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