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 "cgroup-util.h"
  32 #include "clock-util.h"
  33 #include "conf-parser.h"
  34 #include "cpu-set-util.h"
  35 #include "dbus.h"
  36 #include "dbus-manager.h"
  37 #include "def.h"
  38 #include "emergency-action.h"
  39 #include "env-util.h"
  40 #include "exit-status.h"
  41 #include "fd-util.h"
  42 #include "fdset.h"
  43 #include "fileio.h"
  44 #include "format-util.h"
  45 #include "fs-util.h"
  46 #include "hostname-setup.h"
  47 #include "ima-setup.h"
  48 #include "killall.h"
  49 #include "kmod-setup.h"
  50 #include "load-fragment.h"
  51 #include "log.h"
  52 #include "loopback-setup.h"
  53 #include "machine-id-setup.h"
  54 #include "manager.h"
  55 #include "missing.h"
  56 #include "mount-setup.h"
  57 #include "os-util.h"
  58 #include "pager.h"
  59 #include "parse-util.h"
  60 #include "path-util.h"
  61 #include "proc-cmdline.h"
  62 #include "process-util.h"
  63 #include "raw-clone.h"
  64 #include "rlimit-util.h"
  65 #if HAVE_SECCOMP
  66 #include "seccomp-util.h"
  67 #endif
  68 #include "selinux-setup.h"
  69 #include "selinux-util.h"
  70 #include "signal-util.h"
  71 #include "smack-setup.h"
  72 #include "special.h"
  73 #include "stat-util.h"
  74 #include "stdio-util.h"
  75 #include "strv.h"
  76 #include "switch-root.h"
  77 #include "sysctl-util.h"
  78 #include "terminal-util.h"
  79 #include "umask-util.h"
  80 #include "user-util.h"
  81 #include "util.h"
  82 #include "virt.h"
  83 #include "watchdog.h"
  84
  85 static enum {
  86         ACTION_RUN,
  87         ACTION_HELP,
  88         ACTION_VERSION,
  89         ACTION_TEST,
  90         ACTION_DUMP_CONFIGURATION_ITEMS,
  91         ACTION_DUMP_BUS_PROPERTIES,
  92 } arg_action = ACTION_RUN;
  93 static char *arg_default_unit = NULL;
  94 static bool arg_system = false;
  95 static bool arg_dump_core = true;
  96 static int arg_crash_chvt = -1;
  97 static bool arg_crash_shell = false;
  98 static bool arg_crash_reboot = false;
  99 static char *arg_confirm_spawn = NULL;
 100 static ShowStatus arg_show_status = _SHOW_STATUS_INVALID;
 101 static bool arg_switched_root = false;
 102 static PagerFlags arg_pager_flags = 0;
 103 static bool arg_service_watchdogs = true;
 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 int arg_default_cpu_accounting = -1;
 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_exit_or_reboot(void) {
 134
 135         /* If we are running in a contianer, let's prefer exiting, after all we can propagate an exit code to the
 136          * container manager, and thus inform it that something went wrong. */
 137         if (detect_container() > 0) {
 138                 log_emergency("Exiting PID 1...");
 139                 exit(EXIT_EXCEPTION);
 140         }
 141
 142         if (arg_crash_reboot) {
 143                 log_notice("Rebooting in 10s...");
 144                 (void) sleep(10);
 145
 146                 log_notice("Rebooting now...");
 147                 (void) reboot(RB_AUTOBOOT);
 148                 log_emergency_errno(errno, "Failed to reboot: %m");
 149         }
 150
 151         log_emergency("Freezing execution.");
 152         freeze();
 153 }
 154
 155 _noreturn_ static void crash(int sig) {
 156         struct sigaction sa;
 157         pid_t pid;
 158
 159         if (getpid_cached() != 1)
 160                 /* Pass this on immediately, if this is not PID 1 */
 161                 (void) raise(sig);
 162         else if (!arg_dump_core)
 163                 log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig));
 164         else {
 165                 sa = (struct sigaction) {
 166                         .sa_handler = nop_signal_handler,
 167                         .sa_flags = SA_NOCLDSTOP|SA_RESTART,
 168                 };
 169
 170                 /* We want to wait for the core process, hence let's enable SIGCHLD */
 171                 (void) sigaction(SIGCHLD, &sa, NULL);
 172
 173                 pid = raw_clone(SIGCHLD);
 174                 if (pid < 0)
 175                         log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig));
 176                 else if (pid == 0) {
 177                         /* Enable default signal handler for core dump */
 178
 179                         sa = (struct sigaction) {
 180                                 .sa_handler = SIG_DFL,
 181                         };
 182                         (void) sigaction(sig, &sa, NULL);
 183
 184                         /* Don't limit the coredump size */
 185                         (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY));
 186
 187                         /* Just to be sure... */
 188                         (void) chdir("/");
 189
 190                         /* Raise the signal again */
 191                         pid = raw_getpid();
 192                         (void) kill(pid, sig); /* raise() would kill the parent */
 193
 194                         assert_not_reached("We shouldn't be here...");
 195                         _exit(EXIT_EXCEPTION);
 196                 } else {
 197                         siginfo_t status;
 198                         int r;
 199
 200                         /* Order things nicely. */
 201                         r = wait_for_terminate(pid, &status);
 202                         if (r < 0)
 203                                 log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig));
 204                         else if (status.si_code != CLD_DUMPED)
 205                                 log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).",
 206                                               signal_to_string(sig),
 207                                               pid, sigchld_code_to_string(status.si_code),
 208                                               status.si_status,
 209                                               strna(status.si_code == CLD_EXITED
 210                                                     ? exit_status_to_string(status.si_status, EXIT_STATUS_MINIMAL)
 211                                                     : signal_to_string(status.si_status)));
 212                         else
 213                                 log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid);
 214                 }
 215         }
 216
 217         if (arg_crash_chvt >= 0)
 218                 (void) chvt(arg_crash_chvt);
 219
 220         sa = (struct sigaction) {
 221                 .sa_handler = SIG_IGN,
 222                 .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART,
 223         };
 224
 225         /* Let the kernel reap children for us */
 226         (void) sigaction(SIGCHLD, &sa, NULL);
 227
 228         if (arg_crash_shell) {
 229                 log_notice("Executing crash shell in 10s...");
 230                 (void) sleep(10);
 231
 232                 pid = raw_clone(SIGCHLD);
 233                 if (pid < 0)
 234                         log_emergency_errno(errno, "Failed to fork off crash shell: %m");
 235                 else if (pid == 0) {
 236                         (void) setsid();
 237                         (void) make_console_stdio();
 238                         (void) execle("/bin/sh", "/bin/sh", NULL, environ);
 239
 240                         log_emergency_errno(errno, "execle() failed: %m");
 241                         _exit(EXIT_EXCEPTION);
 242                 } else {
 243                         log_info("Spawned crash shell as PID "PID_FMT".", pid);
 244                         (void) wait_for_terminate(pid, NULL);
 245                 }
 246         }
 247
 248         freeze_or_exit_or_reboot();
 249 }
 250
 251 static void install_crash_handler(void) {
 252         static const struct sigaction sa = {
 253                 .sa_handler = crash,
 254                 .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */
 255         };
 256         int r;
 257
 258         /* We ignore the return value here, since, we don't mind if we
 259          * cannot set up a crash handler */
 260         r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1);
 261         if (r < 0)
 262                 log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m");
 263 }
 264
 265 static int console_setup(void) {
 266         _cleanup_close_ int tty_fd = -1;
 267         int r;
 268
 269         tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
 270         if (tty_fd < 0)
 271                 return log_error_errno(tty_fd, "Failed to open /dev/console: %m");
 272
 273         /* We don't want to force text mode.  plymouth may be showing
 274          * pictures already from initrd. */
 275         r = reset_terminal_fd(tty_fd, false);
 276         if (r < 0)
 277                 return log_error_errno(r, "Failed to reset /dev/console: %m");
 278
 279         return 0;
 280 }
 281
 282 static int parse_crash_chvt(const char *value) {
 283         int b;
 284
 285         if (safe_atoi(value, &arg_crash_chvt) >= 0)
 286                 return 0;
 287
 288         b = parse_boolean(value);
 289         if (b < 0)
 290                 return b;
 291
 292         if (b > 0)
 293                 arg_crash_chvt = 0; /* switch to where kmsg goes */
 294         else
 295                 arg_crash_chvt = -1; /* turn off switching */
 296
 297         return 0;
 298 }
 299
 300 static int parse_confirm_spawn(const char *value, char **console) {
 301         char *s;
 302         int r;
 303
 304         r = value ? parse_boolean(value) : 1;
 305         if (r == 0) {
 306                 *console = NULL;
 307                 return 0;
 308         }
 309
 310         if (r > 0) /* on with default tty */
 311                 s = strdup("/dev/console");
 312         else if (is_path(value)) /* on with fully qualified path */
 313                 s = strdup(value);
 314         else /* on with only a tty file name, not a fully qualified path */
 315                 s = strjoin("/dev/", value);
 316         if (!s)
 317                 return -ENOMEM;
 318
 319         *console = s;
 320         return 0;
 321 }
 322
 323 static int set_machine_id(const char *m) {
 324         sd_id128_t t;
 325         assert(m);
 326
 327         if (sd_id128_from_string(m, &t) < 0)
 328                 return -EINVAL;
 329
 330         if (sd_id128_is_null(t))
 331                 return -EINVAL;
 332
 333         arg_machine_id = t;
 334         return 0;
 335 }
 336
 337 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
 338
 339         int r;
 340
 341         assert(key);
 342
 343         if (STR_IN_SET(key, "systemd.unit", "rd.systemd.unit")) {
 344
 345                 if (proc_cmdline_value_missing(key, value))
 346                         return 0;
 347
 348                 if (!unit_name_is_valid(value, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE))
 349                         log_warning("Unit name specified on %s= is not valid, ignoring: %s", key, value);
 350                 else if (in_initrd() == !!startswith(key, "rd.")) {
 351                         if (free_and_strdup(&arg_default_unit, value) < 0)
 352                                 return log_oom();
 353                 }
 354
 355         } else if (proc_cmdline_key_streq(key, "systemd.dump_core")) {
 356
 357                 r = value ? parse_boolean(value) : true;
 358                 if (r < 0)
 359                         log_warning_errno(r, "Failed to parse dump core switch %s, ignoring: %m", value);
 360                 else
 361                         arg_dump_core = r;
 362
 363         } else if (proc_cmdline_key_streq(key, "systemd.early_core_pattern")) {
 364
 365                 if (proc_cmdline_value_missing(key, value))
 366                         return 0;
 367
 368                 if (path_is_absolute(value))
 369                         (void) parse_path_argument_and_warn(value, false, &arg_early_core_pattern);
 370                 else
 371                         log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value);
 372
 373         } else if (proc_cmdline_key_streq(key, "systemd.crash_chvt")) {
 374
 375                 if (!value)
 376                         arg_crash_chvt = 0; /* turn on */
 377                 else {
 378                         r = parse_crash_chvt(value);
 379                         if (r < 0)
 380                                 log_warning_errno(r, "Failed to parse crash chvt switch %s, ignoring: %m", value);
 381                 }
 382
 383         } else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) {
 384
 385                 r = value ? parse_boolean(value) : true;
 386                 if (r < 0)
 387                         log_warning_errno(r, "Failed to parse crash shell switch %s, ignoring: %m", value);
 388                 else
 389                         arg_crash_shell = r;
 390
 391         } else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) {
 392
 393                 r = value ? parse_boolean(value) : true;
 394                 if (r < 0)
 395                         log_warning_errno(r, "Failed to parse crash reboot switch %s, ignoring: %m", value);
 396                 else
 397                         arg_crash_reboot = r;
 398
 399         } else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) {
 400                 char *s;
 401
 402                 r = parse_confirm_spawn(value, &s);
 403                 if (r < 0)
 404                         log_warning_errno(r, "Failed to parse confirm_spawn switch %s, ignoring: %m", value);
 405                 else
 406                         free_and_replace(arg_confirm_spawn, s);
 407
 408         } else if (proc_cmdline_key_streq(key, "systemd.service_watchdogs")) {
 409
 410                 r = value ? parse_boolean(value) : true;
 411                 if (r < 0)
 412                         log_warning_errno(r, "Failed to parse service watchdog switch %s, ignoring: %m", value);
 413                 else
 414                         arg_service_watchdogs = r;
 415
 416         } else if (proc_cmdline_key_streq(key, "systemd.show_status")) {
 417
 418                 if (value) {
 419                         r = parse_show_status(value, &arg_show_status);
 420                         if (r < 0)
 421                                 log_warning_errno(r, "Failed to parse show status switch %s, ignoring: %m", value);
 422                 } else
 423                         arg_show_status = SHOW_STATUS_YES;
 424
 425         } else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) {
 426
 427                 if (proc_cmdline_value_missing(key, value))
 428                         return 0;
 429
 430                 r = exec_output_from_string(value);
 431                 if (r < 0)
 432                         log_warning_errno(r, "Failed to parse default standard output switch %s, ignoring: %m", value);
 433                 else
 434                         arg_default_std_output = r;
 435
 436         } else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) {
 437
 438                 if (proc_cmdline_value_missing(key, value))
 439                         return 0;
 440
 441                 r = exec_output_from_string(value);
 442                 if (r < 0)
 443                         log_warning_errno(r, "Failed to parse default standard error switch %s, ignoring: %m", value);
 444                 else
 445                         arg_default_std_error = r;
 446
 447         } else if (streq(key, "systemd.setenv")) {
 448
 449                 if (proc_cmdline_value_missing(key, value))
 450                         return 0;
 451
 452                 if (env_assignment_is_valid(value)) {
 453                         char **env;
 454
 455                         env = strv_env_set(arg_default_environment, value);
 456                         if (!env)
 457                                 return log_oom();
 458
 459                         arg_default_environment = env;
 460                 } else
 461                         log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
 462
 463         } else if (proc_cmdline_key_streq(key, "systemd.machine_id")) {
 464
 465                 if (proc_cmdline_value_missing(key, value))
 466                         return 0;
 467
 468                 r = set_machine_id(value);
 469                 if (r < 0)
 470                         log_warning_errno(r, "MachineID '%s' is not valid, ignoring: %m", value);
 471
 472         } else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) {
 473
 474                 if (proc_cmdline_value_missing(key, value))
 475                         return 0;
 476
 477                 r = parse_sec(value, &arg_default_timeout_start_usec);
 478                 if (r < 0)
 479                         log_warning_errno(r, "Failed to parse default start timeout '%s', ignoring: %m", value);
 480
 481                 if (arg_default_timeout_start_usec <= 0)
 482                         arg_default_timeout_start_usec = USEC_INFINITY;
 483
 484         } else if (proc_cmdline_key_streq(key, "systemd.watchdog_device")) {
 485
 486                 if (proc_cmdline_value_missing(key, value))
 487                         return 0;
 488
 489                 (void) parse_path_argument_and_warn(value, false, &arg_watchdog_device);
 490
 491         } else if (streq(key, "quiet") && !value) {
 492
 493                 if (arg_show_status == _SHOW_STATUS_INVALID)
 494                         arg_show_status = SHOW_STATUS_AUTO;
 495
 496         } else if (streq(key, "debug") && !value) {
 497
 498                 /* Note that log_parse_environment() handles 'debug'
 499                  * too, and sets the log level to LOG_DEBUG. */
 500
 501                 if (detect_container() > 0)
 502                         log_set_target(LOG_TARGET_CONSOLE);
 503
 504         } else if (!value) {
 505                 const char *target;
 506
 507                 /* SysV compatibility */
 508                 target = runlevel_to_target(key);
 509                 if (target)
 510                         return free_and_strdup(&arg_default_unit, target);
 511         }
 512
 513         return 0;
 514 }
 515
 516 #define DEFINE_SETTER(name, func, descr)                              \
 517         static int name(const char *unit,                             \
 518                         const char *filename,                         \
 519                         unsigned line,                                \
 520                         const char *section,                          \
 521                         unsigned section_line,                        \
 522                         const char *lvalue,                           \
 523                         int ltype,                                    \
 524                         const char *rvalue,                           \
 525                         void *data,                                   \
 526                         void *userdata) {                             \
 527                                                                       \
 528                 int r;                                                \
 529                                                                       \
 530                 assert(filename);                                     \
 531                 assert(lvalue);                                       \
 532                 assert(rvalue);                                       \
 533                                                                       \
 534                 r = func(rvalue);                                     \
 535                 if (r < 0)                                            \
 536                         log_syntax(unit, LOG_ERR, filename, line, r,  \
 537                                    "Invalid " descr "'%s': %m",       \
 538                                    rvalue);                           \
 539                                                                       \
 540                 return 0;                                             \
 541         }
 542
 543 DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level");
 544 DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target");
 545 DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" );
 546 DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location");
 547
 548 static int config_parse_cpu_affinity2(
 549                 const char *unit,
 550                 const char *filename,
 551                 unsigned line,
 552                 const char *section,
 553                 unsigned section_line,
 554                 const char *lvalue,
 555                 int ltype,
 556                 const char *rvalue,
 557                 void *data,
 558                 void *userdata) {
 559
 560         _cleanup_cpu_free_ cpu_set_t *c = NULL;
 561         int ncpus;
 562
 563         ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue);
 564         if (ncpus < 0)
 565                 return ncpus;
 566
 567         if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0)
 568                 log_warning_errno(errno, "Failed to set CPU affinity: %m");
 569
 570         return 0;
 571 }
 572
 573 static int config_parse_show_status(
 574                 const char* unit,
 575                 const char *filename,
 576                 unsigned line,
 577                 const char *section,
 578                 unsigned section_line,
 579                 const char *lvalue,
 580                 int ltype,
 581                 const char *rvalue,
 582                 void *data,
 583                 void *userdata) {
 584
 585         int k;
 586         ShowStatus *b = data;
 587
 588         assert(filename);
 589         assert(lvalue);
 590         assert(rvalue);
 591         assert(data);
 592
 593         k = parse_show_status(rvalue, b);
 594         if (k < 0) {
 595                 log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue);
 596                 return 0;
 597         }
 598
 599         return 0;
 600 }
 601
 602 static int config_parse_output_restricted(
 603                 const char* unit,
 604                 const char *filename,
 605                 unsigned line,
 606                 const char *section,
 607                 unsigned section_line,
 608                 const char *lvalue,
 609                 int ltype,
 610                 const char *rvalue,
 611                 void *data,
 612                 void *userdata) {
 613
 614         ExecOutput t, *eo = data;
 615
 616         assert(filename);
 617         assert(lvalue);
 618         assert(rvalue);
 619         assert(data);
 620
 621         t = exec_output_from_string(rvalue);
 622         if (t < 0) {
 623                 log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse output type, ignoring: %s", rvalue);
 624                 return 0;
 625         }
 626
 627         if (IN_SET(t, EXEC_OUTPUT_SOCKET, EXEC_OUTPUT_NAMED_FD, EXEC_OUTPUT_FILE, EXEC_OUTPUT_FILE_APPEND)) {
 628                 log_syntax(unit, LOG_ERR, filename, line, 0, "Standard output types socket, fd:, file:, append: are not supported as defaults, ignoring: %s", rvalue);
 629                 return 0;
 630         }
 631
 632         *eo = t;
 633         return 0;
 634 }
 635
 636 static int config_parse_crash_chvt(
 637                 const char* unit,
 638                 const char *filename,
 639                 unsigned line,
 640                 const char *section,
 641                 unsigned section_line,
 642                 const char *lvalue,
 643                 int ltype,
 644                 const char *rvalue,
 645                 void *data,
 646                 void *userdata) {
 647
 648         int r;
 649
 650         assert(filename);
 651         assert(lvalue);
 652         assert(rvalue);
 653
 654         r = parse_crash_chvt(rvalue);
 655         if (r < 0) {
 656                 log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue);
 657                 return 0;
 658         }
 659
 660         return 0;
 661 }
 662
 663 static int parse_config_file(void) {
 664
 665         const ConfigTableItem items[] = {
 666                 { "Manager", "LogLevel",                  config_parse_level2,           0, NULL                                   },
 667                 { "Manager", "LogTarget",                 config_parse_target,           0, NULL                                   },
 668                 { "Manager", "LogColor",                  config_parse_color,            0, NULL                                   },
 669                 { "Manager", "LogLocation",               config_parse_location,         0, NULL                                   },
 670                 { "Manager", "DumpCore",                  config_parse_bool,             0, &arg_dump_core                         },
 671                 { "Manager", "CrashChVT", /* legacy */    config_parse_crash_chvt,       0, NULL                                   },
 672                 { "Manager", "CrashChangeVT",             config_parse_crash_chvt,       0, NULL                                   },
 673                 { "Manager", "CrashShell",                config_parse_bool,             0, &arg_crash_shell                       },
 674                 { "Manager", "CrashReboot",               config_parse_bool,             0, &arg_crash_reboot                      },
 675                 { "Manager", "ShowStatus",                config_parse_show_status,      0, &arg_show_status                       },
 676                 { "Manager", "CPUAffinity",               config_parse_cpu_affinity2,    0, NULL                                   },
 677                 { "Manager", "JoinControllers",           config_parse_warn_compat,      DISABLED_CONFIGURATION, NULL              },
 678                 { "Manager", "RuntimeWatchdogSec",        config_parse_sec,              0, &arg_runtime_watchdog                  },
 679                 { "Manager", "ShutdownWatchdogSec",       config_parse_sec,              0, &arg_shutdown_watchdog                 },
 680                 { "Manager", "WatchdogDevice",            config_parse_path,             0, &arg_watchdog_device                   },
 681                 { "Manager", "CapabilityBoundingSet",     config_parse_capability_set,   0, &arg_capability_bounding_set           },
 682                 { "Manager", "NoNewPrivileges",           config_parse_bool,             0, &arg_no_new_privs                      },
 683 #if HAVE_SECCOMP
 684                 { "Manager", "SystemCallArchitectures",   config_parse_syscall_archs,    0, &arg_syscall_archs                     },
 685 #endif
 686                 { "Manager", "TimerSlackNSec",            config_parse_nsec,             0, &arg_timer_slack_nsec                  },
 687                 { "Manager", "DefaultTimerAccuracySec",   config_parse_sec,              0, &arg_default_timer_accuracy_usec       },
 688                 { "Manager", "DefaultStandardOutput",     config_parse_output_restricted,0, &arg_default_std_output                },
 689                 { "Manager", "DefaultStandardError",      config_parse_output_restricted,0, &arg_default_std_error                 },
 690                 { "Manager", "DefaultTimeoutStartSec",    config_parse_sec,              0, &arg_default_timeout_start_usec        },
 691                 { "Manager", "DefaultTimeoutStopSec",     config_parse_sec,              0, &arg_default_timeout_stop_usec         },
 692                 { "Manager", "DefaultRestartSec",         config_parse_sec,              0, &arg_default_restart_usec              },
 693                 { "Manager", "DefaultStartLimitInterval", config_parse_sec,              0, &arg_default_start_limit_interval      }, /* obsolete alias */
 694                 { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec,            0, &arg_default_start_limit_interval      },
 695                 { "Manager", "DefaultStartLimitBurst",    config_parse_unsigned,         0, &arg_default_start_limit_burst         },
 696                 { "Manager", "DefaultEnvironment",        config_parse_environ,          0, &arg_default_environment               },
 697                 { "Manager", "DefaultLimitCPU",           config_parse_rlimit,           RLIMIT_CPU, arg_default_rlimit            },
 698                 { "Manager", "DefaultLimitFSIZE",         config_parse_rlimit,           RLIMIT_FSIZE, arg_default_rlimit          },
 699                 { "Manager", "DefaultLimitDATA",          config_parse_rlimit,           RLIMIT_DATA, arg_default_rlimit           },
 700                 { "Manager", "DefaultLimitSTACK",         config_parse_rlimit,           RLIMIT_STACK, arg_default_rlimit          },
 701                 { "Manager", "DefaultLimitCORE",          config_parse_rlimit,           RLIMIT_CORE, arg_default_rlimit           },
 702                 { "Manager", "DefaultLimitRSS",           config_parse_rlimit,           RLIMIT_RSS, arg_default_rlimit            },
 703                 { "Manager", "DefaultLimitNOFILE",        config_parse_rlimit,           RLIMIT_NOFILE, arg_default_rlimit         },
 704                 { "Manager", "DefaultLimitAS",            config_parse_rlimit,           RLIMIT_AS, arg_default_rlimit             },
 705                 { "Manager", "DefaultLimitNPROC",         config_parse_rlimit,           RLIMIT_NPROC, arg_default_rlimit          },
 706                 { "Manager", "DefaultLimitMEMLOCK",       config_parse_rlimit,           RLIMIT_MEMLOCK, arg_default_rlimit        },
 707                 { "Manager", "DefaultLimitLOCKS",         config_parse_rlimit,           RLIMIT_LOCKS, arg_default_rlimit          },
 708                 { "Manager", "DefaultLimitSIGPENDING",    config_parse_rlimit,           RLIMIT_SIGPENDING, arg_default_rlimit     },
 709                 { "Manager", "DefaultLimitMSGQUEUE",      config_parse_rlimit,           RLIMIT_MSGQUEUE, arg_default_rlimit       },
 710                 { "Manager", "DefaultLimitNICE",          config_parse_rlimit,           RLIMIT_NICE, arg_default_rlimit           },
 711                 { "Manager", "DefaultLimitRTPRIO",        config_parse_rlimit,           RLIMIT_RTPRIO, arg_default_rlimit         },
 712                 { "Manager", "DefaultLimitRTTIME",        config_parse_rlimit,           RLIMIT_RTTIME, arg_default_rlimit         },
 713                 { "Manager", "DefaultCPUAccounting",      config_parse_tristate,         0, &arg_default_cpu_accounting            },
 714                 { "Manager", "DefaultIOAccounting",       config_parse_bool,             0, &arg_default_io_accounting             },
 715                 { "Manager", "DefaultIPAccounting",       config_parse_bool,             0, &arg_default_ip_accounting             },
 716                 { "Manager", "DefaultBlockIOAccounting",  config_parse_bool,             0, &arg_default_blockio_accounting        },
 717                 { "Manager", "DefaultMemoryAccounting",   config_parse_bool,             0, &arg_default_memory_accounting         },
 718                 { "Manager", "DefaultTasksAccounting",    config_parse_bool,             0, &arg_default_tasks_accounting          },
 719                 { "Manager", "DefaultTasksMax",           config_parse_tasks_max,        0, &arg_default_tasks_max                 },
 720                 { "Manager", "CtrlAltDelBurstAction",     config_parse_emergency_action, 0, &arg_cad_burst_action                  },
 721                 {}
 722         };
 723
 724         const char *fn, *conf_dirs_nulstr;
 725
 726         fn = arg_system ?
 727                 PKGSYSCONFDIR "/system.conf" :
 728                 PKGSYSCONFDIR "/user.conf";
 729
 730         conf_dirs_nulstr = arg_system ?
 731                 CONF_PATHS_NULSTR("systemd/system.conf.d") :
 732                 CONF_PATHS_NULSTR("systemd/user.conf.d");
 733
 734         (void) config_parse_many_nulstr(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, CONFIG_PARSE_WARN, NULL);
 735
 736         /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY
 737          * like everywhere else. */
 738         if (arg_default_timeout_start_usec <= 0)
 739                 arg_default_timeout_start_usec = USEC_INFINITY;
 740         if (arg_default_timeout_stop_usec <= 0)
 741                 arg_default_timeout_stop_usec = USEC_INFINITY;
 742
 743         return 0;
 744 }
 745
 746 static void set_manager_defaults(Manager *m) {
 747
 748         assert(m);
 749
 750         /* Propagates the various default unit property settings into the manager object, i.e. properties that do not
 751          * affect the manager itself, but are just what newly allocated units will have set if they haven't set
 752          * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */
 753
 754         m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec;
 755         m->default_std_output = arg_default_std_output;
 756         m->default_std_error = arg_default_std_error;
 757         m->default_timeout_start_usec = arg_default_timeout_start_usec;
 758         m->default_timeout_stop_usec = arg_default_timeout_stop_usec;
 759         m->default_restart_usec = arg_default_restart_usec;
 760         m->default_start_limit_interval = arg_default_start_limit_interval;
 761         m->default_start_limit_burst = arg_default_start_limit_burst;
 762
 763         /* On 4.15+ with unified hierarchy, CPU accounting is essentially free as it doesn't require the CPU
 764          * controller to be enabled, so the default is to enable it unless we got told otherwise. */
 765         if (arg_default_cpu_accounting >= 0)
 766                 m->default_cpu_accounting = arg_default_cpu_accounting;
 767         else
 768                 m->default_cpu_accounting = cpu_accounting_is_cheap();
 769
 770         m->default_io_accounting = arg_default_io_accounting;
 771         m->default_ip_accounting = arg_default_ip_accounting;
 772         m->default_blockio_accounting = arg_default_blockio_accounting;
 773         m->default_memory_accounting = arg_default_memory_accounting;
 774         m->default_tasks_accounting = arg_default_tasks_accounting;
 775         m->default_tasks_max = arg_default_tasks_max;
 776
 777         (void) manager_set_default_rlimits(m, arg_default_rlimit);
 778
 779         (void) manager_default_environment(m);
 780         (void) manager_transient_environment_add(m, arg_default_environment);
 781 }
 782
 783 static void set_manager_settings(Manager *m) {
 784
 785         assert(m);
 786
 787         /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager
 788          * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */
 789
 790         m->confirm_spawn = arg_confirm_spawn;
 791         m->service_watchdogs = arg_service_watchdogs;
 792         m->runtime_watchdog = arg_runtime_watchdog;
 793         m->shutdown_watchdog = arg_shutdown_watchdog;
 794         m->cad_burst_action = arg_cad_burst_action;
 795
 796         manager_set_show_status(m, arg_show_status);
 797 }
 798
 799 static int parse_argv(int argc, char *argv[]) {
 800         enum {
 801                 ARG_LOG_LEVEL = 0x100,
 802                 ARG_LOG_TARGET,
 803                 ARG_LOG_COLOR,
 804                 ARG_LOG_LOCATION,
 805                 ARG_UNIT,
 806                 ARG_SYSTEM,
 807                 ARG_USER,
 808                 ARG_TEST,
 809                 ARG_NO_PAGER,
 810                 ARG_VERSION,
 811                 ARG_DUMP_CONFIGURATION_ITEMS,
 812                 ARG_DUMP_BUS_PROPERTIES,
 813                 ARG_DUMP_CORE,
 814                 ARG_CRASH_CHVT,
 815                 ARG_CRASH_SHELL,
 816                 ARG_CRASH_REBOOT,
 817                 ARG_CONFIRM_SPAWN,
 818                 ARG_SHOW_STATUS,
 819                 ARG_DESERIALIZE,
 820                 ARG_SWITCHED_ROOT,
 821                 ARG_DEFAULT_STD_OUTPUT,
 822                 ARG_DEFAULT_STD_ERROR,
 823                 ARG_MACHINE_ID,
 824                 ARG_SERVICE_WATCHDOGS,
 825         };
 826
 827         static const struct option options[] = {
 828                 { "log-level",                required_argument, NULL, ARG_LOG_LEVEL                },
 829                 { "log-target",               required_argument, NULL, ARG_LOG_TARGET               },
 830                 { "log-color",                optional_argument, NULL, ARG_LOG_COLOR                },
 831                 { "log-location",             optional_argument, NULL, ARG_LOG_LOCATION             },
 832                 { "unit",                     required_argument, NULL, ARG_UNIT                     },
 833                 { "system",                   no_argument,       NULL, ARG_SYSTEM                   },
 834                 { "user",                     no_argument,       NULL, ARG_USER                     },
 835                 { "test",                     no_argument,       NULL, ARG_TEST                     },
 836                 { "no-pager",                 no_argument,       NULL, ARG_NO_PAGER                 },
 837                 { "help",                     no_argument,       NULL, 'h'                          },
 838                 { "version",                  no_argument,       NULL, ARG_VERSION                  },
 839                 { "dump-configuration-items", no_argument,       NULL, ARG_DUMP_CONFIGURATION_ITEMS },
 840                 { "dump-bus-properties",      no_argument,       NULL, ARG_DUMP_BUS_PROPERTIES      },
 841                 { "dump-core",                optional_argument, NULL, ARG_DUMP_CORE                },
 842                 { "crash-chvt",               required_argument, NULL, ARG_CRASH_CHVT               },
 843                 { "crash-shell",              optional_argument, NULL, ARG_CRASH_SHELL              },
 844                 { "crash-reboot",             optional_argument, NULL, ARG_CRASH_REBOOT             },
 845                 { "confirm-spawn",            optional_argument, NULL, ARG_CONFIRM_SPAWN            },
 846                 { "show-status",              optional_argument, NULL, ARG_SHOW_STATUS              },
 847                 { "deserialize",              required_argument, NULL, ARG_DESERIALIZE              },
 848                 { "switched-root",            no_argument,       NULL, ARG_SWITCHED_ROOT            },
 849                 { "default-standard-output",  required_argument, NULL, ARG_DEFAULT_STD_OUTPUT,      },
 850                 { "default-standard-error",   required_argument, NULL, ARG_DEFAULT_STD_ERROR,       },
 851                 { "machine-id",               required_argument, NULL, ARG_MACHINE_ID               },
 852                 { "service-watchdogs",        required_argument, NULL, ARG_SERVICE_WATCHDOGS        },
 853                 {}
 854         };
 855
 856         int c, r;
 857
 858         assert(argc >= 1);
 859         assert(argv);
 860
 861         if (getpid_cached() == 1)
 862                 opterr = 0;
 863
 864         while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0)
 865
 866                 switch (c) {
 867
 868                 case ARG_LOG_LEVEL:
 869                         r = log_set_max_level_from_string(optarg);
 870                         if (r < 0)
 871                                 return log_error_errno(r, "Failed to parse log level \"%s\": %m", optarg);
 872
 873                         break;
 874
 875                 case ARG_LOG_TARGET:
 876                         r = log_set_target_from_string(optarg);
 877                         if (r < 0)
 878                                 return log_error_errno(r, "Failed to parse log target \"%s\": %m", optarg);
 879
 880                         break;
 881
 882                 case ARG_LOG_COLOR:
 883
 884                         if (optarg) {
 885                                 r = log_show_color_from_string(optarg);
 886                                 if (r < 0)
 887                                         return log_error_errno(r, "Failed to parse log color setting \"%s\": %m",
 888                                                                optarg);
 889                         } else
 890                                 log_show_color(true);
 891
 892                         break;
 893
 894                 case ARG_LOG_LOCATION:
 895                         if (optarg) {
 896                                 r = log_show_location_from_string(optarg);
 897                                 if (r < 0)
 898                                         return log_error_errno(r, "Failed to parse log location setting \"%s\": %m",
 899                                                                optarg);
 900                         } else
 901                                 log_show_location(true);
 902
 903                         break;
 904
 905                 case ARG_DEFAULT_STD_OUTPUT:
 906                         r = exec_output_from_string(optarg);
 907                         if (r < 0)
 908                                 return log_error_errno(r, "Failed to parse default standard output setting \"%s\": %m",
 909                                                        optarg);
 910                         arg_default_std_output = r;
 911                         break;
 912
 913                 case ARG_DEFAULT_STD_ERROR:
 914                         r = exec_output_from_string(optarg);
 915                         if (r < 0)
 916                                 return log_error_errno(r, "Failed to parse default standard error output setting \"%s\": %m",
 917                                                        optarg);
 918                         arg_default_std_error = r;
 919                         break;
 920
 921                 case ARG_UNIT:
 922                         r = free_and_strdup(&arg_default_unit, optarg);
 923                         if (r < 0)
 924                                 return log_error_errno(r, "Failed to set default unit \"%s\": %m", optarg);
 925
 926                         break;
 927
 928                 case ARG_SYSTEM:
 929                         arg_system = true;
 930                         break;
 931
 932                 case ARG_USER:
 933                         arg_system = false;
 934                         break;
 935
 936                 case ARG_TEST:
 937                         arg_action = ACTION_TEST;
 938                         break;
 939
 940                 case ARG_NO_PAGER:
 941                         arg_pager_flags |= PAGER_DISABLE;
 942                         break;
 943
 944                 case ARG_VERSION:
 945                         arg_action = ACTION_VERSION;
 946                         break;
 947
 948                 case ARG_DUMP_CONFIGURATION_ITEMS:
 949                         arg_action = ACTION_DUMP_CONFIGURATION_ITEMS;
 950                         break;
 951
 952                 case ARG_DUMP_BUS_PROPERTIES:
 953                         arg_action = ACTION_DUMP_BUS_PROPERTIES;
 954                         break;
 955
 956                 case ARG_DUMP_CORE:
 957                         if (!optarg)
 958                                 arg_dump_core = true;
 959                         else {
 960                                 r = parse_boolean(optarg);
 961                                 if (r < 0)
 962                                         return log_error_errno(r, "Failed to parse dump core boolean: \"%s\": %m",
 963                                                                optarg);
 964                                 arg_dump_core = r;
 965                         }
 966                         break;
 967
 968                 case ARG_CRASH_CHVT:
 969                         r = parse_crash_chvt(optarg);
 970                         if (r < 0)
 971                                 return log_error_errno(r, "Failed to parse crash virtual terminal index: \"%s\": %m",
 972                                                        optarg);
 973                         break;
 974
 975                 case ARG_CRASH_SHELL:
 976                         if (!optarg)
 977                                 arg_crash_shell = true;
 978                         else {
 979                                 r = parse_boolean(optarg);
 980                                 if (r < 0)
 981                                         return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
 982                                                                optarg);
 983                                 arg_crash_shell = r;
 984                         }
 985                         break;
 986
 987                 case ARG_CRASH_REBOOT:
 988                         if (!optarg)
 989                                 arg_crash_reboot = true;
 990                         else {
 991                                 r = parse_boolean(optarg);
 992                                 if (r < 0)
 993                                         return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m",
 994                                                                optarg);
 995                                 arg_crash_reboot = r;
 996                         }
 997                         break;
 998
 999                 case ARG_CONFIRM_SPAWN:
1000                         arg_confirm_spawn = mfree(arg_confirm_spawn);
1001
1002                         r = parse_confirm_spawn(optarg, &arg_confirm_spawn);
1003                         if (r < 0)
1004                                 return log_error_errno(r, "Failed to parse confirm spawn option: \"%s\": %m",
1005                                                        optarg);
1006                         break;
1007
1008                 case ARG_SERVICE_WATCHDOGS:
1009                         r = parse_boolean(optarg);
1010                         if (r < 0)
1011                                 return log_error_errno(r, "Failed to parse service watchdogs boolean: \"%s\": %m",
1012                                                        optarg);
1013                         arg_service_watchdogs = r;
1014                         break;
1015
1016                 case ARG_SHOW_STATUS:
1017                         if (optarg) {
1018                                 r = parse_show_status(optarg, &arg_show_status);
1019                                 if (r < 0)
1020                                         return log_error_errno(r, "Failed to parse show status boolean: \"%s\": %m",
1021                                                                optarg);
1022                         } else
1023                                 arg_show_status = SHOW_STATUS_YES;
1024                         break;
1025
1026                 case ARG_DESERIALIZE: {
1027                         int fd;
1028                         FILE *f;
1029
1030                         r = safe_atoi(optarg, &fd);
1031                         if (r < 0)
1032                                 log_error_errno(r, "Failed to parse deserialize option \"%s\": %m", optarg);
1033                         if (fd < 0) {
1034                                 log_error("Invalid deserialize fd: %d", fd);
1035                                 return -EINVAL;
1036                         }
1037
1038                         (void) fd_cloexec(fd, true);
1039
1040                         f = fdopen(fd, "r");
1041                         if (!f)
1042                                 return log_error_errno(errno, "Failed to open serialization fd %d: %m", fd);
1043
1044                         safe_fclose(arg_serialization);
1045                         arg_serialization = f;
1046
1047                         break;
1048                 }
1049
1050                 case ARG_SWITCHED_ROOT:
1051                         arg_switched_root = true;
1052                         break;
1053
1054                 case ARG_MACHINE_ID:
1055                         r = set_machine_id(optarg);
1056                         if (r < 0)
1057                                 return log_error_errno(r, "MachineID '%s' is not valid: %m", optarg);
1058                         break;
1059
1060                 case 'h':
1061                         arg_action = ACTION_HELP;
1062                         break;
1063
1064                 case 'D':
1065                         log_set_max_level(LOG_DEBUG);
1066                         break;
1067
1068                 case 'b':
1069                 case 's':
1070                 case 'z':
1071                         /* Just to eat away the sysvinit kernel
1072                          * cmdline args without getopt() error
1073                          * messages that we'll parse in
1074                          * parse_proc_cmdline_word() or ignore. */
1075
1076                 case '?':
1077                         if (getpid_cached() != 1)
1078                                 return -EINVAL;
1079                         else
1080                                 return 0;
1081
1082                 default:
1083                         assert_not_reached("Unhandled option code.");
1084                 }
1085
1086         if (optind < argc && getpid_cached() != 1) {
1087                 /* Hmm, when we aren't run as init system
1088                  * let's complain about excess arguments */
1089
1090                 log_error("Excess arguments.");
1091                 return -EINVAL;
1092         }
1093
1094         return 0;
1095 }
1096
1097 static int help(void) {
1098         _cleanup_free_ char *link = NULL;
1099         int r;
1100
1101         r = terminal_urlify_man("systemd", "1", &link);
1102         if (r < 0)
1103                 return log_oom();
1104
1105         printf("%s [OPTIONS...]\n\n"
1106                "Starts up and maintains the system or user services.\n\n"
1107                "  -h --help                      Show this help\n"
1108                "     --version                   Show version\n"
1109                "     --test                      Determine startup sequence, dump it and exit\n"
1110                "     --no-pager                  Do not pipe output into a pager\n"
1111                "     --dump-configuration-items  Dump understood unit configuration items\n"
1112                "     --dump-bus-properties       Dump exposed bus properties\n"
1113                "     --unit=UNIT                 Set default unit\n"
1114                "     --system                    Run a system instance, even if PID != 1\n"
1115                "     --user                      Run a user instance\n"
1116                "     --dump-core[=BOOL]          Dump core on crash\n"
1117                "     --crash-vt=NR               Change to specified VT on crash\n"
1118                "     --crash-reboot[=BOOL]       Reboot on crash\n"
1119                "     --crash-shell[=BOOL]        Run shell on crash\n"
1120                "     --confirm-spawn[=BOOL]      Ask for confirmation when spawning processes\n"
1121                "     --show-status[=BOOL]        Show status updates on the console during bootup\n"
1122                "     --log-target=TARGET         Set log target (console, journal, kmsg, journal-or-kmsg, null)\n"
1123                "     --log-level=LEVEL           Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n"
1124                "     --log-color[=BOOL]          Highlight important log messages\n"
1125                "     --log-location[=BOOL]       Include code location in log messages\n"
1126                "     --default-standard-output=  Set default standard output for services\n"
1127                "     --default-standard-error=   Set default standard error output for services\n"
1128                "\nSee the %s for details.\n"
1129                , program_invocation_short_name
1130                , link
1131         );
1132
1133         return 0;
1134 }
1135
1136 static int prepare_reexecute(
1137                 Manager *m,
1138                 FILE **ret_f,
1139                 FDSet **ret_fds,
1140                 bool switching_root) {
1141
1142         _cleanup_fdset_free_ FDSet *fds = NULL;
1143         _cleanup_fclose_ FILE *f = NULL;
1144         int r;
1145
1146         assert(m);
1147         assert(ret_f);
1148         assert(ret_fds);
1149
1150         r = manager_open_serialization(m, &f);
1151         if (r < 0)
1152                 return log_error_errno(r, "Failed to create serialization file: %m");
1153
1154         /* Make sure nothing is really destructed when we shut down */
1155         m->n_reloading++;
1156         bus_manager_send_reloading(m, true);
1157
1158         fds = fdset_new();
1159         if (!fds)
1160                 return log_oom();
1161
1162         r = manager_serialize(m, f, fds, switching_root);
1163         if (r < 0)
1164                 return r;
1165
1166         if (fseeko(f, 0, SEEK_SET) == (off_t) -1)
1167                 return log_error_errno(errno, "Failed to rewind serialization fd: %m");
1168
1169         r = fd_cloexec(fileno(f), false);
1170         if (r < 0)
1171                 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m");
1172
1173         r = fdset_cloexec(fds, false);
1174         if (r < 0)
1175                 return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m");
1176
1177         *ret_f = TAKE_PTR(f);
1178         *ret_fds = TAKE_PTR(fds);
1179
1180         return 0;
1181 }
1182
1183 static void bump_file_max_and_nr_open(void) {
1184
1185         /* Let's bump fs.file-max and fs.nr_open to their respective maximums. On current kernels large numbers of file
1186          * descriptors are no longer a performance problem and their memory is properly tracked by memcg, thus counting
1187          * them and limiting them in another two layers of limits is unnecessary and just complicates things. This
1188          * function hence turns off 2 of the 4 levels of limits on file descriptors, and makes RLIMIT_NOLIMIT (soft +
1189          * hard) the only ones that really matter. */
1190
1191 #if BUMP_PROC_SYS_FS_FILE_MAX || BUMP_PROC_SYS_FS_NR_OPEN
1192         _cleanup_free_ char *t = NULL;
1193         int r;
1194 #endif
1195
1196 #if BUMP_PROC_SYS_FS_FILE_MAX
1197         /* I so wanted to use STRINGIFY(ULONG_MAX) here, but alas we can't as glibc/gcc define that as
1198          * "(0x7fffffffffffffffL * 2UL + 1UL)". Seriously. 😢 */
1199         if (asprintf(&t, "%lu\n", ULONG_MAX) < 0) {
1200                 log_oom();
1201                 return;
1202         }
1203
1204         r = sysctl_write("fs/file-max", t);
1205         if (r < 0)
1206                 log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.file-max, ignoring: %m");
1207 #endif
1208
1209 #if BUMP_PROC_SYS_FS_FILE_MAX && BUMP_PROC_SYS_FS_NR_OPEN
1210         t = mfree(t);
1211 #endif
1212
1213 #if BUMP_PROC_SYS_FS_NR_OPEN
1214         int v = INT_MAX;
1215
1216         /* Arg! The kernel enforces maximum and minimum values on the fs.nr_open, but we don't really know what they
1217          * are. The expression by which the maximum is determined is dependent on the architecture, and is something we
1218          * don't really want to copy to userspace, as it is dependent on implementation details of the kernel. Since
1219          * the kernel doesn't expose the maximum value to us, we can only try and hope. Hence, let's start with
1220          * INT_MAX, and then keep halving the value until we find one that works. Ugly? Yes, absolutely, but kernel
1221          * APIs are kernel APIs, so what do can we do... 🤯 */
1222
1223         for (;;) {
1224                 int k;
1225
1226                 v &= ~(__SIZEOF_POINTER__ - 1); /* Round down to next multiple of the pointer size */
1227                 if (v < 1024) {
1228                         log_warning("Can't bump fs.nr_open, value too small.");
1229                         break;
1230                 }
1231
1232                 k = read_nr_open();
1233                 if (k < 0) {
1234                         log_error_errno(k, "Failed to read fs.nr_open: %m");
1235                         break;
1236                 }
1237                 if (k >= v) { /* Already larger */
1238                         log_debug("Skipping bump, value is already larger.");
1239                         break;
1240                 }
1241
1242                 if (asprintf(&t, "%i\n", v) < 0) {
1243                         log_oom();
1244                         return;
1245                 }
1246
1247                 r = sysctl_write("fs/nr_open", t);
1248                 t = mfree(t);
1249                 if (r == -EINVAL) {
1250                         log_debug("Couldn't write fs.nr_open as %i, halving it.", v);
1251                         v /= 2;
1252                         continue;
1253                 }
1254                 if (r < 0) {
1255                         log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.nr_open, ignoring: %m");
1256                         break;
1257                 }
1258
1259                 log_debug("Successfully bumped fs.nr_open to %i", v);
1260                 break;
1261         }
1262 #endif
1263 }
1264
1265 static int bump_rlimit_nofile(struct rlimit *saved_rlimit) {
1266         int r, nr;
1267
1268         assert(saved_rlimit);
1269
1270         /* Save the original RLIMIT_NOFILE so that we can reset it later when transitioning from the initrd to the main
1271          * systemd or suchlike. */
1272         if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0)
1273                 return log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m");
1274
1275         /* Get the underlying absolute limit the kernel enforces */
1276         nr = read_nr_open();
1277
1278         /* Make sure forked processes get limits based on the original kernel setting */
1279         if (!arg_default_rlimit[RLIMIT_NOFILE]) {
1280                 struct rlimit *rl;
1281
1282                 rl = newdup(struct rlimit, saved_rlimit, 1);
1283                 if (!rl)
1284                         return log_oom();
1285
1286                 /* Bump the hard limit for system services to a substantially higher value. The default hard limit
1287                  * current kernels set is pretty low (4K), mostly for historical reasons. According to kernel
1288                  * developers, the fd handling in recent kernels has been optimized substantially enough, so that we
1289                  * can bump the limit now, without paying too high a price in memory or performance. Note however that
1290                  * we only bump the hard limit, not the soft limit. That's because select() works the way it works, and
1291                  * chokes on fds >= 1024. If we'd bump the soft limit globally, it might accidentally happen to
1292                  * unexpecting programs that they get fds higher than what they can process using select(). By only
1293                  * bumping the hard limit but leaving the low limit as it is we avoid this pitfall: programs that are
1294                  * written by folks aware of the select() problem in mind (and thus use poll()/epoll instead of
1295                  * select(), the way everybody should) can explicitly opt into high fds by bumping their soft limit
1296                  * beyond 1024, to the hard limit we pass. */
1297                 if (arg_system)
1298                         rl->rlim_max = MIN((rlim_t) nr, MAX(rl->rlim_max, (rlim_t) HIGH_RLIMIT_NOFILE));
1299
1300                 arg_default_rlimit[RLIMIT_NOFILE] = rl;
1301         }
1302
1303         /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for
1304          * both hard and soft. */
1305         r = setrlimit_closest(RLIMIT_NOFILE, &RLIMIT_MAKE_CONST(nr));
1306         if (r < 0)
1307                 return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m");
1308
1309         return 0;
1310 }
1311
1312 static int bump_rlimit_memlock(struct rlimit *saved_rlimit) {
1313         int r;
1314
1315         assert(saved_rlimit);
1316
1317         /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK which should
1318          * normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's bump
1319          * the value high enough for our user. */
1320
1321         if (getrlimit(RLIMIT_MEMLOCK, saved_rlimit) < 0)
1322                 return log_warning_errno(errno, "Reading RLIMIT_MEMLOCK failed, ignoring: %m");
1323
1324         r = setrlimit_closest(RLIMIT_MEMLOCK, &RLIMIT_MAKE_CONST(HIGH_RLIMIT_MEMLOCK));
1325         if (r < 0)
1326                 return log_warning_errno(r, "Setting RLIMIT_MEMLOCK failed, ignoring: %m");
1327
1328         return 0;
1329 }
1330
1331 static void test_usr(void) {
1332
1333         /* Check that /usr is not a separate fs */
1334
1335         if (dir_is_empty("/usr") <= 0)
1336                 return;
1337
1338         log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. "
1339                     "Some things will probably break (sometimes even silently) in mysterious ways. "
1340                     "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information.");
1341 }
1342
1343 static int enforce_syscall_archs(Set *archs) {
1344 #if HAVE_SECCOMP
1345         int r;
1346
1347         if (!is_seccomp_available())
1348                 return 0;
1349
1350         r = seccomp_restrict_archs(arg_syscall_archs);
1351         if (r < 0)
1352                 return log_error_errno(r, "Failed to enforce system call architecture restrication: %m");
1353 #endif
1354         return 0;
1355 }
1356
1357 static int status_welcome(void) {
1358         _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL;
1359         int r;
1360
1361         if (IN_SET(arg_show_status, SHOW_STATUS_NO, SHOW_STATUS_AUTO))
1362                 return 0;
1363
1364         r = parse_os_release(NULL,
1365                              "PRETTY_NAME", &pretty_name,
1366                              "ANSI_COLOR", &ansi_color,
1367                              NULL);
1368         if (r < 0)
1369                 log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r,
1370                                "Failed to read os-release file, ignoring: %m");
1371
1372         if (log_get_show_color())
1373                 return status_printf(NULL, false, false,
1374                                      "\nWelcome to \x1B[%sm%s\x1B[0m!\n",
1375                                      isempty(ansi_color) ? "1" : ansi_color,
1376                                      isempty(pretty_name) ? "Linux" : pretty_name);
1377         else
1378                 return status_printf(NULL, false, false,
1379                                      "\nWelcome to %s!\n",
1380                                      isempty(pretty_name) ? "Linux" : pretty_name);
1381 }
1382
1383 static int write_container_id(void) {
1384         const char *c;
1385         int r;
1386
1387         c = getenv("container");
1388         if (isempty(c))
1389                 return 0;
1390
1391         RUN_WITH_UMASK(0022)
1392                 r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE);
1393         if (r < 0)
1394                 return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m");
1395
1396         return 1;
1397 }
1398
1399 static int bump_unix_max_dgram_qlen(void) {
1400         _cleanup_free_ char *qlen = NULL;
1401         unsigned long v;
1402         int r;
1403
1404         /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value
1405          * really really early during boot, so that it is actually applied to all our sockets, including the
1406          * $NOTIFY_SOCKET one. */
1407
1408         r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen);
1409         if (r < 0)
1410                 return log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r, "Failed to read AF_UNIX datagram queue length, ignoring: %m");
1411
1412         r = safe_atolu(qlen, &v);
1413         if (r < 0)
1414                 return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen);
1415
1416         if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN)
1417                 return 0;
1418
1419         r = write_string_filef("/proc/sys/net/unix/max_dgram_qlen", WRITE_STRING_FILE_DISABLE_BUFFER, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN);
1420         if (r < 0)
1421                 return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r,
1422                                       "Failed to bump AF_UNIX datagram queue length, ignoring: %m");
1423
1424         return 1;
1425 }
1426
1427 static int fixup_environment(void) {
1428         _cleanup_free_ char *term = NULL;
1429         const char *t;
1430         int r;
1431
1432         /* Only fix up the environment when we are started as PID 1 */
1433         if (getpid_cached() != 1)
1434                 return 0;
1435
1436         /* We expect the environment to be set correctly if run inside a container. */
1437         if (detect_container() > 0)
1438                 return 0;
1439
1440         /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend
1441          * device used by the console. We try to make a better guess here since some consoles might not have support
1442          * for color mode for example.
1443          *
1444          * However if TERM was configured through the kernel command line then leave it alone. */
1445         r = proc_cmdline_get_key("TERM", 0, &term);
1446         if (r < 0)
1447                 return r;
1448
1449         t = term ?: default_term_for_tty("/dev/console");
1450
1451         if (setenv("TERM", t, 1) < 0)
1452                 return -errno;
1453
1454         return 0;
1455 }
1456
1457 static void redirect_telinit(int argc, char *argv[]) {
1458
1459         /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */
1460
1461 #if HAVE_SYSV_COMPAT
1462         if (getpid_cached() == 1)
1463                 return;
1464
1465         if (!strstr(program_invocation_short_name, "init"))
1466                 return;
1467
1468         execv(SYSTEMCTL_BINARY_PATH, argv);
1469         log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m");
1470         exit(EXIT_FAILURE);
1471 #endif
1472 }
1473
1474 static int become_shutdown(
1475                 const char *shutdown_verb,
1476                 int retval) {
1477
1478         char log_level[DECIMAL_STR_MAX(int) + 1],
1479                 exit_code[DECIMAL_STR_MAX(uint8_t) + 1],
1480                 timeout[DECIMAL_STR_MAX(usec_t) + 1];
1481
1482         const char* command_line[13] = {
1483                 SYSTEMD_SHUTDOWN_BINARY_PATH,
1484                 shutdown_verb,
1485                 "--timeout", timeout,
1486                 "--log-level", log_level,
1487                 "--log-target",
1488         };
1489
1490         _cleanup_strv_free_ char **env_block = NULL;
1491         size_t pos = 7;
1492         int r;
1493
1494         assert(shutdown_verb);
1495         assert(!command_line[pos]);
1496         env_block = strv_copy(environ);
1497
1498         xsprintf(log_level, "%d", log_get_max_level());
1499         xsprintf(timeout, "%" PRI_USEC "us", arg_default_timeout_stop_usec);
1500
1501         switch (log_get_target()) {
1502
1503         case LOG_TARGET_KMSG:
1504         case LOG_TARGET_JOURNAL_OR_KMSG:
1505         case LOG_TARGET_SYSLOG_OR_KMSG:
1506                 command_line[pos++] = "kmsg";
1507                 break;
1508
1509         case LOG_TARGET_NULL:
1510                 command_line[pos++] = "null";
1511                 break;
1512
1513         case LOG_TARGET_CONSOLE:
1514         default:
1515                 command_line[pos++] = "console";
1516                 break;
1517         };
1518
1519         if (log_get_show_color())
1520                 command_line[pos++] = "--log-color";
1521
1522         if (log_get_show_location())
1523                 command_line[pos++] = "--log-location";
1524
1525         if (streq(shutdown_verb, "exit")) {
1526                 command_line[pos++] = "--exit-code";
1527                 command_line[pos++] = exit_code;
1528                 xsprintf(exit_code, "%d", retval);
1529         }
1530
1531         assert(pos < ELEMENTSOF(command_line));
1532
1533         if (streq(shutdown_verb, "reboot") &&
1534             arg_shutdown_watchdog > 0 &&
1535             arg_shutdown_watchdog != USEC_INFINITY) {
1536
1537                 char *e;
1538
1539                 /* If we reboot let's set the shutdown
1540                  * watchdog and tell the shutdown binary to
1541                  * repeatedly ping it */
1542                 r = watchdog_set_timeout(&arg_shutdown_watchdog);
1543                 watchdog_close(r < 0);
1544
1545                 /* Tell the binary how often to ping, ignore failure */
1546                 if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0)
1547                         (void) strv_consume(&env_block, e);
1548
1549                 if (arg_watchdog_device &&
1550                     asprintf(&e, "WATCHDOG_DEVICE=%s", arg_watchdog_device) > 0)
1551                         (void) strv_consume(&env_block, e);
1552         } else
1553                 watchdog_close(true);
1554
1555         /* Avoid the creation of new processes forked by the
1556          * kernel; at this point, we will not listen to the
1557          * signals anyway */
1558         if (detect_container() <= 0)
1559                 (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER);
1560
1561         execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block);
1562         return -errno;
1563 }
1564
1565 static void initialize_clock(void) {
1566         int r;
1567
1568         if (clock_is_localtime(NULL) > 0) {
1569                 int min;
1570
1571                 /*
1572                  * The very first call of settimeofday() also does a time warp in the kernel.
1573                  *
1574                  * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care
1575                  * of maintaining the RTC and do all adjustments.  This matches the behavior of Windows, which leaves
1576                  * the RTC alone if the registry tells that the RTC runs in UTC.
1577                  */
1578                 r = clock_set_timezone(&min);
1579                 if (r < 0)
1580                         log_error_errno(r, "Failed to apply local time delta, ignoring: %m");
1581                 else
1582                         log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min);
1583
1584         } else if (!in_initrd()) {
1585                 /*
1586                  * Do a dummy very first call to seal the kernel's time warp magic.
1587                  *
1588                  * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the
1589                  * real system could be set up that way. In such case, we need to delay the time-warp or the sealing
1590                  * until we reach the real system.
1591                  *
1592                  * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time
1593                  * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will
1594                  * be treated as UTC that way.
1595                  */
1596                 (void) clock_reset_timewarp();
1597         }
1598
1599         r = clock_apply_epoch();
1600         if (r < 0)
1601                 log_error_errno(r, "Current system time is before build time, but cannot correct: %m");
1602         else if (r > 0)
1603                 log_info("System time before build time, advancing clock.");
1604 }
1605
1606 static void initialize_coredump(bool skip_setup) {
1607 #if ENABLE_COREDUMP
1608         if (getpid_cached() != 1)
1609                 return;
1610
1611         /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit)
1612          * will process core dumps for system services by default. */
1613         if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0)
1614                 log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m");
1615
1616         /* But at the same time, turn off the core_pattern logic by default, so that no
1617          * coredumps are stored until the systemd-coredump tool is enabled via
1618          * sysctl. However it can be changed via the kernel command line later so core
1619          * dumps can still be generated during early startup and in initramfs. */
1620         if (!skip_setup)
1621                 disable_coredumps();
1622 #endif
1623 }
1624
1625 static void initialize_core_pattern(bool skip_setup) {
1626         int r;
1627
1628         if (skip_setup || !arg_early_core_pattern)
1629                 return;
1630
1631         if (getpid_cached() != 1)
1632                 return;
1633
1634         r = write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern, WRITE_STRING_FILE_DISABLE_BUFFER);
1635         if (r < 0)
1636                 log_warning_errno(r, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", arg_early_core_pattern);
1637 }
1638
1639 static void do_reexecute(
1640                 int argc,
1641                 char *argv[],
1642                 const struct rlimit *saved_rlimit_nofile,
1643                 const struct rlimit *saved_rlimit_memlock,
1644                 FDSet *fds,
1645                 const char *switch_root_dir,
1646                 const char *switch_root_init,
1647                 const char **ret_error_message) {
1648
1649         unsigned i, j, args_size;
1650         const char **args;
1651         int r;
1652
1653         assert(saved_rlimit_nofile);
1654         assert(saved_rlimit_memlock);
1655         assert(ret_error_message);
1656
1657         /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while
1658          * we do that */
1659         watchdog_close(true);
1660
1661         /* Reset the RLIMIT_NOFILE to the kernel default, so that the new systemd can pass the kernel default to its
1662          * child processes */
1663
1664         if (saved_rlimit_nofile->rlim_cur > 0)
1665                 (void) setrlimit(RLIMIT_NOFILE, saved_rlimit_nofile);
1666         if (saved_rlimit_memlock->rlim_cur != (rlim_t) -1)
1667                 (void) setrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock);
1668
1669         if (switch_root_dir) {
1670                 /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the
1671                  * SIGCHLD for them after deserializing. */
1672                 broadcast_signal(SIGTERM, false, true, arg_default_timeout_stop_usec);
1673
1674                 /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */
1675                 r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE);
1676                 if (r < 0)
1677                         log_error_errno(r, "Failed to switch root, trying to continue: %m");
1678         }
1679
1680         args_size = MAX(6, argc+1);
1681         args = newa(const char*, args_size);
1682
1683         if (!switch_root_init) {
1684                 char sfd[DECIMAL_STR_MAX(int) + 1];
1685
1686                 /* First try to spawn ourselves with the right path, and with full serialization. We do this only if
1687                  * the user didn't specify an explicit init to spawn. */
1688
1689                 assert(arg_serialization);
1690                 assert(fds);
1691
1692                 xsprintf(sfd, "%i", fileno(arg_serialization));
1693
1694                 i = 0;
1695                 args[i++] = SYSTEMD_BINARY_PATH;
1696                 if (switch_root_dir)
1697                         args[i++] = "--switched-root";
1698                 args[i++] = arg_system ? "--system" : "--user";
1699                 args[i++] = "--deserialize";
1700                 args[i++] = sfd;
1701                 args[i++] = NULL;
1702
1703                 assert(i <= args_size);
1704
1705                 /*
1706                  * We want valgrind to print its memory usage summary before reexecution.  Valgrind won't do this is on
1707                  * its own on exec(), but it will do it on exit().  Hence, to ensure we get a summary here, fork() off
1708                  * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent,
1709                  * before proceeding into the exec().
1710                  */
1711                 valgrind_summary_hack();
1712
1713                 (void) execv(args[0], (char* const*) args);
1714                 log_debug_errno(errno, "Failed to execute our own binary, trying fallback: %m");
1715         }
1716
1717         /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well,
1718          * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that
1719          * doesn't matter.) */
1720
1721         arg_serialization = safe_fclose(arg_serialization);
1722         fds = fdset_free(fds);
1723
1724         /* Reopen the console */
1725         (void) make_console_stdio();
1726
1727         for (j = 1, i = 1; j < (unsigned) argc; j++)
1728                 args[i++] = argv[j];
1729         args[i++] = NULL;
1730         assert(i <= args_size);
1731
1732         /* Reenable any blocked signals, especially important if we switch from initial ramdisk to init=... */
1733         (void) reset_all_signal_handlers();
1734         (void) reset_signal_mask();
1735
1736         if (switch_root_init) {
1737                 args[0] = switch_root_init;
1738                 (void) execv(args[0], (char* const*) args);
1739                 log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m");
1740         }
1741
1742         args[0] = "/sbin/init";
1743         (void) execv(args[0], (char* const*) args);
1744         r = -errno;
1745
1746         manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
1747                               ANSI_HIGHLIGHT_RED "  !!  " ANSI_NORMAL,
1748                               "Failed to execute /sbin/init");
1749
1750         if (r == -ENOENT) {
1751                 log_warning("No /sbin/init, trying fallback");
1752
1753                 args[0] = "/bin/sh";
1754                 args[1] = NULL;
1755                 (void) execv(args[0], (char* const*) args);
1756                 log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m");
1757         } else
1758                 log_warning_errno(r, "Failed to execute /sbin/init, giving up: %m");
1759
1760         *ret_error_message = "Failed to execute fallback shell";
1761 }
1762
1763 static int invoke_main_loop(
1764                 Manager *m,
1765                 bool *ret_reexecute,
1766                 int *ret_retval,                   /* Return parameters relevant for shutting down */
1767                 const char **ret_shutdown_verb,    /* … */
1768                 FDSet **ret_fds,                   /* Return parameters for reexecuting */
1769                 char **ret_switch_root_dir,        /* … */
1770                 char **ret_switch_root_init,       /* … */
1771                 const char **ret_error_message) {
1772
1773         int r;
1774
1775         assert(m);
1776         assert(ret_reexecute);
1777         assert(ret_retval);
1778         assert(ret_shutdown_verb);
1779         assert(ret_fds);
1780         assert(ret_switch_root_dir);
1781         assert(ret_switch_root_init);
1782         assert(ret_error_message);
1783
1784         for (;;) {
1785                 r = manager_loop(m);
1786                 if (r < 0) {
1787                         *ret_error_message = "Failed to run main loop";
1788                         return log_emergency_errno(r, "Failed to run main loop: %m");
1789                 }
1790
1791                 switch ((ManagerObjective) r) {
1792
1793                 case MANAGER_RELOAD: {
1794                         LogTarget saved_log_target;
1795                         int saved_log_level;
1796
1797                         log_info("Reloading.");
1798
1799                         /* First, save any overridden log level/target, then parse the configuration file, which might
1800                          * change the log level to new settings. */
1801
1802                         saved_log_level = m->log_level_overridden ? log_get_max_level() : -1;
1803                         saved_log_target = m->log_target_overridden ? log_get_target() : _LOG_TARGET_INVALID;
1804
1805                         r = parse_config_file();
1806                         if (r < 0)
1807                                 log_warning_errno(r, "Failed to parse config file, ignoring: %m");
1808
1809                         set_manager_defaults(m);
1810
1811                         if (saved_log_level >= 0)
1812                                 manager_override_log_level(m, saved_log_level);
1813                         if (saved_log_target >= 0)
1814                                 manager_override_log_target(m, saved_log_target);
1815
1816                         r = manager_reload(m);
1817                         if (r < 0)
1818                                 /* Reloading failed before the point of no return. Let's continue running as if nothing happened. */
1819                                 m->objective = MANAGER_OK;
1820
1821                         break;
1822                 }
1823
1824                 case MANAGER_REEXECUTE:
1825
1826                         r = prepare_reexecute(m, &arg_serialization, ret_fds, false);
1827                         if (r < 0) {
1828                                 *ret_error_message = "Failed to prepare for reexecution";
1829                                 return r;
1830                         }
1831
1832                         log_notice("Reexecuting.");
1833
1834                         *ret_reexecute = true;
1835                         *ret_retval = EXIT_SUCCESS;
1836                         *ret_shutdown_verb = NULL;
1837                         *ret_switch_root_dir = *ret_switch_root_init = NULL;
1838
1839                         return 0;
1840
1841                 case MANAGER_SWITCH_ROOT:
1842                         if (!m->switch_root_init) {
1843                                 r = prepare_reexecute(m, &arg_serialization, ret_fds, true);
1844                                 if (r < 0) {
1845                                         *ret_error_message = "Failed to prepare for reexecution";
1846                                         return r;
1847                                 }
1848                         } else
1849                                 *ret_fds = NULL;
1850
1851                         log_notice("Switching root.");
1852
1853                         *ret_reexecute = true;
1854                         *ret_retval = EXIT_SUCCESS;
1855                         *ret_shutdown_verb = NULL;
1856
1857                         /* Steal the switch root parameters */
1858                         *ret_switch_root_dir = m->switch_root;
1859                         *ret_switch_root_init = m->switch_root_init;
1860                         m->switch_root = m->switch_root_init = NULL;
1861
1862                         return 0;
1863
1864                 case MANAGER_EXIT:
1865
1866                         if (MANAGER_IS_USER(m)) {
1867                                 log_debug("Exit.");
1868
1869                                 *ret_reexecute = false;
1870                                 *ret_retval = m->return_value;
1871                                 *ret_shutdown_verb = NULL;
1872                                 *ret_fds = NULL;
1873                                 *ret_switch_root_dir = *ret_switch_root_init = NULL;
1874
1875                                 return 0;
1876                         }
1877
1878                         _fallthrough_;
1879                 case MANAGER_REBOOT:
1880                 case MANAGER_POWEROFF:
1881                 case MANAGER_HALT:
1882                 case MANAGER_KEXEC: {
1883                         static const char * const table[_MANAGER_OBJECTIVE_MAX] = {
1884                                 [MANAGER_EXIT]     = "exit",
1885                                 [MANAGER_REBOOT]   = "reboot",
1886                                 [MANAGER_POWEROFF] = "poweroff",
1887                                 [MANAGER_HALT]     = "halt",
1888                                 [MANAGER_KEXEC]    = "kexec",
1889                         };
1890
1891                         log_notice("Shutting down.");
1892
1893                         *ret_reexecute = false;
1894                         *ret_retval = m->return_value;
1895                         assert_se(*ret_shutdown_verb = table[m->objective]);
1896                         *ret_fds = NULL;
1897                         *ret_switch_root_dir = *ret_switch_root_init = NULL;
1898
1899                         return 0;
1900                 }
1901
1902                 default:
1903                         assert_not_reached("Unknown or unexpected manager objective.");
1904                 }
1905         }
1906 }
1907
1908 static void log_execution_mode(bool *ret_first_boot) {
1909         assert(ret_first_boot);
1910
1911         if (arg_system) {
1912                 int v;
1913
1914                 log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")",
1915                          arg_action == ACTION_TEST ? "test " : "" );
1916
1917                 v = detect_virtualization();
1918                 if (v > 0)
1919                         log_info("Detected virtualization %s.", virtualization_to_string(v));
1920
1921                 log_info("Detected architecture %s.", architecture_to_string(uname_architecture()));
1922
1923                 if (in_initrd()) {
1924                         *ret_first_boot = false;
1925                         log_info("Running in initial RAM disk.");
1926                 } else {
1927                         /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use
1928                          * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it
1929                          * doesn't it's unpopulated. This allows container managers and installers to provision a
1930                          * couple of files already. If the container manager wants to provision the machine ID itself
1931                          * it should pass $container_uuid to PID 1. */
1932
1933                         *ret_first_boot = access("/etc/machine-id", F_OK) < 0;
1934                         if (*ret_first_boot)
1935                                 log_info("Running with unpopulated /etc.");
1936                 }
1937         } else {
1938                 if (DEBUG_LOGGING) {
1939                         _cleanup_free_ char *t;
1940
1941                         t = uid_to_name(getuid());
1942                         log_debug(PACKAGE_STRING " running in %suser mode for user " UID_FMT "/%s. (" SYSTEMD_FEATURES ")",
1943                                   arg_action == ACTION_TEST ? " test" : "", getuid(), strna(t));
1944                 }
1945
1946                 *ret_first_boot = false;
1947         }
1948 }
1949
1950 static int initialize_runtime(
1951                 bool skip_setup,
1952                 struct rlimit *saved_rlimit_nofile,
1953                 struct rlimit *saved_rlimit_memlock,
1954                 const char **ret_error_message) {
1955
1956         int r;
1957
1958         assert(ret_error_message);
1959
1960         /* Sets up various runtime parameters. Many of these initializations are conditionalized:
1961          *
1962          * - Some only apply to --system instances
1963          * - Some only apply to --user instances
1964          * - Some only apply when we first start up, but not when we reexecute
1965          */
1966
1967         if (arg_action != ACTION_RUN)
1968                 return 0;
1969
1970         if (arg_system) {
1971                 /* Make sure we leave a core dump without panicing the kernel. */
1972                 install_crash_handler();
1973
1974                 if (!skip_setup) {
1975                         r = mount_cgroup_controllers();
1976                         if (r < 0) {
1977                                 *ret_error_message = "Failed to mount cgroup hierarchies";
1978                                 return r;
1979                         }
1980
1981                         status_welcome();
1982                         hostname_setup();
1983                         machine_id_setup(NULL, arg_machine_id, NULL);
1984                         loopback_setup();
1985                         bump_unix_max_dgram_qlen();
1986                         bump_file_max_and_nr_open();
1987                         test_usr();
1988                         write_container_id();
1989                 }
1990
1991                 if (arg_watchdog_device) {
1992                         r = watchdog_set_device(arg_watchdog_device);
1993                         if (r < 0)
1994                                 log_warning_errno(r, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device);
1995                 }
1996
1997                 if (arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY)
1998                         watchdog_set_timeout(&arg_runtime_watchdog);
1999         }
2000
2001         if (arg_timer_slack_nsec != NSEC_INFINITY)
2002                 if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0)
2003                         log_warning_errno(errno, "Failed to adjust timer slack, ignoring: %m");
2004
2005         if (arg_system && !cap_test_all(arg_capability_bounding_set)) {
2006                 r = capability_bounding_set_drop_usermode(arg_capability_bounding_set);
2007                 if (r < 0) {
2008                         *ret_error_message = "Failed to drop capability bounding set of usermode helpers";
2009                         return log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m");
2010                 }
2011
2012                 r = capability_bounding_set_drop(arg_capability_bounding_set, true);
2013                 if (r < 0) {
2014                         *ret_error_message = "Failed to drop capability bounding set";
2015                         return log_emergency_errno(r, "Failed to drop capability bounding set: %m");
2016                 }
2017         }
2018
2019         if (arg_system && arg_no_new_privs) {
2020                 if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) {
2021                         *ret_error_message = "Failed to disable new privileges";
2022                         return log_emergency_errno(errno, "Failed to disable new privileges: %m");
2023                 }
2024         }
2025
2026         if (arg_syscall_archs) {
2027                 r = enforce_syscall_archs(arg_syscall_archs);
2028                 if (r < 0) {
2029                         *ret_error_message = "Failed to set syscall architectures";
2030                         return r;
2031                 }
2032         }
2033
2034         if (!arg_system)
2035                 /* Become reaper of our children */
2036                 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0)
2037                         log_warning_errno(errno, "Failed to make us a subreaper: %m");
2038
2039         /* Bump up RLIMIT_NOFILE for systemd itself */
2040         (void) bump_rlimit_nofile(saved_rlimit_nofile);
2041         (void) bump_rlimit_memlock(saved_rlimit_memlock);
2042
2043         return 0;
2044 }
2045
2046 static int do_queue_default_job(
2047                 Manager *m,
2048                 const char **ret_error_message) {
2049
2050         _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
2051         Job *default_unit_job;
2052         Unit *target = NULL;
2053         int r;
2054
2055         log_debug("Activating default unit: %s", arg_default_unit);
2056
2057         r = manager_load_startable_unit_or_warn(m, arg_default_unit, NULL, &target);
2058         if (r < 0) {
2059                 log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET);
2060
2061                 r = manager_load_startable_unit_or_warn(m, SPECIAL_RESCUE_TARGET, NULL, &target);
2062                 if (r < 0) {
2063                         *ret_error_message = r == -ERFKILL ? "Rescue target masked"
2064                                                            : "Failed to load rescue target";
2065                         return r;
2066                 }
2067         }
2068
2069         assert(target->load_state == UNIT_LOADED);
2070
2071         r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, &error, &default_unit_job);
2072         if (r == -EPERM) {
2073                 log_debug_errno(r, "Default target could not be isolated, starting instead: %s", bus_error_message(&error, r));
2074
2075                 sd_bus_error_free(&error);
2076
2077                 r = manager_add_job(m, JOB_START, target, JOB_REPLACE, &error, &default_unit_job);
2078                 if (r < 0) {
2079                         *ret_error_message = "Failed to start default target";
2080                         return log_emergency_errno(r, "Failed to start default target: %s", bus_error_message(&error, r));
2081                 }
2082
2083         } else if (r < 0) {
2084                 *ret_error_message = "Failed to isolate default target";
2085                 return log_emergency_errno(r, "Failed to isolate default target: %s", bus_error_message(&error, r));
2086         }
2087
2088         m->default_unit_job_id = default_unit_job->id;
2089
2090         return 0;
2091 }
2092
2093 static void free_arguments(void) {
2094
2095         /* Frees all arg_* variables, with the exception of arg_serialization */
2096         rlimit_free_all(arg_default_rlimit);
2097
2098         arg_default_unit = mfree(arg_default_unit);
2099         arg_confirm_spawn = mfree(arg_confirm_spawn);
2100         arg_default_environment = strv_free(arg_default_environment);
2101         arg_syscall_archs = set_free(arg_syscall_archs);
2102 }
2103
2104 static int load_configuration(int argc, char **argv, const char **ret_error_message) {
2105         int r;
2106
2107         assert(ret_error_message);
2108
2109         arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U);
2110
2111         r = parse_config_file();
2112         if (r < 0) {
2113                 *ret_error_message = "Failed to parse config file";
2114                 return r;
2115         }
2116
2117         if (arg_system) {
2118                 r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
2119                 if (r < 0)
2120                         log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
2121         }
2122
2123         /* Note that this also parses bits from the kernel command line, including "debug". */
2124         log_parse_environment();
2125
2126         r = parse_argv(argc, argv);
2127         if (r < 0) {
2128                 *ret_error_message = "Failed to parse commandline arguments";
2129                 return r;
2130         }
2131
2132         /* Initialize default unit */
2133         if (!arg_default_unit) {
2134                 arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET);
2135                 if (!arg_default_unit) {
2136                         *ret_error_message = "Failed to set default unit";
2137                         return log_oom();
2138                 }
2139         }
2140
2141         /* Initialize the show status setting if it hasn't been set explicitly yet */
2142         if (arg_show_status == _SHOW_STATUS_INVALID)
2143                 arg_show_status = SHOW_STATUS_YES;
2144
2145         return 0;
2146 }
2147
2148 static int safety_checks(void) {
2149
2150         if (getpid_cached() == 1 &&
2151             arg_action != ACTION_RUN) {
2152                 log_error("Unsupported execution mode while PID 1.");
2153                 return -EPERM;
2154         }
2155
2156         if (getpid_cached() == 1 &&
2157             !arg_system) {
2158                 log_error("Can't run --user mode as PID 1.");
2159                 return -EPERM;
2160         }
2161
2162         if (arg_action == ACTION_RUN &&
2163             arg_system &&
2164             getpid_cached() != 1) {
2165                 log_error("Can't run system mode unless PID 1.");
2166                 return -EPERM;
2167         }
2168
2169         if (arg_action == ACTION_TEST &&
2170             geteuid() == 0) {
2171                 log_error("Don't run test mode as root.");
2172                 return -EPERM;
2173         }
2174
2175         if (!arg_system &&
2176             arg_action == ACTION_RUN &&
2177             sd_booted() <= 0) {
2178                 log_error("Trying to run as user instance, but the system has not been booted with systemd.");
2179                 return -EOPNOTSUPP;
2180         }
2181
2182         if (!arg_system &&
2183             arg_action == ACTION_RUN &&
2184             !getenv("XDG_RUNTIME_DIR")) {
2185                 log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set.");
2186                 return -EUNATCH;
2187         }
2188
2189         if (arg_system &&
2190             arg_action == ACTION_RUN &&
2191             running_in_chroot() > 0) {
2192                 log_error("Cannot be run in a chroot() environment.");
2193                 return -EOPNOTSUPP;
2194         }
2195
2196         return 0;
2197 }
2198
2199 static int initialize_security(
2200                 bool *loaded_policy,
2201                 dual_timestamp *security_start_timestamp,
2202                 dual_timestamp *security_finish_timestamp,
2203                 const char **ret_error_message) {
2204
2205         int r;
2206
2207         assert(loaded_policy);
2208         assert(security_start_timestamp);
2209         assert(security_finish_timestamp);
2210         assert(ret_error_message);
2211
2212         dual_timestamp_get(security_start_timestamp);
2213
2214         r = mac_selinux_setup(loaded_policy);
2215         if (r < 0) {
2216                 *ret_error_message = "Failed to load SELinux policy";
2217                 return r;
2218         }
2219
2220         r = mac_smack_setup(loaded_policy);
2221         if (r < 0) {
2222                 *ret_error_message = "Failed to load SMACK policy";
2223                 return r;
2224         }
2225
2226         r = ima_setup();
2227         if (r < 0) {
2228                 *ret_error_message = "Failed to load IMA policy";
2229                 return r;
2230         }
2231
2232         dual_timestamp_get(security_finish_timestamp);
2233         return 0;
2234 }
2235
2236 static void test_summary(Manager *m) {
2237         assert(m);
2238
2239         printf("-> By units:\n");
2240         manager_dump_units(m, stdout, "\t");
2241
2242         printf("-> By jobs:\n");
2243         manager_dump_jobs(m, stdout, "\t");
2244 }
2245
2246 static int collect_fds(FDSet **ret_fds, const char **ret_error_message) {
2247         int r;
2248
2249         assert(ret_fds);
2250         assert(ret_error_message);
2251
2252         r = fdset_new_fill(ret_fds);
2253         if (r < 0) {
2254                 *ret_error_message = "Failed to allocate fd set";
2255                 return log_emergency_errno(r, "Failed to allocate fd set: %m");
2256         }
2257
2258         fdset_cloexec(*ret_fds, true);
2259
2260         if (arg_serialization)
2261                 assert_se(fdset_remove(*ret_fds, fileno(arg_serialization)) >= 0);
2262
2263         return 0;
2264 }
2265
2266 static void setup_console_terminal(bool skip_setup) {
2267
2268         if (!arg_system)
2269                 return;
2270
2271         /* Become a session leader if we aren't one yet. */
2272         (void) setsid();
2273
2274         /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling
2275          * tty. */
2276         (void) release_terminal();
2277
2278         /* Reset the console, but only if this is really init and we are freshly booted */
2279         if (getpid_cached() == 1 && !skip_setup)
2280                 (void) console_setup();
2281 }
2282
2283 static bool early_skip_setup_check(int argc, char *argv[]) {
2284         bool found_deserialize = false;
2285         int i;
2286
2287         /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so
2288          * let's just have a quick peek here. Note that if we have switched root, do all the special setup things
2289          * anyway, even if in that case we also do deserialization. */
2290
2291         for (i = 1; i < argc; i++) {
2292                 if (streq(argv[i], "--switched-root"))
2293                         return false; /* If we switched root, don't skip the setup. */
2294                 else if (streq(argv[i], "--deserialize"))
2295                         found_deserialize = true;
2296         }
2297
2298         return found_deserialize; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */
2299 }
2300
2301 int main(int argc, char *argv[]) {
2302
2303         dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL, userspace_timestamp = DUAL_TIMESTAMP_NULL, kernel_timestamp = DUAL_TIMESTAMP_NULL,
2304                 security_start_timestamp = DUAL_TIMESTAMP_NULL, security_finish_timestamp = DUAL_TIMESTAMP_NULL;
2305         struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0), saved_rlimit_memlock = RLIMIT_MAKE_CONST((rlim_t) -1);
2306         bool skip_setup, loaded_policy = false, queue_default_job = false, first_boot = false, reexecute = false;
2307         char *switch_root_dir = NULL, *switch_root_init = NULL;
2308         usec_t before_startup, after_startup;
2309         static char systemd[] = "systemd";
2310         char timespan[FORMAT_TIMESPAN_MAX];
2311         const char *shutdown_verb = NULL, *error_message = NULL;
2312         int r, retval = EXIT_FAILURE;
2313         Manager *m = NULL;
2314         FDSet *fds = NULL;
2315
2316         /* SysV compatibility: redirect init → telinit */
2317         redirect_telinit(argc, argv);
2318
2319         /* Take timestamps early on */
2320         dual_timestamp_from_monotonic(&kernel_timestamp, 0);
2321         dual_timestamp_get(&userspace_timestamp);
2322
2323         /* Figure out whether we need to do initialize the system, or if we already did that because we are
2324          * reexecuting */
2325         skip_setup = early_skip_setup_check(argc, argv);
2326
2327         /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we
2328          * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */
2329         program_invocation_short_name = systemd;
2330         (void) prctl(PR_SET_NAME, systemd);
2331
2332         /* Save the original command line */
2333         saved_argv = argv;
2334         saved_argc = argc;
2335
2336         /* Make sure that if the user says "syslog" we actually log to the journal. */
2337         log_set_upgrade_syslog_to_journal(true);
2338
2339         if (getpid_cached() == 1) {
2340                 /* When we run as PID 1 force system mode */
2341                 arg_system = true;
2342
2343                 /* Disable the umask logic */
2344                 umask(0);
2345
2346                 /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be
2347                  * activated yet (even though the log socket for it exists). */
2348                 log_set_prohibit_ipc(true);
2349
2350                 /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is
2351                  * important so that we never end up logging to any foreign stderr, for example if we have to log in a
2352                  * child process right before execve()'ing the actual binary, at a point in time where socket
2353                  * activation stderr/stdout area already set up. */
2354                 log_set_always_reopen_console(true);
2355
2356                 if (detect_container() <= 0) {
2357
2358                         /* Running outside of a container as PID 1 */
2359                         log_set_target(LOG_TARGET_KMSG);
2360                         log_open();
2361
2362                         if (in_initrd())
2363                                 initrd_timestamp = userspace_timestamp;
2364
2365                         if (!skip_setup) {
2366                                 r = mount_setup_early();
2367                                 if (r < 0) {
2368                                         error_message = "Failed to mount early API filesystems";
2369                                         goto finish;
2370                                 }
2371
2372                                 r = initialize_security(
2373                                                 &loaded_policy,
2374                                                 &security_start_timestamp,
2375                                                 &security_finish_timestamp,
2376                                                 &error_message);
2377                                 if (r < 0)
2378                                         goto finish;
2379                         }
2380
2381                         if (mac_selinux_init() < 0) {
2382                                 error_message = "Failed to initialize SELinux policy";
2383                                 goto finish;
2384                         }
2385
2386                         if (!skip_setup)
2387                                 initialize_clock();
2388
2389                         /* Set the default for later on, but don't actually open the logs like this for now. Note that
2390                          * if we are transitioning from the initrd there might still be journal fd open, and we
2391                          * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output
2392                          * elsewhere. */
2393                         log_set_target(LOG_TARGET_JOURNAL_OR_KMSG);
2394
2395                 } else {
2396                         /* Running inside a container, as PID 1 */
2397                         log_set_target(LOG_TARGET_CONSOLE);
2398                         log_open();
2399
2400                         /* For later on, see above... */
2401                         log_set_target(LOG_TARGET_JOURNAL);
2402
2403                         /* clear the kernel timestamp,
2404                          * because we are in a container */
2405                         kernel_timestamp = DUAL_TIMESTAMP_NULL;
2406                 }
2407
2408                 initialize_coredump(skip_setup);
2409
2410                 r = fixup_environment();
2411                 if (r < 0) {
2412                         log_emergency_errno(r, "Failed to fix up PID 1 environment: %m");
2413                         error_message = "Failed to fix up PID1 environment";
2414                         goto finish;
2415                 }
2416
2417         } else {
2418                 /* Running as user instance */
2419                 arg_system = false;
2420                 log_set_target(LOG_TARGET_AUTO);
2421                 log_open();
2422
2423                 /* clear the kernel timestamp,
2424                  * because we are not PID 1 */
2425                 kernel_timestamp = DUAL_TIMESTAMP_NULL;
2426         }
2427
2428         if (arg_system) {
2429                 /* Try to figure out if we can use colors with the console. No need to do that for user instances since
2430                  * they never log into the console. */
2431                 log_show_color(colors_enabled());
2432
2433                 r = make_null_stdio();
2434                 if (r < 0)
2435                         log_warning_errno(r, "Failed to redirect standard streams to /dev/null, ignoring: %m");
2436         }
2437
2438         /* Mount /proc, /sys and friends, so that /proc/cmdline and
2439          * /proc/$PID/fd is available. */
2440         if (getpid_cached() == 1) {
2441
2442                 /* Load the kernel modules early. */
2443                 if (!skip_setup)
2444                         kmod_setup();
2445
2446                 r = mount_setup(loaded_policy);
2447                 if (r < 0) {
2448                         error_message = "Failed to mount API filesystems";
2449                         goto finish;
2450                 }
2451         }
2452
2453         /* Reset all signal handlers. */
2454         (void) reset_all_signal_handlers();
2455         (void) ignore_signals(SIGNALS_IGNORE, -1);
2456
2457         r = load_configuration(argc, argv, &error_message);
2458         if (r < 0)
2459                 goto finish;
2460
2461         r = safety_checks();
2462         if (r < 0)
2463                 goto finish;
2464
2465         if (IN_SET(arg_action, ACTION_TEST, ACTION_HELP, ACTION_DUMP_CONFIGURATION_ITEMS, ACTION_DUMP_BUS_PROPERTIES))
2466                 (void) pager_open(arg_pager_flags);
2467
2468         if (arg_action != ACTION_RUN)
2469                 skip_setup = true;
2470
2471         if (arg_action == ACTION_HELP) {
2472                 retval = help() < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
2473                 goto finish;
2474         } else if (arg_action == ACTION_VERSION) {
2475                 retval = version();
2476                 goto finish;
2477         } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) {
2478                 unit_dump_config_items(stdout);
2479                 retval = EXIT_SUCCESS;
2480                 goto finish;
2481         } else if (arg_action == ACTION_DUMP_BUS_PROPERTIES) {
2482                 dump_bus_properties(stdout);
2483                 retval = EXIT_SUCCESS;
2484                 goto finish;
2485         }
2486
2487         assert_se(IN_SET(arg_action, ACTION_RUN, ACTION_TEST));
2488
2489         /* Move out of the way, so that we won't block unmounts */
2490         assert_se(chdir("/") == 0);
2491
2492         if (arg_action == ACTION_RUN) {
2493
2494                 /* A core pattern might have been specified via the cmdline.  */
2495                 initialize_core_pattern(skip_setup);
2496
2497                 /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */
2498                 log_close();
2499
2500                 /* Remember open file descriptors for later deserialization */
2501                 r = collect_fds(&fds, &error_message);
2502                 if (r < 0)
2503                         goto finish;
2504
2505                 /* Give up any control of the console, but make sure its initialized. */
2506                 setup_console_terminal(skip_setup);
2507
2508                 /* Open the logging devices, if possible and necessary */
2509                 log_open();
2510         }
2511
2512         log_execution_mode(&first_boot);
2513
2514         r = initialize_runtime(skip_setup,
2515                                &saved_rlimit_nofile,
2516                                &saved_rlimit_memlock,
2517                                &error_message);
2518         if (r < 0)
2519                 goto finish;
2520
2521         r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER,
2522                         arg_action == ACTION_TEST ? MANAGER_TEST_FULL : 0,
2523                         &m);
2524         if (r < 0) {
2525                 log_emergency_errno(r, "Failed to allocate manager object: %m");
2526                 error_message = "Failed to allocate manager object";
2527                 goto finish;
2528         }
2529
2530         m->timestamps[MANAGER_TIMESTAMP_KERNEL] = kernel_timestamp;
2531         m->timestamps[MANAGER_TIMESTAMP_INITRD] = initrd_timestamp;
2532         m->timestamps[MANAGER_TIMESTAMP_USERSPACE] = userspace_timestamp;
2533         m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START)] = security_start_timestamp;
2534         m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH)] = security_finish_timestamp;
2535
2536         set_manager_defaults(m);
2537         set_manager_settings(m);
2538         manager_set_first_boot(m, first_boot);
2539
2540         /* Remember whether we should queue the default job */
2541         queue_default_job = !arg_serialization || arg_switched_root;
2542
2543         before_startup = now(CLOCK_MONOTONIC);
2544
2545         r = manager_startup(m, arg_serialization, fds);
2546         if (r < 0) {
2547                 error_message = "Failed to start up manager";
2548                 goto finish;
2549         }
2550
2551         /* This will close all file descriptors that were opened, but not claimed by any unit. */
2552         fds = fdset_free(fds);
2553         arg_serialization = safe_fclose(arg_serialization);
2554
2555         if (queue_default_job) {
2556                 r = do_queue_default_job(m, &error_message);
2557                 if (r < 0)
2558                         goto finish;
2559         }
2560
2561         after_startup = now(CLOCK_MONOTONIC);
2562
2563         log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG,
2564                  "Loaded units and determined initial transaction in %s.",
2565                  format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC));
2566
2567         if (arg_action == ACTION_TEST) {
2568                 test_summary(m);
2569                 retval = EXIT_SUCCESS;
2570                 goto finish;
2571         }
2572
2573         (void) invoke_main_loop(m,
2574                                 &reexecute,
2575                                 &retval,
2576                                 &shutdown_verb,
2577                                 &fds,
2578                                 &switch_root_dir,
2579                                 &switch_root_init,
2580                                 &error_message);
2581
2582 finish:
2583         pager_close();
2584
2585         if (m) {
2586                 arg_shutdown_watchdog = m->shutdown_watchdog;
2587                 m = manager_free(m);
2588         }
2589
2590         free_arguments();
2591         mac_selinux_finish();
2592
2593         if (reexecute)
2594                 do_reexecute(argc, argv,
2595                              &saved_rlimit_nofile,
2596                              &saved_rlimit_memlock,
2597                              fds,
2598                              switch_root_dir,
2599                              switch_root_init,
2600                              &error_message); /* This only returns if reexecution failed */
2601
2602         arg_serialization = safe_fclose(arg_serialization);
2603         fds = fdset_free(fds);
2604
2605 #if HAVE_VALGRIND_VALGRIND_H
2606         /* If we are PID 1 and running under valgrind, then let's exit
2607          * here explicitly. valgrind will only generate nice output on
2608          * exit(), not on exec(), hence let's do the former not the
2609          * latter here. */
2610         if (getpid_cached() == 1 && RUNNING_ON_VALGRIND) {
2611                 /* Cleanup watchdog_device strings for valgrind. We need them
2612                  * in become_shutdown() so normally we cannot free them yet. */
2613                 watchdog_free_device();
2614                 arg_watchdog_device = mfree(arg_watchdog_device);
2615                 return retval;
2616         }
2617 #endif
2618
2619         if (shutdown_verb) {
2620                 r = become_shutdown(shutdown_verb, retval);
2621                 log_error_errno(r, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting");
2622                 error_message = "Failed to execute shutdown binary";
2623         }
2624
2625         watchdog_free_device();
2626         arg_watchdog_device = mfree(arg_watchdog_device);
2627
2628         if (getpid_cached() == 1) {
2629                 if (error_message)
2630                         manager_status_printf(NULL, STATUS_TYPE_EMERGENCY,
2631                                               ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL,
2632                                               "%s.", error_message);
2633                 freeze_or_exit_or_reboot();
2634         }
2635
2636         return retval;
2637 }