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/kmsan.h> 38 #include <linux/vmalloc.h> 39 #include <linux/kernel_stat.h> 40 #include <linux/start_kernel.h> 41 #include <linux/security.h> 42 #include <linux/smp.h> 43 #include <linux/profile.h> 44 #include <linux/kfence.h> 45 #include <linux/rcupdate.h> 46 #include <linux/srcu.h> 47 #include <linux/moduleparam.h> 48 #include <linux/kallsyms.h> 49 #include <linux/buildid.h> 50 #include <linux/writeback.h> 51 #include <linux/cpu.h> 52 #include <linux/cpuset.h> 53 #include <linux/cgroup.h> 54 #include <linux/efi.h> 55 #include <linux/tick.h> 56 #include <linux/sched/isolation.h> 57 #include <linux/interrupt.h> 58 #include <linux/taskstats_kern.h> 59 #include <linux/delayacct.h> 60 #include <linux/unistd.h> 61 #include <linux/utsname.h> 62 #include <linux/rmap.h> 63 #include <linux/mempolicy.h> 64 #include <linux/key.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/sched/clock.h> 87 #include <linux/sched/task.h> 88 #include <linux/sched/task_stack.h> 89 #include <linux/context_tracking.h> 90 #include <linux/random.h> 91 #include <linux/moduleloader.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/kcsan.h> 100 #include <linux/init_syscalls.h> 101 #include <linux/stackdepot.h> 102 #include <linux/randomize_kstack.h> 103 #include <linux/pidfs.h> 104 #include <linux/ptdump.h> 105 #include <net/net_namespace.h> 106 107 #include <asm/io.h> 108 #include <asm/setup.h> 109 #include <asm/sections.h> 110 #include <asm/cacheflush.h> 111 112 #define CREATE_TRACE_POINTS 113 #include <trace/events/initcall.h> 114 115 #include <kunit/test.h> 116 117 static int kernel_init(void *); 118 119 /* 120 * Debug helper: via this flag we know that we are in 'early bootup code' 121 * where only the boot processor is running with IRQ disabled. This means 122 * two things - IRQ must not be enabled before the flag is cleared and some 123 * operations which are not allowed with IRQ disabled are allowed while the 124 * flag is set. 125 */ 126 bool early_boot_irqs_disabled __read_mostly; 127 128 enum system_states system_state __read_mostly; 129 EXPORT_SYMBOL(system_state); 130 131 /* 132 * Boot command-line arguments 133 */ 134 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT 135 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT 136 137 /* Default late time init is NULL. archs can override this later. */ 138 void (*__initdata late_time_init)(void); 139 140 /* Untouched command line saved by arch-specific code. */ 141 char __initdata boot_command_line[COMMAND_LINE_SIZE]; 142 /* Untouched saved command line (eg. for /proc) */ 143 char *saved_command_line __ro_after_init; 144 unsigned int saved_command_line_len __ro_after_init; 145 /* Command line for parameter parsing */ 146 static char *static_command_line; 147 /* Untouched extra command line */ 148 static char *extra_command_line; 149 /* Extra init arguments */ 150 static char *extra_init_args; 151 152 #ifdef CONFIG_BOOT_CONFIG 153 /* Is bootconfig on command line? */ 154 static bool bootconfig_found; 155 static size_t initargs_offs; 156 #else 157 # define bootconfig_found false 158 # define initargs_offs 0 159 #endif 160 161 static char *execute_command; 162 static char *ramdisk_execute_command = "/init"; 163 164 /* 165 * Used to generate warnings if static_key manipulation functions are used 166 * before jump_label_init is called. 167 */ 168 bool static_key_initialized __read_mostly; 169 EXPORT_SYMBOL_GPL(static_key_initialized); 170 171 /* 172 * If set, this is an indication to the drivers that reset the underlying 173 * device before going ahead with the initialization otherwise driver might 174 * rely on the BIOS and skip the reset operation. 175 * 176 * This is useful if kernel is booting in an unreliable environment. 177 * For ex. kdump situation where previous kernel has crashed, BIOS has been 178 * skipped and devices will be in unknown state. 179 */ 180 unsigned int reset_devices; 181 EXPORT_SYMBOL(reset_devices); 182 183 static int __init set_reset_devices(char *str) 184 { 185 reset_devices = 1; 186 return 1; 187 } 188 189 __setup("reset_devices", set_reset_devices); 190 191 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; 192 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; 193 static const char *panic_later, *panic_param; 194 195 static bool __init obsolete_checksetup(char *line) 196 { 197 const struct obs_kernel_param *p; 198 bool had_early_param = false; 199 200 p = __setup_start; 201 do { 202 int n = strlen(p->str); 203 if (parameqn(line, p->str, n)) { 204 if (p->early) { 205 /* Already done in parse_early_param? 206 * (Needs exact match on param part). 207 * Keep iterating, as we can have early 208 * params and __setups of same names 8( */ 209 if (line[n] == '\0' || line[n] == '=') 210 had_early_param = true; 211 } else if (!p->setup_func) { 212 pr_warn("Parameter %s is obsolete, ignored\n", 213 p->str); 214 return true; 215 } else if (p->setup_func(line + n)) 216 return true; 217 } 218 p++; 219 } while (p < __setup_end); 220 221 return had_early_param; 222 } 223 224 /* 225 * This should be approx 2 Bo*oMips to start (note initial shift), and will 226 * still work even if initially too large, it will just take slightly longer 227 */ 228 unsigned long loops_per_jiffy = (1<<12); 229 EXPORT_SYMBOL(loops_per_jiffy); 230 231 static int __init debug_kernel(char *str) 232 { 233 console_loglevel = CONSOLE_LOGLEVEL_DEBUG; 234 return 0; 235 } 236 237 static int __init quiet_kernel(char *str) 238 { 239 console_loglevel = CONSOLE_LOGLEVEL_QUIET; 240 return 0; 241 } 242 243 early_param("debug", debug_kernel); 244 early_param("quiet", quiet_kernel); 245 246 static int __init loglevel(char *str) 247 { 248 int newlevel; 249 250 /* 251 * Only update loglevel value when a correct setting was passed, 252 * to prevent blind crashes (when loglevel being set to 0) that 253 * are quite hard to debug 254 */ 255 if (get_option(&str, &newlevel)) { 256 console_loglevel = newlevel; 257 return 0; 258 } 259 260 return -EINVAL; 261 } 262 263 early_param("loglevel", loglevel); 264 265 #ifdef CONFIG_BLK_DEV_INITRD 266 static void * __init get_boot_config_from_initrd(size_t *_size) 267 { 268 u32 size, csum; 269 char *data; 270 u32 *hdr; 271 int i; 272 273 if (!initrd_end) 274 return NULL; 275 276 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN; 277 /* 278 * Since Grub may align the size of initrd to 4, we must 279 * check the preceding 3 bytes as well. 280 */ 281 for (i = 0; i < 4; i++) { 282 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN)) 283 goto found; 284 data--; 285 } 286 return NULL; 287 288 found: 289 hdr = (u32 *)(data - 8); 290 size = le32_to_cpu(hdr[0]); 291 csum = le32_to_cpu(hdr[1]); 292 293 data = ((void *)hdr) - size; 294 if ((unsigned long)data < initrd_start) { 295 pr_err("bootconfig size %d is greater than initrd size %ld\n", 296 size, initrd_end - initrd_start); 297 return NULL; 298 } 299 300 if (xbc_calc_checksum(data, size) != csum) { 301 pr_err("bootconfig checksum failed\n"); 302 return NULL; 303 } 304 305 /* Remove bootconfig from initramfs/initrd */ 306 initrd_end = (unsigned long)data; 307 if (_size) 308 *_size = size; 309 310 return data; 311 } 312 #else 313 static void * __init get_boot_config_from_initrd(size_t *_size) 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, *q; 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 q = strpbrk(val, " \t\r\n") ? "\"" : ""; 349 ret = snprintf(buf, rest(buf, end), "%s=%s%s%s ", 350 xbc_namebuf, q, val, q); 351 if (ret < 0) 352 return ret; 353 buf += ret; 354 } 355 } 356 357 return buf - (end - size); 358 } 359 #undef rest 360 361 /* Make an extra command line under given key word */ 362 static char * __init xbc_make_cmdline(const char *key) 363 { 364 struct xbc_node *root; 365 char *new_cmdline; 366 int ret, len = 0; 367 368 root = xbc_find_node(key); 369 if (!root) 370 return NULL; 371 372 /* Count required buffer size */ 373 len = xbc_snprint_cmdline(NULL, 0, root); 374 if (len <= 0) 375 return NULL; 376 377 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES); 378 if (!new_cmdline) { 379 pr_err("Failed to allocate memory for extra kernel cmdline.\n"); 380 return NULL; 381 } 382 383 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root); 384 if (ret < 0 || ret > len) { 385 pr_err("Failed to print extra kernel cmdline.\n"); 386 memblock_free(new_cmdline, len + 1); 387 return NULL; 388 } 389 390 return new_cmdline; 391 } 392 393 static int __init bootconfig_params(char *param, char *val, 394 const char *unused, void *arg) 395 { 396 if (strcmp(param, "bootconfig") == 0) { 397 bootconfig_found = true; 398 } 399 return 0; 400 } 401 402 static int __init warn_bootconfig(char *str) 403 { 404 /* The 'bootconfig' has been handled by bootconfig_params(). */ 405 return 0; 406 } 407 408 static void __init setup_boot_config(void) 409 { 410 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 411 const char *msg, *data; 412 int pos, ret; 413 size_t size; 414 char *err; 415 416 /* Cut out the bootconfig data even if we have no bootconfig option */ 417 data = get_boot_config_from_initrd(&size); 418 /* If there is no bootconfig in initrd, try embedded one. */ 419 if (!data) 420 data = xbc_get_embedded_bootconfig(&size); 421 422 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 423 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL, 424 bootconfig_params); 425 426 if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE))) 427 return; 428 429 /* parse_args() stops at the next param of '--' and returns an address */ 430 if (err) 431 initargs_offs = err - tmp_cmdline; 432 433 if (!data) { 434 /* If user intended to use bootconfig, show an error level message */ 435 if (bootconfig_found) 436 pr_err("'bootconfig' found on command line, but no bootconfig found\n"); 437 else 438 pr_info("No bootconfig data provided, so skipping bootconfig"); 439 return; 440 } 441 442 if (size >= XBC_DATA_MAX) { 443 pr_err("bootconfig size %ld greater than max size %d\n", 444 (long)size, XBC_DATA_MAX); 445 return; 446 } 447 448 ret = xbc_init(data, size, &msg, &pos); 449 if (ret < 0) { 450 if (pos < 0) 451 pr_err("Failed to init bootconfig: %s.\n", msg); 452 else 453 pr_err("Failed to parse bootconfig: %s at %d.\n", 454 msg, pos); 455 } else { 456 xbc_get_info(&ret, NULL); 457 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret); 458 /* keys starting with "kernel." are passed via cmdline */ 459 extra_command_line = xbc_make_cmdline("kernel"); 460 /* Also, "init." keys are init arguments */ 461 extra_init_args = xbc_make_cmdline("init"); 462 } 463 return; 464 } 465 466 static void __init exit_boot_config(void) 467 { 468 xbc_exit(); 469 } 470 471 #else /* !CONFIG_BOOT_CONFIG */ 472 473 static void __init setup_boot_config(void) 474 { 475 /* Remove bootconfig data from initrd */ 476 get_boot_config_from_initrd(NULL); 477 } 478 479 static int __init warn_bootconfig(char *str) 480 { 481 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n"); 482 return 0; 483 } 484 485 #define exit_boot_config() do {} while (0) 486 487 #endif /* CONFIG_BOOT_CONFIG */ 488 489 early_param("bootconfig", warn_bootconfig); 490 491 bool __init cmdline_has_extra_options(void) 492 { 493 return extra_command_line || extra_init_args; 494 } 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 /* Handle params aliased to sysctls */ 543 if (sysctl_is_alias(param)) 544 return 0; 545 546 repair_env_string(param, val); 547 548 /* Handle obsolete-style parameters */ 549 if (obsolete_checksetup(param)) 550 return 0; 551 552 /* Unused module parameter. */ 553 if (strnchr(param, len, '.')) 554 return 0; 555 556 if (panic_later) 557 return 0; 558 559 if (val) { 560 /* Environment option */ 561 unsigned int i; 562 for (i = 0; envp_init[i]; i++) { 563 if (i == MAX_INIT_ENVS) { 564 panic_later = "env"; 565 panic_param = param; 566 } 567 if (!strncmp(param, envp_init[i], len+1)) 568 break; 569 } 570 envp_init[i] = param; 571 } else { 572 /* Command line option */ 573 unsigned int i; 574 for (i = 0; argv_init[i]; i++) { 575 if (i == MAX_INIT_ARGS) { 576 panic_later = "init"; 577 panic_param = param; 578 } 579 } 580 argv_init[i] = param; 581 } 582 return 0; 583 } 584 585 static int __init init_setup(char *str) 586 { 587 unsigned int i; 588 589 execute_command = str; 590 /* 591 * In case LILO is going to boot us with default command line, 592 * it prepends "auto" before the whole cmdline which makes 593 * the shell think it should execute a script with such name. 594 * So we ignore all arguments entered _before_ init=... [MJ] 595 */ 596 for (i = 1; i < MAX_INIT_ARGS; i++) 597 argv_init[i] = NULL; 598 return 1; 599 } 600 __setup("init=", init_setup); 601 602 static int __init rdinit_setup(char *str) 603 { 604 unsigned int i; 605 606 ramdisk_execute_command = str; 607 /* See "auto" comment in init_setup */ 608 for (i = 1; i < MAX_INIT_ARGS; i++) 609 argv_init[i] = NULL; 610 return 1; 611 } 612 __setup("rdinit=", rdinit_setup); 613 614 #ifndef CONFIG_SMP 615 static inline void setup_nr_cpu_ids(void) { } 616 static inline void smp_prepare_cpus(unsigned int maxcpus) { } 617 #endif 618 619 /* 620 * We need to store the untouched command line for future reference. 621 * We also need to store the touched command line since the parameter 622 * parsing is performed in place, and we should allow a component to 623 * store reference of name/value for future reference. 624 */ 625 static void __init setup_command_line(char *command_line) 626 { 627 size_t len, xlen = 0, ilen = 0; 628 629 if (extra_command_line) 630 xlen = strlen(extra_command_line); 631 if (extra_init_args) { 632 extra_init_args = strim(extra_init_args); /* remove trailing space */ 633 ilen = strlen(extra_init_args) + 4; /* for " -- " */ 634 } 635 636 len = xlen + strlen(boot_command_line) + ilen + 1; 637 638 saved_command_line = memblock_alloc(len, SMP_CACHE_BYTES); 639 if (!saved_command_line) 640 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 641 642 len = xlen + strlen(command_line) + 1; 643 644 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES); 645 if (!static_command_line) 646 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 647 648 if (xlen) { 649 /* 650 * We have to put extra_command_line before boot command 651 * lines because there could be dashes (separator of init 652 * command line) in the command lines. 653 */ 654 strcpy(saved_command_line, extra_command_line); 655 strcpy(static_command_line, extra_command_line); 656 } 657 strcpy(saved_command_line + xlen, boot_command_line); 658 strcpy(static_command_line + xlen, command_line); 659 660 if (ilen) { 661 /* 662 * Append supplemental init boot args to saved_command_line 663 * so that user can check what command line options passed 664 * to init. 665 * The order should always be 666 * " -- "[bootconfig init-param][cmdline init-param] 667 */ 668 if (initargs_offs) { 669 len = xlen + initargs_offs; 670 strcpy(saved_command_line + len, extra_init_args); 671 len += ilen - 4; /* strlen(extra_init_args) */ 672 strcpy(saved_command_line + len, 673 boot_command_line + initargs_offs - 1); 674 } else { 675 len = strlen(saved_command_line); 676 strcpy(saved_command_line + len, " -- "); 677 len += 4; 678 strcpy(saved_command_line + len, extra_init_args); 679 } 680 } 681 682 saved_command_line_len = strlen(saved_command_line); 683 } 684 685 /* 686 * We need to finalize in a non-__init function or else race conditions 687 * between the root thread and the init thread may cause start_kernel to 688 * be reaped by free_initmem before the root thread has proceeded to 689 * cpu_idle. 690 * 691 * gcc-3.4 accidentally inlines this function, so use noinline. 692 */ 693 694 static __initdata DECLARE_COMPLETION(kthreadd_done); 695 696 static noinline void __ref __noreturn rest_init(void) 697 { 698 struct task_struct *tsk; 699 int pid; 700 701 rcu_scheduler_starting(); 702 /* 703 * We need to spawn init first so that it obtains pid 1, however 704 * the init task will end up wanting to create kthreads, which, if 705 * we schedule it before we create kthreadd, will OOPS. 706 */ 707 pid = user_mode_thread(kernel_init, NULL, CLONE_FS); 708 /* 709 * Pin init on the boot CPU. Task migration is not properly working 710 * until sched_init_smp() has been run. It will set the allowed 711 * CPUs for init to the non isolated CPUs. 712 */ 713 rcu_read_lock(); 714 tsk = find_task_by_pid_ns(pid, &init_pid_ns); 715 tsk->flags |= PF_NO_SETAFFINITY; 716 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id())); 717 rcu_read_unlock(); 718 719 numa_default_policy(); 720 pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES); 721 rcu_read_lock(); 722 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); 723 rcu_read_unlock(); 724 725 /* 726 * Enable might_sleep() and smp_processor_id() checks. 727 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y 728 * kernel_thread() would trigger might_sleep() splats. With 729 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled 730 * already, but it's stuck on the kthreadd_done completion. 731 */ 732 system_state = SYSTEM_SCHEDULING; 733 734 complete(&kthreadd_done); 735 736 /* 737 * The boot idle thread must execute schedule() 738 * at least once to get things moving: 739 */ 740 schedule_preempt_disabled(); 741 /* Call into cpu_idle with preempt disabled */ 742 cpu_startup_entry(CPUHP_ONLINE); 743 } 744 745 /* Check for early params. */ 746 static int __init do_early_param(char *param, char *val, 747 const char *unused, void *arg) 748 { 749 const struct obs_kernel_param *p; 750 751 for (p = __setup_start; p < __setup_end; p++) { 752 if ((p->early && parameq(param, p->str)) || 753 (strcmp(param, "console") == 0 && 754 strcmp(p->str, "earlycon") == 0) 755 ) { 756 if (p->setup_func(val) != 0) 757 pr_warn("Malformed early option '%s'\n", param); 758 } 759 } 760 /* We accept everything at this stage. */ 761 return 0; 762 } 763 764 void __init parse_early_options(char *cmdline) 765 { 766 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL, 767 do_early_param); 768 } 769 770 /* Arch code calls this early on, or if not, just before other parsing. */ 771 void __init parse_early_param(void) 772 { 773 static int done __initdata; 774 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 775 776 if (done) 777 return; 778 779 /* All fall through to do_early_param. */ 780 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 781 parse_early_options(tmp_cmdline); 782 done = 1; 783 } 784 785 void __init __weak arch_post_acpi_subsys_init(void) { } 786 787 void __init __weak smp_setup_processor_id(void) 788 { 789 } 790 791 void __init __weak smp_prepare_boot_cpu(void) 792 { 793 } 794 795 # if THREAD_SIZE >= PAGE_SIZE 796 void __init __weak thread_stack_cache_init(void) 797 { 798 } 799 #endif 800 801 void __init __weak poking_init(void) { } 802 803 void __init __weak pgtable_cache_init(void) { } 804 805 void __init __weak trap_init(void) { } 806 807 bool initcall_debug; 808 core_param(initcall_debug, initcall_debug, bool, 0644); 809 810 #ifdef TRACEPOINTS_ENABLED 811 static void __init initcall_debug_enable(void); 812 #else 813 static inline void initcall_debug_enable(void) 814 { 815 } 816 #endif 817 818 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET 819 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT, 820 randomize_kstack_offset); 821 DEFINE_PER_CPU(u32, kstack_offset); 822 823 static int __init early_randomize_kstack_offset(char *buf) 824 { 825 int ret; 826 bool bool_result; 827 828 ret = kstrtobool(buf, &bool_result); 829 if (ret) 830 return ret; 831 832 if (bool_result) 833 static_branch_enable(&randomize_kstack_offset); 834 else 835 static_branch_disable(&randomize_kstack_offset); 836 return 0; 837 } 838 early_param("randomize_kstack_offset", early_randomize_kstack_offset); 839 #endif 840 841 static void __init print_unknown_bootoptions(void) 842 { 843 char *unknown_options; 844 char *end; 845 const char *const *p; 846 size_t len; 847 848 if (panic_later || (!argv_init[1] && !envp_init[2])) 849 return; 850 851 /* 852 * Determine how many options we have to print out, plus a space 853 * before each 854 */ 855 len = 1; /* null terminator */ 856 for (p = &argv_init[1]; *p; p++) { 857 len++; 858 len += strlen(*p); 859 } 860 for (p = &envp_init[2]; *p; p++) { 861 len++; 862 len += strlen(*p); 863 } 864 865 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES); 866 if (!unknown_options) { 867 pr_err("%s: Failed to allocate %zu bytes\n", 868 __func__, len); 869 return; 870 } 871 end = unknown_options; 872 873 for (p = &argv_init[1]; *p; p++) 874 end += sprintf(end, " %s", *p); 875 for (p = &envp_init[2]; *p; p++) 876 end += sprintf(end, " %s", *p); 877 878 /* Start at unknown_options[1] to skip the initial space */ 879 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n", 880 &unknown_options[1]); 881 memblock_free(unknown_options, len); 882 } 883 884 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector 885 void start_kernel(void) 886 { 887 char *command_line; 888 char *after_dashes; 889 890 set_task_stack_end_magic(&init_task); 891 smp_setup_processor_id(); 892 debug_objects_early_init(); 893 init_vmlinux_build_id(); 894 895 cgroup_init_early(); 896 897 local_irq_disable(); 898 early_boot_irqs_disabled = true; 899 900 /* 901 * Interrupts are still disabled. Do necessary setups, then 902 * enable them. 903 */ 904 boot_cpu_init(); 905 page_address_init(); 906 pr_notice("%s", linux_banner); 907 early_security_init(); 908 setup_arch(&command_line); 909 setup_boot_config(); 910 setup_command_line(command_line); 911 setup_nr_cpu_ids(); 912 setup_per_cpu_areas(); 913 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ 914 boot_cpu_hotplug_init(); 915 916 pr_notice("Kernel command line: %s\n", saved_command_line); 917 /* parameters may set static keys */ 918 jump_label_init(); 919 parse_early_param(); 920 after_dashes = parse_args("Booting kernel", 921 static_command_line, __start___param, 922 __stop___param - __start___param, 923 -1, -1, NULL, &unknown_bootoption); 924 print_unknown_bootoptions(); 925 if (!IS_ERR_OR_NULL(after_dashes)) 926 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1, 927 NULL, set_init_arg); 928 if (extra_init_args) 929 parse_args("Setting extra init args", extra_init_args, 930 NULL, 0, -1, -1, NULL, set_init_arg); 931 932 /* Architectural and non-timekeeping rng init, before allocator init */ 933 random_init_early(command_line); 934 935 /* 936 * These use large bootmem allocations and must precede 937 * initalization of page allocator 938 */ 939 setup_log_buf(0); 940 vfs_caches_init_early(); 941 sort_main_extable(); 942 trap_init(); 943 mm_core_init(); 944 poking_init(); 945 ftrace_init(); 946 947 /* trace_printk can be enabled here */ 948 early_trace_init(); 949 950 /* 951 * Set up the scheduler prior starting any interrupts (such as the 952 * timer interrupt). Full topology setup happens at smp_init() 953 * time - but meanwhile we still have a functioning scheduler. 954 */ 955 sched_init(); 956 957 if (WARN(!irqs_disabled(), 958 "Interrupts were enabled *very* early, fixing it\n")) 959 local_irq_disable(); 960 radix_tree_init(); 961 maple_tree_init(); 962 963 /* 964 * Set up housekeeping before setting up workqueues to allow the unbound 965 * workqueue to take non-housekeeping into account. 966 */ 967 housekeeping_init(); 968 969 /* 970 * Allow workqueue creation and work item queueing/cancelling 971 * early. Work item execution depends on kthreads and starts after 972 * workqueue_init(). 973 */ 974 workqueue_init_early(); 975 976 rcu_init(); 977 978 /* Trace events are available after this */ 979 trace_init(); 980 981 if (initcall_debug) 982 initcall_debug_enable(); 983 984 context_tracking_init(); 985 /* init some links before init_ISA_irqs() */ 986 early_irq_init(); 987 init_IRQ(); 988 tick_init(); 989 rcu_init_nohz(); 990 init_timers(); 991 srcu_init(); 992 hrtimers_init(); 993 softirq_init(); 994 timekeeping_init(); 995 time_init(); 996 997 /* This must be after timekeeping is initialized */ 998 random_init(); 999 1000 /* These make use of the fully initialized rng */ 1001 kfence_init(); 1002 boot_init_stack_canary(); 1003 1004 perf_event_init(); 1005 profile_init(); 1006 call_function_init(); 1007 WARN(!irqs_disabled(), "Interrupts were enabled early\n"); 1008 1009 early_boot_irqs_disabled = false; 1010 local_irq_enable(); 1011 1012 kmem_cache_init_late(); 1013 1014 /* 1015 * HACK ALERT! This is early. We're enabling the console before 1016 * we've done PCI setups etc, and console_init() must be aware of 1017 * this. But we do want output early, in case something goes wrong. 1018 */ 1019 console_init(); 1020 if (panic_later) 1021 panic("Too many boot %s vars at `%s'", panic_later, 1022 panic_param); 1023 1024 lockdep_init(); 1025 1026 /* 1027 * Need to run this when irqs are enabled, because it wants 1028 * to self-test [hard/soft]-irqs on/off lock inversion bugs 1029 * too: 1030 */ 1031 locking_selftest(); 1032 1033 #ifdef CONFIG_BLK_DEV_INITRD 1034 if (initrd_start && !initrd_below_start_ok && 1035 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { 1036 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", 1037 page_to_pfn(virt_to_page((void *)initrd_start)), 1038 min_low_pfn); 1039 initrd_start = 0; 1040 } 1041 #endif 1042 setup_per_cpu_pageset(); 1043 numa_policy_init(); 1044 acpi_early_init(); 1045 if (late_time_init) 1046 late_time_init(); 1047 sched_clock_init(); 1048 calibrate_delay(); 1049 1050 arch_cpu_finalize_init(); 1051 1052 pid_idr_init(); 1053 anon_vma_init(); 1054 #ifdef CONFIG_X86 1055 if (efi_enabled(EFI_RUNTIME_SERVICES)) 1056 efi_enter_virtual_mode(); 1057 #endif 1058 thread_stack_cache_init(); 1059 cred_init(); 1060 fork_init(); 1061 proc_caches_init(); 1062 uts_ns_init(); 1063 key_init(); 1064 security_init(); 1065 dbg_late_init(); 1066 net_ns_init(); 1067 vfs_caches_init(); 1068 pagecache_init(); 1069 signals_init(); 1070 seq_file_init(); 1071 proc_root_init(); 1072 nsfs_init(); 1073 pidfs_init(); 1074 cpuset_init(); 1075 cgroup_init(); 1076 taskstats_init_early(); 1077 delayacct_init(); 1078 1079 acpi_subsystem_init(); 1080 arch_post_acpi_subsys_init(); 1081 kcsan_init(); 1082 1083 /* Do the rest non-__init'ed, we're now alive */ 1084 rest_init(); 1085 1086 /* 1087 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10 1088 * and older. 1089 */ 1090 #if !__has_attribute(__no_stack_protector__) 1091 prevent_tail_call_optimization(); 1092 #endif 1093 } 1094 1095 /* Call all constructor functions linked into the kernel. */ 1096 static void __init do_ctors(void) 1097 { 1098 /* 1099 * For UML, the constructors have already been called by the 1100 * normal setup code as it's just a normal ELF binary, so we 1101 * cannot do it again - but we do need CONFIG_CONSTRUCTORS 1102 * even on UML for modules. 1103 */ 1104 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML) 1105 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; 1106 1107 for (; fn < (ctor_fn_t *) __ctors_end; fn++) 1108 (*fn)(); 1109 #endif 1110 } 1111 1112 #ifdef CONFIG_KALLSYMS 1113 struct blacklist_entry { 1114 struct list_head next; 1115 char *buf; 1116 }; 1117 1118 static __initdata_or_module LIST_HEAD(blacklisted_initcalls); 1119 1120 static int __init initcall_blacklist(char *str) 1121 { 1122 char *str_entry; 1123 struct blacklist_entry *entry; 1124 1125 /* str argument is a comma-separated list of functions */ 1126 do { 1127 str_entry = strsep(&str, ","); 1128 if (str_entry) { 1129 pr_debug("blacklisting initcall %s\n", str_entry); 1130 entry = memblock_alloc(sizeof(*entry), 1131 SMP_CACHE_BYTES); 1132 if (!entry) 1133 panic("%s: Failed to allocate %zu bytes\n", 1134 __func__, sizeof(*entry)); 1135 entry->buf = memblock_alloc(strlen(str_entry) + 1, 1136 SMP_CACHE_BYTES); 1137 if (!entry->buf) 1138 panic("%s: Failed to allocate %zu bytes\n", 1139 __func__, strlen(str_entry) + 1); 1140 strcpy(entry->buf, str_entry); 1141 list_add(&entry->next, &blacklisted_initcalls); 1142 } 1143 } while (str_entry); 1144 1145 return 1; 1146 } 1147 1148 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1149 { 1150 struct blacklist_entry *entry; 1151 char fn_name[KSYM_SYMBOL_LEN]; 1152 unsigned long addr; 1153 1154 if (list_empty(&blacklisted_initcalls)) 1155 return false; 1156 1157 addr = (unsigned long) dereference_function_descriptor(fn); 1158 sprint_symbol_no_offset(fn_name, addr); 1159 1160 /* 1161 * fn will be "function_name [module_name]" where [module_name] is not 1162 * displayed for built-in init functions. Strip off the [module_name]. 1163 */ 1164 strreplace(fn_name, ' ', '\0'); 1165 1166 list_for_each_entry(entry, &blacklisted_initcalls, next) { 1167 if (!strcmp(fn_name, entry->buf)) { 1168 pr_debug("initcall %s blacklisted\n", fn_name); 1169 return true; 1170 } 1171 } 1172 1173 return false; 1174 } 1175 #else 1176 static int __init initcall_blacklist(char *str) 1177 { 1178 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); 1179 return 0; 1180 } 1181 1182 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1183 { 1184 return false; 1185 } 1186 #endif 1187 __setup("initcall_blacklist=", initcall_blacklist); 1188 1189 static __init_or_module void 1190 trace_initcall_start_cb(void *data, initcall_t fn) 1191 { 1192 ktime_t *calltime = data; 1193 1194 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current)); 1195 *calltime = ktime_get(); 1196 } 1197 1198 static __init_or_module void 1199 trace_initcall_finish_cb(void *data, initcall_t fn, int ret) 1200 { 1201 ktime_t rettime, *calltime = data; 1202 1203 rettime = ktime_get(); 1204 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n", 1205 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime)); 1206 } 1207 1208 static ktime_t initcall_calltime; 1209 1210 #ifdef TRACEPOINTS_ENABLED 1211 static void __init initcall_debug_enable(void) 1212 { 1213 int ret; 1214 1215 ret = register_trace_initcall_start(trace_initcall_start_cb, 1216 &initcall_calltime); 1217 ret |= register_trace_initcall_finish(trace_initcall_finish_cb, 1218 &initcall_calltime); 1219 WARN(ret, "Failed to register initcall tracepoints\n"); 1220 } 1221 # define do_trace_initcall_start trace_initcall_start 1222 # define do_trace_initcall_finish trace_initcall_finish 1223 #else 1224 static inline void do_trace_initcall_start(initcall_t fn) 1225 { 1226 if (!initcall_debug) 1227 return; 1228 trace_initcall_start_cb(&initcall_calltime, fn); 1229 } 1230 static inline void do_trace_initcall_finish(initcall_t fn, int ret) 1231 { 1232 if (!initcall_debug) 1233 return; 1234 trace_initcall_finish_cb(&initcall_calltime, fn, ret); 1235 } 1236 #endif /* !TRACEPOINTS_ENABLED */ 1237 1238 int __init_or_module do_one_initcall(initcall_t fn) 1239 { 1240 int count = preempt_count(); 1241 char msgbuf[64]; 1242 int ret; 1243 1244 if (initcall_blacklisted(fn)) 1245 return -EPERM; 1246 1247 do_trace_initcall_start(fn); 1248 ret = fn(); 1249 do_trace_initcall_finish(fn, ret); 1250 1251 msgbuf[0] = 0; 1252 1253 if (preempt_count() != count) { 1254 sprintf(msgbuf, "preemption imbalance "); 1255 preempt_count_set(count); 1256 } 1257 if (irqs_disabled()) { 1258 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); 1259 local_irq_enable(); 1260 } 1261 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf); 1262 1263 add_latent_entropy(); 1264 return ret; 1265 } 1266 1267 1268 static initcall_entry_t *initcall_levels[] __initdata = { 1269 __initcall0_start, 1270 __initcall1_start, 1271 __initcall2_start, 1272 __initcall3_start, 1273 __initcall4_start, 1274 __initcall5_start, 1275 __initcall6_start, 1276 __initcall7_start, 1277 __initcall_end, 1278 }; 1279 1280 /* Keep these in sync with initcalls in include/linux/init.h */ 1281 static const char *initcall_level_names[] __initdata = { 1282 "pure", 1283 "core", 1284 "postcore", 1285 "arch", 1286 "subsys", 1287 "fs", 1288 "device", 1289 "late", 1290 }; 1291 1292 static int __init ignore_unknown_bootoption(char *param, char *val, 1293 const char *unused, void *arg) 1294 { 1295 return 0; 1296 } 1297 1298 static void __init do_initcall_level(int level, char *command_line) 1299 { 1300 initcall_entry_t *fn; 1301 1302 parse_args(initcall_level_names[level], 1303 command_line, __start___param, 1304 __stop___param - __start___param, 1305 level, level, 1306 NULL, ignore_unknown_bootoption); 1307 1308 trace_initcall_level(initcall_level_names[level]); 1309 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) 1310 do_one_initcall(initcall_from_entry(fn)); 1311 } 1312 1313 static void __init do_initcalls(void) 1314 { 1315 int level; 1316 size_t len = saved_command_line_len + 1; 1317 char *command_line; 1318 1319 command_line = kzalloc(len, GFP_KERNEL); 1320 if (!command_line) 1321 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 1322 1323 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) { 1324 /* Parser modifies command_line, restore it each time */ 1325 strcpy(command_line, saved_command_line); 1326 do_initcall_level(level, command_line); 1327 } 1328 1329 kfree(command_line); 1330 } 1331 1332 /* 1333 * Ok, the machine is now initialized. None of the devices 1334 * have been touched yet, but the CPU subsystem is up and 1335 * running, and memory and process management works. 1336 * 1337 * Now we can finally start doing some real work.. 1338 */ 1339 static void __init do_basic_setup(void) 1340 { 1341 cpuset_init_smp(); 1342 driver_init(); 1343 init_irq_proc(); 1344 do_ctors(); 1345 do_initcalls(); 1346 } 1347 1348 static void __init do_pre_smp_initcalls(void) 1349 { 1350 initcall_entry_t *fn; 1351 1352 trace_initcall_level("early"); 1353 for (fn = __initcall_start; fn < __initcall0_start; fn++) 1354 do_one_initcall(initcall_from_entry(fn)); 1355 } 1356 1357 static int run_init_process(const char *init_filename) 1358 { 1359 const char *const *p; 1360 1361 argv_init[0] = init_filename; 1362 pr_info("Run %s as init process\n", init_filename); 1363 pr_debug(" with arguments:\n"); 1364 for (p = argv_init; *p; p++) 1365 pr_debug(" %s\n", *p); 1366 pr_debug(" with environment:\n"); 1367 for (p = envp_init; *p; p++) 1368 pr_debug(" %s\n", *p); 1369 return kernel_execve(init_filename, argv_init, envp_init); 1370 } 1371 1372 static int try_to_run_init_process(const char *init_filename) 1373 { 1374 int ret; 1375 1376 ret = run_init_process(init_filename); 1377 1378 if (ret && ret != -ENOENT) { 1379 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", 1380 init_filename, ret); 1381 } 1382 1383 return ret; 1384 } 1385 1386 static noinline void __init kernel_init_freeable(void); 1387 1388 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX) 1389 bool rodata_enabled __ro_after_init = true; 1390 1391 #ifndef arch_parse_debug_rodata 1392 static inline bool arch_parse_debug_rodata(char *str) { return false; } 1393 #endif 1394 1395 static int __init set_debug_rodata(char *str) 1396 { 1397 if (arch_parse_debug_rodata(str)) 1398 return 0; 1399 1400 if (str && !strcmp(str, "on")) 1401 rodata_enabled = true; 1402 else if (str && !strcmp(str, "off")) 1403 rodata_enabled = false; 1404 else 1405 pr_warn("Invalid option string for rodata: '%s'\n", str); 1406 return 0; 1407 } 1408 early_param("rodata", set_debug_rodata); 1409 #endif 1410 1411 static void mark_readonly(void) 1412 { 1413 if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX) && rodata_enabled) { 1414 /* 1415 * load_module() results in W+X mappings, which are cleaned 1416 * up with init_free_wq. Let's make sure that queued work is 1417 * flushed so that we don't hit false positives looking for 1418 * insecure pages which are W+X. 1419 */ 1420 flush_module_init_free_work(); 1421 jump_label_init_ro(); 1422 mark_rodata_ro(); 1423 debug_checkwx(); 1424 rodata_test(); 1425 } else if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) { 1426 pr_info("Kernel memory protection disabled.\n"); 1427 } else if (IS_ENABLED(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)) { 1428 pr_warn("Kernel memory protection not selected by kernel config.\n"); 1429 } else { 1430 pr_warn("This architecture does not have kernel memory protection.\n"); 1431 } 1432 } 1433 1434 void __weak free_initmem(void) 1435 { 1436 free_initmem_default(POISON_FREE_INITMEM); 1437 } 1438 1439 static int __ref kernel_init(void *unused) 1440 { 1441 int ret; 1442 1443 /* 1444 * Wait until kthreadd is all set-up. 1445 */ 1446 wait_for_completion(&kthreadd_done); 1447 1448 kernel_init_freeable(); 1449 /* need to finish all async __init code before freeing the memory */ 1450 async_synchronize_full(); 1451 1452 system_state = SYSTEM_FREEING_INITMEM; 1453 kprobe_free_init_mem(); 1454 ftrace_free_init_mem(); 1455 kgdb_free_init_mem(); 1456 exit_boot_config(); 1457 free_initmem(); 1458 mark_readonly(); 1459 1460 /* 1461 * Kernel mappings are now finalized - update the userspace page-table 1462 * to finalize PTI. 1463 */ 1464 pti_finalize(); 1465 1466 system_state = SYSTEM_RUNNING; 1467 numa_default_policy(); 1468 1469 rcu_end_inkernel_boot(); 1470 1471 do_sysctl_args(); 1472 1473 if (ramdisk_execute_command) { 1474 ret = run_init_process(ramdisk_execute_command); 1475 if (!ret) 1476 return 0; 1477 pr_err("Failed to execute %s (error %d)\n", 1478 ramdisk_execute_command, ret); 1479 } 1480 1481 /* 1482 * We try each of these until one succeeds. 1483 * 1484 * The Bourne shell can be used instead of init if we are 1485 * trying to recover a really broken machine. 1486 */ 1487 if (execute_command) { 1488 ret = run_init_process(execute_command); 1489 if (!ret) 1490 return 0; 1491 panic("Requested init %s failed (error %d).", 1492 execute_command, ret); 1493 } 1494 1495 if (CONFIG_DEFAULT_INIT[0] != '\0') { 1496 ret = run_init_process(CONFIG_DEFAULT_INIT); 1497 if (ret) 1498 pr_err("Default init %s failed (error %d)\n", 1499 CONFIG_DEFAULT_INIT, ret); 1500 else 1501 return 0; 1502 } 1503 1504 if (!try_to_run_init_process("/sbin/init") || 1505 !try_to_run_init_process("/etc/init") || 1506 !try_to_run_init_process("/bin/init") || 1507 !try_to_run_init_process("/bin/sh")) 1508 return 0; 1509 1510 panic("No working init found. Try passing init= option to kernel. " 1511 "See Linux Documentation/admin-guide/init.rst for guidance."); 1512 } 1513 1514 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */ 1515 void __init console_on_rootfs(void) 1516 { 1517 struct file *file = filp_open("/dev/console", O_RDWR, 0); 1518 1519 if (IS_ERR(file)) { 1520 pr_err("Warning: unable to open an initial console.\n"); 1521 return; 1522 } 1523 init_dup(file); 1524 init_dup(file); 1525 init_dup(file); 1526 fput(file); 1527 } 1528 1529 static noinline void __init kernel_init_freeable(void) 1530 { 1531 /* Now the scheduler is fully set up and can do blocking allocations */ 1532 gfp_allowed_mask = __GFP_BITS_MASK; 1533 1534 /* 1535 * init can allocate pages on any node 1536 */ 1537 set_mems_allowed(node_states[N_MEMORY]); 1538 1539 cad_pid = get_pid(task_pid(current)); 1540 1541 smp_prepare_cpus(setup_max_cpus); 1542 1543 workqueue_init(); 1544 1545 init_mm_internals(); 1546 1547 rcu_init_tasks_generic(); 1548 do_pre_smp_initcalls(); 1549 lockup_detector_init(); 1550 1551 smp_init(); 1552 sched_init_smp(); 1553 1554 workqueue_init_topology(); 1555 async_init(); 1556 padata_init(); 1557 page_alloc_init_late(); 1558 1559 do_basic_setup(); 1560 1561 kunit_run_all_tests(); 1562 1563 wait_for_initramfs(); 1564 console_on_rootfs(); 1565 1566 /* 1567 * check if there is an early userspace init. If yes, let it do all 1568 * the work 1569 */ 1570 if (init_eaccess(ramdisk_execute_command) != 0) { 1571 ramdisk_execute_command = NULL; 1572 prepare_namespace(); 1573 } 1574 1575 /* 1576 * Ok, we have completed the initial bootup, and 1577 * we're essentially up and running. Get rid of the 1578 * initmem segments and start the user-mode stuff.. 1579 * 1580 * rootfs is available now, try loading the public keys 1581 * and default modules 1582 */ 1583 1584 integrity_load_keys(); 1585 } 1586