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/export.h> 17 #include <linux/extable.h> 18 #include <linux/module.h> 19 #include <linux/proc_fs.h> 20 #include <linux/binfmts.h> 21 #include <linux/kernel.h> 22 #include <linux/syscalls.h> 23 #include <linux/stackprotector.h> 24 #include <linux/string.h> 25 #include <linux/ctype.h> 26 #include <linux/delay.h> 27 #include <linux/ioport.h> 28 #include <linux/init.h> 29 #include <linux/initrd.h> 30 #include <linux/memblock.h> 31 #include <linux/acpi.h> 32 #include <linux/bootconfig.h> 33 #include <linux/console.h> 34 #include <linux/nmi.h> 35 #include <linux/percpu.h> 36 #include <linux/kmod.h> 37 #include <linux/kprobes.h> 38 #include <linux/kmsan.h> 39 #include <linux/vmalloc.h> 40 #include <linux/kernel_stat.h> 41 #include <linux/start_kernel.h> 42 #include <linux/security.h> 43 #include <linux/smp.h> 44 #include <linux/profile.h> 45 #include <linux/kfence.h> 46 #include <linux/rcupdate.h> 47 #include <linux/srcu.h> 48 #include <linux/moduleparam.h> 49 #include <linux/kallsyms.h> 50 #include <linux/buildid.h> 51 #include <linux/writeback.h> 52 #include <linux/cpu.h> 53 #include <linux/cpuset.h> 54 #include <linux/memcontrol.h> 55 #include <linux/cgroup.h> 56 #include <linux/tick.h> 57 #include <linux/sched/isolation.h> 58 #include <linux/interrupt.h> 59 #include <linux/taskstats_kern.h> 60 #include <linux/delayacct.h> 61 #include <linux/unistd.h> 62 #include <linux/utsname.h> 63 #include <linux/rmap.h> 64 #include <linux/mempolicy.h> 65 #include <linux/key.h> 66 #include <linux/debug_locks.h> 67 #include <linux/debugobjects.h> 68 #include <linux/lockdep.h> 69 #include <linux/kmemleak.h> 70 #include <linux/padata.h> 71 #include <linux/pid_namespace.h> 72 #include <linux/device/driver.h> 73 #include <linux/kthread.h> 74 #include <linux/sched.h> 75 #include <linux/sched/init.h> 76 #include <linux/signal.h> 77 #include <linux/idr.h> 78 #include <linux/kgdb.h> 79 #include <linux/ftrace.h> 80 #include <linux/async.h> 81 #include <linux/shmem_fs.h> 82 #include <linux/slab.h> 83 #include <linux/perf_event.h> 84 #include <linux/ptrace.h> 85 #include <linux/pti.h> 86 #include <linux/blkdev.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/moduleloader.h> 93 #include <linux/list.h> 94 #include <linux/integrity.h> 95 #include <linux/proc_ns.h> 96 #include <linux/io.h> 97 #include <linux/cache.h> 98 #include <linux/rodata_test.h> 99 #include <linux/jump_label.h> 100 #include <linux/kcsan.h> 101 #include <linux/init_syscalls.h> 102 #include <linux/stackdepot.h> 103 #include <linux/randomize_kstack.h> 104 #include <linux/pidfs.h> 105 #include <linux/ptdump.h> 106 #include <linux/time_namespace.h> 107 #include <linux/unaligned.h> 108 #include <net/net_namespace.h> 109 110 #include <asm/io.h> 111 #include <asm/setup.h> 112 #include <asm/sections.h> 113 #include <asm/cacheflush.h> 114 115 #define CREATE_TRACE_POINTS 116 #include <trace/events/initcall.h> 117 118 #include <kunit/test.h> 119 120 static int kernel_init(void *); 121 122 /* 123 * Debug helper: via this flag we know that we are in 'early bootup code' 124 * where only the boot processor is running with IRQ disabled. This means 125 * two things - IRQ must not be enabled before the flag is cleared and some 126 * operations which are not allowed with IRQ disabled are allowed while the 127 * flag is set. 128 */ 129 bool early_boot_irqs_disabled __read_mostly; 130 131 enum system_states system_state __read_mostly; 132 EXPORT_SYMBOL(system_state); 133 134 /* 135 * Boot command-line arguments 136 */ 137 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT 138 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT 139 140 /* Default late time init is NULL. archs can override this later. */ 141 void (*__initdata late_time_init)(void); 142 143 /* Untouched command line saved by arch-specific code. */ 144 char __initdata boot_command_line[COMMAND_LINE_SIZE]; 145 /* Untouched saved command line (eg. for /proc) */ 146 char *saved_command_line __ro_after_init; 147 unsigned int saved_command_line_len __ro_after_init; 148 /* Command line for parameter parsing */ 149 static char *static_command_line; 150 /* Untouched extra command line */ 151 static char *extra_command_line; 152 /* Extra init arguments */ 153 static char *extra_init_args; 154 155 #ifdef CONFIG_BOOT_CONFIG 156 /* Is bootconfig on command line? */ 157 static bool bootconfig_found; 158 static size_t initargs_offs; 159 #else 160 # define bootconfig_found false 161 # define initargs_offs 0 162 #endif 163 164 static char *execute_command; 165 static char *ramdisk_execute_command = "/init"; 166 static bool __initdata ramdisk_execute_command_set; 167 168 /* 169 * Used to generate warnings if static_key manipulation functions are used 170 * before jump_label_init is called. 171 */ 172 bool static_key_initialized __read_mostly; 173 EXPORT_SYMBOL_GPL(static_key_initialized); 174 175 /* 176 * If set, this is an indication to the drivers that reset the underlying 177 * device before going ahead with the initialization otherwise driver might 178 * rely on the BIOS and skip the reset operation. 179 * 180 * This is useful if kernel is booting in an unreliable environment. 181 * For ex. kdump situation where previous kernel has crashed, BIOS has been 182 * skipped and devices will be in unknown state. 183 */ 184 unsigned int reset_devices; 185 EXPORT_SYMBOL(reset_devices); 186 187 static int __init set_reset_devices(char *str) 188 { 189 reset_devices = 1; 190 return 1; 191 } 192 193 __setup("reset_devices", set_reset_devices); 194 195 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; 196 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; 197 static const char *panic_later, *panic_param; 198 199 static bool __init obsolete_checksetup(char *line) 200 { 201 const struct obs_kernel_param *p; 202 bool had_early_param = false; 203 204 p = __setup_start; 205 do { 206 int n = strlen(p->str); 207 if (parameqn(line, p->str, n)) { 208 if (p->early) { 209 /* Already done in parse_early_param? 210 * (Needs exact match on param part). 211 * Keep iterating, as we can have early 212 * params and __setups of same names 8( */ 213 if (line[n] == '\0' || line[n] == '=') 214 had_early_param = true; 215 } else if (!p->setup_func) { 216 pr_warn("Parameter %s is obsolete, ignored\n", 217 p->str); 218 return true; 219 } else if (p->setup_func(line + n)) 220 return true; 221 } 222 p++; 223 } while (p < __setup_end); 224 225 return had_early_param; 226 } 227 228 /* 229 * This should be approx 2 Bo*oMips to start (note initial shift), and will 230 * still work even if initially too large, it will just take slightly longer 231 */ 232 unsigned long loops_per_jiffy = (1<<12); 233 EXPORT_SYMBOL(loops_per_jiffy); 234 235 static int __init debug_kernel(char *str) 236 { 237 console_loglevel = CONSOLE_LOGLEVEL_DEBUG; 238 return 0; 239 } 240 241 static int __init quiet_kernel(char *str) 242 { 243 console_loglevel = CONSOLE_LOGLEVEL_QUIET; 244 return 0; 245 } 246 247 early_param("debug", debug_kernel); 248 early_param("quiet", quiet_kernel); 249 250 static int __init loglevel(char *str) 251 { 252 int newlevel; 253 254 /* 255 * Only update loglevel value when a correct setting was passed, 256 * to prevent blind crashes (when loglevel being set to 0) that 257 * are quite hard to debug 258 */ 259 if (get_option(&str, &newlevel)) { 260 console_loglevel = newlevel; 261 return 0; 262 } 263 264 return -EINVAL; 265 } 266 267 early_param("loglevel", loglevel); 268 269 #ifdef CONFIG_BLK_DEV_INITRD 270 static void * __init get_boot_config_from_initrd(size_t *_size) 271 { 272 u32 size, csum; 273 char *data; 274 u8 *hdr; 275 int i; 276 277 if (!initrd_end) 278 return NULL; 279 280 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN; 281 /* 282 * Since Grub may align the size of initrd to 4, we must 283 * check the preceding 3 bytes as well. 284 */ 285 for (i = 0; i < 4; i++) { 286 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN)) 287 goto found; 288 data--; 289 } 290 return NULL; 291 292 found: 293 hdr = (u8 *)(data - 8); 294 size = get_unaligned_le32(hdr); 295 csum = get_unaligned_le32(hdr + 4); 296 297 data = ((void *)hdr) - size; 298 if ((unsigned long)data < initrd_start) { 299 pr_err("bootconfig size %d is greater than initrd size %ld\n", 300 size, initrd_end - initrd_start); 301 return NULL; 302 } 303 304 if (xbc_calc_checksum(data, size) != csum) { 305 pr_err("bootconfig checksum failed\n"); 306 return NULL; 307 } 308 309 /* Remove bootconfig from initramfs/initrd */ 310 initrd_end = (unsigned long)data; 311 if (_size) 312 *_size = size; 313 314 return data; 315 } 316 #else 317 static void * __init get_boot_config_from_initrd(size_t *_size) 318 { 319 return NULL; 320 } 321 #endif 322 323 #ifdef CONFIG_BOOT_CONFIG 324 325 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata; 326 327 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0) 328 329 static int __init xbc_snprint_cmdline(char *buf, size_t size, 330 struct xbc_node *root) 331 { 332 struct xbc_node *knode, *vnode; 333 char *end = buf + size; 334 const char *val, *q; 335 int ret; 336 337 xbc_node_for_each_key_value(root, knode, val) { 338 ret = xbc_node_compose_key_after(root, knode, 339 xbc_namebuf, XBC_KEYLEN_MAX); 340 if (ret < 0) 341 return ret; 342 343 vnode = xbc_node_get_child(knode); 344 if (!vnode) { 345 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf); 346 if (ret < 0) 347 return ret; 348 buf += ret; 349 continue; 350 } 351 xbc_array_for_each_value(vnode, val) { 352 /* 353 * For prettier and more readable /proc/cmdline, only 354 * quote the value when necessary, i.e. when it contains 355 * whitespace. 356 */ 357 q = strpbrk(val, " \t\r\n") ? "\"" : ""; 358 ret = snprintf(buf, rest(buf, end), "%s=%s%s%s ", 359 xbc_namebuf, q, val, q); 360 if (ret < 0) 361 return ret; 362 buf += ret; 363 } 364 } 365 366 return buf - (end - size); 367 } 368 #undef rest 369 370 /* Make an extra command line under given key word */ 371 static char * __init xbc_make_cmdline(const char *key) 372 { 373 struct xbc_node *root; 374 char *new_cmdline; 375 int ret, len = 0; 376 377 root = xbc_find_node(key); 378 if (!root) 379 return NULL; 380 381 /* Count required buffer size */ 382 len = xbc_snprint_cmdline(NULL, 0, root); 383 if (len <= 0) 384 return NULL; 385 386 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES); 387 if (!new_cmdline) { 388 pr_err("Failed to allocate memory for extra kernel cmdline.\n"); 389 return NULL; 390 } 391 392 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root); 393 if (ret < 0 || ret > len) { 394 pr_err("Failed to print extra kernel cmdline.\n"); 395 memblock_free(new_cmdline, len + 1); 396 return NULL; 397 } 398 399 return new_cmdline; 400 } 401 402 static int __init bootconfig_params(char *param, char *val, 403 const char *unused, void *arg) 404 { 405 if (strcmp(param, "bootconfig") == 0) { 406 bootconfig_found = true; 407 } 408 return 0; 409 } 410 411 static int __init warn_bootconfig(char *str) 412 { 413 /* The 'bootconfig' has been handled by bootconfig_params(). */ 414 return 0; 415 } 416 417 static void __init setup_boot_config(void) 418 { 419 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 420 const char *msg, *data; 421 int pos, ret; 422 size_t size; 423 char *err; 424 425 /* Cut out the bootconfig data even if we have no bootconfig option */ 426 data = get_boot_config_from_initrd(&size); 427 /* If there is no bootconfig in initrd, try embedded one. */ 428 if (!data) 429 data = xbc_get_embedded_bootconfig(&size); 430 431 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 432 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL, 433 bootconfig_params); 434 435 if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE))) 436 return; 437 438 /* parse_args() stops at the next param of '--' and returns an address */ 439 if (err) 440 initargs_offs = err - tmp_cmdline; 441 442 if (!data) { 443 /* If user intended to use bootconfig, show an error level message */ 444 if (bootconfig_found) 445 pr_err("'bootconfig' found on command line, but no bootconfig found\n"); 446 else 447 pr_info("No bootconfig data provided, so skipping bootconfig"); 448 return; 449 } 450 451 if (size >= XBC_DATA_MAX) { 452 pr_err("bootconfig size %ld greater than max size %d\n", 453 (long)size, XBC_DATA_MAX); 454 return; 455 } 456 457 ret = xbc_init(data, size, &msg, &pos); 458 if (ret < 0) { 459 if (pos < 0) 460 pr_err("Failed to init bootconfig: %s.\n", msg); 461 else 462 pr_err("Failed to parse bootconfig: %s at %d.\n", 463 msg, pos); 464 } else { 465 xbc_get_info(&ret, NULL); 466 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret); 467 /* keys starting with "kernel." are passed via cmdline */ 468 extra_command_line = xbc_make_cmdline("kernel"); 469 /* Also, "init." keys are init arguments */ 470 extra_init_args = xbc_make_cmdline("init"); 471 } 472 return; 473 } 474 475 static void __init exit_boot_config(void) 476 { 477 xbc_exit(); 478 } 479 480 #else /* !CONFIG_BOOT_CONFIG */ 481 482 static void __init setup_boot_config(void) 483 { 484 /* Remove bootconfig data from initrd */ 485 get_boot_config_from_initrd(NULL); 486 } 487 488 static int __init warn_bootconfig(char *str) 489 { 490 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n"); 491 return 0; 492 } 493 494 #define exit_boot_config() do {} while (0) 495 496 #endif /* CONFIG_BOOT_CONFIG */ 497 498 early_param("bootconfig", warn_bootconfig); 499 500 bool __init cmdline_has_extra_options(void) 501 { 502 return extra_command_line || extra_init_args; 503 } 504 505 /* Change NUL term back to "=", to make "param" the whole string. */ 506 static void __init repair_env_string(char *param, char *val) 507 { 508 if (val) { 509 /* param=val or param="val"? */ 510 if (val == param+strlen(param)+1) 511 val[-1] = '='; 512 else if (val == param+strlen(param)+2) { 513 val[-2] = '='; 514 memmove(val-1, val, strlen(val)+1); 515 } else 516 BUG(); 517 } 518 } 519 520 /* Anything after -- gets handed straight to init. */ 521 static int __init set_init_arg(char *param, char *val, 522 const char *unused, void *arg) 523 { 524 unsigned int i; 525 526 if (panic_later) 527 return 0; 528 529 repair_env_string(param, val); 530 531 for (i = 0; argv_init[i]; i++) { 532 if (i == MAX_INIT_ARGS) { 533 panic_later = "init"; 534 panic_param = param; 535 return 0; 536 } 537 } 538 argv_init[i] = param; 539 return 0; 540 } 541 542 /* 543 * Unknown boot options get handed to init, unless they look like 544 * unused parameters (modprobe will find them in /proc/cmdline). 545 */ 546 static int __init unknown_bootoption(char *param, char *val, 547 const char *unused, void *arg) 548 { 549 size_t len = strlen(param); 550 /* 551 * Well-known bootloader identifiers: 552 * 1. LILO/Grub pass "BOOT_IMAGE=..."; 553 * 2. kexec/kdump (kexec-tools) pass "kexec". 554 */ 555 const char *bootloader[] = { "BOOT_IMAGE=", "kexec", NULL }; 556 557 /* Handle params aliased to sysctls */ 558 if (sysctl_is_alias(param)) 559 return 0; 560 561 repair_env_string(param, val); 562 563 /* Handle bootloader identifier */ 564 for (int i = 0; bootloader[i]; i++) { 565 if (strstarts(param, bootloader[i])) 566 return 0; 567 } 568 569 /* Handle obsolete-style parameters */ 570 if (obsolete_checksetup(param)) 571 return 0; 572 573 /* Unused module parameter. */ 574 if (strnchr(param, len, '.')) 575 return 0; 576 577 if (panic_later) 578 return 0; 579 580 if (val) { 581 /* Environment option */ 582 unsigned int i; 583 for (i = 0; envp_init[i]; i++) { 584 if (i == MAX_INIT_ENVS) { 585 panic_later = "env"; 586 panic_param = param; 587 } 588 if (!strncmp(param, envp_init[i], len+1)) 589 break; 590 } 591 envp_init[i] = param; 592 } else { 593 /* Command line option */ 594 unsigned int i; 595 for (i = 0; argv_init[i]; i++) { 596 if (i == MAX_INIT_ARGS) { 597 panic_later = "init"; 598 panic_param = param; 599 } 600 } 601 argv_init[i] = param; 602 } 603 return 0; 604 } 605 606 static int __init init_setup(char *str) 607 { 608 unsigned int i; 609 610 execute_command = str; 611 /* 612 * In case LILO is going to boot us with default command line, 613 * it prepends "auto" before the whole cmdline which makes 614 * the shell think it should execute a script with such name. 615 * So we ignore all arguments entered _before_ init=... [MJ] 616 */ 617 for (i = 1; i < MAX_INIT_ARGS; i++) 618 argv_init[i] = NULL; 619 return 1; 620 } 621 __setup("init=", init_setup); 622 623 static int __init rdinit_setup(char *str) 624 { 625 unsigned int i; 626 627 ramdisk_execute_command = str; 628 ramdisk_execute_command_set = true; 629 /* See "auto" comment in init_setup */ 630 for (i = 1; i < MAX_INIT_ARGS; i++) 631 argv_init[i] = NULL; 632 return 1; 633 } 634 __setup("rdinit=", rdinit_setup); 635 636 #ifndef CONFIG_SMP 637 static inline void setup_nr_cpu_ids(void) { } 638 static inline void smp_prepare_cpus(unsigned int maxcpus) { } 639 #endif 640 641 /* 642 * We need to store the untouched command line for future reference. 643 * We also need to store the touched command line since the parameter 644 * parsing is performed in place, and we should allow a component to 645 * store reference of name/value for future reference. 646 */ 647 static void __init setup_command_line(char *command_line) 648 { 649 size_t len, xlen = 0, ilen = 0; 650 651 if (extra_command_line) 652 xlen = strlen(extra_command_line); 653 if (extra_init_args) { 654 extra_init_args = strim(extra_init_args); /* remove trailing space */ 655 ilen = strlen(extra_init_args) + 4; /* for " -- " */ 656 } 657 658 len = xlen + strlen(boot_command_line) + ilen + 1; 659 660 saved_command_line = memblock_alloc_or_panic(len, SMP_CACHE_BYTES); 661 662 len = xlen + strlen(command_line) + 1; 663 664 static_command_line = memblock_alloc_or_panic(len, SMP_CACHE_BYTES); 665 666 if (xlen) { 667 /* 668 * We have to put extra_command_line before boot command 669 * lines because there could be dashes (separator of init 670 * command line) in the command lines. 671 */ 672 strcpy(saved_command_line, extra_command_line); 673 strcpy(static_command_line, extra_command_line); 674 } 675 strcpy(saved_command_line + xlen, boot_command_line); 676 strcpy(static_command_line + xlen, command_line); 677 678 if (ilen) { 679 /* 680 * Append supplemental init boot args to saved_command_line 681 * so that user can check what command line options passed 682 * to init. 683 * The order should always be 684 * " -- "[bootconfig init-param][cmdline init-param] 685 */ 686 if (initargs_offs) { 687 len = xlen + initargs_offs; 688 strcpy(saved_command_line + len, extra_init_args); 689 len += ilen - 4; /* strlen(extra_init_args) */ 690 strcpy(saved_command_line + len, 691 boot_command_line + initargs_offs - 1); 692 } else { 693 len = strlen(saved_command_line); 694 strcpy(saved_command_line + len, " -- "); 695 len += 4; 696 strcpy(saved_command_line + len, extra_init_args); 697 } 698 } 699 700 saved_command_line_len = strlen(saved_command_line); 701 } 702 703 /* 704 * We need to finalize in a non-__init function or else race conditions 705 * between the root thread and the init thread may cause start_kernel to 706 * be reaped by free_initmem before the root thread has proceeded to 707 * cpu_idle. 708 * 709 * gcc-3.4 accidentally inlines this function, so use noinline. 710 */ 711 712 static __initdata DECLARE_COMPLETION(kthreadd_done); 713 714 static noinline void __ref __noreturn rest_init(void) 715 { 716 struct task_struct *tsk; 717 int pid; 718 719 rcu_scheduler_starting(); 720 /* 721 * We need to spawn init first so that it obtains pid 1, however 722 * the init task will end up wanting to create kthreads, which, if 723 * we schedule it before we create kthreadd, will OOPS. 724 */ 725 pid = user_mode_thread(kernel_init, NULL, CLONE_FS); 726 /* 727 * Pin init on the boot CPU. Task migration is not properly working 728 * until sched_init_smp() has been run. It will set the allowed 729 * CPUs for init to the non isolated CPUs. 730 */ 731 rcu_read_lock(); 732 tsk = find_task_by_pid_ns(pid, &init_pid_ns); 733 tsk->flags |= PF_NO_SETAFFINITY; 734 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id())); 735 rcu_read_unlock(); 736 737 numa_default_policy(); 738 pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES); 739 rcu_read_lock(); 740 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); 741 rcu_read_unlock(); 742 743 /* 744 * Enable might_sleep() and smp_processor_id() checks. 745 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y 746 * kernel_thread() would trigger might_sleep() splats. With 747 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled 748 * already, but it's stuck on the kthreadd_done completion. 749 */ 750 system_state = SYSTEM_SCHEDULING; 751 752 complete(&kthreadd_done); 753 754 /* 755 * The boot idle thread must execute schedule() 756 * at least once to get things moving: 757 */ 758 schedule_preempt_disabled(); 759 /* Call into cpu_idle with preempt disabled */ 760 cpu_startup_entry(CPUHP_ONLINE); 761 } 762 763 /* Check for early params. */ 764 static int __init do_early_param(char *param, char *val, 765 const char *unused, void *arg) 766 { 767 const struct obs_kernel_param *p; 768 769 for (p = __setup_start; p < __setup_end; p++) { 770 if (p->early && parameq(param, p->str)) { 771 if (p->setup_func(val) != 0) 772 pr_warn("Malformed early option '%s'\n", param); 773 } 774 } 775 /* We accept everything at this stage. */ 776 return 0; 777 } 778 779 void __init parse_early_options(char *cmdline) 780 { 781 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL, 782 do_early_param); 783 } 784 785 /* Arch code calls this early on, or if not, just before other parsing. */ 786 void __init parse_early_param(void) 787 { 788 static int done __initdata; 789 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 790 791 if (done) 792 return; 793 794 /* All fall through to do_early_param. */ 795 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 796 parse_early_options(tmp_cmdline); 797 done = 1; 798 } 799 800 void __init __weak arch_post_acpi_subsys_init(void) { } 801 802 void __init __weak smp_setup_processor_id(void) 803 { 804 } 805 806 void __init __weak smp_prepare_boot_cpu(void) 807 { 808 } 809 810 # if THREAD_SIZE >= PAGE_SIZE 811 void __init __weak thread_stack_cache_init(void) 812 { 813 } 814 #endif 815 816 void __init __weak poking_init(void) { } 817 818 void __init __weak pgtable_cache_init(void) { } 819 820 void __init __weak trap_init(void) { } 821 822 bool initcall_debug; 823 core_param(initcall_debug, initcall_debug, bool, 0644); 824 825 #ifdef TRACEPOINTS_ENABLED 826 static void __init initcall_debug_enable(void); 827 #else 828 static inline void initcall_debug_enable(void) 829 { 830 } 831 #endif 832 833 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET 834 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT, 835 randomize_kstack_offset); 836 DEFINE_PER_CPU(u32, kstack_offset); 837 838 static int __init early_randomize_kstack_offset(char *buf) 839 { 840 int ret; 841 bool bool_result; 842 843 ret = kstrtobool(buf, &bool_result); 844 if (ret) 845 return ret; 846 847 if (bool_result) 848 static_branch_enable(&randomize_kstack_offset); 849 else 850 static_branch_disable(&randomize_kstack_offset); 851 return 0; 852 } 853 early_param("randomize_kstack_offset", early_randomize_kstack_offset); 854 #endif 855 856 static void __init print_unknown_bootoptions(void) 857 { 858 char *unknown_options; 859 char *end; 860 const char *const *p; 861 size_t len; 862 863 if (panic_later || (!argv_init[1] && !envp_init[2])) 864 return; 865 866 /* 867 * Determine how many options we have to print out, plus a space 868 * before each 869 */ 870 len = 1; /* null terminator */ 871 for (p = &argv_init[1]; *p; p++) { 872 len++; 873 len += strlen(*p); 874 } 875 for (p = &envp_init[2]; *p; p++) { 876 len++; 877 len += strlen(*p); 878 } 879 880 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES); 881 if (!unknown_options) { 882 pr_err("%s: Failed to allocate %zu bytes\n", 883 __func__, len); 884 return; 885 } 886 end = unknown_options; 887 888 for (p = &argv_init[1]; *p; p++) 889 end += sprintf(end, " %s", *p); 890 for (p = &envp_init[2]; *p; p++) 891 end += sprintf(end, " %s", *p); 892 893 /* Start at unknown_options[1] to skip the initial space */ 894 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n", 895 &unknown_options[1]); 896 memblock_free(unknown_options, len); 897 } 898 899 static void __init early_numa_node_init(void) 900 { 901 #ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID 902 #ifndef cpu_to_node 903 int cpu; 904 905 /* The early_cpu_to_node() should be ready here. */ 906 for_each_possible_cpu(cpu) 907 set_cpu_numa_node(cpu, early_cpu_to_node(cpu)); 908 #endif 909 #endif 910 } 911 912 #define KERNEL_CMDLINE_PREFIX "Kernel command line: " 913 #define KERNEL_CMDLINE_PREFIX_LEN (sizeof(KERNEL_CMDLINE_PREFIX) - 1) 914 #define KERNEL_CMDLINE_CONTINUATION " \\" 915 #define KERNEL_CMDLINE_CONTINUATION_LEN (sizeof(KERNEL_CMDLINE_CONTINUATION) - 1) 916 917 #define MIN_CMDLINE_LOG_WRAP_IDEAL_LEN (KERNEL_CMDLINE_PREFIX_LEN + \ 918 KERNEL_CMDLINE_CONTINUATION_LEN) 919 #define CMDLINE_LOG_WRAP_IDEAL_LEN (CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN > \ 920 MIN_CMDLINE_LOG_WRAP_IDEAL_LEN ? \ 921 CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN : \ 922 MIN_CMDLINE_LOG_WRAP_IDEAL_LEN) 923 924 #define IDEAL_CMDLINE_LEN (CMDLINE_LOG_WRAP_IDEAL_LEN - KERNEL_CMDLINE_PREFIX_LEN) 925 #define IDEAL_CMDLINE_SPLIT_LEN (IDEAL_CMDLINE_LEN - KERNEL_CMDLINE_CONTINUATION_LEN) 926 927 /** 928 * print_kernel_cmdline() - Print the kernel cmdline with wrapping. 929 * @cmdline: The cmdline to print. 930 * 931 * Print the kernel command line, trying to wrap based on the Kconfig knob 932 * CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN. 933 * 934 * Wrapping is based on spaces, ignoring quotes. All lines are prefixed 935 * with "Kernel command line: " and lines that are not the last line have 936 * a " \" suffix added to them. The prefix and suffix count towards the 937 * line length for wrapping purposes. The ideal length will be exceeded 938 * if no appropriate place to wrap is found. 939 * 940 * Example output if CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN is 40: 941 * Kernel command line: loglevel=7 \ 942 * Kernel command line: init=/sbin/init \ 943 * Kernel command line: root=PARTUUID=8c3efc1a-768b-6642-8d0c-89eb782f19f0/PARTNROFF=1 \ 944 * Kernel command line: rootwait ro \ 945 * Kernel command line: my_quoted_arg="The \ 946 * Kernel command line: quick brown fox \ 947 * Kernel command line: jumps over the \ 948 * Kernel command line: lazy dog." 949 */ 950 static void __init print_kernel_cmdline(const char *cmdline) 951 { 952 size_t len; 953 954 /* Config option of 0 or anything longer than the max disables wrapping */ 955 if (CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN == 0 || 956 IDEAL_CMDLINE_LEN >= COMMAND_LINE_SIZE - 1) { 957 pr_notice("%s%s\n", KERNEL_CMDLINE_PREFIX, cmdline); 958 return; 959 } 960 961 len = strlen(cmdline); 962 while (len > IDEAL_CMDLINE_LEN) { 963 const char *first_space; 964 const char *prev_cutoff; 965 const char *cutoff; 966 int to_print; 967 size_t used; 968 969 /* Find the last ' ' that wouldn't make the line too long */ 970 prev_cutoff = NULL; 971 cutoff = cmdline; 972 while (true) { 973 cutoff = strchr(cutoff + 1, ' '); 974 if (!cutoff || cutoff - cmdline > IDEAL_CMDLINE_SPLIT_LEN) 975 break; 976 prev_cutoff = cutoff; 977 } 978 if (prev_cutoff) 979 cutoff = prev_cutoff; 980 else if (!cutoff) 981 break; 982 983 /* Find the beginning and end of the string of spaces */ 984 first_space = cutoff; 985 while (first_space > cmdline && first_space[-1] == ' ') 986 first_space--; 987 to_print = first_space - cmdline; 988 while (*cutoff == ' ') 989 cutoff++; 990 used = cutoff - cmdline; 991 992 /* If the whole string is used, break and do the final printout */ 993 if (len == used) 994 break; 995 996 if (to_print) 997 pr_notice("%s%.*s%s\n", KERNEL_CMDLINE_PREFIX, 998 to_print, cmdline, KERNEL_CMDLINE_CONTINUATION); 999 1000 len -= used; 1001 cmdline += used; 1002 } 1003 if (len) 1004 pr_notice("%s%s\n", KERNEL_CMDLINE_PREFIX, cmdline); 1005 } 1006 1007 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector 1008 void start_kernel(void) 1009 { 1010 char *command_line; 1011 char *after_dashes; 1012 1013 set_task_stack_end_magic(&init_task); 1014 smp_setup_processor_id(); 1015 debug_objects_early_init(); 1016 init_vmlinux_build_id(); 1017 1018 cgroup_init_early(); 1019 1020 local_irq_disable(); 1021 early_boot_irqs_disabled = true; 1022 1023 /* 1024 * Interrupts are still disabled. Do necessary setups, then 1025 * enable them. 1026 */ 1027 boot_cpu_init(); 1028 page_address_init(); 1029 pr_notice("%s", linux_banner); 1030 setup_arch(&command_line); 1031 /* Static keys and static calls are needed by LSMs */ 1032 jump_label_init(); 1033 static_call_init(); 1034 early_security_init(); 1035 setup_boot_config(); 1036 setup_command_line(command_line); 1037 setup_nr_cpu_ids(); 1038 setup_per_cpu_areas(); 1039 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ 1040 early_numa_node_init(); 1041 boot_cpu_hotplug_init(); 1042 1043 print_kernel_cmdline(saved_command_line); 1044 /* parameters may set static keys */ 1045 parse_early_param(); 1046 after_dashes = parse_args("Booting kernel", 1047 static_command_line, __start___param, 1048 __stop___param - __start___param, 1049 -1, -1, NULL, &unknown_bootoption); 1050 print_unknown_bootoptions(); 1051 if (!IS_ERR_OR_NULL(after_dashes)) 1052 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1, 1053 NULL, set_init_arg); 1054 if (extra_init_args) 1055 parse_args("Setting extra init args", extra_init_args, 1056 NULL, 0, -1, -1, NULL, set_init_arg); 1057 1058 /* Architectural and non-timekeeping rng init, before allocator init */ 1059 random_init_early(command_line); 1060 1061 /* 1062 * These use large bootmem allocations and must precede 1063 * initalization of page allocator 1064 */ 1065 setup_log_buf(0); 1066 vfs_caches_init_early(); 1067 sort_main_extable(); 1068 trap_init(); 1069 mm_core_init(); 1070 maple_tree_init(); 1071 poking_init(); 1072 ftrace_init(); 1073 1074 /* trace_printk can be enabled here */ 1075 early_trace_init(); 1076 1077 /* 1078 * Set up the scheduler prior starting any interrupts (such as the 1079 * timer interrupt). Full topology setup happens at smp_init() 1080 * time - but meanwhile we still have a functioning scheduler. 1081 */ 1082 sched_init(); 1083 1084 if (WARN(!irqs_disabled(), 1085 "Interrupts were enabled *very* early, fixing it\n")) 1086 local_irq_disable(); 1087 radix_tree_init(); 1088 1089 /* 1090 * Set up housekeeping before setting up workqueues to allow the unbound 1091 * workqueue to take non-housekeeping into account. 1092 */ 1093 housekeeping_init(); 1094 1095 /* 1096 * Allow workqueue creation and work item queueing/cancelling 1097 * early. Work item execution depends on kthreads and starts after 1098 * workqueue_init(). 1099 */ 1100 workqueue_init_early(); 1101 1102 rcu_init(); 1103 kvfree_rcu_init(); 1104 1105 /* Trace events are available after this */ 1106 trace_init(); 1107 1108 if (initcall_debug) 1109 initcall_debug_enable(); 1110 1111 context_tracking_init(); 1112 /* init some links before init_ISA_irqs() */ 1113 early_irq_init(); 1114 init_IRQ(); 1115 tick_init(); 1116 rcu_init_nohz(); 1117 timers_init(); 1118 srcu_init(); 1119 hrtimers_init(); 1120 softirq_init(); 1121 timekeeping_init(); 1122 time_init(); 1123 1124 /* This must be after timekeeping is initialized */ 1125 random_init(); 1126 1127 /* These make use of the fully initialized rng */ 1128 kfence_init(); 1129 boot_init_stack_canary(); 1130 1131 perf_event_init(); 1132 profile_init(); 1133 call_function_init(); 1134 WARN(!irqs_disabled(), "Interrupts were enabled early\n"); 1135 1136 early_boot_irqs_disabled = false; 1137 local_irq_enable(); 1138 1139 kmem_cache_init_late(); 1140 1141 /* 1142 * HACK ALERT! This is early. We're enabling the console before 1143 * we've done PCI setups etc, and console_init() must be aware of 1144 * this. But we do want output early, in case something goes wrong. 1145 */ 1146 console_init(); 1147 if (panic_later) 1148 panic("Too many boot %s vars at `%s'", panic_later, 1149 panic_param); 1150 1151 lockdep_init(); 1152 1153 /* 1154 * Need to run this when irqs are enabled, because it wants 1155 * to self-test [hard/soft]-irqs on/off lock inversion bugs 1156 * too: 1157 */ 1158 locking_selftest(); 1159 1160 #ifdef CONFIG_BLK_DEV_INITRD 1161 if (initrd_start && !initrd_below_start_ok && 1162 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { 1163 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", 1164 page_to_pfn(virt_to_page((void *)initrd_start)), 1165 min_low_pfn); 1166 initrd_start = 0; 1167 } 1168 #endif 1169 setup_per_cpu_pageset(); 1170 numa_policy_init(); 1171 acpi_early_init(); 1172 if (late_time_init) 1173 late_time_init(); 1174 sched_clock_init(); 1175 calibrate_delay(); 1176 1177 arch_cpu_finalize_init(); 1178 1179 pid_idr_init(); 1180 anon_vma_init(); 1181 thread_stack_cache_init(); 1182 cred_init(); 1183 fork_init(); 1184 proc_caches_init(); 1185 uts_ns_init(); 1186 time_ns_init(); 1187 key_init(); 1188 security_init(); 1189 dbg_late_init(); 1190 net_ns_init(); 1191 vfs_caches_init(); 1192 pagecache_init(); 1193 signals_init(); 1194 seq_file_init(); 1195 proc_root_init(); 1196 nsfs_init(); 1197 pidfs_init(); 1198 cpuset_init(); 1199 mem_cgroup_init(); 1200 cgroup_init(); 1201 taskstats_init_early(); 1202 delayacct_init(); 1203 1204 acpi_subsystem_init(); 1205 arch_post_acpi_subsys_init(); 1206 kcsan_init(); 1207 1208 /* Do the rest non-__init'ed, we're now alive */ 1209 rest_init(); 1210 1211 /* 1212 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10 1213 * and older. 1214 */ 1215 #if !__has_attribute(__no_stack_protector__) 1216 prevent_tail_call_optimization(); 1217 #endif 1218 } 1219 1220 /* Call all constructor functions linked into the kernel. */ 1221 static void __init do_ctors(void) 1222 { 1223 /* 1224 * For UML, the constructors have already been called by the 1225 * normal setup code as it's just a normal ELF binary, so we 1226 * cannot do it again - but we do need CONFIG_CONSTRUCTORS 1227 * even on UML for modules. 1228 */ 1229 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML) 1230 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; 1231 1232 for (; fn < (ctor_fn_t *) __ctors_end; fn++) 1233 (*fn)(); 1234 #endif 1235 } 1236 1237 #ifdef CONFIG_KALLSYMS 1238 struct blacklist_entry { 1239 struct list_head next; 1240 char *buf; 1241 }; 1242 1243 static __initdata_or_module LIST_HEAD(blacklisted_initcalls); 1244 1245 static int __init initcall_blacklist(char *str) 1246 { 1247 char *str_entry; 1248 struct blacklist_entry *entry; 1249 1250 /* str argument is a comma-separated list of functions */ 1251 do { 1252 str_entry = strsep(&str, ","); 1253 if (str_entry) { 1254 pr_debug("blacklisting initcall %s\n", str_entry); 1255 entry = memblock_alloc_or_panic(sizeof(*entry), 1256 SMP_CACHE_BYTES); 1257 entry->buf = memblock_alloc_or_panic(strlen(str_entry) + 1, 1258 SMP_CACHE_BYTES); 1259 strcpy(entry->buf, str_entry); 1260 list_add(&entry->next, &blacklisted_initcalls); 1261 } 1262 } while (str_entry); 1263 1264 return 1; 1265 } 1266 1267 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1268 { 1269 struct blacklist_entry *entry; 1270 char fn_name[KSYM_SYMBOL_LEN]; 1271 unsigned long addr; 1272 1273 if (list_empty(&blacklisted_initcalls)) 1274 return false; 1275 1276 addr = (unsigned long) dereference_function_descriptor(fn); 1277 sprint_symbol_no_offset(fn_name, addr); 1278 1279 /* 1280 * fn will be "function_name [module_name]" where [module_name] is not 1281 * displayed for built-in init functions. Strip off the [module_name]. 1282 */ 1283 strreplace(fn_name, ' ', '\0'); 1284 1285 list_for_each_entry(entry, &blacklisted_initcalls, next) { 1286 if (!strcmp(fn_name, entry->buf)) { 1287 pr_debug("initcall %s blacklisted\n", fn_name); 1288 return true; 1289 } 1290 } 1291 1292 return false; 1293 } 1294 #else 1295 static int __init initcall_blacklist(char *str) 1296 { 1297 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); 1298 return 0; 1299 } 1300 1301 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1302 { 1303 return false; 1304 } 1305 #endif 1306 __setup("initcall_blacklist=", initcall_blacklist); 1307 1308 static __init_or_module void 1309 trace_initcall_start_cb(void *data, initcall_t fn) 1310 { 1311 ktime_t *calltime = data; 1312 1313 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current)); 1314 *calltime = ktime_get(); 1315 } 1316 1317 static __init_or_module void 1318 trace_initcall_finish_cb(void *data, initcall_t fn, int ret) 1319 { 1320 ktime_t rettime, *calltime = data; 1321 1322 rettime = ktime_get(); 1323 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n", 1324 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime)); 1325 } 1326 1327 static __init_or_module void 1328 trace_initcall_level_cb(void *data, const char *level) 1329 { 1330 printk(KERN_DEBUG "entering initcall level: %s\n", level); 1331 } 1332 1333 static ktime_t initcall_calltime; 1334 1335 #ifdef TRACEPOINTS_ENABLED 1336 static void __init initcall_debug_enable(void) 1337 { 1338 int ret; 1339 1340 ret = register_trace_initcall_start(trace_initcall_start_cb, 1341 &initcall_calltime); 1342 ret |= register_trace_initcall_finish(trace_initcall_finish_cb, 1343 &initcall_calltime); 1344 ret |= register_trace_initcall_level(trace_initcall_level_cb, NULL); 1345 WARN(ret, "Failed to register initcall tracepoints\n"); 1346 } 1347 # define do_trace_initcall_start trace_initcall_start 1348 # define do_trace_initcall_finish trace_initcall_finish 1349 # define do_trace_initcall_level trace_initcall_level 1350 #else 1351 static inline void do_trace_initcall_start(initcall_t fn) 1352 { 1353 if (!initcall_debug) 1354 return; 1355 trace_initcall_start_cb(&initcall_calltime, fn); 1356 } 1357 static inline void do_trace_initcall_finish(initcall_t fn, int ret) 1358 { 1359 if (!initcall_debug) 1360 return; 1361 trace_initcall_finish_cb(&initcall_calltime, fn, ret); 1362 } 1363 static inline void do_trace_initcall_level(const char *level) 1364 { 1365 if (!initcall_debug) 1366 return; 1367 trace_initcall_level_cb(NULL, level); 1368 } 1369 #endif /* !TRACEPOINTS_ENABLED */ 1370 1371 int __init_or_module do_one_initcall(initcall_t fn) 1372 { 1373 int count = preempt_count(); 1374 char msgbuf[64]; 1375 int ret; 1376 1377 if (initcall_blacklisted(fn)) 1378 return -EPERM; 1379 1380 do_trace_initcall_start(fn); 1381 ret = fn(); 1382 do_trace_initcall_finish(fn, ret); 1383 1384 msgbuf[0] = 0; 1385 1386 if (preempt_count() != count) { 1387 sprintf(msgbuf, "preemption imbalance "); 1388 preempt_count_set(count); 1389 } 1390 if (irqs_disabled()) { 1391 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); 1392 local_irq_enable(); 1393 } 1394 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf); 1395 1396 add_latent_entropy(); 1397 return ret; 1398 } 1399 1400 1401 static initcall_entry_t *initcall_levels[] __initdata = { 1402 __initcall0_start, 1403 __initcall1_start, 1404 __initcall2_start, 1405 __initcall3_start, 1406 __initcall4_start, 1407 __initcall5_start, 1408 __initcall6_start, 1409 __initcall7_start, 1410 __initcall_end, 1411 }; 1412 1413 /* Keep these in sync with initcalls in include/linux/init.h */ 1414 static const char *initcall_level_names[] __initdata = { 1415 "pure", 1416 "core", 1417 "postcore", 1418 "arch", 1419 "subsys", 1420 "fs", 1421 "device", 1422 "late", 1423 }; 1424 1425 static int __init ignore_unknown_bootoption(char *param, char *val, 1426 const char *unused, void *arg) 1427 { 1428 return 0; 1429 } 1430 1431 static void __init do_initcall_level(int level, char *command_line) 1432 { 1433 initcall_entry_t *fn; 1434 1435 parse_args(initcall_level_names[level], 1436 command_line, __start___param, 1437 __stop___param - __start___param, 1438 level, level, 1439 NULL, ignore_unknown_bootoption); 1440 1441 do_trace_initcall_level(initcall_level_names[level]); 1442 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) 1443 do_one_initcall(initcall_from_entry(fn)); 1444 } 1445 1446 static void __init do_initcalls(void) 1447 { 1448 int level; 1449 size_t len = saved_command_line_len + 1; 1450 char *command_line; 1451 1452 command_line = kzalloc(len, GFP_KERNEL); 1453 if (!command_line) 1454 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 1455 1456 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) { 1457 /* Parser modifies command_line, restore it each time */ 1458 strcpy(command_line, saved_command_line); 1459 do_initcall_level(level, command_line); 1460 } 1461 1462 kfree(command_line); 1463 } 1464 1465 /* 1466 * Ok, the machine is now initialized. None of the devices 1467 * have been touched yet, but the CPU subsystem is up and 1468 * running, and memory and process management works. 1469 * 1470 * Now we can finally start doing some real work.. 1471 */ 1472 static void __init do_basic_setup(void) 1473 { 1474 cpuset_init_smp(); 1475 driver_init(); 1476 init_irq_proc(); 1477 do_ctors(); 1478 do_initcalls(); 1479 } 1480 1481 static void __init do_pre_smp_initcalls(void) 1482 { 1483 initcall_entry_t *fn; 1484 1485 do_trace_initcall_level("early"); 1486 for (fn = __initcall_start; fn < __initcall0_start; fn++) 1487 do_one_initcall(initcall_from_entry(fn)); 1488 } 1489 1490 static int run_init_process(const char *init_filename) 1491 { 1492 const char *const *p; 1493 1494 argv_init[0] = init_filename; 1495 pr_info("Run %s as init process\n", init_filename); 1496 pr_debug(" with arguments:\n"); 1497 for (p = argv_init; *p; p++) 1498 pr_debug(" %s\n", *p); 1499 pr_debug(" with environment:\n"); 1500 for (p = envp_init; *p; p++) 1501 pr_debug(" %s\n", *p); 1502 return kernel_execve(init_filename, argv_init, envp_init); 1503 } 1504 1505 static int try_to_run_init_process(const char *init_filename) 1506 { 1507 int ret; 1508 1509 ret = run_init_process(init_filename); 1510 1511 if (ret && ret != -ENOENT) { 1512 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", 1513 init_filename, ret); 1514 } 1515 1516 return ret; 1517 } 1518 1519 static noinline void __init kernel_init_freeable(void); 1520 1521 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX) 1522 bool rodata_enabled __ro_after_init = true; 1523 1524 #ifndef arch_parse_debug_rodata 1525 static inline bool arch_parse_debug_rodata(char *str) { return false; } 1526 #endif 1527 1528 static int __init set_debug_rodata(char *str) 1529 { 1530 if (arch_parse_debug_rodata(str)) 1531 return 0; 1532 1533 if (str && !strcmp(str, "on")) 1534 rodata_enabled = true; 1535 else if (str && !strcmp(str, "off")) 1536 rodata_enabled = false; 1537 else 1538 pr_warn("Invalid option string for rodata: '%s'\n", str); 1539 return 0; 1540 } 1541 early_param("rodata", set_debug_rodata); 1542 #endif 1543 1544 static void mark_readonly(void) 1545 { 1546 if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX) && rodata_enabled) { 1547 /* 1548 * load_module() results in W+X mappings, which are cleaned 1549 * up with init_free_wq. Let's make sure that queued work is 1550 * flushed so that we don't hit false positives looking for 1551 * insecure pages which are W+X. 1552 */ 1553 flush_module_init_free_work(); 1554 jump_label_init_ro(); 1555 mark_rodata_ro(); 1556 debug_checkwx(); 1557 rodata_test(); 1558 } else if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) { 1559 pr_info("Kernel memory protection disabled.\n"); 1560 } else if (IS_ENABLED(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)) { 1561 pr_warn("Kernel memory protection not selected by kernel config.\n"); 1562 } else { 1563 pr_warn("This architecture does not have kernel memory protection.\n"); 1564 } 1565 } 1566 1567 void __weak free_initmem(void) 1568 { 1569 free_initmem_default(POISON_FREE_INITMEM); 1570 } 1571 1572 static int __ref kernel_init(void *unused) 1573 { 1574 int ret; 1575 1576 /* 1577 * Wait until kthreadd is all set-up. 1578 */ 1579 wait_for_completion(&kthreadd_done); 1580 1581 kernel_init_freeable(); 1582 /* need to finish all async __init code before freeing the memory */ 1583 async_synchronize_full(); 1584 1585 system_state = SYSTEM_FREEING_INITMEM; 1586 kprobe_free_init_mem(); 1587 ftrace_free_init_mem(); 1588 kgdb_free_init_mem(); 1589 exit_boot_config(); 1590 free_initmem(); 1591 mark_readonly(); 1592 1593 /* 1594 * Kernel mappings are now finalized - update the userspace page-table 1595 * to finalize PTI. 1596 */ 1597 pti_finalize(); 1598 1599 system_state = SYSTEM_RUNNING; 1600 numa_default_policy(); 1601 1602 rcu_end_inkernel_boot(); 1603 1604 do_sysctl_args(); 1605 1606 if (ramdisk_execute_command) { 1607 ret = run_init_process(ramdisk_execute_command); 1608 if (!ret) 1609 return 0; 1610 pr_err("Failed to execute %s (error %d)\n", 1611 ramdisk_execute_command, ret); 1612 } 1613 1614 /* 1615 * We try each of these until one succeeds. 1616 * 1617 * The Bourne shell can be used instead of init if we are 1618 * trying to recover a really broken machine. 1619 */ 1620 if (execute_command) { 1621 ret = run_init_process(execute_command); 1622 if (!ret) 1623 return 0; 1624 panic("Requested init %s failed (error %d).", 1625 execute_command, ret); 1626 } 1627 1628 if (CONFIG_DEFAULT_INIT[0] != '\0') { 1629 ret = run_init_process(CONFIG_DEFAULT_INIT); 1630 if (ret) 1631 pr_err("Default init %s failed (error %d)\n", 1632 CONFIG_DEFAULT_INIT, ret); 1633 else 1634 return 0; 1635 } 1636 1637 if (!try_to_run_init_process("/sbin/init") || 1638 !try_to_run_init_process("/etc/init") || 1639 !try_to_run_init_process("/bin/init") || 1640 !try_to_run_init_process("/bin/sh")) 1641 return 0; 1642 1643 panic("No working init found. Try passing init= option to kernel. " 1644 "See Linux Documentation/admin-guide/init.rst for guidance."); 1645 } 1646 1647 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */ 1648 void __init console_on_rootfs(void) 1649 { 1650 struct file *file = filp_open("/dev/console", O_RDWR, 0); 1651 1652 if (IS_ERR(file)) { 1653 pr_err("Warning: unable to open an initial console.\n"); 1654 return; 1655 } 1656 init_dup(file); 1657 init_dup(file); 1658 init_dup(file); 1659 fput(file); 1660 } 1661 1662 static noinline void __init kernel_init_freeable(void) 1663 { 1664 /* Now the scheduler is fully set up and can do blocking allocations */ 1665 gfp_allowed_mask = __GFP_BITS_MASK; 1666 1667 /* 1668 * init can allocate pages on any node 1669 */ 1670 set_mems_allowed(node_states[N_MEMORY]); 1671 1672 cad_pid = get_pid(task_pid(current)); 1673 1674 smp_prepare_cpus(setup_max_cpus); 1675 1676 workqueue_init(); 1677 1678 init_mm_internals(); 1679 1680 do_pre_smp_initcalls(); 1681 lockup_detector_init(); 1682 1683 smp_init(); 1684 sched_init_smp(); 1685 1686 workqueue_init_topology(); 1687 async_init(); 1688 padata_init(); 1689 page_alloc_init_late(); 1690 1691 do_basic_setup(); 1692 1693 kunit_run_all_tests(); 1694 1695 wait_for_initramfs(); 1696 console_on_rootfs(); 1697 1698 /* 1699 * check if there is an early userspace init. If yes, let it do all 1700 * the work 1701 */ 1702 int ramdisk_command_access; 1703 ramdisk_command_access = init_eaccess(ramdisk_execute_command); 1704 if (ramdisk_command_access != 0) { 1705 if (ramdisk_execute_command_set) 1706 pr_warn("check access for rdinit=%s failed: %i, ignoring\n", 1707 ramdisk_execute_command, ramdisk_command_access); 1708 ramdisk_execute_command = NULL; 1709 prepare_namespace(); 1710 } 1711 1712 /* 1713 * Ok, we have completed the initial bootup, and 1714 * we're essentially up and running. Get rid of the 1715 * initmem segments and start the user-mode stuff.. 1716 * 1717 * rootfs is available now, try loading the public keys 1718 * and default modules 1719 */ 1720 1721 integrity_load_keys(); 1722 } 1723