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 bool initargs_found; 157 #else 158 # define bootconfig_found false 159 # define initargs_found false 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 '--' and returns an address */ 426 if (err) 427 initargs_found = true; 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 #else 472 473 static void __init setup_boot_config(void) 474 { 475 /* Remove bootconfig data from initrd */ 476 get_boot_config_from_initrd(NULL, 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 #endif 485 early_param("bootconfig", warn_bootconfig); 486 487 /* Change NUL term back to "=", to make "param" the whole string. */ 488 static void __init repair_env_string(char *param, char *val) 489 { 490 if (val) { 491 /* param=val or param="val"? */ 492 if (val == param+strlen(param)+1) 493 val[-1] = '='; 494 else if (val == param+strlen(param)+2) { 495 val[-2] = '='; 496 memmove(val-1, val, strlen(val)+1); 497 } else 498 BUG(); 499 } 500 } 501 502 /* Anything after -- gets handed straight to init. */ 503 static int __init set_init_arg(char *param, char *val, 504 const char *unused, void *arg) 505 { 506 unsigned int i; 507 508 if (panic_later) 509 return 0; 510 511 repair_env_string(param, val); 512 513 for (i = 0; argv_init[i]; i++) { 514 if (i == MAX_INIT_ARGS) { 515 panic_later = "init"; 516 panic_param = param; 517 return 0; 518 } 519 } 520 argv_init[i] = param; 521 return 0; 522 } 523 524 /* 525 * Unknown boot options get handed to init, unless they look like 526 * unused parameters (modprobe will find them in /proc/cmdline). 527 */ 528 static int __init unknown_bootoption(char *param, char *val, 529 const char *unused, void *arg) 530 { 531 size_t len = strlen(param); 532 533 repair_env_string(param, val); 534 535 /* Handle obsolete-style parameters */ 536 if (obsolete_checksetup(param)) 537 return 0; 538 539 /* Unused module parameter. */ 540 if (strnchr(param, len, '.')) 541 return 0; 542 543 if (panic_later) 544 return 0; 545 546 if (val) { 547 /* Environment option */ 548 unsigned int i; 549 for (i = 0; envp_init[i]; i++) { 550 if (i == MAX_INIT_ENVS) { 551 panic_later = "env"; 552 panic_param = param; 553 } 554 if (!strncmp(param, envp_init[i], len+1)) 555 break; 556 } 557 envp_init[i] = param; 558 } else { 559 /* Command line option */ 560 unsigned int i; 561 for (i = 0; argv_init[i]; i++) { 562 if (i == MAX_INIT_ARGS) { 563 panic_later = "init"; 564 panic_param = param; 565 } 566 } 567 argv_init[i] = param; 568 } 569 return 0; 570 } 571 572 static int __init init_setup(char *str) 573 { 574 unsigned int i; 575 576 execute_command = str; 577 /* 578 * In case LILO is going to boot us with default command line, 579 * it prepends "auto" before the whole cmdline which makes 580 * the shell think it should execute a script with such name. 581 * So we ignore all arguments entered _before_ init=... [MJ] 582 */ 583 for (i = 1; i < MAX_INIT_ARGS; i++) 584 argv_init[i] = NULL; 585 return 1; 586 } 587 __setup("init=", init_setup); 588 589 static int __init rdinit_setup(char *str) 590 { 591 unsigned int i; 592 593 ramdisk_execute_command = str; 594 /* See "auto" comment in init_setup */ 595 for (i = 1; i < MAX_INIT_ARGS; i++) 596 argv_init[i] = NULL; 597 return 1; 598 } 599 __setup("rdinit=", rdinit_setup); 600 601 #ifndef CONFIG_SMP 602 static const unsigned int setup_max_cpus = NR_CPUS; 603 static inline void setup_nr_cpu_ids(void) { } 604 static inline void smp_prepare_cpus(unsigned int maxcpus) { } 605 #endif 606 607 /* 608 * We need to store the untouched command line for future reference. 609 * We also need to store the touched command line since the parameter 610 * parsing is performed in place, and we should allow a component to 611 * store reference of name/value for future reference. 612 */ 613 static void __init setup_command_line(char *command_line) 614 { 615 size_t len, xlen = 0, ilen = 0; 616 617 if (extra_command_line) 618 xlen = strlen(extra_command_line); 619 if (extra_init_args) 620 ilen = strlen(extra_init_args) + 4; /* for " -- " */ 621 622 len = xlen + strlen(boot_command_line) + 1; 623 624 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES); 625 if (!saved_command_line) 626 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen); 627 628 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES); 629 if (!static_command_line) 630 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 631 632 if (xlen) { 633 /* 634 * We have to put extra_command_line before boot command 635 * lines because there could be dashes (separator of init 636 * command line) in the command lines. 637 */ 638 strcpy(saved_command_line, extra_command_line); 639 strcpy(static_command_line, extra_command_line); 640 } 641 strcpy(saved_command_line + xlen, boot_command_line); 642 strcpy(static_command_line + xlen, command_line); 643 644 if (ilen) { 645 /* 646 * Append supplemental init boot args to saved_command_line 647 * so that user can check what command line options passed 648 * to init. 649 */ 650 len = strlen(saved_command_line); 651 if (initargs_found) { 652 saved_command_line[len++] = ' '; 653 } else { 654 strcpy(saved_command_line + len, " -- "); 655 len += 4; 656 } 657 658 strcpy(saved_command_line + len, extra_init_args); 659 } 660 } 661 662 /* 663 * We need to finalize in a non-__init function or else race conditions 664 * between the root thread and the init thread may cause start_kernel to 665 * be reaped by free_initmem before the root thread has proceeded to 666 * cpu_idle. 667 * 668 * gcc-3.4 accidentally inlines this function, so use noinline. 669 */ 670 671 static __initdata DECLARE_COMPLETION(kthreadd_done); 672 673 noinline void __ref rest_init(void) 674 { 675 struct task_struct *tsk; 676 int pid; 677 678 rcu_scheduler_starting(); 679 /* 680 * We need to spawn init first so that it obtains pid 1, however 681 * the init task will end up wanting to create kthreads, which, if 682 * we schedule it before we create kthreadd, will OOPS. 683 */ 684 pid = kernel_thread(kernel_init, NULL, CLONE_FS); 685 /* 686 * Pin init on the boot CPU. Task migration is not properly working 687 * until sched_init_smp() has been run. It will set the allowed 688 * CPUs for init to the non isolated CPUs. 689 */ 690 rcu_read_lock(); 691 tsk = find_task_by_pid_ns(pid, &init_pid_ns); 692 tsk->flags |= PF_NO_SETAFFINITY; 693 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id())); 694 rcu_read_unlock(); 695 696 numa_default_policy(); 697 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); 698 rcu_read_lock(); 699 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); 700 rcu_read_unlock(); 701 702 /* 703 * Enable might_sleep() and smp_processor_id() checks. 704 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y 705 * kernel_thread() would trigger might_sleep() splats. With 706 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled 707 * already, but it's stuck on the kthreadd_done completion. 708 */ 709 system_state = SYSTEM_SCHEDULING; 710 711 complete(&kthreadd_done); 712 713 /* 714 * The boot idle thread must execute schedule() 715 * at least once to get things moving: 716 */ 717 schedule_preempt_disabled(); 718 /* Call into cpu_idle with preempt disabled */ 719 cpu_startup_entry(CPUHP_ONLINE); 720 } 721 722 /* Check for early params. */ 723 static int __init do_early_param(char *param, char *val, 724 const char *unused, void *arg) 725 { 726 const struct obs_kernel_param *p; 727 728 for (p = __setup_start; p < __setup_end; p++) { 729 if ((p->early && parameq(param, p->str)) || 730 (strcmp(param, "console") == 0 && 731 strcmp(p->str, "earlycon") == 0) 732 ) { 733 if (p->setup_func(val) != 0) 734 pr_warn("Malformed early option '%s'\n", param); 735 } 736 } 737 /* We accept everything at this stage. */ 738 return 0; 739 } 740 741 void __init parse_early_options(char *cmdline) 742 { 743 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL, 744 do_early_param); 745 } 746 747 /* Arch code calls this early on, or if not, just before other parsing. */ 748 void __init parse_early_param(void) 749 { 750 static int done __initdata; 751 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 752 753 if (done) 754 return; 755 756 /* All fall through to do_early_param. */ 757 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 758 parse_early_options(tmp_cmdline); 759 done = 1; 760 } 761 762 void __init __weak arch_post_acpi_subsys_init(void) { } 763 764 void __init __weak smp_setup_processor_id(void) 765 { 766 } 767 768 # if THREAD_SIZE >= PAGE_SIZE 769 void __init __weak thread_stack_cache_init(void) 770 { 771 } 772 #endif 773 774 void __init __weak mem_encrypt_init(void) { } 775 776 void __init __weak poking_init(void) { } 777 778 void __init __weak pgtable_cache_init(void) { } 779 780 void __init __weak trap_init(void) { } 781 782 bool initcall_debug; 783 core_param(initcall_debug, initcall_debug, bool, 0644); 784 785 #ifdef TRACEPOINTS_ENABLED 786 static void __init initcall_debug_enable(void); 787 #else 788 static inline void initcall_debug_enable(void) 789 { 790 } 791 #endif 792 793 /* Report memory auto-initialization states for this boot. */ 794 static void __init report_meminit(void) 795 { 796 const char *stack; 797 798 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN)) 799 stack = "all(pattern)"; 800 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO)) 801 stack = "all(zero)"; 802 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL)) 803 stack = "byref_all(zero)"; 804 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF)) 805 stack = "byref(zero)"; 806 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER)) 807 stack = "__user(zero)"; 808 else 809 stack = "off"; 810 811 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n", 812 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off", 813 want_init_on_free() ? "on" : "off"); 814 if (want_init_on_free()) 815 pr_info("mem auto-init: clearing system memory may take some time...\n"); 816 } 817 818 /* 819 * Set up kernel memory allocators 820 */ 821 static void __init mm_init(void) 822 { 823 /* 824 * page_ext requires contiguous pages, 825 * bigger than MAX_ORDER unless SPARSEMEM. 826 */ 827 page_ext_init_flatmem(); 828 init_mem_debugging_and_hardening(); 829 kfence_alloc_pool(); 830 report_meminit(); 831 stack_depot_init(); 832 mem_init(); 833 mem_init_print_info(); 834 /* page_owner must be initialized after buddy is ready */ 835 page_ext_init_flatmem_late(); 836 kmem_cache_init(); 837 kmemleak_init(); 838 pgtable_init(); 839 debug_objects_mem_init(); 840 vmalloc_init(); 841 /* Should be run before the first non-init thread is created */ 842 init_espfix_bsp(); 843 /* Should be run after espfix64 is set up. */ 844 pti_init(); 845 } 846 847 #ifdef CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET 848 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT, 849 randomize_kstack_offset); 850 DEFINE_PER_CPU(u32, kstack_offset); 851 852 static int __init early_randomize_kstack_offset(char *buf) 853 { 854 int ret; 855 bool bool_result; 856 857 ret = kstrtobool(buf, &bool_result); 858 if (ret) 859 return ret; 860 861 if (bool_result) 862 static_branch_enable(&randomize_kstack_offset); 863 else 864 static_branch_disable(&randomize_kstack_offset); 865 return 0; 866 } 867 early_param("randomize_kstack_offset", early_randomize_kstack_offset); 868 #endif 869 870 void __init __weak arch_call_rest_init(void) 871 { 872 rest_init(); 873 } 874 875 static void __init print_unknown_bootoptions(void) 876 { 877 char *unknown_options; 878 char *end; 879 const char *const *p; 880 size_t len; 881 882 if (panic_later || (!argv_init[1] && !envp_init[2])) 883 return; 884 885 /* 886 * Determine how many options we have to print out, plus a space 887 * before each 888 */ 889 len = 1; /* null terminator */ 890 for (p = &argv_init[1]; *p; p++) { 891 len++; 892 len += strlen(*p); 893 } 894 for (p = &envp_init[2]; *p; p++) { 895 len++; 896 len += strlen(*p); 897 } 898 899 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES); 900 if (!unknown_options) { 901 pr_err("%s: Failed to allocate %zu bytes\n", 902 __func__, len); 903 return; 904 } 905 end = unknown_options; 906 907 for (p = &argv_init[1]; *p; p++) 908 end += sprintf(end, " %s", *p); 909 for (p = &envp_init[2]; *p; p++) 910 end += sprintf(end, " %s", *p); 911 912 pr_notice("Unknown command line parameters:%s\n", unknown_options); 913 memblock_free(__pa(unknown_options), len); 914 } 915 916 asmlinkage __visible void __init __no_sanitize_address start_kernel(void) 917 { 918 char *command_line; 919 char *after_dashes; 920 921 set_task_stack_end_magic(&init_task); 922 smp_setup_processor_id(); 923 debug_objects_early_init(); 924 init_vmlinux_build_id(); 925 926 cgroup_init_early(); 927 928 local_irq_disable(); 929 early_boot_irqs_disabled = true; 930 931 /* 932 * Interrupts are still disabled. Do necessary setups, then 933 * enable them. 934 */ 935 boot_cpu_init(); 936 page_address_init(); 937 pr_notice("%s", linux_banner); 938 early_security_init(); 939 setup_arch(&command_line); 940 setup_boot_config(); 941 setup_command_line(command_line); 942 setup_nr_cpu_ids(); 943 setup_per_cpu_areas(); 944 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ 945 boot_cpu_hotplug_init(); 946 947 build_all_zonelists(NULL); 948 page_alloc_init(); 949 950 pr_notice("Kernel command line: %s\n", saved_command_line); 951 /* parameters may set static keys */ 952 jump_label_init(); 953 parse_early_param(); 954 after_dashes = parse_args("Booting kernel", 955 static_command_line, __start___param, 956 __stop___param - __start___param, 957 -1, -1, NULL, &unknown_bootoption); 958 print_unknown_bootoptions(); 959 if (!IS_ERR_OR_NULL(after_dashes)) 960 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1, 961 NULL, set_init_arg); 962 if (extra_init_args) 963 parse_args("Setting extra init args", extra_init_args, 964 NULL, 0, -1, -1, NULL, set_init_arg); 965 966 /* 967 * These use large bootmem allocations and must precede 968 * kmem_cache_init() 969 */ 970 setup_log_buf(0); 971 vfs_caches_init_early(); 972 sort_main_extable(); 973 trap_init(); 974 mm_init(); 975 976 ftrace_init(); 977 978 /* trace_printk can be enabled here */ 979 early_trace_init(); 980 981 /* 982 * Set up the scheduler prior starting any interrupts (such as the 983 * timer interrupt). Full topology setup happens at smp_init() 984 * time - but meanwhile we still have a functioning scheduler. 985 */ 986 sched_init(); 987 988 if (WARN(!irqs_disabled(), 989 "Interrupts were enabled *very* early, fixing it\n")) 990 local_irq_disable(); 991 radix_tree_init(); 992 993 /* 994 * Set up housekeeping before setting up workqueues to allow the unbound 995 * workqueue to take non-housekeeping into account. 996 */ 997 housekeeping_init(); 998 999 /* 1000 * Allow workqueue creation and work item queueing/cancelling 1001 * early. Work item execution depends on kthreads and starts after 1002 * workqueue_init(). 1003 */ 1004 workqueue_init_early(); 1005 1006 rcu_init(); 1007 1008 /* Trace events are available after this */ 1009 trace_init(); 1010 1011 if (initcall_debug) 1012 initcall_debug_enable(); 1013 1014 context_tracking_init(); 1015 /* init some links before init_ISA_irqs() */ 1016 early_irq_init(); 1017 init_IRQ(); 1018 tick_init(); 1019 rcu_init_nohz(); 1020 init_timers(); 1021 srcu_init(); 1022 hrtimers_init(); 1023 softirq_init(); 1024 timekeeping_init(); 1025 kfence_init(); 1026 1027 /* 1028 * For best initial stack canary entropy, prepare it after: 1029 * - setup_arch() for any UEFI RNG entropy and boot cmdline access 1030 * - timekeeping_init() for ktime entropy used in rand_initialize() 1031 * - rand_initialize() to get any arch-specific entropy like RDRAND 1032 * - add_latent_entropy() to get any latent entropy 1033 * - adding command line entropy 1034 */ 1035 rand_initialize(); 1036 add_latent_entropy(); 1037 add_device_randomness(command_line, strlen(command_line)); 1038 boot_init_stack_canary(); 1039 1040 time_init(); 1041 perf_event_init(); 1042 profile_init(); 1043 call_function_init(); 1044 WARN(!irqs_disabled(), "Interrupts were enabled early\n"); 1045 1046 early_boot_irqs_disabled = false; 1047 local_irq_enable(); 1048 1049 kmem_cache_init_late(); 1050 1051 /* 1052 * HACK ALERT! This is early. We're enabling the console before 1053 * we've done PCI setups etc, and console_init() must be aware of 1054 * this. But we do want output early, in case something goes wrong. 1055 */ 1056 console_init(); 1057 if (panic_later) 1058 panic("Too many boot %s vars at `%s'", panic_later, 1059 panic_param); 1060 1061 lockdep_init(); 1062 1063 /* 1064 * Need to run this when irqs are enabled, because it wants 1065 * to self-test [hard/soft]-irqs on/off lock inversion bugs 1066 * too: 1067 */ 1068 locking_selftest(); 1069 1070 /* 1071 * This needs to be called before any devices perform DMA 1072 * operations that might use the SWIOTLB bounce buffers. It will 1073 * mark the bounce buffers as decrypted so that their usage will 1074 * not cause "plain-text" data to be decrypted when accessed. 1075 */ 1076 mem_encrypt_init(); 1077 1078 #ifdef CONFIG_BLK_DEV_INITRD 1079 if (initrd_start && !initrd_below_start_ok && 1080 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { 1081 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", 1082 page_to_pfn(virt_to_page((void *)initrd_start)), 1083 min_low_pfn); 1084 initrd_start = 0; 1085 } 1086 #endif 1087 setup_per_cpu_pageset(); 1088 numa_policy_init(); 1089 acpi_early_init(); 1090 if (late_time_init) 1091 late_time_init(); 1092 sched_clock_init(); 1093 calibrate_delay(); 1094 pid_idr_init(); 1095 anon_vma_init(); 1096 #ifdef CONFIG_X86 1097 if (efi_enabled(EFI_RUNTIME_SERVICES)) 1098 efi_enter_virtual_mode(); 1099 #endif 1100 thread_stack_cache_init(); 1101 cred_init(); 1102 fork_init(); 1103 proc_caches_init(); 1104 uts_ns_init(); 1105 key_init(); 1106 security_init(); 1107 dbg_late_init(); 1108 vfs_caches_init(); 1109 pagecache_init(); 1110 signals_init(); 1111 seq_file_init(); 1112 proc_root_init(); 1113 nsfs_init(); 1114 cpuset_init(); 1115 cgroup_init(); 1116 taskstats_init_early(); 1117 delayacct_init(); 1118 1119 poking_init(); 1120 check_bugs(); 1121 1122 acpi_subsystem_init(); 1123 arch_post_acpi_subsys_init(); 1124 kcsan_init(); 1125 1126 /* Do the rest non-__init'ed, we're now alive */ 1127 arch_call_rest_init(); 1128 1129 prevent_tail_call_optimization(); 1130 } 1131 1132 /* Call all constructor functions linked into the kernel. */ 1133 static void __init do_ctors(void) 1134 { 1135 /* 1136 * For UML, the constructors have already been called by the 1137 * normal setup code as it's just a normal ELF binary, so we 1138 * cannot do it again - but we do need CONFIG_CONSTRUCTORS 1139 * even on UML for modules. 1140 */ 1141 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML) 1142 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; 1143 1144 for (; fn < (ctor_fn_t *) __ctors_end; fn++) 1145 (*fn)(); 1146 #endif 1147 } 1148 1149 #ifdef CONFIG_KALLSYMS 1150 struct blacklist_entry { 1151 struct list_head next; 1152 char *buf; 1153 }; 1154 1155 static __initdata_or_module LIST_HEAD(blacklisted_initcalls); 1156 1157 static int __init initcall_blacklist(char *str) 1158 { 1159 char *str_entry; 1160 struct blacklist_entry *entry; 1161 1162 /* str argument is a comma-separated list of functions */ 1163 do { 1164 str_entry = strsep(&str, ","); 1165 if (str_entry) { 1166 pr_debug("blacklisting initcall %s\n", str_entry); 1167 entry = memblock_alloc(sizeof(*entry), 1168 SMP_CACHE_BYTES); 1169 if (!entry) 1170 panic("%s: Failed to allocate %zu bytes\n", 1171 __func__, sizeof(*entry)); 1172 entry->buf = memblock_alloc(strlen(str_entry) + 1, 1173 SMP_CACHE_BYTES); 1174 if (!entry->buf) 1175 panic("%s: Failed to allocate %zu bytes\n", 1176 __func__, strlen(str_entry) + 1); 1177 strcpy(entry->buf, str_entry); 1178 list_add(&entry->next, &blacklisted_initcalls); 1179 } 1180 } while (str_entry); 1181 1182 return 0; 1183 } 1184 1185 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1186 { 1187 struct blacklist_entry *entry; 1188 char fn_name[KSYM_SYMBOL_LEN]; 1189 unsigned long addr; 1190 1191 if (list_empty(&blacklisted_initcalls)) 1192 return false; 1193 1194 addr = (unsigned long) dereference_function_descriptor(fn); 1195 sprint_symbol_no_offset(fn_name, addr); 1196 1197 /* 1198 * fn will be "function_name [module_name]" where [module_name] is not 1199 * displayed for built-in init functions. Strip off the [module_name]. 1200 */ 1201 strreplace(fn_name, ' ', '\0'); 1202 1203 list_for_each_entry(entry, &blacklisted_initcalls, next) { 1204 if (!strcmp(fn_name, entry->buf)) { 1205 pr_debug("initcall %s blacklisted\n", fn_name); 1206 return true; 1207 } 1208 } 1209 1210 return false; 1211 } 1212 #else 1213 static int __init initcall_blacklist(char *str) 1214 { 1215 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); 1216 return 0; 1217 } 1218 1219 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1220 { 1221 return false; 1222 } 1223 #endif 1224 __setup("initcall_blacklist=", initcall_blacklist); 1225 1226 static __init_or_module void 1227 trace_initcall_start_cb(void *data, initcall_t fn) 1228 { 1229 ktime_t *calltime = (ktime_t *)data; 1230 1231 printk(KERN_DEBUG "calling %pS @ %i irqs_disabled() %d\n", fn, task_pid_nr(current), irqs_disabled()); 1232 *calltime = ktime_get(); 1233 } 1234 1235 static __init_or_module void 1236 trace_initcall_finish_cb(void *data, initcall_t fn, int ret) 1237 { 1238 ktime_t *calltime = (ktime_t *)data; 1239 ktime_t delta, rettime; 1240 unsigned long long duration; 1241 1242 rettime = ktime_get(); 1243 delta = ktime_sub(rettime, *calltime); 1244 duration = (unsigned long long) ktime_to_ns(delta) >> 10; 1245 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs, irqs_disabled() %d\n", 1246 fn, ret, duration, irqs_disabled()); 1247 } 1248 1249 static ktime_t initcall_calltime; 1250 1251 #ifdef TRACEPOINTS_ENABLED 1252 static void __init initcall_debug_enable(void) 1253 { 1254 int ret; 1255 1256 ret = register_trace_initcall_start(trace_initcall_start_cb, 1257 &initcall_calltime); 1258 ret |= register_trace_initcall_finish(trace_initcall_finish_cb, 1259 &initcall_calltime); 1260 WARN(ret, "Failed to register initcall tracepoints\n"); 1261 } 1262 # define do_trace_initcall_start trace_initcall_start 1263 # define do_trace_initcall_finish trace_initcall_finish 1264 #else 1265 static inline void do_trace_initcall_start(initcall_t fn) 1266 { 1267 if (!initcall_debug) 1268 return; 1269 trace_initcall_start_cb(&initcall_calltime, fn); 1270 } 1271 static inline void do_trace_initcall_finish(initcall_t fn, int ret) 1272 { 1273 if (!initcall_debug) 1274 return; 1275 trace_initcall_finish_cb(&initcall_calltime, fn, ret); 1276 } 1277 #endif /* !TRACEPOINTS_ENABLED */ 1278 1279 int __init_or_module do_one_initcall(initcall_t fn) 1280 { 1281 int count = preempt_count(); 1282 char msgbuf[64]; 1283 int ret; 1284 1285 if (initcall_blacklisted(fn)) 1286 return -EPERM; 1287 1288 do_trace_initcall_start(fn); 1289 ret = fn(); 1290 do_trace_initcall_finish(fn, ret); 1291 1292 msgbuf[0] = 0; 1293 1294 if (preempt_count() != count) { 1295 sprintf(msgbuf, "preemption imbalance "); 1296 preempt_count_set(count); 1297 } 1298 if (irqs_disabled()) { 1299 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); 1300 local_irq_enable(); 1301 } 1302 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf); 1303 1304 add_latent_entropy(); 1305 return ret; 1306 } 1307 1308 1309 extern initcall_entry_t __initcall_start[]; 1310 extern initcall_entry_t __initcall0_start[]; 1311 extern initcall_entry_t __initcall1_start[]; 1312 extern initcall_entry_t __initcall2_start[]; 1313 extern initcall_entry_t __initcall3_start[]; 1314 extern initcall_entry_t __initcall4_start[]; 1315 extern initcall_entry_t __initcall5_start[]; 1316 extern initcall_entry_t __initcall6_start[]; 1317 extern initcall_entry_t __initcall7_start[]; 1318 extern initcall_entry_t __initcall_end[]; 1319 1320 static initcall_entry_t *initcall_levels[] __initdata = { 1321 __initcall0_start, 1322 __initcall1_start, 1323 __initcall2_start, 1324 __initcall3_start, 1325 __initcall4_start, 1326 __initcall5_start, 1327 __initcall6_start, 1328 __initcall7_start, 1329 __initcall_end, 1330 }; 1331 1332 /* Keep these in sync with initcalls in include/linux/init.h */ 1333 static const char *initcall_level_names[] __initdata = { 1334 "pure", 1335 "core", 1336 "postcore", 1337 "arch", 1338 "subsys", 1339 "fs", 1340 "device", 1341 "late", 1342 }; 1343 1344 static int __init ignore_unknown_bootoption(char *param, char *val, 1345 const char *unused, void *arg) 1346 { 1347 return 0; 1348 } 1349 1350 static void __init do_initcall_level(int level, char *command_line) 1351 { 1352 initcall_entry_t *fn; 1353 1354 parse_args(initcall_level_names[level], 1355 command_line, __start___param, 1356 __stop___param - __start___param, 1357 level, level, 1358 NULL, ignore_unknown_bootoption); 1359 1360 trace_initcall_level(initcall_level_names[level]); 1361 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) 1362 do_one_initcall(initcall_from_entry(fn)); 1363 } 1364 1365 static void __init do_initcalls(void) 1366 { 1367 int level; 1368 size_t len = strlen(saved_command_line) + 1; 1369 char *command_line; 1370 1371 command_line = kzalloc(len, GFP_KERNEL); 1372 if (!command_line) 1373 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 1374 1375 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) { 1376 /* Parser modifies command_line, restore it each time */ 1377 strcpy(command_line, saved_command_line); 1378 do_initcall_level(level, command_line); 1379 } 1380 1381 kfree(command_line); 1382 } 1383 1384 /* 1385 * Ok, the machine is now initialized. None of the devices 1386 * have been touched yet, but the CPU subsystem is up and 1387 * running, and memory and process management works. 1388 * 1389 * Now we can finally start doing some real work.. 1390 */ 1391 static void __init do_basic_setup(void) 1392 { 1393 cpuset_init_smp(); 1394 driver_init(); 1395 init_irq_proc(); 1396 do_ctors(); 1397 do_initcalls(); 1398 } 1399 1400 static void __init do_pre_smp_initcalls(void) 1401 { 1402 initcall_entry_t *fn; 1403 1404 trace_initcall_level("early"); 1405 for (fn = __initcall_start; fn < __initcall0_start; fn++) 1406 do_one_initcall(initcall_from_entry(fn)); 1407 } 1408 1409 static int run_init_process(const char *init_filename) 1410 { 1411 const char *const *p; 1412 1413 argv_init[0] = init_filename; 1414 pr_info("Run %s as init process\n", init_filename); 1415 pr_debug(" with arguments:\n"); 1416 for (p = argv_init; *p; p++) 1417 pr_debug(" %s\n", *p); 1418 pr_debug(" with environment:\n"); 1419 for (p = envp_init; *p; p++) 1420 pr_debug(" %s\n", *p); 1421 return kernel_execve(init_filename, argv_init, envp_init); 1422 } 1423 1424 static int try_to_run_init_process(const char *init_filename) 1425 { 1426 int ret; 1427 1428 ret = run_init_process(init_filename); 1429 1430 if (ret && ret != -ENOENT) { 1431 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", 1432 init_filename, ret); 1433 } 1434 1435 return ret; 1436 } 1437 1438 static noinline void __init kernel_init_freeable(void); 1439 1440 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX) 1441 bool rodata_enabled __ro_after_init = true; 1442 static int __init set_debug_rodata(char *str) 1443 { 1444 return strtobool(str, &rodata_enabled); 1445 } 1446 __setup("rodata=", set_debug_rodata); 1447 #endif 1448 1449 #ifdef CONFIG_STRICT_KERNEL_RWX 1450 static void mark_readonly(void) 1451 { 1452 if (rodata_enabled) { 1453 /* 1454 * load_module() results in W+X mappings, which are cleaned 1455 * up with call_rcu(). Let's make sure that queued work is 1456 * flushed so that we don't hit false positives looking for 1457 * insecure pages which are W+X. 1458 */ 1459 rcu_barrier(); 1460 mark_rodata_ro(); 1461 rodata_test(); 1462 } else 1463 pr_info("Kernel memory protection disabled.\n"); 1464 } 1465 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX) 1466 static inline void mark_readonly(void) 1467 { 1468 pr_warn("Kernel memory protection not selected by kernel config.\n"); 1469 } 1470 #else 1471 static inline void mark_readonly(void) 1472 { 1473 pr_warn("This architecture does not have kernel memory protection.\n"); 1474 } 1475 #endif 1476 1477 void __weak free_initmem(void) 1478 { 1479 free_initmem_default(POISON_FREE_INITMEM); 1480 } 1481 1482 static int __ref kernel_init(void *unused) 1483 { 1484 int ret; 1485 1486 /* 1487 * Wait until kthreadd is all set-up. 1488 */ 1489 wait_for_completion(&kthreadd_done); 1490 1491 kernel_init_freeable(); 1492 /* need to finish all async __init code before freeing the memory */ 1493 async_synchronize_full(); 1494 kprobe_free_init_mem(); 1495 ftrace_free_init_mem(); 1496 kgdb_free_init_mem(); 1497 free_initmem(); 1498 mark_readonly(); 1499 1500 /* 1501 * Kernel mappings are now finalized - update the userspace page-table 1502 * to finalize PTI. 1503 */ 1504 pti_finalize(); 1505 1506 system_state = SYSTEM_RUNNING; 1507 numa_default_policy(); 1508 1509 rcu_end_inkernel_boot(); 1510 1511 do_sysctl_args(); 1512 1513 if (ramdisk_execute_command) { 1514 ret = run_init_process(ramdisk_execute_command); 1515 if (!ret) 1516 return 0; 1517 pr_err("Failed to execute %s (error %d)\n", 1518 ramdisk_execute_command, ret); 1519 } 1520 1521 /* 1522 * We try each of these until one succeeds. 1523 * 1524 * The Bourne shell can be used instead of init if we are 1525 * trying to recover a really broken machine. 1526 */ 1527 if (execute_command) { 1528 ret = run_init_process(execute_command); 1529 if (!ret) 1530 return 0; 1531 panic("Requested init %s failed (error %d).", 1532 execute_command, ret); 1533 } 1534 1535 if (CONFIG_DEFAULT_INIT[0] != '\0') { 1536 ret = run_init_process(CONFIG_DEFAULT_INIT); 1537 if (ret) 1538 pr_err("Default init %s failed (error %d)\n", 1539 CONFIG_DEFAULT_INIT, ret); 1540 else 1541 return 0; 1542 } 1543 1544 if (!try_to_run_init_process("/sbin/init") || 1545 !try_to_run_init_process("/etc/init") || 1546 !try_to_run_init_process("/bin/init") || 1547 !try_to_run_init_process("/bin/sh")) 1548 return 0; 1549 1550 panic("No working init found. Try passing init= option to kernel. " 1551 "See Linux Documentation/admin-guide/init.rst for guidance."); 1552 } 1553 1554 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */ 1555 void __init console_on_rootfs(void) 1556 { 1557 struct file *file = filp_open("/dev/console", O_RDWR, 0); 1558 1559 if (IS_ERR(file)) { 1560 pr_err("Warning: unable to open an initial console.\n"); 1561 return; 1562 } 1563 init_dup(file); 1564 init_dup(file); 1565 init_dup(file); 1566 fput(file); 1567 } 1568 1569 static noinline void __init kernel_init_freeable(void) 1570 { 1571 /* Now the scheduler is fully set up and can do blocking allocations */ 1572 gfp_allowed_mask = __GFP_BITS_MASK; 1573 1574 /* 1575 * init can allocate pages on any node 1576 */ 1577 set_mems_allowed(node_states[N_MEMORY]); 1578 1579 cad_pid = get_pid(task_pid(current)); 1580 1581 smp_prepare_cpus(setup_max_cpus); 1582 1583 workqueue_init(); 1584 1585 init_mm_internals(); 1586 1587 rcu_init_tasks_generic(); 1588 do_pre_smp_initcalls(); 1589 lockup_detector_init(); 1590 1591 smp_init(); 1592 sched_init_smp(); 1593 1594 padata_init(); 1595 page_alloc_init_late(); 1596 /* Initialize page ext after all struct pages are initialized. */ 1597 page_ext_init(); 1598 1599 do_basic_setup(); 1600 1601 kunit_run_all_tests(); 1602 1603 wait_for_initramfs(); 1604 console_on_rootfs(); 1605 1606 /* 1607 * check if there is an early userspace init. If yes, let it do all 1608 * the work 1609 */ 1610 if (init_eaccess(ramdisk_execute_command) != 0) { 1611 ramdisk_execute_command = NULL; 1612 prepare_namespace(); 1613 } 1614 1615 /* 1616 * Ok, we have completed the initial bootup, and 1617 * we're essentially up and running. Get rid of the 1618 * initmem segments and start the user-mode stuff.. 1619 * 1620 * rootfs is available now, try loading the public keys 1621 * and default modules 1622 */ 1623 1624 integrity_load_keys(); 1625 } 1626