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/list.h> 92 #include <linux/integrity.h> 93 #include <linux/proc_ns.h> 94 #include <linux/io.h> 95 #include <linux/cache.h> 96 #include <linux/rodata_test.h> 97 #include <linux/jump_label.h> 98 #include <linux/kcsan.h> 99 #include <linux/init_syscalls.h> 100 #include <linux/stackdepot.h> 101 #include <linux/randomize_kstack.h> 102 #include <net/net_namespace.h> 103 104 #include <asm/io.h> 105 #include <asm/setup.h> 106 #include <asm/sections.h> 107 #include <asm/cacheflush.h> 108 109 #define CREATE_TRACE_POINTS 110 #include <trace/events/initcall.h> 111 112 #include <kunit/test.h> 113 114 static int kernel_init(void *); 115 116 extern void init_IRQ(void); 117 extern void radix_tree_init(void); 118 extern void maple_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 __ro_after_init; 146 unsigned int saved_command_line_len __ro_after_init; 147 /* Command line for parameter parsing */ 148 static char *static_command_line; 149 /* Untouched extra command line */ 150 static char *extra_command_line; 151 /* Extra init arguments */ 152 static char *extra_init_args; 153 154 #ifdef CONFIG_BOOT_CONFIG 155 /* Is bootconfig on command line? */ 156 static bool bootconfig_found; 157 static size_t initargs_offs; 158 #else 159 # define bootconfig_found false 160 # define initargs_offs 0 161 #endif 162 163 static char *execute_command; 164 static char *ramdisk_execute_command = "/init"; 165 166 /* 167 * Used to generate warnings if static_key manipulation functions are used 168 * before jump_label_init is called. 169 */ 170 bool static_key_initialized __read_mostly; 171 EXPORT_SYMBOL_GPL(static_key_initialized); 172 173 /* 174 * If set, this is an indication to the drivers that reset the underlying 175 * device before going ahead with the initialization otherwise driver might 176 * rely on the BIOS and skip the reset operation. 177 * 178 * This is useful if kernel is booting in an unreliable environment. 179 * For ex. kdump situation where previous kernel has crashed, BIOS has been 180 * skipped and devices will be in unknown state. 181 */ 182 unsigned int reset_devices; 183 EXPORT_SYMBOL(reset_devices); 184 185 static int __init set_reset_devices(char *str) 186 { 187 reset_devices = 1; 188 return 1; 189 } 190 191 __setup("reset_devices", set_reset_devices); 192 193 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; 194 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; 195 static const char *panic_later, *panic_param; 196 197 extern const struct obs_kernel_param __setup_start[], __setup_end[]; 198 199 static bool __init obsolete_checksetup(char *line) 200 { 201 const struct obs_kernel_param *p; 202 bool had_early_param = false; 203 204 p = __setup_start; 205 do { 206 int n = strlen(p->str); 207 if (parameqn(line, p->str, n)) { 208 if (p->early) { 209 /* Already done in parse_early_param? 210 * (Needs exact match on param part). 211 * Keep iterating, as we can have early 212 * params and __setups of same names 8( */ 213 if (line[n] == '\0' || line[n] == '=') 214 had_early_param = true; 215 } else if (!p->setup_func) { 216 pr_warn("Parameter %s is obsolete, ignored\n", 217 p->str); 218 return true; 219 } else if (p->setup_func(line + n)) 220 return true; 221 } 222 p++; 223 } while (p < __setup_end); 224 225 return had_early_param; 226 } 227 228 /* 229 * This should be approx 2 Bo*oMips to start (note initial shift), and will 230 * still work even if initially too large, it will just take slightly longer 231 */ 232 unsigned long loops_per_jiffy = (1<<12); 233 EXPORT_SYMBOL(loops_per_jiffy); 234 235 static int __init debug_kernel(char *str) 236 { 237 console_loglevel = CONSOLE_LOGLEVEL_DEBUG; 238 return 0; 239 } 240 241 static int __init quiet_kernel(char *str) 242 { 243 console_loglevel = CONSOLE_LOGLEVEL_QUIET; 244 return 0; 245 } 246 247 early_param("debug", debug_kernel); 248 early_param("quiet", quiet_kernel); 249 250 static int __init loglevel(char *str) 251 { 252 int newlevel; 253 254 /* 255 * Only update loglevel value when a correct setting was passed, 256 * to prevent blind crashes (when loglevel being set to 0) that 257 * are quite hard to debug 258 */ 259 if (get_option(&str, &newlevel)) { 260 console_loglevel = newlevel; 261 return 0; 262 } 263 264 return -EINVAL; 265 } 266 267 early_param("loglevel", loglevel); 268 269 #ifdef CONFIG_BLK_DEV_INITRD 270 static void * __init get_boot_config_from_initrd(size_t *_size) 271 { 272 u32 size, csum; 273 char *data; 274 u32 *hdr; 275 int i; 276 277 if (!initrd_end) 278 return NULL; 279 280 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN; 281 /* 282 * Since Grub may align the size of initrd to 4, we must 283 * check the preceding 3 bytes as well. 284 */ 285 for (i = 0; i < 4; i++) { 286 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN)) 287 goto found; 288 data--; 289 } 290 return NULL; 291 292 found: 293 hdr = (u32 *)(data - 8); 294 size = le32_to_cpu(hdr[0]); 295 csum = le32_to_cpu(hdr[1]); 296 297 data = ((void *)hdr) - size; 298 if ((unsigned long)data < initrd_start) { 299 pr_err("bootconfig size %d is greater than initrd size %ld\n", 300 size, initrd_end - initrd_start); 301 return NULL; 302 } 303 304 if (xbc_calc_checksum(data, size) != csum) { 305 pr_err("bootconfig checksum failed\n"); 306 return NULL; 307 } 308 309 /* Remove bootconfig from initramfs/initrd */ 310 initrd_end = (unsigned long)data; 311 if (_size) 312 *_size = size; 313 314 return data; 315 } 316 #else 317 static void * __init get_boot_config_from_initrd(size_t *_size) 318 { 319 return NULL; 320 } 321 #endif 322 323 #ifdef CONFIG_BOOT_CONFIG 324 325 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata; 326 327 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0) 328 329 static int __init xbc_snprint_cmdline(char *buf, size_t size, 330 struct xbc_node *root) 331 { 332 struct xbc_node *knode, *vnode; 333 char *end = buf + size; 334 const char *val; 335 int ret; 336 337 xbc_node_for_each_key_value(root, knode, val) { 338 ret = xbc_node_compose_key_after(root, knode, 339 xbc_namebuf, XBC_KEYLEN_MAX); 340 if (ret < 0) 341 return ret; 342 343 vnode = xbc_node_get_child(knode); 344 if (!vnode) { 345 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf); 346 if (ret < 0) 347 return ret; 348 buf += ret; 349 continue; 350 } 351 xbc_array_for_each_value(vnode, val) { 352 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ", 353 xbc_namebuf, val); 354 if (ret < 0) 355 return ret; 356 buf += ret; 357 } 358 } 359 360 return buf - (end - size); 361 } 362 #undef rest 363 364 /* Make an extra command line under given key word */ 365 static char * __init xbc_make_cmdline(const char *key) 366 { 367 struct xbc_node *root; 368 char *new_cmdline; 369 int ret, len = 0; 370 371 root = xbc_find_node(key); 372 if (!root) 373 return NULL; 374 375 /* Count required buffer size */ 376 len = xbc_snprint_cmdline(NULL, 0, root); 377 if (len <= 0) 378 return NULL; 379 380 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES); 381 if (!new_cmdline) { 382 pr_err("Failed to allocate memory for extra kernel cmdline.\n"); 383 return NULL; 384 } 385 386 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root); 387 if (ret < 0 || ret > len) { 388 pr_err("Failed to print extra kernel cmdline.\n"); 389 memblock_free(new_cmdline, len + 1); 390 return NULL; 391 } 392 393 return new_cmdline; 394 } 395 396 static int __init bootconfig_params(char *param, char *val, 397 const char *unused, void *arg) 398 { 399 if (strcmp(param, "bootconfig") == 0) { 400 bootconfig_found = true; 401 } 402 return 0; 403 } 404 405 static int __init warn_bootconfig(char *str) 406 { 407 /* The 'bootconfig' has been handled by bootconfig_params(). */ 408 return 0; 409 } 410 411 static void __init setup_boot_config(void) 412 { 413 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 414 const char *msg, *data; 415 int pos, ret; 416 size_t size; 417 char *err; 418 419 /* Cut out the bootconfig data even if we have no bootconfig option */ 420 data = get_boot_config_from_initrd(&size); 421 /* If there is no bootconfig in initrd, try embedded one. */ 422 if (!data) 423 data = xbc_get_embedded_bootconfig(&size); 424 425 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 426 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL, 427 bootconfig_params); 428 429 if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE))) 430 return; 431 432 /* parse_args() stops at the next param of '--' and returns an address */ 433 if (err) 434 initargs_offs = err - tmp_cmdline; 435 436 if (!data) { 437 /* If user intended to use bootconfig, show an error level message */ 438 if (bootconfig_found) 439 pr_err("'bootconfig' found on command line, but no bootconfig found\n"); 440 else 441 pr_info("No bootconfig data provided, so skipping bootconfig"); 442 return; 443 } 444 445 if (size >= XBC_DATA_MAX) { 446 pr_err("bootconfig size %ld greater than max size %d\n", 447 (long)size, XBC_DATA_MAX); 448 return; 449 } 450 451 ret = xbc_init(data, size, &msg, &pos); 452 if (ret < 0) { 453 if (pos < 0) 454 pr_err("Failed to init bootconfig: %s.\n", msg); 455 else 456 pr_err("Failed to parse bootconfig: %s at %d.\n", 457 msg, pos); 458 } else { 459 xbc_get_info(&ret, NULL); 460 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret); 461 /* keys starting with "kernel." are passed via cmdline */ 462 extra_command_line = xbc_make_cmdline("kernel"); 463 /* Also, "init." keys are init arguments */ 464 extra_init_args = xbc_make_cmdline("init"); 465 } 466 return; 467 } 468 469 static void __init exit_boot_config(void) 470 { 471 xbc_exit(); 472 } 473 474 #else /* !CONFIG_BOOT_CONFIG */ 475 476 static void __init setup_boot_config(void) 477 { 478 /* Remove bootconfig data from initrd */ 479 get_boot_config_from_initrd(NULL); 480 } 481 482 static int __init warn_bootconfig(char *str) 483 { 484 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n"); 485 return 0; 486 } 487 488 #define exit_boot_config() do {} while (0) 489 490 #endif /* CONFIG_BOOT_CONFIG */ 491 492 early_param("bootconfig", warn_bootconfig); 493 494 /* Change NUL term back to "=", to make "param" the whole string. */ 495 static void __init repair_env_string(char *param, char *val) 496 { 497 if (val) { 498 /* param=val or param="val"? */ 499 if (val == param+strlen(param)+1) 500 val[-1] = '='; 501 else if (val == param+strlen(param)+2) { 502 val[-2] = '='; 503 memmove(val-1, val, strlen(val)+1); 504 } else 505 BUG(); 506 } 507 } 508 509 /* Anything after -- gets handed straight to init. */ 510 static int __init set_init_arg(char *param, char *val, 511 const char *unused, void *arg) 512 { 513 unsigned int i; 514 515 if (panic_later) 516 return 0; 517 518 repair_env_string(param, val); 519 520 for (i = 0; argv_init[i]; i++) { 521 if (i == MAX_INIT_ARGS) { 522 panic_later = "init"; 523 panic_param = param; 524 return 0; 525 } 526 } 527 argv_init[i] = param; 528 return 0; 529 } 530 531 /* 532 * Unknown boot options get handed to init, unless they look like 533 * unused parameters (modprobe will find them in /proc/cmdline). 534 */ 535 static int __init unknown_bootoption(char *param, char *val, 536 const char *unused, void *arg) 537 { 538 size_t len = strlen(param); 539 540 repair_env_string(param, val); 541 542 /* Handle obsolete-style parameters */ 543 if (obsolete_checksetup(param)) 544 return 0; 545 546 /* Unused module parameter. */ 547 if (strnchr(param, len, '.')) 548 return 0; 549 550 if (panic_later) 551 return 0; 552 553 if (val) { 554 /* Environment option */ 555 unsigned int i; 556 for (i = 0; envp_init[i]; i++) { 557 if (i == MAX_INIT_ENVS) { 558 panic_later = "env"; 559 panic_param = param; 560 } 561 if (!strncmp(param, envp_init[i], len+1)) 562 break; 563 } 564 envp_init[i] = param; 565 } else { 566 /* Command line option */ 567 unsigned int i; 568 for (i = 0; argv_init[i]; i++) { 569 if (i == MAX_INIT_ARGS) { 570 panic_later = "init"; 571 panic_param = param; 572 } 573 } 574 argv_init[i] = param; 575 } 576 return 0; 577 } 578 579 static int __init init_setup(char *str) 580 { 581 unsigned int i; 582 583 execute_command = str; 584 /* 585 * In case LILO is going to boot us with default command line, 586 * it prepends "auto" before the whole cmdline which makes 587 * the shell think it should execute a script with such name. 588 * So we ignore all arguments entered _before_ init=... [MJ] 589 */ 590 for (i = 1; i < MAX_INIT_ARGS; i++) 591 argv_init[i] = NULL; 592 return 1; 593 } 594 __setup("init=", init_setup); 595 596 static int __init rdinit_setup(char *str) 597 { 598 unsigned int i; 599 600 ramdisk_execute_command = str; 601 /* See "auto" comment in init_setup */ 602 for (i = 1; i < MAX_INIT_ARGS; i++) 603 argv_init[i] = NULL; 604 return 1; 605 } 606 __setup("rdinit=", rdinit_setup); 607 608 #ifndef CONFIG_SMP 609 static const unsigned int setup_max_cpus = NR_CPUS; 610 static inline void setup_nr_cpu_ids(void) { } 611 static inline void smp_prepare_cpus(unsigned int maxcpus) { } 612 #endif 613 614 /* 615 * We need to store the untouched command line for future reference. 616 * We also need to store the touched command line since the parameter 617 * parsing is performed in place, and we should allow a component to 618 * store reference of name/value for future reference. 619 */ 620 static void __init setup_command_line(char *command_line) 621 { 622 size_t len, xlen = 0, ilen = 0; 623 624 if (extra_command_line) 625 xlen = strlen(extra_command_line); 626 if (extra_init_args) 627 ilen = strlen(extra_init_args) + 4; /* for " -- " */ 628 629 len = xlen + strlen(boot_command_line) + 1; 630 631 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES); 632 if (!saved_command_line) 633 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen); 634 635 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES); 636 if (!static_command_line) 637 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 638 639 if (xlen) { 640 /* 641 * We have to put extra_command_line before boot command 642 * lines because there could be dashes (separator of init 643 * command line) in the command lines. 644 */ 645 strcpy(saved_command_line, extra_command_line); 646 strcpy(static_command_line, extra_command_line); 647 } 648 strcpy(saved_command_line + xlen, boot_command_line); 649 strcpy(static_command_line + xlen, command_line); 650 651 if (ilen) { 652 /* 653 * Append supplemental init boot args to saved_command_line 654 * so that user can check what command line options passed 655 * to init. 656 * The order should always be 657 * " -- "[bootconfig init-param][cmdline init-param] 658 */ 659 if (initargs_offs) { 660 len = xlen + initargs_offs; 661 strcpy(saved_command_line + len, extra_init_args); 662 len += ilen - 4; /* strlen(extra_init_args) */ 663 strcpy(saved_command_line + len, 664 boot_command_line + initargs_offs - 1); 665 } else { 666 len = strlen(saved_command_line); 667 strcpy(saved_command_line + len, " -- "); 668 len += 4; 669 strcpy(saved_command_line + len, extra_init_args); 670 } 671 } 672 673 saved_command_line_len = strlen(saved_command_line); 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 __noreturn 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 = user_mode_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, 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 strscpy(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 poking_init(void) { } 789 790 void __init __weak pgtable_cache_init(void) { } 791 792 void __init __weak trap_init(void) { } 793 794 bool initcall_debug; 795 core_param(initcall_debug, initcall_debug, bool, 0644); 796 797 #ifdef TRACEPOINTS_ENABLED 798 static void __init initcall_debug_enable(void); 799 #else 800 static inline void initcall_debug_enable(void) 801 { 802 } 803 #endif 804 805 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET 806 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT, 807 randomize_kstack_offset); 808 DEFINE_PER_CPU(u32, kstack_offset); 809 810 static int __init early_randomize_kstack_offset(char *buf) 811 { 812 int ret; 813 bool bool_result; 814 815 ret = kstrtobool(buf, &bool_result); 816 if (ret) 817 return ret; 818 819 if (bool_result) 820 static_branch_enable(&randomize_kstack_offset); 821 else 822 static_branch_disable(&randomize_kstack_offset); 823 return 0; 824 } 825 early_param("randomize_kstack_offset", early_randomize_kstack_offset); 826 #endif 827 828 void __init __weak __noreturn arch_call_rest_init(void) 829 { 830 rest_init(); 831 } 832 833 static void __init print_unknown_bootoptions(void) 834 { 835 char *unknown_options; 836 char *end; 837 const char *const *p; 838 size_t len; 839 840 if (panic_later || (!argv_init[1] && !envp_init[2])) 841 return; 842 843 /* 844 * Determine how many options we have to print out, plus a space 845 * before each 846 */ 847 len = 1; /* null terminator */ 848 for (p = &argv_init[1]; *p; p++) { 849 len++; 850 len += strlen(*p); 851 } 852 for (p = &envp_init[2]; *p; p++) { 853 len++; 854 len += strlen(*p); 855 } 856 857 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES); 858 if (!unknown_options) { 859 pr_err("%s: Failed to allocate %zu bytes\n", 860 __func__, len); 861 return; 862 } 863 end = unknown_options; 864 865 for (p = &argv_init[1]; *p; p++) 866 end += sprintf(end, " %s", *p); 867 for (p = &envp_init[2]; *p; p++) 868 end += sprintf(end, " %s", *p); 869 870 /* Start at unknown_options[1] to skip the initial space */ 871 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n", 872 &unknown_options[1]); 873 memblock_free(unknown_options, len); 874 } 875 876 asmlinkage __visible void __init __no_sanitize_address __noreturn start_kernel(void) 877 { 878 char *command_line; 879 char *after_dashes; 880 881 set_task_stack_end_magic(&init_task); 882 smp_setup_processor_id(); 883 debug_objects_early_init(); 884 init_vmlinux_build_id(); 885 886 cgroup_init_early(); 887 888 local_irq_disable(); 889 early_boot_irqs_disabled = true; 890 891 /* 892 * Interrupts are still disabled. Do necessary setups, then 893 * enable them. 894 */ 895 boot_cpu_init(); 896 page_address_init(); 897 pr_notice("%s", linux_banner); 898 early_security_init(); 899 setup_arch(&command_line); 900 setup_boot_config(); 901 setup_command_line(command_line); 902 setup_nr_cpu_ids(); 903 setup_per_cpu_areas(); 904 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ 905 boot_cpu_hotplug_init(); 906 907 pr_notice("Kernel command line: %s\n", saved_command_line); 908 /* parameters may set static keys */ 909 jump_label_init(); 910 parse_early_param(); 911 after_dashes = parse_args("Booting kernel", 912 static_command_line, __start___param, 913 __stop___param - __start___param, 914 -1, -1, NULL, &unknown_bootoption); 915 print_unknown_bootoptions(); 916 if (!IS_ERR_OR_NULL(after_dashes)) 917 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1, 918 NULL, set_init_arg); 919 if (extra_init_args) 920 parse_args("Setting extra init args", extra_init_args, 921 NULL, 0, -1, -1, NULL, set_init_arg); 922 923 /* Architectural and non-timekeeping rng init, before allocator init */ 924 random_init_early(command_line); 925 926 /* 927 * These use large bootmem allocations and must precede 928 * initalization of page allocator 929 */ 930 setup_log_buf(0); 931 vfs_caches_init_early(); 932 sort_main_extable(); 933 trap_init(); 934 mm_core_init(); 935 poking_init(); 936 ftrace_init(); 937 938 /* trace_printk can be enabled here */ 939 early_trace_init(); 940 941 /* 942 * Set up the scheduler prior starting any interrupts (such as the 943 * timer interrupt). Full topology setup happens at smp_init() 944 * time - but meanwhile we still have a functioning scheduler. 945 */ 946 sched_init(); 947 948 if (WARN(!irqs_disabled(), 949 "Interrupts were enabled *very* early, fixing it\n")) 950 local_irq_disable(); 951 radix_tree_init(); 952 maple_tree_init(); 953 954 /* 955 * Set up housekeeping before setting up workqueues to allow the unbound 956 * workqueue to take non-housekeeping into account. 957 */ 958 housekeeping_init(); 959 960 /* 961 * Allow workqueue creation and work item queueing/cancelling 962 * early. Work item execution depends on kthreads and starts after 963 * workqueue_init(). 964 */ 965 workqueue_init_early(); 966 967 rcu_init(); 968 969 /* Trace events are available after this */ 970 trace_init(); 971 972 if (initcall_debug) 973 initcall_debug_enable(); 974 975 context_tracking_init(); 976 /* init some links before init_ISA_irqs() */ 977 early_irq_init(); 978 init_IRQ(); 979 tick_init(); 980 rcu_init_nohz(); 981 init_timers(); 982 srcu_init(); 983 hrtimers_init(); 984 softirq_init(); 985 timekeeping_init(); 986 time_init(); 987 988 /* This must be after timekeeping is initialized */ 989 random_init(); 990 991 /* These make use of the fully initialized rng */ 992 kfence_init(); 993 boot_init_stack_canary(); 994 995 perf_event_init(); 996 profile_init(); 997 call_function_init(); 998 WARN(!irqs_disabled(), "Interrupts were enabled early\n"); 999 1000 early_boot_irqs_disabled = false; 1001 local_irq_enable(); 1002 1003 kmem_cache_init_late(); 1004 1005 /* 1006 * HACK ALERT! This is early. We're enabling the console before 1007 * we've done PCI setups etc, and console_init() must be aware of 1008 * this. But we do want output early, in case something goes wrong. 1009 */ 1010 console_init(); 1011 if (panic_later) 1012 panic("Too many boot %s vars at `%s'", panic_later, 1013 panic_param); 1014 1015 lockdep_init(); 1016 1017 /* 1018 * Need to run this when irqs are enabled, because it wants 1019 * to self-test [hard/soft]-irqs on/off lock inversion bugs 1020 * too: 1021 */ 1022 locking_selftest(); 1023 1024 #ifdef CONFIG_BLK_DEV_INITRD 1025 if (initrd_start && !initrd_below_start_ok && 1026 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { 1027 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", 1028 page_to_pfn(virt_to_page((void *)initrd_start)), 1029 min_low_pfn); 1030 initrd_start = 0; 1031 } 1032 #endif 1033 setup_per_cpu_pageset(); 1034 numa_policy_init(); 1035 acpi_early_init(); 1036 if (late_time_init) 1037 late_time_init(); 1038 sched_clock_init(); 1039 calibrate_delay(); 1040 1041 arch_cpu_finalize_init(); 1042 1043 pid_idr_init(); 1044 anon_vma_init(); 1045 #ifdef CONFIG_X86 1046 if (efi_enabled(EFI_RUNTIME_SERVICES)) 1047 efi_enter_virtual_mode(); 1048 #endif 1049 thread_stack_cache_init(); 1050 cred_init(); 1051 fork_init(); 1052 proc_caches_init(); 1053 uts_ns_init(); 1054 key_init(); 1055 security_init(); 1056 dbg_late_init(); 1057 net_ns_init(); 1058 vfs_caches_init(); 1059 pagecache_init(); 1060 signals_init(); 1061 seq_file_init(); 1062 proc_root_init(); 1063 nsfs_init(); 1064 cpuset_init(); 1065 cgroup_init(); 1066 taskstats_init_early(); 1067 delayacct_init(); 1068 1069 acpi_subsystem_init(); 1070 arch_post_acpi_subsys_init(); 1071 kcsan_init(); 1072 1073 /* Do the rest non-__init'ed, we're now alive */ 1074 arch_call_rest_init(); 1075 1076 prevent_tail_call_optimization(); 1077 } 1078 1079 /* Call all constructor functions linked into the kernel. */ 1080 static void __init do_ctors(void) 1081 { 1082 /* 1083 * For UML, the constructors have already been called by the 1084 * normal setup code as it's just a normal ELF binary, so we 1085 * cannot do it again - but we do need CONFIG_CONSTRUCTORS 1086 * even on UML for modules. 1087 */ 1088 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML) 1089 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; 1090 1091 for (; fn < (ctor_fn_t *) __ctors_end; fn++) 1092 (*fn)(); 1093 #endif 1094 } 1095 1096 #ifdef CONFIG_KALLSYMS 1097 struct blacklist_entry { 1098 struct list_head next; 1099 char *buf; 1100 }; 1101 1102 static __initdata_or_module LIST_HEAD(blacklisted_initcalls); 1103 1104 static int __init initcall_blacklist(char *str) 1105 { 1106 char *str_entry; 1107 struct blacklist_entry *entry; 1108 1109 /* str argument is a comma-separated list of functions */ 1110 do { 1111 str_entry = strsep(&str, ","); 1112 if (str_entry) { 1113 pr_debug("blacklisting initcall %s\n", str_entry); 1114 entry = memblock_alloc(sizeof(*entry), 1115 SMP_CACHE_BYTES); 1116 if (!entry) 1117 panic("%s: Failed to allocate %zu bytes\n", 1118 __func__, sizeof(*entry)); 1119 entry->buf = memblock_alloc(strlen(str_entry) + 1, 1120 SMP_CACHE_BYTES); 1121 if (!entry->buf) 1122 panic("%s: Failed to allocate %zu bytes\n", 1123 __func__, strlen(str_entry) + 1); 1124 strcpy(entry->buf, str_entry); 1125 list_add(&entry->next, &blacklisted_initcalls); 1126 } 1127 } while (str_entry); 1128 1129 return 1; 1130 } 1131 1132 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1133 { 1134 struct blacklist_entry *entry; 1135 char fn_name[KSYM_SYMBOL_LEN]; 1136 unsigned long addr; 1137 1138 if (list_empty(&blacklisted_initcalls)) 1139 return false; 1140 1141 addr = (unsigned long) dereference_function_descriptor(fn); 1142 sprint_symbol_no_offset(fn_name, addr); 1143 1144 /* 1145 * fn will be "function_name [module_name]" where [module_name] is not 1146 * displayed for built-in init functions. Strip off the [module_name]. 1147 */ 1148 strreplace(fn_name, ' ', '\0'); 1149 1150 list_for_each_entry(entry, &blacklisted_initcalls, next) { 1151 if (!strcmp(fn_name, entry->buf)) { 1152 pr_debug("initcall %s blacklisted\n", fn_name); 1153 return true; 1154 } 1155 } 1156 1157 return false; 1158 } 1159 #else 1160 static int __init initcall_blacklist(char *str) 1161 { 1162 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); 1163 return 0; 1164 } 1165 1166 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1167 { 1168 return false; 1169 } 1170 #endif 1171 __setup("initcall_blacklist=", initcall_blacklist); 1172 1173 static __init_or_module void 1174 trace_initcall_start_cb(void *data, initcall_t fn) 1175 { 1176 ktime_t *calltime = data; 1177 1178 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current)); 1179 *calltime = ktime_get(); 1180 } 1181 1182 static __init_or_module void 1183 trace_initcall_finish_cb(void *data, initcall_t fn, int ret) 1184 { 1185 ktime_t rettime, *calltime = data; 1186 1187 rettime = ktime_get(); 1188 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n", 1189 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime)); 1190 } 1191 1192 static ktime_t initcall_calltime; 1193 1194 #ifdef TRACEPOINTS_ENABLED 1195 static void __init initcall_debug_enable(void) 1196 { 1197 int ret; 1198 1199 ret = register_trace_initcall_start(trace_initcall_start_cb, 1200 &initcall_calltime); 1201 ret |= register_trace_initcall_finish(trace_initcall_finish_cb, 1202 &initcall_calltime); 1203 WARN(ret, "Failed to register initcall tracepoints\n"); 1204 } 1205 # define do_trace_initcall_start trace_initcall_start 1206 # define do_trace_initcall_finish trace_initcall_finish 1207 #else 1208 static inline void do_trace_initcall_start(initcall_t fn) 1209 { 1210 if (!initcall_debug) 1211 return; 1212 trace_initcall_start_cb(&initcall_calltime, fn); 1213 } 1214 static inline void do_trace_initcall_finish(initcall_t fn, int ret) 1215 { 1216 if (!initcall_debug) 1217 return; 1218 trace_initcall_finish_cb(&initcall_calltime, fn, ret); 1219 } 1220 #endif /* !TRACEPOINTS_ENABLED */ 1221 1222 int __init_or_module do_one_initcall(initcall_t fn) 1223 { 1224 int count = preempt_count(); 1225 char msgbuf[64]; 1226 int ret; 1227 1228 if (initcall_blacklisted(fn)) 1229 return -EPERM; 1230 1231 do_trace_initcall_start(fn); 1232 ret = fn(); 1233 do_trace_initcall_finish(fn, ret); 1234 1235 msgbuf[0] = 0; 1236 1237 if (preempt_count() != count) { 1238 sprintf(msgbuf, "preemption imbalance "); 1239 preempt_count_set(count); 1240 } 1241 if (irqs_disabled()) { 1242 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); 1243 local_irq_enable(); 1244 } 1245 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf); 1246 1247 add_latent_entropy(); 1248 return ret; 1249 } 1250 1251 1252 extern initcall_entry_t __initcall_start[]; 1253 extern initcall_entry_t __initcall0_start[]; 1254 extern initcall_entry_t __initcall1_start[]; 1255 extern initcall_entry_t __initcall2_start[]; 1256 extern initcall_entry_t __initcall3_start[]; 1257 extern initcall_entry_t __initcall4_start[]; 1258 extern initcall_entry_t __initcall5_start[]; 1259 extern initcall_entry_t __initcall6_start[]; 1260 extern initcall_entry_t __initcall7_start[]; 1261 extern initcall_entry_t __initcall_end[]; 1262 1263 static initcall_entry_t *initcall_levels[] __initdata = { 1264 __initcall0_start, 1265 __initcall1_start, 1266 __initcall2_start, 1267 __initcall3_start, 1268 __initcall4_start, 1269 __initcall5_start, 1270 __initcall6_start, 1271 __initcall7_start, 1272 __initcall_end, 1273 }; 1274 1275 /* Keep these in sync with initcalls in include/linux/init.h */ 1276 static const char *initcall_level_names[] __initdata = { 1277 "pure", 1278 "core", 1279 "postcore", 1280 "arch", 1281 "subsys", 1282 "fs", 1283 "device", 1284 "late", 1285 }; 1286 1287 static int __init ignore_unknown_bootoption(char *param, char *val, 1288 const char *unused, void *arg) 1289 { 1290 return 0; 1291 } 1292 1293 static void __init do_initcall_level(int level, char *command_line) 1294 { 1295 initcall_entry_t *fn; 1296 1297 parse_args(initcall_level_names[level], 1298 command_line, __start___param, 1299 __stop___param - __start___param, 1300 level, level, 1301 NULL, ignore_unknown_bootoption); 1302 1303 trace_initcall_level(initcall_level_names[level]); 1304 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) 1305 do_one_initcall(initcall_from_entry(fn)); 1306 } 1307 1308 static void __init do_initcalls(void) 1309 { 1310 int level; 1311 size_t len = saved_command_line_len + 1; 1312 char *command_line; 1313 1314 command_line = kzalloc(len, GFP_KERNEL); 1315 if (!command_line) 1316 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 1317 1318 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) { 1319 /* Parser modifies command_line, restore it each time */ 1320 strcpy(command_line, saved_command_line); 1321 do_initcall_level(level, command_line); 1322 } 1323 1324 kfree(command_line); 1325 } 1326 1327 /* 1328 * Ok, the machine is now initialized. None of the devices 1329 * have been touched yet, but the CPU subsystem is up and 1330 * running, and memory and process management works. 1331 * 1332 * Now we can finally start doing some real work.. 1333 */ 1334 static void __init do_basic_setup(void) 1335 { 1336 cpuset_init_smp(); 1337 driver_init(); 1338 init_irq_proc(); 1339 do_ctors(); 1340 do_initcalls(); 1341 } 1342 1343 static void __init do_pre_smp_initcalls(void) 1344 { 1345 initcall_entry_t *fn; 1346 1347 trace_initcall_level("early"); 1348 for (fn = __initcall_start; fn < __initcall0_start; fn++) 1349 do_one_initcall(initcall_from_entry(fn)); 1350 } 1351 1352 static int run_init_process(const char *init_filename) 1353 { 1354 const char *const *p; 1355 1356 argv_init[0] = init_filename; 1357 pr_info("Run %s as init process\n", init_filename); 1358 pr_debug(" with arguments:\n"); 1359 for (p = argv_init; *p; p++) 1360 pr_debug(" %s\n", *p); 1361 pr_debug(" with environment:\n"); 1362 for (p = envp_init; *p; p++) 1363 pr_debug(" %s\n", *p); 1364 return kernel_execve(init_filename, argv_init, envp_init); 1365 } 1366 1367 static int try_to_run_init_process(const char *init_filename) 1368 { 1369 int ret; 1370 1371 ret = run_init_process(init_filename); 1372 1373 if (ret && ret != -ENOENT) { 1374 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", 1375 init_filename, ret); 1376 } 1377 1378 return ret; 1379 } 1380 1381 static noinline void __init kernel_init_freeable(void); 1382 1383 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX) 1384 bool rodata_enabled __ro_after_init = true; 1385 1386 #ifndef arch_parse_debug_rodata 1387 static inline bool arch_parse_debug_rodata(char *str) { return false; } 1388 #endif 1389 1390 static int __init set_debug_rodata(char *str) 1391 { 1392 if (arch_parse_debug_rodata(str)) 1393 return 0; 1394 1395 if (str && !strcmp(str, "on")) 1396 rodata_enabled = true; 1397 else if (str && !strcmp(str, "off")) 1398 rodata_enabled = false; 1399 else 1400 pr_warn("Invalid option string for rodata: '%s'\n", str); 1401 return 0; 1402 } 1403 early_param("rodata", set_debug_rodata); 1404 #endif 1405 1406 #ifdef CONFIG_STRICT_KERNEL_RWX 1407 static void mark_readonly(void) 1408 { 1409 if (rodata_enabled) { 1410 /* 1411 * load_module() results in W+X mappings, which are cleaned 1412 * up with call_rcu(). Let's make sure that queued work is 1413 * flushed so that we don't hit false positives looking for 1414 * insecure pages which are W+X. 1415 */ 1416 rcu_barrier(); 1417 mark_rodata_ro(); 1418 rodata_test(); 1419 } else 1420 pr_info("Kernel memory protection disabled.\n"); 1421 } 1422 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX) 1423 static inline void mark_readonly(void) 1424 { 1425 pr_warn("Kernel memory protection not selected by kernel config.\n"); 1426 } 1427 #else 1428 static inline void mark_readonly(void) 1429 { 1430 pr_warn("This architecture does not have kernel memory protection.\n"); 1431 } 1432 #endif 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 padata_init(); 1555 page_alloc_init_late(); 1556 1557 do_basic_setup(); 1558 1559 kunit_run_all_tests(); 1560 1561 wait_for_initramfs(); 1562 console_on_rootfs(); 1563 1564 /* 1565 * check if there is an early userspace init. If yes, let it do all 1566 * the work 1567 */ 1568 if (init_eaccess(ramdisk_execute_command) != 0) { 1569 ramdisk_execute_command = NULL; 1570 prepare_namespace(); 1571 } 1572 1573 /* 1574 * Ok, we have completed the initial bootup, and 1575 * we're essentially up and running. Get rid of the 1576 * initmem segments and start the user-mode stuff.. 1577 * 1578 * rootfs is available now, try loading the public keys 1579 * and default modules 1580 */ 1581 1582 integrity_load_keys(); 1583 } 1584