | |
143 | } | |
144 | ||
145 | noreturn static void crash(int sig) { | |
146 | struct sigaction sa; | |
147 | pid_t pid; | |
148 | ||
149 | if (getpid() != 1) | |
150 | /* Pass this on immediately, if this is not PID 1 */ | |
151 | (void) raise(sig); | |
152 | else if (!arg_dump_core) | |
153 | log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig)); | |
154 | else { | |
155 | sa = (struct sigaction) { | |
156 | .sa_handler = nop_signal_handler, | |
157 | .sa_flags = SA_NOCLDSTOP|SA_RESTART, | |
158 | }; | |
159 | ||
160 | /* We want to wait for the core process, hence let's enable SIGCHLD */ | |
161 | (void) sigaction(SIGCHLD, &sa, NULL); | |
162 | ||
163 | pid = raw_clone(SIGCHLD, NULL); | |
164 | if (pid < 0) | |
165 | log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig)); | |
166 | else if (pid == 0) { | |
167 | struct rlimit rl = { | |
168 | .rlim_cur = RLIM_INFINITY, | |
169 | .rlim_max = RLIM_INFINITY, | |
170 | }; | |
171 | ||
172 | /* Enable default signal handler for core dump */ | |
173 | sa = (struct sigaction) { | |
174 | .sa_handler = SIG_DFL, | |
175 | }; | |
176 | (void) sigaction(sig, &sa, NULL); | |
177 | ||
178 | /* Don't limit the core dump size */ | |
179 | (void) setrlimit(RLIMIT_CORE, &rl); | |
180 | ||
181 | /* Just to be sure... */ | |
182 | (void) chdir("/"); | |
183 | ||
184 | /* Raise the signal again */ | |
185 | pid = raw_getpid(); | |
186 | (void) kill(pid, sig); /* raise() would kill the parent */ | |
187 | ||
188 | assert_not_reached("We shouldn't be here..."); | |
189 | _exit(EXIT_FAILURE); | |
190 | } else { | |
191 | siginfo_t status; | |
192 | int r; | |
193 | ||
194 | /* Order things nicely. */ | |
195 | r = wait_for_terminate(pid, &status); | |
196 | if (r < 0) | |
197 | log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig)); | |
198 | else if (status.si_code != CLD_DUMPED) | |
199 | log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).", | |
200 | signal_to_string(sig), | |
201 | pid, sigchld_code_to_string(status.si_code), | |
202 | status.si_status, | |
203 | strna(status.si_code == CLD_EXITED | |
204 | ? exit_status_to_string(status.si_status, EXIT_STATUS_FULL) | |
205 | : signal_to_string(status.si_status))); | |
206 | else | |
207 | log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid); | |
208 | } | |
209 | } | |
210 | ||
211 | if (arg_crash_chvt >= 0) | |
212 | (void) chvt(arg_crash_chvt); | |
213 | ||
214 | sa = (struct sigaction) { | |
215 | .sa_handler = SIG_IGN, | |
216 | .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART, | |
217 | }; | |
218 | ||
219 | /* Let the kernel reap children for us */ | |
220 | (void) sigaction(SIGCHLD, &sa, NULL); | |
221 | ||
222 | if (arg_crash_shell) { | |
223 | log_notice("Executing crash shell in 10s..."); | |
224 | (void) sleep(10); | |
225 | ||
226 | pid = raw_clone(SIGCHLD, NULL); | |
227 | if (pid < 0) | |
228 | log_emergency_errno(errno, "Failed to fork off crash shell: %m"); | |
229 | else if (pid == 0) { | |
230 | (void) setsid(); | |
231 | (void) make_console_stdio(); | |
232 | (void) execle("/bin/sh", "/bin/sh", NULL, environ); | |
233 | ||
234 | log_emergency_errno(errno, "execle() failed: %m"); | |
235 | _exit(EXIT_FAILURE); | |
236 | } else { | |
237 | log_info("Spawned crash shell as PID "PID_FMT".", pid); | |
238 | (void) wait_for_terminate(pid, NULL); | |
239 | } | |
240 | } | |
241 | ||
242 | freeze_or_reboot(); | |
243 | } | |
244 | ||
245 | static void install_crash_handler(void) { | |
246 | static const struct sigaction sa = { | |
247 | .sa_handler = crash, | |
248 | .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */ | |
249 | }; | |
250 | int r; | |
251 | ||
252 | /* We ignore the return value here, since, we don't mind if we | |
253 | * cannot set up a crash handler */ | |
254 | r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1); | |
255 | if (r < 0) | |
256 | log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m"); | |
257 | } | |
258 | ||
259 | static int console_setup(void) { | |
260 | _cleanup_close_ int tty_fd = -1; | |
261 | int r; | |
262 | ||
263 | tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC); | |
264 | if (tty_fd < 0) | |
265 | return log_error_errno(tty_fd, "Failed to open /dev/console: %m"); | |
266 | ||
267 | /* We don't want to force text mode. plymouth may be showing | |
268 | * pictures already from initrd. */ | |
269 | r = reset_terminal_fd(tty_fd, false); | |
270 | if (r < 0) | |
271 | return log_error_errno(r, "Failed to reset /dev/console: %m"); | |
272 | ||
273 | return 0; | |
274 | } | |
275 | ||
276 | static int parse_crash_chvt(const char *value) { | |
277 | int b; | |
278 | ||
279 | if (safe_atoi(value, &arg_crash_chvt) >= 0) | |
280 | return 0; | |
281 | ||
282 | b = parse_boolean(value); | |
283 | if (b < 0) | |
284 | return b; | |
285 | ||
286 | if (b > 0) | |
287 | arg_crash_chvt = 0; /* switch to where kmsg goes */ | |
288 | else | |
289 | arg_crash_chvt = -1; /* turn off switching */ | |
290 | ||
291 | return 0; | |
292 | } | |
293 | ||
294 | static int parse_proc_cmdline_item(const char *key, const char *value) { | |
295 | ||
296 | static const char * const rlmap[] = { | |
297 | "emergency", SPECIAL_EMERGENCY_TARGET, | |
298 | "-b", SPECIAL_EMERGENCY_TARGET, | |
299 | "rescue", SPECIAL_RESCUE_TARGET, | |
300 | "single", SPECIAL_RESCUE_TARGET, | |
301 | "-s", SPECIAL_RESCUE_TARGET, | |
302 | "s", SPECIAL_RESCUE_TARGET, | |
303 | "S", SPECIAL_RESCUE_TARGET, | |
304 | "1", SPECIAL_RESCUE_TARGET, | |
305 | "2", SPECIAL_MULTI_USER_TARGET, | |
306 | "3", SPECIAL_MULTI_USER_TARGET, | |
307 | "4", SPECIAL_MULTI_USER_TARGET, | |
308 | "5", SPECIAL_GRAPHICAL_TARGET, | |
309 | }; | |
310 | int r; | |
311 | ||
312 | assert(key); | |
313 | ||
314 | if (streq(key, "systemd.unit") && value) { | |
315 | ||
316 | if (!in_initrd()) | |
317 | return free_and_strdup(&arg_default_unit, value); | |
318 | ||
319 | } else if (streq(key, "rd.systemd.unit") && value) { | |
320 | ||
321 | if (in_initrd()) | |
322 | return free_and_strdup(&arg_default_unit, value); | |
323 | ||
324 | } else if (streq(key, "systemd.dump_core") && value) { | |
325 | ||
326 | r = parse_boolean(value); | |
327 | if (r < 0) | |
328 | log_warning("Failed to parse dump core switch %s. Ignoring.", value); | |
329 | else | |
330 | arg_dump_core = r; | |
331 | ||
332 | } else if (streq(key, "systemd.crash_chvt") && value) { | |
333 | ||
334 | if (parse_crash_chvt(value) < 0) | |
335 | log_warning("Failed to parse crash chvt switch %s. Ignoring.", value); | |
336 | ||
337 | } else if (streq(key, "systemd.crash_shell") && value) { | |
338 | ||
339 | r = parse_boolean(value); | |
340 | if (r < 0) | |
341 | log_warning("Failed to parse crash shell switch %s. Ignoring.", value); | |
342 | else | |
343 | arg_crash_shell = r; | |
344 | ||
345 | } else if (streq(key, "systemd.crash_reboot") && value) { | |
346 | ||
347 | r = parse_boolean(value); | |
348 | if (r < 0) | |
349 | log_warning("Failed to parse crash reboot switch %s. Ignoring.", value); | |
350 | else | |
351 | arg_crash_reboot = r; | |
352 | ||
353 | } else if (streq(key, "systemd.confirm_spawn") && value) { | |
354 | ||
355 | r = parse_boolean(value); | |
356 | if (r < 0) | |
357 | log_warning("Failed to parse confirm spawn switch %s. Ignoring.", value); | |
358 | else | |
359 | arg_confirm_spawn = r; | |
360 | ||
361 | } else if (streq(key, "systemd.show_status") && value) { | |
362 | ||
363 | r = parse_show_status(value, &arg_show_status); | |
364 | if (r < 0) | |
365 | log_warning("Failed to parse show status switch %s. Ignoring.", value); | |
366 | ||
367 | } else if (streq(key, "systemd.default_standard_output") && value) { | |
368 | ||
369 | r = exec_output_from_string(value); | |
370 | if (r < 0) | |
371 | log_warning("Failed to parse default standard output switch %s. Ignoring.", value); | |
372 | else | |
373 | arg_default_std_output = r; | |
374 | ||
375 | } else if (streq(key, "systemd.default_standard_error") && value) { | |
376 | ||
377 | r = exec_output_from_string(value); | |
378 | if (r < 0) | |
379 | log_warning("Failed to parse default standard error switch %s. Ignoring.", value); | |
380 | else | |
381 | arg_default_std_error = r; | |
382 | ||
383 | } else if (streq(key, "systemd.setenv") && value) { | |
384 | ||
385 | if (env_assignment_is_valid(value)) { | |
386 | char **env; | |
387 | ||
388 | env = strv_env_set(arg_default_environment, value); | |
389 | if (env) | |
390 | arg_default_environment = env; | |
391 | else | |
392 | log_warning_errno(ENOMEM, "Setting environment variable '%s' failed, ignoring: %m", value); | |
393 | } else | |
394 | log_warning("Environment variable name '%s' is not valid. Ignoring.", value); | |
395 | ||
396 | } else if (streq(key, "quiet") && !value) { | |
397 | ||
398 | if (arg_show_status == _SHOW_STATUS_UNSET) | |
399 | arg_show_status = SHOW_STATUS_AUTO; | |
400 | ||
401 | } else if (streq(key, "debug") && !value) { | |
402 | ||
403 | /* Note that log_parse_environment() handles 'debug' | |
404 | * too, and sets the log level to LOG_DEBUG. */ | |
405 | ||
406 | if (detect_container() > 0) | |
407 | log_set_target(LOG_TARGET_CONSOLE); | |
408 | ||
409 | } else if (!in_initrd() && !value) { | |
410 | unsigned i; | |
411 | ||
412 | /* SysV compatibility */ | |
413 | for (i = 0; i < ELEMENTSOF(rlmap); i += 2) | |
414 | if (streq(key, rlmap[i])) | |
415 | return free_and_strdup(&arg_default_unit, rlmap[i+1]); | |
416 | } | |
417 | ||
418 | return 0; | |
419 | } | |
420 | ||
421 | #define DEFINE_SETTER(name, func, descr) \ | |
422 | static int name(const char *unit, \ | |
423 | const char *filename, \ | |
424 | unsigned line, \ | |
425 | const char *section, \ | |
426 | unsigned section_line, \ | |
427 | const char *lvalue, \ | |
428 | int ltype, \ | |
429 | const char *rvalue, \ | |
430 | void *data, \ | |
431 | void *userdata) { \ | |
432 | \ | |
433 | int r; \ | |
434 | \ | |
435 | assert(filename); \ | |
436 | assert(lvalue); \ | |
437 | assert(rvalue); \ | |
438 | \ | |
439 | r = func(rvalue); \ | |
440 | if (r < 0) \ | |
441 | log_syntax(unit, LOG_ERR, filename, line, r, \ | |
442 | "Invalid " descr "'%s': %m", \ | |
443 | rvalue); \ | |
444 | \ | |
445 | return 0; \ | |
446 | } | |
447 | ||
448 | DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level") | |
449 | DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target") | |
450 | DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" ) | |
451 | DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location") | |
452 | ||
453 | static int config_parse_cpu_affinity2( | |
454 | const char *unit, | |
455 | const char *filename, | |
456 | unsigned line, | |
457 | const char *section, | |
458 | unsigned section_line, | |
459 | const char *lvalue, | |
460 | int ltype, | |
461 | const char *rvalue, | |
462 | void *data, | |
463 | void *userdata) { | |
464 | ||
465 | _cleanup_cpu_free_ cpu_set_t *c = NULL; | |
466 | int ncpus; | |
467 | ||
468 | ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue); | |
469 | if (ncpus < 0) | |
470 | return ncpus; | |
471 | ||
472 | if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0) | |
473 | log_warning("Failed to set CPU affinity: %m"); | |
474 | ||
475 | return 0; | |
476 | } | |
477 | ||
478 | static int config_parse_show_status( | |
479 | const char* unit, | |
480 | const char *filename, | |
481 | unsigned line, | |
482 | const char *section, | |
483 | unsigned section_line, | |
484 | const char *lvalue, | |
485 | int ltype, | |
486 | const char *rvalue, | |
487 | void *data, | |
488 | void *userdata) { | |
489 | ||
490 | int k; | |
491 | ShowStatus *b = data; | |
492 | ||
493 | assert(filename); | |
494 | assert(lvalue); | |
495 | assert(rvalue); | |
496 | assert(data); | |
497 | ||
498 | k = parse_show_status(rvalue, b); | |
499 | if (k < 0) { | |
500 | log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue); | |
501 | return 0; | |
502 | } | |
503 | ||
504 | return 0; | |
505 | } | |
506 | ||
507 | static int config_parse_crash_chvt( | |
508 | const char* unit, | |
509 | const char *filename, | |
510 | unsigned line, | |
511 | const char *section, | |
512 | unsigned section_line, | |
513 | const char *lvalue, | |
514 | int ltype, | |
515 | const char *rvalue, | |
516 | void *data, | |
517 | void *userdata) { | |
518 | ||
519 | int r; | |
520 | ||
521 | assert(filename); | |
522 | assert(lvalue); | |
523 | assert(rvalue); | |
524 | ||
525 | r = parse_crash_chvt(rvalue); | |
526 | if (r < 0) { | |
527 | log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue); | |
528 | return 0; | |
529 | } | |
530 | ||
531 | return 0; | |
532 | } | |
533 | ||
534 | static int config_parse_join_controllers(const char *unit, | |
535 | const char *filename, | |
536 | unsigned line, | |
537 | const char *section, | |
538 | unsigned section_line, | |
539 | const char *lvalue, | |
540 | int ltype, | |
541 | const char *rvalue, | |
542 | void *data, | |
543 | void *userdata) { | |
544 | ||
545 | const char *whole_rvalue = rvalue; | |
546 | unsigned n = 0; | |
547 | ||
548 | assert(filename); | |
549 | assert(lvalue); | |
550 | assert(rvalue); | |
551 | ||
552 | arg_join_controllers = strv_free_free(arg_join_controllers); | |
553 | ||
554 | for (;;) { | |
555 | _cleanup_free_ char *word = NULL; | |
556 | char **l; | |
557 | int r; | |
558 | ||
559 | r = extract_first_word(&rvalue, &word, WHITESPACE, EXTRACT_QUOTES); | |
560 | if (r < 0) { | |
561 | log_syntax(unit, LOG_ERR, filename, line, r, "Invalid value for %s: %s", lvalue, whole_rvalue); | |
562 | return r; | |
563 | } | |
564 | if (r == 0) | |
565 | break; | |
566 | ||
567 | l = strv_split(word, ","); | |
568 | if (!l) | |
569 | return log_oom(); | |
570 | strv_uniq(l); | |
571 | ||
572 | if (strv_length(l) <= 1) { | |
573 | strv_free(l); | |
574 | continue; | |
575 | } | |
576 | ||
577 | if (!arg_join_controllers) { | |
578 | arg_join_controllers = new(char**, 2); | |
579 | if (!arg_join_controllers) { | |
580 | strv_free(l); | |
581 | return log_oom(); | |
582 | } | |
583 | ||
584 | arg_join_controllers[0] = l; | |
585 | arg_join_controllers[1] = NULL; | |
586 | ||
587 | n = 1; | |
588 | } else { | |
589 | char ***a; | |
590 | char ***t; | |
591 | ||
592 | t = new0(char**, n+2); | |
593 | if (!t) { | |
594 | strv_free(l); | |
595 | return log_oom(); | |
596 | } | |
597 | ||
598 | n = 0; | |
599 | ||
600 | for (a = arg_join_controllers; *a; a++) { | |
601 | ||
602 | if (strv_overlap(*a, l)) { | |
603 | if (strv_extend_strv(&l, *a, false) < 0) { | |
604 | strv_free(l); | |
605 | strv_free_free(t); | |
606 | return log_oom(); | |
607 | } | |
608 | ||
609 | } else { | |
610 | char **c; | |
611 | ||
612 | c = strv_copy(*a); | |
613 | if (!c) { | |
614 | strv_free(l); | |
615 | strv_free_free(t); | |
616 | return log_oom(); | |
617 | } | |
618 | ||
619 | t[n++] = c; | |
620 | } | |
621 | } | |
622 | ||
623 | t[n++] = strv_uniq(l); | |
624 | ||
625 | strv_free_free(arg_join_controllers); | |
626 | arg_join_controllers = t; | |
627 | } | |
628 | } | |
629 | if (!isempty(rvalue)) | |
630 | log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring."); | |
631 | ||
632 | return 0; | |
633 | } | |
634 | ||
635 | static int parse_config_file(void) { | |
636 | ||
637 | const ConfigTableItem items[] = { | |
638 | { "Manager", "LogLevel", config_parse_level2, 0, NULL }, | |
639 | { "Manager", "LogTarget", config_parse_target, 0, NULL }, | |
640 | { "Manager", "LogColor", config_parse_color, 0, NULL }, | |
641 | { "Manager", "LogLocation", config_parse_location, 0, NULL }, | |
642 | { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core }, | |
643 | { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL }, | |
644 | { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL }, | |
645 | { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell }, | |
646 | { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot }, | |
647 | { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status }, | |
648 | { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL }, | |
649 | { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers }, | |
650 | { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog }, | |
651 | { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog }, | |
652 | { "Manager", "CapabilityBoundingSet", config_parse_bounding_set, 0, &arg_capability_bounding_set_drop }, | |
653 | #ifdef HAVE_SECCOMP | |
654 | { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs }, | |
655 | #endif | |
656 | { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec }, | |
657 | { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec }, | |
658 | { "Manager", "DefaultStandardOutput", config_parse_output, 0, &arg_default_std_output }, | |
659 | { "Manager", "DefaultStandardError", config_parse_output, 0, &arg_default_std_error }, | |
660 | { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec }, | |
661 | { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec }, | |
662 | { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec }, | |
663 | { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, | |
664 | { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst }, | |
665 | { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment }, | |
666 | { "Manager", "DefaultLimitCPU", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CPU] }, | |
667 | { "Manager", "DefaultLimitFSIZE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_FSIZE] }, | |
668 | { "Manager", "DefaultLimitDATA", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_DATA] }, | |
669 | { "Manager", "DefaultLimitSTACK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_STACK] }, | |
670 | { "Manager", "DefaultLimitCORE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CORE] }, | |
671 | { "Manager", "DefaultLimitRSS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RSS] }, | |
672 | { "Manager", "DefaultLimitNOFILE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NOFILE] }, | |
673 | { "Manager", "DefaultLimitAS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_AS] }, | |
674 | { "Manager", "DefaultLimitNPROC", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NPROC] }, | |
675 | { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MEMLOCK] }, | |
676 | { "Manager", "DefaultLimitLOCKS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_LOCKS] }, | |
677 | { "Manager", "DefaultLimitSIGPENDING", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_SIGPENDING] }, | |
678 | { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MSGQUEUE] }, | |
679 | { "Manager", "DefaultLimitNICE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NICE] }, | |
680 | { "Manager", "DefaultLimitRTPRIO", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTPRIO] }, | |
681 | { "Manager", "DefaultLimitRTTIME", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTTIME] }, | |
682 | { "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_default_cpu_accounting }, | |
683 | { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting }, | |
684 | { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting }, | |
685 | { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting }, | |
686 | {} | |
687 | }; | |
688 | ||
689 | const char *fn, *conf_dirs_nulstr; | |
690 | ||
691 | fn = arg_running_as == MANAGER_SYSTEM ? PKGSYSCONFDIR "/system.conf" : PKGSYSCONFDIR "/user.conf"; | |
692 | conf_dirs_nulstr = arg_running_as == MANAGER_SYSTEM ? CONF_DIRS_NULSTR("systemd/system.conf") : CONF_DIRS_NULSTR("systemd/user.conf"); | |
693 | config_parse_many(fn, conf_dirs_nulstr, "Manager\0", | |
694 | config_item_table_lookup, items, false, NULL); | |
695 | ||
696 | return 0; | |
697 | } | |
698 | ||
699 | static void manager_set_defaults(Manager *m) { | |
700 | ||
701 | assert(m); | |
702 | ||
703 | m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec; | |
704 | m->default_std_output = arg_default_std_output; | |
705 | m->default_std_error = arg_default_std_error; | |
706 | m->default_timeout_start_usec = arg_default_timeout_start_usec; | |
707 | m->default_timeout_stop_usec = arg_default_timeout_stop_usec; | |
708 | m->default_restart_usec = arg_default_restart_usec; | |
709 | m->default_start_limit_interval = arg_default_start_limit_interval; | |
710 | m->default_start_limit_burst = arg_default_start_limit_burst; | |
711 | m->default_cpu_accounting = arg_default_cpu_accounting; | |
712 | m->default_blockio_accounting = arg_default_blockio_accounting; | |
713 | m->default_memory_accounting = arg_default_memory_accounting; | |
714 | m->default_tasks_accounting = arg_default_tasks_accounting; | |
715 | ||
716 | manager_set_default_rlimits(m, arg_default_rlimit); | |
717 | manager_environment_add(m, NULL, arg_default_environment); | |
718 | } | |
719 | ||
720 | static int parse_argv(int argc, char *argv[]) { | |
721 | ||
722 | enum { | |
723 | ARG_LOG_LEVEL = 0x100, | |
724 | ARG_LOG_TARGET, | |
725 | ARG_LOG_COLOR, | |
726 | ARG_LOG_LOCATION, | |
727 | ARG_UNIT, | |
728 | ARG_SYSTEM, | |
729 | ARG_USER, | |
730 | ARG_TEST, | |
731 | ARG_NO_PAGER, | |
732 | ARG_VERSION, | |
733 | ARG_DUMP_CONFIGURATION_ITEMS, | |
734 | ARG_DUMP_CORE, | |
735 | ARG_CRASH_CHVT, | |
736 | ARG_CRASH_SHELL, | |
737 | ARG_CRASH_REBOOT, | |
738 | ARG_CONFIRM_SPAWN, | |
739 | ARG_SHOW_STATUS, | |
740 | ARG_DESERIALIZE, | |
741 | ARG_SWITCHED_ROOT, | |
742 | ARG_DEFAULT_STD_OUTPUT, | |
743 | ARG_DEFAULT_STD_ERROR | |
744 | }; | |
745 | ||
746 | static const struct option options[] = { | |
747 | { "log-level", required_argument, NULL, ARG_LOG_LEVEL }, | |
748 | { "log-target", required_argument, NULL, ARG_LOG_TARGET }, | |
749 | { "log-color", optional_argument, NULL, ARG_LOG_COLOR }, | |
750 | { "log-location", optional_argument, NULL, ARG_LOG_LOCATION }, | |
751 | { "unit", required_argument, NULL, ARG_UNIT }, | |
752 | { "system", no_argument, NULL, ARG_SYSTEM }, | |
753 | { "user", no_argument, NULL, ARG_USER }, | |
754 | { "test", no_argument, NULL, ARG_TEST }, | |
755 | { "no-pager", no_argument, NULL, ARG_NO_PAGER }, | |
756 | { "help", no_argument, NULL, 'h' }, | |
757 | { "version", no_argument, NULL, ARG_VERSION }, | |
758 | { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS }, | |
759 | { "dump-core", optional_argument, NULL, ARG_DUMP_CORE }, | |
760 | { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT }, | |
761 | { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL }, | |
762 | { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT }, | |
763 | { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN }, | |
764 | { "show-status", optional_argument, NULL, ARG_SHOW_STATUS }, | |
765 | { "deserialize", required_argument, NULL, ARG_DESERIALIZE }, | |
766 | { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT }, | |
767 | { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, }, | |
768 | { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, }, | |
769 | {} | |
770 | }; | |
771 | ||
772 | int c, r; | |
773 | ||
774 | assert(argc >= 1); | |
775 | assert(argv); | |
776 | ||
777 | if (getpid() == 1) | |
778 | opterr = 0; | |
779 | ||
780 | while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0) | |
781 | ||
782 | switch (c) { | |
783 | ||
784 | case ARG_LOG_LEVEL: | |
785 | r = log_set_max_level_from_string(optarg); | |
786 | if (r < 0) { | |
787 | log_error("Failed to parse log level %s.", optarg); | |
788 | return r; | |
789 | } | |
790 | ||
791 | break; | |
792 | ||
793 | case ARG_LOG_TARGET: | |
794 | r = log_set_target_from_string(optarg); | |
795 | if (r < 0) { | |
796 | log_error("Failed to parse log target %s.", optarg); | |
797 | return r; | |
798 | } | |
799 | ||
800 | break; | |
801 | ||
802 | case ARG_LOG_COLOR: | |
803 | ||
804 | if (optarg) { | |
805 | r = log_show_color_from_string(optarg); | |
806 | if (r < 0) { | |
807 | log_error("Failed to parse log color setting %s.", optarg); | |
808 | return r; | |
809 | } | |
810 | } else | |
811 | log_show_color(true); | |
812 | ||
813 | break; | |
814 | ||
815 | case ARG_LOG_LOCATION: | |
816 | if (optarg) { | |
817 | r = log_show_location_from_string(optarg); | |
818 | if (r < 0) { | |
819 | log_error("Failed to parse log location setting %s.", optarg); | |
820 | return r; | |
821 | } | |
822 | } else | |
823 | log_show_location(true); | |
824 | ||
825 | break; | |
826 | ||
827 | case ARG_DEFAULT_STD_OUTPUT: | |
828 | r = exec_output_from_string(optarg); | |
829 | if (r < 0) { | |
830 | log_error("Failed to parse default standard output setting %s.", optarg); | |
831 | return r; | |
832 | } else | |
833 | arg_default_std_output = r; | |
834 | break; | |
835 | ||
836 | case ARG_DEFAULT_STD_ERROR: | |
837 | r = exec_output_from_string(optarg); | |
838 | if (r < 0) { | |
839 | log_error("Failed to parse default standard error output setting %s.", optarg); | |
840 | return r; | |
841 | } else | |
842 | arg_default_std_error = r; | |
843 | break; | |
844 | ||
845 | case ARG_UNIT: | |
846 | ||
847 | r = free_and_strdup(&arg_default_unit, optarg); | |
848 | if (r < 0) | |
849 | return log_error_errno(r, "Failed to set default unit %s: %m", optarg); | |
850 | ||
851 | break; | |
852 | ||
853 | case ARG_SYSTEM: | |
854 | arg_running_as = MANAGER_SYSTEM; | |
855 | break; | |
856 | ||
857 | case ARG_USER: | |
858 | arg_running_as = MANAGER_USER; | |
859 | break; | |
860 | ||
861 | case ARG_TEST: | |
862 | arg_action = ACTION_TEST; | |
863 | if (arg_no_pager < 0) | |
864 | arg_no_pager = true; | |
865 | break; | |
866 | ||
867 | case ARG_NO_PAGER: | |
868 | arg_no_pager = true; | |
869 | break; | |
870 | ||
871 | case ARG_VERSION: | |
872 | arg_action = ACTION_VERSION; | |
873 | break; | |
874 | ||
875 | case ARG_DUMP_CONFIGURATION_ITEMS: | |
876 | arg_action = ACTION_DUMP_CONFIGURATION_ITEMS; | |
877 | break; | |
878 | ||
879 | case ARG_DUMP_CORE: | |
880 | if (!optarg) | |
881 | arg_dump_core = true; | |
882 | else { | |
883 | r = parse_boolean(optarg); | |
884 | if (r < 0) | |
885 | return log_error_errno(r, "Failed to parse dump core boolean: %s", optarg); | |
886 | arg_dump_core = r; | |
887 | } | |
888 | break; | |
889 | ||
890 | case ARG_CRASH_CHVT: | |
891 | r = parse_crash_chvt(optarg); | |
892 | if (r < 0) | |
893 | return log_error_errno(r, "Failed to parse crash virtual terminal index: %s", optarg); | |
894 | break; | |
895 | ||
896 | case ARG_CRASH_SHELL: | |
897 | if (!optarg) | |
898 | arg_crash_shell = true; | |
899 | else { | |
900 | r = parse_boolean(optarg); | |
901 | if (r < 0) | |
902 | return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg); | |
903 | arg_crash_shell = r; | |
904 | } | |
905 | break; | |
906 | ||
907 | case ARG_CRASH_REBOOT: | |
908 | if (!optarg) | |
909 | arg_crash_reboot = true; | |
910 | else { | |
911 | r = parse_boolean(optarg); | |
912 | if (r < 0) | |
913 | return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg); | |
914 | arg_crash_reboot = r; | |
915 | } | |
916 | break; | |
917 | ||
918 | case ARG_CONFIRM_SPAWN: | |
919 | r = optarg ? parse_boolean(optarg) : 1; | |
920 | if (r < 0) { | |
921 | log_error("Failed to parse confirm spawn boolean %s.", optarg); | |
922 | return r; | |
923 | } | |
924 | arg_confirm_spawn = r; | |
925 | break; | |
926 | ||
927 | case ARG_SHOW_STATUS: | |
928 | if (optarg) { | |
929 | r = parse_show_status(optarg, &arg_show_status); | |
930 | if (r < 0) { | |
931 | log_error("Failed to parse show status boolean %s.", optarg); | |
932 | return r; | |
933 | } | |
934 | } else | |
935 | arg_show_status = SHOW_STATUS_YES; | |
936 | break; | |
937 | ||
938 | case ARG_DESERIALIZE: { | |
939 | int fd; | |
940 | FILE *f; | |
941 | ||
942 | r = safe_atoi(optarg, &fd); | |
943 | if (r < 0 || fd < 0) { | |
944 | log_error("Failed to parse deserialize option %s.", optarg); | |
945 | return -EINVAL; | |
946 | } | |
947 | ||
948 | (void) fd_cloexec(fd, true); | |
949 | ||
950 | f = fdopen(fd, "r"); | |
951 | if (!f) | |
952 | return log_error_errno(errno, "Failed to open serialization fd: %m"); | |
953 | ||
954 | safe_fclose(arg_serialization); | |
955 | arg_serialization = f; | |
956 | ||
957 | break; | |
958 | } | |
959 | ||
960 | case ARG_SWITCHED_ROOT: | |
961 | arg_switched_root = true; | |
962 | break; | |
963 | ||
964 | case 'h': | |
965 | arg_action = ACTION_HELP; | |
966 | if (arg_no_pager < 0) | |
967 | arg_no_pager = true; | |
968 | break; | |
969 | ||
970 | case 'D': | |
971 | log_set_max_level(LOG_DEBUG); | |
972 | break; | |
973 | ||
974 | case 'b': | |
975 | case 's': | |
976 | case 'z': | |
977 | /* Just to eat away the sysvinit kernel | |
978 | * cmdline args without getopt() error | |
979 | * messages that we'll parse in | |
980 | * parse_proc_cmdline_word() or ignore. */ | |
981 | ||
982 | case '?': | |
983 | if (getpid() != 1) | |
984 | return -EINVAL; | |
985 | else | |
986 | return 0; | |
987 | ||
988 | default: | |
989 | assert_not_reached("Unhandled option code."); | |
990 | } | |
991 | ||
992 | if (optind < argc && getpid() != 1) { | |
993 | /* Hmm, when we aren't run as init system | |
994 | * let's complain about excess arguments */ | |
995 | ||
996 | log_error("Excess arguments."); | |
997 | return -EINVAL; | |
998 | } | |
999 | ||
1000 | return 0; | |
1001 | } | |
1002 | ||
1003 | static int help(void) { | |
1004 | ||
1005 | printf("%s [OPTIONS...]\n\n" | |
1006 | "Starts up and maintains the system or user services.\n\n" | |
1007 | " -h --help Show this help\n" | |
1008 | " --test Determine startup sequence, dump it and exit\n" | |
1009 | " --no-pager Do not pipe output into a pager\n" | |
1010 | " --dump-configuration-items Dump understood unit configuration items\n" | |
1011 | " --unit=UNIT Set default unit\n" | |
1012 | " --system Run a system instance, even if PID != 1\n" | |
1013 | " --user Run a user instance\n" | |
1014 | " --dump-core[=BOOL] Dump core on crash\n" | |
1015 | " --crash-vt=NR Change to specified VT on crash\n" | |
1016 | " --crash-reboot[=BOOL] Reboot on crash\n" | |
1017 | " --crash-shell[=BOOL] Run shell on crash\n" | |
1018 | " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n" | |
1019 | " --show-status[=BOOL] Show status updates on the console during bootup\n" | |
1020 | " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n" | |
1021 | " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n" | |
1022 | " --log-color[=BOOL] Highlight important log messages\n" | |
1023 | " --log-location[=BOOL] Include code location in log messages\n" | |
1024 | " --default-standard-output= Set default standard output for services\n" | |
1025 | " --default-standard-error= Set default standard error output for services\n", | |
1026 | program_invocation_short_name); | |
1027 | ||
1028 | return 0; | |
1029 | } | |
1030 | ||
1031 | static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) { | |
1032 | _cleanup_fdset_free_ FDSet *fds = NULL; | |
1033 | _cleanup_fclose_ FILE *f = NULL; | |
1034 | int r; | |
1035 | ||
1036 | assert(m); | |
1037 | assert(_f); | |
1038 | assert(_fds); | |
1039 | ||
1040 | r = manager_open_serialization(m, &f); | |
1041 | if (r < 0) | |
1042 | return log_error_errno(r, "Failed to create serialization file: %m"); | |
1043 | ||
1044 | /* Make sure nothing is really destructed when we shut down */ | |
1045 | m->n_reloading ++; | |
1046 | bus_manager_send_reloading(m, true); | |
1047 | ||
1048 | fds = fdset_new(); | |
1049 | if (!fds) | |
1050 | return log_oom(); | |
1051 | ||
1052 | r = manager_serialize(m, f, fds, switching_root); | |
1053 | if (r < 0) | |
1054 | return log_error_errno(r, "Failed to serialize state: %m"); | |
1055 | ||
1056 | if (fseeko(f, 0, SEEK_SET) == (off_t) -1) | |
1057 | return log_error_errno(errno, "Failed to rewind serialization fd: %m"); | |
1058 | ||
1059 | r = fd_cloexec(fileno(f), false); | |
1060 | if (r < 0) | |
1061 | return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m"); | |
1062 | ||
1063 | r = fdset_cloexec(fds, false); | |
1064 | if (r < 0) | |
1065 | return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m"); | |
1066 | ||
1067 | *_f = f; | |
1068 | *_fds = fds; | |
1069 | ||
1070 | f = NULL; | |
1071 | fds = NULL; | |
1072 | ||
1073 | return 0; | |
1074 | } | |
1075 | ||
1076 | static int bump_rlimit_nofile(struct rlimit *saved_rlimit) { | |
1077 | struct rlimit nl; | |
1078 | int r; | |
1079 | ||
1080 | assert(saved_rlimit); | |
1081 | ||
1082 | /* Save the original RLIMIT_NOFILE so that we can reset it | |
1083 | * later when transitioning from the initrd to the main | |
1084 | * systemd or suchlike. */ | |
1085 | if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0) | |
1086 | return log_error_errno(errno, "Reading RLIMIT_NOFILE failed: %m"); | |
1087 | ||
1088 | /* Make sure forked processes get the default kernel setting */ | |
1089 | if (!arg_default_rlimit[RLIMIT_NOFILE]) { | |
1090 | struct rlimit *rl; | |
1091 | ||
1092 | rl = newdup(struct rlimit, saved_rlimit, 1); | |
1093 | if (!rl) | |
1094 | return log_oom(); | |
1095 | ||
1096 | arg_default_rlimit[RLIMIT_NOFILE] = rl; | |
1097 | } | |
1098 | ||
1099 | /* Bump up the resource limit for ourselves substantially */ | |
1100 | nl.rlim_cur = nl.rlim_max = 64*1024; | |
1101 | r = setrlimit_closest(RLIMIT_NOFILE, &nl); | |
1102 | if (r < 0) | |
1103 | return log_error_errno(r, "Setting RLIMIT_NOFILE failed: %m"); | |
1104 | ||
1105 | return 0; | |
1106 | } | |
1107 | ||
1108 | static void test_mtab(void) { | |
1109 | ||
1110 | static const char ok[] = | |
1111 | "/proc/self/mounts\0" | |
1112 | "/proc/mounts\0" | |
1113 | "../proc/self/mounts\0" | |
1114 | "../proc/mounts\0"; | |
1115 | ||
1116 | _cleanup_free_ char *p = NULL; | |
1117 | int r; | |
1118 | ||
1119 | /* Check that /etc/mtab is a symlink to the right place or | |
1120 | * non-existing. But certainly not a file, or a symlink to | |
1121 | * some weird place... */ | |
1122 | ||
1123 | r = readlink_malloc("/etc/mtab", &p); | |
1124 | if (r == -ENOENT) | |
1125 | return; | |
1126 | if (r >= 0 && nulstr_contains(ok, p)) | |
1127 | return; | |
1128 | ||
1129 | log_error("/etc/mtab is not a symlink or not pointing to /proc/self/mounts. " | |
1130 | "This is not supported anymore. " | |
1131 | "Please replace /etc/mtab with a symlink to /proc/self/mounts."); | |
1132 | freeze_or_reboot(); | |
1133 | } | |
1134 | ||
1135 | static void test_usr(void) { | |
1136 | ||
1137 | /* Check that /usr is not a separate fs */ | |
1138 | ||
1139 | if (dir_is_empty("/usr") <= 0) | |
1140 | return; | |
1141 | ||
1142 | log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. " | |
1143 | "Some things will probably break (sometimes even silently) in mysterious ways. " | |
1144 | "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information."); | |
1145 | } | |
1146 | ||
1147 | static int initialize_join_controllers(void) { | |
1148 | /* By default, mount "cpu" + "cpuacct" together, and "net_cls" | |
1149 | * + "net_prio". We'd like to add "cpuset" to the mix, but | |
1150 | * "cpuset" doesn't really work for groups with no initialized | |
1151 | * attributes. */ | |
1152 | ||
1153 | arg_join_controllers = new(char**, 3); | |
1154 | if (!arg_join_controllers) | |
1155 | return -ENOMEM; | |
1156 | ||
1157 | arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL); | |
1158 | if (!arg_join_controllers[0]) | |
1159 | goto oom; | |
1160 | ||
1161 | arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL); | |
1162 | if (!arg_join_controllers[1]) | |
1163 | goto oom; | |
1164 | ||
1165 | arg_join_controllers[2] = NULL; | |
1166 | return 0; | |
1167 | ||
1168 | oom: | |
1169 | arg_join_controllers = strv_free_free(arg_join_controllers); | |
1170 | return -ENOMEM; | |
1171 | } | |
1172 | ||
1173 | static int enforce_syscall_archs(Set *archs) { | |
1174 | #ifdef HAVE_SECCOMP | |
1175 | scmp_filter_ctx *seccomp; | |
1176 | Iterator i; | |
1177 | void *id; | |
1178 | int r; | |
1179 | ||
1180 | seccomp = seccomp_init(SCMP_ACT_ALLOW); | |
1181 | if (!seccomp) | |
1182 | return log_oom(); | |
1183 | ||
1184 | SET_FOREACH(id, arg_syscall_archs, i) { | |
1185 | r = seccomp_arch_add(seccomp, PTR_TO_UINT32(id) - 1); | |
1186 | if (r == -EEXIST) | |
1187 | continue; | |
1188 | if (r < 0) { | |
1189 | log_error_errno(r, "Failed to add architecture to seccomp: %m"); | |
1190 | goto finish; | |
1191 | } | |
1192 | } | |
1193 | ||
1194 | r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0); | |
1195 | if (r < 0) { | |
1196 | log_error_errno(r, "Failed to unset NO_NEW_PRIVS: %m"); | |
1197 | goto finish; | |
1198 | } | |
1199 | ||
1200 | r = seccomp_load(seccomp); | |
1201 | if (r < 0) | |
1202 | log_error_errno(r, "Failed to add install architecture seccomp: %m"); | |
1203 | ||
1204 | finish: | |
1205 | seccomp_release(seccomp); | |
1206 | return r; | |
1207 | #else | |
1208 | return 0; | |
1209 | #endif | |
1210 | } | |
1211 | ||
1212 | static int status_welcome(void) { | |
1213 | _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL; | |
1214 | int r; | |
1215 | ||
1216 | r = parse_env_file("/etc/os-release", NEWLINE, | |
1217 | "PRETTY_NAME", &pretty_name, | |
1218 | "ANSI_COLOR", &ansi_color, | |
1219 | NULL); | |
1220 | if (r == -ENOENT) | |
1221 | r = parse_env_file("/usr/lib/os-release", NEWLINE, | |
1222 | "PRETTY_NAME", &pretty_name, | |
1223 | "ANSI_COLOR", &ansi_color, | |
1224 | NULL); | |
1225 | ||
1226 | if (r < 0 && r != -ENOENT) | |
1227 | log_warning_errno(r, "Failed to read os-release file: %m"); | |
1228 | ||
1229 | return status_printf(NULL, false, false, | |
1230 | "\nWelcome to \x1B[%sm%s\x1B[0m!\n", | |
1231 | isempty(ansi_color) ? "1" : ansi_color, | |
1232 | isempty(pretty_name) ? "Linux" : pretty_name); | |
1233 | } | |
1234 | ||
1235 | static int write_container_id(void) { | |
1236 | const char *c; | |
1237 | ||
1238 | c = getenv("container"); | |
1239 | if (isempty(c)) | |
1240 | return 0; | |
1241 | ||
1242 | return write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE); | |
1243 | } | |
1244 | ||
1245 | int main(int argc, char *argv[]) { | |
1246 | Manager *m = NULL; | |
1247 | int r, retval = EXIT_FAILURE; | |
1248 | usec_t before_startup, after_startup; | |
1249 | char timespan[FORMAT_TIMESPAN_MAX]; | |
1250 | FDSet *fds = NULL; | |
1251 | bool reexecute = false; | |
1252 | const char *shutdown_verb = NULL; | |
1253 | dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL; | |
1254 | dual_timestamp userspace_timestamp = DUAL_TIMESTAMP_NULL; | |
1255 | dual_timestamp kernel_timestamp = DUAL_TIMESTAMP_NULL; | |
1256 | dual_timestamp security_start_timestamp = DUAL_TIMESTAMP_NULL; | |
1257 | dual_timestamp security_finish_timestamp = DUAL_TIMESTAMP_NULL; | |
1258 | static char systemd[] = "systemd"; | |
1259 | bool skip_setup = false; | |
1260 | unsigned j; | |
1261 | bool loaded_policy = false; | |
1262 | bool arm_reboot_watchdog = false; | |
1263 | bool queue_default_job = false; | |
1264 | bool empty_etc = false; | |
1265 | char *switch_root_dir = NULL, *switch_root_init = NULL; | |
1266 | struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0); | |
1267 | const char *error_message = NULL; | |
1268 | ||
1269 | #ifdef HAVE_SYSV_COMPAT | |
1270 | if (getpid() != 1 && strstr(program_invocation_short_name, "init")) { | |
1271 | /* This is compatibility support for SysV, where | |
1272 | * calling init as a user is identical to telinit. */ | |
1273 | ||
1274 | errno = -ENOENT; | |
1275 | execv(SYSTEMCTL_BINARY_PATH, argv); | |
1276 | log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m"); | |
1277 | return 1; | |
1278 | } | |
1279 | #endif | |
1280 | ||
1281 | dual_timestamp_from_monotonic(&kernel_timestamp, 0); | |
1282 | dual_timestamp_get(&userspace_timestamp); | |
1283 | ||
1284 | /* Determine if this is a reexecution or normal bootup. We do | |
1285 | * the full command line parsing much later, so let's just | |
1286 | * have a quick peek here. */ | |
1287 | if (strv_find(argv+1, "--deserialize")) | |
1288 | skip_setup = true; | |
1289 | ||
1290 | /* If we have switched root, do all the special setup | |
1291 | * things */ | |
1292 | if (strv_find(argv+1, "--switched-root")) | |
1293 | skip_setup = false; | |
1294 | ||
1295 | /* If we get started via the /sbin/init symlink then we are | |
1296 | called 'init'. After a subsequent reexecution we are then | |
1297 | called 'systemd'. That is confusing, hence let's call us | |
1298 | systemd right-away. */ | |
1299 | program_invocation_short_name = systemd; | |
1300 | prctl(PR_SET_NAME, systemd); | |
1301 | ||
1302 | saved_argv = argv; | |
1303 | saved_argc = argc; | |
1304 | ||
1305 | log_show_color(isatty(STDERR_FILENO) > 0); | |
1306 | log_set_upgrade_syslog_to_journal(true); | |
1307 | ||
1308 | /* Disable the umask logic */ | |
1309 | if (getpid() == 1) | |
1310 | umask(0); | |
1311 | ||
1312 | if (getpid() == 1 && detect_container() <= 0) { | |
1313 | ||
1314 | /* Running outside of a container as PID 1 */ | |
1315 | arg_running_as = MANAGER_SYSTEM; | |
1316 | make_null_stdio(); | |
1317 | log_set_target(LOG_TARGET_KMSG); | |
1318 | log_open(); | |
1319 | ||
1320 | if (in_initrd()) | |
1321 | initrd_timestamp = userspace_timestamp; | |
1322 | ||
1323 | if (!skip_setup) { | |
1324 | mount_setup_early(); | |
1325 | dual_timestamp_get(&security_start_timestamp); | |
1326 | if (mac_selinux_setup(&loaded_policy) < 0) { | |
1327 | error_message = "Failed to load SELinux policy"; | |
1328 | goto finish; | |
1329 | } else if (ima_setup() < 0) { | |
1330 | error_message = "Failed to load IMA policy"; | |
1331 | goto finish; | |
1332 | } else if (mac_smack_setup(&loaded_policy) < 0) { | |
1333 | error_message = "Failed to load SMACK policy"; | |
1334 | goto finish; | |
1335 | } | |
1336 | dual_timestamp_get(&security_finish_timestamp); | |
1337 | } | |
1338 | ||
1339 | if (mac_selinux_init(NULL) < 0) { | |
1340 | error_message = "Failed to initialize SELinux policy"; | |
1341 | goto finish; | |
1342 | } | |
1343 | ||
1344 | if (!skip_setup) { | |
1345 | if (clock_is_localtime() > 0) { | |
1346 | int min; | |
1347 | ||
1348 | /* | |
1349 | * The very first call of settimeofday() also does a time warp in the kernel. | |
1350 | * | |
1351 | * In the rtc-in-local time mode, we set the kernel's timezone, and rely on | |
1352 | * external tools to take care of maintaining the RTC and do all adjustments. | |
1353 | * This matches the behavior of Windows, which leaves the RTC alone if the | |
1354 | * registry tells that the RTC runs in UTC. | |
1355 | */ | |
1356 | r = clock_set_timezone(&min); | |
1357 | if (r < 0) | |
1358 | log_error_errno(r, "Failed to apply local time delta, ignoring: %m"); | |
1359 | else | |
1360 | log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min); | |
1361 | } else if (!in_initrd()) { | |
1362 | /* | |
1363 | * Do a dummy very first call to seal the kernel's time warp magic. | |
1364 | * | |
1365 | * Do not call this this from inside the initrd. The initrd might not | |
1366 | * carry /etc/adjtime with LOCAL, but the real system could be set up | |
1367 | * that way. In such case, we need to delay the time-warp or the sealing | |
1368 | * until we reach the real system. | |
1369 | * | |
1370 | * Do no set the kernel's timezone. The concept of local time cannot | |
1371 | * be supported reliably, the time will jump or be incorrect at every daylight | |
1372 | * saving time change. All kernel local time concepts will be treated | |
1373 | * as UTC that way. | |
1374 | */ | |
1375 | clock_reset_timewarp(); | |
1376 | } | |
1377 | } | |
1378 | ||
1379 | /* Set the default for later on, but don't actually | |
1380 | * open the logs like this for now. Note that if we | |
1381 | * are transitioning from the initrd there might still | |
1382 | * be journal fd open, and we shouldn't attempt | |
1383 | * opening that before we parsed /proc/cmdline which | |
1384 | * might redirect output elsewhere. */ | |
1385 | log_set_target(LOG_TARGET_JOURNAL_OR_KMSG); | |
1386 | ||
1387 | } else if (getpid() == 1) { | |
1388 | /* Running inside a container, as PID 1 */ | |
1389 | arg_running_as = MANAGER_SYSTEM; | |
1390 | log_set_target(LOG_TARGET_CONSOLE); | |
1391 | log_close_console(); /* force reopen of /dev/console */ | |
1392 | log_open(); | |
1393 | ||
1394 | /* For the later on, see above... */ | |
1395 | log_set_target(LOG_TARGET_JOURNAL); | |
1396 | ||
1397 | /* clear the kernel timestamp, | |
1398 | * because we are in a container */ | |
1399 | kernel_timestamp.monotonic = 0ULL; | |
1400 | kernel_timestamp.realtime = 0ULL; | |
1401 | ||
1402 | } else { | |
1403 | /* Running as user instance */ | |
1404 | arg_running_as = MANAGER_USER; | |
1405 | log_set_target(LOG_TARGET_AUTO); | |
1406 | log_open(); | |
1407 | ||
1408 | /* clear the kernel timestamp, | |
1409 | * because we are not PID 1 */ | |
1410 | kernel_timestamp = DUAL_TIMESTAMP_NULL; | |
1411 | } | |
1412 | ||
1413 | /* Initialize default unit */ | |
1414 | r = free_and_strdup(&arg_default_unit, SPECIAL_DEFAULT_TARGET); | |
1415 | if (r < 0) { | |
1416 | log_emergency_errno(r, "Failed to set default unit %s: %m", SPECIAL_DEFAULT_TARGET); | |
1417 | error_message = "Failed to set default unit"; | |
1418 | goto finish; | |
1419 | } | |
1420 | ||
1421 | r = initialize_join_controllers(); | |
1422 | if (r < 0) { | |
1423 | error_message = "Failed to initialize cgroup controllers"; | |
1424 | goto finish; | |
1425 | } | |
1426 | ||
1427 | /* Mount /proc, /sys and friends, so that /proc/cmdline and | |
1428 | * /proc/$PID/fd is available. */ | |
1429 | if (getpid() == 1) { | |
1430 | ||
1431 | /* Load the kernel modules early, so that we kdbus.ko is loaded before kdbusfs shall be mounted */ | |
1432 | if (!skip_setup) | |
1433 | kmod_setup(); | |
1434 | ||
1435 | r = mount_setup(loaded_policy); | |
1436 | if (r < 0) { | |
1437 | error_message = "Failed to mount API filesystems"; | |
1438 | goto finish; | |
1439 | } | |
1440 | } | |
1441 | ||
1442 | /* Reset all signal handlers. */ | |
1443 | (void) reset_all_signal_handlers(); | |
1444 | (void) ignore_signals(SIGNALS_IGNORE, -1); | |
1445 | ||
1446 | if (parse_config_file() < 0) { | |
1447 | error_message = "Failed to parse config file"; | |
1448 | goto finish; | |
1449 | } | |
1450 | ||
1451 | if (arg_running_as == MANAGER_SYSTEM) { | |
1452 | r = parse_proc_cmdline(parse_proc_cmdline_item); | |
1453 | if (r < 0) | |
1454 | log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m"); | |
1455 | } | |
1456 | ||
1457 | /* Note that this also parses bits from the kernel command | |
1458 | * line, including "debug". */ | |
1459 | log_parse_environment(); | |
1460 | ||
1461 | if (parse_argv(argc, argv) < 0) { | |
1462 | error_message = "Failed to parse commandline arguments"; | |
1463 | goto finish; | |
1464 | } | |
1465 | ||
1466 | if (arg_action == ACTION_TEST && | |
1467 | geteuid() == 0) { | |
1468 | log_error("Don't run test mode as root."); | |
1469 | goto finish; | |
1470 | } | |
1471 | ||
1472 | if (arg_running_as == MANAGER_USER && | |
1473 | arg_action == ACTION_RUN && | |
1474 | sd_booted() <= 0) { | |
1475 | log_error("Trying to run as user instance, but the system has not been booted with systemd."); | |
1476 | goto finish; | |
1477 | } | |
1478 | ||
1479 | if (arg_running_as == MANAGER_SYSTEM && | |
1480 | arg_action == ACTION_RUN && | |
1481 | running_in_chroot() > 0) { | |
1482 | log_error("Cannot be run in a chroot() environment."); | |
1483 | goto finish; | |
1484 | } | |
1485 | ||
1486 | if (arg_action == ACTION_TEST) | |
1487 | skip_setup = true; | |
1488 | ||
1489 | pager_open_if_enabled(); | |
1490 | ||
1491 | if (arg_action == ACTION_HELP) { | |
1492 | retval = help(); | |
1493 | goto finish; | |
1494 | } else if (arg_action == ACTION_VERSION) { | |
1495 | retval = version(); | |
1496 | goto finish; | |
1497 | } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) { | |
1498 | unit_dump_config_items(stdout); | |
1499 | retval = EXIT_SUCCESS; | |
1500 | goto finish; | |
1501 | } else if (arg_action == ACTION_DONE) { | |
1502 | retval = EXIT_SUCCESS; | |
1503 | goto finish; | |
1504 | } | |
1505 | ||
1506 | if (arg_running_as == MANAGER_USER && | |
1507 | !getenv("XDG_RUNTIME_DIR")) { | |
1508 | log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set."); | |
1509 | goto finish; | |
1510 | } | |
1511 | ||
1512 | assert_se(arg_action == ACTION_RUN || arg_action == ACTION_TEST); | |
1513 | ||
1514 | /* Close logging fds, in order not to confuse fdset below */ | |
1515 | log_close(); | |
1516 | ||
1517 | /* Remember open file descriptors for later deserialization */ | |
1518 | r = fdset_new_fill(&fds); | |
1519 | if (r < 0) { | |
1520 | log_emergency_errno(r, "Failed to allocate fd set: %m"); | |
1521 | error_message = "Failed to allocate fd set"; | |
1522 | goto finish; | |
1523 | } else | |
1524 | fdset_cloexec(fds, true); | |
1525 | ||
1526 | if (arg_serialization) | |
1527 | assert_se(fdset_remove(fds, fileno(arg_serialization)) >= 0); | |
1528 | ||
1529 | if (arg_running_as == MANAGER_SYSTEM) | |
1530 | /* Become a session leader if we aren't one yet. */ | |
1531 | setsid(); | |
1532 | ||
1533 | /* Move out of the way, so that we won't block unmounts */ | |
1534 | assert_se(chdir("/") == 0); | |
1535 | ||
1536 | /* Reset the console, but only if this is really init and we | |
1537 | * are freshly booted */ | |
1538 | if (arg_running_as == MANAGER_SYSTEM && arg_action == ACTION_RUN) { | |
1539 | ||
1540 | /* If we are init, we connect stdin/stdout/stderr to | |
1541 | * /dev/null and make sure we don't have a controlling | |
1542 | * tty. */ | |
1543 | release_terminal(); | |
1544 | ||
1545 | if (getpid() == 1 && !skip_setup) | |
1546 | console_setup(); | |
1547 | } | |
1548 | ||
1549 | /* Open the logging devices, if possible and necessary */ | |
1550 | log_open(); | |
1551 | ||
1552 | if (arg_show_status == _SHOW_STATUS_UNSET) | |
1553 | arg_show_status = SHOW_STATUS_YES; | |
1554 | ||
1555 | /* Make sure we leave a core dump without panicing the | |
1556 | * kernel. */ | |
1557 | if (getpid() == 1) { | |
1558 | install_crash_handler(); | |
1559 | ||
1560 | r = mount_cgroup_controllers(arg_join_controllers); | |
1561 | if (r < 0) | |
1562 | goto finish; | |
1563 | } | |
1564 | ||
1565 | if (arg_running_as == MANAGER_SYSTEM) { | |
1566 | int v; | |
1567 | ||
1568 | log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")", | |
1569 | arg_action == ACTION_TEST ? "test " : "" ); | |
1570 | ||
1571 | v = detect_virtualization(); | |
1572 | if (v > 0) | |
1573 | log_info("Detected virtualization %s.", virtualization_to_string(v)); | |
1574 | ||
1575 | write_container_id(); | |
1576 | ||
1577 | log_info("Detected architecture %s.", architecture_to_string(uname_architecture())); | |
1578 | ||
1579 | if (in_initrd()) | |
1580 | log_info("Running in initial RAM disk."); | |
1581 | ||
1582 | /* Let's check whether /etc is already populated. We | |
1583 | * don't actually really check for that, but use | |
1584 | * /etc/machine-id as flag file. This allows container | |
1585 | * managers and installers to provision a couple of | |
1586 | * files already. If the container manager wants to | |
1587 | * provision the machine ID itself it should pass | |
1588 | * $container_uuid to PID 1. */ | |
1589 | ||
1590 | empty_etc = access("/etc/machine-id", F_OK) < 0; | |
1591 | if (empty_etc) | |
1592 | log_info("Running with unpopulated /etc."); | |
1593 | } else { | |
1594 | _cleanup_free_ char *t; | |
1595 | ||
1596 | t = uid_to_name(getuid()); | |
1597 | log_debug(PACKAGE_STRING " running in %suser mode for user "UID_FMT"/%s. (" SYSTEMD_FEATURES ")", | |
1598 | arg_action == ACTION_TEST ? " test" : "", getuid(), t); | |
1599 | } | |
1600 | ||
1601 | if (arg_running_as == MANAGER_SYSTEM && !skip_setup) { | |
1602 | if (arg_show_status > 0) | |
1603 | status_welcome(); | |
1604 | ||
1605 | hostname_setup(); | |
1606 | machine_id_setup(NULL); | |
1607 | loopback_setup(); | |
1608 | ||
1609 | test_mtab(); | |
1610 | test_usr(); | |
1611 | } | |
1612 | ||
1613 | if (arg_running_as == MANAGER_SYSTEM && arg_runtime_watchdog > 0) | |
1614 | watchdog_set_timeout(&arg_runtime_watchdog); | |
1615 | ||
1616 | if (arg_timer_slack_nsec != NSEC_INFINITY) | |
1617 | if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0) | |
1618 | log_error_errno(errno, "Failed to adjust timer slack: %m"); | |
1619 | ||
1620 | if (arg_capability_bounding_set_drop) { | |
1621 | r = capability_bounding_set_drop_usermode(arg_capability_bounding_set_drop); | |
1622 | if (r < 0) { | |
1623 | log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m"); | |
1624 | error_message = "Failed to drop capability bounding set of usermode helpers"; | |
1625 | goto finish; | |
1626 | } | |
1627 | r = capability_bounding_set_drop(arg_capability_bounding_set_drop, true); | |
1628 | if (r < 0) { | |
1629 | log_emergency_errno(r, "Failed to drop capability bounding set: %m"); | |
1630 | error_message = "Failed to drop capability bounding set"; | |
1631 | goto finish; | |
1632 | } | |
1633 | } | |
1634 | ||
1635 | if (arg_syscall_archs) { | |
1636 | r = enforce_syscall_archs(arg_syscall_archs); | |
1637 | if (r < 0) { | |
1638 | error_message = "Failed to set syscall architectures"; | |
1639 | goto finish; | |
1640 | } | |
1641 | } | |
1642 | ||
1643 | if (arg_running_as == MANAGER_USER) | |
1644 | /* Become reaper of our children */ | |
1645 | if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0) | |
1646 | log_warning_errno(errno, "Failed to make us a subreaper: %m"); | |
1647 | ||
1648 | if (arg_running_as == MANAGER_SYSTEM) { | |
1649 | bump_rlimit_nofile(&saved_rlimit_nofile); | |
1650 | ||
1651 | if (empty_etc) { | |
1652 | r = unit_file_preset_all(UNIT_FILE_SYSTEM, false, NULL, UNIT_FILE_PRESET_ENABLE_ONLY, false, NULL, 0); | |
1653 | if (r < 0) | |
1654 | log_warning_errno(r, "Failed to populate /etc with preset unit settings, ignoring: %m"); | |
1655 | else | |
1656 | log_info("Populated /etc with preset unit settings."); | |
1657 | } | |
1658 | } | |
1659 | ||
1660 | r = manager_new(arg_running_as, arg_action == ACTION_TEST, &m); | |
1661 | if (r < 0) { | |
1662 | log_emergency_errno(r, "Failed to allocate manager object: %m"); | |
1663 | error_message = "Failed to allocate manager object"; | |
1664 | goto finish; | |
1665 | } | |
1666 | ||
1667 | m->confirm_spawn = arg_confirm_spawn; | |
1668 | m->runtime_watchdog = arg_runtime_watchdog; | |
1669 | m->shutdown_watchdog = arg_shutdown_watchdog; | |
1670 | m->userspace_timestamp = userspace_timestamp; | |
1671 | m->kernel_timestamp = kernel_timestamp; | |
1672 | m->initrd_timestamp = initrd_timestamp; | |
1673 | m->security_start_timestamp = security_start_timestamp; | |
1674 | m->security_finish_timestamp = security_finish_timestamp; | |
1675 | ||
1676 | manager_set_defaults(m); | |
1677 | manager_set_show_status(m, arg_show_status); | |
1678 | manager_set_first_boot(m, empty_etc); | |
1679 | ||
1680 | /* Remember whether we should queue the default job */ | |
1681 | queue_default_job = !arg_serialization || arg_switched_root; | |
1682 | ||
1683 | before_startup = now(CLOCK_MONOTONIC); | |
1684 | ||
1685 | r = manager_startup(m, arg_serialization, fds); | |
1686 | if (r < 0) | |
1687 | log_error_errno(r, "Failed to fully start up daemon: %m"); | |
1688 | ||
1689 | /* This will close all file descriptors that were opened, but | |
1690 | * not claimed by any unit. */ | |
1691 | fds = fdset_free(fds); | |
1692 | ||
1693 | arg_serialization = safe_fclose(arg_serialization); | |
1694 | ||
1695 | if (queue_default_job) { | |
1696 | _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; | |
1697 | Unit *target = NULL; | |
1698 | Job *default_unit_job; | |
1699 | ||
1700 | log_debug("Activating default unit: %s", arg_default_unit); | |
1701 | ||
1702 | r = manager_load_unit(m, arg_default_unit, NULL, &error, &target); | |
1703 | if (r < 0) | |
1704 | log_error("Failed to load default target: %s", bus_error_message(&error, r)); | |
1705 | else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND) | |
1706 | log_error_errno(target->load_error, "Failed to load default target: %m"); | |
1707 | else if (target->load_state == UNIT_MASKED) | |
1708 | log_error("Default target masked."); | |
1709 | ||
1710 | if (!target || target->load_state != UNIT_LOADED) { | |
1711 | log_info("Trying to load rescue target..."); | |
1712 | ||
1713 | r = manager_load_unit(m, SPECIAL_RESCUE_TARGET, NULL, &error, &target); | |
1714 | if (r < 0) { | |
1715 | log_emergency("Failed to load rescue target: %s", bus_error_message(&error, r)); | |
1716 | error_message = "Failed to load rescue target"; | |
1717 | goto finish; | |
1718 | } else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND) { | |
1719 | log_emergency_errno(target->load_error, "Failed to load rescue target: %m"); | |
1720 | error_message = "Failed to load rescue target"; | |
1721 | goto finish; | |
1722 | } else if (target->load_state == UNIT_MASKED) { | |
1723 | log_emergency("Rescue target masked."); | |
1724 | error_message = "Rescue target masked"; | |
1725 | goto finish; | |
1726 | } | |
1727 | } | |
1728 | ||
1729 | assert(target->load_state == UNIT_LOADED); | |
1730 | ||
1731 | if (arg_action == ACTION_TEST) { | |
1732 | printf("-> By units:\n"); | |
1733 | manager_dump_units(m, stdout, "\t"); | |
1734 | } | |
1735 | ||
1736 | r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, false, &error, &default_unit_job); | |
1737 | if (r == -EPERM) { | |
1738 | log_debug("Default target could not be isolated, starting instead: %s", bus_error_message(&error, r)); | |
1739 | ||
1740 | r = manager_add_job(m, JOB_START, target, JOB_REPLACE, false, &error, &default_unit_job); | |
1741 | if (r < 0) { | |
1742 | log_emergency("Failed to start default target: %s", bus_error_message(&error, r)); | |
1743 | error_message = "Failed to start default target"; | |
1744 | goto finish; | |
1745 | } | |
1746 | } else if (r < 0) { | |
1747 | log_emergency("Failed to isolate default target: %s", bus_error_message(&error, r)); | |
1748 | error_message = "Failed to isolate default target"; | |
1749 | goto finish; | |
1750 | } | |
1751 | ||
1752 | m->default_unit_job_id = default_unit_job->id; | |
1753 | ||
1754 | after_startup = now(CLOCK_MONOTONIC); | |
1755 | log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG, | |
1756 | "Loaded units and determined initial transaction in %s.", | |
1757 | format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC)); | |
1758 | ||
1759 | if (arg_action == ACTION_TEST) { | |
1760 | printf("-> By jobs:\n"); | |
1761 | manager_dump_jobs(m, stdout, "\t"); | |
1762 | retval = EXIT_SUCCESS; | |
1763 | goto finish; | |
1764 | } | |
1765 | } | |
1766 | ||
1767 | for (;;) { | |
1768 | r = manager_loop(m); | |
1769 | if (r < 0) { | |
1770 | log_emergency_errno(r, "Failed to run main loop: %m"); | |
1771 | error_message = "Failed to run main loop"; | |
1772 | goto finish; | |
1773 | } | |
1774 | ||
1775 | switch (m->exit_code) { | |
1776 | ||
1777 | case MANAGER_RELOAD: | |
1778 | log_info("Reloading."); | |
1779 | ||
1780 | r = parse_config_file(); | |
1781 | if (r < 0) | |
1782 | log_error("Failed to parse config file."); | |
1783 | ||
1784 | manager_set_defaults(m); | |
1785 | ||
1786 | r = manager_reload(m); | |
1787 | if (r < 0) | |
1788 | log_error_errno(r, "Failed to reload: %m"); | |
1789 | break; | |
1790 | ||
1791 | case MANAGER_REEXECUTE: | |
1792 | ||
1793 | if (prepare_reexecute(m, &arg_serialization, &fds, false) < 0) { | |
1794 | error_message = "Failed to prepare for reexecution"; | |
1795 | goto finish; | |
1796 | } | |
1797 | ||
1798 | reexecute = true; | |
1799 | log_notice("Reexecuting."); | |
1800 | goto finish; | |
1801 | ||
1802 | case MANAGER_SWITCH_ROOT: | |
1803 | /* Steal the switch root parameters */ | |
1804 | switch_root_dir = m->switch_root; | |
1805 | switch_root_init = m->switch_root_init; | |
1806 | m->switch_root = m->switch_root_init = NULL; | |
1807 | ||
1808 | if (!switch_root_init) | |
1809 | if (prepare_reexecute(m, &arg_serialization, &fds, true) < 0) { | |
1810 | error_message = "Failed to prepare for reexecution"; | |
1811 | goto finish; | |
1812 | } | |
1813 | ||
1814 | reexecute = true; | |
1815 | log_notice("Switching root."); | |
1816 | goto finish; | |
1817 | ||
1818 | case MANAGER_EXIT: | |
1819 | retval = m->return_value; | |
1820 | ||
1821 | if (m->running_as == MANAGER_USER) { | |
1822 | log_debug("Exit."); | |
1823 | goto finish; | |
1824 | } | |
1825 | ||
1826 | /* fallthrough */ | |
1827 | case MANAGER_REBOOT: | |
1828 | case MANAGER_POWEROFF: | |
1829 | case MANAGER_HALT: | |
1830 | case MANAGER_KEXEC: { | |
1831 | static const char * const table[_MANAGER_EXIT_CODE_MAX] = { | |
1832 | [MANAGER_EXIT] = "exit", | |
1833 | [MANAGER_REBOOT] = "reboot", | |
1834 | [MANAGER_POWEROFF] = "poweroff", | |
1835 | [MANAGER_HALT] = "halt", | |
1836 | [MANAGER_KEXEC] = "kexec" | |
1837 | }; | |
1838 | ||
1839 | assert_se(shutdown_verb = table[m->exit_code]); | |
1840 | arm_reboot_watchdog = m->exit_code == MANAGER_REBOOT; | |
1841 | ||
1842 | log_notice("Shutting down."); | |
1843 | goto finish; | |
1844 | } | |
1845 | ||
1846 | default: | |
1847 | assert_not_reached("Unknown exit code."); | |
1848 | } | |
1849 | } | |
1850 | ||
1851 | finish: | |
1852 | pager_close(); | |
1853 | ||
1854 | if (m) | |
1855 | arg_shutdown_watchdog = m->shutdown_watchdog; | |
1856 | ||
1857 | m = manager_free(m); | |
1858 | ||
1859 | for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++) | |
1860 | arg_default_rlimit[j] = mfree(arg_default_rlimit[j]); | |
1861 | ||
1862 | arg_default_unit = mfree(arg_default_unit); | |
1863 | arg_join_controllers = strv_free_free(arg_join_controllers); | |
1864 | arg_default_environment = strv_free(arg_default_environment); | |
1865 | arg_syscall_archs = set_free(arg_syscall_archs); | |
1866 | ||
1867 | mac_selinux_finish(); | |
1868 | ||
1869 | if (reexecute) { | |
1870 | const char **args; | |
1871 | unsigned i, args_size; | |
1872 | ||
1873 | /* Close and disarm the watchdog, so that the new | |
1874 | * instance can reinitialize it, but doesn't get | |
1875 | * rebooted while we do that */ | |
1876 | watchdog_close(true); | |
1877 | ||
1878 | /* Reset the RLIMIT_NOFILE to the kernel default, so | |
1879 | * that the new systemd can pass the kernel default to | |
1880 | * its child processes */ | |
1881 | if (saved_rlimit_nofile.rlim_cur > 0) | |
1882 | (void) setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile); | |
1883 | ||
1884 | if (switch_root_dir) { | |
1885 | /* Kill all remaining processes from the | |
1886 | * initrd, but don't wait for them, so that we | |
1887 | * can handle the SIGCHLD for them after | |
1888 | * deserializing. */ | |
1889 | broadcast_signal(SIGTERM, false, true); | |
1890 | ||
1891 | /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */ | |
1892 | r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE); | |
1893 | if (r < 0) | |
1894 | log_error_errno(r, "Failed to switch root, trying to continue: %m"); | |
1895 | } | |
1896 | ||
1897 | args_size = MAX(6, argc+1); | |
1898 | args = newa(const char*, args_size); | |
1899 | ||
1900 | if (!switch_root_init) { | |
1901 | char sfd[DECIMAL_STR_MAX(int) + 1]; | |
1902 | ||
1903 | /* First try to spawn ourselves with the right | |
1904 | * path, and with full serialization. We do | |
1905 | * this only if the user didn't specify an | |
1906 | * explicit init to spawn. */ | |
1907 | ||
1908 | assert(arg_serialization); | |
1909 | assert(fds); | |
1910 | ||
1911 | xsprintf(sfd, "%i", fileno(arg_serialization)); | |
1912 | ||
1913 | i = 0; | |
1914 | args[i++] = SYSTEMD_BINARY_PATH; | |
1915 | if (switch_root_dir) | |
1916 | args[i++] = "--switched-root"; | |
1917 | args[i++] = arg_running_as == MANAGER_SYSTEM ? "--system" : "--user"; | |
1918 | args[i++] = "--deserialize"; | |
1919 | args[i++] = sfd; | |
1920 | args[i++] = NULL; | |
1921 | ||
1922 | /* do not pass along the environment we inherit from the kernel or initrd */ | |
1923 | if (switch_root_dir) | |
1924 | (void) clearenv(); | |
1925 | ||
1926 | assert(i <= args_size); | |
1927 | (void) execv(args[0], (char* const*) args); | |
1928 | } | |
1929 | ||
1930 | /* Try the fallback, if there is any, without any | |
1931 | * serialization. We pass the original argv[] and | |
1932 | * envp[]. (Well, modulo the ordering changes due to | |
1933 | * getopt() in argv[], and some cleanups in envp[], | |
1934 | * but let's hope that doesn't matter.) */ | |
1935 | ||
1936 | arg_serialization = safe_fclose(arg_serialization); | |
1937 | fds = fdset_free(fds); | |
1938 | ||
1939 | /* Reopen the console */ | |
1940 | (void) make_console_stdio(); | |
1941 | ||
1942 | for (j = 1, i = 1; j < (unsigned) argc; j++) | |
1943 | args[i++] = argv[j]; | |
1944 | args[i++] = NULL; | |
1945 | assert(i <= args_size); | |
1946 | ||
1947 | /* Reenable any blocked signals, especially important | |
1948 | * if we switch from initial ramdisk to init=... */ | |
1949 | (void) reset_all_signal_handlers(); | |
1950 | (void) reset_signal_mask(); | |
1951 | ||
1952 | if (switch_root_init) { | |
1953 | args[0] = switch_root_init; | |
1954 | (void) execv(args[0], (char* const*) args); | |
1955 | log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m"); | |
1956 | } | |
1957 | ||
1958 | args[0] = "/sbin/init"; | |
1959 | (void) execv(args[0], (char* const*) args); | |
1960 | ||
1961 | if (errno == ENOENT) { | |
1962 | log_warning("No /sbin/init, trying fallback"); | |
1963 | ||
1964 | args[0] = "/bin/sh"; | |
1965 | args[1] = NULL; | |
1966 | (void) execv(args[0], (char* const*) args); | |
1967 | log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m"); | |
1968 | } else | |
1969 | log_warning_errno(errno, "Failed to execute /sbin/init, giving up: %m"); | |
1970 | } | |
1971 | ||
1972 | arg_serialization = safe_fclose(arg_serialization); | |
1973 | fds = fdset_free(fds); | |
1974 | ||
1975 | #ifdef HAVE_VALGRIND_VALGRIND_H | |
1976 | /* If we are PID 1 and running under valgrind, then let's exit | |
1977 | * here explicitly. valgrind will only generate nice output on | |
1978 | * exit(), not on exec(), hence let's do the former not the | |
1979 | * latter here. */ | |
1980 | if (getpid() == 1 && RUNNING_ON_VALGRIND) | |
1981 | return 0; | |
1982 | #endif | |
1983 | ||
1984 | if (shutdown_verb) { | |
1985 | char log_level[DECIMAL_STR_MAX(int) + 1]; | |
1986 | char exit_code[DECIMAL_STR_MAX(uint8_t) + 1]; | |
1987 | const char* command_line[11] = { | |
1988 | SYSTEMD_SHUTDOWN_BINARY_PATH, | |
1989 | shutdown_verb, | |
1990 | "--log-level", log_level, | |
1991 | "--log-target", | |
1992 | }; | |
1993 | unsigned pos = 5; | |
1994 | _cleanup_strv_free_ char **env_block = NULL; | |
1995 | ||
1996 | assert(command_line[pos] == NULL); | |
1997 | env_block = strv_copy(environ); | |
1998 | ||
1999 | xsprintf(log_level, "%d", log_get_max_level()); | |
2000 | ||
2001 | switch (log_get_target()) { | |
2002 | ||
2003 | case LOG_TARGET_KMSG: | |
2004 | case LOG_TARGET_JOURNAL_OR_KMSG: | |
2005 | case LOG_TARGET_SYSLOG_OR_KMSG: | |
2006 | command_line[pos++] = "kmsg"; | |
2007 | break; | |
2008 | ||
2009 | case LOG_TARGET_NULL: | |
2010 | command_line[pos++] = "null"; | |
2011 | break; | |
2012 | ||
2013 | case LOG_TARGET_CONSOLE: | |
2014 | default: | |
2015 | command_line[pos++] = "console"; | |
2016 | break; | |
2017 | }; | |
2018 | ||
2019 | if (log_get_show_color()) | |
2020 | command_line[pos++] = "--log-color"; | |
2021 | ||
2022 | if (log_get_show_location()) | |
2023 | command_line[pos++] = "--log-location"; | |
2024 | ||
2025 | if (streq(shutdown_verb, "exit")) { | |
2026 | command_line[pos++] = "--exit-code"; | |
2027 | command_line[pos++] = exit_code; | |
2028 | xsprintf(exit_code, "%d", retval); | |
2029 | } | |
2030 | ||
2031 | assert(pos < ELEMENTSOF(command_line)); | |
2032 | ||
2033 | if (arm_reboot_watchdog && arg_shutdown_watchdog > 0) { | |
2034 | char *e; | |
2035 | ||
2036 | /* If we reboot let's set the shutdown | |
2037 | * watchdog and tell the shutdown binary to | |
2038 | * repeatedly ping it */ | |
2039 | r = watchdog_set_timeout(&arg_shutdown_watchdog); | |
2040 | watchdog_close(r < 0); | |
2041 | ||
2042 | /* Tell the binary how often to ping, ignore failure */ | |
2043 | if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0) | |
2044 | (void) strv_push(&env_block, e); | |
2045 | } else | |
2046 | watchdog_close(true); | |
2047 | ||
2048 | /* Avoid the creation of new processes forked by the | |
2049 | * kernel; at this point, we will not listen to the | |
2050 | * signals anyway */ | |
2051 | if (detect_container() <= 0) | |
2052 | (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER); | |
2053 | ||
2054 | execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block); | |
2055 | log_error_errno(errno, "Failed to execute shutdown binary, %s: %m", | |
2056 | getpid() == 1 ? "freezing" : "quitting"); | |
2057 | } | |
2058 | ||
2059 | if (getpid() == 1) { | |
2060 | if (error_message) | |
2061 | manager_status_printf(NULL, STATUS_TYPE_EMERGENCY, | |
2062 | ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL, | |
2063 | "%s, freezing.", error_message); | |
2064 | freeze_or_reboot(); | |
2065 | } | |
2066 | ||
2067 | return retval; | |
2068 | } |