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