1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * linux/init/main.c 4 * 5 * Copyright (C) 1991, 1992 Linus Torvalds 6 * 7 * GK 2/5/95 - Changed to support mounting root fs via NFS 8 * Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96 9 * Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96 10 * Simplified starting of init: Michael A. Griffith <grif@acm.org> 11 */ 12 13 #define DEBUG /* Enable initcall_debug */ 14 15 #include <linux/types.h> 16 #include <linux/extable.h> 17 #include <linux/module.h> 18 #include <linux/proc_fs.h> 19 #include <linux/binfmts.h> 20 #include <linux/kernel.h> 21 #include <linux/syscalls.h> 22 #include <linux/stackprotector.h> 23 #include <linux/string.h> 24 #include <linux/ctype.h> 25 #include <linux/delay.h> 26 #include <linux/ioport.h> 27 #include <linux/init.h> 28 #include <linux/initrd.h> 29 #include <linux/memblock.h> 30 #include <linux/acpi.h> 31 #include <linux/bootconfig.h> 32 #include <linux/console.h> 33 #include <linux/nmi.h> 34 #include <linux/percpu.h> 35 #include <linux/kmod.h> 36 #include <linux/kprobes.h> 37 #include <linux/vmalloc.h> 38 #include <linux/kernel_stat.h> 39 #include <linux/start_kernel.h> 40 #include <linux/security.h> 41 #include <linux/smp.h> 42 #include <linux/profile.h> 43 #include <linux/kfence.h> 44 #include <linux/rcupdate.h> 45 #include <linux/srcu.h> 46 #include <linux/moduleparam.h> 47 #include <linux/kallsyms.h> 48 #include <linux/buildid.h> 49 #include <linux/writeback.h> 50 #include <linux/cpu.h> 51 #include <linux/cpuset.h> 52 #include <linux/cgroup.h> 53 #include <linux/efi.h> 54 #include <linux/tick.h> 55 #include <linux/sched/isolation.h> 56 #include <linux/interrupt.h> 57 #include <linux/taskstats_kern.h> 58 #include <linux/delayacct.h> 59 #include <linux/unistd.h> 60 #include <linux/utsname.h> 61 #include <linux/rmap.h> 62 #include <linux/mempolicy.h> 63 #include <linux/key.h> 64 #include <linux/page_ext.h> 65 #include <linux/debug_locks.h> 66 #include <linux/debugobjects.h> 67 #include <linux/lockdep.h> 68 #include <linux/kmemleak.h> 69 #include <linux/padata.h> 70 #include <linux/pid_namespace.h> 71 #include <linux/device/driver.h> 72 #include <linux/kthread.h> 73 #include <linux/sched.h> 74 #include <linux/sched/init.h> 75 #include <linux/signal.h> 76 #include <linux/idr.h> 77 #include <linux/kgdb.h> 78 #include <linux/ftrace.h> 79 #include <linux/async.h> 80 #include <linux/shmem_fs.h> 81 #include <linux/slab.h> 82 #include <linux/perf_event.h> 83 #include <linux/ptrace.h> 84 #include <linux/pti.h> 85 #include <linux/blkdev.h> 86 #include <linux/elevator.h> 87 #include <linux/sched/clock.h> 88 #include <linux/sched/task.h> 89 #include <linux/sched/task_stack.h> 90 #include <linux/context_tracking.h> 91 #include <linux/random.h> 92 #include <linux/list.h> 93 #include <linux/integrity.h> 94 #include <linux/proc_ns.h> 95 #include <linux/io.h> 96 #include <linux/cache.h> 97 #include <linux/rodata_test.h> 98 #include <linux/jump_label.h> 99 #include <linux/mem_encrypt.h> 100 #include <linux/kcsan.h> 101 #include <linux/init_syscalls.h> 102 #include <linux/stackdepot.h> 103 104 #include <asm/io.h> 105 #include <asm/bugs.h> 106 #include <asm/setup.h> 107 #include <asm/sections.h> 108 #include <asm/cacheflush.h> 109 110 #define CREATE_TRACE_POINTS 111 #include <trace/events/initcall.h> 112 113 #include <kunit/test.h> 114 115 static int kernel_init(void *); 116 117 extern void init_IRQ(void); 118 extern void radix_tree_init(void); 119 120 /* 121 * Debug helper: via this flag we know that we are in 'early bootup code' 122 * where only the boot processor is running with IRQ disabled. This means 123 * two things - IRQ must not be enabled before the flag is cleared and some 124 * operations which are not allowed with IRQ disabled are allowed while the 125 * flag is set. 126 */ 127 bool early_boot_irqs_disabled __read_mostly; 128 129 enum system_states system_state __read_mostly; 130 EXPORT_SYMBOL(system_state); 131 132 /* 133 * Boot command-line arguments 134 */ 135 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT 136 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT 137 138 extern void time_init(void); 139 /* Default late time init is NULL. archs can override this later. */ 140 void (*__initdata late_time_init)(void); 141 142 /* Untouched command line saved by arch-specific code. */ 143 char __initdata boot_command_line[COMMAND_LINE_SIZE]; 144 /* Untouched saved command line (eg. for /proc) */ 145 char *saved_command_line; 146 /* Command line for parameter parsing */ 147 static char *static_command_line; 148 /* Untouched extra command line */ 149 static char *extra_command_line; 150 /* Extra init arguments */ 151 static char *extra_init_args; 152 153 #ifdef CONFIG_BOOT_CONFIG 154 /* Is bootconfig on command line? */ 155 static bool bootconfig_found; 156 static size_t initargs_offs; 157 #else 158 # define bootconfig_found false 159 # define initargs_offs 0 160 #endif 161 162 static char *execute_command; 163 static char *ramdisk_execute_command = "/init"; 164 165 /* 166 * Used to generate warnings if static_key manipulation functions are used 167 * before jump_label_init is called. 168 */ 169 bool static_key_initialized __read_mostly; 170 EXPORT_SYMBOL_GPL(static_key_initialized); 171 172 /* 173 * If set, this is an indication to the drivers that reset the underlying 174 * device before going ahead with the initialization otherwise driver might 175 * rely on the BIOS and skip the reset operation. 176 * 177 * This is useful if kernel is booting in an unreliable environment. 178 * For ex. kdump situation where previous kernel has crashed, BIOS has been 179 * skipped and devices will be in unknown state. 180 */ 181 unsigned int reset_devices; 182 EXPORT_SYMBOL(reset_devices); 183 184 static int __init set_reset_devices(char *str) 185 { 186 reset_devices = 1; 187 return 1; 188 } 189 190 __setup("reset_devices", set_reset_devices); 191 192 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; 193 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; 194 static const char *panic_later, *panic_param; 195 196 extern const struct obs_kernel_param __setup_start[], __setup_end[]; 197 198 static bool __init obsolete_checksetup(char *line) 199 { 200 const struct obs_kernel_param *p; 201 bool had_early_param = false; 202 203 p = __setup_start; 204 do { 205 int n = strlen(p->str); 206 if (parameqn(line, p->str, n)) { 207 if (p->early) { 208 /* Already done in parse_early_param? 209 * (Needs exact match on param part). 210 * Keep iterating, as we can have early 211 * params and __setups of same names 8( */ 212 if (line[n] == '\0' || line[n] == '=') 213 had_early_param = true; 214 } else if (!p->setup_func) { 215 pr_warn("Parameter %s is obsolete, ignored\n", 216 p->str); 217 return true; 218 } else if (p->setup_func(line + n)) 219 return true; 220 } 221 p++; 222 } while (p < __setup_end); 223 224 return had_early_param; 225 } 226 227 /* 228 * This should be approx 2 Bo*oMips to start (note initial shift), and will 229 * still work even if initially too large, it will just take slightly longer 230 */ 231 unsigned long loops_per_jiffy = (1<<12); 232 EXPORT_SYMBOL(loops_per_jiffy); 233 234 static int __init debug_kernel(char *str) 235 { 236 console_loglevel = CONSOLE_LOGLEVEL_DEBUG; 237 return 0; 238 } 239 240 static int __init quiet_kernel(char *str) 241 { 242 console_loglevel = CONSOLE_LOGLEVEL_QUIET; 243 return 0; 244 } 245 246 early_param("debug", debug_kernel); 247 early_param("quiet", quiet_kernel); 248 249 static int __init loglevel(char *str) 250 { 251 int newlevel; 252 253 /* 254 * Only update loglevel value when a correct setting was passed, 255 * to prevent blind crashes (when loglevel being set to 0) that 256 * are quite hard to debug 257 */ 258 if (get_option(&str, &newlevel)) { 259 console_loglevel = newlevel; 260 return 0; 261 } 262 263 return -EINVAL; 264 } 265 266 early_param("loglevel", loglevel); 267 268 #ifdef CONFIG_BLK_DEV_INITRD 269 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum) 270 { 271 u32 size, csum; 272 char *data; 273 u32 *hdr; 274 int i; 275 276 if (!initrd_end) 277 return NULL; 278 279 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN; 280 /* 281 * Since Grub may align the size of initrd to 4, we must 282 * check the preceding 3 bytes as well. 283 */ 284 for (i = 0; i < 4; i++) { 285 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN)) 286 goto found; 287 data--; 288 } 289 return NULL; 290 291 found: 292 hdr = (u32 *)(data - 8); 293 size = le32_to_cpu(hdr[0]); 294 csum = le32_to_cpu(hdr[1]); 295 296 data = ((void *)hdr) - size; 297 if ((unsigned long)data < initrd_start) { 298 pr_err("bootconfig size %d is greater than initrd size %ld\n", 299 size, initrd_end - initrd_start); 300 return NULL; 301 } 302 303 /* Remove bootconfig from initramfs/initrd */ 304 initrd_end = (unsigned long)data; 305 if (_size) 306 *_size = size; 307 if (_csum) 308 *_csum = csum; 309 310 return data; 311 } 312 #else 313 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum) 314 { 315 return NULL; 316 } 317 #endif 318 319 #ifdef CONFIG_BOOT_CONFIG 320 321 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata; 322 323 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0) 324 325 static int __init xbc_snprint_cmdline(char *buf, size_t size, 326 struct xbc_node *root) 327 { 328 struct xbc_node *knode, *vnode; 329 char *end = buf + size; 330 const char *val; 331 int ret; 332 333 xbc_node_for_each_key_value(root, knode, val) { 334 ret = xbc_node_compose_key_after(root, knode, 335 xbc_namebuf, XBC_KEYLEN_MAX); 336 if (ret < 0) 337 return ret; 338 339 vnode = xbc_node_get_child(knode); 340 if (!vnode) { 341 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf); 342 if (ret < 0) 343 return ret; 344 buf += ret; 345 continue; 346 } 347 xbc_array_for_each_value(vnode, val) { 348 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ", 349 xbc_namebuf, val); 350 if (ret < 0) 351 return ret; 352 buf += ret; 353 } 354 } 355 356 return buf - (end - size); 357 } 358 #undef rest 359 360 /* Make an extra command line under given key word */ 361 static char * __init xbc_make_cmdline(const char *key) 362 { 363 struct xbc_node *root; 364 char *new_cmdline; 365 int ret, len = 0; 366 367 root = xbc_find_node(key); 368 if (!root) 369 return NULL; 370 371 /* Count required buffer size */ 372 len = xbc_snprint_cmdline(NULL, 0, root); 373 if (len <= 0) 374 return NULL; 375 376 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES); 377 if (!new_cmdline) { 378 pr_err("Failed to allocate memory for extra kernel cmdline.\n"); 379 return NULL; 380 } 381 382 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root); 383 if (ret < 0 || ret > len) { 384 pr_err("Failed to print extra kernel cmdline.\n"); 385 return NULL; 386 } 387 388 return new_cmdline; 389 } 390 391 static int __init bootconfig_params(char *param, char *val, 392 const char *unused, void *arg) 393 { 394 if (strcmp(param, "bootconfig") == 0) { 395 bootconfig_found = true; 396 } 397 return 0; 398 } 399 400 static int __init warn_bootconfig(char *str) 401 { 402 /* The 'bootconfig' has been handled by bootconfig_params(). */ 403 return 0; 404 } 405 406 static void __init setup_boot_config(void) 407 { 408 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 409 const char *msg; 410 int pos; 411 u32 size, csum; 412 char *data, *err; 413 int ret; 414 415 /* Cut out the bootconfig data even if we have no bootconfig option */ 416 data = get_boot_config_from_initrd(&size, &csum); 417 418 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 419 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL, 420 bootconfig_params); 421 422 if (IS_ERR(err) || !bootconfig_found) 423 return; 424 425 /* parse_args() stops at the next param of '--' and returns an address */ 426 if (err) 427 initargs_offs = err - tmp_cmdline; 428 429 if (!data) { 430 pr_err("'bootconfig' found on command line, but no bootconfig found\n"); 431 return; 432 } 433 434 if (size >= XBC_DATA_MAX) { 435 pr_err("bootconfig size %d greater than max size %d\n", 436 size, XBC_DATA_MAX); 437 return; 438 } 439 440 if (xbc_calc_checksum(data, size) != csum) { 441 pr_err("bootconfig checksum failed\n"); 442 return; 443 } 444 445 ret = xbc_init(data, size, &msg, &pos); 446 if (ret < 0) { 447 if (pos < 0) 448 pr_err("Failed to init bootconfig: %s.\n", msg); 449 else 450 pr_err("Failed to parse bootconfig: %s at %d.\n", 451 msg, pos); 452 } else { 453 pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret); 454 /* keys starting with "kernel." are passed via cmdline */ 455 extra_command_line = xbc_make_cmdline("kernel"); 456 /* Also, "init." keys are init arguments */ 457 extra_init_args = xbc_make_cmdline("init"); 458 } 459 return; 460 } 461 462 static void __init exit_boot_config(void) 463 { 464 xbc_destroy_all(); 465 } 466 467 #else /* !CONFIG_BOOT_CONFIG */ 468 469 static void __init setup_boot_config(void) 470 { 471 /* Remove bootconfig data from initrd */ 472 get_boot_config_from_initrd(NULL, NULL); 473 } 474 475 static int __init warn_bootconfig(char *str) 476 { 477 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n"); 478 return 0; 479 } 480 481 #define exit_boot_config() do {} while (0) 482 483 #endif /* CONFIG_BOOT_CONFIG */ 484 485 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 * The order should always be 650 * " -- "[bootconfig init-param][cmdline init-param] 651 */ 652 if (initargs_offs) { 653 len = xlen + initargs_offs; 654 strcpy(saved_command_line + len, extra_init_args); 655 len += ilen - 4; /* strlen(extra_init_args) */ 656 strcpy(saved_command_line + len, 657 boot_command_line + initargs_offs - 1); 658 } else { 659 len = strlen(saved_command_line); 660 strcpy(saved_command_line + len, " -- "); 661 len += 4; 662 strcpy(saved_command_line + len, extra_init_args); 663 } 664 } 665 } 666 667 /* 668 * We need to finalize in a non-__init function or else race conditions 669 * between the root thread and the init thread may cause start_kernel to 670 * be reaped by free_initmem before the root thread has proceeded to 671 * cpu_idle. 672 * 673 * gcc-3.4 accidentally inlines this function, so use noinline. 674 */ 675 676 static __initdata DECLARE_COMPLETION(kthreadd_done); 677 678 noinline void __ref rest_init(void) 679 { 680 struct task_struct *tsk; 681 int pid; 682 683 rcu_scheduler_starting(); 684 /* 685 * We need to spawn init first so that it obtains pid 1, however 686 * the init task will end up wanting to create kthreads, which, if 687 * we schedule it before we create kthreadd, will OOPS. 688 */ 689 pid = kernel_thread(kernel_init, NULL, CLONE_FS); 690 /* 691 * Pin init on the boot CPU. Task migration is not properly working 692 * until sched_init_smp() has been run. It will set the allowed 693 * CPUs for init to the non isolated CPUs. 694 */ 695 rcu_read_lock(); 696 tsk = find_task_by_pid_ns(pid, &init_pid_ns); 697 tsk->flags |= PF_NO_SETAFFINITY; 698 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id())); 699 rcu_read_unlock(); 700 701 numa_default_policy(); 702 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); 703 rcu_read_lock(); 704 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); 705 rcu_read_unlock(); 706 707 /* 708 * Enable might_sleep() and smp_processor_id() checks. 709 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y 710 * kernel_thread() would trigger might_sleep() splats. With 711 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled 712 * already, but it's stuck on the kthreadd_done completion. 713 */ 714 system_state = SYSTEM_SCHEDULING; 715 716 complete(&kthreadd_done); 717 718 /* 719 * The boot idle thread must execute schedule() 720 * at least once to get things moving: 721 */ 722 schedule_preempt_disabled(); 723 /* Call into cpu_idle with preempt disabled */ 724 cpu_startup_entry(CPUHP_ONLINE); 725 } 726 727 /* Check for early params. */ 728 static int __init do_early_param(char *param, char *val, 729 const char *unused, void *arg) 730 { 731 const struct obs_kernel_param *p; 732 733 for (p = __setup_start; p < __setup_end; p++) { 734 if ((p->early && parameq(param, p->str)) || 735 (strcmp(param, "console") == 0 && 736 strcmp(p->str, "earlycon") == 0) 737 ) { 738 if (p->setup_func(val) != 0) 739 pr_warn("Malformed early option '%s'\n", param); 740 } 741 } 742 /* We accept everything at this stage. */ 743 return 0; 744 } 745 746 void __init parse_early_options(char *cmdline) 747 { 748 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL, 749 do_early_param); 750 } 751 752 /* Arch code calls this early on, or if not, just before other parsing. */ 753 void __init parse_early_param(void) 754 { 755 static int done __initdata; 756 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 757 758 if (done) 759 return; 760 761 /* All fall through to do_early_param. */ 762 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 763 parse_early_options(tmp_cmdline); 764 done = 1; 765 } 766 767 void __init __weak arch_post_acpi_subsys_init(void) { } 768 769 void __init __weak smp_setup_processor_id(void) 770 { 771 } 772 773 # if THREAD_SIZE >= PAGE_SIZE 774 void __init __weak thread_stack_cache_init(void) 775 { 776 } 777 #endif 778 779 void __init __weak mem_encrypt_init(void) { } 780 781 void __init __weak poking_init(void) { } 782 783 void __init __weak pgtable_cache_init(void) { } 784 785 void __init __weak trap_init(void) { } 786 787 bool initcall_debug; 788 core_param(initcall_debug, initcall_debug, bool, 0644); 789 790 #ifdef TRACEPOINTS_ENABLED 791 static void __init initcall_debug_enable(void); 792 #else 793 static inline void initcall_debug_enable(void) 794 { 795 } 796 #endif 797 798 /* Report memory auto-initialization states for this boot. */ 799 static void __init report_meminit(void) 800 { 801 const char *stack; 802 803 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN)) 804 stack = "all(pattern)"; 805 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO)) 806 stack = "all(zero)"; 807 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL)) 808 stack = "byref_all(zero)"; 809 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF)) 810 stack = "byref(zero)"; 811 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER)) 812 stack = "__user(zero)"; 813 else 814 stack = "off"; 815 816 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n", 817 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off", 818 want_init_on_free() ? "on" : "off"); 819 if (want_init_on_free()) 820 pr_info("mem auto-init: clearing system memory may take some time...\n"); 821 } 822 823 /* 824 * Set up kernel memory allocators 825 */ 826 static void __init mm_init(void) 827 { 828 /* 829 * page_ext requires contiguous pages, 830 * bigger than MAX_ORDER unless SPARSEMEM. 831 */ 832 page_ext_init_flatmem(); 833 init_mem_debugging_and_hardening(); 834 kfence_alloc_pool(); 835 report_meminit(); 836 stack_depot_init(); 837 mem_init(); 838 mem_init_print_info(); 839 /* page_owner must be initialized after buddy is ready */ 840 page_ext_init_flatmem_late(); 841 kmem_cache_init(); 842 kmemleak_init(); 843 pgtable_init(); 844 debug_objects_mem_init(); 845 vmalloc_init(); 846 /* Should be run before the first non-init thread is created */ 847 init_espfix_bsp(); 848 /* Should be run after espfix64 is set up. */ 849 pti_init(); 850 } 851 852 #ifdef CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET 853 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT, 854 randomize_kstack_offset); 855 DEFINE_PER_CPU(u32, kstack_offset); 856 857 static int __init early_randomize_kstack_offset(char *buf) 858 { 859 int ret; 860 bool bool_result; 861 862 ret = kstrtobool(buf, &bool_result); 863 if (ret) 864 return ret; 865 866 if (bool_result) 867 static_branch_enable(&randomize_kstack_offset); 868 else 869 static_branch_disable(&randomize_kstack_offset); 870 return 0; 871 } 872 early_param("randomize_kstack_offset", early_randomize_kstack_offset); 873 #endif 874 875 void __init __weak arch_call_rest_init(void) 876 { 877 rest_init(); 878 } 879 880 static void __init print_unknown_bootoptions(void) 881 { 882 char *unknown_options; 883 char *end; 884 const char *const *p; 885 size_t len; 886 887 if (panic_later || (!argv_init[1] && !envp_init[2])) 888 return; 889 890 /* 891 * Determine how many options we have to print out, plus a space 892 * before each 893 */ 894 len = 1; /* null terminator */ 895 for (p = &argv_init[1]; *p; p++) { 896 len++; 897 len += strlen(*p); 898 } 899 for (p = &envp_init[2]; *p; p++) { 900 len++; 901 len += strlen(*p); 902 } 903 904 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES); 905 if (!unknown_options) { 906 pr_err("%s: Failed to allocate %zu bytes\n", 907 __func__, len); 908 return; 909 } 910 end = unknown_options; 911 912 for (p = &argv_init[1]; *p; p++) 913 end += sprintf(end, " %s", *p); 914 for (p = &envp_init[2]; *p; p++) 915 end += sprintf(end, " %s", *p); 916 917 pr_notice("Unknown command line parameters:%s\n", unknown_options); 918 memblock_free_ptr(unknown_options, len); 919 } 920 921 asmlinkage __visible void __init __no_sanitize_address start_kernel(void) 922 { 923 char *command_line; 924 char *after_dashes; 925 926 set_task_stack_end_magic(&init_task); 927 smp_setup_processor_id(); 928 debug_objects_early_init(); 929 init_vmlinux_build_id(); 930 931 cgroup_init_early(); 932 933 local_irq_disable(); 934 early_boot_irqs_disabled = true; 935 936 /* 937 * Interrupts are still disabled. Do necessary setups, then 938 * enable them. 939 */ 940 boot_cpu_init(); 941 page_address_init(); 942 pr_notice("%s", linux_banner); 943 early_security_init(); 944 setup_arch(&command_line); 945 setup_boot_config(); 946 setup_command_line(command_line); 947 setup_nr_cpu_ids(); 948 setup_per_cpu_areas(); 949 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ 950 boot_cpu_hotplug_init(); 951 952 build_all_zonelists(NULL); 953 page_alloc_init(); 954 955 pr_notice("Kernel command line: %s\n", saved_command_line); 956 /* parameters may set static keys */ 957 jump_label_init(); 958 parse_early_param(); 959 after_dashes = parse_args("Booting kernel", 960 static_command_line, __start___param, 961 __stop___param - __start___param, 962 -1, -1, NULL, &unknown_bootoption); 963 print_unknown_bootoptions(); 964 if (!IS_ERR_OR_NULL(after_dashes)) 965 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1, 966 NULL, set_init_arg); 967 if (extra_init_args) 968 parse_args("Setting extra init args", extra_init_args, 969 NULL, 0, -1, -1, NULL, set_init_arg); 970 971 /* 972 * These use large bootmem allocations and must precede 973 * kmem_cache_init() 974 */ 975 setup_log_buf(0); 976 vfs_caches_init_early(); 977 sort_main_extable(); 978 trap_init(); 979 mm_init(); 980 981 ftrace_init(); 982 983 /* trace_printk can be enabled here */ 984 early_trace_init(); 985 986 /* 987 * Set up the scheduler prior starting any interrupts (such as the 988 * timer interrupt). Full topology setup happens at smp_init() 989 * time - but meanwhile we still have a functioning scheduler. 990 */ 991 sched_init(); 992 993 if (WARN(!irqs_disabled(), 994 "Interrupts were enabled *very* early, fixing it\n")) 995 local_irq_disable(); 996 radix_tree_init(); 997 998 /* 999 * Set up housekeeping before setting up workqueues to allow the unbound 1000 * workqueue to take non-housekeeping into account. 1001 */ 1002 housekeeping_init(); 1003 1004 /* 1005 * Allow workqueue creation and work item queueing/cancelling 1006 * early. Work item execution depends on kthreads and starts after 1007 * workqueue_init(). 1008 */ 1009 workqueue_init_early(); 1010 1011 rcu_init(); 1012 1013 /* Trace events are available after this */ 1014 trace_init(); 1015 1016 if (initcall_debug) 1017 initcall_debug_enable(); 1018 1019 context_tracking_init(); 1020 /* init some links before init_ISA_irqs() */ 1021 early_irq_init(); 1022 init_IRQ(); 1023 tick_init(); 1024 rcu_init_nohz(); 1025 init_timers(); 1026 srcu_init(); 1027 hrtimers_init(); 1028 softirq_init(); 1029 timekeeping_init(); 1030 kfence_init(); 1031 1032 /* 1033 * For best initial stack canary entropy, prepare it after: 1034 * - setup_arch() for any UEFI RNG entropy and boot cmdline access 1035 * - timekeeping_init() for ktime entropy used in rand_initialize() 1036 * - rand_initialize() to get any arch-specific entropy like RDRAND 1037 * - add_latent_entropy() to get any latent entropy 1038 * - adding command line entropy 1039 */ 1040 rand_initialize(); 1041 add_latent_entropy(); 1042 add_device_randomness(command_line, strlen(command_line)); 1043 boot_init_stack_canary(); 1044 1045 time_init(); 1046 perf_event_init(); 1047 profile_init(); 1048 call_function_init(); 1049 WARN(!irqs_disabled(), "Interrupts were enabled early\n"); 1050 1051 early_boot_irqs_disabled = false; 1052 local_irq_enable(); 1053 1054 kmem_cache_init_late(); 1055 1056 /* 1057 * HACK ALERT! This is early. We're enabling the console before 1058 * we've done PCI setups etc, and console_init() must be aware of 1059 * this. But we do want output early, in case something goes wrong. 1060 */ 1061 console_init(); 1062 if (panic_later) 1063 panic("Too many boot %s vars at `%s'", panic_later, 1064 panic_param); 1065 1066 lockdep_init(); 1067 1068 /* 1069 * Need to run this when irqs are enabled, because it wants 1070 * to self-test [hard/soft]-irqs on/off lock inversion bugs 1071 * too: 1072 */ 1073 locking_selftest(); 1074 1075 /* 1076 * This needs to be called before any devices perform DMA 1077 * operations that might use the SWIOTLB bounce buffers. It will 1078 * mark the bounce buffers as decrypted so that their usage will 1079 * not cause "plain-text" data to be decrypted when accessed. 1080 */ 1081 mem_encrypt_init(); 1082 1083 #ifdef CONFIG_BLK_DEV_INITRD 1084 if (initrd_start && !initrd_below_start_ok && 1085 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { 1086 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", 1087 page_to_pfn(virt_to_page((void *)initrd_start)), 1088 min_low_pfn); 1089 initrd_start = 0; 1090 } 1091 #endif 1092 setup_per_cpu_pageset(); 1093 numa_policy_init(); 1094 acpi_early_init(); 1095 if (late_time_init) 1096 late_time_init(); 1097 sched_clock_init(); 1098 calibrate_delay(); 1099 pid_idr_init(); 1100 anon_vma_init(); 1101 #ifdef CONFIG_X86 1102 if (efi_enabled(EFI_RUNTIME_SERVICES)) 1103 efi_enter_virtual_mode(); 1104 #endif 1105 thread_stack_cache_init(); 1106 cred_init(); 1107 fork_init(); 1108 proc_caches_init(); 1109 uts_ns_init(); 1110 key_init(); 1111 security_init(); 1112 dbg_late_init(); 1113 vfs_caches_init(); 1114 pagecache_init(); 1115 signals_init(); 1116 seq_file_init(); 1117 proc_root_init(); 1118 nsfs_init(); 1119 cpuset_init(); 1120 cgroup_init(); 1121 taskstats_init_early(); 1122 delayacct_init(); 1123 1124 poking_init(); 1125 check_bugs(); 1126 1127 acpi_subsystem_init(); 1128 arch_post_acpi_subsys_init(); 1129 kcsan_init(); 1130 1131 /* Do the rest non-__init'ed, we're now alive */ 1132 arch_call_rest_init(); 1133 1134 prevent_tail_call_optimization(); 1135 } 1136 1137 /* Call all constructor functions linked into the kernel. */ 1138 static void __init do_ctors(void) 1139 { 1140 /* 1141 * For UML, the constructors have already been called by the 1142 * normal setup code as it's just a normal ELF binary, so we 1143 * cannot do it again - but we do need CONFIG_CONSTRUCTORS 1144 * even on UML for modules. 1145 */ 1146 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML) 1147 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; 1148 1149 for (; fn < (ctor_fn_t *) __ctors_end; fn++) 1150 (*fn)(); 1151 #endif 1152 } 1153 1154 #ifdef CONFIG_KALLSYMS 1155 struct blacklist_entry { 1156 struct list_head next; 1157 char *buf; 1158 }; 1159 1160 static __initdata_or_module LIST_HEAD(blacklisted_initcalls); 1161 1162 static int __init initcall_blacklist(char *str) 1163 { 1164 char *str_entry; 1165 struct blacklist_entry *entry; 1166 1167 /* str argument is a comma-separated list of functions */ 1168 do { 1169 str_entry = strsep(&str, ","); 1170 if (str_entry) { 1171 pr_debug("blacklisting initcall %s\n", str_entry); 1172 entry = memblock_alloc(sizeof(*entry), 1173 SMP_CACHE_BYTES); 1174 if (!entry) 1175 panic("%s: Failed to allocate %zu bytes\n", 1176 __func__, sizeof(*entry)); 1177 entry->buf = memblock_alloc(strlen(str_entry) + 1, 1178 SMP_CACHE_BYTES); 1179 if (!entry->buf) 1180 panic("%s: Failed to allocate %zu bytes\n", 1181 __func__, strlen(str_entry) + 1); 1182 strcpy(entry->buf, str_entry); 1183 list_add(&entry->next, &blacklisted_initcalls); 1184 } 1185 } while (str_entry); 1186 1187 return 0; 1188 } 1189 1190 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1191 { 1192 struct blacklist_entry *entry; 1193 char fn_name[KSYM_SYMBOL_LEN]; 1194 unsigned long addr; 1195 1196 if (list_empty(&blacklisted_initcalls)) 1197 return false; 1198 1199 addr = (unsigned long) dereference_function_descriptor(fn); 1200 sprint_symbol_no_offset(fn_name, addr); 1201 1202 /* 1203 * fn will be "function_name [module_name]" where [module_name] is not 1204 * displayed for built-in init functions. Strip off the [module_name]. 1205 */ 1206 strreplace(fn_name, ' ', '\0'); 1207 1208 list_for_each_entry(entry, &blacklisted_initcalls, next) { 1209 if (!strcmp(fn_name, entry->buf)) { 1210 pr_debug("initcall %s blacklisted\n", fn_name); 1211 return true; 1212 } 1213 } 1214 1215 return false; 1216 } 1217 #else 1218 static int __init initcall_blacklist(char *str) 1219 { 1220 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); 1221 return 0; 1222 } 1223 1224 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1225 { 1226 return false; 1227 } 1228 #endif 1229 __setup("initcall_blacklist=", initcall_blacklist); 1230 1231 static __init_or_module void 1232 trace_initcall_start_cb(void *data, initcall_t fn) 1233 { 1234 ktime_t *calltime = (ktime_t *)data; 1235 1236 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current)); 1237 *calltime = ktime_get(); 1238 } 1239 1240 static __init_or_module void 1241 trace_initcall_finish_cb(void *data, initcall_t fn, int ret) 1242 { 1243 ktime_t *calltime = (ktime_t *)data; 1244 ktime_t delta, rettime; 1245 unsigned long long duration; 1246 1247 rettime = ktime_get(); 1248 delta = ktime_sub(rettime, *calltime); 1249 duration = (unsigned long long) ktime_to_ns(delta) >> 10; 1250 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n", 1251 fn, ret, duration); 1252 } 1253 1254 static ktime_t initcall_calltime; 1255 1256 #ifdef TRACEPOINTS_ENABLED 1257 static void __init initcall_debug_enable(void) 1258 { 1259 int ret; 1260 1261 ret = register_trace_initcall_start(trace_initcall_start_cb, 1262 &initcall_calltime); 1263 ret |= register_trace_initcall_finish(trace_initcall_finish_cb, 1264 &initcall_calltime); 1265 WARN(ret, "Failed to register initcall tracepoints\n"); 1266 } 1267 # define do_trace_initcall_start trace_initcall_start 1268 # define do_trace_initcall_finish trace_initcall_finish 1269 #else 1270 static inline void do_trace_initcall_start(initcall_t fn) 1271 { 1272 if (!initcall_debug) 1273 return; 1274 trace_initcall_start_cb(&initcall_calltime, fn); 1275 } 1276 static inline void do_trace_initcall_finish(initcall_t fn, int ret) 1277 { 1278 if (!initcall_debug) 1279 return; 1280 trace_initcall_finish_cb(&initcall_calltime, fn, ret); 1281 } 1282 #endif /* !TRACEPOINTS_ENABLED */ 1283 1284 int __init_or_module do_one_initcall(initcall_t fn) 1285 { 1286 int count = preempt_count(); 1287 char msgbuf[64]; 1288 int ret; 1289 1290 if (initcall_blacklisted(fn)) 1291 return -EPERM; 1292 1293 do_trace_initcall_start(fn); 1294 ret = fn(); 1295 do_trace_initcall_finish(fn, ret); 1296 1297 msgbuf[0] = 0; 1298 1299 if (preempt_count() != count) { 1300 sprintf(msgbuf, "preemption imbalance "); 1301 preempt_count_set(count); 1302 } 1303 if (irqs_disabled()) { 1304 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); 1305 local_irq_enable(); 1306 } 1307 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf); 1308 1309 add_latent_entropy(); 1310 return ret; 1311 } 1312 1313 1314 extern initcall_entry_t __initcall_start[]; 1315 extern initcall_entry_t __initcall0_start[]; 1316 extern initcall_entry_t __initcall1_start[]; 1317 extern initcall_entry_t __initcall2_start[]; 1318 extern initcall_entry_t __initcall3_start[]; 1319 extern initcall_entry_t __initcall4_start[]; 1320 extern initcall_entry_t __initcall5_start[]; 1321 extern initcall_entry_t __initcall6_start[]; 1322 extern initcall_entry_t __initcall7_start[]; 1323 extern initcall_entry_t __initcall_end[]; 1324 1325 static initcall_entry_t *initcall_levels[] __initdata = { 1326 __initcall0_start, 1327 __initcall1_start, 1328 __initcall2_start, 1329 __initcall3_start, 1330 __initcall4_start, 1331 __initcall5_start, 1332 __initcall6_start, 1333 __initcall7_start, 1334 __initcall_end, 1335 }; 1336 1337 /* Keep these in sync with initcalls in include/linux/init.h */ 1338 static const char *initcall_level_names[] __initdata = { 1339 "pure", 1340 "core", 1341 "postcore", 1342 "arch", 1343 "subsys", 1344 "fs", 1345 "device", 1346 "late", 1347 }; 1348 1349 static int __init ignore_unknown_bootoption(char *param, char *val, 1350 const char *unused, void *arg) 1351 { 1352 return 0; 1353 } 1354 1355 static void __init do_initcall_level(int level, char *command_line) 1356 { 1357 initcall_entry_t *fn; 1358 1359 parse_args(initcall_level_names[level], 1360 command_line, __start___param, 1361 __stop___param - __start___param, 1362 level, level, 1363 NULL, ignore_unknown_bootoption); 1364 1365 trace_initcall_level(initcall_level_names[level]); 1366 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) 1367 do_one_initcall(initcall_from_entry(fn)); 1368 } 1369 1370 static void __init do_initcalls(void) 1371 { 1372 int level; 1373 size_t len = strlen(saved_command_line) + 1; 1374 char *command_line; 1375 1376 command_line = kzalloc(len, GFP_KERNEL); 1377 if (!command_line) 1378 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 1379 1380 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) { 1381 /* Parser modifies command_line, restore it each time */ 1382 strcpy(command_line, saved_command_line); 1383 do_initcall_level(level, command_line); 1384 } 1385 1386 kfree(command_line); 1387 } 1388 1389 /* 1390 * Ok, the machine is now initialized. None of the devices 1391 * have been touched yet, but the CPU subsystem is up and 1392 * running, and memory and process management works. 1393 * 1394 * Now we can finally start doing some real work.. 1395 */ 1396 static void __init do_basic_setup(void) 1397 { 1398 cpuset_init_smp(); 1399 driver_init(); 1400 init_irq_proc(); 1401 do_ctors(); 1402 do_initcalls(); 1403 } 1404 1405 static void __init do_pre_smp_initcalls(void) 1406 { 1407 initcall_entry_t *fn; 1408 1409 trace_initcall_level("early"); 1410 for (fn = __initcall_start; fn < __initcall0_start; fn++) 1411 do_one_initcall(initcall_from_entry(fn)); 1412 } 1413 1414 static int run_init_process(const char *init_filename) 1415 { 1416 const char *const *p; 1417 1418 argv_init[0] = init_filename; 1419 pr_info("Run %s as init process\n", init_filename); 1420 pr_debug(" with arguments:\n"); 1421 for (p = argv_init; *p; p++) 1422 pr_debug(" %s\n", *p); 1423 pr_debug(" with environment:\n"); 1424 for (p = envp_init; *p; p++) 1425 pr_debug(" %s\n", *p); 1426 return kernel_execve(init_filename, argv_init, envp_init); 1427 } 1428 1429 static int try_to_run_init_process(const char *init_filename) 1430 { 1431 int ret; 1432 1433 ret = run_init_process(init_filename); 1434 1435 if (ret && ret != -ENOENT) { 1436 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", 1437 init_filename, ret); 1438 } 1439 1440 return ret; 1441 } 1442 1443 static noinline void __init kernel_init_freeable(void); 1444 1445 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX) 1446 bool rodata_enabled __ro_after_init = true; 1447 static int __init set_debug_rodata(char *str) 1448 { 1449 return strtobool(str, &rodata_enabled); 1450 } 1451 __setup("rodata=", set_debug_rodata); 1452 #endif 1453 1454 #ifdef CONFIG_STRICT_KERNEL_RWX 1455 static void mark_readonly(void) 1456 { 1457 if (rodata_enabled) { 1458 /* 1459 * load_module() results in W+X mappings, which are cleaned 1460 * up with call_rcu(). Let's make sure that queued work is 1461 * flushed so that we don't hit false positives looking for 1462 * insecure pages which are W+X. 1463 */ 1464 rcu_barrier(); 1465 mark_rodata_ro(); 1466 rodata_test(); 1467 } else 1468 pr_info("Kernel memory protection disabled.\n"); 1469 } 1470 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX) 1471 static inline void mark_readonly(void) 1472 { 1473 pr_warn("Kernel memory protection not selected by kernel config.\n"); 1474 } 1475 #else 1476 static inline void mark_readonly(void) 1477 { 1478 pr_warn("This architecture does not have kernel memory protection.\n"); 1479 } 1480 #endif 1481 1482 void __weak free_initmem(void) 1483 { 1484 free_initmem_default(POISON_FREE_INITMEM); 1485 } 1486 1487 static int __ref kernel_init(void *unused) 1488 { 1489 int ret; 1490 1491 /* 1492 * Wait until kthreadd is all set-up. 1493 */ 1494 wait_for_completion(&kthreadd_done); 1495 1496 kernel_init_freeable(); 1497 /* need to finish all async __init code before freeing the memory */ 1498 async_synchronize_full(); 1499 kprobe_free_init_mem(); 1500 ftrace_free_init_mem(); 1501 kgdb_free_init_mem(); 1502 exit_boot_config(); 1503 free_initmem(); 1504 mark_readonly(); 1505 1506 /* 1507 * Kernel mappings are now finalized - update the userspace page-table 1508 * to finalize PTI. 1509 */ 1510 pti_finalize(); 1511 1512 system_state = SYSTEM_RUNNING; 1513 numa_default_policy(); 1514 1515 rcu_end_inkernel_boot(); 1516 1517 do_sysctl_args(); 1518 1519 if (ramdisk_execute_command) { 1520 ret = run_init_process(ramdisk_execute_command); 1521 if (!ret) 1522 return 0; 1523 pr_err("Failed to execute %s (error %d)\n", 1524 ramdisk_execute_command, ret); 1525 } 1526 1527 /* 1528 * We try each of these until one succeeds. 1529 * 1530 * The Bourne shell can be used instead of init if we are 1531 * trying to recover a really broken machine. 1532 */ 1533 if (execute_command) { 1534 ret = run_init_process(execute_command); 1535 if (!ret) 1536 return 0; 1537 panic("Requested init %s failed (error %d).", 1538 execute_command, ret); 1539 } 1540 1541 if (CONFIG_DEFAULT_INIT[0] != '\0') { 1542 ret = run_init_process(CONFIG_DEFAULT_INIT); 1543 if (ret) 1544 pr_err("Default init %s failed (error %d)\n", 1545 CONFIG_DEFAULT_INIT, ret); 1546 else 1547 return 0; 1548 } 1549 1550 if (!try_to_run_init_process("/sbin/init") || 1551 !try_to_run_init_process("/etc/init") || 1552 !try_to_run_init_process("/bin/init") || 1553 !try_to_run_init_process("/bin/sh")) 1554 return 0; 1555 1556 panic("No working init found. Try passing init= option to kernel. " 1557 "See Linux Documentation/admin-guide/init.rst for guidance."); 1558 } 1559 1560 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */ 1561 void __init console_on_rootfs(void) 1562 { 1563 struct file *file = filp_open("/dev/console", O_RDWR, 0); 1564 1565 if (IS_ERR(file)) { 1566 pr_err("Warning: unable to open an initial console.\n"); 1567 return; 1568 } 1569 init_dup(file); 1570 init_dup(file); 1571 init_dup(file); 1572 fput(file); 1573 } 1574 1575 static noinline void __init kernel_init_freeable(void) 1576 { 1577 /* Now the scheduler is fully set up and can do blocking allocations */ 1578 gfp_allowed_mask = __GFP_BITS_MASK; 1579 1580 /* 1581 * init can allocate pages on any node 1582 */ 1583 set_mems_allowed(node_states[N_MEMORY]); 1584 1585 cad_pid = get_pid(task_pid(current)); 1586 1587 smp_prepare_cpus(setup_max_cpus); 1588 1589 workqueue_init(); 1590 1591 init_mm_internals(); 1592 1593 rcu_init_tasks_generic(); 1594 do_pre_smp_initcalls(); 1595 lockup_detector_init(); 1596 1597 smp_init(); 1598 sched_init_smp(); 1599 1600 padata_init(); 1601 page_alloc_init_late(); 1602 /* Initialize page ext after all struct pages are initialized. */ 1603 page_ext_init(); 1604 1605 do_basic_setup(); 1606 1607 kunit_run_all_tests(); 1608 1609 wait_for_initramfs(); 1610 console_on_rootfs(); 1611 1612 /* 1613 * check if there is an early userspace init. If yes, let it do all 1614 * the work 1615 */ 1616 if (init_eaccess(ramdisk_execute_command) != 0) { 1617 ramdisk_execute_command = NULL; 1618 prepare_namespace(); 1619 } 1620 1621 /* 1622 * Ok, we have completed the initial bootup, and 1623 * we're essentially up and running. Get rid of the 1624 * initmem segments and start the user-mode stuff.. 1625 * 1626 * rootfs is available now, try loading the public keys 1627 * and default modules 1628 */ 1629 1630 integrity_load_keys(); 1631 } 1632