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/kfence.h>
  44 #include <linux/rcupdate.h>
  45 #include <linux/moduleparam.h>
  46 #include <linux/kallsyms.h>
  47 #include <linux/writeback.h>
  48 #include <linux/cpu.h>
  49 #include <linux/cpuset.h>
  50 #include <linux/cgroup.h>
  51 #include <linux/efi.h>
  52 #include <linux/tick.h>
  53 #include <linux/sched/isolation.h>
  54 #include <linux/interrupt.h>
  55 #include <linux/taskstats_kern.h>
  56 #include <linux/delayacct.h>
  57 #include <linux/unistd.h>
  58 #include <linux/utsname.h>
  59 #include <linux/rmap.h>
  60 #include <linux/mempolicy.h>
  61 #include <linux/key.h>
  62 #include <linux/page_ext.h>
  63 #include <linux/debug_locks.h>
  64 #include <linux/debugobjects.h>
  65 #include <linux/lockdep.h>
  66 #include <linux/kmemleak.h>
  67 #include <linux/padata.h>
  68 #include <linux/pid_namespace.h>
  69 #include <linux/device/driver.h>
  70 #include <linux/kthread.h>
  71 #include <linux/sched.h>
  72 #include <linux/sched/init.h>
  73 #include <linux/signal.h>
  74 #include <linux/idr.h>
  75 #include <linux/kgdb.h>
  76 #include <linux/ftrace.h>
  77 #include <linux/async.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 #include <linux/stackdepot.h>
 101
 102 #include <asm/io.h>
 103 #include <asm/bugs.h>
 104 #include <asm/setup.h>
 105 #include <asm/sections.h>
 106 #include <asm/cacheflush.h>
 107
 108 #define CREATE_TRACE_POINTS
 109 #include <trace/events/initcall.h>
 110
 111 #include <kunit/test.h>
 112
 113 static int kernel_init(void *);
 114
 115 extern void init_IRQ(void);
 116 extern void radix_tree_init(void);
 117
 118 /*
 119  * Debug helper: via this flag we know that we are in 'early bootup code'
 120  * where only the boot processor is running with IRQ disabled.  This means
 121  * two things - IRQ must not be enabled before the flag is cleared and some
 122  * operations which are not allowed with IRQ disabled are allowed while the
 123  * flag is set.
 124  */
 125 bool early_boot_irqs_disabled __read_mostly;
 126
 127 enum system_states system_state __read_mostly;
 128 EXPORT_SYMBOL(system_state);
 129
 130 /*
 131  * Boot command-line arguments
 132  */
 133 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 134 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 135
 136 extern void time_init(void);
 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;
 144 /* Command line for parameter parsing */
 145 static char *static_command_line;
 146 /* Untouched extra command line */
 147 static char *extra_command_line;
 148 /* Extra init arguments */
 149 static char *extra_init_args;
 150
 151 #ifdef CONFIG_BOOT_CONFIG
 152 /* Is bootconfig on command line? */
 153 static bool bootconfig_found;
 154 static bool initargs_found;
 155 #else
 156 # define bootconfig_found false
 157 # define initargs_found false
 158 #endif
 159
 160 static char *execute_command;
 161 static char *ramdisk_execute_command = "/init";
 162
 163 /*
 164  * Used to generate warnings if static_key manipulation functions are used
 165  * before jump_label_init is called.
 166  */
 167 bool static_key_initialized __read_mostly;
 168 EXPORT_SYMBOL_GPL(static_key_initialized);
 169
 170 /*
 171  * If set, this is an indication to the drivers that reset the underlying
 172  * device before going ahead with the initialization otherwise driver might
 173  * rely on the BIOS and skip the reset operation.
 174  *
 175  * This is useful if kernel is booting in an unreliable environment.
 176  * For ex. kdump situation where previous kernel has crashed, BIOS has been
 177  * skipped and devices will be in unknown state.
 178  */
 179 unsigned int reset_devices;
 180 EXPORT_SYMBOL(reset_devices);
 181
 182 static int __init set_reset_devices(char *str)
 183 {
 184         reset_devices = 1;
 185         return 1;
 186 }
 187
 188 __setup("reset_devices", set_reset_devices);
 189
 190 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 191 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 192 static const char *panic_later, *panic_param;
 193
 194 extern const struct obs_kernel_param __setup_start[], __setup_end[];
 195
 196 static bool __init obsolete_checksetup(char *line)
 197 {
 198         const struct obs_kernel_param *p;
 199         bool had_early_param = false;
 200
 201         p = __setup_start;
 202         do {
 203                 int n = strlen(p->str);
 204                 if (parameqn(line, p->str, n)) {
 205                         if (p->early) {
 206                                 /* Already done in parse_early_param?
 207                                  * (Needs exact match on param part).
 208                                  * Keep iterating, as we can have early
 209                                  * params and __setups of same names 8( */
 210                                 if (line[n] == '\0' || line[n] == '=')
 211                                         had_early_param = true;
 212                         } else if (!p->setup_func) {
 213                                 pr_warn("Parameter %s is obsolete, ignored\n",
 214                                         p->str);
 215                                 return true;
 216                         } else if (p->setup_func(line + n))
 217                                 return true;
 218                 }
 219                 p++;
 220         } while (p < __setup_end);
 221
 222         return had_early_param;
 223 }
 224
 225 /*
 226  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 227  * still work even if initially too large, it will just take slightly longer
 228  */
 229 unsigned long loops_per_jiffy = (1<<12);
 230 EXPORT_SYMBOL(loops_per_jiffy);
 231
 232 static int __init debug_kernel(char *str)
 233 {
 234         console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
 235         return 0;
 236 }
 237
 238 static int __init quiet_kernel(char *str)
 239 {
 240         console_loglevel = CONSOLE_LOGLEVEL_QUIET;
 241         return 0;
 242 }
 243
 244 early_param("debug", debug_kernel);
 245 early_param("quiet", quiet_kernel);
 246
 247 static int __init loglevel(char *str)
 248 {
 249         int newlevel;
 250
 251         /*
 252          * Only update loglevel value when a correct setting was passed,
 253          * to prevent blind crashes (when loglevel being set to 0) that
 254          * are quite hard to debug
 255          */
 256         if (get_option(&str, &newlevel)) {
 257                 console_loglevel = newlevel;
 258                 return 0;
 259         }
 260
 261         return -EINVAL;
 262 }
 263
 264 early_param("loglevel", loglevel);
 265
 266 #ifdef CONFIG_BLK_DEV_INITRD
 267 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
 268 {
 269         u32 size, csum;
 270         char *data;
 271         u32 *hdr;
 272         int i;
 273
 274         if (!initrd_end)
 275                 return NULL;
 276
 277         data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
 278         /*
 279          * Since Grub may align the size of initrd to 4, we must
 280          * check the preceding 3 bytes as well.
 281          */
 282         for (i = 0; i < 4; i++) {
 283                 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
 284                         goto found;
 285                 data--;
 286         }
 287         return NULL;
 288
 289 found:
 290         hdr = (u32 *)(data - 8);
 291         size = le32_to_cpu(hdr[0]);
 292         csum = le32_to_cpu(hdr[1]);
 293
 294         data = ((void *)hdr) - size;
 295         if ((unsigned long)data < initrd_start) {
 296                 pr_err("bootconfig size %d is greater than initrd size %ld\n",
 297                         size, initrd_end - initrd_start);
 298                 return NULL;
 299         }
 300
 301         /* Remove bootconfig from initramfs/initrd */
 302         initrd_end = (unsigned long)data;
 303         if (_size)
 304                 *_size = size;
 305         if (_csum)
 306                 *_csum = csum;
 307
 308         return data;
 309 }
 310 #else
 311 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
 312 {
 313         return NULL;
 314 }
 315 #endif
 316
 317 #ifdef CONFIG_BOOT_CONFIG
 318
 319 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
 320
 321 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
 322
 323 static int __init xbc_snprint_cmdline(char *buf, size_t size,
 324                                       struct xbc_node *root)
 325 {
 326         struct xbc_node *knode, *vnode;
 327         char *end = buf + size;
 328         const char *val;
 329         int ret;
 330
 331         xbc_node_for_each_key_value(root, knode, val) {
 332                 ret = xbc_node_compose_key_after(root, knode,
 333                                         xbc_namebuf, XBC_KEYLEN_MAX);
 334                 if (ret < 0)
 335                         return ret;
 336
 337                 vnode = xbc_node_get_child(knode);
 338                 if (!vnode) {
 339                         ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
 340                         if (ret < 0)
 341                                 return ret;
 342                         buf += ret;
 343                         continue;
 344                 }
 345                 xbc_array_for_each_value(vnode, val) {
 346                         ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
 347                                        xbc_namebuf, val);
 348                         if (ret < 0)
 349                                 return ret;
 350                         buf += ret;
 351                 }
 352         }
 353
 354         return buf - (end - size);
 355 }
 356 #undef rest
 357
 358 /* Make an extra command line under given key word */
 359 static char * __init xbc_make_cmdline(const char *key)
 360 {
 361         struct xbc_node *root;
 362         char *new_cmdline;
 363         int ret, len = 0;
 364
 365         root = xbc_find_node(key);
 366         if (!root)
 367                 return NULL;
 368
 369         /* Count required buffer size */
 370         len = xbc_snprint_cmdline(NULL, 0, root);
 371         if (len <= 0)
 372                 return NULL;
 373
 374         new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
 375         if (!new_cmdline) {
 376                 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
 377                 return NULL;
 378         }
 379
 380         ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
 381         if (ret < 0 || ret > len) {
 382                 pr_err("Failed to print extra kernel cmdline.\n");
 383                 return NULL;
 384         }
 385
 386         return new_cmdline;
 387 }
 388
 389 static u32 boot_config_checksum(unsigned char *p, u32 size)
 390 {
 391         u32 ret = 0;
 392
 393         while (size--)
 394                 ret += *p++;
 395
 396         return ret;
 397 }
 398
 399 static int __init bootconfig_params(char *param, char *val,
 400                                     const char *unused, void *arg)
 401 {
 402         if (strcmp(param, "bootconfig") == 0) {
 403                 bootconfig_found = true;
 404         }
 405         return 0;
 406 }
 407
 408 static void __init setup_boot_config(const char *cmdline)
 409 {
 410         static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
 411         const char *msg;
 412         int pos;
 413         u32 size, csum;
 414         char *data, *copy, *err;
 415         int ret;
 416
 417         /* Cut out the bootconfig data even if we have no bootconfig option */
 418         data = get_boot_config_from_initrd(&size, &csum);
 419
 420         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 421         err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
 422                          bootconfig_params);
 423
 424         if (IS_ERR(err) || !bootconfig_found)
 425                 return;
 426
 427         /* parse_args() stops at '--' and returns an address */
 428         if (err)
 429                 initargs_found = true;
 430
 431         if (!data) {
 432                 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
 433                 return;
 434         }
 435
 436         if (size >= XBC_DATA_MAX) {
 437                 pr_err("bootconfig size %d greater than max size %d\n",
 438                         size, XBC_DATA_MAX);
 439                 return;
 440         }
 441
 442         if (boot_config_checksum((unsigned char *)data, size) != csum) {
 443                 pr_err("bootconfig checksum failed\n");
 444                 return;
 445         }
 446
 447         copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
 448         if (!copy) {
 449                 pr_err("Failed to allocate memory for bootconfig\n");
 450                 return;
 451         }
 452
 453         memcpy(copy, data, size);
 454         copy[size] = '\0';
 455
 456         ret = xbc_init(copy, &msg, &pos);
 457         if (ret < 0) {
 458                 if (pos < 0)
 459                         pr_err("Failed to init bootconfig: %s.\n", msg);
 460                 else
 461                         pr_err("Failed to parse bootconfig: %s at %d.\n",
 462                                 msg, pos);
 463         } else {
 464                 pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
 465                 /* keys starting with "kernel." are passed via cmdline */
 466                 extra_command_line = xbc_make_cmdline("kernel");
 467                 /* Also, "init." keys are init arguments */
 468                 extra_init_args = xbc_make_cmdline("init");
 469         }
 470         return;
 471 }
 472
 473 #else
 474
 475 static void __init setup_boot_config(const char *cmdline)
 476 {
 477         /* Remove bootconfig data from initrd */
 478         get_boot_config_from_initrd(NULL, NULL);
 479 }
 480
 481 static int __init warn_bootconfig(char *str)
 482 {
 483         pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
 484         return 0;
 485 }
 486 early_param("bootconfig", warn_bootconfig);
 487
 488 #endif
 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         repair_env_string(param, val);
 537
 538         /* Handle obsolete-style parameters */
 539         if (obsolete_checksetup(param))
 540                 return 0;
 541
 542         /* Unused module parameter. */
 543         if (strnchr(param, len, '.'))
 544                 return 0;
 545
 546         if (panic_later)
 547                 return 0;
 548
 549         if (val) {
 550                 /* Environment option */
 551                 unsigned int i;
 552                 for (i = 0; envp_init[i]; i++) {
 553                         if (i == MAX_INIT_ENVS) {
 554                                 panic_later = "env";
 555                                 panic_param = param;
 556                         }
 557                         if (!strncmp(param, envp_init[i], len+1))
 558                                 break;
 559                 }
 560                 envp_init[i] = param;
 561         } else {
 562                 /* Command line option */
 563                 unsigned int i;
 564                 for (i = 0; argv_init[i]; i++) {
 565                         if (i == MAX_INIT_ARGS) {
 566                                 panic_later = "init";
 567                                 panic_param = param;
 568                         }
 569                 }
 570                 argv_init[i] = param;
 571         }
 572         return 0;
 573 }
 574
 575 static int __init init_setup(char *str)
 576 {
 577         unsigned int i;
 578
 579         execute_command = str;
 580         /*
 581          * In case LILO is going to boot us with default command line,
 582          * it prepends "auto" before the whole cmdline which makes
 583          * the shell think it should execute a script with such name.
 584          * So we ignore all arguments entered _before_ init=... [MJ]
 585          */
 586         for (i = 1; i < MAX_INIT_ARGS; i++)
 587                 argv_init[i] = NULL;
 588         return 1;
 589 }
 590 __setup("init=", init_setup);
 591
 592 static int __init rdinit_setup(char *str)
 593 {
 594         unsigned int i;
 595
 596         ramdisk_execute_command = str;
 597         /* See "auto" comment in init_setup */
 598         for (i = 1; i < MAX_INIT_ARGS; i++)
 599                 argv_init[i] = NULL;
 600         return 1;
 601 }
 602 __setup("rdinit=", rdinit_setup);
 603
 604 #ifndef CONFIG_SMP
 605 static const unsigned int setup_max_cpus = NR_CPUS;
 606 static inline void setup_nr_cpu_ids(void) { }
 607 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 608 #endif
 609
 610 /*
 611  * We need to store the untouched command line for future reference.
 612  * We also need to store the touched command line since the parameter
 613  * parsing is performed in place, and we should allow a component to
 614  * store reference of name/value for future reference.
 615  */
 616 static void __init setup_command_line(char *command_line)
 617 {
 618         size_t len, xlen = 0, ilen = 0;
 619
 620         if (extra_command_line)
 621                 xlen = strlen(extra_command_line);
 622         if (extra_init_args)
 623                 ilen = strlen(extra_init_args) + 4; /* for " -- " */
 624
 625         len = xlen + strlen(boot_command_line) + 1;
 626
 627         saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
 628         if (!saved_command_line)
 629                 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
 630
 631         static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
 632         if (!static_command_line)
 633                 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
 634
 635         if (xlen) {
 636                 /*
 637                  * We have to put extra_command_line before boot command
 638                  * lines because there could be dashes (separator of init
 639                  * command line) in the command lines.
 640                  */
 641                 strcpy(saved_command_line, extra_command_line);
 642                 strcpy(static_command_line, extra_command_line);
 643         }
 644         strcpy(saved_command_line + xlen, boot_command_line);
 645         strcpy(static_command_line + xlen, command_line);
 646
 647         if (ilen) {
 648                 /*
 649                  * Append supplemental init boot args to saved_command_line
 650                  * so that user can check what command line options passed
 651                  * to init.
 652                  */
 653                 len = strlen(saved_command_line);
 654                 if (initargs_found) {
 655                         saved_command_line[len++] = ' ';
 656                 } else {
 657                         strcpy(saved_command_line + len, " -- ");
 658                         len += 4;
 659                 }
 660
 661                 strcpy(saved_command_line + len, extra_init_args);
 662         }
 663 }
 664
 665 /*
 666  * We need to finalize in a non-__init function or else race conditions
 667  * between the root thread and the init thread may cause start_kernel to
 668  * be reaped by free_initmem before the root thread has proceeded to
 669  * cpu_idle.
 670  *
 671  * gcc-3.4 accidentally inlines this function, so use noinline.
 672  */
 673
 674 static __initdata DECLARE_COMPLETION(kthreadd_done);
 675
 676 noinline void __ref rest_init(void)
 677 {
 678         struct task_struct *tsk;
 679         int pid;
 680
 681         rcu_scheduler_starting();
 682         /*
 683          * We need to spawn init first so that it obtains pid 1, however
 684          * the init task will end up wanting to create kthreads, which, if
 685          * we schedule it before we create kthreadd, will OOPS.
 686          */
 687         pid = kernel_thread(kernel_init, NULL, CLONE_FS);
 688         /*
 689          * Pin init on the boot CPU. Task migration is not properly working
 690          * until sched_init_smp() has been run. It will set the allowed
 691          * CPUs for init to the non isolated CPUs.
 692          */
 693         rcu_read_lock();
 694         tsk = find_task_by_pid_ns(pid, &init_pid_ns);
 695         set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
 696         rcu_read_unlock();
 697
 698         numa_default_policy();
 699         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 700         rcu_read_lock();
 701         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 702         rcu_read_unlock();
 703
 704         /*
 705          * Enable might_sleep() and smp_processor_id() checks.
 706          * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
 707          * kernel_thread() would trigger might_sleep() splats. With
 708          * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
 709          * already, but it's stuck on the kthreadd_done completion.
 710          */
 711         system_state = SYSTEM_SCHEDULING;
 712
 713         complete(&kthreadd_done);
 714
 715         /*
 716          * The boot idle thread must execute schedule()
 717          * at least once to get things moving:
 718          */
 719         schedule_preempt_disabled();
 720         /* Call into cpu_idle with preempt disabled */
 721         cpu_startup_entry(CPUHP_ONLINE);
 722 }
 723
 724 /* Check for early params. */
 725 static int __init do_early_param(char *param, char *val,
 726                                  const char *unused, void *arg)
 727 {
 728         const struct obs_kernel_param *p;
 729
 730         for (p = __setup_start; p < __setup_end; p++) {
 731                 if ((p->early && parameq(param, p->str)) ||
 732                     (strcmp(param, "console") == 0 &&
 733                      strcmp(p->str, "earlycon") == 0)
 734                 ) {
 735                         if (p->setup_func(val) != 0)
 736                                 pr_warn("Malformed early option '%s'\n", param);
 737                 }
 738         }
 739         /* We accept everything at this stage. */
 740         return 0;
 741 }
 742
 743 void __init parse_early_options(char *cmdline)
 744 {
 745         parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
 746                    do_early_param);
 747 }
 748
 749 /* Arch code calls this early on, or if not, just before other parsing. */
 750 void __init parse_early_param(void)
 751 {
 752         static int done __initdata;
 753         static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
 754
 755         if (done)
 756                 return;
 757
 758         /* All fall through to do_early_param. */
 759         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 760         parse_early_options(tmp_cmdline);
 761         done = 1;
 762 }
 763
 764 void __init __weak arch_post_acpi_subsys_init(void) { }
 765
 766 void __init __weak smp_setup_processor_id(void)
 767 {
 768 }
 769
 770 # if THREAD_SIZE >= PAGE_SIZE
 771 void __init __weak thread_stack_cache_init(void)
 772 {
 773 }
 774 #endif
 775
 776 void __init __weak mem_encrypt_init(void) { }
 777
 778 void __init __weak poking_init(void) { }
 779
 780 void __init __weak pgtable_cache_init(void) { }
 781
 782 bool initcall_debug;
 783 core_param(initcall_debug, initcall_debug, bool, 0644);
 784
 785 #ifdef TRACEPOINTS_ENABLED
 786 static void __init initcall_debug_enable(void);
 787 #else
 788 static inline void initcall_debug_enable(void)
 789 {
 790 }
 791 #endif
 792
 793 /* Report memory auto-initialization states for this boot. */
 794 static void __init report_meminit(void)
 795 {
 796         const char *stack;
 797
 798         if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
 799                 stack = "all(pattern)";
 800         else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
 801                 stack = "all(zero)";
 802         else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
 803                 stack = "byref_all(zero)";
 804         else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
 805                 stack = "byref(zero)";
 806         else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
 807                 stack = "__user(zero)";
 808         else
 809                 stack = "off";
 810
 811         pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
 812                 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
 813                 want_init_on_free() ? "on" : "off");
 814         if (want_init_on_free())
 815                 pr_info("mem auto-init: clearing system memory may take some time...\n");
 816 }
 817
 818 /*
 819  * Set up kernel memory allocators
 820  */
 821 static void __init mm_init(void)
 822 {
 823         /*
 824          * page_ext requires contiguous pages,
 825          * bigger than MAX_ORDER unless SPARSEMEM.
 826          */
 827         page_ext_init_flatmem();
 828         init_mem_debugging_and_hardening();
 829         kfence_alloc_pool();
 830         report_meminit();
 831         stack_depot_init();
 832         mem_init();
 833         /* page_owner must be initialized after buddy is ready */
 834         page_ext_init_flatmem_late();
 835         kmem_cache_init();
 836         kmemleak_init();
 837         pgtable_init();
 838         debug_objects_mem_init();
 839         vmalloc_init();
 840         ioremap_huge_init();
 841         /* Should be run before the first non-init thread is created */
 842         init_espfix_bsp();
 843         /* Should be run after espfix64 is set up. */
 844         pti_init();
 845 }
 846
 847 void __init __weak arch_call_rest_init(void)
 848 {
 849         rest_init();
 850 }
 851
 852 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
 853 {
 854         char *command_line;
 855         char *after_dashes;
 856
 857         set_task_stack_end_magic(&init_task);
 858         smp_setup_processor_id();
 859         debug_objects_early_init();
 860
 861         cgroup_init_early();
 862
 863         local_irq_disable();
 864         early_boot_irqs_disabled = true;
 865
 866         /*
 867          * Interrupts are still disabled. Do necessary setups, then
 868          * enable them.
 869          */
 870         boot_cpu_init();
 871         page_address_init();
 872         pr_notice("%s", linux_banner);
 873         early_security_init();
 874         setup_arch(&command_line);
 875         setup_boot_config(command_line);
 876         setup_command_line(command_line);
 877         setup_nr_cpu_ids();
 878         setup_per_cpu_areas();
 879         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 880         boot_cpu_hotplug_init();
 881
 882         build_all_zonelists(NULL);
 883         page_alloc_init();
 884
 885         pr_notice("Kernel command line: %s\n", saved_command_line);
 886         /* parameters may set static keys */
 887         jump_label_init();
 888         parse_early_param();
 889         after_dashes = parse_args("Booting kernel",
 890                                   static_command_line, __start___param,
 891                                   __stop___param - __start___param,
 892                                   -1, -1, NULL, &unknown_bootoption);
 893         if (!IS_ERR_OR_NULL(after_dashes))
 894                 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
 895                            NULL, set_init_arg);
 896         if (extra_init_args)
 897                 parse_args("Setting extra init args", extra_init_args,
 898                            NULL, 0, -1, -1, NULL, set_init_arg);
 899
 900         /*
 901          * These use large bootmem allocations and must precede
 902          * kmem_cache_init()
 903          */
 904         setup_log_buf(0);
 905         vfs_caches_init_early();
 906         sort_main_extable();
 907         trap_init();
 908         mm_init();
 909
 910         ftrace_init();
 911
 912         /* trace_printk can be enabled here */
 913         early_trace_init();
 914
 915         /*
 916          * Set up the scheduler prior starting any interrupts (such as the
 917          * timer interrupt). Full topology setup happens at smp_init()
 918          * time - but meanwhile we still have a functioning scheduler.
 919          */
 920         sched_init();
 921         /*
 922          * Disable preemption - early bootup scheduling is extremely
 923          * fragile until we cpu_idle() for the first time.
 924          */
 925         preempt_disable();
 926         if (WARN(!irqs_disabled(),
 927                  "Interrupts were enabled *very* early, fixing it\n"))
 928                 local_irq_disable();
 929         radix_tree_init();
 930
 931         /*
 932          * Set up housekeeping before setting up workqueues to allow the unbound
 933          * workqueue to take non-housekeeping into account.
 934          */
 935         housekeeping_init();
 936
 937         /*
 938          * Allow workqueue creation and work item queueing/cancelling
 939          * early.  Work item execution depends on kthreads and starts after
 940          * workqueue_init().
 941          */
 942         workqueue_init_early();
 943
 944         rcu_init();
 945
 946         /* Trace events are available after this */
 947         trace_init();
 948
 949         if (initcall_debug)
 950                 initcall_debug_enable();
 951
 952         context_tracking_init();
 953         /* init some links before init_ISA_irqs() */
 954         early_irq_init();
 955         init_IRQ();
 956         tick_init();
 957         rcu_init_nohz();
 958         init_timers();
 959         hrtimers_init();
 960         softirq_init();
 961         timekeeping_init();
 962         kfence_init();
 963
 964         /*
 965          * For best initial stack canary entropy, prepare it after:
 966          * - setup_arch() for any UEFI RNG entropy and boot cmdline access
 967          * - timekeeping_init() for ktime entropy used in rand_initialize()
 968          * - rand_initialize() to get any arch-specific entropy like RDRAND
 969          * - add_latent_entropy() to get any latent entropy
 970          * - adding command line entropy
 971          */
 972         rand_initialize();
 973         add_latent_entropy();
 974         add_device_randomness(command_line, strlen(command_line));
 975         boot_init_stack_canary();
 976
 977         time_init();
 978         perf_event_init();
 979         profile_init();
 980         call_function_init();
 981         WARN(!irqs_disabled(), "Interrupts were enabled early\n");
 982
 983         early_boot_irqs_disabled = false;
 984         local_irq_enable();
 985
 986         kmem_cache_init_late();
 987
 988         /*
 989          * HACK ALERT! This is early. We're enabling the console before
 990          * we've done PCI setups etc, and console_init() must be aware of
 991          * this. But we do want output early, in case something goes wrong.
 992          */
 993         console_init();
 994         if (panic_later)
 995                 panic("Too many boot %s vars at `%s'", panic_later,
 996                       panic_param);
 997
 998         lockdep_init();
 999
1000         /*
1001          * Need to run this when irqs are enabled, because it wants
1002          * to self-test [hard/soft]-irqs on/off lock inversion bugs
1003          * too:
1004          */
1005         locking_selftest();
1006
1007         /*
1008          * This needs to be called before any devices perform DMA
1009          * operations that might use the SWIOTLB bounce buffers. It will
1010          * mark the bounce buffers as decrypted so that their usage will
1011          * not cause "plain-text" data to be decrypted when accessed.
1012          */
1013         mem_encrypt_init();
1014
1015 #ifdef CONFIG_BLK_DEV_INITRD
1016         if (initrd_start && !initrd_below_start_ok &&
1017             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1018                 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1019                     page_to_pfn(virt_to_page((void *)initrd_start)),
1020                     min_low_pfn);
1021                 initrd_start = 0;
1022         }
1023 #endif
1024         setup_per_cpu_pageset();
1025         numa_policy_init();
1026         acpi_early_init();
1027         if (late_time_init)
1028                 late_time_init();
1029         sched_clock_init();
1030         calibrate_delay();
1031         pid_idr_init();
1032         anon_vma_init();
1033 #ifdef CONFIG_X86
1034         if (efi_enabled(EFI_RUNTIME_SERVICES))
1035                 efi_enter_virtual_mode();
1036 #endif
1037         thread_stack_cache_init();
1038         cred_init();
1039         fork_init();
1040         proc_caches_init();
1041         uts_ns_init();
1042         key_init();
1043         security_init();
1044         dbg_late_init();
1045         vfs_caches_init();
1046         pagecache_init();
1047         signals_init();
1048         seq_file_init();
1049         proc_root_init();
1050         nsfs_init();
1051         cpuset_init();
1052         cgroup_init();
1053         taskstats_init_early();
1054         delayacct_init();
1055
1056         poking_init();
1057         check_bugs();
1058
1059         acpi_subsystem_init();
1060         arch_post_acpi_subsys_init();
1061         kcsan_init();
1062
1063         /* Do the rest non-__init'ed, we're now alive */
1064         arch_call_rest_init();
1065
1066         prevent_tail_call_optimization();
1067 }
1068
1069 /* Call all constructor functions linked into the kernel. */
1070 static void __init do_ctors(void)
1071 {
1072 /*
1073  * For UML, the constructors have already been called by the
1074  * normal setup code as it's just a normal ELF binary, so we
1075  * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1076  * even on UML for modules.
1077  */
1078 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1079         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1080
1081         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1082                 (*fn)();
1083 #endif
1084 }
1085
1086 #ifdef CONFIG_KALLSYMS
1087 struct blacklist_entry {
1088         struct list_head next;
1089         char *buf;
1090 };
1091
1092 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1093
1094 static int __init initcall_blacklist(char *str)
1095 {
1096         char *str_entry;
1097         struct blacklist_entry *entry;
1098
1099         /* str argument is a comma-separated list of functions */
1100         do {
1101                 str_entry = strsep(&str, ",");
1102                 if (str_entry) {
1103                         pr_debug("blacklisting initcall %s\n", str_entry);
1104                         entry = memblock_alloc(sizeof(*entry),
1105                                                SMP_CACHE_BYTES);
1106                         if (!entry)
1107                                 panic("%s: Failed to allocate %zu bytes\n",
1108                                       __func__, sizeof(*entry));
1109                         entry->buf = memblock_alloc(strlen(str_entry) + 1,
1110                                                     SMP_CACHE_BYTES);
1111                         if (!entry->buf)
1112                                 panic("%s: Failed to allocate %zu bytes\n",
1113                                       __func__, strlen(str_entry) + 1);
1114                         strcpy(entry->buf, str_entry);
1115                         list_add(&entry->next, &blacklisted_initcalls);
1116                 }
1117         } while (str_entry);
1118
1119         return 0;
1120 }
1121
1122 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1123 {
1124         struct blacklist_entry *entry;
1125         char fn_name[KSYM_SYMBOL_LEN];
1126         unsigned long addr;
1127
1128         if (list_empty(&blacklisted_initcalls))
1129                 return false;
1130
1131         addr = (unsigned long) dereference_function_descriptor(fn);
1132         sprint_symbol_no_offset(fn_name, addr);
1133
1134         /*
1135          * fn will be "function_name [module_name]" where [module_name] is not
1136          * displayed for built-in init functions.  Strip off the [module_name].
1137          */
1138         strreplace(fn_name, ' ', '\0');
1139
1140         list_for_each_entry(entry, &blacklisted_initcalls, next) {
1141                 if (!strcmp(fn_name, entry->buf)) {
1142                         pr_debug("initcall %s blacklisted\n", fn_name);
1143                         return true;
1144                 }
1145         }
1146
1147         return false;
1148 }
1149 #else
1150 static int __init initcall_blacklist(char *str)
1151 {
1152         pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1153         return 0;
1154 }
1155
1156 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1157 {
1158         return false;
1159 }
1160 #endif
1161 __setup("initcall_blacklist=", initcall_blacklist);
1162
1163 static __init_or_module void
1164 trace_initcall_start_cb(void *data, initcall_t fn)
1165 {
1166         ktime_t *calltime = (ktime_t *)data;
1167
1168         printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1169         *calltime = ktime_get();
1170 }
1171
1172 static __init_or_module void
1173 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1174 {
1175         ktime_t *calltime = (ktime_t *)data;
1176         ktime_t delta, rettime;
1177         unsigned long long duration;
1178
1179         rettime = ktime_get();
1180         delta = ktime_sub(rettime, *calltime);
1181         duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1182         printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1183                  fn, ret, duration);
1184 }
1185
1186 static ktime_t initcall_calltime;
1187
1188 #ifdef TRACEPOINTS_ENABLED
1189 static void __init initcall_debug_enable(void)
1190 {
1191         int ret;
1192
1193         ret = register_trace_initcall_start(trace_initcall_start_cb,
1194                                             &initcall_calltime);
1195         ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1196                                               &initcall_calltime);
1197         WARN(ret, "Failed to register initcall tracepoints\n");
1198 }
1199 # define do_trace_initcall_start        trace_initcall_start
1200 # define do_trace_initcall_finish       trace_initcall_finish
1201 #else
1202 static inline void do_trace_initcall_start(initcall_t fn)
1203 {
1204         if (!initcall_debug)
1205                 return;
1206         trace_initcall_start_cb(&initcall_calltime, fn);
1207 }
1208 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1209 {
1210         if (!initcall_debug)
1211                 return;
1212         trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1213 }
1214 #endif /* !TRACEPOINTS_ENABLED */
1215
1216 int __init_or_module do_one_initcall(initcall_t fn)
1217 {
1218         int count = preempt_count();
1219         char msgbuf[64];
1220         int ret;
1221
1222         if (initcall_blacklisted(fn))
1223                 return -EPERM;
1224
1225         do_trace_initcall_start(fn);
1226         ret = fn();
1227         do_trace_initcall_finish(fn, ret);
1228
1229         msgbuf[0] = 0;
1230
1231         if (preempt_count() != count) {
1232                 sprintf(msgbuf, "preemption imbalance ");
1233                 preempt_count_set(count);
1234         }
1235         if (irqs_disabled()) {
1236                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1237                 local_irq_enable();
1238         }
1239         WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1240
1241         add_latent_entropy();
1242         return ret;
1243 }
1244
1245
1246 extern initcall_entry_t __initcall_start[];
1247 extern initcall_entry_t __initcall0_start[];
1248 extern initcall_entry_t __initcall1_start[];
1249 extern initcall_entry_t __initcall2_start[];
1250 extern initcall_entry_t __initcall3_start[];
1251 extern initcall_entry_t __initcall4_start[];
1252 extern initcall_entry_t __initcall5_start[];
1253 extern initcall_entry_t __initcall6_start[];
1254 extern initcall_entry_t __initcall7_start[];
1255 extern initcall_entry_t __initcall_end[];
1256
1257 static initcall_entry_t *initcall_levels[] __initdata = {
1258         __initcall0_start,
1259         __initcall1_start,
1260         __initcall2_start,
1261         __initcall3_start,
1262         __initcall4_start,
1263         __initcall5_start,
1264         __initcall6_start,
1265         __initcall7_start,
1266         __initcall_end,
1267 };
1268
1269 /* Keep these in sync with initcalls in include/linux/init.h */
1270 static const char *initcall_level_names[] __initdata = {
1271         "pure",
1272         "core",
1273         "postcore",
1274         "arch",
1275         "subsys",
1276         "fs",
1277         "device",
1278         "late",
1279 };
1280
1281 static int __init ignore_unknown_bootoption(char *param, char *val,
1282                                const char *unused, void *arg)
1283 {
1284         return 0;
1285 }
1286
1287 static void __init do_initcall_level(int level, char *command_line)
1288 {
1289         initcall_entry_t *fn;
1290
1291         parse_args(initcall_level_names[level],
1292                    command_line, __start___param,
1293                    __stop___param - __start___param,
1294                    level, level,
1295                    NULL, ignore_unknown_bootoption);
1296
1297         trace_initcall_level(initcall_level_names[level]);
1298         for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1299                 do_one_initcall(initcall_from_entry(fn));
1300 }
1301
1302 static void __init do_initcalls(void)
1303 {
1304         int level;
1305         size_t len = strlen(saved_command_line) + 1;
1306         char *command_line;
1307
1308         command_line = kzalloc(len, GFP_KERNEL);
1309         if (!command_line)
1310                 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1311
1312         for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1313                 /* Parser modifies command_line, restore it each time */
1314                 strcpy(command_line, saved_command_line);
1315                 do_initcall_level(level, command_line);
1316         }
1317
1318         kfree(command_line);
1319 }
1320
1321 /*
1322  * Ok, the machine is now initialized. None of the devices
1323  * have been touched yet, but the CPU subsystem is up and
1324  * running, and memory and process management works.
1325  *
1326  * Now we can finally start doing some real work..
1327  */
1328 static void __init do_basic_setup(void)
1329 {
1330         cpuset_init_smp();
1331         driver_init();
1332         init_irq_proc();
1333         do_ctors();
1334         usermodehelper_enable();
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 static int __init set_debug_rodata(char *str)
1381 {
1382         return strtobool(str, &rodata_enabled);
1383 }
1384 __setup("rodata=", set_debug_rodata);
1385 #endif
1386
1387 #ifdef CONFIG_STRICT_KERNEL_RWX
1388 static void mark_readonly(void)
1389 {
1390         if (rodata_enabled) {
1391                 /*
1392                  * load_module() results in W+X mappings, which are cleaned
1393                  * up with call_rcu().  Let's make sure that queued work is
1394                  * flushed so that we don't hit false positives looking for
1395                  * insecure pages which are W+X.
1396                  */
1397                 rcu_barrier();
1398                 mark_rodata_ro();
1399                 rodata_test();
1400         } else
1401                 pr_info("Kernel memory protection disabled.\n");
1402 }
1403 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1404 static inline void mark_readonly(void)
1405 {
1406         pr_warn("Kernel memory protection not selected by kernel config.\n");
1407 }
1408 #else
1409 static inline void mark_readonly(void)
1410 {
1411         pr_warn("This architecture does not have kernel memory protection.\n");
1412 }
1413 #endif
1414
1415 void __weak free_initmem(void)
1416 {
1417         free_initmem_default(POISON_FREE_INITMEM);
1418 }
1419
1420 static int __ref kernel_init(void *unused)
1421 {
1422         int ret;
1423
1424         kernel_init_freeable();
1425         /* need to finish all async __init code before freeing the memory */
1426         async_synchronize_full();
1427         kprobe_free_init_mem();
1428         ftrace_free_init_mem();
1429         kgdb_free_init_mem();
1430         free_initmem();
1431         mark_readonly();
1432
1433         /*
1434          * Kernel mappings are now finalized - update the userspace page-table
1435          * to finalize PTI.
1436          */
1437         pti_finalize();
1438
1439         system_state = SYSTEM_RUNNING;
1440         numa_default_policy();
1441
1442         rcu_end_inkernel_boot();
1443
1444         do_sysctl_args();
1445
1446         if (ramdisk_execute_command) {
1447                 ret = run_init_process(ramdisk_execute_command);
1448                 if (!ret)
1449                         return 0;
1450                 pr_err("Failed to execute %s (error %d)\n",
1451                        ramdisk_execute_command, ret);
1452         }
1453
1454         /*
1455          * We try each of these until one succeeds.
1456          *
1457          * The Bourne shell can be used instead of init if we are
1458          * trying to recover a really broken machine.
1459          */
1460         if (execute_command) {
1461                 ret = run_init_process(execute_command);
1462                 if (!ret)
1463                         return 0;
1464                 panic("Requested init %s failed (error %d).",
1465                       execute_command, ret);
1466         }
1467
1468         if (CONFIG_DEFAULT_INIT[0] != '\0') {
1469                 ret = run_init_process(CONFIG_DEFAULT_INIT);
1470                 if (ret)
1471                         pr_err("Default init %s failed (error %d)\n",
1472                                CONFIG_DEFAULT_INIT, ret);
1473                 else
1474                         return 0;
1475         }
1476
1477         if (!try_to_run_init_process("/sbin/init") ||
1478             !try_to_run_init_process("/etc/init") ||
1479             !try_to_run_init_process("/bin/init") ||
1480             !try_to_run_init_process("/bin/sh"))
1481                 return 0;
1482
1483         panic("No working init found.  Try passing init= option to kernel. "
1484               "See Linux Documentation/admin-guide/init.rst for guidance.");
1485 }
1486
1487 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1488 void __init console_on_rootfs(void)
1489 {
1490         struct file *file = filp_open("/dev/console", O_RDWR, 0);
1491
1492         if (IS_ERR(file)) {
1493                 pr_err("Warning: unable to open an initial console.\n");
1494                 return;
1495         }
1496         init_dup(file);
1497         init_dup(file);
1498         init_dup(file);
1499         fput(file);
1500 }
1501
1502 static noinline void __init kernel_init_freeable(void)
1503 {
1504         /*
1505          * Wait until kthreadd is all set-up.
1506          */
1507         wait_for_completion(&kthreadd_done);
1508
1509         /* Now the scheduler is fully set up and can do blocking allocations */
1510         gfp_allowed_mask = __GFP_BITS_MASK;
1511
1512         /*
1513          * init can allocate pages on any node
1514          */
1515         set_mems_allowed(node_states[N_MEMORY]);
1516
1517         cad_pid = task_pid(current);
1518
1519         smp_prepare_cpus(setup_max_cpus);
1520
1521         workqueue_init();
1522
1523         init_mm_internals();
1524
1525         rcu_init_tasks_generic();
1526         do_pre_smp_initcalls();
1527         lockup_detector_init();
1528
1529         smp_init();
1530         sched_init_smp();
1531
1532         padata_init();
1533         page_alloc_init_late();
1534         /* Initialize page ext after all struct pages are initialized. */
1535         page_ext_init();
1536
1537         do_basic_setup();
1538
1539         kunit_run_all_tests();
1540
1541         console_on_rootfs();
1542
1543         /*
1544          * check if there is an early userspace init.  If yes, let it do all
1545          * the work
1546          */
1547         if (init_eaccess(ramdisk_execute_command) != 0) {
1548                 ramdisk_execute_command = NULL;
1549                 prepare_namespace();
1550         }
1551
1552         /*
1553          * Ok, we have completed the initial bootup, and
1554          * we're essentially up and running. Get rid of the
1555          * initmem segments and start the user-mode stuff..
1556          *
1557          * rootfs is available now, try loading the public keys
1558          * and default modules
1559          */
1560
1561         integrity_load_keys();
1562 }