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 "sysctl-util.h"
  77 #include "terminal-util.h"
  78 #include "umask-util.h"
  79 #include "user-util.h"
  80 #include "util.h"
  81 #include "virt.h"
  82 #include "watchdog.h"
  83
  84 static enum {
  85         ACTION_RUN,
  86         ACTION_HELP,
  87         ACTION_VERSION,
  88         ACTION_TEST,
  89         ACTION_DUMP_CONFIGURATION_ITEMS,
  90         ACTION_DUMP_BUS_PROPERTIES,
  91 } arg_action = ACTION_RUN;
  92 static char *arg_default_unit = NULL;
  93 static bool arg_system = false;
  94 static bool arg_dump_core = true;
  95 static int arg_crash_chvt = -1;
  96 static bool arg_crash_shell = false;
  97 static bool arg_crash_reboot = false;
  98 static char *arg_confirm_spawn = NULL;
  99 static ShowStatus arg_show_status = _SHOW_STATUS_INVALID;
 100 static bool arg_switched_root = false;
 101 static bool arg_no_pager = false;
 102 static bool arg_service_watchdogs = true;
 103 static char ***arg_join_controllers = NULL;
 104 static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL;
 105 static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT;
 106 static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC;
 107 static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC;
 108 static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC;
 109 static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL;
 110 static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST;
 111 static usec_t arg_runtime_watchdog = 0;
 112 static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE;
 113 static char *arg_early_core_pattern = NULL;
 114 static char *arg_watchdog_device = NULL;
 115 static char **arg_default_environment = NULL;
 116 static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {};
 117 static uint64_t arg_capability_bounding_set = CAP_ALL;
 118 static bool arg_no_new_privs = false;
 119 static nsec_t arg_timer_slack_nsec = NSEC_INFINITY;
 120 static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE;
 121 static Set* arg_syscall_archs = NULL;
 122 static FILE* arg_serialization = NULL;
 123 static bool arg_default_cpu_accounting = false;
 124 static bool arg_default_io_accounting = false;
 125 static bool arg_default_ip_accounting = false;
 126 static bool arg_default_blockio_accounting = false;
 127 static bool arg_default_memory_accounting = MEMORY_ACCOUNTING_DEFAULT;
 128 static bool arg_default_tasks_accounting = true;
 129 static uint64_t arg_default_tasks_max = UINT64_MAX;
 130 static sd_id128_t arg_machine_id = {};
 131 static EmergencyAction arg_cad_burst_action = EMERGENCY_ACTION_REBOOT_FORCE;
 132
 133 _noreturn_ static void freeze_or_reboot(void) {
 134
 135         if (arg_crash_reboot) {
 136                 log_notice("Rebooting in 10s...");
 137                 (void) sleep(10);
 138
 139                 log_notice("Rebooting now...");
 140                 (void) reboot(RB_AUTOBOOT);
 141                 log_emergency_errno(errno, "Failed to reboot: %m");
 142         }
 143
 144         log_emergency("Freezing execution.");
 145         freeze();
 146 }
 147
 148 _noreturn_ static void crash(int sig) {
 149         struct sigaction sa;
 150         pid_t pid;
 151
 152         if (getpid_cached() != 1)
 153                 /* Pass this on immediately, if this is not PID 1 */
 154                 (void) raise(sig);
 155         else if (!arg_dump_core)
 156                 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
 157         else {
 158                 sa = (struct sigaction) {
 159                         .sa_handler = nop_signal_handler,
 160                         .sa_flags = SA_NOCLDSTOP|SA_RESTART,
 161                 };
 162
 163                 /* We want to wait for the core process, hence let's enable SIGCHLD */
 164                 (void) sigaction(SIGCHLD, &sa, NULL);
 165
 166                 pid = raw_clone(SIGCHLD);
 167                 if (pid < 0)
 168                         log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
 169                 else if (pid == 0) {
 170                         /* Enable default signal handler for core dump */
 171
 172                         sa = (struct sigaction) {
 173                                 .sa_handler = SIG_DFL,
 174                         };
 175                         (void) sigaction(sig, &sa, NULL);
 176
 177                         /* Don't limit the coredump size */
 178                         (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
 179
 180                         /* Just to be sure... */
 181                         (void) chdir("/");
 182
 183                         /* Raise the signal again */
 184                         pid = raw_getpid();
 185                         (void) kill(pid, sig); /* raise() would kill the parent */
 186
 187                         assert_not_reached("We shouldn't be here...");
 188                         _exit(EXIT_FAILURE);
 189                 } else {
 190                         siginfo_t status;
 191                         int r;
 192
 193                         /* Order things nicely. */
 194                         r = wait_for_terminate(pid, &status);
 195                         if (r < 0)
 196                                 log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
 197                         else if (status.si_code != CLD_DUMPED)
 198                                 log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
 199                                               signal_to_string(sig),
 200                                               pid, sigchld_code_to_string(status.si_code),
 201                                               status.si_status,
 202                                               strna(status.si_code == CLD_EXITED
 203                                                     ? exit_status_to_string(status.si_status, EXIT_STATUS_MINIMAL)
 204                                                     : signal_to_string(status.si_status)));
 205                         else
 206                                 log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
 207                 }
 208         }
 209
 210         if (arg_crash_chvt >= 0)
 211                 (void) chvt(arg_crash_chvt);
 212
 213         sa = (struct sigaction) {
 214                 .sa_handler = SIG_IGN,
 215                 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
 216         };
 217
 218         /* Let the kernel reap children for us */
 219         (void) sigaction(SIGCHLD, &sa, NULL);
 220
 221         if (arg_crash_shell) {
 222                 log_notice("Executing crash shell in 10s...");
 223                 (void) sleep(10);
 224
 225                 pid = raw_clone(SIGCHLD);
 226                 if (pid < 0)
 227                         log_emergency_errno(errno, "Failed to fork off crash shell: %m");
 228                 else if (pid == 0) {
 229                         (void) setsid();
 230                         (void) make_console_stdio();
 231                         (void) execle("/bin/sh", "/bin/sh", NULL, environ);
 232
 233                         log_emergency_errno(errno, "execle() failed: %m");
 234                         _exit(EXIT_FAILURE);
 235                 } else {
 236                         log_info("Spawned crash shell as PID "PID_FMT".", pid);
 237                         (void) wait_for_terminate(pid, NULL);
 238                 }
 239         }
 240
 241         freeze_or_reboot();
 242 }
 243
 244 static void install_crash_handler(void) {
 245         static const struct sigaction sa = {
 246                 .sa_handler = crash,
 247                 .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
 248         };
 249         int r;
 250
 251         /* We ignore the return value here, since, we don't mind if we
 252          * cannot set up a crash handler */
 253         r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
 254         if (r < 0)
 255                 log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
 256 }
 257
 258 static int console_setup(void) {
 259         _cleanup_close_ int tty_fd = -1;
 260         int r;
 261
 262         tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
 263         if (tty_fd < 0)
 264                 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
 265
 266         /* We don't want to force text mode.  plymouth may be showing
 267          * pictures already from initrd. */
 268         r = reset_terminal_fd(tty_fd, false);
 269         if (r < 0)
 270                 return log_error_errno(r, "Failed to reset /dev/console: %m");
 271
 272         return 0;
 273 }
 274
 275 static int parse_crash_chvt(const char *value) {
 276         int b;
 277
 278         if (safe_atoi(value, &arg_crash_chvt) >= 0)
 279                 return 0;
 280
 281         b = parse_boolean(value);
 282         if (b < 0)
 283                 return b;
 284
 285         if (b > 0)
 286                 arg_crash_chvt = 0; /* switch to where kmsg goes */
 287         else
 288                 arg_crash_chvt = -1; /* turn off switching */
 289
 290         return 0;
 291 }
 292
 293 static int parse_confirm_spawn(const char *value, char **console) {
 294         char *s;
 295         int r;
 296
 297         r = value ? parse_boolean(value) : 1;
 298         if (r == 0) {
 299                 *console = NULL;
 300                 return 0;
 301         }
 302
 303         if (r > 0) /* on with default tty */
 304                 s = strdup("/dev/console");
 305         else if (is_path(value)) /* on with fully qualified path */
 306                 s = strdup(value);
 307         else /* on with only a tty file name, not a fully qualified path */
 308                 s = strjoin("/dev/", value);
 309         if (!s)
 310                 return -ENOMEM;
 311
 312         *console = s;
 313         return 0;
 314 }
 315
 316 static int set_machine_id(const char *m) {
 317         sd_id128_t t;
 318         assert(m);
 319
 320         if (sd_id128_from_string(m, &t) < 0)
 321                 return -EINVAL;
 322
 323         if (sd_id128_is_null(t))
 324                 return -EINVAL;
 325
 326         arg_machine_id = t;
 327         return 0;
 328 }
 329
 330 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
 331
 332         int r;
 333
 334         assert(key);
 335
 336         if (STR_IN_SET(key, "systemd.unit", "rd.systemd.unit")) {
 337
 338                 if (proc_cmdline_value_missing(key, value))
 339                         return 0;
 340
 341                 if (!unit_name_is_valid(value, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE))
 342                         log_warning("Unit name specified on %s= is not valid, ignoring: %s", key, value);
 343                 else if (in_initrd() == !!startswith(key, "rd.")) {
 344                         if (free_and_strdup(&arg_default_unit, value) < 0)
 345                                 return log_oom();
 346                 }
 347
 348         } else if (proc_cmdline_key_streq(key, "systemd.dump_core")) {
 349
 350                 r = value ? parse_boolean(value) : true;
 351                 if (r < 0)
 352                         log_warning_errno(r, "Failed to parse dump core switch %s, ignoring: %m", value);
 353                 else
 354                         arg_dump_core = r;
 355
 356         } else if (proc_cmdline_key_streq(key, "systemd.early_core_pattern")) {
 357
 358                 if (proc_cmdline_value_missing(key, value))
 359                         return 0;
 360
 361                 if (path_is_absolute(value))
 362                         (void) parse_path_argument_and_warn(value, false, &arg_early_core_pattern);
 363                 else
 364                         log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value);
 365
 366         } else if (proc_cmdline_key_streq(key, "systemd.crash_chvt")) {
 367
 368                 if (!value)
 369                         arg_crash_chvt = 0; /* turn on */
 370                 else {
 371                         r = parse_crash_chvt(value);
 372                         if (r < 0)
 373                                 log_warning_errno(r, "Failed to parse crash chvt switch %s, ignoring: %m", value);
 374                 }
 375
 376         } else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) {
 377
 378                 r = value ? parse_boolean(value) : true;
 379                 if (r < 0)
 380                         log_warning_errno(r, "Failed to parse crash shell switch %s, ignoring: %m", value);
 381                 else
 382                         arg_crash_shell = r;
 383
 384         } else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) {
 385
 386                 r = value ? parse_boolean(value) : true;
 387                 if (r < 0)
 388                         log_warning_errno(r, "Failed to parse crash reboot switch %s, ignoring: %m", value);
 389                 else
 390                         arg_crash_reboot = r;
 391
 392         } else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) {
 393                 char *s;
 394
 395                 r = parse_confirm_spawn(value, &s);
 396                 if (r < 0)
 397                         log_warning_errno(r, "Failed to parse confirm_spawn switch %s, ignoring: %m", value);
 398                 else
 399                         free_and_replace(arg_confirm_spawn, s);
 400
 401         } else if (proc_cmdline_key_streq(key, "systemd.service_watchdogs")) {
 402
 403                 r = value ? parse_boolean(value) : true;
 404                 if (r < 0)
 405                         log_warning_errno(r, "Failed to parse service watchdog switch %s, ignoring: %m", value);
 406                 else
 407                         arg_service_watchdogs = r;
 408
 409         } else if (proc_cmdline_key_streq(key, "systemd.show_status")) {
 410
 411                 if (value) {
 412                         r = parse_show_status(value, &arg_show_status);
 413                         if (r < 0)
 414                                 log_warning_errno(r, "Failed to parse show status switch %s, ignoring: %m", value);
 415                 } else
 416                         arg_show_status = SHOW_STATUS_YES;
 417
 418         } else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) {
 419
 420                 if (proc_cmdline_value_missing(key, value))
 421                         return 0;
 422
 423                 r = exec_output_from_string(value);
 424                 if (r < 0)
 425                         log_warning_errno(r, "Failed to parse default standard output switch %s, ignoring: %m", value);
 426                 else
 427                         arg_default_std_output = r;
 428
 429         } else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) {
 430
 431                 if (proc_cmdline_value_missing(key, value))
 432                         return 0;
 433
 434                 r = exec_output_from_string(value);
 435                 if (r < 0)
 436                         log_warning_errno(r, "Failed to parse default standard error switch %s, ignoring: %m", value);
 437                 else
 438                         arg_default_std_error = r;
 439
 440         } else if (streq(key, "systemd.setenv")) {
 441
 442                 if (proc_cmdline_value_missing(key, value))
 443                         return 0;
 444
 445                 if (env_assignment_is_valid(value)) {
 446                         char **env;
 447
 448                         env = strv_env_set(arg_default_environment, value);
 449                         if (!env)
 450                                 return log_oom();
 451
 452                         arg_default_environment = env;
 453                 } else
 454                         log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
 455
 456         } else if (proc_cmdline_key_streq(key, "systemd.machine_id")) {
 457
 458                 if (proc_cmdline_value_missing(key, value))
 459                         return 0;
 460
 461                 r = set_machine_id(value);
 462                 if (r < 0)
 463                         log_warning_errno(r, "MachineID '%s' is not valid, ignoring: %m", value);
 464
 465         } else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) {
 466
 467                 if (proc_cmdline_value_missing(key, value))
 468                         return 0;
 469
 470                 r = parse_sec(value, &arg_default_timeout_start_usec);
 471                 if (r < 0)
 472                         log_warning_errno(r, "Failed to parse default start timeout '%s', ignoring: %m", value);
 473
 474                 if (arg_default_timeout_start_usec <= 0)
 475                         arg_default_timeout_start_usec = USEC_INFINITY;
 476
 477         } else if (proc_cmdline_key_streq(key, "systemd.watchdog_device")) {
 478
 479                 if (proc_cmdline_value_missing(key, value))
 480                         return 0;
 481
 482                 (void) parse_path_argument_and_warn(value, false, &arg_watchdog_device);
 483
 484         } else if (streq(key, "quiet") && !value) {
 485
 486                 if (arg_show_status == _SHOW_STATUS_INVALID)
 487                         arg_show_status = SHOW_STATUS_AUTO;
 488
 489         } else if (streq(key, "debug") && !value) {
 490
 491                 /* Note that log_parse_environment() handles 'debug'
 492                  * too, and sets the log level to LOG_DEBUG. */
 493
 494                 if (detect_container() > 0)
 495                         log_set_target(LOG_TARGET_CONSOLE);
 496
 497         } else if (!value) {
 498                 const char *target;
 499
 500                 /* SysV compatibility */
 501                 target = runlevel_to_target(key);
 502                 if (target)
 503                         return free_and_strdup(&arg_default_unit, target);
 504         }
 505
 506         return 0;
 507 }
 508
 509 #define DEFINE_SETTER(name, func, descr)                              \
 510         static int name(const char *unit,                             \
 511                         const char *filename,                         \
 512                         unsigned line,                                \
 513                         const char *section,                          \
 514                         unsigned section_line,                        \
 515                         const char *lvalue,                           \
 516                         int ltype,                                    \
 517                         const char *rvalue,                           \
 518                         void *data,                                   \
 519                         void *userdata) {                             \
 520                                                                       \
 521                 int r;                                                \
 522                                                                       \
 523                 assert(filename);                                     \
 524                 assert(lvalue);                                       \
 525                 assert(rvalue);                                       \
 526                                                                       \
 527                 r = func(rvalue);                                     \
 528                 if (r < 0)                                            \
 529                         log_syntax(unit, LOG_ERR, filename, line, r,  \
 530                                    "Invalid " descr "'%s': %m",       \
 531                                    rvalue);                           \
 532                                                                       \
 533                 return 0;                                             \
 534         }
 535
 536 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level");
 537 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target");
 538 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" );
 539 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location");
 540
 541 static int config_parse_cpu_affinity2(
 542                 const char *unit,
 543                 const char *filename,
 544                 unsigned line,
 545                 const char *section,
 546                 unsigned section_line,
 547                 const char *lvalue,
 548                 int ltype,
 549                 const char *rvalue,
 550                 void *data,
 551                 void *userdata) {
 552
 553         _cleanup_cpu_free_ cpu_set_t *c = NULL;
 554         int ncpus;
 555
 556         ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue);
 557         if (ncpus < 0)
 558                 return ncpus;
 559
 560         if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
 561                 log_warning_errno(errno, "Failed to set CPU affinity: %m");
 562
 563         return 0;
 564 }
 565
 566 static int config_parse_show_status(
 567                 const char* unit,
 568                 const char *filename,
 569                 unsigned line,
 570                 const char *section,
 571                 unsigned section_line,
 572                 const char *lvalue,
 573                 int ltype,
 574                 const char *rvalue,
 575                 void *data,
 576                 void *userdata) {
 577
 578         int k;
 579         ShowStatus *b = data;
 580
 581         assert(filename);
 582         assert(lvalue);
 583         assert(rvalue);
 584         assert(data);
 585
 586         k = parse_show_status(rvalue, b);
 587         if (k < 0) {
 588                 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
 589                 return 0;
 590         }
 591
 592         return 0;
 593 }
 594
 595 static int config_parse_output_restricted(
 596                 const char* unit,
 597                 const char *filename,
 598                 unsigned line,
 599                 const char *section,
 600                 unsigned section_line,
 601                 const char *lvalue,
 602                 int ltype,
 603                 const char *rvalue,
 604                 void *data,
 605                 void *userdata) {
 606
 607         ExecOutput t, *eo = data;
 608
 609         assert(filename);
 610         assert(lvalue);
 611         assert(rvalue);
 612         assert(data);
 613
 614         t = exec_output_from_string(rvalue);
 615         if (t < 0) {
 616                 log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse output type, ignoring: %s", rvalue);
 617                 return 0;
 618         }
 619
 620         if (IN_SET(t, EXEC_OUTPUT_SOCKET, EXEC_OUTPUT_NAMED_FD, EXEC_OUTPUT_FILE, EXEC_OUTPUT_FILE_APPEND)) {
 621                 log_syntax(unit, LOG_ERR, filename, line, 0, "Standard output types socket, fd:, file:, append: are not supported as defaults, ignoring: %s", rvalue);
 622                 return 0;
 623         }
 624
 625         *eo = t;
 626         return 0;
 627 }
 628
 629 static int config_parse_crash_chvt(
 630                 const char* unit,
 631                 const char *filename,
 632                 unsigned line,
 633                 const char *section,
 634                 unsigned section_line,
 635                 const char *lvalue,
 636                 int ltype,
 637                 const char *rvalue,
 638                 void *data,
 639                 void *userdata) {
 640
 641         int r;
 642
 643         assert(filename);
 644         assert(lvalue);
 645         assert(rvalue);
 646
 647         r = parse_crash_chvt(rvalue);
 648         if (r < 0) {
 649                 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
 650                 return 0;
 651         }
 652
 653         return 0;
 654 }
 655
 656 static int parse_config_file(void) {
 657
 658         const ConfigTableItem items[] = {
 659                 { "Manager", "LogLevel",                  config_parse_level2,           0, NULL                                   },
 660                 { "Manager", "LogTarget",                 config_parse_target,           0, NULL                                   },
 661                 { "Manager", "LogColor",                  config_parse_color,            0, NULL                                   },
 662                 { "Manager", "LogLocation",               config_parse_location,         0, NULL                                   },
 663                 { "Manager", "DumpCore",                  config_parse_bool,             0, &arg_dump_core                         },
 664                 { "Manager", "CrashChVT", /* legacy */    config_parse_crash_chvt,       0, NULL                                   },
 665                 { "Manager", "CrashChangeVT",             config_parse_crash_chvt,       0, NULL                                   },
 666                 { "Manager", "CrashShell",                config_parse_bool,             0, &arg_crash_shell                       },
 667                 { "Manager", "CrashReboot",               config_parse_bool,             0, &arg_crash_reboot                      },
 668                 { "Manager", "ShowStatus",                config_parse_show_status,      0, &arg_show_status                       },
 669                 { "Manager", "CPUAffinity",               config_parse_cpu_affinity2,    0, NULL                                   },
 670                 { "Manager", "JoinControllers",           config_parse_join_controllers, 0, &arg_join_controllers                  },
 671                 { "Manager", "RuntimeWatchdogSec",        config_parse_sec,              0, &arg_runtime_watchdog                  },
 672                 { "Manager", "ShutdownWatchdogSec",       config_parse_sec,              0, &arg_shutdown_watchdog                 },
 673                 { "Manager", "WatchdogDevice",            config_parse_path,             0, &arg_watchdog_device                   },
 674                 { "Manager", "CapabilityBoundingSet",     config_parse_capability_set,   0, &arg_capability_bounding_set           },
 675                 { "Manager", "NoNewPrivileges",           config_parse_bool,             0, &arg_no_new_privs                      },
 676 #if HAVE_SECCOMP
 677                 { "Manager", "SystemCallArchitectures",   config_parse_syscall_archs,    0, &arg_syscall_archs                     },
 678 #endif
 679                 { "Manager", "TimerSlackNSec",            config_parse_nsec,             0, &arg_timer_slack_nsec                  },
 680                 { "Manager", "DefaultTimerAccuracySec",   config_parse_sec,              0, &arg_default_timer_accuracy_usec       },
 681                 { "Manager", "DefaultStandardOutput",     config_parse_output_restricted,0, &arg_default_std_output                },
 682                 { "Manager", "DefaultStandardError",      config_parse_output_restricted,0, &arg_default_std_error                 },
 683                 { "Manager", "DefaultTimeoutStartSec",    config_parse_sec,              0, &arg_default_timeout_start_usec        },
 684                 { "Manager", "DefaultTimeoutStopSec",     config_parse_sec,              0, &arg_default_timeout_stop_usec         },
 685                 { "Manager", "DefaultRestartSec",         config_parse_sec,              0, &arg_default_restart_usec              },
 686                 { "Manager", "DefaultStartLimitInterval", config_parse_sec,              0, &arg_default_start_limit_interval      }, /* obsolete alias */
 687                 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec,            0, &arg_default_start_limit_interval      },
 688                 { "Manager", "DefaultStartLimitBurst",    config_parse_unsigned,         0, &arg_default_start_limit_burst         },
 689                 { "Manager", "DefaultEnvironment",        config_parse_environ,          0, &arg_default_environment               },
 690                 { "Manager", "DefaultLimitCPU",           config_parse_rlimit,           RLIMIT_CPU, arg_default_rlimit            },
 691                 { "Manager", "DefaultLimitFSIZE",         config_parse_rlimit,           RLIMIT_FSIZE, arg_default_rlimit          },
 692                 { "Manager", "DefaultLimitDATA",          config_parse_rlimit,           RLIMIT_DATA, arg_default_rlimit           },
 693                 { "Manager", "DefaultLimitSTACK",         config_parse_rlimit,           RLIMIT_STACK, arg_default_rlimit          },
 694                 { "Manager", "DefaultLimitCORE",          config_parse_rlimit,           RLIMIT_CORE, arg_default_rlimit           },
 695                 { "Manager", "DefaultLimitRSS",           config_parse_rlimit,           RLIMIT_RSS, arg_default_rlimit            },
 696                 { "Manager", "DefaultLimitNOFILE",        config_parse_rlimit,           RLIMIT_NOFILE, arg_default_rlimit         },
 697                 { "Manager", "DefaultLimitAS",            config_parse_rlimit,           RLIMIT_AS, arg_default_rlimit             },
 698                 { "Manager", "DefaultLimitNPROC",         config_parse_rlimit,           RLIMIT_NPROC, arg_default_rlimit          },
 699                 { "Manager", "DefaultLimitMEMLOCK",       config_parse_rlimit,           RLIMIT_MEMLOCK, arg_default_rlimit        },
 700                 { "Manager", "DefaultLimitLOCKS",         config_parse_rlimit,           RLIMIT_LOCKS, arg_default_rlimit          },
 701                 { "Manager", "DefaultLimitSIGPENDING",    config_parse_rlimit,           RLIMIT_SIGPENDING, arg_default_rlimit     },
 702                 { "Manager", "DefaultLimitMSGQUEUE",      config_parse_rlimit,           RLIMIT_MSGQUEUE, arg_default_rlimit       },
 703                 { "Manager", "DefaultLimitNICE",          config_parse_rlimit,           RLIMIT_NICE, arg_default_rlimit           },
 704                 { "Manager", "DefaultLimitRTPRIO",        config_parse_rlimit,           RLIMIT_RTPRIO, arg_default_rlimit         },
 705                 { "Manager", "DefaultLimitRTTIME",        config_parse_rlimit,           RLIMIT_RTTIME, arg_default_rlimit         },
 706                 { "Manager", "DefaultCPUAccounting",      config_parse_bool,             0, &arg_default_cpu_accounting            },
 707                 { "Manager", "DefaultIOAccounting",       config_parse_bool,             0, &arg_default_io_accounting             },
 708                 { "Manager", "DefaultIPAccounting",       config_parse_bool,             0, &arg_default_ip_accounting             },
 709                 { "Manager", "DefaultBlockIOAccounting",  config_parse_bool,             0, &arg_default_blockio_accounting        },
 710                 { "Manager", "DefaultMemoryAccounting",   config_parse_bool,             0, &arg_default_memory_accounting         },
 711                 { "Manager", "DefaultTasksAccounting",    config_parse_bool,             0, &arg_default_tasks_accounting          },
 712                 { "Manager", "DefaultTasksMax",           config_parse_tasks_max,        0, &arg_default_tasks_max                 },
 713                 { "Manager", "CtrlAltDelBurstAction",     config_parse_emergency_action, 0, &arg_cad_burst_action                  },
 714                 {}
 715         };
 716
 717         const char *fn, *conf_dirs_nulstr;
 718
 719         fn = arg_system ?
 720                 PKGSYSCONFDIR "/system.conf" :
 721                 PKGSYSCONFDIR "/user.conf";
 722
 723         conf_dirs_nulstr = arg_system ?
 724                 CONF_PATHS_NULSTR("systemd/system.conf.d") :
 725                 CONF_PATHS_NULSTR("systemd/user.conf.d");
 726
 727         (void) config_parse_many_nulstr(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, CONFIG_PARSE_WARN, NULL);
 728
 729         /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
 730          * like everywhere else. */
 731         if (arg_default_timeout_start_usec <= 0)
 732                 arg_default_timeout_start_usec = USEC_INFINITY;
 733         if (arg_default_timeout_stop_usec <= 0)
 734                 arg_default_timeout_stop_usec = USEC_INFINITY;
 735
 736         return 0;
 737 }
 738
 739 static void set_manager_defaults(Manager *m) {
 740
 741         assert(m);
 742
 743         /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
 744          * affect the manager itself, but are just what newly allocated units will have set if they haven't set
 745          * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
 746
 747         m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
 748         m->default_std_output = arg_default_std_output;
 749         m->default_std_error = arg_default_std_error;
 750         m->default_timeout_start_usec = arg_default_timeout_start_usec;
 751         m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
 752         m->default_restart_usec = arg_default_restart_usec;
 753         m->default_start_limit_interval = arg_default_start_limit_interval;
 754         m->default_start_limit_burst = arg_default_start_limit_burst;
 755         m->default_cpu_accounting = arg_default_cpu_accounting;
 756         m->default_io_accounting = arg_default_io_accounting;
 757         m->default_ip_accounting = arg_default_ip_accounting;
 758         m->default_blockio_accounting = arg_default_blockio_accounting;
 759         m->default_memory_accounting = arg_default_memory_accounting;
 760         m->default_tasks_accounting = arg_default_tasks_accounting;
 761         m->default_tasks_max = arg_default_tasks_max;
 762
 763         manager_set_default_rlimits(m, arg_default_rlimit);
 764         manager_environment_add(m, NULL, arg_default_environment);
 765 }
 766
 767 static void set_manager_settings(Manager *m) {
 768
 769         assert(m);
 770
 771         /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
 772          * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
 773
 774         m->confirm_spawn = arg_confirm_spawn;
 775         m->service_watchdogs = arg_service_watchdogs;
 776         m->runtime_watchdog = arg_runtime_watchdog;
 777         m->shutdown_watchdog = arg_shutdown_watchdog;
 778         m->cad_burst_action = arg_cad_burst_action;
 779
 780         manager_set_show_status(m, arg_show_status);
 781 }
 782
 783 static int parse_argv(int argc, char *argv[]) {
 784         enum {
 785                 ARG_LOG_LEVEL = 0x100,
 786                 ARG_LOG_TARGET,
 787                 ARG_LOG_COLOR,
 788                 ARG_LOG_LOCATION,
 789                 ARG_UNIT,
 790                 ARG_SYSTEM,
 791                 ARG_USER,
 792                 ARG_TEST,
 793                 ARG_NO_PAGER,
 794                 ARG_VERSION,
 795                 ARG_DUMP_CONFIGURATION_ITEMS,
 796                 ARG_DUMP_BUS_PROPERTIES,
 797                 ARG_DUMP_CORE,
 798                 ARG_CRASH_CHVT,
 799                 ARG_CRASH_SHELL,
 800                 ARG_CRASH_REBOOT,
 801                 ARG_CONFIRM_SPAWN,
 802                 ARG_SHOW_STATUS,
 803                 ARG_DESERIALIZE,
 804                 ARG_SWITCHED_ROOT,
 805                 ARG_DEFAULT_STD_OUTPUT,
 806                 ARG_DEFAULT_STD_ERROR,
 807                 ARG_MACHINE_ID,
 808                 ARG_SERVICE_WATCHDOGS,
 809         };
 810
 811         static const struct option options[] = {
 812                 { "log-level",                required_argument, NULL, ARG_LOG_LEVEL                },
 813                 { "log-target",               required_argument, NULL, ARG_LOG_TARGET               },
 814                 { "log-color",                optional_argument, NULL, ARG_LOG_COLOR                },
 815                 { "log-location",             optional_argument, NULL, ARG_LOG_LOCATION             },
 816                 { "unit",                     required_argument, NULL, ARG_UNIT                     },
 817                 { "system",                   no_argument,       NULL, ARG_SYSTEM                   },
 818                 { "user",                     no_argument,       NULL, ARG_USER                     },
 819                 { "test",                     no_argument,       NULL, ARG_TEST                     },
 820                 { "no-pager",                 no_argument,       NULL, ARG_NO_PAGER                 },
 821                 { "help",                     no_argument,       NULL, 'h'                          },
 822                 { "version",                  no_argument,       NULL, ARG_VERSION                  },
 823                 { "dump-configuration-items", no_argument,       NULL, ARG_DUMP_CONFIGURATION_ITEMS },
 824                 { "dump-bus-properties",      no_argument,       NULL, ARG_DUMP_BUS_PROPERTIES      },
 825                 { "dump-core",                optional_argument, NULL, ARG_DUMP_CORE                },
 826                 { "crash-chvt",               required_argument, NULL, ARG_CRASH_CHVT               },
 827                 { "crash-shell",              optional_argument, NULL, ARG_CRASH_SHELL              },
 828                 { "crash-reboot",             optional_argument, NULL, ARG_CRASH_REBOOT             },
 829                 { "confirm-spawn",            optional_argument, NULL, ARG_CONFIRM_SPAWN            },
 830                 { "show-status",              optional_argument, NULL, ARG_SHOW_STATUS              },
 831                 { "deserialize",              required_argument, NULL, ARG_DESERIALIZE              },
 832                 { "switched-root",            no_argument,       NULL, ARG_SWITCHED_ROOT            },
 833                 { "default-standard-output",  required_argument, NULL, ARG_DEFAULT_STD_OUTPUT,      },
 834                 { "default-standard-error",   required_argument, NULL, ARG_DEFAULT_STD_ERROR,       },
 835                 { "machine-id",               required_argument, NULL, ARG_MACHINE_ID               },
 836                 { "service-watchdogs",        required_argument, NULL, ARG_SERVICE_WATCHDOGS        },
 837                 {}
 838         };
 839
 840         int c, r;
 841
 842         assert(argc >= 1);
 843         assert(argv);
 844
 845         if (getpid_cached() == 1)
 846                 opterr = 0;
 847
 848         while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
 849
 850                 switch (c) {
 851
 852                 case ARG_LOG_LEVEL:
 853                         r = log_set_max_level_from_string(optarg);
 854                         if (r < 0)
 855                                 return log_error_errno(r, "Failed to parse log level \"%s\": %m", optarg);
 856
 857                         break;
 858
 859                 case ARG_LOG_TARGET:
 860                         r = log_set_target_from_string(optarg);
 861                         if (r < 0)
 862                                 return log_error_errno(r, "Failed to parse log target \"%s\": %m", optarg);
 863
 864                         break;
 865
 866                 case ARG_LOG_COLOR:
 867
 868                         if (optarg) {
 869                                 r = log_show_color_from_string(optarg);
 870                                 if (r < 0)
 871                                         return log_error_errno(r, "Failed to parse log color setting \"%s\": %m",
 872                                                                optarg);
 873                         } else
 874                                 log_show_color(true);
 875
 876                         break;
 877
 878                 case ARG_LOG_LOCATION:
 879                         if (optarg) {
 880                                 r = log_show_location_from_string(optarg);
 881                                 if (r < 0)
 882                                         return log_error_errno(r, "Failed to parse log location setting \"%s\": %m",
 883                                                                optarg);
 884                         } else
 885                                 log_show_location(true);
 886
 887                         break;
 888
 889                 case ARG_DEFAULT_STD_OUTPUT:
 890                         r = exec_output_from_string(optarg);
 891                         if (r < 0)
 892                                 return log_error_errno(r, "Failed to parse default standard output setting \"%s\": %m",
 893                                                        optarg);
 894                         arg_default_std_output = r;
 895                         break;
 896
 897                 case ARG_DEFAULT_STD_ERROR:
 898                         r = exec_output_from_string(optarg);
 899                         if (r < 0)
 900                                 return log_error_errno(r, "Failed to parse default standard error output setting \"%s\": %m",
 901                                                        optarg);
 902                         arg_default_std_error = r;
 903                         break;
 904
 905                 case ARG_UNIT:
 906                         r = free_and_strdup(&arg_default_unit, optarg);
 907                         if (r < 0)
 908                                 return log_error_errno(r, "Failed to set default unit \"%s\": %m", optarg);
 909
 910                         break;
 911
 912                 case ARG_SYSTEM:
 913                         arg_system = true;
 914                         break;
 915
 916                 case ARG_USER:
 917                         arg_system = false;
 918                         break;
 919
 920                 case ARG_TEST:
 921                         arg_action = ACTION_TEST;
 922                         break;
 923
 924                 case ARG_NO_PAGER:
 925                         arg_no_pager = true;
 926                         break;
 927
 928                 case ARG_VERSION:
 929                         arg_action = ACTION_VERSION;
 930                         break;
 931
 932                 case ARG_DUMP_CONFIGURATION_ITEMS:
 933                         arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
 934                         break;
 935
 936                 case ARG_DUMP_BUS_PROPERTIES:
 937                         arg_action = ACTION_DUMP_BUS_PROPERTIES;
 938                         break;
 939
 940                 case ARG_DUMP_CORE:
 941                         if (!optarg)
 942                                 arg_dump_core = true;
 943                         else {
 944                                 r = parse_boolean(optarg);
 945                                 if (r < 0)
 946                                         return log_error_errno(r, "Failed to parse dump core boolean: \"%s\": %m",
 947                                                                optarg);
 948                                 arg_dump_core = r;
 949                         }
 950                         break;
 951
 952                 case ARG_CRASH_CHVT:
 953                         r = parse_crash_chvt(optarg);
 954                         if (r < 0)
 955                                 return log_error_errno(r, "Failed to parse crash virtual terminal index: \"%s\": %m",
 956                                                        optarg);
 957                         break;
 958
 959                 case ARG_CRASH_SHELL:
 960                         if (!optarg)
 961                                 arg_crash_shell = true;
 962                         else {
 963                                 r = parse_boolean(optarg);
 964                                 if (r < 0)
 965                                         return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
 966                                                                optarg);
 967                                 arg_crash_shell = r;
 968                         }
 969                         break;
 970
 971                 case ARG_CRASH_REBOOT:
 972                         if (!optarg)
 973                                 arg_crash_reboot = true;
 974                         else {
 975                                 r = parse_boolean(optarg);
 976                                 if (r < 0)
 977                                         return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
 978                                                                optarg);
 979                                 arg_crash_reboot = r;
 980                         }
 981                         break;
 982
 983                 case ARG_CONFIRM_SPAWN:
 984                         arg_confirm_spawn = mfree(arg_confirm_spawn);
 985
 986                         r = parse_confirm_spawn(optarg, &arg_confirm_spawn);
 987                         if (r < 0)
 988                                 return log_error_errno(r, "Failed to parse confirm spawn option: \"%s\": %m",
 989                                                        optarg);
 990                         break;
 991
 992                 case ARG_SERVICE_WATCHDOGS:
 993                         r = parse_boolean(optarg);
 994                         if (r < 0)
 995                                 return log_error_errno(r, "Failed to parse service watchdogs boolean: \"%s\": %m",
 996                                                        optarg);
 997                         arg_service_watchdogs = r;
 998                         break;
 999
1000                 case ARG_SHOW_STATUS:
1001                         if (optarg) {
1002                                 r = parse_show_status(optarg, &arg_show_status);
1003                                 if (r < 0)
1004                                         return log_error_errno(r, "Failed to parse show status boolean: \"%s\": %m",
1005                                                                optarg);
1006                         } else
1007                                 arg_show_status = SHOW_STATUS_YES;
1008                         break;
1009
1010                 case ARG_DESERIALIZE: {
1011                         int fd;
1012                         FILE *f;
1013
1014                         r = safe_atoi(optarg, &fd);
1015                         if (r < 0)
1016                                 log_error_errno(r, "Failed to parse deserialize option \"%s\": %m", optarg);
1017                         if (fd < 0) {
1018                                 log_error("Invalid deserialize fd: %d", fd);
1019                                 return -EINVAL;
1020                         }
1021
1022                         (void) fd_cloexec(fd, true);
1023
1024                         f = fdopen(fd, "r");
1025                         if (!f)
1026                                 return log_error_errno(errno, "Failed to open serialization fd %d: %m", fd);
1027
1028                         safe_fclose(arg_serialization);
1029                         arg_serialization = f;
1030
1031                         break;
1032                 }
1033
1034                 case ARG_SWITCHED_ROOT:
1035                         arg_switched_root = true;
1036                         break;
1037
1038                 case ARG_MACHINE_ID:
1039                         r = set_machine_id(optarg);
1040                         if (r < 0)
1041                                 return log_error_errno(r, "MachineID '%s' is not valid: %m", optarg);
1042                         break;
1043
1044                 case 'h':
1045                         arg_action = ACTION_HELP;
1046                         break;
1047
1048                 case 'D':
1049                         log_set_max_level(LOG_DEBUG);
1050                         break;
1051
1052                 case 'b':
1053                 case 's':
1054                 case 'z':
1055                         /* Just to eat away the sysvinit kernel
1056                          * cmdline args without getopt() error
1057                          * messages that we'll parse in
1058                          * parse_proc_cmdline_word() or ignore. */
1059
1060                 case '?':
1061                         if (getpid_cached() != 1)
1062                                 return -EINVAL;
1063                         else
1064                                 return 0;
1065
1066                 default:
1067                         assert_not_reached("Unhandled option code.");
1068                 }
1069
1070         if (optind < argc && getpid_cached() != 1) {
1071                 /* Hmm, when we aren't run as init system
1072                  * let's complain about excess arguments */
1073
1074                 log_error("Excess arguments.");
1075                 return -EINVAL;
1076         }
1077
1078         return 0;
1079 }
1080
1081 static int help(void) {
1082         _cleanup_free_ char *link = NULL;
1083         int r;
1084
1085         r = terminal_urlify_man("systemd", "1", &link);
1086         if (r < 0)
1087                 return log_oom();
1088
1089         printf("%s [OPTIONS...]\n\n"
1090                "Starts up and maintains the system or user services.\n\n"
1091                "  -h --help                      Show this help\n"
1092                "     --version                   Show version\n"
1093                "     --test                      Determine startup sequence, dump it and exit\n"
1094                "     --no-pager                  Do not pipe output into a pager\n"
1095                "     --dump-configuration-items  Dump understood unit configuration items\n"
1096                "     --dump-bus-properties       Dump exposed bus properties\n"
1097                "     --unit=UNIT                 Set default unit\n"
1098                "     --system                    Run a system instance, even if PID != 1\n"
1099                "     --user                      Run a user instance\n"
1100                "     --dump-core[=BOOL]          Dump core on crash\n"
1101                "     --crash-vt=NR               Change to specified VT on crash\n"
1102                "     --crash-reboot[=BOOL]       Reboot on crash\n"
1103                "     --crash-shell[=BOOL]        Run shell on crash\n"
1104                "     --confirm-spawn[=BOOL]      Ask for confirmation when spawning processes\n"
1105                "     --show-status[=BOOL]        Show status updates on the console during bootup\n"
1106                "     --log-target=TARGET         Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1107                "     --log-level=LEVEL           Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1108                "     --log-color[=BOOL]          Highlight important log messages\n"
1109                "     --log-location[=BOOL]       Include code location in log messages\n"
1110                "     --default-standard-output=  Set default standard output for services\n"
1111                "     --default-standard-error=   Set default standard error output for services\n"
1112                "\nSee the %s for details.\n"
1113                , program_invocation_short_name
1114                , link
1115         );
1116
1117         return 0;
1118 }
1119
1120 static int prepare_reexecute(
1121                 Manager *m,
1122                 FILE **ret_f,
1123                 FDSet **ret_fds,
1124                 bool switching_root) {
1125
1126         _cleanup_fdset_free_ FDSet *fds = NULL;
1127         _cleanup_fclose_ FILE *f = NULL;
1128         int r;
1129
1130         assert(m);
1131         assert(ret_f);
1132         assert(ret_fds);
1133
1134         r = manager_open_serialization(m, &f);
1135         if (r < 0)
1136                 return log_error_errno(r, "Failed to create serialization file: %m");
1137
1138         /* Make sure nothing is really destructed when we shut down */
1139         m->n_reloading++;
1140         bus_manager_send_reloading(m, true);
1141
1142         fds = fdset_new();
1143         if (!fds)
1144                 return log_oom();
1145
1146         r = manager_serialize(m, f, fds, switching_root);
1147         if (r < 0)
1148                 return r;
1149
1150         if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1151                 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1152
1153         r = fd_cloexec(fileno(f), false);
1154         if (r < 0)
1155                 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1156
1157         r = fdset_cloexec(fds, false);
1158         if (r < 0)
1159                 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1160
1161         *ret_f = TAKE_PTR(f);
1162         *ret_fds = TAKE_PTR(fds);
1163
1164         return 0;
1165 }
1166
1167 static void bump_file_max_and_nr_open(void) {
1168
1169         /* Let's bump fs.file-max and fs.nr_open to their respective maximums. On current kernels large numbers of file
1170          * descriptors are no longer a performance problem and their memory is properly tracked by memcg, thus counting
1171          * them and limiting them in another two layers of limits is unnecessary and just complicates things. This
1172          * function hence turns off 2 of the 4 levels of limits on file descriptors, and makes RLIMIT_NOLIMIT (soft +
1173          * hard) the only ones that really matter. */
1174
1175 #if BUMP_PROC_SYS_FS_FILE_MAX || BUMP_PROC_SYS_FS_NR_OPEN
1176         _cleanup_free_ char *t = NULL;
1177         int r;
1178 #endif
1179
1180 #if BUMP_PROC_SYS_FS_FILE_MAX
1181         /* I so wanted to use STRINGIFY(ULONG_MAX) here, but alas we can't as glibc/gcc define that as
1182          * "(0x7fffffffffffffffL * 2UL + 1UL)". Seriously. 😢 */
1183         if (asprintf(&t, "%lu\n", ULONG_MAX) < 0) {
1184                 log_oom();
1185                 return;
1186         }
1187
1188         r = sysctl_write("fs/file-max", t);
1189         if (r < 0)
1190                 log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.file-max, ignoring: %m");
1191 #endif
1192
1193 #if BUMP_PROC_SYS_FS_FILE_MAX && BUMP_PROC_SYS_FS_NR_OPEN
1194         t = mfree(t);
1195 #endif
1196
1197 #if BUMP_PROC_SYS_FS_NR_OPEN
1198         int v = INT_MAX;
1199
1200         /* Arg! The kernel enforces maximum and minimum values on the fs.nr_open, but we don't really know what they
1201          * are. The expression by which the maximum is determined is dependent on the architecture, and is something we
1202          * don't really want to copy to userspace, as it is dependent on implementation details of the kernel. Since
1203          * the kernel doesn't expose the maximum value to us, we can only try and hope. Hence, let's start with
1204          * INT_MAX, and then keep halving the value until we find one that works. Ugly? Yes, absolutely, but kernel
1205          * APIs are kernel APIs, so what do can we do... 🤯 */
1206
1207         for (;;) {
1208                 int k;
1209
1210                 v &= ~(__SIZEOF_POINTER__ - 1); /* Round down to next multiple of the pointer size */
1211                 if (v < 1024) {
1212                         log_warning("Can't bump fs.nr_open, value too small.");
1213                         break;
1214                 }
1215
1216                 k = read_nr_open();
1217                 if (k < 0) {
1218                         log_error_errno(k, "Failed to read fs.nr_open: %m");
1219                         break;
1220                 }
1221                 if (k >= v) { /* Already larger */
1222                         log_debug("Skipping bump, value is already larger.");
1223                         break;
1224                 }
1225
1226                 if (asprintf(&t, "%i\n", v) < 0) {
1227                         log_oom();
1228                         return;
1229                 }
1230
1231                 r = sysctl_write("fs/nr_open", t);
1232                 t = mfree(t);
1233                 if (r == -EINVAL) {
1234                         log_debug("Couldn't write fs.nr_open as %i, halving it.", v);
1235                         v /= 2;
1236                         continue;
1237                 }
1238                 if (r < 0) {
1239                         log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.nr_open, ignoring: %m");
1240                         break;
1241                 }
1242
1243                 log_debug("Successfully bumped fs.nr_open to %i", v);
1244                 break;
1245         }
1246 #endif
1247 }
1248
1249 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1250         int r, nr;
1251
1252         assert(saved_rlimit);
1253
1254         /* Save the original RLIMIT_NOFILE so that we can reset it later when transitioning from the initrd to the main
1255          * systemd or suchlike. */
1256         if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1257                 return log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1258
1259         /* Get the underlying absolute limit the kernel enforces */
1260         nr = read_nr_open();
1261
1262         /* Make sure forked processes get limits based on the original kernel setting */
1263         if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1264                 struct rlimit *rl;
1265
1266                 rl = newdup(struct rlimit, saved_rlimit, 1);
1267                 if (!rl)
1268                         return log_oom();
1269
1270                 /* Bump the hard limit for system services to a substantially higher value. The default hard limit
1271                  * current kernels set is pretty low (4K), mostly for historical reasons. According to kernel
1272                  * developers, the fd handling in recent kernels has been optimized substantially enough, so that we
1273                  * can bump the limit now, without paying too high a price in memory or performance. Note however that
1274                  * we only bump the hard limit, not the soft limit. That's because select() works the way it works, and
1275                  * chokes on fds >= 1024. If we'd bump the soft limit globally, it might accidentally happen to
1276                  * unexpecting programs that they get fds higher than what they can process using select(). By only
1277                  * bumping the hard limit but leaving the low limit as it is we avoid this pitfall: programs that are
1278                  * written by folks aware of the select() problem in mind (and thus use poll()/epoll instead of
1279                  * select(), the way everybody should) can explicitly opt into high fds by bumping their soft limit
1280                  * beyond 1024, to the hard limit we pass. */
1281                 if (arg_system)
1282                         rl->rlim_max = MIN((rlim_t) nr, MAX(rl->rlim_max, (rlim_t) HIGH_RLIMIT_NOFILE));
1283
1284                 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1285         }
1286
1287         /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for
1288          * both hard and soft. */
1289         r = setrlimit_closest(RLIMIT_NOFILE, &RLIMIT_MAKE_CONST(nr));
1290         if (r < 0)
1291                 return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1292
1293         return 0;
1294 }
1295
1296 static int bump_rlimit_memlock(struct rlimit *saved_rlimit) {
1297         int r;
1298
1299         assert(saved_rlimit);
1300
1301         /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK which should
1302          * normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's bump
1303          * the value high enough for our user. */
1304
1305         if (getrlimit(RLIMIT_MEMLOCK, saved_rlimit) < 0)
1306                 return log_warning_errno(errno, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1307
1308         r = setrlimit_closest(RLIMIT_MEMLOCK, &RLIMIT_MAKE_CONST(HIGH_RLIMIT_MEMLOCK));
1309         if (r < 0)
1310                 return log_warning_errno(r, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1311
1312         return 0;
1313 }
1314
1315 static void test_usr(void) {
1316
1317         /* Check that /usr is not a separate fs */
1318
1319         if (dir_is_empty("/usr") <= 0)
1320                 return;
1321
1322         log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1323                     "Some things will probably break (sometimes even silently) in mysterious ways. "
1324                     "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1325 }
1326
1327 static int enforce_syscall_archs(Set *archs) {
1328 #if HAVE_SECCOMP
1329         int r;
1330
1331         if (!is_seccomp_available())
1332                 return 0;
1333
1334         r = seccomp_restrict_archs(arg_syscall_archs);
1335         if (r < 0)
1336                 return log_error_errno(r, "Failed to enforce system call architecture restrication: %m");
1337 #endif
1338         return 0;
1339 }
1340
1341 static int status_welcome(void) {
1342         _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1343         int r;
1344
1345         if (IN_SET(arg_show_status, SHOW_STATUS_NO, SHOW_STATUS_AUTO))
1346                 return 0;
1347
1348         r = parse_os_release(NULL,
1349                              "PRETTY_NAME", &pretty_name,
1350                              "ANSI_COLOR", &ansi_color,
1351                              NULL);
1352         if (r < 0)
1353                 log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r,
1354                                "Failed to read os-release file, ignoring: %m");
1355
1356         if (log_get_show_color())
1357                 return status_printf(NULL, false, false,
1358                                      "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1359                                      isempty(ansi_color) ? "1" : ansi_color,
1360                                      isempty(pretty_name) ? "Linux" : pretty_name);
1361         else
1362                 return status_printf(NULL, false, false,
1363                                      "\nWelcome to %s!\n",
1364                                      isempty(pretty_name) ? "Linux" : pretty_name);
1365 }
1366
1367 static int write_container_id(void) {
1368         const char *c;
1369         int r;
1370
1371         c = getenv("container");
1372         if (isempty(c))
1373                 return 0;
1374
1375         RUN_WITH_UMASK(0022)
1376                 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1377         if (r < 0)
1378                 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
1379
1380         return 1;
1381 }
1382
1383 static int bump_unix_max_dgram_qlen(void) {
1384         _cleanup_free_ char *qlen = NULL;
1385         unsigned long v;
1386         int r;
1387
1388         /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1389          * really really early during boot, so that it is actually applied to all our sockets, including the
1390          * $NOTIFY_SOCKET one. */
1391
1392         r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1393         if (r < 0)
1394                 return log_warning_errno(r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1395
1396         r = safe_atolu(qlen, &v);
1397         if (r < 0)
1398                 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen);
1399
1400         if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1401                 return 0;
1402
1403         r = write_string_filef("/proc/sys/net/unix/max_dgram_qlen", 0, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN);
1404         if (r < 0)
1405                 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1406                                       "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1407
1408         return 1;
1409 }
1410
1411 static int fixup_environment(void) {
1412         _cleanup_free_ char *term = NULL;
1413         const char *t;
1414         int r;
1415
1416         /* Only fix up the environment when we are started as PID 1 */
1417         if (getpid_cached() != 1)
1418                 return 0;
1419
1420         /* We expect the environment to be set correctly if run inside a container. */
1421         if (detect_container() > 0)
1422                 return 0;
1423
1424         /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1425          * device used by the console. We try to make a better guess here since some consoles might not have support
1426          * for color mode for example.
1427          *
1428          * However if TERM was configured through the kernel command line then leave it alone. */
1429         r = proc_cmdline_get_key("TERM", 0, &term);
1430         if (r < 0)
1431                 return r;
1432
1433         t = term ?: default_term_for_tty("/dev/console");
1434
1435         if (setenv("TERM", t, 1) < 0)
1436                 return -errno;
1437
1438         return 0;
1439 }
1440
1441 static void redirect_telinit(int argc, char *argv[]) {
1442
1443         /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1444
1445 #if HAVE_SYSV_COMPAT
1446         if (getpid_cached() == 1)
1447                 return;
1448
1449         if (!strstr(program_invocation_short_name, "init"))
1450                 return;
1451
1452         execv(SYSTEMCTL_BINARY_PATH, argv);
1453         log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1454         exit(EXIT_FAILURE);
1455 #endif
1456 }
1457
1458 static int become_shutdown(
1459                 const char *shutdown_verb,
1460                 int retval) {
1461
1462         char log_level[DECIMAL_STR_MAX(int) + 1],
1463                 exit_code[DECIMAL_STR_MAX(uint8_t) + 1],
1464                 timeout[DECIMAL_STR_MAX(usec_t) + 1];
1465
1466         const char* command_line[13] = {
1467                 SYSTEMD_SHUTDOWN_BINARY_PATH,
1468                 shutdown_verb,
1469                 "--timeout", timeout,
1470                 "--log-level", log_level,
1471                 "--log-target",
1472         };
1473
1474         _cleanup_strv_free_ char **env_block = NULL;
1475         size_t pos = 7;
1476         int r;
1477
1478         assert(shutdown_verb);
1479         assert(!command_line[pos]);
1480         env_block = strv_copy(environ);
1481
1482         xsprintf(log_level, "%d", log_get_max_level());
1483         xsprintf(timeout, "%" PRI_USEC "us", arg_default_timeout_stop_usec);
1484
1485         switch (log_get_target()) {
1486
1487         case LOG_TARGET_KMSG:
1488         case LOG_TARGET_JOURNAL_OR_KMSG:
1489         case LOG_TARGET_SYSLOG_OR_KMSG:
1490                 command_line[pos++] = "kmsg";
1491                 break;
1492
1493         case LOG_TARGET_NULL:
1494                 command_line[pos++] = "null";
1495                 break;
1496
1497         case LOG_TARGET_CONSOLE:
1498         default:
1499                 command_line[pos++] = "console";
1500                 break;
1501         };
1502
1503         if (log_get_show_color())
1504                 command_line[pos++] = "--log-color";
1505
1506         if (log_get_show_location())
1507                 command_line[pos++] = "--log-location";
1508
1509         if (streq(shutdown_verb, "exit")) {
1510                 command_line[pos++] = "--exit-code";
1511                 command_line[pos++] = exit_code;
1512                 xsprintf(exit_code, "%d", retval);
1513         }
1514
1515         assert(pos < ELEMENTSOF(command_line));
1516
1517         if (streq(shutdown_verb, "reboot") &&
1518             arg_shutdown_watchdog > 0 &&
1519             arg_shutdown_watchdog != USEC_INFINITY) {
1520
1521                 char *e;
1522
1523                 /* If we reboot let's set the shutdown
1524                  * watchdog and tell the shutdown binary to
1525                  * repeatedly ping it */
1526                 r = watchdog_set_timeout(&arg_shutdown_watchdog);
1527                 watchdog_close(r < 0);
1528
1529                 /* Tell the binary how often to ping, ignore failure */
1530                 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
1531                         (void) strv_consume(&env_block, e);
1532
1533                 if (arg_watchdog_device &&
1534                     asprintf(&e, "WATCHDOG_DEVICE=%s", arg_watchdog_device) > 0)
1535                         (void) strv_consume(&env_block, e);
1536         } else
1537                 watchdog_close(true);
1538
1539         /* Avoid the creation of new processes forked by the
1540          * kernel; at this point, we will not listen to the
1541          * signals anyway */
1542         if (detect_container() <= 0)
1543                 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
1544
1545         execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
1546         return -errno;
1547 }
1548
1549 static void initialize_clock(void) {
1550         int r;
1551
1552         if (clock_is_localtime(NULL) > 0) {
1553                 int min;
1554
1555                 /*
1556                  * The very first call of settimeofday() also does a time warp in the kernel.
1557                  *
1558                  * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1559                  * of maintaining the RTC and do all adjustments.  This matches the behavior of Windows, which leaves
1560                  * the RTC alone if the registry tells that the RTC runs in UTC.
1561                  */
1562                 r = clock_set_timezone(&min);
1563                 if (r < 0)
1564                         log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1565                 else
1566                         log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1567
1568         } else if (!in_initrd()) {
1569                 /*
1570                  * Do a dummy very first call to seal the kernel's time warp magic.
1571                  *
1572                  * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1573                  * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1574                  * until we reach the real system.
1575                  *
1576                  * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1577                  * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1578                  * be treated as UTC that way.
1579                  */
1580                 (void) clock_reset_timewarp();
1581         }
1582
1583         r = clock_apply_epoch();
1584         if (r < 0)
1585                 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1586         else if (r > 0)
1587                 log_info("System time before build time, advancing clock.");
1588 }
1589
1590 static void initialize_coredump(bool skip_setup) {
1591 #if ENABLE_COREDUMP
1592         if (getpid_cached() != 1)
1593                 return;
1594
1595         /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1596          * will process core dumps for system services by default. */
1597         if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0)
1598                 log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m");
1599
1600         /* But at the same time, turn off the core_pattern logic by default, so that no
1601          * coredumps are stored until the systemd-coredump tool is enabled via
1602          * sysctl. However it can be changed via the kernel command line later so core
1603          * dumps can still be generated during early startup and in initramfs. */
1604         if (!skip_setup)
1605                 disable_coredumps();
1606 #endif
1607 }
1608
1609 static void initialize_core_pattern(bool skip_setup) {
1610         int r;
1611
1612         if (skip_setup || !arg_early_core_pattern)
1613                 return;
1614
1615         if (getpid_cached() != 1)
1616                 return;
1617
1618         r = write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern, 0);
1619         if (r < 0)
1620                 log_warning_errno(r, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", arg_early_core_pattern);
1621 }
1622
1623 static void do_reexecute(
1624                 int argc,
1625                 char *argv[],
1626                 const struct rlimit *saved_rlimit_nofile,
1627                 const struct rlimit *saved_rlimit_memlock,
1628                 FDSet *fds,
1629                 const char *switch_root_dir,
1630                 const char *switch_root_init,
1631                 const char **ret_error_message) {
1632
1633         unsigned i, j, args_size;
1634         const char **args;
1635         int r;
1636
1637         assert(saved_rlimit_nofile);
1638         assert(saved_rlimit_memlock);
1639         assert(ret_error_message);
1640
1641         /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1642          * we do that */
1643         watchdog_close(true);
1644
1645         /* Reset the RLIMIT_NOFILE to the kernel default, so that the new systemd can pass the kernel default to its
1646          * child processes */
1647
1648         if (saved_rlimit_nofile->rlim_cur > 0)
1649                 (void) setrlimit(RLIMIT_NOFILE, saved_rlimit_nofile);
1650         if (saved_rlimit_memlock->rlim_cur != (rlim_t) -1)
1651                 (void) setrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock);
1652
1653         if (switch_root_dir) {
1654                 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1655                  * SIGCHLD for them after deserializing. */
1656                 broadcast_signal(SIGTERM, false, true, arg_default_timeout_stop_usec);
1657
1658                 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1659                 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1660                 if (r < 0)
1661                         log_error_errno(r, "Failed to switch root, trying to continue: %m");
1662         }
1663
1664         args_size = MAX(6, argc+1);
1665         args = newa(const char*, args_size);
1666
1667         if (!switch_root_init) {
1668                 char sfd[DECIMAL_STR_MAX(int) + 1];
1669
1670                 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1671                  * the user didn't specify an explicit init to spawn. */
1672
1673                 assert(arg_serialization);
1674                 assert(fds);
1675
1676                 xsprintf(sfd, "%i", fileno(arg_serialization));
1677
1678                 i = 0;
1679                 args[i++] = SYSTEMD_BINARY_PATH;
1680                 if (switch_root_dir)
1681                         args[i++] = "--switched-root";
1682                 args[i++] = arg_system ? "--system" : "--user";
1683                 args[i++] = "--deserialize";
1684                 args[i++] = sfd;
1685                 args[i++] = NULL;
1686
1687                 assert(i <= args_size);
1688
1689                 /*
1690                  * We want valgrind to print its memory usage summary before reexecution.  Valgrind won't do this is on
1691                  * its own on exec(), but it will do it on exit().  Hence, to ensure we get a summary here, fork() off
1692                  * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1693                  * before proceeding into the exec().
1694                  */
1695                 valgrind_summary_hack();
1696
1697                 (void) execv(args[0], (char* const*) args);
1698                 log_debug_errno(errno, "Failed to execute our own binary, trying fallback: %m");
1699         }
1700
1701         /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1702          * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1703          * doesn't matter.) */
1704
1705         arg_serialization = safe_fclose(arg_serialization);
1706         fds = fdset_free(fds);
1707
1708         /* Reopen the console */
1709         (void) make_console_stdio();
1710
1711         for (j = 1, i = 1; j < (unsigned) argc; j++)
1712                 args[i++] = argv[j];
1713         args[i++] = NULL;
1714         assert(i <= args_size);
1715
1716         /* Reenable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1717         (void) reset_all_signal_handlers();
1718         (void) reset_signal_mask();
1719
1720         if (switch_root_init) {
1721                 args[0] = switch_root_init;
1722                 (void) execv(args[0], (char* const*) args);
1723                 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
1724         }
1725
1726         args[0] = "/sbin/init";
1727         (void) execv(args[0], (char* const*) args);
1728         r = -errno;
1729
1730         manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
1731                               ANSI_HIGHLIGHT_RED "  !!  " ANSI_NORMAL,
1732                               "Failed to execute /sbin/init");
1733
1734         if (r == -ENOENT) {
1735                 log_warning("No /sbin/init, trying fallback");
1736
1737                 args[0] = "/bin/sh";
1738                 args[1] = NULL;
1739                 (void) execv(args[0], (char* const*) args);
1740                 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
1741         } else
1742                 log_warning_errno(r, "Failed to execute /sbin/init, giving up: %m");
1743
1744         *ret_error_message = "Failed to execute fallback shell";
1745 }
1746
1747 static int invoke_main_loop(
1748                 Manager *m,
1749                 bool *ret_reexecute,
1750                 int *ret_retval,                   /* Return parameters relevant for shutting down */
1751                 const char **ret_shutdown_verb,    /* … */
1752                 FDSet **ret_fds,                   /* Return parameters for reexecuting */
1753                 char **ret_switch_root_dir,        /* … */
1754                 char **ret_switch_root_init,       /* … */
1755                 const char **ret_error_message) {
1756
1757         int r;
1758
1759         assert(m);
1760         assert(ret_reexecute);
1761         assert(ret_retval);
1762         assert(ret_shutdown_verb);
1763         assert(ret_fds);
1764         assert(ret_switch_root_dir);
1765         assert(ret_switch_root_init);
1766         assert(ret_error_message);
1767
1768         for (;;) {
1769                 r = manager_loop(m);
1770                 if (r < 0) {
1771                         *ret_error_message = "Failed to run main loop";
1772                         return log_emergency_errno(r, "Failed to run main loop: %m");
1773                 }
1774
1775                 switch ((ManagerObjective) r) {
1776
1777                 case MANAGER_RELOAD: {
1778                         LogTarget saved_log_target;
1779                         int saved_log_level;
1780
1781                         log_info("Reloading.");
1782
1783                         /* First, save any overridden log level/target, then parse the configuration file, which might
1784                          * change the log level to new settings. */
1785
1786                         saved_log_level = m->log_level_overridden ? log_get_max_level() : -1;
1787                         saved_log_target = m->log_target_overridden ? log_get_target() : _LOG_TARGET_INVALID;
1788
1789                         r = parse_config_file();
1790                         if (r < 0)
1791                                 log_warning_errno(r, "Failed to parse config file, ignoring: %m");
1792
1793                         set_manager_defaults(m);
1794
1795                         if (saved_log_level >= 0)
1796                                 manager_override_log_level(m, saved_log_level);
1797                         if (saved_log_target >= 0)
1798                                 manager_override_log_target(m, saved_log_target);
1799
1800                         r = manager_reload(m);
1801                         if (r < 0)
1802                                 /* Reloading failed before the point of no return. Let's continue running as if nothing happened. */
1803                                 m->objective = MANAGER_OK;
1804
1805                         break;
1806                 }
1807
1808                 case MANAGER_REEXECUTE:
1809
1810                         r = prepare_reexecute(m, &arg_serialization, ret_fds, false);
1811                         if (r < 0) {
1812                                 *ret_error_message = "Failed to prepare for reexecution";
1813                                 return r;
1814                         }
1815
1816                         log_notice("Reexecuting.");
1817
1818                         *ret_reexecute = true;
1819                         *ret_retval = EXIT_SUCCESS;
1820                         *ret_shutdown_verb = NULL;
1821                         *ret_switch_root_dir = *ret_switch_root_init = NULL;
1822
1823                         return 0;
1824
1825                 case MANAGER_SWITCH_ROOT:
1826                         if (!m->switch_root_init) {
1827                                 r = prepare_reexecute(m, &arg_serialization, ret_fds, true);
1828                                 if (r < 0) {
1829                                         *ret_error_message = "Failed to prepare for reexecution";
1830                                         return r;
1831                                 }
1832                         } else
1833                                 *ret_fds = NULL;
1834
1835                         log_notice("Switching root.");
1836
1837                         *ret_reexecute = true;
1838                         *ret_retval = EXIT_SUCCESS;
1839                         *ret_shutdown_verb = NULL;
1840
1841                         /* Steal the switch root parameters */
1842                         *ret_switch_root_dir = m->switch_root;
1843                         *ret_switch_root_init = m->switch_root_init;
1844                         m->switch_root = m->switch_root_init = NULL;
1845
1846                         return 0;
1847
1848                 case MANAGER_EXIT:
1849
1850                         if (MANAGER_IS_USER(m)) {
1851                                 log_debug("Exit.");
1852
1853                                 *ret_reexecute = false;
1854                                 *ret_retval = m->return_value;
1855                                 *ret_shutdown_verb = NULL;
1856                                 *ret_fds = NULL;
1857                                 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1858
1859                                 return 0;
1860                         }
1861
1862                         _fallthrough_;
1863                 case MANAGER_REBOOT:
1864                 case MANAGER_POWEROFF:
1865                 case MANAGER_HALT:
1866                 case MANAGER_KEXEC: {
1867                         static const char * const table[_MANAGER_OBJECTIVE_MAX] = {
1868                                 [MANAGER_EXIT]     = "exit",
1869                                 [MANAGER_REBOOT]   = "reboot",
1870                                 [MANAGER_POWEROFF] = "poweroff",
1871                                 [MANAGER_HALT]     = "halt",
1872                                 [MANAGER_KEXEC]    = "kexec",
1873                         };
1874
1875                         log_notice("Shutting down.");
1876
1877                         *ret_reexecute = false;
1878                         *ret_retval = m->return_value;
1879                         assert_se(*ret_shutdown_verb = table[m->objective]);
1880                         *ret_fds = NULL;
1881                         *ret_switch_root_dir = *ret_switch_root_init = NULL;
1882
1883                         return 0;
1884                 }
1885
1886                 default:
1887                         assert_not_reached("Unknown or unexpected manager objective.");
1888                 }
1889         }
1890 }
1891
1892 static void log_execution_mode(bool *ret_first_boot) {
1893         assert(ret_first_boot);
1894
1895         if (arg_system) {
1896                 int v;
1897
1898                 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1899                          arg_action == ACTION_TEST ? "test " : "" );
1900
1901                 v = detect_virtualization();
1902                 if (v > 0)
1903                         log_info("Detected virtualization %s.", virtualization_to_string(v));
1904
1905                 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1906
1907                 if (in_initrd()) {
1908                         *ret_first_boot = false;
1909                         log_info("Running in initial RAM disk.");
1910                 } else {
1911                         /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
1912                          * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
1913                          * doesn't it's unpopulated. This allows container managers and installers to provision a
1914                          * couple of files already. If the container manager wants to provision the machine ID itself
1915                          * it should pass $container_uuid to PID 1. */
1916
1917                         *ret_first_boot = access("/etc/machine-id", F_OK) < 0;
1918                         if (*ret_first_boot)
1919                                 log_info("Running with unpopulated /etc.");
1920                 }
1921         } else {
1922                 if (DEBUG_LOGGING) {
1923                         _cleanup_free_ char *t;
1924
1925                         t = uid_to_name(getuid());
1926                         log_debug(PACKAGE_STRING " running in %suser mode for user " UID_FMT "/%s. (" SYSTEMD_FEATURES ")",
1927                                   arg_action == ACTION_TEST ? " test" : "", getuid(), strna(t));
1928                 }
1929
1930                 *ret_first_boot = false;
1931         }
1932 }
1933
1934 static int initialize_runtime(
1935                 bool skip_setup,
1936                 struct rlimit *saved_rlimit_nofile,
1937                 struct rlimit *saved_rlimit_memlock,
1938                 const char **ret_error_message) {
1939
1940         int r;
1941
1942         assert(ret_error_message);
1943
1944         /* Sets up various runtime parameters. Many of these initializations are conditionalized:
1945          *
1946          * - Some only apply to --system instances
1947          * - Some only apply to --user instances
1948          * - Some only apply when we first start up, but not when we reexecute
1949          */
1950
1951         if (arg_action != ACTION_RUN)
1952                 return 0;
1953
1954         if (arg_system) {
1955                 /* Make sure we leave a core dump without panicing the kernel. */
1956                 install_crash_handler();
1957
1958                 if (!skip_setup) {
1959                         r = mount_cgroup_controllers(arg_join_controllers);
1960                         if (r < 0) {
1961                                 *ret_error_message = "Failed to mount cgroup hierarchies";
1962                                 return r;
1963                         }
1964
1965                         status_welcome();
1966                         hostname_setup();
1967                         machine_id_setup(NULL, arg_machine_id, NULL);
1968                         loopback_setup();
1969                         bump_unix_max_dgram_qlen();
1970                         bump_file_max_and_nr_open();
1971                         test_usr();
1972                         write_container_id();
1973                 }
1974
1975                 if (arg_watchdog_device) {
1976                         r = watchdog_set_device(arg_watchdog_device);
1977                         if (r < 0)
1978                                 log_warning_errno(r, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device);
1979                 }
1980
1981                 if (arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY)
1982                         watchdog_set_timeout(&arg_runtime_watchdog);
1983         }
1984
1985         if (arg_timer_slack_nsec != NSEC_INFINITY)
1986                 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
1987                         log_warning_errno(errno, "Failed to adjust timer slack, ignoring: %m");
1988
1989         if (arg_system && !cap_test_all(arg_capability_bounding_set)) {
1990                 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
1991                 if (r < 0) {
1992                         *ret_error_message = "Failed to drop capability bounding set of usermode helpers";
1993                         return log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
1994                 }
1995
1996                 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
1997                 if (r < 0) {
1998                         *ret_error_message = "Failed to drop capability bounding set";
1999                         return log_emergency_errno(r, "Failed to drop capability bounding set: %m");
2000                 }
2001         }
2002
2003         if (arg_system && arg_no_new_privs) {
2004                 if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) {
2005                         *ret_error_message = "Failed to disable new privileges";
2006                         return log_emergency_errno(errno, "Failed to disable new privileges: %m");
2007                 }
2008         }
2009
2010         if (arg_syscall_archs) {
2011                 r = enforce_syscall_archs(arg_syscall_archs);
2012                 if (r < 0) {
2013                         *ret_error_message = "Failed to set syscall architectures";
2014                         return r;
2015                 }
2016         }
2017
2018         if (!arg_system)
2019                 /* Become reaper of our children */
2020                 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
2021                         log_warning_errno(errno, "Failed to make us a subreaper: %m");
2022
2023         /* Bump up RLIMIT_NOFILE for systemd itself */
2024         (void) bump_rlimit_nofile(saved_rlimit_nofile);
2025         (void) bump_rlimit_memlock(saved_rlimit_memlock);
2026
2027         return 0;
2028 }
2029
2030 static int do_queue_default_job(
2031                 Manager *m,
2032                 const char **ret_error_message) {
2033
2034         _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
2035         Job *default_unit_job;
2036         Unit *target = NULL;
2037         int r;
2038
2039         log_debug("Activating default unit: %s", arg_default_unit);
2040
2041         r = manager_load_startable_unit_or_warn(m, arg_default_unit, NULL, &target);
2042         if (r < 0) {
2043                 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET);
2044
2045                 r = manager_load_startable_unit_or_warn(m, SPECIAL_RESCUE_TARGET, NULL, &target);
2046                 if (r < 0) {
2047                         *ret_error_message = r == -ERFKILL ? "Rescue target masked"
2048                                                            : "Failed to load rescue target";
2049                         return r;
2050                 }
2051         }
2052
2053         assert(target->load_state == UNIT_LOADED);
2054
2055         r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, &error, &default_unit_job);
2056         if (r == -EPERM) {
2057                 log_debug_errno(r, "Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
2058
2059                 sd_bus_error_free(&error);
2060
2061                 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, &error, &default_unit_job);
2062                 if (r < 0) {
2063                         *ret_error_message = "Failed to start default target";
2064                         return log_emergency_errno(r, "Failed to start default target: %s", bus_error_message(&error, r));
2065                 }
2066
2067         } else if (r < 0) {
2068                 *ret_error_message = "Failed to isolate default target";
2069                 return log_emergency_errno(r, "Failed to isolate default target: %s", bus_error_message(&error, r));
2070         }
2071
2072         m->default_unit_job_id = default_unit_job->id;
2073
2074         return 0;
2075 }
2076
2077 static void free_arguments(void) {
2078
2079         /* Frees all arg_* variables, with the exception of arg_serialization */
2080         rlimit_free_all(arg_default_rlimit);
2081
2082         arg_default_unit = mfree(arg_default_unit);
2083         arg_confirm_spawn = mfree(arg_confirm_spawn);
2084         arg_join_controllers = strv_free_free(arg_join_controllers);
2085         arg_default_environment = strv_free(arg_default_environment);
2086         arg_syscall_archs = set_free(arg_syscall_archs);
2087 }
2088
2089 static int load_configuration(int argc, char **argv, const char **ret_error_message) {
2090         int r;
2091
2092         assert(ret_error_message);
2093
2094         arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U);
2095
2096         r = parse_config_file();
2097         if (r < 0) {
2098                 *ret_error_message = "Failed to parse config file";
2099                 return r;
2100         }
2101
2102         if (arg_system) {
2103                 r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
2104                 if (r < 0)
2105                         log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
2106         }
2107
2108         /* Note that this also parses bits from the kernel command line, including "debug". */
2109         log_parse_environment();
2110
2111         r = parse_argv(argc, argv);
2112         if (r < 0) {
2113                 *ret_error_message = "Failed to parse commandline arguments";
2114                 return r;
2115         }
2116
2117         /* Initialize default unit */
2118         if (!arg_default_unit) {
2119                 arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET);
2120                 if (!arg_default_unit) {
2121                         *ret_error_message = "Failed to set default unit";
2122                         return log_oom();
2123                 }
2124         }
2125
2126         /* Initialize the show status setting if it hasn't been set explicitly yet */
2127         if (arg_show_status == _SHOW_STATUS_INVALID)
2128                 arg_show_status = SHOW_STATUS_YES;
2129
2130         return 0;
2131 }
2132
2133 static int safety_checks(void) {
2134
2135         if (getpid_cached() == 1 &&
2136             arg_action != ACTION_RUN) {
2137                 log_error("Unsupported execution mode while PID 1.");
2138                 return -EPERM;
2139         }
2140
2141         if (getpid_cached() == 1 &&
2142             !arg_system) {
2143                 log_error("Can't run --user mode as PID 1.");
2144                 return -EPERM;
2145         }
2146
2147         if (arg_action == ACTION_RUN &&
2148             arg_system &&
2149             getpid_cached() != 1) {
2150                 log_error("Can't run system mode unless PID 1.");
2151                 return -EPERM;
2152         }
2153
2154         if (arg_action == ACTION_TEST &&
2155             geteuid() == 0) {
2156                 log_error("Don't run test mode as root.");
2157                 return -EPERM;
2158         }
2159
2160         if (!arg_system &&
2161             arg_action == ACTION_RUN &&
2162             sd_booted() <= 0) {
2163                 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
2164                 return -EOPNOTSUPP;
2165         }
2166
2167         if (!arg_system &&
2168             arg_action == ACTION_RUN &&
2169             !getenv("XDG_RUNTIME_DIR")) {
2170                 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
2171                 return -EUNATCH;
2172         }
2173
2174         if (arg_system &&
2175             arg_action == ACTION_RUN &&
2176             running_in_chroot() > 0) {
2177                 log_error("Cannot be run in a chroot() environment.");
2178                 return -EOPNOTSUPP;
2179         }
2180
2181         return 0;
2182 }
2183
2184 static int initialize_security(
2185                 bool *loaded_policy,
2186                 dual_timestamp *security_start_timestamp,
2187                 dual_timestamp *security_finish_timestamp,
2188                 const char **ret_error_message) {
2189
2190         int r;
2191
2192         assert(loaded_policy);
2193         assert(security_start_timestamp);
2194         assert(security_finish_timestamp);
2195         assert(ret_error_message);
2196
2197         dual_timestamp_get(security_start_timestamp);
2198
2199         r = mac_selinux_setup(loaded_policy);
2200         if (r < 0) {
2201                 *ret_error_message = "Failed to load SELinux policy";
2202                 return r;
2203         }
2204
2205         r = mac_smack_setup(loaded_policy);
2206         if (r < 0) {
2207                 *ret_error_message = "Failed to load SMACK policy";
2208                 return r;
2209         }
2210
2211         r = ima_setup();
2212         if (r < 0) {
2213                 *ret_error_message = "Failed to load IMA policy";
2214                 return r;
2215         }
2216
2217         dual_timestamp_get(security_finish_timestamp);
2218         return 0;
2219 }
2220
2221 static void test_summary(Manager *m) {
2222         assert(m);
2223
2224         printf("-> By units:\n");
2225         manager_dump_units(m, stdout, "\t");
2226
2227         printf("-> By jobs:\n");
2228         manager_dump_jobs(m, stdout, "\t");
2229 }
2230
2231 static int collect_fds(FDSet **ret_fds, const char **ret_error_message) {
2232         int r;
2233
2234         assert(ret_fds);
2235         assert(ret_error_message);
2236
2237         r = fdset_new_fill(ret_fds);
2238         if (r < 0) {
2239                 *ret_error_message = "Failed to allocate fd set";
2240                 return log_emergency_errno(r, "Failed to allocate fd set: %m");
2241         }
2242
2243         fdset_cloexec(*ret_fds, true);
2244
2245         if (arg_serialization)
2246                 assert_se(fdset_remove(*ret_fds, fileno(arg_serialization)) >= 0);
2247
2248         return 0;
2249 }
2250
2251 static void setup_console_terminal(bool skip_setup) {
2252
2253         if (!arg_system)
2254                 return;
2255
2256         /* Become a session leader if we aren't one yet. */
2257         (void) setsid();
2258
2259         /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2260          * tty. */
2261         (void) release_terminal();
2262
2263         /* Reset the console, but only if this is really init and we are freshly booted */
2264         if (getpid_cached() == 1 && !skip_setup)
2265                 (void) console_setup();
2266 }
2267
2268 static bool early_skip_setup_check(int argc, char *argv[]) {
2269         bool found_deserialize = false;
2270         int i;
2271
2272         /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2273          * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2274          * anyway, even if in that case we also do deserialization. */
2275
2276         for (i = 1; i < argc; i++) {
2277                 if (streq(argv[i], "--switched-root"))
2278                         return false; /* If we switched root, don't skip the setup. */
2279                 else if (streq(argv[i], "--deserialize"))
2280                         found_deserialize = true;
2281         }
2282
2283         return found_deserialize; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2284 }
2285
2286 int main(int argc, char *argv[]) {
2287
2288         dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL, userspace_timestamp = DUAL_TIMESTAMP_NULL, kernel_timestamp = DUAL_TIMESTAMP_NULL,
2289                 security_start_timestamp = DUAL_TIMESTAMP_NULL, security_finish_timestamp = DUAL_TIMESTAMP_NULL;
2290         struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0), saved_rlimit_memlock = RLIMIT_MAKE_CONST((rlim_t) -1);
2291         bool skip_setup, loaded_policy = false, queue_default_job = false, first_boot = false, reexecute = false;
2292         char *switch_root_dir = NULL, *switch_root_init = NULL;
2293         usec_t before_startup, after_startup;
2294         static char systemd[] = "systemd";
2295         char timespan[FORMAT_TIMESPAN_MAX];
2296         const char *shutdown_verb = NULL, *error_message = NULL;
2297         int r, retval = EXIT_FAILURE;
2298         Manager *m = NULL;
2299         FDSet *fds = NULL;
2300
2301         /* SysV compatibility: redirect init → telinit */
2302         redirect_telinit(argc, argv);
2303
2304         /* Take timestamps early on */
2305         dual_timestamp_from_monotonic(&kernel_timestamp, 0);
2306         dual_timestamp_get(&userspace_timestamp);
2307
2308         /* Figure out whether we need to do initialize the system, or if we already did that because we are
2309          * reexecuting */
2310         skip_setup = early_skip_setup_check(argc, argv);
2311
2312         /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2313          * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
2314         program_invocation_short_name = systemd;
2315         (void) prctl(PR_SET_NAME, systemd);
2316
2317         /* Save the original command line */
2318         saved_argv = argv;
2319         saved_argc = argc;
2320
2321         /* Make sure that if the user says "syslog" we actually log to the journal. */
2322         log_set_upgrade_syslog_to_journal(true);
2323
2324         if (getpid_cached() == 1) {
2325                 /* When we run as PID 1 force system mode */
2326                 arg_system = true;
2327
2328                 /* Disable the umask logic */
2329                 umask(0);
2330
2331                 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2332                  * activated yet (even though the log socket for it exists). */
2333                 log_set_prohibit_ipc(true);
2334
2335                 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2336                  * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2337                  * child process right before execve()'ing the actual binary, at a point in time where socket
2338                  * activation stderr/stdout area already set up. */
2339                 log_set_always_reopen_console(true);
2340
2341                 if (detect_container() <= 0) {
2342
2343                         /* Running outside of a container as PID 1 */
2344                         log_set_target(LOG_TARGET_KMSG);
2345                         log_open();
2346
2347                         if (in_initrd())
2348                                 initrd_timestamp = userspace_timestamp;
2349
2350                         if (!skip_setup) {
2351                                 r = mount_setup_early();
2352                                 if (r < 0) {
2353                                         error_message = "Failed to mount early API filesystems";
2354                                         goto finish;
2355                                 }
2356
2357                                 r = initialize_security(
2358                                                 &loaded_policy,
2359                                                 &security_start_timestamp,
2360                                                 &security_finish_timestamp,
2361                                                 &error_message);
2362                                 if (r < 0)
2363                                         goto finish;
2364                         }
2365
2366                         if (mac_selinux_init() < 0) {
2367                                 error_message = "Failed to initialize SELinux policy";
2368                                 goto finish;
2369                         }
2370
2371                         if (!skip_setup)
2372                                 initialize_clock();
2373
2374                         /* Set the default for later on, but don't actually open the logs like this for now. Note that
2375                          * if we are transitioning from the initrd there might still be journal fd open, and we
2376                          * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2377                          * elsewhere. */
2378                         log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
2379
2380                 } else {
2381                         /* Running inside a container, as PID 1 */
2382                         log_set_target(LOG_TARGET_CONSOLE);
2383                         log_open();
2384
2385                         /* For later on, see above... */
2386                         log_set_target(LOG_TARGET_JOURNAL);
2387
2388                         /* clear the kernel timestamp,
2389                          * because we are in a container */
2390                         kernel_timestamp = DUAL_TIMESTAMP_NULL;
2391                 }
2392
2393                 initialize_coredump(skip_setup);
2394
2395                 r = fixup_environment();
2396                 if (r < 0) {
2397                         log_emergency_errno(r, "Failed to fix up PID 1 environment: %m");
2398                         error_message = "Failed to fix up PID1 environment";
2399                         goto finish;
2400                 }
2401
2402         } else {
2403                 /* Running as user instance */
2404                 arg_system = false;
2405                 log_set_target(LOG_TARGET_AUTO);
2406                 log_open();
2407
2408                 /* clear the kernel timestamp,
2409                  * because we are not PID 1 */
2410                 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2411         }
2412
2413         if (arg_system) {
2414                 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2415                  * they never log into the console. */
2416                 log_show_color(colors_enabled());
2417
2418                 r = make_null_stdio();
2419                 if (r < 0)
2420                         log_warning_errno(r, "Failed to redirect standard streams to /dev/null, ignoring: %m");
2421         }
2422
2423         /* Mount /proc, /sys and friends, so that /proc/cmdline and
2424          * /proc/$PID/fd is available. */
2425         if (getpid_cached() == 1) {
2426
2427                 /* Load the kernel modules early. */
2428                 if (!skip_setup)
2429                         kmod_setup();
2430
2431                 r = mount_setup(loaded_policy);
2432                 if (r < 0) {
2433                         error_message = "Failed to mount API filesystems";
2434                         goto finish;
2435                 }
2436         }
2437
2438         /* Reset all signal handlers. */
2439         (void) reset_all_signal_handlers();
2440         (void) ignore_signals(SIGNALS_IGNORE, -1);
2441
2442         r = load_configuration(argc, argv, &error_message);
2443         if (r < 0)
2444                 goto finish;
2445
2446         r = safety_checks();
2447         if (r < 0)
2448                 goto finish;
2449
2450         if (IN_SET(arg_action, ACTION_TEST, ACTION_HELP, ACTION_DUMP_CONFIGURATION_ITEMS, ACTION_DUMP_BUS_PROPERTIES))
2451                 (void) pager_open(arg_no_pager, false);
2452
2453         if (arg_action != ACTION_RUN)
2454                 skip_setup = true;
2455
2456         if (arg_action == ACTION_HELP) {
2457                 retval = help() < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
2458                 goto finish;
2459         } else if (arg_action == ACTION_VERSION) {
2460                 retval = version();
2461                 goto finish;
2462         } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
2463                 unit_dump_config_items(stdout);
2464                 retval = EXIT_SUCCESS;
2465                 goto finish;
2466         } else if (arg_action == ACTION_DUMP_BUS_PROPERTIES) {
2467                 dump_bus_properties(stdout);
2468                 retval = EXIT_SUCCESS;
2469                 goto finish;
2470         }
2471
2472         assert_se(IN_SET(arg_action, ACTION_RUN, ACTION_TEST));
2473
2474         /* Move out of the way, so that we won't block unmounts */
2475         assert_se(chdir("/") == 0);
2476
2477         if (arg_action == ACTION_RUN) {
2478
2479                 /* A core pattern might have been specified via the cmdline.  */
2480                 initialize_core_pattern(skip_setup);
2481
2482                 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2483                 log_close();
2484
2485                 /* Remember open file descriptors for later deserialization */
2486                 r = collect_fds(&fds, &error_message);
2487                 if (r < 0)
2488                         goto finish;
2489
2490                 /* Give up any control of the console, but make sure its initialized. */
2491                 setup_console_terminal(skip_setup);
2492
2493                 /* Open the logging devices, if possible and necessary */
2494                 log_open();
2495         }
2496
2497         log_execution_mode(&first_boot);
2498
2499         r = initialize_runtime(skip_setup,
2500                                &saved_rlimit_nofile,
2501                                &saved_rlimit_memlock,
2502                                &error_message);
2503         if (r < 0)
2504                 goto finish;
2505
2506         r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER,
2507                         arg_action == ACTION_TEST ? MANAGER_TEST_FULL : 0,
2508                         &m);
2509         if (r < 0) {
2510                 log_emergency_errno(r, "Failed to allocate manager object: %m");
2511                 error_message = "Failed to allocate manager object";
2512                 goto finish;
2513         }
2514
2515         m->timestamps[MANAGER_TIMESTAMP_KERNEL] = kernel_timestamp;
2516         m->timestamps[MANAGER_TIMESTAMP_INITRD] = initrd_timestamp;
2517         m->timestamps[MANAGER_TIMESTAMP_USERSPACE] = userspace_timestamp;
2518         m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START)] = security_start_timestamp;
2519         m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH)] = security_finish_timestamp;
2520
2521         set_manager_defaults(m);
2522         set_manager_settings(m);
2523         manager_set_first_boot(m, first_boot);
2524
2525         /* Remember whether we should queue the default job */
2526         queue_default_job = !arg_serialization || arg_switched_root;
2527
2528         before_startup = now(CLOCK_MONOTONIC);
2529
2530         r = manager_startup(m, arg_serialization, fds);
2531         if (r < 0) {
2532                 error_message = "Failed to start up manager";
2533                 goto finish;
2534         }
2535
2536         /* This will close all file descriptors that were opened, but not claimed by any unit. */
2537         fds = fdset_free(fds);
2538         arg_serialization = safe_fclose(arg_serialization);
2539
2540         if (queue_default_job) {
2541                 r = do_queue_default_job(m, &error_message);
2542                 if (r < 0)
2543                         goto finish;
2544         }
2545
2546         after_startup = now(CLOCK_MONOTONIC);
2547
2548         log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
2549                  "Loaded units and determined initial transaction in %s.",
2550                  format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
2551
2552         if (arg_action == ACTION_TEST) {
2553                 test_summary(m);
2554                 retval = EXIT_SUCCESS;
2555                 goto finish;
2556         }
2557
2558         (void) invoke_main_loop(m,
2559                                 &reexecute,
2560                                 &retval,
2561                                 &shutdown_verb,
2562                                 &fds,
2563                                 &switch_root_dir,
2564                                 &switch_root_init,
2565                                 &error_message);
2566
2567 finish:
2568         pager_close();
2569
2570         if (m) {
2571                 arg_shutdown_watchdog = m->shutdown_watchdog;
2572                 m = manager_free(m);
2573         }
2574
2575         free_arguments();
2576         mac_selinux_finish();
2577
2578         if (reexecute)
2579                 do_reexecute(argc, argv,
2580                              &saved_rlimit_nofile,
2581                              &saved_rlimit_memlock,
2582                              fds,
2583                              switch_root_dir,
2584                              switch_root_init,
2585                              &error_message); /* This only returns if reexecution failed */
2586
2587         arg_serialization = safe_fclose(arg_serialization);
2588         fds = fdset_free(fds);
2589
2590 #if HAVE_VALGRIND_VALGRIND_H
2591         /* If we are PID 1 and running under valgrind, then let's exit
2592          * here explicitly. valgrind will only generate nice output on
2593          * exit(), not on exec(), hence let's do the former not the
2594          * latter here. */
2595         if (getpid_cached() == 1 && RUNNING_ON_VALGRIND) {
2596                 /* Cleanup watchdog_device strings for valgrind. We need them
2597                  * in become_shutdown() so normally we cannot free them yet. */
2598                 watchdog_free_device();
2599                 arg_watchdog_device = mfree(arg_watchdog_device);
2600                 return retval;
2601         }
2602 #endif
2603
2604         if (shutdown_verb) {
2605                 r = become_shutdown(shutdown_verb, retval);
2606                 log_error_errno(r, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
2607                 error_message = "Failed to execute shutdown binary";
2608         }
2609
2610         watchdog_free_device();
2611         arg_watchdog_device = mfree(arg_watchdog_device);
2612
2613         if (getpid_cached() == 1) {
2614                 if (error_message)
2615                         manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2616                                               ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
2617                                               "%s, freezing.", error_message);
2618                 freeze_or_reboot();
2619         }
2620
2621         return retval;
2622 }