init/main.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  *  linux/init/main.c
   4  *
   5  *  Copyright (C) 1991, 1992  Linus Torvalds
   6  *
   7  *  GK 2/5/95  -  Changed to support mounting root fs via NFS
   8  *  Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
   9  *  Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
  10  *  Simplified starting of init:  Michael A. Griffith <grif@acm.org>
  11  */
  12
  13 #define DEBUG           /* Enable initcall_debug */
  14
  15 #include <linux/types.h>
  16 #include <linux/extable.h>
  17 #include <linux/module.h>
  18 #include <linux/proc_fs.h>
  19 #include <linux/binfmts.h>
  20 #include <linux/kernel.h>
  21 #include <linux/syscalls.h>
  22 #include <linux/stackprotector.h>
  23 #include <linux/string.h>
  24 #include <linux/ctype.h>
  25 #include <linux/delay.h>
  26 #include <linux/ioport.h>
  27 #include <linux/init.h>
  28 #include <linux/initrd.h>
  29 #include <linux/memblock.h>
  30 #include <linux/acpi.h>
  31 #include <linux/bootconfig.h>
  32 #include <linux/console.h>
  33 #include <linux/nmi.h>
  34 #include <linux/percpu.h>
  35 #include <linux/kmod.h>
  36 #include <linux/kprobes.h>
  37 #include <linux/kmsan.h>
  38 #include <linux/vmalloc.h>
  39 #include <linux/kernel_stat.h>
  40 #include <linux/start_kernel.h>
  41 #include <linux/security.h>
  42 #include <linux/smp.h>
  43 #include <linux/profile.h>
  44 #include <linux/kfence.h>
  45 #include <linux/rcupdate.h>
  46 #include <linux/srcu.h>
  47 #include <linux/moduleparam.h>
  48 #include <linux/kallsyms.h>
  49 #include <linux/buildid.h>
  50 #include <linux/writeback.h>
  51 #include <linux/cpu.h>
  52 #include <linux/cpuset.h>
  53 #include <linux/cgroup.h>
  54 #include <linux/efi.h>
  55 #include <linux/tick.h>
  56 #include <linux/sched/isolation.h>
  57 #include <linux/interrupt.h>
  58 #include <linux/taskstats_kern.h>
  59 #include <linux/delayacct.h>
  60 #include <linux/unistd.h>
  61 #include <linux/utsname.h>
  62 #include <linux/rmap.h>
  63 #include <linux/mempolicy.h>
  64 #include <linux/key.h>
  65 #include <linux/debug_locks.h>
  66 #include <linux/debugobjects.h>
  67 #include <linux/lockdep.h>
  68 #include <linux/kmemleak.h>
  69 #include <linux/padata.h>
  70 #include <linux/pid_namespace.h>
  71 #include <linux/device/driver.h>
  72 #include <linux/kthread.h>
  73 #include <linux/sched.h>
  74 #include <linux/sched/init.h>
  75 #include <linux/signal.h>
  76 #include <linux/idr.h>
  77 #include <linux/kgdb.h>
  78 #include <linux/ftrace.h>
  79 #include <linux/async.h>
  80 #include <linux/shmem_fs.h>
  81 #include <linux/slab.h>
  82 #include <linux/perf_event.h>
  83 #include <linux/ptrace.h>
  84 #include <linux/pti.h>
  85 #include <linux/blkdev.h>
  86 #include <linux/sched/clock.h>
  87 #include <linux/sched/task.h>
  88 #include <linux/sched/task_stack.h>
  89 #include <linux/context_tracking.h>
  90 #include <linux/random.h>
  91 #include <linux/moduleloader.h>
  92 #include <linux/list.h>
  93 #include <linux/integrity.h>
  94 #include <linux/proc_ns.h>
  95 #include <linux/io.h>
  96 #include <linux/cache.h>
  97 #include <linux/rodata_test.h>
  98 #include <linux/jump_label.h>
  99 #include <linux/kcsan.h>
 100 #include <linux/init_syscalls.h>
 101 #include <linux/stackdepot.h>
 102 #include <linux/randomize_kstack.h>
 103 #include <linux/pidfs.h>
 104 #include <linux/ptdump.h>
 105 #include <net/net_namespace.h>
 106
 107 #include <asm/io.h>
 108 #include <asm/setup.h>
 109 #include <asm/sections.h>
 110 #include <asm/cacheflush.h>
 111
 112 #define CREATE_TRACE_POINTS
 113 #include <trace/events/initcall.h>
 114
 115 #include <kunit/test.h>
 116
 117 static int kernel_init(void *);
 118
 119 /*
 120  * Debug helper: via this flag we know that we are in 'early bootup code'
 121  * where only the boot processor is running with IRQ disabled.  This means
 122  * two things - IRQ must not be enabled before the flag is cleared and some
 123  * operations which are not allowed with IRQ disabled are allowed while the
 124  * flag is set.
 125  */
 126 bool early_boot_irqs_disabled __read_mostly;
 127
 128 enum system_states system_state __read_mostly;
 129 EXPORT_SYMBOL(system_state);
 130
 131 /*
 132  * Boot command-line arguments
 133  */
 134 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 135 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 136
 137 /* Default late time init is NULL. archs can override this later. */
 138 void (*__initdata late_time_init)(void);
 139
 140 /* Untouched command line saved by arch-specific code. */
 141 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 142 /* Untouched saved command line (eg. for /proc) */
 143 char *saved_command_line __ro_after_init;
 144 unsigned int saved_command_line_len __ro_after_init;
 145 /* Command line for parameter parsing */
 146 static char *static_command_line;
 147 /* Untouched extra command line */
 148 static char *extra_command_line;
 149 /* Extra init arguments */
 150 static char *extra_init_args;
 151
 152 #ifdef CONFIG_BOOT_CONFIG
 153 /* Is bootconfig on command line? */
 154 static bool bootconfig_found;
 155 static size_t initargs_offs;
 156 #else
 157 # define bootconfig_found false
 158 # define initargs_offs 0
 159 #endif
 160
 161 static char *execute_command;
 162 static char *ramdisk_execute_command = "/init";
 163
 164 /*
 165  * Used to generate warnings if static_key manipulation functions are used
 166  * before jump_label_init is called.
 167  */
 168 bool static_key_initialized __read_mostly;
 169 EXPORT_SYMBOL_GPL(static_key_initialized);
 170
 171 /*
 172  * If set, this is an indication to the drivers that reset the underlying
 173  * device before going ahead with the initialization otherwise driver might
 174  * rely on the BIOS and skip the reset operation.
 175  *
 176  * This is useful if kernel is booting in an unreliable environment.
 177  * For ex. kdump situation where previous kernel has crashed, BIOS has been
 178  * skipped and devices will be in unknown state.
 179  */
 180 unsigned int reset_devices;
 181 EXPORT_SYMBOL(reset_devices);
 182
 183 static int __init set_reset_devices(char *str)
 184 {
 185         reset_devices = 1;
 186         return 1;
 187 }
 188
 189 __setup("reset_devices", set_reset_devices);
 190
 191 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 192 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 193 static const char *panic_later, *panic_param;
 194
 195 static bool __init obsolete_checksetup(char *line)
 196 {
 197         const struct obs_kernel_param *p;
 198         bool had_early_param = false;
 199
 200         p = __setup_start;
 201         do {
 202                 int n = strlen(p->str);
 203                 if (parameqn(line, p->str, n)) {
 204                         if (p->early) {
 205                                 /* Already done in parse_early_param?
 206                                  * (Needs exact match on param part).
 207                                  * Keep iterating, as we can have early
 208                                  * params and __setups of same names 8( */
 209                                 if (line[n] == '\0' || line[n] == '=')
 210                                         had_early_param = true;
 211                         } else if (!p->setup_func) {
 212                                 pr_warn("Parameter %s is obsolete, ignored\n",
 213                                         p->str);
 214                                 return true;
 215                         } else if (p->setup_func(line + n))
 216                                 return true;
 217                 }
 218                 p++;
 219         } while (p < __setup_end);
 220
 221         return had_early_param;
 222 }
 223
 224 /*
 225  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 226  * still work even if initially too large, it will just take slightly longer
 227  */
 228 unsigned long loops_per_jiffy = (1<<12);
 229 EXPORT_SYMBOL(loops_per_jiffy);
 230
 231 static int __init debug_kernel(char *str)
 232 {
 233         console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
 234         return 0;
 235 }
 236
 237 static int __init quiet_kernel(char *str)
 238 {
 239         console_loglevel = CONSOLE_LOGLEVEL_QUIET;
 240         return 0;
 241 }
 242
 243 early_param("debug", debug_kernel);
 244 early_param("quiet", quiet_kernel);
 245
 246 static int __init loglevel(char *str)
 247 {
 248         int newlevel;
 249
 250         /*
 251          * Only update loglevel value when a correct setting was passed,
 252          * to prevent blind crashes (when loglevel being set to 0) that
 253          * are quite hard to debug
 254          */
 255         if (get_option(&str, &newlevel)) {
 256                 console_loglevel = newlevel;
 257                 return 0;
 258         }
 259
 260         return -EINVAL;
 261 }
 262
 263 early_param("loglevel", loglevel);
 264
 265 #ifdef CONFIG_BLK_DEV_INITRD
 266 static void * __init get_boot_config_from_initrd(size_t *_size)
 267 {
 268         u32 size, csum;
 269         char *data;
 270         u32 *hdr;
 271         int i;
 272
 273         if (!initrd_end)
 274                 return NULL;
 275
 276         data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
 277         /*
 278          * Since Grub may align the size of initrd to 4, we must
 279          * check the preceding 3 bytes as well.
 280          */
 281         for (i = 0; i < 4; i++) {
 282                 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
 283                         goto found;
 284                 data--;
 285         }
 286         return NULL;
 287
 288 found:
 289         hdr = (u32 *)(data - 8);
 290         size = le32_to_cpu(hdr[0]);
 291         csum = le32_to_cpu(hdr[1]);
 292
 293         data = ((void *)hdr) - size;
 294         if ((unsigned long)data < initrd_start) {
 295                 pr_err("bootconfig size %d is greater than initrd size %ld\n",
 296                         size, initrd_end - initrd_start);
 297                 return NULL;
 298         }
 299
 300         if (xbc_calc_checksum(data, size) != csum) {
 301                 pr_err("bootconfig checksum failed\n");
 302                 return NULL;
 303         }
 304
 305         /* Remove bootconfig from initramfs/initrd */
 306         initrd_end = (unsigned long)data;
 307         if (_size)
 308                 *_size = size;
 309
 310         return data;
 311 }
 312 #else
 313 static void * __init get_boot_config_from_initrd(size_t *_size)
 314 {
 315         return NULL;
 316 }
 317 #endif
 318
 319 #ifdef CONFIG_BOOT_CONFIG
 320
 321 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
 322
 323 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
 324
 325 static int __init xbc_snprint_cmdline(char *buf, size_t size,
 326                                       struct xbc_node *root)
 327 {
 328         struct xbc_node *knode, *vnode;
 329         char *end = buf + size;
 330         const char *val;
 331         int ret;
 332
 333         xbc_node_for_each_key_value(root, knode, val) {
 334                 ret = xbc_node_compose_key_after(root, knode,
 335                                         xbc_namebuf, XBC_KEYLEN_MAX);
 336                 if (ret < 0)
 337                         return ret;
 338
 339                 vnode = xbc_node_get_child(knode);
 340                 if (!vnode) {
 341                         ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
 342                         if (ret < 0)
 343                                 return ret;
 344                         buf += ret;
 345                         continue;
 346                 }
 347                 xbc_array_for_each_value(vnode, val) {
 348                         ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
 349                                        xbc_namebuf, val);
 350                         if (ret < 0)
 351                                 return ret;
 352                         buf += ret;
 353                 }
 354         }
 355
 356         return buf - (end - size);
 357 }
 358 #undef rest
 359
 360 /* Make an extra command line under given key word */
 361 static char * __init xbc_make_cmdline(const char *key)
 362 {
 363         struct xbc_node *root;
 364         char *new_cmdline;
 365         int ret, len = 0;
 366
 367         root = xbc_find_node(key);
 368         if (!root)
 369                 return NULL;
 370
 371         /* Count required buffer size */
 372         len = xbc_snprint_cmdline(NULL, 0, root);
 373         if (len <= 0)
 374                 return NULL;
 375
 376         new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
 377         if (!new_cmdline) {
 378                 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
 379                 return NULL;
 380         }
 381
 382         ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
 383         if (ret < 0 || ret > len) {
 384                 pr_err("Failed to print extra kernel cmdline.\n");
 385                 memblock_free(new_cmdline, len + 1);
 386                 return NULL;
 387         }
 388
 389         return new_cmdline;
 390 }
 391
 392 static int __init bootconfig_params(char *param, char *val,
 393                                     const char *unused, void *arg)
 394 {
 395         if (strcmp(param, "bootconfig") == 0) {
 396                 bootconfig_found = true;
 397         }
 398         return 0;
 399 }
 400
 401 static int __init warn_bootconfig(char *str)
 402 {
 403         /* The 'bootconfig' has been handled by bootconfig_params(). */
 404         return 0;
 405 }
 406
 407 static void __init setup_boot_config(void)
 408 {
 409         static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
 410         const char *msg, *data;
 411         int pos, ret;
 412         size_t size;
 413         char *err;
 414
 415         /* Cut out the bootconfig data even if we have no bootconfig option */
 416         data = get_boot_config_from_initrd(&size);
 417         /* If there is no bootconfig in initrd, try embedded one. */
 418         if (!data)
 419                 data = xbc_get_embedded_bootconfig(&size);
 420
 421         strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 422         err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
 423                          bootconfig_params);
 424
 425         if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE)))
 426                 return;
 427
 428         /* parse_args() stops at the next param of '--' and returns an address */
 429         if (err)
 430                 initargs_offs = err - tmp_cmdline;
 431
 432         if (!data) {
 433                 /* If user intended to use bootconfig, show an error level message */
 434                 if (bootconfig_found)
 435                         pr_err("'bootconfig' found on command line, but no bootconfig found\n");
 436                 else
 437                         pr_info("No bootconfig data provided, so skipping bootconfig");
 438                 return;
 439         }
 440
 441         if (size >= XBC_DATA_MAX) {
 442                 pr_err("bootconfig size %ld greater than max size %d\n",
 443                         (long)size, XBC_DATA_MAX);
 444                 return;
 445         }
 446
 447         ret = xbc_init(data, size, &msg, &pos);
 448         if (ret < 0) {
 449                 if (pos < 0)
 450                         pr_err("Failed to init bootconfig: %s.\n", msg);
 451                 else
 452                         pr_err("Failed to parse bootconfig: %s at %d.\n",
 453                                 msg, pos);
 454         } else {
 455                 xbc_get_info(&ret, NULL);
 456                 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
 457                 /* keys starting with "kernel." are passed via cmdline */
 458                 extra_command_line = xbc_make_cmdline("kernel");
 459                 /* Also, "init." keys are init arguments */
 460                 extra_init_args = xbc_make_cmdline("init");
 461         }
 462         return;
 463 }
 464
 465 static void __init exit_boot_config(void)
 466 {
 467         xbc_exit();
 468 }
 469
 470 #else   /* !CONFIG_BOOT_CONFIG */
 471
 472 static void __init setup_boot_config(void)
 473 {
 474         /* Remove bootconfig data from initrd */
 475         get_boot_config_from_initrd(NULL);
 476 }
 477
 478 static int __init warn_bootconfig(char *str)
 479 {
 480         pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
 481         return 0;
 482 }
 483
 484 #define exit_boot_config()      do {} while (0)
 485
 486 #endif  /* CONFIG_BOOT_CONFIG */
 487
 488 early_param("bootconfig", warn_bootconfig);
 489
 490 /* Change NUL term back to "=", to make "param" the whole string. */
 491 static void __init repair_env_string(char *param, char *val)
 492 {
 493         if (val) {
 494                 /* param=val or param="val"? */
 495                 if (val == param+strlen(param)+1)
 496                         val[-1] = '=';
 497                 else if (val == param+strlen(param)+2) {
 498                         val[-2] = '=';
 499                         memmove(val-1, val, strlen(val)+1);
 500                 } else
 501                         BUG();
 502         }
 503 }
 504
 505 /* Anything after -- gets handed straight to init. */
 506 static int __init set_init_arg(char *param, char *val,
 507                                const char *unused, void *arg)
 508 {
 509         unsigned int i;
 510
 511         if (panic_later)
 512                 return 0;
 513
 514         repair_env_string(param, val);
 515
 516         for (i = 0; argv_init[i]; i++) {
 517                 if (i == MAX_INIT_ARGS) {
 518                         panic_later = "init";
 519                         panic_param = param;
 520                         return 0;
 521                 }
 522         }
 523         argv_init[i] = param;
 524         return 0;
 525 }
 526
 527 /*
 528  * Unknown boot options get handed to init, unless they look like
 529  * unused parameters (modprobe will find them in /proc/cmdline).
 530  */
 531 static int __init unknown_bootoption(char *param, char *val,
 532                                      const char *unused, void *arg)
 533 {
 534         size_t len = strlen(param);
 535
 536         /* Handle params aliased to sysctls */
 537         if (sysctl_is_alias(param))
 538                 return 0;
 539
 540         repair_env_string(param, val);
 541
 542         /* Handle obsolete-style parameters */
 543         if (obsolete_checksetup(param))
 544                 return 0;
 545
 546         /* Unused module parameter. */
 547         if (strnchr(param, len, '.'))
 548                 return 0;
 549
 550         if (panic_later)
 551                 return 0;
 552
 553         if (val) {
 554                 /* Environment option */
 555                 unsigned int i;
 556                 for (i = 0; envp_init[i]; i++) {
 557                         if (i == MAX_INIT_ENVS) {
 558                                 panic_later = "env";
 559                                 panic_param = param;
 560                         }
 561                         if (!strncmp(param, envp_init[i], len+1))
 562                                 break;
 563                 }
 564                 envp_init[i] = param;
 565         } else {
 566                 /* Command line option */
 567                 unsigned int i;
 568                 for (i = 0; argv_init[i]; i++) {
 569                         if (i == MAX_INIT_ARGS) {
 570                                 panic_later = "init";
 571                                 panic_param = param;
 572                         }
 573                 }
 574                 argv_init[i] = param;
 575         }
 576         return 0;
 577 }
 578
 579 static int __init init_setup(char *str)
 580 {
 581         unsigned int i;
 582
 583         execute_command = str;
 584         /*
 585          * In case LILO is going to boot us with default command line,
 586          * it prepends "auto" before the whole cmdline which makes
 587          * the shell think it should execute a script with such name.
 588          * So we ignore all arguments entered _before_ init=... [MJ]
 589          */
 590         for (i = 1; i < MAX_INIT_ARGS; i++)
 591                 argv_init[i] = NULL;
 592         return 1;
 593 }
 594 __setup("init=", init_setup);
 595
 596 static int __init rdinit_setup(char *str)
 597 {
 598         unsigned int i;
 599
 600         ramdisk_execute_command = str;
 601         /* See "auto" comment in init_setup */
 602         for (i = 1; i < MAX_INIT_ARGS; i++)
 603                 argv_init[i] = NULL;
 604         return 1;
 605 }
 606 __setup("rdinit=", rdinit_setup);
 607
 608 #ifndef CONFIG_SMP
 609 static inline void setup_nr_cpu_ids(void) { }
 610 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 611 #endif
 612
 613 /*
 614  * We need to store the untouched command line for future reference.
 615  * We also need to store the touched command line since the parameter
 616  * parsing is performed in place, and we should allow a component to
 617  * store reference of name/value for future reference.
 618  */
 619 static void __init setup_command_line(char *command_line)
 620 {
 621         size_t len, xlen = 0, ilen = 0;
 622
 623         if (extra_command_line)
 624                 xlen = strlen(extra_command_line);
 625         if (extra_init_args)
 626                 ilen = strlen(extra_init_args) + 4; /* for " -- " */
 627
 628         len = xlen + strlen(boot_command_line) + 1;
 629
 630         saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
 631         if (!saved_command_line)
 632                 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
 633
 634         static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
 635         if (!static_command_line)
 636                 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
 637
 638         if (xlen) {
 639                 /*
 640                  * We have to put extra_command_line before boot command
 641                  * lines because there could be dashes (separator of init
 642                  * command line) in the command lines.
 643                  */
 644                 strcpy(saved_command_line, extra_command_line);
 645                 strcpy(static_command_line, extra_command_line);
 646         }
 647         strcpy(saved_command_line + xlen, boot_command_line);
 648         strcpy(static_command_line + xlen, command_line);
 649
 650         if (ilen) {
 651                 /*
 652                  * Append supplemental init boot args to saved_command_line
 653                  * so that user can check what command line options passed
 654                  * to init.
 655                  * The order should always be
 656                  * " -- "[bootconfig init-param][cmdline init-param]
 657                  */
 658                 if (initargs_offs) {
 659                         len = xlen + initargs_offs;
 660                         strcpy(saved_command_line + len, extra_init_args);
 661                         len += ilen - 4;        /* strlen(extra_init_args) */
 662                         strcpy(saved_command_line + len,
 663                                 boot_command_line + initargs_offs - 1);
 664                 } else {
 665                         len = strlen(saved_command_line);
 666                         strcpy(saved_command_line + len, " -- ");
 667                         len += 4;
 668                         strcpy(saved_command_line + len, extra_init_args);
 669                 }
 670         }
 671
 672         saved_command_line_len = strlen(saved_command_line);
 673 }
 674
 675 /*
 676  * We need to finalize in a non-__init function or else race conditions
 677  * between the root thread and the init thread may cause start_kernel to
 678  * be reaped by free_initmem before the root thread has proceeded to
 679  * cpu_idle.
 680  *
 681  * gcc-3.4 accidentally inlines this function, so use noinline.
 682  */
 683
 684 static __initdata DECLARE_COMPLETION(kthreadd_done);
 685
 686 static noinline void __ref __noreturn rest_init(void)
 687 {
 688         struct task_struct *tsk;
 689         int pid;
 690
 691         rcu_scheduler_starting();
 692         /*
 693          * We need to spawn init first so that it obtains pid 1, however
 694          * the init task will end up wanting to create kthreads, which, if
 695          * we schedule it before we create kthreadd, will OOPS.
 696          */
 697         pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
 698         /*
 699          * Pin init on the boot CPU. Task migration is not properly working
 700          * until sched_init_smp() has been run. It will set the allowed
 701          * CPUs for init to the non isolated CPUs.
 702          */
 703         rcu_read_lock();
 704         tsk = find_task_by_pid_ns(pid, &init_pid_ns);
 705         tsk->flags |= PF_NO_SETAFFINITY;
 706         set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
 707         rcu_read_unlock();
 708
 709         numa_default_policy();
 710         pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES);
 711         rcu_read_lock();
 712         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 713         rcu_read_unlock();
 714
 715         /*
 716          * Enable might_sleep() and smp_processor_id() checks.
 717          * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
 718          * kernel_thread() would trigger might_sleep() splats. With
 719          * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
 720          * already, but it's stuck on the kthreadd_done completion.
 721          */
 722         system_state = SYSTEM_SCHEDULING;
 723
 724         complete(&kthreadd_done);
 725
 726         /*
 727          * The boot idle thread must execute schedule()
 728          * at least once to get things moving:
 729          */
 730         schedule_preempt_disabled();
 731         /* Call into cpu_idle with preempt disabled */
 732         cpu_startup_entry(CPUHP_ONLINE);
 733 }
 734
 735 /* Check for early params. */
 736 static int __init do_early_param(char *param, char *val,
 737                                  const char *unused, void *arg)
 738 {
 739         const struct obs_kernel_param *p;
 740
 741         for (p = __setup_start; p < __setup_end; p++) {
 742                 if ((p->early && parameq(param, p->str)) ||
 743                     (strcmp(param, "console") == 0 &&
 744                      strcmp(p->str, "earlycon") == 0)
 745                 ) {
 746                         if (p->setup_func(val) != 0)
 747                                 pr_warn("Malformed early option '%s'\n", param);
 748                 }
 749         }
 750         /* We accept everything at this stage. */
 751         return 0;
 752 }
 753
 754 void __init parse_early_options(char *cmdline)
 755 {
 756         parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
 757                    do_early_param);
 758 }
 759
 760 /* Arch code calls this early on, or if not, just before other parsing. */
 761 void __init parse_early_param(void)
 762 {
 763         static int done __initdata;
 764         static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
 765
 766         if (done)
 767                 return;
 768
 769         /* All fall through to do_early_param. */
 770         strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 771         parse_early_options(tmp_cmdline);
 772         done = 1;
 773 }
 774
 775 void __init __weak arch_post_acpi_subsys_init(void) { }
 776
 777 void __init __weak smp_setup_processor_id(void)
 778 {
 779 }
 780
 781 void __init __weak smp_prepare_boot_cpu(void)
 782 {
 783 }
 784
 785 # if THREAD_SIZE >= PAGE_SIZE
 786 void __init __weak thread_stack_cache_init(void)
 787 {
 788 }
 789 #endif
 790
 791 void __init __weak poking_init(void) { }
 792
 793 void __init __weak pgtable_cache_init(void) { }
 794
 795 void __init __weak trap_init(void) { }
 796
 797 bool initcall_debug;
 798 core_param(initcall_debug, initcall_debug, bool, 0644);
 799
 800 #ifdef TRACEPOINTS_ENABLED
 801 static void __init initcall_debug_enable(void);
 802 #else
 803 static inline void initcall_debug_enable(void)
 804 {
 805 }
 806 #endif
 807
 808 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
 809 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
 810                            randomize_kstack_offset);
 811 DEFINE_PER_CPU(u32, kstack_offset);
 812
 813 static int __init early_randomize_kstack_offset(char *buf)
 814 {
 815         int ret;
 816         bool bool_result;
 817
 818         ret = kstrtobool(buf, &bool_result);
 819         if (ret)
 820                 return ret;
 821
 822         if (bool_result)
 823                 static_branch_enable(&randomize_kstack_offset);
 824         else
 825                 static_branch_disable(&randomize_kstack_offset);
 826         return 0;
 827 }
 828 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
 829 #endif
 830
 831 static void __init print_unknown_bootoptions(void)
 832 {
 833         char *unknown_options;
 834         char *end;
 835         const char *const *p;
 836         size_t len;
 837
 838         if (panic_later || (!argv_init[1] && !envp_init[2]))
 839                 return;
 840
 841         /*
 842          * Determine how many options we have to print out, plus a space
 843          * before each
 844          */
 845         len = 1; /* null terminator */
 846         for (p = &argv_init[1]; *p; p++) {
 847                 len++;
 848                 len += strlen(*p);
 849         }
 850         for (p = &envp_init[2]; *p; p++) {
 851                 len++;
 852                 len += strlen(*p);
 853         }
 854
 855         unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
 856         if (!unknown_options) {
 857                 pr_err("%s: Failed to allocate %zu bytes\n",
 858                         __func__, len);
 859                 return;
 860         }
 861         end = unknown_options;
 862
 863         for (p = &argv_init[1]; *p; p++)
 864                 end += sprintf(end, " %s", *p);
 865         for (p = &envp_init[2]; *p; p++)
 866                 end += sprintf(end, " %s", *p);
 867
 868         /* Start at unknown_options[1] to skip the initial space */
 869         pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
 870                 &unknown_options[1]);
 871         memblock_free(unknown_options, len);
 872 }
 873
 874 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector
 875 void start_kernel(void)
 876 {
 877         char *command_line;
 878         char *after_dashes;
 879
 880         set_task_stack_end_magic(&init_task);
 881         smp_setup_processor_id();
 882         debug_objects_early_init();
 883         init_vmlinux_build_id();
 884
 885         cgroup_init_early();
 886
 887         local_irq_disable();
 888         early_boot_irqs_disabled = true;
 889
 890         /*
 891          * Interrupts are still disabled. Do necessary setups, then
 892          * enable them.
 893          */
 894         boot_cpu_init();
 895         page_address_init();
 896         pr_notice("%s", linux_banner);
 897         early_security_init();
 898         setup_arch(&command_line);
 899         setup_boot_config();
 900         setup_command_line(command_line);
 901         setup_nr_cpu_ids();
 902         setup_per_cpu_areas();
 903         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 904         boot_cpu_hotplug_init();
 905
 906         pr_notice("Kernel command line: %s\n", saved_command_line);
 907         /* parameters may set static keys */
 908         jump_label_init();
 909         parse_early_param();
 910         after_dashes = parse_args("Booting kernel",
 911                                   static_command_line, __start___param,
 912                                   __stop___param - __start___param,
 913                                   -1, -1, NULL, &unknown_bootoption);
 914         print_unknown_bootoptions();
 915         if (!IS_ERR_OR_NULL(after_dashes))
 916                 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
 917                            NULL, set_init_arg);
 918         if (extra_init_args)
 919                 parse_args("Setting extra init args", extra_init_args,
 920                            NULL, 0, -1, -1, NULL, set_init_arg);
 921
 922         /* Architectural and non-timekeeping rng init, before allocator init */
 923         random_init_early(command_line);
 924
 925         /*
 926          * These use large bootmem allocations and must precede
 927          * initalization of page allocator
 928          */
 929         setup_log_buf(0);
 930         vfs_caches_init_early();
 931         sort_main_extable();
 932         trap_init();
 933         mm_core_init();
 934         poking_init();
 935         ftrace_init();
 936
 937         /* trace_printk can be enabled here */
 938         early_trace_init();
 939
 940         /*
 941          * Set up the scheduler prior starting any interrupts (such as the
 942          * timer interrupt). Full topology setup happens at smp_init()
 943          * time - but meanwhile we still have a functioning scheduler.
 944          */
 945         sched_init();
 946
 947         if (WARN(!irqs_disabled(),
 948                  "Interrupts were enabled *very* early, fixing it\n"))
 949                 local_irq_disable();
 950         radix_tree_init();
 951         maple_tree_init();
 952
 953         /*
 954          * Set up housekeeping before setting up workqueues to allow the unbound
 955          * workqueue to take non-housekeeping into account.
 956          */
 957         housekeeping_init();
 958
 959         /*
 960          * Allow workqueue creation and work item queueing/cancelling
 961          * early.  Work item execution depends on kthreads and starts after
 962          * workqueue_init().
 963          */
 964         workqueue_init_early();
 965
 966         rcu_init();
 967
 968         /* Trace events are available after this */
 969         trace_init();
 970
 971         if (initcall_debug)
 972                 initcall_debug_enable();
 973
 974         context_tracking_init();
 975         /* init some links before init_ISA_irqs() */
 976         early_irq_init();
 977         init_IRQ();
 978         tick_init();
 979         rcu_init_nohz();
 980         init_timers();
 981         srcu_init();
 982         hrtimers_init();
 983         softirq_init();
 984         timekeeping_init();
 985         time_init();
 986
 987         /* This must be after timekeeping is initialized */
 988         random_init();
 989
 990         /* These make use of the fully initialized rng */
 991         kfence_init();
 992         boot_init_stack_canary();
 993
 994         perf_event_init();
 995         profile_init();
 996         call_function_init();
 997         WARN(!irqs_disabled(), "Interrupts were enabled early\n");
 998
 999         early_boot_irqs_disabled = false;
1000         local_irq_enable();
1001
1002         kmem_cache_init_late();
1003
1004         /*
1005          * HACK ALERT! This is early. We're enabling the console before
1006          * we've done PCI setups etc, and console_init() must be aware of
1007          * this. But we do want output early, in case something goes wrong.
1008          */
1009         console_init();
1010         if (panic_later)
1011                 panic("Too many boot %s vars at `%s'", panic_later,
1012                       panic_param);
1013
1014         lockdep_init();
1015
1016         /*
1017          * Need to run this when irqs are enabled, because it wants
1018          * to self-test [hard/soft]-irqs on/off lock inversion bugs
1019          * too:
1020          */
1021         locking_selftest();
1022
1023 #ifdef CONFIG_BLK_DEV_INITRD
1024         if (initrd_start && !initrd_below_start_ok &&
1025             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1026                 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1027                     page_to_pfn(virt_to_page((void *)initrd_start)),
1028                     min_low_pfn);
1029                 initrd_start = 0;
1030         }
1031 #endif
1032         setup_per_cpu_pageset();
1033         numa_policy_init();
1034         acpi_early_init();
1035         if (late_time_init)
1036                 late_time_init();
1037         sched_clock_init();
1038         calibrate_delay();
1039
1040         arch_cpu_finalize_init();
1041
1042         pid_idr_init();
1043         anon_vma_init();
1044 #ifdef CONFIG_X86
1045         if (efi_enabled(EFI_RUNTIME_SERVICES))
1046                 efi_enter_virtual_mode();
1047 #endif
1048         thread_stack_cache_init();
1049         cred_init();
1050         fork_init();
1051         proc_caches_init();
1052         uts_ns_init();
1053         key_init();
1054         security_init();
1055         dbg_late_init();
1056         net_ns_init();
1057         vfs_caches_init();
1058         pagecache_init();
1059         signals_init();
1060         seq_file_init();
1061         proc_root_init();
1062         nsfs_init();
1063         pidfs_init();
1064         cpuset_init();
1065         cgroup_init();
1066         taskstats_init_early();
1067         delayacct_init();
1068
1069         acpi_subsystem_init();
1070         arch_post_acpi_subsys_init();
1071         kcsan_init();
1072
1073         /* Do the rest non-__init'ed, we're now alive */
1074         rest_init();
1075
1076         /*
1077          * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10
1078          * and older.
1079          */
1080 #if !__has_attribute(__no_stack_protector__)
1081         prevent_tail_call_optimization();
1082 #endif
1083 }
1084
1085 /* Call all constructor functions linked into the kernel. */
1086 static void __init do_ctors(void)
1087 {
1088 /*
1089  * For UML, the constructors have already been called by the
1090  * normal setup code as it's just a normal ELF binary, so we
1091  * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1092  * even on UML for modules.
1093  */
1094 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1095         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1096
1097         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1098                 (*fn)();
1099 #endif
1100 }
1101
1102 #ifdef CONFIG_KALLSYMS
1103 struct blacklist_entry {
1104         struct list_head next;
1105         char *buf;
1106 };
1107
1108 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1109
1110 static int __init initcall_blacklist(char *str)
1111 {
1112         char *str_entry;
1113         struct blacklist_entry *entry;
1114
1115         /* str argument is a comma-separated list of functions */
1116         do {
1117                 str_entry = strsep(&str, ",");
1118                 if (str_entry) {
1119                         pr_debug("blacklisting initcall %s\n", str_entry);
1120                         entry = memblock_alloc(sizeof(*entry),
1121                                                SMP_CACHE_BYTES);
1122                         if (!entry)
1123                                 panic("%s: Failed to allocate %zu bytes\n",
1124                                       __func__, sizeof(*entry));
1125                         entry->buf = memblock_alloc(strlen(str_entry) + 1,
1126                                                     SMP_CACHE_BYTES);
1127                         if (!entry->buf)
1128                                 panic("%s: Failed to allocate %zu bytes\n",
1129                                       __func__, strlen(str_entry) + 1);
1130                         strcpy(entry->buf, str_entry);
1131                         list_add(&entry->next, &blacklisted_initcalls);
1132                 }
1133         } while (str_entry);
1134
1135         return 1;
1136 }
1137
1138 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1139 {
1140         struct blacklist_entry *entry;
1141         char fn_name[KSYM_SYMBOL_LEN];
1142         unsigned long addr;
1143
1144         if (list_empty(&blacklisted_initcalls))
1145                 return false;
1146
1147         addr = (unsigned long) dereference_function_descriptor(fn);
1148         sprint_symbol_no_offset(fn_name, addr);
1149
1150         /*
1151          * fn will be "function_name [module_name]" where [module_name] is not
1152          * displayed for built-in init functions.  Strip off the [module_name].
1153          */
1154         strreplace(fn_name, ' ', '\0');
1155
1156         list_for_each_entry(entry, &blacklisted_initcalls, next) {
1157                 if (!strcmp(fn_name, entry->buf)) {
1158                         pr_debug("initcall %s blacklisted\n", fn_name);
1159                         return true;
1160                 }
1161         }
1162
1163         return false;
1164 }
1165 #else
1166 static int __init initcall_blacklist(char *str)
1167 {
1168         pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1169         return 0;
1170 }
1171
1172 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1173 {
1174         return false;
1175 }
1176 #endif
1177 __setup("initcall_blacklist=", initcall_blacklist);
1178
1179 static __init_or_module void
1180 trace_initcall_start_cb(void *data, initcall_t fn)
1181 {
1182         ktime_t *calltime = data;
1183
1184         printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1185         *calltime = ktime_get();
1186 }
1187
1188 static __init_or_module void
1189 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1190 {
1191         ktime_t rettime, *calltime = data;
1192
1193         rettime = ktime_get();
1194         printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1195                  fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1196 }
1197
1198 static ktime_t initcall_calltime;
1199
1200 #ifdef TRACEPOINTS_ENABLED
1201 static void __init initcall_debug_enable(void)
1202 {
1203         int ret;
1204
1205         ret = register_trace_initcall_start(trace_initcall_start_cb,
1206                                             &initcall_calltime);
1207         ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1208                                               &initcall_calltime);
1209         WARN(ret, "Failed to register initcall tracepoints\n");
1210 }
1211 # define do_trace_initcall_start        trace_initcall_start
1212 # define do_trace_initcall_finish       trace_initcall_finish
1213 #else
1214 static inline void do_trace_initcall_start(initcall_t fn)
1215 {
1216         if (!initcall_debug)
1217                 return;
1218         trace_initcall_start_cb(&initcall_calltime, fn);
1219 }
1220 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1221 {
1222         if (!initcall_debug)
1223                 return;
1224         trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1225 }
1226 #endif /* !TRACEPOINTS_ENABLED */
1227
1228 int __init_or_module do_one_initcall(initcall_t fn)
1229 {
1230         int count = preempt_count();
1231         char msgbuf[64];
1232         int ret;
1233
1234         if (initcall_blacklisted(fn))
1235                 return -EPERM;
1236
1237         do_trace_initcall_start(fn);
1238         ret = fn();
1239         do_trace_initcall_finish(fn, ret);
1240
1241         msgbuf[0] = 0;
1242
1243         if (preempt_count() != count) {
1244                 sprintf(msgbuf, "preemption imbalance ");
1245                 preempt_count_set(count);
1246         }
1247         if (irqs_disabled()) {
1248                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1249                 local_irq_enable();
1250         }
1251         WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1252
1253         add_latent_entropy();
1254         return ret;
1255 }
1256
1257
1258 static initcall_entry_t *initcall_levels[] __initdata = {
1259         __initcall0_start,
1260         __initcall1_start,
1261         __initcall2_start,
1262         __initcall3_start,
1263         __initcall4_start,
1264         __initcall5_start,
1265         __initcall6_start,
1266         __initcall7_start,
1267         __initcall_end,
1268 };
1269
1270 /* Keep these in sync with initcalls in include/linux/init.h */
1271 static const char *initcall_level_names[] __initdata = {
1272         "pure",
1273         "core",
1274         "postcore",
1275         "arch",
1276         "subsys",
1277         "fs",
1278         "device",
1279         "late",
1280 };
1281
1282 static int __init ignore_unknown_bootoption(char *param, char *val,
1283                                const char *unused, void *arg)
1284 {
1285         return 0;
1286 }
1287
1288 static void __init do_initcall_level(int level, char *command_line)
1289 {
1290         initcall_entry_t *fn;
1291
1292         parse_args(initcall_level_names[level],
1293                    command_line, __start___param,
1294                    __stop___param - __start___param,
1295                    level, level,
1296                    NULL, ignore_unknown_bootoption);
1297
1298         trace_initcall_level(initcall_level_names[level]);
1299         for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1300                 do_one_initcall(initcall_from_entry(fn));
1301 }
1302
1303 static void __init do_initcalls(void)
1304 {
1305         int level;
1306         size_t len = saved_command_line_len + 1;
1307         char *command_line;
1308
1309         command_line = kzalloc(len, GFP_KERNEL);
1310         if (!command_line)
1311                 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1312
1313         for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1314                 /* Parser modifies command_line, restore it each time */
1315                 strcpy(command_line, saved_command_line);
1316                 do_initcall_level(level, command_line);
1317         }
1318
1319         kfree(command_line);
1320 }
1321
1322 /*
1323  * Ok, the machine is now initialized. None of the devices
1324  * have been touched yet, but the CPU subsystem is up and
1325  * running, and memory and process management works.
1326  *
1327  * Now we can finally start doing some real work..
1328  */
1329 static void __init do_basic_setup(void)
1330 {
1331         cpuset_init_smp();
1332         driver_init();
1333         init_irq_proc();
1334         do_ctors();
1335         do_initcalls();
1336 }
1337
1338 static void __init do_pre_smp_initcalls(void)
1339 {
1340         initcall_entry_t *fn;
1341
1342         trace_initcall_level("early");
1343         for (fn = __initcall_start; fn < __initcall0_start; fn++)
1344                 do_one_initcall(initcall_from_entry(fn));
1345 }
1346
1347 static int run_init_process(const char *init_filename)
1348 {
1349         const char *const *p;
1350
1351         argv_init[0] = init_filename;
1352         pr_info("Run %s as init process\n", init_filename);
1353         pr_debug("  with arguments:\n");
1354         for (p = argv_init; *p; p++)
1355                 pr_debug("    %s\n", *p);
1356         pr_debug("  with environment:\n");
1357         for (p = envp_init; *p; p++)
1358                 pr_debug("    %s\n", *p);
1359         return kernel_execve(init_filename, argv_init, envp_init);
1360 }
1361
1362 static int try_to_run_init_process(const char *init_filename)
1363 {
1364         int ret;
1365
1366         ret = run_init_process(init_filename);
1367
1368         if (ret && ret != -ENOENT) {
1369                 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1370                        init_filename, ret);
1371         }
1372
1373         return ret;
1374 }
1375
1376 static noinline void __init kernel_init_freeable(void);
1377
1378 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1379 bool rodata_enabled __ro_after_init = true;
1380
1381 #ifndef arch_parse_debug_rodata
1382 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1383 #endif
1384
1385 static int __init set_debug_rodata(char *str)
1386 {
1387         if (arch_parse_debug_rodata(str))
1388                 return 0;
1389
1390         if (str && !strcmp(str, "on"))
1391                 rodata_enabled = true;
1392         else if (str && !strcmp(str, "off"))
1393                 rodata_enabled = false;
1394         else
1395                 pr_warn("Invalid option string for rodata: '%s'\n", str);
1396         return 0;
1397 }
1398 early_param("rodata", set_debug_rodata);
1399 #endif
1400
1401 static void mark_readonly(void)
1402 {
1403         if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX) && rodata_enabled) {
1404                 /*
1405                  * load_module() results in W+X mappings, which are cleaned
1406                  * up with init_free_wq. Let's make sure that queued work is
1407                  * flushed so that we don't hit false positives looking for
1408                  * insecure pages which are W+X.
1409                  */
1410                 flush_module_init_free_work();
1411                 mark_rodata_ro();
1412                 debug_checkwx();
1413                 rodata_test();
1414         } else if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) {
1415                 pr_info("Kernel memory protection disabled.\n");
1416         } else if (IS_ENABLED(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)) {
1417                 pr_warn("Kernel memory protection not selected by kernel config.\n");
1418         } else {
1419                 pr_warn("This architecture does not have kernel memory protection.\n");
1420         }
1421 }
1422
1423 void __weak free_initmem(void)
1424 {
1425         free_initmem_default(POISON_FREE_INITMEM);
1426 }
1427
1428 static int __ref kernel_init(void *unused)
1429 {
1430         int ret;
1431
1432         /*
1433          * Wait until kthreadd is all set-up.
1434          */
1435         wait_for_completion(&kthreadd_done);
1436
1437         kernel_init_freeable();
1438         /* need to finish all async __init code before freeing the memory */
1439         async_synchronize_full();
1440
1441         system_state = SYSTEM_FREEING_INITMEM;
1442         kprobe_free_init_mem();
1443         ftrace_free_init_mem();
1444         kgdb_free_init_mem();
1445         exit_boot_config();
1446         free_initmem();
1447         mark_readonly();
1448
1449         /*
1450          * Kernel mappings are now finalized - update the userspace page-table
1451          * to finalize PTI.
1452          */
1453         pti_finalize();
1454
1455         system_state = SYSTEM_RUNNING;
1456         numa_default_policy();
1457
1458         rcu_end_inkernel_boot();
1459
1460         do_sysctl_args();
1461
1462         if (ramdisk_execute_command) {
1463                 ret = run_init_process(ramdisk_execute_command);
1464                 if (!ret)
1465                         return 0;
1466                 pr_err("Failed to execute %s (error %d)\n",
1467                        ramdisk_execute_command, ret);
1468         }
1469
1470         /*
1471          * We try each of these until one succeeds.
1472          *
1473          * The Bourne shell can be used instead of init if we are
1474          * trying to recover a really broken machine.
1475          */
1476         if (execute_command) {
1477                 ret = run_init_process(execute_command);
1478                 if (!ret)
1479                         return 0;
1480                 panic("Requested init %s failed (error %d).",
1481                       execute_command, ret);
1482         }
1483
1484         if (CONFIG_DEFAULT_INIT[0] != '\0') {
1485                 ret = run_init_process(CONFIG_DEFAULT_INIT);
1486                 if (ret)
1487                         pr_err("Default init %s failed (error %d)\n",
1488                                CONFIG_DEFAULT_INIT, ret);
1489                 else
1490                         return 0;
1491         }
1492
1493         if (!try_to_run_init_process("/sbin/init") ||
1494             !try_to_run_init_process("/etc/init") ||
1495             !try_to_run_init_process("/bin/init") ||
1496             !try_to_run_init_process("/bin/sh"))
1497                 return 0;
1498
1499         panic("No working init found.  Try passing init= option to kernel. "
1500               "See Linux Documentation/admin-guide/init.rst for guidance.");
1501 }
1502
1503 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1504 void __init console_on_rootfs(void)
1505 {
1506         struct file *file = filp_open("/dev/console", O_RDWR, 0);
1507
1508         if (IS_ERR(file)) {
1509                 pr_err("Warning: unable to open an initial console.\n");
1510                 return;
1511         }
1512         init_dup(file);
1513         init_dup(file);
1514         init_dup(file);
1515         fput(file);
1516 }
1517
1518 static noinline void __init kernel_init_freeable(void)
1519 {
1520         /* Now the scheduler is fully set up and can do blocking allocations */
1521         gfp_allowed_mask = __GFP_BITS_MASK;
1522
1523         /*
1524          * init can allocate pages on any node
1525          */
1526         set_mems_allowed(node_states[N_MEMORY]);
1527
1528         cad_pid = get_pid(task_pid(current));
1529
1530         smp_prepare_cpus(setup_max_cpus);
1531
1532         workqueue_init();
1533
1534         init_mm_internals();
1535
1536         rcu_init_tasks_generic();
1537         do_pre_smp_initcalls();
1538         lockup_detector_init();
1539
1540         smp_init();
1541         sched_init_smp();
1542
1543         workqueue_init_topology();
1544         async_init();
1545         padata_init();
1546         page_alloc_init_late();
1547
1548         do_basic_setup();
1549
1550         kunit_run_all_tests();
1551
1552         wait_for_initramfs();
1553         console_on_rootfs();
1554
1555         /*
1556          * check if there is an early userspace init.  If yes, let it do all
1557          * the work
1558          */
1559         if (init_eaccess(ramdisk_execute_command) != 0) {
1560                 ramdisk_execute_command = NULL;
1561                 prepare_namespace();
1562         }
1563
1564         /*
1565          * Ok, we have completed the initial bootup, and
1566          * we're essentially up and running. Get rid of the
1567          * initmem segments and start the user-mode stuff..
1568          *
1569          * rootfs is available now, try loading the public keys
1570          * and default modules
1571          */
1572
1573         integrity_load_keys();
1574 }