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(void)
 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(void)
 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         tsk->flags |= PF_NO_SETAFFINITY;
 696         set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
 697         rcu_read_unlock();
 698
 699         numa_default_policy();
 700         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 701         rcu_read_lock();
 702         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 703         rcu_read_unlock();
 704
 705         /*
 706          * Enable might_sleep() and smp_processor_id() checks.
 707          * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
 708          * kernel_thread() would trigger might_sleep() splats. With
 709          * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
 710          * already, but it's stuck on the kthreadd_done completion.
 711          */
 712         system_state = SYSTEM_SCHEDULING;
 713
 714         complete(&kthreadd_done);
 715
 716         /*
 717          * The boot idle thread must execute schedule()
 718          * at least once to get things moving:
 719          */
 720         schedule_preempt_disabled();
 721         /* Call into cpu_idle with preempt disabled */
 722         cpu_startup_entry(CPUHP_ONLINE);
 723 }
 724
 725 /* Check for early params. */
 726 static int __init do_early_param(char *param, char *val,
 727                                  const char *unused, void *arg)
 728 {
 729         const struct obs_kernel_param *p;
 730
 731         for (p = __setup_start; p < __setup_end; p++) {
 732                 if ((p->early && parameq(param, p->str)) ||
 733                     (strcmp(param, "console") == 0 &&
 734                      strcmp(p->str, "earlycon") == 0)
 735                 ) {
 736                         if (p->setup_func(val) != 0)
 737                                 pr_warn("Malformed early option '%s'\n", param);
 738                 }
 739         }
 740         /* We accept everything at this stage. */
 741         return 0;
 742 }
 743
 744 void __init parse_early_options(char *cmdline)
 745 {
 746         parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
 747                    do_early_param);
 748 }
 749
 750 /* Arch code calls this early on, or if not, just before other parsing. */
 751 void __init parse_early_param(void)
 752 {
 753         static int done __initdata;
 754         static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
 755
 756         if (done)
 757                 return;
 758
 759         /* All fall through to do_early_param. */
 760         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 761         parse_early_options(tmp_cmdline);
 762         done = 1;
 763 }
 764
 765 void __init __weak arch_post_acpi_subsys_init(void) { }
 766
 767 void __init __weak smp_setup_processor_id(void)
 768 {
 769 }
 770
 771 # if THREAD_SIZE >= PAGE_SIZE
 772 void __init __weak thread_stack_cache_init(void)
 773 {
 774 }
 775 #endif
 776
 777 void __init __weak mem_encrypt_init(void) { }
 778
 779 void __init __weak poking_init(void) { }
 780
 781 void __init __weak pgtable_cache_init(void) { }
 782
 783 bool initcall_debug;
 784 core_param(initcall_debug, initcall_debug, bool, 0644);
 785
 786 #ifdef TRACEPOINTS_ENABLED
 787 static void __init initcall_debug_enable(void);
 788 #else
 789 static inline void initcall_debug_enable(void)
 790 {
 791 }
 792 #endif
 793
 794 /* Report memory auto-initialization states for this boot. */
 795 static void __init report_meminit(void)
 796 {
 797         const char *stack;
 798
 799         if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
 800                 stack = "all(pattern)";
 801         else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
 802                 stack = "all(zero)";
 803         else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
 804                 stack = "byref_all(zero)";
 805         else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
 806                 stack = "byref(zero)";
 807         else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
 808                 stack = "__user(zero)";
 809         else
 810                 stack = "off";
 811
 812         pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
 813                 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
 814                 want_init_on_free() ? "on" : "off");
 815         if (want_init_on_free())
 816                 pr_info("mem auto-init: clearing system memory may take some time...\n");
 817 }
 818
 819 /*
 820  * Set up kernel memory allocators
 821  */
 822 static void __init mm_init(void)
 823 {
 824         /*
 825          * page_ext requires contiguous pages,
 826          * bigger than MAX_ORDER unless SPARSEMEM.
 827          */
 828         page_ext_init_flatmem();
 829         init_mem_debugging_and_hardening();
 830         kfence_alloc_pool();
 831         report_meminit();
 832         stack_depot_init();
 833         mem_init();
 834         mem_init_print_info();
 835         /* page_owner must be initialized after buddy is ready */
 836         page_ext_init_flatmem_late();
 837         kmem_cache_init();
 838         kmemleak_init();
 839         pgtable_init();
 840         debug_objects_mem_init();
 841         vmalloc_init();
 842         /* Should be run before the first non-init thread is created */
 843         init_espfix_bsp();
 844         /* Should be run after espfix64 is set up. */
 845         pti_init();
 846 }
 847
 848 #ifdef CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
 849 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
 850                            randomize_kstack_offset);
 851 DEFINE_PER_CPU(u32, kstack_offset);
 852
 853 static int __init early_randomize_kstack_offset(char *buf)
 854 {
 855         int ret;
 856         bool bool_result;
 857
 858         ret = kstrtobool(buf, &bool_result);
 859         if (ret)
 860                 return ret;
 861
 862         if (bool_result)
 863                 static_branch_enable(&randomize_kstack_offset);
 864         else
 865                 static_branch_disable(&randomize_kstack_offset);
 866         return 0;
 867 }
 868 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
 869 #endif
 870
 871 void __init __weak arch_call_rest_init(void)
 872 {
 873         rest_init();
 874 }
 875
 876 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
 877 {
 878         char *command_line;
 879         char *after_dashes;
 880
 881         set_task_stack_end_magic(&init_task);
 882         smp_setup_processor_id();
 883         debug_objects_early_init();
 884
 885         cgroup_init_early();
 886
 887         local_irq_disable();
 888         early_boot_irqs_disabled = true;
 889
 890         /*
 891          * Interrupts are still disabled. Do necessary setups, then
 892          * enable them.
 893          */
 894         boot_cpu_init();
 895         page_address_init();
 896         pr_notice("%s", linux_banner);
 897         early_security_init();
 898         setup_arch(&command_line);
 899         setup_boot_config();
 900         setup_command_line(command_line);
 901         setup_nr_cpu_ids();
 902         setup_per_cpu_areas();
 903         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 904         boot_cpu_hotplug_init();
 905
 906         build_all_zonelists(NULL);
 907         page_alloc_init();
 908
 909         pr_notice("Kernel command line: %s\n", saved_command_line);
 910         /* parameters may set static keys */
 911         jump_label_init();
 912         parse_early_param();
 913         after_dashes = parse_args("Booting kernel",
 914                                   static_command_line, __start___param,
 915                                   __stop___param - __start___param,
 916                                   -1, -1, NULL, &unknown_bootoption);
 917         if (!IS_ERR_OR_NULL(after_dashes))
 918                 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
 919                            NULL, set_init_arg);
 920         if (extra_init_args)
 921                 parse_args("Setting extra init args", extra_init_args,
 922                            NULL, 0, -1, -1, NULL, set_init_arg);
 923
 924         /*
 925          * These use large bootmem allocations and must precede
 926          * kmem_cache_init()
 927          */
 928         setup_log_buf(0);
 929         vfs_caches_init_early();
 930         sort_main_extable();
 931         trap_init();
 932         mm_init();
 933
 934         ftrace_init();
 935
 936         /* trace_printk can be enabled here */
 937         early_trace_init();
 938
 939         /*
 940          * Set up the scheduler prior starting any interrupts (such as the
 941          * timer interrupt). Full topology setup happens at smp_init()
 942          * time - but meanwhile we still have a functioning scheduler.
 943          */
 944         sched_init();
 945
 946         if (WARN(!irqs_disabled(),
 947                  "Interrupts were enabled *very* early, fixing it\n"))
 948                 local_irq_disable();
 949         radix_tree_init();
 950
 951         /*
 952          * Set up housekeeping before setting up workqueues to allow the unbound
 953          * workqueue to take non-housekeeping into account.
 954          */
 955         housekeeping_init();
 956
 957         /*
 958          * Allow workqueue creation and work item queueing/cancelling
 959          * early.  Work item execution depends on kthreads and starts after
 960          * workqueue_init().
 961          */
 962         workqueue_init_early();
 963
 964         rcu_init();
 965
 966         /* Trace events are available after this */
 967         trace_init();
 968
 969         if (initcall_debug)
 970                 initcall_debug_enable();
 971
 972         context_tracking_init();
 973         /* init some links before init_ISA_irqs() */
 974         early_irq_init();
 975         init_IRQ();
 976         tick_init();
 977         rcu_init_nohz();
 978         init_timers();
 979         hrtimers_init();
 980         softirq_init();
 981         timekeeping_init();
 982         kfence_init();
 983
 984         /*
 985          * For best initial stack canary entropy, prepare it after:
 986          * - setup_arch() for any UEFI RNG entropy and boot cmdline access
 987          * - timekeeping_init() for ktime entropy used in rand_initialize()
 988          * - rand_initialize() to get any arch-specific entropy like RDRAND
 989          * - add_latent_entropy() to get any latent entropy
 990          * - adding command line entropy
 991          */
 992         rand_initialize();
 993         add_latent_entropy();
 994         add_device_randomness(command_line, strlen(command_line));
 995         boot_init_stack_canary();
 996
 997         time_init();
 998         perf_event_init();
 999         profile_init();
1000         call_function_init();
1001         WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1002
1003         early_boot_irqs_disabled = false;
1004         local_irq_enable();
1005
1006         kmem_cache_init_late();
1007
1008         /*
1009          * HACK ALERT! This is early. We're enabling the console before
1010          * we've done PCI setups etc, and console_init() must be aware of
1011          * this. But we do want output early, in case something goes wrong.
1012          */
1013         console_init();
1014         if (panic_later)
1015                 panic("Too many boot %s vars at `%s'", panic_later,
1016                       panic_param);
1017
1018         lockdep_init();
1019
1020         /*
1021          * Need to run this when irqs are enabled, because it wants
1022          * to self-test [hard/soft]-irqs on/off lock inversion bugs
1023          * too:
1024          */
1025         locking_selftest();
1026
1027         /*
1028          * This needs to be called before any devices perform DMA
1029          * operations that might use the SWIOTLB bounce buffers. It will
1030          * mark the bounce buffers as decrypted so that their usage will
1031          * not cause "plain-text" data to be decrypted when accessed.
1032          */
1033         mem_encrypt_init();
1034
1035 #ifdef CONFIG_BLK_DEV_INITRD
1036         if (initrd_start && !initrd_below_start_ok &&
1037             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1038                 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1039                     page_to_pfn(virt_to_page((void *)initrd_start)),
1040                     min_low_pfn);
1041                 initrd_start = 0;
1042         }
1043 #endif
1044         setup_per_cpu_pageset();
1045         numa_policy_init();
1046         acpi_early_init();
1047         if (late_time_init)
1048                 late_time_init();
1049         sched_clock_init();
1050         calibrate_delay();
1051         pid_idr_init();
1052         anon_vma_init();
1053 #ifdef CONFIG_X86
1054         if (efi_enabled(EFI_RUNTIME_SERVICES))
1055                 efi_enter_virtual_mode();
1056 #endif
1057         thread_stack_cache_init();
1058         cred_init();
1059         fork_init();
1060         proc_caches_init();
1061         uts_ns_init();
1062         key_init();
1063         security_init();
1064         dbg_late_init();
1065         vfs_caches_init();
1066         pagecache_init();
1067         signals_init();
1068         seq_file_init();
1069         proc_root_init();
1070         nsfs_init();
1071         cpuset_init();
1072         cgroup_init();
1073         taskstats_init_early();
1074         delayacct_init();
1075
1076         poking_init();
1077         check_bugs();
1078
1079         acpi_subsystem_init();
1080         arch_post_acpi_subsys_init();
1081         kcsan_init();
1082
1083         /* Do the rest non-__init'ed, we're now alive */
1084         arch_call_rest_init();
1085
1086         prevent_tail_call_optimization();
1087 }
1088
1089 /* Call all constructor functions linked into the kernel. */
1090 static void __init do_ctors(void)
1091 {
1092 /*
1093  * For UML, the constructors have already been called by the
1094  * normal setup code as it's just a normal ELF binary, so we
1095  * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1096  * even on UML for modules.
1097  */
1098 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1099         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1100
1101         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1102                 (*fn)();
1103 #endif
1104 }
1105
1106 #ifdef CONFIG_KALLSYMS
1107 struct blacklist_entry {
1108         struct list_head next;
1109         char *buf;
1110 };
1111
1112 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1113
1114 static int __init initcall_blacklist(char *str)
1115 {
1116         char *str_entry;
1117         struct blacklist_entry *entry;
1118
1119         /* str argument is a comma-separated list of functions */
1120         do {
1121                 str_entry = strsep(&str, ",");
1122                 if (str_entry) {
1123                         pr_debug("blacklisting initcall %s\n", str_entry);
1124                         entry = memblock_alloc(sizeof(*entry),
1125                                                SMP_CACHE_BYTES);
1126                         if (!entry)
1127                                 panic("%s: Failed to allocate %zu bytes\n",
1128                                       __func__, sizeof(*entry));
1129                         entry->buf = memblock_alloc(strlen(str_entry) + 1,
1130                                                     SMP_CACHE_BYTES);
1131                         if (!entry->buf)
1132                                 panic("%s: Failed to allocate %zu bytes\n",
1133                                       __func__, strlen(str_entry) + 1);
1134                         strcpy(entry->buf, str_entry);
1135                         list_add(&entry->next, &blacklisted_initcalls);
1136                 }
1137         } while (str_entry);
1138
1139         return 0;
1140 }
1141
1142 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1143 {
1144         struct blacklist_entry *entry;
1145         char fn_name[KSYM_SYMBOL_LEN];
1146         unsigned long addr;
1147
1148         if (list_empty(&blacklisted_initcalls))
1149                 return false;
1150
1151         addr = (unsigned long) dereference_function_descriptor(fn);
1152         sprint_symbol_no_offset(fn_name, addr);
1153
1154         /*
1155          * fn will be "function_name [module_name]" where [module_name] is not
1156          * displayed for built-in init functions.  Strip off the [module_name].
1157          */
1158         strreplace(fn_name, ' ', '\0');
1159
1160         list_for_each_entry(entry, &blacklisted_initcalls, next) {
1161                 if (!strcmp(fn_name, entry->buf)) {
1162                         pr_debug("initcall %s blacklisted\n", fn_name);
1163                         return true;
1164                 }
1165         }
1166
1167         return false;
1168 }
1169 #else
1170 static int __init initcall_blacklist(char *str)
1171 {
1172         pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1173         return 0;
1174 }
1175
1176 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1177 {
1178         return false;
1179 }
1180 #endif
1181 __setup("initcall_blacklist=", initcall_blacklist);
1182
1183 static __init_or_module void
1184 trace_initcall_start_cb(void *data, initcall_t fn)
1185 {
1186         ktime_t *calltime = (ktime_t *)data;
1187
1188         printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1189         *calltime = ktime_get();
1190 }
1191
1192 static __init_or_module void
1193 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1194 {
1195         ktime_t *calltime = (ktime_t *)data;
1196         ktime_t delta, rettime;
1197         unsigned long long duration;
1198
1199         rettime = ktime_get();
1200         delta = ktime_sub(rettime, *calltime);
1201         duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1202         printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1203                  fn, ret, duration);
1204 }
1205
1206 static ktime_t initcall_calltime;
1207
1208 #ifdef TRACEPOINTS_ENABLED
1209 static void __init initcall_debug_enable(void)
1210 {
1211         int ret;
1212
1213         ret = register_trace_initcall_start(trace_initcall_start_cb,
1214                                             &initcall_calltime);
1215         ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1216                                               &initcall_calltime);
1217         WARN(ret, "Failed to register initcall tracepoints\n");
1218 }
1219 # define do_trace_initcall_start        trace_initcall_start
1220 # define do_trace_initcall_finish       trace_initcall_finish
1221 #else
1222 static inline void do_trace_initcall_start(initcall_t fn)
1223 {
1224         if (!initcall_debug)
1225                 return;
1226         trace_initcall_start_cb(&initcall_calltime, fn);
1227 }
1228 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1229 {
1230         if (!initcall_debug)
1231                 return;
1232         trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1233 }
1234 #endif /* !TRACEPOINTS_ENABLED */
1235
1236 int __init_or_module do_one_initcall(initcall_t fn)
1237 {
1238         int count = preempt_count();
1239         char msgbuf[64];
1240         int ret;
1241
1242         if (initcall_blacklisted(fn))
1243                 return -EPERM;
1244
1245         do_trace_initcall_start(fn);
1246         ret = fn();
1247         do_trace_initcall_finish(fn, ret);
1248
1249         msgbuf[0] = 0;
1250
1251         if (preempt_count() != count) {
1252                 sprintf(msgbuf, "preemption imbalance ");
1253                 preempt_count_set(count);
1254         }
1255         if (irqs_disabled()) {
1256                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1257                 local_irq_enable();
1258         }
1259         WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1260
1261         add_latent_entropy();
1262         return ret;
1263 }
1264
1265
1266 extern initcall_entry_t __initcall_start[];
1267 extern initcall_entry_t __initcall0_start[];
1268 extern initcall_entry_t __initcall1_start[];
1269 extern initcall_entry_t __initcall2_start[];
1270 extern initcall_entry_t __initcall3_start[];
1271 extern initcall_entry_t __initcall4_start[];
1272 extern initcall_entry_t __initcall5_start[];
1273 extern initcall_entry_t __initcall6_start[];
1274 extern initcall_entry_t __initcall7_start[];
1275 extern initcall_entry_t __initcall_end[];
1276
1277 static initcall_entry_t *initcall_levels[] __initdata = {
1278         __initcall0_start,
1279         __initcall1_start,
1280         __initcall2_start,
1281         __initcall3_start,
1282         __initcall4_start,
1283         __initcall5_start,
1284         __initcall6_start,
1285         __initcall7_start,
1286         __initcall_end,
1287 };
1288
1289 /* Keep these in sync with initcalls in include/linux/init.h */
1290 static const char *initcall_level_names[] __initdata = {
1291         "pure",
1292         "core",
1293         "postcore",
1294         "arch",
1295         "subsys",
1296         "fs",
1297         "device",
1298         "late",
1299 };
1300
1301 static int __init ignore_unknown_bootoption(char *param, char *val,
1302                                const char *unused, void *arg)
1303 {
1304         return 0;
1305 }
1306
1307 static void __init do_initcall_level(int level, char *command_line)
1308 {
1309         initcall_entry_t *fn;
1310
1311         parse_args(initcall_level_names[level],
1312                    command_line, __start___param,
1313                    __stop___param - __start___param,
1314                    level, level,
1315                    NULL, ignore_unknown_bootoption);
1316
1317         trace_initcall_level(initcall_level_names[level]);
1318         for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1319                 do_one_initcall(initcall_from_entry(fn));
1320 }
1321
1322 static void __init do_initcalls(void)
1323 {
1324         int level;
1325         size_t len = strlen(saved_command_line) + 1;
1326         char *command_line;
1327
1328         command_line = kzalloc(len, GFP_KERNEL);
1329         if (!command_line)
1330                 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1331
1332         for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1333                 /* Parser modifies command_line, restore it each time */
1334                 strcpy(command_line, saved_command_line);
1335                 do_initcall_level(level, command_line);
1336         }
1337
1338         kfree(command_line);
1339 }
1340
1341 /*
1342  * Ok, the machine is now initialized. None of the devices
1343  * have been touched yet, but the CPU subsystem is up and
1344  * running, and memory and process management works.
1345  *
1346  * Now we can finally start doing some real work..
1347  */
1348 static void __init do_basic_setup(void)
1349 {
1350         cpuset_init_smp();
1351         driver_init();
1352         init_irq_proc();
1353         do_ctors();
1354         usermodehelper_enable();
1355         do_initcalls();
1356 }
1357
1358 static void __init do_pre_smp_initcalls(void)
1359 {
1360         initcall_entry_t *fn;
1361
1362         trace_initcall_level("early");
1363         for (fn = __initcall_start; fn < __initcall0_start; fn++)
1364                 do_one_initcall(initcall_from_entry(fn));
1365 }
1366
1367 static int run_init_process(const char *init_filename)
1368 {
1369         const char *const *p;
1370
1371         argv_init[0] = init_filename;
1372         pr_info("Run %s as init process\n", init_filename);
1373         pr_debug("  with arguments:\n");
1374         for (p = argv_init; *p; p++)
1375                 pr_debug("    %s\n", *p);
1376         pr_debug("  with environment:\n");
1377         for (p = envp_init; *p; p++)
1378                 pr_debug("    %s\n", *p);
1379         return kernel_execve(init_filename, argv_init, envp_init);
1380 }
1381
1382 static int try_to_run_init_process(const char *init_filename)
1383 {
1384         int ret;
1385
1386         ret = run_init_process(init_filename);
1387
1388         if (ret && ret != -ENOENT) {
1389                 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1390                        init_filename, ret);
1391         }
1392
1393         return ret;
1394 }
1395
1396 static noinline void __init kernel_init_freeable(void);
1397
1398 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1399 bool rodata_enabled __ro_after_init = true;
1400 static int __init set_debug_rodata(char *str)
1401 {
1402         return strtobool(str, &rodata_enabled);
1403 }
1404 __setup("rodata=", set_debug_rodata);
1405 #endif
1406
1407 #ifdef CONFIG_STRICT_KERNEL_RWX
1408 static void mark_readonly(void)
1409 {
1410         if (rodata_enabled) {
1411                 /*
1412                  * load_module() results in W+X mappings, which are cleaned
1413                  * up with call_rcu().  Let's make sure that queued work is
1414                  * flushed so that we don't hit false positives looking for
1415                  * insecure pages which are W+X.
1416                  */
1417                 rcu_barrier();
1418                 mark_rodata_ro();
1419                 rodata_test();
1420         } else
1421                 pr_info("Kernel memory protection disabled.\n");
1422 }
1423 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1424 static inline void mark_readonly(void)
1425 {
1426         pr_warn("Kernel memory protection not selected by kernel config.\n");
1427 }
1428 #else
1429 static inline void mark_readonly(void)
1430 {
1431         pr_warn("This architecture does not have kernel memory protection.\n");
1432 }
1433 #endif
1434
1435 void __weak free_initmem(void)
1436 {
1437         free_initmem_default(POISON_FREE_INITMEM);
1438 }
1439
1440 static int __ref kernel_init(void *unused)
1441 {
1442         int ret;
1443
1444         /*
1445          * Wait until kthreadd is all set-up.
1446          */
1447         wait_for_completion(&kthreadd_done);
1448
1449         kernel_init_freeable();
1450         /* need to finish all async __init code before freeing the memory */
1451         async_synchronize_full();
1452         kprobe_free_init_mem();
1453         ftrace_free_init_mem();
1454         kgdb_free_init_mem();
1455         free_initmem();
1456         mark_readonly();
1457
1458         /*
1459          * Kernel mappings are now finalized - update the userspace page-table
1460          * to finalize PTI.
1461          */
1462         pti_finalize();
1463
1464         system_state = SYSTEM_RUNNING;
1465         numa_default_policy();
1466
1467         rcu_end_inkernel_boot();
1468
1469         do_sysctl_args();
1470
1471         if (ramdisk_execute_command) {
1472                 ret = run_init_process(ramdisk_execute_command);
1473                 if (!ret)
1474                         return 0;
1475                 pr_err("Failed to execute %s (error %d)\n",
1476                        ramdisk_execute_command, ret);
1477         }
1478
1479         /*
1480          * We try each of these until one succeeds.
1481          *
1482          * The Bourne shell can be used instead of init if we are
1483          * trying to recover a really broken machine.
1484          */
1485         if (execute_command) {
1486                 ret = run_init_process(execute_command);
1487                 if (!ret)
1488                         return 0;
1489                 panic("Requested init %s failed (error %d).",
1490                       execute_command, ret);
1491         }
1492
1493         if (CONFIG_DEFAULT_INIT[0] != '\0') {
1494                 ret = run_init_process(CONFIG_DEFAULT_INIT);
1495                 if (ret)
1496                         pr_err("Default init %s failed (error %d)\n",
1497                                CONFIG_DEFAULT_INIT, ret);
1498                 else
1499                         return 0;
1500         }
1501
1502         if (!try_to_run_init_process("/sbin/init") ||
1503             !try_to_run_init_process("/etc/init") ||
1504             !try_to_run_init_process("/bin/init") ||
1505             !try_to_run_init_process("/bin/sh"))
1506                 return 0;
1507
1508         panic("No working init found.  Try passing init= option to kernel. "
1509               "See Linux Documentation/admin-guide/init.rst for guidance.");
1510 }
1511
1512 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1513 void __init console_on_rootfs(void)
1514 {
1515         struct file *file = filp_open("/dev/console", O_RDWR, 0);
1516
1517         if (IS_ERR(file)) {
1518                 pr_err("Warning: unable to open an initial console.\n");
1519                 return;
1520         }
1521         init_dup(file);
1522         init_dup(file);
1523         init_dup(file);
1524         fput(file);
1525 }
1526
1527 static noinline void __init kernel_init_freeable(void)
1528 {
1529         /* Now the scheduler is fully set up and can do blocking allocations */
1530         gfp_allowed_mask = __GFP_BITS_MASK;
1531
1532         /*
1533          * init can allocate pages on any node
1534          */
1535         set_mems_allowed(node_states[N_MEMORY]);
1536
1537         cad_pid = get_pid(task_pid(current));
1538
1539         smp_prepare_cpus(setup_max_cpus);
1540
1541         workqueue_init();
1542
1543         init_mm_internals();
1544
1545         rcu_init_tasks_generic();
1546         do_pre_smp_initcalls();
1547         lockup_detector_init();
1548
1549         smp_init();
1550         sched_init_smp();
1551
1552         padata_init();
1553         page_alloc_init_late();
1554         /* Initialize page ext after all struct pages are initialized. */
1555         page_ext_init();
1556
1557         do_basic_setup();
1558
1559         kunit_run_all_tests();
1560
1561         wait_for_initramfs();
1562         console_on_rootfs();
1563
1564         /*
1565          * check if there is an early userspace init.  If yes, let it do all
1566          * the work
1567          */
1568         if (init_eaccess(ramdisk_execute_command) != 0) {
1569                 ramdisk_execute_command = NULL;
1570                 prepare_namespace();
1571         }
1572
1573         /*
1574          * Ok, we have completed the initial bootup, and
1575          * we're essentially up and running. Get rid of the
1576          * initmem segments and start the user-mode stuff..
1577          *
1578          * rootfs is available now, try loading the public keys
1579          * and default modules
1580          */
1581
1582         integrity_load_keys();
1583 }