]> git.ipfire.org Git - thirdparty/systemd.git/blob - src/core/main.c
projects / thirdparty / systemd.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
seccomp: fix build again if libseccomp is missing
[thirdparty/systemd.git] / src / core / main.c
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 <stdio.h>
23 #include <errno.h>
24 #include <string.h>
25 #include <unistd.h>
26 #include <sys/types.h>
27 #include <sys/stat.h>
28 #include <getopt.h>
29 #include <signal.h>
30 #include <sys/wait.h>
31 #include <fcntl.h>
32 #include <sys/prctl.h>
33 #include <sys/mount.h>
34
35 #ifdef HAVE_VALGRIND_VALGRIND_H
36 #include <valgrind/valgrind.h>
37 #endif
38
39 #include "sd-daemon.h"
40 #include "sd-messages.h"
41 #include "sd-bus.h"
42 #include "manager.h"
43 #include "log.h"
44 #include "load-fragment.h"
45 #include "fdset.h"
46 #include "special.h"
47 #include "conf-parser.h"
48 #include "missing.h"
49 #include "label.h"
50 #include "build.h"
51 #include "strv.h"
52 #include "def.h"
53 #include "virt.h"
54 #include "watchdog.h"
55 #include "path-util.h"
56 #include "switch-root.h"
57 #include "capability.h"
58 #include "killall.h"
59 #include "env-util.h"
60 #include "hwclock.h"
61 #include "fileio.h"
62 #include "dbus-manager.h"
63 #include "bus-error.h"
64 #include "bus-util.h"
65
66 #include "mount-setup.h"
67 #include "loopback-setup.h"
68 #include "hostname-setup.h"
69 #include "machine-id-setup.h"
70 #include "selinux-setup.h"
71 #include "ima-setup.h"
72 #include "smack-setup.h"
73 #ifdef HAVE_KMOD
74 #include "kmod-setup.h"
75 #endif
76
77 static enum {
78 ACTION_RUN,
79 ACTION_HELP,
80 ACTION_VERSION,
81 ACTION_TEST,
82 ACTION_DUMP_CONFIGURATION_ITEMS,
83 ACTION_DONE
84 } arg_action = ACTION_RUN;
85 static char *arg_default_unit = NULL;
86 static SystemdRunningAs arg_running_as = _SYSTEMD_RUNNING_AS_INVALID;
87 static bool arg_dump_core = true;
88 static bool arg_crash_shell = false;
89 static int arg_crash_chvt = -1;
90 static bool arg_confirm_spawn = false;
91 static ShowStatus arg_show_status = SHOW_STATUS_UNSET;
92 static bool arg_switched_root = false;
93 static char ***arg_join_controllers = NULL;
94 static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
95 static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
96 static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
97 static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
98 static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
99 static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
100 static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
101 static usec_t arg_runtime_watchdog = 0;
102 static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
103 static char **arg_default_environment = NULL;
104 static struct rlimit *arg_default_rlimit[RLIMIT_NLIMITS] = {};
105 static uint64_t arg_capability_bounding_set_drop = 0;
106 static nsec_t arg_timer_slack_nsec = (nsec_t) -1;
107 static Set* arg_syscall_archs = NULL;
108 static FILE* arg_serialization = NULL;
109
110 static void nop_handler(int sig) {}
111
112 noreturn static void crash(int sig) {
113
114 if (getpid() != 1)
115 /* Pass this on immediately, if this is not PID 1 */
116 raise(sig);
117 else if (!arg_dump_core)
118 log_error("Caught <%s>, not dumping core.", signal_to_string(sig));
119 else {
120 struct sigaction sa = {
121 .sa_handler = nop_handler,
122 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
123 };
124 pid_t pid;
125
126 /* We want to wait for the core process, hence let's enable SIGCHLD */
127 sigaction(SIGCHLD, &sa, NULL);
128
129 pid = fork();
130 if (pid < 0)
131 log_error("Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
132
133 else if (pid == 0) {
134 struct rlimit rl = {};
135
136 /* Enable default signal handler for core dump */
137 zero(sa);
138 sa.sa_handler = SIG_DFL;
139 sigaction(sig, &sa, NULL);
140
141 /* Don't limit the core dump size */
142 rl.rlim_cur = RLIM_INFINITY;
143 rl.rlim_max = RLIM_INFINITY;
144 setrlimit(RLIMIT_CORE, &rl);
145
146 /* Just to be sure... */
147 chdir("/");
148
149 /* Raise the signal again */
150 raise(sig);
151
152 assert_not_reached("We shouldn't be here...");
153 _exit(1);
154
155 } else {
156 siginfo_t status;
157 int r;
158
159 /* Order things nicely. */
160 r = wait_for_terminate(pid, &status);
161 if (r < 0)
162 log_error("Caught <%s>, waitpid() failed: %s", signal_to_string(sig), strerror(-r));
163 else if (status.si_code != CLD_DUMPED)
164 log_error("Caught <%s>, core dump failed.", signal_to_string(sig));
165 else
166 log_error("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
167 }
168 }
169
170 if (arg_crash_chvt)
171 chvt(arg_crash_chvt);
172
173 if (arg_crash_shell) {
174 struct sigaction sa = {
175 .sa_handler = SIG_IGN,
176 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
177 };
178 pid_t pid;
179
180 log_info("Executing crash shell in 10s...");
181 sleep(10);
182
183 /* Let the kernel reap children for us */
184 assert_se(sigaction(SIGCHLD, &sa, NULL) == 0);
185
186 pid = fork();
187 if (pid < 0)
188 log_error("Failed to fork off crash shell: %m");
189 else if (pid == 0) {
190 make_console_stdio();
191 execl("/bin/sh", "/bin/sh", NULL);
192
193 log_error("execl() failed: %m");
194 _exit(1);
195 }
196
197 log_info("Successfully spawned crash shell as pid "PID_FMT".", pid);
198 }
199
200 log_info("Freezing execution.");
201 freeze();
202 }
203
204 static void install_crash_handler(void) {
205 struct sigaction sa = {
206 .sa_handler = crash,
207 .sa_flags = SA_NODEFER,
208 };
209
210 sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
211 }
212
213 static int console_setup(bool do_reset) {
214 int tty_fd, r;
215
216 /* If we are init, we connect stdin/stdout/stderr to /dev/null
217 * and make sure we don't have a controlling tty. */
218
219 release_terminal();
220
221 if (!do_reset)
222 return 0;
223
224 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
225 if (tty_fd < 0) {
226 log_error("Failed to open /dev/console: %s", strerror(-tty_fd));
227 return -tty_fd;
228 }
229
230 /* We don't want to force text mode.
231 * plymouth may be showing pictures already from initrd. */
232 r = reset_terminal_fd(tty_fd, false);
233 if (r < 0)
234 log_error("Failed to reset /dev/console: %s", strerror(-r));
235
236 close_nointr_nofail(tty_fd);
237 return r;
238 }
239
240 static int set_default_unit(const char *u) {
241 char *c;
242
243 assert(u);
244
245 c = strdup(u);
246 if (!c)
247 return -ENOMEM;
248
249 free(arg_default_unit);
250 arg_default_unit = c;
251
252 return 0;
253 }
254
255 static int parse_proc_cmdline_word(const char *word) {
256
257 static const char * const rlmap[] = {
258 "emergency", SPECIAL_EMERGENCY_TARGET,
259 "-b", SPECIAL_EMERGENCY_TARGET,
260 "single", SPECIAL_RESCUE_TARGET,
261 "-s", SPECIAL_RESCUE_TARGET,
262 "s", SPECIAL_RESCUE_TARGET,
263 "S", SPECIAL_RESCUE_TARGET,
264 "1", SPECIAL_RESCUE_TARGET,
265 "2", SPECIAL_RUNLEVEL2_TARGET,
266 "3", SPECIAL_RUNLEVEL3_TARGET,
267 "4", SPECIAL_RUNLEVEL4_TARGET,
268 "5", SPECIAL_RUNLEVEL5_TARGET,
269 };
270
271 assert(word);
272
273 if (startswith(word, "systemd.unit=")) {
274
275 if (!in_initrd())
276 return set_default_unit(word + 13);
277
278 } else if (startswith(word, "rd.systemd.unit=")) {
279
280 if (in_initrd())
281 return set_default_unit(word + 16);
282
283 } else if (startswith(word, "systemd.log_target=")) {
284
285 if (log_set_target_from_string(word + 19) < 0)
286 log_warning("Failed to parse log target %s. Ignoring.", word + 19);
287
288 } else if (startswith(word, "systemd.log_level=")) {
289
290 if (log_set_max_level_from_string(word + 18) < 0)
291 log_warning("Failed to parse log level %s. Ignoring.", word + 18);
292
293 } else if (startswith(word, "systemd.log_color=")) {
294
295 if (log_show_color_from_string(word + 18) < 0)
296 log_warning("Failed to parse log color setting %s. Ignoring.", word + 18);
297
298 } else if (startswith(word, "systemd.log_location=")) {
299
300 if (log_show_location_from_string(word + 21) < 0)
301 log_warning("Failed to parse log location setting %s. Ignoring.", word + 21);
302
303 } else if (startswith(word, "systemd.dump_core=")) {
304 int r;
305
306 if ((r = parse_boolean(word + 18)) < 0)
307 log_warning("Failed to parse dump core switch %s. Ignoring.", word + 18);
308 else
309 arg_dump_core = r;
310
311 } else if (startswith(word, "systemd.crash_shell=")) {
312 int r;
313
314 if ((r = parse_boolean(word + 20)) < 0)
315 log_warning("Failed to parse crash shell switch %s. Ignoring.", word + 20);
316 else
317 arg_crash_shell = r;
318
319 } else if (startswith(word, "systemd.confirm_spawn=")) {
320 int r;
321
322 if ((r = parse_boolean(word + 22)) < 0)
323 log_warning("Failed to parse confirm spawn switch %s. Ignoring.", word + 22);
324 else
325 arg_confirm_spawn = r;
326
327 } else if (startswith(word, "systemd.crash_chvt=")) {
328 int k;
329
330 if (safe_atoi(word + 19, &k) < 0)
331 log_warning("Failed to parse crash chvt switch %s. Ignoring.", word + 19);
332 else
333 arg_crash_chvt = k;
334
335 } else if (startswith(word, "systemd.show_status=")) {
336 int r;
337
338 r = parse_show_status(word + 20, &arg_show_status);
339 if (r < 0)
340 log_warning("Failed to parse show status switch %s. Ignoring.", word + 20);
341 } else if (startswith(word, "systemd.default_standard_output=")) {
342 int r;
343
344 if ((r = exec_output_from_string(word + 32)) < 0)
345 log_warning("Failed to parse default standard output switch %s. Ignoring.", word + 32);
346 else
347 arg_default_std_output = r;
348 } else if (startswith(word, "systemd.default_standard_error=")) {
349 int r;
350
351 if ((r = exec_output_from_string(word + 31)) < 0)
352 log_warning("Failed to parse default standard error switch %s. Ignoring.", word + 31);
353 else
354 arg_default_std_error = r;
355 } else if (startswith(word, "systemd.setenv=")) {
356 _cleanup_free_ char *cenv = NULL;
357
358 cenv = strdup(word + 15);
359 if (!cenv)
360 return -ENOMEM;
361
362 if (env_assignment_is_valid(cenv)) {
363 char **env;
364
365 env = strv_env_set(arg_default_environment, cenv);
366 if (env)
367 arg_default_environment = env;
368 else
369 log_warning("Setting environment variable '%s' failed, ignoring: %m", cenv);
370 } else
371 log_warning("Environment variable name '%s' is not valid. Ignoring.", cenv);
372
373 } else if (startswith(word, "systemd.") ||
374 (in_initrd() && startswith(word, "rd.systemd."))) {
375
376 const char *c;
377
378 /* Ignore systemd.journald.xyz and friends */
379 c = word;
380 if (startswith(c, "rd."))
381 c += 3;
382 if (startswith(c, "systemd."))
383 c += 8;
384 if (c[strcspn(c, ".=")] != '.') {
385
386 log_warning("Unknown kernel switch %s. Ignoring.", word);
387
388 log_info("Supported kernel switches:\n"
389 "systemd.unit=UNIT Default unit to start\n"
390 "rd.systemd.unit=UNIT Default unit to start when run in initrd\n"
391 "systemd.dump_core=0|1 Dump core on crash\n"
392 "systemd.crash_shell=0|1 Run shell on crash\n"
393 "systemd.crash_chvt=N Change to VT #N on crash\n"
394 "systemd.confirm_spawn=0|1 Confirm every process spawn\n"
395 "systemd.show_status=0|1|auto Show status updates on the console during bootup\n"
396 "systemd.log_target=console|kmsg|journal|journal-or-kmsg|syslog|syslog-or-kmsg|null\n"
397 " Log target\n"
398 "systemd.log_level=LEVEL Log level\n"
399 "systemd.log_color=0|1 Highlight important log messages\n"
400 "systemd.log_location=0|1 Include code location in log messages\n"
401 "systemd.default_standard_output=null|tty|syslog|syslog+console|kmsg|kmsg+console|journal|journal+console\n"
402 " Set default log output for services\n"
403 "systemd.default_standard_error=null|tty|syslog|syslog+console|kmsg|kmsg+console|journal|journal+console\n"
404 " Set default log error output for services\n"
405 "systemd.setenv=ASSIGNMENT Set an environment variable for all spawned processes\n");
406 }
407
408 } else if (streq(word, "quiet")) {
409 if (arg_show_status == SHOW_STATUS_UNSET)
410 arg_show_status = SHOW_STATUS_AUTO;
411 } else if (streq(word, "debug")) {
412 /* Log to kmsg, the journal socket will fill up before the
413 * journal is started and tools running during that time
414 * will block with every log message for for 60 seconds,
415 * before they give up. */
416 log_set_max_level(LOG_DEBUG);
417 log_set_target(detect_container(NULL) > 0 ? LOG_TARGET_CONSOLE : LOG_TARGET_KMSG);
418 } else if (!in_initrd()) {
419 unsigned i;
420
421 /* SysV compatibility */
422 for (i = 0; i < ELEMENTSOF(rlmap); i += 2)
423 if (streq(word, rlmap[i]))
424 return set_default_unit(rlmap[i+1]);
425 }
426
427 return 0;
428 }
429
430 #define DEFINE_SETTER(name, func, descr) \
431 static int name(const char *unit, \
432 const char *filename, \
433 unsigned line, \
434 const char *section, \
435 unsigned section_line, \
436 const char *lvalue, \
437 int ltype, \
438 const char *rvalue, \
439 void *data, \
440 void *userdata) { \
441 \
442 int r; \
443 \
444 assert(filename); \
445 assert(lvalue); \
446 assert(rvalue); \
447 \
448 r = func(rvalue); \
449 if (r < 0) \
450 log_syntax(unit, LOG_ERR, filename, line, -r, \
451 "Invalid " descr "'%s': %s", \
452 rvalue, strerror(-r)); \
453 \
454 return 0; \
455 }
456
457 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level")
458 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target")
459 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" )
460 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location")
461
462 static int config_parse_cpu_affinity2(const char *unit,
463 const char *filename,
464 unsigned line,
465 const char *section,
466 unsigned section_line,
467 const char *lvalue,
468 int ltype,
469 const char *rvalue,
470 void *data,
471 void *userdata) {
472
473 char *w;
474 size_t l;
475 char *state;
476 cpu_set_t *c = NULL;
477 unsigned ncpus = 0;
478
479 assert(filename);
480 assert(lvalue);
481 assert(rvalue);
482
483 FOREACH_WORD_QUOTED(w, l, rvalue, state) {
484 char *t;
485 int r;
486 unsigned cpu;
487
488 if (!(t = strndup(w, l)))
489 return log_oom();
490
491 r = safe_atou(t, &cpu);
492 free(t);
493
494 if (!c)
495 if (!(c = cpu_set_malloc(&ncpus)))
496 return log_oom();
497
498 if (r < 0 || cpu >= ncpus) {
499 log_syntax(unit, LOG_ERR, filename, line, -r,
500 "Failed to parse CPU affinity '%s'", rvalue);
501 CPU_FREE(c);
502 return -EBADMSG;
503 }
504
505 CPU_SET_S(cpu, CPU_ALLOC_SIZE(ncpus), c);
506 }
507
508 if (c) {
509 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
510 log_warning_unit(unit, "Failed to set CPU affinity: %m");
511
512 CPU_FREE(c);
513 }
514
515 return 0;
516 }
517
518 static void strv_free_free(char ***l) {
519 char ***i;
520
521 if (!l)
522 return;
523
524 for (i = l; *i; i++)
525 strv_free(*i);
526
527 free(l);
528 }
529
530 static void free_join_controllers(void) {
531 strv_free_free(arg_join_controllers);
532 arg_join_controllers = NULL;
533 }
534
535 static int config_parse_join_controllers(const char *unit,
536 const char *filename,
537 unsigned line,
538 const char *section,
539 unsigned section_line,
540 const char *lvalue,
541 int ltype,
542 const char *rvalue,
543 void *data,
544 void *userdata) {
545
546 unsigned n = 0;
547 char *state, *w;
548 size_t length;
549
550 assert(filename);
551 assert(lvalue);
552 assert(rvalue);
553
554 free_join_controllers();
555
556 FOREACH_WORD_QUOTED(w, length, rvalue, state) {
557 char *s, **l;
558
559 s = strndup(w, length);
560 if (!s)
561 return log_oom();
562
563 l = strv_split(s, ",");
564 free(s);
565
566 strv_uniq(l);
567
568 if (strv_length(l) <= 1) {
569 strv_free(l);
570 continue;
571 }
572
573 if (!arg_join_controllers) {
574 arg_join_controllers = new(char**, 2);
575 if (!arg_join_controllers) {
576 strv_free(l);
577 return log_oom();
578 }
579
580 arg_join_controllers[0] = l;
581 arg_join_controllers[1] = NULL;
582
583 n = 1;
584 } else {
585 char ***a;
586 char ***t;
587
588 t = new0(char**, n+2);
589 if (!t) {
590 strv_free(l);
591 return log_oom();
592 }
593
594 n = 0;
595
596 for (a = arg_join_controllers; *a; a++) {
597
598 if (strv_overlap(*a, l)) {
599 if (strv_extend_strv(&l, *a) < 0) {
600 strv_free(l);
601 strv_free_free(t);
602 return log_oom();
603 }
604
605 } else {
606 char **c;
607
608 c = strv_copy(*a);
609 if (!c) {
610 strv_free(l);
611 strv_free_free(t);
612 return log_oom();
613 }
614
615 t[n++] = c;
616 }
617 }
618
619 t[n++] = strv_uniq(l);
620
621 strv_free_free(arg_join_controllers);
622 arg_join_controllers = t;
623 }
624 }
625
626 return 0;
627 }
628
629 static int parse_config_file(void) {
630
631 const ConfigTableItem items[] = {
632 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
633 { "Manager", "LogTarget", config_parse_target, 0, NULL },
634 { "Manager", "LogColor", config_parse_color, 0, NULL },
635 { "Manager", "LogLocation", config_parse_location, 0, NULL },
636 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
637 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
638 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
639 { "Manager", "CrashChVT", config_parse_int, 0, &arg_crash_chvt },
640 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
641 { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
642 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
643 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
644 { "Manager", "CapabilityBoundingSet", config_parse_bounding_set, 0, &arg_capability_bounding_set_drop },
645 #ifdef HAVE_SECCOMP
646 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
647 #endif
648 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
649 { "Manager", "DefaultStandardOutput", config_parse_output, 0, &arg_default_std_output },
650 { "Manager", "DefaultStandardError", config_parse_output, 0, &arg_default_std_error },
651 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
652 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
653 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
654 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval },
655 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
656 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
657 { "Manager", "DefaultLimitCPU", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CPU] },
658 { "Manager", "DefaultLimitFSIZE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_FSIZE] },
659 { "Manager", "DefaultLimitDATA", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_DATA] },
660 { "Manager", "DefaultLimitSTACK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_STACK] },
661 { "Manager", "DefaultLimitCORE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CORE] },
662 { "Manager", "DefaultLimitRSS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RSS] },
663 { "Manager", "DefaultLimitNOFILE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NOFILE] },
664 { "Manager", "DefaultLimitAS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_AS] },
665 { "Manager", "DefaultLimitNPROC", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NPROC] },
666 { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MEMLOCK] },
667 { "Manager", "DefaultLimitLOCKS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_LOCKS] },
668 { "Manager", "DefaultLimitSIGPENDING", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_SIGPENDING] },
669 { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MSGQUEUE] },
670 { "Manager", "DefaultLimitNICE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NICE] },
671 { "Manager", "DefaultLimitRTPRIO", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTPRIO] },
672 { "Manager", "DefaultLimitRTTIME", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTTIME] },
673 {}
674 };
675
676 _cleanup_fclose_ FILE *f;
677 const char *fn;
678 int r;
679
680 fn = arg_running_as == SYSTEMD_SYSTEM ? PKGSYSCONFDIR "/system.conf" : PKGSYSCONFDIR "/user.conf";
681 f = fopen(fn, "re");
682 if (!f) {
683 if (errno == ENOENT)
684 return 0;
685
686 log_warning("Failed to open configuration file '%s': %m", fn);
687 return 0;
688 }
689
690 r = config_parse(NULL, fn, f, "Manager\0", config_item_table_lookup, (void*) items, false, false, NULL);
691 if (r < 0)
692 log_warning("Failed to parse configuration file: %s", strerror(-r));
693
694 return 0;
695 }
696
697 static int parse_proc_cmdline(void) {
698 _cleanup_free_ char *line = NULL;
699 char *w, *state;
700 size_t l;
701 int r;
702
703 r = proc_cmdline(&line);
704 if (r < 0)
705 log_warning("Failed to read /proc/cmdline, ignoring: %s", strerror(-r));
706 if (r <= 0)
707 return 0;
708
709 FOREACH_WORD_QUOTED(w, l, line, state) {
710 _cleanup_free_ char *word;
711
712 word = strndup(w, l);
713 if (!word)
714 return log_oom();
715
716 r = parse_proc_cmdline_word(word);
717 if (r < 0) {
718 log_error("Failed on cmdline argument %s: %s", word, strerror(-r));
719 return r;
720 }
721 }
722
723 return 0;
724 }
725
726 static int parse_argv(int argc, char *argv[]) {
727
728 enum {
729 ARG_LOG_LEVEL = 0x100,
730 ARG_LOG_TARGET,
731 ARG_LOG_COLOR,
732 ARG_LOG_LOCATION,
733 ARG_UNIT,
734 ARG_SYSTEM,
735 ARG_USER,
736 ARG_TEST,
737 ARG_VERSION,
738 ARG_DUMP_CONFIGURATION_ITEMS,
739 ARG_DUMP_CORE,
740 ARG_CRASH_SHELL,
741 ARG_CONFIRM_SPAWN,
742 ARG_SHOW_STATUS,
743 ARG_DESERIALIZE,
744 ARG_SWITCHED_ROOT,
745 ARG_DEFAULT_STD_OUTPUT,
746 ARG_DEFAULT_STD_ERROR
747 };
748
749 static const struct option options[] = {
750 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
751 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
752 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
753 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
754 { "unit", required_argument, NULL, ARG_UNIT },
755 { "system", no_argument, NULL, ARG_SYSTEM },
756 { "user", no_argument, NULL, ARG_USER },
757 { "test", no_argument, NULL, ARG_TEST },
758 { "help", no_argument, NULL, 'h' },
759 { "version", no_argument, NULL, ARG_VERSION },
760 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
761 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
762 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
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 { NULL, 0, NULL, 0 }
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 if ((r = log_set_max_level_from_string(optarg)) < 0) {
786 log_error("Failed to parse log level %s.", optarg);
787 return r;
788 }
789
790 break;
791
792 case ARG_LOG_TARGET:
793
794 if ((r = log_set_target_from_string(optarg)) < 0) {
795 log_error("Failed to parse log target %s.", optarg);
796 return r;
797 }
798
799 break;
800
801 case ARG_LOG_COLOR:
802
803 if (optarg) {
804 if ((r = log_show_color_from_string(optarg)) < 0) {
805 log_error("Failed to parse log color setting %s.", optarg);
806 return r;
807 }
808 } else
809 log_show_color(true);
810
811 break;
812
813 case ARG_LOG_LOCATION:
814
815 if (optarg) {
816 if ((r = log_show_location_from_string(optarg)) < 0) {
817 log_error("Failed to parse log location setting %s.", optarg);
818 return r;
819 }
820 } else
821 log_show_location(true);
822
823 break;
824
825 case ARG_DEFAULT_STD_OUTPUT:
826
827 if ((r = exec_output_from_string(optarg)) < 0) {
828 log_error("Failed to parse default standard output setting %s.", optarg);
829 return r;
830 } else
831 arg_default_std_output = r;
832 break;
833
834 case ARG_DEFAULT_STD_ERROR:
835
836 if ((r = exec_output_from_string(optarg)) < 0) {
837 log_error("Failed to parse default standard error output setting %s.", optarg);
838 return r;
839 } else
840 arg_default_std_error = r;
841 break;
842
843 case ARG_UNIT:
844
845 if ((r = set_default_unit(optarg)) < 0) {
846 log_error("Failed to set default unit %s: %s", optarg, strerror(-r));
847 return r;
848 }
849
850 break;
851
852 case ARG_SYSTEM:
853 arg_running_as = SYSTEMD_SYSTEM;
854 break;
855
856 case ARG_USER:
857 arg_running_as = SYSTEMD_USER;
858 break;
859
860 case ARG_TEST:
861 arg_action = ACTION_TEST;
862 break;
863
864 case ARG_VERSION:
865 arg_action = ACTION_VERSION;
866 break;
867
868 case ARG_DUMP_CONFIGURATION_ITEMS:
869 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
870 break;
871
872 case ARG_DUMP_CORE:
873 r = optarg ? parse_boolean(optarg) : 1;
874 if (r < 0) {
875 log_error("Failed to parse dump core boolean %s.", optarg);
876 return r;
877 }
878 arg_dump_core = r;
879 break;
880
881 case ARG_CRASH_SHELL:
882 r = optarg ? parse_boolean(optarg) : 1;
883 if (r < 0) {
884 log_error("Failed to parse crash shell boolean %s.", optarg);
885 return r;
886 }
887 arg_crash_shell = r;
888 break;
889
890 case ARG_CONFIRM_SPAWN:
891 r = optarg ? parse_boolean(optarg) : 1;
892 if (r < 0) {
893 log_error("Failed to parse confirm spawn boolean %s.", optarg);
894 return r;
895 }
896 arg_confirm_spawn = r;
897 break;
898
899 case ARG_SHOW_STATUS:
900 if (optarg) {
901 r = parse_show_status(optarg, &arg_show_status);
902 if (r < 0) {
903 log_error("Failed to parse show status boolean %s.", optarg);
904 return r;
905 }
906 } else
907 arg_show_status = SHOW_STATUS_YES;
908 break;
909
910 case ARG_DESERIALIZE: {
911 int fd;
912 FILE *f;
913
914 r = safe_atoi(optarg, &fd);
915 if (r < 0 || fd < 0) {
916 log_error("Failed to parse deserialize option %s.", optarg);
917 return r < 0 ? r : -EINVAL;
918 }
919
920 fd_cloexec(fd, true);
921
922 f = fdopen(fd, "r");
923 if (!f) {
924 log_error("Failed to open serialization fd: %m");
925 return -errno;
926 }
927
928 if (arg_serialization)
929 fclose(arg_serialization);
930
931 arg_serialization = f;
932
933 break;
934 }
935
936 case ARG_SWITCHED_ROOT:
937 arg_switched_root = true;
938 break;
939
940 case 'h':
941 arg_action = ACTION_HELP;
942 break;
943
944 case 'D':
945 log_set_max_level(LOG_DEBUG);
946 break;
947
948 case 'b':
949 case 's':
950 case 'z':
951 /* Just to eat away the sysvinit kernel
952 * cmdline args without getopt() error
953 * messages that we'll parse in
954 * parse_proc_cmdline_word() or ignore. */
955
956 case '?':
957 default:
958 if (getpid() != 1) {
959 log_error("Unknown option code %c", c);
960 return -EINVAL;
961 }
962
963 break;
964 }
965
966 if (optind < argc && getpid() != 1) {
967 /* Hmm, when we aren't run as init system
968 * let's complain about excess arguments */
969
970 log_error("Excess arguments.");
971 return -EINVAL;
972 }
973
974 if (detect_container(NULL) > 0) {
975 char **a;
976
977 /* All /proc/cmdline arguments the kernel didn't
978 * understand it passed to us. We're not really
979 * interested in that usually since /proc/cmdline is
980 * more interesting and complete. With one exception:
981 * if we are run in a container /proc/cmdline is not
982 * relevant for the container, hence we rely on argv[]
983 * instead. */
984
985 for (a = argv; a < argv + argc; a++) {
986 r = parse_proc_cmdline_word(*a);
987 if (r < 0) {
988 log_error("Failed on cmdline argument %s: %s", *a, strerror(-r));
989 return r;
990 }
991 }
992 }
993
994 return 0;
995 }
996
997 static int help(void) {
998
999 printf("%s [OPTIONS...]\n\n"
1000 "Starts up and maintains the system or user services.\n\n"
1001 " -h --help Show this help\n"
1002 " --test Determine startup sequence, dump it and exit\n"
1003 " --dump-configuration-items Dump understood unit configuration items\n"
1004 " --unit=UNIT Set default unit\n"
1005 " --system Run a system instance, even if PID != 1\n"
1006 " --user Run a user instance\n"
1007 " --dump-core[=0|1] Dump core on crash\n"
1008 " --crash-shell[=0|1] Run shell on crash\n"
1009 " --confirm-spawn[=0|1] Ask for confirmation when spawning processes\n"
1010 " --show-status[=0|1] Show status updates on the console during bootup\n"
1011 " --log-target=TARGET Set log target (console, journal, syslog, kmsg, journal-or-kmsg, syslog-or-kmsg, null)\n"
1012 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1013 " --log-color[=0|1] Highlight important log messages\n"
1014 " --log-location[=0|1] Include code location in log messages\n"
1015 " --default-standard-output= Set default standard output for services\n"
1016 " --default-standard-error= Set default standard error output for services\n",
1017 program_invocation_short_name);
1018
1019 return 0;
1020 }
1021
1022 static int version(void) {
1023 puts(PACKAGE_STRING);
1024 puts(SYSTEMD_FEATURES);
1025
1026 return 0;
1027 }
1028
1029 static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
1030 FILE *f = NULL;
1031 FDSet *fds = NULL;
1032 int r;
1033
1034 assert(m);
1035 assert(_f);
1036 assert(_fds);
1037
1038 r = manager_open_serialization(m, &f);
1039 if (r < 0) {
1040 log_error("Failed to create serialization file: %s", strerror(-r));
1041 goto fail;
1042 }
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 r = -ENOMEM;
1051 log_error("Failed to allocate fd set: %s", strerror(-r));
1052 goto fail;
1053 }
1054
1055 r = manager_serialize(m, f, fds, switching_root);
1056 if (r < 0) {
1057 log_error("Failed to serialize state: %s", strerror(-r));
1058 goto fail;
1059 }
1060
1061 if (fseeko(f, 0, SEEK_SET) < 0) {
1062 log_error("Failed to rewind serialization fd: %m");
1063 goto fail;
1064 }
1065
1066 r = fd_cloexec(fileno(f), false);
1067 if (r < 0) {
1068 log_error("Failed to disable O_CLOEXEC for serialization: %s", strerror(-r));
1069 goto fail;
1070 }
1071
1072 r = fdset_cloexec(fds, false);
1073 if (r < 0) {
1074 log_error("Failed to disable O_CLOEXEC for serialization fds: %s", strerror(-r));
1075 goto fail;
1076 }
1077
1078 *_f = f;
1079 *_fds = fds;
1080
1081 return 0;
1082
1083 fail:
1084 fdset_free(fds);
1085
1086 if (f)
1087 fclose(f);
1088
1089 return r;
1090 }
1091
1092 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1093 struct rlimit nl;
1094 int r;
1095
1096 assert(saved_rlimit);
1097
1098 /* Save the original RLIMIT_NOFILE so that we can reset it
1099 * later when transitioning from the initrd to the main
1100 * systemd or suchlike. */
1101 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0) {
1102 log_error("Reading RLIMIT_NOFILE failed: %m");
1103 return -errno;
1104 }
1105
1106 /* Make sure forked processes get the default kernel setting */
1107 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1108 struct rlimit *rl;
1109
1110 rl = newdup(struct rlimit, saved_rlimit, 1);
1111 if (!rl)
1112 return log_oom();
1113
1114 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1115 }
1116
1117 /* Bump up the resource limit for ourselves substantially */
1118 nl.rlim_cur = nl.rlim_max = 64*1024;
1119 r = setrlimit_closest(RLIMIT_NOFILE, &nl);
1120 if (r < 0) {
1121 log_error("Setting RLIMIT_NOFILE failed: %s", strerror(-r));
1122 return r;
1123 }
1124
1125 return 0;
1126 }
1127
1128 static void test_mtab(void) {
1129 char *p;
1130
1131 /* Check that /etc/mtab is a symlink */
1132
1133 if (readlink_malloc("/etc/mtab", &p) >= 0) {
1134 bool b;
1135
1136 b = streq(p, "/proc/self/mounts") || streq(p, "/proc/mounts");
1137 free(p);
1138
1139 if (b)
1140 return;
1141 }
1142
1143 log_warning("/etc/mtab is not a symlink or not pointing to /proc/self/mounts. "
1144 "This is not supported anymore. "
1145 "Please make sure to replace this file by a symlink to avoid incorrect or misleading mount(8) output.");
1146 }
1147
1148 static void test_usr(void) {
1149
1150 /* Check that /usr is not a separate fs */
1151
1152 if (dir_is_empty("/usr") <= 0)
1153 return;
1154
1155 log_warning("/usr appears to be on its own filesytem and is not already mounted. This is not a supported setup. "
1156 "Some things will probably break (sometimes even silently) in mysterious ways. "
1157 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1158 }
1159
1160 static void test_cgroups(void) {
1161
1162 if (access("/proc/cgroups", F_OK) >= 0)
1163 return;
1164
1165 log_warning("CONFIG_CGROUPS was not set when your kernel was compiled. "
1166 "Systems without control groups are not supported. "
1167 "We will now sleep for 10s, and then continue boot-up. "
1168 "Expect breakage and please do not file bugs. "
1169 "Instead fix your kernel and enable CONFIG_CGROUPS. "
1170 "Consult http://0pointer.de/blog/projects/cgroups-vs-cgroups.html for more information.");
1171
1172 sleep(10);
1173 }
1174
1175 static int initialize_join_controllers(void) {
1176 /* By default, mount "cpu" + "cpuacct" together, and "net_cls"
1177 * + "net_prio". We'd like to add "cpuset" to the mix, but
1178 * "cpuset" does't really work for groups with no initialized
1179 * attributes. */
1180
1181 arg_join_controllers = new(char**, 3);
1182 if (!arg_join_controllers)
1183 return -ENOMEM;
1184
1185 arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL);
1186 arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL);
1187 arg_join_controllers[2] = NULL;
1188
1189 if (!arg_join_controllers[0] || !arg_join_controllers[1]) {
1190 free_join_controllers();
1191 return -ENOMEM;
1192 }
1193
1194 return 0;
1195 }
1196
1197 static int enforce_syscall_archs(Set *archs) {
1198 #ifdef HAVE_SECCOMP
1199 scmp_filter_ctx *seccomp;
1200 Iterator i;
1201 void *id;
1202 int r;
1203
1204 seccomp = seccomp_init(SCMP_ACT_ALLOW);
1205 if (!seccomp)
1206 return log_oom();
1207
1208 SET_FOREACH(id, arg_syscall_archs, i) {
1209 r = seccomp_arch_add(seccomp, PTR_TO_UINT32(id) - 1);
1210 if (r == -EEXIST)
1211 continue;
1212 if (r < 0) {
1213 log_error("Failed to add architecture to seccomp: %s", strerror(-r));
1214 goto finish;
1215 }
1216 }
1217
1218 r = seccomp_load(seccomp);
1219 if (r < 0)
1220 log_error("Failed to add install architecture seccomp: %s", strerror(-r));
1221
1222 finish:
1223 seccomp_release(seccomp);
1224 return r;
1225 #else
1226 return 0;
1227 #endif
1228 }
1229
1230 int main(int argc, char *argv[]) {
1231 Manager *m = NULL;
1232 int r, retval = EXIT_FAILURE;
1233 usec_t before_startup, after_startup;
1234 char timespan[FORMAT_TIMESPAN_MAX];
1235 FDSet *fds = NULL;
1236 bool reexecute = false;
1237 const char *shutdown_verb = NULL;
1238 dual_timestamp initrd_timestamp = { 0ULL, 0ULL };
1239 dual_timestamp userspace_timestamp = { 0ULL, 0ULL };
1240 dual_timestamp kernel_timestamp = { 0ULL, 0ULL };
1241 dual_timestamp security_start_timestamp = { 0ULL, 0ULL };
1242 dual_timestamp security_finish_timestamp = { 0ULL, 0ULL };
1243 static char systemd[] = "systemd";
1244 bool skip_setup = false;
1245 unsigned j;
1246 bool loaded_policy = false;
1247 bool arm_reboot_watchdog = false;
1248 bool queue_default_job = false;
1249 char *switch_root_dir = NULL, *switch_root_init = NULL;
1250 static struct rlimit saved_rlimit_nofile = { 0, 0 };
1251
1252 #ifdef HAVE_SYSV_COMPAT
1253 if (getpid() != 1 && strstr(program_invocation_short_name, "init")) {
1254 /* This is compatibility support for SysV, where
1255 * calling init as a user is identical to telinit. */
1256
1257 errno = -ENOENT;
1258 execv(SYSTEMCTL_BINARY_PATH, argv);
1259 log_error("Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1260 return 1;
1261 }
1262 #endif
1263
1264 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
1265 dual_timestamp_get(&userspace_timestamp);
1266
1267 /* Determine if this is a reexecution or normal bootup. We do
1268 * the full command line parsing much later, so let's just
1269 * have a quick peek here. */
1270 if (strv_find(argv+1, "--deserialize"))
1271 skip_setup = true;
1272
1273 /* If we have switched root, do all the special setup
1274 * things */
1275 if (strv_find(argv+1, "--switched-root"))
1276 skip_setup = false;
1277
1278 /* If we get started via the /sbin/init symlink then we are
1279 called 'init'. After a subsequent reexecution we are then
1280 called 'systemd'. That is confusing, hence let's call us
1281 systemd right-away. */
1282 program_invocation_short_name = systemd;
1283 prctl(PR_SET_NAME, systemd);
1284
1285 saved_argv = argv;
1286 saved_argc = argc;
1287
1288 log_show_color(isatty(STDERR_FILENO) > 0);
1289
1290 /* Disable the umask logic */
1291 if (getpid() == 1)
1292 umask(0);
1293
1294 if (getpid() == 1 && detect_container(NULL) <= 0) {
1295
1296 /* Running outside of a container as PID 1 */
1297 arg_running_as = SYSTEMD_SYSTEM;
1298 make_null_stdio();
1299 log_set_target(LOG_TARGET_KMSG);
1300 log_open();
1301
1302 if (in_initrd())
1303 initrd_timestamp = userspace_timestamp;
1304
1305 if (!skip_setup) {
1306 mount_setup_early();
1307 dual_timestamp_get(&security_start_timestamp);
1308 if (selinux_setup(&loaded_policy) < 0)
1309 goto finish;
1310 if (ima_setup() < 0)
1311 goto finish;
1312 if (smack_setup() < 0)
1313 goto finish;
1314 dual_timestamp_get(&security_finish_timestamp);
1315 }
1316
1317 if (label_init(NULL) < 0)
1318 goto finish;
1319
1320 if (!skip_setup) {
1321 if (hwclock_is_localtime() > 0) {
1322 int min;
1323
1324 /* The first-time call to settimeofday() does a time warp in the kernel */
1325 r = hwclock_set_timezone(&min);
1326 if (r < 0)
1327 log_error("Failed to apply local time delta, ignoring: %s", strerror(-r));
1328 else
1329 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1330 } else if (!in_initrd()) {
1331 /*
1332 * Do dummy first-time call to seal the kernel's time warp magic
1333 *
1334 * Do not call this this from inside the initrd. The initrd might not
1335 * carry /etc/adjtime with LOCAL, but the real system could be set up
1336 * that way. In such case, we need to delay the time-warp or the sealing
1337 * until we reach the real system.
1338 */
1339 hwclock_reset_timezone();
1340
1341 /* Tell the kernel our timezone */
1342 r = hwclock_set_timezone(NULL);
1343 if (r < 0)
1344 log_error("Failed to set the kernel's timezone, ignoring: %s", strerror(-r));
1345 }
1346 }
1347
1348 /* Set the default for later on, but don't actually
1349 * open the logs like this for now. Note that if we
1350 * are transitioning from the initrd there might still
1351 * be journal fd open, and we shouldn't attempt
1352 * opening that before we parsed /proc/cmdline which
1353 * might redirect output elsewhere. */
1354 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
1355
1356 } else if (getpid() == 1) {
1357 /* Running inside a container, as PID 1 */
1358 arg_running_as = SYSTEMD_SYSTEM;
1359 log_set_target(LOG_TARGET_CONSOLE);
1360 log_close_console(); /* force reopen of /dev/console */
1361 log_open();
1362
1363 /* For the later on, see above... */
1364 log_set_target(LOG_TARGET_JOURNAL);
1365
1366 /* clear the kernel timestamp,
1367 * because we are in a container */
1368 kernel_timestamp.monotonic = 0ULL;
1369 kernel_timestamp.realtime = 0ULL;
1370
1371 } else {
1372 /* Running as user instance */
1373 arg_running_as = SYSTEMD_USER;
1374 log_set_target(LOG_TARGET_AUTO);
1375 log_open();
1376
1377 /* clear the kernel timestamp,
1378 * because we are not PID 1 */
1379 kernel_timestamp.monotonic = 0ULL;
1380 kernel_timestamp.realtime = 0ULL;
1381 }
1382
1383 /* Initialize default unit */
1384 r = set_default_unit(SPECIAL_DEFAULT_TARGET);
1385 if (r < 0) {
1386 log_error("Failed to set default unit %s: %s", SPECIAL_DEFAULT_TARGET, strerror(-r));
1387 goto finish;
1388 }
1389
1390 r = initialize_join_controllers();
1391 if (r < 0)
1392 goto finish;
1393
1394 /* Mount /proc, /sys and friends, so that /proc/cmdline and
1395 * /proc/$PID/fd is available. */
1396 if (getpid() == 1) {
1397 r = mount_setup(loaded_policy);
1398 if (r < 0)
1399 goto finish;
1400 }
1401
1402 /* Reset all signal handlers. */
1403 assert_se(reset_all_signal_handlers() == 0);
1404
1405 ignore_signals(SIGNALS_IGNORE, -1);
1406
1407 if (parse_config_file() < 0)
1408 goto finish;
1409
1410 if (arg_running_as == SYSTEMD_SYSTEM)
1411 if (parse_proc_cmdline() < 0)
1412 goto finish;
1413
1414 log_parse_environment();
1415
1416 if (parse_argv(argc, argv) < 0)
1417 goto finish;
1418
1419 if (arg_action == ACTION_TEST &&
1420 geteuid() == 0) {
1421 log_error("Don't run test mode as root.");
1422 goto finish;
1423 }
1424
1425 if (arg_running_as == SYSTEMD_USER &&
1426 arg_action == ACTION_RUN &&
1427 sd_booted() <= 0) {
1428 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
1429 goto finish;
1430 }
1431
1432 if (arg_running_as == SYSTEMD_SYSTEM &&
1433 arg_action == ACTION_RUN &&
1434 running_in_chroot() > 0) {
1435 log_error("Cannot be run in a chroot() environment.");
1436 goto finish;
1437 }
1438
1439 if (arg_action == ACTION_HELP) {
1440 retval = help();
1441 goto finish;
1442 } else if (arg_action == ACTION_VERSION) {
1443 retval = version();
1444 goto finish;
1445 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
1446 unit_dump_config_items(stdout);
1447 retval = EXIT_SUCCESS;
1448 goto finish;
1449 } else if (arg_action == ACTION_DONE) {
1450 retval = EXIT_SUCCESS;
1451 goto finish;
1452 }
1453
1454 if (arg_running_as == SYSTEMD_USER &&
1455 !getenv("XDG_RUNTIME_DIR")) {
1456 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
1457 goto finish;
1458 }
1459
1460 assert_se(arg_action == ACTION_RUN || arg_action == ACTION_TEST);
1461
1462 /* Close logging fds, in order not to confuse fdset below */
1463 log_close();
1464
1465 /* Remember open file descriptors for later deserialization */
1466 r = fdset_new_fill(&fds);
1467 if (r < 0) {
1468 log_error("Failed to allocate fd set: %s", strerror(-r));
1469 goto finish;
1470 } else
1471 fdset_cloexec(fds, true);
1472
1473 if (arg_serialization)
1474 assert_se(fdset_remove(fds, fileno(arg_serialization)) >= 0);
1475
1476 if (arg_running_as == SYSTEMD_SYSTEM)
1477 /* Become a session leader if we aren't one yet. */
1478 setsid();
1479
1480 /* Move out of the way, so that we won't block unmounts */
1481 assert_se(chdir("/") == 0);
1482
1483 /* Reset the console, but only if this is really init and we
1484 * are freshly booted */
1485 if (arg_running_as == SYSTEMD_SYSTEM && arg_action == ACTION_RUN)
1486 console_setup(getpid() == 1 && !skip_setup);
1487
1488 /* Open the logging devices, if possible and necessary */
1489 log_open();
1490
1491 /* Make sure we leave a core dump without panicing the
1492 * kernel. */
1493 if (getpid() == 1) {
1494 install_crash_handler();
1495
1496 r = mount_cgroup_controllers(arg_join_controllers);
1497 if (r < 0)
1498 goto finish;
1499 }
1500
1501 if (arg_running_as == SYSTEMD_SYSTEM) {
1502 const char *virtualization = NULL;
1503
1504 log_info(PACKAGE_STRING " running in system mode. (" SYSTEMD_FEATURES ")");
1505
1506 detect_virtualization(&virtualization);
1507 if (virtualization)
1508 log_info("Detected virtualization '%s'.", virtualization);
1509
1510 if (in_initrd())
1511 log_info("Running in initial RAM disk.");
1512
1513 } else {
1514 _cleanup_free_ char *t = uid_to_name(getuid());
1515 log_debug(PACKAGE_STRING " running in user mode for user "PID_FMT"/%s. (" SYSTEMD_FEATURES ")",
1516 getuid(), t);
1517 }
1518
1519 if (arg_running_as == SYSTEMD_SYSTEM && !skip_setup) {
1520 if (arg_show_status > 0 || plymouth_running())
1521 status_welcome();
1522
1523 #ifdef HAVE_KMOD
1524 if (detect_container(NULL) <= 0)
1525 kmod_setup();
1526 #endif
1527 hostname_setup();
1528 machine_id_setup();
1529 loopback_setup();
1530
1531 test_mtab();
1532 test_usr();
1533 test_cgroups();
1534 }
1535
1536 if (arg_running_as == SYSTEMD_SYSTEM && arg_runtime_watchdog > 0)
1537 watchdog_set_timeout(&arg_runtime_watchdog);
1538
1539 if (arg_timer_slack_nsec != (nsec_t) -1)
1540 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1541 log_error("Failed to adjust timer slack: %m");
1542
1543 if (arg_capability_bounding_set_drop) {
1544 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set_drop);
1545 if (r < 0) {
1546 log_error("Failed to drop capability bounding set of usermode helpers: %s", strerror(-r));
1547 goto finish;
1548 }
1549 r = capability_bounding_set_drop(arg_capability_bounding_set_drop, true);
1550 if (r < 0) {
1551 log_error("Failed to drop capability bounding set: %s", strerror(-r));
1552 goto finish;
1553 }
1554 }
1555
1556 if (arg_syscall_archs) {
1557 r = enforce_syscall_archs(arg_syscall_archs);
1558 if (r < 0)
1559 goto finish;
1560 }
1561
1562 if (arg_running_as == SYSTEMD_USER) {
1563 /* Become reaper of our children */
1564 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0) {
1565 log_warning("Failed to make us a subreaper: %m");
1566 if (errno == EINVAL)
1567 log_info("Perhaps the kernel version is too old (< 3.4?)");
1568 }
1569 }
1570
1571 if (arg_running_as == SYSTEMD_SYSTEM)
1572 bump_rlimit_nofile(&saved_rlimit_nofile);
1573
1574 r = manager_new(arg_running_as, &m);
1575 if (r < 0) {
1576 log_error("Failed to allocate manager object: %s", strerror(-r));
1577 goto finish;
1578 }
1579
1580 m->confirm_spawn = arg_confirm_spawn;
1581 m->default_std_output = arg_default_std_output;
1582 m->default_std_error = arg_default_std_error;
1583 m->default_restart_usec = arg_default_restart_usec;
1584 m->default_timeout_start_usec = arg_default_timeout_start_usec;
1585 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
1586 m->default_start_limit_interval = arg_default_start_limit_interval;
1587 m->default_start_limit_burst = arg_default_start_limit_burst;
1588 m->runtime_watchdog = arg_runtime_watchdog;
1589 m->shutdown_watchdog = arg_shutdown_watchdog;
1590 m->userspace_timestamp = userspace_timestamp;
1591 m->kernel_timestamp = kernel_timestamp;
1592 m->initrd_timestamp = initrd_timestamp;
1593 m->security_start_timestamp = security_start_timestamp;
1594 m->security_finish_timestamp = security_finish_timestamp;
1595
1596 manager_set_default_rlimits(m, arg_default_rlimit);
1597
1598 if (arg_default_environment)
1599 manager_environment_add(m, NULL, arg_default_environment);
1600
1601 if (arg_show_status == SHOW_STATUS_UNSET)
1602 arg_show_status = SHOW_STATUS_YES;
1603 manager_set_show_status(m, arg_show_status);
1604
1605 /* Remember whether we should queue the default job */
1606 queue_default_job = !arg_serialization || arg_switched_root;
1607
1608 before_startup = now(CLOCK_MONOTONIC);
1609
1610 r = manager_startup(m, arg_serialization, fds);
1611 if (r < 0)
1612 log_error("Failed to fully start up daemon: %s", strerror(-r));
1613
1614 /* This will close all file descriptors that were opened, but
1615 * not claimed by any unit. */
1616 fdset_free(fds);
1617 fds = NULL;
1618
1619 if (arg_serialization) {
1620 fclose(arg_serialization);
1621 arg_serialization = NULL;
1622 }
1623
1624 if (queue_default_job) {
1625 _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL;
1626 Unit *target = NULL;
1627 Job *default_unit_job;
1628
1629 log_debug("Activating default unit: %s", arg_default_unit);
1630
1631 r = manager_load_unit(m, arg_default_unit, NULL, &error, &target);
1632 if (r < 0)
1633 log_error("Failed to load default target: %s", bus_error_message(&error, r));
1634 else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND)
1635 log_error("Failed to load default target: %s", strerror(-target->load_error));
1636 else if (target->load_state == UNIT_MASKED)
1637 log_error("Default target masked.");
1638
1639 if (!target || target->load_state != UNIT_LOADED) {
1640 log_info("Trying to load rescue target...");
1641
1642 r = manager_load_unit(m, SPECIAL_RESCUE_TARGET, NULL, &error, &target);
1643 if (r < 0) {
1644 log_error("Failed to load rescue target: %s", bus_error_message(&error, r));
1645 goto finish;
1646 } else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND) {
1647 log_error("Failed to load rescue target: %s", strerror(-target->load_error));
1648 goto finish;
1649 } else if (target->load_state == UNIT_MASKED) {
1650 log_error("Rescue target masked.");
1651 goto finish;
1652 }
1653 }
1654
1655 assert(target->load_state == UNIT_LOADED);
1656
1657 if (arg_action == ACTION_TEST) {
1658 printf("-> By units:\n");
1659 manager_dump_units(m, stdout, "\t");
1660 }
1661
1662 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, false, &error, &default_unit_job);
1663 if (r == -EPERM) {
1664 log_debug("Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1665
1666 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, false, &error, &default_unit_job);
1667 if (r < 0) {
1668 log_error("Failed to start default target: %s", bus_error_message(&error, r));
1669 goto finish;
1670 }
1671 } else if (r < 0) {
1672 log_error("Failed to isolate default target: %s", bus_error_message(&error, r));
1673 goto finish;
1674 }
1675
1676 m->default_unit_job_id = default_unit_job->id;
1677
1678 after_startup = now(CLOCK_MONOTONIC);
1679 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
1680 "Loaded units and determined initial transaction in %s.",
1681 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 0));
1682
1683 if (arg_action == ACTION_TEST) {
1684 printf("-> By jobs:\n");
1685 manager_dump_jobs(m, stdout, "\t");
1686 retval = EXIT_SUCCESS;
1687 goto finish;
1688 }
1689 }
1690
1691 for (;;) {
1692 r = manager_loop(m);
1693 if (r < 0) {
1694 log_error("Failed to run mainloop: %s", strerror(-r));
1695 goto finish;
1696 }
1697
1698 switch (m->exit_code) {
1699
1700 case MANAGER_EXIT:
1701 retval = EXIT_SUCCESS;
1702 log_debug("Exit.");
1703 goto finish;
1704
1705 case MANAGER_RELOAD:
1706 log_info("Reloading.");
1707 r = manager_reload(m);
1708 if (r < 0)
1709 log_error("Failed to reload: %s", strerror(-r));
1710 break;
1711
1712 case MANAGER_REEXECUTE:
1713
1714 if (prepare_reexecute(m, &arg_serialization, &fds, false) < 0)
1715 goto finish;
1716
1717 reexecute = true;
1718 log_notice("Reexecuting.");
1719 goto finish;
1720
1721 case MANAGER_SWITCH_ROOT:
1722 /* Steal the switch root parameters */
1723 switch_root_dir = m->switch_root;
1724 switch_root_init = m->switch_root_init;
1725 m->switch_root = m->switch_root_init = NULL;
1726
1727 if (!switch_root_init)
1728 if (prepare_reexecute(m, &arg_serialization, &fds, true) < 0)
1729 goto finish;
1730
1731 reexecute = true;
1732 log_notice("Switching root.");
1733 goto finish;
1734
1735 case MANAGER_REBOOT:
1736 case MANAGER_POWEROFF:
1737 case MANAGER_HALT:
1738 case MANAGER_KEXEC: {
1739 static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
1740 [MANAGER_REBOOT] = "reboot",
1741 [MANAGER_POWEROFF] = "poweroff",
1742 [MANAGER_HALT] = "halt",
1743 [MANAGER_KEXEC] = "kexec"
1744 };
1745
1746 assert_se(shutdown_verb = table[m->exit_code]);
1747 arm_reboot_watchdog = m->exit_code == MANAGER_REBOOT;
1748
1749 log_notice("Shutting down.");
1750 goto finish;
1751 }
1752
1753 default:
1754 assert_not_reached("Unknown exit code.");
1755 }
1756 }
1757
1758 finish:
1759 if (m) {
1760 manager_free(m);
1761 m = NULL;
1762 }
1763
1764 for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++) {
1765 free(arg_default_rlimit[j]);
1766 arg_default_rlimit[j] = NULL;
1767 }
1768
1769 free(arg_default_unit);
1770 arg_default_unit = NULL;
1771
1772 free_join_controllers();
1773
1774 strv_free(arg_default_environment);
1775 arg_default_environment = NULL;
1776
1777 set_free(arg_syscall_archs);
1778 arg_syscall_archs = NULL;
1779
1780 label_finish();
1781
1782 if (reexecute) {
1783 const char **args;
1784 unsigned i, args_size;
1785
1786 /* Close and disarm the watchdog, so that the new
1787 * instance can reinitialize it, but doesn't get
1788 * rebooted while we do that */
1789 watchdog_close(true);
1790
1791 /* Reset the RLIMIT_NOFILE to the kernel default, so
1792 * that the new systemd can pass the kernel default to
1793 * its child processes */
1794 if (saved_rlimit_nofile.rlim_cur > 0)
1795 setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile);
1796
1797 if (switch_root_dir) {
1798 /* Kill all remaining processes from the
1799 * initrd, but don't wait for them, so that we
1800 * can handle the SIGCHLD for them after
1801 * deserializing. */
1802 broadcast_signal(SIGTERM, false, true);
1803
1804 /* And switch root */
1805 r = switch_root(switch_root_dir);
1806 if (r < 0)
1807 log_error("Failed to switch root, ignoring: %s", strerror(-r));
1808 }
1809
1810 args_size = MAX(6, argc+1);
1811 args = newa(const char*, args_size);
1812
1813 if (!switch_root_init) {
1814 char sfd[16];
1815
1816 /* First try to spawn ourselves with the right
1817 * path, and with full serialization. We do
1818 * this only if the user didn't specify an
1819 * explicit init to spawn. */
1820
1821 assert(arg_serialization);
1822 assert(fds);
1823
1824 snprintf(sfd, sizeof(sfd), "%i", fileno(arg_serialization));
1825 char_array_0(sfd);
1826
1827 i = 0;
1828 args[i++] = SYSTEMD_BINARY_PATH;
1829 if (switch_root_dir)
1830 args[i++] = "--switched-root";
1831 args[i++] = arg_running_as == SYSTEMD_SYSTEM ? "--system" : "--user";
1832 args[i++] = "--deserialize";
1833 args[i++] = sfd;
1834 args[i++] = NULL;
1835
1836 /* do not pass along the environment we inherit from the kernel or initrd */
1837 if (switch_root_dir)
1838 clearenv();
1839
1840 assert(i <= args_size);
1841 execv(args[0], (char* const*) args);
1842 }
1843
1844 /* Try the fallback, if there is any, without any
1845 * serialization. We pass the original argv[] and
1846 * envp[]. (Well, modulo the ordering changes due to
1847 * getopt() in argv[], and some cleanups in envp[],
1848 * but let's hope that doesn't matter.) */
1849
1850 if (arg_serialization) {
1851 fclose(arg_serialization);
1852 arg_serialization = NULL;
1853 }
1854
1855 if (fds) {
1856 fdset_free(fds);
1857 fds = NULL;
1858 }
1859
1860 /* Reopen the console */
1861 make_console_stdio();
1862
1863 for (j = 1, i = 1; j < (unsigned) argc; j++)
1864 args[i++] = argv[j];
1865 args[i++] = NULL;
1866 assert(i <= args_size);
1867
1868 if (switch_root_init) {
1869 args[0] = switch_root_init;
1870 execv(args[0], (char* const*) args);
1871 log_warning("Failed to execute configured init, trying fallback: %m");
1872 }
1873
1874 args[0] = "/sbin/init";
1875 execv(args[0], (char* const*) args);
1876
1877 if (errno == ENOENT) {
1878 log_warning("No /sbin/init, trying fallback");
1879
1880 args[0] = "/bin/sh";
1881 args[1] = NULL;
1882 execv(args[0], (char* const*) args);
1883 log_error("Failed to execute /bin/sh, giving up: %m");
1884 } else
1885 log_warning("Failed to execute /sbin/init, giving up: %m");
1886 }
1887
1888 if (arg_serialization) {
1889 fclose(arg_serialization);
1890 arg_serialization = NULL;
1891 }
1892
1893 if (fds) {
1894 fdset_free(fds);
1895 fds = NULL;
1896 }
1897
1898 #ifdef HAVE_VALGRIND_VALGRIND_H
1899 /* If we are PID 1 and running under valgrind, then let's exit
1900 * here explicitly. valgrind will only generate nice output on
1901 * exit(), not on exec(), hence let's do the former not the
1902 * latter here. */
1903 if (getpid() == 1 && RUNNING_ON_VALGRIND)
1904 return 0;
1905 #endif
1906
1907 if (shutdown_verb) {
1908 const char * command_line[] = {
1909 SYSTEMD_SHUTDOWN_BINARY_PATH,
1910 shutdown_verb,
1911 NULL
1912 };
1913 _cleanup_strv_free_ char **env_block = NULL;
1914 env_block = strv_copy(environ);
1915
1916 if (arm_reboot_watchdog && arg_shutdown_watchdog > 0) {
1917 char *e;
1918
1919 /* If we reboot let's set the shutdown
1920 * watchdog and tell the shutdown binary to
1921 * repeatedly ping it */
1922 watchdog_set_timeout(&arg_shutdown_watchdog);
1923 watchdog_close(false);
1924
1925 /* Tell the binary how often to ping, ignore failure */
1926 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
1927 strv_push(&env_block, e);
1928 } else
1929 watchdog_close(true);
1930
1931 /* Avoid the creation of new processes forked by the
1932 * kernel; at this point, we will not listen to the
1933 * signals anyway */
1934 if (detect_container(NULL) <= 0)
1935 cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
1936
1937 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
1938 log_error("Failed to execute shutdown binary, freezing: %m");
1939 }
1940
1941 if (getpid() == 1)
1942 freeze();
1943
1944 return retval;
1945 }
Unnamed repository; edit this file 'description' to name the repository.