src/core/main.c

   1 /* SPDX-License-Identifier: LGPL-2.1+ */
   2
   3 #include <errno.h>
   4 #include <fcntl.h>
   5 #include <getopt.h>
   6 #include <signal.h>
   7 #include <stdio.h>
   8 #include <string.h>
   9 #include <sys/mount.h>
  10 #include <sys/prctl.h>
  11 #include <sys/reboot.h>
  12 #include <sys/stat.h>
  13 #include <unistd.h>
  14 #if HAVE_SECCOMP
  15 #include <seccomp.h>
  16 #endif
  17 #if HAVE_VALGRIND_VALGRIND_H
  18 #include <valgrind/valgrind.h>
  19 #endif
  20
  21 #include "sd-bus.h"
  22 #include "sd-daemon.h"
  23 #include "sd-messages.h"
  24
  25 #include "alloc-util.h"
  26 #include "architecture.h"
  27 #include "build.h"
  28 #include "bus-error.h"
  29 #include "bus-util.h"
  30 #include "capability-util.h"
  31 #include "clock-util.h"
  32 #include "conf-parser.h"
  33 #include "cpu-set-util.h"
  34 #include "dbus.h"
  35 #include "dbus-manager.h"
  36 #include "def.h"
  37 #include "emergency-action.h"
  38 #include "env-util.h"
  39 #include "exit-status.h"
  40 #include "fd-util.h"
  41 #include "fdset.h"
  42 #include "fileio.h"
  43 #include "format-util.h"
  44 #include "fs-util.h"
  45 #include "hostname-setup.h"
  46 #include "ima-setup.h"
  47 #include "killall.h"
  48 #include "kmod-setup.h"
  49 #include "load-fragment.h"
  50 #include "log.h"
  51 #include "loopback-setup.h"
  52 #include "machine-id-setup.h"
  53 #include "manager.h"
  54 #include "missing.h"
  55 #include "mount-setup.h"
  56 #include "os-util.h"
  57 #include "pager.h"
  58 #include "parse-util.h"
  59 #include "path-util.h"
  60 #include "proc-cmdline.h"
  61 #include "process-util.h"
  62 #include "raw-clone.h"
  63 #include "rlimit-util.h"
  64 #if HAVE_SECCOMP
  65 #include "seccomp-util.h"
  66 #endif
  67 #include "selinux-setup.h"
  68 #include "selinux-util.h"
  69 #include "signal-util.h"
  70 #include "smack-setup.h"
  71 #include "special.h"
  72 #include "stat-util.h"
  73 #include "stdio-util.h"
  74 #include "strv.h"
  75 #include "switch-root.h"
  76 #include "terminal-util.h"
  77 #include "umask-util.h"
  78 #include "user-util.h"
  79 #include "util.h"
  80 #include "virt.h"
  81 #include "watchdog.h"
  82
  83 static enum {
  84         ACTION_RUN,
  85         ACTION_HELP,
  86         ACTION_VERSION,
  87         ACTION_TEST,
  88         ACTION_DUMP_CONFIGURATION_ITEMS,
  89         ACTION_DUMP_BUS_PROPERTIES,
  90 } arg_action = ACTION_RUN;
  91 static char *arg_default_unit = NULL;
  92 static bool arg_system = false;
  93 static bool arg_dump_core = true;
  94 static int arg_crash_chvt = -1;
  95 static bool arg_crash_shell = false;
  96 static bool arg_crash_reboot = false;
  97 static char *arg_confirm_spawn = NULL;
  98 static ShowStatus arg_show_status = _SHOW_STATUS_INVALID;
  99 static bool arg_switched_root = false;
 100 static bool arg_no_pager = false;
 101 static bool arg_service_watchdogs = true;
 102 static char ***arg_join_controllers = NULL;
 103 static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
 104 static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
 105 static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
 106 static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
 107 static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
 108 static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
 109 static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
 110 static usec_t arg_runtime_watchdog = 0;
 111 static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
 112 static char *arg_early_core_pattern = NULL;
 113 static char *arg_watchdog_device = NULL;
 114 static char **arg_default_environment = NULL;
 115 static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {};
 116 static uint64_t arg_capability_bounding_set = CAP_ALL;
 117 static bool arg_no_new_privs = false;
 118 static nsec_t arg_timer_slack_nsec = NSEC_INFINITY;
 119 static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
 120 static Set* arg_syscall_archs = NULL;
 121 static FILE* arg_serialization = NULL;
 122 static bool arg_default_cpu_accounting = false;
 123 static bool arg_default_io_accounting = false;
 124 static bool arg_default_ip_accounting = false;
 125 static bool arg_default_blockio_accounting = false;
 126 static bool arg_default_memory_accounting = MEMORY_ACCOUNTING_DEFAULT;
 127 static bool arg_default_tasks_accounting = true;
 128 static uint64_t arg_default_tasks_max = UINT64_MAX;
 129 static sd_id128_t arg_machine_id = {};
 130 static EmergencyAction arg_cad_burst_action = EMERGENCY_ACTION_REBOOT_FORCE;
 131
 132 _noreturn_ static void freeze_or_reboot(void) {
 133
 134         if (arg_crash_reboot) {
 135                 log_notice("Rebooting in 10s...");
 136                 (void) sleep(10);
 137
 138                 log_notice("Rebooting now...");
 139                 (void) reboot(RB_AUTOBOOT);
 140                 log_emergency_errno(errno, "Failed to reboot: %m");
 141         }
 142
 143         log_emergency("Freezing execution.");
 144         freeze();
 145 }
 146
 147 _noreturn_ static void crash(int sig) {
 148         struct sigaction sa;
 149         pid_t pid;
 150
 151         if (getpid_cached() != 1)
 152                 /* Pass this on immediately, if this is not PID 1 */
 153                 (void) raise(sig);
 154         else if (!arg_dump_core)
 155                 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
 156         else {
 157                 sa = (struct sigaction) {
 158                         .sa_handler = nop_signal_handler,
 159                         .sa_flags = SA_NOCLDSTOP|SA_RESTART,
 160                 };
 161
 162                 /* We want to wait for the core process, hence let's enable SIGCHLD */
 163                 (void) sigaction(SIGCHLD, &sa, NULL);
 164
 165                 pid = raw_clone(SIGCHLD);
 166                 if (pid < 0)
 167                         log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
 168                 else if (pid == 0) {
 169                         /* Enable default signal handler for core dump */
 170
 171                         sa = (struct sigaction) {
 172                                 .sa_handler = SIG_DFL,
 173                         };
 174                         (void) sigaction(sig, &sa, NULL);
 175
 176                         /* Don't limit the coredump size */
 177                         (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
 178
 179                         /* Just to be sure... */
 180                         (void) chdir("/");
 181
 182                         /* Raise the signal again */
 183                         pid = raw_getpid();
 184                         (void) kill(pid, sig); /* raise() would kill the parent */
 185
 186                         assert_not_reached("We shouldn't be here...");
 187                         _exit(EXIT_FAILURE);
 188                 } else {
 189                         siginfo_t status;
 190                         int r;
 191
 192                         /* Order things nicely. */
 193                         r = wait_for_terminate(pid, &status);
 194                         if (r < 0)
 195                                 log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
 196                         else if (status.si_code != CLD_DUMPED)
 197                                 log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
 198                                               signal_to_string(sig),
 199                                               pid, sigchld_code_to_string(status.si_code),
 200                                               status.si_status,
 201                                               strna(status.si_code == CLD_EXITED
 202                                                     ? exit_status_to_string(status.si_status, EXIT_STATUS_MINIMAL)
 203                                                     : signal_to_string(status.si_status)));
 204                         else
 205                                 log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
 206                 }
 207         }
 208
 209         if (arg_crash_chvt >= 0)
 210                 (void) chvt(arg_crash_chvt);
 211
 212         sa = (struct sigaction) {
 213                 .sa_handler = SIG_IGN,
 214                 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
 215         };
 216
 217         /* Let the kernel reap children for us */
 218         (void) sigaction(SIGCHLD, &sa, NULL);
 219
 220         if (arg_crash_shell) {
 221                 log_notice("Executing crash shell in 10s...");
 222                 (void) sleep(10);
 223
 224                 pid = raw_clone(SIGCHLD);
 225                 if (pid < 0)
 226                         log_emergency_errno(errno, "Failed to fork off crash shell: %m");
 227                 else if (pid == 0) {
 228                         (void) setsid();
 229                         (void) make_console_stdio();
 230                         (void) execle("/bin/sh", "/bin/sh", NULL, environ);
 231
 232                         log_emergency_errno(errno, "execle() failed: %m");
 233                         _exit(EXIT_FAILURE);
 234                 } else {
 235                         log_info("Spawned crash shell as PID "PID_FMT".", pid);
 236                         (void) wait_for_terminate(pid, NULL);
 237                 }
 238         }
 239
 240         freeze_or_reboot();
 241 }
 242
 243 static void install_crash_handler(void) {
 244         static const struct sigaction sa = {
 245                 .sa_handler = crash,
 246                 .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
 247         };
 248         int r;
 249
 250         /* We ignore the return value here, since, we don't mind if we
 251          * cannot set up a crash handler */
 252         r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
 253         if (r < 0)
 254                 log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
 255 }
 256
 257 static int console_setup(void) {
 258         _cleanup_close_ int tty_fd = -1;
 259         int r;
 260
 261         tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
 262         if (tty_fd < 0)
 263                 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
 264
 265         /* We don't want to force text mode.  plymouth may be showing
 266          * pictures already from initrd. */
 267         r = reset_terminal_fd(tty_fd, false);
 268         if (r < 0)
 269                 return log_error_errno(r, "Failed to reset /dev/console: %m");
 270
 271         return 0;
 272 }
 273
 274 static int parse_crash_chvt(const char *value) {
 275         int b;
 276
 277         if (safe_atoi(value, &arg_crash_chvt) >= 0)
 278                 return 0;
 279
 280         b = parse_boolean(value);
 281         if (b < 0)
 282                 return b;
 283
 284         if (b > 0)
 285                 arg_crash_chvt = 0; /* switch to where kmsg goes */
 286         else
 287                 arg_crash_chvt = -1; /* turn off switching */
 288
 289         return 0;
 290 }
 291
 292 static int parse_confirm_spawn(const char *value, char **console) {
 293         char *s;
 294         int r;
 295
 296         r = value ? parse_boolean(value) : 1;
 297         if (r == 0) {
 298                 *console = NULL;
 299                 return 0;
 300         }
 301
 302         if (r > 0) /* on with default tty */
 303                 s = strdup("/dev/console");
 304         else if (is_path(value)) /* on with fully qualified path */
 305                 s = strdup(value);
 306         else /* on with only a tty file name, not a fully qualified path */
 307                 s = strjoin("/dev/", value);
 308         if (!s)
 309                 return -ENOMEM;
 310
 311         *console = s;
 312         return 0;
 313 }
 314
 315 static int set_machine_id(const char *m) {
 316         sd_id128_t t;
 317         assert(m);
 318
 319         if (sd_id128_from_string(m, &t) < 0)
 320                 return -EINVAL;
 321
 322         if (sd_id128_is_null(t))
 323                 return -EINVAL;
 324
 325         arg_machine_id = t;
 326         return 0;
 327 }
 328
 329 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
 330
 331         int r;
 332
 333         assert(key);
 334
 335         if (STR_IN_SET(key, "systemd.unit", "rd.systemd.unit")) {
 336
 337                 if (proc_cmdline_value_missing(key, value))
 338                         return 0;
 339
 340                 if (!unit_name_is_valid(value, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE))
 341                         log_warning("Unit name specified on %s= is not valid, ignoring: %s", key, value);
 342                 else if (in_initrd() == !!startswith(key, "rd.")) {
 343                         if (free_and_strdup(&arg_default_unit, value) < 0)
 344                                 return log_oom();
 345                 }
 346
 347         } else if (proc_cmdline_key_streq(key, "systemd.dump_core")) {
 348
 349                 r = value ? parse_boolean(value) : true;
 350                 if (r < 0)
 351                         log_warning("Failed to parse dump core switch %s. Ignoring.", value);
 352                 else
 353                         arg_dump_core = r;
 354
 355         } else if (proc_cmdline_key_streq(key, "systemd.early_core_pattern")) {
 356
 357                 if (proc_cmdline_value_missing(key, value))
 358                         return 0;
 359
 360                 if (path_is_absolute(value))
 361                         (void) parse_path_argument_and_warn(value, false, &arg_early_core_pattern);
 362                 else
 363                         log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value);
 364
 365         } else if (proc_cmdline_key_streq(key, "systemd.crash_chvt")) {
 366
 367                 if (!value)
 368                         arg_crash_chvt = 0; /* turn on */
 369                 else if (parse_crash_chvt(value) < 0)
 370                         log_warning("Failed to parse crash chvt switch %s. Ignoring.", value);
 371
 372         } else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) {
 373
 374                 r = value ? parse_boolean(value) : true;
 375                 if (r < 0)
 376                         log_warning("Failed to parse crash shell switch %s. Ignoring.", value);
 377                 else
 378                         arg_crash_shell = r;
 379
 380         } else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) {
 381
 382                 r = value ? parse_boolean(value) : true;
 383                 if (r < 0)
 384                         log_warning("Failed to parse crash reboot switch %s. Ignoring.", value);
 385                 else
 386                         arg_crash_reboot = r;
 387
 388         } else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) {
 389                 char *s;
 390
 391                 r = parse_confirm_spawn(value, &s);
 392                 if (r < 0)
 393                         log_warning_errno(r, "Failed to parse confirm_spawn switch %s. Ignoring.", value);
 394                 else {
 395                         free(arg_confirm_spawn);
 396                         arg_confirm_spawn = s;
 397                 }
 398
 399         } else if (proc_cmdline_key_streq(key, "systemd.service_watchdogs")) {
 400
 401                 r = value ? parse_boolean(value) : true;
 402                 if (r < 0)
 403                         log_warning("Failed to parse service watchdog switch %s. Ignoring.", value);
 404                 else
 405                         arg_service_watchdogs = r;
 406
 407         } else if (proc_cmdline_key_streq(key, "systemd.show_status")) {
 408
 409                 if (value) {
 410                         r = parse_show_status(value, &arg_show_status);
 411                         if (r < 0)
 412                                 log_warning("Failed to parse show status switch %s. Ignoring.", value);
 413                 } else
 414                         arg_show_status = SHOW_STATUS_YES;
 415
 416         } else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) {
 417
 418                 if (proc_cmdline_value_missing(key, value))
 419                         return 0;
 420
 421                 r = exec_output_from_string(value);
 422                 if (r < 0)
 423                         log_warning("Failed to parse default standard output switch %s. Ignoring.", value);
 424                 else
 425                         arg_default_std_output = r;
 426
 427         } else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) {
 428
 429                 if (proc_cmdline_value_missing(key, value))
 430                         return 0;
 431
 432                 r = exec_output_from_string(value);
 433                 if (r < 0)
 434                         log_warning("Failed to parse default standard error switch %s. Ignoring.", value);
 435                 else
 436                         arg_default_std_error = r;
 437
 438         } else if (streq(key, "systemd.setenv")) {
 439
 440                 if (proc_cmdline_value_missing(key, value))
 441                         return 0;
 442
 443                 if (env_assignment_is_valid(value)) {
 444                         char **env;
 445
 446                         env = strv_env_set(arg_default_environment, value);
 447                         if (!env)
 448                                 return log_oom();
 449
 450                         arg_default_environment = env;
 451                 } else
 452                         log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
 453
 454         } else if (proc_cmdline_key_streq(key, "systemd.machine_id")) {
 455
 456                 if (proc_cmdline_value_missing(key, value))
 457                         return 0;
 458
 459                 r = set_machine_id(value);
 460                 if (r < 0)
 461                         log_warning("MachineID '%s' is not valid. Ignoring.", value);
 462
 463         } else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) {
 464
 465                 if (proc_cmdline_value_missing(key, value))
 466                         return 0;
 467
 468                 r = parse_sec(value, &arg_default_timeout_start_usec);
 469                 if (r < 0)
 470                         log_warning_errno(r, "Failed to parse default start timeout: %s, ignoring.", value);
 471
 472                 if (arg_default_timeout_start_usec <= 0)
 473                         arg_default_timeout_start_usec = USEC_INFINITY;
 474
 475         } else if (proc_cmdline_key_streq(key, "systemd.watchdog_device")) {
 476
 477                 if (proc_cmdline_value_missing(key, value))
 478                         return 0;
 479
 480                 (void) parse_path_argument_and_warn(value, false, &arg_watchdog_device);
 481
 482         } else if (streq(key, "quiet") && !value) {
 483
 484                 if (arg_show_status == _SHOW_STATUS_INVALID)
 485                         arg_show_status = SHOW_STATUS_AUTO;
 486
 487         } else if (streq(key, "debug") && !value) {
 488
 489                 /* Note that log_parse_environment() handles 'debug'
 490                  * too, and sets the log level to LOG_DEBUG. */
 491
 492                 if (detect_container() > 0)
 493                         log_set_target(LOG_TARGET_CONSOLE);
 494
 495         } else if (!value) {
 496                 const char *target;
 497
 498                 /* SysV compatibility */
 499                 target = runlevel_to_target(key);
 500                 if (target)
 501                         return free_and_strdup(&arg_default_unit, target);
 502         }
 503
 504         return 0;
 505 }
 506
 507 #define DEFINE_SETTER(name, func, descr)                              \
 508         static int name(const char *unit,                             \
 509                         const char *filename,                         \
 510                         unsigned line,                                \
 511                         const char *section,                          \
 512                         unsigned section_line,                        \
 513                         const char *lvalue,                           \
 514                         int ltype,                                    \
 515                         const char *rvalue,                           \
 516                         void *data,                                   \
 517                         void *userdata) {                             \
 518                                                                       \
 519                 int r;                                                \
 520                                                                       \
 521                 assert(filename);                                     \
 522                 assert(lvalue);                                       \
 523                 assert(rvalue);                                       \
 524                                                                       \
 525                 r = func(rvalue);                                     \
 526                 if (r < 0)                                            \
 527                         log_syntax(unit, LOG_ERR, filename, line, r,  \
 528                                    "Invalid " descr "'%s': %m",       \
 529                                    rvalue);                           \
 530                                                                       \
 531                 return 0;                                             \
 532         }
 533
 534 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level");
 535 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target");
 536 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" );
 537 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location");
 538
 539 static int config_parse_cpu_affinity2(
 540                 const char *unit,
 541                 const char *filename,
 542                 unsigned line,
 543                 const char *section,
 544                 unsigned section_line,
 545                 const char *lvalue,
 546                 int ltype,
 547                 const char *rvalue,
 548                 void *data,
 549                 void *userdata) {
 550
 551         _cleanup_cpu_free_ cpu_set_t *c = NULL;
 552         int ncpus;
 553
 554         ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue);
 555         if (ncpus < 0)
 556                 return ncpus;
 557
 558         if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
 559                 log_warning_errno(errno, "Failed to set CPU affinity: %m");
 560
 561         return 0;
 562 }
 563
 564 static int config_parse_show_status(
 565                 const char* unit,
 566                 const char *filename,
 567                 unsigned line,
 568                 const char *section,
 569                 unsigned section_line,
 570                 const char *lvalue,
 571                 int ltype,
 572                 const char *rvalue,
 573                 void *data,
 574                 void *userdata) {
 575
 576         int k;
 577         ShowStatus *b = data;
 578
 579         assert(filename);
 580         assert(lvalue);
 581         assert(rvalue);
 582         assert(data);
 583
 584         k = parse_show_status(rvalue, b);
 585         if (k < 0) {
 586                 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
 587                 return 0;
 588         }
 589
 590         return 0;
 591 }
 592
 593 static int config_parse_output_restricted(
 594                 const char* unit,
 595                 const char *filename,
 596                 unsigned line,
 597                 const char *section,
 598                 unsigned section_line,
 599                 const char *lvalue,
 600                 int ltype,
 601                 const char *rvalue,
 602                 void *data,
 603                 void *userdata) {
 604
 605         ExecOutput t, *eo = data;
 606
 607         assert(filename);
 608         assert(lvalue);
 609         assert(rvalue);
 610         assert(data);
 611
 612         t = exec_output_from_string(rvalue);
 613         if (t < 0) {
 614                 log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse output type, ignoring: %s", rvalue);
 615                 return 0;
 616         }
 617
 618         if (IN_SET(t, EXEC_OUTPUT_SOCKET, EXEC_OUTPUT_NAMED_FD, EXEC_OUTPUT_FILE, EXEC_OUTPUT_FILE_APPEND)) {
 619                 log_syntax(unit, LOG_ERR, filename, line, 0, "Standard output types socket, fd:, file:, append: are not supported as defaults, ignoring: %s", rvalue);
 620                 return 0;
 621         }
 622
 623         *eo = t;
 624         return 0;
 625 }
 626
 627 static int config_parse_crash_chvt(
 628                 const char* unit,
 629                 const char *filename,
 630                 unsigned line,
 631                 const char *section,
 632                 unsigned section_line,
 633                 const char *lvalue,
 634                 int ltype,
 635                 const char *rvalue,
 636                 void *data,
 637                 void *userdata) {
 638
 639         int r;
 640
 641         assert(filename);
 642         assert(lvalue);
 643         assert(rvalue);
 644
 645         r = parse_crash_chvt(rvalue);
 646         if (r < 0) {
 647                 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
 648                 return 0;
 649         }
 650
 651         return 0;
 652 }
 653
 654 static int parse_config_file(void) {
 655
 656         const ConfigTableItem items[] = {
 657                 { "Manager", "LogLevel",                  config_parse_level2,           0, NULL                                   },
 658                 { "Manager", "LogTarget",                 config_parse_target,           0, NULL                                   },
 659                 { "Manager", "LogColor",                  config_parse_color,            0, NULL                                   },
 660                 { "Manager", "LogLocation",               config_parse_location,         0, NULL                                   },
 661                 { "Manager", "DumpCore",                  config_parse_bool,             0, &arg_dump_core                         },
 662                 { "Manager", "CrashChVT", /* legacy */    config_parse_crash_chvt,       0, NULL                                   },
 663                 { "Manager", "CrashChangeVT",             config_parse_crash_chvt,       0, NULL                                   },
 664                 { "Manager", "CrashShell",                config_parse_bool,             0, &arg_crash_shell                       },
 665                 { "Manager", "CrashReboot",               config_parse_bool,             0, &arg_crash_reboot                      },
 666                 { "Manager", "ShowStatus",                config_parse_show_status,      0, &arg_show_status                       },
 667                 { "Manager", "CPUAffinity",               config_parse_cpu_affinity2,    0, NULL                                   },
 668                 { "Manager", "JoinControllers",           config_parse_join_controllers, 0, &arg_join_controllers                  },
 669                 { "Manager", "RuntimeWatchdogSec",        config_parse_sec,              0, &arg_runtime_watchdog                  },
 670                 { "Manager", "ShutdownWatchdogSec",       config_parse_sec,              0, &arg_shutdown_watchdog                 },
 671                 { "Manager", "WatchdogDevice",            config_parse_path,             0, &arg_watchdog_device                   },
 672                 { "Manager", "CapabilityBoundingSet",     config_parse_capability_set,   0, &arg_capability_bounding_set           },
 673                 { "Manager", "NoNewPrivileges",           config_parse_bool,             0, &arg_no_new_privs                      },
 674 #if HAVE_SECCOMP
 675                 { "Manager", "SystemCallArchitectures",   config_parse_syscall_archs,    0, &arg_syscall_archs                     },
 676 #endif
 677                 { "Manager", "TimerSlackNSec",            config_parse_nsec,             0, &arg_timer_slack_nsec                  },
 678                 { "Manager", "DefaultTimerAccuracySec",   config_parse_sec,              0, &arg_default_timer_accuracy_usec       },
 679                 { "Manager", "DefaultStandardOutput",     config_parse_output_restricted,0, &arg_default_std_output                },
 680                 { "Manager", "DefaultStandardError",      config_parse_output_restricted,0, &arg_default_std_error                 },
 681                 { "Manager", "DefaultTimeoutStartSec",    config_parse_sec,              0, &arg_default_timeout_start_usec        },
 682                 { "Manager", "DefaultTimeoutStopSec",     config_parse_sec,              0, &arg_default_timeout_stop_usec         },
 683                 { "Manager", "DefaultRestartSec",         config_parse_sec,              0, &arg_default_restart_usec              },
 684                 { "Manager", "DefaultStartLimitInterval", config_parse_sec,              0, &arg_default_start_limit_interval      }, /* obsolete alias */
 685                 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec,            0, &arg_default_start_limit_interval      },
 686                 { "Manager", "DefaultStartLimitBurst",    config_parse_unsigned,         0, &arg_default_start_limit_burst         },
 687                 { "Manager", "DefaultEnvironment",        config_parse_environ,          0, &arg_default_environment               },
 688                 { "Manager", "DefaultLimitCPU",           config_parse_rlimit,           RLIMIT_CPU, arg_default_rlimit            },
 689                 { "Manager", "DefaultLimitFSIZE",         config_parse_rlimit,           RLIMIT_FSIZE, arg_default_rlimit          },
 690                 { "Manager", "DefaultLimitDATA",          config_parse_rlimit,           RLIMIT_DATA, arg_default_rlimit           },
 691                 { "Manager", "DefaultLimitSTACK",         config_parse_rlimit,           RLIMIT_STACK, arg_default_rlimit          },
 692                 { "Manager", "DefaultLimitCORE",          config_parse_rlimit,           RLIMIT_CORE, arg_default_rlimit           },
 693                 { "Manager", "DefaultLimitRSS",           config_parse_rlimit,           RLIMIT_RSS, arg_default_rlimit            },
 694                 { "Manager", "DefaultLimitNOFILE",        config_parse_rlimit,           RLIMIT_NOFILE, arg_default_rlimit         },
 695                 { "Manager", "DefaultLimitAS",            config_parse_rlimit,           RLIMIT_AS, arg_default_rlimit             },
 696                 { "Manager", "DefaultLimitNPROC",         config_parse_rlimit,           RLIMIT_NPROC, arg_default_rlimit          },
 697                 { "Manager", "DefaultLimitMEMLOCK",       config_parse_rlimit,           RLIMIT_MEMLOCK, arg_default_rlimit        },
 698                 { "Manager", "DefaultLimitLOCKS",         config_parse_rlimit,           RLIMIT_LOCKS, arg_default_rlimit          },
 699                 { "Manager", "DefaultLimitSIGPENDING",    config_parse_rlimit,           RLIMIT_SIGPENDING, arg_default_rlimit     },
 700                 { "Manager", "DefaultLimitMSGQUEUE",      config_parse_rlimit,           RLIMIT_MSGQUEUE, arg_default_rlimit       },
 701                 { "Manager", "DefaultLimitNICE",          config_parse_rlimit,           RLIMIT_NICE, arg_default_rlimit           },
 702                 { "Manager", "DefaultLimitRTPRIO",        config_parse_rlimit,           RLIMIT_RTPRIO, arg_default_rlimit         },
 703                 { "Manager", "DefaultLimitRTTIME",        config_parse_rlimit,           RLIMIT_RTTIME, arg_default_rlimit         },
 704                 { "Manager", "DefaultCPUAccounting",      config_parse_bool,             0, &arg_default_cpu_accounting            },
 705                 { "Manager", "DefaultIOAccounting",       config_parse_bool,             0, &arg_default_io_accounting             },
 706                 { "Manager", "DefaultIPAccounting",       config_parse_bool,             0, &arg_default_ip_accounting             },
 707                 { "Manager", "DefaultBlockIOAccounting",  config_parse_bool,             0, &arg_default_blockio_accounting        },
 708                 { "Manager", "DefaultMemoryAccounting",   config_parse_bool,             0, &arg_default_memory_accounting         },
 709                 { "Manager", "DefaultTasksAccounting",    config_parse_bool,             0, &arg_default_tasks_accounting          },
 710                 { "Manager", "DefaultTasksMax",           config_parse_tasks_max,        0, &arg_default_tasks_max                 },
 711                 { "Manager", "CtrlAltDelBurstAction",     config_parse_emergency_action, 0, &arg_cad_burst_action                  },
 712                 {}
 713         };
 714
 715         const char *fn, *conf_dirs_nulstr;
 716
 717         fn = arg_system ?
 718                 PKGSYSCONFDIR "/system.conf" :
 719                 PKGSYSCONFDIR "/user.conf";
 720
 721         conf_dirs_nulstr = arg_system ?
 722                 CONF_PATHS_NULSTR("systemd/system.conf.d") :
 723                 CONF_PATHS_NULSTR("systemd/user.conf.d");
 724
 725         (void) config_parse_many_nulstr(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, CONFIG_PARSE_WARN, NULL);
 726
 727         /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
 728          * like everywhere else. */
 729         if (arg_default_timeout_start_usec <= 0)
 730                 arg_default_timeout_start_usec = USEC_INFINITY;
 731         if (arg_default_timeout_stop_usec <= 0)
 732                 arg_default_timeout_stop_usec = USEC_INFINITY;
 733
 734         return 0;
 735 }
 736
 737 static void set_manager_defaults(Manager *m) {
 738
 739         assert(m);
 740
 741         /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
 742          * affect the manager itself, but are just what newly allocated units will have set if they haven't set
 743          * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
 744
 745         m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
 746         m->default_std_output = arg_default_std_output;
 747         m->default_std_error = arg_default_std_error;
 748         m->default_timeout_start_usec = arg_default_timeout_start_usec;
 749         m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
 750         m->default_restart_usec = arg_default_restart_usec;
 751         m->default_start_limit_interval = arg_default_start_limit_interval;
 752         m->default_start_limit_burst = arg_default_start_limit_burst;
 753         m->default_cpu_accounting = arg_default_cpu_accounting;
 754         m->default_io_accounting = arg_default_io_accounting;
 755         m->default_ip_accounting = arg_default_ip_accounting;
 756         m->default_blockio_accounting = arg_default_blockio_accounting;
 757         m->default_memory_accounting = arg_default_memory_accounting;
 758         m->default_tasks_accounting = arg_default_tasks_accounting;
 759         m->default_tasks_max = arg_default_tasks_max;
 760
 761         manager_set_default_rlimits(m, arg_default_rlimit);
 762         manager_environment_add(m, NULL, arg_default_environment);
 763 }
 764
 765 static void set_manager_settings(Manager *m) {
 766
 767         assert(m);
 768
 769         /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
 770          * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
 771
 772         m->confirm_spawn = arg_confirm_spawn;
 773         m->service_watchdogs = arg_service_watchdogs;
 774         m->runtime_watchdog = arg_runtime_watchdog;
 775         m->shutdown_watchdog = arg_shutdown_watchdog;
 776         m->cad_burst_action = arg_cad_burst_action;
 777
 778         manager_set_show_status(m, arg_show_status);
 779 }
 780
 781 static int parse_argv(int argc, char *argv[]) {
 782         enum {
 783                 ARG_LOG_LEVEL = 0x100,
 784                 ARG_LOG_TARGET,
 785                 ARG_LOG_COLOR,
 786                 ARG_LOG_LOCATION,
 787                 ARG_UNIT,
 788                 ARG_SYSTEM,
 789                 ARG_USER,
 790                 ARG_TEST,
 791                 ARG_NO_PAGER,
 792                 ARG_VERSION,
 793                 ARG_DUMP_CONFIGURATION_ITEMS,
 794                 ARG_DUMP_BUS_PROPERTIES,
 795                 ARG_DUMP_CORE,
 796                 ARG_CRASH_CHVT,
 797                 ARG_CRASH_SHELL,
 798                 ARG_CRASH_REBOOT,
 799                 ARG_CONFIRM_SPAWN,
 800                 ARG_SHOW_STATUS,
 801                 ARG_DESERIALIZE,
 802                 ARG_SWITCHED_ROOT,
 803                 ARG_DEFAULT_STD_OUTPUT,
 804                 ARG_DEFAULT_STD_ERROR,
 805                 ARG_MACHINE_ID,
 806                 ARG_SERVICE_WATCHDOGS,
 807         };
 808
 809         static const struct option options[] = {
 810                 { "log-level",                required_argument, NULL, ARG_LOG_LEVEL                },
 811                 { "log-target",               required_argument, NULL, ARG_LOG_TARGET               },
 812                 { "log-color",                optional_argument, NULL, ARG_LOG_COLOR                },
 813                 { "log-location",             optional_argument, NULL, ARG_LOG_LOCATION             },
 814                 { "unit",                     required_argument, NULL, ARG_UNIT                     },
 815                 { "system",                   no_argument,       NULL, ARG_SYSTEM                   },
 816                 { "user",                     no_argument,       NULL, ARG_USER                     },
 817                 { "test",                     no_argument,       NULL, ARG_TEST                     },
 818                 { "no-pager",                 no_argument,       NULL, ARG_NO_PAGER                 },
 819                 { "help",                     no_argument,       NULL, 'h'                          },
 820                 { "version",                  no_argument,       NULL, ARG_VERSION                  },
 821                 { "dump-configuration-items", no_argument,       NULL, ARG_DUMP_CONFIGURATION_ITEMS },
 822                 { "dump-bus-properties",      no_argument,       NULL, ARG_DUMP_BUS_PROPERTIES      },
 823                 { "dump-core",                optional_argument, NULL, ARG_DUMP_CORE                },
 824                 { "crash-chvt",               required_argument, NULL, ARG_CRASH_CHVT               },
 825                 { "crash-shell",              optional_argument, NULL, ARG_CRASH_SHELL              },
 826                 { "crash-reboot",             optional_argument, NULL, ARG_CRASH_REBOOT             },
 827                 { "confirm-spawn",            optional_argument, NULL, ARG_CONFIRM_SPAWN            },
 828                 { "show-status",              optional_argument, NULL, ARG_SHOW_STATUS              },
 829                 { "deserialize",              required_argument, NULL, ARG_DESERIALIZE              },
 830                 { "switched-root",            no_argument,       NULL, ARG_SWITCHED_ROOT            },
 831                 { "default-standard-output",  required_argument, NULL, ARG_DEFAULT_STD_OUTPUT,      },
 832                 { "default-standard-error",   required_argument, NULL, ARG_DEFAULT_STD_ERROR,       },
 833                 { "machine-id",               required_argument, NULL, ARG_MACHINE_ID               },
 834                 { "service-watchdogs",        required_argument, NULL, ARG_SERVICE_WATCHDOGS        },
 835                 {}
 836         };
 837
 838         int c, r;
 839
 840         assert(argc >= 1);
 841         assert(argv);
 842
 843         if (getpid_cached() == 1)
 844                 opterr = 0;
 845
 846         while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
 847
 848                 switch (c) {
 849
 850                 case ARG_LOG_LEVEL:
 851                         r = log_set_max_level_from_string(optarg);
 852                         if (r < 0)
 853                                 return log_error_errno(r, "Failed to parse log level \"%s\": %m", optarg);
 854
 855                         break;
 856
 857                 case ARG_LOG_TARGET:
 858                         r = log_set_target_from_string(optarg);
 859                         if (r < 0)
 860                                 return log_error_errno(r, "Failed to parse log target \"%s\": %m", optarg);
 861
 862                         break;
 863
 864                 case ARG_LOG_COLOR:
 865
 866                         if (optarg) {
 867                                 r = log_show_color_from_string(optarg);
 868                                 if (r < 0)
 869                                         return log_error_errno(r, "Failed to parse log color setting \"%s\": %m",
 870                                                                optarg);
 871                         } else
 872                                 log_show_color(true);
 873
 874                         break;
 875
 876                 case ARG_LOG_LOCATION:
 877                         if (optarg) {
 878                                 r = log_show_location_from_string(optarg);
 879                                 if (r < 0)
 880                                         return log_error_errno(r, "Failed to parse log location setting \"%s\": %m",
 881                                                                optarg);
 882                         } else
 883                                 log_show_location(true);
 884
 885                         break;
 886
 887                 case ARG_DEFAULT_STD_OUTPUT:
 888                         r = exec_output_from_string(optarg);
 889                         if (r < 0)
 890                                 return log_error_errno(r, "Failed to parse default standard output setting \"%s\": %m",
 891                                                        optarg);
 892                         arg_default_std_output = r;
 893                         break;
 894
 895                 case ARG_DEFAULT_STD_ERROR:
 896                         r = exec_output_from_string(optarg);
 897                         if (r < 0)
 898                                 return log_error_errno(r, "Failed to parse default standard error output setting \"%s\": %m",
 899                                                        optarg);
 900                         arg_default_std_error = r;
 901                         break;
 902
 903                 case ARG_UNIT:
 904                         r = free_and_strdup(&arg_default_unit, optarg);
 905                         if (r < 0)
 906                                 return log_error_errno(r, "Failed to set default unit \"%s\": %m", optarg);
 907
 908                         break;
 909
 910                 case ARG_SYSTEM:
 911                         arg_system = true;
 912                         break;
 913
 914                 case ARG_USER:
 915                         arg_system = false;
 916                         break;
 917
 918                 case ARG_TEST:
 919                         arg_action = ACTION_TEST;
 920                         break;
 921
 922                 case ARG_NO_PAGER:
 923                         arg_no_pager = true;
 924                         break;
 925
 926                 case ARG_VERSION:
 927                         arg_action = ACTION_VERSION;
 928                         break;
 929
 930                 case ARG_DUMP_CONFIGURATION_ITEMS:
 931                         arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
 932                         break;
 933
 934                 case ARG_DUMP_BUS_PROPERTIES:
 935                         arg_action = ACTION_DUMP_BUS_PROPERTIES;
 936                         break;
 937
 938                 case ARG_DUMP_CORE:
 939                         if (!optarg)
 940                                 arg_dump_core = true;
 941                         else {
 942                                 r = parse_boolean(optarg);
 943                                 if (r < 0)
 944                                         return log_error_errno(r, "Failed to parse dump core boolean: \"%s\": %m",
 945                                                                optarg);
 946                                 arg_dump_core = r;
 947                         }
 948                         break;
 949
 950                 case ARG_CRASH_CHVT:
 951                         r = parse_crash_chvt(optarg);
 952                         if (r < 0)
 953                                 return log_error_errno(r, "Failed to parse crash virtual terminal index: \"%s\": %m",
 954                                                        optarg);
 955                         break;
 956
 957                 case ARG_CRASH_SHELL:
 958                         if (!optarg)
 959                                 arg_crash_shell = true;
 960                         else {
 961                                 r = parse_boolean(optarg);
 962                                 if (r < 0)
 963                                         return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
 964                                                                optarg);
 965                                 arg_crash_shell = r;
 966                         }
 967                         break;
 968
 969                 case ARG_CRASH_REBOOT:
 970                         if (!optarg)
 971                                 arg_crash_reboot = true;
 972                         else {
 973                                 r = parse_boolean(optarg);
 974                                 if (r < 0)
 975                                         return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
 976                                                                optarg);
 977                                 arg_crash_reboot = r;
 978                         }
 979                         break;
 980
 981                 case ARG_CONFIRM_SPAWN:
 982                         arg_confirm_spawn = mfree(arg_confirm_spawn);
 983
 984                         r = parse_confirm_spawn(optarg, &arg_confirm_spawn);
 985                         if (r < 0)
 986                                 return log_error_errno(r, "Failed to parse confirm spawn option: \"%s\": %m",
 987                                                        optarg);
 988                         break;
 989
 990                 case ARG_SERVICE_WATCHDOGS:
 991                         r = parse_boolean(optarg);
 992                         if (r < 0)
 993                                 return log_error_errno(r, "Failed to parse service watchdogs boolean: \"%s\": %m",
 994                                                        optarg);
 995                         arg_service_watchdogs = r;
 996                         break;
 997
 998                 case ARG_SHOW_STATUS:
 999                         if (optarg) {
1000                                 r = parse_show_status(optarg, &arg_show_status);
1001                                 if (r < 0)
1002                                         return log_error_errno(r, "Failed to parse show status boolean: \"%s\": %m",
1003                                                                optarg);
1004                         } else
1005                                 arg_show_status = SHOW_STATUS_YES;
1006                         break;
1007
1008                 case ARG_DESERIALIZE: {
1009                         int fd;
1010                         FILE *f;
1011
1012                         r = safe_atoi(optarg, &fd);
1013                         if (r < 0)
1014                                 log_error_errno(r, "Failed to parse deserialize option \"%s\": %m", optarg);
1015                         if (fd < 0) {
1016                                 log_error("Invalid deserialize fd: %d", fd);
1017                                 return -EINVAL;
1018                         }
1019
1020                         (void) fd_cloexec(fd, true);
1021
1022                         f = fdopen(fd, "r");
1023                         if (!f)
1024                                 return log_error_errno(errno, "Failed to open serialization fd %d: %m", fd);
1025
1026                         safe_fclose(arg_serialization);
1027                         arg_serialization = f;
1028
1029                         break;
1030                 }
1031
1032                 case ARG_SWITCHED_ROOT:
1033                         arg_switched_root = true;
1034                         break;
1035
1036                 case ARG_MACHINE_ID:
1037                         r = set_machine_id(optarg);
1038                         if (r < 0)
1039                                 return log_error_errno(r, "MachineID '%s' is not valid: %m", optarg);
1040                         break;
1041
1042                 case 'h':
1043                         arg_action = ACTION_HELP;
1044                         break;
1045
1046                 case 'D':
1047                         log_set_max_level(LOG_DEBUG);
1048                         break;
1049
1050                 case 'b':
1051                 case 's':
1052                 case 'z':
1053                         /* Just to eat away the sysvinit kernel
1054                          * cmdline args without getopt() error
1055                          * messages that we'll parse in
1056                          * parse_proc_cmdline_word() or ignore. */
1057
1058                 case '?':
1059                         if (getpid_cached() != 1)
1060                                 return -EINVAL;
1061                         else
1062                                 return 0;
1063
1064                 default:
1065                         assert_not_reached("Unhandled option code.");
1066                 }
1067
1068         if (optind < argc && getpid_cached() != 1) {
1069                 /* Hmm, when we aren't run as init system
1070                  * let's complain about excess arguments */
1071
1072                 log_error("Excess arguments.");
1073                 return -EINVAL;
1074         }
1075
1076         return 0;
1077 }
1078
1079 static int help(void) {
1080         _cleanup_free_ char *link = NULL;
1081         int r;
1082
1083         r = terminal_urlify_man("systemd", "1", &link);
1084         if (r < 0)
1085                 return log_oom();
1086
1087         printf("%s [OPTIONS...]\n\n"
1088                "Starts up and maintains the system or user services.\n\n"
1089                "  -h --help                      Show this help\n"
1090                "     --version                   Show version\n"
1091                "     --test                      Determine startup sequence, dump it and exit\n"
1092                "     --no-pager                  Do not pipe output into a pager\n"
1093                "     --dump-configuration-items  Dump understood unit configuration items\n"
1094                "     --dump-bus-properties       Dump exposed bus properties\n"
1095                "     --unit=UNIT                 Set default unit\n"
1096                "     --system                    Run a system instance, even if PID != 1\n"
1097                "     --user                      Run a user instance\n"
1098                "     --dump-core[=BOOL]          Dump core on crash\n"
1099                "     --crash-vt=NR               Change to specified VT on crash\n"
1100                "     --crash-reboot[=BOOL]       Reboot on crash\n"
1101                "     --crash-shell[=BOOL]        Run shell on crash\n"
1102                "     --confirm-spawn[=BOOL]      Ask for confirmation when spawning processes\n"
1103                "     --show-status[=BOOL]        Show status updates on the console during bootup\n"
1104                "     --log-target=TARGET         Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1105                "     --log-level=LEVEL           Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1106                "     --log-color[=BOOL]          Highlight important log messages\n"
1107                "     --log-location[=BOOL]       Include code location in log messages\n"
1108                "     --default-standard-output=  Set default standard output for services\n"
1109                "     --default-standard-error=   Set default standard error output for services\n"
1110                "\nSee the %s for details.\n"
1111                , program_invocation_short_name
1112                , link
1113         );
1114
1115         return 0;
1116 }
1117
1118 static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) {
1119         _cleanup_fdset_free_ FDSet *fds = NULL;
1120         _cleanup_fclose_ FILE *f = NULL;
1121         int r;
1122
1123         assert(m);
1124         assert(_f);
1125         assert(_fds);
1126
1127         r = manager_open_serialization(m, &f);
1128         if (r < 0)
1129                 return log_error_errno(r, "Failed to create serialization file: %m");
1130
1131         /* Make sure nothing is really destructed when we shut down */
1132         m->n_reloading++;
1133         bus_manager_send_reloading(m, true);
1134
1135         fds = fdset_new();
1136         if (!fds)
1137                 return log_oom();
1138
1139         r = manager_serialize(m, f, fds, switching_root);
1140         if (r < 0)
1141                 return log_error_errno(r, "Failed to serialize state: %m");
1142
1143         if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1144                 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1145
1146         r = fd_cloexec(fileno(f), false);
1147         if (r < 0)
1148                 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1149
1150         r = fdset_cloexec(fds, false);
1151         if (r < 0)
1152                 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1153
1154         *_f = TAKE_PTR(f);
1155         *_fds = TAKE_PTR(fds);
1156
1157         return 0;
1158 }
1159
1160 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1161         int r, nr;
1162
1163         assert(saved_rlimit);
1164
1165         /* Save the original RLIMIT_NOFILE so that we can reset it
1166          * later when transitioning from the initrd to the main
1167          * systemd or suchlike. */
1168         if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1169                 return log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1170
1171         /* Make sure forked processes get the default kernel setting */
1172         if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1173                 struct rlimit *rl;
1174
1175                 rl = newdup(struct rlimit, saved_rlimit, 1);
1176                 if (!rl)
1177                         return log_oom();
1178
1179                 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1180         }
1181
1182         /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows */
1183         nr = read_nr_open();
1184         r = setrlimit_closest(RLIMIT_NOFILE, &RLIMIT_MAKE_CONST(nr));
1185         if (r < 0)
1186                 return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1187
1188         return 0;
1189 }
1190
1191 static int bump_rlimit_memlock(struct rlimit *saved_rlimit) {
1192         int r;
1193
1194         assert(saved_rlimit);
1195         assert(getuid() == 0);
1196
1197         /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even though we have CAP_IPC_LOCK which
1198          * should normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's
1199          * bump the value high enough for the root user. */
1200
1201         if (getrlimit(RLIMIT_MEMLOCK, saved_rlimit) < 0)
1202                 return log_warning_errno(errno, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1203
1204         r = setrlimit_closest(RLIMIT_MEMLOCK, &RLIMIT_MAKE_CONST(1024ULL*1024ULL*16ULL));
1205         if (r < 0)
1206                 return log_warning_errno(r, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1207
1208         return 0;
1209 }
1210
1211 static void test_usr(void) {
1212
1213         /* Check that /usr is not a separate fs */
1214
1215         if (dir_is_empty("/usr") <= 0)
1216                 return;
1217
1218         log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1219                     "Some things will probably break (sometimes even silently) in mysterious ways. "
1220                     "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1221 }
1222
1223 static int enforce_syscall_archs(Set *archs) {
1224 #if HAVE_SECCOMP
1225         int r;
1226
1227         if (!is_seccomp_available())
1228                 return 0;
1229
1230         r = seccomp_restrict_archs(arg_syscall_archs);
1231         if (r < 0)
1232                 return log_error_errno(r, "Failed to enforce system call architecture restrication: %m");
1233 #endif
1234         return 0;
1235 }
1236
1237 static int status_welcome(void) {
1238         _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1239         int r;
1240
1241         if (IN_SET(arg_show_status, SHOW_STATUS_NO, SHOW_STATUS_AUTO))
1242                 return 0;
1243
1244         r = parse_os_release(NULL,
1245                              "PRETTY_NAME", &pretty_name,
1246                              "ANSI_COLOR", &ansi_color,
1247                              NULL);
1248         if (r < 0)
1249                 log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r,
1250                                "Failed to read os-release file, ignoring: %m");
1251
1252         if (log_get_show_color())
1253                 return status_printf(NULL, false, false,
1254                                      "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1255                                      isempty(ansi_color) ? "1" : ansi_color,
1256                                      isempty(pretty_name) ? "Linux" : pretty_name);
1257         else
1258                 return status_printf(NULL, false, false,
1259                                      "\nWelcome to %s!\n",
1260                                      isempty(pretty_name) ? "Linux" : pretty_name);
1261 }
1262
1263 static int write_container_id(void) {
1264         const char *c;
1265         int r;
1266
1267         c = getenv("container");
1268         if (isempty(c))
1269                 return 0;
1270
1271         RUN_WITH_UMASK(0022)
1272                 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1273         if (r < 0)
1274                 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
1275
1276         return 1;
1277 }
1278
1279 static int bump_unix_max_dgram_qlen(void) {
1280         _cleanup_free_ char *qlen = NULL;
1281         unsigned long v;
1282         int r;
1283
1284         /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1285          * really really early during boot, so that it is actually applied to all our sockets, including the
1286          * $NOTIFY_SOCKET one. */
1287
1288         r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1289         if (r < 0)
1290                 return log_warning_errno(r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1291
1292         r = safe_atolu(qlen, &v);
1293         if (r < 0)
1294                 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen);
1295
1296         if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1297                 return 0;
1298
1299         r = write_string_filef("/proc/sys/net/unix/max_dgram_qlen", 0, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN);
1300         if (r < 0)
1301                 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1302                                       "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1303
1304         return 1;
1305 }
1306
1307 static int fixup_environment(void) {
1308         _cleanup_free_ char *term = NULL;
1309         const char *t;
1310         int r;
1311
1312         /* Only fix up the environment when we are started as PID 1 */
1313         if (getpid_cached() != 1)
1314                 return 0;
1315
1316         /* We expect the environment to be set correctly if run inside a container. */
1317         if (detect_container() > 0)
1318                 return 0;
1319
1320         /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1321          * device used by the console. We try to make a better guess here since some consoles might not have support
1322          * for color mode for example.
1323          *
1324          * However if TERM was configured through the kernel command line then leave it alone. */
1325         r = proc_cmdline_get_key("TERM", 0, &term);
1326         if (r < 0)
1327                 return r;
1328
1329         t = term ?: default_term_for_tty("/dev/console");
1330
1331         if (setenv("TERM", t, 1) < 0)
1332                 return -errno;
1333
1334         return 0;
1335 }
1336
1337 static void redirect_telinit(int argc, char *argv[]) {
1338
1339         /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1340
1341 #if HAVE_SYSV_COMPAT
1342         if (getpid_cached() == 1)
1343                 return;
1344
1345         if (!strstr(program_invocation_short_name, "init"))
1346                 return;
1347
1348         execv(SYSTEMCTL_BINARY_PATH, argv);
1349         log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1350         exit(EXIT_FAILURE);
1351 #endif
1352 }
1353
1354 static int become_shutdown(
1355                 const char *shutdown_verb,
1356                 int retval) {
1357
1358         char log_level[DECIMAL_STR_MAX(int) + 1],
1359                 exit_code[DECIMAL_STR_MAX(uint8_t) + 1],
1360                 timeout[DECIMAL_STR_MAX(usec_t) + 1];
1361
1362         const char* command_line[13] = {
1363                 SYSTEMD_SHUTDOWN_BINARY_PATH,
1364                 shutdown_verb,
1365                 "--timeout", timeout,
1366                 "--log-level", log_level,
1367                 "--log-target",
1368         };
1369
1370         _cleanup_strv_free_ char **env_block = NULL;
1371         size_t pos = 7;
1372         int r;
1373
1374         assert(shutdown_verb);
1375         assert(!command_line[pos]);
1376         env_block = strv_copy(environ);
1377
1378         xsprintf(log_level, "%d", log_get_max_level());
1379         xsprintf(timeout, "%" PRI_USEC "us", arg_default_timeout_stop_usec);
1380
1381         switch (log_get_target()) {
1382
1383         case LOG_TARGET_KMSG:
1384         case LOG_TARGET_JOURNAL_OR_KMSG:
1385         case LOG_TARGET_SYSLOG_OR_KMSG:
1386                 command_line[pos++] = "kmsg";
1387                 break;
1388
1389         case LOG_TARGET_NULL:
1390                 command_line[pos++] = "null";
1391                 break;
1392
1393         case LOG_TARGET_CONSOLE:
1394         default:
1395                 command_line[pos++] = "console";
1396                 break;
1397         };
1398
1399         if (log_get_show_color())
1400                 command_line[pos++] = "--log-color";
1401
1402         if (log_get_show_location())
1403                 command_line[pos++] = "--log-location";
1404
1405         if (streq(shutdown_verb, "exit")) {
1406                 command_line[pos++] = "--exit-code";
1407                 command_line[pos++] = exit_code;
1408                 xsprintf(exit_code, "%d", retval);
1409         }
1410
1411         assert(pos < ELEMENTSOF(command_line));
1412
1413         if (streq(shutdown_verb, "reboot") &&
1414             arg_shutdown_watchdog > 0 &&
1415             arg_shutdown_watchdog != USEC_INFINITY) {
1416
1417                 char *e;
1418
1419                 /* If we reboot let's set the shutdown
1420                  * watchdog and tell the shutdown binary to
1421                  * repeatedly ping it */
1422                 r = watchdog_set_timeout(&arg_shutdown_watchdog);
1423                 watchdog_close(r < 0);
1424
1425                 /* Tell the binary how often to ping, ignore failure */
1426                 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
1427                         (void) strv_consume(&env_block, e);
1428
1429                 if (arg_watchdog_device &&
1430                     asprintf(&e, "WATCHDOG_DEVICE=%s", arg_watchdog_device) > 0)
1431                         (void) strv_consume(&env_block, e);
1432         } else
1433                 watchdog_close(true);
1434
1435         /* Avoid the creation of new processes forked by the
1436          * kernel; at this point, we will not listen to the
1437          * signals anyway */
1438         if (detect_container() <= 0)
1439                 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
1440
1441         execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
1442         return -errno;
1443 }
1444
1445 static void initialize_clock(void) {
1446         int r;
1447
1448         if (clock_is_localtime(NULL) > 0) {
1449                 int min;
1450
1451                 /*
1452                  * The very first call of settimeofday() also does a time warp in the kernel.
1453                  *
1454                  * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1455                  * of maintaining the RTC and do all adjustments.  This matches the behavior of Windows, which leaves
1456                  * the RTC alone if the registry tells that the RTC runs in UTC.
1457                  */
1458                 r = clock_set_timezone(&min);
1459                 if (r < 0)
1460                         log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1461                 else
1462                         log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1463
1464         } else if (!in_initrd()) {
1465                 /*
1466                  * Do a dummy very first call to seal the kernel's time warp magic.
1467                  *
1468                  * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1469                  * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1470                  * until we reach the real system.
1471                  *
1472                  * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1473                  * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1474                  * be treated as UTC that way.
1475                  */
1476                 (void) clock_reset_timewarp();
1477         }
1478
1479         r = clock_apply_epoch();
1480         if (r < 0)
1481                 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1482         else if (r > 0)
1483                 log_info("System time before build time, advancing clock.");
1484 }
1485
1486 static void initialize_coredump(bool skip_setup) {
1487 #if ENABLE_COREDUMP
1488         if (getpid_cached() != 1)
1489                 return;
1490
1491         /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1492          * will process core dumps for system services by default. */
1493         if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0)
1494                 log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m");
1495
1496         /* But at the same time, turn off the core_pattern logic by default, so that no
1497          * coredumps are stored until the systemd-coredump tool is enabled via
1498          * sysctl. However it can be changed via the kernel command line later so core
1499          * dumps can still be generated during early startup and in initramfs. */
1500         if (!skip_setup)
1501                 disable_coredumps();
1502 #endif
1503 }
1504
1505 static void initialize_core_pattern(bool skip_setup) {
1506         int r;
1507
1508         if (skip_setup || !arg_early_core_pattern)
1509                 return;
1510
1511         if (getpid_cached() != 1)
1512                 return;
1513
1514         r = write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern, 0);
1515         if (r < 0)
1516                 log_warning_errno(r, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", arg_early_core_pattern);
1517 }
1518
1519 static void do_reexecute(
1520                 int argc,
1521                 char *argv[],
1522                 const struct rlimit *saved_rlimit_nofile,
1523                 const struct rlimit *saved_rlimit_memlock,
1524                 FDSet *fds,
1525                 const char *switch_root_dir,
1526                 const char *switch_root_init,
1527                 const char **ret_error_message) {
1528
1529         unsigned i, j, args_size;
1530         const char **args;
1531         int r;
1532
1533         assert(saved_rlimit_nofile);
1534         assert(saved_rlimit_memlock);
1535         assert(ret_error_message);
1536
1537         /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1538          * we do that */
1539         watchdog_close(true);
1540
1541         /* Reset the RLIMIT_NOFILE to the kernel default, so that the new systemd can pass the kernel default to its
1542          * child processes */
1543
1544         if (saved_rlimit_nofile->rlim_cur > 0)
1545                 (void) setrlimit(RLIMIT_NOFILE, saved_rlimit_nofile);
1546         if (saved_rlimit_memlock->rlim_cur != (rlim_t) -1)
1547                 (void) setrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock);
1548
1549         if (switch_root_dir) {
1550                 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1551                  * SIGCHLD for them after deserializing. */
1552                 broadcast_signal(SIGTERM, false, true, arg_default_timeout_stop_usec);
1553
1554                 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1555                 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1556                 if (r < 0)
1557                         log_error_errno(r, "Failed to switch root, trying to continue: %m");
1558         }
1559
1560         args_size = MAX(6, argc+1);
1561         args = newa(const char*, args_size);
1562
1563         if (!switch_root_init) {
1564                 char sfd[DECIMAL_STR_MAX(int) + 1];
1565
1566                 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1567                  * the user didn't specify an explicit init to spawn. */
1568
1569                 assert(arg_serialization);
1570                 assert(fds);
1571
1572                 xsprintf(sfd, "%i", fileno(arg_serialization));
1573
1574                 i = 0;
1575                 args[i++] = SYSTEMD_BINARY_PATH;
1576                 if (switch_root_dir)
1577                         args[i++] = "--switched-root";
1578                 args[i++] = arg_system ? "--system" : "--user";
1579                 args[i++] = "--deserialize";
1580                 args[i++] = sfd;
1581                 args[i++] = NULL;
1582
1583                 assert(i <= args_size);
1584
1585                 /*
1586                  * We want valgrind to print its memory usage summary before reexecution.  Valgrind won't do this is on
1587                  * its own on exec(), but it will do it on exit().  Hence, to ensure we get a summary here, fork() off
1588                  * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1589                  * before proceeding into the exec().
1590                  */
1591                 valgrind_summary_hack();
1592
1593                 (void) execv(args[0], (char* const*) args);
1594                 log_debug_errno(errno, "Failed to execute our own binary, trying fallback: %m");
1595         }
1596
1597         /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1598          * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1599          * doesn't matter.) */
1600
1601         arg_serialization = safe_fclose(arg_serialization);
1602         fds = fdset_free(fds);
1603
1604         /* Reopen the console */
1605         (void) make_console_stdio();
1606
1607         for (j = 1, i = 1; j < (unsigned) argc; j++)
1608                 args[i++] = argv[j];
1609         args[i++] = NULL;
1610         assert(i <= args_size);
1611
1612         /* Reenable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1613         (void) reset_all_signal_handlers();
1614         (void) reset_signal_mask();
1615
1616         if (switch_root_init) {
1617                 args[0] = switch_root_init;
1618                 (void) execv(args[0], (char* const*) args);
1619                 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
1620         }
1621
1622         args[0] = "/sbin/init";
1623         (void) execv(args[0], (char* const*) args);
1624         r = -errno;
1625
1626         manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
1627                               ANSI_HIGHLIGHT_RED "  !!  " ANSI_NORMAL,
1628                               "Failed to execute /sbin/init");
1629
1630         if (r == -ENOENT) {
1631                 log_warning("No /sbin/init, trying fallback");
1632
1633                 args[0] = "/bin/sh";
1634                 args[1] = NULL;
1635                 (void) execv(args[0], (char* const*) args);
1636                 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
1637         } else
1638                 log_warning_errno(r, "Failed to execute /sbin/init, giving up: %m");
1639
1640         *ret_error_message = "Failed to execute fallback shell";
1641 }
1642
1643 static int invoke_main_loop(
1644                 Manager *m,
1645                 bool *ret_reexecute,
1646                 int *ret_retval,                   /* Return parameters relevant for shutting down */
1647                 const char **ret_shutdown_verb,    /* … */
1648                 FDSet **ret_fds,                   /* Return parameters for reexecuting */
1649                 char **ret_switch_root_dir,        /* … */
1650                 char **ret_switch_root_init,       /* … */
1651                 const char **ret_error_message) {
1652
1653         int r;
1654
1655         assert(m);
1656         assert(ret_reexecute);
1657         assert(ret_retval);
1658         assert(ret_shutdown_verb);
1659         assert(ret_fds);
1660         assert(ret_switch_root_dir);
1661         assert(ret_switch_root_init);
1662         assert(ret_error_message);
1663
1664         for (;;) {
1665                 r = manager_loop(m);
1666                 if (r < 0) {
1667                         *ret_error_message = "Failed to run main loop";
1668                         return log_emergency_errno(r, "Failed to run main loop: %m");
1669                 }
1670
1671                 switch (m->exit_code) {
1672
1673                 case MANAGER_RELOAD: {
1674                         LogTarget saved_log_target;
1675                         int saved_log_level;
1676
1677                         log_info("Reloading.");
1678
1679                         /* First, save any overridden log level/target, then parse the configuration file, which might
1680                          * change the log level to new settings. */
1681
1682                         saved_log_level = m->log_level_overridden ? log_get_max_level() : -1;
1683                         saved_log_target = m->log_target_overridden ? log_get_target() : _LOG_TARGET_INVALID;
1684
1685                         r = parse_config_file();
1686                         if (r < 0)
1687                                 log_warning_errno(r, "Failed to parse config file, ignoring: %m");
1688
1689                         set_manager_defaults(m);
1690
1691                         if (saved_log_level >= 0)
1692                                 manager_override_log_level(m, saved_log_level);
1693                         if (saved_log_target >= 0)
1694                                 manager_override_log_target(m, saved_log_target);
1695
1696                         r = manager_reload(m);
1697                         if (r < 0)
1698                                 log_warning_errno(r, "Failed to reload, ignoring: %m");
1699
1700                         break;
1701                 }
1702
1703                 case MANAGER_REEXECUTE:
1704
1705                         r = prepare_reexecute(m, &arg_serialization, ret_fds, false);
1706                         if (r < 0) {
1707                                 *ret_error_message = "Failed to prepare for reexecution";
1708                                 return r;
1709                         }
1710
1711                         log_notice("Reexecuting.");
1712
1713                         *ret_reexecute = true;
1714                         *ret_retval = EXIT_SUCCESS;
1715                         *ret_shutdown_verb = NULL;
1716                         *ret_switch_root_dir = *ret_switch_root_init = NULL;
1717
1718                         return 0;
1719
1720                 case MANAGER_SWITCH_ROOT:
1721                         if (!m->switch_root_init) {
1722                                 r = prepare_reexecute(m, &arg_serialization, ret_fds, true);
1723                                 if (r < 0) {
1724                                         *ret_error_message = "Failed to prepare for reexecution";
1725                                         return r;
1726                                 }
1727                         } else
1728                                 *ret_fds = NULL;
1729
1730                         log_notice("Switching root.");
1731
1732                         *ret_reexecute = true;
1733                         *ret_retval = EXIT_SUCCESS;
1734                         *ret_shutdown_verb = NULL;
1735
1736                         /* Steal the switch root parameters */
1737                         *ret_switch_root_dir = m->switch_root;
1738                         *ret_switch_root_init = m->switch_root_init;
1739                         m->switch_root = m->switch_root_init = NULL;
1740
1741                         return 0;
1742
1743                 case MANAGER_EXIT:
1744
1745                         if (MANAGER_IS_USER(m)) {
1746                                 log_debug("Exit.");
1747
1748                                 *ret_reexecute = false;
1749                                 *ret_retval = m->return_value;
1750                                 *ret_shutdown_verb = NULL;
1751                                 *ret_fds = NULL;
1752                                 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1753
1754                                 return 0;
1755                         }
1756
1757                         _fallthrough_;
1758                 case MANAGER_REBOOT:
1759                 case MANAGER_POWEROFF:
1760                 case MANAGER_HALT:
1761                 case MANAGER_KEXEC: {
1762                         static const char * const table[_MANAGER_EXIT_CODE_MAX] = {
1763                                 [MANAGER_EXIT] = "exit",
1764                                 [MANAGER_REBOOT] = "reboot",
1765                                 [MANAGER_POWEROFF] = "poweroff",
1766                                 [MANAGER_HALT] = "halt",
1767                                 [MANAGER_KEXEC] = "kexec"
1768                         };
1769
1770                         log_notice("Shutting down.");
1771
1772                         *ret_reexecute = false;
1773                         *ret_retval = m->return_value;
1774                         assert_se(*ret_shutdown_verb = table[m->exit_code]);
1775                         *ret_fds = NULL;
1776                         *ret_switch_root_dir = *ret_switch_root_init = NULL;
1777
1778                         return 0;
1779                 }
1780
1781                 default:
1782                         assert_not_reached("Unknown exit code.");
1783                 }
1784         }
1785 }
1786
1787 static void log_execution_mode(bool *ret_first_boot) {
1788         assert(ret_first_boot);
1789
1790         if (arg_system) {
1791                 int v;
1792
1793                 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1794                          arg_action == ACTION_TEST ? "test " : "" );
1795
1796                 v = detect_virtualization();
1797                 if (v > 0)
1798                         log_info("Detected virtualization %s.", virtualization_to_string(v));
1799
1800                 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1801
1802                 if (in_initrd()) {
1803                         *ret_first_boot = false;
1804                         log_info("Running in initial RAM disk.");
1805                 } else {
1806                         /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
1807                          * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
1808                          * doesn't it's unpopulated. This allows container managers and installers to provision a
1809                          * couple of files already. If the container manager wants to provision the machine ID itself
1810                          * it should pass $container_uuid to PID 1. */
1811
1812                         *ret_first_boot = access("/etc/machine-id", F_OK) < 0;
1813                         if (*ret_first_boot)
1814                                 log_info("Running with unpopulated /etc.");
1815                 }
1816         } else {
1817                 if (DEBUG_LOGGING) {
1818                         _cleanup_free_ char *t;
1819
1820                         t = uid_to_name(getuid());
1821                         log_debug(PACKAGE_STRING " running in %suser mode for user " UID_FMT "/%s. (" SYSTEMD_FEATURES ")",
1822                                   arg_action == ACTION_TEST ? " test" : "", getuid(), strna(t));
1823                 }
1824
1825                 *ret_first_boot = false;
1826         }
1827 }
1828
1829 static int initialize_runtime(
1830                 bool skip_setup,
1831                 struct rlimit *saved_rlimit_nofile,
1832                 struct rlimit *saved_rlimit_memlock,
1833                 const char **ret_error_message) {
1834
1835         int r;
1836
1837         assert(ret_error_message);
1838
1839         /* Sets up various runtime parameters. Many of these initializations are conditionalized:
1840          *
1841          * - Some only apply to --system instances
1842          * - Some only apply to --user instances
1843          * - Some only apply when we first start up, but not when we reexecute
1844          */
1845
1846         if (arg_action != ACTION_RUN)
1847                 return 0;
1848
1849         if (arg_system) {
1850                 /* Make sure we leave a core dump without panicing the kernel. */
1851                 install_crash_handler();
1852
1853                 if (!skip_setup) {
1854                         r = mount_cgroup_controllers(arg_join_controllers);
1855                         if (r < 0) {
1856                                 *ret_error_message = "Failed to mount cgroup hierarchies";
1857                                 return r;
1858                         }
1859
1860                         status_welcome();
1861                         hostname_setup();
1862                         machine_id_setup(NULL, arg_machine_id, NULL);
1863                         loopback_setup();
1864                         bump_unix_max_dgram_qlen();
1865                         test_usr();
1866                         write_container_id();
1867                 }
1868
1869                 if (arg_watchdog_device) {
1870                         r = watchdog_set_device(arg_watchdog_device);
1871                         if (r < 0)
1872                                 log_warning_errno(r, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device);
1873                 }
1874
1875                 if (arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY)
1876                         watchdog_set_timeout(&arg_runtime_watchdog);
1877         }
1878
1879         if (arg_timer_slack_nsec != NSEC_INFINITY)
1880                 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1881                         log_warning_errno(errno, "Failed to adjust timer slack, ignoring: %m");
1882
1883         if (arg_system && !cap_test_all(arg_capability_bounding_set)) {
1884                 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
1885                 if (r < 0) {
1886                         *ret_error_message = "Failed to drop capability bounding set of usermode helpers";
1887                         return log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
1888                 }
1889
1890                 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
1891                 if (r < 0) {
1892                         *ret_error_message = "Failed to drop capability bounding set";
1893                         return log_emergency_errno(r, "Failed to drop capability bounding set: %m");
1894                 }
1895         }
1896
1897         if (arg_system && arg_no_new_privs) {
1898                 if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) {
1899                         *ret_error_message = "Failed to disable new privileges";
1900                         return log_emergency_errno(errno, "Failed to disable new privileges: %m");
1901                 }
1902         }
1903
1904         if (arg_syscall_archs) {
1905                 r = enforce_syscall_archs(arg_syscall_archs);
1906                 if (r < 0) {
1907                         *ret_error_message = "Failed to set syscall architectures";
1908                         return r;
1909                 }
1910         }
1911
1912         if (!arg_system)
1913                 /* Become reaper of our children */
1914                 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
1915                         log_warning_errno(errno, "Failed to make us a subreaper: %m");
1916
1917         if (arg_system) {
1918                 /* Bump up RLIMIT_NOFILE for systemd itself */
1919                 (void) bump_rlimit_nofile(saved_rlimit_nofile);
1920                 (void) bump_rlimit_memlock(saved_rlimit_memlock);
1921         }
1922
1923         return 0;
1924 }
1925
1926 static int do_queue_default_job(
1927                 Manager *m,
1928                 const char **ret_error_message) {
1929
1930         _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
1931         Job *default_unit_job;
1932         Unit *target = NULL;
1933         int r;
1934
1935         log_debug("Activating default unit: %s", arg_default_unit);
1936
1937         r = manager_load_startable_unit_or_warn(m, arg_default_unit, NULL, &target);
1938         if (r < 0) {
1939                 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET);
1940
1941                 r = manager_load_startable_unit_or_warn(m, SPECIAL_RESCUE_TARGET, NULL, &target);
1942                 if (r < 0) {
1943                         *ret_error_message = r == -ERFKILL ? "Rescue target masked"
1944                                                            : "Failed to load rescue target";
1945                         return r;
1946                 }
1947         }
1948
1949         assert(target->load_state == UNIT_LOADED);
1950
1951         r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, &error, &default_unit_job);
1952         if (r == -EPERM) {
1953                 log_debug_errno(r, "Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
1954
1955                 sd_bus_error_free(&error);
1956
1957                 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, &error, &default_unit_job);
1958                 if (r < 0) {
1959                         *ret_error_message = "Failed to start default target";
1960                         return log_emergency_errno(r, "Failed to start default target: %s", bus_error_message(&error, r));
1961                 }
1962
1963         } else if (r < 0) {
1964                 *ret_error_message = "Failed to isolate default target";
1965                 return log_emergency_errno(r, "Failed to isolate default target: %s", bus_error_message(&error, r));
1966         }
1967
1968         m->default_unit_job_id = default_unit_job->id;
1969
1970         return 0;
1971 }
1972
1973 static void free_arguments(void) {
1974
1975         /* Frees all arg_* variables, with the exception of arg_serialization */
1976         rlimit_free_all(arg_default_rlimit);
1977
1978         arg_default_unit = mfree(arg_default_unit);
1979         arg_confirm_spawn = mfree(arg_confirm_spawn);
1980         arg_join_controllers = strv_free_free(arg_join_controllers);
1981         arg_default_environment = strv_free(arg_default_environment);
1982         arg_syscall_archs = set_free(arg_syscall_archs);
1983 }
1984
1985 static int load_configuration(int argc, char **argv, const char **ret_error_message) {
1986         int r;
1987
1988         assert(ret_error_message);
1989
1990         arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U);
1991
1992         r = parse_config_file();
1993         if (r < 0) {
1994                 *ret_error_message = "Failed to parse config file";
1995                 return r;
1996         }
1997
1998         if (arg_system) {
1999                 r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
2000                 if (r < 0)
2001                         log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
2002         }
2003
2004         /* Note that this also parses bits from the kernel command line, including "debug". */
2005         log_parse_environment();
2006
2007         r = parse_argv(argc, argv);
2008         if (r < 0) {
2009                 *ret_error_message = "Failed to parse commandline arguments";
2010                 return r;
2011         }
2012
2013         /* Initialize default unit */
2014         if (!arg_default_unit) {
2015                 arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET);
2016                 if (!arg_default_unit) {
2017                         *ret_error_message = "Failed to set default unit";
2018                         return log_oom();
2019                 }
2020         }
2021
2022         /* Initialize the show status setting if it hasn't been set explicitly yet */
2023         if (arg_show_status == _SHOW_STATUS_INVALID)
2024                 arg_show_status = SHOW_STATUS_YES;
2025
2026         return 0;
2027 }
2028
2029 static int safety_checks(void) {
2030
2031         if (getpid_cached() == 1 &&
2032             arg_action != ACTION_RUN) {
2033                 log_error("Unsupported execution mode while PID 1.");
2034                 return -EPERM;
2035         }
2036
2037         if (getpid_cached() == 1 &&
2038             !arg_system) {
2039                 log_error("Can't run --user mode as PID 1.");
2040                 return -EPERM;
2041         }
2042
2043         if (arg_action == ACTION_RUN &&
2044             arg_system &&
2045             getpid_cached() != 1) {
2046                 log_error("Can't run system mode unless PID 1.");
2047                 return -EPERM;
2048         }
2049
2050         if (arg_action == ACTION_TEST &&
2051             geteuid() == 0) {
2052                 log_error("Don't run test mode as root.");
2053                 return -EPERM;
2054         }
2055
2056         if (!arg_system &&
2057             arg_action == ACTION_RUN &&
2058             sd_booted() <= 0) {
2059                 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
2060                 return -EOPNOTSUPP;
2061         }
2062
2063         if (!arg_system &&
2064             arg_action == ACTION_RUN &&
2065             !getenv("XDG_RUNTIME_DIR")) {
2066                 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
2067                 return -EUNATCH;
2068         }
2069
2070         if (arg_system &&
2071             arg_action == ACTION_RUN &&
2072             running_in_chroot() > 0) {
2073                 log_error("Cannot be run in a chroot() environment.");
2074                 return -EOPNOTSUPP;
2075         }
2076
2077         return 0;
2078 }
2079
2080 static int initialize_security(
2081                 bool *loaded_policy,
2082                 dual_timestamp *security_start_timestamp,
2083                 dual_timestamp *security_finish_timestamp,
2084                 const char **ret_error_message) {
2085
2086         int r;
2087
2088         assert(loaded_policy);
2089         assert(security_start_timestamp);
2090         assert(security_finish_timestamp);
2091         assert(ret_error_message);
2092
2093         dual_timestamp_get(security_start_timestamp);
2094
2095         r = mac_selinux_setup(loaded_policy);
2096         if (r < 0) {
2097                 *ret_error_message = "Failed to load SELinux policy";
2098                 return r;
2099         }
2100
2101         r = mac_smack_setup(loaded_policy);
2102         if (r < 0) {
2103                 *ret_error_message = "Failed to load SMACK policy";
2104                 return r;
2105         }
2106
2107         r = ima_setup();
2108         if (r < 0) {
2109                 *ret_error_message = "Failed to load IMA policy";
2110                 return r;
2111         }
2112
2113         dual_timestamp_get(security_finish_timestamp);
2114         return 0;
2115 }
2116
2117 static void test_summary(Manager *m) {
2118         assert(m);
2119
2120         printf("-> By units:\n");
2121         manager_dump_units(m, stdout, "\t");
2122
2123         printf("-> By jobs:\n");
2124         manager_dump_jobs(m, stdout, "\t");
2125 }
2126
2127 static int collect_fds(FDSet **ret_fds, const char **ret_error_message) {
2128         int r;
2129
2130         assert(ret_fds);
2131         assert(ret_error_message);
2132
2133         r = fdset_new_fill(ret_fds);
2134         if (r < 0) {
2135                 *ret_error_message = "Failed to allocate fd set";
2136                 return log_emergency_errno(r, "Failed to allocate fd set: %m");
2137         }
2138
2139         fdset_cloexec(*ret_fds, true);
2140
2141         if (arg_serialization)
2142                 assert_se(fdset_remove(*ret_fds, fileno(arg_serialization)) >= 0);
2143
2144         return 0;
2145 }
2146
2147 static void setup_console_terminal(bool skip_setup) {
2148
2149         if (!arg_system)
2150                 return;
2151
2152         /* Become a session leader if we aren't one yet. */
2153         (void) setsid();
2154
2155         /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2156          * tty. */
2157         (void) release_terminal();
2158
2159         /* Reset the console, but only if this is really init and we are freshly booted */
2160         if (getpid_cached() == 1 && !skip_setup)
2161                 (void) console_setup();
2162 }
2163
2164 static bool early_skip_setup_check(int argc, char *argv[]) {
2165         bool found_deserialize = false;
2166         int i;
2167
2168         /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2169          * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2170          * anyway, even if in that case we also do deserialization. */
2171
2172         for (i = 1; i < argc; i++) {
2173                 if (streq(argv[i], "--switched-root"))
2174                         return false; /* If we switched root, don't skip the setup. */
2175                 else if (streq(argv[i], "--deserialize"))
2176                         found_deserialize = true;
2177         }
2178
2179         return found_deserialize; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2180 }
2181
2182 int main(int argc, char *argv[]) {
2183
2184         dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL, userspace_timestamp = DUAL_TIMESTAMP_NULL, kernel_timestamp = DUAL_TIMESTAMP_NULL,
2185                 security_start_timestamp = DUAL_TIMESTAMP_NULL, security_finish_timestamp = DUAL_TIMESTAMP_NULL;
2186         struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0), saved_rlimit_memlock = RLIMIT_MAKE_CONST((rlim_t) -1);
2187         bool skip_setup, loaded_policy = false, queue_default_job = false, first_boot = false, reexecute = false;
2188         char *switch_root_dir = NULL, *switch_root_init = NULL;
2189         usec_t before_startup, after_startup;
2190         static char systemd[] = "systemd";
2191         char timespan[FORMAT_TIMESPAN_MAX];
2192         const char *shutdown_verb = NULL, *error_message = NULL;
2193         int r, retval = EXIT_FAILURE;
2194         Manager *m = NULL;
2195         FDSet *fds = NULL;
2196
2197         /* SysV compatibility: redirect init → telinit */
2198         redirect_telinit(argc, argv);
2199
2200         /* Take timestamps early on */
2201         dual_timestamp_from_monotonic(&kernel_timestamp, 0);
2202         dual_timestamp_get(&userspace_timestamp);
2203
2204         /* Figure out whether we need to do initialize the system, or if we already did that because we are
2205          * reexecuting */
2206         skip_setup = early_skip_setup_check(argc, argv);
2207
2208         /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2209          * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
2210         program_invocation_short_name = systemd;
2211         (void) prctl(PR_SET_NAME, systemd);
2212
2213         /* Save the original command line */
2214         saved_argv = argv;
2215         saved_argc = argc;
2216
2217         /* Make sure that if the user says "syslog" we actually log to the journal. */
2218         log_set_upgrade_syslog_to_journal(true);
2219
2220         if (getpid_cached() == 1) {
2221                 /* When we run as PID 1 force system mode */
2222                 arg_system = true;
2223
2224                 /* Disable the umask logic */
2225                 umask(0);
2226
2227                 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2228                  * activated yet (even though the log socket for it exists). */
2229                 log_set_prohibit_ipc(true);
2230
2231                 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2232                  * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2233                  * child process right before execve()'ing the actual binary, at a point in time where socket
2234                  * activation stderr/stdout area already set up. */
2235                 log_set_always_reopen_console(true);
2236
2237                 if (detect_container() <= 0) {
2238
2239                         /* Running outside of a container as PID 1 */
2240                         log_set_target(LOG_TARGET_KMSG);
2241                         log_open();
2242
2243                         if (in_initrd())
2244                                 initrd_timestamp = userspace_timestamp;
2245
2246                         if (!skip_setup) {
2247                                 r = mount_setup_early();
2248                                 if (r < 0) {
2249                                         error_message = "Failed to mount early API filesystems";
2250                                         goto finish;
2251                                 }
2252
2253                                 r = initialize_security(
2254                                                 &loaded_policy,
2255                                                 &security_start_timestamp,
2256                                                 &security_finish_timestamp,
2257                                                 &error_message);
2258                                 if (r < 0)
2259                                         goto finish;
2260                         }
2261
2262                         if (mac_selinux_init() < 0) {
2263                                 error_message = "Failed to initialize SELinux policy";
2264                                 goto finish;
2265                         }
2266
2267                         if (!skip_setup)
2268                                 initialize_clock();
2269
2270                         /* Set the default for later on, but don't actually open the logs like this for now. Note that
2271                          * if we are transitioning from the initrd there might still be journal fd open, and we
2272                          * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2273                          * elsewhere. */
2274                         log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
2275
2276                 } else {
2277                         /* Running inside a container, as PID 1 */
2278                         log_set_target(LOG_TARGET_CONSOLE);
2279                         log_open();
2280
2281                         /* For later on, see above... */
2282                         log_set_target(LOG_TARGET_JOURNAL);
2283
2284                         /* clear the kernel timestamp,
2285                          * because we are in a container */
2286                         kernel_timestamp = DUAL_TIMESTAMP_NULL;
2287                 }
2288
2289                 initialize_coredump(skip_setup);
2290
2291                 r = fixup_environment();
2292                 if (r < 0) {
2293                         log_emergency_errno(r, "Failed to fix up PID 1 environment: %m");
2294                         error_message = "Failed to fix up PID1 environment";
2295                         goto finish;
2296                 }
2297
2298         } else {
2299                 /* Running as user instance */
2300                 arg_system = false;
2301                 log_set_target(LOG_TARGET_AUTO);
2302                 log_open();
2303
2304                 /* clear the kernel timestamp,
2305                  * because we are not PID 1 */
2306                 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2307         }
2308
2309         if (arg_system) {
2310                 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2311                  * they never log into the console. */
2312                 log_show_color(colors_enabled());
2313
2314                 r = make_null_stdio();
2315                 if (r < 0)
2316                         log_warning_errno(r, "Failed to redirect standard streams to /dev/null, ignoring: %m");
2317         }
2318
2319         /* Mount /proc, /sys and friends, so that /proc/cmdline and
2320          * /proc/$PID/fd is available. */
2321         if (getpid_cached() == 1) {
2322
2323                 /* Load the kernel modules early. */
2324                 if (!skip_setup)
2325                         kmod_setup();
2326
2327                 r = mount_setup(loaded_policy);
2328                 if (r < 0) {
2329                         error_message = "Failed to mount API filesystems";
2330                         goto finish;
2331                 }
2332         }
2333
2334         /* Reset all signal handlers. */
2335         (void) reset_all_signal_handlers();
2336         (void) ignore_signals(SIGNALS_IGNORE, -1);
2337
2338         r = load_configuration(argc, argv, &error_message);
2339         if (r < 0)
2340                 goto finish;
2341
2342         r = safety_checks();
2343         if (r < 0)
2344                 goto finish;
2345
2346         if (IN_SET(arg_action, ACTION_TEST, ACTION_HELP, ACTION_DUMP_CONFIGURATION_ITEMS, ACTION_DUMP_BUS_PROPERTIES))
2347                 (void) pager_open(arg_no_pager, false);
2348
2349         if (arg_action != ACTION_RUN)
2350                 skip_setup = true;
2351
2352         if (arg_action == ACTION_HELP) {
2353                 retval = help() < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
2354                 goto finish;
2355         } else if (arg_action == ACTION_VERSION) {
2356                 retval = version();
2357                 goto finish;
2358         } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
2359                 unit_dump_config_items(stdout);
2360                 retval = EXIT_SUCCESS;
2361                 goto finish;
2362         } else if (arg_action == ACTION_DUMP_BUS_PROPERTIES) {
2363                 dump_bus_properties(stdout);
2364                 retval = EXIT_SUCCESS;
2365                 goto finish;
2366         }
2367
2368         assert_se(IN_SET(arg_action, ACTION_RUN, ACTION_TEST));
2369
2370         /* Move out of the way, so that we won't block unmounts */
2371         assert_se(chdir("/") == 0);
2372
2373         if (arg_action == ACTION_RUN) {
2374
2375                 /* A core pattern might have been specified via the cmdline.  */
2376                 initialize_core_pattern(skip_setup);
2377
2378                 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2379                 log_close();
2380
2381                 /* Remember open file descriptors for later deserialization */
2382                 r = collect_fds(&fds, &error_message);
2383                 if (r < 0)
2384                         goto finish;
2385
2386                 /* Give up any control of the console, but make sure its initialized. */
2387                 setup_console_terminal(skip_setup);
2388
2389                 /* Open the logging devices, if possible and necessary */
2390                 log_open();
2391         }
2392
2393         log_execution_mode(&first_boot);
2394
2395         r = initialize_runtime(skip_setup,
2396                                &saved_rlimit_nofile,
2397                                &saved_rlimit_memlock,
2398                                &error_message);
2399         if (r < 0)
2400                 goto finish;
2401
2402         r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER,
2403                         arg_action == ACTION_TEST ? MANAGER_TEST_FULL : 0,
2404                         &m);
2405         if (r < 0) {
2406                 log_emergency_errno(r, "Failed to allocate manager object: %m");
2407                 error_message = "Failed to allocate manager object";
2408                 goto finish;
2409         }
2410
2411         m->timestamps[MANAGER_TIMESTAMP_KERNEL] = kernel_timestamp;
2412         m->timestamps[MANAGER_TIMESTAMP_INITRD] = initrd_timestamp;
2413         m->timestamps[MANAGER_TIMESTAMP_USERSPACE] = userspace_timestamp;
2414         m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START)] = security_start_timestamp;
2415         m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH)] = security_finish_timestamp;
2416
2417         set_manager_defaults(m);
2418         set_manager_settings(m);
2419         manager_set_first_boot(m, first_boot);
2420
2421         /* Remember whether we should queue the default job */
2422         queue_default_job = !arg_serialization || arg_switched_root;
2423
2424         before_startup = now(CLOCK_MONOTONIC);
2425
2426         r = manager_startup(m, arg_serialization, fds);
2427         if (r < 0) {
2428                 log_error_errno(r, "Failed to fully start up daemon: %m");
2429                 error_message = "Failed to start up manager";
2430                 goto finish;
2431         }
2432
2433         /* This will close all file descriptors that were opened, but not claimed by any unit. */
2434         fds = fdset_free(fds);
2435         arg_serialization = safe_fclose(arg_serialization);
2436
2437         if (queue_default_job) {
2438                 r = do_queue_default_job(m, &error_message);
2439                 if (r < 0)
2440                         goto finish;
2441         }
2442
2443         after_startup = now(CLOCK_MONOTONIC);
2444
2445         log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
2446                  "Loaded units and determined initial transaction in %s.",
2447                  format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
2448
2449         if (arg_action == ACTION_TEST) {
2450                 test_summary(m);
2451                 retval = EXIT_SUCCESS;
2452                 goto finish;
2453         }
2454
2455         (void) invoke_main_loop(m,
2456                                 &reexecute,
2457                                 &retval,
2458                                 &shutdown_verb,
2459                                 &fds,
2460                                 &switch_root_dir,
2461                                 &switch_root_init,
2462                                 &error_message);
2463
2464 finish:
2465         pager_close();
2466
2467         if (m) {
2468                 arg_shutdown_watchdog = m->shutdown_watchdog;
2469                 m = manager_free(m);
2470         }
2471
2472         free_arguments();
2473         mac_selinux_finish();
2474
2475         if (reexecute)
2476                 do_reexecute(argc, argv,
2477                              &saved_rlimit_nofile,
2478                              &saved_rlimit_memlock,
2479                              fds,
2480                              switch_root_dir,
2481                              switch_root_init,
2482                              &error_message); /* This only returns if reexecution failed */
2483
2484         arg_serialization = safe_fclose(arg_serialization);
2485         fds = fdset_free(fds);
2486
2487 #if HAVE_VALGRIND_VALGRIND_H
2488         /* If we are PID 1 and running under valgrind, then let's exit
2489          * here explicitly. valgrind will only generate nice output on
2490          * exit(), not on exec(), hence let's do the former not the
2491          * latter here. */
2492         if (getpid_cached() == 1 && RUNNING_ON_VALGRIND) {
2493                 /* Cleanup watchdog_device strings for valgrind. We need them
2494                  * in become_shutdown() so normally we cannot free them yet. */
2495                 watchdog_free_device();
2496                 arg_watchdog_device = mfree(arg_watchdog_device);
2497                 return retval;
2498         }
2499 #endif
2500
2501         if (shutdown_verb) {
2502                 r = become_shutdown(shutdown_verb, retval);
2503                 log_error_errno(r, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
2504                 error_message = "Failed to execute shutdown binary";
2505         }
2506
2507         watchdog_free_device();
2508         arg_watchdog_device = mfree(arg_watchdog_device);
2509
2510         if (getpid_cached() == 1) {
2511                 if (error_message)
2512                         manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2513                                               ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
2514                                               "%s, freezing.", error_message);
2515                 freeze_or_reboot();
2516         }
2517
2518         return retval;
2519 }