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/vmalloc.h>
  38 #include <linux/kernel_stat.h>
  39 #include <linux/start_kernel.h>
  40 #include <linux/security.h>
  41 #include <linux/smp.h>
  42 #include <linux/profile.h>
  43 #include <linux/rcupdate.h>
  44 #include <linux/moduleparam.h>
  45 #include <linux/kallsyms.h>
  46 #include <linux/writeback.h>
  47 #include <linux/cpu.h>
  48 #include <linux/cpuset.h>
  49 #include <linux/cgroup.h>
  50 #include <linux/efi.h>
  51 #include <linux/tick.h>
  52 #include <linux/sched/isolation.h>
  53 #include <linux/interrupt.h>
  54 #include <linux/taskstats_kern.h>
  55 #include <linux/delayacct.h>
  56 #include <linux/unistd.h>
  57 #include <linux/utsname.h>
  58 #include <linux/rmap.h>
  59 #include <linux/mempolicy.h>
  60 #include <linux/key.h>
  61 #include <linux/page_ext.h>
  62 #include <linux/debug_locks.h>
  63 #include <linux/debugobjects.h>
  64 #include <linux/lockdep.h>
  65 #include <linux/kmemleak.h>
  66 #include <linux/padata.h>
  67 #include <linux/pid_namespace.h>
  68 #include <linux/device/driver.h>
  69 #include <linux/kthread.h>
  70 #include <linux/sched.h>
  71 #include <linux/sched/init.h>
  72 #include <linux/signal.h>
  73 #include <linux/idr.h>
  74 #include <linux/kgdb.h>
  75 #include <linux/ftrace.h>
  76 #include <linux/async.h>
  77 #include <linux/sfi.h>
  78 #include <linux/shmem_fs.h>
  79 #include <linux/slab.h>
  80 #include <linux/perf_event.h>
  81 #include <linux/ptrace.h>
  82 #include <linux/pti.h>
  83 #include <linux/blkdev.h>
  84 #include <linux/elevator.h>
  85 #include <linux/sched/clock.h>
  86 #include <linux/sched/task.h>
  87 #include <linux/sched/task_stack.h>
  88 #include <linux/context_tracking.h>
  89 #include <linux/random.h>
  90 #include <linux/list.h>
  91 #include <linux/integrity.h>
  92 #include <linux/proc_ns.h>
  93 #include <linux/io.h>
  94 #include <linux/cache.h>
  95 #include <linux/rodata_test.h>
  96 #include <linux/jump_label.h>
  97 #include <linux/mem_encrypt.h>
  98 #include <linux/kcsan.h>
  99 #include <linux/init_syscalls.h>
 100
 101 #include <asm/io.h>
 102 #include <asm/bugs.h>
 103 #include <asm/setup.h>
 104 #include <asm/sections.h>
 105 #include <asm/cacheflush.h>
 106
 107 #define CREATE_TRACE_POINTS
 108 #include <trace/events/initcall.h>
 109
 110 #include <kunit/test.h>
 111
 112 static int kernel_init(void *);
 113
 114 extern void init_IRQ(void);
 115 extern void radix_tree_init(void);
 116
 117 /*
 118  * Debug helper: via this flag we know that we are in 'early bootup code'
 119  * where only the boot processor is running with IRQ disabled.  This means
 120  * two things - IRQ must not be enabled before the flag is cleared and some
 121  * operations which are not allowed with IRQ disabled are allowed while the
 122  * flag is set.
 123  */
 124 bool early_boot_irqs_disabled __read_mostly;
 125
 126 enum system_states system_state __read_mostly;
 127 EXPORT_SYMBOL(system_state);
 128
 129 /*
 130  * Boot command-line arguments
 131  */
 132 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 133 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 134
 135 extern void time_init(void);
 136 /* Default late time init is NULL. archs can override this later. */
 137 void (*__initdata late_time_init)(void);
 138
 139 /* Untouched command line saved by arch-specific code. */
 140 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 141 /* Untouched saved command line (eg. for /proc) */
 142 char *saved_command_line;
 143 /* Command line for parameter parsing */
 144 static char *static_command_line;
 145 /* Untouched extra command line */
 146 static char *extra_command_line;
 147 /* Extra init arguments */
 148 static char *extra_init_args;
 149
 150 #ifdef CONFIG_BOOT_CONFIG
 151 /* Is bootconfig on command line? */
 152 static bool bootconfig_found;
 153 static bool initargs_found;
 154 #else
 155 # define bootconfig_found false
 156 # define initargs_found false
 157 #endif
 158
 159 static char *execute_command;
 160 static char *ramdisk_execute_command = "/init";
 161
 162 /*
 163  * Used to generate warnings if static_key manipulation functions are used
 164  * before jump_label_init is called.
 165  */
 166 bool static_key_initialized __read_mostly;
 167 EXPORT_SYMBOL_GPL(static_key_initialized);
 168
 169 /*
 170  * If set, this is an indication to the drivers that reset the underlying
 171  * device before going ahead with the initialization otherwise driver might
 172  * rely on the BIOS and skip the reset operation.
 173  *
 174  * This is useful if kernel is booting in an unreliable environment.
 175  * For ex. kdump situation where previous kernel has crashed, BIOS has been
 176  * skipped and devices will be in unknown state.
 177  */
 178 unsigned int reset_devices;
 179 EXPORT_SYMBOL(reset_devices);
 180
 181 static int __init set_reset_devices(char *str)
 182 {
 183         reset_devices = 1;
 184         return 1;
 185 }
 186
 187 __setup("reset_devices", set_reset_devices);
 188
 189 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 190 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 191 static const char *panic_later, *panic_param;
 192
 193 extern const struct obs_kernel_param __setup_start[], __setup_end[];
 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(u32 *_size, u32 *_csum)
 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         /* Remove bootconfig from initramfs/initrd */
 301         initrd_end = (unsigned long)data;
 302         if (_size)
 303                 *_size = size;
 304         if (_csum)
 305                 *_csum = csum;
 306
 307         return data;
 308 }
 309 #else
 310 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
 311 {
 312         return NULL;
 313 }
 314 #endif
 315
 316 #ifdef CONFIG_BOOT_CONFIG
 317
 318 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
 319
 320 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
 321
 322 static int __init xbc_snprint_cmdline(char *buf, size_t size,
 323                                       struct xbc_node *root)
 324 {
 325         struct xbc_node *knode, *vnode;
 326         char *end = buf + size;
 327         const char *val;
 328         int ret;
 329
 330         xbc_node_for_each_key_value(root, knode, val) {
 331                 ret = xbc_node_compose_key_after(root, knode,
 332                                         xbc_namebuf, XBC_KEYLEN_MAX);
 333                 if (ret < 0)
 334                         return ret;
 335
 336                 vnode = xbc_node_get_child(knode);
 337                 if (!vnode) {
 338                         ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
 339                         if (ret < 0)
 340                                 return ret;
 341                         buf += ret;
 342                         continue;
 343                 }
 344                 xbc_array_for_each_value(vnode, val) {
 345                         ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
 346                                        xbc_namebuf, val);
 347                         if (ret < 0)
 348                                 return ret;
 349                         buf += ret;
 350                 }
 351         }
 352
 353         return buf - (end - size);
 354 }
 355 #undef rest
 356
 357 /* Make an extra command line under given key word */
 358 static char * __init xbc_make_cmdline(const char *key)
 359 {
 360         struct xbc_node *root;
 361         char *new_cmdline;
 362         int ret, len = 0;
 363
 364         root = xbc_find_node(key);
 365         if (!root)
 366                 return NULL;
 367
 368         /* Count required buffer size */
 369         len = xbc_snprint_cmdline(NULL, 0, root);
 370         if (len <= 0)
 371                 return NULL;
 372
 373         new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
 374         if (!new_cmdline) {
 375                 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
 376                 return NULL;
 377         }
 378
 379         ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
 380         if (ret < 0 || ret > len) {
 381                 pr_err("Failed to print extra kernel cmdline.\n");
 382                 return NULL;
 383         }
 384
 385         return new_cmdline;
 386 }
 387
 388 static u32 boot_config_checksum(unsigned char *p, u32 size)
 389 {
 390         u32 ret = 0;
 391
 392         while (size--)
 393                 ret += *p++;
 394
 395         return ret;
 396 }
 397
 398 static int __init bootconfig_params(char *param, char *val,
 399                                     const char *unused, void *arg)
 400 {
 401         if (strcmp(param, "bootconfig") == 0) {
 402                 bootconfig_found = true;
 403         }
 404         return 0;
 405 }
 406
 407 static void __init setup_boot_config(const char *cmdline)
 408 {
 409         static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
 410         const char *msg;
 411         int pos;
 412         u32 size, csum;
 413         char *data, *copy, *err;
 414         int ret;
 415
 416         /* Cut out the bootconfig data even if we have no bootconfig option */
 417         data = get_boot_config_from_initrd(&size, &csum);
 418
 419         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 420         err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
 421                          bootconfig_params);
 422
 423         if (IS_ERR(err) || !bootconfig_found)
 424                 return;
 425
 426         /* parse_args() stops at '--' and returns an address */
 427         if (err)
 428                 initargs_found = true;
 429
 430         if (!data) {
 431                 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
 432                 return;
 433         }
 434
 435         if (size >= XBC_DATA_MAX) {
 436                 pr_err("bootconfig size %d greater than max size %d\n",
 437                         size, XBC_DATA_MAX);
 438                 return;
 439         }
 440
 441         if (boot_config_checksum((unsigned char *)data, size) != csum) {
 442                 pr_err("bootconfig checksum failed\n");
 443                 return;
 444         }
 445
 446         copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
 447         if (!copy) {
 448                 pr_err("Failed to allocate memory for bootconfig\n");
 449                 return;
 450         }
 451
 452         memcpy(copy, data, size);
 453         copy[size] = '\0';
 454
 455         ret = xbc_init(copy, &msg, &pos);
 456         if (ret < 0) {
 457                 if (pos < 0)
 458                         pr_err("Failed to init bootconfig: %s.\n", msg);
 459                 else
 460                         pr_err("Failed to parse bootconfig: %s at %d.\n",
 461                                 msg, pos);
 462         } else {
 463                 pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
 464                 /* keys starting with "kernel." are passed via cmdline */
 465                 extra_command_line = xbc_make_cmdline("kernel");
 466                 /* Also, "init." keys are init arguments */
 467                 extra_init_args = xbc_make_cmdline("init");
 468         }
 469         return;
 470 }
 471
 472 #else
 473
 474 static void __init setup_boot_config(const char *cmdline)
 475 {
 476         /* Remove bootconfig data from initrd */
 477         get_boot_config_from_initrd(NULL, NULL);
 478 }
 479
 480 static int __init warn_bootconfig(char *str)
 481 {
 482         pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
 483         return 0;
 484 }
 485 early_param("bootconfig", warn_bootconfig);
 486
 487 #endif
 488
 489 /* Change NUL term back to "=", to make "param" the whole string. */
 490 static void __init repair_env_string(char *param, char *val)
 491 {
 492         if (val) {
 493                 /* param=val or param="val"? */
 494                 if (val == param+strlen(param)+1)
 495                         val[-1] = '=';
 496                 else if (val == param+strlen(param)+2) {
 497                         val[-2] = '=';
 498                         memmove(val-1, val, strlen(val)+1);
 499                 } else
 500                         BUG();
 501         }
 502 }
 503
 504 /* Anything after -- gets handed straight to init. */
 505 static int __init set_init_arg(char *param, char *val,
 506                                const char *unused, void *arg)
 507 {
 508         unsigned int i;
 509
 510         if (panic_later)
 511                 return 0;
 512
 513         repair_env_string(param, val);
 514
 515         for (i = 0; argv_init[i]; i++) {
 516                 if (i == MAX_INIT_ARGS) {
 517                         panic_later = "init";
 518                         panic_param = param;
 519                         return 0;
 520                 }
 521         }
 522         argv_init[i] = param;
 523         return 0;
 524 }
 525
 526 /*
 527  * Unknown boot options get handed to init, unless they look like
 528  * unused parameters (modprobe will find them in /proc/cmdline).
 529  */
 530 static int __init unknown_bootoption(char *param, char *val,
 531                                      const char *unused, void *arg)
 532 {
 533         size_t len = strlen(param);
 534
 535         repair_env_string(param, val);
 536
 537         /* Handle obsolete-style parameters */
 538         if (obsolete_checksetup(param))
 539                 return 0;
 540
 541         /* Unused module parameter. */
 542         if (strnchr(param, len, '.'))
 543                 return 0;
 544
 545         if (panic_later)
 546                 return 0;
 547
 548         if (val) {
 549                 /* Environment option */
 550                 unsigned int i;
 551                 for (i = 0; envp_init[i]; i++) {
 552                         if (i == MAX_INIT_ENVS) {
 553                                 panic_later = "env";
 554                                 panic_param = param;
 555                         }
 556                         if (!strncmp(param, envp_init[i], len+1))
 557                                 break;
 558                 }
 559                 envp_init[i] = param;
 560         } else {
 561                 /* Command line option */
 562                 unsigned int i;
 563                 for (i = 0; argv_init[i]; i++) {
 564                         if (i == MAX_INIT_ARGS) {
 565                                 panic_later = "init";
 566                                 panic_param = param;
 567                         }
 568                 }
 569                 argv_init[i] = param;
 570         }
 571         return 0;
 572 }
 573
 574 static int __init init_setup(char *str)
 575 {
 576         unsigned int i;
 577
 578         execute_command = str;
 579         /*
 580          * In case LILO is going to boot us with default command line,
 581          * it prepends "auto" before the whole cmdline which makes
 582          * the shell think it should execute a script with such name.
 583          * So we ignore all arguments entered _before_ init=... [MJ]
 584          */
 585         for (i = 1; i < MAX_INIT_ARGS; i++)
 586                 argv_init[i] = NULL;
 587         return 1;
 588 }
 589 __setup("init=", init_setup);
 590
 591 static int __init rdinit_setup(char *str)
 592 {
 593         unsigned int i;
 594
 595         ramdisk_execute_command = str;
 596         /* See "auto" comment in init_setup */
 597         for (i = 1; i < MAX_INIT_ARGS; i++)
 598                 argv_init[i] = NULL;
 599         return 1;
 600 }
 601 __setup("rdinit=", rdinit_setup);
 602
 603 #ifndef CONFIG_SMP
 604 static const unsigned int setup_max_cpus = NR_CPUS;
 605 static inline void setup_nr_cpu_ids(void) { }
 606 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 607 #endif
 608
 609 /*
 610  * We need to store the untouched command line for future reference.
 611  * We also need to store the touched command line since the parameter
 612  * parsing is performed in place, and we should allow a component to
 613  * store reference of name/value for future reference.
 614  */
 615 static void __init setup_command_line(char *command_line)
 616 {
 617         size_t len, xlen = 0, ilen = 0;
 618
 619         if (extra_command_line)
 620                 xlen = strlen(extra_command_line);
 621         if (extra_init_args)
 622                 ilen = strlen(extra_init_args) + 4; /* for " -- " */
 623
 624         len = xlen + strlen(boot_command_line) + 1;
 625
 626         saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
 627         if (!saved_command_line)
 628                 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
 629
 630         static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
 631         if (!static_command_line)
 632                 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
 633
 634         if (xlen) {
 635                 /*
 636                  * We have to put extra_command_line before boot command
 637                  * lines because there could be dashes (separator of init
 638                  * command line) in the command lines.
 639                  */
 640                 strcpy(saved_command_line, extra_command_line);
 641                 strcpy(static_command_line, extra_command_line);
 642         }
 643         strcpy(saved_command_line + xlen, boot_command_line);
 644         strcpy(static_command_line + xlen, command_line);
 645
 646         if (ilen) {
 647                 /*
 648                  * Append supplemental init boot args to saved_command_line
 649                  * so that user can check what command line options passed
 650                  * to init.
 651                  */
 652                 len = strlen(saved_command_line);
 653                 if (initargs_found) {
 654                         saved_command_line[len++] = ' ';
 655                 } else {
 656                         strcpy(saved_command_line + len, " -- ");
 657                         len += 4;
 658                 }
 659
 660                 strcpy(saved_command_line + len, extra_init_args);
 661         }
 662 }
 663
 664 /*
 665  * We need to finalize in a non-__init function or else race conditions
 666  * between the root thread and the init thread may cause start_kernel to
 667  * be reaped by free_initmem before the root thread has proceeded to
 668  * cpu_idle.
 669  *
 670  * gcc-3.4 accidentally inlines this function, so use noinline.
 671  */
 672
 673 static __initdata DECLARE_COMPLETION(kthreadd_done);
 674
 675 noinline void __ref rest_init(void)
 676 {
 677         struct task_struct *tsk;
 678         int pid;
 679
 680         rcu_scheduler_starting();
 681         /*
 682          * We need to spawn init first so that it obtains pid 1, however
 683          * the init task will end up wanting to create kthreads, which, if
 684          * we schedule it before we create kthreadd, will OOPS.
 685          */
 686         pid = kernel_thread(kernel_init, NULL, CLONE_FS);
 687         /*
 688          * Pin init on the boot CPU. Task migration is not properly working
 689          * until sched_init_smp() has been run. It will set the allowed
 690          * CPUs for init to the non isolated CPUs.
 691          */
 692         rcu_read_lock();
 693         tsk = find_task_by_pid_ns(pid, &init_pid_ns);
 694         set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
 695         rcu_read_unlock();
 696
 697         numa_default_policy();
 698         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 699         rcu_read_lock();
 700         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 701         rcu_read_unlock();
 702
 703         /*
 704          * Enable might_sleep() and smp_processor_id() checks.
 705          * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
 706          * kernel_thread() would trigger might_sleep() splats. With
 707          * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
 708          * already, but it's stuck on the kthreadd_done completion.
 709          */
 710         system_state = SYSTEM_SCHEDULING;
 711
 712         complete(&kthreadd_done);
 713
 714         /*
 715          * The boot idle thread must execute schedule()
 716          * at least once to get things moving:
 717          */
 718         schedule_preempt_disabled();
 719         /* Call into cpu_idle with preempt disabled */
 720         cpu_startup_entry(CPUHP_ONLINE);
 721 }
 722
 723 /* Check for early params. */
 724 static int __init do_early_param(char *param, char *val,
 725                                  const char *unused, void *arg)
 726 {
 727         const struct obs_kernel_param *p;
 728
 729         for (p = __setup_start; p < __setup_end; p++) {
 730                 if ((p->early && parameq(param, p->str)) ||
 731                     (strcmp(param, "console") == 0 &&
 732                      strcmp(p->str, "earlycon") == 0)
 733                 ) {
 734                         if (p->setup_func(val) != 0)
 735                                 pr_warn("Malformed early option '%s'\n", param);
 736                 }
 737         }
 738         /* We accept everything at this stage. */
 739         return 0;
 740 }
 741
 742 void __init parse_early_options(char *cmdline)
 743 {
 744         parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
 745                    do_early_param);
 746 }
 747
 748 /* Arch code calls this early on, or if not, just before other parsing. */
 749 void __init parse_early_param(void)
 750 {
 751         static int done __initdata;
 752         static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
 753
 754         if (done)
 755                 return;
 756
 757         /* All fall through to do_early_param. */
 758         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 759         parse_early_options(tmp_cmdline);
 760         done = 1;
 761 }
 762
 763 void __init __weak arch_post_acpi_subsys_init(void) { }
 764
 765 void __init __weak smp_setup_processor_id(void)
 766 {
 767 }
 768
 769 # if THREAD_SIZE >= PAGE_SIZE
 770 void __init __weak thread_stack_cache_init(void)
 771 {
 772 }
 773 #endif
 774
 775 void __init __weak mem_encrypt_init(void) { }
 776
 777 void __init __weak poking_init(void) { }
 778
 779 void __init __weak pgtable_cache_init(void) { }
 780
 781 bool initcall_debug;
 782 core_param(initcall_debug, initcall_debug, bool, 0644);
 783
 784 #ifdef TRACEPOINTS_ENABLED
 785 static void __init initcall_debug_enable(void);
 786 #else
 787 static inline void initcall_debug_enable(void)
 788 {
 789 }
 790 #endif
 791
 792 /* Report memory auto-initialization states for this boot. */
 793 static void __init report_meminit(void)
 794 {
 795         const char *stack;
 796
 797         if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
 798                 stack = "all(pattern)";
 799         else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
 800                 stack = "all(zero)";
 801         else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
 802                 stack = "byref_all(zero)";
 803         else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
 804                 stack = "byref(zero)";
 805         else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
 806                 stack = "__user(zero)";
 807         else
 808                 stack = "off";
 809
 810         pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
 811                 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
 812                 want_init_on_free() ? "on" : "off");
 813         if (want_init_on_free())
 814                 pr_info("mem auto-init: clearing system memory may take some time...\n");
 815 }
 816
 817 /*
 818  * Set up kernel memory allocators
 819  */
 820 static void __init mm_init(void)
 821 {
 822         /*
 823          * page_ext requires contiguous pages,
 824          * bigger than MAX_ORDER unless SPARSEMEM.
 825          */
 826         page_ext_init_flatmem();
 827         init_mem_debugging_and_hardening();
 828         report_meminit();
 829         mem_init();
 830         /* page_owner must be initialized after buddy is ready */
 831         page_ext_init_flatmem_late();
 832         kmem_cache_init();
 833         kmemleak_init();
 834         pgtable_init();
 835         debug_objects_mem_init();
 836         vmalloc_init();
 837         ioremap_huge_init();
 838         /* Should be run before the first non-init thread is created */
 839         init_espfix_bsp();
 840         /* Should be run after espfix64 is set up. */
 841         pti_init();
 842 }
 843
 844 void __init __weak arch_call_rest_init(void)
 845 {
 846         rest_init();
 847 }
 848
 849 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
 850 {
 851         char *command_line;
 852         char *after_dashes;
 853
 854         set_task_stack_end_magic(&init_task);
 855         smp_setup_processor_id();
 856         debug_objects_early_init();
 857
 858         cgroup_init_early();
 859
 860         local_irq_disable();
 861         early_boot_irqs_disabled = true;
 862
 863         /*
 864          * Interrupts are still disabled. Do necessary setups, then
 865          * enable them.
 866          */
 867         boot_cpu_init();
 868         page_address_init();
 869         pr_notice("%s", linux_banner);
 870         early_security_init();
 871         setup_arch(&command_line);
 872         setup_boot_config(command_line);
 873         setup_command_line(command_line);
 874         setup_nr_cpu_ids();
 875         setup_per_cpu_areas();
 876         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 877         boot_cpu_hotplug_init();
 878
 879         build_all_zonelists(NULL);
 880         page_alloc_init();
 881
 882         pr_notice("Kernel command line: %s\n", saved_command_line);
 883         /* parameters may set static keys */
 884         jump_label_init();
 885         parse_early_param();
 886         after_dashes = parse_args("Booting kernel",
 887                                   static_command_line, __start___param,
 888                                   __stop___param - __start___param,
 889                                   -1, -1, NULL, &unknown_bootoption);
 890         if (!IS_ERR_OR_NULL(after_dashes))
 891                 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
 892                            NULL, set_init_arg);
 893         if (extra_init_args)
 894                 parse_args("Setting extra init args", extra_init_args,
 895                            NULL, 0, -1, -1, NULL, set_init_arg);
 896
 897         /*
 898          * These use large bootmem allocations and must precede
 899          * kmem_cache_init()
 900          */
 901         setup_log_buf(0);
 902         vfs_caches_init_early();
 903         sort_main_extable();
 904         trap_init();
 905         mm_init();
 906
 907         ftrace_init();
 908
 909         /* trace_printk can be enabled here */
 910         early_trace_init();
 911
 912         /*
 913          * Set up the scheduler prior starting any interrupts (such as the
 914          * timer interrupt). Full topology setup happens at smp_init()
 915          * time - but meanwhile we still have a functioning scheduler.
 916          */
 917         sched_init();
 918         /*
 919          * Disable preemption - early bootup scheduling is extremely
 920          * fragile until we cpu_idle() for the first time.
 921          */
 922         preempt_disable();
 923         if (WARN(!irqs_disabled(),
 924                  "Interrupts were enabled *very* early, fixing it\n"))
 925                 local_irq_disable();
 926         radix_tree_init();
 927
 928         /*
 929          * Set up housekeeping before setting up workqueues to allow the unbound
 930          * workqueue to take non-housekeeping into account.
 931          */
 932         housekeeping_init();
 933
 934         /*
 935          * Allow workqueue creation and work item queueing/cancelling
 936          * early.  Work item execution depends on kthreads and starts after
 937          * workqueue_init().
 938          */
 939         workqueue_init_early();
 940
 941         rcu_init();
 942
 943         /* Trace events are available after this */
 944         trace_init();
 945
 946         if (initcall_debug)
 947                 initcall_debug_enable();
 948
 949         context_tracking_init();
 950         /* init some links before init_ISA_irqs() */
 951         early_irq_init();
 952         init_IRQ();
 953         tick_init();
 954         rcu_init_nohz();
 955         init_timers();
 956         hrtimers_init();
 957         softirq_init();
 958         timekeeping_init();
 959
 960         /*
 961          * For best initial stack canary entropy, prepare it after:
 962          * - setup_arch() for any UEFI RNG entropy and boot cmdline access
 963          * - timekeeping_init() for ktime entropy used in rand_initialize()
 964          * - rand_initialize() to get any arch-specific entropy like RDRAND
 965          * - add_latent_entropy() to get any latent entropy
 966          * - adding command line entropy
 967          */
 968         rand_initialize();
 969         add_latent_entropy();
 970         add_device_randomness(command_line, strlen(command_line));
 971         boot_init_stack_canary();
 972
 973         time_init();
 974         perf_event_init();
 975         profile_init();
 976         call_function_init();
 977         WARN(!irqs_disabled(), "Interrupts were enabled early\n");
 978
 979         early_boot_irqs_disabled = false;
 980         local_irq_enable();
 981
 982         kmem_cache_init_late();
 983
 984         /*
 985          * HACK ALERT! This is early. We're enabling the console before
 986          * we've done PCI setups etc, and console_init() must be aware of
 987          * this. But we do want output early, in case something goes wrong.
 988          */
 989         console_init();
 990         if (panic_later)
 991                 panic("Too many boot %s vars at `%s'", panic_later,
 992                       panic_param);
 993
 994         lockdep_init();
 995
 996         /*
 997          * Need to run this when irqs are enabled, because it wants
 998          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 999          * too:
1000          */
1001         locking_selftest();
1002
1003         /*
1004          * This needs to be called before any devices perform DMA
1005          * operations that might use the SWIOTLB bounce buffers. It will
1006          * mark the bounce buffers as decrypted so that their usage will
1007          * not cause "plain-text" data to be decrypted when accessed.
1008          */
1009         mem_encrypt_init();
1010
1011 #ifdef CONFIG_BLK_DEV_INITRD
1012         if (initrd_start && !initrd_below_start_ok &&
1013             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1014                 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1015                     page_to_pfn(virt_to_page((void *)initrd_start)),
1016                     min_low_pfn);
1017                 initrd_start = 0;
1018         }
1019 #endif
1020         setup_per_cpu_pageset();
1021         numa_policy_init();
1022         acpi_early_init();
1023         if (late_time_init)
1024                 late_time_init();
1025         sched_clock_init();
1026         calibrate_delay();
1027         pid_idr_init();
1028         anon_vma_init();
1029 #ifdef CONFIG_X86
1030         if (efi_enabled(EFI_RUNTIME_SERVICES))
1031                 efi_enter_virtual_mode();
1032 #endif
1033         thread_stack_cache_init();
1034         cred_init();
1035         fork_init();
1036         proc_caches_init();
1037         uts_ns_init();
1038         key_init();
1039         security_init();
1040         dbg_late_init();
1041         vfs_caches_init();
1042         pagecache_init();
1043         signals_init();
1044         seq_file_init();
1045         proc_root_init();
1046         nsfs_init();
1047         cpuset_init();
1048         cgroup_init();
1049         taskstats_init_early();
1050         delayacct_init();
1051
1052         poking_init();
1053         check_bugs();
1054
1055         acpi_subsystem_init();
1056         arch_post_acpi_subsys_init();
1057         sfi_init_late();
1058         kcsan_init();
1059
1060         /* Do the rest non-__init'ed, we're now alive */
1061         arch_call_rest_init();
1062
1063         prevent_tail_call_optimization();
1064 }
1065
1066 /* Call all constructor functions linked into the kernel. */
1067 static void __init do_ctors(void)
1068 {
1069 #ifdef CONFIG_CONSTRUCTORS
1070         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1071
1072         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1073                 (*fn)();
1074 #endif
1075 }
1076
1077 #ifdef CONFIG_KALLSYMS
1078 struct blacklist_entry {
1079         struct list_head next;
1080         char *buf;
1081 };
1082
1083 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1084
1085 static int __init initcall_blacklist(char *str)
1086 {
1087         char *str_entry;
1088         struct blacklist_entry *entry;
1089
1090         /* str argument is a comma-separated list of functions */
1091         do {
1092                 str_entry = strsep(&str, ",");
1093                 if (str_entry) {
1094                         pr_debug("blacklisting initcall %s\n", str_entry);
1095                         entry = memblock_alloc(sizeof(*entry),
1096                                                SMP_CACHE_BYTES);
1097                         if (!entry)
1098                                 panic("%s: Failed to allocate %zu bytes\n",
1099                                       __func__, sizeof(*entry));
1100                         entry->buf = memblock_alloc(strlen(str_entry) + 1,
1101                                                     SMP_CACHE_BYTES);
1102                         if (!entry->buf)
1103                                 panic("%s: Failed to allocate %zu bytes\n",
1104                                       __func__, strlen(str_entry) + 1);
1105                         strcpy(entry->buf, str_entry);
1106                         list_add(&entry->next, &blacklisted_initcalls);
1107                 }
1108         } while (str_entry);
1109
1110         return 0;
1111 }
1112
1113 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1114 {
1115         struct blacklist_entry *entry;
1116         char fn_name[KSYM_SYMBOL_LEN];
1117         unsigned long addr;
1118
1119         if (list_empty(&blacklisted_initcalls))
1120                 return false;
1121
1122         addr = (unsigned long) dereference_function_descriptor(fn);
1123         sprint_symbol_no_offset(fn_name, addr);
1124
1125         /*
1126          * fn will be "function_name [module_name]" where [module_name] is not
1127          * displayed for built-in init functions.  Strip off the [module_name].
1128          */
1129         strreplace(fn_name, ' ', '\0');
1130
1131         list_for_each_entry(entry, &blacklisted_initcalls, next) {
1132                 if (!strcmp(fn_name, entry->buf)) {
1133                         pr_debug("initcall %s blacklisted\n", fn_name);
1134                         return true;
1135                 }
1136         }
1137
1138         return false;
1139 }
1140 #else
1141 static int __init initcall_blacklist(char *str)
1142 {
1143         pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1144         return 0;
1145 }
1146
1147 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1148 {
1149         return false;
1150 }
1151 #endif
1152 __setup("initcall_blacklist=", initcall_blacklist);
1153
1154 static __init_or_module void
1155 trace_initcall_start_cb(void *data, initcall_t fn)
1156 {
1157         ktime_t *calltime = (ktime_t *)data;
1158
1159         printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1160         *calltime = ktime_get();
1161 }
1162
1163 static __init_or_module void
1164 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1165 {
1166         ktime_t *calltime = (ktime_t *)data;
1167         ktime_t delta, rettime;
1168         unsigned long long duration;
1169
1170         rettime = ktime_get();
1171         delta = ktime_sub(rettime, *calltime);
1172         duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1173         printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1174                  fn, ret, duration);
1175 }
1176
1177 static ktime_t initcall_calltime;
1178
1179 #ifdef TRACEPOINTS_ENABLED
1180 static void __init initcall_debug_enable(void)
1181 {
1182         int ret;
1183
1184         ret = register_trace_initcall_start(trace_initcall_start_cb,
1185                                             &initcall_calltime);
1186         ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1187                                               &initcall_calltime);
1188         WARN(ret, "Failed to register initcall tracepoints\n");
1189 }
1190 # define do_trace_initcall_start        trace_initcall_start
1191 # define do_trace_initcall_finish       trace_initcall_finish
1192 #else
1193 static inline void do_trace_initcall_start(initcall_t fn)
1194 {
1195         if (!initcall_debug)
1196                 return;
1197         trace_initcall_start_cb(&initcall_calltime, fn);
1198 }
1199 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1200 {
1201         if (!initcall_debug)
1202                 return;
1203         trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1204 }
1205 #endif /* !TRACEPOINTS_ENABLED */
1206
1207 int __init_or_module do_one_initcall(initcall_t fn)
1208 {
1209         int count = preempt_count();
1210         char msgbuf[64];
1211         int ret;
1212
1213         if (initcall_blacklisted(fn))
1214                 return -EPERM;
1215
1216         do_trace_initcall_start(fn);
1217         ret = fn();
1218         do_trace_initcall_finish(fn, ret);
1219
1220         msgbuf[0] = 0;
1221
1222         if (preempt_count() != count) {
1223                 sprintf(msgbuf, "preemption imbalance ");
1224                 preempt_count_set(count);
1225         }
1226         if (irqs_disabled()) {
1227                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1228                 local_irq_enable();
1229         }
1230         WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1231
1232         add_latent_entropy();
1233         return ret;
1234 }
1235
1236
1237 extern initcall_entry_t __initcall_start[];
1238 extern initcall_entry_t __initcall0_start[];
1239 extern initcall_entry_t __initcall1_start[];
1240 extern initcall_entry_t __initcall2_start[];
1241 extern initcall_entry_t __initcall3_start[];
1242 extern initcall_entry_t __initcall4_start[];
1243 extern initcall_entry_t __initcall5_start[];
1244 extern initcall_entry_t __initcall6_start[];
1245 extern initcall_entry_t __initcall7_start[];
1246 extern initcall_entry_t __initcall_end[];
1247
1248 static initcall_entry_t *initcall_levels[] __initdata = {
1249         __initcall0_start,
1250         __initcall1_start,
1251         __initcall2_start,
1252         __initcall3_start,
1253         __initcall4_start,
1254         __initcall5_start,
1255         __initcall6_start,
1256         __initcall7_start,
1257         __initcall_end,
1258 };
1259
1260 /* Keep these in sync with initcalls in include/linux/init.h */
1261 static const char *initcall_level_names[] __initdata = {
1262         "pure",
1263         "core",
1264         "postcore",
1265         "arch",
1266         "subsys",
1267         "fs",
1268         "device",
1269         "late",
1270 };
1271
1272 static int __init ignore_unknown_bootoption(char *param, char *val,
1273                                const char *unused, void *arg)
1274 {
1275         return 0;
1276 }
1277
1278 static void __init do_initcall_level(int level, char *command_line)
1279 {
1280         initcall_entry_t *fn;
1281
1282         parse_args(initcall_level_names[level],
1283                    command_line, __start___param,
1284                    __stop___param - __start___param,
1285                    level, level,
1286                    NULL, ignore_unknown_bootoption);
1287
1288         trace_initcall_level(initcall_level_names[level]);
1289         for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1290                 do_one_initcall(initcall_from_entry(fn));
1291 }
1292
1293 static void __init do_initcalls(void)
1294 {
1295         int level;
1296         size_t len = strlen(saved_command_line) + 1;
1297         char *command_line;
1298
1299         command_line = kzalloc(len, GFP_KERNEL);
1300         if (!command_line)
1301                 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1302
1303         for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1304                 /* Parser modifies command_line, restore it each time */
1305                 strcpy(command_line, saved_command_line);
1306                 do_initcall_level(level, command_line);
1307         }
1308
1309         kfree(command_line);
1310 }
1311
1312 /*
1313  * Ok, the machine is now initialized. None of the devices
1314  * have been touched yet, but the CPU subsystem is up and
1315  * running, and memory and process management works.
1316  *
1317  * Now we can finally start doing some real work..
1318  */
1319 static void __init do_basic_setup(void)
1320 {
1321         cpuset_init_smp();
1322         driver_init();
1323         init_irq_proc();
1324         do_ctors();
1325         usermodehelper_enable();
1326         do_initcalls();
1327 }
1328
1329 static void __init do_pre_smp_initcalls(void)
1330 {
1331         initcall_entry_t *fn;
1332
1333         trace_initcall_level("early");
1334         for (fn = __initcall_start; fn < __initcall0_start; fn++)
1335                 do_one_initcall(initcall_from_entry(fn));
1336 }
1337
1338 static int run_init_process(const char *init_filename)
1339 {
1340         const char *const *p;
1341
1342         argv_init[0] = init_filename;
1343         pr_info("Run %s as init process\n", init_filename);
1344         pr_debug("  with arguments:\n");
1345         for (p = argv_init; *p; p++)
1346                 pr_debug("    %s\n", *p);
1347         pr_debug("  with environment:\n");
1348         for (p = envp_init; *p; p++)
1349                 pr_debug("    %s\n", *p);
1350         return kernel_execve(init_filename, argv_init, envp_init);
1351 }
1352
1353 static int try_to_run_init_process(const char *init_filename)
1354 {
1355         int ret;
1356
1357         ret = run_init_process(init_filename);
1358
1359         if (ret && ret != -ENOENT) {
1360                 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1361                        init_filename, ret);
1362         }
1363
1364         return ret;
1365 }
1366
1367 static noinline void __init kernel_init_freeable(void);
1368
1369 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1370 bool rodata_enabled __ro_after_init = true;
1371 static int __init set_debug_rodata(char *str)
1372 {
1373         return strtobool(str, &rodata_enabled);
1374 }
1375 __setup("rodata=", set_debug_rodata);
1376 #endif
1377
1378 #ifdef CONFIG_STRICT_KERNEL_RWX
1379 static void mark_readonly(void)
1380 {
1381         if (rodata_enabled) {
1382                 /*
1383                  * load_module() results in W+X mappings, which are cleaned
1384                  * up with call_rcu().  Let's make sure that queued work is
1385                  * flushed so that we don't hit false positives looking for
1386                  * insecure pages which are W+X.
1387                  */
1388                 rcu_barrier();
1389                 mark_rodata_ro();
1390                 rodata_test();
1391         } else
1392                 pr_info("Kernel memory protection disabled.\n");
1393 }
1394 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1395 static inline void mark_readonly(void)
1396 {
1397         pr_warn("Kernel memory protection not selected by kernel config.\n");
1398 }
1399 #else
1400 static inline void mark_readonly(void)
1401 {
1402         pr_warn("This architecture does not have kernel memory protection.\n");
1403 }
1404 #endif
1405
1406 void __weak free_initmem(void)
1407 {
1408         free_initmem_default(POISON_FREE_INITMEM);
1409 }
1410
1411 static int __ref kernel_init(void *unused)
1412 {
1413         int ret;
1414
1415         kernel_init_freeable();
1416         /* need to finish all async __init code before freeing the memory */
1417         async_synchronize_full();
1418         kprobe_free_init_mem();
1419         ftrace_free_init_mem();
1420         free_initmem();
1421         mark_readonly();
1422
1423         /*
1424          * Kernel mappings are now finalized - update the userspace page-table
1425          * to finalize PTI.
1426          */
1427         pti_finalize();
1428
1429         system_state = SYSTEM_RUNNING;
1430         numa_default_policy();
1431
1432         rcu_end_inkernel_boot();
1433
1434         do_sysctl_args();
1435
1436         if (ramdisk_execute_command) {
1437                 ret = run_init_process(ramdisk_execute_command);
1438                 if (!ret)
1439                         return 0;
1440                 pr_err("Failed to execute %s (error %d)\n",
1441                        ramdisk_execute_command, ret);
1442         }
1443
1444         /*
1445          * We try each of these until one succeeds.
1446          *
1447          * The Bourne shell can be used instead of init if we are
1448          * trying to recover a really broken machine.
1449          */
1450         if (execute_command) {
1451                 ret = run_init_process(execute_command);
1452                 if (!ret)
1453                         return 0;
1454                 panic("Requested init %s failed (error %d).",
1455                       execute_command, ret);
1456         }
1457
1458         if (CONFIG_DEFAULT_INIT[0] != '\0') {
1459                 ret = run_init_process(CONFIG_DEFAULT_INIT);
1460                 if (ret)
1461                         pr_err("Default init %s failed (error %d)\n",
1462                                CONFIG_DEFAULT_INIT, ret);
1463                 else
1464                         return 0;
1465         }
1466
1467         if (!try_to_run_init_process("/sbin/init") ||
1468             !try_to_run_init_process("/etc/init") ||
1469             !try_to_run_init_process("/bin/init") ||
1470             !try_to_run_init_process("/bin/sh"))
1471                 return 0;
1472
1473         panic("No working init found.  Try passing init= option to kernel. "
1474               "See Linux Documentation/admin-guide/init.rst for guidance.");
1475 }
1476
1477 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1478 void __init console_on_rootfs(void)
1479 {
1480         struct file *file = filp_open("/dev/console", O_RDWR, 0);
1481
1482         if (IS_ERR(file)) {
1483                 pr_err("Warning: unable to open an initial console. Fallback to ttynull.\n");
1484                 register_ttynull_console();
1485
1486                 file = filp_open("/dev/console", O_RDWR, 0);
1487                 if (IS_ERR(file)) {
1488                         pr_err("Warning: Failed to add ttynull console. No stdin, stdout, and stderr for the init process!\n");
1489                         return;
1490                 }
1491         }
1492         init_dup(file);
1493         init_dup(file);
1494         init_dup(file);
1495         fput(file);
1496 }
1497
1498 static noinline void __init kernel_init_freeable(void)
1499 {
1500         /*
1501          * Wait until kthreadd is all set-up.
1502          */
1503         wait_for_completion(&kthreadd_done);
1504
1505         /* Now the scheduler is fully set up and can do blocking allocations */
1506         gfp_allowed_mask = __GFP_BITS_MASK;
1507
1508         /*
1509          * init can allocate pages on any node
1510          */
1511         set_mems_allowed(node_states[N_MEMORY]);
1512
1513         cad_pid = task_pid(current);
1514
1515         smp_prepare_cpus(setup_max_cpus);
1516
1517         workqueue_init();
1518
1519         init_mm_internals();
1520
1521         rcu_init_tasks_generic();
1522         do_pre_smp_initcalls();
1523         lockup_detector_init();
1524
1525         smp_init();
1526         sched_init_smp();
1527
1528         padata_init();
1529         page_alloc_init_late();
1530         /* Initialize page ext after all struct pages are initialized. */
1531         page_ext_init();
1532
1533         do_basic_setup();
1534
1535         kunit_run_all_tests();
1536
1537         console_on_rootfs();
1538
1539         /*
1540          * check if there is an early userspace init.  If yes, let it do all
1541          * the work
1542          */
1543         if (init_eaccess(ramdisk_execute_command) != 0) {
1544                 ramdisk_execute_command = NULL;
1545                 prepare_namespace();
1546         }
1547
1548         /*
1549          * Ok, we have completed the initial bootup, and
1550          * we're essentially up and running. Get rid of the
1551          * initmem segments and start the user-mode stuff..
1552          *
1553          * rootfs is available now, try loading the public keys
1554          * and default modules
1555          */
1556
1557         integrity_load_keys();
1558 }