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/sched/clock.h> 87 #include <linux/sched/task.h> 88 #include <linux/sched/task_stack.h> 89 #include <linux/context_tracking.h> 90 #include <linux/random.h> 91 #include <linux/list.h> 92 #include <linux/integrity.h> 93 #include <linux/proc_ns.h> 94 #include <linux/io.h> 95 #include <linux/cache.h> 96 #include <linux/rodata_test.h> 97 #include <linux/jump_label.h> 98 #include <linux/mem_encrypt.h> 99 #include <linux/kcsan.h> 100 #include <linux/init_syscalls.h> 101 #include <linux/stackdepot.h> 102 #include <linux/randomize_kstack.h> 103 #include <net/net_namespace.h> 104 105 #include <asm/io.h> 106 #include <asm/bugs.h> 107 #include <asm/setup.h> 108 #include <asm/sections.h> 109 #include <asm/cacheflush.h> 110 111 #define CREATE_TRACE_POINTS 112 #include <trace/events/initcall.h> 113 114 #include <kunit/test.h> 115 116 static int kernel_init(void *); 117 118 extern void init_IRQ(void); 119 extern void radix_tree_init(void); 120 extern void maple_tree_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 extern void time_init(void); 141 /* Default late time init is NULL. archs can override this later. */ 142 void (*__initdata late_time_init)(void); 143 144 /* Untouched command line saved by arch-specific code. */ 145 char __initdata boot_command_line[COMMAND_LINE_SIZE]; 146 /* Untouched saved command line (eg. for /proc) */ 147 char *saved_command_line; 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"; 166 167 /* 168 * Used to generate warnings if static_key manipulation functions are used 169 * before jump_label_init is called. 170 */ 171 bool static_key_initialized __read_mostly; 172 EXPORT_SYMBOL_GPL(static_key_initialized); 173 174 /* 175 * If set, this is an indication to the drivers that reset the underlying 176 * device before going ahead with the initialization otherwise driver might 177 * rely on the BIOS and skip the reset operation. 178 * 179 * This is useful if kernel is booting in an unreliable environment. 180 * For ex. kdump situation where previous kernel has crashed, BIOS has been 181 * skipped and devices will be in unknown state. 182 */ 183 unsigned int reset_devices; 184 EXPORT_SYMBOL(reset_devices); 185 186 static int __init set_reset_devices(char *str) 187 { 188 reset_devices = 1; 189 return 1; 190 } 191 192 __setup("reset_devices", set_reset_devices); 193 194 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; 195 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; 196 static const char *panic_later, *panic_param; 197 198 extern const struct obs_kernel_param __setup_start[], __setup_end[]; 199 200 static bool __init obsolete_checksetup(char *line) 201 { 202 const struct obs_kernel_param *p; 203 bool had_early_param = false; 204 205 p = __setup_start; 206 do { 207 int n = strlen(p->str); 208 if (parameqn(line, p->str, n)) { 209 if (p->early) { 210 /* Already done in parse_early_param? 211 * (Needs exact match on param part). 212 * Keep iterating, as we can have early 213 * params and __setups of same names 8( */ 214 if (line[n] == '\0' || line[n] == '=') 215 had_early_param = true; 216 } else if (!p->setup_func) { 217 pr_warn("Parameter %s is obsolete, ignored\n", 218 p->str); 219 return true; 220 } else if (p->setup_func(line + n)) 221 return true; 222 } 223 p++; 224 } while (p < __setup_end); 225 226 return had_early_param; 227 } 228 229 /* 230 * This should be approx 2 Bo*oMips to start (note initial shift), and will 231 * still work even if initially too large, it will just take slightly longer 232 */ 233 unsigned long loops_per_jiffy = (1<<12); 234 EXPORT_SYMBOL(loops_per_jiffy); 235 236 static int __init debug_kernel(char *str) 237 { 238 console_loglevel = CONSOLE_LOGLEVEL_DEBUG; 239 return 0; 240 } 241 242 static int __init quiet_kernel(char *str) 243 { 244 console_loglevel = CONSOLE_LOGLEVEL_QUIET; 245 return 0; 246 } 247 248 early_param("debug", debug_kernel); 249 early_param("quiet", quiet_kernel); 250 251 static int __init loglevel(char *str) 252 { 253 int newlevel; 254 255 /* 256 * Only update loglevel value when a correct setting was passed, 257 * to prevent blind crashes (when loglevel being set to 0) that 258 * are quite hard to debug 259 */ 260 if (get_option(&str, &newlevel)) { 261 console_loglevel = newlevel; 262 return 0; 263 } 264 265 return -EINVAL; 266 } 267 268 early_param("loglevel", loglevel); 269 270 #ifdef CONFIG_BLK_DEV_INITRD 271 static void * __init get_boot_config_from_initrd(size_t *_size) 272 { 273 u32 size, csum; 274 char *data; 275 u32 *hdr; 276 int i; 277 278 if (!initrd_end) 279 return NULL; 280 281 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN; 282 /* 283 * Since Grub may align the size of initrd to 4, we must 284 * check the preceding 3 bytes as well. 285 */ 286 for (i = 0; i < 4; i++) { 287 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN)) 288 goto found; 289 data--; 290 } 291 return NULL; 292 293 found: 294 hdr = (u32 *)(data - 8); 295 size = le32_to_cpu(hdr[0]); 296 csum = le32_to_cpu(hdr[1]); 297 298 data = ((void *)hdr) - size; 299 if ((unsigned long)data < initrd_start) { 300 pr_err("bootconfig size %d is greater than initrd size %ld\n", 301 size, initrd_end - initrd_start); 302 return NULL; 303 } 304 305 if (xbc_calc_checksum(data, size) != csum) { 306 pr_err("bootconfig checksum failed\n"); 307 return NULL; 308 } 309 310 /* Remove bootconfig from initramfs/initrd */ 311 initrd_end = (unsigned long)data; 312 if (_size) 313 *_size = size; 314 315 return data; 316 } 317 #else 318 static void * __init get_boot_config_from_initrd(size_t *_size) 319 { 320 return NULL; 321 } 322 #endif 323 324 #ifdef CONFIG_BOOT_CONFIG 325 326 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata; 327 328 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0) 329 330 static int __init xbc_snprint_cmdline(char *buf, size_t size, 331 struct xbc_node *root) 332 { 333 struct xbc_node *knode, *vnode; 334 char *end = buf + size; 335 const char *val; 336 int ret; 337 338 xbc_node_for_each_key_value(root, knode, val) { 339 ret = xbc_node_compose_key_after(root, knode, 340 xbc_namebuf, XBC_KEYLEN_MAX); 341 if (ret < 0) 342 return ret; 343 344 vnode = xbc_node_get_child(knode); 345 if (!vnode) { 346 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf); 347 if (ret < 0) 348 return ret; 349 buf += ret; 350 continue; 351 } 352 xbc_array_for_each_value(vnode, val) { 353 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ", 354 xbc_namebuf, val); 355 if (ret < 0) 356 return ret; 357 buf += ret; 358 } 359 } 360 361 return buf - (end - size); 362 } 363 #undef rest 364 365 /* Make an extra command line under given key word */ 366 static char * __init xbc_make_cmdline(const char *key) 367 { 368 struct xbc_node *root; 369 char *new_cmdline; 370 int ret, len = 0; 371 372 root = xbc_find_node(key); 373 if (!root) 374 return NULL; 375 376 /* Count required buffer size */ 377 len = xbc_snprint_cmdline(NULL, 0, root); 378 if (len <= 0) 379 return NULL; 380 381 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES); 382 if (!new_cmdline) { 383 pr_err("Failed to allocate memory for extra kernel cmdline.\n"); 384 return NULL; 385 } 386 387 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root); 388 if (ret < 0 || ret > len) { 389 pr_err("Failed to print extra kernel cmdline.\n"); 390 memblock_free(new_cmdline, len + 1); 391 return NULL; 392 } 393 394 return new_cmdline; 395 } 396 397 static int __init bootconfig_params(char *param, char *val, 398 const char *unused, void *arg) 399 { 400 if (strcmp(param, "bootconfig") == 0) { 401 bootconfig_found = true; 402 } 403 return 0; 404 } 405 406 static int __init warn_bootconfig(char *str) 407 { 408 /* The 'bootconfig' has been handled by bootconfig_params(). */ 409 return 0; 410 } 411 412 static void __init setup_boot_config(void) 413 { 414 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 415 const char *msg, *data; 416 int pos, ret; 417 size_t size; 418 char *err; 419 420 /* Cut out the bootconfig data even if we have no bootconfig option */ 421 data = get_boot_config_from_initrd(&size); 422 /* If there is no bootconfig in initrd, try embedded one. */ 423 if (!data) 424 data = xbc_get_embedded_bootconfig(&size); 425 426 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 427 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL, 428 bootconfig_params); 429 430 if (IS_ERR(err) || !bootconfig_found) 431 return; 432 433 /* parse_args() stops at the next param of '--' and returns an address */ 434 if (err) 435 initargs_offs = err - tmp_cmdline; 436 437 if (!data) { 438 pr_err("'bootconfig' found on command line, but no bootconfig found\n"); 439 return; 440 } 441 442 if (size >= XBC_DATA_MAX) { 443 pr_err("bootconfig size %ld greater than max size %d\n", 444 (long)size, XBC_DATA_MAX); 445 return; 446 } 447 448 ret = xbc_init(data, size, &msg, &pos); 449 if (ret < 0) { 450 if (pos < 0) 451 pr_err("Failed to init bootconfig: %s.\n", msg); 452 else 453 pr_err("Failed to parse bootconfig: %s at %d.\n", 454 msg, pos); 455 } else { 456 xbc_get_info(&ret, NULL); 457 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret); 458 /* keys starting with "kernel." are passed via cmdline */ 459 extra_command_line = xbc_make_cmdline("kernel"); 460 /* Also, "init." keys are init arguments */ 461 extra_init_args = xbc_make_cmdline("init"); 462 } 463 return; 464 } 465 466 static void __init exit_boot_config(void) 467 { 468 xbc_exit(); 469 } 470 471 #else /* !CONFIG_BOOT_CONFIG */ 472 473 static void __init setup_boot_config(void) 474 { 475 /* Remove bootconfig data from initrd */ 476 get_boot_config_from_initrd(NULL); 477 } 478 479 static int __init warn_bootconfig(char *str) 480 { 481 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n"); 482 return 0; 483 } 484 485 #define exit_boot_config() do {} while (0) 486 487 #endif /* CONFIG_BOOT_CONFIG */ 488 489 early_param("bootconfig", warn_bootconfig); 490 491 /* Change NUL term back to "=", to make "param" the whole string. */ 492 static void __init repair_env_string(char *param, char *val) 493 { 494 if (val) { 495 /* param=val or param="val"? */ 496 if (val == param+strlen(param)+1) 497 val[-1] = '='; 498 else if (val == param+strlen(param)+2) { 499 val[-2] = '='; 500 memmove(val-1, val, strlen(val)+1); 501 } else 502 BUG(); 503 } 504 } 505 506 /* Anything after -- gets handed straight to init. */ 507 static int __init set_init_arg(char *param, char *val, 508 const char *unused, void *arg) 509 { 510 unsigned int i; 511 512 if (panic_later) 513 return 0; 514 515 repair_env_string(param, val); 516 517 for (i = 0; argv_init[i]; i++) { 518 if (i == MAX_INIT_ARGS) { 519 panic_later = "init"; 520 panic_param = param; 521 return 0; 522 } 523 } 524 argv_init[i] = param; 525 return 0; 526 } 527 528 /* 529 * Unknown boot options get handed to init, unless they look like 530 * unused parameters (modprobe will find them in /proc/cmdline). 531 */ 532 static int __init unknown_bootoption(char *param, char *val, 533 const char *unused, void *arg) 534 { 535 size_t len = strlen(param); 536 537 repair_env_string(param, val); 538 539 /* Handle obsolete-style parameters */ 540 if (obsolete_checksetup(param)) 541 return 0; 542 543 /* Unused module parameter. */ 544 if (strnchr(param, len, '.')) 545 return 0; 546 547 if (panic_later) 548 return 0; 549 550 if (val) { 551 /* Environment option */ 552 unsigned int i; 553 for (i = 0; envp_init[i]; i++) { 554 if (i == MAX_INIT_ENVS) { 555 panic_later = "env"; 556 panic_param = param; 557 } 558 if (!strncmp(param, envp_init[i], len+1)) 559 break; 560 } 561 envp_init[i] = param; 562 } else { 563 /* Command line option */ 564 unsigned int i; 565 for (i = 0; argv_init[i]; i++) { 566 if (i == MAX_INIT_ARGS) { 567 panic_later = "init"; 568 panic_param = param; 569 } 570 } 571 argv_init[i] = param; 572 } 573 return 0; 574 } 575 576 static int __init init_setup(char *str) 577 { 578 unsigned int i; 579 580 execute_command = str; 581 /* 582 * In case LILO is going to boot us with default command line, 583 * it prepends "auto" before the whole cmdline which makes 584 * the shell think it should execute a script with such name. 585 * So we ignore all arguments entered _before_ init=... [MJ] 586 */ 587 for (i = 1; i < MAX_INIT_ARGS; i++) 588 argv_init[i] = NULL; 589 return 1; 590 } 591 __setup("init=", init_setup); 592 593 static int __init rdinit_setup(char *str) 594 { 595 unsigned int i; 596 597 ramdisk_execute_command = str; 598 /* See "auto" comment in init_setup */ 599 for (i = 1; i < MAX_INIT_ARGS; i++) 600 argv_init[i] = NULL; 601 return 1; 602 } 603 __setup("rdinit=", rdinit_setup); 604 605 #ifndef CONFIG_SMP 606 static const unsigned int setup_max_cpus = NR_CPUS; 607 static inline void setup_nr_cpu_ids(void) { } 608 static inline void smp_prepare_cpus(unsigned int maxcpus) { } 609 #endif 610 611 /* 612 * We need to store the untouched command line for future reference. 613 * We also need to store the touched command line since the parameter 614 * parsing is performed in place, and we should allow a component to 615 * store reference of name/value for future reference. 616 */ 617 static void __init setup_command_line(char *command_line) 618 { 619 size_t len, xlen = 0, ilen = 0; 620 621 if (extra_command_line) 622 xlen = strlen(extra_command_line); 623 if (extra_init_args) 624 ilen = strlen(extra_init_args) + 4; /* for " -- " */ 625 626 len = xlen + strlen(boot_command_line) + 1; 627 628 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES); 629 if (!saved_command_line) 630 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen); 631 632 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES); 633 if (!static_command_line) 634 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 635 636 if (xlen) { 637 /* 638 * We have to put extra_command_line before boot command 639 * lines because there could be dashes (separator of init 640 * command line) in the command lines. 641 */ 642 strcpy(saved_command_line, extra_command_line); 643 strcpy(static_command_line, extra_command_line); 644 } 645 strcpy(saved_command_line + xlen, boot_command_line); 646 strcpy(static_command_line + xlen, command_line); 647 648 if (ilen) { 649 /* 650 * Append supplemental init boot args to saved_command_line 651 * so that user can check what command line options passed 652 * to init. 653 * The order should always be 654 * " -- "[bootconfig init-param][cmdline init-param] 655 */ 656 if (initargs_offs) { 657 len = xlen + initargs_offs; 658 strcpy(saved_command_line + len, extra_init_args); 659 len += ilen - 4; /* strlen(extra_init_args) */ 660 strcpy(saved_command_line + len, 661 boot_command_line + initargs_offs - 1); 662 } else { 663 len = strlen(saved_command_line); 664 strcpy(saved_command_line + len, " -- "); 665 len += 4; 666 strcpy(saved_command_line + len, extra_init_args); 667 } 668 } 669 } 670 671 /* 672 * We need to finalize in a non-__init function or else race conditions 673 * between the root thread and the init thread may cause start_kernel to 674 * be reaped by free_initmem before the root thread has proceeded to 675 * cpu_idle. 676 * 677 * gcc-3.4 accidentally inlines this function, so use noinline. 678 */ 679 680 static __initdata DECLARE_COMPLETION(kthreadd_done); 681 682 noinline void __ref rest_init(void) 683 { 684 struct task_struct *tsk; 685 int pid; 686 687 rcu_scheduler_starting(); 688 /* 689 * We need to spawn init first so that it obtains pid 1, however 690 * the init task will end up wanting to create kthreads, which, if 691 * we schedule it before we create kthreadd, will OOPS. 692 */ 693 pid = user_mode_thread(kernel_init, NULL, CLONE_FS); 694 /* 695 * Pin init on the boot CPU. Task migration is not properly working 696 * until sched_init_smp() has been run. It will set the allowed 697 * CPUs for init to the non isolated CPUs. 698 */ 699 rcu_read_lock(); 700 tsk = find_task_by_pid_ns(pid, &init_pid_ns); 701 tsk->flags |= PF_NO_SETAFFINITY; 702 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id())); 703 rcu_read_unlock(); 704 705 numa_default_policy(); 706 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); 707 rcu_read_lock(); 708 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); 709 rcu_read_unlock(); 710 711 /* 712 * Enable might_sleep() and smp_processor_id() checks. 713 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y 714 * kernel_thread() would trigger might_sleep() splats. With 715 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled 716 * already, but it's stuck on the kthreadd_done completion. 717 */ 718 system_state = SYSTEM_SCHEDULING; 719 720 complete(&kthreadd_done); 721 722 /* 723 * The boot idle thread must execute schedule() 724 * at least once to get things moving: 725 */ 726 schedule_preempt_disabled(); 727 /* Call into cpu_idle with preempt disabled */ 728 cpu_startup_entry(CPUHP_ONLINE); 729 } 730 731 /* Check for early params. */ 732 static int __init do_early_param(char *param, char *val, 733 const char *unused, void *arg) 734 { 735 const struct obs_kernel_param *p; 736 737 for (p = __setup_start; p < __setup_end; p++) { 738 if ((p->early && parameq(param, p->str)) || 739 (strcmp(param, "console") == 0 && 740 strcmp(p->str, "earlycon") == 0) 741 ) { 742 if (p->setup_func(val) != 0) 743 pr_warn("Malformed early option '%s'\n", param); 744 } 745 } 746 /* We accept everything at this stage. */ 747 return 0; 748 } 749 750 void __init parse_early_options(char *cmdline) 751 { 752 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL, 753 do_early_param); 754 } 755 756 /* Arch code calls this early on, or if not, just before other parsing. */ 757 void __init parse_early_param(void) 758 { 759 static int done __initdata; 760 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 761 762 if (done) 763 return; 764 765 /* All fall through to do_early_param. */ 766 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 767 parse_early_options(tmp_cmdline); 768 done = 1; 769 } 770 771 void __init __weak arch_post_acpi_subsys_init(void) { } 772 773 void __init __weak smp_setup_processor_id(void) 774 { 775 } 776 777 # if THREAD_SIZE >= PAGE_SIZE 778 void __init __weak thread_stack_cache_init(void) 779 { 780 } 781 #endif 782 783 void __init __weak mem_encrypt_init(void) { } 784 785 void __init __weak poking_init(void) { } 786 787 void __init __weak pgtable_cache_init(void) { } 788 789 void __init __weak trap_init(void) { } 790 791 bool initcall_debug; 792 core_param(initcall_debug, initcall_debug, bool, 0644); 793 794 #ifdef TRACEPOINTS_ENABLED 795 static void __init initcall_debug_enable(void); 796 #else 797 static inline void initcall_debug_enable(void) 798 { 799 } 800 #endif 801 802 /* Report memory auto-initialization states for this boot. */ 803 static void __init report_meminit(void) 804 { 805 const char *stack; 806 807 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN)) 808 stack = "all(pattern)"; 809 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO)) 810 stack = "all(zero)"; 811 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL)) 812 stack = "byref_all(zero)"; 813 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF)) 814 stack = "byref(zero)"; 815 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER)) 816 stack = "__user(zero)"; 817 else 818 stack = "off"; 819 820 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n", 821 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off", 822 want_init_on_free() ? "on" : "off"); 823 if (want_init_on_free()) 824 pr_info("mem auto-init: clearing system memory may take some time...\n"); 825 } 826 827 /* 828 * Set up kernel memory allocators 829 */ 830 static void __init mm_init(void) 831 { 832 /* 833 * page_ext requires contiguous pages, 834 * bigger than MAX_ORDER unless SPARSEMEM. 835 */ 836 page_ext_init_flatmem(); 837 init_mem_debugging_and_hardening(); 838 kfence_alloc_pool(); 839 report_meminit(); 840 stack_depot_early_init(); 841 mem_init(); 842 mem_init_print_info(); 843 kmem_cache_init(); 844 /* 845 * page_owner must be initialized after buddy is ready, and also after 846 * slab is ready so that stack_depot_init() works properly 847 */ 848 page_ext_init_flatmem_late(); 849 kmemleak_init(); 850 pgtable_init(); 851 debug_objects_mem_init(); 852 vmalloc_init(); 853 /* Should be run after vmap initialization */ 854 if (early_page_ext_enabled()) 855 page_ext_init(); 856 /* Should be run before the first non-init thread is created */ 857 init_espfix_bsp(); 858 /* Should be run after espfix64 is set up. */ 859 pti_init(); 860 } 861 862 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET 863 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT, 864 randomize_kstack_offset); 865 DEFINE_PER_CPU(u32, kstack_offset); 866 867 static int __init early_randomize_kstack_offset(char *buf) 868 { 869 int ret; 870 bool bool_result; 871 872 ret = kstrtobool(buf, &bool_result); 873 if (ret) 874 return ret; 875 876 if (bool_result) 877 static_branch_enable(&randomize_kstack_offset); 878 else 879 static_branch_disable(&randomize_kstack_offset); 880 return 0; 881 } 882 early_param("randomize_kstack_offset", early_randomize_kstack_offset); 883 #endif 884 885 void __init __weak arch_call_rest_init(void) 886 { 887 rest_init(); 888 } 889 890 static void __init print_unknown_bootoptions(void) 891 { 892 char *unknown_options; 893 char *end; 894 const char *const *p; 895 size_t len; 896 897 if (panic_later || (!argv_init[1] && !envp_init[2])) 898 return; 899 900 /* 901 * Determine how many options we have to print out, plus a space 902 * before each 903 */ 904 len = 1; /* null terminator */ 905 for (p = &argv_init[1]; *p; p++) { 906 len++; 907 len += strlen(*p); 908 } 909 for (p = &envp_init[2]; *p; p++) { 910 len++; 911 len += strlen(*p); 912 } 913 914 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES); 915 if (!unknown_options) { 916 pr_err("%s: Failed to allocate %zu bytes\n", 917 __func__, len); 918 return; 919 } 920 end = unknown_options; 921 922 for (p = &argv_init[1]; *p; p++) 923 end += sprintf(end, " %s", *p); 924 for (p = &envp_init[2]; *p; p++) 925 end += sprintf(end, " %s", *p); 926 927 /* Start at unknown_options[1] to skip the initial space */ 928 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n", 929 &unknown_options[1]); 930 memblock_free(unknown_options, len); 931 } 932 933 asmlinkage __visible void __init __no_sanitize_address start_kernel(void) 934 { 935 char *command_line; 936 char *after_dashes; 937 938 set_task_stack_end_magic(&init_task); 939 smp_setup_processor_id(); 940 debug_objects_early_init(); 941 init_vmlinux_build_id(); 942 943 cgroup_init_early(); 944 945 local_irq_disable(); 946 early_boot_irqs_disabled = true; 947 948 /* 949 * Interrupts are still disabled. Do necessary setups, then 950 * enable them. 951 */ 952 boot_cpu_init(); 953 page_address_init(); 954 pr_notice("%s", linux_banner); 955 early_security_init(); 956 setup_arch(&command_line); 957 setup_boot_config(); 958 setup_command_line(command_line); 959 setup_nr_cpu_ids(); 960 setup_per_cpu_areas(); 961 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ 962 boot_cpu_hotplug_init(); 963 964 build_all_zonelists(NULL); 965 page_alloc_init(); 966 967 pr_notice("Kernel command line: %s\n", saved_command_line); 968 /* parameters may set static keys */ 969 jump_label_init(); 970 parse_early_param(); 971 after_dashes = parse_args("Booting kernel", 972 static_command_line, __start___param, 973 __stop___param - __start___param, 974 -1, -1, NULL, &unknown_bootoption); 975 print_unknown_bootoptions(); 976 if (!IS_ERR_OR_NULL(after_dashes)) 977 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1, 978 NULL, set_init_arg); 979 if (extra_init_args) 980 parse_args("Setting extra init args", extra_init_args, 981 NULL, 0, -1, -1, NULL, set_init_arg); 982 983 /* 984 * These use large bootmem allocations and must precede 985 * kmem_cache_init() 986 */ 987 setup_log_buf(0); 988 vfs_caches_init_early(); 989 sort_main_extable(); 990 trap_init(); 991 mm_init(); 992 993 ftrace_init(); 994 995 /* trace_printk can be enabled here */ 996 early_trace_init(); 997 998 /* 999 * Set up the scheduler prior starting any interrupts (such as the 1000 * timer interrupt). Full topology setup happens at smp_init() 1001 * time - but meanwhile we still have a functioning scheduler. 1002 */ 1003 sched_init(); 1004 1005 if (WARN(!irqs_disabled(), 1006 "Interrupts were enabled *very* early, fixing it\n")) 1007 local_irq_disable(); 1008 radix_tree_init(); 1009 maple_tree_init(); 1010 1011 /* 1012 * Set up housekeeping before setting up workqueues to allow the unbound 1013 * workqueue to take non-housekeeping into account. 1014 */ 1015 housekeeping_init(); 1016 1017 /* 1018 * Allow workqueue creation and work item queueing/cancelling 1019 * early. Work item execution depends on kthreads and starts after 1020 * workqueue_init(). 1021 */ 1022 workqueue_init_early(); 1023 1024 rcu_init(); 1025 1026 /* Trace events are available after this */ 1027 trace_init(); 1028 1029 if (initcall_debug) 1030 initcall_debug_enable(); 1031 1032 context_tracking_init(); 1033 /* init some links before init_ISA_irqs() */ 1034 early_irq_init(); 1035 init_IRQ(); 1036 tick_init(); 1037 rcu_init_nohz(); 1038 init_timers(); 1039 srcu_init(); 1040 hrtimers_init(); 1041 softirq_init(); 1042 timekeeping_init(); 1043 kfence_init(); 1044 time_init(); 1045 1046 /* 1047 * For best initial stack canary entropy, prepare it after: 1048 * - setup_arch() for any UEFI RNG entropy and boot cmdline access 1049 * - timekeeping_init() for ktime entropy used in random_init() 1050 * - time_init() for making random_get_entropy() work on some platforms 1051 * - random_init() to initialize the RNG from from early entropy sources 1052 */ 1053 random_init(command_line); 1054 boot_init_stack_canary(); 1055 1056 perf_event_init(); 1057 profile_init(); 1058 call_function_init(); 1059 WARN(!irqs_disabled(), "Interrupts were enabled early\n"); 1060 1061 early_boot_irqs_disabled = false; 1062 local_irq_enable(); 1063 1064 kmem_cache_init_late(); 1065 1066 /* 1067 * HACK ALERT! This is early. We're enabling the console before 1068 * we've done PCI setups etc, and console_init() must be aware of 1069 * this. But we do want output early, in case something goes wrong. 1070 */ 1071 console_init(); 1072 if (panic_later) 1073 panic("Too many boot %s vars at `%s'", panic_later, 1074 panic_param); 1075 1076 lockdep_init(); 1077 1078 /* 1079 * Need to run this when irqs are enabled, because it wants 1080 * to self-test [hard/soft]-irqs on/off lock inversion bugs 1081 * too: 1082 */ 1083 locking_selftest(); 1084 1085 /* 1086 * This needs to be called before any devices perform DMA 1087 * operations that might use the SWIOTLB bounce buffers. It will 1088 * mark the bounce buffers as decrypted so that their usage will 1089 * not cause "plain-text" data to be decrypted when accessed. 1090 */ 1091 mem_encrypt_init(); 1092 1093 #ifdef CONFIG_BLK_DEV_INITRD 1094 if (initrd_start && !initrd_below_start_ok && 1095 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { 1096 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", 1097 page_to_pfn(virt_to_page((void *)initrd_start)), 1098 min_low_pfn); 1099 initrd_start = 0; 1100 } 1101 #endif 1102 setup_per_cpu_pageset(); 1103 numa_policy_init(); 1104 acpi_early_init(); 1105 if (late_time_init) 1106 late_time_init(); 1107 sched_clock_init(); 1108 calibrate_delay(); 1109 pid_idr_init(); 1110 anon_vma_init(); 1111 #ifdef CONFIG_X86 1112 if (efi_enabled(EFI_RUNTIME_SERVICES)) 1113 efi_enter_virtual_mode(); 1114 #endif 1115 thread_stack_cache_init(); 1116 cred_init(); 1117 fork_init(); 1118 proc_caches_init(); 1119 uts_ns_init(); 1120 key_init(); 1121 security_init(); 1122 dbg_late_init(); 1123 net_ns_init(); 1124 vfs_caches_init(); 1125 pagecache_init(); 1126 signals_init(); 1127 seq_file_init(); 1128 proc_root_init(); 1129 nsfs_init(); 1130 cpuset_init(); 1131 cgroup_init(); 1132 taskstats_init_early(); 1133 delayacct_init(); 1134 1135 poking_init(); 1136 check_bugs(); 1137 1138 acpi_subsystem_init(); 1139 arch_post_acpi_subsys_init(); 1140 kcsan_init(); 1141 1142 /* Do the rest non-__init'ed, we're now alive */ 1143 arch_call_rest_init(); 1144 1145 prevent_tail_call_optimization(); 1146 } 1147 1148 /* Call all constructor functions linked into the kernel. */ 1149 static void __init do_ctors(void) 1150 { 1151 /* 1152 * For UML, the constructors have already been called by the 1153 * normal setup code as it's just a normal ELF binary, so we 1154 * cannot do it again - but we do need CONFIG_CONSTRUCTORS 1155 * even on UML for modules. 1156 */ 1157 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML) 1158 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; 1159 1160 for (; fn < (ctor_fn_t *) __ctors_end; fn++) 1161 (*fn)(); 1162 #endif 1163 } 1164 1165 #ifdef CONFIG_KALLSYMS 1166 struct blacklist_entry { 1167 struct list_head next; 1168 char *buf; 1169 }; 1170 1171 static __initdata_or_module LIST_HEAD(blacklisted_initcalls); 1172 1173 static int __init initcall_blacklist(char *str) 1174 { 1175 char *str_entry; 1176 struct blacklist_entry *entry; 1177 1178 /* str argument is a comma-separated list of functions */ 1179 do { 1180 str_entry = strsep(&str, ","); 1181 if (str_entry) { 1182 pr_debug("blacklisting initcall %s\n", str_entry); 1183 entry = memblock_alloc(sizeof(*entry), 1184 SMP_CACHE_BYTES); 1185 if (!entry) 1186 panic("%s: Failed to allocate %zu bytes\n", 1187 __func__, sizeof(*entry)); 1188 entry->buf = memblock_alloc(strlen(str_entry) + 1, 1189 SMP_CACHE_BYTES); 1190 if (!entry->buf) 1191 panic("%s: Failed to allocate %zu bytes\n", 1192 __func__, strlen(str_entry) + 1); 1193 strcpy(entry->buf, str_entry); 1194 list_add(&entry->next, &blacklisted_initcalls); 1195 } 1196 } while (str_entry); 1197 1198 return 1; 1199 } 1200 1201 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1202 { 1203 struct blacklist_entry *entry; 1204 char fn_name[KSYM_SYMBOL_LEN]; 1205 unsigned long addr; 1206 1207 if (list_empty(&blacklisted_initcalls)) 1208 return false; 1209 1210 addr = (unsigned long) dereference_function_descriptor(fn); 1211 sprint_symbol_no_offset(fn_name, addr); 1212 1213 /* 1214 * fn will be "function_name [module_name]" where [module_name] is not 1215 * displayed for built-in init functions. Strip off the [module_name]. 1216 */ 1217 strreplace(fn_name, ' ', '\0'); 1218 1219 list_for_each_entry(entry, &blacklisted_initcalls, next) { 1220 if (!strcmp(fn_name, entry->buf)) { 1221 pr_debug("initcall %s blacklisted\n", fn_name); 1222 return true; 1223 } 1224 } 1225 1226 return false; 1227 } 1228 #else 1229 static int __init initcall_blacklist(char *str) 1230 { 1231 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); 1232 return 0; 1233 } 1234 1235 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1236 { 1237 return false; 1238 } 1239 #endif 1240 __setup("initcall_blacklist=", initcall_blacklist); 1241 1242 static __init_or_module void 1243 trace_initcall_start_cb(void *data, initcall_t fn) 1244 { 1245 ktime_t *calltime = (ktime_t *)data; 1246 1247 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current)); 1248 *calltime = ktime_get(); 1249 } 1250 1251 static __init_or_module void 1252 trace_initcall_finish_cb(void *data, initcall_t fn, int ret) 1253 { 1254 ktime_t rettime, *calltime = (ktime_t *)data; 1255 1256 rettime = ktime_get(); 1257 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n", 1258 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime)); 1259 } 1260 1261 static ktime_t initcall_calltime; 1262 1263 #ifdef TRACEPOINTS_ENABLED 1264 static void __init initcall_debug_enable(void) 1265 { 1266 int ret; 1267 1268 ret = register_trace_initcall_start(trace_initcall_start_cb, 1269 &initcall_calltime); 1270 ret |= register_trace_initcall_finish(trace_initcall_finish_cb, 1271 &initcall_calltime); 1272 WARN(ret, "Failed to register initcall tracepoints\n"); 1273 } 1274 # define do_trace_initcall_start trace_initcall_start 1275 # define do_trace_initcall_finish trace_initcall_finish 1276 #else 1277 static inline void do_trace_initcall_start(initcall_t fn) 1278 { 1279 if (!initcall_debug) 1280 return; 1281 trace_initcall_start_cb(&initcall_calltime, fn); 1282 } 1283 static inline void do_trace_initcall_finish(initcall_t fn, int ret) 1284 { 1285 if (!initcall_debug) 1286 return; 1287 trace_initcall_finish_cb(&initcall_calltime, fn, ret); 1288 } 1289 #endif /* !TRACEPOINTS_ENABLED */ 1290 1291 int __init_or_module do_one_initcall(initcall_t fn) 1292 { 1293 int count = preempt_count(); 1294 char msgbuf[64]; 1295 int ret; 1296 1297 if (initcall_blacklisted(fn)) 1298 return -EPERM; 1299 1300 do_trace_initcall_start(fn); 1301 ret = fn(); 1302 do_trace_initcall_finish(fn, ret); 1303 1304 msgbuf[0] = 0; 1305 1306 if (preempt_count() != count) { 1307 sprintf(msgbuf, "preemption imbalance "); 1308 preempt_count_set(count); 1309 } 1310 if (irqs_disabled()) { 1311 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); 1312 local_irq_enable(); 1313 } 1314 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf); 1315 1316 add_latent_entropy(); 1317 return ret; 1318 } 1319 1320 1321 extern initcall_entry_t __initcall_start[]; 1322 extern initcall_entry_t __initcall0_start[]; 1323 extern initcall_entry_t __initcall1_start[]; 1324 extern initcall_entry_t __initcall2_start[]; 1325 extern initcall_entry_t __initcall3_start[]; 1326 extern initcall_entry_t __initcall4_start[]; 1327 extern initcall_entry_t __initcall5_start[]; 1328 extern initcall_entry_t __initcall6_start[]; 1329 extern initcall_entry_t __initcall7_start[]; 1330 extern initcall_entry_t __initcall_end[]; 1331 1332 static initcall_entry_t *initcall_levels[] __initdata = { 1333 __initcall0_start, 1334 __initcall1_start, 1335 __initcall2_start, 1336 __initcall3_start, 1337 __initcall4_start, 1338 __initcall5_start, 1339 __initcall6_start, 1340 __initcall7_start, 1341 __initcall_end, 1342 }; 1343 1344 /* Keep these in sync with initcalls in include/linux/init.h */ 1345 static const char *initcall_level_names[] __initdata = { 1346 "pure", 1347 "core", 1348 "postcore", 1349 "arch", 1350 "subsys", 1351 "fs", 1352 "device", 1353 "late", 1354 }; 1355 1356 static int __init ignore_unknown_bootoption(char *param, char *val, 1357 const char *unused, void *arg) 1358 { 1359 return 0; 1360 } 1361 1362 static void __init do_initcall_level(int level, char *command_line) 1363 { 1364 initcall_entry_t *fn; 1365 1366 parse_args(initcall_level_names[level], 1367 command_line, __start___param, 1368 __stop___param - __start___param, 1369 level, level, 1370 NULL, ignore_unknown_bootoption); 1371 1372 trace_initcall_level(initcall_level_names[level]); 1373 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) 1374 do_one_initcall(initcall_from_entry(fn)); 1375 } 1376 1377 static void __init do_initcalls(void) 1378 { 1379 int level; 1380 size_t len = strlen(saved_command_line) + 1; 1381 char *command_line; 1382 1383 command_line = kzalloc(len, GFP_KERNEL); 1384 if (!command_line) 1385 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 1386 1387 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) { 1388 /* Parser modifies command_line, restore it each time */ 1389 strcpy(command_line, saved_command_line); 1390 do_initcall_level(level, command_line); 1391 } 1392 1393 kfree(command_line); 1394 } 1395 1396 /* 1397 * Ok, the machine is now initialized. None of the devices 1398 * have been touched yet, but the CPU subsystem is up and 1399 * running, and memory and process management works. 1400 * 1401 * Now we can finally start doing some real work.. 1402 */ 1403 static void __init do_basic_setup(void) 1404 { 1405 cpuset_init_smp(); 1406 driver_init(); 1407 init_irq_proc(); 1408 do_ctors(); 1409 do_initcalls(); 1410 } 1411 1412 static void __init do_pre_smp_initcalls(void) 1413 { 1414 initcall_entry_t *fn; 1415 1416 trace_initcall_level("early"); 1417 for (fn = __initcall_start; fn < __initcall0_start; fn++) 1418 do_one_initcall(initcall_from_entry(fn)); 1419 } 1420 1421 static int run_init_process(const char *init_filename) 1422 { 1423 const char *const *p; 1424 1425 argv_init[0] = init_filename; 1426 pr_info("Run %s as init process\n", init_filename); 1427 pr_debug(" with arguments:\n"); 1428 for (p = argv_init; *p; p++) 1429 pr_debug(" %s\n", *p); 1430 pr_debug(" with environment:\n"); 1431 for (p = envp_init; *p; p++) 1432 pr_debug(" %s\n", *p); 1433 return kernel_execve(init_filename, argv_init, envp_init); 1434 } 1435 1436 static int try_to_run_init_process(const char *init_filename) 1437 { 1438 int ret; 1439 1440 ret = run_init_process(init_filename); 1441 1442 if (ret && ret != -ENOENT) { 1443 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", 1444 init_filename, ret); 1445 } 1446 1447 return ret; 1448 } 1449 1450 static noinline void __init kernel_init_freeable(void); 1451 1452 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX) 1453 bool rodata_enabled __ro_after_init = true; 1454 1455 #ifndef arch_parse_debug_rodata 1456 static inline bool arch_parse_debug_rodata(char *str) { return false; } 1457 #endif 1458 1459 static int __init set_debug_rodata(char *str) 1460 { 1461 if (arch_parse_debug_rodata(str)) 1462 return 0; 1463 1464 if (str && !strcmp(str, "on")) 1465 rodata_enabled = true; 1466 else if (str && !strcmp(str, "off")) 1467 rodata_enabled = false; 1468 else 1469 pr_warn("Invalid option string for rodata: '%s'\n", str); 1470 return 0; 1471 } 1472 early_param("rodata", set_debug_rodata); 1473 #endif 1474 1475 #ifdef CONFIG_STRICT_KERNEL_RWX 1476 static void mark_readonly(void) 1477 { 1478 if (rodata_enabled) { 1479 /* 1480 * load_module() results in W+X mappings, which are cleaned 1481 * up with call_rcu(). Let's make sure that queued work is 1482 * flushed so that we don't hit false positives looking for 1483 * insecure pages which are W+X. 1484 */ 1485 rcu_barrier(); 1486 mark_rodata_ro(); 1487 rodata_test(); 1488 } else 1489 pr_info("Kernel memory protection disabled.\n"); 1490 } 1491 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX) 1492 static inline void mark_readonly(void) 1493 { 1494 pr_warn("Kernel memory protection not selected by kernel config.\n"); 1495 } 1496 #else 1497 static inline void mark_readonly(void) 1498 { 1499 pr_warn("This architecture does not have kernel memory protection.\n"); 1500 } 1501 #endif 1502 1503 void __weak free_initmem(void) 1504 { 1505 free_initmem_default(POISON_FREE_INITMEM); 1506 } 1507 1508 static int __ref kernel_init(void *unused) 1509 { 1510 int ret; 1511 1512 /* 1513 * Wait until kthreadd is all set-up. 1514 */ 1515 wait_for_completion(&kthreadd_done); 1516 1517 kernel_init_freeable(); 1518 /* need to finish all async __init code before freeing the memory */ 1519 async_synchronize_full(); 1520 1521 system_state = SYSTEM_FREEING_INITMEM; 1522 kprobe_free_init_mem(); 1523 ftrace_free_init_mem(); 1524 kgdb_free_init_mem(); 1525 exit_boot_config(); 1526 free_initmem(); 1527 mark_readonly(); 1528 1529 /* 1530 * Kernel mappings are now finalized - update the userspace page-table 1531 * to finalize PTI. 1532 */ 1533 pti_finalize(); 1534 1535 system_state = SYSTEM_RUNNING; 1536 numa_default_policy(); 1537 1538 rcu_end_inkernel_boot(); 1539 1540 do_sysctl_args(); 1541 1542 if (ramdisk_execute_command) { 1543 ret = run_init_process(ramdisk_execute_command); 1544 if (!ret) 1545 return 0; 1546 pr_err("Failed to execute %s (error %d)\n", 1547 ramdisk_execute_command, ret); 1548 } 1549 1550 /* 1551 * We try each of these until one succeeds. 1552 * 1553 * The Bourne shell can be used instead of init if we are 1554 * trying to recover a really broken machine. 1555 */ 1556 if (execute_command) { 1557 ret = run_init_process(execute_command); 1558 if (!ret) 1559 return 0; 1560 panic("Requested init %s failed (error %d).", 1561 execute_command, ret); 1562 } 1563 1564 if (CONFIG_DEFAULT_INIT[0] != '\0') { 1565 ret = run_init_process(CONFIG_DEFAULT_INIT); 1566 if (ret) 1567 pr_err("Default init %s failed (error %d)\n", 1568 CONFIG_DEFAULT_INIT, ret); 1569 else 1570 return 0; 1571 } 1572 1573 if (!try_to_run_init_process("/sbin/init") || 1574 !try_to_run_init_process("/etc/init") || 1575 !try_to_run_init_process("/bin/init") || 1576 !try_to_run_init_process("/bin/sh")) 1577 return 0; 1578 1579 panic("No working init found. Try passing init= option to kernel. " 1580 "See Linux Documentation/admin-guide/init.rst for guidance."); 1581 } 1582 1583 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */ 1584 void __init console_on_rootfs(void) 1585 { 1586 struct file *file = filp_open("/dev/console", O_RDWR, 0); 1587 1588 if (IS_ERR(file)) { 1589 pr_err("Warning: unable to open an initial console.\n"); 1590 return; 1591 } 1592 init_dup(file); 1593 init_dup(file); 1594 init_dup(file); 1595 fput(file); 1596 } 1597 1598 static noinline void __init kernel_init_freeable(void) 1599 { 1600 /* Now the scheduler is fully set up and can do blocking allocations */ 1601 gfp_allowed_mask = __GFP_BITS_MASK; 1602 1603 /* 1604 * init can allocate pages on any node 1605 */ 1606 set_mems_allowed(node_states[N_MEMORY]); 1607 1608 cad_pid = get_pid(task_pid(current)); 1609 1610 smp_prepare_cpus(setup_max_cpus); 1611 1612 workqueue_init(); 1613 1614 init_mm_internals(); 1615 1616 rcu_init_tasks_generic(); 1617 do_pre_smp_initcalls(); 1618 lockup_detector_init(); 1619 1620 smp_init(); 1621 sched_init_smp(); 1622 1623 padata_init(); 1624 page_alloc_init_late(); 1625 /* Initialize page ext after all struct pages are initialized. */ 1626 if (!early_page_ext_enabled()) 1627 page_ext_init(); 1628 1629 do_basic_setup(); 1630 1631 kunit_run_all_tests(); 1632 1633 wait_for_initramfs(); 1634 console_on_rootfs(); 1635 1636 /* 1637 * check if there is an early userspace init. If yes, let it do all 1638 * the work 1639 */ 1640 if (init_eaccess(ramdisk_execute_command) != 0) { 1641 ramdisk_execute_command = NULL; 1642 prepare_namespace(); 1643 } 1644 1645 /* 1646 * Ok, we have completed the initial bootup, and 1647 * we're essentially up and running. Get rid of the 1648 * initmem segments and start the user-mode stuff.. 1649 * 1650 * rootfs is available now, try loading the public keys 1651 * and default modules 1652 */ 1653 1654 integrity_load_keys(); 1655 } 1656