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/vmalloc.h> 37 #include <linux/kernel_stat.h> 38 #include <linux/start_kernel.h> 39 #include <linux/security.h> 40 #include <linux/smp.h> 41 #include <linux/profile.h> 42 #include <linux/rcupdate.h> 43 #include <linux/moduleparam.h> 44 #include <linux/kallsyms.h> 45 #include <linux/writeback.h> 46 #include <linux/cpu.h> 47 #include <linux/cpuset.h> 48 #include <linux/cgroup.h> 49 #include <linux/efi.h> 50 #include <linux/tick.h> 51 #include <linux/sched/isolation.h> 52 #include <linux/interrupt.h> 53 #include <linux/taskstats_kern.h> 54 #include <linux/delayacct.h> 55 #include <linux/unistd.h> 56 #include <linux/utsname.h> 57 #include <linux/rmap.h> 58 #include <linux/mempolicy.h> 59 #include <linux/key.h> 60 #include <linux/buffer_head.h> 61 #include <linux/page_ext.h> 62 #include <linux/debug_locks.h> 63 #include <linux/debugobjects.h> 64 #include <linux/lockdep.h> 65 #include <linux/kmemleak.h> 66 #include <linux/pid_namespace.h> 67 #include <linux/device/driver.h> 68 #include <linux/kthread.h> 69 #include <linux/sched.h> 70 #include <linux/sched/init.h> 71 #include <linux/signal.h> 72 #include <linux/idr.h> 73 #include <linux/kgdb.h> 74 #include <linux/ftrace.h> 75 #include <linux/async.h> 76 #include <linux/sfi.h> 77 #include <linux/shmem_fs.h> 78 #include <linux/slab.h> 79 #include <linux/perf_event.h> 80 #include <linux/ptrace.h> 81 #include <linux/pti.h> 82 #include <linux/blkdev.h> 83 #include <linux/elevator.h> 84 #include <linux/sched/clock.h> 85 #include <linux/sched/task.h> 86 #include <linux/sched/task_stack.h> 87 #include <linux/context_tracking.h> 88 #include <linux/random.h> 89 #include <linux/list.h> 90 #include <linux/integrity.h> 91 #include <linux/proc_ns.h> 92 #include <linux/io.h> 93 #include <linux/cache.h> 94 #include <linux/rodata_test.h> 95 #include <linux/jump_label.h> 96 #include <linux/mem_encrypt.h> 97 98 #include <asm/io.h> 99 #include <asm/bugs.h> 100 #include <asm/setup.h> 101 #include <asm/sections.h> 102 #include <asm/cacheflush.h> 103 104 #define CREATE_TRACE_POINTS 105 #include <trace/events/initcall.h> 106 107 static int kernel_init(void *); 108 109 extern void init_IRQ(void); 110 extern void radix_tree_init(void); 111 112 /* 113 * Debug helper: via this flag we know that we are in 'early bootup code' 114 * where only the boot processor is running with IRQ disabled. This means 115 * two things - IRQ must not be enabled before the flag is cleared and some 116 * operations which are not allowed with IRQ disabled are allowed while the 117 * flag is set. 118 */ 119 bool early_boot_irqs_disabled __read_mostly; 120 121 enum system_states system_state __read_mostly; 122 EXPORT_SYMBOL(system_state); 123 124 /* 125 * Boot command-line arguments 126 */ 127 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT 128 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT 129 130 extern void time_init(void); 131 /* Default late time init is NULL. archs can override this later. */ 132 void (*__initdata late_time_init)(void); 133 134 /* Untouched command line saved by arch-specific code. */ 135 char __initdata boot_command_line[COMMAND_LINE_SIZE]; 136 /* Untouched saved command line (eg. for /proc) */ 137 char *saved_command_line; 138 /* Command line for parameter parsing */ 139 static char *static_command_line; 140 /* Untouched extra command line */ 141 static char *extra_command_line; 142 /* Extra init arguments */ 143 static char *extra_init_args; 144 145 #ifdef CONFIG_BOOT_CONFIG 146 /* Is bootconfig on command line? */ 147 static bool bootconfig_found; 148 static bool initargs_found; 149 #else 150 # define bootconfig_found false 151 # define initargs_found false 152 #endif 153 154 static char *execute_command; 155 static char *ramdisk_execute_command; 156 157 /* 158 * Used to generate warnings if static_key manipulation functions are used 159 * before jump_label_init is called. 160 */ 161 bool static_key_initialized __read_mostly; 162 EXPORT_SYMBOL_GPL(static_key_initialized); 163 164 /* 165 * If set, this is an indication to the drivers that reset the underlying 166 * device before going ahead with the initialization otherwise driver might 167 * rely on the BIOS and skip the reset operation. 168 * 169 * This is useful if kernel is booting in an unreliable environment. 170 * For ex. kdump situation where previous kernel has crashed, BIOS has been 171 * skipped and devices will be in unknown state. 172 */ 173 unsigned int reset_devices; 174 EXPORT_SYMBOL(reset_devices); 175 176 static int __init set_reset_devices(char *str) 177 { 178 reset_devices = 1; 179 return 1; 180 } 181 182 __setup("reset_devices", set_reset_devices); 183 184 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; 185 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; 186 static const char *panic_later, *panic_param; 187 188 extern const struct obs_kernel_param __setup_start[], __setup_end[]; 189 190 static bool __init obsolete_checksetup(char *line) 191 { 192 const struct obs_kernel_param *p; 193 bool had_early_param = false; 194 195 p = __setup_start; 196 do { 197 int n = strlen(p->str); 198 if (parameqn(line, p->str, n)) { 199 if (p->early) { 200 /* Already done in parse_early_param? 201 * (Needs exact match on param part). 202 * Keep iterating, as we can have early 203 * params and __setups of same names 8( */ 204 if (line[n] == '\0' || line[n] == '=') 205 had_early_param = true; 206 } else if (!p->setup_func) { 207 pr_warn("Parameter %s is obsolete, ignored\n", 208 p->str); 209 return true; 210 } else if (p->setup_func(line + n)) 211 return true; 212 } 213 p++; 214 } while (p < __setup_end); 215 216 return had_early_param; 217 } 218 219 /* 220 * This should be approx 2 Bo*oMips to start (note initial shift), and will 221 * still work even if initially too large, it will just take slightly longer 222 */ 223 unsigned long loops_per_jiffy = (1<<12); 224 EXPORT_SYMBOL(loops_per_jiffy); 225 226 static int __init debug_kernel(char *str) 227 { 228 console_loglevel = CONSOLE_LOGLEVEL_DEBUG; 229 return 0; 230 } 231 232 static int __init quiet_kernel(char *str) 233 { 234 console_loglevel = CONSOLE_LOGLEVEL_QUIET; 235 return 0; 236 } 237 238 early_param("debug", debug_kernel); 239 early_param("quiet", quiet_kernel); 240 241 static int __init loglevel(char *str) 242 { 243 int newlevel; 244 245 /* 246 * Only update loglevel value when a correct setting was passed, 247 * to prevent blind crashes (when loglevel being set to 0) that 248 * are quite hard to debug 249 */ 250 if (get_option(&str, &newlevel)) { 251 console_loglevel = newlevel; 252 return 0; 253 } 254 255 return -EINVAL; 256 } 257 258 early_param("loglevel", loglevel); 259 260 #ifdef CONFIG_BLK_DEV_INITRD 261 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum) 262 { 263 u32 size, csum; 264 char *data; 265 u32 *hdr; 266 267 if (!initrd_end) 268 return NULL; 269 270 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN; 271 if (memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN)) 272 return NULL; 273 274 hdr = (u32 *)(data - 8); 275 size = hdr[0]; 276 csum = hdr[1]; 277 278 data = ((void *)hdr) - size; 279 if ((unsigned long)data < initrd_start) { 280 pr_err("bootconfig size %d is greater than initrd size %ld\n", 281 size, initrd_end - initrd_start); 282 return NULL; 283 } 284 285 /* Remove bootconfig from initramfs/initrd */ 286 initrd_end = (unsigned long)data; 287 if (_size) 288 *_size = size; 289 if (_csum) 290 *_csum = csum; 291 292 return data; 293 } 294 #else 295 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum) 296 { 297 return NULL; 298 } 299 #endif 300 301 #ifdef CONFIG_BOOT_CONFIG 302 303 char xbc_namebuf[XBC_KEYLEN_MAX] __initdata; 304 305 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0) 306 307 static int __init xbc_snprint_cmdline(char *buf, size_t size, 308 struct xbc_node *root) 309 { 310 struct xbc_node *knode, *vnode; 311 char *end = buf + size; 312 const char *val; 313 int ret; 314 315 xbc_node_for_each_key_value(root, knode, val) { 316 ret = xbc_node_compose_key_after(root, knode, 317 xbc_namebuf, XBC_KEYLEN_MAX); 318 if (ret < 0) 319 return ret; 320 321 vnode = xbc_node_get_child(knode); 322 if (!vnode) { 323 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf); 324 if (ret < 0) 325 return ret; 326 buf += ret; 327 continue; 328 } 329 xbc_array_for_each_value(vnode, val) { 330 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ", 331 xbc_namebuf, val); 332 if (ret < 0) 333 return ret; 334 buf += ret; 335 } 336 } 337 338 return buf - (end - size); 339 } 340 #undef rest 341 342 /* Make an extra command line under given key word */ 343 static char * __init xbc_make_cmdline(const char *key) 344 { 345 struct xbc_node *root; 346 char *new_cmdline; 347 int ret, len = 0; 348 349 root = xbc_find_node(key); 350 if (!root) 351 return NULL; 352 353 /* Count required buffer size */ 354 len = xbc_snprint_cmdline(NULL, 0, root); 355 if (len <= 0) 356 return NULL; 357 358 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES); 359 if (!new_cmdline) { 360 pr_err("Failed to allocate memory for extra kernel cmdline.\n"); 361 return NULL; 362 } 363 364 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root); 365 if (ret < 0 || ret > len) { 366 pr_err("Failed to print extra kernel cmdline.\n"); 367 return NULL; 368 } 369 370 return new_cmdline; 371 } 372 373 static u32 boot_config_checksum(unsigned char *p, u32 size) 374 { 375 u32 ret = 0; 376 377 while (size--) 378 ret += *p++; 379 380 return ret; 381 } 382 383 static int __init bootconfig_params(char *param, char *val, 384 const char *unused, void *arg) 385 { 386 if (strcmp(param, "bootconfig") == 0) { 387 bootconfig_found = true; 388 } else if (strcmp(param, "--") == 0) { 389 initargs_found = true; 390 } 391 return 0; 392 } 393 394 static void __init setup_boot_config(const char *cmdline) 395 { 396 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 397 const char *msg; 398 int pos; 399 u32 size, csum; 400 char *data, *copy; 401 int ret; 402 403 /* Cut out the bootconfig data even if we have no bootconfig option */ 404 data = get_boot_config_from_initrd(&size, &csum); 405 406 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 407 parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL, 408 bootconfig_params); 409 410 if (!bootconfig_found) 411 return; 412 413 if (!data) { 414 pr_err("'bootconfig' found on command line, but no bootconfig found\n"); 415 return; 416 } 417 418 if (size >= XBC_DATA_MAX) { 419 pr_err("bootconfig size %d greater than max size %d\n", 420 size, XBC_DATA_MAX); 421 return; 422 } 423 424 if (boot_config_checksum((unsigned char *)data, size) != csum) { 425 pr_err("bootconfig checksum failed\n"); 426 return; 427 } 428 429 copy = memblock_alloc(size + 1, SMP_CACHE_BYTES); 430 if (!copy) { 431 pr_err("Failed to allocate memory for bootconfig\n"); 432 return; 433 } 434 435 memcpy(copy, data, size); 436 copy[size] = '\0'; 437 438 ret = xbc_init(copy, &msg, &pos); 439 if (ret < 0) { 440 if (pos < 0) 441 pr_err("Failed to init bootconfig: %s.\n", msg); 442 else 443 pr_err("Failed to parse bootconfig: %s at %d.\n", 444 msg, pos); 445 } else { 446 pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret); 447 /* keys starting with "kernel." are passed via cmdline */ 448 extra_command_line = xbc_make_cmdline("kernel"); 449 /* Also, "init." keys are init arguments */ 450 extra_init_args = xbc_make_cmdline("init"); 451 } 452 return; 453 } 454 455 #else 456 457 static void __init setup_boot_config(const char *cmdline) 458 { 459 /* Remove bootconfig data from initrd */ 460 get_boot_config_from_initrd(NULL, NULL); 461 } 462 463 static int __init warn_bootconfig(char *str) 464 { 465 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOTCONFIG is not set.\n"); 466 return 0; 467 } 468 early_param("bootconfig", warn_bootconfig); 469 470 #endif 471 472 /* Change NUL term back to "=", to make "param" the whole string. */ 473 static void __init repair_env_string(char *param, char *val) 474 { 475 if (val) { 476 /* param=val or param="val"? */ 477 if (val == param+strlen(param)+1) 478 val[-1] = '='; 479 else if (val == param+strlen(param)+2) { 480 val[-2] = '='; 481 memmove(val-1, val, strlen(val)+1); 482 } else 483 BUG(); 484 } 485 } 486 487 /* Anything after -- gets handed straight to init. */ 488 static int __init set_init_arg(char *param, char *val, 489 const char *unused, void *arg) 490 { 491 unsigned int i; 492 493 if (panic_later) 494 return 0; 495 496 repair_env_string(param, val); 497 498 for (i = 0; argv_init[i]; i++) { 499 if (i == MAX_INIT_ARGS) { 500 panic_later = "init"; 501 panic_param = param; 502 return 0; 503 } 504 } 505 argv_init[i] = param; 506 return 0; 507 } 508 509 /* 510 * Unknown boot options get handed to init, unless they look like 511 * unused parameters (modprobe will find them in /proc/cmdline). 512 */ 513 static int __init unknown_bootoption(char *param, char *val, 514 const char *unused, void *arg) 515 { 516 size_t len = strlen(param); 517 518 repair_env_string(param, val); 519 520 /* Handle obsolete-style parameters */ 521 if (obsolete_checksetup(param)) 522 return 0; 523 524 /* Unused module parameter. */ 525 if (strnchr(param, len, '.')) 526 return 0; 527 528 if (panic_later) 529 return 0; 530 531 if (val) { 532 /* Environment option */ 533 unsigned int i; 534 for (i = 0; envp_init[i]; i++) { 535 if (i == MAX_INIT_ENVS) { 536 panic_later = "env"; 537 panic_param = param; 538 } 539 if (!strncmp(param, envp_init[i], len+1)) 540 break; 541 } 542 envp_init[i] = param; 543 } else { 544 /* Command line option */ 545 unsigned int i; 546 for (i = 0; argv_init[i]; i++) { 547 if (i == MAX_INIT_ARGS) { 548 panic_later = "init"; 549 panic_param = param; 550 } 551 } 552 argv_init[i] = param; 553 } 554 return 0; 555 } 556 557 static int __init init_setup(char *str) 558 { 559 unsigned int i; 560 561 execute_command = str; 562 /* 563 * In case LILO is going to boot us with default command line, 564 * it prepends "auto" before the whole cmdline which makes 565 * the shell think it should execute a script with such name. 566 * So we ignore all arguments entered _before_ init=... [MJ] 567 */ 568 for (i = 1; i < MAX_INIT_ARGS; i++) 569 argv_init[i] = NULL; 570 return 1; 571 } 572 __setup("init=", init_setup); 573 574 static int __init rdinit_setup(char *str) 575 { 576 unsigned int i; 577 578 ramdisk_execute_command = str; 579 /* See "auto" comment in init_setup */ 580 for (i = 1; i < MAX_INIT_ARGS; i++) 581 argv_init[i] = NULL; 582 return 1; 583 } 584 __setup("rdinit=", rdinit_setup); 585 586 #ifndef CONFIG_SMP 587 static const unsigned int setup_max_cpus = NR_CPUS; 588 static inline void setup_nr_cpu_ids(void) { } 589 static inline void smp_prepare_cpus(unsigned int maxcpus) { } 590 #endif 591 592 /* 593 * We need to store the untouched command line for future reference. 594 * We also need to store the touched command line since the parameter 595 * parsing is performed in place, and we should allow a component to 596 * store reference of name/value for future reference. 597 */ 598 static void __init setup_command_line(char *command_line) 599 { 600 size_t len, xlen = 0, ilen = 0; 601 602 if (extra_command_line) 603 xlen = strlen(extra_command_line); 604 if (extra_init_args) 605 ilen = strlen(extra_init_args) + 4; /* for " -- " */ 606 607 len = xlen + strlen(boot_command_line) + 1; 608 609 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES); 610 if (!saved_command_line) 611 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen); 612 613 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES); 614 if (!static_command_line) 615 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 616 617 if (xlen) { 618 /* 619 * We have to put extra_command_line before boot command 620 * lines because there could be dashes (separator of init 621 * command line) in the command lines. 622 */ 623 strcpy(saved_command_line, extra_command_line); 624 strcpy(static_command_line, extra_command_line); 625 } 626 strcpy(saved_command_line + xlen, boot_command_line); 627 strcpy(static_command_line + xlen, command_line); 628 629 if (ilen) { 630 /* 631 * Append supplemental init boot args to saved_command_line 632 * so that user can check what command line options passed 633 * to init. 634 */ 635 len = strlen(saved_command_line); 636 if (initargs_found) { 637 saved_command_line[len++] = ' '; 638 } else { 639 strcpy(saved_command_line + len, " -- "); 640 len += 4; 641 } 642 643 strcpy(saved_command_line + len, extra_init_args); 644 } 645 } 646 647 /* 648 * We need to finalize in a non-__init function or else race conditions 649 * between the root thread and the init thread may cause start_kernel to 650 * be reaped by free_initmem before the root thread has proceeded to 651 * cpu_idle. 652 * 653 * gcc-3.4 accidentally inlines this function, so use noinline. 654 */ 655 656 static __initdata DECLARE_COMPLETION(kthreadd_done); 657 658 noinline void __ref rest_init(void) 659 { 660 struct task_struct *tsk; 661 int pid; 662 663 rcu_scheduler_starting(); 664 /* 665 * We need to spawn init first so that it obtains pid 1, however 666 * the init task will end up wanting to create kthreads, which, if 667 * we schedule it before we create kthreadd, will OOPS. 668 */ 669 pid = kernel_thread(kernel_init, NULL, CLONE_FS); 670 /* 671 * Pin init on the boot CPU. Task migration is not properly working 672 * until sched_init_smp() has been run. It will set the allowed 673 * CPUs for init to the non isolated CPUs. 674 */ 675 rcu_read_lock(); 676 tsk = find_task_by_pid_ns(pid, &init_pid_ns); 677 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id())); 678 rcu_read_unlock(); 679 680 numa_default_policy(); 681 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); 682 rcu_read_lock(); 683 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); 684 rcu_read_unlock(); 685 686 /* 687 * Enable might_sleep() and smp_processor_id() checks. 688 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y 689 * kernel_thread() would trigger might_sleep() splats. With 690 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled 691 * already, but it's stuck on the kthreadd_done completion. 692 */ 693 system_state = SYSTEM_SCHEDULING; 694 695 complete(&kthreadd_done); 696 697 /* 698 * The boot idle thread must execute schedule() 699 * at least once to get things moving: 700 */ 701 schedule_preempt_disabled(); 702 /* Call into cpu_idle with preempt disabled */ 703 cpu_startup_entry(CPUHP_ONLINE); 704 } 705 706 /* Check for early params. */ 707 static int __init do_early_param(char *param, char *val, 708 const char *unused, void *arg) 709 { 710 const struct obs_kernel_param *p; 711 712 for (p = __setup_start; p < __setup_end; p++) { 713 if ((p->early && parameq(param, p->str)) || 714 (strcmp(param, "console") == 0 && 715 strcmp(p->str, "earlycon") == 0) 716 ) { 717 if (p->setup_func(val) != 0) 718 pr_warn("Malformed early option '%s'\n", param); 719 } 720 } 721 /* We accept everything at this stage. */ 722 return 0; 723 } 724 725 void __init parse_early_options(char *cmdline) 726 { 727 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL, 728 do_early_param); 729 } 730 731 /* Arch code calls this early on, or if not, just before other parsing. */ 732 void __init parse_early_param(void) 733 { 734 static int done __initdata; 735 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; 736 737 if (done) 738 return; 739 740 /* All fall through to do_early_param. */ 741 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); 742 parse_early_options(tmp_cmdline); 743 done = 1; 744 } 745 746 void __init __weak arch_post_acpi_subsys_init(void) { } 747 748 void __init __weak smp_setup_processor_id(void) 749 { 750 } 751 752 # if THREAD_SIZE >= PAGE_SIZE 753 void __init __weak thread_stack_cache_init(void) 754 { 755 } 756 #endif 757 758 void __init __weak mem_encrypt_init(void) { } 759 760 void __init __weak poking_init(void) { } 761 762 void __init __weak pgtable_cache_init(void) { } 763 764 bool initcall_debug; 765 core_param(initcall_debug, initcall_debug, bool, 0644); 766 767 #ifdef TRACEPOINTS_ENABLED 768 static void __init initcall_debug_enable(void); 769 #else 770 static inline void initcall_debug_enable(void) 771 { 772 } 773 #endif 774 775 /* Report memory auto-initialization states for this boot. */ 776 static void __init report_meminit(void) 777 { 778 const char *stack; 779 780 if (IS_ENABLED(CONFIG_INIT_STACK_ALL)) 781 stack = "all"; 782 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL)) 783 stack = "byref_all"; 784 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF)) 785 stack = "byref"; 786 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER)) 787 stack = "__user"; 788 else 789 stack = "off"; 790 791 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n", 792 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off", 793 want_init_on_free() ? "on" : "off"); 794 if (want_init_on_free()) 795 pr_info("mem auto-init: clearing system memory may take some time...\n"); 796 } 797 798 /* 799 * Set up kernel memory allocators 800 */ 801 static void __init mm_init(void) 802 { 803 /* 804 * page_ext requires contiguous pages, 805 * bigger than MAX_ORDER unless SPARSEMEM. 806 */ 807 page_ext_init_flatmem(); 808 init_debug_pagealloc(); 809 report_meminit(); 810 mem_init(); 811 kmem_cache_init(); 812 kmemleak_init(); 813 pgtable_init(); 814 debug_objects_mem_init(); 815 vmalloc_init(); 816 ioremap_huge_init(); 817 /* Should be run before the first non-init thread is created */ 818 init_espfix_bsp(); 819 /* Should be run after espfix64 is set up. */ 820 pti_init(); 821 } 822 823 void __init __weak arch_call_rest_init(void) 824 { 825 rest_init(); 826 } 827 828 asmlinkage __visible void __init start_kernel(void) 829 { 830 char *command_line; 831 char *after_dashes; 832 833 set_task_stack_end_magic(&init_task); 834 smp_setup_processor_id(); 835 debug_objects_early_init(); 836 837 cgroup_init_early(); 838 839 local_irq_disable(); 840 early_boot_irqs_disabled = true; 841 842 /* 843 * Interrupts are still disabled. Do necessary setups, then 844 * enable them. 845 */ 846 boot_cpu_init(); 847 page_address_init(); 848 pr_notice("%s", linux_banner); 849 early_security_init(); 850 setup_arch(&command_line); 851 setup_boot_config(command_line); 852 setup_command_line(command_line); 853 setup_nr_cpu_ids(); 854 setup_per_cpu_areas(); 855 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ 856 boot_cpu_hotplug_init(); 857 858 build_all_zonelists(NULL); 859 page_alloc_init(); 860 861 pr_notice("Kernel command line: %s\n", saved_command_line); 862 /* parameters may set static keys */ 863 jump_label_init(); 864 parse_early_param(); 865 after_dashes = parse_args("Booting kernel", 866 static_command_line, __start___param, 867 __stop___param - __start___param, 868 -1, -1, NULL, &unknown_bootoption); 869 if (!IS_ERR_OR_NULL(after_dashes)) 870 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1, 871 NULL, set_init_arg); 872 if (extra_init_args) 873 parse_args("Setting extra init args", extra_init_args, 874 NULL, 0, -1, -1, NULL, set_init_arg); 875 876 /* 877 * These use large bootmem allocations and must precede 878 * kmem_cache_init() 879 */ 880 setup_log_buf(0); 881 vfs_caches_init_early(); 882 sort_main_extable(); 883 trap_init(); 884 mm_init(); 885 886 ftrace_init(); 887 888 /* trace_printk can be enabled here */ 889 early_trace_init(); 890 891 /* 892 * Set up the scheduler prior starting any interrupts (such as the 893 * timer interrupt). Full topology setup happens at smp_init() 894 * time - but meanwhile we still have a functioning scheduler. 895 */ 896 sched_init(); 897 /* 898 * Disable preemption - early bootup scheduling is extremely 899 * fragile until we cpu_idle() for the first time. 900 */ 901 preempt_disable(); 902 if (WARN(!irqs_disabled(), 903 "Interrupts were enabled *very* early, fixing it\n")) 904 local_irq_disable(); 905 radix_tree_init(); 906 907 /* 908 * Set up housekeeping before setting up workqueues to allow the unbound 909 * workqueue to take non-housekeeping into account. 910 */ 911 housekeeping_init(); 912 913 /* 914 * Allow workqueue creation and work item queueing/cancelling 915 * early. Work item execution depends on kthreads and starts after 916 * workqueue_init(). 917 */ 918 workqueue_init_early(); 919 920 rcu_init(); 921 922 /* Trace events are available after this */ 923 trace_init(); 924 925 if (initcall_debug) 926 initcall_debug_enable(); 927 928 context_tracking_init(); 929 /* init some links before init_ISA_irqs() */ 930 early_irq_init(); 931 init_IRQ(); 932 tick_init(); 933 rcu_init_nohz(); 934 init_timers(); 935 hrtimers_init(); 936 softirq_init(); 937 timekeeping_init(); 938 939 /* 940 * For best initial stack canary entropy, prepare it after: 941 * - setup_arch() for any UEFI RNG entropy and boot cmdline access 942 * - timekeeping_init() for ktime entropy used in rand_initialize() 943 * - rand_initialize() to get any arch-specific entropy like RDRAND 944 * - add_latent_entropy() to get any latent entropy 945 * - adding command line entropy 946 */ 947 rand_initialize(); 948 add_latent_entropy(); 949 add_device_randomness(command_line, strlen(command_line)); 950 boot_init_stack_canary(); 951 952 time_init(); 953 perf_event_init(); 954 profile_init(); 955 call_function_init(); 956 WARN(!irqs_disabled(), "Interrupts were enabled early\n"); 957 958 early_boot_irqs_disabled = false; 959 local_irq_enable(); 960 961 kmem_cache_init_late(); 962 963 /* 964 * HACK ALERT! This is early. We're enabling the console before 965 * we've done PCI setups etc, and console_init() must be aware of 966 * this. But we do want output early, in case something goes wrong. 967 */ 968 console_init(); 969 if (panic_later) 970 panic("Too many boot %s vars at `%s'", panic_later, 971 panic_param); 972 973 lockdep_init(); 974 975 /* 976 * Need to run this when irqs are enabled, because it wants 977 * to self-test [hard/soft]-irqs on/off lock inversion bugs 978 * too: 979 */ 980 locking_selftest(); 981 982 /* 983 * This needs to be called before any devices perform DMA 984 * operations that might use the SWIOTLB bounce buffers. It will 985 * mark the bounce buffers as decrypted so that their usage will 986 * not cause "plain-text" data to be decrypted when accessed. 987 */ 988 mem_encrypt_init(); 989 990 #ifdef CONFIG_BLK_DEV_INITRD 991 if (initrd_start && !initrd_below_start_ok && 992 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { 993 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", 994 page_to_pfn(virt_to_page((void *)initrd_start)), 995 min_low_pfn); 996 initrd_start = 0; 997 } 998 #endif 999 setup_per_cpu_pageset(); 1000 numa_policy_init(); 1001 acpi_early_init(); 1002 if (late_time_init) 1003 late_time_init(); 1004 sched_clock_init(); 1005 calibrate_delay(); 1006 pid_idr_init(); 1007 anon_vma_init(); 1008 #ifdef CONFIG_X86 1009 if (efi_enabled(EFI_RUNTIME_SERVICES)) 1010 efi_enter_virtual_mode(); 1011 #endif 1012 thread_stack_cache_init(); 1013 cred_init(); 1014 fork_init(); 1015 proc_caches_init(); 1016 uts_ns_init(); 1017 buffer_init(); 1018 key_init(); 1019 security_init(); 1020 dbg_late_init(); 1021 vfs_caches_init(); 1022 pagecache_init(); 1023 signals_init(); 1024 seq_file_init(); 1025 proc_root_init(); 1026 nsfs_init(); 1027 cpuset_init(); 1028 cgroup_init(); 1029 taskstats_init_early(); 1030 delayacct_init(); 1031 1032 poking_init(); 1033 check_bugs(); 1034 1035 acpi_subsystem_init(); 1036 arch_post_acpi_subsys_init(); 1037 sfi_init_late(); 1038 1039 /* Do the rest non-__init'ed, we're now alive */ 1040 arch_call_rest_init(); 1041 1042 prevent_tail_call_optimization(); 1043 } 1044 1045 /* Call all constructor functions linked into the kernel. */ 1046 static void __init do_ctors(void) 1047 { 1048 #ifdef CONFIG_CONSTRUCTORS 1049 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; 1050 1051 for (; fn < (ctor_fn_t *) __ctors_end; fn++) 1052 (*fn)(); 1053 #endif 1054 } 1055 1056 #ifdef CONFIG_KALLSYMS 1057 struct blacklist_entry { 1058 struct list_head next; 1059 char *buf; 1060 }; 1061 1062 static __initdata_or_module LIST_HEAD(blacklisted_initcalls); 1063 1064 static int __init initcall_blacklist(char *str) 1065 { 1066 char *str_entry; 1067 struct blacklist_entry *entry; 1068 1069 /* str argument is a comma-separated list of functions */ 1070 do { 1071 str_entry = strsep(&str, ","); 1072 if (str_entry) { 1073 pr_debug("blacklisting initcall %s\n", str_entry); 1074 entry = memblock_alloc(sizeof(*entry), 1075 SMP_CACHE_BYTES); 1076 if (!entry) 1077 panic("%s: Failed to allocate %zu bytes\n", 1078 __func__, sizeof(*entry)); 1079 entry->buf = memblock_alloc(strlen(str_entry) + 1, 1080 SMP_CACHE_BYTES); 1081 if (!entry->buf) 1082 panic("%s: Failed to allocate %zu bytes\n", 1083 __func__, strlen(str_entry) + 1); 1084 strcpy(entry->buf, str_entry); 1085 list_add(&entry->next, &blacklisted_initcalls); 1086 } 1087 } while (str_entry); 1088 1089 return 0; 1090 } 1091 1092 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1093 { 1094 struct blacklist_entry *entry; 1095 char fn_name[KSYM_SYMBOL_LEN]; 1096 unsigned long addr; 1097 1098 if (list_empty(&blacklisted_initcalls)) 1099 return false; 1100 1101 addr = (unsigned long) dereference_function_descriptor(fn); 1102 sprint_symbol_no_offset(fn_name, addr); 1103 1104 /* 1105 * fn will be "function_name [module_name]" where [module_name] is not 1106 * displayed for built-in init functions. Strip off the [module_name]. 1107 */ 1108 strreplace(fn_name, ' ', '\0'); 1109 1110 list_for_each_entry(entry, &blacklisted_initcalls, next) { 1111 if (!strcmp(fn_name, entry->buf)) { 1112 pr_debug("initcall %s blacklisted\n", fn_name); 1113 return true; 1114 } 1115 } 1116 1117 return false; 1118 } 1119 #else 1120 static int __init initcall_blacklist(char *str) 1121 { 1122 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); 1123 return 0; 1124 } 1125 1126 static bool __init_or_module initcall_blacklisted(initcall_t fn) 1127 { 1128 return false; 1129 } 1130 #endif 1131 __setup("initcall_blacklist=", initcall_blacklist); 1132 1133 static __init_or_module void 1134 trace_initcall_start_cb(void *data, initcall_t fn) 1135 { 1136 ktime_t *calltime = (ktime_t *)data; 1137 1138 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current)); 1139 *calltime = ktime_get(); 1140 } 1141 1142 static __init_or_module void 1143 trace_initcall_finish_cb(void *data, initcall_t fn, int ret) 1144 { 1145 ktime_t *calltime = (ktime_t *)data; 1146 ktime_t delta, rettime; 1147 unsigned long long duration; 1148 1149 rettime = ktime_get(); 1150 delta = ktime_sub(rettime, *calltime); 1151 duration = (unsigned long long) ktime_to_ns(delta) >> 10; 1152 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n", 1153 fn, ret, duration); 1154 } 1155 1156 static ktime_t initcall_calltime; 1157 1158 #ifdef TRACEPOINTS_ENABLED 1159 static void __init initcall_debug_enable(void) 1160 { 1161 int ret; 1162 1163 ret = register_trace_initcall_start(trace_initcall_start_cb, 1164 &initcall_calltime); 1165 ret |= register_trace_initcall_finish(trace_initcall_finish_cb, 1166 &initcall_calltime); 1167 WARN(ret, "Failed to register initcall tracepoints\n"); 1168 } 1169 # define do_trace_initcall_start trace_initcall_start 1170 # define do_trace_initcall_finish trace_initcall_finish 1171 #else 1172 static inline void do_trace_initcall_start(initcall_t fn) 1173 { 1174 if (!initcall_debug) 1175 return; 1176 trace_initcall_start_cb(&initcall_calltime, fn); 1177 } 1178 static inline void do_trace_initcall_finish(initcall_t fn, int ret) 1179 { 1180 if (!initcall_debug) 1181 return; 1182 trace_initcall_finish_cb(&initcall_calltime, fn, ret); 1183 } 1184 #endif /* !TRACEPOINTS_ENABLED */ 1185 1186 int __init_or_module do_one_initcall(initcall_t fn) 1187 { 1188 int count = preempt_count(); 1189 char msgbuf[64]; 1190 int ret; 1191 1192 if (initcall_blacklisted(fn)) 1193 return -EPERM; 1194 1195 do_trace_initcall_start(fn); 1196 ret = fn(); 1197 do_trace_initcall_finish(fn, ret); 1198 1199 msgbuf[0] = 0; 1200 1201 if (preempt_count() != count) { 1202 sprintf(msgbuf, "preemption imbalance "); 1203 preempt_count_set(count); 1204 } 1205 if (irqs_disabled()) { 1206 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); 1207 local_irq_enable(); 1208 } 1209 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf); 1210 1211 add_latent_entropy(); 1212 return ret; 1213 } 1214 1215 1216 extern initcall_entry_t __initcall_start[]; 1217 extern initcall_entry_t __initcall0_start[]; 1218 extern initcall_entry_t __initcall1_start[]; 1219 extern initcall_entry_t __initcall2_start[]; 1220 extern initcall_entry_t __initcall3_start[]; 1221 extern initcall_entry_t __initcall4_start[]; 1222 extern initcall_entry_t __initcall5_start[]; 1223 extern initcall_entry_t __initcall6_start[]; 1224 extern initcall_entry_t __initcall7_start[]; 1225 extern initcall_entry_t __initcall_end[]; 1226 1227 static initcall_entry_t *initcall_levels[] __initdata = { 1228 __initcall0_start, 1229 __initcall1_start, 1230 __initcall2_start, 1231 __initcall3_start, 1232 __initcall4_start, 1233 __initcall5_start, 1234 __initcall6_start, 1235 __initcall7_start, 1236 __initcall_end, 1237 }; 1238 1239 /* Keep these in sync with initcalls in include/linux/init.h */ 1240 static const char *initcall_level_names[] __initdata = { 1241 "pure", 1242 "core", 1243 "postcore", 1244 "arch", 1245 "subsys", 1246 "fs", 1247 "device", 1248 "late", 1249 }; 1250 1251 static int __init ignore_unknown_bootoption(char *param, char *val, 1252 const char *unused, void *arg) 1253 { 1254 return 0; 1255 } 1256 1257 static void __init do_initcall_level(int level, char *command_line) 1258 { 1259 initcall_entry_t *fn; 1260 1261 parse_args(initcall_level_names[level], 1262 command_line, __start___param, 1263 __stop___param - __start___param, 1264 level, level, 1265 NULL, ignore_unknown_bootoption); 1266 1267 trace_initcall_level(initcall_level_names[level]); 1268 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) 1269 do_one_initcall(initcall_from_entry(fn)); 1270 } 1271 1272 static void __init do_initcalls(void) 1273 { 1274 int level; 1275 size_t len = strlen(saved_command_line) + 1; 1276 char *command_line; 1277 1278 command_line = kzalloc(len, GFP_KERNEL); 1279 if (!command_line) 1280 panic("%s: Failed to allocate %zu bytes\n", __func__, len); 1281 1282 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) { 1283 /* Parser modifies command_line, restore it each time */ 1284 strcpy(command_line, saved_command_line); 1285 do_initcall_level(level, command_line); 1286 } 1287 1288 kfree(command_line); 1289 } 1290 1291 /* 1292 * Ok, the machine is now initialized. None of the devices 1293 * have been touched yet, but the CPU subsystem is up and 1294 * running, and memory and process management works. 1295 * 1296 * Now we can finally start doing some real work.. 1297 */ 1298 static void __init do_basic_setup(void) 1299 { 1300 cpuset_init_smp(); 1301 driver_init(); 1302 init_irq_proc(); 1303 do_ctors(); 1304 usermodehelper_enable(); 1305 do_initcalls(); 1306 } 1307 1308 static void __init do_pre_smp_initcalls(void) 1309 { 1310 initcall_entry_t *fn; 1311 1312 trace_initcall_level("early"); 1313 for (fn = __initcall_start; fn < __initcall0_start; fn++) 1314 do_one_initcall(initcall_from_entry(fn)); 1315 } 1316 1317 static int run_init_process(const char *init_filename) 1318 { 1319 const char *const *p; 1320 1321 argv_init[0] = init_filename; 1322 pr_info("Run %s as init process\n", init_filename); 1323 pr_debug(" with arguments:\n"); 1324 for (p = argv_init; *p; p++) 1325 pr_debug(" %s\n", *p); 1326 pr_debug(" with environment:\n"); 1327 for (p = envp_init; *p; p++) 1328 pr_debug(" %s\n", *p); 1329 return do_execve(getname_kernel(init_filename), 1330 (const char __user *const __user *)argv_init, 1331 (const char __user *const __user *)envp_init); 1332 } 1333 1334 static int try_to_run_init_process(const char *init_filename) 1335 { 1336 int ret; 1337 1338 ret = run_init_process(init_filename); 1339 1340 if (ret && ret != -ENOENT) { 1341 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", 1342 init_filename, ret); 1343 } 1344 1345 return ret; 1346 } 1347 1348 static noinline void __init kernel_init_freeable(void); 1349 1350 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX) 1351 bool rodata_enabled __ro_after_init = true; 1352 static int __init set_debug_rodata(char *str) 1353 { 1354 return strtobool(str, &rodata_enabled); 1355 } 1356 __setup("rodata=", set_debug_rodata); 1357 #endif 1358 1359 #ifdef CONFIG_STRICT_KERNEL_RWX 1360 static void mark_readonly(void) 1361 { 1362 if (rodata_enabled) { 1363 /* 1364 * load_module() results in W+X mappings, which are cleaned 1365 * up with call_rcu(). Let's make sure that queued work is 1366 * flushed so that we don't hit false positives looking for 1367 * insecure pages which are W+X. 1368 */ 1369 rcu_barrier(); 1370 mark_rodata_ro(); 1371 rodata_test(); 1372 } else 1373 pr_info("Kernel memory protection disabled.\n"); 1374 } 1375 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX) 1376 static inline void mark_readonly(void) 1377 { 1378 pr_warn("Kernel memory protection not selected by kernel config.\n"); 1379 } 1380 #else 1381 static inline void mark_readonly(void) 1382 { 1383 pr_warn("This architecture does not have kernel memory protection.\n"); 1384 } 1385 #endif 1386 1387 void __weak free_initmem(void) 1388 { 1389 free_initmem_default(POISON_FREE_INITMEM); 1390 } 1391 1392 static int __ref kernel_init(void *unused) 1393 { 1394 int ret; 1395 1396 kernel_init_freeable(); 1397 /* need to finish all async __init code before freeing the memory */ 1398 async_synchronize_full(); 1399 ftrace_free_init_mem(); 1400 free_initmem(); 1401 mark_readonly(); 1402 1403 /* 1404 * Kernel mappings are now finalized - update the userspace page-table 1405 * to finalize PTI. 1406 */ 1407 pti_finalize(); 1408 1409 system_state = SYSTEM_RUNNING; 1410 numa_default_policy(); 1411 1412 rcu_end_inkernel_boot(); 1413 1414 if (ramdisk_execute_command) { 1415 ret = run_init_process(ramdisk_execute_command); 1416 if (!ret) 1417 return 0; 1418 pr_err("Failed to execute %s (error %d)\n", 1419 ramdisk_execute_command, ret); 1420 } 1421 1422 /* 1423 * We try each of these until one succeeds. 1424 * 1425 * The Bourne shell can be used instead of init if we are 1426 * trying to recover a really broken machine. 1427 */ 1428 if (execute_command) { 1429 ret = run_init_process(execute_command); 1430 if (!ret) 1431 return 0; 1432 panic("Requested init %s failed (error %d).", 1433 execute_command, ret); 1434 } 1435 if (!try_to_run_init_process("/sbin/init") || 1436 !try_to_run_init_process("/etc/init") || 1437 !try_to_run_init_process("/bin/init") || 1438 !try_to_run_init_process("/bin/sh")) 1439 return 0; 1440 1441 panic("No working init found. Try passing init= option to kernel. " 1442 "See Linux Documentation/admin-guide/init.rst for guidance."); 1443 } 1444 1445 void console_on_rootfs(void) 1446 { 1447 /* Open the /dev/console as stdin, this should never fail */ 1448 if (ksys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) 1449 pr_err("Warning: unable to open an initial console.\n"); 1450 1451 /* create stdout/stderr */ 1452 (void) ksys_dup(0); 1453 (void) ksys_dup(0); 1454 } 1455 1456 static noinline void __init kernel_init_freeable(void) 1457 { 1458 /* 1459 * Wait until kthreadd is all set-up. 1460 */ 1461 wait_for_completion(&kthreadd_done); 1462 1463 /* Now the scheduler is fully set up and can do blocking allocations */ 1464 gfp_allowed_mask = __GFP_BITS_MASK; 1465 1466 /* 1467 * init can allocate pages on any node 1468 */ 1469 set_mems_allowed(node_states[N_MEMORY]); 1470 1471 cad_pid = task_pid(current); 1472 1473 smp_prepare_cpus(setup_max_cpus); 1474 1475 workqueue_init(); 1476 1477 init_mm_internals(); 1478 1479 do_pre_smp_initcalls(); 1480 lockup_detector_init(); 1481 1482 smp_init(); 1483 sched_init_smp(); 1484 1485 page_alloc_init_late(); 1486 /* Initialize page ext after all struct pages are initialized. */ 1487 page_ext_init(); 1488 1489 do_basic_setup(); 1490 1491 console_on_rootfs(); 1492 1493 /* 1494 * check if there is an early userspace init. If yes, let it do all 1495 * the work 1496 */ 1497 1498 if (!ramdisk_execute_command) 1499 ramdisk_execute_command = "/init"; 1500 1501 if (ksys_access((const char __user *) 1502 ramdisk_execute_command, 0) != 0) { 1503 ramdisk_execute_command = NULL; 1504 prepare_namespace(); 1505 } 1506 1507 /* 1508 * Ok, we have completed the initial bootup, and 1509 * we're essentially up and running. Get rid of the 1510 * initmem segments and start the user-mode stuff.. 1511 * 1512 * rootfs is available now, try loading the public keys 1513 * and default modules 1514 */ 1515 1516 integrity_load_keys(); 1517 } 1518