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/vmalloc.h>
  37 #include <linux/kernel_stat.h>
  38 #include <linux/start_kernel.h>
  39 #include <linux/security.h>
  40 #include <linux/smp.h>
  41 #include <linux/profile.h>
  42 #include <linux/rcupdate.h>
  43 #include <linux/moduleparam.h>
  44 #include <linux/kallsyms.h>
  45 #include <linux/writeback.h>
  46 #include <linux/cpu.h>
  47 #include <linux/cpuset.h>
  48 #include <linux/cgroup.h>
  49 #include <linux/efi.h>
  50 #include <linux/tick.h>
  51 #include <linux/sched/isolation.h>
  52 #include <linux/interrupt.h>
  53 #include <linux/taskstats_kern.h>
  54 #include <linux/delayacct.h>
  55 #include <linux/unistd.h>
  56 #include <linux/utsname.h>
  57 #include <linux/rmap.h>
  58 #include <linux/mempolicy.h>
  59 #include <linux/key.h>
  60 #include <linux/buffer_head.h>
  61 #include <linux/page_ext.h>
  62 #include <linux/debug_locks.h>
  63 #include <linux/debugobjects.h>
  64 #include <linux/lockdep.h>
  65 #include <linux/kmemleak.h>
  66 #include <linux/pid_namespace.h>
  67 #include <linux/device/driver.h>
  68 #include <linux/kthread.h>
  69 #include <linux/sched.h>
  70 #include <linux/sched/init.h>
  71 #include <linux/signal.h>
  72 #include <linux/idr.h>
  73 #include <linux/kgdb.h>
  74 #include <linux/ftrace.h>
  75 #include <linux/async.h>
  76 #include <linux/sfi.h>
  77 #include <linux/shmem_fs.h>
  78 #include <linux/slab.h>
  79 #include <linux/perf_event.h>
  80 #include <linux/ptrace.h>
  81 #include <linux/pti.h>
  82 #include <linux/blkdev.h>
  83 #include <linux/elevator.h>
  84 #include <linux/sched/clock.h>
  85 #include <linux/sched/task.h>
  86 #include <linux/sched/task_stack.h>
  87 #include <linux/context_tracking.h>
  88 #include <linux/random.h>
  89 #include <linux/list.h>
  90 #include <linux/integrity.h>
  91 #include <linux/proc_ns.h>
  92 #include <linux/io.h>
  93 #include <linux/cache.h>
  94 #include <linux/rodata_test.h>
  95 #include <linux/jump_label.h>
  96 #include <linux/mem_encrypt.h>
  97
  98 #include <asm/io.h>
  99 #include <asm/bugs.h>
 100 #include <asm/setup.h>
 101 #include <asm/sections.h>
 102 #include <asm/cacheflush.h>
 103
 104 #define CREATE_TRACE_POINTS
 105 #include <trace/events/initcall.h>
 106
 107 static int kernel_init(void *);
 108
 109 extern void init_IRQ(void);
 110 extern void radix_tree_init(void);
 111
 112 /*
 113  * Debug helper: via this flag we know that we are in 'early bootup code'
 114  * where only the boot processor is running with IRQ disabled.  This means
 115  * two things - IRQ must not be enabled before the flag is cleared and some
 116  * operations which are not allowed with IRQ disabled are allowed while the
 117  * flag is set.
 118  */
 119 bool early_boot_irqs_disabled __read_mostly;
 120
 121 enum system_states system_state __read_mostly;
 122 EXPORT_SYMBOL(system_state);
 123
 124 /*
 125  * Boot command-line arguments
 126  */
 127 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
 128 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
 129
 130 extern void time_init(void);
 131 /* Default late time init is NULL. archs can override this later. */
 132 void (*__initdata late_time_init)(void);
 133
 134 /* Untouched command line saved by arch-specific code. */
 135 char __initdata boot_command_line[COMMAND_LINE_SIZE];
 136 /* Untouched saved command line (eg. for /proc) */
 137 char *saved_command_line;
 138 /* Command line for parameter parsing */
 139 static char *static_command_line;
 140 /* Untouched extra command line */
 141 static char *extra_command_line;
 142 /* Extra init arguments */
 143 static char *extra_init_args;
 144
 145 static char *execute_command;
 146 static char *ramdisk_execute_command;
 147
 148 /*
 149  * Used to generate warnings if static_key manipulation functions are used
 150  * before jump_label_init is called.
 151  */
 152 bool static_key_initialized __read_mostly;
 153 EXPORT_SYMBOL_GPL(static_key_initialized);
 154
 155 /*
 156  * If set, this is an indication to the drivers that reset the underlying
 157  * device before going ahead with the initialization otherwise driver might
 158  * rely on the BIOS and skip the reset operation.
 159  *
 160  * This is useful if kernel is booting in an unreliable environment.
 161  * For ex. kdump situation where previous kernel has crashed, BIOS has been
 162  * skipped and devices will be in unknown state.
 163  */
 164 unsigned int reset_devices;
 165 EXPORT_SYMBOL(reset_devices);
 166
 167 static int __init set_reset_devices(char *str)
 168 {
 169         reset_devices = 1;
 170         return 1;
 171 }
 172
 173 __setup("reset_devices", set_reset_devices);
 174
 175 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
 176 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
 177 static const char *panic_later, *panic_param;
 178
 179 extern const struct obs_kernel_param __setup_start[], __setup_end[];
 180
 181 static bool __init obsolete_checksetup(char *line)
 182 {
 183         const struct obs_kernel_param *p;
 184         bool had_early_param = false;
 185
 186         p = __setup_start;
 187         do {
 188                 int n = strlen(p->str);
 189                 if (parameqn(line, p->str, n)) {
 190                         if (p->early) {
 191                                 /* Already done in parse_early_param?
 192                                  * (Needs exact match on param part).
 193                                  * Keep iterating, as we can have early
 194                                  * params and __setups of same names 8( */
 195                                 if (line[n] == '\0' || line[n] == '=')
 196                                         had_early_param = true;
 197                         } else if (!p->setup_func) {
 198                                 pr_warn("Parameter %s is obsolete, ignored\n",
 199                                         p->str);
 200                                 return true;
 201                         } else if (p->setup_func(line + n))
 202                                 return true;
 203                 }
 204                 p++;
 205         } while (p < __setup_end);
 206
 207         return had_early_param;
 208 }
 209
 210 /*
 211  * This should be approx 2 Bo*oMips to start (note initial shift), and will
 212  * still work even if initially too large, it will just take slightly longer
 213  */
 214 unsigned long loops_per_jiffy = (1<<12);
 215 EXPORT_SYMBOL(loops_per_jiffy);
 216
 217 static int __init debug_kernel(char *str)
 218 {
 219         console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
 220         return 0;
 221 }
 222
 223 static int __init quiet_kernel(char *str)
 224 {
 225         console_loglevel = CONSOLE_LOGLEVEL_QUIET;
 226         return 0;
 227 }
 228
 229 early_param("debug", debug_kernel);
 230 early_param("quiet", quiet_kernel);
 231
 232 static int __init loglevel(char *str)
 233 {
 234         int newlevel;
 235
 236         /*
 237          * Only update loglevel value when a correct setting was passed,
 238          * to prevent blind crashes (when loglevel being set to 0) that
 239          * are quite hard to debug
 240          */
 241         if (get_option(&str, &newlevel)) {
 242                 console_loglevel = newlevel;
 243                 return 0;
 244         }
 245
 246         return -EINVAL;
 247 }
 248
 249 early_param("loglevel", loglevel);
 250
 251 #ifdef CONFIG_BOOT_CONFIG
 252
 253 char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
 254
 255 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
 256
 257 static int __init xbc_snprint_cmdline(char *buf, size_t size,
 258                                       struct xbc_node *root)
 259 {
 260         struct xbc_node *knode, *vnode;
 261         char *end = buf + size;
 262         char c = '\"';
 263         const char *val;
 264         int ret;
 265
 266         xbc_node_for_each_key_value(root, knode, val) {
 267                 ret = xbc_node_compose_key_after(root, knode,
 268                                         xbc_namebuf, XBC_KEYLEN_MAX);
 269                 if (ret < 0)
 270                         return ret;
 271
 272                 vnode = xbc_node_get_child(knode);
 273                 ret = snprintf(buf, rest(buf, end), "%s%c", xbc_namebuf,
 274                                 vnode ? '=' : ' ');
 275                 if (ret < 0)
 276                         return ret;
 277                 buf += ret;
 278                 if (!vnode)
 279                         continue;
 280
 281                 c = '\"';
 282                 xbc_array_for_each_value(vnode, val) {
 283                         ret = snprintf(buf, rest(buf, end), "%c%s", c, val);
 284                         if (ret < 0)
 285                                 return ret;
 286                         buf += ret;
 287                         c = ',';
 288                 }
 289                 if (rest(buf, end) > 2)
 290                         strcpy(buf, "\" ");
 291                 buf += 2;
 292         }
 293
 294         return buf - (end - size);
 295 }
 296 #undef rest
 297
 298 /* Make an extra command line under given key word */
 299 static char * __init xbc_make_cmdline(const char *key)
 300 {
 301         struct xbc_node *root;
 302         char *new_cmdline;
 303         int ret, len = 0;
 304
 305         root = xbc_find_node(key);
 306         if (!root)
 307                 return NULL;
 308
 309         /* Count required buffer size */
 310         len = xbc_snprint_cmdline(NULL, 0, root);
 311         if (len <= 0)
 312                 return NULL;
 313
 314         new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
 315         if (!new_cmdline) {
 316                 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
 317                 return NULL;
 318         }
 319
 320         ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
 321         if (ret < 0 || ret > len) {
 322                 pr_err("Failed to print extra kernel cmdline.\n");
 323                 return NULL;
 324         }
 325
 326         return new_cmdline;
 327 }
 328
 329 u32 boot_config_checksum(unsigned char *p, u32 size)
 330 {
 331         u32 ret = 0;
 332
 333         while (size--)
 334                 ret += *p++;
 335
 336         return ret;
 337 }
 338
 339 static void __init setup_boot_config(const char *cmdline)
 340 {
 341         u32 size, csum;
 342         char *data, *copy;
 343         const char *p;
 344         u32 *hdr;
 345         int ret;
 346
 347         p = strstr(cmdline, "bootconfig");
 348         if (!p || (p != cmdline && !isspace(*(p-1))) ||
 349             (p[10] && !isspace(p[10])))
 350                 return;
 351
 352         if (!initrd_end)
 353                 goto not_found;
 354
 355         hdr = (u32 *)(initrd_end - 8);
 356         size = hdr[0];
 357         csum = hdr[1];
 358
 359         if (size >= XBC_DATA_MAX) {
 360                 pr_err("bootconfig size %d greater than max size %d\n",
 361                         size, XBC_DATA_MAX);
 362                 return;
 363         }
 364
 365         data = ((void *)hdr) - size;
 366         if ((unsigned long)data < initrd_start)
 367                 goto not_found;
 368
 369         if (boot_config_checksum((unsigned char *)data, size) != csum) {
 370                 pr_err("bootconfig checksum failed\n");
 371                 return;
 372         }
 373
 374         copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
 375         if (!copy) {
 376                 pr_err("Failed to allocate memory for bootconfig\n");
 377                 return;
 378         }
 379
 380         memcpy(copy, data, size);
 381         copy[size] = '\0';
 382
 383         ret = xbc_init(copy);
 384         if (ret < 0)
 385                 pr_err("Failed to parse bootconfig\n");
 386         else {
 387                 pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
 388                 /* keys starting with "kernel." are passed via cmdline */
 389                 extra_command_line = xbc_make_cmdline("kernel");
 390                 /* Also, "init." keys are init arguments */
 391                 extra_init_args = xbc_make_cmdline("init");
 392         }
 393         return;
 394 not_found:
 395         pr_err("'bootconfig' found on command line, but no bootconfig found\n");
 396 }
 397 #else
 398 #define setup_boot_config(cmdline)      do { } while (0)
 399 #endif
 400
 401 /* Change NUL term back to "=", to make "param" the whole string. */
 402 static void __init repair_env_string(char *param, char *val)
 403 {
 404         if (val) {
 405                 /* param=val or param="val"? */
 406                 if (val == param+strlen(param)+1)
 407                         val[-1] = '=';
 408                 else if (val == param+strlen(param)+2) {
 409                         val[-2] = '=';
 410                         memmove(val-1, val, strlen(val)+1);
 411                 } else
 412                         BUG();
 413         }
 414 }
 415
 416 /* Anything after -- gets handed straight to init. */
 417 static int __init set_init_arg(char *param, char *val,
 418                                const char *unused, void *arg)
 419 {
 420         unsigned int i;
 421
 422         if (panic_later)
 423                 return 0;
 424
 425         repair_env_string(param, val);
 426
 427         for (i = 0; argv_init[i]; i++) {
 428                 if (i == MAX_INIT_ARGS) {
 429                         panic_later = "init";
 430                         panic_param = param;
 431                         return 0;
 432                 }
 433         }
 434         argv_init[i] = param;
 435         return 0;
 436 }
 437
 438 /*
 439  * Unknown boot options get handed to init, unless they look like
 440  * unused parameters (modprobe will find them in /proc/cmdline).
 441  */
 442 static int __init unknown_bootoption(char *param, char *val,
 443                                      const char *unused, void *arg)
 444 {
 445         size_t len = strlen(param);
 446
 447         repair_env_string(param, val);
 448
 449         /* Handle obsolete-style parameters */
 450         if (obsolete_checksetup(param))
 451                 return 0;
 452
 453         /* Unused module parameter. */
 454         if (strnchr(param, len, '.'))
 455                 return 0;
 456
 457         if (panic_later)
 458                 return 0;
 459
 460         if (val) {
 461                 /* Environment option */
 462                 unsigned int i;
 463                 for (i = 0; envp_init[i]; i++) {
 464                         if (i == MAX_INIT_ENVS) {
 465                                 panic_later = "env";
 466                                 panic_param = param;
 467                         }
 468                         if (!strncmp(param, envp_init[i], len+1))
 469                                 break;
 470                 }
 471                 envp_init[i] = param;
 472         } else {
 473                 /* Command line option */
 474                 unsigned int i;
 475                 for (i = 0; argv_init[i]; i++) {
 476                         if (i == MAX_INIT_ARGS) {
 477                                 panic_later = "init";
 478                                 panic_param = param;
 479                         }
 480                 }
 481                 argv_init[i] = param;
 482         }
 483         return 0;
 484 }
 485
 486 static int __init init_setup(char *str)
 487 {
 488         unsigned int i;
 489
 490         execute_command = str;
 491         /*
 492          * In case LILO is going to boot us with default command line,
 493          * it prepends "auto" before the whole cmdline which makes
 494          * the shell think it should execute a script with such name.
 495          * So we ignore all arguments entered _before_ init=... [MJ]
 496          */
 497         for (i = 1; i < MAX_INIT_ARGS; i++)
 498                 argv_init[i] = NULL;
 499         return 1;
 500 }
 501 __setup("init=", init_setup);
 502
 503 static int __init rdinit_setup(char *str)
 504 {
 505         unsigned int i;
 506
 507         ramdisk_execute_command = str;
 508         /* See "auto" comment in init_setup */
 509         for (i = 1; i < MAX_INIT_ARGS; i++)
 510                 argv_init[i] = NULL;
 511         return 1;
 512 }
 513 __setup("rdinit=", rdinit_setup);
 514
 515 #ifndef CONFIG_SMP
 516 static const unsigned int setup_max_cpus = NR_CPUS;
 517 static inline void setup_nr_cpu_ids(void) { }
 518 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
 519 #endif
 520
 521 /*
 522  * We need to store the untouched command line for future reference.
 523  * We also need to store the touched command line since the parameter
 524  * parsing is performed in place, and we should allow a component to
 525  * store reference of name/value for future reference.
 526  */
 527 static void __init setup_command_line(char *command_line)
 528 {
 529         size_t len, xlen = 0, ilen = 0;
 530
 531         if (extra_command_line)
 532                 xlen = strlen(extra_command_line);
 533         if (extra_init_args)
 534                 ilen = strlen(extra_init_args) + 4; /* for " -- " */
 535
 536         len = xlen + strlen(boot_command_line) + 1;
 537
 538         saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
 539         if (!saved_command_line)
 540                 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
 541
 542         static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
 543         if (!static_command_line)
 544                 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
 545
 546         if (xlen) {
 547                 /*
 548                  * We have to put extra_command_line before boot command
 549                  * lines because there could be dashes (separator of init
 550                  * command line) in the command lines.
 551                  */
 552                 strcpy(saved_command_line, extra_command_line);
 553                 strcpy(static_command_line, extra_command_line);
 554         }
 555         strcpy(saved_command_line + xlen, boot_command_line);
 556         strcpy(static_command_line + xlen, command_line);
 557
 558         if (ilen) {
 559                 /*
 560                  * Append supplemental init boot args to saved_command_line
 561                  * so that user can check what command line options passed
 562                  * to init.
 563                  */
 564                 len = strlen(saved_command_line);
 565                 if (!strstr(boot_command_line, " -- ")) {
 566                         strcpy(saved_command_line + len, " -- ");
 567                         len += 4;
 568                 } else
 569                         saved_command_line[len++] = ' ';
 570
 571                 strcpy(saved_command_line + len, extra_init_args);
 572         }
 573 }
 574
 575 /*
 576  * We need to finalize in a non-__init function or else race conditions
 577  * between the root thread and the init thread may cause start_kernel to
 578  * be reaped by free_initmem before the root thread has proceeded to
 579  * cpu_idle.
 580  *
 581  * gcc-3.4 accidentally inlines this function, so use noinline.
 582  */
 583
 584 static __initdata DECLARE_COMPLETION(kthreadd_done);
 585
 586 noinline void __ref rest_init(void)
 587 {
 588         struct task_struct *tsk;
 589         int pid;
 590
 591         rcu_scheduler_starting();
 592         /*
 593          * We need to spawn init first so that it obtains pid 1, however
 594          * the init task will end up wanting to create kthreads, which, if
 595          * we schedule it before we create kthreadd, will OOPS.
 596          */
 597         pid = kernel_thread(kernel_init, NULL, CLONE_FS);
 598         /*
 599          * Pin init on the boot CPU. Task migration is not properly working
 600          * until sched_init_smp() has been run. It will set the allowed
 601          * CPUs for init to the non isolated CPUs.
 602          */
 603         rcu_read_lock();
 604         tsk = find_task_by_pid_ns(pid, &init_pid_ns);
 605         set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
 606         rcu_read_unlock();
 607
 608         numa_default_policy();
 609         pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
 610         rcu_read_lock();
 611         kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
 612         rcu_read_unlock();
 613
 614         /*
 615          * Enable might_sleep() and smp_processor_id() checks.
 616          * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
 617          * kernel_thread() would trigger might_sleep() splats. With
 618          * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
 619          * already, but it's stuck on the kthreadd_done completion.
 620          */
 621         system_state = SYSTEM_SCHEDULING;
 622
 623         complete(&kthreadd_done);
 624
 625         /*
 626          * The boot idle thread must execute schedule()
 627          * at least once to get things moving:
 628          */
 629         schedule_preempt_disabled();
 630         /* Call into cpu_idle with preempt disabled */
 631         cpu_startup_entry(CPUHP_ONLINE);
 632 }
 633
 634 /* Check for early params. */
 635 static int __init do_early_param(char *param, char *val,
 636                                  const char *unused, void *arg)
 637 {
 638         const struct obs_kernel_param *p;
 639
 640         for (p = __setup_start; p < __setup_end; p++) {
 641                 if ((p->early && parameq(param, p->str)) ||
 642                     (strcmp(param, "console") == 0 &&
 643                      strcmp(p->str, "earlycon") == 0)
 644                 ) {
 645                         if (p->setup_func(val) != 0)
 646                                 pr_warn("Malformed early option '%s'\n", param);
 647                 }
 648         }
 649         /* We accept everything at this stage. */
 650         return 0;
 651 }
 652
 653 void __init parse_early_options(char *cmdline)
 654 {
 655         parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
 656                    do_early_param);
 657 }
 658
 659 /* Arch code calls this early on, or if not, just before other parsing. */
 660 void __init parse_early_param(void)
 661 {
 662         static int done __initdata;
 663         static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
 664
 665         if (done)
 666                 return;
 667
 668         /* All fall through to do_early_param. */
 669         strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 670         parse_early_options(tmp_cmdline);
 671         done = 1;
 672 }
 673
 674 void __init __weak arch_post_acpi_subsys_init(void) { }
 675
 676 void __init __weak smp_setup_processor_id(void)
 677 {
 678 }
 679
 680 # if THREAD_SIZE >= PAGE_SIZE
 681 void __init __weak thread_stack_cache_init(void)
 682 {
 683 }
 684 #endif
 685
 686 void __init __weak mem_encrypt_init(void) { }
 687
 688 void __init __weak poking_init(void) { }
 689
 690 void __init __weak pgtable_cache_init(void) { }
 691
 692 bool initcall_debug;
 693 core_param(initcall_debug, initcall_debug, bool, 0644);
 694
 695 #ifdef TRACEPOINTS_ENABLED
 696 static void __init initcall_debug_enable(void);
 697 #else
 698 static inline void initcall_debug_enable(void)
 699 {
 700 }
 701 #endif
 702
 703 /* Report memory auto-initialization states for this boot. */
 704 static void __init report_meminit(void)
 705 {
 706         const char *stack;
 707
 708         if (IS_ENABLED(CONFIG_INIT_STACK_ALL))
 709                 stack = "all";
 710         else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
 711                 stack = "byref_all";
 712         else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
 713                 stack = "byref";
 714         else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
 715                 stack = "__user";
 716         else
 717                 stack = "off";
 718
 719         pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
 720                 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
 721                 want_init_on_free() ? "on" : "off");
 722         if (want_init_on_free())
 723                 pr_info("mem auto-init: clearing system memory may take some time...\n");
 724 }
 725
 726 /*
 727  * Set up kernel memory allocators
 728  */
 729 static void __init mm_init(void)
 730 {
 731         /*
 732          * page_ext requires contiguous pages,
 733          * bigger than MAX_ORDER unless SPARSEMEM.
 734          */
 735         page_ext_init_flatmem();
 736         init_debug_pagealloc();
 737         report_meminit();
 738         mem_init();
 739         kmem_cache_init();
 740         kmemleak_init();
 741         pgtable_init();
 742         debug_objects_mem_init();
 743         vmalloc_init();
 744         ioremap_huge_init();
 745         /* Should be run before the first non-init thread is created */
 746         init_espfix_bsp();
 747         /* Should be run after espfix64 is set up. */
 748         pti_init();
 749 }
 750
 751 void __init __weak arch_call_rest_init(void)
 752 {
 753         rest_init();
 754 }
 755
 756 asmlinkage __visible void __init start_kernel(void)
 757 {
 758         char *command_line;
 759         char *after_dashes;
 760
 761         set_task_stack_end_magic(&init_task);
 762         smp_setup_processor_id();
 763         debug_objects_early_init();
 764
 765         cgroup_init_early();
 766
 767         local_irq_disable();
 768         early_boot_irqs_disabled = true;
 769
 770         /*
 771          * Interrupts are still disabled. Do necessary setups, then
 772          * enable them.
 773          */
 774         boot_cpu_init();
 775         page_address_init();
 776         pr_notice("%s", linux_banner);
 777         early_security_init();
 778         setup_arch(&command_line);
 779         setup_boot_config(command_line);
 780         setup_command_line(command_line);
 781         setup_nr_cpu_ids();
 782         setup_per_cpu_areas();
 783         smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
 784         boot_cpu_hotplug_init();
 785
 786         build_all_zonelists(NULL);
 787         page_alloc_init();
 788
 789         pr_notice("Kernel command line: %s\n", saved_command_line);
 790         /* parameters may set static keys */
 791         jump_label_init();
 792         parse_early_param();
 793         after_dashes = parse_args("Booting kernel",
 794                                   static_command_line, __start___param,
 795                                   __stop___param - __start___param,
 796                                   -1, -1, NULL, &unknown_bootoption);
 797         if (!IS_ERR_OR_NULL(after_dashes))
 798                 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
 799                            NULL, set_init_arg);
 800         if (extra_init_args)
 801                 parse_args("Setting extra init args", extra_init_args,
 802                            NULL, 0, -1, -1, NULL, set_init_arg);
 803
 804         /*
 805          * These use large bootmem allocations and must precede
 806          * kmem_cache_init()
 807          */
 808         setup_log_buf(0);
 809         vfs_caches_init_early();
 810         sort_main_extable();
 811         trap_init();
 812         mm_init();
 813
 814         ftrace_init();
 815
 816         /* trace_printk can be enabled here */
 817         early_trace_init();
 818
 819         /*
 820          * Set up the scheduler prior starting any interrupts (such as the
 821          * timer interrupt). Full topology setup happens at smp_init()
 822          * time - but meanwhile we still have a functioning scheduler.
 823          */
 824         sched_init();
 825         /*
 826          * Disable preemption - early bootup scheduling is extremely
 827          * fragile until we cpu_idle() for the first time.
 828          */
 829         preempt_disable();
 830         if (WARN(!irqs_disabled(),
 831                  "Interrupts were enabled *very* early, fixing it\n"))
 832                 local_irq_disable();
 833         radix_tree_init();
 834
 835         /*
 836          * Set up housekeeping before setting up workqueues to allow the unbound
 837          * workqueue to take non-housekeeping into account.
 838          */
 839         housekeeping_init();
 840
 841         /*
 842          * Allow workqueue creation and work item queueing/cancelling
 843          * early.  Work item execution depends on kthreads and starts after
 844          * workqueue_init().
 845          */
 846         workqueue_init_early();
 847
 848         rcu_init();
 849
 850         /* Trace events are available after this */
 851         trace_init();
 852
 853         if (initcall_debug)
 854                 initcall_debug_enable();
 855
 856         context_tracking_init();
 857         /* init some links before init_ISA_irqs() */
 858         early_irq_init();
 859         init_IRQ();
 860         tick_init();
 861         rcu_init_nohz();
 862         init_timers();
 863         hrtimers_init();
 864         softirq_init();
 865         timekeeping_init();
 866
 867         /*
 868          * For best initial stack canary entropy, prepare it after:
 869          * - setup_arch() for any UEFI RNG entropy and boot cmdline access
 870          * - timekeeping_init() for ktime entropy used in rand_initialize()
 871          * - rand_initialize() to get any arch-specific entropy like RDRAND
 872          * - add_latent_entropy() to get any latent entropy
 873          * - adding command line entropy
 874          */
 875         rand_initialize();
 876         add_latent_entropy();
 877         add_device_randomness(command_line, strlen(command_line));
 878         boot_init_stack_canary();
 879
 880         time_init();
 881         printk_safe_init();
 882         perf_event_init();
 883         profile_init();
 884         call_function_init();
 885         WARN(!irqs_disabled(), "Interrupts were enabled early\n");
 886
 887         early_boot_irqs_disabled = false;
 888         local_irq_enable();
 889
 890         kmem_cache_init_late();
 891
 892         /*
 893          * HACK ALERT! This is early. We're enabling the console before
 894          * we've done PCI setups etc, and console_init() must be aware of
 895          * this. But we do want output early, in case something goes wrong.
 896          */
 897         console_init();
 898         if (panic_later)
 899                 panic("Too many boot %s vars at `%s'", panic_later,
 900                       panic_param);
 901
 902         lockdep_init();
 903
 904         /*
 905          * Need to run this when irqs are enabled, because it wants
 906          * to self-test [hard/soft]-irqs on/off lock inversion bugs
 907          * too:
 908          */
 909         locking_selftest();
 910
 911         /*
 912          * This needs to be called before any devices perform DMA
 913          * operations that might use the SWIOTLB bounce buffers. It will
 914          * mark the bounce buffers as decrypted so that their usage will
 915          * not cause "plain-text" data to be decrypted when accessed.
 916          */
 917         mem_encrypt_init();
 918
 919 #ifdef CONFIG_BLK_DEV_INITRD
 920         if (initrd_start && !initrd_below_start_ok &&
 921             page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
 922                 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
 923                     page_to_pfn(virt_to_page((void *)initrd_start)),
 924                     min_low_pfn);
 925                 initrd_start = 0;
 926         }
 927 #endif
 928         setup_per_cpu_pageset();
 929         numa_policy_init();
 930         acpi_early_init();
 931         if (late_time_init)
 932                 late_time_init();
 933         sched_clock_init();
 934         calibrate_delay();
 935         pid_idr_init();
 936         anon_vma_init();
 937 #ifdef CONFIG_X86
 938         if (efi_enabled(EFI_RUNTIME_SERVICES))
 939                 efi_enter_virtual_mode();
 940 #endif
 941         thread_stack_cache_init();
 942         cred_init();
 943         fork_init();
 944         proc_caches_init();
 945         uts_ns_init();
 946         buffer_init();
 947         key_init();
 948         security_init();
 949         dbg_late_init();
 950         vfs_caches_init();
 951         pagecache_init();
 952         signals_init();
 953         seq_file_init();
 954         proc_root_init();
 955         nsfs_init();
 956         cpuset_init();
 957         cgroup_init();
 958         taskstats_init_early();
 959         delayacct_init();
 960
 961         poking_init();
 962         check_bugs();
 963
 964         acpi_subsystem_init();
 965         arch_post_acpi_subsys_init();
 966         sfi_init_late();
 967
 968         /* Do the rest non-__init'ed, we're now alive */
 969         arch_call_rest_init();
 970 }
 971
 972 /* Call all constructor functions linked into the kernel. */
 973 static void __init do_ctors(void)
 974 {
 975 #ifdef CONFIG_CONSTRUCTORS
 976         ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
 977
 978         for (; fn < (ctor_fn_t *) __ctors_end; fn++)
 979                 (*fn)();
 980 #endif
 981 }
 982
 983 #ifdef CONFIG_KALLSYMS
 984 struct blacklist_entry {
 985         struct list_head next;
 986         char *buf;
 987 };
 988
 989 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
 990
 991 static int __init initcall_blacklist(char *str)
 992 {
 993         char *str_entry;
 994         struct blacklist_entry *entry;
 995
 996         /* str argument is a comma-separated list of functions */
 997         do {
 998                 str_entry = strsep(&str, ",");
 999                 if (str_entry) {
1000                         pr_debug("blacklisting initcall %s\n", str_entry);
1001                         entry = memblock_alloc(sizeof(*entry),
1002                                                SMP_CACHE_BYTES);
1003                         if (!entry)
1004                                 panic("%s: Failed to allocate %zu bytes\n",
1005                                       __func__, sizeof(*entry));
1006                         entry->buf = memblock_alloc(strlen(str_entry) + 1,
1007                                                     SMP_CACHE_BYTES);
1008                         if (!entry->buf)
1009                                 panic("%s: Failed to allocate %zu bytes\n",
1010                                       __func__, strlen(str_entry) + 1);
1011                         strcpy(entry->buf, str_entry);
1012                         list_add(&entry->next, &blacklisted_initcalls);
1013                 }
1014         } while (str_entry);
1015
1016         return 0;
1017 }
1018
1019 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1020 {
1021         struct blacklist_entry *entry;
1022         char fn_name[KSYM_SYMBOL_LEN];
1023         unsigned long addr;
1024
1025         if (list_empty(&blacklisted_initcalls))
1026                 return false;
1027
1028         addr = (unsigned long) dereference_function_descriptor(fn);
1029         sprint_symbol_no_offset(fn_name, addr);
1030
1031         /*
1032          * fn will be "function_name [module_name]" where [module_name] is not
1033          * displayed for built-in init functions.  Strip off the [module_name].
1034          */
1035         strreplace(fn_name, ' ', '\0');
1036
1037         list_for_each_entry(entry, &blacklisted_initcalls, next) {
1038                 if (!strcmp(fn_name, entry->buf)) {
1039                         pr_debug("initcall %s blacklisted\n", fn_name);
1040                         return true;
1041                 }
1042         }
1043
1044         return false;
1045 }
1046 #else
1047 static int __init initcall_blacklist(char *str)
1048 {
1049         pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1050         return 0;
1051 }
1052
1053 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1054 {
1055         return false;
1056 }
1057 #endif
1058 __setup("initcall_blacklist=", initcall_blacklist);
1059
1060 static __init_or_module void
1061 trace_initcall_start_cb(void *data, initcall_t fn)
1062 {
1063         ktime_t *calltime = (ktime_t *)data;
1064
1065         printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1066         *calltime = ktime_get();
1067 }
1068
1069 static __init_or_module void
1070 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1071 {
1072         ktime_t *calltime = (ktime_t *)data;
1073         ktime_t delta, rettime;
1074         unsigned long long duration;
1075
1076         rettime = ktime_get();
1077         delta = ktime_sub(rettime, *calltime);
1078         duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1079         printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1080                  fn, ret, duration);
1081 }
1082
1083 static ktime_t initcall_calltime;
1084
1085 #ifdef TRACEPOINTS_ENABLED
1086 static void __init initcall_debug_enable(void)
1087 {
1088         int ret;
1089
1090         ret = register_trace_initcall_start(trace_initcall_start_cb,
1091                                             &initcall_calltime);
1092         ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1093                                               &initcall_calltime);
1094         WARN(ret, "Failed to register initcall tracepoints\n");
1095 }
1096 # define do_trace_initcall_start        trace_initcall_start
1097 # define do_trace_initcall_finish       trace_initcall_finish
1098 #else
1099 static inline void do_trace_initcall_start(initcall_t fn)
1100 {
1101         if (!initcall_debug)
1102                 return;
1103         trace_initcall_start_cb(&initcall_calltime, fn);
1104 }
1105 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1106 {
1107         if (!initcall_debug)
1108                 return;
1109         trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1110 }
1111 #endif /* !TRACEPOINTS_ENABLED */
1112
1113 int __init_or_module do_one_initcall(initcall_t fn)
1114 {
1115         int count = preempt_count();
1116         char msgbuf[64];
1117         int ret;
1118
1119         if (initcall_blacklisted(fn))
1120                 return -EPERM;
1121
1122         do_trace_initcall_start(fn);
1123         ret = fn();
1124         do_trace_initcall_finish(fn, ret);
1125
1126         msgbuf[0] = 0;
1127
1128         if (preempt_count() != count) {
1129                 sprintf(msgbuf, "preemption imbalance ");
1130                 preempt_count_set(count);
1131         }
1132         if (irqs_disabled()) {
1133                 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1134                 local_irq_enable();
1135         }
1136         WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1137
1138         add_latent_entropy();
1139         return ret;
1140 }
1141
1142
1143 extern initcall_entry_t __initcall_start[];
1144 extern initcall_entry_t __initcall0_start[];
1145 extern initcall_entry_t __initcall1_start[];
1146 extern initcall_entry_t __initcall2_start[];
1147 extern initcall_entry_t __initcall3_start[];
1148 extern initcall_entry_t __initcall4_start[];
1149 extern initcall_entry_t __initcall5_start[];
1150 extern initcall_entry_t __initcall6_start[];
1151 extern initcall_entry_t __initcall7_start[];
1152 extern initcall_entry_t __initcall_end[];
1153
1154 static initcall_entry_t *initcall_levels[] __initdata = {
1155         __initcall0_start,
1156         __initcall1_start,
1157         __initcall2_start,
1158         __initcall3_start,
1159         __initcall4_start,
1160         __initcall5_start,
1161         __initcall6_start,
1162         __initcall7_start,
1163         __initcall_end,
1164 };
1165
1166 /* Keep these in sync with initcalls in include/linux/init.h */
1167 static const char *initcall_level_names[] __initdata = {
1168         "pure",
1169         "core",
1170         "postcore",
1171         "arch",
1172         "subsys",
1173         "fs",
1174         "device",
1175         "late",
1176 };
1177
1178 static int __init ignore_unknown_bootoption(char *param, char *val,
1179                                const char *unused, void *arg)
1180 {
1181         return 0;
1182 }
1183
1184 static void __init do_initcall_level(int level, char *command_line)
1185 {
1186         initcall_entry_t *fn;
1187
1188         parse_args(initcall_level_names[level],
1189                    command_line, __start___param,
1190                    __stop___param - __start___param,
1191                    level, level,
1192                    NULL, ignore_unknown_bootoption);
1193
1194         trace_initcall_level(initcall_level_names[level]);
1195         for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1196                 do_one_initcall(initcall_from_entry(fn));
1197 }
1198
1199 static void __init do_initcalls(void)
1200 {
1201         int level;
1202         size_t len = strlen(saved_command_line) + 1;
1203         char *command_line;
1204
1205         command_line = kzalloc(len, GFP_KERNEL);
1206         if (!command_line)
1207                 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1208
1209         for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1210                 /* Parser modifies command_line, restore it each time */
1211                 strcpy(command_line, saved_command_line);
1212                 do_initcall_level(level, command_line);
1213         }
1214
1215         kfree(command_line);
1216 }
1217
1218 /*
1219  * Ok, the machine is now initialized. None of the devices
1220  * have been touched yet, but the CPU subsystem is up and
1221  * running, and memory and process management works.
1222  *
1223  * Now we can finally start doing some real work..
1224  */
1225 static void __init do_basic_setup(void)
1226 {
1227         cpuset_init_smp();
1228         driver_init();
1229         init_irq_proc();
1230         do_ctors();
1231         usermodehelper_enable();
1232         do_initcalls();
1233 }
1234
1235 static void __init do_pre_smp_initcalls(void)
1236 {
1237         initcall_entry_t *fn;
1238
1239         trace_initcall_level("early");
1240         for (fn = __initcall_start; fn < __initcall0_start; fn++)
1241                 do_one_initcall(initcall_from_entry(fn));
1242 }
1243
1244 static int run_init_process(const char *init_filename)
1245 {
1246         const char *const *p;
1247
1248         argv_init[0] = init_filename;
1249         pr_info("Run %s as init process\n", init_filename);
1250         pr_debug("  with arguments:\n");
1251         for (p = argv_init; *p; p++)
1252                 pr_debug("    %s\n", *p);
1253         pr_debug("  with environment:\n");
1254         for (p = envp_init; *p; p++)
1255                 pr_debug("    %s\n", *p);
1256         return do_execve(getname_kernel(init_filename),
1257                 (const char __user *const __user *)argv_init,
1258                 (const char __user *const __user *)envp_init);
1259 }
1260
1261 static int try_to_run_init_process(const char *init_filename)
1262 {
1263         int ret;
1264
1265         ret = run_init_process(init_filename);
1266
1267         if (ret && ret != -ENOENT) {
1268                 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1269                        init_filename, ret);
1270         }
1271
1272         return ret;
1273 }
1274
1275 static noinline void __init kernel_init_freeable(void);
1276
1277 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1278 bool rodata_enabled __ro_after_init = true;
1279 static int __init set_debug_rodata(char *str)
1280 {
1281         return strtobool(str, &rodata_enabled);
1282 }
1283 __setup("rodata=", set_debug_rodata);
1284 #endif
1285
1286 #ifdef CONFIG_STRICT_KERNEL_RWX
1287 static void mark_readonly(void)
1288 {
1289         if (rodata_enabled) {
1290                 /*
1291                  * load_module() results in W+X mappings, which are cleaned
1292                  * up with call_rcu().  Let's make sure that queued work is
1293                  * flushed so that we don't hit false positives looking for
1294                  * insecure pages which are W+X.
1295                  */
1296                 rcu_barrier();
1297                 mark_rodata_ro();
1298                 rodata_test();
1299         } else
1300                 pr_info("Kernel memory protection disabled.\n");
1301 }
1302 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1303 static inline void mark_readonly(void)
1304 {
1305         pr_warn("Kernel memory protection not selected by kernel config.\n");
1306 }
1307 #else
1308 static inline void mark_readonly(void)
1309 {
1310         pr_warn("This architecture does not have kernel memory protection.\n");
1311 }
1312 #endif
1313
1314 void __weak free_initmem(void)
1315 {
1316         free_initmem_default(POISON_FREE_INITMEM);
1317 }
1318
1319 static int __ref kernel_init(void *unused)
1320 {
1321         int ret;
1322
1323         kernel_init_freeable();
1324         /* need to finish all async __init code before freeing the memory */
1325         async_synchronize_full();
1326         ftrace_free_init_mem();
1327         free_initmem();
1328         mark_readonly();
1329
1330         /*
1331          * Kernel mappings are now finalized - update the userspace page-table
1332          * to finalize PTI.
1333          */
1334         pti_finalize();
1335
1336         system_state = SYSTEM_RUNNING;
1337         numa_default_policy();
1338
1339         rcu_end_inkernel_boot();
1340
1341         if (ramdisk_execute_command) {
1342                 ret = run_init_process(ramdisk_execute_command);
1343                 if (!ret)
1344                         return 0;
1345                 pr_err("Failed to execute %s (error %d)\n",
1346                        ramdisk_execute_command, ret);
1347         }
1348
1349         /*
1350          * We try each of these until one succeeds.
1351          *
1352          * The Bourne shell can be used instead of init if we are
1353          * trying to recover a really broken machine.
1354          */
1355         if (execute_command) {
1356                 ret = run_init_process(execute_command);
1357                 if (!ret)
1358                         return 0;
1359                 panic("Requested init %s failed (error %d).",
1360                       execute_command, ret);
1361         }
1362         if (!try_to_run_init_process("/sbin/init") ||
1363             !try_to_run_init_process("/etc/init") ||
1364             !try_to_run_init_process("/bin/init") ||
1365             !try_to_run_init_process("/bin/sh"))
1366                 return 0;
1367
1368         panic("No working init found.  Try passing init= option to kernel. "
1369               "See Linux Documentation/admin-guide/init.rst for guidance.");
1370 }
1371
1372 void console_on_rootfs(void)
1373 {
1374         /* Open the /dev/console as stdin, this should never fail */
1375         if (ksys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1376                 pr_err("Warning: unable to open an initial console.\n");
1377
1378         /* create stdout/stderr */
1379         (void) ksys_dup(0);
1380         (void) ksys_dup(0);
1381 }
1382
1383 static noinline void __init kernel_init_freeable(void)
1384 {
1385         /*
1386          * Wait until kthreadd is all set-up.
1387          */
1388         wait_for_completion(&kthreadd_done);
1389
1390         /* Now the scheduler is fully set up and can do blocking allocations */
1391         gfp_allowed_mask = __GFP_BITS_MASK;
1392
1393         /*
1394          * init can allocate pages on any node
1395          */
1396         set_mems_allowed(node_states[N_MEMORY]);
1397
1398         cad_pid = task_pid(current);
1399
1400         smp_prepare_cpus(setup_max_cpus);
1401
1402         workqueue_init();
1403
1404         init_mm_internals();
1405
1406         do_pre_smp_initcalls();
1407         lockup_detector_init();
1408
1409         smp_init();
1410         sched_init_smp();
1411
1412         page_alloc_init_late();
1413         /* Initialize page ext after all struct pages are initialized. */
1414         page_ext_init();
1415
1416         do_basic_setup();
1417
1418         console_on_rootfs();
1419
1420         /*
1421          * check if there is an early userspace init.  If yes, let it do all
1422          * the work
1423          */
1424
1425         if (!ramdisk_execute_command)
1426                 ramdisk_execute_command = "/init";
1427
1428         if (ksys_access((const char __user *)
1429                         ramdisk_execute_command, 0) != 0) {
1430                 ramdisk_execute_command = NULL;
1431                 prepare_namespace();
1432         }
1433
1434         /*
1435          * Ok, we have completed the initial bootup, and
1436          * we're essentially up and running. Get rid of the
1437          * initmem segments and start the user-mode stuff..
1438          *
1439          * rootfs is available now, try loading the public keys
1440          * and default modules
1441          */
1442
1443         integrity_load_keys();
1444 }