]> git.ipfire.org Git - thirdparty/systemd.git/blame_incremental - src/core/main.c
projects / thirdparty / systemd.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
util: add realloc_multiply() helper
[thirdparty/systemd.git] / src / core / main.c
... / ...
CommitLineData
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#ifdef HAVE_SECCOMP
39#include <seccomp.h>
40#endif
41
42#include "sd-daemon.h"
43#include "sd-messages.h"
44#include "sd-bus.h"
45#include "manager.h"
46#include "log.h"
47#include "load-fragment.h"
48#include "fdset.h"
49#include "special.h"
50#include "conf-parser.h"
51#include "missing.h"
52#include "label.h"
53#include "build.h"
54#include "strv.h"
55#include "def.h"
56#include "virt.h"
57#include "architecture.h"
58#include "watchdog.h"
59#include "path-util.h"
60#include "switch-root.h"
61#include "capability.h"
62#include "killall.h"
63#include "env-util.h"
64#include "clock-util.h"
65#include "fileio.h"
66#include "dbus-manager.h"
67#include "bus-error.h"
68#include "bus-util.h"
69
70#include "mount-setup.h"
71#include "loopback-setup.h"
72#include "hostname-setup.h"
73#include "machine-id-setup.h"
74#include "selinux-setup.h"
75#include "ima-setup.h"
76#include "smack-setup.h"
77#ifdef HAVE_KMOD
78#include "kmod-setup.h"
79#endif
80
81static enum {
82 ACTION_RUN,
83 ACTION_HELP,
84 ACTION_VERSION,
85 ACTION_TEST,
86 ACTION_DUMP_CONFIGURATION_ITEMS,
87 ACTION_DONE
88} arg_action = ACTION_RUN;
89static char *arg_default_unit = NULL;
90static SystemdRunningAs arg_running_as = _SYSTEMD_RUNNING_AS_INVALID;
91static bool arg_dump_core = true;
92static bool arg_crash_shell = false;
93static int arg_crash_chvt = -1;
94static bool arg_confirm_spawn = false;
95static ShowStatus arg_show_status = _SHOW_STATUS_UNSET;
96static bool arg_switched_root = false;
97static char ***arg_join_controllers = NULL;
98static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
99static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
100static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
101static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
102static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
103static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
104static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
105static usec_t arg_runtime_watchdog = 0;
106static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
107static char **arg_default_environment = NULL;
108static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {};
109static uint64_t arg_capability_bounding_set_drop = 0;
110static nsec_t arg_timer_slack_nsec = (nsec_t) -1;
111static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
112static Set* arg_syscall_archs = NULL;
113static FILE* arg_serialization = NULL;
114static bool arg_default_cpu_accounting = false;
115static bool arg_default_blockio_accounting = false;
116static bool arg_default_memory_accounting = false;
117
118static void nop_handler(int sig) {}
119
120noreturn static void crash(int sig) {
121
122 if (getpid() != 1)
123 /* Pass this on immediately, if this is not PID 1 */
124 raise(sig);
125 else if (!arg_dump_core)
126 log_error("Caught <%s>, not dumping core.", signal_to_string(sig));
127 else {
128 struct sigaction sa = {
129 .sa_handler = nop_handler,
130 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
131 };
132 pid_t pid;
133
134 /* We want to wait for the core process, hence let's enable SIGCHLD */
135 sigaction(SIGCHLD, &sa, NULL);
136
137 pid = fork();
138 if (pid < 0)
139 log_error("Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
140
141 else if (pid == 0) {
142 struct rlimit rl = {};
143
144 /* Enable default signal handler for core dump */
145 zero(sa);
146 sa.sa_handler = SIG_DFL;
147 sigaction(sig, &sa, NULL);
148
149 /* Don't limit the core dump size */
150 rl.rlim_cur = RLIM_INFINITY;
151 rl.rlim_max = RLIM_INFINITY;
152 setrlimit(RLIMIT_CORE, &rl);
153
154 /* Just to be sure... */
155 chdir("/");
156
157 /* Raise the signal again */
158 raise(sig);
159
160 assert_not_reached("We shouldn't be here...");
161 _exit(1);
162
163 } else {
164 siginfo_t status;
165 int r;
166
167 /* Order things nicely. */
168 r = wait_for_terminate(pid, &status);
169 if (r < 0)
170 log_error("Caught <%s>, waitpid() failed: %s", signal_to_string(sig), strerror(-r));
171 else if (status.si_code != CLD_DUMPED)
172 log_error("Caught <%s>, core dump failed.", signal_to_string(sig));
173 else
174 log_error("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
175 }
176 }
177
178 if (arg_crash_chvt)
179 chvt(arg_crash_chvt);
180
181 if (arg_crash_shell) {
182 struct sigaction sa = {
183 .sa_handler = SIG_IGN,
184 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
185 };
186 pid_t pid;
187
188 log_info("Executing crash shell in 10s...");
189 sleep(10);
190
191 /* Let the kernel reap children for us */
192 assert_se(sigaction(SIGCHLD, &sa, NULL) == 0);
193
194 pid = fork();
195 if (pid < 0)
196 log_error("Failed to fork off crash shell: %m");
197 else if (pid == 0) {
198 make_console_stdio();
199 execl("/bin/sh", "/bin/sh", NULL);
200
201 log_error("execl() failed: %m");
202 _exit(1);
203 }
204
205 log_info("Successfully spawned crash shell as pid "PID_FMT".", pid);
206 }
207
208 log_info("Freezing execution.");
209 freeze();
210}
211
212static void install_crash_handler(void) {
213 struct sigaction sa = {
214 .sa_handler = crash,
215 .sa_flags = SA_NODEFER,
216 };
217
218 sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
219}
220
221static int console_setup(bool do_reset) {
222 int tty_fd, r;
223
224 /* If we are init, we connect stdin/stdout/stderr to /dev/null
225 * and make sure we don't have a controlling tty. */
226
227 release_terminal();
228
229 if (!do_reset)
230 return 0;
231
232 tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
233 if (tty_fd < 0) {
234 log_error("Failed to open /dev/console: %s", strerror(-tty_fd));
235 return -tty_fd;
236 }
237
238 /* We don't want to force text mode.
239 * plymouth may be showing pictures already from initrd. */
240 r = reset_terminal_fd(tty_fd, false);
241 if (r < 0)
242 log_error("Failed to reset /dev/console: %s", strerror(-r));
243
244 safe_close(tty_fd);
245 return r;
246}
247
248static int set_default_unit(const char *u) {
249 char *c;
250
251 assert(u);
252
253 c = strdup(u);
254 if (!c)
255 return -ENOMEM;
256
257 free(arg_default_unit);
258 arg_default_unit = c;
259
260 return 0;
261}
262
263static int parse_proc_cmdline_item(const char *key, const char *value) {
264
265 static const char * const rlmap[] = {
266 "emergency", SPECIAL_EMERGENCY_TARGET,
267 "-b", SPECIAL_EMERGENCY_TARGET,
268 "single", SPECIAL_RESCUE_TARGET,
269 "-s", SPECIAL_RESCUE_TARGET,
270 "s", SPECIAL_RESCUE_TARGET,
271 "S", SPECIAL_RESCUE_TARGET,
272 "1", SPECIAL_RESCUE_TARGET,
273 "2", SPECIAL_RUNLEVEL2_TARGET,
274 "3", SPECIAL_RUNLEVEL3_TARGET,
275 "4", SPECIAL_RUNLEVEL4_TARGET,
276 "5", SPECIAL_RUNLEVEL5_TARGET,
277 };
278 int r;
279
280 assert(key);
281
282 if (streq(key, "systemd.unit") && value) {
283
284 if (!in_initrd())
285 return set_default_unit(value);
286
287 } else if (streq(key, "rd.systemd.unit") && value) {
288
289 return set_default_unit(value);
290
291 } else if (streq(key, "systemd.log_target") && value) {
292
293 if (log_set_target_from_string(value) < 0)
294 log_warning("Failed to parse log target %s. Ignoring.", value);
295
296 } else if (streq(key, "systemd.log_level") && value) {
297
298 if (log_set_max_level_from_string(value) < 0)
299 log_warning("Failed to parse log level %s. Ignoring.", value);
300
301 } else if (streq(key, "systemd.log_color") && value) {
302
303 if (log_show_color_from_string(value) < 0)
304 log_warning("Failed to parse log color setting %s. Ignoring.", value);
305
306 } else if (streq(key, "systemd.log_location") && value) {
307
308 if (log_show_location_from_string(value) < 0)
309 log_warning("Failed to parse log location setting %s. Ignoring.", value);
310
311 } else if (streq(key, "systemd.dump_core") && value) {
312
313 r = parse_boolean(value);
314 if (r < 0)
315 log_warning("Failed to parse dump core switch %s. Ignoring.", value);
316 else
317 arg_dump_core = r;
318
319 } else if (streq(key, "systemd.crash_shell") && value) {
320
321 r = parse_boolean(value);
322 if (r < 0)
323 log_warning("Failed to parse crash shell switch %s. Ignoring.", value);
324 else
325 arg_crash_shell = r;
326
327 } else if (streq(key, "systemd.crash_chvt") && value) {
328
329 if (safe_atoi(value, &r) < 0)
330 log_warning("Failed to parse crash chvt switch %s. Ignoring.", value);
331 else
332 arg_crash_chvt = r;
333
334 } else if (streq(key, "systemd.confirm_spawn") && value) {
335
336 r = parse_boolean(value);
337 if (r < 0)
338 log_warning("Failed to parse confirm spawn switch %s. Ignoring.", value);
339 else
340 arg_confirm_spawn = r;
341
342 } else if (streq(key, "systemd.show_status") && value) {
343
344 r = parse_show_status(value, &arg_show_status);
345 if (r < 0)
346 log_warning("Failed to parse show status switch %s. Ignoring.", value);
347
348 } else if (streq(key, "systemd.default_standard_output") && value) {
349
350 r = exec_output_from_string(value);
351 if (r < 0)
352 log_warning("Failed to parse default standard output switch %s. Ignoring.", value);
353 else
354 arg_default_std_output = r;
355
356 } else if (streq(key, "systemd.default_standard_error") && value) {
357
358 r = exec_output_from_string(value);
359 if (r < 0)
360 log_warning("Failed to parse default standard error switch %s. Ignoring.", value);
361 else
362 arg_default_std_error = r;
363
364 } else if (streq(key, "systemd.setenv") && value) {
365
366 if (env_assignment_is_valid(value)) {
367 char **env;
368
369 env = strv_env_set(arg_default_environment, value);
370 if (env)
371 arg_default_environment = env;
372 else
373 log_warning("Setting environment variable '%s' failed, ignoring: %s", value, strerror(ENOMEM));
374 } else
375 log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
376
377 } else if (!streq(key, "systemd.restore_state") &&
378 !streq(key, "systemd.gpt_auto") &&
379 (startswith(key, "systemd.") || startswith(key, "rd.systemd."))) {
380
381 const char *c;
382
383 /* Ignore systemd.journald.xyz and friends */
384 c = key;
385 if (startswith(c, "rd."))
386 c += 3;
387 if (startswith(c, "systemd."))
388 c += 8;
389 if (c[strcspn(c, ".=")] != '.') {
390
391 log_warning("Unknown kernel switch %s. Ignoring.", key);
392
393 log_info("Supported kernel switches:\n"
394 "systemd.unit=UNIT Default unit to start\n"
395 "rd.systemd.unit=UNIT Default unit to start when run in initrd\n"
396 "systemd.dump_core=0|1 Dump core on crash\n"
397 "systemd.crash_shell=0|1 Run shell on crash\n"
398 "systemd.crash_chvt=N Change to VT #N on crash\n"
399 "systemd.confirm_spawn=0|1 Confirm every process spawn\n"
400 "systemd.show_status=0|1|auto Show status updates on the console during bootup\n"
401 "systemd.log_target=console|kmsg|journal|journal-or-kmsg|syslog|syslog-or-kmsg|null\n"
402 " Log target\n"
403 "systemd.log_level=LEVEL Log level\n"
404 "systemd.log_color=0|1 Highlight important log messages\n"
405 "systemd.log_location=0|1 Include code location in log messages\n"
406 "systemd.default_standard_output=null|tty|syslog|syslog+console|kmsg|kmsg+console|journal|journal+console\n"
407 " Set default log output for services\n"
408 "systemd.default_standard_error=null|tty|syslog|syslog+console|kmsg|kmsg+console|journal|journal+console\n"
409 " Set default log error output for services\n"
410 "systemd.setenv=ASSIGNMENT Set an environment variable for all spawned processes\n"
411 "systemd.restore_state=0|1 Restore backlight/rfkill state at boot\n");
412 }
413
414 } else if (streq(key, "quiet") && !value) {
415 if (arg_show_status == _SHOW_STATUS_UNSET)
416 arg_show_status = SHOW_STATUS_AUTO;
417
418 } else if (streq(key, "debug") && !value) {
419 log_set_max_level(LOG_DEBUG);
420 if (detect_container(NULL) > 0)
421 log_set_target(LOG_TARGET_CONSOLE);
422
423 } else if (!in_initrd() && !value) {
424 unsigned i;
425
426 /* SysV compatibility */
427 for (i = 0; i < ELEMENTSOF(rlmap); i += 2)
428 if (streq(key, rlmap[i]))
429 return set_default_unit(rlmap[i+1]);
430 }
431
432 return 0;
433}
434
435#define DEFINE_SETTER(name, func, descr) \
436 static int name(const char *unit, \
437 const char *filename, \
438 unsigned line, \
439 const char *section, \
440 unsigned section_line, \
441 const char *lvalue, \
442 int ltype, \
443 const char *rvalue, \
444 void *data, \
445 void *userdata) { \
446 \
447 int r; \
448 \
449 assert(filename); \
450 assert(lvalue); \
451 assert(rvalue); \
452 \
453 r = func(rvalue); \
454 if (r < 0) \
455 log_syntax(unit, LOG_ERR, filename, line, -r, \
456 "Invalid " descr "'%s': %s", \
457 rvalue, strerror(-r)); \
458 \
459 return 0; \
460 }
461
462DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level")
463DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target")
464DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" )
465DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location")
466
467static int config_parse_cpu_affinity2(
468 const char *unit,
469 const char *filename,
470 unsigned line,
471 const char *section,
472 unsigned section_line,
473 const char *lvalue,
474 int ltype,
475 const char *rvalue,
476 void *data,
477 void *userdata) {
478
479 char *w;
480 size_t l;
481 char *state;
482 cpu_set_t *c = NULL;
483 unsigned ncpus = 0;
484
485 assert(filename);
486 assert(lvalue);
487 assert(rvalue);
488
489 FOREACH_WORD_QUOTED(w, l, rvalue, state) {
490 char *t;
491 int r;
492 unsigned cpu;
493
494 if (!(t = strndup(w, l)))
495 return log_oom();
496
497 r = safe_atou(t, &cpu);
498 free(t);
499
500 if (!c)
501 if (!(c = cpu_set_malloc(&ncpus)))
502 return log_oom();
503
504 if (r < 0 || cpu >= ncpus) {
505 log_syntax(unit, LOG_ERR, filename, line, -r,
506 "Failed to parse CPU affinity '%s'", rvalue);
507 CPU_FREE(c);
508 return -EBADMSG;
509 }
510
511 CPU_SET_S(cpu, CPU_ALLOC_SIZE(ncpus), c);
512 }
513
514 if (c) {
515 if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
516 log_warning_unit(unit, "Failed to set CPU affinity: %m");
517
518 CPU_FREE(c);
519 }
520
521 return 0;
522}
523
524static int config_parse_show_status(
525 const char* unit,
526 const char *filename,
527 unsigned line,
528 const char *section,
529 unsigned section_line,
530 const char *lvalue,
531 int ltype,
532 const char *rvalue,
533 void *data,
534 void *userdata) {
535
536 int k;
537 ShowStatus *b = data;
538
539 assert(filename);
540 assert(lvalue);
541 assert(rvalue);
542 assert(data);
543
544 k = parse_show_status(rvalue, b);
545 if (k < 0) {
546 log_syntax(unit, LOG_ERR, filename, line, -k,
547 "Failed to parse show status setting, ignoring: %s", rvalue);
548 return 0;
549 }
550
551 return 0;
552}
553
554static void strv_free_free(char ***l) {
555 char ***i;
556
557 if (!l)
558 return;
559
560 for (i = l; *i; i++)
561 strv_free(*i);
562
563 free(l);
564}
565
566static void free_join_controllers(void) {
567 strv_free_free(arg_join_controllers);
568 arg_join_controllers = NULL;
569}
570
571static int config_parse_join_controllers(const char *unit,
572 const char *filename,
573 unsigned line,
574 const char *section,
575 unsigned section_line,
576 const char *lvalue,
577 int ltype,
578 const char *rvalue,
579 void *data,
580 void *userdata) {
581
582 unsigned n = 0;
583 char *state, *w;
584 size_t length;
585
586 assert(filename);
587 assert(lvalue);
588 assert(rvalue);
589
590 free_join_controllers();
591
592 FOREACH_WORD_QUOTED(w, length, rvalue, state) {
593 char *s, **l;
594
595 s = strndup(w, length);
596 if (!s)
597 return log_oom();
598
599 l = strv_split(s, ",");
600 free(s);
601
602 strv_uniq(l);
603
604 if (strv_length(l) <= 1) {
605 strv_free(l);
606 continue;
607 }
608
609 if (!arg_join_controllers) {
610 arg_join_controllers = new(char**, 2);
611 if (!arg_join_controllers) {
612 strv_free(l);
613 return log_oom();
614 }
615
616 arg_join_controllers[0] = l;
617 arg_join_controllers[1] = NULL;
618
619 n = 1;
620 } else {
621 char ***a;
622 char ***t;
623
624 t = new0(char**, n+2);
625 if (!t) {
626 strv_free(l);
627 return log_oom();
628 }
629
630 n = 0;
631
632 for (a = arg_join_controllers; *a; a++) {
633
634 if (strv_overlap(*a, l)) {
635 if (strv_extend_strv(&l, *a) < 0) {
636 strv_free(l);
637 strv_free_free(t);
638 return log_oom();
639 }
640
641 } else {
642 char **c;
643
644 c = strv_copy(*a);
645 if (!c) {
646 strv_free(l);
647 strv_free_free(t);
648 return log_oom();
649 }
650
651 t[n++] = c;
652 }
653 }
654
655 t[n++] = strv_uniq(l);
656
657 strv_free_free(arg_join_controllers);
658 arg_join_controllers = t;
659 }
660 }
661
662 return 0;
663}
664
665static int parse_config_file(void) {
666
667 const ConfigTableItem items[] = {
668 { "Manager", "LogLevel", config_parse_level2, 0, NULL },
669 { "Manager", "LogTarget", config_parse_target, 0, NULL },
670 { "Manager", "LogColor", config_parse_color, 0, NULL },
671 { "Manager", "LogLocation", config_parse_location, 0, NULL },
672 { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core },
673 { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell },
674 { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status },
675 { "Manager", "CrashChVT", config_parse_int, 0, &arg_crash_chvt },
676 { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL },
677 { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers },
678 { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog },
679 { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog },
680 { "Manager", "CapabilityBoundingSet", config_parse_bounding_set, 0, &arg_capability_bounding_set_drop },
681#ifdef HAVE_SECCOMP
682 { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs },
683#endif
684 { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec },
685 { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec },
686 { "Manager", "DefaultStandardOutput", config_parse_output, 0, &arg_default_std_output },
687 { "Manager", "DefaultStandardError", config_parse_output, 0, &arg_default_std_error },
688 { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec },
689 { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec },
690 { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec },
691 { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval },
692 { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst },
693 { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment },
694 { "Manager", "DefaultLimitCPU", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CPU] },
695 { "Manager", "DefaultLimitFSIZE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_FSIZE] },
696 { "Manager", "DefaultLimitDATA", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_DATA] },
697 { "Manager", "DefaultLimitSTACK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_STACK] },
698 { "Manager", "DefaultLimitCORE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CORE] },
699 { "Manager", "DefaultLimitRSS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RSS] },
700 { "Manager", "DefaultLimitNOFILE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NOFILE] },
701 { "Manager", "DefaultLimitAS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_AS] },
702 { "Manager", "DefaultLimitNPROC", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NPROC] },
703 { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MEMLOCK] },
704 { "Manager", "DefaultLimitLOCKS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_LOCKS] },
705 { "Manager", "DefaultLimitSIGPENDING", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_SIGPENDING] },
706 { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MSGQUEUE] },
707 { "Manager", "DefaultLimitNICE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NICE] },
708 { "Manager", "DefaultLimitRTPRIO", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTPRIO] },
709 { "Manager", "DefaultLimitRTTIME", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTTIME] },
710 { "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_default_cpu_accounting },
711 { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting },
712 { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting },
713 {}
714 };
715
716 _cleanup_fclose_ FILE *f;
717 const char *fn;
718 int r;
719
720 fn = arg_running_as == SYSTEMD_SYSTEM ? PKGSYSCONFDIR "/system.conf" : PKGSYSCONFDIR "/user.conf";
721 f = fopen(fn, "re");
722 if (!f) {
723 if (errno == ENOENT)
724 return 0;
725
726 log_warning("Failed to open configuration file '%s': %m", fn);
727 return 0;
728 }
729
730 r = config_parse(NULL, fn, f, "Manager\0", config_item_table_lookup, (void*) items, false, false, NULL);
731 if (r < 0)
732 log_warning("Failed to parse configuration file: %s", strerror(-r));
733
734 return 0;
735}
736
737static int parse_argv(int argc, char *argv[]) {
738
739 enum {
740 ARG_LOG_LEVEL = 0x100,
741 ARG_LOG_TARGET,
742 ARG_LOG_COLOR,
743 ARG_LOG_LOCATION,
744 ARG_UNIT,
745 ARG_SYSTEM,
746 ARG_USER,
747 ARG_TEST,
748 ARG_VERSION,
749 ARG_DUMP_CONFIGURATION_ITEMS,
750 ARG_DUMP_CORE,
751 ARG_CRASH_SHELL,
752 ARG_CONFIRM_SPAWN,
753 ARG_SHOW_STATUS,
754 ARG_DESERIALIZE,
755 ARG_SWITCHED_ROOT,
756 ARG_DEFAULT_STD_OUTPUT,
757 ARG_DEFAULT_STD_ERROR
758 };
759
760 static const struct option options[] = {
761 { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
762 { "log-target", required_argument, NULL, ARG_LOG_TARGET },
763 { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
764 { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
765 { "unit", required_argument, NULL, ARG_UNIT },
766 { "system", no_argument, NULL, ARG_SYSTEM },
767 { "user", no_argument, NULL, ARG_USER },
768 { "test", no_argument, NULL, ARG_TEST },
769 { "help", no_argument, NULL, 'h' },
770 { "version", no_argument, NULL, ARG_VERSION },
771 { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS },
772 { "dump-core", optional_argument, NULL, ARG_DUMP_CORE },
773 { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL },
774 { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN },
775 { "show-status", optional_argument, NULL, ARG_SHOW_STATUS },
776 { "deserialize", required_argument, NULL, ARG_DESERIALIZE },
777 { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT },
778 { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, },
779 { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, },
780 {}
781 };
782
783 int c, r;
784
785 assert(argc >= 1);
786 assert(argv);
787
788 if (getpid() == 1)
789 opterr = 0;
790
791 while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
792
793 switch (c) {
794
795 case ARG_LOG_LEVEL:
796 r = log_set_max_level_from_string(optarg);
797 if (r < 0) {
798 log_error("Failed to parse log level %s.", optarg);
799 return r;
800 }
801
802 break;
803
804 case ARG_LOG_TARGET:
805 r = log_set_target_from_string(optarg);
806 if (r < 0) {
807 log_error("Failed to parse log target %s.", optarg);
808 return r;
809 }
810
811 break;
812
813 case ARG_LOG_COLOR:
814
815 if (optarg) {
816 r = log_show_color_from_string(optarg);
817 if (r < 0) {
818 log_error("Failed to parse log color setting %s.", optarg);
819 return r;
820 }
821 } else
822 log_show_color(true);
823
824 break;
825
826 case ARG_LOG_LOCATION:
827 if (optarg) {
828 r = log_show_location_from_string(optarg);
829 if (r < 0) {
830 log_error("Failed to parse log location setting %s.", optarg);
831 return r;
832 }
833 } else
834 log_show_location(true);
835
836 break;
837
838 case ARG_DEFAULT_STD_OUTPUT:
839 r = exec_output_from_string(optarg);
840 if (r < 0) {
841 log_error("Failed to parse default standard output setting %s.", optarg);
842 return r;
843 } else
844 arg_default_std_output = r;
845 break;
846
847 case ARG_DEFAULT_STD_ERROR:
848 r = exec_output_from_string(optarg);
849 if (r < 0) {
850 log_error("Failed to parse default standard error output setting %s.", optarg);
851 return r;
852 } else
853 arg_default_std_error = r;
854 break;
855
856 case ARG_UNIT:
857
858 r = set_default_unit(optarg);
859 if (r < 0) {
860 log_error("Failed to set default unit %s: %s", optarg, strerror(-r));
861 return r;
862 }
863
864 break;
865
866 case ARG_SYSTEM:
867 arg_running_as = SYSTEMD_SYSTEM;
868 break;
869
870 case ARG_USER:
871 arg_running_as = SYSTEMD_USER;
872 break;
873
874 case ARG_TEST:
875 arg_action = ACTION_TEST;
876 break;
877
878 case ARG_VERSION:
879 arg_action = ACTION_VERSION;
880 break;
881
882 case ARG_DUMP_CONFIGURATION_ITEMS:
883 arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
884 break;
885
886 case ARG_DUMP_CORE:
887 r = optarg ? parse_boolean(optarg) : 1;
888 if (r < 0) {
889 log_error("Failed to parse dump core boolean %s.", optarg);
890 return r;
891 }
892 arg_dump_core = r;
893 break;
894
895 case ARG_CRASH_SHELL:
896 r = optarg ? parse_boolean(optarg) : 1;
897 if (r < 0) {
898 log_error("Failed to parse crash shell boolean %s.", optarg);
899 return r;
900 }
901 arg_crash_shell = r;
902 break;
903
904 case ARG_CONFIRM_SPAWN:
905 r = optarg ? parse_boolean(optarg) : 1;
906 if (r < 0) {
907 log_error("Failed to parse confirm spawn boolean %s.", optarg);
908 return r;
909 }
910 arg_confirm_spawn = r;
911 break;
912
913 case ARG_SHOW_STATUS:
914 if (optarg) {
915 r = parse_show_status(optarg, &arg_show_status);
916 if (r < 0) {
917 log_error("Failed to parse show status boolean %s.", optarg);
918 return r;
919 }
920 } else
921 arg_show_status = SHOW_STATUS_YES;
922 break;
923
924 case ARG_DESERIALIZE: {
925 int fd;
926 FILE *f;
927
928 r = safe_atoi(optarg, &fd);
929 if (r < 0 || fd < 0) {
930 log_error("Failed to parse deserialize option %s.", optarg);
931 return r < 0 ? r : -EINVAL;
932 }
933
934 fd_cloexec(fd, true);
935
936 f = fdopen(fd, "r");
937 if (!f) {
938 log_error("Failed to open serialization fd: %m");
939 return -errno;
940 }
941
942 if (arg_serialization)
943 fclose(arg_serialization);
944
945 arg_serialization = f;
946
947 break;
948 }
949
950 case ARG_SWITCHED_ROOT:
951 arg_switched_root = true;
952 break;
953
954 case 'h':
955 arg_action = ACTION_HELP;
956 break;
957
958 case 'D':
959 log_set_max_level(LOG_DEBUG);
960 break;
961
962 case 'b':
963 case 's':
964 case 'z':
965 /* Just to eat away the sysvinit kernel
966 * cmdline args without getopt() error
967 * messages that we'll parse in
968 * parse_proc_cmdline_word() or ignore. */
969
970 case '?':
971 default:
972 if (getpid() != 1) {
973 log_error("Unknown option code %c", c);
974 return -EINVAL;
975 }
976
977 break;
978 }
979
980 if (optind < argc && getpid() != 1) {
981 /* Hmm, when we aren't run as init system
982 * let's complain about excess arguments */
983
984 log_error("Excess arguments.");
985 return -EINVAL;
986 }
987
988 if (detect_container(NULL) > 0) {
989 char **a;
990
991 /* All /proc/cmdline arguments the kernel didn't
992 * understand it passed to us. We're not really
993 * interested in that usually since /proc/cmdline is
994 * more interesting and complete. With one exception:
995 * if we are run in a container /proc/cmdline is not
996 * relevant for the container, hence we rely on argv[]
997 * instead. */
998
999 for (a = argv; a < argv + argc; a++) {
1000 _cleanup_free_ char *w;
1001 char *value;
1002
1003 w = strdup(*a);
1004 if (!w)
1005 return log_oom();
1006
1007 value = strchr(w, '=');
1008 if (value)
1009 *(value++) = 0;
1010
1011 r = parse_proc_cmdline_item(w, value);
1012 if (r < 0) {
1013 log_error("Failed on cmdline argument %s: %s", *a, strerror(-r));
1014 return r;
1015 }
1016 }
1017 }
1018
1019 return 0;
1020}
1021
1022static int help(void) {
1023
1024 printf("%s [OPTIONS...]\n\n"
1025 "Starts up and maintains the system or user services.\n\n"
1026 " -h --help Show this help\n"
1027 " --test Determine startup sequence, dump it and exit\n"
1028 " --dump-configuration-items Dump understood unit configuration items\n"
1029 " --unit=UNIT Set default unit\n"
1030 " --system Run a system instance, even if PID != 1\n"
1031 " --user Run a user instance\n"
1032 " --dump-core[=0|1] Dump core on crash\n"
1033 " --crash-shell[=0|1] Run shell on crash\n"
1034 " --confirm-spawn[=0|1] Ask for confirmation when spawning processes\n"
1035 " --show-status[=0|1] Show status updates on the console during bootup\n"
1036 " --log-target=TARGET Set log target (console, journal, syslog, kmsg, journal-or-kmsg, syslog-or-kmsg, null)\n"
1037 " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1038 " --log-color[=0|1] Highlight important log messages\n"
1039 " --log-location[=0|1] Include code location in log messages\n"
1040 " --default-standard-output= Set default standard output for services\n"
1041 " --default-standard-error= Set default standard error output for services\n",
1042 program_invocation_short_name);
1043
1044 return 0;
1045}
1046
1047static int version(void) {
1048 puts(PACKAGE_STRING);
1049 puts(SYSTEMD_FEATURES);
1050
1051 return 0;
1052}
1053
1054static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
1055 FILE *f = NULL;
1056 FDSet *fds = NULL;
1057 int r;
1058
1059 assert(m);
1060 assert(_f);
1061 assert(_fds);
1062
1063 r = manager_open_serialization(m, &f);
1064 if (r < 0) {
1065 log_error("Failed to create serialization file: %s", strerror(-r));
1066 goto fail;
1067 }
1068
1069 /* Make sure nothing is really destructed when we shut down */
1070 m->n_reloading ++;
1071 bus_manager_send_reloading(m, true);
1072
1073 fds = fdset_new();
1074 if (!fds) {
1075 r = -ENOMEM;
1076 log_error("Failed to allocate fd set: %s", strerror(-r));
1077 goto fail;
1078 }
1079
1080 r = manager_serialize(m, f, fds, switching_root);
1081 if (r < 0) {
1082 log_error("Failed to serialize state: %s", strerror(-r));
1083 goto fail;
1084 }
1085
1086 if (fseeko(f, 0, SEEK_SET) < 0) {
1087 log_error("Failed to rewind serialization fd: %m");
1088 goto fail;
1089 }
1090
1091 r = fd_cloexec(fileno(f), false);
1092 if (r < 0) {
1093 log_error("Failed to disable O_CLOEXEC for serialization: %s", strerror(-r));
1094 goto fail;
1095 }
1096
1097 r = fdset_cloexec(fds, false);
1098 if (r < 0) {
1099 log_error("Failed to disable O_CLOEXEC for serialization fds: %s", strerror(-r));
1100 goto fail;
1101 }
1102
1103 *_f = f;
1104 *_fds = fds;
1105
1106 return 0;
1107
1108fail:
1109 fdset_free(fds);
1110
1111 if (f)
1112 fclose(f);
1113
1114 return r;
1115}
1116
1117static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1118 struct rlimit nl;
1119 int r;
1120
1121 assert(saved_rlimit);
1122
1123 /* Save the original RLIMIT_NOFILE so that we can reset it
1124 * later when transitioning from the initrd to the main
1125 * systemd or suchlike. */
1126 if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0) {
1127 log_error("Reading RLIMIT_NOFILE failed: %m");
1128 return -errno;
1129 }
1130
1131 /* Make sure forked processes get the default kernel setting */
1132 if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1133 struct rlimit *rl;
1134
1135 rl = newdup(struct rlimit, saved_rlimit, 1);
1136 if (!rl)
1137 return log_oom();
1138
1139 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1140 }
1141
1142 /* Bump up the resource limit for ourselves substantially */
1143 nl.rlim_cur = nl.rlim_max = 64*1024;
1144 r = setrlimit_closest(RLIMIT_NOFILE, &nl);
1145 if (r < 0) {
1146 log_error("Setting RLIMIT_NOFILE failed: %s", strerror(-r));
1147 return r;
1148 }
1149
1150 return 0;
1151}
1152
1153static void test_mtab(void) {
1154
1155 static const char ok[] =
1156 "/proc/self/mounts\0"
1157 "/proc/mounts\0"
1158 "../proc/self/mounts\0"
1159 "../proc/mounts\0";
1160
1161 _cleanup_free_ char *p = NULL;
1162 int r;
1163
1164 /* Check that /etc/mtab is a symlink to the right place or
1165 * non-existing. But certainly not a file, or a symlink to
1166 * some weird place... */
1167
1168 r = readlink_malloc("/etc/mtab", &p);
1169 if (r == -ENOENT)
1170 return;
1171 if (r >= 0 && nulstr_contains(ok, p))
1172 return;
1173
1174 log_warning("/etc/mtab is not a symlink or not pointing to /proc/self/mounts. "
1175 "This is not supported anymore. "
1176 "Please make sure to replace this file by a symlink to avoid incorrect or misleading mount(8) output.");
1177}
1178
1179static void test_usr(void) {
1180
1181 /* Check that /usr is not a separate fs */
1182
1183 if (dir_is_empty("/usr") <= 0)
1184 return;
1185
1186 log_warning("/usr appears to be on its own filesytem and is not already mounted. This is not a supported setup. "
1187 "Some things will probably break (sometimes even silently) in mysterious ways. "
1188 "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1189}
1190
1191static int initialize_join_controllers(void) {
1192 /* By default, mount "cpu" + "cpuacct" together, and "net_cls"
1193 * + "net_prio". We'd like to add "cpuset" to the mix, but
1194 * "cpuset" does't really work for groups with no initialized
1195 * attributes. */
1196
1197 arg_join_controllers = new(char**, 3);
1198 if (!arg_join_controllers)
1199 return -ENOMEM;
1200
1201 arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL);
1202 arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL);
1203 arg_join_controllers[2] = NULL;
1204
1205 if (!arg_join_controllers[0] || !arg_join_controllers[1]) {
1206 free_join_controllers();
1207 return -ENOMEM;
1208 }
1209
1210 return 0;
1211}
1212
1213static int enforce_syscall_archs(Set *archs) {
1214#ifdef HAVE_SECCOMP
1215 scmp_filter_ctx *seccomp;
1216 Iterator i;
1217 void *id;
1218 int r;
1219
1220 seccomp = seccomp_init(SCMP_ACT_ALLOW);
1221 if (!seccomp)
1222 return log_oom();
1223
1224 SET_FOREACH(id, arg_syscall_archs, i) {
1225 r = seccomp_arch_add(seccomp, PTR_TO_UINT32(id) - 1);
1226 if (r == -EEXIST)
1227 continue;
1228 if (r < 0) {
1229 log_error("Failed to add architecture to seccomp: %s", strerror(-r));
1230 goto finish;
1231 }
1232 }
1233
1234 r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
1235 if (r < 0) {
1236 log_error("Failed to unset NO_NEW_PRIVS: %s", strerror(-r));
1237 goto finish;
1238 }
1239
1240 r = seccomp_load(seccomp);
1241 if (r < 0)
1242 log_error("Failed to add install architecture seccomp: %s", strerror(-r));
1243
1244finish:
1245 seccomp_release(seccomp);
1246 return r;
1247#else
1248 return 0;
1249#endif
1250}
1251
1252static int status_welcome(void) {
1253 _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1254 int r;
1255
1256 r = parse_env_file("/etc/os-release", NEWLINE,
1257 "PRETTY_NAME", &pretty_name,
1258 "ANSI_COLOR", &ansi_color,
1259 NULL);
1260 if (r == -ENOENT) {
1261 r = parse_env_file("/usr/lib/os-release", NEWLINE,
1262 "PRETTY_NAME", &pretty_name,
1263 "ANSI_COLOR", &ansi_color,
1264 NULL);
1265 }
1266
1267 if (r < 0 && r != -ENOENT)
1268 log_warning("Failed to read os-release file: %s", strerror(-r));
1269
1270 return status_printf(NULL, false, false,
1271 "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1272 isempty(ansi_color) ? "1" : ansi_color,
1273 isempty(pretty_name) ? "Linux" : pretty_name);
1274}
1275
1276static int write_container_id(void) {
1277 const char *c;
1278
1279 c = getenv("container");
1280 if (isempty(c))
1281 return 0;
1282
1283 return write_string_file("/run/systemd/container", c);
1284}
1285
1286int main(int argc, char *argv[]) {
1287 Manager *m = NULL;
1288 int r, retval = EXIT_FAILURE;
1289 usec_t before_startup, after_startup;
1290 char timespan[FORMAT_TIMESPAN_MAX];
1291 FDSet *fds = NULL;
1292 bool reexecute = false;
1293 const char *shutdown_verb = NULL;
1294 dual_timestamp initrd_timestamp = { 0ULL, 0ULL };
1295 dual_timestamp userspace_timestamp = { 0ULL, 0ULL };
1296 dual_timestamp kernel_timestamp = { 0ULL, 0ULL };
1297 dual_timestamp security_start_timestamp = { 0ULL, 0ULL };
1298 dual_timestamp security_finish_timestamp = { 0ULL, 0ULL };
1299 static char systemd[] = "systemd";
1300 bool skip_setup = false;
1301 unsigned j;
1302 bool loaded_policy = false;
1303 bool arm_reboot_watchdog = false;
1304 bool queue_default_job = false;
1305 char *switch_root_dir = NULL, *switch_root_init = NULL;
1306 static struct rlimit saved_rlimit_nofile = { 0, 0 };
1307
1308#ifdef HAVE_SYSV_COMPAT
1309 if (getpid() != 1 && strstr(program_invocation_short_name, "init")) {
1310 /* This is compatibility support for SysV, where
1311 * calling init as a user is identical to telinit. */
1312
1313 errno = -ENOENT;
1314 execv(SYSTEMCTL_BINARY_PATH, argv);
1315 log_error("Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1316 return 1;
1317 }
1318#endif
1319
1320 dual_timestamp_from_monotonic(&kernel_timestamp, 0);
1321 dual_timestamp_get(&userspace_timestamp);
1322
1323 /* Determine if this is a reexecution or normal bootup. We do
1324 * the full command line parsing much later, so let's just
1325 * have a quick peek here. */
1326 if (strv_find(argv+1, "--deserialize"))
1327 skip_setup = true;
1328
1329 /* If we have switched root, do all the special setup
1330 * things */
1331 if (strv_find(argv+1, "--switched-root"))
1332 skip_setup = false;
1333
1334 /* If we get started via the /sbin/init symlink then we are
1335 called 'init'. After a subsequent reexecution we are then
1336 called 'systemd'. That is confusing, hence let's call us
1337 systemd right-away. */
1338 program_invocation_short_name = systemd;
1339 prctl(PR_SET_NAME, systemd);
1340
1341 saved_argv = argv;
1342 saved_argc = argc;
1343
1344 log_show_color(isatty(STDERR_FILENO) > 0);
1345
1346 /* Disable the umask logic */
1347 if (getpid() == 1)
1348 umask(0);
1349
1350 if (getpid() == 1 && detect_container(NULL) <= 0) {
1351
1352 /* Running outside of a container as PID 1 */
1353 arg_running_as = SYSTEMD_SYSTEM;
1354 make_null_stdio();
1355 log_set_target(LOG_TARGET_KMSG);
1356 log_open();
1357
1358 if (in_initrd())
1359 initrd_timestamp = userspace_timestamp;
1360
1361 if (!skip_setup) {
1362 mount_setup_early();
1363 dual_timestamp_get(&security_start_timestamp);
1364 if (selinux_setup(&loaded_policy) < 0)
1365 goto finish;
1366 if (ima_setup() < 0)
1367 goto finish;
1368 if (smack_setup(&loaded_policy) < 0)
1369 goto finish;
1370 dual_timestamp_get(&security_finish_timestamp);
1371 }
1372
1373 if (label_init(NULL) < 0)
1374 goto finish;
1375
1376 if (!skip_setup) {
1377 if (clock_is_localtime() > 0) {
1378 int min;
1379
1380 /* The first-time call to settimeofday() does a time warp in the kernel */
1381 r = clock_set_timezone(&min);
1382 if (r < 0)
1383 log_error("Failed to apply local time delta, ignoring: %s", strerror(-r));
1384 else
1385 log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1386 } else if (!in_initrd()) {
1387 /*
1388 * Do dummy first-time call to seal the kernel's time warp magic
1389 *
1390 * Do not call this this from inside the initrd. The initrd might not
1391 * carry /etc/adjtime with LOCAL, but the real system could be set up
1392 * that way. In such case, we need to delay the time-warp or the sealing
1393 * until we reach the real system.
1394 */
1395 clock_reset_timezone();
1396
1397 /* Tell the kernel our timezone */
1398 r = clock_set_timezone(NULL);
1399 if (r < 0)
1400 log_error("Failed to set the kernel's timezone, ignoring: %s", strerror(-r));
1401 }
1402 }
1403
1404 /* Set the default for later on, but don't actually
1405 * open the logs like this for now. Note that if we
1406 * are transitioning from the initrd there might still
1407 * be journal fd open, and we shouldn't attempt
1408 * opening that before we parsed /proc/cmdline which
1409 * might redirect output elsewhere. */
1410 log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
1411
1412 } else if (getpid() == 1) {
1413 /* Running inside a container, as PID 1 */
1414 arg_running_as = SYSTEMD_SYSTEM;
1415 log_set_target(LOG_TARGET_CONSOLE);
1416 log_close_console(); /* force reopen of /dev/console */
1417 log_open();
1418
1419 /* For the later on, see above... */
1420 log_set_target(LOG_TARGET_JOURNAL);
1421
1422 /* clear the kernel timestamp,
1423 * because we are in a container */
1424 kernel_timestamp.monotonic = 0ULL;
1425 kernel_timestamp.realtime = 0ULL;
1426
1427 } else {
1428 /* Running as user instance */
1429 arg_running_as = SYSTEMD_USER;
1430 log_set_target(LOG_TARGET_AUTO);
1431 log_open();
1432
1433 /* clear the kernel timestamp,
1434 * because we are not PID 1 */
1435 kernel_timestamp.monotonic = 0ULL;
1436 kernel_timestamp.realtime = 0ULL;
1437 }
1438
1439 /* Initialize default unit */
1440 r = set_default_unit(SPECIAL_DEFAULT_TARGET);
1441 if (r < 0) {
1442 log_error("Failed to set default unit %s: %s", SPECIAL_DEFAULT_TARGET, strerror(-r));
1443 goto finish;
1444 }
1445
1446 r = initialize_join_controllers();
1447 if (r < 0)
1448 goto finish;
1449
1450 /* Mount /proc, /sys and friends, so that /proc/cmdline and
1451 * /proc/$PID/fd is available. */
1452 if (getpid() == 1) {
1453 r = mount_setup(loaded_policy);
1454 if (r < 0)
1455 goto finish;
1456 }
1457
1458 /* Reset all signal handlers. */
1459 assert_se(reset_all_signal_handlers() == 0);
1460
1461 ignore_signals(SIGNALS_IGNORE, -1);
1462
1463 if (parse_config_file() < 0)
1464 goto finish;
1465
1466 if (arg_running_as == SYSTEMD_SYSTEM)
1467 if (parse_proc_cmdline(parse_proc_cmdline_item) < 0)
1468 goto finish;
1469
1470 log_parse_environment();
1471
1472 if (parse_argv(argc, argv) < 0)
1473 goto finish;
1474
1475 if (arg_action == ACTION_TEST &&
1476 geteuid() == 0) {
1477 log_error("Don't run test mode as root.");
1478 goto finish;
1479 }
1480
1481 if (arg_running_as == SYSTEMD_USER &&
1482 arg_action == ACTION_RUN &&
1483 sd_booted() <= 0) {
1484 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
1485 goto finish;
1486 }
1487
1488 if (arg_running_as == SYSTEMD_SYSTEM &&
1489 arg_action == ACTION_RUN &&
1490 running_in_chroot() > 0) {
1491 log_error("Cannot be run in a chroot() environment.");
1492 goto finish;
1493 }
1494
1495 if (arg_action == ACTION_HELP) {
1496 retval = help();
1497 goto finish;
1498 } else if (arg_action == ACTION_VERSION) {
1499 retval = version();
1500 goto finish;
1501 } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
1502 unit_dump_config_items(stdout);
1503 retval = EXIT_SUCCESS;
1504 goto finish;
1505 } else if (arg_action == ACTION_DONE) {
1506 retval = EXIT_SUCCESS;
1507 goto finish;
1508 }
1509
1510 if (arg_running_as == SYSTEMD_USER &&
1511 !getenv("XDG_RUNTIME_DIR")) {
1512 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
1513 goto finish;
1514 }
1515
1516 assert_se(arg_action == ACTION_RUN || arg_action == ACTION_TEST);
1517
1518 /* Close logging fds, in order not to confuse fdset below */
1519 log_close();
1520
1521 /* Remember open file descriptors for later deserialization */
1522 r = fdset_new_fill(&fds);
1523 if (r < 0) {
1524 log_error("Failed to allocate fd set: %s", strerror(-r));
1525 goto finish;
1526 } else
1527 fdset_cloexec(fds, true);
1528
1529 if (arg_serialization)
1530 assert_se(fdset_remove(fds, fileno(arg_serialization)) >= 0);
1531
1532 if (arg_running_as == SYSTEMD_SYSTEM)
1533 /* Become a session leader if we aren't one yet. */
1534 setsid();
1535
1536 /* Move out of the way, so that we won't block unmounts */
1537 assert_se(chdir("/") == 0);
1538
1539 /* Reset the console, but only if this is really init and we
1540 * are freshly booted */
1541 if (arg_running_as == SYSTEMD_SYSTEM && arg_action == ACTION_RUN)
1542 console_setup(getpid() == 1 && !skip_setup);
1543
1544 /* Open the logging devices, if possible and necessary */
1545 log_open();
1546
1547 if (arg_show_status == _SHOW_STATUS_UNSET)
1548 arg_show_status = SHOW_STATUS_YES;
1549
1550 /* Make sure we leave a core dump without panicing the
1551 * kernel. */
1552 if (getpid() == 1) {
1553 install_crash_handler();
1554
1555 r = mount_cgroup_controllers(arg_join_controllers);
1556 if (r < 0)
1557 goto finish;
1558 }
1559
1560 if (arg_running_as == SYSTEMD_SYSTEM) {
1561 const char *virtualization = NULL;
1562
1563 log_info(PACKAGE_STRING " running in system mode. (" SYSTEMD_FEATURES ")");
1564
1565 detect_virtualization(&virtualization);
1566 if (virtualization)
1567 log_info("Detected virtualization '%s'.", virtualization);
1568
1569 write_container_id();
1570
1571 log_info("Detected architecture '%s'.", architecture_to_string(uname_architecture()));
1572
1573 if (in_initrd())
1574 log_info("Running in initial RAM disk.");
1575
1576 } else {
1577 _cleanup_free_ char *t = uid_to_name(getuid());
1578 log_debug(PACKAGE_STRING " running in user mode for user "PID_FMT"/%s. (" SYSTEMD_FEATURES ")",
1579 getuid(), t);
1580 }
1581
1582 if (arg_running_as == SYSTEMD_SYSTEM && !skip_setup) {
1583 if (arg_show_status > 0 || plymouth_running())
1584 status_welcome();
1585
1586#ifdef HAVE_KMOD
1587 if (detect_container(NULL) <= 0)
1588 kmod_setup();
1589#endif
1590 hostname_setup();
1591 machine_id_setup("");
1592 loopback_setup();
1593
1594 test_mtab();
1595 test_usr();
1596 }
1597
1598 if (arg_running_as == SYSTEMD_SYSTEM && arg_runtime_watchdog > 0)
1599 watchdog_set_timeout(&arg_runtime_watchdog);
1600
1601 if (arg_timer_slack_nsec != (nsec_t) -1)
1602 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1603 log_error("Failed to adjust timer slack: %m");
1604
1605 if (arg_capability_bounding_set_drop) {
1606 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set_drop);
1607 if (r < 0) {
1608 log_error("Failed to drop capability bounding set of usermode helpers: %s", strerror(-r));
1609 goto finish;
1610 }
1611 r = capability_bounding_set_drop(arg_capability_bounding_set_drop, true);
1612 if (r < 0) {
1613 log_error("Failed to drop capability bounding set: %s", strerror(-r));
1614 goto finish;
1615 }
1616 }
1617
1618 if (arg_syscall_archs) {
1619 r = enforce_syscall_archs(arg_syscall_archs);
1620 if (r < 0)
1621 goto finish;
1622 }
1623
1624 if (arg_running_as == SYSTEMD_USER) {
1625 /* Become reaper of our children */
1626 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0) {
1627 log_warning("Failed to make us a subreaper: %m");
1628 if (errno == EINVAL)
1629 log_info("Perhaps the kernel version is too old (< 3.4?)");
1630 }
1631 }
1632
1633 if (arg_running_as == SYSTEMD_SYSTEM)
1634 bump_rlimit_nofile(&saved_rlimit_nofile);
1635
1636 r = manager_new(arg_running_as, &m);
1637 if (r < 0) {
1638 log_error("Failed to allocate manager object: %s", strerror(-r));
1639 goto finish;
1640 }
1641
1642 m->confirm_spawn = arg_confirm_spawn;
1643 m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
1644 m->default_std_output = arg_default_std_output;
1645 m->default_std_error = arg_default_std_error;
1646 m->default_restart_usec = arg_default_restart_usec;
1647 m->default_timeout_start_usec = arg_default_timeout_start_usec;
1648 m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
1649 m->default_start_limit_interval = arg_default_start_limit_interval;
1650 m->default_start_limit_burst = arg_default_start_limit_burst;
1651 m->default_cpu_accounting = arg_default_cpu_accounting;
1652 m->default_blockio_accounting = arg_default_blockio_accounting;
1653 m->default_memory_accounting = arg_default_memory_accounting;
1654 m->runtime_watchdog = arg_runtime_watchdog;
1655 m->shutdown_watchdog = arg_shutdown_watchdog;
1656 m->userspace_timestamp = userspace_timestamp;
1657 m->kernel_timestamp = kernel_timestamp;
1658 m->initrd_timestamp = initrd_timestamp;
1659 m->security_start_timestamp = security_start_timestamp;
1660 m->security_finish_timestamp = security_finish_timestamp;
1661
1662 manager_set_default_rlimits(m, arg_default_rlimit);
1663 manager_environment_add(m, NULL, arg_default_environment);
1664 manager_set_show_status(m, arg_show_status);
1665
1666 /* Remember whether we should queue the default job */
1667 queue_default_job = !arg_serialization || arg_switched_root;
1668
1669 before_startup = now(CLOCK_MONOTONIC);
1670
1671 r = manager_startup(m, arg_serialization, fds);
1672 if (r < 0)
1673 log_error("Failed to fully start up daemon: %s", strerror(-r));
1674
1675 /* This will close all file descriptors that were opened, but
1676 * not claimed by any unit. */
1677 fdset_free(fds);
1678 fds = NULL;
1679
1680 if (arg_serialization) {
1681 fclose(arg_serialization);
1682 arg_serialization = NULL;
1683 }
1684
1685 if (queue_default_job) {
1686 _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL;
1687 Unit *target = NULL;
1688 Job *default_unit_job;
1689
1690 log_debug("Activating default unit: %s", arg_default_unit);
1691
1692 r = manager_load_unit(m, arg_default_unit, NULL, &error, &target);
1693 if (r < 0)
1694 log_error("Failed to load default target: %s", bus_error_message(&error, r));
1695 else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND)
1696 log_error("Failed to load default target: %s", strerror(-target->load_error));
1697 else if (target->load_state == UNIT_MASKED)
1698 log_error("Default target masked.");
1699
1700 if (!target || target->load_state != UNIT_LOADED) {
1701 log_info("Trying to load rescue target...");
1702
1703 r = manager_load_unit(m, SPECIAL_RESCUE_TARGET, NULL, &error, &target);
1704 if (r < 0) {
1705 log_error("Failed to load rescue target: %s", bus_error_message(&error, r));
1706 goto finish;
1707 } else if (target->load_state == UNIT_ERROR || target->load_state == UNIT_NOT_FOUND) {
1708 log_error("Failed to load rescue target: %s", strerror(-target->load_error));
1709 goto finish;
1710 } else if (target->load_state == UNIT_MASKED) {
1711 log_error("Rescue target masked.");
1712 goto finish;
1713 }
1714 }
1715
1716 assert(target->load_state == UNIT_LOADED);
1717
1718 if (arg_action == ACTION_TEST) {
1719 printf("-> By units:\n");
1720 manager_dump_units(m, stdout, "\t");
1721 }
1722
1723 r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, false, &error, &default_unit_job);
1724 if (r == -EPERM) {
1725 log_debug("Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1726
1727 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, false, &error, &default_unit_job);
1728 if (r < 0) {
1729 log_error("Failed to start default target: %s", bus_error_message(&error, r));
1730 goto finish;
1731 }
1732 } else if (r < 0) {
1733 log_error("Failed to isolate default target: %s", bus_error_message(&error, r));
1734 goto finish;
1735 }
1736
1737 m->default_unit_job_id = default_unit_job->id;
1738
1739 after_startup = now(CLOCK_MONOTONIC);
1740 log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
1741 "Loaded units and determined initial transaction in %s.",
1742 format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 0));
1743
1744 if (arg_action == ACTION_TEST) {
1745 printf("-> By jobs:\n");
1746 manager_dump_jobs(m, stdout, "\t");
1747 retval = EXIT_SUCCESS;
1748 goto finish;
1749 }
1750 }
1751
1752 for (;;) {
1753 r = manager_loop(m);
1754 if (r < 0) {
1755 log_error("Failed to run mainloop: %s", strerror(-r));
1756 goto finish;
1757 }
1758
1759 switch (m->exit_code) {
1760
1761 case MANAGER_EXIT:
1762 retval = EXIT_SUCCESS;
1763 log_debug("Exit.");
1764 goto finish;
1765
1766 case MANAGER_RELOAD:
1767 log_info("Reloading.");
1768 r = manager_reload(m);
1769 if (r < 0)
1770 log_error("Failed to reload: %s", strerror(-r));
1771 break;
1772
1773 case MANAGER_REEXECUTE:
1774
1775 if (prepare_reexecute(m, &arg_serialization, &fds, false) < 0)
1776 goto finish;
1777
1778 reexecute = true;
1779 log_notice("Reexecuting.");
1780 goto finish;
1781
1782 case MANAGER_SWITCH_ROOT:
1783 /* Steal the switch root parameters */
1784 switch_root_dir = m->switch_root;
1785 switch_root_init = m->switch_root_init;
1786 m->switch_root = m->switch_root_init = NULL;
1787
1788 if (!switch_root_init)
1789 if (prepare_reexecute(m, &arg_serialization, &fds, true) < 0)
1790 goto finish;
1791
1792 reexecute = true;
1793 log_notice("Switching root.");
1794 goto finish;
1795
1796 case MANAGER_REBOOT:
1797 case MANAGER_POWEROFF:
1798 case MANAGER_HALT:
1799 case MANAGER_KEXEC: {
1800 static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
1801 [MANAGER_REBOOT] = "reboot",
1802 [MANAGER_POWEROFF] = "poweroff",
1803 [MANAGER_HALT] = "halt",
1804 [MANAGER_KEXEC] = "kexec"
1805 };
1806
1807 assert_se(shutdown_verb = table[m->exit_code]);
1808 arm_reboot_watchdog = m->exit_code == MANAGER_REBOOT;
1809
1810 log_notice("Shutting down.");
1811 goto finish;
1812 }
1813
1814 default:
1815 assert_not_reached("Unknown exit code.");
1816 }
1817 }
1818
1819finish:
1820 if (m) {
1821 manager_free(m);
1822 m = NULL;
1823 }
1824
1825 for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++) {
1826 free(arg_default_rlimit[j]);
1827 arg_default_rlimit[j] = NULL;
1828 }
1829
1830 free(arg_default_unit);
1831 arg_default_unit = NULL;
1832
1833 free_join_controllers();
1834
1835 strv_free(arg_default_environment);
1836 arg_default_environment = NULL;
1837
1838 set_free(arg_syscall_archs);
1839 arg_syscall_archs = NULL;
1840
1841 label_finish();
1842
1843 if (reexecute) {
1844 const char **args;
1845 unsigned i, args_size;
1846
1847 /* Close and disarm the watchdog, so that the new
1848 * instance can reinitialize it, but doesn't get
1849 * rebooted while we do that */
1850 watchdog_close(true);
1851
1852 /* Reset the RLIMIT_NOFILE to the kernel default, so
1853 * that the new systemd can pass the kernel default to
1854 * its child processes */
1855 if (saved_rlimit_nofile.rlim_cur > 0)
1856 setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile);
1857
1858 if (switch_root_dir) {
1859 /* Kill all remaining processes from the
1860 * initrd, but don't wait for them, so that we
1861 * can handle the SIGCHLD for them after
1862 * deserializing. */
1863 broadcast_signal(SIGTERM, false, true);
1864
1865 /* And switch root */
1866 r = switch_root(switch_root_dir);
1867 if (r < 0)
1868 log_error("Failed to switch root, ignoring: %s", strerror(-r));
1869 }
1870
1871 args_size = MAX(6, argc+1);
1872 args = newa(const char*, args_size);
1873
1874 if (!switch_root_init) {
1875 char sfd[16];
1876
1877 /* First try to spawn ourselves with the right
1878 * path, and with full serialization. We do
1879 * this only if the user didn't specify an
1880 * explicit init to spawn. */
1881
1882 assert(arg_serialization);
1883 assert(fds);
1884
1885 snprintf(sfd, sizeof(sfd), "%i", fileno(arg_serialization));
1886 char_array_0(sfd);
1887
1888 i = 0;
1889 args[i++] = SYSTEMD_BINARY_PATH;
1890 if (switch_root_dir)
1891 args[i++] = "--switched-root";
1892 args[i++] = arg_running_as == SYSTEMD_SYSTEM ? "--system" : "--user";
1893 args[i++] = "--deserialize";
1894 args[i++] = sfd;
1895 args[i++] = NULL;
1896
1897 /* do not pass along the environment we inherit from the kernel or initrd */
1898 if (switch_root_dir)
1899 clearenv();
1900
1901 assert(i <= args_size);
1902 execv(args[0], (char* const*) args);
1903 }
1904
1905 /* Try the fallback, if there is any, without any
1906 * serialization. We pass the original argv[] and
1907 * envp[]. (Well, modulo the ordering changes due to
1908 * getopt() in argv[], and some cleanups in envp[],
1909 * but let's hope that doesn't matter.) */
1910
1911 if (arg_serialization) {
1912 fclose(arg_serialization);
1913 arg_serialization = NULL;
1914 }
1915
1916 if (fds) {
1917 fdset_free(fds);
1918 fds = NULL;
1919 }
1920
1921 /* Reopen the console */
1922 make_console_stdio();
1923
1924 for (j = 1, i = 1; j < (unsigned) argc; j++)
1925 args[i++] = argv[j];
1926 args[i++] = NULL;
1927 assert(i <= args_size);
1928
1929 if (switch_root_init) {
1930 args[0] = switch_root_init;
1931 execv(args[0], (char* const*) args);
1932 log_warning("Failed to execute configured init, trying fallback: %m");
1933 }
1934
1935 args[0] = "/sbin/init";
1936 execv(args[0], (char* const*) args);
1937
1938 if (errno == ENOENT) {
1939 log_warning("No /sbin/init, trying fallback");
1940
1941 args[0] = "/bin/sh";
1942 args[1] = NULL;
1943 execv(args[0], (char* const*) args);
1944 log_error("Failed to execute /bin/sh, giving up: %m");
1945 } else
1946 log_warning("Failed to execute /sbin/init, giving up: %m");
1947 }
1948
1949 if (arg_serialization) {
1950 fclose(arg_serialization);
1951 arg_serialization = NULL;
1952 }
1953
1954 if (fds) {
1955 fdset_free(fds);
1956 fds = NULL;
1957 }
1958
1959#ifdef HAVE_VALGRIND_VALGRIND_H
1960 /* If we are PID 1 and running under valgrind, then let's exit
1961 * here explicitly. valgrind will only generate nice output on
1962 * exit(), not on exec(), hence let's do the former not the
1963 * latter here. */
1964 if (getpid() == 1 && RUNNING_ON_VALGRIND)
1965 return 0;
1966#endif
1967
1968 if (shutdown_verb) {
1969 char log_level[DECIMAL_STR_MAX(int) + 1];
1970 const char* command_line[9] = {
1971 SYSTEMD_SHUTDOWN_BINARY_PATH,
1972 shutdown_verb,
1973 "--log-level", log_level,
1974 "--log-target",
1975 };
1976 unsigned pos = 5;
1977 _cleanup_strv_free_ char **env_block = NULL;
1978
1979 assert(command_line[pos] == NULL);
1980 env_block = strv_copy(environ);
1981
1982 snprintf(log_level, sizeof(log_level), "%d", log_get_max_level());
1983
1984 switch (log_get_target()) {
1985 case LOG_TARGET_KMSG:
1986 case LOG_TARGET_JOURNAL_OR_KMSG:
1987 case LOG_TARGET_SYSLOG_OR_KMSG:
1988 command_line[pos++] = "kmsg";
1989 break;
1990
1991 case LOG_TARGET_CONSOLE:
1992 default:
1993 command_line[pos++] = "console";
1994 break;
1995 };
1996
1997 if (log_get_show_color())
1998 command_line[pos++] = "--log-color";
1999
2000 if (log_get_show_location())
2001 command_line[pos++] = "--log-location";
2002
2003 assert(pos < ELEMENTSOF(command_line));
2004
2005 if (arm_reboot_watchdog && arg_shutdown_watchdog > 0) {
2006 char *e;
2007
2008 /* If we reboot let's set the shutdown
2009 * watchdog and tell the shutdown binary to
2010 * repeatedly ping it */
2011 watchdog_set_timeout(&arg_shutdown_watchdog);
2012 watchdog_close(false);
2013
2014 /* Tell the binary how often to ping, ignore failure */
2015 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
2016 strv_push(&env_block, e);
2017 } else
2018 watchdog_close(true);
2019
2020 /* Avoid the creation of new processes forked by the
2021 * kernel; at this point, we will not listen to the
2022 * signals anyway */
2023 if (detect_container(NULL) <= 0)
2024 cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
2025
2026 execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
2027 log_error("Failed to execute shutdown binary, %s: %m",
2028 getpid() == 1 ? "freezing" : "quitting");
2029 }
2030
2031 if (getpid() == 1)
2032 freeze();
2033
2034 return retval;
2035}
Unnamed repository; edit this file 'description' to name the repository.