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/srcu.h> 46 #include <linux/moduleparam.h> 47 #include <linux/kallsyms.h> 48 #include <linux/buildid.h> 49 #include <linux/writeback.h> 50 #include <linux/cpu.h> 51 #include <linux/cpuset.h> 52 #include <linux/cgroup.h> 53 #include <linux/efi.h> 54 #include <linux/tick.h> 55 #include <linux/sched/isolation.h> 56 #include <linux/interrupt.h> 57 #include <linux/taskstats_kern.h> 58 #include <linux/delayacct.h> 59 #include <linux/unistd.h> 60 #include <linux/utsname.h> 61 #include <linux/rmap.h> 62 #include <linux/mempolicy.h> 63 #include <linux/key.h> 64 #include <linux/page_ext.h> 65 #include <linux/debug_locks.h> 66 #include <linux/debugobjects.h> 67 #include <linux/lockdep.h> 68 #include <linux/kmemleak.h> 69 #include <linux/padata.h> 70 #include <linux/pid_namespace.h> 71 #include <linux/device/driver.h> 72 #include <linux/kthread.h> 73 #include <linux/sched.h> 74 #include <linux/sched/init.h> 75 #include <linux/signal.h> 76 #include <linux/idr.h> 77 #include <linux/kgdb.h> 78 #include <linux/ftrace.h> 79 #include <linux/async.h> 80 #include <linux/shmem_fs.h> 81 #include <linux/slab.h> 82 #include <linux/perf_event.h> 83 #include <linux/ptrace.h> 84 #include <linux/pti.h> 85 #include <linux/blkdev.h> 86 #include <linux/elevator.h> 87 #include <linux/sched/clock.h> 88 #include <linux/sched/task.h> 89 #include <linux/sched/task_stack.h> 90 #include <linux/context_tracking.h> 91 #include <linux/random.h> 92 #include <linux/list.h> 93 #include <linux/integrity.h> 94 #include <linux/proc_ns.h> 95 #include <linux/io.h> 96 #include <linux/cache.h> 97 #include <linux/rodata_test.h> 98 #include <linux/jump_label.h> 99 #include <linux/mem_encrypt.h> 100 #include <linux/kcsan.h> 101 #include <linux/init_syscalls.h> 102 #include <linux/stackdepot.h> 103 104 #include <asm/io.h> 105 #include <asm/bugs.h> 106 #include <asm/setup.h> 107 #include <asm/sections.h> 108 #include <asm/cacheflush.h> 109 110 #define CREATE_TRACE_POINTS 111 #include <trace/events/initcall.h> 112 113 #include <kunit/test.h> 114 115 static int kernel_init(void *); 116 117 extern void init_IRQ(void); 118 extern void radix_tree_init(void); 119 120 /* 121 * Debug helper: via this flag we know that we are in 'early bootup code' 122 * where only the boot processor is running with IRQ disabled. This means 123 * two things - IRQ must not be enabled before the flag is cleared and some 124 * operations which are not allowed with IRQ disabled are allowed while the 125 * flag is set. 126 */ 127 bool early_boot_irqs_disabled __read_mostly; 128 129 enum system_states system_state __read_mostly; 130 EXPORT_SYMBOL(system_state); 131 132 /* 133 * Boot command-line arguments 134 */ 135 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT 136 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT 137 138 extern void time_init(void); 139 /* Default late time init is NULL. archs can override this later. */ 140 void (*__initdata late_time_init)(void); 141 142 /* Untouched command line saved by arch-specific code. */ 143 char __initdata boot_command_line[COMMAND_LINE_SIZE]; 144 /* Untouched saved command line (eg. for /proc) */ 145 char *saved_command_line; 146 /* Command line for parameter parsing */ 147 static char *static_command_line; 148 /* Untouched extra command line */ 149 static char *extra_command_line; 150 /* Extra init arguments */ 151 static char *extra_init_args; 152 153 #ifdef CONFIG_BOOT_CONFIG 154 /* Is bootconfig on command line? */ 155 static bool bootconfig_found; 156 static size_t initargs_offs; 157 #else 158 # define bootconfig_found false 159 # define initargs_offs 0 160 #endif 161 162 static char *execute_command; 163 static char *ramdisk_execute_command = "/init"; 164 165 /* 166 * Used to generate warnings if static_key manipulation functions are used 167 * before jump_label_init is called. 168 */ 169 bool static_key_initialized __read_mostly; 170 EXPORT_SYMBOL_GPL(static_key_initialized); 171 172 /* 173 * If set, this is an indication to the drivers that reset the underlying 174 * device before going ahead with the initialization otherwise driver might 175 * rely on the BIOS and skip the reset operation. 176 * 177 * This is useful if kernel is booting in an unreliable environment. 178 * For ex. kdump situation where previous kernel has crashed, BIOS has been 179 * skipped and devices will be in unknown state. 180 */ 181 unsigned int reset_devices; 182 EXPORT_SYMBOL(reset_devices); 183 184 static int __init set_reset_devices(char *str) 185 { 186 reset_devices = 1; 187 return 1; 188 } 189 190 __setup("reset_devices", set_reset_devices); 191 192 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; 193 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; 194 static const char *panic_later, *panic_param; 195 196 extern const struct obs_kernel_param __setup_start[], __setup_end[]; 197 198 static bool __init obsolete_checksetup(char *line) 199 { 200 const struct obs_kernel_param *p; 201 bool had_early_param = false; 202 203 p = __setup_start; 204 do { 205 int n = strlen(p->str); 206 if (parameqn(line, p->str, n)) { 207 if (p->early) { 208 /* Already done in parse_early_param? 209 * (Needs exact match on param part). 210 * Keep iterating, as we can have early 211 * params and __setups of same names 8( */ 212 if (line[n] == '\0' || line[n] == '=') 213 had_early_param = true; 214 } else if (!p->setup_func) { 215 pr_warn("Parameter %s is obsolete, ignored\n", 216 p->str); 217 return true; 218 } else if (p->setup_func(line + n)) 219 return true; 220 } 221 p++; 222 } while (p < __setup_end); 223 224 return had_early_param; 225 } 226 227 /* 228 * This should be approx 2 Bo*oMips to start (note initial shift), and will 229 * still work even if initially too large, it will just take slightly longer 230 */ 231 unsigned long loops_per_jiffy = (1<<12); 232 EXPORT_SYMBOL(loops_per_jiffy); 233 234 static int __init debug_kernel(char *str) 235 { 236 console_loglevel = CONSOLE_LOGLEVEL_DEBUG; 237 return 0; 238 } 239 240 static int __init quiet_kernel(char *str) 241 { 242 console_loglevel = CONSOLE_LOGLEVEL_QUIET; 243 return 0; 244 } 245 246 early_param("debug", debug_kernel); 247 early_param("quiet", quiet_kernel); 248 249 static int __init loglevel(char *str) 250 { 251 int newlevel; 252 253 /* 254 * Only update loglevel value when a correct setting was passed, 255 * to prevent blind crashes (when loglevel being set to 0) that 256 * are quite hard to debug 257 */ 258 if (get_option(&str, &newlevel)) { 259 console_loglevel = newlevel; 260 return 0; 261 } 262 263 return -EINVAL; 264 } 265 266 early_param("loglevel", loglevel); 267 268 #ifdef CONFIG_BLK_DEV_INITRD 269 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum) 270 { 271 u32 size, csum; 272 char *data; 273 u32 *hdr; 274 int i; 275 276 if (!initrd_end) 277 return NULL; 278 279 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN; 280 /* 281 * Since Grub may align the size of initrd to 4, we must 282 * check the preceding 3 bytes as well. 283 */ 284 for (i = 0; i < 4; i++) { 285 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN)) 286 goto found; 287 data--; 288 } 289 return NULL; 290 291 found: 292 hdr = (u32 *)(data - 8); 293 size = le32_to_cpu(hdr[0]); 294 csum = le32_to_cpu(hdr[1]); 295 296 data = ((void *)hdr) - size; 297 if ((unsigned long)data < initrd_start) { 298 pr_err("bootconfig size %d is greater than initrd size %ld\n", 299 size, initrd_end - initrd_start); 300 return NULL; 301 } 302 303 /* Remove bootconfig from initramfs/initrd */ 304 initrd_end = (unsigned long)data; 305 if (_size) 306 *_size = size; 307 if (_csum) 308 *_csum = csum; 309 310 return data; 311 } 312 #else 313 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum) 314 { 315 return NULL; 316 } 317 #endif 318 319 #ifdef CONFIG_BOOT_CONFIG 320 321 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata; 322 323 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0) 324 325 static int __init xbc_snprint_cmdline(char *buf, size_t size, 326 struct xbc_node *root) 327 { 328 struct xbc_node *knode, *vnode; 329 char *end = buf + size; 330 const char *val; 331 int ret; 332 333 xbc_node_for_each_key_value(root, knode, val) { 334 ret = xbc_node_compose_key_after(root, knode, 335 xbc_namebuf, XBC_KEYLEN_MAX); 336 if (ret < 0) 337 return ret; 338 339 vnode = xbc_node_get_child(knode); 340 if (!vnode) { 341 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf); 342 if (ret < 0) 343 return ret; 344 buf += ret; 345 continue; 346 } 347 xbc_array_for_each_value(vnode, val) { 348 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ", 349 xbc_namebuf, val); 350 if (ret < 0) 351 return ret; 352 buf += ret; 353 } 354 } 355 356 return buf - (end - size); 357 } 358 #undef rest 359 360 /* Make an extra command line under given key word */ 361 static char * __init xbc_make_cmdline(const char *key) 362 { 363 struct xbc_node *root; 364 char *new_cmdline; 365 int ret, len = 0; 366 367 root = xbc_find_node(key); 368 if (!root) 369 return NULL; 370 371 /* Count required buffer size */ 372 len = xbc_snprint_cmdline(NULL, 0, root); 373 if (len <= 0) 374 return NULL; 375 376 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES); 377 if (!new_cmdline) { 378 pr_err("Failed to allocate memory for extra kernel cmdline.\n"); 379 return NULL; 380 } 381 382 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root); 383 if (ret < 0 || ret > len) { 384 pr_err("Failed to print extra kernel cmdline.\n"); 385 return NULL; 386 } 387 388 return new_cmdline; 389 } 390 391 static int __init bootconfig_params(char *param, char *val, 392 const char *unused, void *arg) 393 { 394 if (strcmp(param, "bootconfig") == 0) { 395 bootconfig_found = true; 396 } 397 return 0; 398 } 399 400 static int __init warn_bootconfig(char *str) 401 { 402 /* The 'bootconfig' has been handled by bootconfig_params(). */ 403 return 0; 404 } 405 406 static void __init setup_boot_config(void) 407 { 408 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 409 const char *msg; 410 int pos; 411 u32 size, csum; 412 char *data, *copy, *err; 413 int ret; 414 415 /* Cut out the bootconfig data even if we have no bootconfig option */ 416 data = get_boot_config_from_initrd(&size, &csum); 417 418 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 419 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL, 420 bootconfig_params); 421 422 if (IS_ERR(err) || !bootconfig_found) 423 return; 424 425 /* parse_args() stops at the next param of '--' and returns an address */ 426 if (err) 427 initargs_offs = err - tmp_cmdline; 428 429 if (!data) { 430 pr_err("'bootconfig' found on command line, but no bootconfig found\n"); 431 return; 432 } 433 434 if (size >= XBC_DATA_MAX) { 435 pr_err("bootconfig size %d greater than max size %d\n", 436 size, XBC_DATA_MAX); 437 return; 438 } 439 440 if (xbc_calc_checksum(data, size) != csum) { 441 pr_err("bootconfig checksum failed\n"); 442 return; 443 } 444 445 copy = memblock_alloc(size + 1, SMP_CACHE_BYTES); 446 if (!copy) { 447 pr_err("Failed to allocate memory for bootconfig\n"); 448 return; 449 } 450 451 memcpy(copy, data, size); 452 copy[size] = '\0'; 453 454 ret = xbc_init(copy, &msg, &pos); 455 if (ret < 0) { 456 if (pos < 0) 457 pr_err("Failed to init bootconfig: %s.\n", msg); 458 else 459 pr_err("Failed to parse bootconfig: %s at %d.\n", 460 msg, pos); 461 } else { 462 pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret); 463 /* keys starting with "kernel." are passed via cmdline */ 464 extra_command_line = xbc_make_cmdline("kernel"); 465 /* Also, "init." keys are init arguments */ 466 extra_init_args = xbc_make_cmdline("init"); 467 } 468 return; 469 } 470 471 static void __init exit_boot_config(void) 472 { 473 xbc_destroy_all(); 474 } 475 476 #else /* !CONFIG_BOOT_CONFIG */ 477 478 static void __init setup_boot_config(void) 479 { 480 /* Remove bootconfig data from initrd */ 481 get_boot_config_from_initrd(NULL, NULL); 482 } 483 484 static int __init warn_bootconfig(char *str) 485 { 486 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n"); 487 return 0; 488 } 489 490 #define exit_boot_config() do {} while (0) 491 492 #endif /* CONFIG_BOOT_CONFIG */ 493 494 early_param("bootconfig", warn_bootconfig); 495 496 /* Change NUL term back to "=", to make "param" the whole string. */ 497 static void __init repair_env_string(char *param, char *val) 498 { 499 if (val) { 500 /* param=val or param="val"? */ 501 if (val == param+strlen(param)+1) 502 val[-1] = '='; 503 else if (val == param+strlen(param)+2) { 504 val[-2] = '='; 505 memmove(val-1, val, strlen(val)+1); 506 } else 507 BUG(); 508 } 509 } 510 511 /* Anything after -- gets handed straight to init. */ 512 static int __init set_init_arg(char *param, char *val, 513 const char *unused, void *arg) 514 { 515 unsigned int i; 516 517 if (panic_later) 518 return 0; 519 520 repair_env_string(param, val); 521 522 for (i = 0; argv_init[i]; i++) { 523 if (i == MAX_INIT_ARGS) { 524 panic_later = "init"; 525 panic_param = param; 526 return 0; 527 } 528 } 529 argv_init[i] = param; 530 return 0; 531 } 532 533 /* 534 * Unknown boot options get handed to init, unless they look like 535 * unused parameters (modprobe will find them in /proc/cmdline). 536 */ 537 static int __init unknown_bootoption(char *param, char *val, 538 const char *unused, void *arg) 539 { 540 size_t len = strlen(param); 541 542 repair_env_string(param, val); 543 544 /* Handle obsolete-style parameters */ 545 if (obsolete_checksetup(param)) 546 return 0; 547 548 /* Unused module parameter. */ 549 if (strnchr(param, len, '.')) 550 return 0; 551 552 if (panic_later) 553 return 0; 554 555 if (val) { 556 /* Environment option */ 557 unsigned int i; 558 for (i = 0; envp_init[i]; i++) { 559 if (i == MAX_INIT_ENVS) { 560 panic_later = "env"; 561 panic_param = param; 562 } 563 if (!strncmp(param, envp_init[i], len+1)) 564 break; 565 } 566 envp_init[i] = param; 567 } else { 568 /* Command line option */ 569 unsigned int i; 570 for (i = 0; argv_init[i]; i++) { 571 if (i == MAX_INIT_ARGS) { 572 panic_later = "init"; 573 panic_param = param; 574 } 575 } 576 argv_init[i] = param; 577 } 578 return 0; 579 } 580 581 static int __init init_setup(char *str) 582 { 583 unsigned int i; 584 585 execute_command = str; 586 /* 587 * In case LILO is going to boot us with default command line, 588 * it prepends "auto" before the whole cmdline which makes 589 * the shell think it should execute a script with such name. 590 * So we ignore all arguments entered _before_ init=... [MJ] 591 */ 592 for (i = 1; i < MAX_INIT_ARGS; i++) 593 argv_init[i] = NULL; 594 return 1; 595 } 596 __setup("init=", init_setup); 597 598 static int __init rdinit_setup(char *str) 599 { 600 unsigned int i; 601 602 ramdisk_execute_command = str; 603 /* See "auto" comment in init_setup */ 604 for (i = 1; i < MAX_INIT_ARGS; i++) 605 argv_init[i] = NULL; 606 return 1; 607 } 608 __setup("rdinit=", rdinit_setup); 609 610 #ifndef CONFIG_SMP 611 static const unsigned int setup_max_cpus = NR_CPUS; 612 static inline void setup_nr_cpu_ids(void) { } 613 static inline void smp_prepare_cpus(unsigned int maxcpus) { } 614 #endif 615 616 /* 617 * We need to store the untouched command line for future reference. 618 * We also need to store the touched command line since the parameter 619 * parsing is performed in place, and we should allow a component to 620 * store reference of name/value for future reference. 621 */ 622 static void __init setup_command_line(char *command_line) 623 { 624 size_t len, xlen = 0, ilen = 0; 625 626 if (extra_command_line) 627 xlen = strlen(extra_command_line); 628 if (extra_init_args) 629 ilen = strlen(extra_init_args) + 4; /* for " -- " */ 630 631 len = xlen + strlen(boot_command_line) + 1; 632 633 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES); 634 if (!saved_command_line) 635 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen); 636 637 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES); 638 if (!static_command_line) 639 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 640 641 if (xlen) { 642 /* 643 * We have to put extra_command_line before boot command 644 * lines because there could be dashes (separator of init 645 * command line) in the command lines. 646 */ 647 strcpy(saved_command_line, extra_command_line); 648 strcpy(static_command_line, extra_command_line); 649 } 650 strcpy(saved_command_line + xlen, boot_command_line); 651 strcpy(static_command_line + xlen, command_line); 652 653 if (ilen) { 654 /* 655 * Append supplemental init boot args to saved_command_line 656 * so that user can check what command line options passed 657 * to init. 658 * The order should always be 659 * " -- "[bootconfig init-param][cmdline init-param] 660 */ 661 if (initargs_offs) { 662 len = xlen + initargs_offs; 663 strcpy(saved_command_line + len, extra_init_args); 664 len += ilen - 4; /* strlen(extra_init_args) */ 665 strcpy(saved_command_line + len, 666 boot_command_line + initargs_offs - 1); 667 } else { 668 len = strlen(saved_command_line); 669 strcpy(saved_command_line + len, " -- "); 670 len += 4; 671 strcpy(saved_command_line + len, extra_init_args); 672 } 673 } 674 } 675 676 /* 677 * We need to finalize in a non-__init function or else race conditions 678 * between the root thread and the init thread may cause start_kernel to 679 * be reaped by free_initmem before the root thread has proceeded to 680 * cpu_idle. 681 * 682 * gcc-3.4 accidentally inlines this function, so use noinline. 683 */ 684 685 static __initdata DECLARE_COMPLETION(kthreadd_done); 686 687 noinline void __ref rest_init(void) 688 { 689 struct task_struct *tsk; 690 int pid; 691 692 rcu_scheduler_starting(); 693 /* 694 * We need to spawn init first so that it obtains pid 1, however 695 * the init task will end up wanting to create kthreads, which, if 696 * we schedule it before we create kthreadd, will OOPS. 697 */ 698 pid = kernel_thread(kernel_init, NULL, CLONE_FS); 699 /* 700 * Pin init on the boot CPU. Task migration is not properly working 701 * until sched_init_smp() has been run. It will set the allowed 702 * CPUs for init to the non isolated CPUs. 703 */ 704 rcu_read_lock(); 705 tsk = find_task_by_pid_ns(pid, &init_pid_ns); 706 tsk->flags |= PF_NO_SETAFFINITY; 707 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id())); 708 rcu_read_unlock(); 709 710 numa_default_policy(); 711 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); 712 rcu_read_lock(); 713 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); 714 rcu_read_unlock(); 715 716 /* 717 * Enable might_sleep() and smp_processor_id() checks. 718 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y 719 * kernel_thread() would trigger might_sleep() splats. With 720 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled 721 * already, but it's stuck on the kthreadd_done completion. 722 */ 723 system_state = SYSTEM_SCHEDULING; 724 725 complete(&kthreadd_done); 726 727 /* 728 * The boot idle thread must execute schedule() 729 * at least once to get things moving: 730 */ 731 schedule_preempt_disabled(); 732 /* Call into cpu_idle with preempt disabled */ 733 cpu_startup_entry(CPUHP_ONLINE); 734 } 735 736 /* Check for early params. */ 737 static int __init do_early_param(char *param, char *val, 738 const char *unused, void *arg) 739 { 740 const struct obs_kernel_param *p; 741 742 for (p = __setup_start; p < __setup_end; p++) { 743 if ((p->early && parameq(param, p->str)) || 744 (strcmp(param, "console") == 0 && 745 strcmp(p->str, "earlycon") == 0) 746 ) { 747 if (p->setup_func(val) != 0) 748 pr_warn("Malformed early option '%s'\n", param); 749 } 750 } 751 /* We accept everything at this stage. */ 752 return 0; 753 } 754 755 void __init parse_early_options(char *cmdline) 756 { 757 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL, 758 do_early_param); 759 } 760 761 /* Arch code calls this early on, or if not, just before other parsing. */ 762 void __init parse_early_param(void) 763 { 764 static int done __initdata; 765 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 766 767 if (done) 768 return; 769 770 /* All fall through to do_early_param. */ 771 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 772 parse_early_options(tmp_cmdline); 773 done = 1; 774 } 775 776 void __init __weak arch_post_acpi_subsys_init(void) { } 777 778 void __init __weak smp_setup_processor_id(void) 779 { 780 } 781 782 # if THREAD_SIZE >= PAGE_SIZE 783 void __init __weak thread_stack_cache_init(void) 784 { 785 } 786 #endif 787 788 void __init __weak mem_encrypt_init(void) { } 789 790 void __init __weak poking_init(void) { } 791 792 void __init __weak pgtable_cache_init(void) { } 793 794 void __init __weak trap_init(void) { } 795 796 bool initcall_debug; 797 core_param(initcall_debug, initcall_debug, bool, 0644); 798 799 #ifdef TRACEPOINTS_ENABLED 800 static void __init initcall_debug_enable(void); 801 #else 802 static inline void initcall_debug_enable(void) 803 { 804 } 805 #endif 806 807 /* Report memory auto-initialization states for this boot. */ 808 static void __init report_meminit(void) 809 { 810 const char *stack; 811 812 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN)) 813 stack = "all(pattern)"; 814 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO)) 815 stack = "all(zero)"; 816 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL)) 817 stack = "byref_all(zero)"; 818 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF)) 819 stack = "byref(zero)"; 820 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER)) 821 stack = "__user(zero)"; 822 else 823 stack = "off"; 824 825 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n", 826 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off", 827 want_init_on_free() ? "on" : "off"); 828 if (want_init_on_free()) 829 pr_info("mem auto-init: clearing system memory may take some time...\n"); 830 } 831 832 /* 833 * Set up kernel memory allocators 834 */ 835 static void __init mm_init(void) 836 { 837 /* 838 * page_ext requires contiguous pages, 839 * bigger than MAX_ORDER unless SPARSEMEM. 840 */ 841 page_ext_init_flatmem(); 842 init_mem_debugging_and_hardening(); 843 kfence_alloc_pool(); 844 report_meminit(); 845 stack_depot_init(); 846 mem_init(); 847 mem_init_print_info(); 848 /* page_owner must be initialized after buddy is ready */ 849 page_ext_init_flatmem_late(); 850 kmem_cache_init(); 851 kmemleak_init(); 852 pgtable_init(); 853 debug_objects_mem_init(); 854 vmalloc_init(); 855 /* Should be run before the first non-init thread is created */ 856 init_espfix_bsp(); 857 /* Should be run after espfix64 is set up. */ 858 pti_init(); 859 } 860 861 #ifdef CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET 862 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT, 863 randomize_kstack_offset); 864 DEFINE_PER_CPU(u32, kstack_offset); 865 866 static int __init early_randomize_kstack_offset(char *buf) 867 { 868 int ret; 869 bool bool_result; 870 871 ret = kstrtobool(buf, &bool_result); 872 if (ret) 873 return ret; 874 875 if (bool_result) 876 static_branch_enable(&randomize_kstack_offset); 877 else 878 static_branch_disable(&randomize_kstack_offset); 879 return 0; 880 } 881 early_param("randomize_kstack_offset", early_randomize_kstack_offset); 882 #endif 883 884 void __init __weak arch_call_rest_init(void) 885 { 886 rest_init(); 887 } 888 889 static void __init print_unknown_bootoptions(void) 890 { 891 char *unknown_options; 892 char *end; 893 const char *const *p; 894 size_t len; 895 896 if (panic_later || (!argv_init[1] && !envp_init[2])) 897 return; 898 899 /* 900 * Determine how many options we have to print out, plus a space 901 * before each 902 */ 903 len = 1; /* null terminator */ 904 for (p = &argv_init[1]; *p; p++) { 905 len++; 906 len += strlen(*p); 907 } 908 for (p = &envp_init[2]; *p; p++) { 909 len++; 910 len += strlen(*p); 911 } 912 913 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES); 914 if (!unknown_options) { 915 pr_err("%s: Failed to allocate %zu bytes\n", 916 __func__, len); 917 return; 918 } 919 end = unknown_options; 920 921 for (p = &argv_init[1]; *p; p++) 922 end += sprintf(end, " %s", *p); 923 for (p = &envp_init[2]; *p; p++) 924 end += sprintf(end, " %s", *p); 925 926 pr_notice("Unknown command line parameters:%s\n", unknown_options); 927 memblock_free_ptr(unknown_options, len); 928 } 929 930 asmlinkage __visible void __init __no_sanitize_address start_kernel(void) 931 { 932 char *command_line; 933 char *after_dashes; 934 935 set_task_stack_end_magic(&init_task); 936 smp_setup_processor_id(); 937 debug_objects_early_init(); 938 init_vmlinux_build_id(); 939 940 cgroup_init_early(); 941 942 local_irq_disable(); 943 early_boot_irqs_disabled = true; 944 945 /* 946 * Interrupts are still disabled. Do necessary setups, then 947 * enable them. 948 */ 949 boot_cpu_init(); 950 page_address_init(); 951 pr_notice("%s", linux_banner); 952 early_security_init(); 953 setup_arch(&command_line); 954 setup_boot_config(); 955 setup_command_line(command_line); 956 setup_nr_cpu_ids(); 957 setup_per_cpu_areas(); 958 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ 959 boot_cpu_hotplug_init(); 960 961 build_all_zonelists(NULL); 962 page_alloc_init(); 963 964 pr_notice("Kernel command line: %s\n", saved_command_line); 965 /* parameters may set static keys */ 966 jump_label_init(); 967 parse_early_param(); 968 after_dashes = parse_args("Booting kernel", 969 static_command_line, __start___param, 970 __stop___param - __start___param, 971 -1, -1, NULL, &unknown_bootoption); 972 print_unknown_bootoptions(); 973 if (!IS_ERR_OR_NULL(after_dashes)) 974 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1, 975 NULL, set_init_arg); 976 if (extra_init_args) 977 parse_args("Setting extra init args", extra_init_args, 978 NULL, 0, -1, -1, NULL, set_init_arg); 979 980 /* 981 * These use large bootmem allocations and must precede 982 * kmem_cache_init() 983 */ 984 setup_log_buf(0); 985 vfs_caches_init_early(); 986 sort_main_extable(); 987 trap_init(); 988 mm_init(); 989 990 ftrace_init(); 991 992 /* trace_printk can be enabled here */ 993 early_trace_init(); 994 995 /* 996 * Set up the scheduler prior starting any interrupts (such as the 997 * timer interrupt). Full topology setup happens at smp_init() 998 * time - but meanwhile we still have a functioning scheduler. 999 */ 1000 sched_init(); 1001 1002 if (WARN(!irqs_disabled(), 1003 "Interrupts were enabled *very* early, fixing it\n")) 1004 local_irq_disable(); 1005 radix_tree_init(); 1006 1007 /* 1008 * Set up housekeeping before setting up workqueues to allow the unbound 1009 * workqueue to take non-housekeeping into account. 1010 */ 1011 housekeeping_init(); 1012 1013 /* 1014 * Allow workqueue creation and work item queueing/cancelling 1015 * early. Work item execution depends on kthreads and starts after 1016 * workqueue_init(). 1017 */ 1018 workqueue_init_early(); 1019 1020 rcu_init(); 1021 1022 /* Trace events are available after this */ 1023 trace_init(); 1024 1025 if (initcall_debug) 1026 initcall_debug_enable(); 1027 1028 context_tracking_init(); 1029 /* init some links before init_ISA_irqs() */ 1030 early_irq_init(); 1031 init_IRQ(); 1032 tick_init(); 1033 rcu_init_nohz(); 1034 init_timers(); 1035 srcu_init(); 1036 hrtimers_init(); 1037 softirq_init(); 1038 timekeeping_init(); 1039 kfence_init(); 1040 1041 /* 1042 * For best initial stack canary entropy, prepare it after: 1043 * - setup_arch() for any UEFI RNG entropy and boot cmdline access 1044 * - timekeeping_init() for ktime entropy used in rand_initialize() 1045 * - rand_initialize() to get any arch-specific entropy like RDRAND 1046 * - add_latent_entropy() to get any latent entropy 1047 * - adding command line entropy 1048 */ 1049 rand_initialize(); 1050 add_latent_entropy(); 1051 add_device_randomness(command_line, strlen(command_line)); 1052 boot_init_stack_canary(); 1053 1054 time_init(); 1055 perf_event_init(); 1056 profile_init(); 1057 call_function_init(); 1058 WARN(!irqs_disabled(), "Interrupts were enabled early\n"); 1059 1060 early_boot_irqs_disabled = false; 1061 local_irq_enable(); 1062 1063 kmem_cache_init_late(); 1064 1065 /* 1066 * HACK ALERT! This is early. We're enabling the console before 1067 * we've done PCI setups etc, and console_init() must be aware of 1068 * this. But we do want output early, in case something goes wrong. 1069 */ 1070 console_init(); 1071 if (panic_later) 1072 panic("Too many boot %s vars at `%s'", panic_later, 1073 panic_param); 1074 1075 lockdep_init(); 1076 1077 /* 1078 * Need to run this when irqs are enabled, because it wants 1079 * to self-test [hard/soft]-irqs on/off lock inversion bugs 1080 * too: 1081 */ 1082 locking_selftest(); 1083 1084 /* 1085 * This needs to be called before any devices perform DMA 1086 * operations that might use the SWIOTLB bounce buffers. It will 1087 * mark the bounce buffers as decrypted so that their usage will 1088 * not cause "plain-text" data to be decrypted when accessed. 1089 */ 1090 mem_encrypt_init(); 1091 1092 #ifdef CONFIG_BLK_DEV_INITRD 1093 if (initrd_start && !initrd_below_start_ok && 1094 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { 1095 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", 1096 page_to_pfn(virt_to_page((void *)initrd_start)), 1097 min_low_pfn); 1098 initrd_start = 0; 1099 } 1100 #endif 1101 setup_per_cpu_pageset(); 1102 numa_policy_init(); 1103 acpi_early_init(); 1104 if (late_time_init) 1105 late_time_init(); 1106 sched_clock_init(); 1107 calibrate_delay(); 1108 pid_idr_init(); 1109 anon_vma_init(); 1110 #ifdef CONFIG_X86 1111 if (efi_enabled(EFI_RUNTIME_SERVICES)) 1112 efi_enter_virtual_mode(); 1113 #endif 1114 thread_stack_cache_init(); 1115 cred_init(); 1116 fork_init(); 1117 proc_caches_init(); 1118 uts_ns_init(); 1119 key_init(); 1120 security_init(); 1121 dbg_late_init(); 1122 vfs_caches_init(); 1123 pagecache_init(); 1124 signals_init(); 1125 seq_file_init(); 1126 proc_root_init(); 1127 nsfs_init(); 1128 cpuset_init(); 1129 cgroup_init(); 1130 taskstats_init_early(); 1131 delayacct_init(); 1132 1133 poking_init(); 1134 check_bugs(); 1135 1136 acpi_subsystem_init(); 1137 arch_post_acpi_subsys_init(); 1138 kcsan_init(); 1139 1140 /* Do the rest non-__init'ed, we're now alive */ 1141 arch_call_rest_init(); 1142 1143 prevent_tail_call_optimization(); 1144 } 1145 1146 /* Call all constructor functions linked into the kernel. */ 1147 static void __init do_ctors(void) 1148 { 1149 /* 1150 * For UML, the constructors have already been called by the 1151 * normal setup code as it's just a normal ELF binary, so we 1152 * cannot do it again - but we do need CONFIG_CONSTRUCTORS 1153 * even on UML for modules. 1154 */ 1155 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML) 1156 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; 1157 1158 for (; fn < (ctor_fn_t *) __ctors_end; fn++) 1159 (*fn)(); 1160 #endif 1161 } 1162 1163 #ifdef CONFIG_KALLSYMS 1164 struct blacklist_entry { 1165 struct list_head next; 1166 char *buf; 1167 }; 1168 1169 static __initdata_or_module LIST_HEAD(blacklisted_initcalls); 1170 1171 static int __init initcall_blacklist(char *str) 1172 { 1173 char *str_entry; 1174 struct blacklist_entry *entry; 1175 1176 /* str argument is a comma-separated list of functions */ 1177 do { 1178 str_entry = strsep(&str, ","); 1179 if (str_entry) { 1180 pr_debug("blacklisting initcall %s\n", str_entry); 1181 entry = memblock_alloc(sizeof(*entry), 1182 SMP_CACHE_BYTES); 1183 if (!entry) 1184 panic("%s: Failed to allocate %zu bytes\n", 1185 __func__, sizeof(*entry)); 1186 entry->buf = memblock_alloc(strlen(str_entry) + 1, 1187 SMP_CACHE_BYTES); 1188 if (!entry->buf) 1189 panic("%s: Failed to allocate %zu bytes\n", 1190 __func__, strlen(str_entry) + 1); 1191 strcpy(entry->buf, str_entry); 1192 list_add(&entry->next, &blacklisted_initcalls); 1193 } 1194 } while (str_entry); 1195 1196 return 0; 1197 } 1198 1199 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1200 { 1201 struct blacklist_entry *entry; 1202 char fn_name[KSYM_SYMBOL_LEN]; 1203 unsigned long addr; 1204 1205 if (list_empty(&blacklisted_initcalls)) 1206 return false; 1207 1208 addr = (unsigned long) dereference_function_descriptor(fn); 1209 sprint_symbol_no_offset(fn_name, addr); 1210 1211 /* 1212 * fn will be "function_name [module_name]" where [module_name] is not 1213 * displayed for built-in init functions. Strip off the [module_name]. 1214 */ 1215 strreplace(fn_name, ' ', '\0'); 1216 1217 list_for_each_entry(entry, &blacklisted_initcalls, next) { 1218 if (!strcmp(fn_name, entry->buf)) { 1219 pr_debug("initcall %s blacklisted\n", fn_name); 1220 return true; 1221 } 1222 } 1223 1224 return false; 1225 } 1226 #else 1227 static int __init initcall_blacklist(char *str) 1228 { 1229 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); 1230 return 0; 1231 } 1232 1233 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1234 { 1235 return false; 1236 } 1237 #endif 1238 __setup("initcall_blacklist=", initcall_blacklist); 1239 1240 static __init_or_module void 1241 trace_initcall_start_cb(void *data, initcall_t fn) 1242 { 1243 ktime_t *calltime = (ktime_t *)data; 1244 1245 printk(KERN_DEBUG "calling %pS @ %i irqs_disabled() %d\n", fn, task_pid_nr(current), irqs_disabled()); 1246 *calltime = ktime_get(); 1247 } 1248 1249 static __init_or_module void 1250 trace_initcall_finish_cb(void *data, initcall_t fn, int ret) 1251 { 1252 ktime_t *calltime = (ktime_t *)data; 1253 ktime_t delta, rettime; 1254 unsigned long long duration; 1255 1256 rettime = ktime_get(); 1257 delta = ktime_sub(rettime, *calltime); 1258 duration = (unsigned long long) ktime_to_ns(delta) >> 10; 1259 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs, irqs_disabled() %d\n", 1260 fn, ret, duration, irqs_disabled()); 1261 } 1262 1263 static ktime_t initcall_calltime; 1264 1265 #ifdef TRACEPOINTS_ENABLED 1266 static void __init initcall_debug_enable(void) 1267 { 1268 int ret; 1269 1270 ret = register_trace_initcall_start(trace_initcall_start_cb, 1271 &initcall_calltime); 1272 ret |= register_trace_initcall_finish(trace_initcall_finish_cb, 1273 &initcall_calltime); 1274 WARN(ret, "Failed to register initcall tracepoints\n"); 1275 } 1276 # define do_trace_initcall_start trace_initcall_start 1277 # define do_trace_initcall_finish trace_initcall_finish 1278 #else 1279 static inline void do_trace_initcall_start(initcall_t fn) 1280 { 1281 if (!initcall_debug) 1282 return; 1283 trace_initcall_start_cb(&initcall_calltime, fn); 1284 } 1285 static inline void do_trace_initcall_finish(initcall_t fn, int ret) 1286 { 1287 if (!initcall_debug) 1288 return; 1289 trace_initcall_finish_cb(&initcall_calltime, fn, ret); 1290 } 1291 #endif /* !TRACEPOINTS_ENABLED */ 1292 1293 int __init_or_module do_one_initcall(initcall_t fn) 1294 { 1295 int count = preempt_count(); 1296 char msgbuf[64]; 1297 int ret; 1298 1299 if (initcall_blacklisted(fn)) 1300 return -EPERM; 1301 1302 do_trace_initcall_start(fn); 1303 ret = fn(); 1304 do_trace_initcall_finish(fn, ret); 1305 1306 msgbuf[0] = 0; 1307 1308 if (preempt_count() != count) { 1309 sprintf(msgbuf, "preemption imbalance "); 1310 preempt_count_set(count); 1311 } 1312 if (irqs_disabled()) { 1313 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); 1314 local_irq_enable(); 1315 } 1316 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf); 1317 1318 add_latent_entropy(); 1319 return ret; 1320 } 1321 1322 1323 extern initcall_entry_t __initcall_start[]; 1324 extern initcall_entry_t __initcall0_start[]; 1325 extern initcall_entry_t __initcall1_start[]; 1326 extern initcall_entry_t __initcall2_start[]; 1327 extern initcall_entry_t __initcall3_start[]; 1328 extern initcall_entry_t __initcall4_start[]; 1329 extern initcall_entry_t __initcall5_start[]; 1330 extern initcall_entry_t __initcall6_start[]; 1331 extern initcall_entry_t __initcall7_start[]; 1332 extern initcall_entry_t __initcall_end[]; 1333 1334 static initcall_entry_t *initcall_levels[] __initdata = { 1335 __initcall0_start, 1336 __initcall1_start, 1337 __initcall2_start, 1338 __initcall3_start, 1339 __initcall4_start, 1340 __initcall5_start, 1341 __initcall6_start, 1342 __initcall7_start, 1343 __initcall_end, 1344 }; 1345 1346 /* Keep these in sync with initcalls in include/linux/init.h */ 1347 static const char *initcall_level_names[] __initdata = { 1348 "pure", 1349 "core", 1350 "postcore", 1351 "arch", 1352 "subsys", 1353 "fs", 1354 "device", 1355 "late", 1356 }; 1357 1358 static int __init ignore_unknown_bootoption(char *param, char *val, 1359 const char *unused, void *arg) 1360 { 1361 return 0; 1362 } 1363 1364 static void __init do_initcall_level(int level, char *command_line) 1365 { 1366 initcall_entry_t *fn; 1367 1368 parse_args(initcall_level_names[level], 1369 command_line, __start___param, 1370 __stop___param - __start___param, 1371 level, level, 1372 NULL, ignore_unknown_bootoption); 1373 1374 trace_initcall_level(initcall_level_names[level]); 1375 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) 1376 do_one_initcall(initcall_from_entry(fn)); 1377 } 1378 1379 static void __init do_initcalls(void) 1380 { 1381 int level; 1382 size_t len = strlen(saved_command_line) + 1; 1383 char *command_line; 1384 1385 command_line = kzalloc(len, GFP_KERNEL); 1386 if (!command_line) 1387 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 1388 1389 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) { 1390 /* Parser modifies command_line, restore it each time */ 1391 strcpy(command_line, saved_command_line); 1392 do_initcall_level(level, command_line); 1393 } 1394 1395 kfree(command_line); 1396 } 1397 1398 /* 1399 * Ok, the machine is now initialized. None of the devices 1400 * have been touched yet, but the CPU subsystem is up and 1401 * running, and memory and process management works. 1402 * 1403 * Now we can finally start doing some real work.. 1404 */ 1405 static void __init do_basic_setup(void) 1406 { 1407 cpuset_init_smp(); 1408 driver_init(); 1409 init_irq_proc(); 1410 do_ctors(); 1411 do_initcalls(); 1412 } 1413 1414 static void __init do_pre_smp_initcalls(void) 1415 { 1416 initcall_entry_t *fn; 1417 1418 trace_initcall_level("early"); 1419 for (fn = __initcall_start; fn < __initcall0_start; fn++) 1420 do_one_initcall(initcall_from_entry(fn)); 1421 } 1422 1423 static int run_init_process(const char *init_filename) 1424 { 1425 const char *const *p; 1426 1427 argv_init[0] = init_filename; 1428 pr_info("Run %s as init process\n", init_filename); 1429 pr_debug(" with arguments:\n"); 1430 for (p = argv_init; *p; p++) 1431 pr_debug(" %s\n", *p); 1432 pr_debug(" with environment:\n"); 1433 for (p = envp_init; *p; p++) 1434 pr_debug(" %s\n", *p); 1435 return kernel_execve(init_filename, argv_init, envp_init); 1436 } 1437 1438 static int try_to_run_init_process(const char *init_filename) 1439 { 1440 int ret; 1441 1442 ret = run_init_process(init_filename); 1443 1444 if (ret && ret != -ENOENT) { 1445 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", 1446 init_filename, ret); 1447 } 1448 1449 return ret; 1450 } 1451 1452 static noinline void __init kernel_init_freeable(void); 1453 1454 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX) 1455 bool rodata_enabled __ro_after_init = true; 1456 static int __init set_debug_rodata(char *str) 1457 { 1458 return strtobool(str, &rodata_enabled); 1459 } 1460 __setup("rodata=", set_debug_rodata); 1461 #endif 1462 1463 #ifdef CONFIG_STRICT_KERNEL_RWX 1464 static void mark_readonly(void) 1465 { 1466 if (rodata_enabled) { 1467 /* 1468 * load_module() results in W+X mappings, which are cleaned 1469 * up with call_rcu(). Let's make sure that queued work is 1470 * flushed so that we don't hit false positives looking for 1471 * insecure pages which are W+X. 1472 */ 1473 rcu_barrier(); 1474 mark_rodata_ro(); 1475 rodata_test(); 1476 } else 1477 pr_info("Kernel memory protection disabled.\n"); 1478 } 1479 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX) 1480 static inline void mark_readonly(void) 1481 { 1482 pr_warn("Kernel memory protection not selected by kernel config.\n"); 1483 } 1484 #else 1485 static inline void mark_readonly(void) 1486 { 1487 pr_warn("This architecture does not have kernel memory protection.\n"); 1488 } 1489 #endif 1490 1491 void __weak free_initmem(void) 1492 { 1493 free_initmem_default(POISON_FREE_INITMEM); 1494 } 1495 1496 static int __ref kernel_init(void *unused) 1497 { 1498 int ret; 1499 1500 /* 1501 * Wait until kthreadd is all set-up. 1502 */ 1503 wait_for_completion(&kthreadd_done); 1504 1505 kernel_init_freeable(); 1506 /* need to finish all async __init code before freeing the memory */ 1507 async_synchronize_full(); 1508 kprobe_free_init_mem(); 1509 ftrace_free_init_mem(); 1510 kgdb_free_init_mem(); 1511 exit_boot_config(); 1512 free_initmem(); 1513 mark_readonly(); 1514 1515 /* 1516 * Kernel mappings are now finalized - update the userspace page-table 1517 * to finalize PTI. 1518 */ 1519 pti_finalize(); 1520 1521 system_state = SYSTEM_RUNNING; 1522 numa_default_policy(); 1523 1524 rcu_end_inkernel_boot(); 1525 1526 do_sysctl_args(); 1527 1528 if (ramdisk_execute_command) { 1529 ret = run_init_process(ramdisk_execute_command); 1530 if (!ret) 1531 return 0; 1532 pr_err("Failed to execute %s (error %d)\n", 1533 ramdisk_execute_command, ret); 1534 } 1535 1536 /* 1537 * We try each of these until one succeeds. 1538 * 1539 * The Bourne shell can be used instead of init if we are 1540 * trying to recover a really broken machine. 1541 */ 1542 if (execute_command) { 1543 ret = run_init_process(execute_command); 1544 if (!ret) 1545 return 0; 1546 panic("Requested init %s failed (error %d).", 1547 execute_command, ret); 1548 } 1549 1550 if (CONFIG_DEFAULT_INIT[0] != '\0') { 1551 ret = run_init_process(CONFIG_DEFAULT_INIT); 1552 if (ret) 1553 pr_err("Default init %s failed (error %d)\n", 1554 CONFIG_DEFAULT_INIT, ret); 1555 else 1556 return 0; 1557 } 1558 1559 if (!try_to_run_init_process("/sbin/init") || 1560 !try_to_run_init_process("/etc/init") || 1561 !try_to_run_init_process("/bin/init") || 1562 !try_to_run_init_process("/bin/sh")) 1563 return 0; 1564 1565 panic("No working init found. Try passing init= option to kernel. " 1566 "See Linux Documentation/admin-guide/init.rst for guidance."); 1567 } 1568 1569 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */ 1570 void __init console_on_rootfs(void) 1571 { 1572 struct file *file = filp_open("/dev/console", O_RDWR, 0); 1573 1574 if (IS_ERR(file)) { 1575 pr_err("Warning: unable to open an initial console.\n"); 1576 return; 1577 } 1578 init_dup(file); 1579 init_dup(file); 1580 init_dup(file); 1581 fput(file); 1582 } 1583 1584 static noinline void __init kernel_init_freeable(void) 1585 { 1586 /* Now the scheduler is fully set up and can do blocking allocations */ 1587 gfp_allowed_mask = __GFP_BITS_MASK; 1588 1589 /* 1590 * init can allocate pages on any node 1591 */ 1592 set_mems_allowed(node_states[N_MEMORY]); 1593 1594 cad_pid = get_pid(task_pid(current)); 1595 1596 smp_prepare_cpus(setup_max_cpus); 1597 1598 workqueue_init(); 1599 1600 init_mm_internals(); 1601 1602 rcu_init_tasks_generic(); 1603 do_pre_smp_initcalls(); 1604 lockup_detector_init(); 1605 1606 smp_init(); 1607 sched_init_smp(); 1608 1609 padata_init(); 1610 page_alloc_init_late(); 1611 /* Initialize page ext after all struct pages are initialized. */ 1612 page_ext_init(); 1613 1614 do_basic_setup(); 1615 1616 kunit_run_all_tests(); 1617 1618 wait_for_initramfs(); 1619 console_on_rootfs(); 1620 1621 /* 1622 * check if there is an early userspace init. If yes, let it do all 1623 * the work 1624 */ 1625 if (init_eaccess(ramdisk_execute_command) != 0) { 1626 ramdisk_execute_command = NULL; 1627 prepare_namespace(); 1628 } 1629 1630 /* 1631 * Ok, we have completed the initial bootup, and 1632 * we're essentially up and running. Get rid of the 1633 * initmem segments and start the user-mode stuff.. 1634 * 1635 * rootfs is available now, try loading the public keys 1636 * and default modules 1637 */ 1638 1639 integrity_load_keys(); 1640 } 1641