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/kmsan.h> 38 #include <linux/vmalloc.h> 39 #include <linux/kernel_stat.h> 40 #include <linux/start_kernel.h> 41 #include <linux/security.h> 42 #include <linux/smp.h> 43 #include <linux/profile.h> 44 #include <linux/kfence.h> 45 #include <linux/rcupdate.h> 46 #include <linux/srcu.h> 47 #include <linux/moduleparam.h> 48 #include <linux/kallsyms.h> 49 #include <linux/buildid.h> 50 #include <linux/writeback.h> 51 #include <linux/cpu.h> 52 #include <linux/cpuset.h> 53 #include <linux/cgroup.h> 54 #include <linux/efi.h> 55 #include <linux/tick.h> 56 #include <linux/sched/isolation.h> 57 #include <linux/interrupt.h> 58 #include <linux/taskstats_kern.h> 59 #include <linux/delayacct.h> 60 #include <linux/unistd.h> 61 #include <linux/utsname.h> 62 #include <linux/rmap.h> 63 #include <linux/mempolicy.h> 64 #include <linux/key.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/kcsan.h> 99 #include <linux/init_syscalls.h> 100 #include <linux/stackdepot.h> 101 #include <linux/randomize_kstack.h> 102 #include <linux/pidfs.h> 103 #include <net/net_namespace.h> 104 105 #include <asm/io.h> 106 #include <asm/setup.h> 107 #include <asm/sections.h> 108 #include <asm/cacheflush.h> 109 110 #define CREATE_TRACE_POINTS 111 #include <trace/events/initcall.h> 112 113 #include <kunit/test.h> 114 115 static int kernel_init(void *); 116 117 /* 118 * Debug helper: via this flag we know that we are in 'early bootup code' 119 * where only the boot processor is running with IRQ disabled. This means 120 * two things - IRQ must not be enabled before the flag is cleared and some 121 * operations which are not allowed with IRQ disabled are allowed while the 122 * flag is set. 123 */ 124 bool early_boot_irqs_disabled __read_mostly; 125 126 enum system_states system_state __read_mostly; 127 EXPORT_SYMBOL(system_state); 128 129 /* 130 * Boot command-line arguments 131 */ 132 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT 133 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT 134 135 /* Default late time init is NULL. archs can override this later. */ 136 void (*__initdata late_time_init)(void); 137 138 /* Untouched command line saved by arch-specific code. */ 139 char __initdata boot_command_line[COMMAND_LINE_SIZE]; 140 /* Untouched saved command line (eg. for /proc) */ 141 char *saved_command_line __ro_after_init; 142 unsigned int saved_command_line_len __ro_after_init; 143 /* Command line for parameter parsing */ 144 static char *static_command_line; 145 /* Untouched extra command line */ 146 static char *extra_command_line; 147 /* Extra init arguments */ 148 static char *extra_init_args; 149 150 #ifdef CONFIG_BOOT_CONFIG 151 /* Is bootconfig on command line? */ 152 static bool bootconfig_found; 153 static size_t initargs_offs; 154 #else 155 # define bootconfig_found false 156 # define initargs_offs 0 157 #endif 158 159 static char *execute_command; 160 static char *ramdisk_execute_command = "/init"; 161 162 /* 163 * Used to generate warnings if static_key manipulation functions are used 164 * before jump_label_init is called. 165 */ 166 bool static_key_initialized __read_mostly; 167 EXPORT_SYMBOL_GPL(static_key_initialized); 168 169 /* 170 * If set, this is an indication to the drivers that reset the underlying 171 * device before going ahead with the initialization otherwise driver might 172 * rely on the BIOS and skip the reset operation. 173 * 174 * This is useful if kernel is booting in an unreliable environment. 175 * For ex. kdump situation where previous kernel has crashed, BIOS has been 176 * skipped and devices will be in unknown state. 177 */ 178 unsigned int reset_devices; 179 EXPORT_SYMBOL(reset_devices); 180 181 static int __init set_reset_devices(char *str) 182 { 183 reset_devices = 1; 184 return 1; 185 } 186 187 __setup("reset_devices", set_reset_devices); 188 189 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; 190 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; 191 static const char *panic_later, *panic_param; 192 193 static bool __init obsolete_checksetup(char *line) 194 { 195 const struct obs_kernel_param *p; 196 bool had_early_param = false; 197 198 p = __setup_start; 199 do { 200 int n = strlen(p->str); 201 if (parameqn(line, p->str, n)) { 202 if (p->early) { 203 /* Already done in parse_early_param? 204 * (Needs exact match on param part). 205 * Keep iterating, as we can have early 206 * params and __setups of same names 8( */ 207 if (line[n] == '\0' || line[n] == '=') 208 had_early_param = true; 209 } else if (!p->setup_func) { 210 pr_warn("Parameter %s is obsolete, ignored\n", 211 p->str); 212 return true; 213 } else if (p->setup_func(line + n)) 214 return true; 215 } 216 p++; 217 } while (p < __setup_end); 218 219 return had_early_param; 220 } 221 222 /* 223 * This should be approx 2 Bo*oMips to start (note initial shift), and will 224 * still work even if initially too large, it will just take slightly longer 225 */ 226 unsigned long loops_per_jiffy = (1<<12); 227 EXPORT_SYMBOL(loops_per_jiffy); 228 229 static int __init debug_kernel(char *str) 230 { 231 console_loglevel = CONSOLE_LOGLEVEL_DEBUG; 232 return 0; 233 } 234 235 static int __init quiet_kernel(char *str) 236 { 237 console_loglevel = CONSOLE_LOGLEVEL_QUIET; 238 return 0; 239 } 240 241 early_param("debug", debug_kernel); 242 early_param("quiet", quiet_kernel); 243 244 static int __init loglevel(char *str) 245 { 246 int newlevel; 247 248 /* 249 * Only update loglevel value when a correct setting was passed, 250 * to prevent blind crashes (when loglevel being set to 0) that 251 * are quite hard to debug 252 */ 253 if (get_option(&str, &newlevel)) { 254 console_loglevel = newlevel; 255 return 0; 256 } 257 258 return -EINVAL; 259 } 260 261 early_param("loglevel", loglevel); 262 263 #ifdef CONFIG_BLK_DEV_INITRD 264 static void * __init get_boot_config_from_initrd(size_t *_size) 265 { 266 u32 size, csum; 267 char *data; 268 u32 *hdr; 269 int i; 270 271 if (!initrd_end) 272 return NULL; 273 274 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN; 275 /* 276 * Since Grub may align the size of initrd to 4, we must 277 * check the preceding 3 bytes as well. 278 */ 279 for (i = 0; i < 4; i++) { 280 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN)) 281 goto found; 282 data--; 283 } 284 return NULL; 285 286 found: 287 hdr = (u32 *)(data - 8); 288 size = le32_to_cpu(hdr[0]); 289 csum = le32_to_cpu(hdr[1]); 290 291 data = ((void *)hdr) - size; 292 if ((unsigned long)data < initrd_start) { 293 pr_err("bootconfig size %d is greater than initrd size %ld\n", 294 size, initrd_end - initrd_start); 295 return NULL; 296 } 297 298 if (xbc_calc_checksum(data, size) != csum) { 299 pr_err("bootconfig checksum failed\n"); 300 return NULL; 301 } 302 303 /* Remove bootconfig from initramfs/initrd */ 304 initrd_end = (unsigned long)data; 305 if (_size) 306 *_size = size; 307 308 return data; 309 } 310 #else 311 static void * __init get_boot_config_from_initrd(size_t *_size) 312 { 313 return NULL; 314 } 315 #endif 316 317 #ifdef CONFIG_BOOT_CONFIG 318 319 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata; 320 321 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0) 322 323 static int __init xbc_snprint_cmdline(char *buf, size_t size, 324 struct xbc_node *root) 325 { 326 struct xbc_node *knode, *vnode; 327 char *end = buf + size; 328 const char *val; 329 int ret; 330 331 xbc_node_for_each_key_value(root, knode, val) { 332 ret = xbc_node_compose_key_after(root, knode, 333 xbc_namebuf, XBC_KEYLEN_MAX); 334 if (ret < 0) 335 return ret; 336 337 vnode = xbc_node_get_child(knode); 338 if (!vnode) { 339 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf); 340 if (ret < 0) 341 return ret; 342 buf += ret; 343 continue; 344 } 345 xbc_array_for_each_value(vnode, val) { 346 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ", 347 xbc_namebuf, val); 348 if (ret < 0) 349 return ret; 350 buf += ret; 351 } 352 } 353 354 return buf - (end - size); 355 } 356 #undef rest 357 358 /* Make an extra command line under given key word */ 359 static char * __init xbc_make_cmdline(const char *key) 360 { 361 struct xbc_node *root; 362 char *new_cmdline; 363 int ret, len = 0; 364 365 root = xbc_find_node(key); 366 if (!root) 367 return NULL; 368 369 /* Count required buffer size */ 370 len = xbc_snprint_cmdline(NULL, 0, root); 371 if (len <= 0) 372 return NULL; 373 374 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES); 375 if (!new_cmdline) { 376 pr_err("Failed to allocate memory for extra kernel cmdline.\n"); 377 return NULL; 378 } 379 380 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root); 381 if (ret < 0 || ret > len) { 382 pr_err("Failed to print extra kernel cmdline.\n"); 383 memblock_free(new_cmdline, len + 1); 384 return NULL; 385 } 386 387 return new_cmdline; 388 } 389 390 static int __init bootconfig_params(char *param, char *val, 391 const char *unused, void *arg) 392 { 393 if (strcmp(param, "bootconfig") == 0) { 394 bootconfig_found = true; 395 } 396 return 0; 397 } 398 399 static int __init warn_bootconfig(char *str) 400 { 401 /* The 'bootconfig' has been handled by bootconfig_params(). */ 402 return 0; 403 } 404 405 static void __init setup_boot_config(void) 406 { 407 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 408 const char *msg, *data; 409 int pos, ret; 410 size_t size; 411 char *err; 412 413 /* Cut out the bootconfig data even if we have no bootconfig option */ 414 data = get_boot_config_from_initrd(&size); 415 /* If there is no bootconfig in initrd, try embedded one. */ 416 if (!data) 417 data = xbc_get_embedded_bootconfig(&size); 418 419 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 420 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL, 421 bootconfig_params); 422 423 if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE))) 424 return; 425 426 /* parse_args() stops at the next param of '--' and returns an address */ 427 if (err) 428 initargs_offs = err - tmp_cmdline; 429 430 if (!data) { 431 /* If user intended to use bootconfig, show an error level message */ 432 if (bootconfig_found) 433 pr_err("'bootconfig' found on command line, but no bootconfig found\n"); 434 else 435 pr_info("No bootconfig data provided, so skipping bootconfig"); 436 return; 437 } 438 439 if (size >= XBC_DATA_MAX) { 440 pr_err("bootconfig size %ld greater than max size %d\n", 441 (long)size, XBC_DATA_MAX); 442 return; 443 } 444 445 ret = xbc_init(data, size, &msg, &pos); 446 if (ret < 0) { 447 if (pos < 0) 448 pr_err("Failed to init bootconfig: %s.\n", msg); 449 else 450 pr_err("Failed to parse bootconfig: %s at %d.\n", 451 msg, pos); 452 } else { 453 xbc_get_info(&ret, NULL); 454 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret); 455 /* keys starting with "kernel." are passed via cmdline */ 456 extra_command_line = xbc_make_cmdline("kernel"); 457 /* Also, "init." keys are init arguments */ 458 extra_init_args = xbc_make_cmdline("init"); 459 } 460 return; 461 } 462 463 static void __init exit_boot_config(void) 464 { 465 xbc_exit(); 466 } 467 468 #else /* !CONFIG_BOOT_CONFIG */ 469 470 static void __init setup_boot_config(void) 471 { 472 /* Remove bootconfig data from initrd */ 473 get_boot_config_from_initrd(NULL); 474 } 475 476 static int __init warn_bootconfig(char *str) 477 { 478 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n"); 479 return 0; 480 } 481 482 #define exit_boot_config() do {} while (0) 483 484 #endif /* CONFIG_BOOT_CONFIG */ 485 486 early_param("bootconfig", warn_bootconfig); 487 488 /* Change NUL term back to "=", to make "param" the whole string. */ 489 static void __init repair_env_string(char *param, char *val) 490 { 491 if (val) { 492 /* param=val or param="val"? */ 493 if (val == param+strlen(param)+1) 494 val[-1] = '='; 495 else if (val == param+strlen(param)+2) { 496 val[-2] = '='; 497 memmove(val-1, val, strlen(val)+1); 498 } else 499 BUG(); 500 } 501 } 502 503 /* Anything after -- gets handed straight to init. */ 504 static int __init set_init_arg(char *param, char *val, 505 const char *unused, void *arg) 506 { 507 unsigned int i; 508 509 if (panic_later) 510 return 0; 511 512 repair_env_string(param, val); 513 514 for (i = 0; argv_init[i]; i++) { 515 if (i == MAX_INIT_ARGS) { 516 panic_later = "init"; 517 panic_param = param; 518 return 0; 519 } 520 } 521 argv_init[i] = param; 522 return 0; 523 } 524 525 /* 526 * Unknown boot options get handed to init, unless they look like 527 * unused parameters (modprobe will find them in /proc/cmdline). 528 */ 529 static int __init unknown_bootoption(char *param, char *val, 530 const char *unused, void *arg) 531 { 532 size_t len = strlen(param); 533 534 /* Handle params aliased to sysctls */ 535 if (sysctl_is_alias(param)) 536 return 0; 537 538 repair_env_string(param, val); 539 540 /* Handle obsolete-style parameters */ 541 if (obsolete_checksetup(param)) 542 return 0; 543 544 /* Unused module parameter. */ 545 if (strnchr(param, len, '.')) 546 return 0; 547 548 if (panic_later) 549 return 0; 550 551 if (val) { 552 /* Environment option */ 553 unsigned int i; 554 for (i = 0; envp_init[i]; i++) { 555 if (i == MAX_INIT_ENVS) { 556 panic_later = "env"; 557 panic_param = param; 558 } 559 if (!strncmp(param, envp_init[i], len+1)) 560 break; 561 } 562 envp_init[i] = param; 563 } else { 564 /* Command line option */ 565 unsigned int i; 566 for (i = 0; argv_init[i]; i++) { 567 if (i == MAX_INIT_ARGS) { 568 panic_later = "init"; 569 panic_param = param; 570 } 571 } 572 argv_init[i] = param; 573 } 574 return 0; 575 } 576 577 static int __init init_setup(char *str) 578 { 579 unsigned int i; 580 581 execute_command = str; 582 /* 583 * In case LILO is going to boot us with default command line, 584 * it prepends "auto" before the whole cmdline which makes 585 * the shell think it should execute a script with such name. 586 * So we ignore all arguments entered _before_ init=... [MJ] 587 */ 588 for (i = 1; i < MAX_INIT_ARGS; i++) 589 argv_init[i] = NULL; 590 return 1; 591 } 592 __setup("init=", init_setup); 593 594 static int __init rdinit_setup(char *str) 595 { 596 unsigned int i; 597 598 ramdisk_execute_command = str; 599 /* See "auto" comment in init_setup */ 600 for (i = 1; i < MAX_INIT_ARGS; i++) 601 argv_init[i] = NULL; 602 return 1; 603 } 604 __setup("rdinit=", rdinit_setup); 605 606 #ifndef CONFIG_SMP 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 saved_command_line_len = strlen(saved_command_line); 671 } 672 673 /* 674 * We need to finalize in a non-__init function or else race conditions 675 * between the root thread and the init thread may cause start_kernel to 676 * be reaped by free_initmem before the root thread has proceeded to 677 * cpu_idle. 678 * 679 * gcc-3.4 accidentally inlines this function, so use noinline. 680 */ 681 682 static __initdata DECLARE_COMPLETION(kthreadd_done); 683 684 noinline void __ref __noreturn rest_init(void) 685 { 686 struct task_struct *tsk; 687 int pid; 688 689 rcu_scheduler_starting(); 690 /* 691 * We need to spawn init first so that it obtains pid 1, however 692 * the init task will end up wanting to create kthreads, which, if 693 * we schedule it before we create kthreadd, will OOPS. 694 */ 695 pid = user_mode_thread(kernel_init, NULL, CLONE_FS); 696 /* 697 * Pin init on the boot CPU. Task migration is not properly working 698 * until sched_init_smp() has been run. It will set the allowed 699 * CPUs for init to the non isolated CPUs. 700 */ 701 rcu_read_lock(); 702 tsk = find_task_by_pid_ns(pid, &init_pid_ns); 703 tsk->flags |= PF_NO_SETAFFINITY; 704 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id())); 705 rcu_read_unlock(); 706 707 numa_default_policy(); 708 pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES); 709 rcu_read_lock(); 710 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); 711 rcu_read_unlock(); 712 713 /* 714 * Enable might_sleep() and smp_processor_id() checks. 715 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y 716 * kernel_thread() would trigger might_sleep() splats. With 717 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled 718 * already, but it's stuck on the kthreadd_done completion. 719 */ 720 system_state = SYSTEM_SCHEDULING; 721 722 complete(&kthreadd_done); 723 724 /* 725 * The boot idle thread must execute schedule() 726 * at least once to get things moving: 727 */ 728 schedule_preempt_disabled(); 729 /* Call into cpu_idle with preempt disabled */ 730 cpu_startup_entry(CPUHP_ONLINE); 731 } 732 733 /* Check for early params. */ 734 static int __init do_early_param(char *param, char *val, 735 const char *unused, void *arg) 736 { 737 const struct obs_kernel_param *p; 738 739 for (p = __setup_start; p < __setup_end; p++) { 740 if ((p->early && parameq(param, p->str)) || 741 (strcmp(param, "console") == 0 && 742 strcmp(p->str, "earlycon") == 0) 743 ) { 744 if (p->setup_func(val) != 0) 745 pr_warn("Malformed early option '%s'\n", param); 746 } 747 } 748 /* We accept everything at this stage. */ 749 return 0; 750 } 751 752 void __init parse_early_options(char *cmdline) 753 { 754 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL, 755 do_early_param); 756 } 757 758 /* Arch code calls this early on, or if not, just before other parsing. */ 759 void __init parse_early_param(void) 760 { 761 static int done __initdata; 762 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 763 764 if (done) 765 return; 766 767 /* All fall through to do_early_param. */ 768 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 769 parse_early_options(tmp_cmdline); 770 done = 1; 771 } 772 773 void __init __weak arch_post_acpi_subsys_init(void) { } 774 775 void __init __weak smp_setup_processor_id(void) 776 { 777 } 778 779 void __init __weak smp_prepare_boot_cpu(void) 780 { 781 } 782 783 # if THREAD_SIZE >= PAGE_SIZE 784 void __init __weak thread_stack_cache_init(void) 785 { 786 } 787 #endif 788 789 void __init __weak poking_init(void) { } 790 791 void __init __weak pgtable_cache_init(void) { } 792 793 void __init __weak trap_init(void) { } 794 795 bool initcall_debug; 796 core_param(initcall_debug, initcall_debug, bool, 0644); 797 798 #ifdef TRACEPOINTS_ENABLED 799 static void __init initcall_debug_enable(void); 800 #else 801 static inline void initcall_debug_enable(void) 802 { 803 } 804 #endif 805 806 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET 807 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT, 808 randomize_kstack_offset); 809 DEFINE_PER_CPU(u32, kstack_offset); 810 811 static int __init early_randomize_kstack_offset(char *buf) 812 { 813 int ret; 814 bool bool_result; 815 816 ret = kstrtobool(buf, &bool_result); 817 if (ret) 818 return ret; 819 820 if (bool_result) 821 static_branch_enable(&randomize_kstack_offset); 822 else 823 static_branch_disable(&randomize_kstack_offset); 824 return 0; 825 } 826 early_param("randomize_kstack_offset", early_randomize_kstack_offset); 827 #endif 828 829 void __init __weak __noreturn arch_call_rest_init(void) 830 { 831 rest_init(); 832 } 833 834 static void __init print_unknown_bootoptions(void) 835 { 836 char *unknown_options; 837 char *end; 838 const char *const *p; 839 size_t len; 840 841 if (panic_later || (!argv_init[1] && !envp_init[2])) 842 return; 843 844 /* 845 * Determine how many options we have to print out, plus a space 846 * before each 847 */ 848 len = 1; /* null terminator */ 849 for (p = &argv_init[1]; *p; p++) { 850 len++; 851 len += strlen(*p); 852 } 853 for (p = &envp_init[2]; *p; p++) { 854 len++; 855 len += strlen(*p); 856 } 857 858 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES); 859 if (!unknown_options) { 860 pr_err("%s: Failed to allocate %zu bytes\n", 861 __func__, len); 862 return; 863 } 864 end = unknown_options; 865 866 for (p = &argv_init[1]; *p; p++) 867 end += sprintf(end, " %s", *p); 868 for (p = &envp_init[2]; *p; p++) 869 end += sprintf(end, " %s", *p); 870 871 /* Start at unknown_options[1] to skip the initial space */ 872 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n", 873 &unknown_options[1]); 874 memblock_free(unknown_options, len); 875 } 876 877 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector 878 void start_kernel(void) 879 { 880 char *command_line; 881 char *after_dashes; 882 883 set_task_stack_end_magic(&init_task); 884 smp_setup_processor_id(); 885 debug_objects_early_init(); 886 init_vmlinux_build_id(); 887 888 cgroup_init_early(); 889 890 local_irq_disable(); 891 early_boot_irqs_disabled = true; 892 893 /* 894 * Interrupts are still disabled. Do necessary setups, then 895 * enable them. 896 */ 897 boot_cpu_init(); 898 page_address_init(); 899 pr_notice("%s", linux_banner); 900 early_security_init(); 901 setup_arch(&command_line); 902 setup_boot_config(); 903 setup_command_line(command_line); 904 setup_nr_cpu_ids(); 905 setup_per_cpu_areas(); 906 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ 907 boot_cpu_hotplug_init(); 908 909 pr_notice("Kernel command line: %s\n", saved_command_line); 910 /* parameters may set static keys */ 911 jump_label_init(); 912 parse_early_param(); 913 after_dashes = parse_args("Booting kernel", 914 static_command_line, __start___param, 915 __stop___param - __start___param, 916 -1, -1, NULL, &unknown_bootoption); 917 print_unknown_bootoptions(); 918 if (!IS_ERR_OR_NULL(after_dashes)) 919 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1, 920 NULL, set_init_arg); 921 if (extra_init_args) 922 parse_args("Setting extra init args", extra_init_args, 923 NULL, 0, -1, -1, NULL, set_init_arg); 924 925 /* Architectural and non-timekeeping rng init, before allocator init */ 926 random_init_early(command_line); 927 928 /* 929 * These use large bootmem allocations and must precede 930 * initalization of page allocator 931 */ 932 setup_log_buf(0); 933 vfs_caches_init_early(); 934 sort_main_extable(); 935 trap_init(); 936 mm_core_init(); 937 poking_init(); 938 ftrace_init(); 939 940 /* trace_printk can be enabled here */ 941 early_trace_init(); 942 943 /* 944 * Set up the scheduler prior starting any interrupts (such as the 945 * timer interrupt). Full topology setup happens at smp_init() 946 * time - but meanwhile we still have a functioning scheduler. 947 */ 948 sched_init(); 949 950 if (WARN(!irqs_disabled(), 951 "Interrupts were enabled *very* early, fixing it\n")) 952 local_irq_disable(); 953 radix_tree_init(); 954 maple_tree_init(); 955 956 /* 957 * Set up housekeeping before setting up workqueues to allow the unbound 958 * workqueue to take non-housekeeping into account. 959 */ 960 housekeeping_init(); 961 962 /* 963 * Allow workqueue creation and work item queueing/cancelling 964 * early. Work item execution depends on kthreads and starts after 965 * workqueue_init(). 966 */ 967 workqueue_init_early(); 968 969 rcu_init(); 970 971 /* Trace events are available after this */ 972 trace_init(); 973 974 if (initcall_debug) 975 initcall_debug_enable(); 976 977 context_tracking_init(); 978 /* init some links before init_ISA_irqs() */ 979 early_irq_init(); 980 init_IRQ(); 981 tick_init(); 982 rcu_init_nohz(); 983 init_timers(); 984 srcu_init(); 985 hrtimers_init(); 986 softirq_init(); 987 timekeeping_init(); 988 time_init(); 989 990 /* This must be after timekeeping is initialized */ 991 random_init(); 992 993 /* These make use of the fully initialized rng */ 994 kfence_init(); 995 boot_init_stack_canary(); 996 997 perf_event_init(); 998 profile_init(); 999 call_function_init(); 1000 WARN(!irqs_disabled(), "Interrupts were enabled early\n"); 1001 1002 early_boot_irqs_disabled = false; 1003 local_irq_enable(); 1004 1005 kmem_cache_init_late(); 1006 1007 /* 1008 * HACK ALERT! This is early. We're enabling the console before 1009 * we've done PCI setups etc, and console_init() must be aware of 1010 * this. But we do want output early, in case something goes wrong. 1011 */ 1012 console_init(); 1013 if (panic_later) 1014 panic("Too many boot %s vars at `%s'", panic_later, 1015 panic_param); 1016 1017 lockdep_init(); 1018 1019 /* 1020 * Need to run this when irqs are enabled, because it wants 1021 * to self-test [hard/soft]-irqs on/off lock inversion bugs 1022 * too: 1023 */ 1024 locking_selftest(); 1025 1026 #ifdef CONFIG_BLK_DEV_INITRD 1027 if (initrd_start && !initrd_below_start_ok && 1028 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { 1029 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", 1030 page_to_pfn(virt_to_page((void *)initrd_start)), 1031 min_low_pfn); 1032 initrd_start = 0; 1033 } 1034 #endif 1035 setup_per_cpu_pageset(); 1036 numa_policy_init(); 1037 acpi_early_init(); 1038 if (late_time_init) 1039 late_time_init(); 1040 sched_clock_init(); 1041 calibrate_delay(); 1042 1043 arch_cpu_finalize_init(); 1044 1045 pid_idr_init(); 1046 anon_vma_init(); 1047 #ifdef CONFIG_X86 1048 if (efi_enabled(EFI_RUNTIME_SERVICES)) 1049 efi_enter_virtual_mode(); 1050 #endif 1051 thread_stack_cache_init(); 1052 cred_init(); 1053 fork_init(); 1054 proc_caches_init(); 1055 uts_ns_init(); 1056 key_init(); 1057 security_init(); 1058 dbg_late_init(); 1059 net_ns_init(); 1060 vfs_caches_init(); 1061 pagecache_init(); 1062 signals_init(); 1063 seq_file_init(); 1064 proc_root_init(); 1065 nsfs_init(); 1066 pidfs_init(); 1067 cpuset_init(); 1068 cgroup_init(); 1069 taskstats_init_early(); 1070 delayacct_init(); 1071 1072 acpi_subsystem_init(); 1073 arch_post_acpi_subsys_init(); 1074 kcsan_init(); 1075 1076 /* Do the rest non-__init'ed, we're now alive */ 1077 arch_call_rest_init(); 1078 1079 /* 1080 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10 1081 * and older. 1082 */ 1083 #if !__has_attribute(__no_stack_protector__) 1084 prevent_tail_call_optimization(); 1085 #endif 1086 } 1087 1088 /* Call all constructor functions linked into the kernel. */ 1089 static void __init do_ctors(void) 1090 { 1091 /* 1092 * For UML, the constructors have already been called by the 1093 * normal setup code as it's just a normal ELF binary, so we 1094 * cannot do it again - but we do need CONFIG_CONSTRUCTORS 1095 * even on UML for modules. 1096 */ 1097 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML) 1098 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; 1099 1100 for (; fn < (ctor_fn_t *) __ctors_end; fn++) 1101 (*fn)(); 1102 #endif 1103 } 1104 1105 #ifdef CONFIG_KALLSYMS 1106 struct blacklist_entry { 1107 struct list_head next; 1108 char *buf; 1109 }; 1110 1111 static __initdata_or_module LIST_HEAD(blacklisted_initcalls); 1112 1113 static int __init initcall_blacklist(char *str) 1114 { 1115 char *str_entry; 1116 struct blacklist_entry *entry; 1117 1118 /* str argument is a comma-separated list of functions */ 1119 do { 1120 str_entry = strsep(&str, ","); 1121 if (str_entry) { 1122 pr_debug("blacklisting initcall %s\n", str_entry); 1123 entry = memblock_alloc(sizeof(*entry), 1124 SMP_CACHE_BYTES); 1125 if (!entry) 1126 panic("%s: Failed to allocate %zu bytes\n", 1127 __func__, sizeof(*entry)); 1128 entry->buf = memblock_alloc(strlen(str_entry) + 1, 1129 SMP_CACHE_BYTES); 1130 if (!entry->buf) 1131 panic("%s: Failed to allocate %zu bytes\n", 1132 __func__, strlen(str_entry) + 1); 1133 strcpy(entry->buf, str_entry); 1134 list_add(&entry->next, &blacklisted_initcalls); 1135 } 1136 } while (str_entry); 1137 1138 return 1; 1139 } 1140 1141 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1142 { 1143 struct blacklist_entry *entry; 1144 char fn_name[KSYM_SYMBOL_LEN]; 1145 unsigned long addr; 1146 1147 if (list_empty(&blacklisted_initcalls)) 1148 return false; 1149 1150 addr = (unsigned long) dereference_function_descriptor(fn); 1151 sprint_symbol_no_offset(fn_name, addr); 1152 1153 /* 1154 * fn will be "function_name [module_name]" where [module_name] is not 1155 * displayed for built-in init functions. Strip off the [module_name]. 1156 */ 1157 strreplace(fn_name, ' ', '\0'); 1158 1159 list_for_each_entry(entry, &blacklisted_initcalls, next) { 1160 if (!strcmp(fn_name, entry->buf)) { 1161 pr_debug("initcall %s blacklisted\n", fn_name); 1162 return true; 1163 } 1164 } 1165 1166 return false; 1167 } 1168 #else 1169 static int __init initcall_blacklist(char *str) 1170 { 1171 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); 1172 return 0; 1173 } 1174 1175 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1176 { 1177 return false; 1178 } 1179 #endif 1180 __setup("initcall_blacklist=", initcall_blacklist); 1181 1182 static __init_or_module void 1183 trace_initcall_start_cb(void *data, initcall_t fn) 1184 { 1185 ktime_t *calltime = data; 1186 1187 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current)); 1188 *calltime = ktime_get(); 1189 } 1190 1191 static __init_or_module void 1192 trace_initcall_finish_cb(void *data, initcall_t fn, int ret) 1193 { 1194 ktime_t rettime, *calltime = data; 1195 1196 rettime = ktime_get(); 1197 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n", 1198 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime)); 1199 } 1200 1201 static ktime_t initcall_calltime; 1202 1203 #ifdef TRACEPOINTS_ENABLED 1204 static void __init initcall_debug_enable(void) 1205 { 1206 int ret; 1207 1208 ret = register_trace_initcall_start(trace_initcall_start_cb, 1209 &initcall_calltime); 1210 ret |= register_trace_initcall_finish(trace_initcall_finish_cb, 1211 &initcall_calltime); 1212 WARN(ret, "Failed to register initcall tracepoints\n"); 1213 } 1214 # define do_trace_initcall_start trace_initcall_start 1215 # define do_trace_initcall_finish trace_initcall_finish 1216 #else 1217 static inline void do_trace_initcall_start(initcall_t fn) 1218 { 1219 if (!initcall_debug) 1220 return; 1221 trace_initcall_start_cb(&initcall_calltime, fn); 1222 } 1223 static inline void do_trace_initcall_finish(initcall_t fn, int ret) 1224 { 1225 if (!initcall_debug) 1226 return; 1227 trace_initcall_finish_cb(&initcall_calltime, fn, ret); 1228 } 1229 #endif /* !TRACEPOINTS_ENABLED */ 1230 1231 int __init_or_module do_one_initcall(initcall_t fn) 1232 { 1233 int count = preempt_count(); 1234 char msgbuf[64]; 1235 int ret; 1236 1237 if (initcall_blacklisted(fn)) 1238 return -EPERM; 1239 1240 do_trace_initcall_start(fn); 1241 ret = fn(); 1242 do_trace_initcall_finish(fn, ret); 1243 1244 msgbuf[0] = 0; 1245 1246 if (preempt_count() != count) { 1247 sprintf(msgbuf, "preemption imbalance "); 1248 preempt_count_set(count); 1249 } 1250 if (irqs_disabled()) { 1251 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); 1252 local_irq_enable(); 1253 } 1254 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf); 1255 1256 add_latent_entropy(); 1257 return ret; 1258 } 1259 1260 1261 static initcall_entry_t *initcall_levels[] __initdata = { 1262 __initcall0_start, 1263 __initcall1_start, 1264 __initcall2_start, 1265 __initcall3_start, 1266 __initcall4_start, 1267 __initcall5_start, 1268 __initcall6_start, 1269 __initcall7_start, 1270 __initcall_end, 1271 }; 1272 1273 /* Keep these in sync with initcalls in include/linux/init.h */ 1274 static const char *initcall_level_names[] __initdata = { 1275 "pure", 1276 "core", 1277 "postcore", 1278 "arch", 1279 "subsys", 1280 "fs", 1281 "device", 1282 "late", 1283 }; 1284 1285 static int __init ignore_unknown_bootoption(char *param, char *val, 1286 const char *unused, void *arg) 1287 { 1288 return 0; 1289 } 1290 1291 static void __init do_initcall_level(int level, char *command_line) 1292 { 1293 initcall_entry_t *fn; 1294 1295 parse_args(initcall_level_names[level], 1296 command_line, __start___param, 1297 __stop___param - __start___param, 1298 level, level, 1299 NULL, ignore_unknown_bootoption); 1300 1301 trace_initcall_level(initcall_level_names[level]); 1302 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) 1303 do_one_initcall(initcall_from_entry(fn)); 1304 } 1305 1306 static void __init do_initcalls(void) 1307 { 1308 int level; 1309 size_t len = saved_command_line_len + 1; 1310 char *command_line; 1311 1312 command_line = kzalloc(len, GFP_KERNEL); 1313 if (!command_line) 1314 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 1315 1316 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) { 1317 /* Parser modifies command_line, restore it each time */ 1318 strcpy(command_line, saved_command_line); 1319 do_initcall_level(level, command_line); 1320 } 1321 1322 kfree(command_line); 1323 } 1324 1325 /* 1326 * Ok, the machine is now initialized. None of the devices 1327 * have been touched yet, but the CPU subsystem is up and 1328 * running, and memory and process management works. 1329 * 1330 * Now we can finally start doing some real work.. 1331 */ 1332 static void __init do_basic_setup(void) 1333 { 1334 cpuset_init_smp(); 1335 driver_init(); 1336 init_irq_proc(); 1337 do_ctors(); 1338 do_initcalls(); 1339 } 1340 1341 static void __init do_pre_smp_initcalls(void) 1342 { 1343 initcall_entry_t *fn; 1344 1345 trace_initcall_level("early"); 1346 for (fn = __initcall_start; fn < __initcall0_start; fn++) 1347 do_one_initcall(initcall_from_entry(fn)); 1348 } 1349 1350 static int run_init_process(const char *init_filename) 1351 { 1352 const char *const *p; 1353 1354 argv_init[0] = init_filename; 1355 pr_info("Run %s as init process\n", init_filename); 1356 pr_debug(" with arguments:\n"); 1357 for (p = argv_init; *p; p++) 1358 pr_debug(" %s\n", *p); 1359 pr_debug(" with environment:\n"); 1360 for (p = envp_init; *p; p++) 1361 pr_debug(" %s\n", *p); 1362 return kernel_execve(init_filename, argv_init, envp_init); 1363 } 1364 1365 static int try_to_run_init_process(const char *init_filename) 1366 { 1367 int ret; 1368 1369 ret = run_init_process(init_filename); 1370 1371 if (ret && ret != -ENOENT) { 1372 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", 1373 init_filename, ret); 1374 } 1375 1376 return ret; 1377 } 1378 1379 static noinline void __init kernel_init_freeable(void); 1380 1381 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX) 1382 bool rodata_enabled __ro_after_init = true; 1383 1384 #ifndef arch_parse_debug_rodata 1385 static inline bool arch_parse_debug_rodata(char *str) { return false; } 1386 #endif 1387 1388 static int __init set_debug_rodata(char *str) 1389 { 1390 if (arch_parse_debug_rodata(str)) 1391 return 0; 1392 1393 if (str && !strcmp(str, "on")) 1394 rodata_enabled = true; 1395 else if (str && !strcmp(str, "off")) 1396 rodata_enabled = false; 1397 else 1398 pr_warn("Invalid option string for rodata: '%s'\n", str); 1399 return 0; 1400 } 1401 early_param("rodata", set_debug_rodata); 1402 #endif 1403 1404 #ifdef CONFIG_STRICT_KERNEL_RWX 1405 static void mark_readonly(void) 1406 { 1407 if (rodata_enabled) { 1408 /* 1409 * load_module() results in W+X mappings, which are cleaned 1410 * up with call_rcu(). Let's make sure that queued work is 1411 * flushed so that we don't hit false positives looking for 1412 * insecure pages which are W+X. 1413 */ 1414 rcu_barrier(); 1415 mark_rodata_ro(); 1416 rodata_test(); 1417 } else 1418 pr_info("Kernel memory protection disabled.\n"); 1419 } 1420 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX) 1421 static inline void mark_readonly(void) 1422 { 1423 pr_warn("Kernel memory protection not selected by kernel config.\n"); 1424 } 1425 #else 1426 static inline void mark_readonly(void) 1427 { 1428 pr_warn("This architecture does not have kernel memory protection.\n"); 1429 } 1430 #endif 1431 1432 void __weak free_initmem(void) 1433 { 1434 free_initmem_default(POISON_FREE_INITMEM); 1435 } 1436 1437 static int __ref kernel_init(void *unused) 1438 { 1439 int ret; 1440 1441 /* 1442 * Wait until kthreadd is all set-up. 1443 */ 1444 wait_for_completion(&kthreadd_done); 1445 1446 kernel_init_freeable(); 1447 /* need to finish all async __init code before freeing the memory */ 1448 async_synchronize_full(); 1449 1450 system_state = SYSTEM_FREEING_INITMEM; 1451 kprobe_free_init_mem(); 1452 ftrace_free_init_mem(); 1453 kgdb_free_init_mem(); 1454 exit_boot_config(); 1455 free_initmem(); 1456 mark_readonly(); 1457 1458 /* 1459 * Kernel mappings are now finalized - update the userspace page-table 1460 * to finalize PTI. 1461 */ 1462 pti_finalize(); 1463 1464 system_state = SYSTEM_RUNNING; 1465 numa_default_policy(); 1466 1467 rcu_end_inkernel_boot(); 1468 1469 do_sysctl_args(); 1470 1471 if (ramdisk_execute_command) { 1472 ret = run_init_process(ramdisk_execute_command); 1473 if (!ret) 1474 return 0; 1475 pr_err("Failed to execute %s (error %d)\n", 1476 ramdisk_execute_command, ret); 1477 } 1478 1479 /* 1480 * We try each of these until one succeeds. 1481 * 1482 * The Bourne shell can be used instead of init if we are 1483 * trying to recover a really broken machine. 1484 */ 1485 if (execute_command) { 1486 ret = run_init_process(execute_command); 1487 if (!ret) 1488 return 0; 1489 panic("Requested init %s failed (error %d).", 1490 execute_command, ret); 1491 } 1492 1493 if (CONFIG_DEFAULT_INIT[0] != '\0') { 1494 ret = run_init_process(CONFIG_DEFAULT_INIT); 1495 if (ret) 1496 pr_err("Default init %s failed (error %d)\n", 1497 CONFIG_DEFAULT_INIT, ret); 1498 else 1499 return 0; 1500 } 1501 1502 if (!try_to_run_init_process("/sbin/init") || 1503 !try_to_run_init_process("/etc/init") || 1504 !try_to_run_init_process("/bin/init") || 1505 !try_to_run_init_process("/bin/sh")) 1506 return 0; 1507 1508 panic("No working init found. Try passing init= option to kernel. " 1509 "See Linux Documentation/admin-guide/init.rst for guidance."); 1510 } 1511 1512 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */ 1513 void __init console_on_rootfs(void) 1514 { 1515 struct file *file = filp_open("/dev/console", O_RDWR, 0); 1516 1517 if (IS_ERR(file)) { 1518 pr_err("Warning: unable to open an initial console.\n"); 1519 return; 1520 } 1521 init_dup(file); 1522 init_dup(file); 1523 init_dup(file); 1524 fput(file); 1525 } 1526 1527 static noinline void __init kernel_init_freeable(void) 1528 { 1529 /* Now the scheduler is fully set up and can do blocking allocations */ 1530 gfp_allowed_mask = __GFP_BITS_MASK; 1531 1532 /* 1533 * init can allocate pages on any node 1534 */ 1535 set_mems_allowed(node_states[N_MEMORY]); 1536 1537 cad_pid = get_pid(task_pid(current)); 1538 1539 smp_prepare_cpus(setup_max_cpus); 1540 1541 workqueue_init(); 1542 1543 init_mm_internals(); 1544 1545 rcu_init_tasks_generic(); 1546 do_pre_smp_initcalls(); 1547 lockup_detector_init(); 1548 1549 smp_init(); 1550 sched_init_smp(); 1551 1552 workqueue_init_topology(); 1553 async_init(); 1554 padata_init(); 1555 page_alloc_init_late(); 1556 1557 do_basic_setup(); 1558 1559 kunit_run_all_tests(); 1560 1561 wait_for_initramfs(); 1562 console_on_rootfs(); 1563 1564 /* 1565 * check if there is an early userspace init. If yes, let it do all 1566 * the work 1567 */ 1568 if (init_eaccess(ramdisk_execute_command) != 0) { 1569 ramdisk_execute_command = NULL; 1570 prepare_namespace(); 1571 } 1572 1573 /* 1574 * Ok, we have completed the initial bootup, and 1575 * we're essentially up and running. Get rid of the 1576 * initmem segments and start the user-mode stuff.. 1577 * 1578 * rootfs is available now, try loading the public keys 1579 * and default modules 1580 */ 1581 1582 integrity_load_keys(); 1583 } 1584